diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000000000000000000000000000000000000..f5e10357d6d6d2ef04c264fafd7653256cefa4bd --- /dev/null +++ b/.gitignore @@ -0,0 +1,81 @@ +# no IntelliJ files +.idea + +# don't upload macOS folder info +*.DS_Store + +# don't upload node_modules from npm test +node_modules/* +flow-typed/* + +# potential files generated by golang +bin/ + +# don't upload webpack bundle file +app/dist/ + +# potential integration testing data directory +# test_data/ +/data + +#python +*.pyc +__pycache__/ + +# pytype +.pytype + +# vscode sftp settings +.vscode/sftp.json + +# vscode launch settings +.vscode/launch.json + +# redis +*.rdb + +# mypy +.mypy_cache + +# jest coverage cache +coverage/ + +# downloaded repos and models +scalabel/bot/experimental/* + + +# python virtual environment +env/ + +# vscode workspace configuration +*.code-workspace + +# sphinx build folder +_build/ + +# media files are not in this repo +doc/media + +# ignore rope db cache +.vscode/.ropeproject + +# python build +build/ +dist/ + +# coverage +.coverage* + +# package default workspace +/output + +*.tmp +*.zip + +# local test logs and scripts +log/ +/*.sh +wandb/ + +# No lightning logs +lightning_logs/ diff --git a/52715.error b/52715.error new file mode 100644 index 0000000000000000000000000000000000000000..bacdfe52b63ca97653a4c20c0e481b5eb2579f33 --- /dev/null +++ b/52715.error @@ -0,0 +1,309 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-zcubaqpyrbpe +error libmamba Could not lock non-existing path '/scratch/yuqian_fu/micromamba/pkgs' +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-zcubaqpyrbpe + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 hd590300_5 conda-forge + + ca-certificates 2024.6.2 hbcca054_0 conda-forge + + certifi 2024.6.2 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.0.4 0 nvidia + + libcurand 10.3.6.39 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 13.2.0 h77fa898_13 conda-forge + + libgfortran-ng 13.2.0 h69a702a_13 conda-forge + + libgfortran5 13.2.0 h3d2ce59_13 conda-forge + + libhwloc 2.10.0 default_h5622ce7_1001 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 13.2.0 hc0a3c3a_13 conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4ab18f5_1 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.3.1 py3.11_cuda12.1_cudnn8.9.2_0 pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 70.1.1 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.12.1 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h297d8ca_1 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 2.3.1 py311 pytorch + + torchvision 0.18.1 py311_cu121 pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.22.0 py311hb6f056b_1 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 118 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.6.2-hbcca054_0 +Linking libgcc-ng-13.2.0-h77fa898_13 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4ab18f5_1 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-13.2.0-h3d2ce59_13 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-13.2.0-hc0a3c3a_13 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-hd590300_5 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-13.2.0-h69a702a_13 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.10.0-default_h5622ce7_1001 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h297d8ca_1 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.39-0 +Linking libcufile-1.10.0.4-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-70.1.1-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.6.2-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.22.0-py311hb6f056b_1 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.12.1-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.3.1-py3.11_cuda12.1_cudnn8.9.2_0 +Linking torchtriton-2.3.1-py311 +Linking torchvision-0.18.1-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-zcubaqpyrbpe + +Or to execute a single command in this environment, use: + + mamba run -n auto-zcubaqpyrbpe mycommand + +Installing pip packages +WARNING: The candidate selected for download or install is a yanked version: 'opencv-python' candidate (version 4.5.5.62 at https://files.pythonhosted.org/packages/9d/98/36bfcbff30da27dd6922ed73ca7802c37d87f77daf4c569da3dcb87b4296/opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from https://pypi.org/simple/opencv-python/) (requires-python:>=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/metatrain_CausalStyle_RN.py", line 124, in + base_loader = base_datamgr.get_data_loader( base_file , aug = params.train_aug ) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/data/datamgr.py", line 137, in get_data_loader + dataset = SetDataset( data_file , self.batch_size, transform ) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/data/dataset.py", line 62, in __init__ + with open(data_file, 'r') as f: + ^^^^^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/scratch/yuqian_fu/Data/CDFSL/miniImagenet/base.json' +srun: error: gcp-eu-2: task 0: Exited with exit code 1 diff --git a/52715.log b/52715.log new file mode 100644 index 0000000000000000000000000000000000000000..e2dab5e5e97b17c6726913cd802593d7624926d6 --- /dev/null +++ b/52715.log @@ -0,0 +1,114 @@ +Collecting h5py>=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 8.8 MB/s eta 0:00:00 + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 8.1 MB/s eta 0:00:00 +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 16.0 MB/s eta 0:00:00 +Requirement already satisfied: numpy>=1.21.2 in ./lib/python3.11/site-packages (from opencv-python==4.5.5.62) (2.0.0) +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.64.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (70.1.1) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.3.1) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.18.1) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.23.4-py3-none-any.whl.metadata (12 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.0-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 22.1 MB/s eta 0:00:00 +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.12.1) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.6.2) +Requirement already satisfied: mpmath<1.4.0,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 91.0 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 183.1 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 107.5 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 30.9 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 36.0 MB/s eta 0:00:00 +Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 137.1 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 51.3 MB/s eta 0:00:00 +Downloading grpcio-1.64.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.6/5.6 MB 88.2 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 38.1 MB/s eta 0:00:00 +Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl (294 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 50.7 MB/s eta 0:00:00 +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 173.4 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 73.1 MB/s eta 0:00:00 +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.23.4-py3-none-any.whl (402 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 402.6/402.6 kB 61.1 MB/s eta 0:00:00 +Downloading packaging-24.1-py3-none-any.whl (53 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 21.8 MB/s eta 0:00:00 +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 122.2 MB/s eta 0:00:00 +Downloading fsspec-2024.6.0-py3-none-any.whl (176 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 176.9/176.9 kB 62.7 MB/s eta 0:00:00 +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 27.8 MB/s eta 0:00:00 +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94508 sha256=d89d1d746d60ee7c5ccd906afd932a6369bd5c90b009d4e595ac300929458aa5 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.0 grpcio-1.64.1 h5py-3.11.0 huggingface_hub-0.23.4 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.3 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 +backbone: maml: False +hi this is causal style +set seed = 0 + +--- prepare dataloader --- + train with single seen domain miniImagenet + +--- build model --- diff --git a/52729.error b/52729.error new file mode 100644 index 0000000000000000000000000000000000000000..6bb69d733d0e6354df15e75d8b5e56eef42c386e --- /dev/null +++ b/52729.error @@ -0,0 +1,309 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-zcubaqpyrbpe +error libmamba Could not lock non-existing path '/scratch/yuqian_fu/micromamba/pkgs' +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-zcubaqpyrbpe + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 hd590300_5 conda-forge + + ca-certificates 2024.6.2 hbcca054_0 conda-forge + + certifi 2024.6.2 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.0.4 0 nvidia + + libcurand 10.3.6.39 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 13.2.0 h77fa898_13 conda-forge + + libgfortran-ng 13.2.0 h69a702a_13 conda-forge + + libgfortran5 13.2.0 h3d2ce59_13 conda-forge + + libhwloc 2.10.0 default_h5622ce7_1001 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 13.2.0 hc0a3c3a_13 conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4ab18f5_1 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.3.1 py3.11_cuda12.1_cudnn8.9.2_0 pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 70.1.1 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.12.1 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h297d8ca_1 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 2.3.1 py311 pytorch + + torchvision 0.18.1 py311_cu121 pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.22.0 py311hb6f056b_1 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 118 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.6.2-hbcca054_0 +Linking libgcc-ng-13.2.0-h77fa898_13 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4ab18f5_1 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-13.2.0-h3d2ce59_13 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-13.2.0-hc0a3c3a_13 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-hd590300_5 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-13.2.0-h69a702a_13 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.10.0-default_h5622ce7_1001 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h297d8ca_1 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.39-0 +Linking libcufile-1.10.0.4-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-70.1.1-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.6.2-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.22.0-py311hb6f056b_1 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.12.1-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.3.1-py3.11_cuda12.1_cudnn8.9.2_0 +Linking torchtriton-2.3.1-py311 +Linking torchvision-0.18.1-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-zcubaqpyrbpe + +Or to execute a single command in this environment, use: + + mamba run -n auto-zcubaqpyrbpe mycommand + +Installing pip packages +WARNING: The candidate selected for download or install is a yanked version: 'opencv-python' candidate (version 4.5.5.62 at https://files.pythonhosted.org/packages/9d/98/36bfcbff30da27dd6922ed73ca7802c37d87f77daf4c569da3dcb87b4296/opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from https://pypi.org/simple/opencv-python/) (requires-python:>=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/metatrain_CausalStyle_RN.py", line 124, in + base_loader = base_datamgr.get_data_loader( base_file , aug = params.train_aug ) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/data/datamgr.py", line 137, in get_data_loader + dataset = SetDataset( data_file , self.batch_size, transform ) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/data/dataset.py", line 62, in __init__ + with open(data_file, 'r') as f: + ^^^^^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/scratch/yuqian_fu/Data/CDFSL/miniImagenet/base.json' +srun: error: gcpl4-eu-2: task 0: Exited with exit code 1 diff --git a/52729.log b/52729.log new file mode 100644 index 0000000000000000000000000000000000000000..2e5387778a5566fec5cdfbab38bcbeafe1e5d4d7 --- /dev/null +++ b/52729.log @@ -0,0 +1,114 @@ +Collecting h5py>=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 8.5 MB/s eta 0:00:00 + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 10.4 MB/s eta 0:00:00 +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 16.7 MB/s eta 0:00:00 +Requirement already satisfied: numpy>=1.21.2 in ./lib/python3.11/site-packages (from opencv-python==4.5.5.62) (2.0.0) +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.64.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (70.1.1) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.3.1) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.18.1) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.23.4-py3-none-any.whl.metadata (12 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.0-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 24.9 MB/s eta 0:00:00 +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.12.1) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.6.2) +Requirement already satisfied: mpmath<1.4.0,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 72.8 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 209.0 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 115.1 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 211.7 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 38.7 MB/s eta 0:00:00 +Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 181.5 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 52.8 MB/s eta 0:00:00 +Downloading grpcio-1.64.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.6/5.6 MB 212.4 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 40.3 MB/s eta 0:00:00 +Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl (294 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 101.1 MB/s eta 0:00:00 +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 214.7 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 74.9 MB/s eta 0:00:00 +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.23.4-py3-none-any.whl (402 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 402.6/402.6 kB 109.9 MB/s eta 0:00:00 +Downloading packaging-24.1-py3-none-any.whl (53 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 21.8 MB/s eta 0:00:00 +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 170.9 MB/s eta 0:00:00 +Downloading fsspec-2024.6.0-py3-none-any.whl (176 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 176.9/176.9 kB 62.5 MB/s eta 0:00:00 +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 30.5 MB/s eta 0:00:00 +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94508 sha256=2e320bb7bf02566bf671fd943ea8dfe7cb6c35a1fab523a080d4ab487706ca51 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.0 grpcio-1.64.1 h5py-3.11.0 huggingface_hub-0.23.4 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.3 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 +backbone: maml: False +hi this is causal style +set seed = 0 + +--- prepare dataloader --- + train with single seen domain miniImagenet + +--- build model --- diff --git a/Meta-causal/code-stage1-pipeline/56451.error b/Meta-causal/code-stage1-pipeline/56451.error new file mode 100644 index 0000000000000000000000000000000000000000..2551fb8116caf9cb33618556d7b8f611e0d7465d --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56451.error @@ -0,0 +1,297 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem +error libmamba Could not lock non-existing path '/scratch/yuqian_fu/micromamba/pkgs' +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 hd590300_5 conda-forge + + ca-certificates 2024.6.2 hbcca054_0 conda-forge + + certifi 2024.6.2 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + click 8.1.7 unix_pyh707e725_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.0.4 0 nvidia + + libcurand 10.3.6.39 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 14.1.0 h77fa898_0 conda-forge + + libgfortran-ng 14.1.0 h69a702a_0 conda-forge + + libgfortran5 14.1.0 hc5f4f2c_0 conda-forge + + libhwloc 2.10.0 default_h5622ce7_1001 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 14.1.0 hc0a3c3a_0 conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4ab18f5_1 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.3.1 py3.11_cuda12.1_cudnn8.9.2_0 pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 70.1.1 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.12.1 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h297d8ca_1 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 2.3.1 py311 pytorch + + torchvision 0.18.1 py311_cu121 pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.22.0 py311hb6f056b_1 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 119 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.6.2-hbcca054_0 +Linking libgcc-ng-14.1.0-h77fa898_0 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4ab18f5_1 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-14.1.0-hc5f4f2c_0 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-14.1.0-hc0a3c3a_0 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-hd590300_5 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-14.1.0-h69a702a_0 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.10.0-default_h5622ce7_1001 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h297d8ca_1 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.39-0 +Linking libcufile-1.10.0.4-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-70.1.1-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.6.2-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking click-8.1.7-unix_pyh707e725_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.22.0-py311hb6f056b_1 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.12.1-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.3.1-py3.11_cuda12.1_cudnn8.9.2_0 +Linking torchtriton-2.3.1-py311 +Linking torchvision-0.18.1-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-uvapqvk3mmem + +Or to execute a single command in this environment, use: + + mamba run -n auto-uvapqvk3mmem mycommand + +slurmstepd: error: *** JOB 56451 ON gcpl4-eu-1 CANCELLED AT 2024-07-03T18:51:16 *** diff --git a/Meta-causal/code-stage1-pipeline/56451.log b/Meta-causal/code-stage1-pipeline/56451.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-stage1-pipeline/56452.error b/Meta-causal/code-stage1-pipeline/56452.error new file mode 100644 index 0000000000000000000000000000000000000000..eb5687437cf8e2e4cb8a88d2216f2c4477d146f8 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56452.error @@ -0,0 +1,302 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 hd590300_5 conda-forge + + ca-certificates 2024.6.2 hbcca054_0 conda-forge + + certifi 2024.6.2 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + click 8.1.7 unix_pyh707e725_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.0.4 0 nvidia + + libcurand 10.3.6.39 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 14.1.0 h77fa898_0 conda-forge + + libgfortran-ng 14.1.0 h69a702a_0 conda-forge + + libgfortran5 14.1.0 hc5f4f2c_0 conda-forge + + libhwloc 2.10.0 default_h5622ce7_1001 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 14.1.0 hc0a3c3a_0 conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4ab18f5_1 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.3.1 py3.11_cuda12.1_cudnn8.9.2_0 pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 70.1.1 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.12.1 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h297d8ca_1 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 2.3.1 py311 pytorch + + torchvision 0.18.1 py311_cu121 pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.22.0 py311hb6f056b_1 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 119 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.6.2-hbcca054_0 +Linking libgcc-ng-14.1.0-h77fa898_0 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4ab18f5_1 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-14.1.0-hc5f4f2c_0 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-14.1.0-hc0a3c3a_0 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-hd590300_5 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-14.1.0-h69a702a_0 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.10.0-default_h5622ce7_1001 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h297d8ca_1 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.39-0 +Linking libcufile-1.10.0.4-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-70.1.1-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.6.2-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking click-8.1.7-unix_pyh707e725_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.22.0-py311hb6f056b_1 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.12.1-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.3.1-py3.11_cuda12.1_cudnn8.9.2_0 +Linking torchtriton-2.3.1-py311 +Linking torchvision-0.18.1-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-uvapqvk3mmem + +Or to execute a single command in this environment, use: + + mamba run -n auto-uvapqvk3mmem mycommand + +Installing pip packages +WARNING: The candidate selected for download or install is a yanked version: 'opencv-python' candidate (version 4.5.5.62 at https://files.pythonhosted.org/packages/9d/98/36bfcbff30da27dd6922ed73ca7802c37d87f77daf4c569da3dcb87b4296/opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from https://pypi.org/simple/opencv-python/) (requires-python:>=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +Downloading: "https://download.pytorch.org/models/resnet18-5c106cde.pth" to /home/yuqian_fu/.cache/torch/hub/checkpoints/resnet18-5c106cde.pth + 0%| | 0.00/44.7M [00:00=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 7.6 MB/s eta 0:00:00 + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.0) +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 8.2 MB/s eta 0:00:00 +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 10.6 MB/s eta 0:00:00 +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.64.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (70.1.1) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.3.1) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.18.1) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.23.4-py3-none-any.whl.metadata (12 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 16.4 MB/s eta 0:00:00 +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.12.1) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.6.2) +Requirement already satisfied: mpmath<1.4.0,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 78.5 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 187.2 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 105.5 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 189.1 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 23.7 MB/s eta 0:00:00 +Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 138.0 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 34.2 MB/s eta 0:00:00 +Downloading grpcio-1.64.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.6/5.6 MB 47.4 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 26.1 MB/s eta 0:00:00 +Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl (294 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 76.8 MB/s eta 0:00:00 +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 184.7 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 52.0 MB/s eta 0:00:00 +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.23.4-py3-none-any.whl (402 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 402.6/402.6 kB 82.1 MB/s eta 0:00:00 +Downloading packaging-24.1-py3-none-any.whl (53 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 14.7 MB/s eta 0:00:00 +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 139.2 MB/s eta 0:00:00 +Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 177.6/177.6 kB 39.5 MB/s eta 0:00:00 +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 18.7 MB/s eta 0:00:00 +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94508 sha256=ed551f29efc119726e34db83e826c7054abc35ca00823e92370fbc0d102e27d3 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.64.1 h5py-3.11.0 huggingface_hub-0.23.4 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.3 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +cls_loss: tensor(1.8677, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5274, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2724, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7454, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7351, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(19.6074, device='cuda:0', grad_fn=) +cls_loss: tensor(23.8932, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(28.2583, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(19.5065, device='cuda:0', grad_fn=) +cls_loss: tensor(22.3359, device='cuda:0', grad_fn=) +cls_loss: tensor(28.7865, device='cuda:0', grad_fn=) +cls_loss: tensor(15.5003, device='cuda:0', grad_fn=) +cls_loss: tensor(27.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3876, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5369, device='cuda:0', grad_fn=) +cls_loss: tensor(17.4609, device='cuda:0', grad_fn=) +cls_loss: tensor(22.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(13.6418, device='cuda:0', grad_fn=) +cls_loss: tensor(13.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5850, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4558, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8523, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7253, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8304, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(12.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(13.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9399, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3218, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8511, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4748, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6276, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8281, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5759, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8670, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6841, device='cuda:0', grad_fn=) +cls_loss: tensor(22.9951, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8101, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3748, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9532, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2864, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6035, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7227, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2335, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8568, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(15.2824, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4733, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6382, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4119, device='cuda:0', grad_fn=) +cls_loss: tensor(12.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6377, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6395, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2512, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5969, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4717, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1980, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4515, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5948, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9618, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3908, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7495, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(13.9140, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9928, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3869, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7747, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4823, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4743, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8799, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2122, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3178, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6303, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5418, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(13.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2512, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9743, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3078, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5521, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2450, device='cuda:0', grad_fn=) +cls_loss: tensor(30.5912, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(32.6253, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7701, device='cuda:0', grad_fn=) +cls_loss: tensor(16.4349, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(14.8165, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(16.9219, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3618, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8140, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7055, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7781, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(28.6595, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4610, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3654, device='cuda:0', grad_fn=) +cls_loss: tensor(26.3878, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8438, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7386, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6859, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(27.7709, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4917, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3954, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5443, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2637, device='cuda:0', grad_fn=) +cls_loss: tensor(11.7891, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8556, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7957, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5208, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6721, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5635, device='cuda:0', grad_fn=) +cls_loss: tensor(12.7279, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3374, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(10.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5833, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4670, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9811, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6888, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7663, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9148, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7839, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3827, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8335, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5385, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8363, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7214, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8316, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6947, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9106, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5374, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7467, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5571, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4619, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5257, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9370, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3424, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3971, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(12.9838, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6745, device='cuda:0', grad_fn=) +cls_loss: tensor(17.8411, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(21.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8903, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7702, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7995, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6107, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2865, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8125, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(13.8633, device='cuda:0', grad_fn=) +cls_loss: tensor(30.7253, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1873, device='cuda:0', grad_fn=) +cls_loss: tensor(18.3563, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5029, device='cuda:0', grad_fn=) +cls_loss: tensor(18.3669, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4372, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6957, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3639, device='cuda:0', grad_fn=) +cls_loss: tensor(16.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6319, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4059, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(13.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4266, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(17.7253, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7390, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3624, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6455, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6290, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(10.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9396, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(16.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(12.2853, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7637, device='cuda:0', grad_fn=) +cls_loss: tensor(20.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4255, device='cuda:0', grad_fn=) +cls_loss: tensor(14.9134, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 296.16, cls_loss 6.9189 +306 +cls_loss: tensor(6.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6393, device='cuda:0', grad_fn=) +cls_loss: tensor(16.9254, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9640, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(17.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5752, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7122, device='cuda:0', grad_fn=) +cls_loss: tensor(11.5632, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(31.9297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3857, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7765, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5456, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(14.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9997, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(12.5495, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9035, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4453, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8842, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5899, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8408, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3622, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5247, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6808, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6878, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5247, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8314, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3802, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7747, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3682, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8264, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9121, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7459, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(13.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(19.4871, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9396, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7344, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6937, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6211, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9950, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(24.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4063, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6615, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3763, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3221, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7031, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3477, device='cuda:0', grad_fn=) +cls_loss: tensor(22.6382, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7207, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7938, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4337, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(19.7530, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(11.6979, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(11.8449, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8123, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4502, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6549, device='cuda:0', grad_fn=) +cls_loss: tensor(18.2136, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(21.3053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7774, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5023, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(20.6413, device='cuda:0', grad_fn=) +cls_loss: tensor(11.5389, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4310, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4346, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8451, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8858, device='cuda:0', grad_fn=) +cls_loss: tensor(13.7500, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7048, device='cuda:0', grad_fn=) +cls_loss: tensor(13.1950, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(16.4442, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7812, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6981, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5546, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2949, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4180, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2578, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9661, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9688, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7450, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9196, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9195, device='cuda:0', grad_fn=) +cls_loss: tensor(17.5592, device='cuda:0', grad_fn=) +cls_loss: tensor(15.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7760, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4530, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9727, device='cuda:0', grad_fn=) +cls_loss: tensor(15.5319, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7071, device='cuda:0', grad_fn=) +cls_loss: tensor(24.5208, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8910, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1912, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6824, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1745, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2082, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7356, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7450, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6276, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6017, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4591, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(20.8785, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(22.6927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8420, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(27.6635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3336, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7500, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(19.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8439, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9730, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(12.5365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2710, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4475, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4921, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2827, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7057, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2812, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8569, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2350, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5521, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4861, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9331, device='cuda:0', grad_fn=) +9.994965332706573e-05 +changing lr +epoch 1, time 305.00, cls_loss 4.7944 +306 +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6212, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7223, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6408, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4090, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7480, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5130, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9505, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7084, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8700, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9600, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6317, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9554, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8973, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7747, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6836, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3855, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4648, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4356, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4202, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5091, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5110, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2930, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8164, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8958, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3945, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4557, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8604, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3287, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4400, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6982, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(15.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8916, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9158, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7920, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3186, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(12.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4870, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9398, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3468, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7098, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6364, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2660, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3360, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5329, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9283, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6562, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7605, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8872, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4518, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4785, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9435, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2741, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9573, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9740, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8268, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5970, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3830, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3260, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8992, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6237, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5330, device='cuda:0', grad_fn=) +cls_loss: tensor(10.3138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6589, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5856, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6621, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9229, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9467, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7670, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9922, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8952, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9362, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8444, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8551, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6776, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(11.8453, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4567, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5544, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7513, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3997, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7809, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1938, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5475, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3503, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7511, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5345, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5400, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4372, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2842, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9781, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(20.5430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3415, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9740, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9635, device='cuda:0', grad_fn=) +9.979871469976196e-05 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 358.04, cls_loss 2.7433 +306 +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8438, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9303, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9068, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7981, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3783, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7448, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8804, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8320, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7562, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8789, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7832, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7746, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5094, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(10.6749, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4309, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5586, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2178, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8205, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9202, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6470, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(17.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9486, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3156, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2650, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6198, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8674, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6260, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(10.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5928, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3988, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9833, device='cuda:0', grad_fn=) +cls_loss: tensor(10.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2467, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9310, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5236, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5938, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5078, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3501, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3770, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5087, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3910, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9382, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3000, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9893, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8937, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5553, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9251, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8989, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6901, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4160, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6473, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3900, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7695, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(12.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8594, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4178, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3259, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6777, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2070, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3097, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6628, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5117, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9310, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.954748808839674e-05 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 369.87, cls_loss 1.7181 +306 +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7214, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4611, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5794, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5056, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5615, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7240, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3980, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9878, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5648, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8600, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3245, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4410, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2475, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6285, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3163, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3763, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8281, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6729, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6460, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7715, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2793, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3848, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1870, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3194, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6228, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8646, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6442, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8730, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8070e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4632, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8135, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7444, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5869, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9113, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8375, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7760, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6455, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5081, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9173, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2308, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6436, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4759, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6087, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4464, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8594, device='cuda:0', grad_fn=) +9.919647942993148e-05 +changing lr +epoch 4, time 372.12, cls_loss 0.8746 +306 +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3662, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9564, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9844, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9860, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8331, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7716, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7891, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3758, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8075, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6120, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(10.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9630, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4749, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(13.5286, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8627, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9948, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6654, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3817, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(16.4116, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3711, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2773, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6191, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9704, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3973, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8333, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6927, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5451, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3356, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6556, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9622, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5159, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3356, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(12.3050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2839, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3801, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +9.874639560909117e-05 +changing lr +epoch 5, time 370.13, cls_loss 1.0098 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(12.1962, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9928, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7713, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8266, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8385, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9674, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3216, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2804, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4851e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8190, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5072, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2662, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8151, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7575, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9320, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4715, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2723, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7324, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7902, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5680, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7656, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9904, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8493, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8125, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2526, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7331, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5581, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7090, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6927, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2268, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5360, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3680, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9818, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9844, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7575, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3296, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5312, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4489, device='cuda:0', grad_fn=) +9.819814303479267e-05 +changing lr +epoch 6, time 370.75, cls_loss 0.7924 +306 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9782, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9564, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1777, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6901, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7760, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7500, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9355, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5209, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7031, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7849, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5973, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5521, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5732, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1692, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3290, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9245, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(15.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5352, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2318, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3434, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8509, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3096, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9798, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9662, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9621, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2650, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4468, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8516, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2649, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8282, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8399, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3161, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3161, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2232, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3362, device='cuda:0', grad_fn=) +9.755282581475769e-05 +changing lr +epoch 7, time 370.66, cls_loss 0.7084 +306 +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7383, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4538, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5998, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5794, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2922, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6417, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9486, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2077, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7431, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9230, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6602, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2969, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6386, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5036e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9785, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4870, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4261, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4550, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5511, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9844, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5280, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3346, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5298, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5755, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6745, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4629, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3856, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3648, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4064, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5839, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7695, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2334, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +9.681174353198687e-05 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 370.71, cls_loss 0.4985 +306 +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3172, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5179, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4206, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2535, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5948, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7485, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5244, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6824, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4958, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4362, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7448, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3916, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4101, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7366, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9243, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9682, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9870, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4584, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3880, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9642, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7162, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8294, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9447, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4417, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7526, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5521, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2765, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5477, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.597638862757255e-05 +changing lr +epoch 9, time 370.77, cls_loss 0.3858 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7461, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4322, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6699, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7292, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9701, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8749, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6452, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(13.3325, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9909, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8958, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5844e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4762, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9992, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5924, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3516, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4158, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4500, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2814, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2124, device='cuda:0', grad_fn=) +9.504844339512095e-05 +changing lr +epoch 10, time 379.50, cls_loss 0.4715 +306 +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8783, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4935, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2493, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7578, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2845, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5638, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5329, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5859, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8152, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7337, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.40297765928369e-05 +changing lr +epoch 11, time 371.48, cls_loss 0.1305 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6250, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4531, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2861, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5381, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7383, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4294, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5807, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8997, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9303, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4349, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8913, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8516, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6510, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3988, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2463, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6844, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.292243968009331e-05 +changing lr +epoch 12, time 369.81, cls_loss 0.1468 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6966, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4932, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4604, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3318, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4761, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3604, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2293, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5679, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3669, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2410, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8070e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4467, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9590, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5658, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +9.172866268606513e-05 +changing lr +epoch 13, time 369.00, cls_loss 0.1700 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2058, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8525, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5299, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2432, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2067, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4394e-05, device='cuda:0', grad_fn=) +9.045084971874738e-05 +changing lr +epoch 14, time 370.38, cls_loss 0.1292 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5623, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2858, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9653, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2736, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5124, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7695, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7682, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3277, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3870, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9219, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3359, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5537, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6979, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2988, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(13.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7363, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1473e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9642, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4434, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4528, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6517, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9531, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6328, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8840, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9837, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +8.90915741234015e-05 +changing lr +epoch 15, time 371.67, cls_loss 0.2955 +306 +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6953, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3308, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3886, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1745, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7540, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5195, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4349, device='cuda:0', grad_fn=) +8.765357330018056e-05 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 372.24, cls_loss 0.1137 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4157, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5745, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3340, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4535, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6315, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8769, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6671, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8306, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2922, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8529, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +8.613974319136958e-05 +changing lr +epoch 17, time 371.27, cls_loss 0.0677 +306 +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2409, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4736, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5954, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5709, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9401, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9772, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7735, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +8.455313244934324e-05 +changing lr +epoch 18, time 372.50, cls_loss 0.1072 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9720, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9440, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8997, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6085, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3371, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2246, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4486, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5740, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +8.289693629698562e-05 +changing lr +epoch 19, time 371.56, cls_loss 0.0832 +306 +cls_loss: tensor(0.8887, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3916, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +8.117449009293667e-05 +changing lr +epoch 20, time 369.09, cls_loss 0.0275 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6484, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6237, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9694, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7233, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7780, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +7.938926261462365e-05 +changing lr +epoch 21, time 368.36, cls_loss 0.0585 +306 +cls_loss: tensor(7.2360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7396, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6992, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3819, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4635, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2596, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +7.754484907260511e-05 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 367.15, cls_loss 0.0616 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8216, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4892, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1966, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8472, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3099, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +7.56449638702953e-05 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 368.69, cls_loss 0.0542 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8118, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4564, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4251, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7467, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6972, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9954, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6000, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4648, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6401e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +7.369343312364992e-05 +changing lr +epoch 24, time 372.23, cls_loss 0.0459 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4358, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2423, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2856, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +7.169418695587788e-05 +changing lr +epoch 25, time 371.99, cls_loss 0.0211 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4023, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2248, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5355, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3441, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.965125158269616e-05 +changing lr +epoch 26, time 366.50, cls_loss 0.0463 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3253, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6520, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6390, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3398, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +6.756874120406712e-05 +changing lr +epoch 27, time 373.44, cls_loss 0.0311 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6159, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8939, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3408, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6364, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7650, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.545084971874736e-05 +changing lr +epoch 28, time 369.89, cls_loss 0.0291 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4903e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2433, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.330184227833373e-05 +changing lr +epoch 29, time 380.65, cls_loss 0.0019 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.11260466978157e-05 +changing lr +epoch 30, time 372.52, cls_loss 0.0014 +306 +cls_loss: tensor(4.1107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.892784473993181e-05 +changing lr +---------------------saving model at epoch 31---------------------------------------------------- +epoch 31, time 368.20, cls_loss 0.0031 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9863, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9895, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.6711663290882756e-05 +changing lr +epoch 32, time 371.78, cls_loss 0.0068 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5632, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7933, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2443, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.448196544517166e-05 +changing lr +epoch 33, time 370.50, cls_loss 0.0207 +306 +cls_loss: tensor(5.3445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7852, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8320, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2215, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3328, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9355, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2943e-05, device='cuda:0', grad_fn=) +5.2243241517525733e-05 +changing lr +epoch 34, time 368.99, cls_loss 0.0115 +306 +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6279, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.999999999999998e-05 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 371.09, cls_loss 0.0023 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6556, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2861, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.7756758482474244e-05 +changing lr +epoch 36, time 370.52, cls_loss 0.0033 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.551803455482832e-05 +changing lr +epoch 37, time 373.66, cls_loss 0.0003 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4226, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.328833670911722e-05 +changing lr +epoch 38, time 370.52, cls_loss 0.0022 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4945, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2814, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.1072155260068164e-05 +changing lr +epoch 39, time 368.60, cls_loss 0.0333 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.8873953302184275e-05 +changing lr +epoch 40, time 372.40, cls_loss 0.0009 +306 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2400, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.669815772166625e-05 +changing lr +epoch 41, time 370.21, cls_loss 0.0017 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.454915028125263e-05 +changing lr +epoch 42, time 371.34, cls_loss 0.0008 +306 +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.2431258795932867e-05 +changing lr +epoch 43, time 371.28, cls_loss 0.0003 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6527, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5505, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.0348748417303827e-05 +changing lr +epoch 44, time 369.58, cls_loss 0.0184 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2645, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4310, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.830581304412209e-05 +changing lr +epoch 45, time 371.66, cls_loss 0.0106 +306 +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.6306566876350062e-05 +changing lr +epoch 46, time 370.93, cls_loss 0.0004 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5869, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3923, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8846, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +2.4355036129704693e-05 +changing lr +---------------------saving model at epoch 47---------------------------------------------------- +epoch 47, time 371.02, cls_loss 0.0094 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8372, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3216, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.2455150927394874e-05 +changing lr +epoch 48, time 372.32, cls_loss 0.0075 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.0610737385376345e-05 +changing lr +epoch 49, time 370.91, cls_loss 0.0004 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8851e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2657, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8713, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5108, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5698e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2016e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.8825509907063323e-05 +changing lr +epoch 50, time 371.81, cls_loss 0.0091 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6416, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2433, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.710306370301437e-05 +changing lr +epoch 51, time 375.01, cls_loss 0.0040 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6901, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5754, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5486e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4372, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +1.5446867550656765e-05 +changing lr +epoch 52, time 370.33, cls_loss 0.0134 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4450, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5050e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9421, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.3860256808630425e-05 +changing lr +epoch 53, time 367.86, cls_loss 0.0062 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2710, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.2346426699819455e-05 +changing lr +epoch 54, time 372.13, cls_loss 0.0010 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2922, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.0908425876598507e-05 +changing lr +epoch 55, time 372.19, cls_loss 0.0017 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.549150281252631e-06 +changing lr +epoch 56, time 374.32, cls_loss 0.0014 +306 +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9588, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +8.271337313934865e-06 +changing lr +epoch 57, time 375.12, cls_loss 0.0065 +306 +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +7.077560319906693e-06 +changing lr +epoch 58, time 369.93, cls_loss 0.0002 +306 +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5183, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3956, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.970223407163098e-06 +changing lr +epoch 59, time 371.25, cls_loss 0.0097 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2596e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.9515566048790464e-06 +changing lr +epoch 60, time 369.02, cls_loss 0.0020 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +4.0236113724274705e-06 +changing lr +epoch 61, time 371.04, cls_loss 0.0006 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9142e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +3.188256468013139e-06 +changing lr +epoch 62, time 370.44, cls_loss 0.0022 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.4471741852423225e-06 +changing lr +epoch 63, time 369.65, cls_loss 0.0012 +306 +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +1.8018569652073373e-06 +changing lr +epoch 64, time 370.77, cls_loss 0.0120 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3499, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.2536043909088185e-06 +changing lr +epoch 65, time 372.14, cls_loss 0.0013 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +8.035205700685162e-07 +changing lr +epoch 66, time 368.52, cls_loss 0.0000 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.525119116032647e-07 +changing lr +epoch 67, time 373.91, cls_loss 0.0013 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.0128530023804648e-07 +changing lr +epoch 68, time 370.90, cls_loss 0.0011 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2596e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5963e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +5.03466729342705e-08 +changing lr +epoch 69, time 343.07, cls_loss 0.0001 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +x.shape: (2048, 227, 227, 3) +x_aug test here torch.Size([2048, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +x.shape: (2344, 227, 227, 3) +x_aug test here torch.Size([2344, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +x.shape: (1670, 227, 227, 3) +x_aug test here torch.Size([1670, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +x.shape: (3929, 227, 227, 3) +x_aug test here torch.Size([3929, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 8.984375 5.759386 11.137725 15.474675 10.790595 diff --git a/Meta-causal/code-stage1-pipeline/56454.error b/Meta-causal/code-stage1-pipeline/56454.error new file mode 100644 index 0000000000000000000000000000000000000000..4d4eb4cfa24b0c733d487e584e09bc5d6db9f3f9 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56454.error @@ -0,0 +1,3 @@ +slurmstepd: error: *** JOB 56454 ON gcpl4-eu-1 CANCELLED AT 2024-07-04T06:57:02 DUE TO TIME LIMIT *** +slurmstepd: error: *** STEP 56454.0 ON gcpl4-eu-1 CANCELLED AT 2024-07-04T06:57:02 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-stage1-pipeline/56454.log b/Meta-causal/code-stage1-pipeline/56454.log new file mode 100644 index 0000000000000000000000000000000000000000..0916114b9207bb4c78f7d8d859de666c1add3cc5 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56454.log @@ -0,0 +1,24485 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +cls_loss: tensor(2.2979, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3093, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2847, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2577, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2842, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2863, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2263, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1963, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2168, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9801, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9552, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8921, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9290, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5482, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6989, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5818, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4784, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4934, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4230, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8324, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9257, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6746, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9380, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6683, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4493, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 184.94, cls_loss 1.0915 +100 +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1861, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 185.54, cls_loss 0.3618 +100 +cls_loss: tensor(0.2670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2480, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 185.39, cls_loss 0.2422 +100 +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 186.34, cls_loss 0.1883 +100 +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2930, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 185.66, cls_loss 0.1516 +100 +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 185.84, cls_loss 0.1194 +100 +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 185.38, cls_loss 0.1092 +100 +cls_loss: tensor(0.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 185.55, cls_loss 0.0984 +100 +cls_loss: tensor(0.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 185.48, cls_loss 0.0902 +100 +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 9, time 185.10, cls_loss 0.0674 +100 +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 186.23, cls_loss 0.0662 +100 +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 185.79, cls_loss 0.0599 +100 +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 185.75, cls_loss 0.0600 +100 +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 13, time 185.55, cls_loss 0.0650 +100 +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 14, time 184.99, cls_loss 0.0570 +100 +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 15, time 185.17, cls_loss 0.0396 +100 +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 184.75, cls_loss 0.0403 +100 +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 17, time 183.70, cls_loss 0.0370 +100 +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 18, time 183.93, cls_loss 0.0422 +100 +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 19, time 183.43, cls_loss 0.0434 +100 +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 183.27, cls_loss 0.0352 +100 +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 21, time 183.36, cls_loss 0.0297 +100 +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 22, time 183.12, cls_loss 0.0239 +100 +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 23, time 183.42, cls_loss 0.0314 +100 +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 24, time 183.44, cls_loss 0.0295 +100 +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 183.89, cls_loss 0.0145 +100 +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 183.47, cls_loss 0.0102 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1840e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 27, time 183.01, cls_loss 0.0175 +100 +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 28, time 183.09, cls_loss 0.0356 +100 +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 29, time 182.78, cls_loss 0.0161 +100 +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9165e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 30, time 183.12, cls_loss 0.0169 +100 +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 31, time 183.03, cls_loss 0.0115 +100 +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4276e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 32, time 183.43, cls_loss 0.0193 +100 +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1997e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 33, time 183.03, cls_loss 0.0195 +100 +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 34, time 183.14, cls_loss 0.0103 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 35, time 182.90, cls_loss 0.0126 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8654e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 36, time 183.16, cls_loss 0.0095 +100 +cls_loss: tensor(8.6002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7754e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0931e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 37, time 183.79, cls_loss 0.0082 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0350e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2454e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 38, time 183.14, cls_loss 0.0161 +100 +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2270e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4112e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 39, time 183.19, cls_loss 0.0133 +100 +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4493e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9906e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5250e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 40, time 183.48, cls_loss 0.0055 +100 +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 41, time 183.25, cls_loss 0.0050 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5164e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7840e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4812e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 183.26, cls_loss 0.0034 +100 +cls_loss: tensor(1.9077e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0634e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0144e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 43, time 183.54, cls_loss 0.0097 +100 +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0311e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2846e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 44, time 183.51, cls_loss 0.0057 +100 +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7469e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5237e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2063e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5005e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 45, time 183.32, cls_loss 0.0031 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4176e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4666e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4343e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 46, time 183.18, cls_loss 0.0010 +100 +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6224e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2217e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7390e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 47, time 183.73, cls_loss 0.0091 +100 +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3092e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8653e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6184e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3377e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8375e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8256e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9499e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 48, time 183.68, cls_loss 0.0036 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4872e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7216e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2609e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8013e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9726e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6196e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7258e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 49, time 183.41, cls_loss 0.0090 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 50, time 183.60, cls_loss 0.0302 +100 +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4362e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2143e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 51, time 183.34, cls_loss 0.0138 +100 +cls_loss: tensor(7.3526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5654e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5647e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 52, time 183.53, cls_loss 0.0038 +100 +cls_loss: tensor(3.4284e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8682e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7932e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3938e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0118e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8772e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 53, time 183.15, cls_loss 0.0060 +100 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7428e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5845e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0541e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6885e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 54, time 183.39, cls_loss 0.0054 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8589e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3906e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1993e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5436e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7322e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3713e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0812e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 55, time 183.71, cls_loss 0.0068 +100 +cls_loss: tensor(2.9791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2586e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5198e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4814e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8582e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8688e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1800e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6667e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2792e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 56, time 183.66, cls_loss 0.0037 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3979e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7845e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9547e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5204e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 57, time 183.49, cls_loss 0.0132 +100 +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7966e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6329e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3931e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9600e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6867e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 58, time 183.39, cls_loss 0.0024 +100 +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0275e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7548e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0753e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6448e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9276e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1183e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 59, time 183.85, cls_loss 0.0037 +100 +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4772e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7939e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5965e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0633e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5820e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3005e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 60, time 183.40, cls_loss 0.0053 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9581e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9807e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6554e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5197e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0172e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9251e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8965e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7125e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9866e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5911e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3375e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0330e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 61, time 183.48, cls_loss 0.0066 +100 +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5621e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4561e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5442e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0243e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0436e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1713e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1561e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8666e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2912e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3051e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 62, time 183.83, cls_loss 0.0050 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9998e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3309e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0978e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0807e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9893e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2741e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5939e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2662e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 63---------------------------------------------------- +epoch 63, time 184.00, cls_loss 0.0017 +100 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2799e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5177e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7588e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6761e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7906e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3931e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5779e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1178e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9422e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2668e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 64, time 183.92, cls_loss 0.0046 +100 +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9244e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9647e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2144e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 65, time 183.72, cls_loss 0.0122 +100 +cls_loss: tensor(9.3438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3847e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5509e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6699e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1628e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0031e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2437e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 66, time 183.80, cls_loss 0.0045 +100 +cls_loss: tensor(7.6279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3945e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3304e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0197e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4760e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9006e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1727e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8971e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8142e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1191e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1832e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1953e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 67, time 183.63, cls_loss 0.0041 +100 +cls_loss: tensor(7.7050e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5045e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9337e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2093, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5144e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3185e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3383e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1906e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 68, time 183.36, cls_loss 0.0139 +100 +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8038e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8470e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8980e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8079e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8634e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7654e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3289e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0939e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 69, time 182.95, cls_loss 0.0066 +100 +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9628e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9111e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7721e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 70, time 183.30, cls_loss 0.0065 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9109e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5406e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2950e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7953e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9242e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2406e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5079e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8488e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7270e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 71, time 183.29, cls_loss 0.0053 +100 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5225e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0812e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6891e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6237e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 72, time 183.13, cls_loss 0.0131 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8841e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7666e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5853e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2084e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8205e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6455e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2282e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 73, time 183.25, cls_loss 0.0029 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5806e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9152e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5344e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6131e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6430e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6838e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2926e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6375e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3076e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7713e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4503e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9606e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9733e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 74, time 183.26, cls_loss 0.0017 +100 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9282e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0576e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0852e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8847e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1116e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1370e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7912e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6171e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0589e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7582e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4568e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2203e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2999e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8719e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3032e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5129e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2820e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0735e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0124e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 75, time 183.19, cls_loss 0.0008 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5475e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0424e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5389e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5812e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9741e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9920e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0284e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8686e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8441e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2282e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6436e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1185e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2667e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7603e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8470e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8721e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4899e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4382e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5032e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9945e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6218e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2203e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7747e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 76, time 183.32, cls_loss 0.0005 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1660e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2911e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1383e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3534e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8759e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6853e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1833e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5343e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1038e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7125e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7119e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5297e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1754e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6694e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2308e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5311e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3541e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5317e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 77, time 183.08, cls_loss 0.0056 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3019e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1747e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1337e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6688e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6026e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3521e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5481e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2700e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5806e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4980e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1761e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7927e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2747e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 78, time 183.10, cls_loss 0.0029 +100 +cls_loss: tensor(4.1511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5972e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3283e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5576e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1105e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5800e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1646e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0085e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3171e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7701e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4528e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8747e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6514e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8727e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 79, time 182.70, cls_loss 0.0097 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2059e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8574e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5032e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0735e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8430e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3979e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6879e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2720e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9945e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4762e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 80, time 183.14, cls_loss 0.0045 +100 +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7795e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9311e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7251e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9011e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7912e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7463e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3382e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3211e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2536e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2767e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9742e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3568e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 81, time 183.31, cls_loss 0.0067 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0323e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6449e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7284e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7255e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5282e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2648e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7501e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9335e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4283e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5130e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7397e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2607e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5699e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 82, time 183.07, cls_loss 0.0021 +100 +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6536e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0336e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6634e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8939e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9004e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2760e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2487e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1800e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1554e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 83, time 183.04, cls_loss 0.0046 +100 +cls_loss: tensor(8.0541e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1515e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4693e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2118e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6238e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5097e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1053e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7197e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6225e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6423e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3814e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7032e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1194e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 84, time 183.35, cls_loss 0.0011 +100 +cls_loss: tensor(3.2660e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2985e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8536e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8712e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1940e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1613e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6084e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8184e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5204e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2978e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8820e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2701e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5847e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3244e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2204e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1456e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2242e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4085e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3143e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 85, time 183.16, cls_loss 0.0036 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5288e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3867e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7215e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8124e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3423e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3792e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4549e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8156e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7377e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4112e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5799e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6992e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7410e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8912e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0079e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7828e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1854e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1951e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7151e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1388e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2363e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9847e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4547e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7065e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 86, time 183.46, cls_loss 0.0006 +100 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1296e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3277e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4574e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7198e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6343e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0349e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3474e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6720e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0278e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4309e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4999e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0827e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3556e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3161e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 87---------------------------------------------------- +epoch 87, time 183.24, cls_loss 0.0002 +100 +cls_loss: tensor(1.1124e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9806e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9527e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5039e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9004e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0699e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1331e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7084e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8727e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8255e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7176e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2686e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9712e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7767e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0886e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0865e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9521e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5509e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2621e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0834e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1044e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2984e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 88, time 183.05, cls_loss 0.0001 +100 +cls_loss: tensor(5.3123e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6581e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8269e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1204e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2329e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5110e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1739e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1940e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4250e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4997e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1535e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6066e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9237e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2762e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4569e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0371e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2112e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0960e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5218e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9019e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9874e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6622e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 89, time 182.97, cls_loss 0.0001 +100 +cls_loss: tensor(4.2133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5270e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6628e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9297e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3954e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8712e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4720e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0509e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5291e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2515e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3351e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7834e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6589e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2205e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9740e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7554e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1947e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3541e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4104e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8343e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8924e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4302e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5779e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3112e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1157e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3525e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 90, time 183.28, cls_loss 0.0001 +100 +cls_loss: tensor(4.1187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9204e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8780e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5687e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0635e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0951e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5927e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6998e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3653e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7866e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8211e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6950e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5103e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0084e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1037e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6753e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7455e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1271e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7158e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2000e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5534e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9620e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4522e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2171e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8454e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3058e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7515e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9542e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0390e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5523e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0429e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7528e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 91, time 182.74, cls_loss 0.0001 +100 +cls_loss: tensor(4.3765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1249e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6719e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5779e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1085e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9190e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4350e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0059e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5311e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0324e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6098e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7601e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5528e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2536e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0205e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5806e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5607e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5773e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8163e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9077e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5363e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2647e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5740e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8066e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0622e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0330e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6858e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4070e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9132e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0798e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 92, time 183.03, cls_loss 0.0001 +100 +cls_loss: tensor(2.3969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4953e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0899e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2277e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2832e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5311e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2774e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8759e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4324e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1288e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6129e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8396e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4993e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2164e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4569e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5712e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1404e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5959e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6980e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1271e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1156e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8820e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4550e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2871e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5661e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1225e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1052e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9971e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2825e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8389e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8142e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0502e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9799e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4303e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8694e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2244e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4501e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 93, time 183.15, cls_loss 0.0000 +100 +cls_loss: tensor(1.5900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3973e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4974e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9549e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5772e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0436e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2946e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2606e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1642e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4503e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1255e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2311e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1020e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5660e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5628e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7602e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9297e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3851e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8666e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3581e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3091e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0429e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5230e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7258e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3991e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7302e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8967e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3601e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7150e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8203e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9891e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2198e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9051e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9211e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5615e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8251e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7932e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5077e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5278e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8024e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3432e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 94---------------------------------------------------- +epoch 94, time 183.35, cls_loss 0.0000 +100 +cls_loss: tensor(1.6153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3574e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0064e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8761e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5204e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0832e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1931e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9826e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2653e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7217e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9648e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3978e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8640e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3806e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7655e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2867e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5437e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3746e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5311e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1537e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1754e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0481e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8165e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3607e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2145e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7316e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1079e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 95, time 183.62, cls_loss 0.0000 +100 +cls_loss: tensor(5.5034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5091e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2170e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8806e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1654e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9328e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4322e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3416e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8887e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8177e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2110e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5442e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3906e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0401e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4303e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0237e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8515e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3819e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0124e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8992e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0256e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2513e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8541e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8152e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0839e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2569e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6620e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0580e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1197e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4104e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7125e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1250e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5291e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8403e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3691e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 96, time 183.37, cls_loss 0.0000 +100 +cls_loss: tensor(2.0012e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8338e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7183e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6416e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8653e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9620e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7839e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0602e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6470e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9030e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6859e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4111e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5330e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1085e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9819e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3860e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0338e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7813e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9705e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7655e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9211e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7975e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3746e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5289e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2701e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0669e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5985e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8276e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4476e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1893e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5363e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 97, time 182.83, cls_loss 0.0000 +100 +cls_loss: tensor(9.9763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0163e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5481e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8767e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6581e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0727e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2165e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0951e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7911e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9349e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9363e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2866e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5244e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6550e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6487e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7165e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9614e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6905e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7290e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9534e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6284e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0971e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8991e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9720e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5151e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1316e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3767e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1004e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0899e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8142e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8507e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8368e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3607e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2846e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 98, time 183.28, cls_loss 0.0000 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1414e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3079e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6024e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6580e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4269e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0218e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9621e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7269e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4885e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5609e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8275e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7364e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9582e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7826e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1500e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9728e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6773e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8317e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1441e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3728e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6125e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6343e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2808e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2486e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5954e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3255e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3991e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0543e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3998e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1917e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 99, time 183.73, cls_loss 0.0000 +100 +cls_loss: tensor(2.5071e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5470e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4470e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8819e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2701e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9428e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9823e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9547e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6800e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2144e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7714e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8636e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8748e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7391e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1462e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5963e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1773e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2448e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9733e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9210e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3521e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7854e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6673e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3011e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5614e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0515e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2152e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9104e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2634e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4719e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2971e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0185e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7535e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1218e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5491e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 100, time 183.31, cls_loss 0.0000 +100 +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1748e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3926e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2966e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3084e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7614e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4899e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8507e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5183e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4887e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0576e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2878e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0528e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2819e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5501e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2634e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7726e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3991e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9355e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8119e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0844e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0337e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5674e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2920e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7176e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8291e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0734e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1092e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2480e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9302e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7840e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6040e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0064e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2812e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4090e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2144e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7083e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1980e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 101, time 183.45, cls_loss 0.0000 +100 +cls_loss: tensor(3.4980e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7368e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4044e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1197e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5455e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0568e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4767e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5423e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1183e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7549e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6514e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2185e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1085e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8827e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6291e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0026e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7721e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7356e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3132e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0839e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3251e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7606e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8024e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7753e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4132e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2813e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5006e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1250e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0600e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5965e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2953e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9269e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6853e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6853e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5667e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0627e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6806e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9966e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1119e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7393e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 102, time 183.33, cls_loss 0.0000 +100 +cls_loss: tensor(9.2324e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0997e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0554e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5832e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3090e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2416e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1488e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8774e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0457e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0006e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5058e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6654e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5885e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5111e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3271e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4044e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2164e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8369e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5555e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2454e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1488e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8607e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8343e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9269e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9705e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2457e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2476e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0602e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5289e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5290e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2894e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3176e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4436e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2787e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7679e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 103, time 183.20, cls_loss 0.0000 +100 +cls_loss: tensor(7.2941e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0734e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2957e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8210e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3608e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1749e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2277e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5958e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1772e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4203e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8840e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2229e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5655e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3872e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8694e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0688e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4237e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8238e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8343e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7807e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1912e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0721e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7217e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4032e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2834e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7768e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6098e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8893e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3607e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2957e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3145e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8205e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4715e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7959e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7493e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 104, time 183.37, cls_loss 0.0000 +100 +cls_loss: tensor(8.2701e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0734e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4746e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8238e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7397e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3150e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0415e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4342e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9390e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6561e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7344e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6827e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3613e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4428e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5065e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6827e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1158e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9595e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0701e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4103e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4674e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8859e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2827e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5719e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4178e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1269e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6766e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4315e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7975e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2230e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8461e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4172e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8308e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5667e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3596e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2613e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5667e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4788e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3633e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 105---------------------------------------------------- +epoch 105, time 183.46, cls_loss 0.0000 +100 +cls_loss: tensor(7.5147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7502e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7700e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5330e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5993e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3170e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8024e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2483e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8958e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8773e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2111e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4932e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0289e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0792e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0196e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8615e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1547e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5395e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5481e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0726e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2596e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9695e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4827e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0494e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3843e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 106, time 183.21, cls_loss 0.0000 +100 +cls_loss: tensor(9.6485e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6801e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4198e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4091e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4144e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3481e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6302e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3085e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2701e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6699e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2971e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7693e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2957e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7344e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2031e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2549e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3052e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0238e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6872e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3963e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4699e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2613e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6628e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4018e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4177e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0429e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1362e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0079e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0043e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 107---------------------------------------------------- +epoch 107, time 183.49, cls_loss 0.0000 +100 +cls_loss: tensor(3.3081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7583e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2825e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1429e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5832e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9707e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5917e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7595e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9030e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6682e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9118e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1828e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8077e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8583e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9072e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3894e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5304e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8103e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6397e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1090e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1250e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5428e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2389e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7470e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1767e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 108, time 183.56, cls_loss 0.0000 +100 +cls_loss: tensor(2.3767e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3567e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5747e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2442e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2457e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5204e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1090e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6480e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0813e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2808e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9137e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6396e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2527e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0587e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3826e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8044e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0170e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8177e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8515e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9337e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2461e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7679e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6177e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5986e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9446e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1761e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 109, time 183.51, cls_loss 0.0000 +100 +cls_loss: tensor(1.6030e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0541e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0502e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6603e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0436e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5295e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9609e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6794e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6548e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1940e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1404e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2933e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0675e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2984e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7588e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4946e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9402e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4699e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6375e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5516e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5779e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6317e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6582e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8388e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7621e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3746e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 110, time 183.34, cls_loss 0.0000 +100 +cls_loss: tensor(3.1032e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6383e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1404e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5719e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9085e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1250e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7993e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4533e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9998e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9428e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7114e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6098e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3541e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7828e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5151e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4608e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9446e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3376e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1905e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5721e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9334e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8105e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9051e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2799e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8177e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6613e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9370e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1569e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9376e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8924e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 111, time 183.70, cls_loss 0.0000 +100 +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8589e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5740e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0509e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4550e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9695e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7812e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5667e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7753e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0256e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5481e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3370e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0461e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5693e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1171e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3596e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0159e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7583e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5832e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5568e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4682e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0659e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0019e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5129e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4927e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7746e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6335e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6462e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 112, time 183.42, cls_loss 0.0000 +100 +cls_loss: tensor(3.3565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1720e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0874e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3693e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6503e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8429e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1441e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6838e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7083e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4091e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5719e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4103e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6838e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9712e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6476e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3190e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9465e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9370e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1726e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5295e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1137e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2115e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 113, time 183.41, cls_loss 0.0000 +100 +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3362e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4205e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8210e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6613e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8376e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2489e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5884e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6368e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3304e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2878e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0237e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9847e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9465e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5516e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1819e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1563e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3165e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1927e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3826e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0461e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4997e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0483e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3507e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 114, time 183.47, cls_loss 0.0000 +100 +cls_loss: tensor(4.0010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7880e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9390e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8349e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4065e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6859e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1872e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4331e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3507e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9951e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8403e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8322e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9733e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6748e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8641e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6913e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9819e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4005e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9707e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2044e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4119e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2457e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5050e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4012e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4475e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1488e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0364e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7489e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1973e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3854e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0589e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7050e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 115, time 183.40, cls_loss 0.0000 +100 +cls_loss: tensor(1.8775e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1761e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8834e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0897e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9781e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4231e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7807e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5721e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9237e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8794e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8450e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1540e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0659e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2013e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1903e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5735e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0185e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0859e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9898e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0159e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0673e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2521e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5768e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0740e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0324e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 116, time 183.99, cls_loss 0.0000 +100 +cls_loss: tensor(1.0759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9132e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0401e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8424e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1265e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8636e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3814e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3197e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5131e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7211e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1414e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2662e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6336e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0397e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3984e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6242e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4522e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4715e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8887e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2634e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7912e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9111e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9838e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3026e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8534e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3714e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0897e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 117, time 184.01, cls_loss 0.0000 +100 +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7321e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4355e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0138e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2964e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1845e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5278e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7278e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9020e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6513e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1891e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3304e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3940e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5779e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4065e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4468e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4324e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4059e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9190e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1899e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3230e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5144e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4475e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 118, time 183.71, cls_loss 0.0000 +100 +cls_loss: tensor(1.1511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4825e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3892e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6138e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9951e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5837e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1131e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5348e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0541e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4401e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1441e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9893e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7083e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0642e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1500e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7865e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3076e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3972e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8183e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7109e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3416e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3123e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3979e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1137e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2715e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2441e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 119, time 183.40, cls_loss 0.0000 +100 +cls_loss: tensor(2.6524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4342e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9222e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3210e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3947e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3772e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2084e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7807e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4608e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9509e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3158e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9465e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9104e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3877e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7700e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6641e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3991e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8012e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7975e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7295e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8636e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9349e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9092e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3926e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7742e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3516e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3805e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5912e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8024e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 120, time 183.55, cls_loss 0.0000 +100 +cls_loss: tensor(4.7721e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7083e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0423e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2762e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1516e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1872e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2000e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7199e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1383e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7125e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7476e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8243e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2513e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8759e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1143e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3877e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5039e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7114e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5502e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0031e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6317e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5832e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1931e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2894e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0063e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6636e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0334e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 121, time 183.63, cls_loss 0.0000 +100 +cls_loss: tensor(2.9728e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7523e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4682e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2389e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5740e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6615e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6315e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6282e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0329e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6343e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4171e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6112e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8813e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4448e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9402e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9951e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6263e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7402e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1893e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8701e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6454e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8199e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0499e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2825e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1707e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9521e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8328e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4953e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5251e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 122, time 183.47, cls_loss 0.0000 +100 +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8775e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5567e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3490e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8179e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4198e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9951e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0543e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0606e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3947e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9355e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1020e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9705e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9972e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4715e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9020e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4985e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3826e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9334e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6838e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6338e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7050e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9781e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2373e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 123, time 183.69, cls_loss 0.0000 +100 +cls_loss: tensor(3.7216e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5688e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1861e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5408e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4303e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3111e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8019e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1111e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3150e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4454e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8476e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2112e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4932e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2224e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2171e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8030e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3954e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2091e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3090e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9153e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8184e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 124, time 183.26, cls_loss 0.0000 +100 +cls_loss: tensor(7.7933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1383e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6827e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5146e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6375e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9376e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9179e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4000e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5740e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5416e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7237e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9781e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9621e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5589e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9754e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1429e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7125e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6859e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3602e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8275e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9036e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7344e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5581e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3057e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1009e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 125, time 183.70, cls_loss 0.0000 +100 +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6487e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2415e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1588e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5854e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0739e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6322e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4119e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1053e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1270e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5779e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5125e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0118e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7603e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0063e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5183e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6038e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7588e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4995e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3138e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 126, time 183.73, cls_loss 0.0000 +100 +cls_loss: tensor(3.2447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0415e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3979e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0077e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7807e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2063e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0904e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7092e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0297e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8461e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3171e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2170e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3076e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4177e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7807e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3191e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1202e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6561e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4987e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 127, time 183.40, cls_loss 0.0000 +100 +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2000e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0774e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4995e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6322e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4682e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1675e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4119e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7476e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6534e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8536e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1270e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2203e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9779e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0026e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5555e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1255e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1288e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6091e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 128, time 183.55, cls_loss 0.0000 +100 +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9707e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3851e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2985e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5805e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4177e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9595e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3262e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5742e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9446e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7603e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9435e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7721e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0217e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4794e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1011e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4124e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1250e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4368e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4443e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9334e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0026e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3649e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 129, time 184.01, cls_loss 0.0000 +100 +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4205e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2144e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9616e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0520e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2701e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2799e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2613e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0543e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1137e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4720e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2946e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3947e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5204e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7092e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6343e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4044e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0170e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2927e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6582e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2825e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6443e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1085e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1718e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 130, time 183.55, cls_loss 0.0000 +100 +cls_loss: tensor(2.8722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3430e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7114e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9104e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8217e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5369e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9707e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1331e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5278e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6336e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2415e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1255e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5602e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3358e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5204e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4133e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4289e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3788e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2634e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8414e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7679e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7397e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0755e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4347e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 131, time 183.49, cls_loss 0.0000 +100 +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6913e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1211e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6918e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8839e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9349e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8482e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3448e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1330e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2957e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5534e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2298e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9647e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8887e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4815e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2687e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1164e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6859e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5321e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3507e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4475e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6561e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6396e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1474e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3304e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 132, time 183.32, cls_loss 0.0000 +100 +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6615e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0813e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2329e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9567e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9242e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7807e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1335e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0487e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2580e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4720e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2165e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5826e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5050e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3537e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0892e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2734e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3819e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3500e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5251e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4145e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6242e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8813e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2489e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8338e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9483e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4363e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3984e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8887e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2159e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 133, time 183.66, cls_loss 0.0000 +100 +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9614e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7695e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2762e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8780e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4257e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2442e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5262e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2792e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5358e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1404e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7114e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1569e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6364e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5251e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9898e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 134, time 183.55, cls_loss 0.0000 +100 +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0622e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1270e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3574e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7018e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3767e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7434e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8072e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6152e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9779e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6485e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7157e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6040e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7434e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6587e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4065e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6720e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0659e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2733e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6748e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5726e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5854e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9595e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 135, time 183.62, cls_loss 0.0000 +100 +cls_loss: tensor(2.6338e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0960e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2329e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9158e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7269e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4548e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3283e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1218e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2091e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4145e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0957e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2708e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1143e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8913e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9779e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6874e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8701e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9302e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0675e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0499e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 136, time 183.57, cls_loss 0.0000 +100 +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2480e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3395e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3490e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3549e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5742e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2964e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5428e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4145e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1038e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2760e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5086e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9046e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2229e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2489e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2000e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0494e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9893e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 137, time 183.42, cls_loss 0.0000 +100 +cls_loss: tensor(1.4268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2904e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9838e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3262e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3668e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0543e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0734e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2203e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0159e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9779e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0834e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8310e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0005e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0932e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2554e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3171e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8291e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4289e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 138, time 184.18, cls_loss 0.0000 +100 +cls_loss: tensor(1.7956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2329e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1712e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8217e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5358e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1621e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9004e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2925e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4289e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7828e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8967e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7753e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9823e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4827e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3877e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4699e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1569e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7902e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 139, time 183.36, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3693e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8589e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6971e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4448e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1926e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8813e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8515e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9111e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5726e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1083e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3230e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9563e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7544e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3409e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9222e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9781e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 140, time 183.67, cls_loss 0.0000 +100 +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2894e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9435e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7046e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6485e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7679e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1628e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8946e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6913e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2613e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6368e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9781e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7695e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1211e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1707e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6368e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2144e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 141, time 183.11, cls_loss 0.0000 +100 +cls_loss: tensor(2.2128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9819e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6284e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7859e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1919e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4885e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7104e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9349e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6838e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7795e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2760e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3150e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7434e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 142, time 183.21, cls_loss 0.0000 +100 +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3474e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4277e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1474e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0026e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2277e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7742e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1675e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2659e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0759e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2112e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 143, time 184.42, cls_loss 0.0000 +100 +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3150e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3230e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0079e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3877e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0580e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4715e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0170e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7178e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2144e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8759e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2005e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3421e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 144, time 183.15, cls_loss 0.0000 +100 +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7721e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9297e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7230e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9344e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6971e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2522e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7269e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9072e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6007e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7854e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1053e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2867e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 145, time 183.32, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4885e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3076e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4044e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6992e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4103e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9185e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9216e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2787e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7178e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8424e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0382e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8887e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 146, time 183.25, cls_loss 0.0000 +100 +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0543e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1772e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1011e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4827e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3283e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2957e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3076e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4401e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8291e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3076e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2760e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4044e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6662e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 147, time 183.20, cls_loss 0.0000 +100 +cls_loss: tensor(2.1309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4995e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0871e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2575e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8328e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8759e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3262e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8296e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5805e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0275e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 148, time 183.04, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8291e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2985e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7476e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4448e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6375e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9521e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7104e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6024e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5832e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5523e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7211e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 149, time 183.56, cls_loss 0.0000 +100 +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6152e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0191e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7232e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8701e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9334e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0308e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1288e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1772e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8871e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 150, time 183.73, cls_loss 0.0000 +100 +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2277e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0170e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0580e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9465e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8887e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5257e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1495e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9349e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3819e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 151, time 183.99, cls_loss 0.0000 +100 +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6242e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2033e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2964e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4533e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2760e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6801e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8515e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2329e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8030e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3954e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 152, time 183.16, cls_loss 0.0000 +100 +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1846e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0675e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4715e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4454e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1793e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3877e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 153, time 183.79, cls_loss 0.0000 +100 +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3283e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5570e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1309e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0706e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0899e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9297e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8217e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0079e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5555e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 154, time 183.70, cls_loss 0.0000 +100 +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9465e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1772e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9712e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4133e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8536e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8647e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9898e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3523e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 155, time 183.49, cls_loss 0.0000 +100 +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8210e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6485e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6485e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4133e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3991e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2389e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0957e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7695e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 156, time 183.23, cls_loss 0.0000 +100 +cls_loss: tensor(1.6764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3746e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7774e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2144e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1474e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7172e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1761e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6242e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1926e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4401e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6615e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 157, time 183.25, cls_loss 0.0000 +100 +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3262e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7603e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2964e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2927e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7455e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9838e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 158, time 183.20, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7046e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5907e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1772e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5581e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 159, time 183.05, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3819e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2762e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9838e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 160, time 183.27, cls_loss 0.0000 +100 +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1362e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2957e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6799e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0675e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5050e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 161, time 182.81, cls_loss 0.0000 +100 +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1211e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7621e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4661e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6054e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5965e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 162, time 182.84, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0468e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7099e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6683e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7859e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7025e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7269e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 163, time 183.19, cls_loss 0.0000 +100 +cls_loss: tensor(2.3283e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1467e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5721e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2091e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8720e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1793e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7742e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 164, time 182.87, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3388e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7975e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4906e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4363e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4231e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 165, time 182.71, cls_loss 0.0000 +100 +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8403e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0291e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5623e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0897e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 166, time 182.87, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7859e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7603e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6252e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8366e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 167, time 182.78, cls_loss 0.0000 +100 +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6615e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6485e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3016e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6503e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0415e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 168, time 183.09, cls_loss 0.0000 +100 +cls_loss: tensor(1.2703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8589e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1211e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9663e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 169, time 183.05, cls_loss 0.0000 +100 +cls_loss: tensor(8.1584e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1270e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0005e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3581e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6368e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 170, time 183.05, cls_loss 0.0000 +100 +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7742e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7114e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5740e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1197e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 171, time 183.27, cls_loss 0.0000 +100 +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1036e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 172, time 182.75, cls_loss 0.0000 +100 +cls_loss: tensor(1.0319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8860e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6997e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5740e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 173, time 183.05, cls_loss 0.0000 +100 +cls_loss: tensor(7.4133e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2643e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1840e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9838e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2554e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9838e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4878e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 174, time 182.69, cls_loss 0.0000 +100 +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0580e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8701e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3821e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3150e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 175, time 183.35, cls_loss 0.0000 +100 +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5309e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 176, time 183.02, cls_loss 0.0000 +100 +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0035e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1211e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 177, time 183.08, cls_loss 0.0000 +100 +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2271e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7603e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7323e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2212e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0245e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 178, time 183.13, cls_loss 0.0000 +100 +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3819e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8918e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8662e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6741e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 179, time 183.49, cls_loss 0.0000 +100 +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6503e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4762e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1362e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5821e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 180, time 183.09, cls_loss 0.0000 +100 +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1660e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4331e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 181, time 183.00, cls_loss 0.0000 +100 +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2927e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 182, time 182.58, cls_loss 0.0000 +100 +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4342e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5134e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 183, time 182.91, cls_loss 0.0000 +100 +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0952e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9093e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8604e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1409e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9349e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 184, time 182.94, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8801e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 185, time 183.15, cls_loss 0.0000 +100 +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9488e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 186, time 183.24, cls_loss 0.0000 +100 +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1351e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9861e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 187, time 183.02, cls_loss 0.0000 +100 +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9779e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7828e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2554e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 188, time 183.02, cls_loss 0.0000 +100 +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7998e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 189, time 183.04, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3586e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5251e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 190, time 182.72, cls_loss 0.0000 +100 +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6135e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5938e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 191, time 182.98, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2037e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2468e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 192, time 182.67, cls_loss 0.0000 +100 +cls_loss: tensor(5.4017e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9546e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 193, time 183.35, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6939e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8999e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 194, time 183.48, cls_loss 0.0000 +100 +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4389e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 195, time 182.89, cls_loss 0.0000 +100 +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0350e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3097e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9034e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 196, time 183.43, cls_loss 0.0000 +100 +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5076e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 197, time 183.22, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 198, time 183.38, cls_loss 0.0000 +100 +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 199, time 183.30, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9919e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4645e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 200, time 183.08, cls_loss 0.0000 +100 +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 201, time 183.06, cls_loss 0.0000 +100 +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9116e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 202, time 182.94, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0547e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7311e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 203, time 183.22, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 204, time 183.02, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5018e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 205, time 183.61, cls_loss 0.0000 +100 +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 206, time 183.08, cls_loss 0.0000 +100 +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3900e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1665e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 207, time 183.14, cls_loss 0.0000 +100 +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9744e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1607e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 208, time 182.97, cls_loss 0.0000 +100 +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9430e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4587e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 209, time 183.18, cls_loss 0.0000 +100 +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 210, time 183.34, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 211, time 182.80, cls_loss 0.0000 +100 +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2410e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7567e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 212, time 183.16, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8254e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 213, time 182.93, cls_loss 0.0000 +100 +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 214, time 182.98, cls_loss 0.0000 +100 +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 215, time 183.00, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6019e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 216, time 182.80, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 217---------------------------------------------------- +epoch 217, time 182.90, cls_loss 0.0000 +100 +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2293e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 218, time 182.38, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3155e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 219, time 182.69, cls_loss 0.0000 +100 +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4156e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 220, time 182.68, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 221, time 182.94, cls_loss 0.0000 +100 +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7136e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5682e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 222, time 182.99, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 223, time 182.86, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 224, time 182.57, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 225, time 182.60, cls_loss 0.0000 +100 +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1956e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 226, time 182.72, cls_loss 0.0000 +100 +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 227, time 182.96, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5879e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0058e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 228, time 183.29, cls_loss 0.0000 +100 +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 229, time 182.80, cls_loss 0.0000 +100 +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3330e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 230, time 183.33, cls_loss 0.0000 +100 +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 231, time 182.99, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5274e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 232, time 183.17, cls_loss 0.0000 +100 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3132e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8429e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6764e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 233, time 182.75, cls_loss 0.0000 +100 +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6077e-08, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 234, time 182.95, cls_loss 0.0000 +100 +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-09, device='cuda:0', grad_fn=) diff --git a/Meta-causal/code-stage1-pipeline/56455.error b/Meta-causal/code-stage1-pipeline/56455.error new file mode 100644 index 0000000000000000000000000000000000000000..38d934231566fdb9cd98197f76b6dd49f332d93a --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56455.error @@ -0,0 +1,4 @@ +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) diff --git a/Meta-causal/code-stage1-pipeline/56455.log b/Meta-causal/code-stage1-pipeline/56455.log new file mode 100644 index 0000000000000000000000000000000000000000..56a2ee207832bca2331b885d1cc197f495694c99 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56455.log @@ -0,0 +1,21748 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +x.shape: (1840, 227, 227, 3) +x_aug train here torch.Size([1840, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +x.shape: (208, 227, 227, 3) +x_aug test here torch.Size([208, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +cls_loss: tensor(2.2935, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2858, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7793, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5242, device='cuda:0', grad_fn=) +cls_loss: tensor(18.2740, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2471, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1980, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3535, device='cuda:0', grad_fn=) +cls_loss: tensor(22.8802, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4238, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2368, device='cuda:0', grad_fn=) +cls_loss: tensor(30.8763, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7465, device='cuda:0', grad_fn=) +cls_loss: tensor(15.3426, device='cuda:0', grad_fn=) +cls_loss: tensor(24.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(22.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(16.8876, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7633, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5395, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3681, device='cuda:0', grad_fn=) +cls_loss: tensor(14.5060, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2463, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6156, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2663, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5333, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2918, device='cuda:0', grad_fn=) +cls_loss: tensor(13.7344, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8238, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4837, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3485, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7990, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6686, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5339, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5646, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2536, device='cuda:0', grad_fn=) +cls_loss: tensor(32.3005, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7116, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5924, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9130, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6592, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6772, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6541, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7866, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7919, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4239, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9542, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5696, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6715, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6591, device='cuda:0', grad_fn=) +cls_loss: tensor(23.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2596, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5277, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5132, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5606, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6145, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4831, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3048, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8227, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5807, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4805, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6686, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2433, device='cuda:0', grad_fn=) +cls_loss: tensor(15.3711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8917, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(14.5354, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3483, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7110, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6338, device='cuda:0', grad_fn=) +cls_loss: tensor(35.7786, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6535, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9492, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2592, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9883, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7067, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4258, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(19.8615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3566, device='cuda:0', grad_fn=) +cls_loss: tensor(16.8438, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2898, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7071, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9359, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5880, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7292, device='cuda:0', grad_fn=) +cls_loss: tensor(15.7689, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(16.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7127, device='cuda:0', grad_fn=) +cls_loss: tensor(28.6545, device='cuda:0', grad_fn=) +cls_loss: tensor(20.7057, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5473, device='cuda:0', grad_fn=) +cls_loss: tensor(17.5304, device='cuda:0', grad_fn=) +cls_loss: tensor(23.4637, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3529, device='cuda:0', grad_fn=) +cls_loss: tensor(16.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(21.6462, device='cuda:0', grad_fn=) +cls_loss: tensor(16.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(15.4636, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6375, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2438, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(19.9086, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3116, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5767, device='cuda:0', grad_fn=) +cls_loss: tensor(15.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(18.1963, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(21.9622, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4477, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8626, device='cuda:0', grad_fn=) +cls_loss: tensor(22.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8924, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2720, device='cuda:0', grad_fn=) +cls_loss: tensor(21.7604, device='cuda:0', grad_fn=) +cls_loss: tensor(22.6425, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2670, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6418, device='cuda:0', grad_fn=) +cls_loss: tensor(23.8737, device='cuda:0', grad_fn=) +cls_loss: tensor(23.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(15.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6768, device='cuda:0', grad_fn=) +cls_loss: tensor(18.6380, device='cuda:0', grad_fn=) +cls_loss: tensor(15.8815, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9714, device='cuda:0', grad_fn=) +cls_loss: tensor(36.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8483, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6608, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4831, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5443, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(22.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5573, device='cuda:0', grad_fn=) +cls_loss: tensor(22.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(21.6570, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7983, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9805, device='cuda:0', grad_fn=) +cls_loss: tensor(16.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9316, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2359, device='cuda:0', grad_fn=) +cls_loss: tensor(18.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5645, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6550, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9648, device='cuda:0', grad_fn=) +cls_loss: tensor(25.5935, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4616, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5394, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3388, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7676, device='cuda:0', grad_fn=) +cls_loss: tensor(11.6200, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5615, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7409, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(13.9711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4583, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9310, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1976, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4964, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3540, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6148, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(13.7392, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8260, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(15.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3953, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4560, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6074, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2597, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4432, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5127, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(18.8507, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9632, device='cuda:0', grad_fn=) +cls_loss: tensor(27.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(25.3352, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7556, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9127, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8202, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8961, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4439, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8843, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4674, device='cuda:0', grad_fn=) +cls_loss: tensor(24.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5400, device='cuda:0', grad_fn=) +cls_loss: tensor(14.7572, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6254, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6815, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6294, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6929, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6918, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4524, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3878, device='cuda:0', grad_fn=) +cls_loss: tensor(13.9297, device='cuda:0', grad_fn=) +cls_loss: tensor(15.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8737, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4522, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8331, device='cuda:0', grad_fn=) +cls_loss: tensor(16.5143, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6966, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6194, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9896, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6211, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8257, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(16.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(12.9343, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6281, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9135, device='cuda:0', grad_fn=) +cls_loss: tensor(12.9583, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7833, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7396, device='cuda:0', grad_fn=) +cls_loss: tensor(20.7474, device='cuda:0', grad_fn=) +cls_loss: tensor(12.5770, device='cuda:0', grad_fn=) +cls_loss: tensor(15.6935, device='cuda:0', grad_fn=) +cls_loss: tensor(13.9870, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6599, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8792, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4656, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5304, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8819, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5363, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7025, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 329.69, cls_loss 9.0417 +306 +cls_loss: tensor(11.7513, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5858, device='cuda:0', grad_fn=) +cls_loss: tensor(25.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3963, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3412, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3737, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6683, device='cuda:0', grad_fn=) +cls_loss: tensor(1., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4508, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9634, device='cuda:0', grad_fn=) +cls_loss: tensor(13.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5938, device='cuda:0', grad_fn=) +cls_loss: tensor(16.6397, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4795, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8984, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6746, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5677, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6180, device='cuda:0', grad_fn=) +cls_loss: tensor(14.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7717, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3349, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8467, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6442, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(11.7009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(11.8664, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8697, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8673, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8930, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3060, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3919, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9326, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(15.8626, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9971, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4440, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4298, device='cuda:0', grad_fn=) +cls_loss: tensor(14.5384, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3147, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5954, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7924, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8438, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6395, device='cuda:0', grad_fn=) +cls_loss: tensor(18.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9702, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2031, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6228, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3112, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8034, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7113, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2743, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4714, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3882, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4451, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9799, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6413, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6640, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1941, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6510, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5975, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1476, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8712, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2478, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2097, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9688, device='cuda:0', grad_fn=) +cls_loss: tensor(22.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6532, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4733, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4891, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8434, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4929, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7852, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8465, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4226, device='cuda:0', grad_fn=) +cls_loss: tensor(12.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6929, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6439, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2871, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2485, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7856, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4085, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7292, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2884, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9042, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7905, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4439, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5203, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2961, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1995, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6726, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3671, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2744, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8702, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8305, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2063, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2025, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7927, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(13.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4564, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5992, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9473, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7230, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9518, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5143, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3282, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5887, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9669, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5222, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9622, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9868, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3735, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4391, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5443, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8513, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5460, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3369, device='cuda:0', grad_fn=) +cls_loss: tensor(15.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8587, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3589, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9562, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9977, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8757, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4481, device='cuda:0', grad_fn=) +cls_loss: tensor(17.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7166, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7644, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3680, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3547, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2626, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8995, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7760, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8818, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(19.7578, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5216, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6152, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4978, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6004, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7148, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6655, device='cuda:0', grad_fn=) +cls_loss: tensor(16.9719, device='cuda:0', grad_fn=) +cls_loss: tensor(13.8009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9433, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2560, device='cuda:0', grad_fn=) +cls_loss: tensor(18.4814, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9872, device='cuda:0', grad_fn=) +cls_loss: tensor(12.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4977, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(15.6644, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9265, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8037, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3372, device='cuda:0', grad_fn=) +cls_loss: tensor(14.7178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2789, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3239, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6081, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4523, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3825, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3672, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2898, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7176, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3679, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1543, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5835, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9294, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5901, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8289, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7026, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4831, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5354, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6640, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7059, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7082, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5568, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4647, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7613, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3796, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9888, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7546, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3891, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2903, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5960, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6260, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7917, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4195, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4440, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2416, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9753, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3217, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5893, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6595, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3892, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +9.994965332706573e-05 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 359.68, cls_loss 5.2840 +306 +cls_loss: tensor(4.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(18.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9527, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7201, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9018, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4564, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9733, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8075, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5371, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4890, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3892, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7513, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9043, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5197, device='cuda:0', grad_fn=) +cls_loss: tensor(17.6563, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9535, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9661, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5154, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3395, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7267, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8293, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4554, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7233, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6647, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3291, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4698, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7439, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4531, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7253, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7917, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3112, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6922, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3156, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4175, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6609, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7383, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4315, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4275e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7180, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9834, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3737, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9362, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3106, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4318, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5574, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6218, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5427, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1918, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6238, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4277, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6668, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4163, device='cuda:0', grad_fn=) +cls_loss: tensor(23.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3890, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2776, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5921, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8676, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5537, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6854, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7518, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5695, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8334, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4036, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5941, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4365, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3669, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3553, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4818, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8445, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8661, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7623, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2723, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9333, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3979, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2911, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9208, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7165, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2890, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6632, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8350, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9732, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9526, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2122, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9873, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7555, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2920, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5839, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4355, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4658, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9144, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5426, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3563, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6396, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3656, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4375, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6837, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4456, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8783, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3419, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7488, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9503, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4697, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3474, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6390, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9421, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2569, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8340, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3245, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4557, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4206, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7426, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9577, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5373, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4457, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4724, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6822, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6164, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7462, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2069, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6196, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7313, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0000, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4252, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3613, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +9.979871469976196e-05 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 358.51, cls_loss 2.9761 +306 +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4946, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9401, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2587, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8107, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2992, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2457, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6772, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4826, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5631, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8100, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5709, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9710, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7950, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7234, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8062, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5356, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5576, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4742, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5028, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6198, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6087, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5248, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7891, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6838, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3789, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4290, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8666, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2978, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4821, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9357, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7187, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8070, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9719, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8874, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8210, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6058, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5576, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8929, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8066, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6661, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5615, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3268, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1729, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3751, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6548, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1966, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2745, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4375, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6211, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8164, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3884, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6213, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6065, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5652, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9499, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3973, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7993, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9766, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7598, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9497, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5078, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3425, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6807, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5688, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2547, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9466, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7602, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7506, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3379, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0209, device='cuda:0', grad_fn=) +9.954748808839674e-05 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 357.65, cls_loss 1.7143 +306 +cls_loss: tensor(1.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3650, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9296, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9838, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5425, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7494, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4331, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5483, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9107, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3923, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4797, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8021, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8210, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5716, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9979, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7574, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4226, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9645, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7293, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3786, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7422, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7454, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6803, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3758, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6864, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3972, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2873, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9671, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8283, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5140, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4099, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6804, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6352, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3169, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8577, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3290, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9726, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5095, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4475, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2647, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8483, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8582, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1974, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9610, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9820, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6374, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8713, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5967, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5426, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5642, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8281, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9924, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9126, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +9.919647942993148e-05 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 360.90, cls_loss 1.1780 +306 +cls_loss: tensor(2.0266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5977, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7585, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4850, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5827, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8213, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6796, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6728, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1839, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3812, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2584, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4482, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5963e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8845, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5542, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7868, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5614, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3180, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7376, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2886, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9493e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6205, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7616, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6377, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8426, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5199, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6751, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9437, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4824, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9828, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4092, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9669, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5690, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3783, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4721, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6373, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5433, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4841, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6342, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8622, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6715, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9799, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5841, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4038, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9428, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3446, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9307, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4830, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7195, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8451, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7812, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5583, device='cuda:0', grad_fn=) +9.874639560909117e-05 +changing lr +epoch 5, time 357.06, cls_loss 0.9013 +306 +cls_loss: tensor(0.5313, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4755, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3986, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8151, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6320, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3757, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4429, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2710, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8611, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8968, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4225, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3096, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9903, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8119, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2217, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2381, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7416, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3609, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1638, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6810, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6833, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6211, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7355, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3817, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8399, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3896, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6599, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3449, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3313, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5247, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8027, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9178, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9925, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8092, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3359, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2791, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4389, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9060, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6001e-05, device='cuda:0', grad_fn=) +9.819814303479267e-05 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 355.43, cls_loss 0.7067 +306 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5811, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7354, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7422, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5485, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5230, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7240, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9316, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9144, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4084, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3121, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3864, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9643, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5782, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5993, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3861, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7643, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9060, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6668, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5507, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6603, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4674, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2555, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2874, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2694, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1635, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3127e-05, device='cuda:0', grad_fn=) +9.755282581475769e-05 +changing lr +epoch 7, time 354.12, cls_loss 0.4559 +306 +cls_loss: tensor(0.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8073, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3897, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9997e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7902, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9347, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3610, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3392, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8848, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4962, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6192, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3145, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4056, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4772, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3942, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3988, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2560, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7756e-05, device='cuda:0', grad_fn=) +9.681174353198687e-05 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 358.80, cls_loss 0.3400 +306 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4342, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9971, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2899, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7840, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9082, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5859, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7451, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4602, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6712, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3590, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5934, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5882, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2431, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8573, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2813, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8520, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3726, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6986, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3206, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2611, device='cuda:0', grad_fn=) +9.597638862757255e-05 +changing lr +epoch 9, time 371.01, cls_loss 0.2145 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5150, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4959, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8857, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4271, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8667, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4418, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9786, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6086, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2128, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1952, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6431, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6244, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5844, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6842, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7956, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5156, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4433, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9779, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6239, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4562, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4823, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3685, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8088, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +9.504844339512095e-05 +changing lr +epoch 10, time 357.40, cls_loss 0.3486 +306 +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6061, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8122, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9397, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9831, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9356, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7428, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2501, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5331, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6335, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5746, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4129, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9310, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5633, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2443, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4395, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4938, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +9.40297765928369e-05 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 358.94, cls_loss 0.1780 +306 +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3231, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6416, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3144, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3705, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2759, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8737, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2297, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7507, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2529, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4769, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3581, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9362, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6335, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9095, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4958, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3406, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3159, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4535, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4437, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3382, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2690, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1910, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8269, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6992, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9798e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3679, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4352, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7227, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3803, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +9.292243968009331e-05 +changing lr +epoch 12, time 356.09, cls_loss 0.2509 +306 +cls_loss: tensor(8.3804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2942, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7708, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7453, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5453, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5646e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5150, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2779, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3975, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5499, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2926, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3444, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6398, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7754, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3219, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4276, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2670, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4126, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1716, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7743, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6418, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3228, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9940, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1879, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3092, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3045, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3041, device='cuda:0', grad_fn=) +9.172866268606513e-05 +changing lr +epoch 13, time 358.04, cls_loss 0.2684 +306 +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6300, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6436, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5581, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5625, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4988, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4419, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6821, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2849, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9087, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2549, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3455, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4586, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6827, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.045084971874738e-05 +changing lr +epoch 14, time 357.34, cls_loss 0.1243 +306 +cls_loss: tensor(1.6669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1637, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5815, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3273, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9450, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7728, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2642, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5129, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5414, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8473e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7467, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3104, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7492, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6543, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8022, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +8.90915741234015e-05 +changing lr +epoch 15, time 357.89, cls_loss 0.1362 +306 +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5378, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4570, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6491, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4541, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2264, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2777, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9505, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4194, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +8.765357330018056e-05 +changing lr +epoch 16, time 360.08, cls_loss 0.0830 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3477, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7910, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7516, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2151, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5531, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3146, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7246, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9513e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1768, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7150, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5120, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3652, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3339, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7760, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5744e-05, device='cuda:0', grad_fn=) +8.613974319136958e-05 +changing lr +epoch 17, time 358.06, cls_loss 0.1141 +306 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6240, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2077, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9775, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9918, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6656, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4377, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +8.455313244934324e-05 +changing lr +epoch 18, time 357.52, cls_loss 0.0474 +306 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4052, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3801, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2905, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4382, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4862, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8555, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2298, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8473, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8601, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +8.289693629698562e-05 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 356.57, cls_loss 0.0406 +306 +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5844e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8189, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2900, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3408, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7798e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2494, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +8.117449009293667e-05 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 357.61, cls_loss 0.0392 +306 +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9306, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3406e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5999, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2688, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1987, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3571, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2330, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9180, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9180, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5809, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +7.938926261462365e-05 +changing lr +epoch 21, time 358.54, cls_loss 0.0412 +306 +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7533, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +7.754484907260511e-05 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 357.59, cls_loss 0.0210 +306 +cls_loss: tensor(5.3247e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3529, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9232, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3581, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2513, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6505e-05, device='cuda:0', grad_fn=) +7.56449638702953e-05 +changing lr +epoch 23, time 360.75, cls_loss 0.0187 +306 +cls_loss: tensor(3.4332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2909, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9375, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2314, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3314, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4976, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +7.369343312364992e-05 +changing lr +epoch 24, time 356.58, cls_loss 0.0428 +306 +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8075, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6346, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7819, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4486, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3486, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5062, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +7.169418695587788e-05 +changing lr +epoch 25, time 365.69, cls_loss 0.0531 +306 +cls_loss: tensor(7.1824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5657, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8123, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4821, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4606, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6934, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8678, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3685, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +6.965125158269616e-05 +changing lr +epoch 26, time 358.04, cls_loss 0.0377 +306 +cls_loss: tensor(0.7696, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1550, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5844e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +6.756874120406712e-05 +changing lr +epoch 27, time 357.21, cls_loss 0.0154 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8050e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9548, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1625, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9913, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5036e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5274, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7275e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5051, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0339e-05, device='cuda:0', grad_fn=) +6.545084971874736e-05 +changing lr +epoch 28, time 358.61, cls_loss 0.0293 +306 +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2347, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1918, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5798, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2678, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +6.330184227833373e-05 +changing lr +epoch 29, time 355.96, cls_loss 0.0227 +306 +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6957e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6369, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6957e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3789, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +6.11260466978157e-05 +changing lr +epoch 30, time 354.42, cls_loss 0.0082 +306 +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3519, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5116, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +5.892784473993181e-05 +changing lr +epoch 31, time 357.66, cls_loss 0.0135 +306 +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8308e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8561, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +5.6711663290882756e-05 +changing lr +epoch 32, time 359.43, cls_loss 0.0168 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1794, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4275e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5698e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1968, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3060, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4837e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2624, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5554, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +5.448196544517166e-05 +changing lr +epoch 33, time 358.22, cls_loss 0.0171 +306 +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5932, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +5.2243241517525733e-05 +changing lr +epoch 34, time 357.39, cls_loss 0.0185 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5447, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2626, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2910, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2430, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +4.999999999999998e-05 +changing lr +epoch 35, time 360.38, cls_loss 0.0109 +306 +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2016e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9883e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +4.7756758482474244e-05 +changing lr +epoch 36, time 356.36, cls_loss 0.0026 +306 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7285, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3686, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4163, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +4.551803455482832e-05 +changing lr +epoch 37, time 356.94, cls_loss 0.0074 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4254, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8282, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.328833670911722e-05 +changing lr +epoch 38, time 361.18, cls_loss 0.0128 +306 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5622, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4593e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +4.1072155260068164e-05 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 358.80, cls_loss 0.0086 +306 +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8247e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +3.8873953302184275e-05 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 361.67, cls_loss 0.0076 +306 +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1831, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4886, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3366, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-05, device='cuda:0', grad_fn=) +3.669815772166625e-05 +changing lr +epoch 41, time 357.39, cls_loss 0.0055 +306 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4551, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8030e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4838, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1853, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +3.454915028125263e-05 +changing lr +epoch 42, time 356.34, cls_loss 0.0075 +306 +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2844e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +3.2431258795932867e-05 +changing lr +epoch 43, time 357.10, cls_loss 0.0085 +306 +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4837e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8493e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9142e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +3.0348748417303827e-05 +changing lr +epoch 44, time 356.52, cls_loss 0.0033 +306 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5963e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0016e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5068, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3386, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8507e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4866, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +2.830581304412209e-05 +changing lr +epoch 45, time 358.60, cls_loss 0.0138 +306 +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6595, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1841e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7120e-05, device='cuda:0', grad_fn=) +2.6306566876350062e-05 +changing lr +epoch 46, time 355.85, cls_loss 0.0049 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +2.4355036129704693e-05 +changing lr +epoch 47, time 358.34, cls_loss 0.0063 +306 +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2694, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2415, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +2.2455150927394874e-05 +changing lr +epoch 48, time 354.26, cls_loss 0.0070 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8666e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9606e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6920, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +2.0610737385376345e-05 +changing lr +epoch 49, time 358.20, cls_loss 0.0036 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5520, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6883e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3997e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2213, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +1.8825509907063323e-05 +changing lr +epoch 50, time 356.73, cls_loss 0.0042 +306 +cls_loss: tensor(6.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3838e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3223, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1016e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +1.710306370301437e-05 +changing lr +epoch 51, time 356.93, cls_loss 0.0081 +306 +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2304, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.5446867550656765e-05 +changing lr +epoch 52, time 357.03, cls_loss 0.0026 +306 +cls_loss: tensor(2.4100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3657, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +1.3860256808630425e-05 +changing lr +epoch 53, time 357.45, cls_loss 0.0030 +306 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7145, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9016e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9606e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1769e-05, device='cuda:0', grad_fn=) +1.2346426699819455e-05 +changing lr +epoch 54, time 356.20, cls_loss 0.0051 +306 +cls_loss: tensor(9.9719e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7904, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8694, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3486e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +1.0908425876598507e-05 +changing lr +epoch 55, time 356.09, cls_loss 0.0136 +306 +cls_loss: tensor(3.6001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5698e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7285, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3977e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3108e-05, device='cuda:0', grad_fn=) +9.549150281252631e-06 +changing lr +epoch 56, time 355.81, cls_loss 0.0036 +306 +cls_loss: tensor(2.1438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1968, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2864, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2734, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +8.271337313934865e-06 +changing lr +epoch 57, time 361.23, cls_loss 0.0051 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2637, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8825e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +7.077560319906693e-06 +changing lr +epoch 58, time 364.07, cls_loss 0.0025 +306 +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7247e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3026, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6823e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +5.970223407163098e-06 +changing lr +epoch 59, time 356.49, cls_loss 0.0036 +306 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2596e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1841e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8348e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4066, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +4.9515566048790464e-06 +changing lr +epoch 60, time 355.95, cls_loss 0.0089 +306 +cls_loss: tensor(2.6246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0473e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5036e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5698e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.0236113724274705e-06 +changing lr +epoch 61, time 354.62, cls_loss 0.0018 +306 +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6823e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7613e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.188256468013139e-06 +changing lr +epoch 62, time 356.37, cls_loss 0.0015 +306 +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2732, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6401e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +2.4471741852423225e-06 +changing lr +epoch 63, time 356.08, cls_loss 0.0047 +306 +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1275e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +1.8018569652073373e-06 +changing lr +epoch 64, time 360.71, cls_loss 0.0015 +306 +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5487e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2282, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8586e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2870, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3774, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +1.2536043909088185e-06 +changing lr +epoch 65, time 356.73, cls_loss 0.0081 +306 +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014e-05, device='cuda:0', grad_fn=) +8.035205700685162e-07 +changing lr +epoch 66, time 355.86, cls_loss 0.0052 +306 +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +4.525119116032647e-07 +changing lr +epoch 67, time 360.93, cls_loss 0.0015 +306 +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5036e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2446e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +2.0128530023804648e-07 +changing lr +epoch 68, time 357.71, cls_loss 0.0022 +306 +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.03466729342705e-08 +changing lr +epoch 69, time 357.42, cls_loss 0.0029 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +x.shape: (2048, 227, 227, 3) +x_aug test here torch.Size([2048, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +x.shape: (2344, 227, 227, 3) +x_aug test here torch.Size([2344, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +x.shape: (1670, 227, 227, 3) +x_aug test here torch.Size([1670, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +x.shape: (3929, 227, 227, 3) +x_aug test here torch.Size([3929, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 88.916016 58.020478 81.497006 50.572665 63.363383 diff --git a/Meta-causal/code-stage1-pipeline/56456.error b/Meta-causal/code-stage1-pipeline/56456.error new file mode 100644 index 0000000000000000000000000000000000000000..b85c3d299f0714adaa84ad7c6716dd4d92fba1c2 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56456.error @@ -0,0 +1,3 @@ +slurmstepd: error: *** JOB 56456 ON gcpl4-eu-1 CANCELLED AT 2024-07-04T07:05:01 DUE TO TIME LIMIT *** +slurmstepd: error: *** STEP 56456.0 ON gcpl4-eu-1 CANCELLED AT 2024-07-04T07:05:01 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-stage1-pipeline/56456.log b/Meta-causal/code-stage1-pipeline/56456.log new file mode 100644 index 0000000000000000000000000000000000000000..0219e07a418a32f1d9a7d4547364102ea57bbb24 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56456.log @@ -0,0 +1,24798 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +x.shape: torch.Size([10000, 3, 32, 32]) +x_aug train here torch.Size([10000, 3, 32, 32]) +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +cls_loss: tensor(2.2975, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3231, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3059, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3236, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2853, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3042, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3150, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2800, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2253, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2585, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2845, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2509, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2468, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2679, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1811, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2562, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2691, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2551, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1938, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9462, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9986, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1789, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8458, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7802, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9741, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3087, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6669, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9964, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9678, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7691, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9197, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3908, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8969, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8429, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8275, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8880, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8747, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8223, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9431, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7663, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7526, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8128, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8826, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8587, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8089, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7867, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6653, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6441, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2972, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4918, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4246, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7241, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9828, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5358, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 184.48, cls_loss 2.0572 +100 +cls_loss: tensor(1.5558, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7339, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6572, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7955, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5656, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6640, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6369, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7916, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8687, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8726, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7140, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6356, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4700, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7803, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7543, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6353, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7704, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7423, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8626, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8885, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5303, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7471, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6653, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7721, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6090, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8727, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6788, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5714, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7308, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4991, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7237, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9436, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5769, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4804, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8779, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6535, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4428, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6484, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2972, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4363, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5876, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5480, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4144, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6745, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7365, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7442, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6890, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3278, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4912, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2433, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8144, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5466, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5886, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4970, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7452, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3988, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9478, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5276, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5203, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6601, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6790, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5301, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6815, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4725, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6399, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8933, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 1, time 182.25, cls_loss 1.6321 +100 +cls_loss: tensor(1.5194, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6739, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5537, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5348, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6299, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2347, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6907, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5375, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3416, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4267, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7357, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5554, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2837, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5432, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2508, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6813, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6465, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3554, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7530, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5418, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5738, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5523, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3862, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3293, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5324, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4950, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6649, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5249, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3481, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2489, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4359, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6905, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7705, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3680, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3642, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3348, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3994, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6713, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2854, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2764, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5958, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8064, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5644, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3137, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5517, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7766, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2651, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3798, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3716, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2471, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2960, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5738, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3468, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4361, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4701, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2084, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 2, time 183.56, cls_loss 1.4163 +100 +cls_loss: tensor(1.2069, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5396, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3433, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2811, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5569, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2881, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2713, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5636, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5399, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3915, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3287, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3215, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4508, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3977, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9564, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3217, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2509, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3645, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2879, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7460, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3826, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9851, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3677, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3825, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2529, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3824, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4124, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9410, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9933, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4434, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2759, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4408, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2818, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4261, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5453, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1887, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9539, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8422, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2927, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8426, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7444, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8914, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4134, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9158, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5259, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6527, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3991, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 3, time 183.41, cls_loss 1.2737 +100 +cls_loss: tensor(1.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3728, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3733, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2809, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1401, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2173, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4186, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4988, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3607, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5702, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2513, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2564, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3455, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3432, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3374, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2739, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4722, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9820, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6374, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3555, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4506, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4717, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1870, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2545, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4576, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1791, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9187, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4598, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3579, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8718, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3409, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4540, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9445, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4626, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3885, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9774, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2804, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3710, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8588, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5531, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 4, time 182.84, cls_loss 1.2137 +100 +cls_loss: tensor(1.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7465, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3836, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5348, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3485, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9572, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2449, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6198, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5188, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4449, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9741, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2913, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9545, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3584, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4447, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7857, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9891, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9698, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9213, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5602, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8956, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8956, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9971, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6908, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2425, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9286, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9249, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1763, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7522, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3682, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9903, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2986, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3592, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8798, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1425, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 5, time 182.60, cls_loss 1.1255 +100 +cls_loss: tensor(0.9153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3387, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2859, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9628, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1789, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7546, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5997, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7247, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1431, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6286, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8442, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8719, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8416, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8896, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8991, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9545, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7481, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8663, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9185, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6829, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8510, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4671, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9226, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7333, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8559, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9152, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0637, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 6, time 182.96, cls_loss 1.0495 +100 +cls_loss: tensor(1.4299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7472, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8519, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7780, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8589, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6793, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9476, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8188, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8239, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9168, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3160, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8309, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8981, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7736, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8981, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3133, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7757, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7635, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9410, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9382, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9530, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2697, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5791, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9980, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 7, time 182.96, cls_loss 1.0033 +100 +cls_loss: tensor(0.8280, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5899, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8529, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8575, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8397, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8940, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7841, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9476, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9450, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6166, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9523, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5962, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7363, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7894, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7233, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7336, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3301, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2532, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8574, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7841, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9277, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 8, time 183.71, cls_loss 0.9300 +100 +cls_loss: tensor(1.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4587, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8432, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9095, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9196, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8194, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8196, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6628, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6995, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5288, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9283, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9595, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6140, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9192, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8246, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8172, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1883, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 9, time 183.26, cls_loss 0.8857 +100 +cls_loss: tensor(0.8764, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2856, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2707, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6473, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8751, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6308, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4957, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7619, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6814, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5358, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9228, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9651, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8448, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8344, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7179, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 10, time 183.13, cls_loss 0.8474 +100 +cls_loss: tensor(0.5607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7123, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6641, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8456, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7634, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5377, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8640, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8392, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7986, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7308, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7741, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5433, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7300, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 183.26, cls_loss 0.8124 +100 +cls_loss: tensor(0.8014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8058, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4960, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8641, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6800, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6644, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7287, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 12, time 183.80, cls_loss 0.7501 +100 +cls_loss: tensor(1.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9742, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7283, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7378, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5614, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9151, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6537, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 13, time 183.47, cls_loss 0.7426 +100 +cls_loss: tensor(0.9129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4397, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4866, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5814, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8074, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 14, time 182.69, cls_loss 0.6840 +100 +cls_loss: tensor(0.4868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9876, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7337, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 15, time 182.20, cls_loss 0.6795 +100 +cls_loss: tensor(0.6640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5949, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4358, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 16, time 182.74, cls_loss 0.5845 +100 +cls_loss: tensor(0.6800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3621, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4684, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7715, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 17, time 182.30, cls_loss 0.5823 +100 +cls_loss: tensor(0.5717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4763, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7051, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 18, time 182.94, cls_loss 0.5565 +100 +cls_loss: tensor(0.8004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5268, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6726, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 19, time 183.04, cls_loss 0.5575 +100 +cls_loss: tensor(0.6149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4592, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 20, time 182.50, cls_loss 0.5078 +100 +cls_loss: tensor(0.3351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4991, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 21, time 182.53, cls_loss 0.4780 +100 +cls_loss: tensor(0.5946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7080, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 22, time 182.15, cls_loss 0.4635 +100 +cls_loss: tensor(0.4381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2347, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 23, time 180.67, cls_loss 0.4375 +100 +cls_loss: tensor(0.3127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3886, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 24, time 180.55, cls_loss 0.4255 +100 +cls_loss: tensor(0.4367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 25, time 180.21, cls_loss 0.3821 +100 +cls_loss: tensor(0.4193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3341, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 26, time 180.73, cls_loss 0.3916 +100 +cls_loss: tensor(0.3168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2318, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 27, time 180.85, cls_loss 0.3633 +100 +cls_loss: tensor(0.4510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1950, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 28, time 181.01, cls_loss 0.3164 +100 +cls_loss: tensor(0.4032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4149, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 29, time 180.94, cls_loss 0.3405 +100 +cls_loss: tensor(0.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2549, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 30, time 180.76, cls_loss 0.3251 +100 +cls_loss: tensor(0.3253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 31, time 180.98, cls_loss 0.3065 +100 +cls_loss: tensor(0.2163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2159, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 32, time 180.92, cls_loss 0.2573 +100 +cls_loss: tensor(0.5187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 33, time 180.55, cls_loss 0.2864 +100 +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 34, time 181.01, cls_loss 0.2477 +100 +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2051, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 35, time 180.79, cls_loss 0.2699 +100 +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 36, time 180.57, cls_loss 0.2638 +100 +cls_loss: tensor(0.4190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2787, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 37, time 180.99, cls_loss 0.2365 +100 +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 38, time 181.18, cls_loss 0.2396 +100 +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 39, time 180.88, cls_loss 0.2110 +100 +cls_loss: tensor(0.3090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 40, time 180.62, cls_loss 0.1737 +100 +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2040, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 41, time 180.57, cls_loss 0.2038 +100 +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 42, time 180.72, cls_loss 0.1968 +100 +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 43, time 180.74, cls_loss 0.1963 +100 +cls_loss: tensor(0.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 44, time 180.85, cls_loss 0.1777 +100 +cls_loss: tensor(0.2576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 45, time 180.80, cls_loss 0.1613 +100 +cls_loss: tensor(0.3567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 46, time 181.06, cls_loss 0.1712 +100 +cls_loss: tensor(0.2654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 47, time 181.14, cls_loss 0.1416 +100 +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 48, time 180.71, cls_loss 0.1491 +100 +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1755, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 49, time 180.66, cls_loss 0.1595 +100 +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1836, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 50, time 180.78, cls_loss 0.1485 +100 +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 51, time 180.70, cls_loss 0.1875 +100 +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 52, time 180.53, cls_loss 0.1500 +100 +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 53, time 180.72, cls_loss 0.1404 +100 +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2732, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 54, time 180.79, cls_loss 0.1201 +100 +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 55, time 180.78, cls_loss 0.1276 +100 +cls_loss: tensor(0.1962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 56, time 180.85, cls_loss 0.1206 +100 +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 57, time 180.62, cls_loss 0.1216 +100 +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 58, time 180.77, cls_loss 0.1272 +100 +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 59, time 180.92, cls_loss 0.1357 +100 +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 60, time 180.26, cls_loss 0.1066 +100 +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3367, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 61, time 180.80, cls_loss 0.1077 +100 +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 62, time 180.84, cls_loss 0.1309 +100 +cls_loss: tensor(0.3423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1401, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 63, time 180.60, cls_loss 0.1120 +100 +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 64, time 181.00, cls_loss 0.1222 +100 +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 65, time 180.81, cls_loss 0.1110 +100 +cls_loss: tensor(0.2858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 66, time 180.77, cls_loss 0.1078 +100 +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 67, time 180.73, cls_loss 0.1101 +100 +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 68, time 180.92, cls_loss 0.0807 +100 +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 69, time 180.74, cls_loss 0.0841 +100 +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 70, time 180.29, cls_loss 0.0776 +100 +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 71, time 180.49, cls_loss 0.0906 +100 +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 72, time 180.89, cls_loss 0.1042 +100 +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 73, time 180.75, cls_loss 0.0955 +100 +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 74, time 180.40, cls_loss 0.0924 +100 +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1636, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 75, time 180.41, cls_loss 0.0954 +100 +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 76, time 180.39, cls_loss 0.0967 +100 +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 77, time 180.43, cls_loss 0.0833 +100 +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 78, time 180.16, cls_loss 0.0689 +100 +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 79, time 180.22, cls_loss 0.0866 +100 +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 80, time 180.45, cls_loss 0.0876 +100 +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 81, time 180.23, cls_loss 0.0722 +100 +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 82, time 180.44, cls_loss 0.0746 +100 +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 83, time 180.69, cls_loss 0.0826 +100 +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 84, time 180.71, cls_loss 0.0726 +100 +cls_loss: tensor(0.1768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 85, time 180.60, cls_loss 0.0607 +100 +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 86, time 180.74, cls_loss 0.0742 +100 +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 87, time 180.07, cls_loss 0.0698 +100 +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 88, time 180.15, cls_loss 0.0637 +100 +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 89, time 180.65, cls_loss 0.0693 +100 +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 90, time 180.47, cls_loss 0.0744 +100 +cls_loss: tensor(0.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 91, time 180.53, cls_loss 0.0857 +100 +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 92, time 180.45, cls_loss 0.0753 +100 +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 93, time 180.53, cls_loss 0.0573 +100 +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 94, time 180.19, cls_loss 0.0624 +100 +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 95, time 180.07, cls_loss 0.0549 +100 +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 96, time 180.14, cls_loss 0.0583 +100 +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 97, time 180.57, cls_loss 0.0686 +100 +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 98, time 180.56, cls_loss 0.0737 +100 +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 99, time 180.29, cls_loss 0.0943 +100 +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 100, time 180.17, cls_loss 0.0965 +100 +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 101, time 180.48, cls_loss 0.0919 +100 +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 102, time 180.69, cls_loss 0.0643 +100 +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 103, time 180.79, cls_loss 0.0469 +100 +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 104, time 180.41, cls_loss 0.0439 +100 +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 105, time 180.22, cls_loss 0.0617 +100 +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 106, time 180.38, cls_loss 0.0562 +100 +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 107, time 180.43, cls_loss 0.0510 +100 +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 108, time 180.59, cls_loss 0.0550 +100 +cls_loss: tensor(0.2173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 109, time 180.65, cls_loss 0.0556 +100 +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 110, time 180.21, cls_loss 0.0514 +100 +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 111, time 180.58, cls_loss 0.0699 +100 +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 112, time 180.32, cls_loss 0.0705 +100 +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 113, time 180.52, cls_loss 0.0669 +100 +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 114, time 180.25, cls_loss 0.0758 +100 +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 115, time 180.34, cls_loss 0.0879 +100 +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 116, time 180.64, cls_loss 0.0596 +100 +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 117, time 180.49, cls_loss 0.0614 +100 +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 118, time 180.43, cls_loss 0.0585 +100 +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 119, time 180.82, cls_loss 0.0526 +100 +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 120, time 180.33, cls_loss 0.0412 +100 +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 121, time 180.50, cls_loss 0.0300 +100 +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 122, time 180.39, cls_loss 0.0516 +100 +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 123, time 180.64, cls_loss 0.0569 +100 +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1806, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 124, time 180.72, cls_loss 0.0822 +100 +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 125, time 180.57, cls_loss 0.0957 +100 +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 126, time 180.33, cls_loss 0.0616 +100 +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 127, time 180.60, cls_loss 0.0773 +100 +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 128, time 180.13, cls_loss 0.0676 +100 +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 129, time 180.46, cls_loss 0.0579 +100 +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 130, time 180.34, cls_loss 0.0499 +100 +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 131, time 180.47, cls_loss 0.0486 +100 +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 132---------------------------------------------------- +epoch 132, time 180.57, cls_loss 0.0416 +100 +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 133, time 180.39, cls_loss 0.0443 +100 +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 134, time 180.42, cls_loss 0.0399 +100 +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 135, time 180.67, cls_loss 0.0452 +100 +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 136, time 180.61, cls_loss 0.0362 +100 +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1289, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 137, time 180.79, cls_loss 0.0508 +100 +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 138, time 181.22, cls_loss 0.0511 +100 +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 139, time 180.59, cls_loss 0.0421 +100 +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 140, time 180.32, cls_loss 0.0275 +100 +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 141, time 180.31, cls_loss 0.0324 +100 +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 142, time 180.23, cls_loss 0.0257 +100 +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 143, time 181.43, cls_loss 0.0335 +100 +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 144, time 180.43, cls_loss 0.0344 +100 +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 145, time 180.23, cls_loss 0.0589 +100 +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 146, time 180.44, cls_loss 0.0921 +100 +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1676, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 147, time 180.31, cls_loss 0.0741 +100 +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 148, time 180.17, cls_loss 0.0496 +100 +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 149, time 180.29, cls_loss 0.0786 +100 +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2305, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 150, time 180.50, cls_loss 0.0562 +100 +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 151, time 180.18, cls_loss 0.0525 +100 +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 152, time 180.08, cls_loss 0.0569 +100 +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 153, time 180.41, cls_loss 0.0456 +100 +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 154, time 180.49, cls_loss 0.0422 +100 +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 155, time 180.16, cls_loss 0.0341 +100 +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 156, time 180.28, cls_loss 0.0499 +100 +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 157, time 180.11, cls_loss 0.0500 +100 +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 158, time 179.91, cls_loss 0.0545 +100 +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 159, time 180.14, cls_loss 0.0288 +100 +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 160, time 180.46, cls_loss 0.0381 +100 +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 161, time 180.24, cls_loss 0.0295 +100 +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7064e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 162, time 180.45, cls_loss 0.0276 +100 +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 163, time 180.50, cls_loss 0.0598 +100 +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 164, time 179.98, cls_loss 0.0356 +100 +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 165, time 180.25, cls_loss 0.0409 +100 +cls_loss: tensor(0.0956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 166, time 180.21, cls_loss 0.0437 +100 +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 167, time 180.17, cls_loss 0.0596 +100 +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 168, time 180.37, cls_loss 0.0458 +100 +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 169, time 180.19, cls_loss 0.0429 +100 +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 170, time 180.15, cls_loss 0.0411 +100 +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 171, time 180.45, cls_loss 0.0581 +100 +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 172, time 180.02, cls_loss 0.0511 +100 +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 173, time 180.13, cls_loss 0.0536 +100 +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 174, time 180.37, cls_loss 0.0752 +100 +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 175, time 180.28, cls_loss 0.0458 +100 +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 176, time 180.26, cls_loss 0.0505 +100 +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 177, time 180.48, cls_loss 0.0529 +100 +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 178, time 180.32, cls_loss 0.0321 +100 +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 179, time 180.33, cls_loss 0.0502 +100 +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 180, time 180.28, cls_loss 0.0428 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5527e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1761e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 181, time 180.35, cls_loss 0.0185 +100 +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1700e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 182, time 180.06, cls_loss 0.0326 +100 +cls_loss: tensor(9.0167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 183, time 180.02, cls_loss 0.0301 +100 +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 184, time 180.22, cls_loss 0.0315 +100 +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 185, time 180.08, cls_loss 0.0269 +100 +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 186, time 180.11, cls_loss 0.0360 +100 +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5480, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 187, time 180.14, cls_loss 0.0498 +100 +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8137e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7596e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 188---------------------------------------------------- +epoch 188, time 180.24, cls_loss 0.0643 +100 +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 189, time 180.16, cls_loss 0.0654 +100 +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 190, time 180.39, cls_loss 0.0545 +100 +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 191, time 180.57, cls_loss 0.0361 +100 +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 192, time 180.43, cls_loss 0.0263 +100 +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0865e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 193, time 180.56, cls_loss 0.0433 +100 +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 194, time 180.25, cls_loss 0.0324 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 195, time 180.48, cls_loss 0.0177 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5646e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0243e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 196, time 180.02, cls_loss 0.0236 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0073e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 197, time 180.89, cls_loss 0.0201 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6946e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8163e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6428e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4242e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3556e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2960e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 198, time 181.50, cls_loss 0.0188 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6448e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 199, time 180.80, cls_loss 0.0253 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6422e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5747e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7474e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 200, time 180.90, cls_loss 0.0206 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0878e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7488e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 201, time 181.38, cls_loss 0.0196 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4576e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 202, time 181.31, cls_loss 0.0256 +100 +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1304e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5706e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0998e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1569e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 203, time 181.27, cls_loss 0.0327 +100 +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 204---------------------------------------------------- +epoch 204, time 181.55, cls_loss 0.0568 +100 +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 205, time 180.91, cls_loss 0.0628 +100 +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 206, time 181.19, cls_loss 0.0607 +100 +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 207, time 180.78, cls_loss 0.0429 +100 +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 208, time 181.24, cls_loss 0.0450 +100 +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 209, time 180.83, cls_loss 0.0522 +100 +cls_loss: tensor(0.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 210, time 180.63, cls_loss 0.0631 +100 +cls_loss: tensor(4.4130e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2714e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 211, time 180.97, cls_loss 0.0368 +100 +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 212, time 181.14, cls_loss 0.0274 +100 +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 213, time 180.93, cls_loss 0.0278 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3441e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 214, time 180.98, cls_loss 0.0331 +100 +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 215, time 181.10, cls_loss 0.0366 +100 +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6397e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 216, time 181.17, cls_loss 0.0303 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2012e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8852e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5887e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4728e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 217, time 181.11, cls_loss 0.0236 +100 +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1825e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 218, time 180.81, cls_loss 0.0342 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2364e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4635e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 219, time 181.15, cls_loss 0.0340 +100 +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6927e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 220, time 181.03, cls_loss 0.0428 +100 +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5706e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7092e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 221, time 181.32, cls_loss 0.0363 +100 +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6323e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 222, time 181.27, cls_loss 0.0367 +100 +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7999e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 223, time 180.99, cls_loss 0.0284 +100 +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5613e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3589e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3607e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 224, time 181.11, cls_loss 0.0263 +100 +cls_loss: tensor(0.2828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1586e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2958e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 225, time 180.71, cls_loss 0.0252 +100 +cls_loss: tensor(6.4332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 226, time 180.76, cls_loss 0.0326 +100 +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2374, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2999e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5554e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4599, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 227, time 180.91, cls_loss 0.0383 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 228, time 181.09, cls_loss 0.0225 +100 +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5291e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5138e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 229, time 180.89, cls_loss 0.0257 +100 +cls_loss: tensor(4.4182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 230---------------------------------------------------- +epoch 230, time 181.23, cls_loss 0.0499 +100 +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 231, time 180.47, cls_loss 0.0463 +100 +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4176e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 232, time 180.35, cls_loss 0.0335 +100 +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1283e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 233, time 180.16, cls_loss 0.0408 +100 +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 234, time 180.29, cls_loss 0.0530 +100 +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2348e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 235, time 180.13, cls_loss 0.0467 +100 +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 236, time 179.66, cls_loss 0.0509 +100 +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 237, time 179.65, cls_loss 0.0516 +100 +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) diff --git a/Meta-causal/code-stage1-pipeline/56457.error b/Meta-causal/code-stage1-pipeline/56457.error new file mode 100644 index 0000000000000000000000000000000000000000..38d934231566fdb9cd98197f76b6dd49f332d93a --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56457.error @@ -0,0 +1,4 @@ +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) diff --git a/Meta-causal/code-stage1-pipeline/56457.log b/Meta-causal/code-stage1-pipeline/56457.log new file mode 100644 index 0000000000000000000000000000000000000000..39719eeab01cfbb421df95d8ac145de4111e49a6 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56457.log @@ -0,0 +1,21742 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +x.shape: (1840, 227, 227, 3) +x_aug train here torch.Size([1840, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +x.shape: (208, 227, 227, 3) +x_aug test here torch.Size([208, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +cls_loss: tensor(1.8001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9299, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8794, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5091, device='cuda:0', grad_fn=) +cls_loss: tensor(25.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(25.7894, device='cuda:0', grad_fn=) +cls_loss: tensor(23.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(28.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(17.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7904, device='cuda:0', grad_fn=) +cls_loss: tensor(23.8932, device='cuda:0', grad_fn=) +cls_loss: tensor(27.6354, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1585, device='cuda:0', grad_fn=) +cls_loss: tensor(24.6112, device='cuda:0', grad_fn=) +cls_loss: tensor(29.7782, device='cuda:0', grad_fn=) +cls_loss: tensor(23.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(19.5303, device='cuda:0', grad_fn=) +cls_loss: tensor(19.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6807, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3949, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5959, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4301, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5330, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4234, device='cuda:0', grad_fn=) +cls_loss: tensor(15.7894, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9794, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3270, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8557, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2788, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(19.3808, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9408, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9294, device='cuda:0', grad_fn=) +cls_loss: tensor(15.5034, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(13.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4702, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3672, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3439, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4944, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2801, device='cuda:0', grad_fn=) +cls_loss: tensor(16.3740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9396, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2803, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5613, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2403, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4300, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(13.3665, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6107, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7223, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4867, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6742, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7194, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4206, device='cuda:0', grad_fn=) +cls_loss: tensor(19.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9691, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(25.3668, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7656, device='cuda:0', grad_fn=) +cls_loss: tensor(16.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4841, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3048, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6151, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8525, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2786, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4486, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4277, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5532, device='cuda:0', grad_fn=) +cls_loss: tensor(12.5544, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9449, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8611, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4457, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6114, device='cuda:0', grad_fn=) +cls_loss: tensor(11.6261, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6582, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4405, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6596, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(14.7926, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9336, device='cuda:0', grad_fn=) +cls_loss: tensor(12.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6289, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2154, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6102, device='cuda:0', grad_fn=) +cls_loss: tensor(17.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(12.3890, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(17.6497, device='cuda:0', grad_fn=) +cls_loss: tensor(19.8324, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6579, device='cuda:0', grad_fn=) +cls_loss: tensor(22.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(12.5198, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5854, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8422, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8657, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3396, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7211, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9114, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4384, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3797, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9216, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6545, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2565, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4142, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9802, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2859, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4630, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5263, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7301, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(15.2033, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4877, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6162, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7433, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7440, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6465, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4256, device='cuda:0', grad_fn=) +cls_loss: tensor(16.8358, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3587, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3293, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(25.3555, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5823, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6908, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2976, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5063, device='cuda:0', grad_fn=) +cls_loss: tensor(21.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4762, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4065, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5736, device='cuda:0', grad_fn=) +cls_loss: tensor(25.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4488, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9688, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7394, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5130, device='cuda:0', grad_fn=) +cls_loss: tensor(17., device='cuda:0', grad_fn=) +cls_loss: tensor(12.3238, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2204, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7247, device='cuda:0', grad_fn=) +cls_loss: tensor(13.3807, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(17.6940, device='cuda:0', grad_fn=) +cls_loss: tensor(17.4583, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7437, device='cuda:0', grad_fn=) +cls_loss: tensor(21.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2457, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5077, device='cuda:0', grad_fn=) +cls_loss: tensor(12.2505, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7308, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8458, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6159, device='cuda:0', grad_fn=) +cls_loss: tensor(12.9400, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4021, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2427, device='cuda:0', grad_fn=) +cls_loss: tensor(12.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3137, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7487, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2251, device='cuda:0', grad_fn=) +cls_loss: tensor(18.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(15.6161, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2637, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(15.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4290, device='cuda:0', grad_fn=) +cls_loss: tensor(16.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(26.5755, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8587, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6341, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4594, device='cuda:0', grad_fn=) +cls_loss: tensor(19.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7146, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3693, device='cuda:0', grad_fn=) +cls_loss: tensor(14.7995, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7735, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(11.5618, device='cuda:0', grad_fn=) +cls_loss: tensor(24.7995, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7440, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8859, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2269, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7375, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5599, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4910, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6345, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(13.3747, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5560, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7812, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6732, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8348, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4532, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4954, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8372, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8015, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6430, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2600, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4367, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3777, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4704, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6166, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6781, device='cuda:0', grad_fn=) +cls_loss: tensor(14.3766, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7828, device='cuda:0', grad_fn=) +cls_loss: tensor(11.7823, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4476, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9189, device='cuda:0', grad_fn=) +cls_loss: tensor(15.6693, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3355, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3373, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3844, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(13.7246, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7033, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3232, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7478, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4453, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7487, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9839, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3760, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4967, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3087, device='cuda:0', grad_fn=) +cls_loss: tensor(12.7029, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5404, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(10.3973, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 356.45, cls_loss 8.9120 +306 +cls_loss: tensor(12.5783, device='cuda:0', grad_fn=) +cls_loss: tensor(22.7766, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9323, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8330, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2163, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8922, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8695, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5859, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4982, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6702, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5626, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3901, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9973, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6489, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7367, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8149, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(26.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2807, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9345, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6540, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6501, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5496, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3737, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5476, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2122, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6587, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6156, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8307, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3903, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5666, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9927, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4385, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8707, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2324, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9331, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5441, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5386, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5826, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9526, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5171, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8998, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8954, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7513, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6133, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2598, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9991, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3633, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7163, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8482, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9323, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8767, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7917, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5763, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4408, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9213, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5732, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8871, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6235, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5885, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7035, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1399, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(14.8154, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3601, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3393, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8820, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2289, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7128, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6846, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(15.9635, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6947, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3711, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4599, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6156, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7633, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4592, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4883, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8040, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9838, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3702, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3389, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2246, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8317, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4335, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8617, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2567, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8420, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8324, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5365, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7856, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6839, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7114, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4601, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7890, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6108, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7713, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3770, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6908, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9342, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4126, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7928, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3644, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2582, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7899, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9704, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6732, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7110, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9154, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2632, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8044, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4193, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5802, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9354, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4477, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4794, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9518, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9652, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9329, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7445, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7990, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9779, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7214, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9364, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2801, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(10.8749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5449, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4098, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3384, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4398, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6305, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7559, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1887, device='cuda:0', grad_fn=) +cls_loss: tensor(15.2556, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3162, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6454, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5326, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9311, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9320, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5410, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6575, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6130, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4415, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8564, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7948, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6105, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7356, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4574, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2589, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8604, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8568, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9463, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7113, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5820, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3990, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5347, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4487, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2481, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4912, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9030, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6894, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5722, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3936, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3058, device='cuda:0', grad_fn=) +cls_loss: tensor(16.4623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6103, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6277, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9784, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7820, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3140, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9751, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7773, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9333, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4382, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6959, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(14.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9720, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9323, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9613, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8262, device='cuda:0', grad_fn=) +9.994965332706573e-05 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 373.22, cls_loss 4.8684 +306 +cls_loss: tensor(6.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7882, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3629, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9083, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3573, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7995, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9142, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9375, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3154, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5774, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4465, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8646, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3848, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3726, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4811, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8902, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9981, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2633, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7520, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6224, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9418, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6975, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7996, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3513, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2398, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6458, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3804, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3376, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5143, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5841, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5504, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4321, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6785, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6365, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7574, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5109, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5246, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7541, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6344, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5434, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9860, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9245, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9559, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6334, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8965, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8599, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9258, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4474, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4870, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4116, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5391, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4707, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4521, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8062, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8150, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7877, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1951, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5468, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8897, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3348, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4188, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(10.2733, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9387, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3305, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4222, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6645, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5771, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3361, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9867, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9570, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4741, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9220, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1592, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5677, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6237, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8459, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7244, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6690, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4662, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9333, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4509, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7645, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9559, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4033, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(17.3685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4135, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3727, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9391, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7061, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7064, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9017, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9208, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4548, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3556, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3601, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8664, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9316, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2784, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5548, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6626, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7637, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4199, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8817, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8763, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9830, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6348, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4459, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6888, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9313, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9805, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7759, device='cuda:0', grad_fn=) +9.979871469976196e-05 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 368.73, cls_loss 2.6936 +306 +cls_loss: tensor(2.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3620, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3187, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3892, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7732, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8465, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6966, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5840, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2713, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3739, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9073, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5326, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4838, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5873, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7632, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9507, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2887, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3034, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8521, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6338, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6533, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9062, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5681, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4870, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2670, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3672, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3676, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5205, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7178, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8306, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5777, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7142, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8495, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9914, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7738, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4067, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5664, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8327, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6696, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7822, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8301, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6590, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8909, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2426, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4818, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8729, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9746, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7868, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5840, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9102, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7000, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6290, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2768, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4765, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3956, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7313, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3857, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5573, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7967, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3860, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4243, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7123, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7910, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3882, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6215, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6108, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3992, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4994, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2471, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7378, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3758, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4928, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3324, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8192, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5856, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8607, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3800, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3252, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8561, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5081, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4244, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4708, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6334, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8596, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5527, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3879, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5658, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3258, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2720, device='cuda:0', grad_fn=) +9.954748808839674e-05 +changing lr +epoch 3, time 371.62, cls_loss 1.9643 +306 +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4955, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7761, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9546, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6588, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4942, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3750, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9220, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6196, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6732, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6517, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3160, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5677, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7045, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5078, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6413, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6834, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6988, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3664, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3886, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4856, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4050, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1956, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3913, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8607, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9811, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2583, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6879, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5929, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3491, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5107, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3112, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5596, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2371, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9195, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6973, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4575, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5773, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5998, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7244, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4897, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3882, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6589, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5179, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6559, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6249, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2371, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6852, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9499, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3549, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3321, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7363, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8671, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9885, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3058, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9751, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3395, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8771, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4997, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6061, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6904, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3154, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9772, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5676, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8757, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8931, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8451, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2832, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3280, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7871, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2494, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5124, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3916, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4736, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4603, device='cuda:0', grad_fn=) +9.919647942993148e-05 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 371.66, cls_loss 1.5969 +306 +cls_loss: tensor(0.7379, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3251, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9422, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9325, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6675, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3453, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7650, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6141, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1873, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9162, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9642, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9523, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8049, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6289, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3124, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4782, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8302, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6458, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9867, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7135, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3476, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4755, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4809, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3950, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6681, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4561, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8820, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8464, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4619, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4665, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3327, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5289, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4425, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4610, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6112, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2481, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6885, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4362, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2980, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2845, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5422, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6723, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8704, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8646, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8230, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6576, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8988, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2333, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2168, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3879, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4055, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4936, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8269, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4071, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5268, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7255, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5999, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3796, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3907, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6432, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3529, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5380, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8378, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9968, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6455, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7119, device='cuda:0', grad_fn=) +9.874639560909117e-05 +changing lr +epoch 5, time 370.53, cls_loss 1.0565 +306 +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4704, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6284, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8815, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3532, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4989, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5817, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8132, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7227, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7969, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7520, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5446, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4604, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3601, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5645, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9999, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7855, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5363, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4222, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5836, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7148, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5293, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4432, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3297, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5791, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5149, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3587, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8307, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5866, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6430, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5203, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2793, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3015, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5827, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5170, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8464, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5616, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2279, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8568, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1975, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7350, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6148, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3233, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +9.819814303479267e-05 +changing lr +epoch 6, time 372.51, cls_loss 0.7236 +306 +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4891, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2893, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4693, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7143, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5989, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1607, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1633, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9350, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4566, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5608, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3034, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7371, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2535, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7201, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3713, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6816, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6395, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6485, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9551, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7676, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3259, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6608, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4606, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5923, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6921, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6828, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5986, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7494, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7882, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7876, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3852, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8505, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3425, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +9.755282581475769e-05 +changing lr +epoch 7, time 371.05, cls_loss 0.6909 +306 +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5534, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2462, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7178, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5870, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3164, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6846, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7538, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7284, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8653, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4545, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3892, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4657, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8629, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2942, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9810, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2931, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1760, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6531, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9118, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8818, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4858, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7869, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1768, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4207, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7897, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3157, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5588, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4199, device='cuda:0', grad_fn=) +9.681174353198687e-05 +changing lr +epoch 8, time 374.19, cls_loss 0.5058 +306 +cls_loss: tensor(2.2836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9899, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9248, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7754, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3515, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6783, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8734, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5872, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7546, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6351, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7067, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7682, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7997, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4225, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6838e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4952, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5833, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4935, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3287, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5850, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4829, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5910, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3799, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4259, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4801, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2641, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3600, device='cuda:0', grad_fn=) +9.597638862757255e-05 +changing lr +epoch 9, time 370.81, cls_loss 0.3589 +306 +cls_loss: tensor(0.4434, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4238, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4958, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3473, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2897, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6230, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1492, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9844, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4649, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5605, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7751, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6512, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7982, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8751, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6914, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6784, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7505, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9340, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2402, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8626, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3496, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8221, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7572, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4089, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5976, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +9.504844339512095e-05 +changing lr +epoch 10, time 374.15, cls_loss 0.3164 +306 +cls_loss: tensor(0.2835, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8126, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6530, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3513, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5823, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4249, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8998, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6833, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7943, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8119, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4553, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5668, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4250, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4761, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4552, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5602, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9281, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4430, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2930, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3140, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6946, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9564, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8048, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7262, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1750, device='cuda:0', grad_fn=) +9.40297765928369e-05 +changing lr +epoch 11, time 368.41, cls_loss 0.4055 +306 +cls_loss: tensor(0.8117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9958, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5628, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6383, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2831, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4782, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5400, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6562, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9311, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6387, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2378, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8972, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1712e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2562, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9134, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8900, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6287, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5524, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0599, device='cuda:0', grad_fn=) +9.292243968009331e-05 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 370.29, cls_loss 0.2203 +306 +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4489, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5058, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8125, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5920, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9486, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5573, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9605, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1853, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2647, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2920, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4296, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9870, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2415, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7663, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9875, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5477, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3153, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6491, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5772, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9951, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3223, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3732, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629e-05, device='cuda:0', grad_fn=) +9.172866268606513e-05 +changing lr +epoch 13, time 370.78, cls_loss 0.2437 +306 +cls_loss: tensor(1.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7253, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6120, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6155, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3678, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3443, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8774, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2722, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7217, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5193, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2343, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3977, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5798, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4404, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7943, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4199, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7078, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +9.045084971874738e-05 +changing lr +epoch 14, time 369.04, cls_loss 0.1421 +306 +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4079, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7526, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6367, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3926, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2082, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6823e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5870, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5907, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4402, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +8.90915741234015e-05 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 371.33, cls_loss 0.0941 +306 +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7702, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6075, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3677, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2213, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5700, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5820, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3154, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +8.765357330018056e-05 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 370.20, cls_loss 0.0437 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4757, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6772, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4542, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2092, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1773, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3734, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4949, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3868, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4869, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +8.613974319136958e-05 +changing lr +epoch 17, time 367.09, cls_loss 0.0791 +306 +cls_loss: tensor(0.2980, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3395, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1988, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7249, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7079, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4903e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7257, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7641, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5182, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6352, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6771e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3877, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +8.455313244934324e-05 +changing lr +epoch 18, time 371.84, cls_loss 0.0770 +306 +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3975, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7579, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5583, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6672, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4922, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6941, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2472, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8553, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8399, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3524, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1536, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6980, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6419, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1374e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7622, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5489, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7207, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2796, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5938, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +8.289693629698562e-05 +changing lr +epoch 19, time 371.53, cls_loss 0.1129 +306 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1695, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5982, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9821, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6160e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3169, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4352, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2523, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +8.117449009293667e-05 +changing lr +epoch 20, time 370.30, cls_loss 0.0432 +306 +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1811e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2724, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4196, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8754, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3079, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6568, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1503, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7773, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4344, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6483, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8247e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2303, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9486, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3675, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2851, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7481, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2443, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8141, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7109, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0220e-05, device='cuda:0', grad_fn=) +7.938926261462365e-05 +changing lr +epoch 21, time 371.94, cls_loss 0.0887 +306 +cls_loss: tensor(2.9723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4940, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7911e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2739, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7051, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2374e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3633, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2218, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1444, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7925, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +7.754484907260511e-05 +changing lr +epoch 22, time 371.89, cls_loss 0.0312 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9994, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4274, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7096e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3482, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4656, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6858e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4059, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +7.56449638702953e-05 +changing lr +epoch 23, time 370.29, cls_loss 0.0272 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2596e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5296, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8763, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5406e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9349, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6586e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6494, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3183, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8217, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5537, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8139, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2535, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6977, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +7.369343312364992e-05 +changing lr +epoch 24, time 381.37, cls_loss 0.0446 +306 +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4184, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6140, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7317, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4704, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9345, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6158, device='cuda:0', grad_fn=) +7.169418695587788e-05 +changing lr +epoch 25, time 371.11, cls_loss 0.0312 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7767, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2774, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5665, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2922, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6369, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9750, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6095, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5831, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4287, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4772, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6240, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3233, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +6.965125158269616e-05 +changing lr +epoch 26, time 369.76, cls_loss 0.0675 +306 +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4228, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7918, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3480, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1327e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2777, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6222e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5632e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6642, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.756874120406712e-05 +changing lr +epoch 27, time 378.67, cls_loss 0.0209 +306 +cls_loss: tensor(4.3790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5924, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +6.545084971874736e-05 +changing lr +epoch 28, time 371.40, cls_loss 0.0175 +306 +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5311, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.330184227833373e-05 +changing lr +epoch 29, time 370.74, cls_loss 0.0151 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6077, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6248, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4485, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3291, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4408, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2614, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +6.11260466978157e-05 +changing lr +epoch 30, time 370.81, cls_loss 0.0388 +306 +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6544, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4404, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4919, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2218, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7363, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4791, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1948, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1076e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8379, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2513, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +5.892784473993181e-05 +changing lr +epoch 31, time 371.78, cls_loss 0.0343 +306 +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3447, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2402, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2857, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9625, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4066, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2493e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9825, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2043, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +5.6711663290882756e-05 +changing lr +epoch 32, time 369.09, cls_loss 0.0272 +306 +cls_loss: tensor(2.1696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5605, device='cuda:0', grad_fn=) +5.448196544517166e-05 +changing lr +epoch 33, time 367.91, cls_loss 0.0076 +306 +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5487e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8063, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4757, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1928, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6980, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +5.2243241517525733e-05 +changing lr +epoch 34, time 369.64, cls_loss 0.0462 +306 +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6956, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3036e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2720, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0996e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2713, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2323, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +4.999999999999998e-05 +changing lr +epoch 35, time 368.34, cls_loss 0.0160 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6308e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3224, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3405e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2944e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3636, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1473e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8020, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +4.7756758482474244e-05 +changing lr +epoch 36, time 368.79, cls_loss 0.0156 +306 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3406e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3777e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7575, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4851, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4883e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +4.551803455482832e-05 +changing lr +epoch 37, time 369.07, cls_loss 0.0176 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5239, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5507e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1723, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +4.328833670911722e-05 +changing lr +epoch 38, time 370.45, cls_loss 0.0107 +306 +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +4.1072155260068164e-05 +changing lr +epoch 39, time 368.45, cls_loss 0.0066 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9142e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7585, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9954, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6798e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6509, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +3.8873953302184275e-05 +changing lr +epoch 40, time 371.13, cls_loss 0.0125 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5617, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1950e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5199, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4903e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1289, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +3.669815772166625e-05 +changing lr +epoch 41, time 369.79, cls_loss 0.0184 +306 +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9803e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7220, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6823e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4374e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +3.454915028125263e-05 +changing lr +epoch 42, time 370.74, cls_loss 0.0064 +306 +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2817, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4969e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3677, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4274, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +3.2431258795932867e-05 +changing lr +epoch 43, time 371.94, cls_loss 0.0071 +306 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1658e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2720, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4950e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6957e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +3.0348748417303827e-05 +changing lr +epoch 44, time 373.41, cls_loss 0.0029 +306 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2986, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5698e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9606e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4619, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4704e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-06, device='cuda:0', grad_fn=) +2.830581304412209e-05 +changing lr +epoch 45, time 369.63, cls_loss 0.0047 +306 +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8877e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7089e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3313e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2253e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3699e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4945, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941e-05, device='cuda:0', grad_fn=) +2.6306566876350062e-05 +changing lr +epoch 46, time 373.36, cls_loss 0.0037 +306 +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9737e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +2.4355036129704693e-05 +changing lr +epoch 47, time 369.08, cls_loss 0.0015 +306 +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0808e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8810e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4241e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0202e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4770e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1790e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1662, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6425e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2766, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +2.2455150927394874e-05 +changing lr +epoch 48, time 368.30, cls_loss 0.0068 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1247e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8944e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4222, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6791e-05, device='cuda:0', grad_fn=) +2.0610737385376345e-05 +changing lr +epoch 49, time 368.14, cls_loss 0.0027 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8492, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8513e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9406e-06, device='cuda:0', grad_fn=) +1.8825509907063323e-05 +changing lr +epoch 50, time 372.71, cls_loss 0.0070 +306 +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2603, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4261, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5249, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.710306370301437e-05 +changing lr +epoch 51, time 372.36, cls_loss 0.0120 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4903e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3855, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7178, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0844e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +1.5446867550656765e-05 +changing lr +epoch 52, time 374.38, cls_loss 0.0065 +306 +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0731e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0400e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5301e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7096e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8023, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +1.3860256808630425e-05 +changing lr +epoch 53, time 372.56, cls_loss 0.0064 +306 +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2189e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2838, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2903e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9268e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9738e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6054e-05, device='cuda:0', grad_fn=) +1.2346426699819455e-05 +changing lr +epoch 54, time 366.99, cls_loss 0.0043 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3805, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4771e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1404e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1856e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0711e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3834, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1923e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5036e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +1.0908425876598507e-05 +changing lr +epoch 55, time 369.57, cls_loss 0.0075 +306 +cls_loss: tensor(1.3928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5102e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6028e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5738e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8268e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7722e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0895e-05, device='cuda:0', grad_fn=) +9.549150281252631e-06 +changing lr +epoch 56, time 369.69, cls_loss 0.0030 +306 +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4903e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3711e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2095, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2159e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4798e-05, device='cuda:0', grad_fn=) +8.271337313934865e-06 +changing lr +epoch 57, time 371.73, cls_loss 0.0023 +306 +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2030e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1675, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8347e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5379e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3470, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9208e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3180e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9274e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9672e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8412e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0268e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7882e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2268e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5963e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0155e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +7.077560319906693e-06 +changing lr +epoch 58, time 372.81, cls_loss 0.0023 +306 +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9598e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9937e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4306e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5235e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6493e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1872, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2312e-05, device='cuda:0', grad_fn=) +5.970223407163098e-06 +changing lr +epoch 59, time 369.63, cls_loss 0.0032 +306 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4851e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2754, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1261e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6625e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4197, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6512e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +4.9515566048790464e-06 +changing lr +epoch 60, time 370.26, cls_loss 0.0047 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3963e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.0222e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6758e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4742e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8366e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8896e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3619e-05, device='cuda:0', grad_fn=) +4.0236113724274705e-06 +changing lr +epoch 61, time 368.57, cls_loss 0.0017 +306 +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7221e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9883e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6029e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8744e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6824e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2833, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-06, device='cuda:0', grad_fn=) +3.188256468013139e-06 +changing lr +epoch 62, time 373.65, cls_loss 0.0035 +306 +cls_loss: tensor(0.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3301e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1702e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6540e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3947e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7818e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3762, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7752e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6559e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +2.4471741852423225e-06 +changing lr +epoch 63, time 373.02, cls_loss 0.0045 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4016e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5141e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1731e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1618e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9141e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9530e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3182e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8956e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8811e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7950e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0798e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3181e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5169e-05, device='cuda:0', grad_fn=) +1.8018569652073373e-06 +changing lr +epoch 64, time 369.82, cls_loss 0.0030 +306 +cls_loss: tensor(5.8492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3115e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8359e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5564e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0053e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6558e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1473e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2425, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6001e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6491e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2559, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +1.2536043909088185e-06 +changing lr +epoch 65, time 368.64, cls_loss 0.0050 +306 +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1255e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1981e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0333e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5644e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3818e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9339e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7948e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8943e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5233e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0862e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6848e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6427e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0604e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2651e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8414e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3696, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7553e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4837e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +8.035205700685162e-07 +changing lr +epoch 66, time 370.25, cls_loss 0.0089 +306 +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3697e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6093e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1525e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6207e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3809e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7120e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9606e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6227e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4374, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +4.525119116032647e-07 +changing lr +epoch 67, time 328.19, cls_loss 0.0088 +306 +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7287e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8544e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8803e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4663e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2866, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8213e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4440e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6756e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5830e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8452e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5636e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6228e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5897e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +2.0128530023804648e-07 +changing lr +epoch 68, time 301.22, cls_loss 0.0030 +306 +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5255e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3312e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1060e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1696e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6948e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2492e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4108e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4769e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2134e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4054e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2453e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3379e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4638e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0465e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +5.03466729342705e-08 +changing lr +epoch 69, time 302.19, cls_loss 0.0016 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +x.shape: (2048, 227, 227, 3) +x_aug test here torch.Size([2048, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +x.shape: (2344, 227, 227, 3) +x_aug test here torch.Size([2344, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +x.shape: (1670, 227, 227, 3) +x_aug test here torch.Size([1670, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +x.shape: (3929, 227, 227, 3) +x_aug test here torch.Size([3929, 3, 227, 227]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 94.873047 55.332765 92.035928 45.405956 64.258216 diff --git a/Meta-causal/code-stage1-pipeline/56458.error b/Meta-causal/code-stage1-pipeline/56458.error new file mode 100644 index 0000000000000000000000000000000000000000..551e8156b629e823284bff1b15239ab90e9d86f5 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56458.error @@ -0,0 +1,3 @@ +slurmstepd: error: *** JOB 56458 ON gcpl4-eu-1 CANCELLED AT 2024-07-04T07:07:32 DUE TO TIME LIMIT *** +slurmstepd: error: *** STEP 56458.0 ON gcpl4-eu-1 CANCELLED AT 2024-07-04T07:07:32 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-stage1-pipeline/56458.log b/Meta-causal/code-stage1-pipeline/56458.log new file mode 100644 index 0000000000000000000000000000000000000000..37d3fcaee1f3c1e4f1ecc7be70d61a7fb1d798d4 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56458.log @@ -0,0 +1,24674 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +x.shape: torch.Size([10000, 3, 32, 32]) +x_aug train here torch.Size([10000, 3, 32, 32]) +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +cls_loss: tensor(2.3079, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2931, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3058, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2999, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2840, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2934, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2733, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3294, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2662, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2903, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2722, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2537, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3054, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2793, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3136, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2623, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3207, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2251, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2884, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2494, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2540, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2831, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2834, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2752, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2482, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3054, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2457, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2605, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2169, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2065, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3337, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2359, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9895, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1839, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2786, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9758, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9965, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8568, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9605, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9957, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9389, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8199, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9608, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1692, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8549, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8658, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 182.97, cls_loss 2.1788 +100 +cls_loss: tensor(1.9016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7707, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7044, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9385, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9929, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8472, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8293, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8809, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9756, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8752, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9647, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9836, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7100, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8504, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8809, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8848, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8328, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9843, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9197, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5955, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6149, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9997, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8156, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7500, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6487, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9658, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8108, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8701, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8169, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7855, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8321, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7273, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6239, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6693, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6372, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6851, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8000, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6925, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8394, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7388, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8530, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7313, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7628, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8970, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7781, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7949, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5280, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5867, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9102, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6814, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8362, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6488, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8687, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5871, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7681, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4927, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7462, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7127, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6863, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2979, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6581, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6178, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6773, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6223, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7810, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 183.12, cls_loss 1.7921 +100 +cls_loss: tensor(1.9749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8196, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7885, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7805, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5744, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5467, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2567, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4973, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6400, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9441, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8220, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8152, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7497, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8375, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4668, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9177, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4990, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5710, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5976, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5976, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4461, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4521, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3662, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4533, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3670, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8689, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4935, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3970, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7139, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4610, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3537, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3739, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7381, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5404, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5081, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5063, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5032, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5949, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4891, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7150, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7156, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7877, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8415, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3960, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3092, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3552, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5214, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6576, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6414, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4892, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2881, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5793, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5382, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6121, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4836, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4468, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3301, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3933, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4268, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5500, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4401, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4446, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2978, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4087, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4928, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1187, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 183.70, cls_loss 1.5671 +100 +cls_loss: tensor(1.8093, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2399, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3824, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3487, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3840, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8727, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6536, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5675, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4422, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6113, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3095, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4354, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4324, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3647, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9147, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4698, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6610, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4665, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6213, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4228, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5127, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5536, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4303, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6869, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6858, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4859, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3819, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8291, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5160, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5682, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7920, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6071, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8427, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1980, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6937, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2955, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5322, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7210, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7717, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3903, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2257, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2940, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3082, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3144, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7590, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2211, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2758, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3074, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3348, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2605, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5915, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7922, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4113, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2883, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5853, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9655, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4354, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7935, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2084, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3416, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3374, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3902, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6203, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7762, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7499, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4446, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 183.81, cls_loss 1.4443 +100 +cls_loss: tensor(1.2070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7850, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6301, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2639, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3629, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3554, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4849, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6227, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4975, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5691, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2949, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2990, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5546, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2424, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4482, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4771, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3120, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2696, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4293, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4597, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9670, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3432, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4851, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3994, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6883, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6221, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9418, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4539, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6384, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2820, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7306, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4735, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9438, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6882, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6977, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4287, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3915, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3186, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9954, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2802, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3168, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5962, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4917, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4790, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8765, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3072, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2825, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3297, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2095, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3620, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5508, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 183.27, cls_loss 1.3402 +100 +cls_loss: tensor(1.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5493, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6666, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5408, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1483, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6016, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3700, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2906, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5546, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6420, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6809, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4977, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9271, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3535, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2672, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8806, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2284, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4810, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4888, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2757, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2567, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2602, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3557, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2099, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5560, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3029, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4196, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4460, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6541, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2777, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3059, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3976, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3478, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4436, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4072, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5270, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3488, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2325, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2727, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2300, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5975, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8613, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4934, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3161, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5520, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3613, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3774, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0341, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 182.60, cls_loss 1.2952 +100 +cls_loss: tensor(0.9599, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4692, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2765, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2686, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5453, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1995, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2652, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4814, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3812, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8304, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1583, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9853, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4382, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9827, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2304, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6106, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8292, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2522, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2215, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4200, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3431, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5388, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6887, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2457, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9575, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3216, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4130, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3331, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3286, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8362, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9577, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2899, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4097, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8443, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4308, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2717, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9875, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9412, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2622, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 182.83, cls_loss 1.2123 +100 +cls_loss: tensor(1.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6974, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7744, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2374, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3979, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1955, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3936, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8393, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2298, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2273, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9755, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2784, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2472, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2967, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8918, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4696, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9190, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5090, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9568, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2596, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5278, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9468, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3838, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6612, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9149, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9196, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9907, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9868, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2757, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9928, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0050, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 183.06, cls_loss 1.1647 +100 +cls_loss: tensor(1.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4165, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5983, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2850, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3446, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3684, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3338, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4648, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7872, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7503, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8053, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8487, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4807, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5603, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2538, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3990, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8742, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8661, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2643, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9608, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9643, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8102, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9162, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2506, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9464, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9465, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9412, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8585, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2855, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1592, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9848, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9387, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8638, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2839, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 8, time 183.03, cls_loss 1.1313 +100 +cls_loss: tensor(0.9350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8663, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8170, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1880, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3222, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8763, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2723, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9345, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2825, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8702, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7274, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8283, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1618, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7304, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6967, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3867, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8985, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9250, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9566, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8790, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7383, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9414, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8446, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8352, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8752, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7966, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8707, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 9, time 182.42, cls_loss 1.0235 +100 +cls_loss: tensor(0.8801, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3744, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9355, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2733, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2722, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9178, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8784, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6921, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8221, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2274, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8483, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9948, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9493, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2198, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9325, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2374, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3684, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4619, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8252, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9712, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9793, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3401, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2284, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2982, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2649, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3136, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3619, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4557, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7597, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8343, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4434, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9204, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3511, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8155, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7914, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9857, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 182.93, cls_loss 1.0756 +100 +cls_loss: tensor(0.6990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9202, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8905, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8759, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2520, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7898, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7986, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8232, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8767, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9959, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9895, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8294, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9041, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6558, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7759, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9440, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9730, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8590, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7835, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9390, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7484, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2111, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 11, time 183.17, cls_loss 0.9875 +100 +cls_loss: tensor(1.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7440, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8436, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8556, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5354, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8797, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9235, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3526, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7522, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9987, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7389, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7792, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7150, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4996, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6125, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6407, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8519, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9472, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8363, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8518, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9104, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9767, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 12, time 183.37, cls_loss 0.9463 +100 +cls_loss: tensor(1.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9954, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8617, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9452, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8538, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8935, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9574, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7731, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6753, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2888, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6991, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2046, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3055, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8857, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8907, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2172, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7959, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6749, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9816, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 13, time 182.55, cls_loss 0.9335 +100 +cls_loss: tensor(0.9881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9269, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7764, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8132, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6135, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9796, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6860, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9065, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7840, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8930, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7533, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9999, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8816, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6737, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6539, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8128, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9373, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7902, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8832, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6736, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8305, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6765, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9685, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 14, time 182.43, cls_loss 0.9064 +100 +cls_loss: tensor(1.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9621, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9672, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9643, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8298, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4563, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6721, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8298, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9066, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8642, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 15, time 182.11, cls_loss 0.8350 +100 +cls_loss: tensor(0.9005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8191, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8449, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9588, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9145, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7948, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9839, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9180, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9604, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7581, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7094, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6881, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9420, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5464, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9606, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8559, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5906, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5919, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2059, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 16, time 182.31, cls_loss 0.8641 +100 +cls_loss: tensor(0.7301, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2775, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6499, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6941, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8028, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9624, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5526, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5264, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5295, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8794, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5572, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6776, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8045, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6533, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5253, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6341, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9244, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8239, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9864, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 182.36, cls_loss 0.8358 +100 +cls_loss: tensor(0.5601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8241, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7924, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7265, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7530, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7879, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7383, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5784, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4384, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5840, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 18, time 182.33, cls_loss 0.7596 +100 +cls_loss: tensor(0.8550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7577, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8534, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9753, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8367, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7567, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6243, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7403, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6774, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 19, time 182.50, cls_loss 0.7861 +100 +cls_loss: tensor(0.6278, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7122, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5953, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5189, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8537, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6156, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8144, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7625, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7470, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 20, time 182.37, cls_loss 0.7305 +100 +cls_loss: tensor(0.7701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8080, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7883, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6714, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6688, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9482, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7300, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7246, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 21, time 182.61, cls_loss 0.7229 +100 +cls_loss: tensor(0.5952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5847, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7952, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 22, time 182.62, cls_loss 0.6799 +100 +cls_loss: tensor(0.5969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4367, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7232, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 23, time 182.25, cls_loss 0.6738 +100 +cls_loss: tensor(0.9265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6981, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8159, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6246, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6390, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8722, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8792, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 182.46, cls_loss 0.6676 +100 +cls_loss: tensor(0.6122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6093, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 182.63, cls_loss 0.6359 +100 +cls_loss: tensor(0.6362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4981, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 26, time 182.69, cls_loss 0.6211 +100 +cls_loss: tensor(0.7790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4546, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6933, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5803, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 27, time 182.69, cls_loss 0.6134 +100 +cls_loss: tensor(0.7295, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7240, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7901, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3067, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4491, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 182.95, cls_loss 0.5875 +100 +cls_loss: tensor(0.6105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3424, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3070, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5978, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 182.97, cls_loss 0.5629 +100 +cls_loss: tensor(0.5179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4934, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 30, time 182.71, cls_loss 0.5500 +100 +cls_loss: tensor(0.5226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5280, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 31, time 182.49, cls_loss 0.5263 +100 +cls_loss: tensor(0.3079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3217, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 32, time 182.04, cls_loss 0.5264 +100 +cls_loss: tensor(0.2748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7491, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 33, time 181.62, cls_loss 0.4934 +100 +cls_loss: tensor(0.4167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2698, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3229, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 34, time 182.03, cls_loss 0.4776 +100 +cls_loss: tensor(0.8007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3388, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 35, time 182.11, cls_loss 0.4587 +100 +cls_loss: tensor(0.3386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5650, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 36, time 182.03, cls_loss 0.4303 +100 +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4439, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 37, time 182.14, cls_loss 0.4102 +100 +cls_loss: tensor(0.6017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3504, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 38, time 181.69, cls_loss 0.4135 +100 +cls_loss: tensor(0.3605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3669, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 39, time 181.99, cls_loss 0.3995 +100 +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 40, time 181.76, cls_loss 0.3803 +100 +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7923, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 41, time 181.85, cls_loss 0.3892 +100 +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3293, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 181.68, cls_loss 0.3526 +100 +cls_loss: tensor(0.3209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3763, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 43, time 181.70, cls_loss 0.3728 +100 +cls_loss: tensor(0.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3491, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 44, time 181.75, cls_loss 0.3538 +100 +cls_loss: tensor(0.4396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 45, time 182.09, cls_loss 0.3350 +100 +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5294, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 46, time 181.65, cls_loss 0.3007 +100 +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3026, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 47---------------------------------------------------- +epoch 47, time 182.27, cls_loss 0.2936 +100 +cls_loss: tensor(0.1747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6032, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 48, time 181.69, cls_loss 0.3251 +100 +cls_loss: tensor(0.3330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4153, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 49, time 181.59, cls_loss 0.2937 +100 +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2568, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 50---------------------------------------------------- +epoch 50, time 182.08, cls_loss 0.2901 +100 +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2459, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 51, time 181.65, cls_loss 0.2588 +100 +cls_loss: tensor(0.2445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3418, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 52, time 181.75, cls_loss 0.2353 +100 +cls_loss: tensor(0.4048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4930, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 53, time 182.10, cls_loss 0.2449 +100 +cls_loss: tensor(0.5333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 54, time 181.99, cls_loss 0.2617 +100 +cls_loss: tensor(0.3551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4579, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 55, time 182.03, cls_loss 0.2681 +100 +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3024, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 56, time 181.95, cls_loss 0.2233 +100 +cls_loss: tensor(0.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 57, time 181.83, cls_loss 0.2456 +100 +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 58, time 181.97, cls_loss 0.2123 +100 +cls_loss: tensor(0.1692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4286, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 59, time 181.86, cls_loss 0.2182 +100 +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2365, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 60, time 182.00, cls_loss 0.2060 +100 +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2435, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 61, time 182.01, cls_loss 0.2162 +100 +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 62, time 181.87, cls_loss 0.1768 +100 +cls_loss: tensor(0.3925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 63, time 181.99, cls_loss 0.1934 +100 +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 64, time 182.12, cls_loss 0.1679 +100 +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 65, time 181.79, cls_loss 0.1513 +100 +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2375, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 66, time 181.81, cls_loss 0.1830 +100 +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 67, time 181.63, cls_loss 0.1695 +100 +cls_loss: tensor(0.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 68, time 181.65, cls_loss 0.2151 +100 +cls_loss: tensor(0.2077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 69, time 181.73, cls_loss 0.1747 +100 +cls_loss: tensor(0.1804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 70, time 181.89, cls_loss 0.1611 +100 +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 71, time 181.78, cls_loss 0.1409 +100 +cls_loss: tensor(0.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3360, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 72, time 182.02, cls_loss 0.1552 +100 +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 73, time 181.96, cls_loss 0.1751 +100 +cls_loss: tensor(0.1695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 74, time 181.98, cls_loss 0.1424 +100 +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 75, time 181.81, cls_loss 0.1479 +100 +cls_loss: tensor(0.2626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 76, time 181.52, cls_loss 0.1632 +100 +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 77---------------------------------------------------- +epoch 77, time 182.15, cls_loss 0.1436 +100 +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 78, time 181.86, cls_loss 0.1121 +100 +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 79, time 181.90, cls_loss 0.1170 +100 +cls_loss: tensor(0.2415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 80, time 181.83, cls_loss 0.1138 +100 +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 81, time 181.98, cls_loss 0.1567 +100 +cls_loss: tensor(0.2511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 82, time 181.87, cls_loss 0.1422 +100 +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 83, time 182.15, cls_loss 0.1114 +100 +cls_loss: tensor(0.4540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 84, time 181.75, cls_loss 0.1159 +100 +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 85, time 181.65, cls_loss 0.1011 +100 +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 86, time 182.02, cls_loss 0.1193 +100 +cls_loss: tensor(0.2521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2108, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 87, time 181.87, cls_loss 0.1497 +100 +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 88, time 181.76, cls_loss 0.1331 +100 +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 89, time 182.19, cls_loss 0.1006 +100 +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 90, time 182.36, cls_loss 0.0891 +100 +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1663, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 91, time 182.32, cls_loss 0.1021 +100 +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2214, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 92, time 182.17, cls_loss 0.1216 +100 +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 93, time 181.97, cls_loss 0.1505 +100 +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1692, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 94, time 181.86, cls_loss 0.1161 +100 +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 95, time 181.74, cls_loss 0.0986 +100 +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2319, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 96, time 181.78, cls_loss 0.0813 +100 +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 97, time 181.82, cls_loss 0.0745 +100 +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 98, time 182.03, cls_loss 0.0993 +100 +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 99, time 181.99, cls_loss 0.0871 +100 +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 100, time 182.06, cls_loss 0.0818 +100 +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 101, time 182.08, cls_loss 0.0989 +100 +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 102, time 181.69, cls_loss 0.0943 +100 +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 103, time 181.95, cls_loss 0.1023 +100 +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 104, time 182.08, cls_loss 0.1027 +100 +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 105, time 182.06, cls_loss 0.1148 +100 +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 106, time 182.04, cls_loss 0.1094 +100 +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 107, time 182.13, cls_loss 0.0729 +100 +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 108, time 182.04, cls_loss 0.0695 +100 +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 109, time 182.10, cls_loss 0.0663 +100 +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 110, time 182.12, cls_loss 0.0816 +100 +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1690, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 111, time 182.20, cls_loss 0.0832 +100 +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 112, time 181.79, cls_loss 0.1207 +100 +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 113, time 182.12, cls_loss 0.0922 +100 +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 114, time 181.64, cls_loss 0.0860 +100 +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 115, time 181.92, cls_loss 0.0823 +100 +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 116, time 182.03, cls_loss 0.0861 +100 +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 117, time 181.99, cls_loss 0.0853 +100 +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 118, time 181.73, cls_loss 0.0851 +100 +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 119, time 181.94, cls_loss 0.0954 +100 +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 120, time 181.68, cls_loss 0.0792 +100 +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 121, time 181.87, cls_loss 0.0544 +100 +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 122, time 181.93, cls_loss 0.0714 +100 +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0892, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 123, time 181.92, cls_loss 0.0593 +100 +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 124, time 181.94, cls_loss 0.0714 +100 +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 125, time 181.98, cls_loss 0.0573 +100 +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 126, time 181.92, cls_loss 0.0685 +100 +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 127, time 182.19, cls_loss 0.0751 +100 +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 128, time 181.68, cls_loss 0.0752 +100 +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 129, time 181.97, cls_loss 0.1064 +100 +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 130, time 182.13, cls_loss 0.0621 +100 +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0961, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 131, time 182.13, cls_loss 0.0419 +100 +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2931, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 132, time 181.99, cls_loss 0.0582 +100 +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 133, time 182.23, cls_loss 0.0569 +100 +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 134, time 181.73, cls_loss 0.0607 +100 +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6910, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 135, time 182.09, cls_loss 0.0796 +100 +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 136, time 182.58, cls_loss 0.0635 +100 +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 137, time 182.15, cls_loss 0.0795 +100 +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 138, time 181.64, cls_loss 0.0925 +100 +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2037, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 139, time 181.95, cls_loss 0.1050 +100 +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 140, time 182.05, cls_loss 0.0962 +100 +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 141, time 182.89, cls_loss 0.0766 +100 +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 142, time 181.95, cls_loss 0.0578 +100 +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 143, time 181.86, cls_loss 0.0473 +100 +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 144, time 182.05, cls_loss 0.0472 +100 +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 145, time 181.68, cls_loss 0.0378 +100 +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 146, time 181.78, cls_loss 0.0419 +100 +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 147, time 182.07, cls_loss 0.0474 +100 +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 148, time 182.19, cls_loss 0.0349 +100 +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0747, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6098, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 149, time 181.86, cls_loss 0.0639 +100 +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 150, time 182.00, cls_loss 0.0508 +100 +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 151, time 182.04, cls_loss 0.0721 +100 +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2490, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1906, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 152, time 181.96, cls_loss 0.1119 +100 +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 153, time 181.61, cls_loss 0.0857 +100 +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 154, time 181.55, cls_loss 0.0740 +100 +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 155, time 181.59, cls_loss 0.0413 +100 +cls_loss: tensor(0.1669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2116, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 156, time 181.81, cls_loss 0.0624 +100 +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0895, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 157, time 181.92, cls_loss 0.0348 +100 +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 158, time 182.11, cls_loss 0.0519 +100 +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 159, time 182.25, cls_loss 0.0464 +100 +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 160, time 182.23, cls_loss 0.0377 +100 +cls_loss: tensor(0.2825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 161, time 181.71, cls_loss 0.0451 +100 +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1480, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 162, time 181.75, cls_loss 0.0378 +100 +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 163, time 182.01, cls_loss 0.0551 +100 +cls_loss: tensor(0.2520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 164, time 181.71, cls_loss 0.0456 +100 +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 165, time 180.88, cls_loss 0.0434 +100 +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0757, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 166, time 180.51, cls_loss 0.0322 +100 +cls_loss: tensor(0.1234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 167, time 180.31, cls_loss 0.0448 +100 +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 168, time 180.23, cls_loss 0.0628 +100 +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 169, time 180.07, cls_loss 0.0702 +100 +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 170, time 180.03, cls_loss 0.0696 +100 +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 171, time 179.98, cls_loss 0.0670 +100 +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 172, time 180.04, cls_loss 0.0921 +100 +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 173, time 179.82, cls_loss 0.0772 +100 +cls_loss: tensor(0.1601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0914, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 174, time 180.21, cls_loss 0.0590 +100 +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 175---------------------------------------------------- +epoch 175, time 180.57, cls_loss 0.0403 +100 +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1919, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 176, time 179.85, cls_loss 0.0464 +100 +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 177, time 180.69, cls_loss 0.0419 +100 +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 178, time 180.24, cls_loss 0.0298 +100 +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 179, time 180.23, cls_loss 0.0243 +100 +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 180, time 180.04, cls_loss 0.0261 +100 +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0638, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 181, time 180.07, cls_loss 0.0405 +100 +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 182---------------------------------------------------- +epoch 182, time 180.41, cls_loss 0.0311 +100 +cls_loss: tensor(0.1585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1886, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1858e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 183, time 180.20, cls_loss 0.0313 +100 +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 184, time 180.37, cls_loss 0.0380 +100 +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0982, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 185, time 180.42, cls_loss 0.0429 +100 +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2700, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 186, time 180.40, cls_loss 0.0528 +100 +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0342, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 187, time 180.59, cls_loss 0.0500 +100 +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 188, time 180.53, cls_loss 0.0367 +100 +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 189, time 180.38, cls_loss 0.0366 +100 +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 190, time 180.34, cls_loss 0.0799 +100 +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1988, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2884, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 191, time 180.65, cls_loss 0.1287 +100 +cls_loss: tensor(0.1760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 192, time 180.41, cls_loss 0.0843 +100 +cls_loss: tensor(0.3001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 193, time 180.63, cls_loss 0.0771 +100 +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0958, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 194, time 180.78, cls_loss 0.0617 +100 +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 195, time 180.76, cls_loss 0.0339 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 196, time 180.70, cls_loss 0.0396 +100 +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 197, time 180.52, cls_loss 0.0287 +100 +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0890, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0922, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 198, time 180.62, cls_loss 0.0239 +100 +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 199, time 180.11, cls_loss 0.0215 +100 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 200, time 180.20, cls_loss 0.0221 +100 +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 201, time 180.58, cls_loss 0.0254 +100 +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 202, time 180.73, cls_loss 0.0227 +100 +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 203, time 180.40, cls_loss 0.0427 +100 +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0979, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 204, time 180.62, cls_loss 0.0425 +100 +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 205, time 180.38, cls_loss 0.0369 +100 +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6243e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 206, time 180.36, cls_loss 0.0462 +100 +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0541, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0980e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 207, time 180.11, cls_loss 0.0838 +100 +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0735, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1718, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0550, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 208, time 180.44, cls_loss 0.0864 +100 +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 209, time 180.48, cls_loss 0.0424 +100 +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0791, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1957, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0441, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2102, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 210, time 180.65, cls_loss 0.0462 +100 +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0880, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0512, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 211, time 180.57, cls_loss 0.0445 +100 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0765, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 212, time 180.82, cls_loss 0.0446 +100 +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0637, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 213, time 180.64, cls_loss 0.0399 +100 +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0915, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1704, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0867, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 214, time 180.52, cls_loss 0.0660 +100 +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0992, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0893, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0888, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2032, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 215, time 180.71, cls_loss 0.0597 +100 +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1475, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1439, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 216, time 180.60, cls_loss 0.0576 +100 +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 217, time 180.62, cls_loss 0.0425 +100 +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0537, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0871, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 218, time 180.49, cls_loss 0.0314 +100 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 219, time 180.31, cls_loss 0.0316 +100 +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 220, time 180.70, cls_loss 0.0204 +100 +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 221, time 180.66, cls_loss 0.0195 +100 +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1191, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 222, time 180.36, cls_loss 0.0201 +100 +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0768, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4615e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0618, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0858, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7018e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 223, time 180.81, cls_loss 0.0143 +100 +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1669, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0785, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 224, time 180.23, cls_loss 0.0279 +100 +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8787e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 225, time 180.58, cls_loss 0.0268 +100 +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1633e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0970, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0986, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3583e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 226, time 180.38, cls_loss 0.0274 +100 +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0916, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 227, time 180.45, cls_loss 0.0416 +100 +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0503, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 228, time 180.37, cls_loss 0.0355 +100 +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0968, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0649, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0996, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 229, time 180.61, cls_loss 0.0416 +100 +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1752, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0513, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1819, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 230, time 180.77, cls_loss 0.0597 +100 +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0837, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0918, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0831, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 231, time 180.49, cls_loss 0.0821 +100 +cls_loss: tensor(0.1716, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0748, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 232, time 180.36, cls_loss 0.0642 +100 +cls_loss: tensor(0.3302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 233, time 180.74, cls_loss 0.0522 +100 +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1949, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0778, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1897, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0578, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 234, time 180.17, cls_loss 0.0561 +100 +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 235, time 179.98, cls_loss 0.0445 +100 +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1981, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) diff --git a/Meta-causal/code-stage1-pipeline/56526.error b/Meta-causal/code-stage1-pipeline/56526.error new file mode 100644 index 0000000000000000000000000000000000000000..dba98d58d7937963999306a8f7218b42cac8d5ee --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56526.error @@ -0,0 +1,31 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/main_test_pacs_v13.py", line 86, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/main_test_pacs_v13.py", line 29, in main + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/main_test_pacs_v13.py", line 42, in evaluate_pacs + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/torch/serialization.py", line 997, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/torch/serialization.py", line 444, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/torch/serialization.py", line 425, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/best_cls_net.pkl' +srun: error: gcpl4-eu-1: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-stage1-pipeline/56526.log b/Meta-causal/code-stage1-pipeline/56526.log new file mode 100644 index 0000000000000000000000000000000000000000..1ece7503c1ad1338d6f4d957f617386d42cebd82 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56526.log @@ -0,0 +1,4 @@ +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best diff --git a/Meta-causal/code-stage1-pipeline/56527.error b/Meta-causal/code-stage1-pipeline/56527.error new file mode 100644 index 0000000000000000000000000000000000000000..2bd579eb38978e7fa1e67d92618d75bfaabd44b3 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56527.error @@ -0,0 +1,31 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/main_test_digit_v13.py", line 84, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/main_test_digit_v13.py", line 28, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/main_test_digit_v13.py", line 42, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/torch/serialization.py", line 997, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/torch/serialization.py", line 444, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/torch/serialization.py", line 425, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/best_cls_net.pkl' +srun: error: gcpl4-eu-1: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-stage1-pipeline/56527.log b/Meta-causal/code-stage1-pipeline/56527.log new file mode 100644 index 0000000000000000000000000000000000000000..3fb3c2fbae12943f0379a7bccbc53cc9cd261cb1 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56527.log @@ -0,0 +1,3 @@ +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best diff --git a/Meta-causal/code-stage1-pipeline/56528.error b/Meta-causal/code-stage1-pipeline/56528.error new file mode 100644 index 0000000000000000000000000000000000000000..89474217af8112f2e4e5d05a7103bba0ccb52ddd --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56528.error @@ -0,0 +1,3 @@ +run_my_joint_v13_test.sh: line 25: ndm: command not found +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) diff --git a/Meta-causal/code-stage1-pipeline/56528.log b/Meta-causal/code-stage1-pipeline/56528.log new file mode 100644 index 0000000000000000000000000000000000000000..ff3259a1c37b7fa77b234d82797257b99df418de --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56528.log @@ -0,0 +1,9332 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 30, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +cls_loss: tensor(2.3446, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6243, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5887, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5546, device='cuda:0', grad_fn=) +cls_loss: tensor(17.9310, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2499, device='cuda:0', grad_fn=) +cls_loss: tensor(36.5010, device='cuda:0', grad_fn=) +cls_loss: tensor(35.8023, device='cuda:0', grad_fn=) +cls_loss: tensor(19.3028, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(19.2803, device='cuda:0', grad_fn=) +cls_loss: tensor(17.4525, device='cuda:0', grad_fn=) +cls_loss: tensor(15.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3067, device='cuda:0', grad_fn=) +cls_loss: tensor(20.6821, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4511, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7550, device='cuda:0', grad_fn=) +cls_loss: tensor(10.6936, device='cuda:0', grad_fn=) +cls_loss: tensor(12.9821, device='cuda:0', grad_fn=) +cls_loss: tensor(19.3366, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8249, device='cuda:0', grad_fn=) +cls_loss: tensor(17.5163, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3895, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4897, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3366, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7183, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2559, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3325, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5716, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8314, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6634, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6389, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0640, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3700, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3503, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(14.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(11.7174, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6397, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2224, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(22.5807, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5660, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1732, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4415, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8072, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7344, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7357, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(23.5156, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9980, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5980, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2853, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2396, device='cuda:0', grad_fn=) +cls_loss: tensor(17.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7125, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8796, device='cuda:0', grad_fn=) +cls_loss: tensor(17.2038, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7163, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7873, device='cuda:0', grad_fn=) +cls_loss: tensor(11.6703, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1760, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6429, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3490, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9469, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3686, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4827, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(11.8472, device='cuda:0', grad_fn=) +cls_loss: tensor(11.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0734, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4192, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8991, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(16.5872, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7119, device='cuda:0', grad_fn=) +cls_loss: tensor(19.4049, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4810, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3760, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(11.4950, device='cuda:0', grad_fn=) +cls_loss: tensor(24.6309, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7244, device='cuda:0', grad_fn=) +cls_loss: tensor(15.6605, device='cuda:0', grad_fn=) +cls_loss: tensor(16.7064, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1648, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3370, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8855, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3114, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1925, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7053, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4371, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9393, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6745, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7712, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5479, device='cuda:0', grad_fn=) +cls_loss: tensor(23.6750, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9434, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9200, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(39.4583, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(31.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8249, device='cuda:0', grad_fn=) +cls_loss: tensor(16.3997, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8835, device='cuda:0', grad_fn=) +cls_loss: tensor(11.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5312, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4784, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1875, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4934, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6799, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5466, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8312, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9097, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(21.6758, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(18.8994, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6459, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7805, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(17.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5316, device='cuda:0', grad_fn=) +cls_loss: tensor(11.5190, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2225, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8672, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(13.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7787, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4677, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7448, device='cuda:0', grad_fn=) +cls_loss: tensor(17.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2741, device='cuda:0', grad_fn=) +cls_loss: tensor(12.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6225, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6533, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9635, device='cuda:0', grad_fn=) +cls_loss: tensor(22.5258, device='cuda:0', grad_fn=) +cls_loss: tensor(12.2950, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5653, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(13.3756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6576, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4893, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4193, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(14.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8191, device='cuda:0', grad_fn=) +cls_loss: tensor(14.9922, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1056, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7598, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3733, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9471, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3914, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2969, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8442, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5560, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(29.8678, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4195, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6443, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8333, device='cuda:0', grad_fn=) +cls_loss: tensor(20.5466, device='cuda:0', grad_fn=) +cls_loss: tensor(10.5044, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0892, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2054, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5381, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(15.9594, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(13.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6675, device='cuda:0', grad_fn=) +cls_loss: tensor(39.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6380, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(16.5840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(18.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1920, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4916, device='cuda:0', grad_fn=) +cls_loss: tensor(12.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8933, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5143, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4958, device='cuda:0', grad_fn=) +cls_loss: tensor(14.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(15.8158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0397, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7152, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6850, device='cuda:0', grad_fn=) +cls_loss: tensor(15.2419, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5430, device='cuda:0', grad_fn=) +cls_loss: tensor(16.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(13.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5299, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5645, device='cuda:0', grad_fn=) +cls_loss: tensor(22.6590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6448, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4255, device='cuda:0', grad_fn=) +cls_loss: tensor(13.2031, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4945, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(12.3591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5732, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4883, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(22.2168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4457, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6354, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6914, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(18.9355, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2566, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 301.10, cls_loss 7.5080 +306 +cls_loss: tensor(7.7917, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(14.9401, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8073, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3574, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(13.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9460, device='cuda:0', grad_fn=) +cls_loss: tensor(18.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6548, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4225, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3176, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3341, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6943, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5430, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1967, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2949, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2457, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3717, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7396, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4334, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9540, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3763, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9319, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7163, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1289, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9583, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(12.1877, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2576, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6297, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5313, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8864e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1559, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0486, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5632, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7656, device='cuda:0', grad_fn=) +cls_loss: tensor(14.1272, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1772, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7578, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9219, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(13.7507, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7761, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3906, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5392, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3158, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6563, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9075, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8809e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5326, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0495, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9219, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8492, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5484, device='cuda:0', grad_fn=) +cls_loss: tensor(13.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9141, device='cuda:0', grad_fn=) +cls_loss: tensor(13.5834, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6560, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8112, device='cuda:0', grad_fn=) +cls_loss: tensor(10.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9975, device='cuda:0', grad_fn=) +cls_loss: tensor(24.5313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8932, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7881, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9349, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2376, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1979, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9690, device='cuda:0', grad_fn=) +cls_loss: tensor(11.5286, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5070, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9595, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2480, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(23.9271, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2474, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4676, device='cuda:0', grad_fn=) +cls_loss: tensor(14.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0716, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3307, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7630, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2637, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(26.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(12.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5213, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6719, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2114, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7852, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4118, device='cuda:0', grad_fn=) +cls_loss: tensor(21.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7891, device='cuda:0', grad_fn=) +cls_loss: tensor(12.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8132, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6341, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3276, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(11.8021, device='cuda:0', grad_fn=) +cls_loss: tensor(18.4313, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2346, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5917, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8208, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1525, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3751, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5768, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5744, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1478, device='cuda:0', grad_fn=) +cls_loss: tensor(13.9935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4038, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9688, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8076, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5833, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4753, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9914, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5803, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8802, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4227, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3586, device='cuda:0', grad_fn=) +cls_loss: tensor(14.4421, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2292, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5339, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4094, device='cuda:0', grad_fn=) +cls_loss: tensor(12.4375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8285, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(14.8646, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8740, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6159, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6510, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6966, device='cuda:0', grad_fn=) +cls_loss: tensor(20.5742, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5234, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9888, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6074, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(26.7812, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2814, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2644, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7331, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3796, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4758, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8083, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8438, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9129, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6513, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1841, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5387, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4033, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5260, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9896, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1502, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7318, device='cuda:0', grad_fn=) +9.972609476841367e-05 +changing lr +epoch 1, time 297.68, cls_loss 4.1508 +306 +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8325, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2112, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9980, device='cuda:0', grad_fn=) +cls_loss: tensor(11.9948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1828, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8362, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4723, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5868, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(12.8879, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0903, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8294, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5808, device='cuda:0', grad_fn=) +cls_loss: tensor(34.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(10.9610, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7292, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3022, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3031, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(11.3542, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3346, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8574, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3857, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4072, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4336, device='cuda:0', grad_fn=) +cls_loss: tensor(16.3691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7267, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6085, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5280, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4736, device='cuda:0', grad_fn=) +cls_loss: tensor(20.6589, device='cuda:0', grad_fn=) +cls_loss: tensor(12.6606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(26.3333, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4416, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5863, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4082, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3400, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7673, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9740, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6592, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6525, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3636, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8925, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4518, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9447, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8487, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8158, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1768, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7682, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3803, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3228, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4922, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2487, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7625, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5446, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2302, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9811, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5488, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(13.5260, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4427, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(14.6703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3853, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7396, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6862, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6423, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5300e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0645, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9225, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9618, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(10.4076, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1636, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5573, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2363, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5538, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3078, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5080, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6495, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3190, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2734, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1110, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4635, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9069, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3655, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2896, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0730e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7415, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2951, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(15.3555, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7957, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5846, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9918, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3101, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8350, device='cuda:0', grad_fn=) +cls_loss: tensor(14.7599, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5966, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2946, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6989, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2075, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9297, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8294, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2867, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3126, device='cuda:0', grad_fn=) +cls_loss: tensor(18.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7201, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4688, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(10.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4635, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8181, device='cuda:0', grad_fn=) +9.890738003669029e-05 +changing lr +epoch 2, time 302.80, cls_loss 2.9470 +306 +cls_loss: tensor(0.4344, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6693, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(30.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1673, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6298, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6458, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5156, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(12.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6296, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5586, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7904, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4779, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2755, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9080, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8652, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4622, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7708, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6783e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8147e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(20.8870, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5605, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8887, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5573, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2682, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1853, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9909, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5957, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2075, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5000, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7774, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4219, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2813, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4395, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3283, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8389, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9544, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4631, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8952, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2812, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8098, device='cuda:0', grad_fn=) +cls_loss: tensor(17.1893, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4775, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1808, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0540, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6215, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1836, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5469, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2731, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4217, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1888, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3405, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0934, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(11.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4962, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5286, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5782, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7751, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3705, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1481, device='cuda:0', grad_fn=) +cls_loss: tensor(14.8685, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8203, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3541, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4141, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6257, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7311, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8868, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5924, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1797, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2563, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8705, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5996, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3262, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4808, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9414, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6725, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4974e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5706, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8818, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9389, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9922, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9699, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3699, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5329, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8461, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7110, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(16.4661, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7605, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4776, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7389, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4212, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2211, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9609, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4805, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0679, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5185, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(10.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8655, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9593, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1347, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8880, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4609, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.755282581475769e-05 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 300.19, cls_loss 1.8753 +306 +cls_loss: tensor(6.3925, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3613, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4795, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1849, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2368, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2588, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7135, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3099, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3293, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6751, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4544, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8267, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6496, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3880, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1929, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2318, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3298, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3669, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6253, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9590, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4376, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6823, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3945, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1406, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9912, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8959, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6417, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9746, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1693, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9225, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6338, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9882, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1719, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3687, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8179, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5625, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6354, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7031, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5964, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8499, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5365, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3307, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2992, device='cuda:0', grad_fn=) +cls_loss: tensor(15.9395, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7422, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4818, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(11.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9157, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0929e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3075, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4395, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8835, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4458, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3838, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8779, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7813, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3268, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4034, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3226, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4451e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7174, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8724, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2709, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8730, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4573e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3372, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8736, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5799, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2614, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2786, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7734, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1706, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5885, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5889, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5312, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9225, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1549, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8977, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5311, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5386e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7318, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0912, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9303, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8169e-05, device='cuda:0', grad_fn=) +9.567727288213003e-05 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 302.37, cls_loss 1.6264 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4857, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5498, device='cuda:0', grad_fn=) +cls_loss: tensor(15.3538, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6250, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(13.2866, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1895, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8255, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7397, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1407, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8600, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(11.7006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4710, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9870, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5986, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7799, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6922, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0911, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2656, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2772, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3511e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1381, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6439, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8105, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1380, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9876, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5837, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6400e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2075, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(13.1302, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5306, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3249, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7174, device='cuda:0', grad_fn=) +cls_loss: tensor(13.4271, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8294, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8611e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7689, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7266, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8783, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1146, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9798, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6680, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2586e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2760, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6515, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9307, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8854, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1953, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6035, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9333, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5934, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5254, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3021, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1156, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5130, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2096, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5262, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1878, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7995, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2762, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3649, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2616e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9414, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2584e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.330127018922194e-05 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 321.56, cls_loss 0.9863 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8346, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8546e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4206, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6094, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5560, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5934, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8486, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2865, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5685, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6289, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9939, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5444, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1914, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7061, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6995, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(10.6368, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6953, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5732, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7630, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7813, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2751, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9974, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7658e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4844, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8776, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3727, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3771, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8281e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5088, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3568, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1432, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8529, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2448, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0956, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6849, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5812, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2777, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7161, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6668, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2293, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3385, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9792, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6263, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0759, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8125, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9402, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1816, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9453, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8458e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6821, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9391e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6250, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8906, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1751, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6771, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5286, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2715, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9062, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4974, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7148, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.045084971874738e-05 +changing lr +epoch 6, time 329.11, cls_loss 0.6325 +306 +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3682, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9271, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7138, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2686, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2785e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4062, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5599, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2604, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9813, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1341, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(11.2057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2503, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1292, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5156, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5029, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8099, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2787, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5356, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3151, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1799, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8260, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0643, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4824, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6042, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1784, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6184, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1544, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(15.1355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4794, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1289, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5339, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0944, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4042e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1546, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2188, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2281, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8984, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4189, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2189, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2878, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7676, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8160e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4167, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3307, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1857e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8802, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8745e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5610, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(10.7670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2318, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0874, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6392, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8542, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5050e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +8.715724127386972e-05 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 328.84, cls_loss 0.6575 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5312, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4811, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1291, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5934, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1641, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3932, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3984, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2454, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8359, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9505, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2786, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9844, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1825, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5209, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1484, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6498, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7624, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5678, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0964, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8802, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9920, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2272, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4167, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6276, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1656e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5781, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1489, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9762, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7799, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0547, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6042, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2835, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2668, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6544, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8229, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +8.345653031794292e-05 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 330.88, cls_loss 0.3828 +306 +cls_loss: tensor(0.7207, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0814, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9850, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4837, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6422, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.6361e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0522, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5326, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7943, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6364, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4799, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9818, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2863, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9909, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7560, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3712, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4970e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3206, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6647, device='cuda:0', grad_fn=) +cls_loss: tensor(16.8367, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6549, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0940, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9400, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5280, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2414, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3643e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4102e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3659, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4889, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5505, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7552, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7786, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2301, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1680, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7214, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5081, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1984, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2648, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1924, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3626, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3854, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8880, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2226, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0625, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5065, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1042, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7057, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7126, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8281, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9154, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8711, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +7.938926261462366e-05 +changing lr +epoch 9, time 331.22, cls_loss 0.5072 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4261, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3894, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5110, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4316, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5582, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9206, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2344, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0863e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1344, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3698, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2546, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6836, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3247e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2968e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3561, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7816e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8333, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3167, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0960, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8503, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6294e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6382, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4154, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9694, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3963, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6751, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +7.500000000000001e-05 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 332.14, cls_loss 0.2086 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6888e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4836e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.8490, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2784e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2109, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7476, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5961e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0930e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7044, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0067e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3333, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4948, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2132, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6940, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6598, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7665e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0995, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1934, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3672, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1453, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3185, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7533, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9596, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3551e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9368, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7109, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3177, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2246, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8841, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4740, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +7.033683215379003e-05 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 327.33, cls_loss 0.2434 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2280e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5249, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1383, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5417, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6250, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1822e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1195e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5130, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3184, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5432e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7703, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7754, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8016e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5280, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4209, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3317, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6309, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7715, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6195, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4749, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0711, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6507, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4175e-06, device='cuda:0', grad_fn=) +6.545084971874738e-05 +changing lr +epoch 12, time 329.78, cls_loss 0.1202 +306 +cls_loss: tensor(8.0665e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9512, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4128, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1750, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0398e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1518, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3942, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2874, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6814, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7240, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4297, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2625, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1113, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4615, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2174e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7433e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4792, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.4107e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8138, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7083, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1326e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3049e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2708, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9121, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.039558454088797e-05 +changing lr +epoch 13, time 329.34, cls_loss 0.0916 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7842e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2912, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4464e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5934, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8320, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9225, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.4333e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9121, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2377, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7370, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2817, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5517, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2463, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.4240e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1913, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6875, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0475, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3687, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1630e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2319e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0257, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.522642316338269e-05 +changing lr +epoch 14, time 333.44, cls_loss 0.0604 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.8030e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9103e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6397, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3724, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9870, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0841, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7734, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4909, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7087e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1634, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.3619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7751e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7352e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2039, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0715, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +5.000000000000002e-05 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 329.08, cls_loss 0.0349 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6159, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8828, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7259, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9746, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6725, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0664e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6146, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6098, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3429, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2451e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2900, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8942e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8610e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9219, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +4.4773576836617344e-05 +changing lr +epoch 16, time 329.02, cls_loss 0.0865 +306 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2964, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4808, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5157, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8214e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3818, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8876e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8047, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2655, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8676e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4813, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1689, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0510e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0894, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0977, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.9604415459112035e-05 +changing lr +epoch 17, time 329.91, cls_loss 0.0235 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2689, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.6890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1740, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6690e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4583, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2652e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.2347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4443, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1325e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6569, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6113, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9074e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.6067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4479, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3828, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +3.4549150281252636e-05 +changing lr +epoch 18, time 327.45, cls_loss 0.0265 +306 +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8413e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5267, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4375, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6095e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2314, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4625e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.6346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7022e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0796e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1524e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0994e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +2.966316784621e-05 +changing lr +epoch 19, time 330.49, cls_loss 0.0079 +306 +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8374e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9460, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9133, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3438, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2637, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0288, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1640, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5579, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1901e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1724e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1305, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3444e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6967, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7484e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9313e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9173, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.6757e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2254e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.5000000000000015e-05 +changing lr +epoch 20, time 329.35, cls_loss 0.0503 +306 +cls_loss: tensor(6.5843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0455, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7286e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3058, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9804e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0533, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4460, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2120e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9479, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7611, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5052, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.5169, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3047e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8239e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7881e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1668, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8594, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4359, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8516, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +2.0610737385376352e-05 +changing lr +epoch 21, time 326.17, cls_loss 0.0363 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3242, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1696, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3988, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4372e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5656e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.4506e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.0531e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4258, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6094e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9140e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4710, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3447e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.0135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +1.654346968205711e-05 +changing lr +epoch 22, time 327.71, cls_loss 0.0188 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8358e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7721, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8080e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0201e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3578e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6436, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2277, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0476, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6279, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.4639e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5100e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +1.2842758726130304e-05 +changing lr +epoch 23, time 330.17, cls_loss 0.0121 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1591e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.7181, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6292e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9751, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0925, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1667, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4239e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.8949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1127e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9008e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2581, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.3910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7156e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5499e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1414, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +9.549150281252636e-06 +changing lr +epoch 24, time 328.04, cls_loss 0.0214 +306 +cls_loss: tensor(2.6623e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2121e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0782, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0200e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3512e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3844e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5221e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6988, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.3944, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.8122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1230, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5924, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7552e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.2518e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.6559, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5167e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +6.698729810778068e-06 +changing lr +epoch 25, time 329.09, cls_loss 0.0270 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.3976e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.0267e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7188, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.6689e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1965, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5697e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7365e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0001e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.1094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.8279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3381e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.2386e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7021e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5698e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +4.322727117869953e-06 +changing lr +epoch 26, time 328.98, cls_loss 0.0144 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.4637e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1062e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2318e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.8731, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.2981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.5566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0972, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5332e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3644e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7155e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.1870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7683e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.9539e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.5763e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9668, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.7760, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5696e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3351e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.9010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3842e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.5055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.3645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.5817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.447174185242324e-06 +changing lr +epoch 27, time 334.18, cls_loss 0.0122 +306 +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7442, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9802e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5385e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.1855e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0530e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2452e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3454, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.3710e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5367e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.5829e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.6175, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.5433e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0815, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7750e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5631e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0134e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.7354e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3289, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.0663e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.9720, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3908e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0755, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1592e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.9341e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.2591, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.6149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2716e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.4493e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7551e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.1657e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +1.092619963309716e-06 +changing lr +epoch 28, time 332.14, cls_loss 0.0115 +306 +cls_loss: tensor(3.3776e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.9671e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.1921e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9207e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0447, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.2849e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8477e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.0339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0598e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.8148e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1348, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1789e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(9.1394e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.7618e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(6.0002e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.8014e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.4571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7815e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0883, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1460e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.9027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.3862e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.1860, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.8612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0416, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9473e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5914e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1922e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.7829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.5558e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.2040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7420e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.1723e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(2.7617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(8.7023e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.4504e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.9736e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(3.8743e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.9868e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(1.5895e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-08, device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +cls_loss: tensor(0., device='cuda:0', grad_fn=) +2.7390523158633003e-07 +changing lr +epoch 29, time 333.24, cls_loss 0.0053 +---------------------saving last model at epoch 29---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 18.505859 16.595563 11.317365 19.648766 15.853898 diff --git a/Meta-causal/code-stage1-pipeline/56529.error b/Meta-causal/code-stage1-pipeline/56529.error new file mode 100644 index 0000000000000000000000000000000000000000..6719d172fffc992e3a94bb0e7c93ae4f696e52cd --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56529.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 24: randm: command not found diff --git a/Meta-causal/code-stage1-pipeline/56529.log b/Meta-causal/code-stage1-pipeline/56529.log new file mode 100644 index 0000000000000000000000000000000000000000..37ddaf98d8e631bf6ee50e8252e96f04757d2a0c --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56529.log @@ -0,0 +1,10437 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 100, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +cls_loss: tensor(2.3018, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2967, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2914, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2888, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2421, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2609, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2621, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2491, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2366, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2654, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1621, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2152, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1575, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1013, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0939, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9794, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8126, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9371, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8364, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9253, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7372, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7174, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5920, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5893, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4467, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5729, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3543, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1536, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3137, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2285, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9117, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.9361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6077, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8701, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6843, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5725, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4988, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 182.72, cls_loss 1.1179 +100 +cls_loss: tensor(0.6869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.7101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.8156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3784, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4575, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4822, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3789, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4645, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2827, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1476, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2374, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2830, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1787, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2944, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2945, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1723, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 181.61, cls_loss 0.3343 +100 +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1866, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1647, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1711, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4902, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1920, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3517, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1758, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1834, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1858, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1815, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3965, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1398, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1845, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0864, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0932, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1733, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4873, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 181.39, cls_loss 0.2231 +100 +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4841, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1923, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4641, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1533, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2955, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1490, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4826, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2966, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2347, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1779, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3870, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 180.81, cls_loss 0.1847 +100 +cls_loss: tensor(0.1245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0938, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1793, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0803, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0948, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2381, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3650, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3569, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1553, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0666, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2640, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1662, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 180.77, cls_loss 0.1422 +100 +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0878, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0770, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2492, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0271, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0745, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0798, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2605, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1547, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0985, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1556, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.6162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0989, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1869, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1494, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0302, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1760, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0756, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3796, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1589, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 5, time 180.23, cls_loss 0.1221 +100 +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0729, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1865, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0690, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0971, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0623, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0799, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3636, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0839, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0349, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2863, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0983, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0899, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1574, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1998, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1687, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1683, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1871, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1557, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2597, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 179.72, cls_loss 0.1258 +100 +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0905, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1000, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2734, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0411, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0994, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0921, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0592, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0990, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1628, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2706, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2776, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2881, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0739, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4801, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0340, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 178.45, cls_loss 0.0844 +100 +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0908, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0428, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0624, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1567, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0444, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1986, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0626, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0963, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1810, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0795, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0786, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.5539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1927, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0673, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2527, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0804, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0952, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0472, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0442, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0561, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 178.44, cls_loss 0.0747 +100 +cls_loss: tensor(0.2122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1975, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1288, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0710, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0850, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0777, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0896, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0811, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0523, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2149, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0721, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0997, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0859, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0467, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0469, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0542, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2915, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 9, time 178.73, cls_loss 0.0658 +100 +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0973, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0510, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0764, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0681, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1807, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1767, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0675, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1898, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0612, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0247, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0697, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0551, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0594, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0847, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 178.65, cls_loss 0.0630 +100 +cls_loss: tensor(0.1233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1615, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0913, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1582, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0263, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1749, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0431, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1455, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0351, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0978, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0930, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0346, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0728, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2269, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 11, time 178.47, cls_loss 0.0731 +100 +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1590, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0402, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0823, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0781, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2191, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0954, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0497, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0583, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1685, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0805, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0369, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0456, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2754, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0751, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2606, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0573, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0423, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0857, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0408, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0678, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0598, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0693, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0322, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3584, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 178.58, cls_loss 0.0598 +100 +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0771, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0820, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0368, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1514, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0653, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0487, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0450, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0738, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1792, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0696, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0500, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1907, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0750, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0630, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 178.67, cls_loss 0.0462 +100 +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0477, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0443, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0633, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4964, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0509, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0836, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1607, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0900, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1485, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0621, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1576, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0763, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0671, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0727, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0318, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0238, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1812, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 14, time 179.05, cls_loss 0.0505 +100 +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0506, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0917, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0294, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0695, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0634, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0817, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0564, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0414, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0667, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0644, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0731, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1676, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 15, time 179.05, cls_loss 0.0410 +100 +cls_loss: tensor(0.0254, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0354, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0499, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0462, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0891, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0535, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0658, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1933, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0324, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0855, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0536, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0672, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0483, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0545, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0478, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 179.14, cls_loss 0.0373 +100 +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0445, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0387, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1486, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0482, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0941, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0720, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0401, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0703, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1852, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0463, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0192, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0244, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0538, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0587, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0364, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 17, time 179.44, cls_loss 0.0374 +100 +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0762, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0308, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0848, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1345, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0732, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0849, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1946, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0507, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0375, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 18, time 179.42, cls_loss 0.0304 +100 +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1774, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0554, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1635, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0980, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0419, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0586, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0656, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0929, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0563, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1519, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0904, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0730, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1829, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 19, time 179.24, cls_loss 0.0376 +100 +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0449, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0164, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0742, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0328, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0440, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0452, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1909, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0363, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0790, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0424, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0824, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0886, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1846, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0589, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0532, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 20, time 179.34, cls_loss 0.0289 +100 +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0766, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0331, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0544, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1661, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3797, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0723, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0744, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0937, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0474, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0840, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0882, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0311, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0741, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 21, time 179.59, cls_loss 0.0318 +100 +cls_loss: tensor(0.0655, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0832, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0931, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1425, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0418, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0310, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0875, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0258, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0717, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 22, time 179.67, cls_loss 0.0259 +100 +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0562, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0856, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0740, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0530, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0461, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0394, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0806, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0501, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0664, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0999, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1654, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0390, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0600, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0465, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0828, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 179.59, cls_loss 0.0215 +100 +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0775, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0266, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0395, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0458, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0708, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0366, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1686, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0868, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0385, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0446, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0769, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0680, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0312, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0773, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0464, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0552, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 24, time 179.65, cls_loss 0.0217 +100 +cls_loss: tensor(0.2044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1722, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0603, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1736, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0272, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0489, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0595, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0209, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0802, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0976, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0838, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0309, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 25, time 179.39, cls_loss 0.0219 +100 +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0362, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0659, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0698, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0339, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0307, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0613, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0737, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2524, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0599, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0885, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0853, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0434, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 26, time 179.46, cls_loss 0.0198 +100 +cls_loss: tensor(0.1287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0370, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0835, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0549, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0350, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0677, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0147, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0252, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0316, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0129, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0330, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0317, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0531, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 27, time 178.94, cls_loss 0.0217 +100 +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0632, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0225, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0984, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0584, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0275, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0376, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0772, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0279, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0780, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0229, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0596, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0335, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 28, time 178.77, cls_loss 0.0193 +100 +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0410, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0185, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0488, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0355, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0448, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0304, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0451, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0480, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0198, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0403, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0521, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0230, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0699, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1276, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0993, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0182, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 29, time 178.99, cls_loss 0.0216 +100 +cls_loss: tensor(0.0459, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0457, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0131, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2611, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0273, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0284, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 30, time 179.12, cls_loss 0.0125 +100 +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0248, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0692, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0579, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1280, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0319, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0161, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0329, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0246, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 31, time 178.96, cls_loss 0.0122 +100 +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0534, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6131e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0901, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0936, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0861, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0876, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0399, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0150, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0566, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0269, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0484, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0872, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0844, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0184, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 179.24, cls_loss 0.0167 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0407, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0372, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0783, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0326, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0188, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0336, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0620, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0245, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0657, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0227, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0253, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0145, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4602e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 179.48, cls_loss 0.0097 +100 +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0226, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0862, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0173, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0338, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0358, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0343, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0295, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0926, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0601, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 179.46, cls_loss 0.0098 +100 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0642, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0684, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0265, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0639, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1821e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0190, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1290, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0422, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0297, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1429e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0616, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2367, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0268, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 35, time 179.65, cls_loss 0.0115 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0221, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0539, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0714, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0726, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0546, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0429, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0436, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0689, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0558, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0608, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0396, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0196, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1447, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0357, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0842, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0125, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0321, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 36, time 179.32, cls_loss 0.0192 +100 +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0388, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0176, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0332, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0202, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0591, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0565, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1281, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0148, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0460, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0943, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0334, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 37, time 179.63, cls_loss 0.0165 +100 +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0113, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0380, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0794, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0713, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0529, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0528, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0453, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 38, time 179.50, cls_loss 0.0099 +100 +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0581, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0195, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0674, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1208, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0473, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0987, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7368e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1593, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0130, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0854, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 39, time 178.62, cls_loss 0.0099 +100 +cls_loss: tensor(0.1314, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0421, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4593e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3297e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0170, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4548e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 40, time 178.17, cls_loss 0.0038 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0256, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1299, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0816, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1590, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7971e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0525, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0712, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1595e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0231, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0560, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0224, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0251, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0285, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0158, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0207, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 41, time 179.17, cls_loss 0.0132 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2566e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0430, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0627, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8892e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0139, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3918e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 179.63, cls_loss 0.0041 +100 +cls_loss: tensor(7.9356e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7414e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6645e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0526, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8860e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0305e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0409, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1411e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7620e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0260, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0204, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2196e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0491, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 43, time 179.33, cls_loss 0.0037 +100 +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0097, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0691, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0809, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0991, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0413, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0619, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0235, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0108, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0291, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3249e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2920e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6244e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9668e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4490e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0355e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0608e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4481e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 44, time 179.66, cls_loss 0.0074 +100 +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4196e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0753, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0094, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4758e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0384, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5024e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6022e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0255, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0165, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0427, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0415, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 45, time 179.02, cls_loss 0.0087 +100 +cls_loss: tensor(6.9752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0167, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0663, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0278, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0106, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0264, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0629e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1410e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 46, time 179.06, cls_loss 0.0054 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1833, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0670, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0325, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1883, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0341, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0187, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0181, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0320, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1520, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0135, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4256e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3073e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1223e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6840e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1228, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5801e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 47, time 179.17, cls_loss 0.0134 +100 +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0199, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0212, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0432, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0614, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0887, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5059e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4574e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0511, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0845e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2257e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3485e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 48, time 179.12, cls_loss 0.0055 +100 +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6150e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0282, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3999e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0360, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0296, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3741e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0496, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9487e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6620e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1610, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0122, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 49, time 179.50, cls_loss 0.0066 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1388e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0391, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0652, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3654e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0588, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0274, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0879, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0102, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0694, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 50, time 179.28, cls_loss 0.0065 +100 +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0417, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6335e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9271e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0193, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0925e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0309e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0243, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7865e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6519e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1761, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5851e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0580, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0175, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0361, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9323e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0386, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 51, time 179.13, cls_loss 0.0070 +100 +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0237, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0404, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0210, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0962, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7536e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7523e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0219, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2668, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0481, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0120, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4157e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0392, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0283, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0234, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5984e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 52, time 179.50, cls_loss 0.0088 +100 +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1270, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0405, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4760e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0911e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0206, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0383, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0646, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9206e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0068, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0293, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4575e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0860, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1105, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0172, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0242, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 53, time 179.20, cls_loss 0.0100 +100 +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0420, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0377, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0629, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0947, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0178, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0236, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8242e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0063, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1219e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0072, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6118e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0329e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0461e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1176e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 54, time 180.20, cls_loss 0.0050 +100 +cls_loss: tensor(1.5691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0935, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1018e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2393, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0406, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2000e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2927e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9052e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0426, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7529e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8398e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7819e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4357e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2240e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0151, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 55, time 178.71, cls_loss 0.0061 +100 +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0736e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0438, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7210e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6191e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0037, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3130e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0138, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0162, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4202e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 56, time 179.30, cls_loss 0.0064 +100 +cls_loss: tensor(4.4063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0117, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0215, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1110e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9259e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3984e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8046e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9227e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9298e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0124, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7430e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1004e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6740e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1397e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0144, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0702, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5191e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4198e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6786e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 57, time 179.10, cls_loss 0.0028 +100 +cls_loss: tensor(3.8136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3076e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0466, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0079, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6907e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0578e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0281, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8342e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2831e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9461e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0141, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0121, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0059, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0504, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6336e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5257e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9127e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9575e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3702e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0315, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0140, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4572e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9397e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 58, time 179.17, cls_loss 0.0044 +100 +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7065e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1011e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0654, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8501e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0171, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0543, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6229e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0327, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0303, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6594e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0487e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1776e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0241, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0800, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1471, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0352, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0217, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0154, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5177e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0505, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0136, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0201, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0058, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0398, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 59, time 179.42, cls_loss 0.0115 +100 +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0070, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9767e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1183, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6423e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0877, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0437, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1707, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1122e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0091, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.3951, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1189, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2337, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0602, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0062, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0153, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0112, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0118, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0087, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9247e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 60, time 179.30, cls_loss 0.0156 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0235e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0470, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9608e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1556e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0169, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0053, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4881e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0222, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0267, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7540e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4631e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8728e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0179, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0286, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 61, time 179.43, cls_loss 0.0031 +100 +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4258e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4217e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4157e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0617, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0200, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3905e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0075, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7438e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2883e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8476e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0353, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2059e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3326e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9188e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1393e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7481e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2039e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0604, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0152, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 62, time 179.66, cls_loss 0.0034 +100 +cls_loss: tensor(3.2596e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0232, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2617e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1493, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1665, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1854e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0194, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0240, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2705, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0359, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0186, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0132, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0609, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7106e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0724, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9577e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 63, time 179.26, cls_loss 0.0154 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9715e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0306, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0051, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1720e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4133e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0052, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5837e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2980e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2119e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1067e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6660e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2841e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5667e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8245e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0180, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4272e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0159, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2118e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 64, time 179.39, cls_loss 0.0016 +100 +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0813e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0400, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6924e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8534e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2223e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9271e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6046e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2706e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0110, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3268e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5192e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5969e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4378e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0136e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1265e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0792e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9150e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0379, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0146, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0156, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0572, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4906e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7679e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1388e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 65, time 179.12, cls_loss 0.0026 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7296e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5248e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9669e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9004e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6137e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7891e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6462e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9357e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1581e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0047, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3469e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1198e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2485e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3993e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0984e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5605e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 66, time 179.32, cls_loss 0.0010 +100 +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5831e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0577, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7534e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5928e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6801e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0220, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0920e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1147e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4892e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0344, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8230e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0433, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0067, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9965e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4470e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1090e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5419e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0897, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6040e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1085e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0515, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1160, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5865e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0548, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1838e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4096e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0115, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0157, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0111, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0301, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0205, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0155, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0060, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0373, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0722e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3185e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 67, time 179.10, cls_loss 0.0074 +100 +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8329e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0300, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0081, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0163, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7893e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0928, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0233, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9801e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0378, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0218, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4950e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8273e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6114e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0516, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6040e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0042, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0669, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4596e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3130e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7145e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0292, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0127, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0974, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0214, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3462e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 68, time 179.10, cls_loss 0.0060 +100 +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0082, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2436e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1349, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0323, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6360e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0371, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0166, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3429e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2632e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6587e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0126, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9536e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0239, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5430e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0562e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5013e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5050e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4166e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4115e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0090e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4096e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0519, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3838e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0313, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3521e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0064, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0261, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0137, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5038e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2233, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2226e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0076, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 69, time 179.34, cls_loss 0.0071 +100 +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0041, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2165e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0597, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0036, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7554e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7857e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0571, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6091e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8290e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.2435, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1133, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0287, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0568, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0061, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0851, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1942e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0412, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0259, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0223, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0128, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0045, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0362e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1268e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0090, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0389, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 70, time 179.09, cls_loss 0.0111 +100 +cls_loss: tensor(8.5853e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0382, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0069, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0249, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0038, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2733e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0074, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0039, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0093, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5810e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0098, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0048, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3061e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0508, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4182e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0143, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0055, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.4213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0527, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0109, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0085, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4110e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0216, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0719, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0050, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0585, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0174, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0029, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0298, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0570, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 71, time 178.65, cls_loss 0.0118 +100 +cls_loss: tensor(0.0116, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0066, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0100, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0743, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1197, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0788, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0028, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5483e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5587e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0104, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0142, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0468, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0333, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0078, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0054, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0651, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0262, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5561e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0134, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0114, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0168, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0096, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2187e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0046, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0071, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0365, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0043, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6352e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0289, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0211, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5309e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5094e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 72, time 178.69, cls_loss 0.0074 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0149, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2748e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0077, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0092, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0660, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0177, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0101, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0065, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6336e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0095, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0031, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2670e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0123, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0084, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0044, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1773e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9140e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0160, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9137e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0305, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0056, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0213, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0038e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1502e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9050e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0218e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0277, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0073, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6230e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5096e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3682e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 73---------------------------------------------------- +epoch 73, time 178.81, cls_loss 0.0035 +100 +cls_loss: tensor(4.3157e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0030, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9966e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8124e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0088, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4523e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5103e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9754e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1898e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9951e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2929e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4959e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7854e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5199e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8587e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8829e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0330e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0165e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5317e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6005e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1749e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0834e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0807e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8599e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2420e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0250, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2308e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6911e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1191e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1263, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5271e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 74---------------------------------------------------- +epoch 74, time 178.84, cls_loss 0.0019 +100 +cls_loss: tensor(7.1581e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0152e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0022, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0083, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0089, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.1688, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0356, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0107, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6719e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2056e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7425e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4156e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0030e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2348e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7233e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7170e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0034, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0026, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0032, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7349e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4612e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0746e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0057, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4794e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8335e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6915e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9796e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0203, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0027, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8121e-05, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 75, time 178.30, cls_loss 0.0032 +100 +cls_loss: tensor(6.8396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6774e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6142e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0017, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1567e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2116e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6756e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2930e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0025, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4169e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4906e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3191e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0119, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0601e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4719e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4607e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2255e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0080, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5889e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4890e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7896e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7078e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4933e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1781e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0040, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0018, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7793e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4464e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2594e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9777e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2701e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 76, time 179.04, cls_loss 0.0006 +100 +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0020, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0099, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1377e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7353e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4145e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0591e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2946e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6794e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1765e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0560e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3068e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3274e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2541e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0014, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6893e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1600e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0049, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6578e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2970e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3913e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9584e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4283e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0156e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6503e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2164e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1048e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5516e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1278e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2964e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4022e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +---------------------saving model at epoch 77---------------------------------------------------- +epoch 77, time 178.92, cls_loss 0.0004 +100 +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7900e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6502e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6636e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9203e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2592e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1921e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0525e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4917e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5091e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0103, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6310e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9938e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1701e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1874e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0033, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3058e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8897e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8497e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7416e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0019, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2610e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9263e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1232e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4438e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0554e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2343e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4423e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0035, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0827e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7621e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4192e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9887e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8124e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5287e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9104e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1754e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6877e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2840e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2256e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2069e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 78, time 178.94, cls_loss 0.0004 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0340e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3471e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6083e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6038e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9858e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4567e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4382e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1388e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5290e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2469e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0787e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1042e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2341e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6459e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6560e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1211e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2590e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5230e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1105e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9594e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9232e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8642e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0528e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6561e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2932e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8041e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3447e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0023, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4775e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9521e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1581e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9748e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9749e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0087e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7244e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5449e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9009e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3337e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7305e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6454e-06, device='cuda:0', grad_fn=) +0.0001 +changing lr +epoch 79, time 178.91, cls_loss 0.0002 +100 +cls_loss: tensor(7.8112e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6466e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6928e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8086e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8523e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5893e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3028e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7999e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1340e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6976e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7839e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8450e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2015e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6173e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0186e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6710e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8323e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3958e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3422e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4527e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2405e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2569e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4000e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8575e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4315e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0021, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3093e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5507e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0080e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9052e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4059e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4815e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 80, time 178.67, cls_loss 0.0001 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2595e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0011, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5982e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7538e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8872e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6624e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3242e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4105e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6219e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1383e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9307e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0125e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3272e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4286e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8056e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4050e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0741e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0290e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4698e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2480e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8296e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5705e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8902e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7046e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9668e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0416e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1979e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4650e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6200e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6280e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3623e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1712e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2362e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6801e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5991e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6971e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7572e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8158e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1580e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1064e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1383e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5528e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3860e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3341e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5619e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0991e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0967e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6070e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7791e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8685e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3576e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3647e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8131e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4990e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 81, time 178.92, cls_loss 0.0001 +100 +cls_loss: tensor(1.3568e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3031e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9926e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1012e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2580e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8040e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1478e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3346e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7293e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0966e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3106e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9527e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6846e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1176e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9316e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2059e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5005e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3074e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5602e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6997e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5384e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2123e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1472e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0016, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3672e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4480e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5825e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7050e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4149e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1577e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3602e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0013, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4176e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0010, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3968e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1126e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8408e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6317e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8110e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4152e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5162e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0392e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0211e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6038e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5854e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7502e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9761e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0207e-06, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 82, time 178.87, cls_loss 0.0001 +100 +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2175e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2314e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9117e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7535e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9569e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9543e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6863e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5224e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7248e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5321e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1213e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3284e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4746e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1523e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0489e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8126e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9088e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4199e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0474e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6574e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4716e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9605e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1197e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5246e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4773e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7996e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0111e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1006e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7412e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2909e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2912e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1932e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4566e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 83, time 179.10, cls_loss 0.0001 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0645e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4952e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7608e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5349e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9977e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0606e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1547e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6263e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7746e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0978e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8179e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6887e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0086, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4111e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2680e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2311e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7540e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2390e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2918e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3310e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5153e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0356e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6387e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5380e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9444e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6848e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0024, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1443e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9717e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7291e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8726e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9765e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7497e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8913e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0892e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4820e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5390e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5322e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6566e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3463e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1782e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8533e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9542e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1902e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1828e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3755e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0707e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 84, time 179.33, cls_loss 0.0002 +100 +cls_loss: tensor(1.1861e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3899e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7577e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6315e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2816e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2708e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0835e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3135e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7798e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7660e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8422e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0206e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6587e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4514e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2691e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3782e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4750e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8060e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2787e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8864e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7993e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1555e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2238e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1962e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8366e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2266e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0005e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0674e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4342e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4074e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4789e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9721e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6168e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8253e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1456e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1526e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8563e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8427e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8165e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6317e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5609e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7635e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4373e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1733e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1342e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6778e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8727e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 85, time 178.58, cls_loss 0.0001 +100 +cls_loss: tensor(4.1053e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7365e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9172e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0289e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8236e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4794e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3767e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1926e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0216e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9397e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0175e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9374e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5769e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5624e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0465e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6905e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8699e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1735e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8498e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6205e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9985e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1835e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3154e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2276e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2629e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6609e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7146e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3950e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0755e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9588e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4685e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6822e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1989e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4501e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5262e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0342e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1859e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9252e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8913e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9190e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2778e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3784e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0677e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3877e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8852e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9878e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6241e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0088e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3693e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6508e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6184e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 86, time 178.76, cls_loss 0.0001 +100 +cls_loss: tensor(4.2751e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4275e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4345e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2805e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4841e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4922e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7703e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6375e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1786e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7083e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0684e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8508e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0683e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1833e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3646e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7195e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9529e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7885e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6202e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8764e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4043e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8430e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7110e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1527e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4075e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1174e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2899e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5220e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5805e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9416e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9210e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7743e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5655e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2507e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1690e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5113e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7508e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1292e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5190e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4801e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 87, time 178.82, cls_loss 0.0001 +100 +cls_loss: tensor(2.1681e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3335e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2528e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1983e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5552e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7100e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3741e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8762e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4450e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1244e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6981e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5279e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1469e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9376e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2361e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4506e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4095e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0278e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7667e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4699e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6066e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2440e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3866e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3694e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0082e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2151e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6692e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2564e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5991e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3849e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6471e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8899e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3020e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7705e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9407e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3039e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1079e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8855e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1569e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8557e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4417e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1588e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4640e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6306e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0479e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3594e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3184e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3358e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 88, time 178.67, cls_loss 0.0001 +100 +cls_loss: tensor(0.0015, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9834e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0628e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3381e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8183e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7048e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7742e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2243e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7878e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5300e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6157e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6364e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0510e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1866e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0814e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9630e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9873e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1262e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9185e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4250e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8040e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5568e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6368e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0220e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4650e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2603e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5746e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2960e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6065e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6734e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8019e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8727e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2546e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5071e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3680e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8348e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5842e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3916e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8012e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3761e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5688e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2202e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2229e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7290e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8068e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9521e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8392e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2517e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8775e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6131e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2282e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1360e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1697e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7344e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1269e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4624e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 89, time 178.88, cls_loss 0.0001 +100 +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7912e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7604e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4193e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6626e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7956e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2990e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3225e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5453e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1780e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8194e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8110e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6439e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4703e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2783e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5785e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0792e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8370e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5949e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0275e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8173e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0445e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4012e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9222e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9858e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2832e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2329e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4823e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0419e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4764e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9593e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6209e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8977e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7660e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6475e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0518e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2680e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8667e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8869e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3868e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2318e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2692e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7193e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0859e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5267e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1459e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1479e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8092e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9494e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1336e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0012, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5257e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8871e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3819e-07, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 90, time 179.28, cls_loss 0.0001 +100 +cls_loss: tensor(6.8072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4317e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6161e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8008e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2023e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7407e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3505e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2569e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2414e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7104e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4789e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8763e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7283e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6615e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0649e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1914e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3463e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8152e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4611e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1038e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8161e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7886e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2807e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3824e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5677e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9545e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9622e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2113e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3528e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4228e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9556e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1057e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8659e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7367e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2585e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7178e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7418e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0668e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3975e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8302e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6309e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6536e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5214e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4394e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9402e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3448e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3887e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3382e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1614e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0249e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1265e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1978e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.5256e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4564e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3246e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7875e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6194e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3830e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1910e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2121e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2527e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9898e-06, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 91, time 179.30, cls_loss 0.0001 +100 +cls_loss: tensor(4.7684e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0936e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1010e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5475e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1676e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3299e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5063e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8498e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4338e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8413e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3937e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2054e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4927e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7805e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3678e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7472e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6450e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6661e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5086e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2473e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8110e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6571e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9486e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0395e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6525e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4499e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4024e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7299e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5457e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4901e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9353e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1396e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0951e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3786e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7369e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0007, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4258e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1898e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9643e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9595e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6806e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9045e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0075e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4320e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8157e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6132e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7869e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9158e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0308e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4500e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0817e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3348e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0634e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5070e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6343e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2826e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9965e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4089e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 92, time 179.17, cls_loss 0.0001 +100 +cls_loss: tensor(8.4154e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0009, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1027e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8662e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7583e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0052e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0639e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3171e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2538e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2975e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0117e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9024e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0853e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4627e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0062e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3404e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6913e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5460e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4736e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3670e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6637e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4399e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2415e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1441e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5092e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9357e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.8836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6856e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9047e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3384e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0646e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9204e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0992e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7049e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3795e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5936e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6112e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7329e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7715e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7044e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7621e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3375e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9504e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3597e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1231e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2349e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2539e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.9435e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0962e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2655e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9078e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6571e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2971e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7208e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3484e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4792e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4108e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3370e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4339e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4503e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9616e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7817e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3496e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8328e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2888e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0009e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9072e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6954e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1720e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0812e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 93, time 179.46, cls_loss 0.0001 +100 +cls_loss: tensor(5.9307e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9290e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0713e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1548e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4260e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9813e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6693e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0008, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0035e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7880e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2251e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2995e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6801e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8323e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8934e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1833e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8168e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5935e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3840e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4469e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7439e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1745e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5034e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0878e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1180e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6131e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5932e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9145e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2024e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2081e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7959e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2884e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2666e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1122e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0495e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9955e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7699e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2493e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0494e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7799e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.8768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5128e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1780e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1125e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1372e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0355e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1732e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6955e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4573e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4799e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3633e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1988e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3371e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5270e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 94, time 179.61, cls_loss 0.0001 +100 +cls_loss: tensor(6.8359e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7097e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4448e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1456e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1642e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5512e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3802e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1399e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1713e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8876e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5744e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3628e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2960e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8266e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4334e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7509e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6941e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4421e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2989e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5133e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3858e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3269e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9792e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6212e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9850e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0910e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2478e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0701e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1402e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4214e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8505e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6657e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4995e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2766e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.8231e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2298e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9804e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2426e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3798e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4238e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4421e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8986e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4913e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0508e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2489e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1886e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7170e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8949e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5065e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2689e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6297e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4815e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1752e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.9423e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1234e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1202e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4622e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1590e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9482e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.6746e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3586e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5337e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0466e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4887e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2797e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1038e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 95, time 179.34, cls_loss 0.0001 +100 +cls_loss: tensor(1.4622e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.9332e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5528e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5995e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6082e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3839e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1845e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7486e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3256e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5799e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4445e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6238e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6528e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4894e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0247e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2096e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9700e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9041e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6619e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.7524e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7638e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6426e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1064e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6403e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6834e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2025e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4149e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0511e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9169e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9607e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5714e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6544e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1579e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4380e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7017e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3055e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4529e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7839e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7865e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0502e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3411e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6190e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9961e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6021e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7742e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6473e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8476e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7800e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5943e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9723e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1890e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(8.1733e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5704e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0261e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0973e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.5406e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7303e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6003e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8741e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0259e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5474e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7037e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5273e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3597e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7187e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2965e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.2264e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3944e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2709e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6043e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9164e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0843e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1456e-06, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 96, time 179.23, cls_loss 0.0001 +100 +cls_loss: tensor(6.0860e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9057e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8289e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2859e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2923e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.6880e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2297e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.9320e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0005e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.5841e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1820e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6859e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.8467e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3146e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3945e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0431e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.0152e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6156e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2288e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3815e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5189e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0883e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6033e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5461e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3442e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2990e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5899e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1681e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3873e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8342e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5565e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.0310e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.5231e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4652e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1281e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3994e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0870e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4223e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0203e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1312e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1747e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3143e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2836e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7940e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3605e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7904e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3954e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8934e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.1254e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.4250e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2966e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8740e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9781e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2314e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7014e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6201e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.6622e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.3768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6302e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7162e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5520e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8107e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5337e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0109e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0325e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.0908e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4937e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3875e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8871e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.6385e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0129e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5788e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8959e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 97, time 179.40, cls_loss 0.0001 +100 +cls_loss: tensor(1.2740e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1946e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1730e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(4.6570e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.7181e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.8679e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0617e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4770e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1458e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(6.3542e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3648e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.7428e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1456e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.1197e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4980e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4537e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7993e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9841e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9347e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8007e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9737e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(5.3234e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(9.3020e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9759e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4957e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2277e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1101e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6725e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.2029e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1408e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0915e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(8.3152e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3015e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4754e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0717e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.4521e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7582e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8651e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6599e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4553e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(8.7354e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0005, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0729e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3437e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0006, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0718e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.9935e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8889e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2002e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6051e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0143e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.4090e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2003e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.4128e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2960e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3291e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2919e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8487e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.8450e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0114e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3346e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.5996e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4173e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6477e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.8215e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 98, time 178.99, cls_loss 0.0001 +100 +cls_loss: tensor(2.8454e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8163e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0148e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.6323e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.4434e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7084e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3139e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.6933e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7285e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.9691e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(5.7481e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9327e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.2101e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.7725e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5790e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.3285e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2521e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(7.4401e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0004, device='cuda:0', grad_fn=) +cls_loss: tensor(2.4814e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.1585e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3664e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4827e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.2882e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2331e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0397e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.5615e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.7588e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2532e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.1668e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.5867e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3757e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8845e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7531e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.0407e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2089e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9614e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3600e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.0355e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2105e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.8290e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(8.2295e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.2852e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(2.2460e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6131e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.9279e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(4.3768e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5188e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3027e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(6.8545e-07, device='cuda:0', grad_fn=) +cls_loss: tensor(1.9424e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.3724e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.7579e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0002, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2987e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.3209e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(5.1491e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.5646e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(7.0408e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1404e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(3.2317e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.0502e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(1.1098e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.6855e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(5.9046e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(4.5791e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(9.4797e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(4.0319e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(2.7966e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.1367e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0003, device='cuda:0', grad_fn=) +cls_loss: tensor(1.4432e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(0.0001, device='cuda:0', grad_fn=) +cls_loss: tensor(5.2132e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(1.6056e-06, device='cuda:0', grad_fn=) +cls_loss: tensor(9.1378e-05, device='cuda:0', grad_fn=) +cls_loss: tensor(3.0231e-05, device='cuda:0', grad_fn=) +1e-05 +changing lr +epoch 99, time 179.35, cls_loss 0.0001 +---------------------saving last model at epoch 99---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist svhn ... usps Avg +w/o do (original x) 98.76 28.284419 ... 80.568012 51.575641 + +[1 rows x 6 columns] diff --git a/Meta-causal/code-stage1-pipeline/56540.error b/Meta-causal/code-stage1-pipeline/56540.error new file mode 100644 index 0000000000000000000000000000000000000000..38d934231566fdb9cd98197f76b6dd49f332d93a --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56540.error @@ -0,0 +1,4 @@ +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py:426: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) diff --git a/Meta-causal/code-stage1-pipeline/56540.log b/Meta-causal/code-stage1-pipeline/56540.log new file mode 100644 index 0000000000000000000000000000000000000000..aa21cfd02e90b3e6efc8d3fb59385cc17555ae0e --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56540.log @@ -0,0 +1,151 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 30, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_pipelineAugWoNorm', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 328.72, cls_loss 8.5176 +306 +9.972609476841367e-05 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 323.89, cls_loss 4.1852 +306 +9.890738003669029e-05 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 321.19, cls_loss 1.7210 +306 +9.755282581475769e-05 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 321.77, cls_loss 0.8687 +306 +9.567727288213003e-05 +changing lr +epoch 4, time 322.73, cls_loss 0.5508 +306 +9.330127018922194e-05 +changing lr +epoch 5, time 321.65, cls_loss 0.4191 +306 +9.045084971874738e-05 +changing lr +epoch 6, time 323.65, cls_loss 0.3904 +306 +8.715724127386972e-05 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 324.01, cls_loss 0.2998 +306 +8.345653031794292e-05 +changing lr +epoch 8, time 327.47, cls_loss 0.2084 +306 +7.938926261462366e-05 +changing lr +epoch 9, time 326.22, cls_loss 0.1815 +306 +7.500000000000001e-05 +changing lr +epoch 10, time 325.03, cls_loss 0.1476 +306 +7.033683215379003e-05 +changing lr +epoch 11, time 325.92, cls_loss 0.1094 +306 +6.545084971874738e-05 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 322.71, cls_loss 0.0653 +306 +6.039558454088797e-05 +changing lr +epoch 13, time 321.20, cls_loss 0.0639 +306 +5.522642316338269e-05 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 322.13, cls_loss 0.0464 +306 +5.000000000000002e-05 +changing lr +epoch 15, time 324.52, cls_loss 0.0402 +306 +4.4773576836617344e-05 +changing lr +epoch 16, time 322.56, cls_loss 0.0452 +306 +3.9604415459112035e-05 +changing lr +epoch 17, time 322.39, cls_loss 0.0403 +306 +3.4549150281252636e-05 +changing lr +epoch 18, time 324.36, cls_loss 0.0190 +306 +2.966316784621e-05 +changing lr +epoch 19, time 327.79, cls_loss 0.0250 +306 +2.5000000000000015e-05 +changing lr +epoch 20, time 322.28, cls_loss 0.0416 +306 +2.0610737385376352e-05 +changing lr +epoch 21, time 322.98, cls_loss 0.0203 +306 +1.654346968205711e-05 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 325.56, cls_loss 0.0271 +306 +1.2842758726130304e-05 +changing lr +epoch 23, time 321.58, cls_loss 0.0190 +306 +9.549150281252636e-06 +changing lr +epoch 24, time 327.21, cls_loss 0.0236 +306 +6.698729810778068e-06 +changing lr +epoch 25, time 301.36, cls_loss 0.0107 +306 +4.322727117869953e-06 +changing lr +epoch 26, time 295.79, cls_loss 0.0165 +306 +2.447174185242324e-06 +changing lr +epoch 27, time 296.53, cls_loss 0.0218 +306 +1.092619963309716e-06 +changing lr +epoch 28, time 297.90, cls_loss 0.0198 +306 +2.7390523158633003e-07 +changing lr +epoch 29, time 299.73, cls_loss 0.0094 +---------------------saving last model at epoch 29---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_pipelineAugWoNorm', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep30_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_pipelineAugWoNorm/art_painting_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 93.017578 53.412969 88.203593 45.838636 62.485066 diff --git a/Meta-causal/code-stage1-pipeline/56541.error b/Meta-causal/code-stage1-pipeline/56541.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-stage1-pipeline/56541.log b/Meta-causal/code-stage1-pipeline/56541.log new file mode 100644 index 0000000000000000000000000000000000000000..df2e118cf0696d37e84e6dee5b38c9fa6814d806 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/56541.log @@ -0,0 +1,432 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 100, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_pipelineAugWoNorm', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 183.25, cls_loss 2.1515 +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 183.70, cls_loss 1.7865 +100 +0.0001 +changing lr +epoch 2, time 183.59, cls_loss 1.5733 +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 183.31, cls_loss 1.4407 +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 183.01, cls_loss 1.3369 +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 183.85, cls_loss 1.3080 +100 +0.0001 +changing lr +epoch 6, time 182.43, cls_loss 1.2082 +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 182.60, cls_loss 1.1517 +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 183.05, cls_loss 1.0938 +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 183.04, cls_loss 1.0485 +100 +0.0001 +changing lr +epoch 10, time 182.31, cls_loss 1.0636 +100 +0.0001 +changing lr +epoch 11, time 182.08, cls_loss 0.9913 +100 +0.0001 +changing lr +epoch 12, time 182.44, cls_loss 0.9240 +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 182.56, cls_loss 0.8962 +100 +0.0001 +changing lr +epoch 14, time 182.83, cls_loss 0.8474 +100 +0.0001 +changing lr +epoch 15, time 182.24, cls_loss 0.8730 +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 182.40, cls_loss 0.8184 +100 +0.0001 +changing lr +epoch 17, time 182.12, cls_loss 0.8083 +100 +0.0001 +changing lr +epoch 18, time 182.02, cls_loss 0.7381 +100 +0.0001 +changing lr +epoch 19, time 182.19, cls_loss 0.7326 +100 +0.0001 +changing lr +epoch 20, time 181.69, cls_loss 0.6649 +100 +0.0001 +changing lr +epoch 21, time 181.62, cls_loss 0.6849 +100 +0.0001 +changing lr +epoch 22, time 181.68, cls_loss 0.6675 +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 182.29, cls_loss 0.6101 +100 +0.0001 +changing lr +epoch 24, time 182.13, cls_loss 0.6237 +100 +0.0001 +changing lr +epoch 25, time 182.23, cls_loss 0.6229 +100 +0.0001 +changing lr +epoch 26, time 182.24, cls_loss 0.5664 +100 +0.0001 +changing lr +epoch 27, time 182.13, cls_loss 0.5588 +100 +0.0001 +changing lr +epoch 28, time 182.14, cls_loss 0.5539 +100 +0.0001 +changing lr +epoch 29, time 182.35, cls_loss 0.5198 +100 +0.0001 +changing lr +epoch 30, time 182.22, cls_loss 0.5153 +100 +0.0001 +changing lr +epoch 31, time 182.36, cls_loss 0.4764 +100 +0.0001 +changing lr +epoch 32, time 182.13, cls_loss 0.4748 +100 +0.0001 +changing lr +epoch 33, time 181.83, cls_loss 0.4448 +100 +0.0001 +changing lr +epoch 34, time 182.32, cls_loss 0.4358 +100 +0.0001 +changing lr +epoch 35, time 181.92, cls_loss 0.4201 +100 +0.0001 +changing lr +epoch 36, time 181.91, cls_loss 0.3949 +100 +0.0001 +changing lr +epoch 37, time 182.01, cls_loss 0.3818 +100 +0.0001 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 182.02, cls_loss 0.3651 +100 +0.0001 +changing lr +epoch 39, time 182.07, cls_loss 0.3656 +100 +0.0001 +changing lr +epoch 40, time 181.87, cls_loss 0.3864 +100 +0.0001 +changing lr +epoch 41, time 182.33, cls_loss 0.3647 +100 +0.0001 +changing lr +epoch 42, time 182.58, cls_loss 0.3301 +100 +0.0001 +changing lr +---------------------saving model at epoch 43---------------------------------------------------- +epoch 43, time 182.56, cls_loss 0.3279 +100 +0.0001 +changing lr +epoch 44, time 185.15, cls_loss 0.3470 +100 +0.0001 +changing lr +epoch 45, time 182.28, cls_loss 0.2938 +100 +0.0001 +changing lr +epoch 46, time 182.03, cls_loss 0.2920 +100 +0.0001 +changing lr +epoch 47, time 182.53, cls_loss 0.2780 +100 +0.0001 +changing lr +epoch 48, time 182.87, cls_loss 0.2592 +100 +0.0001 +changing lr +epoch 49, time 182.61, cls_loss 0.2725 +100 +0.0001 +changing lr +epoch 50, time 182.34, cls_loss 0.2344 +100 +0.0001 +changing lr +epoch 51, time 182.13, cls_loss 0.2686 +100 +0.0001 +changing lr +epoch 52, time 183.03, cls_loss 0.2475 +100 +0.0001 +changing lr +epoch 53, time 182.25, cls_loss 0.2359 +100 +0.0001 +changing lr +epoch 54, time 182.39, cls_loss 0.2279 +100 +0.0001 +changing lr +epoch 55, time 182.38, cls_loss 0.2340 +100 +0.0001 +changing lr +epoch 56, time 182.19, cls_loss 0.2217 +100 +0.0001 +changing lr +epoch 57, time 182.01, cls_loss 0.2188 +100 +0.0001 +changing lr +epoch 58, time 182.23, cls_loss 0.2269 +100 +0.0001 +changing lr +epoch 59, time 182.47, cls_loss 0.2212 +100 +0.0001 +changing lr +epoch 60, time 182.34, cls_loss 0.1887 +100 +0.0001 +changing lr +epoch 61, time 182.11, cls_loss 0.1859 +100 +0.0001 +changing lr +epoch 62, time 182.40, cls_loss 0.2021 +100 +0.0001 +changing lr +epoch 63, time 182.09, cls_loss 0.1756 +100 +0.0001 +changing lr +epoch 64, time 182.38, cls_loss 0.1737 +100 +0.0001 +changing lr +epoch 65, time 182.21, cls_loss 0.1648 +100 +0.0001 +changing lr +epoch 66, time 182.02, cls_loss 0.1613 +100 +0.0001 +changing lr +epoch 67, time 182.29, cls_loss 0.1569 +100 +0.0001 +changing lr +epoch 68, time 182.29, cls_loss 0.1487 +100 +0.0001 +changing lr +---------------------saving model at epoch 69---------------------------------------------------- +epoch 69, time 182.61, cls_loss 0.1538 +100 +0.0001 +changing lr +epoch 70, time 182.28, cls_loss 0.1653 +100 +0.0001 +changing lr +epoch 71, time 181.94, cls_loss 0.1639 +100 +0.0001 +changing lr +epoch 72, time 181.84, cls_loss 0.1784 +100 +0.0001 +changing lr +epoch 73, time 181.70, cls_loss 0.1843 +100 +0.0001 +changing lr +epoch 74, time 180.53, cls_loss 0.1832 +100 +0.0001 +changing lr +epoch 75, time 180.51, cls_loss 0.1421 +100 +0.0001 +changing lr +epoch 76, time 180.07, cls_loss 0.1224 +100 +0.0001 +changing lr +epoch 77, time 180.21, cls_loss 0.1187 +100 +0.0001 +changing lr +epoch 78, time 180.07, cls_loss 0.1058 +100 +0.0001 +changing lr +epoch 79, time 180.76, cls_loss 0.1301 +100 +1e-05 +changing lr +---------------------saving model at epoch 80---------------------------------------------------- +epoch 80, time 181.07, cls_loss 0.0915 +100 +1e-05 +changing lr +epoch 81, time 180.00, cls_loss 0.0845 +100 +1e-05 +changing lr +epoch 82, time 180.09, cls_loss 0.0767 +100 +1e-05 +changing lr +epoch 83, time 180.14, cls_loss 0.0711 +100 +1e-05 +changing lr +epoch 84, time 180.25, cls_loss 0.0698 +100 +1e-05 +changing lr +epoch 85, time 180.12, cls_loss 0.0682 +100 +1e-05 +changing lr +epoch 86, time 179.91, cls_loss 0.0590 +100 +1e-05 +changing lr +epoch 87, time 179.84, cls_loss 0.0607 +100 +1e-05 +changing lr +epoch 88, time 179.82, cls_loss 0.0634 +100 +1e-05 +changing lr +epoch 89, time 180.04, cls_loss 0.0718 +100 +1e-05 +changing lr +epoch 90, time 179.62, cls_loss 0.0704 +100 +1e-05 +changing lr +epoch 91, time 179.77, cls_loss 0.0669 +100 +1e-05 +changing lr +epoch 92, time 179.87, cls_loss 0.0574 +100 +1e-05 +changing lr +epoch 93, time 179.66, cls_loss 0.0556 +100 +1e-05 +changing lr +epoch 94, time 179.87, cls_loss 0.0631 +100 +1e-05 +changing lr +epoch 95, time 179.67, cls_loss 0.0525 +100 +1e-05 +changing lr +epoch 96, time 179.69, cls_loss 0.0473 +100 +1e-05 +changing lr +epoch 97, time 179.39, cls_loss 0.0470 +100 +1e-05 +changing lr +epoch 98, time 179.75, cls_loss 0.0529 +100 +1e-05 +changing lr +epoch 99, time 180.06, cls_loss 0.0541 +---------------------saving last model at epoch 99---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_pipelineAugWoNorm', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep100_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_pipelineAugWoNorm/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist svhn ... usps Avg +w/o do (original x) 93.89 13.579441 ... 89.436971 40.16719 + +[1 rows x 6 columns] diff --git a/Meta-causal/code-stage1-pipeline/AllEpochs_test_digit_v13.py b/Meta-causal/code-stage1-pipeline/AllEpochs_test_digit_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..822168206b9fb4eaa051f71ca1918c000717354d --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/AllEpochs_test_digit_v13.py @@ -0,0 +1,101 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import mnist_net_my as mnist_net +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--channels', type=int, default=3) +@click.option('--factor_num', type=int, default=16) +@click.option('--stride', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--eval_mapping', type=bool, default=True, help='是否查看mapping学习效果') +def main(gpu, svroot, svpath, channels, factor_num,stride, epoch, eval_mapping): + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + +def evaluate_digit(gpu, svroot, svpath, channels=3, factor_num=16,stride=5,epoch='best', eval_mapping=True): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if channels == 3: + cls_net = mnist_net.ConvNet().cuda() + elif channels == 1: + cls_net = mnist_net.ConvNet(imdim=channels).cuda() + + + epoch_list = [] + file_list = os.listdir(svroot) + for file in file_list: + if('.pkl' in file): + epoch_list.append(file) + print('epoch_list:', epoch_list) + + ''' + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + ''' + + for epoch_file in epoch_list: + print("loading weight of %s"%(epoch_file)) + saved_weight = torch.load(os.path.join(svroot, epoch_file)) + + cls_net.load_state_dict(saved_weight) + cls_net.eval() + + # 测试 + str2fun = { + 'mnist': data_loader.load_mnist, + 'mnist_m': data_loader.load_mnist_m, + 'usps': data_loader.load_usps, + 'svhn': data_loader.load_svhn, + 'syndigit': data_loader.load_syndigit, + } + + columns = ['mnist', 'svhn', 'mnist_m', 'syndigit','usps'] + target = ['svhn', 'mnist_m', 'syndigit','usps'] + + index = ['w/o do (original x)'] + data_result = {} + + for idx, data in enumerate(columns): + teset = str2fun[data]('test', channels=channels) + teloader = DataLoader(teset, batch_size=8, num_workers=0) + # 计算评价指标 + teacc = evaluate(cls_net, teloader) + if data == 'mnist': + acc_avg = np.zeros(teacc.shape) + else: + acc_avg = acc_avg + teacc + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + + data_result['Avg'] = acc_avg + + df = pd.DataFrame(data_result,index = index) + print(df) + if svpath is not None: + df.to_csv(svpath) + +if __name__=='__main__': + main() + diff --git a/Meta-causal/code-stage1-pipeline/AllEpochs_test_pacs_v13.py b/Meta-causal/code-stage1-pipeline/AllEpochs_test_pacs_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..9a54c12cb4164746c76957a03f660880a973b8c7 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/AllEpochs_test_pacs_v13.py @@ -0,0 +1,103 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import resnet as resnet +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--source_domain', type=str, default='art_painting', help='source domain') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--factor_num', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--stride', type=int, default=5) +@click.option('--eval_mapping', type=bool, default=False, help='是否查看mapping学习效果') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') +def main(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network): + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + +def evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num=16, epoch='best', stride=5,eval_mapping=False, network='resnet18'): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if network == 'resnet18': + cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + input_dim = 2048 + + epoch_list = [] + file_list = os.listdir(svroot) + for file in file_list: + if('.pkl' in file): + epoch_list.append(file) + print('epoch_list:', epoch_list) + + ''' + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + ''' + + for epoch_file in epoch_list: + print("loading weight of %s"%(epoch_file)) + saved_weight = torch.load(os.path.join(svroot, epoch_file)) + + cls_net.load_state_dict(saved_weight) + cls_net.eval() + + + columns = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in columns if i!=source_domain] + columns = [source_domain] + target + print("columns:",columns) + + + index = ['w/o do (original x)'] + + data_result = {} + data_result_ours = {} + + for idx, data in enumerate(columns): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=4, num_workers=0) + # 计算评价指标 + acc = evaluate(cls_net, teloader) + data_result_ours[data] = acc + + teacc = evaluate(cls_net, teloader) + if data == source_domain: + acc_avg = np.zeros(teacc.shape) + else: + acc_avg = acc_avg + teacc + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + + data_result['Avg'] = acc_avg + + df = pd.DataFrame(data_result,index = index) + print(df) + + if svpath is not None: + df.to_csv(svpath) + +if __name__=='__main__': + main() + + + diff --git a/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py b/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py new file mode 100644 index 0000000000000000000000000000000000000000..c3012d69bfaecec18db3d7888170baa4b35e7b4d --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/data_loader_joint_v3.py @@ -0,0 +1,861 @@ +''' Digit 实验 +''' +import torch +import torch.nn.functional as F +from torch.utils.data import Dataset, TensorDataset +from torchvision import transforms +from torchvision.datasets import MNIST, SVHN, CIFAR10, STL10, USPS + +import os +import pickle +import numpy as np +import h5py +#import cv2 +from scipy.io import loadmat +from PIL import Image + +from tools.autoaugment import SVHNPolicy, CIFAR10Policy +from tools.randaugment import RandAugment +from tools.causalaugment_v3 import RandAugment_incausal, FactualAugment_incausal, CounterfactualAugment_incausal, MultiCounterfactualAugment_incausal + +from PIL import ImageEnhance + + +transformtypedict=dict(Brightness=ImageEnhance.Brightness, Contrast=ImageEnhance.Contrast, Sharpness=ImageEnhance.Sharpness, Color=ImageEnhance.Color) + +class ImageJitterforX(object): + ''' + from StyleAdv dataaug + ''' + def __init__(self, transformdict): + self.transforms = [(transformtypedict[k], transformdict[k]) for k in transformdict] + + + def __call__(self, img): + out = img + randtensor = torch.rand(len(self.transforms)) + + for i, (transformer, alpha) in enumerate(self.transforms): + r = alpha*(randtensor[i]*2.0 -1.0) + 1 + out = transformer(out).enhance(r).convert('RGB') + + return out + +class TransformLoaderforX: + ''' + from StyleAdv dataaug + ''' + def __init__(self, image_size, + normalize_param = dict(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), + jitter_param = dict(Brightness=0.4, Contrast=0.4, Color=0.4)): + self.image_size = image_size + self.normalize_param = normalize_param + self.jitter_param = jitter_param + + def parse_transform(self, transform_type): + if transform_type=='ImageJitter': + method = ImageJitterforX( self.jitter_param ) + return method + method = getattr(transforms, transform_type) + + if transform_type=='RandomResizedCrop': + return method(self.image_size) + elif transform_type=='CenterCrop': + return method(self.image_size) + elif transform_type=='Resize': + return method([int(self.image_size*1.15), int(self.image_size*1.15)]) + elif transform_type=='Normalize': + return method(**self.normalize_param ) + else: + return method() + + + def get_composed_transform(self, aug = False): + if aug: + #transform_list = ['RandomResizedCrop', 'ImageJitter', 'RandomHorizontalFlip', 'ToTensor', 'Normalize'] + transform_list = ['RandomResizedCrop', 'ImageJitter', 'RandomHorizontalFlip', 'ToTensor'] + else: + #transform_list = ['Resize','CenterCrop', 'ToTensor', 'Normalize'] + #transform_list = ['ToTensor', 'Normalize'] + transform_list = ['ToTensor'] + + tranform0 = [transforms.ToPILImage()] + transform_funcs = [ self.parse_transform(x) for x in transform_list] + tranform_all = tranform0 + transform_funcs + transform = transforms.Compose(tranform_all) + return transform + + +class myTensorDataset(Dataset): + def __init__(self, x, y, transform=None, transform2=None, transform3=None, twox=False): + self.x = x + self.y = y + self.transform = transform + self.transform2 = transform2 + self.transform3 = transform3 + self.twox = twox + def __len__(self): + return len(self.x) + + def __getitem__(self, index): + x = self.x[index] + y = self.y[index] + c, h, w =x.shape + if self.transform is not None: + x_RA = self.transform(x) + # print("x_RA.shape:",x_RA.shape) + if self.transform3 is not None: + x_CA = self.transform3(x_RA) + x_CA = x_CA.reshape(-1,c,h,w) + # print("x_CA.shape:",x_CA.shape) + if self.transform2 is not None: + x_FA = self.transform2(x) + # x_FA = x_FA.view(c,13,h,w) + x_FA = x_FA.reshape(-1,c,h,w) + # print("x_FA_in getitem.shape:",x_FA.shape) + # print("x_FA.shape:",x_FA.shape) + + return (x, x_RA, x_FA, x_CA), y + else: + return (x, x_RA, x_CA), y + else: + if self.transform2 is not None: + x_FA = self.transform2(x) + x_FA = x_FA.reshape(-1,c,h,w) + return (x, x_RA, x_FA), y + else: + if self.twox: + return (x, x_RA), y + else: + x_RA = self.transform(x) + return x_RA, y + + +HOME = os.environ['HOME'] +print(HOME) +def resize_imgs(x, size): + ''' 目前只能处理单通道 + x [n, 28, 28] + size int + ''' + resize_x = np.zeros([x.shape[0], size, size]) + for i, im in enumerate(x): + im = Image.fromarray(im) + im = im.resize([size, size], Image.ANTIALIAS) + resize_x[i] = np.asarray(im) + return resize_x + +def load_mnist(split='train', translate=None, twox=False, ntr=None, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + ''' + autoaug == 'AA', AutoAugment + 'FastAA', Fast AutoAugment + 'RA', RandAugment + channels == 3 默认返回 rgb 3通道图像 + 1 返回单通道图像 + ''' + #path = f'data/mnist-{split}.pkl' + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/minst-{split}.pkl' + if not os.path.exists(path): + dataset = MNIST(f'{HOME}/.pytorch/MNIST', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + if split=='train': + x, y = x[0:10000], y[0:10000] + x = torch.tensor(resize_imgs(x.numpy(), 32)) + x = (x.float()/255.).unsqueeze(1).repeat(1,3,1,1) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + with open(path, 'rb') as f: + # print("reading!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!") + x, y = pickle.load(f) + if channels == 1: + x = x[:,0:1,:,:] + + if ntr is not None: + x, y = x[0:ntr], y[0:ntr] + + # 如果没有数据增强 + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + + + #fuyuqian: add styleadv-style aug + transform_x_train = TransformLoaderforX((x.shape[-2], x.shape[-1])).get_composed_transform(aug=True) + transform_x_test = TransformLoaderforX((x.shape[-2], x.shape[-1])).get_composed_transform(aug=False) + if(split == 'train'): + transformed_images = [] + for img in x: + img = transform_x_train(img) # Apply transform to each image + transformed_images.append(img) + x = torch.stack(transformed_images) + #print('x_aug train here', x.shape) + else: + transformed_images = [] + for img in x: + img = transform_x_test(img) # Apply transform to each image + transformed_images.append(img) + x = torch.stack(transformed_images) + #print('x_aug test here', x.shape) + + + + transform = [transforms.ToPILImage()] + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def load_cifar10(split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + dataset = CIFAR10(f'{HOME}/.pytorch/CIFAR10', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + x = x.transpose(0,3,1,2) + + x, y = torch.tensor(x), torch.tensor(y) + x = x.float()/255. + print(x.shape,y.shape) + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset +def load_IMG(task='S-U', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + # path = f'data/img2vid/{domain}/stanford40_12.npz' + if task == 'S-U': + path = f'data/img2vid/{task}/stanford40_12.npz' + elif task == 'E-H': + path = f'data/img2vid/{task}/EAD50_13.npz' + print(path) + dataset = np.load(path) + x, y = dataset['x'], dataset['y'] + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def load_VID(task='S-U',split='1'): + if task == 'S-U': + path = f'data/img2vid/{task}/ucf101_12_frame_sample8_{split}.npz' + elif task == 'E-H': + path = f'data/img2vid/{task}/hmdb51_13_frame_sample8_{split}.npz' + dataset = np.load(path) + print(path) + x, y = dataset['x'], dataset['y'] + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + dataset = TensorDataset(x, y) + return dataset + +def load_pacs(domain='photo', split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + #path = f'data/PACS/{domain}_{split}.hdf5' + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/{domain}_{split}.hdf5' + dataset = h5py.File(path, 'r') + x, y = dataset['images'], dataset['labels'] + #for i in range(20): + # cv2.imwrite('data/PACS/debug_images/img_cv2_'+domain+'_'+split+'_'+str(i)+'.png', x[i]) + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + + #x = x.transpose(0,3,1,2) + # Convert image data to uint8 + + + #fuyuqian: add styleadv-style aug + x = x.astype(np.uint8) + transform_x_train = TransformLoaderforX((x.shape[-3], x.shape[-2])).get_composed_transform(aug=True) + transform_x_test = TransformLoaderforX((x.shape[-3], x.shape[-2])).get_composed_transform(aug=False) + if(split == 'train'): + transformed_images = [] + for img in x: + img = transform_x_train(img) # Apply transform to each image + transformed_images.append(img) + x = torch.stack(transformed_images) + #print('x_aug train here', x.shape) + else: + transformed_images = [] + for img in x: + img = transform_x_test(img) # Apply transform to each image + transformed_images.append(img) + x = torch.stack(transformed_images) + #print('x_aug test here', x.shape) + + + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + + y = y - 1 + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def read_dataset(domain, split): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/{domain}_{split}.hdf5' + dataset = h5py.File(path, 'r') + x_temp, y_temp = dataset['images'], dataset['labels'] + b, g, r = np.split(x_temp,3,axis=-1) + x_temp = np.concatenate((r,g,b),axis=-1) + x_temp = x_temp.transpose(0,3,1,2) + x_temp, y_temp = torch.tensor(x_temp), torch.tensor(y_temp, dtype=torch.long) + y_temp = y_temp - 1 + x_temp = x_temp.float()/255. + return x_temp, y_temp + +def load_pacs_multi(target_domain=['photo'], split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + domains = ['art_painting', 'cartoon', 'photo', 'sketch'] + source_domain = [i for i in domains if i != target_domain] + for i in range(len(source_domain)): + x_temp, y_temp = read_dataset(source_domain[i],split=split) + print(x_temp.shape,y_temp.shape) + if i == 0: + x = x_temp.clone() + y = y_temp.clone() + else: + x = torch.cat([x,x_temp],0) + y = torch.cat([y,y_temp],0) + print(x.shape,y.shape) + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + + +def load_cifar10_c_level1(dataroot): + path = f'data/cifar10_c_level1.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level1") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[0:10000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level1") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level2(dataroot): + path = f'data/cifar10_c_level2.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level2") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[10000:20000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level2") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level3(dataroot): + path = f'data/cifar10_c_level3.pkl' + if not os.path.exists(path): + print("generating cifar10_c_level3") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[20000:30000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level3") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level4(dataroot): + path = f'data/cifar10_c_level4.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level4") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[30000:40000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level4") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level5(dataroot): + path = f'data/cifar10_c_level5.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level5") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[40000:50000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level5") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c(dataroot): + y = np.load(os.path.join(dataroot, 'labels.npy')) + print("y.shape:",y.shape) + y_single = y[0:10000] + x1 = torch.zeros((190000,3,32,32)) + x2 = torch.zeros((190000,3,32,32)) + x3 = torch.zeros((190000,3,32,32)) + x4 = torch.zeros((190000,3,32,32)) + x5 = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y_total = y_single + else: + y_total = np.hstack((y_total,y_single)) + print("y_total.shape:",y_total.shape) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + x = np.load(os.path.join(dataroot,filename)) + x = x.transpose(0,3,1,2) + x = torch.tensor(x) + x = x.float()/255. + print(x.shape) + x1[index*10000:(index+1)*10000] = x[0:10000] + x2[index*10000:(index+1)*10000] = x[10000:20000] + x3[index*10000:(index+1)*10000] = x[20000:30000] + x4[index*10000:(index+1)*10000] = x[30000:40000] + x5[index*10000:(index+1)*10000] = x[40000:50000] + index = index + 1 + # x1, x2, x3, x4, x5, y_total = torch.tensor(x1), torch.tensor(x2), torch.tensor(x3),\ + # torch.tensor(x4),torch.tensor(x5),torch.tensor(y_total) + y_total = torch.tensor(y_total) + dataset1 = TensorDataset(x1, y_total) + dataset2 = TensorDataset(x2, y_total) + dataset3 = TensorDataset(x3, y_total) + dataset4 = TensorDataset(x4, y_total) + dataset5 = TensorDataset(x5, y_total) + return dataset1,dataset2,dataset3,dataset4,dataset5 + +def load_cifar10_c_class(dataroot,CORRUPTIONS): + y = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = y[0:10000] + y_single = torch.tensor(y_single) + print("y.shape:",y.shape) + x = np.load(os.path.join(dataroot,CORRUPTIONS+'.npy')) + print("loading data of",os.path.join(dataroot,CORRUPTIONS+'.npy')) + x = x.transpose(0,3,1,2) + x = torch.tensor(x) + x = x.float()/255. + dataset = [] + for i in range(5): + x_single = x[i*10000:(i+1)*10000] + dataset.append(TensorDataset(x_single, y_single)) + return dataset + +def load_usps(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/usps-{split}.pkl' + if not os.path.exists(path): + dataset = USPS(f'{HOME}/.pytorch/USPS', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + x = torch.tensor(resize_imgs(x, 32)) + x = (x.float()/255.).unsqueeze(1).repeat(1,3,1,1) + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + with open(path, 'rb') as f: + x, y = pickle.load(f) + if channels == 1: + x = x[:,0:1,:,:] + dataset = TensorDataset(x, y) + return dataset + +def load_svhn(split='train', channels=3): + dataset = SVHN(f'{HOME}/.pytorch/SVHN', split=split, download=True) + x, y = dataset.data, dataset.labels + x = x.astype('float32')/255. + x, y = torch.tensor(x), torch.tensor(y) + if channels == 1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + + +def load_syndigit(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/synth_{split}_32x32.mat' + data = loadmat(path) + x, y = data['X'], data['y'] + x = np.transpose(x, [3, 2, 0, 1]).astype('float32')/255. + y = y.squeeze() + x, y = torch.tensor(x), torch.tensor(y) + if channels == 1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + +def load_mnist_m(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/mnist_m-{split}.pkl' + with open(path, 'rb') as f: + x, y = pickle.load(f) + x, y = torch.tensor(x.astype('float32')/255.), torch.tensor(y) + if channels==1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + +if __name__=='__main__': + dataset = load_mnist(split='train') + print('mnist train', len(dataset)) + dataset = load_mnist('test') + print('mnist test', len(dataset)) + dataset = load_mnist_m('test') + print('mnsit_m test', len(dataset)) + dataset = load_svhn(split='test') + print('svhn', len(dataset)) + dataset = load_usps(split='test') + print('usps', len(dataset)) + dataset = load_syndigit(split='test') + print('syndigit', len(dataset)) + diff --git a/Meta-causal/code-stage1-pipeline/env.yaml b/Meta-causal/code-stage1-pipeline/env.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b0bd424fb7c5aa818f10a82173549eb0dd3199c7 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/env.yaml @@ -0,0 +1,119 @@ +name: Py3.7_torch1.8 +channels: + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ + - conda-forge + - bioconda + - defaults +dependencies: + - _libgcc_mutex=0.1=main + - asn1crypto=1.2.0=py37_0 + - blas=1.0=mkl + - bottleneck=1.3.2=py37heb32a55_1 + - bzip2=1.0.8=h7b6447c_0 + - ca-certificates=2021.10.8=ha878542_0 + - cairo=1.14.12=h8948797_3 + - certifi=2021.10.8=py37h89c1867_1 + - cffi=1.13.0=py37h2e261b9_0 + - chardet=3.0.4=py37_1003 + - click=8.0.3=pyhd3eb1b0_0 + - conda-package-handling=1.6.0=py37h7b6447c_0 + - cryptography=2.8=py37h1ba5d50_0 + - ffmpeg=4.0=hcdf2ecd_0 + - fontconfig=2.13.0=h9420a91_0 + - freeglut=3.0.0=hf484d3e_5 + - freetype=2.11.0=h70c0345_0 + - glib=2.63.1=h5a9c865_0 + - graphite2=1.3.14=h23475e2_0 + - h5py=2.8.0=py37h3010b51_1003 + - harfbuzz=1.8.8=hffaf4a1_0 + - hdf5=1.10.2=hba1933b_1 + - icu=58.2=he6710b0_3 + - idna=2.8=py37_0 + - intel-openmp=2021.3.0=h06a4308_3350 + - jasper=2.0.14=hd8c5072_2 + - jpeg=9d=h7f8727e_0 + - libedit=3.1.20181209=hc058e9b_0 + - libffi=3.2.1=hd88cf55_4 + - libgcc-ng=9.1.0=hdf63c60_0 + - libgfortran-ng=7.5.0=ha8ba4b0_17 + - libgfortran4=7.5.0=ha8ba4b0_17 + - libglu=9.0.0=hf484d3e_1 + - libopencv=3.4.2=hb342d67_1 + - libopus=1.3.1=h7b6447c_0 + - libpng=1.6.37=hbc83047_0 + - libprotobuf=3.17.2=h4ff587b_1 + - libstdcxx-ng=9.1.0=hdf63c60_0 + - libtiff=4.1.0=h2733197_0 + - libuuid=1.0.3=h7f8727e_2 + - libvpx=1.7.0=h439df22_0 + - libxcb=1.14=h7b6447c_0 + - libxml2=2.9.9=hea5a465_1 + - mkl=2021.3.0=h06a4308_520 + - mkl-service=2.4.0=py37h7f8727e_0 + - mkl_fft=1.3.1=py37hd3c417c_0 + - mkl_random=1.2.2=py37h51133e4_0 + - ncurses=6.1=he6710b0_1 + - numexpr=2.7.3=py37h22e1b3c_1 + - numpy-base=1.21.2=py37h79a1101_0 + - opencv=3.4.2=py37h6fd60c2_1 + - openssl=1.1.1h=h516909a_0 + - pandas=1.3.3=py37h8c16a72_0 + - pcre=8.45=h295c915_0 + - pip=19.3.1=py37_0 + - pixman=0.40.0=h7f8727e_1 + - protobuf=3.17.2=py37h295c915_0 + - py-opencv=3.4.2=py37hb342d67_1 + - pycosat=0.6.3=py37h14c3975_0 + - pycparser=2.19=py37_0 + - pyopenssl=19.0.0=py37_0 + - pysocks=1.7.1=py37_0 + - python=3.7.4=h265db76_1 + - python-dateutil=2.8.2=pyhd3eb1b0_0 + - python_abi=3.7=2_cp37m + - pytz=2021.3=pyhd3eb1b0_0 + - readline=7.0=h7b6447c_5 + - requests=2.22.0=py37_0 + - ruamel_yaml=0.15.46=py37h14c3975_0 + - scipy=1.7.1=py37h292c36d_2 + - setuptools=41.4.0=py37_0 + - six=1.12.0=py37_0 + - sqlite=3.30.0=h7b6447c_0 + - tensorboardx=2.2=pyhd3eb1b0_0 + - tk=8.6.8=hbc83047_0 + - tqdm=4.36.1=py_0 + - urllib3=1.24.2=py37_0 + - wheel=0.33.6=py37_0 + - xz=5.2.4=h14c3975_4 + - yaml=0.1.7=had09818_2 + - zlib=1.2.11=h7b6447c_3 + - zstd=1.3.7=h0b5b093_0 + - pip: + - absl-py==1.0.0 + - cachetools==4.2.4 + - conda-pack==0.6.0 + - google-auth==2.3.3 + - google-auth-oauthlib==0.4.6 + - grpcio==1.42.0 + - importlib-metadata==4.8.2 + - markdown==3.3.6 + - numpy==1.21.3 + - oauthlib==3.1.1 + - pillow==8.4.0 + - pyasn1==0.4.8 + - pyasn1-modules==0.2.8 + - requests-oauthlib==1.3.0 + - rsa==4.8 + - tensorboard==2.7.0 + - tensorboard-data-server==0.6.1 + - tensorboard-plugin-wit==1.8.0 + - torch==1.8.1+cu111 + - torchvision==0.9.1+cu111 + - typing-extensions==3.10.0.2 + - werkzeug==2.0.2 + - zipp==3.6.0 +prefix: /home/chenjin/miniconda3/envs/Py3.7_torch1.8 diff --git a/Meta-causal/code-stage1-pipeline/main_my_joint_v13_auto.py b/Meta-causal/code-stage1-pipeline/main_my_joint_v13_auto.py new file mode 100644 index 0000000000000000000000000000000000000000..6c7dbc66ee116fe2049d41145424827488c827bc --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/main_my_joint_v13_auto.py @@ -0,0 +1,279 @@ + +''' +训练 base 模型 +''' + +import torch +import torch.nn as nn +import torch.nn.functional as F +import itertools +from torch import optim +from torch.utils.data import DataLoader, RandomSampler +from torchvision import models +from torchvision.datasets import CIFAR10 +from torchvision.utils import make_grid +import torchvision.transforms as transforms +from tensorboardX import SummaryWriter +from torch.cuda.amp import autocast,GradScaler + +import os +import click +import time +import numpy as np + +from network import mnist_net_my as mnist_net +from network import wideresnet as wideresnet +from network import resnet as resnet +from network import adaptor_v2 + +from tools import causalaugment_v3 as causalaugment +import data_loader_joint_v3 as data_loader +# from utils import set_requires_grad + +HOME = os.environ['HOME'] + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择gpu') +@click.option('--data', type=str, default='mnist', help='数据集名称') +@click.option('--ntr', type=int, default=None, help='选择训练集前ntr个样本') +@click.option('--translate', type=float, default=None, help='随机平移数据增强') +@click.option('--autoaug', type=str, default=None, help='AA FastAA RA') +@click.option('--n', type=int, default=3, help='选择多少个factor生成RA') +@click.option('--stride', type=int, default=5, help='if autoaug==CA_multiple, stride is used') +@click.option('--factor_num', type=int, default=16, help='the first n factors') +@click.option('--epochs', type=int, default=100) +@click.option('--nbatch', type=int, default=100, help='每个epoch中batch的数量') +@click.option('--batchsize', type=int, default=128, help='每个batch中样本的数量') +@click.option('--lr', type=float, default=1e-3) +@click.option('--lr_scheduler', type=str, default='none', help='是否选择学习率衰减策略') +@click.option('--svroot', type=str, default='./saved', help='项目文件保存路径') +@click.option('--clsadapt', type=bool, default=True, help='映射后是否用分类损失') +@click.option('--lambda_causal', type=float, default=1, help='the weight of reconstruction during mapping and causal ') +@click.option('--lambda_re', type=float, default=1, help='the weight of reconstruction during mapping and causal ') +@click.option('--randm', type=bool, default=True, help='m取值是否randm') +@click.option('--randn', type=bool, default=False, help='原始特征是否detach') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') + + +def experiment(gpu, data, ntr, translate, autoaug, n, stride, factor_num, epochs, nbatch, batchsize, lr, lr_scheduler, svroot, clsadapt, lambda_causal,lambda_re,randm,randn,network): + settings = locals().copy() + print(settings) + + # 全局设置 + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + if not os.path.exists(svroot): + os.makedirs(svroot) + log_file = open(svroot+os.sep+'log.log',"w") + log_file.write(str(settings)+'\n') + writer = SummaryWriter(svroot) + + # 加载数据集和模型 + if data in ['mnist', 'mnist_t']: + if data == 'mnist': + trset = data_loader.load_mnist('train', translate=translate,twox=True, ntr=ntr, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + elif data == 'mnist_t': + trset = data_loader.load_mnist_t('train', translate=translate, ntr=ntr) + teset = data_loader.load_mnist('test') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=0, \ + sampler=RandomSampler(trset, True, nbatch*batchsize)) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=0, shuffle=False) + cls_net = mnist_net.ConvNet().cuda() + + parameter_list = [] + parameter_list.append({'params':cls_net.parameters(),'lr':lr}) + opt = optim.Adam(parameter_list, lr=lr) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.95) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=int(epochs*0.8)) + + elif data == 'cifar10': + # 加载数据集 + trset = data_loader.load_cifar10(split='train',twox=True, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + teset = data_loader.load_cifar10(split='test') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=4, shuffle=True, drop_last=True) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=4, shuffle=False) + cls_net = wideresnet.WideResNet(16, 10, 4).cuda() + # cls_opt = optim.SGD(cls_net.parameters(), lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(256,512,256,4).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + parameter_list.append({'params':cls_net.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + #print("---------------------------------------------------------------------------------------") + # opt = optim.Adam(parameter_list) + opt = optim.SGD(parameter_list, lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.95) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=int(epochs*0.8)) + + + elif data in ['art_painting', 'cartoon', 'photo', 'sketch']: + # 加载数据集 + trset = data_loader.load_pacs(domain=data, split='train', twox=True, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + teset = data_loader.load_pacs(domain=data, split='val') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=4, shuffle=True, drop_last=True) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=4, shuffle=False) + if network == 'resnet18': + cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + + classifier_param = list(map(id, cls_net.class_classifier.parameters())) + backbone_param = filter(lambda p: id(p) not in classifier_param and p.requires_grad, cls_net.parameters()) + + parameter_list = [] + parameter_list.append({'params':backbone_param,'lr':0.01*lr}) + parameter_list.append({'params':cls_net.class_classifier.parameters(),'lr':lr}) + + opt = optim.SGD(parameter_list, momentum=0.9, nesterov=True, weight_decay=5e-4) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.99999) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=15) + + cls_criterion = nn.CrossEntropyLoss() + + # 开始训练 + best_acc = 0 + best_acc_t = 0 + scaler = GradScaler() + for epoch in range(epochs): + t1 = time.time() + loss_list = [] + cls_net.train() + print(len(trloader)) + for i, (x_four,y) in enumerate(trloader): + x, x_RA, x_FA, x_CA, y = x_four[0].cuda(), x_four[1].cuda(), x_four[2].cuda(), x_four[3].cuda(), y.cuda() + #print('x:', x.shape, 'x_RA:', x_RA.shape, 'x_FA:', x_FA.shape, 'x_CA:', x_CA.shape, 'y:', y.shape) + b, c, h, w = x.shape + with autocast(): + p,f = cls_net(x) + #print('p:', p.size(), 'f:', f.size()) + + cls_loss = cls_criterion(p, y) + #print('cls_loss:', cls_loss) + + loss = cls_loss + + opt.zero_grad() + scaler.scale(loss).backward() + scaler.step(opt) + scaler.update() + #loss_list.append([cls_loss.item(), cls_loss_mapping.item(),cls_loss_causal.item(), re_mapping.item(), re_causal.item()]) + loss_list.append(cls_loss.item()) + + # 调整学习率 + if lr_scheduler in ['cosine', 'Exp', 'Step']: + writer.add_scalar('scalar/lr', opt.param_groups[0]["lr"], epoch) + print(opt.param_groups[0]["lr"]) + print("changing lr") + scheduler.step() + #cls_loss, cls_loss_mapping, cls_loss_causal, re_mapping, re_causal = np.mean(loss_list, 0) + cls_loss = np.mean(loss_list) + + # 测试，并保存最优模型 + cls_net.eval() + if data in ['mnist', 'mnist_t', 'cifar10', 'mnistvis', 'art_painting', 'cartoon', 'photo', 'sketch']: + teacc = evaluate(cls_net, teloader) + + if best_acc < teacc: + print(f'---------------------saving model at epoch {epoch}----------------------------------------------------') + log_file.write(f'saving model at epoch {epoch}\n') + + best_acc = teacc + torch.save(cls_net.state_dict(),os.path.join(svroot, 'best_cls_net.pkl')) + + if ((epoch+1)%5==0): + torch.save(cls_net.state_dict(),os.path.join(svroot, f'epoch{epoch}_cls_net.pkl')) + + # 保存日志 + t2 = time.time() + #print(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f} cls_loss_mapping {cls_loss_mapping:.4f} cls_loss_causal {cls_loss_causal:.4f} re_mapping {re_mapping:.4f} re_causal {re_causal:.4f} /// teacc {teacc:2.2f} lr {opt.param_groups[0]["lr"]:.8f}') + print(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f}') + + #log_file.write(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f} cls_loss_mapping {cls_loss_mapping:.4f} cls_loss_causal {cls_loss_causal:.4f} re_mapping {re_mapping:.4f} re_causal {re_causal:.4f} /// teacc {teacc:2.2f} lr {opt.param_groups[0]["lr"]:.8f} \n') + log_file.write(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f}') + + writer.add_scalar('scalar/cls_loss', cls_loss, epoch) + #writer.add_scalar('scalar/cls_loss_mapping', cls_loss_mapping, epoch) + #writer.add_scalar('scalar/cls_loss_causal', cls_loss_causal, epoch) + #writer.add_scalar('scalar/re_mapping', re_mapping, epoch) + #writer.add_scalar('scalar/re_causal', re_causal, epoch) + writer.add_scalar('scalar/teacc', teacc, epoch) + + print(f'---------------------saving last model at epoch {epoch}----------------------------------------------------') + log_file.write(f'saving last model at epoch {epoch}\n') + torch.save(cls_net.state_dict(),os.path.join(svroot, 'last_cls_net.pkl')) + writer.close() + + +def evalute_pacs(source_domain,cls_net): + cls_net.eval() + data_total = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in data_total if i!=source_domain] + acc = np.zeros(len(target)) + for idx, data in enumerate(target): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=6, num_workers=0) + # 计算评价指标 + acc[idx] = evaluate(cls_net, teloader) + acc_avg = sum(acc)/len(target) + return acc_avg,acc + +def evaluate(net, teloader): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + with torch.no_grad(): + x1 = x1.cuda() + p1,_ = net(x1, mode='fc') + p1 = p1.argmax(dim=1) + ps.append(p1.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +def extract_feature(net, teloader, savedir): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + img_class = y1[0].cpu().numpy() + save_path = os.path.join(savedir,str(img_class)) + if not os.path.exists(save_path): + os.makedirs(save_path) + + with torch.no_grad(): + x1 = x1.cuda() + p1,f1 = net(x1, mode='fc') + save_name = save_path+os.sep+str(i)+'.npy' + np.save(save_name,f1.cpu()) + p1 = p1.argmax(dim=1) + ps.append(p1.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + + + +if __name__=='__main__': + experiment() \ No newline at end of file diff --git a/Meta-causal/code-stage1-pipeline/main_test_digit_v13.py b/Meta-causal/code-stage1-pipeline/main_test_digit_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..fcba8e67b2ac2671b34d90590d87d62fa7ef7da9 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/main_test_digit_v13.py @@ -0,0 +1,85 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import mnist_net_my as mnist_net +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--channels', type=int, default=3) +@click.option('--factor_num', type=int, default=16) +@click.option('--stride', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--eval_mapping', type=bool, default=True, help='是否查看mapping学习效果') +def main(gpu, svroot, svpath, channels, factor_num,stride, epoch, eval_mapping): + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + +def evaluate_digit(gpu, svroot, svpath, channels=3, factor_num=16,stride=5,epoch='best', eval_mapping=True): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if channels == 3: + cls_net = mnist_net.ConvNet().cuda() + elif channels == 1: + cls_net = mnist_net.ConvNet(imdim=channels).cuda() + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + cls_net.load_state_dict(saved_weight) + cls_net.eval() + + # 测试 + str2fun = { + 'mnist': data_loader.load_mnist, + 'mnist_m': data_loader.load_mnist_m, + 'usps': data_loader.load_usps, + 'svhn': data_loader.load_svhn, + 'syndigit': data_loader.load_syndigit, + } + + columns = ['mnist', 'svhn', 'mnist_m', 'syndigit','usps'] + target = ['svhn', 'mnist_m', 'syndigit','usps'] + + index = ['w/o do (original x)'] + data_result = {} + + for idx, data in enumerate(columns): + teset = str2fun[data]('test', channels=channels) + teloader = DataLoader(teset, batch_size=8, num_workers=0) + # 计算评价指标 + teacc = evaluate(cls_net, teloader) + if data == 'mnist': + acc_avg = np.zeros(teacc.shape) + else: + acc_avg = acc_avg + teacc + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + + data_result['Avg'] = acc_avg + + df = pd.DataFrame(data_result,index = index) + print(df) + if svpath is not None: + df.to_csv(svpath) + +if __name__=='__main__': + main() + diff --git a/Meta-causal/code-stage1-pipeline/main_test_pacs_v13.py b/Meta-causal/code-stage1-pipeline/main_test_pacs_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..9a8b05675e6319cea1651da2048fbf4720a4dae7 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/main_test_pacs_v13.py @@ -0,0 +1,89 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import resnet as resnet +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--source_domain', type=str, default='art_painting', help='source domain') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--factor_num', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--stride', type=int, default=5) +@click.option('--eval_mapping', type=bool, default=False, help='是否查看mapping学习效果') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') +def main(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network): + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + +def evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num=16, epoch='best', stride=5,eval_mapping=False, network='resnet18'): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if network == 'resnet18': + cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + input_dim = 2048 + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + + cls_net.load_state_dict(saved_weight) + cls_net.eval() + + + columns = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in columns if i!=source_domain] + columns = [source_domain] + target + print("columns:",columns) + + + index = ['w/o do (original x)'] + + data_result = {} + data_result_ours = {} + + for idx, data in enumerate(columns): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=4, num_workers=0) + # 计算评价指标 + acc = evaluate(cls_net, teloader) + data_result_ours[data] = acc + + teacc = evaluate(cls_net, teloader) + if data == source_domain: + acc_avg = np.zeros(teacc.shape) + else: + acc_avg = acc_avg + teacc + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + + data_result['Avg'] = acc_avg + + df = pd.DataFrame(data_result,index = index) + print(df) + + if svpath is not None: + df.to_csv(svpath) + +if __name__=='__main__': + main() + + + diff --git a/Meta-causal/code-stage1-pipeline/network/adaptor_v2.py b/Meta-causal/code-stage1-pipeline/network/adaptor_v2.py new file mode 100644 index 0000000000000000000000000000000000000000..ce47dbd1a24f9e2f741d8a82061b62b86d3dba41 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/network/adaptor_v2.py @@ -0,0 +1,63 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +import numpy as np + +class mapping(nn.Module): + def __init__(self, input_dim=1024, hidden_dim = 512, out_dim=1024, layernum=4): + ''' + ''' + super().__init__() + self.layernum = layernum + if layernum == 4: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim) + self.fc3 = nn.Linear(hidden_dim, hidden_dim) + self.fc4 = nn.Linear(hidden_dim, out_dim) + elif layernum == 2: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, out_dim) + self.relu = nn.ReLU(inplace=True) + + def forward(self, x): + ''' x ''' + if self.layernum == 4: + x = self.relu(self.fc1(x)) + x = self.relu(self.fc2(x)) + x = self.relu(self.fc3(x)) + x = self.fc4(x) + elif self.layernum == 2: + x = self.relu(self.fc1(x)) + x = self.fc2(x) + return x + + +class effect_to_weight(nn.Module): + def __init__(self, input_dim = 512, hidden_dim = 256, out_dim = 1, layernum=2, hidden_dim2 = 128): + ''' + ''' + super().__init__() + + self.layernum = layernum + if layernum == 2: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, out_dim) + elif layernum == 3: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim2) + self.fc3 = nn.Linear(hidden_dim2, out_dim) + self.relu = nn.ReLU(inplace=True) + + def forward(self, x): + ''' x ''' + if self.layernum == 2: + x = self.relu(self.fc1(x)) + x = self.fc2(x) + else: + x = self.relu(self.fc1(x)) + x = self.relu(self.fc2(x)) + x = self.fc3(x) + return x + + diff --git a/Meta-causal/code-stage1-pipeline/network/mnist_net_my.py b/Meta-causal/code-stage1-pipeline/network/mnist_net_my.py new file mode 100644 index 0000000000000000000000000000000000000000..15e2e677280fdd2211b559f9f1bafd2fb66b5ef4 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/network/mnist_net_my.py @@ -0,0 +1,104 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F + +class ConvNet(nn.Module): + ''' 网络结构和cvpr2020的 M-ADA 方法一致 ''' + def __init__(self, imdim=3): + super(ConvNet, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 1024) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(1024, 10) + # self.cls_head_tgt = nn.Linear(1024, 10) + # self.pro_head = nn.Linear(1024, 128) + + def forward(self, x, mode='fc'): + if mode == 'c': + out4 = self.relu4(x) + p = self.cls_head_src(out4) + return p + elif mode == 'fc': + in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + out2 = self.mp(self.relu2(self.conv2(out1))) + out2 = out2.view(in_size, -1) + out3 = self.relu3(self.fc1(out2)) + out4_worelu = self.fc2(out3) + out4 = self.relu4(out4_worelu) + p = self.cls_head_src(out4) + return p, out4_worelu + + # if mode == 'test': + # p = self.cls_head_src(out4) + # return p + # elif mode == 'train': + # p = self.cls_head_src(out4) + # # z = self.pro_head(out4) + # # z = F.normalize(z) + # return p,out4_worelu + # elif mode == 'p_f': + # p = self.cls_head_src(out4) + # return p, out4 + #elif mode == 'target': + # p = self.cls_head_tgt(out4) + # z = self.pro_head(out4) + # z = F.normalize(z) + # return p,z + +class ConvNetVis(nn.Module): + ''' 方便可视化，特征提取器输出2-d特征 + ''' + def __init__(self, imdim=3): + super(ConvNetVis, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 2) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(2, 10) + self.cls_head_tgt = nn.Linear(2, 10) + self.pro_head = nn.Linear(2, 128) + + def forward(self, x, mode='test'): + + in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + out2 = self.mp(self.relu2(self.conv2(out1))) + out2 = out2.view(in_size, -1) + out3 = self.relu3(self.fc1(out2)) + out4 = self.relu4(self.fc2(out3)) + + if mode == 'test': + p = self.cls_head_src(out4) + return p + elif mode == 'train': + p = self.cls_head_src(out4) + z = self.pro_head(out4) + z = F.normalize(z) + return p,z + elif mode == 'p_f': + p = self.cls_head_src(out4) + return p, out4 + #elif mode == 'target': + # p = self.cls_head_tgt(out4) + # z = self.pro_head(out4) + # z = F.normalize(z) + # return p,z + + diff --git a/Meta-causal/code-stage1-pipeline/network/resnet.py b/Meta-causal/code-stage1-pipeline/network/resnet.py new file mode 100644 index 0000000000000000000000000000000000000000..925410b6cc064aba01d1f86efa8eb7fdd592ecee --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/network/resnet.py @@ -0,0 +1,101 @@ +from torch import nn +from torch.utils import model_zoo +#from torchvision.models.resnet import BasicBlock, model_urls, Bottleneck +from torchvision.models.resnet import BasicBlock, Bottleneck + +import torch +import ssl +# from torch import nn as nn +# from utils.util import * + +ssl._create_default_https_context = ssl._create_unverified_context + +all = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101','resnet152'] + +model_urls = { +'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', +'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', +'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', +'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', +'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth', +} + + +class ResNet(nn.Module): + def __init__(self, block, layers,classes=7,c_dim=512): + self.inplanes = 64 + super(ResNet, self).__init__() + self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, + bias=False) + self.bn1 = nn.BatchNorm2d(64) + self.relu = nn.ReLU(inplace=True) + self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + self.layer1 = self._make_layer(block, 64, layers[0]) + self.layer2 = self._make_layer(block, 128, layers[1], stride=2) + self.layer3 = self._make_layer(block, 256, layers[2], stride=2) + self.layer4 = self._make_layer(block, 512, layers[3], stride=2) + self.avgpool = nn.AvgPool2d(7, stride=1) + self.class_classifier = nn.Linear(c_dim, classes) + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + nn.init.constant_(m.weight, 1) + nn.init.constant_(m.bias, 0) + + def _make_layer(self, block, planes, blocks, stride=1): + downsample = None + if stride != 1 or self.inplanes != planes * block.expansion: + downsample = nn.Sequential( + nn.Conv2d(self.inplanes, planes * block.expansion, + kernel_size=1, stride=stride, bias=False), + nn.BatchNorm2d(planes * block.expansion), + ) + + layers = [] + layers.append(block(self.inplanes, planes, stride, downsample)) + self.inplanes = planes * block.expansion + for i in range(1, blocks): + layers.append(block(self.inplanes, planes)) + + return nn.Sequential(*layers) + def forward(self, x, mode='fc'): + if mode == 'c': + return self.class_classifier(x) + else: + x = self.conv1(x) + x = self.bn1(x) + x = self.relu(x) + x = self.maxpool(x) + + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + x = self.avgpool(x) + x = x.view(x.size(0), -1) + # print("x.shape:",x.shape) + return self.class_classifier(x), x + + +def resnet18(pretrained=True, **kwargs): + """Constructs a ResNet-18 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet18']), strict=False) + return model + +def resnet50(pretrained=True, **kwargs): + """Constructs a ResNet-50 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet50']), strict=False) + return model diff --git a/Meta-causal/code-stage1-pipeline/network/wideresnet.py b/Meta-causal/code-stage1-pipeline/network/wideresnet.py new file mode 100644 index 0000000000000000000000000000000000000000..1ca130a5f278c3b63f43b589db6ebd18d6e91593 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/network/wideresnet.py @@ -0,0 +1,86 @@ +import math +import torch +import torch.nn as nn +import torch.nn.functional as F + + +class BasicBlock(nn.Module): + def __init__(self, in_planes, out_planes, stride, dropRate=0.0): + super(BasicBlock, self).__init__() + self.bn1 = nn.BatchNorm2d(in_planes) + self.relu1 = nn.ReLU(inplace=True) + self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, + padding=1, bias=False) + self.bn2 = nn.BatchNorm2d(out_planes) + self.relu2 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=3, stride=1, + padding=1, bias=False) + self.droprate = dropRate + self.equalInOut = (in_planes == out_planes) + self.convShortcut = (not self.equalInOut) and nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, + padding=0, bias=False) or None + def forward(self, x): + if not self.equalInOut: + x = self.relu1(self.bn1(x)) + else: + out = self.relu1(self.bn1(x)) + out = self.relu2(self.bn2(self.conv1(out if self.equalInOut else x))) + if self.droprate > 0: + out = F.dropout(out, p=self.droprate, training=self.training) + out = self.conv2(out) + return torch.add(x if self.equalInOut else self.convShortcut(x), out) + +class NetworkBlock(nn.Module): + def __init__(self, nb_layers, in_planes, out_planes, block, stride, dropRate=0.0): + super(NetworkBlock, self).__init__() + self.layer = self._make_layer(block, in_planes, out_planes, nb_layers, stride, dropRate) + def _make_layer(self, block, in_planes, out_planes, nb_layers, stride, dropRate): + layers = [] + for i in range(int(nb_layers)): + layers.append(block(i == 0 and in_planes or out_planes, out_planes, i == 0 and stride or 1, dropRate)) + return nn.Sequential(*layers) + def forward(self, x): + return self.layer(x) + +class WideResNet(nn.Module): + def __init__(self, depth, num_classes, widen_factor=1, dropRate=0.0): + super(WideResNet, self).__init__() + nChannels = [16, 16*widen_factor, 32*widen_factor, 64*widen_factor] + assert((depth - 4) % 6 == 0) + n = (depth - 4) / 6 + block = BasicBlock + # 1st conv before any network block + self.conv1 = nn.Conv2d(3, nChannels[0], kernel_size=3, stride=1, + padding=1, bias=False) + # 1st block + self.block1 = NetworkBlock(n, nChannels[0], nChannels[1], block, 1, dropRate) + # 2nd block + self.block2 = NetworkBlock(n, nChannels[1], nChannels[2], block, 2, dropRate) + # 3rd block + self.block3 = NetworkBlock(n, nChannels[2], nChannels[3], block, 2, dropRate) + # global average pooling and classifier + self.bn1 = nn.BatchNorm2d(nChannels[3]) + self.relu = nn.ReLU(inplace=True) + self.fc = nn.Linear(nChannels[3], num_classes) + self.nChannels = nChannels[3] + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + m.weight.data.fill_(1) + m.bias.data.zero_() + elif isinstance(m, nn.Linear): + m.bias.data.zero_() + def forward(self, x, mode='fc'): + if mode == 'c': + return self.fc(x) + else: + out = self.conv1(x) + out = self.block1(out) + out = self.block2(out) + out = self.block3(out) + out = self.relu(self.bn1(out)) + out = F.avg_pool2d(out, 8) + out = out.view(-1, self.nChannels) + return self.fc(out), out diff --git a/Meta-causal/code-stage1-pipeline/run_PACS/run_my_joint_v13_test.sh b/Meta-causal/code-stage1-pipeline/run_PACS/run_my_joint_v13_test.sh new file mode 100644 index 0000000000000000000000000000000000000000..aa31ce0c1616cc80133bf211bfe621836131b1c4 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/run_PACS/run_my_joint_v13_test.sh @@ -0,0 +1,35 @@ + +# $1 gpuid +# $2 runid + +# base方法 +cd .. +epochs=30 +clsadapt=True +lr=0.01 +factor_num=16 +lr_scheduler=cosine +lambda_causal=1 +lambda_re=1 +batchsize=6 +stride=5 +randm=True +randn=True +autoaug=CA_multiple +network=resnet18 +UniqueExpName=pipelineAugWoNorm + +root=/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS/ +data=art_painting +svroot=$root/${data}/${autoaug}_${factor_num}fa_v2_ep${epochs}_lr${lr}_${lr_scheduler}_base0.01_bs${batchsize}_lamCa_${lambda_causal}_lamRe${lambda_re}_adt4_cls1_EW2_70_rm${randm}_rn${randn}_str${stride}_${UniqueExpName} +#python3 main_my_joint_v13_auto.py --gpu $1 --data ${data} --epochs $epochs --autoaug $autoaug --lambda_causal ${lambda_causal} --lambda_re ${lambda_re} --lr $lr --svroot $svroot --clsadapt $clsadapt --factor_num $factor_num --lr_scheduler ${lr_scheduler} --batchsize ${batchsize} --network ${network} --randm ${randm} --randn ${randn} --stride ${stride} + +test_epoch=best +#python3 main_test_pacs_v13.py --gpu $1 --source_domain $data --svroot $svroot --svpath $svroot/${data}_${factor_num}factor_${test_epoch}_test_check.csv --factor_num $factor_num --epoch $test_epoch --network ${network} --stride ${stride} + +python3 AllEpochs_test_pacs_v13.py --gpu $1 --source_domain $data --svroot $svroot --svpath $svroot/${data}_${factor_num}factor_${test_epoch}_test_check.csv --factor_num $factor_num --epoch $test_epoch --network ${network} --stride ${stride} + + + + + diff --git a/Meta-causal/code-stage1-pipeline/run_digits/run_my_joint_test.sh b/Meta-causal/code-stage1-pipeline/run_digits/run_my_joint_test.sh new file mode 100644 index 0000000000000000000000000000000000000000..7cb85797bdb805f9fee2432529ce99fc8303a8c4 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/run_digits/run_my_joint_test.sh @@ -0,0 +1,34 @@ + +# $1 gpuid + +cd .. +epochs=100 +clsadapt=True +lr=1e-4 +lr_scheduler=Step +factor_num=14 +test_epoch=best +lambda_causal=1 +lambda_re=1 +batchsize=32 +stride=3 +randm=True +randn=True +autoaug=CA_multiple +UniqueExpName='pipelineAugWoNorm' + + +root=/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit +svroot=$root/${autoaug}_${factor_num}fa_all_ep${epochs}_lr${lr}_lr_scheduler${lr_scheduler}0.8_bs${batchsize}_lamCa_${lambda_causal}_lamRe_${lambda_re}_cls1_adt2_EW2_100_rm${randm}_rn${randn}_str${stride}_${UniqueExpName} + +#python3 main_my_joint_v13_auto.py --gpu $1 --data mnist --epochs $epochs --autoaug $autoaug --lambda_causal ${lambda_causal} --lambda_re ${lambda_re} --lr $lr --lr_scheduler $lr_scheduler --svroot $svroot --clsadapt $clsadapt --factor_num $factor_num --batchsize ${batchsize} --randm ${randm} --randn ${randn} --stride ${stride} + +#python3 main_test_digit_v13.py --gpu $1 --svroot $svroot --svpath $svroot/${factor_num}factor_${test_epoch}.csv --factor_num $factor_num --epoch $test_epoch --stride ${stride} + +python3 AllEpochs_test_digit_v13.py --gpu $1 --svroot $svroot --svpath $svroot/${factor_num}factor_${test_epoch}.csv --factor_num $factor_num --epoch $test_epoch --stride ${stride} + + + + + + diff --git a/Meta-causal/code-stage1-pipeline/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala b/Meta-causal/code-stage1-pipeline/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala new file mode 100644 index 0000000000000000000000000000000000000000..aa44ae0c513b57a8501e9bb1af27dc442b72f7d7 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a44a49f0a1b3c59b9763c67ea85708ef8b56cae5fe4336f0383f5f71ba0dac84 +size 40 diff --git a/Meta-causal/code-stage1-pipeline/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log b/Meta-causal/code-stage1-pipeline/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log new file mode 100644 index 0000000000000000000000000000000000000000..f26feaaef352ae5821e49b7fbc4b1720f8735f38 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log @@ -0,0 +1 @@ +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': 'saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} diff --git a/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala new file mode 100644 index 0000000000000000000000000000000000000000..5ec21b3afdf0e11651cc768f4f55ea6269b887f5 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7946f93077ec2136f75fc090a5762ce810be71cc78d5201e8a671217a678c563 +size 40 diff --git a/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala new file mode 100644 index 0000000000000000000000000000000000000000..620f9ba109e77ed90b7676c138933f814245e7f1 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2021c61739fbe1f9c066067b4e5903d8d2d6c1c44865e1e9c61449eb3d90327 +size 40 diff --git a/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala new file mode 100644 index 0000000000000000000000000000000000000000..3144b1448112cff1aa0c26e0d825b50698f41d65 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fafb4b17d350157735eb6174ff44bafcea7ab8bf86948df3421447ef45ffcae3 +size 40 diff --git a/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log new file mode 100644 index 0000000000000000000000000000000000000000..f4c211545f0d2b537d3dcf980579f604a33419a7 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log @@ -0,0 +1 @@ +{'gpu': '0çç', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': 'saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} diff --git a/Meta-causal/code-stage1-pipeline/submit_v1-2.sh b/Meta-causal/code-stage1-pipeline/submit_v1-2.sh new file mode 100644 index 0000000000000000000000000000000000000000..bda242b4573754859457a05d8e0f778917c95a13 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/submit_v1-2.sh @@ -0,0 +1,22 @@ +#!/bin/bash +#SBATCH --job-name=metatrainRN +#SBATCH --nodes=1 # Request 1 node +#SBATCH --ntasks=1 # Number of tasks (total) +#SBATCH --cpus-per-task=8 # Number of CPU cores (threads) per task +#SBATCH --mem-per-cpu=4G # Memory limit per CPU core (there is no --mem-per-task) +#SBATCH --time=12:00:00 # Job timeout +#SBATCH --gpus-per-node=l4-24g:1 +#SBATCH --nodelist=gcpl4-eu-1 +#SBATCH --output=%j.log # Redirect stdout to a log file +#SBATCH --error=%j.error # Redirect stderr to a separate error log file + +srun --nodes 1 --ntasks-per-node 1 -- \ +mkenv -f ../env_mc.yml -- \ +sh -c "cd run_digits +bash run_my_joint_test.sh 0 +" + + + + + diff --git a/Meta-causal/code-stage1-pipeline/submit_v1.sh b/Meta-causal/code-stage1-pipeline/submit_v1.sh new file mode 100644 index 0000000000000000000000000000000000000000..6f434f0b5acdd336596daaba7fafcce8464dc41a --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/submit_v1.sh @@ -0,0 +1,22 @@ +#!/bin/bash +#SBATCH --job-name=metatrainRN +#SBATCH --nodes=1 # Request 1 node +#SBATCH --ntasks=1 # Number of tasks (total) +#SBATCH --cpus-per-task=8 # Number of CPU cores (threads) per task +#SBATCH --mem-per-cpu=4G # Memory limit per CPU core (there is no --mem-per-task) +#SBATCH --time=12:00:00 # Job timeout +#SBATCH --gpus-per-node=l4-24g:1 +#SBATCH --nodelist=gcpl4-eu-1 +#SBATCH --output=%j.log # Redirect stdout to a log file +#SBATCH --error=%j.error # Redirect stderr to a separate error log file + +srun --nodes 1 --ntasks-per-node 1 -- \ +mkenv -f ../env_mc.yml -- \ +sh -c "cd run_PACS +bash run_my_joint_v13_test.sh 0 +" + + + + + diff --git a/Meta-causal/code-stage1-pipeline/tools/autoaugment.py b/Meta-causal/code-stage1-pipeline/tools/autoaugment.py new file mode 100644 index 0000000000000000000000000000000000000000..76c6bc4ebd5c59b76a58a8dca196f22d41fbf114 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/tools/autoaugment.py @@ -0,0 +1,234 @@ +from PIL import Image, ImageEnhance, ImageOps +import numpy as np +import random + + +class ImageNetPolicy(object): + """ Randomly choose one of the best 24 Sub-policies on ImageNet. + + Example: + >>> policy = ImageNetPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> ImageNetPolicy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.4, "posterize", 8, 0.6, "rotate", 9, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "posterize", 7, 0.6, "posterize", 6, fillcolor), + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + + SubPolicy(0.4, "equalize", 4, 0.8, "rotate", 8, fillcolor), + SubPolicy(0.6, "solarize", 3, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.8, "posterize", 5, 1.0, "equalize", 2, fillcolor), + SubPolicy(0.2, "rotate", 3, 0.6, "solarize", 8, fillcolor), + SubPolicy(0.6, "equalize", 8, 0.4, "posterize", 6, fillcolor), + + SubPolicy(0.8, "rotate", 8, 0.4, "color", 0, fillcolor), + SubPolicy(0.4, "rotate", 9, 0.6, "equalize", 2, fillcolor), + SubPolicy(0.0, "equalize", 7, 0.8, "equalize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + + SubPolicy(0.8, "rotate", 8, 1.0, "color", 2, fillcolor), + SubPolicy(0.8, "color", 8, 0.8, "solarize", 7, fillcolor), + SubPolicy(0.4, "sharpness", 7, 0.6, "invert", 8, fillcolor), + SubPolicy(0.6, "shearX", 5, 1.0, "equalize", 9, fillcolor), + SubPolicy(0.4, "color", 0, 0.6, "equalize", 3, fillcolor), + + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment ImageNet Policy" + + +class CIFAR10Policy(object): + """ Randomly choose one of the best 25 Sub-policies on CIFAR10. + + Example: + >>> policy = CIFAR10Policy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> CIFAR10Policy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.1, "invert", 7, 0.2, "contrast", 6, fillcolor), + SubPolicy(0.7, "rotate", 2, 0.3, "translateX", 9, fillcolor), + SubPolicy(0.8, "sharpness", 1, 0.9, "sharpness", 3, fillcolor), + SubPolicy(0.5, "shearY", 8, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.5, "autocontrast", 8, 0.9, "equalize", 2, fillcolor), + + SubPolicy(0.2, "shearY", 7, 0.3, "posterize", 7, fillcolor), + SubPolicy(0.4, "color", 3, 0.6, "brightness", 7, fillcolor), + SubPolicy(0.3, "sharpness", 9, 0.7, "brightness", 9, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.5, "equalize", 1, fillcolor), + SubPolicy(0.6, "contrast", 7, 0.6, "sharpness", 5, fillcolor), + + SubPolicy(0.7, "color", 7, 0.5, "translateX", 8, fillcolor), + SubPolicy(0.3, "equalize", 7, 0.4, "autocontrast", 8, fillcolor), + SubPolicy(0.4, "translateY", 3, 0.2, "sharpness", 6, fillcolor), + SubPolicy(0.9, "brightness", 6, 0.2, "color", 8, fillcolor), + SubPolicy(0.5, "solarize", 2, 0.0, "invert", 3, fillcolor), + + SubPolicy(0.2, "equalize", 0, 0.6, "autocontrast", 0, fillcolor), + SubPolicy(0.2, "equalize", 8, 0.6, "equalize", 4, fillcolor), + SubPolicy(0.9, "color", 9, 0.6, "equalize", 6, fillcolor), + SubPolicy(0.8, "autocontrast", 4, 0.2, "solarize", 8, fillcolor), + SubPolicy(0.1, "brightness", 3, 0.7, "color", 0, fillcolor), + + SubPolicy(0.4, "solarize", 5, 0.9, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "translateY", 9, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.9, "autocontrast", 2, 0.8, "solarize", 3, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.1, "invert", 3, fillcolor), + SubPolicy(0.7, "translateY", 9, 0.9, "autocontrast", 1, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment CIFAR10 Policy" + + +class SVHNPolicy(object): + """ Randomly choose one of the best 25 Sub-policies on SVHN. + + Example: + >>> policy = SVHNPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> SVHNPolicy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.9, "shearX", 4, 0.2, "invert", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.7, "invert", 5, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.6, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 3, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "equalize", 1, 0.9, "rotate", 3, fillcolor), + + SubPolicy(0.9, "shearX", 4, 0.8, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.4, "invert", 5, fillcolor), + SubPolicy(0.9, "shearY", 5, 0.2, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 6, 0.8, "autocontrast", 1, fillcolor), + SubPolicy(0.6, "equalize", 3, 0.9, "rotate", 3, fillcolor), + + SubPolicy(0.9, "shearX", 4, 0.3, "solarize", 3, fillcolor), + SubPolicy(0.8, "shearY", 8, 0.7, "invert", 4, fillcolor), + SubPolicy(0.9, "equalize", 5, 0.6, "translateY", 6, fillcolor), + SubPolicy(0.9, "invert", 4, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.3, "contrast", 3, 0.8, "rotate", 4, fillcolor), + + SubPolicy(0.8, "invert", 5, 0.0, "translateY", 2, fillcolor), + SubPolicy(0.7, "shearY", 6, 0.4, "solarize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 0.8, "rotate", 4, fillcolor), + SubPolicy(0.3, "shearY", 7, 0.9, "translateX", 3, fillcolor), + SubPolicy(0.1, "shearX", 6, 0.6, "invert", 5, fillcolor), + + SubPolicy(0.7, "solarize", 2, 0.6, "translateY", 7, fillcolor), + SubPolicy(0.8, "shearY", 4, 0.8, "invert", 8, fillcolor), + SubPolicy(0.7, "shearX", 9, 0.8, "translateY", 3, fillcolor), + SubPolicy(0.8, "shearY", 5, 0.7, "autocontrast", 3, fillcolor), + SubPolicy(0.7, "shearX", 2, 0.1, "invert", 5, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment SVHN Policy" + + +class SubPolicy(object): + def __init__(self, p1, operation1, magnitude_idx1, p2, operation2, magnitude_idx2, fillcolor=(128, 128, 128)): + ranges = { + "shearX": np.linspace(0, 0.3, 10), + "shearY": np.linspace(0, 0.3, 10), + "translateX": np.linspace(0, 150 / 331, 10), + "translateY": np.linspace(0, 150 / 331, 10), + "rotate": np.linspace(0, 30, 10), + "color": np.linspace(0.0, 0.9, 10), + "posterize": np.round(np.linspace(8, 4, 10), 0).astype(np.int), + "solarize": np.linspace(256, 0, 10), + "contrast": np.linspace(0.0, 0.9, 10), + "sharpness": np.linspace(0.0, 0.9, 10), + "brightness": np.linspace(0.0, 0.9, 10), + "autocontrast": [0] * 10, + "equalize": [0] * 10, + "invert": [0] * 10 + } + + # from https://stackoverflow.com/questions/5252170/specify-image-filling-color-when-rotating-in-python-with-pil-and-setting-expand + def rotate_with_fill(img, magnitude): + rot = img.convert("RGBA").rotate(magnitude) + return Image.composite(rot, Image.new("RGBA", rot.size, (128,) * 4), rot).convert(img.mode) + + func = { + "shearX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, magnitude * random.choice([-1, 1]), 0, 0, 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "shearY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, magnitude * random.choice([-1, 1]), 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "translateX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, magnitude * img.size[0] * random.choice([-1, 1]), 0, 1, 0), + fillcolor=fillcolor), + "translateY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, 0, 1, magnitude * img.size[1] * random.choice([-1, 1])), + fillcolor=fillcolor), + "rotate": lambda img, magnitude: rotate_with_fill(img, magnitude), + "color": lambda img, magnitude: ImageEnhance.Color(img).enhance(1 + magnitude * random.choice([-1, 1])), + "posterize": lambda img, magnitude: ImageOps.posterize(img, magnitude), + "solarize": lambda img, magnitude: ImageOps.solarize(img, magnitude), + "contrast": lambda img, magnitude: ImageEnhance.Contrast(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "sharpness": lambda img, magnitude: ImageEnhance.Sharpness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "brightness": lambda img, magnitude: ImageEnhance.Brightness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "autocontrast": lambda img, magnitude: ImageOps.autocontrast(img), + "equalize": lambda img, magnitude: ImageOps.equalize(img), + "invert": lambda img, magnitude: ImageOps.invert(img) + } + + self.p1 = p1 + self.operation1 = func[operation1] + self.magnitude1 = ranges[operation1][magnitude_idx1] + self.p2 = p2 + self.operation2 = func[operation2] + self.magnitude2 = ranges[operation2][magnitude_idx2] + + + def __call__(self, img): + if random.random() < self.p1: img = self.operation1(img, self.magnitude1) + if random.random() < self.p2: img = self.operation2(img, self.magnitude2) + return img \ No newline at end of file diff --git a/Meta-causal/code-stage1-pipeline/tools/causalaugment_v3.py b/Meta-causal/code-stage1-pipeline/tools/causalaugment_v3.py new file mode 100644 index 0000000000000000000000000000000000000000..a375b7ebe5a83c3dba5b88f48f23a4326dec77e1 --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/tools/causalaugment_v3.py @@ -0,0 +1,694 @@ +# code in this file is adpated from rpmcruz/autoaugment +# https://github.com/rpmcruz/autoaugment/blob/master/transformations.py +import random + +import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw +import numpy as np +import torch +from PIL import Image,ImageStat +#import cv2 +from torchvision import transforms + +# def tensor2img(tensor): +# transform = transforms.Compose() + +def ShearX(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + +def DoShearX(img, v): # [-0.3, 0.3] + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + +def ShearY(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + +def DoShearY(img, v): # [-0.3, 0.3] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + +def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[0] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + +def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) +def DoTranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + +def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[1] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) +def DoTranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + +def Rotate(img, v): # [-30, 30] + assert -30 <= v <= 30 + if random.random() > 0.5: + v = -v + return img.rotate(v) +def DoRotate(img, v): # [-30, 30] + return img.rotate(v) + + +def AutoContrast(img, v): + return PIL.ImageOps.autocontrast(img, v) +def DoAutoContrast(img, v): + return PIL.ImageOps.autocontrast(img, v) + +def Invert(img, _): + return PIL.ImageOps.invert(img) +def DoInvert(img, _): + return PIL.ImageOps.invert(img) + + +def Equalize(img, _): + return PIL.ImageOps.equalize(img) +def DoEqualize(img, _): + return PIL.ImageOps.equalize(img) + +def Flip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + +def DoFlip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + + +def Solarize(img, v): # [0, 256] + assert 0 <= v <= 256 + return PIL.ImageOps.solarize(img, v) +def DoSolarize(img, v): # [0, 256] + return PIL.ImageOps.solarize(img, v) + +def SolarizeAdd(img, addition=0, threshold=128): + #img_np = np.array(img).astype(np.int) + img_np = np.array(img).astype(np.int32) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) +def DoSolarizeAdd(img, addition=0, threshold=128): + #img_np = np.array(img).astype(np.int) + img_np = np.array(img).astype(np.int32) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) + +def Posterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) +def DoPosterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) + + +def Contrast(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Contrast(img).enhance(v) + +def DoContrast(img, v): + return PIL.ImageEnhance.Contrast(img).enhance(v) + +def Color(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Color(img).enhance(v) + +def DoColor(img, v): + stat =ImageStat.Stat(img) + return PIL.ImageEnhance.Color(img).enhance(v) + + +def Brightness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Brightness(img).enhance(v) + +def DoBrightness(img, v): # obtain the brightness of image + return PIL.ImageEnhance.Brightness(img).enhance(v) + + +def Sharpness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Sharpness(img).enhance(v) + +def DoSharpness(img, v): + return PIL.ImageEnhance.Sharpness(img).enhance(v) + +def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] + assert 0.0 <= v <= 0.2 + if v <= 0.: + return img + + v = v * img.size[0] + return CutoutAbs(img, v) + + +def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img +def DoCutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img + + +def SamplePairing(imgs): # [0, 0.4] + def f(img1, v): + i = np.random.choice(len(imgs)) + img2 = PIL.Image.fromarray(imgs[i]) + return PIL.Image.blend(img1, img2, v) + + return f + + +def Identity(img, v): + return img + +def NoiseSalt(img, noise_rate): + """增加椒盐噪声 + args: + noise_rate (float): noise rate + """ + img_ = np.array(img).copy() + h, w, c = img_.shape + signal_pct = 1 - noise_rate + mask = np.random.choice((0, 1, 2), size=(h, w, 1), p=[signal_pct, noise_rate/2., noise_rate/2.]) + mask = np.repeat(mask, c, axis=2) + img_[mask == 1] = 255 # 盐噪声 + img_[mask == 2] = 0 # 椒噪声 + return Image.fromarray(img_.astype('uint8')) + +def DoNoiseSalt(img, noise_rate): + """增加椒盐噪声 + args: + noise_rate (float): noise rate + """ + img_ = np.array(img).copy() + h, w, c = img_.shape + signal_pct = 1 - noise_rate + mask = np.random.choice((0, 1, 2), size=(h, w, 1), p=[signal_pct, noise_rate/2., noise_rate/2.]) + mask = np.repeat(mask, c, axis=2) + img_[mask == 1] = 255 # 盐噪声 + img_[mask == 2] = 0 # 椒噪声 + return Image.fromarray(img_.astype('uint8')) +def NoiseGaussian(img, sigma): + """增加高斯噪声 + 传入: + img : 原图 + mean : 均值默认0 + sigma : 标准差 + 返回: + gaussian_out : 噪声处理后的图片 + """ + # 将图片灰度标准化 + img_ = np.array(img).copy() + img_ = img_ / 255.0 + # 产生高斯 noise + noise = np.random.normal(0, sigma, img_.shape) + # 将噪声和图片叠加 + gaussian_out = img_ + noise + # 将超过 1 的置 1，低于 0 的置 0 + gaussian_out = np.clip(gaussian_out, 0, 1) + # 将图片灰度范围的恢复为 0-255 + gaussian_out = np.uint8(gaussian_out*255) + # 将噪声范围搞为 0-255 + # noise = np.uint8(noise*255) + return Image.fromarray(gaussian_out) + +def DoNoiseGaussian(img, sigma): + """增加高斯噪声 + 传入: + img : 原图 + mean : 均值默认0 + sigma : 标准差 + 返回: + gaussian_out : 噪声处理后的图片 + """ + # 将图片灰度标准化 + img_ = np.array(img).copy() + img_ = img_ / 255.0 + # 产生高斯 noise + noise = np.random.normal(0, sigma, img_.shape) + # 将噪声和图片叠加 + gaussian_out = img_ + noise + # 将超过 1 的置 1，低于 0 的置 0 + gaussian_out = np.clip(gaussian_out, 0, 1) + # 将图片灰度范围的恢复为 0-255 + gaussian_out = np.uint8(gaussian_out*255) + # 将噪声范围搞为 0-255 + # noise = np.uint8(noise*255) + return Image.fromarray(gaussian_out) + +# def factor_list(factor_num): +# l = [ +# 'AutoContrast', +# 'Invert', +# 'Equalize', +# 'Solarize', +# 'SolarizeAdd', +# 'Posterize', +# 'Contrast', +# 'Color', +# 'Brightness', +# 'Sharpness', +# 'NoiseSalt', +# 'NoiseGaussian', +# ] +# return l[:factor_num] + +# def causal_list(factor_num): # 16 oeprations and their ranges +# l = [ +# (AutoContrast, 0, 100), +# (Invert, 0, 1), +# (Equalize, 0, 1), +# (Solarize, 0, 256), +# (SolarizeAdd, 0, 110), +# (Posterize, 0, 4), +# (Contrast, 0.1, 1.9), +# (Color, 0.1, 1.9), +# (Brightness, 0.1, 1.9), +# (Sharpness, 0.1, 1.9), +# (NoiseSalt,0.0,0.1), +# (NoiseGaussian,0.0,0.1), +# ] + +# return l[:factor_num] + + +# def factor_list(factor_num): +# l = [ +# 'ShearX', +# 'ShearY', +# 'Rotate', +# 'Flip' +# ] +# return l[:factor_num] + +# def causal_list(factor_num): # 16 oeprations and their ranges +# l = [ +# (ShearX, 0., 0.3), +# (ShearY, 0., 0.3), +# (Rotate, 0, 30), +# (Flip, 0, 1), +# ] + +# return l[:factor_num] + +def factor_list(factor_num): + l = [ + 'ShearX', + 'ShearY', + 'AutoContrast', + 'Invert', + 'Equalize', + 'Solarize', + 'SolarizeAdd', + 'Posterize', + 'Contrast', + 'Color', + 'Brightness', + 'Sharpness', + 'NoiseSalt', + 'NoiseGaussian', + 'Rotate', + 'Flip' + ] + return l[:factor_num] + +def causal_list(factor_num): # 16 oeprations and their ranges + l = [ + (ShearX, 0., 0.3), + (ShearY, 0., 0.3), + (AutoContrast, 0, 100), + (Invert, 0, 1), + (Equalize, 0, 1), + (Solarize, 0, 256), + (SolarizeAdd, 0, 110), + (Posterize, 0, 4), + (Contrast, 0.1, 1.9), + (Color, 0.1, 1.9), + (Brightness, 0.1, 1.9), + (Sharpness, 0.1, 1.9), + (NoiseSalt,0.0,0.1), + (NoiseGaussian,0.0,0.1), + (Rotate, 0, 30), + (Flip, 0, 1), + ] + + return l[:factor_num] + +class Lighting(object): + """Lighting noise(AlexNet - style PCA - based noise)""" + + def __init__(self, alphastd, eigval, eigvec): + self.alphastd = alphastd + self.eigval = torch.Tensor(eigval) + self.eigvec = torch.Tensor(eigvec) + + def __call__(self, img): + if self.alphastd == 0: + return img + + alpha = img.new().resize_(3).normal_(0, self.alphastd) + rgb = self.eigvec.type_as(img).clone() \ + .mul(alpha.view(1, 3).expand(3, 3)) \ + .mul(self.eigval.view(1, 3).expand(3, 3)) \ + .sum(1).squeeze() + + return img.add(rgb.view(3, 1, 1).expand_as(img)) + + +class CutoutDefault(object): + """ + Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py + """ + def __init__(self, length): + self.length = length + + def __call__(self, img): + h, w = img.size(1), img.size(2) + mask = np.ones((h, w), np.float32) + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + mask[y1: y2, x1: x2] = 0. + mask = torch.from_numpy(mask) + mask = mask.expand_as(img) + img *= mask + return img + + +class RandAugment_incausal: + def __init__(self, n, m, factor_num, randm=False, randn=False): + self.n = n + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + self.randn = randn + self.factor_num = factor_num + print("randm:",self.randm) + print("randn:",self.randn) + print("n:",self.n) + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + if self.randn: + self.n = random.randint(1,self.factor_num) + + ops = random.choices(self.causal_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img +class RandAugment_all: + def __init__(self, m, factor_num, randm=False): + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + self.factor_num = factor_num + + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + factor_choice = np.random.randint(0,2,self.factor_num) + # ops = random.choices(self.causal_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + if factor_choice[index] == 0: + continue + else: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img +class RandAugment_incausal_label: + def __init__(self, n, m, factor_num, randm=False): + self.n = n + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + self.factor_num = factor_num + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + print("randm:",self.randm) + + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + #op_labels = np.random.randint(0,self.factor_num-1,self.n) + op_labels = random.sample(range(0, self.factor_num), self.n) + ops = [li for index, li in enumerate(self.causal_list) if index in op_labels] + #ops = random.choices(self.causal_list, k=self.n) + # print(self.causal_list) + # print("op_labels:",op_labels) + # print("select_op:",ops) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img, np.array(op_labels) +class FactualAugment_incausal: + def __init__(self, m, factor_num, randm=False): + self.m = m + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.randm = randm + print("randm:",self.randm) + def __call__(self, img): + # ops = random.choices(self.causal_list, k=1) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + val = (float(self.m) / 30) * float(maxval - minval) + minval + if index == 0: + imgs = np.array(op(img, val)) + else: + imgs = np.concatenate((imgs, op(img, val)),-1) + # print("imgs",imgs.shape) + return imgs +class CounterfactualAugment_incausal: + def __init__(self,factor_num): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + def __call__(self, img): + # index = 0 + # b, c, h, w = img.shape + # imgs = torch.zeros(b*self.factor_num, c, h, w) + # for b_ in range(32): + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + if index == 0: + imgs = np.array(op(img, maxval)) + else: + imgs = np.concatenate((imgs, op(img, maxval)),-1) + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs +class MultiCounterfactualAugment_incausal: + def __init__(self, factor_num, stride): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.stride = stride + + def __call__(self, img): + # index = 0 + # b, c, h, w = img.shape + # imgs = torch.zeros(b*self.factor_num, c, h, w) + # for b_ in range(32): + # 0,5,10,15,20,25,30 + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + for i in range(0, 31, self.stride): + val = (float(i) / 30) * float(maxval - minval) + minval + if index == 0 and i == 0: + imgs = np.array(op(img, val)) + else: + imgs = np.concatenate((imgs, op(img, val)),-1) + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs +class MultiCounterfactualAugment: + def __init__(self, factor_num, stride=5): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.stride = stride + self.var_num = len(list(range(0, 31, self.stride))) + print("stride:",stride) + def __call__(self, img): + # index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num*self.var_num, c, h, w) + # for b_ in range(32): + # 0,5,10,15,20,25,30 + # print(img.shape) + for b_ in range(b): + img0 = transforms.ToPILImage()(imgs[b_]) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + i_index = 0 + for i in range(0, 31, self.stride): + val = (float(i) / 30) * float(maxval - minval) + minval + img1 = op(img0, val) + img1 = transforms.ToTensor()(img1) + #print(f'batch {b_} factor {index} stride {i} i_index {i_index} total {b_*self.factor_num*self.var_num+index*self.var_num+i_index}') + imgs[b_*self.factor_num*self.var_num+index*self.var_num+i_index] = img1 + i_index = i_index + 1 + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs + + +class FactualAugment: + def __init__(self, m, factor_num, randm=False): + self.m = m + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.randm = randm + print("randm:",randm) + def __call__(self, img): + index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num, c, h, w) + + img = img.cpu() + for b_ in range(b): + imgs[b_*self.factor_num:(b_+1)*self.factor_num] = self.get_item(img[b_]) + return imgs + def get_item(self, img): + index = 0 + # print("input_dim:",img.shape) + c, h, w = img.shape + imgs = torch.zeros(self.factor_num, c, h, w) + # img = img.squeeze(0) + # print(img.shape) + img = transforms.ToPILImage()(img) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval(self.factor_list[index]) + val = (float(self.m) / 30) * float(maxval - minval) + minval + img1 = op(img, val) + img1 = transforms.ToTensor()(img1) + imgs[index] = img1 + return imgs +class CounterfactualAugment: + def __init__(self,factor_num): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + + def __call__(self, img): + index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num, c, h, w) + + img = img.cpu() + for b_ in range(b): + imgs[b_*self.factor_num:(b_+1)*self.factor_num] = self.get_item(img[b_]) + return imgs + def get_item(self, img): + index = 0 + c, h, w = img.shape + imgs = torch.ones(self.factor_num, c, h, w) + # img = img.squeeze(0) + img = transforms.ToPILImage()(img) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + img1 = op(img, maxval) + # img1.save('test'+str(index)+'.png') + img1 = transforms.ToTensor()(img1) + imgs[index] = img1 + return imgs + +class Avg_statistic: + def __init__(self): + self.do_list = do_list() + self.statistic_num = len(self.do_list) + self.avg_val = np.zeros(self.statistic_num) + self.img_num = 0 + + def get_item(self,img): + # ops = self.statistic_list + do_index = 0 + for op in self.do_list: + val=op(img) + self.avg_val[do_index] += val + self.img_num = self.img_num + 1 + + def compute_average(self): + self.avg_val = self.avg_val/self.img_num + + def get_infor(self): + return self.avg_val, self.img_num + + + + diff --git a/Meta-causal/code-stage1-pipeline/tools/randaugment.py b/Meta-causal/code-stage1-pipeline/tools/randaugment.py new file mode 100644 index 0000000000000000000000000000000000000000..f3bbdf11541df078144fa0ced8d693d4c98507ad --- /dev/null +++ b/Meta-causal/code-stage1-pipeline/tools/randaugment.py @@ -0,0 +1,248 @@ +# code in this file is adpated from rpmcruz/autoaugment +# https://github.com/rpmcruz/autoaugment/blob/master/transformations.py +import random + +import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw +import numpy as np +import torch +from PIL import Image + + +def ShearX(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + + +def ShearY(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + + +def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[0] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + + +def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + + +def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[1] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def Rotate(img, v): # [-30, 30] + assert -30 <= v <= 30 + if random.random() > 0.5: + v = -v + return img.rotate(v) + + +def AutoContrast(img, _): + return PIL.ImageOps.autocontrast(img) + + +def Invert(img, _): + return PIL.ImageOps.invert(img) + + +def Equalize(img, _): + return PIL.ImageOps.equalize(img) + + +def Flip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + + +def Solarize(img, v): # [0, 256] + assert 0 <= v <= 256 + return PIL.ImageOps.solarize(img, v) + + +def SolarizeAdd(img, addition=0, threshold=128): + img_np = np.array(img).astype(np.int) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) + + +def Posterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) + + +def Contrast(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Contrast(img).enhance(v) + + +def Color(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Color(img).enhance(v) + + +def Brightness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Brightness(img).enhance(v) + + +def Sharpness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Sharpness(img).enhance(v) + + +def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] + assert 0.0 <= v <= 0.2 + if v <= 0.: + return img + + v = v * img.size[0] + return CutoutAbs(img, v) + + +def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img + + +def SamplePairing(imgs): # [0, 0.4] + def f(img1, v): + i = np.random.choice(len(imgs)) + img2 = PIL.Image.fromarray(imgs[i]) + return PIL.Image.blend(img1, img2, v) + + return f + + +def Identity(img, v): + return img + + +def augment_list(): # 16 oeprations and their ranges + + # https://github.com/tensorflow/tpu/blob/8462d083dd89489a79e3200bcc8d4063bf362186/models/official/efficientnet/autoaugment.py#L505 + l = [ + (AutoContrast, 0, 1), + (Equalize, 0, 1), + (Invert, 0, 1), + (Rotate, 0, 30), + (Posterize, 0, 4), + (Solarize, 0, 256), + (SolarizeAdd, 0, 110), + (Color, 0.1, 1.9), + (Contrast, 0.1, 1.9), + (Brightness, 0.1, 1.9), + (Sharpness, 0.1, 1.9), + (ShearX, 0., 0.3), + (ShearY, 0., 0.3), + (CutoutAbs, 0, 40), + (TranslateXabs, 0., 100), + (TranslateYabs, 0., 100), + ] + + return l + + +class Lighting(object): + """Lighting noise(AlexNet - style PCA - based noise)""" + + def __init__(self, alphastd, eigval, eigvec): + self.alphastd = alphastd + self.eigval = torch.Tensor(eigval) + self.eigvec = torch.Tensor(eigvec) + + def __call__(self, img): + if self.alphastd == 0: + return img + + alpha = img.new().resize_(3).normal_(0, self.alphastd) + rgb = self.eigvec.type_as(img).clone() \ + .mul(alpha.view(1, 3).expand(3, 3)) \ + .mul(self.eigval.view(1, 3).expand(3, 3)) \ + .sum(1).squeeze() + + return img.add(rgb.view(3, 1, 1).expand_as(img)) + + +class CutoutDefault(object): + """ + Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py + """ + def __init__(self, length): + self.length = length + + def __call__(self, img): + h, w = img.size(1), img.size(2) + mask = np.ones((h, w), np.float32) + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + mask[y1: y2, x1: x2] = 0. + mask = torch.from_numpy(mask) + mask = mask.expand_as(img) + img *= mask + return img + + +class RandAugment: + def __init__(self, n, m, randm=False): + self.n = n + self.m = m # [0, 30] + self.augment_list = augment_list() + self.randm = randm + + def __call__(self, img): + ops = random.choices(self.augment_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + + return img diff --git a/Meta-causal/code-withStyleAttack/56717.error b/Meta-causal/code-withStyleAttack/56717.error new file mode 100644 index 0000000000000000000000000000000000000000..f4d95947c7a86339e1d04481c9ef0f88fee09876 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56717.error @@ -0,0 +1 @@ +run_my_joint_v13_test.sh: line 27: m}: command not found diff --git a/Meta-causal/code-withStyleAttack/56717.log b/Meta-causal/code-withStyleAttack/56717.log new file mode 100644 index 0000000000000000000000000000000000000000..342d449cbbc0cf96ab603cfcc9a39a8178c93297 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56717.log @@ -0,0 +1,334 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 396.56, cls_loss 6.7564 cls_loss_mapping 1.5193 cls_loss_causal 1.7521 re_mapping 1.0575 re_causal 1.0584 /// teacc 81.25 lr 0.00999497 +306 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 415.32, cls_loss 2.1970 cls_loss_mapping 0.9096 cls_loss_causal 1.4403 re_mapping 0.7024 re_causal 0.7051 /// teacc 83.65 lr 0.00997987 +306 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 457.96, cls_loss 1.3065 cls_loss_mapping 0.6322 cls_loss_causal 1.2780 re_mapping 0.6032 re_causal 0.6057 /// teacc 88.46 lr 0.00995475 +306 +0.009954748808839675 +changing lr +epoch 3, time 451.75, cls_loss 0.5818 cls_loss_mapping 0.5055 cls_loss_causal 1.1465 re_mapping 0.5267 re_causal 0.5293 /// teacc 87.02 lr 0.00991965 +306 +0.009919647942993149 +changing lr +epoch 4, time 451.48, cls_loss 0.3909 cls_loss_mapping 0.4012 cls_loss_causal 1.0889 re_mapping 0.4649 re_causal 0.4683 /// teacc 84.62 lr 0.00987464 +306 +0.009874639560909117 +changing lr +epoch 5, time 441.59, cls_loss 0.3191 cls_loss_mapping 0.3555 cls_loss_causal 1.0670 re_mapping 0.3968 re_causal 0.4013 /// teacc 86.06 lr 0.00981981 +306 +0.009819814303479266 +changing lr +epoch 6, time 432.93, cls_loss 0.1327 cls_loss_mapping 0.2760 cls_loss_causal 1.0002 re_mapping 0.3232 re_causal 0.3278 /// teacc 83.17 lr 0.00975528 +306 +0.009755282581475767 +changing lr +epoch 7, time 444.85, cls_loss 0.0411 cls_loss_mapping 0.2236 cls_loss_causal 0.9368 re_mapping 0.2592 re_causal 0.2641 /// teacc 88.46 lr 0.00968117 +306 +0.009681174353198686 +changing lr +epoch 8, time 448.36, cls_loss 0.0723 cls_loss_mapping 0.2492 cls_loss_causal 0.9911 re_mapping 0.2174 re_causal 0.2224 /// teacc 86.54 lr 0.00959764 +306 +0.009597638862757255 +changing lr +epoch 9, time 446.26, cls_loss 0.0174 cls_loss_mapping 0.1853 cls_loss_causal 0.8733 re_mapping 0.1873 re_causal 0.1925 /// teacc 86.54 lr 0.00950484 +306 +0.009504844339512096 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 457.12, cls_loss 0.0358 cls_loss_mapping 0.1781 cls_loss_causal 0.8735 re_mapping 0.1610 re_causal 0.1661 /// teacc 89.90 lr 0.00940298 +306 +0.009402977659283692 +changing lr +epoch 11, time 443.50, cls_loss 0.0162 cls_loss_mapping 0.1514 cls_loss_causal 0.8453 re_mapping 0.1432 re_causal 0.1486 /// teacc 89.90 lr 0.00929224 +306 +0.009292243968009333 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 453.53, cls_loss 0.0101 cls_loss_mapping 0.1383 cls_loss_causal 0.8002 re_mapping 0.1270 re_causal 0.1328 /// teacc 90.87 lr 0.00917287 +306 +0.009172866268606516 +changing lr +epoch 13, time 466.47, cls_loss 0.0092 cls_loss_mapping 0.1432 cls_loss_causal 0.8412 re_mapping 0.1167 re_causal 0.1224 /// teacc 90.38 lr 0.00904508 +306 +0.00904508497187474 +changing lr +epoch 14, time 448.78, cls_loss 0.0063 cls_loss_mapping 0.1207 cls_loss_causal 0.7912 re_mapping 0.1077 re_causal 0.1140 /// teacc 90.38 lr 0.00890916 +306 +0.008909157412340152 +changing lr +epoch 15, time 442.01, cls_loss 0.0075 cls_loss_mapping 0.1148 cls_loss_causal 0.7640 re_mapping 0.0982 re_causal 0.1047 /// teacc 89.90 lr 0.00876536 +306 +0.00876535733001806 +changing lr +epoch 16, time 451.70, cls_loss 0.0050 cls_loss_mapping 0.1000 cls_loss_causal 0.7562 re_mapping 0.0898 re_causal 0.0964 /// teacc 90.87 lr 0.00861397 +306 +0.008613974319136962 +changing lr +epoch 17, time 454.36, cls_loss 0.0084 cls_loss_mapping 0.0986 cls_loss_causal 0.7422 re_mapping 0.0817 re_causal 0.0883 /// teacc 89.90 lr 0.00845531 +306 +0.008455313244934327 +changing lr +epoch 18, time 450.70, cls_loss 0.0033 cls_loss_mapping 0.0951 cls_loss_causal 0.7426 re_mapping 0.0760 re_causal 0.0827 /// teacc 89.42 lr 0.00828969 +306 +0.008289693629698565 +changing lr +epoch 19, time 456.56, cls_loss 0.0051 cls_loss_mapping 0.0938 cls_loss_causal 0.7288 re_mapping 0.0711 re_causal 0.0787 /// teacc 88.94 lr 0.00811745 +306 +0.00811744900929367 +changing lr +epoch 20, time 444.31, cls_loss 0.0025 cls_loss_mapping 0.0920 cls_loss_causal 0.7432 re_mapping 0.0652 re_causal 0.0723 /// teacc 89.90 lr 0.00793893 +306 +0.007938926261462368 +changing lr +epoch 21, time 436.20, cls_loss 0.0028 cls_loss_mapping 0.0782 cls_loss_causal 0.7226 re_mapping 0.0605 re_causal 0.0677 /// teacc 90.87 lr 0.00775448 +306 +0.007754484907260515 +changing lr +epoch 22, time 447.42, cls_loss 0.0020 cls_loss_mapping 0.0778 cls_loss_causal 0.6694 re_mapping 0.0571 re_causal 0.0641 /// teacc 90.38 lr 0.00756450 +306 +0.007564496387029534 +changing lr +epoch 23, time 443.40, cls_loss 0.0019 cls_loss_mapping 0.0766 cls_loss_causal 0.7606 re_mapping 0.0533 re_causal 0.0621 /// teacc 89.42 lr 0.00736934 +306 +0.007369343312364995 +changing lr +epoch 24, time 439.80, cls_loss 0.0045 cls_loss_mapping 0.0782 cls_loss_causal 0.7261 re_mapping 0.0521 re_causal 0.0608 /// teacc 90.38 lr 0.00716942 +306 +0.0071694186955877925 +changing lr +epoch 25, time 430.50, cls_loss 0.0020 cls_loss_mapping 0.0645 cls_loss_causal 0.7059 re_mapping 0.0500 re_causal 0.0593 /// teacc 90.87 lr 0.00696513 +306 +0.0069651251582696205 +changing lr +epoch 26, time 444.21, cls_loss 0.0008 cls_loss_mapping 0.0529 cls_loss_causal 0.6660 re_mapping 0.0448 re_causal 0.0527 /// teacc 90.87 lr 0.00675687 +306 +0.006756874120406716 +changing lr +epoch 27, time 451.19, cls_loss 0.0027 cls_loss_mapping 0.0633 cls_loss_causal 0.7457 re_mapping 0.0430 re_causal 0.0520 /// teacc 90.87 lr 0.00654508 +306 +0.00654508497187474 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 444.91, cls_loss 0.0045 cls_loss_mapping 0.0630 cls_loss_causal 0.6839 re_mapping 0.0409 re_causal 0.0485 /// teacc 91.35 lr 0.00633018 +306 +0.006330184227833378 +changing lr +epoch 29, time 454.38, cls_loss 0.0030 cls_loss_mapping 0.0528 cls_loss_causal 0.6373 re_mapping 0.0388 re_causal 0.0468 /// teacc 88.94 lr 0.00611260 +306 +0.006112604669781575 +changing lr +epoch 30, time 455.36, cls_loss 0.0023 cls_loss_mapping 0.0479 cls_loss_causal 0.6459 re_mapping 0.0382 re_causal 0.0462 /// teacc 91.35 lr 0.00589278 +306 +0.005892784473993186 +changing lr +epoch 31, time 447.61, cls_loss 0.0014 cls_loss_mapping 0.0532 cls_loss_causal 0.6553 re_mapping 0.0365 re_causal 0.0447 /// teacc 91.35 lr 0.00567117 +306 +0.00567116632908828 +changing lr +epoch 32, time 455.74, cls_loss 0.0019 cls_loss_mapping 0.0470 cls_loss_causal 0.6156 re_mapping 0.0346 re_causal 0.0422 /// teacc 90.38 lr 0.00544820 +306 +0.00544819654451717 +changing lr +epoch 33, time 458.62, cls_loss 0.0026 cls_loss_mapping 0.0475 cls_loss_causal 0.6128 re_mapping 0.0336 re_causal 0.0415 /// teacc 91.35 lr 0.00522432 +306 +0.005224324151752577 +changing lr +epoch 34, time 443.89, cls_loss 0.0034 cls_loss_mapping 0.0503 cls_loss_causal 0.6216 re_mapping 0.0331 re_causal 0.0412 /// teacc 90.87 lr 0.00500000 +306 +0.005000000000000003 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 474.23, cls_loss 0.0025 cls_loss_mapping 0.0398 cls_loss_causal 0.5884 re_mapping 0.0317 re_causal 0.0397 /// teacc 91.83 lr 0.00477568 +306 +0.004775675848247429 +changing lr +epoch 36, time 456.46, cls_loss 0.0023 cls_loss_mapping 0.0434 cls_loss_causal 0.6319 re_mapping 0.0308 re_causal 0.0386 /// teacc 91.35 lr 0.00455180 +306 +0.004551803455482836 +changing lr +epoch 37, time 460.36, cls_loss 0.0024 cls_loss_mapping 0.0376 cls_loss_causal 0.6052 re_mapping 0.0290 re_causal 0.0364 /// teacc 90.87 lr 0.00432883 +306 +0.004328833670911726 +changing lr +epoch 38, time 456.58, cls_loss 0.0013 cls_loss_mapping 0.0368 cls_loss_causal 0.6265 re_mapping 0.0276 re_causal 0.0354 /// teacc 90.38 lr 0.00410722 +306 +0.0041072155260068206 +changing lr +epoch 39, time 468.90, cls_loss 0.0019 cls_loss_mapping 0.0310 cls_loss_causal 0.6240 re_mapping 0.0264 re_causal 0.0344 /// teacc 90.87 lr 0.00388740 +306 +0.0038873953302184317 +changing lr +epoch 40, time 457.96, cls_loss 0.0020 cls_loss_mapping 0.0328 cls_loss_causal 0.6230 re_mapping 0.0257 re_causal 0.0335 /// teacc 90.87 lr 0.00366982 +306 +0.003669815772166629 +changing lr +---------------------saving model at epoch 41---------------------------------------------------- +epoch 41, time 469.29, cls_loss 0.0023 cls_loss_mapping 0.0376 cls_loss_causal 0.6061 re_mapping 0.0249 re_causal 0.0320 /// teacc 92.31 lr 0.00345492 +306 +0.0034549150281252667 +changing lr +epoch 42, time 475.72, cls_loss 0.0025 cls_loss_mapping 0.0311 cls_loss_causal 0.6195 re_mapping 0.0243 re_causal 0.0322 /// teacc 90.87 lr 0.00324313 +306 +0.0032431258795932905 +changing lr +epoch 43, time 450.85, cls_loss 0.0018 cls_loss_mapping 0.0341 cls_loss_causal 0.6223 re_mapping 0.0235 re_causal 0.0310 /// teacc 90.87 lr 0.00303487 +306 +0.0030348748417303863 +changing lr +epoch 44, time 441.78, cls_loss 0.0019 cls_loss_mapping 0.0317 cls_loss_causal 0.6072 re_mapping 0.0228 re_causal 0.0304 /// teacc 90.38 lr 0.00283058 +306 +0.0028305813044122124 +changing lr +---------------------saving model at epoch 45---------------------------------------------------- +epoch 45, time 462.98, cls_loss 0.0013 cls_loss_mapping 0.0307 cls_loss_causal 0.5641 re_mapping 0.0222 re_causal 0.0291 /// teacc 93.75 lr 0.00263066 +306 +0.0026306566876350096 +changing lr +epoch 46, time 474.81, cls_loss 0.0028 cls_loss_mapping 0.0323 cls_loss_causal 0.6004 re_mapping 0.0218 re_causal 0.0287 /// teacc 91.83 lr 0.00243550 +306 +0.0024355036129704724 +changing lr +epoch 47, time 465.56, cls_loss 0.0013 cls_loss_mapping 0.0291 cls_loss_causal 0.6082 re_mapping 0.0213 re_causal 0.0289 /// teacc 92.31 lr 0.00224552 +306 +0.00224551509273949 +changing lr +epoch 48, time 458.33, cls_loss 0.0011 cls_loss_mapping 0.0269 cls_loss_causal 0.6051 re_mapping 0.0208 re_causal 0.0289 /// teacc 91.35 lr 0.00206107 +306 +0.002061073738537637 +changing lr +epoch 49, time 450.51, cls_loss 0.0012 cls_loss_mapping 0.0242 cls_loss_causal 0.5558 re_mapping 0.0200 re_causal 0.0273 /// teacc 91.35 lr 0.00188255 +306 +0.0018825509907063344 +changing lr +epoch 50, time 462.46, cls_loss 0.0009 cls_loss_mapping 0.0237 cls_loss_causal 0.5775 re_mapping 0.0194 re_causal 0.0261 /// teacc 90.38 lr 0.00171031 +306 +0.0017103063703014388 +changing lr +epoch 51, time 458.67, cls_loss 0.0017 cls_loss_mapping 0.0239 cls_loss_causal 0.5359 re_mapping 0.0184 re_causal 0.0244 /// teacc 91.35 lr 0.00154469 +306 +0.0015446867550656784 +changing lr +epoch 52, time 439.55, cls_loss 0.0016 cls_loss_mapping 0.0239 cls_loss_causal 0.5782 re_mapping 0.0180 re_causal 0.0248 /// teacc 92.31 lr 0.00138603 +306 +0.001386025680863044 +changing lr +epoch 53, time 468.39, cls_loss 0.0011 cls_loss_mapping 0.0221 cls_loss_causal 0.5797 re_mapping 0.0174 re_causal 0.0241 /// teacc 90.38 lr 0.00123464 +306 +0.0012346426699819469 +changing lr +epoch 54, time 478.52, cls_loss 0.0011 cls_loss_mapping 0.0208 cls_loss_causal 0.5323 re_mapping 0.0171 re_causal 0.0233 /// teacc 91.35 lr 0.00109084 +306 +0.0010908425876598518 +changing lr +epoch 55, time 451.23, cls_loss 0.0018 cls_loss_mapping 0.0228 cls_loss_causal 0.5217 re_mapping 0.0167 re_causal 0.0227 /// teacc 91.35 lr 0.00095492 +306 +0.000954915028125264 +changing lr +epoch 56, time 455.62, cls_loss 0.0008 cls_loss_mapping 0.0185 cls_loss_causal 0.5520 re_mapping 0.0165 re_causal 0.0225 /// teacc 90.87 lr 0.00082713 +306 +0.0008271337313934874 +changing lr +epoch 57, time 455.64, cls_loss 0.0015 cls_loss_mapping 0.0242 cls_loss_causal 0.5776 re_mapping 0.0162 re_causal 0.0225 /// teacc 90.87 lr 0.00070776 +306 +0.00070775603199067 +changing lr +epoch 58, time 446.78, cls_loss 0.0009 cls_loss_mapping 0.0185 cls_loss_causal 0.5541 re_mapping 0.0158 re_causal 0.0221 /// teacc 91.35 lr 0.00059702 +306 +0.0005970223407163104 +changing lr +epoch 59, time 451.88, cls_loss 0.0025 cls_loss_mapping 0.0193 cls_loss_causal 0.5280 re_mapping 0.0156 re_causal 0.0217 /// teacc 92.31 lr 0.00049516 +306 +0.0004951556604879052 +changing lr +epoch 60, time 459.80, cls_loss 0.0019 cls_loss_mapping 0.0191 cls_loss_causal 0.5650 re_mapping 0.0154 re_causal 0.0212 /// teacc 91.83 lr 0.00040236 +306 +0.00040236113724274745 +changing lr +epoch 61, time 456.30, cls_loss 0.0013 cls_loss_mapping 0.0195 cls_loss_causal 0.5573 re_mapping 0.0151 re_causal 0.0209 /// teacc 90.87 lr 0.00031883 +306 +0.00031882564680131423 +changing lr +epoch 62, time 461.25, cls_loss 0.0016 cls_loss_mapping 0.0184 cls_loss_causal 0.5320 re_mapping 0.0149 re_causal 0.0203 /// teacc 91.83 lr 0.00024472 +306 +0.0002447174185242325 +changing lr +epoch 63, time 461.95, cls_loss 0.0025 cls_loss_mapping 0.0234 cls_loss_causal 0.5478 re_mapping 0.0148 re_causal 0.0203 /// teacc 91.35 lr 0.00018019 +306 +0.0001801856965207339 +changing lr +epoch 64, time 443.04, cls_loss 0.0012 cls_loss_mapping 0.0208 cls_loss_causal 0.5022 re_mapping 0.0147 re_causal 0.0200 /// teacc 91.35 lr 0.00012536 +306 +0.000125360439090882 +changing lr +epoch 65, time 454.35, cls_loss 0.0012 cls_loss_mapping 0.0176 cls_loss_causal 0.5745 re_mapping 0.0147 re_causal 0.0203 /// teacc 91.83 lr 0.00008035 +306 +8.03520570068517e-05 +changing lr +epoch 66, time 462.74, cls_loss 0.0018 cls_loss_mapping 0.0228 cls_loss_causal 0.5579 re_mapping 0.0147 re_causal 0.0201 /// teacc 91.35 lr 0.00004525 +306 +4.5251191160326525e-05 +changing lr +epoch 67, time 470.10, cls_loss 0.0012 cls_loss_mapping 0.0186 cls_loss_causal 0.5288 re_mapping 0.0147 re_causal 0.0205 /// teacc 92.31 lr 0.00002013 +306 +2.0128530023804673e-05 +changing lr +epoch 68, time 446.31, cls_loss 0.0011 cls_loss_mapping 0.0165 cls_loss_causal 0.5339 re_mapping 0.0146 re_causal 0.0202 /// teacc 89.42 lr 0.00000503 +306 +5.034667293427056e-06 +changing lr +epoch 69, time 458.08, cls_loss 0.0013 cls_loss_mapping 0.0148 cls_loss_causal 0.5422 re_mapping 0.0146 re_causal 0.0204 /// teacc 92.31 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/art_painting_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 99.169922 65.784983 95.209581 64.596589 75.197051 + art_painting cartoon photo sketch Avg +do 99.072266 64.803754 95.269461 64.342072 74.805096 diff --git a/Meta-causal/code-withStyleAttack/56718.error b/Meta-causal/code-withStyleAttack/56718.error new file mode 100644 index 0000000000000000000000000000000000000000..f26c68e6c5fa980b508c7bd532627e6b75b149fa --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56718.error @@ -0,0 +1,2 @@ +bash: run_my_joint_v13_test.sh: No such file or directory +srun: error: gcpl4-eu-1: task 0: Exited with exit code 127 diff --git a/Meta-causal/code-withStyleAttack/56718.log b/Meta-causal/code-withStyleAttack/56718.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/56719.error b/Meta-causal/code-withStyleAttack/56719.error new file mode 100644 index 0000000000000000000000000000000000000000..444e676738c3b4b1c880f3c832cec125757b1b1b --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56719.error @@ -0,0 +1,25 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/main_test_digit_v13.py", line 142, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/main_test_digit_v13.py", line 28, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/main_test_digit_v13.py", line 101, in evaluate_digit + teset = str2fun[data]('test', channels=channels) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/data_loader_joint_v3.py", line 722, in load_mnist_m + with open(path, 'rb') as f: + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: 'data/mnist_m-test.pkl' +srun: error: gcpl4-eu-1: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-withStyleAttack/56719.log b/Meta-causal/code-withStyleAttack/56719.log new file mode 100644 index 0000000000000000000000000000000000000000..877231eeeaf1d132b8ddf33e5c8762b82226e64b --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56719.log @@ -0,0 +1,2066 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 260.22, cls_loss 1.1168 cls_loss_mapping 1.7217 cls_loss_causal 2.1730 re_mapping 0.1107 re_causal 0.1210 /// teacc 88.60 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 265.91, cls_loss 0.3310 cls_loss_mapping 0.6635 cls_loss_causal 1.7775 re_mapping 0.1227 re_causal 0.1643 /// teacc 94.13 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 274.59, cls_loss 0.2190 cls_loss_mapping 0.3836 cls_loss_causal 1.5398 re_mapping 0.0889 re_causal 0.1349 /// teacc 95.73 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 274.87, cls_loss 0.1538 cls_loss_mapping 0.2464 cls_loss_causal 1.3205 re_mapping 0.0726 re_causal 0.1133 /// teacc 96.67 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 274.40, cls_loss 0.1333 cls_loss_mapping 0.2005 cls_loss_causal 1.2889 re_mapping 0.0565 re_causal 0.0967 /// teacc 96.76 lr 0.00010000 +100 +0.0001 +changing lr +epoch 5, time 273.58, cls_loss 0.1192 cls_loss_mapping 0.1777 cls_loss_causal 1.1780 re_mapping 0.0494 re_causal 0.0858 /// teacc 96.69 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 274.48, cls_loss 0.1015 cls_loss_mapping 0.1485 cls_loss_causal 1.1906 re_mapping 0.0407 re_causal 0.0792 /// teacc 97.65 lr 0.00010000 +100 +0.0001 +changing lr +epoch 7, time 273.51, cls_loss 0.0994 cls_loss_mapping 0.1401 cls_loss_causal 1.0640 re_mapping 0.0373 re_causal 0.0706 /// teacc 97.62 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 274.20, cls_loss 0.0851 cls_loss_mapping 0.1189 cls_loss_causal 1.0603 re_mapping 0.0328 re_causal 0.0659 /// teacc 97.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 9, time 273.45, cls_loss 0.0854 cls_loss_mapping 0.1226 cls_loss_causal 1.0207 re_mapping 0.0298 re_causal 0.0623 /// teacc 97.90 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 273.93, cls_loss 0.0650 cls_loss_mapping 0.0935 cls_loss_causal 0.9621 re_mapping 0.0281 re_causal 0.0602 /// teacc 98.02 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 274.43, cls_loss 0.0669 cls_loss_mapping 0.0951 cls_loss_causal 0.9560 re_mapping 0.0255 re_causal 0.0558 /// teacc 98.22 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 274.42, cls_loss 0.0667 cls_loss_mapping 0.0970 cls_loss_causal 0.9466 re_mapping 0.0245 re_causal 0.0554 /// teacc 98.28 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 273.96, cls_loss 0.0591 cls_loss_mapping 0.0844 cls_loss_causal 0.9433 re_mapping 0.0231 re_causal 0.0545 /// teacc 98.31 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 273.58, cls_loss 0.0548 cls_loss_mapping 0.0830 cls_loss_causal 0.8947 re_mapping 0.0220 re_causal 0.0519 /// teacc 98.41 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 269.38, cls_loss 0.0418 cls_loss_mapping 0.0628 cls_loss_causal 0.9005 re_mapping 0.0207 re_causal 0.0518 /// teacc 98.44 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 265.41, cls_loss 0.0544 cls_loss_mapping 0.0769 cls_loss_causal 0.8831 re_mapping 0.0197 re_causal 0.0493 /// teacc 98.48 lr 0.00010000 +100 +0.0001 +changing lr +epoch 17, time 261.91, cls_loss 0.0525 cls_loss_mapping 0.0776 cls_loss_causal 0.8870 re_mapping 0.0197 re_causal 0.0493 /// teacc 98.32 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 262.02, cls_loss 0.0382 cls_loss_mapping 0.0581 cls_loss_causal 0.8764 re_mapping 0.0184 re_causal 0.0472 /// teacc 98.51 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 262.11, cls_loss 0.0374 cls_loss_mapping 0.0573 cls_loss_causal 0.7987 re_mapping 0.0184 re_causal 0.0452 /// teacc 98.54 lr 0.00010000 +100 +0.0001 +changing lr +epoch 20, time 261.89, cls_loss 0.0342 cls_loss_mapping 0.0538 cls_loss_causal 0.7636 re_mapping 0.0178 re_causal 0.0453 /// teacc 98.52 lr 0.00010000 +100 +0.0001 +changing lr +epoch 21, time 261.79, cls_loss 0.0292 cls_loss_mapping 0.0457 cls_loss_causal 0.7961 re_mapping 0.0171 re_causal 0.0436 /// teacc 98.43 lr 0.00010000 +100 +0.0001 +changing lr +epoch 22, time 261.58, cls_loss 0.0277 cls_loss_mapping 0.0426 cls_loss_causal 0.8074 re_mapping 0.0162 re_causal 0.0421 /// teacc 98.49 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 262.62, cls_loss 0.0333 cls_loss_mapping 0.0530 cls_loss_causal 0.7916 re_mapping 0.0156 re_causal 0.0414 /// teacc 98.64 lr 0.00010000 +100 +0.0001 +changing lr +epoch 24, time 262.08, cls_loss 0.0296 cls_loss_mapping 0.0474 cls_loss_causal 0.7989 re_mapping 0.0151 re_causal 0.0402 /// teacc 98.45 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 263.12, cls_loss 0.0246 cls_loss_mapping 0.0418 cls_loss_causal 0.7816 re_mapping 0.0149 re_causal 0.0393 /// teacc 98.75 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 262.71, cls_loss 0.0229 cls_loss_mapping 0.0378 cls_loss_causal 0.7518 re_mapping 0.0141 re_causal 0.0374 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +epoch 27, time 261.78, cls_loss 0.0247 cls_loss_mapping 0.0419 cls_loss_causal 0.7570 re_mapping 0.0147 re_causal 0.0376 /// teacc 98.74 lr 0.00010000 +100 +0.0001 +changing lr +epoch 28, time 262.28, cls_loss 0.0212 cls_loss_mapping 0.0304 cls_loss_causal 0.7520 re_mapping 0.0141 re_causal 0.0367 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 262.94, cls_loss 0.0295 cls_loss_mapping 0.0448 cls_loss_causal 0.7504 re_mapping 0.0136 re_causal 0.0360 /// teacc 98.83 lr 0.00010000 +100 +0.0001 +changing lr +epoch 30, time 262.56, cls_loss 0.0240 cls_loss_mapping 0.0389 cls_loss_causal 0.7479 re_mapping 0.0136 re_causal 0.0364 /// teacc 98.76 lr 0.00010000 +100 +0.0001 +changing lr +epoch 31, time 262.55, cls_loss 0.0208 cls_loss_mapping 0.0348 cls_loss_causal 0.7169 re_mapping 0.0130 re_causal 0.0347 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 32, time 262.24, cls_loss 0.0193 cls_loss_mapping 0.0329 cls_loss_causal 0.6995 re_mapping 0.0122 re_causal 0.0327 /// teacc 98.66 lr 0.00010000 +100 +0.0001 +changing lr +epoch 33, time 262.86, cls_loss 0.0189 cls_loss_mapping 0.0334 cls_loss_causal 0.7307 re_mapping 0.0124 re_causal 0.0343 /// teacc 98.57 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 263.86, cls_loss 0.0187 cls_loss_mapping 0.0314 cls_loss_causal 0.7412 re_mapping 0.0121 re_causal 0.0325 /// teacc 98.88 lr 0.00010000 +100 +0.0001 +changing lr +epoch 35, time 262.90, cls_loss 0.0162 cls_loss_mapping 0.0290 cls_loss_causal 0.7096 re_mapping 0.0120 re_causal 0.0328 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 36, time 263.18, cls_loss 0.0130 cls_loss_mapping 0.0216 cls_loss_causal 0.6816 re_mapping 0.0117 re_causal 0.0312 /// teacc 98.71 lr 0.00010000 +100 +0.0001 +changing lr +epoch 37, time 263.13, cls_loss 0.0150 cls_loss_mapping 0.0245 cls_loss_causal 0.6711 re_mapping 0.0119 re_causal 0.0316 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +epoch 38, time 262.71, cls_loss 0.0171 cls_loss_mapping 0.0291 cls_loss_causal 0.6826 re_mapping 0.0114 re_causal 0.0303 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 263.42, cls_loss 0.0148 cls_loss_mapping 0.0251 cls_loss_causal 0.6789 re_mapping 0.0111 re_causal 0.0298 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 263.43, cls_loss 0.0165 cls_loss_mapping 0.0301 cls_loss_causal 0.6877 re_mapping 0.0113 re_causal 0.0297 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 41, time 262.57, cls_loss 0.0161 cls_loss_mapping 0.0290 cls_loss_causal 0.6867 re_mapping 0.0103 re_causal 0.0283 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 42, time 261.54, cls_loss 0.0124 cls_loss_mapping 0.0221 cls_loss_causal 0.6524 re_mapping 0.0104 re_causal 0.0276 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 43, time 261.64, cls_loss 0.0121 cls_loss_mapping 0.0236 cls_loss_causal 0.6499 re_mapping 0.0107 re_causal 0.0281 /// teacc 98.87 lr 0.00010000 +100 +0.0001 +changing lr +epoch 44, time 261.70, cls_loss 0.0126 cls_loss_mapping 0.0222 cls_loss_causal 0.6472 re_mapping 0.0107 re_causal 0.0277 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 45, time 262.23, cls_loss 0.0139 cls_loss_mapping 0.0248 cls_loss_causal 0.6458 re_mapping 0.0097 re_causal 0.0267 /// teacc 98.72 lr 0.00010000 +100 +0.0001 +changing lr +epoch 46, time 262.35, cls_loss 0.0128 cls_loss_mapping 0.0236 cls_loss_causal 0.6192 re_mapping 0.0103 re_causal 0.0264 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +epoch 47, time 262.98, cls_loss 0.0120 cls_loss_mapping 0.0198 cls_loss_causal 0.6455 re_mapping 0.0097 re_causal 0.0258 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 48, time 262.58, cls_loss 0.0116 cls_loss_mapping 0.0229 cls_loss_causal 0.6623 re_mapping 0.0099 re_causal 0.0264 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 49, time 262.10, cls_loss 0.0109 cls_loss_mapping 0.0222 cls_loss_causal 0.6632 re_mapping 0.0094 re_causal 0.0260 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 50, time 262.49, cls_loss 0.0107 cls_loss_mapping 0.0186 cls_loss_causal 0.6425 re_mapping 0.0094 re_causal 0.0260 /// teacc 98.74 lr 0.00010000 +100 +0.0001 +changing lr +epoch 51, time 261.98, cls_loss 0.0105 cls_loss_mapping 0.0196 cls_loss_causal 0.6062 re_mapping 0.0099 re_causal 0.0249 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 52---------------------------------------------------- +epoch 52, time 262.70, cls_loss 0.0123 cls_loss_mapping 0.0222 cls_loss_causal 0.6539 re_mapping 0.0090 re_causal 0.0243 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 53, time 262.04, cls_loss 0.0082 cls_loss_mapping 0.0165 cls_loss_causal 0.5830 re_mapping 0.0095 re_causal 0.0243 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 54, time 262.03, cls_loss 0.0134 cls_loss_mapping 0.0238 cls_loss_causal 0.6506 re_mapping 0.0092 re_causal 0.0241 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 55, time 262.23, cls_loss 0.0092 cls_loss_mapping 0.0175 cls_loss_causal 0.6151 re_mapping 0.0094 re_causal 0.0241 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 56, time 261.68, cls_loss 0.0083 cls_loss_mapping 0.0146 cls_loss_causal 0.6247 re_mapping 0.0093 re_causal 0.0250 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 57, time 262.04, cls_loss 0.0094 cls_loss_mapping 0.0173 cls_loss_causal 0.6450 re_mapping 0.0082 re_causal 0.0236 /// teacc 98.80 lr 0.00010000 +100 +0.0001 +changing lr +epoch 58, time 262.22, cls_loss 0.0082 cls_loss_mapping 0.0182 cls_loss_causal 0.5940 re_mapping 0.0090 re_causal 0.0236 /// teacc 98.76 lr 0.00010000 +100 +0.0001 +changing lr +epoch 59, time 261.98, cls_loss 0.0107 cls_loss_mapping 0.0187 cls_loss_causal 0.6018 re_mapping 0.0082 re_causal 0.0217 /// teacc 98.85 lr 0.00010000 +100 +0.0001 +changing lr +epoch 60, time 262.64, cls_loss 0.0096 cls_loss_mapping 0.0165 cls_loss_causal 0.6197 re_mapping 0.0079 re_causal 0.0227 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 61, time 262.62, cls_loss 0.0077 cls_loss_mapping 0.0133 cls_loss_causal 0.6104 re_mapping 0.0077 re_causal 0.0216 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 62, time 261.55, cls_loss 0.0094 cls_loss_mapping 0.0177 cls_loss_causal 0.6325 re_mapping 0.0077 re_causal 0.0211 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 63, time 261.66, cls_loss 0.0096 cls_loss_mapping 0.0173 cls_loss_causal 0.6390 re_mapping 0.0075 re_causal 0.0211 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 64, time 262.39, cls_loss 0.0089 cls_loss_mapping 0.0176 cls_loss_causal 0.6220 re_mapping 0.0080 re_causal 0.0211 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 65, time 262.81, cls_loss 0.0054 cls_loss_mapping 0.0089 cls_loss_causal 0.5919 re_mapping 0.0081 re_causal 0.0215 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 66, time 262.62, cls_loss 0.0072 cls_loss_mapping 0.0145 cls_loss_causal 0.5995 re_mapping 0.0079 re_causal 0.0213 /// teacc 98.80 lr 0.00010000 +100 +0.0001 +changing lr +epoch 67, time 262.54, cls_loss 0.0065 cls_loss_mapping 0.0115 cls_loss_causal 0.5839 re_mapping 0.0082 re_causal 0.0214 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 68, time 262.66, cls_loss 0.0082 cls_loss_mapping 0.0151 cls_loss_causal 0.6010 re_mapping 0.0072 re_causal 0.0203 /// teacc 98.68 lr 0.00010000 +100 +0.0001 +changing lr +epoch 69, time 262.04, cls_loss 0.0072 cls_loss_mapping 0.0131 cls_loss_causal 0.5964 re_mapping 0.0075 re_causal 0.0202 /// teacc 98.86 lr 0.00010000 +100 +0.0001 +changing lr +epoch 70, time 262.42, cls_loss 0.0068 cls_loss_mapping 0.0141 cls_loss_causal 0.6231 re_mapping 0.0076 re_causal 0.0214 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 71, time 261.65, cls_loss 0.0077 cls_loss_mapping 0.0151 cls_loss_causal 0.5752 re_mapping 0.0073 re_causal 0.0197 /// teacc 98.89 lr 0.00010000 +100 +0.0001 +changing lr +epoch 72, time 262.42, cls_loss 0.0077 cls_loss_mapping 0.0130 cls_loss_causal 0.5860 re_mapping 0.0073 re_causal 0.0197 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 73---------------------------------------------------- +epoch 73, time 263.56, cls_loss 0.0074 cls_loss_mapping 0.0145 cls_loss_causal 0.5783 re_mapping 0.0071 re_causal 0.0192 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 74, time 262.97, cls_loss 0.0064 cls_loss_mapping 0.0111 cls_loss_causal 0.5752 re_mapping 0.0073 re_causal 0.0198 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 75, time 262.29, cls_loss 0.0067 cls_loss_mapping 0.0128 cls_loss_causal 0.5738 re_mapping 0.0072 re_causal 0.0190 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 76, time 262.02, cls_loss 0.0060 cls_loss_mapping 0.0107 cls_loss_causal 0.5944 re_mapping 0.0071 re_causal 0.0198 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 77, time 262.25, cls_loss 0.0055 cls_loss_mapping 0.0086 cls_loss_causal 0.5748 re_mapping 0.0066 re_causal 0.0185 /// teacc 98.83 lr 0.00010000 +100 +0.0001 +changing lr +epoch 78, time 262.33, cls_loss 0.0048 cls_loss_mapping 0.0105 cls_loss_causal 0.5729 re_mapping 0.0071 re_causal 0.0200 /// teacc 98.86 lr 0.00010000 +100 +0.0001 +changing lr +epoch 79, time 262.67, cls_loss 0.0054 cls_loss_mapping 0.0086 cls_loss_causal 0.5782 re_mapping 0.0067 re_causal 0.0187 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 80, time 262.54, cls_loss 0.0048 cls_loss_mapping 0.0092 cls_loss_causal 0.5620 re_mapping 0.0067 re_causal 0.0185 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 81, time 262.50, cls_loss 0.0063 cls_loss_mapping 0.0131 cls_loss_causal 0.6240 re_mapping 0.0067 re_causal 0.0190 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 82, time 261.76, cls_loss 0.0077 cls_loss_mapping 0.0136 cls_loss_causal 0.5922 re_mapping 0.0067 re_causal 0.0178 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 83, time 262.17, cls_loss 0.0064 cls_loss_mapping 0.0120 cls_loss_causal 0.5514 re_mapping 0.0073 re_causal 0.0188 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 84, time 262.19, cls_loss 0.0056 cls_loss_mapping 0.0093 cls_loss_causal 0.5766 re_mapping 0.0065 re_causal 0.0180 /// teacc 98.85 lr 0.00010000 +100 +0.0001 +changing lr +epoch 85, time 262.14, cls_loss 0.0050 cls_loss_mapping 0.0080 cls_loss_causal 0.5528 re_mapping 0.0063 re_causal 0.0174 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 86, time 261.86, cls_loss 0.0051 cls_loss_mapping 0.0088 cls_loss_causal 0.5929 re_mapping 0.0063 re_causal 0.0178 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 87, time 261.74, cls_loss 0.0050 cls_loss_mapping 0.0087 cls_loss_causal 0.5941 re_mapping 0.0063 re_causal 0.0177 /// teacc 98.80 lr 0.00010000 +100 +0.0001 +changing lr +epoch 88, time 260.51, cls_loss 0.0048 cls_loss_mapping 0.0085 cls_loss_causal 0.5624 re_mapping 0.0064 re_causal 0.0177 /// teacc 98.88 lr 0.00010000 +100 +0.0001 +changing lr +epoch 89, time 250.06, cls_loss 0.0047 cls_loss_mapping 0.0084 cls_loss_causal 0.5650 re_mapping 0.0065 re_causal 0.0173 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 90, time 250.78, cls_loss 0.0049 cls_loss_mapping 0.0091 cls_loss_causal 0.5613 re_mapping 0.0060 re_causal 0.0167 /// teacc 98.88 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 91---------------------------------------------------- +epoch 91, time 251.20, cls_loss 0.0041 cls_loss_mapping 0.0070 cls_loss_causal 0.5382 re_mapping 0.0064 re_causal 0.0165 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 92, time 251.25, cls_loss 0.0051 cls_loss_mapping 0.0108 cls_loss_causal 0.6002 re_mapping 0.0059 re_causal 0.0168 /// teacc 98.89 lr 0.00010000 +100 +0.0001 +changing lr +epoch 93, time 250.57, cls_loss 0.0049 cls_loss_mapping 0.0096 cls_loss_causal 0.5548 re_mapping 0.0063 re_causal 0.0168 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 94, time 250.77, cls_loss 0.0047 cls_loss_mapping 0.0096 cls_loss_causal 0.5460 re_mapping 0.0063 re_causal 0.0163 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 95, time 250.37, cls_loss 0.0045 cls_loss_mapping 0.0074 cls_loss_causal 0.5265 re_mapping 0.0064 re_causal 0.0160 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 96, time 249.18, cls_loss 0.0037 cls_loss_mapping 0.0063 cls_loss_causal 0.5633 re_mapping 0.0062 re_causal 0.0172 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 97, time 249.34, cls_loss 0.0051 cls_loss_mapping 0.0080 cls_loss_causal 0.5467 re_mapping 0.0057 re_causal 0.0156 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 98, time 249.24, cls_loss 0.0043 cls_loss_mapping 0.0077 cls_loss_causal 0.5665 re_mapping 0.0061 re_causal 0.0163 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 99, time 247.66, cls_loss 0.0042 cls_loss_mapping 0.0055 cls_loss_causal 0.5559 re_mapping 0.0059 re_causal 0.0160 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 100, time 247.43, cls_loss 0.0039 cls_loss_mapping 0.0072 cls_loss_causal 0.5491 re_mapping 0.0059 re_causal 0.0159 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 101, time 247.17, cls_loss 0.0036 cls_loss_mapping 0.0062 cls_loss_causal 0.5947 re_mapping 0.0058 re_causal 0.0166 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 102, time 247.21, cls_loss 0.0041 cls_loss_mapping 0.0065 cls_loss_causal 0.5484 re_mapping 0.0057 re_causal 0.0155 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 103, time 247.20, cls_loss 0.0047 cls_loss_mapping 0.0077 cls_loss_causal 0.5315 re_mapping 0.0056 re_causal 0.0149 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 104, time 247.26, cls_loss 0.0047 cls_loss_mapping 0.0076 cls_loss_causal 0.5507 re_mapping 0.0055 re_causal 0.0148 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 105, time 247.09, cls_loss 0.0040 cls_loss_mapping 0.0063 cls_loss_causal 0.5417 re_mapping 0.0052 re_causal 0.0150 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 106, time 247.22, cls_loss 0.0046 cls_loss_mapping 0.0085 cls_loss_causal 0.5688 re_mapping 0.0053 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 107, time 247.40, cls_loss 0.0039 cls_loss_mapping 0.0085 cls_loss_causal 0.5396 re_mapping 0.0057 re_causal 0.0155 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 108, time 247.52, cls_loss 0.0047 cls_loss_mapping 0.0094 cls_loss_causal 0.5722 re_mapping 0.0056 re_causal 0.0150 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 109, time 247.28, cls_loss 0.0036 cls_loss_mapping 0.0055 cls_loss_causal 0.5219 re_mapping 0.0055 re_causal 0.0145 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 110, time 247.31, cls_loss 0.0033 cls_loss_mapping 0.0053 cls_loss_causal 0.5339 re_mapping 0.0056 re_causal 0.0153 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 111---------------------------------------------------- +epoch 111, time 248.20, cls_loss 0.0044 cls_loss_mapping 0.0070 cls_loss_causal 0.5686 re_mapping 0.0051 re_causal 0.0146 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 112, time 247.44, cls_loss 0.0037 cls_loss_mapping 0.0064 cls_loss_causal 0.5641 re_mapping 0.0053 re_causal 0.0150 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 113, time 247.39, cls_loss 0.0037 cls_loss_mapping 0.0063 cls_loss_causal 0.5414 re_mapping 0.0054 re_causal 0.0149 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 114, time 247.52, cls_loss 0.0039 cls_loss_mapping 0.0082 cls_loss_causal 0.5541 re_mapping 0.0051 re_causal 0.0144 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 115, time 247.69, cls_loss 0.0040 cls_loss_mapping 0.0066 cls_loss_causal 0.5456 re_mapping 0.0054 re_causal 0.0145 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 116, time 247.22, cls_loss 0.0031 cls_loss_mapping 0.0053 cls_loss_causal 0.5168 re_mapping 0.0053 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 117---------------------------------------------------- +epoch 117, time 248.08, cls_loss 0.0053 cls_loss_mapping 0.0090 cls_loss_causal 0.5568 re_mapping 0.0053 re_causal 0.0148 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 118, time 247.60, cls_loss 0.0033 cls_loss_mapping 0.0064 cls_loss_causal 0.5252 re_mapping 0.0052 re_causal 0.0147 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 119, time 247.65, cls_loss 0.0033 cls_loss_mapping 0.0068 cls_loss_causal 0.5163 re_mapping 0.0053 re_causal 0.0149 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 120---------------------------------------------------- +epoch 120, time 249.12, cls_loss 0.0041 cls_loss_mapping 0.0073 cls_loss_causal 0.5428 re_mapping 0.0048 re_causal 0.0140 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 121, time 247.46, cls_loss 0.0038 cls_loss_mapping 0.0055 cls_loss_causal 0.5502 re_mapping 0.0047 re_causal 0.0134 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 122, time 247.34, cls_loss 0.0040 cls_loss_mapping 0.0070 cls_loss_causal 0.5413 re_mapping 0.0049 re_causal 0.0141 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 123, time 247.75, cls_loss 0.0029 cls_loss_mapping 0.0051 cls_loss_causal 0.5293 re_mapping 0.0052 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 124, time 247.49, cls_loss 0.0039 cls_loss_mapping 0.0059 cls_loss_causal 0.5299 re_mapping 0.0048 re_causal 0.0137 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 125, time 247.82, cls_loss 0.0035 cls_loss_mapping 0.0055 cls_loss_causal 0.5164 re_mapping 0.0052 re_causal 0.0143 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 126, time 247.56, cls_loss 0.0033 cls_loss_mapping 0.0056 cls_loss_causal 0.5298 re_mapping 0.0050 re_causal 0.0141 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 127, time 247.90, cls_loss 0.0032 cls_loss_mapping 0.0058 cls_loss_causal 0.5069 re_mapping 0.0051 re_causal 0.0140 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 128, time 247.44, cls_loss 0.0035 cls_loss_mapping 0.0061 cls_loss_causal 0.5469 re_mapping 0.0046 re_causal 0.0133 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 129, time 247.22, cls_loss 0.0035 cls_loss_mapping 0.0046 cls_loss_causal 0.5124 re_mapping 0.0049 re_causal 0.0131 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 130, time 247.53, cls_loss 0.0041 cls_loss_mapping 0.0070 cls_loss_causal 0.5574 re_mapping 0.0048 re_causal 0.0133 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 131, time 247.26, cls_loss 0.0034 cls_loss_mapping 0.0052 cls_loss_causal 0.5246 re_mapping 0.0049 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 132, time 247.37, cls_loss 0.0034 cls_loss_mapping 0.0064 cls_loss_causal 0.5529 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 133, time 247.41, cls_loss 0.0035 cls_loss_mapping 0.0043 cls_loss_causal 0.5204 re_mapping 0.0046 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 134, time 247.49, cls_loss 0.0033 cls_loss_mapping 0.0055 cls_loss_causal 0.5262 re_mapping 0.0045 re_causal 0.0127 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 135, time 247.43, cls_loss 0.0031 cls_loss_mapping 0.0054 cls_loss_causal 0.5655 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 136, time 247.17, cls_loss 0.0030 cls_loss_mapping 0.0045 cls_loss_causal 0.5369 re_mapping 0.0046 re_causal 0.0129 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 137---------------------------------------------------- +epoch 137, time 247.90, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.4877 re_mapping 0.0049 re_causal 0.0136 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 138, time 247.42, cls_loss 0.0034 cls_loss_mapping 0.0051 cls_loss_causal 0.5592 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 139, time 247.30, cls_loss 0.0035 cls_loss_mapping 0.0068 cls_loss_causal 0.4932 re_mapping 0.0048 re_causal 0.0133 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 140, time 247.43, cls_loss 0.0030 cls_loss_mapping 0.0041 cls_loss_causal 0.5293 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 141, time 247.95, cls_loss 0.0024 cls_loss_mapping 0.0036 cls_loss_causal 0.5366 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 142, time 247.33, cls_loss 0.0033 cls_loss_mapping 0.0058 cls_loss_causal 0.5003 re_mapping 0.0042 re_causal 0.0122 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 143, time 247.18, cls_loss 0.0030 cls_loss_mapping 0.0053 cls_loss_causal 0.5321 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 144, time 247.52, cls_loss 0.0032 cls_loss_mapping 0.0051 cls_loss_causal 0.4899 re_mapping 0.0044 re_causal 0.0121 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 145, time 247.52, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.5202 re_mapping 0.0047 re_causal 0.0134 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 146, time 248.01, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.4945 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 147, time 247.07, cls_loss 0.0025 cls_loss_mapping 0.0037 cls_loss_causal 0.5273 re_mapping 0.0047 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 148, time 247.44, cls_loss 0.0029 cls_loss_mapping 0.0042 cls_loss_causal 0.5309 re_mapping 0.0046 re_causal 0.0126 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 149, time 246.82, cls_loss 0.0030 cls_loss_mapping 0.0043 cls_loss_causal 0.5280 re_mapping 0.0042 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 150, time 247.07, cls_loss 0.0025 cls_loss_mapping 0.0038 cls_loss_causal 0.5050 re_mapping 0.0042 re_causal 0.0121 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 151, time 247.12, cls_loss 0.0030 cls_loss_mapping 0.0058 cls_loss_causal 0.5175 re_mapping 0.0042 re_causal 0.0118 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 152, time 247.31, cls_loss 0.0028 cls_loss_mapping 0.0039 cls_loss_causal 0.5003 re_mapping 0.0041 re_causal 0.0114 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 153, time 247.30, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.5100 re_mapping 0.0043 re_causal 0.0120 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 154, time 247.18, cls_loss 0.0028 cls_loss_mapping 0.0035 cls_loss_causal 0.5038 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 155, time 247.28, cls_loss 0.0030 cls_loss_mapping 0.0046 cls_loss_causal 0.5092 re_mapping 0.0045 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 156, time 247.53, cls_loss 0.0050 cls_loss_mapping 0.0085 cls_loss_causal 0.5153 re_mapping 0.0044 re_causal 0.0121 /// teacc 98.89 lr 0.00010000 +100 +0.0001 +changing lr +epoch 157, time 247.52, cls_loss 0.0027 cls_loss_mapping 0.0043 cls_loss_causal 0.5363 re_mapping 0.0044 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 158, time 247.53, cls_loss 0.0020 cls_loss_mapping 0.0042 cls_loss_causal 0.4788 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 159, time 247.44, cls_loss 0.0027 cls_loss_mapping 0.0053 cls_loss_causal 0.5289 re_mapping 0.0040 re_causal 0.0117 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 160, time 247.25, cls_loss 0.0031 cls_loss_mapping 0.0043 cls_loss_causal 0.4845 re_mapping 0.0040 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 161, time 247.20, cls_loss 0.0023 cls_loss_mapping 0.0031 cls_loss_causal 0.5342 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 162, time 247.36, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.5377 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 163, time 247.20, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5306 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 164, time 247.35, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5117 re_mapping 0.0042 re_causal 0.0119 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 165, time 247.03, cls_loss 0.0026 cls_loss_mapping 0.0040 cls_loss_causal 0.5038 re_mapping 0.0040 re_causal 0.0114 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 166, time 247.07, cls_loss 0.0025 cls_loss_mapping 0.0041 cls_loss_causal 0.5101 re_mapping 0.0042 re_causal 0.0118 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 167, time 247.31, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.5069 re_mapping 0.0042 re_causal 0.0115 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 168, time 247.20, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5038 re_mapping 0.0041 re_causal 0.0114 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 169, time 247.25, cls_loss 0.0029 cls_loss_mapping 0.0037 cls_loss_causal 0.5111 re_mapping 0.0041 re_causal 0.0109 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 170, time 247.40, cls_loss 0.0030 cls_loss_mapping 0.0044 cls_loss_causal 0.5222 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 171, time 247.19, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.5095 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 172, time 247.11, cls_loss 0.0023 cls_loss_mapping 0.0030 cls_loss_causal 0.5020 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 173, time 246.80, cls_loss 0.0024 cls_loss_mapping 0.0035 cls_loss_causal 0.5326 re_mapping 0.0038 re_causal 0.0111 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 174, time 247.13, cls_loss 0.0024 cls_loss_mapping 0.0026 cls_loss_causal 0.5236 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 175, time 247.24, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.4945 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 176, time 247.09, cls_loss 0.0024 cls_loss_mapping 0.0042 cls_loss_causal 0.5163 re_mapping 0.0039 re_causal 0.0114 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 177, time 246.73, cls_loss 0.0025 cls_loss_mapping 0.0033 cls_loss_causal 0.5106 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 178, time 247.11, cls_loss 0.0022 cls_loss_mapping 0.0025 cls_loss_causal 0.4798 re_mapping 0.0040 re_causal 0.0111 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 179, time 246.93, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.5108 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 180, time 247.21, cls_loss 0.0030 cls_loss_mapping 0.0037 cls_loss_causal 0.5233 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 181, time 247.09, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.5065 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 182, time 246.91, cls_loss 0.0024 cls_loss_mapping 0.0037 cls_loss_causal 0.5588 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 183, time 247.64, cls_loss 0.0026 cls_loss_mapping 0.0037 cls_loss_causal 0.5331 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 184, time 247.11, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5064 re_mapping 0.0039 re_causal 0.0110 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 185, time 247.25, cls_loss 0.0025 cls_loss_mapping 0.0035 cls_loss_causal 0.4997 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 186, time 247.56, cls_loss 0.0023 cls_loss_mapping 0.0033 cls_loss_causal 0.5319 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 187, time 246.96, cls_loss 0.0027 cls_loss_mapping 0.0039 cls_loss_causal 0.5077 re_mapping 0.0035 re_causal 0.0098 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 188, time 247.08, cls_loss 0.0018 cls_loss_mapping 0.0029 cls_loss_causal 0.4799 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 189, time 247.25, cls_loss 0.0017 cls_loss_mapping 0.0028 cls_loss_causal 0.4788 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 190, time 246.87, cls_loss 0.0025 cls_loss_mapping 0.0032 cls_loss_causal 0.4994 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 191, time 247.17, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.4759 re_mapping 0.0039 re_causal 0.0106 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 192, time 246.82, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5043 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 193, time 247.13, cls_loss 0.0025 cls_loss_mapping 0.0043 cls_loss_causal 0.5180 re_mapping 0.0035 re_causal 0.0103 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 194, time 246.74, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5164 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 195, time 246.90, cls_loss 0.0028 cls_loss_mapping 0.0044 cls_loss_causal 0.5003 re_mapping 0.0037 re_causal 0.0105 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 196, time 246.96, cls_loss 0.0024 cls_loss_mapping 0.0043 cls_loss_causal 0.5004 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 197, time 247.11, cls_loss 0.0020 cls_loss_mapping 0.0031 cls_loss_causal 0.4946 re_mapping 0.0038 re_causal 0.0107 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 198, time 247.15, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5043 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 199, time 246.86, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.4853 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 200, time 247.43, cls_loss 0.0021 cls_loss_mapping 0.0037 cls_loss_causal 0.4856 re_mapping 0.0034 re_causal 0.0103 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 201, time 247.24, cls_loss 0.0019 cls_loss_mapping 0.0023 cls_loss_causal 0.5071 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 202, time 247.65, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.5178 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 203, time 247.52, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4825 re_mapping 0.0033 re_causal 0.0094 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 204, time 247.32, cls_loss 0.0018 cls_loss_mapping 0.0021 cls_loss_causal 0.4940 re_mapping 0.0036 re_causal 0.0101 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 205, time 247.61, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.5107 re_mapping 0.0039 re_causal 0.0105 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 206, time 247.53, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.4936 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 207, time 247.27, cls_loss 0.0024 cls_loss_mapping 0.0033 cls_loss_causal 0.4938 re_mapping 0.0033 re_causal 0.0095 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 208, time 247.31, cls_loss 0.0020 cls_loss_mapping 0.0024 cls_loss_causal 0.4524 re_mapping 0.0035 re_causal 0.0101 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 209, time 247.73, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4973 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 210, time 247.40, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4702 re_mapping 0.0036 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 211, time 246.89, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5196 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 212, time 247.35, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.4901 re_mapping 0.0034 re_causal 0.0096 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 213, time 247.22, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5019 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 214, time 247.39, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4779 re_mapping 0.0034 re_causal 0.0100 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 215, time 247.27, cls_loss 0.0024 cls_loss_mapping 0.0040 cls_loss_causal 0.4916 re_mapping 0.0033 re_causal 0.0095 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 216, time 247.37, cls_loss 0.0016 cls_loss_mapping 0.0017 cls_loss_causal 0.4976 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 217, time 247.45, cls_loss 0.0023 cls_loss_mapping 0.0035 cls_loss_causal 0.4677 re_mapping 0.0034 re_causal 0.0095 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 218, time 247.74, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4740 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 219, time 247.00, cls_loss 0.0020 cls_loss_mapping 0.0024 cls_loss_causal 0.4902 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 220, time 247.26, cls_loss 0.0015 cls_loss_mapping 0.0015 cls_loss_causal 0.4984 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 221, time 247.21, cls_loss 0.0018 cls_loss_mapping 0.0018 cls_loss_causal 0.4791 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 222, time 247.27, cls_loss 0.0019 cls_loss_mapping 0.0025 cls_loss_causal 0.4897 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 223, time 247.18, cls_loss 0.0022 cls_loss_mapping 0.0027 cls_loss_causal 0.5187 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 224, time 247.18, cls_loss 0.0020 cls_loss_mapping 0.0025 cls_loss_causal 0.4952 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 225, time 247.25, cls_loss 0.0021 cls_loss_mapping 0.0027 cls_loss_causal 0.4951 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 226, time 247.42, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.5013 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 227, time 248.14, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.5144 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 228, time 247.56, cls_loss 0.0020 cls_loss_mapping 0.0027 cls_loss_causal 0.5000 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 229---------------------------------------------------- +epoch 229, time 247.85, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5045 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.21 lr 0.00010000 +100 +0.0001 +changing lr +epoch 230, time 247.45, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5028 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 231, time 247.60, cls_loss 0.0024 cls_loss_mapping 0.0033 cls_loss_causal 0.5090 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 232, time 247.14, cls_loss 0.0024 cls_loss_mapping 0.0037 cls_loss_causal 0.4987 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.18 lr 0.00010000 +100 +0.0001 +changing lr +epoch 233, time 247.34, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5306 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 234, time 247.70, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4880 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 235, time 247.42, cls_loss 0.0017 cls_loss_mapping 0.0020 cls_loss_causal 0.4734 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 236, time 248.01, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.4746 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 237, time 247.40, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4826 re_mapping 0.0029 re_causal 0.0088 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 238, time 247.38, cls_loss 0.0017 cls_loss_mapping 0.0016 cls_loss_causal 0.5047 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 239, time 248.73, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.5064 re_mapping 0.0030 re_causal 0.0092 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 240, time 248.15, cls_loss 0.0018 cls_loss_mapping 0.0034 cls_loss_causal 0.5029 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 241, time 247.23, cls_loss 0.0017 cls_loss_mapping 0.0023 cls_loss_causal 0.4986 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 242, time 247.18, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.4912 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 243, time 247.37, cls_loss 0.0017 cls_loss_mapping 0.0024 cls_loss_causal 0.4714 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 244, time 247.36, cls_loss 0.0018 cls_loss_mapping 0.0030 cls_loss_causal 0.4707 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 245, time 247.07, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4907 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 246, time 247.63, cls_loss 0.0017 cls_loss_mapping 0.0039 cls_loss_causal 0.5042 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 247, time 247.28, cls_loss 0.0026 cls_loss_mapping 0.0037 cls_loss_causal 0.4860 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 248, time 247.15, cls_loss 0.0021 cls_loss_mapping 0.0031 cls_loss_causal 0.4790 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 249, time 247.23, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4878 re_mapping 0.0031 re_causal 0.0088 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 250, time 247.02, cls_loss 0.0015 cls_loss_mapping 0.0016 cls_loss_causal 0.4962 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 251, time 247.51, cls_loss 0.0018 cls_loss_mapping 0.0028 cls_loss_causal 0.4979 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 252, time 247.35, cls_loss 0.0020 cls_loss_mapping 0.0024 cls_loss_causal 0.4525 re_mapping 0.0031 re_causal 0.0088 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 253, time 247.04, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4552 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 254, time 247.26, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.4710 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 255, time 247.16, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4512 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 256, time 247.38, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4975 re_mapping 0.0028 re_causal 0.0086 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 257, time 247.20, cls_loss 0.0016 cls_loss_mapping 0.0018 cls_loss_causal 0.4632 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 258, time 247.12, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4664 re_mapping 0.0032 re_causal 0.0091 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 259, time 247.32, cls_loss 0.0018 cls_loss_mapping 0.0025 cls_loss_causal 0.4997 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 260, time 247.19, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4738 re_mapping 0.0029 re_causal 0.0088 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 261, time 246.87, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4996 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 262, time 246.91, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5120 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 263, time 247.18, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4762 re_mapping 0.0031 re_causal 0.0087 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 264, time 246.94, cls_loss 0.0015 cls_loss_mapping 0.0019 cls_loss_causal 0.4728 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 265, time 246.92, cls_loss 0.0017 cls_loss_mapping 0.0016 cls_loss_causal 0.4729 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 266, time 246.97, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.4830 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 267, time 247.19, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.4905 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 268, time 247.41, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4688 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 269, time 247.14, cls_loss 0.0021 cls_loss_mapping 0.0027 cls_loss_causal 0.5079 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.18 lr 0.00010000 +100 +0.0001 +changing lr +epoch 270, time 247.08, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4751 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 271, time 247.23, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.4807 re_mapping 0.0029 re_causal 0.0087 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 272, time 247.25, cls_loss 0.0021 cls_loss_mapping 0.0029 cls_loss_causal 0.4811 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 273, time 247.12, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4693 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 274, time 247.52, cls_loss 0.0018 cls_loss_mapping 0.0019 cls_loss_causal 0.4625 re_mapping 0.0030 re_causal 0.0081 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 275, time 247.32, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4594 re_mapping 0.0030 re_causal 0.0086 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 276, time 247.31, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.4717 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 277, time 247.59, cls_loss 0.0021 cls_loss_mapping 0.0022 cls_loss_causal 0.4800 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 278, time 247.25, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.4832 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 279, time 247.30, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4871 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 280---------------------------------------------------- +epoch 280, time 248.43, cls_loss 0.0016 cls_loss_mapping 0.0017 cls_loss_causal 0.4657 re_mapping 0.0030 re_causal 0.0084 /// teacc 99.26 lr 0.00010000 +100 +0.0001 +changing lr +epoch 281, time 247.76, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4639 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.25 lr 0.00010000 +100 +0.0001 +changing lr +epoch 282, time 247.74, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.4466 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 283, time 247.60, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4532 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 284, time 247.58, cls_loss 0.0020 cls_loss_mapping 0.0031 cls_loss_causal 0.4614 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 285, time 247.57, cls_loss 0.0022 cls_loss_mapping 0.0026 cls_loss_causal 0.5009 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 286, time 247.67, cls_loss 0.0017 cls_loss_mapping 0.0019 cls_loss_causal 0.4442 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.15 lr 0.00010000 +100 +0.0001 +changing lr +epoch 287, time 247.52, cls_loss 0.0018 cls_loss_mapping 0.0041 cls_loss_causal 0.4619 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 288, time 246.94, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4668 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 289, time 247.41, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.4698 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 290, time 247.31, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4558 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 291, time 247.85, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4896 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 292, time 247.58, cls_loss 0.0018 cls_loss_mapping 0.0017 cls_loss_causal 0.4845 re_mapping 0.0025 re_causal 0.0075 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 293, time 246.85, cls_loss 0.0018 cls_loss_mapping 0.0019 cls_loss_causal 0.4797 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 294, time 247.06, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4528 re_mapping 0.0029 re_causal 0.0080 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 295---------------------------------------------------- +epoch 295, time 248.20, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4663 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.27 lr 0.00010000 +100 +0.0001 +changing lr +epoch 296, time 247.05, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4457 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.19 lr 0.00010000 +100 +0.0001 +changing lr +epoch 297, time 247.01, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4646 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 298, time 247.03, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4582 re_mapping 0.0027 re_causal 0.0081 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 299, time 247.08, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.4958 re_mapping 0.0026 re_causal 0.0083 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 300, time 247.04, cls_loss 0.0012 cls_loss_mapping 0.0012 cls_loss_causal 0.4689 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 301, time 247.00, cls_loss 0.0016 cls_loss_mapping 0.0018 cls_loss_causal 0.4784 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 302, time 246.87, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4964 re_mapping 0.0026 re_causal 0.0082 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 303, time 246.97, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.4547 re_mapping 0.0027 re_causal 0.0077 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 304, time 246.93, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4448 re_mapping 0.0027 re_causal 0.0081 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 305, time 247.00, cls_loss 0.0011 cls_loss_mapping 0.0009 cls_loss_causal 0.4617 re_mapping 0.0027 re_causal 0.0080 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 306, time 246.97, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4790 re_mapping 0.0027 re_causal 0.0085 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 307, time 247.08, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4900 re_mapping 0.0026 re_causal 0.0081 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 308, time 246.70, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4600 re_mapping 0.0026 re_causal 0.0078 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 309, time 246.77, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4756 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 310, time 247.15, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.4717 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 311, time 247.54, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.4607 re_mapping 0.0027 re_causal 0.0078 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 312, time 247.41, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4517 re_mapping 0.0027 re_causal 0.0080 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 313, time 247.18, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4551 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 314, time 247.40, cls_loss 0.0012 cls_loss_mapping 0.0013 cls_loss_causal 0.4355 re_mapping 0.0028 re_causal 0.0079 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 315, time 247.58, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4555 re_mapping 0.0026 re_causal 0.0075 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 316, time 247.22, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4448 re_mapping 0.0026 re_causal 0.0075 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 317, time 247.06, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4914 re_mapping 0.0027 re_causal 0.0078 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 318, time 246.94, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.4779 re_mapping 0.0027 re_causal 0.0080 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 319, time 247.06, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4348 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 320, time 247.23, cls_loss 0.0015 cls_loss_mapping 0.0018 cls_loss_causal 0.4390 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 321, time 246.84, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4578 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 322, time 247.00, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4698 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 323, time 247.23, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4447 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 324, time 246.90, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4695 re_mapping 0.0028 re_causal 0.0077 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 325, time 246.85, cls_loss 0.0016 cls_loss_mapping 0.0019 cls_loss_causal 0.4536 re_mapping 0.0027 re_causal 0.0078 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 326, time 247.13, cls_loss 0.0018 cls_loss_mapping 0.0017 cls_loss_causal 0.4503 re_mapping 0.0026 re_causal 0.0073 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 327, time 246.92, cls_loss 0.0014 cls_loss_mapping 0.0014 cls_loss_causal 0.4610 re_mapping 0.0027 re_causal 0.0078 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 328, time 247.01, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4952 re_mapping 0.0026 re_causal 0.0081 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 329, time 247.26, cls_loss 0.0012 cls_loss_mapping 0.0013 cls_loss_causal 0.4556 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 330, time 247.18, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4647 re_mapping 0.0024 re_causal 0.0073 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 331, time 247.30, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4686 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 332, time 247.31, cls_loss 0.0012 cls_loss_mapping 0.0011 cls_loss_causal 0.4722 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 333, time 247.49, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4423 re_mapping 0.0026 re_causal 0.0077 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 334, time 247.27, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4602 re_mapping 0.0023 re_causal 0.0074 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 335, time 247.17, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4384 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 336, time 247.21, cls_loss 0.0021 cls_loss_mapping 0.0031 cls_loss_causal 0.4611 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.20 lr 0.00010000 +100 +0.0001 +changing lr +epoch 337, time 247.10, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4444 re_mapping 0.0025 re_causal 0.0077 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 338, time 247.26, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4533 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 339, time 247.32, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4566 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 340, time 247.16, cls_loss 0.0010 cls_loss_mapping 0.0010 cls_loss_causal 0.4598 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 341, time 247.15, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4526 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 342, time 247.51, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.5016 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 343, time 247.38, cls_loss 0.0015 cls_loss_mapping 0.0015 cls_loss_causal 0.4960 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 344, time 247.23, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4559 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 345, time 247.50, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4610 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 346, time 247.38, cls_loss 0.0017 cls_loss_mapping 0.0023 cls_loss_causal 0.4869 re_mapping 0.0025 re_causal 0.0075 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 347, time 247.37, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4465 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 348, time 247.38, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4634 re_mapping 0.0026 re_causal 0.0077 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 349, time 247.41, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4596 re_mapping 0.0025 re_causal 0.0073 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 350, time 246.99, cls_loss 0.0013 cls_loss_mapping 0.0013 cls_loss_causal 0.4557 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 351, time 247.25, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4623 re_mapping 0.0025 re_causal 0.0075 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 352, time 247.18, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4614 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.15 lr 0.00010000 +100 +0.0001 +changing lr +epoch 353, time 246.88, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4559 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 354, time 247.59, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4315 re_mapping 0.0025 re_causal 0.0079 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 355, time 247.21, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4696 re_mapping 0.0025 re_causal 0.0078 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 356, time 247.56, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.4666 re_mapping 0.0025 re_causal 0.0077 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 357, time 247.29, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4475 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 358, time 247.22, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4710 re_mapping 0.0025 re_causal 0.0077 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 359, time 247.26, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4392 re_mapping 0.0024 re_causal 0.0073 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 360, time 247.41, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4256 re_mapping 0.0024 re_causal 0.0071 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 361, time 247.29, cls_loss 0.0011 cls_loss_mapping 0.0011 cls_loss_causal 0.4301 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 362, time 247.23, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4589 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 363, time 247.09, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4652 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 364, time 246.92, cls_loss 0.0014 cls_loss_mapping 0.0011 cls_loss_causal 0.4869 re_mapping 0.0024 re_causal 0.0071 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 365, time 247.20, cls_loss 0.0020 cls_loss_mapping 0.0026 cls_loss_causal 0.4712 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 366, time 247.23, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4724 re_mapping 0.0025 re_causal 0.0078 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 367, time 247.13, cls_loss 0.0015 cls_loss_mapping 0.0021 cls_loss_causal 0.4755 re_mapping 0.0023 re_causal 0.0072 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 368, time 247.27, cls_loss 0.0020 cls_loss_mapping 0.0022 cls_loss_causal 0.4718 re_mapping 0.0024 re_causal 0.0073 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 369, time 247.12, cls_loss 0.0019 cls_loss_mapping 0.0024 cls_loss_causal 0.4716 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 370, time 246.83, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4717 re_mapping 0.0026 re_causal 0.0074 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 371, time 246.80, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4637 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 372, time 246.84, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4744 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 373, time 246.58, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4420 re_mapping 0.0024 re_causal 0.0071 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 374, time 246.67, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4422 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 375, time 246.70, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4266 re_mapping 0.0024 re_causal 0.0072 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 376, time 247.08, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4453 re_mapping 0.0023 re_causal 0.0071 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 377, time 247.12, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4735 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 378, time 247.23, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4365 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 379, time 247.06, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4635 re_mapping 0.0023 re_causal 0.0070 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 380, time 247.45, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4711 re_mapping 0.0024 re_causal 0.0078 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 381, time 247.57, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4574 re_mapping 0.0022 re_causal 0.0069 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 382, time 247.44, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4559 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 383, time 247.33, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.4570 re_mapping 0.0026 re_causal 0.0074 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 384, time 247.28, cls_loss 0.0020 cls_loss_mapping 0.0025 cls_loss_causal 0.4635 re_mapping 0.0024 re_causal 0.0074 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 385, time 247.44, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4457 re_mapping 0.0024 re_causal 0.0070 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 386, time 247.08, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4753 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 387, time 247.01, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4634 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 388, time 247.20, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4897 re_mapping 0.0023 re_causal 0.0072 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 389, time 247.17, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4658 re_mapping 0.0022 re_causal 0.0068 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 390, time 247.07, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4615 re_mapping 0.0023 re_causal 0.0072 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 391, time 247.09, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4522 re_mapping 0.0022 re_causal 0.0068 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 392, time 247.11, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4690 re_mapping 0.0023 re_causal 0.0074 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 393, time 246.80, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4457 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 394, time 247.16, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4379 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 395, time 247.64, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4345 re_mapping 0.0024 re_causal 0.0074 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 396, time 247.37, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4401 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 397, time 247.54, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4345 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 398, time 247.34, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4310 re_mapping 0.0023 re_causal 0.0071 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 399, time 246.86, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4484 re_mapping 0.0023 re_causal 0.0074 /// teacc 99.11 lr 0.00001000 +100 +1e-05 +changing lr +epoch 400, time 247.11, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4272 re_mapping 0.0023 re_causal 0.0071 /// teacc 99.14 lr 0.00001000 +100 +1e-05 +changing lr +epoch 401, time 247.02, cls_loss 0.0009 cls_loss_mapping 0.0009 cls_loss_causal 0.4241 re_mapping 0.0022 re_causal 0.0068 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 402, time 247.20, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4582 re_mapping 0.0020 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +100 +1e-05 +changing lr +epoch 403, time 247.43, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4785 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 404, time 246.87, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4336 re_mapping 0.0020 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 405, time 247.03, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4119 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.18 lr 0.00001000 +100 +1e-05 +changing lr +epoch 406, time 247.20, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4492 re_mapping 0.0020 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +100 +1e-05 +changing lr +epoch 407, time 247.14, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4428 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 408, time 246.95, cls_loss 0.0008 cls_loss_mapping 0.0006 cls_loss_causal 0.4292 re_mapping 0.0019 re_causal 0.0065 /// teacc 99.19 lr 0.00001000 +100 +1e-05 +changing lr +epoch 409, time 247.07, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4267 re_mapping 0.0019 re_causal 0.0063 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 410, time 246.79, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4532 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 411, time 247.28, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4191 re_mapping 0.0018 re_causal 0.0065 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 412, time 247.11, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4227 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 413, time 247.24, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4239 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 414, time 246.99, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4236 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 415, time 247.09, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4173 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 416, time 246.96, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4130 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 417, time 247.21, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4219 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 418, time 247.16, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4107 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 419, time 247.16, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4217 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 420, time 247.03, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4252 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 421, time 246.83, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4170 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 422, time 247.05, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4611 re_mapping 0.0017 re_causal 0.0067 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 423, time 247.12, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4178 re_mapping 0.0017 re_causal 0.0064 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 424, time 246.46, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4336 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 425, time 246.57, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4000 re_mapping 0.0017 re_causal 0.0060 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 426, time 246.61, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4221 re_mapping 0.0017 re_causal 0.0062 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 427, time 246.74, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4261 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 428, time 247.08, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4231 re_mapping 0.0017 re_causal 0.0062 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 429, time 247.14, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4139 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 430, time 247.05, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4485 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 431, time 246.93, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.3967 re_mapping 0.0016 re_causal 0.0059 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 432, time 246.91, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4022 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 433, time 246.92, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4053 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 434, time 246.75, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4163 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 435, time 247.10, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4253 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 436, time 247.14, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4303 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 437, time 246.97, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4144 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 438, time 247.00, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4147 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 439, time 247.17, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4282 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 440, time 246.86, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.3817 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 441, time 246.73, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4127 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 442, time 247.03, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4147 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 443, time 247.06, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4173 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 444, time 246.70, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4060 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 445, time 246.69, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3875 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +---------------------saving model at epoch 446---------------------------------------------------- +epoch 446, time 247.46, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4331 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 447, time 246.73, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4148 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 448, time 247.05, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4421 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 449, time 246.60, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3929 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +---------------------saving model at epoch 450---------------------------------------------------- +epoch 450, time 247.44, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4173 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.31 lr 0.00001000 +100 +1e-05 +changing lr +epoch 451, time 246.85, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4352 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 452, time 246.56, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4248 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 453, time 246.71, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4208 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 454, time 247.02, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4349 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 455, time 246.71, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4133 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 456, time 246.84, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4097 re_mapping 0.0015 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 457, time 246.87, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3905 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 458, time 246.86, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.3884 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 459, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3983 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 460, time 246.95, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4410 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 461, time 246.76, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4042 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 462, time 247.25, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4222 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 463, time 247.13, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4288 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 464, time 247.10, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4350 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 465, time 247.08, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4409 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.29 lr 0.00001000 +100 +1e-05 +changing lr +epoch 466, time 246.90, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.3901 re_mapping 0.0015 re_causal 0.0056 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 467, time 247.04, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4076 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 468, time 246.84, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.3978 re_mapping 0.0015 re_causal 0.0056 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 469, time 246.98, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4259 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 470, time 246.99, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4459 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 471, time 247.05, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4229 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 472, time 246.98, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4319 re_mapping 0.0014 re_causal 0.0058 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 473, time 246.98, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4292 re_mapping 0.0014 re_causal 0.0058 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 474, time 247.17, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4197 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 475, time 247.03, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.3734 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.29 lr 0.00001000 +100 +1e-05 +changing lr +epoch 476, time 247.30, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3885 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 477, time 248.54, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4097 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 478, time 247.02, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4113 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 479, time 247.12, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.3727 re_mapping 0.0014 re_causal 0.0054 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 480, time 246.85, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4400 re_mapping 0.0015 re_causal 0.0060 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 481, time 246.83, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4142 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 482, time 246.97, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4060 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 483, time 246.78, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4004 re_mapping 0.0014 re_causal 0.0055 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 484, time 246.96, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4422 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 485, time 246.70, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3912 re_mapping 0.0014 re_causal 0.0055 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 486, time 247.02, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3976 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 487, time 246.77, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4313 re_mapping 0.0015 re_causal 0.0060 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 488, time 246.95, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4221 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 489, time 246.76, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4082 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 490, time 246.83, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4112 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 491, time 246.79, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3886 re_mapping 0.0015 re_causal 0.0055 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 492, time 246.78, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4168 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 493, time 246.88, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4221 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 494, time 246.71, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4312 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.19 lr 0.00001000 +100 +1e-05 +changing lr +epoch 495, time 246.90, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4413 re_mapping 0.0014 re_causal 0.0059 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 496, time 246.63, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4036 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 497, time 246.91, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4220 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.18 lr 0.00001000 +100 +1e-05 +changing lr +epoch 498, time 247.19, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4161 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.17 lr 0.00001000 +100 +1e-05 +changing lr +epoch 499, time 247.07, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3912 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.19 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_ReProduceMetaCausal', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_ReProduceMetaCausal/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat diff --git a/Meta-causal/code-withStyleAttack/56720.error b/Meta-causal/code-withStyleAttack/56720.error new file mode 100644 index 0000000000000000000000000000000000000000..4c741962fda5fd145618ae7373555295b05ff9de --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56720.error @@ -0,0 +1 @@ +run_my_joint_v13_test.sh: line 28: andn}: command not found diff --git a/Meta-causal/code-withStyleAttack/56720.log b/Meta-causal/code-withStyleAttack/56720.log new file mode 100644 index 0000000000000000000000000000000000000000..5e281586589ea1303704167b5264e28ec7b696f5 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56720.log @@ -0,0 +1,336 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'cartoon', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_train.hdf5 torch.Size([2107, 3, 227, 227]) torch.Size([2107]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_val.hdf5 torch.Size([237, 3, 227, 227]) torch.Size([237]) +-------------------------------------loading pretrain weights---------------------------------- +351 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 500.68, cls_loss 5.0126 cls_loss_mapping 1.4019 cls_loss_causal 1.7210 re_mapping 1.0578 re_causal 1.0584 /// teacc 83.12 lr 0.00999497 +351 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 530.56, cls_loss 2.0306 cls_loss_mapping 0.7946 cls_loss_causal 1.3288 re_mapping 0.6527 re_causal 0.6538 /// teacc 87.76 lr 0.00997987 +351 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 536.46, cls_loss 0.6382 cls_loss_mapping 0.4834 cls_loss_causal 1.1278 re_mapping 0.3952 re_causal 0.3957 /// teacc 91.98 lr 0.00995475 +351 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 514.73, cls_loss 0.2115 cls_loss_mapping 0.3043 cls_loss_causal 0.9479 re_mapping 0.2605 re_causal 0.2608 /// teacc 92.41 lr 0.00991965 +351 +0.009919647942993149 +changing lr +epoch 4, time 518.58, cls_loss 0.1048 cls_loss_mapping 0.2504 cls_loss_causal 0.8913 re_mapping 0.2075 re_causal 0.2080 /// teacc 92.41 lr 0.00987464 +351 +0.009874639560909117 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 522.88, cls_loss 0.0517 cls_loss_mapping 0.2038 cls_loss_causal 0.8571 re_mapping 0.1746 re_causal 0.1753 /// teacc 95.36 lr 0.00981981 +351 +0.009819814303479266 +changing lr +epoch 6, time 515.51, cls_loss 0.0244 cls_loss_mapping 0.1830 cls_loss_causal 0.7905 re_mapping 0.1502 re_causal 0.1512 /// teacc 94.51 lr 0.00975528 +351 +0.009755282581475767 +changing lr +epoch 7, time 516.68, cls_loss 0.0226 cls_loss_mapping 0.1536 cls_loss_causal 0.7386 re_mapping 0.1335 re_causal 0.1347 /// teacc 94.94 lr 0.00968117 +351 +0.009681174353198686 +changing lr +epoch 8, time 512.83, cls_loss 0.0311 cls_loss_mapping 0.1488 cls_loss_causal 0.7284 re_mapping 0.1200 re_causal 0.1218 /// teacc 91.56 lr 0.00959764 +351 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 515.75, cls_loss 0.0257 cls_loss_mapping 0.1258 cls_loss_causal 0.7038 re_mapping 0.1090 re_causal 0.1110 /// teacc 95.78 lr 0.00950484 +351 +0.009504844339512096 +changing lr +epoch 10, time 508.97, cls_loss 0.0086 cls_loss_mapping 0.1049 cls_loss_causal 0.7078 re_mapping 0.0973 re_causal 0.0997 /// teacc 94.09 lr 0.00940298 +351 +0.009402977659283692 +changing lr +epoch 11, time 522.72, cls_loss 0.0121 cls_loss_mapping 0.1017 cls_loss_causal 0.6880 re_mapping 0.0899 re_causal 0.0929 /// teacc 95.36 lr 0.00929224 +351 +0.009292243968009333 +changing lr +epoch 12, time 511.13, cls_loss 0.0138 cls_loss_mapping 0.0946 cls_loss_causal 0.7011 re_mapping 0.0820 re_causal 0.0855 /// teacc 94.94 lr 0.00917287 +351 +0.009172866268606516 +changing lr +epoch 13, time 522.62, cls_loss 0.0104 cls_loss_mapping 0.0844 cls_loss_causal 0.6675 re_mapping 0.0747 re_causal 0.0784 /// teacc 94.51 lr 0.00904508 +351 +0.00904508497187474 +changing lr +epoch 14, time 530.18, cls_loss 0.0122 cls_loss_mapping 0.0736 cls_loss_causal 0.6363 re_mapping 0.0698 re_causal 0.0745 /// teacc 95.78 lr 0.00890916 +351 +0.008909157412340152 +changing lr +epoch 15, time 513.03, cls_loss 0.0108 cls_loss_mapping 0.0735 cls_loss_causal 0.6055 re_mapping 0.0623 re_causal 0.0673 /// teacc 94.94 lr 0.00876536 +351 +0.00876535733001806 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 515.62, cls_loss 0.0097 cls_loss_mapping 0.0626 cls_loss_causal 0.6328 re_mapping 0.0572 re_causal 0.0629 /// teacc 97.05 lr 0.00861397 +351 +0.008613974319136962 +changing lr +epoch 17, time 534.26, cls_loss 0.0145 cls_loss_mapping 0.0706 cls_loss_causal 0.6484 re_mapping 0.0533 re_causal 0.0603 /// teacc 96.20 lr 0.00845531 +351 +0.008455313244934327 +changing lr +epoch 18, time 532.03, cls_loss 0.0106 cls_loss_mapping 0.0571 cls_loss_causal 0.5705 re_mapping 0.0492 re_causal 0.0567 /// teacc 96.62 lr 0.00828969 +351 +0.008289693629698565 +changing lr +epoch 19, time 518.07, cls_loss 0.0076 cls_loss_mapping 0.0474 cls_loss_causal 0.5525 re_mapping 0.0441 re_causal 0.0513 /// teacc 95.78 lr 0.00811745 +351 +0.00811744900929367 +changing lr +epoch 20, time 530.26, cls_loss 0.0081 cls_loss_mapping 0.0546 cls_loss_causal 0.5926 re_mapping 0.0409 re_causal 0.0491 /// teacc 97.05 lr 0.00793893 +351 +0.007938926261462368 +changing lr +epoch 21, time 534.00, cls_loss 0.0104 cls_loss_mapping 0.0511 cls_loss_causal 0.5469 re_mapping 0.0373 re_causal 0.0451 /// teacc 95.78 lr 0.00775448 +351 +0.007754484907260515 +changing lr +epoch 22, time 534.87, cls_loss 0.0148 cls_loss_mapping 0.0474 cls_loss_causal 0.5694 re_mapping 0.0353 re_causal 0.0430 /// teacc 95.36 lr 0.00756450 +351 +0.007564496387029534 +changing lr +epoch 23, time 515.04, cls_loss 0.0053 cls_loss_mapping 0.0395 cls_loss_causal 0.5557 re_mapping 0.0324 re_causal 0.0409 /// teacc 96.62 lr 0.00736934 +351 +0.007369343312364995 +changing lr +epoch 24, time 527.73, cls_loss 0.0083 cls_loss_mapping 0.0487 cls_loss_causal 0.5594 re_mapping 0.0306 re_causal 0.0402 /// teacc 94.94 lr 0.00716942 +351 +0.0071694186955877925 +changing lr +epoch 25, time 521.10, cls_loss 0.0080 cls_loss_mapping 0.0392 cls_loss_causal 0.5600 re_mapping 0.0291 re_causal 0.0390 /// teacc 97.05 lr 0.00696513 +351 +0.0069651251582696205 +changing lr +epoch 26, time 528.30, cls_loss 0.0054 cls_loss_mapping 0.0316 cls_loss_causal 0.5380 re_mapping 0.0270 re_causal 0.0366 /// teacc 96.20 lr 0.00675687 +351 +0.006756874120406716 +changing lr +epoch 27, time 526.16, cls_loss 0.0075 cls_loss_mapping 0.0354 cls_loss_causal 0.5384 re_mapping 0.0251 re_causal 0.0347 /// teacc 97.05 lr 0.00654508 +351 +0.00654508497187474 +changing lr +epoch 28, time 520.46, cls_loss 0.0066 cls_loss_mapping 0.0281 cls_loss_causal 0.5043 re_mapping 0.0240 re_causal 0.0354 /// teacc 96.62 lr 0.00633018 +351 +0.006330184227833378 +changing lr +epoch 29, time 536.80, cls_loss 0.0074 cls_loss_mapping 0.0305 cls_loss_causal 0.5296 re_mapping 0.0227 re_causal 0.0341 /// teacc 95.78 lr 0.00611260 +351 +0.006112604669781575 +changing lr +epoch 30, time 530.96, cls_loss 0.0062 cls_loss_mapping 0.0301 cls_loss_causal 0.5251 re_mapping 0.0214 re_causal 0.0317 /// teacc 96.20 lr 0.00589278 +351 +0.005892784473993186 +changing lr +epoch 31, time 528.74, cls_loss 0.0051 cls_loss_mapping 0.0263 cls_loss_causal 0.5350 re_mapping 0.0205 re_causal 0.0317 /// teacc 96.20 lr 0.00567117 +351 +0.00567116632908828 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 517.42, cls_loss 0.0051 cls_loss_mapping 0.0225 cls_loss_causal 0.5060 re_mapping 0.0197 re_causal 0.0305 /// teacc 97.47 lr 0.00544820 +351 +0.00544819654451717 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 532.29, cls_loss 0.0050 cls_loss_mapping 0.0196 cls_loss_causal 0.5099 re_mapping 0.0185 re_causal 0.0291 /// teacc 97.89 lr 0.00522432 +351 +0.005224324151752577 +changing lr +epoch 34, time 521.23, cls_loss 0.0079 cls_loss_mapping 0.0235 cls_loss_causal 0.5058 re_mapping 0.0177 re_causal 0.0285 /// teacc 97.89 lr 0.00500000 +351 +0.005000000000000003 +changing lr +epoch 35, time 521.42, cls_loss 0.0054 cls_loss_mapping 0.0236 cls_loss_causal 0.4683 re_mapping 0.0178 re_causal 0.0281 /// teacc 97.05 lr 0.00477568 +351 +0.004775675848247429 +changing lr +epoch 36, time 526.29, cls_loss 0.0057 cls_loss_mapping 0.0231 cls_loss_causal 0.5159 re_mapping 0.0172 re_causal 0.0278 /// teacc 97.05 lr 0.00455180 +351 +0.004551803455482836 +changing lr +epoch 37, time 535.59, cls_loss 0.0063 cls_loss_mapping 0.0199 cls_loss_causal 0.4658 re_mapping 0.0163 re_causal 0.0267 /// teacc 97.47 lr 0.00432883 +351 +0.004328833670911726 +changing lr +epoch 38, time 512.58, cls_loss 0.0045 cls_loss_mapping 0.0199 cls_loss_causal 0.4925 re_mapping 0.0155 re_causal 0.0258 /// teacc 97.05 lr 0.00410722 +351 +0.0041072155260068206 +changing lr +epoch 39, time 532.69, cls_loss 0.0056 cls_loss_mapping 0.0220 cls_loss_causal 0.4772 re_mapping 0.0150 re_causal 0.0253 /// teacc 97.47 lr 0.00388740 +351 +0.0038873953302184317 +changing lr +epoch 40, time 536.18, cls_loss 0.0044 cls_loss_mapping 0.0185 cls_loss_causal 0.4992 re_mapping 0.0146 re_causal 0.0241 /// teacc 97.47 lr 0.00366982 +351 +0.003669815772166629 +changing lr +epoch 41, time 531.87, cls_loss 0.0044 cls_loss_mapping 0.0147 cls_loss_causal 0.4840 re_mapping 0.0144 re_causal 0.0246 /// teacc 97.89 lr 0.00345492 +351 +0.0034549150281252667 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 509.65, cls_loss 0.0045 cls_loss_mapping 0.0164 cls_loss_causal 0.4600 re_mapping 0.0136 re_causal 0.0224 /// teacc 98.31 lr 0.00324313 +351 +0.0032431258795932905 +changing lr +epoch 43, time 520.56, cls_loss 0.0051 cls_loss_mapping 0.0169 cls_loss_causal 0.5021 re_mapping 0.0137 re_causal 0.0235 /// teacc 97.47 lr 0.00303487 +351 +0.0030348748417303863 +changing lr +---------------------saving model at epoch 44---------------------------------------------------- +epoch 44, time 532.35, cls_loss 0.0042 cls_loss_mapping 0.0153 cls_loss_causal 0.4512 re_mapping 0.0131 re_causal 0.0230 /// teacc 98.73 lr 0.00283058 +351 +0.0028305813044122124 +changing lr +epoch 45, time 523.83, cls_loss 0.0053 cls_loss_mapping 0.0159 cls_loss_causal 0.4523 re_mapping 0.0130 re_causal 0.0219 /// teacc 97.89 lr 0.00263066 +351 +0.0026306566876350096 +changing lr +epoch 46, time 536.05, cls_loss 0.0050 cls_loss_mapping 0.0148 cls_loss_causal 0.4521 re_mapping 0.0125 re_causal 0.0215 /// teacc 96.62 lr 0.00243550 +351 +0.0024355036129704724 +changing lr +epoch 47, time 509.13, cls_loss 0.0043 cls_loss_mapping 0.0159 cls_loss_causal 0.4864 re_mapping 0.0121 re_causal 0.0214 /// teacc 97.89 lr 0.00224552 +351 +0.00224551509273949 +changing lr +epoch 48, time 524.58, cls_loss 0.0037 cls_loss_mapping 0.0109 cls_loss_causal 0.4474 re_mapping 0.0120 re_causal 0.0208 /// teacc 98.31 lr 0.00206107 +351 +0.002061073738537637 +changing lr +epoch 49, time 517.27, cls_loss 0.0033 cls_loss_mapping 0.0125 cls_loss_causal 0.4527 re_mapping 0.0117 re_causal 0.0205 /// teacc 97.89 lr 0.00188255 +351 +0.0018825509907063344 +changing lr +epoch 50, time 516.76, cls_loss 0.0039 cls_loss_mapping 0.0142 cls_loss_causal 0.4602 re_mapping 0.0116 re_causal 0.0204 /// teacc 97.47 lr 0.00171031 +351 +0.0017103063703014388 +changing lr +epoch 51, time 513.81, cls_loss 0.0025 cls_loss_mapping 0.0098 cls_loss_causal 0.4081 re_mapping 0.0116 re_causal 0.0197 /// teacc 98.31 lr 0.00154469 +351 +0.0015446867550656784 +changing lr +epoch 52, time 514.01, cls_loss 0.0042 cls_loss_mapping 0.0125 cls_loss_causal 0.4603 re_mapping 0.0114 re_causal 0.0195 /// teacc 97.89 lr 0.00138603 +351 +0.001386025680863044 +changing lr +epoch 53, time 524.35, cls_loss 0.0051 cls_loss_mapping 0.0127 cls_loss_causal 0.4572 re_mapping 0.0111 re_causal 0.0193 /// teacc 97.89 lr 0.00123464 +351 +0.0012346426699819469 +changing lr +epoch 54, time 514.44, cls_loss 0.0044 cls_loss_mapping 0.0127 cls_loss_causal 0.4353 re_mapping 0.0111 re_causal 0.0187 /// teacc 97.47 lr 0.00109084 +351 +0.0010908425876598518 +changing lr +epoch 55, time 522.77, cls_loss 0.0037 cls_loss_mapping 0.0112 cls_loss_causal 0.4375 re_mapping 0.0109 re_causal 0.0188 /// teacc 98.31 lr 0.00095492 +351 +0.000954915028125264 +changing lr +epoch 56, time 523.02, cls_loss 0.0041 cls_loss_mapping 0.0109 cls_loss_causal 0.4403 re_mapping 0.0108 re_causal 0.0186 /// teacc 97.05 lr 0.00082713 +351 +0.0008271337313934874 +changing lr +epoch 57, time 527.11, cls_loss 0.0028 cls_loss_mapping 0.0091 cls_loss_causal 0.4157 re_mapping 0.0108 re_causal 0.0176 /// teacc 97.47 lr 0.00070776 +351 +0.00070775603199067 +changing lr +epoch 58, time 504.49, cls_loss 0.0031 cls_loss_mapping 0.0086 cls_loss_causal 0.4095 re_mapping 0.0108 re_causal 0.0171 /// teacc 97.89 lr 0.00059702 +351 +0.0005970223407163104 +changing lr +epoch 59, time 497.53, cls_loss 0.0053 cls_loss_mapping 0.0115 cls_loss_causal 0.4429 re_mapping 0.0105 re_causal 0.0172 /// teacc 97.05 lr 0.00049516 +351 +0.0004951556604879052 +changing lr +epoch 60, time 507.85, cls_loss 0.0043 cls_loss_mapping 0.0108 cls_loss_causal 0.4240 re_mapping 0.0103 re_causal 0.0166 /// teacc 98.31 lr 0.00040236 +351 +0.00040236113724274745 +changing lr +epoch 61, time 489.10, cls_loss 0.0040 cls_loss_mapping 0.0104 cls_loss_causal 0.4613 re_mapping 0.0103 re_causal 0.0175 /// teacc 97.05 lr 0.00031883 +351 +0.00031882564680131423 +changing lr +epoch 62, time 487.44, cls_loss 0.0040 cls_loss_mapping 0.0101 cls_loss_causal 0.4445 re_mapping 0.0102 re_causal 0.0167 /// teacc 98.31 lr 0.00024472 +351 +0.0002447174185242325 +changing lr +epoch 63, time 492.60, cls_loss 0.0030 cls_loss_mapping 0.0067 cls_loss_causal 0.3786 re_mapping 0.0102 re_causal 0.0165 /// teacc 97.89 lr 0.00018019 +351 +0.0001801856965207339 +changing lr +epoch 64, time 493.59, cls_loss 0.0040 cls_loss_mapping 0.0106 cls_loss_causal 0.4459 re_mapping 0.0101 re_causal 0.0165 /// teacc 96.62 lr 0.00012536 +351 +0.000125360439090882 +changing lr +epoch 65, time 485.22, cls_loss 0.0051 cls_loss_mapping 0.0094 cls_loss_causal 0.4355 re_mapping 0.0101 re_causal 0.0162 /// teacc 97.47 lr 0.00008035 +351 +8.03520570068517e-05 +changing lr +epoch 66, time 475.77, cls_loss 0.0036 cls_loss_mapping 0.0086 cls_loss_causal 0.4274 re_mapping 0.0101 re_causal 0.0165 /// teacc 97.89 lr 0.00004525 +351 +4.5251191160326525e-05 +changing lr +epoch 67, time 483.43, cls_loss 0.0043 cls_loss_mapping 0.0107 cls_loss_causal 0.4531 re_mapping 0.0102 re_causal 0.0168 /// teacc 96.62 lr 0.00002013 +351 +2.0128530023804673e-05 +changing lr +epoch 68, time 484.93, cls_loss 0.0030 cls_loss_mapping 0.0073 cls_loss_causal 0.4376 re_mapping 0.0102 re_causal 0.0166 /// teacc 97.89 lr 0.00000503 +351 +5.034667293427056e-06 +changing lr +epoch 69, time 479.93, cls_loss 0.0041 cls_loss_mapping 0.0089 cls_loss_causal 0.4412 re_mapping 0.0100 re_causal 0.0165 /// teacc 96.20 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'cartoon', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/cartoon_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['cartoon', 'art_painting', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + cartoon art_painting photo sketch Avg +w/o do (original x) 99.616041 76.806641 89.700599 72.613897 79.707045 + cartoon art_painting photo sketch Avg +do 99.573379 75.537109 89.760479 73.631967 79.643185 diff --git a/Meta-causal/code-withStyleAttack/56721.error b/Meta-causal/code-withStyleAttack/56721.error new file mode 100644 index 0000000000000000000000000000000000000000..fd313270a3ba847b383c7eb4ae546600fd872b6a --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56721.error @@ -0,0 +1 @@ +run_my_joint_v13_test.sh: line 29: de: command not found diff --git a/Meta-causal/code-withStyleAttack/56721.log b/Meta-causal/code-withStyleAttack/56721.log new file mode 100644 index 0000000000000000000000000000000000000000..e0e3660d35bfb4347dcbaf5d7e601b60419518bc --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56721.log @@ -0,0 +1,329 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'photo', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_train.hdf5 torch.Size([1499, 3, 227, 227]) torch.Size([1499]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_val.hdf5 torch.Size([171, 3, 227, 227]) torch.Size([171]) +-------------------------------------loading pretrain weights---------------------------------- +249 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 330.26, cls_loss 2.1901 cls_loss_mapping 1.0966 cls_loss_causal 1.5411 re_mapping 1.2660 re_causal 1.2689 /// teacc 95.32 lr 0.00999497 +249 +0.009994965332706574 +changing lr +epoch 1, time 329.57, cls_loss 1.1155 cls_loss_mapping 0.7787 cls_loss_causal 1.4901 re_mapping 0.9558 re_causal 0.9646 /// teacc 93.57 lr 0.00997987 +249 +0.009979871469976196 +changing lr +epoch 2, time 330.39, cls_loss 0.9288 cls_loss_mapping 0.7032 cls_loss_causal 1.4747 re_mapping 0.7723 re_causal 0.7847 /// teacc 75.44 lr 0.00995475 +249 +0.009954748808839675 +changing lr +epoch 3, time 328.90, cls_loss 1.2627 cls_loss_mapping 0.6321 cls_loss_causal 1.5390 re_mapping 0.6502 re_causal 0.6690 /// teacc 85.96 lr 0.00991965 +249 +0.009919647942993149 +changing lr +epoch 4, time 327.35, cls_loss 0.9500 cls_loss_mapping 0.7241 cls_loss_causal 1.5912 re_mapping 0.6164 re_causal 0.6345 /// teacc 93.57 lr 0.00987464 +249 +0.009874639560909117 +changing lr +epoch 5, time 326.62, cls_loss 1.4824 cls_loss_mapping 0.9000 cls_loss_causal 1.6461 re_mapping 0.5127 re_causal 0.5278 /// teacc 79.53 lr 0.00981981 +249 +0.009819814303479266 +changing lr +epoch 6, time 329.86, cls_loss 0.5391 cls_loss_mapping 0.6994 cls_loss_causal 1.5297 re_mapping 0.4445 re_causal 0.4580 /// teacc 91.23 lr 0.00975528 +249 +0.009755282581475767 +changing lr +epoch 7, time 326.96, cls_loss 0.4282 cls_loss_mapping 0.7985 cls_loss_causal 1.5762 re_mapping 0.4031 re_causal 0.4239 /// teacc 90.06 lr 0.00968117 +249 +0.009681174353198686 +changing lr +epoch 8, time 324.18, cls_loss 0.4582 cls_loss_mapping 0.6437 cls_loss_causal 1.4929 re_mapping 0.3704 re_causal 0.4016 /// teacc 92.40 lr 0.00959764 +249 +0.009597638862757255 +changing lr +epoch 9, time 329.22, cls_loss 0.4196 cls_loss_mapping 0.6434 cls_loss_causal 1.5206 re_mapping 0.3366 re_causal 0.3732 /// teacc 91.23 lr 0.00950484 +249 +0.009504844339512096 +changing lr +epoch 10, time 326.22, cls_loss 0.6899 cls_loss_mapping 0.6931 cls_loss_causal 1.4531 re_mapping 0.3138 re_causal 0.3465 /// teacc 91.81 lr 0.00940298 +249 +0.009402977659283692 +changing lr +epoch 11, time 332.53, cls_loss 0.2100 cls_loss_mapping 0.5305 cls_loss_causal 1.2812 re_mapping 0.2652 re_causal 0.3015 /// teacc 94.15 lr 0.00929224 +249 +0.009292243968009333 +changing lr +epoch 12, time 328.24, cls_loss 1.9157 cls_loss_mapping 1.1807 cls_loss_causal 1.8542 re_mapping 0.2875 re_causal 0.3153 /// teacc 81.29 lr 0.00917287 +249 +0.009172866268606516 +changing lr +epoch 13, time 330.89, cls_loss 0.5559 cls_loss_mapping 0.9412 cls_loss_causal 1.5863 re_mapping 0.2804 re_causal 0.3010 /// teacc 88.30 lr 0.00904508 +249 +0.00904508497187474 +changing lr +epoch 14, time 327.08, cls_loss 0.2945 cls_loss_mapping 0.7027 cls_loss_causal 1.4399 re_mapping 0.2493 re_causal 0.2637 /// teacc 89.47 lr 0.00890916 +249 +0.008909157412340152 +changing lr +epoch 15, time 327.50, cls_loss 0.1556 cls_loss_mapping 0.5735 cls_loss_causal 1.3348 re_mapping 0.2367 re_causal 0.2499 /// teacc 90.64 lr 0.00876536 +249 +0.00876535733001806 +changing lr +epoch 16, time 325.93, cls_loss 0.5865 cls_loss_mapping 0.6469 cls_loss_causal 1.4535 re_mapping 0.2249 re_causal 0.2442 /// teacc 83.63 lr 0.00861397 +249 +0.008613974319136962 +changing lr +epoch 17, time 325.67, cls_loss 0.2541 cls_loss_mapping 0.5530 cls_loss_causal 1.3152 re_mapping 0.1981 re_causal 0.2108 /// teacc 90.64 lr 0.00845531 +249 +0.008455313244934327 +changing lr +epoch 18, time 328.07, cls_loss 0.1021 cls_loss_mapping 0.4746 cls_loss_causal 1.2840 re_mapping 0.1724 re_causal 0.1940 /// teacc 91.81 lr 0.00828969 +249 +0.008289693629698565 +changing lr +epoch 19, time 327.66, cls_loss 0.2583 cls_loss_mapping 0.4658 cls_loss_causal 1.3477 re_mapping 0.1511 re_causal 0.1725 /// teacc 85.38 lr 0.00811745 +249 +0.00811744900929367 +changing lr +epoch 20, time 330.32, cls_loss 0.2436 cls_loss_mapping 0.4640 cls_loss_causal 1.2885 re_mapping 0.1358 re_causal 0.1612 /// teacc 89.47 lr 0.00793893 +249 +0.007938926261462368 +changing lr +epoch 21, time 327.13, cls_loss 0.0809 cls_loss_mapping 0.3624 cls_loss_causal 1.1645 re_mapping 0.1276 re_causal 0.1497 /// teacc 92.40 lr 0.00775448 +249 +0.007754484907260515 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 327.51, cls_loss 0.0782 cls_loss_mapping 0.2983 cls_loss_causal 1.0244 re_mapping 0.1161 re_causal 0.1302 /// teacc 95.91 lr 0.00756450 +249 +0.007564496387029534 +changing lr +epoch 23, time 328.44, cls_loss 0.0508 cls_loss_mapping 0.2665 cls_loss_causal 1.0062 re_mapping 0.1035 re_causal 0.1238 /// teacc 92.40 lr 0.00736934 +249 +0.007369343312364995 +changing lr +epoch 24, time 326.85, cls_loss 0.0439 cls_loss_mapping 0.2489 cls_loss_causal 0.9377 re_mapping 0.0935 re_causal 0.1083 /// teacc 93.57 lr 0.00716942 +249 +0.0071694186955877925 +changing lr +epoch 25, time 328.40, cls_loss 0.0447 cls_loss_mapping 0.2510 cls_loss_causal 0.9697 re_mapping 0.0891 re_causal 0.1042 /// teacc 95.32 lr 0.00696513 +249 +0.0069651251582696205 +changing lr +epoch 26, time 326.49, cls_loss 0.0183 cls_loss_mapping 0.2090 cls_loss_causal 0.9070 re_mapping 0.0889 re_causal 0.1054 /// teacc 94.15 lr 0.00675687 +249 +0.006756874120406716 +changing lr +epoch 27, time 329.18, cls_loss 0.0199 cls_loss_mapping 0.2252 cls_loss_causal 0.9563 re_mapping 0.0849 re_causal 0.1040 /// teacc 92.40 lr 0.00654508 +249 +0.00654508497187474 +changing lr +epoch 28, time 331.00, cls_loss 0.0349 cls_loss_mapping 0.1811 cls_loss_causal 0.8829 re_mapping 0.0737 re_causal 0.0947 /// teacc 94.15 lr 0.00633018 +249 +0.006330184227833378 +changing lr +epoch 29, time 330.93, cls_loss 0.0173 cls_loss_mapping 0.1582 cls_loss_causal 0.8307 re_mapping 0.0685 re_causal 0.0870 /// teacc 95.91 lr 0.00611260 +249 +0.006112604669781575 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 333.04, cls_loss 0.0136 cls_loss_mapping 0.1520 cls_loss_causal 0.8025 re_mapping 0.0632 re_causal 0.0809 /// teacc 97.08 lr 0.00589278 +249 +0.005892784473993186 +changing lr +epoch 31, time 328.71, cls_loss 0.0093 cls_loss_mapping 0.1464 cls_loss_causal 0.7705 re_mapping 0.0664 re_causal 0.0860 /// teacc 95.32 lr 0.00567117 +249 +0.00567116632908828 +changing lr +epoch 32, time 331.65, cls_loss 0.0048 cls_loss_mapping 0.1322 cls_loss_causal 0.7072 re_mapping 0.0552 re_causal 0.0736 /// teacc 95.32 lr 0.00544820 +249 +0.00544819654451717 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 331.93, cls_loss 0.0196 cls_loss_mapping 0.1406 cls_loss_causal 0.7016 re_mapping 0.0551 re_causal 0.0790 /// teacc 97.66 lr 0.00522432 +249 +0.005224324151752577 +changing lr +epoch 34, time 326.51, cls_loss 0.0110 cls_loss_mapping 0.1272 cls_loss_causal 0.7379 re_mapping 0.0532 re_causal 0.0753 /// teacc 95.91 lr 0.00500000 +249 +0.005000000000000003 +changing lr +epoch 35, time 326.79, cls_loss 0.0039 cls_loss_mapping 0.1204 cls_loss_causal 0.7016 re_mapping 0.0500 re_causal 0.0750 /// teacc 96.49 lr 0.00477568 +249 +0.004775675848247429 +changing lr +epoch 36, time 328.75, cls_loss 0.0098 cls_loss_mapping 0.1122 cls_loss_causal 0.6372 re_mapping 0.0458 re_causal 0.0661 /// teacc 95.32 lr 0.00455180 +249 +0.004551803455482836 +changing lr +epoch 37, time 333.82, cls_loss 0.0088 cls_loss_mapping 0.1083 cls_loss_causal 0.6648 re_mapping 0.0459 re_causal 0.0701 /// teacc 95.91 lr 0.00432883 +249 +0.004328833670911726 +changing lr +epoch 38, time 328.36, cls_loss 0.0111 cls_loss_mapping 0.1082 cls_loss_causal 0.6774 re_mapping 0.0479 re_causal 0.0716 /// teacc 94.74 lr 0.00410722 +249 +0.0041072155260068206 +changing lr +epoch 39, time 329.81, cls_loss 0.0019 cls_loss_mapping 0.0890 cls_loss_causal 0.6447 re_mapping 0.0461 re_causal 0.0699 /// teacc 95.32 lr 0.00388740 +249 +0.0038873953302184317 +changing lr +epoch 40, time 329.55, cls_loss 0.0031 cls_loss_mapping 0.0853 cls_loss_causal 0.5882 re_mapping 0.0445 re_causal 0.0632 /// teacc 94.74 lr 0.00366982 +249 +0.003669815772166629 +changing lr +epoch 41, time 330.31, cls_loss 0.0050 cls_loss_mapping 0.0811 cls_loss_causal 0.5662 re_mapping 0.0384 re_causal 0.0568 /// teacc 95.32 lr 0.00345492 +249 +0.0034549150281252667 +changing lr +epoch 42, time 333.18, cls_loss 0.0062 cls_loss_mapping 0.0839 cls_loss_causal 0.6104 re_mapping 0.0375 re_causal 0.0582 /// teacc 95.91 lr 0.00324313 +249 +0.0032431258795932905 +changing lr +epoch 43, time 329.10, cls_loss 0.0014 cls_loss_mapping 0.0792 cls_loss_causal 0.5998 re_mapping 0.0385 re_causal 0.0578 /// teacc 96.49 lr 0.00303487 +249 +0.0030348748417303863 +changing lr +epoch 44, time 327.44, cls_loss 0.0038 cls_loss_mapping 0.0816 cls_loss_causal 0.5993 re_mapping 0.0363 re_causal 0.0564 /// teacc 96.49 lr 0.00283058 +249 +0.0028305813044122124 +changing lr +epoch 45, time 328.69, cls_loss 0.0064 cls_loss_mapping 0.0724 cls_loss_causal 0.5434 re_mapping 0.0350 re_causal 0.0566 /// teacc 97.08 lr 0.00263066 +249 +0.0026306566876350096 +changing lr +epoch 46, time 329.11, cls_loss 0.0036 cls_loss_mapping 0.0732 cls_loss_causal 0.6550 re_mapping 0.0336 re_causal 0.0560 /// teacc 97.66 lr 0.00243550 +249 +0.0024355036129704724 +changing lr +epoch 47, time 330.95, cls_loss 0.0028 cls_loss_mapping 0.0696 cls_loss_causal 0.5213 re_mapping 0.0347 re_causal 0.0540 /// teacc 95.32 lr 0.00224552 +249 +0.00224551509273949 +changing lr +epoch 48, time 329.49, cls_loss 0.0022 cls_loss_mapping 0.0614 cls_loss_causal 0.5186 re_mapping 0.0319 re_causal 0.0531 /// teacc 97.08 lr 0.00206107 +249 +0.002061073738537637 +changing lr +epoch 49, time 327.39, cls_loss 0.0030 cls_loss_mapping 0.0631 cls_loss_causal 0.5368 re_mapping 0.0315 re_causal 0.0477 /// teacc 97.08 lr 0.00188255 +249 +0.0018825509907063344 +changing lr +epoch 50, time 330.68, cls_loss 0.0025 cls_loss_mapping 0.0624 cls_loss_causal 0.5418 re_mapping 0.0308 re_causal 0.0501 /// teacc 95.91 lr 0.00171031 +249 +0.0017103063703014388 +changing lr +epoch 51, time 331.11, cls_loss 0.0024 cls_loss_mapping 0.0666 cls_loss_causal 0.6219 re_mapping 0.0303 re_causal 0.0463 /// teacc 94.15 lr 0.00154469 +249 +0.0015446867550656784 +changing lr +epoch 52, time 329.80, cls_loss 0.0037 cls_loss_mapping 0.0624 cls_loss_causal 0.5204 re_mapping 0.0305 re_causal 0.0459 /// teacc 95.91 lr 0.00138603 +249 +0.001386025680863044 +changing lr +epoch 53, time 330.30, cls_loss 0.0021 cls_loss_mapping 0.0573 cls_loss_causal 0.4976 re_mapping 0.0330 re_causal 0.0522 /// teacc 96.49 lr 0.00123464 +249 +0.0012346426699819469 +changing lr +epoch 54, time 328.15, cls_loss 0.0037 cls_loss_mapping 0.0636 cls_loss_causal 0.5476 re_mapping 0.0300 re_causal 0.0478 /// teacc 94.74 lr 0.00109084 +249 +0.0010908425876598518 +changing lr +epoch 55, time 330.82, cls_loss 0.0019 cls_loss_mapping 0.0573 cls_loss_causal 0.4965 re_mapping 0.0298 re_causal 0.0464 /// teacc 94.74 lr 0.00095492 +249 +0.000954915028125264 +changing lr +epoch 56, time 327.34, cls_loss 0.0026 cls_loss_mapping 0.0569 cls_loss_causal 0.5251 re_mapping 0.0303 re_causal 0.0466 /// teacc 95.91 lr 0.00082713 +249 +0.0008271337313934874 +changing lr +epoch 57, time 333.58, cls_loss 0.0042 cls_loss_mapping 0.0546 cls_loss_causal 0.5309 re_mapping 0.0287 re_causal 0.0428 /// teacc 95.32 lr 0.00070776 +249 +0.00070775603199067 +changing lr +epoch 58, time 328.86, cls_loss 0.0031 cls_loss_mapping 0.0587 cls_loss_causal 0.5149 re_mapping 0.0288 re_causal 0.0456 /// teacc 96.49 lr 0.00059702 +249 +0.0005970223407163104 +changing lr +epoch 59, time 328.86, cls_loss 0.0046 cls_loss_mapping 0.0559 cls_loss_causal 0.5242 re_mapping 0.0292 re_causal 0.0461 /// teacc 95.32 lr 0.00049516 +249 +0.0004951556604879052 +changing lr +epoch 60, time 329.33, cls_loss 0.0035 cls_loss_mapping 0.0531 cls_loss_causal 0.5105 re_mapping 0.0286 re_causal 0.0415 /// teacc 94.74 lr 0.00040236 +249 +0.00040236113724274745 +changing lr +epoch 61, time 329.57, cls_loss 0.0024 cls_loss_mapping 0.0552 cls_loss_causal 0.5395 re_mapping 0.0269 re_causal 0.0440 /// teacc 95.91 lr 0.00031883 +249 +0.00031882564680131423 +changing lr +epoch 62, time 333.79, cls_loss 0.0025 cls_loss_mapping 0.0505 cls_loss_causal 0.5307 re_mapping 0.0257 re_causal 0.0430 /// teacc 95.32 lr 0.00024472 +249 +0.0002447174185242325 +changing lr +epoch 63, time 325.49, cls_loss 0.0033 cls_loss_mapping 0.0561 cls_loss_causal 0.5009 re_mapping 0.0285 re_causal 0.0429 /// teacc 96.49 lr 0.00018019 +249 +0.0001801856965207339 +changing lr +epoch 64, time 325.60, cls_loss 0.0020 cls_loss_mapping 0.0478 cls_loss_causal 0.5195 re_mapping 0.0274 re_causal 0.0416 /// teacc 95.32 lr 0.00012536 +249 +0.000125360439090882 +changing lr +epoch 65, time 329.45, cls_loss 0.0022 cls_loss_mapping 0.0502 cls_loss_causal 0.4924 re_mapping 0.0274 re_causal 0.0425 /// teacc 94.15 lr 0.00008035 +249 +8.03520570068517e-05 +changing lr +epoch 66, time 331.82, cls_loss 0.0036 cls_loss_mapping 0.0536 cls_loss_causal 0.5226 re_mapping 0.0276 re_causal 0.0429 /// teacc 95.91 lr 0.00004525 +249 +4.5251191160326525e-05 +changing lr +epoch 67, time 328.42, cls_loss 0.0030 cls_loss_mapping 0.0563 cls_loss_causal 0.5390 re_mapping 0.0282 re_causal 0.0435 /// teacc 95.32 lr 0.00002013 +249 +2.0128530023804673e-05 +changing lr +epoch 68, time 331.35, cls_loss 0.0034 cls_loss_mapping 0.0501 cls_loss_causal 0.5100 re_mapping 0.0269 re_causal 0.0424 /// teacc 95.32 lr 0.00000503 +249 +5.034667293427056e-06 +changing lr +epoch 69, time 332.59, cls_loss 0.0023 cls_loss_mapping 0.0540 cls_loss_causal 0.5166 re_mapping 0.0279 re_causal 0.0451 /// teacc 97.08 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'photo', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/photo_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['photo', 'art_painting', 'cartoon', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + photo art_painting cartoon sketch Avg +w/o do (original x) 99.700599 60.253906 43.515358 57.724612 53.831292 + photo art_painting cartoon sketch Avg +do 99.760479 60.15625 49.274744 60.57521 56.668735 diff --git a/Meta-causal/code-withStyleAttack/56722.error b/Meta-causal/code-withStyleAttack/56722.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/56722.log b/Meta-causal/code-withStyleAttack/56722.log new file mode 100644 index 0000000000000000000000000000000000000000..7f87ef44ae593d26a6c9da1bc6c04bcc59b5b290 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/56722.log @@ -0,0 +1,333 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398]) +-------------------------------------loading pretrain weights---------------------------------- +588 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 845.95, cls_loss 3.6738 cls_loss_mapping 1.1243 cls_loss_causal 1.4459 re_mapping 0.6948 re_causal 0.6950 /// teacc 87.69 lr 0.00999497 +588 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 854.64, cls_loss 0.5577 cls_loss_mapping 0.4581 cls_loss_causal 1.0021 re_mapping 0.2545 re_causal 0.2541 /// teacc 90.20 lr 0.00997987 +588 +0.009979871469976196 +changing lr +epoch 2, time 878.75, cls_loss 0.1988 cls_loss_mapping 0.2884 cls_loss_causal 0.8433 re_mapping 0.1588 re_causal 0.1584 /// teacc 89.70 lr 0.00995475 +588 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 892.14, cls_loss 0.1337 cls_loss_mapping 0.2165 cls_loss_causal 0.7708 re_mapping 0.1261 re_causal 0.1263 /// teacc 92.96 lr 0.00991965 +588 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 852.27, cls_loss 0.0720 cls_loss_mapping 0.1604 cls_loss_causal 0.6995 re_mapping 0.1031 re_causal 0.1040 /// teacc 93.22 lr 0.00987464 +588 +0.009874639560909117 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 847.00, cls_loss 0.0390 cls_loss_mapping 0.1253 cls_loss_causal 0.6512 re_mapping 0.0839 re_causal 0.0858 /// teacc 93.72 lr 0.00981981 +588 +0.009819814303479266 +changing lr +epoch 6, time 868.82, cls_loss 0.0280 cls_loss_mapping 0.1074 cls_loss_causal 0.6153 re_mapping 0.0698 re_causal 0.0724 /// teacc 92.46 lr 0.00975528 +588 +0.009755282581475767 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 845.22, cls_loss 0.0251 cls_loss_mapping 0.0936 cls_loss_causal 0.5946 re_mapping 0.0582 re_causal 0.0616 /// teacc 93.97 lr 0.00968117 +588 +0.009681174353198686 +changing lr +epoch 8, time 850.90, cls_loss 0.0209 cls_loss_mapping 0.0815 cls_loss_causal 0.5643 re_mapping 0.0500 re_causal 0.0549 /// teacc 92.71 lr 0.00959764 +588 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 861.45, cls_loss 0.0196 cls_loss_mapping 0.0758 cls_loss_causal 0.5799 re_mapping 0.0425 re_causal 0.0492 /// teacc 94.97 lr 0.00950484 +588 +0.009504844339512096 +changing lr +epoch 10, time 887.24, cls_loss 0.0124 cls_loss_mapping 0.0557 cls_loss_causal 0.5474 re_mapping 0.0349 re_causal 0.0424 /// teacc 94.22 lr 0.00940298 +588 +0.009402977659283692 +changing lr +epoch 11, time 860.82, cls_loss 0.0155 cls_loss_mapping 0.0521 cls_loss_causal 0.5327 re_mapping 0.0303 re_causal 0.0397 /// teacc 92.96 lr 0.00929224 +588 +0.009292243968009333 +changing lr +epoch 12, time 882.17, cls_loss 0.0122 cls_loss_mapping 0.0491 cls_loss_causal 0.5306 re_mapping 0.0254 re_causal 0.0353 /// teacc 92.46 lr 0.00917287 +588 +0.009172866268606516 +changing lr +epoch 13, time 886.31, cls_loss 0.0123 cls_loss_mapping 0.0488 cls_loss_causal 0.5114 re_mapping 0.0231 re_causal 0.0341 /// teacc 93.97 lr 0.00904508 +588 +0.00904508497187474 +changing lr +epoch 14, time 860.16, cls_loss 0.0097 cls_loss_mapping 0.0391 cls_loss_causal 0.5384 re_mapping 0.0201 re_causal 0.0320 /// teacc 94.47 lr 0.00890916 +588 +0.008909157412340152 +changing lr +epoch 15, time 851.33, cls_loss 0.0083 cls_loss_mapping 0.0374 cls_loss_causal 0.4965 re_mapping 0.0180 re_causal 0.0299 /// teacc 92.71 lr 0.00876536 +588 +0.00876535733001806 +changing lr +epoch 16, time 856.91, cls_loss 0.0087 cls_loss_mapping 0.0368 cls_loss_causal 0.4716 re_mapping 0.0163 re_causal 0.0278 /// teacc 91.96 lr 0.00861397 +588 +0.008613974319136962 +changing lr +epoch 17, time 881.17, cls_loss 0.0070 cls_loss_mapping 0.0307 cls_loss_causal 0.4890 re_mapping 0.0146 re_causal 0.0269 /// teacc 93.47 lr 0.00845531 +588 +0.008455313244934327 +changing lr +epoch 18, time 868.14, cls_loss 0.0060 cls_loss_mapping 0.0262 cls_loss_causal 0.4835 re_mapping 0.0132 re_causal 0.0256 /// teacc 93.47 lr 0.00828969 +588 +0.008289693629698565 +changing lr +epoch 19, time 883.13, cls_loss 0.0067 cls_loss_mapping 0.0258 cls_loss_causal 0.4667 re_mapping 0.0123 re_causal 0.0250 /// teacc 93.72 lr 0.00811745 +588 +0.00811744900929367 +changing lr +epoch 20, time 884.58, cls_loss 0.0064 cls_loss_mapping 0.0251 cls_loss_causal 0.4629 re_mapping 0.0117 re_causal 0.0239 /// teacc 94.72 lr 0.00793893 +588 +0.007938926261462368 +changing lr +epoch 21, time 875.05, cls_loss 0.0051 cls_loss_mapping 0.0202 cls_loss_causal 0.4715 re_mapping 0.0107 re_causal 0.0234 /// teacc 92.46 lr 0.00775448 +588 +0.007754484907260515 +changing lr +epoch 22, time 896.60, cls_loss 0.0054 cls_loss_mapping 0.0194 cls_loss_causal 0.4351 re_mapping 0.0099 re_causal 0.0214 /// teacc 94.97 lr 0.00756450 +588 +0.007564496387029534 +changing lr +epoch 23, time 860.93, cls_loss 0.0049 cls_loss_mapping 0.0175 cls_loss_causal 0.4279 re_mapping 0.0094 re_causal 0.0210 /// teacc 92.71 lr 0.00736934 +588 +0.007369343312364995 +changing lr +epoch 24, time 870.01, cls_loss 0.0046 cls_loss_mapping 0.0183 cls_loss_causal 0.4499 re_mapping 0.0094 re_causal 0.0216 /// teacc 94.22 lr 0.00716942 +588 +0.0071694186955877925 +changing lr +epoch 25, time 881.12, cls_loss 0.0059 cls_loss_mapping 0.0212 cls_loss_causal 0.4502 re_mapping 0.0092 re_causal 0.0210 /// teacc 94.22 lr 0.00696513 +588 +0.0069651251582696205 +changing lr +epoch 26, time 883.73, cls_loss 0.0052 cls_loss_mapping 0.0151 cls_loss_causal 0.4330 re_mapping 0.0088 re_causal 0.0207 /// teacc 94.47 lr 0.00675687 +588 +0.006756874120406716 +changing lr +epoch 27, time 876.67, cls_loss 0.0050 cls_loss_mapping 0.0183 cls_loss_causal 0.4334 re_mapping 0.0082 re_causal 0.0200 /// teacc 93.22 lr 0.00654508 +588 +0.00654508497187474 +changing lr +epoch 28, time 849.09, cls_loss 0.0067 cls_loss_mapping 0.0154 cls_loss_causal 0.4283 re_mapping 0.0084 re_causal 0.0204 /// teacc 93.47 lr 0.00633018 +588 +0.006330184227833378 +changing lr +epoch 29, time 851.32, cls_loss 0.0044 cls_loss_mapping 0.0147 cls_loss_causal 0.3901 re_mapping 0.0077 re_causal 0.0185 /// teacc 92.96 lr 0.00611260 +588 +0.006112604669781575 +changing lr +epoch 30, time 854.67, cls_loss 0.0034 cls_loss_mapping 0.0126 cls_loss_causal 0.4241 re_mapping 0.0076 re_causal 0.0193 /// teacc 93.72 lr 0.00589278 +588 +0.005892784473993186 +changing lr +epoch 31, time 861.52, cls_loss 0.0048 cls_loss_mapping 0.0151 cls_loss_causal 0.4106 re_mapping 0.0072 re_causal 0.0186 /// teacc 93.22 lr 0.00567117 +588 +0.00567116632908828 +changing lr +epoch 32, time 886.70, cls_loss 0.0034 cls_loss_mapping 0.0119 cls_loss_causal 0.4174 re_mapping 0.0070 re_causal 0.0183 /// teacc 93.72 lr 0.00544820 +588 +0.00544819654451717 +changing lr +epoch 33, time 865.62, cls_loss 0.0038 cls_loss_mapping 0.0111 cls_loss_causal 0.4096 re_mapping 0.0068 re_causal 0.0178 /// teacc 92.96 lr 0.00522432 +588 +0.005224324151752577 +changing lr +epoch 34, time 853.80, cls_loss 0.0039 cls_loss_mapping 0.0117 cls_loss_causal 0.4176 re_mapping 0.0066 re_causal 0.0176 /// teacc 93.22 lr 0.00500000 +588 +0.005000000000000003 +changing lr +epoch 35, time 873.63, cls_loss 0.0043 cls_loss_mapping 0.0126 cls_loss_causal 0.4324 re_mapping 0.0065 re_causal 0.0176 /// teacc 93.22 lr 0.00477568 +588 +0.004775675848247429 +changing lr +epoch 36, time 847.64, cls_loss 0.0035 cls_loss_mapping 0.0099 cls_loss_causal 0.4156 re_mapping 0.0062 re_causal 0.0166 /// teacc 93.72 lr 0.00455180 +588 +0.004551803455482836 +changing lr +epoch 37, time 821.88, cls_loss 0.0038 cls_loss_mapping 0.0099 cls_loss_causal 0.4130 re_mapping 0.0059 re_causal 0.0165 /// teacc 94.22 lr 0.00432883 +588 +0.004328833670911726 +changing lr +epoch 38, time 833.52, cls_loss 0.0039 cls_loss_mapping 0.0113 cls_loss_causal 0.3887 re_mapping 0.0059 re_causal 0.0166 /// teacc 94.97 lr 0.00410722 +588 +0.0041072155260068206 +changing lr +epoch 39, time 803.24, cls_loss 0.0032 cls_loss_mapping 0.0079 cls_loss_causal 0.4193 re_mapping 0.0058 re_causal 0.0165 /// teacc 94.72 lr 0.00388740 +588 +0.0038873953302184317 +changing lr +epoch 40, time 810.38, cls_loss 0.0034 cls_loss_mapping 0.0082 cls_loss_causal 0.3832 re_mapping 0.0056 re_causal 0.0154 /// teacc 93.47 lr 0.00366982 +588 +0.003669815772166629 +changing lr +epoch 41, time 798.30, cls_loss 0.0038 cls_loss_mapping 0.0093 cls_loss_causal 0.3853 re_mapping 0.0054 re_causal 0.0152 /// teacc 93.72 lr 0.00345492 +588 +0.0034549150281252667 +changing lr +epoch 42, time 770.71, cls_loss 0.0038 cls_loss_mapping 0.0078 cls_loss_causal 0.4206 re_mapping 0.0052 re_causal 0.0155 /// teacc 93.22 lr 0.00324313 +588 +0.0032431258795932905 +changing lr +epoch 43, time 769.73, cls_loss 0.0032 cls_loss_mapping 0.0085 cls_loss_causal 0.3786 re_mapping 0.0052 re_causal 0.0147 /// teacc 94.22 lr 0.00303487 +588 +0.0030348748417303863 +changing lr +epoch 44, time 781.30, cls_loss 0.0030 cls_loss_mapping 0.0066 cls_loss_causal 0.3762 re_mapping 0.0052 re_causal 0.0141 /// teacc 92.96 lr 0.00283058 +588 +0.0028305813044122124 +changing lr +epoch 45, time 763.23, cls_loss 0.0028 cls_loss_mapping 0.0060 cls_loss_causal 0.3935 re_mapping 0.0050 re_causal 0.0143 /// teacc 93.97 lr 0.00263066 +588 +0.0026306566876350096 +changing lr +epoch 46, time 756.78, cls_loss 0.0030 cls_loss_mapping 0.0072 cls_loss_causal 0.3847 re_mapping 0.0049 re_causal 0.0141 /// teacc 94.47 lr 0.00243550 +588 +0.0024355036129704724 +changing lr +epoch 47, time 753.45, cls_loss 0.0027 cls_loss_mapping 0.0062 cls_loss_causal 0.3732 re_mapping 0.0048 re_causal 0.0134 /// teacc 93.72 lr 0.00224552 +588 +0.00224551509273949 +changing lr +epoch 48, time 753.93, cls_loss 0.0029 cls_loss_mapping 0.0050 cls_loss_causal 0.3621 re_mapping 0.0047 re_causal 0.0131 /// teacc 94.22 lr 0.00206107 +588 +0.002061073738537637 +changing lr +epoch 49, time 760.37, cls_loss 0.0028 cls_loss_mapping 0.0057 cls_loss_causal 0.3736 re_mapping 0.0048 re_causal 0.0132 /// teacc 93.47 lr 0.00188255 +588 +0.0018825509907063344 +changing lr +---------------------saving model at epoch 50---------------------------------------------------- +epoch 50, time 761.62, cls_loss 0.0025 cls_loss_mapping 0.0047 cls_loss_causal 0.3886 re_mapping 0.0047 re_causal 0.0133 /// teacc 95.48 lr 0.00171031 +588 +0.0017103063703014388 +changing lr +epoch 51, time 757.83, cls_loss 0.0026 cls_loss_mapping 0.0051 cls_loss_causal 0.3723 re_mapping 0.0047 re_causal 0.0131 /// teacc 92.21 lr 0.00154469 +588 +0.0015446867550656784 +changing lr +epoch 52, time 756.68, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.3874 re_mapping 0.0046 re_causal 0.0131 /// teacc 92.71 lr 0.00138603 +588 +0.001386025680863044 +changing lr +epoch 53, time 758.80, cls_loss 0.0027 cls_loss_mapping 0.0049 cls_loss_causal 0.3915 re_mapping 0.0046 re_causal 0.0130 /// teacc 93.72 lr 0.00123464 +588 +0.0012346426699819469 +changing lr +epoch 54, time 754.88, cls_loss 0.0026 cls_loss_mapping 0.0049 cls_loss_causal 0.3825 re_mapping 0.0045 re_causal 0.0130 /// teacc 94.22 lr 0.00109084 +588 +0.0010908425876598518 +changing lr +epoch 55, time 758.50, cls_loss 0.0030 cls_loss_mapping 0.0050 cls_loss_causal 0.3839 re_mapping 0.0045 re_causal 0.0126 /// teacc 93.22 lr 0.00095492 +588 +0.000954915028125264 +changing lr +epoch 56, time 757.06, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.3577 re_mapping 0.0045 re_causal 0.0122 /// teacc 94.22 lr 0.00082713 +588 +0.0008271337313934874 +changing lr +epoch 57, time 758.09, cls_loss 0.0023 cls_loss_mapping 0.0046 cls_loss_causal 0.3461 re_mapping 0.0046 re_causal 0.0118 /// teacc 93.47 lr 0.00070776 +588 +0.00070775603199067 +changing lr +epoch 58, time 760.13, cls_loss 0.0023 cls_loss_mapping 0.0039 cls_loss_causal 0.3523 re_mapping 0.0046 re_causal 0.0118 /// teacc 94.47 lr 0.00059702 +588 +0.0005970223407163104 +changing lr +epoch 59, time 756.85, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.3762 re_mapping 0.0045 re_causal 0.0121 /// teacc 93.22 lr 0.00049516 +588 +0.0004951556604879052 +changing lr +epoch 60, time 754.41, cls_loss 0.0023 cls_loss_mapping 0.0041 cls_loss_causal 0.3579 re_mapping 0.0044 re_causal 0.0115 /// teacc 92.71 lr 0.00040236 +588 +0.00040236113724274745 +changing lr +epoch 61, time 758.79, cls_loss 0.0026 cls_loss_mapping 0.0042 cls_loss_causal 0.3682 re_mapping 0.0044 re_causal 0.0114 /// teacc 93.72 lr 0.00031883 +588 +0.00031882564680131423 +changing lr +epoch 62, time 752.52, cls_loss 0.0025 cls_loss_mapping 0.0039 cls_loss_causal 0.3746 re_mapping 0.0044 re_causal 0.0117 /// teacc 92.46 lr 0.00024472 +588 +0.0002447174185242325 +changing lr +epoch 63, time 758.14, cls_loss 0.0025 cls_loss_mapping 0.0034 cls_loss_causal 0.3751 re_mapping 0.0044 re_causal 0.0115 /// teacc 93.72 lr 0.00018019 +588 +0.0001801856965207339 +changing lr +epoch 64, time 751.51, cls_loss 0.0024 cls_loss_mapping 0.0034 cls_loss_causal 0.3590 re_mapping 0.0044 re_causal 0.0114 /// teacc 93.47 lr 0.00012536 +588 +0.000125360439090882 +changing lr +epoch 65, time 760.25, cls_loss 0.0030 cls_loss_mapping 0.0049 cls_loss_causal 0.3519 re_mapping 0.0044 re_causal 0.0112 /// teacc 93.97 lr 0.00008035 +588 +8.03520570068517e-05 +changing lr +epoch 66, time 759.68, cls_loss 0.0023 cls_loss_mapping 0.0036 cls_loss_causal 0.3571 re_mapping 0.0044 re_causal 0.0114 /// teacc 92.96 lr 0.00004525 +588 +4.5251191160326525e-05 +changing lr +epoch 67, time 758.83, cls_loss 0.0025 cls_loss_mapping 0.0043 cls_loss_causal 0.3541 re_mapping 0.0044 re_causal 0.0113 /// teacc 92.21 lr 0.00002013 +588 +2.0128530023804673e-05 +changing lr +epoch 68, time 755.82, cls_loss 0.0021 cls_loss_mapping 0.0032 cls_loss_causal 0.3385 re_mapping 0.0044 re_causal 0.0113 /// teacc 93.22 lr 0.00000503 +588 +5.034667293427056e-06 +changing lr +epoch 69, time 757.80, cls_loss 0.0027 cls_loss_mapping 0.0040 cls_loss_causal 0.3563 re_mapping 0.0044 re_causal 0.0113 /// teacc 93.22 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.312802 55.029297 67.491468 57.964072 60.161612 + sketch art_painting cartoon photo Avg +do 99.312802 49.804688 63.90785 55.449102 56.387213 diff --git a/Meta-causal/code-withStyleAttack/64943.error b/Meta-causal/code-withStyleAttack/64943.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/64943.log b/Meta-causal/code-withStyleAttack/64943.log new file mode 100644 index 0000000000000000000000000000000000000000..39c489604f15f0be2dc04cfe377194415fb1e9bb --- /dev/null +++ b/Meta-causal/code-withStyleAttack/64943.log @@ -0,0 +1,1914 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[ 0.0106, -0.0051, 0.0193, ..., 0.0075, 0.0158, -0.0062], + [-0.0130, -0.0094, -0.0199, ..., 0.0148, 0.0091, 0.0089], + [ 0.0217, 0.0123, -0.0198, ..., -0.0208, 0.0086, -0.0179], + ..., + [-0.0085, -0.0153, 0.0125, ..., 0.0016, 0.0065, 0.0184], + [ 0.0111, -0.0125, 0.0214, ..., -0.0080, 0.0129, -0.0186], + [-0.0095, 0.0164, 0.0024, ..., -0.0037, 0.0123, 0.0207]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0156, -0.0099, -0.0060, -0.0064, -0.0070, 0.0143, 0.0059], + device='cuda:0'), grad: None +306 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 421.76, cls_loss 11.6044 cls_loss_mapping 1.8986 cls_loss_causal 1.9230 re_mapping 0.4058 re_causal 0.4051 /// teacc 43.27 lr 0.00999497 +Epoch 2, weight, value: tensor([[ 0.1878, 0.1646, 0.2100, ..., -0.0196, 0.0020, -0.0184], + [-0.0649, -0.0708, -0.0772, ..., 0.0933, 0.0843, 0.0784], + [-0.0386, -0.0153, -0.0629, ..., 0.0813, 0.0574, 0.0432], + ..., + [-0.0308, -0.0403, -0.0028, ..., 0.0012, 0.0620, 0.0384], + [ 0.0231, 0.0282, 0.0095, ..., -0.1716, -0.1842, -0.2046], + [-0.0140, 0.0138, -0.0027, ..., -0.0242, -0.0221, 0.0152]], + device='cuda:0'), grad: tensor([[-0.2456, -0.2186, -0.2211, ..., -0.0522, -0.0388, -0.0158], + [-0.0894, -0.0504, -0.0532, ..., -0.0684, -0.0511, -0.0178], + [ 0.0267, 0.0195, 0.0190, ..., 0.0144, 0.0101, 0.0034], + ..., + [-0.0347, -0.0211, -0.0188, ..., -0.0300, -0.0179, -0.0050], + [ 0.0645, 0.0405, 0.0429, ..., 0.0440, 0.0329, 0.0115], + [ 0.2252, 0.1888, 0.1958, ..., 0.0678, 0.0501, 0.0192]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0260, 0.0482, -0.0075, -0.0372, 0.0386, -0.0156, 0.0055], + device='cuda:0'), grad: tensor([-0.2336, -0.2844, 0.0637, 0.0666, -0.1208, 0.2062, 0.3025], + device='cuda:0') +306 +0.009994965332706574 +changing lr +epoch 1, time 424.69, cls_loss 2.0222 cls_loss_mapping 1.6048 cls_loss_causal 1.8042 re_mapping 0.1042 re_causal 0.1037 /// teacc 42.31 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 0.1727, 0.1568, 0.1972, ..., -0.0276, -0.0070, -0.0236], + [-0.1169, -0.1227, -0.1246, ..., 0.0940, 0.0914, 0.0822], + [-0.0560, -0.0360, -0.0763, ..., 0.0755, 0.0478, 0.0340], + ..., + [-0.0177, -0.0277, 0.0238, ..., -0.0073, 0.0621, 0.0368], + [ 0.0245, 0.0343, 0.0103, ..., -0.1734, -0.1889, -0.2067], + [ 0.0536, 0.0736, 0.0462, ..., -0.0229, -0.0228, 0.0129]], + device='cuda:0'), grad: tensor([[ 4.2572e-02, 3.2379e-02, 3.3173e-02, ..., 1.3580e-02, + 9.7656e-03, 4.5853e-03], + [ 2.6875e-03, 1.4639e-03, 1.4391e-03, ..., 1.2617e-03, + 9.5606e-04, 5.2929e-04], + [ 1.7932e-01, 1.0278e-01, 1.1157e-01, ..., 6.3354e-02, + 4.5898e-02, 2.2125e-02], + ..., + [-2.3206e-01, -1.4355e-01, -1.5369e-01, ..., -7.8003e-02, + -5.5969e-02, -2.5665e-02], + [ 7.9679e-04, 4.0364e-04, 3.8266e-04, ..., 3.4571e-04, + 2.4199e-04, 1.1581e-04], + [ 4.0680e-02, 2.0706e-02, 2.0401e-02, ..., 1.9073e-02, + 1.4320e-02, 7.4272e-03]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0599, 0.0352, -0.0025, -0.0539, 0.0827, -0.0219, 0.0262], + device='cuda:0'), grad: tensor([ 0.0748, 0.0074, 0.2949, -0.1169, -0.3730, 0.0019, 0.1111], + device='cuda:0') +306 +0.009979871469976196 +changing lr +epoch 2, time 425.75, cls_loss 1.2357 cls_loss_mapping 1.2540 cls_loss_causal 1.6276 re_mapping 0.0906 re_causal 0.0899 /// teacc 37.50 lr 0.00995475 +Epoch 4, weight, value: tensor([[ 0.1852, 0.1727, 0.2084, ..., -0.0183, -0.0016, -0.0191], + [-0.1073, -0.1069, -0.1134, ..., 0.0933, 0.0911, 0.0836], + [-0.0701, -0.0514, -0.0840, ..., 0.0735, 0.0460, 0.0309], + ..., + [-0.0694, -0.0779, -0.0236, ..., -0.0143, 0.0575, 0.0329], + [ 0.0138, 0.0213, -0.0054, ..., -0.1757, -0.1922, -0.2086], + [ 0.1033, 0.1200, 0.0955, ..., -0.0222, -0.0174, 0.0177]], + device='cuda:0'), grad: tensor([[-2.4323e-02, -1.7715e-02, -1.8951e-02, ..., -3.5667e-03, + -2.6474e-03, -1.3180e-03], + [ 2.2745e-04, 1.7011e-04, 1.8191e-04, ..., 2.9743e-05, + 2.2605e-05, 1.0028e-05], + [ 2.1515e-03, 1.2884e-03, 1.3847e-03, ..., 5.4646e-04, + 3.7622e-04, 2.5415e-04], + ..., + [ 1.3123e-02, 9.5825e-03, 1.0254e-02, ..., 1.9026e-03, + 1.4153e-03, 6.9475e-04], + [ 3.9124e-04, 3.5381e-04, 3.7313e-04, ..., 8.3372e-06, + 1.3404e-05, 1.1101e-06], + [ 6.6795e-03, 5.0278e-03, 5.3711e-03, ..., 8.4639e-04, + 6.4564e-04, 2.7847e-04]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0368, 0.0249, -0.0431, -0.0514, 0.0937, -0.0193, 0.0377], + device='cuda:0'), grad: tensor([-0.0337, 0.0003, 0.0039, 0.0023, 0.0181, 0.0003, 0.0088], + device='cuda:0') +306 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 431.13, cls_loss 0.8660 cls_loss_mapping 0.9373 cls_loss_causal 1.4483 re_mapping 0.0863 re_causal 0.0854 /// teacc 48.56 lr 0.00991965 +Epoch 5, weight, value: tensor([[ 0.1969, 0.1932, 0.2240, ..., -0.0106, 0.0083, -0.0084], + [-0.1106, -0.1103, -0.1220, ..., 0.0942, 0.0883, 0.0799], + [-0.0832, -0.0670, -0.0937, ..., 0.0688, 0.0420, 0.0268], + ..., + [-0.0814, -0.0917, -0.0321, ..., -0.0199, 0.0522, 0.0261], + [ 0.0065, 0.0142, -0.0115, ..., -0.1784, -0.1959, -0.2123], + [ 0.1163, 0.1315, 0.1102, ..., -0.0206, -0.0112, 0.0235]], + device='cuda:0'), grad: tensor([[ 5.9433e-03, 2.8019e-03, 2.9926e-03, ..., 1.8187e-03, + 1.3828e-03, 1.1148e-03], + [-5.0812e-02, -2.6550e-02, -2.8870e-02, ..., -1.4725e-02, + -1.0307e-02, -8.5678e-03], + [ 4.2009e-04, 1.5187e-04, 1.5056e-04, ..., 1.4293e-04, + 1.2743e-04, 9.7096e-05], + ..., + [ 5.1514e-02, 2.4612e-02, 2.6321e-02, ..., 1.5656e-02, + 1.1826e-02, 9.5596e-03], + [-1.1978e-02, -2.7523e-03, -2.3022e-03, ..., -4.5547e-03, + -4.4975e-03, -3.3283e-03], + [ 3.6449e-03, 1.3599e-03, 1.3704e-03, ..., 1.2245e-03, + 1.0624e-03, 8.1873e-04]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0341, 0.0254, -0.0705, -0.0229, 0.1050, -0.0339, 0.0366], + device='cuda:0'), grad: tensor([ 0.0132, -0.0865, 0.0014, 0.0052, 0.1113, -0.0564, 0.0118], + device='cuda:0') +306 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 432.30, cls_loss 0.6247 cls_loss_mapping 0.7047 cls_loss_causal 1.2669 re_mapping 0.0851 re_causal 0.0840 /// teacc 72.12 lr 0.00987464 +Epoch 6, weight, value: tensor([[ 0.1957, 0.1933, 0.2247, ..., -0.0115, 0.0068, -0.0090], + [-0.1062, -0.1071, -0.1227, ..., 0.0857, 0.0783, 0.0707], + [-0.0818, -0.0668, -0.0912, ..., 0.0748, 0.0483, 0.0343], + ..., + [-0.0890, -0.1005, -0.0373, ..., -0.0144, 0.0582, 0.0325], + [-0.0086, 0.0016, -0.0259, ..., -0.1826, -0.1993, -0.2163], + [ 0.1395, 0.1505, 0.1312, ..., -0.0195, -0.0080, 0.0255]], + device='cuda:0'), grad: tensor([[ 0.0750, 0.0472, 0.0440, ..., 0.0164, 0.0148, 0.0158], + [-0.0565, -0.0350, -0.0352, ..., -0.0148, -0.0146, -0.0162], + [ 0.0156, 0.0057, 0.0061, ..., 0.0040, 0.0032, 0.0034], + ..., + [ 0.0461, 0.0297, 0.0238, ..., 0.0068, 0.0044, 0.0037], + [-0.0116, -0.0024, -0.0030, ..., -0.0031, -0.0021, -0.0022], + [-0.0692, -0.0453, -0.0359, ..., -0.0096, -0.0058, -0.0047]], + device='cuda:0') +Epoch 6, bias, value: tensor([-0.0279, 0.0292, -0.0618, -0.0331, 0.0900, -0.0454, 0.0549], + device='cuda:0'), grad: tensor([ 0.1162, -0.0898, 0.0411, 0.0019, 0.0696, -0.0381, -0.1008], + device='cuda:0') +306 +0.009874639560909117 +changing lr +epoch 5, time 426.44, cls_loss 0.4380 cls_loss_mapping 0.5281 cls_loss_causal 1.2226 re_mapping 0.0861 re_causal 0.0850 /// teacc 42.31 lr 0.00981981 +Epoch 7, weight, value: tensor([[ 0.2098, 0.2093, 0.2438, ..., -0.0112, 0.0055, -0.0106], + [-0.0974, -0.0969, -0.1146, ..., 0.0915, 0.0849, 0.0769], + [-0.0816, -0.0701, -0.0943, ..., 0.0685, 0.0410, 0.0278], + ..., + [-0.1026, -0.1140, -0.0502, ..., -0.0106, 0.0625, 0.0370], + [-0.0079, 0.0012, -0.0266, ..., -0.1819, -0.1995, -0.2156], + [ 0.1374, 0.1491, 0.1289, ..., -0.0233, -0.0101, 0.0223]], + device='cuda:0'), grad: tensor([[-0.0735, -0.0327, -0.0339, ..., -0.0254, -0.0235, -0.0245], + [ 0.0145, 0.0052, 0.0056, ..., 0.0056, 0.0052, 0.0054], + [-0.0057, -0.0010, -0.0008, ..., -0.0023, -0.0022, -0.0022], + ..., + [ 0.0115, 0.0060, 0.0061, ..., 0.0034, 0.0032, 0.0033], + [ 0.0025, 0.0004, 0.0004, ..., 0.0010, 0.0009, 0.0009], + [ 0.0484, 0.0214, 0.0222, ..., 0.0167, 0.0155, 0.0161]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0215, 0.0466, -0.0830, -0.0234, 0.0787, -0.0530, 0.0614], + device='cuda:0'), grad: tensor([-0.1722, 0.0393, -0.0172, 0.0070, 0.0224, 0.0075, 0.1133], + device='cuda:0') +306 +0.009819814303479266 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 427.73, cls_loss 0.3658 cls_loss_mapping 0.4601 cls_loss_causal 1.1575 re_mapping 0.0819 re_causal 0.0808 /// teacc 78.85 lr 0.00975528 +Epoch 8, weight, value: tensor([[ 0.2240, 0.2268, 0.2634, ..., -0.0166, 0.0008, -0.0141], + [-0.0867, -0.0825, -0.1027, ..., 0.0913, 0.0859, 0.0786], + [-0.0880, -0.0760, -0.0994, ..., 0.0664, 0.0399, 0.0277], + ..., + [-0.1128, -0.1233, -0.0609, ..., -0.0075, 0.0650, 0.0393], + [-0.0067, 0.0008, -0.0251, ..., -0.1829, -0.1997, -0.2169], + [ 0.1252, 0.1315, 0.1129, ..., -0.0217, -0.0112, 0.0210]], + device='cuda:0'), grad: tensor([[-1.6870e-03, -1.3332e-03, -9.8896e-04, ..., -2.5082e-04, + -2.2459e-04, -2.6464e-04], + [ 2.4211e-04, 6.1810e-05, 4.8459e-05, ..., 1.4257e-04, + 1.3053e-04, 1.4758e-04], + [ 2.3880e-03, 1.4086e-03, 1.0519e-03, ..., 7.5197e-04, + 6.8378e-04, 7.8249e-04], + ..., + [ 8.5688e-04, 1.7285e-04, 1.3876e-04, ..., 5.4407e-04, + 4.9782e-04, 5.6219e-04], + [-1.4257e-04, -3.1739e-05, -2.5257e-06, ..., -5.1796e-05, + -5.4002e-05, -5.2422e-05], + [ 4.9734e-04, 1.1754e-04, 9.1851e-05, ..., 3.0231e-04, + 2.7633e-04, 3.1233e-04]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0263, 0.0549, -0.0846, -0.0161, 0.0702, -0.0534, 0.0610], + device='cuda:0'), grad: tensor([-0.0018, 0.0007, 0.0042, -0.0068, 0.0026, -0.0004, 0.0015], + device='cuda:0') +306 +0.009755282581475767 +changing lr +epoch 7, time 428.13, cls_loss 0.3124 cls_loss_mapping 0.3850 cls_loss_causal 1.0907 re_mapping 0.0829 re_causal 0.0819 /// teacc 77.88 lr 0.00968117 +Epoch 9, weight, value: tensor([[ 0.2525, 0.2555, 0.2922, ..., -0.0049, 0.0127, -0.0029], + [-0.0902, -0.0873, -0.1086, ..., 0.0879, 0.0829, 0.0759], + [-0.0888, -0.0797, -0.1016, ..., 0.0671, 0.0397, 0.0283], + ..., + [-0.1185, -0.1264, -0.0642, ..., -0.0080, 0.0637, 0.0385], + [-0.0113, -0.0028, -0.0307, ..., -0.1846, -0.2024, -0.2191], + [ 0.1178, 0.1233, 0.1060, ..., -0.0286, -0.0175, 0.0142]], + device='cuda:0'), grad: tensor([[ 2.7314e-05, -1.1764e-05, -1.4231e-05, ..., 2.3901e-05, + 2.5332e-05, 2.7269e-05], + [ 3.2067e-04, 1.0395e-04, 6.3479e-05, ..., 1.3578e-04, + 1.5271e-04, 1.6749e-04], + [-8.8024e-04, -2.0969e-04, -1.2898e-04, ..., -3.6907e-04, + -3.9673e-04, -4.4155e-04], + ..., + [ 2.5272e-05, 4.2580e-06, 2.9281e-06, ..., 1.1168e-05, + 1.1407e-05, 1.2547e-05], + [ 5.6177e-05, 1.7658e-05, 1.0453e-05, ..., 2.3872e-05, + 2.6867e-05, 2.9534e-05], + [ 8.3685e-05, 2.6420e-05, 2.0012e-05, ..., 3.2485e-05, + 3.4392e-05, 3.7998e-05]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0317, 0.0606, -0.0707, -0.0167, 0.0594, -0.0630, 0.0678], + device='cuda:0'), grad: tensor([ 1.6165e-04, 1.2770e-03, -3.2120e-03, 1.1988e-03, 8.2850e-05, + 2.2674e-04, 2.6727e-04], device='cuda:0') +306 +0.009681174353198686 +changing lr +epoch 8, time 424.95, cls_loss 0.2002 cls_loss_mapping 0.3516 cls_loss_causal 1.0047 re_mapping 0.0820 re_causal 0.0811 /// teacc 33.65 lr 0.00959764 +Epoch 10, weight, value: tensor([[ 0.2614, 0.2600, 0.2987, ..., -0.0098, 0.0075, -0.0080], + [-0.0923, -0.0862, -0.1080, ..., 0.0822, 0.0782, 0.0712], + [-0.0955, -0.0838, -0.1060, ..., 0.0645, 0.0378, 0.0267], + ..., + [-0.1192, -0.1275, -0.0653, ..., -0.0048, 0.0652, 0.0409], + [-0.0104, -0.0028, -0.0313, ..., -0.1847, -0.2024, -0.2193], + [ 0.1168, 0.1226, 0.1041, ..., -0.0272, -0.0155, 0.0157]], + device='cuda:0'), grad: tensor([[ 2.3193e-03, 1.1292e-03, 1.1568e-03, ..., 1.0481e-03, + 1.2121e-03, 1.2636e-03], + [ 2.1305e-03, 1.9920e-04, 2.6274e-04, ..., 1.1711e-03, + 1.3361e-03, 1.3800e-03], + [-5.6152e-03, -1.1911e-03, -1.2016e-03, ..., -2.9316e-03, + -3.5038e-03, -3.5591e-03], + ..., + [-1.7195e-03, -1.0481e-03, -1.0672e-03, ..., -7.2527e-04, + -8.3351e-04, -8.7643e-04], + [-7.3552e-05, -1.5363e-05, -6.2399e-06, ..., -2.7969e-05, + -3.7313e-05, -3.1382e-05], + [ 4.4250e-04, 7.7963e-05, 8.1897e-05, ..., 2.3389e-04, + 2.7561e-04, 2.8086e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0325, 0.0560, -0.0762, -0.0128, 0.0681, -0.0589, 0.0619], + device='cuda:0'), grad: tensor([ 0.0045, 0.0079, -0.0192, 0.0078, -0.0023, -0.0003, 0.0016], + device='cuda:0') +306 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 428.66, cls_loss 0.1527 cls_loss_mapping 0.2936 cls_loss_causal 0.9805 re_mapping 0.0773 re_causal 0.0766 /// teacc 84.13 lr 0.00950484 +Epoch 11, weight, value: tensor([[ 0.2543, 0.2539, 0.2933, ..., -0.0140, 0.0034, -0.0122], + [-0.0843, -0.0777, -0.1006, ..., 0.0825, 0.0791, 0.0716], + [-0.0936, -0.0842, -0.1065, ..., 0.0610, 0.0347, 0.0243], + ..., + [-0.1203, -0.1270, -0.0658, ..., -0.0010, 0.0682, 0.0443], + [-0.0038, 0.0005, -0.0281, ..., -0.1811, -0.1990, -0.2157], + [ 0.1156, 0.1199, 0.1026, ..., -0.0263, -0.0146, 0.0164]], + device='cuda:0'), grad: tensor([[ 2.4765e-02, 6.7558e-03, 6.9962e-03, ..., 1.8234e-02, + 1.8661e-02, 1.9272e-02], + [ 4.1313e-03, 1.1272e-03, 1.1683e-03, ..., 3.0441e-03, + 3.1166e-03, 3.2177e-03], + [ 7.8964e-03, 2.1553e-03, 2.2316e-03, ..., 5.8098e-03, + 5.9509e-03, 6.1417e-03], + ..., + [-3.7140e-02, -1.0139e-02, -1.0498e-02, ..., -2.7344e-02, + -2.7985e-02, -2.8900e-02], + [ 1.7428e-04, 4.7594e-05, 4.9263e-05, ..., 1.2815e-04, + 1.3125e-04, 1.3554e-04], + [ 8.5831e-05, 2.3320e-05, 2.4110e-05, ..., 6.3360e-05, + 6.4850e-05, 6.6936e-05]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0377, 0.0606, -0.0670, -0.0292, 0.0683, -0.0484, 0.0588], + device='cuda:0'), grad: tensor([ 0.0754, 0.0126, 0.0241, 0.0003, -0.1132, 0.0005, 0.0003], + device='cuda:0') +306 +0.009504844339512096 +changing lr +epoch 10, time 432.92, cls_loss 0.1024 cls_loss_mapping 0.2664 cls_loss_causal 0.9671 re_mapping 0.0781 re_causal 0.0776 /// teacc 78.85 lr 0.00940298 +Epoch 12, weight, value: tensor([[ 0.2556, 0.2549, 0.2939, ..., -0.0154, 0.0019, -0.0134], + [-0.0869, -0.0772, -0.1012, ..., 0.0757, 0.0722, 0.0644], + [-0.0960, -0.0876, -0.1101, ..., 0.0613, 0.0361, 0.0255], + ..., + [-0.1176, -0.1258, -0.0646, ..., 0.0041, 0.0723, 0.0490], + [-0.0108, -0.0027, -0.0305, ..., -0.1790, -0.1969, -0.2136], + [ 0.1190, 0.1213, 0.1047, ..., -0.0270, -0.0154, 0.0159]], + device='cuda:0'), grad: tensor([[ 4.6939e-06, -6.5342e-06, -7.6517e-06, ..., 2.5146e-06, + 3.9227e-06, 4.4741e-06], + [ 2.5302e-05, 8.0988e-06, 5.8897e-06, ..., 4.1500e-06, + 5.7928e-06, 6.3777e-06], + [ 6.5279e-04, 9.4473e-05, 5.7399e-05, ..., 3.0136e-04, + 3.3832e-04, 3.9124e-04], + ..., + [-3.4866e-03, -3.1996e-04, -1.4079e-04, ..., -1.8482e-03, + -2.0466e-03, -2.3766e-03], + [-6.5207e-05, -2.3693e-05, -1.9684e-05, ..., -6.7875e-06, + -3.9041e-06, -1.0878e-06], + [-1.3351e-04, -4.1932e-05, -2.8417e-05, ..., -1.0617e-05, + -1.9982e-05, -2.1055e-05]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0359, 0.0466, -0.0613, -0.0207, 0.0722, -0.0636, 0.0679], + device='cuda:0'), grad: tensor([ 3.7521e-05, 6.2823e-05, 1.5831e-03, 7.4615e-03, -8.5754e-03, + -2.2757e-04, -3.4595e-04], device='cuda:0') +306 +0.009402977659283692 +changing lr +epoch 11, time 428.82, cls_loss 0.1045 cls_loss_mapping 0.2426 cls_loss_causal 0.9590 re_mapping 0.0746 re_causal 0.0742 /// teacc 73.08 lr 0.00929224 +Epoch 13, weight, value: tensor([[ 0.2544, 0.2549, 0.2941, ..., -0.0189, -0.0027, -0.0172], + [-0.0718, -0.0682, -0.0925, ..., 0.0745, 0.0723, 0.0643], + [-0.0980, -0.0885, -0.1105, ..., 0.0600, 0.0352, 0.0247], + ..., + [-0.1210, -0.1289, -0.0685, ..., 0.0058, 0.0737, 0.0503], + [-0.0109, -0.0042, -0.0327, ..., -0.1759, -0.1932, -0.2096], + [ 0.1122, 0.1174, 0.1006, ..., -0.0270, -0.0160, 0.0152]], + device='cuda:0'), grad: tensor([[-3.5858e-03, -1.9474e-03, -1.8644e-03, ..., -1.1988e-03, + -1.4086e-03, -1.4277e-03], + [ 3.4409e-03, 1.7090e-03, 1.6346e-03, ..., 1.1806e-03, + 1.3924e-03, 1.4219e-03], + [-1.8692e-03, -2.7037e-04, -2.5535e-04, ..., -6.8378e-04, + -8.3160e-04, -9.0551e-04], + ..., + [ 3.8457e-04, 1.0127e-04, 9.6619e-05, ..., 1.2803e-04, + 1.5509e-04, 1.6749e-04], + [ 8.3780e-04, 1.4365e-04, 1.3626e-04, ..., 3.0255e-04, + 3.6740e-04, 3.9887e-04], + [ 1.8275e-04, 4.3809e-05, 4.1515e-05, ..., 6.1333e-05, + 7.4029e-05, 8.0884e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0398, 0.0663, -0.0684, -0.0271, 0.0654, -0.0569, 0.0658], + device='cuda:0'), grad: tensor([-0.0050, 0.0056, -0.0061, 0.0013, 0.0010, 0.0026, 0.0005], + device='cuda:0') +306 +0.009292243968009333 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 433.28, cls_loss 0.1435 cls_loss_mapping 0.2366 cls_loss_causal 0.9508 re_mapping 0.0761 re_causal 0.0760 /// teacc 84.62 lr 0.00917287 +Epoch 14, weight, value: tensor([[ 0.2648, 0.2631, 0.3023, ..., -0.0203, -0.0039, -0.0183], + [-0.0708, -0.0647, -0.0908, ..., 0.0738, 0.0713, 0.0633], + [-0.1016, -0.0902, -0.1106, ..., 0.0580, 0.0337, 0.0228], + ..., + [-0.1377, -0.1389, -0.0794, ..., 0.0060, 0.0728, 0.0497], + [-0.0093, -0.0050, -0.0338, ..., -0.1737, -0.1908, -0.2077], + [ 0.1134, 0.1155, 0.1006, ..., -0.0270, -0.0161, 0.0155]], + device='cuda:0'), grad: tensor([[-1.0452e-02, -4.9820e-03, -4.7569e-03, ..., -1.1320e-03, + -1.4286e-03, -1.5659e-03], + [ 9.8109e-05, 4.7505e-05, 4.5270e-05, ..., 1.0222e-05, + 1.2979e-05, 1.4357e-05], + [ 4.1217e-05, 1.9357e-05, 1.8522e-05, ..., 4.6566e-06, + 5.8375e-06, 6.3404e-06], + ..., + [ 1.2827e-03, 6.1560e-04, 5.8699e-04, ..., 1.3649e-04, + 1.7273e-04, 1.9014e-04], + [ 6.7101e-03, 3.0994e-03, 2.9678e-03, ..., 7.8487e-04, + 9.7942e-04, 1.0557e-03], + [ 1.6441e-03, 8.8120e-04, 8.3113e-04, ..., 1.2231e-04, + 1.6546e-04, 1.9848e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0348, 0.0573, -0.0797, -0.0097, 0.0495, -0.0497, 0.0722], + device='cuda:0'), grad: tensor([-1.6800e-02, 1.5473e-04, 6.7532e-05, 1.0681e-03, 2.0447e-03, + 1.1192e-02, 2.2488e-03], device='cuda:0') +306 +0.009172866268606516 +changing lr +epoch 13, time 428.41, cls_loss 0.0724 cls_loss_mapping 0.2087 cls_loss_causal 0.9581 re_mapping 0.0757 re_causal 0.0758 /// teacc 82.21 lr 0.00904508 +Epoch 15, weight, value: tensor([[ 0.2678, 0.2653, 0.3044, ..., -0.0203, -0.0042, -0.0183], + [-0.0688, -0.0640, -0.0901, ..., 0.0728, 0.0709, 0.0626], + [-0.1001, -0.0903, -0.1106, ..., 0.0604, 0.0366, 0.0265], + ..., + [-0.1351, -0.1369, -0.0781, ..., 0.0080, 0.0743, 0.0514], + [-0.0108, -0.0065, -0.0349, ..., -0.1723, -0.1900, -0.2064], + [ 0.1121, 0.1141, 0.0995, ..., -0.0277, -0.0169, 0.0143]], + device='cuda:0'), grad: tensor([[ 1.1940e-02, 4.6654e-03, 3.6144e-03, ..., 3.5648e-03, + 3.6526e-03, 4.1809e-03], + [ 7.1526e-04, 2.5606e-04, 1.9670e-04, ..., 2.4796e-04, + 2.5439e-04, 2.9278e-04], + [ 3.0756e-04, 9.9123e-05, 7.3135e-05, ..., 1.2970e-04, + 1.3614e-04, 1.5116e-04], + ..., + [-1.6800e-02, -6.4583e-03, -4.9934e-03, ..., -5.1765e-03, + -5.3024e-03, -6.0768e-03], + [ 3.9649e-04, 1.4973e-04, 1.1551e-04, ..., 1.2648e-04, + 1.2982e-04, 1.4853e-04], + [ 3.3360e-03, 1.2503e-03, 9.6607e-04, ..., 1.0710e-03, + 1.0958e-03, 1.2627e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0334, 0.0568, -0.0657, -0.0270, 0.0542, -0.0515, 0.0717], + device='cuda:0'), grad: tensor([ 0.0220, 0.0015, 0.0007, 0.0002, -0.0316, 0.0008, 0.0065], + device='cuda:0') +306 +0.00904508497187474 +changing lr +epoch 14, time 427.18, cls_loss 0.0752 cls_loss_mapping 0.2063 cls_loss_causal 0.9385 re_mapping 0.0745 re_causal 0.0746 /// teacc 77.88 lr 0.00890916 +Epoch 16, weight, value: tensor([[ 0.2597, 0.2606, 0.2997, ..., -0.0224, -0.0069, -0.0210], + [-0.0762, -0.0676, -0.0934, ..., 0.0694, 0.0677, 0.0592], + [-0.0925, -0.0868, -0.1069, ..., 0.0608, 0.0379, 0.0282], + ..., + [-0.1296, -0.1333, -0.0759, ..., 0.0107, 0.0755, 0.0534], + [-0.0116, -0.0077, -0.0362, ..., -0.1713, -0.1888, -0.2052], + [ 0.1154, 0.1160, 0.1024, ..., -0.0267, -0.0156, 0.0152]], + device='cuda:0'), grad: tensor([[-1.1276e-02, -7.5569e-03, -7.1335e-03, ..., -9.6512e-04, + -1.2064e-03, -1.4009e-03], + [ 9.4399e-06, 6.8583e-06, 6.3330e-06, ..., 1.8068e-06, + 1.8124e-06, 1.7043e-06], + [ 3.1776e-03, 1.9855e-03, 1.8578e-03, ..., 3.5858e-04, + 4.3797e-04, 4.9114e-04], + ..., + [-5.6791e-04, -1.5104e-04, -1.1533e-04, ..., -1.8704e-04, + -2.1851e-04, -2.2566e-04], + [ 1.1760e-04, 4.9055e-05, 4.1932e-05, ..., 2.5496e-05, + 3.0220e-05, 3.1352e-05], + [ 8.4915e-03, 5.6458e-03, 5.3215e-03, ..., 7.5531e-04, + 9.4271e-04, 1.0891e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0451, 0.0468, -0.0566, -0.0274, 0.0629, -0.0538, 0.0782], + device='cuda:0'), grad: tensor([-1.2550e-02, 7.4729e-06, 4.0016e-03, 7.3373e-05, -1.3828e-03, + 2.2531e-04, 9.6130e-03], device='cuda:0') +306 +0.008909157412340152 +changing lr +epoch 15, time 429.27, cls_loss 0.0493 cls_loss_mapping 0.1644 cls_loss_causal 0.8087 re_mapping 0.0737 re_causal 0.0738 /// teacc 84.13 lr 0.00876536 +Epoch 17, weight, value: tensor([[ 0.2622, 0.2626, 0.3013, ..., -0.0197, -0.0045, -0.0183], + [-0.0737, -0.0665, -0.0920, ..., 0.0687, 0.0676, 0.0594], + [-0.1002, -0.0888, -0.1083, ..., 0.0566, 0.0336, 0.0238], + ..., + [-0.1308, -0.1338, -0.0773, ..., 0.0105, 0.0745, 0.0524], + [-0.0091, -0.0070, -0.0355, ..., -0.1685, -0.1859, -0.2023], + [ 0.1094, 0.1124, 0.0990, ..., -0.0280, -0.0173, 0.0129]], + device='cuda:0'), grad: tensor([[ 3.2349e-03, 9.7322e-04, 1.0643e-03, ..., 9.9182e-04, + 1.1444e-03, 1.1063e-03], + [ 1.0931e-04, 3.7700e-05, 2.9683e-05, ..., 3.7044e-05, + 4.3064e-05, 4.2886e-05], + [ 3.2806e-04, 1.1384e-04, 1.1349e-04, ..., 1.0669e-04, + 1.2398e-04, 1.2791e-04], + ..., + [-9.7275e-03, -3.2368e-03, -3.2558e-03, ..., -3.1281e-03, + -3.6716e-03, -3.7193e-03], + [ 2.3975e-03, 7.9107e-04, 8.0633e-04, ..., 7.6532e-04, + 8.9788e-04, 9.0694e-04], + [ 4.2582e-04, 1.5390e-04, 1.4496e-04, ..., 1.4293e-04, + 1.7035e-04, 1.7917e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0407, 0.0508, -0.0761, -0.0104, 0.0615, -0.0460, 0.0660], + device='cuda:0'), grad: tensor([ 0.0076, 0.0002, 0.0007, 0.0070, -0.0219, 0.0054, 0.0009], + device='cuda:0') +306 +0.00876535733001806 +changing lr +epoch 16, time 430.23, cls_loss 0.0479 cls_loss_mapping 0.1632 cls_loss_causal 0.8588 re_mapping 0.0722 re_causal 0.0726 /// teacc 83.65 lr 0.00861397 +Epoch 18, weight, value: tensor([[ 0.2592, 0.2595, 0.2979, ..., -0.0207, -0.0055, -0.0192], + [-0.0728, -0.0663, -0.0919, ..., 0.0676, 0.0669, 0.0586], + [-0.0975, -0.0859, -0.1046, ..., 0.0558, 0.0330, 0.0236], + ..., + [-0.1267, -0.1304, -0.0751, ..., 0.0126, 0.0759, 0.0540], + [-0.0114, -0.0086, -0.0370, ..., -0.1673, -0.1846, -0.2010], + [ 0.1103, 0.1118, 0.0986, ..., -0.0272, -0.0167, 0.0132]], + device='cuda:0'), grad: tensor([[ 1.6737e-04, 5.5701e-05, 5.5939e-05, ..., 5.0575e-05, + 5.4538e-05, 6.0648e-05], + [ 4.0913e-04, 1.3793e-04, 1.3995e-04, ..., 1.1837e-04, + 1.2434e-04, 1.3864e-04], + [-6.3133e-03, -1.0328e-03, -1.0738e-03, ..., -2.9068e-03, + -2.9488e-03, -3.3131e-03], + ..., + [ 5.3711e-03, 8.8596e-04, 9.2173e-04, ..., 2.4643e-03, + 2.5005e-03, 2.8095e-03], + [ 1.8072e-04, 4.8667e-05, 4.9621e-05, ..., 6.4075e-05, + 6.6817e-05, 7.4565e-05], + [-3.5214e-04, -2.0921e-04, -2.0885e-04, ..., -1.7747e-05, + -3.2753e-05, -3.3498e-05]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0402, 0.0505, -0.0748, -0.0158, 0.0623, -0.0484, 0.0714], + device='cuda:0'), grad: tensor([ 0.0004, 0.0009, -0.0178, 0.0014, 0.0151, 0.0004, -0.0005], + device='cuda:0') +306 +0.008613974319136962 +changing lr +epoch 17, time 429.92, cls_loss 0.0243 cls_loss_mapping 0.1391 cls_loss_causal 0.8789 re_mapping 0.0726 re_causal 0.0731 /// teacc 83.17 lr 0.00845531 +Epoch 19, weight, value: tensor([[ 0.2583, 0.2586, 0.2964, ..., -0.0196, -0.0048, -0.0185], + [-0.0752, -0.0660, -0.0914, ..., 0.0652, 0.0644, 0.0560], + [-0.0976, -0.0868, -0.1054, ..., 0.0534, 0.0304, 0.0213], + ..., + [-0.1255, -0.1298, -0.0747, ..., 0.0132, 0.0766, 0.0549], + [-0.0113, -0.0088, -0.0373, ..., -0.1651, -0.1825, -0.1985], + [ 0.1100, 0.1122, 0.0997, ..., -0.0276, -0.0173, 0.0123]], + device='cuda:0'), grad: tensor([[ 1.8403e-06, -8.6240e-07, -1.1362e-06, ..., 1.7639e-06, + 1.7975e-06, 2.0228e-06], + [ 8.0653e-07, 2.6636e-07, 1.9558e-07, ..., 2.4587e-07, + 2.5518e-07, 2.9616e-07], + [ 1.5311e-06, 3.5390e-07, 1.6950e-07, ..., 6.5006e-07, + 5.9977e-07, 7.3202e-07], + ..., + [-3.5584e-05, -6.4559e-06, -2.1216e-06, ..., -1.6674e-05, + -1.6719e-05, -1.9029e-05], + [-7.9721e-07, -1.6205e-07, -3.1292e-07, ..., 3.2410e-07, + 2.8685e-07, 2.4214e-07], + [ 9.9838e-07, 6.4634e-07, 5.4948e-07, ..., 1.3597e-07, + 1.3970e-07, 1.5274e-07]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0405, 0.0421, -0.0691, -0.0113, 0.0613, -0.0455, 0.0680], + device='cuda:0'), grad: tensor([ 8.4639e-06, 1.9073e-06, 3.9414e-06, 8.4400e-05, -9.7275e-05, + -2.9840e-06, 1.5069e-06], device='cuda:0') +306 +0.008455313244934327 +changing lr +epoch 18, time 426.88, cls_loss 0.0267 cls_loss_mapping 0.1306 cls_loss_causal 0.8283 re_mapping 0.0700 re_causal 0.0706 /// teacc 84.62 lr 0.00828969 +Epoch 20, weight, value: tensor([[ 0.2555, 0.2564, 0.2938, ..., -0.0193, -0.0046, -0.0181], + [-0.0724, -0.0644, -0.0896, ..., 0.0648, 0.0644, 0.0561], + [-0.0933, -0.0848, -0.1034, ..., 0.0538, 0.0313, 0.0223], + ..., + [-0.1238, -0.1286, -0.0743, ..., 0.0136, 0.0762, 0.0548], + [-0.0132, -0.0105, -0.0388, ..., -0.1628, -0.1805, -0.1963], + [ 0.1110, 0.1128, 0.1005, ..., -0.0283, -0.0180, 0.0112]], + device='cuda:0'), grad: tensor([[ 2.0943e-03, 6.0558e-04, 6.5231e-04, ..., 1.0023e-03, + 9.3126e-04, 1.0624e-03], + [ 3.7360e-04, 1.0639e-04, 1.0651e-04, ..., 1.6189e-04, + 1.5295e-04, 1.7846e-04], + [ 3.8662e-03, 9.4938e-04, 7.3290e-04, ..., 1.3113e-03, + 1.5354e-03, 1.6270e-03], + ..., + [-8.3771e-03, -2.2507e-03, -2.1477e-03, ..., -3.5439e-03, + -3.6964e-03, -4.0207e-03], + [ 4.0007e-04, 1.0943e-04, 1.0467e-04, ..., 1.6475e-04, + 1.6594e-04, 1.8454e-04], + [ 8.8573e-05, 3.6508e-05, 7.8022e-05, ..., 1.6296e-04, + 1.9848e-04, 1.7130e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0415, 0.0456, -0.0612, -0.0222, 0.0607, -0.0479, 0.0713], + device='cuda:0'), grad: tensor([ 0.0055, 0.0010, 0.0094, 0.0041, -0.0212, 0.0010, 0.0003], + device='cuda:0') +306 +0.008289693629698565 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 432.12, cls_loss 0.0267 cls_loss_mapping 0.1288 cls_loss_causal 0.8148 re_mapping 0.0707 re_causal 0.0716 /// teacc 86.06 lr 0.00811745 +Epoch 21, weight, value: tensor([[ 0.2600, 0.2587, 0.2960, ..., -0.0177, -0.0030, -0.0163], + [-0.0687, -0.0632, -0.0889, ..., 0.0651, 0.0654, 0.0569], + [-0.0958, -0.0865, -0.1049, ..., 0.0533, 0.0306, 0.0220], + ..., + [-0.1248, -0.1282, -0.0745, ..., 0.0130, 0.0747, 0.0536], + [-0.0146, -0.0111, -0.0389, ..., -0.1617, -0.1793, -0.1949], + [ 0.1092, 0.1118, 0.0995, ..., -0.0295, -0.0196, 0.0092]], + device='cuda:0'), grad: tensor([[ 1.8910e-05, 6.5379e-06, 4.4592e-06, ..., 6.7987e-06, + 7.1824e-06, 7.8008e-06], + [-1.4128e-06, -1.0086e-06, -1.0710e-06, ..., -6.9104e-07, + -1.0608e-06, -9.7603e-07], + [ 4.8101e-05, 1.2361e-05, 5.9381e-06, ..., 8.7246e-06, + 5.6922e-06, 8.7470e-06], + ..., + [-6.1803e-06, -3.6955e-06, -3.3882e-06, ..., -5.7295e-06, + -7.3798e-06, -7.0296e-06], + [-1.5056e-04, -3.7521e-05, -1.7092e-05, ..., -2.5272e-05, + -1.4730e-05, -2.4661e-05], + [ 3.8624e-05, 1.0028e-05, 4.9323e-06, ..., 6.9328e-06, + 4.5486e-06, 6.9775e-06]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0320, 0.0540, -0.0647, -0.0240, 0.0552, -0.0512, 0.0674], + device='cuda:0'), grad: tensor([ 4.3154e-05, -3.6228e-07, 1.2267e-04, 1.3447e-04, -1.0043e-05, + -3.8791e-04, 9.7752e-05], device='cuda:0') +306 +0.00811744900929367 +changing lr +epoch 20, time 429.61, cls_loss 0.0334 cls_loss_mapping 0.1398 cls_loss_causal 0.8032 re_mapping 0.0706 re_causal 0.0716 /// teacc 84.62 lr 0.00793893 +Epoch 22, weight, value: tensor([[ 0.2576, 0.2569, 0.2942, ..., -0.0177, -0.0028, -0.0160], + [-0.0645, -0.0590, -0.0853, ..., 0.0635, 0.0634, 0.0551], + [-0.0925, -0.0848, -0.1032, ..., 0.0532, 0.0312, 0.0227], + ..., + [-0.1247, -0.1288, -0.0756, ..., 0.0144, 0.0759, 0.0548], + [-0.0171, -0.0135, -0.0409, ..., -0.1600, -0.1781, -0.1935], + [ 0.1069, 0.1103, 0.0983, ..., -0.0300, -0.0205, 0.0080]], + device='cuda:0'), grad: tensor([[-8.5533e-05, -6.3598e-05, -6.1393e-05, ..., -1.3635e-05, + -1.5028e-05, -1.5661e-05], + [-9.6858e-05, -2.0891e-05, -1.3158e-05, ..., -1.9193e-05, + -2.6330e-05, -3.3528e-05], + [ 1.8907e-04, 4.3601e-05, 3.0041e-05, ..., 4.0948e-05, + 4.7296e-05, 5.7191e-05], + ..., + [ 8.9645e-05, 4.6551e-05, 4.2140e-05, ..., 1.4164e-05, + 1.6123e-05, 1.9029e-05], + [ 3.4887e-06, 1.4324e-06, 1.4063e-06, ..., -2.2911e-07, + 7.2271e-07, 9.7789e-07], + [-1.1152e-04, -9.6262e-06, -6.3889e-07, ..., -2.4766e-05, + -2.5764e-05, -3.1590e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0325, 0.0521, -0.0610, -0.0267, 0.0597, -0.0502, 0.0632], + device='cuda:0'), grad: tensor([-9.4354e-05, -2.4581e-04, 4.6301e-04, 2.9758e-05, 1.5485e-04, + 6.5416e-06, -3.1376e-04], device='cuda:0') +306 +0.007938926261462368 +changing lr +epoch 21, time 426.58, cls_loss 0.0379 cls_loss_mapping 0.1226 cls_loss_causal 0.8209 re_mapping 0.0672 re_causal 0.0683 /// teacc 80.77 lr 0.00775448 +Epoch 23, weight, value: tensor([[ 0.2543, 0.2548, 0.2918, ..., -0.0181, -0.0035, -0.0166], + [-0.0610, -0.0557, -0.0815, ..., 0.0627, 0.0633, 0.0549], + [-0.0944, -0.0856, -0.1037, ..., 0.0523, 0.0305, 0.0222], + ..., + [-0.1257, -0.1288, -0.0764, ..., 0.0143, 0.0749, 0.0540], + [-0.0155, -0.0136, -0.0409, ..., -0.1578, -0.1759, -0.1909], + [ 0.1065, 0.1089, 0.0971, ..., -0.0297, -0.0205, 0.0078]], + device='cuda:0'), grad: tensor([[-2.0206e-05, -1.4797e-05, -1.2912e-05, ..., -2.2482e-06, + -2.2426e-06, -2.8517e-06], + [ 3.0920e-07, 3.2224e-07, 3.0361e-07, ..., -7.0781e-08, + -1.7323e-07, -1.7136e-07], + [ 8.7991e-06, 3.8929e-06, 3.2075e-06, ..., 2.0005e-06, + 2.0061e-06, 2.2966e-06], + ..., + [-2.0675e-07, 1.2089e-06, 1.0785e-06, ..., -2.2911e-06, + -2.1961e-06, -2.2668e-06], + [-6.8881e-06, -1.1437e-06, -6.6683e-07, ..., -1.4920e-06, + -1.4491e-06, -1.6559e-06], + [ 1.3739e-05, 9.4026e-06, 8.1286e-06, ..., 1.7323e-06, + 1.7434e-06, 2.1514e-06]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0375, 0.0531, -0.0654, -0.0235, 0.0556, -0.0433, 0.0657], + device='cuda:0'), grad: tensor([-1.7449e-05, -2.8312e-07, 1.7524e-05, 1.3448e-05, -7.6182e-06, + -2.0295e-05, 1.4588e-05], device='cuda:0') +306 +0.007754484907260515 +changing lr +epoch 22, time 426.67, cls_loss 0.0169 cls_loss_mapping 0.1011 cls_loss_causal 0.7419 re_mapping 0.0658 re_causal 0.0668 /// teacc 85.58 lr 0.00756450 +Epoch 24, weight, value: tensor([[ 0.2518, 0.2533, 0.2901, ..., -0.0181, -0.0037, -0.0170], + [-0.0619, -0.0557, -0.0812, ..., 0.0615, 0.0621, 0.0538], + [-0.0946, -0.0862, -0.1040, ..., 0.0516, 0.0301, 0.0220], + ..., + [-0.1201, -0.1249, -0.0734, ..., 0.0155, 0.0755, 0.0550], + [-0.0152, -0.0134, -0.0403, ..., -0.1562, -0.1744, -0.1893], + [ 0.1044, 0.1070, 0.0952, ..., -0.0296, -0.0204, 0.0076]], + device='cuda:0'), grad: tensor([[ 3.6269e-05, 1.0885e-05, 5.3830e-06, ..., 2.0787e-05, + 2.2650e-05, 2.5585e-05], + [ 7.8753e-06, 2.4289e-06, 1.2554e-06, ..., 4.4480e-06, + 4.8541e-06, 5.4799e-06], + [ 9.0837e-05, 2.9549e-05, 1.5378e-05, ..., 5.3287e-05, + 5.8591e-05, 6.5625e-05], + ..., + [-1.1673e-03, -3.6597e-04, -1.8895e-04, ..., -6.7091e-04, + -7.3385e-04, -8.2588e-04], + [ 8.7166e-04, 2.7275e-04, 1.4091e-04, ..., 5.0020e-04, + 5.4693e-04, 6.1560e-04], + [ 4.8757e-05, 1.5154e-05, 7.8380e-06, ..., 2.7969e-05, + 3.0503e-05, 3.4362e-05]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0378, 0.0497, -0.0652, -0.0239, 0.0594, -0.0413, 0.0638], + device='cuda:0'), grad: tensor([ 1.0443e-04, 2.2441e-05, 2.6131e-04, 3.2115e-04, -3.3417e-03, + 2.4929e-03, 1.3912e-04], device='cuda:0') +306 +0.007564496387029534 +changing lr +epoch 23, time 429.30, cls_loss 0.0200 cls_loss_mapping 0.1184 cls_loss_causal 0.7899 re_mapping 0.0647 re_causal 0.0658 /// teacc 85.58 lr 0.00736934 +Epoch 25, weight, value: tensor([[ 0.2588, 0.2577, 0.2946, ..., -0.0171, -0.0030, -0.0160], + [-0.0635, -0.0567, -0.0819, ..., 0.0609, 0.0619, 0.0534], + [-0.0968, -0.0871, -0.1049, ..., 0.0503, 0.0291, 0.0212], + ..., + [-0.1225, -0.1247, -0.0735, ..., 0.0152, 0.0745, 0.0544], + [-0.0159, -0.0147, -0.0414, ..., -0.1544, -0.1725, -0.1874], + [ 0.1051, 0.1059, 0.0938, ..., -0.0290, -0.0201, 0.0078]], + device='cuda:0'), grad: tensor([[-7.3481e-04, -4.4537e-04, -4.3368e-04, ..., -7.9334e-05, + -8.5056e-05, -9.8109e-05], + [ 3.8147e-05, 1.6361e-05, 1.4454e-05, ..., 9.4771e-06, + 1.0312e-05, 1.1541e-05], + [ 2.7347e-04, 1.6248e-04, 1.5748e-04, ..., 3.1322e-05, + 3.3647e-05, 3.8832e-05], + ..., + [ 1.1367e-04, 6.1572e-05, 5.8204e-05, ..., 1.6898e-05, + 1.8463e-05, 2.0921e-05], + [ 1.5259e-04, 1.1063e-04, 1.1152e-04, ..., 3.8184e-06, + 4.9807e-06, 5.8860e-06], + [ 6.7055e-05, 4.6581e-05, 4.6998e-05, ..., 3.6433e-06, + 2.3656e-06, 3.5577e-06]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0299, 0.0463, -0.0686, -0.0247, 0.0534, -0.0402, 0.0682], + device='cuda:0'), grad: tensor([-9.9564e-04, 7.6473e-05, 3.8290e-04, 1.4734e-04, 1.8203e-04, + 1.3793e-04, 6.9320e-05], device='cuda:0') +306 +0.007369343312364995 +changing lr +epoch 24, time 426.07, cls_loss 0.0251 cls_loss_mapping 0.1051 cls_loss_causal 0.7912 re_mapping 0.0639 re_causal 0.0653 /// teacc 86.06 lr 0.00716942 +Epoch 26, weight, value: tensor([[ 0.2537, 0.2548, 0.2915, ..., -0.0177, -0.0040, -0.0170], + [-0.0597, -0.0541, -0.0789, ..., 0.0608, 0.0620, 0.0536], + [-0.0957, -0.0868, -0.1045, ..., 0.0499, 0.0289, 0.0212], + ..., + [-0.1195, -0.1230, -0.0726, ..., 0.0159, 0.0747, 0.0549], + [-0.0175, -0.0150, -0.0414, ..., -0.1536, -0.1717, -0.1865], + [ 0.1010, 0.1035, 0.0916, ..., -0.0302, -0.0214, 0.0061]], + device='cuda:0'), grad: tensor([[ 1.2159e-04, 5.0753e-05, 4.7743e-05, ..., 3.1084e-05, + 3.1054e-05, 3.8862e-05], + [ 2.5213e-05, 1.1273e-05, 1.0602e-05, ..., 5.8673e-06, + 5.9754e-06, 7.4059e-06], + [ 1.0826e-05, 5.6848e-06, 5.4725e-06, ..., 2.8498e-06, + 3.1032e-06, 3.4459e-06], + ..., + [-3.8218e-04, -1.7190e-04, -1.6272e-04, ..., -9.3102e-05, + -9.3400e-05, -1.1605e-04], + [ 5.5507e-06, 4.1053e-06, 4.2394e-06, ..., 1.8217e-06, + 1.5423e-06, 1.8962e-06], + [ 3.5197e-05, 1.7121e-05, 1.6272e-05, ..., 7.8604e-06, + 7.9349e-06, 9.8199e-06]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0365, 0.0497, -0.0662, -0.0146, 0.0565, -0.0447, 0.0602], + device='cuda:0'), grad: tensor([ 2.1684e-04, 4.3213e-05, 1.8209e-05, 3.1137e-04, -6.5041e-04, + 4.7274e-06, 5.6416e-05], device='cuda:0') +306 +0.0071694186955877925 +changing lr +epoch 25, time 425.98, cls_loss 0.0207 cls_loss_mapping 0.1023 cls_loss_causal 0.7798 re_mapping 0.0635 re_causal 0.0652 /// teacc 85.10 lr 0.00696513 +Epoch 27, weight, value: tensor([[ 0.2559, 0.2547, 0.2909, ..., -0.0169, -0.0032, -0.0158], + [-0.0587, -0.0531, -0.0774, ..., 0.0610, 0.0621, 0.0538], + [-0.0948, -0.0862, -0.1037, ..., 0.0492, 0.0285, 0.0209], + ..., + [-0.1254, -0.1251, -0.0754, ..., 0.0145, 0.0725, 0.0526], + [-0.0178, -0.0154, -0.0415, ..., -0.1522, -0.1701, -0.1848], + [ 0.1037, 0.1041, 0.0924, ..., -0.0293, -0.0205, 0.0069]], + device='cuda:0'), grad: tensor([[ 0.0110, 0.0050, 0.0045, ..., 0.0032, 0.0036, 0.0042], + [ 0.0048, 0.0021, 0.0020, ..., 0.0013, 0.0015, 0.0018], + [-0.0494, -0.0227, -0.0208, ..., -0.0142, -0.0161, -0.0184], + ..., + [-0.0014, -0.0005, -0.0005, ..., -0.0004, -0.0004, -0.0005], + [ 0.0221, 0.0103, 0.0094, ..., 0.0062, 0.0070, 0.0081], + [ 0.0116, 0.0054, 0.0049, ..., 0.0033, 0.0037, 0.0043]], + device='cuda:0') +Epoch 27, bias, value: tensor([-0.0291, 0.0499, -0.0656, -0.0168, 0.0440, -0.0445, 0.0666], + device='cuda:0'), grad: tensor([ 0.0222, 0.0096, -0.0975, 0.0028, -0.0031, 0.0432, 0.0229], + device='cuda:0') +306 +0.0069651251582696205 +changing lr +epoch 26, time 430.16, cls_loss 0.0187 cls_loss_mapping 0.0969 cls_loss_causal 0.7636 re_mapping 0.0622 re_causal 0.0637 /// teacc 81.25 lr 0.00675687 +Epoch 28, weight, value: tensor([[ 0.2514, 0.2523, 0.2883, ..., -0.0173, -0.0038, -0.0165], + [-0.0548, -0.0502, -0.0744, ..., 0.0596, 0.0609, 0.0527], + [-0.0947, -0.0856, -0.1030, ..., 0.0477, 0.0272, 0.0195], + ..., + [-0.1235, -0.1248, -0.0757, ..., 0.0164, 0.0739, 0.0545], + [-0.0161, -0.0152, -0.0411, ..., -0.1507, -0.1689, -0.1832], + [ 0.1032, 0.1029, 0.0916, ..., -0.0280, -0.0190, 0.0081]], + device='cuda:0'), grad: tensor([[-6.0415e-04, -3.0947e-04, -3.3808e-04, ..., -1.3041e-04, + -1.2791e-04, -1.2898e-04], + [ 4.1664e-05, 2.1353e-05, 2.3395e-05, ..., 9.0152e-06, + 8.7842e-06, 8.8438e-06], + [ 1.7655e-04, 9.0182e-05, 9.8526e-05, ..., 3.8236e-05, + 3.7462e-05, 3.7819e-05], + ..., + [ 1.4520e-04, 7.4148e-05, 8.1003e-05, ..., 3.1382e-05, + 3.0756e-05, 3.1054e-05], + [ 1.5426e-04, 7.8619e-05, 8.5890e-05, ..., 3.3408e-05, + 3.2693e-05, 3.3081e-05], + [ 3.6776e-05, 1.9848e-05, 2.1681e-05, ..., 7.6070e-06, + 7.5847e-06, 7.5288e-06]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0358, 0.0521, -0.0681, -0.0243, 0.0515, -0.0394, 0.0684], + device='cuda:0'), grad: tensor([-9.4175e-04, 6.4969e-05, 2.7657e-04, 7.8797e-05, 2.2757e-04, + 2.4247e-04, 5.2065e-05], device='cuda:0') +306 +0.006756874120406716 +changing lr +epoch 27, time 426.89, cls_loss 0.0093 cls_loss_mapping 0.0824 cls_loss_causal 0.7365 re_mapping 0.0618 re_causal 0.0635 /// teacc 83.17 lr 0.00654508 +Epoch 29, weight, value: tensor([[ 0.2525, 0.2524, 0.2881, ..., -0.0166, -0.0033, -0.0158], + [-0.0550, -0.0497, -0.0737, ..., 0.0594, 0.0608, 0.0527], + [-0.0931, -0.0852, -0.1024, ..., 0.0472, 0.0270, 0.0195], + ..., + [-0.1213, -0.1236, -0.0752, ..., 0.0161, 0.0732, 0.0540], + [-0.0171, -0.0156, -0.0414, ..., -0.1496, -0.1678, -0.1820], + [ 0.0988, 0.1004, 0.0894, ..., -0.0284, -0.0196, 0.0071]], + device='cuda:0'), grad: tensor([[ 8.8453e-04, 5.9783e-05, 7.9036e-05, ..., 3.7479e-04, + 4.0603e-04, 4.2248e-04], + [ 8.6021e-04, 1.4782e-04, 1.5628e-04, ..., 3.4332e-04, + 3.6836e-04, 3.8695e-04], + [-3.3932e-03, -5.7173e-04, -6.1417e-04, ..., -1.2398e-03, + -1.3380e-03, -1.4076e-03], + ..., + [ 7.9203e-04, 1.7011e-04, 1.7536e-04, ..., 2.8706e-04, + 3.0875e-04, 3.2568e-04], + [ 9.2268e-04, 1.6189e-04, 1.6904e-04, ..., 3.9434e-04, + 4.2653e-04, 4.4417e-04], + [ 6.0749e-04, 1.1957e-04, 1.2189e-04, ..., 2.7061e-04, + 2.9230e-04, 3.0398e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0325, 0.0499, -0.0641, -0.0240, 0.0539, -0.0409, 0.0621], + device='cuda:0'), grad: tensor([ 0.0026, 0.0023, -0.0089, -0.0022, 0.0020, 0.0025, 0.0017], + device='cuda:0') +306 +0.00654508497187474 +changing lr +epoch 28, time 428.99, cls_loss 0.0094 cls_loss_mapping 0.0876 cls_loss_causal 0.7353 re_mapping 0.0604 re_causal 0.0621 /// teacc 80.77 lr 0.00633018 +Epoch 30, weight, value: tensor([[ 0.2499, 0.2511, 0.2864, ..., -0.0170, -0.0040, -0.0164], + [-0.0549, -0.0491, -0.0729, ..., 0.0583, 0.0598, 0.0516], + [-0.0926, -0.0852, -0.1023, ..., 0.0469, 0.0268, 0.0194], + ..., + [-0.1186, -0.1221, -0.0742, ..., 0.0167, 0.0734, 0.0544], + [-0.0186, -0.0162, -0.0417, ..., -0.1486, -0.1666, -0.1807], + [ 0.0981, 0.0995, 0.0886, ..., -0.0280, -0.0192, 0.0073]], + device='cuda:0'), grad: tensor([[-1.0252e-04, -6.0618e-05, -6.3598e-05, ..., -1.8939e-05, + -2.0102e-05, -2.1175e-05], + [ 1.9372e-07, 9.1270e-07, 1.0356e-06, ..., -1.5274e-07, + -1.6391e-07, -1.3039e-07], + [ 8.8155e-05, 4.9591e-05, 5.1677e-05, ..., 1.5959e-05, + 1.6943e-05, 1.8016e-05], + ..., + [ 2.2665e-05, 5.7630e-06, 4.2319e-06, ..., 5.5581e-06, + 6.2585e-06, 7.0035e-06], + [-1.6429e-06, 1.9372e-07, 3.2037e-07, ..., 1.7881e-07, + 2.0862e-07, -2.9802e-08], + [-8.9109e-06, 3.0212e-06, 5.1446e-06, ..., -2.9318e-06, + -3.5129e-06, -4.0941e-06]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0365, 0.0475, -0.0623, -0.0214, 0.0576, -0.0432, 0.0627], + device='cuda:0'), grad: tensor([-1.2743e-04, -6.9290e-07, 1.1796e-04, 3.1106e-06, 5.4836e-05, + -1.0312e-05, -3.7611e-05], device='cuda:0') +306 +0.006330184227833378 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 434.47, cls_loss 0.0094 cls_loss_mapping 0.0785 cls_loss_causal 0.7117 re_mapping 0.0606 re_causal 0.0626 /// teacc 87.98 lr 0.00611260 +Epoch 31, weight, value: tensor([[ 0.2503, 0.2510, 0.2860, ..., -0.0165, -0.0035, -0.0157], + [-0.0539, -0.0484, -0.0720, ..., 0.0578, 0.0594, 0.0514], + [-0.0942, -0.0859, -0.1026, ..., 0.0460, 0.0261, 0.0188], + ..., + [-0.1188, -0.1221, -0.0747, ..., 0.0166, 0.0727, 0.0539], + [-0.0162, -0.0152, -0.0408, ..., -0.1465, -0.1646, -0.1787], + [ 0.0972, 0.0984, 0.0877, ..., -0.0281, -0.0194, 0.0069]], + device='cuda:0'), grad: tensor([[-4.4346e-04, -2.6488e-04, -2.6155e-04, ..., -9.3997e-05, + -1.1146e-04, -1.1992e-04], + [-1.3721e-04, -4.3064e-05, -4.4137e-05, ..., -5.4270e-05, + -5.3287e-05, -5.9724e-05], + [ 3.0375e-04, 1.5330e-04, 1.5008e-04, ..., 8.3685e-05, + 9.2566e-05, 1.0157e-04], + ..., + [ 1.5867e-04, 9.5129e-05, 9.6679e-05, ..., 3.1501e-05, + 3.5971e-05, 3.8385e-05], + [ 3.0577e-05, 1.7941e-05, 1.7643e-05, ..., 7.5772e-06, + 8.9332e-06, 9.7007e-06], + [ 5.4836e-05, 2.5928e-05, 2.5794e-05, ..., 1.6108e-05, + 1.7121e-05, 1.8835e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0348, 0.0479, -0.0655, -0.0245, 0.0561, -0.0370, 0.0620], + device='cuda:0'), grad: tensor([-5.3549e-04, -4.0984e-04, 5.2595e-04, 6.1989e-05, 2.0742e-04, + 4.1455e-05, 1.0896e-04], device='cuda:0') +306 +0.006112604669781575 +changing lr +epoch 30, time 430.27, cls_loss 0.0094 cls_loss_mapping 0.0758 cls_loss_causal 0.7057 re_mapping 0.0586 re_causal 0.0606 /// teacc 87.98 lr 0.00589278 +Epoch 32, weight, value: tensor([[ 0.2509, 0.2513, 0.2860, ..., -0.0162, -0.0034, -0.0155], + [-0.0541, -0.0483, -0.0717, ..., 0.0573, 0.0590, 0.0510], + [-0.0944, -0.0854, -0.1020, ..., 0.0452, 0.0255, 0.0182], + ..., + [-0.1200, -0.1226, -0.0757, ..., 0.0161, 0.0717, 0.0530], + [-0.0130, -0.0142, -0.0394, ..., -0.1452, -0.1632, -0.1771], + [ 0.0928, 0.0963, 0.0856, ..., -0.0284, -0.0198, 0.0063]], + device='cuda:0'), grad: tensor([[ 1.1826e-03, 5.4789e-04, 4.1938e-04, ..., 3.5620e-04, + 4.4441e-04, 4.4823e-04], + [-3.4008e-03, -1.2341e-03, -9.7752e-04, ..., -9.1982e-04, + -1.2970e-03, -1.2903e-03], + [ 1.4651e-04, 4.3333e-05, 2.9847e-05, ..., 3.4630e-05, + 6.0350e-05, 5.6565e-05], + ..., + [ 8.8120e-04, 2.8610e-04, 2.3878e-04, ..., 1.9145e-04, + 3.2616e-04, 3.1114e-04], + [ 3.8791e-04, 9.6798e-05, 7.8917e-05, ..., 1.2290e-04, + 1.5807e-04, 1.6606e-04], + [ 5.0402e-04, 1.7214e-04, 1.4174e-04, ..., 1.1754e-04, + 1.8585e-04, 1.8072e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0336, 0.0461, -0.0671, -0.0183, 0.0533, -0.0309, 0.0549], + device='cuda:0'), grad: tensor([ 0.0021, -0.0077, 0.0004, 0.0007, 0.0023, 0.0010, 0.0012], + device='cuda:0') +306 +0.005892784473993186 +changing lr +---------------------saving model at epoch 31---------------------------------------------------- +epoch 31, time 433.10, cls_loss 0.0076 cls_loss_mapping 0.0626 cls_loss_causal 0.6597 re_mapping 0.0570 re_causal 0.0588 /// teacc 88.46 lr 0.00567117 +Epoch 33, weight, value: tensor([[ 0.2517, 0.2513, 0.2859, ..., -0.0157, -0.0029, -0.0148], + [-0.0529, -0.0478, -0.0712, ..., 0.0578, 0.0593, 0.0514], + [-0.0946, -0.0857, -0.1020, ..., 0.0441, 0.0248, 0.0176], + ..., + [-0.1194, -0.1216, -0.0752, ..., 0.0161, 0.0713, 0.0527], + [-0.0172, -0.0160, -0.0411, ..., -0.1449, -0.1629, -0.1767], + [ 0.0949, 0.0966, 0.0861, ..., -0.0278, -0.0193, 0.0065]], + device='cuda:0'), grad: tensor([[-1.3514e-03, -7.8011e-04, -7.8249e-04, ..., -2.5344e-04, + -2.6321e-04, -3.0375e-04], + [ 2.7561e-04, 1.6153e-04, 1.6010e-04, ..., 5.0634e-05, + 5.1320e-05, 6.0827e-05], + [ 2.5606e-04, 1.3661e-04, 1.3423e-04, ..., 5.1290e-05, + 5.2512e-05, 6.0558e-05], + ..., + [ 3.1090e-04, 1.7726e-04, 1.7977e-04, ..., 5.7518e-05, + 6.1095e-05, 6.9439e-05], + [ 5.1171e-05, 4.8816e-05, 5.0664e-05, ..., 1.1854e-05, + 1.3553e-05, 1.4424e-05], + [ 2.8563e-04, 1.6415e-04, 1.6391e-04, ..., 5.2303e-05, + 5.2512e-05, 6.1929e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0312, 0.0491, -0.0674, -0.0199, 0.0518, -0.0381, 0.0600], + device='cuda:0'), grad: tensor([-1.9341e-03, 3.7980e-04, 4.0054e-04, 2.6917e-04, 4.5681e-04, + 2.0683e-05, 4.0674e-04], device='cuda:0') +306 +0.00567116632908828 +changing lr +epoch 32, time 426.99, cls_loss 0.0106 cls_loss_mapping 0.0678 cls_loss_causal 0.6951 re_mapping 0.0560 re_causal 0.0581 /// teacc 82.69 lr 0.00544820 +Epoch 34, weight, value: tensor([[ 0.2521, 0.2514, 0.2857, ..., -0.0157, -0.0028, -0.0148], + [-0.0508, -0.0470, -0.0703, ..., 0.0579, 0.0594, 0.0517], + [-0.0956, -0.0860, -0.1020, ..., 0.0434, 0.0242, 0.0169], + ..., + [-0.1187, -0.1210, -0.0750, ..., 0.0163, 0.0710, 0.0527], + [-0.0184, -0.0165, -0.0415, ..., -0.1441, -0.1620, -0.1757], + [ 0.0944, 0.0958, 0.0854, ..., -0.0276, -0.0192, 0.0065]], + device='cuda:0'), grad: tensor([[ 5.7459e-04, 2.7680e-04, 2.7418e-04, ..., 5.6207e-05, + 6.7353e-05, 1.0335e-04], + [ 1.3304e-04, 6.0856e-05, 5.9932e-05, ..., 1.6704e-05, + 1.9088e-05, 2.7254e-05], + [-1.2275e-06, 2.3041e-06, 2.5667e-06, ..., -1.2163e-06, + -2.1253e-06, -1.9837e-06], + ..., + [ 2.1264e-05, 1.0088e-05, 1.0043e-05, ..., 2.2147e-06, + 2.6375e-06, 3.8743e-06], + [ 1.0014e-04, 4.5031e-05, 4.4435e-05, ..., 1.4551e-05, + 1.6063e-05, 2.1860e-05], + [-8.6927e-04, -4.2367e-04, -4.2057e-04, ..., -8.1062e-05, + -9.7871e-05, -1.5199e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0304, 0.0528, -0.0694, -0.0211, 0.0513, -0.0399, 0.0607], + device='cuda:0'), grad: tensor([ 1.0033e-03, 2.4414e-04, -6.8992e-06, 3.7402e-05, 3.8028e-05, + 1.8930e-04, -1.5049e-03], device='cuda:0') +306 +0.00544819654451717 +changing lr +epoch 33, time 427.05, cls_loss 0.0080 cls_loss_mapping 0.0678 cls_loss_causal 0.7349 re_mapping 0.0560 re_causal 0.0582 /// teacc 86.54 lr 0.00522432 +Epoch 35, weight, value: tensor([[ 0.2496, 0.2498, 0.2841, ..., -0.0159, -0.0032, -0.0152], + [-0.0509, -0.0468, -0.0699, ..., 0.0572, 0.0588, 0.0512], + [-0.0940, -0.0851, -0.1011, ..., 0.0432, 0.0242, 0.0171], + ..., + [-0.1182, -0.1201, -0.0745, ..., 0.0161, 0.0704, 0.0523], + [-0.0183, -0.0166, -0.0416, ..., -0.1430, -0.1609, -0.1745], + [ 0.0933, 0.0948, 0.0846, ..., -0.0276, -0.0192, 0.0062]], + device='cuda:0'), grad: tensor([[-2.4395e-03, -1.4315e-03, -1.4219e-03, ..., -6.6614e-04, + -6.5660e-04, -7.2098e-04], + [ 4.0746e-04, 1.6570e-04, 1.5414e-04, ..., 8.9586e-05, + 9.1910e-05, 1.0550e-04], + [ 4.8685e-04, 2.4652e-04, 2.3901e-04, ..., 1.1039e-04, + 1.1253e-04, 1.2529e-04], + ..., + [ 8.7929e-04, 5.0354e-04, 4.9829e-04, ..., 2.5129e-04, + 2.4438e-04, 2.7108e-04], + [-3.9077e-04, 3.8408e-06, 4.0323e-05, ..., -4.2140e-05, + -4.7147e-05, -6.9320e-05], + [ 8.9645e-04, 4.3797e-04, 4.2105e-04, ..., 2.1875e-04, + 2.1732e-04, 2.4557e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0333, 0.0510, -0.0670, -0.0178, 0.0498, -0.0383, 0.0596], + device='cuda:0'), grad: tensor([-0.0039, 0.0008, 0.0008, 0.0003, 0.0015, -0.0012, 0.0016], + device='cuda:0') +306 +0.005224324151752577 +changing lr +epoch 34, time 428.33, cls_loss 0.0042 cls_loss_mapping 0.0575 cls_loss_causal 0.6963 re_mapping 0.0553 re_causal 0.0577 /// teacc 85.58 lr 0.00500000 +Epoch 36, weight, value: tensor([[ 0.2492, 0.2493, 0.2833, ..., -0.0158, -0.0032, -0.0151], + [-0.0521, -0.0472, -0.0701, ..., 0.0566, 0.0582, 0.0505], + [-0.0929, -0.0847, -0.1006, ..., 0.0430, 0.0242, 0.0171], + ..., + [-0.1164, -0.1190, -0.0738, ..., 0.0165, 0.0704, 0.0524], + [-0.0172, -0.0164, -0.0414, ..., -0.1417, -0.1595, -0.1728], + [ 0.0920, 0.0938, 0.0838, ..., -0.0276, -0.0193, 0.0059]], + device='cuda:0'), grad: tensor([[-2.1362e-04, -1.1706e-04, -1.1629e-04, ..., -5.4300e-05, + -5.6475e-05, -5.9038e-05], + [ 2.7761e-05, 1.4283e-05, 1.4096e-05, ..., 6.8471e-06, + 7.1414e-06, 7.5735e-06], + [ 1.2493e-04, 6.5327e-05, 6.4611e-05, ..., 3.1233e-05, + 3.2574e-05, 3.4422e-05], + ..., + [ 1.7971e-05, 9.3132e-06, 9.1493e-06, ..., 3.6974e-06, + 3.9116e-06, 4.1649e-06], + [ 4.7088e-05, 2.2233e-05, 2.1875e-05, ..., 1.0654e-05, + 1.1109e-05, 1.2226e-05], + [-1.9550e-05, -2.3656e-06, -1.5628e-06, ..., -1.5832e-06, + -1.8850e-06, -3.2447e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0330, 0.0474, -0.0645, -0.0211, 0.0514, -0.0336, 0.0575], + device='cuda:0'), grad: tensor([-3.4666e-04, 4.8816e-05, 2.1541e-04, 2.4825e-05, 3.2306e-05, + 9.2864e-05, -6.8188e-05], device='cuda:0') +306 +0.005000000000000003 +changing lr +epoch 35, time 428.68, cls_loss 0.0057 cls_loss_mapping 0.0564 cls_loss_causal 0.6790 re_mapping 0.0534 re_causal 0.0557 /// teacc 86.54 lr 0.00477568 +Epoch 37, weight, value: tensor([[ 0.2490, 0.2492, 0.2831, ..., -0.0157, -0.0032, -0.0150], + [-0.0533, -0.0474, -0.0702, ..., 0.0559, 0.0575, 0.0498], + [-0.0930, -0.0845, -0.1003, ..., 0.0425, 0.0239, 0.0169], + ..., + [-0.1153, -0.1182, -0.0733, ..., 0.0167, 0.0703, 0.0524], + [-0.0177, -0.0169, -0.0417, ..., -0.1408, -0.1588, -0.1719], + [ 0.0928, 0.0935, 0.0835, ..., -0.0270, -0.0188, 0.0063]], + device='cuda:0'), grad: tensor([[-1.9026e-04, -1.2153e-04, -1.2040e-04, ..., -2.5526e-05, + -2.8968e-05, -3.4362e-05], + [ 1.2696e-04, 4.9949e-05, 4.9144e-05, ..., 3.4660e-05, + 4.0740e-05, 4.1753e-05], + [ 8.4221e-05, 4.2409e-05, 4.1664e-05, ..., 1.6063e-05, + 1.8761e-05, 2.1011e-05], + ..., + [-4.0221e-04, -1.1349e-04, -1.1182e-04, ..., -1.4126e-04, + -1.6749e-04, -1.6463e-04], + [ 1.6600e-05, 1.8686e-05, 2.0117e-05, ..., 3.7812e-07, + 2.9206e-06, 2.1532e-06], + [ 1.7178e-04, 6.0230e-05, 5.8681e-05, ..., 5.3525e-05, + 6.0827e-05, 6.0886e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0331, 0.0438, -0.0652, -0.0216, 0.0517, -0.0324, 0.0609], + device='cuda:0'), grad: tensor([-2.2066e-04, 2.5630e-04, 1.3638e-04, 4.3583e-04, -9.7466e-04, + -8.4341e-06, 3.7503e-04], device='cuda:0') +306 +0.004775675848247429 +changing lr +epoch 36, time 426.09, cls_loss 0.0056 cls_loss_mapping 0.0548 cls_loss_causal 0.6659 re_mapping 0.0529 re_causal 0.0553 /// teacc 84.62 lr 0.00455180 +Epoch 38, weight, value: tensor([[ 0.2494, 0.2492, 0.2830, ..., -0.0157, -0.0032, -0.0149], + [-0.0519, -0.0468, -0.0695, ..., 0.0557, 0.0575, 0.0499], + [-0.0936, -0.0847, -0.1003, ..., 0.0419, 0.0234, 0.0164], + ..., + [-0.1138, -0.1172, -0.0726, ..., 0.0171, 0.0703, 0.0526], + [-0.0186, -0.0172, -0.0419, ..., -0.1402, -0.1581, -0.1712], + [ 0.0912, 0.0921, 0.0822, ..., -0.0268, -0.0187, 0.0063]], + device='cuda:0'), grad: tensor([[ 3.4750e-05, 1.2450e-05, 1.1384e-05, ..., 8.5309e-06, + 9.8050e-06, 1.1593e-05], + [-1.0271e-03, -4.1318e-04, -3.9172e-04, ..., -1.4448e-04, + -1.7440e-04, -2.1589e-04], + [ 5.4985e-06, 2.5854e-06, 2.5835e-06, ..., 2.0489e-08, + 1.8626e-08, 2.0862e-07], + ..., + [-4.6492e-06, -3.4589e-06, -3.4049e-06, ..., -4.2692e-06, + -5.1111e-06, -5.2936e-06], + [-1.3888e-05, -8.7917e-07, -6.4634e-07, ..., -3.5372e-06, + -2.5574e-06, -3.6340e-06], + [ 9.9373e-04, 3.9911e-04, 3.7861e-04, ..., 1.4126e-04, + 1.6975e-04, 2.0993e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0323, 0.0466, -0.0664, -0.0230, 0.0530, -0.0339, 0.0600], + device='cuda:0'), grad: tensor([ 7.5400e-05, -1.8406e-03, 8.1286e-06, 3.1173e-05, -1.1809e-05, + -4.4286e-05, 1.7853e-03], device='cuda:0') +306 +0.004551803455482836 +changing lr +epoch 37, time 425.12, cls_loss 0.0061 cls_loss_mapping 0.0590 cls_loss_causal 0.6774 re_mapping 0.0525 re_causal 0.0551 /// teacc 83.17 lr 0.00432883 +Epoch 39, weight, value: tensor([[ 0.2483, 0.2485, 0.2821, ..., -0.0158, -0.0035, -0.0151], + [-0.0507, -0.0462, -0.0687, ..., 0.0556, 0.0575, 0.0499], + [-0.0933, -0.0845, -0.1000, ..., 0.0415, 0.0232, 0.0162], + ..., + [-0.1130, -0.1164, -0.0722, ..., 0.0172, 0.0700, 0.0525], + [-0.0192, -0.0175, -0.0421, ..., -0.1396, -0.1574, -0.1704], + [ 0.0900, 0.0911, 0.0812, ..., -0.0268, -0.0187, 0.0061]], + device='cuda:0'), grad: tensor([[-1.6487e-04, -1.0222e-04, -1.0294e-04, ..., -1.8775e-05, + -1.5110e-05, -1.9088e-05], + [ 2.0981e-05, 3.2540e-06, 2.7195e-06, ..., 5.4128e-06, + 6.1393e-06, 7.4096e-06], + [ 1.8418e-04, 1.0049e-04, 1.0014e-04, ..., 2.5466e-05, + 2.3037e-05, 2.8580e-05], + ..., + [ 4.7654e-05, 9.2238e-06, 7.8902e-06, ..., 1.1139e-05, + 1.2316e-05, 1.5184e-05], + [-1.4281e-04, -1.4052e-05, -9.4771e-06, ..., -3.4928e-05, + -3.9577e-05, -4.9472e-05], + [ 8.9034e-06, -2.5947e-06, -2.8443e-06, ..., 2.9244e-07, + 4.1351e-07, 1.5832e-06]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0338, 0.0484, -0.0660, -0.0216, 0.0525, -0.0344, 0.0589], + device='cuda:0'), grad: tensor([-1.9264e-04, 6.2644e-05, 2.6774e-04, 1.4007e-04, 1.3340e-04, + -4.5323e-04, 4.2528e-05], device='cuda:0') +306 +0.004328833670911726 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 430.89, cls_loss 0.0048 cls_loss_mapping 0.0498 cls_loss_causal 0.6565 re_mapping 0.0502 re_causal 0.0527 /// teacc 90.38 lr 0.00410722 +Epoch 40, weight, value: tensor([[ 0.2478, 0.2481, 0.2815, ..., -0.0158, -0.0035, -0.0151], + [-0.0505, -0.0459, -0.0683, ..., 0.0551, 0.0570, 0.0494], + [-0.0923, -0.0841, -0.0996, ..., 0.0414, 0.0233, 0.0164], + ..., + [-0.1132, -0.1161, -0.0721, ..., 0.0171, 0.0696, 0.0521], + [-0.0200, -0.0178, -0.0422, ..., -0.1390, -0.1568, -0.1697], + [ 0.0894, 0.0902, 0.0805, ..., -0.0267, -0.0188, 0.0059]], + device='cuda:0'), grad: tensor([[-3.8218e-04, -2.5368e-04, -2.5129e-04, ..., -3.4750e-05, + -4.4465e-05, -5.1945e-05], + [ 9.3281e-05, 5.9783e-05, 5.9187e-05, ..., 1.0341e-05, + 1.3016e-05, 1.4789e-05], + [ 1.3657e-05, 1.0043e-05, 9.9838e-06, ..., 1.3318e-06, + 1.8254e-06, 1.9278e-06], + ..., + [ 5.6207e-05, 3.2425e-05, 3.2097e-05, ..., 8.9929e-06, + 1.1019e-05, 1.1913e-05], + [ 4.3213e-06, 8.9258e-06, 9.3654e-06, ..., -5.5321e-07, + 8.8103e-07, 4.0978e-07], + [ 2.1148e-04, 1.3483e-04, 1.3328e-04, ..., 2.1800e-05, + 2.6941e-05, 3.1203e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0341, 0.0476, -0.0636, -0.0197, 0.0505, -0.0361, 0.0592], + device='cuda:0'), grad: tensor([-3.9268e-04, 1.0526e-04, 1.1511e-05, -2.8744e-05, 7.9453e-05, + -1.2942e-05, 2.3806e-04], device='cuda:0') +306 +0.0041072155260068206 +changing lr +epoch 39, time 431.70, cls_loss 0.0051 cls_loss_mapping 0.0535 cls_loss_causal 0.7021 re_mapping 0.0495 re_causal 0.0520 /// teacc 86.06 lr 0.00388740 +Epoch 41, weight, value: tensor([[ 0.2513, 0.2500, 0.2832, ..., -0.0149, -0.0026, -0.0140], + [-0.0511, -0.0459, -0.0682, ..., 0.0546, 0.0565, 0.0489], + [-0.0920, -0.0841, -0.0995, ..., 0.0412, 0.0232, 0.0164], + ..., + [-0.1122, -0.1155, -0.0718, ..., 0.0172, 0.0693, 0.0519], + [-0.0206, -0.0180, -0.0424, ..., -0.1384, -0.1561, -0.1690], + [ 0.0860, 0.0879, 0.0783, ..., -0.0271, -0.0193, 0.0051]], + device='cuda:0'), grad: tensor([[-9.6977e-05, -7.2122e-05, -7.0632e-05, ..., -4.6343e-06, + -6.8955e-06, -6.5528e-06], + [ 7.1883e-05, 1.8924e-05, 1.8641e-05, ..., 2.4423e-05, + 2.8446e-05, 3.0786e-05], + [-1.3530e-04, -5.0068e-06, -4.9546e-06, ..., -5.1588e-05, + -5.1260e-05, -5.7846e-05], + ..., + [ 1.7941e-04, 4.3571e-05, 4.2886e-05, ..., 5.6475e-05, + 6.2644e-05, 6.8307e-05], + [-4.5598e-05, 6.7055e-08, 9.5367e-07, ..., -1.6466e-05, + -2.1920e-05, -2.2978e-05], + [-3.1441e-05, 5.8487e-07, -5.1036e-07, ..., -2.8342e-05, + -3.4481e-05, -3.7223e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0283, 0.0453, -0.0618, -0.0206, 0.0508, -0.0370, 0.0554], + device='cuda:0'), grad: tensor([-7.4744e-05, 1.7619e-04, -4.0269e-04, 1.4830e-04, 4.4298e-04, + -2.0444e-04, -8.5533e-05], device='cuda:0') +306 +0.0038873953302184317 +changing lr +epoch 40, time 429.47, cls_loss 0.0052 cls_loss_mapping 0.0468 cls_loss_causal 0.6742 re_mapping 0.0483 re_causal 0.0508 /// teacc 87.02 lr 0.00366982 +Epoch 42, weight, value: tensor([[ 0.2490, 0.2490, 0.2821, ..., -0.0153, -0.0030, -0.0145], + [-0.0508, -0.0456, -0.0678, ..., 0.0542, 0.0562, 0.0486], + [-0.0907, -0.0838, -0.0991, ..., 0.0412, 0.0233, 0.0167], + ..., + [-0.1121, -0.1149, -0.0715, ..., 0.0172, 0.0689, 0.0516], + [-0.0202, -0.0181, -0.0423, ..., -0.1376, -0.1552, -0.1681], + [ 0.0861, 0.0873, 0.0777, ..., -0.0269, -0.0191, 0.0052]], + device='cuda:0'), grad: tensor([[ 7.0524e-04, 2.4354e-04, 2.1636e-04, ..., 9.5367e-05, + 1.0043e-04, 1.0616e-04], + [ 8.6203e-06, -1.9744e-06, -2.6431e-06, ..., -6.4969e-06, + -6.2473e-06, -7.8455e-06], + [ 1.5199e-04, 4.8757e-05, 4.3005e-05, ..., 3.3259e-05, + 3.6716e-05, 3.8743e-05], + ..., + [ 9.5427e-05, 3.2485e-05, 2.9102e-05, ..., 1.6078e-05, + 1.7956e-05, 1.9237e-05], + [ 7.1466e-05, 2.4527e-05, 2.2009e-05, ..., 1.1332e-05, + 1.2673e-05, 1.3590e-05], + [-9.4128e-04, -3.2496e-04, -2.8896e-04, ..., -1.0651e-04, + -1.1295e-04, -1.1897e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0322, 0.0448, -0.0585, -0.0213, 0.0490, -0.0356, 0.0576], + device='cuda:0'), grad: tensor([ 1.4572e-03, 3.4839e-05, 3.1948e-04, -2.0826e-04, 1.9515e-04, + 1.4567e-04, -1.9464e-03], device='cuda:0') +306 +0.003669815772166629 +changing lr +epoch 41, time 427.88, cls_loss 0.0065 cls_loss_mapping 0.0534 cls_loss_causal 0.6753 re_mapping 0.0489 re_causal 0.0515 /// teacc 86.54 lr 0.00345492 +Epoch 43, weight, value: tensor([[ 0.2492, 0.2492, 0.2821, ..., -0.0153, -0.0032, -0.0146], + [-0.0510, -0.0455, -0.0676, ..., 0.0537, 0.0557, 0.0482], + [-0.0909, -0.0837, -0.0989, ..., 0.0408, 0.0231, 0.0165], + ..., + [-0.1110, -0.1143, -0.0712, ..., 0.0175, 0.0690, 0.0519], + [-0.0209, -0.0183, -0.0424, ..., -0.1370, -0.1546, -0.1674], + [ 0.0852, 0.0864, 0.0768, ..., -0.0269, -0.0192, 0.0050]], + device='cuda:0'), grad: tensor([[ 1.2779e-04, 4.0859e-05, 3.5703e-05, ..., 3.8385e-05, + 4.5091e-05, 5.1558e-05], + [ 7.3612e-06, 2.3488e-06, 2.0191e-06, ..., 2.3954e-06, + 2.9132e-06, 3.2187e-06], + [-6.0797e-05, -1.8999e-05, -1.8507e-05, ..., -4.4629e-06, + -9.8720e-07, -7.0706e-06], + ..., + [-1.0753e-04, -3.7402e-05, -3.1054e-05, ..., -4.6730e-05, + -5.9545e-05, -6.1870e-05], + [-4.8988e-06, 8.9221e-07, 1.3467e-06, ..., 1.0617e-07, + -4.2282e-07, -1.7136e-07], + [ 1.0051e-05, 3.7905e-06, 3.2373e-06, ..., 1.7025e-06, + 2.3823e-06, 2.7604e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0323, 0.0438, -0.0593, -0.0198, 0.0506, -0.0368, 0.0577], + device='cuda:0'), grad: tensor([ 2.9945e-04, 1.6332e-05, -1.7369e-04, 6.4611e-05, -2.1088e-04, + -1.8358e-05, 2.2486e-05], device='cuda:0') +306 +0.0034549150281252667 +changing lr +epoch 42, time 431.83, cls_loss 0.0043 cls_loss_mapping 0.0461 cls_loss_causal 0.6401 re_mapping 0.0486 re_causal 0.0514 /// teacc 84.62 lr 0.00324313 +Epoch 44, weight, value: tensor([[ 0.2495, 0.2492, 0.2820, ..., -0.0153, -0.0032, -0.0145], + [-0.0501, -0.0451, -0.0671, ..., 0.0536, 0.0556, 0.0481], + [-0.0910, -0.0837, -0.0989, ..., 0.0405, 0.0229, 0.0163], + ..., + [-0.1111, -0.1140, -0.0711, ..., 0.0175, 0.0687, 0.0517], + [-0.0213, -0.0185, -0.0426, ..., -0.1364, -0.1540, -0.1667], + [ 0.0847, 0.0858, 0.0763, ..., -0.0268, -0.0192, 0.0050]], + device='cuda:0'), grad: tensor([[-1.4976e-05, -1.8388e-05, -1.7866e-05, ..., 1.8664e-06, + 2.2594e-06, 2.4643e-06], + [ 1.0014e-04, 9.9018e-06, 6.0499e-06, ..., 3.9488e-05, + 4.5002e-05, 4.9442e-05], + [ 2.6536e-04, 2.3797e-05, 1.4089e-05, ..., 1.0502e-04, + 1.2076e-04, 1.3196e-04], + ..., + [-6.3515e-04, -4.6462e-05, -2.4155e-05, ..., -2.5392e-04, + -2.9373e-04, -3.2043e-04], + [ 5.5671e-05, 5.0776e-06, 3.2317e-06, ..., 2.2560e-05, + 2.5466e-05, 2.7850e-05], + [ 1.9002e-04, 2.3529e-05, 1.6689e-05, ..., 7.2062e-05, + 8.2552e-05, 9.0301e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0314, 0.0445, -0.0594, -0.0199, 0.0491, -0.0369, 0.0577], + device='cuda:0'), grad: tensor([ 1.7658e-05, 3.0422e-04, 8.1301e-04, 1.1760e-04, -1.9779e-03, + 1.6999e-04, 5.5742e-04], device='cuda:0') +306 +0.0032431258795932905 +changing lr +epoch 43, time 427.82, cls_loss 0.0039 cls_loss_mapping 0.0450 cls_loss_causal 0.6249 re_mapping 0.0474 re_causal 0.0502 /// teacc 86.54 lr 0.00303487 +Epoch 45, weight, value: tensor([[ 0.2480, 0.2484, 0.2811, ..., -0.0155, -0.0035, -0.0148], + [-0.0493, -0.0446, -0.0666, ..., 0.0534, 0.0555, 0.0480], + [-0.0911, -0.0836, -0.0987, ..., 0.0401, 0.0227, 0.0161], + ..., + [-0.1105, -0.1136, -0.0709, ..., 0.0176, 0.0687, 0.0517], + [-0.0208, -0.0185, -0.0425, ..., -0.1358, -0.1533, -0.1659], + [ 0.0842, 0.0853, 0.0759, ..., -0.0268, -0.0191, 0.0049]], + device='cuda:0'), grad: tensor([[ 3.8862e-04, 1.0622e-04, 1.0890e-04, ..., 1.5783e-04, + 1.7059e-04, 1.8311e-04], + [ 3.5614e-05, 1.0513e-05, 8.5682e-06, ..., 1.9521e-05, + 2.2337e-05, 2.3574e-05], + [-2.2268e-04, -3.7193e-05, -4.0859e-05, ..., -7.3195e-05, + -7.8142e-05, -9.1970e-05], + ..., + [-3.6907e-04, -1.2648e-04, -1.1647e-04, ..., -1.8442e-04, + -2.0373e-04, -2.0933e-04], + [ 4.9770e-05, 1.2524e-05, 1.2390e-05, ..., 2.1741e-05, + 2.4214e-05, 2.6256e-05], + [ 3.8534e-05, 1.1489e-05, 9.8050e-06, ..., 1.9610e-05, + 2.1860e-05, 2.2933e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0337, 0.0451, -0.0600, -0.0192, 0.0491, -0.0350, 0.0576], + device='cuda:0'), grad: tensor([ 8.6308e-04, 7.6950e-05, -5.8699e-04, 1.7869e-04, -7.2908e-04, + 1.1402e-04, 8.3148e-05], device='cuda:0') +306 +0.0030348748417303863 +changing lr +epoch 44, time 426.53, cls_loss 0.0031 cls_loss_mapping 0.0422 cls_loss_causal 0.6510 re_mapping 0.0467 re_causal 0.0495 /// teacc 86.06 lr 0.00283058 +Epoch 46, weight, value: tensor([[ 0.2471, 0.2479, 0.2805, ..., -0.0157, -0.0037, -0.0151], + [-0.0492, -0.0445, -0.0663, ..., 0.0532, 0.0554, 0.0479], + [-0.0904, -0.0834, -0.0984, ..., 0.0401, 0.0227, 0.0162], + ..., + [-0.1096, -0.1130, -0.0704, ..., 0.0179, 0.0687, 0.0519], + [-0.0199, -0.0182, -0.0422, ..., -0.1351, -0.1525, -0.1650], + [ 0.0826, 0.0844, 0.0751, ..., -0.0269, -0.0193, 0.0046]], + device='cuda:0'), grad: tensor([[ 2.2233e-05, 4.0531e-06, 3.6173e-06, ..., 7.3127e-06, + 8.8513e-06, 9.5591e-06], + [ 2.4855e-05, 4.6380e-06, 4.7497e-06, ..., 1.1235e-05, + 1.2487e-05, 1.3188e-05], + [-1.9088e-05, -3.9786e-06, -2.7493e-06, ..., 3.8594e-06, + 1.6009e-06, -4.5635e-07], + ..., + [ 7.0989e-05, 8.3148e-06, 9.3728e-06, ..., 3.5614e-05, + 3.9279e-05, 4.1932e-05], + [ 1.1116e-04, 1.2904e-05, 1.5192e-05, ..., 6.3360e-05, + 6.8307e-05, 7.0989e-05], + [-3.1441e-05, -1.5251e-05, -1.2942e-05, ..., -6.5006e-07, + -3.1181e-06, -3.5092e-06]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0350, 0.0450, -0.0580, -0.0206, 0.0497, -0.0327, 0.0553], + device='cuda:0'), grad: tensor([ 6.0409e-05, 6.4909e-05, -5.6982e-05, -5.0735e-04, 1.9813e-04, + 3.0398e-04, -6.2764e-05], device='cuda:0') +306 +0.0028305813044122124 +changing lr +epoch 45, time 428.43, cls_loss 0.0033 cls_loss_mapping 0.0372 cls_loss_causal 0.6389 re_mapping 0.0461 re_causal 0.0489 /// teacc 86.06 lr 0.00263066 +Epoch 47, weight, value: tensor([[ 0.2468, 0.2476, 0.2801, ..., -0.0156, -0.0037, -0.0151], + [-0.0488, -0.0443, -0.0660, ..., 0.0530, 0.0552, 0.0478], + [-0.0904, -0.0833, -0.0983, ..., 0.0398, 0.0225, 0.0160], + ..., + [-0.1086, -0.1124, -0.0699, ..., 0.0182, 0.0688, 0.0521], + [-0.0199, -0.0184, -0.0423, ..., -0.1346, -0.1520, -0.1645], + [ 0.0815, 0.0837, 0.0744, ..., -0.0270, -0.0195, 0.0043]], + device='cuda:0'), grad: tensor([[ 2.9951e-05, -2.6792e-05, -3.1501e-05, ..., 8.2031e-06, + 8.5086e-06, 1.1154e-05], + [-1.4961e-04, -6.3956e-05, -6.0052e-05, ..., -5.1737e-05, + -6.2943e-05, -6.9201e-05], + [-9.4604e-04, -1.0973e-04, -5.1677e-05, ..., -1.7416e-04, + -1.8346e-04, -2.0885e-04], + ..., + [ 3.8338e-04, 8.6546e-05, 6.5982e-05, ..., 9.0778e-05, + 1.0258e-04, 1.1402e-04], + [ 4.9114e-04, 6.0260e-05, 3.3230e-05, ..., 8.3089e-05, + 8.5413e-05, 9.7692e-05], + [ 1.1992e-04, 3.7611e-05, 3.2216e-05, ..., 2.8685e-05, + 3.3110e-05, 3.6567e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0350, 0.0456, -0.0582, -0.0206, 0.0506, -0.0320, 0.0536], + device='cuda:0'), grad: tensor([ 0.0002, -0.0003, -0.0027, 0.0002, 0.0010, 0.0014, 0.0003], + device='cuda:0') +306 +0.0026306566876350096 +changing lr +epoch 46, time 470.73, cls_loss 0.0037 cls_loss_mapping 0.0379 cls_loss_causal 0.6401 re_mapping 0.0456 re_causal 0.0483 /// teacc 86.06 lr 0.00243550 +Epoch 48, weight, value: tensor([[ 0.2481, 0.2483, 0.2807, ..., -0.0153, -0.0034, -0.0146], + [-0.0491, -0.0443, -0.0660, ..., 0.0528, 0.0550, 0.0475], + [-0.0905, -0.0833, -0.0982, ..., 0.0396, 0.0224, 0.0159], + ..., + [-0.1091, -0.1123, -0.0700, ..., 0.0180, 0.0684, 0.0517], + [-0.0202, -0.0186, -0.0425, ..., -0.1343, -0.1516, -0.1641], + [ 0.0814, 0.0831, 0.0739, ..., -0.0267, -0.0193, 0.0045]], + device='cuda:0'), grad: tensor([[-6.1356e-06, -1.0031e-04, -1.0329e-04, ..., -3.7737e-06, + 4.6641e-06, -4.1816e-07], + [ 1.4558e-05, 8.7991e-06, 8.2701e-06, ..., 1.1362e-07, + 1.1306e-06, 1.6019e-06], + [-2.8872e-04, -2.8953e-05, -2.7329e-05, ..., -4.5776e-05, + -5.9903e-05, -6.5863e-05], + ..., + [ 1.9741e-04, 9.6142e-05, 9.7275e-05, ..., 3.3945e-05, + 3.6329e-05, 4.4256e-05], + [ 5.7846e-05, 1.5557e-05, 1.5691e-05, ..., 1.1273e-05, + 1.2152e-05, 1.4022e-05], + [ 1.4775e-05, 4.5337e-06, 5.0813e-06, ..., 3.1181e-06, + 4.2319e-06, 4.7013e-06]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0331, 0.0446, -0.0586, -0.0211, 0.0488, -0.0323, 0.0554], + device='cuda:0'), grad: tensor([ 2.4533e-04, 1.8641e-05, -8.0538e-04, 1.8939e-05, 3.4881e-04, + 1.3614e-04, 3.7193e-05], device='cuda:0') +306 +0.0024355036129704724 +changing lr +epoch 47, time 429.21, cls_loss 0.0029 cls_loss_mapping 0.0359 cls_loss_causal 0.6213 re_mapping 0.0454 re_causal 0.0483 /// teacc 85.58 lr 0.00224552 +Epoch 49, weight, value: tensor([[ 0.2480, 0.2482, 0.2805, ..., -0.0153, -0.0034, -0.0146], + [-0.0491, -0.0442, -0.0658, ..., 0.0525, 0.0547, 0.0473], + [-0.0902, -0.0832, -0.0981, ..., 0.0395, 0.0223, 0.0158], + ..., + [-0.1089, -0.1120, -0.0699, ..., 0.0180, 0.0682, 0.0516], + [-0.0203, -0.0188, -0.0426, ..., -0.1339, -0.1512, -0.1637], + [ 0.0814, 0.0828, 0.0736, ..., -0.0265, -0.0191, 0.0047]], + device='cuda:0'), grad: tensor([[-2.2233e-04, -1.4102e-04, -1.4091e-04, ..., -3.0786e-05, + -3.1620e-05, -3.5584e-05], + [ 7.0989e-05, 1.6347e-05, 1.2673e-05, ..., 1.3053e-05, + 1.4298e-05, 1.8820e-05], + [ 9.7334e-05, 4.0352e-05, 3.7849e-05, ..., 1.6138e-05, + 1.6898e-05, 2.0996e-05], + ..., + [ 1.5485e-04, 8.2076e-05, 8.0466e-05, ..., 2.3916e-05, + 2.5123e-05, 2.9176e-05], + [-1.0288e-04, -8.0243e-06, 7.8324e-07, ..., -2.0772e-05, + -2.2560e-05, -3.1799e-05], + [-3.6329e-05, 1.5283e-06, 2.0489e-06, ..., -9.4473e-06, + -1.0073e-05, -1.0923e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0331, 0.0438, -0.0579, -0.0215, 0.0480, -0.0320, 0.0564], + device='cuda:0'), grad: tensor([-0.0002, 0.0002, 0.0002, 0.0001, 0.0002, -0.0004, -0.0001], + device='cuda:0') +306 +0.00224551509273949 +changing lr +epoch 48, time 427.84, cls_loss 0.0030 cls_loss_mapping 0.0376 cls_loss_causal 0.6181 re_mapping 0.0457 re_causal 0.0488 /// teacc 87.98 lr 0.00206107 +Epoch 50, weight, value: tensor([[ 0.2472, 0.2477, 0.2799, ..., -0.0154, -0.0036, -0.0148], + [-0.0493, -0.0442, -0.0657, ..., 0.0523, 0.0545, 0.0471], + [-0.0901, -0.0832, -0.0980, ..., 0.0394, 0.0223, 0.0158], + ..., + [-0.1086, -0.1115, -0.0695, ..., 0.0177, 0.0678, 0.0513], + [-0.0209, -0.0190, -0.0428, ..., -0.1336, -0.1509, -0.1634], + [ 0.0821, 0.0828, 0.0736, ..., -0.0262, -0.0188, 0.0050]], + device='cuda:0'), grad: tensor([[ 1.4400e-04, 4.0770e-05, 3.5912e-05, ..., 4.9591e-05, + 5.7667e-05, 5.9694e-05], + [-2.9826e-04, -7.5936e-05, -6.6042e-05, ..., -1.1307e-04, + -1.3459e-04, -1.3673e-04], + [ 1.9684e-05, 1.6894e-06, 8.9593e-07, ..., 1.1876e-05, + 1.4775e-05, 1.4365e-05], + ..., + [ 2.1636e-05, 5.9158e-06, 5.1521e-06, ..., 7.8231e-06, + 9.0152e-06, 9.3728e-06], + [ 7.5512e-06, -1.0384e-06, -7.2550e-07, ..., 5.0589e-06, + 8.4639e-06, 7.0296e-06], + [ 9.3520e-05, 2.5243e-05, 2.1860e-05, ..., 3.4422e-05, + 3.9756e-05, 4.1187e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0340, 0.0430, -0.0577, -0.0202, 0.0473, -0.0331, 0.0585], + device='cuda:0'), grad: tensor([ 3.0828e-04, -6.5708e-04, 5.1737e-05, 2.4900e-05, 4.6939e-05, + 2.0772e-05, 2.0397e-04], device='cuda:0') +306 +0.002061073738537637 +changing lr +epoch 49, time 427.43, cls_loss 0.0027 cls_loss_mapping 0.0358 cls_loss_causal 0.6541 re_mapping 0.0448 re_causal 0.0479 /// teacc 88.46 lr 0.00188255 +Epoch 51, weight, value: tensor([[ 0.2476, 0.2479, 0.2800, ..., -0.0152, -0.0035, -0.0146], + [-0.0483, -0.0439, -0.0654, ..., 0.0525, 0.0547, 0.0474], + [-0.0902, -0.0832, -0.0979, ..., 0.0392, 0.0221, 0.0157], + ..., + [-0.1085, -0.1113, -0.0694, ..., 0.0177, 0.0677, 0.0511], + [-0.0210, -0.0191, -0.0428, ..., -0.1333, -0.1506, -0.1629], + [ 0.0808, 0.0820, 0.0729, ..., -0.0264, -0.0190, 0.0046]], + device='cuda:0'), grad: tensor([[-7.3761e-06, -6.8903e-05, -6.1393e-05, ..., 3.7342e-05, + 4.6521e-05, 4.6879e-05], + [-4.7827e-04, -1.3089e-04, -1.4806e-04, ..., -1.6093e-04, + -1.8084e-04, -1.7262e-04], + [ 9.6321e-05, 4.4078e-05, 4.3571e-05, ..., 2.0713e-05, + 2.2292e-05, 2.2203e-05], + ..., + [ 1.0103e-04, 4.2081e-05, 4.1813e-05, ..., 2.4498e-05, + 2.7284e-05, 2.7090e-05], + [-7.2420e-05, -5.0105e-06, -3.0901e-06, ..., -3.3885e-05, + -3.8385e-05, -3.9697e-05], + [ 3.3355e-04, 1.0943e-04, 1.1837e-04, ..., 1.0401e-04, + 1.1396e-04, 1.0717e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0331, 0.0450, -0.0578, -0.0209, 0.0469, -0.0330, 0.0567], + device='cuda:0'), grad: tensor([ 1.2612e-04, -9.6321e-04, 1.7142e-04, 6.3956e-05, 1.8120e-04, + -2.7323e-04, 6.9284e-04], device='cuda:0') +306 +0.0018825509907063344 +changing lr +epoch 50, time 426.95, cls_loss 0.0033 cls_loss_mapping 0.0344 cls_loss_causal 0.6313 re_mapping 0.0445 re_causal 0.0478 /// teacc 86.54 lr 0.00171031 +Epoch 52, weight, value: tensor([[ 0.2476, 0.2479, 0.2798, ..., -0.0152, -0.0035, -0.0146], + [-0.0481, -0.0439, -0.0653, ..., 0.0525, 0.0548, 0.0474], + [-0.0902, -0.0832, -0.0979, ..., 0.0390, 0.0220, 0.0156], + ..., + [-0.1074, -0.1107, -0.0689, ..., 0.0179, 0.0678, 0.0512], + [-0.0213, -0.0192, -0.0429, ..., -0.1330, -0.1503, -0.1626], + [ 0.0802, 0.0816, 0.0724, ..., -0.0264, -0.0191, 0.0045]], + device='cuda:0'), grad: tensor([[ 1.1891e-04, 3.1561e-05, 2.2739e-05, ..., 3.7551e-05, + 2.8744e-05, 3.4332e-05], + [ 2.6679e-04, 6.5207e-05, 4.4525e-05, ..., 7.9095e-05, + 5.8681e-05, 7.3135e-05], + [ 7.6234e-05, 1.8701e-05, 1.0729e-05, ..., 2.2352e-05, + 1.4529e-05, 1.7613e-05], + ..., + [ 3.4004e-05, 5.2080e-06, 1.2452e-06, ..., 9.4548e-06, + 6.0275e-06, 6.8285e-06], + [-1.1311e-03, -2.8086e-04, -1.7083e-04, ..., -3.4928e-04, + -2.4164e-04, -2.8658e-04], + [ 3.1185e-04, 7.9513e-05, 4.2439e-05, ..., 1.0026e-04, + 6.4373e-05, 7.1526e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0330, 0.0449, -0.0578, -0.0217, 0.0485, -0.0332, 0.0560], + device='cuda:0'), grad: tensor([ 0.0003, 0.0008, 0.0002, 0.0009, 0.0001, -0.0033, 0.0009], + device='cuda:0') +306 +0.0017103063703014388 +changing lr +epoch 51, time 429.18, cls_loss 0.0033 cls_loss_mapping 0.0337 cls_loss_causal 0.6361 re_mapping 0.0436 re_causal 0.0467 /// teacc 83.65 lr 0.00154469 +Epoch 53, weight, value: tensor([[ 0.2471, 0.2476, 0.2795, ..., -0.0152, -0.0035, -0.0147], + [-0.0487, -0.0439, -0.0653, ..., 0.0522, 0.0545, 0.0471], + [-0.0902, -0.0831, -0.0978, ..., 0.0388, 0.0219, 0.0155], + ..., + [-0.1067, -0.1102, -0.0686, ..., 0.0180, 0.0678, 0.0513], + [-0.0216, -0.0193, -0.0429, ..., -0.1328, -0.1500, -0.1623], + [ 0.0808, 0.0815, 0.0723, ..., -0.0261, -0.0189, 0.0047]], + device='cuda:0'), grad: tensor([[-2.8858e-03, -1.5879e-03, -1.5993e-03, ..., -4.1032e-04, + -5.5933e-04, -5.7220e-04], + [ 2.7132e-04, 1.4031e-04, 1.3936e-04, ..., 5.5403e-05, + 6.7294e-05, 6.7949e-05], + [ 7.6151e-04, 4.0030e-04, 4.0221e-04, ..., 1.1492e-04, + 1.5497e-04, 1.5962e-04], + ..., + [ 1.3828e-03, 7.1383e-04, 7.1812e-04, ..., 1.9825e-04, + 2.7800e-04, 2.9016e-04], + [-2.2149e-04, 8.4192e-06, 1.9401e-05, ..., -1.2338e-04, + -1.3256e-04, -1.3959e-04], + [ 5.5552e-04, 2.9540e-04, 2.9325e-04, ..., 1.1337e-04, + 1.3423e-04, 1.3423e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0338, 0.0429, -0.0577, -0.0219, 0.0494, -0.0337, 0.0586], + device='cuda:0'), grad: tensor([-0.0040, 0.0004, 0.0011, 0.0004, 0.0021, -0.0009, 0.0009], + device='cuda:0') +306 +0.0015446867550656784 +changing lr +epoch 52, time 430.64, cls_loss 0.0035 cls_loss_mapping 0.0352 cls_loss_causal 0.6378 re_mapping 0.0435 re_causal 0.0466 /// teacc 85.58 lr 0.00138603 +Epoch 54, weight, value: tensor([[ 0.2474, 0.2477, 0.2796, ..., -0.0150, -0.0034, -0.0145], + [-0.0488, -0.0439, -0.0652, ..., 0.0520, 0.0543, 0.0469], + [-0.0901, -0.0832, -0.0979, ..., 0.0387, 0.0218, 0.0155], + ..., + [-0.1063, -0.1099, -0.0683, ..., 0.0181, 0.0678, 0.0514], + [-0.0212, -0.0193, -0.0429, ..., -0.1324, -0.1496, -0.1619], + [ 0.0799, 0.0810, 0.0718, ..., -0.0262, -0.0191, 0.0045]], + device='cuda:0'), grad: tensor([[ 4.4417e-04, 1.6558e-04, 1.5855e-04, ..., 2.1517e-04, + 2.1636e-04, 2.3675e-04], + [ 9.7871e-05, 3.1441e-05, 2.6315e-05, ..., 5.4896e-05, + 5.1409e-05, 5.7578e-05], + [ 5.0247e-05, 2.3901e-05, 2.0757e-05, ..., 3.8058e-05, + 3.3408e-05, 3.6001e-05], + ..., + [-2.1362e-04, -1.0002e-04, -1.0782e-04, ..., -8.4937e-05, + -9.6142e-05, -1.0109e-04], + [ 1.8752e-04, 5.8442e-05, 4.9949e-05, ..., 9.8109e-05, + 9.3877e-05, 1.0526e-04], + [ 1.3046e-03, 4.1962e-04, 3.4881e-04, ..., 7.3671e-04, + 6.8760e-04, 7.7057e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0333, 0.0422, -0.0572, -0.0223, 0.0495, -0.0324, 0.0572], + device='cuda:0'), grad: tensor([ 0.0009, 0.0002, 0.0001, -0.0045, -0.0004, 0.0004, 0.0031], + device='cuda:0') +306 +0.001386025680863044 +changing lr +epoch 53, time 427.28, cls_loss 0.0021 cls_loss_mapping 0.0327 cls_loss_causal 0.5944 re_mapping 0.0434 re_causal 0.0465 /// teacc 88.94 lr 0.00123464 +Epoch 55, weight, value: tensor([[ 0.2475, 0.2476, 0.2795, ..., -0.0149, -0.0033, -0.0144], + [-0.0489, -0.0439, -0.0652, ..., 0.0518, 0.0542, 0.0468], + [-0.0900, -0.0831, -0.0978, ..., 0.0387, 0.0218, 0.0155], + ..., + [-0.1068, -0.1099, -0.0684, ..., 0.0180, 0.0676, 0.0511], + [-0.0210, -0.0193, -0.0429, ..., -0.1321, -0.1493, -0.1615], + [ 0.0799, 0.0808, 0.0717, ..., -0.0261, -0.0190, 0.0045]], + device='cuda:0'), grad: tensor([[ 3.1680e-05, 1.2154e-06, -2.8219e-07, ..., 1.5661e-05, + 1.7688e-05, 1.8716e-05], + [ 2.3949e-04, 2.4959e-05, 1.8969e-05, ..., 1.3852e-04, + 1.4877e-04, 1.5247e-04], + [ 1.6904e-04, 3.0249e-05, 1.9073e-05, ..., 8.3625e-05, + 9.3639e-05, 9.9063e-05], + ..., + [ 3.6716e-04, 5.6565e-05, 3.5137e-05, ..., 1.9920e-04, + 2.1935e-04, 2.2936e-04], + [ 4.8161e-05, 3.6024e-06, -6.8396e-06, ..., 5.0902e-05, + 5.4926e-05, 5.8174e-05], + [ 2.0003e-04, 3.6567e-05, 2.5868e-05, ..., 9.4771e-05, + 1.0562e-04, 1.1134e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0327, 0.0418, -0.0569, -0.0223, 0.0481, -0.0316, 0.0574], + device='cuda:0'), grad: tensor([ 9.8288e-05, 6.6566e-04, 4.4632e-04, -2.8362e-03, 9.7656e-04, + 1.2803e-04, 5.2500e-04], device='cuda:0') +306 +0.0012346426699819469 +changing lr +epoch 54, time 429.05, cls_loss 0.0017 cls_loss_mapping 0.0296 cls_loss_causal 0.6024 re_mapping 0.0432 re_causal 0.0464 /// teacc 87.50 lr 0.00109084 +Epoch 56, weight, value: tensor([[ 0.2472, 0.2475, 0.2793, ..., -0.0149, -0.0033, -0.0144], + [-0.0489, -0.0439, -0.0651, ..., 0.0517, 0.0541, 0.0467], + [-0.0899, -0.0831, -0.0977, ..., 0.0386, 0.0218, 0.0155], + ..., + [-0.1066, -0.1098, -0.0684, ..., 0.0180, 0.0675, 0.0511], + [-0.0209, -0.0193, -0.0429, ..., -0.1319, -0.1491, -0.1613], + [ 0.0798, 0.0807, 0.0716, ..., -0.0261, -0.0190, 0.0045]], + device='cuda:0'), grad: tensor([[-3.3474e-03, -2.1229e-03, -2.0885e-03, ..., -2.0981e-04, + -3.1662e-04, -3.3784e-04], + [ 2.4605e-04, 1.4138e-04, 1.3793e-04, ..., 3.1292e-05, + 3.9667e-05, 4.2975e-05], + [ 1.2693e-03, 7.1526e-04, 6.9714e-04, ..., 1.6069e-04, + 2.0373e-04, 2.1863e-04], + ..., + [ 6.1178e-04, 3.5381e-04, 3.4547e-04, ..., 7.1764e-05, + 9.2328e-05, 9.9242e-05], + [ 6.0844e-04, 2.5725e-04, 2.4486e-04, ..., 1.1837e-04, + 1.4007e-04, 1.4651e-04], + [ 1.3056e-03, 7.2861e-04, 7.1001e-04, ..., 1.7309e-04, + 2.1768e-04, 2.3365e-04]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0329, 0.0416, -0.0569, -0.0224, 0.0480, -0.0312, 0.0576], + device='cuda:0'), grad: tensor([-0.0040, 0.0004, 0.0019, -0.0025, 0.0009, 0.0013, 0.0020], + device='cuda:0') +306 +0.0010908425876598518 +changing lr +epoch 55, time 426.80, cls_loss 0.0022 cls_loss_mapping 0.0329 cls_loss_causal 0.6184 re_mapping 0.0427 re_causal 0.0458 /// teacc 86.06 lr 0.00095492 +Epoch 57, weight, value: tensor([[ 0.2475, 0.2476, 0.2794, ..., -0.0149, -0.0033, -0.0143], + [-0.0490, -0.0439, -0.0651, ..., 0.0516, 0.0539, 0.0466], + [-0.0899, -0.0830, -0.0976, ..., 0.0385, 0.0217, 0.0154], + ..., + [-0.1065, -0.1096, -0.0683, ..., 0.0180, 0.0674, 0.0511], + [-0.0207, -0.0193, -0.0429, ..., -0.1316, -0.1488, -0.1610], + [ 0.0791, 0.0803, 0.0713, ..., -0.0262, -0.0191, 0.0043]], + device='cuda:0'), grad: tensor([[ 4.2391e-04, 1.2326e-04, 1.0896e-04, ..., 1.5497e-04, + 1.8597e-04, 2.0492e-04], + [-1.5807e-04, -5.4806e-05, -4.8727e-05, ..., -3.1233e-05, + -3.6001e-05, -4.9263e-05], + [ 2.2995e-04, 7.9930e-05, 6.9261e-05, ..., 8.5711e-05, + 1.0091e-04, 9.5487e-05], + ..., + [-5.3787e-04, -1.6129e-04, -1.4102e-04, ..., -2.2840e-04, + -2.7609e-04, -2.7633e-04], + [ 2.2441e-05, 1.0282e-05, 1.1332e-05, ..., 7.6592e-06, + 1.0222e-05, 1.0461e-05], + [ 9.7454e-06, 8.0746e-07, -2.0936e-06, ..., 9.2760e-06, + 1.2159e-05, 1.1526e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0323, 0.0411, -0.0570, -0.0218, 0.0477, -0.0305, 0.0565], + device='cuda:0'), grad: tensor([ 1.1177e-03, -3.0947e-04, 5.4264e-04, 2.3723e-05, -1.4601e-03, + 4.5985e-05, 4.0382e-05], device='cuda:0') +306 +0.000954915028125264 +changing lr +epoch 56, time 428.67, cls_loss 0.0014 cls_loss_mapping 0.0284 cls_loss_causal 0.5831 re_mapping 0.0422 re_causal 0.0451 /// teacc 89.42 lr 0.00082713 +Epoch 58, weight, value: tensor([[ 0.2476, 0.2476, 0.2793, ..., -0.0148, -0.0033, -0.0143], + [-0.0488, -0.0438, -0.0650, ..., 0.0515, 0.0539, 0.0466], + [-0.0900, -0.0830, -0.0976, ..., 0.0384, 0.0216, 0.0153], + ..., + [-0.1065, -0.1095, -0.0683, ..., 0.0180, 0.0674, 0.0510], + [-0.0208, -0.0194, -0.0429, ..., -0.1315, -0.1486, -0.1608], + [ 0.0788, 0.0800, 0.0710, ..., -0.0262, -0.0191, 0.0043]], + device='cuda:0'), grad: tensor([[ 1.0282e-05, -2.3227e-06, -4.9472e-06, ..., 5.3830e-06, + 5.2825e-06, 6.8247e-06], + [-1.4150e-04, -3.4779e-05, -2.1845e-05, ..., -4.9174e-05, + -5.9396e-05, -6.2883e-05], + [ 5.2303e-05, 1.5765e-05, 1.3009e-05, ..., 1.6302e-05, + 1.9863e-05, 2.0280e-05], + ..., + [ 2.0757e-05, 5.6326e-06, 4.7274e-06, ..., 6.5118e-06, + 7.2829e-06, 7.7114e-06], + [-4.5560e-06, 1.7434e-06, 8.5402e-07, ..., -4.2468e-06, + -4.0457e-06, -3.2615e-06], + [ 5.4955e-05, 1.2606e-05, 7.1153e-06, ..., 2.2396e-05, + 2.7567e-05, 2.8089e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0320, 0.0412, -0.0571, -0.0216, 0.0475, -0.0304, 0.0562], + device='cuda:0'), grad: tensor([ 5.3912e-05, -3.7479e-04, 1.2082e-04, 2.0117e-05, 5.1171e-05, + -1.5542e-05, 1.4389e-04], device='cuda:0') +306 +0.0008271337313934874 +changing lr +epoch 57, time 423.71, cls_loss 0.0016 cls_loss_mapping 0.0269 cls_loss_causal 0.5705 re_mapping 0.0421 re_causal 0.0451 /// teacc 87.98 lr 0.00070776 +Epoch 59, weight, value: tensor([[ 0.2474, 0.2474, 0.2791, ..., -0.0149, -0.0033, -0.0143], + [-0.0487, -0.0437, -0.0649, ..., 0.0515, 0.0538, 0.0465], + [-0.0899, -0.0830, -0.0975, ..., 0.0383, 0.0216, 0.0153], + ..., + [-0.1063, -0.1094, -0.0682, ..., 0.0181, 0.0674, 0.0510], + [-0.0207, -0.0194, -0.0429, ..., -0.1313, -0.1485, -0.1606], + [ 0.0785, 0.0799, 0.0709, ..., -0.0262, -0.0192, 0.0042]], + device='cuda:0'), grad: tensor([[-9.9182e-05, -6.8128e-05, -6.8247e-05, ..., -9.6485e-06, + -9.5665e-06, -1.1303e-05], + [-1.1706e-04, -3.6120e-05, -3.3170e-05, ..., -3.1769e-05, + -3.9518e-05, -4.3839e-05], + [ 1.0276e-04, 3.3408e-05, 3.0905e-05, ..., 3.1918e-05, + 3.6389e-05, 3.8296e-05], + ..., + [-8.3864e-05, 3.0044e-06, 7.2084e-06, ..., -4.6194e-05, + -5.0426e-05, -5.1051e-05], + [ 8.7619e-05, 2.3827e-05, 2.1413e-05, ..., 3.0056e-05, + 3.3259e-05, 3.5048e-05], + [ 8.0526e-05, 3.5167e-05, 3.3796e-05, ..., 1.6659e-05, + 1.9476e-05, 2.1666e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0322, 0.0413, -0.0569, -0.0217, 0.0476, -0.0303, 0.0558], + device='cuda:0'), grad: tensor([-1.0520e-04, -2.8086e-04, 2.3329e-04, 6.8605e-05, -2.8443e-04, + 2.1255e-04, 1.5569e-04], device='cuda:0') +306 +0.00070775603199067 +changing lr +epoch 58, time 421.70, cls_loss 0.0019 cls_loss_mapping 0.0311 cls_loss_causal 0.5978 re_mapping 0.0416 re_causal 0.0446 /// teacc 88.46 lr 0.00059702 +Epoch 60, weight, value: tensor([[ 0.2471, 0.2473, 0.2789, ..., -0.0149, -0.0034, -0.0144], + [-0.0483, -0.0436, -0.0647, ..., 0.0515, 0.0539, 0.0466], + [-0.0900, -0.0830, -0.0975, ..., 0.0382, 0.0215, 0.0152], + ..., + [-0.1060, -0.1092, -0.0680, ..., 0.0181, 0.0674, 0.0510], + [-0.0210, -0.0194, -0.0429, ..., -0.1313, -0.1484, -0.1605], + [ 0.0782, 0.0797, 0.0707, ..., -0.0263, -0.0192, 0.0042]], + device='cuda:0'), grad: tensor([[ 9.6798e-05, 3.5226e-05, 3.2395e-05, ..., 2.2635e-05, + 2.4319e-05, 2.8834e-05], + [-1.2958e-04, -2.9504e-05, -2.5034e-05, ..., -5.7399e-05, + -6.4015e-05, -7.1168e-05], + [ 2.7716e-05, 7.5512e-06, 6.8322e-06, ..., 8.3372e-06, + 8.8662e-06, 1.0163e-05], + ..., + [ 1.3828e-05, 3.9861e-06, 3.5781e-06, ..., 7.3947e-06, + 5.8748e-06, 6.3442e-06], + [-9.2566e-05, -2.2456e-05, -2.1636e-05, ..., -2.6435e-05, + -1.8954e-05, -2.1636e-05], + [-9.8124e-06, -1.6779e-05, -1.5572e-05, ..., 9.1493e-06, + 7.4469e-06, 6.7241e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0326, 0.0420, -0.0573, -0.0211, 0.0478, -0.0307, 0.0556], + device='cuda:0'), grad: tensor([ 2.0659e-04, -3.7932e-04, 6.7472e-05, 2.5487e-04, 4.3511e-05, + -2.1946e-04, 2.6584e-05], device='cuda:0') +306 +0.0005970223407163104 +changing lr +epoch 59, time 419.85, cls_loss 0.0019 cls_loss_mapping 0.0281 cls_loss_causal 0.6170 re_mapping 0.0414 re_causal 0.0446 /// teacc 86.54 lr 0.00049516 +Epoch 61, weight, value: tensor([[ 0.2471, 0.2472, 0.2789, ..., -0.0149, -0.0034, -0.0144], + [-0.0482, -0.0435, -0.0646, ..., 0.0515, 0.0538, 0.0465], + [-0.0900, -0.0830, -0.0975, ..., 0.0382, 0.0215, 0.0152], + ..., + [-0.1060, -0.1092, -0.0680, ..., 0.0181, 0.0673, 0.0510], + [-0.0211, -0.0195, -0.0430, ..., -0.1312, -0.1483, -0.1604], + [ 0.0781, 0.0796, 0.0706, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[ 3.0851e-04, 8.7559e-05, 8.5950e-05, ..., 8.6188e-05, + 9.1970e-05, 9.9599e-05], + [ 9.4950e-05, 2.9132e-05, 2.7582e-05, ..., 3.2783e-05, + 3.2723e-05, 3.5316e-05], + [ 2.3806e-04, 6.7055e-05, 6.4909e-05, ..., 6.5923e-05, + 6.9439e-05, 7.5936e-05], + ..., + [-6.3276e-04, -1.7273e-04, -1.7107e-04, ..., -1.5140e-04, + -1.6677e-04, -1.8263e-04], + [ 4.5270e-05, 1.0677e-05, 1.3418e-05, ..., 2.3127e-05, + 2.5794e-05, 2.4602e-05], + [ 1.1426e-04, 3.2634e-05, 3.1173e-05, ..., 3.8356e-05, + 4.0770e-05, 4.3154e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0325, 0.0420, -0.0571, -0.0209, 0.0476, -0.0309, 0.0554], + device='cuda:0'), grad: tensor([ 6.9857e-04, 2.1350e-04, 5.4169e-04, -3.6597e-04, -1.4334e-03, + 8.5652e-05, 2.5988e-04], device='cuda:0') +306 +0.0004951556604879052 +changing lr +epoch 60, time 418.06, cls_loss 0.0022 cls_loss_mapping 0.0259 cls_loss_causal 0.6041 re_mapping 0.0412 re_causal 0.0443 /// teacc 86.54 lr 0.00040236 +Epoch 62, weight, value: tensor([[ 0.2470, 0.2472, 0.2788, ..., -0.0149, -0.0034, -0.0144], + [-0.0483, -0.0435, -0.0646, ..., 0.0514, 0.0538, 0.0465], + [-0.0900, -0.0830, -0.0975, ..., 0.0382, 0.0215, 0.0152], + ..., + [-0.1060, -0.1091, -0.0680, ..., 0.0181, 0.0673, 0.0510], + [-0.0210, -0.0195, -0.0430, ..., -0.1311, -0.1482, -0.1603], + [ 0.0780, 0.0795, 0.0705, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[-1.7226e-04, -1.2827e-04, -1.2863e-04, ..., -2.8118e-05, + -2.7418e-05, -3.1918e-05], + [ 7.6115e-05, 2.7344e-05, 2.4825e-05, ..., 2.2009e-05, + 1.9506e-05, 2.3991e-05], + [-1.7416e-04, -4.3243e-05, -3.6567e-05, ..., -5.5730e-05, + -6.6042e-05, -7.8082e-05], + ..., + [ 1.4949e-04, 5.6535e-05, 5.2422e-05, ..., 3.9339e-05, + 4.3750e-05, 5.2392e-05], + [ 3.4302e-05, 1.2308e-05, 1.1645e-05, ..., -4.5598e-06, + 4.0010e-06, 5.4576e-06], + [ 2.7195e-05, 5.8621e-05, 6.1750e-05, ..., 8.6352e-06, + 8.2105e-06, 5.7817e-06]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0326, 0.0417, -0.0571, -0.0207, 0.0475, -0.0307, 0.0555], + device='cuda:0'), grad: tensor([-1.1986e-04, 1.8978e-04, -5.1594e-04, 1.7297e-04, 3.4547e-04, + 1.6943e-05, -8.9169e-05], device='cuda:0') +306 +0.00040236113724274745 +changing lr +epoch 61, time 417.60, cls_loss 0.0014 cls_loss_mapping 0.0258 cls_loss_causal 0.5725 re_mapping 0.0409 re_causal 0.0440 /// teacc 87.50 lr 0.00031883 +Epoch 63, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0144], + [-0.0483, -0.0435, -0.0645, ..., 0.0514, 0.0538, 0.0465], + [-0.0899, -0.0830, -0.0975, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1060, -0.1090, -0.0679, ..., 0.0181, 0.0672, 0.0509], + [-0.0210, -0.0195, -0.0430, ..., -0.1310, -0.1482, -0.1602], + [ 0.0780, 0.0794, 0.0704, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[-3.8743e-06, -7.7784e-06, -8.5309e-06, ..., 9.4250e-07, + 1.3597e-06, 1.5721e-06], + [ 1.4007e-05, 4.0494e-06, 3.5055e-06, ..., 4.4778e-06, + 4.9956e-06, 5.0962e-06], + [-1.5117e-05, -8.3074e-07, 8.9593e-07, ..., -2.8815e-06, + -3.7625e-06, -4.4890e-06], + ..., + [ 3.1926e-06, 2.3656e-07, -2.8312e-07, ..., 2.5332e-07, + 3.2783e-07, 6.7800e-07], + [-1.6674e-05, -3.1460e-06, -2.5742e-06, ..., -6.2101e-06, + -7.0110e-06, -7.0184e-06], + [ 1.7866e-05, 6.9812e-06, 6.5118e-06, ..., 4.3884e-06, + 4.9882e-06, 5.0478e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0326, 0.0417, -0.0570, -0.0206, 0.0473, -0.0306, 0.0555], + device='cuda:0'), grad: tensor([ 9.9242e-06, 3.3975e-05, -4.4495e-05, 3.4831e-07, 9.3728e-06, + -4.6074e-05, 3.7163e-05], device='cuda:0') +306 +0.00031882564680131423 +changing lr +epoch 62, time 418.99, cls_loss 0.0020 cls_loss_mapping 0.0323 cls_loss_causal 0.5851 re_mapping 0.0406 re_causal 0.0436 /// teacc 88.46 lr 0.00024472 +Epoch 64, weight, value: tensor([[ 0.2469, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0144], + [-0.0482, -0.0434, -0.0645, ..., 0.0514, 0.0538, 0.0465], + [-0.0899, -0.0829, -0.0975, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1060, -0.1090, -0.0679, ..., 0.0181, 0.0672, 0.0509], + [-0.0210, -0.0195, -0.0430, ..., -0.1310, -0.1481, -0.1602], + [ 0.0778, 0.0793, 0.0704, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[ 1.2884e-03, 1.9336e-04, 1.0115e-04, ..., 3.2020e-04, + 3.5286e-04, 3.9339e-04], + [ 6.8521e-04, 1.0777e-04, 5.8651e-05, ..., 1.6892e-04, + 1.8585e-04, 2.0766e-04], + [ 1.1377e-03, 1.6427e-04, 7.5102e-05, ..., 2.8133e-04, + 3.1304e-04, 3.5000e-04], + ..., + [ 2.3975e-03, 3.2902e-04, 1.3816e-04, ..., 5.9986e-04, + 6.6805e-04, 7.4720e-04], + [-8.9111e-03, -1.2150e-03, -5.0020e-04, ..., -2.2259e-03, + -2.4834e-03, -2.7771e-03], + [ 3.9291e-04, 1.2249e-05, -3.6687e-05, ..., 1.0353e-04, + 1.2201e-04, 1.3936e-04]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0327, 0.0418, -0.0569, -0.0207, 0.0473, -0.0304, 0.0553], + device='cuda:0'), grad: tensor([ 0.0043, 0.0023, 0.0039, 0.0104, 0.0083, -0.0308, 0.0017], + device='cuda:0') +306 +0.0002447174185242325 +changing lr +epoch 63, time 414.73, cls_loss 0.0017 cls_loss_mapping 0.0244 cls_loss_causal 0.6037 re_mapping 0.0406 re_causal 0.0438 /// teacc 86.06 lr 0.00018019 +Epoch 65, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0144], + [-0.0482, -0.0434, -0.0645, ..., 0.0514, 0.0538, 0.0465], + [-0.0899, -0.0829, -0.0975, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1059, -0.1090, -0.0679, ..., 0.0181, 0.0672, 0.0509], + [-0.0209, -0.0195, -0.0429, ..., -0.1309, -0.1480, -0.1601], + [ 0.0777, 0.0792, 0.0703, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[-4.4964e-06, -1.0490e-05, -1.0796e-05, ..., 1.2871e-06, + 1.4622e-06, 2.0340e-06], + [ 7.4387e-05, 1.5885e-05, 1.3448e-05, ..., 1.7881e-05, + 1.8552e-05, 2.2933e-05], + [ 9.9063e-05, 2.2203e-05, 1.9148e-05, ..., 2.7061e-05, + 2.4825e-05, 3.1203e-05], + ..., + [ 2.1532e-05, 5.4277e-06, 4.3362e-06, ..., -1.9260e-06, + -1.6876e-06, -8.2701e-07], + [ 1.1361e-04, 2.1651e-05, 1.7896e-05, ..., 3.0607e-05, + 2.9683e-05, 3.7044e-05], + [-7.2420e-05, -2.1517e-05, -1.9461e-05, ..., -1.2890e-05, + -1.8656e-05, -2.1428e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0326, 0.0418, -0.0570, -0.0208, 0.0472, -0.0302, 0.0552], + device='cuda:0'), grad: tensor([ 2.3663e-05, 1.9288e-04, 2.5082e-04, -6.5517e-04, 5.3644e-05, + 3.0255e-04, -1.6916e-04], device='cuda:0') +306 +0.0001801856965207339 +changing lr +epoch 64, time 413.22, cls_loss 0.0016 cls_loss_mapping 0.0248 cls_loss_causal 0.6161 re_mapping 0.0405 re_causal 0.0437 /// teacc 88.94 lr 0.00012536 +Epoch 66, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0144], + [-0.0481, -0.0434, -0.0645, ..., 0.0513, 0.0538, 0.0465], + [-0.0898, -0.0829, -0.0974, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1059, -0.1090, -0.0679, ..., 0.0180, 0.0672, 0.0509], + [-0.0209, -0.0195, -0.0430, ..., -0.1309, -0.1480, -0.1601], + [ 0.0777, 0.0792, 0.0703, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[-7.0512e-05, -4.7058e-05, -5.0157e-05, ..., -8.2478e-06, + -8.2850e-06, -9.5665e-06], + [ 2.0787e-05, 9.5218e-06, 9.8273e-06, ..., 3.1665e-06, + 3.2280e-06, 4.0270e-06], + [-4.8801e-06, 1.4380e-06, 2.3991e-06, ..., -1.5441e-06, + -1.7043e-06, -3.0026e-06], + ..., + [ 7.8917e-05, 3.0667e-05, 3.0696e-05, ..., 1.4104e-05, + 1.4886e-05, 1.8507e-05], + [ 4.3333e-05, 1.1683e-05, 1.1191e-05, ..., 9.6634e-06, + 9.5069e-06, 1.1683e-05], + [-9.7632e-05, -2.0340e-05, -1.8418e-05, ..., -2.1979e-05, + -2.2665e-05, -2.7910e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0326, 0.0419, -0.0568, -0.0208, 0.0471, -0.0302, 0.0551], + device='cuda:0'), grad: tensor([-8.1718e-05, 3.6746e-05, -2.0251e-05, 5.1826e-05, 1.5628e-04, + 1.0252e-04, -2.4605e-04], device='cuda:0') +306 +0.000125360439090882 +changing lr +epoch 65, time 415.02, cls_loss 0.0015 cls_loss_mapping 0.0254 cls_loss_causal 0.5788 re_mapping 0.0404 re_causal 0.0434 /// teacc 87.50 lr 0.00008035 +Epoch 67, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0144], + [-0.0481, -0.0434, -0.0645, ..., 0.0513, 0.0537, 0.0465], + [-0.0898, -0.0829, -0.0974, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1060, -0.1090, -0.0679, ..., 0.0180, 0.0672, 0.0509], + [-0.0209, -0.0195, -0.0430, ..., -0.1309, -0.1480, -0.1600], + [ 0.0777, 0.0792, 0.0702, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[ 4.1676e-04, 1.7440e-04, 1.6701e-04, ..., 7.9632e-05, + 8.6963e-05, 9.7990e-05], + [ 1.0794e-04, 4.8518e-05, 4.6700e-05, ..., 2.0698e-05, + 2.1353e-05, 2.3842e-05], + [ 1.4484e-04, 6.3241e-05, 6.0648e-05, ..., 2.8014e-05, + 3.1024e-05, 3.4124e-05], + ..., + [-2.1820e-03, -9.4128e-04, -9.0265e-04, ..., -4.2129e-04, + -4.6349e-04, -5.1355e-04], + [ 8.1873e-04, 3.5405e-04, 3.3951e-04, ..., 1.5831e-04, + 1.7440e-04, 1.9288e-04], + [ 6.5470e-04, 2.8372e-04, 2.7227e-04, ..., 1.2708e-04, + 1.4138e-04, 1.5545e-04]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0325, 0.0418, -0.0568, -0.0208, 0.0470, -0.0302, 0.0551], + device='cuda:0'), grad: tensor([ 9.2602e-04, 2.3615e-04, 3.1257e-04, 8.7500e-05, -4.7569e-03, + 1.7815e-03, 1.4172e-03], device='cuda:0') +306 +8.03520570068517e-05 +changing lr +epoch 66, time 411.65, cls_loss 0.0014 cls_loss_mapping 0.0245 cls_loss_causal 0.5859 re_mapping 0.0403 re_causal 0.0433 /// teacc 88.46 lr 0.00004525 +Epoch 68, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0143], + [-0.0482, -0.0434, -0.0645, ..., 0.0513, 0.0537, 0.0464], + [-0.0898, -0.0829, -0.0974, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1060, -0.1089, -0.0679, ..., 0.0180, 0.0671, 0.0509], + [-0.0209, -0.0195, -0.0430, ..., -0.1309, -0.1480, -0.1600], + [ 0.0776, 0.0792, 0.0702, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[ 6.9320e-05, 2.5973e-05, 2.5213e-05, ..., 1.3977e-05, + 1.4402e-05, 1.9848e-05], + [ 9.1970e-05, 1.9833e-05, 1.6674e-05, ..., 5.2363e-05, + 5.4508e-05, 6.0260e-05], + [ 1.2094e-04, 4.5180e-05, 4.3631e-05, ..., 2.4304e-05, + 2.4974e-05, 3.4660e-05], + ..., + [ 1.0721e-05, 3.2503e-06, 2.9113e-06, ..., 4.1090e-06, + 4.3735e-06, 5.1148e-06], + [-1.9260e-06, 8.3819e-08, 8.9779e-07, ..., 8.9593e-07, + 7.9162e-07, 1.0133e-06], + [-2.2829e-04, -8.7917e-05, -8.5890e-05, ..., -4.2439e-05, + -4.3660e-05, -6.2108e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0325, 0.0418, -0.0568, -0.0208, 0.0470, -0.0302, 0.0551], + device='cuda:0'), grad: tensor([ 1.3232e-04, 2.3067e-04, 2.3127e-04, -1.8311e-04, 2.3663e-05, + -9.8050e-06, -4.2439e-04], device='cuda:0') +306 +4.5251191160326525e-05 +changing lr +epoch 67, time 411.01, cls_loss 0.0014 cls_loss_mapping 0.0238 cls_loss_causal 0.6263 re_mapping 0.0403 re_causal 0.0434 /// teacc 86.06 lr 0.00002013 +Epoch 69, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0143], + [-0.0482, -0.0434, -0.0645, ..., 0.0513, 0.0537, 0.0464], + [-0.0898, -0.0829, -0.0974, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1060, -0.1089, -0.0679, ..., 0.0180, 0.0671, 0.0509], + [-0.0209, -0.0195, -0.0430, ..., -0.1309, -0.1479, -0.1600], + [ 0.0776, 0.0792, 0.0702, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[-1.0309e-03, -6.2466e-04, -6.1512e-04, ..., -9.6560e-05, + -1.5414e-04, -1.9073e-04], + [ 5.1165e-04, 2.3520e-04, 2.2519e-04, ..., 1.5378e-04, + 1.6558e-04, 1.8859e-04], + [ 3.4738e-04, 1.6952e-04, 1.6379e-04, ..., 7.6950e-05, + 8.8632e-05, 1.0359e-04], + ..., + [-6.6471e-04, -1.8024e-04, -1.5926e-04, ..., -3.7694e-04, + -3.6836e-04, -4.0412e-04], + [ 2.2173e-04, 7.6950e-05, 7.1943e-05, ..., 7.9215e-05, + 8.2910e-05, 9.3460e-05], + [ 4.4799e-04, 2.6631e-04, 2.6083e-04, ..., 8.7559e-05, + 1.0908e-04, 1.2445e-04]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0325, 0.0418, -0.0568, -0.0208, 0.0470, -0.0301, 0.0551], + device='cuda:0'), grad: tensor([-0.0012, 0.0010, 0.0006, 0.0004, -0.0018, 0.0005, 0.0005], + device='cuda:0') +306 +2.0128530023804673e-05 +changing lr +epoch 68, time 410.85, cls_loss 0.0015 cls_loss_mapping 0.0246 cls_loss_causal 0.5761 re_mapping 0.0403 re_causal 0.0433 /// teacc 89.42 lr 0.00000503 +Epoch 70, weight, value: tensor([[ 0.2470, 0.2471, 0.2787, ..., -0.0149, -0.0034, -0.0143], + [-0.0482, -0.0434, -0.0645, ..., 0.0513, 0.0537, 0.0464], + [-0.0898, -0.0829, -0.0974, ..., 0.0381, 0.0214, 0.0152], + ..., + [-0.1059, -0.1089, -0.0679, ..., 0.0180, 0.0671, 0.0509], + [-0.0209, -0.0195, -0.0430, ..., -0.1309, -0.1479, -0.1600], + [ 0.0776, 0.0791, 0.0702, ..., -0.0262, -0.0192, 0.0041]], + device='cuda:0'), grad: tensor([[ 3.0696e-05, 8.4490e-06, 8.9705e-06, ..., 7.9796e-06, + 8.0392e-06, 9.2313e-06], + [ 2.1577e-05, 5.4277e-06, 5.2042e-06, ..., 7.7486e-06, + 7.9274e-06, 8.6129e-06], + [ 1.7853e-06, 1.0766e-06, 1.6764e-08, ..., 1.8803e-06, + 2.0973e-06, 1.9260e-06], + ..., + [ 7.2084e-06, 2.0005e-06, 1.8878e-06, ..., 1.7714e-06, + 1.9129e-06, 2.2277e-06], + [-7.9811e-05, -2.2769e-05, -2.2128e-05, ..., -2.4080e-05, + -2.4781e-05, -2.7478e-05], + [ 1.5169e-05, 3.6657e-06, 3.6526e-06, ..., 5.2936e-06, + 5.4389e-06, 5.9977e-06]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0325, 0.0418, -0.0568, -0.0208, 0.0470, -0.0301, 0.0551], + device='cuda:0'), grad: tensor([ 8.1778e-05, 6.2466e-05, 9.5889e-06, 1.1146e-05, 2.0608e-05, + -2.3401e-04, 4.8488e-05], device='cuda:0') +306 +5.034667293427056e-06 +changing lr +epoch 69, time 414.33, cls_loss 0.0018 cls_loss_mapping 0.0269 cls_loss_causal 0.6085 re_mapping 0.0402 re_causal 0.0433 /// teacc 88.46 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1/art_painting_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 99.023438 69.795222 90.778443 73.199287 77.924317 + art_painting cartoon photo sketch Avg +do 99.023438 72.1843 92.035928 72.410283 78.876837 diff --git a/Meta-causal/code-withStyleAttack/64945.error b/Meta-causal/code-withStyleAttack/64945.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/64945.log b/Meta-causal/code-withStyleAttack/64945.log new file mode 100644 index 0000000000000000000000000000000000000000..fded65bae35888cad880615b022fe0045d7b6eea --- /dev/null +++ b/Meta-causal/code-withStyleAttack/64945.log @@ -0,0 +1,1954 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'cartoon', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_train.hdf5 torch.Size([2107, 3, 227, 227]) torch.Size([2107]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_val.hdf5 torch.Size([237, 3, 227, 227]) torch.Size([237]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[-0.0100, 0.0028, 0.0031, ..., -0.0029, -0.0129, 0.0067], + [ 0.0146, 0.0136, 0.0209, ..., 0.0182, -0.0162, -0.0063], + [ 0.0144, -0.0044, -0.0032, ..., 0.0196, 0.0067, 0.0059], + ..., + [-0.0163, -0.0022, 0.0082, ..., 0.0012, 0.0086, -0.0212], + [-0.0167, -0.0119, 0.0066, ..., -0.0117, 0.0125, -0.0117], + [-0.0118, -0.0128, -0.0087, ..., -0.0020, 0.0197, 0.0048]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0072, -0.0193, 0.0095, 0.0119, 0.0171, 0.0063, -0.0017], + device='cuda:0'), grad: None +351 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 485.47, cls_loss 16.9782 cls_loss_mapping 2.0665 cls_loss_causal 2.0978 re_mapping 0.2883 re_causal 0.2878 /// teacc 12.66 lr 0.00999497 +Epoch 2, weight, value: tensor([[ 0.0028, -0.0005, 0.0340, ..., -0.0002, 0.0233, 0.0462], + [ 0.0658, 0.0861, 0.0585, ..., 0.0947, 0.0172, 0.0346], + [ 0.0262, -0.0145, 0.0102, ..., 0.0781, 0.0733, 0.0588], + ..., + [-0.1058, -0.0852, -0.0944, ..., 0.0447, 0.0601, 0.0102], + [ 0.0011, 0.0218, 0.0048, ..., -0.0536, -0.0391, -0.0657], + [-0.0314, -0.0301, 0.0041, ..., -0.0928, -0.0691, -0.0766]], + device='cuda:0'), grad: tensor([[-0.0698, -0.0076, -0.0194, ..., -0.0092, -0.0196, -0.0026], + [-0.0486, -0.0072, -0.0131, ..., -0.0160, -0.0171, -0.0081], + [ 0.0616, 0.0076, 0.0175, ..., 0.0109, 0.0186, 0.0045], + ..., + [ 0.0679, 0.0086, 0.0195, ..., 0.0142, 0.0205, 0.0061], + [ 0.0363, 0.0044, 0.0101, ..., 0.0059, 0.0107, 0.0021], + [ 0.0342, 0.0043, 0.0096, ..., 0.0069, 0.0106, 0.0030]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0115, -0.2892, 0.0229, -0.1414, -0.0596, 0.3315, 0.1640], + device='cuda:0'), grad: tensor([-0.1210, -0.2117, 0.1382, -0.1489, 0.1758, 0.0772, 0.0903], + device='cuda:0') +351 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 485.94, cls_loss 2.0030 cls_loss_mapping 1.9396 cls_loss_causal 1.9387 re_mapping 0.0647 re_causal 0.0644 /// teacc 41.77 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 3.4078e-04, 1.6000e-03, 3.1084e-02, ..., -1.0942e-03, + 2.9201e-02, 5.4861e-02], + [ 6.9312e-02, 8.8419e-02, 5.7617e-02, ..., 9.5143e-02, + 1.9121e-02, 3.5263e-02], + [ 2.8979e-02, -1.5773e-02, 2.5921e-02, ..., 6.5189e-02, + 6.9842e-02, 5.3385e-02], + ..., + [-1.1333e-01, -9.1322e-02, -1.0055e-01, ..., 5.0898e-02, + 6.5492e-02, 1.1030e-02], + [-5.0384e-03, 1.4584e-02, 4.9783e-04, ..., -4.5509e-02, + -4.2445e-02, -7.5379e-02], + [-2.9434e-02, -2.5451e-02, 6.7909e-05, ..., -9.6225e-02, + -7.2842e-02, -7.1301e-02]], device='cuda:0'), grad: tensor([[ 0.0200, 0.0067, 0.0079, ..., 0.0014, 0.0034, -0.0002], + [ 0.0065, 0.0008, 0.0016, ..., 0.0011, 0.0014, 0.0007], + [ 0.0329, 0.0065, 0.0102, ..., 0.0055, 0.0080, 0.0035], + ..., + [ 0.0395, 0.0060, 0.0104, ..., 0.0067, 0.0091, 0.0043], + [-0.0082, 0.0013, -0.0012, ..., -0.0020, -0.0016, -0.0008], + [-0.0192, -0.0043, -0.0041, ..., -0.0010, -0.0025, -0.0002]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0612, -0.3493, 0.0748, -0.2091, -0.0425, 0.3946, 0.2087], + device='cuda:0'), grad: tensor([ 0.0385, 0.0165, 0.0712, -0.1410, 0.0951, -0.0314, -0.0487], + device='cuda:0') +351 +0.009979871469976196 +changing lr +epoch 2, time 484.44, cls_loss 1.5550 cls_loss_mapping 1.8738 cls_loss_causal 1.9254 re_mapping 0.0608 re_causal 0.0606 /// teacc 13.08 lr 0.00995475 +Epoch 4, weight, value: tensor([[-0.0137, -0.0117, 0.0201, ..., -0.0003, 0.0411, 0.0727], + [ 0.0566, 0.0884, 0.0561, ..., 0.0828, 0.0134, 0.0221], + [ 0.0523, 0.0052, 0.0546, ..., 0.0551, 0.0651, 0.0509], + ..., + [-0.1301, -0.1026, -0.1092, ..., 0.0596, 0.0749, 0.0162], + [-0.0134, 0.0029, -0.0080, ..., -0.0267, -0.0355, -0.0760], + [-0.0125, -0.0185, -0.0061, ..., -0.1114, -0.0908, -0.0794]], + device='cuda:0'), grad: tensor([[-0.1074, -0.0163, -0.0219, ..., -0.0164, -0.0173, -0.0177], + [-0.0030, -0.0003, -0.0006, ..., -0.0013, -0.0013, -0.0013], + [ 0.0052, 0.0004, 0.0009, ..., 0.0014, 0.0014, 0.0014], + ..., + [ 0.0270, 0.0045, 0.0053, ..., 0.0024, 0.0023, 0.0031], + [ 0.0175, 0.0033, 0.0036, ..., 0.0030, 0.0029, 0.0029], + [ 0.0384, 0.0056, 0.0080, ..., 0.0066, 0.0071, 0.0070]], + device='cuda:0') +Epoch 4, bias, value: tensor([-0.0258, -0.3557, 0.0860, -0.2310, -0.0076, 0.3735, 0.1763], + device='cuda:0'), grad: tensor([-0.2710, -0.0077, 0.0140, 0.0579, 0.0644, 0.0446, 0.0978], + device='cuda:0') +351 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 488.33, cls_loss 1.3546 cls_loss_mapping 1.7404 cls_loss_causal 1.8961 re_mapping 0.0588 re_causal 0.0585 /// teacc 53.59 lr 0.00991965 +Epoch 5, weight, value: tensor([[-0.0074, -0.0123, 0.0209, ..., -0.0032, 0.0541, 0.0888], + [ 0.0611, 0.0895, 0.0580, ..., 0.0690, 0.0110, 0.0133], + [ 0.0503, 0.0057, 0.0587, ..., 0.0498, 0.0610, 0.0490], + ..., + [-0.1420, -0.1151, -0.1146, ..., 0.0570, 0.0713, 0.0154], + [-0.0224, -0.0038, -0.0130, ..., -0.0028, -0.0191, -0.0638], + [-0.0048, -0.0111, -0.0072, ..., -0.1203, -0.1105, -0.0966]], + device='cuda:0'), grad: tensor([[ 1.4026e-01, 1.3710e-02, 3.0167e-02, ..., 3.7689e-02, + 3.3051e-02, 3.6926e-02], + [ 8.0729e-04, 8.0884e-05, 1.7869e-04, ..., 1.8716e-04, + 1.3447e-04, 1.5807e-04], + [ 2.3224e-02, 2.2202e-03, 5.0621e-03, ..., 5.9090e-03, + 5.0125e-03, 6.0768e-03], + ..., + [-8.1116e-02, -8.6365e-03, -1.6846e-02, ..., -1.8463e-02, + -1.6403e-02, -1.9104e-02], + [ 4.2725e-03, 4.3392e-04, 9.1553e-04, ..., 1.0796e-03, + 7.8964e-04, 9.8324e-04], + [-9.1919e-02, -8.2321e-03, -2.0401e-02, ..., -2.7817e-02, + -2.3865e-02, -2.6306e-02]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0100, -0.3700, 0.0890, -0.2639, 0.0169, 0.3545, 0.1994], + device='cuda:0'), grad: tensor([ 0.4014, 0.0023, 0.0675, 0.0128, -0.2286, 0.0120, -0.2673], + device='cuda:0') +351 +0.009919647942993149 +changing lr +epoch 4, time 490.81, cls_loss 1.1636 cls_loss_mapping 1.6025 cls_loss_causal 1.8435 re_mapping 0.0581 re_causal 0.0578 /// teacc 47.26 lr 0.00987464 +Epoch 6, weight, value: tensor([[-0.0092, -0.0115, 0.0200, ..., -0.0068, 0.0601, 0.0934], + [ 0.0657, 0.0900, 0.0604, ..., 0.0629, 0.0111, 0.0086], + [ 0.0585, 0.0162, 0.0694, ..., 0.0436, 0.0533, 0.0450], + ..., + [-0.1502, -0.1233, -0.1191, ..., 0.0550, 0.0668, 0.0162], + [-0.0251, -0.0028, -0.0127, ..., 0.0180, -0.0031, -0.0522], + [-0.0082, -0.0168, -0.0138, ..., -0.1336, -0.1261, -0.1108]], + device='cuda:0'), grad: tensor([[ 0.0056, 0.0005, 0.0013, ..., 0.0016, 0.0015, 0.0011], + [-0.0229, -0.0027, -0.0048, ..., -0.0075, -0.0068, -0.0052], + [ 0.0082, 0.0008, 0.0021, ..., 0.0037, 0.0028, 0.0022], + ..., + [-0.0790, -0.0061, -0.0199, ..., -0.0227, -0.0231, -0.0172], + [ 0.0002, 0.0001, 0.0004, ..., -0.0002, 0.0006, 0.0005], + [ 0.0731, 0.0060, 0.0168, ..., 0.0200, 0.0208, 0.0149]], + device='cuda:0') +Epoch 6, bias, value: tensor([-0.0260, -0.3347, 0.0891, -0.2536, 0.0157, 0.3505, 0.1747], + device='cuda:0'), grad: tensor([ 0.0153, -0.0533, 0.0237, 0.0438, -0.2416, 0.0005, 0.2114], + device='cuda:0') +351 +0.009874639560909117 +changing lr +epoch 5, time 491.16, cls_loss 0.9659 cls_loss_mapping 1.4805 cls_loss_causal 1.8159 re_mapping 0.0580 re_causal 0.0577 /// teacc 44.30 lr 0.00981981 +Epoch 7, weight, value: tensor([[-0.0087, -0.0102, 0.0192, ..., -0.0007, 0.0825, 0.1130], + [ 0.0642, 0.0912, 0.0624, ..., 0.0528, 0.0038, -0.0031], + [ 0.0583, 0.0211, 0.0742, ..., 0.0440, 0.0475, 0.0432], + ..., + [-0.1536, -0.1256, -0.1215, ..., 0.0537, 0.0628, 0.0130], + [-0.0222, -0.0026, -0.0132, ..., 0.0380, 0.0139, -0.0401], + [-0.0087, -0.0225, -0.0157, ..., -0.1544, -0.1509, -0.1300]], + device='cuda:0'), grad: tensor([[ 2.9419e-02, 3.1052e-03, 8.2626e-03, ..., 1.6357e-02, + 1.6663e-02, 1.5762e-02], + [ 1.7862e-03, 1.0294e-04, 4.2748e-04, ..., 4.0174e-04, + 4.5538e-04, 3.9983e-04], + [ 6.8115e-02, 1.1337e-02, 2.0950e-02, ..., 1.4877e-02, + 1.6281e-02, 1.6724e-02], + ..., + [-7.9285e-02, -1.3260e-02, -2.5253e-02, ..., -2.6566e-02, + -2.7786e-02, -2.7908e-02], + [-8.7678e-05, 6.4820e-07, -4.7654e-05, ..., -4.4513e-04, + -1.5175e-04, -1.6284e-04], + [-4.0710e-02, -2.8210e-03, -9.1019e-03, ..., -1.0178e-02, + -1.1513e-02, -1.0269e-02]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0170, -0.3173, 0.0795, -0.2454, 0.0139, 0.3444, 0.1572], + device='cuda:0'), grad: tensor([ 0.0901, 0.0056, 0.1676, 0.0627, -0.2017, -0.0006, -0.1240], + device='cuda:0') +351 +0.009819814303479266 +changing lr +epoch 6, time 485.87, cls_loss 0.8556 cls_loss_mapping 1.3801 cls_loss_causal 1.7626 re_mapping 0.0591 re_causal 0.0588 /// teacc 16.46 lr 0.00975528 +Epoch 8, weight, value: tensor([[-0.0206, -0.0165, 0.0098, ..., -0.0012, 0.0856, 0.1165], + [ 0.0669, 0.0905, 0.0646, ..., 0.0561, 0.0145, 0.0017], + [ 0.0613, 0.0221, 0.0739, ..., 0.0347, 0.0348, 0.0349], + ..., + [-0.1449, -0.1215, -0.1127, ..., 0.0389, 0.0514, 0.0033], + [-0.0210, -0.0022, -0.0147, ..., 0.0576, 0.0264, -0.0269], + [-0.0144, -0.0258, -0.0177, ..., -0.1550, -0.1571, -0.1350]], + device='cuda:0'), grad: tensor([[-1.8387e-02, 2.7108e-04, -4.0817e-03, ..., 1.6880e-03, + -1.1263e-03, 6.9761e-04], + [ 2.3499e-02, 4.3144e-03, 4.4899e-03, ..., 8.8425e-03, + 8.3237e-03, 5.8784e-03], + [ 1.4435e-02, 1.4095e-03, 2.9545e-03, ..., 3.0537e-03, + 3.5877e-03, 2.1877e-03], + ..., + [-4.3671e-02, -9.4910e-03, -8.1940e-03, ..., -1.9760e-02, + -1.8127e-02, -1.3351e-02], + [-3.3302e-03, -8.7172e-06, -5.5170e-04, ..., -8.2636e-04, + -2.4056e-04, -1.7405e-04], + [ 1.0490e-02, 1.2846e-03, 1.9875e-03, ..., 2.7065e-03, + 2.8782e-03, 1.7948e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0110, -0.3280, 0.0902, -0.2517, 0.0217, 0.3492, 0.1446], + device='cuda:0'), grad: tensor([-0.0444, 0.0637, 0.0409, 0.0447, -0.1157, -0.0180, 0.0287], + device='cuda:0') +351 +0.009755282581475767 +changing lr +epoch 7, time 487.93, cls_loss 0.7146 cls_loss_mapping 1.2877 cls_loss_causal 1.6994 re_mapping 0.0594 re_causal 0.0592 /// teacc 34.60 lr 0.00968117 +Epoch 9, weight, value: tensor([[-0.0208, -0.0210, 0.0043, ..., -0.0036, 0.0879, 0.1206], + [ 0.0655, 0.0872, 0.0628, ..., 0.0514, 0.0141, -0.0047], + [ 0.0660, 0.0297, 0.0799, ..., 0.0356, 0.0282, 0.0334], + ..., + [-0.1432, -0.1176, -0.1069, ..., 0.0358, 0.0539, 0.0041], + [-0.0209, -0.0021, -0.0152, ..., 0.0649, 0.0277, -0.0243], + [-0.0217, -0.0287, -0.0192, ..., -0.1577, -0.1638, -0.1406]], + device='cuda:0'), grad: tensor([[-7.3004e-04, -3.0255e-04, -2.6321e-04, ..., 3.6263e-04, + 1.9729e-04, 2.9969e-04], + [ 4.7493e-04, 2.4632e-05, 8.4996e-05, ..., 2.2411e-04, + 2.1279e-04, 2.3365e-04], + [-1.7941e-05, -1.2837e-05, -1.4558e-05, ..., 2.2948e-06, + -4.7162e-06, -4.2170e-06], + ..., + [-1.8625e-03, -1.7643e-05, -2.9802e-04, ..., -1.2522e-03, + -1.2255e-03, -1.3466e-03], + [ 4.2328e-02, 1.7605e-03, 7.6981e-03, ..., 1.1963e-02, + 7.0267e-03, 8.6975e-03], + [-4.2358e-02, -1.7443e-03, -7.6790e-03, ..., -1.1917e-02, + -6.9313e-03, -8.6136e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0195, -0.3211, 0.0706, -0.2405, 0.0008, 0.3643, 0.1601], + device='cuda:0'), grad: tensor([-1.0853e-03, 1.5802e-03, -1.3411e-05, 6.2561e-03, -6.5804e-03, + 1.4355e-01, -1.4368e-01], device='cuda:0') +351 +0.009681174353198686 +changing lr +epoch 8, time 487.83, cls_loss 0.6065 cls_loss_mapping 1.1835 cls_loss_causal 1.6608 re_mapping 0.0617 re_causal 0.0615 /// teacc 49.37 lr 0.00959764 +Epoch 10, weight, value: tensor([[-2.4050e-02, -1.7761e-02, 3.4237e-03, ..., -1.7390e-03, + 8.9277e-02, 1.2650e-01], + [ 6.3247e-02, 8.4007e-02, 6.1289e-02, ..., 5.2398e-02, + 1.3658e-02, -8.1415e-03], + [ 6.9869e-02, 3.1879e-02, 8.0394e-02, ..., 2.8270e-02, + 1.8491e-02, 2.4165e-02], + ..., + [-1.4282e-01, -1.2039e-01, -1.0256e-01, ..., 2.5484e-02, + 5.2426e-02, 1.4157e-04], + [-2.1257e-02, -1.3096e-03, -1.7762e-02, ..., 7.3701e-02, + 2.8826e-02, -2.2097e-02], + [-2.1198e-02, -2.8250e-02, -1.8889e-02, ..., -1.6167e-01, + -1.6683e-01, -1.4144e-01]], device='cuda:0'), grad: tensor([[ 1.7500e-03, 3.7360e-04, 5.3549e-04, ..., 3.4451e-04, + 4.9019e-04, 4.7493e-04], + [ 2.2078e-04, 5.2124e-05, 7.2539e-05, ..., 4.3780e-05, + 6.1393e-05, 5.5730e-05], + [-4.3671e-02, -8.8120e-03, -1.5808e-02, ..., -1.7059e-02, + -1.6403e-02, -1.5350e-02], + ..., + [ 3.6896e-02, 7.3662e-03, 1.3657e-02, ..., 1.5488e-02, + 1.4389e-02, 1.3420e-02], + [-5.0688e-04, 7.3314e-06, -6.0052e-05, ..., -1.6499e-04, + -7.8738e-05, -2.9728e-05], + [ 1.5955e-03, 2.3997e-04, 4.1461e-04, ..., 3.8505e-04, + 3.8671e-04, 3.2830e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0185, -0.3178, 0.0688, -0.2319, 0.0088, 0.3629, 0.1423], + device='cuda:0'), grad: tensor([ 0.0046, 0.0006, -0.1209, 0.0099, 0.1028, -0.0015, 0.0044], + device='cuda:0') +351 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 496.62, cls_loss 0.5204 cls_loss_mapping 1.0647 cls_loss_causal 1.5735 re_mapping 0.0619 re_causal 0.0617 /// teacc 78.48 lr 0.00950484 +Epoch 11, weight, value: tensor([[-0.0143, -0.0127, 0.0037, ..., 0.0025, 0.0978, 0.1330], + [ 0.0605, 0.0804, 0.0595, ..., 0.0587, 0.0218, -0.0022], + [ 0.0698, 0.0287, 0.0795, ..., 0.0239, 0.0185, 0.0256], + ..., + [-0.1494, -0.1267, -0.1035, ..., 0.0184, 0.0450, -0.0050], + [-0.0252, 0.0017, -0.0171, ..., 0.0766, 0.0278, -0.0212], + [-0.0184, -0.0282, -0.0203, ..., -0.1655, -0.1755, -0.1504]], + device='cuda:0'), grad: tensor([[ 1.2459e-02, 1.3876e-03, 2.9373e-03, ..., 2.5539e-03, + 2.7657e-03, 3.2463e-03], + [ 1.9388e-03, 1.8775e-04, 4.2844e-04, ..., 4.3082e-04, + 4.6372e-04, 5.4216e-04], + [ 5.2612e-02, 1.4328e-02, 2.0798e-02, ..., 4.3564e-03, + 5.9853e-03, 6.3286e-03], + ..., + [ 2.6886e-02, 2.6340e-03, 6.0768e-03, ..., 5.0621e-03, + 5.5580e-03, 6.7787e-03], + [ 6.6996e-05, 9.0525e-06, 1.7419e-05, ..., 3.0063e-06, + 5.5954e-06, 9.2536e-06], + [-9.5337e-02, -1.8707e-02, -3.0579e-02, ..., -1.2718e-02, + -1.5114e-02, -1.7288e-02]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0015, -0.3192, 0.0555, -0.2433, 0.0173, 0.3439, 0.1586], + device='cuda:0'), grad: tensor([ 4.2358e-02, 6.6528e-03, 1.5479e-01, 4.8218e-03, 9.2896e-02, + 2.2638e-04, -3.0176e-01], device='cuda:0') +351 +0.009504844339512096 +changing lr +epoch 10, time 488.71, cls_loss 0.4416 cls_loss_mapping 0.9614 cls_loss_causal 1.5411 re_mapping 0.0639 re_causal 0.0637 /// teacc 43.88 lr 0.00940298 +Epoch 12, weight, value: tensor([[-0.0199, -0.0105, 0.0022, ..., 0.0044, 0.1020, 0.1346], + [ 0.0602, 0.0779, 0.0589, ..., 0.0587, 0.0236, -0.0019], + [ 0.0649, 0.0265, 0.0755, ..., 0.0258, 0.0215, 0.0291], + ..., + [-0.1561, -0.1310, -0.1052, ..., 0.0121, 0.0383, -0.0080], + [-0.0179, 0.0020, -0.0169, ..., 0.0814, 0.0301, -0.0190], + [-0.0118, -0.0252, -0.0170, ..., -0.1677, -0.1806, -0.1532]], + device='cuda:0'), grad: tensor([[ 6.2752e-04, 8.8215e-05, 1.4699e-04, ..., 1.3494e-04, + 8.3327e-05, 4.4316e-05], + [ 3.3450e-04, 4.9382e-05, 9.3460e-05, ..., 3.2276e-05, + -1.6272e-05, -3.8147e-05], + [-1.0910e-02, -1.5612e-03, -2.5730e-03, ..., -2.1000e-03, + -1.0977e-03, -4.0007e-04], + ..., + [ 5.4455e-04, 8.6308e-05, 1.0788e-04, ..., 6.6996e-05, + 9.3579e-06, -3.2425e-05], + [ 3.1796e-03, 4.5466e-04, 7.5006e-04, ..., 6.1369e-04, + 3.2234e-04, 1.1897e-04], + [ 3.2673e-03, 4.6039e-04, 7.7772e-04, ..., 6.8188e-04, + 4.0054e-04, 1.9753e-04]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0175, -0.3137, 0.0387, -0.2354, -0.0062, 0.3748, 0.1735], + device='cuda:0'), grad: tensor([ 0.0018, 0.0009, -0.0317, 0.0086, 0.0015, 0.0092, 0.0096], + device='cuda:0') +351 +0.009402977659283692 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 494.18, cls_loss 0.3891 cls_loss_mapping 0.8157 cls_loss_causal 1.4446 re_mapping 0.0666 re_causal 0.0663 /// teacc 78.90 lr 0.00929224 +Epoch 13, weight, value: tensor([[-0.0114, -0.0099, 0.0014, ..., 0.0046, 0.1073, 0.1377], + [ 0.0535, 0.0753, 0.0569, ..., 0.0620, 0.0234, -0.0039], + [ 0.0735, 0.0337, 0.0822, ..., 0.0240, 0.0212, 0.0305], + ..., + [-0.1620, -0.1316, -0.1037, ..., 0.0099, 0.0368, -0.0077], + [-0.0194, 0.0019, -0.0192, ..., 0.0848, 0.0300, -0.0185], + [-0.0162, -0.0297, -0.0210, ..., -0.1704, -0.1856, -0.1569]], + device='cuda:0'), grad: tensor([[ 4.6611e-05, 1.1183e-05, 1.9923e-05, ..., -2.2575e-05, + -2.5004e-05, -2.1055e-05], + [ 1.0766e-05, 3.0249e-06, 4.0717e-06, ..., 2.8312e-06, + 2.3656e-06, 2.8126e-06], + [-1.6193e-03, -5.1022e-04, -6.5613e-04, ..., -5.1498e-04, + -3.5930e-04, -4.0555e-04], + ..., + [ 1.3876e-04, 3.0369e-05, 4.1813e-05, ..., 6.6936e-05, + 7.4625e-05, 8.1360e-05], + [ 2.7239e-05, 7.3612e-06, 1.0043e-05, ..., 7.8529e-06, + 6.9402e-06, 8.0913e-06], + [ 6.2734e-06, 2.1681e-05, 1.8269e-05, ..., 1.8656e-05, + -9.0450e-06, -1.5110e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([-9.4110e-05, -3.2179e-01, 3.0043e-02, -2.2424e-01, -2.7306e-02, + 3.7682e-01, 1.8037e-01], device='cuda:0'), grad: tensor([ 1.3185e-04, 2.7537e-05, -3.9215e-03, 3.3684e-03, 3.8481e-04, + 7.0870e-05, -6.4135e-05], device='cuda:0') +351 +0.009292243968009333 +changing lr +epoch 12, time 487.57, cls_loss 0.3192 cls_loss_mapping 0.7209 cls_loss_causal 1.3775 re_mapping 0.0656 re_causal 0.0653 /// teacc 63.71 lr 0.00917287 +Epoch 14, weight, value: tensor([[-0.0128, -0.0094, 0.0006, ..., 0.0067, 0.1094, 0.1410], + [ 0.0535, 0.0740, 0.0570, ..., 0.0602, 0.0213, -0.0071], + [ 0.0651, 0.0308, 0.0764, ..., 0.0198, 0.0181, 0.0274], + ..., + [-0.1485, -0.1289, -0.0992, ..., 0.0107, 0.0365, -0.0067], + [-0.0197, 0.0018, -0.0197, ..., 0.0855, 0.0319, -0.0186], + [-0.0218, -0.0307, -0.0218, ..., -0.1687, -0.1871, -0.1557]], + device='cuda:0'), grad: tensor([[ 3.2940e-03, 4.0150e-04, 5.7077e-04, ..., 5.7507e-04, + 1.0185e-03, 9.1982e-04], + [ 1.8632e-04, 2.0161e-05, 5.4568e-05, ..., 1.0991e-04, + 1.2201e-04, 1.3220e-04], + [ 1.4639e-04, 1.1936e-05, 3.6836e-05, ..., 5.2154e-05, + 4.9412e-05, 5.9545e-05], + ..., + [ 3.3045e-04, 9.2089e-06, 7.9036e-05, ..., 4.6194e-05, + -1.2025e-05, 2.9653e-05], + [ 4.0016e-03, 2.8110e-04, 1.0900e-03, ..., 1.6613e-03, + 1.4162e-03, 1.8120e-03], + [-8.0948e-03, -7.3528e-04, -1.8692e-03, ..., -2.5005e-03, + -2.6455e-03, -3.0155e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0019, -0.3214, 0.0035, -0.2192, 0.0107, 0.3626, 0.1791], + device='cuda:0'), grad: tensor([ 0.0094, 0.0007, 0.0006, 0.0006, 0.0016, 0.0171, -0.0299], + device='cuda:0') +351 +0.009172866268606516 +changing lr +epoch 13, time 489.06, cls_loss 0.2306 cls_loss_mapping 0.6268 cls_loss_causal 1.2624 re_mapping 0.0687 re_causal 0.0684 /// teacc 16.46 lr 0.00904508 +Epoch 15, weight, value: tensor([[-0.0146, -0.0088, -0.0035, ..., 0.0025, 0.1056, 0.1395], + [ 0.0602, 0.0743, 0.0604, ..., 0.0627, 0.0234, -0.0063], + [ 0.0683, 0.0323, 0.0766, ..., 0.0169, 0.0138, 0.0250], + ..., + [-0.1461, -0.1274, -0.0961, ..., 0.0092, 0.0416, -0.0031], + [-0.0188, 0.0011, -0.0206, ..., 0.0915, 0.0352, -0.0144], + [-0.0272, -0.0327, -0.0228, ..., -0.1682, -0.1902, -0.1596]], + device='cuda:0'), grad: tensor([[ 0.0088, 0.0010, 0.0025, ..., 0.0017, 0.0029, 0.0036], + [ 0.0007, 0.0002, 0.0003, ..., 0.0004, 0.0003, 0.0004], + [-0.0638, -0.0165, -0.0286, ..., -0.0233, -0.0222, -0.0276], + ..., + [ 0.0265, 0.0029, 0.0075, ..., 0.0056, 0.0092, 0.0112], + [ 0.0006, 0.0003, 0.0004, ..., 0.0004, 0.0002, 0.0003], + [ 0.0212, 0.0095, 0.0139, ..., 0.0119, 0.0074, 0.0095]], + device='cuda:0') +Epoch 15, bias, value: tensor([-0.0048, -0.3123, 0.0152, -0.2279, 0.0035, 0.3635, 0.1762], + device='cuda:0'), grad: tensor([ 0.0277, 0.0026, -0.2307, 0.0250, 0.0837, 0.0027, 0.0891], + device='cuda:0') +351 +0.00904508497187474 +changing lr +epoch 14, time 487.91, cls_loss 0.2457 cls_loss_mapping 0.5984 cls_loss_causal 1.2677 re_mapping 0.0668 re_causal 0.0666 /// teacc 56.12 lr 0.00890916 +Epoch 16, weight, value: tensor([[-0.0190, -0.0126, -0.0079, ..., -0.0027, 0.1032, 0.1370], + [ 0.0610, 0.0744, 0.0602, ..., 0.0630, 0.0232, -0.0067], + [ 0.0704, 0.0346, 0.0775, ..., 0.0177, 0.0127, 0.0243], + ..., + [-0.1485, -0.1284, -0.0956, ..., 0.0113, 0.0454, 0.0003], + [-0.0218, 0.0008, -0.0216, ..., 0.0921, 0.0343, -0.0144], + [-0.0231, -0.0305, -0.0202, ..., -0.1639, -0.1877, -0.1558]], + device='cuda:0'), grad: tensor([[ 3.6216e-04, 6.3956e-05, 7.0810e-05, ..., 1.5664e-04, + 1.2243e-04, 1.6391e-04], + [ 3.8025e-02, 7.5226e-03, 1.1467e-02, ..., 1.0361e-02, + 1.5282e-02, 1.5373e-02], + [-3.8666e-02, -7.6523e-03, -1.1673e-02, ..., -1.0529e-02, + -1.5549e-02, -1.5640e-02], + ..., + [ 1.2600e-04, 4.2140e-05, 8.7678e-05, ..., -1.9401e-05, + 7.9334e-05, 3.8594e-05], + [ 1.6558e-04, 3.2425e-05, 4.8608e-05, ..., 4.7058e-05, + 6.5863e-05, 6.7592e-05], + [ 6.8665e-05, 1.3933e-05, 2.0951e-05, ..., 1.8701e-05, + 2.7299e-05, 2.7537e-05]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0068, -0.2992, 0.0154, -0.2199, -0.0004, 0.3446, 0.1795], + device='cuda:0'), grad: tensor([ 1.2054e-03, 1.0278e-01, -1.0455e-01, -1.5152e-04, 6.4135e-05, + 4.5562e-04, 1.8466e-04], device='cuda:0') +351 +0.008909157412340152 +changing lr +epoch 15, time 487.09, cls_loss 0.1644 cls_loss_mapping 0.5142 cls_loss_causal 1.1892 re_mapping 0.0672 re_causal 0.0669 /// teacc 66.67 lr 0.00876536 +Epoch 17, weight, value: tensor([[-0.0198, -0.0133, -0.0073, ..., -0.0034, 0.1047, 0.1385], + [ 0.0540, 0.0720, 0.0575, ..., 0.0566, 0.0184, -0.0133], + [ 0.0707, 0.0354, 0.0767, ..., 0.0200, 0.0152, 0.0280], + ..., + [-0.1433, -0.1255, -0.0916, ..., 0.0082, 0.0417, -0.0034], + [-0.0192, 0.0005, -0.0220, ..., 0.0945, 0.0362, -0.0125], + [-0.0222, -0.0315, -0.0223, ..., -0.1588, -0.1861, -0.1524]], + device='cuda:0'), grad: tensor([[ 5.0049e-03, 6.1369e-04, 1.1215e-03, ..., 2.3670e-03, + 2.7428e-03, 3.0746e-03], + [-5.3482e-03, -6.4516e-04, -1.1797e-03, ..., -2.5311e-03, + -2.9335e-03, -3.2787e-03], + [ 1.3359e-05, 2.4792e-06, 3.2727e-06, ..., 4.5523e-06, + 4.6417e-06, 5.2229e-06], + ..., + [ 4.4912e-05, 8.4862e-06, 1.0416e-05, ..., 1.3746e-05, + 1.2174e-05, 1.3910e-05], + [ 2.6345e-04, 5.5492e-05, 6.3419e-05, ..., 6.0797e-05, + 4.9591e-05, 5.6654e-05], + [ 1.0535e-05, -3.6955e-05, -2.1502e-05, ..., 8.2850e-05, + 1.2177e-04, 1.2362e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0110, -0.3091, 0.0176, -0.2245, 0.0073, 0.3497, 0.1830], + device='cuda:0'), grad: tensor([ 1.5442e-02, -1.6510e-02, 3.9279e-05, 2.3291e-05, 1.3268e-04, + 7.6675e-04, 9.7513e-05], device='cuda:0') +351 +0.00876535733001806 +changing lr +epoch 16, time 486.79, cls_loss 0.1451 cls_loss_mapping 0.4625 cls_loss_causal 1.1703 re_mapping 0.0665 re_causal 0.0663 /// teacc 75.95 lr 0.00861397 +Epoch 18, weight, value: tensor([[-0.0251, -0.0154, -0.0094, ..., -0.0082, 0.1041, 0.1376], + [ 0.0554, 0.0698, 0.0561, ..., 0.0582, 0.0230, -0.0094], + [ 0.0726, 0.0386, 0.0782, ..., 0.0188, 0.0155, 0.0284], + ..., + [-0.1426, -0.1242, -0.0895, ..., 0.0081, 0.0359, -0.0078], + [-0.0184, 0.0003, -0.0225, ..., 0.0962, 0.0364, -0.0117], + [-0.0208, -0.0312, -0.0229, ..., -0.1566, -0.1876, -0.1535]], + device='cuda:0'), grad: tensor([[-3.5076e-03, -3.7932e-04, -9.4175e-04, ..., -8.6021e-04, + -2.0180e-03, -2.1801e-03], + [ 3.5357e-04, 4.5151e-05, 9.7215e-05, ..., 8.0526e-05, + 1.9038e-04, 2.0647e-04], + [-2.5787e-03, -8.8978e-04, -1.0452e-03, ..., -4.4012e-04, + -3.0732e-04, -2.9087e-04], + ..., + [ 3.7956e-04, 7.5400e-05, 1.2279e-04, ..., 8.1301e-05, + 1.5175e-04, 1.6046e-04], + [ 4.0770e-04, 1.2398e-04, 1.5557e-04, ..., 7.5758e-05, + 7.9870e-05, 8.0645e-05], + [ 3.2368e-03, 4.5490e-04, 9.3269e-04, ..., 7.6056e-04, + 1.6651e-03, 1.7910e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0241, -0.2958, 0.0095, -0.2232, 0.0036, 0.3491, 0.1936], + device='cuda:0'), grad: tensor([-0.0110, 0.0011, -0.0058, 0.0039, 0.0011, 0.0010, 0.0097], + device='cuda:0') +351 +0.008613974319136962 +changing lr +epoch 17, time 490.87, cls_loss 0.1359 cls_loss_mapping 0.4107 cls_loss_causal 1.1271 re_mapping 0.0690 re_causal 0.0689 /// teacc 31.65 lr 0.00845531 +Epoch 19, weight, value: tensor([[-0.0262, -0.0149, -0.0089, ..., -0.0061, 0.1036, 0.1387], + [ 0.0530, 0.0664, 0.0550, ..., 0.0562, 0.0261, -0.0083], + [ 0.0759, 0.0417, 0.0804, ..., 0.0192, 0.0141, 0.0275], + ..., + [-0.1386, -0.1222, -0.0873, ..., 0.0064, 0.0339, -0.0098], + [-0.0211, -0.0006, -0.0238, ..., 0.0952, 0.0354, -0.0127], + [-0.0220, -0.0315, -0.0255, ..., -0.1548, -0.1872, -0.1521]], + device='cuda:0'), grad: tensor([[ 2.1145e-05, 1.6410e-06, 2.7213e-06, ..., 1.2532e-05, + 9.9614e-06, 1.2040e-05], + [ 4.3586e-06, 5.0291e-07, 8.1584e-07, ..., 2.3693e-06, + 1.9353e-06, 2.2948e-06], + [-1.2871e-06, -4.6566e-07, -8.1956e-07, ..., 7.4506e-09, + -4.6380e-07, -6.3702e-07], + ..., + [-3.9369e-05, -3.1348e-06, -5.4277e-06, ..., -2.2262e-05, + -1.7211e-05, -2.0772e-05], + [ 7.0855e-06, 7.5996e-07, 1.2480e-06, ..., 3.7327e-06, + 2.9169e-06, 3.5390e-06], + [ 8.8438e-06, 1.6782e-06, 2.3842e-06, ..., 4.1127e-06, + 3.3025e-06, 4.1015e-06]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0223, -0.2936, 0.0166, -0.2197, 0.0082, 0.3351, 0.1881], + device='cuda:0'), grad: tensor([ 6.7830e-05, 1.3463e-05, -3.2894e-06, 3.5204e-07, -1.2648e-04, + 2.2292e-05, 2.5839e-05], device='cuda:0') +351 +0.008455313244934327 +changing lr +epoch 18, time 489.60, cls_loss 0.0976 cls_loss_mapping 0.3568 cls_loss_causal 1.0910 re_mapping 0.0673 re_causal 0.0673 /// teacc 77.22 lr 0.00828969 +Epoch 20, weight, value: tensor([[-0.0217, -0.0144, -0.0075, ..., -0.0047, 0.1070, 0.1415], + [ 0.0498, 0.0659, 0.0546, ..., 0.0559, 0.0258, -0.0093], + [ 0.0745, 0.0411, 0.0786, ..., 0.0179, 0.0144, 0.0282], + ..., + [-0.1379, -0.1209, -0.0862, ..., 0.0031, 0.0304, -0.0127], + [-0.0152, 0.0002, -0.0226, ..., 0.0979, 0.0375, -0.0094], + [-0.0275, -0.0336, -0.0282, ..., -0.1550, -0.1903, -0.1554]], + device='cuda:0'), grad: tensor([[ 6.8140e-04, 9.6321e-05, 2.1148e-04, ..., 1.2398e-04, + 2.8491e-04, 2.6059e-04], + [ 8.2016e-05, 1.1437e-05, 2.5108e-05, ..., 1.6108e-05, + 3.4273e-05, 3.1531e-05], + [ 7.6714e-03, 1.0843e-03, 2.3823e-03, ..., 1.3924e-03, + 3.2101e-03, 2.9316e-03], + ..., + [-8.6060e-03, -1.2188e-03, -2.6779e-03, ..., -1.5488e-03, + -3.6030e-03, -3.2883e-03], + [ 8.0839e-06, 2.9299e-06, 6.7875e-06, ..., -1.3940e-05, + 3.6377e-06, 3.3900e-07], + [ 8.2791e-05, 1.1660e-05, 2.5585e-05, ..., 1.5527e-05, + 3.4660e-05, 3.1769e-05]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0104, -0.3015, 0.0106, -0.2195, 0.0064, 0.3531, 0.1737], + device='cuda:0'), grad: tensor([ 2.2831e-03, 2.7514e-04, 2.5711e-02, 2.7728e-04, -2.8854e-02, + 2.4706e-05, 2.7752e-04], device='cuda:0') +351 +0.008289693629698565 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 497.06, cls_loss 0.1102 cls_loss_mapping 0.3219 cls_loss_causal 1.0679 re_mapping 0.0690 re_causal 0.0690 /// teacc 85.65 lr 0.00811745 +Epoch 21, weight, value: tensor([[-0.0224, -0.0154, -0.0073, ..., -0.0084, 0.1079, 0.1397], + [ 0.0464, 0.0648, 0.0531, ..., 0.0535, 0.0237, -0.0116], + [ 0.0759, 0.0387, 0.0755, ..., 0.0178, 0.0120, 0.0267], + ..., + [-0.1408, -0.1212, -0.0868, ..., 0.0026, 0.0276, -0.0143], + [-0.0176, -0.0003, -0.0230, ..., 0.0976, 0.0375, -0.0091], + [-0.0218, -0.0310, -0.0260, ..., -0.1495, -0.1864, -0.1501]], + device='cuda:0'), grad: tensor([[ 4.0054e-05, 2.4103e-06, 7.0930e-06, ..., 1.6317e-05, + 9.4175e-06, 1.6287e-05], + [ 2.0787e-06, 9.4995e-08, 3.2783e-07, ..., 8.7358e-07, + 4.6566e-07, 8.5682e-07], + [ 8.4378e-07, 3.7253e-08, 1.4342e-07, ..., 3.5577e-07, + 2.1048e-07, 3.4459e-07], + ..., + [-1.4752e-06, -2.1793e-07, -4.3400e-07, ..., -1.2107e-06, + -1.3504e-06, -1.4361e-06], + [ 1.5274e-06, 6.7055e-08, 2.4587e-07, ..., 6.0722e-07, + 3.1851e-07, 6.0722e-07], + [-4.6432e-05, -3.0883e-06, -8.5682e-06, ..., -1.7300e-05, + -9.3505e-06, -1.7017e-05]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0157, -0.3083, 0.0252, -0.2177, -0.0058, 0.3422, 0.1923], + device='cuda:0'), grad: tensor([ 1.3244e-04, 6.9551e-06, 2.8443e-06, 9.5814e-06, -4.6417e-06, + 5.1148e-06, -1.5259e-04], device='cuda:0') +351 +0.00811744900929367 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 492.87, cls_loss 0.0760 cls_loss_mapping 0.2840 cls_loss_causal 1.0343 re_mapping 0.0687 re_causal 0.0688 /// teacc 88.61 lr 0.00793893 +Epoch 22, weight, value: tensor([[-2.5705e-02, -1.5634e-02, -7.1052e-03, ..., -1.0747e-02, + 1.0806e-01, 1.3788e-01], + [ 5.1844e-02, 6.5237e-02, 5.4339e-02, ..., 5.6280e-02, + 2.6188e-02, -9.1537e-03], + [ 6.8620e-02, 3.6508e-02, 7.1406e-02, ..., 1.6478e-02, + 9.5231e-03, 2.5444e-02], + ..., + [-1.3768e-01, -1.1991e-01, -8.5010e-02, ..., 2.3305e-03, + 2.6403e-02, -1.5035e-02], + [-1.5751e-02, -7.5948e-05, -2.3335e-02, ..., 9.7384e-02, + 3.7156e-02, -9.2214e-03], + [-2.1356e-02, -3.0200e-02, -2.5406e-02, ..., -1.4869e-01, + -1.8599e-01, -1.4877e-01]], device='cuda:0'), grad: tensor([[ 2.1867e-06, 2.5891e-07, 4.9360e-07, ..., 2.1048e-07, + 5.0291e-08, 2.0303e-07], + [ 5.4315e-06, 6.2026e-07, 1.2163e-06, ..., 5.7369e-07, + 2.3656e-07, 6.0908e-07], + [ 5.8375e-06, 2.1812e-06, 3.0119e-06, ..., 1.9372e-06, + 2.6189e-06, 2.4084e-06], + ..., + [ 5.4896e-05, 2.3603e-05, 3.2067e-05, ..., 1.9833e-05, + 2.8804e-05, 2.5585e-05], + [ 2.3600e-06, 1.9185e-07, 4.9174e-07, ..., -3.5390e-08, + -1.6019e-07, 3.3528e-08], + [-1.3016e-05, -1.3337e-06, -2.7455e-06, ..., -1.2256e-06, + -3.1479e-07, -1.2722e-06]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0246, -0.2882, 0.0090, -0.2145, -0.0027, 0.3421, 0.1908], + device='cuda:0'), grad: tensor([ 7.1079e-06, 1.7703e-05, 1.3880e-05, -1.2386e-04, 1.2010e-04, + 7.9721e-06, -4.2945e-05], device='cuda:0') +351 +0.007938926261462368 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 495.37, cls_loss 0.0555 cls_loss_mapping 0.2618 cls_loss_causal 1.0042 re_mapping 0.0691 re_causal 0.0694 /// teacc 89.03 lr 0.00775448 +Epoch 23, weight, value: tensor([[-0.0250, -0.0152, -0.0076, ..., -0.0115, 0.1060, 0.1355], + [ 0.0525, 0.0647, 0.0542, ..., 0.0553, 0.0270, -0.0090], + [ 0.0669, 0.0349, 0.0696, ..., 0.0153, 0.0090, 0.0249], + ..., + [-0.1370, -0.1186, -0.0835, ..., 0.0030, 0.0262, -0.0145], + [-0.0132, -0.0003, -0.0236, ..., 0.0985, 0.0381, -0.0077], + [-0.0240, -0.0302, -0.0255, ..., -0.1479, -0.1849, -0.1475]], + device='cuda:0'), grad: tensor([[-5.6386e-05, -6.1654e-06, -1.6123e-05, ..., -5.9873e-05, + -7.1406e-05, -7.0572e-05], + [ 2.3186e-04, 3.1978e-05, 7.4446e-05, ..., 1.0973e-04, + 9.6023e-05, 1.0985e-04], + [-6.2808e-06, -3.4869e-06, -7.7672e-07, ..., 2.3004e-06, + 1.0543e-06, 1.2200e-06], + ..., + [ 5.5641e-05, 1.2450e-05, 1.6242e-05, ..., 1.5393e-05, + 9.5591e-06, 1.1526e-05], + [ 5.9242e-03, 7.5436e-04, 2.0351e-03, ..., 1.7452e-03, + 8.3017e-04, 1.3838e-03], + [-6.1455e-03, -7.8821e-04, -2.1076e-03, ..., -1.8110e-03, + -8.6451e-04, -1.4343e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0237, -0.2811, 0.0093, -0.2139, -0.0032, 0.3464, 0.1781], + device='cuda:0'), grad: tensor([-1.8227e-04, 7.1716e-04, -3.7491e-05, -1.1344e-06, 1.5235e-04, + 1.8356e-02, -1.8997e-02], device='cuda:0') +351 +0.007754484907260515 +changing lr +epoch 22, time 489.20, cls_loss 0.0471 cls_loss_mapping 0.2469 cls_loss_causal 0.9975 re_mapping 0.0680 re_causal 0.0683 /// teacc 82.70 lr 0.00756450 +Epoch 24, weight, value: tensor([[-0.0221, -0.0143, -0.0068, ..., -0.0097, 0.1085, 0.1372], + [ 0.0452, 0.0622, 0.0522, ..., 0.0519, 0.0224, -0.0141], + [ 0.0691, 0.0355, 0.0695, ..., 0.0148, 0.0087, 0.0250], + ..., + [-0.1330, -0.1165, -0.0813, ..., 0.0051, 0.0275, -0.0122], + [-0.0186, -0.0018, -0.0254, ..., 0.0961, 0.0369, -0.0089], + [-0.0209, -0.0295, -0.0248, ..., -0.1462, -0.1833, -0.1452]], + device='cuda:0'), grad: tensor([[ 1.4174e-04, 1.3284e-05, 2.9057e-05, ..., 1.0335e-04, + 7.4446e-05, 7.3075e-05], + [ 1.5712e-04, 1.6555e-05, 2.9877e-05, ..., 8.7261e-05, + 7.5936e-05, 7.2062e-05], + [ 1.4429e-03, 1.5485e-04, 2.7990e-04, ..., 8.6403e-04, + 7.5531e-04, 7.2527e-04], + ..., + [ 1.9875e-03, 2.1029e-04, 3.7670e-04, ..., 1.0900e-03, + 9.5558e-04, 9.0551e-04], + [-4.4212e-03, -4.6921e-04, -8.4877e-04, ..., -2.5463e-03, + -2.2163e-03, -2.1152e-03], + [ 4.1676e-04, 4.4823e-05, 7.9811e-05, ..., 2.3985e-04, + 2.1195e-04, 2.0242e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0165, -0.2987, 0.0153, -0.2091, 0.0057, 0.3272, 0.1878], + device='cuda:0'), grad: tensor([ 0.0006, 0.0006, 0.0057, 0.0011, 0.0080, -0.0178, 0.0017], + device='cuda:0') +351 +0.007564496387029534 +changing lr +epoch 23, time 487.18, cls_loss 0.0468 cls_loss_mapping 0.2215 cls_loss_causal 0.9253 re_mapping 0.0658 re_causal 0.0661 /// teacc 74.68 lr 0.00736934 +Epoch 25, weight, value: tensor([[-0.0234, -0.0146, -0.0073, ..., -0.0096, 0.1084, 0.1368], + [ 0.0481, 0.0616, 0.0525, ..., 0.0520, 0.0238, -0.0126], + [ 0.0674, 0.0362, 0.0695, ..., 0.0141, 0.0085, 0.0248], + ..., + [-0.1308, -0.1160, -0.0815, ..., 0.0056, 0.0258, -0.0134], + [-0.0168, -0.0017, -0.0253, ..., 0.0961, 0.0370, -0.0083], + [-0.0243, -0.0299, -0.0251, ..., -0.1461, -0.1824, -0.1447]], + device='cuda:0'), grad: tensor([[ 6.6817e-05, 1.1541e-05, 1.8194e-05, ..., 4.5657e-05, + 4.5270e-05, 4.9025e-05], + [ 6.1035e-05, 1.0975e-05, 1.7136e-05, ..., 4.0889e-05, + 4.0948e-05, 4.4376e-05], + [-1.2368e-05, -4.3958e-06, -4.6641e-06, ..., -1.9595e-06, + -1.4678e-06, -2.7642e-06], + ..., + [-5.5218e-04, -9.5844e-05, -1.5271e-04, ..., -3.7932e-04, + -3.8004e-04, -4.1032e-04], + [ 3.8296e-05, 7.4282e-06, 1.0990e-05, ..., 2.3633e-05, + 2.3440e-05, 2.5749e-05], + [ 4.2021e-06, 1.1772e-06, 1.3858e-06, ..., 1.8775e-06, + 1.8701e-06, 2.2650e-06]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0207, -0.2837, 0.0064, -0.2068, 0.0094, 0.3279, 0.1792], + device='cuda:0'), grad: tensor([ 1.8096e-04, 1.6356e-04, -3.0279e-05, 1.0567e-03, -1.4839e-03, + 1.0246e-04, 1.0207e-05], device='cuda:0') +351 +0.007369343312364995 +changing lr +epoch 24, time 489.43, cls_loss 0.0561 cls_loss_mapping 0.2156 cls_loss_causal 0.9517 re_mapping 0.0632 re_causal 0.0636 /// teacc 79.75 lr 0.00716942 +Epoch 26, weight, value: tensor([[-0.0244, -0.0149, -0.0076, ..., -0.0097, 0.1082, 0.1363], + [ 0.0462, 0.0608, 0.0517, ..., 0.0509, 0.0229, -0.0135], + [ 0.0703, 0.0368, 0.0696, ..., 0.0148, 0.0095, 0.0267], + ..., + [-0.1327, -0.1161, -0.0814, ..., 0.0048, 0.0241, -0.0150], + [-0.0165, -0.0017, -0.0252, ..., 0.0958, 0.0370, -0.0079], + [-0.0249, -0.0295, -0.0252, ..., -0.1451, -0.1814, -0.1440]], + device='cuda:0'), grad: tensor([[ 5.2631e-05, 7.4133e-06, 9.4920e-06, ..., 2.3797e-05, + 2.7403e-05, 2.1726e-05], + [ 1.2308e-05, 2.1905e-06, 8.9407e-08, ..., 5.4166e-06, + 1.6391e-06, -1.0058e-06], + [ 6.5804e-04, 8.4043e-05, 1.1456e-04, ..., 1.8108e-04, + 1.8704e-04, 1.4186e-04], + ..., + [ 1.3752e-03, 2.1374e-04, 1.9753e-04, ..., 6.8903e-04, + 6.8665e-04, 4.8661e-04], + [-1.8311e-03, -2.8539e-04, -2.6178e-04, ..., -9.2459e-04, + -9.2125e-04, -6.5231e-04], + [-4.1652e-04, -4.5180e-05, -8.1360e-05, ..., -4.9084e-05, + -5.5760e-05, -4.9591e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0235, -0.2851, 0.0169, -0.1963, 0.0014, 0.3240, 0.1741], + device='cuda:0'), grad: tensor([ 1.5664e-04, 3.3677e-05, 2.0599e-03, 4.2963e-04, 3.9406e-03, + -5.2376e-03, -1.3800e-03], device='cuda:0') +351 +0.0071694186955877925 +changing lr +epoch 25, time 486.77, cls_loss 0.0445 cls_loss_mapping 0.2002 cls_loss_causal 0.9541 re_mapping 0.0637 re_causal 0.0641 /// teacc 86.08 lr 0.00696513 +Epoch 27, weight, value: tensor([[-0.0224, -0.0148, -0.0069, ..., -0.0090, 0.1102, 0.1378], + [ 0.0452, 0.0599, 0.0507, ..., 0.0500, 0.0225, -0.0140], + [ 0.0656, 0.0353, 0.0671, ..., 0.0134, 0.0080, 0.0254], + ..., + [-0.1328, -0.1142, -0.0798, ..., 0.0032, 0.0210, -0.0175], + [-0.0156, -0.0017, -0.0253, ..., 0.0967, 0.0379, -0.0066], + [-0.0209, -0.0287, -0.0242, ..., -0.1424, -0.1791, -0.1420]], + device='cuda:0'), grad: tensor([[ 2.1248e-03, 3.3641e-04, 6.0511e-04, ..., 6.1703e-04, + 1.2503e-03, 1.2083e-03], + [-2.7657e-03, -4.6110e-04, -7.8583e-04, ..., -7.3004e-04, + -1.3189e-03, -1.2741e-03], + [ 3.1161e-04, 5.1141e-05, 9.0480e-05, ..., 1.1814e-04, + 1.1945e-04, 1.3459e-04], + ..., + [-6.3467e-04, -1.0312e-04, -1.8442e-04, ..., -2.4652e-04, + -2.5082e-04, -2.8372e-04], + [ 2.0123e-04, 3.3110e-05, 5.8502e-05, ..., 7.1347e-05, + 7.2539e-05, 8.2135e-05], + [ 7.2813e-04, 1.3781e-04, 2.0635e-04, ..., 1.5700e-04, + 1.1438e-04, 1.1861e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0163, -0.2837, 0.0050, -0.1952, -0.0045, 0.3244, 0.1816], + device='cuda:0'), grad: tensor([ 0.0065, -0.0087, 0.0010, 0.0001, -0.0021, 0.0007, 0.0025], + device='cuda:0') +351 +0.0069651251582696205 +changing lr +epoch 26, time 487.73, cls_loss 0.0368 cls_loss_mapping 0.1828 cls_loss_causal 0.8843 re_mapping 0.0628 re_causal 0.0632 /// teacc 86.92 lr 0.00675687 +Epoch 28, weight, value: tensor([[-0.0236, -0.0151, -0.0075, ..., -0.0090, 0.1096, 0.1365], + [ 0.0429, 0.0591, 0.0491, ..., 0.0491, 0.0215, -0.0149], + [ 0.0693, 0.0359, 0.0674, ..., 0.0138, 0.0074, 0.0253], + ..., + [-0.1272, -0.1129, -0.0777, ..., 0.0041, 0.0222, -0.0156], + [-0.0171, -0.0017, -0.0255, ..., 0.0952, 0.0373, -0.0070], + [-0.0240, -0.0291, -0.0241, ..., -0.1416, -0.1779, -0.1408]], + device='cuda:0'), grad: tensor([[ 6.0856e-05, 2.1964e-05, 2.7269e-05, ..., 6.5342e-06, + -1.4685e-05, -5.9530e-06], + [-6.1631e-05, -9.1717e-06, -6.6832e-06, ..., -4.4763e-05, + -1.7956e-05, -2.1353e-05], + [ 9.1195e-05, 2.1666e-05, 5.9724e-05, ..., 3.2812e-05, + 1.0960e-05, 2.4229e-05], + ..., + [-7.8297e-04, -1.6737e-04, -3.1304e-04, ..., -1.6844e-04, + -1.1462e-04, -1.4091e-04], + [ 7.5996e-05, 1.4573e-05, 2.2784e-05, ..., 1.7464e-05, + 1.4804e-05, 1.4916e-05], + [ 2.5916e-04, 5.2959e-05, 8.9407e-05, ..., 5.7548e-05, + 4.3720e-05, 4.9055e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0193, -0.2885, 0.0172, -0.1947, 0.0115, 0.3121, 0.1729], + device='cuda:0'), grad: tensor([ 0.0004, -0.0001, 0.0008, 0.0014, -0.0038, 0.0003, 0.0011], + device='cuda:0') +351 +0.006756874120406716 +changing lr +epoch 27, time 487.15, cls_loss 0.0301 cls_loss_mapping 0.1562 cls_loss_causal 0.8664 re_mapping 0.0620 re_causal 0.0625 /// teacc 86.50 lr 0.00654508 +Epoch 29, weight, value: tensor([[-0.0220, -0.0148, -0.0074, ..., -0.0088, 0.1090, 0.1355], + [ 0.0451, 0.0592, 0.0495, ..., 0.0491, 0.0212, -0.0149], + [ 0.0643, 0.0347, 0.0653, ..., 0.0122, 0.0056, 0.0231], + ..., + [-0.1302, -0.1122, -0.0772, ..., 0.0024, 0.0208, -0.0162], + [-0.0137, -0.0015, -0.0252, ..., 0.0964, 0.0387, -0.0057], + [-0.0215, -0.0287, -0.0234, ..., -0.1391, -0.1752, -0.1381]], + device='cuda:0'), grad: tensor([[-3.0112e-04, -3.0011e-05, -8.4877e-05, ..., -1.0002e-04, + -1.4138e-04, -1.4341e-04], + [-2.8610e-05, -2.1160e-06, -7.0333e-06, ..., -5.9530e-06, + -9.5963e-06, -8.3521e-06], + [-8.0168e-05, -2.8774e-05, -3.0190e-05, ..., -7.9945e-06, + -3.1590e-06, -7.7263e-06], + ..., + [ 1.4508e-04, 1.4298e-05, 4.1276e-05, ..., 4.0680e-05, + 6.0260e-05, 6.1095e-05], + [ 4.0799e-05, 8.3372e-06, 1.2673e-05, ..., 9.0152e-06, + 1.1228e-05, 1.1653e-05], + [ 1.9717e-04, 3.2246e-05, 5.9605e-05, ..., 5.6714e-05, + 7.3969e-05, 7.7367e-05]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0145, -0.2783, 0.0044, -0.1967, -0.0009, 0.3187, 0.1783], + device='cuda:0'), grad: tensor([-1.0118e-03, -8.4460e-05, -1.9419e-04, 7.6890e-05, 4.8494e-04, + 1.1617e-04, 6.1226e-04], device='cuda:0') +351 +0.00654508497187474 +changing lr +epoch 28, time 488.45, cls_loss 0.0259 cls_loss_mapping 0.1594 cls_loss_causal 0.8755 re_mapping 0.0618 re_causal 0.0624 /// teacc 83.54 lr 0.00633018 +Epoch 30, weight, value: tensor([[-0.0226, -0.0141, -0.0070, ..., -0.0097, 0.1078, 0.1339], + [ 0.0451, 0.0585, 0.0489, ..., 0.0486, 0.0208, -0.0149], + [ 0.0660, 0.0352, 0.0655, ..., 0.0120, 0.0055, 0.0230], + ..., + [-0.1285, -0.1112, -0.0761, ..., 0.0028, 0.0212, -0.0155], + [-0.0149, -0.0017, -0.0254, ..., 0.0952, 0.0380, -0.0061], + [-0.0242, -0.0300, -0.0249, ..., -0.1378, -0.1735, -0.1366]], + device='cuda:0'), grad: tensor([[-2.3043e-04, -2.5898e-05, -3.6359e-05, ..., -6.9618e-05, + -6.9380e-05, -6.5029e-05], + [ 1.2040e-05, 9.2015e-07, 2.3283e-06, ..., 4.3623e-06, + 7.4543e-06, 6.0536e-06], + [ 6.5751e-06, 5.9977e-07, 1.3076e-06, ..., 2.2091e-06, + 3.5539e-06, 2.9653e-06], + ..., + [-2.4462e-04, -1.8463e-05, -5.0098e-05, ..., -1.2875e-04, + -9.0241e-05, -9.0182e-05], + [ 1.3925e-05, 1.1101e-06, 2.8498e-06, ..., 6.5938e-06, + 5.9046e-06, 5.5246e-06], + [ 2.1410e-04, 2.4378e-05, 3.3259e-05, ..., 6.6340e-05, + 5.7548e-05, 5.6148e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0197, -0.2738, 0.0092, -0.1897, 0.0017, 0.3110, 0.1723], + device='cuda:0'), grad: tensor([-7.7391e-04, 4.0770e-05, 2.1964e-05, 8.1396e-04, -8.7309e-04, + 4.8757e-05, 7.2145e-04], device='cuda:0') +351 +0.006330184227833378 +changing lr +epoch 29, time 490.75, cls_loss 0.0292 cls_loss_mapping 0.1531 cls_loss_causal 0.8828 re_mapping 0.0614 re_causal 0.0621 /// teacc 86.92 lr 0.00611260 +Epoch 31, weight, value: tensor([[-0.0218, -0.0140, -0.0068, ..., -0.0089, 0.1091, 0.1349], + [ 0.0433, 0.0577, 0.0479, ..., 0.0471, 0.0192, -0.0167], + [ 0.0632, 0.0351, 0.0644, ..., 0.0113, 0.0043, 0.0219], + ..., + [-0.1234, -0.1095, -0.0737, ..., 0.0036, 0.0218, -0.0143], + [-0.0140, -0.0017, -0.0254, ..., 0.0953, 0.0381, -0.0057], + [-0.0247, -0.0296, -0.0247, ..., -0.1371, -0.1730, -0.1360]], + device='cuda:0'), grad: tensor([[ 1.1892e-03, 1.4639e-04, 3.5691e-04, ..., 1.6296e-04, + 1.0908e-04, 1.8024e-04], + [ 1.6034e-04, 2.6032e-05, 4.6492e-05, ..., 1.7181e-05, + 9.0823e-06, 1.5661e-05], + [ 2.4438e-04, 3.5912e-05, 7.3314e-05, ..., 3.0607e-05, + 1.9833e-05, 3.1680e-05], + ..., + [ 1.6680e-03, 2.5249e-04, 4.9782e-04, ..., 1.9598e-04, + 1.2219e-04, 1.9646e-04], + [ 1.1700e-04, 1.9327e-05, 3.4690e-05, ..., 1.2390e-05, + 7.2531e-06, 1.1504e-05], + [-3.3855e-03, -4.7874e-04, -1.0099e-03, ..., -4.1747e-04, + -2.6464e-04, -4.3321e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0185, -0.2735, 0.0005, -0.1896, 0.0133, 0.3104, 0.1683], + device='cuda:0'), grad: tensor([ 3.7479e-03, 4.8637e-04, 7.4959e-04, 2.9132e-05, 5.0888e-03, + 3.5143e-04, -1.0460e-02], device='cuda:0') +351 +0.006112604669781575 +changing lr +epoch 30, time 493.20, cls_loss 0.0291 cls_loss_mapping 0.1489 cls_loss_causal 0.8701 re_mapping 0.0617 re_causal 0.0625 /// teacc 74.68 lr 0.00589278 +Epoch 32, weight, value: tensor([[-0.0220, -0.0139, -0.0067, ..., -0.0089, 0.1086, 0.1337], + [ 0.0467, 0.0574, 0.0485, ..., 0.0477, 0.0197, -0.0158], + [ 0.0662, 0.0358, 0.0649, ..., 0.0117, 0.0051, 0.0228], + ..., + [-0.1264, -0.1089, -0.0743, ..., 0.0026, 0.0201, -0.0159], + [-0.0151, -0.0019, -0.0255, ..., 0.0943, 0.0375, -0.0061], + [-0.0254, -0.0300, -0.0251, ..., -0.1358, -0.1713, -0.1342]], + device='cuda:0'), grad: tensor([[-1.3389e-05, -1.6242e-06, -1.9707e-06, ..., -2.7418e-06, + -9.4846e-06, -9.3952e-06], + [-1.7524e-05, -2.8014e-06, -3.5129e-06, ..., -1.0431e-06, + 4.0978e-07, 1.3411e-07], + [ 6.8881e-06, 1.2890e-06, 1.4454e-06, ..., 2.3954e-06, + 2.4922e-06, 2.6710e-06], + ..., + [ 1.1548e-05, 1.7248e-06, 1.8440e-06, ..., 2.1644e-06, + 5.9679e-06, 5.9269e-06], + [-3.1888e-06, -1.3709e-06, -8.5682e-07, ..., -4.1500e-06, + -3.0473e-06, -3.2410e-06], + [ 1.1280e-05, 1.9930e-06, 2.2314e-06, ..., 2.0228e-06, + 1.6540e-06, 1.8477e-06]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0184, -0.2584, 0.0100, -0.1898, -0.0010, 0.3036, 0.1646], + device='cuda:0'), grad: tensor([-4.1336e-05, -5.4091e-05, 2.0459e-05, 1.2487e-05, 3.4273e-05, + -5.7705e-06, 3.3677e-05], device='cuda:0') +351 +0.005892784473993186 +changing lr +epoch 31, time 488.22, cls_loss 0.0271 cls_loss_mapping 0.1293 cls_loss_causal 0.8446 re_mapping 0.0604 re_causal 0.0613 /// teacc 82.70 lr 0.00567117 +Epoch 33, weight, value: tensor([[-0.0197, -0.0135, -0.0061, ..., -0.0079, 0.1092, 0.1343], + [ 0.0433, 0.0561, 0.0470, ..., 0.0464, 0.0185, -0.0169], + [ 0.0662, 0.0359, 0.0648, ..., 0.0113, 0.0046, 0.0225], + ..., + [-0.1251, -0.1077, -0.0734, ..., 0.0020, 0.0190, -0.0164], + [-0.0145, -0.0019, -0.0255, ..., 0.0947, 0.0381, -0.0053], + [-0.0254, -0.0299, -0.0248, ..., -0.1351, -0.1700, -0.1333]], + device='cuda:0'), grad: tensor([[ 7.5288e-06, 2.6189e-06, 3.0771e-06, ..., 2.4885e-06, + 1.8999e-06, 2.2314e-06], + [-1.2450e-05, -1.6689e-06, -3.1479e-06, ..., -3.0622e-06, + -2.6710e-06, -4.0866e-06], + [ 1.7472e-06, 7.2643e-07, 7.4133e-07, ..., 7.0408e-07, + 4.8056e-07, 4.6566e-07], + ..., + [ 4.3586e-07, 2.6822e-07, 2.6822e-07, ..., -1.4901e-08, + 6.3330e-08, 4.8429e-08], + [ 2.8573e-06, 9.9838e-07, 1.1586e-06, ..., 9.1642e-07, + 7.1153e-07, 8.3074e-07], + [ 1.7941e-05, 4.2692e-06, 5.8673e-06, ..., 5.0589e-06, + 4.1910e-06, 5.6028e-06]], device='cuda:0') +Epoch 33, bias, value: tensor([-1.1186e-02, -2.6577e-01, 1.0681e-02, -1.8894e-01, -2.5201e-04, + 3.0202e-01, 1.6403e-01], device='cuda:0'), grad: tensor([ 1.7270e-05, -3.9309e-05, 3.2783e-06, -3.6538e-05, 1.3039e-07, + 6.4075e-06, 4.8697e-05], device='cuda:0') +351 +0.00567116632908828 +changing lr +epoch 32, time 494.20, cls_loss 0.0140 cls_loss_mapping 0.1146 cls_loss_causal 0.8057 re_mapping 0.0611 re_causal 0.0621 /// teacc 83.12 lr 0.00544820 +Epoch 34, weight, value: tensor([[-0.0195, -0.0134, -0.0060, ..., -0.0085, 0.1079, 0.1328], + [ 0.0439, 0.0556, 0.0467, ..., 0.0464, 0.0186, -0.0164], + [ 0.0643, 0.0355, 0.0638, ..., 0.0109, 0.0038, 0.0215], + ..., + [-0.1249, -0.1069, -0.0729, ..., 0.0022, 0.0192, -0.0160], + [-0.0141, -0.0020, -0.0255, ..., 0.0944, 0.0382, -0.0049], + [-0.0244, -0.0296, -0.0242, ..., -0.1340, -0.1684, -0.1319]], + device='cuda:0'), grad: tensor([[ 1.2890e-06, 1.9372e-07, 2.7195e-07, ..., 2.6450e-07, + 2.2352e-08, 1.7509e-07], + [ 9.8720e-07, 5.5879e-08, 1.4529e-07, ..., 1.4156e-07, + 2.5705e-07, 3.3528e-07], + [ 1.4901e-07, -1.4901e-08, -7.4506e-09, ..., 7.4506e-08, + 5.9605e-08, 6.7055e-08], + ..., + [ 1.3672e-06, 4.4703e-07, 5.2899e-07, ..., 6.1840e-07, + 5.4017e-07, 6.1095e-07], + [ 7.1228e-06, 6.7800e-07, 1.1809e-06, ..., 1.3076e-06, + 9.6112e-07, 1.6615e-06], + [-9.7528e-06, -8.9779e-07, -1.5907e-06, ..., -1.7695e-06, + -1.2815e-06, -2.2501e-06]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0111, -0.2602, 0.0059, -0.1873, -0.0030, 0.3006, 0.1656], + device='cuda:0'), grad: tensor([ 4.3325e-06, 3.6284e-06, 6.5938e-07, -2.7381e-06, 3.4980e-06, + 2.5198e-05, -3.4690e-05], device='cuda:0') +351 +0.00544819654451717 +changing lr +epoch 33, time 487.32, cls_loss 0.0128 cls_loss_mapping 0.1002 cls_loss_causal 0.7890 re_mapping 0.0586 re_causal 0.0596 /// teacc 80.17 lr 0.00522432 +Epoch 35, weight, value: tensor([[-0.0213, -0.0138, -0.0065, ..., -0.0094, 0.1059, 0.1304], + [ 0.0420, 0.0549, 0.0458, ..., 0.0458, 0.0181, -0.0168], + [ 0.0661, 0.0362, 0.0639, ..., 0.0114, 0.0046, 0.0225], + ..., + [-0.1238, -0.1061, -0.0723, ..., 0.0023, 0.0190, -0.0159], + [-0.0145, -0.0020, -0.0256, ..., 0.0938, 0.0380, -0.0048], + [-0.0227, -0.0296, -0.0236, ..., -0.1327, -0.1667, -0.1304]], + device='cuda:0'), grad: tensor([[ 1.6081e-04, 1.3858e-05, 3.4720e-05, ..., 4.6670e-05, + 2.1726e-05, 4.5151e-05], + [ 1.0207e-05, 5.5134e-07, 2.1979e-06, ..., 2.7604e-06, + 1.0394e-06, 2.5928e-06], + [ 2.6911e-05, 1.4007e-06, 5.7444e-06, ..., 7.3723e-06, + 2.7679e-06, 6.8769e-06], + ..., + [-6.6936e-05, -9.1121e-06, -1.4573e-05, ..., -2.1294e-05, + -1.2442e-05, -2.1517e-05], + [ 1.9655e-05, 1.0766e-06, 4.2394e-06, ..., 5.3495e-06, + 2.0340e-06, 5.0180e-06], + [-1.6916e-04, -8.7768e-06, -3.6329e-05, ..., -4.5896e-05, + -1.7032e-05, -4.2856e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0173, -0.2628, 0.0116, -0.1854, -0.0026, 0.2957, 0.1711], + device='cuda:0'), grad: tensor([ 5.9175e-04, 4.2528e-05, 1.1235e-04, 7.7784e-05, -1.9562e-04, + 8.1778e-05, -7.1049e-04], device='cuda:0') +351 +0.005224324151752577 +changing lr +epoch 34, time 489.98, cls_loss 0.0186 cls_loss_mapping 0.1047 cls_loss_causal 0.7736 re_mapping 0.0585 re_causal 0.0595 /// teacc 83.54 lr 0.00500000 +Epoch 36, weight, value: tensor([[-0.0216, -0.0141, -0.0069, ..., -0.0092, 0.1056, 0.1298], + [ 0.0414, 0.0545, 0.0456, ..., 0.0451, 0.0177, -0.0170], + [ 0.0655, 0.0358, 0.0632, ..., 0.0109, 0.0041, 0.0221], + ..., + [-0.1214, -0.1049, -0.0712, ..., 0.0022, 0.0184, -0.0160], + [-0.0141, -0.0021, -0.0256, ..., 0.0938, 0.0382, -0.0044], + [-0.0250, -0.0298, -0.0240, ..., -0.1324, -0.1659, -0.1300]], + device='cuda:0'), grad: tensor([[ 3.3408e-05, 6.4932e-06, 9.7007e-06, ..., 1.7822e-05, + 2.0593e-05, 2.3276e-05], + [-1.0021e-06, -6.2026e-07, -3.3528e-07, ..., 6.7614e-07, + 3.2391e-06, 3.2559e-06], + [ 2.3395e-05, 4.2990e-06, 7.3910e-06, ..., 1.0505e-05, + 1.4037e-05, 1.5646e-05], + ..., + [-7.9870e-05, -1.5073e-05, -2.4453e-05, ..., -3.7611e-05, + -5.0575e-05, -5.6446e-05], + [-1.4920e-06, -2.2538e-07, -2.7567e-07, ..., -1.7807e-06, + -7.6927e-07, -5.2899e-07], + [ 1.7151e-05, 3.1851e-06, 5.3979e-06, ..., 7.2606e-06, + 9.5218e-06, 1.0468e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0171, -0.2617, 0.0107, -0.1806, 0.0014, 0.2941, 0.1635], + device='cuda:0'), grad: tensor([ 1.2863e-04, 1.5218e-06, 8.2374e-05, 2.6941e-05, -2.9230e-04, + -4.2953e-06, 5.7399e-05], device='cuda:0') +351 +0.005000000000000003 +changing lr +epoch 35, time 489.59, cls_loss 0.0186 cls_loss_mapping 0.0992 cls_loss_causal 0.7708 re_mapping 0.0567 re_causal 0.0578 /// teacc 86.08 lr 0.00477568 +Epoch 37, weight, value: tensor([[-0.0225, -0.0142, -0.0072, ..., -0.0097, 0.1046, 0.1286], + [ 0.0419, 0.0541, 0.0454, ..., 0.0452, 0.0181, -0.0164], + [ 0.0645, 0.0356, 0.0626, ..., 0.0108, 0.0036, 0.0217], + ..., + [-0.1206, -0.1042, -0.0707, ..., 0.0022, 0.0184, -0.0158], + [-0.0141, -0.0022, -0.0255, ..., 0.0934, 0.0381, -0.0043], + [-0.0259, -0.0298, -0.0239, ..., -0.1321, -0.1653, -0.1296]], + device='cuda:0'), grad: tensor([[ 7.0870e-05, 1.1154e-05, 2.2084e-05, ..., 2.9728e-05, + 2.9624e-05, 3.3706e-05], + [-1.6823e-05, -2.8890e-06, -5.2638e-06, ..., -1.3612e-05, + -1.8016e-05, -1.7941e-05], + [-8.9824e-05, -2.4319e-05, -2.8491e-05, ..., -6.5714e-06, + -1.2629e-05, -1.4260e-05], + ..., + [ 4.0799e-05, 7.8306e-06, 1.2830e-05, ..., 1.1452e-05, + 1.1101e-05, 1.3262e-05], + [ 4.5776e-05, 7.6741e-06, 1.4372e-05, ..., 1.4342e-05, + 1.1854e-05, 1.5013e-05], + [-1.1605e-04, -1.4335e-05, -3.6120e-05, ..., -4.6641e-05, + -3.4481e-05, -4.4852e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0201, -0.2563, 0.0100, -0.1743, 0.0009, 0.2915, 0.1585], + device='cuda:0'), grad: tensor([ 2.4629e-04, -4.9651e-05, -2.0874e-04, 1.8287e-04, 1.3006e-04, + 1.5986e-04, -4.6062e-04], device='cuda:0') +351 +0.004775675848247429 +changing lr +epoch 36, time 488.84, cls_loss 0.0166 cls_loss_mapping 0.0919 cls_loss_causal 0.7734 re_mapping 0.0554 re_causal 0.0565 /// teacc 65.40 lr 0.00455180 +Epoch 38, weight, value: tensor([[-0.0223, -0.0141, -0.0071, ..., -0.0098, 0.1041, 0.1278], + [ 0.0414, 0.0536, 0.0449, ..., 0.0448, 0.0178, -0.0165], + [ 0.0643, 0.0358, 0.0624, ..., 0.0106, 0.0035, 0.0215], + ..., + [-0.1181, -0.1033, -0.0697, ..., 0.0027, 0.0188, -0.0149], + [-0.0145, -0.0024, -0.0256, ..., 0.0929, 0.0379, -0.0042], + [-0.0246, -0.0293, -0.0234, ..., -0.1310, -0.1640, -0.1285]], + device='cuda:0'), grad: tensor([[ 1.3316e-04, 1.7837e-05, 2.2829e-05, ..., 1.4164e-05, + 1.2748e-05, 2.1756e-05], + [ 5.2564e-06, 4.0606e-07, 8.4192e-07, ..., 5.1595e-07, + 6.2212e-07, 8.6613e-07], + [ 3.3081e-05, 2.6152e-06, 5.2229e-06, ..., 3.0957e-06, + 3.8035e-06, 5.3719e-06], + ..., + [-5.8189e-06, -4.7125e-07, -1.5218e-06, ..., -1.4678e-06, + -1.4100e-06, -1.4715e-06], + [ 2.6207e-06, 1.7136e-07, 3.7812e-07, ..., 3.9116e-08, + 2.0675e-07, 3.4086e-07], + [-1.6868e-04, -2.0489e-05, -2.7731e-05, ..., -1.6347e-05, + -1.6004e-05, -2.6926e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0192, -0.2548, 0.0095, -0.1794, 0.0065, 0.2877, 0.1599], + device='cuda:0'), grad: tensor([ 4.4608e-04, 1.7881e-05, 1.1235e-04, 1.7788e-06, -1.9386e-05, + 9.0301e-06, -5.6791e-04], device='cuda:0') +351 +0.004551803455482836 +changing lr +epoch 37, time 487.50, cls_loss 0.0129 cls_loss_mapping 0.0849 cls_loss_causal 0.7819 re_mapping 0.0550 re_causal 0.0562 /// teacc 89.03 lr 0.00432883 +Epoch 39, weight, value: tensor([[-0.0228, -0.0142, -0.0072, ..., -0.0100, 0.1035, 0.1270], + [ 0.0410, 0.0531, 0.0445, ..., 0.0444, 0.0176, -0.0165], + [ 0.0628, 0.0354, 0.0617, ..., 0.0099, 0.0029, 0.0208], + ..., + [-0.1189, -0.1028, -0.0695, ..., 0.0021, 0.0180, -0.0154], + [-0.0129, -0.0021, -0.0254, ..., 0.0936, 0.0385, -0.0033], + [-0.0235, -0.0291, -0.0230, ..., -0.1298, -0.1629, -0.1274]], + device='cuda:0'), grad: tensor([[ 1.1024e-03, 1.6773e-04, 1.4496e-04, ..., 3.4833e-04, + 2.5558e-04, 3.3259e-04], + [ 4.5925e-05, 1.2770e-05, 1.7360e-06, ..., 1.1012e-05, + 1.0118e-05, 1.5013e-05], + [-1.8396e-03, -5.5504e-04, -3.8934e-04, ..., -4.4203e-04, + -5.5504e-04, -6.4707e-04], + ..., + [-9.4032e-04, -4.8190e-05, -7.2837e-05, ..., -3.4523e-04, + -1.7154e-04, -2.5058e-04], + [ 2.1040e-04, 2.8461e-05, 2.7418e-05, ..., 6.7711e-05, + 4.3333e-05, 5.9068e-05], + [ 8.0061e-04, 2.2531e-04, 1.6654e-04, ..., 2.0170e-04, + 2.3818e-04, 2.7800e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0210, -0.2532, 0.0062, -0.1753, 0.0016, 0.2897, 0.1620], + device='cuda:0'), grad: tensor([ 0.0037, 0.0001, -0.0050, 0.0018, -0.0036, 0.0007, 0.0022], + device='cuda:0') +351 +0.004328833670911726 +changing lr +epoch 38, time 486.95, cls_loss 0.0092 cls_loss_mapping 0.0883 cls_loss_causal 0.7727 re_mapping 0.0537 re_causal 0.0550 /// teacc 78.06 lr 0.00410722 +Epoch 40, weight, value: tensor([[-0.0224, -0.0141, -0.0071, ..., -0.0100, 0.1032, 0.1266], + [ 0.0395, 0.0524, 0.0439, ..., 0.0437, 0.0169, -0.0171], + [ 0.0637, 0.0356, 0.0616, ..., 0.0100, 0.0029, 0.0208], + ..., + [-0.1182, -0.1022, -0.0690, ..., 0.0022, 0.0181, -0.0152], + [-0.0136, -0.0023, -0.0255, ..., 0.0932, 0.0385, -0.0032], + [-0.0221, -0.0287, -0.0226, ..., -0.1290, -0.1619, -0.1264]], + device='cuda:0'), grad: tensor([[ 1.0431e-04, 8.6427e-06, 8.2403e-06, ..., 3.6299e-05, + 2.9549e-05, 2.5302e-05], + [ 2.6047e-05, 1.6242e-06, 3.4273e-07, ..., 8.7395e-06, + 7.0557e-06, 4.9658e-06], + [ 9.1696e-04, 7.5698e-05, 5.7697e-05, ..., 3.6097e-04, + 2.8181e-04, 2.4402e-04], + ..., + [ 1.0424e-03, 7.0632e-05, 5.5134e-05, ..., 3.4261e-04, + 2.7657e-04, 2.2483e-04], + [-2.4719e-03, -1.8394e-04, -1.4257e-04, ..., -8.9836e-04, + -7.1335e-04, -6.0034e-04], + [ 2.6083e-04, 1.5914e-05, 1.1899e-05, ..., 9.9719e-05, + 7.9751e-05, 6.6876e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0200, -0.2550, 0.0099, -0.1766, 0.0015, 0.2851, 0.1650], + device='cuda:0'), grad: tensor([ 3.3784e-04, 8.6427e-05, 2.9964e-03, 3.9721e-04, 3.4218e-03, + -8.0948e-03, 8.5974e-04], device='cuda:0') +351 +0.0041072155260068206 +changing lr +epoch 39, time 489.22, cls_loss 0.0097 cls_loss_mapping 0.0775 cls_loss_causal 0.7592 re_mapping 0.0543 re_causal 0.0557 /// teacc 82.28 lr 0.00388740 +Epoch 41, weight, value: tensor([[-0.0227, -0.0141, -0.0072, ..., -0.0101, 0.1026, 0.1259], + [ 0.0403, 0.0520, 0.0439, ..., 0.0437, 0.0173, -0.0165], + [ 0.0628, 0.0355, 0.0611, ..., 0.0095, 0.0024, 0.0202], + ..., + [-0.1180, -0.1017, -0.0687, ..., 0.0020, 0.0177, -0.0154], + [-0.0123, -0.0022, -0.0253, ..., 0.0937, 0.0388, -0.0026], + [-0.0233, -0.0287, -0.0227, ..., -0.1287, -0.1614, -0.1262]], + device='cuda:0'), grad: tensor([[ 4.6196e-03, 4.9496e-04, 5.7077e-04, ..., 1.1282e-03, + 2.0866e-03, 2.3994e-03], + [ 2.5436e-05, 1.1943e-05, 7.2420e-06, ..., -7.5214e-06, + -7.6517e-06, -4.2208e-06], + [-5.7106e-03, -7.9346e-04, -8.9979e-04, ..., -1.3742e-03, + -2.3251e-03, -2.7313e-03], + ..., + [ 2.3162e-04, 5.5492e-05, 6.3598e-05, ..., 5.3585e-05, + 6.6578e-05, 8.4639e-05], + [ 5.7983e-04, 1.6105e-04, 1.8013e-04, ..., 1.3936e-04, + 1.2082e-04, 1.7250e-04], + [ 1.0353e-04, 2.8998e-05, 3.2276e-05, ..., 2.4587e-05, + 2.3708e-05, 3.2336e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([-2.1788e-02, -2.5003e-01, 8.0240e-03, -1.7373e-01, -1.4076e-04, + 2.8711e-01, 1.6039e-01], device='cuda:0'), grad: tensor([ 1.5137e-02, 5.2959e-05, -1.8082e-02, 4.0960e-04, 6.5422e-04, + 1.5545e-03, 2.7752e-04], device='cuda:0') +351 +0.0038873953302184317 +changing lr +epoch 40, time 485.05, cls_loss 0.0115 cls_loss_mapping 0.0751 cls_loss_causal 0.7536 re_mapping 0.0549 re_causal 0.0564 /// teacc 88.61 lr 0.00366982 +Epoch 42, weight, value: tensor([[-0.0213, -0.0140, -0.0069, ..., -0.0095, 0.1029, 0.1261], + [ 0.0404, 0.0518, 0.0437, ..., 0.0434, 0.0173, -0.0165], + [ 0.0623, 0.0354, 0.0608, ..., 0.0093, 0.0022, 0.0200], + ..., + [-0.1180, -0.1012, -0.0685, ..., 0.0017, 0.0173, -0.0155], + [-0.0132, -0.0024, -0.0255, ..., 0.0930, 0.0385, -0.0028], + [-0.0228, -0.0285, -0.0225, ..., -0.1279, -0.1605, -0.1255]], + device='cuda:0'), grad: tensor([[ 5.6076e-03, 1.2617e-03, 1.4315e-03, ..., 2.1935e-03, + 1.7796e-03, 3.0479e-03], + [ 4.9362e-03, 2.6488e-04, 7.0047e-04, ..., 1.4935e-03, + 1.3771e-03, 1.4277e-03], + [ 1.0651e-04, -7.0781e-07, 8.5086e-06, ..., 3.8654e-05, + 3.2961e-05, 3.8892e-05], + ..., + [ 3.0041e-04, 2.2113e-05, 4.7088e-05, ..., 9.2387e-05, + 8.4519e-05, 9.2983e-05], + [ 9.2316e-03, 3.9482e-04, 1.2436e-03, ..., 2.7428e-03, + 2.5539e-03, 2.5234e-03], + [-2.0370e-02, -1.9531e-03, -3.4580e-03, ..., -6.6147e-03, + -5.8784e-03, -7.1831e-03]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0175, -0.2477, 0.0069, -0.1734, -0.0019, 0.2824, 0.1610], + device='cuda:0'), grad: tensor([ 0.0323, 0.0152, 0.0005, 0.0005, 0.0010, 0.0268, -0.0762], + device='cuda:0') +351 +0.003669815772166629 +changing lr +epoch 41, time 487.61, cls_loss 0.0074 cls_loss_mapping 0.0727 cls_loss_causal 0.7325 re_mapping 0.0550 re_causal 0.0565 /// teacc 85.23 lr 0.00345492 +Epoch 43, weight, value: tensor([[-0.0212, -0.0139, -0.0070, ..., -0.0095, 0.1025, 0.1256], + [ 0.0400, 0.0514, 0.0434, ..., 0.0431, 0.0171, -0.0166], + [ 0.0619, 0.0353, 0.0605, ..., 0.0092, 0.0022, 0.0200], + ..., + [-0.1162, -0.1007, -0.0678, ..., 0.0020, 0.0174, -0.0152], + [-0.0136, -0.0025, -0.0255, ..., 0.0928, 0.0385, -0.0026], + [-0.0236, -0.0285, -0.0226, ..., -0.1276, -0.1601, -0.1253]], + device='cuda:0'), grad: tensor([[ 1.8328e-05, 3.5167e-06, 4.0494e-06, ..., 6.2659e-06, + 3.9861e-06, 5.5842e-06], + [ 3.2261e-06, 8.4192e-07, 1.1250e-06, ..., 1.3001e-06, + 1.0841e-06, 1.3933e-06], + [ 2.5794e-05, 9.8720e-06, 1.1310e-05, ..., 9.2685e-06, + 9.9540e-06, 1.1086e-05], + ..., + [ 2.9549e-05, 1.0736e-05, 1.2115e-05, ..., 1.0125e-05, + 1.0498e-05, 1.1921e-05], + [ 1.4000e-05, 2.1495e-06, 3.7774e-06, ..., 5.8562e-06, + 3.5763e-06, 5.4203e-06], + [-3.0845e-05, -3.8296e-06, -5.8822e-06, ..., -1.1764e-05, + -6.1207e-06, -9.7081e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0182, -0.2470, 0.0065, -0.1719, 0.0034, 0.2799, 0.1570], + device='cuda:0'), grad: tensor([ 5.5403e-05, 8.9854e-06, 5.5552e-05, -1.2755e-04, 6.5923e-05, + 4.7505e-05, -1.0574e-04], device='cuda:0') +351 +0.0034549150281252667 +changing lr +epoch 42, time 486.22, cls_loss 0.0069 cls_loss_mapping 0.0656 cls_loss_causal 0.7226 re_mapping 0.0538 re_causal 0.0554 /// teacc 68.35 lr 0.00324313 +Epoch 44, weight, value: tensor([[-0.0216, -0.0140, -0.0071, ..., -0.0096, 0.1021, 0.1251], + [ 0.0402, 0.0512, 0.0433, ..., 0.0430, 0.0172, -0.0164], + [ 0.0620, 0.0353, 0.0603, ..., 0.0091, 0.0021, 0.0199], + ..., + [-0.1160, -0.1002, -0.0676, ..., 0.0017, 0.0170, -0.0154], + [-0.0131, -0.0025, -0.0254, ..., 0.0929, 0.0387, -0.0023], + [-0.0234, -0.0283, -0.0224, ..., -0.1271, -0.1595, -0.1248]], + device='cuda:0'), grad: tensor([[ 1.6963e-04, 2.2471e-05, 2.5943e-05, ..., 8.0049e-05, + 6.7055e-05, 7.7963e-05], + [-3.7432e-05, -6.4336e-06, -1.2316e-05, ..., -4.7833e-05, + -5.9873e-05, -6.4135e-05], + [-1.2279e-04, -5.3316e-05, -5.8621e-05, ..., -1.4551e-05, + -3.0726e-05, -4.2588e-05], + ..., + [ 3.2067e-04, 7.4446e-05, 8.3625e-05, ..., 9.1612e-05, + 9.7215e-05, 1.2577e-04], + [ 5.6219e-04, 5.9366e-05, 6.4909e-05, ..., 1.5295e-04, + 1.1712e-04, 1.6892e-04], + [-9.2030e-04, -9.9599e-05, -1.0693e-04, ..., -2.6965e-04, + -1.9670e-04, -2.7466e-04]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0191, -0.2443, 0.0069, -0.1719, 0.0013, 0.2796, 0.1572], + device='cuda:0'), grad: tensor([ 6.8665e-04, 2.4974e-05, -2.1517e-04, 9.7215e-05, 9.1743e-04, + 2.0123e-03, -3.5248e-03], device='cuda:0') +351 +0.0032431258795932905 +changing lr +epoch 43, time 489.72, cls_loss 0.0105 cls_loss_mapping 0.0591 cls_loss_causal 0.7122 re_mapping 0.0533 re_causal 0.0549 /// teacc 88.61 lr 0.00303487 +Epoch 45, weight, value: tensor([[-0.0222, -0.0140, -0.0073, ..., -0.0099, 0.1017, 0.1245], + [ 0.0400, 0.0509, 0.0431, ..., 0.0428, 0.0171, -0.0163], + [ 0.0616, 0.0353, 0.0601, ..., 0.0087, 0.0019, 0.0196], + ..., + [-0.1159, -0.0999, -0.0675, ..., 0.0017, 0.0166, -0.0157], + [-0.0133, -0.0026, -0.0255, ..., 0.0928, 0.0388, -0.0020], + [-0.0227, -0.0282, -0.0221, ..., -0.1265, -0.1588, -0.1241]], + device='cuda:0'), grad: tensor([[-8.6203e-06, -1.0803e-06, -2.7604e-06, ..., -4.3027e-06, + -4.8839e-06, -7.2680e-06], + [-3.8981e-04, -2.3216e-05, -9.8407e-05, ..., -1.4544e-04, + -2.0409e-04, -1.9073e-04], + [-3.0082e-06, -1.6894e-06, -1.5181e-06, ..., 3.5763e-07, + 3.6880e-07, 1.2107e-07], + ..., + [ 3.8028e-04, 2.3350e-05, 9.6440e-05, ..., 1.4150e-04, + 1.9836e-04, 1.8644e-04], + [ 6.8918e-06, 5.0664e-07, 1.7155e-06, ..., 2.4550e-06, + 3.3509e-06, 3.2224e-06], + [ 8.7097e-06, 1.1623e-06, 2.7977e-06, ..., 3.7644e-06, + 4.5709e-06, 5.8189e-06]], device='cuda:0') +Epoch 45, bias, value: tensor([-2.0831e-02, -2.4265e-01, 5.8777e-03, -1.6848e-01, 1.9903e-05, + 2.7736e-01, 1.5830e-01], device='cuda:0'), grad: tensor([-3.1412e-05, -1.2970e-03, -5.6811e-06, 1.4350e-05, 1.2646e-03, + 2.2843e-05, 2.9579e-05], device='cuda:0') +351 +0.0030348748417303863 +changing lr +epoch 44, time 487.09, cls_loss 0.0084 cls_loss_mapping 0.0627 cls_loss_causal 0.7363 re_mapping 0.0523 re_causal 0.0540 /// teacc 72.15 lr 0.00283058 +Epoch 46, weight, value: tensor([[-0.0226, -0.0141, -0.0073, ..., -0.0101, 0.1011, 0.1239], + [ 0.0396, 0.0506, 0.0428, ..., 0.0426, 0.0171, -0.0163], + [ 0.0621, 0.0355, 0.0601, ..., 0.0086, 0.0018, 0.0196], + ..., + [-0.1147, -0.0993, -0.0669, ..., 0.0018, 0.0167, -0.0154], + [-0.0123, -0.0025, -0.0252, ..., 0.0933, 0.0392, -0.0013], + [-0.0237, -0.0283, -0.0223, ..., -0.1264, -0.1586, -0.1242]], + device='cuda:0'), grad: tensor([[ 6.1005e-05, 8.5235e-06, 2.0206e-05, ..., 3.7134e-05, + 4.8548e-05, 5.1439e-05], + [-5.2303e-05, -7.2159e-06, -1.8224e-05, ..., -3.3587e-05, + -4.5151e-05, -4.7475e-05], + [-2.4978e-06, -8.5682e-07, -1.0654e-06, ..., 8.8662e-07, + -8.7544e-08, -2.1793e-07], + ..., + [ 2.4624e-06, 2.3469e-07, 5.1782e-07, ..., 1.7583e-06, + 1.6093e-06, 1.7621e-06], + [-6.5379e-07, -3.5018e-07, 9.3132e-08, ..., -8.7172e-06, + -6.4522e-06, -6.6496e-06], + [-8.6203e-06, -1.7323e-07, -1.4603e-06, ..., 1.7472e-06, + 8.6240e-07, 4.0419e-07]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0229, -0.2422, 0.0083, -0.1695, 0.0023, 0.2799, 0.1536], + device='cuda:0'), grad: tensor([ 2.0087e-04, -1.7345e-04, -5.4091e-06, 2.8592e-06, 8.5533e-06, + -5.4836e-06, -2.7969e-05], device='cuda:0') +351 +0.0028305813044122124 +changing lr +---------------------saving model at epoch 45---------------------------------------------------- +epoch 45, time 492.10, cls_loss 0.0091 cls_loss_mapping 0.0608 cls_loss_causal 0.7182 re_mapping 0.0522 re_causal 0.0539 /// teacc 90.30 lr 0.00263066 +Epoch 47, weight, value: tensor([[-0.0212, -0.0139, -0.0069, ..., -0.0095, 0.1019, 0.1245], + [ 0.0388, 0.0502, 0.0424, ..., 0.0422, 0.0167, -0.0167], + [ 0.0622, 0.0355, 0.0600, ..., 0.0085, 0.0017, 0.0195], + ..., + [-0.1149, -0.0990, -0.0669, ..., 0.0016, 0.0161, -0.0159], + [-0.0125, -0.0025, -0.0252, ..., 0.0931, 0.0391, -0.0013], + [-0.0236, -0.0282, -0.0222, ..., -0.1260, -0.1581, -0.1238]], + device='cuda:0'), grad: tensor([[-2.3961e-04, -1.5020e-05, -4.5061e-05, ..., -7.1526e-05, + -1.4293e-04, -1.5366e-04], + [ 4.7952e-05, 6.9402e-06, 1.2204e-05, ..., 1.8135e-05, + 2.4453e-05, 2.5079e-05], + [-3.4273e-06, -3.3509e-06, -2.6412e-06, ..., 1.6391e-06, + 3.9525e-06, 3.6862e-06], + ..., + [ 1.8501e-04, 1.4961e-05, 3.5077e-05, ..., 5.6356e-05, + 1.0079e-04, 1.0961e-04], + [-6.5863e-05, -2.5541e-05, -2.4706e-05, ..., -6.4194e-05, + -4.3005e-05, -4.3362e-05], + [-2.0619e-06, 1.1697e-06, -4.8615e-07, ..., 8.0764e-06, + 8.3670e-06, 9.6112e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([-1.8068e-02, -2.4327e-01, 9.5686e-03, -1.6941e-01, -7.8014e-05, + 2.7772e-01, 1.5303e-01], device='cuda:0'), grad: tensor([-8.3208e-04, 1.5223e-04, 4.2468e-07, 2.0671e-04, 6.2895e-04, + -1.4079e-04, -1.5587e-05], device='cuda:0') +351 +0.0026306566876350096 +changing lr +epoch 46, time 489.43, cls_loss 0.0071 cls_loss_mapping 0.0539 cls_loss_causal 0.6687 re_mapping 0.0523 re_causal 0.0540 /// teacc 81.86 lr 0.00243550 +Epoch 48, weight, value: tensor([[-0.0214, -0.0139, -0.0070, ..., -0.0097, 0.1014, 0.1239], + [ 0.0393, 0.0501, 0.0423, ..., 0.0422, 0.0167, -0.0165], + [ 0.0616, 0.0354, 0.0597, ..., 0.0082, 0.0015, 0.0192], + ..., + [-0.1144, -0.0986, -0.0665, ..., 0.0016, 0.0162, -0.0156], + [-0.0122, -0.0024, -0.0251, ..., 0.0932, 0.0393, -0.0009], + [-0.0237, -0.0282, -0.0221, ..., -0.1258, -0.1577, -0.1235]], + device='cuda:0'), grad: tensor([[ 4.4298e-04, 6.1333e-05, 1.1182e-04, ..., 2.1458e-04, + 1.8668e-04, 1.7023e-04], + [ 1.8826e-03, 2.6131e-04, 4.8971e-04, ..., 1.0023e-03, + 8.5163e-04, 7.7057e-04], + [ 2.8515e-04, 3.9458e-05, 6.5804e-05, ..., 7.5281e-05, + 7.8142e-05, 8.3566e-05], + ..., + [-2.8276e-04, -4.5151e-05, -1.1039e-04, ..., -2.2113e-04, + -1.5962e-04, -2.0111e-04], + [ 5.2691e-04, 7.3135e-05, 1.3399e-04, ..., 2.4748e-04, + 2.1160e-04, 2.0254e-04], + [-2.9583e-03, -4.0483e-04, -7.1955e-04, ..., -1.3676e-03, + -1.2093e-03, -1.0691e-03]], device='cuda:0') +Epoch 48, bias, value: tensor([-1.8857e-02, -2.4018e-01, 8.1123e-03, -1.6882e-01, -1.7916e-04, + 2.7774e-01, 1.5164e-01], device='cuda:0'), grad: tensor([ 0.0014, 0.0060, 0.0009, 0.0003, -0.0010, 0.0017, -0.0095], + device='cuda:0') +351 +0.0024355036129704724 +changing lr +epoch 47, time 487.96, cls_loss 0.0077 cls_loss_mapping 0.0550 cls_loss_causal 0.6829 re_mapping 0.0517 re_causal 0.0534 /// teacc 81.01 lr 0.00224552 +Epoch 49, weight, value: tensor([[-0.0218, -0.0139, -0.0071, ..., -0.0099, 0.1009, 0.1234], + [ 0.0387, 0.0498, 0.0420, ..., 0.0420, 0.0165, -0.0166], + [ 0.0617, 0.0354, 0.0596, ..., 0.0082, 0.0014, 0.0191], + ..., + [-0.1136, -0.0983, -0.0662, ..., 0.0017, 0.0162, -0.0154], + [-0.0129, -0.0026, -0.0253, ..., 0.0928, 0.0391, -0.0010], + [-0.0224, -0.0280, -0.0218, ..., -0.1249, -0.1569, -0.1228]], + device='cuda:0'), grad: tensor([[ 1.1139e-05, 3.9563e-06, 4.2059e-06, ..., 5.1297e-06, + -3.5092e-06, -1.6186e-06], + [ 1.8522e-05, 5.1521e-06, 6.6422e-06, ..., 5.1931e-06, + 4.4443e-06, 4.9062e-06], + [-1.5345e-03, -5.2452e-04, -6.5041e-04, ..., -2.7299e-04, + -2.6488e-04, -3.1328e-04], + ..., + [ 2.2805e-04, 7.3135e-05, 9.2268e-05, ..., 4.5002e-05, + 4.8876e-05, 5.4628e-05], + [ 1.2875e-05, 2.3752e-05, 2.5123e-05, ..., -2.8685e-05, + -1.7419e-05, -1.6347e-05], + [ 1.1272e-03, 3.7432e-04, 4.6635e-04, ..., 2.1875e-04, + 2.0671e-04, 2.4176e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0203, -0.2404, 0.0085, -0.1690, 0.0013, 0.2743, 0.1549], + device='cuda:0'), grad: tensor([ 2.9311e-05, 4.7684e-05, -3.3817e-03, 3.1281e-04, 5.2595e-04, + -8.5831e-05, 2.5520e-03], device='cuda:0') +351 +0.00224551509273949 +changing lr +epoch 48, time 489.20, cls_loss 0.0051 cls_loss_mapping 0.0480 cls_loss_causal 0.6975 re_mapping 0.0510 re_causal 0.0528 /// teacc 82.28 lr 0.00206107 +Epoch 50, weight, value: tensor([[-0.0211, -0.0138, -0.0069, ..., -0.0097, 0.1009, 0.1233], + [ 0.0382, 0.0495, 0.0417, ..., 0.0417, 0.0164, -0.0167], + [ 0.0627, 0.0359, 0.0599, ..., 0.0083, 0.0016, 0.0193], + ..., + [-0.1137, -0.0980, -0.0661, ..., 0.0015, 0.0159, -0.0155], + [-0.0126, -0.0026, -0.0252, ..., 0.0929, 0.0392, -0.0008], + [-0.0227, -0.0280, -0.0217, ..., -0.1247, -0.1566, -0.1226]], + device='cuda:0'), grad: tensor([[ 1.1253e-03, 1.8346e-04, 4.4370e-04, ..., 3.2568e-04, + 3.2568e-04, 3.5405e-04], + [-1.6794e-03, -2.7108e-04, -6.5708e-04, ..., -4.8637e-04, + -4.8971e-04, -5.3120e-04], + [ 1.3448e-05, 1.6382e-06, 4.7870e-06, ..., 4.1053e-06, + 4.4927e-06, 4.7088e-06], + ..., + [ 6.8322e-06, 7.2643e-07, 1.7742e-06, ..., 1.8915e-06, + 2.3209e-06, 2.2016e-06], + [ 1.9342e-05, 3.0510e-06, 7.3537e-06, ..., 5.6028e-06, + 5.7891e-06, 6.2510e-06], + [ 5.0974e-04, 8.1480e-05, 1.9753e-04, ..., 1.4758e-04, + 1.5008e-04, 1.6248e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0180, -0.2408, 0.0113, -0.1703, -0.0003, 0.2742, 0.1533], + device='cuda:0'), grad: tensor([ 3.3379e-03, -4.9934e-03, 4.1991e-05, 8.9183e-06, 2.1189e-05, + 5.7757e-05, 1.5182e-03], device='cuda:0') +351 +0.002061073738537637 +changing lr +epoch 49, time 484.91, cls_loss 0.0058 cls_loss_mapping 0.0522 cls_loss_causal 0.6878 re_mapping 0.0512 re_causal 0.0531 /// teacc 75.53 lr 0.00188255 +Epoch 51, weight, value: tensor([[-0.0213, -0.0139, -0.0069, ..., -0.0098, 0.1006, 0.1229], + [ 0.0383, 0.0494, 0.0416, ..., 0.0417, 0.0165, -0.0165], + [ 0.0622, 0.0358, 0.0597, ..., 0.0082, 0.0014, 0.0190], + ..., + [-0.1135, -0.0978, -0.0659, ..., 0.0015, 0.0158, -0.0155], + [-0.0126, -0.0026, -0.0252, ..., 0.0928, 0.0393, -0.0006], + [-0.0230, -0.0279, -0.0217, ..., -0.1246, -0.1563, -0.1224]], + device='cuda:0'), grad: tensor([[ 3.5453e-04, 4.2230e-05, 8.2910e-05, ..., 1.2863e-04, + 1.9825e-04, 1.6248e-04], + [ 4.3213e-05, 5.1521e-06, 8.0839e-06, ..., 2.7746e-05, + 2.8342e-05, 2.6956e-05], + [ 2.9758e-05, 3.7849e-06, 5.3681e-06, ..., 1.8895e-05, + 1.9521e-05, 1.8239e-05], + ..., + [-3.7861e-04, -3.9101e-05, -9.9599e-05, ..., -7.7963e-05, + -2.0134e-04, -1.4389e-04], + [-1.8530e-03, -3.1042e-04, -3.1281e-04, ..., -1.2970e-03, + -1.1034e-03, -1.1406e-03], + [ 1.6336e-03, 2.7657e-04, 2.7704e-04, ..., 1.1320e-03, + 9.6369e-04, 9.9659e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0185, -0.2392, 0.0103, -0.1683, -0.0005, 0.2735, 0.1519], + device='cuda:0'), grad: tensor([ 1.2074e-03, 1.3626e-04, 9.5785e-05, 5.7364e-04, -1.3552e-03, + -5.5962e-03, 4.9400e-03], device='cuda:0') +351 +0.0018825509907063344 +changing lr +epoch 50, time 487.79, cls_loss 0.0047 cls_loss_mapping 0.0466 cls_loss_causal 0.6943 re_mapping 0.0511 re_causal 0.0530 /// teacc 78.90 lr 0.00171031 +Epoch 52, weight, value: tensor([[-0.0216, -0.0139, -0.0070, ..., -0.0099, 0.1002, 0.1225], + [ 0.0385, 0.0493, 0.0415, ..., 0.0417, 0.0165, -0.0164], + [ 0.0622, 0.0358, 0.0596, ..., 0.0081, 0.0013, 0.0190], + ..., + [-0.1131, -0.0975, -0.0657, ..., 0.0015, 0.0157, -0.0155], + [-0.0127, -0.0027, -0.0252, ..., 0.0927, 0.0393, -0.0006], + [-0.0229, -0.0279, -0.0216, ..., -0.1242, -0.1559, -0.1221]], + device='cuda:0'), grad: tensor([[-1.2982e-04, -1.2673e-05, -4.6909e-05, ..., -6.3837e-05, + -1.1736e-04, -1.3304e-04], + [-2.3592e-04, -4.2081e-05, -3.9279e-05, ..., -8.7023e-05, + -8.8811e-05, -6.4790e-05], + [-3.4034e-05, -2.2024e-05, -2.2888e-05, ..., 6.2622e-06, + 9.6485e-06, 3.5074e-06], + ..., + [ 2.4235e-04, 3.5316e-05, 6.6459e-05, ..., 1.0478e-04, + 1.4913e-04, 1.5426e-04], + [-9.1434e-05, 1.8887e-06, -5.2340e-06, ..., -7.2837e-05, + -4.9442e-05, -4.6045e-05], + [ 1.8239e-04, 2.4110e-05, 3.0428e-05, ..., 9.0301e-05, + 7.6056e-05, 6.6817e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0198, -0.2375, 0.0106, -0.1674, -0.0003, 0.2722, 0.1516], + device='cuda:0'), grad: tensor([-5.4979e-04, -6.8712e-04, -4.6283e-05, 1.8716e-04, 8.5592e-04, + -3.2711e-04, 5.6744e-04], device='cuda:0') +351 +0.0017103063703014388 +changing lr +epoch 51, time 488.35, cls_loss 0.0062 cls_loss_mapping 0.0457 cls_loss_causal 0.6795 re_mapping 0.0503 re_causal 0.0522 /// teacc 78.48 lr 0.00154469 +Epoch 53, weight, value: tensor([[-0.0217, -0.0139, -0.0071, ..., -0.0100, 0.1000, 0.1222], + [ 0.0387, 0.0492, 0.0415, ..., 0.0417, 0.0166, -0.0162], + [ 0.0617, 0.0357, 0.0594, ..., 0.0078, 0.0011, 0.0187], + ..., + [-0.1128, -0.0972, -0.0655, ..., 0.0015, 0.0156, -0.0155], + [-0.0125, -0.0027, -0.0252, ..., 0.0927, 0.0394, -0.0004], + [-0.0228, -0.0278, -0.0215, ..., -0.1240, -0.1556, -0.1218]], + device='cuda:0'), grad: tensor([[-4.3184e-05, -7.6443e-06, -1.2971e-05, ..., -8.9481e-06, + -2.2054e-05, -2.4438e-05], + [ 3.1084e-05, 5.4389e-06, 8.3521e-06, ..., 8.5458e-06, + 1.3798e-05, 1.4573e-05], + [ 1.3590e-04, 1.5318e-05, 2.5660e-05, ..., 6.3658e-05, + 4.5538e-05, 4.2319e-05], + ..., + [-1.6212e-04, -1.7196e-05, -2.8029e-05, ..., -8.2672e-05, + -5.1290e-05, -4.5180e-05], + [ 2.1487e-05, 2.5854e-06, 4.3064e-06, ..., 9.0301e-06, + 6.7316e-06, 6.4820e-06], + [-1.1288e-05, -1.6820e-06, -2.6543e-06, ..., -2.7064e-06, + -1.9800e-06, -2.3786e-06]], device='cuda:0') +Epoch 53, bias, value: tensor([-2.0178e-02, -2.3575e-01, 9.2648e-03, -1.6716e-01, -2.3270e-04, + 2.7194e-01, 1.5140e-01], device='cuda:0'), grad: tensor([-1.4365e-04, 9.9480e-05, 4.1962e-04, 8.7023e-05, -4.9305e-04, + 6.7174e-05, -3.6746e-05], device='cuda:0') +351 +0.0015446867550656784 +changing lr +epoch 52, time 486.84, cls_loss 0.0055 cls_loss_mapping 0.0471 cls_loss_causal 0.6569 re_mapping 0.0497 re_causal 0.0515 /// teacc 85.23 lr 0.00138603 +Epoch 54, weight, value: tensor([[-2.1269e-02, -1.3865e-02, -6.9552e-03, ..., -9.8699e-03, + 1.0012e-01, 1.2224e-01], + [ 3.8476e-02, 4.8979e-02, 4.1318e-02, ..., 4.1490e-02, + 1.6448e-02, -1.6356e-02], + [ 6.1428e-02, 3.5670e-02, 5.9260e-02, ..., 7.6740e-03, + 9.0968e-04, 1.8529e-02], + ..., + [-1.1242e-01, -9.6968e-02, -6.5334e-02, ..., 1.5697e-03, + 1.5542e-02, -1.5468e-02], + [-1.2210e-02, -2.6671e-03, -2.5152e-02, ..., 9.2897e-02, + 3.9566e-02, -1.3846e-04], + [-2.3079e-02, -2.7743e-02, -2.1526e-02, ..., -1.2388e-01, + -1.5544e-01, -1.2169e-01]], device='cuda:0'), grad: tensor([[-6.1512e-04, -2.2143e-05, -1.5163e-04, ..., -1.7321e-04, + -3.0017e-04, -3.2377e-04], + [ 3.2759e-04, 6.2287e-05, 9.6262e-05, ..., 7.9751e-05, + 8.0824e-05, 9.2447e-05], + [ 1.1148e-03, 3.7837e-04, 4.0412e-04, ..., 3.4690e-04, + 2.2399e-04, 3.0065e-04], + ..., + [ 6.1750e-04, 8.4043e-05, 1.8024e-04, ..., 1.6785e-04, + 2.0778e-04, 2.3592e-04], + [ 6.5422e-04, 2.2161e-04, 2.3806e-04, ..., 1.9085e-04, + 1.0508e-04, 1.4663e-04], + [ 2.6894e-04, 1.5521e-04, 1.3745e-04, ..., 1.1659e-04, + 3.2067e-05, 6.8426e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([-1.8857e-02, -2.3560e-01, 8.6756e-03, -1.6740e-01, 2.6472e-05, + 2.7247e-01, 1.4994e-01], device='cuda:0'), grad: tensor([-0.0022, 0.0010, 0.0029, -0.0059, 0.0020, 0.0017, 0.0005], + device='cuda:0') +351 +0.001386025680863044 +changing lr +---------------------saving model at epoch 53---------------------------------------------------- +epoch 53, time 490.61, cls_loss 0.0052 cls_loss_mapping 0.0430 cls_loss_causal 0.7053 re_mapping 0.0491 re_causal 0.0511 /// teacc 90.72 lr 0.00123464 +Epoch 55, weight, value: tensor([[-2.1294e-02, -1.3864e-02, -6.9845e-03, ..., -9.8950e-03, + 9.9935e-02, 1.2198e-01], + [ 3.8158e-02, 4.8834e-02, 4.1167e-02, ..., 4.1333e-02, + 1.6349e-02, -1.6418e-02], + [ 6.1134e-02, 3.5595e-02, 5.9100e-02, ..., 7.4893e-03, + 7.9670e-04, 1.8389e-02], + ..., + [-1.1226e-01, -9.6775e-02, -6.5204e-02, ..., 1.5158e-03, + 1.5462e-02, -1.5474e-02], + [-1.2280e-02, -2.6849e-03, -2.5158e-02, ..., 9.2897e-02, + 3.9609e-02, -5.1898e-05], + [-2.2534e-02, -2.7644e-02, -2.1351e-02, ..., -1.2353e-01, + -1.5505e-01, -1.2130e-01]], device='cuda:0'), grad: tensor([[ 3.6740e-04, 1.1533e-04, 1.4138e-04, ..., 7.1287e-05, + 6.3181e-05, 8.7380e-05], + [ 1.0300e-04, 1.9923e-05, 2.7791e-05, ..., 2.7940e-05, + 3.5048e-05, 3.9577e-05], + [-4.7874e-03, -1.3676e-03, -1.7347e-03, ..., -1.0529e-03, + -1.1044e-03, -1.3723e-03], + ..., + [ 2.4045e-04, 5.3853e-05, 6.8605e-05, ..., 5.7757e-05, + 6.1929e-05, 7.4923e-05], + [ 7.4482e-04, 1.7750e-04, 2.0897e-04, ..., 1.5354e-04, + 1.3340e-04, 1.7190e-04], + [ 1.6146e-03, 5.3740e-04, 6.7186e-04, ..., 3.2306e-04, + 3.2139e-04, 4.0817e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([-1.9001e-02, -2.3573e-01, 8.0842e-03, -1.6670e-01, -2.0496e-04, + 2.7154e-01, 1.5125e-01], device='cuda:0'), grad: tensor([ 0.0010, 0.0003, -0.0144, 0.0048, 0.0007, 0.0022, 0.0053], + device='cuda:0') +351 +0.0012346426699819469 +changing lr +epoch 54, time 488.59, cls_loss 0.0049 cls_loss_mapping 0.0387 cls_loss_causal 0.6318 re_mapping 0.0490 re_causal 0.0508 /// teacc 81.86 lr 0.00109084 +Epoch 56, weight, value: tensor([[-2.1449e-02, -1.3913e-02, -7.0644e-03, ..., -9.9782e-03, + 9.9706e-02, 1.2171e-01], + [ 3.8039e-02, 4.8727e-02, 4.1080e-02, ..., 4.1237e-02, + 1.6312e-02, -1.6407e-02], + [ 6.1227e-02, 3.5646e-02, 5.9084e-02, ..., 7.4554e-03, + 7.6650e-04, 1.8350e-02], + ..., + [-1.1180e-01, -9.6573e-02, -6.5017e-02, ..., 1.5741e-03, + 1.5466e-02, -1.5377e-02], + [-1.2142e-02, -2.6935e-03, -2.5135e-02, ..., 9.2940e-02, + 3.9678e-02, 7.4000e-05], + [-2.2755e-02, -2.7644e-02, -2.1360e-02, ..., -1.2344e-01, + -1.5488e-01, -1.2119e-01]], device='cuda:0'), grad: tensor([[ 1.8072e-04, 2.8282e-05, 2.8521e-05, ..., 1.1289e-04, + 1.1140e-04, 1.0514e-04], + [ 8.2701e-06, 2.1160e-06, -5.6066e-07, ..., 1.0535e-05, + 9.0897e-06, 9.0003e-06], + [ 2.6870e-04, 4.2528e-05, 4.1068e-05, ..., 1.8346e-04, + 1.8668e-04, 1.6940e-04], + ..., + [ 1.0824e-04, 2.4393e-05, 2.8685e-05, ..., 4.8369e-05, + 4.9680e-05, 4.8935e-05], + [-6.2132e-04, -8.6784e-05, -7.6830e-05, ..., -4.2725e-04, + -4.1485e-04, -3.9148e-04], + [ 1.1188e-04, 1.4149e-05, 1.2986e-05, ..., 7.1883e-05, + 6.5327e-05, 6.6519e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0195, -0.2354, 0.0085, -0.1666, 0.0006, 0.2715, 0.1501], + device='cuda:0'), grad: tensor([ 5.4550e-04, 2.2292e-05, 7.8392e-04, -1.4389e-04, 3.1972e-04, + -1.8892e-03, 3.6216e-04], device='cuda:0') +351 +0.0010908425876598518 +changing lr +epoch 55, time 487.74, cls_loss 0.0051 cls_loss_mapping 0.0425 cls_loss_causal 0.6995 re_mapping 0.0486 re_causal 0.0505 /// teacc 75.95 lr 0.00095492 +Epoch 57, weight, value: tensor([[-2.1218e-02, -1.3897e-02, -7.0272e-03, ..., -9.9669e-03, + 9.9634e-02, 1.2159e-01], + [ 3.7903e-02, 4.8628e-02, 4.0986e-02, ..., 4.1183e-02, + 1.6288e-02, -1.6388e-02], + [ 6.1276e-02, 3.5676e-02, 5.9051e-02, ..., 7.4110e-03, + 7.6182e-04, 1.8332e-02], + ..., + [-1.1156e-01, -9.6415e-02, -6.4905e-02, ..., 1.6178e-03, + 1.5434e-02, -1.5334e-02], + [-1.2345e-02, -2.7407e-03, -2.5176e-02, ..., 9.2830e-02, + 3.9633e-02, 7.1650e-05], + [-2.2691e-02, -2.7605e-02, -2.1297e-02, ..., -1.2326e-01, + -1.5467e-01, -1.2102e-01]], device='cuda:0'), grad: tensor([[ 3.3617e-04, 3.0011e-05, 6.1452e-05, ..., 1.2374e-04, + 1.1009e-04, 1.2720e-04], + [-1.7416e-04, -1.3128e-05, -3.3319e-05, ..., -8.8274e-05, + -1.1045e-04, -1.1855e-04], + [-4.5091e-05, -2.1130e-05, -1.8314e-05, ..., -1.2778e-05, + -1.1884e-05, -1.6809e-05], + ..., + [-3.4451e-04, -2.5943e-05, -5.5701e-05, ..., -9.2685e-05, + -3.5495e-05, -5.2899e-05], + [ 5.7518e-05, 9.0301e-06, 1.2822e-05, ..., 1.7747e-05, + 1.5348e-05, 1.8820e-05], + [ 1.1241e-04, 1.0774e-05, 1.9923e-05, ..., 3.4451e-05, + 2.0042e-05, 2.6032e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0187, -0.2352, 0.0087, -0.1668, 0.0009, 0.2703, 0.1501], + device='cuda:0'), grad: tensor([ 0.0011, -0.0005, -0.0001, 0.0002, -0.0012, 0.0002, 0.0004], + device='cuda:0') +351 +0.000954915028125264 +changing lr +epoch 56, time 487.53, cls_loss 0.0046 cls_loss_mapping 0.0386 cls_loss_causal 0.6371 re_mapping 0.0485 re_causal 0.0503 /// teacc 89.45 lr 0.00082713 +Epoch 58, weight, value: tensor([[-2.1349e-02, -1.3977e-02, -7.1024e-03, ..., -1.0061e-02, + 9.9467e-02, 1.2139e-01], + [ 3.7762e-02, 4.8537e-02, 4.0907e-02, ..., 4.1115e-02, + 1.6256e-02, -1.6382e-02], + [ 6.0902e-02, 3.5573e-02, 5.8886e-02, ..., 7.3153e-03, + 6.9126e-04, 1.8243e-02], + ..., + [-1.1138e-01, -9.6263e-02, -6.4804e-02, ..., 1.6073e-03, + 1.5345e-02, -1.5362e-02], + [-1.2528e-02, -2.7840e-03, -2.5215e-02, ..., 9.2736e-02, + 3.9597e-02, 7.3678e-05], + [-2.2541e-02, -2.7571e-02, -2.1249e-02, ..., -1.2309e-01, + -1.5450e-01, -1.2087e-01]], device='cuda:0'), grad: tensor([[ 8.9228e-05, 8.5235e-06, 1.4402e-05, ..., 4.6790e-05, + 3.9130e-05, 4.3839e-05], + [-4.0841e-04, -3.3736e-05, -6.2108e-05, ..., -1.9312e-04, + -1.0115e-04, -1.4365e-04], + [ 2.2307e-05, 1.9856e-06, 3.4813e-06, ..., 1.3821e-05, + 1.5780e-05, 1.5527e-05], + ..., + [ 5.1230e-05, 4.4182e-06, 7.8455e-06, ..., 2.4691e-05, + 1.4320e-05, 1.9193e-05], + [-8.1837e-05, -8.5905e-06, -1.2361e-05, ..., -6.3300e-05, + -8.9049e-05, -8.0585e-05], + [ 3.2496e-04, 2.7806e-05, 4.8935e-05, ..., 1.6892e-04, + 1.1927e-04, 1.4389e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0190, -0.2351, 0.0078, -0.1651, 0.0010, 0.2693, 0.1504], + device='cuda:0'), grad: tensor([ 3.0637e-04, -1.2379e-03, 9.0182e-05, 1.1995e-05, 1.5855e-04, + -3.9029e-04, 1.0614e-03], device='cuda:0') +351 +0.0008271337313934874 +changing lr +epoch 57, time 488.10, cls_loss 0.0042 cls_loss_mapping 0.0381 cls_loss_causal 0.6417 re_mapping 0.0488 re_causal 0.0507 /// teacc 79.32 lr 0.00070776 +Epoch 59, weight, value: tensor([[-2.1408e-02, -1.3989e-02, -7.1333e-03, ..., -1.0104e-02, + 9.9326e-02, 1.2121e-01], + [ 3.7841e-02, 4.8474e-02, 4.0876e-02, ..., 4.1118e-02, + 1.6304e-02, -1.6304e-02], + [ 6.1134e-02, 3.5647e-02, 5.8916e-02, ..., 7.3222e-03, + 6.9682e-04, 1.8255e-02], + ..., + [-1.1137e-01, -9.6154e-02, -6.4747e-02, ..., 1.5936e-03, + 1.5274e-02, -1.5383e-02], + [-1.2516e-02, -2.8023e-03, -2.5213e-02, ..., 9.2714e-02, + 3.9614e-02, 1.2627e-04], + [-2.2560e-02, -2.7561e-02, -2.1224e-02, ..., -1.2300e-01, + -1.5434e-01, -1.2075e-01]], device='cuda:0'), grad: tensor([[-6.8173e-06, 4.6566e-09, -1.0632e-05, ..., -4.0606e-07, + -1.4566e-05, -1.8001e-05], + [ 2.6083e-04, 4.6819e-05, 7.5758e-05, ..., 6.0946e-05, + 6.7711e-05, 6.5386e-05], + [ 6.2656e-04, 6.6280e-05, 7.1764e-05, ..., 1.4627e-04, + 1.2958e-04, 1.4138e-04], + ..., + [ 2.2799e-05, 2.8126e-06, 5.0887e-06, ..., 3.4329e-06, + 6.0424e-06, 6.2250e-06], + [ 1.7595e-04, 2.1204e-05, 2.5839e-05, ..., 4.1038e-05, + 3.7640e-05, 3.9846e-05], + [-1.1568e-03, -1.4627e-04, -1.7905e-04, ..., -2.6917e-04, + -2.4295e-04, -2.5249e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0193, -0.2344, 0.0084, -0.1653, 0.0006, 0.2690, 0.1501], + device='cuda:0'), grad: tensor([-1.3560e-05, 7.8678e-04, 1.9627e-03, 2.4247e-04, 6.6400e-05, + 5.4741e-04, -3.5896e-03], device='cuda:0') +351 +0.00070775603199067 +changing lr +epoch 58, time 484.91, cls_loss 0.0032 cls_loss_mapping 0.0359 cls_loss_causal 0.6249 re_mapping 0.0487 re_causal 0.0506 /// teacc 86.92 lr 0.00059702 +Epoch 60, weight, value: tensor([[-0.0212, -0.0140, -0.0071, ..., -0.0101, 0.0993, 0.1212], + [ 0.0379, 0.0484, 0.0408, ..., 0.0411, 0.0163, -0.0163], + [ 0.0611, 0.0356, 0.0589, ..., 0.0073, 0.0007, 0.0182], + ..., + [-0.1116, -0.0961, -0.0648, ..., 0.0015, 0.0151, -0.0155], + [-0.0125, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0002], + [-0.0226, -0.0275, -0.0212, ..., -0.1229, -0.1542, -0.1207]], + device='cuda:0'), grad: tensor([[ 4.3243e-05, 3.6228e-06, 9.9689e-06, ..., 4.7050e-06, + 9.7156e-06, 7.1637e-06], + [ 1.1969e-04, 9.7975e-06, 2.7657e-05, ..., 1.1198e-05, + 2.6122e-05, 1.8969e-05], + [ 1.9163e-05, 1.2685e-06, 4.2096e-06, ..., 2.0973e-06, + 4.4741e-06, 3.3099e-06], + ..., + [ 1.0237e-05, 1.1064e-06, 2.3302e-06, ..., 2.5518e-06, + 3.0324e-06, 2.6748e-06], + [ 6.7830e-05, 5.7332e-06, 1.5661e-05, ..., 7.5512e-06, + 1.5438e-05, 1.1593e-05], + [-2.8563e-04, -2.3916e-05, -6.5863e-05, ..., -3.0905e-05, + -6.4492e-05, -4.7982e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0185, -0.2338, 0.0084, -0.1651, -0.0005, 0.2689, 0.1499], + device='cuda:0'), grad: tensor([ 1.4329e-04, 3.9768e-04, 6.4731e-05, 8.4817e-05, 3.3200e-05, + 2.2471e-04, -9.4748e-04], device='cuda:0') +351 +0.0005970223407163104 +changing lr +epoch 59, time 492.38, cls_loss 0.0040 cls_loss_mapping 0.0384 cls_loss_causal 0.6520 re_mapping 0.0485 re_causal 0.0505 /// teacc 82.70 lr 0.00049516 +Epoch 61, weight, value: tensor([[-0.0212, -0.0140, -0.0071, ..., -0.0101, 0.0992, 0.1211], + [ 0.0379, 0.0484, 0.0408, ..., 0.0411, 0.0163, -0.0162], + [ 0.0611, 0.0357, 0.0589, ..., 0.0073, 0.0006, 0.0182], + ..., + [-0.1115, -0.0960, -0.0647, ..., 0.0015, 0.0151, -0.0155], + [-0.0122, -0.0028, -0.0252, ..., 0.0928, 0.0397, 0.0003], + [-0.0227, -0.0275, -0.0212, ..., -0.1229, -0.1541, -0.1206]], + device='cuda:0'), grad: tensor([[-2.4581e-04, -2.5958e-05, -5.0008e-05, ..., -4.5031e-05, + -1.1921e-04, -8.9407e-05], + [ 8.8871e-05, 5.2899e-06, 7.9200e-06, ..., 3.9667e-05, + 2.7984e-05, 2.2709e-05], + [ 2.7013e-04, 3.2693e-05, 3.7163e-05, ..., 1.3626e-04, + 1.0562e-04, 8.9467e-05], + ..., + [ 5.0497e-04, 4.4256e-05, 7.9632e-05, ..., 1.6642e-04, + 2.0587e-04, 1.6856e-04], + [-3.8967e-03, -2.2078e-04, -4.4274e-04, ..., -1.4009e-03, + -1.1187e-03, -9.2793e-04], + [ 3.1357e-03, 1.4865e-04, 3.4857e-04, ..., 1.0357e-03, + 8.4352e-04, 6.8998e-04]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0187, -0.2335, 0.0085, -0.1652, -0.0006, 0.2693, 0.1494], + device='cuda:0'), grad: tensor([-0.0008, 0.0003, 0.0008, 0.0005, 0.0016, -0.0127, 0.0103], + device='cuda:0') +351 +0.0004951556604879052 +changing lr +epoch 60, time 489.26, cls_loss 0.0038 cls_loss_mapping 0.0370 cls_loss_causal 0.6870 re_mapping 0.0484 re_causal 0.0504 /// teacc 82.28 lr 0.00040236 +Epoch 62, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0991, 0.1210], + [ 0.0379, 0.0483, 0.0408, ..., 0.0411, 0.0163, -0.0162], + [ 0.0610, 0.0357, 0.0588, ..., 0.0072, 0.0006, 0.0182], + ..., + [-0.1114, -0.0959, -0.0647, ..., 0.0015, 0.0151, -0.0155], + [-0.0122, -0.0028, -0.0252, ..., 0.0928, 0.0397, 0.0003], + [-0.0226, -0.0275, -0.0212, ..., -0.1228, -0.1540, -0.1205]], + device='cuda:0'), grad: tensor([[ 2.2188e-05, 8.2105e-06, 8.0541e-06, ..., 3.5763e-06, + 9.4622e-07, 1.9614e-06], + [-4.1723e-05, -2.1998e-06, -4.1761e-06, ..., -1.7449e-05, + -1.4827e-05, -1.3500e-05], + [-4.8637e-04, -1.8167e-04, -1.8668e-04, ..., -9.2447e-05, + -6.3539e-05, -8.6546e-05], + ..., + [ 7.5512e-06, 1.9222e-06, 2.1514e-06, ..., 2.1532e-06, + 1.8720e-06, 1.9912e-06], + [ 1.2040e-04, 4.2289e-05, 4.3929e-05, ..., 2.5019e-05, + 1.8135e-05, 2.3142e-05], + [ 1.1921e-04, 3.4183e-05, 3.6806e-05, ..., 2.8491e-05, + 2.1860e-05, 2.5377e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0192, -0.2331, 0.0083, -0.1652, -0.0005, 0.2692, 0.1497], + device='cuda:0'), grad: tensor([ 5.2631e-05, -1.3030e-04, -1.1740e-03, 6.2513e-04, 2.0266e-05, + 2.9707e-04, 3.0971e-04], device='cuda:0') +351 +0.00040236113724274745 +changing lr +epoch 61, time 489.43, cls_loss 0.0045 cls_loss_mapping 0.0367 cls_loss_causal 0.6441 re_mapping 0.0480 re_causal 0.0499 /// teacc 78.48 lr 0.00031883 +Epoch 63, weight, value: tensor([[-0.0213, -0.0140, -0.0071, ..., -0.0101, 0.0991, 0.1209], + [ 0.0378, 0.0483, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0609, 0.0356, 0.0588, ..., 0.0072, 0.0006, 0.0181], + ..., + [-0.1114, -0.0959, -0.0646, ..., 0.0015, 0.0151, -0.0155], + [-0.0123, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0224, -0.0275, -0.0211, ..., -0.1227, -0.1539, -0.1204]], + device='cuda:0'), grad: tensor([[ 1.2529e-04, 2.2218e-05, 2.8625e-05, ..., 3.4153e-05, + 2.4438e-05, 2.4959e-05], + [ 1.5825e-05, 1.6261e-06, 1.8962e-06, ..., 8.0094e-06, + 6.0126e-06, 6.2659e-06], + [-2.7156e-04, -8.5413e-05, -9.2268e-05, ..., -2.8670e-05, + -1.1556e-05, -9.0450e-06], + ..., + [-2.2966e-06, 7.2680e-06, 4.5262e-06, ..., -5.1968e-07, + 2.3246e-06, -6.3702e-07], + [-3.3855e-05, 2.4676e-05, 1.8314e-05, ..., -9.7752e-05, + -9.1434e-05, -9.2506e-05], + [ 4.2379e-05, 4.2804e-06, 8.4341e-06, ..., 4.6492e-05, + 4.1276e-05, 4.2289e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0190, -0.2331, 0.0080, -0.1650, -0.0006, 0.2687, 0.1501], + device='cuda:0'), grad: tensor([ 3.8457e-04, 5.1320e-05, -7.3528e-04, 3.6812e-04, -5.8979e-05, + -1.7655e-04, 1.6725e-04], device='cuda:0') +351 +0.00031882564680131423 +changing lr +epoch 62, time 483.86, cls_loss 0.0042 cls_loss_mapping 0.0371 cls_loss_causal 0.6606 re_mapping 0.0480 re_causal 0.0500 /// teacc 87.76 lr 0.00024472 +Epoch 64, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0483, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0608, 0.0356, 0.0587, ..., 0.0072, 0.0006, 0.0181], + ..., + [-0.1112, -0.0958, -0.0646, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0224, -0.0275, -0.0211, ..., -0.1227, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 2.7698e-06, 5.7667e-06, 3.1590e-06, ..., -3.6303e-06, + -1.3135e-05, -1.1802e-05], + [ 4.9829e-05, 7.4059e-06, 1.1750e-05, ..., 9.0599e-06, + 1.1452e-05, 1.4067e-05], + [-1.3173e-04, -5.1230e-05, -5.5671e-05, ..., -1.5497e-05, + -5.6326e-06, -1.2428e-05], + ..., + [ 2.7090e-05, 5.1446e-06, 6.8285e-06, ..., 3.5875e-06, + 2.9411e-06, 4.6305e-06], + [ 5.3972e-05, 1.4357e-05, 1.7211e-05, ..., 3.2391e-06, + 7.2457e-07, 4.1276e-06], + [-5.3346e-05, 9.4622e-07, -2.8946e-06, ..., -4.3772e-06, + -1.2368e-06, -6.2324e-06]], device='cuda:0') +Epoch 64, bias, value: tensor([-1.9339e-02, -2.3265e-01, 7.7978e-03, -1.6499e-01, -7.8075e-05, + 2.6835e-01, 1.5003e-01], device='cuda:0'), grad: tensor([-2.4382e-06, 1.5533e-04, -3.2139e-04, 1.3471e-04, 8.2672e-05, + 1.5342e-04, -2.0194e-04], device='cuda:0') +351 +0.0002447174185242325 +changing lr +epoch 63, time 490.22, cls_loss 0.0036 cls_loss_mapping 0.0347 cls_loss_causal 0.6470 re_mapping 0.0480 re_causal 0.0500 /// teacc 83.12 lr 0.00018019 +Epoch 65, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0483, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0608, 0.0356, 0.0587, ..., 0.0071, 0.0006, 0.0181], + ..., + [-0.1111, -0.0958, -0.0645, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0224, -0.0275, -0.0211, ..., -0.1226, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 5.2881e-04, 4.0114e-05, 1.0240e-04, ..., 1.3661e-04, + 1.6046e-04, 1.8704e-04], + [-8.0919e-04, -6.5684e-05, -7.5638e-05, ..., -4.0150e-04, + -2.9635e-04, -3.4237e-04], + [ 7.9060e-04, 5.5760e-05, 1.4138e-04, ..., 2.0337e-04, + 2.2817e-04, 2.6584e-04], + ..., + [-1.2274e-03, -8.4937e-05, -2.8634e-04, ..., -2.2125e-04, + -3.6216e-04, -4.2987e-04], + [ 3.4070e-04, 2.7388e-05, 4.6611e-05, ..., 1.2219e-04, + 1.0574e-04, 1.2124e-04], + [ 1.5497e-04, 1.1295e-05, 2.6509e-05, ..., 1.0812e-04, + 9.7275e-05, 1.2004e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([-1.9334e-02, -2.3244e-01, 7.7977e-03, -1.6502e-01, 1.9134e-05, + 2.6818e-01, 1.4991e-01], device='cuda:0'), grad: tensor([ 0.0018, -0.0027, 0.0026, 0.0007, -0.0042, 0.0011, 0.0006], + device='cuda:0') +351 +0.0001801856965207339 +changing lr +epoch 64, time 492.97, cls_loss 0.0040 cls_loss_mapping 0.0365 cls_loss_causal 0.6452 re_mapping 0.0477 re_causal 0.0498 /// teacc 82.28 lr 0.00012536 +Epoch 66, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0482, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0608, 0.0356, 0.0587, ..., 0.0071, 0.0006, 0.0181], + ..., + [-0.1110, -0.0957, -0.0645, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0224, -0.0275, -0.0211, ..., -0.1226, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 4.5419e-05, 6.3926e-06, 1.3947e-05, ..., 1.1124e-05, + 1.1191e-05, 1.2450e-05], + [-1.0870e-05, -2.7269e-06, -2.4606e-06, ..., -3.9451e-06, + -4.3772e-06, -5.6140e-06], + [ 4.3422e-05, 5.9567e-06, 1.4521e-05, ..., 7.3612e-06, + 5.9381e-06, 7.0743e-06], + ..., + [ 1.1817e-05, 2.1253e-06, 4.8801e-06, ..., -5.6345e-07, + -2.2799e-06, -1.7304e-06], + [ 2.2992e-05, 3.1963e-06, 7.7263e-06, ..., 3.9227e-06, + 3.1628e-06, 3.7048e-06], + [-1.1885e-04, -1.5810e-05, -4.0680e-05, ..., -1.8969e-05, + -1.4514e-05, -1.6913e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-1.9370e-02, -2.3244e-01, 7.9069e-03, -1.6501e-01, 9.5384e-05, + 2.6809e-01, 1.4984e-01], device='cuda:0'), grad: tensor([ 1.8620e-04, -2.0117e-05, 1.9896e-04, 2.7701e-05, 6.4075e-05, + 1.0514e-04, -5.6219e-04], device='cuda:0') +351 +0.000125360439090882 +changing lr +epoch 65, time 491.54, cls_loss 0.0036 cls_loss_mapping 0.0321 cls_loss_causal 0.6499 re_mapping 0.0478 re_causal 0.0498 /// teacc 85.65 lr 0.00008035 +Epoch 67, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0482, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0608, 0.0356, 0.0587, ..., 0.0071, 0.0005, 0.0181], + ..., + [-0.1110, -0.0957, -0.0645, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0224, -0.0275, -0.0211, ..., -0.1226, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 7.2360e-05, 5.2638e-06, 1.7837e-05, ..., 1.5587e-05, + 2.2396e-05, 2.0936e-05], + [ 9.2804e-05, 4.1537e-06, 2.0787e-05, ..., 2.1264e-05, + 3.0354e-05, 2.8640e-05], + [-9.5293e-06, -6.4522e-06, -6.5379e-06, ..., -1.0915e-06, + -2.5313e-06, -3.1143e-06], + ..., + [ 9.6187e-06, 5.1688e-07, 2.1979e-06, ..., 1.9427e-06, + 3.2876e-06, 2.9653e-06], + [ 7.5519e-05, 4.3809e-06, 1.7673e-05, ..., 1.7107e-05, + 2.4602e-05, 2.3276e-05], + [-2.6488e-04, -9.6783e-06, -5.7906e-05, ..., -6.0499e-05, + -8.6486e-05, -8.0824e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([-1.9302e-02, -2.3238e-01, 7.8917e-03, -1.6504e-01, 1.5820e-04, + 2.6802e-01, 1.4977e-01], device='cuda:0'), grad: tensor([ 2.1708e-04, 2.8634e-04, -1.5043e-05, 7.3016e-05, 2.8566e-05, + 2.3031e-04, -8.1968e-04], device='cuda:0') +351 +8.03520570068517e-05 +changing lr +epoch 66, time 488.28, cls_loss 0.0043 cls_loss_mapping 0.0363 cls_loss_causal 0.6060 re_mapping 0.0477 re_causal 0.0496 /// teacc 80.17 lr 0.00004525 +Epoch 68, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0482, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0609, 0.0356, 0.0587, ..., 0.0071, 0.0006, 0.0181], + ..., + [-0.1110, -0.0957, -0.0645, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0225, -0.0275, -0.0211, ..., -0.1226, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 1.7738e-04, 3.6508e-05, 4.7892e-05, ..., 3.3021e-05, + 5.1409e-05, 5.8830e-05], + [-1.1176e-04, -1.2986e-05, -2.6107e-05, ..., -6.1005e-05, + -7.1406e-05, -7.7963e-05], + [-1.8620e-04, -4.3780e-05, -5.1647e-05, ..., -1.6481e-05, + -4.3839e-05, -4.8906e-05], + ..., + [ 9.0122e-05, 1.1779e-05, 2.0489e-05, ..., 2.9355e-05, + 3.2455e-05, 3.6657e-05], + [ 1.6510e-04, 2.1622e-05, 3.7253e-05, ..., 5.1230e-05, + 5.6177e-05, 6.3777e-05], + [-2.3258e-04, -2.8268e-05, -5.0306e-05, ..., -5.5730e-05, + -4.7892e-05, -5.9724e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([-1.9269e-02, -2.3232e-01, 8.0016e-03, -1.6504e-01, 1.6669e-04, + 2.6794e-01, 1.4963e-01], device='cuda:0'), grad: tensor([ 0.0005, -0.0004, -0.0005, 0.0003, 0.0003, 0.0005, -0.0007], + device='cuda:0') +351 +4.5251191160326525e-05 +changing lr +epoch 67, time 486.30, cls_loss 0.0037 cls_loss_mapping 0.0329 cls_loss_causal 0.6237 re_mapping 0.0478 re_causal 0.0498 /// teacc 78.48 lr 0.00002013 +Epoch 69, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0482, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0609, 0.0356, 0.0587, ..., 0.0071, 0.0006, 0.0181], + ..., + [-0.1110, -0.0957, -0.0645, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0225, -0.0275, -0.0211, ..., -0.1226, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 8.3113e-04, 7.4923e-05, 1.7071e-04, ..., 2.6250e-04, + 3.9792e-04, 3.2210e-04], + [-9.2602e-04, -8.4460e-05, -1.9145e-04, ..., -2.9302e-04, + -4.3583e-04, -3.5167e-04], + [-5.4128e-06, -2.5220e-06, -2.1048e-06, ..., 1.5944e-06, + 1.8477e-06, 1.1027e-06], + ..., + [ 2.8223e-05, 2.8014e-06, 6.0685e-06, ..., 8.5831e-06, + 1.2398e-05, 1.0081e-05], + [ 2.3380e-05, 2.6226e-06, 5.3160e-06, ..., 7.4133e-06, + 8.4266e-06, 6.3777e-06], + [ 4.0412e-05, 4.8652e-06, 9.2611e-06, ..., 1.1846e-05, + 1.4052e-05, 1.0982e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-1.9275e-02, -2.3228e-01, 8.0206e-03, -1.6505e-01, 1.7925e-04, + 2.6789e-01, 1.4963e-01], device='cuda:0'), grad: tensor([ 2.4948e-03, -2.7771e-03, -1.3143e-05, 2.4006e-05, 8.4937e-05, + 6.9320e-05, 1.1951e-04], device='cuda:0') +351 +2.0128530023804673e-05 +changing lr +epoch 68, time 488.59, cls_loss 0.0038 cls_loss_mapping 0.0348 cls_loss_causal 0.6464 re_mapping 0.0477 re_causal 0.0497 /// teacc 81.01 lr 0.00000503 +Epoch 70, weight, value: tensor([[-0.0214, -0.0140, -0.0071, ..., -0.0102, 0.0990, 0.1208], + [ 0.0379, 0.0482, 0.0407, ..., 0.0410, 0.0163, -0.0162], + [ 0.0609, 0.0356, 0.0587, ..., 0.0071, 0.0006, 0.0181], + ..., + [-0.1110, -0.0957, -0.0645, ..., 0.0016, 0.0151, -0.0154], + [-0.0124, -0.0028, -0.0252, ..., 0.0927, 0.0397, 0.0003], + [-0.0224, -0.0275, -0.0211, ..., -0.1226, -0.1538, -0.1203]], + device='cuda:0'), grad: tensor([[ 2.3559e-05, 6.6012e-06, 1.9409e-06, ..., -3.4779e-05, + -7.4029e-05, -6.5267e-05], + [ 2.2388e-04, 2.8431e-05, 4.1455e-05, ..., 3.1710e-05, + 5.0277e-05, 5.2512e-05], + [-5.5850e-05, -4.7028e-05, -3.0845e-05, ..., -2.1219e-05, + 1.6883e-05, 5.6773e-06], + ..., + [ 7.6115e-05, 1.3202e-05, 1.9178e-05, ..., 2.7522e-05, + 4.1485e-05, 4.3154e-05], + [ 4.0293e-04, 5.9634e-05, 8.0347e-05, ..., 5.4657e-05, + 7.3731e-05, 7.9453e-05], + [-8.0967e-04, -9.9838e-05, -1.5163e-04, ..., -9.5963e-05, + -1.4210e-04, -1.5724e-04]], device='cuda:0') +Epoch 70, bias, value: tensor([-1.9308e-02, -2.3226e-01, 8.0159e-03, -1.6505e-01, 1.7581e-04, + 2.6787e-01, 1.4967e-01], device='cuda:0'), grad: tensor([ 5.3346e-05, 7.1144e-04, -1.0407e-04, 4.0460e-04, 2.4176e-04, + 1.2579e-03, -2.5673e-03], device='cuda:0') +351 +5.034667293427056e-06 +changing lr +epoch 69, time 490.70, cls_loss 0.0033 cls_loss_mapping 0.0336 cls_loss_causal 0.6347 re_mapping 0.0477 re_causal 0.0497 /// teacc 83.54 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'source_domain': 'cartoon', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1/cartoon_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['cartoon', 'art_painting', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + cartoon art_painting photo sketch Avg +w/o do (original x) 99.061433 54.6875 66.047904 60.066175 60.267193 + cartoon art_painting photo sketch Avg +do 99.146758 55.712891 72.45509 63.425808 63.864596 diff --git a/Meta-causal/code-withStyleAttack/64946.error b/Meta-causal/code-withStyleAttack/64946.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/64946.log b/Meta-causal/code-withStyleAttack/64946.log new file mode 100644 index 0000000000000000000000000000000000000000..7fb9e7a47f327a6500968fbe621b218e91069882 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/64946.log @@ -0,0 +1,1950 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'photo', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_train.hdf5 torch.Size([1499, 3, 227, 227]) torch.Size([1499]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_val.hdf5 torch.Size([171, 3, 227, 227]) torch.Size([171]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[ 1.3781e-02, -2.1813e-02, -1.7643e-02, ..., -3.4786e-03, + 7.0747e-03, 1.6051e-02], + [ 1.6921e-02, 1.6796e-02, 3.9765e-03, ..., 6.6386e-03, + 2.4196e-03, 9.7383e-03], + [ 1.8894e-03, -6.7549e-03, 2.1032e-02, ..., -1.9916e-02, + -1.0781e-02, 1.7924e-02], + ..., + [ 4.6711e-04, -7.8218e-03, 8.8535e-03, ..., -1.7695e-02, + -1.9742e-02, -5.5186e-03], + [-1.4236e-02, 1.5676e-02, -6.0038e-03, ..., 1.4177e-02, + 1.0025e-02, -3.0311e-03], + [-1.4947e-02, 1.6332e-02, 1.3555e-02, ..., 1.0778e-02, + 9.8178e-05, -1.3844e-02]], device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0165, -0.0171, 0.0209, -0.0155, 0.0020, -0.0164, 0.0005], + device='cuda:0'), grad: None +249 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 348.21, cls_loss 9.8684 cls_loss_mapping 1.6022 cls_loss_causal 1.9054 re_mapping 1.0159 re_causal 1.0189 /// teacc 69.59 lr 0.00999497 +Epoch 2, weight, value: tensor([[ 0.0774, 0.0641, 0.0080, ..., -0.0354, -0.0471, -0.0399], + [-0.0260, -0.0458, -0.0257, ..., -0.0312, -0.0105, -0.0277], + [-0.0029, -0.0232, 0.0263, ..., -0.0459, -0.0319, -0.0043], + ..., + [ 0.0248, 0.0389, 0.0264, ..., 0.0109, -0.0060, 0.0109], + [-0.0430, -0.0148, -0.0446, ..., -0.0595, -0.0504, -0.0497], + [ 0.0564, 0.0909, 0.0901, ..., 0.0745, 0.0518, 0.0574]], + device='cuda:0'), grad: tensor([[-1.0231e-02, -3.4180e-03, -1.1459e-02, ..., 7.9918e-04, + -4.7612e-04, -1.6251e-03], + [ 1.5697e-03, 1.3151e-03, 1.7748e-03, ..., 1.2712e-03, + 1.7195e-03, 1.3046e-03], + [ 7.8049e-03, 2.6760e-03, 9.8877e-03, ..., 4.4975e-03, + 4.5662e-03, 3.6087e-03], + ..., + [ 2.2011e-03, 1.6279e-03, 2.6894e-03, ..., 2.0332e-03, + 2.4071e-03, 1.8253e-03], + [ 1.2942e-05, 4.3437e-06, 1.5251e-05, ..., 2.3097e-06, + 3.3155e-06, 3.5837e-06], + [-3.1967e-03, -2.8076e-03, -3.6583e-03, ..., -2.7180e-03, + -3.6221e-03, -2.7809e-03]], device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0328, -0.0842, 0.2049, 0.0465, 0.0064, -0.1213, -0.0942], + device='cuda:0'), grad: tensor([-0.0866, 0.0029, 0.1262, -0.0544, 0.0152, 0.0001, -0.0035], + device='cuda:0') +249 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 347.28, cls_loss 3.2468 cls_loss_mapping 1.3975 cls_loss_causal 1.8708 re_mapping 0.7421 re_causal 0.7474 /// teacc 80.70 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 0.0785, 0.0747, 0.0011, ..., -0.0510, -0.0911, -0.0519], + [-0.0322, -0.0574, -0.0140, ..., -0.0477, -0.0259, -0.0421], + [-0.0056, -0.0181, 0.0213, ..., -0.0507, -0.0635, -0.0161], + ..., + [-0.0080, 0.0068, -0.0124, ..., -0.0106, -0.0226, -0.0205], + [-0.0517, -0.0187, -0.0551, ..., -0.0376, -0.0179, -0.0239], + [ 0.0674, 0.1059, 0.1026, ..., 0.1048, 0.0980, 0.0961]], + device='cuda:0'), grad: tensor([[ 3.9101e-05, 3.8743e-05, 1.3423e-04, ..., 1.6487e-04, + 1.4436e-04, 1.5497e-04], + [ 1.8263e-03, 1.4849e-03, 3.7289e-03, ..., 2.7828e-03, + 2.1725e-03, 2.8553e-03], + [ 4.3368e-04, 9.7847e-04, 3.6945e-03, ..., 7.3967e-03, + 7.4463e-03, 7.1487e-03], + ..., + [-2.4109e-03, -1.7910e-03, -5.5389e-03, ..., -3.6659e-03, + -2.2316e-03, -3.3226e-03], + [ 1.4687e-04, 1.4591e-04, 7.2622e-04, ..., 9.2077e-04, + 7.4911e-04, 8.0156e-04], + [ 3.0994e-06, 2.9132e-06, 1.0602e-05, ..., 1.2219e-05, + 1.0259e-05, 1.1235e-05]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0016, -0.0449, 0.2193, 0.0254, 0.0386, -0.1846, -0.0613], + device='cuda:0'), grad: tensor([ 0.0033, 0.0715, 0.1316, -0.1288, -0.0958, 0.0180, 0.0003], + device='cuda:0') +249 +0.009979871469976196 +changing lr +epoch 2, time 346.20, cls_loss 3.2564 cls_loss_mapping 1.4641 cls_loss_causal 1.8684 re_mapping 0.6795 re_causal 0.6891 /// teacc 63.16 lr 0.00995475 +Epoch 4, weight, value: tensor([[ 0.1029, 0.0873, 0.0093, ..., -0.0784, -0.1080, -0.0719], + [-0.0437, -0.0644, -0.0257, ..., -0.0425, -0.0232, -0.0373], + [-0.0310, -0.0245, -0.0142, ..., -0.0414, -0.0979, -0.0245], + ..., + [-0.0239, -0.0071, -0.0055, ..., -0.0321, -0.0321, -0.0325], + [-0.0364, -0.0142, -0.0472, ..., -0.0315, -0.0031, -0.0247], + [ 0.0723, 0.1110, 0.0884, ..., 0.1055, 0.1089, 0.1103]], + device='cuda:0'), grad: tensor([[-7.9393e-04, -1.0939e-03, -4.1733e-03, ..., -1.1044e-03, + -8.3113e-04, -1.1721e-03], + [ 2.8372e-04, 3.9291e-04, 2.0332e-03, ..., 5.7697e-04, + 4.7493e-04, 6.3467e-04], + [ 1.4377e-04, 1.9968e-04, 1.2112e-03, ..., 3.5334e-04, + 3.0017e-04, 3.9363e-04], + ..., + [ 4.0245e-04, 5.5695e-04, 2.8667e-03, ..., 8.1253e-04, + 6.6805e-04, 8.9312e-04], + [ 6.9797e-05, 9.6500e-05, 5.6171e-04, ..., 1.6296e-04, + 1.3769e-04, 1.8096e-04], + [-1.5214e-05, 1.3635e-06, 7.7248e-05, ..., 1.8284e-05, + -2.4140e-06, 2.1279e-05]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0196, 0.1142, 0.2746, 0.0413, -0.0684, -0.2400, -0.1112], + device='cuda:0'), grad: tensor([-0.0831, 0.0550, 0.0363, -0.1049, 0.0771, 0.0164, 0.0032], + device='cuda:0') +249 +0.009954748808839675 +changing lr +epoch 3, time 349.42, cls_loss 2.7298 cls_loss_mapping 1.5867 cls_loss_causal 2.0432 re_mapping 0.5943 re_causal 0.6005 /// teacc 77.78 lr 0.00991965 +Epoch 5, weight, value: tensor([[ 0.1081, 0.0900, 0.0151, ..., -0.0868, -0.1245, -0.0882], + [-0.0605, -0.0741, -0.0279, ..., -0.0441, -0.0276, -0.0324], + [-0.0019, -0.0298, 0.0355, ..., -0.0761, -0.1008, -0.0539], + ..., + [-0.0150, -0.0192, -0.0049, ..., -0.0336, -0.0156, -0.0245], + [-0.0533, -0.0477, -0.0658, ..., -0.0337, -0.0129, -0.0378], + [ 0.1000, 0.1785, 0.0972, ..., 0.1097, 0.1032, 0.1183]], + device='cuda:0'), grad: tensor([[-1.2827e-03, -1.2982e-04, -2.0485e-03, ..., -5.8632e-03, + -4.2610e-03, -4.8866e-03], + [ 7.7248e-04, 4.5151e-05, 1.0691e-03, ..., 3.7918e-03, + 3.3112e-03, 3.3703e-03], + [ 1.0872e-04, 1.4193e-05, 1.6677e-04, ..., 4.7517e-04, + 3.9291e-04, 4.0770e-04], + ..., + [ 5.5408e-04, 1.0329e-04, 8.8072e-04, ..., 2.1973e-03, + 1.8730e-03, 1.8682e-03], + [ 1.1146e-05, 1.7807e-06, 1.7256e-05, ..., 4.6402e-05, + 3.9756e-05, 3.9905e-05], + [ 1.4134e-05, 2.3544e-06, 1.9610e-05, ..., 5.8502e-05, + 6.0886e-05, 5.3585e-05]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0686, 0.2019, 0.1705, 0.0255, 0.1121, -0.3466, -0.2412], + device='cuda:0'), grad: tensor([-0.1044, 0.0511, 0.0085, -0.0029, 0.0458, 0.0009, 0.0010], + device='cuda:0') +249 +0.009919647942993149 +changing lr +epoch 4, time 346.66, cls_loss 3.5903 cls_loss_mapping 1.6275 cls_loss_causal 2.0572 re_mapping 0.5120 re_causal 0.5143 /// teacc 35.09 lr 0.00987464 +Epoch 6, weight, value: tensor([[ 0.0966, 0.1068, 0.0187, ..., -0.0904, -0.1378, -0.0916], + [-0.0324, -0.0771, 0.0030, ..., -0.0663, -0.0303, -0.0512], + [ 0.0144, -0.0241, 0.0409, ..., -0.0741, -0.0996, -0.0597], + ..., + [-0.0197, -0.0390, -0.0097, ..., -0.0399, -0.0355, -0.0279], + [-0.0444, -0.0515, -0.0625, ..., -0.0262, 0.0110, -0.0326], + [ 0.0648, 0.2060, 0.0403, ..., 0.1140, 0.0686, 0.1289]], + device='cuda:0'), grad: tensor([[ 3.0975e-03, 6.9201e-05, 4.6806e-03, ..., 4.5848e-04, + 1.0166e-03, 3.1519e-04], + [ 1.7853e-02, 2.6822e-04, 2.4368e-02, ..., 1.7824e-03, + 5.9357e-03, 9.9754e-04], + [-3.5362e-03, -1.0973e-04, -6.5269e-03, ..., -1.0843e-03, + 3.8576e-04, 3.3927e-04], + ..., + [-1.7624e-02, -7.7367e-05, -1.9379e-02, ..., 6.7770e-05, + -6.5308e-03, 1.6078e-05], + [-3.2139e-03, -2.4211e-04, -8.5297e-03, ..., -1.7824e-03, + -1.9932e-03, -2.1458e-03], + [ 1.1911e-03, 2.1294e-05, 1.6680e-03, ..., 1.2493e-04, + 4.1246e-04, 9.4652e-05]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0241, 0.2085, 0.0226, 0.1522, 0.1168, -0.1597, -0.3738], + device='cuda:0'), grad: tensor([ 0.0638, 0.2766, -0.0494, 0.0591, -0.0909, -0.2791, 0.0196], + device='cuda:0') +249 +0.009874639560909117 +changing lr +epoch 5, time 347.35, cls_loss 3.8974 cls_loss_mapping 1.7399 cls_loss_causal 2.0707 re_mapping 0.4239 re_causal 0.4300 /// teacc 49.12 lr 0.00981981 +Epoch 7, weight, value: tensor([[ 0.0807, 0.1011, 0.0078, ..., -0.0864, -0.1360, -0.0869], + [-0.0310, -0.0750, 0.0037, ..., -0.0671, -0.0333, -0.0504], + [ 0.0193, -0.0341, 0.0472, ..., -0.1011, -0.0952, -0.0832], + ..., + [-0.0081, -0.0604, 0.0070, ..., -0.0446, -0.0309, -0.0428], + [-0.0455, -0.0510, -0.0497, ..., -0.0187, 0.0106, -0.0391], + [ 0.0654, 0.2450, 0.0348, ..., 0.1289, 0.0681, 0.1527]], + device='cuda:0'), grad: tensor([[ 2.0275e-03, 4.0388e-04, 5.3253e-03, ..., 4.7417e-03, + 5.2872e-03, 4.8218e-03], + [-2.3365e-03, -5.3227e-05, -5.9280e-03, ..., -1.7691e-03, + -9.7513e-04, -1.3580e-03], + [ 9.0122e-04, 1.6940e-04, 2.2984e-03, ..., 1.9093e-03, + 2.0332e-03, 1.8892e-03], + ..., + [ 1.4031e-04, 2.3878e-04, 8.2350e-04, ..., -4.3373e-03, + -4.6272e-03, -4.6959e-03], + [-1.5945e-03, -9.1267e-04, -4.8599e-03, ..., -2.5234e-03, + -3.9253e-03, -2.6951e-03], + [ 3.6925e-05, 6.7800e-06, 9.6202e-05, ..., 7.4565e-05, + 8.3208e-05, 7.4804e-05]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0282, 0.2684, 0.0592, 0.0523, 0.3811, -0.0482, -0.6935], + device='cuda:0'), grad: tensor([ 0.1420, -0.1486, 0.0656, 0.0565, 0.0802, -0.1981, 0.0025], + device='cuda:0') +249 +0.009819814303479266 +changing lr +epoch 6, time 348.88, cls_loss 2.1611 cls_loss_mapping 1.5626 cls_loss_causal 2.0639 re_mapping 0.4027 re_causal 0.4165 /// teacc 53.80 lr 0.00975528 +Epoch 8, weight, value: tensor([[ 0.0680, 0.0850, -0.0052, ..., -0.0881, -0.1487, -0.0975], + [-0.0428, -0.0737, 0.0020, ..., -0.0586, -0.0247, -0.0390], + [ 0.0264, -0.0334, 0.0473, ..., -0.1011, -0.1059, -0.0839], + ..., + [-0.0076, -0.0591, 0.0107, ..., -0.0418, -0.0316, -0.0460], + [-0.0439, -0.0489, -0.0498, ..., -0.0385, -0.0017, -0.0530], + [ 0.0812, 0.2544, 0.0455, ..., 0.1305, 0.0698, 0.1522]], + device='cuda:0'), grad: tensor([[ 5.3692e-04, 2.7537e-05, 2.0752e-03, ..., -8.0633e-04, + -4.9686e-04, -8.4734e-04], + [-4.8578e-05, 6.2227e-05, -8.9824e-05, ..., 7.8249e-04, + 6.9737e-06, 7.3862e-04], + [ 3.7956e-04, 1.3542e-04, 1.1082e-03, ..., 8.0729e-04, + 7.6675e-04, 7.0953e-04], + ..., + [-6.8426e-04, -2.0921e-05, -2.8839e-03, ..., -1.2426e-03, + -8.3065e-04, -1.3056e-03], + [-5.6648e-04, -3.2020e-04, -1.4095e-03, ..., -2.0278e-04, + -7.4446e-05, 1.4007e-04], + [ 1.3745e-04, 3.2336e-05, 4.5538e-04, ..., 2.6965e-04, + 2.5702e-04, 2.5105e-04]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.1659, 0.0559, 0.0831, -0.0525, 0.2321, 0.0564, -0.5498], + device='cuda:0'), grad: tensor([ 0.0200, 0.0316, 0.0316, 0.0189, -0.0889, -0.0259, 0.0126], + device='cuda:0') +249 +0.009755282581475767 +changing lr +epoch 7, time 347.51, cls_loss 1.8054 cls_loss_mapping 1.4719 cls_loss_causal 1.9194 re_mapping 0.3440 re_causal 0.3578 /// teacc 60.23 lr 0.00968117 +Epoch 9, weight, value: tensor([[ 0.0677, 0.0792, 0.0004, ..., -0.0899, -0.1448, -0.1001], + [-0.0454, -0.0682, -0.0072, ..., -0.0556, -0.0266, -0.0361], + [ 0.0299, -0.0330, 0.0590, ..., -0.0957, -0.0870, -0.0778], + ..., + [-0.0150, -0.0534, 0.0009, ..., -0.0312, -0.0406, -0.0365], + [-0.0395, -0.0464, -0.0477, ..., -0.0495, -0.0057, -0.0608], + [ 0.0828, 0.2522, 0.0470, ..., 0.1284, 0.0701, 0.1498]], + device='cuda:0'), grad: tensor([[ 9.4354e-05, 2.9802e-08, 5.6028e-04, ..., 2.4009e-04, + 2.0373e-04, 1.9014e-04], + [ 3.6687e-05, 0.0000e+00, 8.0585e-05, ..., 9.3341e-05, + 6.4671e-05, 7.3910e-05], + [ 8.9228e-05, 0.0000e+00, 5.1832e-04, ..., 2.2709e-04, + 1.9145e-04, 1.7989e-04], + ..., + [ 1.3673e-04, 0.0000e+00, 8.4925e-04, ..., 3.4785e-04, + 2.9922e-04, 2.7561e-04], + [-4.1175e-04, 0.0000e+00, -2.3327e-03, ..., -1.0481e-03, + -8.7738e-04, -8.2970e-04], + [ 1.2487e-05, -2.9802e-08, 7.4208e-05, ..., 3.1859e-05, + 2.6986e-05, 2.5228e-05]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.2562, 0.0744, 0.0247, -0.0490, 0.3360, -0.0729, -0.5780], + device='cuda:0'), grad: tensor([ 0.0390, -0.0083, 0.0350, 0.0175, 0.0630, -0.1512, 0.0052], + device='cuda:0') +249 +0.009681174353198686 +changing lr +epoch 8, time 344.34, cls_loss 1.1558 cls_loss_mapping 1.3434 cls_loss_causal 1.8095 re_mapping 0.3223 re_causal 0.3378 /// teacc 71.35 lr 0.00959764 +Epoch 10, weight, value: tensor([[ 0.0637, 0.0797, -0.0172, ..., -0.0936, -0.1668, -0.1064], + [-0.0445, -0.0665, -0.0026, ..., -0.0505, -0.0229, -0.0326], + [ 0.0233, -0.0360, 0.0487, ..., -0.1026, -0.0940, -0.0814], + ..., + [-0.0145, -0.0573, 0.0077, ..., -0.0440, -0.0493, -0.0489], + [-0.0382, -0.0424, -0.0457, ..., -0.0357, 0.0062, -0.0461], + [ 0.0888, 0.2514, 0.0565, ..., 0.1306, 0.0868, 0.1515]], + device='cuda:0'), grad: tensor([[-1.7223e-03, 3.1531e-05, -6.3858e-03, ..., -2.5387e-03, + -4.1237e-03, -2.4738e-03], + [ 3.2926e-04, 4.4525e-05, 6.4516e-04, ..., -1.1349e-03, + -1.8263e-03, -1.7786e-03], + [ 6.6805e-04, 3.7193e-05, 2.1992e-03, ..., 1.1168e-03, + 1.1148e-03, 6.0797e-04], + ..., + [-7.1383e-04, -2.6226e-04, -6.6233e-04, ..., -3.1734e-04, + 3.4962e-03, 3.0727e-03], + [ 7.5817e-04, 8.6546e-05, 2.1133e-03, ..., 1.5717e-03, + 4.4703e-04, 1.1426e-04], + [ 1.2231e-04, 9.2536e-06, 3.8886e-04, ..., 2.2089e-04, + 1.8620e-04, 1.0014e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.1999, 0.0502, -0.0396, 0.0345, 0.3870, -0.0554, -0.5850], + device='cuda:0'), grad: tensor([-0.1238, 0.0103, 0.0461, 0.0386, -0.0331, 0.0534, 0.0084], + device='cuda:0') +249 +0.009597638862757255 +changing lr +epoch 9, time 346.82, cls_loss 1.0152 cls_loss_mapping 1.2447 cls_loss_causal 1.7866 re_mapping 0.2933 re_causal 0.2910 /// teacc 71.93 lr 0.00950484 +Epoch 11, weight, value: tensor([[ 6.1647e-02, 7.1985e-02, -6.6718e-03, ..., -8.4501e-02, + -1.5132e-01, -9.8548e-02], + [-4.4163e-02, -6.3831e-02, -2.0746e-05, ..., -5.3221e-02, + -2.4077e-02, -3.4561e-02], + [ 2.2785e-02, -3.1338e-02, 4.7170e-02, ..., -1.0755e-01, + -1.0055e-01, -8.6465e-02], + ..., + [-1.1486e-02, -5.7438e-02, 8.9108e-03, ..., -4.2754e-02, + -4.6038e-02, -4.9284e-02], + [-3.6297e-02, -4.0168e-02, -5.0431e-02, ..., -3.7683e-02, + 1.5160e-04, -4.6720e-02], + [ 8.4497e-02, 2.4673e-01, 4.9078e-02, ..., 1.2770e-01, + 8.1287e-02, 1.4858e-01]], device='cuda:0'), grad: tensor([[-7.1973e-06, -1.4484e-05, -4.7159e-04, ..., -8.3542e-04, + -3.4523e-04, -3.3140e-04], + [ 5.9009e-05, 1.4454e-06, 4.6563e-04, ..., 1.5724e-04, + 2.6941e-04, 1.0693e-04], + [ 7.9334e-05, 3.0696e-06, 6.5899e-04, ..., 2.7585e-04, + 3.8648e-04, 1.6868e-04], + ..., + [-2.1803e-04, 7.0184e-06, -1.3599e-03, ..., 1.3018e-04, + -7.1812e-04, -1.1444e-04], + [ 2.7567e-05, 1.3560e-06, 2.3830e-04, ..., 1.1498e-04, + 1.4412e-04, 6.7711e-05], + [ 1.1414e-05, 3.4273e-07, 9.2506e-05, ..., 3.5554e-05, + 5.5403e-05, 2.3544e-05]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.3634, -0.0167, -0.1066, 0.0363, 0.4038, -0.0840, -0.6044], + device='cuda:0'), grad: tensor([-0.0597, 0.0193, 0.0304, 0.0157, -0.0216, 0.0118, 0.0040], + device='cuda:0') +249 +0.009504844339512096 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 346.95, cls_loss 0.7606 cls_loss_mapping 1.2027 cls_loss_causal 1.7721 re_mapping 0.2745 re_causal 0.2954 /// teacc 83.04 lr 0.00940298 +Epoch 12, weight, value: tensor([[ 0.0608, 0.0733, -0.0066, ..., -0.0808, -0.1450, -0.0952], + [-0.0455, -0.0656, -0.0015, ..., -0.0516, -0.0244, -0.0344], + [ 0.0261, -0.0287, 0.0484, ..., -0.1044, -0.0978, -0.0841], + ..., + [-0.0148, -0.0566, 0.0036, ..., -0.0439, -0.0561, -0.0477], + [-0.0346, -0.0400, -0.0446, ..., -0.0398, 0.0050, -0.0485], + [ 0.0834, 0.2440, 0.0476, ..., 0.1261, 0.0795, 0.1467]], + device='cuda:0'), grad: tensor([[ 2.9125e-03, 2.1839e-03, 1.1055e-02, ..., 5.8022e-03, + 1.0353e-02, 4.9248e-03], + [ 6.7568e-04, 3.6812e-04, 2.3785e-03, ..., 8.6689e-04, + 1.7233e-03, 8.5974e-04], + [ 2.8276e-04, 1.4174e-04, 1.0948e-03, ..., 5.7745e-04, + 9.5844e-04, 4.5371e-04], + ..., + [-4.9820e-03, -3.3283e-03, -1.8692e-02, ..., -9.2621e-03, + -1.6571e-02, -7.9346e-03], + [ 2.0456e-04, 1.1301e-04, 7.2622e-04, ..., 2.7800e-04, + 5.3978e-04, 2.6870e-04], + [ 5.4210e-05, 3.0264e-05, 2.0385e-04, ..., 9.9778e-05, + 1.7345e-04, 8.3029e-05]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.2281, 0.0016, -0.0989, 0.1680, 0.3630, -0.0823, -0.5875], + device='cuda:0'), grad: tensor([ 0.1564, 0.0358, 0.0146, 0.0443, -0.2646, 0.0108, 0.0028], + device='cuda:0') +249 +0.009402977659283692 +changing lr +epoch 11, time 348.26, cls_loss 0.8984 cls_loss_mapping 1.2257 cls_loss_causal 1.7599 re_mapping 0.2440 re_causal 0.2828 /// teacc 64.91 lr 0.00929224 +Epoch 13, weight, value: tensor([[ 0.0547, 0.0676, -0.0145, ..., -0.0803, -0.1415, -0.0967], + [-0.0441, -0.0585, -0.0022, ..., -0.0485, -0.0291, -0.0319], + [ 0.0192, -0.0320, 0.0357, ..., -0.1065, -0.1117, -0.0865], + ..., + [-0.0133, -0.0545, 0.0126, ..., -0.0470, -0.0490, -0.0475], + [-0.0264, -0.0354, -0.0327, ..., -0.0376, 0.0110, -0.0460], + [ 0.0860, 0.2429, 0.0513, ..., 0.1244, 0.0811, 0.1448]], + device='cuda:0'), grad: tensor([[ 8.0948e-03, 4.7188e-03, 2.4841e-02, ..., 6.6223e-03, + 2.3666e-02, 5.8098e-03], + [-1.4984e-02, -7.6065e-03, -5.0842e-02, ..., -1.6510e-02, + -5.0995e-02, -1.5388e-02], + [ 3.6883e-04, 9.5189e-05, 1.7767e-03, ..., 6.6376e-04, + 1.8787e-03, 6.5994e-04], + ..., + [ 4.9591e-03, 2.2087e-03, 1.6663e-02, ..., 5.1651e-03, + 1.6617e-02, 4.8523e-03], + [-4.7541e-04, 1.4246e-04, -1.5488e-03, ..., 4.7541e-04, + -9.6083e-04, 4.7874e-04], + [ 5.4026e-04, 1.0842e-04, 2.3575e-03, ..., 7.8869e-04, + 2.4357e-03, 7.8535e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.2046, 0.0119, -0.1103, 0.1370, 0.3917, -0.0697, -0.5732], + device='cuda:0'), grad: tensor([ 0.1096, -0.2756, 0.0134, 0.0499, 0.0839, 0.0035, 0.0153], + device='cuda:0') +249 +0.009292243968009333 +changing lr +epoch 12, time 346.66, cls_loss 0.8350 cls_loss_mapping 1.2619 cls_loss_causal 1.8637 re_mapping 0.2095 re_causal 0.2569 /// teacc 75.44 lr 0.00917287 +Epoch 14, weight, value: tensor([[ 0.0415, 0.0553, -0.0194, ..., -0.0846, -0.1425, -0.1021], + [-0.0380, -0.0485, -0.0086, ..., -0.0332, -0.0208, -0.0135], + [ 0.0273, -0.0239, 0.0513, ..., -0.1055, -0.0992, -0.0865], + ..., + [-0.0156, -0.0573, 0.0202, ..., -0.0573, -0.0553, -0.0609], + [-0.0235, -0.0368, -0.0384, ..., -0.0439, -0.0034, -0.0513], + [ 0.0841, 0.2409, 0.0416, ..., 0.1204, 0.0700, 0.1413]], + device='cuda:0'), grad: tensor([[ 5.5084e-03, 3.4237e-03, 2.1042e-02, ..., 5.4054e-03, + 2.2049e-02, 3.4885e-03], + [ 2.6642e-02, 8.3694e-03, 8.7708e-02, ..., 2.2125e-02, + 9.0027e-02, 1.5335e-02], + [-2.5024e-03, -1.3599e-03, -6.5117e-03, ..., -8.6164e-04, + -6.2943e-03, -9.2697e-04], + ..., + [-2.9495e-02, -1.0429e-02, -1.0193e-01, ..., -2.6520e-02, + -1.0535e-01, -1.7776e-02], + [-4.4012e-04, -1.0741e-04, -1.1673e-03, ..., -3.8886e-04, + -1.3380e-03, -3.0255e-04], + [ 2.0817e-05, 5.5730e-06, 5.6982e-05, ..., 1.8135e-05, + 6.4135e-05, 1.3933e-05]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.3738, -0.1402, -0.0957, 0.1669, 0.4431, -0.1405, -0.6153], + device='cuda:0'), grad: tensor([ 7.6782e-02, 2.4097e-01, -1.0109e-02, 2.5806e-03, -3.0664e-01, + -3.6335e-03, 1.7524e-04], device='cuda:0') +249 +0.009172866268606516 +changing lr +epoch 13, time 350.62, cls_loss 0.6228 cls_loss_mapping 1.0631 cls_loss_causal 1.7189 re_mapping 0.2041 re_causal 0.2671 /// teacc 82.46 lr 0.00904508 +Epoch 15, weight, value: tensor([[ 0.0415, 0.0524, -0.0151, ..., -0.0856, -0.1345, -0.1026], + [-0.0382, -0.0479, -0.0106, ..., -0.0319, -0.0248, -0.0121], + [ 0.0226, -0.0257, 0.0433, ..., -0.1051, -0.1057, -0.0868], + ..., + [-0.0088, -0.0529, 0.0357, ..., -0.0457, -0.0336, -0.0496], + [-0.0274, -0.0357, -0.0453, ..., -0.0479, -0.0161, -0.0550], + [ 0.0828, 0.2384, 0.0393, ..., 0.1179, 0.0668, 0.1387]], + device='cuda:0'), grad: tensor([[ 1.0033e-03, 3.1304e-04, 4.0283e-03, ..., 3.7632e-03, + 6.0463e-03, 3.6373e-03], + [ 4.2419e-03, 8.4877e-04, 2.1423e-02, ..., 8.8348e-03, + 1.9821e-02, 5.6648e-03], + [ 1.6034e-04, 7.2122e-05, 6.4850e-04, ..., 5.4169e-04, + 8.7976e-04, 5.1785e-04], + ..., + [-5.4092e-03, -7.6914e-04, -2.6001e-02, ..., -1.4404e-02, + -2.8580e-02, -1.1177e-02], + [-8.4305e-04, -7.4387e-04, -3.4904e-03, ..., -1.8578e-03, + -3.1853e-03, -1.6642e-03], + [ 1.1188e-04, 4.2140e-05, 4.5061e-04, ..., 4.0054e-04, + 6.4659e-04, 3.8576e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.3647, -0.0599, -0.0534, 0.0527, 0.4296, -0.1354, -0.6058], + device='cuda:0'), grad: tensor([ 0.0326, 0.1567, 0.0054, 0.0238, -0.1915, -0.0308, 0.0037], + device='cuda:0') +249 +0.00904508497187474 +changing lr +epoch 14, time 345.66, cls_loss 0.4806 cls_loss_mapping 0.9650 cls_loss_causal 1.6150 re_mapping 0.1894 re_causal 0.2416 /// teacc 77.19 lr 0.00890916 +Epoch 16, weight, value: tensor([[ 0.0310, 0.0436, -0.0326, ..., -0.0893, -0.1463, -0.1070], + [-0.0336, -0.0450, -0.0082, ..., -0.0312, -0.0237, -0.0113], + [ 0.0246, -0.0232, 0.0485, ..., -0.1049, -0.0998, -0.0867], + ..., + [-0.0121, -0.0519, 0.0299, ..., -0.0440, -0.0457, -0.0475], + [-0.0270, -0.0393, -0.0407, ..., -0.0581, -0.0167, -0.0640], + [ 0.0877, 0.2420, 0.0476, ..., 0.1297, 0.0832, 0.1521]], + device='cuda:0'), grad: tensor([[-2.5749e-03, -2.3155e-03, -2.6566e-02, ..., -1.3832e-02, + -3.0853e-02, -1.3977e-02], + [ 1.1644e-03, 1.1311e-03, 1.2085e-02, ..., 7.2594e-03, + 1.5259e-02, 7.4463e-03], + [ 1.4566e-05, 1.0222e-05, 8.0287e-05, ..., 4.5151e-05, + 8.7559e-05, 4.4107e-05], + ..., + [ 1.1015e-03, 9.5701e-04, 1.2695e-02, ..., 5.4970e-03, + 1.3535e-02, 5.4626e-03], + [ 1.5274e-06, 1.0803e-06, 7.8306e-06, ..., 5.0291e-06, + 8.9854e-06, 4.9099e-06], + [ 1.0610e-04, 7.4029e-05, 5.2881e-04, ..., 3.1948e-04, + 5.9080e-04, 3.1304e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.3305, -0.0721, -0.0274, 0.0394, 0.4979, -0.1767, -0.5990], + device='cuda:0'), grad: tensor([-2.5024e-01, 1.2463e-01, 6.4611e-04, 1.0529e-02, 1.1017e-01, + 6.5565e-05, 4.2992e-03], device='cuda:0') +249 +0.008909157412340152 +changing lr +epoch 15, time 350.72, cls_loss 0.4993 cls_loss_mapping 0.9868 cls_loss_causal 1.6746 re_mapping 0.1866 re_causal 0.2383 /// teacc 80.70 lr 0.00876536 +Epoch 17, weight, value: tensor([[ 0.0314, 0.0435, -0.0261, ..., -0.0824, -0.1354, -0.1003], + [-0.0328, -0.0435, -0.0109, ..., -0.0273, -0.0238, -0.0068], + [ 0.0247, -0.0233, 0.0477, ..., -0.1042, -0.1016, -0.0861], + ..., + [-0.0157, -0.0528, 0.0222, ..., -0.0466, -0.0491, -0.0507], + [-0.0192, -0.0351, -0.0167, ..., -0.0552, 0.0035, -0.0613], + [ 0.0835, 0.2381, 0.0352, ..., 0.1240, 0.0684, 0.1463]], + device='cuda:0'), grad: tensor([[-2.4104e-04, -2.0754e-04, -8.3399e-04, ..., -1.1355e-04, + -6.2180e-04, -2.6512e-04], + [ 2.5183e-05, 2.0832e-05, 1.2672e-04, ..., 7.1526e-05, + 1.3828e-04, 7.4565e-05], + [ 5.3018e-05, 3.4392e-05, 4.8375e-04, ..., 4.2129e-04, + 6.4516e-04, 3.6454e-04], + ..., + [ 1.0633e-04, 1.0949e-04, -1.2743e-04, ..., -6.1083e-04, + -5.7030e-04, -3.9434e-04], + [ 7.9200e-06, 6.3106e-06, 4.0919e-05, ..., 2.1562e-05, + 4.3303e-05, 2.2501e-05], + [ 1.1288e-05, 8.6278e-06, 6.8188e-05, ..., 4.3750e-05, + 7.8440e-05, 4.2111e-05]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.4034, -0.0693, -0.0067, -0.0020, 0.4321, -0.1292, -0.6354], + device='cuda:0'), grad: tensor([-0.0068, 0.0015, 0.0075, 0.0033, -0.0069, 0.0005, 0.0009], + device='cuda:0') +249 +0.00876535733001806 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 349.48, cls_loss 0.4967 cls_loss_mapping 0.9364 cls_loss_causal 1.6148 re_mapping 0.1727 re_causal 0.2309 /// teacc 86.55 lr 0.00861397 +Epoch 18, weight, value: tensor([[ 0.0292, 0.0414, -0.0274, ..., -0.0836, -0.1345, -0.1007], + [-0.0355, -0.0440, -0.0144, ..., -0.0241, -0.0254, -0.0043], + [ 0.0262, -0.0213, 0.0525, ..., -0.0980, -0.0898, -0.0800], + ..., + [-0.0131, -0.0515, 0.0331, ..., -0.0464, -0.0410, -0.0511], + [-0.0188, -0.0325, -0.0242, ..., -0.0596, -0.0079, -0.0654], + [ 0.0825, 0.2355, 0.0323, ..., 0.1215, 0.0647, 0.1435]], + device='cuda:0'), grad: tensor([[-1.1539e-04, 5.9366e-04, -4.9591e-03, ..., 2.3193e-03, + -3.9482e-03, 1.3437e-03], + [ 1.4591e-04, 1.9825e-04, 1.5459e-03, ..., 2.2554e-04, + 1.3132e-03, 2.3484e-04], + [ 6.3598e-05, 2.2009e-05, 8.8978e-04, ..., 1.7858e-04, + 8.3733e-04, 1.6987e-04], + ..., + [-5.5456e-04, -8.7023e-04, -4.5128e-03, ..., -3.4504e-03, + -4.6577e-03, -2.6188e-03], + [ 2.9516e-04, 2.9728e-05, 4.5509e-03, ..., 4.1389e-04, + 4.1580e-03, 5.2357e-04], + [ 1.5050e-05, 3.7774e-06, 2.1791e-04, ..., 3.6508e-05, + 2.0361e-04, 3.6627e-05]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.4269, 0.0323, -0.0444, -0.0081, 0.4233, -0.1873, -0.6500], + device='cuda:0'), grad: tensor([-0.0424, 0.0184, 0.0097, 0.0242, -0.0605, 0.0482, 0.0024], + device='cuda:0') +249 +0.008613974319136962 +changing lr +epoch 17, time 349.89, cls_loss 0.4311 cls_loss_mapping 0.8407 cls_loss_causal 1.5558 re_mapping 0.1541 re_causal 0.2379 /// teacc 77.78 lr 0.00845531 +Epoch 19, weight, value: tensor([[ 0.0282, 0.0388, -0.0318, ..., -0.0847, -0.1366, -0.1021], + [-0.0396, -0.0462, -0.0243, ..., -0.0225, -0.0308, -0.0026], + [ 0.0270, -0.0203, 0.0469, ..., -0.1002, -0.0983, -0.0844], + ..., + [-0.0153, -0.0498, 0.0313, ..., -0.0474, -0.0472, -0.0527], + [-0.0148, -0.0298, -0.0123, ..., -0.0554, 0.0059, -0.0606], + [ 0.0825, 0.2338, 0.0333, ..., 0.1194, 0.0642, 0.1420]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.8238e-06, ..., 3.3155e-06, + 5.3868e-06, 3.4422e-06], + [ 0.0000e+00, 0.0000e+00, -5.2392e-05, ..., -6.1572e-05, + -1.0008e-04, -6.3956e-05], + [-0.0000e+00, -0.0000e+00, 4.7684e-07, ..., 5.7369e-07, + 9.2387e-07, 5.9605e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.4107e-05, ..., 5.1796e-05, + 8.4221e-05, 5.3823e-05], + [ 0.0000e+00, 0.0000e+00, 2.0340e-06, ..., 2.3842e-06, + 3.8743e-06, 2.4736e-06], + [ 0.0000e+00, 0.0000e+00, 9.5367e-07, ..., 1.1250e-06, + 1.8254e-06, 1.1697e-06]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.3902, -0.0056, -0.0767, -0.0025, 0.4757, -0.1395, -0.6489], + device='cuda:0'), grad: tensor([ 7.1049e-05, -1.3189e-03, 1.2085e-05, 4.8608e-05, 1.1101e-03, + 5.1111e-05, 2.4080e-05], device='cuda:0') +249 +0.008455313244934327 +changing lr +epoch 18, time 349.75, cls_loss 0.2844 cls_loss_mapping 0.8367 cls_loss_causal 1.5332 re_mapping 0.1430 re_causal 0.2201 /// teacc 86.55 lr 0.00828969 +Epoch 20, weight, value: tensor([[ 0.0257, 0.0360, -0.0373, ..., -0.0822, -0.1377, -0.0995], + [-0.0379, -0.0446, -0.0158, ..., -0.0208, -0.0206, -0.0008], + [ 0.0269, -0.0205, 0.0475, ..., -0.1011, -0.0969, -0.0853], + ..., + [-0.0122, -0.0477, 0.0364, ..., -0.0440, -0.0426, -0.0496], + [-0.0130, -0.0280, -0.0092, ..., -0.0557, 0.0070, -0.0598], + [ 0.0787, 0.2307, 0.0253, ..., 0.1169, 0.0551, 0.1387]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.0565e-04, ..., 6.5416e-06, + 1.8394e-04, 6.3702e-06], + [ 7.4506e-09, 7.4506e-09, -9.8419e-04, ..., -1.6794e-05, + -5.8889e-04, -1.6853e-05], + [ 7.4506e-09, 7.4506e-09, 2.3067e-05, ..., 1.3933e-06, + 1.4633e-05, 1.2591e-06], + ..., + [ 1.4901e-08, 7.4506e-09, 6.2418e-04, ..., 1.3180e-05, + 3.7551e-04, 1.2860e-05], + [-2.9802e-08, -2.2352e-08, 1.2152e-05, ..., 1.8477e-06, + 8.6427e-06, 1.6168e-06], + [ 0.0000e+00, 0.0000e+00, 5.0813e-06, ..., 1.1027e-06, + 3.8818e-06, 9.6112e-07]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.4344, 0.0165, -0.0767, -0.0469, 0.4770, -0.1534, -0.6582], + device='cuda:0'), grad: tensor([ 0.0161, -0.0517, 0.0012, 0.0008, 0.0328, 0.0006, 0.0003], + device='cuda:0') +249 +0.008289693629698565 +changing lr +epoch 19, time 345.25, cls_loss 0.4219 cls_loss_mapping 0.7627 cls_loss_causal 1.4024 re_mapping 0.1221 re_causal 0.1921 /// teacc 85.38 lr 0.00811745 +Epoch 21, weight, value: tensor([[ 0.0246, 0.0335, -0.0398, ..., -0.0794, -0.1339, -0.0970], + [-0.0375, -0.0443, -0.0158, ..., -0.0173, -0.0177, 0.0023], + [ 0.0239, -0.0216, 0.0428, ..., -0.1032, -0.1000, -0.0871], + ..., + [-0.0065, -0.0431, 0.0462, ..., -0.0464, -0.0408, -0.0515], + [-0.0135, -0.0270, -0.0098, ..., -0.0579, 0.0038, -0.0615], + [ 0.0774, 0.2280, 0.0237, ..., 0.1151, 0.0532, 0.1367]], + device='cuda:0'), grad: tensor([[ 3.1769e-05, -1.3721e-04, -6.9380e-05, ..., 1.2875e-04, + -1.4472e-04, 1.0723e-04], + [-6.3133e-04, -3.1090e-04, -3.1681e-03, ..., -6.6280e-04, + -3.2272e-03, -6.7329e-04], + [ 1.6415e-04, 7.4446e-05, 1.1692e-03, ..., 2.9182e-04, + 1.1387e-03, 2.0289e-04], + ..., + [ 2.4462e-04, 2.8396e-04, 3.3855e-04, ..., -3.3498e-04, + 5.1117e-04, -5.7593e-06], + [ 1.1869e-05, 5.7518e-06, 7.9215e-05, ..., 1.9401e-05, + 7.8201e-05, 1.4484e-05], + [ 2.7493e-06, 1.6987e-06, 2.1398e-05, ..., 7.3612e-06, + 2.2352e-05, 5.3942e-06]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.3833, 0.0472, -0.0680, -0.0074, 0.4733, -0.1728, -0.6628], + device='cuda:0'), grad: tensor([-0.0019, -0.0153, 0.0072, 0.0116, -0.0022, 0.0005, 0.0001], + device='cuda:0') +249 +0.00811744900929367 +changing lr +epoch 20, time 346.79, cls_loss 0.2886 cls_loss_mapping 0.6949 cls_loss_causal 1.3418 re_mapping 0.1059 re_causal 0.1685 /// teacc 84.80 lr 0.00793893 +Epoch 22, weight, value: tensor([[ 0.0233, 0.0313, -0.0416, ..., -0.0790, -0.1330, -0.0969], + [-0.0378, -0.0435, -0.0128, ..., -0.0185, -0.0118, 0.0010], + [ 0.0220, -0.0224, 0.0384, ..., -0.1022, -0.1030, -0.0863], + ..., + [-0.0045, -0.0410, 0.0448, ..., -0.0452, -0.0463, -0.0504], + [-0.0108, -0.0260, -0.0016, ..., -0.0557, 0.0125, -0.0590], + [ 0.0770, 0.2260, 0.0238, ..., 0.1138, 0.0528, 0.1352]], + device='cuda:0'), grad: tensor([[-1.5383e-03, -1.8444e-03, -2.1469e-02, ..., -1.0399e-02, + -2.5879e-02, -9.4986e-03], + [ 2.7609e-04, 3.4833e-04, 4.5853e-03, ..., 2.2602e-03, + 5.6686e-03, 2.0962e-03], + [ 7.6443e-06, 6.0983e-06, 1.0163e-04, ..., 5.5999e-05, + 1.0902e-04, 4.1693e-05], + ..., + [ 1.1730e-03, 1.4095e-03, 1.6357e-02, ..., 7.8964e-03, + 1.9699e-02, 7.2136e-03], + [ 3.9965e-05, 3.9756e-05, 2.0468e-04, ..., 8.4817e-05, + 1.8144e-04, 6.2823e-05], + [ 1.2055e-05, 1.1958e-05, 6.2764e-05, ..., 2.6688e-05, + 5.6267e-05, 1.9848e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.3787, 0.0097, -0.0376, 0.0187, 0.4699, -0.1906, -0.6559], + device='cuda:0'), grad: tensor([-0.2013, 0.0428, 0.0012, 0.0016, 0.1532, 0.0020, 0.0006], + device='cuda:0') +249 +0.007938926261462368 +changing lr +epoch 21, time 346.47, cls_loss 0.2398 cls_loss_mapping 0.5935 cls_loss_causal 1.1991 re_mapping 0.0881 re_causal 0.1494 /// teacc 82.46 lr 0.00775448 +Epoch 23, weight, value: tensor([[ 0.0253, 0.0321, -0.0354, ..., -0.0819, -0.1262, -0.0989], + [-0.0420, -0.0463, -0.0222, ..., -0.0196, -0.0175, -0.0004], + [ 0.0206, -0.0235, 0.0366, ..., -0.1000, -0.1013, -0.0838], + ..., + [-0.0007, -0.0374, 0.0553, ..., -0.0389, -0.0385, -0.0454], + [-0.0104, -0.0250, -0.0011, ..., -0.0555, 0.0127, -0.0587], + [ 0.0764, 0.2240, 0.0234, ..., 0.1123, 0.0518, 0.1337]], + device='cuda:0'), grad: tensor([[-5.4389e-05, -7.5281e-05, 5.6076e-04, ..., 1.3840e-04, + 6.6090e-04, 8.7142e-05], + [ 2.1443e-05, 2.7061e-05, 3.8767e-04, ..., 8.2433e-05, + 3.6097e-04, 7.1824e-05], + [ 5.1558e-06, 3.9563e-06, 4.6968e-05, ..., 9.5740e-06, + 4.2051e-05, 7.6294e-06], + ..., + [ 2.5108e-05, 4.2379e-05, -1.0281e-03, ..., -2.3723e-04, + -1.0948e-03, -1.7190e-04], + [ 7.1526e-07, 4.7684e-07, 8.8960e-06, ..., 1.8738e-06, + 8.3223e-06, 1.4454e-06], + [ 5.3644e-07, 4.7684e-07, 3.5278e-06, ..., 6.8918e-07, + 2.9393e-06, 5.9232e-07]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.3388, -0.0143, -0.0064, 0.0110, 0.5262, -0.2086, -0.6536], + device='cuda:0'), grad: tensor([ 4.0855e-03, 3.0880e-03, 3.2306e-04, 1.3936e-04, -7.7248e-03, + 6.1810e-05, 2.4408e-05], device='cuda:0') +249 +0.007754484907260515 +changing lr +epoch 22, time 343.89, cls_loss 0.2083 cls_loss_mapping 0.5953 cls_loss_causal 1.3084 re_mapping 0.0874 re_causal 0.1513 /// teacc 80.12 lr 0.00756450 +Epoch 24, weight, value: tensor([[ 0.0259, 0.0315, -0.0369, ..., -0.0825, -0.1337, -0.1016], + [-0.0418, -0.0450, -0.0224, ..., -0.0180, -0.0164, 0.0007], + [ 0.0223, -0.0211, 0.0504, ..., -0.0949, -0.0819, -0.0779], + ..., + [-0.0006, -0.0368, 0.0494, ..., -0.0380, -0.0434, -0.0445], + [-0.0110, -0.0256, -0.0048, ..., -0.0579, 0.0065, -0.0612], + [ 0.0736, 0.2195, 0.0208, ..., 0.1090, 0.0525, 0.1319]], + device='cuda:0'), grad: tensor([[-9.3132e-08, -1.9372e-07, -6.5938e-07, ..., -2.3469e-07, + -7.3388e-07, -2.3097e-07], + [ 7.4506e-09, 1.4901e-08, 4.8429e-08, ..., 1.8626e-08, + 5.5879e-08, 1.8626e-08], + [ 1.1176e-08, 7.4506e-09, 5.2154e-08, ..., 2.2352e-08, + 7.0781e-08, 2.6077e-08], + ..., + [ 8.5682e-08, 1.7509e-07, 6.0350e-07, ..., 2.1607e-07, + 6.7055e-07, 2.1234e-07], + [-1.1176e-08, -7.4506e-09, -5.2154e-08, ..., -1.8626e-08, + -6.7055e-08, -2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.3312, -0.0065, 0.0246, 0.0131, 0.5065, -0.2245, -0.6513], + device='cuda:0'), grad: tensor([-2.1271e-06, 1.5646e-07, 1.3411e-07, 0.0000e+00, 1.9483e-06, + -1.1921e-07, 0.0000e+00], device='cuda:0') +249 +0.007564496387029534 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 350.19, cls_loss 0.2486 cls_loss_mapping 0.5770 cls_loss_causal 1.2395 re_mapping 0.0702 re_causal 0.1356 /// teacc 87.72 lr 0.00736934 +Epoch 25, weight, value: tensor([[ 0.0267, 0.0314, -0.0344, ..., -0.0778, -0.1292, -0.0977], + [-0.0408, -0.0440, -0.0236, ..., -0.0215, -0.0203, -0.0027], + [ 0.0218, -0.0228, 0.0440, ..., -0.0951, -0.0880, -0.0785], + ..., + [-0.0041, -0.0360, 0.0495, ..., -0.0334, -0.0376, -0.0394], + [-0.0122, -0.0253, -0.0072, ..., -0.0591, 0.0014, -0.0627], + [ 0.0729, 0.2174, 0.0201, ..., 0.1076, 0.0512, 0.1303]], + device='cuda:0'), grad: tensor([[ 2.5064e-05, 6.3181e-06, 5.1546e-04, ..., 1.9991e-04, + 5.3024e-04, 2.0480e-04], + [ 5.9462e-04, 3.4310e-06, 1.1429e-02, ..., 1.4524e-03, + 8.9111e-03, 1.6298e-03], + [ 1.6734e-05, 6.5640e-06, 2.9111e-04, ..., 1.4937e-04, + 3.1781e-04, 1.5056e-04], + ..., + [-6.4468e-04, -1.8567e-05, -1.2383e-02, ..., -1.8587e-03, + -9.9106e-03, -2.0428e-03], + [ 3.2559e-06, 7.9721e-07, 5.6028e-05, ..., 2.0042e-05, + 5.3853e-05, 2.0519e-05], + [ 1.3672e-06, 3.1665e-07, 2.4319e-05, ..., 8.2403e-06, + 2.2992e-05, 8.4564e-06]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.3551, -0.0608, 0.0083, -0.0027, 0.5596, -0.2145, -0.6520], + device='cuda:0'), grad: tensor([ 0.0057, 0.1221, 0.0036, 0.0008, -0.1331, 0.0007, 0.0003], + device='cuda:0') +249 +0.007369343312364995 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 349.95, cls_loss 0.1266 cls_loss_mapping 0.4742 cls_loss_causal 1.1183 re_mapping 0.0595 re_causal 0.1259 /// teacc 88.30 lr 0.00716942 +Epoch 26, weight, value: tensor([[ 2.6065e-02, 3.1333e-02, -3.2591e-02, ..., -7.8665e-02, + -1.2607e-01, -9.7870e-02], + [-4.1028e-02, -4.3888e-02, -2.8106e-02, ..., -2.2541e-02, + -2.3828e-02, -3.8757e-03], + [ 2.0624e-02, -2.3310e-02, 3.7576e-02, ..., -9.6356e-02, + -9.4617e-02, -8.0039e-02], + ..., + [-3.9162e-03, -3.5971e-02, 4.8947e-02, ..., -3.0996e-02, + -3.8739e-02, -3.7762e-02], + [-1.0823e-02, -2.4368e-02, -2.1525e-04, ..., -5.7820e-02, + 8.0542e-03, -6.0702e-02], + [ 7.2154e-02, 2.1539e-01, 1.9567e-02, ..., 1.0641e-01, + 5.0375e-02, 1.2895e-01]], device='cuda:0'), grad: tensor([[ 8.8072e-04, 5.1594e-04, 4.9973e-03, ..., 1.3981e-03, + 6.2180e-03, 2.2316e-03], + [-1.3784e-07, -5.2899e-07, -1.6801e-06, ..., 4.0233e-07, + -1.3262e-06, 4.5076e-07], + [ 1.2982e-04, 7.1049e-05, 7.2861e-04, ..., 1.9443e-04, + 9.0170e-04, 3.2187e-04], + ..., + [-1.8549e-03, -1.0757e-03, -1.0483e-02, ..., -2.9087e-03, + -1.3023e-02, -4.6730e-03], + [ 6.5744e-05, 3.9637e-05, 3.7694e-04, ..., 1.0800e-04, + 4.7064e-04, 1.6940e-04], + [ 1.5274e-07, 1.1176e-07, 1.0990e-06, ..., 3.8370e-07, + 1.4305e-06, 5.2527e-07]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.3469, -0.1051, -0.0202, 0.0171, 0.5930, -0.1889, -0.6496], + device='cuda:0'), grad: tensor([ 9.0256e-03, -6.0312e-06, 1.2798e-03, 7.8125e-03, -1.8814e-02, + 6.9237e-04, 2.4401e-06], device='cuda:0') +249 +0.0071694186955877925 +changing lr +epoch 25, time 346.90, cls_loss 0.1057 cls_loss_mapping 0.4313 cls_loss_causal 1.1527 re_mapping 0.0601 re_causal 0.1291 /// teacc 87.72 lr 0.00696513 +Epoch 27, weight, value: tensor([[ 0.0240, 0.0297, -0.0393, ..., -0.0773, -0.1310, -0.0966], + [-0.0409, -0.0436, -0.0290, ..., -0.0234, -0.0260, -0.0050], + [ 0.0212, -0.0224, 0.0399, ..., -0.0959, -0.0926, -0.0799], + ..., + [-0.0022, -0.0348, 0.0558, ..., -0.0309, -0.0303, -0.0371], + [-0.0113, -0.0246, -0.0006, ..., -0.0579, 0.0076, -0.0603], + [ 0.0715, 0.2134, 0.0192, ..., 0.1053, 0.0496, 0.1276]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 6.7055e-08, 8.5682e-07, ..., 2.0489e-07, + 6.3702e-07, 2.1979e-07], + [-2.1234e-07, -9.6112e-07, -1.2949e-05, ..., -8.9407e-08, + -6.5267e-06, -3.5390e-07], + [ 9.3132e-08, 9.6858e-08, 9.4250e-07, ..., 6.2585e-07, + 1.1213e-06, 6.5938e-07], + ..., + [ 2.7940e-07, 9.5367e-07, 1.2428e-05, ..., 7.0408e-07, + 6.9104e-06, 9.6858e-07], + [ 1.2256e-06, 1.0133e-06, 8.2999e-06, ..., 8.7470e-06, + 1.3277e-05, 9.1121e-06], + [ 1.0058e-07, 8.1956e-08, 6.8918e-07, ..., 7.0781e-07, + 1.0803e-06, 7.3761e-07]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.3509, -0.1067, -0.0124, 0.0109, 0.5675, -0.1701, -0.6468], + device='cuda:0'), grad: tensor([ 1.0535e-05, -1.7989e-04, 9.7156e-06, -8.5473e-05, 1.6975e-04, + 6.9559e-05, 5.8375e-06], device='cuda:0') +249 +0.0069651251582696205 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 350.28, cls_loss 0.1538 cls_loss_mapping 0.4250 cls_loss_causal 1.0776 re_mapping 0.0632 re_causal 0.1255 /// teacc 90.64 lr 0.00675687 +Epoch 28, weight, value: tensor([[ 0.0238, 0.0302, -0.0388, ..., -0.0751, -0.1280, -0.0947], + [-0.0399, -0.0428, -0.0256, ..., -0.0245, -0.0237, -0.0060], + [ 0.0207, -0.0227, 0.0378, ..., -0.0960, -0.0952, -0.0803], + ..., + [-0.0025, -0.0350, 0.0536, ..., -0.0307, -0.0333, -0.0368], + [-0.0113, -0.0245, -0.0016, ..., -0.0580, 0.0047, -0.0603], + [ 0.0711, 0.2119, 0.0192, ..., 0.1043, 0.0492, 0.1265]], + device='cuda:0'), grad: tensor([[ 4.2057e-04, 4.4632e-04, 1.9722e-03, ..., 1.6460e-03, + 2.2316e-03, 1.5688e-03], + [-2.1400e-03, -2.7809e-03, -1.2062e-02, ..., -2.9259e-03, + -8.4991e-03, -3.2768e-03], + [ 1.3411e-04, 1.1194e-04, 5.8365e-04, ..., 5.7316e-04, + 7.2050e-04, 5.3406e-04], + ..., + [ 1.4725e-03, 2.1286e-03, 9.0027e-03, ..., 2.7800e-04, + 4.9706e-03, 7.7105e-04], + [ 3.4392e-05, 3.2216e-05, 1.6570e-04, ..., 1.0234e-04, + 1.6093e-04, 9.8109e-05], + [ 1.6198e-05, 1.4491e-05, 7.2718e-05, ..., 6.4850e-05, + 8.5056e-05, 6.0856e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.3468, -0.1148, -0.0048, 0.0030, 0.5668, -0.1701, -0.6335], + device='cuda:0'), grad: tensor([ 0.0176, -0.1155, 0.0043, 0.0018, 0.0899, 0.0013, 0.0006], + device='cuda:0') +249 +0.006756874120406716 +changing lr +epoch 27, time 347.40, cls_loss 0.1100 cls_loss_mapping 0.3891 cls_loss_causal 1.0223 re_mapping 0.0498 re_causal 0.1108 /// teacc 86.55 lr 0.00654508 +Epoch 29, weight, value: tensor([[ 0.0251, 0.0313, -0.0354, ..., -0.0745, -0.1234, -0.0937], + [-0.0389, -0.0416, -0.0225, ..., -0.0243, -0.0201, -0.0057], + [ 0.0203, -0.0227, 0.0378, ..., -0.0959, -0.0940, -0.0799], + ..., + [-0.0031, -0.0347, 0.0522, ..., -0.0271, -0.0322, -0.0340], + [-0.0118, -0.0255, -0.0050, ..., -0.0594, -0.0014, -0.0617], + [ 0.0705, 0.2101, 0.0188, ..., 0.1032, 0.0484, 0.1253]], + device='cuda:0'), grad: tensor([[ 9.3341e-05, 2.8419e-04, -1.9245e-03, ..., -8.4925e-04, + -1.9855e-03, -7.6866e-04], + [ 8.4639e-06, 6.5193e-07, 1.0949e-04, ..., 4.6611e-05, + 1.0580e-04, 4.3750e-05], + [-4.2582e-04, -4.3154e-04, -9.9754e-04, ..., -3.8195e-04, + -7.8344e-04, -4.0030e-04], + ..., + [ 3.0875e-04, 1.4150e-04, 2.6627e-03, ..., 1.1206e-03, + 2.5196e-03, 1.0653e-03], + [ 7.2010e-06, 3.9637e-06, 5.4628e-05, ..., 2.2888e-05, + 5.1260e-05, 2.1860e-05], + [ 2.4959e-06, 8.6799e-07, 2.4691e-05, ..., 1.0438e-05, + 2.3559e-05, 9.8720e-06]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.3258, -0.1182, -0.0050, -0.0046, 0.5885, -0.1621, -0.6312], + device='cuda:0'), grad: tensor([-0.0291, 0.0014, -0.0080, 0.0009, 0.0337, 0.0007, 0.0003], + device='cuda:0') +249 +0.00654508497187474 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 350.09, cls_loss 0.0846 cls_loss_mapping 0.3395 cls_loss_causal 1.0291 re_mapping 0.0492 re_causal 0.0999 /// teacc 91.23 lr 0.00633018 +Epoch 30, weight, value: tensor([[ 0.0245, 0.0303, -0.0354, ..., -0.0729, -0.1203, -0.0921], + [-0.0388, -0.0409, -0.0223, ..., -0.0253, -0.0204, -0.0068], + [ 0.0205, -0.0223, 0.0394, ..., -0.0949, -0.0916, -0.0788], + ..., + [-0.0039, -0.0353, 0.0485, ..., -0.0274, -0.0367, -0.0344], + [-0.0116, -0.0252, -0.0048, ..., -0.0589, -0.0014, -0.0611], + [ 0.0707, 0.2090, 0.0198, ..., 0.1023, 0.0487, 0.1243]], + device='cuda:0'), grad: tensor([[ 1.7681e-03, 4.8485e-03, 2.3468e-02, ..., 1.2390e-02, + 3.2959e-02, 1.2566e-02], + [ 7.0073e-06, 1.9208e-05, 9.2983e-05, ..., 4.9084e-05, + 1.3053e-04, 4.9770e-05], + [ 2.7776e-05, 7.6175e-05, 3.6860e-04, ..., 1.9467e-04, + 5.1785e-04, 1.9753e-04], + ..., + [-1.8711e-03, -5.1308e-03, -2.4826e-02, ..., -1.3115e-02, + -3.4882e-02, -1.3306e-02], + [ 6.5006e-07, 1.8254e-06, 8.8066e-06, ..., 4.6603e-06, + 1.2383e-05, 4.7237e-06], + [ 8.1956e-08, 2.2165e-07, 1.0766e-06, ..., 5.6811e-07, + 1.5125e-06, 5.7742e-07]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.3038, -0.1232, 0.0082, -0.0091, 0.5983, -0.1606, -0.6240], + device='cuda:0'), grad: tensor([ 7.3120e-02, 2.8968e-04, 1.1492e-03, 2.8114e-03, -7.7393e-02, + 2.7552e-05, 3.3565e-06], device='cuda:0') +249 +0.006330184227833378 +changing lr +epoch 29, time 347.50, cls_loss 0.0666 cls_loss_mapping 0.3109 cls_loss_causal 1.0087 re_mapping 0.0451 re_causal 0.1122 /// teacc 88.89 lr 0.00611260 +Epoch 31, weight, value: tensor([[ 0.0241, 0.0298, -0.0359, ..., -0.0724, -0.1209, -0.0915], + [-0.0386, -0.0409, -0.0225, ..., -0.0254, -0.0200, -0.0071], + [ 0.0206, -0.0218, 0.0393, ..., -0.0944, -0.0911, -0.0786], + ..., + [-0.0044, -0.0350, 0.0488, ..., -0.0263, -0.0352, -0.0333], + [-0.0110, -0.0247, -0.0016, ..., -0.0572, 0.0027, -0.0592], + [ 0.0708, 0.2078, 0.0205, ..., 0.1016, 0.0490, 0.1235]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 3.3528e-07, 2.7847e-06, ..., 2.5500e-06, + 4.0308e-06, 2.2575e-06], + [ 8.9407e-08, 4.4703e-08, 4.9360e-07, ..., 1.6578e-07, + 4.1164e-07, 1.8440e-07], + [ 6.7055e-08, 6.3330e-08, 6.0722e-07, ..., 3.5018e-07, + 6.5938e-07, 3.3528e-07], + ..., + [ 1.2107e-07, -2.7753e-07, -2.0042e-06, ..., -2.3581e-06, + -3.4366e-06, -1.9912e-06], + [-3.6135e-07, -2.0303e-07, -2.2780e-06, ..., -8.9034e-07, + -2.0526e-06, -9.7975e-07], + [ 4.6566e-08, 2.7940e-08, 2.9802e-07, ..., 1.2666e-07, + 2.7753e-07, 1.3597e-07]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.3060, -0.1185, 0.0095, -0.0269, 0.5951, -0.1523, -0.6192], + device='cuda:0'), grad: tensor([ 5.0515e-05, 3.3751e-06, 6.9812e-06, 1.1921e-06, -4.6074e-05, + -1.8597e-05, 2.5593e-06], device='cuda:0') +249 +0.006112604669781575 +changing lr +epoch 30, time 348.46, cls_loss 0.0678 cls_loss_mapping 0.3041 cls_loss_causal 1.0036 re_mapping 0.0422 re_causal 0.0991 /// teacc 87.72 lr 0.00589278 +Epoch 32, weight, value: tensor([[ 0.0240, 0.0295, -0.0348, ..., -0.0705, -0.1179, -0.0899], + [-0.0389, -0.0411, -0.0259, ..., -0.0262, -0.0234, -0.0080], + [ 0.0193, -0.0227, 0.0352, ..., -0.0947, -0.0946, -0.0793], + ..., + [-0.0031, -0.0331, 0.0529, ..., -0.0269, -0.0323, -0.0332], + [-0.0111, -0.0248, -0.0027, ..., -0.0571, 0.0015, -0.0591], + [ 0.0704, 0.2064, 0.0204, ..., 0.1008, 0.0485, 0.1226]], + device='cuda:0'), grad: tensor([[ 5.2303e-06, 1.8626e-09, 7.0512e-05, ..., 1.5959e-05, + 4.7475e-05, 1.5870e-05], + [ 1.6205e-07, 3.7253e-09, 2.1849e-06, ..., 4.0792e-07, + 1.4659e-06, 4.2841e-07], + [ 3.4273e-07, 0.0000e+00, 4.6194e-06, ..., 1.0468e-06, + 3.1106e-06, 1.0412e-06], + ..., + [-6.1430e-06, -5.5879e-09, -8.2850e-05, ..., -1.8671e-05, + -5.5790e-05, -1.8582e-05], + [ 6.1467e-08, 0.0000e+00, 8.2888e-07, ..., 1.8813e-07, + 5.5879e-07, 1.8626e-07], + [ 1.0431e-07, 0.0000e+00, 1.4119e-06, ..., 3.2037e-07, + 9.5181e-07, 3.1851e-07]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.2986, -0.1310, 0.0074, -0.0110, 0.5969, -0.1544, -0.6129], + device='cuda:0'), grad: tensor([ 8.8167e-04, 2.6435e-05, 5.7757e-05, 4.0114e-05, -1.0347e-03, + 1.0364e-05, 1.7658e-05], device='cuda:0') +249 +0.005892784473993186 +changing lr +epoch 31, time 347.30, cls_loss 0.0762 cls_loss_mapping 0.2870 cls_loss_causal 0.9118 re_mapping 0.0449 re_causal 0.0952 /// teacc 86.55 lr 0.00567117 +Epoch 33, weight, value: tensor([[ 0.0234, 0.0290, -0.0348, ..., -0.0716, -0.1188, -0.0907], + [-0.0386, -0.0408, -0.0268, ..., -0.0260, -0.0243, -0.0082], + [ 0.0210, -0.0213, 0.0404, ..., -0.0924, -0.0888, -0.0774], + ..., + [-0.0047, -0.0339, 0.0466, ..., -0.0284, -0.0371, -0.0344], + [-0.0111, -0.0246, -0.0016, ..., -0.0565, 0.0029, -0.0583], + [ 0.0698, 0.2048, 0.0199, ..., 0.0998, 0.0478, 0.1215]], + device='cuda:0'), grad: tensor([[ 1.2469e-04, 3.1799e-05, 1.4935e-03, ..., 7.4530e-04, + 1.5926e-03, 7.6580e-04], + [ 1.1921e-04, 3.4094e-05, 1.4572e-03, ..., 7.0524e-04, + 1.5268e-03, 7.2336e-04], + [ 1.2279e-04, 3.2246e-05, 1.4772e-03, ..., 7.3195e-04, + 1.5697e-03, 7.5197e-04], + ..., + [-4.2677e-04, -1.1748e-04, -5.1804e-03, ..., -2.5349e-03, + -5.4626e-03, -2.6016e-03], + [ 2.8372e-05, 9.4175e-06, 3.5691e-04, ..., 1.6570e-04, + 3.6502e-04, 1.6952e-04], + [ 9.0078e-06, 2.6394e-06, 1.1051e-04, ..., 5.3346e-05, + 1.1563e-04, 5.4717e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.3002, -0.1302, 0.0179, 0.0061, 0.5638, -0.1528, -0.6115], + device='cuda:0'), grad: tensor([ 0.0169, 0.0172, 0.0169, 0.0035, -0.0601, 0.0044, 0.0013], + device='cuda:0') +249 +0.00567116632908828 +changing lr +epoch 32, time 344.88, cls_loss 0.0540 cls_loss_mapping 0.2908 cls_loss_causal 0.9360 re_mapping 0.0471 re_causal 0.0968 /// teacc 91.23 lr 0.00544820 +Epoch 34, weight, value: tensor([[ 0.0228, 0.0284, -0.0369, ..., -0.0712, -0.1194, -0.0903], + [-0.0384, -0.0407, -0.0253, ..., -0.0263, -0.0227, -0.0082], + [ 0.0209, -0.0210, 0.0396, ..., -0.0922, -0.0895, -0.0773], + ..., + [-0.0041, -0.0327, 0.0500, ..., -0.0258, -0.0327, -0.0321], + [-0.0109, -0.0245, -0.0009, ..., -0.0557, 0.0040, -0.0574], + [ 0.0694, 0.2034, 0.0198, ..., 0.0990, 0.0474, 0.1206]], + device='cuda:0'), grad: tensor([[-2.0489e-07, -1.2107e-07, -1.1250e-06, ..., -2.6077e-07, + -7.8417e-07, -2.5891e-07], + [ 2.7940e-08, 2.0489e-08, 1.2480e-07, ..., 3.3528e-08, + 8.9407e-08, 3.3528e-08], + [-1.8999e-07, -1.6950e-07, -4.6194e-07, ..., -1.8440e-07, + -3.6508e-07, -1.8626e-07], + ..., + [ 3.1851e-07, 2.3097e-07, 1.3150e-06, ..., 3.6135e-07, + 9.4995e-07, 3.6322e-07], + [ 2.4214e-08, 2.0489e-08, 7.0781e-08, ..., 2.4214e-08, + 5.4017e-08, 2.4214e-08], + [ 1.6764e-08, 1.4901e-08, 5.5879e-08, ..., 1.8626e-08, + 4.2841e-08, 1.8626e-08]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.2845, -0.1277, 0.0171, -0.0108, 0.5882, -0.1503, -0.6074], + device='cuda:0'), grad: tensor([-8.8513e-06, 9.9279e-07, -3.8035e-06, 1.6205e-07, 1.0483e-05, + 5.7369e-07, 4.5635e-07], device='cuda:0') +249 +0.00544819654451717 +changing lr +epoch 33, time 347.79, cls_loss 0.0434 cls_loss_mapping 0.2676 cls_loss_causal 0.8887 re_mapping 0.0472 re_causal 0.0963 /// teacc 90.06 lr 0.00522432 +Epoch 35, weight, value: tensor([[ 0.0221, 0.0274, -0.0383, ..., -0.0715, -0.1193, -0.0896], + [-0.0379, -0.0402, -0.0244, ..., -0.0261, -0.0218, -0.0081], + [ 0.0206, -0.0210, 0.0386, ..., -0.0917, -0.0898, -0.0771], + ..., + [-0.0036, -0.0317, 0.0507, ..., -0.0256, -0.0323, -0.0328], + [-0.0106, -0.0242, 0.0007, ..., -0.0543, 0.0056, -0.0559], + [ 0.0689, 0.2020, 0.0196, ..., 0.0982, 0.0469, 0.1197]], + device='cuda:0'), grad: tensor([[ 3.2410e-07, 2.4214e-07, 5.4874e-06, ..., 4.0196e-06, + 6.5789e-06, 2.8349e-06], + [ 1.8626e-09, 1.8626e-09, 2.6077e-08, ..., 1.6764e-08, + 2.7940e-08, 1.1176e-08], + [ 3.3528e-08, 2.4214e-08, 5.6624e-07, ..., 4.1351e-07, + 6.7614e-07, 2.9057e-07], + ..., + [-3.6322e-07, -2.7195e-07, -6.1877e-06, ..., -4.5486e-06, + -7.4320e-06, -3.2056e-06], + [-3.7253e-09, -0.0000e+00, -1.8626e-09, ..., 1.6764e-08, + 1.4901e-08, 1.1176e-08], + [ 3.7253e-09, 1.8626e-09, 4.4703e-08, ..., 3.1665e-08, + 5.2154e-08, 2.2352e-08]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.2687, -0.1285, 0.0204, -0.0102, 0.5901, -0.1419, -0.6049], + device='cuda:0'), grad: tensor([ 5.6118e-05, 2.5332e-07, 5.7742e-06, 5.8673e-07, -6.3360e-05, + 7.8231e-08, 4.5076e-07], device='cuda:0') +249 +0.005224324151752577 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 349.39, cls_loss 0.0430 cls_loss_mapping 0.2367 cls_loss_causal 0.8582 re_mapping 0.0450 re_causal 0.0985 /// teacc 92.40 lr 0.00500000 +Epoch 36, weight, value: tensor([[ 0.0217, 0.0271, -0.0392, ..., -0.0714, -0.1198, -0.0895], + [-0.0401, -0.0425, -0.0308, ..., -0.0291, -0.0304, -0.0152], + [ 0.0203, -0.0210, 0.0371, ..., -0.0920, -0.0909, -0.0774], + ..., + [-0.0033, -0.0313, 0.0507, ..., -0.0255, -0.0322, -0.0326], + [-0.0104, -0.0240, 0.0017, ..., -0.0535, 0.0067, -0.0551], + [ 0.0708, 0.2032, 0.0263, ..., 0.1006, 0.0557, 0.1260]], + device='cuda:0'), grad: tensor([[ 2.0303e-07, 3.7253e-09, 3.2932e-06, ..., 1.4752e-06, + 3.1497e-06, 9.6299e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 7.4506e-09, 1.8626e-09], + [ 5.5879e-09, -1.6764e-08, 3.4459e-07, ..., 1.5274e-07, + 3.3528e-07, 9.6858e-08], + ..., + [-2.3469e-07, -3.7253e-09, -3.8296e-06, ..., -1.7155e-06, + -3.6620e-06, -1.1194e-06], + [ 1.8626e-08, 1.4901e-08, 8.7544e-08, ..., 4.0978e-08, + 7.8231e-08, 2.9802e-08], + [ 3.7253e-09, 0.0000e+00, 3.7253e-08, ..., 1.6764e-08, + 3.5390e-08, 1.1176e-08]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.2641, -0.1327, 0.0109, -0.0064, 0.5864, -0.1347, -0.5939], + device='cuda:0'), grad: tensor([ 2.8789e-05, 7.4506e-08, 3.0044e-06, 4.7125e-07, -3.3468e-05, + 7.7486e-07, 3.1851e-07], device='cuda:0') +249 +0.005000000000000003 +changing lr +epoch 35, time 348.58, cls_loss 0.0310 cls_loss_mapping 0.2276 cls_loss_causal 0.8888 re_mapping 0.0457 re_causal 0.1002 /// teacc 91.23 lr 0.00477568 +Epoch 37, weight, value: tensor([[ 0.0220, 0.0273, -0.0368, ..., -0.0704, -0.1163, -0.0880], + [-0.0400, -0.0422, -0.0303, ..., -0.0289, -0.0301, -0.0155], + [ 0.0201, -0.0209, 0.0377, ..., -0.0911, -0.0898, -0.0768], + ..., + [-0.0037, -0.0315, 0.0476, ..., -0.0260, -0.0347, -0.0332], + [-0.0104, -0.0239, 0.0009, ..., -0.0536, 0.0057, -0.0551], + [ 0.0705, 0.2020, 0.0262, ..., 0.1000, 0.0555, 0.1256]], + device='cuda:0'), grad: tensor([[ 3.0566e-06, 4.5821e-06, 3.4451e-05, ..., 3.8564e-05, + 5.6267e-05, 3.7909e-05], + [-8.5682e-08, -3.3528e-08, 6.5193e-08, ..., 5.1409e-07, + 5.1036e-07, 4.9546e-07], + [ 5.9605e-08, 1.0431e-07, 9.4622e-07, ..., 1.1232e-06, + 1.6093e-06, 1.1008e-06], + ..., + [ 5.9232e-06, 8.9034e-06, 6.6996e-05, ..., 7.5161e-05, + 1.0961e-04, 7.3850e-05], + [ 3.1758e-06, 2.3935e-06, 1.6183e-05, ..., 1.4700e-05, + 2.1026e-05, 1.3165e-05], + [ 7.0781e-07, 7.9162e-07, 5.7667e-06, ..., 6.0536e-06, + 8.8066e-06, 5.8040e-06]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.2656, -0.1274, 0.0195, -0.0015, 0.5698, -0.1398, -0.5924], + device='cuda:0'), grad: tensor([ 3.1233e-04, 7.2457e-07, 8.8885e-06, -1.0853e-03, 6.0797e-04, + 1.0645e-04, 4.7445e-05], device='cuda:0') +249 +0.004775675848247429 +changing lr +---------------------saving model at epoch 36---------------------------------------------------- +epoch 36, time 353.79, cls_loss 0.0258 cls_loss_mapping 0.2135 cls_loss_causal 0.8457 re_mapping 0.0440 re_causal 0.0889 /// teacc 92.98 lr 0.00455180 +Epoch 38, weight, value: tensor([[ 0.0216, 0.0268, -0.0387, ..., -0.0706, -0.1180, -0.0882], + [-0.0392, -0.0414, -0.0269, ..., -0.0278, -0.0259, -0.0145], + [ 0.0201, -0.0205, 0.0382, ..., -0.0908, -0.0885, -0.0765], + ..., + [-0.0037, -0.0314, 0.0478, ..., -0.0250, -0.0340, -0.0322], + [-0.0105, -0.0239, -0.0002, ..., -0.0537, 0.0045, -0.0552], + [ 0.0700, 0.2008, 0.0257, ..., 0.0992, 0.0547, 0.1247]], + device='cuda:0'), grad: tensor([[ 1.7226e-04, 2.0072e-05, 1.4858e-03, ..., 5.2261e-04, + 1.4391e-03, 5.3930e-04], + [ 3.0667e-05, 3.6322e-06, 2.6464e-04, ..., 9.3102e-05, + 2.5630e-04, 9.6083e-05], + [ 5.8323e-05, 6.7316e-06, 5.0306e-04, ..., 1.7703e-04, + 4.8757e-04, 1.8275e-04], + ..., + [ 1.9684e-03, 2.2709e-04, 1.6953e-02, ..., 5.9662e-03, + 1.6418e-02, 6.1569e-03], + [-2.4223e-03, -2.7990e-04, -2.0889e-02, ..., -7.3471e-03, + -2.0233e-02, -7.5836e-03], + [ 2.8446e-05, 3.3136e-06, 2.4533e-04, ..., 8.6308e-05, + 2.3758e-04, 8.9049e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.2553, -0.1169, 0.0185, -0.0010, 0.5754, -0.1466, -0.5910], + device='cuda:0'), grad: tensor([ 0.0071, 0.0013, 0.0024, 0.0068, 0.0804, -0.0991, 0.0012], + device='cuda:0') +249 +0.004551803455482836 +changing lr +epoch 37, time 345.70, cls_loss 0.0392 cls_loss_mapping 0.2031 cls_loss_causal 0.7827 re_mapping 0.0426 re_causal 0.0847 /// teacc 91.81 lr 0.00432883 +Epoch 39, weight, value: tensor([[ 0.0215, 0.0267, -0.0387, ..., -0.0700, -0.1176, -0.0875], + [-0.0390, -0.0412, -0.0273, ..., -0.0277, -0.0263, -0.0145], + [ 0.0200, -0.0202, 0.0388, ..., -0.0900, -0.0872, -0.0757], + ..., + [-0.0041, -0.0317, 0.0457, ..., -0.0263, -0.0372, -0.0337], + [-0.0103, -0.0238, 0.0006, ..., -0.0531, 0.0052, -0.0546], + [ 0.0696, 0.1996, 0.0253, ..., 0.0986, 0.0543, 0.1239]], + device='cuda:0'), grad: tensor([[-5.1069e-04, -5.7316e-04, -1.7967e-03, ..., -6.3467e-04, + -1.9684e-03, -6.0940e-04], + [-2.6338e-06, -2.0098e-06, -2.4334e-05, ..., -3.8091e-06, + -2.3484e-05, -4.0159e-06], + [ 3.3885e-05, 1.1325e-04, 6.0272e-04, ..., 9.6262e-05, + 7.1335e-04, 1.5843e-04], + ..., + [ 4.6396e-04, 4.4703e-04, 1.1797e-03, ..., 5.2500e-04, + 1.2379e-03, 4.4084e-04], + [-1.1455e-07, -2.5146e-08, -4.9360e-07, ..., -8.6613e-08, + -5.4948e-07, -1.2945e-07], + [ 1.5736e-05, 1.5169e-05, 3.9428e-05, ..., 1.7777e-05, + 4.1455e-05, 1.4901e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.2524, -0.1184, 0.0241, 0.0039, 0.5624, -0.1418, -0.5888], + device='cuda:0'), grad: tensor([-5.0392e-03, -7.8619e-05, 2.1935e-03, 2.6729e-06, 2.8324e-03, + -8.0094e-07, 9.3877e-05], device='cuda:0') +249 +0.004328833670911726 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 348.97, cls_loss 0.0203 cls_loss_mapping 0.1966 cls_loss_causal 0.8061 re_mapping 0.0423 re_causal 0.0850 /// teacc 94.74 lr 0.00410722 +Epoch 40, weight, value: tensor([[ 0.0215, 0.0266, -0.0386, ..., -0.0695, -0.1169, -0.0868], + [-0.0386, -0.0406, -0.0256, ..., -0.0275, -0.0248, -0.0143], + [ 0.0196, -0.0203, 0.0365, ..., -0.0902, -0.0895, -0.0764], + ..., + [-0.0039, -0.0316, 0.0467, ..., -0.0254, -0.0348, -0.0325], + [-0.0105, -0.0238, -0.0011, ..., -0.0535, 0.0031, -0.0552], + [ 0.0690, 0.1984, 0.0247, ..., 0.0979, 0.0535, 0.1230]], + device='cuda:0'), grad: tensor([[-7.0781e-08, -1.9558e-08, -9.7137e-07, ..., -4.9174e-07, + -1.0999e-06, -4.7404e-07], + [ 1.1176e-08, 9.3132e-10, 9.7789e-08, ..., 2.5146e-08, + 1.0058e-07, 2.6077e-08], + [ 1.3039e-08, 2.7940e-09, 1.3132e-07, ..., 4.0978e-08, + 1.3690e-07, 4.0978e-08], + ..., + [ 7.5437e-08, 1.6764e-08, 8.1304e-07, ..., 2.9150e-07, + 8.6334e-07, 2.8964e-07], + [-1.0058e-07, -1.3970e-08, -7.5065e-07, ..., -4.4703e-08, + -6.9104e-07, -6.6124e-08], + [ 3.4459e-08, 5.5879e-09, 3.0361e-07, ..., 6.6124e-08, + 3.0268e-07, 6.9849e-08]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.2518, -0.1152, 0.0227, 0.0056, 0.5624, -0.1456, -0.5879], + device='cuda:0'), grad: tensor([-6.8396e-06, 5.5786e-07, 8.0466e-07, 2.2724e-06, 5.1595e-06, + -3.6880e-06, 1.7229e-06], device='cuda:0') +249 +0.0041072155260068206 +changing lr +epoch 39, time 348.75, cls_loss 0.0249 cls_loss_mapping 0.1840 cls_loss_causal 0.8273 re_mapping 0.0398 re_causal 0.0809 /// teacc 90.64 lr 0.00388740 +Epoch 41, weight, value: tensor([[ 0.0218, 0.0268, -0.0371, ..., -0.0690, -0.1148, -0.0860], + [-0.0383, -0.0403, -0.0250, ..., -0.0273, -0.0241, -0.0142], + [ 0.0195, -0.0203, 0.0359, ..., -0.0899, -0.0896, -0.0763], + ..., + [-0.0044, -0.0319, 0.0448, ..., -0.0254, -0.0368, -0.0328], + [-0.0102, -0.0233, 0.0007, ..., -0.0526, 0.0053, -0.0540], + [ 0.0687, 0.1974, 0.0243, ..., 0.0973, 0.0529, 0.1223]], + device='cuda:0'), grad: tensor([[2.2352e-06, 6.1840e-07, 1.8418e-05, ..., 1.3851e-05, 2.5854e-05, + 1.2666e-05], + [2.2817e-07, 7.8231e-08, 1.3914e-06, ..., 8.5030e-07, 1.7053e-06, + 8.2236e-07], + [1.6764e-05, 5.5805e-06, 7.7784e-05, ..., 3.8773e-05, 8.3685e-05, + 3.9458e-05], + ..., + [8.1360e-06, 2.7195e-06, 3.8117e-05, ..., 1.9133e-05, 4.1217e-05, + 1.9446e-05], + [4.0419e-06, 1.3923e-06, 2.1443e-05, ..., 1.1943e-05, 2.4766e-05, + 1.1817e-05], + [3.3919e-06, 1.2238e-06, 1.7583e-05, ..., 9.4399e-06, 1.9893e-05, + 9.4622e-06]], device='cuda:0') +Epoch 41, bias, value: tensor([ 2.4937e-01, -1.1424e-01, 2.2541e-02, 4.9674e-04, 5.6182e-01, + -1.4016e-01, -5.8597e-01], device='cuda:0'), grad: tensor([ 1.0884e-04, 8.0839e-06, 3.7861e-04, -8.8835e-04, 1.8752e-04, + 1.1289e-04, 9.2149e-05], device='cuda:0') +249 +0.0038873953302184317 +changing lr +epoch 40, time 347.58, cls_loss 0.0184 cls_loss_mapping 0.1822 cls_loss_causal 0.8379 re_mapping 0.0405 re_causal 0.0762 /// teacc 93.57 lr 0.00366982 +Epoch 42, weight, value: tensor([[ 0.0217, 0.0267, -0.0374, ..., -0.0689, -0.1149, -0.0858], + [-0.0381, -0.0401, -0.0244, ..., -0.0270, -0.0234, -0.0139], + [ 0.0194, -0.0202, 0.0359, ..., -0.0894, -0.0891, -0.0759], + ..., + [-0.0035, -0.0310, 0.0476, ..., -0.0241, -0.0331, -0.0314], + [-0.0109, -0.0238, -0.0021, ..., -0.0534, 0.0022, -0.0548], + [ 0.0683, 0.1964, 0.0239, ..., 0.0967, 0.0524, 0.1216]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 1.3970e-08, ..., 4.6566e-09, + 1.3970e-08, 4.6566e-09], + [-1.8626e-09, -2.7940e-09, -2.7008e-08, ..., -8.3819e-09, + -2.6077e-08, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 3.7253e-09, 9.3132e-10], + ..., + [ 1.2107e-08, 5.5879e-09, 9.8720e-08, ..., 3.0734e-08, + 1.0617e-07, 4.3772e-08], + [-1.5832e-08, -6.5193e-09, -1.2200e-07, ..., -3.8184e-08, + -1.3225e-07, -5.5879e-08], + [ 1.8626e-09, 9.3132e-10, 1.7695e-08, ..., 5.5879e-09, + 1.9558e-08, 8.3819e-09]], device='cuda:0') +Epoch 42, bias, value: tensor([ 2.4822e-01, -1.1041e-01, 2.3572e-02, -2.5686e-04, 5.6161e-01, + -1.4442e-01, -5.8441e-01], device='cuda:0'), grad: tensor([ 1.2759e-07, -2.8498e-07, 1.4901e-08, 7.6368e-08, 5.7649e-07, + -5.9418e-07, 9.1270e-08], device='cuda:0') +249 +0.003669815772166629 +changing lr +epoch 41, time 349.76, cls_loss 0.0273 cls_loss_mapping 0.1802 cls_loss_causal 0.8285 re_mapping 0.0382 re_causal 0.0830 /// teacc 91.81 lr 0.00345492 +Epoch 43, weight, value: tensor([[ 0.0215, 0.0264, -0.0379, ..., -0.0689, -0.1151, -0.0857], + [-0.0373, -0.0394, -0.0207, ..., -0.0266, -0.0203, -0.0134], + [ 0.0193, -0.0201, 0.0359, ..., -0.0890, -0.0885, -0.0755], + ..., + [-0.0041, -0.0313, 0.0442, ..., -0.0241, -0.0355, -0.0316], + [-0.0105, -0.0235, 0.0002, ..., -0.0525, 0.0048, -0.0537], + [ 0.0679, 0.1955, 0.0237, ..., 0.0962, 0.0519, 0.1210]], + device='cuda:0'), grad: tensor([[ 1.7416e-07, 8.6613e-08, 3.6266e-06, ..., 8.9873e-07, + 3.3863e-06, 7.4785e-07], + [-5.5879e-09, -0.0000e+00, -8.6613e-08, ..., -4.6566e-09, + -6.0536e-08, -2.7940e-09], + [ 3.7253e-09, 9.3132e-10, 1.0710e-07, ..., 2.6077e-08, + 1.1269e-07, 3.3528e-08], + ..., + [-1.6857e-07, -8.7544e-08, -3.5223e-06, ..., -8.9034e-07, + -3.2913e-06, -7.2736e-07], + [-5.5879e-09, 9.3132e-10, -2.1793e-07, ..., -5.3085e-08, + -2.4587e-07, -8.1956e-08], + [ 2.7940e-09, 0.0000e+00, 8.1025e-08, ..., 1.9558e-08, + 8.7544e-08, 2.7940e-08]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.2411, -0.1017, 0.0264, -0.0056, 0.5517, -0.1354, -0.5827], + device='cuda:0'), grad: tensor([ 2.3931e-05, -6.6031e-07, 6.8638e-07, 8.9407e-08, -2.3156e-05, + -1.3653e-06, 5.1130e-07], device='cuda:0') +249 +0.0034549150281252667 +changing lr +epoch 42, time 349.43, cls_loss 0.0260 cls_loss_mapping 0.1887 cls_loss_causal 0.7472 re_mapping 0.0344 re_causal 0.0714 /// teacc 92.40 lr 0.00324313 +Epoch 44, weight, value: tensor([[ 0.0212, 0.0261, -0.0394, ..., -0.0695, -0.1166, -0.0861], + [-0.0372, -0.0394, -0.0214, ..., -0.0270, -0.0215, -0.0139], + [ 0.0194, -0.0199, 0.0365, ..., -0.0884, -0.0874, -0.0750], + ..., + [-0.0035, -0.0308, 0.0474, ..., -0.0223, -0.0309, -0.0298], + [-0.0106, -0.0235, -0.0005, ..., -0.0525, 0.0040, -0.0537], + [ 0.0676, 0.1946, 0.0233, ..., 0.0957, 0.0514, 0.1203]], + device='cuda:0'), grad: tensor([[-1.3959e-04, -9.1136e-05, -1.1940e-03, ..., -1.2386e-04, + -1.0395e-03, -1.3590e-04], + [ 1.1966e-05, 1.1869e-05, 1.3447e-04, ..., 1.7852e-05, + 1.3018e-04, 1.9625e-05], + [ 2.1867e-06, 1.9129e-06, 2.2575e-05, ..., 2.8014e-06, + 2.1234e-05, 3.0827e-06], + ..., + [ 1.2386e-04, 7.6175e-05, 1.0223e-03, ..., 1.0151e-04, + 8.7500e-04, 1.1128e-04], + [ 5.2806e-07, 4.9081e-07, 5.6736e-06, ..., 7.2829e-07, + 5.4091e-06, 8.0094e-07], + [ 8.0373e-07, 5.4482e-07, 7.0147e-06, ..., 7.4971e-07, + 6.1691e-06, 8.2050e-07]], device='cuda:0') +Epoch 44, bias, value: tensor([ 2.2749e-01, -1.0423e-01, 3.1431e-02, 1.9333e-04, 5.5837e-01, + -1.3793e-01, -5.8138e-01], device='cuda:0'), grad: tensor([-5.3062e-03, 6.4611e-04, 1.0616e-04, 1.3657e-05, 4.4861e-03, + 2.6956e-05, 3.1412e-05], device='cuda:0') +249 +0.0032431258795932905 +changing lr +epoch 43, time 347.80, cls_loss 0.0220 cls_loss_mapping 0.1586 cls_loss_causal 0.7720 re_mapping 0.0327 re_causal 0.0697 /// teacc 92.98 lr 0.00303487 +Epoch 45, weight, value: tensor([[ 0.0212, 0.0260, -0.0388, ..., -0.0691, -0.1155, -0.0856], + [-0.0370, -0.0391, -0.0209, ..., -0.0267, -0.0210, -0.0137], + [ 0.0193, -0.0198, 0.0367, ..., -0.0879, -0.0868, -0.0746], + ..., + [-0.0036, -0.0307, 0.0466, ..., -0.0223, -0.0314, -0.0298], + [-0.0106, -0.0234, -0.0010, ..., -0.0529, 0.0031, -0.0540], + [ 0.0673, 0.1938, 0.0231, ..., 0.0952, 0.0510, 0.1198]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, 6.5193e-09, 2.7753e-07, ..., 1.2107e-08, + 2.2911e-07, 4.1910e-08], + [ 6.3330e-08, 9.3132e-09, 4.2841e-07, ..., 1.8626e-08, + 3.5483e-07, 6.6124e-08], + [ 7.0781e-08, 7.4506e-09, 4.8056e-07, ..., 1.6764e-08, + 3.9861e-07, 7.2643e-08], + ..., + [ 1.8440e-07, 2.7008e-08, 1.2359e-06, ..., 5.4017e-08, + 1.0226e-06, 1.8999e-07], + [-6.0163e-07, -8.5682e-08, -4.0382e-06, ..., -1.6857e-07, + -3.3397e-06, -6.1560e-07], + [ 1.5274e-07, 2.2352e-08, 1.0235e-06, ..., 4.5635e-08, + 8.4843e-07, 1.5832e-07]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.2288, -0.1013, 0.0338, 0.0020, 0.5523, -0.1420, -0.5797], + device='cuda:0'), grad: tensor([ 1.2945e-06, 2.0079e-06, 2.1942e-06, 2.7083e-06, 5.8040e-06, + -1.8835e-05, 4.7907e-06], device='cuda:0') +249 +0.0030348748417303863 +changing lr +epoch 44, time 348.82, cls_loss 0.0162 cls_loss_mapping 0.1627 cls_loss_causal 0.7833 re_mapping 0.0324 re_causal 0.0694 /// teacc 92.40 lr 0.00283058 +Epoch 46, weight, value: tensor([[ 0.0214, 0.0261, -0.0369, ..., -0.0681, -0.1128, -0.0845], + [-0.0369, -0.0390, -0.0210, ..., -0.0267, -0.0211, -0.0138], + [ 0.0192, -0.0197, 0.0366, ..., -0.0876, -0.0864, -0.0743], + ..., + [-0.0037, -0.0307, 0.0451, ..., -0.0229, -0.0332, -0.0304], + [-0.0105, -0.0233, -0.0010, ..., -0.0527, 0.0031, -0.0538], + [ 0.0670, 0.1930, 0.0228, ..., 0.0948, 0.0506, 0.1193]], + device='cuda:0'), grad: tensor([[-4.6976e-06, -1.3476e-06, -2.8268e-05, ..., -5.4240e-06, + -2.3216e-05, -6.0126e-06], + [ 9.7230e-07, 2.8871e-07, 5.8413e-06, ..., 1.1381e-06, + 4.8243e-06, 1.2657e-06], + [ 1.6019e-07, 6.9384e-08, 9.8627e-07, ..., 2.4773e-07, + 8.8522e-07, 2.7753e-07], + ..., + [ 3.2540e-06, 8.7917e-07, 1.9431e-05, ..., 3.5278e-06, + 1.5780e-05, 3.9414e-06], + [-8.3353e-08, -5.3085e-08, -3.2783e-07, ..., -4.0978e-08, + -3.4273e-07, -1.0384e-07], + [ 3.0082e-07, 1.1595e-07, 1.7677e-06, ..., 3.8277e-07, + 1.5311e-06, 4.4005e-07]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.2330, -0.1023, 0.0352, 0.0026, 0.5458, -0.1421, -0.5782], + device='cuda:0'), grad: tensor([-1.0890e-04, 2.2680e-05, 4.4368e-06, 2.9523e-06, 7.2658e-05, + -9.5367e-07, 7.2941e-06], device='cuda:0') +249 +0.0028305813044122124 +changing lr +epoch 45, time 347.20, cls_loss 0.0141 cls_loss_mapping 0.1489 cls_loss_causal 0.7867 re_mapping 0.0319 re_causal 0.0649 /// teacc 91.81 lr 0.00263066 +Epoch 47, weight, value: tensor([[ 0.0213, 0.0260, -0.0365, ..., -0.0677, -0.1122, -0.0841], + [-0.0368, -0.0389, -0.0212, ..., -0.0267, -0.0213, -0.0138], + [ 0.0191, -0.0197, 0.0361, ..., -0.0874, -0.0865, -0.0742], + ..., + [-0.0037, -0.0305, 0.0450, ..., -0.0228, -0.0331, -0.0303], + [-0.0104, -0.0231, -0.0005, ..., -0.0524, 0.0036, -0.0535], + [ 0.0668, 0.1923, 0.0226, ..., 0.0944, 0.0503, 0.1188]], + device='cuda:0'), grad: tensor([[ 8.6725e-06, 5.0217e-06, 4.3958e-05, ..., 5.4464e-06, + 3.4899e-05, 6.7502e-06], + [ 7.4506e-07, 4.2841e-07, 3.7719e-06, ..., 4.8149e-07, + 3.0138e-06, 5.9186e-07], + [ 1.7090e-07, 9.6858e-08, 8.6566e-07, ..., 1.2107e-07, + 7.0408e-07, 1.4529e-07], + ..., + [-9.7156e-06, -5.6326e-06, -4.9174e-05, ..., -5.9083e-06, + -3.8832e-05, -7.3835e-06], + [-1.1781e-07, -2.3283e-08, -5.2340e-07, ..., -1.5181e-07, + -5.5414e-07, -1.5413e-07], + [ 1.7742e-07, 7.4971e-08, 8.8243e-07, ..., 2.3749e-07, + 8.6799e-07, 2.5053e-07]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.2331, -0.1031, 0.0329, 0.0037, 0.5448, -0.1405, -0.5769], + device='cuda:0'), grad: tensor([ 1.4853e-04, 1.2763e-05, 2.9393e-06, 2.2631e-07, -1.6594e-04, + -1.7229e-06, 3.0603e-06], device='cuda:0') +249 +0.0026306566876350096 +changing lr +epoch 46, time 351.40, cls_loss 0.0152 cls_loss_mapping 0.1497 cls_loss_causal 0.7486 re_mapping 0.0335 re_causal 0.0688 /// teacc 93.57 lr 0.00243550 +Epoch 48, weight, value: tensor([[ 0.0212, 0.0259, -0.0366, ..., -0.0673, -0.1120, -0.0837], + [-0.0368, -0.0388, -0.0214, ..., -0.0268, -0.0216, -0.0139], + [ 0.0191, -0.0196, 0.0359, ..., -0.0873, -0.0864, -0.0741], + ..., + [-0.0035, -0.0303, 0.0454, ..., -0.0227, -0.0324, -0.0301], + [-0.0103, -0.0231, -0.0005, ..., -0.0522, 0.0037, -0.0533], + [ 0.0665, 0.1917, 0.0224, ..., 0.0940, 0.0500, 0.1183]], + device='cuda:0'), grad: tensor([[-3.7253e-08, -2.8871e-08, -4.2934e-07, ..., -1.4808e-07, + -5.2014e-07, -1.5087e-07], + [ 3.4925e-08, 2.1886e-08, 3.6834e-07, ..., 1.2806e-07, + 4.2701e-07, 1.3644e-07], + [ 6.6590e-08, 2.9802e-08, 4.0792e-07, ..., 1.6857e-07, + 4.2701e-07, 1.7509e-07], + ..., + [-6.9849e-08, -3.3993e-08, -4.3446e-07, ..., -2.0536e-07, + -4.4284e-07, -2.0489e-07], + [-2.5611e-08, 1.3970e-09, -1.7043e-07, ..., -7.9162e-09, + -1.7602e-07, -4.1910e-08], + [ 1.8161e-08, 4.6566e-09, 1.5087e-07, ..., 3.3993e-08, + 1.6671e-07, 4.8429e-08]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.2353, -0.1035, 0.0320, 0.0043, 0.5411, -0.1397, -0.5755], + device='cuda:0'), grad: tensor([-2.6878e-06, 2.3711e-06, 2.2128e-06, 5.7975e-07, -2.6841e-06, + -5.8161e-07, 7.8138e-07], device='cuda:0') +249 +0.0024355036129704724 +changing lr +epoch 47, time 348.26, cls_loss 0.0167 cls_loss_mapping 0.1378 cls_loss_causal 0.7175 re_mapping 0.0324 re_causal 0.0644 /// teacc 92.40 lr 0.00224552 +Epoch 49, weight, value: tensor([[ 0.0212, 0.0258, -0.0364, ..., -0.0671, -0.1116, -0.0834], + [-0.0367, -0.0387, -0.0215, ..., -0.0267, -0.0217, -0.0139], + [ 0.0190, -0.0195, 0.0362, ..., -0.0869, -0.0856, -0.0737], + ..., + [-0.0036, -0.0302, 0.0452, ..., -0.0227, -0.0324, -0.0301], + [-0.0103, -0.0230, -0.0005, ..., -0.0521, 0.0036, -0.0532], + [ 0.0663, 0.1911, 0.0223, ..., 0.0937, 0.0498, 0.1180]], + device='cuda:0'), grad: tensor([[ 1.5041e-07, 1.8626e-08, 2.8182e-06, ..., 2.5146e-07, + 2.3022e-06, 2.9523e-07], + [-4.1649e-06, -3.5809e-07, -8.1778e-05, ..., -5.4538e-06, + -6.5923e-05, -7.3649e-06], + [-3.5390e-07, -2.3004e-07, -1.1837e-06, ..., -6.8778e-07, + -1.3318e-06, -6.9663e-07], + ..., + [ 4.3176e-06, 5.4855e-07, 7.9632e-05, ..., 5.8115e-06, + 6.4492e-05, 7.6815e-06], + [ 1.8626e-08, 1.0710e-08, 1.4761e-07, ..., 3.2596e-08, + 1.3225e-07, 3.3993e-08], + [ 1.1176e-08, 6.0536e-09, 1.0571e-07, ..., 1.8626e-08, + 9.1735e-08, 2.0023e-08]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.2346, -0.1037, 0.0338, 0.0039, 0.5387, -0.1391, -0.5740], + device='cuda:0'), grad: tensor([ 2.5690e-05, -7.4625e-04, -8.5533e-06, 1.1157e-06, 7.2432e-04, + 1.2517e-06, 9.1363e-07], device='cuda:0') +249 +0.00224551509273949 +changing lr +epoch 48, time 349.54, cls_loss 0.0157 cls_loss_mapping 0.1420 cls_loss_causal 0.6908 re_mapping 0.0317 re_causal 0.0606 /// teacc 92.98 lr 0.00206107 +Epoch 50, weight, value: tensor([[ 0.0211, 0.0257, -0.0366, ..., -0.0671, -0.1117, -0.0833], + [-0.0365, -0.0386, -0.0207, ..., -0.0266, -0.0210, -0.0138], + [ 0.0189, -0.0195, 0.0356, ..., -0.0868, -0.0859, -0.0738], + ..., + [-0.0037, -0.0303, 0.0446, ..., -0.0229, -0.0328, -0.0302], + [-0.0102, -0.0229, -0.0003, ..., -0.0517, 0.0038, -0.0527], + [ 0.0661, 0.1906, 0.0221, ..., 0.0934, 0.0496, 0.1176]], + device='cuda:0'), grad: tensor([[ 9.1374e-05, 2.3050e-07, 1.6365e-03, ..., 1.0624e-03, + 2.1915e-03, 1.0624e-03], + [ 1.9059e-05, 2.4680e-08, 3.4165e-04, ..., 2.2185e-04, + 4.5753e-04, 2.2185e-04], + [ 1.7571e-04, 4.6566e-09, 3.1509e-03, ..., 2.0466e-03, + 4.2229e-03, 2.0466e-03], + ..., + [ 3.8236e-05, -2.6356e-07, 6.8855e-04, ..., 4.4870e-04, + 9.2459e-04, 4.4870e-04], + [ 2.2948e-05, 0.0000e+00, 4.1151e-04, ..., 2.6751e-04, + 5.5122e-04, 2.6751e-04], + [ 1.2010e-05, 2.3283e-09, 2.1529e-04, ..., 1.3983e-04, + 2.8849e-04, 1.3983e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.2324, -0.1012, 0.0313, 0.0066, 0.5357, -0.1379, -0.5729], + device='cuda:0'), grad: tensor([ 0.0201, 0.0042, 0.0387, -0.0791, 0.0085, 0.0051, 0.0026], + device='cuda:0') +249 +0.002061073738537637 +changing lr +epoch 49, time 349.14, cls_loss 0.0116 cls_loss_mapping 0.1371 cls_loss_causal 0.6570 re_mapping 0.0313 re_causal 0.0632 /// teacc 90.06 lr 0.00188255 +Epoch 51, weight, value: tensor([[ 0.0212, 0.0257, -0.0357, ..., -0.0667, -0.1105, -0.0829], + [-0.0363, -0.0384, -0.0197, ..., -0.0265, -0.0200, -0.0137], + [ 0.0188, -0.0195, 0.0353, ..., -0.0867, -0.0859, -0.0736], + ..., + [-0.0039, -0.0303, 0.0433, ..., -0.0230, -0.0339, -0.0304], + [-0.0103, -0.0229, -0.0008, ..., -0.0518, 0.0032, -0.0528], + [ 0.0659, 0.1901, 0.0220, ..., 0.0932, 0.0494, 0.1173]], + device='cuda:0'), grad: tensor([[ 2.1094e-07, 1.3877e-07, 1.2722e-06, ..., 5.4948e-07, + 1.3569e-06, 4.4703e-07], + [ 4.6566e-10, -4.6566e-09, -1.8626e-08, ..., 3.8650e-08, + 1.1642e-08, 1.9092e-08], + [ 5.5879e-09, 2.3283e-09, 3.3993e-08, ..., 2.4680e-08, + 4.3306e-08, 1.9558e-08], + ..., + [-1.2573e-07, -1.0757e-07, -7.4692e-07, ..., -1.7555e-07, + -6.9290e-07, -1.4016e-07], + [ 4.6082e-06, 1.4938e-06, 2.7388e-05, ..., 2.1830e-05, + 3.6180e-05, 1.7256e-05], + [ 1.1269e-07, 3.7253e-08, 6.7614e-07, ..., 5.3132e-07, + 8.8988e-07, 4.2561e-07]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.2339, -0.0990, 0.0304, 0.0086, 0.5305, -0.1386, -0.5717], + device='cuda:0'), grad: tensor([ 5.8785e-06, -1.5087e-07, 1.8068e-07, -1.5676e-04, -2.9448e-06, + 1.5032e-04, 3.7197e-06], device='cuda:0') +249 +0.0018825509907063344 +changing lr +epoch 50, time 347.08, cls_loss 0.0089 cls_loss_mapping 0.1230 cls_loss_causal 0.6582 re_mapping 0.0311 re_causal 0.0606 /// teacc 92.98 lr 0.00171031 +Epoch 52, weight, value: tensor([[ 0.0211, 0.0257, -0.0356, ..., -0.0665, -0.1102, -0.0827], + [-0.0362, -0.0383, -0.0197, ..., -0.0264, -0.0200, -0.0137], + [ 0.0188, -0.0195, 0.0351, ..., -0.0865, -0.0858, -0.0735], + ..., + [-0.0038, -0.0302, 0.0433, ..., -0.0230, -0.0337, -0.0303], + [-0.0103, -0.0229, -0.0008, ..., -0.0516, 0.0032, -0.0526], + [ 0.0658, 0.1896, 0.0219, ..., 0.0929, 0.0492, 0.1170]], + device='cuda:0'), grad: tensor([[ 5.9232e-06, 1.7181e-05, 1.1671e-04, ..., 4.6760e-05, + 1.3542e-04, 5.2303e-05], + [ 3.1404e-06, 8.9407e-06, 6.0827e-05, ..., 2.4393e-05, + 7.0512e-05, 2.7254e-05], + [-1.4760e-05, -4.3690e-05, -2.9564e-04, ..., -1.1837e-04, + -3.4332e-04, -1.3256e-04], + ..., + [ 2.0899e-06, 5.6662e-06, 3.8713e-05, ..., 1.5587e-05, + 4.4852e-05, 1.7360e-05], + [ 3.9935e-06, 8.7619e-06, 6.1870e-05, ..., 2.5079e-05, + 7.0989e-05, 2.7627e-05], + [ 1.7686e-06, 3.5763e-06, 2.5600e-05, ..., 1.0423e-05, + 2.9281e-05, 1.1422e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.2331, -0.0991, 0.0303, 0.0088, 0.5297, -0.1382, -0.5705], + device='cuda:0'), grad: tensor([ 9.2030e-04, 4.7851e-04, -2.3384e-03, -1.7226e-05, 3.0231e-04, + 4.6611e-04, 1.9002e-04], device='cuda:0') +249 +0.0017103063703014388 +changing lr +epoch 51, time 350.62, cls_loss 0.0133 cls_loss_mapping 0.1291 cls_loss_causal 0.6928 re_mapping 0.0306 re_causal 0.0583 /// teacc 92.40 lr 0.00154469 +Epoch 53, weight, value: tensor([[ 0.0209, 0.0255, -0.0364, ..., -0.0666, -0.1109, -0.0828], + [-0.0361, -0.0383, -0.0199, ..., -0.0264, -0.0202, -0.0137], + [ 0.0189, -0.0192, 0.0361, ..., -0.0861, -0.0845, -0.0731], + ..., + [-0.0038, -0.0301, 0.0433, ..., -0.0228, -0.0335, -0.0301], + [-0.0102, -0.0228, -0.0007, ..., -0.0515, 0.0033, -0.0525], + [ 0.0656, 0.1892, 0.0218, ..., 0.0927, 0.0490, 0.1167]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, 1.2247e-07, ..., 1.2107e-08, + 9.7323e-08, 9.7789e-09], + [-1.0245e-08, -2.7940e-09, -2.2631e-07, ..., -2.4680e-08, + -1.8207e-07, -1.9558e-08], + [ 9.3132e-10, 4.6566e-10, 1.0710e-08, ..., 9.3132e-10, + 8.8476e-09, 1.3970e-09], + ..., + [ 1.4901e-08, 4.6566e-09, 2.0303e-07, ..., 1.9092e-08, + 1.6438e-07, 2.3283e-08], + [-1.4435e-08, -4.6566e-09, -1.4110e-07, ..., -9.7789e-09, + -1.1316e-07, -1.9092e-08], + [ 2.3283e-09, 9.3132e-10, 2.3749e-08, ..., 1.8626e-09, + 1.9092e-08, 3.2596e-09]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.2295, -0.0996, 0.0344, 0.0079, 0.5294, -0.1378, -0.5696], + device='cuda:0'), grad: tensor([ 8.1165e-07, -1.5292e-06, 4.8894e-08, 4.0513e-08, 1.1073e-06, + -5.8394e-07, 1.0803e-07], device='cuda:0') +249 +0.0015446867550656784 +changing lr +epoch 52, time 348.17, cls_loss 0.0145 cls_loss_mapping 0.1320 cls_loss_causal 0.7023 re_mapping 0.0281 re_causal 0.0541 /// teacc 91.81 lr 0.00138603 +Epoch 54, weight, value: tensor([[ 0.0208, 0.0253, -0.0370, ..., -0.0667, -0.1116, -0.0828], + [-0.0361, -0.0382, -0.0199, ..., -0.0264, -0.0202, -0.0137], + [ 0.0189, -0.0191, 0.0367, ..., -0.0856, -0.0834, -0.0727], + ..., + [-0.0038, -0.0300, 0.0433, ..., -0.0229, -0.0335, -0.0302], + [-0.0102, -0.0228, -0.0007, ..., -0.0514, 0.0033, -0.0524], + [ 0.0655, 0.1888, 0.0217, ..., 0.0925, 0.0489, 0.1165]], + device='cuda:0'), grad: tensor([[ 3.2522e-06, 1.6252e-06, 1.9357e-05, ..., 1.2005e-06, + 1.5438e-05, 1.0198e-06], + [ 4.3004e-07, 1.4878e-07, 2.4345e-06, ..., 1.5367e-07, + 1.9297e-06, 1.2107e-07], + [ 3.3993e-07, 1.6950e-07, 2.0228e-06, ..., 1.2713e-07, + 1.6158e-06, 1.0803e-07], + ..., + [-4.1947e-06, -2.0079e-06, -2.4796e-05, ..., -1.5376e-06, + -1.9774e-05, -1.2927e-06], + [ 8.4285e-08, 2.9569e-08, 4.8243e-07, ..., 5.0990e-08, + 4.0536e-07, 4.4703e-08], + [ 3.1665e-08, 1.3039e-08, 1.8463e-07, ..., 1.7928e-08, + 1.5344e-07, 1.5832e-08]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.2269, -0.0993, 0.0366, 0.0082, 0.5280, -0.1374, -0.5688], + device='cuda:0'), grad: tensor([ 5.9217e-05, 7.1488e-06, 6.1877e-06, 7.3621e-07, -7.5459e-05, + 1.4594e-06, 5.6159e-07], device='cuda:0') +249 +0.001386025680863044 +changing lr +epoch 53, time 348.79, cls_loss 0.0133 cls_loss_mapping 0.1284 cls_loss_causal 0.7003 re_mapping 0.0296 re_causal 0.0593 /// teacc 91.81 lr 0.00123464 +Epoch 55, weight, value: tensor([[ 0.0208, 0.0252, -0.0373, ..., -0.0668, -0.1118, -0.0828], + [-0.0360, -0.0381, -0.0197, ..., -0.0263, -0.0200, -0.0136], + [ 0.0189, -0.0190, 0.0370, ..., -0.0854, -0.0829, -0.0725], + ..., + [-0.0038, -0.0300, 0.0432, ..., -0.0228, -0.0334, -0.0301], + [-0.0102, -0.0228, -0.0010, ..., -0.0515, 0.0028, -0.0524], + [ 0.0654, 0.1885, 0.0217, ..., 0.0923, 0.0488, 0.1162]], + device='cuda:0'), grad: tensor([[-3.7253e-09, -3.4925e-09, -1.8626e-08, ..., -5.1223e-09, + -2.2585e-08, -5.1223e-09], + [-7.6834e-09, -4.6566e-09, -5.3551e-08, ..., 2.3283e-09, + -5.0291e-08, -3.0268e-09], + [ 5.1688e-08, 2.7008e-08, 2.6380e-07, ..., 1.2666e-07, + 3.8091e-07, 1.4249e-07], + ..., + [ 1.4435e-08, 8.8476e-09, 9.0804e-08, ..., 8.8476e-09, + 9.7556e-08, 1.7462e-08], + [-1.3737e-08, -6.9849e-09, -8.3586e-08, ..., -1.1409e-08, + -9.4064e-08, -2.3516e-08], + [ 1.1642e-08, 6.5193e-09, 6.9384e-08, ..., 2.3516e-08, + 9.2434e-08, 3.0734e-08]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.2246, -0.0989, 0.0374, 0.0097, 0.5275, -0.1382, -0.5680], + device='cuda:0'), grad: tensor([-7.2876e-08, -1.2759e-07, 6.0722e-07, -5.7183e-07, 2.2934e-07, + -1.8883e-07, 1.6927e-07], device='cuda:0') +249 +0.0012346426699819469 +changing lr +epoch 54, time 350.60, cls_loss 0.0113 cls_loss_mapping 0.1171 cls_loss_causal 0.6751 re_mapping 0.0298 re_causal 0.0561 /// teacc 92.40 lr 0.00109084 +Epoch 56, weight, value: tensor([[ 0.0207, 0.0252, -0.0373, ..., -0.0667, -0.1117, -0.0827], + [-0.0360, -0.0381, -0.0198, ..., -0.0263, -0.0201, -0.0137], + [ 0.0189, -0.0189, 0.0372, ..., -0.0851, -0.0823, -0.0722], + ..., + [-0.0037, -0.0299, 0.0433, ..., -0.0228, -0.0333, -0.0301], + [-0.0102, -0.0228, -0.0010, ..., -0.0514, 0.0028, -0.0524], + [ 0.0653, 0.1882, 0.0216, ..., 0.0922, 0.0486, 0.1161]], + device='cuda:0'), grad: tensor([[-1.0012e-08, -6.0536e-09, -3.8184e-08, ..., -4.1910e-09, + -2.0023e-08, -4.8894e-09], + [-1.8626e-09, 0.0000e+00, -2.3516e-08, ..., 0.0000e+00, + -1.7462e-08, -4.6566e-10], + [ 4.8894e-09, 2.7940e-09, 1.8394e-08, ..., 2.0955e-09, + 9.7789e-09, 2.3283e-09], + ..., + [ 4.6566e-09, 1.8626e-09, 3.2596e-08, ..., 1.3970e-09, + 2.1420e-08, 1.8626e-09], + [ 4.6566e-10, 2.3283e-10, 1.8626e-09, ..., 2.3283e-10, + 1.1642e-09, 4.6566e-10], + [ 1.3970e-09, 9.3132e-10, 5.5879e-09, ..., 6.9849e-10, + 3.0268e-09, 6.9849e-10]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.2242, -0.0992, 0.0382, 0.0095, 0.5265, -0.1378, -0.5674], + device='cuda:0'), grad: tensor([-3.2573e-07, -1.0384e-07, 1.5576e-07, 3.0268e-09, 1.9046e-07, + 1.3504e-08, 4.7730e-08], device='cuda:0') +249 +0.0010908425876598518 +changing lr +epoch 55, time 347.23, cls_loss 0.0155 cls_loss_mapping 0.1176 cls_loss_causal 0.7039 re_mapping 0.0292 re_causal 0.0602 /// teacc 91.81 lr 0.00095492 +Epoch 57, weight, value: tensor([[ 0.0207, 0.0251, -0.0374, ..., -0.0666, -0.1116, -0.0826], + [-0.0359, -0.0380, -0.0198, ..., -0.0263, -0.0201, -0.0137], + [ 0.0188, -0.0189, 0.0371, ..., -0.0850, -0.0822, -0.0722], + ..., + [-0.0037, -0.0298, 0.0435, ..., -0.0227, -0.0330, -0.0300], + [-0.0106, -0.0229, -0.0022, ..., -0.0516, 0.0016, -0.0526], + [ 0.0652, 0.1880, 0.0215, ..., 0.0920, 0.0486, 0.1159]], + device='cuda:0'), grad: tensor([[-3.4750e-05, -3.1918e-05, -1.2326e-04, ..., -5.0396e-05, + -1.2255e-04, -5.0694e-05], + [ 6.5193e-08, 1.5763e-07, -2.6054e-07, ..., 1.1874e-07, + -1.5344e-07, 9.9652e-08], + [ 3.3677e-06, 3.0696e-06, 1.2785e-05, ..., 5.1595e-06, + 1.2890e-05, 5.2527e-06], + ..., + [ 2.9519e-05, 2.7031e-05, 1.0395e-04, ..., 4.2289e-05, + 1.0300e-04, 4.2528e-05], + [ 1.1106e-07, 1.3853e-07, -4.8429e-08, ..., 1.5693e-07, + -1.3364e-07, -2.3283e-10], + [ 1.0636e-06, 9.4343e-07, 4.4405e-06, ..., 1.7015e-06, + 4.5374e-06, 1.8021e-06]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.2236, -0.0991, 0.0380, 0.0114, 0.5264, -0.1394, -0.5667], + device='cuda:0'), grad: tensor([-7.5626e-04, -9.3132e-10, 7.8321e-05, 1.4409e-05, 6.3515e-04, + 1.3635e-06, 2.6494e-05], device='cuda:0') +249 +0.000954915028125264 +changing lr +epoch 56, time 347.41, cls_loss 0.0119 cls_loss_mapping 0.1096 cls_loss_causal 0.6021 re_mapping 0.0282 re_causal 0.0535 /// teacc 92.98 lr 0.00082713 +Epoch 58, weight, value: tensor([[ 0.0207, 0.0251, -0.0369, ..., -0.0663, -0.1110, -0.0823], + [-0.0359, -0.0380, -0.0198, ..., -0.0263, -0.0202, -0.0137], + [ 0.0188, -0.0189, 0.0372, ..., -0.0849, -0.0819, -0.0721], + ..., + [-0.0037, -0.0298, 0.0430, ..., -0.0228, -0.0334, -0.0301], + [-0.0105, -0.0229, -0.0022, ..., -0.0515, 0.0016, -0.0525], + [ 0.0651, 0.1877, 0.0215, ..., 0.0919, 0.0484, 0.1157]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 6.9849e-10, 2.5611e-09, ..., 2.3283e-10, + 2.0955e-09, 2.3283e-10], + [ 2.0955e-09, 2.3283e-09, 7.6834e-09, ..., 6.9849e-10, + 6.5193e-09, 9.3132e-10], + [-1.6531e-08, -1.6997e-08, -5.6112e-08, ..., -4.6566e-09, + -4.7032e-08, -6.5193e-09], + ..., + [ 1.8626e-09, 1.8626e-09, 6.5193e-09, ..., 6.9849e-10, + 5.5879e-09, 9.3132e-10], + [ 7.2177e-09, 7.4506e-09, 2.4680e-08, ..., 2.0955e-09, + 2.0722e-08, 3.0268e-09], + [ 6.9849e-10, 6.9849e-10, 2.7940e-09, ..., 2.3283e-10, + 2.5611e-09, 4.6566e-10]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.2244, -0.0993, 0.0384, 0.0119, 0.5244, -0.1393, -0.5663], + device='cuda:0'), grad: tensor([ 5.3551e-09, 1.6298e-08, -1.1828e-07, 3.1665e-08, 1.4901e-08, + 5.2154e-08, 6.7521e-09], device='cuda:0') +249 +0.0008271337313934874 +changing lr +epoch 57, time 376.49, cls_loss 0.0116 cls_loss_mapping 0.1136 cls_loss_causal 0.6296 re_mapping 0.0275 re_causal 0.0539 /// teacc 92.40 lr 0.00070776 +Epoch 59, weight, value: tensor([[ 0.0208, 0.0253, -0.0363, ..., -0.0659, -0.1101, -0.0819], + [-0.0360, -0.0381, -0.0201, ..., -0.0264, -0.0205, -0.0138], + [ 0.0188, -0.0189, 0.0371, ..., -0.0849, -0.0819, -0.0720], + ..., + [-0.0037, -0.0298, 0.0429, ..., -0.0228, -0.0335, -0.0301], + [-0.0105, -0.0228, -0.0022, ..., -0.0515, 0.0016, -0.0524], + [ 0.0650, 0.1875, 0.0214, ..., 0.0918, 0.0483, 0.1155]], + device='cuda:0'), grad: tensor([[ 2.8219e-07, 3.6857e-07, 2.9244e-06, ..., -5.0757e-07, + 2.0470e-06, -4.2375e-07], + [-8.3912e-07, -7.2038e-07, -7.1563e-06, ..., -8.2701e-07, + -6.3777e-06, -9.5321e-07], + [-2.9076e-06, -2.4326e-06, -1.1869e-05, ..., -3.7178e-06, + -1.2442e-05, -4.3623e-06], + ..., + [ 2.9034e-07, 2.2585e-07, 2.2016e-06, ..., 9.3365e-07, + 2.5388e-06, 9.7509e-07], + [ 2.8107e-06, 2.2501e-06, 1.1228e-05, ..., 2.5090e-06, + 1.0677e-05, 3.0845e-06], + [ 1.0035e-07, 6.4261e-08, 6.9477e-07, ..., 3.0361e-07, + 7.8604e-07, 3.1339e-07]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.2261, -0.1007, 0.0381, 0.0114, 0.5241, -0.1388, -0.5660], + device='cuda:0'), grad: tensor([ 1.1176e-05, -3.3289e-05, -3.5226e-05, 9.2983e-06, 1.0848e-05, + 3.3617e-05, 3.5930e-06], device='cuda:0') +249 +0.00070775603199067 +changing lr +epoch 58, time 348.35, cls_loss 0.0117 cls_loss_mapping 0.1144 cls_loss_causal 0.6328 re_mapping 0.0285 re_causal 0.0533 /// teacc 92.40 lr 0.00059702 +Epoch 60, weight, value: tensor([[ 0.0208, 0.0253, -0.0363, ..., -0.0659, -0.1100, -0.0818], + [-0.0360, -0.0380, -0.0203, ..., -0.0264, -0.0206, -0.0138], + [ 0.0187, -0.0189, 0.0369, ..., -0.0849, -0.0820, -0.0720], + ..., + [-0.0037, -0.0298, 0.0431, ..., -0.0228, -0.0332, -0.0300], + [-0.0106, -0.0228, -0.0023, ..., -0.0515, 0.0015, -0.0524], + [ 0.0650, 0.1874, 0.0213, ..., 0.0917, 0.0482, 0.1154]], + device='cuda:0'), grad: tensor([[ 1.0012e-08, 7.2177e-09, 2.1653e-08, ..., 3.2596e-09, + 1.8394e-08, 4.8894e-09], + [ 6.9849e-10, 0.0000e+00, 1.1874e-08, ..., 2.7940e-09, + 1.2806e-08, 3.7253e-09], + [ 1.8626e-08, 0.0000e+00, 2.6240e-07, ..., 9.2434e-08, + 3.1875e-07, 1.0966e-07], + ..., + [ 3.0268e-09, 0.0000e+00, 5.2387e-08, ..., 1.4668e-08, + 5.8906e-08, 1.8394e-08], + [-2.4447e-08, 0.0000e+00, -3.6880e-07, ..., -1.1991e-07, + -4.3539e-07, -1.4459e-07], + [-9.3132e-09, -7.9162e-09, 3.7253e-09, ..., 3.9581e-09, + 1.1176e-08, 4.4238e-09]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.2262, -0.1010, 0.0375, 0.0116, 0.5244, -0.1389, -0.5656], + device='cuda:0'), grad: tensor([ 5.0524e-08, 7.7067e-08, 1.6429e-06, 9.4064e-08, 3.3434e-07, + -2.3246e-06, 1.1711e-07], device='cuda:0') +249 +0.0005970223407163104 +changing lr +epoch 59, time 345.73, cls_loss 0.0119 cls_loss_mapping 0.1103 cls_loss_causal 0.6765 re_mapping 0.0283 re_causal 0.0535 /// teacc 91.81 lr 0.00049516 +Epoch 61, weight, value: tensor([[ 0.0208, 0.0252, -0.0365, ..., -0.0659, -0.1102, -0.0818], + [-0.0359, -0.0380, -0.0201, ..., -0.0263, -0.0205, -0.0138], + [ 0.0187, -0.0189, 0.0370, ..., -0.0848, -0.0819, -0.0719], + ..., + [-0.0036, -0.0297, 0.0436, ..., -0.0226, -0.0326, -0.0298], + [-0.0105, -0.0228, -0.0023, ..., -0.0515, 0.0014, -0.0524], + [ 0.0649, 0.1872, 0.0213, ..., 0.0916, 0.0481, 0.1153]], + device='cuda:0'), grad: tensor([[ 4.6253e-05, 6.1572e-05, 5.0497e-04, ..., 2.6250e-04, + 5.8270e-04, 1.9765e-04], + [ 4.1835e-06, 3.6862e-06, 2.5585e-05, ..., 4.8950e-06, + 2.0474e-05, 4.6268e-06], + [ 2.0102e-05, 1.8865e-05, 1.2565e-04, ..., 3.6746e-05, + 1.1319e-04, 3.0190e-05], + ..., + [-9.2268e-05, -1.0306e-04, -7.8583e-04, ..., -3.2687e-04, + -8.1778e-04, -2.5463e-04], + [ 1.0476e-05, 8.9481e-06, 6.1095e-05, ..., 9.4324e-06, + 4.6521e-05, 9.6112e-06], + [ 7.7114e-06, 6.6534e-06, 4.5508e-05, ..., 7.6964e-06, + 3.5346e-05, 7.5698e-06]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.2248, -0.1002, 0.0378, 0.0107, 0.5253, -0.1390, -0.5653], + device='cuda:0'), grad: tensor([ 0.0040, 0.0001, 0.0008, 0.0001, -0.0057, 0.0003, 0.0003], + device='cuda:0') +249 +0.0004951556604879052 +changing lr +epoch 60, time 352.12, cls_loss 0.0114 cls_loss_mapping 0.1141 cls_loss_causal 0.6381 re_mapping 0.0275 re_causal 0.0531 /// teacc 92.98 lr 0.00040236 +Epoch 62, weight, value: tensor([[ 0.0208, 0.0252, -0.0365, ..., -0.0659, -0.1102, -0.0818], + [-0.0359, -0.0379, -0.0199, ..., -0.0263, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0848, -0.0819, -0.0719], + ..., + [-0.0036, -0.0297, 0.0433, ..., -0.0226, -0.0328, -0.0299], + [-0.0105, -0.0228, -0.0023, ..., -0.0514, 0.0014, -0.0524], + [ 0.0649, 0.1871, 0.0213, ..., 0.0915, 0.0481, 0.1153]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 3.7253e-09, 1.2573e-07, ..., 2.3283e-09, + 8.9407e-08, 8.8476e-09], + [-1.2107e-08, -4.1910e-09, -1.4901e-07, ..., -1.3970e-09, + -1.0338e-07, -8.8476e-09], + [ 4.6566e-10, 0.0000e+00, 6.0536e-09, ..., 1.8626e-09, + 6.5193e-09, 1.8626e-09], + ..., + [ 2.7940e-09, 9.3132e-10, 3.4459e-08, ..., 1.8626e-09, + 2.6543e-08, 3.7253e-09], + [-3.7253e-09, 0.0000e+00, -4.6566e-08, ..., -1.4435e-08, + -5.3085e-08, -1.5832e-08], + [ 9.3132e-10, 0.0000e+00, 1.4435e-08, ..., 4.1910e-09, + 1.6298e-08, 4.6566e-09]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.2244, -0.0993, 0.0373, 0.0113, 0.5244, -0.1389, -0.5651], + device='cuda:0'), grad: tensor([ 1.2144e-06, -1.4575e-06, 4.0513e-08, 1.2666e-07, 3.1944e-07, + -2.9523e-07, 9.4064e-08], device='cuda:0') +249 +0.00040236113724274745 +changing lr +epoch 61, time 346.83, cls_loss 0.0095 cls_loss_mapping 0.1049 cls_loss_causal 0.6761 re_mapping 0.0273 re_causal 0.0569 /// teacc 92.40 lr 0.00031883 +Epoch 63, weight, value: tensor([[ 0.0208, 0.0252, -0.0365, ..., -0.0659, -0.1101, -0.0817], + [-0.0359, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0847, -0.0819, -0.0719], + ..., + [-0.0036, -0.0297, 0.0434, ..., -0.0226, -0.0328, -0.0298], + [-0.0105, -0.0228, -0.0023, ..., -0.0514, 0.0014, -0.0523], + [ 0.0648, 0.1870, 0.0212, ..., 0.0915, 0.0481, 0.1152]], + device='cuda:0'), grad: tensor([[ 3.4785e-07, 5.5879e-08, 3.0063e-06, ..., 2.0433e-06, + 4.2580e-06, 1.8263e-06], + [ 6.0536e-09, -3.2596e-09, 4.8429e-08, ..., 5.4948e-08, + 9.0804e-08, 4.7497e-08], + [ 2.3749e-08, 1.3970e-09, 2.2352e-07, ..., 1.5460e-07, + 3.2084e-07, 1.3690e-07], + ..., + [-4.3260e-07, -6.4727e-08, -3.7365e-06, ..., -2.5649e-06, + -5.3197e-06, -2.2911e-06], + [ 1.6764e-08, 3.2596e-09, 1.4296e-07, ..., 9.6858e-08, + 2.0163e-07, 8.6613e-08], + [ 1.5832e-08, 3.2596e-09, 1.3271e-07, ..., 8.9407e-08, + 1.8626e-07, 8.0094e-08]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.2242, -0.0994, 0.0373, 0.0114, 0.5243, -0.1388, -0.5648], + device='cuda:0'), grad: tensor([ 1.7613e-05, 2.6822e-07, 1.3281e-06, 1.0775e-06, -2.1890e-05, + 8.3307e-07, 7.7114e-07], device='cuda:0') +249 +0.00031882564680131423 +changing lr +epoch 62, time 350.05, cls_loss 0.0106 cls_loss_mapping 0.1023 cls_loss_causal 0.6294 re_mapping 0.0283 re_causal 0.0536 /// teacc 94.15 lr 0.00024472 +Epoch 64, weight, value: tensor([[ 0.0208, 0.0252, -0.0364, ..., -0.0658, -0.1100, -0.0816], + [-0.0359, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0847, -0.0818, -0.0719], + ..., + [-0.0036, -0.0297, 0.0433, ..., -0.0226, -0.0328, -0.0299], + [-0.0105, -0.0228, -0.0023, ..., -0.0514, 0.0014, -0.0523], + [ 0.0648, 0.1869, 0.0212, ..., 0.0914, 0.0480, 0.1152]], + device='cuda:0'), grad: tensor([[ 6.7167e-06, 5.4836e-06, 5.4657e-05, ..., 2.2575e-05, + 6.0171e-05, 2.3410e-05], + [-4.5478e-05, -3.7491e-05, -3.7265e-04, ..., -1.5616e-04, + -4.1270e-04, -1.6129e-04], + [-3.0193e-06, -2.0340e-06, -2.1368e-05, ..., -5.6773e-06, + -1.9714e-05, -6.5938e-06], + ..., + [ 3.9160e-05, 3.2187e-05, 3.1996e-04, ..., 1.3328e-04, + 3.5357e-04, 1.3781e-04], + [ 1.5711e-06, 1.0915e-06, 1.1586e-05, ..., 3.5428e-06, + 1.1206e-05, 3.9823e-06], + [ 3.3574e-07, 2.4680e-07, 2.5537e-06, ..., 8.7265e-07, + 2.5854e-06, 9.4902e-07]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.2244, -0.0992, 0.0373, 0.0111, 0.5240, -0.1387, -0.5647], + device='cuda:0'), grad: tensor([ 3.9625e-04, -2.6951e-03, -1.6081e-04, 3.8087e-05, 2.3155e-03, + 8.8155e-05, 1.9044e-05], device='cuda:0') +249 +0.0002447174185242325 +changing lr +epoch 63, time 349.59, cls_loss 0.0104 cls_loss_mapping 0.1057 cls_loss_causal 0.6085 re_mapping 0.0285 re_causal 0.0531 /// teacc 92.40 lr 0.00018019 +Epoch 65, weight, value: tensor([[ 0.0208, 0.0252, -0.0364, ..., -0.0658, -0.1100, -0.0816], + [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0202, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0847, -0.0817, -0.0718], + ..., + [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0329, -0.0299], + [-0.0105, -0.0228, -0.0022, ..., -0.0514, 0.0015, -0.0523], + [ 0.0648, 0.1869, 0.0212, ..., 0.0914, 0.0480, 0.1151]], + device='cuda:0'), grad: tensor([[-2.6748e-06, -2.0582e-06, -1.2316e-05, ..., -4.2096e-06, + -1.1683e-05, -3.3285e-06], + [ 5.5740e-07, 4.2049e-07, 2.5891e-06, ..., 9.0618e-07, + 2.4661e-06, 7.1293e-07], + [ 8.3819e-09, 7.4506e-09, 3.5390e-08, ..., 9.3132e-09, + 3.1665e-08, 7.4506e-09], + ..., + [ 2.0694e-06, 1.5972e-06, 9.5293e-06, ..., 3.2485e-06, + 9.0301e-06, 2.5686e-06], + [ 1.3970e-09, 9.3132e-10, 5.5879e-09, ..., 1.8626e-09, + 5.1223e-09, 1.3970e-09], + [ 1.6298e-08, 1.5832e-08, 6.5193e-08, ..., 1.3504e-08, + 5.7276e-08, 1.2107e-08]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.2243, -0.0992, 0.0374, 0.0111, 0.5236, -0.1383, -0.5645], + device='cuda:0'), grad: tensor([-3.7402e-05, 7.9349e-06, 9.6392e-08, 2.9430e-07, 2.8893e-05, + 1.6764e-08, 1.6717e-07], device='cuda:0') +249 +0.0001801856965207339 +changing lr +epoch 64, time 344.76, cls_loss 0.0114 cls_loss_mapping 0.1071 cls_loss_causal 0.6610 re_mapping 0.0270 re_causal 0.0537 /// teacc 92.98 lr 0.00012536 +Epoch 66, weight, value: tensor([[ 0.0208, 0.0252, -0.0364, ..., -0.0658, -0.1099, -0.0816], + [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], + ..., + [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0329, -0.0299], + [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], + [ 0.0648, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1151]], + device='cuda:0'), grad: tensor([[-9.0804e-07, -9.2201e-08, -9.3803e-06, ..., -6.5751e-06, + -1.2361e-05, -6.6906e-06], + [ 1.5413e-07, 3.9581e-08, 1.1614e-06, ..., 7.3295e-07, + 1.4473e-06, 7.4739e-07], + [-1.1558e-06, -5.9744e-07, -3.2261e-06, ..., -6.3749e-07, + -2.5984e-06, -6.8406e-07], + ..., + [ 5.5227e-07, 1.1967e-07, 4.5411e-06, ..., 2.9691e-06, + 5.7630e-06, 3.0249e-06], + [ 5.0338e-07, 2.4168e-07, 1.7425e-06, ..., 5.7649e-07, + 1.6401e-06, 6.0070e-07], + [ 3.6415e-07, 1.2945e-07, 2.0899e-06, ..., 1.1530e-06, + 2.4363e-06, 1.1791e-06]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.2242, -0.0993, 0.0375, 0.0111, 0.5233, -0.1382, -0.5644], + device='cuda:0'), grad: tensor([-7.1764e-05, 8.9332e-06, -2.5719e-05, 2.3767e-05, 3.4869e-05, + 1.3739e-05, 1.6198e-05], device='cuda:0') +249 +0.000125360439090882 +changing lr +epoch 65, time 347.47, cls_loss 0.0074 cls_loss_mapping 0.1005 cls_loss_causal 0.6283 re_mapping 0.0267 re_causal 0.0555 /// teacc 92.98 lr 0.00008035 +Epoch 67, weight, value: tensor([[ 0.0208, 0.0252, -0.0363, ..., -0.0658, -0.1099, -0.0816], + [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], + ..., + [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0329, -0.0299], + [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], + [ 0.0648, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1151]], + device='cuda:0'), grad: tensor([[-8.9873e-08, -8.8476e-08, -5.7323e-07, ..., -3.3528e-08, + -5.2247e-07, -5.7742e-08], + [-1.3970e-08, -2.0955e-08, -9.2667e-08, ..., -1.9558e-08, + -8.4750e-08, -1.8626e-08], + [ 3.7253e-09, 2.7940e-09, 2.0489e-08, ..., 1.3970e-09, + 1.7695e-08, 2.3283e-09], + ..., + [ 9.4529e-08, 9.4529e-08, 5.8860e-07, ..., 4.7032e-08, + 5.3411e-07, 6.8918e-08], + [-1.9092e-08, -2.7940e-09, -6.6590e-08, ..., -2.7940e-09, + -4.8894e-08, -8.8476e-09], + [ 2.2817e-08, 1.3039e-08, 1.1083e-07, ..., 9.3132e-09, + 9.4995e-08, 1.5367e-08]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.2242, -0.0993, 0.0374, 0.0111, 0.5233, -0.1382, -0.5644], + device='cuda:0'), grad: tensor([-2.8703e-06, -4.4191e-07, 8.5682e-08, 8.1025e-08, 2.8610e-06, + -9.1735e-08, 4.0047e-07], device='cuda:0') +249 +8.03520570068517e-05 +changing lr +epoch 66, time 350.79, cls_loss 0.0102 cls_loss_mapping 0.1035 cls_loss_causal 0.6503 re_mapping 0.0282 re_causal 0.0544 /// teacc 92.98 lr 0.00004525 +Epoch 68, weight, value: tensor([[ 0.0208, 0.0252, -0.0363, ..., -0.0657, -0.1098, -0.0816], + [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], + ..., + [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0330, -0.0299], + [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], + [ 0.0647, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1151]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, -1.6578e-07, 3.4040e-07, ..., 8.8708e-07, + 1.9390e-06, 1.2936e-06], + [ 5.5656e-06, 2.6505e-06, 3.2008e-05, ..., 6.8583e-06, + 3.1292e-05, 9.4473e-06], + [ 1.8822e-06, -2.0154e-06, -5.2452e-06, ..., -6.8061e-06, + -1.0513e-05, -6.3330e-06], + ..., + [ 7.3723e-06, 3.7421e-06, 4.2647e-05, ..., 8.7246e-06, + 4.0591e-05, 1.1817e-05], + [-2.2918e-05, -8.0541e-06, -1.1629e-04, ..., -1.9357e-05, + -1.0854e-04, -2.9698e-05], + [ 6.1579e-06, 2.8070e-06, 3.4779e-05, ..., 7.3351e-06, + 3.3885e-05, 1.0177e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.2243, -0.0993, 0.0374, 0.0110, 0.5232, -0.1381, -0.5643], + device='cuda:0'), grad: tensor([ 4.2133e-06, 1.1146e-04, -1.1331e-04, 4.3392e-05, 1.4925e-04, + -3.1281e-04, 1.1742e-04], device='cuda:0') +249 +4.5251191160326525e-05 +changing lr +epoch 67, time 348.82, cls_loss 0.0106 cls_loss_mapping 0.1088 cls_loss_causal 0.6864 re_mapping 0.0277 re_causal 0.0531 /// teacc 94.15 lr 0.00002013 +Epoch 69, weight, value: tensor([[ 0.0208, 0.0252, -0.0362, ..., -0.0657, -0.1098, -0.0816], + [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], + ..., + [-0.0037, -0.0297, 0.0431, ..., -0.0227, -0.0330, -0.0299], + [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], + [ 0.0647, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1150]], + device='cuda:0'), grad: tensor([[-1.3970e-09, -1.8626e-09, -8.8476e-09, ..., -2.3283e-09, + -8.3819e-09, -2.3283e-09], + [ 2.7940e-09, 1.3970e-09, 9.3132e-09, ..., 1.8626e-09, + 7.4506e-09, 2.3283e-09], + [ 6.0536e-09, 2.7940e-09, 2.1886e-08, ..., 3.7253e-09, + 1.7695e-08, 5.1223e-09], + ..., + [ 1.9558e-08, 9.7789e-09, 7.5437e-08, ..., 1.3970e-08, + 6.1933e-08, 1.7229e-08], + [-2.1560e-07, -9.8720e-08, -7.9814e-07, ..., -1.4249e-07, + -6.4913e-07, -1.8161e-07], + [ 3.6322e-08, 1.6764e-08, 1.3597e-07, ..., 2.4214e-08, + 1.1083e-07, 3.1199e-08]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.2244, -0.0993, 0.0374, 0.0110, 0.5230, -0.1381, -0.5643], + device='cuda:0'), grad: tensor([-7.5437e-08, 1.5367e-08, 6.5193e-08, 1.6261e-06, 2.5705e-07, + -2.2799e-06, 3.9907e-07], device='cuda:0') +249 +2.0128530023804673e-05 +changing lr +epoch 68, time 348.35, cls_loss 0.0081 cls_loss_mapping 0.0988 cls_loss_causal 0.6166 re_mapping 0.0275 re_causal 0.0561 /// teacc 92.98 lr 0.00000503 +Epoch 70, weight, value: tensor([[ 0.0208, 0.0252, -0.0362, ..., -0.0657, -0.1098, -0.0816], + [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], + [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], + ..., + [-0.0036, -0.0297, 0.0431, ..., -0.0226, -0.0330, -0.0299], + [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], + [ 0.0647, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1150]], + device='cuda:0'), grad: tensor([[ 1.6321e-07, 1.0757e-07, 8.1025e-07, ..., 8.5915e-08, + 8.3540e-07, 1.4552e-07], + [-1.7928e-07, -8.5915e-08, -8.9034e-07, ..., -8.5915e-08, + -9.1083e-07, -1.5227e-07], + [ 6.4261e-08, 6.8452e-08, 3.1642e-07, ..., 4.1211e-08, + 3.3132e-07, 6.4494e-08], + ..., + [-6.8918e-08, -9.8720e-08, -3.4203e-07, ..., -4.8662e-08, + -3.6345e-07, -7.3109e-08], + [ 1.4435e-08, 5.8208e-09, 7.5903e-08, ..., 4.6566e-09, + 7.7533e-08, 1.0012e-08], + [ 4.1910e-09, 3.0268e-09, 1.9558e-08, ..., 3.0268e-09, + 2.0023e-08, 4.4238e-09]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.2244, -0.0993, 0.0374, 0.0110, 0.5230, -0.1381, -0.5643], + device='cuda:0'), grad: tensor([ 1.9502e-06, -2.0340e-06, 8.4704e-07, 5.7044e-08, -1.0040e-06, + 1.7532e-07, 4.5868e-08], device='cuda:0') +249 +5.034667293427056e-06 +changing lr +epoch 69, time 350.06, cls_loss 0.0124 cls_loss_mapping 0.1044 cls_loss_causal 0.6741 re_mapping 0.0288 re_causal 0.0534 /// teacc 91.81 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'source_domain': 'photo', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1/photo_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['photo', 'art_painting', 'cartoon', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + photo art_painting cartoon sketch Avg +w/o do (original x) 99.461078 55.761719 49.616041 61.28786 55.555206 + photo art_painting cartoon sketch Avg +do 98.802395 57.617188 50.0 60.269789 55.962325 diff --git a/Meta-causal/code-withStyleAttack/64947.error b/Meta-causal/code-withStyleAttack/64947.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/64947.log b/Meta-causal/code-withStyleAttack/64947.log new file mode 100644 index 0000000000000000000000000000000000000000..6c545012ab5f4d2f633171db662ccc1fee29cadd --- /dev/null +++ b/Meta-causal/code-withStyleAttack/64947.log @@ -0,0 +1,1939 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[-0.0208, -0.0220, 0.0164, ..., 0.0019, 0.0060, 0.0055], + [ 0.0100, 0.0072, 0.0083, ..., -0.0045, -0.0209, -0.0166], + [ 0.0103, 0.0061, 0.0040, ..., -0.0083, -0.0102, 0.0187], + ..., + [ 0.0074, 0.0004, -0.0085, ..., -0.0055, 0.0045, 0.0045], + [ 0.0123, -0.0002, -0.0053, ..., 0.0055, -0.0196, -0.0166], + [-0.0206, 0.0139, 0.0016, ..., -0.0135, 0.0219, -0.0163]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0178, 0.0114, 0.0058, -0.0179, 0.0187, 0.0121, -0.0002], + device='cuda:0'), grad: None +588 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 808.00, cls_loss 7.4242 cls_loss_mapping 1.5280 cls_loss_causal 1.6920 re_mapping 0.2361 re_causal 0.2355 /// teacc 58.79 lr 0.00999497 +Epoch 2, weight, value: tensor([[-0.0031, -0.0020, 0.0128, ..., 0.0158, 0.0300, 0.0269], + [ 0.1318, 0.1076, 0.0753, ..., -0.0284, -0.0293, -0.0291], + [-0.0311, -0.0348, 0.0314, ..., 0.0709, 0.0060, 0.0512], + ..., + [-0.1643, -0.1202, -0.1104, ..., 0.0602, 0.0656, 0.0368], + [ 0.1364, 0.0869, 0.0833, ..., -0.0678, -0.0868, -0.0781], + [-0.0035, 0.0195, -0.0150, ..., -0.0955, -0.0366, -0.0639]], + device='cuda:0'), grad: tensor([[ 1.4246e-01, 4.9591e-02, 4.4464e-02, ..., 4.6326e-02, + 4.4983e-02, 1.6815e-02], + [-4.0009e-02, -1.2909e-02, -7.5607e-03, ..., -1.5297e-02, + -1.6342e-02, -6.8398e-03], + [-9.7168e-02, -3.3356e-02, -3.5645e-02, ..., -3.4088e-02, + -2.8824e-02, -9.9335e-03], + ..., + [-3.7323e-02, -1.3939e-02, -1.0986e-02, ..., -6.9771e-03, + -8.4991e-03, -2.9697e-03], + [ 1.8387e-03, 6.6328e-04, 5.3406e-04, ..., 6.3944e-04, + 6.3848e-04, 2.6083e-04], + [ 1.7262e-04, 5.7131e-05, 5.0813e-05, ..., 5.4508e-05, + 4.8101e-05, 1.6570e-05]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0211, -0.0330, 0.0198, 0.0471, -0.0429, 0.0361, 0.0054], + device='cuda:0'), grad: tensor([ 0.3125, -0.1070, -0.2090, 0.0684, -0.0692, 0.0041, 0.0004], + device='cuda:0') +588 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 814.99, cls_loss 0.9246 cls_loss_mapping 0.9319 cls_loss_causal 1.3704 re_mapping 0.0927 re_causal 0.0920 /// teacc 63.57 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 0.0241, 0.0225, 0.0199, ..., 0.0375, 0.0633, 0.0597], + [ 0.0934, 0.0721, 0.0373, ..., -0.0349, -0.0409, -0.0381], + [-0.0102, -0.0210, 0.0503, ..., 0.0576, -0.0119, 0.0319], + ..., + [-0.1723, -0.1231, -0.1008, ..., 0.0570, 0.0629, 0.0307], + [ 0.1450, 0.0950, 0.0944, ..., -0.0800, -0.0971, -0.0869], + [-0.0020, 0.0219, -0.0180, ..., -0.0981, -0.0372, -0.0616]], + device='cuda:0'), grad: tensor([[-4.3732e-02, -1.3214e-02, -7.5569e-03, ..., -2.2293e-02, + -2.5269e-02, -1.8829e-02], + [ 4.2152e-03, 1.6422e-03, 1.0777e-03, ..., 1.3571e-03, + 1.5516e-03, 1.1444e-03], + [ 1.9958e-02, 5.8060e-03, 3.2330e-03, ..., 1.0559e-02, + 1.1971e-02, 8.9111e-03], + ..., + [ 1.5152e-02, 4.4861e-03, 2.5291e-03, ..., 8.0414e-03, + 9.1019e-03, 6.8016e-03], + [ 2.3627e-04, 6.9618e-05, 3.9160e-05, ..., 1.2517e-04, + 1.4174e-04, 1.0562e-04], + [ 2.0713e-05, 4.4778e-06, 2.8610e-06, ..., 1.4633e-05, + 1.6212e-05, 1.1973e-05]], device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0072, -0.0282, 0.0243, 0.0413, -0.0554, 0.0151, 0.0064], + device='cuda:0'), grad: tensor([-1.1499e-01, 9.4757e-03, 5.3375e-02, 1.1086e-02, 4.0375e-02, + 6.2943e-04, 6.2585e-05], device='cuda:0') +588 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 817.05, cls_loss 0.5564 cls_loss_mapping 0.5713 cls_loss_causal 1.1146 re_mapping 0.0838 re_causal 0.0829 /// teacc 74.37 lr 0.00995475 +Epoch 4, weight, value: tensor([[ 0.0254, 0.0233, 0.0175, ..., 0.0372, 0.0729, 0.0651], + [ 0.0889, 0.0631, 0.0286, ..., -0.0262, -0.0341, -0.0294], + [-0.0140, -0.0197, 0.0500, ..., 0.0424, -0.0304, 0.0150], + ..., + [-0.1693, -0.1204, -0.0955, ..., 0.0552, 0.0656, 0.0328], + [ 0.1554, 0.1025, 0.1018, ..., -0.0833, -0.1015, -0.0903], + [-0.0044, 0.0212, -0.0197, ..., -0.0954, -0.0348, -0.0581]], + device='cuda:0'), grad: tensor([[-9.0551e-04, -1.7595e-04, -1.8907e-04, ..., -2.0266e-04, + -2.7013e-04, -2.0099e-04], + [ 8.4281e-05, 1.9148e-05, 1.9908e-05, ..., 1.7017e-05, + 2.3603e-05, 1.7703e-05], + [ 5.8413e-05, 1.0587e-05, 1.1586e-05, ..., 1.3597e-05, + 1.7911e-05, 1.3277e-05], + ..., + [ 6.3276e-04, 1.1927e-04, 1.2910e-04, ..., 1.4400e-04, + 1.9073e-04, 1.4174e-04], + [ 1.1668e-05, 2.5332e-06, 2.6636e-06, ..., 2.4624e-06, + 3.3677e-06, 2.5183e-06], + [ 3.7141e-06, 7.7859e-07, 8.2701e-07, ..., 7.9721e-07, + 1.0841e-06, 8.0839e-07]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0101, -0.0139, 0.0060, 0.0420, -0.0526, 0.0196, -0.0008], + device='cuda:0'), grad: tensor([-2.4834e-03, 2.2125e-04, 1.6344e-04, 3.0804e-04, 1.7481e-03, + 3.1203e-05, 1.0014e-05], device='cuda:0') +588 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 816.59, cls_loss 0.4075 cls_loss_mapping 0.3747 cls_loss_causal 0.9471 re_mapping 0.0793 re_causal 0.0785 /// teacc 82.91 lr 0.00991965 +Epoch 5, weight, value: tensor([[ 0.0587, 0.0530, 0.0429, ..., 0.0455, 0.0862, 0.0758], + [ 0.0638, 0.0423, 0.0081, ..., -0.0265, -0.0333, -0.0300], + [-0.0198, -0.0299, 0.0373, ..., 0.0311, -0.0395, 0.0039], + ..., + [-0.1696, -0.1202, -0.0889, ..., 0.0488, 0.0549, 0.0276], + [ 0.1541, 0.1041, 0.1031, ..., -0.0847, -0.1032, -0.0916], + [ 0.0019, 0.0266, -0.0159, ..., -0.0918, -0.0314, -0.0543]], + device='cuda:0'), grad: tensor([[-0.0616, -0.0126, -0.0163, ..., -0.0148, -0.0252, -0.0245], + [ 0.0148, 0.0030, 0.0038, ..., 0.0036, 0.0058, 0.0057], + [ 0.0296, 0.0063, 0.0077, ..., 0.0072, 0.0109, 0.0107], + ..., + [ 0.0063, 0.0008, 0.0018, ..., 0.0015, 0.0047, 0.0045], + [ 0.0024, 0.0005, 0.0006, ..., 0.0006, 0.0010, 0.0010], + [ 0.0011, 0.0002, 0.0003, ..., 0.0003, 0.0004, 0.0004]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0308, -0.0303, 0.0234, 0.0388, -0.0666, 0.0088, 0.0053], + device='cuda:0'), grad: tensor([-0.1545, 0.0386, 0.0775, 0.0186, 0.0105, 0.0064, 0.0030], + device='cuda:0') +588 +0.009919647942993149 +changing lr +epoch 4, time 813.70, cls_loss 0.3134 cls_loss_mapping 0.2826 cls_loss_causal 0.8661 re_mapping 0.0756 re_causal 0.0750 /// teacc 79.65 lr 0.00987464 +Epoch 6, weight, value: tensor([[ 0.0543, 0.0522, 0.0399, ..., 0.0406, 0.0876, 0.0772], + [ 0.0528, 0.0340, 0.0004, ..., -0.0232, -0.0285, -0.0257], + [-0.0175, -0.0272, 0.0388, ..., 0.0211, -0.0511, -0.0094], + ..., + [-0.1690, -0.1242, -0.0903, ..., 0.0546, 0.0608, 0.0338], + [ 0.1551, 0.1075, 0.1070, ..., -0.0853, -0.1042, -0.0924], + [ 0.0127, 0.0348, -0.0096, ..., -0.0896, -0.0303, -0.0522]], + device='cuda:0'), grad: tensor([[ 2.3499e-03, 1.0033e-03, 9.9850e-04, ..., 1.0929e-03, + 1.1692e-03, 1.2064e-03], + [-2.9125e-03, -1.2484e-03, -1.2417e-03, ..., -1.3580e-03, + -1.4515e-03, -1.4982e-03], + [-1.9181e-04, -7.5772e-06, -1.7881e-05, ..., -2.9355e-05, + -5.0336e-05, -5.4300e-05], + ..., + [ 9.0778e-05, 1.0014e-05, 1.3940e-05, ..., 1.8895e-05, + 2.7582e-05, 2.9400e-05], + [ 2.1681e-06, 8.0466e-07, 7.9907e-07, ..., 8.4005e-07, + 9.4809e-07, 9.8534e-07], + [ 2.9460e-05, 1.1928e-05, 1.1489e-05, ..., 1.1235e-05, + 1.2793e-05, 1.3262e-05]], device='cuda:0') +Epoch 6, bias, value: tensor([ 2.3044e-02, -3.1049e-02, 1.9318e-02, 4.3537e-02, -6.0893e-02, + -9.9219e-05, 1.5851e-02], device='cuda:0'), grad: tensor([ 4.4403e-03, -5.4855e-03, -5.7840e-04, 1.3084e-03, 2.5487e-04, + 4.4256e-06, 5.7220e-05], device='cuda:0') +588 +0.009874639560909117 +changing lr +epoch 5, time 816.31, cls_loss 0.2576 cls_loss_mapping 0.2196 cls_loss_causal 0.7947 re_mapping 0.0733 re_causal 0.0730 /// teacc 68.09 lr 0.00981981 +Epoch 7, weight, value: tensor([[ 0.0654, 0.0580, 0.0449, ..., 0.0393, 0.0916, 0.0802], + [ 0.0489, 0.0304, -0.0033, ..., -0.0207, -0.0244, -0.0211], + [-0.0205, -0.0309, 0.0346, ..., 0.0150, -0.0583, -0.0180], + ..., + [-0.1776, -0.1272, -0.0928, ..., 0.0570, 0.0628, 0.0360], + [ 0.1572, 0.1097, 0.1087, ..., -0.0837, -0.1025, -0.0906], + [ 0.0154, 0.0374, -0.0072, ..., -0.0894, -0.0337, -0.0543]], + device='cuda:0'), grad: tensor([[ 2.8763e-03, 1.4992e-03, 1.5965e-03, ..., 4.7517e-04, + 7.0667e-04, 7.4673e-04], + [-2.7885e-03, -1.4744e-03, -1.5659e-03, ..., -4.5371e-04, + -6.7568e-04, -7.1335e-04], + [-4.0770e-05, -1.8165e-05, -1.9968e-05, ..., -1.0021e-05, + -1.5117e-05, -1.5944e-05], + ..., + [-4.1306e-05, -2.6934e-06, -8.1509e-06, ..., -1.0774e-05, + -1.5028e-05, -1.6659e-05], + [ 1.3821e-06, 5.4017e-07, 5.9605e-07, ..., 2.9802e-07, + 4.5076e-07, 4.8056e-07], + [-8.7321e-06, -5.0887e-06, -4.1500e-06, ..., -3.6508e-07, + -9.2760e-07, -1.0021e-06]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0437, -0.0316, 0.0258, 0.0446, -0.0881, 0.0026, 0.0124], + device='cuda:0'), grad: tensor([ 4.6577e-03, -4.4403e-03, -8.0109e-05, -2.6338e-06, -1.3149e-04, + 2.9169e-06, -1.2100e-05], device='cuda:0') +588 +0.009819814303479266 +changing lr +epoch 6, time 818.25, cls_loss 0.1873 cls_loss_mapping 0.1654 cls_loss_causal 0.7497 re_mapping 0.0694 re_causal 0.0692 /// teacc 82.66 lr 0.00975528 +Epoch 8, weight, value: tensor([[ 0.0525, 0.0526, 0.0366, ..., 0.0360, 0.0902, 0.0794], + [ 0.0595, 0.0406, 0.0080, ..., -0.0211, -0.0242, -0.0207], + [-0.0158, -0.0296, 0.0351, ..., 0.0148, -0.0563, -0.0175], + ..., + [-0.1736, -0.1281, -0.0912, ..., 0.0596, 0.0646, 0.0378], + [ 0.1559, 0.1095, 0.1088, ..., -0.0824, -0.1009, -0.0893], + [ 0.0144, 0.0344, -0.0106, ..., -0.0869, -0.0334, -0.0533]], + device='cuda:0'), grad: tensor([[-0.0395, -0.0124, -0.0165, ..., -0.0090, -0.0090, -0.0092], + [ 0.0177, 0.0055, 0.0073, ..., 0.0041, 0.0041, 0.0042], + [ 0.0131, 0.0041, 0.0054, ..., 0.0029, 0.0029, 0.0030], + ..., + [ 0.0060, 0.0019, 0.0027, ..., 0.0015, 0.0015, 0.0015], + [ 0.0008, 0.0002, 0.0003, ..., 0.0002, 0.0002, 0.0002], + [ 0.0016, 0.0005, 0.0006, ..., 0.0004, 0.0004, 0.0004]], + device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0272, -0.0320, 0.0361, 0.0317, -0.0745, 0.0015, 0.0192], + device='cuda:0'), grad: tensor([-0.0837, 0.0378, 0.0279, 0.0006, 0.0122, 0.0018, 0.0034], + device='cuda:0') +588 +0.009755282581475767 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 819.90, cls_loss 0.1601 cls_loss_mapping 0.1472 cls_loss_causal 0.7095 re_mapping 0.0678 re_causal 0.0680 /// teacc 87.19 lr 0.00968117 +Epoch 9, weight, value: tensor([[ 0.0682, 0.0640, 0.0488, ..., 0.0328, 0.0870, 0.0772], + [ 0.0578, 0.0391, 0.0079, ..., -0.0195, -0.0213, -0.0182], + [-0.0168, -0.0309, 0.0323, ..., 0.0155, -0.0549, -0.0170], + ..., + [-0.1874, -0.1370, -0.0998, ..., 0.0593, 0.0642, 0.0377], + [ 0.1556, 0.1098, 0.1089, ..., -0.0811, -0.0990, -0.0878], + [ 0.0138, 0.0337, -0.0127, ..., -0.0853, -0.0336, -0.0529]], + device='cuda:0'), grad: tensor([[ 2.3198e-04, 4.4197e-05, 4.7863e-05, ..., 8.1241e-05, + 7.5579e-05, 8.2672e-05], + [ 1.6642e-04, 2.4870e-05, 2.9132e-05, ..., 3.4600e-05, + 3.7372e-05, 3.7581e-05], + [-8.1396e-04, -1.6522e-04, -1.7905e-04, ..., -2.9278e-04, + -2.5034e-04, -2.8634e-04], + ..., + [ 2.6441e-04, 6.7830e-05, 7.0870e-05, ..., 1.3185e-04, + 9.2685e-05, 1.1873e-04], + [ 6.4850e-05, 1.1407e-05, 1.2659e-05, ..., 1.8835e-05, + 1.8939e-05, 1.9878e-05], + [ 6.4135e-05, 1.2577e-05, 1.3813e-05, ..., 1.9103e-05, + 1.8761e-05, 2.0012e-05]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0404, -0.0317, 0.0415, 0.0336, -0.0923, 0.0013, 0.0160], + device='cuda:0'), grad: tensor([ 5.8842e-04, 5.0735e-04, -1.9464e-03, 5.7876e-05, 4.4751e-04, + 1.7881e-04, 1.6797e-04], device='cuda:0') +588 +0.009681174353198686 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 816.90, cls_loss 0.0972 cls_loss_mapping 0.1247 cls_loss_causal 0.6614 re_mapping 0.0653 re_causal 0.0658 /// teacc 87.69 lr 0.00959764 +Epoch 10, weight, value: tensor([[ 0.0649, 0.0594, 0.0458, ..., 0.0358, 0.0910, 0.0805], + [ 0.0608, 0.0405, 0.0098, ..., -0.0244, -0.0258, -0.0222], + [-0.0181, -0.0319, 0.0302, ..., 0.0187, -0.0479, -0.0123], + ..., + [-0.1769, -0.1286, -0.0939, ..., 0.0563, 0.0585, 0.0340], + [ 0.1486, 0.1067, 0.1055, ..., -0.0797, -0.0975, -0.0865], + [ 0.0117, 0.0320, -0.0136, ..., -0.0836, -0.0337, -0.0524]], + device='cuda:0'), grad: tensor([[ 1.3912e-04, 1.3083e-05, 3.1382e-05, ..., 1.8790e-05, + 2.7537e-05, 2.7105e-05], + [ 1.4760e-05, 3.1404e-06, 4.3884e-06, ..., 3.0473e-06, + 4.1239e-06, 3.8967e-06], + [-2.3866e-04, -2.5272e-05, -5.5760e-05, ..., -3.4124e-05, + -4.9949e-05, -4.8786e-05], + ..., + [ 6.2108e-05, 7.3090e-06, 1.5132e-05, ..., 8.9258e-06, + 1.3202e-05, 1.2845e-05], + [ 3.4384e-06, 5.6624e-07, 8.8289e-07, ..., 4.7311e-07, + 6.9290e-07, 6.6683e-07], + [ 1.0289e-05, 1.9372e-07, 2.1122e-06, ..., 1.8664e-06, + 2.6971e-06, 2.6338e-06]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0402, -0.0282, 0.0481, 0.0325, -0.0892, -0.0080, 0.0130], + device='cuda:0'), grad: tensor([ 3.9959e-04, 3.6925e-05, -6.7711e-04, 2.5898e-05, 1.7405e-04, + 9.0674e-06, 3.2365e-05], device='cuda:0') +588 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 819.66, cls_loss 0.0605 cls_loss_mapping 0.0894 cls_loss_causal 0.6396 re_mapping 0.0631 re_causal 0.0640 /// teacc 87.94 lr 0.00950484 +Epoch 11, weight, value: tensor([[ 0.0652, 0.0586, 0.0463, ..., 0.0348, 0.0899, 0.0799], + [ 0.0594, 0.0395, 0.0094, ..., -0.0237, -0.0246, -0.0211], + [-0.0221, -0.0334, 0.0274, ..., 0.0183, -0.0476, -0.0128], + ..., + [-0.1754, -0.1277, -0.0936, ..., 0.0539, 0.0566, 0.0324], + [ 0.1493, 0.1067, 0.1054, ..., -0.0776, -0.0948, -0.0839], + [ 0.0100, 0.0321, -0.0137, ..., -0.0817, -0.0339, -0.0521]], + device='cuda:0'), grad: tensor([[ 4.1342e-04, 6.2287e-05, 5.8174e-05, ..., 1.6308e-04, + 2.1768e-04, 2.0707e-04], + [ 5.8487e-06, 1.2424e-06, 1.2908e-06, ..., 1.1362e-06, + 1.2964e-06, 1.3020e-06], + [-1.8632e-04, -3.6508e-05, -4.0621e-05, ..., -2.4974e-05, + -3.2604e-05, -3.2872e-05], + ..., + [-3.2640e-04, -4.5002e-05, -3.9011e-05, ..., -1.5128e-04, + -2.0230e-04, -1.9169e-04], + [ 1.6093e-05, 3.6769e-06, 3.7700e-06, ..., 2.2780e-06, + 3.0100e-06, 3.0380e-06], + [-4.5374e-06, -2.0377e-06, -1.6317e-06, ..., -6.0350e-07, + -9.1456e-07, -9.5554e-07]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0427, -0.0284, 0.0403, 0.0371, -0.0876, -0.0022, 0.0062], + device='cuda:0'), grad: tensor([ 1.0977e-03, 1.4700e-05, -4.6778e-04, 2.0730e-04, -8.7881e-04, + 3.9816e-05, -1.0781e-05], device='cuda:0') +588 +0.009504844339512096 +changing lr +epoch 10, time 821.27, cls_loss 0.0561 cls_loss_mapping 0.0788 cls_loss_causal 0.6070 re_mapping 0.0611 re_causal 0.0623 /// teacc 83.92 lr 0.00940298 +Epoch 12, weight, value: tensor([[ 0.0616, 0.0580, 0.0463, ..., 0.0337, 0.0886, 0.0786], + [ 0.0605, 0.0394, 0.0097, ..., -0.0220, -0.0226, -0.0193], + [-0.0253, -0.0349, 0.0243, ..., 0.0168, -0.0476, -0.0138], + ..., + [-0.1707, -0.1254, -0.0913, ..., 0.0518, 0.0543, 0.0307], + [ 0.1450, 0.1049, 0.1035, ..., -0.0758, -0.0927, -0.0822], + [ 0.0142, 0.0328, -0.0124, ..., -0.0793, -0.0332, -0.0505]], + device='cuda:0'), grad: tensor([[-1.8275e-04, -1.1110e-04, -1.1367e-04, ..., -4.3213e-05, + -6.9678e-05, -6.4313e-05], + [ 1.5283e-04, 8.8334e-05, 9.0718e-05, ..., 3.4511e-05, + 5.3734e-05, 5.0426e-05], + [-3.7491e-05, -1.5181e-06, -3.3714e-06, ..., -3.9898e-06, + -3.8147e-06, -6.1728e-06], + ..., + [ 1.9655e-05, 6.2659e-06, 6.9514e-06, ..., 2.5034e-06, + 3.4645e-06, 3.9563e-06], + [-3.6597e-05, -1.8775e-05, -1.9938e-05, ..., -9.4250e-07, + -1.6708e-06, -2.5108e-06], + [ 5.4181e-05, 2.5660e-05, 2.7269e-05, ..., 5.6326e-06, + 9.1419e-06, 9.7007e-06]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0346, -0.0208, 0.0361, 0.0345, -0.0834, -0.0068, 0.0138], + device='cuda:0'), grad: tensor([-2.5177e-04, 2.2542e-04, -1.1235e-04, 6.1154e-05, 4.2081e-05, + -5.0873e-05, 8.6904e-05], device='cuda:0') +588 +0.009402977659283692 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 821.31, cls_loss 0.0570 cls_loss_mapping 0.0703 cls_loss_causal 0.5692 re_mapping 0.0586 re_causal 0.0600 /// teacc 90.45 lr 0.00929224 +Epoch 13, weight, value: tensor([[ 0.0644, 0.0581, 0.0467, ..., 0.0352, 0.0904, 0.0803], + [ 0.0619, 0.0420, 0.0125, ..., -0.0235, -0.0242, -0.0208], + [-0.0252, -0.0338, 0.0236, ..., 0.0161, -0.0469, -0.0139], + ..., + [-0.1647, -0.1242, -0.0901, ..., 0.0510, 0.0538, 0.0308], + [ 0.1384, 0.1016, 0.1000, ..., -0.0744, -0.0911, -0.0808], + [ 0.0093, 0.0302, -0.0140, ..., -0.0776, -0.0332, -0.0500]], + device='cuda:0'), grad: tensor([[ 8.0392e-06, -8.7544e-08, 2.9430e-07, ..., 4.3623e-06, + 5.4911e-06, 5.5321e-06], + [ 1.2279e-05, 9.0711e-07, 1.6913e-06, ..., 7.1265e-06, + 8.9929e-06, 8.9481e-06], + [-3.6621e-04, -2.8491e-05, -4.9472e-05, ..., -2.0599e-04, + -2.6464e-04, -2.6321e-04], + ..., + [ 3.5000e-04, 3.4899e-05, 5.5104e-05, ..., 1.8859e-04, + 2.4211e-04, 2.4068e-04], + [-2.7314e-05, -1.5214e-05, -1.6838e-05, ..., -3.7253e-07, + -7.0594e-07, -7.6555e-07], + [ 1.8537e-05, 7.1526e-06, 8.2403e-06, ..., 3.8929e-06, + 5.1036e-06, 5.0962e-06]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0464, -0.0207, 0.0333, 0.0331, -0.0736, -0.0147, 0.0041], + device='cuda:0'), grad: tensor([ 3.1501e-05, 4.5478e-05, -1.3533e-03, 1.9565e-05, 1.2522e-03, + -3.3081e-05, 3.8832e-05], device='cuda:0') +588 +0.009292243968009333 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 825.67, cls_loss 0.0419 cls_loss_mapping 0.0680 cls_loss_causal 0.5785 re_mapping 0.0564 re_causal 0.0583 /// teacc 90.95 lr 0.00917287 +Epoch 14, weight, value: tensor([[ 0.0640, 0.0587, 0.0481, ..., 0.0346, 0.0896, 0.0796], + [ 0.0618, 0.0423, 0.0138, ..., -0.0229, -0.0235, -0.0202], + [-0.0253, -0.0344, 0.0214, ..., 0.0166, -0.0444, -0.0121], + ..., + [-0.1670, -0.1243, -0.0911, ..., 0.0478, 0.0499, 0.0273], + [ 0.1387, 0.0999, 0.0983, ..., -0.0718, -0.0881, -0.0780], + [ 0.0112, 0.0304, -0.0132, ..., -0.0752, -0.0325, -0.0486]], + device='cuda:0'), grad: tensor([[ 2.0409e-03, 9.6655e-04, 1.0223e-03, ..., 8.1360e-05, + 2.6798e-04, 2.6321e-04], + [ 1.8731e-05, 3.2540e-06, 3.3826e-06, ..., 1.0967e-05, + 1.3873e-05, 1.2882e-05], + [-3.7594e-03, -1.2169e-03, -1.4153e-03, ..., -4.4370e-04, + -8.7452e-04, -8.0872e-04], + ..., + [ 1.4982e-03, 2.1255e-04, 3.4595e-04, ..., 2.7776e-04, + 4.8876e-04, 4.3702e-04], + [ 9.2164e-06, 2.5686e-06, 2.9355e-06, ..., 1.6699e-06, + 2.7679e-06, 2.5965e-06], + [ 5.2750e-05, 9.8124e-06, 1.0490e-05, ..., 2.8387e-05, + 3.6836e-05, 3.4153e-05]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0433, -0.0196, 0.0364, 0.0285, -0.0805, -0.0071, 0.0064], + device='cuda:0'), grad: tensor([ 3.7994e-03, 5.9754e-05, -8.5068e-03, 3.9625e-04, 4.0627e-03, + 2.2739e-05, 1.6463e-04], device='cuda:0') +588 +0.009172866268606516 +changing lr +epoch 13, time 817.11, cls_loss 0.0311 cls_loss_mapping 0.0553 cls_loss_causal 0.5720 re_mapping 0.0535 re_causal 0.0557 /// teacc 90.70 lr 0.00904508 +Epoch 15, weight, value: tensor([[ 0.0683, 0.0621, 0.0518, ..., 0.0324, 0.0866, 0.0768], + [ 0.0620, 0.0419, 0.0141, ..., -0.0221, -0.0225, -0.0192], + [-0.0247, -0.0339, 0.0208, ..., 0.0148, -0.0450, -0.0135], + ..., + [-0.1660, -0.1239, -0.0918, ..., 0.0477, 0.0502, 0.0282], + [ 0.1343, 0.0975, 0.0958, ..., -0.0702, -0.0863, -0.0765], + [ 0.0075, 0.0282, -0.0144, ..., -0.0735, -0.0320, -0.0477]], + device='cuda:0'), grad: tensor([[ 3.0651e-03, 5.9986e-04, 5.0879e-04, ..., 5.8985e-04, + 8.3113e-04, 7.6103e-04], + [ 4.5419e-04, 8.9407e-05, 8.0764e-05, ..., 8.5473e-05, + 1.2624e-04, 1.1367e-04], + [-8.7509e-03, -1.7099e-03, -1.4582e-03, ..., -1.6956e-03, + -2.3994e-03, -2.1935e-03], + ..., + [ 2.7809e-03, 5.3740e-04, 4.3988e-04, ..., 5.4169e-04, + 7.4339e-04, 6.8808e-04], + [ 1.7011e-04, 3.3975e-05, 2.9579e-05, ..., 3.2157e-05, + 4.6343e-05, 4.2140e-05], + [ 4.3726e-04, 8.6069e-05, 7.3731e-05, ..., 8.3447e-05, + 1.1867e-04, 1.0842e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 4.2851e-02, -1.4903e-02, 3.4590e-02, 3.1522e-02, -7.6456e-02, + -1.0301e-02, 1.8977e-05], device='cuda:0'), grad: tensor([ 0.0080, 0.0012, -0.0229, 0.0049, 0.0072, 0.0004, 0.0011], + device='cuda:0') +588 +0.00904508497187474 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 826.21, cls_loss 0.0221 cls_loss_mapping 0.0496 cls_loss_causal 0.5465 re_mapping 0.0512 re_causal 0.0539 /// teacc 91.71 lr 0.00890916 +Epoch 16, weight, value: tensor([[ 0.0712, 0.0653, 0.0555, ..., 0.0318, 0.0849, 0.0754], + [ 0.0595, 0.0405, 0.0135, ..., -0.0222, -0.0225, -0.0193], + [-0.0267, -0.0348, 0.0188, ..., 0.0135, -0.0448, -0.0141], + ..., + [-0.1610, -0.1225, -0.0914, ..., 0.0478, 0.0508, 0.0292], + [ 0.1298, 0.0948, 0.0930, ..., -0.0687, -0.0847, -0.0750], + [ 0.0085, 0.0278, -0.0139, ..., -0.0716, -0.0310, -0.0463]], + device='cuda:0'), grad: tensor([[ 1.7147e-03, 3.0088e-04, 3.9959e-04, ..., 4.1032e-04, + 6.5947e-04, 6.2037e-04], + [ 1.4715e-05, 4.4033e-06, 4.8503e-06, ..., 2.9411e-06, + 4.1239e-06, 4.0755e-06], + [-2.1495e-06, 1.5786e-06, 2.2314e-06, ..., -4.2655e-06, + -2.8461e-06, -3.6471e-06], + ..., + [-1.7195e-03, -2.9278e-04, -3.9339e-04, ..., -4.1509e-04, + -6.6853e-04, -6.2752e-04], + [ 2.4036e-05, 1.1437e-05, 1.1727e-05, ..., 2.2370e-06, + 3.2187e-06, 3.4068e-06], + [-5.2631e-05, -2.8491e-05, -2.8670e-05, ..., -3.0771e-06, + -4.1500e-06, -4.8578e-06]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0440, -0.0177, 0.0296, 0.0272, -0.0665, -0.0132, 0.0038], + device='cuda:0'), grad: tensor([ 4.1809e-03, 3.2425e-05, -1.2070e-05, 5.0902e-05, -4.2076e-03, + 4.3541e-05, -8.7917e-05], device='cuda:0') +588 +0.008909157412340152 +changing lr +epoch 15, time 813.34, cls_loss 0.0266 cls_loss_mapping 0.0508 cls_loss_causal 0.5462 re_mapping 0.0484 re_causal 0.0513 /// teacc 91.71 lr 0.00876536 +Epoch 17, weight, value: tensor([[ 0.0715, 0.0649, 0.0553, ..., 0.0318, 0.0838, 0.0750], + [ 0.0587, 0.0413, 0.0148, ..., -0.0222, -0.0224, -0.0192], + [-0.0274, -0.0342, 0.0180, ..., 0.0116, -0.0450, -0.0154], + ..., + [-0.1618, -0.1227, -0.0923, ..., 0.0472, 0.0501, 0.0288], + [ 0.1293, 0.0936, 0.0919, ..., -0.0669, -0.0824, -0.0730], + [ 0.0088, 0.0271, -0.0136, ..., -0.0700, -0.0305, -0.0454]], + device='cuda:0'), grad: tensor([[ 8.0541e-06, 3.9004e-06, 2.7288e-06, ..., 2.1141e-06, + 1.3644e-06, 1.4128e-06], + [-5.5343e-05, -2.2486e-05, -2.2516e-05, ..., -4.5486e-06, + -6.7763e-06, -7.1824e-06], + [ 1.2174e-05, 2.0973e-06, 3.1088e-06, ..., 2.5313e-06, + 3.2149e-06, 3.2037e-06], + ..., + [ 4.1649e-06, 2.9355e-06, 3.1441e-06, ..., -4.7032e-07, + -3.0920e-07, -2.5053e-07], + [ 5.8673e-06, 1.9930e-06, 2.1309e-06, ..., 5.8115e-07, + 9.0804e-07, 9.4622e-07], + [ 2.0936e-05, 7.8082e-06, 8.0466e-06, ..., 1.8356e-06, + 2.7772e-06, 2.9653e-06]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0460, -0.0192, 0.0244, 0.0282, -0.0692, -0.0082, 0.0049], + device='cuda:0'), grad: tensor([ 1.7703e-05, -1.0329e-04, 3.1352e-05, 2.3097e-06, -1.1222e-06, + 1.2226e-05, 4.0591e-05], device='cuda:0') +588 +0.00876535733001806 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 816.99, cls_loss 0.0211 cls_loss_mapping 0.0483 cls_loss_causal 0.5618 re_mapping 0.0473 re_causal 0.0506 /// teacc 91.96 lr 0.00861397 +Epoch 18, weight, value: tensor([[ 0.0775, 0.0697, 0.0607, ..., 0.0303, 0.0809, 0.0726], + [ 0.0547, 0.0376, 0.0118, ..., -0.0215, -0.0212, -0.0183], + [-0.0281, -0.0353, 0.0158, ..., 0.0125, -0.0429, -0.0142], + ..., + [-0.1592, -0.1212, -0.0916, ..., 0.0467, 0.0500, 0.0292], + [ 0.1258, 0.0916, 0.0898, ..., -0.0655, -0.0808, -0.0716], + [ 0.0085, 0.0268, -0.0133, ..., -0.0684, -0.0301, -0.0447]], + device='cuda:0'), grad: tensor([[ 1.4365e-04, 5.3078e-05, 5.5641e-05, ..., 8.5086e-06, + 1.4983e-05, 1.8373e-05], + [ 4.0591e-05, 1.3851e-05, 1.5192e-05, ..., 4.5113e-06, + 6.4410e-06, 7.2829e-06], + [ 1.2779e-04, 3.4571e-05, 3.9786e-05, ..., 2.5526e-05, + 2.9683e-05, 3.1978e-05], + ..., + [ 1.4191e-03, 5.2643e-04, 5.5504e-04, ..., 8.7798e-05, + 1.5748e-04, 1.9002e-04], + [-5.7945e-03, -2.1477e-03, -2.2564e-03, ..., -3.5357e-04, + -6.3562e-04, -7.6866e-04], + [ 3.9864e-03, 1.4734e-03, 1.5469e-03, ..., 2.4414e-04, + 4.3726e-04, 5.2881e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0497, -0.0176, 0.0255, 0.0224, -0.0677, -0.0103, 0.0047], + device='cuda:0'), grad: tensor([ 2.6631e-04, 7.9513e-05, 2.7847e-04, 1.0753e-04, 2.6379e-03, + -1.0796e-02, 7.4348e-03], device='cuda:0') +588 +0.008613974319136962 +changing lr +epoch 17, time 818.76, cls_loss 0.0146 cls_loss_mapping 0.0373 cls_loss_causal 0.5409 re_mapping 0.0453 re_causal 0.0490 /// teacc 91.71 lr 0.00845531 +Epoch 19, weight, value: tensor([[ 0.0752, 0.0691, 0.0600, ..., 0.0290, 0.0787, 0.0707], + [ 0.0547, 0.0370, 0.0118, ..., -0.0211, -0.0202, -0.0175], + [-0.0266, -0.0350, 0.0150, ..., 0.0123, -0.0417, -0.0137], + ..., + [-0.1582, -0.1196, -0.0907, ..., 0.0456, 0.0486, 0.0283], + [ 0.1242, 0.0905, 0.0887, ..., -0.0641, -0.0790, -0.0700], + [ 0.0071, 0.0258, -0.0134, ..., -0.0670, -0.0298, -0.0441]], + device='cuda:0'), grad: tensor([[-1.0096e-06, -1.2126e-06, -1.1018e-06, ..., 3.4366e-07, + 2.8312e-07, 1.9837e-07], + [ 7.3351e-06, 5.9605e-07, 1.2675e-06, ..., 3.3565e-06, + 4.1500e-06, 4.0606e-06], + [-5.8636e-06, 4.5262e-07, -1.9465e-07, ..., -4.3362e-06, + -5.1446e-06, -5.1446e-06], + ..., + [ 3.3751e-06, 1.1064e-06, 1.2377e-06, ..., 1.1427e-06, + 1.4827e-06, 1.4231e-06], + [ 3.0696e-06, 1.0030e-06, 1.0589e-06, ..., 3.2783e-07, + 5.2527e-07, 5.0385e-07], + [-3.3528e-06, -1.4165e-06, -1.3765e-06, ..., 1.2759e-07, + -2.5146e-08, 1.8626e-09]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0439, -0.0119, 0.0294, 0.0249, -0.0726, -0.0098, 0.0027], + device='cuda:0'), grad: tensor([ 5.2713e-07, 1.9491e-05, -1.6972e-05, -8.5086e-06, 4.9025e-06, + 5.8562e-06, -5.3048e-06], device='cuda:0') +588 +0.008455313244934327 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 814.80, cls_loss 0.0140 cls_loss_mapping 0.0346 cls_loss_causal 0.5277 re_mapping 0.0424 re_causal 0.0464 /// teacc 92.46 lr 0.00828969 +Epoch 20, weight, value: tensor([[ 0.0781, 0.0697, 0.0611, ..., 0.0294, 0.0788, 0.0709], + [ 0.0527, 0.0364, 0.0118, ..., -0.0212, -0.0206, -0.0179], + [-0.0272, -0.0350, 0.0138, ..., 0.0120, -0.0409, -0.0136], + ..., + [-0.1548, -0.1181, -0.0897, ..., 0.0448, 0.0481, 0.0283], + [ 0.1220, 0.0894, 0.0875, ..., -0.0628, -0.0776, -0.0687], + [ 0.0063, 0.0250, -0.0135, ..., -0.0657, -0.0294, -0.0434]], + device='cuda:0'), grad: tensor([[ 4.7296e-05, 1.2435e-05, 1.3180e-05, ..., 8.9556e-06, + 1.2085e-05, 1.2808e-05], + [ 2.1249e-05, 3.6787e-06, 5.5023e-06, ..., 7.8008e-06, + 8.6352e-06, 9.1121e-06], + [-9.3341e-05, -2.8506e-05, -2.8655e-05, ..., -1.3091e-05, + -1.9684e-05, -2.0817e-05], + ..., + [ 2.8729e-05, 9.0674e-06, 9.4250e-06, ..., 3.9823e-06, + 4.9956e-06, 5.6550e-06], + [ 8.0541e-06, 2.3283e-06, 2.3935e-06, ..., 1.2824e-06, + 1.7788e-06, 1.9064e-06], + [ 5.2862e-06, 1.3188e-06, 1.4612e-06, ..., 1.1791e-06, + 1.5628e-06, 1.6280e-06]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0500, -0.0137, 0.0276, 0.0195, -0.0668, -0.0109, 0.0007], + device='cuda:0'), grad: tensor([ 1.1319e-04, 5.6565e-05, -2.1183e-04, -5.3734e-05, 6.3956e-05, + 1.8641e-05, 1.2890e-05], device='cuda:0') +588 +0.008289693629698565 +changing lr +epoch 19, time 817.70, cls_loss 0.0107 cls_loss_mapping 0.0303 cls_loss_causal 0.4927 re_mapping 0.0397 re_causal 0.0443 /// teacc 91.96 lr 0.00811745 +Epoch 21, weight, value: tensor([[ 0.0791, 0.0704, 0.0623, ..., 0.0290, 0.0777, 0.0699], + [ 0.0500, 0.0352, 0.0110, ..., -0.0210, -0.0206, -0.0179], + [-0.0248, -0.0341, 0.0136, ..., 0.0118, -0.0396, -0.0130], + ..., + [-0.1541, -0.1172, -0.0894, ..., 0.0437, 0.0465, 0.0273], + [ 0.1196, 0.0874, 0.0854, ..., -0.0615, -0.0760, -0.0673], + [ 0.0064, 0.0247, -0.0130, ..., -0.0642, -0.0286, -0.0423]], + device='cuda:0'), grad: tensor([[-3.9160e-05, -1.7390e-05, -2.0787e-05, ..., -1.6809e-05, + -2.4855e-05, -2.4080e-05], + [ 3.4142e-04, 8.4937e-05, 9.1851e-05, ..., 4.8280e-05, + 8.7559e-05, 8.2612e-05], + [-1.4208e-05, 3.8259e-06, 3.5446e-06, ..., -7.3127e-06, + -8.6874e-06, -8.7619e-06], + ..., + [ 1.3396e-05, 3.8035e-06, 4.4145e-06, ..., 3.3602e-06, + 5.0776e-06, 4.9211e-06], + [-2.9411e-06, -3.5223e-06, -4.0568e-06, ..., 6.3144e-07, + 1.1791e-06, 1.0673e-06], + [-3.4285e-04, -7.9155e-05, -8.3864e-05, ..., -4.0293e-05, + -7.7784e-05, -7.2777e-05]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0515, -0.0168, 0.0326, 0.0178, -0.0700, -0.0094, 0.0005], + device='cuda:0'), grad: tensor([-7.2360e-05, 8.0204e-04, -5.6326e-05, 1.1629e-04, 3.0488e-05, + 1.3709e-06, -8.2064e-04], device='cuda:0') +588 +0.00811744900929367 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 819.64, cls_loss 0.0094 cls_loss_mapping 0.0289 cls_loss_causal 0.4921 re_mapping 0.0375 re_causal 0.0426 /// teacc 93.97 lr 0.00793893 +Epoch 22, weight, value: tensor([[ 0.0755, 0.0697, 0.0617, ..., 0.0272, 0.0746, 0.0672], + [ 0.0486, 0.0338, 0.0103, ..., -0.0204, -0.0197, -0.0172], + [-0.0257, -0.0337, 0.0127, ..., 0.0108, -0.0398, -0.0138], + ..., + [-0.1503, -0.1157, -0.0884, ..., 0.0436, 0.0470, 0.0282], + [ 0.1188, 0.0868, 0.0848, ..., -0.0600, -0.0743, -0.0659], + [ 0.0069, 0.0244, -0.0126, ..., -0.0629, -0.0281, -0.0414]], + device='cuda:0'), grad: tensor([[-1.3173e-04, -6.0469e-05, -6.1989e-05, ..., -9.6112e-06, + -2.5392e-05, -2.3678e-05], + [-4.2051e-05, -3.4384e-06, -4.9137e-06, ..., -1.7628e-05, + -2.1622e-05, -2.1964e-05], + [ 2.7984e-05, 1.1958e-05, 1.2219e-05, ..., 3.5465e-06, + 6.8322e-06, 6.5304e-06], + ..., + [ 9.0480e-05, 3.0965e-05, 3.3319e-05, ..., 1.5631e-05, + 2.5705e-05, 2.5153e-05], + [ 1.5885e-05, 5.9344e-06, 6.1505e-06, ..., 2.3320e-06, + 4.1239e-06, 3.9600e-06], + [ 3.3468e-05, 1.1489e-05, 1.2018e-05, ..., 5.7481e-06, + 9.4101e-06, 9.1642e-06]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0416, -0.0155, 0.0282, 0.0205, -0.0621, -0.0081, 0.0015], + device='cuda:0'), grad: tensor([-2.2805e-04, -1.2910e-04, 5.2780e-05, 7.2978e-06, 1.9133e-04, + 3.2693e-05, 7.3075e-05], device='cuda:0') +588 +0.007938926261462368 +changing lr +epoch 21, time 817.08, cls_loss 0.0062 cls_loss_mapping 0.0266 cls_loss_causal 0.4959 re_mapping 0.0337 re_causal 0.0396 /// teacc 92.46 lr 0.00775448 +Epoch 23, weight, value: tensor([[ 0.0774, 0.0699, 0.0624, ..., 0.0271, 0.0741, 0.0668], + [ 0.0476, 0.0338, 0.0106, ..., -0.0205, -0.0197, -0.0172], + [-0.0258, -0.0335, 0.0118, ..., 0.0105, -0.0391, -0.0137], + ..., + [-0.1503, -0.1148, -0.0882, ..., 0.0425, 0.0455, 0.0272], + [ 0.1166, 0.0855, 0.0835, ..., -0.0589, -0.0731, -0.0647], + [ 0.0065, 0.0236, -0.0126, ..., -0.0617, -0.0276, -0.0406]], + device='cuda:0'), grad: tensor([[-2.8193e-05, -1.7360e-05, -2.0295e-05, ..., 4.8876e-06, + 2.8685e-07, 0.0000e+00], + [ 3.0443e-05, 5.6103e-06, 8.6799e-06, ..., 1.1146e-05, + 1.4178e-05, 1.3806e-05], + [ 5.9158e-06, 1.6000e-06, 2.3283e-06, ..., 3.2876e-06, + 3.8818e-06, 3.7644e-06], + ..., + [ 2.7075e-05, 8.2180e-06, 1.0900e-05, ..., 4.9397e-06, + 8.1584e-06, 8.0317e-06], + [ 1.0602e-05, 2.5760e-06, 3.6433e-06, ..., 3.0547e-06, + 4.2059e-06, 4.1127e-06], + [ 1.5199e-05, 4.4219e-06, 5.9418e-06, ..., 3.2336e-06, + 5.0031e-06, 4.9137e-06]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0455, -0.0160, 0.0273, 0.0216, -0.0653, -0.0090, 0.0019], + device='cuda:0'), grad: tensor([-3.4958e-05, 7.6354e-05, 1.3001e-05, -1.7130e-04, 5.8532e-05, + 2.4796e-05, 3.3438e-05], device='cuda:0') +588 +0.007754484907260515 +changing lr +epoch 22, time 813.28, cls_loss 0.0129 cls_loss_mapping 0.0309 cls_loss_causal 0.4926 re_mapping 0.0328 re_causal 0.0391 /// teacc 92.21 lr 0.00756450 +Epoch 24, weight, value: tensor([[ 0.0758, 0.0697, 0.0622, ..., 0.0260, 0.0721, 0.0649], + [ 0.0471, 0.0335, 0.0109, ..., -0.0200, -0.0191, -0.0167], + [-0.0247, -0.0333, 0.0110, ..., 0.0097, -0.0386, -0.0138], + ..., + [-0.1486, -0.1136, -0.0874, ..., 0.0423, 0.0453, 0.0273], + [ 0.1150, 0.0843, 0.0823, ..., -0.0578, -0.0717, -0.0635], + [ 0.0080, 0.0234, -0.0121, ..., -0.0597, -0.0263, -0.0389]], + device='cuda:0'), grad: tensor([[-2.2335e-03, -7.6771e-04, -8.5497e-04, ..., -2.6250e-04, + -6.4802e-04, -6.1846e-04], + [ 5.7936e-04, 2.0504e-04, 2.0194e-04, ..., 7.2360e-05, + 1.7726e-04, 1.7560e-04], + [ 1.0309e-03, 3.1829e-04, 3.8576e-04, ..., 1.1569e-04, + 2.8825e-04, 2.7323e-04], + ..., + [ 5.9754e-05, 5.4240e-05, 6.3658e-05, ..., 6.6273e-06, + 1.5646e-05, 7.1041e-06], + [ 7.4506e-05, 2.3946e-05, 2.7284e-05, ..., 8.6576e-06, + 2.1428e-05, 2.0623e-05], + [ 3.7026e-04, 1.2684e-04, 1.2946e-04, ..., 4.5449e-05, + 1.1164e-04, 1.1009e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0413, -0.0152, 0.0281, 0.0178, -0.0638, -0.0087, 0.0063], + device='cuda:0'), grad: tensor([-4.5090e-03, 1.1473e-03, 2.1763e-03, 2.4605e-04, 3.8803e-05, + 1.5461e-04, 7.4577e-04], device='cuda:0') +588 +0.007564496387029534 +changing lr +epoch 23, time 838.63, cls_loss 0.0068 cls_loss_mapping 0.0268 cls_loss_causal 0.4974 re_mapping 0.0309 re_causal 0.0379 /// teacc 93.47 lr 0.00736934 +Epoch 25, weight, value: tensor([[ 0.0772, 0.0701, 0.0627, ..., 0.0253, 0.0711, 0.0641], + [ 0.0474, 0.0329, 0.0109, ..., -0.0192, -0.0180, -0.0156], + [-0.0240, -0.0325, 0.0112, ..., 0.0092, -0.0379, -0.0136], + ..., + [-0.1461, -0.1120, -0.0863, ..., 0.0420, 0.0448, 0.0272], + [ 0.1120, 0.0828, 0.0807, ..., -0.0569, -0.0707, -0.0627], + [ 0.0052, 0.0222, -0.0129, ..., -0.0591, -0.0268, -0.0390]], + device='cuda:0'), grad: tensor([[-5.1880e-04, -1.9383e-04, -1.9336e-04, ..., -4.0913e-04, + -5.2977e-04, -5.5504e-04], + [ 2.3937e-04, 4.2588e-05, 5.4538e-05, ..., 1.3030e-04, + 1.8048e-04, 1.8859e-04], + [ 6.7043e-04, 5.6118e-05, 1.1438e-04, ..., 1.1480e-04, + 2.1446e-04, 2.1780e-04], + ..., + [-7.0429e-04, 3.7283e-05, -5.2094e-05, ..., 6.7055e-05, + -1.4283e-05, -4.5113e-06], + [ 8.7440e-05, 1.3962e-05, 1.9893e-05, ..., 2.1994e-05, + 3.5554e-05, 3.6418e-05], + [ 1.3733e-04, 2.6837e-05, 3.4779e-05, ..., 4.2856e-05, + 6.5029e-05, 6.6936e-05]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0444, -0.0117, 0.0283, 0.0172, -0.0612, -0.0112, -0.0002], + device='cuda:0'), grad: tensor([-0.0020, 0.0009, 0.0019, 0.0003, -0.0017, 0.0002, 0.0004], + device='cuda:0') +588 +0.007369343312364995 +changing lr +epoch 24, time 812.60, cls_loss 0.0083 cls_loss_mapping 0.0233 cls_loss_causal 0.4944 re_mapping 0.0281 re_causal 0.0354 /// teacc 92.46 lr 0.00716942 +Epoch 26, weight, value: tensor([[ 0.0793, 0.0713, 0.0640, ..., 0.0250, 0.0705, 0.0636], + [ 0.0455, 0.0320, 0.0104, ..., -0.0190, -0.0178, -0.0154], + [-0.0262, -0.0326, 0.0101, ..., 0.0075, -0.0389, -0.0151], + ..., + [-0.1435, -0.1111, -0.0858, ..., 0.0423, 0.0453, 0.0279], + [ 0.1096, 0.0815, 0.0795, ..., -0.0560, -0.0697, -0.0618], + [ 0.0047, 0.0217, -0.0128, ..., -0.0581, -0.0266, -0.0386]], + device='cuda:0'), grad: tensor([[-1.9336e-04, -8.9705e-05, -1.1009e-04, ..., -4.8935e-05, + -6.3360e-05, -5.6565e-05], + [ 1.6642e-04, 6.4254e-05, 7.9572e-05, ..., 4.3362e-05, + 5.6803e-05, 5.0813e-05], + [-5.6922e-06, -6.2399e-08, -2.3283e-07, ..., -2.2445e-07, + -1.1520e-06, -1.0673e-06], + ..., + [ 2.4468e-05, 1.0535e-05, 1.3068e-05, ..., 6.4000e-06, + 8.0764e-06, 7.2680e-06], + [ 2.3529e-05, 9.8720e-06, 1.1981e-05, ..., 5.3756e-06, + 7.2233e-06, 6.5342e-06], + [-4.8280e-06, -1.2685e-06, -8.4285e-07, ..., 8.8289e-07, + 5.6252e-07, 1.9465e-07]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0483, -0.0136, 0.0214, 0.0208, -0.0566, -0.0131, -0.0017], + device='cuda:0'), grad: tensor([-5.5456e-04, 4.6062e-04, -7.7561e-06, -2.8417e-05, 6.7294e-05, + 6.2883e-05, -7.5717e-07], device='cuda:0') +588 +0.0071694186955877925 +changing lr +epoch 25, time 815.09, cls_loss 0.0045 cls_loss_mapping 0.0209 cls_loss_causal 0.4697 re_mapping 0.0255 re_causal 0.0334 /// teacc 90.95 lr 0.00696513 +Epoch 27, weight, value: tensor([[ 0.0799, 0.0712, 0.0642, ..., 0.0250, 0.0700, 0.0632], + [ 0.0423, 0.0301, 0.0090, ..., -0.0189, -0.0177, -0.0154], + [-0.0227, -0.0313, 0.0106, ..., 0.0076, -0.0376, -0.0143], + ..., + [-0.1438, -0.1101, -0.0854, ..., 0.0410, 0.0437, 0.0268], + [ 0.1093, 0.0812, 0.0792, ..., -0.0550, -0.0684, -0.0607], + [ 0.0045, 0.0212, -0.0128, ..., -0.0570, -0.0263, -0.0380]], + device='cuda:0'), grad: tensor([[ 7.5102e-04, 2.4223e-04, 2.5082e-04, ..., 8.1003e-05, + 1.7059e-04, 1.6427e-04], + [-2.1534e-03, -4.9210e-04, -5.3501e-04, ..., -4.0579e-04, + -7.5388e-04, -7.0095e-04], + [ 6.5565e-04, 1.3733e-04, 1.5342e-04, ..., 1.4067e-04, + 2.4140e-04, 2.2197e-04], + ..., + [ 4.6134e-04, 1.2887e-04, 1.3399e-04, ..., 6.0886e-05, + 1.3816e-04, 1.3185e-04], + [ 1.4603e-04, 4.7207e-05, 4.9174e-05, ..., 1.7181e-05, + 3.4273e-05, 3.2783e-05], + [-2.2185e-04, -1.3363e-04, -1.3316e-04, ..., 1.8224e-05, + 2.9132e-05, 2.1964e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0511, -0.0171, 0.0282, 0.0184, -0.0614, -0.0114, -0.0023], + device='cuda:0'), grad: tensor([ 0.0015, -0.0049, 0.0015, 0.0008, 0.0010, 0.0003, -0.0003], + device='cuda:0') +588 +0.0069651251582696205 +changing lr +epoch 26, time 811.56, cls_loss 0.0048 cls_loss_mapping 0.0252 cls_loss_causal 0.4728 re_mapping 0.0231 re_causal 0.0307 /// teacc 92.21 lr 0.00675687 +Epoch 28, weight, value: tensor([[ 0.0781, 0.0708, 0.0640, ..., 0.0244, 0.0687, 0.0620], + [ 0.0419, 0.0297, 0.0090, ..., -0.0184, -0.0171, -0.0149], + [-0.0219, -0.0308, 0.0105, ..., 0.0071, -0.0371, -0.0143], + ..., + [-0.1406, -0.1084, -0.0842, ..., 0.0407, 0.0434, 0.0268], + [ 0.1079, 0.0803, 0.0783, ..., -0.0541, -0.0674, -0.0597], + [ 0.0038, 0.0204, -0.0132, ..., -0.0560, -0.0258, -0.0374]], + device='cuda:0'), grad: tensor([[ 2.9993e-04, 5.0426e-05, 7.1168e-05, ..., 6.6698e-05, + 1.1802e-04, 1.0926e-04], + [ 6.8903e-05, 8.4564e-06, 1.6898e-05, ..., 2.1636e-05, + 3.6865e-05, 3.4124e-05], + [-2.8992e-04, -4.8310e-05, -7.9393e-05, ..., -7.9989e-05, + -1.5891e-04, -1.5199e-04], + ..., + [-1.2755e-04, -2.4274e-05, -2.2009e-05, ..., -1.2144e-05, + -2.3648e-05, -2.0996e-05], + [ 3.1620e-05, 5.1260e-06, 8.0764e-06, ..., 8.0988e-06, + 1.3381e-05, 1.2286e-05], + [ 5.0277e-05, 7.9572e-06, 1.3076e-05, ..., 1.3307e-05, + 2.3946e-05, 2.2426e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0468, -0.0160, 0.0291, 0.0164, -0.0566, -0.0112, -0.0030], + device='cuda:0'), grad: tensor([ 8.6927e-04, 2.4223e-04, -1.1597e-03, 3.6865e-05, -2.4700e-04, + 9.1076e-05, 1.6832e-04], device='cuda:0') +588 +0.006756874120406716 +changing lr +epoch 27, time 812.10, cls_loss 0.0076 cls_loss_mapping 0.0221 cls_loss_causal 0.4726 re_mapping 0.0227 re_causal 0.0314 /// teacc 93.22 lr 0.00654508 +Epoch 29, weight, value: tensor([[ 0.0767, 0.0703, 0.0636, ..., 0.0237, 0.0673, 0.0607], + [ 0.0415, 0.0293, 0.0090, ..., -0.0181, -0.0167, -0.0145], + [-0.0233, -0.0309, 0.0096, ..., 0.0065, -0.0370, -0.0145], + ..., + [-0.1376, -0.1069, -0.0831, ..., 0.0406, 0.0437, 0.0272], + [ 0.1074, 0.0792, 0.0773, ..., -0.0532, -0.0663, -0.0588], + [ 0.0036, 0.0202, -0.0129, ..., -0.0553, -0.0257, -0.0370]], + device='cuda:0'), grad: tensor([[ 4.3094e-05, 5.2899e-06, 8.3968e-06, ..., 3.5278e-06, + 6.4224e-06, 6.6943e-06], + [ 1.5283e-04, 2.0862e-05, 3.1590e-05, ..., 1.8626e-05, + 2.8476e-05, 2.9594e-05], + [ 3.7730e-05, 5.0962e-06, 7.6368e-06, ..., 5.5395e-06, + 8.8438e-06, 8.7544e-06], + ..., + [-1.3523e-06, 7.4506e-08, 1.4901e-07, ..., -4.0233e-06, + -7.0371e-06, -5.8599e-06], + [ 4.2766e-05, 6.2548e-06, 9.1791e-06, ..., 2.9728e-06, + 5.7705e-06, 6.0350e-06], + [ 9.9599e-05, 1.4491e-05, 2.1368e-05, ..., 5.8301e-06, + 1.1764e-05, 1.2614e-05]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0437, -0.0154, 0.0251, 0.0160, -0.0519, -0.0080, -0.0043], + device='cuda:0'), grad: tensor([ 1.0687e-04, 3.7146e-04, 9.4533e-05, -9.0218e-04, -1.3970e-05, + 1.0341e-04, 2.3937e-04], device='cuda:0') +588 +0.00654508497187474 +changing lr +epoch 28, time 813.77, cls_loss 0.0030 cls_loss_mapping 0.0186 cls_loss_causal 0.4803 re_mapping 0.0193 re_causal 0.0285 /// teacc 93.72 lr 0.00633018 +Epoch 30, weight, value: tensor([[ 0.0781, 0.0712, 0.0646, ..., 0.0232, 0.0664, 0.0599], + [ 0.0420, 0.0290, 0.0091, ..., -0.0173, -0.0157, -0.0134], + [-0.0223, -0.0306, 0.0091, ..., 0.0059, -0.0366, -0.0144], + ..., + [-0.1383, -0.1062, -0.0829, ..., 0.0398, 0.0426, 0.0264], + [ 0.1040, 0.0778, 0.0758, ..., -0.0526, -0.0658, -0.0584], + [ 0.0040, 0.0198, -0.0128, ..., -0.0544, -0.0254, -0.0364]], + device='cuda:0'), grad: tensor([[-4.5508e-05, -2.1264e-05, -2.1949e-05, ..., -2.3674e-06, + -5.2936e-06, -4.8615e-06], + [-2.9489e-05, -6.8396e-06, -6.5006e-06, ..., -4.3102e-06, + -1.1213e-05, -1.0550e-05], + [-2.6766e-06, -2.5257e-06, -2.3134e-06, ..., 2.0117e-07, + 1.0058e-07, -2.6077e-08], + ..., + [ 3.0696e-05, 1.1295e-05, 1.2062e-05, ..., 3.9376e-06, + 6.7912e-06, 6.2697e-06], + [ 1.0431e-05, 3.5726e-06, 3.8259e-06, ..., 1.7378e-06, + 2.6803e-06, 2.4978e-06], + [ 4.9114e-05, 1.3009e-05, 1.3947e-05, ..., 9.2536e-06, + 1.6272e-05, 1.5147e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0453, -0.0113, 0.0267, 0.0166, -0.0566, -0.0127, -0.0029], + device='cuda:0'), grad: tensor([-7.7009e-05, -7.4744e-05, -4.2170e-06, -4.6015e-05, 6.2525e-05, + 2.2516e-05, 1.1718e-04], device='cuda:0') +588 +0.006330184227833378 +changing lr +epoch 29, time 812.07, cls_loss 0.0059 cls_loss_mapping 0.0181 cls_loss_causal 0.4874 re_mapping 0.0183 re_causal 0.0279 /// teacc 93.97 lr 0.00611260 +Epoch 31, weight, value: tensor([[ 0.0769, 0.0709, 0.0645, ..., 0.0224, 0.0651, 0.0587], + [ 0.0414, 0.0288, 0.0094, ..., -0.0171, -0.0156, -0.0134], + [-0.0222, -0.0306, 0.0084, ..., 0.0058, -0.0357, -0.0140], + ..., + [-0.1354, -0.1048, -0.0819, ..., 0.0395, 0.0425, 0.0266], + [ 0.1018, 0.0766, 0.0746, ..., -0.0519, -0.0650, -0.0577], + [ 0.0035, 0.0195, -0.0127, ..., -0.0537, -0.0253, -0.0361]], + device='cuda:0'), grad: tensor([[ 3.9139e-03, 5.4932e-04, 3.8695e-04, ..., 3.5524e-04, + 7.6008e-04, 8.3113e-04], + [-7.1466e-05, -4.0025e-05, -4.5657e-05, ..., -2.9542e-06, + -7.9945e-06, -5.8375e-06], + [ 1.5774e-03, 2.2268e-04, 1.6809e-04, ..., 1.9121e-04, + 3.5977e-04, 3.7146e-04], + ..., + [-5.7487e-03, -7.9298e-04, -5.6410e-04, ..., -5.7840e-04, + -1.1806e-03, -1.2665e-03], + [ 4.4435e-05, 7.5325e-06, 6.5975e-06, ..., 4.9621e-06, + 9.7528e-06, 9.9167e-06], + [ 1.0777e-04, 2.2024e-05, 2.0474e-05, ..., 1.0796e-05, + 2.1920e-05, 2.2098e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0428, -0.0124, 0.0279, 0.0186, -0.0526, -0.0146, -0.0046], + device='cuda:0'), grad: tensor([ 1.0109e-02, -8.3566e-05, 4.1885e-03, 4.4632e-04, -1.5038e-02, + 1.1331e-04, 2.6035e-04], device='cuda:0') +588 +0.006112604669781575 +changing lr +epoch 30, time 810.53, cls_loss 0.0028 cls_loss_mapping 0.0157 cls_loss_causal 0.4478 re_mapping 0.0176 re_causal 0.0272 /// teacc 89.95 lr 0.00589278 +Epoch 32, weight, value: tensor([[ 0.0778, 0.0712, 0.0649, ..., 0.0224, 0.0645, 0.0582], + [ 0.0394, 0.0280, 0.0090, ..., -0.0171, -0.0157, -0.0135], + [-0.0223, -0.0304, 0.0079, ..., 0.0056, -0.0352, -0.0139], + ..., + [-0.1338, -0.1039, -0.0814, ..., 0.0390, 0.0420, 0.0264], + [ 0.1008, 0.0758, 0.0739, ..., -0.0511, -0.0640, -0.0568], + [ 0.0047, 0.0196, -0.0122, ..., -0.0528, -0.0248, -0.0354]], + device='cuda:0'), grad: tensor([[ 1.5440e-03, 2.1601e-04, 3.2449e-04, ..., 5.2166e-04, + 7.2908e-04, 7.3481e-04], + [ 8.5890e-05, 3.0577e-05, 2.9802e-05, ..., 9.7379e-06, + 1.2994e-05, 1.3396e-05], + [ 1.4953e-05, 3.4757e-06, 2.5313e-06, ..., 4.3176e-06, + 5.6326e-06, 5.4464e-06], + ..., + [-1.4734e-03, -1.8299e-04, -2.9349e-04, ..., -5.2118e-04, + -7.2861e-04, -7.3338e-04], + [ 5.3227e-05, 2.0131e-05, 2.0579e-05, ..., 4.8876e-06, + 7.8529e-06, 7.9647e-06], + [-1.9717e-04, -8.7500e-05, -8.7082e-05, ..., -3.6918e-06, + -1.1519e-05, -1.1921e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0444, -0.0149, 0.0270, 0.0153, -0.0508, -0.0138, -0.0022], + device='cuda:0'), grad: tensor([ 4.4670e-03, 1.7548e-04, 4.0948e-05, -9.1374e-05, -4.3602e-03, + 1.0496e-04, -3.3760e-04], device='cuda:0') +588 +0.005892784473993186 +changing lr +epoch 31, time 813.01, cls_loss 0.0045 cls_loss_mapping 0.0195 cls_loss_causal 0.4568 re_mapping 0.0153 re_causal 0.0249 /// teacc 93.72 lr 0.00567117 +Epoch 33, weight, value: tensor([[ 0.0805, 0.0723, 0.0664, ..., 0.0227, 0.0647, 0.0585], + [ 0.0396, 0.0276, 0.0089, ..., -0.0171, -0.0156, -0.0135], + [-0.0233, -0.0305, 0.0070, ..., 0.0051, -0.0351, -0.0142], + ..., + [-0.1332, -0.1033, -0.0812, ..., 0.0383, 0.0413, 0.0259], + [ 0.0997, 0.0750, 0.0731, ..., -0.0504, -0.0631, -0.0561], + [ 0.0031, 0.0189, -0.0125, ..., -0.0522, -0.0249, -0.0353]], + device='cuda:0'), grad: tensor([[-1.4362e-03, -3.2687e-04, -3.6669e-04, ..., -3.9530e-04, + -6.0272e-04, -6.3896e-04], + [ 8.1539e-05, 1.5169e-05, 1.7136e-05, ..., 1.8433e-05, + 3.4511e-05, 3.5882e-05], + [ 2.3198e-04, 6.6042e-05, 6.8784e-05, ..., 7.9215e-05, + 1.0860e-04, 1.1557e-04], + ..., + [ 2.8539e-04, 5.8323e-05, 6.8307e-05, ..., 6.3658e-05, + 1.0639e-04, 1.1194e-04], + [ 3.8475e-05, 8.2403e-06, 9.5218e-06, ..., 7.7859e-06, + 1.3344e-05, 1.3851e-05], + [ 1.0186e-04, 2.2292e-05, 2.5555e-05, ..., 2.1353e-05, + 3.5822e-05, 3.7313e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0503, -0.0127, 0.0242, 0.0137, -0.0515, -0.0136, -0.0056], + device='cuda:0'), grad: tensor([-3.3913e-03, 1.8573e-04, 5.0020e-04, 1.6565e-03, 7.0381e-04, + 9.3997e-05, 2.4843e-04], device='cuda:0') +588 +0.00567116632908828 +changing lr +epoch 32, time 813.20, cls_loss 0.0044 cls_loss_mapping 0.0188 cls_loss_causal 0.4588 re_mapping 0.0156 re_causal 0.0258 /// teacc 91.96 lr 0.00544820 +Epoch 34, weight, value: tensor([[ 0.0806, 0.0727, 0.0669, ..., 0.0224, 0.0639, 0.0578], + [ 0.0380, 0.0269, 0.0086, ..., -0.0171, -0.0157, -0.0136], + [-0.0225, -0.0304, 0.0066, ..., 0.0050, -0.0343, -0.0138], + ..., + [-0.1327, -0.1026, -0.0809, ..., 0.0376, 0.0405, 0.0254], + [ 0.0984, 0.0742, 0.0722, ..., -0.0497, -0.0624, -0.0554], + [ 0.0033, 0.0186, -0.0124, ..., -0.0514, -0.0246, -0.0348]], + device='cuda:0'), grad: tensor([[ 7.3204e-03, 2.9449e-03, 3.4027e-03, ..., 8.6594e-04, + 1.3466e-03, 1.3685e-03], + [-8.4152e-03, -3.3875e-03, -3.9139e-03, ..., -9.9945e-04, + -1.5469e-03, -1.5726e-03], + [ 2.3198e-04, 7.9691e-05, 9.4295e-05, ..., 2.4244e-05, + 4.5002e-05, 4.4882e-05], + ..., + [ 2.0623e-05, 3.5644e-05, 3.6538e-05, ..., 1.2584e-05, + 2.5295e-06, 4.0717e-06], + [ 2.9469e-04, 1.1790e-04, 1.3626e-04, ..., 3.4988e-05, + 5.4449e-05, 5.5283e-05], + [ 2.4796e-04, 9.5606e-05, 1.1057e-04, ..., 2.9609e-05, + 4.6521e-05, 4.6968e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0493, -0.0150, 0.0264, 0.0159, -0.0530, -0.0139, -0.0048], + device='cuda:0'), grad: tensor([ 1.3618e-02, -1.5640e-02, 4.8470e-04, 5.6648e-04, -6.4909e-05, + 5.5075e-04, 4.7946e-04], device='cuda:0') +588 +0.00544819654451717 +changing lr +epoch 33, time 818.78, cls_loss 0.0023 cls_loss_mapping 0.0129 cls_loss_causal 0.4522 re_mapping 0.0148 re_causal 0.0254 /// teacc 91.96 lr 0.00522432 +Epoch 35, weight, value: tensor([[ 0.0796, 0.0724, 0.0668, ..., 0.0219, 0.0628, 0.0569], + [ 0.0381, 0.0267, 0.0087, ..., -0.0168, -0.0154, -0.0133], + [-0.0228, -0.0304, 0.0060, ..., 0.0050, -0.0337, -0.0136], + ..., + [-0.1316, -0.1018, -0.0804, ..., 0.0372, 0.0401, 0.0253], + [ 0.0989, 0.0742, 0.0721, ..., -0.0490, -0.0614, -0.0546], + [ 0.0030, 0.0181, -0.0125, ..., -0.0507, -0.0242, -0.0343]], + device='cuda:0'), grad: tensor([[ 2.1183e-04, 5.8502e-05, 6.9916e-05, ..., 1.9699e-05, + 5.2303e-05, 4.9174e-05], + [-8.1635e-04, -2.5177e-04, -3.0231e-04, ..., -5.6922e-05, + -1.7893e-04, -1.6677e-04], + [ 1.2612e-04, 2.9176e-05, 3.7014e-05, ..., 2.0146e-05, + 3.7521e-05, 3.6120e-05], + ..., + [-2.9594e-05, 7.0855e-06, 7.8827e-06, ..., -1.1660e-05, + -1.6466e-05, -1.6406e-05], + [ 7.7128e-05, 2.2039e-05, 2.6911e-05, ..., 6.6794e-06, + 1.7837e-05, 1.6809e-05], + [ 3.9768e-04, 1.2183e-04, 1.4615e-04, ..., 2.8759e-05, + 8.8274e-05, 8.2433e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0468, -0.0135, 0.0256, 0.0136, -0.0524, -0.0108, -0.0047], + device='cuda:0'), grad: tensor([ 4.9591e-04, -1.8606e-03, 3.1376e-04, 7.2539e-05, -1.0914e-04, + 1.8167e-04, 9.0694e-04], device='cuda:0') +588 +0.005224324151752577 +changing lr +epoch 34, time 809.73, cls_loss 0.0030 cls_loss_mapping 0.0131 cls_loss_causal 0.4321 re_mapping 0.0133 re_causal 0.0234 /// teacc 93.72 lr 0.00500000 +Epoch 36, weight, value: tensor([[ 0.0792, 0.0726, 0.0671, ..., 0.0215, 0.0620, 0.0561], + [ 0.0384, 0.0266, 0.0090, ..., -0.0166, -0.0150, -0.0129], + [-0.0234, -0.0303, 0.0054, ..., 0.0045, -0.0336, -0.0138], + ..., + [-0.1299, -0.1010, -0.0799, ..., 0.0368, 0.0399, 0.0252], + [ 0.0972, 0.0733, 0.0713, ..., -0.0484, -0.0608, -0.0540], + [ 0.0034, 0.0177, -0.0125, ..., -0.0499, -0.0239, -0.0338]], + device='cuda:0'), grad: tensor([[-3.4839e-05, -1.7911e-05, -1.8388e-05, ..., -9.1717e-06, + -1.5274e-05, -1.3106e-05], + [-4.1604e-05, -1.1638e-05, -1.3761e-05, ..., -7.8678e-06, + -1.2323e-05, -1.2226e-05], + [-2.9266e-05, -5.5470e-06, -6.1207e-06, ..., -4.1202e-06, + -8.7023e-06, -8.1137e-06], + ..., + [ 3.9935e-05, 1.6361e-05, 1.7539e-05, ..., 9.4622e-06, + 1.5900e-05, 1.4067e-05], + [ 9.3356e-06, 2.7418e-06, 3.0492e-06, ..., 1.7481e-06, + 2.9411e-06, 2.8014e-06], + [ 2.9460e-05, 9.2387e-06, 1.0267e-05, ..., 5.3309e-06, + 9.0525e-06, 8.5756e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0453, -0.0114, 0.0233, 0.0128, -0.0499, -0.0123, -0.0032], + device='cuda:0'), grad: tensor([-7.5638e-05, -9.5069e-05, -7.8142e-05, 6.7353e-05, 8.9824e-05, + 2.2471e-05, 6.9141e-05], device='cuda:0') +588 +0.005000000000000003 +changing lr +epoch 35, time 808.54, cls_loss 0.0037 cls_loss_mapping 0.0152 cls_loss_causal 0.4379 re_mapping 0.0126 re_causal 0.0233 /// teacc 92.71 lr 0.00477568 +Epoch 37, weight, value: tensor([[ 0.0787, 0.0727, 0.0673, ..., 0.0212, 0.0613, 0.0555], + [ 0.0376, 0.0263, 0.0089, ..., -0.0166, -0.0151, -0.0130], + [-0.0220, -0.0300, 0.0053, ..., 0.0042, -0.0331, -0.0135], + ..., + [-0.1292, -0.1005, -0.0797, ..., 0.0367, 0.0396, 0.0252], + [ 0.0957, 0.0724, 0.0704, ..., -0.0479, -0.0602, -0.0536], + [ 0.0029, 0.0175, -0.0124, ..., -0.0495, -0.0241, -0.0338]], + device='cuda:0'), grad: tensor([[ 1.1081e-04, 9.8497e-06, 2.6390e-05, ..., 5.8085e-05, + 7.8619e-05, 7.5281e-05], + [ 2.9993e-04, 3.1650e-05, 7.5817e-05, ..., 1.5366e-04, + 2.0933e-04, 2.0015e-04], + [ 6.7055e-05, 7.0333e-06, 1.7270e-05, ..., 3.8087e-05, + 5.1022e-05, 4.8697e-05], + ..., + [ 3.2514e-05, 4.0419e-06, 8.8885e-06, ..., 1.5900e-05, + 2.1756e-05, 2.0742e-05], + [ 2.3872e-05, 2.6692e-06, 6.0759e-06, ..., 1.1623e-05, + 1.6004e-05, 1.5318e-05], + [ 9.1270e-06, -7.8883e-07, 1.5451e-06, ..., 7.1563e-06, + 9.1866e-06, 8.8215e-06]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0436, -0.0122, 0.0267, 0.0142, -0.0493, -0.0133, -0.0051], + device='cuda:0'), grad: tensor([ 3.1781e-04, 8.4877e-04, 1.8942e-04, -1.5469e-03, 9.0361e-05, + 6.7234e-05, 3.1173e-05], device='cuda:0') +588 +0.004775675848247429 +changing lr +epoch 36, time 818.63, cls_loss 0.0022 cls_loss_mapping 0.0121 cls_loss_causal 0.4217 re_mapping 0.0124 re_causal 0.0234 /// teacc 93.47 lr 0.00455180 +Epoch 38, weight, value: tensor([[ 0.0800, 0.0731, 0.0679, ..., 0.0211, 0.0610, 0.0552], + [ 0.0368, 0.0258, 0.0087, ..., -0.0164, -0.0150, -0.0129], + [-0.0223, -0.0300, 0.0047, ..., 0.0039, -0.0329, -0.0136], + ..., + [-0.1282, -0.0996, -0.0792, ..., 0.0364, 0.0393, 0.0251], + [ 0.0949, 0.0719, 0.0699, ..., -0.0473, -0.0595, -0.0529], + [ 0.0025, 0.0170, -0.0125, ..., -0.0490, -0.0238, -0.0334]], + device='cuda:0'), grad: tensor([[-7.7605e-05, -4.7684e-05, -4.8161e-05, ..., -6.0908e-07, + -3.0603e-06, -4.1500e-06], + [ 4.1366e-05, 2.1577e-05, 2.2545e-05, ..., 2.4587e-06, + 4.7572e-06, 4.7386e-06], + [-4.3094e-05, -7.4506e-06, -1.1407e-05, ..., -1.2629e-05, + -1.9446e-05, -1.6555e-05], + ..., + [ 9.5442e-06, 3.5148e-06, 4.1686e-06, ..., 1.9278e-06, + 2.9113e-06, 2.4792e-06], + [ 1.3210e-05, 6.4522e-06, 6.8471e-06, ..., 8.5309e-07, + 1.6466e-06, 1.6224e-06], + [ 2.2411e-05, 8.8289e-06, 1.0006e-05, ..., 2.2650e-06, + 4.0941e-06, 3.7327e-06]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0465, -0.0125, 0.0255, 0.0122, -0.0490, -0.0131, -0.0052], + device='cuda:0'), grad: tensor([-8.8215e-05, 6.2108e-05, -1.2684e-04, 6.6817e-05, 2.1145e-05, + 2.1160e-05, 4.3809e-05], device='cuda:0') +588 +0.004551803455482836 +changing lr +epoch 37, time 812.52, cls_loss 0.0024 cls_loss_mapping 0.0128 cls_loss_causal 0.4274 re_mapping 0.0113 re_causal 0.0217 /// teacc 93.97 lr 0.00432883 +Epoch 39, weight, value: tensor([[ 0.0801, 0.0732, 0.0682, ..., 0.0209, 0.0604, 0.0547], + [ 0.0360, 0.0253, 0.0084, ..., -0.0163, -0.0149, -0.0129], + [-0.0223, -0.0298, 0.0044, ..., 0.0036, -0.0327, -0.0137], + ..., + [-0.1261, -0.0988, -0.0787, ..., 0.0363, 0.0394, 0.0253], + [ 0.0938, 0.0713, 0.0693, ..., -0.0468, -0.0590, -0.0524], + [ 0.0012, 0.0164, -0.0128, ..., -0.0486, -0.0240, -0.0334]], + device='cuda:0'), grad: tensor([[ 1.3208e-03, 2.7132e-04, 2.7966e-04, ..., 2.1148e-04, + 3.4261e-04, 3.1471e-04], + [ 4.4847e-04, 6.2883e-05, 8.6010e-05, ..., 1.6391e-04, + 2.0814e-04, 1.9217e-04], + [-2.8467e-04, -5.8651e-05, -8.5235e-05, ..., -9.0823e-06, + -5.7757e-05, -7.6950e-05], + ..., + [-1.1253e-03, -2.3699e-04, -2.0397e-04, ..., -1.2493e-04, + -2.4390e-04, -2.0444e-04], + [ 1.0937e-04, 1.8716e-05, 2.3216e-05, ..., 2.8864e-05, + 4.0889e-05, 3.8743e-05], + [ 3.8415e-05, -2.9765e-06, 2.0452e-06, ..., 1.7777e-05, + 2.2992e-05, 2.1994e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0464, -0.0130, 0.0248, 0.0131, -0.0451, -0.0139, -0.0080], + device='cuda:0'), grad: tensor([ 0.0036, 0.0013, -0.0009, -0.0013, -0.0031, 0.0003, 0.0002], + device='cuda:0') +588 +0.004328833670911726 +changing lr +epoch 38, time 816.59, cls_loss 0.0036 cls_loss_mapping 0.0166 cls_loss_causal 0.4544 re_mapping 0.0109 re_causal 0.0217 /// teacc 92.96 lr 0.00410722 +Epoch 40, weight, value: tensor([[ 0.0799, 0.0733, 0.0683, ..., 0.0205, 0.0596, 0.0540], + [ 0.0372, 0.0255, 0.0089, ..., -0.0160, -0.0144, -0.0124], + [-0.0221, -0.0297, 0.0042, ..., 0.0036, -0.0322, -0.0135], + ..., + [-0.1258, -0.0986, -0.0787, ..., 0.0360, 0.0391, 0.0252], + [ 0.0927, 0.0707, 0.0687, ..., -0.0464, -0.0585, -0.0520], + [ 0.0008, 0.0161, -0.0128, ..., -0.0483, -0.0240, -0.0333]], + device='cuda:0'), grad: tensor([[ 9.0182e-05, 7.5512e-06, 4.5821e-06, ..., 1.2167e-05, + 2.6450e-05, 2.8521e-05], + [ 5.2541e-05, 8.3819e-06, 1.0535e-05, ..., 1.2502e-05, + 2.0817e-05, 1.9684e-05], + [-3.5405e-04, -8.3625e-05, -8.7738e-05, ..., -1.2740e-05, + -4.4912e-05, -5.0306e-05], + ..., + [ 9.5487e-05, 4.2170e-05, 3.9607e-05, ..., -3.0175e-05, + -3.2872e-05, -2.7359e-05], + [ 2.5019e-05, 5.5432e-06, 7.0781e-06, ..., 4.3400e-06, + 7.1600e-06, 6.7949e-06], + [ 4.0144e-05, 8.8736e-06, 1.1116e-05, ..., 5.5581e-06, + 9.4995e-06, 9.3132e-06]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0456, -0.0094, 0.0255, 0.0118, -0.0453, -0.0149, -0.0090], + device='cuda:0'), grad: tensor([ 2.6417e-04, 1.5450e-04, -8.4782e-04, 1.2887e-04, 1.3530e-04, + 6.4671e-05, 1.0115e-04], device='cuda:0') +588 +0.0041072155260068206 +changing lr +epoch 39, time 810.16, cls_loss 0.0023 cls_loss_mapping 0.0103 cls_loss_causal 0.4366 re_mapping 0.0109 re_causal 0.0218 /// teacc 92.96 lr 0.00388740 +Epoch 41, weight, value: tensor([[ 0.0813, 0.0739, 0.0691, ..., 0.0206, 0.0597, 0.0541], + [ 0.0351, 0.0247, 0.0081, ..., -0.0162, -0.0148, -0.0127], + [-0.0220, -0.0297, 0.0037, ..., 0.0036, -0.0318, -0.0133], + ..., + [-0.1254, -0.0982, -0.0785, ..., 0.0355, 0.0385, 0.0248], + [ 0.0919, 0.0703, 0.0683, ..., -0.0460, -0.0580, -0.0516], + [ 0.0014, 0.0162, -0.0125, ..., -0.0478, -0.0238, -0.0330]], + device='cuda:0'), grad: tensor([[ 3.0786e-05, 6.2510e-06, 6.6385e-06, ..., 6.4000e-06, + 1.0177e-05, 9.1493e-06], + [ 5.1297e-06, 7.0129e-07, 9.4157e-07, ..., 2.5406e-06, + 3.5912e-06, 3.3006e-06], + [-1.6153e-04, -3.6359e-05, -4.1485e-05, ..., -3.7223e-05, + -5.7846e-05, -5.2512e-05], + ..., + [ 4.1604e-05, 9.7454e-06, 1.1414e-05, ..., 4.6752e-06, + 9.4995e-06, 8.5384e-06], + [ 5.7332e-06, 1.5143e-06, 1.7388e-06, ..., 9.4809e-07, + 1.5926e-06, 1.4519e-06], + [ 1.7926e-05, 4.1164e-06, 4.8168e-06, ..., 3.5632e-06, + 5.8189e-06, 5.2750e-06]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0489, -0.0132, 0.0257, 0.0117, -0.0461, -0.0150, -0.0077], + device='cuda:0'), grad: tensor([ 8.4579e-05, 1.6034e-05, -4.2963e-04, 1.6403e-04, 1.0425e-04, + 1.4126e-05, 4.6611e-05], device='cuda:0') +588 +0.0038873953302184317 +changing lr +epoch 40, time 817.72, cls_loss 0.0031 cls_loss_mapping 0.0110 cls_loss_causal 0.4158 re_mapping 0.0106 re_causal 0.0213 /// teacc 93.72 lr 0.00366982 +Epoch 42, weight, value: tensor([[ 0.0811, 0.0739, 0.0692, ..., 0.0203, 0.0591, 0.0535], + [ 0.0352, 0.0245, 0.0082, ..., -0.0161, -0.0146, -0.0125], + [-0.0223, -0.0296, 0.0034, ..., 0.0033, -0.0318, -0.0134], + ..., + [-0.1249, -0.0977, -0.0782, ..., 0.0354, 0.0383, 0.0248], + [ 0.0923, 0.0700, 0.0680, ..., -0.0454, -0.0573, -0.0509], + [ 0.0008, 0.0158, -0.0126, ..., -0.0474, -0.0237, -0.0328]], + device='cuda:0'), grad: tensor([[-5.3596e-04, -1.5938e-04, -1.5640e-04, ..., -1.0812e-04, + -1.7965e-04, -1.6940e-04], + [ 2.6524e-05, 8.6874e-06, 8.8885e-06, ..., 5.1670e-06, + 8.2627e-06, 7.8455e-06], + [ 3.5048e-04, 1.0979e-04, 1.0973e-04, ..., 6.7711e-05, + 1.1122e-04, 1.0532e-04], + ..., + [ 5.9277e-05, 2.0459e-05, 2.0370e-05, ..., 9.4101e-06, + 1.5780e-05, 1.4931e-05], + [-2.8157e-04, -1.4019e-04, -1.5092e-04, ..., -6.7241e-06, + -9.8944e-06, -1.2219e-05], + [ 2.2864e-04, 1.0979e-04, 1.1772e-04, ..., 8.9556e-06, + 1.3918e-05, 1.5289e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0479, -0.0116, 0.0243, 0.0106, -0.0460, -0.0122, -0.0089], + device='cuda:0'), grad: tensor([-1.2474e-03, 5.8174e-05, 7.9107e-04, 3.3116e-04, 1.2589e-04, + -4.0770e-04, 3.4833e-04], device='cuda:0') +588 +0.003669815772166629 +changing lr +epoch 41, time 817.28, cls_loss 0.0020 cls_loss_mapping 0.0085 cls_loss_causal 0.4236 re_mapping 0.0098 re_causal 0.0200 /// teacc 92.71 lr 0.00345492 +Epoch 43, weight, value: tensor([[ 0.0810, 0.0739, 0.0693, ..., 0.0202, 0.0586, 0.0532], + [ 0.0355, 0.0243, 0.0082, ..., -0.0159, -0.0144, -0.0124], + [-0.0226, -0.0296, 0.0030, ..., 0.0031, -0.0316, -0.0135], + ..., + [-0.1245, -0.0972, -0.0780, ..., 0.0350, 0.0379, 0.0244], + [ 0.0916, 0.0696, 0.0677, ..., -0.0450, -0.0569, -0.0506], + [ 0.0011, 0.0156, -0.0125, ..., -0.0469, -0.0234, -0.0324]], + device='cuda:0'), grad: tensor([[ 5.7995e-05, -2.0489e-07, 3.4031e-06, ..., 3.5316e-05, + 2.9922e-05, 3.0935e-05], + [ 1.4651e-04, 1.6674e-05, 2.3216e-05, ..., 5.4210e-05, + 5.6028e-05, 5.6356e-05], + [ 2.0862e-04, 3.3289e-05, 4.1366e-05, ..., 6.0529e-05, + 7.0274e-05, 6.9916e-05], + ..., + [ 1.3411e-04, 1.4596e-05, 2.4140e-05, ..., 5.3406e-05, + 5.9366e-05, 6.0856e-05], + [ 5.0902e-05, 5.4762e-06, 8.8289e-06, ..., 2.0519e-05, + 2.2277e-05, 2.2873e-05], + [-3.6061e-05, -1.9625e-05, -1.4506e-05, ..., 1.0744e-05, + 6.2138e-06, 7.1488e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0477, -0.0095, 0.0231, 0.0105, -0.0466, -0.0127, -0.0083], + device='cuda:0'), grad: tensor([ 1.6332e-04, 4.1175e-04, 5.6458e-04, -1.6212e-03, 3.8433e-04, + 1.4436e-04, -4.7505e-05], device='cuda:0') +588 +0.0034549150281252667 +changing lr +epoch 42, time 812.58, cls_loss 0.0022 cls_loss_mapping 0.0107 cls_loss_causal 0.4164 re_mapping 0.0095 re_causal 0.0194 /// teacc 93.47 lr 0.00324313 +Epoch 44, weight, value: tensor([[ 0.0800, 0.0737, 0.0691, ..., 0.0197, 0.0578, 0.0524], + [ 0.0340, 0.0236, 0.0077, ..., -0.0159, -0.0144, -0.0124], + [-0.0221, -0.0294, 0.0029, ..., 0.0031, -0.0312, -0.0132], + ..., + [-0.1233, -0.0966, -0.0776, ..., 0.0348, 0.0378, 0.0244], + [ 0.0914, 0.0695, 0.0676, ..., -0.0447, -0.0565, -0.0502], + [ 0.0014, 0.0155, -0.0124, ..., -0.0465, -0.0231, -0.0320]], + device='cuda:0'), grad: tensor([[-3.1042e-04, -1.0139e-04, -1.1402e-04, ..., -6.8963e-05, + -1.1438e-04, -1.0639e-04], + [ 5.1528e-05, 1.2212e-05, 1.5438e-05, ..., 1.2323e-05, + 2.0087e-05, 1.9476e-05], + [ 3.5435e-05, 9.5963e-06, 1.1921e-05, ..., 7.0035e-06, + 1.2219e-05, 1.1384e-05], + ..., + [ 4.0025e-05, 1.4938e-05, 1.6630e-05, ..., 5.5954e-06, + 1.2115e-05, 1.0990e-05], + [-9.4846e-06, -7.3239e-06, -6.9775e-06, ..., 7.0781e-07, + 1.2163e-06, 1.0170e-06], + [ 8.6725e-05, 4.3094e-05, 4.1217e-05, ..., 2.2948e-05, + 3.1769e-05, 2.8774e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0451, -0.0115, 0.0242, 0.0110, -0.0447, -0.0125, -0.0075], + device='cuda:0'), grad: tensor([-7.4959e-04, 1.3137e-04, 9.0301e-05, 2.6941e-04, 9.0837e-05, + -4.8429e-06, 1.7333e-04], device='cuda:0') +588 +0.0032431258795932905 +changing lr +epoch 43, time 804.92, cls_loss 0.0019 cls_loss_mapping 0.0109 cls_loss_causal 0.4262 re_mapping 0.0091 re_causal 0.0198 /// teacc 93.47 lr 0.00303487 +Epoch 45, weight, value: tensor([[ 0.0801, 0.0737, 0.0692, ..., 0.0195, 0.0574, 0.0520], + [ 0.0328, 0.0230, 0.0073, ..., -0.0160, -0.0145, -0.0125], + [-0.0228, -0.0296, 0.0025, ..., 0.0029, -0.0311, -0.0133], + ..., + [-0.1231, -0.0964, -0.0774, ..., 0.0345, 0.0374, 0.0242], + [ 0.0925, 0.0699, 0.0679, ..., -0.0442, -0.0558, -0.0496], + [ 0.0019, 0.0154, -0.0123, ..., -0.0459, -0.0228, -0.0316]], + device='cuda:0'), grad: tensor([[-1.1873e-04, -5.6744e-05, -5.7548e-05, ..., -4.7311e-06, + -1.7673e-05, -1.6883e-05], + [-4.3958e-05, -7.3984e-06, -9.5293e-06, ..., -1.9416e-05, + -2.5347e-05, -2.5064e-05], + [ 6.7830e-05, 2.6092e-05, 2.6479e-05, ..., 5.3681e-06, + 1.3813e-05, 1.3068e-05], + ..., + [ 5.3883e-05, 2.4945e-05, 2.5496e-05, ..., 8.3223e-06, + 1.2457e-05, 1.2666e-05], + [ 2.9638e-05, 1.0327e-05, 1.0647e-05, ..., 4.1239e-06, + 7.9423e-06, 7.6257e-06], + [-1.5184e-05, -6.3851e-06, -4.7423e-06, ..., 1.8915e-06, + 8.3167e-07, 1.0170e-06]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0453, -0.0129, 0.0222, 0.0099, -0.0452, -0.0095, -0.0056], + device='cuda:0'), grad: tensor([-2.0373e-04, -1.1629e-04, 1.4174e-04, 5.9426e-05, 8.1062e-05, + 6.4790e-05, -2.6986e-05], device='cuda:0') +588 +0.0030348748417303863 +changing lr +epoch 44, time 804.95, cls_loss 0.0013 cls_loss_mapping 0.0102 cls_loss_causal 0.4315 re_mapping 0.0091 re_causal 0.0199 /// teacc 93.97 lr 0.00283058 +Epoch 46, weight, value: tensor([[ 0.0796, 0.0736, 0.0691, ..., 0.0193, 0.0569, 0.0516], + [ 0.0330, 0.0230, 0.0074, ..., -0.0159, -0.0143, -0.0123], + [-0.0223, -0.0294, 0.0024, ..., 0.0028, -0.0309, -0.0132], + ..., + [-0.1218, -0.0958, -0.0770, ..., 0.0343, 0.0374, 0.0242], + [ 0.0917, 0.0694, 0.0674, ..., -0.0440, -0.0556, -0.0494], + [ 0.0009, 0.0151, -0.0125, ..., -0.0457, -0.0228, -0.0316]], + device='cuda:0'), grad: tensor([[ 8.4102e-05, 1.2226e-05, 1.5602e-05, ..., 2.7627e-05, + 3.4750e-05, 3.2485e-05], + [ 4.2558e-05, 6.7316e-06, 8.8438e-06, ..., 1.4171e-05, + 1.5810e-05, 1.5028e-05], + [-5.0277e-05, -7.6517e-06, -8.5086e-06, ..., -2.5943e-05, + -3.3945e-05, -3.1084e-05], + ..., + [-1.1601e-05, -1.3327e-06, -7.1526e-07, ..., -3.8464e-07, + -3.6620e-06, -3.3043e-06], + [ 2.0102e-05, 3.0212e-06, 3.9376e-06, ..., 6.6571e-06, + 8.0541e-06, 7.5586e-06], + [ 3.5107e-05, 6.1132e-06, 7.5251e-06, ..., 1.1280e-05, + 1.3128e-05, 1.2472e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0443, -0.0117, 0.0232, 0.0096, -0.0433, -0.0102, -0.0078], + device='cuda:0'), grad: tensor([ 2.4033e-04, 1.1915e-04, -2.0254e-04, -2.9182e-04, -2.3723e-05, + 5.7220e-05, 1.0157e-04], device='cuda:0') +588 +0.0028305813044122124 +changing lr +---------------------saving model at epoch 45---------------------------------------------------- +epoch 45, time 802.96, cls_loss 0.0019 cls_loss_mapping 0.0101 cls_loss_causal 0.4210 re_mapping 0.0088 re_causal 0.0195 /// teacc 95.23 lr 0.00263066 +Epoch 47, weight, value: tensor([[ 0.0792, 0.0736, 0.0691, ..., 0.0191, 0.0565, 0.0512], + [ 0.0332, 0.0229, 0.0075, ..., -0.0157, -0.0141, -0.0122], + [-0.0215, -0.0292, 0.0023, ..., 0.0029, -0.0303, -0.0129], + ..., + [-0.1216, -0.0955, -0.0769, ..., 0.0340, 0.0371, 0.0240], + [ 0.0911, 0.0690, 0.0670, ..., -0.0437, -0.0553, -0.0491], + [ 0.0004, 0.0148, -0.0126, ..., -0.0455, -0.0229, -0.0315]], + device='cuda:0'), grad: tensor([[-1.8418e-05, -7.2941e-06, -5.7556e-06, ..., 1.8589e-06, + 5.3830e-07, -3.7253e-08], + [-7.3433e-05, -2.5347e-05, -2.9951e-05, ..., -9.6112e-06, + -1.9461e-05, -1.8835e-05], + [ 3.6359e-05, 8.6427e-06, 9.7007e-06, ..., 1.0766e-05, + 1.2040e-05, 1.2040e-05], + ..., + [ 5.8889e-05, 1.8105e-05, 2.0280e-05, ..., 9.3952e-06, + 1.4931e-05, 1.5140e-05], + [ 1.7866e-05, 5.5432e-06, 6.1654e-06, ..., 3.0100e-06, + 4.7572e-06, 4.6492e-06], + [ 2.5898e-05, 7.6815e-06, 8.3670e-06, ..., 5.1856e-06, + 7.3612e-06, 7.0445e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0431, -0.0104, 0.0253, 0.0092, -0.0437, -0.0105, -0.0089], + device='cuda:0'), grad: tensor([-3.0026e-05, -1.9670e-04, 8.9884e-05, -1.2034e-04, 1.3995e-04, + 4.6521e-05, 7.1287e-05], device='cuda:0') +588 +0.0026306566876350096 +changing lr +epoch 46, time 796.28, cls_loss 0.0014 cls_loss_mapping 0.0082 cls_loss_causal 0.4007 re_mapping 0.0084 re_causal 0.0181 /// teacc 93.97 lr 0.00243550 +Epoch 48, weight, value: tensor([[ 0.0800, 0.0739, 0.0695, ..., 0.0190, 0.0563, 0.0510], + [ 0.0325, 0.0226, 0.0074, ..., -0.0158, -0.0142, -0.0122], + [-0.0215, -0.0291, 0.0021, ..., 0.0028, -0.0302, -0.0128], + ..., + [-0.1211, -0.0952, -0.0767, ..., 0.0339, 0.0370, 0.0240], + [ 0.0904, 0.0686, 0.0666, ..., -0.0434, -0.0550, -0.0489], + [ 0.0002, 0.0146, -0.0126, ..., -0.0452, -0.0228, -0.0314]], + device='cuda:0'), grad: tensor([[-9.3728e-06, -8.9779e-06, -1.1541e-05, ..., -3.2373e-06, + -2.8387e-06, -2.4326e-06], + [-2.7701e-05, -5.1409e-06, -5.3905e-06, ..., -1.6578e-06, + -6.9477e-06, -6.6012e-06], + [-2.6628e-05, -4.9174e-06, -5.9009e-06, ..., -5.7742e-07, + -5.8301e-07, -1.8738e-06], + ..., + [ 1.3180e-05, 5.2154e-06, 5.7928e-06, ..., 8.4750e-07, + -2.2538e-07, -2.4959e-07], + [ 1.4670e-05, 4.6529e-06, 5.8711e-06, ..., 2.4792e-06, + 3.8445e-06, 4.0159e-06], + [ 2.3812e-05, 6.0014e-06, 7.0669e-06, ..., 2.8089e-06, + 5.7369e-06, 5.8338e-06]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0447, -0.0115, 0.0252, 0.0092, -0.0433, -0.0108, -0.0094], + device='cuda:0'), grad: tensor([ 3.0175e-06, -8.2195e-05, -6.7174e-05, 3.0354e-05, 2.5079e-05, + 3.1263e-05, 5.9664e-05], device='cuda:0') +588 +0.0024355036129704724 +changing lr +epoch 47, time 792.94, cls_loss 0.0016 cls_loss_mapping 0.0101 cls_loss_causal 0.4051 re_mapping 0.0084 re_causal 0.0183 /// teacc 94.97 lr 0.00224552 +Epoch 49, weight, value: tensor([[ 0.0804, 0.0740, 0.0697, ..., 0.0189, 0.0560, 0.0508], + [ 0.0324, 0.0225, 0.0073, ..., -0.0157, -0.0141, -0.0122], + [-0.0214, -0.0291, 0.0019, ..., 0.0027, -0.0300, -0.0128], + ..., + [-0.1210, -0.0949, -0.0766, ..., 0.0338, 0.0368, 0.0239], + [ 0.0896, 0.0683, 0.0662, ..., -0.0432, -0.0548, -0.0487], + [ 0.0002, 0.0144, -0.0126, ..., -0.0449, -0.0227, -0.0312]], + device='cuda:0'), grad: tensor([[-4.8965e-05, -4.8161e-05, -3.5852e-05, ..., 9.9897e-05, + 1.1665e-04, 1.0639e-04], + [ 3.0613e-04, 4.1783e-05, 6.2168e-05, ..., 1.5008e-04, + 1.8573e-04, 1.8001e-04], + [ 1.8191e-04, 2.0936e-05, 4.0591e-05, ..., 1.1367e-04, + 1.3721e-04, 1.3256e-04], + ..., + [-3.9673e-04, -1.5929e-05, -6.0380e-05, ..., -3.6240e-04, + -4.4727e-04, -4.2868e-04], + [ 8.0705e-05, 1.0327e-05, 1.5706e-05, ..., 3.7074e-05, + 4.5300e-05, 4.4018e-05], + [ 7.8142e-05, 1.2510e-05, 1.6153e-05, ..., 2.5809e-05, + 3.1680e-05, 3.1292e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0453, -0.0111, 0.0253, 0.0092, -0.0439, -0.0116, -0.0092], + device='cuda:0'), grad: tensor([ 1.4640e-05, 8.7214e-04, 5.3501e-04, -5.8937e-04, -1.2760e-03, + 2.3127e-04, 2.1350e-04], device='cuda:0') +588 +0.00224551509273949 +changing lr +epoch 48, time 794.79, cls_loss 0.0011 cls_loss_mapping 0.0088 cls_loss_causal 0.3953 re_mapping 0.0084 re_causal 0.0184 /// teacc 93.47 lr 0.00206107 +Epoch 50, weight, value: tensor([[ 0.0810, 0.0742, 0.0699, ..., 0.0188, 0.0559, 0.0507], + [ 0.0323, 0.0223, 0.0073, ..., -0.0156, -0.0140, -0.0120], + [-0.0219, -0.0292, 0.0016, ..., 0.0026, -0.0300, -0.0129], + ..., + [-0.1206, -0.0946, -0.0764, ..., 0.0336, 0.0366, 0.0238], + [ 0.0894, 0.0681, 0.0661, ..., -0.0430, -0.0545, -0.0484], + [-0.0002, 0.0142, -0.0127, ..., -0.0447, -0.0226, -0.0312]], + device='cuda:0'), grad: tensor([[ 3.0398e-04, 5.4359e-05, 5.4806e-05, ..., 4.3452e-05, + 8.6904e-05, 8.8394e-05], + [ 1.2064e-04, 2.2903e-05, 2.6226e-05, ..., 3.2097e-05, + 4.6730e-05, 4.1187e-05], + [ 9.8765e-05, 1.6436e-05, 2.0295e-05, ..., 2.4199e-05, + 3.8147e-05, 3.3438e-05], + ..., + [-5.0640e-04, -8.4877e-05, -8.9824e-05, ..., -8.7798e-05, + -1.6844e-04, -1.5855e-04], + [ 4.8965e-05, 8.0839e-06, 9.2164e-06, ..., 1.0423e-05, + 1.8060e-05, 1.6093e-05], + [ 2.6956e-05, 4.8652e-06, 5.8636e-06, ..., 6.7279e-06, + 9.9689e-06, 8.8587e-06]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0467, -0.0107, 0.0239, 0.0093, -0.0439, -0.0116, -0.0097], + device='cuda:0'), grad: tensor([ 8.0776e-04, 3.3808e-04, 2.8300e-04, -2.2209e-04, -1.4238e-03, + 1.4186e-04, 7.4983e-05], device='cuda:0') +588 +0.002061073738537637 +changing lr +epoch 49, time 791.10, cls_loss 0.0014 cls_loss_mapping 0.0094 cls_loss_causal 0.4036 re_mapping 0.0082 re_causal 0.0180 /// teacc 95.23 lr 0.00188255 +Epoch 51, weight, value: tensor([[ 8.0744e-02, 7.4140e-02, 6.9905e-02, ..., 1.8685e-02, + 5.5529e-02, 5.0404e-02], + [ 3.1746e-02, 2.2061e-02, 7.1410e-03, ..., -1.5593e-02, + -1.3971e-02, -1.2023e-02], + [-2.2164e-02, -2.9223e-02, 1.3682e-03, ..., 2.5026e-03, + -2.9860e-02, -1.2878e-02], + ..., + [-1.1973e-01, -9.4191e-02, -7.6101e-02, ..., 3.3470e-02, + 3.6480e-02, 2.3775e-02], + [ 8.9263e-02, 6.8041e-02, 6.5975e-02, ..., -4.2811e-02, + -5.4277e-02, -4.8199e-02], + [-3.9018e-05, 1.4104e-02, -1.2632e-02, ..., -4.4504e-02, + -2.2524e-02, -3.0979e-02]], device='cuda:0'), grad: tensor([[ 6.5947e-04, 1.4544e-04, 1.5628e-04, ..., 1.0252e-04, + 1.5032e-04, 1.4842e-04], + [-2.4295e-04, -9.3803e-06, -1.9312e-05, ..., -1.9222e-05, + -4.0501e-05, -4.5538e-05], + [-2.1343e-03, -4.5037e-04, -4.8351e-04, ..., -4.4036e-04, + -6.3276e-04, -5.9366e-04], + ..., + [ 8.5306e-04, 1.6356e-04, 1.7834e-04, ..., 1.7262e-04, + 2.5058e-04, 2.3615e-04], + [ 2.0075e-04, 3.8862e-05, 4.3094e-05, ..., 3.9220e-05, + 5.7250e-05, 5.3853e-05], + [ 3.0589e-04, 5.1260e-05, 5.7101e-05, ..., 6.2466e-05, + 9.1553e-05, 8.6486e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0460, -0.0112, 0.0234, 0.0090, -0.0426, -0.0114, -0.0092], + device='cuda:0'), grad: tensor([ 0.0016, -0.0008, -0.0052, 0.0009, 0.0022, 0.0005, 0.0008], + device='cuda:0') +588 +0.0018825509907063344 +changing lr +epoch 50, time 795.94, cls_loss 0.0011 cls_loss_mapping 0.0068 cls_loss_causal 0.4129 re_mapping 0.0080 re_causal 0.0179 /// teacc 94.97 lr 0.00171031 +Epoch 52, weight, value: tensor([[ 0.0813, 0.0744, 0.0702, ..., 0.0186, 0.0554, 0.0503], + [ 0.0316, 0.0219, 0.0071, ..., -0.0156, -0.0139, -0.0120], + [-0.0220, -0.0292, 0.0012, ..., 0.0025, -0.0297, -0.0128], + ..., + [-0.1198, -0.0940, -0.0760, ..., 0.0332, 0.0362, 0.0235], + [ 0.0887, 0.0678, 0.0657, ..., -0.0427, -0.0541, -0.0481], + [-0.0004, 0.0139, -0.0127, ..., -0.0444, -0.0225, -0.0309]], + device='cuda:0'), grad: tensor([[ 1.1241e-04, 1.8403e-05, 2.2218e-05, ..., 9.9093e-06, + 1.7971e-05, 1.4402e-05], + [ 6.7770e-05, 1.4529e-05, 1.5959e-05, ..., 8.6650e-06, + 1.3560e-05, 1.1742e-05], + [-4.6802e-04, -1.0204e-04, -1.1444e-04, ..., -4.8071e-05, + -8.4698e-05, -7.0930e-05], + ..., + [ 1.3340e-04, 3.3885e-05, 3.7134e-05, ..., 2.2039e-05, + 3.2812e-05, 2.8864e-05], + [ 3.4034e-05, 7.4208e-06, 8.2627e-06, ..., 3.6862e-06, + 6.2957e-06, 5.3234e-06], + [ 7.0393e-05, 1.4640e-05, 1.6451e-05, ..., 6.6310e-06, + 1.1899e-05, 9.8571e-06]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0470, -0.0109, 0.0237, 0.0099, -0.0437, -0.0121, -0.0100], + device='cuda:0'), grad: tensor([ 2.9802e-04, 1.7095e-04, -1.1902e-03, 1.1951e-04, 3.3593e-04, + 8.6248e-05, 1.7893e-04], device='cuda:0') +588 +0.0017103063703014388 +changing lr +epoch 51, time 792.04, cls_loss 0.0011 cls_loss_mapping 0.0071 cls_loss_causal 0.3982 re_mapping 0.0078 re_causal 0.0171 /// teacc 94.22 lr 0.00154469 +Epoch 53, weight, value: tensor([[ 0.0812, 0.0743, 0.0702, ..., 0.0185, 0.0551, 0.0501], + [ 0.0315, 0.0219, 0.0071, ..., -0.0155, -0.0139, -0.0119], + [-0.0223, -0.0292, 0.0011, ..., 0.0024, -0.0296, -0.0128], + ..., + [-0.1192, -0.0937, -0.0758, ..., 0.0331, 0.0361, 0.0235], + [ 0.0883, 0.0675, 0.0655, ..., -0.0425, -0.0539, -0.0479], + [-0.0003, 0.0138, -0.0127, ..., -0.0441, -0.0224, -0.0308]], + device='cuda:0'), grad: tensor([[ 1.5008e-04, 5.9873e-05, 6.0588e-05, ..., 1.9684e-05, + 2.8446e-05, 2.8625e-05], + [ 7.9155e-05, 2.2858e-05, 2.1115e-05, ..., 1.0148e-05, + 1.1407e-05, 1.2696e-05], + [ 2.7150e-05, 3.9190e-05, 3.9279e-05, ..., -2.3127e-05, + -2.2963e-05, -2.4423e-05], + ..., + [ 4.2111e-05, 2.3544e-05, 2.2352e-05, ..., -1.0498e-05, + -1.4298e-05, -1.1802e-05], + [-1.4181e-03, -6.0606e-04, -6.0272e-04, ..., -9.5427e-05, + -1.4818e-04, -1.5831e-04], + [ 6.4373e-04, 2.7061e-04, 2.6965e-04, ..., 4.9829e-05, + 7.5281e-05, 7.9691e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0467, -0.0108, 0.0230, 0.0096, -0.0428, -0.0121, -0.0098], + device='cuda:0'), grad: tensor([ 2.8634e-04, 1.4770e-04, -2.5302e-05, 8.5211e-04, 1.3009e-05, + -2.3708e-03, 1.0977e-03], device='cuda:0') +588 +0.0015446867550656784 +changing lr +epoch 52, time 795.64, cls_loss 0.0012 cls_loss_mapping 0.0080 cls_loss_causal 0.4011 re_mapping 0.0074 re_causal 0.0164 /// teacc 94.72 lr 0.00138603 +Epoch 54, weight, value: tensor([[ 0.0809, 0.0743, 0.0701, ..., 0.0183, 0.0548, 0.0498], + [ 0.0313, 0.0217, 0.0071, ..., -0.0155, -0.0138, -0.0119], + [-0.0224, -0.0292, 0.0009, ..., 0.0023, -0.0295, -0.0128], + ..., + [-0.1189, -0.0935, -0.0757, ..., 0.0330, 0.0360, 0.0235], + [ 0.0875, 0.0672, 0.0651, ..., -0.0424, -0.0538, -0.0478], + [ 0.0007, 0.0141, -0.0123, ..., -0.0439, -0.0222, -0.0305]], + device='cuda:0'), grad: tensor([[ 5.2261e-04, 1.0097e-04, 1.3101e-04, ..., 6.6817e-05, + 1.1981e-04, 1.1861e-04], + [-1.7338e-03, -2.5797e-04, -3.6550e-04, ..., -4.3845e-04, + -7.2861e-04, -6.7663e-04], + [ 2.9945e-04, 2.9072e-05, 5.4806e-05, ..., 1.1230e-04, + 1.8954e-04, 1.6987e-04], + ..., + [ 2.6298e-04, 4.7296e-05, 6.0678e-05, ..., 4.7565e-05, + 7.9811e-05, 7.6592e-05], + [ 6.3241e-05, 1.1340e-05, 1.4648e-05, ..., 1.3143e-05, + 2.2039e-05, 2.0698e-05], + [ 3.9071e-05, 6.1700e-07, 3.9712e-06, ..., 1.5177e-05, + 2.3156e-05, 2.1845e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0460, -0.0107, 0.0226, 0.0094, -0.0424, -0.0133, -0.0077], + device='cuda:0'), grad: tensor([ 0.0013, -0.0054, 0.0011, 0.0019, 0.0007, 0.0002, 0.0001], + device='cuda:0') +588 +0.001386025680863044 +changing lr +epoch 53, time 793.18, cls_loss 0.0015 cls_loss_mapping 0.0074 cls_loss_causal 0.4159 re_mapping 0.0073 re_causal 0.0167 /// teacc 95.23 lr 0.00123464 +Epoch 55, weight, value: tensor([[ 0.0808, 0.0742, 0.0700, ..., 0.0183, 0.0547, 0.0497], + [ 0.0314, 0.0217, 0.0072, ..., -0.0154, -0.0137, -0.0118], + [-0.0221, -0.0291, 0.0009, ..., 0.0024, -0.0293, -0.0127], + ..., + [-0.1188, -0.0934, -0.0757, ..., 0.0329, 0.0359, 0.0234], + [ 0.0875, 0.0670, 0.0650, ..., -0.0422, -0.0536, -0.0476], + [ 0.0005, 0.0141, -0.0122, ..., -0.0438, -0.0222, -0.0305]], + device='cuda:0'), grad: tensor([[ 9.7215e-05, 3.1769e-05, 3.2514e-05, ..., 6.7428e-06, + 1.2338e-05, 1.2912e-05], + [-1.1349e-04, -5.8800e-05, -5.8651e-05, ..., -9.7454e-06, + -1.0319e-05, -1.2718e-05], + [ 3.9130e-05, 6.7838e-06, 8.2403e-06, ..., 3.8221e-06, + 1.0595e-05, 9.8199e-06], + ..., + [-1.5461e-04, -1.5020e-05, -2.1026e-05, ..., -1.6153e-05, + -4.4107e-05, -4.0203e-05], + [ 4.2558e-05, 7.0110e-06, 8.5458e-06, ..., 4.8019e-06, + 1.1727e-05, 1.0826e-05], + [ 5.9634e-05, 1.9446e-05, 2.0742e-05, ..., 6.4783e-06, + 1.2368e-05, 1.2167e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0458, -0.0104, 0.0234, 0.0088, -0.0427, -0.0127, -0.0083], + device='cuda:0'), grad: tensor([ 1.7858e-04, -1.4818e-04, 1.0729e-04, 6.5923e-05, -4.4537e-04, + 1.1533e-04, 1.2648e-04], device='cuda:0') +588 +0.0012346426699819469 +changing lr +epoch 54, time 790.30, cls_loss 0.0012 cls_loss_mapping 0.0074 cls_loss_causal 0.4004 re_mapping 0.0072 re_causal 0.0161 /// teacc 91.46 lr 0.00109084 +Epoch 56, weight, value: tensor([[ 0.0806, 0.0741, 0.0700, ..., 0.0182, 0.0546, 0.0496], + [ 0.0316, 0.0217, 0.0073, ..., -0.0154, -0.0137, -0.0117], + [-0.0220, -0.0291, 0.0008, ..., 0.0023, -0.0292, -0.0126], + ..., + [-0.1187, -0.0933, -0.0756, ..., 0.0328, 0.0357, 0.0233], + [ 0.0872, 0.0669, 0.0649, ..., -0.0421, -0.0535, -0.0475], + [ 0.0006, 0.0140, -0.0122, ..., -0.0436, -0.0221, -0.0304]], + device='cuda:0'), grad: tensor([[ 1.4591e-04, 3.2336e-05, 3.3647e-05, ..., 1.9282e-05, + 3.5226e-05, 3.2991e-05], + [-2.3139e-04, -5.2631e-05, -5.8651e-05, ..., -1.4164e-05, + -5.5045e-05, -5.0008e-05], + [ 1.2636e-04, 2.8759e-05, 2.8357e-05, ..., 2.2590e-05, + 2.8297e-05, 2.6703e-05], + ..., + [ 5.3108e-05, 1.2226e-05, 1.3150e-05, ..., 4.5523e-06, + 1.2390e-05, 1.1265e-05], + [ 3.8326e-05, 8.5011e-06, 9.1940e-06, ..., 3.8184e-06, + 9.3132e-06, 8.6427e-06], + [ 6.7830e-05, 1.5043e-05, 1.6212e-05, ..., 7.2420e-06, + 1.6958e-05, 1.5751e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0456, -0.0096, 0.0236, 0.0082, -0.0428, -0.0130, -0.0080], + device='cuda:0'), grad: tensor([ 0.0004, -0.0008, 0.0003, -0.0005, 0.0002, 0.0001, 0.0002], + device='cuda:0') +588 +0.0010908425876598518 +changing lr +epoch 55, time 791.04, cls_loss 0.0013 cls_loss_mapping 0.0064 cls_loss_causal 0.4006 re_mapping 0.0070 re_causal 0.0157 /// teacc 92.96 lr 0.00095492 +Epoch 57, weight, value: tensor([[ 0.0808, 0.0742, 0.0701, ..., 0.0182, 0.0545, 0.0495], + [ 0.0315, 0.0217, 0.0073, ..., -0.0154, -0.0136, -0.0117], + [-0.0221, -0.0291, 0.0007, ..., 0.0023, -0.0292, -0.0127], + ..., + [-0.1184, -0.0931, -0.0755, ..., 0.0327, 0.0357, 0.0233], + [ 0.0870, 0.0668, 0.0647, ..., -0.0420, -0.0534, -0.0474], + [ 0.0003, 0.0139, -0.0123, ..., -0.0435, -0.0222, -0.0304]], + device='cuda:0'), grad: tensor([[-1.2696e-05, -2.6584e-05, -2.5883e-05, ..., 2.2784e-05, + 1.7941e-05, 1.9476e-05], + [ 3.1024e-05, -3.7383e-06, -1.1204e-06, ..., 2.1696e-05, + 2.3231e-05, 2.0608e-05], + [ 7.4089e-05, 8.7768e-06, 1.0565e-05, ..., 2.9594e-05, + 3.3528e-05, 3.1263e-05], + ..., + [ 1.1605e-04, 3.8505e-05, 4.2140e-05, ..., 2.1189e-05, + 3.2544e-05, 2.7776e-05], + [ 2.7940e-05, 4.8168e-06, 5.5768e-06, ..., 9.3654e-06, + 1.1064e-05, 1.0468e-05], + [ 4.4197e-05, 7.8306e-06, 8.4937e-06, ..., 1.2912e-05, + 1.5251e-05, 1.4573e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0457, -0.0097, 0.0232, 0.0087, -0.0423, -0.0132, -0.0086], + device='cuda:0'), grad: tensor([ 2.4304e-05, 1.1349e-04, 2.1267e-04, -8.1396e-04, 2.6894e-04, + 7.5221e-05, 1.1951e-04], device='cuda:0') +588 +0.000954915028125264 +changing lr +epoch 56, time 792.00, cls_loss 0.0014 cls_loss_mapping 0.0058 cls_loss_causal 0.3947 re_mapping 0.0070 re_causal 0.0153 /// teacc 93.97 lr 0.00082713 +Epoch 58, weight, value: tensor([[ 0.0812, 0.0743, 0.0703, ..., 0.0182, 0.0545, 0.0495], + [ 0.0314, 0.0216, 0.0072, ..., -0.0153, -0.0136, -0.0117], + [-0.0222, -0.0291, 0.0006, ..., 0.0023, -0.0291, -0.0126], + ..., + [-0.1183, -0.0930, -0.0755, ..., 0.0326, 0.0356, 0.0232], + [ 0.0867, 0.0667, 0.0646, ..., -0.0420, -0.0533, -0.0474], + [ 0.0002, 0.0138, -0.0123, ..., -0.0435, -0.0221, -0.0303]], + device='cuda:0'), grad: tensor([[ 2.2018e-04, 4.3094e-05, 4.2945e-05, ..., 3.9011e-05, + 6.7592e-05, 6.3181e-05], + [ 8.5890e-05, 1.3866e-05, 1.4685e-05, ..., 2.1249e-05, + 3.2306e-05, 3.1620e-05], + [-4.9263e-05, -8.6352e-06, -7.9125e-06, ..., -1.0364e-05, + -1.6361e-05, -1.6630e-05], + ..., + [-2.1291e-04, -1.9759e-05, -2.5615e-05, ..., -6.2048e-05, + -1.0628e-04, -9.5904e-05], + [ 2.1315e-04, 7.1287e-05, 6.4909e-05, ..., 1.3031e-05, + 1.5482e-05, 2.2084e-05], + [-3.4785e-04, -1.2279e-04, -1.1086e-04, ..., -1.4894e-05, + -1.4380e-05, -2.6733e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0465, -0.0097, 0.0230, 0.0087, -0.0424, -0.0135, -0.0087], + device='cuda:0'), grad: tensor([ 0.0006, 0.0002, -0.0001, 0.0002, -0.0007, 0.0004, -0.0007], + device='cuda:0') +588 +0.0008271337313934874 +changing lr +epoch 57, time 785.47, cls_loss 0.0014 cls_loss_mapping 0.0066 cls_loss_causal 0.3959 re_mapping 0.0070 re_causal 0.0155 /// teacc 93.72 lr 0.00070776 +Epoch 59, weight, value: tensor([[ 8.1379e-02, 7.4439e-02, 7.0417e-02, ..., 1.8152e-02, + 5.4398e-02, 4.9417e-02], + [ 3.1237e-02, 2.1533e-02, 7.2049e-03, ..., -1.5339e-02, + -1.3612e-02, -1.1715e-02], + [-2.2203e-02, -2.9107e-02, 5.2069e-04, ..., 2.2342e-03, + -2.9068e-02, -1.2622e-02], + ..., + [-1.1797e-01, -9.2928e-02, -7.5440e-02, ..., 3.2604e-02, + 3.5572e-02, 2.3227e-02], + [ 8.6427e-02, 6.6554e-02, 6.4479e-02, ..., -4.1890e-02, + -5.3248e-02, -4.7302e-02], + [ 3.9185e-05, 1.3749e-02, -1.2354e-02, ..., -4.3392e-02, + -2.2147e-02, -3.0321e-02]], device='cuda:0'), grad: tensor([[-1.1235e-04, -3.5435e-05, -3.2008e-05, ..., -6.7614e-06, + -2.2739e-05, -2.2516e-05], + [ 2.6822e-07, -1.0356e-05, -1.0908e-05, ..., 5.3830e-06, + 3.4962e-06, 6.2659e-06], + [ 2.7850e-05, 5.6587e-06, 7.6592e-06, ..., 5.3123e-06, + 9.2909e-06, 9.0823e-06], + ..., + [ 8.6054e-07, 1.7315e-05, 9.5293e-06, ..., -1.4305e-05, + -1.3761e-05, -1.5251e-05], + [ 2.3812e-05, 5.7258e-06, 6.6236e-06, ..., 3.6899e-06, + 7.1824e-06, 6.9141e-06], + [ 2.9907e-05, 8.1360e-06, 9.1270e-06, ..., 3.8054e-06, + 8.0764e-06, 7.7859e-06]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0468, -0.0098, 0.0229, 0.0086, -0.0417, -0.0137, -0.0092], + device='cuda:0'), grad: tensor([-2.5082e-04, 2.2411e-05, 7.2360e-05, 6.9439e-05, -4.3809e-05, + 5.9098e-05, 7.1585e-05], device='cuda:0') +588 +0.00070775603199067 +changing lr +epoch 58, time 790.13, cls_loss 0.0015 cls_loss_mapping 0.0079 cls_loss_causal 0.3930 re_mapping 0.0070 re_causal 0.0153 /// teacc 92.71 lr 0.00059702 +Epoch 60, weight, value: tensor([[ 0.0813, 0.0744, 0.0704, ..., 0.0181, 0.0543, 0.0493], + [ 0.0312, 0.0215, 0.0072, ..., -0.0153, -0.0136, -0.0117], + [-0.0220, -0.0291, 0.0005, ..., 0.0022, -0.0290, -0.0126], + ..., + [-0.1179, -0.0928, -0.0754, ..., 0.0325, 0.0355, 0.0232], + [ 0.0865, 0.0665, 0.0644, ..., -0.0418, -0.0531, -0.0472], + [-0.0002, 0.0137, -0.0124, ..., -0.0433, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[-1.3614e-04, -4.7654e-05, -4.9800e-05, ..., -9.0376e-06, + -2.8431e-05, -3.0428e-05], + [ 2.2963e-05, -1.4622e-07, 5.2713e-07, ..., 7.8231e-06, + 9.9093e-06, 9.8273e-06], + [ 2.0003e-04, 4.9949e-05, 5.4389e-05, ..., 3.1143e-05, + 4.8310e-05, 4.9323e-05], + ..., + [-3.0428e-05, 3.3360e-06, 2.5891e-06, ..., -6.1467e-06, + -7.1004e-06, -7.6443e-06], + [ 1.8820e-05, 4.6901e-06, 5.1148e-06, ..., 3.0342e-06, + 4.5076e-06, 4.5747e-06], + [ 3.3736e-05, 9.0003e-06, 9.7901e-06, ..., 5.9679e-06, + 8.0764e-06, 8.0466e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0465, -0.0098, 0.0233, 0.0084, -0.0418, -0.0132, -0.0096], + device='cuda:0'), grad: tensor([-2.8539e-04, 7.4804e-05, 4.8685e-04, -2.9206e-04, -1.1086e-04, + 4.5896e-05, 8.0407e-05], device='cuda:0') +588 +0.0005970223407163104 +changing lr +epoch 59, time 787.39, cls_loss 0.0013 cls_loss_mapping 0.0065 cls_loss_causal 0.4018 re_mapping 0.0070 re_causal 0.0152 /// teacc 91.46 lr 0.00049516 +Epoch 61, weight, value: tensor([[ 0.0813, 0.0745, 0.0705, ..., 0.0181, 0.0542, 0.0493], + [ 0.0313, 0.0215, 0.0072, ..., -0.0153, -0.0135, -0.0116], + [-0.0220, -0.0291, 0.0005, ..., 0.0022, -0.0289, -0.0125], + ..., + [-0.1179, -0.0928, -0.0754, ..., 0.0324, 0.0354, 0.0231], + [ 0.0865, 0.0664, 0.0644, ..., -0.0417, -0.0530, -0.0471], + [-0.0003, 0.0136, -0.0124, ..., -0.0433, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[ 6.6161e-05, 2.5108e-05, 2.2292e-05, ..., -2.0247e-06, + 2.7250e-06, 2.9895e-06], + [ 1.3210e-05, 1.6913e-05, 1.7300e-05, ..., -6.8434e-06, + -1.0744e-05, -8.6725e-06], + [ 5.3674e-05, 3.2604e-05, 3.2365e-05, ..., 5.1893e-06, + 6.1728e-06, 6.3814e-06], + ..., + [-1.4462e-05, 5.9456e-06, 9.0003e-06, ..., 2.1271e-06, + -2.9467e-06, -3.6284e-06], + [-7.7200e-04, -3.9530e-04, -4.0317e-04, ..., -4.7147e-05, + -6.9857e-05, -7.2420e-05], + [ 4.9162e-04, 2.4366e-04, 2.4891e-04, ..., 3.4362e-05, + 5.1647e-05, 5.2601e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0466, -0.0095, 0.0235, 0.0083, -0.0421, -0.0130, -0.0101], + device='cuda:0'), grad: tensor([ 1.3614e-04, -1.2711e-05, 6.5029e-05, 2.9325e-04, -7.4446e-05, + -1.1940e-03, 7.8773e-04], device='cuda:0') +588 +0.0004951556604879052 +changing lr +epoch 60, time 786.59, cls_loss 0.0013 cls_loss_mapping 0.0053 cls_loss_causal 0.4019 re_mapping 0.0070 re_causal 0.0155 /// teacc 93.22 lr 0.00040236 +Epoch 62, weight, value: tensor([[ 0.0815, 0.0745, 0.0705, ..., 0.0181, 0.0542, 0.0493], + [ 0.0313, 0.0215, 0.0073, ..., -0.0153, -0.0135, -0.0116], + [-0.0220, -0.0291, 0.0004, ..., 0.0022, -0.0289, -0.0125], + ..., + [-0.1179, -0.0927, -0.0754, ..., 0.0324, 0.0353, 0.0230], + [ 0.0863, 0.0664, 0.0643, ..., -0.0417, -0.0530, -0.0471], + [-0.0004, 0.0135, -0.0125, ..., -0.0433, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[-1.4246e-04, -1.9640e-05, -3.0264e-05, ..., -6.9857e-05, + -1.2910e-04, -1.1581e-04], + [-1.0908e-05, -6.3218e-06, -5.1782e-06, ..., 2.3730e-06, + 3.9525e-06, 3.3639e-06], + [ 2.5630e-05, 3.5446e-06, 5.0627e-06, ..., 1.0781e-05, + 2.0429e-05, 1.8463e-05], + ..., + [ 3.6359e-05, 7.1079e-06, 9.3430e-06, ..., 1.7747e-05, + 3.2604e-05, 2.9057e-05], + [ 1.3679e-05, 2.3674e-06, 3.1535e-06, ..., 5.2191e-06, + 9.7379e-06, 8.8066e-06], + [ 2.7865e-05, 5.3830e-06, 6.8247e-06, ..., 1.0662e-05, + 1.9848e-05, 1.7881e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0468, -0.0092, 0.0235, 0.0086, -0.0424, -0.0132, -0.0102], + device='cuda:0'), grad: tensor([-5.7840e-04, 9.1270e-07, 9.4593e-05, 1.9586e-04, 1.4210e-04, + 4.8250e-05, 9.6798e-05], device='cuda:0') +588 +0.00040236113724274745 +changing lr +---------------------saving model at epoch 61---------------------------------------------------- +epoch 61, time 798.02, cls_loss 0.0014 cls_loss_mapping 0.0060 cls_loss_causal 0.3726 re_mapping 0.0069 re_causal 0.0147 /// teacc 95.73 lr 0.00031883 +Epoch 63, weight, value: tensor([[ 0.0815, 0.0745, 0.0706, ..., 0.0181, 0.0542, 0.0492], + [ 0.0313, 0.0215, 0.0073, ..., -0.0153, -0.0135, -0.0116], + [-0.0219, -0.0290, 0.0004, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1179, -0.0927, -0.0753, ..., 0.0323, 0.0353, 0.0230], + [ 0.0863, 0.0663, 0.0642, ..., -0.0416, -0.0529, -0.0470], + [-0.0005, 0.0135, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[-4.3929e-05, -3.4690e-05, -4.0025e-05, ..., -3.5614e-06, + -5.4650e-06, -4.4927e-06], + [ 5.9128e-05, 2.5079e-05, 2.8685e-05, ..., 1.2778e-05, + 1.7628e-05, 1.6913e-05], + [-8.0585e-05, -1.8463e-05, -1.8775e-05, ..., -5.4836e-05, + -6.4671e-05, -6.2823e-05], + ..., + [-6.9857e-05, -3.1013e-06, -4.7386e-06, ..., 1.1753e-06, + -6.4522e-06, -6.5900e-06], + [ 2.0623e-05, 6.4410e-06, 7.3016e-06, ..., 4.1425e-06, + 6.3218e-06, 6.1989e-06], + [ 3.7521e-05, 1.0610e-05, 1.1824e-05, ..., 1.3597e-05, + 1.7479e-05, 1.6838e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0468, -0.0093, 0.0235, 0.0086, -0.0424, -0.0130, -0.0104], + device='cuda:0'), grad: tensor([-1.0625e-05, 1.1456e-04, -2.4128e-04, 2.0921e-04, -2.1255e-04, + 4.9680e-05, 9.0718e-05], device='cuda:0') +588 +0.00031882564680131423 +changing lr +epoch 62, time 791.56, cls_loss 0.0014 cls_loss_mapping 0.0056 cls_loss_causal 0.3860 re_mapping 0.0069 re_causal 0.0145 /// teacc 94.97 lr 0.00024472 +Epoch 64, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0541, 0.0492], + [ 0.0313, 0.0215, 0.0073, ..., -0.0153, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0004, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1178, -0.0926, -0.0753, ..., 0.0323, 0.0353, 0.0230], + [ 0.0862, 0.0663, 0.0642, ..., -0.0416, -0.0529, -0.0470], + [-0.0006, 0.0135, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[ 4.8518e-04, 1.4091e-04, 1.4400e-04, ..., 5.1767e-05, + 1.0324e-04, 1.0949e-04], + [-1.6241e-03, -4.9543e-04, -4.9257e-04, ..., -1.2589e-04, + -2.7442e-04, -3.0184e-04], + [-9.9719e-05, -6.7130e-06, -1.5736e-05, ..., -2.8729e-05, + -4.7058e-05, -3.9220e-05], + ..., + [ 1.5032e-04, 4.4167e-05, 4.5151e-05, ..., 1.7181e-05, + 3.1292e-05, 3.0518e-05], + [ 7.5758e-05, 2.1458e-05, 2.2143e-05, ..., 7.0743e-06, + 1.4804e-05, 1.5438e-05], + [ 2.1255e-04, 7.0274e-05, 7.1526e-05, ..., 1.7896e-05, + 3.7909e-05, 3.9816e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0468, -0.0092, 0.0235, 0.0086, -0.0423, -0.0131, -0.0105], + device='cuda:0'), grad: tensor([ 0.0011, -0.0037, -0.0003, 0.0019, 0.0003, 0.0002, 0.0005], + device='cuda:0') +588 +0.0002447174185242325 +changing lr +epoch 63, time 792.66, cls_loss 0.0014 cls_loss_mapping 0.0055 cls_loss_causal 0.3893 re_mapping 0.0069 re_causal 0.0150 /// teacc 92.96 lr 0.00018019 +Epoch 65, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0541, 0.0491], + [ 0.0312, 0.0214, 0.0073, ..., -0.0153, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0003, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1177, -0.0926, -0.0753, ..., 0.0323, 0.0352, 0.0230], + [ 0.0862, 0.0662, 0.0642, ..., -0.0416, -0.0529, -0.0470], + [-0.0006, 0.0135, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[ 8.9228e-05, 1.2606e-05, 1.9163e-05, ..., 2.3171e-05, + 3.9995e-05, 3.6895e-05], + [ 2.4647e-05, 4.6715e-06, 5.2266e-06, ..., 1.1235e-05, + 1.3925e-05, 1.3247e-05], + [-1.8505e-06, -1.0058e-06, -1.2983e-06, ..., 3.5521e-06, + 4.2692e-06, 3.7104e-06], + ..., + [-1.0407e-04, -1.2234e-05, -2.0415e-05, ..., -3.8534e-05, + -6.0707e-05, -5.6356e-05], + [ 2.6360e-05, 9.4771e-06, 9.5293e-06, ..., 3.1702e-06, + 6.7726e-06, 6.0052e-06], + [-2.6450e-05, -1.1444e-05, -1.0967e-05, ..., 1.0896e-07, + -3.2596e-06, -2.4661e-06]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0467, -0.0093, 0.0235, 0.0087, -0.0422, -0.0130, -0.0106], + device='cuda:0'), grad: tensor([ 2.6202e-04, 7.3969e-05, 4.8243e-06, -1.3702e-05, -3.3689e-04, + 5.6893e-05, -4.6700e-05], device='cuda:0') +588 +0.0001801856965207339 +changing lr +epoch 64, time 789.94, cls_loss 0.0014 cls_loss_mapping 0.0060 cls_loss_causal 0.3751 re_mapping 0.0069 re_causal 0.0144 /// teacc 94.47 lr 0.00012536 +Epoch 66, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0541, 0.0491], + [ 0.0312, 0.0214, 0.0072, ..., -0.0152, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0003, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1177, -0.0926, -0.0753, ..., 0.0323, 0.0352, 0.0230], + [ 0.0862, 0.0662, 0.0642, ..., -0.0415, -0.0528, -0.0469], + [-0.0006, 0.0135, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[ 2.0289e-04, 4.7714e-05, 4.6968e-05, ..., 2.2978e-05, + 5.6028e-05, 5.4181e-05], + [-1.6952e-04, -6.3956e-05, -7.1347e-05, ..., -3.0875e-05, + -5.2601e-05, -5.0157e-05], + [-2.3925e-04, -3.6091e-05, -2.8566e-05, ..., -1.7762e-05, + -6.2704e-05, -6.1035e-05], + ..., + [ 5.8234e-05, 1.5043e-05, 1.5035e-05, ..., 6.8173e-06, + 1.5691e-05, 1.5207e-05], + [ 3.3915e-05, 7.8082e-06, 7.6219e-06, ..., 3.8110e-06, + 9.4622e-06, 9.1419e-06], + [ 4.7415e-05, 1.1221e-05, 1.1034e-05, ..., 5.9307e-06, + 1.4342e-05, 1.3761e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0466, -0.0093, 0.0235, 0.0087, -0.0421, -0.0130, -0.0106], + device='cuda:0'), grad: tensor([ 5.1594e-04, -3.5167e-04, -6.7663e-04, 1.6046e-04, 1.4317e-04, + 8.6904e-05, 1.2153e-04], device='cuda:0') +588 +0.000125360439090882 +changing lr +epoch 65, time 792.08, cls_loss 0.0013 cls_loss_mapping 0.0047 cls_loss_causal 0.4086 re_mapping 0.0070 re_causal 0.0149 /// teacc 94.47 lr 0.00008035 +Epoch 67, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0541, 0.0491], + [ 0.0312, 0.0214, 0.0072, ..., -0.0152, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0003, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1176, -0.0926, -0.0752, ..., 0.0323, 0.0352, 0.0230], + [ 0.0861, 0.0662, 0.0641, ..., -0.0415, -0.0528, -0.0469], + [-0.0006, 0.0134, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[ 8.9169e-05, 1.9014e-05, 2.5526e-05, ..., 8.1286e-06, + 2.1189e-05, 2.0862e-05], + [-7.4327e-05, -1.5467e-05, -2.2471e-05, ..., -1.4305e-06, + -1.2703e-05, -1.2293e-05], + [ 6.7241e-06, 6.5006e-07, 7.9162e-07, ..., -1.1548e-06, + -6.4261e-07, -3.6508e-07], + ..., + [-1.8179e-05, 2.3656e-07, -1.3821e-06, ..., -5.5097e-06, + -9.0227e-06, -9.2089e-06], + [ 1.1787e-05, 3.2131e-06, 3.6694e-06, ..., 1.5590e-06, + 3.1739e-06, 3.1292e-06], + [-2.1517e-05, -9.9093e-06, -8.9705e-06, ..., -1.8366e-06, + -3.9525e-06, -4.0568e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0467, -0.0093, 0.0235, 0.0086, -0.0421, -0.0130, -0.0106], + device='cuda:0'), grad: tensor([ 2.2435e-04, -1.8561e-04, 1.4521e-05, 1.4886e-05, -6.1214e-05, + 2.7463e-05, -3.4809e-05], device='cuda:0') +588 +8.03520570068517e-05 +changing lr +epoch 66, time 791.60, cls_loss 0.0012 cls_loss_mapping 0.0048 cls_loss_causal 0.3824 re_mapping 0.0069 re_causal 0.0144 /// teacc 92.71 lr 0.00004525 +Epoch 68, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0540, 0.0491], + [ 0.0312, 0.0214, 0.0072, ..., -0.0152, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0003, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1176, -0.0926, -0.0752, ..., 0.0323, 0.0352, 0.0230], + [ 0.0861, 0.0662, 0.0641, ..., -0.0415, -0.0528, -0.0469], + [-0.0006, 0.0134, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[-3.4499e-04, -1.0622e-04, -1.1998e-04, ..., -5.3972e-05, + -8.1837e-05, -8.7082e-05], + [ 3.6597e-05, 1.3076e-05, 1.3515e-05, ..., -1.1846e-05, + -9.7007e-06, -7.8529e-06], + [ 2.3949e-04, 6.9976e-05, 7.8917e-05, ..., 4.0054e-05, + 6.0946e-05, 6.2346e-05], + ..., + [ 8.7678e-05, 2.6032e-05, 2.8893e-05, ..., 1.6496e-05, + 2.5570e-05, 2.4974e-05], + [ 6.9022e-05, 2.1338e-05, 2.3648e-05, ..., 9.4622e-06, + 1.6078e-05, 1.6108e-05], + [-8.6308e-05, -3.0845e-05, -3.1888e-05, ..., 2.0768e-06, + -9.3579e-06, -7.0520e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0467, -0.0093, 0.0235, 0.0086, -0.0420, -0.0130, -0.0106], + device='cuda:0'), grad: tensor([-7.2241e-04, 5.9903e-05, 5.2357e-04, -2.9445e-05, 1.9670e-04, + 1.4806e-04, -1.7786e-04], device='cuda:0') +588 +4.5251191160326525e-05 +changing lr +epoch 67, time 790.33, cls_loss 0.0011 cls_loss_mapping 0.0054 cls_loss_causal 0.4041 re_mapping 0.0069 re_causal 0.0150 /// teacc 94.72 lr 0.00002013 +Epoch 69, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0540, 0.0491], + [ 0.0312, 0.0214, 0.0072, ..., -0.0152, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0003, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1176, -0.0925, -0.0752, ..., 0.0323, 0.0352, 0.0230], + [ 0.0861, 0.0662, 0.0641, ..., -0.0415, -0.0528, -0.0469], + [-0.0006, 0.0134, -0.0125, ..., -0.0432, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[ 3.5071e-04, 1.1247e-04, 1.2410e-04, ..., 2.8133e-05, + 6.1333e-05, 5.7489e-05], + [-6.8474e-04, -2.2268e-04, -2.4223e-04, ..., -5.7817e-05, + -1.1975e-04, -1.1617e-04], + [-9.4235e-05, -2.9683e-05, -3.3438e-05, ..., -1.4827e-05, + -2.4989e-05, -2.4036e-05], + ..., + [ 1.2064e-04, 3.9995e-05, 4.2647e-05, ..., 1.1832e-05, + 2.3708e-05, 2.3514e-05], + [ 5.0396e-05, 1.6794e-05, 1.8194e-05, ..., 4.7646e-06, + 9.4473e-06, 9.3356e-06], + [ 1.3649e-04, 4.4614e-05, 4.8399e-05, ..., 1.2234e-05, + 2.4438e-05, 2.3961e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0467, -0.0093, 0.0234, 0.0086, -0.0420, -0.0131, -0.0106], + device='cuda:0'), grad: tensor([ 0.0008, -0.0015, -0.0002, 0.0003, 0.0003, 0.0001, 0.0003], + device='cuda:0') +588 +2.0128530023804673e-05 +changing lr +epoch 68, time 787.75, cls_loss 0.0012 cls_loss_mapping 0.0053 cls_loss_causal 0.3859 re_mapping 0.0069 re_causal 0.0147 /// teacc 94.47 lr 0.00000503 +Epoch 70, weight, value: tensor([[ 0.0815, 0.0746, 0.0706, ..., 0.0180, 0.0540, 0.0491], + [ 0.0312, 0.0214, 0.0072, ..., -0.0152, -0.0135, -0.0116], + [-0.0220, -0.0290, 0.0003, ..., 0.0022, -0.0288, -0.0125], + ..., + [-0.1176, -0.0925, -0.0752, ..., 0.0323, 0.0352, 0.0230], + [ 0.0861, 0.0662, 0.0641, ..., -0.0415, -0.0528, -0.0469], + [-0.0006, 0.0134, -0.0125, ..., -0.0431, -0.0222, -0.0303]], + device='cuda:0'), grad: tensor([[-1.6153e-04, -5.3585e-05, -5.9724e-05, ..., -3.2395e-05, + -5.6535e-05, -5.7220e-05], + [ 8.6248e-05, 2.2650e-05, 2.6375e-05, ..., 2.1592e-05, + 3.2693e-05, 3.2932e-05], + [ 4.7624e-05, 1.4886e-05, 1.6645e-05, ..., 1.0297e-05, + 1.6779e-05, 1.7002e-05], + ..., + [ 4.1991e-05, 1.4745e-05, 1.6540e-05, ..., 7.4059e-06, + 1.2279e-05, 1.2465e-05], + [-8.6069e-05, -4.3333e-05, -4.7565e-05, ..., -1.9930e-06, + -4.6790e-06, -6.3106e-06], + [ 7.8261e-05, 3.6418e-05, 4.0323e-05, ..., 4.9211e-06, + 8.8066e-06, 9.9540e-06]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0467, -0.0093, 0.0234, 0.0086, -0.0420, -0.0131, -0.0106], + device='cuda:0'), grad: tensor([-3.4833e-04, 1.9848e-04, 1.0222e-04, -4.1157e-05, 8.4817e-05, + -1.1569e-04, 1.1945e-04], device='cuda:0') +588 +5.034667293427056e-06 +changing lr +epoch 69, time 799.21, cls_loss 0.0015 cls_loss_mapping 0.0062 cls_loss_causal 0.3843 re_mapping 0.0069 re_causal 0.0145 /// teacc 93.47 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.338254 43.212891 56.186007 46.467066 48.621988 + sketch art_painting cartoon photo Avg +do 99.28735 46.630859 60.025597 52.035928 52.897462 diff --git a/Meta-causal/code-withStyleAttack/65654.error b/Meta-causal/code-withStyleAttack/65654.error new file mode 100644 index 0000000000000000000000000000000000000000..c8925473af6329b0a06ae76e97c6ab027644bc98 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/65654.error @@ -0,0 +1,3 @@ +slurmstepd: error: *** STEP 65654.0 ON gcpl4-eu-1 CANCELLED AT 2024-07-19T07:11:27 *** +slurmstepd: error: *** JOB 65654 ON gcpl4-eu-1 CANCELLED AT 2024-07-19T07:11:27 *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-withStyleAttack/65654.log b/Meta-causal/code-withStyleAttack/65654.log new file mode 100644 index 0000000000000000000000000000000000000000..47375e47e56437d95c5b818b3dd56955e45a39c6 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/65654.log @@ -0,0 +1,7457 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0271, 0.0278, 0.0225, ..., -0.0296, -0.0126, 0.0133], + [ 0.0281, -0.0014, 0.0019, ..., 0.0049, -0.0001, -0.0167], + [ 0.0096, -0.0095, -0.0208, ..., -0.0112, 0.0086, 0.0155], + ..., + [ 0.0241, 0.0234, -0.0310, ..., -0.0023, 0.0109, -0.0108], + [-0.0104, 0.0021, 0.0085, ..., -0.0012, -0.0071, -0.0235], + [ 0.0238, -0.0081, 0.0041, ..., -0.0056, 0.0007, 0.0226]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0091, 0.0009, 0.0029, 0.0177, 0.0243, 0.0294, -0.0206, 0.0056, + -0.0307, -0.0182], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 223.22, cls_loss 1.6736 cls_loss_mapping 2.0181 cls_loss_causal 2.2459 re_mapping 0.0816 re_causal 0.0818 /// teacc 73.46 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0191, 0.0323, 0.0264, ..., -0.0359, -0.0128, 0.0174], + [ 0.0360, -0.0041, 0.0007, ..., 0.0067, -0.0003, -0.0264], + [ 0.0048, -0.0081, -0.0186, ..., -0.0090, 0.0080, 0.0122], + ..., + [ 0.0229, 0.0276, -0.0377, ..., -0.0113, 0.0093, -0.0189], + [-0.0084, -0.0019, 0.0107, ..., 0.0022, -0.0053, -0.0228], + [ 0.0194, -0.0109, -0.0004, ..., -0.0055, -0.0014, 0.0232]], + device='cuda:0'), grad: tensor([[ 2.0035e-02, -1.3054e-02, -1.5045e-02, ..., 7.2479e-03, + 1.3113e-04, -1.1559e-02], + [-6.2988e-02, 2.7943e-03, -1.8673e-03, ..., -4.6967e-02, + 8.8736e-06, -1.4580e-02], + [ 2.7679e-02, 3.3607e-03, 1.9638e-02, ..., 2.5070e-02, + 1.5184e-05, 1.3123e-02], + ..., + [-2.3865e-02, -1.8097e-02, 7.0152e-03, ..., -5.5695e-03, + 3.0845e-05, 2.3499e-03], + [-3.2921e-03, 7.3509e-03, -9.6130e-04, ..., -2.1133e-03, + 1.0669e-04, 1.1459e-02], + [ 5.2582e-02, 1.3435e-02, 3.2013e-02, ..., 7.6172e-02, + 9.8348e-05, 3.1250e-02]], device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0069, 0.0036, 0.0026, 0.0171, 0.0236, 0.0295, -0.0215, 0.0064, + -0.0314, -0.0176], device='cuda:0'), grad: tensor([ 0.0031, -0.0215, 0.0242, 0.0185, -0.0174, -0.0472, -0.0401, -0.0387, + 0.0231, 0.0958], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 220.40, cls_loss 0.5839 cls_loss_mapping 0.8899 cls_loss_causal 1.9196 re_mapping 0.2132 re_causal 0.2487 /// teacc 88.40 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 1.5019e-02, 3.5892e-02, 2.8696e-02, ..., -3.9882e-02, + -1.5826e-02, 1.8511e-02], + [ 3.8934e-02, -5.4322e-03, -1.9550e-03, ..., 8.0060e-03, + -4.0296e-04, -2.8752e-02], + [ 5.3168e-05, -6.9198e-03, -1.8001e-02, ..., -1.0121e-02, + 6.0515e-04, 9.4711e-03], + ..., + [ 2.0476e-02, 3.2725e-02, -4.1894e-02, ..., -1.5440e-02, + 8.8445e-03, -2.1740e-02], + [-7.3172e-03, -5.6156e-03, 1.3275e-02, ..., 5.6482e-03, + -1.0707e-02, -2.2726e-02], + [ 1.8082e-02, -1.1282e-02, -1.0400e-03, ..., -9.5011e-03, + -4.7546e-03, 2.3836e-02]], device='cuda:0'), grad: tensor([[-0.0201, -0.0018, -0.0141, ..., -0.0006, -0.0001, -0.0125], + [-0.0225, 0.0020, 0.0056, ..., -0.0228, -0.0046, -0.0055], + [-0.0170, 0.0011, -0.0056, ..., -0.0105, -0.0029, 0.0010], + ..., + [ 0.0154, -0.0052, 0.0065, ..., 0.0234, 0.0011, 0.0092], + [ 0.0093, -0.0120, -0.0126, ..., -0.0094, 0.0024, -0.0142], + [-0.0535, 0.0031, -0.0219, ..., -0.0405, -0.0099, -0.0302]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0060, 0.0037, 0.0024, 0.0170, 0.0235, 0.0311, -0.0226, 0.0060, + -0.0307, -0.0176], device='cuda:0'), grad: tensor([-0.0033, -0.0064, -0.0138, 0.0236, 0.0052, 0.0347, 0.0149, 0.0178, + -0.0105, -0.0623], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 221.40, cls_loss 0.3490 cls_loss_mapping 0.5111 cls_loss_causal 1.6934 re_mapping 0.1598 re_causal 0.2406 /// teacc 93.28 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0128, 0.0385, 0.0302, ..., -0.0418, -0.0186, 0.0193], + [ 0.0411, -0.0065, -0.0034, ..., 0.0094, 0.0028, -0.0307], + [-0.0016, -0.0057, -0.0179, ..., -0.0114, -0.0039, 0.0077], + ..., + [ 0.0176, 0.0352, -0.0448, ..., -0.0192, 0.0066, -0.0247], + [-0.0074, -0.0083, 0.0139, ..., 0.0072, -0.0168, -0.0230], + [ 0.0178, -0.0127, 0.0003, ..., -0.0107, -0.0075, 0.0257]], + device='cuda:0'), grad: tensor([[ 2.1667e-03, -1.6556e-03, -8.8406e-04, ..., 2.3785e-03, + -8.8811e-05, 1.1215e-03], + [-3.7903e-02, 9.2793e-04, -2.2621e-03, ..., -2.9175e-02, + -5.3635e-03, -1.1063e-03], + [ 1.2840e-02, 2.3148e-02, 3.6743e-02, ..., 8.3160e-03, + 1.0786e-03, 1.8982e-02], + ..., + [ 8.4782e-04, -2.5757e-02, -1.1475e-02, ..., 5.2986e-03, + -8.3566e-05, 2.4052e-03], + [ 2.3956e-03, 2.5101e-03, -2.2717e-03, ..., 9.8228e-05, + 4.2915e-03, -4.1389e-03], + [ 2.4929e-03, 7.8659e-03, -1.2722e-03, ..., 2.6436e-03, + 1.9798e-03, 3.5858e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0061, 0.0039, 0.0026, 0.0171, 0.0231, 0.0317, -0.0228, 0.0055, + -0.0311, -0.0173], device='cuda:0'), grad: tensor([ 0.0012, -0.0161, 0.0436, -0.0147, -0.0163, 0.0077, 0.0087, -0.0279, + 0.0087, 0.0052], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 226.06, cls_loss 0.2774 cls_loss_mapping 0.3965 cls_loss_causal 1.5685 re_mapping 0.1232 re_causal 0.2210 /// teacc 94.53 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 0.0110, 0.0411, 0.0317, ..., -0.0440, -0.0202, 0.0198], + [ 0.0435, -0.0077, -0.0047, ..., 0.0119, 0.0073, -0.0326], + [-0.0040, -0.0042, -0.0183, ..., -0.0126, -0.0063, 0.0055], + ..., + [ 0.0148, 0.0371, -0.0472, ..., -0.0220, 0.0069, -0.0272], + [-0.0070, -0.0103, 0.0147, ..., 0.0082, -0.0222, -0.0230], + [ 0.0170, -0.0142, 0.0010, ..., -0.0117, -0.0113, 0.0266]], + device='cuda:0'), grad: tensor([[ 0.0053, -0.0012, 0.0075, ..., 0.0070, 0.0017, 0.0064], + [ 0.0100, 0.0007, 0.0122, ..., 0.0086, 0.0056, 0.0073], + [ 0.0088, 0.0049, 0.0098, ..., 0.0096, 0.0011, 0.0059], + ..., + [ 0.0021, -0.0108, 0.0007, ..., 0.0030, 0.0003, -0.0003], + [-0.0210, -0.0032, -0.0249, ..., -0.0383, -0.0084, -0.0205], + [-0.0124, 0.0045, -0.0043, ..., -0.0104, 0.0005, 0.0046]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0063, 0.0044, 0.0029, 0.0172, 0.0232, 0.0316, -0.0234, 0.0052, + -0.0313, -0.0174], device='cuda:0'), grad: tensor([ 0.0081, 0.0140, 0.0146, 0.0090, 0.0138, -0.0153, 0.0077, -0.0058, + -0.0338, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 220.46, cls_loss 0.1985 cls_loss_mapping 0.2859 cls_loss_causal 1.4268 re_mapping 0.1054 re_causal 0.2123 /// teacc 95.74 lr 0.00010000 +Epoch 6, weight, value: tensor([[ 0.0102, 0.0432, 0.0323, ..., -0.0454, -0.0223, 0.0199], + [ 0.0447, -0.0086, -0.0058, ..., 0.0129, 0.0089, -0.0346], + [-0.0060, -0.0034, -0.0182, ..., -0.0142, -0.0080, 0.0036], + ..., + [ 0.0133, 0.0393, -0.0497, ..., -0.0241, 0.0056, -0.0296], + [-0.0066, -0.0124, 0.0154, ..., 0.0097, -0.0244, -0.0235], + [ 0.0167, -0.0154, 0.0021, ..., -0.0120, -0.0112, 0.0287]], + device='cuda:0'), grad: tensor([[ 1.4601e-03, -1.3130e-02, -2.2583e-02, ..., 1.3094e-03, + -4.4799e-04, -2.4612e-02], + [-1.5251e-02, -1.5140e-04, 8.5640e-04, ..., -1.1124e-02, + -4.9400e-04, 4.7040e-04], + [ 3.2482e-03, 7.5912e-03, 1.2321e-02, ..., 3.3531e-03, + 5.0640e-04, 1.4908e-02], + ..., + [-1.5125e-03, -8.2092e-03, -1.5507e-03, ..., -2.8706e-03, + 8.3327e-05, -1.2074e-03], + [ 1.1047e-02, 2.0676e-03, 3.0403e-03, ..., 7.8735e-03, + 1.4143e-03, 5.1842e-03], + [ 2.3499e-03, 5.7602e-03, 3.4561e-03, ..., 1.1539e-03, + 2.4486e-04, 3.8681e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0066, 0.0043, 0.0030, 0.0170, 0.0233, 0.0313, -0.0236, 0.0054, + -0.0313, -0.0171], device='cuda:0'), grad: tensor([-0.0211, -0.0063, 0.0145, 0.0088, 0.0003, -0.0048, 0.0039, -0.0125, + 0.0109, 0.0064], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 222.11, cls_loss 0.1691 cls_loss_mapping 0.2370 cls_loss_causal 1.3333 re_mapping 0.0893 re_causal 0.1940 /// teacc 96.15 lr 0.00010000 +Epoch 7, weight, value: tensor([[ 0.0096, 0.0449, 0.0328, ..., -0.0469, -0.0243, 0.0205], + [ 0.0462, -0.0079, -0.0069, ..., 0.0142, 0.0102, -0.0361], + [-0.0077, -0.0027, -0.0187, ..., -0.0156, -0.0081, 0.0016], + ..., + [ 0.0117, 0.0412, -0.0513, ..., -0.0260, 0.0059, -0.0314], + [-0.0066, -0.0143, 0.0159, ..., 0.0108, -0.0281, -0.0238], + [ 0.0155, -0.0170, 0.0025, ..., -0.0129, -0.0132, 0.0298]], + device='cuda:0'), grad: tensor([[-4.7989e-03, 2.9755e-02, 1.6830e-02, ..., 7.0429e-04, + 5.4896e-05, 1.9928e-02], + [-5.1689e-03, -8.4877e-04, 4.2572e-03, ..., -9.7275e-03, + -3.8452e-03, 4.0092e-03], + [ 3.7861e-03, 1.0368e-02, 8.0414e-03, ..., 2.2278e-03, + 3.1161e-04, 2.1763e-03], + ..., + [ 2.1324e-03, 4.6730e-03, 6.3324e-03, ..., 1.4000e-03, + 2.8062e-04, 5.9605e-04], + [ 1.7262e-04, 5.6038e-03, 8.4991e-03, ..., 3.6087e-03, + 4.4785e-03, 3.9101e-03], + [ 4.1866e-04, -3.2837e-02, -2.7374e-02, ..., -1.5717e-03, + -2.7522e-05, -2.7679e-02]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0068, 0.0043, 0.0028, 0.0169, 0.0232, 0.0313, -0.0237, 0.0057, + -0.0310, -0.0174], device='cuda:0'), grad: tensor([ 0.0241, -0.0030, 0.0233, -0.0246, -0.0155, -0.0007, 0.0012, 0.0155, + 0.0117, -0.0320], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 221.10, cls_loss 0.1594 cls_loss_mapping 0.2062 cls_loss_causal 1.3062 re_mapping 0.0763 re_causal 0.1682 /// teacc 96.48 lr 0.00010000 +Epoch 8, weight, value: tensor([[ 9.5645e-03, 4.6679e-02, 3.3759e-02, ..., -4.8533e-02, + -2.6581e-02, 2.0594e-02], + [ 4.7984e-02, -6.7997e-03, -7.5239e-03, ..., 1.5814e-02, + 1.2059e-02, -3.6902e-02], + [-9.2853e-03, -2.6466e-03, -1.9201e-02, ..., -1.6556e-02, + -7.1986e-03, 5.3548e-05], + ..., + [ 1.0127e-02, 4.3045e-02, -5.3172e-02, ..., -2.7381e-02, + 3.9840e-03, -3.3275e-02], + [-6.3788e-03, -1.6351e-02, 1.7163e-02, ..., 1.1576e-02, + -2.9850e-02, -2.3237e-02], + [ 1.3993e-02, -1.7653e-02, 2.8311e-03, ..., -1.3822e-02, + -1.4593e-02, 3.1092e-02]], device='cuda:0'), grad: tensor([[ 5.3930e-04, -1.5097e-03, -5.8174e-04, ..., 1.1902e-03, + 1.3375e-04, -1.5438e-04], + [ 2.4242e-03, 2.9621e-03, 2.5196e-03, ..., 3.6716e-03, + -5.2005e-05, 1.4477e-03], + [ 1.0633e-03, 1.8406e-03, 1.4019e-03, ..., 1.6718e-03, + 2.4527e-05, 9.9277e-04], + ..., + [ 2.4433e-03, -4.7455e-03, 3.3417e-03, ..., 3.3054e-03, + 2.6658e-05, 4.9782e-03], + [-8.4839e-03, 2.2907e-03, -6.9160e-03, ..., -1.0422e-02, + 4.7255e-04, -9.0551e-04], + [ 8.1921e-04, -5.2147e-03, -5.4474e-03, ..., -2.7847e-03, + 2.2471e-04, -8.9340e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0071, 0.0049, 0.0026, 0.0170, 0.0233, 0.0308, -0.0240, 0.0054, + -0.0309, -0.0173], device='cuda:0'), grad: tensor([ 0.0004, 0.0062, 0.0017, 0.0059, 0.0057, -0.0002, -0.0008, 0.0009, + -0.0055, -0.0144], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 220.63, cls_loss 0.1296 cls_loss_mapping 0.1668 cls_loss_causal 1.2145 re_mapping 0.0685 re_causal 0.1564 /// teacc 96.98 lr 0.00010000 +Epoch 9, weight, value: tensor([[ 0.0090, 0.0485, 0.0343, ..., -0.0498, -0.0278, 0.0206], + [ 0.0487, -0.0064, -0.0080, ..., 0.0163, 0.0147, -0.0386], + [-0.0105, -0.0028, -0.0199, ..., -0.0175, -0.0079, -0.0011], + ..., + [ 0.0092, 0.0445, -0.0551, ..., -0.0282, 0.0038, -0.0348], + [-0.0064, -0.0180, 0.0177, ..., 0.0123, -0.0317, -0.0236], + [ 0.0138, -0.0185, 0.0036, ..., -0.0137, -0.0161, 0.0322]], + device='cuda:0'), grad: tensor([[ 4.2367e-04, 4.2796e-04, -5.7459e-05, ..., 1.0729e-03, + 1.3173e-04, 1.1797e-03], + [-3.7937e-03, 4.6444e-04, 3.4308e-04, ..., -2.3613e-03, + -8.3447e-04, 5.1880e-04], + [ 6.2990e-04, -2.7370e-03, -1.7395e-03, ..., 1.1911e-03, + -9.4986e-04, -8.8596e-04], + ..., + [ 1.0519e-03, -7.4625e-05, 1.2188e-03, ..., 1.6823e-03, + 9.6226e-04, 1.5535e-03], + [ 1.1902e-03, 9.8610e-04, 1.7996e-03, ..., 5.0354e-04, + 5.4073e-04, 2.3155e-03], + [ 4.9782e-04, 4.2114e-03, 5.4502e-04, ..., 1.0979e-02, + -1.5008e-04, 2.7657e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0072, 0.0048, 0.0022, 0.0171, 0.0233, 0.0306, -0.0240, 0.0057, + -0.0308, -0.0171], device='cuda:0'), grad: tensor([ 0.0021, -0.0009, -0.0110, -0.0023, -0.0227, 0.0058, 0.0010, 0.0061, + 0.0062, 0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 8, time 219.99, cls_loss 0.1144 cls_loss_mapping 0.1485 cls_loss_causal 1.1824 re_mapping 0.0658 re_causal 0.1525 /// teacc 96.50 lr 0.00010000 +Epoch 10, weight, value: tensor([[ 0.0083, 0.0493, 0.0348, ..., -0.0516, -0.0292, 0.0210], + [ 0.0496, -0.0067, -0.0091, ..., 0.0169, 0.0169, -0.0400], + [-0.0112, -0.0026, -0.0206, ..., -0.0180, -0.0079, -0.0019], + ..., + [ 0.0081, 0.0461, -0.0564, ..., -0.0293, 0.0026, -0.0361], + [-0.0059, -0.0189, 0.0180, ..., 0.0134, -0.0330, -0.0242], + [ 0.0130, -0.0192, 0.0042, ..., -0.0143, -0.0175, 0.0331]], + device='cuda:0'), grad: tensor([[ 3.9196e-04, -8.0967e-04, -2.7847e-04, ..., 1.8787e-03, + -4.2945e-05, 4.9686e-04], + [-1.0902e-04, 3.3712e-04, 7.0238e-04, ..., 4.1270e-04, + -8.4996e-05, 5.3406e-04], + [ 5.8594e-03, 1.2207e-02, 7.1411e-03, ..., 2.6016e-03, + 1.2375e-02, 1.0185e-03], + ..., + [-4.7760e-03, -1.2985e-02, -4.4022e-03, ..., 1.5850e-03, + -1.2268e-02, 4.8923e-04], + [ 2.0294e-03, 1.9526e-04, -4.1008e-03, ..., -5.0201e-03, + 1.3628e-03, -1.4772e-03], + [ 1.1253e-03, -2.6989e-03, -3.3021e-04, ..., -1.6678e-02, + 7.8773e-04, 1.5488e-03]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0070, 0.0049, 0.0022, 0.0174, 0.0232, 0.0306, -0.0243, 0.0057, + -0.0307, -0.0173], device='cuda:0'), grad: tensor([ 0.0013, 0.0016, 0.0327, -0.0082, 0.0449, -0.0026, -0.0027, -0.0279, + -0.0060, -0.0330], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 220.70, cls_loss 0.1127 cls_loss_mapping 0.1464 cls_loss_causal 1.1127 re_mapping 0.0605 re_causal 0.1393 /// teacc 97.29 lr 0.00010000 +Epoch 11, weight, value: tensor([[ 0.0076, 0.0506, 0.0353, ..., -0.0530, -0.0307, 0.0217], + [ 0.0500, -0.0070, -0.0104, ..., 0.0172, 0.0171, -0.0415], + [-0.0123, -0.0018, -0.0207, ..., -0.0185, -0.0075, -0.0026], + ..., + [ 0.0074, 0.0470, -0.0578, ..., -0.0300, 0.0033, -0.0372], + [-0.0059, -0.0200, 0.0187, ..., 0.0141, -0.0346, -0.0247], + [ 0.0125, -0.0205, 0.0052, ..., -0.0148, -0.0168, 0.0342]], + device='cuda:0'), grad: tensor([[-6.7902e-04, 3.5667e-03, 9.3174e-04, ..., 5.9080e-04, + 5.4240e-05, 4.0016e-03], + [ 1.6699e-03, 9.3126e-04, 4.1604e-04, ..., 1.9464e-03, + 9.5367e-05, 3.6311e-04], + [ 3.6359e-04, -7.3099e-04, 1.8120e-03, ..., 1.6575e-03, + -1.2457e-04, 2.0657e-03], + ..., + [ 1.0335e-04, -2.3132e-02, -2.3289e-03, ..., -2.6569e-03, + 4.6134e-04, -1.1581e-02], + [ 1.5011e-03, 1.9627e-03, -2.9202e-03, ..., -2.2507e-03, + 1.5557e-04, -3.5152e-03], + [ 2.3675e-04, 1.0742e-02, 3.9597e-03, ..., 2.7390e-03, + 1.4015e-05, 8.4457e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0071, 0.0045, 0.0025, 0.0171, 0.0231, 0.0308, -0.0242, 0.0058, + -0.0306, -0.0172], device='cuda:0'), grad: tensor([ 0.0059, 0.0036, 0.0040, 0.0004, 0.0058, 0.0013, -0.0023, -0.0410, + -0.0020, 0.0243], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 219.69, cls_loss 0.1089 cls_loss_mapping 0.1431 cls_loss_causal 1.1138 re_mapping 0.0560 re_causal 0.1323 /// teacc 97.06 lr 0.00010000 +Epoch 12, weight, value: tensor([[ 0.0077, 0.0518, 0.0356, ..., -0.0537, -0.0312, 0.0216], + [ 0.0506, -0.0070, -0.0111, ..., 0.0176, 0.0193, -0.0429], + [-0.0134, -0.0019, -0.0214, ..., -0.0194, -0.0083, -0.0037], + ..., + [ 0.0065, 0.0486, -0.0593, ..., -0.0311, 0.0032, -0.0380], + [-0.0055, -0.0220, 0.0193, ..., 0.0151, -0.0354, -0.0247], + [ 0.0122, -0.0215, 0.0052, ..., -0.0152, -0.0179, 0.0346]], + device='cuda:0'), grad: tensor([[-0.0003, -0.0039, -0.0028, ..., 0.0002, 0.0001, -0.0018], + [-0.0016, -0.0002, 0.0006, ..., -0.0008, -0.0003, 0.0003], + [ 0.0009, 0.0016, 0.0038, ..., 0.0002, 0.0001, 0.0014], + ..., + [ 0.0008, 0.0009, 0.0046, ..., 0.0003, 0.0002, 0.0028], + [ 0.0011, 0.0010, 0.0049, ..., 0.0004, 0.0003, 0.0026], + [ 0.0004, -0.0004, -0.0016, ..., -0.0005, -0.0002, -0.0031]], + device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0072, 0.0044, 0.0026, 0.0175, 0.0231, 0.0305, -0.0242, 0.0059, + -0.0305, -0.0176], device='cuda:0'), grad: tensor([-4.0779e-03, 5.7295e-06, 4.3793e-03, -1.5244e-02, 1.0633e-03, + -9.2363e-04, 2.3613e-03, 7.5340e-03, 7.4158e-03, -2.5177e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 220.20, cls_loss 0.0997 cls_loss_mapping 0.1288 cls_loss_causal 1.0871 re_mapping 0.0517 re_causal 0.1222 /// teacc 97.69 lr 0.00010000 +Epoch 13, weight, value: tensor([[ 0.0073, 0.0526, 0.0364, ..., -0.0547, -0.0317, 0.0218], + [ 0.0517, -0.0070, -0.0118, ..., 0.0185, 0.0220, -0.0439], + [-0.0148, -0.0023, -0.0219, ..., -0.0201, -0.0085, -0.0048], + ..., + [ 0.0060, 0.0496, -0.0604, ..., -0.0317, 0.0039, -0.0396], + [-0.0053, -0.0234, 0.0195, ..., 0.0158, -0.0373, -0.0247], + [ 0.0121, -0.0217, 0.0063, ..., -0.0155, -0.0180, 0.0363]], + device='cuda:0'), grad: tensor([[ 0.0070, 0.0002, 0.0089, ..., 0.0008, 0.0001, 0.0064], + [ 0.0012, 0.0012, 0.0007, ..., 0.0004, 0.0001, 0.0010], + [ 0.0008, -0.0001, 0.0007, ..., 0.0004, -0.0005, 0.0007], + ..., + [-0.0053, -0.0071, -0.0018, ..., 0.0003, 0.0002, -0.0041], + [-0.0077, 0.0008, 0.0077, ..., -0.0116, -0.0051, -0.0032], + [ 0.0052, 0.0049, 0.0050, ..., 0.0007, 0.0002, 0.0056]], + device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0074, 0.0046, 0.0023, 0.0171, 0.0231, 0.0301, -0.0242, 0.0059, + -0.0303, -0.0171], device='cuda:0'), grad: tensor([ 0.0065, 0.0027, -0.0007, -0.0220, -0.0017, 0.0090, 0.0061, -0.0085, + -0.0006, 0.0091], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 221.66, cls_loss 0.0806 cls_loss_mapping 0.1054 cls_loss_causal 1.0424 re_mapping 0.0504 re_causal 0.1206 /// teacc 97.65 lr 0.00010000 +Epoch 14, weight, value: tensor([[ 0.0070, 0.0535, 0.0368, ..., -0.0556, -0.0326, 0.0221], + [ 0.0523, -0.0063, -0.0123, ..., 0.0188, 0.0229, -0.0454], + [-0.0150, -0.0025, -0.0219, ..., -0.0206, -0.0076, -0.0047], + ..., + [ 0.0051, 0.0506, -0.0622, ..., -0.0327, 0.0042, -0.0408], + [-0.0053, -0.0244, 0.0199, ..., 0.0165, -0.0383, -0.0247], + [ 0.0117, -0.0224, 0.0068, ..., -0.0159, -0.0191, 0.0370]], + device='cuda:0'), grad: tensor([[-9.2506e-04, -1.1635e-03, -6.9332e-04, ..., 2.3746e-04, + 2.9787e-05, -6.1417e-04], + [ 9.6893e-04, 1.1892e-03, 1.0309e-03, ..., 9.8038e-04, + -9.0182e-05, 7.1526e-04], + [ 8.8024e-04, 9.7275e-04, -2.8327e-05, ..., -2.7537e-04, + -1.9813e-04, 4.1389e-04], + ..., + [ 1.0471e-03, -5.1079e-03, 1.0364e-05, ..., 1.8263e-03, + 3.2485e-05, 8.4114e-04], + [ 2.6360e-03, 2.3403e-03, 2.8839e-03, ..., 2.4719e-03, + 5.4359e-04, 2.9182e-03], + [-4.7684e-03, -1.0042e-03, -6.2065e-03, ..., -2.3708e-03, + 3.6985e-05, -5.0354e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0073, 0.0045, 0.0026, 0.0173, 0.0232, 0.0298, -0.0243, 0.0058, + -0.0302, -0.0173], device='cuda:0'), grad: tensor([-0.0007, 0.0032, 0.0016, 0.0061, -0.0024, -0.0009, 0.0003, -0.0083, + 0.0086, -0.0075], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 219.62, cls_loss 0.0816 cls_loss_mapping 0.1040 cls_loss_causal 1.0737 re_mapping 0.0457 re_causal 0.1136 /// teacc 97.55 lr 0.00010000 +Epoch 15, weight, value: tensor([[ 0.0067, 0.0543, 0.0372, ..., -0.0563, -0.0337, 0.0223], + [ 0.0533, -0.0060, -0.0130, ..., 0.0194, 0.0232, -0.0467], + [-0.0158, -0.0027, -0.0224, ..., -0.0211, -0.0064, -0.0057], + ..., + [ 0.0042, 0.0513, -0.0634, ..., -0.0343, 0.0050, -0.0415], + [-0.0051, -0.0251, 0.0202, ..., 0.0170, -0.0395, -0.0248], + [ 0.0109, -0.0228, 0.0082, ..., -0.0163, -0.0202, 0.0386]], + device='cuda:0'), grad: tensor([[ 1.4293e-04, 1.9598e-04, 1.1196e-03, ..., 9.8050e-05, + 7.2479e-05, 1.2302e-03], + [-3.0708e-04, 2.3723e-04, 2.9993e-04, ..., -2.1541e-04, + -1.0449e-04, 1.4913e-04], + [ 1.1468e-04, 2.3162e-04, 4.6110e-04, ..., -5.3525e-05, + 4.9978e-05, 2.6751e-04], + ..., + [-6.3002e-05, -1.8501e-03, 9.9182e-04, ..., 9.4771e-05, + -1.0914e-04, 1.2093e-03], + [ 3.6502e-04, 3.5429e-04, 1.9035e-03, ..., 5.9307e-05, + 1.1194e-04, 7.9918e-04], + [ 5.1403e-04, 1.8489e-04, -4.7398e-04, ..., 9.0933e-04, + 3.6526e-04, -2.7199e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0075, 0.0043, 0.0026, 0.0170, 0.0233, 0.0298, -0.0246, 0.0059, + -0.0300, -0.0170], device='cuda:0'), grad: tensor([ 1.3027e-03, 4.9686e-04, -1.8072e-04, -3.7308e-03, -1.5554e-03, + 3.6621e-04, 2.4527e-05, -1.4362e-03, 2.7523e-03, 1.9627e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 220.42, cls_loss 0.0733 cls_loss_mapping 0.0957 cls_loss_causal 0.9991 re_mapping 0.0436 re_causal 0.1045 /// teacc 97.41 lr 0.00010000 +Epoch 16, weight, value: tensor([[ 0.0067, 0.0553, 0.0374, ..., -0.0572, -0.0349, 0.0224], + [ 0.0541, -0.0060, -0.0134, ..., 0.0199, 0.0238, -0.0478], + [-0.0164, -0.0030, -0.0227, ..., -0.0219, -0.0061, -0.0061], + ..., + [ 0.0035, 0.0523, -0.0648, ..., -0.0351, 0.0050, -0.0426], + [-0.0051, -0.0259, 0.0206, ..., 0.0177, -0.0403, -0.0249], + [ 0.0106, -0.0233, 0.0086, ..., -0.0167, -0.0209, 0.0396]], + device='cuda:0'), grad: tensor([[ 2.1982e-04, -7.0524e-04, -5.6458e-04, ..., 3.3593e-04, + 6.4522e-06, -4.7755e-04], + [ 7.4482e-04, 4.0197e-04, 1.3170e-03, ..., 4.9353e-04, + 4.0078e-04, 1.3733e-03], + [ 1.2219e-04, 1.3423e-04, 1.3447e-04, ..., 1.1021e-04, + 4.5337e-06, 1.6093e-04], + ..., + [ 3.1137e-04, -2.1400e-03, 9.5606e-04, ..., 3.8457e-04, + -7.2765e-04, 9.2149e-05], + [ 6.2828e-03, 2.0838e-04, 1.0574e-02, ..., 6.7101e-03, + 1.3466e-03, 1.0597e-02], + [-1.7300e-05, 1.1148e-03, -3.2825e-03, ..., 1.1212e-04, + 7.1335e-04, -1.5526e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0075, 0.0042, 0.0027, 0.0172, 0.0233, 0.0296, -0.0248, 0.0059, + -0.0298, -0.0173], device='cuda:0'), grad: tensor([-0.0005, 0.0019, 0.0002, -0.0092, 0.0021, 0.0296, -0.0307, -0.0017, + 0.0108, -0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 220.13, cls_loss 0.0584 cls_loss_mapping 0.0774 cls_loss_causal 0.9628 re_mapping 0.0421 re_causal 0.1043 /// teacc 97.49 lr 0.00010000 +Epoch 17, weight, value: tensor([[ 0.0063, 0.0563, 0.0378, ..., -0.0581, -0.0361, 0.0227], + [ 0.0547, -0.0062, -0.0134, ..., 0.0208, 0.0256, -0.0488], + [-0.0172, -0.0033, -0.0231, ..., -0.0224, -0.0070, -0.0066], + ..., + [ 0.0030, 0.0531, -0.0657, ..., -0.0355, 0.0054, -0.0435], + [-0.0049, -0.0262, 0.0209, ..., 0.0180, -0.0416, -0.0251], + [ 0.0103, -0.0238, 0.0093, ..., -0.0172, -0.0219, 0.0406]], + device='cuda:0'), grad: tensor([[ 5.8794e-04, 4.2558e-05, 2.7442e-04, ..., 4.2367e-04, + 5.1081e-05, 4.0245e-04], + [ 1.0933e-02, -6.7651e-05, 2.2621e-03, ..., 6.9389e-03, + 3.3903e-04, 3.5973e-03], + [ 1.0300e-03, 2.2531e-04, 2.4700e-04, ..., 5.6505e-04, + -3.1185e-04, 4.2701e-04], + ..., + [ 3.7313e-04, -8.0681e-04, 4.6039e-04, ..., 4.2105e-04, + 1.1581e-04, 5.7983e-04], + [-1.2350e-03, 1.3733e-04, -1.1473e-03, ..., -1.3304e-03, + 1.4567e-04, -1.6260e-03], + [ 2.9877e-05, 1.2922e-04, -3.0231e-03, ..., 1.2369e-03, + 1.1597e-03, -4.2419e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0076, 0.0045, 0.0025, 0.0171, 0.0234, 0.0295, -0.0250, 0.0062, + -0.0299, -0.0173], device='cuda:0'), grad: tensor([ 7.7820e-04, 1.0666e-02, -2.1696e-04, 3.1586e-03, -2.0084e-03, + 4.2076e-03, -1.6663e-02, 3.8457e-04, 1.8179e-05, -3.2973e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 220.42, cls_loss 0.0681 cls_loss_mapping 0.0863 cls_loss_causal 0.9701 re_mapping 0.0395 re_causal 0.0956 /// teacc 97.89 lr 0.00010000 +Epoch 18, weight, value: tensor([[ 0.0058, 0.0570, 0.0378, ..., -0.0590, -0.0369, 0.0225], + [ 0.0553, -0.0061, -0.0142, ..., 0.0214, 0.0267, -0.0502], + [-0.0180, -0.0035, -0.0230, ..., -0.0226, -0.0075, -0.0070], + ..., + [ 0.0023, 0.0543, -0.0668, ..., -0.0364, 0.0057, -0.0446], + [-0.0047, -0.0268, 0.0210, ..., 0.0182, -0.0422, -0.0255], + [ 0.0096, -0.0247, 0.0097, ..., -0.0173, -0.0220, 0.0418]], + device='cuda:0'), grad: tensor([[-1.6606e-04, -2.2507e-03, -2.5892e-04, ..., 1.1331e-04, + 3.0205e-05, -9.4175e-04], + [-1.0264e-04, 1.7176e-03, 3.6168e-04, ..., -3.0780e-04, + 2.7394e-04, 9.9599e-05], + [-3.8624e-04, 2.9297e-03, 5.4169e-04, ..., 6.2399e-07, + 9.6321e-04, 1.1337e-04], + ..., + [ 2.6369e-04, 5.0774e-03, 2.3155e-03, ..., 1.6797e-04, + 1.3809e-03, 1.4105e-03], + [ 5.1165e-04, 1.4029e-03, 7.2908e-04, ..., 3.1853e-04, + 2.9731e-04, 1.0271e-03], + [ 1.1438e-04, -1.8034e-03, -2.6817e-03, ..., -1.7774e-04, + 4.6402e-05, -2.5082e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0074, 0.0043, 0.0027, 0.0170, 0.0231, 0.0298, -0.0250, 0.0064, + -0.0299, -0.0172], device='cuda:0'), grad: tensor([-0.0031, 0.0063, 0.0029, -0.0224, 0.0013, 0.0007, 0.0010, 0.0132, + 0.0042, -0.0042], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 219.83, cls_loss 0.0489 cls_loss_mapping 0.0716 cls_loss_causal 0.9451 re_mapping 0.0377 re_causal 0.0944 /// teacc 98.13 lr 0.00010000 +Epoch 19, weight, value: tensor([[ 0.0056, 0.0584, 0.0385, ..., -0.0598, -0.0374, 0.0231], + [ 0.0556, -0.0063, -0.0146, ..., 0.0213, 0.0277, -0.0511], + [-0.0189, -0.0038, -0.0234, ..., -0.0231, -0.0075, -0.0080], + ..., + [ 0.0018, 0.0550, -0.0680, ..., -0.0369, 0.0069, -0.0456], + [-0.0041, -0.0277, 0.0213, ..., 0.0190, -0.0429, -0.0256], + [ 0.0091, -0.0252, 0.0102, ..., -0.0176, -0.0224, 0.0425]], + device='cuda:0'), grad: tensor([[ 1.1051e-04, -2.0874e-04, -1.0979e-04, ..., 1.3673e-04, + 4.6015e-05, -6.8069e-05], + [-2.0638e-03, -9.9421e-05, 4.3869e-05, ..., -1.7042e-03, + -6.4707e-04, 5.6326e-05], + [ 7.4768e-04, 2.9773e-05, 1.3733e-04, ..., 8.0013e-04, + 6.0987e-04, 1.9515e-04], + ..., + [ 1.4973e-04, -1.7166e-04, 2.4748e-04, ..., 2.0123e-04, + 5.3167e-05, 2.1780e-04], + [ 1.5160e-02, 1.7381e-04, 1.0971e-02, ..., 2.0920e-02, + 1.3718e-02, 1.0208e-02], + [ 3.6979e-04, -2.7463e-05, -1.4043e-04, ..., 5.1165e-04, + 3.5214e-04, 5.3465e-05]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0080, 0.0041, 0.0025, 0.0170, 0.0232, 0.0298, -0.0253, 0.0066, + -0.0298, -0.0174], device='cuda:0'), grad: tensor([-6.8367e-05, -1.7681e-03, 7.6818e-04, 4.3416e-04, 3.6407e-04, + -2.9572e-02, 6.0987e-04, 2.7752e-04, 2.9022e-02, -6.4850e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 219.10, cls_loss 0.0536 cls_loss_mapping 0.0717 cls_loss_causal 0.9451 re_mapping 0.0348 re_causal 0.0893 /// teacc 98.08 lr 0.00010000 +Epoch 20, weight, value: tensor([[ 0.0055, 0.0591, 0.0385, ..., -0.0605, -0.0389, 0.0230], + [ 0.0562, -0.0061, -0.0152, ..., 0.0215, 0.0289, -0.0518], + [-0.0198, -0.0042, -0.0237, ..., -0.0235, -0.0078, -0.0085], + ..., + [ 0.0009, 0.0560, -0.0693, ..., -0.0377, 0.0073, -0.0461], + [-0.0038, -0.0285, 0.0214, ..., 0.0195, -0.0442, -0.0259], + [ 0.0094, -0.0254, 0.0110, ..., -0.0177, -0.0222, 0.0437]], + device='cuda:0'), grad: tensor([[ 4.0591e-05, -2.1148e-04, -9.0837e-05, ..., 2.0587e-04, + 1.8328e-05, -9.5546e-05], + [-2.4147e-03, 1.0036e-05, -9.7215e-05, ..., -2.6913e-03, + -2.0161e-03, 3.1084e-05], + [ 1.7223e-03, 4.3726e-04, 2.2066e-04, ..., 2.5406e-03, + 1.5469e-03, 6.5684e-05], + ..., + [ 2.0754e-04, -1.1021e-04, 2.1553e-04, ..., 2.2411e-04, + 1.0091e-04, 2.4438e-04], + [-3.0231e-03, 6.5863e-05, -2.1629e-03, ..., -2.6226e-03, + 1.1224e-04, -3.9673e-03], + [ 1.1224e-04, -1.6436e-05, -7.6151e-04, ..., 2.0301e-04, + 1.9401e-05, -9.8324e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0079, 0.0039, 0.0026, 0.0168, 0.0234, 0.0295, -0.0255, 0.0067, + -0.0298, -0.0170], device='cuda:0'), grad: tensor([ 0.0001, -0.0035, 0.0035, 0.0005, -0.0154, 0.0029, 0.0163, 0.0005, + -0.0042, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 219.26, cls_loss 0.0496 cls_loss_mapping 0.0627 cls_loss_causal 0.9332 re_mapping 0.0336 re_causal 0.0876 /// teacc 98.05 lr 0.00010000 +Epoch 21, weight, value: tensor([[ 0.0058, 0.0597, 0.0387, ..., -0.0612, -0.0399, 0.0231], + [ 0.0569, -0.0062, -0.0156, ..., 0.0219, 0.0299, -0.0524], + [-0.0207, -0.0047, -0.0242, ..., -0.0241, -0.0073, -0.0092], + ..., + [ 0.0002, 0.0567, -0.0705, ..., -0.0381, 0.0070, -0.0472], + [-0.0038, -0.0295, 0.0215, ..., 0.0199, -0.0455, -0.0261], + [ 0.0086, -0.0252, 0.0117, ..., -0.0183, -0.0229, 0.0444]], + device='cuda:0'), grad: tensor([[ 1.1665e-04, 6.6936e-05, 1.3566e-04, ..., 6.4433e-05, + 4.7892e-05, 8.9705e-05], + [-3.3951e-03, 1.2290e-04, 2.7990e-04, ..., -2.9411e-03, + -9.0170e-04, 8.5652e-05], + [-1.5503e-02, -2.0523e-03, 2.6627e-03, ..., -1.3382e-02, + 1.7557e-03, 9.4509e-04], + ..., + [ 3.3665e-04, -2.7409e-03, -1.6785e-03, ..., 7.9393e-04, + -6.5565e-04, -2.2964e-03], + [ 1.6739e-02, 1.9207e-03, -4.5609e-04, ..., 1.4687e-02, + 8.9788e-04, 1.8358e-04], + [ 6.4564e-04, 2.1019e-03, 5.5237e-03, ..., 3.7932e-04, + 2.9430e-03, 3.1948e-03]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0077, 0.0038, 0.0024, 0.0171, 0.0234, 0.0298, -0.0252, 0.0064, + -0.0301, -0.0168], device='cuda:0'), grad: tensor([ 0.0004, -0.0020, -0.0226, 0.0039, -0.0104, -0.0001, 0.0004, -0.0071, + 0.0267, 0.0109], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 219.80, cls_loss 0.0453 cls_loss_mapping 0.0621 cls_loss_causal 0.9226 re_mapping 0.0327 re_causal 0.0852 /// teacc 98.40 lr 0.00010000 +Epoch 22, weight, value: tensor([[ 0.0056, 0.0604, 0.0388, ..., -0.0619, -0.0400, 0.0234], + [ 0.0575, -0.0059, -0.0155, ..., 0.0227, 0.0304, -0.0530], + [-0.0211, -0.0047, -0.0245, ..., -0.0246, -0.0076, -0.0098], + ..., + [-0.0004, 0.0575, -0.0717, ..., -0.0393, 0.0077, -0.0480], + [-0.0035, -0.0300, 0.0217, ..., 0.0205, -0.0456, -0.0258], + [ 0.0084, -0.0261, 0.0124, ..., -0.0186, -0.0231, 0.0453]], + device='cuda:0'), grad: tensor([[-4.5985e-05, -6.9046e-04, -3.7909e-04, ..., 1.9145e-04, + 2.6543e-06, -3.5024e-04], + [-2.0170e-04, 5.9932e-05, 1.0216e-04, ..., -4.5627e-05, + -3.0935e-05, 7.9393e-05], + [ 1.9825e-04, 3.2425e-04, 2.2137e-04, ..., 2.4307e-04, + -1.8924e-05, 2.0266e-04], + ..., + [ 8.8751e-05, -2.0337e-04, 1.0478e-04, ..., 1.3888e-04, + 1.1943e-05, 9.8467e-05], + [ 4.4727e-04, 1.3149e-04, 1.7762e-04, ..., 5.2977e-04, + 1.7032e-05, 1.2341e-03], + [ 1.6820e-04, 2.4343e-04, 2.5139e-03, ..., 4.7836e-03, + 4.0084e-06, 2.3632e-03]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0077, 0.0040, 0.0026, 0.0167, 0.0234, 0.0295, -0.0252, 0.0066, + -0.0299, -0.0168], device='cuda:0'), grad: tensor([-7.6818e-04, 6.9678e-05, 4.7684e-04, 6.8188e-04, -9.1553e-03, + 4.3607e-04, -2.4433e-03, 9.4771e-05, 1.2293e-03, 9.3842e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 218.49, cls_loss 0.0501 cls_loss_mapping 0.0673 cls_loss_causal 0.8941 re_mapping 0.0318 re_causal 0.0784 /// teacc 98.07 lr 0.00010000 +Epoch 23, weight, value: tensor([[ 0.0055, 0.0611, 0.0393, ..., -0.0624, -0.0404, 0.0241], + [ 0.0580, -0.0059, -0.0157, ..., 0.0230, 0.0310, -0.0537], + [-0.0213, -0.0052, -0.0252, ..., -0.0251, -0.0074, -0.0109], + ..., + [-0.0010, 0.0587, -0.0726, ..., -0.0400, 0.0074, -0.0488], + [-0.0035, -0.0305, 0.0218, ..., 0.0210, -0.0465, -0.0261], + [ 0.0077, -0.0267, 0.0126, ..., -0.0195, -0.0236, 0.0456]], + device='cuda:0'), grad: tensor([[ 4.9305e-04, 1.2808e-03, 5.1403e-04, ..., 6.4087e-04, + 3.5584e-05, 1.3647e-03], + [-5.2929e-04, 1.5032e-04, 1.4496e-04, ..., -4.6825e-04, + -2.1553e-04, 1.2052e-04], + [ 4.8709e-04, 3.5524e-05, 5.8460e-04, ..., 1.0735e-04, + -2.2745e-04, 4.8494e-04], + ..., + [ 2.9516e-04, -1.2426e-03, 2.2507e-04, ..., 2.0361e-04, + 1.9252e-04, 2.1100e-04], + [ 1.1044e-03, 2.8062e-04, 1.5516e-03, ..., 2.0099e-04, + 1.5306e-04, 1.3542e-03], + [ 1.1617e-04, 1.0508e-04, 4.7177e-05, ..., 1.1814e-04, + 4.0948e-05, -9.6202e-05]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0081, 0.0041, 0.0023, 0.0165, 0.0232, 0.0302, -0.0251, 0.0067, + -0.0298, -0.0174], device='cuda:0'), grad: tensor([ 0.0023, -0.0002, 0.0003, -0.0031, 0.0002, 0.0013, -0.0024, -0.0013, + 0.0026, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 218.35, cls_loss 0.0504 cls_loss_mapping 0.0670 cls_loss_causal 0.8771 re_mapping 0.0303 re_causal 0.0771 /// teacc 98.16 lr 0.00010000 +Epoch 24, weight, value: tensor([[ 0.0051, 0.0611, 0.0390, ..., -0.0632, -0.0408, 0.0235], + [ 0.0586, -0.0060, -0.0163, ..., 0.0230, 0.0322, -0.0546], + [-0.0221, -0.0051, -0.0253, ..., -0.0254, -0.0077, -0.0111], + ..., + [-0.0015, 0.0595, -0.0738, ..., -0.0407, 0.0073, -0.0499], + [-0.0035, -0.0305, 0.0221, ..., 0.0213, -0.0472, -0.0262], + [ 0.0070, -0.0269, 0.0132, ..., -0.0197, -0.0232, 0.0466]], + device='cuda:0'), grad: tensor([[-3.4118e-04, -8.9645e-04, -2.0123e-04, ..., 1.4699e-04, + 8.8662e-06, 1.3888e-04], + [-8.0109e-04, 3.9220e-04, 7.1406e-05, ..., -5.9652e-04, + -8.5413e-05, 5.3406e-05], + [ 2.1875e-04, 5.2977e-04, 2.3663e-04, ..., 2.7204e-04, + -2.5049e-05, 1.7333e-04], + ..., + [ 1.6189e-04, 2.8687e-03, 1.5569e-04, ..., 2.8553e-03, + 4.5031e-05, 2.6917e-04], + [ 1.7428e-04, 9.2649e-04, -1.7080e-03, ..., -1.9574e-04, + 6.3121e-05, -9.7132e-04], + [ 1.5223e-04, 2.2697e-03, 1.3363e-04, ..., 1.2646e-03, + 5.6893e-05, -2.6298e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0072, 0.0038, 0.0025, 0.0168, 0.0233, 0.0303, -0.0252, 0.0066, + -0.0298, -0.0170], device='cuda:0'), grad: tensor([-1.0109e-03, -1.7083e-04, 1.0939e-03, 1.4820e-03, -1.2550e-02, + 3.3927e-04, -2.3469e-05, 7.5531e-03, -4.8733e-04, 3.7708e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 219.35, cls_loss 0.0496 cls_loss_mapping 0.0664 cls_loss_causal 0.8789 re_mapping 0.0292 re_causal 0.0773 /// teacc 98.13 lr 0.00010000 +Epoch 25, weight, value: tensor([[ 0.0054, 0.0615, 0.0394, ..., -0.0639, -0.0413, 0.0239], + [ 0.0593, -0.0056, -0.0164, ..., 0.0237, 0.0332, -0.0548], + [-0.0227, -0.0055, -0.0257, ..., -0.0262, -0.0076, -0.0118], + ..., + [-0.0024, 0.0608, -0.0746, ..., -0.0412, 0.0087, -0.0506], + [-0.0036, -0.0314, 0.0223, ..., 0.0215, -0.0483, -0.0262], + [ 0.0065, -0.0274, 0.0135, ..., -0.0201, -0.0241, 0.0470]], + device='cuda:0'), grad: tensor([[ 2.5773e-04, -2.8515e-04, -1.2553e-04, ..., 1.9753e-04, + 3.1918e-05, -2.6393e-04], + [ 2.7390e-03, 1.0118e-05, 1.3244e-04, ..., 2.4414e-03, + 1.0872e-03, 7.9441e-04], + [ 1.3554e-04, 3.7968e-05, 1.0228e-04, ..., -3.5000e-04, + -6.9714e-04, 8.9407e-05], + ..., + [ 1.1563e-04, -1.6689e-05, 9.6083e-05, ..., 1.2267e-04, + 1.2243e-04, 1.0562e-04], + [-7.4539e-03, 5.1409e-05, -6.0409e-05, ..., -5.7449e-03, + -1.5554e-03, -2.6073e-03], + [ 2.7585e-04, 1.6236e-04, -1.0881e-03, ..., -3.6335e-04, + -6.9559e-05, -1.4200e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0069, 0.0040, 0.0019, 0.0167, 0.0234, 0.0302, -0.0249, 0.0073, + -0.0297, -0.0174], device='cuda:0'), grad: tensor([-3.2961e-05, 3.4847e-03, -2.0485e-03, -3.2568e-04, 4.3335e-03, + 3.2005e-03, -1.7290e-03, 6.8903e-04, -5.4932e-03, -2.0771e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 218.92, cls_loss 0.0489 cls_loss_mapping 0.0642 cls_loss_causal 0.8523 re_mapping 0.0289 re_causal 0.0755 /// teacc 98.18 lr 0.00010000 +Epoch 26, weight, value: tensor([[ 0.0057, 0.0623, 0.0403, ..., -0.0645, -0.0410, 0.0244], + [ 0.0597, -0.0055, -0.0171, ..., 0.0240, 0.0339, -0.0561], + [-0.0233, -0.0057, -0.0260, ..., -0.0267, -0.0066, -0.0123], + ..., + [-0.0032, 0.0613, -0.0757, ..., -0.0415, 0.0080, -0.0511], + [-0.0035, -0.0320, 0.0225, ..., 0.0220, -0.0492, -0.0266], + [ 0.0063, -0.0284, 0.0139, ..., -0.0200, -0.0230, 0.0477]], + device='cuda:0'), grad: tensor([[ 4.1342e-04, 1.8346e-04, 7.9536e-04, ..., 1.0500e-03, + 8.3637e-04, 6.2943e-04], + [-7.8773e-04, 4.5395e-04, 2.0790e-04, ..., -3.2711e-04, + -6.1870e-05, 1.8167e-04], + [ 1.4954e-03, 5.5933e-04, 9.4528e-03, ..., 1.2751e-03, + 5.8823e-03, 4.7607e-03], + ..., + [ 2.0564e-04, -2.8324e-03, 4.4799e-04, ..., -1.3232e-04, + -1.1339e-03, 2.8324e-04], + [ 3.8090e-03, 1.3046e-03, 2.3289e-03, ..., 3.5496e-03, + 6.2180e-03, 1.3952e-03], + [ 2.1076e-04, 7.9441e-04, 1.5936e-03, ..., 1.9474e-03, + 4.0388e-04, 2.0862e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0078, 0.0037, 0.0025, 0.0166, 0.0232, 0.0300, -0.0250, 0.0069, + -0.0297, -0.0174], device='cuda:0'), grad: tensor([ 2.2564e-03, 9.7561e-04, 2.4323e-02, -2.2736e-02, -8.3506e-05, + -1.1223e-02, -2.0866e-03, -6.9542e-03, 1.0719e-02, 4.8141e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 219.04, cls_loss 0.0429 cls_loss_mapping 0.0577 cls_loss_causal 0.8779 re_mapping 0.0278 re_causal 0.0738 /// teacc 98.31 lr 0.00010000 +Epoch 27, weight, value: tensor([[ 0.0051, 0.0629, 0.0405, ..., -0.0655, -0.0417, 0.0247], + [ 0.0607, -0.0057, -0.0170, ..., 0.0246, 0.0348, -0.0565], + [-0.0245, -0.0057, -0.0264, ..., -0.0274, -0.0077, -0.0129], + ..., + [-0.0036, 0.0619, -0.0770, ..., -0.0419, 0.0088, -0.0518], + [-0.0038, -0.0329, 0.0221, ..., 0.0222, -0.0505, -0.0272], + [ 0.0056, -0.0286, 0.0143, ..., -0.0204, -0.0237, 0.0483]], + device='cuda:0'), grad: tensor([[ 1.0714e-05, 1.8954e-04, 2.3139e-04, ..., 1.3053e-05, + 9.5814e-06, 3.1638e-04], + [-3.6049e-04, 6.8724e-05, 8.7544e-06, ..., -1.9324e-04, + -1.0598e-04, -1.8775e-06], + [ 4.5180e-05, -2.4343e-04, -1.5521e-04, ..., 2.1398e-05, + -3.5137e-05, 1.3745e-04], + ..., + [ 7.7069e-05, -1.7910e-03, 5.6076e-04, ..., 6.7651e-05, + -2.8476e-05, -1.4293e-04], + [ 2.0251e-05, 1.1104e-04, 1.3793e-04, ..., -9.6619e-05, + 5.4181e-05, 2.0134e-04], + [ 9.9242e-05, 1.0691e-03, -1.7738e-03, ..., 7.6175e-05, + -4.6641e-06, -1.4668e-03]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0077, 0.0040, 0.0024, 0.0167, 0.0236, 0.0305, -0.0252, 0.0067, + -0.0302, -0.0173], device='cuda:0'), grad: tensor([ 4.5276e-04, -1.5342e-04, -4.5562e-04, 1.2884e-03, 3.6907e-04, + 2.5654e-04, 1.0781e-05, -1.7262e-03, 2.4211e-04, -2.8563e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 218.59, cls_loss 0.0436 cls_loss_mapping 0.0513 cls_loss_causal 0.8530 re_mapping 0.0279 re_causal 0.0697 /// teacc 98.14 lr 0.00010000 +Epoch 28, weight, value: tensor([[ 0.0046, 0.0633, 0.0408, ..., -0.0665, -0.0427, 0.0250], + [ 0.0611, -0.0056, -0.0174, ..., 0.0250, 0.0352, -0.0572], + [-0.0248, -0.0059, -0.0266, ..., -0.0278, -0.0068, -0.0136], + ..., + [-0.0042, 0.0627, -0.0780, ..., -0.0429, 0.0086, -0.0525], + [-0.0034, -0.0336, 0.0226, ..., 0.0231, -0.0512, -0.0273], + [ 0.0046, -0.0292, 0.0150, ..., -0.0211, -0.0242, 0.0488]], + device='cuda:0'), grad: tensor([[ 5.3263e-04, 1.0365e-04, 3.3069e-04, ..., 2.9778e-04, + 7.8231e-06, 3.8052e-04], + [ 3.3712e-04, 5.7161e-05, 1.8036e-04, ..., 6.1274e-04, + 1.0198e-04, 9.8884e-05], + [ 1.1663e-03, 2.9147e-05, 8.0442e-04, ..., 1.4949e-04, + 7.5054e-04, 8.9765e-05], + ..., + [-1.5581e-04, -3.8586e-03, -7.3433e-04, ..., -5.2404e-04, + 9.2834e-06, -2.1648e-03], + [ 4.4060e-03, 3.0732e-04, 3.6812e-03, ..., 2.7981e-03, + 3.4356e-04, 4.6730e-03], + [ 2.0733e-03, 4.5319e-03, 1.6689e-03, ..., 5.6190e-03, + 1.0215e-05, 3.5381e-03]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0075, 0.0039, 0.0031, 0.0164, 0.0236, 0.0306, -0.0253, 0.0065, + -0.0299, -0.0176], device='cuda:0'), grad: tensor([ 0.0006, 0.0011, 0.0027, -0.0029, -0.0068, 0.0031, -0.0106, -0.0065, + 0.0063, 0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 218.74, cls_loss 0.0394 cls_loss_mapping 0.0515 cls_loss_causal 0.8938 re_mapping 0.0268 re_causal 0.0723 /// teacc 98.38 lr 0.00010000 +Epoch 29, weight, value: tensor([[ 0.0045, 0.0643, 0.0412, ..., -0.0673, -0.0435, 0.0254], + [ 0.0613, -0.0063, -0.0180, ..., 0.0251, 0.0359, -0.0577], + [-0.0258, -0.0062, -0.0269, ..., -0.0286, -0.0075, -0.0136], + ..., + [-0.0042, 0.0637, -0.0786, ..., -0.0431, 0.0094, -0.0533], + [-0.0029, -0.0347, 0.0226, ..., 0.0242, -0.0516, -0.0276], + [ 0.0040, -0.0291, 0.0150, ..., -0.0217, -0.0247, 0.0493]], + device='cuda:0'), grad: tensor([[ 1.1379e-04, -1.2283e-03, -1.2350e-03, ..., -2.2805e-04, + 6.3181e-05, -1.0834e-03], + [-2.5902e-03, -1.5068e-04, -4.2129e-04, ..., -1.1053e-03, + -1.3523e-03, 3.5548e-04], + [ 1.4172e-03, 6.7520e-04, 5.2738e-04, ..., 6.2418e-04, + 1.2420e-05, 7.2813e-04], + ..., + [ 4.0674e-04, 8.8930e-05, 3.2926e-04, ..., 1.8537e-04, + 2.1052e-04, 3.8338e-04], + [ 1.4410e-03, 2.1482e-04, 5.8174e-04, ..., 9.0981e-04, + 4.3321e-04, 4.8709e-04], + [-6.9737e-06, -4.3422e-05, -3.6716e-03, ..., 6.1703e-04, + 1.1218e-04, -3.2043e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0077, 0.0035, 0.0028, 0.0168, 0.0234, 0.0304, -0.0256, 0.0066, + -0.0297, -0.0174], device='cuda:0'), grad: tensor([-2.9716e-03, -3.2349e-03, 2.4872e-03, 3.4313e-03, -3.2429e-06, + 1.8711e-03, -2.1114e-03, 1.0519e-03, 1.9484e-03, -2.4719e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 218.66, cls_loss 0.0316 cls_loss_mapping 0.0431 cls_loss_causal 0.7947 re_mapping 0.0261 re_causal 0.0710 /// teacc 98.39 lr 0.00010000 +Epoch 30, weight, value: tensor([[ 0.0040, 0.0645, 0.0415, ..., -0.0676, -0.0439, 0.0254], + [ 0.0618, -0.0052, -0.0181, ..., 0.0255, 0.0362, -0.0585], + [-0.0257, -0.0064, -0.0268, ..., -0.0289, -0.0076, -0.0135], + ..., + [-0.0049, 0.0640, -0.0797, ..., -0.0437, 0.0097, -0.0542], + [-0.0035, -0.0356, 0.0227, ..., 0.0242, -0.0527, -0.0278], + [ 0.0040, -0.0288, 0.0149, ..., -0.0223, -0.0248, 0.0496]], + device='cuda:0'), grad: tensor([[-3.5137e-05, -1.0848e-04, -7.3195e-05, ..., 3.4660e-05, + 4.0054e-05, -8.4460e-05], + [ 2.9579e-05, -1.2308e-05, 2.2918e-05, ..., 1.5807e-04, + 5.3972e-05, 4.7356e-05], + [-4.3464e-04, -1.3262e-05, 1.2398e-05, ..., -2.4378e-04, + -5.8413e-04, 2.7508e-05], + ..., + [ 8.1062e-05, -1.7896e-05, 4.3213e-05, ..., 2.1577e-04, + 1.5163e-04, 5.0724e-05], + [-6.5446e-05, 2.3305e-05, -6.3419e-04, ..., -1.0853e-03, + -4.7684e-05, -6.3086e-04], + [ 5.7399e-05, 4.3839e-05, 1.8799e-04, ..., 3.3832e-04, + 1.0180e-04, 1.3852e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0074, 0.0034, 0.0029, 0.0167, 0.0236, 0.0310, -0.0258, 0.0066, + -0.0299, -0.0175], device='cuda:0'), grad: tensor([-4.0770e-05, 3.1471e-04, -2.3918e-03, 5.7364e-04, -3.2353e-04, + 5.0926e-04, 2.5368e-04, 5.4216e-04, 6.5565e-05, 4.9639e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 219.74, cls_loss 0.0331 cls_loss_mapping 0.0461 cls_loss_causal 0.8036 re_mapping 0.0244 re_causal 0.0668 /// teacc 98.42 lr 0.00010000 +Epoch 31, weight, value: tensor([[ 0.0035, 0.0650, 0.0417, ..., -0.0685, -0.0443, 0.0255], + [ 0.0628, -0.0056, -0.0189, ..., 0.0268, 0.0374, -0.0592], + [-0.0262, -0.0061, -0.0268, ..., -0.0298, -0.0079, -0.0138], + ..., + [-0.0057, 0.0648, -0.0807, ..., -0.0448, 0.0100, -0.0551], + [-0.0034, -0.0360, 0.0233, ..., 0.0246, -0.0533, -0.0278], + [ 0.0030, -0.0296, 0.0153, ..., -0.0229, -0.0250, 0.0502]], + device='cuda:0'), grad: tensor([[-9.3460e-05, -9.6619e-05, -5.5283e-05, ..., -3.2157e-05, + 8.0541e-06, -1.4734e-04], + [-1.9574e-04, 1.3494e-04, 1.4567e-04, ..., -1.3435e-04, + -5.5492e-05, 5.2303e-05], + [ 1.0145e-04, 2.1725e-03, 1.7262e-03, ..., 1.0467e-04, + -1.0812e-04, 3.0518e-05], + ..., + [ 5.5969e-05, -5.1613e-03, -3.7727e-03, ..., 1.0347e-04, + 5.2720e-05, 3.4660e-05], + [-2.0027e-04, 3.5429e-04, -3.5346e-05, ..., -4.4584e-04, + 3.8564e-05, -2.1338e-04], + [ 1.0240e-04, 2.0337e-04, 2.9659e-04, ..., 1.5402e-03, + 1.0677e-05, 1.0681e-03]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0074, 0.0034, 0.0031, 0.0164, 0.0234, 0.0312, -0.0258, 0.0068, + -0.0295, -0.0179], device='cuda:0'), grad: tensor([-6.3896e-05, 4.3225e-04, 6.6223e-03, 6.5002e-03, -2.4605e-03, + 6.2895e-04, 2.0218e-04, -1.5556e-02, 5.1355e-04, 3.1776e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 218.43, cls_loss 0.0352 cls_loss_mapping 0.0457 cls_loss_causal 0.8238 re_mapping 0.0239 re_causal 0.0642 /// teacc 98.36 lr 0.00010000 +Epoch 32, weight, value: tensor([[ 0.0034, 0.0656, 0.0420, ..., -0.0692, -0.0448, 0.0259], + [ 0.0630, -0.0055, -0.0194, ..., 0.0267, 0.0378, -0.0604], + [-0.0267, -0.0064, -0.0270, ..., -0.0303, -0.0076, -0.0139], + ..., + [-0.0070, 0.0654, -0.0820, ..., -0.0455, 0.0095, -0.0563], + [-0.0030, -0.0361, 0.0235, ..., 0.0252, -0.0538, -0.0280], + [ 0.0029, -0.0298, 0.0158, ..., -0.0232, -0.0253, 0.0510]], + device='cuda:0'), grad: tensor([[ 8.2627e-06, -8.9049e-05, -8.9228e-05, ..., 1.4074e-05, + 9.9838e-06, -9.8944e-05], + [-2.9488e-03, 1.8612e-05, 1.5028e-05, ..., -3.6716e-03, + -3.6812e-03, 1.2040e-05], + [ 2.7924e-03, 8.8453e-05, 1.9640e-05, ..., 3.5038e-03, + 3.4485e-03, 3.1948e-05], + ..., + [ 3.8207e-05, -1.4210e-04, 1.9002e-04, ..., 3.9577e-05, + 3.2485e-05, 1.3947e-04], + [ 1.0353e-04, 2.9549e-05, 1.5903e-04, ..., 9.7036e-05, + 7.0989e-05, 1.6773e-04], + [ 4.3809e-05, -2.9638e-05, -4.8089e-04, ..., 5.7928e-07, + 8.7172e-06, -3.7050e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0074, 0.0030, 0.0032, 0.0168, 0.0234, 0.0310, -0.0259, 0.0065, + -0.0293, -0.0178], device='cuda:0'), grad: tensor([-1.3328e-04, -4.5815e-03, 4.3068e-03, 2.7716e-05, 3.1161e-04, + 9.2328e-05, 4.7088e-05, 9.3162e-05, 3.2449e-04, -4.8804e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 218.53, cls_loss 0.0307 cls_loss_mapping 0.0406 cls_loss_causal 0.7918 re_mapping 0.0234 re_causal 0.0632 /// teacc 98.26 lr 0.00010000 +Epoch 33, weight, value: tensor([[ 0.0032, 0.0660, 0.0422, ..., -0.0698, -0.0454, 0.0260], + [ 0.0638, -0.0061, -0.0199, ..., 0.0271, 0.0393, -0.0611], + [-0.0277, -0.0066, -0.0269, ..., -0.0309, -0.0080, -0.0142], + ..., + [-0.0073, 0.0660, -0.0828, ..., -0.0456, 0.0101, -0.0568], + [-0.0028, -0.0368, 0.0237, ..., 0.0258, -0.0547, -0.0278], + [ 0.0025, -0.0298, 0.0165, ..., -0.0238, -0.0255, 0.0516]], + device='cuda:0'), grad: tensor([[ 2.4274e-05, -1.8358e-04, -1.2338e-04, ..., 2.2754e-05, + 6.1579e-06, -2.0647e-04], + [-2.0885e-04, 9.9465e-06, 5.4032e-05, ..., 1.3065e-04, + -1.2302e-04, 3.4302e-05], + [ 7.6175e-05, 2.2054e-05, 9.0003e-05, ..., 6.0141e-05, + 1.7554e-05, 9.6381e-05], + ..., + [ 1.3494e-04, -3.3522e-04, 6.8486e-05, ..., 5.1677e-05, + 6.8665e-05, 5.1945e-05], + [ 1.9300e-04, 2.6286e-05, 3.2663e-04, ..., 8.6248e-05, + 6.4015e-05, 3.8505e-04], + [-1.7285e-05, 2.9612e-04, -2.4166e-03, ..., 3.8838e-04, + 2.0516e-04, -1.7834e-03]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0075, 0.0031, 0.0032, 0.0163, 0.0234, 0.0308, -0.0258, 0.0069, + -0.0292, -0.0177], device='cuda:0'), grad: tensor([-2.8419e-04, 6.2287e-05, 1.1539e-04, 1.8969e-03, -1.8969e-03, + 3.6329e-05, 3.0994e-05, -1.5354e-04, 4.5967e-04, -2.6608e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 218.64, cls_loss 0.0322 cls_loss_mapping 0.0418 cls_loss_causal 0.8160 re_mapping 0.0227 re_causal 0.0610 /// teacc 98.24 lr 0.00010000 +Epoch 34, weight, value: tensor([[ 0.0031, 0.0662, 0.0421, ..., -0.0705, -0.0463, 0.0252], + [ 0.0640, -0.0061, -0.0207, ..., 0.0273, 0.0400, -0.0619], + [-0.0279, -0.0070, -0.0272, ..., -0.0315, -0.0083, -0.0149], + ..., + [-0.0079, 0.0670, -0.0836, ..., -0.0461, 0.0093, -0.0572], + [-0.0029, -0.0373, 0.0235, ..., 0.0262, -0.0554, -0.0285], + [ 0.0017, -0.0303, 0.0172, ..., -0.0240, -0.0259, 0.0528]], + device='cuda:0'), grad: tensor([[ 6.0558e-05, -1.8165e-05, 8.3745e-05, ..., 2.9728e-05, + 7.1377e-06, 2.0340e-05], + [-9.8884e-05, 4.9055e-05, 1.8671e-05, ..., -6.3837e-05, + -4.3064e-05, 9.1866e-06], + [ 8.0645e-05, 1.5748e-04, 2.7537e-04, ..., 1.5366e-04, + 2.0668e-05, 1.2469e-04], + ..., + [ 3.6478e-05, -8.7690e-04, 1.8287e-04, ..., 8.3327e-05, + -3.5107e-05, 1.3268e-04], + [-9.4557e-04, 6.4850e-05, -2.5349e-03, ..., -3.7718e-04, + 1.7598e-05, -1.5869e-03], + [ 4.9257e-04, 1.6963e-04, 1.3065e-03, ..., 2.9993e-04, + 1.7449e-05, 8.1110e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0070, 0.0029, 0.0031, 0.0167, 0.0233, 0.0310, -0.0255, 0.0069, + -0.0296, -0.0177], device='cuda:0'), grad: tensor([ 1.2600e-04, 9.1136e-05, 7.2098e-04, 2.4080e-04, 9.2804e-05, + 2.1470e-04, 2.4533e-04, -1.1606e-03, -2.5730e-03, 2.0027e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 218.62, cls_loss 0.0356 cls_loss_mapping 0.0402 cls_loss_causal 0.7848 re_mapping 0.0237 re_causal 0.0591 /// teacc 98.37 lr 0.00010000 +Epoch 35, weight, value: tensor([[ 0.0025, 0.0667, 0.0423, ..., -0.0713, -0.0475, 0.0252], + [ 0.0645, -0.0067, -0.0207, ..., 0.0275, 0.0415, -0.0626], + [-0.0287, -0.0073, -0.0273, ..., -0.0322, -0.0090, -0.0154], + ..., + [-0.0085, 0.0682, -0.0849, ..., -0.0469, 0.0099, -0.0581], + [-0.0021, -0.0380, 0.0237, ..., 0.0271, -0.0559, -0.0282], + [ 0.0014, -0.0307, 0.0176, ..., -0.0247, -0.0265, 0.0536]], + device='cuda:0'), grad: tensor([[ 4.2367e-04, 2.9063e-04, 1.2755e-04, ..., 3.2449e-04, + 2.8834e-06, 4.5490e-04], + [ 4.0889e-05, 2.9445e-04, 2.5392e-04, ..., 1.0020e-04, + 4.5747e-06, 2.0993e-04], + [ 2.3574e-05, -1.8921e-03, 4.1664e-05, ..., 2.9624e-05, + -3.6812e-04, 2.7224e-05], + ..., + [ 1.9684e-05, 3.5686e-03, 2.0123e-03, ..., 5.1022e-04, + 3.3069e-04, 1.7958e-03], + [-1.0133e-04, 1.8060e-04, 1.2374e-04, ..., -1.5974e-04, + 3.2093e-06, 2.0695e-04], + [-1.4573e-05, -2.6340e-03, -2.8706e-03, ..., -5.9128e-04, + 3.2019e-06, -2.6112e-03]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0068, 0.0029, 0.0030, 0.0167, 0.0238, 0.0309, -0.0262, 0.0070, + -0.0292, -0.0177], device='cuda:0'), grad: tensor([ 0.0010, 0.0013, -0.0041, 0.0005, 0.0009, -0.0002, -0.0007, 0.0137, + 0.0004, -0.0128], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 218.48, cls_loss 0.0307 cls_loss_mapping 0.0392 cls_loss_causal 0.7970 re_mapping 0.0231 re_causal 0.0635 /// teacc 98.41 lr 0.00010000 +Epoch 36, weight, value: tensor([[ 0.0026, 0.0678, 0.0428, ..., -0.0720, -0.0482, 0.0257], + [ 0.0649, -0.0071, -0.0212, ..., 0.0278, 0.0420, -0.0631], + [-0.0292, -0.0075, -0.0277, ..., -0.0328, -0.0093, -0.0160], + ..., + [-0.0089, 0.0684, -0.0862, ..., -0.0474, 0.0109, -0.0589], + [-0.0022, -0.0385, 0.0238, ..., 0.0274, -0.0566, -0.0284], + [ 0.0008, -0.0305, 0.0178, ..., -0.0253, -0.0263, 0.0536]], + device='cuda:0'), grad: tensor([[ 3.9965e-05, 1.2743e-04, 2.9847e-05, ..., 1.9360e-04, + 3.8967e-06, 3.5226e-05], + [-2.7478e-05, 2.0429e-05, 5.0247e-05, ..., 7.5400e-05, + -1.7613e-05, 1.3098e-05], + [ 2.2352e-05, 2.2620e-05, 1.7941e-05, ..., 7.3791e-05, + -1.6633e-06, 5.0515e-06], + ..., + [ 1.5542e-05, -8.1003e-05, 5.7817e-05, ..., 1.0389e-04, + 4.1239e-06, 1.6868e-05], + [ 8.5294e-05, 2.4036e-05, 1.6952e-04, ..., 1.7142e-04, + 1.0967e-05, 1.0186e-04], + [ 4.7386e-05, 6.0201e-05, 8.1730e-04, ..., 1.5917e-03, + 2.8573e-06, -3.0309e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0074, 0.0027, 0.0029, 0.0166, 0.0238, 0.0318, -0.0261, 0.0064, + -0.0295, -0.0177], device='cuda:0'), grad: tensor([ 3.3855e-04, 1.8120e-04, 8.0913e-06, -6.4373e-06, -5.7449e-03, + -5.1594e-04, 2.1839e-03, 1.2255e-04, 4.3869e-04, 2.9964e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 218.10, cls_loss 0.0274 cls_loss_mapping 0.0356 cls_loss_causal 0.7697 re_mapping 0.0220 re_causal 0.0590 /// teacc 98.36 lr 0.00010000 +Epoch 37, weight, value: tensor([[ 0.0025, 0.0683, 0.0430, ..., -0.0727, -0.0484, 0.0260], + [ 0.0656, -0.0065, -0.0216, ..., 0.0288, 0.0421, -0.0634], + [-0.0297, -0.0078, -0.0277, ..., -0.0334, -0.0087, -0.0162], + ..., + [-0.0097, 0.0696, -0.0862, ..., -0.0486, 0.0109, -0.0591], + [-0.0021, -0.0394, 0.0239, ..., 0.0277, -0.0568, -0.0286], + [ 0.0002, -0.0313, 0.0178, ..., -0.0260, -0.0266, 0.0541]], + device='cuda:0'), grad: tensor([[-1.6451e-05, -2.3377e-04, 1.1355e-05, ..., -1.0777e-04, + 8.7380e-05, -2.8682e-04], + [ 2.1946e-04, 2.0254e-04, 2.2805e-04, ..., 2.0730e-04, + 9.6738e-05, 6.2361e-06], + [-1.6470e-03, -2.9370e-05, -8.6212e-03, ..., 4.2468e-05, + -6.3820e-03, 9.6709e-06], + ..., + [-2.0635e-04, -1.2732e-03, -4.9210e-04, ..., 1.4198e-04, + -2.0098e-06, 5.2452e-05], + [ 1.7905e-04, 8.5771e-05, 1.3089e-04, ..., 2.1112e-04, + 3.2932e-05, 4.8310e-05], + [ 7.9012e-04, 6.9332e-04, 1.0748e-03, ..., 1.9798e-03, + 9.8765e-05, 2.3007e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0074, 0.0030, 0.0028, 0.0166, 0.0236, 0.0317, -0.0259, 0.0069, + -0.0296, -0.0180], device='cuda:0'), grad: tensor([-7.6413e-05, 1.0233e-03, -1.6785e-02, 1.6418e-02, -4.9248e-03, + 6.7854e-04, 2.6450e-05, -1.9798e-03, 6.5660e-04, 4.9744e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 219.42, cls_loss 0.0297 cls_loss_mapping 0.0412 cls_loss_causal 0.7410 re_mapping 0.0203 re_causal 0.0544 /// teacc 98.37 lr 0.00010000 +Epoch 38, weight, value: tensor([[ 0.0021, 0.0687, 0.0430, ..., -0.0734, -0.0503, 0.0260], + [ 0.0662, -0.0068, -0.0221, ..., 0.0294, 0.0422, -0.0641], + [-0.0298, -0.0082, -0.0272, ..., -0.0340, -0.0083, -0.0165], + ..., + [-0.0104, 0.0705, -0.0867, ..., -0.0493, 0.0112, -0.0596], + [-0.0018, -0.0398, 0.0239, ..., 0.0283, -0.0572, -0.0285], + [-0.0003, -0.0315, 0.0183, ..., -0.0269, -0.0257, 0.0548]], + device='cuda:0'), grad: tensor([[-4.3821e-04, -1.0414e-03, -7.4291e-04, ..., -1.4734e-04, + 3.1143e-06, -5.2309e-04], + [ 2.6211e-05, 2.1353e-05, 5.8860e-06, ..., 5.8115e-05, + 2.6405e-05, 6.1803e-06], + [ 3.4183e-05, 3.6716e-04, 5.6237e-05, ..., 5.9396e-05, + -5.4777e-05, 5.2154e-05], + ..., + [ 1.6227e-05, -4.7588e-04, 5.1886e-05, ..., 1.1265e-04, + 2.1428e-05, 7.1824e-05], + [-7.9036e-05, 7.8619e-05, 7.1704e-05, ..., 1.5306e-04, + 2.4110e-05, 1.3530e-04], + [ 1.6108e-05, 2.4176e-04, -1.7190e-04, ..., -3.3112e-03, + 3.7819e-05, -3.8218e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0073, 0.0028, 0.0031, 0.0166, 0.0238, 0.0318, -0.0263, 0.0069, + -0.0296, -0.0181], device='cuda:0'), grad: tensor([-0.0019, 0.0003, 0.0002, 0.0007, 0.0045, -0.0004, 0.0020, -0.0004, + 0.0005, -0.0055], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 219.45, cls_loss 0.0286 cls_loss_mapping 0.0346 cls_loss_causal 0.7683 re_mapping 0.0214 re_causal 0.0582 /// teacc 98.52 lr 0.00010000 +Epoch 39, weight, value: tensor([[ 0.0019, 0.0691, 0.0431, ..., -0.0738, -0.0508, 0.0259], + [ 0.0672, -0.0072, -0.0213, ..., 0.0302, 0.0437, -0.0641], + [-0.0304, -0.0085, -0.0273, ..., -0.0345, -0.0093, -0.0166], + ..., + [-0.0109, 0.0718, -0.0875, ..., -0.0501, 0.0112, -0.0602], + [-0.0027, -0.0404, 0.0237, ..., 0.0282, -0.0580, -0.0289], + [-0.0002, -0.0321, 0.0189, ..., -0.0269, -0.0254, 0.0556]], + device='cuda:0'), grad: tensor([[ 4.1336e-05, -6.5947e-04, -5.8031e-04, ..., -4.5204e-04, + 1.5453e-05, -9.8515e-04], + [-7.1585e-05, 2.7306e-06, 1.7464e-05, ..., -3.6657e-05, + 2.9966e-05, 2.5719e-05], + [ 1.6019e-05, 7.4744e-05, 6.0052e-05, ..., 5.5790e-05, + -1.5545e-04, 1.0175e-04], + ..., + [ 2.9862e-05, 4.9323e-05, 8.3566e-05, ..., 1.4782e-04, + 5.5462e-05, 1.2022e-04], + [ 4.2021e-05, 4.3809e-05, 9.5308e-05, ..., 9.2983e-05, + 9.5144e-06, 1.5080e-04], + [ 2.7731e-05, 3.0732e-04, 5.1111e-06, ..., 2.3484e-04, + 1.6198e-05, 1.5485e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0070, 0.0034, 0.0032, 0.0167, 0.0236, 0.0320, -0.0268, 0.0071, + -0.0302, -0.0178], device='cuda:0'), grad: tensor([-1.6642e-03, 8.8453e-05, -1.3864e-04, 7.1859e-04, -1.2898e-04, + 5.7459e-04, -5.7602e-04, 3.5071e-04, 3.0899e-04, 4.6635e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 218.12, cls_loss 0.0238 cls_loss_mapping 0.0290 cls_loss_causal 0.7601 re_mapping 0.0200 re_causal 0.0557 /// teacc 98.32 lr 0.00010000 +Epoch 40, weight, value: tensor([[ 0.0007, 0.0692, 0.0437, ..., -0.0752, -0.0516, 0.0259], + [ 0.0668, -0.0073, -0.0216, ..., 0.0301, 0.0438, -0.0648], + [-0.0306, -0.0088, -0.0276, ..., -0.0348, -0.0096, -0.0169], + ..., + [-0.0119, 0.0728, -0.0884, ..., -0.0510, 0.0111, -0.0611], + [-0.0024, -0.0407, 0.0239, ..., 0.0288, -0.0587, -0.0288], + [-0.0007, -0.0327, 0.0188, ..., -0.0272, -0.0257, 0.0558]], + device='cuda:0'), grad: tensor([[ 3.0398e-05, -2.0429e-05, 7.2908e-04, ..., 1.4687e-04, + 2.9374e-06, 6.4230e-04], + [-6.2799e-04, 1.2871e-06, 4.5866e-05, ..., -5.1355e-04, + -1.7190e-04, 3.1263e-05], + [ 1.7118e-04, 6.6906e-06, 3.0732e-04, ..., 1.4853e-04, + 4.1008e-05, 2.6202e-04], + ..., + [ 5.0664e-05, -6.0499e-06, 3.9577e-05, ..., 3.3826e-05, + 1.0937e-05, 3.2306e-05], + [ 6.3896e-04, 2.0396e-06, 3.4561e-03, ..., 8.9884e-04, + 2.9922e-05, 3.0060e-03], + [ 9.0897e-05, 6.0014e-06, -4.5319e-03, ..., -7.8773e-04, + 6.4075e-06, -4.0207e-03]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0071, 0.0029, 0.0032, 0.0171, 0.0236, 0.0319, -0.0261, 0.0071, + -0.0302, -0.0182], device='cuda:0'), grad: tensor([ 0.0008, -0.0005, 0.0002, -0.0034, 0.0002, 0.0027, 0.0003, 0.0002, + 0.0042, -0.0047], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 218.26, cls_loss 0.0254 cls_loss_mapping 0.0353 cls_loss_causal 0.7439 re_mapping 0.0200 re_causal 0.0565 /// teacc 98.34 lr 0.00010000 +Epoch 41, weight, value: tensor([[ 0.0004, 0.0694, 0.0438, ..., -0.0758, -0.0519, 0.0258], + [ 0.0671, -0.0058, -0.0218, ..., 0.0300, 0.0446, -0.0653], + [-0.0312, -0.0091, -0.0277, ..., -0.0352, -0.0099, -0.0172], + ..., + [-0.0133, 0.0732, -0.0890, ..., -0.0514, 0.0109, -0.0615], + [-0.0021, -0.0414, 0.0238, ..., 0.0292, -0.0589, -0.0291], + [-0.0015, -0.0337, 0.0195, ..., -0.0276, -0.0256, 0.0569]], + device='cuda:0'), grad: tensor([[ 1.6623e-03, 2.2755e-03, -2.2149e-04, ..., 2.9755e-03, + 4.3400e-06, -1.4400e-03], + [-2.0504e-04, 5.5218e-04, 1.3268e-04, ..., -2.8586e-04, + -1.5366e-04, 2.2089e-04], + [ 1.7369e-04, -1.9002e-04, 1.3673e-04, ..., 1.5283e-04, + 2.2128e-05, 8.7738e-05], + ..., + [-1.1139e-03, -2.1496e-03, -4.1097e-05, ..., -1.3262e-05, + 1.5900e-05, -5.7697e-04], + [-1.8425e-03, -2.5978e-03, -4.3440e-04, ..., -3.4637e-03, + 2.8044e-05, 1.1530e-03], + [ 5.9700e-04, 1.3371e-03, -5.6171e-04, ..., -7.4863e-05, + 1.3992e-05, -9.6917e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0066, 0.0031, 0.0033, 0.0177, 0.0234, 0.0315, -0.0257, 0.0067, + -0.0303, -0.0181], device='cuda:0'), grad: tensor([ 0.0027, 0.0008, -0.0011, 0.0003, 0.0016, 0.0003, 0.0008, -0.0035, + -0.0035, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 219.06, cls_loss 0.0203 cls_loss_mapping 0.0270 cls_loss_causal 0.7160 re_mapping 0.0194 re_causal 0.0542 /// teacc 98.54 lr 0.00010000 +Epoch 42, weight, value: tensor([[ 0.0002, 0.0698, 0.0439, ..., -0.0765, -0.0522, 0.0256], + [ 0.0681, -0.0053, -0.0219, ..., 0.0308, 0.0452, -0.0657], + [-0.0314, -0.0092, -0.0278, ..., -0.0357, -0.0100, -0.0173], + ..., + [-0.0140, 0.0736, -0.0898, ..., -0.0525, 0.0110, -0.0623], + [-0.0024, -0.0421, 0.0237, ..., 0.0293, -0.0603, -0.0293], + [-0.0023, -0.0341, 0.0199, ..., -0.0280, -0.0260, 0.0576]], + device='cuda:0'), grad: tensor([[-3.6061e-05, -3.3927e-04, -6.3598e-05, ..., 1.3605e-05, + 1.1340e-05, -1.8716e-04], + [ 2.7370e-04, 6.5982e-05, 1.5354e-03, ..., 7.5519e-05, + 2.5654e-04, 1.2517e-05], + [-5.7650e-04, 2.3899e-03, -2.1992e-03, ..., -1.3363e-04, + -5.3024e-04, 3.3796e-05], + ..., + [ 1.5318e-04, -2.8458e-03, 3.0017e-04, ..., 6.1691e-05, + 9.1791e-05, 1.4022e-05], + [-6.3038e-04, 8.7678e-05, -3.5453e-04, ..., -7.0953e-04, + 5.3316e-05, 2.8446e-05], + [ 1.0175e-04, 2.6917e-04, 2.0421e-04, ..., 7.8440e-05, + 5.5045e-05, 6.8784e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0065, 0.0037, 0.0034, 0.0174, 0.0235, 0.0318, -0.0256, 0.0066, + -0.0309, -0.0179], device='cuda:0'), grad: tensor([-0.0006, 0.0046, -0.0022, 0.0011, 0.0003, 0.0002, 0.0001, -0.0045, + -0.0003, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 217.77, cls_loss 0.0242 cls_loss_mapping 0.0313 cls_loss_causal 0.7690 re_mapping 0.0187 re_causal 0.0536 /// teacc 98.39 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0003, 0.0705, 0.0442, ..., -0.0770, -0.0524, 0.0258], + [ 0.0681, -0.0058, -0.0223, ..., 0.0304, 0.0455, -0.0661], + [-0.0312, -0.0098, -0.0278, ..., -0.0353, -0.0096, -0.0176], + ..., + [-0.0141, 0.0743, -0.0905, ..., -0.0530, 0.0113, -0.0629], + [-0.0021, -0.0422, 0.0239, ..., 0.0300, -0.0609, -0.0294], + [-0.0025, -0.0342, 0.0200, ..., -0.0286, -0.0262, 0.0578]], + device='cuda:0'), grad: tensor([[ 1.6403e-04, 4.2707e-05, 8.7559e-05, ..., 1.3316e-04, + 3.0790e-06, 5.4091e-06], + [ 1.0222e-04, 2.7448e-05, 8.5115e-05, ..., 1.3530e-04, + -7.5810e-06, 1.2055e-05], + [ 1.8430e-04, -8.0585e-05, 2.4408e-05, ..., 1.8322e-04, + -1.8448e-05, 6.3851e-06], + ..., + [-3.8445e-06, -1.6379e-04, 2.7955e-05, ..., 8.7321e-05, + 4.2394e-06, 7.8678e-06], + [ 1.2070e-04, 1.1027e-04, -8.0395e-04, ..., -1.5335e-03, + 3.9935e-06, -4.1986e-04], + [ 1.6892e-04, 1.3876e-04, -1.7178e-04, ..., 1.1339e-03, + 1.8388e-05, -1.7107e-04]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0068, 0.0032, 0.0034, 0.0173, 0.0240, 0.0316, -0.0262, 0.0067, + -0.0306, -0.0179], device='cuda:0'), grad: tensor([ 3.9792e-04, 3.2473e-04, -9.0301e-05, 1.5354e-03, -6.2180e-04, + 6.3181e-04, -3.3607e-03, -1.8671e-05, -3.3212e-04, 1.5345e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 217.79, cls_loss 0.0241 cls_loss_mapping 0.0309 cls_loss_causal 0.7237 re_mapping 0.0188 re_causal 0.0513 /// teacc 98.58 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0007, 0.0715, 0.0444, ..., -0.0777, -0.0531, 0.0257], + [ 0.0690, -0.0062, -0.0220, ..., 0.0311, 0.0460, -0.0665], + [-0.0323, -0.0102, -0.0281, ..., -0.0361, -0.0099, -0.0180], + ..., + [-0.0146, 0.0750, -0.0909, ..., -0.0537, 0.0111, -0.0630], + [-0.0022, -0.0428, 0.0237, ..., 0.0303, -0.0619, -0.0297], + [-0.0035, -0.0344, 0.0201, ..., -0.0294, -0.0264, 0.0583]], + device='cuda:0'), grad: tensor([[ 3.9482e-04, -1.2755e-04, 6.7329e-04, ..., 1.1170e-04, + 1.2368e-05, 9.6560e-05], + [ 6.8605e-05, 6.0424e-06, 8.4996e-05, ..., 5.1528e-05, + -7.6741e-06, 4.6611e-05], + [ 5.5599e-04, 4.3392e-05, 9.5940e-04, ..., 2.9874e-04, + 1.5274e-05, 1.7571e-04], + ..., + [ 9.4414e-05, 2.6926e-05, 1.8644e-04, ..., 1.3864e-04, + 1.2755e-05, 1.4472e-04], + [-4.7569e-03, -2.0087e-04, 6.6376e-04, ..., -3.5419e-03, + 1.0639e-04, -2.3537e-03], + [ 1.5545e-04, -5.4687e-05, -5.4455e-04, ..., -1.0706e-05, + 7.0110e-06, -5.9223e-04]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0070, 0.0034, 0.0033, 0.0171, 0.0239, 0.0326, -0.0261, 0.0065, + -0.0309, -0.0182], device='cuda:0'), grad: tensor([ 0.0009, 0.0002, 0.0016, -0.0037, 0.0004, 0.0043, 0.0001, 0.0005, + -0.0041, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 43---------------------------------------------------- +epoch 43, time 217.75, cls_loss 0.0242 cls_loss_mapping 0.0309 cls_loss_causal 0.7113 re_mapping 0.0189 re_causal 0.0505 /// teacc 98.64 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0021, 0.0717, 0.0450, ..., -0.0787, -0.0539, 0.0257], + [ 0.0698, -0.0058, -0.0223, ..., 0.0315, 0.0463, -0.0666], + [-0.0323, -0.0107, -0.0282, ..., -0.0360, -0.0097, -0.0182], + ..., + [-0.0154, 0.0760, -0.0918, ..., -0.0546, 0.0116, -0.0637], + [-0.0024, -0.0434, 0.0242, ..., 0.0307, -0.0633, -0.0294], + [-0.0040, -0.0349, 0.0208, ..., -0.0299, -0.0269, 0.0592]], + device='cuda:0'), grad: tensor([[ 9.9763e-06, -7.0214e-05, -4.3809e-05, ..., 8.0407e-05, + 5.6066e-06, -5.5522e-05], + [ 9.1270e-08, 4.2059e-06, 6.7204e-06, ..., 6.2466e-05, + -2.3935e-07, 5.1036e-06], + [ 5.1670e-06, 1.1794e-05, 1.1558e-06, ..., 8.6308e-05, + 5.5991e-06, 1.1981e-05], + ..., + [ 2.8219e-06, -2.9981e-05, 4.8168e-06, ..., 3.4571e-05, + 1.4566e-05, 6.3069e-06], + [ 5.4270e-05, 2.0176e-05, 2.8268e-05, ..., 1.3888e-04, + 3.2723e-05, 7.4267e-05], + [ 6.4895e-06, 3.0607e-05, -6.2287e-05, ..., 9.6440e-05, + 3.1263e-05, -6.9559e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0068, 0.0033, 0.0034, 0.0168, 0.0246, 0.0323, -0.0260, 0.0069, + -0.0308, -0.0187], device='cuda:0'), grad: tensor([ 2.1830e-05, 1.0622e-04, 1.1861e-04, 2.0683e-05, -6.6109e-03, + 2.7895e-04, 5.6572e-03, 3.1590e-05, 2.7728e-04, 9.7513e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 217.05, cls_loss 0.0227 cls_loss_mapping 0.0265 cls_loss_causal 0.7031 re_mapping 0.0179 re_causal 0.0505 /// teacc 98.39 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0031, 0.0720, 0.0453, ..., -0.0793, -0.0552, 0.0255], + [ 0.0702, -0.0061, -0.0226, ..., 0.0318, 0.0465, -0.0670], + [-0.0330, -0.0108, -0.0285, ..., -0.0364, -0.0096, -0.0187], + ..., + [-0.0148, 0.0763, -0.0928, ..., -0.0545, 0.0126, -0.0639], + [-0.0027, -0.0439, 0.0243, ..., 0.0306, -0.0647, -0.0295], + [-0.0043, -0.0351, 0.0215, ..., -0.0301, -0.0273, 0.0601]], + device='cuda:0'), grad: tensor([[-9.3699e-05, -1.5771e-04, 4.3362e-05, ..., 3.6061e-05, + 7.5884e-06, 1.0085e-04], + [-2.4706e-05, 2.0728e-05, 4.1038e-05, ..., -5.0962e-05, + -3.2157e-05, 2.3067e-05], + [ 1.4770e-04, 1.1049e-05, 1.8990e-04, ..., 3.4958e-05, + -2.8443e-04, 4.0203e-05], + ..., + [ 1.3876e-04, -8.1182e-05, 2.3961e-04, ..., 1.2279e-04, + 1.8895e-04, 2.0278e-04], + [-3.1948e-04, 2.5094e-05, -2.9659e-04, ..., -4.5109e-04, + 1.3255e-05, -1.9896e-04], + [ 2.4092e-04, 4.8041e-05, -1.6689e-06, ..., 1.6201e-04, + 1.2256e-05, -9.1195e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0067, 0.0031, 0.0032, 0.0170, 0.0242, 0.0328, -0.0260, 0.0070, + -0.0312, -0.0183], device='cuda:0'), grad: tensor([-7.9125e-06, 2.9355e-05, -2.5481e-05, 9.0218e-04, 1.8322e-04, + -1.1930e-03, 1.3006e-04, 5.7697e-04, -7.7152e-04, 1.7452e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 216.97, cls_loss 0.0265 cls_loss_mapping 0.0345 cls_loss_causal 0.7272 re_mapping 0.0184 re_causal 0.0503 /// teacc 98.41 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0036, 0.0726, 0.0454, ..., -0.0800, -0.0557, 0.0250], + [ 0.0705, -0.0072, -0.0231, ..., 0.0319, 0.0469, -0.0675], + [-0.0329, -0.0106, -0.0278, ..., -0.0371, -0.0086, -0.0188], + ..., + [-0.0143, 0.0770, -0.0941, ..., -0.0542, 0.0126, -0.0647], + [-0.0020, -0.0447, 0.0244, ..., 0.0317, -0.0658, -0.0289], + [-0.0044, -0.0345, 0.0220, ..., -0.0308, -0.0282, 0.0609]], + device='cuda:0'), grad: tensor([[ 1.3731e-05, -3.6992e-06, 7.3314e-06, ..., 2.9981e-05, + 1.4082e-05, 6.9141e-06], + [ 4.3917e-04, -6.3516e-07, 7.6056e-04, ..., 4.1275e-03, + 1.9321e-03, 1.9484e-03], + [ 2.8819e-05, 2.2769e-04, 7.6592e-05, ..., 1.1462e-04, + 5.4300e-05, 1.7732e-05], + ..., + [ 1.1271e-04, -4.1676e-04, 7.0751e-05, ..., 3.9244e-04, + 1.8013e-04, 1.8811e-04], + [ 6.9737e-05, 1.3813e-05, 3.6627e-05, ..., 7.1347e-05, + 3.4034e-05, 4.0054e-05], + [-7.6962e-04, -4.7028e-05, -1.1253e-03, ..., -5.3978e-03, + -2.6531e-03, -2.6207e-03]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0062, 0.0026, 0.0042, 0.0163, 0.0244, 0.0328, -0.0269, 0.0068, + -0.0308, -0.0177], device='cuda:0'), grad: tensor([ 5.7906e-05, 8.2474e-03, 3.2401e-04, 3.4976e-04, 1.3905e-03, + -1.0359e-04, 7.1108e-05, 3.6979e-04, 2.2626e-04, -1.0925e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 217.48, cls_loss 0.0184 cls_loss_mapping 0.0263 cls_loss_causal 0.7013 re_mapping 0.0177 re_causal 0.0504 /// teacc 98.45 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0034, 0.0732, 0.0463, ..., -0.0804, -0.0566, 0.0258], + [ 0.0708, -0.0069, -0.0231, ..., 0.0326, 0.0476, -0.0680], + [-0.0334, -0.0110, -0.0280, ..., -0.0378, -0.0084, -0.0188], + ..., + [-0.0148, 0.0778, -0.0949, ..., -0.0555, 0.0122, -0.0653], + [-0.0024, -0.0450, 0.0243, ..., 0.0317, -0.0681, -0.0290], + [-0.0045, -0.0350, 0.0219, ..., -0.0314, -0.0287, 0.0610]], + device='cuda:0'), grad: tensor([[ 3.1567e-04, 1.3709e-04, 4.5896e-05, ..., 1.1444e-04, + 5.1856e-06, 7.6830e-05], + [-4.6372e-05, 1.8731e-05, 2.8417e-05, ..., 4.1239e-06, + -6.7335e-07, 1.0237e-05], + [-1.7762e-05, 1.9163e-05, 1.7226e-05, ..., 2.5421e-05, + -5.3138e-05, 9.7007e-06], + ..., + [ 1.7107e-05, -1.4901e-04, 1.1146e-05, ..., 2.3648e-05, + -3.5405e-05, 5.0329e-06], + [ 6.4015e-05, 2.8819e-05, 5.2214e-05, ..., 8.7082e-05, + 2.2888e-05, 4.7803e-05], + [ 2.2799e-05, 6.9261e-05, 4.6968e-04, ..., 1.1606e-03, + 5.5283e-05, 1.3041e-04]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0068, 0.0029, 0.0041, 0.0164, 0.0251, 0.0326, -0.0269, 0.0066, + -0.0311, -0.0180], device='cuda:0'), grad: tensor([ 3.6931e-04, 1.1069e-04, -4.9233e-05, 2.0361e-04, -2.0313e-03, + -3.6806e-05, -4.2701e-04, -2.3913e-04, 1.9526e-04, 1.9035e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 216.97, cls_loss 0.0158 cls_loss_mapping 0.0218 cls_loss_causal 0.7106 re_mapping 0.0174 re_causal 0.0497 /// teacc 98.54 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0029, 0.0739, 0.0466, ..., -0.0809, -0.0570, 0.0260], + [ 0.0708, -0.0076, -0.0231, ..., 0.0329, 0.0483, -0.0685], + [-0.0340, -0.0111, -0.0283, ..., -0.0384, -0.0084, -0.0190], + ..., + [-0.0148, 0.0789, -0.0952, ..., -0.0554, 0.0119, -0.0658], + [-0.0023, -0.0455, 0.0244, ..., 0.0321, -0.0684, -0.0292], + [-0.0051, -0.0355, 0.0223, ..., -0.0317, -0.0289, 0.0615]], + device='cuda:0'), grad: tensor([[ 5.5194e-05, 3.0041e-05, 3.7313e-05, ..., 4.7714e-05, + 9.9987e-06, 5.0664e-05], + [-1.7524e-04, -4.5151e-05, 4.4644e-05, ..., -2.7999e-05, + -1.1241e-04, 1.4521e-05], + [ 7.4267e-05, 3.7044e-05, 2.4602e-05, ..., 7.3373e-05, + 2.9132e-05, 1.2510e-05], + ..., + [ 1.8135e-05, -1.5945e-03, -4.3464e-04, ..., 7.0870e-05, + -5.6058e-05, -5.5504e-04], + [ 5.5656e-06, 4.5627e-05, 5.7161e-05, ..., -6.4313e-05, + 4.9621e-06, 1.4973e-04], + [ 1.7381e-04, 1.4324e-03, 8.6594e-04, ..., 6.5041e-04, + 7.8201e-05, 6.6280e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0071, 0.0026, 0.0041, 0.0161, 0.0247, 0.0328, -0.0268, 0.0069, + -0.0311, -0.0180], device='cuda:0'), grad: tensor([ 1.5402e-04, -5.0128e-05, 4.0203e-05, 6.3515e-04, -1.1225e-03, + -5.4693e-04, -4.0483e-04, -1.9951e-03, 1.9336e-04, 3.0975e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 217.41, cls_loss 0.0186 cls_loss_mapping 0.0244 cls_loss_causal 0.6991 re_mapping 0.0162 re_causal 0.0460 /// teacc 98.57 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0030, 0.0747, 0.0466, ..., -0.0816, -0.0569, 0.0259], + [ 0.0712, -0.0077, -0.0233, ..., 0.0333, 0.0493, -0.0689], + [-0.0344, -0.0114, -0.0282, ..., -0.0390, -0.0079, -0.0191], + ..., + [-0.0154, 0.0795, -0.0958, ..., -0.0560, 0.0117, -0.0663], + [-0.0020, -0.0460, 0.0243, ..., 0.0326, -0.0688, -0.0296], + [-0.0057, -0.0360, 0.0228, ..., -0.0322, -0.0297, 0.0623]], + device='cuda:0'), grad: tensor([[ 6.4850e-05, -4.9993e-06, 2.7046e-05, ..., 4.6968e-05, + 1.5080e-05, 2.6837e-05], + [-1.4663e-04, 3.7700e-05, 2.5511e-05, ..., -1.9968e-04, + -7.0930e-05, 2.0057e-05], + [ 8.3208e-05, 2.4283e-04, 3.3528e-05, ..., 8.3685e-05, + 8.4162e-05, 9.5367e-06], + ..., + [ 4.3005e-05, -1.1797e-03, -2.6989e-04, ..., 4.5419e-05, + -1.8084e-04, 1.8567e-05], + [ 3.4881e-04, 3.3170e-05, 2.1935e-04, ..., 1.3721e-04, + 1.3769e-04, 1.9288e-04], + [ 8.1241e-05, 8.9407e-05, 8.7857e-05, ..., 5.9247e-05, + 3.6210e-05, 5.7369e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0069, 0.0027, 0.0048, 0.0160, 0.0246, 0.0328, -0.0265, 0.0067, + -0.0313, -0.0182], device='cuda:0'), grad: tensor([ 1.1390e-04, -2.4930e-05, 7.5102e-04, 2.6703e-03, 9.5248e-05, + -1.2064e-03, -1.0401e-04, -3.2883e-03, 5.8270e-04, 4.1199e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 49---------------------------------------------------- +epoch 49, time 217.63, cls_loss 0.0230 cls_loss_mapping 0.0288 cls_loss_causal 0.7104 re_mapping 0.0164 re_causal 0.0465 /// teacc 98.75 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0041, 0.0750, 0.0473, ..., -0.0827, -0.0594, 0.0258], + [ 0.0714, -0.0069, -0.0243, ..., 0.0332, 0.0504, -0.0691], + [-0.0341, -0.0124, -0.0285, ..., -0.0384, -0.0081, -0.0198], + ..., + [-0.0148, 0.0802, -0.0960, ..., -0.0560, 0.0123, -0.0668], + [-0.0022, -0.0464, 0.0246, ..., 0.0331, -0.0699, -0.0298], + [-0.0056, -0.0364, 0.0234, ..., -0.0327, -0.0303, 0.0634]], + device='cuda:0'), grad: tensor([[-7.7784e-05, -6.6459e-05, -1.2422e-04, ..., 1.7035e-04, + 5.6550e-06, -1.5748e-04], + [-3.3528e-06, 1.8239e-04, 3.8773e-05, ..., 1.0747e-06, + -2.8312e-05, 2.8566e-05], + [ 4.6253e-05, 1.4794e-04, 1.4722e-04, ..., 4.5180e-05, + -1.6270e-06, 1.8120e-05], + ..., + [ 3.2336e-05, -1.2178e-03, -5.4121e-04, ..., 3.1471e-05, + 9.9018e-06, -1.6618e-04], + [-1.0699e-05, -1.4797e-05, -1.4460e-04, ..., 5.2899e-06, + 1.9029e-05, 4.0472e-05], + [ 1.8612e-05, 1.0624e-03, 5.6887e-04, ..., 1.1921e-04, + 3.5614e-06, 2.5177e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0068, 0.0024, 0.0044, 0.0157, 0.0242, 0.0327, -0.0264, 0.0077, + -0.0314, -0.0178], device='cuda:0'), grad: tensor([-4.6939e-05, 2.6488e-04, 7.3528e-04, 2.2638e-04, -4.7073e-03, + 1.0324e-04, 4.0016e-03, -2.3308e-03, -5.5933e-04, 2.3117e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 217.45, cls_loss 0.0172 cls_loss_mapping 0.0253 cls_loss_causal 0.7170 re_mapping 0.0156 re_causal 0.0456 /// teacc 98.49 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0042, 0.0755, 0.0474, ..., -0.0832, -0.0600, 0.0260], + [ 0.0721, -0.0067, -0.0250, ..., 0.0340, 0.0517, -0.0700], + [-0.0352, -0.0127, -0.0285, ..., -0.0392, -0.0089, -0.0200], + ..., + [-0.0159, 0.0810, -0.0966, ..., -0.0574, 0.0121, -0.0674], + [-0.0022, -0.0473, 0.0253, ..., 0.0339, -0.0710, -0.0294], + [-0.0061, -0.0369, 0.0236, ..., -0.0334, -0.0300, 0.0636]], + device='cuda:0'), grad: tensor([[ 2.2864e-04, 2.2564e-03, -1.4141e-05, ..., 2.4602e-05, + 4.4751e-04, 3.1686e-04], + [-6.8426e-05, 1.7989e-04, 3.4750e-05, ..., -5.0068e-05, + -2.1815e-05, 5.0366e-05], + [ 7.3850e-05, 1.0931e-04, 2.0236e-05, ..., 5.2780e-05, + 5.8919e-05, 2.9698e-05], + ..., + [-3.7766e-04, -5.1804e-03, 1.1235e-05, ..., -4.0054e-05, + -9.0361e-04, -6.6090e-04], + [ 3.7932e-04, 2.4605e-04, 7.0035e-05, ..., 7.9811e-05, + 3.6597e-04, 1.7750e-04], + [ 1.8597e-04, 1.9741e-03, -1.4091e-04, ..., -8.7440e-05, + 3.3545e-04, 1.4007e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0069, 0.0026, 0.0042, 0.0158, 0.0242, 0.0322, -0.0260, 0.0075, + -0.0309, -0.0180], device='cuda:0'), grad: tensor([ 0.0033, 0.0002, 0.0003, 0.0002, 0.0004, -0.0019, 0.0013, -0.0073, + 0.0010, 0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 216.88, cls_loss 0.0178 cls_loss_mapping 0.0227 cls_loss_causal 0.7162 re_mapping 0.0165 re_causal 0.0473 /// teacc 98.68 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0043, 0.0759, 0.0476, ..., -0.0838, -0.0603, 0.0261], + [ 0.0727, -0.0073, -0.0252, ..., 0.0345, 0.0520, -0.0708], + [-0.0357, -0.0129, -0.0285, ..., -0.0398, -0.0088, -0.0201], + ..., + [-0.0160, 0.0819, -0.0972, ..., -0.0579, 0.0120, -0.0685], + [-0.0019, -0.0479, 0.0254, ..., 0.0343, -0.0711, -0.0294], + [-0.0062, -0.0369, 0.0240, ..., -0.0337, -0.0301, 0.0646]], + device='cuda:0'), grad: tensor([[ 1.1206e-04, 8.6486e-05, 2.0218e-04, ..., 1.5068e-04, + 6.2585e-06, 2.8038e-04], + [-6.2943e-05, 9.2909e-06, 3.9488e-05, ..., -5.6595e-05, + -2.3648e-05, 3.2097e-05], + [ 1.2660e-04, 1.5274e-05, 2.4509e-04, ..., 8.1286e-06, + -6.9499e-05, 1.5914e-04], + ..., + [ 4.9233e-05, -2.2447e-04, 6.1274e-05, ..., 5.3883e-05, + -1.8239e-04, 5.3525e-05], + [-2.3532e-04, 7.8157e-06, -1.5867e-04, ..., -3.2592e-04, + 2.1413e-05, -1.2732e-04], + [ 7.5769e-04, 6.1214e-05, 6.0463e-03, ..., 2.6631e-04, + 5.8353e-05, 4.4098e-03]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0065, 0.0026, 0.0041, 0.0154, 0.0244, 0.0325, -0.0264, 0.0076, + -0.0310, -0.0175], device='cuda:0'), grad: tensor([ 7.2098e-04, 9.5367e-06, 3.9697e-05, -8.7509e-03, 1.9002e-04, + 1.3380e-03, -6.5184e-04, -3.0541e-04, -3.7456e-04, 7.7820e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 217.10, cls_loss 0.0146 cls_loss_mapping 0.0207 cls_loss_causal 0.6966 re_mapping 0.0164 re_causal 0.0463 /// teacc 98.66 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0048, 0.0761, 0.0478, ..., -0.0846, -0.0605, 0.0261], + [ 0.0732, -0.0070, -0.0255, ..., 0.0350, 0.0527, -0.0712], + [-0.0362, -0.0132, -0.0287, ..., -0.0404, -0.0092, -0.0205], + ..., + [-0.0163, 0.0826, -0.0980, ..., -0.0582, 0.0121, -0.0694], + [-0.0019, -0.0484, 0.0253, ..., 0.0343, -0.0714, -0.0299], + [-0.0067, -0.0374, 0.0239, ..., -0.0339, -0.0305, 0.0650]], + device='cuda:0'), grad: tensor([[-2.5153e-05, -4.3106e-04, -2.5734e-05, ..., 2.9132e-06, + 6.7614e-07, -5.6982e-04], + [ 2.1219e-05, 1.8871e-04, 5.0552e-06, ..., 8.9347e-05, + -5.9903e-06, 3.0845e-05], + [ 2.6405e-05, 6.7949e-05, 1.0088e-05, ..., 2.2948e-05, + 2.4904e-06, 3.4273e-05], + ..., + [ 2.4691e-05, -5.0163e-04, 1.9282e-05, ..., -2.1148e-04, + 2.7120e-06, 5.6982e-05], + [ 9.9540e-05, 6.4313e-05, 1.6972e-05, ..., 1.0896e-04, + 2.6450e-06, 7.3791e-05], + [ 3.0324e-05, 6.9559e-05, -7.3910e-06, ..., 7.8440e-05, + 1.5078e-06, 2.5272e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0064, 0.0029, 0.0038, 0.0159, 0.0240, 0.0326, -0.0261, 0.0075, + -0.0315, -0.0174], device='cuda:0'), grad: tensor([-9.4986e-04, 4.6253e-04, 1.4639e-04, 1.3599e-03, -3.4273e-05, + -1.1873e-03, 7.6580e-04, -1.0948e-03, 3.0923e-04, 2.2280e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 216.68, cls_loss 0.0171 cls_loss_mapping 0.0225 cls_loss_causal 0.6922 re_mapping 0.0162 re_causal 0.0455 /// teacc 98.58 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0058, 0.0763, 0.0480, ..., -0.0850, -0.0611, 0.0266], + [ 0.0737, -0.0078, -0.0259, ..., 0.0353, 0.0537, -0.0721], + [-0.0370, -0.0131, -0.0289, ..., -0.0411, -0.0098, -0.0209], + ..., + [-0.0164, 0.0837, -0.0989, ..., -0.0577, 0.0122, -0.0700], + [-0.0016, -0.0489, 0.0258, ..., 0.0348, -0.0710, -0.0297], + [-0.0068, -0.0381, 0.0240, ..., -0.0349, -0.0309, 0.0650]], + device='cuda:0'), grad: tensor([[ 1.3700e-06, -1.9562e-04, -6.9857e-05, ..., 1.9759e-05, + 3.6597e-05, -7.3835e-06], + [-2.0707e-04, 2.7232e-06, 2.5295e-06, ..., -3.2282e-04, + -4.4256e-05, 7.6368e-06], + [ 2.2665e-05, 3.5256e-05, 1.2226e-05, ..., 2.5883e-05, + 1.4953e-05, 1.5706e-05], + ..., + [ 4.9949e-05, 2.9489e-05, 1.7309e-04, ..., 3.2139e-04, + 8.3029e-05, 2.1482e-04], + [ 2.9653e-05, 3.3081e-05, -1.5087e-07, ..., 4.4316e-05, + 1.8731e-05, 2.0027e-05], + [ 2.3663e-05, -2.5183e-05, -1.8644e-04, ..., 8.9943e-05, + 6.6817e-05, -1.8764e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0065, 0.0028, 0.0036, 0.0159, 0.0241, 0.0329, -0.0267, 0.0082, + -0.0310, -0.0182], device='cuda:0'), grad: tensor([ 8.6129e-05, -3.1328e-04, 1.3995e-04, 1.5926e-04, -1.9951e-03, + 2.5249e-04, 1.5712e-04, 1.0376e-03, 1.4603e-04, 3.2949e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 217.25, cls_loss 0.0172 cls_loss_mapping 0.0246 cls_loss_causal 0.6925 re_mapping 0.0164 re_causal 0.0448 /// teacc 98.74 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0057, 0.0773, 0.0481, ..., -0.0854, -0.0617, 0.0266], + [ 0.0737, -0.0080, -0.0266, ..., 0.0351, 0.0537, -0.0735], + [-0.0368, -0.0134, -0.0282, ..., -0.0410, -0.0090, -0.0201], + ..., + [-0.0171, 0.0843, -0.1000, ..., -0.0584, 0.0122, -0.0701], + [-0.0012, -0.0494, 0.0260, ..., 0.0355, -0.0714, -0.0297], + [-0.0068, -0.0387, 0.0246, ..., -0.0348, -0.0309, 0.0661]], + device='cuda:0'), grad: tensor([[ 9.7454e-06, -1.2624e-04, -1.7032e-05, ..., 1.5736e-05, + 2.8424e-06, -1.6451e-05], + [-4.3571e-05, 1.6421e-05, 8.8215e-06, ..., -3.5256e-05, + -2.0832e-05, 8.6576e-06], + [ 1.2405e-05, -1.5467e-05, 3.4183e-05, ..., 1.0386e-05, + -2.2411e-05, 4.4376e-05], + ..., + [ 2.1636e-05, -5.6219e-04, -1.0401e-04, ..., -5.9783e-05, + 2.2262e-05, -4.5866e-05], + [-6.4373e-05, 5.4032e-05, -4.8965e-05, ..., -6.9201e-05, + 4.4331e-06, 9.1940e-06], + [ 4.8399e-05, 4.9925e-04, 4.9770e-05, ..., 9.8228e-05, + 3.0901e-06, -1.5414e-04]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0070, 0.0023, 0.0048, 0.0155, 0.0231, 0.0333, -0.0264, 0.0077, + -0.0309, -0.0181], device='cuda:0'), grad: tensor([-1.6797e-04, 2.4661e-06, -3.0935e-05, 1.9833e-05, 1.1218e-04, + 1.8322e-04, -1.4246e-04, -9.7847e-04, 4.2766e-05, 9.5892e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 217.64, cls_loss 0.0152 cls_loss_mapping 0.0221 cls_loss_causal 0.6769 re_mapping 0.0147 re_causal 0.0434 /// teacc 98.39 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0061, 0.0778, 0.0483, ..., -0.0865, -0.0623, 0.0266], + [ 0.0744, -0.0077, -0.0269, ..., 0.0359, 0.0547, -0.0740], + [-0.0370, -0.0135, -0.0283, ..., -0.0411, -0.0090, -0.0203], + ..., + [-0.0180, 0.0850, -0.1004, ..., -0.0596, 0.0120, -0.0706], + [-0.0017, -0.0502, 0.0257, ..., 0.0354, -0.0721, -0.0303], + [-0.0074, -0.0391, 0.0242, ..., -0.0353, -0.0311, 0.0658]], + device='cuda:0'), grad: tensor([[-3.7700e-05, -5.6887e-04, -3.5238e-04, ..., -1.5700e-04, + 7.9628e-07, -5.5027e-04], + [-5.8651e-05, 6.4336e-06, 2.8871e-06, ..., -3.9399e-05, + -3.4660e-05, 4.5933e-06], + [ 2.7224e-05, 4.3750e-05, 1.4402e-05, ..., 2.7642e-05, + 1.0014e-05, 3.1263e-05], + ..., + [ 1.2755e-05, -1.4015e-05, 1.3880e-05, ..., 2.2650e-05, + 5.7071e-06, 1.6466e-05], + [ 1.9923e-05, 2.1070e-05, -5.4762e-06, ..., 1.4119e-05, + 1.2733e-05, 4.1693e-05], + [-1.4650e-06, 2.9340e-05, -7.1824e-05, ..., 1.9699e-05, + 1.0617e-06, -5.6028e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0069, 0.0026, 0.0054, 0.0164, 0.0233, 0.0330, -0.0264, 0.0075, + -0.0316, -0.0188], device='cuda:0'), grad: tensor([-1.1654e-03, -5.0247e-05, 1.2130e-04, 3.6454e-04, 6.3419e-05, + -2.0337e-04, 8.6641e-04, 3.7849e-05, 8.4996e-05, -1.1998e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 216.82, cls_loss 0.0140 cls_loss_mapping 0.0178 cls_loss_causal 0.7035 re_mapping 0.0155 re_causal 0.0441 /// teacc 98.67 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0063, 0.0782, 0.0482, ..., -0.0870, -0.0625, 0.0265], + [ 0.0747, -0.0080, -0.0270, ..., 0.0361, 0.0550, -0.0744], + [-0.0371, -0.0136, -0.0285, ..., -0.0413, -0.0092, -0.0204], + ..., + [-0.0182, 0.0856, -0.1010, ..., -0.0602, 0.0129, -0.0713], + [-0.0018, -0.0506, 0.0254, ..., 0.0353, -0.0725, -0.0307], + [-0.0073, -0.0397, 0.0251, ..., -0.0356, -0.0310, 0.0673]], + device='cuda:0'), grad: tensor([[ 7.4245e-06, -7.7039e-06, -4.6119e-06, ..., 1.8656e-05, + 4.4703e-07, -1.4104e-05], + [ 1.2207e-04, 8.3968e-06, 1.5805e-06, ..., 3.3545e-04, + -2.0280e-05, 7.0035e-07], + [ 1.5795e-05, 2.3327e-03, 1.8024e-04, ..., 3.9756e-05, + 4.1485e-04, 5.2676e-06], + ..., + [ 1.4886e-05, -2.4052e-03, -1.7929e-04, ..., 8.6129e-05, + -4.1842e-04, 1.2545e-06], + [ 1.2887e-04, 2.2784e-05, 2.0951e-05, ..., 2.2483e-04, + 1.0483e-05, 2.3425e-05], + [ 1.3240e-05, 8.8453e-05, 3.8654e-05, ..., 6.5660e-04, + 1.5451e-06, -1.2599e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0068, 0.0023, 0.0054, 0.0163, 0.0229, 0.0334, -0.0262, 0.0080, + -0.0321, -0.0186], device='cuda:0'), grad: tensor([ 4.0978e-05, 4.1986e-04, 1.2688e-02, 2.4700e-04, -3.2578e-03, + 4.8327e-04, -1.5581e-04, -1.2627e-02, 4.1580e-04, 1.7433e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 217.10, cls_loss 0.0141 cls_loss_mapping 0.0213 cls_loss_causal 0.7167 re_mapping 0.0146 re_causal 0.0437 /// teacc 98.74 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0062, 0.0788, 0.0485, ..., -0.0881, -0.0632, 0.0266], + [ 0.0745, -0.0083, -0.0274, ..., 0.0359, 0.0554, -0.0753], + [-0.0369, -0.0139, -0.0282, ..., -0.0411, -0.0092, -0.0201], + ..., + [-0.0185, 0.0860, -0.1017, ..., -0.0607, 0.0126, -0.0722], + [-0.0019, -0.0511, 0.0253, ..., 0.0355, -0.0730, -0.0310], + [-0.0076, -0.0401, 0.0256, ..., -0.0363, -0.0312, 0.0678]], + device='cuda:0'), grad: tensor([[-1.0264e-04, -7.8321e-05, -1.4067e-04, ..., 9.2089e-06, + 2.2516e-05, -8.8334e-05], + [-3.2991e-05, 2.6911e-05, 5.8040e-06, ..., 4.9025e-05, + 3.4022e-04, 5.6848e-06], + [ 3.0845e-05, 2.8706e-04, 3.5793e-05, ..., -6.6638e-05, + -2.7418e-04, 3.1292e-05], + ..., + [ 1.2018e-05, -7.9012e-04, 5.2303e-06, ..., -5.0694e-05, + -2.7752e-04, -4.3809e-05], + [ 2.9564e-05, 6.5088e-05, 3.9190e-05, ..., 1.8075e-05, + 9.6560e-06, 3.4600e-05], + [ 8.2031e-06, 8.7380e-05, -2.1964e-05, ..., 1.9699e-05, + 8.3074e-06, -1.8343e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0071, 0.0019, 0.0055, 0.0161, 0.0241, 0.0343, -0.0266, 0.0077, + -0.0326, -0.0190], device='cuda:0'), grad: tensor([-0.0003, 0.0010, -0.0006, 0.0002, 0.0003, 0.0003, 0.0001, -0.0014, + 0.0002, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 216.93, cls_loss 0.0141 cls_loss_mapping 0.0184 cls_loss_causal 0.6714 re_mapping 0.0145 re_causal 0.0426 /// teacc 98.66 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0059, 0.0800, 0.0488, ..., -0.0887, -0.0636, 0.0264], + [ 0.0749, -0.0084, -0.0276, ..., 0.0366, 0.0560, -0.0762], + [-0.0377, -0.0142, -0.0287, ..., -0.0420, -0.0095, -0.0203], + ..., + [-0.0190, 0.0856, -0.1028, ..., -0.0615, 0.0127, -0.0728], + [-0.0015, -0.0508, 0.0254, ..., 0.0360, -0.0734, -0.0311], + [-0.0079, -0.0397, 0.0267, ..., -0.0365, -0.0312, 0.0688]], + device='cuda:0'), grad: tensor([[ 2.7549e-06, -4.5598e-05, -2.8729e-05, ..., 2.9355e-06, + 1.2685e-06, -3.1739e-05], + [-1.2183e-04, 7.7724e-05, 3.3434e-06, ..., -1.0091e-04, + -8.9645e-05, 7.8604e-07], + [ 1.0264e-04, 1.5706e-05, 2.6245e-06, ..., 1.3316e-04, + 5.3167e-05, 8.5831e-06], + ..., + [-1.1331e-04, -2.0719e-04, 1.0490e-05, ..., -2.5153e-04, + 2.9560e-06, 5.5209e-06], + [ 6.9797e-05, 4.7773e-05, 5.2035e-05, ..., 7.9691e-05, + 1.6987e-05, 2.7657e-05], + [ 1.8865e-05, 3.5524e-05, -1.0617e-06, ..., 2.4870e-05, + 2.7437e-06, -1.1690e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0079, 0.0020, 0.0051, 0.0161, 0.0241, 0.0340, -0.0268, 0.0065, + -0.0323, -0.0181], device='cuda:0'), grad: tensor([-9.3937e-05, -4.9174e-07, 1.3185e-04, 1.3161e-03, 1.7309e-04, + -1.3561e-03, 8.8573e-05, -6.1941e-04, 2.7013e-04, 8.9347e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 217.49, cls_loss 0.0158 cls_loss_mapping 0.0236 cls_loss_causal 0.7000 re_mapping 0.0148 re_causal 0.0413 /// teacc 98.57 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0064, 0.0803, 0.0485, ..., -0.0896, -0.0641, 0.0264], + [ 0.0757, -0.0085, -0.0272, ..., 0.0376, 0.0563, -0.0758], + [-0.0378, -0.0144, -0.0286, ..., -0.0424, -0.0100, -0.0205], + ..., + [-0.0191, 0.0865, -0.1037, ..., -0.0619, 0.0148, -0.0736], + [-0.0017, -0.0513, 0.0254, ..., 0.0361, -0.0738, -0.0313], + [-0.0087, -0.0402, 0.0273, ..., -0.0378, -0.0318, 0.0695]], + device='cuda:0'), grad: tensor([[-3.6001e-05, -4.8876e-05, -2.0847e-05, ..., 2.5630e-05, + 1.0058e-05, 1.0848e-05], + [-1.5214e-05, 1.0318e-04, 1.1158e-04, ..., 1.3828e-04, + 9.2864e-05, 5.5492e-05], + [ 7.4580e-06, 3.4213e-05, -2.0909e-04, ..., -2.2459e-04, + -2.7394e-04, -8.7738e-05], + ..., + [-2.1189e-05, -8.4686e-04, 6.3241e-05, ..., 5.7906e-05, + 9.3997e-05, 4.6194e-05], + [ 6.6906e-06, 2.2754e-05, 1.8895e-05, ..., 4.0054e-05, + 1.1064e-05, 2.0280e-05], + [ 2.4870e-05, 5.4979e-04, -5.8532e-05, ..., 3.5954e-04, + 1.7509e-07, -7.2896e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0075, 0.0022, 0.0052, 0.0156, 0.0244, 0.0335, -0.0263, 0.0078, + -0.0329, -0.0184], device='cuda:0'), grad: tensor([ 1.6555e-05, 8.6069e-04, -1.5326e-03, 3.6836e-04, -4.9496e-04, + 8.4281e-05, 3.4332e-05, -6.1321e-04, 1.3590e-04, 1.1377e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 217.04, cls_loss 0.0131 cls_loss_mapping 0.0191 cls_loss_causal 0.6444 re_mapping 0.0146 re_causal 0.0409 /// teacc 98.56 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0057, 0.0814, 0.0496, ..., -0.0903, -0.0650, 0.0265], + [ 0.0764, -0.0090, -0.0274, ..., 0.0382, 0.0570, -0.0762], + [-0.0386, -0.0150, -0.0290, ..., -0.0432, -0.0104, -0.0210], + ..., + [-0.0193, 0.0874, -0.1044, ..., -0.0625, 0.0144, -0.0744], + [-0.0020, -0.0527, 0.0252, ..., 0.0362, -0.0743, -0.0318], + [-0.0094, -0.0403, 0.0273, ..., -0.0388, -0.0320, 0.0703]], + device='cuda:0'), grad: tensor([[ 7.6115e-05, -3.1024e-05, 1.0109e-04, ..., 7.2300e-05, + -1.8366e-06, 1.8024e-04], + [ 4.7231e-04, 2.7388e-05, 8.5354e-05, ..., 3.2473e-04, + 3.9721e-04, 1.2493e-04], + [-3.5000e-04, 2.7984e-05, 7.5996e-05, ..., -1.1760e-04, + -4.0269e-04, 1.2982e-04], + ..., + [ 3.3200e-05, -5.7928e-07, 6.1333e-05, ..., 8.2314e-05, + 3.0492e-06, 1.0407e-04], + [-2.3574e-05, 2.0832e-05, 3.5822e-05, ..., -6.3181e-05, + 1.6242e-06, 9.0718e-05], + [-8.9359e-04, -4.6730e-04, -1.7204e-03, ..., -1.0118e-03, + 7.9349e-07, -2.9564e-03]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0074, 0.0025, 0.0047, 0.0162, 0.0246, 0.0335, -0.0264, 0.0077, + -0.0333, -0.0184], device='cuda:0'), grad: tensor([ 3.3760e-04, 1.3380e-03, -6.9761e-04, 2.0027e-05, 3.7498e-03, + 7.6830e-05, 1.3101e-04, 2.8086e-04, 5.6058e-05, -5.2910e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 217.28, cls_loss 0.0139 cls_loss_mapping 0.0198 cls_loss_causal 0.7196 re_mapping 0.0132 re_causal 0.0402 /// teacc 98.61 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0061, 0.0817, 0.0495, ..., -0.0919, -0.0660, 0.0264], + [ 0.0766, -0.0088, -0.0279, ..., 0.0386, 0.0576, -0.0777], + [-0.0389, -0.0149, -0.0293, ..., -0.0438, -0.0101, -0.0214], + ..., + [-0.0199, 0.0882, -0.1050, ..., -0.0635, 0.0141, -0.0752], + [-0.0022, -0.0538, 0.0251, ..., 0.0364, -0.0745, -0.0327], + [-0.0094, -0.0411, 0.0278, ..., -0.0393, -0.0319, 0.0711]], + device='cuda:0'), grad: tensor([[-2.7943e-04, -4.9829e-04, -5.8556e-04, ..., 9.2909e-06, + 1.7975e-06, -3.7694e-04], + [-6.9666e-04, 1.5702e-06, 3.4496e-06, ..., -3.9101e-04, + -5.6934e-04, 3.7476e-06], + [-3.1257e-04, 5.3465e-05, 5.9664e-05, ..., 2.9802e-04, + 5.3263e-04, 3.7163e-05], + ..., + [ 3.3885e-05, 8.4788e-06, 1.1921e-05, ..., 2.7940e-05, + 1.9848e-05, 1.1757e-05], + [ 1.0443e-03, 3.4958e-05, 6.8307e-05, ..., 1.3506e-04, + 1.8120e-05, 1.0216e-04], + [ 2.1100e-05, 5.4777e-05, 4.6879e-05, ..., 4.5240e-05, + 2.7418e-06, 2.7537e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0069, 0.0023, 0.0047, 0.0165, 0.0247, 0.0333, -0.0252, 0.0077, + -0.0336, -0.0187], device='cuda:0'), grad: tensor([-1.0357e-03, -1.1702e-03, -5.3501e-04, 1.4496e-04, -6.0946e-05, + -9.3746e-04, 1.5068e-03, 8.7798e-05, 1.8692e-03, 1.3113e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 217.46, cls_loss 0.0137 cls_loss_mapping 0.0196 cls_loss_causal 0.6784 re_mapping 0.0147 re_causal 0.0403 /// teacc 98.62 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0063, 0.0821, 0.0497, ..., -0.0928, -0.0667, 0.0263], + [ 0.0774, -0.0088, -0.0282, ..., 0.0393, 0.0582, -0.0785], + [-0.0388, -0.0155, -0.0294, ..., -0.0437, -0.0097, -0.0213], + ..., + [-0.0203, 0.0890, -0.1059, ..., -0.0642, 0.0144, -0.0757], + [-0.0020, -0.0538, 0.0249, ..., 0.0368, -0.0756, -0.0326], + [-0.0102, -0.0414, 0.0283, ..., -0.0398, -0.0327, 0.0717]], + device='cuda:0'), grad: tensor([[ 1.0885e-05, -1.4317e-04, -1.2624e-04, ..., 1.1876e-05, + 4.7311e-07, -1.3590e-04], + [-5.2154e-05, -5.8077e-06, 9.4175e-06, ..., -3.3021e-05, + -2.3827e-05, 5.2825e-06], + [ 1.4387e-05, 1.9521e-05, 2.7299e-05, ..., 6.9886e-06, + 2.9150e-06, 1.8775e-05], + ..., + [ 3.1650e-05, -7.1898e-06, 4.3184e-05, ..., 1.3903e-05, + 8.9481e-06, 2.5570e-05], + [ 2.8163e-05, 5.3719e-06, 6.1333e-05, ..., 1.8656e-05, + 8.5235e-06, 6.2883e-05], + [-3.3188e-04, 1.2837e-05, -1.1730e-03, ..., -4.1294e-04, + 2.5332e-06, -1.0061e-03]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0069, 0.0026, 0.0053, 0.0162, 0.0246, 0.0337, -0.0256, 0.0078, + -0.0342, -0.0188], device='cuda:0'), grad: tensor([-3.3283e-04, -4.1038e-05, 5.9336e-05, 2.8634e-04, 1.2703e-03, + 7.3481e-04, 6.8069e-05, 8.5473e-05, 1.0389e-04, -2.2354e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 217.58, cls_loss 0.0159 cls_loss_mapping 0.0185 cls_loss_causal 0.6907 re_mapping 0.0142 re_causal 0.0397 /// teacc 98.58 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0067, 0.0835, 0.0504, ..., -0.0939, -0.0670, 0.0266], + [ 0.0786, -0.0071, -0.0291, ..., 0.0400, 0.0591, -0.0796], + [-0.0394, -0.0160, -0.0296, ..., -0.0441, -0.0098, -0.0213], + ..., + [-0.0220, 0.0886, -0.1073, ..., -0.0651, 0.0135, -0.0763], + [-0.0007, -0.0532, 0.0262, ..., 0.0384, -0.0760, -0.0321], + [-0.0111, -0.0413, 0.0287, ..., -0.0406, -0.0324, 0.0721]], + device='cuda:0'), grad: tensor([[ 6.2108e-05, -1.3180e-05, 8.1062e-05, ..., 6.0737e-05, + 4.2729e-06, 9.2089e-05], + [ 1.8373e-05, 1.9372e-04, 5.7638e-05, ..., -3.5077e-05, + 3.9905e-05, 3.6895e-05], + [ 3.0947e-04, -4.1628e-04, 4.4227e-04, ..., 1.9014e-04, + 1.9521e-05, 3.7313e-04], + ..., + [-5.9754e-06, 1.9526e-04, 1.0705e-04, ..., 4.7445e-05, + -8.5711e-05, 7.5042e-05], + [ 7.3433e-04, 1.4782e-04, 8.2493e-04, ..., 4.6515e-04, + 2.5123e-05, 7.3671e-04], + [-1.2434e-04, 2.4348e-05, -4.8780e-04, ..., -8.9526e-05, + 3.8832e-05, -4.0293e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0076, 0.0037, 0.0051, 0.0154, 0.0246, 0.0334, -0.0268, 0.0068, + -0.0326, -0.0187], device='cuda:0'), grad: tensor([ 0.0002, 0.0004, -0.0003, 0.0087, 0.0012, -0.0111, -0.0009, 0.0009, + 0.0018, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 217.59, cls_loss 0.0151 cls_loss_mapping 0.0191 cls_loss_causal 0.6702 re_mapping 0.0140 re_causal 0.0397 /// teacc 98.72 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0059, 0.0844, 0.0504, ..., -0.0948, -0.0673, 0.0266], + [ 0.0783, -0.0088, -0.0293, ..., 0.0399, 0.0596, -0.0797], + [-0.0400, -0.0165, -0.0296, ..., -0.0446, -0.0098, -0.0215], + ..., + [-0.0209, 0.0892, -0.1088, ..., -0.0652, 0.0136, -0.0766], + [-0.0009, -0.0538, 0.0265, ..., 0.0382, -0.0768, -0.0320], + [-0.0124, -0.0419, 0.0289, ..., -0.0413, -0.0329, 0.0723]], + device='cuda:0'), grad: tensor([[ 8.7738e-05, -6.5029e-05, 2.3276e-05, ..., 8.7082e-05, + 1.0896e-06, 1.6510e-05], + [-2.5511e-04, -1.3387e-04, -3.3021e-05, ..., -1.3030e-04, + -1.2052e-04, 1.1414e-05], + [ 7.6473e-05, 1.3256e-04, 3.6538e-05, ..., 6.0320e-05, + 2.7046e-05, 9.2313e-06], + ..., + [ 1.2195e-04, -6.8378e-04, 3.7700e-06, ..., 1.1122e-04, + 5.1707e-05, 9.2909e-06], + [ 8.0109e-03, 4.5300e-05, 4.1618e-03, ..., 6.7596e-03, + 1.1295e-05, 4.8904e-03], + [ 1.2696e-04, 1.0109e-04, 2.4376e-03, ..., 2.9640e-03, + 2.0027e-05, 9.9659e-04]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0079, 0.0029, 0.0043, 0.0164, 0.0248, 0.0332, -0.0260, 0.0073, + -0.0331, -0.0189], device='cuda:0'), grad: tensor([ 5.0336e-05, -6.1035e-04, 3.8457e-04, 9.7656e-04, -5.9509e-03, + 8.2111e-04, -1.2543e-02, -9.3555e-04, 1.1742e-02, 6.0616e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 65---------------------------------------------------- +epoch 65, time 218.36, cls_loss 0.0114 cls_loss_mapping 0.0160 cls_loss_causal 0.6463 re_mapping 0.0145 re_causal 0.0400 /// teacc 98.81 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0064, 0.0848, 0.0500, ..., -0.0959, -0.0678, 0.0265], + [ 0.0787, -0.0092, -0.0298, ..., 0.0400, 0.0600, -0.0805], + [-0.0405, -0.0169, -0.0296, ..., -0.0453, -0.0100, -0.0215], + ..., + [-0.0207, 0.0903, -0.1092, ..., -0.0658, 0.0138, -0.0772], + [-0.0010, -0.0542, 0.0264, ..., 0.0384, -0.0772, -0.0324], + [-0.0123, -0.0421, 0.0296, ..., -0.0421, -0.0330, 0.0728]], + device='cuda:0'), grad: tensor([[-4.0829e-05, -3.4600e-05, -9.2387e-05, ..., -8.8885e-06, + 8.4005e-07, -1.4961e-04], + [-4.6700e-05, -2.5667e-06, -7.4804e-06, ..., -4.0352e-05, + -2.0251e-05, 3.1628e-06], + [-5.2806e-07, 2.0117e-06, 2.0284e-06, ..., -5.1744e-06, + 1.4510e-06, 8.6650e-06], + ..., + [ 3.6895e-05, 1.3039e-07, 9.8869e-06, ..., 2.7433e-05, + 6.1281e-06, 9.8497e-06], + [ 1.5616e-04, 1.7822e-05, 1.1545e-04, ..., 1.2434e-04, + 8.0615e-06, 1.3888e-04], + [ 3.5882e-05, 7.1041e-06, 2.3231e-05, ..., 2.7299e-05, + 2.4233e-06, 4.9502e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0076, 0.0023, 0.0042, 0.0159, 0.0258, 0.0335, -0.0260, 0.0077, + -0.0334, -0.0188], device='cuda:0'), grad: tensor([ 8.9705e-05, -4.6015e-05, -5.7369e-05, 1.8806e-03, -1.0958e-03, + -2.1915e-03, 1.5533e-04, 7.4983e-05, 3.8362e-04, 8.0585e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 217.29, cls_loss 0.0129 cls_loss_mapping 0.0183 cls_loss_causal 0.6694 re_mapping 0.0137 re_causal 0.0393 /// teacc 98.53 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0067, 0.0847, 0.0504, ..., -0.0963, -0.0683, 0.0259], + [ 0.0792, -0.0088, -0.0299, ..., 0.0404, 0.0611, -0.0807], + [-0.0406, -0.0175, -0.0297, ..., -0.0455, -0.0099, -0.0220], + ..., + [-0.0216, 0.0904, -0.1104, ..., -0.0662, 0.0130, -0.0775], + [-0.0004, -0.0543, 0.0266, ..., 0.0389, -0.0774, -0.0323], + [-0.0130, -0.0413, 0.0303, ..., -0.0427, -0.0342, 0.0739]], + device='cuda:0'), grad: tensor([[-1.1879e-04, -4.3607e-04, -4.4179e-04, ..., -1.1015e-04, + 4.8727e-06, -5.3358e-04], + [-1.5218e-06, 2.1234e-06, 1.0572e-05, ..., -4.4703e-06, + -6.7018e-06, 8.8513e-06], + [-5.8174e-05, 1.3530e-04, -3.5191e-04, ..., -1.5998e-04, + -4.3392e-05, -2.0611e-04], + ..., + [ 2.7394e-04, 7.0691e-05, 1.4305e-04, ..., 4.0710e-05, + 4.3735e-06, 2.2125e-04], + [ 3.5810e-04, -1.0556e-04, 6.7711e-05, ..., 3.6907e-04, + 4.3821e-04, -9.7632e-05], + [ 1.4353e-04, 2.9850e-04, 6.4039e-04, ..., 1.3578e-04, + 1.2949e-05, 7.1764e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0068, 0.0028, 0.0046, 0.0153, 0.0250, 0.0336, -0.0260, 0.0073, + -0.0333, -0.0178], device='cuda:0'), grad: tensor([-8.6784e-04, 3.1382e-05, -8.8644e-04, 2.1648e-04, 5.0403e-06, + -9.8801e-04, 2.2995e-04, 8.0776e-04, 8.7082e-05, 1.3628e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 217.70, cls_loss 0.0099 cls_loss_mapping 0.0130 cls_loss_causal 0.6772 re_mapping 0.0132 re_causal 0.0392 /// teacc 98.64 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0061, 0.0856, 0.0505, ..., -0.0967, -0.0683, 0.0261], + [ 0.0789, -0.0103, -0.0302, ..., 0.0403, 0.0613, -0.0813], + [-0.0409, -0.0172, -0.0301, ..., -0.0458, -0.0100, -0.0222], + ..., + [-0.0211, 0.0915, -0.1106, ..., -0.0665, 0.0129, -0.0779], + [-0.0005, -0.0546, 0.0263, ..., 0.0390, -0.0782, -0.0326], + [-0.0136, -0.0420, 0.0303, ..., -0.0437, -0.0353, 0.0742]], + device='cuda:0'), grad: tensor([[ 7.1079e-06, -5.7220e-06, -5.5790e-05, ..., 4.2468e-06, + 5.6205e-07, -7.3850e-05], + [ 4.1574e-06, 2.0172e-06, 7.4804e-06, ..., 1.6354e-06, + -9.4529e-08, 3.7067e-06], + [ 1.2867e-05, 5.4948e-07, 2.1771e-05, ..., 1.7434e-06, + -1.0645e-06, 9.2089e-06], + ..., + [ 3.5688e-06, -7.2718e-05, 8.7246e-06, ..., 1.7844e-06, + 3.2922e-07, -8.6352e-06], + [ 3.6627e-05, 3.1404e-06, 6.6459e-05, ..., 3.8743e-06, + 1.8049e-06, 1.9982e-05], + [ 1.4193e-05, 6.3896e-05, 5.2840e-05, ..., 7.6741e-06, + 8.7684e-07, 6.5565e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0071, 0.0018, 0.0049, 0.0155, 0.0256, 0.0345, -0.0262, 0.0077, + -0.0337, -0.0186], device='cuda:0'), grad: tensor([-6.5327e-05, 1.7598e-05, 3.3289e-05, -2.3615e-04, 7.6964e-06, + 6.0469e-05, -4.7758e-06, -1.5426e-04, 1.1778e-04, 2.2399e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 217.55, cls_loss 0.0102 cls_loss_mapping 0.0148 cls_loss_causal 0.6548 re_mapping 0.0137 re_causal 0.0374 /// teacc 98.75 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0061, 0.0859, 0.0508, ..., -0.0975, -0.0685, 0.0264], + [ 0.0794, -0.0104, -0.0310, ..., 0.0408, 0.0621, -0.0818], + [-0.0414, -0.0170, -0.0297, ..., -0.0465, -0.0103, -0.0224], + ..., + [-0.0217, 0.0919, -0.1116, ..., -0.0673, 0.0133, -0.0785], + [-0.0004, -0.0549, 0.0266, ..., 0.0396, -0.0791, -0.0326], + [-0.0141, -0.0424, 0.0312, ..., -0.0442, -0.0355, 0.0747]], + device='cuda:0'), grad: tensor([[ 1.4611e-05, -3.7206e-07, 1.1660e-06, ..., 1.4313e-05, + 1.3430e-06, 1.7239e-06], + [ 5.8460e-04, 6.6757e-06, 2.5947e-06, ..., 4.4775e-04, + -1.4141e-05, 5.1886e-05], + [ 3.6955e-05, 8.0466e-06, 3.2187e-06, ..., 3.3587e-05, + 1.3188e-05, 1.5497e-06], + ..., + [ 6.6683e-06, -9.3162e-05, 1.5765e-05, ..., 1.6510e-05, + 5.7667e-06, 2.1487e-05], + [-9.0361e-04, 3.8929e-06, -1.6287e-05, ..., -6.6948e-04, + -2.9117e-05, -5.1916e-05], + [ 3.4273e-05, 2.2680e-05, -1.1677e-04, ..., -2.5090e-06, + 2.6971e-06, -1.1867e-04]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0071, 0.0019, 0.0054, 0.0151, 0.0247, 0.0346, -0.0260, 0.0076, + -0.0337, -0.0182], device='cuda:0'), grad: tensor([ 1.5244e-05, 5.9795e-04, -9.0078e-06, 2.0039e-04, 1.3351e-04, + -1.3447e-04, 2.6369e-04, -9.4414e-05, -8.7166e-04, -1.0192e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 217.81, cls_loss 0.0091 cls_loss_mapping 0.0133 cls_loss_causal 0.6313 re_mapping 0.0134 re_causal 0.0387 /// teacc 98.74 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0069, 0.0862, 0.0505, ..., -0.0986, -0.0698, 0.0260], + [ 0.0802, -0.0104, -0.0310, ..., 0.0418, 0.0631, -0.0820], + [-0.0417, -0.0172, -0.0298, ..., -0.0470, -0.0102, -0.0225], + ..., + [-0.0228, 0.0929, -0.1125, ..., -0.0682, 0.0130, -0.0789], + [-0.0002, -0.0549, 0.0265, ..., 0.0396, -0.0802, -0.0328], + [-0.0141, -0.0429, 0.0318, ..., -0.0443, -0.0359, 0.0756]], + device='cuda:0'), grad: tensor([[ 1.2174e-05, -1.6883e-05, 3.7476e-06, ..., 2.1532e-05, + 2.6263e-07, 8.5831e-06], + [ 1.7667e-04, 2.3925e-04, 6.5845e-07, ..., 1.0449e-04, + 6.3837e-05, 6.2957e-07], + [ 1.5073e-05, 8.5086e-06, 5.4054e-06, ..., 1.7419e-05, + 1.1502e-06, 2.0955e-06], + ..., + [-3.0017e-04, -3.9744e-04, 9.9279e-07, ..., -1.7309e-04, + -1.1367e-04, 8.9640e-07], + [ 5.6118e-05, 8.1599e-05, -2.5555e-06, ..., 3.1948e-05, + 2.5213e-05, 1.6093e-05], + [ 2.8118e-05, 4.0352e-05, -3.1460e-06, ..., 3.4064e-05, + 9.3207e-06, 2.2352e-07]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0060, 0.0025, 0.0055, 0.0147, 0.0244, 0.0354, -0.0262, 0.0075, + -0.0338, -0.0179], device='cuda:0'), grad: tensor([ 1.5227e-07, 5.0735e-04, 3.9279e-05, 3.6538e-05, -2.3767e-05, + 1.3840e-04, -1.3018e-04, -8.3303e-04, 1.6761e-04, 9.7513e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 217.40, cls_loss 0.0126 cls_loss_mapping 0.0166 cls_loss_causal 0.6516 re_mapping 0.0130 re_causal 0.0374 /// teacc 98.71 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0067, 0.0869, 0.0529, ..., -0.0992, -0.0705, 0.0273], + [ 0.0809, -0.0106, -0.0307, ..., 0.0424, 0.0632, -0.0820], + [-0.0422, -0.0176, -0.0301, ..., -0.0474, -0.0099, -0.0228], + ..., + [-0.0225, 0.0936, -0.1124, ..., -0.0690, 0.0130, -0.0793], + [-0.0008, -0.0556, 0.0262, ..., 0.0393, -0.0819, -0.0334], + [-0.0148, -0.0439, 0.0310, ..., -0.0449, -0.0365, 0.0752]], + device='cuda:0'), grad: tensor([[ 9.8497e-06, -3.2037e-06, -6.2250e-06, ..., 1.1481e-05, + 2.3060e-06, -5.7779e-06], + [-6.1572e-05, -4.6343e-06, 4.9993e-06, ..., -5.3078e-05, + -3.4392e-05, 5.6177e-06], + [ 1.5140e-05, 2.1495e-06, 4.1164e-06, ..., 1.6630e-05, + 6.0350e-06, 1.0896e-06], + ..., + [ 6.4313e-05, 3.0845e-06, 7.2241e-05, ..., 1.0329e-04, + 2.1175e-05, 5.4091e-05], + [-6.4135e-05, 5.6773e-06, -8.5356e-07, ..., -6.0588e-05, + 9.4250e-06, 1.5572e-05], + [ 1.2189e-05, -8.6427e-06, -9.5069e-05, ..., -2.7508e-05, + -1.5453e-05, -8.5175e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0071, 0.0024, 0.0056, 0.0147, 0.0245, 0.0361, -0.0265, 0.0079, + -0.0348, -0.0188], device='cuda:0'), grad: tensor([ 7.4040e-08, -6.1214e-05, 3.0249e-05, 1.6212e-05, 4.0233e-05, + 3.2842e-05, -2.7925e-05, 3.0589e-04, -7.6711e-05, -2.5964e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 71---------------------------------------------------- +epoch 71, time 217.41, cls_loss 0.0109 cls_loss_mapping 0.0170 cls_loss_causal 0.6640 re_mapping 0.0126 re_causal 0.0361 /// teacc 98.87 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0070, 0.0869, 0.0538, ..., -0.0998, -0.0714, 0.0272], + [ 0.0812, -0.0113, -0.0318, ..., 0.0426, 0.0634, -0.0830], + [-0.0425, -0.0180, -0.0307, ..., -0.0477, -0.0095, -0.0232], + ..., + [-0.0229, 0.0946, -0.1128, ..., -0.0704, 0.0135, -0.0801], + [-0.0006, -0.0555, 0.0264, ..., 0.0398, -0.0828, -0.0334], + [-0.0151, -0.0442, 0.0311, ..., -0.0451, -0.0373, 0.0765]], + device='cuda:0'), grad: tensor([[-1.0394e-05, -3.7611e-05, -2.0280e-05, ..., -2.5451e-05, + 7.8510e-07, -4.9055e-05], + [-1.4329e-04, 3.8818e-06, -6.0126e-06, ..., -1.4424e-04, + -5.3257e-05, -5.5045e-05], + [ 5.5842e-06, 1.2495e-05, 3.7551e-06, ..., 4.8056e-06, + 8.3214e-07, 2.5835e-06], + ..., + [ 1.9431e-05, -8.4400e-05, 2.6636e-06, ..., 2.3872e-05, + 6.2808e-06, 8.7023e-06], + [ 6.5386e-05, 8.8662e-06, 5.5321e-06, ..., 6.2704e-05, + 2.3872e-05, 3.3528e-05], + [ 3.3885e-05, 2.2873e-05, 6.9961e-06, ..., 4.6581e-05, + 9.6411e-06, 1.9372e-05]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0065, 0.0018, 0.0056, 0.0146, 0.0240, 0.0366, -0.0268, 0.0081, + -0.0344, -0.0182], device='cuda:0'), grad: tensor([-1.2529e-04, -1.6105e-04, -5.3570e-06, 4.1306e-05, 1.2541e-04, + 7.3135e-05, 2.2769e-05, -2.3019e-04, 1.3590e-04, 1.2326e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 216.82, cls_loss 0.0108 cls_loss_mapping 0.0152 cls_loss_causal 0.6341 re_mapping 0.0127 re_causal 0.0352 /// teacc 98.71 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0074, 0.0874, 0.0539, ..., -0.1007, -0.0720, 0.0272], + [ 0.0823, -0.0116, -0.0312, ..., 0.0439, 0.0648, -0.0829], + [-0.0431, -0.0183, -0.0310, ..., -0.0482, -0.0100, -0.0233], + ..., + [-0.0233, 0.0948, -0.1136, ..., -0.0713, 0.0145, -0.0805], + [-0.0009, -0.0559, 0.0262, ..., 0.0398, -0.0836, -0.0336], + [-0.0154, -0.0439, 0.0316, ..., -0.0460, -0.0369, 0.0768]], + device='cuda:0'), grad: tensor([[ 9.8228e-05, -4.9639e-07, 1.1140e-04, ..., 6.6638e-05, + 1.7500e-06, 4.6313e-05], + [ 3.4552e-06, 4.7609e-06, 1.4238e-05, ..., 8.5607e-06, + 3.4183e-05, 1.1042e-05], + [ 4.8131e-05, 6.1393e-06, 3.2008e-05, ..., -4.3735e-06, + -5.2482e-05, 3.5733e-05], + ..., + [ 8.7619e-06, -9.7513e-05, 1.2994e-05, ..., 1.0863e-05, + 7.2531e-06, 1.2279e-05], + [-3.9291e-04, 2.0303e-06, -3.9721e-04, ..., -1.5020e-04, + 1.8835e-05, -2.1887e-04], + [ 1.7822e-04, 4.2766e-05, 7.4446e-05, ..., -8.3372e-06, + -1.1139e-05, 9.0450e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0065, 0.0027, 0.0054, 0.0147, 0.0241, 0.0363, -0.0269, 0.0080, + -0.0350, -0.0182], device='cuda:0'), grad: tensor([ 5.6952e-05, 2.4819e-04, -1.5855e-04, 1.7858e-04, 2.2602e-04, + -1.7196e-05, 3.1292e-05, -1.2177e-04, -4.9973e-04, 5.4419e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 216.84, cls_loss 0.0106 cls_loss_mapping 0.0136 cls_loss_causal 0.6510 re_mapping 0.0127 re_causal 0.0354 /// teacc 98.71 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0081, 0.0876, 0.0538, ..., -0.1032, -0.0736, 0.0268], + [ 0.0829, -0.0119, -0.0313, ..., 0.0445, 0.0652, -0.0837], + [-0.0441, -0.0187, -0.0313, ..., -0.0495, -0.0102, -0.0234], + ..., + [-0.0236, 0.0958, -0.1143, ..., -0.0723, 0.0149, -0.0811], + [-0.0006, -0.0563, 0.0268, ..., 0.0409, -0.0844, -0.0335], + [-0.0160, -0.0445, 0.0321, ..., -0.0464, -0.0354, 0.0778]], + device='cuda:0'), grad: tensor([[ 2.3139e-04, 9.2030e-05, 1.0246e-04, ..., 2.6774e-04, + 3.4180e-07, 2.2006e-04], + [-1.0937e-05, 1.4137e-06, 1.5378e-05, ..., -2.8774e-05, + -2.2203e-05, 9.5442e-06], + [ 1.4804e-05, 8.8988e-07, 2.1607e-05, ..., 1.0617e-05, + 4.1462e-06, 1.2010e-05], + ..., + [ 4.6849e-05, -9.2685e-06, 6.4194e-05, ..., 3.7998e-05, + 8.6278e-06, 5.5254e-05], + [ 3.7527e-04, 1.5366e-04, 1.7643e-04, ..., 4.2915e-04, + 1.8552e-06, 3.5620e-04], + [ 6.9141e-05, 1.7673e-05, 1.2338e-04, ..., 5.2959e-05, + 1.6727e-06, 1.1003e-04]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0060, 0.0026, 0.0047, 0.0154, 0.0244, 0.0351, -0.0266, 0.0084, + -0.0345, -0.0181], device='cuda:0'), grad: tensor([ 5.8842e-04, 1.5143e-06, 6.8724e-05, -2.9335e-03, 7.7009e-05, + -4.3335e-03, 4.8714e-03, 2.4652e-04, 9.7942e-04, 4.3201e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 217.00, cls_loss 0.0095 cls_loss_mapping 0.0162 cls_loss_causal 0.6392 re_mapping 0.0128 re_causal 0.0361 /// teacc 98.85 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0084, 0.0879, 0.0538, ..., -0.1038, -0.0738, 0.0265], + [ 0.0834, -0.0118, -0.0318, ..., 0.0456, 0.0662, -0.0840], + [-0.0445, -0.0190, -0.0313, ..., -0.0499, -0.0106, -0.0232], + ..., + [-0.0241, 0.0964, -0.1151, ..., -0.0740, 0.0146, -0.0817], + [-0.0004, -0.0567, 0.0272, ..., 0.0410, -0.0847, -0.0336], + [-0.0166, -0.0449, 0.0323, ..., -0.0478, -0.0358, 0.0785]], + device='cuda:0'), grad: tensor([[ 1.3940e-05, -2.0057e-05, 4.6119e-06, ..., 1.8671e-05, + 3.4366e-06, -5.1633e-06], + [-2.8685e-06, 3.0901e-06, 5.4203e-06, ..., 7.7933e-06, + -1.1008e-06, 7.3053e-06], + [ 8.2105e-06, 3.9563e-06, -8.0019e-06, ..., -5.2601e-05, + 1.2830e-05, -1.0818e-05], + ..., + [ 2.1100e-05, -2.0355e-05, 4.2498e-05, ..., 3.1829e-05, + 1.0930e-05, 4.6939e-05], + [ 3.0726e-05, 3.0138e-06, 4.5061e-05, ..., 4.4256e-05, + 7.9796e-06, 4.3958e-05], + [ 2.4819e-04, 1.9744e-05, 2.7561e-04, ..., 2.1684e-04, + 1.4400e-04, 2.8157e-04]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0054, 0.0032, 0.0047, 0.0163, 0.0251, 0.0344, -0.0268, 0.0082, + -0.0343, -0.0187], device='cuda:0'), grad: tensor([ 2.9460e-05, 5.8413e-05, -2.1577e-04, 4.3392e-04, -5.6458e-04, + -1.0576e-03, -4.9084e-05, 1.3280e-04, 1.9884e-04, 1.0309e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 217.29, cls_loss 0.0096 cls_loss_mapping 0.0140 cls_loss_causal 0.6486 re_mapping 0.0126 re_causal 0.0357 /// teacc 98.76 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0085, 0.0886, 0.0543, ..., -0.1046, -0.0745, 0.0267], + [ 0.0835, -0.0120, -0.0321, ..., 0.0459, 0.0660, -0.0846], + [-0.0447, -0.0196, -0.0316, ..., -0.0501, -0.0103, -0.0234], + ..., + [-0.0245, 0.0977, -0.1156, ..., -0.0746, 0.0149, -0.0817], + [ 0.0003, -0.0572, 0.0276, ..., 0.0416, -0.0849, -0.0337], + [-0.0173, -0.0460, 0.0322, ..., -0.0492, -0.0357, 0.0785]], + device='cuda:0'), grad: tensor([[ 1.4165e-06, -2.6301e-05, -2.2992e-05, ..., 2.3115e-06, + 2.3656e-07, -4.3571e-05], + [-1.0222e-05, -8.0606e-07, 2.8722e-06, ..., -1.6570e-05, + -8.2701e-06, 1.1288e-06], + [ 2.6617e-06, 4.9695e-06, 9.7379e-06, ..., -2.0564e-05, + -5.7742e-06, 3.6545e-06], + ..., + [ 9.0823e-06, -1.4079e-04, 1.2197e-05, ..., 1.7300e-05, + 4.8019e-06, 1.1131e-05], + [-2.6917e-04, 5.0142e-06, -5.6297e-05, ..., -3.5739e-04, + 6.1356e-06, -9.9003e-05], + [ 5.2601e-06, 2.6122e-05, 9.3877e-06, ..., 1.0945e-05, + -4.8131e-06, 2.5909e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0058, 0.0027, 0.0044, 0.0160, 0.0260, 0.0346, -0.0267, 0.0089, + -0.0340, -0.0197], device='cuda:0'), grad: tensor([-4.1217e-05, -5.1633e-06, -1.2226e-05, 6.2764e-05, 1.9014e-05, + 6.3848e-04, 1.8641e-05, -1.8847e-04, -5.4932e-04, 5.7161e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 216.57, cls_loss 0.0104 cls_loss_mapping 0.0168 cls_loss_causal 0.6174 re_mapping 0.0129 re_causal 0.0360 /// teacc 98.65 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0091, 0.0889, 0.0544, ..., -0.1057, -0.0762, 0.0265], + [ 0.0843, -0.0121, -0.0326, ..., 0.0463, 0.0677, -0.0853], + [-0.0460, -0.0202, -0.0318, ..., -0.0513, -0.0111, -0.0237], + ..., + [-0.0254, 0.0990, -0.1157, ..., -0.0757, 0.0142, -0.0820], + [ 0.0009, -0.0581, 0.0280, ..., 0.0424, -0.0852, -0.0338], + [-0.0174, -0.0472, 0.0330, ..., -0.0497, -0.0355, 0.0801]], + device='cuda:0'), grad: tensor([[ 4.5188e-06, -6.0024e-07, 3.0637e-05, ..., 1.2532e-05, + 9.7416e-07, 2.9087e-05], + [-1.1463e-03, -2.5344e-04, -3.0184e-04, ..., -9.2793e-04, + -2.8443e-04, -1.7715e-04], + [ 8.7246e-06, -4.3362e-06, 3.0976e-06, ..., -1.8300e-06, + 3.1292e-07, -3.8683e-05], + ..., + [ 3.1614e-04, 6.8665e-05, 9.2983e-05, ..., 2.7251e-04, + 8.6129e-05, 6.3777e-05], + [ 1.1998e-04, 2.4259e-05, 2.3103e-04, ..., 1.4997e-04, + 3.0339e-05, 2.0719e-04], + [ 5.8842e-04, 1.4615e-04, -8.7214e-04, ..., 1.7083e-04, + 1.4830e-04, -8.8787e-04]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0055, 0.0032, 0.0041, 0.0153, 0.0251, 0.0347, -0.0262, 0.0094, + -0.0340, -0.0194], device='cuda:0'), grad: tensor([ 3.9250e-05, -2.1858e-03, -1.3316e-04, 5.0068e-04, 7.9632e-05, + 4.1676e-04, -4.9453e-07, 6.8426e-04, 4.3297e-04, 1.6403e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 217.60, cls_loss 0.0095 cls_loss_mapping 0.0137 cls_loss_causal 0.6158 re_mapping 0.0123 re_causal 0.0350 /// teacc 98.73 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0097, 0.0887, 0.0542, ..., -0.1064, -0.0778, 0.0266], + [ 0.0837, -0.0125, -0.0338, ..., 0.0456, 0.0672, -0.0860], + [-0.0449, -0.0196, -0.0311, ..., -0.0502, -0.0099, -0.0239], + ..., + [-0.0252, 0.1001, -0.1160, ..., -0.0754, 0.0147, -0.0828], + [ 0.0015, -0.0592, 0.0284, ..., 0.0432, -0.0850, -0.0337], + [-0.0179, -0.0476, 0.0336, ..., -0.0500, -0.0362, 0.0811]], + device='cuda:0'), grad: tensor([[ 2.1711e-05, -7.3425e-06, 4.4316e-05, ..., 5.6177e-05, + 2.5565e-07, 5.5999e-05], + [-2.2516e-05, 7.8836e-07, 3.2708e-06, ..., -3.6448e-05, + -1.9744e-05, 3.2894e-06], + [ 1.1593e-05, 2.5511e-05, 2.0906e-05, ..., 2.1711e-05, + 2.9318e-06, 2.1636e-05], + ..., + [ 1.3456e-05, -8.2135e-05, 2.0832e-05, ..., 2.7165e-05, + 5.4426e-06, 2.3559e-05], + [-6.6578e-05, -2.8297e-05, -1.8561e-04, ..., -1.5283e-04, + 4.7982e-06, -2.0170e-04], + [ 2.0444e-05, 3.1531e-05, 2.4423e-05, ..., 4.8578e-05, + 1.3113e-06, 3.3349e-05]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0051, 0.0019, 0.0058, 0.0154, 0.0246, 0.0347, -0.0265, 0.0097, + -0.0339, -0.0192], device='cuda:0'), grad: tensor([ 1.3912e-04, -3.9041e-05, 9.0420e-05, 1.6725e-04, 3.3438e-05, + -4.0792e-06, 3.8654e-05, -3.7044e-05, -5.2691e-04, 1.3828e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 217.57, cls_loss 0.0110 cls_loss_mapping 0.0143 cls_loss_causal 0.6278 re_mapping 0.0127 re_causal 0.0334 /// teacc 98.75 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0100, 0.0893, 0.0541, ..., -0.1078, -0.0785, 0.0264], + [ 0.0843, -0.0124, -0.0352, ..., 0.0456, 0.0680, -0.0881], + [-0.0456, -0.0198, -0.0315, ..., -0.0512, -0.0107, -0.0245], + ..., + [-0.0257, 0.1007, -0.1167, ..., -0.0767, 0.0141, -0.0837], + [ 0.0009, -0.0604, 0.0286, ..., 0.0429, -0.0855, -0.0340], + [-0.0181, -0.0478, 0.0341, ..., -0.0499, -0.0342, 0.0821]], + device='cuda:0'), grad: tensor([[ 3.7951e-07, -1.3128e-05, 4.3437e-06, ..., 2.1562e-05, + 5.7340e-05, -4.5672e-06], + [-2.3112e-05, 3.4012e-06, 1.8254e-06, ..., -3.5912e-05, + -1.5378e-05, 1.0030e-06], + [ 6.8359e-06, 1.2085e-05, 3.5465e-05, ..., 3.3677e-05, + 2.5198e-05, 2.0247e-06], + ..., + [ 6.5081e-06, -7.9632e-05, 3.0071e-05, ..., 3.2842e-05, + -2.0891e-05, -1.4484e-05], + [-1.2845e-05, 1.1995e-05, 1.7673e-05, ..., -1.7375e-05, + 2.6673e-05, -1.0766e-05], + [ 4.9286e-06, 4.4644e-05, 1.1519e-05, ..., 3.2604e-05, + 7.8857e-05, 6.8061e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0048, 0.0018, 0.0056, 0.0154, 0.0250, 0.0343, -0.0243, 0.0093, + -0.0347, -0.0191], device='cuda:0'), grad: tensor([ 3.8791e-04, 3.2727e-06, 1.8442e-04, -2.7347e-04, -1.5287e-03, + 1.7726e-04, 4.8971e-04, -3.8713e-05, 1.0628e-04, 4.9210e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 217.34, cls_loss 0.0108 cls_loss_mapping 0.0158 cls_loss_causal 0.6251 re_mapping 0.0123 re_causal 0.0342 /// teacc 98.72 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0095, 0.0906, 0.0545, ..., -0.1085, -0.0787, 0.0266], + [ 0.0850, -0.0126, -0.0358, ..., 0.0467, 0.0695, -0.0885], + [-0.0464, -0.0183, -0.0318, ..., -0.0525, -0.0116, -0.0248], + ..., + [-0.0261, 0.1003, -0.1172, ..., -0.0774, 0.0137, -0.0843], + [-0.0003, -0.0622, 0.0275, ..., 0.0421, -0.0870, -0.0353], + [-0.0181, -0.0482, 0.0350, ..., -0.0495, -0.0343, 0.0829]], + device='cuda:0'), grad: tensor([[ 7.7903e-05, -1.5050e-06, 5.2266e-06, ..., 9.4891e-05, + 2.4796e-07, 3.6675e-06], + [ 3.2759e-04, -1.5914e-05, 7.4133e-07, ..., 3.9077e-04, + -1.3627e-05, 4.1490e-07], + [ 6.3479e-05, 4.7833e-06, 1.7658e-06, ..., 7.8201e-05, + 1.2694e-06, 9.2201e-07], + ..., + [ 1.9237e-05, -1.7869e-04, 9.7871e-05, ..., 2.8521e-05, + 9.5218e-06, 8.8394e-05], + [ 6.4611e-04, 2.9560e-06, -8.2478e-06, ..., 7.6485e-04, + 6.5845e-07, 1.0379e-05], + [ 8.9407e-06, -1.0836e-04, -2.8348e-04, ..., 1.5482e-05, + 8.3819e-07, -2.6131e-04]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0052, 0.0024, 0.0058, 0.0151, 0.0246, 0.0343, -0.0240, 0.0087, + -0.0357, -0.0185], device='cuda:0'), grad: tensor([ 0.0002, 0.0008, 0.0002, 0.0003, 0.0010, 0.0004, -0.0032, -0.0008, + 0.0015, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 217.63, cls_loss 0.0111 cls_loss_mapping 0.0158 cls_loss_causal 0.6414 re_mapping 0.0120 re_causal 0.0342 /// teacc 98.79 lr 0.00010000 +Epoch 82, weight, value: tensor([[-9.8741e-03, 9.1382e-02, 5.4812e-02, ..., -1.0929e-01, + -7.9147e-02, 2.6775e-02], + [ 8.5780e-02, -1.1860e-02, -3.6087e-02, ..., 4.7225e-02, + 7.1365e-02, -8.8860e-02], + [-4.7703e-02, -1.8855e-02, -3.1960e-02, ..., -5.2953e-02, + -1.3034e-02, -2.5070e-02], + ..., + [-2.5541e-02, 1.0060e-01, -1.1791e-01, ..., -7.7870e-02, + 1.3934e-02, -8.5318e-02], + [ 1.1617e-04, -6.1831e-02, 2.7713e-02, ..., 4.2588e-02, + -8.7229e-02, -3.5074e-02], + [-1.8341e-02, -4.9051e-02, 3.5217e-02, ..., -5.0428e-02, + -3.4711e-02, 8.3518e-02]], device='cuda:0'), grad: tensor([[ 7.0445e-06, -1.6227e-05, 2.1551e-06, ..., 4.8652e-06, + 1.6242e-06, -8.4564e-07], + [-1.5959e-05, 3.1814e-06, 8.4378e-07, ..., -1.7837e-05, + -1.3530e-05, 1.3821e-06], + [ 4.3474e-06, -9.6202e-05, 1.0515e-06, ..., 4.3549e-06, + 1.3281e-06, 1.2033e-06], + ..., + [ 5.9269e-06, -1.6248e-04, 5.2676e-06, ..., 7.1302e-06, + 4.7274e-06, -6.4313e-05], + [ 5.5104e-05, 8.3074e-06, 2.0027e-05, ..., 2.2218e-05, + 8.6501e-06, 2.0131e-05], + [ 4.5560e-06, 2.3437e-04, -1.3232e-05, ..., 4.9323e-06, + 5.9977e-07, 5.3346e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0055, 0.0031, 0.0052, 0.0148, 0.0253, 0.0340, -0.0248, 0.0091, + -0.0353, -0.0190], device='cuda:0'), grad: tensor([-1.2301e-05, -6.9141e-06, -5.9891e-04, 2.2605e-05, 4.0740e-05, + 1.3638e-04, -2.0611e-04, 3.5286e-05, 1.0598e-04, 4.8208e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 217.35, cls_loss 0.0097 cls_loss_mapping 0.0159 cls_loss_causal 0.6430 re_mapping 0.0126 re_causal 0.0342 /// teacc 98.81 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0101, 0.0917, 0.0546, ..., -0.1099, -0.0802, 0.0266], + [ 0.0855, -0.0111, -0.0363, ..., 0.0463, 0.0706, -0.0892], + [-0.0472, -0.0190, -0.0327, ..., -0.0518, -0.0115, -0.0254], + ..., + [-0.0259, 0.1012, -0.1194, ..., -0.0784, 0.0135, -0.0856], + [ 0.0004, -0.0628, 0.0282, ..., 0.0431, -0.0878, -0.0352], + [-0.0189, -0.0497, 0.0354, ..., -0.0509, -0.0355, 0.0838]], + device='cuda:0'), grad: tensor([[ 2.7612e-05, -1.7613e-05, 1.4925e-04, ..., 1.8120e-05, + 1.8060e-05, -1.2212e-05], + [ 8.4221e-05, 3.6025e-04, 8.7768e-06, ..., 7.6517e-06, + 1.8254e-05, 9.0711e-07], + [-1.7762e-05, 1.0520e-05, 3.5334e-04, ..., 3.9935e-06, + -8.5533e-06, 7.3947e-07], + ..., + [-8.3625e-05, -5.2166e-04, 2.1219e-05, ..., 1.2055e-05, + -2.3082e-05, 1.4100e-06], + [ 4.1038e-05, 1.2636e-05, 1.4596e-05, ..., 6.0499e-05, + 9.8497e-06, 6.9924e-06], + [ 1.5661e-05, 4.2439e-05, 3.4034e-05, ..., -8.1444e-04, + 8.0988e-06, -2.0713e-06]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0051, 0.0025, 0.0063, 0.0144, 0.0256, 0.0349, -0.0250, 0.0087, + -0.0351, -0.0193], device='cuda:0'), grad: tensor([ 0.0006, 0.0005, 0.0011, -0.0021, 0.0037, -0.0044, 0.0042, -0.0006, + 0.0002, -0.0032], device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 217.01, cls_loss 0.0084 cls_loss_mapping 0.0114 cls_loss_causal 0.6572 re_mapping 0.0119 re_causal 0.0344 /// teacc 98.65 lr 0.00010000 +Epoch 84, weight, value: tensor([[-1.0296e-02, 9.2249e-02, 5.4699e-02, ..., -1.1042e-01, + -8.0827e-02, 2.6515e-02], + [ 8.6181e-02, -9.8126e-03, -3.6275e-02, ..., 4.6982e-02, + 7.1194e-02, -8.9245e-02], + [-4.7421e-02, -1.9725e-02, -3.3079e-02, ..., -5.2103e-02, + -1.1407e-02, -2.5273e-02], + ..., + [-2.6926e-02, 1.0159e-01, -1.1968e-01, ..., -8.0110e-02, + 1.3139e-02, -8.6026e-02], + [ 4.3944e-05, -6.3522e-02, 2.7892e-02, ..., 4.3039e-02, + -8.8442e-02, -3.5265e-02], + [-1.9875e-02, -5.0545e-02, 3.5718e-02, ..., -5.0521e-02, + -3.5832e-02, 8.4657e-02]], device='cuda:0'), grad: tensor([[ 7.4096e-06, -1.6883e-05, -8.7172e-07, ..., 7.0035e-06, + 1.6969e-06, -3.0873e-07], + [-8.4734e-04, 5.9977e-07, 1.0896e-06, ..., -9.0361e-04, + -1.7090e-03, 1.4454e-06], + [ 6.4802e-04, 3.6452e-06, -3.6918e-06, ..., 6.8808e-04, + 1.2655e-03, 4.6454e-06], + ..., + [ 2.0027e-04, 6.1933e-07, 6.7875e-06, ..., 3.0947e-04, + 3.9220e-04, 9.4995e-06], + [ 4.3929e-05, -7.4273e-08, 7.9930e-05, ..., 3.4839e-05, + -4.0457e-06, 7.1824e-05], + [ 3.7074e-05, 7.0333e-06, 3.8296e-05, ..., 8.1182e-05, + 1.4175e-06, 5.4538e-05]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0048, 0.0030, 0.0062, 0.0145, 0.0245, 0.0348, -0.0246, 0.0087, + -0.0355, -0.0187], device='cuda:0'), grad: tensor([-2.1338e-05, -1.9894e-03, 1.4706e-03, 5.0688e-04, -3.0947e-04, + -7.4053e-04, -2.6766e-06, 6.7377e-04, 1.8239e-04, 2.2733e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 217.40, cls_loss 0.0072 cls_loss_mapping 0.0099 cls_loss_causal 0.6295 re_mapping 0.0117 re_causal 0.0352 /// teacc 98.79 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0100, 0.0928, 0.0543, ..., -0.1108, -0.0829, 0.0260], + [ 0.0865, -0.0101, -0.0364, ..., 0.0473, 0.0713, -0.0896], + [-0.0479, -0.0206, -0.0334, ..., -0.0526, -0.0116, -0.0256], + ..., + [-0.0274, 0.1022, -0.1208, ..., -0.0806, 0.0136, -0.0862], + [ 0.0002, -0.0638, 0.0279, ..., 0.0433, -0.0886, -0.0355], + [-0.0203, -0.0505, 0.0364, ..., -0.0508, -0.0353, 0.0853]], + device='cuda:0'), grad: tensor([[ 2.6450e-06, -5.2713e-07, 8.3148e-06, ..., 4.6529e-06, + 1.4110e-06, 7.6815e-06], + [-1.2726e-05, 5.9232e-07, 2.9560e-06, ..., -2.3365e-05, + -8.6874e-06, 1.6410e-06], + [ 3.3751e-06, 3.2410e-06, 1.9208e-05, ..., 3.7197e-06, + 1.6699e-06, 2.1607e-06], + ..., + [ 1.4424e-05, -1.6782e-06, 4.7445e-05, ..., 1.5102e-05, + 6.6720e-06, 2.5392e-05], + [ 1.8314e-05, 5.3421e-06, 4.2319e-05, ..., 1.7241e-05, + 1.3314e-05, 2.1994e-05], + [ 9.2566e-05, 5.5619e-06, 1.4699e-04, ..., 9.3162e-05, + 8.1658e-06, 2.7394e-04]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0046, 0.0027, 0.0058, 0.0147, 0.0243, 0.0349, -0.0248, 0.0089, + -0.0356, -0.0182], device='cuda:0'), grad: tensor([ 1.6943e-05, -1.6868e-05, 3.8475e-05, 1.8730e-03, 2.8849e-05, + -2.5635e-03, 2.8133e-05, 9.6917e-05, 1.0794e-04, 3.8934e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 217.03, cls_loss 0.0080 cls_loss_mapping 0.0105 cls_loss_causal 0.6125 re_mapping 0.0114 re_causal 0.0329 /// teacc 98.83 lr 0.00010000 +Epoch 86, weight, value: tensor([[-1.0266e-02, 9.3217e-02, 5.4418e-02, ..., -1.1161e-01, + -8.3075e-02, 2.5764e-02], + [ 8.7239e-02, -1.0392e-02, -3.5655e-02, ..., 4.8504e-02, + 7.1378e-02, -8.9176e-02], + [-4.8153e-02, -2.0935e-02, -3.3708e-02, ..., -5.2898e-02, + -1.1542e-02, -2.5909e-02], + ..., + [-2.7471e-02, 1.0312e-01, -1.2112e-01, ..., -8.0917e-02, + 1.3959e-02, -8.7156e-02], + [-9.8744e-05, -6.4129e-02, 2.8179e-02, ..., 4.3266e-02, + -8.8892e-02, -3.5836e-02], + [-2.1463e-02, -5.1019e-02, 3.6275e-02, ..., -5.2047e-02, + -3.5488e-02, 8.5811e-02]], device='cuda:0'), grad: tensor([[-5.1260e-06, -4.2707e-05, -3.2216e-05, ..., -2.6748e-06, + 1.8813e-07, -4.8190e-05], + [-8.3353e-07, -1.3439e-06, 2.7250e-06, ..., -1.9316e-06, + -1.3784e-06, 3.3118e-06], + [ 8.3521e-06, 5.9120e-06, 1.6075e-06, ..., 5.6922e-06, + 5.8673e-08, 1.8571e-06], + ..., + [ 4.1336e-05, 1.1042e-05, 3.9279e-05, ..., 3.2961e-05, + 1.0990e-05, 7.0333e-05], + [-9.5814e-06, 2.2408e-06, 2.0802e-04, ..., 1.5765e-05, + 7.2643e-07, 2.1946e-04], + [-1.4946e-05, -1.8612e-05, -3.5238e-04, ..., -7.5161e-05, + -2.3410e-05, -4.3917e-04]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0042, 0.0029, 0.0060, 0.0136, 0.0243, 0.0358, -0.0242, 0.0093, + -0.0358, -0.0187], device='cuda:0'), grad: tensor([-6.3181e-05, 2.7180e-05, -1.3137e-04, -1.9848e-04, 3.6025e-04, + 2.1350e-04, 9.5546e-05, 2.9778e-04, 5.6601e-04, -1.1673e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 217.61, cls_loss 0.0078 cls_loss_mapping 0.0114 cls_loss_causal 0.5988 re_mapping 0.0116 re_causal 0.0325 /// teacc 98.81 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0104, 0.0943, 0.0554, ..., -0.1120, -0.0833, 0.0262], + [ 0.0875, -0.0101, -0.0365, ..., 0.0483, 0.0715, -0.0904], + [-0.0489, -0.0214, -0.0342, ..., -0.0538, -0.0115, -0.0264], + ..., + [-0.0281, 0.1035, -0.1215, ..., -0.0815, 0.0137, -0.0877], + [ 0.0009, -0.0635, 0.0284, ..., 0.0441, -0.0892, -0.0356], + [-0.0212, -0.0516, 0.0369, ..., -0.0518, -0.0350, 0.0869]], + device='cuda:0'), grad: tensor([[ 4.1090e-06, 2.1476e-06, -2.8033e-07, ..., 5.6587e-06, + 1.9046e-07, -4.0093e-07], + [-7.1153e-06, 2.0694e-06, 5.3691e-07, ..., -5.7854e-06, + -2.3060e-06, 2.2352e-07], + [ 7.0855e-06, 1.6659e-05, 6.5751e-07, ..., 8.5458e-06, + -1.8692e-06, 2.6450e-07], + ..., + [ 3.4980e-06, -3.9101e-05, 1.3951e-06, ..., 8.7991e-06, + 2.9840e-06, 1.6354e-06], + [ 2.6047e-05, 2.4103e-06, 4.5866e-05, ..., 4.0442e-05, + 2.5835e-06, 1.0294e-04], + [ 1.0878e-05, 6.6720e-06, 1.2167e-05, ..., 2.8193e-05, + 8.0932e-07, 1.6972e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0050, 0.0028, 0.0055, 0.0143, 0.0242, 0.0353, -0.0255, 0.0091, + -0.0351, -0.0183], device='cuda:0'), grad: tensor([ 1.3091e-05, 3.9637e-06, 5.7556e-07, -3.2425e-05, -3.4750e-05, + -1.1396e-04, -4.4238e-07, -5.7444e-06, 1.0848e-04, 6.1452e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 217.72, cls_loss 0.0074 cls_loss_mapping 0.0093 cls_loss_causal 0.6044 re_mapping 0.0114 re_causal 0.0332 /// teacc 98.72 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0104, 0.0951, 0.0553, ..., -0.1133, -0.0849, 0.0259], + [ 0.0879, -0.0102, -0.0366, ..., 0.0487, 0.0716, -0.0906], + [-0.0487, -0.0213, -0.0330, ..., -0.0535, -0.0108, -0.0265], + ..., + [-0.0286, 0.1037, -0.1220, ..., -0.0825, 0.0136, -0.0890], + [ 0.0015, -0.0629, 0.0287, ..., 0.0449, -0.0898, -0.0348], + [-0.0215, -0.0521, 0.0373, ..., -0.0521, -0.0353, 0.0874]], + device='cuda:0'), grad: tensor([[ 5.7608e-05, 1.6168e-05, 4.6790e-06, ..., 4.8310e-05, + 1.1940e-06, 8.3307e-07], + [ 3.2842e-05, 7.1041e-06, 4.2655e-06, ..., 4.2140e-05, + -5.9791e-07, 2.4810e-06], + [ 5.3197e-05, 1.7490e-06, 3.1441e-05, ..., 3.6120e-05, + 1.8492e-05, 4.9593e-07], + ..., + [ 5.5991e-06, -1.4864e-05, 5.4985e-06, ..., 3.8683e-05, + 4.7646e-06, 3.2224e-06], + [ 1.0532e-04, 4.0799e-05, -5.6863e-05, ..., 4.8697e-05, + 5.1707e-06, -7.4744e-05], + [ 4.8816e-05, 8.6576e-06, 6.9320e-05, ..., 1.7290e-03, + 1.4789e-06, 1.9038e-04]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0049, 0.0027, 0.0068, 0.0136, 0.0243, 0.0350, -0.0255, 0.0088, + -0.0345, -0.0185], device='cuda:0'), grad: tensor([ 1.3769e-04, 1.2314e-04, 1.6916e-04, -1.6093e-04, -3.6316e-03, + 8.6308e-05, -7.8535e-04, 8.0168e-05, 2.3484e-04, 3.7441e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 217.26, cls_loss 0.0073 cls_loss_mapping 0.0096 cls_loss_causal 0.5807 re_mapping 0.0113 re_causal 0.0307 /// teacc 98.83 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0107, 0.0958, 0.0555, ..., -0.1142, -0.0850, 0.0260], + [ 0.0885, -0.0102, -0.0369, ..., 0.0491, 0.0727, -0.0909], + [-0.0489, -0.0215, -0.0330, ..., -0.0532, -0.0113, -0.0271], + ..., + [-0.0289, 0.1038, -0.1221, ..., -0.0834, 0.0133, -0.0888], + [ 0.0013, -0.0632, 0.0285, ..., 0.0449, -0.0903, -0.0354], + [-0.0222, -0.0534, 0.0376, ..., -0.0527, -0.0355, 0.0878]], + device='cuda:0'), grad: tensor([[ 6.2920e-06, -9.9465e-07, 1.7565e-06, ..., 7.7859e-06, + 2.4168e-07, 1.8403e-06], + [-2.2203e-05, 3.1758e-07, 7.6974e-07, ..., -1.0230e-05, + -2.3291e-05, 6.6822e-07], + [ 4.9919e-05, 4.1910e-07, 3.3751e-06, ..., 7.7307e-05, + 1.6123e-05, 9.1409e-07], + ..., + [ 5.2266e-06, -4.1351e-07, 6.7987e-07, ..., 8.1509e-06, + 3.8259e-06, 5.9186e-07], + [-1.6168e-05, 2.9430e-06, -3.4958e-05, ..., -2.3022e-05, + 1.2945e-06, -1.1824e-05], + [ 8.8364e-06, 5.4343e-07, 7.9125e-06, ..., 1.0565e-05, + 8.5728e-07, 5.3942e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0050, 0.0033, 0.0068, 0.0130, 0.0252, 0.0356, -0.0255, 0.0085, + -0.0351, -0.0191], device='cuda:0'), grad: tensor([ 1.4596e-05, -4.2409e-05, 1.5056e-04, 3.0577e-05, 3.3951e-04, + 3.0667e-05, -5.1451e-04, 1.7881e-05, -5.1022e-05, 2.4498e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 88---------------------------------------------------- +epoch 88, time 218.18, cls_loss 0.0058 cls_loss_mapping 0.0090 cls_loss_causal 0.5834 re_mapping 0.0115 re_causal 0.0322 /// teacc 98.89 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0120, 0.0956, 0.0559, ..., -0.1162, -0.0873, 0.0258], + [ 0.0885, -0.0103, -0.0370, ..., 0.0493, 0.0731, -0.0911], + [-0.0492, -0.0221, -0.0332, ..., -0.0536, -0.0115, -0.0274], + ..., + [-0.0286, 0.1045, -0.1226, ..., -0.0833, 0.0133, -0.0895], + [ 0.0013, -0.0635, 0.0284, ..., 0.0451, -0.0906, -0.0356], + [-0.0226, -0.0538, 0.0377, ..., -0.0529, -0.0357, 0.0882]], + device='cuda:0'), grad: tensor([[ 1.0297e-05, -3.5204e-06, -3.4273e-07, ..., 8.7991e-06, + 2.1607e-07, 1.8533e-06], + [-2.9653e-05, 1.3812e-06, 1.9539e-06, ..., -4.7356e-05, + -2.4945e-05, 2.1569e-06], + [ 1.1601e-05, 2.6733e-05, 3.3546e-06, ..., 1.2681e-05, + 4.2021e-06, 2.8089e-06], + ..., + [ 4.5419e-05, -3.4332e-05, 2.4214e-05, ..., 4.3333e-05, + 9.6709e-06, 2.3663e-05], + [ 4.9204e-05, 7.0361e-07, 2.2814e-05, ..., 4.1485e-05, + 3.5148e-06, 2.6822e-05], + [ 1.8075e-05, 4.2617e-06, 1.3318e-07, ..., 1.5751e-05, + 1.7527e-06, 2.2855e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0045, 0.0029, 0.0065, 0.0131, 0.0251, 0.0358, -0.0249, 0.0090, + -0.0353, -0.0192], device='cuda:0'), grad: tensor([ 2.0474e-05, -6.5982e-05, 8.4996e-05, 9.5889e-06, 2.2545e-05, + -3.9673e-04, 5.4568e-05, 5.9754e-05, 1.6046e-04, 5.0098e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 217.02, cls_loss 0.0056 cls_loss_mapping 0.0090 cls_loss_causal 0.5937 re_mapping 0.0112 re_causal 0.0314 /// teacc 98.82 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0122, 0.0965, 0.0567, ..., -0.1159, -0.0872, 0.0268], + [ 0.0891, -0.0095, -0.0375, ..., 0.0500, 0.0736, -0.0920], + [-0.0493, -0.0224, -0.0335, ..., -0.0538, -0.0117, -0.0277], + ..., + [-0.0294, 0.1044, -0.1235, ..., -0.0847, 0.0127, -0.0899], + [ 0.0013, -0.0637, 0.0283, ..., 0.0451, -0.0907, -0.0360], + [-0.0227, -0.0543, 0.0383, ..., -0.0525, -0.0354, 0.0892]], + device='cuda:0'), grad: tensor([[ 2.7996e-06, -3.0417e-06, 3.4366e-07, ..., 5.7667e-06, + 8.5868e-07, 2.9569e-07], + [-6.1572e-05, 9.1866e-06, 4.3847e-06, ..., -1.1069e-04, + -3.7163e-05, 3.0026e-06], + [ 6.0678e-05, -1.4710e-04, 1.2502e-05, ..., 1.0681e-04, + 3.1948e-05, 1.9707e-06], + ..., + [ 8.0317e-06, 1.2553e-04, 6.9477e-06, ..., 1.7300e-05, + 1.6987e-06, 4.4331e-06], + [ 4.4793e-05, 4.8503e-06, 5.3436e-05, ..., 6.4850e-05, + 2.5518e-06, 3.9250e-05], + [ 2.7061e-05, 2.5667e-06, 2.6584e-05, ..., 4.7445e-05, + 3.7253e-07, 1.6823e-05]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0053, 0.0035, 0.0065, 0.0133, 0.0247, 0.0356, -0.0251, 0.0083, + -0.0355, -0.0188], device='cuda:0'), grad: tensor([ 8.7842e-06, -2.0802e-04, -6.1464e-04, 7.8499e-05, -8.2910e-05, + -2.5034e-04, -1.9506e-05, 8.4162e-04, 1.6427e-04, 8.2910e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 90---------------------------------------------------- +epoch 90, time 217.68, cls_loss 0.0084 cls_loss_mapping 0.0124 cls_loss_causal 0.6120 re_mapping 0.0107 re_causal 0.0299 /// teacc 98.90 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0124, 0.0976, 0.0570, ..., -0.1167, -0.0875, 0.0270], + [ 0.0891, -0.0101, -0.0378, ..., 0.0497, 0.0736, -0.0921], + [-0.0489, -0.0230, -0.0335, ..., -0.0530, -0.0117, -0.0281], + ..., + [-0.0301, 0.1058, -0.1246, ..., -0.0851, 0.0143, -0.0903], + [ 0.0018, -0.0643, 0.0284, ..., 0.0454, -0.0914, -0.0362], + [-0.0239, -0.0549, 0.0389, ..., -0.0524, -0.0363, 0.0896]], + device='cuda:0'), grad: tensor([[ 3.6173e-06, -1.8895e-05, -7.5903e-07, ..., 1.5525e-06, + 1.7673e-05, -2.3752e-05], + [-1.3880e-05, 7.1712e-06, 5.0068e-06, ..., -2.3186e-05, + -1.4953e-05, 2.7381e-06], + [ 9.7454e-06, 7.4655e-06, -7.2360e-05, ..., 1.6600e-05, + -9.6321e-05, 5.7481e-06], + ..., + [ 3.0156e-06, -2.2423e-04, -7.9155e-05, ..., 4.8801e-06, + 2.6152e-05, -5.2780e-05], + [-5.4449e-05, 2.8834e-05, 1.4842e-05, ..., -6.3837e-05, + 3.4515e-06, 9.4250e-06], + [ 8.1733e-06, 1.7118e-04, 9.1016e-05, ..., 1.4223e-05, + 1.7688e-05, 4.6641e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0056, 0.0026, 0.0071, 0.0130, 0.0245, 0.0357, -0.0257, 0.0086, + -0.0349, -0.0186], device='cuda:0'), grad: tensor([-7.8455e-06, -5.0850e-06, -1.9073e-04, 6.1929e-05, 1.3039e-05, + 5.5194e-05, 6.0260e-05, -3.5691e-04, -2.5854e-06, 3.7289e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 216.99, cls_loss 0.0086 cls_loss_mapping 0.0121 cls_loss_causal 0.6155 re_mapping 0.0109 re_causal 0.0293 /// teacc 98.84 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0123, 0.0979, 0.0571, ..., -0.1177, -0.0874, 0.0263], + [ 0.0894, -0.0103, -0.0381, ..., 0.0502, 0.0748, -0.0927], + [-0.0494, -0.0240, -0.0337, ..., -0.0537, -0.0121, -0.0286], + ..., + [-0.0298, 0.1067, -0.1251, ..., -0.0853, 0.0136, -0.0895], + [ 0.0020, -0.0654, 0.0284, ..., 0.0459, -0.0917, -0.0364], + [-0.0247, -0.0545, 0.0396, ..., -0.0531, -0.0367, 0.0910]], + device='cuda:0'), grad: tensor([[ 2.0131e-05, 1.0468e-05, 4.7013e-06, ..., 1.4156e-05, + 1.9036e-06, 5.6922e-06], + [ 8.1211e-06, 2.7999e-05, 3.4142e-06, ..., -2.8424e-06, + -1.8580e-06, 3.8929e-07], + [ 1.5661e-05, 4.4778e-06, 1.2569e-05, ..., 1.4246e-05, + -2.9970e-06, 7.5670e-07], + ..., + [-9.2760e-06, -5.8353e-05, 5.2750e-06, ..., 5.9605e-06, + 1.2824e-06, 7.9907e-07], + [-9.5367e-07, 2.0768e-06, 6.1356e-06, ..., -2.8446e-05, + 1.7285e-06, 6.1840e-07], + [ 1.2249e-05, 2.4885e-05, 6.1169e-06, ..., 6.2510e-06, + 9.2480e-07, 1.9968e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0046, 0.0026, 0.0060, 0.0128, 0.0243, 0.0353, -0.0256, 0.0096, + -0.0345, -0.0180], device='cuda:0'), grad: tensor([ 4.9710e-05, 6.1989e-05, 2.8670e-05, -1.9989e-03, 7.0743e-06, + 1.9064e-03, -3.5554e-05, -8.0168e-05, -8.9630e-06, 6.9201e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 217.13, cls_loss 0.0060 cls_loss_mapping 0.0106 cls_loss_causal 0.5980 re_mapping 0.0103 re_causal 0.0308 /// teacc 98.83 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0127, 0.0982, 0.0573, ..., -0.1186, -0.0872, 0.0265], + [ 0.0897, -0.0105, -0.0384, ..., 0.0503, 0.0751, -0.0930], + [-0.0500, -0.0244, -0.0337, ..., -0.0540, -0.0124, -0.0287], + ..., + [-0.0302, 0.1079, -0.1255, ..., -0.0862, 0.0133, -0.0906], + [ 0.0022, -0.0659, 0.0286, ..., 0.0466, -0.0921, -0.0360], + [-0.0251, -0.0555, 0.0397, ..., -0.0547, -0.0373, 0.0903]], + device='cuda:0'), grad: tensor([[ 1.7300e-05, -3.9153e-06, 6.3069e-06, ..., 3.9153e-06, + 7.5437e-06, -7.2643e-06], + [-2.6315e-05, 3.1479e-07, 2.3134e-06, ..., -3.9160e-05, + -1.2182e-05, 1.2666e-06], + [ 4.5061e-05, 8.2999e-06, 3.4332e-05, ..., 1.3210e-05, + -1.2420e-05, 3.8277e-07], + ..., + [ 2.0579e-05, -3.6061e-06, 2.5079e-05, ..., 2.3678e-05, + 8.9183e-06, 2.1160e-05], + [-1.5700e-04, 3.6731e-06, -1.2851e-04, ..., -3.1978e-05, + 5.7705e-06, 9.3877e-06], + [ 4.1258e-07, -1.2264e-05, -7.6473e-05, ..., 2.2016e-06, + 8.6380e-07, -8.5115e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0045, 0.0025, 0.0057, 0.0136, 0.0249, 0.0350, -0.0250, 0.0097, + -0.0342, -0.0193], device='cuda:0'), grad: tensor([ 1.5998e-04, -2.8551e-05, -1.9407e-04, 3.1137e-04, 9.2566e-05, + 3.8534e-05, -1.2405e-06, 6.1631e-05, -2.9635e-04, -1.4400e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 93---------------------------------------------------- +epoch 93, time 217.87, cls_loss 0.0058 cls_loss_mapping 0.0095 cls_loss_causal 0.6052 re_mapping 0.0102 re_causal 0.0312 /// teacc 98.95 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0131, 0.0986, 0.0572, ..., -0.1193, -0.0876, 0.0266], + [ 0.0901, -0.0106, -0.0385, ..., 0.0507, 0.0753, -0.0933], + [-0.0497, -0.0248, -0.0339, ..., -0.0538, -0.0119, -0.0290], + ..., + [-0.0305, 0.1086, -0.1261, ..., -0.0867, 0.0130, -0.0914], + [ 0.0017, -0.0667, 0.0279, ..., 0.0459, -0.0925, -0.0375], + [-0.0252, -0.0561, 0.0404, ..., -0.0549, -0.0374, 0.0910]], + device='cuda:0'), grad: tensor([[ 1.2359e-06, -1.8075e-05, -2.8446e-05, ..., 1.2051e-06, + -1.1466e-05, -2.5585e-05], + [-8.4829e-04, -5.2881e-04, 9.2620e-07, ..., -8.3494e-04, + -2.0754e-04, 6.9709e-07], + [ 2.0400e-05, 1.7479e-05, 4.5486e-06, ..., 1.8254e-05, + 3.0234e-05, 3.6974e-06], + ..., + [ 6.2704e-04, 3.7241e-04, 1.6522e-06, ..., 6.1798e-04, + 1.3912e-04, 1.8086e-06], + [ 1.4037e-05, 1.1742e-05, 5.4017e-06, ..., 1.4015e-05, + 5.1744e-06, 1.0625e-05], + [ 1.5724e-04, 1.1396e-04, 9.4995e-06, ..., 1.5748e-04, + 2.8402e-05, 9.8497e-06]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0044, 0.0025, 0.0062, 0.0135, 0.0251, 0.0351, -0.0249, 0.0098, + -0.0350, -0.0193], device='cuda:0'), grad: tensor([-8.0526e-05, -2.0885e-03, 6.7115e-05, 4.6015e-05, 6.6102e-05, + -2.6584e-05, 4.0114e-05, 1.5030e-03, 5.9217e-05, 4.1389e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 217.29, cls_loss 0.0072 cls_loss_mapping 0.0108 cls_loss_causal 0.6246 re_mapping 0.0104 re_causal 0.0296 /// teacc 98.93 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0131, 0.1002, 0.0583, ..., -0.1192, -0.0877, 0.0274], + [ 0.0910, -0.0100, -0.0389, ..., 0.0514, 0.0757, -0.0946], + [-0.0501, -0.0252, -0.0343, ..., -0.0541, -0.0122, -0.0296], + ..., + [-0.0317, 0.1091, -0.1273, ..., -0.0881, 0.0130, -0.0921], + [ 0.0020, -0.0678, 0.0281, ..., 0.0465, -0.0916, -0.0376], + [-0.0261, -0.0568, 0.0407, ..., -0.0556, -0.0382, 0.0916]], + device='cuda:0'), grad: tensor([[ 1.0338e-06, -3.8594e-06, -1.0096e-06, ..., 1.3225e-06, + 4.6100e-08, -2.2911e-06], + [ 1.9297e-06, 8.2031e-06, 3.8221e-06, ..., 5.0142e-06, + -5.4389e-07, 3.5111e-07], + [ 8.5160e-06, 5.7757e-05, 9.4995e-06, ..., 5.4464e-06, + 2.3795e-07, 7.9209e-07], + ..., + [ 1.5851e-06, -1.3125e-04, -8.2795e-07, ..., 4.0494e-06, + 2.3236e-07, 6.2631e-07], + [-5.9186e-07, 5.2340e-06, -4.1835e-06, ..., -2.5444e-06, + 1.0710e-08, -7.6229e-07], + [ 1.5935e-06, 1.3046e-05, 3.4999e-06, ..., 1.5959e-05, + 1.0664e-07, 2.0713e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0050, 0.0028, 0.0062, 0.0130, 0.0248, 0.0354, -0.0250, 0.0097, + -0.0351, -0.0193], device='cuda:0'), grad: tensor([-1.3672e-06, 2.2426e-05, 8.3208e-05, 4.3422e-05, -6.3360e-05, + -1.0595e-05, -1.3649e-05, -1.3638e-04, 1.7837e-05, 5.8651e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 95---------------------------------------------------- +epoch 95, time 218.73, cls_loss 0.0065 cls_loss_mapping 0.0098 cls_loss_causal 0.5547 re_mapping 0.0104 re_causal 0.0286 /// teacc 98.98 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0155, 0.0994, 0.0587, ..., -0.1206, -0.0905, 0.0265], + [ 0.0913, -0.0097, -0.0397, ..., 0.0514, 0.0765, -0.0953], + [-0.0507, -0.0249, -0.0348, ..., -0.0545, -0.0125, -0.0296], + ..., + [-0.0324, 0.1091, -0.1277, ..., -0.0887, 0.0128, -0.0926], + [ 0.0022, -0.0684, 0.0283, ..., 0.0468, -0.0932, -0.0377], + [-0.0268, -0.0572, 0.0412, ..., -0.0565, -0.0383, 0.0919]], + device='cuda:0'), grad: tensor([[ 1.3597e-06, -3.5688e-06, -4.3064e-06, ..., 1.3448e-06, + 1.1176e-07, -5.6587e-06], + [ 1.2137e-05, 1.0186e-04, 4.6846e-07, ..., -2.0444e-05, + -4.6156e-06, 3.8790e-07], + [ 7.1377e-06, 1.5646e-05, 3.7551e-06, ..., 4.8243e-06, + 1.4715e-07, 5.7463e-07], + ..., + [-1.4448e-04, -6.0987e-04, 1.7118e-06, ..., 2.2296e-06, + 3.6415e-07, 1.6298e-06], + [ 1.4395e-05, 8.6904e-05, -1.7360e-06, ..., -4.8317e-06, + 2.4140e-06, 7.5661e-06], + [ 3.6001e-05, 1.4269e-04, -8.1882e-06, ..., 3.6117e-06, + -1.7099e-06, -1.3180e-05]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0040, 0.0029, 0.0062, 0.0128, 0.0258, 0.0357, -0.0240, 0.0094, + -0.0352, -0.0201], device='cuda:0'), grad: tensor([ 5.7071e-06, 3.3998e-04, 6.6645e-06, 8.5926e-04, 6.2644e-05, + 1.6272e-05, 2.9698e-05, -2.0905e-03, 3.0065e-04, 4.6921e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 217.34, cls_loss 0.0059 cls_loss_mapping 0.0083 cls_loss_causal 0.5662 re_mapping 0.0103 re_causal 0.0302 /// teacc 98.83 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0161, 0.0994, 0.0589, ..., -0.1221, -0.0910, 0.0263], + [ 0.0914, -0.0110, -0.0408, ..., 0.0520, 0.0758, -0.0959], + [-0.0504, -0.0252, -0.0340, ..., -0.0540, -0.0113, -0.0299], + ..., + [-0.0316, 0.1106, -0.1280, ..., -0.0887, 0.0144, -0.0932], + [ 0.0019, -0.0689, 0.0285, ..., 0.0465, -0.0929, -0.0378], + [-0.0277, -0.0576, 0.0412, ..., -0.0571, -0.0395, 0.0923]], + device='cuda:0'), grad: tensor([[ 2.2771e-07, 4.1444e-07, 1.6158e-07, ..., 2.7567e-07, + 1.4296e-07, 2.1281e-07], + [-8.3493e-07, 3.3528e-07, 1.0384e-07, ..., -1.1558e-06, + -1.8487e-07, -1.1362e-07], + [ 5.4203e-07, -3.2857e-06, 4.3493e-07, ..., 2.1188e-07, + -1.6652e-06, 1.0617e-07], + ..., + [ 5.1642e-07, -1.9483e-06, 7.9488e-07, ..., 9.5554e-07, + 4.8522e-07, 1.2852e-06], + [-1.0934e-06, 8.8476e-08, 4.4703e-07, ..., -1.0803e-06, + 1.0990e-07, 1.2312e-06], + [ 1.1576e-06, 2.8927e-06, -1.0701e-06, ..., 1.9446e-06, + 6.1048e-07, -5.6475e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0037, 0.0023, 0.0073, 0.0125, 0.0253, 0.0356, -0.0238, 0.0110, + -0.0358, -0.0206], device='cuda:0'), grad: tensor([ 2.0582e-06, -3.8650e-08, -1.2361e-05, -2.4755e-06, 8.6753e-07, + 2.1346e-06, 1.4901e-07, 2.4103e-06, 4.3027e-07, 6.8434e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 218.27, cls_loss 0.0059 cls_loss_mapping 0.0100 cls_loss_causal 0.6173 re_mapping 0.0104 re_causal 0.0301 /// teacc 98.86 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0164, 0.0995, 0.0589, ..., -0.1237, -0.0912, 0.0263], + [ 0.0919, -0.0116, -0.0411, ..., 0.0525, 0.0768, -0.0962], + [-0.0514, -0.0254, -0.0337, ..., -0.0547, -0.0119, -0.0297], + ..., + [-0.0316, 0.1114, -0.1285, ..., -0.0890, 0.0142, -0.0939], + [ 0.0015, -0.0697, 0.0282, ..., 0.0466, -0.0937, -0.0381], + [-0.0283, -0.0581, 0.0414, ..., -0.0580, -0.0393, 0.0922]], + device='cuda:0'), grad: tensor([[-4.4331e-07, -2.7895e-05, -8.8960e-06, ..., 1.2666e-06, + 2.0070e-07, -1.0841e-06], + [-5.6401e-06, -3.1777e-06, 2.4103e-06, ..., -1.8571e-06, + -3.2969e-06, 2.7381e-06], + [ 1.9632e-06, 5.3532e-06, 2.0228e-06, ..., 2.8443e-06, + 3.0315e-07, 1.5367e-06], + ..., + [ 5.7109e-06, 3.0734e-06, 3.6657e-06, ..., 9.2760e-06, + 2.5034e-06, 7.5139e-06], + [-1.1340e-05, 1.6857e-06, -4.9965e-07, ..., -2.5705e-06, + 9.4343e-07, 1.3039e-05], + [-1.6969e-06, 3.0305e-06, -5.9158e-05, ..., -9.0742e-04, + -3.0473e-06, -1.2264e-03]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0033, 0.0022, 0.0071, 0.0125, 0.0263, 0.0360, -0.0234, 0.0113, + -0.0362, -0.0218], device='cuda:0'), grad: tensor([-6.5744e-05, 3.3267e-06, 5.8152e-06, 1.7151e-05, 4.1771e-03, + 1.3836e-05, 6.9022e-05, 3.3945e-05, 3.1084e-05, -4.2877e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 217.43, cls_loss 0.0044 cls_loss_mapping 0.0072 cls_loss_causal 0.5994 re_mapping 0.0102 re_causal 0.0298 /// teacc 98.85 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0168, 0.0996, 0.0589, ..., -0.1244, -0.0918, 0.0261], + [ 0.0919, -0.0116, -0.0413, ..., 0.0528, 0.0769, -0.0964], + [-0.0515, -0.0255, -0.0337, ..., -0.0550, -0.0120, -0.0294], + ..., + [-0.0320, 0.1118, -0.1293, ..., -0.0898, 0.0139, -0.0947], + [ 0.0017, -0.0703, 0.0281, ..., 0.0471, -0.0938, -0.0384], + [-0.0283, -0.0585, 0.0419, ..., -0.0583, -0.0397, 0.0928]], + device='cuda:0'), grad: tensor([[ 1.1727e-05, 4.3362e-06, 6.1132e-06, ..., 1.6242e-05, + 1.6391e-07, 1.7472e-06], + [-1.3781e-03, -1.5345e-03, 4.6939e-07, ..., -8.9502e-04, + -2.5947e-06, 3.0966e-07], + [ 6.9477e-06, 5.6028e-06, 1.9148e-06, ..., 6.7689e-06, + 3.6927e-07, 5.6671e-07], + ..., + [ 1.1168e-03, 1.2169e-03, 1.5132e-05, ..., 7.2432e-04, + 8.9174e-07, 2.1741e-05], + [ 9.6411e-06, 4.1686e-06, 6.0834e-06, ..., 1.1824e-05, + 5.1642e-07, 4.9621e-06], + [ 1.5175e-04, 1.9717e-04, -1.6063e-05, ..., 9.8825e-05, + 2.9989e-07, -2.7046e-05]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0029, 0.0023, 0.0071, 0.0126, 0.0264, 0.0363, -0.0235, 0.0111, + -0.0361, -0.0218], device='cuda:0'), grad: tensor([ 3.6180e-05, -3.8548e-03, 1.8761e-05, 5.5671e-05, 2.7323e-04, + -3.1382e-05, -7.4267e-05, 3.1204e-03, 3.0518e-05, 4.2391e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 221.61, cls_loss 0.0072 cls_loss_mapping 0.0106 cls_loss_causal 0.5654 re_mapping 0.0104 re_causal 0.0290 /// teacc 98.89 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0169, 0.1004, 0.0600, ..., -0.1255, -0.0919, 0.0267], + [ 0.0918, -0.0112, -0.0430, ..., 0.0519, 0.0771, -0.0991], + [-0.0523, -0.0260, -0.0333, ..., -0.0561, -0.0120, -0.0298], + ..., + [-0.0332, 0.1117, -0.1308, ..., -0.0922, 0.0137, -0.0952], + [ 0.0015, -0.0704, 0.0277, ..., 0.0471, -0.0945, -0.0388], + [-0.0267, -0.0590, 0.0445, ..., -0.0555, -0.0399, 0.0947]], + device='cuda:0'), grad: tensor([[ 1.9558e-06, -3.2913e-06, 3.9022e-07, ..., 3.2410e-06, + 2.9989e-07, 9.9838e-07], + [ 6.3777e-06, 2.6394e-06, 5.2489e-06, ..., 1.2763e-05, + -8.8476e-08, 3.0342e-06], + [ 5.7230e-07, 7.6834e-07, 8.3586e-07, ..., 1.4193e-06, + 4.1910e-09, 5.8208e-07], + ..., + [ 2.7176e-06, -9.7007e-06, 2.7083e-06, ..., 5.8487e-06, + 5.0943e-07, 3.4384e-06], + [-7.2680e-06, 7.3994e-07, -1.7434e-06, ..., -1.1712e-05, + 6.9663e-07, 3.6061e-06], + [ 4.3325e-06, 4.1835e-06, -5.5274e-07, ..., 7.5698e-06, + 2.2491e-07, -2.0787e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0035, 0.0014, 0.0072, 0.0124, 0.0255, 0.0366, -0.0230, 0.0100, + -0.0369, -0.0196], device='cuda:0'), grad: tensor([ 1.2908e-06, 2.8968e-05, -1.9670e-06, 2.0757e-05, -1.0289e-05, + -7.7546e-05, 4.5121e-05, -2.0474e-05, -5.3942e-06, 1.9670e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 220.01, cls_loss 0.0048 cls_loss_mapping 0.0064 cls_loss_causal 0.6021 re_mapping 0.0101 re_causal 0.0298 /// teacc 98.72 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0170, 0.1011, 0.0603, ..., -0.1261, -0.0923, 0.0266], + [ 0.0928, -0.0111, -0.0426, ..., 0.0528, 0.0773, -0.0991], + [-0.0530, -0.0265, -0.0337, ..., -0.0568, -0.0122, -0.0301], + ..., + [-0.0337, 0.1122, -0.1315, ..., -0.0932, 0.0139, -0.0955], + [ 0.0010, -0.0707, 0.0270, ..., 0.0466, -0.0946, -0.0394], + [-0.0274, -0.0594, 0.0450, ..., -0.0559, -0.0400, 0.0954]], + device='cuda:0'), grad: tensor([[ 4.9686e-07, -7.3537e-06, -2.5947e-06, ..., 2.2501e-06, + 2.1886e-08, -3.2000e-06], + [ 2.7418e-04, 2.7241e-07, 2.5658e-07, ..., 9.8515e-04, + -2.7427e-07, 2.0349e-07], + [ 8.1398e-07, 5.9884e-07, 3.6415e-07, ..., 2.2352e-06, + 5.1688e-08, 2.5332e-07], + ..., + [ 4.6119e-06, 1.5087e-07, 6.2771e-07, ..., 1.6108e-05, + 8.8010e-08, 7.2224e-07], + [ 5.0664e-06, 2.5472e-07, 1.1679e-06, ..., 1.6883e-05, + 3.4459e-08, 1.1679e-06], + [ 4.5151e-06, 1.3215e-06, -6.5984e-07, ..., 1.7345e-05, + 3.2131e-08, -3.4049e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0038, 0.0019, 0.0067, 0.0126, 0.0258, 0.0369, -0.0236, 0.0099, + -0.0378, -0.0195], device='cuda:0'), grad: tensor([-8.3670e-06, 1.1654e-03, 3.8296e-06, -2.7437e-06, -1.2550e-03, + 8.3074e-06, 2.5705e-05, 2.0385e-05, 2.2292e-05, 1.9699e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 219.74, cls_loss 0.0046 cls_loss_mapping 0.0077 cls_loss_causal 0.5650 re_mapping 0.0098 re_causal 0.0287 /// teacc 98.82 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0171, 0.1016, 0.0605, ..., -0.1268, -0.0924, 0.0266], + [ 0.0935, -0.0108, -0.0433, ..., 0.0532, 0.0773, -0.0994], + [-0.0533, -0.0270, -0.0337, ..., -0.0569, -0.0120, -0.0300], + ..., + [-0.0342, 0.1127, -0.1309, ..., -0.0948, 0.0139, -0.0959], + [ 0.0014, -0.0709, 0.0273, ..., 0.0470, -0.0946, -0.0395], + [-0.0289, -0.0603, 0.0449, ..., -0.0565, -0.0403, 0.0959]], + device='cuda:0'), grad: tensor([[ 1.7975e-06, 1.3253e-06, 2.4009e-06, ..., 2.1122e-06, + 7.9162e-09, 6.3051e-07], + [ 9.6671e-07, 1.2880e-06, 7.9125e-06, ..., 1.3910e-05, + -1.8720e-07, 5.5432e-06], + [ 3.3975e-06, 5.9977e-06, 1.1530e-06, ..., 2.6841e-06, + 1.2573e-08, 2.3609e-07], + ..., + [ 5.5227e-07, -1.5005e-05, 1.4221e-06, ..., 2.6915e-06, + 8.0559e-08, 1.0971e-06], + [-1.1253e-04, 5.8487e-07, -1.6117e-04, ..., -1.1420e-04, + 5.9605e-08, -2.7984e-05], + [ 8.4341e-06, 3.3583e-06, 9.7811e-05, ..., 1.9884e-04, + 5.8673e-08, 5.7727e-05]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0039, 0.0017, 0.0070, 0.0126, 0.0260, 0.0366, -0.0236, 0.0105, + -0.0376, -0.0202], device='cuda:0'), grad: tensor([ 7.4431e-06, 3.0071e-05, 1.6794e-05, 1.7321e-04, -3.9005e-04, + 3.4660e-05, 7.3165e-06, -1.4804e-05, -2.4796e-04, 3.8242e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 219.44, cls_loss 0.0048 cls_loss_mapping 0.0085 cls_loss_causal 0.5868 re_mapping 0.0100 re_causal 0.0291 /// teacc 98.92 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0173, 0.1022, 0.0611, ..., -0.1273, -0.0924, 0.0270], + [ 0.0951, -0.0115, -0.0426, ..., 0.0543, 0.0785, -0.0994], + [-0.0542, -0.0273, -0.0343, ..., -0.0574, -0.0126, -0.0303], + ..., + [-0.0354, 0.1146, -0.1313, ..., -0.0953, 0.0133, -0.0961], + [ 0.0017, -0.0711, 0.0274, ..., 0.0472, -0.0946, -0.0395], + [-0.0299, -0.0627, 0.0447, ..., -0.0575, -0.0407, 0.0960]], + device='cuda:0'), grad: tensor([[ 4.6333e-07, -7.4692e-06, -4.6045e-06, ..., 2.9728e-06, + 3.2876e-07, -1.7677e-06], + [ 1.0706e-05, 2.3216e-05, 1.3523e-06, ..., 1.3873e-05, + -4.6939e-07, 5.9791e-07], + [ 5.3868e-06, 2.6524e-06, 3.0342e-06, ..., -5.6744e-05, + -3.2084e-07, 3.8929e-07], + ..., + [-1.7226e-05, -4.7892e-05, 2.6375e-06, ..., -2.2694e-05, + 2.2491e-07, 1.8198e-06], + [-8.4657e-07, 4.7795e-06, 6.5938e-06, ..., -5.8953e-07, + -1.4016e-07, 8.1509e-06], + [ 1.9325e-07, 7.1116e-06, -1.5110e-05, ..., 2.5965e-06, + 1.9511e-07, -1.5840e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0044, 0.0026, 0.0065, 0.0127, 0.0263, 0.0366, -0.0241, 0.0108, + -0.0373, -0.0212], device='cuda:0'), grad: tensor([ 2.3589e-05, 5.8800e-05, -8.2445e-04, 3.7074e-05, 7.0477e-04, + 7.8976e-06, 6.1810e-05, -1.1665e-04, 4.1664e-05, 5.0440e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 219.90, cls_loss 0.0053 cls_loss_mapping 0.0072 cls_loss_causal 0.6032 re_mapping 0.0099 re_causal 0.0287 /// teacc 98.91 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0172, 0.1028, 0.0608, ..., -0.1273, -0.0924, 0.0263], + [ 0.0963, -0.0116, -0.0424, ..., 0.0556, 0.0796, -0.0996], + [-0.0550, -0.0280, -0.0347, ..., -0.0581, -0.0133, -0.0309], + ..., + [-0.0359, 0.1150, -0.1318, ..., -0.0962, 0.0129, -0.0970], + [ 0.0004, -0.0716, 0.0268, ..., 0.0465, -0.0963, -0.0397], + [-0.0306, -0.0623, 0.0458, ..., -0.0577, -0.0408, 0.0974]], + device='cuda:0'), grad: tensor([[ 1.4668e-06, -3.8603e-07, 1.3269e-05, ..., 1.1837e-06, + 2.9709e-07, 1.9908e-05], + [ 2.8051e-06, 7.5903e-08, 9.9018e-06, ..., -1.8207e-07, + 2.7753e-06, 7.3165e-06], + [ 2.2966e-06, 5.4482e-08, 9.5218e-06, ..., 4.6287e-07, + 2.6450e-06, 9.6038e-06], + ..., + [ 9.0897e-06, -1.5507e-07, 2.7344e-05, ..., 9.7603e-07, + 4.6641e-06, 8.5011e-06], + [ 2.2762e-06, 1.0757e-07, 1.0706e-05, ..., 4.2422e-07, + 2.7362e-06, 1.3269e-05], + [ 4.1388e-06, 1.9604e-07, -9.3043e-05, ..., 1.0384e-06, + 9.0478e-07, -1.5414e-04]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0040, 0.0034, 0.0062, 0.0123, 0.0261, 0.0378, -0.0248, 0.0106, + -0.0387, -0.0205], device='cuda:0'), grad: tensor([ 3.0339e-05, 2.6643e-05, 2.2814e-05, -8.0526e-05, 2.0504e-05, + -2.2042e-04, 2.7108e-04, 8.4877e-05, 2.7955e-05, -1.8370e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 219.66, cls_loss 0.0041 cls_loss_mapping 0.0064 cls_loss_causal 0.5925 re_mapping 0.0096 re_causal 0.0281 /// teacc 98.91 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0173, 0.1034, 0.0614, ..., -0.1283, -0.0924, 0.0265], + [ 0.0967, -0.0117, -0.0424, ..., 0.0558, 0.0800, -0.0997], + [-0.0551, -0.0280, -0.0347, ..., -0.0586, -0.0133, -0.0312], + ..., + [-0.0361, 0.1154, -0.1323, ..., -0.0966, 0.0132, -0.0980], + [ 0.0003, -0.0721, 0.0268, ..., 0.0465, -0.0971, -0.0398], + [-0.0307, -0.0626, 0.0458, ..., -0.0582, -0.0414, 0.0977]], + device='cuda:0'), grad: tensor([[ 1.3486e-06, -3.2014e-07, 1.3225e-06, ..., 1.3718e-06, + 1.3504e-08, 9.7789e-09], + [ 6.2678e-07, 2.0536e-07, 1.5395e-06, ..., 8.8662e-07, + -2.1001e-07, 1.9558e-07], + [ 7.0743e-06, 1.0221e-07, 8.0615e-06, ..., 5.6326e-06, + -6.7521e-09, 3.1781e-07], + ..., + [ 2.5295e-06, -2.5835e-06, 3.2522e-06, ..., 7.5996e-06, + -2.9965e-07, 1.2964e-06], + [ 2.2631e-06, 1.4598e-07, 1.7183e-06, ..., -3.0547e-06, + 6.7987e-08, 2.3395e-06], + [ 1.7351e-06, 1.2266e-06, -2.9746e-06, ..., 1.5404e-06, + 2.3609e-07, -1.1131e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0043, 0.0034, 0.0063, 0.0121, 0.0265, 0.0380, -0.0251, 0.0104, + -0.0391, -0.0205], device='cuda:0'), grad: tensor([ 5.5991e-06, 6.6496e-06, 2.3454e-05, -7.0453e-05, -7.4580e-06, + 1.2569e-05, -1.8254e-07, 1.4909e-05, 1.1951e-05, 3.0063e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 219.54, cls_loss 0.0060 cls_loss_mapping 0.0076 cls_loss_causal 0.5872 re_mapping 0.0099 re_causal 0.0296 /// teacc 98.81 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0177, 0.1038, 0.0618, ..., -0.1294, -0.0925, 0.0266], + [ 0.0961, -0.0117, -0.0434, ..., 0.0547, 0.0799, -0.1014], + [-0.0553, -0.0285, -0.0345, ..., -0.0587, -0.0129, -0.0306], + ..., + [-0.0362, 0.1158, -0.1328, ..., -0.0969, 0.0133, -0.0988], + [ 0.0013, -0.0723, 0.0277, ..., 0.0484, -0.0970, -0.0388], + [-0.0311, -0.0630, 0.0454, ..., -0.0584, -0.0417, 0.0978]], + device='cuda:0'), grad: tensor([[ 2.7753e-06, 5.6744e-04, 2.6226e-03, ..., 3.0026e-06, + 4.1910e-09, 4.9210e-03], + [-2.1304e-07, 1.9446e-06, 2.9318e-06, ..., 4.9174e-06, + -1.6298e-07, 5.2564e-06], + [ 3.5157e-08, 6.1020e-06, 1.7121e-05, ..., -1.7524e-05, + 1.4203e-08, 3.0205e-05], + ..., + [ 2.1537e-07, -3.7216e-06, 9.0674e-06, ..., 1.5572e-06, + 2.8405e-08, 1.6674e-05], + [ 2.3302e-06, 9.3551e-07, 5.4538e-06, ..., 6.3591e-06, + 6.1933e-08, 4.2357e-06], + [ 5.5321e-07, -5.8031e-04, -2.6760e-03, ..., 1.0438e-05, + 1.6764e-08, -5.0316e-03]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0045, 0.0028, 0.0066, 0.0141, 0.0262, 0.0365, -0.0252, 0.0101, + -0.0380, -0.0207], device='cuda:0'), grad: tensor([ 8.3618e-03, 2.0325e-05, 1.2279e-05, 3.0547e-05, 2.7061e-05, + 2.2352e-05, 1.0967e-05, 2.1651e-05, 2.2978e-05, -8.5220e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 219.85, cls_loss 0.0065 cls_loss_mapping 0.0107 cls_loss_causal 0.5779 re_mapping 0.0098 re_causal 0.0289 /// teacc 98.82 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0176, 0.1046, 0.0618, ..., -0.1298, -0.0926, 0.0264], + [ 0.0951, -0.0103, -0.0428, ..., 0.0540, 0.0801, -0.1019], + [-0.0557, -0.0289, -0.0351, ..., -0.0596, -0.0129, -0.0310], + ..., + [-0.0343, 0.1151, -0.1336, ..., -0.0949, 0.0135, -0.0996], + [ 0.0019, -0.0721, 0.0280, ..., 0.0492, -0.0974, -0.0385], + [-0.0311, -0.0638, 0.0459, ..., -0.0590, -0.0419, 0.0985]], + device='cuda:0'), grad: tensor([[ 6.0629e-07, 4.0382e-06, 1.0803e-06, ..., 1.7285e-06, + 2.9476e-07, 2.6952e-06], + [-7.0155e-05, 8.7693e-06, -1.3545e-05, ..., -3.4183e-05, + -4.5389e-05, 5.8673e-07], + [ 3.1814e-06, -7.0035e-05, -5.3681e-06, ..., -2.4494e-06, + -1.5758e-06, 2.8964e-07], + ..., + [ 5.3585e-05, -1.9848e-05, 1.1884e-05, ..., 2.9594e-05, + 3.5077e-05, -1.6898e-05], + [-1.1921e-06, 3.4515e-06, -5.4436e-07, ..., -2.7753e-06, + 9.7789e-07, 1.2806e-08], + [ 2.0750e-06, 4.4674e-05, 1.6466e-06, ..., 2.1383e-06, + 1.1567e-06, 1.4976e-05]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0043, 0.0026, 0.0061, 0.0131, 0.0267, 0.0372, -0.0261, 0.0107, + -0.0377, -0.0208], device='cuda:0'), grad: tensor([ 1.2018e-05, -9.4950e-05, -5.3930e-04, 2.4104e-04, 8.3074e-07, + -1.8418e-05, 3.2838e-06, 2.8896e-04, 2.7955e-05, 7.9453e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 219.36, cls_loss 0.0056 cls_loss_mapping 0.0071 cls_loss_causal 0.5808 re_mapping 0.0096 re_causal 0.0274 /// teacc 98.85 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0184, 0.1053, 0.0624, ..., -0.1315, -0.0932, 0.0265], + [ 0.0953, -0.0102, -0.0431, ..., 0.0541, 0.0803, -0.1022], + [-0.0559, -0.0296, -0.0357, ..., -0.0596, -0.0131, -0.0317], + ..., + [-0.0346, 0.1154, -0.1353, ..., -0.0954, 0.0135, -0.1012], + [ 0.0024, -0.0725, 0.0286, ..., 0.0501, -0.0974, -0.0382], + [-0.0313, -0.0638, 0.0472, ..., -0.0593, -0.0420, 0.0997]], + device='cuda:0'), grad: tensor([[ 3.1441e-05, 8.4788e-06, -1.5944e-05, ..., 4.3750e-05, + 7.9162e-09, 9.5144e-06], + [-9.5606e-05, 8.8438e-06, 4.4517e-07, ..., -1.7083e-04, + -1.6438e-07, 7.1665e-07], + [ 5.2117e-06, 9.3162e-05, 2.6003e-05, ..., 2.8118e-05, + 6.2399e-08, 2.6971e-05], + ..., + [ 8.2031e-06, -1.1736e-04, 4.5309e-07, ..., 1.6093e-05, + 3.4925e-08, 5.6531e-07], + [ 1.9297e-05, 7.1004e-06, 1.3560e-06, ..., 3.1859e-05, + 1.2107e-08, 5.3346e-06], + [ 1.4052e-05, 3.8773e-05, 1.1377e-05, ..., 2.4021e-05, + 1.3970e-08, 1.7196e-05]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0044, 0.0025, 0.0061, 0.0123, 0.0266, 0.0368, -0.0261, 0.0106, + -0.0371, -0.0201], device='cuda:0'), grad: tensor([ 4.8965e-05, -1.7607e-04, 2.4939e-04, 4.4882e-05, 5.7608e-05, + 5.6416e-05, -2.1255e-04, -2.3341e-04, 5.3197e-05, 1.1170e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 219.56, cls_loss 0.0046 cls_loss_mapping 0.0074 cls_loss_causal 0.5783 re_mapping 0.0103 re_causal 0.0284 /// teacc 98.85 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0189, 0.1053, 0.0623, ..., -0.1319, -0.0937, 0.0257], + [ 0.0956, -0.0099, -0.0431, ..., 0.0545, 0.0807, -0.1024], + [-0.0562, -0.0305, -0.0356, ..., -0.0599, -0.0133, -0.0316], + ..., + [-0.0348, 0.1163, -0.1351, ..., -0.0960, 0.0134, -0.1004], + [ 0.0026, -0.0727, 0.0290, ..., 0.0507, -0.0975, -0.0381], + [-0.0316, -0.0653, 0.0475, ..., -0.0597, -0.0423, 0.1006]], + device='cuda:0'), grad: tensor([[ 4.5970e-06, -7.9945e-06, -5.0887e-06, ..., 1.2867e-05, + 3.3062e-08, -6.7428e-06], + [ 6.3218e-06, 2.4319e-05, 1.3728e-06, ..., 2.7046e-05, + -9.8627e-07, 5.2061e-07], + [ 9.3430e-06, 2.4050e-05, 2.6766e-06, ..., 2.4766e-05, + 3.8929e-07, 5.3551e-07], + ..., + [ 1.6302e-05, 1.0550e-05, 1.0051e-05, ..., 5.1767e-05, + 2.0070e-07, 1.3784e-05], + [ 2.5019e-05, 2.4378e-05, 2.2560e-05, ..., 4.9859e-05, + 6.2771e-07, 1.7956e-05], + [ 7.1116e-06, 1.7568e-05, -2.2367e-05, ..., 1.8790e-05, + 1.7509e-07, -2.9162e-05]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0034, 0.0027, 0.0061, 0.0121, 0.0267, 0.0365, -0.0260, 0.0110, + -0.0368, -0.0203], device='cuda:0'), grad: tensor([-5.7518e-06, 7.5877e-05, 7.1168e-05, -1.9029e-05, -5.5504e-04, + 1.0920e-04, 4.8101e-05, 1.0926e-04, 1.5938e-04, 6.6683e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 219.88, cls_loss 0.0057 cls_loss_mapping 0.0080 cls_loss_causal 0.5823 re_mapping 0.0094 re_causal 0.0280 /// teacc 98.81 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0191, 0.1060, 0.0630, ..., -0.1320, -0.0937, 0.0258], + [ 0.0974, -0.0100, -0.0437, ..., 0.0552, 0.0827, -0.1031], + [-0.0584, -0.0311, -0.0364, ..., -0.0609, -0.0152, -0.0328], + ..., + [-0.0351, 0.1168, -0.1366, ..., -0.0967, 0.0132, -0.1023], + [ 0.0024, -0.0731, 0.0287, ..., 0.0507, -0.0977, -0.0384], + [-0.0319, -0.0656, 0.0486, ..., -0.0598, -0.0426, 0.1018]], + device='cuda:0'), grad: tensor([[ 2.6058e-06, 6.0536e-08, 7.6462e-07, ..., 1.9856e-06, + 3.3947e-07, 3.9395e-07], + [ 6.0955e-07, 3.7951e-07, 1.4948e-07, ..., -7.7952e-07, + -5.5879e-07, 7.2177e-08], + [ 1.0088e-05, 6.5193e-07, 1.3160e-06, ..., 6.4038e-06, + 1.9297e-06, 6.5193e-07], + ..., + [ 1.0487e-06, -3.6806e-06, 8.4750e-07, ..., 6.0871e-06, + 2.7195e-07, 9.2341e-07], + [-2.3052e-05, 2.4401e-07, -1.3545e-05, ..., -1.3284e-05, + 5.7509e-07, -6.5975e-06], + [ 1.4946e-05, 1.3970e-06, 7.1600e-06, ..., 1.1750e-05, + 2.1840e-07, 2.2501e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0036, 0.0033, 0.0051, 0.0125, 0.0268, 0.0364, -0.0265, 0.0105, + -0.0374, -0.0195], device='cuda:0'), grad: tensor([ 7.1228e-06, 5.9083e-06, 2.0504e-05, -1.2517e-06, -1.9684e-05, + 1.2189e-05, -2.5257e-05, 1.2934e-05, -5.5313e-05, 4.2737e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 219.79, cls_loss 0.0053 cls_loss_mapping 0.0084 cls_loss_causal 0.5818 re_mapping 0.0097 re_causal 0.0273 /// teacc 98.92 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0192, 0.1064, 0.0634, ..., -0.1324, -0.0940, 0.0260], + [ 0.0981, -0.0102, -0.0434, ..., 0.0558, 0.0826, -0.1033], + [-0.0585, -0.0315, -0.0356, ..., -0.0608, -0.0149, -0.0329], + ..., + [-0.0353, 0.1170, -0.1374, ..., -0.0976, 0.0136, -0.1033], + [ 0.0026, -0.0722, 0.0285, ..., 0.0512, -0.0980, -0.0386], + [-0.0321, -0.0665, 0.0494, ..., -0.0605, -0.0429, 0.1025]], + device='cuda:0'), grad: tensor([[ 9.2462e-06, 1.2927e-06, -1.7602e-06, ..., 1.0513e-05, + 7.1758e-07, -2.1085e-06], + [ 1.9614e-06, 2.6412e-06, 7.2829e-07, ..., 3.2298e-06, + 5.7695e-07, 3.5809e-07], + [ 1.6410e-06, 1.2834e-06, 4.2794e-07, ..., 2.0862e-06, + -4.8354e-06, 5.8906e-07], + ..., + [ 4.9872e-07, -9.7454e-06, 1.4910e-06, ..., 1.1325e-06, + 1.2675e-06, 1.2117e-06], + [ 1.3001e-05, 3.1590e-06, 5.6922e-06, ..., 1.3441e-05, + 1.6503e-06, 9.0674e-06], + [ 1.9278e-06, 5.8711e-06, 2.8498e-07, ..., 3.0920e-06, + 6.0862e-07, 8.2888e-07]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0038, 0.0032, 0.0055, 0.0126, 0.0269, 0.0365, -0.0274, 0.0102, + -0.0367, -0.0197], device='cuda:0'), grad: tensor([ 1.7300e-05, 1.2845e-05, -1.3098e-05, -2.1122e-06, 3.1024e-05, + -3.7074e-05, -5.5254e-05, -5.9046e-06, 3.8177e-05, 1.4044e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 219.68, cls_loss 0.0042 cls_loss_mapping 0.0064 cls_loss_causal 0.5716 re_mapping 0.0092 re_causal 0.0261 /// teacc 98.85 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0194, 0.1066, 0.0642, ..., -0.1330, -0.0941, 0.0264], + [ 0.0989, -0.0102, -0.0431, ..., 0.0566, 0.0831, -0.1030], + [-0.0588, -0.0319, -0.0351, ..., -0.0606, -0.0147, -0.0330], + ..., + [-0.0355, 0.1172, -0.1390, ..., -0.0981, 0.0136, -0.1042], + [ 0.0024, -0.0727, 0.0281, ..., 0.0509, -0.0985, -0.0393], + [-0.0326, -0.0667, 0.0499, ..., -0.0610, -0.0435, 0.1031]], + device='cuda:0'), grad: tensor([[ 9.5740e-07, -2.4308e-06, -7.4357e-06, ..., 2.9579e-06, + 5.1688e-08, -1.3359e-05], + [-1.7555e-07, 5.8487e-07, 4.7609e-06, ..., 8.2143e-07, + -1.4016e-07, 4.5411e-06], + [ 4.0652e-07, 4.9500e-07, 1.0990e-05, ..., 1.9651e-06, + 1.9092e-08, 2.9616e-06], + ..., + [ 2.7288e-07, -1.6475e-06, 6.2548e-06, ..., 1.3988e-06, + 1.8068e-07, 6.5491e-06], + [-8.3447e-07, 9.6299e-07, 8.5309e-06, ..., -6.6543e-07, + 6.8452e-08, 1.8030e-06], + [ 3.6974e-07, 1.1772e-06, 1.1409e-06, ..., 1.8757e-06, + -2.9383e-07, -1.0528e-05]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0042, 0.0034, 0.0058, 0.0132, 0.0268, 0.0361, -0.0273, 0.0099, + -0.0372, -0.0198], device='cuda:0'), grad: tensor([-1.5974e-05, 1.5333e-05, 2.9907e-05, -1.0639e-04, 8.8632e-05, + 1.6347e-05, -9.0003e-05, 1.5393e-05, 3.4779e-05, 1.2077e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 219.36, cls_loss 0.0071 cls_loss_mapping 0.0102 cls_loss_causal 0.5708 re_mapping 0.0101 re_causal 0.0273 /// teacc 98.86 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0196, 0.1060, 0.0635, ..., -0.1339, -0.0941, 0.0250], + [ 0.0991, -0.0104, -0.0437, ..., 0.0572, 0.0833, -0.1035], + [-0.0590, -0.0326, -0.0351, ..., -0.0608, -0.0146, -0.0323], + ..., + [-0.0357, 0.1184, -0.1392, ..., -0.0987, 0.0135, -0.1042], + [ 0.0012, -0.0747, 0.0264, ..., 0.0499, -0.0985, -0.0418], + [-0.0334, -0.0664, 0.0513, ..., -0.0614, -0.0437, 0.1050]], + device='cuda:0'), grad: tensor([[-1.4435e-08, -2.9188e-06, -5.7742e-06, ..., 4.1677e-07, + 3.7719e-08, -6.1393e-06], + [-1.1455e-06, 5.8115e-06, 1.7183e-07, ..., -1.2843e-06, + -2.5472e-07, 1.2526e-07], + [ 5.9791e-07, 3.9712e-06, 2.4633e-07, ..., 4.2561e-07, + -5.4017e-08, 2.0443e-07], + ..., + [ 1.0533e-06, -2.3648e-05, 5.2713e-07, ..., 1.4389e-06, + 3.3947e-07, 6.3935e-07], + [ 6.4308e-07, 5.9605e-07, 1.0198e-06, ..., 1.8068e-07, + 3.1199e-07, 1.1856e-06], + [ 4.4983e-07, 6.1989e-06, 2.0824e-06, ..., 7.9488e-07, + 7.7765e-08, 2.7604e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0029, 0.0034, 0.0059, 0.0135, 0.0259, 0.0353, -0.0262, 0.0104, + -0.0383, -0.0187], device='cuda:0'), grad: tensor([-8.9556e-06, 5.8040e-06, 6.4820e-07, 1.5333e-05, 1.5134e-07, + -1.7807e-06, -2.7083e-06, -2.2605e-05, 3.2391e-06, 1.0870e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 219.56, cls_loss 0.0049 cls_loss_mapping 0.0064 cls_loss_causal 0.5931 re_mapping 0.0091 re_causal 0.0264 /// teacc 98.98 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0200, 0.1065, 0.0639, ..., -0.1346, -0.0942, 0.0254], + [ 0.0999, -0.0096, -0.0439, ..., 0.0577, 0.0836, -0.1036], + [-0.0594, -0.0323, -0.0352, ..., -0.0613, -0.0150, -0.0330], + ..., + [-0.0361, 0.1180, -0.1396, ..., -0.0991, 0.0139, -0.1050], + [ 0.0003, -0.0752, 0.0261, ..., 0.0493, -0.0997, -0.0423], + [-0.0339, -0.0671, 0.0517, ..., -0.0621, -0.0440, 0.1052]], + device='cuda:0'), grad: tensor([[ 7.8045e-07, -6.8499e-07, -1.4063e-06, ..., 6.5900e-06, + 1.9139e-07, -3.5670e-07], + [ 3.2485e-05, -4.5635e-08, 3.3557e-05, ..., 1.3614e-04, + 1.0863e-05, 5.5760e-05], + [-3.3341e-06, 4.6566e-08, 5.2340e-07, ..., -1.3256e-04, + 1.5507e-07, 7.4133e-07], + ..., + [ 2.3581e-06, 3.5483e-07, 5.6550e-06, ..., 2.0191e-05, + 2.6356e-07, 6.2995e-06], + [-3.1963e-06, 8.6613e-08, 3.1013e-07, ..., -1.9193e-05, + 2.6859e-06, 1.1370e-05], + [ 1.7300e-05, -8.6613e-07, 3.6899e-06, ..., 2.3082e-05, + 6.3479e-06, 1.3009e-05]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0031, 0.0040, 0.0066, 0.0129, 0.0263, 0.0362, -0.0260, 0.0095, + -0.0393, -0.0190], device='cuda:0'), grad: tensor([ 3.0607e-05, 6.5756e-04, -7.4720e-04, 5.4747e-05, 4.7982e-05, + -1.3673e-04, 2.1473e-05, 5.7638e-05, 6.6590e-08, 1.2949e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 219.59, cls_loss 0.0036 cls_loss_mapping 0.0066 cls_loss_causal 0.6067 re_mapping 0.0088 re_causal 0.0276 /// teacc 98.87 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0201, 0.1067, 0.0641, ..., -0.1354, -0.0942, 0.0256], + [ 0.1001, -0.0093, -0.0437, ..., 0.0579, 0.0837, -0.1037], + [-0.0595, -0.0321, -0.0349, ..., -0.0615, -0.0151, -0.0336], + ..., + [-0.0365, 0.1179, -0.1407, ..., -0.0995, 0.0137, -0.1067], + [ 0.0009, -0.0755, 0.0265, ..., 0.0502, -0.0997, -0.0419], + [-0.0343, -0.0672, 0.0514, ..., -0.0634, -0.0442, 0.1051]], + device='cuda:0'), grad: tensor([[ 3.9535e-07, -3.3174e-06, -1.1399e-06, ..., 1.4165e-06, + 8.2050e-07, -1.0170e-06], + [-3.0268e-07, 1.0334e-05, 8.0699e-07, ..., 5.5470e-06, + -2.6310e-07, 4.0093e-07], + [ 4.3726e-07, 2.1607e-06, 1.2442e-06, ..., 1.5106e-06, + 7.3947e-07, 4.1351e-07], + ..., + [ 4.8056e-07, -2.2784e-05, -3.7719e-08, ..., 1.0654e-05, + 2.2678e-07, 1.1381e-06], + [-3.5055e-06, 4.2515e-07, -6.1616e-06, ..., -4.1649e-06, + 1.9874e-06, -7.9162e-07], + [ 3.3947e-07, 8.7544e-06, -2.5257e-06, ..., 8.8513e-06, + 8.6147e-08, -3.6266e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0032, 0.0042, 0.0068, 0.0132, 0.0267, 0.0359, -0.0262, 0.0090, + -0.0387, -0.0193], device='cuda:0'), grad: tensor([-1.9688e-06, 2.5228e-05, 9.2536e-06, 2.4170e-05, -2.9847e-05, + -2.1115e-05, 8.6278e-06, -3.7998e-05, -3.3118e-06, 2.6822e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 220.37, cls_loss 0.0049 cls_loss_mapping 0.0067 cls_loss_causal 0.5790 re_mapping 0.0086 re_causal 0.0257 /// teacc 98.89 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0205, 0.1073, 0.0644, ..., -0.1364, -0.0944, 0.0255], + [ 0.1006, -0.0094, -0.0435, ..., 0.0581, 0.0840, -0.1040], + [-0.0598, -0.0325, -0.0356, ..., -0.0619, -0.0155, -0.0346], + ..., + [-0.0367, 0.1187, -0.1411, ..., -0.0997, 0.0143, -0.1074], + [ 0.0010, -0.0759, 0.0283, ..., 0.0514, -0.1006, -0.0403], + [-0.0351, -0.0681, 0.0513, ..., -0.0638, -0.0446, 0.1055]], + device='cuda:0'), grad: tensor([[ 1.2899e-06, 1.8813e-06, 1.6149e-06, ..., 7.6555e-07, + 6.3330e-08, 1.7863e-06], + [ 1.1316e-07, 1.2657e-06, 3.3677e-05, ..., 7.2643e-08, + -8.5216e-08, 5.1081e-05], + [ 4.1490e-07, -3.3695e-06, 3.2876e-07, ..., 2.3423e-07, + 1.4016e-07, 3.5344e-07], + ..., + [ 8.4937e-07, -2.6915e-06, 6.3851e-06, ..., 8.8476e-07, + 6.7987e-08, 5.6848e-06], + [-7.9907e-07, 1.8124e-06, 3.5524e-05, ..., 2.6599e-06, + 2.9709e-07, 2.8685e-05], + [ 3.5111e-07, 1.2890e-06, -1.7154e-04, ..., -6.9402e-06, + 2.7940e-08, -2.2888e-04]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0031, 0.0042, 0.0065, 0.0129, 0.0270, 0.0356, -0.0269, 0.0094, + -0.0375, -0.0197], device='cuda:0'), grad: tensor([ 3.2693e-05, 8.6784e-05, -9.1553e-05, 4.7795e-06, 2.0564e-04, + 1.2971e-05, -8.1956e-06, 3.1173e-05, 9.4533e-05, -3.6883e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 219.06, cls_loss 0.0044 cls_loss_mapping 0.0067 cls_loss_causal 0.5838 re_mapping 0.0092 re_causal 0.0264 /// teacc 98.85 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0206, 0.1075, 0.0644, ..., -0.1368, -0.0944, 0.0253], + [ 0.1008, -0.0088, -0.0438, ..., 0.0581, 0.0844, -0.1047], + [-0.0601, -0.0328, -0.0360, ..., -0.0624, -0.0156, -0.0352], + ..., + [-0.0370, 0.1186, -0.1418, ..., -0.0998, 0.0141, -0.1084], + [ 0.0008, -0.0763, 0.0283, ..., 0.0517, -0.1010, -0.0405], + [-0.0354, -0.0685, 0.0523, ..., -0.0637, -0.0447, 0.1069]], + device='cuda:0'), grad: tensor([[ 3.9227e-06, -2.6375e-06, 6.1728e-06, ..., 6.0201e-06, + 1.9278e-07, 2.0731e-06], + [-1.3433e-05, 4.2003e-07, 6.1234e-07, ..., -1.9163e-05, + -5.6289e-06, 6.6916e-07], + [ 4.8913e-06, 1.6624e-07, 2.0210e-07, ..., 6.3516e-06, + 2.0210e-06, 1.3178e-06], + ..., + [ 7.3295e-07, -8.6194e-07, 6.1188e-07, ..., 2.1439e-06, + 1.5367e-07, 5.0990e-07], + [-3.7313e-05, 1.3178e-07, -4.5598e-05, ..., -4.7117e-05, + -6.8024e-06, -3.0354e-05], + [ 6.8024e-06, 2.4196e-06, 1.8537e-05, ..., 3.0816e-05, + 3.0082e-07, 1.6078e-05]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0029, 0.0045, 0.0066, 0.0120, 0.0265, 0.0358, -0.0264, 0.0093, + -0.0378, -0.0191], device='cuda:0'), grad: tensor([ 1.1124e-05, -3.0816e-05, -2.0891e-05, 2.6301e-05, -2.2948e-05, + 3.4757e-06, 6.9439e-05, 3.2410e-06, -1.0979e-04, 7.0810e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 219.96, cls_loss 0.0040 cls_loss_mapping 0.0069 cls_loss_causal 0.5664 re_mapping 0.0090 re_causal 0.0264 /// teacc 98.84 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0214, 0.1078, 0.0646, ..., -0.1386, -0.0951, 0.0253], + [ 0.1011, -0.0093, -0.0444, ..., 0.0586, 0.0846, -0.1049], + [-0.0604, -0.0331, -0.0365, ..., -0.0628, -0.0157, -0.0357], + ..., + [-0.0367, 0.1194, -0.1415, ..., -0.1001, 0.0140, -0.1095], + [ 0.0004, -0.0769, 0.0281, ..., 0.0516, -0.1012, -0.0407], + [-0.0358, -0.0690, 0.0522, ..., -0.0642, -0.0448, 0.1072]], + device='cuda:0'), grad: tensor([[ 4.7088e-06, -5.7276e-08, 6.8285e-06, ..., 4.6976e-06, + 2.6962e-07, 2.3786e-06], + [ 1.9185e-06, 1.0654e-06, 2.5872e-06, ..., 4.6901e-06, + 9.4995e-08, 1.0785e-06], + [ 3.8557e-06, 3.0175e-07, 1.2524e-05, ..., 2.9504e-06, + 2.2203e-06, 9.7230e-07], + ..., + [ 1.0356e-06, -1.3113e-05, 2.0098e-06, ..., 2.9095e-06, + 2.3749e-07, 1.0580e-06], + [-2.3395e-05, 1.3746e-06, -2.1607e-05, ..., -2.1994e-05, + 2.8033e-07, -3.2373e-06], + [ 4.3921e-06, 9.0003e-06, -2.9374e-06, ..., -1.7393e-04, + 6.3796e-08, -1.7858e-04]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0023, 0.0043, 0.0065, 0.0117, 0.0268, 0.0363, -0.0256, 0.0098, + -0.0383, -0.0195], device='cuda:0'), grad: tensor([ 1.7047e-05, 1.5959e-05, 8.4162e-05, -8.6129e-05, 6.5088e-04, + 4.6007e-06, 1.4141e-05, -1.1981e-05, -3.0056e-05, -6.5851e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 219.59, cls_loss 0.0044 cls_loss_mapping 0.0060 cls_loss_causal 0.6092 re_mapping 0.0087 re_causal 0.0259 /// teacc 98.97 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0220, 0.1085, 0.0647, ..., -0.1402, -0.0952, 0.0252], + [ 0.1013, -0.0097, -0.0456, ..., 0.0587, 0.0847, -0.1051], + [-0.0608, -0.0335, -0.0372, ..., -0.0634, -0.0158, -0.0360], + ..., + [-0.0367, 0.1200, -0.1419, ..., -0.1006, 0.0139, -0.1103], + [ 0.0017, -0.0760, 0.0298, ..., 0.0529, -0.1012, -0.0400], + [-0.0364, -0.0695, 0.0525, ..., -0.0644, -0.0450, 0.1077]], + device='cuda:0'), grad: tensor([[ 7.9768e-07, -8.9034e-07, -1.1083e-07, ..., 3.8892e-06, + 1.5367e-08, -1.6484e-07], + [-2.4214e-06, 1.0140e-05, 4.3958e-07, ..., 1.5097e-06, + -1.2666e-07, 2.6403e-07], + [ 5.7230e-07, 9.5367e-07, 6.6170e-07, ..., 1.8969e-05, + -2.0210e-07, 1.6438e-07], + ..., + [ 1.2973e-06, -1.8612e-05, 1.1455e-06, ..., 3.4273e-06, + 2.3516e-07, 9.2387e-07], + [ 7.7710e-06, 9.3691e-07, 6.2473e-06, ..., 1.4640e-05, + 3.6834e-07, 7.3351e-06], + [ 1.6633e-06, 5.9530e-06, 1.1288e-06, ..., 8.8364e-06, + 6.1002e-08, 4.5262e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0023, 0.0040, 0.0061, 0.0113, 0.0270, 0.0366, -0.0268, 0.0098, + -0.0363, -0.0195], device='cuda:0'), grad: tensor([ 5.0962e-06, 2.4304e-05, 2.8253e-05, -8.3819e-06, -5.0020e-04, + -2.0623e-05, 4.2486e-04, -2.1994e-05, 3.9548e-05, 2.8834e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 219.88, cls_loss 0.0036 cls_loss_mapping 0.0061 cls_loss_causal 0.5716 re_mapping 0.0086 re_causal 0.0256 /// teacc 98.91 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0222, 0.1091, 0.0647, ..., -0.1409, -0.0953, 0.0250], + [ 0.1017, -0.0097, -0.0458, ..., 0.0589, 0.0849, -0.1052], + [-0.0611, -0.0338, -0.0374, ..., -0.0638, -0.0160, -0.0362], + ..., + [-0.0370, 0.1206, -0.1423, ..., -0.1009, 0.0138, -0.1105], + [ 0.0016, -0.0769, 0.0295, ..., 0.0528, -0.1015, -0.0402], + [-0.0366, -0.0702, 0.0532, ..., -0.0654, -0.0450, 0.1083]], + device='cuda:0'), grad: tensor([[ 5.5879e-07, -2.3544e-06, 2.5472e-07, ..., 6.0629e-07, + 1.8114e-07, -5.0897e-07], + [-1.4314e-06, 1.8720e-07, 2.3730e-06, ..., -2.0154e-06, + -1.3690e-07, 1.2619e-07], + [ 9.0674e-06, 4.9593e-07, 1.6928e-05, ..., 1.7434e-06, + 7.3984e-06, 2.9104e-07], + ..., + [ 6.4913e-07, -8.3400e-07, 6.8685e-07, ..., 6.5845e-07, + 2.4866e-07, 2.9709e-07], + [ 1.0114e-06, 2.9849e-07, 2.8126e-06, ..., -1.1977e-06, + 1.3402e-06, 5.4529e-07], + [ 7.3249e-07, 1.3020e-06, 5.6345e-07, ..., 1.1306e-06, + 2.1886e-07, 3.7346e-07]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0023, 0.0041, 0.0060, 0.0113, 0.0276, 0.0364, -0.0266, 0.0098, + -0.0367, -0.0198], device='cuda:0'), grad: tensor([-2.4326e-06, -2.2119e-07, 2.9728e-05, -2.8282e-05, 4.0000e-07, + -4.1813e-05, 3.3528e-05, 1.3784e-06, 3.9376e-06, 3.9265e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 219.47, cls_loss 0.0043 cls_loss_mapping 0.0082 cls_loss_causal 0.5748 re_mapping 0.0089 re_causal 0.0253 /// teacc 98.91 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0222, 0.1102, 0.0650, ..., -0.1412, -0.0953, 0.0252], + [ 0.1020, -0.0100, -0.0464, ..., 0.0593, 0.0854, -0.1056], + [-0.0612, -0.0346, -0.0375, ..., -0.0640, -0.0162, -0.0366], + ..., + [-0.0371, 0.1225, -0.1424, ..., -0.1018, 0.0133, -0.1120], + [ 0.0016, -0.0793, 0.0295, ..., 0.0531, -0.1016, -0.0403], + [-0.0369, -0.0713, 0.0539, ..., -0.0652, -0.0450, 0.1094]], + device='cuda:0'), grad: tensor([[ 1.9185e-06, -1.8165e-05, -9.3877e-06, ..., 3.2205e-06, + 1.2713e-07, -1.0721e-05], + [-2.3562e-07, 2.9672e-06, 2.0862e-06, ..., 3.9935e-06, + 8.3353e-07, 4.0885e-07], + [-2.6952e-06, 8.5458e-06, -3.8594e-05, ..., -7.5391e-07, + 8.5309e-07, -3.4392e-05], + ..., + [ 2.8592e-07, -7.4625e-05, 6.9523e-07, ..., 7.5512e-06, + -3.5226e-05, 7.5996e-07], + [-8.1165e-07, 6.7532e-05, 3.4217e-06, ..., -1.5143e-06, + 2.9087e-05, 2.3879e-06], + [ 4.7591e-07, 4.3549e-06, 1.2822e-05, ..., 8.7470e-06, + 8.9174e-07, 1.3933e-05]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0026, 0.0040, 0.0061, 0.0107, 0.0271, 0.0363, -0.0270, 0.0108, + -0.0374, -0.0195], device='cuda:0'), grad: tensor([-7.1883e-05, 2.5645e-05, -1.3089e-04, 9.9778e-05, -3.7849e-05, + 2.6390e-05, 6.6981e-06, -2.0576e-04, 2.2829e-04, 5.9545e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 219.12, cls_loss 0.0038 cls_loss_mapping 0.0070 cls_loss_causal 0.5679 re_mapping 0.0087 re_causal 0.0257 /// teacc 98.96 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0224, 0.1109, 0.0653, ..., -0.1418, -0.0953, 0.0253], + [ 0.1013, -0.0102, -0.0466, ..., 0.0580, 0.0845, -0.1056], + [-0.0598, -0.0349, -0.0378, ..., -0.0618, -0.0152, -0.0368], + ..., + [-0.0372, 0.1234, -0.1433, ..., -0.1021, 0.0132, -0.1125], + [ 0.0014, -0.0803, 0.0297, ..., 0.0531, -0.1018, -0.0402], + [-0.0375, -0.0724, 0.0540, ..., -0.0657, -0.0451, 0.1096]], + device='cuda:0'), grad: tensor([[ 9.2201e-08, -6.8452e-08, 9.3132e-10, ..., 1.3132e-07, + 7.9162e-09, 9.3132e-10], + [-8.7731e-07, 3.7812e-07, 7.7300e-08, ..., -1.7649e-06, + -6.9384e-08, 4.6100e-08], + [ 8.2515e-07, 1.5348e-06, 5.6485e-07, ..., 8.9174e-07, + -2.4680e-08, 3.9581e-08], + ..., + [ 5.3504e-07, -4.7609e-06, 2.1933e-07, ..., 8.5123e-07, + 8.5216e-08, 1.4342e-07], + [-1.8626e-07, 6.3470e-07, -3.8184e-08, ..., -6.6124e-08, + -4.3772e-08, 1.2312e-06], + [ 2.6776e-07, 1.4426e-06, -3.0780e-07, ..., 6.8359e-07, + 1.0245e-08, -3.9907e-07]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0028, 0.0029, 0.0077, 0.0117, 0.0266, 0.0359, -0.0269, 0.0108, + -0.0376, -0.0196], device='cuda:0'), grad: tensor([ 7.6042e-07, 6.1700e-07, -4.9114e-05, 9.7677e-06, 1.2135e-06, + -3.4440e-06, 1.2722e-06, 2.8789e-05, 7.8827e-06, 2.2184e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 218.99, cls_loss 0.0045 cls_loss_mapping 0.0079 cls_loss_causal 0.5839 re_mapping 0.0086 re_causal 0.0256 /// teacc 98.93 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0235, 0.1116, 0.0657, ..., -0.1435, -0.0962, 0.0253], + [ 0.1023, -0.0084, -0.0471, ..., 0.0592, 0.0847, -0.1060], + [-0.0600, -0.0359, -0.0381, ..., -0.0620, -0.0153, -0.0370], + ..., + [-0.0388, 0.1231, -0.1432, ..., -0.1043, 0.0131, -0.1127], + [ 0.0007, -0.0824, 0.0289, ..., 0.0522, -0.1020, -0.0414], + [-0.0381, -0.0744, 0.0529, ..., -0.0665, -0.0455, 0.1091]], + device='cuda:0'), grad: tensor([[ 1.4640e-06, -1.6652e-06, -6.6031e-07, ..., 1.6252e-06, + -3.9581e-08, -1.5358e-06], + [-3.1758e-06, 1.3374e-06, 1.1846e-06, ..., -3.8780e-06, + -1.3765e-06, 1.1288e-06], + [ 2.9393e-06, 7.5018e-07, 1.3672e-06, ..., 3.3211e-06, + 1.5739e-07, 3.0873e-07], + ..., + [ 1.9222e-06, -7.5288e-06, 5.9642e-06, ..., 3.3937e-06, + 4.7637e-07, 6.3218e-06], + [-1.1757e-05, 8.1863e-07, -6.3404e-06, ..., -1.3158e-05, + 2.3935e-07, 1.0626e-06], + [ 8.3912e-07, 3.4235e-06, -1.5721e-05, ..., 1.0453e-05, + 1.2992e-07, -1.7345e-05]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0029, 0.0037, 0.0077, 0.0119, 0.0265, 0.0369, -0.0265, 0.0105, + -0.0386, -0.0205], device='cuda:0'), grad: tensor([ 1.7388e-06, -8.1863e-07, 3.5800e-06, 5.1767e-05, -6.8545e-06, + -3.5882e-05, 1.4342e-05, 6.0797e-06, -2.8431e-05, -5.4576e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 218.59, cls_loss 0.0033 cls_loss_mapping 0.0055 cls_loss_causal 0.5471 re_mapping 0.0088 re_causal 0.0248 /// teacc 98.92 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0238, 0.1118, 0.0659, ..., -0.1442, -0.0961, 0.0254], + [ 0.1025, -0.0085, -0.0475, ..., 0.0595, 0.0848, -0.1064], + [-0.0603, -0.0359, -0.0383, ..., -0.0623, -0.0152, -0.0371], + ..., + [-0.0391, 0.1236, -0.1438, ..., -0.1047, 0.0129, -0.1131], + [ 0.0009, -0.0827, 0.0290, ..., 0.0525, -0.1026, -0.0414], + [-0.0384, -0.0754, 0.0531, ..., -0.0675, -0.0458, 0.1094]], + device='cuda:0'), grad: tensor([[ 1.4948e-07, -3.9535e-07, -2.5425e-07, ..., 1.6205e-07, + 2.7940e-08, -4.2655e-07], + [-3.2783e-05, 4.5123e-07, 3.6508e-07, ..., -5.5999e-05, + -3.1233e-05, 3.1292e-07], + [ 1.5482e-05, 1.0291e-07, 1.2945e-06, ..., 2.5585e-05, + 1.4320e-05, 5.2154e-08], + ..., + [ 1.7107e-05, -2.0266e-06, 6.2473e-06, ..., 3.4153e-05, + 1.5423e-05, 1.5810e-05], + [ 5.0925e-06, 2.6310e-07, 1.1683e-05, ..., 2.7865e-06, + 1.3271e-07, 4.4703e-06], + [ 2.8545e-07, 6.1374e-07, -1.2413e-05, ..., -2.7522e-05, + 6.4261e-08, -7.6711e-05]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0029, 0.0037, 0.0079, 0.0116, 0.0269, 0.0372, -0.0265, 0.0106, + -0.0386, -0.0211], device='cuda:0'), grad: tensor([-5.3318e-07, -5.4687e-05, 2.8312e-05, -3.1829e-05, 1.0687e-04, + 5.9977e-06, -1.3318e-07, 5.7817e-05, 3.3587e-05, -1.4532e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 218.14, cls_loss 0.0042 cls_loss_mapping 0.0059 cls_loss_causal 0.5615 re_mapping 0.0086 re_causal 0.0247 /// teacc 98.88 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0246, 0.1120, 0.0663, ..., -0.1458, -0.0966, 0.0256], + [ 0.1020, -0.0086, -0.0486, ..., 0.0591, 0.0847, -0.1078], + [-0.0594, -0.0363, -0.0387, ..., -0.0618, -0.0149, -0.0373], + ..., + [-0.0393, 0.1240, -0.1443, ..., -0.1048, 0.0127, -0.1141], + [ 0.0024, -0.0824, 0.0283, ..., 0.0539, -0.1030, -0.0423], + [-0.0389, -0.0757, 0.0538, ..., -0.0676, -0.0455, 0.1104]], + device='cuda:0'), grad: tensor([[ 3.8091e-07, -6.9337e-07, -4.3539e-07, ..., 6.2538e-07, + 5.7742e-08, -7.7533e-07], + [ 1.8135e-05, 2.9756e-07, 5.1083e-07, ..., 3.5077e-05, + 5.0291e-06, 1.6764e-08], + [ 1.3262e-06, 2.4773e-07, -5.3318e-07, ..., 2.0117e-06, + 1.4529e-07, 5.4948e-08], + ..., + [ 4.6045e-06, 3.3528e-06, 2.3156e-05, ..., 7.1758e-07, + 8.5682e-08, 8.6147e-08], + [-2.0638e-05, 1.2433e-07, 1.2824e-06, ..., -4.0710e-05, + -5.7295e-06, 8.0094e-08], + [ 5.6485e-07, 9.6858e-07, 2.2668e-06, ..., 2.2352e-07, + 2.3283e-08, 3.1991e-07]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0029, 0.0026, 0.0092, 0.0119, 0.0269, 0.0372, -0.0279, 0.0107, + -0.0378, -0.0209], device='cuda:0'), grad: tensor([-3.0827e-07, 4.6611e-05, 2.5611e-07, -5.9783e-05, 1.5385e-06, + 5.9158e-06, -1.0626e-06, 4.9382e-05, -4.8250e-05, 5.8077e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 218.17, cls_loss 0.0039 cls_loss_mapping 0.0057 cls_loss_causal 0.5538 re_mapping 0.0086 re_causal 0.0249 /// teacc 98.92 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0251, 0.1112, 0.0653, ..., -0.1468, -0.0969, 0.0242], + [ 0.1028, -0.0088, -0.0487, ..., 0.0596, 0.0855, -0.1080], + [-0.0600, -0.0372, -0.0391, ..., -0.0623, -0.0155, -0.0375], + ..., + [-0.0396, 0.1247, -0.1449, ..., -0.1051, 0.0123, -0.1148], + [ 0.0022, -0.0826, 0.0282, ..., 0.0538, -0.1034, -0.0424], + [-0.0396, -0.0753, 0.0548, ..., -0.0678, -0.0457, 0.1119]], + device='cuda:0'), grad: tensor([[ 2.4140e-06, -1.9193e-05, -2.0206e-05, ..., 1.9092e-06, + 2.7940e-09, -2.2322e-05], + [ 6.9849e-08, 1.5851e-06, 7.2550e-07, ..., -2.2305e-07, + -6.8452e-08, 1.0151e-06], + [ 8.1304e-07, -8.3074e-06, 7.4506e-07, ..., 6.1430e-06, + 2.1886e-08, 6.4820e-07], + ..., + [ 4.6426e-07, 6.7279e-06, 1.1493e-06, ..., 7.0594e-07, + 4.3306e-08, 9.2611e-06], + [ 7.6741e-07, 2.0415e-06, 2.8722e-06, ..., 7.8091e-07, + 8.7079e-08, 2.0023e-06], + [ 7.0874e-07, 1.6332e-05, 1.3016e-05, ..., 1.6745e-06, + 1.2107e-08, 2.1840e-07]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0012, 0.0029, 0.0087, 0.0120, 0.0266, 0.0376, -0.0278, 0.0109, + -0.0381, -0.0198], device='cuda:0'), grad: tensor([-7.2718e-05, 8.9705e-06, -9.2268e-05, -2.0079e-06, 4.1455e-05, + 6.0052e-06, -3.6448e-05, 1.2124e-04, 9.4771e-06, 1.6317e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 218.89, cls_loss 0.0037 cls_loss_mapping 0.0046 cls_loss_causal 0.5610 re_mapping 0.0089 re_causal 0.0255 /// teacc 98.86 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0249, 0.1119, 0.0663, ..., -0.1465, -0.0972, 0.0244], + [ 0.1024, -0.0090, -0.0508, ..., 0.0585, 0.0857, -0.1093], + [-0.0600, -0.0380, -0.0376, ..., -0.0613, -0.0155, -0.0375], + ..., + [-0.0398, 0.1255, -0.1445, ..., -0.1053, 0.0121, -0.1138], + [ 0.0020, -0.0832, 0.0280, ..., 0.0536, -0.1038, -0.0428], + [-0.0400, -0.0764, 0.0542, ..., -0.0683, -0.0458, 0.1118]], + device='cuda:0'), grad: tensor([[-3.9302e-06, -7.8261e-05, -1.0943e-04, ..., -4.2468e-05, + 4.6566e-10, -1.0145e-04], + [ 3.1805e-07, 8.9360e-07, 1.0598e-06, ..., 2.1644e-06, + -1.3970e-09, 1.0319e-06], + [ 4.7078e-07, 6.8638e-07, -6.9011e-07, ..., 2.1514e-06, + 9.3132e-10, 3.5297e-07], + ..., + [ 1.5041e-06, 2.5854e-05, 2.9709e-06, ..., 8.7917e-05, + 1.8626e-09, 3.1572e-06], + [ 6.0320e-05, 1.9029e-05, 1.9506e-05, ..., 6.1929e-05, + 1.3039e-08, 3.7670e-05], + [ 1.0662e-05, 5.3495e-05, 5.6088e-05, ..., 7.2002e-05, + 0.0000e+00, 4.7833e-05]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0016, 0.0020, 0.0094, 0.0121, 0.0262, 0.0380, -0.0275, 0.0118, + -0.0384, -0.0206], device='cuda:0'), grad: tensor([-2.0099e-04, 6.1840e-06, 1.5507e-06, 1.3866e-05, -3.1114e-04, + -3.9554e-04, 3.7408e-04, 2.0599e-04, 9.9599e-05, 2.0742e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 217.69, cls_loss 0.0038 cls_loss_mapping 0.0059 cls_loss_causal 0.5678 re_mapping 0.0080 re_causal 0.0242 /// teacc 98.82 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0255, 0.1126, 0.0667, ..., -0.1473, -0.0978, 0.0246], + [ 0.1028, -0.0096, -0.0518, ..., 0.0590, 0.0860, -0.1096], + [-0.0605, -0.0388, -0.0374, ..., -0.0618, -0.0157, -0.0379], + ..., + [-0.0399, 0.1263, -0.1450, ..., -0.1056, 0.0121, -0.1144], + [ 0.0027, -0.0821, 0.0295, ..., 0.0548, -0.1052, -0.0420], + [-0.0412, -0.0770, 0.0542, ..., -0.0691, -0.0459, 0.1118]], + device='cuda:0'), grad: tensor([[ 1.8510e-07, -3.0873e-07, -1.2550e-07, ..., 3.6694e-07, + 7.9721e-07, -4.1444e-08], + [-1.3009e-05, 2.4904e-06, -2.5844e-08, ..., -2.1622e-05, + 3.7672e-07, 1.3690e-07], + [ 2.2110e-06, 6.6590e-07, 1.3527e-07, ..., 3.7793e-06, + -1.2226e-05, 3.5623e-08], + ..., + [ 1.6429e-06, -5.6103e-06, 6.4913e-07, ..., 1.7891e-06, + 1.3225e-06, 8.8476e-07], + [ 1.2584e-05, 3.0990e-07, 4.8764e-06, ..., 2.3305e-05, + 3.6019e-07, 1.0192e-05], + [ 3.0990e-07, 1.3523e-06, -2.1309e-06, ..., 1.6671e-06, + 3.8906e-07, -3.0492e-06]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0018, 0.0019, 0.0093, 0.0118, 0.0262, 0.0380, -0.0278, 0.0120, + -0.0378, -0.0208], device='cuda:0'), grad: tensor([ 1.7077e-05, -3.5074e-06, -2.6202e-04, 1.8990e-04, 3.2596e-06, + -1.7792e-05, 8.9258e-06, 2.4125e-05, 3.1859e-05, 8.0690e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 216.95, cls_loss 0.0045 cls_loss_mapping 0.0078 cls_loss_causal 0.5343 re_mapping 0.0086 re_causal 0.0235 /// teacc 98.98 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0260, 0.1134, 0.0674, ..., -0.1482, -0.0981, 0.0251], + [ 0.1046, -0.0096, -0.0511, ..., 0.0607, 0.0884, -0.1087], + [-0.0612, -0.0397, -0.0375, ..., -0.0624, -0.0172, -0.0394], + ..., + [-0.0409, 0.1270, -0.1454, ..., -0.1066, 0.0104, -0.1151], + [ 0.0023, -0.0817, 0.0295, ..., 0.0542, -0.1056, -0.0423], + [-0.0425, -0.0781, 0.0548, ..., -0.0673, -0.0462, 0.1134]], + device='cuda:0'), grad: tensor([[ 7.7067e-08, -4.2953e-06, -2.7511e-06, ..., 2.3469e-07, + 1.6904e-07, -4.8503e-06], + [-2.1188e-08, 2.9020e-06, 9.4064e-07, ..., 1.2480e-07, + 4.4308e-07, 2.9965e-07], + [ 2.0373e-07, 7.8827e-06, 3.8370e-06, ..., 8.6054e-07, + 1.2387e-06, 8.7544e-07], + ..., + [ 4.2259e-07, -2.8282e-05, 8.1211e-06, ..., 2.1756e-06, + 4.9220e-07, -2.7083e-06], + [ 6.1374e-07, 7.1479e-07, 2.3879e-06, ..., 1.1232e-06, + 8.3260e-07, 9.2667e-07], + [ 1.1967e-07, 8.9109e-06, 2.9784e-06, ..., 2.7902e-06, + 2.0228e-06, 5.1335e-06]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0022, 0.0031, 0.0089, 0.0114, 0.0240, 0.0379, -0.0271, 0.0118, + -0.0387, -0.0199], device='cuda:0'), grad: tensor([-9.1046e-06, 4.5240e-05, -2.1458e-05, -3.5763e-05, 4.1425e-06, + 1.0140e-05, 2.7083e-06, -3.0190e-05, 7.5623e-06, 2.6688e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 217.09, cls_loss 0.0030 cls_loss_mapping 0.0053 cls_loss_causal 0.6039 re_mapping 0.0082 re_causal 0.0251 /// teacc 98.94 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0270, 0.1138, 0.0676, ..., -0.1489, -0.0989, 0.0252], + [ 0.1051, -0.0099, -0.0512, ..., 0.0612, 0.0889, -0.1086], + [-0.0615, -0.0404, -0.0375, ..., -0.0625, -0.0173, -0.0399], + ..., + [-0.0410, 0.1276, -0.1462, ..., -0.1067, 0.0102, -0.1163], + [ 0.0020, -0.0824, 0.0292, ..., 0.0539, -0.1058, -0.0426], + [-0.0430, -0.0785, 0.0551, ..., -0.0675, -0.0464, 0.1138]], + device='cuda:0'), grad: tensor([[ 2.7511e-06, -1.0068e-06, -1.2224e-07, ..., 3.1292e-06, + 3.4692e-08, -3.9674e-07], + [-1.3188e-05, 9.6392e-07, 2.8014e-05, ..., 3.2205e-06, + -3.7462e-05, 2.9713e-05], + [ 2.0698e-05, 5.6252e-07, 1.6112e-06, ..., 1.7837e-05, + 3.2395e-05, 8.9966e-07], + ..., + [ 3.9004e-06, -3.4571e-06, 1.6931e-06, ..., 4.2245e-06, + 4.8839e-06, 1.3681e-06], + [ 1.9848e-04, 2.0908e-07, 1.3411e-06, ..., 2.1315e-04, + 1.8883e-07, 3.9011e-05], + [-1.8299e-05, 1.6401e-06, -6.7651e-05, ..., -4.7237e-05, + 2.2328e-07, -7.1585e-05]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0021, 0.0031, 0.0088, 0.0116, 0.0239, 0.0382, -0.0269, 0.0120, + -0.0392, -0.0200], device='cuda:0'), grad: tensor([ 4.2766e-06, 1.9580e-05, 5.0694e-05, -1.3374e-05, 9.0659e-05, + 4.9710e-05, -4.8566e-04, 8.1733e-06, 4.2439e-04, -1.4806e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 217.19, cls_loss 0.0041 cls_loss_mapping 0.0070 cls_loss_causal 0.5723 re_mapping 0.0082 re_causal 0.0247 /// teacc 98.85 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0271, 0.1122, 0.0667, ..., -0.1495, -0.0990, 0.0231], + [ 0.1055, -0.0104, -0.0514, ..., 0.0614, 0.0894, -0.1091], + [-0.0619, -0.0410, -0.0384, ..., -0.0628, -0.0177, -0.0406], + ..., + [-0.0411, 0.1283, -0.1466, ..., -0.1066, 0.0098, -0.1173], + [ 0.0025, -0.0828, 0.0294, ..., 0.0543, -0.1060, -0.0426], + [-0.0435, -0.0772, 0.0564, ..., -0.0678, -0.0466, 0.1163]], + device='cuda:0'), grad: tensor([[ 5.5227e-07, -7.4785e-07, 2.5239e-07, ..., 7.8557e-07, + 1.2107e-08, -2.4633e-07], + [-4.1090e-06, 6.9709e-07, 3.8510e-07, ..., -9.4026e-06, + -1.5199e-06, 3.4040e-07], + [ 2.1886e-06, 2.4540e-07, 5.0571e-07, ..., 4.3884e-06, + 8.1258e-07, 2.5565e-07], + ..., + [ 8.0094e-07, -2.2799e-06, 1.7285e-06, ..., 1.3858e-06, + 1.8207e-07, 2.5909e-06], + [ 4.3027e-07, 3.3341e-07, -4.3958e-07, ..., 9.5926e-08, + 1.5413e-07, 5.4240e-06], + [ 2.1607e-06, 7.6648e-07, 1.9670e-06, ..., 2.8424e-06, + 4.6566e-08, 9.3598e-07]], device='cuda:0') +Epoch 132, bias, value: tensor([ 8.0441e-05, 2.7230e-03, 8.1190e-03, 1.2262e-02, 2.3700e-02, + 3.8310e-02, -2.7228e-02, 1.2495e-02, -3.8886e-02, -1.8500e-02], + device='cuda:0'), grad: tensor([ 3.1106e-06, -7.9423e-06, -1.9789e-05, 3.1263e-05, 3.1777e-06, + -3.4034e-05, 6.0871e-06, 9.4548e-06, 3.8035e-06, 4.8205e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 217.26, cls_loss 0.0040 cls_loss_mapping 0.0060 cls_loss_causal 0.5742 re_mapping 0.0083 re_causal 0.0240 /// teacc 98.91 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0264, 0.1138, 0.0674, ..., -0.1497, -0.0995, 0.0236], + [ 0.1060, -0.0107, -0.0514, ..., 0.0621, 0.0898, -0.1087], + [-0.0621, -0.0439, -0.0411, ..., -0.0638, -0.0181, -0.0442], + ..., + [-0.0413, 0.1290, -0.1472, ..., -0.1072, 0.0095, -0.1181], + [ 0.0020, -0.0833, 0.0286, ..., 0.0539, -0.1067, -0.0436], + [-0.0440, -0.0780, 0.0565, ..., -0.0684, -0.0473, 0.1163]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, -5.7191e-05, -2.0385e-05, ..., 2.1281e-07, + 3.7253e-09, -8.0228e-05], + [-3.1590e-06, 2.8312e-07, 1.9046e-07, ..., -2.7977e-06, + -4.7823e-07, 2.8312e-07], + [ 7.0082e-07, 2.1085e-06, 7.2876e-07, ..., 1.3653e-06, + 8.8476e-09, 2.7698e-06], + ..., + [ 1.2852e-06, 1.6570e-05, 6.2510e-06, ..., 9.7007e-06, + 1.1874e-07, 2.4512e-05], + [ 1.3644e-07, 3.7178e-06, 1.9241e-06, ..., 3.0547e-07, + 1.1642e-08, 6.2175e-06], + [ 7.6881e-07, 3.1721e-06, -3.0287e-06, ..., 1.3381e-05, + 2.9989e-07, -2.9709e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0010, 0.0027, 0.0065, 0.0127, 0.0241, 0.0400, -0.0275, 0.0125, + -0.0397, -0.0189], device='cuda:0'), grad: tensor([-1.7560e-04, -3.2764e-06, 8.3372e-06, 8.3372e-06, -4.1604e-05, + 6.7353e-05, 2.6450e-05, 7.0333e-05, 1.3143e-05, 2.6584e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 217.19, cls_loss 0.0032 cls_loss_mapping 0.0061 cls_loss_causal 0.5685 re_mapping 0.0084 re_causal 0.0237 /// teacc 98.97 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0269, 0.1142, 0.0676, ..., -0.1506, -0.0996, 0.0238], + [ 0.1065, -0.0108, -0.0515, ..., 0.0628, 0.0902, -0.1089], + [-0.0623, -0.0447, -0.0413, ..., -0.0643, -0.0181, -0.0448], + ..., + [-0.0417, 0.1296, -0.1478, ..., -0.1078, 0.0087, -0.1190], + [ 0.0020, -0.0835, 0.0289, ..., 0.0539, -0.1067, -0.0437], + [-0.0445, -0.0784, 0.0566, ..., -0.0693, -0.0475, 0.1164]], + device='cuda:0'), grad: tensor([[ 2.7800e-07, -3.4068e-06, -5.1893e-06, ..., 6.6590e-07, + 3.6741e-07, -9.1642e-06], + [-7.6771e-05, 3.1479e-06, 1.1409e-07, ..., -1.9920e-04, + -1.6487e-04, 1.4855e-07], + [ 6.8665e-05, 6.7940e-07, 2.0256e-07, ..., 1.8036e-04, + 1.4806e-04, 2.1188e-07], + ..., + [ 6.3740e-06, -6.1169e-06, 2.3469e-07, ..., 1.3106e-05, + 1.2286e-05, 3.2922e-07], + [ 4.1910e-09, 1.2089e-06, -1.4016e-07, ..., 3.3947e-07, + 6.1281e-07, 3.3621e-07], + [ 2.4913e-07, 1.6131e-06, 2.2724e-06, ..., 6.8359e-07, + 3.8277e-07, 3.6061e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0011, 0.0026, 0.0064, 0.0122, 0.0246, 0.0400, -0.0271, 0.0126, + -0.0396, -0.0193], device='cuda:0'), grad: tensor([-1.2517e-05, -5.7459e-04, 5.2214e-04, 6.5453e-06, 4.6901e-06, + 1.7080e-06, 7.1190e-06, 3.3140e-05, 3.5297e-06, 7.4692e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 217.92, cls_loss 0.0039 cls_loss_mapping 0.0068 cls_loss_causal 0.5475 re_mapping 0.0081 re_causal 0.0236 /// teacc 98.95 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0271, 0.1147, 0.0681, ..., -0.1512, -0.0998, 0.0241], + [ 0.1084, -0.0106, -0.0490, ..., 0.0650, 0.0907, -0.1066], + [-0.0629, -0.0450, -0.0414, ..., -0.0653, -0.0188, -0.0456], + ..., + [-0.0425, 0.1295, -0.1498, ..., -0.1092, 0.0091, -0.1195], + [ 0.0015, -0.0841, 0.0282, ..., 0.0539, -0.1060, -0.0444], + [-0.0470, -0.0790, 0.0560, ..., -0.0702, -0.0481, 0.1162]], + device='cuda:0'), grad: tensor([[ 6.0815e-07, 4.6380e-07, 4.5868e-07, ..., 1.0924e-06, + 4.0233e-07, 9.8534e-07], + [ 1.8720e-07, 8.4043e-06, 1.2685e-06, ..., 7.6974e-07, + 3.7067e-07, 2.3693e-06], + [ 2.9663e-07, 3.9876e-05, -3.9116e-07, ..., 3.0734e-07, + -1.4538e-06, 2.1327e-07], + ..., + [ 8.8150e-07, -7.5877e-05, -4.6240e-07, ..., 1.3057e-06, + -9.4855e-07, -3.4785e-07], + [ 3.4094e-04, 9.0718e-05, 3.1233e-04, ..., 4.1032e-04, + 8.3447e-07, 3.8600e-04], + [ 1.0291e-06, 1.1154e-05, 3.7905e-07, ..., 4.3809e-06, + 1.1288e-06, -3.1246e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0014, 0.0038, 0.0061, 0.0114, 0.0244, 0.0404, -0.0265, 0.0121, + -0.0396, -0.0200], device='cuda:0'), grad: tensor([ 5.9083e-06, 2.9042e-05, 1.3523e-05, 5.7578e-05, -5.3011e-06, + -8.1348e-04, 7.4744e-05, -1.4639e-04, 7.5722e-04, 2.7373e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 217.31, cls_loss 0.0027 cls_loss_mapping 0.0050 cls_loss_causal 0.5486 re_mapping 0.0082 re_causal 0.0238 /// teacc 98.84 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0276, 0.1150, 0.0680, ..., -0.1521, -0.1000, 0.0241], + [ 0.1089, -0.0106, -0.0495, ..., 0.0652, 0.0910, -0.1075], + [-0.0632, -0.0452, -0.0412, ..., -0.0655, -0.0188, -0.0460], + ..., + [-0.0428, 0.1297, -0.1508, ..., -0.1099, 0.0091, -0.1215], + [ 0.0015, -0.0847, 0.0281, ..., 0.0539, -0.1065, -0.0449], + [-0.0471, -0.0791, 0.0570, ..., -0.0702, -0.0484, 0.1173]], + device='cuda:0'), grad: tensor([[ 4.0978e-06, -7.4739e-07, 2.0534e-05, ..., 1.7658e-05, + 9.7789e-09, 1.6615e-05], + [-4.5542e-07, 1.8347e-07, 1.2089e-06, ..., 1.3597e-07, + -2.4028e-07, 1.3355e-06], + [ 4.4703e-07, 7.4040e-08, 1.4529e-07, ..., 1.3169e-06, + 1.1642e-08, 1.1474e-06], + ..., + [ 3.5111e-07, -1.1697e-06, 6.8471e-06, ..., 3.2876e-06, + 8.8941e-08, 1.2293e-05], + [-6.3956e-05, 1.6158e-07, -3.4881e-04, ..., -2.8896e-04, + 4.0047e-08, -2.7919e-04], + [ 1.6943e-05, 1.2396e-06, 7.4089e-05, ..., 6.9022e-05, + 6.2399e-08, 4.1902e-05]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0013, 0.0038, 0.0064, 0.0109, 0.0245, 0.0403, -0.0265, 0.0116, + -0.0400, -0.0193], device='cuda:0'), grad: tensor([ 3.9220e-05, 2.0694e-06, -1.0081e-05, 1.2226e-05, 2.0802e-05, + 4.1747e-04, 2.3201e-05, 2.1100e-05, -6.4754e-04, 1.2147e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 217.41, cls_loss 0.0034 cls_loss_mapping 0.0045 cls_loss_causal 0.5399 re_mapping 0.0085 re_causal 0.0240 /// teacc 98.85 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0278, 0.1151, 0.0682, ..., -0.1528, -0.1000, 0.0242], + [ 0.1094, -0.0107, -0.0505, ..., 0.0654, 0.0911, -0.1078], + [-0.0635, -0.0455, -0.0416, ..., -0.0658, -0.0190, -0.0461], + ..., + [-0.0435, 0.1293, -0.1533, ..., -0.1107, 0.0096, -0.1208], + [ 0.0014, -0.0851, 0.0280, ..., 0.0540, -0.1069, -0.0451], + [-0.0476, -0.0782, 0.0589, ..., -0.0705, -0.0497, 0.1175]], + device='cuda:0'), grad: tensor([[ 1.9401e-05, 3.9525e-06, 9.4697e-06, ..., 7.3910e-06, + 2.0489e-08, 6.6757e-06], + [-2.6226e-04, -7.3314e-05, -1.0854e-04, ..., -1.0622e-04, + -1.2062e-05, -7.1883e-05], + [ 4.3325e-06, 9.7416e-07, 2.1271e-06, ..., 1.8841e-06, + 8.1956e-08, 1.3188e-06], + ..., + [ 6.6042e-05, 2.6807e-05, 1.6123e-05, ..., 3.3021e-05, + 1.0431e-05, 1.1280e-05], + [ 1.2022e-04, 2.5049e-05, 5.9277e-05, ..., 4.5061e-05, + 4.9872e-07, 4.2379e-05], + [ 4.2558e-05, 1.3739e-05, 5.3458e-07, ..., 1.8045e-05, + 5.3877e-07, -1.8030e-05]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0013, 0.0034, 0.0062, 0.0118, 0.0247, 0.0398, -0.0266, 0.0109, + -0.0401, -0.0183], device='cuda:0'), grad: tensor([ 3.9309e-05, -5.4216e-04, 9.8124e-06, 7.9051e-06, -4.8280e-06, + 3.0160e-05, 1.0924e-06, 1.4102e-04, 2.4533e-04, 7.2002e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 217.30, cls_loss 0.0032 cls_loss_mapping 0.0055 cls_loss_causal 0.5481 re_mapping 0.0075 re_causal 0.0227 /// teacc 98.98 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0283, 0.1154, 0.0684, ..., -0.1536, -0.0999, 0.0243], + [ 0.1101, -0.0110, -0.0509, ..., 0.0656, 0.0920, -0.1081], + [-0.0636, -0.0457, -0.0403, ..., -0.0659, -0.0191, -0.0448], + ..., + [-0.0439, 0.1299, -0.1545, ..., -0.1114, 0.0086, -0.1222], + [ 0.0013, -0.0854, 0.0278, ..., 0.0541, -0.1069, -0.0451], + [-0.0485, -0.0786, 0.0592, ..., -0.0707, -0.0503, 0.1181]], + device='cuda:0'), grad: tensor([[ 3.8464e-07, 4.1910e-08, 1.8114e-07, ..., 3.6974e-07, + -1.1409e-07, -2.4028e-07], + [-1.1809e-05, 2.0996e-05, 1.8226e-06, ..., 3.8631e-06, + -3.3379e-06, 6.7195e-07], + [ 2.3190e-06, 4.0531e-06, 3.3826e-06, ..., 7.6741e-07, + 8.5961e-07, 2.8592e-07], + ..., + [ 1.2279e-05, -2.7850e-05, 3.0398e-06, ..., 1.6317e-05, + 1.4994e-06, 7.9116e-07], + [-1.4268e-05, 9.7323e-08, -7.6145e-06, ..., -2.4617e-05, + 1.9837e-07, -2.0221e-05], + [ 1.7092e-05, 7.4320e-07, 2.1309e-05, ..., 2.9400e-05, + 5.0850e-07, 1.3776e-05]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0013, 0.0034, 0.0070, 0.0120, 0.0248, 0.0400, -0.0266, 0.0104, + -0.0406, -0.0183], device='cuda:0'), grad: tensor([ 1.1725e-06, 3.8117e-05, 1.6496e-05, -7.5340e-05, -5.6535e-05, + 1.9744e-05, 6.1169e-06, -7.5214e-06, -2.1741e-05, 7.9334e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 217.11, cls_loss 0.0046 cls_loss_mapping 0.0070 cls_loss_causal 0.5767 re_mapping 0.0080 re_causal 0.0235 /// teacc 98.92 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0285, 0.1155, 0.0694, ..., -0.1560, -0.1001, 0.0245], + [ 0.1112, -0.0112, -0.0504, ..., 0.0665, 0.0922, -0.1077], + [-0.0640, -0.0462, -0.0407, ..., -0.0664, -0.0188, -0.0451], + ..., + [-0.0445, 0.1311, -0.1542, ..., -0.1121, 0.0085, -0.1219], + [ 0.0034, -0.0856, 0.0331, ..., 0.0560, -0.1063, -0.0402], + [-0.0522, -0.0799, 0.0554, ..., -0.0707, -0.0521, 0.1158]], + device='cuda:0'), grad: tensor([[ 6.5677e-06, 1.0245e-06, 5.8785e-06, ..., 8.3148e-06, + 1.0245e-08, 4.3958e-06], + [ 3.1898e-07, 8.8988e-07, 1.3988e-06, ..., 1.4454e-06, + -1.3737e-07, 1.2890e-06], + [ 9.3551e-07, 4.1118e-07, 1.3877e-06, ..., 1.3793e-06, + 7.4971e-08, 7.7765e-07], + ..., + [ 6.8313e-07, -4.7199e-06, 7.7710e-06, ..., 3.1628e-06, + 1.3132e-07, 1.1645e-05], + [ 4.1444e-07, 1.7043e-07, -4.9025e-06, ..., -4.2059e-06, + 6.0862e-07, -2.7139e-06], + [ 4.7637e-07, 3.4496e-06, -1.2487e-05, ..., 2.6934e-06, + 5.1223e-08, -1.9103e-05]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0015, 0.0036, 0.0071, 0.0119, 0.0229, 0.0400, -0.0274, 0.0108, + -0.0362, -0.0208], device='cuda:0'), grad: tensor([ 3.5256e-05, 7.7784e-06, -7.5400e-05, -2.4419e-06, 3.5554e-05, + 1.8522e-05, -1.2554e-05, 2.8431e-05, -2.8927e-06, -3.2395e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 217.23, cls_loss 0.0034 cls_loss_mapping 0.0063 cls_loss_causal 0.5555 re_mapping 0.0078 re_causal 0.0228 /// teacc 98.81 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0288, 0.1160, 0.0697, ..., -0.1570, -0.1003, 0.0245], + [ 0.1116, -0.0116, -0.0495, ..., 0.0662, 0.0936, -0.1074], + [-0.0643, -0.0465, -0.0409, ..., -0.0664, -0.0189, -0.0453], + ..., + [-0.0447, 0.1324, -0.1533, ..., -0.1119, 0.0064, -0.1214], + [ 0.0033, -0.0865, 0.0331, ..., 0.0560, -0.1068, -0.0402], + [-0.0536, -0.0812, 0.0551, ..., -0.0708, -0.0522, 0.1159]], + device='cuda:0'), grad: tensor([[ 1.7136e-07, -1.4171e-05, -3.5971e-05, ..., -6.3218e-06, + 2.5611e-08, -2.3663e-05], + [-3.6089e-07, 9.4390e-07, 1.0207e-06, ..., -2.5984e-07, + -2.0489e-07, 8.4704e-07], + [ 1.2806e-07, 4.8168e-06, 1.3277e-05, ..., 2.5481e-06, + 2.1886e-08, 8.7470e-06], + ..., + [ 3.6228e-07, -4.0419e-06, 2.2240e-06, ..., 5.1456e-07, + 1.0896e-07, 2.5369e-06], + [ 4.6305e-06, 1.5898e-06, 2.6956e-05, ..., 1.1235e-05, + 6.3796e-08, 4.1813e-05], + [-1.4724e-06, 4.0568e-06, -2.0549e-05, ..., -5.9940e-06, + 6.1933e-08, -4.3929e-05]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0015, 0.0030, 0.0076, 0.0124, 0.0225, 0.0396, -0.0270, 0.0116, + -0.0363, -0.0211], device='cuda:0'), grad: tensor([-8.1003e-05, 2.6524e-06, 2.9922e-05, 2.6748e-06, 7.6294e-06, + 4.0904e-06, 1.7151e-05, -8.1677e-07, 5.1975e-05, -3.4362e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 217.21, cls_loss 0.0042 cls_loss_mapping 0.0055 cls_loss_causal 0.5839 re_mapping 0.0083 re_causal 0.0231 /// teacc 98.69 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0294, 0.1163, 0.0699, ..., -0.1582, -0.1015, 0.0245], + [ 0.1112, -0.0119, -0.0499, ..., 0.0658, 0.0929, -0.1079], + [-0.0646, -0.0477, -0.0423, ..., -0.0669, -0.0189, -0.0443], + ..., + [-0.0450, 0.1334, -0.1535, ..., -0.1125, 0.0062, -0.1224], + [ 0.0031, -0.0871, 0.0334, ..., 0.0563, -0.1049, -0.0406], + [-0.0541, -0.0817, 0.0551, ..., -0.0714, -0.0525, 0.1161]], + device='cuda:0'), grad: tensor([[ 3.7253e-07, -1.5572e-05, -2.5071e-06, ..., 4.1816e-07, + -1.3888e-05, -8.5309e-06], + [-1.3039e-06, 3.6182e-07, 6.4354e-07, ..., 4.0270e-06, + 4.3213e-07, 1.1222e-07], + [ 1.4007e-06, 1.8589e-06, -6.8173e-07, ..., 6.6310e-07, + -1.2845e-05, 3.4412e-07], + ..., + [ 2.9281e-06, 1.9111e-06, 5.6662e-06, ..., 1.1295e-05, + 1.8794e-06, 4.7591e-07], + [ 5.7295e-06, 1.5516e-06, 3.8035e-06, ..., 6.1393e-06, + 6.9849e-06, 6.8396e-06], + [ 8.2003e-07, 2.9374e-06, 1.1744e-06, ..., 5.4277e-06, + 2.1644e-06, 4.7637e-07]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0014, 0.0023, 0.0062, 0.0121, 0.0225, 0.0397, -0.0261, 0.0118, + -0.0350, -0.0215], device='cuda:0'), grad: tensor([-2.2560e-05, 9.2268e-05, -3.3321e-03, 4.2558e-04, -3.2216e-05, + -4.9233e-05, 2.8908e-05, 2.7657e-03, 9.8765e-05, 2.2471e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 217.61, cls_loss 0.0063 cls_loss_mapping 0.0072 cls_loss_causal 0.5785 re_mapping 0.0085 re_causal 0.0238 /// teacc 98.90 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0297, 0.1171, 0.0702, ..., -0.1590, -0.1013, 0.0247], + [ 0.1124, -0.0122, -0.0502, ..., 0.0662, 0.0932, -0.1082], + [-0.0652, -0.0504, -0.0442, ..., -0.0679, -0.0193, -0.0443], + ..., + [-0.0462, 0.1345, -0.1539, ..., -0.1139, 0.0059, -0.1230], + [ 0.0028, -0.0875, 0.0359, ..., 0.0560, -0.1041, -0.0382], + [-0.0544, -0.0820, 0.0533, ..., -0.0719, -0.0526, 0.1146]], + device='cuda:0'), grad: tensor([[ 1.9372e-07, -1.1129e-07, 4.0652e-07, ..., 4.1630e-07, + 3.3062e-08, 3.0082e-07], + [-1.3504e-08, 3.1851e-07, 3.4971e-07, ..., -1.0915e-05, + -2.5555e-06, 2.8918e-07], + [ 1.3083e-05, 9.5461e-08, 3.6415e-07, ..., 1.7852e-05, + 5.4948e-07, 4.2841e-08], + ..., + [ 7.4506e-08, -4.9919e-06, 2.5239e-07, ..., 5.1558e-06, + 1.1474e-06, 1.8720e-07], + [ 4.7451e-07, 6.3796e-08, 5.5954e-06, ..., 2.3320e-06, + 2.6263e-07, 5.6773e-06], + [ 1.6019e-07, 4.2431e-06, -1.8120e-05, ..., -2.9467e-06, + 6.6124e-08, -1.6928e-05]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0016, 0.0021, 0.0055, 0.0118, 0.0231, 0.0395, -0.0252, 0.0118, + -0.0329, -0.0232], device='cuda:0'), grad: tensor([ 1.1930e-06, -1.5527e-05, 6.1333e-05, 6.8173e-06, 1.9342e-05, + -4.5300e-06, -6.2048e-05, 1.7779e-06, 1.0975e-05, -1.9431e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 217.74, cls_loss 0.0039 cls_loss_mapping 0.0046 cls_loss_causal 0.5521 re_mapping 0.0088 re_causal 0.0244 /// teacc 98.90 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0298, 0.1175, 0.0703, ..., -0.1592, -0.1014, 0.0248], + [ 0.1132, -0.0124, -0.0498, ..., 0.0669, 0.0934, -0.1081], + [-0.0652, -0.0491, -0.0471, ..., -0.0702, -0.0211, -0.0445], + ..., + [-0.0465, 0.1341, -0.1547, ..., -0.1142, 0.0057, -0.1235], + [ 0.0026, -0.0881, 0.0356, ..., 0.0556, -0.1042, -0.0384], + [-0.0547, -0.0818, 0.0538, ..., -0.0720, -0.0528, 0.1152]], + device='cuda:0'), grad: tensor([[ 2.8424e-06, -5.7695e-07, 4.0457e-06, ..., 9.7156e-06, + 4.6566e-10, 3.7160e-06], + [ 4.1910e-09, 2.6748e-06, 5.9698e-07, ..., 3.6554e-07, + -3.6787e-08, 6.6869e-07], + [ 2.5295e-06, 5.5414e-07, 4.3586e-06, ..., 8.6948e-06, + 4.6566e-09, 4.0643e-06], + ..., + [-2.4214e-08, -2.3916e-05, 1.4193e-06, ..., 4.7218e-07, + 1.8626e-08, 1.6587e-06], + [ 4.8392e-06, 5.3346e-06, 1.3806e-05, ..., 1.6183e-05, + 4.1910e-09, 1.7002e-05], + [ 1.4389e-07, 7.0967e-07, -1.2435e-05, ..., 3.2820e-06, + 2.7940e-09, -1.7881e-05]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0016, 0.0023, 0.0041, 0.0117, 0.0227, 0.0418, -0.0250, 0.0110, + -0.0334, -0.0227], device='cuda:0'), grad: tensor([ 1.9073e-05, 6.5342e-06, 1.8761e-05, 2.7806e-05, 1.8422e-06, + 4.8399e-04, -5.5647e-04, -3.9190e-05, 5.9187e-05, -2.1219e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 217.68, cls_loss 0.0041 cls_loss_mapping 0.0053 cls_loss_causal 0.5561 re_mapping 0.0079 re_causal 0.0230 /// teacc 98.88 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0305, 0.1180, 0.0706, ..., -0.1600, -0.1030, 0.0248], + [ 0.1133, -0.0125, -0.0504, ..., 0.0669, 0.0933, -0.1083], + [-0.0650, -0.0491, -0.0469, ..., -0.0701, -0.0211, -0.0447], + ..., + [-0.0466, 0.1350, -0.1548, ..., -0.1149, 0.0059, -0.1239], + [ 0.0027, -0.0886, 0.0356, ..., 0.0559, -0.1042, -0.0385], + [-0.0549, -0.0824, 0.0540, ..., -0.0726, -0.0525, 0.1154]], + device='cuda:0'), grad: tensor([[ 6.7139e-04, 3.6812e-04, -3.7253e-08, ..., 6.9551e-06, + 1.2178e-03, 1.9836e-04], + [-1.1265e-05, 3.2410e-07, 9.0338e-08, ..., -1.8269e-05, + 6.5612e-07, 2.1746e-07], + [ 7.7635e-06, 1.0133e-06, 1.0245e-07, ..., 9.6485e-06, + 2.7008e-06, 5.4669e-07], + ..., + [ 4.2915e-06, -5.8068e-07, 1.5367e-07, ..., 6.7279e-06, + 1.4435e-07, 1.0012e-07], + [ 7.4971e-07, 1.0421e-06, -1.9390e-06, ..., -4.2170e-06, + 1.6969e-06, -2.8200e-06], + [ 2.1942e-06, 6.7195e-07, 2.1886e-06, ..., 6.2101e-06, + 8.2795e-07, 3.1181e-06]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0013, 0.0018, 0.0044, 0.0129, 0.0233, 0.0404, -0.0246, 0.0116, + -0.0334, -0.0231], device='cuda:0'), grad: tensor([ 2.3117e-03, -5.1469e-05, 3.2634e-05, 1.6832e-03, 1.2731e-06, + -1.6518e-03, -2.3556e-03, 1.9148e-05, -6.6590e-08, 1.3053e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 217.10, cls_loss 0.0029 cls_loss_mapping 0.0051 cls_loss_causal 0.5340 re_mapping 0.0079 re_causal 0.0231 /// teacc 98.89 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0314, 0.1182, 0.0712, ..., -0.1604, -0.1044, 0.0251], + [ 0.1139, -0.0127, -0.0504, ..., 0.0670, 0.0942, -0.1084], + [-0.0656, -0.0488, -0.0477, ..., -0.0702, -0.0218, -0.0450], + ..., + [-0.0472, 0.1352, -0.1552, ..., -0.1155, 0.0057, -0.1245], + [ 0.0027, -0.0889, 0.0359, ..., 0.0560, -0.1043, -0.0386], + [-0.0554, -0.0827, 0.0539, ..., -0.0737, -0.0527, 0.1154]], + device='cuda:0'), grad: tensor([[ 3.4692e-07, -2.7902e-06, -4.6343e-06, ..., 5.4203e-07, + 8.5216e-08, -5.3346e-06], + [ 2.9877e-06, 2.6003e-06, 3.2391e-06, ..., 3.0864e-06, + -1.9977e-07, 4.4517e-07], + [ 3.3975e-06, 2.9011e-07, 2.8368e-06, ..., 2.4978e-06, + -2.6077e-08, 2.8173e-07], + ..., + [ 1.6019e-05, -4.6790e-06, 1.4462e-05, ..., 1.2591e-05, + 4.2375e-07, 2.6580e-06], + [ 1.1779e-05, -1.5646e-06, 5.8532e-05, ..., 2.2382e-05, + 1.1228e-05, 8.0943e-05], + [ 1.2200e-06, 1.6391e-06, -4.7803e-05, ..., -1.7449e-05, + -1.2986e-05, -9.1970e-05]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0013, 0.0019, 0.0035, 0.0128, 0.0245, 0.0404, -0.0244, 0.0115, + -0.0327, -0.0237], device='cuda:0'), grad: tensor([-1.4074e-05, 2.1204e-05, 1.4484e-05, -1.9777e-04, 2.4494e-07, + 4.4674e-05, 3.6284e-06, 6.2048e-05, 1.5354e-04, -8.8036e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 217.13, cls_loss 0.0030 cls_loss_mapping 0.0042 cls_loss_causal 0.5391 re_mapping 0.0082 re_causal 0.0230 /// teacc 98.87 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0316, 0.1193, 0.0702, ..., -0.1608, -0.1052, 0.0240], + [ 0.1146, -0.0126, -0.0511, ..., 0.0675, 0.0943, -0.1096], + [-0.0656, -0.0486, -0.0478, ..., -0.0705, -0.0218, -0.0452], + ..., + [-0.0479, 0.1353, -0.1557, ..., -0.1160, 0.0056, -0.1250], + [ 0.0025, -0.0898, 0.0358, ..., 0.0558, -0.1045, -0.0387], + [-0.0559, -0.0829, 0.0543, ..., -0.0740, -0.0520, 0.1160]], + device='cuda:0'), grad: tensor([[ 3.4459e-07, -1.8813e-07, -8.9407e-08, ..., 3.8417e-07, + -4.7497e-08, -2.2398e-07], + [-4.1910e-08, 1.4259e-06, 1.3271e-07, ..., 8.1956e-07, + -8.1491e-08, 9.9652e-08], + [ 5.6345e-08, 5.4948e-08, 1.2061e-07, ..., 2.0210e-07, + -4.0978e-08, 1.8161e-08], + ..., + [ 3.6787e-08, -2.5071e-06, 5.3318e-07, ..., 3.0231e-06, + 1.3271e-07, 3.7346e-07], + [ 5.8068e-07, 1.7835e-07, 2.3562e-06, ..., 5.8953e-07, + 1.4203e-07, 2.3898e-06], + [ 2.2538e-07, 1.3951e-06, -2.0489e-06, ..., 4.2468e-06, + 7.1712e-08, -2.9393e-06]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0005, 0.0019, 0.0036, 0.0128, 0.0245, 0.0409, -0.0246, 0.0110, + -0.0329, -0.0233], device='cuda:0'), grad: tensor([ 7.1526e-07, 4.6678e-06, -2.4401e-07, 4.0233e-06, -1.3381e-05, + -8.2999e-06, -2.4941e-06, 2.4829e-06, 6.3553e-06, 6.1840e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 217.50, cls_loss 0.0024 cls_loss_mapping 0.0041 cls_loss_causal 0.5600 re_mapping 0.0076 re_causal 0.0233 /// teacc 98.90 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0316, 0.1197, 0.0706, ..., -0.1611, -0.1053, 0.0242], + [ 0.1149, -0.0126, -0.0513, ..., 0.0678, 0.0946, -0.1098], + [-0.0659, -0.0488, -0.0481, ..., -0.0707, -0.0220, -0.0452], + ..., + [-0.0483, 0.1358, -0.1558, ..., -0.1164, 0.0054, -0.1254], + [ 0.0025, -0.0903, 0.0357, ..., 0.0558, -0.1045, -0.0388], + [-0.0562, -0.0833, 0.0543, ..., -0.0743, -0.0521, 0.1162]], + device='cuda:0'), grad: tensor([[ 2.1718e-06, 3.2224e-06, -1.9325e-07, ..., 2.2119e-07, + 6.5193e-09, 4.6985e-07], + [-4.6194e-07, 2.4913e-07, 1.6261e-06, ..., -2.3842e-07, + -1.3970e-07, 1.6820e-06], + [ 1.1995e-06, 1.8869e-06, 2.2799e-06, ..., 2.7288e-07, + 1.5367e-08, 1.7649e-07], + ..., + [ 2.2687e-06, 3.5036e-06, 6.4932e-06, ..., 4.4284e-07, + 7.2177e-08, 1.6149e-06], + [ 1.2666e-06, 6.5705e-07, 3.0454e-06, ..., 4.1910e-07, + 2.8871e-08, 4.1574e-06], + [ 4.1537e-07, 7.1526e-07, -1.3433e-05, ..., 6.0583e-07, + 3.4925e-08, -1.8448e-05]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0007, 0.0020, 0.0033, 0.0132, 0.0246, 0.0408, -0.0243, 0.0108, + -0.0330, -0.0233], device='cuda:0'), grad: tensor([ 1.1347e-05, 4.8354e-06, 7.3388e-06, -2.3901e-05, 2.6077e-05, + 8.1677e-07, -1.7479e-05, 1.7807e-05, 1.0617e-05, -3.7491e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 217.21, cls_loss 0.0027 cls_loss_mapping 0.0049 cls_loss_causal 0.5621 re_mapping 0.0076 re_causal 0.0230 /// teacc 98.92 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0299, 0.1212, 0.0708, ..., -0.1613, -0.1053, 0.0242], + [ 0.1147, -0.0130, -0.0516, ..., 0.0678, 0.0950, -0.1102], + [-0.0662, -0.0489, -0.0482, ..., -0.0708, -0.0223, -0.0458], + ..., + [-0.0486, 0.1365, -0.1561, ..., -0.1176, 0.0052, -0.1261], + [ 0.0024, -0.0906, 0.0356, ..., 0.0557, -0.1047, -0.0389], + [-0.0564, -0.0843, 0.0546, ..., -0.0746, -0.0523, 0.1167]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, -2.1793e-06, -1.1623e-06, ..., 7.1712e-08, + -1.2107e-08, -1.4994e-06], + [-3.5437e-07, 1.6810e-07, 3.0641e-07, ..., -5.5647e-07, + -8.4285e-08, 3.7253e-08], + [ 5.8208e-08, 2.3469e-07, 1.3178e-06, ..., 2.1933e-07, + 2.2817e-08, 6.5658e-08], + ..., + [ 8.3819e-08, -4.4098e-07, 2.2277e-06, ..., 1.9884e-07, + 1.8626e-08, 7.7300e-07], + [-4.4238e-08, 1.6950e-07, 2.9001e-06, ..., -6.1328e-07, + 3.7253e-08, 2.3432e-06], + [ 4.7032e-08, 1.8133e-06, -1.0710e-06, ..., 1.8626e-07, + 2.9337e-08, -2.3693e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0018, 0.0015, 0.0033, 0.0137, 0.0247, 0.0403, -0.0246, 0.0105, + -0.0332, -0.0230], device='cuda:0'), grad: tensor([-4.0531e-06, 1.3933e-06, -1.0729e-05, -2.3812e-05, 5.5693e-06, + 6.8322e-06, 7.8380e-06, 7.9796e-06, 8.5235e-06, 5.4203e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 216.86, cls_loss 0.0031 cls_loss_mapping 0.0047 cls_loss_causal 0.5443 re_mapping 0.0077 re_causal 0.0223 /// teacc 98.92 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0300, 0.1235, 0.0722, ..., -0.1619, -0.1054, 0.0248], + [ 0.1148, -0.0140, -0.0517, ..., 0.0679, 0.0952, -0.1103], + [-0.0664, -0.0493, -0.0485, ..., -0.0709, -0.0225, -0.0460], + ..., + [-0.0486, 0.1378, -0.1566, ..., -0.1176, 0.0052, -0.1270], + [ 0.0028, -0.0920, 0.0355, ..., 0.0558, -0.1047, -0.0391], + [-0.0567, -0.0849, 0.0548, ..., -0.0742, -0.0525, 0.1172]], + device='cuda:0'), grad: tensor([[ 8.6101e-07, 2.4308e-07, 1.7881e-07, ..., 1.8124e-06, + 7.6834e-08, 2.6124e-07], + [-1.0945e-05, -7.5484e-07, -2.8759e-06, ..., 7.9691e-05, + -7.3016e-06, 2.3097e-07], + [ 3.2634e-06, 9.1270e-08, 9.9279e-07, ..., 3.3714e-06, + 3.1106e-06, 1.4156e-07], + ..., + [ 2.9448e-06, 2.6682e-07, 6.2305e-07, ..., 7.7188e-06, + 1.0980e-06, 8.4285e-08], + [ 1.8124e-06, 3.4180e-07, 6.6264e-07, ..., 9.5665e-06, + 4.4750e-07, 5.6531e-07], + [ 1.5050e-06, 1.3784e-07, 3.6694e-07, ..., 1.5843e-04, + 1.0338e-06, 7.2829e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0031, 0.0010, 0.0032, 0.0127, 0.0240, 0.0413, -0.0250, 0.0109, + -0.0334, -0.0228], device='cuda:0'), grad: tensor([ 4.5896e-06, 1.4436e-04, 6.9775e-06, 2.9244e-06, -4.9686e-04, + 6.8620e-06, 2.2929e-06, 1.5572e-05, 1.9208e-05, 2.9469e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 217.35, cls_loss 0.0026 cls_loss_mapping 0.0044 cls_loss_causal 0.5725 re_mapping 0.0075 re_causal 0.0234 /// teacc 98.93 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0309, 0.1238, 0.0723, ..., -0.1636, -0.1076, 0.0248], + [ 0.1150, -0.0144, -0.0519, ..., 0.0678, 0.0958, -0.1106], + [-0.0665, -0.0492, -0.0487, ..., -0.0711, -0.0227, -0.0460], + ..., + [-0.0489, 0.1384, -0.1568, ..., -0.1182, 0.0046, -0.1272], + [ 0.0026, -0.0927, 0.0354, ..., 0.0556, -0.1051, -0.0392], + [-0.0569, -0.0853, 0.0549, ..., -0.0746, -0.0526, 0.1174]], + device='cuda:0'), grad: tensor([[ 6.2771e-07, -6.5193e-08, 3.7346e-07, ..., 6.5658e-07, + 1.6578e-07, 2.8405e-07], + [ 1.7405e-05, 3.6368e-07, 9.1493e-06, ..., 1.0677e-05, + 3.5390e-08, 4.6715e-06], + [ 1.4501e-06, 2.1234e-07, 7.6741e-07, ..., 9.5274e-07, + 2.9802e-08, 3.1292e-07], + ..., + [ 4.9826e-07, -1.1651e-06, 9.3784e-07, ..., 4.7032e-07, + 8.8941e-08, 6.8778e-07], + [-3.0786e-05, 1.6904e-07, -7.6890e-06, ..., -2.0549e-05, + 1.3709e-06, -2.6692e-06], + [ 6.4541e-07, 4.1304e-07, -3.7365e-06, ..., 1.1474e-06, + 2.3004e-07, -3.5372e-06]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0022, 0.0006, 0.0033, 0.0124, 0.0250, 0.0420, -0.0251, 0.0110, + -0.0337, -0.0230], device='cuda:0'), grad: tensor([ 1.5134e-06, 2.4244e-05, 3.2000e-06, 6.0163e-06, 7.7020e-07, + -9.8050e-06, 7.3798e-06, 6.0210e-07, -2.9281e-05, -4.6194e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.01, cls_loss 0.0028 cls_loss_mapping 0.0040 cls_loss_causal 0.5482 re_mapping 0.0075 re_causal 0.0220 /// teacc 98.95 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0311, 0.1242, 0.0724, ..., -0.1644, -0.1081, 0.0248], + [ 0.1152, -0.0144, -0.0521, ..., 0.0676, 0.0961, -0.1108], + [-0.0668, -0.0489, -0.0486, ..., -0.0712, -0.0228, -0.0461], + ..., + [-0.0496, 0.1387, -0.1571, ..., -0.1186, 0.0043, -0.1276], + [ 0.0028, -0.0935, 0.0353, ..., 0.0558, -0.1053, -0.0394], + [-0.0572, -0.0855, 0.0551, ..., -0.0749, -0.0527, 0.1176]], + device='cuda:0'), grad: tensor([[ 1.2247e-07, 5.3970e-07, 4.2878e-06, ..., 1.5050e-06, + 7.9162e-09, 4.5449e-06], + [-4.0419e-06, -2.6077e-07, 1.2107e-07, ..., -3.3807e-06, + -9.1735e-08, 1.1269e-07], + [ 1.5367e-07, 2.4140e-06, 6.0536e-08, ..., 1.8720e-07, + 2.5146e-08, 6.0536e-08], + ..., + [ 2.7064e-06, -4.1962e-05, 8.0885e-07, ..., 2.5220e-06, + 1.9791e-07, 7.8510e-07], + [ 3.5577e-07, 2.2259e-06, 2.6554e-05, ..., 8.7768e-06, + -1.5600e-07, 2.8074e-05], + [ 7.3668e-07, 3.3557e-05, -3.5048e-05, ..., -9.2462e-06, + 3.4925e-08, -3.6210e-05]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0019, 0.0005, 0.0035, 0.0113, 0.0251, 0.0428, -0.0251, 0.0108, + -0.0338, -0.0229], device='cuda:0'), grad: tensor([ 1.0140e-05, -7.1898e-06, 3.0100e-06, 6.3293e-06, 4.2506e-06, + 6.2771e-07, 5.6904e-07, -6.0588e-05, 5.7846e-05, -1.4879e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 217.20, cls_loss 0.0023 cls_loss_mapping 0.0044 cls_loss_causal 0.5272 re_mapping 0.0073 re_causal 0.0221 /// teacc 98.98 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0317, 0.1247, 0.0726, ..., -0.1650, -0.1085, 0.0249], + [ 0.1160, -0.0144, -0.0522, ..., 0.0679, 0.0965, -0.1111], + [-0.0671, -0.0489, -0.0486, ..., -0.0713, -0.0230, -0.0461], + ..., + [-0.0501, 0.1389, -0.1574, ..., -0.1190, 0.0042, -0.1282], + [ 0.0010, -0.0964, 0.0350, ..., 0.0544, -0.1054, -0.0398], + [-0.0573, -0.0857, 0.0551, ..., -0.0762, -0.0527, 0.1178]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -5.3868e-06, -2.9989e-06, ..., 4.0280e-08, + -8.9873e-07, -4.2282e-06], + [-2.5798e-07, 2.2724e-07, 9.7789e-09, ..., -3.2317e-07, + -1.1642e-08, 4.4703e-08], + [ 1.2130e-07, 1.5763e-07, 6.9849e-08, ..., 2.2096e-07, + -5.3551e-09, 5.9605e-08], + ..., + [ 1.1059e-07, -4.2934e-07, 8.1258e-08, ..., 2.3679e-07, + 2.1188e-08, 1.4924e-07], + [-2.5379e-08, 1.3225e-07, 6.5425e-08, ..., 3.4925e-09, + 1.8161e-08, 2.0908e-07], + [ 2.5844e-08, 3.8147e-06, 2.1942e-06, ..., 1.7439e-07, + 7.2550e-07, 2.8461e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0019, 0.0006, 0.0034, 0.0114, 0.0262, 0.0435, -0.0250, 0.0108, + -0.0346, -0.0235], device='cuda:0'), grad: tensor([-1.3314e-05, 3.7951e-08, 6.7148e-07, 5.3225e-07, 4.8149e-07, + 1.2051e-06, 1.0692e-06, 1.4971e-07, 3.1316e-07, 8.8215e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 217.66, cls_loss 0.0025 cls_loss_mapping 0.0048 cls_loss_causal 0.5431 re_mapping 0.0076 re_causal 0.0212 /// teacc 98.96 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0319, 0.1250, 0.0728, ..., -0.1654, -0.1085, 0.0250], + [ 0.1162, -0.0146, -0.0524, ..., 0.0678, 0.0966, -0.1114], + [-0.0674, -0.0487, -0.0487, ..., -0.0715, -0.0230, -0.0459], + ..., + [-0.0503, 0.1392, -0.1577, ..., -0.1195, 0.0042, -0.1285], + [ 0.0010, -0.0967, 0.0350, ..., 0.0545, -0.1054, -0.0399], + [-0.0576, -0.0859, 0.0553, ..., -0.0767, -0.0528, 0.1181]], + device='cuda:0'), grad: tensor([[ 2.9267e-07, -1.7406e-06, -7.1852e-07, ..., 5.5972e-07, + -1.0813e-06, -1.1697e-06], + [ 6.2492e-07, -7.1945e-08, 4.4121e-07, ..., -8.2189e-08, + -7.2643e-07, -2.0745e-07], + [ 5.5693e-07, 1.2736e-07, 1.2107e-07, ..., 5.2992e-07, + 1.6112e-07, 8.8476e-08], + ..., + [ 1.3104e-06, -4.8243e-07, 2.0466e-07, ..., 1.0617e-06, + 4.5565e-07, 2.7148e-07], + [-2.2158e-05, 9.8487e-08, -3.3788e-06, ..., -1.1146e-05, + -6.5658e-07, 1.5125e-06], + [ 1.0235e-06, 1.2415e-06, 4.1467e-07, ..., 9.3970e-07, + 9.7509e-07, 6.3330e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0020, 0.0003, 0.0039, 0.0118, 0.0266, 0.0432, -0.0249, 0.0105, + -0.0346, -0.0236], device='cuda:0'), grad: tensor([-5.0776e-06, 4.4773e-07, 1.8235e-06, 2.1845e-05, 1.5693e-07, + 6.4671e-06, 9.1502e-07, 2.4382e-06, -3.5048e-05, 6.0722e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 217.32, cls_loss 0.0028 cls_loss_mapping 0.0047 cls_loss_causal 0.5532 re_mapping 0.0073 re_causal 0.0214 /// teacc 98.98 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0322, 0.1253, 0.0722, ..., -0.1662, -0.1087, 0.0242], + [ 0.1168, -0.0148, -0.0530, ..., 0.0681, 0.0968, -0.1117], + [-0.0675, -0.0488, -0.0487, ..., -0.0715, -0.0228, -0.0461], + ..., + [-0.0505, 0.1398, -0.1579, ..., -0.1200, 0.0040, -0.1294], + [ 0.0008, -0.0973, 0.0350, ..., 0.0542, -0.1055, -0.0401], + [-0.0582, -0.0861, 0.0556, ..., -0.0770, -0.0529, 0.1187]], + device='cuda:0'), grad: tensor([[ 5.5600e-07, 2.7940e-09, 2.4289e-06, ..., 4.8522e-07, + 4.1910e-09, 3.1223e-07], + [-6.9663e-06, -1.1008e-06, 2.4196e-06, ..., -7.6666e-06, + -5.1223e-09, 1.7360e-06], + [ 1.9744e-06, 1.6019e-07, 4.2245e-06, ..., 2.4550e-06, + 3.2829e-08, 6.1002e-07], + ..., + [ 3.1758e-06, -1.6764e-08, 1.3495e-06, ..., 2.4941e-06, + 3.6089e-08, 3.7719e-07], + [-1.4909e-05, 1.3062e-07, -1.5393e-05, ..., -4.3392e-05, + 5.9372e-08, -3.7521e-05], + [ 1.2189e-05, 5.9791e-07, 1.6123e-05, ..., 2.9460e-05, + 7.6834e-09, 2.3499e-05]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0013, 0.0002, 0.0044, 0.0115, 0.0267, 0.0434, -0.0245, 0.0103, + -0.0351, -0.0234], device='cuda:0'), grad: tensor([ 7.4171e-06, -1.5393e-05, 1.5527e-05, -3.0851e-04, 4.9770e-06, + 2.7394e-04, 6.8992e-06, 1.1832e-05, -5.8442e-05, 6.1333e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 217.54, cls_loss 0.0031 cls_loss_mapping 0.0066 cls_loss_causal 0.5615 re_mapping 0.0076 re_causal 0.0217 /// teacc 98.96 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0323, 0.1255, 0.0721, ..., -0.1674, -0.1088, 0.0241], + [ 0.1172, -0.0149, -0.0537, ..., 0.0683, 0.0969, -0.1125], + [-0.0669, -0.0490, -0.0491, ..., -0.0721, -0.0227, -0.0462], + ..., + [-0.0511, 0.1403, -0.1584, ..., -0.1206, 0.0037, -0.1304], + [ 0.0007, -0.0979, 0.0349, ..., 0.0545, -0.1056, -0.0403], + [-0.0589, -0.0864, 0.0560, ..., -0.0779, -0.0531, 0.1193]], + device='cuda:0'), grad: tensor([[ 6.1933e-08, -9.9763e-06, -7.6592e-06, ..., -2.4065e-06, + 3.7253e-09, -3.1590e-06], + [-1.5516e-06, 2.1383e-06, 1.7765e-07, ..., -1.8245e-06, + -1.6065e-08, 9.9186e-08], + [ 3.3039e-07, 4.9882e-06, 3.8464e-07, ..., 4.8568e-07, + -7.0548e-08, 1.6391e-07], + ..., + [ 6.2305e-07, -2.0042e-05, 2.6799e-07, ..., 8.3307e-07, + 1.8626e-08, 2.1770e-07], + [ 3.6135e-07, 7.5810e-07, 9.8348e-07, ..., 2.2631e-07, + 1.1013e-07, 1.0142e-06], + [ 3.4273e-07, 1.0572e-05, 9.1642e-07, ..., 2.1514e-06, + 6.0536e-09, -9.5926e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0012, -0.0004, 0.0049, 0.0105, 0.0272, 0.0433, -0.0230, 0.0104, + -0.0357, -0.0234], device='cuda:0'), grad: tensor([-2.9042e-05, 6.6962e-07, 9.6932e-06, 6.1542e-06, 4.3325e-06, + 2.5313e-06, 1.4752e-05, -3.7402e-05, 4.2208e-06, 2.4021e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 217.20, cls_loss 0.0028 cls_loss_mapping 0.0040 cls_loss_causal 0.5425 re_mapping 0.0075 re_causal 0.0216 /// teacc 98.87 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0327, 0.1261, 0.0721, ..., -0.1681, -0.1089, 0.0241], + [ 0.1211, -0.0151, -0.0516, ..., 0.0707, 0.0975, -0.1127], + [-0.0681, -0.0488, -0.0496, ..., -0.0725, -0.0233, -0.0462], + ..., + [-0.0516, 0.1408, -0.1587, ..., -0.1210, 0.0036, -0.1307], + [-0.0009, -0.0983, 0.0343, ..., 0.0537, -0.1057, -0.0404], + [-0.0594, -0.0867, 0.0561, ..., -0.0781, -0.0533, 0.1196]], + device='cuda:0'), grad: tensor([[-2.8443e-04, -3.0208e-04, -1.5981e-06, ..., 5.3924e-07, + 2.8498e-07, -2.7958e-06], + [ 1.0121e-04, 1.2082e-04, 8.7777e-08, ..., -1.3448e-05, + -1.0140e-05, 6.7987e-08], + [ 3.1590e-05, 2.9370e-05, 1.7858e-07, ..., -3.5256e-05, + -6.6102e-05, 2.5705e-07], + ..., + [ 8.0168e-06, 3.1628e-06, 3.1199e-07, ..., 5.5917e-06, + 4.5598e-06, 4.2538e-07], + [ 3.8054e-06, 3.7868e-06, 4.4610e-07, ..., 1.1679e-06, + 4.2585e-07, 9.9745e-07], + [ 2.2519e-06, 2.5295e-06, -3.5716e-07, ..., 1.1921e-06, + 9.2713e-07, -3.8720e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0012, 0.0015, 0.0052, 0.0105, 0.0271, 0.0433, -0.0236, 0.0103, + -0.0368, -0.0234], device='cuda:0'), grad: tensor([-9.6464e-04, 2.9707e-04, -1.8072e-04, 2.1577e-05, 2.2918e-05, + -2.6941e-05, 7.4673e-04, 5.4181e-05, 1.6510e-05, 1.3798e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 155---------------------------------------------------- +epoch 155, time 217.97, cls_loss 0.0032 cls_loss_mapping 0.0052 cls_loss_causal 0.5341 re_mapping 0.0076 re_causal 0.0215 /// teacc 99.00 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0322, 0.1270, 0.0718, ..., -0.1686, -0.1089, 0.0240], + [ 0.1222, -0.0141, -0.0519, ..., 0.0726, 0.0976, -0.1131], + [-0.0686, -0.0493, -0.0500, ..., -0.0728, -0.0232, -0.0464], + ..., + [-0.0535, 0.1407, -0.1591, ..., -0.1235, 0.0036, -0.1311], + [-0.0012, -0.0989, 0.0344, ..., 0.0534, -0.1058, -0.0406], + [-0.0604, -0.0874, 0.0566, ..., -0.0774, -0.0535, 0.1206]], + device='cuda:0'), grad: tensor([[ 1.2387e-07, 5.0943e-07, -2.1583e-07, ..., 3.8277e-07, + 4.0745e-08, -2.5891e-07], + [-4.6613e-07, 8.9593e-07, 1.3388e-07, ..., 6.1840e-07, + 2.6077e-08, 6.3796e-08], + [ 9.3831e-08, 2.2165e-07, -2.2096e-07, ..., -5.6671e-07, + 7.2177e-09, 2.7940e-08], + ..., + [ 2.0070e-07, 1.6633e-06, 3.8999e-07, ..., 4.6134e-05, + -1.0189e-06, -2.4587e-07], + [ 1.4552e-07, 1.8859e-07, 5.0664e-07, ..., 2.0433e-06, + 8.1491e-09, 1.6391e-07], + [ 7.8231e-08, 5.1521e-06, -4.7474e-07, ..., 1.4290e-05, + 2.0210e-07, -8.6799e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0014, 0.0020, 0.0049, 0.0100, 0.0258, 0.0439, -0.0232, 0.0099, + -0.0369, -0.0227], device='cuda:0'), grad: tensor([ 1.0589e-06, 3.5968e-06, -1.8729e-06, -2.4438e-06, -1.6296e-04, + 1.3679e-05, -3.1237e-06, 1.0687e-04, 5.7369e-06, 3.9458e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 217.45, cls_loss 0.0029 cls_loss_mapping 0.0053 cls_loss_causal 0.5657 re_mapping 0.0073 re_causal 0.0213 /// teacc 98.96 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0323, 0.1284, 0.0717, ..., -0.1687, -0.1090, 0.0235], + [ 0.1224, -0.0144, -0.0524, ..., 0.0727, 0.0979, -0.1136], + [-0.0698, -0.0507, -0.0505, ..., -0.0729, -0.0233, -0.0470], + ..., + [-0.0540, 0.1412, -0.1599, ..., -0.1238, 0.0034, -0.1319], + [-0.0010, -0.0992, 0.0342, ..., 0.0535, -0.1059, -0.0408], + [-0.0611, -0.0882, 0.0574, ..., -0.0777, -0.0537, 0.1216]], + device='cuda:0'), grad: tensor([[ 5.9530e-06, 1.5907e-06, 7.7263e-06, ..., 5.0552e-06, + 0.0000e+00, 2.5965e-06], + [-1.3828e-05, -2.2203e-06, -3.8883e-08, ..., -1.8731e-05, + 0.0000e+00, 4.3493e-07], + [-5.3830e-07, -1.0468e-05, -1.6242e-05, ..., 9.3058e-06, + 0.0000e+00, 1.8673e-07], + ..., + [ 8.0317e-06, 3.5553e-07, 1.5106e-06, ..., 7.9647e-06, + 0.0000e+00, 2.4354e-07], + [-7.0445e-06, 7.6815e-06, 1.9325e-08, ..., -1.0669e-05, + 0.0000e+00, -5.4799e-06], + [ 2.2110e-06, 1.0319e-06, 7.9721e-07, ..., 1.1064e-06, + 0.0000e+00, -4.6706e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0016, 0.0017, 0.0049, 0.0111, 0.0256, 0.0435, -0.0241, 0.0098, + -0.0370, -0.0222], device='cuda:0'), grad: tensor([ 2.1622e-05, -2.5481e-05, -1.1981e-04, 4.8168e-06, 5.5553e-07, + 1.4216e-05, 8.3260e-07, 1.5453e-05, 8.3029e-05, 4.8578e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 217.00, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5195 re_mapping 0.0072 re_causal 0.0210 /// teacc 98.96 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0325, 0.1293, 0.0725, ..., -0.1693, -0.1089, 0.0241], + [ 0.1233, -0.0145, -0.0531, ..., 0.0731, 0.0983, -0.1144], + [-0.0716, -0.0511, -0.0503, ..., -0.0736, -0.0237, -0.0473], + ..., + [-0.0544, 0.1418, -0.1602, ..., -0.1241, 0.0033, -0.1330], + [-0.0007, -0.0996, 0.0343, ..., 0.0538, -0.1059, -0.0408], + [-0.0616, -0.0885, 0.0573, ..., -0.0779, -0.0539, 0.1217]], + device='cuda:0'), grad: tensor([[ 2.9523e-07, -1.2824e-06, -7.4040e-07, ..., 4.3912e-07, + 9.3132e-10, -1.7937e-06], + [-2.3767e-06, 3.2503e-07, -1.0943e-08, ..., -2.1048e-06, + 4.4238e-09, -6.8219e-08], + [-3.1311e-06, -3.5129e-06, -2.9560e-06, ..., 6.1002e-07, + 6.9849e-10, 2.5379e-07], + ..., + [ 7.4320e-07, -4.9397e-06, 4.6007e-07, ..., 6.4448e-07, + 6.5193e-09, 5.5786e-07], + [ 5.0776e-06, 3.8408e-06, 2.7642e-06, ..., 4.0187e-07, + 4.0280e-08, 1.5479e-06], + [ 5.7369e-07, 4.2245e-06, -6.5472e-07, ..., 9.4622e-07, + 2.3283e-09, -3.7253e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0022, 0.0025, 0.0042, 0.0113, 0.0257, 0.0430, -0.0241, 0.0098, + -0.0370, -0.0224], device='cuda:0'), grad: tensor([-3.6284e-06, -2.1346e-06, -6.9797e-05, 8.6427e-06, 2.7828e-06, + -6.9737e-06, 1.9986e-06, -4.0494e-06, 6.6698e-05, 6.3963e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 217.31, cls_loss 0.0020 cls_loss_mapping 0.0042 cls_loss_causal 0.5469 re_mapping 0.0070 re_causal 0.0215 /// teacc 98.97 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0331, 0.1297, 0.0724, ..., -0.1705, -0.1089, 0.0242], + [ 0.1249, -0.0139, -0.0531, ..., 0.0743, 0.0991, -0.1144], + [-0.0734, -0.0519, -0.0503, ..., -0.0745, -0.0244, -0.0479], + ..., + [-0.0556, 0.1417, -0.1604, ..., -0.1254, 0.0032, -0.1340], + [-0.0006, -0.1000, 0.0343, ..., 0.0537, -0.1059, -0.0408], + [-0.0624, -0.0883, 0.0574, ..., -0.0782, -0.0540, 0.1220]], + device='cuda:0'), grad: tensor([[ 1.1874e-06, 2.7427e-07, 6.5845e-07, ..., 1.4799e-06, + 0.0000e+00, -1.0454e-07], + [ 1.9360e-04, 6.6340e-05, 2.1956e-07, ..., 2.1112e-04, + -5.5879e-09, 4.6333e-08], + [ 1.0572e-05, 3.6359e-06, -3.5893e-06, ..., 7.4245e-06, + 2.3283e-10, 3.4692e-08], + ..., + [ 2.3302e-06, -7.0501e-07, 2.5891e-07, ..., 3.1032e-06, + 2.0955e-09, 1.4226e-07], + [-2.0945e-04, -7.1585e-05, 2.7437e-06, ..., -2.2602e-04, + 2.3283e-10, 5.8860e-07], + [ 4.8010e-07, 9.2015e-07, -2.5565e-07, ..., 1.0375e-06, + 1.3970e-09, -9.3551e-07]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0021, 0.0033, 0.0039, 0.0111, 0.0259, 0.0429, -0.0239, 0.0092, + -0.0371, -0.0222], device='cuda:0'), grad: tensor([ 5.3942e-06, 5.2118e-04, -1.1958e-06, -1.4730e-05, 2.7893e-07, + 2.4661e-05, 5.9828e-06, 4.0904e-06, -5.4693e-04, 2.3022e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 217.56, cls_loss 0.0024 cls_loss_mapping 0.0045 cls_loss_causal 0.5345 re_mapping 0.0070 re_causal 0.0214 /// teacc 98.96 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0333, 0.1301, 0.0724, ..., -0.1712, -0.1089, 0.0242], + [ 0.1257, -0.0134, -0.0530, ..., 0.0745, 0.0997, -0.1145], + [-0.0737, -0.0521, -0.0503, ..., -0.0745, -0.0244, -0.0483], + ..., + [-0.0567, 0.1422, -0.1609, ..., -0.1258, 0.0024, -0.1355], + [-0.0006, -0.1004, 0.0341, ..., 0.0537, -0.1059, -0.0410], + [-0.0628, -0.0893, 0.0578, ..., -0.0783, -0.0543, 0.1226]], + device='cuda:0'), grad: tensor([[ 2.6310e-08, 3.9209e-07, -2.6985e-07, ..., 4.5169e-08, + 8.6147e-09, -2.7008e-07], + [-2.5844e-08, 3.3751e-06, 3.9348e-08, ..., -3.7486e-08, + -1.4668e-08, 6.8452e-08], + [ 2.0256e-08, 4.4145e-07, 1.6810e-07, ..., 1.4668e-08, + 1.3970e-09, 1.4203e-08], + ..., + [ 4.6566e-09, -3.6049e-04, 6.2399e-08, ..., 5.6811e-08, + 5.5879e-09, -5.6513e-06], + [ 6.4727e-08, 6.9197e-07, 4.0629e-07, ..., 4.8196e-08, + 5.3551e-09, 2.1956e-07], + [ 2.6077e-08, 3.5238e-04, -2.7381e-07, ..., 2.7241e-08, + 2.3283e-09, 4.9025e-06]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0021, 0.0036, 0.0040, 0.0112, 0.0257, 0.0427, -0.0240, 0.0090, + -0.0372, -0.0219], device='cuda:0'), grad: tensor([ 4.8336e-07, 6.0834e-06, 8.7451e-07, -7.9442e-07, 4.5449e-06, + 2.2147e-06, -2.0396e-07, -6.5851e-04, 2.1234e-06, 6.4373e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 217.33, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.5651 re_mapping 0.0072 re_causal 0.0216 /// teacc 98.89 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0337, 0.1302, 0.0723, ..., -0.1720, -0.1090, 0.0241], + [ 0.1260, -0.0140, -0.0534, ..., 0.0747, 0.1000, -0.1152], + [-0.0740, -0.0526, -0.0502, ..., -0.0749, -0.0245, -0.0485], + ..., + [-0.0567, 0.1434, -0.1617, ..., -0.1260, 0.0023, -0.1373], + [-0.0006, -0.1013, 0.0342, ..., 0.0539, -0.1060, -0.0409], + [-0.0632, -0.0895, 0.0579, ..., -0.0785, -0.0544, 0.1229]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, 1.1572e-07, -4.4471e-08, ..., 1.1688e-07, + 1.2340e-08, -1.1828e-07], + [-1.0571e-07, 3.5297e-06, 5.3132e-07, ..., 2.2608e-07, + 6.5193e-09, 5.6578e-08], + [ 1.8068e-07, 3.0234e-05, 3.6284e-06, ..., 3.1921e-07, + -1.1711e-07, 6.8452e-08], + ..., + [ 2.2235e-07, -9.6917e-05, -1.1854e-05, ..., 4.8522e-07, + 2.3050e-08, 5.8906e-08], + [-1.6410e-06, 6.2585e-07, -1.1660e-06, ..., -1.9427e-06, + 1.0012e-08, -3.9791e-07], + [ 2.0163e-07, 5.5730e-05, 6.8992e-06, ..., 2.1905e-06, + 9.5461e-09, -2.4005e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0019, 0.0033, 0.0042, 0.0110, 0.0259, 0.0426, -0.0241, 0.0091, + -0.0373, -0.0216], device='cuda:0'), grad: tensor([ 9.9000e-07, 9.6709e-06, 7.0751e-05, 6.2697e-06, 4.5560e-06, + 6.6943e-06, 2.4959e-06, -2.4676e-04, -2.2221e-06, 1.4734e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 217.28, cls_loss 0.0028 cls_loss_mapping 0.0051 cls_loss_causal 0.5337 re_mapping 0.0075 re_causal 0.0210 /// teacc 98.89 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0339, 0.1315, 0.0727, ..., -0.1729, -0.1094, 0.0242], + [ 0.1261, -0.0142, -0.0539, ..., 0.0745, 0.1000, -0.1159], + [-0.0740, -0.0539, -0.0506, ..., -0.0746, -0.0243, -0.0488], + ..., + [-0.0570, 0.1448, -0.1620, ..., -0.1263, 0.0022, -0.1382], + [-0.0010, -0.1011, 0.0342, ..., 0.0536, -0.1060, -0.0412], + [-0.0639, -0.0912, 0.0580, ..., -0.0792, -0.0546, 0.1232]], + device='cuda:0'), grad: tensor([[ 5.4808e-07, -1.8075e-05, -1.3702e-05, ..., 2.1495e-06, + 4.5169e-08, -2.0623e-05], + [ 2.4870e-05, 3.2574e-05, 1.8045e-05, ..., 3.2157e-05, + 1.1958e-05, 1.5318e-05], + [ 4.0643e-06, 5.9698e-07, 1.8142e-06, ..., 3.2276e-05, + 2.2873e-06, 8.9873e-07], + ..., + [-1.2949e-05, -4.3571e-05, 4.3982e-07, ..., -6.3032e-06, + -2.1949e-05, 3.2503e-07], + [-1.4581e-05, 6.3255e-06, -2.1324e-05, ..., -2.7180e-05, + 3.2932e-06, -1.2994e-05], + [ 1.5367e-06, 8.7321e-06, 6.3255e-06, ..., 1.5944e-06, + 7.0501e-07, 8.8066e-06]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0024, 0.0029, 0.0040, 0.0095, 0.0261, 0.0434, -0.0233, 0.0094, + -0.0372, -0.0221], device='cuda:0'), grad: tensor([-5.2303e-05, 1.4746e-04, 5.8472e-05, 1.0245e-05, -3.8564e-05, + 1.4633e-05, -1.2062e-05, -1.2517e-04, -3.3200e-05, 3.0667e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 217.42, cls_loss 0.0036 cls_loss_mapping 0.0061 cls_loss_causal 0.5676 re_mapping 0.0075 re_causal 0.0212 /// teacc 98.99 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0345, 0.1319, 0.0727, ..., -0.1739, -0.1094, 0.0244], + [ 0.1257, -0.0162, -0.0561, ..., 0.0750, 0.1004, -0.1164], + [-0.0743, -0.0536, -0.0508, ..., -0.0751, -0.0244, -0.0491], + ..., + [-0.0558, 0.1465, -0.1628, ..., -0.1266, 0.0023, -0.1379], + [-0.0011, -0.1018, 0.0342, ..., 0.0533, -0.1063, -0.0412], + [-0.0655, -0.0911, 0.0594, ..., -0.0812, -0.0550, 0.1238]], + device='cuda:0'), grad: tensor([[-5.8208e-09, -2.9076e-06, -3.8669e-06, ..., 1.8161e-08, + -2.3423e-07, -5.2564e-06], + [-1.0920e-07, 3.0082e-07, 1.3388e-07, ..., -8.4285e-08, + -1.0943e-08, 1.6578e-07], + [ 2.1420e-08, 2.0862e-07, 1.0384e-07, ..., 2.1188e-08, + -2.3283e-08, 1.2410e-07], + ..., + [ 4.8662e-08, -8.0885e-07, 3.7253e-07, ..., 8.5682e-08, + 3.7253e-08, 6.4075e-07], + [-2.1188e-08, 5.8673e-07, 8.4005e-07, ..., -8.1258e-08, + 2.3516e-08, 1.1018e-06], + [ 3.5623e-08, 1.8142e-06, 1.3411e-06, ..., 1.5227e-07, + 1.4156e-07, 1.3849e-06]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0024, 0.0013, 0.0043, 0.0125, 0.0264, 0.0410, -0.0230, 0.0100, + -0.0374, -0.0219], device='cuda:0'), grad: tensor([-1.4581e-05, 1.3094e-06, -1.6186e-06, 4.3400e-07, 7.3388e-07, + 8.5495e-07, 2.4308e-06, 8.4797e-07, 3.0920e-06, 6.4671e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 217.61, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.5235 re_mapping 0.0074 re_causal 0.0205 /// teacc 98.98 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0354, 0.1313, 0.0730, ..., -0.1746, -0.1095, 0.0242], + [ 0.1263, -0.0160, -0.0560, ..., 0.0757, 0.1006, -0.1165], + [-0.0745, -0.0540, -0.0508, ..., -0.0752, -0.0245, -0.0493], + ..., + [-0.0565, 0.1468, -0.1630, ..., -0.1273, 0.0023, -0.1380], + [-0.0010, -0.1022, 0.0341, ..., 0.0534, -0.1064, -0.0411], + [-0.0666, -0.0917, 0.0595, ..., -0.0814, -0.0556, 0.1243]], + device='cuda:0'), grad: tensor([[ 5.6578e-08, -2.2096e-07, -8.0168e-06, ..., 1.2573e-07, + 4.8196e-08, -1.0945e-05], + [ 2.0070e-07, 1.3039e-08, 1.1828e-07, ..., 1.4985e-06, + 8.2422e-07, 1.4622e-07], + [ 1.9167e-06, 5.9837e-08, 8.2888e-08, ..., 8.2999e-06, + 4.0755e-06, 1.2619e-07], + ..., + [ 1.6093e-06, -5.1502e-07, 2.1141e-07, ..., 7.0557e-06, + 3.2447e-06, 3.3597e-07], + [-4.9174e-06, 1.0710e-07, -4.1984e-06, ..., -1.9819e-05, + -1.0185e-05, -4.4107e-06], + [ 7.7533e-07, 5.0105e-07, 7.0222e-06, ..., 4.0978e-07, + 1.6065e-08, 8.1807e-06]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0019, 0.0016, 0.0044, 0.0124, 0.0261, 0.0409, -0.0221, 0.0101, + -0.0375, -0.0220], device='cuda:0'), grad: tensor([-2.1785e-05, 3.5278e-06, 1.5825e-05, 2.7101e-06, 2.2352e-06, + 1.7853e-06, 1.3232e-05, 1.3448e-05, -5.5045e-05, 2.4095e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 217.33, cls_loss 0.0027 cls_loss_mapping 0.0048 cls_loss_causal 0.5711 re_mapping 0.0067 re_causal 0.0206 /// teacc 99.00 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0350, 0.1322, 0.0733, ..., -0.1748, -0.1096, 0.0242], + [ 0.1268, -0.0159, -0.0561, ..., 0.0760, 0.1005, -0.1168], + [-0.0743, -0.0539, -0.0509, ..., -0.0751, -0.0250, -0.0497], + ..., + [-0.0570, 0.1469, -0.1639, ..., -0.1278, 0.0032, -0.1398], + [-0.0010, -0.1024, 0.0341, ..., 0.0538, -0.1067, -0.0412], + [-0.0678, -0.0921, 0.0598, ..., -0.0820, -0.0560, 0.1249]], + device='cuda:0'), grad: tensor([[ 2.2142e-07, -2.4755e-06, -6.5845e-07, ..., 3.7719e-07, + 1.6298e-08, -2.0247e-06], + [-3.4878e-07, 8.2655e-07, 2.8592e-07, ..., 4.3749e-07, + -1.9791e-07, 4.1095e-07], + [ 5.1223e-07, 3.2387e-07, 1.6508e-07, ..., 9.3784e-07, + 1.6321e-07, 2.6054e-07], + ..., + [ 1.8300e-07, -3.4701e-06, 5.5460e-07, ..., 6.8871e-07, + 4.5868e-08, 9.3644e-07], + [ 1.1325e-06, 3.2131e-07, 3.0193e-06, ..., 1.6233e-06, + 1.3062e-07, 3.8296e-06], + [ 6.6310e-07, 3.4980e-06, -2.7549e-06, ..., 3.8981e-05, + 2.0023e-08, 1.1269e-06]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0023, 0.0013, 0.0049, 0.0123, 0.0264, 0.0410, -0.0224, 0.0098, + -0.0376, -0.0221], device='cuda:0'), grad: tensor([-7.6741e-06, 2.7884e-06, 2.9001e-06, 8.3447e-06, -9.5487e-05, + -1.4305e-05, 4.2580e-06, -1.8328e-06, 1.0461e-05, 9.0361e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 165---------------------------------------------------- +epoch 165, time 218.57, cls_loss 0.0024 cls_loss_mapping 0.0040 cls_loss_causal 0.5314 re_mapping 0.0070 re_causal 0.0198 /// teacc 99.02 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0353, 0.1328, 0.0731, ..., -0.1757, -0.1096, 0.0238], + [ 0.1273, -0.0159, -0.0557, ..., 0.0767, 0.1006, -0.1167], + [-0.0745, -0.0547, -0.0510, ..., -0.0754, -0.0252, -0.0500], + ..., + [-0.0579, 0.1468, -0.1641, ..., -0.1296, 0.0035, -0.1403], + [-0.0008, -0.1005, 0.0339, ..., 0.0545, -0.1069, -0.0415], + [-0.0688, -0.0925, 0.0600, ..., -0.0827, -0.0562, 0.1254]], + device='cuda:0'), grad: tensor([[ 2.7940e-07, -2.1681e-06, -7.8045e-07, ..., 3.2061e-07, + 3.6322e-08, -4.4145e-06], + [-1.4305e-06, -1.0384e-07, 4.0978e-08, ..., -1.2247e-06, + -3.7812e-07, 1.0221e-07], + [ 1.5809e-07, 3.4459e-08, -2.1327e-07, ..., 2.2748e-07, + 6.3563e-08, 1.8394e-07], + ..., + [ 4.4145e-07, -1.1828e-07, 9.6392e-08, ..., 6.5239e-07, + 1.1688e-07, 2.0000e-07], + [ 7.6368e-08, 1.7090e-07, 3.8277e-07, ..., 2.7684e-07, + 2.6543e-08, 4.5076e-07], + [ 9.5926e-08, 1.6512e-06, 3.2154e-07, ..., 4.7614e-07, + 2.1188e-08, 2.8275e-06]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0021, 0.0015, 0.0048, 0.0118, 0.0267, 0.0415, -0.0225, 0.0091, + -0.0371, -0.0221], device='cuda:0'), grad: tensor([-8.4490e-06, -2.1700e-06, -4.4890e-07, 7.6834e-07, -2.1365e-06, + 6.3609e-07, 1.2415e-06, 1.3420e-06, 2.0899e-06, 7.1190e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 217.43, cls_loss 0.0025 cls_loss_mapping 0.0035 cls_loss_causal 0.5349 re_mapping 0.0071 re_causal 0.0197 /// teacc 98.97 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0357, 0.1335, 0.0734, ..., -0.1774, -0.1099, 0.0240], + [ 0.1275, -0.0163, -0.0560, ..., 0.0767, 0.1008, -0.1169], + [-0.0743, -0.0547, -0.0511, ..., -0.0757, -0.0248, -0.0502], + ..., + [-0.0581, 0.1475, -0.1645, ..., -0.1299, 0.0035, -0.1406], + [-0.0004, -0.1006, 0.0340, ..., 0.0551, -0.1072, -0.0416], + [-0.0692, -0.0931, 0.0602, ..., -0.0830, -0.0564, 0.1256]], + device='cuda:0'), grad: tensor([[ 5.7835e-07, 2.8173e-08, 1.7788e-06, ..., 8.5123e-07, + 3.7719e-08, 2.1718e-06], + [-1.6876e-06, 6.5845e-07, 1.1846e-06, ..., -1.6391e-07, + -6.8359e-07, 7.0268e-07], + [ 6.7279e-06, 5.0478e-06, 8.7395e-06, ..., 8.6725e-06, + 6.6869e-07, 1.0617e-07], + ..., + [ 1.1250e-06, -1.4696e-06, 2.3730e-06, ..., 1.5050e-06, + 2.5122e-07, 3.1646e-06], + [-6.4671e-05, -3.6627e-05, -6.4552e-05, ..., -1.1295e-04, + -6.5519e-07, 9.6262e-06], + [ 3.2387e-07, 8.3167e-07, -2.1085e-05, ..., 3.2131e-06, + 4.7032e-08, -3.9071e-05]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0022, 0.0011, 0.0059, 0.0118, 0.0267, 0.0413, -0.0227, 0.0092, + -0.0375, -0.0222], device='cuda:0'), grad: tensor([ 6.4224e-06, -1.2377e-06, 4.0740e-05, 3.4094e-05, -5.6997e-06, + 2.6965e-04, 5.0426e-05, 9.2760e-06, -3.3832e-04, -6.5327e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 217.87, cls_loss 0.0027 cls_loss_mapping 0.0046 cls_loss_causal 0.5459 re_mapping 0.0072 re_causal 0.0200 /// teacc 99.00 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0361, 0.1339, 0.0732, ..., -0.1791, -0.1100, 0.0239], + [ 0.1284, -0.0151, -0.0563, ..., 0.0767, 0.1030, -0.1174], + [-0.0745, -0.0530, -0.0512, ..., -0.0760, -0.0248, -0.0508], + ..., + [-0.0601, 0.1460, -0.1649, ..., -0.1305, 0.0011, -0.1412], + [ 0.0012, -0.1000, 0.0348, ..., 0.0574, -0.1074, -0.0409], + [-0.0696, -0.0936, 0.0605, ..., -0.0830, -0.0568, 0.1262]], + device='cuda:0'), grad: tensor([[ 1.4901e-07, 1.1129e-07, 2.9802e-08, ..., 3.2806e-07, + 1.4738e-07, 1.9325e-08], + [ 7.2410e-08, 3.3132e-07, 1.4016e-07, ..., 5.5227e-07, + 2.4214e-07, 4.3074e-08], + [ 1.3155e-07, -1.5842e-06, 1.4692e-07, ..., 1.7439e-07, + 2.5379e-08, 1.8394e-08], + ..., + [ 1.7518e-06, 1.6182e-07, 2.9104e-07, ..., 4.0978e-06, + 3.0454e-06, 6.2399e-08], + [ 5.4315e-06, 2.4051e-07, 8.6846e-08, ..., 1.2390e-05, + 9.3356e-06, -1.3132e-07], + [ 3.7905e-07, 9.1689e-07, -1.0338e-07, ..., 5.8003e-06, + 4.1211e-08, -2.1118e-07]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0018, 0.0022, 0.0074, 0.0114, 0.0261, 0.0410, -0.0234, 0.0070, + -0.0360, -0.0219], device='cuda:0'), grad: tensor([ 1.3430e-06, 2.9858e-06, -9.8273e-06, 1.7220e-06, -1.0192e-05, + -5.8234e-05, 1.2293e-06, 2.1070e-05, 3.8564e-05, 1.1407e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 168---------------------------------------------------- +epoch 168, time 217.92, cls_loss 0.0019 cls_loss_mapping 0.0041 cls_loss_causal 0.5386 re_mapping 0.0074 re_causal 0.0211 /// teacc 99.03 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0363, 0.1344, 0.0733, ..., -0.1801, -0.1100, 0.0241], + [ 0.1282, -0.0162, -0.0583, ..., 0.0766, 0.1029, -0.1193], + [-0.0741, -0.0530, -0.0503, ..., -0.0759, -0.0244, -0.0513], + ..., + [-0.0599, 0.1473, -0.1648, ..., -0.1304, 0.0011, -0.1415], + [ 0.0013, -0.1003, 0.0346, ..., 0.0573, -0.1075, -0.0411], + [-0.0704, -0.0940, 0.0608, ..., -0.0834, -0.0570, 0.1267]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, -2.3516e-07, -1.1874e-07, ..., 6.8219e-08, + 1.6298e-09, -4.2701e-07], + [-6.9104e-07, 8.2655e-08, 5.4017e-08, ..., -6.0722e-07, + -3.6554e-08, 1.1199e-07], + [ 3.7532e-07, 5.6345e-08, 2.6077e-08, ..., 6.6683e-07, + 1.8394e-08, 3.7719e-08], + ..., + [ 1.3504e-07, -1.3635e-06, 1.1316e-07, ..., 7.5530e-07, + 3.7253e-09, 2.3120e-07], + [-2.8461e-06, 4.6333e-08, -1.0608e-06, ..., -3.0342e-06, + 6.9849e-09, -2.1625e-06], + [ 1.9409e-06, 1.0459e-06, 7.1991e-07, ..., 1.7792e-05, + 1.3970e-09, 1.5534e-06]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0019, 0.0008, 0.0082, 0.0113, 0.0261, 0.0409, -0.0228, 0.0075, + -0.0362, -0.0218], device='cuda:0'), grad: tensor([-6.5472e-07, -4.5076e-07, 8.0978e-07, 3.4459e-07, -2.8580e-05, + 1.6969e-06, 1.3411e-06, 5.7276e-08, -7.1600e-06, 3.2574e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 216.63, cls_loss 0.0022 cls_loss_mapping 0.0031 cls_loss_causal 0.5380 re_mapping 0.0070 re_causal 0.0200 /// teacc 99.02 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0364, 0.1349, 0.0734, ..., -0.1808, -0.1101, 0.0240], + [ 0.1282, -0.0165, -0.0589, ..., 0.0765, 0.1026, -0.1202], + [-0.0744, -0.0534, -0.0504, ..., -0.0764, -0.0250, -0.0517], + ..., + [-0.0599, 0.1480, -0.1654, ..., -0.1305, 0.0021, -0.1420], + [ 0.0014, -0.1007, 0.0347, ..., 0.0577, -0.1077, -0.0410], + [-0.0708, -0.0942, 0.0612, ..., -0.0836, -0.0575, 0.1271]], + device='cuda:0'), grad: tensor([[ 1.0375e-06, -3.1060e-07, 1.7742e-07, ..., 2.9299e-06, + 3.2596e-09, -7.1712e-08], + [-5.7602e-07, 4.0233e-07, 2.9057e-07, ..., -3.9767e-07, + -2.3749e-08, 1.7276e-07], + [ 4.6706e-07, 2.1467e-07, 1.4659e-06, ..., 1.2135e-06, + -5.5879e-08, 1.3784e-07], + ..., + [ 9.7789e-08, -1.2666e-06, 1.0822e-06, ..., 3.6135e-07, + 1.2107e-08, 1.1269e-07], + [ 1.0803e-06, 9.4995e-08, 7.8753e-06, ..., 2.7865e-06, + 3.2131e-08, 7.7439e-07], + [ 1.1595e-07, 5.6298e-07, -2.1011e-05, ..., 4.1490e-07, + 4.6566e-09, -1.8448e-05]], device='cuda:0') +Epoch 171, bias, value: tensor([ 1.8011e-03, 7.5177e-05, 8.0893e-03, 1.1034e-02, 2.6050e-02, + 4.0835e-02, -2.2656e-02, 8.3115e-03, -3.6115e-02, -2.1732e-02], + device='cuda:0'), grad: tensor([ 4.4107e-06, 3.5483e-07, 2.5071e-06, -4.0866e-06, 1.0453e-05, + 1.2271e-05, -2.3261e-05, 1.2871e-06, 2.2933e-05, -2.6867e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 216.84, cls_loss 0.0023 cls_loss_mapping 0.0037 cls_loss_causal 0.5338 re_mapping 0.0066 re_causal 0.0197 /// teacc 98.96 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0366, 0.1358, 0.0737, ..., -0.1818, -0.1103, 0.0243], + [ 0.1303, -0.0137, -0.0591, ..., 0.0796, 0.1025, -0.1207], + [-0.0748, -0.0523, -0.0506, ..., -0.0770, -0.0254, -0.0525], + ..., + [-0.0624, 0.1454, -0.1656, ..., -0.1343, 0.0025, -0.1422], + [ 0.0016, -0.1011, 0.0346, ..., 0.0581, -0.1077, -0.0410], + [-0.0714, -0.0947, 0.0614, ..., -0.0840, -0.0577, 0.1274]], + device='cuda:0'), grad: tensor([[-1.1399e-06, -2.2501e-05, -1.1101e-05, ..., -3.4180e-06, + 9.3132e-10, -1.9491e-05], + [-4.7684e-07, -4.6566e-09, 4.5169e-08, ..., -4.8894e-07, + -1.3970e-09, 6.1002e-08], + [ 5.5879e-08, -4.3493e-07, 1.0664e-07, ..., -6.6124e-08, + -9.3132e-10, 6.7521e-08], + ..., + [ 3.6322e-07, 1.2573e-08, 1.3597e-07, ..., 4.9639e-07, + 1.3970e-09, 9.8255e-08], + [-3.4451e-05, 6.5425e-07, 5.9046e-07, ..., -6.5029e-05, + 1.2573e-08, -1.0967e-05], + [ 8.1491e-08, 4.2887e-07, 1.1316e-07, ..., 1.4901e-07, + 2.3283e-09, 1.6904e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0020, 0.0017, 0.0081, 0.0113, 0.0262, 0.0402, -0.0218, 0.0068, + -0.0361, -0.0219], device='cuda:0'), grad: tensor([-4.7833e-05, -2.7614e-07, -5.1633e-06, 1.5469e-06, 1.2433e-07, + 5.3644e-05, 8.9228e-05, 2.4866e-06, -9.4652e-05, 9.4390e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 217.17, cls_loss 0.0017 cls_loss_mapping 0.0036 cls_loss_causal 0.5369 re_mapping 0.0068 re_causal 0.0200 /// teacc 98.99 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0368, 0.1365, 0.0742, ..., -0.1826, -0.1103, 0.0246], + [ 0.1301, -0.0141, -0.0595, ..., 0.0796, 0.1028, -0.1212], + [-0.0752, -0.0526, -0.0506, ..., -0.0775, -0.0257, -0.0531], + ..., + [-0.0625, 0.1459, -0.1661, ..., -0.1345, 0.0027, -0.1436], + [ 0.0018, -0.1014, 0.0347, ..., 0.0584, -0.1079, -0.0411], + [-0.0718, -0.0946, 0.0615, ..., -0.0842, -0.0582, 0.1277]], + device='cuda:0'), grad: tensor([[ 1.6112e-07, -1.2685e-06, -1.0468e-06, ..., 1.5134e-07, + 1.8161e-08, -1.7826e-06], + [-1.1884e-06, 1.8906e-07, -3.6741e-07, ..., -6.1095e-07, + -1.2191e-06, 3.3062e-08], + [ 8.6101e-07, 2.5053e-07, 3.2596e-07, ..., 7.6927e-07, + 8.9267e-07, 3.0734e-08], + ..., + [ 2.2957e-07, -5.6392e-07, 6.4727e-08, ..., 4.5821e-07, + 2.0675e-07, 4.5914e-07], + [ 9.4995e-08, 3.4273e-07, 4.5542e-07, ..., 1.7881e-07, + 1.9092e-08, 6.7381e-07], + [ 1.9418e-07, 1.0412e-06, 5.7695e-07, ..., 3.9907e-07, + 2.0768e-07, 8.8522e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0023, 0.0014, 0.0082, 0.0116, 0.0263, 0.0398, -0.0218, 0.0069, + -0.0360, -0.0219], device='cuda:0'), grad: tensor([-3.8035e-06, -1.5264e-06, 2.4624e-06, -4.9025e-06, -2.2016e-06, + 3.9451e-06, 2.6589e-07, 5.0524e-07, 1.6503e-06, 3.5893e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 217.46, cls_loss 0.0020 cls_loss_mapping 0.0035 cls_loss_causal 0.5576 re_mapping 0.0071 re_causal 0.0206 /// teacc 98.97 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0373, 0.1370, 0.0753, ..., -0.1841, -0.1095, 0.0251], + [ 0.1303, -0.0145, -0.0600, ..., 0.0797, 0.1035, -0.1225], + [-0.0758, -0.0531, -0.0507, ..., -0.0779, -0.0266, -0.0537], + ..., + [-0.0623, 0.1468, -0.1664, ..., -0.1345, 0.0027, -0.1441], + [ 0.0017, -0.1020, 0.0348, ..., 0.0584, -0.1080, -0.0411], + [-0.0724, -0.0951, 0.0612, ..., -0.0847, -0.0601, 0.1276]], + device='cuda:0'), grad: tensor([[ 1.2713e-07, -1.9409e-06, -1.3104e-06, ..., 5.2433e-07, + 2.2817e-08, -1.6671e-06], + [ 1.3076e-05, 5.1688e-08, 7.4971e-08, ..., 2.9683e-05, + 2.9244e-06, 8.7544e-08], + [ 6.9337e-07, -1.2480e-07, 2.6822e-07, ..., 2.0731e-06, + 9.3132e-08, 2.8312e-07], + ..., + [ 9.4483e-07, -1.8533e-07, 8.5216e-08, ..., 2.0619e-06, + 1.2247e-07, 1.2852e-07], + [-1.6540e-05, 2.0256e-07, -9.3132e-08, ..., -3.4750e-05, + -3.4645e-06, 4.3726e-07], + [ 1.4296e-07, 9.3179e-07, 1.0943e-07, ..., 9.0385e-07, + 1.8626e-08, -9.7975e-07]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0029, 0.0014, 0.0078, 0.0117, 0.0264, 0.0399, -0.0218, 0.0072, + -0.0360, -0.0223], device='cuda:0'), grad: tensor([-4.6454e-06, 5.5760e-05, 2.8703e-06, 1.0934e-06, -1.3642e-05, + 1.8273e-06, 1.1794e-05, 5.0329e-06, -6.2823e-05, 2.7288e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 173---------------------------------------------------- +epoch 173, time 217.45, cls_loss 0.0023 cls_loss_mapping 0.0031 cls_loss_causal 0.5156 re_mapping 0.0068 re_causal 0.0190 /// teacc 99.04 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0378, 0.1381, 0.0757, ..., -0.1859, -0.1094, 0.0253], + [ 0.1305, -0.0146, -0.0618, ..., 0.0795, 0.1039, -0.1266], + [-0.0762, -0.0537, -0.0508, ..., -0.0783, -0.0271, -0.0543], + ..., + [-0.0627, 0.1471, -0.1669, ..., -0.1349, 0.0028, -0.1456], + [ 0.0027, -0.1012, 0.0348, ..., 0.0595, -0.1082, -0.0406], + [-0.0720, -0.0951, 0.0617, ..., -0.0842, -0.0607, 0.1295]], + device='cuda:0'), grad: tensor([[ 2.4028e-06, 1.5367e-08, -8.3819e-09, ..., 9.3356e-06, + 0.0000e+00, -1.2573e-08], + [-2.1271e-06, -1.2042e-06, 2.7008e-08, ..., -5.1521e-06, + 0.0000e+00, 3.3993e-08], + [ 2.4196e-06, 1.1735e-06, 1.0477e-07, ..., 6.0759e-06, + 0.0000e+00, 9.4529e-08], + ..., + [ 6.3796e-07, -1.0803e-06, 3.0734e-08, ..., 1.2126e-06, + 0.0000e+00, 4.2841e-08], + [ 3.6918e-06, 2.2585e-07, -1.8952e-07, ..., 5.4240e-06, + 0.0000e+00, -4.5355e-07], + [ 2.5099e-07, 4.9500e-07, 1.5367e-08, ..., 9.5274e-07, + 0.0000e+00, 1.6764e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0032, 0.0009, 0.0075, 0.0125, 0.0259, 0.0391, -0.0220, 0.0067, + -0.0355, -0.0213], device='cuda:0'), grad: tensor([ 1.6049e-05, -1.3143e-05, 1.5259e-05, 1.8887e-06, 3.3736e-05, + 3.8818e-06, -7.4983e-05, 1.1083e-06, 1.3642e-05, 2.5425e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 217.55, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5165 re_mapping 0.0064 re_causal 0.0184 /// teacc 98.99 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0394, 0.1376, 0.0758, ..., -0.1889, -0.1097, 0.0252], + [ 0.1317, -0.0149, -0.0622, ..., 0.0806, 0.1043, -0.1272], + [-0.0767, -0.0542, -0.0509, ..., -0.0787, -0.0275, -0.0547], + ..., + [-0.0629, 0.1485, -0.1671, ..., -0.1351, 0.0027, -0.1466], + [ 0.0022, -0.1024, 0.0344, ..., 0.0589, -0.1083, -0.0411], + [-0.0714, -0.0960, 0.0623, ..., -0.0845, -0.0609, 0.1305]], + device='cuda:0'), grad: tensor([[ 7.8697e-08, -1.6578e-07, 7.8082e-06, ..., 4.0494e-06, + 0.0000e+00, 8.9109e-06], + [-1.6997e-07, 8.0094e-08, 1.1828e-07, ..., -9.0804e-08, + -6.0536e-09, 1.3225e-07], + [ 3.6787e-08, 8.4750e-08, 7.3109e-08, ..., 8.8010e-08, + 9.3132e-10, 7.8231e-08], + ..., + [ 1.4063e-07, -3.6694e-07, 5.3085e-08, ..., 2.2212e-07, + 2.3283e-09, 5.8673e-08], + [ 1.1548e-07, 7.4971e-08, 1.4508e-04, ..., 7.0274e-05, + 1.3970e-09, 1.6499e-04], + [ 4.5169e-08, 3.1106e-07, -1.5461e-04, ..., -7.4089e-05, + 9.3132e-10, -1.7560e-04]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0026, 0.0012, 0.0072, 0.0122, 0.0259, 0.0395, -0.0221, 0.0072, + -0.0362, -0.0211], device='cuda:0'), grad: tensor([ 1.8090e-05, 4.9360e-08, 3.4226e-07, 1.6317e-06, 7.2783e-07, + -1.5199e-06, 2.9337e-08, 5.1688e-08, 3.3307e-04, -3.5262e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 217.39, cls_loss 0.0028 cls_loss_mapping 0.0042 cls_loss_causal 0.5232 re_mapping 0.0068 re_causal 0.0199 /// teacc 99.02 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0398, 0.1381, 0.0759, ..., -0.1898, -0.1097, 0.0253], + [ 0.1319, -0.0149, -0.0636, ..., 0.0805, 0.1034, -0.1280], + [-0.0765, -0.0540, -0.0497, ..., -0.0784, -0.0266, -0.0552], + ..., + [-0.0631, 0.1488, -0.1677, ..., -0.1353, 0.0034, -0.1476], + [ 0.0020, -0.1032, 0.0339, ..., 0.0588, -0.1085, -0.0418], + [-0.0722, -0.0967, 0.0627, ..., -0.0850, -0.0613, 0.1312]], + device='cuda:0'), grad: tensor([[ 1.7369e-07, -1.2619e-07, 2.0117e-07, ..., 1.2992e-07, + 1.3970e-09, -8.0094e-08], + [-4.5076e-07, -1.6298e-08, 2.4308e-07, ..., -6.5938e-07, + -3.9116e-08, -9.7789e-09], + [ 2.2305e-07, 4.6100e-08, 2.0349e-07, ..., 1.0943e-07, + 6.0536e-09, 1.4435e-08], + ..., + [ 2.2165e-07, -3.1991e-07, 3.4133e-07, ..., 1.6158e-07, + 6.9849e-09, 9.7789e-09], + [ 1.2871e-06, 4.8429e-08, 2.1290e-06, ..., 9.4343e-07, + 6.9849e-09, 1.7835e-07], + [ 1.8813e-07, 2.2491e-07, 1.8021e-07, ..., 2.0256e-07, + 7.9162e-09, -1.1642e-08]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0026, 0.0001, 0.0085, 0.0091, 0.0257, 0.0430, -0.0231, 0.0075, + -0.0369, -0.0210], device='cuda:0'), grad: tensor([ 6.1421e-07, -3.7625e-07, 8.0373e-07, -1.0860e-04, 2.1001e-07, + 9.8705e-05, -1.3988e-06, 7.6694e-07, 8.0466e-06, 1.2396e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 217.16, cls_loss 0.0028 cls_loss_mapping 0.0041 cls_loss_causal 0.5138 re_mapping 0.0065 re_causal 0.0188 /// teacc 98.97 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0426, 0.1362, 0.0754, ..., -0.1922, -0.1132, 0.0244], + [ 0.1329, -0.0151, -0.0627, ..., 0.0818, 0.1041, -0.1274], + [-0.0775, -0.0529, -0.0497, ..., -0.0798, -0.0275, -0.0557], + ..., + [-0.0632, 0.1491, -0.1679, ..., -0.1359, 0.0036, -0.1479], + [ 0.0021, -0.1034, 0.0338, ..., 0.0590, -0.1086, -0.0420], + [-0.0748, -0.0974, 0.0629, ..., -0.0859, -0.0618, 0.1317]], + device='cuda:0'), grad: tensor([[-2.0582e-07, -1.3746e-06, -9.0245e-07, ..., -6.8918e-07, + -4.1956e-07, -3.8520e-06], + [ 4.7032e-08, 1.9558e-08, 5.2620e-08, ..., 2.6263e-07, + 2.5611e-08, 1.1362e-07], + [ 2.7474e-08, 3.3062e-08, 3.9581e-08, ..., 1.6978e-06, + 1.2573e-08, 4.9360e-08], + ..., + [ 2.9337e-08, -4.7963e-08, 4.8429e-08, ..., 7.9628e-07, + 7.9162e-09, 1.1595e-07], + [ 9.7416e-07, 1.9558e-08, 5.4343e-07, ..., -2.0210e-07, + 1.0245e-07, 1.7844e-06], + [ 7.5437e-08, 4.0047e-08, -6.3330e-08, ..., 1.5991e-06, + 2.9802e-08, -2.5565e-07]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0001, 0.0004, 0.0092, 0.0083, 0.0266, 0.0438, -0.0226, 0.0075, + -0.0373, -0.0214], device='cuda:0'), grad: tensor([-2.7508e-05, 8.1724e-07, 9.4026e-06, 1.3094e-06, -4.9807e-06, + -4.3996e-06, 2.7567e-05, 1.8477e-06, -7.2196e-06, 3.1553e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 217.13, cls_loss 0.0023 cls_loss_mapping 0.0026 cls_loss_causal 0.5396 re_mapping 0.0065 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0429, 0.1379, 0.0743, ..., -0.1924, -0.1131, 0.0227], + [ 0.1334, -0.0153, -0.0630, ..., 0.0820, 0.1044, -0.1278], + [-0.0780, -0.0508, -0.0496, ..., -0.0802, -0.0273, -0.0550], + ..., + [-0.0634, 0.1477, -0.1682, ..., -0.1363, 0.0035, -0.1485], + [ 0.0020, -0.1040, 0.0336, ..., 0.0589, -0.1088, -0.0423], + [-0.0755, -0.0979, 0.0635, ..., -0.0867, -0.0623, 0.1330]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, -3.5577e-07, -1.2619e-07, ..., 3.9116e-08, + 4.6566e-09, -1.2200e-07], + [-8.7079e-08, 1.5367e-08, 3.9581e-08, ..., -1.0058e-07, + 5.1223e-09, 8.0094e-08], + [ 2.6543e-08, 8.1956e-08, 4.0513e-08, ..., 3.6787e-08, + 9.3132e-10, 4.8429e-08], + ..., + [ 2.1420e-08, -6.5193e-08, 1.1129e-07, ..., 3.1199e-08, + 3.2596e-09, 1.9092e-07], + [ 5.7742e-08, 2.9802e-08, 7.3202e-07, ..., 1.0384e-07, + 2.8871e-08, 1.2880e-06], + [ 3.8650e-08, 9.5461e-08, -1.3756e-06, ..., 6.8918e-08, + 8.3819e-09, -2.3767e-06]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0005, 0.0003, 0.0110, 0.0087, 0.0272, 0.0434, -0.0230, 0.0063, + -0.0377, -0.0211], device='cuda:0'), grad: tensor([-7.9675e-07, -4.6566e-10, 9.7323e-08, 5.4343e-07, 1.5870e-06, + -1.1194e-06, 8.9919e-07, 2.8918e-07, 2.0564e-06, -3.5651e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.81, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.4831 re_mapping 0.0066 re_causal 0.0186 /// teacc 98.81 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0429, 0.1385, 0.0746, ..., -0.1927, -0.1131, 0.0228], + [ 0.1352, -0.0147, -0.0633, ..., 0.0841, 0.1045, -0.1281], + [-0.0787, -0.0507, -0.0497, ..., -0.0806, -0.0275, -0.0551], + ..., + [-0.0643, 0.1480, -0.1685, ..., -0.1377, 0.0042, -0.1486], + [ 0.0015, -0.1051, 0.0334, ..., 0.0586, -0.1094, -0.0425], + [-0.0777, -0.0990, 0.0639, ..., -0.0877, -0.0624, 0.1335]], + device='cuda:0'), grad: tensor([[ 1.3225e-07, -1.8813e-07, -6.7521e-08, ..., 1.1083e-07, + 2.3283e-09, -4.5169e-08], + [ 1.7714e-06, 4.9919e-06, 3.5763e-07, ..., 1.7509e-07, + 3.0734e-08, 4.4005e-07], + [ 4.4703e-08, 2.3562e-07, -6.0536e-09, ..., 7.5903e-08, + 5.1223e-09, 2.3749e-08], + ..., + [-3.4124e-06, -1.2845e-05, -1.9139e-07, ..., 1.6158e-07, + -1.2573e-07, 7.1712e-08], + [ 3.5074e-06, 1.1269e-07, 1.8738e-06, ..., 1.5516e-06, + 1.0431e-07, 3.2485e-06], + [ 2.2221e-06, 6.9663e-06, 4.0513e-07, ..., 6.5938e-07, + 1.2200e-07, 3.3667e-07]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0003, 0.0008, 0.0111, 0.0089, 0.0271, 0.0433, -0.0230, 0.0065, + -0.0384, -0.0212], device='cuda:0'), grad: tensor([-3.7858e-07, 8.5309e-06, -7.6788e-07, 8.0233e-07, -1.9977e-07, + -9.6858e-06, 3.1665e-06, -1.9491e-05, 5.5023e-06, 1.2480e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 216.95, cls_loss 0.0021 cls_loss_mapping 0.0029 cls_loss_causal 0.5251 re_mapping 0.0068 re_causal 0.0198 /// teacc 98.94 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0431, 0.1391, 0.0740, ..., -0.1933, -0.1131, 0.0222], + [ 0.1357, -0.0151, -0.0630, ..., 0.0847, 0.1049, -0.1285], + [-0.0780, -0.0510, -0.0499, ..., -0.0812, -0.0276, -0.0557], + ..., + [-0.0645, 0.1488, -0.1689, ..., -0.1380, 0.0044, -0.1490], + [ 0.0015, -0.1054, 0.0333, ..., 0.0588, -0.1094, -0.0427], + [-0.0782, -0.0994, 0.0643, ..., -0.0882, -0.0631, 0.1343]], + device='cuda:0'), grad: tensor([[ 5.1223e-08, -8.8885e-06, -8.8811e-06, ..., -2.0005e-06, + 4.3306e-08, -1.0006e-05], + [-4.8190e-05, 4.0531e-06, 2.6636e-07, ..., -1.5438e-05, + -9.1970e-05, 3.2270e-07], + [ 2.4483e-05, 1.2936e-06, 9.9000e-07, ..., 8.2329e-06, + 4.6700e-05, 1.0943e-06], + ..., + [ 2.2918e-05, -9.6709e-06, 1.1269e-07, ..., 7.4431e-06, + 4.2677e-05, 1.1455e-07], + [-2.5928e-06, 3.3006e-06, -2.5500e-06, ..., -4.0904e-06, + 6.7893e-07, -3.5260e-06], + [ 2.4661e-06, 4.5337e-06, 4.8727e-06, ..., 3.9861e-06, + 9.3179e-07, 6.1542e-06]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0005, 0.0006, 0.0114, 0.0090, 0.0271, 0.0430, -0.0237, 0.0069, + -0.0385, -0.0211], device='cuda:0'), grad: tensor([-2.8849e-05, -1.5187e-04, 8.5890e-05, 2.1942e-06, 1.0571e-06, + 1.8245e-06, 1.4797e-05, 5.4926e-05, -3.2000e-06, 2.3350e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 217.46, cls_loss 0.0017 cls_loss_mapping 0.0041 cls_loss_causal 0.5551 re_mapping 0.0065 re_causal 0.0190 /// teacc 99.01 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0433, 0.1395, 0.0743, ..., -0.1938, -0.1132, 0.0224], + [ 0.1365, -0.0151, -0.0642, ..., 0.0851, 0.1056, -0.1305], + [-0.0794, -0.0511, -0.0502, ..., -0.0823, -0.0282, -0.0570], + ..., + [-0.0646, 0.1493, -0.1692, ..., -0.1382, 0.0041, -0.1481], + [ 0.0016, -0.1056, 0.0333, ..., 0.0590, -0.1095, -0.0427], + [-0.0793, -0.1004, 0.0651, ..., -0.0881, -0.0632, 0.1355]], + device='cuda:0'), grad: tensor([[ 3.0082e-07, 1.8254e-07, 1.2480e-07, ..., 1.0105e-07, + 0.0000e+00, 5.5274e-07], + [ 4.7730e-07, 2.0117e-07, 3.2037e-07, ..., 4.8941e-07, + -1.8626e-09, 1.4529e-07], + [ 3.6396e-06, 7.8790e-07, 1.4342e-06, ..., 3.7104e-06, + -4.6566e-10, 2.3143e-07], + ..., + [ 1.1735e-06, 9.4203e-07, 1.5646e-06, ..., 3.4552e-07, + 1.3970e-09, 1.1660e-06], + [-1.1586e-05, 3.4738e-07, -1.8962e-06, ..., -1.4767e-05, + 0.0000e+00, 2.5984e-06], + [ 1.7229e-07, 9.6858e-08, -4.2319e-06, ..., 3.0966e-07, + 0.0000e+00, -5.9828e-06]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0003, 0.0005, 0.0108, 0.0092, 0.0269, 0.0429, -0.0238, 0.0072, + -0.0384, -0.0207], device='cuda:0'), grad: tensor([ 1.8682e-06, 2.3823e-06, 1.3389e-05, 3.4422e-06, 2.5239e-06, + 9.9391e-06, 3.0287e-06, 7.7486e-06, -3.4869e-05, -9.4101e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 217.56, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5393 re_mapping 0.0061 re_causal 0.0190 /// teacc 99.00 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0436, 0.1398, 0.0748, ..., -0.1942, -0.1131, 0.0228], + [ 0.1375, -0.0147, -0.0647, ..., 0.0856, 0.1060, -0.1313], + [-0.0798, -0.0513, -0.0505, ..., -0.0827, -0.0284, -0.0577], + ..., + [-0.0658, 0.1493, -0.1695, ..., -0.1389, 0.0039, -0.1490], + [ 0.0020, -0.1059, 0.0334, ..., 0.0593, -0.1096, -0.0427], + [-0.0800, -0.1012, 0.0655, ..., -0.0887, -0.0634, 0.1359]], + device='cuda:0'), grad: tensor([[ 1.3411e-07, -7.6881e-07, -8.3959e-07, ..., 8.9873e-08, + -7.0315e-08, -1.1865e-06], + [-2.7893e-07, 6.7912e-06, -1.0952e-06, ..., -2.5518e-06, + -4.1910e-09, -7.5065e-07], + [ 3.0315e-07, 5.4669e-07, 9.4064e-08, ..., 1.9465e-07, + 1.3970e-09, 5.3085e-08], + ..., + [-1.6652e-06, -7.7188e-05, -4.4070e-06, ..., 1.6997e-07, + 2.3283e-09, 5.4017e-08], + [ 7.3807e-07, 1.5106e-06, 5.1735e-07, ..., 1.1362e-07, + 9.3132e-10, 1.1269e-07], + [ 1.7509e-06, 6.7592e-05, 6.2920e-06, ..., 2.1830e-06, + 4.2375e-08, 1.4547e-06]], device='cuda:0') +Epoch 183, bias, value: tensor([-2.5532e-05, 7.0022e-04, 1.0523e-02, 9.2398e-03, 2.7203e-02, + 4.3542e-02, -2.5544e-02, 6.8184e-03, -3.8187e-02, -2.0872e-02], + device='cuda:0'), grad: tensor([-2.3451e-06, 7.9349e-06, 1.4678e-06, -9.8161e-07, 1.4585e-06, + 1.2666e-07, -6.2725e-07, -1.2887e-04, 3.7923e-06, 1.1808e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 217.45, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5278 re_mapping 0.0064 re_causal 0.0196 /// teacc 98.98 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0442, 0.1403, 0.0751, ..., -0.1951, -0.1133, 0.0229], + [ 0.1385, -0.0144, -0.0646, ..., 0.0861, 0.1062, -0.1316], + [-0.0802, -0.0515, -0.0508, ..., -0.0829, -0.0286, -0.0583], + ..., + [-0.0660, 0.1504, -0.1687, ..., -0.1392, 0.0038, -0.1489], + [ 0.0017, -0.1073, 0.0334, ..., 0.0593, -0.1097, -0.0428], + [-0.0821, -0.1035, 0.0653, ..., -0.0893, -0.0635, 0.1358]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, -1.3504e-08, 2.9337e-08, ..., 2.1886e-08, + 4.6566e-10, 2.7474e-08], + [-3.7253e-08, -4.6566e-10, 1.0524e-07, ..., -1.8161e-08, + -7.4506e-09, 1.2200e-07], + [ 2.3283e-08, 4.6566e-09, 4.4238e-08, ..., 2.0489e-08, + 3.2596e-09, 1.2107e-08], + ..., + [ 4.1910e-08, 3.3528e-08, 2.7474e-07, ..., 5.5414e-08, + 7.4506e-09, 2.6077e-07], + [ 1.2806e-07, 2.9802e-08, 3.4925e-07, ..., 1.4016e-07, + 1.3970e-09, 4.6985e-07], + [ 4.9127e-07, -9.4529e-08, -2.2398e-07, ..., 6.8778e-07, + 9.3132e-10, 2.3888e-07]], device='cuda:0') +Epoch 184, bias, value: tensor([ 9.9534e-05, 9.5707e-04, 1.0190e-02, 9.1858e-03, 2.7522e-02, + 4.3569e-02, -2.5344e-02, 7.6505e-03, -3.8411e-02, -2.1817e-02], + device='cuda:0'), grad: tensor([ 9.7323e-08, 1.8580e-07, -4.0978e-08, -9.7789e-08, 3.3574e-07, + -2.7232e-06, 9.4716e-07, 7.4273e-07, 8.0420e-07, -2.4447e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 217.52, cls_loss 0.0029 cls_loss_mapping 0.0046 cls_loss_causal 0.5287 re_mapping 0.0065 re_causal 0.0185 /// teacc 98.92 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0457, 0.1401, 0.0750, ..., -0.1976, -0.1135, 0.0228], + [ 0.1411, -0.0128, -0.0654, ..., 0.0874, 0.1075, -0.1324], + [-0.0816, -0.0517, -0.0511, ..., -0.0838, -0.0296, -0.0590], + ..., + [-0.0698, 0.1474, -0.1693, ..., -0.1424, 0.0036, -0.1497], + [ 0.0047, -0.1045, 0.0336, ..., 0.0615, -0.1099, -0.0428], + [-0.0841, -0.1046, 0.0658, ..., -0.0903, -0.0637, 0.1376]], + device='cuda:0'), grad: tensor([[ 7.0035e-07, -2.3190e-07, -5.2107e-07, ..., 3.8883e-07, + 1.1642e-08, -1.0859e-06], + [-3.1497e-06, 1.2321e-06, 3.5111e-07, ..., -2.2203e-06, + -1.3225e-07, 4.5402e-07], + [ 1.8207e-07, 3.5437e-07, 7.2177e-08, ..., 3.5437e-07, + -1.8626e-09, 4.9360e-08], + ..., + [ 9.8720e-08, -1.5780e-05, -2.7865e-06, ..., 2.7847e-07, + 5.1223e-09, 6.7055e-08], + [ 1.5572e-06, 6.0629e-07, 2.6217e-07, ..., 1.8291e-06, + 1.0850e-07, 5.8021e-07], + [ 6.2119e-07, 3.7439e-06, 1.7621e-06, ..., 7.7114e-06, + 5.1223e-09, 4.4145e-06]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0005, 0.0025, 0.0098, 0.0091, 0.0280, 0.0436, -0.0250, 0.0047, + -0.0356, -0.0221], device='cuda:0'), grad: tensor([-6.3516e-07, -6.2026e-07, 1.1632e-06, 2.3231e-05, -2.0981e-05, + -7.3433e-05, 7.5936e-05, -3.2812e-05, 4.9658e-06, 2.3171e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 217.35, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5288 re_mapping 0.0067 re_causal 0.0195 /// teacc 98.96 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0462, 0.1401, 0.0750, ..., -0.1985, -0.1136, 0.0228], + [ 0.1413, -0.0127, -0.0665, ..., 0.0873, 0.1077, -0.1331], + [-0.0820, -0.0516, -0.0504, ..., -0.0837, -0.0295, -0.0599], + ..., + [-0.0699, 0.1476, -0.1695, ..., -0.1426, 0.0039, -0.1496], + [ 0.0049, -0.1045, 0.0338, ..., 0.0622, -0.1101, -0.0427], + [-0.0846, -0.1056, 0.0662, ..., -0.0902, -0.0642, 0.1381]], + device='cuda:0'), grad: tensor([[ 1.5879e-07, -6.6683e-06, -3.1069e-06, ..., 1.3690e-07, + -1.8161e-08, -2.1905e-06], + [-5.9307e-06, -4.3446e-07, 3.8184e-08, ..., -9.6411e-06, + 4.6566e-10, 1.3970e-08], + [ 5.5088e-07, 9.7752e-06, 2.6729e-06, ..., 7.7672e-07, + 2.3283e-09, 1.8943e-06], + ..., + [ 9.9465e-07, -5.4017e-06, 6.2399e-08, ..., 1.4044e-06, + 4.6566e-10, 1.9092e-08], + [ 3.8408e-06, 1.3430e-06, 2.2305e-07, ..., 6.5416e-06, + 2.7940e-09, 3.6228e-07], + [ 1.3830e-07, 8.6427e-07, 2.8452e-07, ..., 2.1886e-07, + 6.5193e-09, 1.7881e-07]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0006, 0.0023, 0.0104, 0.0090, 0.0277, 0.0436, -0.0251, 0.0047, + -0.0355, -0.0219], device='cuda:0'), grad: tensor([-1.7717e-05, -1.1146e-05, 1.1548e-05, -1.1129e-06, 1.1679e-06, + 3.8818e-06, -1.2061e-07, -1.7369e-06, 1.2234e-05, 2.9765e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 216.97, cls_loss 0.0021 cls_loss_mapping 0.0033 cls_loss_causal 0.5310 re_mapping 0.0066 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0464, 0.1398, 0.0749, ..., -0.1992, -0.1136, 0.0228], + [ 0.1415, -0.0131, -0.0667, ..., 0.0873, 0.1082, -0.1332], + [-0.0825, -0.0518, -0.0505, ..., -0.0840, -0.0299, -0.0603], + ..., + [-0.0700, 0.1487, -0.1687, ..., -0.1427, 0.0037, -0.1488], + [ 0.0055, -0.1044, 0.0343, ..., 0.0634, -0.1101, -0.0424], + [-0.0851, -0.1087, 0.0659, ..., -0.0904, -0.0644, 0.1382]], + device='cuda:0'), grad: tensor([[ 1.5507e-07, 7.4040e-08, 4.1910e-09, ..., 2.6077e-07, + 4.6566e-10, 1.3039e-08], + [-1.6065e-07, 5.6345e-08, 2.8871e-08, ..., -2.1281e-07, + 4.8429e-08, 5.0757e-08], + [-1.3225e-07, 2.9802e-08, 2.3749e-08, ..., 1.2293e-07, + -1.4715e-07, 1.2573e-08], + ..., + [ 1.5879e-07, -2.8079e-07, 4.4703e-08, ..., 1.0850e-07, + 4.7963e-08, 8.0094e-08], + [ 2.6952e-06, 1.6969e-06, 2.1234e-07, ..., 4.8093e-06, + 2.7940e-09, 4.5029e-07], + [ 3.1665e-08, 1.8207e-07, -9.5041e-07, ..., 5.2154e-08, + 9.3132e-10, -1.8040e-06]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0009, 0.0021, 0.0102, 0.0088, 0.0278, 0.0438, -0.0266, 0.0058, + -0.0349, -0.0229], device='cuda:0'), grad: tensor([ 5.4156e-07, 2.2855e-06, -7.0594e-06, 1.2079e-06, 1.9781e-06, + 4.3400e-07, -9.3579e-06, 2.3395e-06, 9.8422e-06, -2.2054e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 217.30, cls_loss 0.0022 cls_loss_mapping 0.0029 cls_loss_causal 0.5668 re_mapping 0.0068 re_causal 0.0195 /// teacc 98.91 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0468, 0.1403, 0.0750, ..., -0.2001, -0.1136, 0.0230], + [ 0.1420, -0.0137, -0.0668, ..., 0.0874, 0.1088, -0.1333], + [-0.0833, -0.0524, -0.0508, ..., -0.0845, -0.0302, -0.0611], + ..., + [-0.0699, 0.1494, -0.1696, ..., -0.1429, 0.0034, -0.1495], + [ 0.0054, -0.1046, 0.0344, ..., 0.0635, -0.1102, -0.0423], + [-0.0864, -0.1094, 0.0662, ..., -0.0908, -0.0645, 0.1383]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -1.5944e-06, -5.1968e-07, ..., 2.3749e-08, + 0.0000e+00, -2.4624e-06], + [-2.0256e-07, 1.0571e-07, 6.3796e-08, ..., -1.2945e-07, + -4.6566e-10, 9.8720e-08], + [ 5.2620e-08, 8.0094e-08, 7.2177e-08, ..., 4.2841e-08, + 0.0000e+00, 1.3504e-07], + ..., + [ 1.0943e-07, -3.0641e-07, 1.0477e-07, ..., 1.4296e-07, + 0.0000e+00, 2.1374e-07], + [ 1.7649e-07, 1.2480e-07, 8.8476e-08, ..., 2.4354e-07, + 0.0000e+00, 3.3667e-07], + [ 2.5611e-08, 6.3563e-07, 1.3225e-07, ..., 6.1747e-07, + 0.0000e+00, 6.5565e-07]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0008, 0.0017, 0.0093, 0.0087, 0.0280, 0.0440, -0.0265, 0.0062, + -0.0351, -0.0231], device='cuda:0'), grad: tensor([-7.4133e-06, 2.4633e-07, 1.8161e-07, -6.3796e-08, -1.4175e-06, + 1.9390e-06, 1.2014e-06, 5.9465e-07, 1.1530e-06, 3.5670e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 217.38, cls_loss 0.0026 cls_loss_mapping 0.0046 cls_loss_causal 0.5645 re_mapping 0.0064 re_causal 0.0190 /// teacc 98.98 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0474, 0.1403, 0.0749, ..., -0.2009, -0.1136, 0.0230], + [ 0.1425, -0.0142, -0.0671, ..., 0.0876, 0.1095, -0.1334], + [-0.0854, -0.0525, -0.0516, ..., -0.0856, -0.0311, -0.0616], + ..., + [-0.0700, 0.1499, -0.1712, ..., -0.1431, 0.0033, -0.1499], + [ 0.0055, -0.1046, 0.0346, ..., 0.0634, -0.1103, -0.0425], + [-0.0871, -0.1103, 0.0669, ..., -0.0911, -0.0646, 0.1387]], + device='cuda:0'), grad: tensor([[ 1.5181e-06, 7.9349e-07, 9.3132e-10, ..., 1.1828e-06, + 4.6566e-09, -4.0047e-08], + [-6.5193e-08, 1.5786e-07, 2.6543e-08, ..., 3.0734e-08, + 1.7229e-08, 1.3039e-08], + [ 4.6566e-08, 3.6787e-08, 1.1176e-08, ..., 7.0315e-08, + 3.2596e-09, 2.3283e-09], + ..., + [ 1.7695e-08, -7.7672e-07, 3.6787e-08, ..., 7.5437e-08, + -2.8033e-07, 4.2841e-08], + [-4.3306e-07, 7.8697e-08, 4.5169e-08, ..., -7.7300e-07, + 1.6298e-08, 3.5297e-07], + [ 3.9581e-08, 2.4168e-07, -3.6880e-07, ..., 1.9046e-07, + 1.1455e-07, -5.0105e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0009, 0.0016, 0.0085, 0.0113, 0.0279, 0.0415, -0.0264, 0.0063, + -0.0349, -0.0228], device='cuda:0'), grad: tensor([ 9.6783e-06, 3.3006e-06, -1.6004e-05, 3.7905e-06, 8.7079e-08, + 1.9837e-06, -8.1882e-06, 3.2280e-06, 1.5013e-06, 5.8161e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 216.92, cls_loss 0.0021 cls_loss_mapping 0.0036 cls_loss_causal 0.5321 re_mapping 0.0062 re_causal 0.0186 /// teacc 98.88 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0473, 0.1409, 0.0747, ..., -0.2012, -0.1136, 0.0225], + [ 0.1429, -0.0145, -0.0672, ..., 0.0878, 0.1098, -0.1336], + [-0.0861, -0.0526, -0.0519, ..., -0.0863, -0.0314, -0.0622], + ..., + [-0.0700, 0.1502, -0.1719, ..., -0.1432, 0.0033, -0.1498], + [ 0.0052, -0.1048, 0.0344, ..., 0.0629, -0.1104, -0.0432], + [-0.0885, -0.1108, 0.0672, ..., -0.0920, -0.0646, 0.1387]], + device='cuda:0'), grad: tensor([[ 1.7323e-07, 2.5611e-08, -1.1176e-08, ..., 2.2724e-07, + 1.8626e-09, 6.7055e-08], + [-2.5379e-07, 5.0757e-08, 3.8184e-08, ..., -3.4086e-07, + -1.5367e-08, 8.7079e-08], + [ 3.8603e-07, -1.3970e-09, 3.8650e-08, ..., 5.3737e-07, + 1.4435e-08, 1.0012e-07], + ..., + [ 1.6345e-07, -4.8755e-07, 2.3516e-07, ..., 2.2165e-07, + 9.7789e-09, 6.3190e-07], + [ 1.4529e-06, 3.0734e-08, 1.8533e-07, ..., 1.7788e-06, + 2.0023e-08, 8.3493e-07], + [ 2.1327e-07, 7.7300e-08, -9.9279e-07, ..., 3.1525e-07, + 4.6566e-09, -2.9616e-06]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0010, 0.0015, 0.0082, 0.0110, 0.0287, 0.0419, -0.0259, 0.0066, + -0.0353, -0.0235], device='cuda:0'), grad: tensor([ 8.8708e-07, -2.0163e-07, -3.2922e-07, 4.8848e-07, 5.3495e-06, + 1.5870e-05, -2.2128e-05, 1.2135e-06, 2.8741e-06, -4.0084e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 216.85, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5367 re_mapping 0.0062 re_causal 0.0185 /// teacc 98.95 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0476, 0.1414, 0.0753, ..., -0.2019, -0.1136, 0.0228], + [ 0.1434, -0.0145, -0.0671, ..., 0.0883, 0.1099, -0.1333], + [-0.0866, -0.0527, -0.0519, ..., -0.0867, -0.0313, -0.0625], + ..., + [-0.0701, 0.1503, -0.1729, ..., -0.1433, 0.0033, -0.1502], + [ 0.0050, -0.1049, 0.0342, ..., 0.0627, -0.1105, -0.0435], + [-0.0891, -0.1109, 0.0677, ..., -0.0921, -0.0647, 0.1389]], + device='cuda:0'), grad: tensor([[ 3.2084e-07, -2.5611e-09, 1.9316e-06, ..., 1.3532e-06, + 1.4901e-08, 1.7975e-06], + [ 4.7311e-07, 2.9244e-07, 1.0310e-06, ..., 1.0701e-06, + 5.5879e-09, 6.1374e-07], + [ 1.2284e-06, 1.4538e-06, 9.6038e-06, ..., 5.1223e-06, + 1.2876e-07, 5.3123e-06], + ..., + [ 2.2259e-07, -2.8163e-06, 1.7639e-06, ..., 9.4436e-07, + 2.5611e-08, 1.1381e-06], + [-5.4464e-06, 6.1700e-08, -1.6186e-06, ..., -8.2105e-06, + 1.1409e-07, -9.9000e-07], + [ 5.4110e-07, 1.0082e-07, 2.6487e-06, ..., 1.9036e-06, + 3.4925e-08, 1.1902e-06]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0003, 0.0014, 0.0091, 0.0112, 0.0282, 0.0415, -0.0258, 0.0067, + -0.0355, -0.0234], device='cuda:0'), grad: tensor([ 7.0743e-06, 5.4725e-06, 3.1948e-05, 3.6746e-05, 1.2629e-06, + -6.0916e-05, 1.8720e-06, 1.5823e-06, -3.4779e-05, 9.6634e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 217.16, cls_loss 0.0019 cls_loss_mapping 0.0035 cls_loss_causal 0.5090 re_mapping 0.0060 re_causal 0.0181 /// teacc 98.90 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0477, 0.1416, 0.0753, ..., -0.2025, -0.1136, 0.0230], + [ 0.1431, -0.0147, -0.0679, ..., 0.0878, 0.1101, -0.1341], + [-0.0869, -0.0527, -0.0519, ..., -0.0871, -0.0315, -0.0630], + ..., + [-0.0701, 0.1507, -0.1734, ..., -0.1435, 0.0030, -0.1505], + [ 0.0054, -0.1049, 0.0353, ..., 0.0640, -0.1102, -0.0421], + [-0.0913, -0.1113, 0.0679, ..., -0.0924, -0.0649, 0.1390]], + device='cuda:0'), grad: tensor([[ 1.3388e-07, 2.4750e-07, -1.4901e-08, ..., 2.1816e-07, + 4.6566e-10, -2.0256e-08], + [-1.3458e-07, 2.6776e-08, 3.3993e-08, ..., 1.0524e-06, + -2.9569e-08, 5.8906e-08], + [ 5.3085e-08, 6.0303e-08, 1.0012e-08, ..., 1.8300e-07, + 3.0268e-09, 1.8161e-08], + ..., + [ 2.9197e-07, 4.5169e-08, 2.3516e-07, ..., 1.0636e-06, + 8.3819e-09, 1.2037e-07], + [-4.6031e-07, 1.9860e-07, -1.3551e-07, ..., -1.4175e-06, + 1.2340e-08, -6.7241e-07], + [ 4.9639e-07, 8.1956e-08, -3.6485e-07, ..., 3.9749e-06, + 1.3970e-09, -1.2782e-07]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0001, 0.0008, 0.0092, 0.0111, 0.0279, 0.0415, -0.0255, 0.0069, + -0.0351, -0.0234], device='cuda:0'), grad: tensor([ 5.2806e-07, 2.5257e-06, 2.7660e-07, -9.0990e-07, -1.1027e-05, + 4.5002e-06, -4.6864e-06, 3.4459e-06, -1.8831e-06, 7.2271e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 217.21, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5340 re_mapping 0.0060 re_causal 0.0182 /// teacc 98.91 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0484, 0.1415, 0.0751, ..., -0.2035, -0.1138, 0.0228], + [ 0.1439, -0.0145, -0.0691, ..., 0.0880, 0.1107, -0.1355], + [-0.0881, -0.0531, -0.0523, ..., -0.0887, -0.0322, -0.0641], + ..., + [-0.0703, 0.1510, -0.1742, ..., -0.1438, 0.0026, -0.1515], + [ 0.0054, -0.1050, 0.0355, ..., 0.0637, -0.1094, -0.0422], + [-0.0921, -0.1118, 0.0685, ..., -0.0930, -0.0651, 0.1393]], + device='cuda:0'), grad: tensor([[ 1.0943e-08, 6.8024e-06, 1.6764e-08, ..., 7.9498e-06, + 0.0000e+00, 1.5469e-06], + [-6.1747e-07, 1.0729e-06, 1.4901e-08, ..., -3.4925e-08, + -3.4925e-09, 9.0571e-08], + [ 2.4214e-08, 1.1278e-06, 3.9581e-08, ..., 2.0908e-07, + -2.0955e-09, 3.9348e-08], + ..., + [ 2.9732e-07, -3.6880e-06, 3.4692e-08, ..., 3.8208e-07, + 2.3283e-09, 4.0745e-08], + [-1.1688e-07, 9.7789e-07, 3.7509e-07, ..., 8.4750e-07, + 2.3283e-10, 8.7032e-07], + [ 4.1234e-07, 1.3579e-06, -7.5903e-08, ..., 1.6186e-06, + 6.9849e-10, 1.5297e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0005, 0.0008, 0.0085, 0.0112, 0.0285, 0.0414, -0.0250, 0.0071, + -0.0352, -0.0237], device='cuda:0'), grad: tensor([ 3.1888e-05, 1.6261e-06, 1.6196e-06, 1.4864e-06, 2.4602e-05, + 6.2399e-07, -6.8486e-05, -3.6880e-06, 4.5076e-06, 5.8189e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 192---------------------------------------------------- +epoch 192, time 217.92, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5147 re_mapping 0.0058 re_causal 0.0172 /// teacc 99.05 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0493, 0.1398, 0.0756, ..., -0.2064, -0.1150, 0.0232], + [ 0.1442, -0.0160, -0.0693, ..., 0.0876, 0.1107, -0.1360], + [-0.0884, -0.0534, -0.0525, ..., -0.0890, -0.0323, -0.0652], + ..., + [-0.0703, 0.1515, -0.1744, ..., -0.1432, 0.0027, -0.1517], + [ 0.0053, -0.1048, 0.0353, ..., 0.0625, -0.1095, -0.0425], + [-0.0926, -0.1121, 0.0686, ..., -0.0933, -0.0651, 0.1395]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, 2.2352e-08, 2.7940e-09, ..., 1.5367e-08, + 0.0000e+00, -3.2596e-09], + [ 2.6403e-07, 5.5740e-07, 1.9139e-07, ..., 6.7754e-07, + -1.3970e-09, 3.3202e-07], + [ 1.8626e-08, 9.8161e-07, 2.7008e-08, ..., 3.2131e-08, + 0.0000e+00, 6.5193e-09], + ..., + [ 6.8452e-08, -4.9509e-06, 2.3982e-07, ..., 2.0256e-07, + 9.3132e-10, 4.2981e-07], + [-4.2468e-07, 3.4878e-07, 1.1129e-07, ..., -8.4797e-07, + 0.0000e+00, 1.7555e-07], + [ 4.8429e-08, 9.5135e-07, -1.5143e-06, ..., -3.1525e-07, + 0.0000e+00, -2.7865e-06]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0016, -0.0007, 0.0085, 0.0111, 0.0286, 0.0416, -0.0238, 0.0076, + -0.0353, -0.0239], device='cuda:0'), grad: tensor([ 1.1921e-07, 2.7232e-06, 1.3541e-06, 2.4270e-06, 3.9786e-06, + -3.8967e-06, 4.3325e-06, -7.0743e-06, -3.8417e-07, -3.6322e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 217.25, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.4958 re_mapping 0.0064 re_causal 0.0177 /// teacc 99.01 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0494, 0.1409, 0.0749, ..., -0.2068, -0.1150, 0.0224], + [ 0.1445, -0.0165, -0.0710, ..., 0.0866, 0.1123, -0.1377], + [-0.0893, -0.0537, -0.0512, ..., -0.0890, -0.0339, -0.0632], + ..., + [-0.0703, 0.1521, -0.1756, ..., -0.1432, 0.0025, -0.1526], + [ 0.0053, -0.1050, 0.0352, ..., 0.0621, -0.1095, -0.0426], + [-0.0929, -0.1125, 0.0702, ..., -0.0925, -0.0653, 0.1409]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, -1.8254e-07, -1.3039e-08, ..., 2.5146e-08, + 9.3132e-10, -3.1199e-08], + [-1.6112e-07, 3.9581e-08, 3.5856e-08, ..., -2.5006e-07, + 4.6566e-10, 4.0978e-08], + [ 1.4435e-08, 1.0850e-07, 1.2573e-08, ..., 2.9802e-08, + -1.2107e-08, 1.2107e-08], + ..., + [ 2.9802e-08, -1.5274e-07, 1.3830e-07, ..., 9.0338e-08, + 2.3283e-09, 1.9558e-07], + [ 3.1199e-08, 5.2154e-08, 2.2072e-07, ..., 6.5658e-08, + 3.7253e-09, 3.2689e-07], + [ 6.1002e-08, 6.5658e-08, -4.8056e-07, ..., 1.2014e-07, + 4.6566e-10, -6.9849e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0023, -0.0022, 0.0093, 0.0110, 0.0270, 0.0415, -0.0232, 0.0078, + -0.0357, -0.0220], device='cuda:0'), grad: tensor([-1.7788e-07, -2.8033e-07, 3.0268e-08, -6.2305e-07, 8.8010e-08, + 7.3574e-07, -6.9849e-08, 3.4878e-07, 8.6939e-07, -9.3831e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 194---------------------------------------------------- +epoch 194, time 218.21, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5179 re_mapping 0.0063 re_causal 0.0183 /// teacc 99.09 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0497, 0.1411, 0.0745, ..., -0.2076, -0.1150, 0.0219], + [ 0.1457, -0.0165, -0.0708, ..., 0.0874, 0.1126, -0.1379], + [-0.0897, -0.0540, -0.0512, ..., -0.0891, -0.0342, -0.0629], + ..., + [-0.0705, 0.1523, -0.1762, ..., -0.1436, 0.0025, -0.1536], + [ 0.0053, -0.1051, 0.0352, ..., 0.0622, -0.1096, -0.0422], + [-0.0940, -0.1126, 0.0705, ..., -0.0929, -0.0654, 0.1411]], + device='cuda:0'), grad: tensor([[ 4.4238e-07, 8.7544e-08, -6.5193e-09, ..., 5.6485e-07, + 9.3132e-10, 1.2247e-07], + [-4.5076e-06, -1.1139e-06, -2.1374e-07, ..., -7.1004e-06, + 5.1223e-09, 2.2817e-08], + [ 5.0291e-07, 3.2736e-07, 1.3039e-08, ..., 6.5891e-07, + -1.3970e-08, 1.6764e-08], + ..., + [ 1.3821e-06, -3.7719e-07, 1.4668e-07, ..., 2.7362e-06, + 1.8626e-09, 1.6764e-08], + [ 5.2340e-06, 2.1420e-06, 3.2131e-08, ..., 6.5118e-06, + 2.3283e-09, 2.4885e-06], + [ 4.8196e-07, 5.1409e-07, 1.1642e-08, ..., 9.7509e-07, + 4.6566e-10, -1.8161e-08]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0028, -0.0022, 0.0094, 0.0109, 0.0270, 0.0413, -0.0229, 0.0078, + -0.0356, -0.0218], device='cuda:0'), grad: tensor([ 9.1922e-07, -1.4685e-05, 7.7439e-07, 3.7113e-07, 9.2061e-07, + 6.7830e-05, -7.5161e-05, 5.1744e-06, 1.1533e-05, 2.5332e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 217.14, cls_loss 0.0025 cls_loss_mapping 0.0050 cls_loss_causal 0.5187 re_mapping 0.0066 re_causal 0.0175 /// teacc 98.95 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0503, 0.1413, 0.0747, ..., -0.2087, -0.1154, 0.0221], + [ 0.1467, -0.0167, -0.0708, ..., 0.0881, 0.1134, -0.1381], + [-0.0906, -0.0571, -0.0514, ..., -0.0897, -0.0346, -0.0631], + ..., + [-0.0706, 0.1546, -0.1766, ..., -0.1439, 0.0021, -0.1546], + [ 0.0054, -0.1052, 0.0349, ..., 0.0626, -0.1096, -0.0431], + [-0.0947, -0.1131, 0.0712, ..., -0.0925, -0.0658, 0.1418]], + device='cuda:0'), grad: tensor([[-2.0489e-08, -4.8196e-07, 3.3155e-07, ..., 9.4995e-08, + 4.6566e-10, 8.4750e-08], + [-1.8850e-06, 1.2852e-07, 9.9186e-08, ..., -2.5053e-06, + -5.1223e-09, 3.4366e-07], + [ 3.5670e-07, 1.3504e-08, 2.5379e-07, ..., 4.8708e-07, + 4.6566e-10, 1.0198e-07], + ..., + [ 3.5297e-07, -5.9605e-08, 7.0734e-07, ..., 6.0070e-07, + 1.8626e-09, 8.6613e-07], + [-3.8669e-06, -1.3970e-08, 3.3043e-06, ..., -6.9216e-06, + 4.6566e-10, -3.7607e-06], + [ 2.9188e-06, 8.7544e-08, 6.5304e-06, ..., 6.6757e-06, + 9.3132e-10, 6.2287e-06]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0030, -0.0019, 0.0085, 0.0107, 0.0266, 0.0415, -0.0238, 0.0091, + -0.0361, -0.0213], device='cuda:0'), grad: tensor([ 4.3772e-07, -3.9712e-06, 8.0932e-07, -3.4004e-05, -1.6876e-06, + 4.3511e-06, 4.2319e-06, 3.2298e-06, -2.6450e-06, 2.9191e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 216.86, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5055 re_mapping 0.0065 re_causal 0.0190 /// teacc 99.01 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0504, 0.1418, 0.0746, ..., -0.2091, -0.1155, 0.0222], + [ 0.1474, -0.0169, -0.0709, ..., 0.0883, 0.1140, -0.1383], + [-0.0911, -0.0571, -0.0511, ..., -0.0900, -0.0350, -0.0631], + ..., + [-0.0708, 0.1547, -0.1770, ..., -0.1443, 0.0018, -0.1548], + [ 0.0056, -0.1054, 0.0348, ..., 0.0631, -0.1087, -0.0434], + [-0.0952, -0.1131, 0.0714, ..., -0.0924, -0.0662, 0.1421]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 8.3353e-07, 1.8869e-06, ..., 1.1232e-06, + 3.7253e-08, 2.3544e-06], + [-1.6764e-08, 2.6748e-05, 1.3836e-05, ..., 7.9023e-07, + 2.7940e-09, 1.7444e-06], + [ 1.8626e-09, 1.1735e-05, 9.9167e-06, ..., 4.0904e-06, + 1.3039e-08, 7.1302e-06], + ..., + [ 1.4901e-08, 7.1907e-04, 3.4547e-04, ..., 4.7311e-07, + 1.3504e-08, 1.5944e-06], + [-9.3132e-09, 9.5218e-06, 5.8152e-06, ..., 7.7672e-07, + 2.9337e-08, 2.0936e-06], + [ 4.1910e-09, -7.7438e-04, -3.3593e-04, ..., 2.7388e-05, + 8.3819e-09, 5.9813e-05]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0029, -0.0018, 0.0086, 0.0107, 0.0262, 0.0413, -0.0241, 0.0089, + -0.0358, -0.0209], device='cuda:0'), grad: tensor([ 7.8902e-06, 1.2732e-04, 6.6400e-05, 2.2733e-04, 2.2218e-05, + -3.2783e-04, 3.3304e-06, 3.3607e-03, 4.7296e-05, -3.5362e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 216.86, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5165 re_mapping 0.0064 re_causal 0.0180 /// teacc 99.02 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0504, 0.1427, 0.0751, ..., -0.2094, -0.1155, 0.0225], + [ 0.1475, -0.0178, -0.0712, ..., 0.0884, 0.1142, -0.1387], + [-0.0914, -0.0569, -0.0512, ..., -0.0903, -0.0356, -0.0633], + ..., + [-0.0708, 0.1545, -0.1789, ..., -0.1444, 0.0026, -0.1567], + [ 0.0057, -0.1055, 0.0348, ..., 0.0634, -0.1089, -0.0433], + [-0.0957, -0.1113, 0.0709, ..., -0.0933, -0.0664, 0.1416]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, -3.1199e-08, -1.3039e-08, ..., 1.5134e-07, + 1.8626e-09, -1.9558e-08], + [ 1.4435e-08, 1.1036e-07, 8.3819e-09, ..., 4.3772e-08, + 4.6566e-10, 2.1420e-08], + [ 1.3504e-08, 1.3039e-08, 6.0536e-09, ..., -2.9523e-07, + -3.8184e-08, 8.8476e-09], + ..., + [-1.8626e-09, -2.4401e-07, 1.3970e-08, ..., 5.1688e-08, + 3.2596e-09, 3.0268e-08], + [ 2.5658e-07, 5.7742e-08, 1.1688e-07, ..., 3.1432e-07, + 2.5611e-08, 3.4133e-07], + [ 4.4238e-08, 1.2806e-07, -1.1409e-07, ..., 1.3923e-07, + 9.3132e-10, -1.7788e-07]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0027, -0.0023, 0.0088, 0.0107, 0.0265, 0.0413, -0.0235, 0.0081, + -0.0358, -0.0204], device='cuda:0'), grad: tensor([ 3.7625e-07, 4.3772e-07, -2.8051e-06, 1.0692e-06, 1.6108e-05, + -1.7546e-06, -1.5661e-05, 8.2748e-07, 1.3653e-06, 2.3283e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 216.85, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.4966 re_mapping 0.0064 re_causal 0.0181 /// teacc 98.99 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0507, 0.1431, 0.0754, ..., -0.2101, -0.1155, 0.0227], + [ 0.1477, -0.0180, -0.0714, ..., 0.0886, 0.1144, -0.1389], + [-0.0920, -0.0569, -0.0513, ..., -0.0905, -0.0358, -0.0635], + ..., + [-0.0708, 0.1547, -0.1790, ..., -0.1447, 0.0025, -0.1569], + [ 0.0059, -0.1055, 0.0348, ..., 0.0638, -0.1090, -0.0435], + [-0.0961, -0.1116, 0.0712, ..., -0.0939, -0.0664, 0.1419]], + device='cuda:0'), grad: tensor([[ 2.8405e-08, -1.3039e-08, 1.2107e-07, ..., 1.1502e-07, + 0.0000e+00, 2.0862e-07], + [-2.1253e-06, 1.1222e-07, 4.4238e-08, ..., -2.5351e-06, + 0.0000e+00, 7.6368e-08], + [ 1.8086e-06, 6.2166e-07, 2.3376e-07, ..., 2.5742e-06, + 0.0000e+00, 3.2736e-07], + ..., + [ 1.8114e-07, -8.8243e-07, 7.2177e-08, ..., 3.8277e-07, + 0.0000e+00, 1.2480e-07], + [ 6.6590e-08, 2.0489e-08, 8.7824e-07, ..., 2.0489e-07, + 0.0000e+00, 1.6317e-06], + [ 4.2841e-08, 1.0058e-07, 3.3900e-06, ..., 2.0750e-06, + 0.0000e+00, 6.1169e-06]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0026, -0.0023, 0.0088, 0.0107, 0.0268, 0.0416, -0.0240, 0.0081, + -0.0358, -0.0206], device='cuda:0'), grad: tensor([ 5.3924e-07, -4.6045e-06, 4.9360e-06, 8.6203e-06, -4.7497e-08, + -2.2501e-05, 6.7009e-07, -1.6531e-07, 2.4941e-06, 1.0096e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 216.92, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5240 re_mapping 0.0060 re_causal 0.0182 /// teacc 99.08 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0514, 0.1431, 0.0759, ..., -0.2107, -0.1163, 0.0230], + [ 0.1490, -0.0178, -0.0715, ..., 0.0892, 0.1154, -0.1391], + [-0.0925, -0.0568, -0.0514, ..., -0.0908, -0.0360, -0.0637], + ..., + [-0.0713, 0.1547, -0.1791, ..., -0.1456, 0.0012, -0.1571], + [ 0.0058, -0.1056, 0.0346, ..., 0.0637, -0.1092, -0.0439], + [-0.0964, -0.1120, 0.0711, ..., -0.0942, -0.0668, 0.1419]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, -6.9849e-08, 1.1176e-07, ..., 2.0824e-06, + 1.3970e-09, -1.9092e-08], + [ 3.8370e-07, 1.6065e-07, 2.2259e-07, ..., 1.0636e-06, + 1.2573e-08, 1.7928e-07], + [ 4.6566e-08, 7.9628e-08, 5.1036e-07, ..., 3.6508e-07, + 5.0757e-08, 1.4110e-07], + ..., + [ 6.2864e-08, -2.1886e-07, 1.6764e-07, ..., 2.0349e-07, + -9.7789e-09, 8.1025e-08], + [-1.3284e-05, -4.5635e-07, -7.5214e-06, ..., -3.0175e-05, + 3.7253e-09, -6.0573e-06], + [ 1.7043e-07, 9.4064e-08, -1.7881e-07, ..., 4.5858e-06, + 5.5879e-09, -2.7753e-07]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0025, -0.0017, 0.0088, 0.0109, 0.0269, 0.0416, -0.0242, 0.0077, + -0.0359, -0.0209], device='cuda:0'), grad: tensor([ 9.0674e-06, 1.9334e-06, 2.7902e-06, 8.7684e-07, -3.6150e-05, + 2.2322e-05, 2.4587e-05, 6.2445e-07, -4.4048e-05, 1.8016e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 200---------------------------------------------------- +epoch 200, time 217.82, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.4658 re_mapping 0.0062 re_causal 0.0179 /// teacc 99.11 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0547, 0.1418, 0.0762, ..., -0.2116, -0.1191, 0.0232], + [ 0.1490, -0.0179, -0.0716, ..., 0.0892, 0.1155, -0.1393], + [-0.0926, -0.0569, -0.0515, ..., -0.0908, -0.0362, -0.0639], + ..., + [-0.0713, 0.1549, -0.1794, ..., -0.1456, 0.0012, -0.1576], + [ 0.0062, -0.1057, 0.0345, ..., 0.0645, -0.1090, -0.0438], + [-0.0970, -0.1122, 0.0713, ..., -0.0945, -0.0670, 0.1421]], + device='cuda:0'), grad: tensor([[-6.2818e-07, -1.3560e-06, 8.8476e-09, ..., 2.2911e-07, + 9.5926e-08, -4.3120e-07], + [ 1.8254e-06, 1.1921e-07, 1.1642e-08, ..., 2.1197e-06, + 9.8627e-07, 2.8592e-07], + [ 5.3458e-06, 6.3796e-08, 4.0559e-07, ..., 6.3293e-06, + 3.3788e-06, 6.5006e-07], + ..., + [ 1.0924e-06, -1.4435e-08, 4.0978e-08, ..., 1.3830e-06, + 6.0257e-07, 1.4156e-07], + [-9.9361e-05, 2.1514e-07, -9.9372e-07, ..., -1.1927e-04, + -5.4091e-05, -1.0602e-05], + [ 2.0629e-07, 3.6787e-08, -8.5682e-08, ..., 1.0431e-06, + 9.4529e-08, -1.0245e-07]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0048, -0.0019, 0.0089, 0.0107, 0.0270, 0.0417, -0.0228, 0.0077, + -0.0357, -0.0209], device='cuda:0'), grad: tensor([-2.7120e-06, 5.7705e-06, 1.7390e-05, 1.1604e-06, -1.1222e-06, + 2.6488e-04, 9.7677e-06, 3.4925e-06, -3.0184e-04, 2.6524e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 216.93, cls_loss 0.0016 cls_loss_mapping 0.0042 cls_loss_causal 0.5273 re_mapping 0.0060 re_causal 0.0176 /// teacc 99.01 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0548, 0.1423, 0.0768, ..., -0.2120, -0.1192, 0.0237], + [ 0.1497, -0.0179, -0.0720, ..., 0.0895, 0.1164, -0.1399], + [-0.0931, -0.0569, -0.0517, ..., -0.0909, -0.0367, -0.0643], + ..., + [-0.0716, 0.1550, -0.1802, ..., -0.1462, 0.0004, -0.1580], + [ 0.0063, -0.1058, 0.0341, ..., 0.0648, -0.1081, -0.0441], + [-0.0974, -0.1123, 0.0719, ..., -0.0943, -0.0685, 0.1425]], + device='cuda:0'), grad: tensor([[ 2.0210e-07, 4.3306e-08, 1.3504e-07, ..., 8.5589e-07, + 3.7253e-09, -3.1898e-07], + [ 3.5763e-07, 2.2519e-06, 5.7276e-08, ..., 1.3178e-07, + 1.8915e-06, 2.9337e-08], + [-3.2736e-07, 3.4459e-08, -1.2554e-06, ..., 8.4145e-07, + -2.0117e-06, -1.8813e-06], + ..., + [ 2.5611e-08, -9.8050e-06, 3.2783e-07, ..., 2.2352e-08, + 1.8347e-07, 6.2399e-08], + [ 3.0510e-06, 6.8452e-08, 1.4883e-06, ..., 7.2233e-06, + 6.9384e-08, 1.9651e-06], + [ 3.2596e-09, 7.5027e-06, 4.9407e-07, ..., 2.2305e-07, + 2.1886e-08, 4.8010e-07]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0046, -0.0015, 0.0088, 0.0107, 0.0266, 0.0417, -0.0228, 0.0074, + -0.0359, -0.0205], device='cuda:0'), grad: tensor([ 4.2133e-06, 1.5229e-05, -1.3381e-05, -8.2701e-06, 3.0175e-06, + 1.0960e-05, -3.9428e-05, -1.3448e-05, 2.8595e-05, 1.2577e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 217.26, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.4971 re_mapping 0.0063 re_causal 0.0176 /// teacc 98.99 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0547, 0.1441, 0.0764, ..., -0.2124, -0.1193, 0.0235], + [ 0.1505, -0.0176, -0.0725, ..., 0.0902, 0.1165, -0.1407], + [-0.0935, -0.0573, -0.0518, ..., -0.0911, -0.0368, -0.0645], + ..., + [-0.0721, 0.1548, -0.1805, ..., -0.1473, 0.0007, -0.1591], + [ 0.0065, -0.1059, 0.0339, ..., 0.0649, -0.1082, -0.0445], + [-0.0967, -0.1118, 0.0730, ..., -0.0939, -0.0677, 0.1436]], + device='cuda:0'), grad: tensor([[ 1.2442e-06, 1.4482e-07, 1.4473e-06, ..., 1.0580e-06, + 4.6566e-10, 1.5730e-06], + [-5.9139e-08, 4.3306e-08, 3.0082e-07, ..., 5.4156e-07, + -5.5879e-09, 5.5460e-07], + [ 2.5146e-08, 2.0489e-08, 3.9116e-08, ..., 3.9581e-08, + 9.3132e-10, 7.7300e-08], + ..., + [ 4.5169e-08, -5.6345e-08, 5.0198e-07, ..., 1.7649e-07, + 9.3132e-10, 1.4184e-06], + [-4.1723e-06, -8.9360e-07, -3.7923e-06, ..., -2.4792e-06, + 1.8626e-09, -3.7216e-06], + [ 6.3004e-07, 1.6345e-07, -3.2373e-06, ..., 3.5521e-06, + 9.3132e-10, -8.3447e-06]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0036, -0.0017, 0.0089, 0.0107, 0.0261, 0.0415, -0.0228, 0.0058, + -0.0361, -0.0182], device='cuda:0'), grad: tensor([ 3.6098e-06, 1.8440e-06, 1.6578e-07, 2.4214e-06, 3.3043e-06, + 4.6603e-06, 6.8499e-07, 3.1497e-06, -7.2867e-06, -1.2532e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 216.92, cls_loss 0.0012 cls_loss_mapping 0.0029 cls_loss_causal 0.4860 re_mapping 0.0062 re_causal 0.0180 /// teacc 98.89 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0549, 0.1442, 0.0765, ..., -0.2132, -0.1193, 0.0236], + [ 0.1508, -0.0186, -0.0725, ..., 0.0904, 0.1170, -0.1409], + [-0.0939, -0.0573, -0.0520, ..., -0.0913, -0.0370, -0.0647], + ..., + [-0.0721, 0.1556, -0.1807, ..., -0.1474, 0.0014, -0.1597], + [ 0.0063, -0.1059, 0.0336, ..., 0.0648, -0.1083, -0.0451], + [-0.0963, -0.1121, 0.0742, ..., -0.0940, -0.0681, 0.1448]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -1.0850e-07, -3.8929e-07, ..., 1.7695e-08, + 0.0000e+00, -5.2107e-07], + [-1.2619e-07, 2.2817e-08, 4.8429e-08, ..., -1.8626e-07, + -4.6566e-10, 8.9873e-08], + [ 6.5193e-09, 1.8207e-07, 9.5461e-08, ..., -6.0536e-09, + 0.0000e+00, 1.0896e-07], + ..., + [ 7.5437e-08, -9.2993e-07, 1.7369e-07, ..., 2.6729e-07, + -4.6566e-10, 2.8918e-07], + [ 2.9197e-07, 2.9337e-08, 2.2398e-07, ..., 6.4075e-07, + 4.6566e-10, 1.0002e-06], + [ 4.5635e-08, 6.7707e-07, -1.5348e-05, ..., -1.2174e-05, + 1.3970e-09, -2.9042e-05]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0037, -0.0020, 0.0088, 0.0106, 0.0262, 0.0411, -0.0226, 0.0062, + -0.0364, -0.0179], device='cuda:0'), grad: tensor([-1.1437e-06, -1.5600e-07, 3.9069e-07, 5.2992e-07, 5.0455e-05, + -5.9232e-07, 6.4820e-07, -5.0850e-07, 9.2434e-07, -5.0515e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 216.71, cls_loss 0.0022 cls_loss_mapping 0.0041 cls_loss_causal 0.5371 re_mapping 0.0061 re_causal 0.0173 /// teacc 99.03 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0568, 0.1415, 0.0751, ..., -0.2169, -0.1197, 0.0221], + [ 0.1531, -0.0175, -0.0734, ..., 0.0917, 0.1176, -0.1430], + [-0.0943, -0.0574, -0.0521, ..., -0.0915, -0.0372, -0.0649], + ..., + [-0.0732, 0.1553, -0.1822, ..., -0.1497, 0.0008, -0.1603], + [ 0.0062, -0.1060, 0.0335, ..., 0.0647, -0.1085, -0.0454], + [-0.0966, -0.1122, 0.0757, ..., -0.0939, -0.0663, 0.1457]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.7121e-05, -2.9244e-07, ..., 1.1176e-08, + 0.0000e+00, -5.6252e-07], + [ 1.9930e-07, 6.8871e-07, 1.0664e-07, ..., 3.4412e-07, + 0.0000e+00, 1.3504e-07], + [ 6.6590e-08, -4.6611e-05, 4.3772e-08, ..., 4.4703e-08, + 0.0000e+00, 6.7055e-08], + ..., + [ 1.6298e-08, 1.6853e-05, 1.3504e-08, ..., 7.1246e-08, + 0.0000e+00, 1.3970e-08], + [-7.3249e-07, 7.8678e-06, -2.2724e-07, ..., -1.0170e-06, + 0.0000e+00, -2.3609e-07], + [ 1.1548e-07, 3.2932e-06, 1.8766e-07, ..., 2.8824e-07, + 0.0000e+00, 3.2876e-07]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0059, -0.0015, 0.0087, 0.0109, 0.0261, 0.0411, -0.0210, 0.0055, + -0.0365, -0.0175], device='cuda:0'), grad: tensor([ 4.0978e-05, 2.0079e-06, -1.1355e-04, 6.0489e-07, -2.1840e-07, + 9.4017e-07, 1.0943e-06, 4.4584e-05, 1.7777e-05, 5.8338e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 217.40, cls_loss 0.0020 cls_loss_mapping 0.0037 cls_loss_causal 0.5254 re_mapping 0.0057 re_causal 0.0167 /// teacc 98.96 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0569, 0.1424, 0.0760, ..., -0.2173, -0.1197, 0.0234], + [ 0.1535, -0.0184, -0.0738, ..., 0.0915, 0.1183, -0.1435], + [-0.0944, -0.0572, -0.0523, ..., -0.0915, -0.0376, -0.0651], + ..., + [-0.0734, 0.1566, -0.1825, ..., -0.1499, 0.0003, -0.1614], + [ 0.0063, -0.1063, 0.0334, ..., 0.0647, -0.1088, -0.0453], + [-0.0976, -0.1146, 0.0757, ..., -0.0946, -0.0668, 0.1456]], + device='cuda:0'), grad: tensor([[ 1.0878e-06, -1.2685e-06, -1.6708e-06, ..., 6.6124e-08, + -6.0536e-09, -2.8964e-07], + [-1.8626e-08, 4.7684e-07, 1.4855e-07, ..., -1.4901e-08, + -1.8161e-08, 2.2398e-07], + [ 3.2689e-07, -2.6226e-06, 1.6764e-08, ..., 3.0175e-07, + -3.1199e-08, 8.8802e-07], + ..., + [ 2.9104e-07, 2.1756e-06, 4.1118e-07, ..., 3.9767e-07, + 1.5367e-08, 4.9500e-07], + [-7.4646e-07, -1.1455e-07, -2.4866e-07, ..., -1.2303e-06, + 4.6566e-09, 5.9651e-07], + [ 1.3551e-07, 1.7649e-06, 3.8138e-07, ..., 5.3644e-07, + 1.4435e-08, 1.3188e-06]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0049, -0.0021, 0.0088, 0.0109, 0.0265, 0.0412, -0.0214, 0.0062, + -0.0367, -0.0184], device='cuda:0'), grad: tensor([ 2.6934e-06, 1.7062e-06, -9.0078e-06, 8.5915e-07, 3.6741e-07, + 4.8690e-06, -1.7226e-05, 9.9242e-06, -1.0990e-06, 6.9141e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 217.29, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5063 re_mapping 0.0061 re_causal 0.0171 /// teacc 99.01 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0574, 0.1429, 0.0760, ..., -0.2178, -0.1200, 0.0234], + [ 0.1549, -0.0186, -0.0743, ..., 0.0921, 0.1182, -0.1444], + [-0.0947, -0.0574, -0.0521, ..., -0.0918, -0.0378, -0.0653], + ..., + [-0.0738, 0.1573, -0.1834, ..., -0.1503, 0.0006, -0.1622], + [ 0.0063, -0.1065, 0.0333, ..., 0.0647, -0.1092, -0.0456], + [-0.1001, -0.1156, 0.0756, ..., -0.0958, -0.0659, 0.1455]], + device='cuda:0'), grad: tensor([[ 3.8650e-08, 1.2992e-07, 1.3327e-06, ..., 3.2131e-07, + 3.7253e-09, 3.9907e-07], + [-6.1747e-07, 2.4475e-06, 2.5285e-07, ..., -8.7498e-07, + -3.2596e-09, 1.2899e-07], + [ 1.1642e-07, 3.1330e-06, 2.4527e-05, ..., 8.9630e-06, + -4.1910e-08, 1.5095e-05], + ..., + [ 1.9791e-07, -1.9163e-05, 2.1793e-07, ..., 3.5483e-07, + -1.1362e-07, 1.2759e-07], + [-2.7940e-06, -3.0771e-06, -3.4332e-05, ..., -1.3240e-05, + 2.9337e-08, -2.0653e-05], + [ 1.0151e-07, 1.3970e-05, 3.8184e-06, ..., 2.3637e-06, + 1.2573e-08, 2.4438e-06]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0048, -0.0020, 0.0087, 0.0110, 0.0273, 0.0416, -0.0218, 0.0065, + -0.0370, -0.0193], device='cuda:0'), grad: tensor([ 3.3602e-06, 3.0454e-06, 5.1290e-05, 6.6459e-06, -1.4361e-06, + 6.8620e-06, 8.1258e-07, -3.0518e-05, -7.2420e-05, 3.2336e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 217.06, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5266 re_mapping 0.0059 re_causal 0.0169 /// teacc 98.93 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.0576, 0.1434, 0.0756, ..., -0.2181, -0.1199, 0.0232], + [ 0.1549, -0.0185, -0.0760, ..., 0.0916, 0.1186, -0.1468], + [-0.0953, -0.0574, -0.0535, ..., -0.0921, -0.0380, -0.0666], + ..., + [-0.0743, 0.1577, -0.1857, ..., -0.1509, 0.0004, -0.1639], + [ 0.0083, -0.1067, 0.0355, ..., 0.0674, -0.1100, -0.0426], + [-0.1023, -0.1160, 0.0760, ..., -0.0965, -0.0663, 0.1457]], + device='cuda:0'), grad: tensor([[ 4.5635e-08, -1.2051e-06, 5.1688e-08, ..., 8.4750e-08, + -7.9488e-07, -1.1213e-06], + [-3.5716e-07, 4.5672e-06, 7.6368e-08, ..., 2.5276e-06, + 2.4447e-07, 5.3085e-08], + [ 4.8894e-08, 1.9651e-07, -5.3877e-07, ..., 1.0058e-07, + 2.9802e-08, 5.0757e-08], + ..., + [-4.5775e-07, -5.0999e-06, 2.2119e-07, ..., 1.0319e-06, + -2.2538e-07, 3.0175e-07], + [-2.8126e-06, -6.4261e-08, 1.7304e-06, ..., -1.7174e-06, + 8.8010e-08, 3.4608e-06], + [ 1.7742e-07, 3.6228e-07, -4.9621e-06, ..., 1.1977e-06, + 1.5786e-07, -8.9332e-06]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0048, -0.0021, 0.0085, 0.0121, 0.0273, 0.0405, -0.0218, 0.0062, + -0.0359, -0.0194], device='cuda:0'), grad: tensor([-3.0678e-06, 1.2301e-05, -6.4746e-06, 4.9267e-07, -3.5930e-06, + 6.3442e-06, 3.7961e-06, -3.8147e-06, 1.1176e-05, -1.7270e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 217.29, cls_loss 0.0020 cls_loss_mapping 0.0028 cls_loss_causal 0.4938 re_mapping 0.0058 re_causal 0.0168 /// teacc 99.06 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.0577, 0.1443, 0.0762, ..., -0.2187, -0.1199, 0.0242], + [ 0.1563, -0.0183, -0.0771, ..., 0.0924, 0.1196, -0.1476], + [-0.0955, -0.0576, -0.0540, ..., -0.0924, -0.0381, -0.0672], + ..., + [-0.0749, 0.1581, -0.1874, ..., -0.1513, -0.0009, -0.1660], + [ 0.0083, -0.1069, 0.0357, ..., 0.0675, -0.1101, -0.0424], + [-0.1032, -0.1169, 0.0777, ..., -0.0968, -0.0664, 0.1464]], + device='cuda:0'), grad: tensor([[ 1.3690e-07, -8.9407e-06, -1.1310e-05, ..., 2.1141e-07, + 0.0000e+00, -2.0474e-05], + [ 1.7388e-06, 1.2387e-07, 8.2981e-07, ..., 2.6301e-06, + 0.0000e+00, 6.1980e-07], + [ 1.5292e-06, 2.2398e-07, 2.3097e-07, ..., 2.4550e-06, + 0.0000e+00, 7.5949e-07], + ..., + [ 3.8091e-07, -4.2655e-07, 1.5786e-07, ..., 5.7276e-07, + 0.0000e+00, 3.6461e-07], + [-9.5740e-06, 1.2321e-06, 4.4564e-07, ..., -1.3642e-05, + 0.0000e+00, 1.1502e-07], + [ 1.4855e-07, 1.9576e-06, 4.6194e-06, ..., 4.9639e-07, + 0.0000e+00, 8.0168e-06]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0043, -0.0017, 0.0084, 0.0121, 0.0268, 0.0392, -0.0199, 0.0059, + -0.0360, -0.0190], device='cuda:0'), grad: tensor([-4.0472e-05, 6.4224e-06, 4.7386e-06, 3.1069e-06, -1.6671e-06, + 1.8135e-05, 1.8030e-05, 1.2983e-06, -2.7463e-05, 1.7866e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 217.16, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5452 re_mapping 0.0058 re_causal 0.0186 /// teacc 99.03 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0576, 0.1458, 0.0771, ..., -0.2182, -0.1199, 0.0250], + [ 0.1566, -0.0200, -0.0774, ..., 0.0927, 0.1203, -0.1478], + [-0.0986, -0.0577, -0.0543, ..., -0.0931, -0.0384, -0.0677], + ..., + [-0.0748, 0.1590, -0.1878, ..., -0.1518, -0.0015, -0.1664], + [ 0.0081, -0.1071, 0.0355, ..., 0.0673, -0.1102, -0.0427], + [-0.1037, -0.1174, 0.0776, ..., -0.0975, -0.0665, 0.1463]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 5.4482e-08, -9.7789e-09, ..., 1.0710e-08, + 0.0000e+00, -1.3504e-08], + [-3.9116e-08, 4.9826e-08, 3.2596e-09, ..., -5.8673e-08, + 0.0000e+00, 2.7940e-09], + [ 5.1223e-09, -4.1090e-06, 1.2573e-08, ..., 3.6787e-08, + 0.0000e+00, 1.3970e-09], + ..., + [ 1.2573e-08, 3.6769e-06, 6.9849e-09, ..., 4.1910e-08, + 0.0000e+00, 5.5879e-09], + [ 4.2375e-08, 4.2375e-08, 4.6566e-09, ..., 3.0268e-08, + 0.0000e+00, 1.9092e-08], + [ 9.7789e-09, 4.9360e-08, -6.2305e-07, ..., 1.8021e-07, + 0.0000e+00, -6.9616e-07]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0034, -0.0023, 0.0071, 0.0122, 0.0273, 0.0391, -0.0183, 0.0061, + -0.0363, -0.0195], device='cuda:0'), grad: tensor([ 1.4529e-07, 3.7253e-08, -1.0625e-05, 4.8801e-07, 9.3644e-07, + 5.9139e-08, -1.1316e-07, 9.7528e-06, 1.7043e-07, -8.2469e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 217.21, cls_loss 0.0015 cls_loss_mapping 0.0027 cls_loss_causal 0.5085 re_mapping 0.0060 re_causal 0.0176 /// teacc 99.08 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0579, 0.1453, 0.0758, ..., -0.2194, -0.1199, 0.0245], + [ 0.1575, -0.0205, -0.0776, ..., 0.0935, 0.1205, -0.1481], + [-0.0988, -0.0578, -0.0545, ..., -0.0933, -0.0384, -0.0679], + ..., + [-0.0750, 0.1596, -0.1880, ..., -0.1522, -0.0016, -0.1668], + [ 0.0081, -0.1072, 0.0358, ..., 0.0674, -0.1104, -0.0426], + [-0.1043, -0.1181, 0.0775, ..., -0.0980, -0.0665, 0.1463]], + device='cuda:0'), grad: tensor([[ 2.2817e-08, -1.2107e-08, 6.3796e-08, ..., 2.5611e-08, + 0.0000e+00, 5.3085e-08], + [-2.3935e-07, 2.7809e-06, -8.4285e-08, ..., -2.5379e-07, + 0.0000e+00, -2.8871e-08], + [ 5.4948e-08, 5.7295e-06, 3.6974e-07, ..., 7.0315e-08, + 0.0000e+00, 2.4494e-07], + ..., + [ 2.8405e-08, -9.2387e-06, 5.4017e-08, ..., 3.5996e-07, + 0.0000e+00, 2.5611e-08], + [ 2.4447e-07, 2.1607e-07, 2.3982e-07, ..., 2.2817e-07, + 0.0000e+00, 2.3609e-07], + [ 1.6950e-07, 2.8126e-07, -5.0059e-07, ..., 4.2049e-07, + 0.0000e+00, -7.1386e-07]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0040, -0.0019, 0.0070, 0.0121, 0.0277, 0.0393, -0.0184, 0.0063, + -0.0363, -0.0200], device='cuda:0'), grad: tensor([ 1.5646e-07, 3.5241e-06, 9.0152e-06, -7.4552e-07, -2.1216e-06, + 3.2550e-07, -4.2608e-07, -1.1712e-05, 1.5069e-06, 4.6939e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 217.24, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.5065 re_mapping 0.0059 re_causal 0.0176 /// teacc 98.98 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0580, 0.1455, 0.0761, ..., -0.2197, -0.1199, 0.0251], + [ 0.1574, -0.0206, -0.0789, ..., 0.0934, 0.1211, -0.1486], + [-0.0990, -0.0577, -0.0531, ..., -0.0935, -0.0387, -0.0675], + ..., + [-0.0751, 0.1602, -0.1878, ..., -0.1526, -0.0021, -0.1673], + [ 0.0085, -0.1074, 0.0362, ..., 0.0682, -0.1105, -0.0428], + [-0.1046, -0.1188, 0.0775, ..., -0.0990, -0.0665, 0.1465]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.6566e-10, 1.1828e-07, ..., 7.9162e-09, + 0.0000e+00, -3.2596e-08], + [-3.2596e-09, 1.2107e-08, 4.0047e-08, ..., 7.6368e-07, + 0.0000e+00, 1.7229e-08], + [ 4.1910e-09, 1.4435e-08, 3.9581e-08, ..., 2.5611e-08, + 9.3132e-10, 1.2107e-08], + ..., + [ 1.8626e-09, 4.7823e-07, 9.4436e-07, ..., 1.0012e-07, + 0.0000e+00, 8.3866e-07], + [-4.6566e-09, 4.1910e-08, 2.3469e-07, ..., 8.0559e-08, + 0.0000e+00, 1.4948e-07], + [ 1.3970e-09, -7.0967e-07, -8.7079e-07, ..., 1.7602e-07, + 0.0000e+00, -2.1365e-06]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0038, -0.0022, 0.0073, 0.0121, 0.0286, 0.0393, -0.0187, 0.0066, + -0.0360, -0.0208], device='cuda:0'), grad: tensor([ 1.8347e-07, 1.1437e-06, 1.2526e-07, -2.4084e-06, 1.0226e-06, + 2.9989e-07, 1.3597e-07, 3.4738e-06, 7.1805e-07, -4.7013e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 217.28, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5026 re_mapping 0.0062 re_causal 0.0173 /// teacc 99.04 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0580, 0.1459, 0.0757, ..., -0.2200, -0.1200, 0.0252], + [ 0.1579, -0.0213, -0.0799, ..., 0.0941, 0.1226, -0.1490], + [-0.0992, -0.0579, -0.0532, ..., -0.0936, -0.0392, -0.0678], + ..., + [-0.0755, 0.1610, -0.1874, ..., -0.1537, -0.0037, -0.1680], + [ 0.0087, -0.1075, 0.0365, ..., 0.0684, -0.1106, -0.0430], + [-0.1053, -0.1191, 0.0777, ..., -0.0996, -0.0667, 0.1467]], + device='cuda:0'), grad: tensor([[ 6.7055e-07, -7.5111e-07, 6.5193e-09, ..., 1.3215e-06, + 0.0000e+00, -1.6158e-07], + [ 4.4495e-05, 8.0559e-08, 1.7928e-08, ..., 7.9334e-05, + 0.0000e+00, 1.1869e-05], + [ 7.6368e-07, 3.4925e-08, 3.6089e-08, ..., 1.6834e-07, + 0.0000e+00, 2.2794e-07], + ..., + [ 8.5449e-08, 7.1479e-07, 6.0908e-07, ..., 2.0280e-07, + 0.0000e+00, 1.0561e-06], + [ 5.9232e-06, 3.5437e-07, 4.7777e-07, ..., 1.2480e-05, + 0.0000e+00, 2.3134e-06], + [ 1.0733e-07, -3.5670e-07, -6.2957e-07, ..., 7.3574e-07, + 0.0000e+00, -8.0839e-07]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0038, -0.0026, 0.0074, 0.0111, 0.0290, 0.0400, -0.0189, 0.0070, + -0.0359, -0.0213], device='cuda:0'), grad: tensor([ 1.9744e-06, 1.9920e-04, -1.6198e-05, -1.6508e-07, 2.5928e-06, + 5.9009e-05, -2.9635e-04, 3.3285e-06, 4.7743e-05, -7.9628e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 217.33, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.5260 re_mapping 0.0059 re_causal 0.0177 /// teacc 99.01 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0581, 0.1459, 0.0754, ..., -0.2209, -0.1200, 0.0248], + [ 0.1584, -0.0212, -0.0799, ..., 0.0943, 0.1233, -0.1493], + [-0.0993, -0.0580, -0.0533, ..., -0.0938, -0.0393, -0.0683], + ..., + [-0.0759, 0.1612, -0.1879, ..., -0.1545, -0.0045, -0.1688], + [ 0.0088, -0.1076, 0.0360, ..., 0.0684, -0.1107, -0.0440], + [-0.1058, -0.1193, 0.0793, ..., -0.0996, -0.0668, 0.1482]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, -5.1223e-09, 6.9849e-09, ..., 1.1176e-08, + 2.3283e-10, -1.4668e-08], + [-3.6787e-08, 6.7148e-07, 2.0256e-08, ..., -4.5402e-08, + -1.1642e-09, 2.3283e-09], + [ 4.3074e-08, 6.2399e-08, 2.5425e-07, ..., 6.6124e-08, + 2.3283e-10, 1.8626e-09], + ..., + [ 1.6065e-08, -1.0207e-06, 1.5087e-07, ..., 3.2363e-08, + 4.6566e-10, 1.0710e-08], + [ 1.3271e-08, 6.5891e-08, 1.3900e-07, ..., 3.1432e-08, + 0.0000e+00, 4.9127e-08], + [ 4.4238e-09, 1.7066e-07, 4.7497e-08, ..., 2.5053e-07, + 2.3283e-10, -6.7521e-09]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0040, -0.0022, 0.0074, 0.0111, 0.0289, 0.0403, -0.0194, 0.0067, + -0.0363, -0.0206], device='cuda:0'), grad: tensor([ 6.8452e-08, 1.2163e-06, 6.5612e-07, -2.9728e-06, -2.4564e-07, + 2.3167e-07, -1.9791e-07, -5.0291e-07, 8.5495e-07, 8.9873e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 217.31, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5305 re_mapping 0.0057 re_causal 0.0175 /// teacc 99.09 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.0582, 0.1460, 0.0744, ..., -0.2211, -0.1200, 0.0240], + [ 0.1592, -0.0212, -0.0800, ..., 0.0948, 0.1236, -0.1494], + [-0.0994, -0.0582, -0.0535, ..., -0.0939, -0.0395, -0.0685], + ..., + [-0.0760, 0.1616, -0.1886, ..., -0.1549, -0.0049, -0.1697], + [ 0.0086, -0.1076, 0.0359, ..., 0.0681, -0.1110, -0.0445], + [-0.1067, -0.1198, 0.0800, ..., -0.1000, -0.0668, 0.1489]], + device='cuda:0'), grad: tensor([[ 1.1711e-07, -2.0023e-08, 1.6321e-07, ..., 3.9535e-07, + 0.0000e+00, 5.3830e-07], + [ 3.0082e-07, 1.0594e-07, 1.2293e-07, ..., 1.7392e-07, + -9.3132e-10, 2.8149e-07], + [ 2.3702e-07, 2.1420e-08, 1.0408e-07, ..., 2.5798e-07, + 2.3283e-10, 2.6077e-07], + ..., + [ 1.6857e-07, -3.9628e-07, 4.7032e-08, ..., 1.5530e-07, + 2.3283e-10, 9.8022e-08], + [ 4.7982e-05, 2.5146e-08, 1.2510e-05, ..., 3.9101e-05, + 2.3283e-10, 3.2097e-05], + [ 4.9034e-07, 1.4575e-07, 2.0093e-07, ..., 4.8289e-07, + 2.3283e-10, 3.6089e-07]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0045, -0.0019, 0.0073, 0.0113, 0.0291, 0.0403, -0.0196, 0.0065, + -0.0367, -0.0205], device='cuda:0'), grad: tensor([ 1.1399e-06, 2.3097e-06, -6.9337e-07, -2.4915e-05, 1.6298e-07, + -1.2577e-04, 2.8014e-05, -3.2340e-07, 1.1843e-04, 1.7611e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 217.47, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5191 re_mapping 0.0059 re_causal 0.0176 /// teacc 99.05 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0583, 0.1463, 0.0746, ..., -0.2214, -0.1200, 0.0243], + [ 0.1586, -0.0222, -0.0823, ..., 0.0918, 0.1211, -0.1503], + [-0.0984, -0.0585, -0.0512, ..., -0.0911, -0.0373, -0.0687], + ..., + [-0.0761, 0.1625, -0.1887, ..., -0.1550, -0.0050, -0.1701], + [ 0.0085, -0.1077, 0.0356, ..., 0.0678, -0.1113, -0.0451], + [-0.1072, -0.1202, 0.0799, ..., -0.1027, -0.0665, 0.1482]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -6.4261e-08, 2.9802e-08, ..., 2.2352e-08, + 3.8184e-08, -1.7229e-08], + [-9.1735e-08, 6.8918e-08, 2.0955e-08, ..., -1.1409e-07, + 2.7940e-09, 1.6764e-08], + [ 6.6124e-08, 6.1002e-08, 6.7055e-08, ..., 3.4599e-07, + -2.5798e-07, 7.5437e-08], + ..., + [ 3.7719e-08, -2.6822e-07, 1.5832e-07, ..., 9.7323e-08, + 6.1467e-08, 2.5006e-07], + [-8.4937e-06, 4.5635e-08, 1.0571e-07, ..., -1.7956e-05, + 7.5903e-08, -4.2021e-06], + [ 1.0710e-08, 9.5926e-08, -3.4971e-07, ..., 1.7416e-07, + 2.8871e-08, -6.3144e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0042, -0.0051, 0.0099, 0.0113, 0.0316, 0.0411, -0.0205, 0.0070, + -0.0371, -0.0224], device='cuda:0'), grad: tensor([ 1.1129e-07, 1.2573e-07, -1.1213e-06, -2.1979e-07, 7.2177e-08, + 8.1137e-06, 2.1562e-05, 6.1048e-07, -2.8655e-05, -5.8347e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 217.48, cls_loss 0.0029 cls_loss_mapping 0.0025 cls_loss_causal 0.4983 re_mapping 0.0060 re_causal 0.0169 /// teacc 98.98 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0588, 0.1465, 0.0747, ..., -0.2231, -0.1200, 0.0246], + [ 0.1577, -0.0219, -0.0854, ..., 0.0887, 0.1181, -0.1537], + [-0.0963, -0.0584, -0.0481, ..., -0.0881, -0.0345, -0.0692], + ..., + [-0.0763, 0.1626, -0.1892, ..., -0.1558, -0.0050, -0.1703], + [ 0.0077, -0.1079, 0.0354, ..., 0.0672, -0.1118, -0.0454], + [-0.1082, -0.1205, 0.0803, ..., -0.1018, -0.0615, 0.1497]], + device='cuda:0'), grad: tensor([[ 1.7043e-07, -2.3060e-06, 4.6566e-09, ..., 6.6450e-07, + 2.3283e-09, -5.5600e-07], + [-1.1846e-06, 4.0419e-07, 1.0710e-08, ..., -1.0356e-06, + -1.2107e-06, 2.1001e-07], + [ 1.4342e-06, 1.7462e-07, 4.4238e-08, ..., 1.3430e-06, + 1.1157e-06, 3.5111e-07], + ..., + [ 1.7835e-07, 4.6892e-07, 1.8161e-08, ..., 3.5390e-07, + 1.7788e-07, 8.8708e-07], + [-2.5287e-05, 9.1363e-07, 2.3609e-07, ..., -3.4094e-05, + 2.3283e-09, -2.4036e-05], + [ 2.9895e-07, 1.0151e-07, 1.6345e-07, ..., 4.0326e-07, + 6.5193e-09, 3.0268e-07]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0043, -0.0081, 0.0128, 0.0114, 0.0314, 0.0411, -0.0205, 0.0068, + -0.0380, -0.0213], device='cuda:0'), grad: tensor([-3.6750e-06, -3.9674e-07, 2.6599e-06, -3.8333e-06, -7.2177e-08, + 1.3970e-05, 6.4492e-05, 2.4028e-06, -7.7188e-05, 1.6298e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 217.23, cls_loss 0.0020 cls_loss_mapping 0.0029 cls_loss_causal 0.5147 re_mapping 0.0058 re_causal 0.0169 /// teacc 98.94 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0590, 0.1468, 0.0749, ..., -0.2261, -0.1228, 0.0230], + [ 0.1576, -0.0221, -0.0856, ..., 0.0887, 0.1181, -0.1546], + [-0.0965, -0.0586, -0.0481, ..., -0.0882, -0.0345, -0.0699], + ..., + [-0.0769, 0.1625, -0.1904, ..., -0.1561, -0.0051, -0.1729], + [ 0.0086, -0.1072, 0.0357, ..., 0.0679, -0.1119, -0.0451], + [-0.1085, -0.1205, 0.0814, ..., -0.1025, -0.0612, 0.1505]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, 4.6566e-09, 5.5879e-09, ..., 6.0070e-08, + 0.0000e+00, 1.5832e-08], + [-5.5740e-07, 6.6590e-08, 5.5879e-09, ..., -1.1241e-06, + -2.7940e-09, 7.6368e-08], + [ 5.2154e-08, 5.4948e-08, 8.6147e-08, ..., 2.2585e-07, + 3.2596e-09, 2.0955e-08], + ..., + [ 6.9849e-08, -1.4529e-07, 1.5413e-07, ..., 2.3888e-07, + 2.3283e-09, 4.7572e-06], + [ 1.9092e-08, 1.0431e-07, 1.2573e-08, ..., 2.5239e-07, + 2.7940e-09, 2.5611e-07], + [ 2.2352e-08, -2.9709e-07, -5.3039e-07, ..., 1.2442e-05, + 4.6566e-10, -3.2485e-06]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0065, -0.0082, 0.0127, 0.0114, 0.0327, 0.0414, -0.0192, 0.0056, + -0.0366, -0.0214], device='cuda:0'), grad: tensor([ 1.4761e-07, -1.3076e-06, 6.6636e-07, -8.6613e-08, -9.4846e-06, + -8.1025e-08, -4.9546e-06, 1.0252e-05, 9.8161e-07, 3.8594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 216.62, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.5047 re_mapping 0.0059 re_causal 0.0172 /// teacc 99.02 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.0591, 0.1469, 0.0748, ..., -0.2263, -0.1228, 0.0231], + [ 0.1577, -0.0228, -0.0858, ..., 0.0887, 0.1182, -0.1546], + [-0.0966, -0.0587, -0.0481, ..., -0.0882, -0.0346, -0.0699], + ..., + [-0.0770, 0.1628, -0.1917, ..., -0.1564, -0.0051, -0.1730], + [ 0.0085, -0.1071, 0.0358, ..., 0.0677, -0.1121, -0.0455], + [-0.1087, -0.1204, 0.0823, ..., -0.1030, -0.0612, 0.1506]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.8626e-08, 1.7043e-07, ..., 2.1327e-07, + 0.0000e+00, 1.0524e-07], + [-1.4761e-07, 1.0300e-06, 8.3353e-08, ..., -2.1560e-07, + -5.1223e-09, 1.6904e-07], + [ 2.5611e-08, 1.8068e-07, 7.5903e-08, ..., 6.6124e-08, + 1.3970e-09, 1.1409e-07], + ..., + [ 3.1199e-08, -5.9716e-06, 2.9113e-06, ..., 1.1502e-07, + 2.3283e-09, 4.4852e-06], + [ 6.7521e-08, 3.5251e-07, 1.6810e-07, ..., -9.2089e-06, + 0.0000e+00, 1.9632e-06], + [ 1.8626e-08, 3.9823e-06, -7.2345e-06, ..., 4.2289e-05, + 9.3132e-10, -1.5348e-05]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0065, -0.0082, 0.0127, 0.0113, 0.0330, 0.0414, -0.0189, 0.0054, + -0.0365, -0.0213], device='cuda:0'), grad: tensor([ 3.3621e-07, 2.1905e-06, 7.5437e-07, -2.1281e-07, -7.6354e-05, + 8.5607e-06, 8.5123e-07, -6.0536e-06, -2.1532e-06, 7.1824e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 216.92, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5224 re_mapping 0.0055 re_causal 0.0169 /// teacc 99.06 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.0591, 0.1474, 0.0750, ..., -0.2264, -0.1228, 0.0233], + [ 0.1587, -0.0205, -0.0858, ..., 0.0888, 0.1182, -0.1548], + [-0.0967, -0.0590, -0.0481, ..., -0.0882, -0.0346, -0.0701], + ..., + [-0.0782, 0.1624, -0.1918, ..., -0.1591, -0.0052, -0.1732], + [ 0.0084, -0.1072, 0.0360, ..., 0.0678, -0.1128, -0.0459], + [-0.1096, -0.1214, 0.0827, ..., -0.1037, -0.0612, 0.1508]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, -7.8231e-08, -8.8476e-08, ..., 1.0245e-08, + 9.3132e-10, -2.5565e-07], + [-2.9383e-07, 3.3528e-08, 3.7253e-09, ..., -4.1910e-07, + -2.2817e-08, 2.7940e-09], + [ 1.3318e-07, 4.8429e-08, 2.3283e-09, ..., 2.2212e-07, + -6.0536e-09, 7.4506e-09], + ..., + [ 6.7987e-08, -3.6322e-07, 1.0245e-08, ..., 1.0710e-07, + 1.3970e-08, 1.2107e-08], + [-2.8405e-08, 2.0023e-08, 4.1910e-09, ..., -4.1910e-08, + 5.5879e-09, 1.6298e-08], + [ 2.5146e-08, 3.0687e-07, -4.0978e-08, ..., 1.7090e-07, + 2.3283e-09, 8.6613e-08]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0064, -0.0081, 0.0127, 0.0111, 0.0343, 0.0417, -0.0200, 0.0050, + -0.0366, -0.0218], device='cuda:0'), grad: tensor([-4.2375e-07, -7.7160e-07, 1.0477e-07, 3.2596e-08, 0.0000e+00, + 1.5227e-07, 1.8161e-07, -1.5041e-07, 7.8697e-08, 8.1956e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 216.54, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.4930 re_mapping 0.0055 re_causal 0.0160 /// teacc 99.03 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0592, 0.1480, 0.0733, ..., -0.2265, -0.1228, 0.0216], + [ 0.1596, -0.0203, -0.0857, ..., 0.0889, 0.1184, -0.1548], + [-0.0971, -0.0593, -0.0481, ..., -0.0882, -0.0347, -0.0704], + ..., + [-0.0786, 0.1628, -0.1920, ..., -0.1598, -0.0058, -0.1736], + [ 0.0084, -0.1073, 0.0359, ..., 0.0678, -0.1129, -0.0460], + [-0.1101, -0.1218, 0.0842, ..., -0.1038, -0.0612, 0.1519]], + device='cuda:0'), grad: tensor([[ 9.8720e-08, -8.2422e-08, 4.1910e-09, ..., 1.2666e-07, + 0.0000e+00, -1.4901e-08], + [-1.0282e-06, 3.5856e-08, 8.3819e-09, ..., -1.4063e-06, + 0.0000e+00, 1.1176e-08], + [ 2.6729e-07, 2.3283e-08, 1.2573e-08, ..., 3.6322e-07, + 4.6566e-10, 1.9092e-08], + ..., + [ 7.4692e-07, -1.4622e-07, 3.2596e-08, ..., 1.0179e-06, + 4.6566e-10, 4.4703e-08], + [ 7.0920e-07, 3.3993e-08, 1.0524e-07, ..., 8.4937e-07, + 9.3132e-10, 2.3004e-07], + [ 7.9628e-08, 5.6345e-08, -2.9942e-07, ..., 1.0943e-07, + 0.0000e+00, -4.8755e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0073, -0.0079, 0.0126, 0.0108, 0.0341, 0.0421, -0.0204, 0.0051, + -0.0367, -0.0211], device='cuda:0'), grad: tensor([ 1.1921e-07, -3.4031e-06, 1.0990e-06, -2.9337e-08, 2.9542e-06, + 1.5656e-06, -6.7167e-06, 2.4159e-06, 2.4382e-06, -4.2794e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 217.14, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.5020 re_mapping 0.0057 re_causal 0.0165 /// teacc 99.03 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0594, 0.1482, 0.0713, ..., -0.2269, -0.1228, 0.0198], + [ 0.1602, -0.0204, -0.0858, ..., 0.0890, 0.1184, -0.1549], + [-0.0973, -0.0594, -0.0481, ..., -0.0883, -0.0347, -0.0706], + ..., + [-0.0787, 0.1630, -0.1935, ..., -0.1600, -0.0063, -0.1739], + [ 0.0082, -0.1074, 0.0353, ..., 0.0676, -0.1130, -0.0465], + [-0.1105, -0.1219, 0.0856, ..., -0.1047, -0.0612, 0.1529]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -2.5425e-07, -2.9244e-07, ..., 2.7008e-08, + 0.0000e+00, -5.3644e-07], + [-2.1420e-08, 7.4506e-08, 1.3970e-08, ..., -2.5146e-08, + -1.8626e-09, 2.4214e-08], + [ 1.8626e-07, 4.5728e-07, 1.2387e-07, ..., 2.9057e-07, + 9.3132e-10, 1.3132e-07], + ..., + [ 2.3283e-08, -8.2608e-07, 5.3085e-08, ..., 5.6811e-08, + 9.3132e-10, 9.2201e-08], + [-1.7043e-07, 6.2399e-08, 1.4910e-06, ..., -2.1141e-07, + 0.0000e+00, 3.4273e-06], + [ 7.3574e-08, 3.9488e-07, -1.4240e-06, ..., 6.2492e-07, + 0.0000e+00, -3.1367e-06]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0088, -0.0079, 0.0125, 0.0109, 0.0351, 0.0422, -0.0203, 0.0046, + -0.0370, -0.0205], device='cuda:0'), grad: tensor([-2.6040e-06, 1.4249e-07, 1.6605e-06, 1.0179e-06, -4.1164e-07, + -1.1949e-06, -8.3819e-09, -9.9372e-07, 6.6608e-06, -4.2729e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 217.06, cls_loss 0.0022 cls_loss_mapping 0.0034 cls_loss_causal 0.5202 re_mapping 0.0056 re_causal 0.0163 /// teacc 98.91 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0595, 0.1483, 0.0718, ..., -0.2272, -0.1228, 0.0199], + [ 0.1603, -0.0203, -0.0863, ..., 0.0890, 0.1184, -0.1556], + [-0.0977, -0.0599, -0.0480, ..., -0.0883, -0.0347, -0.0711], + ..., + [-0.0791, 0.1639, -0.1938, ..., -0.1605, -0.0064, -0.1743], + [ 0.0091, -0.1077, 0.0362, ..., 0.0685, -0.1132, -0.0462], + [-0.1124, -0.1228, 0.0852, ..., -0.1069, -0.0612, 0.1519]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, -8.3819e-09, 1.6578e-07, ..., 2.4214e-08, + 2.7940e-09, 6.5193e-09], + [-8.4750e-08, 6.2399e-08, 7.8231e-08, ..., -1.4342e-07, + -3.3528e-08, 9.3132e-09], + [ 1.0803e-07, 6.0536e-08, 3.5763e-07, ..., 1.8906e-07, + 5.5879e-09, 3.7253e-08], + ..., + [ 4.0047e-08, -5.3458e-07, 4.4424e-07, ..., 7.2643e-08, + -1.3970e-08, 3.1665e-08], + [-5.4352e-06, 1.5832e-08, -1.2763e-05, ..., -1.0870e-05, + 1.0245e-08, -6.8173e-06], + [ 1.8533e-07, 3.6415e-07, 1.5302e-06, ..., 3.6601e-07, + 1.9558e-08, 2.4866e-07]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0086, -0.0080, 0.0127, 0.0107, 0.0376, 0.0420, -0.0203, 0.0048, + -0.0368, -0.0227], device='cuda:0'), grad: tensor([ 3.6787e-07, -5.8673e-08, 5.6811e-07, -1.5125e-06, 1.5832e-07, + 2.3916e-05, -3.8743e-06, 4.0233e-07, -2.3559e-05, 3.6154e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 216.75, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4844 re_mapping 0.0060 re_causal 0.0168 /// teacc 98.87 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.0596, 0.1493, 0.0725, ..., -0.2275, -0.1228, 0.0206], + [ 0.1616, -0.0204, -0.0864, ..., 0.0891, 0.1186, -0.1557], + [-0.0982, -0.0603, -0.0481, ..., -0.0884, -0.0348, -0.0714], + ..., + [-0.0789, 0.1686, -0.1908, ..., -0.1608, -0.0063, -0.1748], + [ 0.0091, -0.1078, 0.0362, ..., 0.0686, -0.1134, -0.0465], + [-0.1167, -0.1290, 0.0826, ..., -0.1074, -0.0613, 0.1521]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -1.2387e-07, -4.0047e-08, ..., 1.3970e-08, + 1.3970e-08, -7.4506e-08], + [ 3.6322e-08, 8.8476e-08, 3.3528e-08, ..., 5.5879e-09, + 5.2154e-08, 4.6566e-08], + [ 2.6077e-08, 2.2631e-07, 2.9802e-08, ..., 1.9558e-08, + 1.2387e-07, 2.4214e-08], + ..., + [ 8.3819e-09, -1.8105e-06, 7.8231e-08, ..., 9.3132e-09, + -8.2981e-07, 1.1455e-07], + [ 1.4249e-07, 8.1025e-08, 1.5926e-07, ..., 8.0094e-08, + 1.0803e-07, 1.7695e-07], + [ 3.3528e-08, 7.5437e-08, -1.7602e-07, ..., 3.2596e-08, + 2.7008e-08, -2.6450e-07]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0080, -0.0079, 0.0126, 0.0106, 0.0375, 0.0421, -0.0204, 0.0089, + -0.0370, -0.0266], device='cuda:0'), grad: tensor([-2.7567e-07, 4.3213e-07, 2.2724e-07, 5.4650e-06, 1.9670e-06, + -5.6587e-06, 1.7416e-07, -2.9486e-06, 7.4785e-07, -1.5087e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 217.68, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5144 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.99 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.0597, 0.1498, 0.0721, ..., -0.2278, -0.1228, 0.0192], + [ 0.1633, -0.0186, -0.0864, ..., 0.0891, 0.1186, -0.1558], + [-0.0984, -0.0599, -0.0481, ..., -0.0884, -0.0349, -0.0736], + ..., + [-0.0811, 0.1683, -0.1910, ..., -0.1621, -0.0060, -0.1767], + [ 0.0096, -0.1080, 0.0367, ..., 0.0696, -0.1142, -0.0464], + [-0.1175, -0.1290, 0.0829, ..., -0.1081, -0.0615, 0.1538]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, -1.4529e-07, -1.8906e-07, ..., 4.1910e-08, + 6.5193e-09, -1.5646e-07], + [-1.1828e-07, 6.7987e-08, 7.4506e-09, ..., -1.9185e-07, + -1.8626e-08, 2.6077e-08], + [ 6.4261e-08, 1.5553e-07, 1.0803e-07, ..., 1.0431e-07, + 1.2107e-08, 1.0803e-07], + ..., + [ 1.0058e-07, -2.0955e-07, 1.6764e-08, ..., 1.4342e-07, + 6.2399e-08, 2.3097e-07], + [ 5.3644e-07, 1.8626e-08, 5.3830e-07, ..., 7.3109e-07, + 1.1548e-07, 1.7788e-06], + [ 9.2201e-08, 7.6368e-08, -1.5832e-08, ..., 1.4901e-07, + 2.8871e-08, -1.3039e-08]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0090, -0.0075, 0.0126, 0.0106, 0.0370, 0.0419, -0.0201, 0.0083, + -0.0369, -0.0261], device='cuda:0'), grad: tensor([-2.6636e-06, -1.8254e-07, 1.9558e-06, 1.1483e-06, 7.0781e-08, + -3.1590e-05, 2.7820e-05, 5.9605e-08, 2.8759e-06, 4.7684e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 218.10, cls_loss 0.0019 cls_loss_mapping 0.0031 cls_loss_causal 0.4940 re_mapping 0.0056 re_causal 0.0161 /// teacc 98.94 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0598, 0.1510, 0.0723, ..., -0.2282, -0.1228, 0.0195], + [ 0.1639, -0.0189, -0.0865, ..., 0.0892, 0.1187, -0.1568], + [-0.0989, -0.0605, -0.0482, ..., -0.0884, -0.0350, -0.0738], + ..., + [-0.0826, 0.1684, -0.1911, ..., -0.1624, -0.0058, -0.1786], + [ 0.0108, -0.1070, 0.0359, ..., 0.0689, -0.1149, -0.0475], + [-0.1177, -0.1291, 0.0834, ..., -0.1082, -0.0610, 0.1550]], + device='cuda:0'), grad: tensor([[ 5.8766e-07, -1.5832e-08, 5.4017e-08, ..., 4.4052e-07, + 1.8626e-08, 1.0058e-07], + [-5.1260e-06, 1.6764e-08, 2.0489e-08, ..., -4.6566e-06, + -2.6450e-07, 7.6368e-08], + [ 1.2713e-06, 5.3085e-08, 1.7416e-07, ..., 1.2089e-06, + 1.7043e-07, 5.4017e-08], + ..., + [ 1.1120e-06, -5.0291e-08, 1.9744e-07, ..., 7.7300e-07, + 7.0781e-08, 1.5274e-06], + [ 9.8813e-07, 4.7497e-08, -4.5076e-07, ..., -6.6869e-07, + 2.8871e-08, -2.0489e-08], + [-1.4435e-07, 4.0978e-08, 2.9616e-07, ..., 1.1800e-06, + 3.1665e-08, -3.4831e-06]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0086, -0.0077, 0.0126, 0.0112, 0.0366, 0.0416, -0.0200, 0.0076, + -0.0349, -0.0259], device='cuda:0'), grad: tensor([ 1.7267e-06, -1.5303e-05, 4.9621e-06, -3.6377e-06, 4.6417e-06, + 3.3434e-07, 2.3767e-06, 5.4948e-06, 1.8617e-06, -2.5108e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 217.40, cls_loss 0.0018 cls_loss_mapping 0.0026 cls_loss_causal 0.4829 re_mapping 0.0056 re_causal 0.0158 /// teacc 98.94 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0601, 0.1515, 0.0722, ..., -0.2285, -0.1228, 0.0194], + [ 0.1644, -0.0199, -0.0862, ..., 0.0894, 0.1189, -0.1562], + [-0.0998, -0.0607, -0.0484, ..., -0.0885, -0.0352, -0.0763], + ..., + [-0.0824, 0.1687, -0.1912, ..., -0.1631, -0.0053, -0.1791], + [ 0.0109, -0.1073, 0.0362, ..., 0.0693, -0.1169, -0.0475], + [-0.1181, -0.1291, 0.0837, ..., -0.1082, -0.0613, 0.1560]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, -9.8720e-08, 1.4808e-07, ..., 3.0734e-08, + 2.1420e-08, 3.9395e-07], + [-2.1651e-05, 2.0582e-07, 1.9558e-08, ..., -3.2306e-05, + -2.4185e-05, 3.7253e-08], + [ 1.9923e-05, -3.1386e-07, -1.3784e-07, ..., 2.9758e-05, + 2.2277e-05, 1.8999e-07], + ..., + [ 1.0384e-06, -6.7614e-07, 6.0536e-08, ..., 1.5581e-06, + 1.1111e-06, 9.4064e-08], + [ 1.4715e-07, 1.0710e-07, 8.4564e-07, ..., 3.4086e-07, + 9.1270e-08, 1.4752e-06], + [ 5.7742e-08, 7.0687e-07, -1.2834e-06, ..., 9.4995e-08, + 8.6613e-08, -2.5295e-06]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0087, -0.0078, 0.0125, 0.0112, 0.0361, 0.0412, -0.0198, 0.0079, + -0.0350, -0.0258], device='cuda:0'), grad: tensor([ 6.1467e-07, -9.4473e-05, 7.0930e-05, 1.4389e-06, 1.0312e-05, + 5.4669e-07, 2.8275e-06, 5.6662e-06, 5.8040e-06, -3.8035e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 217.62, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5094 re_mapping 0.0058 re_causal 0.0163 /// teacc 98.94 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0599, 0.1530, 0.0729, ..., -0.2287, -0.1229, 0.0197], + [ 0.1647, -0.0201, -0.0863, ..., 0.0895, 0.1189, -0.1564], + [-0.1001, -0.0611, -0.0484, ..., -0.0886, -0.0353, -0.0764], + ..., + [-0.0824, 0.1688, -0.1913, ..., -0.1643, -0.0053, -0.1806], + [ 0.0107, -0.1076, 0.0363, ..., 0.0692, -0.1176, -0.0479], + [-0.1183, -0.1291, 0.0836, ..., -0.1086, -0.0614, 0.1568]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 8.3819e-09, 1.0245e-08, ..., 6.5193e-09, + 5.5879e-09, 9.3132e-10], + [-2.0768e-07, 6.8247e-05, 9.3132e-10, ..., -4.0140e-07, + 5.2243e-05, 0.0000e+00], + [ 1.7416e-07, 1.8999e-07, 1.9558e-08, ..., 3.8277e-07, + 3.4925e-07, 1.8626e-09], + ..., + [ 8.3819e-09, -7.4744e-05, 9.3132e-09, ..., 3.1665e-08, + -5.7399e-05, 2.7940e-09], + [ 1.1176e-08, 5.4985e-06, 7.4506e-09, ..., 1.8626e-08, + 4.2245e-06, 5.5879e-09], + [ 4.6566e-09, 1.9185e-07, -4.6566e-09, ..., 5.3085e-08, + 1.3132e-07, -1.9558e-08]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0079, -0.0077, 0.0125, 0.0108, 0.0356, 0.0427, -0.0207, 0.0079, + -0.0353, -0.0256], device='cuda:0'), grad: tensor([ 3.7812e-07, 3.7384e-04, -8.3148e-06, 2.3358e-06, 1.8254e-07, + 9.9745e-07, 1.0896e-06, -4.0436e-04, 3.1590e-05, 1.6131e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 217.74, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.5074 re_mapping 0.0057 re_causal 0.0163 /// teacc 99.01 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0601, 0.1533, 0.0731, ..., -0.2288, -0.1229, 0.0199], + [ 0.1652, -0.0205, -0.0864, ..., 0.0896, 0.1189, -0.1566], + [-0.1004, -0.0609, -0.0485, ..., -0.0887, -0.0353, -0.0772], + ..., + [-0.0827, 0.1692, -0.1913, ..., -0.1636, -0.0047, -0.1814], + [ 0.0109, -0.1091, 0.0364, ..., 0.0690, -0.1188, -0.0477], + [-0.1190, -0.1291, 0.0841, ..., -0.1089, -0.0614, 0.1574]], + device='cuda:0'), grad: tensor([[ 1.4529e-07, -1.5339e-06, -3.9712e-06, ..., 1.4901e-07, + -3.4180e-06, -5.3197e-06], + [-2.4214e-06, -2.6077e-08, 9.4995e-08, ..., -2.4997e-06, + -2.3562e-07, 8.1956e-08], + [ 1.0887e-06, 2.9057e-07, 3.7178e-06, ..., 1.1204e-06, + 4.3958e-07, 2.4904e-06], + ..., + [ 3.5018e-07, -4.5821e-07, 2.0582e-07, ..., 3.6880e-07, + 7.6368e-08, 1.5926e-07], + [ 5.4669e-07, 1.1176e-07, 2.2091e-06, ..., 5.6904e-07, + 1.3132e-07, 1.4277e-06], + [ 1.3225e-07, 5.0850e-07, -9.3803e-06, ..., 2.2724e-07, + 2.5146e-07, -5.4054e-06]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0078, -0.0078, 0.0125, 0.0107, 0.0350, 0.0427, -0.0208, 0.0082, + -0.0358, -0.0255], device='cuda:0'), grad: tensor([-2.2903e-05, -6.9961e-06, 1.1533e-05, 1.0030e-06, 3.4496e-06, + 1.8671e-05, 8.9593e-07, 7.5530e-07, 5.8673e-06, -1.2323e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 216.99, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4928 re_mapping 0.0053 re_causal 0.0158 /// teacc 99.11 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0604, 0.1536, 0.0737, ..., -0.2292, -0.1228, 0.0202], + [ 0.1655, -0.0204, -0.0866, ..., 0.0896, 0.1190, -0.1571], + [-0.1009, -0.0612, -0.0486, ..., -0.0888, -0.0354, -0.0775], + ..., + [-0.0830, 0.1693, -0.1914, ..., -0.1642, -0.0045, -0.1816], + [ 0.0114, -0.1092, 0.0369, ..., 0.0697, -0.1201, -0.0476], + [-0.1190, -0.1292, 0.0848, ..., -0.1086, -0.0615, 0.1580]], + device='cuda:0'), grad: tensor([[ 7.1712e-08, 1.2051e-06, -3.1758e-07, ..., -7.2364e-07, + 4.2841e-07, -1.3057e-06], + [-7.1377e-06, -5.2974e-06, 1.3970e-08, ..., -5.1335e-06, + 2.5146e-08, 9.3132e-09], + [ 4.0140e-07, 5.6252e-07, 9.0338e-08, ..., 3.1944e-07, + 1.4156e-07, 1.0245e-08], + ..., + [ 6.2585e-06, -6.7428e-07, 3.2037e-07, ..., 4.1947e-06, + -1.7844e-06, 8.1025e-08], + [ 1.5646e-07, 4.2468e-07, 1.2480e-07, ..., 3.1572e-07, + 1.2852e-07, 1.6671e-07], + [ 6.4727e-07, 2.6971e-06, 2.7940e-08, ..., 8.0466e-07, + 7.4040e-07, 8.8476e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0075, -0.0078, 0.0124, 0.0110, 0.0340, 0.0424, -0.0207, 0.0081, + -0.0358, -0.0251], device='cuda:0'), grad: tensor([-7.7300e-07, -3.8117e-05, 3.0696e-06, -2.9877e-06, 6.8638e-07, + 1.4370e-06, 2.0042e-06, 2.4036e-05, 2.1011e-06, 8.5756e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 216.99, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.4933 re_mapping 0.0056 re_causal 0.0155 /// teacc 99.10 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.0609, 0.1548, 0.0742, ..., -0.2297, -0.1228, 0.0207], + [ 0.1665, -0.0212, -0.0869, ..., 0.0899, 0.1194, -0.1582], + [-0.1017, -0.0613, -0.0486, ..., -0.0889, -0.0356, -0.0777], + ..., + [-0.0832, 0.1696, -0.1914, ..., -0.1657, -0.0053, -0.1818], + [ 0.0113, -0.1096, 0.0368, ..., 0.0697, -0.1215, -0.0479], + [-0.1195, -0.1292, 0.0851, ..., -0.1087, -0.0619, 0.1585]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -6.4261e-08, -1.0617e-07, ..., -3.1665e-08, + -2.7940e-08, -2.0396e-07], + [-1.7509e-07, -3.0734e-08, -6.5193e-09, ..., -3.7160e-07, + -4.5635e-08, 2.7940e-09], + [ 1.6764e-08, 1.4901e-08, 1.3039e-08, ..., -6.2399e-08, + 1.2107e-08, 1.9558e-08], + ..., + [ 8.1025e-08, -9.3132e-10, 2.7940e-08, ..., 1.7881e-07, + 2.5146e-08, 1.8626e-08], + [ 6.2399e-08, 3.7253e-08, 8.4750e-08, ..., 2.4308e-07, + 2.1420e-08, 1.3784e-07], + [ 9.3132e-09, 2.5146e-08, -1.0338e-07, ..., 2.8871e-08, + 7.4506e-09, -1.7136e-07]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0070, -0.0078, 0.0123, 0.0111, 0.0337, 0.0423, -0.0210, 0.0084, + -0.0361, -0.0250], device='cuda:0'), grad: tensor([-5.2899e-07, -6.1933e-07, -6.8825e-07, -1.3225e-07, 2.4121e-07, + 8.7544e-08, 4.0978e-08, 4.5169e-07, 1.2917e-06, -1.2480e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 217.25, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.5177 re_mapping 0.0054 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.0613, 0.1559, 0.0746, ..., -0.2300, -0.1229, 0.0210], + [ 0.1667, -0.0226, -0.0870, ..., 0.0900, 0.1196, -0.1583], + [-0.1020, -0.0615, -0.0487, ..., -0.0891, -0.0358, -0.0781], + ..., + [-0.0833, 0.1700, -0.1916, ..., -0.1658, -0.0053, -0.1826], + [ 0.0118, -0.1097, 0.0372, ..., 0.0704, -0.1221, -0.0479], + [-0.1199, -0.1292, 0.0854, ..., -0.1092, -0.0621, 0.1587]], + device='cuda:0'), grad: tensor([[ 5.4948e-08, -7.2550e-07, -2.1048e-07, ..., 1.4622e-07, + 9.3132e-10, -3.2410e-07], + [ 6.2287e-06, 3.7160e-07, 3.2783e-06, ..., 1.0237e-05, + -2.0489e-08, 3.6098e-06], + [ 1.1642e-07, 1.2107e-07, 1.7788e-07, ..., 3.3621e-07, + 2.0582e-07, 7.1712e-08], + ..., + [-1.5832e-08, -3.5856e-07, 2.6077e-08, ..., 3.9674e-07, + 2.0489e-08, 1.9558e-08], + [-8.1658e-06, 1.5087e-07, -4.1835e-06, ..., -1.3486e-05, + 1.8626e-08, -4.6119e-06], + [ 1.7881e-07, 1.9744e-07, 1.3597e-07, ..., 1.3644e-06, + 5.5879e-09, 1.6112e-07]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0068, -0.0076, 0.0120, 0.0120, 0.0336, 0.0413, -0.0211, 0.0086, + -0.0360, -0.0250], device='cuda:0'), grad: tensor([-1.7602e-06, 1.8716e-05, 2.5742e-06, -1.2228e-06, -4.3437e-06, + 2.0899e-06, 1.9670e-06, 7.9814e-07, -2.3156e-05, 4.3213e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 216.79, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4906 re_mapping 0.0054 re_causal 0.0156 /// teacc 98.94 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.0614, 0.1566, 0.0749, ..., -0.2302, -0.1229, 0.0211], + [ 0.1669, -0.0231, -0.0870, ..., 0.0901, 0.1197, -0.1584], + [-0.1024, -0.0616, -0.0487, ..., -0.0892, -0.0358, -0.0783], + ..., + [-0.0834, 0.1704, -0.1918, ..., -0.1661, -0.0051, -0.1853], + [ 0.0117, -0.1104, 0.0374, ..., 0.0706, -0.1226, -0.0480], + [-0.1202, -0.1294, 0.0855, ..., -0.1096, -0.0622, 0.1595]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 7.4506e-09, 0.0000e+00, ..., 3.9116e-08, + 6.5193e-09, -4.6566e-09], + [-2.7105e-05, 2.4214e-08, 9.3132e-10, ..., -1.4082e-05, + -4.2945e-05, 7.4506e-09], + [ 2.4542e-05, 3.4459e-08, 9.3132e-10, ..., 1.4089e-05, + 3.8922e-05, 9.3132e-10], + ..., + [ 2.2389e-06, -2.2892e-06, 1.7695e-08, ..., 1.8952e-06, + 3.5372e-06, 1.7323e-07], + [ 2.7940e-08, 2.8871e-08, 7.4506e-09, ..., 3.9823e-06, + 1.9558e-08, 2.8871e-08], + [ 1.6298e-07, 2.1867e-06, -2.3283e-08, ..., 1.0319e-05, + 2.3749e-07, -2.4959e-07]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0067, -0.0077, 0.0119, 0.0134, 0.0334, 0.0401, -0.0212, 0.0087, + -0.0362, -0.0249], device='cuda:0'), grad: tensor([ 1.0245e-07, -9.7692e-05, 9.0778e-05, 1.5181e-07, -3.3855e-05, + -3.1665e-08, 8.9332e-06, 6.0797e-06, 6.2250e-06, 1.9163e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 216.78, cls_loss 0.0016 cls_loss_mapping 0.0036 cls_loss_causal 0.5011 re_mapping 0.0060 re_causal 0.0166 /// teacc 99.01 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.0624, 0.1587, 0.0760, ..., -0.2308, -0.1231, 0.0225], + [ 0.1686, -0.0221, -0.0871, ..., 0.0907, 0.1199, -0.1581], + [-0.1028, -0.0606, -0.0486, ..., -0.0892, -0.0360, -0.0786], + ..., + [-0.0837, 0.1700, -0.1918, ..., -0.1683, -0.0064, -0.1862], + [ 0.0116, -0.1107, 0.0373, ..., 0.0703, -0.1228, -0.0481], + [-0.1208, -0.1294, 0.0854, ..., -0.1105, -0.0623, 0.1595]], + device='cuda:0'), grad: tensor([[ 3.0734e-08, 1.9558e-08, 4.6566e-09, ..., 4.4703e-08, + 9.3132e-10, 3.7253e-09], + [-3.2131e-07, -2.0303e-07, 1.0245e-08, ..., -5.6531e-07, + -7.8231e-08, 5.5879e-09], + [ 8.5682e-08, 6.5193e-08, 1.6764e-08, ..., 4.2748e-07, + 1.8626e-09, 2.7940e-09], + ..., + [ 1.8626e-07, 7.3574e-08, 5.4948e-08, ..., 3.7346e-07, + 4.6566e-08, 9.4064e-08], + [ 1.3448e-06, 6.2399e-08, 2.6785e-06, ..., 5.7928e-07, + 1.4901e-08, 1.4417e-06], + [ 1.0058e-07, 3.4459e-08, 5.3085e-08, ..., 1.2852e-07, + 9.3132e-09, -3.3062e-07]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0058, -0.0070, 0.0124, 0.0109, 0.0335, 0.0418, -0.0229, 0.0079, + -0.0364, -0.0249], device='cuda:0'), grad: tensor([ 4.4424e-07, -1.0617e-06, -3.5483e-07, -8.2478e-06, 1.5832e-08, + 1.9111e-06, -4.3586e-07, 1.0589e-06, 6.4895e-06, 1.7695e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 234---------------------------------------------------- +epoch 234, time 217.92, cls_loss 0.0014 cls_loss_mapping 0.0021 cls_loss_causal 0.5009 re_mapping 0.0057 re_causal 0.0163 /// teacc 99.12 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.0626, 0.1589, 0.0775, ..., -0.2311, -0.1231, 0.0236], + [ 0.1688, -0.0226, -0.0872, ..., 0.0908, 0.1199, -0.1585], + [-0.1030, -0.0606, -0.0487, ..., -0.0893, -0.0360, -0.0789], + ..., + [-0.0838, 0.1702, -0.1919, ..., -0.1689, -0.0058, -0.1874], + [ 0.0128, -0.1109, 0.0372, ..., 0.0724, -0.1220, -0.0490], + [-0.1213, -0.1294, 0.0852, ..., -0.1107, -0.0624, 0.1595]], + device='cuda:0'), grad: tensor([[ 6.3330e-08, 2.3283e-08, 1.3607e-06, ..., 6.4261e-08, + 9.3132e-10, 1.9558e-08], + [ 1.2107e-08, 6.5193e-09, 3.3528e-08, ..., 1.4901e-08, + 1.8626e-09, 3.7253e-09], + [ 1.3411e-07, 2.1420e-08, -1.8729e-06, ..., 1.4715e-07, + 8.3819e-09, 2.7008e-08], + ..., + [ 2.2352e-08, -1.2107e-08, 8.2888e-08, ..., 2.0489e-08, + 3.7253e-09, 2.3283e-08], + [-2.8778e-07, -3.6322e-08, 2.0675e-07, ..., -3.3993e-07, + 6.8918e-08, -3.5390e-08], + [ 1.4901e-08, 1.4901e-08, 2.7008e-08, ..., 2.4214e-08, + 1.8626e-09, -4.8429e-08]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0050, -0.0070, 0.0124, 0.0106, 0.0333, 0.0418, -0.0229, 0.0079, + -0.0358, -0.0250], device='cuda:0'), grad: tensor([ 1.4380e-05, 1.9372e-07, -2.1130e-05, -9.9242e-06, 4.9360e-08, + 1.2182e-05, -7.5437e-07, 3.9581e-07, 4.2729e-06, 2.8964e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 216.99, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4868 re_mapping 0.0056 re_causal 0.0156 /// teacc 99.05 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.0630, 0.1592, 0.0779, ..., -0.2314, -0.1232, 0.0238], + [ 0.1694, -0.0224, -0.0872, ..., 0.0910, 0.1201, -0.1586], + [-0.1033, -0.0607, -0.0487, ..., -0.0894, -0.0361, -0.0790], + ..., + [-0.0840, 0.1703, -0.1919, ..., -0.1693, -0.0063, -0.1880], + [ 0.0124, -0.1113, 0.0370, ..., 0.0718, -0.1231, -0.0493], + [-0.1215, -0.1294, 0.0852, ..., -0.1111, -0.0626, 0.1597]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 3.0603e-06, 1.8626e-09, ..., 9.3132e-09, + 0.0000e+00, -3.2783e-07], + [ 7.4506e-09, 1.4558e-05, 1.8626e-09, ..., 1.4901e-08, + 0.0000e+00, 1.0245e-08], + [ 2.7008e-08, 4.9546e-07, 1.7695e-08, ..., 7.0781e-08, + -1.8626e-09, 2.4214e-08], + ..., + [ 3.7253e-09, -2.3916e-05, 1.3039e-08, ..., 1.7695e-08, + 9.3132e-10, 7.9162e-08], + [-1.3039e-08, 4.6939e-07, -1.9558e-08, ..., -1.2852e-07, + 0.0000e+00, 6.4261e-08], + [ 6.5193e-09, 3.0212e-06, -8.9407e-08, ..., 2.0396e-07, + 0.0000e+00, -1.8906e-07]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0049, -0.0068, 0.0123, 0.0078, 0.0332, 0.0446, -0.0222, 0.0078, + -0.0363, -0.0251], device='cuda:0'), grad: tensor([ 5.5060e-06, 2.6748e-05, 1.0496e-06, 4.7591e-07, 2.2296e-06, + 1.3039e-06, 1.8254e-07, -4.3690e-05, 6.1374e-07, 5.6289e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 216.96, cls_loss 0.0018 cls_loss_mapping 0.0022 cls_loss_causal 0.5370 re_mapping 0.0053 re_causal 0.0160 /// teacc 99.10 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.0636, 0.1589, 0.0777, ..., -0.2321, -0.1232, 0.0237], + [ 0.1702, -0.0216, -0.0871, ..., 0.0911, 0.1202, -0.1585], + [-0.1039, -0.0619, -0.0487, ..., -0.0894, -0.0361, -0.0789], + ..., + [-0.0842, 0.1704, -0.1922, ..., -0.1698, -0.0063, -0.1889], + [ 0.0116, -0.1124, 0.0360, ..., 0.0711, -0.1235, -0.0499], + [-0.1222, -0.1294, 0.0854, ..., -0.1140, -0.0629, 0.1597]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 8.5682e-08, -7.4506e-09, ..., 7.4506e-09, + 0.0000e+00, -1.8999e-07], + [ 7.4506e-09, 4.2561e-07, 1.8626e-09, ..., 8.3819e-09, + 0.0000e+00, 2.0489e-08], + [ 2.3283e-08, 6.1691e-06, 5.5879e-09, ..., 5.7742e-08, + 0.0000e+00, 2.7940e-08], + ..., + [ 6.2399e-08, -7.3537e-06, 1.1362e-07, ..., 1.5367e-07, + 0.0000e+00, 8.8196e-07], + [-1.3225e-07, 9.4064e-08, 3.7253e-09, ..., -3.1851e-07, + 0.0000e+00, 2.5146e-08], + [ 8.3819e-09, -2.3283e-06, -1.0859e-06, ..., 4.8429e-08, + 0.0000e+00, -8.5086e-06]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0051, -0.0066, 0.0121, 0.0079, 0.0356, 0.0448, -0.0215, 0.0077, + -0.0374, -0.0261], device='cuda:0'), grad: tensor([-1.9744e-07, 7.7300e-07, 9.5963e-06, 1.8440e-07, 1.7568e-05, + 4.2841e-07, 2.9895e-07, -9.3803e-06, -4.4703e-07, -1.8805e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 216.70, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.5184 re_mapping 0.0056 re_causal 0.0160 /// teacc 99.12 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.0643, 0.1589, 0.0776, ..., -0.2333, -0.1232, 0.0236], + [ 0.1708, -0.0209, -0.0872, ..., 0.0913, 0.1201, -0.1587], + [-0.1042, -0.0627, -0.0487, ..., -0.0895, -0.0361, -0.0796], + ..., + [-0.0848, 0.1705, -0.1923, ..., -0.1715, -0.0059, -0.1902], + [ 0.0117, -0.1128, 0.0356, ..., 0.0712, -0.1237, -0.0509], + [-0.1227, -0.1294, 0.0857, ..., -0.1140, -0.0630, 0.1607]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -2.5146e-08, 9.3132e-10, ..., 1.5832e-08, + 0.0000e+00, -1.0245e-08], + [-5.2154e-08, 2.1420e-08, 7.4506e-09, ..., -1.7323e-07, + 0.0000e+00, 1.3970e-08], + [ 4.2841e-08, 7.4506e-09, 4.6566e-09, ..., 1.3877e-07, + -1.8626e-09, 5.5879e-09], + ..., + [ 7.4506e-09, -5.4017e-08, 2.7940e-09, ..., 2.0489e-08, + 0.0000e+00, 1.1176e-08], + [ 2.4214e-08, 1.3039e-08, 2.1420e-08, ..., 3.6322e-08, + 0.0000e+00, 6.1467e-08], + [ 8.3819e-09, 2.2352e-08, 3.7253e-09, ..., 2.5146e-08, + 0.0000e+00, -5.9605e-08]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0054, -0.0065, 0.0120, 0.0080, 0.0354, 0.0450, -0.0224, 0.0077, + -0.0377, -0.0259], device='cuda:0'), grad: tensor([-2.6077e-08, -2.4214e-07, 1.9744e-07, 6.7521e-07, 9.4064e-08, + -7.4599e-07, -3.5390e-08, -2.6077e-08, 1.4715e-07, -3.9116e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 216.69, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5046 re_mapping 0.0055 re_causal 0.0156 /// teacc 99.02 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.0647, 0.1596, 0.0778, ..., -0.2336, -0.1233, 0.0238], + [ 0.1718, -0.0219, -0.0870, ..., 0.0916, 0.1203, -0.1580], + [-0.1047, -0.0601, -0.0489, ..., -0.0896, -0.0362, -0.0808], + ..., + [-0.0847, 0.1703, -0.1923, ..., -0.1716, -0.0060, -0.1912], + [ 0.0124, -0.1130, 0.0353, ..., 0.0720, -0.1231, -0.0512], + [-0.1233, -0.1297, 0.0859, ..., -0.1143, -0.0630, 0.1613]], + device='cuda:0'), grad: tensor([[ 2.5239e-07, 1.4901e-08, 3.0827e-07, ..., 4.0513e-07, + 0.0000e+00, 3.7253e-09], + [ 8.9034e-07, 1.2107e-07, 1.3039e-07, ..., 5.4482e-07, + -9.3132e-10, 1.8626e-09], + [ 2.2799e-06, 5.9977e-07, 2.5295e-06, ..., 4.1015e-06, + 7.4506e-09, 2.7940e-09], + ..., + [ 4.0978e-08, -8.7358e-07, 2.2072e-07, ..., 1.0896e-07, + 0.0000e+00, 6.4261e-08], + [-3.2093e-06, -1.3039e-08, -3.7812e-06, ..., -1.0379e-05, + -9.3132e-09, 1.5367e-07], + [ 1.0896e-07, 3.6322e-08, 7.2177e-07, ..., 8.2050e-07, + 0.0000e+00, -2.4214e-08]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0053, -0.0064, 0.0126, 0.0087, 0.0354, 0.0441, -0.0228, 0.0074, + -0.0375, -0.0260], device='cuda:0'), grad: tensor([ 1.9073e-06, 2.8536e-06, 2.1964e-05, -9.5546e-05, 8.3260e-07, + 1.0413e-04, -5.9698e-07, -2.2538e-07, -3.9160e-05, 3.8128e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 217.40, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4959 re_mapping 0.0057 re_causal 0.0162 /// teacc 99.02 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.0651, 0.1597, 0.0778, ..., -0.2341, -0.1233, 0.0238], + [ 0.1719, -0.0220, -0.0871, ..., 0.0917, 0.1203, -0.1582], + [-0.1049, -0.0602, -0.0490, ..., -0.0896, -0.0363, -0.0807], + ..., + [-0.0847, 0.1704, -0.1925, ..., -0.1717, -0.0059, -0.1920], + [ 0.0129, -0.1134, 0.0349, ..., 0.0724, -0.1234, -0.0514], + [-0.1235, -0.1297, 0.0862, ..., -0.1143, -0.0631, 0.1618]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -2.5425e-07, -1.3784e-07, ..., 6.5193e-09, + 1.3970e-08, -3.1479e-07], + [ 1.0245e-08, 1.6764e-08, 1.9558e-07, ..., 2.8592e-07, + 4.5635e-08, 1.6298e-07], + [ 9.3132e-09, -4.0047e-08, 8.6892e-07, ..., 2.7008e-08, + 4.5411e-06, 1.7695e-08], + ..., + [ 4.6566e-09, 4.2841e-08, 2.6915e-07, ..., 1.7695e-08, + 1.3364e-06, 2.2352e-08], + [-2.5332e-07, 4.8429e-08, -1.5125e-06, ..., -2.8834e-06, + 5.4017e-08, -1.2238e-06], + [ 1.7975e-07, 8.1956e-08, 1.0198e-06, ..., 1.9260e-06, + 2.6077e-08, 7.3481e-07]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0053, -0.0064, 0.0126, 0.0087, 0.0353, 0.0441, -0.0230, 0.0073, + -0.0376, -0.0258], device='cuda:0'), grad: tensor([-7.7579e-07, 7.7393e-07, 3.1441e-05, -4.3392e-05, 8.2795e-07, + 1.6317e-06, 3.4273e-07, 9.7007e-06, -3.5129e-06, 2.9802e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 217.24, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4897 re_mapping 0.0053 re_causal 0.0151 /// teacc 98.99 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.0657, 0.1594, 0.0776, ..., -0.2349, -0.1233, 0.0236], + [ 0.1723, -0.0221, -0.0871, ..., 0.0918, 0.1203, -0.1582], + [-0.1053, -0.0605, -0.0491, ..., -0.0897, -0.0363, -0.0813], + ..., + [-0.0849, 0.1707, -0.1929, ..., -0.1723, -0.0059, -0.1926], + [ 0.0131, -0.1134, 0.0345, ..., 0.0726, -0.1241, -0.0516], + [-0.1239, -0.1299, 0.0868, ..., -0.1144, -0.0631, 0.1622]], + device='cuda:0'), grad: tensor([[ 2.5518e-07, 7.3574e-08, 3.6508e-07, ..., 4.4424e-07, + 0.0000e+00, 7.5903e-07], + [ 3.8296e-06, 1.1930e-06, 1.0803e-07, ..., 4.9248e-06, + -2.7940e-09, 1.2042e-06], + [ 7.4506e-08, 2.7940e-08, 1.2852e-07, ..., 1.6764e-07, + 9.3132e-10, 2.3842e-07], + ..., + [ 4.1910e-08, -4.3772e-08, 3.5856e-07, ..., 2.1048e-07, + 9.3132e-10, 6.3609e-07], + [ 4.0308e-06, 3.6880e-07, 6.0678e-05, ..., 2.1815e-05, + 0.0000e+00, 1.1122e-04], + [ 7.4506e-07, 1.4901e-08, -5.8800e-05, ..., -2.2098e-05, + 9.3132e-10, -1.2612e-04]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0056, -0.0064, 0.0125, 0.0088, 0.0352, 0.0440, -0.0227, 0.0073, + -0.0378, -0.0256], device='cuda:0'), grad: tensor([ 1.7304e-06, 7.8306e-06, 3.2410e-07, -5.0783e-05, 1.5432e-06, + 6.8903e-05, -3.3736e-05, 1.2880e-06, 1.9443e-04, -1.9145e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 217.09, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5291 re_mapping 0.0056 re_causal 0.0165 /// teacc 99.10 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.0661, 0.1593, 0.0775, ..., -0.2354, -0.1233, 0.0235], + [ 0.1727, -0.0222, -0.0872, ..., 0.0918, 0.1204, -0.1583], + [-0.1056, -0.0606, -0.0490, ..., -0.0897, -0.0363, -0.0805], + ..., + [-0.0850, 0.1708, -0.1935, ..., -0.1727, -0.0059, -0.1934], + [ 0.0138, -0.1132, 0.0336, ..., 0.0735, -0.1245, -0.0519], + [-0.1245, -0.1299, 0.0882, ..., -0.1141, -0.0632, 0.1641]], + device='cuda:0'), grad: tensor([[ 5.6811e-08, -3.8045e-07, -1.8161e-08, ..., 4.3772e-08, + 0.0000e+00, -2.3982e-07], + [-2.0396e-07, 4.0513e-08, -1.8161e-08, ..., -2.4680e-07, + 0.0000e+00, -1.7229e-08], + [ 6.3796e-08, -1.9744e-07, 2.3283e-08, ..., 7.1246e-08, + -4.6566e-10, 2.6077e-08], + ..., + [ 2.4680e-08, 2.1188e-07, 3.1199e-08, ..., 7.8697e-08, + 0.0000e+00, 4.3306e-08], + [ 1.4482e-07, 2.3516e-07, 3.3528e-08, ..., 1.8580e-07, + 0.0000e+00, 1.6252e-07], + [ 2.8871e-08, 1.2433e-07, -7.5763e-07, ..., 1.2154e-07, + 0.0000e+00, -1.2517e-06]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0058, -0.0064, 0.0125, 0.0088, 0.0340, 0.0439, -0.0229, 0.0072, + -0.0378, -0.0250], device='cuda:0'), grad: tensor([-7.0827e-07, -3.8417e-07, -4.4964e-06, -2.2491e-07, 2.7008e-06, + 3.0966e-07, -3.9069e-07, 4.3176e-06, 1.4119e-06, -2.5369e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 217.17, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.5190 re_mapping 0.0057 re_causal 0.0163 /// teacc 98.96 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.0669, 0.1590, 0.0774, ..., -0.2364, -0.1234, 0.0234], + [ 0.1736, -0.0220, -0.0872, ..., 0.0918, 0.1209, -0.1584], + [-0.1061, -0.0610, -0.0491, ..., -0.0898, -0.0365, -0.0807], + ..., + [-0.0858, 0.1710, -0.1939, ..., -0.1729, -0.0081, -0.1947], + [ 0.0139, -0.1134, 0.0335, ..., 0.0737, -0.1246, -0.0518], + [-0.1248, -0.1299, 0.0887, ..., -0.1139, -0.0633, 0.1650]], + device='cuda:0'), grad: tensor([[ 1.3784e-06, 2.1234e-06, -1.3597e-07, ..., 2.6841e-06, + 4.6566e-10, 1.5190e-06], + [-1.9558e-08, 2.3283e-08, 7.9162e-09, ..., -1.3504e-08, + -4.1910e-09, 2.4214e-08], + [ 5.7276e-08, 7.5903e-08, 8.3819e-09, ..., 9.9652e-08, + -4.1910e-09, 6.0536e-08], + ..., + [ 1.5832e-08, 9.3132e-10, 1.2247e-07, ..., 1.7229e-08, + 5.1223e-09, 1.6298e-07], + [ 1.2340e-07, 1.9837e-07, 1.4715e-07, ..., 1.9092e-07, + 1.8626e-09, 3.5390e-07], + [ 5.4482e-08, 7.2643e-08, -9.7323e-08, ..., 1.5227e-07, + 1.8626e-09, -7.8697e-08]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0061, -0.0059, 0.0124, 0.0090, 0.0333, 0.0439, -0.0227, 0.0066, + -0.0379, -0.0246], device='cuda:0'), grad: tensor([ 8.8811e-06, 3.8184e-08, 2.3702e-07, 1.2107e-08, 8.3679e-07, + 8.3353e-08, -1.1668e-05, 4.9546e-07, 9.8906e-07, 8.8010e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 216.89, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.5106 re_mapping 0.0057 re_causal 0.0160 /// teacc 98.96 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.0674, 0.1593, 0.0774, ..., -0.2367, -0.1235, 0.0235], + [ 0.1739, -0.0224, -0.0873, ..., 0.0919, 0.1210, -0.1588], + [-0.1066, -0.0611, -0.0492, ..., -0.0898, -0.0365, -0.0808], + ..., + [-0.0858, 0.1712, -0.1940, ..., -0.1732, -0.0082, -0.1949], + [ 0.0138, -0.1136, 0.0330, ..., 0.0737, -0.1251, -0.0522], + [-0.1250, -0.1300, 0.0890, ..., -0.1145, -0.0635, 0.1654]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -6.1467e-08, -1.8626e-09, ..., 5.5879e-09, + 0.0000e+00, -1.3039e-08], + [ 5.5507e-07, 6.5193e-08, 1.0245e-08, ..., 4.7311e-07, + -9.3132e-10, 4.9360e-08], + [ 2.5146e-08, 1.6764e-08, 6.5193e-09, ..., 2.3283e-08, + -9.3132e-10, 6.5193e-09], + ..., + [ 4.2841e-08, -1.5087e-07, 3.9116e-08, ..., 3.8184e-08, + 0.0000e+00, 6.0536e-08], + [-8.2608e-07, 2.3283e-08, 9.5926e-08, ..., -7.0874e-07, + 0.0000e+00, 1.2573e-07], + [ 4.8429e-08, 7.5437e-08, -2.3656e-07, ..., 4.3772e-08, + 9.3132e-10, -4.9546e-07]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0060, -0.0059, 0.0123, 0.0089, 0.0337, 0.0440, -0.0227, 0.0067, + -0.0381, -0.0248], device='cuda:0'), grad: tensor([-2.5798e-07, 1.2284e-06, 7.7300e-08, 9.3784e-07, 5.8487e-07, + -1.0412e-06, 2.8871e-07, 6.9849e-08, -1.1967e-06, -6.9197e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 217.16, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4824 re_mapping 0.0056 re_causal 0.0158 /// teacc 99.09 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.0677, 0.1596, 0.0773, ..., -0.2370, -0.1236, 0.0234], + [ 0.1748, -0.0225, -0.0872, ..., 0.0921, 0.1211, -0.1585], + [-0.1075, -0.0613, -0.0493, ..., -0.0899, -0.0367, -0.0810], + ..., + [-0.0860, 0.1713, -0.1942, ..., -0.1737, -0.0082, -0.1957], + [ 0.0133, -0.1140, 0.0323, ..., 0.0730, -0.1254, -0.0527], + [-0.1256, -0.1300, 0.0893, ..., -0.1173, -0.0637, 0.1649]], + device='cuda:0'), grad: tensor([[ 4.0047e-08, 1.3690e-07, 6.5193e-09, ..., 4.3772e-08, + 8.0094e-08, -9.3132e-10], + [-8.2888e-08, 1.9092e-07, 6.5193e-09, ..., 4.6566e-08, + 6.6124e-08, 7.4506e-09], + [ 6.7987e-08, 6.4261e-08, 8.3819e-09, ..., 8.1956e-08, + 3.4459e-08, 3.7253e-09], + ..., + [ 1.4901e-08, -1.1828e-06, 1.2107e-08, ..., 1.0058e-07, + -4.2748e-07, 1.3039e-08], + [ 8.6799e-07, 1.4156e-07, 2.9802e-07, ..., 6.9570e-07, + 4.9919e-07, 3.9861e-07], + [ 6.5193e-09, 4.8801e-07, -2.2352e-08, ..., 1.8934e-06, + 1.8161e-07, -1.0896e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0061, -0.0057, 0.0122, 0.0088, 0.0363, 0.0442, -0.0223, 0.0066, + -0.0386, -0.0265], device='cuda:0'), grad: tensor([ 3.1572e-07, 5.0012e-07, -8.0094e-07, 3.5297e-07, -3.9749e-06, + -1.5236e-06, -1.7695e-08, -1.7723e-06, 2.6673e-06, 4.2692e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 216.47, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.4911 re_mapping 0.0055 re_causal 0.0152 /// teacc 98.95 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.0678, 0.1599, 0.0774, ..., -0.2373, -0.1236, 0.0235], + [ 0.1773, -0.0232, -0.0872, ..., 0.0930, 0.1240, -0.1587], + [-0.1113, -0.0621, -0.0493, ..., -0.0909, -0.0393, -0.0812], + ..., + [-0.0863, 0.1719, -0.1942, ..., -0.1741, -0.0099, -0.1960], + [ 0.0132, -0.1142, 0.0322, ..., 0.0731, -0.1253, -0.0530], + [-0.1256, -0.1300, 0.0902, ..., -0.1164, -0.0640, 0.1675]], + device='cuda:0'), grad: tensor([[ 1.0524e-07, -1.1176e-08, -6.5193e-09, ..., 1.0151e-07, + 1.7695e-08, -8.3819e-09], + [-1.3616e-06, 5.5879e-09, 9.5926e-08, ..., -1.1949e-06, + -3.0361e-07, 9.0338e-08], + [ 1.2079e-06, 2.7940e-09, 2.7940e-09, ..., 1.4268e-06, + 1.0896e-07, 1.0245e-08], + ..., + [ 1.8161e-07, -1.7695e-08, 3.2596e-08, ..., 2.0396e-07, + 2.7008e-08, 7.7300e-08], + [-2.5295e-06, 2.7940e-09, -7.1339e-07, ..., -3.0678e-06, + 7.1712e-08, -6.8359e-07], + [ 2.3376e-07, 9.3132e-09, -4.8243e-07, ..., 3.8929e-07, + 4.0047e-08, -1.1781e-06]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0061, -0.0049, 0.0111, 0.0087, 0.0351, 0.0440, -0.0220, 0.0069, + -0.0388, -0.0255], device='cuda:0'), grad: tensor([ 2.3469e-07, -3.0808e-06, 4.4666e-06, 3.9209e-07, 1.5646e-06, + 3.8147e-06, 1.8030e-06, 6.8638e-07, -8.4490e-06, -1.4678e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 216.55, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.5105 re_mapping 0.0057 re_causal 0.0164 /// teacc 99.03 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.0687, 0.1592, 0.0765, ..., -0.2384, -0.1237, 0.0228], + [ 0.1798, -0.0208, -0.0873, ..., 0.0934, 0.1248, -0.1589], + [-0.1136, -0.0644, -0.0494, ..., -0.0911, -0.0398, -0.0814], + ..., + [-0.0874, 0.1719, -0.1943, ..., -0.1753, -0.0114, -0.1963], + [ 0.0136, -0.1143, 0.0325, ..., 0.0736, -0.1259, -0.0528], + [-0.1263, -0.1301, 0.0907, ..., -0.1165, -0.0645, 0.1683]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, 1.0245e-08, -5.5879e-09, ..., 3.5390e-08, + 1.0245e-08, -7.4506e-09], + [-2.0489e-08, 3.7253e-09, 2.7940e-09, ..., -5.9605e-08, + 0.0000e+00, 2.7940e-09], + [ 1.4901e-08, 4.6566e-09, 3.2596e-08, ..., 3.5390e-08, + 0.0000e+00, 3.7253e-09], + ..., + [ 5.5879e-09, -7.4506e-09, 4.6566e-09, ..., 1.3039e-08, + 0.0000e+00, 1.8626e-09], + [ 2.2352e-08, 3.7253e-09, 2.4214e-08, ..., -4.6566e-09, + 0.0000e+00, 3.3528e-08], + [ 2.2352e-08, 5.5879e-09, 1.8626e-09, ..., 5.8673e-08, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0068, -0.0038, 0.0103, 0.0093, 0.0348, 0.0434, -0.0213, 0.0066, + -0.0386, -0.0254], device='cuda:0'), grad: tensor([ 4.1910e-08, -1.2573e-07, 2.0489e-07, 8.9128e-07, 7.4506e-09, + -1.2610e-06, 9.3132e-10, 5.8673e-08, 6.3330e-08, 1.1083e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 216.59, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.5057 re_mapping 0.0056 re_causal 0.0167 /// teacc 99.09 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.0694, 0.1594, 0.0763, ..., -0.2392, -0.1237, 0.0228], + [ 0.1792, -0.0226, -0.0872, ..., 0.0935, 0.1249, -0.1589], + [-0.1139, -0.0645, -0.0494, ..., -0.0912, -0.0398, -0.0819], + ..., + [-0.0858, 0.1729, -0.1944, ..., -0.1755, -0.0114, -0.1966], + [ 0.0135, -0.1145, 0.0324, ..., 0.0736, -0.1261, -0.0531], + [-0.1267, -0.1301, 0.0909, ..., -0.1165, -0.0647, 0.1686]], + device='cuda:0'), grad: tensor([[ 3.2037e-07, -1.8626e-09, 2.9802e-07, ..., 3.8929e-07, + 9.3132e-10, 2.0303e-07], + [-1.1176e-08, 1.8440e-07, 1.6764e-08, ..., 5.5879e-09, + -1.6764e-08, 1.2107e-08], + [ 4.7497e-08, 3.1386e-07, 2.8871e-08, ..., 6.1467e-08, + 3.7253e-09, 2.1420e-08], + ..., + [ 3.4459e-08, -8.3540e-07, 2.3283e-08, ..., 1.7136e-07, + -5.5879e-09, 2.0489e-08], + [-1.5711e-06, 2.5518e-07, -1.3327e-06, ..., -1.9018e-06, + 5.5879e-09, -9.0990e-07], + [ 4.4890e-07, 9.9652e-08, 2.1327e-07, ..., 6.2957e-07, + 9.3132e-09, 9.0338e-08]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0069, -0.0045, 0.0102, 0.0093, 0.0347, 0.0433, -0.0204, 0.0075, + -0.0388, -0.0254], device='cuda:0'), grad: tensor([ 1.7770e-06, 3.2131e-07, 7.0687e-07, 2.3656e-06, -2.4028e-07, + 1.5516e-06, 1.4901e-07, -1.0431e-06, -7.9423e-06, 2.3562e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 217.40, cls_loss 0.0017 cls_loss_mapping 0.0021 cls_loss_causal 0.5201 re_mapping 0.0053 re_causal 0.0151 /// teacc 99.06 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.0697, 0.1601, 0.0763, ..., -0.2399, -0.1237, 0.0228], + [ 0.1794, -0.0229, -0.0874, ..., 0.0935, 0.1249, -0.1595], + [-0.1140, -0.0646, -0.0496, ..., -0.0912, -0.0399, -0.0822], + ..., + [-0.0860, 0.1729, -0.1944, ..., -0.1767, -0.0114, -0.1977], + [ 0.0135, -0.1148, 0.0322, ..., 0.0739, -0.1264, -0.0537], + [-0.1261, -0.1301, 0.0920, ..., -0.1172, -0.0648, 0.1703]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 5.5879e-09, 1.3039e-08, ..., 1.7695e-08, + 2.7940e-09, 2.7940e-09], + [-3.3587e-05, -3.0145e-05, 1.3039e-08, ..., -4.3005e-05, + -3.0734e-08, 1.8626e-09], + [ 3.7253e-08, 3.3528e-08, 7.5437e-08, ..., 4.2841e-08, + 1.1176e-08, 1.8626e-09], + ..., + [ 2.8029e-05, 2.5049e-05, 7.2643e-08, ..., 3.5971e-05, + 8.3819e-09, 2.9802e-08], + [ 2.4680e-07, 2.4401e-07, 1.2107e-08, ..., 3.0547e-07, + 9.3132e-09, -1.8626e-09], + [ 3.1739e-06, 2.9132e-06, -1.6671e-07, ..., 4.0941e-06, + 4.6566e-09, -3.5577e-07]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0068, -0.0046, 0.0102, 0.0093, 0.0349, 0.0433, -0.0211, 0.0073, + -0.0388, -0.0251], device='cuda:0'), grad: tensor([ 3.1386e-07, -1.0931e-04, -1.4342e-06, -2.2501e-06, 7.2122e-06, + 8.1118e-07, 9.8068e-07, 9.2387e-05, 1.2508e-06, 1.0066e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 217.34, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5086 re_mapping 0.0052 re_causal 0.0152 /// teacc 99.01 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.0706, 0.1596, 0.0761, ..., -0.2408, -0.1242, 0.0227], + [ 0.1802, -0.0233, -0.0875, ..., 0.0936, 0.1255, -0.1597], + [-0.1142, -0.0647, -0.0497, ..., -0.0913, -0.0400, -0.0826], + ..., + [-0.0868, 0.1736, -0.1945, ..., -0.1772, -0.0136, -0.1979], + [ 0.0135, -0.1154, 0.0325, ..., 0.0742, -0.1267, -0.0540], + [-0.1269, -0.1303, 0.0923, ..., -0.1172, -0.0654, 0.1708]], + device='cuda:0'), grad: tensor([[ 8.0373e-07, 2.5090e-06, 1.1055e-06, ..., 2.3954e-06, + 9.3132e-10, 7.5996e-07], + [ 5.5879e-09, 9.3132e-09, 5.8673e-08, ..., 1.5832e-08, + -1.6764e-08, 2.6077e-08], + [ 3.0734e-08, 7.4506e-09, -6.2399e-07, ..., 8.6613e-08, + 3.7253e-09, 8.3819e-09], + ..., + [ 2.5146e-08, -9.3132e-09, 2.4214e-08, ..., 5.0291e-08, + 5.5879e-09, 1.6764e-08], + [-2.2613e-06, -3.2783e-07, -1.9372e-06, ..., -2.0452e-06, + 2.7940e-09, -1.2731e-06], + [ 4.0419e-07, 6.9849e-08, 3.5204e-07, ..., 3.7719e-07, + 2.7940e-09, 1.7602e-07]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0074, -0.0045, 0.0101, 0.0094, 0.0347, 0.0433, -0.0210, 0.0075, + -0.0388, -0.0251], device='cuda:0'), grad: tensor([ 1.1213e-05, 2.5798e-07, -3.9265e-06, 3.4906e-06, -2.0582e-07, + 1.7053e-06, -7.2569e-06, 2.1979e-07, -6.9179e-06, 1.4063e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 217.23, cls_loss 0.0014 cls_loss_mapping 0.0026 cls_loss_causal 0.5175 re_mapping 0.0055 re_causal 0.0159 /// teacc 99.04 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.0711, 0.1599, 0.0760, ..., -0.2414, -0.1243, 0.0226], + [ 0.1804, -0.0236, -0.0880, ..., 0.0937, 0.1261, -0.1599], + [-0.1153, -0.0648, -0.0499, ..., -0.0915, -0.0406, -0.0829], + ..., + [-0.0868, 0.1745, -0.1945, ..., -0.1763, -0.0139, -0.1984], + [ 0.0133, -0.1177, 0.0326, ..., 0.0736, -0.1284, -0.0539], + [-0.1281, -0.1306, 0.0930, ..., -0.1171, -0.0658, 0.1720]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 1.0245e-08, 1.3970e-08, ..., 1.0245e-08, + 0.0000e+00, -5.5879e-09], + [ 5.7369e-07, 4.9174e-06, 1.4808e-07, ..., 1.0747e-06, + 0.0000e+00, 1.5832e-07], + [ 1.5832e-08, 4.1910e-08, 5.5879e-09, ..., 9.9652e-08, + 0.0000e+00, 1.7695e-08], + ..., + [-1.0403e-06, -9.9316e-06, 1.0245e-08, ..., 1.5553e-07, + 0.0000e+00, 1.3970e-08], + [-2.6077e-08, 6.2399e-08, 1.9558e-08, ..., -1.4957e-06, + 0.0000e+00, 2.3283e-08], + [ 4.9081e-07, 4.6901e-06, -4.8243e-07, ..., 1.3970e-08, + 0.0000e+00, -5.3924e-07]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0075, -0.0044, 0.0097, 0.0094, 0.0341, 0.0433, -0.0203, 0.0081, + -0.0400, -0.0247], device='cuda:0'), grad: tensor([ 7.4506e-08, 9.9093e-06, 2.3283e-07, 1.0431e-07, 5.0478e-07, + 8.7731e-07, -2.5891e-07, -1.4395e-05, -3.0696e-06, 6.0126e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 216.99, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5002 re_mapping 0.0052 re_causal 0.0150 /// teacc 99.05 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.0714, 0.1604, 0.0753, ..., -0.2418, -0.1245, 0.0219], + [ 0.1822, -0.0239, -0.0882, ..., 0.0939, 0.1279, -0.1600], + [-0.1173, -0.0648, -0.0497, ..., -0.0918, -0.0423, -0.0836], + ..., + [-0.0870, 0.1750, -0.1946, ..., -0.1763, -0.0145, -0.1987], + [ 0.0135, -0.1180, 0.0328, ..., 0.0741, -0.1294, -0.0537], + [-0.1284, -0.1309, 0.0936, ..., -0.1173, -0.0660, 0.1727]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -5.0478e-07, -4.3772e-08, ..., 7.4506e-09, + -9.3132e-09, -9.1735e-07], + [-2.0489e-08, 1.2666e-07, 2.0489e-08, ..., -2.8871e-08, + -3.7253e-08, 1.1269e-07], + [ 4.9360e-08, 3.3528e-08, 7.4506e-09, ..., 6.3330e-08, + -5.3085e-08, 3.7253e-08], + ..., + [ 3.4180e-07, -1.1176e-07, 1.6764e-07, ..., 2.7940e-08, + 8.1956e-08, 9.8068e-07], + [-2.8871e-08, 1.5832e-08, -3.9116e-08, ..., -7.1712e-08, + 3.7253e-09, 1.9558e-08], + [-5.1316e-07, 9.4995e-08, -2.5332e-07, ..., 6.6124e-08, + 5.5879e-09, -1.5320e-06]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0079, -0.0036, 0.0089, 0.0095, 0.0341, 0.0432, -0.0201, 0.0082, + -0.0401, -0.0248], device='cuda:0'), grad: tensor([-3.8147e-06, 4.9267e-07, -8.6240e-07, 1.9558e-07, 2.0638e-06, + 1.7416e-07, 2.2985e-06, 4.0196e-06, -7.1712e-08, -4.4890e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 218.08, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4895 re_mapping 0.0052 re_causal 0.0150 /// teacc 98.93 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.0719, 0.1627, 0.0754, ..., -0.2424, -0.1246, 0.0220], + [ 0.1830, -0.0238, -0.0887, ..., 0.0942, 0.1286, -0.1603], + [-0.1178, -0.0649, -0.0492, ..., -0.0920, -0.0426, -0.0836], + ..., + [-0.0878, 0.1750, -0.1948, ..., -0.1770, -0.0162, -0.1994], + [ 0.0132, -0.1183, 0.0328, ..., 0.0741, -0.1303, -0.0542], + [-0.1294, -0.1310, 0.0940, ..., -0.1174, -0.0664, 0.1734]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.9595e-06, 5.5879e-09, ..., 4.6566e-09, + -9.3132e-10, 2.7940e-09], + [ 7.5437e-08, 3.6322e-07, 6.9849e-08, ..., 6.9849e-08, + -4.6566e-09, 1.1176e-08], + [ 9.2480e-07, 8.5682e-08, 7.4785e-07, ..., 8.6334e-07, + 3.7253e-09, 9.3132e-10], + ..., + [ 7.9162e-08, -1.5087e-07, 1.0431e-07, ..., 8.5682e-08, + 9.3132e-10, 6.5193e-08], + [-2.1905e-06, 1.0617e-07, -1.7239e-06, ..., -2.0452e-06, + 0.0000e+00, 6.1467e-08], + [ 2.5146e-08, 2.6077e-08, -5.8580e-07, ..., 3.5390e-08, + 9.3132e-10, -1.0859e-06]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0069, -0.0034, 0.0088, 0.0096, 0.0340, 0.0431, -0.0205, 0.0080, + -0.0403, -0.0247], device='cuda:0'), grad: tensor([-7.9572e-06, 1.7174e-06, 8.0541e-06, -1.9088e-05, 2.4997e-06, + 3.5074e-06, 6.5081e-06, 1.5050e-05, -7.6964e-06, -2.6077e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 216.90, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4777 re_mapping 0.0056 re_causal 0.0154 /// teacc 99.07 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.0724, 0.1644, 0.0751, ..., -0.2433, -0.1254, 0.0215], + [ 0.1830, -0.0240, -0.0901, ..., 0.0941, 0.1281, -0.1604], + [-0.1176, -0.0647, -0.0479, ..., -0.0919, -0.0420, -0.0847], + ..., + [-0.0878, 0.1752, -0.1950, ..., -0.1774, -0.0160, -0.2012], + [ 0.0126, -0.1188, 0.0323, ..., 0.0736, -0.1327, -0.0550], + [-0.1299, -0.1312, 0.0949, ..., -0.1175, -0.0668, 0.1748]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.0338e-08, 4.0978e-08, ..., 3.3528e-08, + 1.8626e-09, 2.8871e-08], + [-2.3190e-07, 2.4624e-06, 3.7253e-09, ..., -2.9150e-07, + 6.8080e-07, 4.6566e-09], + [ 4.8429e-08, 5.0850e-07, 9.3132e-09, ..., 6.7055e-08, + 1.3504e-07, 6.5193e-09], + ..., + [ 8.3819e-08, -5.7593e-06, 5.5879e-09, ..., 8.1956e-08, + -1.6745e-06, 4.6566e-09], + [ 6.3330e-08, 2.5593e-06, 1.9558e-08, ..., 1.0151e-07, + 7.7393e-07, 7.5437e-08], + [ 4.2841e-08, 2.7847e-07, 1.0245e-08, ..., 1.5553e-07, + 6.6124e-08, -2.6077e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0067, -0.0039, 0.0094, 0.0104, 0.0338, 0.0422, -0.0203, 0.0077, + -0.0409, -0.0243], device='cuda:0'), grad: tensor([-2.2352e-08, 3.6582e-06, 1.0226e-06, 4.1015e-06, -2.3842e-07, + -4.1462e-06, -8.1025e-08, -9.8348e-06, 4.7460e-06, 7.8231e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 217.28, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.5136 re_mapping 0.0055 re_causal 0.0148 /// teacc 99.02 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.0733, 0.1656, 0.0749, ..., -0.2445, -0.1253, 0.0219], + [ 0.1831, -0.0240, -0.0915, ..., 0.0941, 0.1281, -0.1610], + [-0.1181, -0.0650, -0.0467, ..., -0.0920, -0.0419, -0.0852], + ..., + [-0.0884, 0.1754, -0.1955, ..., -0.1782, -0.0163, -0.2016], + [ 0.0133, -0.1191, 0.0330, ..., 0.0744, -0.1331, -0.0552], + [-0.1304, -0.1312, 0.0955, ..., -0.1177, -0.0673, 0.1751]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, -1.9222e-06, -1.9278e-07, ..., 3.1665e-08, + 9.3132e-10, -1.7807e-06], + [-2.5146e-08, 5.4017e-08, 5.5879e-09, ..., -5.5879e-09, + -9.3132e-10, 3.9116e-08], + [ 3.5390e-08, 2.0489e-08, 5.6811e-08, ..., -8.3819e-09, + -3.3528e-08, 1.3970e-08], + ..., + [ 1.3039e-08, 1.8626e-08, 8.3819e-09, ..., 2.0862e-07, + 3.7253e-09, 2.5146e-08], + [ 1.3970e-07, 4.8429e-08, 8.0094e-08, ..., 1.7881e-07, + 1.8626e-09, 1.8626e-08], + [ 1.1176e-08, 1.6950e-06, 1.6950e-07, ..., 9.3877e-07, + 1.8626e-09, 1.5628e-06]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0064, -0.0041, 0.0095, 0.0101, 0.0338, 0.0421, -0.0186, 0.0076, + -0.0407, -0.0242], device='cuda:0'), grad: tensor([-6.1169e-06, 3.1572e-07, -4.4797e-07, -2.8498e-07, -1.8813e-06, + 4.9267e-07, -3.4273e-07, 4.8615e-07, 7.1153e-07, 7.0743e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 217.54, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4901 re_mapping 0.0054 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.0739, 0.1671, 0.0753, ..., -0.2449, -0.1253, 0.0229], + [ 0.1816, -0.0265, -0.0920, ..., 0.0942, 0.1281, -0.1614], + [-0.1183, -0.0651, -0.0464, ..., -0.0920, -0.0419, -0.0856], + ..., + [-0.0865, 0.1774, -0.1955, ..., -0.1795, -0.0159, -0.2020], + [ 0.0138, -0.1195, 0.0334, ..., 0.0751, -0.1336, -0.0551], + [-0.1315, -0.1314, 0.0956, ..., -0.1188, -0.0675, 0.1750]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.0822e-06, -3.0734e-08, ..., -2.7660e-07, + 0.0000e+00, -1.0226e-06], + [ 8.3819e-09, 7.4506e-09, 3.0734e-08, ..., 2.7940e-09, + 0.0000e+00, 5.4948e-08], + [ 1.8626e-09, 5.5879e-09, 4.6566e-09, ..., 1.8626e-09, + 0.0000e+00, 5.5879e-09], + ..., + [ 1.9558e-08, 1.4901e-08, 1.1269e-07, ..., 8.3819e-09, + 0.0000e+00, 2.5425e-07], + [-4.4703e-08, 2.0489e-08, 1.7695e-08, ..., -6.3330e-08, + 0.0000e+00, 3.6322e-08], + [-4.6566e-08, 1.7975e-07, -2.2445e-07, ..., 2.2352e-08, + 0.0000e+00, -3.7812e-07]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0058, -0.0054, 0.0097, 0.0100, 0.0344, 0.0420, -0.0185, 0.0091, + -0.0407, -0.0248], device='cuda:0'), grad: tensor([-3.2298e-06, 1.6205e-07, 8.3819e-09, -5.5879e-09, 4.7684e-07, + 2.4345e-06, 3.3528e-07, 7.4599e-07, 4.2841e-08, -9.5740e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 217.53, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.5167 re_mapping 0.0052 re_causal 0.0154 /// teacc 99.05 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.0743, 0.1677, 0.0751, ..., -0.2451, -0.1254, 0.0229], + [ 0.1813, -0.0272, -0.0922, ..., 0.0942, 0.1282, -0.1617], + [-0.1184, -0.0648, -0.0463, ..., -0.0921, -0.0421, -0.0857], + ..., + [-0.0879, 0.1766, -0.1957, ..., -0.1813, -0.0160, -0.2026], + [ 0.0170, -0.1166, 0.0330, ..., 0.0768, -0.1348, -0.0552], + [-0.1318, -0.1315, 0.0960, ..., -0.1203, -0.0676, 0.1753]], + device='cuda:0'), grad: tensor([[ 4.6566e-08, -1.8347e-07, 6.8918e-08, ..., 1.0245e-07, + 0.0000e+00, -8.7544e-08], + [-1.1176e-08, 1.3970e-08, 2.9802e-08, ..., -1.5832e-08, + -1.8626e-09, 3.0734e-08], + [ 4.2841e-08, 5.4948e-08, 5.3085e-08, ..., 8.7544e-08, + -5.5879e-09, 4.2841e-08], + ..., + [ 2.1420e-08, -1.3318e-07, 8.2888e-08, ..., 5.2154e-08, + 1.8626e-09, 1.0524e-07], + [-1.2293e-07, 1.3039e-07, -4.3493e-07, ..., -1.0934e-06, + 2.7940e-09, -2.7753e-07], + [ 1.4901e-07, 7.8231e-08, -1.8720e-07, ..., 3.3248e-07, + 9.3132e-10, -5.5321e-07]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0057, -0.0057, 0.0098, 0.0104, 0.0353, 0.0418, -0.0188, 0.0083, + -0.0382, -0.0254], device='cuda:0'), grad: tensor([-4.2189e-07, 1.5181e-07, 2.4308e-07, 1.7099e-06, 1.5600e-06, + 3.1851e-07, -1.0431e-07, 3.7625e-07, -3.4049e-06, -4.4145e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 216.89, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4978 re_mapping 0.0053 re_causal 0.0149 /// teacc 99.02 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.0755, 0.1685, 0.0754, ..., -0.2461, -0.1254, 0.0233], + [ 0.1816, -0.0272, -0.0925, ..., 0.0944, 0.1284, -0.1631], + [-0.1189, -0.0648, -0.0463, ..., -0.0924, -0.0422, -0.0862], + ..., + [-0.0881, 0.1764, -0.1970, ..., -0.1822, -0.0164, -0.2056], + [ 0.0170, -0.1167, 0.0329, ..., 0.0770, -0.1352, -0.0553], + [-0.1325, -0.1312, 0.0978, ..., -0.1197, -0.0676, 0.1780]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.2107e-08, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 1.8626e-09], + [-8.3260e-07, -5.2713e-07, 0.0000e+00, ..., -7.2643e-07, + 0.0000e+00, 2.0489e-08], + [ 5.5879e-09, -1.3551e-06, 0.0000e+00, ..., 1.5739e-07, + 0.0000e+00, 2.3283e-08], + ..., + [ 6.7241e-07, 9.1176e-07, 2.7940e-09, ..., 1.4035e-06, + 0.0000e+00, 9.6858e-08], + [ 1.6764e-08, 4.7497e-08, 1.8626e-09, ..., 2.7008e-08, + 0.0000e+00, 7.4506e-09], + [ 1.3597e-07, 8.4098e-07, -2.8871e-08, ..., 5.9485e-05, + 0.0000e+00, 9.3505e-06]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0056, -0.0056, 0.0095, 0.0104, 0.0344, 0.0418, -0.0177, 0.0078, + -0.0383, -0.0244], device='cuda:0'), grad: tensor([ 3.9116e-08, -1.2731e-06, -2.4661e-06, 1.2573e-07, -1.6522e-04, + -9.0338e-08, 2.0303e-07, 3.5502e-06, 1.6205e-07, 1.6499e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 217.23, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.4974 re_mapping 0.0052 re_causal 0.0153 /// teacc 98.99 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.0758, 0.1687, 0.0754, ..., -0.2467, -0.1254, 0.0234], + [ 0.1819, -0.0270, -0.0926, ..., 0.0946, 0.1286, -0.1633], + [-0.1191, -0.0672, -0.0463, ..., -0.0925, -0.0423, -0.0864], + ..., + [-0.0882, 0.1776, -0.1972, ..., -0.1831, -0.0166, -0.2058], + [ 0.0166, -0.1169, 0.0319, ..., 0.0762, -0.1377, -0.0562], + [-0.1323, -0.1313, 0.0984, ..., -0.1219, -0.0678, 0.1784]], + device='cuda:0'), grad: tensor([[ 8.1956e-08, -6.7055e-08, 1.6764e-08, ..., 1.8068e-07, + 0.0000e+00, 2.1420e-08], + [ 2.8592e-07, 5.9605e-08, 1.7136e-07, ..., 3.9767e-07, + 0.0000e+00, 1.7695e-07], + [ 4.7497e-08, 6.7055e-08, 3.5390e-08, ..., 6.9849e-08, + 0.0000e+00, 2.1420e-08], + ..., + [ 1.9651e-07, -4.7497e-08, 8.1956e-08, ..., 3.0920e-07, + 0.0000e+00, 8.0094e-08], + [-3.1423e-06, -3.7160e-07, -1.0645e-06, ..., -4.8727e-06, + 0.0000e+00, -1.6764e-06], + [ 7.6555e-07, 2.4308e-07, 3.8277e-07, ..., 1.1856e-06, + 0.0000e+00, 3.5297e-07]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0056, -0.0055, 0.0088, 0.0105, 0.0364, 0.0419, -0.0178, 0.0084, + -0.0391, -0.0260], device='cuda:0'), grad: tensor([ 2.8126e-07, 1.0878e-06, 3.0454e-07, 7.1898e-07, 4.8336e-07, + 3.6880e-06, 1.2200e-06, 5.9139e-07, -1.1377e-05, 2.9951e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 217.25, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.5104 re_mapping 0.0053 re_causal 0.0153 /// teacc 98.92 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.0761, 0.1691, 0.0754, ..., -0.2471, -0.1255, 0.0234], + [ 0.1821, -0.0271, -0.0927, ..., 0.0948, 0.1290, -0.1633], + [-0.1194, -0.0670, -0.0463, ..., -0.0927, -0.0425, -0.0871], + ..., + [-0.0884, 0.1776, -0.1974, ..., -0.1837, -0.0171, -0.2060], + [ 0.0168, -0.1170, 0.0319, ..., 0.0769, -0.1382, -0.0562], + [-0.1331, -0.1313, 0.0987, ..., -0.1221, -0.0680, 0.1788]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, -3.7253e-09, 1.5832e-08, ..., 2.5146e-08, + 0.0000e+00, 1.3039e-08], + [ 1.8626e-09, 4.2003e-07, 1.3039e-08, ..., 2.2352e-08, + 2.9802e-08, 1.3970e-08], + [ 5.5879e-09, 3.7253e-09, 6.5193e-09, ..., 1.2573e-07, + 0.0000e+00, 5.5879e-09], + ..., + [ 4.6566e-09, -4.6846e-07, 5.1223e-08, ..., 4.6566e-08, + -3.1665e-08, 7.3574e-08], + [-1.9372e-07, 1.8626e-09, -3.2876e-07, ..., -6.2585e-07, + 0.0000e+00, -3.0641e-07], + [ 1.1362e-07, 4.0047e-08, 2.3283e-08, ..., 2.4308e-07, + 9.3132e-10, -7.3574e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0055, -0.0053, 0.0091, 0.0108, 0.0363, 0.0417, -0.0177, 0.0079, + -0.0391, -0.0260], device='cuda:0'), grad: tensor([ 8.5682e-08, 9.6392e-07, -2.7847e-07, 1.3039e-07, 4.5169e-07, + 2.9895e-07, 2.5425e-07, -3.0734e-07, -1.7826e-06, 2.0117e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 219.29, cls_loss 0.0019 cls_loss_mapping 0.0031 cls_loss_causal 0.4680 re_mapping 0.0054 re_causal 0.0142 /// teacc 98.94 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.0774, 0.1694, 0.0748, ..., -0.2480, -0.1255, 0.0231], + [ 0.1822, -0.0272, -0.0928, ..., 0.0947, 0.1295, -0.1635], + [-0.1200, -0.0670, -0.0462, ..., -0.0929, -0.0429, -0.0875], + ..., + [-0.0885, 0.1785, -0.1978, ..., -0.1846, -0.0177, -0.2063], + [ 0.0177, -0.1173, 0.0320, ..., 0.0791, -0.1383, -0.0558], + [-0.1343, -0.1314, 0.0994, ..., -0.1246, -0.0685, 0.1797]], + device='cuda:0'), grad: tensor([[ 1.8161e-07, -6.5193e-09, 2.9895e-07, ..., 1.4156e-07, + 1.3039e-08, 3.7998e-07], + [-3.5949e-07, 6.7055e-08, 4.7497e-08, ..., -3.8184e-07, + -2.7288e-07, 9.6858e-08], + [ 1.3225e-07, 1.9558e-08, 3.4459e-08, ..., 1.7788e-07, + 3.3528e-08, 4.4703e-08], + ..., + [ 1.2945e-07, -3.2037e-07, 1.1092e-06, ..., 2.4773e-07, + 6.1467e-08, 1.5972e-06], + [-8.1956e-08, 3.7253e-08, 3.3528e-08, ..., -9.6206e-07, + 3.5390e-08, 3.8743e-07], + [ 2.2352e-07, 2.0489e-07, -4.5821e-06, ..., 1.1520e-06, + 7.4506e-08, -7.2792e-06]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0059, -0.0054, 0.0093, 0.0091, 0.0381, 0.0413, -0.0171, 0.0094, + -0.0388, -0.0276], device='cuda:0'), grad: tensor([ 1.4091e-06, -3.8557e-07, 2.0210e-07, 7.3574e-07, 1.1601e-05, + 3.0994e-06, -1.0114e-06, 4.9546e-06, -1.2415e-06, -1.9401e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 222.27, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4975 re_mapping 0.0057 re_causal 0.0160 /// teacc 99.08 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.0780, 0.1700, 0.0747, ..., -0.2486, -0.1260, 0.0227], + [ 0.1822, -0.0276, -0.0929, ..., 0.0948, 0.1296, -0.1637], + [-0.1204, -0.0669, -0.0463, ..., -0.0931, -0.0431, -0.0877], + ..., + [-0.0884, 0.1788, -0.1981, ..., -0.1855, -0.0174, -0.2068], + [ 0.0178, -0.1174, 0.0321, ..., 0.0795, -0.1389, -0.0558], + [-0.1345, -0.1314, 0.1011, ..., -0.1246, -0.0689, 0.1812]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.7940e-09, 1.0245e-08, ..., 2.7940e-09, + 0.0000e+00, -1.8626e-09], + [-1.6764e-08, 6.7055e-08, 9.3132e-09, ..., -2.4214e-08, + -7.4506e-09, 9.3132e-10], + [ 8.3819e-09, 6.6124e-08, 1.3039e-08, ..., 1.7695e-08, + 1.0245e-08, 0.0000e+00], + ..., + [ 1.8626e-08, -3.9302e-07, 3.9116e-08, ..., 5.0291e-08, + -3.3528e-08, 1.8626e-09], + [-2.0023e-07, 2.2259e-07, 1.1083e-07, ..., -4.7870e-07, + 2.8871e-08, -9.3132e-09], + [ 1.3970e-08, 1.5832e-08, 7.4506e-09, ..., 1.0803e-07, + 1.8626e-09, -3.6322e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0060, -0.0057, 0.0093, 0.0092, 0.0377, 0.0412, -0.0164, 0.0093, + -0.0387, -0.0272], device='cuda:0'), grad: tensor([ 3.1665e-08, 9.7789e-08, 5.7649e-07, -1.3277e-05, -3.1944e-07, + 1.1794e-05, 3.3062e-07, 3.9954e-07, 7.6368e-08, 2.7381e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 221.87, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4670 re_mapping 0.0052 re_causal 0.0147 /// teacc 99.03 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.0803, 0.1704, 0.0744, ..., -0.2508, -0.1275, 0.0224], + [ 0.1814, -0.0280, -0.0929, ..., 0.0946, 0.1291, -0.1638], + [-0.1206, -0.0670, -0.0463, ..., -0.0931, -0.0432, -0.0879], + ..., + [-0.0881, 0.1790, -0.1994, ..., -0.1863, -0.0167, -0.2085], + [ 0.0178, -0.1175, 0.0305, ..., 0.0797, -0.1394, -0.0570], + [-0.1354, -0.1314, 0.1014, ..., -0.1250, -0.0689, 0.1815]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, -1.2675e-06, -3.7346e-07, ..., 4.5635e-08, + 5.5879e-09, -4.7777e-07], + [-7.1712e-08, 8.2888e-08, 9.3132e-08, ..., -1.7695e-07, + -8.1025e-08, 1.1735e-07], + [ 4.2841e-08, 2.5332e-07, 1.4808e-07, ..., 6.9849e-08, + 5.2154e-08, 1.0151e-07], + ..., + [ 1.6950e-07, -9.4064e-08, 1.5832e-07, ..., 3.0268e-07, + 3.9116e-08, 1.5832e-07], + [ 1.1921e-07, 3.0827e-07, 7.4506e-09, ..., -1.2107e-08, + 9.8720e-08, 3.3714e-07], + [ 1.8906e-07, 2.8405e-07, 1.9837e-07, ..., 2.9150e-07, + 9.7789e-08, 2.9057e-07]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0068, -0.0063, 0.0093, 0.0097, 0.0377, 0.0410, -0.0142, 0.0092, + -0.0390, -0.0271], device='cuda:0'), grad: tensor([-6.4075e-06, 4.9360e-08, 1.5637e-06, -8.1025e-08, 1.0524e-07, + -1.6484e-06, 2.6133e-06, 7.4878e-07, 1.1744e-06, 1.8906e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 220.78, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5002 re_mapping 0.0053 re_causal 0.0152 /// teacc 98.99 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.0805, 0.1710, 0.0750, ..., -0.2517, -0.1275, 0.0224], + [ 0.1825, -0.0274, -0.0931, ..., 0.0954, 0.1295, -0.1638], + [-0.1208, -0.0676, -0.0461, ..., -0.0933, -0.0432, -0.0884], + ..., + [-0.0893, 0.1791, -0.1998, ..., -0.1907, -0.0184, -0.2090], + [ 0.0170, -0.1178, 0.0293, ..., 0.0769, -0.1399, -0.0603], + [-0.1369, -0.1317, 0.1019, ..., -0.1255, -0.0698, 0.1819]], + device='cuda:0'), grad: tensor([[ 4.2375e-08, -6.6124e-08, 2.3283e-09, ..., 4.8894e-08, + 0.0000e+00, -2.7940e-09], + [ 3.7719e-08, 2.2305e-07, 5.0757e-08, ..., 4.7963e-08, + 0.0000e+00, 4.5169e-08], + [ 1.3039e-08, -6.7521e-08, 1.8161e-08, ..., 2.0955e-08, + 0.0000e+00, 8.3819e-09], + ..., + [ 5.0757e-08, -3.6368e-07, 7.5158e-07, ..., 8.0094e-08, + 0.0000e+00, 7.2317e-07], + [-3.6508e-07, 1.5087e-07, 2.0489e-08, ..., -4.0093e-07, + 0.0000e+00, 5.8673e-08], + [ 1.0524e-07, -2.6077e-08, -2.0154e-06, ..., 1.1176e-07, + 0.0000e+00, -2.1234e-06]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0065, -0.0055, 0.0093, 0.0108, 0.0380, 0.0404, -0.0140, 0.0086, + -0.0407, -0.0274], device='cuda:0'), grad: tensor([ 2.2585e-07, 6.9151e-07, -3.8138e-07, 1.1418e-06, 3.9749e-06, + -3.2689e-07, 3.9209e-07, 2.9579e-06, -1.7658e-06, -6.9141e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 222.38, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.5126 re_mapping 0.0051 re_causal 0.0147 /// teacc 99.01 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.0806, 0.1719, 0.0751, ..., -0.2522, -0.1275, 0.0225], + [ 0.1825, -0.0276, -0.0933, ..., 0.0954, 0.1295, -0.1641], + [-0.1209, -0.0676, -0.0466, ..., -0.0933, -0.0433, -0.0886], + ..., + [-0.0893, 0.1794, -0.2009, ..., -0.1911, -0.0185, -0.2091], + [ 0.0181, -0.1178, 0.0294, ..., 0.0789, -0.1400, -0.0590], + [-0.1371, -0.1319, 0.1022, ..., -0.1256, -0.0698, 0.1823]], + device='cuda:0'), grad: tensor([[ 1.0338e-07, 2.0489e-08, 1.3504e-08, ..., 7.5437e-08, + 1.3970e-09, 4.6566e-09], + [-4.9472e-06, -4.6100e-07, -3.8650e-08, ..., -7.7337e-06, + -3.8892e-06, 1.0710e-08], + [ 3.3751e-06, 2.1886e-08, 1.2107e-08, ..., 6.2287e-06, + 3.5241e-06, 7.4506e-09], + ..., + [ 5.1735e-07, 2.1420e-08, 2.3749e-08, ..., 6.6636e-07, + 2.8824e-07, 1.3039e-08], + [ 6.1886e-07, 1.4482e-07, 1.9651e-07, ..., 6.9151e-07, + 2.3283e-09, 2.1001e-07], + [ 4.7917e-07, 1.8766e-07, 1.0803e-07, ..., 4.1211e-07, + 5.5879e-09, 9.6858e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0061, -0.0057, 0.0092, 0.0117, 0.0379, 0.0402, -0.0154, 0.0082, + -0.0397, -0.0274], device='cuda:0'), grad: tensor([ 2.0396e-07, -1.8522e-05, 1.4029e-05, 5.4855e-07, 1.6065e-07, + -8.8513e-06, 7.7784e-06, 1.6447e-06, 1.7984e-06, 1.2144e-06], + device='cuda:0') +100 +0.0001 +changing lr diff --git a/Meta-causal/code-withStyleAttack/65662.error b/Meta-causal/code-withStyleAttack/65662.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/65662.log b/Meta-causal/code-withStyleAttack/65662.log new file mode 100644 index 0000000000000000000000000000000000000000..c30ac72a3bdb1ad4985ed47eb8c4e80a2e2b08f4 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/65662.log @@ -0,0 +1,7153 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 250, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep250_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_epoch250', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0118, 0.0228, -0.0024, ..., -0.0263, 0.0294, 0.0225], + [ 0.0117, 0.0125, -0.0312, ..., 0.0006, -0.0133, 0.0132], + [ 0.0204, -0.0145, -0.0123, ..., -0.0209, 0.0127, -0.0073], + ..., + [-0.0015, -0.0013, 0.0121, ..., 0.0043, 0.0199, 0.0109], + [ 0.0096, 0.0208, -0.0142, ..., 0.0160, -0.0174, 0.0288], + [-0.0055, -0.0161, 0.0242, ..., -0.0260, -0.0230, 0.0003]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0002, 0.0236, 0.0062, 0.0051, 0.0040, -0.0094, 0.0250, -0.0272, + 0.0250, 0.0046], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 221.75, cls_loss 1.6842 cls_loss_mapping 2.0049 cls_loss_causal 2.2496 re_mapping 0.0767 re_causal 0.0774 /// teacc 84.37 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0126, 0.0228, 0.0004, ..., -0.0319, 0.0314, 0.0219], + [ 0.0086, 0.0125, -0.0421, ..., 0.0074, -0.0105, 0.0126], + [ 0.0192, -0.0145, -0.0187, ..., -0.0247, 0.0152, -0.0079], + ..., + [ 0.0011, -0.0013, 0.0107, ..., 0.0045, 0.0185, 0.0114], + [ 0.0097, 0.0208, -0.0128, ..., 0.0117, -0.0165, 0.0295], + [-0.0047, -0.0161, 0.0288, ..., -0.0276, -0.0288, -0.0004]], + device='cuda:0'), grad: tensor([[ 0.0072, 0.0000, 0.0089, ..., 0.0002, 0.0204, 0.0000], + [ 0.0124, 0.0000, 0.0125, ..., 0.0015, 0.0347, 0.0000], + [-0.0183, 0.0000, 0.0107, ..., 0.0003, -0.0353, 0.0000], + ..., + [-0.0045, 0.0000, -0.0565, ..., -0.0074, -0.0013, 0.0000], + [-0.0344, 0.0000, -0.0058, ..., -0.0060, -0.0693, 0.0000], + [ 0.0173, 0.0000, 0.1266, ..., 0.0083, 0.0386, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0037, 0.0263, 0.0055, 0.0059, 0.0041, -0.0085, 0.0241, -0.0262, + 0.0237, 0.0050], device='cuda:0'), grad: tensor([ 0.0261, 0.0339, -0.0185, 0.0468, -0.0235, -0.0387, -0.0398, -0.0343, + -0.0600, 0.1082], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 222.81, cls_loss 0.5751 cls_loss_mapping 0.8903 cls_loss_causal 1.9459 re_mapping 0.2095 re_causal 0.2474 /// teacc 90.15 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0133, 0.0228, -0.0016, ..., -0.0345, 0.0322, 0.0219], + [ 0.0073, 0.0125, -0.0446, ..., 0.0099, -0.0113, 0.0126], + [ 0.0210, -0.0145, -0.0211, ..., -0.0259, 0.0167, -0.0079], + ..., + [ 0.0014, -0.0013, 0.0108, ..., 0.0048, 0.0192, 0.0114], + [ 0.0087, 0.0208, -0.0127, ..., 0.0093, -0.0153, 0.0295], + [-0.0047, -0.0161, 0.0294, ..., -0.0308, -0.0306, -0.0004]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0000, 0.0062, ..., 0.0003, 0.0099, 0.0000], + [ 0.0026, 0.0000, 0.0009, ..., 0.0002, 0.0028, 0.0000], + [-0.0282, 0.0000, -0.0045, ..., 0.0004, -0.0375, 0.0000], + ..., + [-0.0017, 0.0000, 0.0044, ..., 0.0007, -0.0014, 0.0000], + [ 0.0175, 0.0000, 0.0013, ..., 0.0011, 0.0173, 0.0000], + [ 0.0029, 0.0000, 0.0058, ..., 0.0027, 0.0027, 0.0000]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0044, 0.0265, 0.0052, 0.0057, 0.0041, -0.0070, 0.0236, -0.0264, + 0.0234, 0.0052], device='cuda:0'), grad: tensor([ 0.0199, 0.0046, -0.0508, -0.0047, 0.0078, -0.0029, -0.0100, 0.0019, + 0.0240, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 220.88, cls_loss 0.3477 cls_loss_mapping 0.5253 cls_loss_causal 1.7447 re_mapping 0.1589 re_causal 0.2344 /// teacc 93.42 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0136, 0.0228, -0.0035, ..., -0.0363, 0.0330, 0.0219], + [ 0.0068, 0.0125, -0.0460, ..., 0.0108, -0.0103, 0.0126], + [ 0.0225, -0.0145, -0.0216, ..., -0.0257, 0.0167, -0.0079], + ..., + [ 0.0025, -0.0013, 0.0104, ..., 0.0053, 0.0195, 0.0114], + [ 0.0083, 0.0208, -0.0141, ..., 0.0069, -0.0150, 0.0295], + [-0.0057, -0.0161, 0.0304, ..., -0.0329, -0.0310, -0.0004]], + device='cuda:0'), grad: tensor([[-0.0008, 0.0000, 0.0006, ..., 0.0000, -0.0026, 0.0000], + [ 0.0009, 0.0000, 0.0005, ..., 0.0000, -0.0022, 0.0000], + [-0.0275, 0.0000, 0.0009, ..., 0.0000, -0.0058, 0.0000], + ..., + [ 0.0023, 0.0000, 0.0018, ..., 0.0000, 0.0020, 0.0000], + [ 0.0611, 0.0000, 0.0223, ..., 0.0000, 0.0236, 0.0000], + [ 0.0045, 0.0000, -0.0014, ..., 0.0000, 0.0022, 0.0000]], + device='cuda:0') +Epoch 4, bias, value: tensor([-0.0041, 0.0269, 0.0052, 0.0057, 0.0037, -0.0064, 0.0235, -0.0267, + 0.0230, 0.0052], device='cuda:0'), grad: tensor([-0.0077, -0.0008, -0.0145, -0.0323, 0.0042, -0.0293, 0.0009, 0.0038, + 0.0704, 0.0052], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 218.72, cls_loss 0.2539 cls_loss_mapping 0.3669 cls_loss_causal 1.5234 re_mapping 0.1287 re_causal 0.2210 /// teacc 94.90 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0134, 0.0228, -0.0046, ..., -0.0408, 0.0336, 0.0219], + [ 0.0065, 0.0125, -0.0470, ..., 0.0115, -0.0087, 0.0126], + [ 0.0238, -0.0145, -0.0223, ..., -0.0229, 0.0165, -0.0079], + ..., + [ 0.0047, -0.0013, 0.0098, ..., 0.0070, 0.0199, 0.0114], + [ 0.0074, 0.0208, -0.0146, ..., 0.0038, -0.0147, 0.0295], + [-0.0073, -0.0161, 0.0309, ..., -0.0338, -0.0308, -0.0004]], + device='cuda:0'), grad: tensor([[ 2.4948e-03, 0.0000e+00, 2.3232e-03, ..., 5.9843e-05, + 8.5449e-03, 0.0000e+00], + [-3.3360e-03, 0.0000e+00, -6.3744e-03, ..., -1.0071e-02, + -2.1877e-03, 0.0000e+00], + [ 2.0355e-02, 0.0000e+00, 1.5106e-03, ..., 8.4114e-04, + 1.0719e-02, 0.0000e+00], + ..., + [ 1.0414e-02, 0.0000e+00, 7.0229e-03, ..., -1.8120e-04, + 6.8588e-03, 0.0000e+00], + [-4.2389e-02, 0.0000e+00, 1.3069e-02, ..., 2.7485e-03, + -1.3840e-02, 0.0000e+00], + [ 4.2953e-03, 0.0000e+00, -3.7781e-02, ..., 6.1703e-04, + -8.6594e-03, 0.0000e+00]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0039, 0.0276, 0.0055, 0.0055, 0.0038, -0.0067, 0.0232, -0.0267, + 0.0230, 0.0051], device='cuda:0'), grad: tensor([ 0.0161, -0.0300, 0.0190, 0.0335, -0.0035, -0.0140, -0.0066, 0.0149, + -0.0064, -0.0230], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 219.21, cls_loss 0.2005 cls_loss_mapping 0.2738 cls_loss_causal 1.4344 re_mapping 0.1058 re_causal 0.2003 /// teacc 95.39 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0138, 0.0228, -0.0060, ..., -0.0444, 0.0334, 0.0219], + [ 0.0058, 0.0125, -0.0481, ..., 0.0117, -0.0080, 0.0126], + [ 0.0248, -0.0145, -0.0229, ..., -0.0204, 0.0163, -0.0079], + ..., + [ 0.0057, -0.0013, 0.0094, ..., 0.0076, 0.0200, 0.0114], + [ 0.0074, 0.0208, -0.0151, ..., 0.0024, -0.0143, 0.0295], + [-0.0078, -0.0161, 0.0315, ..., -0.0345, -0.0306, -0.0004]], + device='cuda:0'), grad: tensor([[-2.9697e-03, 0.0000e+00, 6.4507e-03, ..., 2.4348e-05, + -9.5177e-04, 0.0000e+00], + [ 1.1158e-03, 0.0000e+00, 4.9877e-04, ..., 7.8022e-05, + 9.6560e-04, 0.0000e+00], + [ 1.1053e-03, 0.0000e+00, 1.8187e-03, ..., 1.6415e-04, + -8.0795e-03, 0.0000e+00], + ..., + [-1.9119e-02, 0.0000e+00, -5.7793e-04, ..., 5.4866e-05, + -1.5533e-02, 0.0000e+00], + [ 2.9640e-03, 0.0000e+00, -6.6185e-03, ..., 3.2568e-04, + 4.6883e-03, 0.0000e+00], + [ 1.2794e-02, 0.0000e+00, 8.0338e-03, ..., 2.3575e-03, + 1.3672e-02, 0.0000e+00]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0039, 0.0276, 0.0055, 0.0052, 0.0041, -0.0066, 0.0229, -0.0268, + 0.0231, 0.0053], device='cuda:0'), grad: tensor([ 0.0018, 0.0014, -0.0032, 0.0058, -0.0147, 0.0065, 0.0040, -0.0266, + -0.0003, 0.0253], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 219.32, cls_loss 0.1731 cls_loss_mapping 0.2329 cls_loss_causal 1.3032 re_mapping 0.0907 re_causal 0.1857 /// teacc 95.98 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0144, 0.0228, -0.0067, ..., -0.0481, 0.0336, 0.0219], + [ 0.0057, 0.0125, -0.0497, ..., 0.0112, -0.0073, 0.0126], + [ 0.0258, -0.0145, -0.0238, ..., -0.0200, 0.0160, -0.0079], + ..., + [ 0.0062, -0.0013, 0.0088, ..., 0.0078, 0.0200, 0.0114], + [ 0.0069, 0.0208, -0.0155, ..., 0.0013, -0.0143, 0.0295], + [-0.0090, -0.0161, 0.0321, ..., -0.0352, -0.0308, -0.0004]], + device='cuda:0'), grad: tensor([[-5.1041e-03, 0.0000e+00, 1.3676e-03, ..., 2.3358e-06, + -3.9597e-03, 0.0000e+00], + [ 1.2379e-03, 0.0000e+00, 9.1839e-04, ..., 1.4175e-06, + 6.3438e-03, 0.0000e+00], + [ 5.2166e-04, 0.0000e+00, 2.2125e-03, ..., 9.5833e-07, + -5.6763e-03, 0.0000e+00], + ..., + [ 1.5430e-03, 0.0000e+00, 1.4477e-03, ..., -1.9759e-05, + 4.8304e-04, 0.0000e+00], + [ 4.6120e-03, 0.0000e+00, 2.7122e-03, ..., 3.3388e-07, + 3.3855e-03, 0.0000e+00], + [ 1.3089e-04, 0.0000e+00, -1.5572e-02, ..., 1.0930e-05, + -2.2449e-03, 0.0000e+00]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0034, 0.0276, 0.0053, 0.0051, 0.0045, -0.0070, 0.0225, -0.0269, + 0.0232, 0.0054], device='cuda:0'), grad: tensor([-0.0089, 0.0094, -0.0060, -0.0251, 0.0048, 0.0266, -0.0009, 0.0030, + 0.0094, -0.0122], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 219.21, cls_loss 0.1524 cls_loss_mapping 0.2100 cls_loss_causal 1.2993 re_mapping 0.0762 re_causal 0.1680 /// teacc 96.07 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0149, 0.0228, -0.0079, ..., -0.0489, 0.0333, 0.0219], + [ 0.0052, 0.0125, -0.0507, ..., 0.0111, -0.0067, 0.0126], + [ 0.0266, -0.0145, -0.0243, ..., -0.0201, 0.0152, -0.0079], + ..., + [ 0.0067, -0.0013, 0.0084, ..., 0.0079, 0.0200, 0.0114], + [ 0.0065, 0.0208, -0.0156, ..., 0.0013, -0.0144, 0.0295], + [-0.0094, -0.0161, 0.0328, ..., -0.0354, -0.0297, -0.0004]], + device='cuda:0'), grad: tensor([[-0.0003, 0.0000, 0.0005, ..., 0.0000, -0.0020, 0.0000], + [ 0.0006, 0.0000, -0.0210, ..., 0.0000, -0.0089, 0.0000], + [-0.0083, 0.0000, 0.0005, ..., 0.0000, -0.0036, 0.0000], + ..., + [ 0.0036, 0.0000, 0.0120, ..., 0.0000, 0.0046, 0.0000], + [ 0.0022, 0.0000, 0.0033, ..., 0.0000, 0.0036, 0.0000], + [-0.0108, 0.0000, -0.0057, ..., 0.0000, -0.0091, 0.0000]], + device='cuda:0') +Epoch 8, bias, value: tensor([-0.0035, 0.0279, 0.0055, 0.0049, 0.0042, -0.0072, 0.0225, -0.0270, + 0.0231, 0.0058], device='cuda:0'), grad: tensor([-0.0048, -0.0293, -0.0114, 0.0277, 0.0001, 0.0094, -0.0057, 0.0235, + 0.0070, -0.0167], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 219.46, cls_loss 0.1335 cls_loss_mapping 0.1743 cls_loss_causal 1.2317 re_mapping 0.0690 re_causal 0.1570 /// teacc 96.80 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0148, 0.0228, -0.0085, ..., -0.0495, 0.0332, 0.0219], + [ 0.0047, 0.0125, -0.0512, ..., 0.0108, -0.0057, 0.0126], + [ 0.0269, -0.0145, -0.0251, ..., -0.0202, 0.0141, -0.0079], + ..., + [ 0.0080, -0.0013, 0.0080, ..., 0.0082, 0.0203, 0.0114], + [ 0.0060, 0.0208, -0.0156, ..., 0.0017, -0.0143, 0.0295], + [-0.0108, -0.0161, 0.0333, ..., -0.0362, -0.0297, -0.0004]], + device='cuda:0'), grad: tensor([[ 2.0733e-03, 0.0000e+00, 2.9349e-04, ..., 0.0000e+00, + -7.9453e-05, 0.0000e+00], + [ 1.6060e-03, 0.0000e+00, 8.5402e-04, ..., 0.0000e+00, + 9.1362e-04, 0.0000e+00], + [ 5.8055e-05, 0.0000e+00, 3.3522e-04, ..., 0.0000e+00, + 9.9277e-04, 0.0000e+00], + ..., + [ 4.6692e-03, 0.0000e+00, 2.4109e-03, ..., 0.0000e+00, + 8.7357e-04, 0.0000e+00], + [-1.6518e-03, 0.0000e+00, 1.9779e-03, ..., 0.0000e+00, + -1.9665e-03, 0.0000e+00], + [ 2.2774e-03, 0.0000e+00, 4.0283e-03, ..., 0.0000e+00, + -9.0265e-04, 0.0000e+00]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0035, 0.0284, 0.0051, 0.0052, 0.0044, -0.0077, 0.0223, -0.0268, + 0.0235, 0.0056], device='cuda:0'), grad: tensor([ 0.0006, 0.0042, 0.0017, 0.0190, -0.0242, -0.0178, -0.0019, 0.0090, + 0.0014, 0.0079], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 219.68, cls_loss 0.1251 cls_loss_mapping 0.1543 cls_loss_causal 1.1917 re_mapping 0.0646 re_causal 0.1443 /// teacc 97.04 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0155, 0.0228, -0.0092, ..., -0.0506, 0.0329, 0.0219], + [ 0.0038, 0.0125, -0.0519, ..., 0.0105, -0.0049, 0.0126], + [ 0.0274, -0.0145, -0.0256, ..., -0.0202, 0.0133, -0.0079], + ..., + [ 0.0088, -0.0013, 0.0078, ..., 0.0089, 0.0203, 0.0114], + [ 0.0060, 0.0208, -0.0159, ..., 0.0020, -0.0140, 0.0295], + [-0.0118, -0.0161, 0.0339, ..., -0.0364, -0.0292, -0.0004]], + device='cuda:0'), grad: tensor([[ 3.0828e-04, 0.0000e+00, 4.2892e-04, ..., 4.5970e-06, + 4.0579e-04, 0.0000e+00], + [ 6.8617e-04, 0.0000e+00, 8.0824e-05, ..., 1.7239e-06, + -4.3917e-04, 0.0000e+00], + [ 1.7424e-03, 0.0000e+00, 5.7983e-04, ..., 1.4137e-06, + 2.6398e-03, 0.0000e+00], + ..., + [ 1.7726e-04, 0.0000e+00, 2.1839e-04, ..., -1.5914e-05, + 5.3644e-04, 0.0000e+00], + [-4.1962e-03, 0.0000e+00, -5.4979e-04, ..., 1.0096e-06, + 2.3484e-05, 0.0000e+00], + [ 9.2316e-04, 0.0000e+00, 4.6015e-04, ..., 4.0270e-06, + 1.1492e-04, 0.0000e+00]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0037, 0.0280, 0.0050, 0.0049, 0.0042, -0.0077, 0.0222, -0.0264, + 0.0240, 0.0057], device='cuda:0'), grad: tensor([ 0.0023, 0.0010, 0.0195, -0.0186, 0.0017, 0.0015, -0.0014, 0.0021, + -0.0106, 0.0024], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 218.62, cls_loss 0.1113 cls_loss_mapping 0.1363 cls_loss_causal 1.1221 re_mapping 0.0603 re_causal 0.1350 /// teacc 96.99 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0164, 0.0228, -0.0096, ..., -0.0521, 0.0328, 0.0219], + [ 0.0033, 0.0125, -0.0523, ..., 0.0103, -0.0042, 0.0126], + [ 0.0280, -0.0145, -0.0261, ..., -0.0204, 0.0129, -0.0079], + ..., + [ 0.0089, -0.0013, 0.0071, ..., 0.0104, 0.0199, 0.0114], + [ 0.0057, 0.0208, -0.0162, ..., 0.0015, -0.0138, 0.0295], + [-0.0123, -0.0161, 0.0344, ..., -0.0364, -0.0289, -0.0004]], + device='cuda:0'), grad: tensor([[ 1.7178e-04, 0.0000e+00, 1.3912e-04, ..., 2.2531e-05, + 3.3593e-04, 0.0000e+00], + [ 3.1853e-04, 0.0000e+00, 1.2231e-04, ..., 5.9724e-05, + -3.1519e-04, 0.0000e+00], + [ 1.0300e-03, 0.0000e+00, 1.1057e-04, ..., 3.7074e-04, + 4.4799e-04, 0.0000e+00], + ..., + [-2.0523e-03, 0.0000e+00, 4.9686e-04, ..., -3.4976e-04, + -1.0376e-03, 0.0000e+00], + [ 2.0905e-03, 0.0000e+00, 3.2921e-03, ..., 2.2125e-04, + 2.3098e-03, 0.0000e+00], + [-1.5465e-02, 0.0000e+00, -5.7793e-03, ..., 1.4937e-04, + -1.5442e-02, 0.0000e+00]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0038, 0.0280, 0.0052, 0.0048, 0.0044, -0.0077, 0.0219, -0.0265, + 0.0243, 0.0056], device='cuda:0'), grad: tensor([ 6.6853e-04, -5.5701e-05, 1.0920e-03, 1.9730e-02, 7.5483e-04, + -5.8441e-03, -1.4508e-04, -1.8644e-03, 5.8098e-03, -2.0142e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 221.32, cls_loss 0.0999 cls_loss_mapping 0.1227 cls_loss_causal 1.1094 re_mapping 0.0549 re_causal 0.1267 /// teacc 97.24 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0171, 0.0228, -0.0102, ..., -0.0525, 0.0324, 0.0219], + [ 0.0031, 0.0125, -0.0536, ..., 0.0116, -0.0032, 0.0126], + [ 0.0284, -0.0145, -0.0262, ..., -0.0203, 0.0123, -0.0079], + ..., + [ 0.0094, -0.0013, 0.0069, ..., 0.0110, 0.0197, 0.0114], + [ 0.0054, 0.0208, -0.0166, ..., 0.0004, -0.0138, 0.0295], + [-0.0129, -0.0161, 0.0346, ..., -0.0370, -0.0284, -0.0004]], + device='cuda:0'), grad: tensor([[ 2.4557e-04, 0.0000e+00, -2.1343e-03, ..., 1.0300e-06, + -1.6947e-03, 0.0000e+00], + [ 2.1577e-04, 0.0000e+00, 3.1209e-04, ..., 1.6671e-07, + -8.0013e-04, 0.0000e+00], + [-1.9207e-03, 0.0000e+00, 4.3845e-04, ..., 1.9372e-07, + 2.2519e-04, 0.0000e+00], + ..., + [ 2.8992e-04, 0.0000e+00, 1.5678e-03, ..., 1.7462e-07, + 1.4315e-03, 0.0000e+00], + [-1.1641e-04, 0.0000e+00, -3.7231e-03, ..., 4.0419e-06, + -3.1757e-03, 0.0000e+00], + [ 1.1234e-03, 0.0000e+00, 7.3776e-03, ..., 1.6848e-06, + 2.5864e-03, 0.0000e+00]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0040, 0.0282, 0.0049, 0.0049, 0.0042, -0.0078, 0.0220, -0.0264, + 0.0244, 0.0056], device='cuda:0'), grad: tensor([-1.0231e-02, -1.9920e-04, -1.6049e-05, 4.3144e-03, -4.3869e-03, + 3.5877e-03, 1.2624e-04, 5.4092e-03, -1.2894e-02, 1.4290e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 218.79, cls_loss 0.0856 cls_loss_mapping 0.1040 cls_loss_causal 1.0154 re_mapping 0.0538 re_causal 0.1211 /// teacc 97.46 lr 0.00010000 +Epoch 13, weight, value: tensor([[-1.7175e-02, 2.2769e-02, -1.0488e-02, ..., -5.3541e-02, + 3.2494e-02, 2.1893e-02], + [ 2.9271e-03, 1.2500e-02, -5.4046e-02, ..., 1.1676e-02, + -2.6134e-03, 1.2576e-02], + [ 2.8715e-02, -1.4480e-02, -2.6587e-02, ..., -2.0364e-02, + 1.1720e-02, -7.9110e-03], + ..., + [ 1.0327e-02, -1.2941e-03, 6.8708e-03, ..., 1.1770e-02, + 1.9598e-02, 1.1425e-02], + [ 4.7974e-03, 2.0836e-02, -1.6912e-02, ..., 4.7806e-05, + -1.3859e-02, 2.9502e-02], + [-1.3850e-02, -1.6141e-02, 3.5095e-02, ..., -3.7387e-02, + -2.7884e-02, -3.8447e-04]], device='cuda:0'), grad: tensor([[ 4.3821e-04, 0.0000e+00, 6.7174e-05, ..., 5.4501e-06, + 9.9778e-05, 0.0000e+00], + [ 1.0386e-03, 0.0000e+00, 1.5581e-04, ..., 1.2964e-05, + -3.2783e-04, 0.0000e+00], + [ 1.9627e-03, 0.0000e+00, 2.3377e-04, ..., 1.3493e-05, + 1.2789e-03, 0.0000e+00], + ..., + [-1.0918e-02, 0.0000e+00, -2.1954e-03, ..., -1.4901e-04, + -1.3571e-03, 0.0000e+00], + [ 2.9602e-03, 0.0000e+00, 9.1553e-05, ..., 1.5095e-05, + -7.1478e-04, 0.0000e+00], + [-3.8795e-03, 0.0000e+00, -1.3718e-02, ..., 4.4942e-05, + 7.5340e-04, 0.0000e+00]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0035, 0.0282, 0.0049, 0.0050, 0.0041, -0.0079, 0.0217, -0.0261, + 0.0245, 0.0054], device='cuda:0'), grad: tensor([-0.0006, 0.0010, -0.0040, 0.0029, 0.0187, 0.0026, -0.0006, -0.0105, + 0.0057, -0.0152], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 219.25, cls_loss 0.0909 cls_loss_mapping 0.1133 cls_loss_causal 1.0595 re_mapping 0.0484 re_causal 0.1174 /// teacc 97.57 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0174, 0.0223, -0.0110, ..., -0.0539, 0.0324, 0.0219], + [ 0.0026, 0.0108, -0.0538, ..., 0.0116, -0.0016, 0.0126], + [ 0.0294, -0.0126, -0.0267, ..., -0.0198, 0.0113, -0.0079], + ..., + [ 0.0107, -0.0024, 0.0064, ..., 0.0118, 0.0192, 0.0114], + [ 0.0041, 0.0201, -0.0172, ..., -0.0004, -0.0138, 0.0295], + [-0.0141, -0.0164, 0.0356, ..., -0.0377, -0.0274, -0.0004]], + device='cuda:0'), grad: tensor([[-1.1490e-02, 0.0000e+00, -4.9531e-05, ..., 0.0000e+00, + -1.7609e-02, 0.0000e+00], + [ 8.0109e-03, 0.0000e+00, 3.1352e-04, ..., 0.0000e+00, + 1.1238e-02, 0.0000e+00], + [-5.4665e-03, 0.0000e+00, 1.9860e-04, ..., 0.0000e+00, + 4.8375e-04, 0.0000e+00], + ..., + [ 3.0017e-04, 0.0000e+00, 1.4007e-04, ..., 0.0000e+00, + 4.7827e-04, 0.0000e+00], + [ 5.2948e-03, 0.0000e+00, 6.5279e-04, ..., 0.0000e+00, + 1.3485e-03, 0.0000e+00], + [ 3.5019e-03, 0.0000e+00, 1.9302e-03, ..., 0.0000e+00, + 2.7733e-03, 0.0000e+00]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0034, 0.0282, 0.0052, 0.0048, 0.0041, -0.0080, 0.0213, -0.0263, + 0.0245, 0.0056], device='cuda:0'), grad: tensor([-0.0307, 0.0211, -0.0127, -0.0020, -0.0030, -0.0025, 0.0077, 0.0009, + 0.0143, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 218.77, cls_loss 0.0798 cls_loss_mapping 0.0973 cls_loss_causal 1.0558 re_mapping 0.0448 re_causal 0.1108 /// teacc 97.50 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0175, 0.0202, -0.0115, ..., -0.0566, 0.0324, 0.0219], + [ 0.0023, 0.0079, -0.0546, ..., 0.0113, -0.0007, 0.0126], + [ 0.0293, -0.0095, -0.0275, ..., -0.0198, 0.0108, -0.0079], + ..., + [ 0.0112, -0.0049, 0.0064, ..., 0.0129, 0.0189, 0.0114], + [ 0.0036, 0.0175, -0.0176, ..., -0.0007, -0.0135, 0.0295], + [-0.0153, -0.0178, 0.0362, ..., -0.0381, -0.0271, -0.0004]], + device='cuda:0'), grad: tensor([[ 1.3185e-04, 0.0000e+00, -1.1963e-04, ..., 1.0490e-05, + 2.2864e-04, 0.0000e+00], + [-1.4439e-03, 0.0000e+00, 6.2943e-05, ..., 1.4651e-04, + -2.2297e-03, 0.0000e+00], + [ 2.1629e-03, 0.0000e+00, 7.2241e-05, ..., 7.5197e-04, + 1.1454e-03, 0.0000e+00], + ..., + [ 3.1403e-02, 0.0000e+00, 3.8028e-04, ..., 1.6312e-02, + 4.7922e-04, 0.0000e+00], + [ 4.9973e-04, 0.0000e+00, 1.9255e-03, ..., 1.0657e-04, + 3.2711e-04, 0.0000e+00], + [ 1.1492e-03, 0.0000e+00, 1.2932e-03, ..., 5.6684e-05, + 3.3975e-04, 0.0000e+00]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0033, 0.0283, 0.0051, 0.0051, 0.0043, -0.0082, 0.0211, -0.0261, + 0.0245, 0.0054], device='cuda:0'), grad: tensor([-0.0007, -0.0028, 0.0026, -0.0197, -0.0010, -0.0102, 0.0006, 0.0254, + 0.0029, 0.0028], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 217.64, cls_loss 0.0826 cls_loss_mapping 0.1031 cls_loss_causal 1.0196 re_mapping 0.0416 re_causal 0.0998 /// teacc 97.44 lr 0.00010000 +Epoch 16, weight, value: tensor([[-1.7718e-02, 1.2063e-02, -1.2141e-02, ..., -5.6871e-02, + 3.2121e-02, 2.1893e-02], + [ 1.5121e-03, -5.6718e-04, -5.4833e-02, ..., 1.1570e-02, + -2.3863e-05, 1.2576e-02], + [ 2.9192e-02, -1.5483e-03, -2.7864e-02, ..., -2.0202e-02, + 1.0512e-02, -7.9110e-03], + ..., + [ 1.2028e-02, -1.3074e-02, 5.7202e-03, ..., 1.3618e-02, + 1.8653e-02, 1.1425e-02], + [ 4.3767e-03, 1.0639e-02, -1.7824e-02, ..., -8.9181e-04, + -1.3371e-02, 2.9502e-02], + [-1.6387e-02, -2.5177e-02, 3.6844e-02, ..., -3.8102e-02, + -2.6650e-02, -3.8447e-04]], device='cuda:0'), grad: tensor([[ 5.1260e-05, 0.0000e+00, -8.0913e-06, ..., 3.9674e-07, + -1.6284e-04, 0.0000e+00], + [ 3.3450e-04, 0.0000e+00, 2.2471e-04, ..., 2.7921e-06, + -3.1233e-04, 0.0000e+00], + [ 1.9503e-04, 0.0000e+00, 1.0413e-04, ..., 1.9163e-05, + 1.1146e-04, 0.0000e+00], + ..., + [-9.7370e-04, 0.0000e+00, 1.0544e-04, ..., 2.6613e-05, + -6.1214e-05, 0.0000e+00], + [ 3.0208e-04, 0.0000e+00, 2.2686e-04, ..., 1.3120e-05, + 1.8597e-04, 0.0000e+00], + [ 2.8443e-04, 0.0000e+00, -5.9414e-04, ..., 8.5756e-06, + -4.8470e-04, 0.0000e+00]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0035, 0.0285, 0.0049, 0.0049, 0.0045, -0.0085, 0.0211, -0.0261, + 0.0251, 0.0052], device='cuda:0'), grad: tensor([-1.1375e-02, 1.7452e-04, 6.6805e-04, 1.0559e-02, 7.7844e-05, + -3.0689e-03, 3.1452e-03, -7.8106e-04, 9.4700e-04, -3.5429e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 217.96, cls_loss 0.0682 cls_loss_mapping 0.0822 cls_loss_causal 0.9587 re_mapping 0.0406 re_causal 0.0976 /// teacc 97.77 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0181, 0.0045, -0.0123, ..., -0.0573, 0.0318, 0.0219], + [ 0.0012, 0.0009, -0.0553, ..., 0.0113, 0.0007, 0.0126], + [ 0.0296, -0.0034, -0.0283, ..., -0.0206, 0.0101, -0.0079], + ..., + [ 0.0129, -0.0164, 0.0054, ..., 0.0138, 0.0187, 0.0114], + [ 0.0039, 0.0084, -0.0182, ..., -0.0005, -0.0135, 0.0295], + [-0.0172, -0.0323, 0.0373, ..., -0.0381, -0.0264, -0.0004]], + device='cuda:0'), grad: tensor([[ 1.4687e-04, 0.0000e+00, 1.2720e-04, ..., 2.6543e-07, + 1.8859e-04, 0.0000e+00], + [ 1.4913e-04, 0.0000e+00, 1.7488e-04, ..., 1.1977e-06, + -1.1253e-03, 0.0000e+00], + [-6.3324e-04, 0.0000e+00, 1.1492e-04, ..., 3.8277e-07, + 1.6022e-04, 0.0000e+00], + ..., + [ 2.5234e-03, 0.0000e+00, 5.4646e-04, ..., -6.7838e-06, + 8.2111e-04, 0.0000e+00], + [ 3.8171e-04, 0.0000e+00, 1.2569e-03, ..., 3.0734e-07, + 7.5245e-04, 0.0000e+00], + [ 3.7074e-04, 0.0000e+00, -4.2114e-03, ..., 2.6487e-06, + -5.6601e-04, 0.0000e+00]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0036, 0.0284, 0.0050, 0.0047, 0.0045, -0.0085, 0.0211, -0.0256, + 0.0248, 0.0053], device='cuda:0'), grad: tensor([ 0.0004, -0.0008, -0.0018, 0.0008, 0.0044, -0.0139, 0.0076, 0.0042, + 0.0035, -0.0042], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 217.24, cls_loss 0.0656 cls_loss_mapping 0.0820 cls_loss_causal 0.9977 re_mapping 0.0380 re_causal 0.0964 /// teacc 97.29 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0176, -0.0005, -0.0122, ..., -0.0576, 0.0316, 0.0219], + [ 0.0006, 0.0024, -0.0561, ..., 0.0109, 0.0013, 0.0126], + [ 0.0301, -0.0043, -0.0290, ..., -0.0205, 0.0099, -0.0079], + ..., + [ 0.0131, -0.0207, 0.0057, ..., 0.0136, 0.0186, 0.0114], + [ 0.0035, 0.0050, -0.0187, ..., -0.0001, -0.0133, 0.0295], + [-0.0181, -0.0383, 0.0374, ..., -0.0383, -0.0263, -0.0004]], + device='cuda:0'), grad: tensor([[ 8.4162e-05, 0.0000e+00, 6.0856e-05, ..., 2.8871e-07, + 5.6237e-05, 0.0000e+00], + [-8.5354e-04, 0.0000e+00, 8.1241e-05, ..., 4.4331e-07, + -1.0977e-03, 0.0000e+00], + [-2.4509e-04, 0.0000e+00, 7.7248e-05, ..., 7.7300e-07, + 1.1319e-04, 0.0000e+00], + ..., + [-6.0678e-05, 0.0000e+00, 8.0490e-04, ..., -5.2229e-06, + 2.4199e-04, 0.0000e+00], + [ 5.5408e-04, 0.0000e+00, 1.1454e-03, ..., 2.2631e-07, + 2.9325e-04, 0.0000e+00], + [-8.4305e-04, 0.0000e+00, -4.2801e-03, ..., 1.4668e-06, + 7.2896e-05, 0.0000e+00]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0032, 0.0283, 0.0051, 0.0047, 0.0047, -0.0088, 0.0209, -0.0256, + 0.0248, 0.0053], device='cuda:0'), grad: tensor([-0.0002, -0.0039, 0.0004, 0.0022, 0.0046, -0.0015, 0.0001, 0.0017, + 0.0029, -0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 217.25, cls_loss 0.0763 cls_loss_mapping 0.0965 cls_loss_causal 0.9589 re_mapping 0.0377 re_causal 0.0918 /// teacc 97.72 lr 0.00010000 +Epoch 19, weight, value: tensor([[-1.8015e-02, -1.0963e-02, -1.2383e-02, ..., -5.7778e-02, + 3.1339e-02, 2.1893e-02], + [ 8.5515e-05, 4.4174e-03, -5.6827e-02, ..., 1.0723e-02, + 2.0767e-03, 1.2576e-02], + [ 3.0601e-02, -3.0445e-03, -2.8312e-02, ..., -2.0565e-02, + 9.7862e-03, -7.9110e-03], + ..., + [ 1.2963e-02, -2.1107e-02, 5.4822e-03, ..., 1.4224e-02, + 1.8137e-02, 1.1425e-02], + [ 3.4132e-03, -6.8826e-03, -1.9315e-02, ..., -2.9777e-04, + -1.3033e-02, 2.9502e-02], + [-1.8927e-02, -5.1199e-02, 3.7845e-02, ..., -3.8537e-02, + -2.5991e-02, -3.8447e-04]], device='cuda:0'), grad: tensor([[ 3.1853e-04, 2.1532e-06, 1.5855e-04, ..., 2.0117e-07, + 1.8072e-04, 0.0000e+00], + [ 2.7370e-04, 7.7114e-06, 6.9857e-05, ..., -1.7434e-05, + -6.2704e-05, 0.0000e+00], + [-6.3896e-04, -2.4572e-05, 1.4186e-04, ..., 1.6643e-06, + 3.7342e-05, 0.0000e+00], + ..., + [-8.5115e-04, 2.7381e-06, 1.6773e-04, ..., 7.1079e-06, + -7.5245e-04, 0.0000e+00], + [ 3.0422e-04, 3.9712e-06, 5.8651e-04, ..., 1.0869e-06, + 8.0168e-05, 0.0000e+00], + [ 7.1096e-04, 3.1386e-07, 2.6274e-04, ..., 7.3947e-07, + 4.4203e-04, 0.0000e+00]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0031, 0.0282, 0.0061, 0.0049, 0.0044, -0.0090, 0.0207, -0.0260, + 0.0248, 0.0051], device='cuda:0'), grad: tensor([ 1.1177e-03, 3.6025e-04, -2.2340e-04, -1.9180e-02, 8.4519e-05, + 1.4404e-02, 1.5745e-03, -2.0638e-03, 1.7977e-03, 2.1400e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 218.21, cls_loss 0.0601 cls_loss_mapping 0.0775 cls_loss_causal 0.9205 re_mapping 0.0346 re_causal 0.0893 /// teacc 97.79 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0183, -0.0115, -0.0128, ..., -0.0579, 0.0312, 0.0219], + [-0.0007, 0.0045, -0.0574, ..., 0.0107, 0.0024, 0.0126], + [ 0.0308, -0.0015, -0.0285, ..., -0.0207, 0.0096, -0.0079], + ..., + [ 0.0134, -0.0245, 0.0053, ..., 0.0145, 0.0179, 0.0114], + [ 0.0032, -0.0086, -0.0197, ..., -0.0002, -0.0130, 0.0295], + [-0.0196, -0.0530, 0.0381, ..., -0.0390, -0.0255, -0.0004]], + device='cuda:0'), grad: tensor([[ 1.0884e-04, 0.0000e+00, 1.6153e-05, ..., 1.6764e-08, + 6.9022e-05, 0.0000e+00], + [ 1.8191e-04, 0.0000e+00, 2.9355e-05, ..., 1.6857e-07, + 1.8477e-04, 0.0000e+00], + [-1.3094e-03, 0.0000e+00, -3.0130e-05, ..., 1.5832e-07, + 1.6534e-04, 0.0000e+00], + ..., + [ 1.2052e-04, 0.0000e+00, 2.3365e-05, ..., -1.0924e-06, + -1.6861e-03, 0.0000e+00], + [-1.8060e-05, 0.0000e+00, 1.0139e-04, ..., 1.0245e-07, + 6.1655e-04, 0.0000e+00], + [-9.4414e-05, 0.0000e+00, 6.0320e-04, ..., 2.3562e-07, + -2.7871e-04, 0.0000e+00]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0034, 0.0280, 0.0061, 0.0053, 0.0048, -0.0089, 0.0204, -0.0259, + 0.0248, 0.0049], device='cuda:0'), grad: tensor([ 4.3273e-04, 2.6550e-03, -2.4376e-03, 1.8759e-03, 7.3075e-05, + 1.4553e-03, 1.5039e-03, -1.5472e-02, 1.0910e-02, -1.0080e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 218.04, cls_loss 0.0640 cls_loss_mapping 0.0817 cls_loss_causal 0.8950 re_mapping 0.0349 re_causal 0.0851 /// teacc 98.11 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0184, -0.0115, -0.0130, ..., -0.0580, 0.0308, 0.0219], + [-0.0009, 0.0049, -0.0574, ..., 0.0107, 0.0033, 0.0126], + [ 0.0312, -0.0017, -0.0292, ..., -0.0208, 0.0092, -0.0079], + ..., + [ 0.0143, -0.0247, 0.0050, ..., 0.0148, 0.0179, 0.0114], + [ 0.0026, -0.0089, -0.0199, ..., -0.0002, -0.0128, 0.0295], + [-0.0204, -0.0534, 0.0383, ..., -0.0392, -0.0253, -0.0004]], + device='cuda:0'), grad: tensor([[-5.2547e-04, 1.7667e-06, 5.3942e-05, ..., 5.7742e-08, + 4.2856e-05, 0.0000e+00], + [ 3.6860e-04, 1.1669e-06, 1.4949e-04, ..., 1.7779e-06, + -6.7472e-05, 0.0000e+00], + [-7.4244e-04, -1.0341e-05, 3.2067e-05, ..., 3.9022e-07, + 2.6047e-05, 0.0000e+00], + ..., + [-3.2663e-04, 2.5127e-06, 1.7512e-04, ..., -5.9791e-06, + 1.5795e-04, 0.0000e+00], + [ 2.0874e-04, 1.8450e-06, 1.1164e-04, ..., 2.2445e-07, + -2.4092e-04, 0.0000e+00], + [-1.0830e-04, 3.3621e-07, 3.2806e-04, ..., 1.4734e-06, + -2.7418e-04, 0.0000e+00]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0038, 0.0283, 0.0060, 0.0048, 0.0048, -0.0084, 0.0200, -0.0256, + 0.0249, 0.0049], device='cuda:0'), grad: tensor([-0.0028, 0.0007, -0.0006, 0.0019, -0.0031, 0.0019, 0.0008, 0.0007, + -0.0004, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 217.31, cls_loss 0.0542 cls_loss_mapping 0.0705 cls_loss_causal 0.9315 re_mapping 0.0321 re_causal 0.0833 /// teacc 98.00 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0184, -0.0117, -0.0132, ..., -0.0585, 0.0304, 0.0219], + [-0.0010, 0.0058, -0.0580, ..., 0.0102, 0.0037, 0.0126], + [ 0.0317, -0.0022, -0.0298, ..., -0.0210, 0.0090, -0.0079], + ..., + [ 0.0145, -0.0249, 0.0048, ..., 0.0151, 0.0174, 0.0114], + [ 0.0020, -0.0101, -0.0203, ..., -0.0005, -0.0128, 0.0295], + [-0.0208, -0.0541, 0.0386, ..., -0.0394, -0.0246, -0.0004]], + device='cuda:0'), grad: tensor([[-9.7454e-05, 3.9972e-06, -3.6925e-05, ..., 2.0768e-07, + 3.5435e-05, 0.0000e+00], + [ 1.3280e-04, 1.9222e-05, 2.7806e-05, ..., 1.4547e-06, + -6.2227e-04, 0.0000e+00], + [-6.7043e-04, -1.7178e-04, 6.2108e-05, ..., 1.5860e-06, + 1.1045e-04, 0.0000e+00], + ..., + [ 6.2799e-04, 1.4037e-05, 9.3579e-05, ..., -7.2643e-06, + 3.0923e-04, 0.0000e+00], + [ 7.3671e-04, 8.6010e-05, 4.7803e-04, ..., 5.4855e-07, + -2.6751e-04, 0.0000e+00], + [ 5.8937e-04, 1.4156e-06, -2.1660e-04, ..., 1.5246e-06, + -2.0134e-04, 0.0000e+00]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0040, 0.0282, 0.0062, 0.0045, 0.0049, -0.0084, 0.0202, -0.0260, + 0.0250, 0.0051], device='cuda:0'), grad: tensor([-0.0005, -0.0007, -0.0005, 0.0026, 0.0002, -0.0050, 0.0004, 0.0015, + 0.0013, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 218.45, cls_loss 0.0558 cls_loss_mapping 0.0664 cls_loss_causal 0.8972 re_mapping 0.0329 re_causal 0.0796 /// teacc 98.12 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0186, -0.0118, -0.0133, ..., -0.0587, 0.0301, 0.0219], + [-0.0009, 0.0076, -0.0593, ..., 0.0102, 0.0041, 0.0126], + [ 0.0318, -0.0036, -0.0302, ..., -0.0211, 0.0086, -0.0079], + ..., + [ 0.0148, -0.0253, 0.0048, ..., 0.0162, 0.0172, 0.0114], + [ 0.0021, -0.0110, -0.0203, ..., -0.0007, -0.0126, 0.0295], + [-0.0218, -0.0545, 0.0391, ..., -0.0396, -0.0243, -0.0004]], + device='cuda:0'), grad: tensor([[ 3.8981e-04, 0.0000e+00, 1.4699e-04, ..., 2.1756e-06, + 2.6536e-04, 0.0000e+00], + [ 6.7186e-04, 0.0000e+00, 3.4070e-04, ..., 6.9365e-06, + -9.4461e-04, 0.0000e+00], + [-3.8490e-03, 0.0000e+00, 8.8835e-04, ..., 6.3591e-06, + -6.7091e-04, 0.0000e+00], + ..., + [-5.2261e-04, 0.0000e+00, 3.4733e-03, ..., 1.2529e-04, + 5.5164e-05, 0.0000e+00], + [ 3.1433e-03, 0.0000e+00, 1.5602e-03, ..., 4.8876e-05, + 1.4668e-03, 0.0000e+00], + [ 7.0763e-04, 0.0000e+00, 1.0300e-03, ..., 2.1428e-05, + 9.9838e-05, 0.0000e+00]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0037, 0.0282, 0.0060, 0.0042, 0.0050, -0.0085, 0.0200, -0.0257, + 0.0254, 0.0050], device='cuda:0'), grad: tensor([ 0.0011, 0.0006, -0.0045, 0.0013, -0.0178, 0.0050, -0.0032, 0.0054, + 0.0092, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 217.54, cls_loss 0.0515 cls_loss_mapping 0.0723 cls_loss_causal 0.9218 re_mapping 0.0304 re_causal 0.0818 /// teacc 98.03 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0190, -0.0121, -0.0134, ..., -0.0590, 0.0299, 0.0219], + [-0.0009, 0.0083, -0.0598, ..., 0.0101, 0.0048, 0.0126], + [ 0.0321, -0.0038, -0.0306, ..., -0.0207, 0.0081, -0.0079], + ..., + [ 0.0150, -0.0254, 0.0046, ..., 0.0170, 0.0171, 0.0114], + [ 0.0021, -0.0132, -0.0203, ..., -0.0009, -0.0127, 0.0295], + [-0.0229, -0.0565, 0.0390, ..., -0.0399, -0.0241, -0.0004]], + device='cuda:0'), grad: tensor([[ 8.8394e-05, 4.2655e-06, 3.4682e-06, ..., 5.5693e-06, + 2.7239e-05, 0.0000e+00], + [-1.1295e-04, -7.3290e-04, -4.3720e-05, ..., -3.6359e-05, + -5.7220e-04, 0.0000e+00], + [ 4.0527e-02, 4.5562e-04, 1.8632e-04, ..., 6.4278e-03, + 1.5032e-04, 0.0000e+00], + ..., + [ 6.4468e-04, 7.9751e-05, 3.1877e-04, ..., 3.8838e-04, + 3.8862e-04, 0.0000e+00], + [-6.1512e-04, 7.6592e-05, -3.2444e-03, ..., 3.0696e-05, + -7.5388e-04, 0.0000e+00], + [ 2.3174e-03, 1.2986e-05, 2.5196e-03, ..., 1.5192e-05, + 5.9557e-04, 0.0000e+00]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0039, 0.0285, 0.0059, 0.0042, 0.0053, -0.0083, 0.0200, -0.0256, + 0.0251, 0.0047], device='cuda:0'), grad: tensor([-0.0012, -0.0017, 0.0298, -0.0299, 0.0003, 0.0004, 0.0009, 0.0002, + -0.0079, 0.0092], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 217.11, cls_loss 0.0458 cls_loss_mapping 0.0581 cls_loss_causal 0.8614 re_mapping 0.0302 re_causal 0.0783 /// teacc 98.01 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0196, -0.0126, -0.0135, ..., -0.0592, 0.0296, 0.0219], + [-0.0011, 0.0091, -0.0608, ..., 0.0096, 0.0052, 0.0126], + [ 0.0319, -0.0037, -0.0306, ..., -0.0209, 0.0079, -0.0079], + ..., + [ 0.0156, -0.0260, 0.0052, ..., 0.0189, 0.0168, 0.0114], + [ 0.0015, -0.0175, -0.0203, ..., -0.0011, -0.0129, 0.0295], + [-0.0235, -0.0598, 0.0393, ..., -0.0400, -0.0235, -0.0004]], + device='cuda:0'), grad: tensor([[ 4.8935e-05, 0.0000e+00, 2.1607e-05, ..., 1.3225e-07, + 3.8505e-05, 0.0000e+00], + [ 1.8549e-04, 0.0000e+00, 1.1384e-05, ..., 1.1250e-06, + -2.6122e-05, 0.0000e+00], + [-8.5258e-04, 0.0000e+00, 8.4117e-06, ..., 1.3690e-06, + 5.0306e-05, 0.0000e+00], + ..., + [ 4.9067e-04, 0.0000e+00, 1.0548e-03, ..., -1.5140e-05, + 1.9670e-04, 0.0000e+00], + [ 2.7752e-04, 0.0000e+00, 5.5969e-05, ..., 8.2888e-07, + 1.3232e-04, 0.0000e+00], + [ 1.8609e-04, 0.0000e+00, -1.1787e-03, ..., 6.0052e-06, + -9.8109e-05, 0.0000e+00]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0035, 0.0285, 0.0059, 0.0044, 0.0050, -0.0086, 0.0200, -0.0250, + 0.0248, 0.0046], device='cuda:0'), grad: tensor([ 8.8573e-05, 2.4557e-04, -1.3609e-03, -4.1275e-03, 2.1636e-04, + 3.8681e-03, -4.5490e-04, 2.5425e-03, 5.7983e-04, -1.5926e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 218.00, cls_loss 0.0482 cls_loss_mapping 0.0591 cls_loss_causal 0.9073 re_mapping 0.0287 re_causal 0.0757 /// teacc 98.21 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0200, -0.0126, -0.0137, ..., -0.0600, 0.0294, 0.0219], + [-0.0016, 0.0096, -0.0609, ..., 0.0104, 0.0054, 0.0126], + [ 0.0320, -0.0038, -0.0307, ..., -0.0209, 0.0078, -0.0079], + ..., + [ 0.0163, -0.0261, 0.0048, ..., 0.0181, 0.0170, 0.0114], + [ 0.0009, -0.0183, -0.0208, ..., -0.0012, -0.0128, 0.0295], + [-0.0241, -0.0604, 0.0394, ..., -0.0390, -0.0238, -0.0004]], + device='cuda:0'), grad: tensor([[ 9.5725e-05, 0.0000e+00, 8.0466e-05, ..., 1.1489e-05, + 4.8697e-05, 0.0000e+00], + [-2.5034e-04, 0.0000e+00, 1.3340e-04, ..., -2.1994e-04, + -6.4373e-04, 0.0000e+00], + [ 8.7690e-04, 0.0000e+00, 5.6028e-04, ..., 9.1791e-05, + 7.7188e-05, 0.0000e+00], + ..., + [-3.2005e-03, 0.0000e+00, -2.3712e-02, ..., -4.6501e-03, + 3.2640e-04, 0.0000e+00], + [-3.3665e-04, 0.0000e+00, -1.1625e-03, ..., 1.3635e-05, + 9.6321e-05, 0.0000e+00], + [-9.8944e-05, 0.0000e+00, -9.7084e-04, ..., 2.3186e-05, + 8.5235e-05, 0.0000e+00]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0037, 0.0280, 0.0060, 0.0041, 0.0053, -0.0081, 0.0202, -0.0248, + 0.0245, 0.0045], device='cuda:0'), grad: tensor([-0.0005, -0.0022, 0.0017, 0.0016, 0.0291, 0.0009, -0.0002, -0.0281, + -0.0015, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 217.30, cls_loss 0.0344 cls_loss_mapping 0.0494 cls_loss_causal 0.8657 re_mapping 0.0287 re_causal 0.0770 /// teacc 98.14 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0204, -0.0127, -0.0142, ..., -0.0601, 0.0290, 0.0219], + [-0.0020, 0.0102, -0.0611, ..., 0.0110, 0.0059, 0.0126], + [ 0.0326, -0.0040, -0.0305, ..., -0.0214, 0.0074, -0.0079], + ..., + [ 0.0168, -0.0271, 0.0049, ..., 0.0186, 0.0166, 0.0114], + [ 0.0006, -0.0189, -0.0212, ..., -0.0013, -0.0127, 0.0295], + [-0.0250, -0.0607, 0.0398, ..., -0.0388, -0.0233, -0.0004]], + device='cuda:0'), grad: tensor([[-1.2600e-04, 0.0000e+00, -5.7258e-06, ..., 2.7940e-08, + 8.4005e-07, 0.0000e+00], + [-2.2388e-04, 0.0000e+00, 2.0713e-05, ..., -2.2314e-06, + -1.1387e-03, 0.0000e+00], + [ 9.0837e-05, 0.0000e+00, 1.2839e-04, ..., 1.0990e-07, + 8.0287e-05, 0.0000e+00], + ..., + [ 7.8827e-06, 0.0000e+00, 1.3566e-04, ..., 2.3469e-07, + 1.9968e-04, 0.0000e+00], + [ 2.4009e-04, 0.0000e+00, -1.1176e-04, ..., 2.1607e-07, + 3.5644e-04, 0.0000e+00], + [ 1.2875e-04, 0.0000e+00, -2.6059e-04, ..., 4.6007e-07, + 1.1408e-04, 0.0000e+00]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0038, 0.0279, 0.0062, 0.0042, 0.0051, -0.0086, 0.0203, -0.0245, + 0.0244, 0.0047], device='cuda:0'), grad: tensor([-5.7173e-04, -2.7485e-03, 6.3372e-04, -4.7188e-03, 1.0300e-03, + 4.5738e-03, 5.1320e-05, 8.1778e-04, 7.6199e-04, 1.6844e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 218.42, cls_loss 0.0370 cls_loss_mapping 0.0525 cls_loss_causal 0.8234 re_mapping 0.0278 re_causal 0.0727 /// teacc 98.29 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0206, -0.0127, -0.0145, ..., -0.0603, 0.0286, 0.0219], + [-0.0021, 0.0102, -0.0619, ..., 0.0110, 0.0065, 0.0126], + [ 0.0332, -0.0039, -0.0310, ..., -0.0213, 0.0069, -0.0079], + ..., + [ 0.0168, -0.0272, 0.0049, ..., 0.0193, 0.0162, 0.0114], + [ 0.0002, -0.0190, -0.0212, ..., -0.0014, -0.0128, 0.0295], + [-0.0255, -0.0607, 0.0398, ..., -0.0389, -0.0230, -0.0004]], + device='cuda:0'), grad: tensor([[ 1.9801e-04, 0.0000e+00, 1.0413e-04, ..., 4.6566e-08, + 6.3300e-05, 0.0000e+00], + [ 3.9062e-03, 0.0000e+00, 1.3137e-04, ..., -2.6841e-06, + 4.3106e-04, 0.0000e+00], + [-4.9362e-03, 0.0000e+00, -7.3195e-05, ..., 8.4937e-07, + -1.2693e-03, 0.0000e+00], + ..., + [-2.3804e-03, 0.0000e+00, -2.9869e-03, ..., 1.0431e-07, + -1.0738e-03, 0.0000e+00], + [ 1.0395e-03, 0.0000e+00, 5.9652e-04, ..., 4.0792e-07, + 3.7146e-04, 0.0000e+00], + [ 2.0847e-03, 0.0000e+00, 3.2883e-03, ..., 2.0489e-08, + 8.9359e-04, 0.0000e+00]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0038, 0.0280, 0.0064, 0.0041, 0.0049, -0.0080, 0.0204, -0.0248, + 0.0242, 0.0047], device='cuda:0'), grad: tensor([ 0.0017, 0.0159, -0.0332, 0.0026, -0.0008, 0.0006, 0.0009, -0.0072, + 0.0081, 0.0114], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 217.22, cls_loss 0.0374 cls_loss_mapping 0.0512 cls_loss_causal 0.8379 re_mapping 0.0267 re_causal 0.0732 /// teacc 98.14 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0209, -0.0128, -0.0149, ..., -0.0606, 0.0283, 0.0219], + [-0.0025, 0.0107, -0.0627, ..., 0.0110, 0.0073, 0.0126], + [ 0.0328, -0.0042, -0.0315, ..., -0.0215, 0.0066, -0.0079], + ..., + [ 0.0173, -0.0273, 0.0052, ..., 0.0194, 0.0156, 0.0114], + [-0.0002, -0.0196, -0.0214, ..., -0.0016, -0.0127, 0.0295], + [-0.0266, -0.0618, 0.0400, ..., -0.0389, -0.0226, -0.0004]], + device='cuda:0'), grad: tensor([[ 1.1474e-04, 2.0489e-08, 1.8191e-04, ..., 2.9244e-07, + 1.1253e-04, 0.0000e+00], + [ 4.4346e-04, 6.4261e-07, 2.1696e-04, ..., 2.5276e-06, + 2.4587e-05, 0.0000e+00], + [ 2.0778e-04, 4.6194e-07, 8.2016e-04, ..., 1.9707e-06, + 9.1195e-05, 0.0000e+00], + ..., + [ 1.2541e-04, -1.9316e-06, 3.4761e-04, ..., -2.0005e-06, + 4.7833e-05, 0.0000e+00], + [ 4.6062e-04, 9.8720e-08, 6.0797e-04, ..., 1.9651e-06, + -3.4833e-04, 0.0000e+00], + [ 3.0637e-04, 2.3097e-07, 8.7786e-04, ..., 2.0489e-06, + 1.5116e-04, 0.0000e+00]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0038, 0.0281, 0.0061, 0.0046, 0.0048, -0.0084, 0.0205, -0.0246, + 0.0241, 0.0046], device='cuda:0'), grad: tensor([ 5.0974e-04, 1.0605e-03, 1.8349e-03, -2.7809e-03, -2.4204e-03, + -1.8835e-03, 1.2693e-03, 7.3528e-04, 8.3804e-05, 1.5898e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 218.04, cls_loss 0.0366 cls_loss_mapping 0.0434 cls_loss_causal 0.8025 re_mapping 0.0265 re_causal 0.0670 /// teacc 98.47 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0212, -0.0130, -0.0154, ..., -0.0621, 0.0280, 0.0219], + [-0.0031, 0.0109, -0.0637, ..., 0.0105, 0.0080, 0.0126], + [ 0.0331, -0.0036, -0.0323, ..., -0.0217, 0.0064, -0.0079], + ..., + [ 0.0176, -0.0284, 0.0047, ..., 0.0189, 0.0152, 0.0114], + [-0.0006, -0.0206, -0.0218, ..., -0.0004, -0.0124, 0.0295], + [-0.0269, -0.0667, 0.0405, ..., -0.0392, -0.0221, -0.0004]], + device='cuda:0'), grad: tensor([[ 9.5308e-05, 0.0000e+00, 1.9133e-05, ..., 0.0000e+00, + 9.1851e-05, 0.0000e+00], + [ 1.1027e-04, 0.0000e+00, 1.6570e-05, ..., 0.0000e+00, + -6.4299e-06, 0.0000e+00], + [ 4.9210e-04, 0.0000e+00, 2.5392e-05, ..., 0.0000e+00, + 3.9411e-04, 0.0000e+00], + ..., + [-1.8203e-04, 0.0000e+00, -7.4729e-06, ..., 0.0000e+00, + 2.4661e-05, 0.0000e+00], + [ 2.7347e-04, 0.0000e+00, 1.6868e-04, ..., 0.0000e+00, + -6.8092e-04, 0.0000e+00], + [-1.5056e-04, 0.0000e+00, -3.8624e-04, ..., 0.0000e+00, + -5.8889e-05, 0.0000e+00]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0039, 0.0277, 0.0064, 0.0043, 0.0051, -0.0079, 0.0200, -0.0249, + 0.0242, 0.0047], device='cuda:0'), grad: tensor([ 2.2161e-04, 1.6427e-04, 1.3714e-03, -5.7745e-04, 8.2445e-04, + -5.4550e-04, 4.0054e-05, -2.6464e-04, -6.6137e-04, -5.7173e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 217.56, cls_loss 0.0332 cls_loss_mapping 0.0423 cls_loss_causal 0.8212 re_mapping 0.0253 re_causal 0.0703 /// teacc 98.34 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0215, -0.0131, -0.0149, ..., -0.0626, 0.0278, 0.0219], + [-0.0034, 0.0115, -0.0641, ..., 0.0104, 0.0086, 0.0126], + [ 0.0336, -0.0041, -0.0330, ..., -0.0219, 0.0061, -0.0079], + ..., + [ 0.0178, -0.0280, 0.0046, ..., 0.0190, 0.0150, 0.0114], + [-0.0012, -0.0212, -0.0222, ..., -0.0007, -0.0124, 0.0295], + [-0.0278, -0.0695, 0.0408, ..., -0.0395, -0.0219, -0.0004]], + device='cuda:0'), grad: tensor([[-7.0989e-05, 3.7253e-08, -8.7738e-05, ..., 3.7439e-07, + 3.6150e-05, 0.0000e+00], + [ 2.2411e-04, -1.8403e-06, 1.5962e-04, ..., 1.1474e-06, + 9.7752e-05, 0.0000e+00], + [-2.5868e-04, 4.3772e-07, 5.0992e-05, ..., 6.6496e-07, + 6.1274e-05, 0.0000e+00], + ..., + [-6.8069e-05, -5.2527e-07, -2.1726e-05, ..., -1.6779e-05, + 9.8199e-06, 0.0000e+00], + [ 2.5764e-05, 1.5832e-07, 5.1200e-05, ..., 2.1979e-07, + 3.4332e-05, 0.0000e+00], + [ 7.9811e-05, 3.2037e-07, 5.6028e-04, ..., 1.0386e-05, + 6.8806e-06, 0.0000e+00]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0038, 0.0277, 0.0066, 0.0045, 0.0050, -0.0076, 0.0199, -0.0249, + 0.0237, 0.0048], device='cuda:0'), grad: tensor([-2.2163e-03, 8.8215e-04, 1.9491e-05, 1.6749e-04, -8.3923e-04, + 8.0490e-04, -4.4894e-04, 1.7673e-05, 3.0446e-04, 1.3094e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 217.41, cls_loss 0.0350 cls_loss_mapping 0.0488 cls_loss_causal 0.7968 re_mapping 0.0245 re_causal 0.0674 /// teacc 97.97 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0212, -0.0132, -0.0150, ..., -0.0627, 0.0275, 0.0219], + [-0.0046, 0.0114, -0.0647, ..., 0.0103, 0.0087, 0.0126], + [ 0.0341, -0.0038, -0.0330, ..., -0.0221, 0.0058, -0.0079], + ..., + [ 0.0181, -0.0281, 0.0045, ..., 0.0193, 0.0153, 0.0114], + [-0.0012, -0.0216, -0.0222, ..., -0.0006, -0.0122, 0.0295], + [-0.0295, -0.0710, 0.0408, ..., -0.0396, -0.0222, -0.0004]], + device='cuda:0'), grad: tensor([[ 4.1604e-05, 2.1793e-07, 3.0303e-04, ..., 0.0000e+00, + 1.8805e-05, 0.0000e+00], + [ 6.8367e-05, 1.9558e-06, -6.0940e-04, ..., 0.0000e+00, + -3.4356e-04, 0.0000e+00], + [-2.9182e-04, 2.5127e-06, 1.1837e-04, ..., 0.0000e+00, + -9.7334e-05, 0.0000e+00], + ..., + [-1.2648e-04, -9.6485e-06, 2.4891e-04, ..., 0.0000e+00, + 1.1402e-04, 0.0000e+00], + [ 1.9145e-04, 6.1281e-07, 4.1699e-04, ..., 0.0000e+00, + 1.5914e-04, 0.0000e+00], + [-8.3590e-04, 4.8243e-07, -1.3245e-02, ..., 0.0000e+00, + -1.0490e-03, 0.0000e+00]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0040, 0.0270, 0.0071, 0.0045, 0.0047, -0.0075, 0.0204, -0.0248, + 0.0241, 0.0044], device='cuda:0'), grad: tensor([ 0.0004, -0.0014, -0.0006, 0.0012, 0.0153, 0.0024, 0.0020, 0.0004, + 0.0011, -0.0207], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 217.14, cls_loss 0.0289 cls_loss_mapping 0.0428 cls_loss_causal 0.8166 re_mapping 0.0234 re_causal 0.0661 /// teacc 97.91 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0214, -0.0132, -0.0152, ..., -0.0628, 0.0274, 0.0219], + [-0.0048, 0.0114, -0.0644, ..., 0.0109, 0.0093, 0.0126], + [ 0.0341, -0.0038, -0.0331, ..., -0.0221, 0.0053, -0.0079], + ..., + [ 0.0187, -0.0280, 0.0044, ..., 0.0194, 0.0150, 0.0114], + [-0.0014, -0.0218, -0.0225, ..., -0.0006, -0.0120, 0.0295], + [-0.0298, -0.0713, 0.0406, ..., -0.0400, -0.0220, -0.0004]], + device='cuda:0'), grad: tensor([[-6.7186e-04, 0.0000e+00, 6.1020e-06, ..., 0.0000e+00, + -2.2709e-04, 0.0000e+00], + [-1.3947e-04, 0.0000e+00, 2.1249e-05, ..., 0.0000e+00, + -7.4196e-04, 0.0000e+00], + [ 6.3801e-04, 0.0000e+00, 1.8165e-05, ..., 0.0000e+00, + 3.7980e-04, 0.0000e+00], + ..., + [-5.5361e-04, 0.0000e+00, 2.7210e-05, ..., 0.0000e+00, + 7.0453e-05, 0.0000e+00], + [ 6.3300e-05, 0.0000e+00, -5.3346e-05, ..., 0.0000e+00, + 1.2684e-04, 0.0000e+00], + [ 1.3009e-05, 0.0000e+00, -1.7568e-05, ..., 0.0000e+00, + -8.2791e-05, 0.0000e+00]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0037, 0.0273, 0.0070, 0.0044, 0.0052, -0.0077, 0.0198, -0.0251, + 0.0244, 0.0042], device='cuda:0'), grad: tensor([-4.0855e-03, -8.3303e-04, 1.0691e-03, 1.0443e-03, 2.8133e-04, + -1.4186e-04, 2.7294e-03, -1.8871e-04, 6.5625e-05, 5.9724e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 217.28, cls_loss 0.0318 cls_loss_mapping 0.0407 cls_loss_causal 0.7904 re_mapping 0.0228 re_causal 0.0645 /// teacc 98.26 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0207, -0.0132, -0.0152, ..., -0.0629, 0.0278, 0.0213], + [-0.0056, 0.0116, -0.0641, ..., 0.0116, 0.0101, 0.0051], + [ 0.0342, -0.0039, -0.0335, ..., -0.0221, 0.0048, -0.0112], + ..., + [ 0.0189, -0.0280, 0.0043, ..., 0.0194, 0.0145, 0.0135], + [-0.0010, -0.0219, -0.0227, ..., -0.0006, -0.0121, 0.0260], + [-0.0304, -0.0714, 0.0409, ..., -0.0401, -0.0215, -0.0013]], + device='cuda:0'), grad: tensor([[ 1.5482e-05, 0.0000e+00, 3.5241e-06, ..., 0.0000e+00, + 1.1146e-05, 0.0000e+00], + [ 1.8328e-05, 0.0000e+00, 1.4216e-05, ..., 0.0000e+00, + -2.6524e-05, 0.0000e+00], + [-1.5581e-04, 0.0000e+00, 6.1542e-06, ..., 0.0000e+00, + -5.9605e-08, 0.0000e+00], + ..., + [ 3.0220e-05, 0.0000e+00, 1.3852e-04, ..., 0.0000e+00, + 1.9014e-04, 0.0000e+00], + [ 1.1468e-04, 0.0000e+00, 1.1104e-04, ..., 0.0000e+00, + 7.5102e-05, 0.0000e+00], + [ 5.3495e-05, 0.0000e+00, -2.1905e-05, ..., 0.0000e+00, + -2.3496e-04, 0.0000e+00]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0036, 0.0275, 0.0067, 0.0043, 0.0046, -0.0077, 0.0200, -0.0251, + 0.0245, 0.0046], device='cuda:0'), grad: tensor([-2.5272e-05, 1.2353e-05, -2.8682e-04, 1.2231e-04, -2.1112e-04, + -3.1734e-04, -3.1590e-05, 4.6444e-04, 4.1318e-04, -1.4067e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 217.36, cls_loss 0.0300 cls_loss_mapping 0.0402 cls_loss_causal 0.7865 re_mapping 0.0223 re_causal 0.0598 /// teacc 98.47 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0213, -0.0133, -0.0151, ..., -0.0633, 0.0277, 0.0212], + [-0.0064, 0.0115, -0.0647, ..., 0.0116, 0.0104, 0.0050], + [ 0.0345, -0.0037, -0.0340, ..., -0.0221, 0.0044, -0.0113], + ..., + [ 0.0197, -0.0281, 0.0043, ..., 0.0200, 0.0144, 0.0136], + [-0.0013, -0.0219, -0.0228, ..., -0.0007, -0.0119, 0.0259], + [-0.0314, -0.0715, 0.0409, ..., -0.0402, -0.0214, -0.0013]], + device='cuda:0'), grad: tensor([[ 5.9217e-05, 0.0000e+00, 8.9183e-06, ..., 3.8184e-07, + 4.2111e-05, 0.0000e+00], + [ 6.6757e-05, 0.0000e+00, 1.1727e-05, ..., 3.6694e-07, + -6.7115e-05, 0.0000e+00], + [ 6.3002e-05, 0.0000e+00, 1.0349e-05, ..., 9.0338e-07, + 3.3230e-05, 0.0000e+00], + ..., + [-3.2973e-04, 0.0000e+00, 1.2197e-05, ..., 1.5423e-06, + 8.5086e-06, 0.0000e+00], + [ 8.0228e-05, 0.0000e+00, 3.7044e-05, ..., 6.1430e-06, + 3.6687e-05, 0.0000e+00], + [ 1.8787e-04, 0.0000e+00, 3.0732e-04, ..., 4.8615e-06, + -6.1095e-05, 0.0000e+00]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0034, 0.0271, 0.0068, 0.0039, 0.0048, -0.0075, 0.0202, -0.0246, + 0.0249, 0.0040], device='cuda:0'), grad: tensor([ 1.2267e-04, -1.3083e-05, 1.1319e-04, -8.4782e-04, -4.7731e-04, + 7.5340e-04, -6.8307e-05, -3.9458e-04, 1.8954e-04, 6.2323e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 217.43, cls_loss 0.0290 cls_loss_mapping 0.0377 cls_loss_causal 0.7935 re_mapping 0.0221 re_causal 0.0606 /// teacc 98.46 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0215, -0.0133, -0.0156, ..., -0.0633, 0.0272, 0.0212], + [-0.0070, 0.0115, -0.0653, ..., 0.0117, 0.0107, 0.0012], + [ 0.0346, -0.0037, -0.0343, ..., -0.0222, 0.0038, -0.0143], + ..., + [ 0.0203, -0.0281, 0.0042, ..., 0.0201, 0.0145, 0.0167], + [-0.0013, -0.0220, -0.0233, ..., -0.0007, -0.0117, 0.0255], + [-0.0322, -0.0715, 0.0413, ..., -0.0402, -0.0213, -0.0013]], + device='cuda:0'), grad: tensor([[-9.2566e-05, 0.0000e+00, -6.7890e-05, ..., 0.0000e+00, + -1.5020e-04, 2.5146e-07], + [ 1.9088e-05, 0.0000e+00, 8.3447e-06, ..., 0.0000e+00, + -1.4699e-04, 5.5507e-07], + [-8.8394e-05, 0.0000e+00, 1.3910e-05, ..., 0.0000e+00, + 3.2872e-05, 3.9116e-08], + ..., + [-5.6922e-06, 0.0000e+00, 3.2634e-05, ..., 0.0000e+00, + 1.2732e-04, -3.9786e-06], + [ 1.0467e-04, 0.0000e+00, 2.1607e-05, ..., 0.0000e+00, + 1.4806e-04, 5.2154e-08], + [ 1.1663e-03, 0.0000e+00, 5.0694e-05, ..., 0.0000e+00, + 1.3676e-03, 2.2352e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0035, 0.0269, 0.0066, 0.0038, 0.0050, -0.0077, 0.0205, -0.0244, + 0.0250, 0.0038], device='cuda:0'), grad: tensor([-7.6962e-04, -1.7488e-04, -9.6112e-06, -2.7161e-03, 1.9491e-05, + 3.1209e-04, 2.2793e-04, 2.0015e-04, 3.3307e-04, 2.5768e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 218.07, cls_loss 0.0257 cls_loss_mapping 0.0350 cls_loss_causal 0.7962 re_mapping 0.0212 re_causal 0.0601 /// teacc 98.51 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0213, -0.0133, -0.0156, ..., -0.0634, 0.0269, 0.0212], + [-0.0077, 0.0115, -0.0658, ..., 0.0117, 0.0118, -0.0022], + [ 0.0351, -0.0037, -0.0346, ..., -0.0222, 0.0034, -0.0154], + ..., + [ 0.0205, -0.0281, 0.0044, ..., 0.0203, 0.0138, 0.0195], + [-0.0016, -0.0220, -0.0236, ..., -0.0007, -0.0118, 0.0251], + [-0.0328, -0.0715, 0.0414, ..., -0.0403, -0.0207, -0.0014]], + device='cuda:0'), grad: tensor([[ 1.1921e-04, 0.0000e+00, 2.6897e-05, ..., 0.0000e+00, + 4.2379e-05, 0.0000e+00], + [ 1.5955e-03, 0.0000e+00, 5.2512e-05, ..., 0.0000e+00, + 4.6206e-04, 0.0000e+00], + [-2.1667e-03, 0.0000e+00, 1.1832e-05, ..., 0.0000e+00, + -4.3035e-04, 0.0000e+00], + ..., + [ 2.1863e-04, 0.0000e+00, 5.2023e-04, ..., 0.0000e+00, + 3.4404e-04, 0.0000e+00], + [ 5.0497e-04, 0.0000e+00, 2.1124e-04, ..., 0.0000e+00, + 1.7059e-04, 0.0000e+00], + [-2.3887e-05, 0.0000e+00, -9.7942e-04, ..., 0.0000e+00, + -7.6103e-04, 0.0000e+00]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0031, 0.0273, 0.0069, 0.0038, 0.0047, -0.0076, 0.0200, -0.0246, + 0.0248, 0.0040], device='cuda:0'), grad: tensor([ 0.0002, 0.0050, -0.0057, 0.0034, 0.0006, -0.0041, 0.0001, 0.0013, + 0.0011, -0.0020], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 36---------------------------------------------------- +epoch 36, time 218.13, cls_loss 0.0268 cls_loss_mapping 0.0352 cls_loss_causal 0.7471 re_mapping 0.0214 re_causal 0.0572 /// teacc 98.63 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0217, -0.0133, -0.0153, ..., -0.0639, 0.0265, 0.0212], + [-0.0086, 0.0113, -0.0664, ..., 0.0117, 0.0120, -0.0027], + [ 0.0356, -0.0033, -0.0352, ..., -0.0223, 0.0031, -0.0160], + ..., + [ 0.0208, -0.0283, 0.0040, ..., 0.0208, 0.0135, 0.0202], + [-0.0017, -0.0222, -0.0241, ..., -0.0008, -0.0119, 0.0249], + [-0.0337, -0.0716, 0.0412, ..., -0.0407, -0.0203, -0.0014]], + device='cuda:0'), grad: tensor([[ 4.2975e-05, 0.0000e+00, 3.8356e-05, ..., 0.0000e+00, + 1.5348e-05, 0.0000e+00], + [ 1.4806e-04, 0.0000e+00, 4.3094e-05, ..., 0.0000e+00, + -3.4600e-05, 0.0000e+00], + [ 1.9860e-04, 0.0000e+00, 6.9320e-05, ..., 0.0000e+00, + 7.6115e-05, 0.0000e+00], + ..., + [-9.0742e-04, 0.0000e+00, -1.0812e-04, ..., 0.0000e+00, + -7.3731e-05, 0.0000e+00], + [ 1.5688e-04, 0.0000e+00, 2.8682e-04, ..., 0.0000e+00, + 1.5330e-04, 0.0000e+00], + [-4.9829e-05, 0.0000e+00, -3.7217e-04, ..., 0.0000e+00, + -2.8253e-04, 0.0000e+00]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0029, 0.0267, 0.0069, 0.0039, 0.0051, -0.0077, 0.0204, -0.0243, + 0.0246, 0.0036], device='cuda:0'), grad: tensor([ 5.4747e-05, 2.1505e-04, 3.8195e-04, 6.9141e-04, 2.5535e-04, + -5.3883e-04, 2.3174e-04, -1.3437e-03, 6.3276e-04, -5.8079e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 217.49, cls_loss 0.0266 cls_loss_mapping 0.0349 cls_loss_causal 0.7600 re_mapping 0.0200 re_causal 0.0567 /// teacc 98.43 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0225, -0.0134, -0.0157, ..., -0.0642, 0.0258, 0.0212], + [-0.0089, 0.0112, -0.0672, ..., 0.0116, 0.0124, -0.0028], + [ 0.0359, -0.0031, -0.0360, ..., -0.0225, 0.0028, -0.0160], + ..., + [ 0.0210, -0.0286, 0.0036, ..., 0.0205, 0.0130, 0.0203], + [-0.0019, -0.0223, -0.0243, ..., -0.0009, -0.0117, 0.0249], + [-0.0341, -0.0718, 0.0415, ..., -0.0407, -0.0194, -0.0014]], + device='cuda:0'), grad: tensor([[-2.5177e-04, 0.0000e+00, -1.1501e-03, ..., 7.4506e-08, + -1.1665e-04, 0.0000e+00], + [ 1.7837e-05, 0.0000e+00, 6.3360e-05, ..., 1.5087e-07, + -1.1736e-04, 0.0000e+00], + [-5.0992e-05, 0.0000e+00, 1.1599e-04, ..., 2.6450e-07, + 2.2128e-05, 0.0000e+00], + ..., + [ 7.1466e-05, 0.0000e+00, 2.2244e-04, ..., 7.7859e-07, + 1.9407e-04, 0.0000e+00], + [ 1.2863e-04, 0.0000e+00, 3.6979e-04, ..., 1.5087e-07, + 2.0814e-04, 0.0000e+00], + [ 7.9632e-05, 0.0000e+00, 9.6703e-04, ..., 2.8312e-07, + -3.3236e-04, 0.0000e+00]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0038, 0.0265, 0.0064, 0.0037, 0.0051, -0.0071, 0.0204, -0.0245, + 0.0249, 0.0043], device='cuda:0'), grad: tensor([-2.2202e-03, -3.1859e-05, 1.6952e-04, -2.0468e-04, -1.1339e-03, + 4.4107e-04, 2.9254e-04, 6.3801e-04, 8.7500e-04, 1.1740e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 217.23, cls_loss 0.0263 cls_loss_mapping 0.0371 cls_loss_causal 0.7553 re_mapping 0.0206 re_causal 0.0557 /// teacc 98.47 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0214, -0.0135, -0.0156, ..., -0.0643, 0.0260, 0.0212], + [-0.0097, 0.0114, -0.0677, ..., 0.0116, 0.0128, -0.0036], + [ 0.0359, -0.0032, -0.0363, ..., -0.0226, 0.0023, -0.0161], + ..., + [ 0.0220, -0.0287, 0.0035, ..., 0.0212, 0.0127, 0.0204], + [-0.0022, -0.0224, -0.0251, ..., -0.0010, -0.0117, 0.0249], + [-0.0352, -0.0721, 0.0420, ..., -0.0409, -0.0191, -0.0014]], + device='cuda:0'), grad: tensor([[ 7.6145e-06, 0.0000e+00, 3.9190e-05, ..., 0.0000e+00, + -6.0111e-05, 0.0000e+00], + [ 5.1069e-04, 0.0000e+00, 3.2663e-05, ..., 0.0000e+00, + 1.2982e-04, 0.0000e+00], + [ 5.7936e-04, 0.0000e+00, 1.4983e-05, ..., 0.0000e+00, + 4.1693e-05, 0.0000e+00], + ..., + [-1.4238e-03, 0.0000e+00, 2.0385e-04, ..., 0.0000e+00, + 1.4901e-07, 0.0000e+00], + [ 1.3757e-04, 0.0000e+00, 1.2898e-04, ..., 0.0000e+00, + 2.5320e-04, 0.0000e+00], + [ 2.9755e-04, 0.0000e+00, -1.0824e-03, ..., 0.0000e+00, + 3.3617e-05, 0.0000e+00]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0032, 0.0264, 0.0060, 0.0038, 0.0051, -0.0071, 0.0203, -0.0242, + 0.0246, 0.0043], device='cuda:0'), grad: tensor([-0.0004, 0.0009, 0.0009, -0.0002, 0.0012, 0.0003, -0.0004, -0.0015, + 0.0009, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 217.27, cls_loss 0.0205 cls_loss_mapping 0.0267 cls_loss_causal 0.7271 re_mapping 0.0203 re_causal 0.0569 /// teacc 98.48 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0212, -0.0136, -0.0148, ..., -0.0649, 0.0257, 0.0211], + [-0.0100, 0.0114, -0.0680, ..., 0.0118, 0.0136, -0.0063], + [ 0.0361, -0.0032, -0.0363, ..., -0.0230, 0.0018, -0.0191], + ..., + [ 0.0217, -0.0287, 0.0034, ..., 0.0215, 0.0125, 0.0233], + [-0.0023, -0.0225, -0.0255, ..., -0.0010, -0.0118, 0.0228], + [-0.0360, -0.0722, 0.0422, ..., -0.0411, -0.0187, -0.0014]], + device='cuda:0'), grad: tensor([[ 7.8857e-05, 0.0000e+00, 2.6524e-05, ..., 0.0000e+00, + 7.1645e-05, 7.6927e-07], + [ 4.2826e-05, 0.0000e+00, 3.3021e-05, ..., 0.0000e+00, + 4.4823e-05, 8.7917e-07], + [-2.6917e-04, 0.0000e+00, 2.1899e-04, ..., 0.0000e+00, + 6.7353e-05, 1.0021e-05], + ..., + [-6.6566e-04, 0.0000e+00, -1.1083e-06, ..., 0.0000e+00, + -4.2379e-05, -2.4244e-05], + [ 1.4257e-04, 0.0000e+00, 6.3479e-06, ..., 0.0000e+00, + 2.2042e-04, 6.6869e-07], + [ 1.9717e-04, 0.0000e+00, -5.4455e-04, ..., 0.0000e+00, + -2.4307e-04, 8.1956e-07]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0029, 0.0266, 0.0060, 0.0037, 0.0049, -0.0067, 0.0204, -0.0242, + 0.0243, 0.0041], device='cuda:0'), grad: tensor([ 1.5390e-04, 1.7715e-04, 1.4389e-04, 8.8358e-04, 2.3401e-04, + 1.8346e-04, -3.2663e-04, -9.5177e-04, -5.2661e-05, -4.4346e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 217.22, cls_loss 0.0278 cls_loss_mapping 0.0355 cls_loss_causal 0.7402 re_mapping 0.0194 re_causal 0.0535 /// teacc 98.50 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0217, -0.0137, -0.0155, ..., -0.0652, 0.0252, 0.0212], + [-0.0104, 0.0124, -0.0687, ..., 0.0118, 0.0139, -0.0082], + [ 0.0365, -0.0038, -0.0374, ..., -0.0231, 0.0014, -0.0194], + ..., + [ 0.0220, -0.0299, 0.0028, ..., 0.0216, 0.0117, 0.0233], + [-0.0028, -0.0230, -0.0258, ..., -0.0010, -0.0123, 0.0218], + [-0.0360, -0.0725, 0.0420, ..., -0.0412, -0.0172, -0.0015]], + device='cuda:0'), grad: tensor([[ 1.2231e-04, 0.0000e+00, 1.9178e-05, ..., 2.5705e-07, + 3.7551e-05, 0.0000e+00], + [-2.3580e-04, 0.0000e+00, -1.3723e-03, ..., 5.2713e-07, + -1.8854e-03, 0.0000e+00], + [ 7.8058e-04, 0.0000e+00, 8.6844e-05, ..., 8.9034e-07, + 8.6546e-05, 0.0000e+00], + ..., + [ 2.9421e-04, 0.0000e+00, 1.8239e-04, ..., -4.6901e-06, + 1.4985e-04, 0.0000e+00], + [ 2.9325e-04, 0.0000e+00, 1.4365e-04, ..., 9.1270e-08, + 1.1625e-03, 0.0000e+00], + [ 4.1318e-04, 0.0000e+00, 1.2839e-04, ..., 3.1665e-07, + 3.6240e-04, 0.0000e+00]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0038, 0.0265, 0.0060, 0.0041, 0.0053, -0.0069, 0.0206, -0.0245, + 0.0242, 0.0043], device='cuda:0'), grad: tensor([ 0.0002, -0.0046, 0.0013, -0.0137, 0.0022, 0.0096, 0.0004, 0.0008, + 0.0024, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 217.42, cls_loss 0.0263 cls_loss_mapping 0.0359 cls_loss_causal 0.7653 re_mapping 0.0191 re_causal 0.0547 /// teacc 98.61 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0218, -0.0137, -0.0155, ..., -0.0653, 0.0249, 0.0212], + [-0.0102, 0.0125, -0.0683, ..., 0.0121, 0.0146, -0.0108], + [ 0.0366, -0.0037, -0.0374, ..., -0.0231, 0.0007, -0.0196], + ..., + [ 0.0223, -0.0300, 0.0030, ..., 0.0220, 0.0110, 0.0240], + [-0.0028, -0.0232, -0.0262, ..., -0.0011, -0.0122, 0.0206], + [-0.0373, -0.0726, 0.0423, ..., -0.0409, -0.0166, -0.0016]], + device='cuda:0'), grad: tensor([[ 1.1790e-04, 0.0000e+00, -1.4424e-04, ..., 7.0035e-07, + -5.1409e-05, 1.6764e-08], + [ 5.7489e-05, 0.0000e+00, 4.3094e-05, ..., 3.3993e-06, + -4.2230e-05, 1.1176e-08], + [-2.4452e-03, 0.0000e+00, -6.9332e-04, ..., 2.9132e-06, + 2.3738e-05, -4.3213e-07], + ..., + [ 1.6584e-03, 0.0000e+00, 1.9569e-03, ..., 1.1140e-04, + 2.1420e-07, 2.7008e-07], + [ 1.0443e-04, 0.0000e+00, 4.3392e-05, ..., 1.0524e-06, + 6.2585e-05, 5.9605e-08], + [ 1.3113e-04, 0.0000e+00, 5.0783e-04, ..., 5.1931e-06, + 3.3259e-05, 1.8626e-09]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0035, 0.0271, 0.0062, 0.0038, 0.0049, -0.0061, 0.0194, -0.0246, + 0.0241, 0.0045], device='cuda:0'), grad: tensor([-5.6553e-04, 4.1902e-05, -3.3340e-03, 4.6396e-04, -1.9264e-03, + 7.8506e-03, -7.3128e-03, 3.6144e-03, 2.8491e-04, 8.7690e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 217.23, cls_loss 0.0251 cls_loss_mapping 0.0325 cls_loss_causal 0.7533 re_mapping 0.0186 re_causal 0.0523 /// teacc 98.16 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0216, -0.0139, -0.0156, ..., -0.0655, 0.0245, 0.0213], + [-0.0109, 0.0137, -0.0686, ..., 0.0123, 0.0155, -0.0119], + [ 0.0368, -0.0045, -0.0379, ..., -0.0232, 0.0003, -0.0195], + ..., + [ 0.0226, -0.0307, 0.0028, ..., 0.0224, 0.0106, 0.0241], + [-0.0035, -0.0238, -0.0263, ..., -0.0011, -0.0123, 0.0202], + [-0.0377, -0.0730, 0.0421, ..., -0.0409, -0.0160, -0.0016]], + device='cuda:0'), grad: tensor([[-1.4746e-04, 0.0000e+00, 1.7494e-05, ..., 1.8626e-09, + 1.7494e-05, 1.1362e-07], + [ 3.8475e-05, 0.0000e+00, 2.7761e-05, ..., -2.1979e-07, + -1.2405e-05, 3.2410e-07], + [ 3.8415e-05, 0.0000e+00, 1.3888e-05, ..., 1.8626e-08, + 1.3970e-05, 3.3528e-08], + ..., + [ 5.1558e-05, 0.0000e+00, 9.4235e-05, ..., 1.1735e-07, + 1.3721e-04, -5.2080e-06], + [-4.7445e-04, 0.0000e+00, -2.2697e-04, ..., 3.3528e-08, + -2.0921e-05, 9.4995e-08], + [-5.0604e-05, 0.0000e+00, -3.7098e-04, ..., 1.3039e-08, + -3.2878e-04, 4.0196e-06]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0039, 0.0273, 0.0063, 0.0042, 0.0051, -0.0066, 0.0193, -0.0245, + 0.0241, 0.0044], device='cuda:0'), grad: tensor([-0.0003, 0.0001, 0.0003, 0.0012, 0.0005, -0.0001, 0.0003, 0.0008, + -0.0011, -0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 43---------------------------------------------------- +epoch 43, time 218.05, cls_loss 0.0216 cls_loss_mapping 0.0296 cls_loss_causal 0.7488 re_mapping 0.0182 re_causal 0.0537 /// teacc 98.70 lr 0.00010000 +Epoch 45, weight, value: tensor([[-2.1662e-02, -1.3942e-02, -1.5540e-02, ..., -6.5847e-02, + 2.4035e-02, 2.0561e-02], + [-1.1763e-02, 1.3926e-02, -6.9201e-02, ..., 1.3015e-02, + 1.5959e-02, -1.9042e-02], + [ 3.7467e-02, -4.6319e-03, -3.7659e-02, ..., -2.3430e-02, + 3.5998e-05, -1.6238e-02], + ..., + [ 2.2948e-02, -3.1033e-02, 2.3940e-03, ..., 2.2459e-02, + 1.0317e-02, 2.4452e-02], + [-3.4721e-03, -2.3909e-02, -2.6404e-02, ..., -1.1909e-03, + -1.2683e-02, 1.4079e-02], + [-3.8343e-02, -7.3186e-02, 4.2162e-02, ..., -4.1136e-02, + -1.5362e-02, -1.8112e-03]], device='cuda:0'), grad: tensor([[ 1.6704e-05, 0.0000e+00, 7.9453e-05, ..., 1.8626e-08, + 9.8169e-05, -5.2266e-06], + [ 5.4359e-05, 0.0000e+00, 2.2650e-06, ..., -2.9579e-06, + -9.3818e-05, 4.8522e-07], + [ 7.5674e-04, 0.0000e+00, 3.5539e-06, ..., 6.1467e-08, + 2.9892e-05, 1.1362e-06], + ..., + [ 1.8448e-02, 0.0000e+00, -6.4559e-06, ..., -4.6752e-07, + 1.9878e-05, 1.4808e-06], + [ 1.7715e-04, 0.0000e+00, 7.3731e-05, ..., 1.0990e-06, + 1.0115e-04, 4.2468e-07], + [ 1.3864e-04, 0.0000e+00, 4.4018e-05, ..., 1.0347e-06, + 2.5824e-05, 1.1446e-06]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0041, 0.0270, 0.0068, 0.0040, 0.0051, -0.0067, 0.0195, -0.0247, + 0.0241, 0.0047], device='cuda:0'), grad: tensor([ 3.6454e-04, -7.8022e-05, 7.7581e-04, -2.2430e-02, 1.1027e-04, + 8.7309e-04, -1.9178e-03, 2.1240e-02, 7.0715e-04, 3.5405e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 217.20, cls_loss 0.0268 cls_loss_mapping 0.0348 cls_loss_causal 0.7402 re_mapping 0.0183 re_causal 0.0492 /// teacc 98.68 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0223, -0.0140, -0.0157, ..., -0.0668, 0.0235, 0.0201], + [-0.0123, 0.0147, -0.0699, ..., 0.0135, 0.0164, -0.0214], + [ 0.0382, -0.0053, -0.0381, ..., -0.0237, -0.0004, -0.0166], + ..., + [ 0.0234, -0.0312, 0.0023, ..., 0.0235, 0.0099, 0.0263], + [-0.0044, -0.0240, -0.0268, ..., -0.0015, -0.0125, 0.0133], + [-0.0390, -0.0734, 0.0420, ..., -0.0419, -0.0142, -0.0015]], + device='cuda:0'), grad: tensor([[ 1.0669e-05, 0.0000e+00, 4.6045e-06, ..., 2.3190e-07, + 1.3091e-05, 0.0000e+00], + [ 1.4126e-04, 0.0000e+00, 4.9621e-06, ..., 7.6089e-07, + 2.8700e-05, 0.0000e+00], + [ 3.1638e-04, 0.0000e+00, 1.7406e-06, ..., 1.5926e-07, + 3.5882e-05, 0.0000e+00], + ..., + [-6.7902e-04, 0.0000e+00, 1.1735e-07, ..., -2.6859e-06, + -8.9183e-06, 0.0000e+00], + [ 2.8312e-05, 0.0000e+00, 1.7777e-05, ..., 8.1025e-08, + 2.1660e-04, 0.0000e+00], + [ 6.2466e-05, 0.0000e+00, -4.7497e-06, ..., 2.1793e-07, + -2.8229e-04, 0.0000e+00]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0042, 0.0267, 0.0071, 0.0033, 0.0050, -0.0061, 0.0196, -0.0244, + 0.0241, 0.0045], device='cuda:0'), grad: tensor([-6.1035e-04, 3.5381e-04, 7.6532e-04, 4.0317e-04, 6.7353e-05, + 3.9721e-04, 1.4015e-05, -9.4509e-04, -1.1098e-04, -3.3498e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 217.33, cls_loss 0.0208 cls_loss_mapping 0.0290 cls_loss_causal 0.7351 re_mapping 0.0180 re_causal 0.0513 /// teacc 98.69 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0228, -0.0141, -0.0160, ..., -0.0670, 0.0233, 0.0196], + [-0.0126, 0.0150, -0.0703, ..., 0.0137, 0.0169, -0.0235], + [ 0.0384, -0.0053, -0.0385, ..., -0.0238, -0.0008, -0.0166], + ..., + [ 0.0239, -0.0319, 0.0024, ..., 0.0238, 0.0096, 0.0267], + [-0.0047, -0.0242, -0.0272, ..., -0.0015, -0.0126, 0.0130], + [-0.0400, -0.0736, 0.0419, ..., -0.0418, -0.0144, -0.0011]], + device='cuda:0'), grad: tensor([[ 8.3447e-06, 1.8626e-08, 1.8269e-05, ..., 1.1269e-07, + 1.8924e-05, 1.1893e-06], + [ 1.6347e-05, 8.1956e-08, 1.2800e-05, ..., -6.3181e-06, + -5.4389e-05, 8.4285e-07], + [-1.5616e-05, -6.2399e-07, 4.8965e-05, ..., 1.2694e-06, + 4.0561e-05, -6.8657e-06], + ..., + [-2.3615e-04, 4.6566e-07, 1.6987e-05, ..., 1.6680e-06, + -8.1837e-05, -3.3863e-06], + [ 3.1978e-05, 1.6764e-08, -4.4346e-05, ..., 5.5321e-07, + -6.6936e-05, 4.3400e-06], + [ 1.6749e-04, 2.7940e-09, 6.3753e-04, ..., 3.1479e-07, + 7.5817e-05, 2.1495e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0041, 0.0264, 0.0074, 0.0033, 0.0052, -0.0056, 0.0195, -0.0239, + 0.0235, 0.0040], device='cuda:0'), grad: tensor([-3.1424e-04, 1.0319e-06, 1.4210e-04, 1.0484e-04, -1.3399e-03, + 1.2827e-04, 2.6202e-04, -4.3726e-04, -2.3150e-04, 1.6851e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 217.28, cls_loss 0.0208 cls_loss_mapping 0.0304 cls_loss_causal 0.7148 re_mapping 0.0173 re_causal 0.0485 /// teacc 98.64 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0234, -0.0142, -0.0162, ..., -0.0672, 0.0230, 0.0196], + [-0.0131, 0.0151, -0.0706, ..., 0.0152, 0.0175, -0.0253], + [ 0.0389, -0.0050, -0.0388, ..., -0.0239, -0.0016, -0.0169], + ..., + [ 0.0238, -0.0329, 0.0023, ..., 0.0241, 0.0090, 0.0277], + [-0.0052, -0.0244, -0.0277, ..., -0.0017, -0.0124, 0.0122], + [-0.0410, -0.0738, 0.0418, ..., -0.0422, -0.0139, -0.0014]], + device='cuda:0'), grad: tensor([[ 1.7062e-05, 0.0000e+00, 2.8804e-05, ..., 2.7940e-09, + 3.5763e-05, 0.0000e+00], + [ 8.6799e-06, 0.0000e+00, 1.0848e-05, ..., 3.2596e-08, + -3.9190e-05, 0.0000e+00], + [ 4.5896e-05, 0.0000e+00, 3.2723e-05, ..., 1.3970e-08, + 2.1017e-04, 0.0000e+00], + ..., + [ 1.2845e-05, 0.0000e+00, 8.6963e-05, ..., -1.5739e-07, + 3.2812e-05, 0.0000e+00], + [ 7.5638e-05, 0.0000e+00, -5.3406e-05, ..., 8.3819e-09, + -3.9005e-04, 0.0000e+00], + [ 1.3161e-04, 0.0000e+00, -6.5625e-05, ..., 6.7987e-08, + -5.2527e-06, 0.0000e+00]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0047, 0.0262, 0.0074, 0.0039, 0.0061, -0.0060, 0.0197, -0.0240, + 0.0235, 0.0036], device='cuda:0'), grad: tensor([-2.6345e-04, 1.5981e-06, 9.1696e-04, -2.0218e-03, 6.6423e-04, + -1.0996e-03, 3.2692e-03, 3.5763e-04, -2.2240e-03, 4.0007e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 217.62, cls_loss 0.0205 cls_loss_mapping 0.0240 cls_loss_causal 0.7104 re_mapping 0.0168 re_causal 0.0484 /// teacc 98.68 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0235, -0.0142, -0.0164, ..., -0.0674, 0.0224, 0.0165], + [-0.0133, 0.0151, -0.0709, ..., 0.0155, 0.0183, -0.0294], + [ 0.0385, -0.0049, -0.0396, ..., -0.0241, -0.0021, -0.0166], + ..., + [ 0.0241, -0.0333, 0.0022, ..., 0.0244, 0.0084, 0.0283], + [-0.0050, -0.0247, -0.0281, ..., -0.0018, -0.0123, 0.0110], + [-0.0410, -0.0740, 0.0420, ..., -0.0424, -0.0130, 0.0020]], + device='cuda:0'), grad: tensor([[ 3.2723e-05, 0.0000e+00, 6.3896e-05, ..., 0.0000e+00, + 2.8729e-05, 5.4110e-07], + [ 1.4234e-04, 0.0000e+00, 8.8066e-06, ..., 0.0000e+00, + -3.2149e-06, 8.9593e-07], + [ 4.6778e-04, 0.0000e+00, 5.3532e-06, ..., 0.0000e+00, + 1.2696e-04, 1.5814e-06], + ..., + [ 7.8559e-05, 0.0000e+00, 9.2626e-05, ..., 0.0000e+00, + 1.0234e-04, 1.0051e-05], + [ 1.8215e-04, 0.0000e+00, 1.2577e-05, ..., 0.0000e+00, + 4.3094e-05, 3.9581e-07], + [ 9.0778e-05, 0.0000e+00, -2.1911e-04, ..., 0.0000e+00, + -1.9205e-04, -1.6764e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0053, 0.0265, 0.0061, 0.0034, 0.0063, -0.0054, 0.0199, -0.0241, + 0.0236, 0.0044], device='cuda:0'), grad: tensor([ 0.0003, 0.0001, 0.0006, -0.0014, 0.0006, -0.0008, 0.0002, 0.0004, + 0.0003, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 217.58, cls_loss 0.0171 cls_loss_mapping 0.0232 cls_loss_causal 0.7122 re_mapping 0.0174 re_causal 0.0492 /// teacc 98.60 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0229, -0.0144, -0.0163, ..., -0.0677, 0.0216, 0.0158], + [-0.0135, 0.0156, -0.0712, ..., 0.0171, 0.0185, -0.0293], + [ 0.0386, -0.0052, -0.0398, ..., -0.0242, -0.0027, -0.0153], + ..., + [ 0.0244, -0.0335, 0.0021, ..., 0.0249, 0.0083, 0.0310], + [-0.0054, -0.0249, -0.0284, ..., -0.0019, -0.0120, 0.0078], + [-0.0418, -0.0750, 0.0423, ..., -0.0418, -0.0124, 0.0005]], + device='cuda:0'), grad: tensor([[ 5.0142e-06, 0.0000e+00, 6.1020e-06, ..., 2.1420e-08, + -1.4156e-05, 9.3132e-10], + [-6.5379e-07, 0.0000e+00, 4.9174e-06, ..., -2.3823e-06, + -2.0295e-05, 9.3132e-10], + [ 6.5267e-06, 0.0000e+00, 1.5005e-05, ..., 1.9372e-07, + 1.5408e-05, -1.2107e-08], + ..., + [ 1.2331e-05, 0.0000e+00, 2.1517e-05, ..., 1.3039e-06, + 1.7717e-05, 2.7940e-09], + [-9.4116e-05, 0.0000e+00, -2.7895e-05, ..., 1.8533e-07, + 8.9705e-05, 2.7940e-09], + [-2.3156e-05, 0.0000e+00, -5.6237e-05, ..., 2.6729e-07, + -1.2684e-04, 0.0000e+00]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0048, 0.0265, 0.0059, 0.0037, 0.0057, -0.0055, 0.0199, -0.0240, + 0.0235, 0.0046], device='cuda:0'), grad: tensor([-4.3488e-04, -5.6811e-06, 1.0598e-04, 9.5308e-05, 4.9978e-05, + 4.1246e-04, 2.6131e-04, 1.4341e-04, -4.8876e-04, -1.3745e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 49---------------------------------------------------- +epoch 49, time 218.07, cls_loss 0.0172 cls_loss_mapping 0.0231 cls_loss_causal 0.7000 re_mapping 0.0167 re_causal 0.0471 /// teacc 98.72 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0231, -0.0144, -0.0166, ..., -0.0680, 0.0211, 0.0147], + [-0.0139, 0.0156, -0.0716, ..., 0.0181, 0.0188, -0.0318], + [ 0.0396, -0.0049, -0.0398, ..., -0.0246, -0.0024, -0.0150], + ..., + [ 0.0245, -0.0342, 0.0018, ..., 0.0246, 0.0076, 0.0297], + [-0.0057, -0.0250, -0.0284, ..., -0.0021, -0.0119, 0.0064], + [-0.0426, -0.0752, 0.0422, ..., -0.0421, -0.0116, 0.0011]], + device='cuda:0'), grad: tensor([[-8.6594e-04, 0.0000e+00, 1.1921e-05, ..., 0.0000e+00, + 5.4687e-06, 0.0000e+00], + [ 4.1068e-05, 0.0000e+00, 4.1574e-05, ..., 0.0000e+00, + -7.4089e-05, 0.0000e+00], + [ 1.1700e-04, 0.0000e+00, 6.5446e-05, ..., 0.0000e+00, + 1.3456e-05, -2.7940e-09], + ..., + [-2.2620e-05, 0.0000e+00, 3.8862e-05, ..., 0.0000e+00, + 4.2856e-05, 0.0000e+00], + [ 7.3814e-04, 0.0000e+00, -8.7082e-05, ..., 0.0000e+00, + 2.8610e-05, 9.3132e-10], + [-2.6718e-05, 0.0000e+00, 5.5805e-06, ..., 0.0000e+00, + -1.3721e-04, 0.0000e+00]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0046, 0.0264, 0.0070, 0.0030, 0.0056, -0.0055, 0.0200, -0.0245, + 0.0235, 0.0046], device='cuda:0'), grad: tensor([-5.1727e-03, 7.4387e-05, 6.0892e-04, 3.2425e-03, 3.1328e-04, + -3.0842e-03, 1.1027e-04, 3.2401e-04, 3.7804e-03, -1.9515e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 217.49, cls_loss 0.0194 cls_loss_mapping 0.0260 cls_loss_causal 0.6824 re_mapping 0.0163 re_causal 0.0452 /// teacc 98.56 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0237, -0.0151, -0.0168, ..., -0.0683, 0.0208, 0.0146], + [-0.0139, 0.0150, -0.0722, ..., 0.0181, 0.0196, -0.0351], + [ 0.0401, -0.0038, -0.0403, ..., -0.0247, -0.0028, -0.0152], + ..., + [ 0.0245, -0.0348, 0.0019, ..., 0.0255, 0.0076, 0.0301], + [-0.0063, -0.0263, -0.0286, ..., -0.0021, -0.0122, 0.0049], + [-0.0436, -0.0768, 0.0415, ..., -0.0425, -0.0117, 0.0009]], + device='cuda:0'), grad: tensor([[-6.1452e-05, 0.0000e+00, -4.2200e-05, ..., 0.0000e+00, + 7.9721e-06, 1.6764e-08], + [ 4.0352e-05, 0.0000e+00, 2.9981e-05, ..., 0.0000e+00, + 8.1509e-06, 3.9116e-08], + [ 2.7388e-05, 0.0000e+00, 1.7679e-04, ..., 0.0000e+00, + 1.1191e-05, -3.2131e-07], + ..., + [-3.9756e-05, 0.0000e+00, 3.2693e-05, ..., 0.0000e+00, + 5.7667e-05, 5.7742e-08], + [ 1.0622e-04, 0.0000e+00, 1.6844e-04, ..., 0.0000e+00, + -1.8692e-04, 5.2154e-08], + [ 3.5346e-05, 0.0000e+00, 1.6868e-05, ..., 0.0000e+00, + -8.7202e-05, 3.7253e-09]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0044, 0.0265, 0.0073, 0.0034, 0.0061, -0.0054, 0.0201, -0.0242, + 0.0230, 0.0035], device='cuda:0'), grad: tensor([-2.1553e-04, 1.7560e-04, 1.6892e-04, 1.2817e-03, -1.4615e-04, + -3.2043e-03, 1.0386e-03, 1.3685e-04, 6.8235e-04, 8.4579e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 217.38, cls_loss 0.0181 cls_loss_mapping 0.0232 cls_loss_causal 0.6852 re_mapping 0.0160 re_causal 0.0452 /// teacc 98.68 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0240, -0.0152, -0.0172, ..., -0.0684, 0.0205, 0.0145], + [-0.0140, 0.0152, -0.0721, ..., 0.0189, 0.0205, -0.0359], + [ 0.0398, -0.0039, -0.0409, ..., -0.0248, -0.0036, -0.0153], + ..., + [ 0.0249, -0.0349, 0.0021, ..., 0.0253, 0.0072, 0.0312], + [-0.0060, -0.0263, -0.0290, ..., -0.0022, -0.0128, 0.0040], + [-0.0438, -0.0769, 0.0416, ..., -0.0423, -0.0112, 0.0007]], + device='cuda:0'), grad: tensor([[ 1.5512e-05, 1.1791e-06, 9.5218e-06, ..., 0.0000e+00, + 1.8209e-05, 6.9477e-07], + [ 2.7180e-05, 8.4750e-08, 3.7942e-06, ..., 0.0000e+00, + -6.8605e-05, 2.7940e-07], + [-1.0514e-04, -1.0476e-05, 1.2526e-06, ..., 0.0000e+00, + 2.2516e-05, -1.9297e-06], + ..., + [-8.1122e-05, 8.3633e-07, 4.9248e-06, ..., 0.0000e+00, + 1.2346e-05, 4.5076e-07], + [ 1.5162e-06, 5.1688e-07, 5.5701e-05, ..., 0.0000e+00, + 8.7023e-05, 4.5002e-06], + [ 4.0025e-05, 1.7881e-07, 4.8578e-06, ..., 0.0000e+00, + 6.1989e-06, 8.6799e-07]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0037, 0.0273, 0.0067, 0.0031, 0.0058, -0.0056, 0.0206, -0.0238, + 0.0227, 0.0034], device='cuda:0'), grad: tensor([ 5.4300e-05, -5.0664e-05, -1.0252e-04, 1.6451e-04, 4.5300e-05, + -7.2670e-03, 6.9962e-03, -1.2708e-04, 1.9574e-04, 8.8096e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 217.56, cls_loss 0.0145 cls_loss_mapping 0.0191 cls_loss_causal 0.7120 re_mapping 0.0158 re_causal 0.0456 /// teacc 98.68 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0241, -0.0153, -0.0176, ..., -0.0685, 0.0203, 0.0144], + [-0.0138, 0.0154, -0.0725, ..., 0.0190, 0.0211, -0.0371], + [ 0.0397, -0.0038, -0.0413, ..., -0.0249, -0.0041, -0.0159], + ..., + [ 0.0248, -0.0359, 0.0021, ..., 0.0253, 0.0066, 0.0319], + [-0.0059, -0.0269, -0.0288, ..., -0.0023, -0.0124, 0.0032], + [-0.0443, -0.0789, 0.0415, ..., -0.0426, -0.0106, 0.0006]], + device='cuda:0'), grad: tensor([[-8.9034e-07, 9.3132e-10, 3.8929e-06, ..., 8.7265e-07, + 1.2085e-05, 6.5193e-09], + [ 4.7952e-05, -6.5193e-08, 2.8968e-05, ..., 1.3120e-05, + -9.6142e-05, 3.0734e-08], + [-5.5321e-06, 8.3819e-09, 3.2540e-06, ..., 5.8860e-07, + 1.3672e-05, 3.3528e-08], + ..., + [-8.9288e-05, 4.6566e-09, -2.4661e-05, ..., -3.7491e-05, + 3.8385e-05, 7.0781e-08], + [ 3.9227e-06, 3.6322e-08, -9.1419e-06, ..., 7.3947e-07, + 1.2167e-05, 5.5879e-09], + [-1.1146e-04, 9.3132e-10, -1.2898e-04, ..., 1.4231e-05, + -9.1851e-05, 6.1467e-08]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0036, 0.0274, 0.0066, 0.0028, 0.0057, -0.0053, 0.0204, -0.0241, + 0.0234, 0.0033], device='cuda:0'), grad: tensor([-2.7001e-05, -1.7390e-05, 1.7717e-05, 8.5473e-05, 3.6478e-04, + 5.4330e-05, 8.6904e-05, -7.1585e-05, -1.2189e-05, -4.8113e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 217.60, cls_loss 0.0164 cls_loss_mapping 0.0233 cls_loss_causal 0.6930 re_mapping 0.0159 re_causal 0.0448 /// teacc 98.62 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0245, -0.0155, -0.0180, ..., -0.0689, 0.0198, 0.0144], + [-0.0141, 0.0152, -0.0729, ..., 0.0195, 0.0218, -0.0375], + [ 0.0400, -0.0033, -0.0416, ..., -0.0252, -0.0046, -0.0159], + ..., + [ 0.0251, -0.0366, 0.0018, ..., 0.0248, 0.0064, 0.0318], + [-0.0060, -0.0278, -0.0297, ..., -0.0027, -0.0125, 0.0026], + [-0.0445, -0.0798, 0.0418, ..., -0.0417, -0.0100, 0.0015]], + device='cuda:0'), grad: tensor([[ 5.4855e-07, 1.8626e-09, 2.0880e-06, ..., 9.3132e-10, + 7.0147e-06, 1.8626e-09], + [ 1.0625e-05, -3.6322e-08, 4.1723e-05, ..., 2.7008e-08, + 1.2495e-05, 2.7940e-09], + [-1.1854e-05, 2.3283e-08, 1.8954e-05, ..., 3.7253e-09, + 5.8934e-06, -2.9802e-08], + ..., + [-3.2354e-06, -7.6368e-08, 4.4376e-05, ..., -7.4506e-08, + 1.7107e-05, 1.0245e-08], + [-4.6849e-05, 1.3039e-08, 7.5586e-06, ..., 1.8626e-09, + 2.4036e-05, 6.5193e-09], + [ 9.4324e-06, 3.0734e-08, 2.6035e-04, ..., 2.2352e-08, + 8.3864e-05, 0.0000e+00]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0039, 0.0272, 0.0065, 0.0021, 0.0061, -0.0050, 0.0206, -0.0239, + 0.0228, 0.0038], device='cuda:0'), grad: tensor([-3.8713e-05, 1.0645e-04, 2.7493e-05, 6.2644e-05, -8.4352e-04, + 2.0945e-04, -2.6345e-04, 1.0085e-04, -3.8370e-07, 6.3944e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 217.37, cls_loss 0.0150 cls_loss_mapping 0.0199 cls_loss_causal 0.7068 re_mapping 0.0156 re_causal 0.0453 /// teacc 98.60 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0247, -0.0156, -0.0177, ..., -0.0693, 0.0197, 0.0145], + [-0.0148, 0.0154, -0.0735, ..., 0.0200, 0.0215, -0.0382], + [ 0.0410, -0.0032, -0.0419, ..., -0.0252, -0.0052, -0.0155], + ..., + [ 0.0249, -0.0371, 0.0017, ..., 0.0249, 0.0060, 0.0317], + [-0.0065, -0.0281, -0.0298, ..., -0.0029, -0.0117, 0.0022], + [-0.0451, -0.0802, 0.0417, ..., -0.0417, -0.0097, 0.0015]], + device='cuda:0'), grad: tensor([[ 1.1310e-05, 0.0000e+00, 1.1489e-05, ..., 2.4214e-07, + 8.5384e-06, 2.0489e-07], + [ 1.0908e-05, 0.0000e+00, 9.8133e-04, ..., 1.5553e-07, + -1.4871e-05, 1.3039e-07], + [-6.3062e-05, 0.0000e+00, 1.6928e-05, ..., 8.8476e-08, + 1.0617e-05, 7.4506e-08], + ..., + [ 7.4267e-05, 0.0000e+00, 1.1963e-04, ..., -8.7079e-07, + 2.7373e-05, -7.2457e-07], + [-7.1704e-05, 0.0000e+00, -1.0040e-06, ..., 1.3970e-08, + -5.5641e-05, 1.3970e-08], + [-2.6345e-04, 0.0000e+00, -1.4257e-04, ..., 1.2480e-07, + -1.0115e-04, 8.6613e-08]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0034, 0.0261, 0.0076, 0.0022, 0.0060, -0.0051, 0.0208, -0.0239, + 0.0226, 0.0035], device='cuda:0'), grad: tensor([-6.1512e-05, 1.6642e-03, 1.2144e-05, 6.1369e-04, -2.0256e-03, + 1.1718e-04, 2.0981e-04, 3.4213e-04, -2.2316e-04, -6.4898e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 55---------------------------------------------------- +epoch 55, time 218.09, cls_loss 0.0158 cls_loss_mapping 0.0216 cls_loss_causal 0.6807 re_mapping 0.0153 re_causal 0.0438 /// teacc 98.75 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0248, -0.0158, -0.0183, ..., -0.0694, 0.0195, 0.0146], + [-0.0153, 0.0157, -0.0749, ..., 0.0201, 0.0222, -0.0391], + [ 0.0409, -0.0034, -0.0421, ..., -0.0253, -0.0060, -0.0154], + ..., + [ 0.0251, -0.0371, 0.0017, ..., 0.0247, 0.0056, 0.0319], + [-0.0071, -0.0287, -0.0297, ..., -0.0029, -0.0114, 0.0019], + [-0.0456, -0.0806, 0.0416, ..., -0.0414, -0.0094, 0.0013]], + device='cuda:0'), grad: tensor([[ 4.9710e-05, 0.0000e+00, 4.1053e-06, ..., 0.0000e+00, + 8.2478e-06, 2.0489e-08], + [ 6.7830e-05, 0.0000e+00, 8.4657e-07, ..., 0.0000e+00, + -6.4932e-06, 3.7253e-09], + [ 1.3816e-04, 0.0000e+00, 2.0117e-06, ..., 0.0000e+00, + 6.0797e-05, -1.6019e-07], + ..., + [ 1.2720e-04, 0.0000e+00, 1.6131e-06, ..., 0.0000e+00, + 2.3380e-05, 2.4214e-08], + [ 3.1528e-03, 0.0000e+00, 1.2845e-05, ..., 0.0000e+00, + 5.5122e-04, 4.5635e-08], + [ 4.6790e-05, 0.0000e+00, -7.7952e-07, ..., 0.0000e+00, + -5.1558e-06, 1.8626e-09]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0044, 0.0262, 0.0070, 0.0028, 0.0058, -0.0049, 0.0213, -0.0238, + 0.0228, 0.0034], device='cuda:0'), grad: tensor([-4.8339e-05, 9.0957e-05, 3.6979e-04, -6.9618e-03, 2.9474e-05, + 6.2561e-04, -7.4267e-05, 2.2662e-04, 5.6686e-03, 7.0810e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 217.63, cls_loss 0.0186 cls_loss_mapping 0.0258 cls_loss_causal 0.7043 re_mapping 0.0158 re_causal 0.0433 /// teacc 98.57 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0250, -0.0158, -0.0180, ..., -0.0695, 0.0192, 0.0145], + [-0.0156, 0.0161, -0.0754, ..., 0.0215, 0.0232, -0.0394], + [ 0.0417, -0.0034, -0.0439, ..., -0.0253, -0.0064, -0.0157], + ..., + [ 0.0254, -0.0379, 0.0012, ..., 0.0248, 0.0045, 0.0322], + [-0.0078, -0.0289, -0.0301, ..., -0.0030, -0.0117, 0.0004], + [-0.0466, -0.0810, 0.0413, ..., -0.0424, -0.0096, 0.0033]], + device='cuda:0'), grad: tensor([[ 1.6138e-05, 0.0000e+00, 3.1888e-05, ..., 4.6566e-09, + 4.7117e-05, 0.0000e+00], + [-9.5248e-05, 0.0000e+00, 3.1412e-05, ..., 4.0978e-08, + -2.5892e-04, 0.0000e+00], + [ 2.8133e-04, 0.0000e+00, 4.5395e-04, ..., 8.3819e-09, + 2.1851e-04, 0.0000e+00], + ..., + [-4.9233e-05, 0.0000e+00, 5.6535e-05, ..., -1.4808e-07, + 1.3351e-04, 0.0000e+00], + [-4.2677e-04, 0.0000e+00, -5.2357e-04, ..., 3.7253e-09, + -8.2636e-04, 0.0000e+00], + [ 1.4293e-04, 0.0000e+00, 2.3687e-04, ..., 4.3772e-08, + 3.1137e-04, 0.0000e+00]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0039, 0.0267, 0.0069, 0.0029, 0.0070, -0.0045, 0.0206, -0.0240, + 0.0224, 0.0025], device='cuda:0'), grad: tensor([ 1.6534e-04, -4.2868e-04, 1.7815e-03, 7.1716e-04, -1.0710e-03, + 5.3740e-04, 1.2405e-05, 2.7108e-04, -3.0842e-03, 1.1015e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 57---------------------------------------------------- +epoch 57, time 218.30, cls_loss 0.0125 cls_loss_mapping 0.0167 cls_loss_causal 0.6787 re_mapping 0.0154 re_causal 0.0436 /// teacc 98.78 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0252, -0.0158, -0.0180, ..., -0.0697, 0.0189, 0.0145], + [-0.0163, 0.0161, -0.0760, ..., 0.0215, 0.0231, -0.0400], + [ 0.0415, -0.0034, -0.0442, ..., -0.0254, -0.0066, -0.0156], + ..., + [ 0.0263, -0.0379, 0.0009, ..., 0.0247, 0.0043, 0.0323], + [-0.0084, -0.0290, -0.0303, ..., -0.0030, -0.0117, -0.0010], + [-0.0472, -0.0810, 0.0410, ..., -0.0421, -0.0095, 0.0040]], + device='cuda:0'), grad: tensor([[ 2.0528e-04, 0.0000e+00, 8.5640e-04, ..., 1.2200e-07, + 8.0228e-05, 2.2352e-08], + [ 3.8892e-05, 0.0000e+00, 4.8578e-06, ..., 7.1153e-07, + -3.3855e-05, 2.9802e-08], + [-1.9681e-04, 0.0000e+00, 1.1034e-05, ..., 1.6112e-07, + 3.5137e-05, -4.0792e-07], + ..., + [ 2.4706e-05, 0.0000e+00, 3.2902e-05, ..., -3.7421e-06, + 1.9372e-05, 4.4703e-08], + [ 6.1870e-05, 0.0000e+00, 1.7241e-05, ..., 3.6322e-08, + 3.2216e-05, 1.0710e-07], + [ 1.3340e-04, 0.0000e+00, -2.8968e-05, ..., 1.3616e-06, + 3.5495e-05, 5.6811e-08]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0037, 0.0260, 0.0064, 0.0036, 0.0073, -0.0051, 0.0210, -0.0234, + 0.0222, 0.0023], device='cuda:0'), grad: tensor([ 6.5384e-03, 3.2693e-05, -2.4128e-04, -1.9169e-03, 6.7711e-05, + -5.8861e-03, 7.1907e-04, 1.6999e-04, 1.9586e-04, 3.2806e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 217.54, cls_loss 0.0172 cls_loss_mapping 0.0211 cls_loss_causal 0.6828 re_mapping 0.0150 re_causal 0.0416 /// teacc 98.75 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0257, -0.0160, -0.0187, ..., -0.0699, 0.0184, 0.0121], + [-0.0173, 0.0160, -0.0761, ..., 0.0232, 0.0238, -0.0407], + [ 0.0416, -0.0029, -0.0445, ..., -0.0256, -0.0071, -0.0150], + ..., + [ 0.0271, -0.0385, 0.0006, ..., 0.0239, 0.0037, 0.0321], + [-0.0088, -0.0295, -0.0307, ..., -0.0032, -0.0118, -0.0034], + [-0.0476, -0.0819, 0.0410, ..., -0.0422, -0.0087, 0.0062]], + device='cuda:0'), grad: tensor([[ 1.1347e-05, 0.0000e+00, 2.9624e-05, ..., 5.4296e-07, + 1.6227e-05, 2.1514e-07], + [ 2.5064e-05, 0.0000e+00, 3.1590e-05, ..., 2.9374e-06, + -8.3625e-05, 4.2841e-08], + [ 2.3520e-04, 0.0000e+00, 1.5926e-04, ..., 1.3541e-06, + 6.3181e-05, -2.0284e-06], + ..., + [-7.4530e-04, 0.0000e+00, -2.9373e-04, ..., -1.1764e-05, + -1.0717e-04, 5.2806e-07], + [ 1.2094e-04, 0.0000e+00, 5.5695e-04, ..., 2.0768e-07, + 2.5153e-04, 4.9639e-07], + [ 3.8266e-04, 0.0000e+00, 3.5048e-04, ..., 1.9874e-06, + 1.6320e-04, 5.0291e-08]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0046, 0.0262, 0.0061, 0.0037, 0.0078, -0.0049, 0.0213, -0.0230, + 0.0215, 0.0025], device='cuda:0'), grad: tensor([-7.2300e-05, -8.2433e-05, 6.9761e-04, 3.4237e-03, 1.0049e-04, + -6.0387e-03, 7.3135e-05, -1.3237e-03, 1.8244e-03, 1.3971e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 217.55, cls_loss 0.0166 cls_loss_mapping 0.0204 cls_loss_causal 0.6826 re_mapping 0.0141 re_causal 0.0387 /// teacc 98.65 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0254, -0.0160, -0.0193, ..., -0.0705, 0.0180, 0.0111], + [-0.0169, 0.0160, -0.0763, ..., 0.0229, 0.0251, -0.0412], + [ 0.0423, -0.0028, -0.0438, ..., -0.0259, -0.0079, -0.0139], + ..., + [ 0.0263, -0.0387, 0.0005, ..., 0.0245, 0.0034, 0.0320], + [-0.0092, -0.0297, -0.0309, ..., -0.0033, -0.0117, -0.0040], + [-0.0483, -0.0820, 0.0406, ..., -0.0423, -0.0077, 0.0070]], + device='cuda:0'), grad: tensor([[-3.3736e-05, 0.0000e+00, 2.2724e-06, ..., 1.7416e-07, + 3.0883e-06, 0.0000e+00], + [ 1.7837e-05, 0.0000e+00, 2.3052e-05, ..., 3.1590e-06, + 2.2829e-05, 0.0000e+00], + [-1.8287e-04, 0.0000e+00, 2.1011e-06, ..., 1.8347e-07, + 6.0201e-06, 0.0000e+00], + ..., + [ 1.4937e-04, 0.0000e+00, 9.5516e-06, ..., 1.8999e-07, + 4.2841e-06, 0.0000e+00], + [ 1.1183e-05, 0.0000e+00, 1.4611e-05, ..., 2.2724e-07, + 8.9109e-06, 0.0000e+00], + [ 6.5155e-06, 0.0000e+00, 2.1946e-04, ..., 3.0637e-05, + 2.4462e-04, 0.0000e+00]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0037, 0.0270, 0.0065, 0.0033, 0.0075, -0.0048, 0.0207, -0.0236, + 0.0217, 0.0021], device='cuda:0'), grad: tensor([-1.5318e-04, 1.0294e-04, -2.0874e-04, 4.4703e-05, -2.3991e-05, + -3.2872e-05, -7.9489e-04, 2.3258e-04, 7.4685e-05, 7.5865e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 217.42, cls_loss 0.0137 cls_loss_mapping 0.0218 cls_loss_causal 0.6739 re_mapping 0.0142 re_causal 0.0422 /// teacc 98.61 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0256, -0.0160, -0.0195, ..., -0.0706, 0.0177, 0.0111], + [-0.0171, 0.0160, -0.0770, ..., 0.0231, 0.0256, -0.0414], + [ 0.0417, -0.0027, -0.0439, ..., -0.0260, -0.0082, -0.0138], + ..., + [ 0.0272, -0.0389, 0.0004, ..., 0.0248, 0.0031, 0.0320], + [-0.0095, -0.0297, -0.0307, ..., -0.0034, -0.0116, -0.0042], + [-0.0486, -0.0820, 0.0405, ..., -0.0426, -0.0075, 0.0070]], + device='cuda:0'), grad: tensor([[ 2.0191e-05, 0.0000e+00, 1.8001e-05, ..., 1.0617e-05, + 9.7305e-06, 0.0000e+00], + [ 2.0862e-04, 0.0000e+00, 2.8238e-05, ..., -3.7265e-04, + -2.3079e-04, 0.0000e+00], + [ 2.2903e-05, 0.0000e+00, 6.5118e-06, ..., 1.6198e-05, + 4.0196e-06, 0.0000e+00], + ..., + [-1.2789e-03, 0.0000e+00, 8.1062e-05, ..., -3.9506e-04, + 1.4150e-04, 0.0000e+00], + [ 1.8105e-05, 0.0000e+00, 9.6202e-05, ..., 1.5512e-05, + 1.6421e-05, 0.0000e+00], + [ 7.0238e-04, 0.0000e+00, -1.8911e-03, ..., 4.2415e-04, + -2.1935e-05, 0.0000e+00]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0035, 0.0276, 0.0059, 0.0031, 0.0076, -0.0045, 0.0199, -0.0230, + 0.0217, 0.0017], device='cuda:0'), grad: tensor([ 8.7202e-05, -2.6779e-03, 1.0395e-04, 2.7990e-04, 4.8065e-03, + -1.0052e-03, 8.5497e-04, -1.5268e-03, 2.6345e-04, -1.1806e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 217.58, cls_loss 0.0129 cls_loss_mapping 0.0180 cls_loss_causal 0.6543 re_mapping 0.0144 re_causal 0.0404 /// teacc 98.73 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0264, -0.0161, -0.0200, ..., -0.0708, 0.0174, 0.0110], + [-0.0186, 0.0160, -0.0776, ..., 0.0244, 0.0262, -0.0416], + [ 0.0426, -0.0027, -0.0440, ..., -0.0261, -0.0085, -0.0137], + ..., + [ 0.0280, -0.0389, 0.0004, ..., 0.0248, 0.0028, 0.0320], + [-0.0099, -0.0298, -0.0309, ..., -0.0035, -0.0125, -0.0043], + [-0.0492, -0.0821, 0.0408, ..., -0.0434, -0.0069, 0.0070]], + device='cuda:0'), grad: tensor([[ 1.1988e-05, 0.0000e+00, 6.2250e-06, ..., 4.4219e-06, + 1.3508e-05, 9.5461e-07], + [ 2.5082e-04, 0.0000e+00, 1.0999e-06, ..., 1.0228e-04, + -3.6192e-04, 3.7812e-07], + [ 1.3149e-04, 0.0000e+00, 2.9393e-06, ..., 2.8592e-06, + 5.7220e-05, 2.0675e-07], + ..., + [-2.6727e-04, 0.0000e+00, 9.2462e-06, ..., -4.9973e-04, + -3.0804e-04, 6.6962e-07], + [ 8.1360e-06, 0.0000e+00, 1.0498e-05, ..., 5.7146e-06, + 2.4289e-05, 1.7351e-06], + [ 1.4722e-04, 0.0000e+00, 1.9461e-05, ..., 1.4126e-04, + 1.3447e-04, 3.5390e-08]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0045, 0.0273, 0.0065, 0.0025, 0.0070, -0.0043, 0.0207, -0.0226, + 0.0212, 0.0024], device='cuda:0'), grad: tensor([ 5.1260e-05, -3.9196e-04, 2.0123e-04, -3.1924e-04, 1.1377e-03, + -4.2605e-04, 3.1304e-04, -1.0786e-03, 1.2651e-05, 5.0116e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 217.42, cls_loss 0.0098 cls_loss_mapping 0.0155 cls_loss_causal 0.6317 re_mapping 0.0139 re_causal 0.0408 /// teacc 98.67 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0267, -0.0161, -0.0201, ..., -0.0709, 0.0170, 0.0106], + [-0.0182, 0.0160, -0.0781, ..., 0.0245, 0.0267, -0.0451], + [ 0.0432, -0.0027, -0.0442, ..., -0.0263, -0.0086, -0.0112], + ..., + [ 0.0278, -0.0389, 0.0005, ..., 0.0252, 0.0024, 0.0331], + [-0.0105, -0.0298, -0.0310, ..., -0.0036, -0.0123, -0.0077], + [-0.0498, -0.0821, 0.0407, ..., -0.0436, -0.0066, 0.0070]], + device='cuda:0'), grad: tensor([[ 6.0536e-07, 0.0000e+00, -3.2447e-06, ..., 2.3562e-07, + 2.0694e-06, 0.0000e+00], + [ 2.2396e-05, 0.0000e+00, 1.0796e-05, ..., 4.9584e-06, + 9.0897e-07, 0.0000e+00], + [-2.8551e-05, 0.0000e+00, 3.1125e-06, ..., 6.1560e-07, + 3.4608e-06, 0.0000e+00], + ..., + [-3.1620e-05, 0.0000e+00, 8.2999e-06, ..., -9.4622e-06, + 3.0138e-06, 0.0000e+00], + [ 5.6699e-06, 0.0000e+00, 6.8732e-06, ..., 1.3225e-07, + -5.4669e-07, 0.0000e+00], + [ 1.3776e-05, 0.0000e+00, 6.4913e-07, ..., 2.0433e-06, + -9.7081e-06, 0.0000e+00]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0043, 0.0275, 0.0069, 0.0023, 0.0069, -0.0045, 0.0206, -0.0225, + 0.0213, 0.0020], device='cuda:0'), grad: tensor([-6.9439e-05, 5.7817e-05, -2.4870e-05, 1.9029e-05, -3.3706e-05, + 2.5228e-05, 1.1884e-05, -2.7746e-05, 1.3225e-05, 2.8580e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 217.33, cls_loss 0.0105 cls_loss_mapping 0.0174 cls_loss_causal 0.6285 re_mapping 0.0139 re_causal 0.0398 /// teacc 98.61 lr 0.00010000 +Epoch 65, weight, value: tensor([[-2.7072e-02, -1.6177e-02, -2.0683e-02, ..., -7.1142e-02, + 1.6065e-02, 1.0269e-02], + [-1.8219e-02, 1.6318e-02, -7.8408e-02, ..., 2.4330e-02, + 2.6914e-02, -4.5978e-02], + [ 4.3092e-02, -3.2435e-03, -4.4140e-02, ..., -2.6427e-02, + -9.2436e-03, -1.0578e-02], + ..., + [ 2.8469e-02, -3.7164e-02, 9.7241e-05, ..., 2.5689e-02, + 2.2677e-03, 3.3124e-02], + [-1.1044e-02, -2.8373e-02, -3.1328e-02, ..., -3.7202e-03, + -1.2169e-02, -8.0873e-03], + [-5.0665e-02, -8.2488e-02, 4.0945e-02, ..., -4.3980e-02, + -5.8393e-03, 7.2448e-03]], device='cuda:0'), grad: tensor([[-4.5806e-05, 0.0000e+00, -6.1654e-06, ..., 0.0000e+00, + -8.8692e-05, 2.5891e-07], + [-2.2078e-04, 0.0000e+00, 3.5428e-06, ..., -7.4506e-09, + -1.3590e-04, 8.4750e-08], + [ 1.2517e-04, 0.0000e+00, 9.6038e-06, ..., 9.3132e-10, + 8.9407e-05, -2.4959e-06], + ..., + [-1.9395e-04, 0.0000e+00, 1.1474e-05, ..., 2.7940e-09, + 2.5004e-05, 2.8219e-07], + [ 3.0190e-05, 0.0000e+00, -9.6262e-05, ..., 9.3132e-10, + 2.9400e-05, 9.4995e-07], + [ 9.4056e-05, 0.0000e+00, 3.0756e-05, ..., 9.3132e-10, + 1.1668e-05, 2.9430e-07]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0056, 0.0274, 0.0066, 0.0022, 0.0069, -0.0040, 0.0208, -0.0223, + 0.0213, 0.0025], device='cuda:0'), grad: tensor([-6.9189e-04, -4.8566e-04, 6.0892e-04, 4.1437e-04, -1.0639e-05, + 3.4142e-04, 4.0531e-05, -2.5105e-04, -3.0065e-04, 3.3569e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 217.47, cls_loss 0.0115 cls_loss_mapping 0.0163 cls_loss_causal 0.6585 re_mapping 0.0138 re_causal 0.0398 /// teacc 98.60 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0272, -0.0163, -0.0210, ..., -0.0715, 0.0157, 0.0102], + [-0.0181, 0.0168, -0.0790, ..., 0.0249, 0.0274, -0.0474], + [ 0.0430, -0.0037, -0.0443, ..., -0.0267, -0.0095, -0.0101], + ..., + [ 0.0285, -0.0373, -0.0003, ..., 0.0255, 0.0015, 0.0334], + [-0.0114, -0.0279, -0.0316, ..., -0.0040, -0.0121, -0.0090], + [-0.0507, -0.0830, 0.0411, ..., -0.0438, -0.0049, 0.0073]], + device='cuda:0'), grad: tensor([[ 3.2187e-05, 0.0000e+00, 7.4133e-07, ..., 4.4703e-08, + 6.4336e-06, 3.7253e-09], + [ 1.0520e-04, 0.0000e+00, -1.7323e-07, ..., 3.7067e-07, + -4.6074e-05, 1.8626e-09], + [-1.4615e-04, 0.0000e+00, 3.5223e-06, ..., 6.5193e-08, + 1.0267e-05, 0.0000e+00], + ..., + [-1.4150e-04, 0.0000e+00, 3.3677e-06, ..., 3.0547e-07, + -3.5204e-07, 1.8626e-09], + [ 1.1258e-05, 0.0000e+00, -1.8906e-06, ..., 8.1956e-08, + 2.2426e-05, 1.6764e-08], + [ 4.5061e-05, 0.0000e+00, 9.7603e-06, ..., 2.1700e-06, + 2.8387e-06, 7.4506e-09]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0053, 0.0279, 0.0062, 0.0022, 0.0073, -0.0040, 0.0208, -0.0230, + 0.0211, 0.0029], device='cuda:0'), grad: tensor([ 4.7386e-05, 1.1557e-04, -1.2052e-04, 1.2934e-04, 3.9935e-06, + -3.7272e-06, 4.7654e-05, -3.1686e-04, 5.4836e-05, 4.2111e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 217.73, cls_loss 0.0131 cls_loss_mapping 0.0205 cls_loss_causal 0.6611 re_mapping 0.0136 re_causal 0.0398 /// teacc 98.61 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0276, -0.0163, -0.0225, ..., -0.0717, 0.0154, 0.0086], + [-0.0182, 0.0173, -0.0797, ..., 0.0261, 0.0281, -0.0498], + [ 0.0430, -0.0041, -0.0445, ..., -0.0268, -0.0098, -0.0102], + ..., + [ 0.0291, -0.0373, -0.0002, ..., 0.0245, 0.0011, 0.0325], + [-0.0117, -0.0280, -0.0316, ..., -0.0040, -0.0123, -0.0114], + [-0.0516, -0.0832, 0.0411, ..., -0.0444, -0.0043, 0.0091]], + device='cuda:0'), grad: tensor([[ 2.9281e-06, 0.0000e+00, -4.0010e-06, ..., 2.9802e-08, + 9.4622e-07, 0.0000e+00], + [ 5.2080e-06, 0.0000e+00, 8.5607e-06, ..., 2.2911e-07, + -5.0589e-06, 0.0000e+00], + [ 1.6773e-04, 0.0000e+00, 1.8571e-06, ..., 1.4901e-08, + 4.1246e-05, 0.0000e+00], + ..., + [ 4.6015e-04, 0.0000e+00, 4.7415e-05, ..., -8.4005e-07, + 1.1665e-04, 0.0000e+00], + [ 2.2560e-05, 0.0000e+00, 4.4592e-06, ..., 9.3132e-09, + 1.9707e-06, 0.0000e+00], + [-4.7588e-04, 0.0000e+00, 9.3818e-05, ..., 4.3772e-07, + -1.2022e-04, 0.0000e+00]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0062, 0.0282, 0.0058, 0.0026, 0.0072, -0.0043, 0.0209, -0.0223, + 0.0210, 0.0028], device='cuda:0'), grad: tensor([-4.2653e-04, 3.5793e-05, 2.2519e-04, -1.6630e-04, -3.9911e-04, + -6.3926e-06, 4.6432e-05, 1.2360e-03, 3.9428e-05, -5.8317e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 217.57, cls_loss 0.0108 cls_loss_mapping 0.0163 cls_loss_causal 0.6815 re_mapping 0.0131 re_causal 0.0391 /// teacc 98.44 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0278, -0.0164, -0.0225, ..., -0.0718, 0.0154, 0.0085], + [-0.0188, 0.0172, -0.0803, ..., 0.0262, 0.0280, -0.0518], + [ 0.0430, -0.0038, -0.0447, ..., -0.0269, -0.0103, -0.0097], + ..., + [ 0.0290, -0.0376, -0.0004, ..., 0.0245, 0.0009, 0.0319], + [-0.0116, -0.0282, -0.0319, ..., -0.0040, -0.0119, -0.0129], + [-0.0521, -0.0835, 0.0407, ..., -0.0446, -0.0044, 0.0092]], + device='cuda:0'), grad: tensor([[-4.2558e-05, 0.0000e+00, -1.0401e-04, ..., 1.7323e-06, + 1.0937e-05, 0.0000e+00], + [ 2.0385e-05, 0.0000e+00, 2.5779e-06, ..., 7.1898e-07, + 7.6070e-06, 0.0000e+00], + [-4.4405e-06, 0.0000e+00, 2.7329e-05, ..., 2.0117e-07, + 1.4424e-05, 0.0000e+00], + ..., + [ 1.9610e-05, 0.0000e+00, 6.6310e-06, ..., 4.5672e-06, + 4.9099e-06, 0.0000e+00], + [ 4.4078e-05, 0.0000e+00, 2.2978e-05, ..., 5.1223e-07, + 3.1263e-05, 0.0000e+00], + [ 1.4924e-05, 0.0000e+00, 8.5980e-06, ..., 1.3504e-06, + -5.0217e-05, 0.0000e+00]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0052, 0.0277, 0.0054, 0.0029, 0.0079, -0.0038, 0.0199, -0.0224, + 0.0219, 0.0017], device='cuda:0'), grad: tensor([-2.2864e-04, 4.4703e-05, 6.5088e-05, -1.7500e-04, 6.8188e-04, + 8.3596e-06, -5.8270e-04, 7.8678e-05, 1.7023e-04, -6.3062e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 217.62, cls_loss 0.0130 cls_loss_mapping 0.0167 cls_loss_causal 0.6240 re_mapping 0.0135 re_causal 0.0360 /// teacc 98.68 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0273, -0.0171, -0.0232, ..., -0.0719, 0.0148, 0.0083], + [-0.0195, 0.0175, -0.0807, ..., 0.0266, 0.0282, -0.0539], + [ 0.0436, -0.0041, -0.0447, ..., -0.0273, -0.0102, -0.0084], + ..., + [ 0.0291, -0.0375, -0.0004, ..., 0.0244, 0.0005, 0.0320], + [-0.0122, -0.0285, -0.0323, ..., -0.0042, -0.0117, -0.0147], + [-0.0524, -0.0865, 0.0411, ..., -0.0446, -0.0034, 0.0093]], + device='cuda:0'), grad: tensor([[ 8.6054e-07, 0.0000e+00, 5.7556e-07, ..., 0.0000e+00, + 4.5784e-06, 1.4901e-08], + [ 8.1062e-06, 0.0000e+00, 7.7114e-07, ..., 0.0000e+00, + 2.4028e-06, 3.7253e-09], + [-4.3273e-05, 0.0000e+00, 6.4075e-07, ..., 0.0000e+00, + -8.1807e-06, -1.8999e-07], + ..., + [-7.8604e-06, 0.0000e+00, 1.1940e-06, ..., 0.0000e+00, + 1.1660e-06, 2.6077e-08], + [ 7.7963e-05, 0.0000e+00, -3.0361e-07, ..., 0.0000e+00, + -4.7296e-05, 6.7055e-08], + [ 7.4506e-06, 0.0000e+00, 2.8443e-06, ..., 0.0000e+00, + -2.8815e-06, 1.1176e-08]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0056, 0.0271, 0.0061, 0.0027, 0.0071, -0.0043, 0.0207, -0.0226, + 0.0216, 0.0029], device='cuda:0'), grad: tensor([ 4.1723e-06, 3.2455e-05, -3.0112e-04, -6.1929e-05, 3.6150e-05, + 2.8145e-06, 5.7995e-05, 2.3823e-06, 2.0814e-04, 1.8746e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 217.61, cls_loss 0.0105 cls_loss_mapping 0.0142 cls_loss_causal 0.6285 re_mapping 0.0132 re_causal 0.0378 /// teacc 98.70 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0276, -0.0177, -0.0233, ..., -0.0720, 0.0147, 0.0083], + [-0.0199, 0.0174, -0.0809, ..., 0.0267, 0.0285, -0.0540], + [ 0.0452, -0.0029, -0.0452, ..., -0.0276, -0.0104, -0.0083], + ..., + [ 0.0280, -0.0391, -0.0008, ..., 0.0243, 0.0003, 0.0320], + [-0.0127, -0.0294, -0.0326, ..., -0.0043, -0.0117, -0.0150], + [-0.0531, -0.0898, 0.0410, ..., -0.0447, -0.0032, 0.0093]], + device='cuda:0'), grad: tensor([[-2.1040e-05, 1.4901e-08, 2.5287e-05, ..., 5.8301e-07, + 1.3039e-06, 0.0000e+00], + [-3.9265e-06, 1.8626e-09, 3.8929e-06, ..., 2.6822e-07, + -2.9355e-05, 0.0000e+00], + [ 2.5421e-05, -1.9185e-07, 1.0908e-05, ..., 2.4028e-07, + 6.5304e-06, 0.0000e+00], + ..., + [-3.5763e-05, 5.7742e-08, 3.8091e-06, ..., 4.4703e-07, + 1.3202e-05, 0.0000e+00], + [ 4.1649e-06, 3.5390e-08, 2.1327e-06, ..., 6.3889e-07, + 1.5311e-06, 0.0000e+00], + [ 9.0748e-06, 1.8626e-09, 2.4605e-04, ..., 9.0152e-07, + -4.3884e-06, 0.0000e+00]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0059, 0.0271, 0.0073, 0.0026, 0.0070, -0.0035, 0.0207, -0.0235, + 0.0211, 0.0028], device='cuda:0'), grad: tensor([-5.0545e-05, -5.3406e-05, 7.9989e-05, 2.6679e-04, -6.3086e-04, + -1.8489e-04, 1.1188e-04, -2.3812e-05, -5.2899e-06, 4.9019e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 217.57, cls_loss 0.0124 cls_loss_mapping 0.0166 cls_loss_causal 0.6367 re_mapping 0.0129 re_causal 0.0360 /// teacc 98.55 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0278, -0.0180, -0.0226, ..., -0.0724, 0.0145, 0.0083], + [-0.0206, 0.0190, -0.0813, ..., 0.0269, 0.0288, -0.0541], + [ 0.0452, -0.0032, -0.0453, ..., -0.0278, -0.0111, -0.0083], + ..., + [ 0.0292, -0.0418, -0.0010, ..., 0.0238, 0.0003, 0.0320], + [-0.0134, -0.0293, -0.0328, ..., -0.0043, -0.0119, -0.0150], + [-0.0540, -0.0909, 0.0407, ..., -0.0438, -0.0030, 0.0093]], + device='cuda:0'), grad: tensor([[ 9.9465e-06, 1.1735e-07, 5.9120e-06, ..., 0.0000e+00, + 3.9116e-06, 0.0000e+00], + [ 5.7316e-04, 3.5204e-07, 1.4305e-05, ..., 0.0000e+00, + 1.9228e-04, 0.0000e+00], + [ 1.3344e-05, -3.5763e-06, 9.8124e-06, ..., 0.0000e+00, + 4.1306e-05, 0.0000e+00], + ..., + [-6.3467e-04, 2.5537e-06, -1.3620e-05, ..., 0.0000e+00, + -2.1434e-04, 0.0000e+00], + [ 4.4584e-05, 2.5891e-07, 3.2842e-05, ..., 0.0000e+00, + 8.7693e-06, 0.0000e+00], + [ 1.7524e-05, 7.6368e-08, 7.8306e-06, ..., 0.0000e+00, + -2.5794e-05, 0.0000e+00]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0054, 0.0273, 0.0069, 0.0026, 0.0068, -0.0034, 0.0209, -0.0228, + 0.0209, 0.0021], device='cuda:0'), grad: tensor([ 2.7493e-05, 1.0395e-03, -3.6693e-04, 1.1736e-04, 5.2404e-04, + -1.9684e-03, 1.3933e-03, -1.0023e-03, 2.4009e-04, -2.1681e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 217.65, cls_loss 0.0096 cls_loss_mapping 0.0137 cls_loss_causal 0.6347 re_mapping 0.0125 re_causal 0.0354 /// teacc 98.60 lr 0.00010000 +Epoch 72, weight, value: tensor([[-2.8076e-02, -1.8209e-02, -2.2606e-02, ..., -7.2785e-02, + 1.3884e-02, 8.2790e-03], + [-2.1000e-02, 1.8953e-02, -8.1715e-02, ..., 2.7560e-02, + 2.9278e-02, -5.4273e-02], + [ 4.5328e-02, -2.9603e-03, -4.5887e-02, ..., -2.8830e-02, + -1.1658e-02, -7.9245e-03], + ..., + [ 2.9706e-02, -4.2213e-02, -5.4599e-05, ..., 2.3662e-02, + -8.8294e-05, 3.1973e-02], + [-1.3631e-02, -2.9744e-02, -3.3184e-02, ..., -4.5730e-03, + -1.1871e-02, -1.5075e-02], + [-5.4218e-02, -9.1713e-02, 4.1074e-02, ..., -4.3564e-02, + -2.3413e-03, 9.3247e-03]], device='cuda:0'), grad: tensor([[-5.3257e-05, 0.0000e+00, 4.0978e-05, ..., 1.2480e-07, + 6.3553e-06, 0.0000e+00], + [ 5.7034e-06, 0.0000e+00, 9.4920e-06, ..., -1.6868e-05, + -2.9728e-05, 0.0000e+00], + [ 8.7470e-06, 0.0000e+00, 1.0051e-05, ..., 1.1176e-06, + 5.6326e-06, -9.3132e-09], + ..., + [ 5.7891e-06, 0.0000e+00, 5.3674e-05, ..., 3.6340e-06, + 5.5254e-05, 0.0000e+00], + [ 2.6226e-05, 0.0000e+00, 6.0536e-06, ..., 2.4773e-07, + -4.4554e-06, 1.8626e-09], + [ 6.3255e-06, 0.0000e+00, 6.3837e-05, ..., 9.2015e-07, + -4.6134e-05, 0.0000e+00]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0056, 0.0270, 0.0067, 0.0025, 0.0060, -0.0033, 0.0209, -0.0222, + 0.0209, 0.0024], device='cuda:0'), grad: tensor([-3.5524e-04, 6.8918e-08, 1.2147e-04, 3.8385e-04, -1.1522e-04, + -6.2704e-04, -1.8954e-05, 2.8110e-04, 2.0301e-04, 1.2720e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 217.80, cls_loss 0.0099 cls_loss_mapping 0.0126 cls_loss_causal 0.6549 re_mapping 0.0128 re_causal 0.0372 /// teacc 98.77 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0276, -0.0184, -0.0229, ..., -0.0733, 0.0139, 0.0083], + [-0.0213, 0.0189, -0.0816, ..., 0.0293, 0.0302, -0.0545], + [ 0.0446, -0.0027, -0.0462, ..., -0.0308, -0.0128, -0.0083], + ..., + [ 0.0303, -0.0423, -0.0002, ..., 0.0235, -0.0001, 0.0324], + [-0.0142, -0.0300, -0.0337, ..., -0.0053, -0.0121, -0.0152], + [-0.0547, -0.0924, 0.0411, ..., -0.0437, -0.0021, 0.0093]], + device='cuda:0'), grad: tensor([[ 1.6242e-06, 0.0000e+00, -1.7297e-04, ..., 5.3272e-07, + 1.2666e-06, 0.0000e+00], + [ 5.1230e-05, 0.0000e+00, 5.2564e-06, ..., 4.6402e-05, + -1.1206e-05, 0.0000e+00], + [-5.2661e-05, 0.0000e+00, 4.1425e-06, ..., 6.0163e-07, + 8.9779e-07, 0.0000e+00], + ..., + [-5.5909e-05, 0.0000e+00, 1.0967e-05, ..., -5.7161e-05, + 9.8050e-06, 0.0000e+00], + [ 1.1679e-06, 0.0000e+00, 1.2025e-05, ..., 3.9674e-07, + 2.4810e-06, 0.0000e+00], + [ 4.8950e-06, 0.0000e+00, 1.2124e-04, ..., 6.9030e-06, + -1.0848e-05, 0.0000e+00]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0045, 0.0273, 0.0058, 0.0027, 0.0059, -0.0034, 0.0206, -0.0217, + 0.0205, 0.0024], device='cuda:0'), grad: tensor([-7.5912e-04, 5.9128e-04, -6.1035e-05, 9.1970e-05, -3.3379e-04, + 3.0175e-05, 6.6280e-04, -6.4182e-04, 5.1618e-05, 3.6740e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 217.57, cls_loss 0.0129 cls_loss_mapping 0.0171 cls_loss_causal 0.6663 re_mapping 0.0127 re_causal 0.0349 /// teacc 98.62 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0277, -0.0190, -0.0246, ..., -0.0738, 0.0134, 0.0078], + [-0.0223, 0.0185, -0.0821, ..., 0.0293, 0.0309, -0.0553], + [ 0.0448, -0.0023, -0.0465, ..., -0.0308, -0.0133, -0.0077], + ..., + [ 0.0310, -0.0423, -0.0001, ..., 0.0239, -0.0003, 0.0325], + [-0.0142, -0.0309, -0.0344, ..., -0.0055, -0.0121, -0.0163], + [-0.0555, -0.0943, 0.0415, ..., -0.0436, -0.0013, 0.0097]], + device='cuda:0'), grad: tensor([[ 6.0387e-06, 3.1665e-08, 1.0455e-04, ..., 1.9558e-07, + 1.3933e-06, 4.6566e-08], + [-2.6073e-03, -4.9174e-05, 8.5458e-06, ..., 2.5313e-06, + 2.6897e-06, 5.5879e-09], + [ 2.2202e-03, 4.1455e-05, 1.2144e-05, ..., 2.7753e-07, + 4.5113e-06, -4.3958e-07], + ..., + [ 2.5654e-04, 5.5470e-06, 1.0699e-05, ..., -1.8671e-05, + 2.0545e-06, 2.1048e-07], + [ 1.6987e-05, 1.9744e-07, 1.4409e-05, ..., 8.3819e-08, + -6.2399e-06, 4.0978e-08], + [ 5.2214e-05, 4.0978e-08, 1.0767e-03, ..., 1.3493e-05, + -2.0359e-06, 7.4506e-09]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0050, 0.0271, 0.0057, 0.0025, 0.0058, -0.0030, 0.0199, -0.0212, + 0.0201, 0.0029], device='cuda:0'), grad: tensor([ 1.8048e-04, -4.6577e-03, 4.0016e-03, 8.1718e-05, 3.7625e-06, + -1.9464e-03, 2.7511e-06, 4.8470e-04, 3.1322e-05, 1.8177e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 217.83, cls_loss 0.0099 cls_loss_mapping 0.0152 cls_loss_causal 0.6255 re_mapping 0.0126 re_causal 0.0362 /// teacc 98.67 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0281, -0.0192, -0.0255, ..., -0.0741, 0.0130, 0.0078], + [-0.0216, 0.0191, -0.0827, ..., 0.0298, 0.0315, -0.0558], + [ 0.0450, -0.0026, -0.0463, ..., -0.0315, -0.0137, -0.0075], + ..., + [ 0.0308, -0.0425, -0.0002, ..., 0.0239, -0.0010, 0.0326], + [-0.0145, -0.0312, -0.0355, ..., -0.0056, -0.0123, -0.0165], + [-0.0560, -0.0947, 0.0413, ..., -0.0436, -0.0015, 0.0097]], + device='cuda:0'), grad: tensor([[ 4.6119e-06, 0.0000e+00, 6.6981e-06, ..., 6.7055e-08, + 3.1859e-05, 2.0489e-07], + [ 5.7220e-06, 0.0000e+00, 2.3358e-06, ..., 1.9558e-07, + 3.3903e-04, 2.4214e-08], + [ 1.8284e-05, 0.0000e+00, 1.1176e-08, ..., 2.5518e-07, + 2.1055e-05, -6.8732e-07], + ..., + [-6.1058e-06, 0.0000e+00, -2.6077e-07, ..., 8.4005e-07, + 5.3197e-05, 9.8720e-08], + [ 1.2532e-05, 0.0000e+00, 1.9684e-05, ..., 1.2852e-07, + -3.5381e-04, 1.0990e-07], + [ 1.6481e-05, 0.0000e+00, 4.2245e-06, ..., 3.0231e-06, + -7.0453e-05, 6.5193e-08]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0054, 0.0278, 0.0057, 0.0021, 0.0069, -0.0031, 0.0206, -0.0215, + 0.0197, 0.0025], device='cuda:0'), grad: tensor([ 1.7381e-04, 8.4591e-04, 6.5923e-05, -6.0111e-05, 8.6784e-05, + -4.0359e-03, 3.4924e-03, 1.2183e-04, -4.9400e-04, -1.9872e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 217.65, cls_loss 0.0095 cls_loss_mapping 0.0135 cls_loss_causal 0.6296 re_mapping 0.0118 re_causal 0.0349 /// teacc 98.65 lr 0.00010000 +Epoch 76, weight, value: tensor([[-2.8399e-02, -1.9514e-02, -2.5600e-02, ..., -7.4149e-02, + 1.2587e-02, 7.4560e-03], + [-2.1964e-02, 2.1534e-02, -8.3086e-02, ..., 2.9884e-02, + 3.1886e-02, -5.7684e-02], + [ 4.5360e-02, -4.0404e-03, -4.5648e-02, ..., -3.1736e-02, + -1.4179e-02, -6.7006e-03], + ..., + [ 3.1116e-02, -4.3402e-02, 9.1840e-05, ..., 2.4230e-02, + -9.5036e-04, 3.2503e-02], + [-1.5044e-02, -3.1580e-02, -3.5740e-02, ..., -5.6757e-03, + -1.3008e-02, -1.7315e-02], + [-5.6663e-02, -9.5329e-02, 4.0913e-02, ..., -4.3756e-02, + -9.6210e-04, 9.7934e-03]], device='cuda:0'), grad: tensor([[ 7.1004e-06, 0.0000e+00, 8.8848e-07, ..., 3.9116e-07, + 1.0788e-05, 1.7881e-07], + [ 3.9116e-06, 0.0000e+00, 1.3895e-06, ..., -3.4243e-05, + -2.4652e-04, 9.8720e-08], + [ 2.9707e-04, 0.0000e+00, 1.1399e-06, ..., 5.3085e-07, + 4.3124e-05, -2.5798e-06], + ..., + [ 2.3693e-05, 0.0000e+00, 2.4587e-06, ..., 3.8706e-06, + 3.9846e-05, 4.7684e-07], + [-1.6236e-04, 0.0000e+00, 9.8720e-07, ..., 2.9802e-07, + -2.1088e-04, 7.6741e-07], + [ 8.5160e-06, 0.0000e+00, 4.1574e-06, ..., 4.7944e-06, + -3.5197e-05, 7.8231e-08]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0049, 0.0282, 0.0057, 0.0022, 0.0063, -0.0026, 0.0197, -0.0212, + 0.0190, 0.0024], device='cuda:0'), grad: tensor([-1.1748e-04, -4.1556e-04, 4.7827e-04, -2.1830e-06, 9.2149e-05, + -3.1948e-05, 2.3785e-03, 1.1194e-04, -2.4166e-03, -7.4744e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 217.53, cls_loss 0.0075 cls_loss_mapping 0.0109 cls_loss_causal 0.6068 re_mapping 0.0120 re_causal 0.0356 /// teacc 98.64 lr 0.00010000 +Epoch 77, weight, value: tensor([[-2.8559e-02, -1.9567e-02, -2.5859e-02, ..., -7.4169e-02, + 1.2412e-02, 7.1361e-03], + [-2.1615e-02, 2.1567e-02, -8.3421e-02, ..., 3.0023e-02, + 3.2002e-02, -5.9072e-02], + [ 4.5470e-02, -3.6383e-03, -4.6237e-02, ..., -3.1782e-02, + -1.4606e-02, -6.3415e-03], + ..., + [ 3.1047e-02, -4.3923e-02, 1.2460e-05, ..., 2.4196e-02, + -1.2116e-03, 3.2991e-02], + [-1.5367e-02, -3.1813e-02, -3.5822e-02, ..., -5.6927e-03, + -1.2372e-02, -1.8245e-02], + [-5.7009e-02, -9.5381e-02, 4.0660e-02, ..., -4.3789e-02, + -5.2460e-04, 1.0220e-02]], device='cuda:0'), grad: tensor([[ 2.3916e-06, 0.0000e+00, -1.6969e-06, ..., 0.0000e+00, + 1.8533e-06, 1.4156e-07], + [ 1.2130e-05, 0.0000e+00, 4.8429e-07, ..., 0.0000e+00, + -6.6012e-06, 2.4214e-08], + [ 6.9700e-06, 0.0000e+00, 1.1772e-06, ..., 0.0000e+00, + 2.9169e-06, 3.1665e-08], + ..., + [-1.7917e-04, 0.0000e+00, 1.1697e-06, ..., 0.0000e+00, + 5.1945e-05, 5.5879e-09], + [ 2.0468e-04, 0.0000e+00, 3.3528e-05, ..., 0.0000e+00, + 4.0650e-05, 2.4773e-07], + [ 2.4766e-05, 0.0000e+00, -1.6280e-06, ..., 0.0000e+00, + -6.6042e-05, 2.0489e-08]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0048, 0.0285, 0.0054, 0.0022, 0.0076, -0.0024, 0.0174, -0.0213, + 0.0196, 0.0022], device='cuda:0'), grad: tensor([-9.1791e-06, 1.7196e-05, 4.1872e-05, -8.0395e-04, 9.6023e-05, + 3.4380e-04, -1.3918e-05, -2.5916e-04, 5.5361e-04, 3.5286e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 217.60, cls_loss 0.0090 cls_loss_mapping 0.0118 cls_loss_causal 0.6183 re_mapping 0.0120 re_causal 0.0344 /// teacc 98.77 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0289, -0.0196, -0.0262, ..., -0.0742, 0.0118, 0.0068], + [-0.0222, 0.0217, -0.0851, ..., 0.0300, 0.0325, -0.0612], + [ 0.0466, -0.0037, -0.0468, ..., -0.0318, -0.0142, -0.0054], + ..., + [ 0.0313, -0.0438, 0.0001, ..., 0.0243, -0.0014, 0.0335], + [-0.0163, -0.0320, -0.0355, ..., -0.0057, -0.0129, -0.0190], + [-0.0577, -0.0956, 0.0403, ..., -0.0438, -0.0003, 0.0103]], + device='cuda:0'), grad: tensor([[ 2.5295e-06, 7.6368e-08, 1.7136e-06, ..., 7.4506e-09, + 4.4405e-06, 2.4773e-07], + [-3.1924e-04, -1.0008e-04, 1.5330e-06, ..., -4.2096e-07, + -6.0129e-04, 4.8429e-08], + [ 3.0065e-04, 9.2387e-05, 1.8813e-06, ..., 4.0978e-08, + 5.3644e-04, 2.7195e-07], + ..., + [ 1.4678e-05, 4.0904e-06, 2.4028e-06, ..., 1.7695e-07, + 3.1292e-05, 1.3039e-08], + [ 3.8333e-06, 2.7008e-07, -2.9337e-06, ..., 3.7253e-08, + 2.9244e-06, 1.6261e-06], + [ 1.4510e-06, 7.8231e-08, 1.1213e-05, ..., 2.2352e-08, + 6.5528e-06, 3.5390e-08]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0048, 0.0283, 0.0063, 0.0019, 0.0073, -0.0020, 0.0177, -0.0210, + 0.0189, 0.0017], device='cuda:0'), grad: tensor([ 4.7758e-06, -2.2602e-03, 2.0752e-03, -1.2435e-05, 7.8321e-05, + 7.6950e-05, -4.9859e-05, 1.1832e-04, -9.5487e-05, 6.1035e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 217.53, cls_loss 0.0098 cls_loss_mapping 0.0129 cls_loss_causal 0.6230 re_mapping 0.0113 re_causal 0.0339 /// teacc 98.74 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0286, -0.0196, -0.0263, ..., -0.0747, 0.0114, 0.0067], + [-0.0218, 0.0221, -0.0858, ..., 0.0331, 0.0339, -0.0616], + [ 0.0466, -0.0041, -0.0468, ..., -0.0320, -0.0141, -0.0048], + ..., + [ 0.0310, -0.0438, -0.0003, ..., 0.0228, -0.0034, 0.0334], + [-0.0166, -0.0321, -0.0360, ..., -0.0059, -0.0131, -0.0194], + [-0.0586, -0.0956, 0.0403, ..., -0.0447, 0.0010, 0.0103]], + device='cuda:0'), grad: tensor([[ 9.2328e-05, 0.0000e+00, 1.8060e-05, ..., 1.2480e-07, + 1.3292e-04, 3.7253e-09], + [ 3.3855e-05, 0.0000e+00, 1.5460e-06, ..., 1.7025e-06, + 8.7991e-06, 0.0000e+00], + [-4.5121e-05, 0.0000e+00, 6.0461e-06, ..., 8.3633e-07, + 6.4932e-06, 0.0000e+00], + ..., + [-7.1228e-06, 0.0000e+00, -4.9695e-06, ..., 1.5073e-05, + 4.0442e-05, 0.0000e+00], + [ 2.5570e-05, 0.0000e+00, -4.3809e-06, ..., 6.4634e-07, + -3.2812e-05, 1.4901e-08], + [-8.4221e-05, 0.0000e+00, -1.4782e-04, ..., 2.6580e-06, + -2.7251e-04, 0.0000e+00]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0045, 0.0297, 0.0063, 0.0016, 0.0076, -0.0024, 0.0181, -0.0226, + 0.0188, 0.0018], device='cuda:0'), grad: tensor([ 4.7183e-04, 6.1333e-05, -6.2108e-05, -1.4615e-04, 6.9666e-04, + 6.6876e-05, 9.8869e-06, 3.1382e-05, -5.8174e-05, -1.0710e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 78---------------------------------------------------- +epoch 78, time 218.26, cls_loss 0.0097 cls_loss_mapping 0.0133 cls_loss_causal 0.6282 re_mapping 0.0121 re_causal 0.0346 /// teacc 98.79 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0289, -0.0198, -0.0275, ..., -0.0754, 0.0105, 0.0049], + [-0.0217, 0.0221, -0.0861, ..., 0.0336, 0.0338, -0.0620], + [ 0.0463, -0.0040, -0.0467, ..., -0.0322, -0.0146, -0.0046], + ..., + [ 0.0314, -0.0439, -0.0003, ..., 0.0226, -0.0035, 0.0333], + [-0.0166, -0.0324, -0.0359, ..., -0.0063, -0.0126, -0.0197], + [-0.0595, -0.0957, 0.0409, ..., -0.0447, 0.0018, 0.0122]], + device='cuda:0'), grad: tensor([[ 7.0930e-05, 1.9558e-07, -5.0217e-06, ..., 1.3813e-05, + 3.9786e-06, 3.7253e-09], + [ 1.4639e-04, -1.7136e-06, 3.3509e-06, ..., 1.9018e-06, + -8.3148e-05, 2.4214e-08], + [-7.3957e-04, -7.9274e-06, 1.2778e-06, ..., 2.4438e-06, + 4.7535e-05, 2.7940e-08], + ..., + [ 7.0858e-04, 6.4857e-06, 1.8418e-05, ..., 3.2043e-04, + 6.5744e-05, -1.3970e-07], + [ 7.5758e-05, 1.1027e-06, 2.8443e-06, ..., 1.7229e-06, + -7.4692e-07, 5.5879e-09], + [-5.4598e-04, 1.9372e-07, -3.1471e-05, ..., -3.9339e-04, + -8.2970e-05, 1.4901e-08]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0055, 0.0298, 0.0057, 0.0021, 0.0071, -0.0028, 0.0181, -0.0223, + 0.0192, 0.0028], device='cuda:0'), grad: tensor([ 9.1434e-05, -9.4593e-05, -1.1358e-03, 4.4203e-04, 1.4544e-04, + 1.0180e-04, 4.5955e-05, 1.7052e-03, 1.6081e-04, -1.4639e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 217.32, cls_loss 0.0091 cls_loss_mapping 0.0132 cls_loss_causal 0.6258 re_mapping 0.0123 re_causal 0.0346 /// teacc 98.70 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0290, -0.0202, -0.0279, ..., -0.0756, 0.0100, 0.0050], + [-0.0223, 0.0217, -0.0868, ..., 0.0338, 0.0339, -0.0622], + [ 0.0470, -0.0040, -0.0466, ..., -0.0323, -0.0145, -0.0045], + ..., + [ 0.0315, -0.0419, -0.0005, ..., 0.0224, -0.0039, 0.0332], + [-0.0165, -0.0328, -0.0365, ..., -0.0064, -0.0124, -0.0198], + [-0.0605, -0.0962, 0.0407, ..., -0.0441, 0.0022, 0.0122]], + device='cuda:0'), grad: tensor([[ 2.7776e-05, 8.9407e-08, -2.6405e-05, ..., 0.0000e+00, + 6.1616e-06, 0.0000e+00], + [ 1.4842e-04, 3.5390e-08, 6.4820e-06, ..., 0.0000e+00, + 6.1333e-05, 0.0000e+00], + [-1.3828e-04, 1.0803e-07, 4.7684e-06, ..., 0.0000e+00, + 1.9372e-05, 0.0000e+00], + ..., + [-1.0990e-07, 1.4342e-07, 5.3495e-06, ..., 0.0000e+00, + 4.9882e-06, 0.0000e+00], + [-5.6219e-04, 1.0617e-07, 1.2136e-04, ..., 0.0000e+00, + -2.1923e-04, 0.0000e+00], + [-4.3064e-05, 3.5949e-07, 6.5923e-05, ..., 0.0000e+00, + -1.8394e-04, 0.0000e+00]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0056, 0.0293, 0.0064, 0.0021, 0.0071, -0.0025, 0.0179, -0.0224, + 0.0191, 0.0028], device='cuda:0'), grad: tensor([-1.6165e-04, 3.9744e-04, -1.3697e-04, 1.2474e-03, 2.3350e-05, + -1.8339e-03, 8.6641e-04, 6.3539e-05, -4.9400e-04, 2.9072e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 80---------------------------------------------------- +epoch 80, time 218.07, cls_loss 0.0083 cls_loss_mapping 0.0122 cls_loss_causal 0.5981 re_mapping 0.0119 re_causal 0.0344 /// teacc 98.88 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0295, -0.0203, -0.0289, ..., -0.0757, 0.0097, 0.0042], + [-0.0221, 0.0221, -0.0875, ..., 0.0342, 0.0343, -0.0623], + [ 0.0472, -0.0043, -0.0468, ..., -0.0335, -0.0148, -0.0045], + ..., + [ 0.0320, -0.0418, -0.0002, ..., 0.0227, -0.0041, 0.0332], + [-0.0169, -0.0341, -0.0370, ..., -0.0065, -0.0128, -0.0199], + [-0.0610, -0.0966, 0.0413, ..., -0.0442, 0.0028, 0.0130]], + device='cuda:0'), grad: tensor([[ 5.7220e-06, 1.7695e-07, -3.6713e-06, ..., 2.3805e-06, + 1.2200e-06, 3.7253e-09], + [ 4.9382e-05, 9.1456e-07, 2.1867e-06, ..., 1.8761e-05, + -5.6028e-06, 3.7253e-09], + [ 1.0651e-04, 4.0755e-06, 1.1295e-05, ..., 5.3078e-05, + 1.7025e-06, -6.5193e-08], + ..., + [ 1.6308e-04, 2.2724e-05, 6.4746e-06, ..., 3.0184e-04, + 2.1812e-06, 3.7253e-09], + [ 3.4869e-05, 1.2256e-06, 7.9334e-05, ..., 1.4484e-05, + 3.1199e-06, 2.0489e-08], + [ 1.7846e-04, 1.9930e-07, -4.9400e-03, ..., 3.4031e-06, + -1.0654e-06, 9.3132e-09]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0060, 0.0293, 0.0065, 0.0020, 0.0067, -0.0030, 0.0182, -0.0220, + 0.0185, 0.0036], device='cuda:0'), grad: tensor([-1.7142e-04, 7.7486e-05, 1.7238e-04, -8.3065e-04, 6.1455e-03, + -1.7178e-04, 4.5568e-05, -2.8342e-05, 2.2936e-04, -5.4665e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 217.48, cls_loss 0.0085 cls_loss_mapping 0.0113 cls_loss_causal 0.6304 re_mapping 0.0117 re_causal 0.0343 /// teacc 98.83 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0297, -0.0208, -0.0294, ..., -0.0767, 0.0092, 0.0033], + [-0.0226, 0.0220, -0.0880, ..., 0.0343, 0.0345, -0.0626], + [ 0.0475, -0.0028, -0.0478, ..., -0.0341, -0.0153, -0.0042], + ..., + [ 0.0323, -0.0433, -0.0006, ..., 0.0221, -0.0041, 0.0332], + [-0.0169, -0.0308, -0.0373, ..., -0.0069, -0.0139, -0.0200], + [-0.0618, -0.0982, 0.0415, ..., -0.0443, 0.0032, 0.0138]], + device='cuda:0'), grad: tensor([[ 1.3448e-06, 3.7253e-08, -6.2063e-06, ..., 3.7253e-09, + -1.6429e-06, 7.8231e-08], + [ 4.7348e-06, 5.5879e-09, 1.9372e-07, ..., 1.8626e-09, + 7.8604e-07, 3.1665e-08], + [-5.2378e-06, -8.8662e-07, 2.8312e-07, ..., 0.0000e+00, + 4.0755e-06, -9.8720e-07], + ..., + [-9.8944e-06, 8.0094e-08, 4.0792e-07, ..., 3.7253e-09, + 8.3074e-07, 1.9744e-07], + [ 1.0118e-05, 4.9919e-07, 2.5332e-07, ..., 5.5879e-09, + -7.5065e-06, 5.2713e-07], + [ 1.1526e-05, 1.1176e-08, 3.8259e-06, ..., 1.1176e-08, + 4.9993e-06, 1.1176e-08]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0063, 0.0290, 0.0062, 0.0021, 0.0065, -0.0021, 0.0192, -0.0217, + 0.0179, 0.0033], device='cuda:0'), grad: tensor([-9.0420e-05, 1.2666e-05, 2.2035e-06, -5.9414e-04, 3.4347e-06, + 5.9795e-04, 2.0131e-05, -1.6093e-05, -5.9605e-06, 6.9916e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 217.31, cls_loss 0.0095 cls_loss_mapping 0.0135 cls_loss_causal 0.6320 re_mapping 0.0115 re_causal 0.0318 /// teacc 98.85 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0307, -0.0219, -0.0295, ..., -0.0770, 0.0084, 0.0033], + [-0.0229, 0.0219, -0.0883, ..., 0.0352, 0.0349, -0.0627], + [ 0.0477, -0.0026, -0.0476, ..., -0.0344, -0.0156, -0.0041], + ..., + [ 0.0325, -0.0433, -0.0002, ..., 0.0227, -0.0041, 0.0332], + [-0.0162, -0.0286, -0.0376, ..., -0.0074, -0.0139, -0.0201], + [-0.0632, -0.0995, 0.0413, ..., -0.0446, 0.0031, 0.0140]], + device='cuda:0'), grad: tensor([[ 7.9051e-06, 0.0000e+00, 5.6066e-07, ..., 1.1027e-06, + 1.5385e-06, 1.8626e-09], + [ 3.9554e-04, 0.0000e+00, 7.1339e-07, ..., 9.8348e-05, + -3.7253e-07, 1.8626e-09], + [-3.3236e-04, 0.0000e+00, 2.4214e-06, ..., -1.3435e-04, + 7.2829e-07, -3.3528e-08], + ..., + [-4.0102e-04, 0.0000e+00, -4.8429e-06, ..., 8.5086e-06, + 1.6112e-06, 3.7253e-09], + [ 1.6546e-04, 0.0000e+00, 5.8860e-06, ..., 2.3358e-06, + 3.5595e-06, 1.8626e-09], + [ 1.1906e-05, 0.0000e+00, 1.7118e-06, ..., 6.2957e-07, + -3.7514e-06, 0.0000e+00]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0060, 0.0289, 0.0062, 0.0016, 0.0061, -0.0018, 0.0198, -0.0207, + 0.0182, 0.0022], device='cuda:0'), grad: tensor([ 6.0052e-06, 7.1144e-04, -5.6505e-04, 2.3878e-04, 3.9935e-05, + -3.9458e-05, 2.0295e-05, -7.6151e-04, 3.2234e-04, 2.7657e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 217.22, cls_loss 0.0081 cls_loss_mapping 0.0107 cls_loss_causal 0.6117 re_mapping 0.0115 re_causal 0.0324 /// teacc 98.82 lr 0.00010000 +Epoch 85, weight, value: tensor([[-3.1685e-02, -2.2533e-02, -2.9514e-02, ..., -7.7276e-02, + 7.8431e-03, 8.4042e-04], + [-2.2637e-02, 2.1747e-02, -8.8562e-02, ..., 3.5310e-02, + 3.5274e-02, -6.3473e-02], + [ 4.9551e-02, -4.4498e-04, -4.7656e-02, ..., -3.4386e-02, + -1.5733e-02, -3.9914e-03], + ..., + [ 3.0931e-02, -4.5407e-02, -4.3666e-05, ..., 2.3128e-02, + -4.6640e-03, 3.3222e-02], + [-1.6874e-02, -2.8735e-02, -3.7578e-02, ..., -7.5971e-03, + -1.3717e-02, -2.0330e-02], + [-6.3224e-02, -1.0058e-01, 4.1021e-02, ..., -4.4998e-02, + 3.1767e-03, 1.6374e-02]], device='cuda:0'), grad: tensor([[-2.7150e-05, 3.7253e-09, 1.3225e-07, ..., 1.3039e-08, + 6.4224e-06, 3.7253e-09], + [ 5.2787e-06, 3.7253e-09, 1.2293e-07, ..., -2.7940e-08, + 5.6601e-04, 1.8626e-09], + [-2.3559e-05, -7.0781e-08, 2.4214e-08, ..., 3.5390e-08, + 1.1571e-05, -3.9116e-08], + ..., + [ 7.3686e-06, 1.8626e-08, 2.0117e-07, ..., 3.3528e-08, + 3.8669e-06, 7.4506e-09], + [ 2.4468e-05, 2.0489e-08, 3.0156e-06, ..., 4.2841e-08, + -7.7391e-04, 1.3039e-08], + [ 4.1910e-06, 0.0000e+00, 1.2629e-06, ..., 3.5390e-08, + 6.4783e-06, 0.0000e+00]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0068, 0.0291, 0.0075, 0.0017, 0.0062, -0.0014, 0.0201, -0.0219, + 0.0181, 0.0021], device='cuda:0'), grad: tensor([-7.9930e-05, 1.6823e-03, 1.3895e-05, 4.8667e-05, 2.1145e-05, + 8.5354e-05, 3.5524e-04, 2.8700e-05, -2.1915e-03, 3.5554e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 217.28, cls_loss 0.0073 cls_loss_mapping 0.0105 cls_loss_causal 0.6257 re_mapping 0.0116 re_causal 0.0334 /// teacc 98.88 lr 0.00010000 +Epoch 86, weight, value: tensor([[-3.1825e-02, -2.3469e-02, -2.9618e-02, ..., -7.7369e-02, + 7.4945e-03, 5.2248e-04], + [-2.2788e-02, 2.1964e-02, -8.8911e-02, ..., 3.5498e-02, + 3.5503e-02, -6.3982e-02], + [ 4.9456e-02, 4.7805e-05, -4.8111e-02, ..., -3.4768e-02, + -1.6050e-02, -3.4699e-03], + ..., + [ 3.1461e-02, -4.5332e-02, -9.5766e-05, ..., 2.3146e-02, + -4.7482e-03, 3.3152e-02], + [-1.6918e-02, -2.9669e-02, -3.7855e-02, ..., -7.7859e-03, + -1.3847e-02, -2.0583e-02], + [-6.3803e-02, -1.0353e-01, 4.1037e-02, ..., -4.5113e-02, + 3.5997e-03, 1.6679e-02]], device='cuda:0'), grad: tensor([[ 1.1260e-06, 1.7695e-08, 1.9503e-04, ..., 6.5193e-09, + 3.2596e-07, 3.6322e-08], + [ 1.2983e-06, 6.2399e-08, 6.6198e-06, ..., -2.1420e-08, + -4.6790e-06, 1.2107e-08], + [-1.0736e-05, -7.5437e-07, 4.1395e-05, ..., 3.2596e-08, + 1.0859e-06, -6.5099e-07], + ..., + [-2.7418e-06, 2.0489e-08, 1.0639e-05, ..., 4.2841e-08, + 2.0191e-06, 2.0489e-08], + [ 5.7966e-06, 3.8277e-07, 8.2552e-05, ..., 5.3085e-08, + 2.1383e-06, 3.4086e-07], + [ 1.7062e-06, 5.5879e-09, -1.0139e-04, ..., 1.1269e-07, + -1.9521e-06, 4.0978e-08]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0065, 0.0290, 0.0072, 0.0018, 0.0061, -0.0014, 0.0200, -0.0214, + 0.0178, 0.0020], device='cuda:0'), grad: tensor([ 3.5858e-04, 4.9174e-06, 6.3777e-05, 4.7207e-05, -8.9359e-04, + 2.5129e-04, 1.4913e-04, 2.2173e-05, 1.9062e-04, -1.9395e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 217.40, cls_loss 0.0081 cls_loss_mapping 0.0098 cls_loss_causal 0.5964 re_mapping 0.0113 re_causal 0.0326 /// teacc 98.49 lr 0.00010000 +Epoch 87, weight, value: tensor([[-3.1343e-02, -2.4470e-02, -2.9745e-02, ..., -7.7509e-02, + 7.6419e-03, 5.0189e-04], + [-2.2495e-02, 2.0924e-02, -8.9341e-02, ..., 3.6480e-02, + 3.6347e-02, -6.4996e-02], + [ 4.9868e-02, 1.4075e-03, -4.8820e-02, ..., -3.4958e-02, + -1.5985e-02, -3.3829e-03], + ..., + [ 3.1334e-02, -4.5443e-02, 5.2919e-05, ..., 2.3273e-02, + -5.9362e-03, 3.3070e-02], + [-1.6709e-02, -3.0405e-02, -3.7853e-02, ..., -7.9055e-03, + -1.3858e-02, -2.2335e-02], + [-6.4422e-02, -1.0611e-01, 4.0836e-02, ..., -4.5450e-02, + 3.9890e-03, 1.6682e-02]], device='cuda:0'), grad: tensor([[ 4.0978e-07, 9.3132e-10, -1.0170e-05, ..., 0.0000e+00, + 5.4296e-07, 7.4506e-09], + [ 1.5544e-06, 9.3132e-10, 2.5164e-06, ..., 0.0000e+00, + -3.7178e-06, 6.5193e-09], + [-1.3791e-05, -1.3970e-08, 5.5507e-07, ..., 0.0000e+00, + 8.4005e-07, -9.8720e-08], + ..., + [-8.5030e-07, 9.3132e-09, 1.6736e-06, ..., 0.0000e+00, + 5.3823e-05, 4.9360e-08], + [ 7.9200e-06, 9.3132e-10, 6.5506e-05, ..., 0.0000e+00, + 9.7379e-06, 1.0245e-08], + [ 1.4817e-06, 0.0000e+00, 1.8880e-05, ..., 0.0000e+00, + -1.5497e-04, 1.8626e-09]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0061, 0.0293, 0.0075, 0.0013, 0.0064, -0.0016, 0.0196, -0.0218, + 0.0183, 0.0017], device='cuda:0'), grad: tensor([-1.3685e-04, 2.7679e-06, -4.1686e-06, 8.9228e-05, 3.0661e-04, + -2.4045e-04, 2.9087e-05, 2.0981e-04, 2.5105e-04, -5.0783e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 217.68, cls_loss 0.0067 cls_loss_mapping 0.0099 cls_loss_causal 0.5950 re_mapping 0.0114 re_causal 0.0325 /// teacc 98.70 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0312, -0.0246, -0.0295, ..., -0.0782, 0.0075, 0.0002], + [-0.0228, 0.0209, -0.0900, ..., 0.0364, 0.0365, -0.0667], + [ 0.0502, 0.0013, -0.0498, ..., -0.0350, -0.0165, -0.0029], + ..., + [ 0.0315, -0.0452, 0.0002, ..., 0.0235, -0.0060, 0.0331], + [-0.0170, -0.0305, -0.0380, ..., -0.0079, -0.0142, -0.0247], + [-0.0646, -0.1065, 0.0410, ..., -0.0454, 0.0050, 0.0170]], + device='cuda:0'), grad: tensor([[ 4.7907e-06, 0.0000e+00, -5.7649e-07, ..., 0.0000e+00, + 1.0449e-06, 6.6496e-07], + [ 1.1303e-05, 0.0000e+00, 2.1413e-05, ..., 4.6566e-09, + -1.8990e-06, 2.6338e-06], + [-3.0100e-05, 0.0000e+00, 4.1813e-05, ..., 1.8626e-09, + -6.7204e-06, 4.8764e-06], + ..., + [ 1.7524e-05, 0.0000e+00, 3.8296e-06, ..., 3.2596e-08, + 4.2953e-06, 1.5181e-07], + [ 8.2701e-06, 0.0000e+00, -2.0396e-06, ..., 9.3132e-10, + -4.3735e-06, 3.8464e-07], + [ 2.4214e-06, 0.0000e+00, 6.9737e-05, ..., 8.4843e-07, + -4.8727e-06, 1.0524e-07]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0055, 0.0290, 0.0079, 0.0007, 0.0065, -0.0022, 0.0192, -0.0216, + 0.0183, 0.0020], device='cuda:0'), grad: tensor([ 1.0412e-06, 7.9870e-05, 1.3269e-05, 2.9147e-05, 1.2331e-03, + 7.5459e-05, -1.6127e-03, 5.8949e-05, 8.2776e-06, 1.1331e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 217.48, cls_loss 0.0085 cls_loss_mapping 0.0108 cls_loss_causal 0.6280 re_mapping 0.0110 re_causal 0.0322 /// teacc 98.77 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0314, -0.0251, -0.0293, ..., -0.0785, 0.0071, 0.0002], + [-0.0230, 0.0212, -0.0907, ..., 0.0371, 0.0365, -0.0686], + [ 0.0491, 0.0004, -0.0504, ..., -0.0356, -0.0174, -0.0023], + ..., + [ 0.0328, -0.0442, 0.0004, ..., 0.0232, -0.0067, 0.0329], + [-0.0175, -0.0311, -0.0381, ..., -0.0081, -0.0145, -0.0269], + [-0.0650, -0.1072, 0.0408, ..., -0.0456, 0.0066, 0.0170]], + device='cuda:0'), grad: tensor([[ 1.4510e-06, 6.4261e-08, 5.1036e-07, ..., 0.0000e+00, + -3.2574e-05, 4.5635e-08], + [ 3.4392e-05, 2.6412e-06, 1.1539e-06, ..., 0.0000e+00, + -6.3963e-06, 2.1420e-08], + [-2.8417e-05, -4.3958e-06, 6.0070e-07, ..., 0.0000e+00, + 5.9754e-06, -2.8312e-07], + ..., + [ 2.1935e-05, 1.0682e-06, 4.6752e-07, ..., 0.0000e+00, + 4.7982e-06, 2.9802e-08], + [ 1.4707e-05, 6.3330e-08, -4.3139e-06, ..., 0.0000e+00, + 1.6503e-06, 3.7253e-08], + [ 4.8801e-06, 4.8429e-08, 3.6657e-05, ..., 0.0000e+00, + 1.2271e-05, 9.3132e-09]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0057, 0.0289, 0.0068, 0.0005, 0.0063, -0.0010, 0.0189, -0.0212, + 0.0179, 0.0025], device='cuda:0'), grad: tensor([-6.7234e-04, 4.2140e-05, -4.7266e-05, -4.3631e-05, -1.5020e-04, + 5.5313e-05, 5.9032e-04, 4.2647e-05, 1.5251e-05, 1.6809e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 217.44, cls_loss 0.0074 cls_loss_mapping 0.0116 cls_loss_causal 0.5915 re_mapping 0.0114 re_causal 0.0316 /// teacc 98.73 lr 0.00010000 +Epoch 90, weight, value: tensor([[-3.1579e-02, -2.6152e-02, -2.9042e-02, ..., -7.8652e-02, + 6.8017e-03, 1.4344e-04], + [-2.3562e-02, 2.0799e-02, -9.0943e-02, ..., 3.7086e-02, + 3.6621e-02, -7.0911e-02], + [ 4.7967e-02, -5.5387e-04, -5.0581e-02, ..., -3.5739e-02, + -1.8544e-02, -1.5847e-03], + ..., + [ 3.4704e-02, -4.2498e-02, 4.0799e-05, ..., 2.3258e-02, + -6.1374e-03, 3.2744e-02], + [-1.8080e-02, -3.3078e-02, -3.8440e-02, ..., -8.2613e-03, + -1.4512e-02, -2.7978e-02], + [-6.5961e-02, -1.0942e-01, 4.0714e-02, ..., -4.5746e-02, + 7.1932e-03, 1.6938e-02]], device='cuda:0'), grad: tensor([[ 9.8124e-06, 1.8347e-07, 2.8126e-07, ..., 6.7707e-07, + 3.8259e-06, 7.4506e-09], + [ 4.4203e-04, 1.8571e-06, 2.7195e-07, ..., 7.1898e-06, + 8.2016e-05, 1.8626e-09], + [-1.6975e-04, -7.3671e-05, 8.1956e-08, ..., 3.5483e-07, + 2.9877e-05, -5.5879e-09], + ..., + [-2.3403e-03, 7.0214e-05, 1.0729e-06, ..., -3.7074e-05, + -1.5855e-04, 4.6566e-09], + [ 8.0824e-05, 6.4634e-07, 8.0653e-07, ..., 4.1258e-07, + 2.5138e-05, 2.0489e-08], + [ 1.4281e-04, 3.3528e-08, 9.7230e-07, ..., 2.3738e-05, + 2.8461e-05, 3.7253e-09]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0057, 0.0286, 0.0057, 0.0009, 0.0060, -0.0014, 0.0198, -0.0198, + 0.0174, 0.0025], device='cuda:0'), grad: tensor([-7.1563e-06, 1.0080e-03, -9.0539e-05, 2.0084e-03, 1.7658e-05, + 2.0057e-05, -1.9744e-06, -3.4561e-03, 1.9920e-04, 3.0518e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 217.67, cls_loss 0.0073 cls_loss_mapping 0.0093 cls_loss_causal 0.5837 re_mapping 0.0107 re_causal 0.0308 /// teacc 98.78 lr 0.00010000 +Epoch 91, weight, value: tensor([[-3.1746e-02, -2.7109e-02, -2.9686e-02, ..., -7.8891e-02, + 6.5501e-03, -1.6570e-03], + [-2.3629e-02, 2.0780e-02, -9.1575e-02, ..., 3.6932e-02, + 3.7027e-02, -7.1353e-02], + [ 4.8177e-02, 2.5558e-05, -5.0932e-02, ..., -3.5760e-02, + -1.8929e-02, -1.5281e-03], + ..., + [ 3.4894e-02, -4.3069e-02, 1.3605e-04, ..., 2.3350e-02, + -6.3779e-03, 3.2673e-02], + [-1.8058e-02, -3.6473e-02, -3.9193e-02, ..., -8.4988e-03, + -1.4524e-02, -2.8189e-02], + [-6.6684e-02, -1.1092e-01, 4.1127e-02, ..., -4.6283e-02, + 7.5552e-03, 1.8712e-02]], device='cuda:0'), grad: tensor([[ 1.2424e-06, 2.1420e-08, 2.6003e-06, ..., 0.0000e+00, + -2.1964e-05, 2.1607e-07], + [ 1.0744e-05, 6.5193e-09, 2.0824e-06, ..., 0.0000e+00, + 6.2399e-07, 2.3283e-08], + [-2.1346e-06, -2.1234e-07, 1.6196e-06, ..., 0.0000e+00, + 4.5374e-06, 3.0361e-07], + ..., + [-1.4091e-04, 6.5193e-08, 1.2927e-06, ..., 0.0000e+00, + -6.7055e-05, 4.4703e-08], + [ 5.2713e-06, 8.1025e-08, 1.5963e-06, ..., 0.0000e+00, + 5.4426e-06, 3.8184e-08], + [ 1.2803e-04, 2.7940e-09, 1.3679e-05, ..., 0.0000e+00, + 7.4387e-05, -7.7114e-07]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0062, 0.0285, 0.0056, 0.0005, 0.0059, -0.0013, 0.0196, -0.0199, + 0.0176, 0.0032], device='cuda:0'), grad: tensor([-2.4843e-04, 2.7999e-05, 1.0908e-05, 1.1347e-05, -4.0501e-05, + 1.3694e-05, 1.4402e-05, -4.5466e-04, 3.1620e-05, 6.3324e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 217.48, cls_loss 0.0071 cls_loss_mapping 0.0088 cls_loss_causal 0.6120 re_mapping 0.0111 re_causal 0.0317 /// teacc 98.75 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0322, -0.0274, -0.0297, ..., -0.0794, 0.0062, -0.0017], + [-0.0244, 0.0209, -0.0918, ..., 0.0371, 0.0372, -0.0720], + [ 0.0483, 0.0001, -0.0511, ..., -0.0358, -0.0193, -0.0015], + ..., + [ 0.0347, -0.0432, 0.0004, ..., 0.0243, -0.0064, 0.0330], + [-0.0170, -0.0365, -0.0396, ..., -0.0088, -0.0141, -0.0284], + [-0.0680, -0.1112, 0.0409, ..., -0.0468, 0.0070, 0.0187]], + device='cuda:0'), grad: tensor([[ 1.1742e-05, 0.0000e+00, 1.6717e-06, ..., 0.0000e+00, + 2.2665e-05, 4.1723e-07], + [ 3.9458e-05, 0.0000e+00, 2.5239e-06, ..., 0.0000e+00, + 9.7901e-06, 1.8626e-09], + [-3.8058e-05, 0.0000e+00, 9.5833e-07, ..., 0.0000e+00, + 3.7700e-06, 2.7940e-09], + ..., + [ 2.9221e-05, 0.0000e+00, 2.1160e-06, ..., 0.0000e+00, + 5.3123e-06, 9.3132e-10], + [ 1.7121e-05, 0.0000e+00, 6.3255e-06, ..., 0.0000e+00, + -2.2724e-06, 1.1176e-08], + [ 3.3647e-05, 0.0000e+00, 4.2655e-07, ..., 0.0000e+00, + -2.1141e-06, 1.1176e-08]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0058, 0.0281, 0.0056, 0.0013, 0.0061, -0.0017, 0.0193, -0.0199, + 0.0184, 0.0025], device='cuda:0'), grad: tensor([ 2.1122e-06, 1.1557e-04, 6.3404e-06, 1.5147e-05, -7.5400e-05, + 2.9594e-05, -2.5630e-04, 6.0707e-05, 4.5806e-05, 5.6356e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 217.71, cls_loss 0.0082 cls_loss_mapping 0.0113 cls_loss_causal 0.5898 re_mapping 0.0106 re_causal 0.0300 /// teacc 98.74 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0327, -0.0276, -0.0300, ..., -0.0796, 0.0058, -0.0017], + [-0.0256, 0.0210, -0.0923, ..., 0.0383, 0.0380, -0.0738], + [ 0.0484, 0.0001, -0.0511, ..., -0.0359, -0.0198, -0.0015], + ..., + [ 0.0346, -0.0431, 0.0007, ..., 0.0239, -0.0064, 0.0339], + [-0.0152, -0.0369, -0.0400, ..., -0.0089, -0.0150, -0.0288], + [-0.0696, -0.1117, 0.0409, ..., -0.0471, 0.0072, 0.0187]], + device='cuda:0'), grad: tensor([[ 1.6345e-06, 0.0000e+00, -2.1327e-07, ..., 0.0000e+00, + 9.4995e-07, 1.0245e-08], + [ 9.8720e-06, 0.0000e+00, 2.3022e-06, ..., 0.0000e+00, + 5.2676e-06, 8.3819e-08], + [ 1.1444e-05, 0.0000e+00, 3.8836e-07, ..., 0.0000e+00, + 2.6487e-06, 1.3970e-08], + ..., + [-7.4387e-05, 0.0000e+00, -2.9467e-06, ..., 0.0000e+00, + -1.2696e-05, 6.3330e-08], + [ 7.1935e-06, 0.0000e+00, 2.0880e-06, ..., 0.0000e+00, + 2.8312e-05, 6.5193e-09], + [ 4.0323e-05, 0.0000e+00, -8.0705e-05, ..., 0.0000e+00, + -2.6989e-04, 1.1269e-07]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0055, 0.0282, 0.0057, 0.0009, 0.0062, -0.0014, 0.0195, -0.0204, + 0.0192, 0.0019], device='cuda:0'), grad: tensor([-3.8457e-04, 3.9786e-05, 2.4736e-05, -1.7537e-06, 8.3256e-04, + 2.4509e-04, 1.2743e-04, -1.9467e-04, 9.7811e-05, -7.8630e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 217.78, cls_loss 0.0079 cls_loss_mapping 0.0109 cls_loss_causal 0.6241 re_mapping 0.0104 re_causal 0.0300 /// teacc 98.67 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0335, -0.0284, -0.0298, ..., -0.0800, 0.0075, -0.0018], + [-0.0254, 0.0211, -0.0932, ..., 0.0390, 0.0385, -0.0767], + [ 0.0486, 0.0006, -0.0513, ..., -0.0360, -0.0200, -0.0013], + ..., + [ 0.0347, -0.0436, 0.0005, ..., 0.0237, -0.0067, 0.0343], + [-0.0154, -0.0376, -0.0412, ..., -0.0090, -0.0151, -0.0310], + [-0.0708, -0.1126, 0.0407, ..., -0.0479, 0.0077, 0.0189]], + device='cuda:0'), grad: tensor([[ 3.9227e-06, 0.0000e+00, 1.1586e-05, ..., 1.3132e-07, + 4.2245e-06, 0.0000e+00], + [ 2.0355e-05, 0.0000e+00, 6.8992e-06, ..., 2.9057e-07, + 2.5891e-07, 0.0000e+00], + [-2.3365e-05, 0.0000e+00, -2.0117e-06, ..., 6.9849e-08, + 6.9384e-07, 0.0000e+00], + ..., + [-2.1830e-05, 0.0000e+00, 7.8380e-06, ..., 2.4494e-07, + 6.2808e-06, 0.0000e+00], + [ 5.6356e-05, 0.0000e+00, 1.0610e-05, ..., 5.4203e-07, + 1.2092e-05, 0.0000e+00], + [-8.9686e-07, 0.0000e+00, 3.0994e-05, ..., 8.9407e-08, + -2.6330e-05, 0.0000e+00]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0044, 0.0283, 0.0060, 0.0006, 0.0068, -0.0011, 0.0183, -0.0203, + 0.0190, 0.0012], device='cuda:0'), grad: tensor([ 6.1356e-06, 5.4240e-05, -3.8385e-05, 3.0786e-05, 2.3693e-05, + -1.9336e-04, 2.2739e-05, -1.4365e-05, 1.3149e-04, -2.3007e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 217.53, cls_loss 0.0064 cls_loss_mapping 0.0085 cls_loss_causal 0.5821 re_mapping 0.0112 re_causal 0.0301 /// teacc 98.77 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0337, -0.0289, -0.0299, ..., -0.0801, 0.0071, -0.0018], + [-0.0257, 0.0209, -0.0937, ..., 0.0391, 0.0383, -0.0771], + [ 0.0490, -0.0002, -0.0512, ..., -0.0360, -0.0202, -0.0012], + ..., + [ 0.0351, -0.0427, 0.0004, ..., 0.0236, -0.0069, 0.0343], + [-0.0163, -0.0364, -0.0413, ..., -0.0091, -0.0152, -0.0313], + [-0.0713, -0.1136, 0.0402, ..., -0.0481, 0.0087, 0.0189]], + device='cuda:0'), grad: tensor([[ 2.8685e-06, 0.0000e+00, -4.3027e-07, ..., 0.0000e+00, + 4.1053e-06, 0.0000e+00], + [ 5.6252e-06, 0.0000e+00, 1.0505e-06, ..., 0.0000e+00, + 3.9935e-06, 0.0000e+00], + [-2.7657e-04, 0.0000e+00, 1.4780e-06, ..., 0.0000e+00, + -5.3763e-05, 0.0000e+00], + ..., + [ 5.6103e-06, 0.0000e+00, 3.8296e-06, ..., 0.0000e+00, + 2.5593e-06, 0.0000e+00], + [ 4.9353e-05, 0.0000e+00, 5.8208e-07, ..., 0.0000e+00, + -5.4762e-06, 0.0000e+00], + [ 1.3048e-06, 0.0000e+00, 9.8441e-07, ..., 0.0000e+00, + -5.4687e-06, 0.0000e+00]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0041, 0.0277, 0.0060, 0.0004, 0.0066, -0.0012, 0.0192, -0.0197, + 0.0181, 0.0013], device='cuda:0'), grad: tensor([-1.1310e-05, 2.5257e-05, -3.2830e-04, 2.7680e-04, 8.7693e-06, + 1.5780e-05, -4.0382e-05, 5.8264e-06, 2.8074e-05, 1.9282e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 217.70, cls_loss 0.0068 cls_loss_mapping 0.0090 cls_loss_causal 0.5953 re_mapping 0.0108 re_causal 0.0296 /// teacc 98.76 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0332, -0.0295, -0.0301, ..., -0.0801, 0.0071, -0.0027], + [-0.0264, 0.0209, -0.0953, ..., 0.0391, 0.0382, -0.0777], + [ 0.0489, -0.0002, -0.0499, ..., -0.0361, -0.0207, -0.0007], + ..., + [ 0.0356, -0.0426, 0.0003, ..., 0.0236, -0.0074, 0.0342], + [-0.0162, -0.0369, -0.0418, ..., -0.0092, -0.0156, -0.0314], + [-0.0722, -0.1155, 0.0398, ..., -0.0483, 0.0097, 0.0198]], + device='cuda:0'), grad: tensor([[ 1.5683e-06, 0.0000e+00, 9.8534e-07, ..., 0.0000e+00, + 3.6806e-06, 2.4214e-08], + [-9.3639e-05, 0.0000e+00, 3.6042e-07, ..., 0.0000e+00, + -2.0456e-04, 1.8626e-09], + [-2.2817e-06, 0.0000e+00, 1.1735e-07, ..., 0.0000e+00, + 6.6198e-06, -9.6858e-08], + ..., + [-1.4678e-05, 0.0000e+00, 2.6356e-07, ..., 0.0000e+00, + 1.8373e-05, 2.0489e-08], + [ 7.4446e-05, 0.0000e+00, 1.0848e-05, ..., 0.0000e+00, + 1.9288e-04, 1.8626e-08], + [ 3.4552e-06, 0.0000e+00, 1.7444e-06, ..., 0.0000e+00, + -5.3734e-05, 6.5193e-09]], device='cuda:0') +Epoch 96, bias, value: tensor([-3.8285e-03, 2.7058e-02, 5.7196e-03, 4.5091e-05, 6.8383e-03, + -1.6129e-03, 1.9928e-02, -1.9388e-02, 1.8173e-02, 1.3574e-03], + device='cuda:0'), grad: tensor([ 1.0476e-05, -4.3154e-04, 7.2643e-06, 4.7475e-05, 3.8713e-05, + -1.9848e-04, 1.8823e-04, 2.1607e-05, 4.4274e-04, -1.2684e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 217.70, cls_loss 0.0067 cls_loss_mapping 0.0095 cls_loss_causal 0.5823 re_mapping 0.0103 re_causal 0.0293 /// teacc 98.85 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0331, -0.0296, -0.0303, ..., -0.0802, 0.0070, -0.0033], + [-0.0267, 0.0209, -0.0961, ..., 0.0392, 0.0389, -0.0779], + [ 0.0491, -0.0002, -0.0502, ..., -0.0361, -0.0209, -0.0006], + ..., + [ 0.0356, -0.0426, 0.0006, ..., 0.0236, -0.0087, 0.0342], + [-0.0166, -0.0374, -0.0426, ..., -0.0093, -0.0163, -0.0314], + [-0.0717, -0.1157, 0.0396, ..., -0.0484, 0.0107, 0.0204]], + device='cuda:0'), grad: tensor([[ 5.3905e-06, 9.3132e-09, -1.1951e-05, ..., 0.0000e+00, + 1.0353e-04, 0.0000e+00], + [ 3.4533e-06, 2.1420e-08, 5.6438e-07, ..., 0.0000e+00, + -4.2629e-04, 0.0000e+00], + [-3.0383e-05, -5.0571e-07, 2.6450e-07, ..., 0.0000e+00, + 3.0875e-05, 0.0000e+00], + ..., + [ 6.2659e-06, 3.0547e-07, 1.0934e-06, ..., 0.0000e+00, + 1.9431e-05, 0.0000e+00], + [ 1.3940e-05, 1.0151e-07, 1.5991e-06, ..., 0.0000e+00, + 1.8382e-04, 0.0000e+00], + [ 1.6302e-05, 9.3132e-10, 1.0524e-06, ..., 0.0000e+00, + 1.2964e-05, 0.0000e+00]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0044, 0.0272, 0.0058, -0.0002, 0.0070, -0.0013, 0.0195, -0.0194, + 0.0173, 0.0024], device='cuda:0'), grad: tensor([ 1.3947e-04, -8.3208e-04, 2.7224e-05, 4.8429e-07, 5.4121e-05, + 5.3167e-05, 1.0347e-04, 5.0157e-05, 3.6764e-04, 3.7253e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 218.00, cls_loss 0.0065 cls_loss_mapping 0.0105 cls_loss_causal 0.6096 re_mapping 0.0104 re_causal 0.0302 /// teacc 98.85 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0333, -0.0296, -0.0303, ..., -0.0803, 0.0063, -0.0033], + [-0.0270, 0.0209, -0.0973, ..., 0.0392, 0.0411, -0.0783], + [ 0.0490, -0.0002, -0.0507, ..., -0.0361, -0.0218, -0.0003], + ..., + [ 0.0359, -0.0426, 0.0009, ..., 0.0240, -0.0096, 0.0342], + [-0.0169, -0.0373, -0.0425, ..., -0.0093, -0.0166, -0.0316], + [-0.0730, -0.1157, 0.0383, ..., -0.0485, 0.0098, 0.0204]], + device='cuda:0'), grad: tensor([[ 9.4157e-07, 0.0000e+00, 2.0452e-06, ..., 0.0000e+00, + 1.8045e-05, 5.8934e-06], + [ 1.5616e-05, 0.0000e+00, 8.4005e-07, ..., 0.0000e+00, + -5.6811e-06, 3.9116e-08], + [-2.0695e-03, 0.0000e+00, 5.0385e-07, ..., 0.0000e+00, + 2.2054e-06, 4.9360e-08], + ..., + [ 3.4332e-05, 0.0000e+00, 3.7346e-07, ..., 0.0000e+00, + 4.5411e-06, 1.5832e-08], + [ 2.0199e-03, 0.0000e+00, 1.4147e-06, ..., 0.0000e+00, + 5.1633e-06, 2.9150e-07], + [ 4.5076e-06, 0.0000e+00, 7.2829e-06, ..., 0.0000e+00, + 5.1744e-06, 2.1141e-07]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0042, 0.0278, 0.0052, -0.0019, 0.0079, 0.0004, 0.0197, -0.0191, + 0.0174, 0.0009], device='cuda:0'), grad: tensor([ 7.0572e-05, 6.8724e-05, -1.0391e-02, -1.7971e-05, 8.3745e-06, + 2.9594e-05, -9.5189e-05, 1.7679e-04, 1.0132e-02, 2.8670e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 219.00, cls_loss 0.0060 cls_loss_mapping 0.0092 cls_loss_causal 0.6126 re_mapping 0.0106 re_causal 0.0308 /// teacc 98.75 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0339, -0.0298, -0.0312, ..., -0.0804, 0.0058, -0.0038], + [-0.0272, 0.0208, -0.0988, ..., 0.0392, 0.0420, -0.0788], + [ 0.0487, -0.0002, -0.0513, ..., -0.0361, -0.0223, 0.0007], + ..., + [ 0.0366, -0.0425, 0.0013, ..., 0.0241, -0.0114, 0.0342], + [-0.0174, -0.0375, -0.0420, ..., -0.0093, -0.0159, -0.0318], + [-0.0740, -0.1160, 0.0385, ..., -0.0486, 0.0104, 0.0209]], + device='cuda:0'), grad: tensor([[ 1.1455e-06, 0.0000e+00, 3.3118e-06, ..., 0.0000e+00, + 2.7604e-06, 0.0000e+00], + [ 1.1334e-06, 0.0000e+00, 2.2903e-05, ..., 0.0000e+00, + 2.7925e-05, 0.0000e+00], + [-2.0444e-05, 0.0000e+00, 2.8647e-06, ..., 0.0000e+00, + 3.4068e-06, 0.0000e+00], + ..., + [ 1.0906e-06, 0.0000e+00, 1.6332e-05, ..., 0.0000e+00, + 1.2338e-05, 0.0000e+00], + [ 9.9763e-06, 0.0000e+00, 2.6286e-05, ..., 0.0000e+00, + 3.4183e-05, 0.0000e+00], + [-9.5218e-06, 0.0000e+00, -1.2856e-03, ..., 0.0000e+00, + -1.5278e-03, 0.0000e+00]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0046, 0.0282, 0.0053, -0.0019, 0.0082, 0.0003, 0.0200, -0.0192, + 0.0171, 0.0006], device='cuda:0'), grad: tensor([ 1.3858e-05, 1.1659e-04, -2.6584e-05, 2.7001e-05, 5.9814e-03, + 2.6464e-05, 5.9083e-06, 6.7651e-05, 1.5199e-04, -6.3629e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 220.62, cls_loss 0.0061 cls_loss_mapping 0.0087 cls_loss_causal 0.5846 re_mapping 0.0107 re_causal 0.0294 /// teacc 98.87 lr 0.00010000 +Epoch 100, weight, value: tensor([[-3.3953e-02, -3.0393e-02, -3.1401e-02, ..., -8.0508e-02, + 5.6541e-03, -3.9072e-03], + [-2.8621e-02, 2.0676e-02, -9.9945e-02, ..., 3.9273e-02, + 4.1239e-02, -7.9373e-02], + [ 4.9674e-02, 1.0520e-04, -5.1749e-02, ..., -3.6169e-02, + -2.1992e-02, 3.4256e-04], + ..., + [ 3.6689e-02, -4.3312e-02, 6.7844e-04, ..., 2.4578e-02, + -1.0763e-02, 3.4235e-02], + [-1.7645e-02, -3.4443e-02, -4.2634e-02, ..., -9.3818e-03, + -1.5728e-02, -3.1914e-02], + [-7.5739e-02, -1.1693e-01, 3.9002e-02, ..., -4.8760e-02, + 1.0871e-02, 2.0944e-02]], device='cuda:0'), grad: tensor([[ 4.7684e-06, 0.0000e+00, -7.3668e-07, ..., 0.0000e+00, + 3.2447e-06, 0.0000e+00], + [ 2.9672e-06, 0.0000e+00, 8.9407e-08, ..., 0.0000e+00, + -2.3264e-06, 9.3132e-10], + [ 3.5405e-05, 0.0000e+00, 2.9709e-07, ..., 0.0000e+00, + 2.1495e-06, 9.3132e-10], + ..., + [ 4.1038e-05, 0.0000e+00, 2.0862e-07, ..., 0.0000e+00, + 2.0117e-06, -3.7253e-09], + [ 3.2037e-05, 0.0000e+00, -1.1316e-06, ..., 0.0000e+00, + -1.0459e-06, 0.0000e+00], + [ 1.9386e-05, 0.0000e+00, 1.7621e-06, ..., 0.0000e+00, + -1.4435e-07, 0.0000e+00]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0045, 0.0270, 0.0061, -0.0019, 0.0081, 0.0003, 0.0199, -0.0191, + 0.0172, 0.0007], device='cuda:0'), grad: tensor([ 9.5665e-06, 1.8738e-06, 6.9201e-05, -2.7299e-04, 1.2249e-05, + 2.7418e-05, -2.0176e-05, 7.9393e-05, 5.6714e-05, 3.6359e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 219.76, cls_loss 0.0061 cls_loss_mapping 0.0098 cls_loss_causal 0.5776 re_mapping 0.0097 re_causal 0.0281 /// teacc 98.79 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0344, -0.0308, -0.0316, ..., -0.0808, 0.0052, -0.0039], + [-0.0280, 0.0221, -0.1008, ..., 0.0393, 0.0423, -0.0805], + [ 0.0494, -0.0005, -0.0518, ..., -0.0363, -0.0229, 0.0003], + ..., + [ 0.0369, -0.0435, 0.0007, ..., 0.0246, -0.0113, 0.0342], + [-0.0179, -0.0344, -0.0427, ..., -0.0095, -0.0154, -0.0321], + [-0.0764, -0.1177, 0.0383, ..., -0.0488, 0.0093, 0.0210]], + device='cuda:0'), grad: tensor([[ 8.3353e-07, 0.0000e+00, 6.8918e-08, ..., 9.3132e-10, + 2.0228e-06, 0.0000e+00], + [ 9.1866e-06, 0.0000e+00, 1.9092e-07, ..., 4.6566e-09, + -4.9263e-05, 0.0000e+00], + [ 7.2643e-06, 0.0000e+00, 6.3330e-08, ..., 9.3132e-10, + 1.4089e-05, 0.0000e+00], + ..., + [-1.6123e-05, 0.0000e+00, -3.3062e-07, ..., 6.5193e-09, + 6.4969e-06, 0.0000e+00], + [-7.9162e-07, 0.0000e+00, 6.7055e-08, ..., 0.0000e+00, + -1.4335e-05, 0.0000e+00], + [ 5.4426e-06, 0.0000e+00, 1.1902e-06, ..., 1.8626e-09, + 2.3276e-05, 0.0000e+00]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0043, 0.0280, 0.0055, -0.0019, 0.0094, 0.0004, 0.0195, -0.0191, + 0.0174, -0.0006], device='cuda:0'), grad: tensor([ 8.6520e-07, -1.1861e-04, 4.8667e-05, -3.8683e-05, 2.2829e-05, + 4.7565e-05, 1.4812e-05, -1.3679e-05, -3.6657e-05, 7.2837e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 220.23, cls_loss 0.0049 cls_loss_mapping 0.0077 cls_loss_causal 0.5981 re_mapping 0.0095 re_causal 0.0302 /// teacc 98.85 lr 0.00010000 +Epoch 102, weight, value: tensor([[-3.6489e-02, -3.0842e-02, -3.2005e-02, ..., -8.1052e-02, + 4.7590e-03, -5.4737e-03], + [-2.8206e-02, 2.2080e-02, -1.0224e-01, ..., 3.9480e-02, + 4.2716e-02, -8.1424e-02], + [ 4.9432e-02, -5.1822e-04, -5.2023e-02, ..., -3.6279e-02, + -2.3179e-02, 7.6837e-05], + ..., + [ 3.7007e-02, -4.3526e-02, 6.6154e-04, ..., 2.4375e-02, + -1.1564e-02, 3.4171e-02], + [-1.8206e-02, -3.4415e-02, -4.2652e-02, ..., -9.5949e-03, + -1.5960e-02, -3.3009e-02], + [-7.6831e-02, -1.1773e-01, 3.8692e-02, ..., -4.9303e-02, + 9.9592e-03, 2.2451e-02]], device='cuda:0'), grad: tensor([[ 6.8545e-07, 0.0000e+00, 4.5355e-07, ..., 0.0000e+00, + -2.2911e-07, 9.3132e-10], + [ 6.9514e-06, 0.0000e+00, 3.3323e-06, ..., 0.0000e+00, + 6.7391e-06, 0.0000e+00], + [ 4.8131e-06, 0.0000e+00, 9.9558e-07, ..., 0.0000e+00, + 1.6466e-05, 0.0000e+00], + ..., + [-2.0906e-05, 0.0000e+00, 4.2111e-05, ..., 0.0000e+00, + 1.5534e-06, 9.3132e-10], + [ 4.7199e-06, 0.0000e+00, 1.2564e-06, ..., 0.0000e+00, + -6.0886e-05, 9.3132e-09], + [ 1.3299e-06, 0.0000e+00, 1.7929e-04, ..., 0.0000e+00, + -5.2899e-07, 1.8626e-09]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0053, 0.0281, 0.0052, -0.0015, 0.0089, 0.0005, 0.0198, -0.0191, + 0.0170, 0.0002], device='cuda:0'), grad: tensor([-7.5847e-06, 5.0664e-05, 5.9903e-05, -2.1353e-05, -3.3379e-04, + 1.0902e-04, 2.3350e-05, 3.3051e-05, -1.8108e-04, 2.6751e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 219.97, cls_loss 0.0055 cls_loss_mapping 0.0081 cls_loss_causal 0.5910 re_mapping 0.0098 re_causal 0.0284 /// teacc 98.82 lr 0.00010000 +Epoch 103, weight, value: tensor([[-3.6805e-02, -3.1012e-02, -3.2107e-02, ..., -8.1267e-02, + 4.7487e-03, -5.6299e-03], + [-2.8118e-02, 2.2084e-02, -1.0252e-01, ..., 3.9501e-02, + 4.2437e-02, -8.1695e-02], + [ 4.8978e-02, -2.9018e-04, -5.1800e-02, ..., -3.6337e-02, + -2.3783e-02, 9.4734e-05], + ..., + [ 3.7675e-02, -4.3732e-02, 5.3099e-04, ..., 2.4381e-02, + -1.1898e-02, 3.4208e-02], + [-1.8497e-02, -3.4439e-02, -4.2968e-02, ..., -9.6682e-03, + -1.6326e-02, -3.3269e-02], + [-7.7103e-02, -1.1787e-01, 3.8125e-02, ..., -4.9589e-02, + 1.1430e-02, 2.2593e-02]], device='cuda:0'), grad: tensor([[ 7.4320e-07, 0.0000e+00, 1.9997e-05, ..., 1.2573e-07, + 3.8058e-05, 3.7253e-09], + [ 3.3472e-06, 0.0000e+00, 3.8370e-06, ..., 3.5763e-07, + 1.8710e-06, 3.7253e-09], + [-4.1537e-06, 0.0000e+00, 5.2862e-06, ..., 7.7300e-08, + 4.3996e-06, -8.4750e-08], + ..., + [-1.9163e-05, 0.0000e+00, 2.0787e-06, ..., -4.3698e-06, + 3.8818e-06, 1.3039e-08], + [ 1.2266e-06, 0.0000e+00, 4.3772e-07, ..., 4.9360e-08, + 1.7673e-05, 3.0734e-08], + [ 1.4603e-05, 0.0000e+00, -8.8096e-05, ..., 3.9563e-06, + -1.3924e-04, 3.7253e-09]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0048, 0.0281, 0.0043, -0.0017, 0.0088, 0.0003, 0.0202, -0.0185, + 0.0166, 0.0005], device='cuda:0'), grad: tensor([ 2.3580e-04, 3.7789e-05, 1.9640e-05, 2.2024e-05, 6.1655e-04, + 1.6823e-05, 2.2396e-05, -6.7279e-06, 4.7803e-05, -1.0118e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 220.05, cls_loss 0.0062 cls_loss_mapping 0.0108 cls_loss_causal 0.5879 re_mapping 0.0097 re_causal 0.0290 /// teacc 98.87 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0372, -0.0314, -0.0326, ..., -0.0813, 0.0045, -0.0057], + [-0.0287, 0.0221, -0.1038, ..., 0.0397, 0.0423, -0.0833], + [ 0.0489, -0.0010, -0.0521, ..., -0.0364, -0.0239, 0.0015], + ..., + [ 0.0382, -0.0427, 0.0004, ..., 0.0244, -0.0124, 0.0340], + [-0.0191, -0.0345, -0.0431, ..., -0.0097, -0.0165, -0.0338], + [-0.0773, -0.1185, 0.0383, ..., -0.0499, 0.0114, 0.0225]], + device='cuda:0'), grad: tensor([[ 4.8801e-06, 0.0000e+00, 3.4366e-07, ..., 0.0000e+00, + 1.7863e-06, -6.0629e-07], + [ 4.2021e-05, 0.0000e+00, 1.4063e-07, ..., 0.0000e+00, + -2.1011e-06, 3.6322e-08], + [-5.3585e-05, 0.0000e+00, 3.0827e-07, ..., 0.0000e+00, + 8.3167e-07, 1.5367e-07], + ..., + [-3.4332e-05, 0.0000e+00, 2.3656e-07, ..., 0.0000e+00, + 1.2573e-06, 9.2201e-08], + [ 1.2048e-05, 0.0000e+00, 1.1548e-07, ..., 0.0000e+00, + 2.6934e-06, 9.5926e-08], + [ 1.2487e-05, 0.0000e+00, 6.4913e-07, ..., 0.0000e+00, + 3.5763e-07, 4.9360e-08]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0036, 0.0277, 0.0041, -0.0008, 0.0091, -0.0004, 0.0206, -0.0183, + 0.0162, -0.0001], device='cuda:0'), grad: tensor([-3.6210e-06, 9.2983e-05, -1.3793e-04, -1.1981e-05, 1.0997e-05, + 1.8626e-05, 2.5213e-05, -5.8264e-05, 2.3052e-05, 4.0591e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 219.73, cls_loss 0.0055 cls_loss_mapping 0.0080 cls_loss_causal 0.5779 re_mapping 0.0097 re_causal 0.0281 /// teacc 98.79 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0377, -0.0315, -0.0327, ..., -0.0816, 0.0042, -0.0059], + [-0.0282, 0.0222, -0.1054, ..., 0.0401, 0.0427, -0.0848], + [ 0.0493, -0.0010, -0.0513, ..., -0.0364, -0.0243, 0.0016], + ..., + [ 0.0382, -0.0427, 0.0005, ..., 0.0247, -0.0134, 0.0341], + [-0.0195, -0.0345, -0.0432, ..., -0.0098, -0.0164, -0.0341], + [-0.0782, -0.1185, 0.0380, ..., -0.0498, 0.0116, 0.0225]], + device='cuda:0'), grad: tensor([[ 9.7603e-07, 0.0000e+00, 6.4075e-07, ..., 0.0000e+00, + 3.6448e-05, 2.1420e-08], + [-9.6798e-05, 0.0000e+00, 1.9465e-07, ..., 0.0000e+00, + -2.4629e-04, 7.4506e-09], + [ 7.4357e-06, 0.0000e+00, 3.0641e-07, ..., 0.0000e+00, + 1.6451e-05, 3.7253e-09], + ..., + [ 6.8307e-05, 0.0000e+00, 2.0862e-07, ..., 0.0000e+00, + 1.7691e-04, 1.5832e-08], + [ 3.5577e-06, 0.0000e+00, 9.3691e-07, ..., 0.0000e+00, + 3.7938e-05, 1.9185e-07], + [ 8.3745e-06, 0.0000e+00, 2.2054e-06, ..., 0.0000e+00, + -4.7654e-05, 1.6764e-07]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0042, 0.0281, 0.0044, -0.0022, 0.0098, 0.0007, 0.0211, -0.0186, + 0.0160, -0.0006], device='cuda:0'), grad: tensor([ 3.2902e-04, -7.4434e-04, 1.5211e-04, 9.5963e-05, 1.7428e-04, + 9.2313e-06, -8.7991e-06, 5.7507e-04, -1.1958e-05, -5.7125e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 219.80, cls_loss 0.0053 cls_loss_mapping 0.0086 cls_loss_causal 0.6177 re_mapping 0.0097 re_causal 0.0289 /// teacc 98.84 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0375, -0.0318, -0.0335, ..., -0.0818, 0.0043, -0.0078], + [-0.0294, 0.0224, -0.1056, ..., 0.0403, 0.0426, -0.0851], + [ 0.0495, -0.0010, -0.0514, ..., -0.0367, -0.0229, 0.0016], + ..., + [ 0.0389, -0.0428, 0.0004, ..., 0.0246, -0.0140, 0.0341], + [-0.0201, -0.0335, -0.0428, ..., -0.0099, -0.0162, -0.0342], + [-0.0784, -0.1199, 0.0380, ..., -0.0498, 0.0120, 0.0244]], + device='cuda:0'), grad: tensor([[ 2.0377e-06, 3.2596e-09, 1.2163e-06, ..., 2.6962e-07, + 5.1316e-07, 1.3504e-08], + [ 1.1802e-05, 4.7963e-08, 1.4426e-06, ..., 1.4808e-06, + -2.5108e-06, 2.2817e-08], + [-1.8716e-05, -2.3795e-07, 3.4971e-07, ..., 2.9616e-07, + 4.3586e-07, -1.3085e-07], + ..., + [-8.3983e-05, 1.5227e-07, -1.7285e-05, ..., -1.8656e-05, + 2.1756e-06, 2.5611e-08], + [ 2.4855e-05, 8.3819e-09, 6.1421e-07, ..., 2.2911e-07, + -2.1867e-06, 6.0536e-09], + [ 4.7535e-05, 9.3132e-10, 8.0317e-06, ..., 7.7635e-06, + 2.5565e-07, 5.1223e-09]], device='cuda:0') +Epoch 106, bias, value: tensor([-4.4666e-03, 2.7202e-02, 4.9677e-03, -1.8957e-03, 9.6344e-03, + -4.6174e-04, 2.1111e-02, -1.8173e-02, 1.6226e-02, -7.8943e-05], + device='cuda:0'), grad: tensor([-8.6927e-04, 3.1650e-05, -3.9786e-06, 3.0935e-05, 3.7134e-05, + 7.3686e-06, 1.0520e-05, -2.5725e-04, 1.6838e-05, 9.9659e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 219.41, cls_loss 0.0056 cls_loss_mapping 0.0066 cls_loss_causal 0.5660 re_mapping 0.0099 re_causal 0.0279 /// teacc 98.81 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0380, -0.0338, -0.0338, ..., -0.0819, 0.0038, -0.0078], + [-0.0297, 0.0233, -0.1064, ..., 0.0406, 0.0433, -0.0856], + [ 0.0496, -0.0015, -0.0515, ..., -0.0367, -0.0233, 0.0018], + ..., + [ 0.0392, -0.0427, 0.0006, ..., 0.0247, -0.0144, 0.0343], + [-0.0206, -0.0329, -0.0432, ..., -0.0100, -0.0165, -0.0344], + [-0.0791, -0.1209, 0.0382, ..., -0.0500, 0.0121, 0.0244]], + device='cuda:0'), grad: tensor([[ 7.0035e-07, 2.7474e-08, 4.5775e-07, ..., 0.0000e+00, + 2.5127e-06, 2.9802e-08], + [ 3.5707e-06, 2.2817e-08, 4.5868e-07, ..., 0.0000e+00, + -2.4401e-06, 8.5216e-08], + [-3.9548e-05, -5.2452e-06, 9.3598e-08, ..., 0.0000e+00, + 2.2724e-07, -1.7229e-07], + ..., + [ 1.9401e-05, 5.1335e-06, 2.0443e-07, ..., 0.0000e+00, + 1.3933e-06, -2.0443e-07], + [ 3.3937e-06, 3.6322e-08, 2.0474e-05, ..., 0.0000e+00, + 9.7603e-06, 5.7742e-08], + [ 7.6666e-06, 1.3970e-09, 5.3365e-07, ..., 0.0000e+00, + -1.6429e-06, 6.2864e-08]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0046, 0.0273, 0.0049, -0.0016, 0.0094, -0.0007, 0.0190, -0.0179, + 0.0183, -0.0001], device='cuda:0'), grad: tensor([-1.5404e-06, 9.8720e-07, -5.4210e-05, 2.7239e-05, 1.2495e-05, + -1.6248e-04, 5.5850e-05, 2.2382e-05, 9.7394e-05, 1.6484e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 219.79, cls_loss 0.0052 cls_loss_mapping 0.0092 cls_loss_causal 0.5927 re_mapping 0.0095 re_causal 0.0290 /// teacc 98.65 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0382, -0.0345, -0.0339, ..., -0.0820, 0.0035, -0.0079], + [-0.0309, 0.0235, -0.1067, ..., 0.0407, 0.0445, -0.0868], + [ 0.0506, -0.0014, -0.0512, ..., -0.0368, -0.0235, 0.0022], + ..., + [ 0.0392, -0.0429, 0.0004, ..., 0.0250, -0.0150, 0.0342], + [-0.0208, -0.0336, -0.0445, ..., -0.0100, -0.0169, -0.0345], + [-0.0795, -0.1217, 0.0382, ..., -0.0508, 0.0120, 0.0244]], + device='cuda:0'), grad: tensor([[ 1.3961e-06, 0.0000e+00, 4.7497e-08, ..., 0.0000e+00, + 2.9981e-05, 1.3292e-05], + [ 7.4413e-07, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + -6.9849e-07, 6.2399e-08], + [ 4.8578e-06, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 4.6268e-06, 1.6568e-06], + ..., + [ 1.1353e-06, 0.0000e+00, 3.5390e-08, ..., 0.0000e+00, + 1.3700e-06, 9.3132e-09], + [ 2.0396e-06, 0.0000e+00, 3.5297e-07, ..., 0.0000e+00, + 2.5164e-06, 2.6729e-07], + [ 2.6077e-06, 0.0000e+00, 3.6322e-08, ..., 0.0000e+00, + -1.6456e-06, 3.5390e-08]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0045, 0.0273, 0.0062, -0.0019, 0.0109, -0.0006, 0.0191, -0.0187, + 0.0177, -0.0013], device='cuda:0'), grad: tensor([ 1.6916e-04, 1.0170e-06, 3.1114e-05, -2.7850e-05, 8.5160e-06, + 2.6435e-05, -2.2793e-04, 5.0329e-06, 1.3761e-05, 8.7079e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 219.86, cls_loss 0.0052 cls_loss_mapping 0.0074 cls_loss_causal 0.5872 re_mapping 0.0096 re_causal 0.0278 /// teacc 98.86 lr 0.00010000 +Epoch 109, weight, value: tensor([[-3.8403e-02, -3.5559e-02, -3.4272e-02, ..., -8.2115e-02, + 2.9623e-03, -8.0154e-03], + [-3.1084e-02, 2.2822e-02, -1.0705e-01, ..., 3.8417e-02, + 4.4395e-02, -8.6931e-02], + [ 5.0280e-02, -1.7909e-03, -5.1552e-02, ..., -3.6777e-02, + -2.4747e-02, 2.1256e-03], + ..., + [ 3.9916e-02, -4.1949e-02, 1.0697e-04, ..., 2.5264e-02, + -1.5121e-02, 3.4228e-02], + [-2.1378e-02, -3.5603e-02, -4.5783e-02, ..., -1.0111e-02, + -1.7135e-02, -3.4575e-02], + [-7.9854e-02, -1.2301e-01, 3.7948e-02, ..., -4.8472e-02, + 1.2831e-02, 2.4600e-02]], device='cuda:0'), grad: tensor([[ 5.8301e-07, -2.1532e-06, 7.0594e-07, ..., 0.0000e+00, + 1.6959e-06, 0.0000e+00], + [ 1.4745e-05, 6.0536e-08, 1.3905e-06, ..., 0.0000e+00, + 6.9849e-07, 0.0000e+00], + [-2.2218e-05, 8.7544e-08, -9.7789e-08, ..., 0.0000e+00, + -6.2399e-07, 0.0000e+00], + ..., + [-2.6971e-06, 3.1292e-07, 2.0228e-06, ..., 0.0000e+00, + 6.8918e-07, 0.0000e+00], + [ 2.8759e-06, 1.5274e-07, 6.7167e-06, ..., 0.0000e+00, + 2.2873e-06, 0.0000e+00], + [ 2.0433e-06, 2.3656e-07, -1.3580e-03, ..., 0.0000e+00, + -3.6788e-04, 0.0000e+00]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0042, 0.0271, 0.0052, -0.0021, 0.0114, -0.0006, 0.0191, -0.0176, + 0.0169, -0.0011], device='cuda:0'), grad: tensor([-1.9848e-05, 3.2067e-05, -3.6597e-05, 2.2620e-05, 3.3932e-03, + 2.1175e-05, -4.8071e-05, 1.7164e-06, 3.0056e-05, -3.3989e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 219.60, cls_loss 0.0059 cls_loss_mapping 0.0100 cls_loss_causal 0.5958 re_mapping 0.0092 re_causal 0.0278 /// teacc 98.78 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0397, -0.0375, -0.0346, ..., -0.0836, 0.0025, -0.0081], + [-0.0318, 0.0207, -0.1075, ..., 0.0384, 0.0443, -0.0871], + [ 0.0500, -0.0008, -0.0536, ..., -0.0373, -0.0245, 0.0021], + ..., + [ 0.0403, -0.0417, -0.0009, ..., 0.0246, -0.0151, 0.0343], + [-0.0218, -0.0337, -0.0461, ..., -0.0106, -0.0170, -0.0346], + [-0.0807, -0.1270, 0.0384, ..., -0.0486, 0.0131, 0.0246]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, 2.0489e-08, 3.4254e-06, ..., 4.1258e-07, + 2.0452e-06, 0.0000e+00], + [ 2.3656e-07, 1.0245e-08, 7.1153e-07, ..., 2.0489e-08, + -3.0994e-06, 0.0000e+00], + [-8.2608e-07, -7.0874e-07, 5.8208e-07, ..., 7.4506e-09, + 3.9581e-07, 0.0000e+00], + ..., + [-7.5437e-08, 2.5518e-07, 6.0722e-06, ..., 5.5879e-08, + 1.3607e-06, 0.0000e+00], + [ 7.8231e-07, 1.6112e-07, 1.4920e-06, ..., 9.3132e-08, + 1.9316e-06, 0.0000e+00], + [ 7.1526e-07, 1.2107e-08, 1.4193e-05, ..., 1.0617e-07, + 4.4703e-08, 0.0000e+00]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0047, 0.0257, 0.0052, 0.0003, 0.0111, -0.0010, 0.0188, -0.0171, + 0.0169, -0.0009], device='cuda:0'), grad: tensor([-2.3380e-05, -2.4252e-06, 3.3677e-06, 3.6340e-06, -8.4877e-05, + -2.2605e-05, 3.2932e-06, 4.2081e-05, 1.0341e-05, 7.0393e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 219.83, cls_loss 0.0051 cls_loss_mapping 0.0073 cls_loss_causal 0.6049 re_mapping 0.0095 re_causal 0.0287 /// teacc 98.85 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0401, -0.0383, -0.0351, ..., -0.0838, 0.0019, -0.0081], + [-0.0320, 0.0204, -0.1081, ..., 0.0385, 0.0450, -0.0875], + [ 0.0501, -0.0009, -0.0542, ..., -0.0374, -0.0250, 0.0025], + ..., + [ 0.0405, -0.0414, -0.0002, ..., 0.0248, -0.0157, 0.0342], + [-0.0220, -0.0337, -0.0465, ..., -0.0106, -0.0177, -0.0348], + [-0.0817, -0.1276, 0.0384, ..., -0.0487, 0.0133, 0.0247]], + device='cuda:0'), grad: tensor([[ 2.1830e-06, 1.3504e-07, 3.1628e-06, ..., 0.0000e+00, + 7.7300e-07, 4.3772e-07], + [ 1.8273e-06, 5.7742e-08, 7.6368e-07, ..., 0.0000e+00, + -2.2873e-06, 1.3411e-07], + [-1.3113e-06, -1.6158e-06, 5.5656e-06, ..., 0.0000e+00, + 2.1569e-06, -8.2795e-07], + ..., + [ 1.4260e-05, 1.2852e-07, 1.4743e-06, ..., 0.0000e+00, + 7.3388e-06, 2.8033e-07], + [ 4.4182e-06, 2.0210e-07, 4.3809e-06, ..., 0.0000e+00, + 2.0806e-06, 9.9465e-07], + [ 5.1379e-05, 1.5832e-08, 5.2825e-06, ..., 0.0000e+00, + 1.9237e-05, 3.3900e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([-5.3301e-03, 2.5805e-02, 4.9322e-03, 3.8411e-05, 1.1153e-02, + -4.7907e-04, 1.9546e-02, -1.6959e-02, 1.6641e-02, -1.1235e-03], + device='cuda:0'), grad: tensor([-5.1051e-05, -3.2093e-06, 2.5973e-05, -1.5163e-04, -3.4308e-04, + -1.0085e-04, 4.3035e-04, 5.0932e-05, 3.2067e-05, 1.0979e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 220.27, cls_loss 0.0051 cls_loss_mapping 0.0084 cls_loss_causal 0.5764 re_mapping 0.0097 re_causal 0.0273 /// teacc 98.81 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0405, -0.0396, -0.0347, ..., -0.0838, 0.0017, -0.0086], + [-0.0311, 0.0213, -0.1087, ..., 0.0385, 0.0462, -0.0880], + [ 0.0502, -0.0020, -0.0536, ..., -0.0374, -0.0254, 0.0026], + ..., + [ 0.0405, -0.0406, 0.0002, ..., 0.0251, -0.0170, 0.0342], + [-0.0222, -0.0338, -0.0465, ..., -0.0106, -0.0174, -0.0356], + [-0.0823, -0.1288, 0.0387, ..., -0.0487, 0.0137, 0.0247]], + device='cuda:0'), grad: tensor([[ 2.3246e-06, 8.3819e-09, 1.0341e-05, ..., 0.0000e+00, + 2.5574e-06, 0.0000e+00], + [ 8.8010e-07, 1.8626e-09, 3.6843e-06, ..., 0.0000e+00, + -2.1979e-07, 0.0000e+00], + [-7.9256e-07, 9.3132e-10, 2.4289e-06, ..., 0.0000e+00, + 1.6838e-06, 0.0000e+00], + ..., + [-1.6140e-06, 1.8626e-09, 1.4961e-05, ..., 0.0000e+00, + 6.4820e-07, 0.0000e+00], + [-1.1418e-06, 5.2154e-08, 4.6864e-06, ..., 0.0000e+00, + -1.4819e-05, 0.0000e+00], + [ 1.9986e-06, 8.3819e-09, 1.9264e-04, ..., 0.0000e+00, + -5.7183e-07, 0.0000e+00]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0049, 0.0272, 0.0048, -0.0007, 0.0104, -0.0015, 0.0205, -0.0174, + 0.0169, -0.0010], device='cuda:0'), grad: tensor([ 3.0816e-05, 2.3514e-05, 8.3089e-05, 2.1005e-04, -1.1311e-03, + -1.1855e-04, 1.1645e-05, 1.3542e-04, -3.0446e-04, 1.0605e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 219.73, cls_loss 0.0060 cls_loss_mapping 0.0070 cls_loss_causal 0.6029 re_mapping 0.0092 re_causal 0.0268 /// teacc 98.88 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0410, -0.0416, -0.0349, ..., -0.0840, 0.0015, -0.0088], + [-0.0308, 0.0216, -0.1101, ..., 0.0388, 0.0468, -0.0901], + [ 0.0507, -0.0021, -0.0546, ..., -0.0374, -0.0257, 0.0020], + ..., + [ 0.0405, -0.0407, 0.0014, ..., 0.0249, -0.0180, 0.0341], + [-0.0225, -0.0344, -0.0476, ..., -0.0107, -0.0170, -0.0371], + [-0.0834, -0.1302, 0.0389, ..., -0.0488, 0.0139, 0.0247]], + device='cuda:0'), grad: tensor([[ 1.9558e-07, 0.0000e+00, 4.5821e-06, ..., 0.0000e+00, + 1.8282e-06, 0.0000e+00], + [ 2.3190e-07, 0.0000e+00, -1.5914e-05, ..., 1.8626e-09, + -3.9071e-05, 0.0000e+00], + [-3.0361e-06, 0.0000e+00, 1.4994e-07, ..., 9.3132e-10, + 1.5991e-06, 0.0000e+00], + ..., + [ 9.9279e-07, 0.0000e+00, 3.0920e-06, ..., 8.3819e-09, + 5.6997e-06, 0.0000e+00], + [-5.7146e-06, 0.0000e+00, 5.6982e-05, ..., 9.3132e-10, + 2.2531e-05, 0.0000e+00], + [ 3.7812e-07, 0.0000e+00, -4.5347e-04, ..., 1.1176e-08, + -2.0897e-04, 0.0000e+00]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0051, 0.0274, 0.0048, -0.0015, 0.0101, -0.0007, 0.0202, -0.0174, + 0.0173, -0.0010], device='cuda:0'), grad: tensor([ 1.4007e-06, -1.2732e-04, 1.7891e-06, 7.3624e-04, 1.2189e-04, + 3.4404e-04, 4.4614e-05, 2.5123e-05, 9.4354e-05, -1.2417e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 112---------------------------------------------------- +epoch 112, time 220.48, cls_loss 0.0040 cls_loss_mapping 0.0059 cls_loss_causal 0.5629 re_mapping 0.0095 re_causal 0.0286 /// teacc 98.89 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0416, -0.0419, -0.0361, ..., -0.0842, 0.0012, -0.0088], + [-0.0308, 0.0217, -0.1108, ..., 0.0389, 0.0475, -0.0909], + [ 0.0509, -0.0021, -0.0534, ..., -0.0377, -0.0263, 0.0035], + ..., + [ 0.0407, -0.0408, 0.0013, ..., 0.0249, -0.0181, 0.0340], + [-0.0225, -0.0343, -0.0476, ..., -0.0108, -0.0173, -0.0379], + [-0.0839, -0.1303, 0.0393, ..., -0.0489, 0.0144, 0.0246]], + device='cuda:0'), grad: tensor([[ 3.4459e-07, 4.6566e-09, 3.3993e-07, ..., 0.0000e+00, + 4.8801e-07, 6.3330e-08], + [ 4.1574e-06, 9.3132e-09, 2.7064e-06, ..., 0.0000e+00, + -8.3968e-06, 2.4214e-08], + [-1.7732e-05, -1.5739e-07, 1.5255e-06, ..., 0.0000e+00, + 9.2015e-07, 3.5390e-08], + ..., + [-3.6098e-06, 1.5832e-08, 1.9614e-06, ..., 0.0000e+00, + 1.7537e-06, 9.3132e-10], + [ 7.2606e-06, 7.0781e-08, 1.9046e-06, ..., 0.0000e+00, + 4.4405e-06, 2.9802e-08], + [ 1.5832e-06, 1.8626e-09, -8.3260e-07, ..., 0.0000e+00, + -5.6550e-06, 1.8626e-09]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0057, 0.0279, 0.0047, -0.0011, 0.0098, -0.0015, 0.0200, -0.0174, + 0.0176, -0.0008], device='cuda:0'), grad: tensor([ 2.4587e-06, -5.3048e-06, -2.0295e-05, 8.5086e-06, -1.9427e-06, + 9.2015e-07, 2.9840e-06, 1.7984e-06, 2.1785e-05, -1.0937e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 219.65, cls_loss 0.0048 cls_loss_mapping 0.0087 cls_loss_causal 0.5868 re_mapping 0.0096 re_causal 0.0280 /// teacc 98.83 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0419, -0.0422, -0.0371, ..., -0.0843, 0.0009, -0.0088], + [-0.0309, 0.0223, -0.1121, ..., 0.0390, 0.0478, -0.0915], + [ 0.0512, -0.0020, -0.0527, ..., -0.0381, -0.0266, 0.0045], + ..., + [ 0.0407, -0.0411, 0.0011, ..., 0.0253, -0.0188, 0.0340], + [-0.0232, -0.0344, -0.0470, ..., -0.0108, -0.0183, -0.0381], + [-0.0845, -0.1310, 0.0389, ..., -0.0491, 0.0151, 0.0246]], + device='cuda:0'), grad: tensor([[ 4.8801e-07, 0.0000e+00, 5.6550e-06, ..., 4.0978e-08, + 4.6287e-07, 2.7940e-09], + [ 8.4117e-06, 0.0000e+00, 7.6666e-06, ..., 2.4214e-08, + 3.3155e-07, 9.3132e-10], + [ 1.0967e-05, 0.0000e+00, 8.8885e-06, ..., 3.7253e-09, + 8.9686e-07, -5.4017e-08], + ..., + [-1.7602e-06, 0.0000e+00, 7.7710e-06, ..., -4.6566e-08, + 7.3053e-06, 7.4506e-09], + [ 2.0191e-05, 0.0000e+00, 9.5963e-06, ..., 8.3819e-08, + -6.0461e-06, 2.4214e-08], + [ 8.5868e-07, 0.0000e+00, 1.5557e-05, ..., 4.6566e-08, + -5.7220e-06, 9.3132e-10]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0051, 0.0280, 0.0047, -0.0013, 0.0102, -0.0013, 0.0198, -0.0177, + 0.0173, -0.0006], device='cuda:0'), grad: tensor([-1.1814e-04, 4.5747e-05, 5.8800e-05, -5.1022e-05, -3.1257e-04, + -6.2406e-05, 2.9612e-04, 1.3113e-04, 5.0783e-05, -3.8475e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 114---------------------------------------------------- +epoch 114, time 220.54, cls_loss 0.0040 cls_loss_mapping 0.0069 cls_loss_causal 0.5432 re_mapping 0.0096 re_causal 0.0265 /// teacc 98.92 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0423, -0.0426, -0.0373, ..., -0.0843, 0.0007, -0.0092], + [-0.0312, 0.0222, -0.1125, ..., 0.0390, 0.0470, -0.0932], + [ 0.0517, -0.0015, -0.0525, ..., -0.0384, -0.0267, 0.0045], + ..., + [ 0.0407, -0.0413, 0.0012, ..., 0.0253, -0.0177, 0.0339], + [-0.0237, -0.0342, -0.0470, ..., -0.0109, -0.0186, -0.0386], + [-0.0853, -0.1316, 0.0386, ..., -0.0491, 0.0151, 0.0247]], + device='cuda:0'), grad: tensor([[ 4.0606e-07, 1.8626e-09, 3.8370e-07, ..., 9.3132e-10, + 3.8557e-07, 0.0000e+00], + [ 9.6038e-06, 1.6764e-08, 2.0832e-05, ..., 1.8626e-09, + 1.8403e-05, 0.0000e+00], + [ 3.4273e-06, -8.5682e-08, 5.6159e-07, ..., 4.6566e-09, + 1.6233e-06, 0.0000e+00], + ..., + [-7.3835e-06, 1.3970e-08, -9.6619e-05, ..., 1.6764e-08, + -6.5923e-05, 0.0000e+00], + [-9.4891e-05, 3.8184e-08, -4.2260e-05, ..., 1.8626e-09, + -8.6606e-05, 0.0000e+00], + [ 2.0843e-06, 1.8626e-09, 7.0691e-05, ..., 2.7940e-09, + 4.8637e-05, 0.0000e+00]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0052, 0.0264, 0.0050, -0.0018, 0.0099, -0.0006, 0.0199, -0.0164, + 0.0176, -0.0011], device='cuda:0'), grad: tensor([ 1.8645e-06, 1.5104e-04, 1.2010e-05, 2.0695e-04, 3.3528e-05, + 5.3078e-05, 1.7393e-04, -5.8126e-04, -4.6754e-04, 4.1628e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 219.85, cls_loss 0.0056 cls_loss_mapping 0.0099 cls_loss_causal 0.5739 re_mapping 0.0091 re_causal 0.0261 /// teacc 98.77 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0407, -0.0432, -0.0374, ..., -0.0845, 0.0027, -0.0093], + [-0.0325, 0.0221, -0.1119, ..., 0.0402, 0.0472, -0.0946], + [ 0.0514, -0.0006, -0.0528, ..., -0.0384, -0.0271, 0.0048], + ..., + [ 0.0411, -0.0412, 0.0010, ..., 0.0249, -0.0189, 0.0341], + [-0.0247, -0.0362, -0.0474, ..., -0.0110, -0.0186, -0.0394], + [-0.0860, -0.1327, 0.0384, ..., -0.0494, 0.0161, 0.0249]], + device='cuda:0'), grad: tensor([[ 1.3076e-06, 1.2107e-08, 1.0677e-05, ..., 0.0000e+00, + 1.1912e-06, 3.4180e-07], + [ 1.3467e-06, 1.0245e-08, 2.3469e-07, ..., 0.0000e+00, + -6.3404e-06, 6.9849e-08], + [-7.3854e-07, -1.8347e-07, 3.0026e-06, ..., 0.0000e+00, + 1.0692e-06, -1.3132e-07], + ..., + [-3.8855e-06, 2.4214e-08, 5.8487e-07, ..., 0.0000e+00, + 1.8552e-06, 5.3085e-08], + [ 1.3523e-06, 4.4703e-08, 1.0077e-06, ..., 0.0000e+00, + 5.0105e-06, 2.4587e-07], + [ 1.0654e-05, 2.7940e-09, -2.0608e-05, ..., 0.0000e+00, + 1.6108e-05, 1.3970e-08]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0032, 0.0256, 0.0052, -0.0014, 0.0094, -0.0002, 0.0200, -0.0167, + 0.0170, -0.0013], device='cuda:0'), grad: tensor([ 4.9114e-05, -9.1717e-06, 1.4983e-05, -4.3780e-05, 1.0595e-05, + 1.4469e-05, -6.2212e-06, -1.5805e-06, 1.5303e-05, -4.3660e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 219.65, cls_loss 0.0058 cls_loss_mapping 0.0087 cls_loss_causal 0.5783 re_mapping 0.0091 re_causal 0.0268 /// teacc 98.89 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0412, -0.0444, -0.0362, ..., -0.0850, 0.0034, -0.0121], + [-0.0339, 0.0230, -0.1138, ..., 0.0404, 0.0474, -0.0954], + [ 0.0513, -0.0005, -0.0519, ..., -0.0386, -0.0276, 0.0050], + ..., + [ 0.0422, -0.0417, 0.0007, ..., 0.0250, -0.0193, 0.0343], + [-0.0249, -0.0364, -0.0479, ..., -0.0112, -0.0190, -0.0396], + [-0.0869, -0.1341, 0.0369, ..., -0.0498, 0.0159, 0.0277]], + device='cuda:0'), grad: tensor([[ 9.4399e-06, 0.0000e+00, 2.0802e-05, ..., 0.0000e+00, + 6.4261e-08, 0.0000e+00], + [ 1.6140e-06, 0.0000e+00, 2.6338e-06, ..., 0.0000e+00, + -2.1979e-07, 0.0000e+00], + [ 4.3437e-06, 0.0000e+00, 7.7337e-06, ..., 0.0000e+00, + 2.7101e-07, -1.8626e-09], + ..., + [ 8.1304e-07, 0.0000e+00, 9.3505e-06, ..., 0.0000e+00, + 2.5332e-07, 0.0000e+00], + [ 1.5739e-06, 0.0000e+00, 1.3579e-06, ..., 0.0000e+00, + 7.8231e-08, 0.0000e+00], + [ 5.5805e-06, 0.0000e+00, 1.0952e-05, ..., 0.0000e+00, + -9.4995e-07, 0.0000e+00]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0037, 0.0250, 0.0049, -0.0013, 0.0106, -0.0005, 0.0202, -0.0159, + 0.0166, -0.0014], device='cuda:0'), grad: tensor([ 4.8786e-05, 8.2701e-06, 2.3931e-05, 1.1943e-05, -1.4257e-04, + -1.7434e-05, 3.3349e-05, 6.0797e-06, -1.7202e-06, 2.9370e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 219.61, cls_loss 0.0035 cls_loss_mapping 0.0055 cls_loss_causal 0.5805 re_mapping 0.0087 re_causal 0.0272 /// teacc 98.87 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0414, -0.0448, -0.0364, ..., -0.0850, 0.0033, -0.0122], + [-0.0341, 0.0231, -0.1143, ..., 0.0404, 0.0475, -0.0961], + [ 0.0523, -0.0004, -0.0503, ..., -0.0386, -0.0279, 0.0050], + ..., + [ 0.0422, -0.0416, 0.0007, ..., 0.0251, -0.0198, 0.0347], + [-0.0262, -0.0368, -0.0491, ..., -0.0112, -0.0194, -0.0397], + [-0.0871, -0.1347, 0.0367, ..., -0.0498, 0.0169, 0.0277]], + device='cuda:0'), grad: tensor([[ 6.3144e-07, 0.0000e+00, 5.0198e-07, ..., 0.0000e+00, + 2.3749e-07, 4.6566e-09], + [ 5.6066e-06, 0.0000e+00, 9.0525e-07, ..., 0.0000e+00, + -4.2841e-08, 4.6566e-09], + [-4.0114e-05, 0.0000e+00, 1.9064e-06, ..., 0.0000e+00, + 7.3947e-07, 2.7940e-09], + ..., + [ 4.6939e-05, 0.0000e+00, 1.7196e-05, ..., 0.0000e+00, + 2.2911e-07, 2.2352e-08], + [ 7.9349e-06, 0.0000e+00, 1.1995e-06, ..., 0.0000e+00, + 9.8720e-07, 2.7940e-09], + [ 4.9993e-06, 0.0000e+00, 6.9998e-06, ..., 0.0000e+00, + -1.8999e-07, 7.4506e-09]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0032, 0.0248, 0.0056, -0.0015, 0.0106, -0.0004, 0.0204, -0.0163, + 0.0156, -0.0013], device='cuda:0'), grad: tensor([ 2.1793e-06, 1.2830e-05, -3.2961e-05, -3.2187e-05, -9.1553e-05, + 3.6769e-06, 7.6108e-06, 8.7440e-05, 1.3083e-05, 2.9802e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 219.85, cls_loss 0.0037 cls_loss_mapping 0.0055 cls_loss_causal 0.5427 re_mapping 0.0089 re_causal 0.0256 /// teacc 98.83 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0416, -0.0450, -0.0365, ..., -0.0850, 0.0025, -0.0136], + [-0.0344, 0.0231, -0.1147, ..., 0.0404, 0.0476, -0.0969], + [ 0.0525, -0.0004, -0.0504, ..., -0.0387, -0.0277, 0.0049], + ..., + [ 0.0423, -0.0416, 0.0007, ..., 0.0251, -0.0199, 0.0344], + [-0.0266, -0.0368, -0.0503, ..., -0.0112, -0.0201, -0.0400], + [-0.0880, -0.1349, 0.0374, ..., -0.0498, 0.0172, 0.0292]], + device='cuda:0'), grad: tensor([[ 1.9968e-06, 0.0000e+00, 2.7493e-06, ..., 6.7707e-07, + 1.9558e-06, 1.8626e-09], + [-2.8300e-04, 0.0000e+00, 8.0094e-07, ..., -1.1557e-04, + -8.0764e-05, 9.3132e-10], + [ 2.4331e-04, 0.0000e+00, 3.8818e-06, ..., 9.9182e-05, + 6.3241e-05, -2.5146e-08], + ..., + [ 2.2314e-06, 0.0000e+00, 8.8196e-07, ..., 8.5160e-06, + 5.8040e-06, 3.7253e-09], + [ 3.5837e-06, 0.0000e+00, 1.9558e-06, ..., 1.8626e-07, + 1.0006e-05, 1.8626e-09], + [ 1.0699e-05, 0.0000e+00, 2.6608e-04, ..., 1.7043e-07, + 2.2724e-07, 1.8626e-09]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0037, 0.0247, 0.0058, -0.0014, 0.0098, -0.0005, 0.0215, -0.0163, + 0.0147, -0.0005], device='cuda:0'), grad: tensor([-2.1830e-06, -8.9788e-04, 7.7772e-04, 2.5436e-05, -1.2054e-03, + 2.3901e-05, -9.4324e-06, 2.6017e-05, 4.8608e-05, 1.2121e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 219.49, cls_loss 0.0045 cls_loss_mapping 0.0060 cls_loss_causal 0.5151 re_mapping 0.0089 re_causal 0.0245 /// teacc 98.91 lr 0.00010000 +Epoch 121, weight, value: tensor([[-4.1891e-02, -4.5554e-02, -3.6993e-02, ..., -8.5104e-02, + 2.3789e-03, -1.3753e-02], + [-3.4267e-02, 2.3360e-02, -1.1493e-01, ..., 4.0970e-02, + 4.8083e-02, -9.7092e-02], + [ 5.2753e-02, -7.1994e-06, -5.0606e-02, ..., -3.9629e-02, + -2.8378e-02, 5.0982e-03], + ..., + [ 4.1737e-02, -4.2261e-02, 3.1517e-04, ..., 2.4773e-02, + -2.1074e-02, 3.4351e-02], + [-2.6481e-02, -3.6210e-02, -5.0717e-02, ..., -1.1237e-02, + -1.9154e-02, -4.0000e-02], + [-8.8695e-02, -1.3654e-01, 3.6906e-02, ..., -4.9851e-02, + 1.7697e-02, 2.9283e-02]], device='cuda:0'), grad: tensor([[ 7.0315e-07, 0.0000e+00, 1.1744e-06, ..., 0.0000e+00, + 2.7101e-06, 0.0000e+00], + [ 4.6380e-07, 0.0000e+00, 6.3423e-07, ..., 0.0000e+00, + -2.8815e-06, 0.0000e+00], + [-1.1571e-05, 0.0000e+00, -8.7917e-07, ..., 0.0000e+00, + 3.9861e-06, 0.0000e+00], + ..., + [-4.3958e-07, 0.0000e+00, 1.5646e-06, ..., 0.0000e+00, + 1.2740e-06, 0.0000e+00], + [ 2.5108e-06, 0.0000e+00, 5.3108e-05, ..., 0.0000e+00, + -5.8532e-05, 0.0000e+00], + [ 2.4866e-07, 0.0000e+00, -7.8738e-05, ..., 0.0000e+00, + 5.6505e-05, 0.0000e+00]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0034, 0.0249, 0.0060, -0.0009, 0.0099, -0.0005, 0.0210, -0.0172, + 0.0154, -0.0007], device='cuda:0'), grad: tensor([ 4.8786e-05, -6.2473e-06, 3.3826e-05, 1.0192e-04, 3.7503e-04, + -2.4390e-04, 5.5581e-05, 1.4514e-05, 3.8177e-05, -4.1676e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 219.65, cls_loss 0.0042 cls_loss_mapping 0.0060 cls_loss_causal 0.5517 re_mapping 0.0085 re_causal 0.0251 /// teacc 98.84 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0423, -0.0464, -0.0373, ..., -0.0852, 0.0020, -0.0138], + [-0.0338, 0.0234, -0.1153, ..., 0.0410, 0.0483, -0.0971], + [ 0.0527, 0.0008, -0.0507, ..., -0.0397, -0.0288, 0.0051], + ..., + [ 0.0415, -0.0431, 0.0005, ..., 0.0248, -0.0214, 0.0344], + [-0.0267, -0.0355, -0.0510, ..., -0.0113, -0.0192, -0.0401], + [-0.0895, -0.1379, 0.0366, ..., -0.0499, 0.0181, 0.0293]], + device='cuda:0'), grad: tensor([[ 3.2224e-07, 0.0000e+00, 2.0862e-07, ..., 0.0000e+00, + 3.1851e-07, 0.0000e+00], + [ 3.3900e-06, 0.0000e+00, 1.2843e-06, ..., 0.0000e+00, + 2.7381e-07, 0.0000e+00], + [ 5.3346e-06, 0.0000e+00, 1.3132e-07, ..., 0.0000e+00, + 7.0259e-06, 0.0000e+00], + ..., + [-1.0841e-05, 0.0000e+00, 4.3120e-07, ..., 0.0000e+00, + 3.8557e-07, 0.0000e+00], + [-1.0632e-05, 0.0000e+00, 7.9256e-07, ..., 0.0000e+00, + -2.5615e-05, 0.0000e+00], + [ 6.6869e-07, 0.0000e+00, 9.8526e-05, ..., 0.0000e+00, + 3.0287e-06, 0.0000e+00]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0036, 0.0253, 0.0059, -0.0012, 0.0097, 0.0006, 0.0206, -0.0174, + 0.0154, -0.0008], device='cuda:0'), grad: tensor([-1.7909e-06, 1.2085e-05, 4.6521e-05, 8.1480e-05, -3.6263e-04, + 1.0304e-05, 9.7752e-06, -1.3247e-05, -1.5104e-04, 3.6764e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 218.27, cls_loss 0.0038 cls_loss_mapping 0.0058 cls_loss_causal 0.5796 re_mapping 0.0081 re_causal 0.0251 /// teacc 98.73 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0426, -0.0470, -0.0376, ..., -0.0853, 0.0020, -0.0138], + [-0.0337, 0.0231, -0.1159, ..., 0.0410, 0.0484, -0.0973], + [ 0.0524, 0.0007, -0.0509, ..., -0.0398, -0.0290, 0.0054], + ..., + [ 0.0417, -0.0429, 0.0005, ..., 0.0249, -0.0216, 0.0344], + [-0.0266, -0.0341, -0.0521, ..., -0.0113, -0.0193, -0.0402], + [-0.0903, -0.1387, 0.0365, ..., -0.0499, 0.0185, 0.0293]], + device='cuda:0'), grad: tensor([[ 6.6124e-08, 0.0000e+00, 1.0338e-07, ..., 0.0000e+00, + 1.7602e-07, 6.8918e-08], + [ 7.8324e-07, 0.0000e+00, 1.8068e-07, ..., 2.7940e-09, + -6.5845e-07, 1.7695e-08], + [-1.5525e-06, 0.0000e+00, 1.4622e-07, ..., 9.3132e-10, + 2.9244e-07, -4.8429e-08], + ..., + [-4.4703e-07, 0.0000e+00, 6.1560e-07, ..., 8.3819e-09, + 1.0319e-06, 3.6322e-08], + [ 2.1048e-07, 0.0000e+00, 1.1146e-05, ..., 0.0000e+00, + 5.3924e-07, 4.0978e-08], + [ 7.4599e-07, 0.0000e+00, -9.6336e-06, ..., 9.3132e-10, + -1.3925e-05, 1.7136e-07]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0035, 0.0253, 0.0053, -0.0015, 0.0101, 0.0009, 0.0206, -0.0171, + 0.0153, -0.0009], device='cuda:0'), grad: tensor([-1.8924e-05, 2.5779e-06, 1.6484e-06, 3.5524e-05, 8.7440e-05, + -2.3335e-05, 4.7311e-06, 5.7630e-06, 2.2635e-05, -1.1802e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 217.89, cls_loss 0.0045 cls_loss_mapping 0.0068 cls_loss_causal 0.5693 re_mapping 0.0082 re_causal 0.0244 /// teacc 98.82 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0426, -0.0489, -0.0372, ..., -0.0853, 0.0019, -0.0138], + [-0.0349, 0.0233, -0.1164, ..., 0.0410, 0.0484, -0.0974], + [ 0.0517, 0.0010, -0.0507, ..., -0.0398, -0.0296, 0.0055], + ..., + [ 0.0434, -0.0433, 0.0004, ..., 0.0249, -0.0216, 0.0345], + [-0.0270, -0.0343, -0.0527, ..., -0.0113, -0.0191, -0.0402], + [-0.0913, -0.1407, 0.0360, ..., -0.0499, 0.0186, 0.0294]], + device='cuda:0'), grad: tensor([[ 6.2678e-07, 5.0291e-08, 3.0190e-05, ..., 0.0000e+00, + 4.0555e-04, 1.0395e-03], + [ 2.1532e-06, 6.4448e-07, 3.4459e-07, ..., 0.0000e+00, + -6.1281e-06, 1.1707e-06], + [ 6.0797e-06, 2.2855e-06, 4.6194e-07, ..., 0.0000e+00, + 7.1488e-06, 1.3039e-05], + ..., + [-1.4164e-05, -4.1462e-06, 1.9465e-07, ..., 0.0000e+00, + 4.1015e-06, 6.4299e-06], + [ 1.7900e-06, 5.0571e-07, 9.4436e-07, ..., 0.0000e+00, + 3.4794e-06, 4.2096e-07], + [ 2.3730e-06, 2.8871e-08, -3.0905e-05, ..., 0.0000e+00, + -4.2057e-04, -1.0691e-03]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0031, 0.0245, 0.0046, -0.0014, 0.0098, 0.0005, 0.0216, -0.0160, + 0.0155, -0.0015], device='cuda:0'), grad: tensor([ 3.3512e-03, -6.5193e-06, 7.2837e-05, 1.5944e-05, 2.4661e-05, + 9.9763e-06, -6.9775e-06, -3.8624e-05, 2.0161e-05, -3.4447e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 217.94, cls_loss 0.0048 cls_loss_mapping 0.0090 cls_loss_causal 0.5585 re_mapping 0.0081 re_causal 0.0244 /// teacc 98.73 lr 0.00010000 +Epoch 125, weight, value: tensor([[-4.2627e-02, -5.0596e-02, -3.7549e-02, ..., -8.5329e-02, + 1.2055e-03, -1.4830e-02], + [-3.6399e-02, 2.5624e-02, -1.1669e-01, ..., 4.1077e-02, + 4.8046e-02, -9.7861e-02], + [ 5.3142e-02, 5.3213e-05, -5.0821e-02, ..., -3.9793e-02, + -2.7590e-02, 5.0983e-03], + ..., + [ 4.3337e-02, -4.3581e-02, 1.4466e-03, ..., 2.4874e-02, + -2.3025e-02, 3.4299e-02], + [-2.7399e-02, -3.4579e-02, -5.3095e-02, ..., -1.1379e-02, + -1.9393e-02, -4.0246e-02], + [-9.2012e-02, -1.4119e-01, 3.4798e-02, ..., -4.9943e-02, + 1.9133e-02, 3.0366e-02]], device='cuda:0'), grad: tensor([[ 2.9523e-07, 9.3132e-10, 7.2643e-08, ..., 0.0000e+00, + 4.5002e-06, 0.0000e+00], + [ 9.2573e-07, 5.5879e-09, 2.2259e-07, ..., 9.3132e-10, + -5.0198e-07, 0.0000e+00], + [ 1.3959e-04, 1.1176e-08, 6.3330e-08, ..., 0.0000e+00, + 5.2482e-05, 0.0000e+00], + ..., + [ 1.1493e-06, -2.7940e-08, 1.4473e-06, ..., 1.3970e-08, + 3.7681e-06, 0.0000e+00], + [ 5.0142e-06, -2.7940e-09, 1.2480e-07, ..., 0.0000e+00, + 1.8567e-05, 0.0000e+00], + [ 1.0822e-06, 9.3132e-10, -2.6263e-07, ..., 1.8626e-09, + -1.7047e-05, 0.0000e+00]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0032, 0.0235, 0.0065, -0.0008, 0.0110, -0.0002, 0.0218, -0.0160, + 0.0151, -0.0026], device='cuda:0'), grad: tensor([ 9.6262e-06, 1.2014e-06, 2.0218e-04, -2.0838e-04, 2.0593e-05, + 1.2290e-04, -1.7917e-04, 1.5825e-05, 1.0389e-04, -8.8751e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 218.06, cls_loss 0.0034 cls_loss_mapping 0.0056 cls_loss_causal 0.5950 re_mapping 0.0083 re_causal 0.0250 /// teacc 98.79 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0428, -0.0509, -0.0376, ..., -0.0866, 0.0010, -0.0148], + [-0.0362, 0.0260, -0.1168, ..., 0.0423, 0.0494, -0.0980], + [ 0.0527, -0.0002, -0.0503, ..., -0.0405, -0.0284, 0.0052], + ..., + [ 0.0435, -0.0438, 0.0012, ..., 0.0238, -0.0230, 0.0343], + [-0.0281, -0.0339, -0.0535, ..., -0.0116, -0.0212, -0.0403], + [-0.0922, -0.1420, 0.0347, ..., -0.0500, 0.0193, 0.0304]], + device='cuda:0'), grad: tensor([[ 6.7875e-06, 0.0000e+00, 4.1537e-07, ..., 3.5390e-08, + 5.9139e-07, 1.6764e-08], + [ 3.1050e-06, 9.3132e-10, 2.1774e-06, ..., 6.9290e-07, + -1.7852e-05, 1.8626e-09], + [ 1.0759e-05, 9.3132e-10, 4.9081e-07, ..., 3.0734e-08, + 1.4435e-07, 1.8626e-09], + ..., + [ 1.6838e-05, 0.0000e+00, 1.6131e-06, ..., 5.0291e-08, + 8.5309e-07, 0.0000e+00], + [ 5.2489e-06, -5.5879e-09, 2.3097e-07, ..., 2.7940e-09, + 6.3889e-07, 3.6322e-08], + [ 6.8024e-06, 0.0000e+00, 3.1479e-06, ..., 1.2666e-07, + -5.5879e-09, 1.8626e-09]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0029, 0.0244, 0.0063, -0.0001, 0.0111, -0.0007, 0.0218, -0.0160, + 0.0137, -0.0028], device='cuda:0'), grad: tensor([ 3.1926e-06, -8.8871e-05, 2.6375e-05, -7.8917e-05, 7.8261e-05, + 6.2212e-06, 3.8929e-06, 2.9132e-05, -3.8520e-06, 2.4319e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 217.57, cls_loss 0.0039 cls_loss_mapping 0.0057 cls_loss_causal 0.5592 re_mapping 0.0083 re_causal 0.0252 /// teacc 98.84 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0426, -0.0521, -0.0376, ..., -0.0870, 0.0008, -0.0148], + [-0.0368, 0.0249, -0.1158, ..., 0.0442, 0.0492, -0.0984], + [ 0.0525, -0.0003, -0.0502, ..., -0.0409, -0.0288, 0.0055], + ..., + [ 0.0444, -0.0424, 0.0007, ..., 0.0229, -0.0225, 0.0340], + [-0.0284, -0.0344, -0.0547, ..., -0.0118, -0.0204, -0.0404], + [-0.0932, -0.1440, 0.0345, ..., -0.0502, 0.0194, 0.0304]], + device='cuda:0'), grad: tensor([[ 1.8310e-06, 0.0000e+00, 1.2293e-07, ..., 0.0000e+00, + 1.4342e-07, 0.0000e+00], + [ 1.1725e-06, 0.0000e+00, 1.0338e-07, ..., 0.0000e+00, + -2.3246e-06, 9.3132e-10], + [-1.0657e-04, 0.0000e+00, 1.7695e-08, ..., 0.0000e+00, + 3.2410e-07, -6.5193e-09], + ..., + [ 8.5354e-05, 0.0000e+00, 3.0454e-07, ..., 0.0000e+00, + 1.3057e-06, 1.8626e-09], + [ 4.9137e-06, 0.0000e+00, 5.5693e-07, ..., 0.0000e+00, + 1.3243e-06, 1.8626e-09], + [ 1.0684e-05, 0.0000e+00, -2.9616e-06, ..., 0.0000e+00, + -7.9125e-06, 0.0000e+00]], device='cuda:0') +Epoch 127, bias, value: tensor([-2.7319e-03, 2.4492e-02, 6.0260e-03, -1.9379e-05, 1.0846e-02, + -5.6764e-04, 2.1557e-02, -1.5500e-02, 1.3877e-02, -3.1854e-03], + device='cuda:0'), grad: tensor([-6.3851e-06, -2.0042e-06, -1.6820e-04, 1.6257e-05, 1.7643e-05, + -8.6054e-06, 7.4655e-06, 1.4496e-04, 1.2785e-05, -1.4186e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 217.37, cls_loss 0.0034 cls_loss_mapping 0.0062 cls_loss_causal 0.5403 re_mapping 0.0081 re_causal 0.0253 /// teacc 98.90 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0425, -0.0528, -0.0379, ..., -0.0872, 0.0007, -0.0148], + [-0.0369, 0.0253, -0.1164, ..., 0.0444, 0.0497, -0.0987], + [ 0.0513, -0.0005, -0.0504, ..., -0.0414, -0.0289, 0.0055], + ..., + [ 0.0457, -0.0424, 0.0005, ..., 0.0229, -0.0225, 0.0340], + [-0.0285, -0.0345, -0.0547, ..., -0.0119, -0.0210, -0.0404], + [-0.0939, -0.1444, 0.0344, ..., -0.0503, 0.0195, 0.0304]], + device='cuda:0'), grad: tensor([[ 1.8422e-06, 0.0000e+00, 1.8338e-06, ..., 4.6566e-10, + 1.2480e-07, 2.7940e-09], + [ 1.0021e-05, 1.8626e-09, 2.9337e-06, ..., 2.3283e-09, + -1.0785e-06, 2.0955e-08], + [ 2.4457e-06, 9.3132e-10, 1.3243e-06, ..., 1.8626e-09, + 1.7649e-07, 3.4925e-08], + ..., + [-2.2396e-05, -6.5193e-09, -3.8091e-06, ..., 4.6566e-09, + 4.9965e-07, -1.0431e-07], + [ 1.6659e-05, 4.6566e-10, 1.1057e-05, ..., 7.9162e-09, + -3.0734e-08, 4.1910e-09], + [ 2.5872e-06, 4.6566e-10, 8.7731e-07, ..., 1.1642e-08, + 4.6100e-08, 2.5611e-08]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0025, 0.0247, 0.0048, -0.0004, 0.0110, -0.0006, 0.0213, -0.0144, + 0.0139, -0.0034], device='cuda:0'), grad: tensor([ 1.6913e-05, 3.0816e-05, 1.3649e-05, 3.4094e-05, 1.5363e-05, + 2.3520e-04, -3.9649e-04, -5.6863e-05, 9.9123e-05, 8.1956e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 217.16, cls_loss 0.0054 cls_loss_mapping 0.0067 cls_loss_causal 0.5697 re_mapping 0.0086 re_causal 0.0245 /// teacc 98.79 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0431, -0.0539, -0.0384, ..., -0.0876, 0.0008, -0.0148], + [-0.0363, 0.0256, -0.1174, ..., 0.0447, 0.0508, -0.0988], + [ 0.0513, 0.0008, -0.0493, ..., -0.0422, -0.0292, 0.0056], + ..., + [ 0.0455, -0.0440, 0.0005, ..., 0.0229, -0.0243, 0.0340], + [-0.0289, -0.0348, -0.0554, ..., -0.0121, -0.0216, -0.0404], + [-0.0930, -0.1463, 0.0342, ..., -0.0503, 0.0203, 0.0304]], + device='cuda:0'), grad: tensor([[ 4.1258e-07, 0.0000e+00, 2.1420e-07, ..., 0.0000e+00, + 6.1467e-08, 9.3132e-10], + [ 9.4533e-05, 0.0000e+00, 6.4727e-07, ..., 4.6566e-09, + -2.1420e-08, 5.5879e-09], + [ 5.4762e-07, 0.0000e+00, 3.4459e-07, ..., 9.3132e-10, + 1.0896e-07, 4.6566e-09], + ..., + [-7.2527e-04, 0.0000e+00, 6.6161e-05, ..., 2.7008e-08, + 8.5831e-06, -2.6077e-08], + [ 1.7481e-06, 0.0000e+00, 5.9828e-06, ..., 0.0000e+00, + 1.6633e-06, 3.7253e-09], + [ 5.7745e-04, 0.0000e+00, -9.6083e-05, ..., 0.0000e+00, + -1.3061e-05, 8.3819e-09]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0022, 0.0255, 0.0050, -0.0011, 0.0105, -0.0012, 0.0227, -0.0155, + 0.0130, -0.0025], device='cuda:0'), grad: tensor([ 1.6550e-06, 2.4092e-04, 4.4219e-06, 1.1832e-04, 8.6546e-05, + -1.9282e-05, 4.3474e-06, -1.5459e-03, 1.9282e-05, 1.0891e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 217.40, cls_loss 0.0045 cls_loss_mapping 0.0068 cls_loss_causal 0.5977 re_mapping 0.0082 re_causal 0.0249 /// teacc 98.87 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0434, -0.0553, -0.0403, ..., -0.0878, -0.0003, -0.0160], + [-0.0365, 0.0258, -0.1181, ..., 0.0450, 0.0512, -0.0995], + [ 0.0518, 0.0008, -0.0492, ..., -0.0426, -0.0294, 0.0051], + ..., + [ 0.0457, -0.0443, 0.0003, ..., 0.0227, -0.0246, 0.0338], + [-0.0284, -0.0349, -0.0574, ..., -0.0122, -0.0218, -0.0405], + [-0.0943, -0.1489, 0.0365, ..., -0.0505, 0.0213, 0.0316]], + device='cuda:0'), grad: tensor([[ 1.4072e-06, 2.8219e-07, 1.3411e-07, ..., 0.0000e+00, + 3.3267e-06, 7.4506e-09], + [ 1.6105e-04, 3.3975e-05, 1.5926e-07, ..., 0.0000e+00, + -1.9222e-06, 9.3132e-10], + [-2.0921e-04, -4.4346e-05, 3.9209e-07, ..., 0.0000e+00, + 1.2405e-06, -4.0978e-08], + ..., + [ 8.6427e-06, 2.0191e-06, 1.5460e-07, ..., 0.0000e+00, + 4.3306e-07, 1.1176e-08], + [ 4.1910e-06, 7.6555e-07, -2.9746e-06, ..., 0.0000e+00, + 1.5181e-07, 9.3132e-10], + [ 1.1116e-05, 2.2855e-06, 2.7418e-06, ..., 0.0000e+00, + 1.0040e-06, 1.4901e-08]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0031, 0.0255, 0.0054, -0.0015, 0.0102, -0.0016, 0.0219, -0.0154, + 0.0128, -0.0010], device='cuda:0'), grad: tensor([ 9.9242e-06, 2.6059e-04, -3.4046e-04, 2.0459e-05, 4.7684e-06, + 1.5073e-05, -1.1809e-05, 1.6093e-05, -4.4331e-06, 2.9758e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 217.34, cls_loss 0.0038 cls_loss_mapping 0.0056 cls_loss_causal 0.5527 re_mapping 0.0082 re_causal 0.0253 /// teacc 98.82 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0437, -0.0563, -0.0406, ..., -0.0878, -0.0007, -0.0165], + [-0.0375, 0.0238, -0.1184, ..., 0.0451, 0.0515, -0.0998], + [ 0.0517, 0.0022, -0.0495, ..., -0.0427, -0.0294, 0.0050], + ..., + [ 0.0467, -0.0440, 0.0016, ..., 0.0227, -0.0247, 0.0336], + [-0.0289, -0.0351, -0.0576, ..., -0.0122, -0.0222, -0.0405], + [-0.0954, -0.1510, 0.0365, ..., -0.0505, 0.0218, 0.0321]], + device='cuda:0'), grad: tensor([[ 7.7300e-07, 1.0151e-07, 1.3877e-07, ..., 0.0000e+00, + 3.6974e-07, 0.0000e+00], + [ 8.5533e-06, 1.4249e-07, 3.5763e-07, ..., 0.0000e+00, + -3.4198e-06, 0.0000e+00], + [-5.0664e-06, -2.1160e-06, 1.4063e-07, ..., 0.0000e+00, + 8.2608e-07, -5.5879e-09], + ..., + [ 7.8380e-06, 5.4482e-07, 1.8522e-05, ..., 0.0000e+00, + 4.0568e-06, 2.7940e-09], + [ 9.5144e-06, 1.1539e-06, 1.5851e-06, ..., 0.0000e+00, + 3.0361e-07, 9.3132e-10], + [ 7.9036e-05, 8.3819e-09, 1.2070e-04, ..., 0.0000e+00, + -4.2245e-06, 0.0000e+00]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0032, 0.0251, 0.0053, -0.0016, 0.0097, -0.0019, 0.0216, -0.0143, + 0.0131, -0.0011], device='cuda:0'), grad: tensor([ 2.6338e-06, 9.9242e-06, -4.4778e-06, 1.5542e-05, -4.4227e-04, + 3.9674e-06, 9.7975e-07, 7.2122e-05, 2.2560e-05, 3.1805e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 217.21, cls_loss 0.0041 cls_loss_mapping 0.0064 cls_loss_causal 0.5552 re_mapping 0.0082 re_causal 0.0238 /// teacc 98.89 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0466, -0.0579, -0.0408, ..., -0.0879, -0.0012, -0.0166], + [-0.0385, 0.0230, -0.1188, ..., 0.0449, 0.0518, -0.1009], + [ 0.0519, 0.0021, -0.0489, ..., -0.0428, -0.0295, 0.0092], + ..., + [ 0.0473, -0.0432, 0.0017, ..., 0.0227, -0.0250, 0.0335], + [-0.0294, -0.0348, -0.0575, ..., -0.0123, -0.0221, -0.0409], + [-0.0963, -0.1540, 0.0359, ..., -0.0500, 0.0220, 0.0320]], + device='cuda:0'), grad: tensor([[ 1.5646e-07, 2.7940e-09, 4.4424e-07, ..., 0.0000e+00, + 1.2480e-07, 2.4214e-08], + [ 4.7311e-07, 4.0047e-08, 1.7583e-06, ..., 0.0000e+00, + -3.4831e-06, 7.4506e-09], + [-5.1036e-07, 2.8871e-08, 7.6182e-07, ..., 0.0000e+00, + 2.4866e-07, -2.4866e-07], + ..., + [-3.5688e-06, -1.7975e-07, 6.4913e-07, ..., 0.0000e+00, + 1.5320e-06, 1.8068e-07], + [ 4.8243e-07, 2.7940e-09, 4.0233e-07, ..., 0.0000e+00, + 9.6671e-07, 2.1420e-08], + [ 1.5311e-06, 5.2154e-08, 4.7326e-05, ..., 0.0000e+00, + -9.8348e-07, 1.8626e-09]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0034, 0.0246, 0.0053, -0.0010, 0.0100, -0.0024, 0.0197, -0.0138, + 0.0151, -0.0015], device='cuda:0'), grad: tensor([ 2.4825e-05, -1.5842e-06, 8.5384e-06, 4.1090e-06, -8.9109e-05, + 3.8072e-06, -8.0407e-05, 2.0396e-07, 8.9332e-06, 1.2082e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 217.19, cls_loss 0.0032 cls_loss_mapping 0.0047 cls_loss_causal 0.5390 re_mapping 0.0083 re_causal 0.0238 /// teacc 98.80 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0463, -0.0582, -0.0412, ..., -0.0892, -0.0010, -0.0166], + [-0.0383, 0.0232, -0.1192, ..., 0.0449, 0.0529, -0.1012], + [ 0.0523, 0.0023, -0.0494, ..., -0.0434, -0.0297, 0.0095], + ..., + [ 0.0471, -0.0435, 0.0015, ..., 0.0225, -0.0259, 0.0332], + [-0.0303, -0.0349, -0.0581, ..., -0.0126, -0.0239, -0.0412], + [-0.0964, -0.1553, 0.0365, ..., -0.0498, 0.0228, 0.0320]], + device='cuda:0'), grad: tensor([[ 4.6566e-07, 9.3132e-10, 8.1956e-08, ..., 2.7195e-07, + 6.0629e-07, 0.0000e+00], + [-2.2069e-05, 2.4214e-08, 9.7789e-08, ..., -2.5630e-05, + -5.0306e-05, 0.0000e+00], + [ 1.6153e-05, 4.6566e-09, 2.3376e-07, ..., 2.0161e-05, + 3.9101e-05, 0.0000e+00], + ..., + [ 1.3942e-06, -7.1712e-08, -9.7416e-07, ..., 2.8163e-06, + 5.8860e-06, 0.0000e+00], + [ 8.4471e-07, 3.7253e-09, 6.8266e-07, ..., 5.1223e-08, + 4.0885e-07, 0.0000e+00], + [ 7.3668e-07, 1.5832e-08, 2.9337e-07, ..., 1.0990e-07, + -6.2492e-07, 0.0000e+00]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0031, 0.0250, 0.0054, -0.0017, 0.0094, -0.0017, 0.0194, -0.0142, + 0.0142, -0.0007], device='cuda:0'), grad: tensor([-4.5240e-05, -2.7251e-04, 2.1565e-04, 4.6715e-06, 2.0668e-05, + 2.1495e-06, 4.8101e-05, 1.2137e-05, 4.9211e-06, 9.2387e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 217.25, cls_loss 0.0039 cls_loss_mapping 0.0056 cls_loss_causal 0.5650 re_mapping 0.0080 re_causal 0.0237 /// teacc 98.77 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0463, -0.0613, -0.0417, ..., -0.0893, -0.0013, -0.0187], + [-0.0384, 0.0235, -0.1194, ..., 0.0455, 0.0533, -0.1029], + [ 0.0525, 0.0023, -0.0488, ..., -0.0442, -0.0299, 0.0089], + ..., + [ 0.0473, -0.0432, 0.0013, ..., 0.0221, -0.0259, 0.0329], + [-0.0310, -0.0349, -0.0587, ..., -0.0128, -0.0245, -0.0417], + [-0.0979, -0.1640, 0.0360, ..., -0.0499, 0.0225, 0.0341]], + device='cuda:0'), grad: tensor([[ 3.4198e-06, 1.8626e-09, 1.8626e-07, ..., 7.4506e-09, + 6.5193e-08, 3.3434e-06], + [ 1.7323e-07, 1.8626e-09, 1.8394e-06, ..., 3.1665e-07, + -1.7062e-06, 4.3772e-08], + [-5.4911e-06, -4.1910e-08, 4.7218e-07, ..., 4.2841e-08, + 1.4994e-07, -5.1446e-06], + ..., + [ 7.6462e-07, 3.5390e-08, 9.9931e-07, ..., 6.5193e-08, + 7.1619e-07, 6.5845e-07], + [ 2.3376e-07, -2.7940e-09, 9.4250e-07, ..., 4.6566e-09, + 7.9256e-07, 1.2200e-07], + [ 4.0885e-07, 9.3132e-10, 1.3942e-06, ..., 5.5879e-08, + -5.8766e-07, 2.5798e-07]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0042, 0.0252, 0.0058, -0.0018, 0.0096, -0.0012, 0.0194, -0.0140, + 0.0137, -0.0006], device='cuda:0'), grad: tensor([-3.5260e-06, 2.3991e-06, -1.3016e-05, 1.7034e-06, -1.1824e-05, + 1.2685e-06, 8.4788e-06, 6.2846e-06, 3.9376e-06, 4.2804e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 217.18, cls_loss 0.0043 cls_loss_mapping 0.0053 cls_loss_causal 0.5260 re_mapping 0.0081 re_causal 0.0236 /// teacc 98.89 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0463, -0.0633, -0.0420, ..., -0.0893, -0.0027, -0.0206], + [-0.0399, 0.0244, -0.1198, ..., 0.0464, 0.0537, -0.1077], + [ 0.0537, 0.0029, -0.0491, ..., -0.0443, -0.0300, 0.0102], + ..., + [ 0.0469, -0.0448, 0.0012, ..., 0.0213, -0.0263, 0.0317], + [-0.0315, -0.0351, -0.0591, ..., -0.0130, -0.0244, -0.0432], + [-0.0995, -0.1688, 0.0354, ..., -0.0501, 0.0233, 0.0359]], + device='cuda:0'), grad: tensor([[ 2.6636e-07, 1.6764e-08, 3.1665e-08, ..., 0.0000e+00, + -1.0364e-05, 0.0000e+00], + [ 9.0618e-07, 3.4273e-07, 8.3819e-08, ..., 0.0000e+00, + -3.4496e-06, 0.0000e+00], + [-2.2911e-06, 3.1944e-07, 5.3085e-08, ..., 9.3132e-10, + 1.6820e-06, -9.3132e-10], + ..., + [ 9.4995e-06, -1.3513e-06, 3.7216e-06, ..., 9.3132e-10, + 4.5169e-07, 9.3132e-10], + [ 1.9260e-06, 6.8918e-08, 1.1735e-07, ..., 9.3132e-10, + 3.4153e-05, 0.0000e+00], + [-5.7369e-05, 1.4901e-08, -1.3188e-05, ..., 0.0000e+00, + 1.6633e-06, 0.0000e+00]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0055, 0.0245, 0.0065, -0.0011, 0.0104, -0.0005, 0.0197, -0.0143, + 0.0131, -0.0003], device='cuda:0'), grad: tensor([-4.0740e-05, -4.6194e-06, 3.1628e-06, 8.1360e-06, 3.3259e-04, + 5.8270e-04, -6.5899e-04, 1.1539e-04, 1.3030e-04, -4.6730e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 217.67, cls_loss 0.0039 cls_loss_mapping 0.0075 cls_loss_causal 0.5445 re_mapping 0.0081 re_causal 0.0231 /// teacc 98.92 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0446, -0.0650, -0.0424, ..., -0.0896, -0.0028, -0.0208], + [-0.0403, 0.0243, -0.1201, ..., 0.0471, 0.0548, -0.1087], + [ 0.0539, 0.0028, -0.0515, ..., -0.0444, -0.0302, 0.0104], + ..., + [ 0.0469, -0.0449, 0.0010, ..., 0.0210, -0.0274, 0.0314], + [-0.0317, -0.0344, -0.0599, ..., -0.0131, -0.0246, -0.0438], + [-0.1025, -0.1704, 0.0346, ..., -0.0506, 0.0234, 0.0363]], + device='cuda:0'), grad: tensor([[ 1.3322e-05, 0.0000e+00, 5.4911e-06, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + [ 5.3458e-06, 0.0000e+00, 2.2128e-06, ..., 0.0000e+00, + -1.4529e-07, 9.3132e-10], + [ 1.5602e-05, 0.0000e+00, 6.6981e-06, ..., 0.0000e+00, + 5.5879e-08, 4.6566e-09], + ..., + [ 4.7497e-08, 0.0000e+00, 1.9744e-07, ..., 0.0000e+00, + 4.2841e-08, -7.4506e-09], + [ 1.3858e-06, 0.0000e+00, 8.2701e-07, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [-1.6898e-05, 0.0000e+00, -6.8434e-06, ..., 0.0000e+00, + -2.0489e-08, 9.3132e-10]], device='cuda:0') +Epoch 136, bias, value: tensor([-4.9428e-03, 2.4967e-02, 6.1505e-03, -4.1212e-04, 1.0852e-02, + 9.9237e-05, 1.9116e-02, -1.4833e-02, 1.2888e-02, -9.0948e-04], + device='cuda:0'), grad: tensor([ 4.9740e-05, 1.3009e-05, 4.1425e-05, 1.8209e-05, 3.4142e-04, + 1.0274e-05, -4.0984e-04, 1.1306e-06, 3.4682e-06, -6.9201e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 217.64, cls_loss 0.0030 cls_loss_mapping 0.0049 cls_loss_causal 0.5488 re_mapping 0.0080 re_causal 0.0247 /// teacc 98.76 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0448, -0.0658, -0.0425, ..., -0.0896, -0.0030, -0.0208], + [-0.0404, 0.0244, -0.1203, ..., 0.0471, 0.0550, -0.1090], + [ 0.0538, 0.0026, -0.0521, ..., -0.0444, -0.0305, 0.0103], + ..., + [ 0.0471, -0.0449, 0.0011, ..., 0.0211, -0.0276, 0.0316], + [-0.0315, -0.0326, -0.0603, ..., -0.0132, -0.0247, -0.0440], + [-0.1029, -0.1710, 0.0333, ..., -0.0515, 0.0235, 0.0363]], + device='cuda:0'), grad: tensor([[ 2.5332e-07, 0.0000e+00, 4.4703e-08, ..., 0.0000e+00, + 2.0396e-07, 1.5739e-07], + [ 1.1683e-05, 0.0000e+00, 1.1083e-07, ..., 0.0000e+00, + -2.6245e-06, 1.6810e-06], + [ 2.2486e-05, 0.0000e+00, -1.3867e-06, ..., 0.0000e+00, + 1.5507e-06, 3.6452e-06], + ..., + [-5.7071e-05, 0.0000e+00, 4.0978e-08, ..., 0.0000e+00, + 6.8173e-07, -8.5905e-06], + [ 3.2987e-06, 0.0000e+00, 1.6764e-07, ..., 0.0000e+00, + 4.4145e-07, 3.7067e-07], + [ 1.4596e-05, 0.0000e+00, 7.7300e-08, ..., 0.0000e+00, + -2.3954e-06, 2.1141e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([-4.9095e-03, 2.4978e-02, 5.8470e-03, 8.5718e-05, 1.1930e-02, + -4.4005e-04, 1.9489e-02, -1.4733e-02, 1.3005e-02, -1.7870e-03], + device='cuda:0'), grad: tensor([-1.5870e-05, 3.3855e-05, 9.0301e-05, 1.9558e-06, 1.3597e-05, + 7.1041e-06, 3.5763e-06, -2.0278e-04, 1.1377e-05, 5.6952e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 217.36, cls_loss 0.0033 cls_loss_mapping 0.0059 cls_loss_causal 0.5427 re_mapping 0.0076 re_causal 0.0228 /// teacc 98.84 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0449, -0.0669, -0.0431, ..., -0.0898, -0.0024, -0.0211], + [-0.0418, 0.0247, -0.1206, ..., 0.0471, 0.0555, -0.1102], + [ 0.0541, 0.0025, -0.0514, ..., -0.0444, -0.0306, 0.0113], + ..., + [ 0.0481, -0.0450, 0.0007, ..., 0.0214, -0.0281, 0.0316], + [-0.0318, -0.0334, -0.0608, ..., -0.0134, -0.0248, -0.0449], + [-0.1037, -0.1713, 0.0326, ..., -0.0523, 0.0237, 0.0365]], + device='cuda:0'), grad: tensor([[-1.4043e-04, 9.3132e-10, 6.2212e-07, ..., 1.8626e-09, + -2.4009e-06, 9.8720e-08], + [ 3.7961e-06, -4.7497e-08, 3.4086e-07, ..., 3.4459e-08, + -3.8370e-06, 5.4948e-08], + [-1.2718e-05, 8.3819e-09, -4.4443e-06, ..., 3.7253e-09, + -1.1260e-06, -1.3718e-06], + ..., + [ 4.2692e-06, 1.0245e-08, 4.8708e-07, ..., 1.2107e-08, + 6.2026e-07, 5.4017e-08], + [ 6.9402e-06, 1.3970e-08, 2.5183e-06, ..., 1.8626e-09, + 2.3618e-06, 3.9116e-08], + [ 1.2279e-04, 9.3132e-10, 1.3933e-06, ..., 8.3819e-09, + 2.7195e-06, 7.5437e-08]], device='cuda:0') +Epoch 138, bias, value: tensor([-4.3489e-03, 2.3705e-02, 5.6880e-03, -1.3175e-04, 1.2197e-02, + -5.3660e-05, 1.9176e-02, -1.3496e-02, 1.3220e-02, -2.2984e-03], + device='cuda:0'), grad: tensor([-7.7009e-04, 1.1409e-06, -1.7881e-05, 5.5611e-05, 6.0759e-06, + 5.9903e-06, 9.7975e-06, 7.3388e-06, 3.2365e-05, 6.6900e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 217.46, cls_loss 0.0042 cls_loss_mapping 0.0060 cls_loss_causal 0.5544 re_mapping 0.0078 re_causal 0.0232 /// teacc 98.87 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0448, -0.0679, -0.0437, ..., -0.0912, -0.0026, -0.0211], + [-0.0423, 0.0250, -0.1206, ..., 0.0474, 0.0550, -0.1124], + [ 0.0543, 0.0024, -0.0517, ..., -0.0447, -0.0295, 0.0122], + ..., + [ 0.0486, -0.0451, 0.0005, ..., 0.0205, -0.0287, 0.0302], + [-0.0324, -0.0338, -0.0610, ..., -0.0138, -0.0253, -0.0470], + [-0.1049, -0.1715, 0.0322, ..., -0.0523, 0.0242, 0.0366]], + device='cuda:0'), grad: tensor([[ 2.2072e-07, 0.0000e+00, 1.9558e-08, ..., 0.0000e+00, + 2.0899e-06, 1.8626e-08], + [ 5.0757e-07, 5.5879e-09, 2.4214e-08, ..., 0.0000e+00, + 5.5879e-08, 1.3970e-08], + [ 7.1526e-06, 9.3132e-10, 4.1910e-08, ..., 0.0000e+00, + 1.5870e-06, 3.7253e-09], + ..., + [ 1.2174e-05, -9.3132e-09, 1.0710e-07, ..., 0.0000e+00, + 1.0626e-06, 3.9116e-08], + [ 6.2585e-07, 9.3132e-10, 2.1141e-07, ..., 0.0000e+00, + 2.4289e-06, 1.2014e-07], + [ 6.1095e-07, 9.3132e-10, -6.9197e-07, ..., 0.0000e+00, + -6.6385e-06, -3.9302e-07]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0043, 0.0236, 0.0059, 0.0019, 0.0121, -0.0017, 0.0196, -0.0133, + 0.0130, -0.0027], device='cuda:0'), grad: tensor([ 4.6156e-06, 1.5106e-06, 1.3456e-05, -2.1562e-05, 2.1249e-05, + 3.4962e-06, -1.4141e-05, 2.0027e-05, 7.1302e-06, -3.5793e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 217.48, cls_loss 0.0047 cls_loss_mapping 0.0060 cls_loss_causal 0.5589 re_mapping 0.0077 re_causal 0.0233 /// teacc 98.84 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0451, -0.0694, -0.0438, ..., -0.0915, -0.0029, -0.0212], + [-0.0424, 0.0250, -0.1212, ..., 0.0492, 0.0561, -0.1182], + [ 0.0536, 0.0024, -0.0518, ..., -0.0449, -0.0290, 0.0139], + ..., + [ 0.0495, -0.0451, 0.0025, ..., 0.0231, -0.0302, 0.0292], + [-0.0327, -0.0336, -0.0619, ..., -0.0139, -0.0265, -0.0526], + [-0.1056, -0.1719, 0.0334, ..., -0.0525, 0.0227, 0.0365]], + device='cuda:0'), grad: tensor([[-2.9244e-07, 9.3132e-10, 3.9581e-07, ..., 0.0000e+00, + 5.7742e-08, 2.1420e-08], + [ 1.7285e-06, 3.7253e-09, 2.3823e-06, ..., 1.7788e-07, + -7.9162e-08, 1.3132e-07], + [-9.8813e-07, 7.4506e-09, 6.0350e-07, ..., 9.3132e-10, + 1.6298e-07, 3.4459e-08], + ..., + [-2.2240e-06, -1.9558e-08, 7.5027e-06, ..., -2.0675e-07, + 4.1723e-07, 4.1444e-07], + [-7.4413e-07, 9.3132e-10, 6.8359e-07, ..., 9.3132e-10, + -3.6769e-06, 2.7008e-08], + [ 7.7859e-07, 9.3132e-10, 3.5837e-06, ..., 2.1420e-08, + -2.4047e-06, 2.9150e-07]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0044, 0.0243, 0.0056, 0.0009, 0.0124, -0.0006, 0.0198, -0.0127, + 0.0120, -0.0035], device='cuda:0'), grad: tensor([ 1.6764e-07, 8.1733e-06, 1.2890e-06, 1.4327e-05, -2.9519e-05, + 8.2105e-06, 1.5181e-06, 1.1414e-05, -2.0623e-05, 5.0142e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 139---------------------------------------------------- +epoch 139, time 218.25, cls_loss 0.0035 cls_loss_mapping 0.0048 cls_loss_causal 0.5540 re_mapping 0.0076 re_causal 0.0230 /// teacc 98.93 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0454, -0.0696, -0.0441, ..., -0.0916, -0.0033, -0.0213], + [-0.0428, 0.0251, -0.1215, ..., 0.0494, 0.0568, -0.1189], + [ 0.0539, 0.0017, -0.0528, ..., -0.0449, -0.0294, 0.0139], + ..., + [ 0.0491, -0.0452, 0.0023, ..., 0.0229, -0.0309, 0.0282], + [-0.0330, -0.0320, -0.0622, ..., -0.0140, -0.0268, -0.0538], + [-0.1058, -0.1722, 0.0339, ..., -0.0523, 0.0233, 0.0367]], + device='cuda:0'), grad: tensor([[-2.4121e-06, 1.8626e-09, 3.4180e-07, ..., 0.0000e+00, + 2.7101e-07, 4.1910e-08], + [ 3.8277e-07, 2.7940e-09, 1.4901e-07, ..., 0.0000e+00, + -5.5879e-09, 1.0245e-08], + [-3.5297e-07, -2.9802e-08, 2.8536e-06, ..., 0.0000e+00, + 5.9325e-07, -4.6566e-09], + ..., + [-5.1036e-07, 9.3132e-09, 1.4435e-07, ..., 0.0000e+00, + 3.7625e-07, 7.4506e-09], + [ 3.1851e-07, 4.6566e-09, -8.1584e-07, ..., 0.0000e+00, + 2.1271e-06, 3.3900e-07], + [ 1.5451e-06, 9.3132e-10, 2.3656e-07, ..., 0.0000e+00, + -1.5395e-06, 1.8626e-09]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0046, 0.0242, 0.0054, 0.0019, 0.0122, -0.0005, 0.0196, -0.0133, + 0.0120, -0.0029], device='cuda:0'), grad: tensor([-5.8636e-06, 1.3923e-06, 1.5125e-05, 3.0808e-06, 3.4496e-06, + 7.8797e-05, -1.0431e-04, 1.5749e-06, 8.1062e-06, -1.5376e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 217.41, cls_loss 0.0032 cls_loss_mapping 0.0053 cls_loss_causal 0.5313 re_mapping 0.0077 re_causal 0.0235 /// teacc 98.85 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0447, -0.0701, -0.0445, ..., -0.0917, -0.0038, -0.0215], + [-0.0425, 0.0258, -0.1222, ..., 0.0500, 0.0573, -0.1193], + [ 0.0538, 0.0013, -0.0520, ..., -0.0469, -0.0297, 0.0162], + ..., + [ 0.0490, -0.0456, 0.0023, ..., 0.0228, -0.0317, 0.0284], + [-0.0329, -0.0321, -0.0626, ..., -0.0140, -0.0270, -0.0592], + [-0.1066, -0.1723, 0.0340, ..., -0.0524, 0.0241, 0.0365]], + device='cuda:0'), grad: tensor([[-2.8033e-07, 0.0000e+00, 8.9034e-07, ..., 0.0000e+00, + 2.0675e-07, 6.5193e-08], + [ 1.0421e-06, 0.0000e+00, 1.2983e-06, ..., 0.0000e+00, + -1.0375e-06, 9.2201e-08], + [ 4.2692e-06, 0.0000e+00, 4.3064e-06, ..., 0.0000e+00, + 1.2480e-07, 3.8464e-07], + ..., + [-1.5497e-06, 0.0000e+00, 2.4289e-06, ..., 0.0000e+00, + 4.1258e-07, -8.6706e-07], + [ 1.0962e-06, 0.0000e+00, 1.9185e-07, ..., 0.0000e+00, + 5.1223e-08, 2.5984e-07], + [ 7.5717e-07, 0.0000e+00, 6.8545e-06, ..., 0.0000e+00, + 4.9733e-07, 6.8918e-08]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0046, 0.0246, 0.0050, 0.0019, 0.0117, -0.0004, 0.0194, -0.0136, + 0.0123, -0.0024], device='cuda:0'), grad: tensor([-4.0792e-06, 1.3718e-06, 1.4804e-05, 2.8163e-06, -3.7611e-05, + -7.2550e-07, 5.1558e-06, 1.1157e-06, -5.2247e-07, 1.7643e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 217.42, cls_loss 0.0035 cls_loss_mapping 0.0060 cls_loss_causal 0.5447 re_mapping 0.0077 re_causal 0.0229 /// teacc 98.83 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0448, -0.0706, -0.0453, ..., -0.0917, -0.0041, -0.0215], + [-0.0414, 0.0260, -0.1232, ..., 0.0500, 0.0595, -0.1196], + [ 0.0543, 0.0015, -0.0515, ..., -0.0469, -0.0298, 0.0175], + ..., + [ 0.0485, -0.0459, 0.0021, ..., 0.0228, -0.0325, 0.0282], + [-0.0333, -0.0323, -0.0634, ..., -0.0140, -0.0274, -0.0595], + [-0.1097, -0.1725, 0.0332, ..., -0.0524, 0.0228, 0.0364]], + device='cuda:0'), grad: tensor([[ 1.1083e-06, 4.6566e-09, 2.0768e-07, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + [ 7.3314e-06, 2.0768e-07, 1.2480e-07, ..., 0.0000e+00, + 3.3807e-07, 0.0000e+00], + [-3.9369e-05, 4.2841e-08, -5.3197e-06, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + ..., + [-9.8944e-06, -5.2340e-07, 1.1297e-06, ..., 0.0000e+00, + -1.5935e-06, 0.0000e+00], + [ 2.1383e-05, 3.7253e-09, 1.2234e-05, ..., 0.0000e+00, + 7.1116e-06, 0.0000e+00], + [ 6.4820e-06, 1.1548e-07, 2.1048e-07, ..., 0.0000e+00, + 6.9663e-07, 0.0000e+00]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0046, 0.0259, 0.0053, 0.0023, 0.0116, -0.0006, 0.0198, -0.0144, + 0.0118, -0.0031], device='cuda:0'), grad: tensor([ 1.6829e-06, 1.2398e-05, -7.4327e-05, 1.2413e-05, 8.0094e-06, + 1.1623e-04, -1.9383e-04, -1.9014e-05, 1.2445e-04, 1.1809e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 217.58, cls_loss 0.0028 cls_loss_mapping 0.0043 cls_loss_causal 0.5187 re_mapping 0.0077 re_causal 0.0228 /// teacc 98.89 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0450, -0.0708, -0.0456, ..., -0.0917, -0.0038, -0.0215], + [-0.0414, 0.0262, -0.1243, ..., 0.0500, 0.0585, -0.1197], + [ 0.0548, 0.0012, -0.0511, ..., -0.0470, -0.0299, 0.0174], + ..., + [ 0.0485, -0.0460, 0.0019, ..., 0.0228, -0.0327, 0.0282], + [-0.0336, -0.0314, -0.0640, ..., -0.0140, -0.0252, -0.0596], + [-0.1108, -0.1727, 0.0348, ..., -0.0524, 0.0231, 0.0364]], + device='cuda:0'), grad: tensor([[-6.2678e-07, 1.8626e-09, 2.8498e-07, ..., 2.3283e-08, + 1.2573e-07, 0.0000e+00], + [ 6.0070e-07, -8.2888e-07, 1.0720e-06, ..., 8.8476e-08, + -6.0052e-06, 0.0000e+00], + [ 6.2361e-06, 5.9884e-07, 5.0385e-07, ..., 3.3528e-08, + 4.2655e-06, 0.0000e+00], + ..., + [ 1.3024e-05, 5.5879e-08, -4.5568e-05, ..., -1.4752e-05, + 4.6939e-06, 0.0000e+00], + [ 2.4572e-05, 1.3877e-07, -1.7630e-06, ..., 2.9802e-08, + 1.2793e-05, 0.0000e+00], + [ 1.3085e-06, 1.1176e-08, 3.5018e-06, ..., 1.3504e-07, + -2.5518e-07, 0.0000e+00]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0042, 0.0249, 0.0056, 0.0021, 0.0100, -0.0005, 0.0199, -0.0144, + 0.0132, -0.0019], device='cuda:0'), grad: tensor([-4.2543e-06, -1.4521e-05, 3.1054e-05, -7.2896e-05, 2.6584e-04, + 1.2696e-05, -7.4983e-05, -2.7537e-04, 1.1462e-04, 1.8358e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 217.66, cls_loss 0.0027 cls_loss_mapping 0.0044 cls_loss_causal 0.5373 re_mapping 0.0075 re_causal 0.0226 /// teacc 98.84 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0444, -0.0713, -0.0461, ..., -0.0917, -0.0041, -0.0215], + [-0.0419, 0.0264, -0.1247, ..., 0.0500, 0.0575, -0.1198], + [ 0.0551, 0.0011, -0.0513, ..., -0.0470, -0.0301, 0.0174], + ..., + [ 0.0489, -0.0460, 0.0010, ..., 0.0229, -0.0309, 0.0276], + [-0.0344, -0.0313, -0.0642, ..., -0.0140, -0.0254, -0.0596], + [-0.1114, -0.1728, 0.0348, ..., -0.0524, 0.0231, 0.0364]], + device='cuda:0'), grad: tensor([[-4.3306e-07, 2.2352e-08, 1.2293e-07, ..., 0.0000e+00, + 1.9372e-07, 0.0000e+00], + [ 2.6915e-07, -2.9523e-06, 3.0082e-07, ..., 0.0000e+00, + -5.8301e-06, 0.0000e+00], + [ 3.0547e-06, 2.6356e-07, 1.5274e-07, ..., 0.0000e+00, + 2.2277e-06, 0.0000e+00], + ..., + [-1.4156e-05, 8.3819e-08, 7.1712e-08, ..., 0.0000e+00, + -3.4869e-06, 0.0000e+00], + [ 2.6524e-06, 2.1607e-06, 1.0151e-07, ..., 0.0000e+00, + 5.1558e-06, 9.3132e-10], + [ 7.3574e-06, 1.8626e-08, 2.2724e-06, ..., 0.0000e+00, + 9.8627e-07, 9.3132e-10]], device='cuda:0') +Epoch 145, bias, value: tensor([-4.0226e-03, 2.4210e-02, 5.7599e-03, 1.7476e-03, 1.0116e-02, + 1.7832e-05, 1.9924e-02, -1.3649e-02, 1.2958e-02, -2.1085e-03], + device='cuda:0'), grad: tensor([-3.8035e-06, -1.1526e-05, 1.4603e-05, 5.5581e-06, -5.6103e-06, + -2.2613e-06, -2.2873e-06, -4.0263e-05, 1.9819e-05, 2.5690e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 217.57, cls_loss 0.0024 cls_loss_mapping 0.0042 cls_loss_causal 0.5331 re_mapping 0.0073 re_causal 0.0216 /// teacc 98.89 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0445, -0.0720, -0.0463, ..., -0.0917, -0.0044, -0.0215], + [-0.0419, 0.0265, -0.1249, ..., 0.0500, 0.0578, -0.1200], + [ 0.0554, 0.0008, -0.0515, ..., -0.0470, -0.0303, 0.0174], + ..., + [ 0.0488, -0.0460, 0.0014, ..., 0.0229, -0.0311, 0.0273], + [-0.0350, -0.0306, -0.0640, ..., -0.0140, -0.0262, -0.0597], + [-0.1121, -0.1734, 0.0342, ..., -0.0524, 0.0235, 0.0364]], + device='cuda:0'), grad: tensor([[ 2.8126e-07, 0.0000e+00, 1.6671e-07, ..., 0.0000e+00, + 1.1176e-07, 0.0000e+00], + [ 9.6112e-07, 0.0000e+00, 7.8324e-07, ..., 0.0000e+00, + -5.8673e-08, 0.0000e+00], + [-2.6673e-06, 0.0000e+00, 1.0710e-07, ..., 0.0000e+00, + 2.9430e-07, 0.0000e+00], + ..., + [ 9.1866e-06, 0.0000e+00, 1.7554e-05, ..., 0.0000e+00, + 4.7535e-06, 0.0000e+00], + [ 2.7195e-06, 0.0000e+00, 4.5449e-07, ..., 0.0000e+00, + 8.2236e-07, 0.0000e+00], + [ 1.2228e-06, 0.0000e+00, 4.8369e-05, ..., 0.0000e+00, + 7.8231e-06, 0.0000e+00]], device='cuda:0') +Epoch 146, bias, value: tensor([-3.6336e-03, 2.4358e-02, 5.8552e-03, 1.8479e-03, 1.0120e-02, + 1.4250e-05, 2.0323e-02, -1.3637e-02, 1.2483e-02, -2.4146e-03], + device='cuda:0'), grad: tensor([-2.7902e-06, 3.3528e-06, -4.1872e-06, 3.2540e-06, -1.9431e-04, + -1.9789e-05, -1.9651e-06, 6.4611e-05, 8.3670e-06, 1.4329e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 217.47, cls_loss 0.0028 cls_loss_mapping 0.0051 cls_loss_causal 0.5687 re_mapping 0.0077 re_causal 0.0228 /// teacc 98.85 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0445, -0.0724, -0.0466, ..., -0.0919, -0.0047, -0.0215], + [-0.0420, 0.0266, -0.1255, ..., 0.0500, 0.0585, -0.1201], + [ 0.0554, 0.0007, -0.0519, ..., -0.0470, -0.0305, 0.0174], + ..., + [ 0.0490, -0.0461, 0.0010, ..., 0.0228, -0.0315, 0.0276], + [-0.0355, -0.0295, -0.0640, ..., -0.0141, -0.0263, -0.0598], + [-0.1123, -0.1737, 0.0340, ..., -0.0524, 0.0241, 0.0364]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, 0.0000e+00, 5.7649e-07, ..., 0.0000e+00, + 7.3947e-07, 2.0489e-07], + [ 1.0477e-06, -2.8871e-08, 5.6811e-08, ..., 0.0000e+00, + 1.5367e-06, 4.5728e-07], + [-3.5316e-06, 1.3970e-08, 3.2503e-07, ..., 0.0000e+00, + -2.1290e-06, 3.4459e-08], + ..., + [-7.1339e-07, 5.5879e-09, 4.7497e-08, ..., 0.0000e+00, + -9.7826e-06, -3.1497e-06], + [ 2.7083e-06, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 2.0787e-06, 3.5390e-08], + [ 2.0675e-07, 0.0000e+00, 1.0738e-06, ..., 0.0000e+00, + 6.0797e-06, 2.2184e-06]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0035, 0.0246, 0.0056, 0.0007, 0.0101, 0.0006, 0.0204, -0.0136, + 0.0126, -0.0024], device='cuda:0'), grad: tensor([ 4.9248e-06, 9.6038e-06, -8.0019e-06, 7.2643e-07, -3.5390e-08, + 7.0706e-06, -4.2841e-06, -5.3465e-05, 8.5384e-06, 3.4958e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 217.52, cls_loss 0.0028 cls_loss_mapping 0.0055 cls_loss_causal 0.5228 re_mapping 0.0077 re_causal 0.0230 /// teacc 98.90 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0446, -0.0726, -0.0467, ..., -0.0919, -0.0049, -0.0215], + [-0.0424, 0.0266, -0.1268, ..., 0.0500, 0.0589, -0.1203], + [ 0.0559, 0.0009, -0.0518, ..., -0.0470, -0.0298, 0.0173], + ..., + [ 0.0491, -0.0462, 0.0009, ..., 0.0228, -0.0323, 0.0277], + [-0.0360, -0.0287, -0.0648, ..., -0.0141, -0.0271, -0.0599], + [-0.1123, -0.1738, 0.0340, ..., -0.0524, 0.0246, 0.0364]], + device='cuda:0'), grad: tensor([[ 6.2399e-08, 0.0000e+00, -9.7096e-05, ..., 0.0000e+00, + 1.6484e-07, 3.7253e-09], + [-7.2271e-07, 0.0000e+00, 1.0459e-06, ..., 0.0000e+00, + -1.3262e-06, 9.3132e-10], + [ 1.0934e-06, 0.0000e+00, 5.0180e-06, ..., 0.0000e+00, + 1.6801e-06, 9.3132e-10], + ..., + [-1.6009e-06, 0.0000e+00, 3.1572e-07, ..., 0.0000e+00, + -9.3132e-08, 0.0000e+00], + [ 1.2759e-07, 0.0000e+00, -4.2439e-05, ..., 0.0000e+00, + -2.7016e-05, 2.7940e-09], + [ 7.0874e-07, 0.0000e+00, 8.7798e-05, ..., 0.0000e+00, + 2.2233e-05, 0.0000e+00]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0034, 0.0245, 0.0062, 0.0007, 0.0101, 0.0008, 0.0203, -0.0139, + 0.0123, -0.0022], device='cuda:0'), grad: tensor([-5.8651e-04, -1.1176e-08, 3.8773e-05, 1.9059e-05, 5.6326e-06, + 3.4153e-05, 2.1720e-04, -1.6810e-06, -2.0349e-04, 4.7636e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 147---------------------------------------------------- +epoch 147, time 218.16, cls_loss 0.0030 cls_loss_mapping 0.0047 cls_loss_causal 0.5137 re_mapping 0.0077 re_causal 0.0221 /// teacc 98.94 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0451, -0.0729, -0.0466, ..., -0.0920, -0.0053, -0.0216], + [-0.0416, 0.0270, -0.1280, ..., 0.0502, 0.0595, -0.1204], + [ 0.0555, 0.0004, -0.0523, ..., -0.0478, -0.0306, 0.0174], + ..., + [ 0.0492, -0.0463, 0.0007, ..., 0.0228, -0.0320, 0.0277], + [-0.0368, -0.0287, -0.0642, ..., -0.0141, -0.0265, -0.0599], + [-0.1132, -0.1739, 0.0333, ..., -0.0525, 0.0244, 0.0364]], + device='cuda:0'), grad: tensor([[ 4.1444e-08, 0.0000e+00, 1.2247e-07, ..., 0.0000e+00, + 4.8950e-06, 2.1607e-07], + [-6.0573e-06, 0.0000e+00, 1.4761e-07, ..., 0.0000e+00, + -7.9051e-06, 1.8626e-09], + [ 1.8906e-07, 0.0000e+00, 3.5390e-08, ..., 0.0000e+00, + 2.4866e-07, 0.0000e+00], + ..., + [ 5.6587e-06, 0.0000e+00, 2.9802e-07, ..., 0.0000e+00, + 7.3090e-06, 4.6566e-10], + [ 2.9337e-07, 0.0000e+00, 5.3644e-07, ..., 0.0000e+00, + 1.0747e-06, 3.2596e-08], + [ 1.3970e-07, 0.0000e+00, 1.0170e-06, ..., 0.0000e+00, + -2.0061e-06, 4.6566e-10]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0035, 0.0253, 0.0051, -0.0010, 0.0101, 0.0020, 0.0202, -0.0137, + 0.0128, -0.0027], device='cuda:0'), grad: tensor([ 2.5809e-05, -2.0534e-05, 8.4052e-07, 1.2070e-06, 2.6580e-06, + 3.0659e-06, -3.2723e-05, 2.0325e-05, 5.8934e-06, -6.6124e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 217.82, cls_loss 0.0030 cls_loss_mapping 0.0062 cls_loss_causal 0.5130 re_mapping 0.0080 re_causal 0.0223 /// teacc 98.84 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0456, -0.0736, -0.0468, ..., -0.0921, -0.0058, -0.0216], + [-0.0415, 0.0272, -0.1290, ..., 0.0505, 0.0600, -0.1207], + [ 0.0559, 0.0014, -0.0526, ..., -0.0492, -0.0309, 0.0180], + ..., + [ 0.0490, -0.0470, 0.0006, ..., 0.0228, -0.0323, 0.0276], + [-0.0379, -0.0287, -0.0651, ..., -0.0141, -0.0267, -0.0600], + [-0.1134, -0.1741, 0.0333, ..., -0.0525, 0.0251, 0.0365]], + device='cuda:0'), grad: tensor([[ 9.3598e-08, 0.0000e+00, 2.8405e-08, ..., 0.0000e+00, + 1.2387e-07, 4.6566e-10], + [ 1.7229e-08, 0.0000e+00, 6.5658e-08, ..., 0.0000e+00, + -2.5313e-06, 4.6566e-10], + [ 1.1213e-05, 0.0000e+00, 1.3392e-06, ..., 0.0000e+00, + 2.0443e-07, -2.3283e-09], + ..., + [-1.2457e-05, 0.0000e+00, -1.4557e-06, ..., 0.0000e+00, + 1.4324e-06, 4.6566e-10], + [ 7.2131e-07, 0.0000e+00, 1.0477e-07, ..., 0.0000e+00, + 2.4401e-07, 9.3132e-10], + [ 2.0862e-07, 0.0000e+00, 3.4040e-07, ..., 0.0000e+00, + 8.8476e-08, 0.0000e+00]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0036, 0.0254, 0.0049, -0.0022, 0.0102, 0.0039, 0.0198, -0.0137, + 0.0123, -0.0026], device='cuda:0'), grad: tensor([ 5.7230e-07, -7.2159e-06, 1.9610e-05, -1.2107e-07, 7.1526e-07, + 9.0711e-07, -3.8138e-07, -1.6734e-05, 1.9874e-06, 6.5379e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.74, cls_loss 0.0027 cls_loss_mapping 0.0036 cls_loss_causal 0.5477 re_mapping 0.0077 re_causal 0.0229 /// teacc 98.92 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0455, -0.0740, -0.0453, ..., -0.0921, -0.0061, -0.0216], + [-0.0412, 0.0272, -0.1294, ..., 0.0505, 0.0605, -0.1209], + [ 0.0557, 0.0012, -0.0527, ..., -0.0492, -0.0311, 0.0180], + ..., + [ 0.0488, -0.0470, 0.0019, ..., 0.0228, -0.0343, 0.0275], + [-0.0375, -0.0283, -0.0651, ..., -0.0141, -0.0265, -0.0600], + [-0.1138, -0.1742, 0.0332, ..., -0.0525, 0.0268, 0.0365]], + device='cuda:0'), grad: tensor([[ 1.5320e-07, 0.0000e+00, -1.2154e-07, ..., 0.0000e+00, + 8.1956e-08, 0.0000e+00], + [ 6.7987e-08, 0.0000e+00, 1.1390e-06, ..., 0.0000e+00, + 8.0233e-07, 0.0000e+00], + [-1.0170e-06, 0.0000e+00, -4.3772e-08, ..., 0.0000e+00, + 1.0058e-07, -9.3132e-10], + ..., + [ 4.6147e-07, 0.0000e+00, 3.6974e-07, ..., 0.0000e+00, + 4.5588e-07, 0.0000e+00], + [ 1.5134e-07, 0.0000e+00, 5.9884e-07, ..., 0.0000e+00, + 5.1549e-07, 0.0000e+00], + [ 3.4226e-07, 0.0000e+00, -3.8072e-06, ..., 0.0000e+00, + -5.7556e-06, 0.0000e+00]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0030, 0.0257, 0.0046, -0.0019, 0.0100, 0.0037, 0.0193, -0.0145, + 0.0127, -0.0022], device='cuda:0'), grad: tensor([-5.2117e-06, 9.8422e-06, -8.2934e-07, 2.6803e-06, 2.2963e-05, + 4.6901e-06, 6.5845e-07, 3.6135e-06, 2.5406e-06, -4.0919e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 217.61, cls_loss 0.0034 cls_loss_mapping 0.0058 cls_loss_causal 0.5523 re_mapping 0.0072 re_causal 0.0217 /// teacc 98.87 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0445, -0.0740, -0.0457, ..., -0.0921, -0.0081, -0.0241], + [-0.0425, 0.0272, -0.1311, ..., 0.0505, 0.0605, -0.1214], + [ 0.0560, 0.0012, -0.0515, ..., -0.0492, -0.0312, 0.0180], + ..., + [ 0.0498, -0.0470, 0.0016, ..., 0.0228, -0.0335, 0.0272], + [-0.0373, -0.0284, -0.0660, ..., -0.0141, -0.0267, -0.0603], + [-0.1154, -0.1743, 0.0332, ..., -0.0525, 0.0274, 0.0390]], + device='cuda:0'), grad: tensor([[ 5.5321e-07, 0.0000e+00, 1.2293e-07, ..., 0.0000e+00, + 8.2888e-08, 1.8626e-09], + [ 1.1642e-06, 0.0000e+00, 1.1735e-07, ..., 0.0000e+00, + -6.0163e-07, 1.8626e-09], + [ 1.7092e-05, 0.0000e+00, 4.2506e-06, ..., 0.0000e+00, + 1.6298e-07, -8.3819e-09], + ..., + [ 5.6289e-06, 0.0000e+00, 1.4324e-06, ..., 0.0000e+00, + 1.8068e-07, 9.3132e-10], + [ 4.2468e-06, 0.0000e+00, 6.9663e-06, ..., 0.0000e+00, + 2.7977e-06, 0.0000e+00], + [ 1.0999e-06, 0.0000e+00, 8.0373e-07, ..., 0.0000e+00, + -3.2037e-07, 0.0000e+00]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0042, 0.0249, 0.0047, -0.0017, 0.0101, 0.0030, 0.0206, -0.0134, + 0.0126, -0.0022], device='cuda:0'), grad: tensor([-5.5023e-06, 2.7940e-09, 1.6540e-05, -2.8536e-05, -6.6217e-07, + -3.9376e-06, -3.1441e-05, 3.6415e-06, 4.4644e-05, 5.1484e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 217.86, cls_loss 0.0033 cls_loss_mapping 0.0058 cls_loss_causal 0.5633 re_mapping 0.0074 re_causal 0.0217 /// teacc 98.83 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0442, -0.0743, -0.0464, ..., -0.0921, -0.0063, -0.0247], + [-0.0426, 0.0273, -0.1324, ..., 0.0505, 0.0607, -0.1218], + [ 0.0571, 0.0012, -0.0523, ..., -0.0493, -0.0312, 0.0172], + ..., + [ 0.0493, -0.0470, 0.0018, ..., 0.0228, -0.0339, 0.0266], + [-0.0379, -0.0284, -0.0657, ..., -0.0142, -0.0270, -0.0605], + [-0.1168, -0.1744, 0.0330, ..., -0.0525, 0.0267, 0.0395]], + device='cuda:0'), grad: tensor([[-8.3819e-09, 0.0000e+00, 4.0047e-08, ..., 0.0000e+00, + 8.4750e-08, 2.1420e-08], + [ 2.3209e-06, 0.0000e+00, 1.2601e-06, ..., 1.8626e-09, + -1.3690e-07, 2.1420e-08], + [-1.2904e-05, 0.0000e+00, 3.6322e-07, ..., 1.8626e-09, + 1.9278e-07, -2.5146e-08], + ..., + [-4.9453e-07, 0.0000e+00, 7.5437e-07, ..., 1.8626e-09, + 7.5996e-07, 1.8813e-07], + [-4.2841e-08, 0.0000e+00, 6.7875e-06, ..., 1.8626e-09, + 1.3355e-06, 1.1083e-07], + [ 3.1386e-07, 0.0000e+00, -4.7591e-07, ..., 9.3132e-10, + -4.3288e-06, -7.0035e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0027, 0.0247, 0.0057, -0.0012, 0.0097, 0.0036, 0.0201, -0.0138, + 0.0126, -0.0031], device='cuda:0'), grad: tensor([-2.0005e-06, 9.4622e-06, -1.2368e-05, 1.4052e-05, -5.3942e-06, + -1.1943e-05, 3.2075e-06, 4.1127e-06, 2.1353e-05, -2.0593e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 218.05, cls_loss 0.0026 cls_loss_mapping 0.0047 cls_loss_causal 0.5490 re_mapping 0.0073 re_causal 0.0216 /// teacc 98.87 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0443, -0.0746, -0.0463, ..., -0.0921, -0.0066, -0.0247], + [-0.0428, 0.0278, -0.1330, ..., 0.0505, 0.0608, -0.1222], + [ 0.0596, 0.0007, -0.0517, ..., -0.0493, -0.0312, 0.0171], + ..., + [ 0.0474, -0.0472, 0.0017, ..., 0.0229, -0.0339, 0.0264], + [-0.0384, -0.0285, -0.0663, ..., -0.0142, -0.0270, -0.0610], + [-0.1173, -0.1745, 0.0331, ..., -0.0526, 0.0269, 0.0395]], + device='cuda:0'), grad: tensor([[ 3.0547e-07, 0.0000e+00, 1.3504e-07, ..., 9.3132e-10, + 5.8450e-06, 1.5926e-06], + [-4.0978e-08, 0.0000e+00, 4.6566e-09, ..., 5.5879e-09, + -1.5795e-06, 2.6077e-08], + [-7.4506e-09, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, + 9.6764e-07, 5.3085e-08], + ..., + [-2.9802e-07, 0.0000e+00, 0.0000e+00, ..., -1.3039e-08, + 3.2224e-07, 1.8626e-09], + [ 4.6473e-07, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 3.2224e-07, 5.0291e-08], + [ 2.9523e-07, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 4.9360e-08, 1.3970e-08]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0023, 0.0246, 0.0080, -0.0016, 0.0094, 0.0042, 0.0197, -0.0155, + 0.0126, -0.0033], device='cuda:0'), grad: tensor([ 3.6716e-05, -3.2857e-06, 1.7313e-06, -2.9318e-06, 3.0212e-06, + 4.8950e-06, -4.3869e-05, -9.3132e-08, 2.6077e-06, 1.1381e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 217.75, cls_loss 0.0029 cls_loss_mapping 0.0044 cls_loss_causal 0.5572 re_mapping 0.0070 re_causal 0.0218 /// teacc 98.85 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0440, -0.0747, -0.0463, ..., -0.0923, -0.0064, -0.0247], + [-0.0425, 0.0278, -0.1318, ..., 0.0505, 0.0614, -0.1225], + [ 0.0598, 0.0006, -0.0519, ..., -0.0494, -0.0320, 0.0171], + ..., + [ 0.0472, -0.0472, 0.0020, ..., 0.0238, -0.0336, 0.0259], + [-0.0389, -0.0284, -0.0666, ..., -0.0142, -0.0272, -0.0611], + [-0.1190, -0.1746, 0.0330, ..., -0.0527, 0.0269, 0.0397]], + device='cuda:0'), grad: tensor([[ 8.6240e-07, 0.0000e+00, 3.3528e-08, ..., 0.0000e+00, + 9.9931e-07, 0.0000e+00], + [ 4.8988e-07, 1.8626e-09, 1.8440e-07, ..., 0.0000e+00, + -1.5181e-07, 2.7940e-09], + [ 4.7684e-05, -3.7253e-09, 4.3772e-08, ..., 0.0000e+00, + 1.3970e-08, 9.3132e-10], + ..., + [-4.5076e-06, 1.8626e-09, 5.5321e-07, ..., 0.0000e+00, + 7.5437e-08, 8.3819e-09], + [-4.6343e-05, 0.0000e+00, 2.0768e-07, ..., 0.0000e+00, + 1.2200e-07, 0.0000e+00], + [ 2.4773e-07, 0.0000e+00, -1.4901e-07, ..., 0.0000e+00, + -5.4482e-07, 2.7940e-09]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0017, 0.0251, 0.0078, -0.0010, 0.0092, 0.0033, 0.0202, -0.0153, + 0.0124, -0.0038], device='cuda:0'), grad: tensor([ 6.0201e-06, 1.1977e-06, 1.4019e-04, 4.9621e-06, 3.3155e-07, + 8.0187e-07, -4.4703e-06, -3.9376e-06, -1.4472e-04, -4.8988e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 217.72, cls_loss 0.0034 cls_loss_mapping 0.0058 cls_loss_causal 0.5530 re_mapping 0.0069 re_causal 0.0213 /// teacc 98.91 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0441, -0.0755, -0.0469, ..., -0.0925, -0.0068, -0.0247], + [-0.0421, 0.0279, -0.1352, ..., 0.0510, 0.0634, -0.1232], + [ 0.0586, -0.0022, -0.0530, ..., -0.0507, -0.0325, 0.0169], + ..., + [ 0.0483, -0.0449, 0.0015, ..., 0.0237, -0.0354, 0.0235], + [-0.0393, -0.0283, -0.0660, ..., -0.0143, -0.0255, -0.0614], + [-0.1201, -0.1748, 0.0330, ..., -0.0529, 0.0270, 0.0398]], + device='cuda:0'), grad: tensor([[ 6.5472e-07, 1.8626e-09, 3.2876e-07, ..., 0.0000e+00, + -2.4699e-06, 9.3132e-10], + [ 1.1828e-07, 5.5879e-09, 9.9652e-08, ..., 0.0000e+00, + -1.4529e-07, 0.0000e+00], + [-3.1412e-05, 3.9116e-08, -2.3603e-05, ..., 0.0000e+00, + 2.4866e-07, -2.7940e-09], + ..., + [ 1.5367e-07, 1.7695e-08, 1.7602e-07, ..., 0.0000e+00, + 2.9337e-07, 9.3132e-10], + [ 2.7269e-05, -1.5646e-07, 2.1785e-05, ..., 0.0000e+00, + 5.4725e-06, 9.3132e-10], + [ 4.0978e-07, 9.3132e-10, -8.5961e-07, ..., 0.0000e+00, + -4.4852e-06, 0.0000e+00]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0023, 0.0256, 0.0065, -0.0002, 0.0093, 0.0019, 0.0202, -0.0149, + 0.0141, -0.0037], device='cuda:0'), grad: tensor([-1.4089e-05, 4.7125e-07, -9.9063e-05, 6.6571e-06, 5.4203e-06, + 2.8536e-06, -9.4809e-07, 1.4780e-06, 1.1295e-04, -1.5602e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 217.80, cls_loss 0.0030 cls_loss_mapping 0.0058 cls_loss_causal 0.5448 re_mapping 0.0072 re_causal 0.0215 /// teacc 98.93 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0442, -0.0758, -0.0471, ..., -0.0928, -0.0067, -0.0247], + [-0.0446, 0.0279, -0.1359, ..., 0.0509, 0.0622, -0.1236], + [ 0.0598, -0.0022, -0.0530, ..., -0.0508, -0.0304, 0.0169], + ..., + [ 0.0487, -0.0449, 0.0009, ..., 0.0237, -0.0353, 0.0232], + [-0.0398, -0.0279, -0.0662, ..., -0.0144, -0.0263, -0.0615], + [-0.1212, -0.1750, 0.0325, ..., -0.0528, 0.0269, 0.0398]], + device='cuda:0'), grad: tensor([[-8.3353e-07, 0.0000e+00, -2.5611e-07, ..., 0.0000e+00, + 2.4959e-07, 0.0000e+00], + [-2.0601e-06, 0.0000e+00, 1.2815e-06, ..., 0.0000e+00, + -7.8827e-06, 0.0000e+00], + [ 2.9802e-06, 0.0000e+00, 1.8394e-06, ..., 0.0000e+00, + 4.3586e-06, 0.0000e+00], + ..., + [ 8.6613e-07, 0.0000e+00, 9.4064e-08, ..., 0.0000e+00, + 4.1686e-06, 0.0000e+00], + [-4.0233e-06, 0.0000e+00, -5.5581e-06, ..., 0.0000e+00, + -3.3081e-06, 0.0000e+00], + [ 4.6287e-07, 0.0000e+00, 1.1818e-06, ..., 0.0000e+00, + 1.7416e-07, 0.0000e+00]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0022, 0.0239, 0.0079, -0.0004, 0.0096, 0.0022, 0.0199, -0.0145, + 0.0135, -0.0039], device='cuda:0'), grad: tensor([-8.4639e-06, -2.0847e-05, 3.1441e-05, 1.3344e-05, 1.7360e-06, + 1.5423e-05, 2.3842e-06, 1.7628e-05, -6.0380e-05, 7.6443e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 217.43, cls_loss 0.0032 cls_loss_mapping 0.0045 cls_loss_causal 0.5750 re_mapping 0.0072 re_causal 0.0213 /// teacc 98.92 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0445, -0.0761, -0.0478, ..., -0.0931, -0.0068, -0.0247], + [-0.0424, 0.0283, -0.1368, ..., 0.0509, 0.0646, -0.1236], + [ 0.0591, -0.0025, -0.0529, ..., -0.0508, -0.0307, 0.0169], + ..., + [ 0.0478, -0.0450, 0.0009, ..., 0.0237, -0.0373, 0.0232], + [-0.0399, -0.0279, -0.0667, ..., -0.0145, -0.0266, -0.0615], + [-0.1223, -0.1751, 0.0321, ..., -0.0528, 0.0263, 0.0398]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, 8.3819e-09, 4.1444e-07, ..., 0.0000e+00, + 4.8429e-08, 0.0000e+00], + [-9.1735e-07, -1.6214e-06, 1.2435e-05, ..., 0.0000e+00, + -2.4028e-07, 0.0000e+00], + [ 1.1344e-06, 1.1222e-06, 2.9057e-07, ..., 0.0000e+00, + 9.4995e-08, 0.0000e+00], + ..., + [-9.2294e-07, 4.2841e-08, 8.0187e-07, ..., 0.0000e+00, + 1.9837e-07, 0.0000e+00], + [-4.9360e-08, 7.9162e-08, -1.8626e-08, ..., 0.0000e+00, + -8.0280e-07, 0.0000e+00], + [ 3.3807e-07, 1.8626e-09, 2.1353e-05, ..., 0.0000e+00, + 5.6997e-07, 0.0000e+00]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0020, 0.0261, 0.0071, -0.0005, 0.0094, 0.0025, 0.0201, -0.0156, + 0.0134, -0.0044], device='cuda:0'), grad: tensor([-8.5160e-06, 2.6196e-05, 7.8380e-06, 6.6683e-07, -1.3053e-04, + 6.0834e-06, 3.9905e-05, 1.7984e-06, -4.0755e-06, 6.0737e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 157---------------------------------------------------- +epoch 157, time 218.21, cls_loss 0.0034 cls_loss_mapping 0.0053 cls_loss_causal 0.5586 re_mapping 0.0067 re_causal 0.0209 /// teacc 98.96 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0452, -0.0764, -0.0487, ..., -0.0932, -0.0071, -0.0247], + [-0.0435, 0.0286, -0.1380, ..., 0.0509, 0.0646, -0.1237], + [ 0.0579, -0.0022, -0.0525, ..., -0.0509, -0.0309, 0.0169], + ..., + [ 0.0498, -0.0454, 0.0009, ..., 0.0237, -0.0373, 0.0232], + [-0.0412, -0.0280, -0.0670, ..., -0.0145, -0.0266, -0.0615], + [-0.1213, -0.1752, 0.0311, ..., -0.0528, 0.0275, 0.0398]], + device='cuda:0'), grad: tensor([[ 1.1735e-07, 0.0000e+00, 8.8476e-08, ..., 0.0000e+00, + 2.1327e-07, 0.0000e+00], + [ 3.7253e-08, 0.0000e+00, 1.9185e-07, ..., 0.0000e+00, + -2.2098e-05, 0.0000e+00], + [-7.9162e-08, 0.0000e+00, 3.6322e-08, ..., 0.0000e+00, + 1.7295e-06, 0.0000e+00], + ..., + [-5.5879e-08, 0.0000e+00, 4.5169e-07, ..., 0.0000e+00, + 1.1539e-06, 0.0000e+00], + [ 3.6508e-06, 0.0000e+00, 1.2517e-05, ..., 0.0000e+00, + 1.9237e-05, 0.0000e+00], + [ 2.6077e-08, 0.0000e+00, -2.4773e-07, ..., 0.0000e+00, + -1.0366e-06, 0.0000e+00]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0024, 0.0252, 0.0058, -0.0021, 0.0101, 0.0026, 0.0212, -0.0139, + 0.0133, -0.0044], device='cuda:0'), grad: tensor([ 6.1747e-07, -6.8307e-05, 4.9807e-06, 7.4580e-06, 1.5581e-06, + -4.1008e-05, 2.7660e-07, 8.9258e-06, 9.3997e-05, -8.7172e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 158---------------------------------------------------- +epoch 158, time 218.38, cls_loss 0.0036 cls_loss_mapping 0.0043 cls_loss_causal 0.5536 re_mapping 0.0071 re_causal 0.0211 /// teacc 98.97 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0455, -0.0767, -0.0508, ..., -0.0932, -0.0089, -0.0255], + [-0.0436, 0.0289, -0.1382, ..., 0.0509, 0.0648, -0.1238], + [ 0.0579, -0.0023, -0.0526, ..., -0.0509, -0.0311, 0.0169], + ..., + [ 0.0478, -0.0455, 0.0009, ..., 0.0237, -0.0372, 0.0231], + [-0.0415, -0.0277, -0.0684, ..., -0.0145, -0.0270, -0.0615], + [-0.1229, -0.1755, 0.0311, ..., -0.0528, 0.0283, 0.0406]], + device='cuda:0'), grad: tensor([[ 5.8077e-06, 0.0000e+00, 2.2762e-06, ..., 0.0000e+00, + 8.5216e-07, 0.0000e+00], + [ 2.7582e-05, 0.0000e+00, 9.1456e-07, ..., 0.0000e+00, + 6.4448e-07, 0.0000e+00], + [-9.2015e-06, 0.0000e+00, 3.0734e-08, ..., 0.0000e+00, + -2.0396e-07, 0.0000e+00], + ..., + [-5.7548e-05, 0.0000e+00, 3.1292e-06, ..., 0.0000e+00, + 1.1148e-06, 0.0000e+00], + [ 6.4597e-06, 0.0000e+00, 2.8014e-06, ..., 0.0000e+00, + 3.8520e-06, 0.0000e+00], + [ 2.0847e-05, 0.0000e+00, 2.8461e-06, ..., 0.0000e+00, + -1.0237e-05, 0.0000e+00]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0036, 0.0253, 0.0057, -0.0008, 0.0101, 0.0041, 0.0214, -0.0156, + 0.0128, -0.0043], device='cuda:0'), grad: tensor([-1.9324e-04, 5.5164e-05, 2.5973e-05, 8.0407e-05, 2.9713e-05, + 3.0939e-06, 4.7356e-05, -1.2052e-04, 5.6177e-05, 1.5780e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 217.57, cls_loss 0.0030 cls_loss_mapping 0.0042 cls_loss_causal 0.5736 re_mapping 0.0069 re_causal 0.0218 /// teacc 98.88 lr 0.00010000 +Epoch 161, weight, value: tensor([[-4.5857e-02, -7.6750e-02, -5.1790e-02, ..., -9.3160e-02, + -9.3876e-03, -2.5657e-02], + [-4.3888e-02, 2.8944e-02, -1.4007e-01, ..., 5.0875e-02, + 6.4678e-02, -1.2382e-01], + [ 5.7464e-02, -2.3744e-03, -5.2539e-02, ..., -5.0855e-02, + -3.1399e-02, 1.6877e-02], + ..., + [ 4.8401e-02, -4.5495e-02, 1.2720e-04, ..., 2.3744e-02, + -3.6993e-02, 2.3039e-02], + [-4.1970e-02, -2.7749e-02, -7.1137e-02, ..., -1.4514e-02, + -2.7987e-02, -6.1501e-02], + [-1.2395e-01, -1.7548e-01, 3.0484e-02, ..., -5.2835e-02, + 2.8416e-02, 4.0770e-02]], device='cuda:0'), grad: tensor([[ 1.0151e-06, 7.4506e-09, 9.3132e-08, ..., 0.0000e+00, + 2.9393e-06, 0.0000e+00], + [-4.5672e-06, -5.4762e-06, 3.1833e-06, ..., 0.0000e+00, + -2.2188e-05, 0.0000e+00], + [ 2.8685e-06, 4.0717e-06, 1.9930e-07, ..., 0.0000e+00, + 7.0706e-06, 0.0000e+00], + ..., + [ 4.5337e-06, 1.1846e-06, 2.2110e-06, ..., 0.0000e+00, + 1.1772e-05, 0.0000e+00], + [ 8.3074e-06, 8.3819e-09, 8.1770e-07, ..., 0.0000e+00, + 6.1467e-07, 0.0000e+00], + [-1.1269e-06, 2.7940e-09, 4.8503e-06, ..., 0.0000e+00, + -1.8068e-06, 0.0000e+00]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0040, 0.0249, 0.0052, -0.0010, 0.0104, 0.0048, 0.0224, -0.0150, + 0.0115, -0.0046], device='cuda:0'), grad: tensor([ 2.6301e-06, -4.5180e-05, 2.3752e-05, -2.2769e-05, -3.8683e-05, + 3.5968e-06, 1.8431e-06, 4.0680e-05, 2.0668e-05, 1.3426e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 160---------------------------------------------------- +epoch 160, time 218.11, cls_loss 0.0026 cls_loss_mapping 0.0038 cls_loss_causal 0.5599 re_mapping 0.0070 re_causal 0.0206 /// teacc 98.99 lr 0.00010000 +Epoch 162, weight, value: tensor([[-4.5117e-02, -7.6832e-02, -5.1337e-02, ..., -9.3163e-02, + -9.3608e-03, -2.5657e-02], + [-4.4441e-02, 2.8980e-02, -1.4044e-01, ..., 5.0872e-02, + 6.4983e-02, -1.2392e-01], + [ 5.7667e-02, -2.3949e-03, -5.3077e-02, ..., -5.0859e-02, + -3.1665e-02, 1.6904e-02], + ..., + [ 4.8568e-02, -4.5502e-02, 1.5208e-04, ..., 2.3794e-02, + -3.7037e-02, 2.3019e-02], + [-4.2857e-02, -2.7810e-02, -7.1684e-02, ..., -1.4516e-02, + -2.8775e-02, -6.1504e-02], + [-1.2449e-01, -1.7553e-01, 3.0229e-02, ..., -5.2838e-02, + 2.8715e-02, 4.0770e-02]], device='cuda:0'), grad: tensor([[ 1.3709e-06, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 6.4168e-07, 0.0000e+00], + [ 5.2974e-06, 0.0000e+00, 2.9802e-08, ..., 0.0000e+00, + 5.8766e-07, 0.0000e+00], + [ 6.5845e-07, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 9.8813e-07, 0.0000e+00], + ..., + [-2.5891e-07, 0.0000e+00, 2.6077e-08, ..., 0.0000e+00, + 1.1558e-06, 0.0000e+00], + [ 3.4049e-06, 0.0000e+00, 2.4214e-08, ..., 0.0000e+00, + -4.3493e-07, 0.0000e+00], + [ 9.9838e-06, 0.0000e+00, 2.9802e-08, ..., 0.0000e+00, + 3.1516e-06, 0.0000e+00]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0034, 0.0246, 0.0054, -0.0009, 0.0102, 0.0047, 0.0226, -0.0148, + 0.0106, -0.0046], device='cuda:0'), grad: tensor([ 4.8019e-06, 1.5199e-05, 5.3905e-06, -7.6711e-05, 5.9754e-06, + 1.5259e-05, -6.6385e-06, 1.4622e-07, -1.3849e-06, 3.7879e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 217.40, cls_loss 0.0030 cls_loss_mapping 0.0047 cls_loss_causal 0.5517 re_mapping 0.0069 re_causal 0.0209 /// teacc 98.89 lr 0.00010000 +Epoch 163, weight, value: tensor([[-4.5946e-02, -7.7002e-02, -5.2057e-02, ..., -9.3185e-02, + -9.5410e-03, -2.5657e-02], + [-4.3536e-02, 2.9060e-02, -1.4104e-01, ..., 5.0846e-02, + 6.6817e-02, -1.2396e-01], + [ 5.7733e-02, -2.4102e-03, -5.4186e-02, ..., -5.0873e-02, + -3.1812e-02, 1.6908e-02], + ..., + [ 4.8195e-02, -4.5548e-02, 1.2527e-04, ..., 2.3880e-02, + -3.9076e-02, 2.3016e-02], + [-4.4463e-02, -2.7671e-02, -7.3438e-02, ..., -1.4531e-02, + -2.8740e-02, -6.1509e-02], + [-1.2521e-01, -1.7565e-01, 3.0465e-02, ..., -5.2845e-02, + 2.8975e-02, 4.0770e-02]], device='cuda:0'), grad: tensor([[ 1.9837e-07, 0.0000e+00, 1.5832e-08, ..., 0.0000e+00, + 1.0896e-07, 0.0000e+00], + [ 5.7463e-07, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + -5.2303e-06, 0.0000e+00], + [-2.1923e-06, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.2573e-07, 0.0000e+00], + ..., + [ 7.8231e-08, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 4.8149e-07, 0.0000e+00], + [ 5.2713e-07, 4.6566e-09, 1.9185e-07, ..., 0.0000e+00, + 4.2878e-06, 0.0000e+00], + [ 6.7987e-08, 0.0000e+00, 1.2387e-07, ..., 0.0000e+00, + 1.3132e-07, 0.0000e+00]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0037, 0.0260, 0.0052, -0.0010, 0.0106, 0.0054, 0.0222, -0.0158, + 0.0100, -0.0045], device='cuda:0'), grad: tensor([ 2.0303e-07, -1.0371e-05, -2.8480e-06, 1.0291e-06, 9.9652e-08, + 4.3027e-07, -3.8184e-07, 1.2992e-06, 1.0043e-05, 4.8894e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 217.49, cls_loss 0.0033 cls_loss_mapping 0.0041 cls_loss_causal 0.5250 re_mapping 0.0072 re_causal 0.0202 /// teacc 98.97 lr 0.00010000 +Epoch 164, weight, value: tensor([[-4.6336e-02, -7.7186e-02, -5.3196e-02, ..., -9.3215e-02, + -1.2330e-02, -2.8327e-02], + [-4.4323e-02, 2.9254e-02, -1.4132e-01, ..., 5.0783e-02, + 6.7174e-02, -1.2457e-01], + [ 5.7638e-02, -2.2848e-03, -5.4184e-02, ..., -5.0889e-02, + -3.2044e-02, 1.6707e-02], + ..., + [ 4.8958e-02, -4.5794e-02, -6.4190e-05, ..., 2.3987e-02, + -3.9013e-02, 2.2711e-02], + [-4.6292e-02, -2.7411e-02, -7.3815e-02, ..., -1.4554e-02, + -2.9622e-02, -6.1524e-02], + [-1.2614e-01, -1.7583e-01, 3.0799e-02, ..., -5.2888e-02, + 3.1135e-02, 4.3435e-02]], device='cuda:0'), grad: tensor([[ 2.2016e-06, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 6.7592e-05, 5.5879e-09, 1.4901e-08, ..., 0.0000e+00, + -4.9639e-07, 0.0000e+00], + [-2.7880e-05, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + ..., + [-5.8919e-05, -8.3819e-09, 2.7008e-08, ..., 0.0000e+00, + 1.5460e-07, 0.0000e+00], + [ 4.2021e-06, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 2.2911e-07, 0.0000e+00], + [ 1.5069e-06, 0.0000e+00, 2.4214e-08, ..., 0.0000e+00, + -2.3935e-07, 0.0000e+00]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0054, 0.0258, 0.0051, -0.0011, 0.0105, 0.0055, 0.0230, -0.0152, + 0.0087, -0.0037], device='cuda:0'), grad: tensor([-4.3511e-06, 1.3995e-04, -4.7803e-05, 1.2711e-05, 2.4773e-06, + 5.5246e-06, 5.1185e-06, -1.2827e-04, 8.6054e-06, 6.0871e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 163---------------------------------------------------- +epoch 163, time 218.26, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5264 re_mapping 0.0069 re_causal 0.0208 /// teacc 99.05 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0463, -0.0772, -0.0532, ..., -0.0934, -0.0123, -0.0283], + [-0.0447, 0.0293, -0.1418, ..., 0.0502, 0.0672, -0.1246], + [ 0.0577, -0.0023, -0.0545, ..., -0.0509, -0.0323, 0.0167], + ..., + [ 0.0492, -0.0458, -0.0002, ..., 0.0250, -0.0388, 0.0227], + [-0.0464, -0.0274, -0.0735, ..., -0.0146, -0.0297, -0.0615], + [-0.1270, -0.1759, 0.0305, ..., -0.0544, 0.0312, 0.0434]], + device='cuda:0'), grad: tensor([[ 5.1502e-07, 0.0000e+00, 5.2620e-07, ..., 8.3819e-09, + 2.3376e-07, 0.0000e+00], + [-8.2999e-06, 0.0000e+00, 4.5635e-06, ..., 4.0978e-08, + -1.5542e-05, 0.0000e+00], + [ 4.5151e-05, 0.0000e+00, 1.9968e-04, ..., 1.3970e-08, + 8.4657e-07, -2.7940e-09], + ..., + [ 6.4671e-06, 0.0000e+00, 3.3695e-06, ..., 5.5879e-08, + 1.2413e-05, 9.3132e-10], + [ 1.2284e-06, 0.0000e+00, 2.8443e-06, ..., 2.7940e-09, + 2.0675e-07, 0.0000e+00], + [ 1.8030e-06, 0.0000e+00, 7.8790e-07, ..., 8.7544e-08, + 1.3337e-06, 0.0000e+00]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0048, 0.0256, 0.0050, -0.0013, 0.0105, 0.0055, 0.0232, -0.0149, + 0.0089, -0.0041], device='cuda:0'), grad: tensor([-3.7365e-06, -4.0889e-05, 3.4976e-04, -2.4773e-06, -3.7122e-04, + 1.1437e-06, 3.6266e-06, 4.4227e-05, 6.7800e-06, 1.3463e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 217.47, cls_loss 0.0029 cls_loss_mapping 0.0033 cls_loss_causal 0.5503 re_mapping 0.0070 re_causal 0.0209 /// teacc 98.99 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0468, -0.0773, -0.0539, ..., -0.0943, -0.0124, -0.0283], + [-0.0449, 0.0296, -0.1427, ..., 0.0519, 0.0680, -0.1246], + [ 0.0571, -0.0024, -0.0553, ..., -0.0510, -0.0325, 0.0167], + ..., + [ 0.0497, -0.0459, -0.0006, ..., 0.0241, -0.0395, 0.0227], + [-0.0472, -0.0274, -0.0739, ..., -0.0151, -0.0301, -0.0615], + [-0.1277, -0.1759, 0.0296, ..., -0.0546, 0.0314, 0.0434]], + device='cuda:0'), grad: tensor([[ 1.6671e-07, 0.0000e+00, 6.5472e-07, ..., 9.5926e-08, + 3.2596e-07, 0.0000e+00], + [ 1.8766e-06, 0.0000e+00, 1.8179e-06, ..., 3.0175e-07, + 3.1665e-08, 0.0000e+00], + [-1.8999e-06, 0.0000e+00, -2.9244e-07, ..., 6.5193e-08, + 7.9162e-08, 0.0000e+00], + ..., + [-1.3970e-08, 0.0000e+00, 2.4438e-06, ..., 4.0047e-07, + 1.9837e-07, 0.0000e+00], + [ 7.9721e-07, 0.0000e+00, 4.8727e-06, ..., 5.8953e-07, + 3.4153e-05, 0.0000e+00], + [ 6.2063e-06, 0.0000e+00, 5.7429e-05, ..., 5.1595e-06, + 1.0207e-06, 0.0000e+00]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0045, 0.0259, 0.0043, -0.0012, 0.0108, 0.0054, 0.0237, -0.0147, + 0.0084, -0.0046], device='cuda:0'), grad: tensor([ 3.2187e-06, 1.0543e-05, -3.2801e-06, 1.3709e-06, -1.9014e-04, + 4.7708e-04, -6.4754e-04, 3.2224e-06, 1.8990e-04, 1.5438e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 217.56, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5648 re_mapping 0.0069 re_causal 0.0207 /// teacc 98.99 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0462, -0.0773, -0.0544, ..., -0.0945, -0.0121, -0.0283], + [-0.0451, 0.0296, -0.1436, ..., 0.0521, 0.0683, -0.1247], + [ 0.0571, -0.0024, -0.0551, ..., -0.0510, -0.0327, 0.0166], + ..., + [ 0.0499, -0.0459, -0.0008, ..., 0.0240, -0.0399, 0.0227], + [-0.0475, -0.0274, -0.0741, ..., -0.0153, -0.0304, -0.0615], + [-0.1284, -0.1759, 0.0292, ..., -0.0548, 0.0316, 0.0434]], + device='cuda:0'), grad: tensor([[ 2.4773e-07, 0.0000e+00, -1.5497e-06, ..., 1.1176e-08, + 1.8533e-07, 0.0000e+00], + [ 4.2375e-07, 0.0000e+00, 5.3179e-07, ..., 4.0978e-08, + 7.4506e-09, 0.0000e+00], + [-1.9297e-06, 0.0000e+00, 7.7207e-07, ..., 1.1176e-08, + 1.4529e-07, 0.0000e+00], + ..., + [-3.6228e-07, 0.0000e+00, 5.0385e-07, ..., 1.0058e-07, + 1.0068e-06, 0.0000e+00], + [ 5.1875e-07, 0.0000e+00, 5.8860e-07, ..., 9.3132e-09, + 5.0887e-06, 0.0000e+00], + [ 3.9581e-07, 0.0000e+00, 2.1905e-06, ..., 1.5553e-07, + -1.2159e-05, 0.0000e+00]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0041, 0.0257, 0.0043, -0.0019, 0.0108, 0.0064, 0.0238, -0.0147, + 0.0082, -0.0046], device='cuda:0'), grad: tensor([-1.0706e-05, 2.3842e-06, -3.2131e-07, 2.1532e-06, 6.7279e-06, + 6.0461e-06, 4.2543e-06, 2.3358e-06, 2.2218e-05, -3.5197e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 217.68, cls_loss 0.0028 cls_loss_mapping 0.0046 cls_loss_causal 0.5522 re_mapping 0.0065 re_causal 0.0200 /// teacc 98.98 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0465, -0.0773, -0.0549, ..., -0.0960, -0.0125, -0.0283], + [-0.0452, 0.0297, -0.1446, ..., 0.0515, 0.0685, -0.1249], + [ 0.0573, -0.0025, -0.0549, ..., -0.0512, -0.0331, 0.0166], + ..., + [ 0.0499, -0.0459, -0.0009, ..., 0.0262, -0.0399, 0.0225], + [-0.0479, -0.0275, -0.0744, ..., -0.0158, -0.0305, -0.0615], + [-0.1296, -0.1759, 0.0293, ..., -0.0575, 0.0319, 0.0435]], + device='cuda:0'), grad: tensor([[ 6.1467e-08, 2.1420e-08, 2.9150e-07, ..., 4.6566e-09, + 6.6534e-06, 0.0000e+00], + [ 1.5553e-07, 6.7987e-08, 1.4249e-07, ..., 2.9802e-08, + -8.5402e-07, 0.0000e+00], + [ 9.9093e-07, 6.5193e-09, 3.1199e-07, ..., 1.3970e-08, + 5.0850e-06, 0.0000e+00], + ..., + [ 2.1234e-06, 4.6566e-09, 1.7043e-07, ..., 6.0536e-08, + 1.1092e-06, 0.0000e+00], + [ 1.1921e-07, 4.6566e-09, -1.3048e-06, ..., 2.7940e-09, + -3.4750e-05, 0.0000e+00], + [ 1.1483e-06, 4.6566e-09, 1.8943e-06, ..., 2.1327e-07, + 3.7979e-06, 0.0000e+00]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0040, 0.0256, 0.0044, -0.0017, 0.0106, 0.0061, 0.0245, -0.0145, + 0.0083, -0.0051], device='cuda:0'), grad: tensor([ 2.5034e-05, -6.6496e-07, 2.5898e-05, 1.3337e-05, -4.2319e-06, + -1.2740e-06, 6.6221e-05, 3.5781e-06, -1.4889e-04, 2.0996e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 217.51, cls_loss 0.0034 cls_loss_mapping 0.0045 cls_loss_causal 0.5187 re_mapping 0.0066 re_causal 0.0194 /// teacc 98.93 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0469, -0.0774, -0.0553, ..., -0.0961, -0.0143, -0.0287], + [-0.0476, 0.0297, -0.1454, ..., 0.0514, 0.0674, -0.1250], + [ 0.0574, -0.0024, -0.0551, ..., -0.0512, -0.0335, 0.0165], + ..., + [ 0.0512, -0.0459, -0.0023, ..., 0.0264, -0.0385, 0.0224], + [-0.0487, -0.0275, -0.0748, ..., -0.0159, -0.0305, -0.0615], + [-0.1304, -0.1759, 0.0293, ..., -0.0578, 0.0336, 0.0438]], + device='cuda:0'), grad: tensor([[-2.4233e-06, 0.0000e+00, 1.2107e-08, ..., 0.0000e+00, + -1.5125e-05, 1.1176e-08], + [ 3.9786e-06, 3.7253e-09, 1.7695e-08, ..., 0.0000e+00, + -1.9893e-06, 2.7940e-09], + [-2.8834e-05, 9.3132e-10, -1.7695e-08, ..., 0.0000e+00, + 2.2445e-06, -1.0431e-07], + ..., + [ 3.2317e-07, -8.3819e-09, 2.6077e-08, ..., 0.0000e+00, + 1.1111e-06, 2.1420e-08], + [ 1.3247e-05, 9.3132e-10, 4.4703e-08, ..., 0.0000e+00, + 3.7625e-07, 7.4506e-09], + [ 4.9546e-06, 0.0000e+00, 1.7509e-07, ..., 0.0000e+00, + 1.2144e-05, 9.3132e-09]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0052, 0.0233, 0.0042, -0.0015, 0.0106, 0.0059, 0.0247, -0.0128, + 0.0085, -0.0044], device='cuda:0'), grad: tensor([-1.3173e-04, 6.2864e-07, -4.6223e-05, 1.4842e-05, 3.0678e-06, + 1.9725e-06, 8.0094e-06, 3.7253e-06, 3.2663e-05, 1.1313e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 217.62, cls_loss 0.0027 cls_loss_mapping 0.0058 cls_loss_causal 0.5603 re_mapping 0.0067 re_causal 0.0196 /// teacc 98.95 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0471, -0.0774, -0.0554, ..., -0.0962, -0.0147, -0.0288], + [-0.0479, 0.0298, -0.1472, ..., 0.0513, 0.0672, -0.1251], + [ 0.0579, -0.0025, -0.0546, ..., -0.0512, -0.0326, 0.0165], + ..., + [ 0.0510, -0.0460, -0.0052, ..., 0.0265, -0.0400, 0.0224], + [-0.0491, -0.0275, -0.0763, ..., -0.0159, -0.0311, -0.0616], + [-0.1292, -0.1760, 0.0293, ..., -0.0579, 0.0357, 0.0440]], + device='cuda:0'), grad: tensor([[ 2.8312e-07, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + -3.1665e-08, 0.0000e+00], + [ 4.8801e-06, 0.0000e+00, 2.1793e-07, ..., 0.0000e+00, + 1.3746e-06, 0.0000e+00], + [-4.7348e-06, 0.0000e+00, -2.1420e-08, ..., 0.0000e+00, + 1.7509e-07, 0.0000e+00], + ..., + [ 6.8955e-06, 0.0000e+00, 7.3574e-08, ..., 0.0000e+00, + 2.1793e-06, 0.0000e+00], + [ 1.3672e-06, 0.0000e+00, 3.8557e-07, ..., 0.0000e+00, + 2.7493e-06, 0.0000e+00], + [-1.1928e-05, 0.0000e+00, 4.7870e-06, ..., 0.0000e+00, + -5.4687e-06, 0.0000e+00]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0050, 0.0230, 0.0049, -0.0016, 0.0106, 0.0060, 0.0244, -0.0135, + 0.0080, -0.0033], device='cuda:0'), grad: tensor([-8.3745e-05, 1.4491e-05, -1.7378e-06, 1.4357e-05, -9.7230e-06, + 2.9951e-06, -1.1288e-05, 3.9488e-05, 1.7539e-05, 1.7643e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 217.55, cls_loss 0.0025 cls_loss_mapping 0.0033 cls_loss_causal 0.5709 re_mapping 0.0070 re_causal 0.0216 /// teacc 98.99 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0474, -0.0775, -0.0557, ..., -0.0965, -0.0148, -0.0288], + [-0.0480, 0.0298, -0.1475, ..., 0.0513, 0.0675, -0.1251], + [ 0.0585, -0.0025, -0.0548, ..., -0.0513, -0.0328, 0.0165], + ..., + [ 0.0508, -0.0460, -0.0056, ..., 0.0267, -0.0401, 0.0224], + [-0.0490, -0.0275, -0.0768, ..., -0.0161, -0.0308, -0.0616], + [-0.1308, -0.1760, 0.0286, ..., -0.0579, 0.0356, 0.0440]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, 0.0000e+00, 4.3772e-08, ..., 0.0000e+00, + 6.8918e-08, 0.0000e+00], + [ 3.3528e-08, 0.0000e+00, 3.2224e-07, ..., 0.0000e+00, + 3.8091e-07, 0.0000e+00], + [-2.6077e-08, 0.0000e+00, 2.0396e-07, ..., 0.0000e+00, + 4.0978e-07, 0.0000e+00], + ..., + [ 2.4214e-08, 0.0000e+00, 3.5856e-07, ..., 0.0000e+00, + 3.7905e-07, 0.0000e+00], + [ 3.1665e-08, 0.0000e+00, 3.1199e-07, ..., 0.0000e+00, + 5.0291e-08, 0.0000e+00], + [ 1.8626e-08, 0.0000e+00, 1.3508e-05, ..., 0.0000e+00, + -1.8384e-06, 0.0000e+00]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0050, 0.0232, 0.0051, -0.0016, 0.0115, 0.0060, 0.0249, -0.0138, + 0.0082, -0.0040], device='cuda:0'), grad: tensor([-5.3085e-08, 2.4959e-06, 1.7369e-06, 3.7998e-07, -3.4571e-05, + -3.2876e-07, -2.5034e-06, 1.9297e-06, 1.6764e-08, 3.0965e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 217.37, cls_loss 0.0055 cls_loss_mapping 0.0079 cls_loss_causal 0.5374 re_mapping 0.0066 re_causal 0.0194 /// teacc 98.98 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0486, -0.0777, -0.0559, ..., -0.0975, -0.0149, -0.0288], + [-0.0497, 0.0299, -0.1479, ..., 0.0565, 0.0661, -0.1251], + [ 0.0581, -0.0025, -0.0548, ..., -0.0515, -0.0343, 0.0165], + ..., + [ 0.0525, -0.0460, -0.0057, ..., 0.0216, -0.0383, 0.0224], + [-0.0497, -0.0268, -0.0770, ..., -0.0170, -0.0318, -0.0616], + [-0.1322, -0.1763, 0.0288, ..., -0.0568, 0.0354, 0.0440]], + device='cuda:0'), grad: tensor([[ 4.7497e-08, 0.0000e+00, -4.0559e-07, ..., 8.8476e-09, + 1.0245e-07, 0.0000e+00], + [ 8.3726e-07, 0.0000e+00, 1.1362e-07, ..., 9.7789e-09, + -4.4936e-07, 0.0000e+00], + [-1.4361e-06, 0.0000e+00, 1.1688e-07, ..., 2.1886e-08, + 9.0338e-08, 0.0000e+00], + ..., + [ 6.9663e-07, 0.0000e+00, 1.1921e-07, ..., 5.4482e-08, + 2.4028e-07, 0.0000e+00], + [ 1.2852e-07, 0.0000e+00, 1.9511e-07, ..., 1.9092e-08, + 3.2363e-07, 0.0000e+00], + [ 4.5681e-07, 0.0000e+00, 8.3912e-07, ..., 3.1665e-08, + -3.1292e-07, 0.0000e+00]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0048, 0.0222, 0.0039, -0.0015, 0.0110, 0.0055, 0.0248, -0.0124, + 0.0078, -0.0041], device='cuda:0'), grad: tensor([-3.6210e-06, 3.9814e-07, -1.5739e-06, 9.3654e-06, -1.7276e-07, + -7.1637e-06, -2.6654e-06, 1.4063e-06, 1.7658e-06, 2.2445e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 217.23, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5282 re_mapping 0.0066 re_causal 0.0205 /// teacc 98.98 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0492, -0.0784, -0.0559, ..., -0.0977, -0.0150, -0.0288], + [-0.0500, 0.0305, -0.1486, ..., 0.0564, 0.0662, -0.1251], + [ 0.0576, -0.0027, -0.0546, ..., -0.0516, -0.0349, 0.0165], + ..., + [ 0.0531, -0.0461, -0.0057, ..., 0.0218, -0.0383, 0.0224], + [-0.0500, -0.0266, -0.0769, ..., -0.0171, -0.0315, -0.0616], + [-0.1324, -0.1766, 0.0287, ..., -0.0576, 0.0360, 0.0440]], + device='cuda:0'), grad: tensor([[ 3.9581e-08, 0.0000e+00, -1.8207e-07, ..., 0.0000e+00, + -4.6846e-07, 0.0000e+00], + [ 4.2142e-07, 1.1176e-08, 7.5437e-08, ..., 0.0000e+00, + 2.0023e-08, 0.0000e+00], + [ 1.2061e-07, 3.9581e-08, 9.2201e-08, ..., 0.0000e+00, + 1.6298e-08, 0.0000e+00], + ..., + [ 2.5146e-06, -5.5879e-08, 6.0070e-08, ..., 0.0000e+00, + 1.0878e-06, 0.0000e+00], + [ 2.6729e-07, 2.3283e-09, -6.8499e-07, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 8.3074e-07, 0.0000e+00, 2.3330e-07, ..., 0.0000e+00, + 1.4761e-07, 0.0000e+00]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0048, 0.0220, 0.0034, -0.0016, 0.0107, 0.0057, 0.0240, -0.0121, + 0.0081, -0.0040], device='cuda:0'), grad: tensor([-6.1929e-05, 9.0478e-07, 5.3085e-07, -4.6268e-06, -1.6671e-07, + 7.5735e-06, 5.0366e-05, 3.3751e-06, 2.9197e-07, 3.5688e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 217.41, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.5382 re_mapping 0.0066 re_causal 0.0206 /// teacc 98.99 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0498, -0.0789, -0.0559, ..., -0.0978, -0.0150, -0.0288], + [-0.0502, 0.0306, -0.1494, ..., 0.0565, 0.0661, -0.1251], + [ 0.0578, -0.0027, -0.0547, ..., -0.0516, -0.0348, 0.0165], + ..., + [ 0.0532, -0.0461, -0.0061, ..., 0.0218, -0.0384, 0.0224], + [-0.0506, -0.0267, -0.0774, ..., -0.0171, -0.0316, -0.0616], + [-0.1330, -0.1775, 0.0287, ..., -0.0577, 0.0363, 0.0440]], + device='cuda:0'), grad: tensor([[ 1.5367e-07, 0.0000e+00, 4.1444e-08, ..., 4.6566e-10, + 6.8452e-08, 0.0000e+00], + [ 4.4843e-07, 0.0000e+00, 2.9802e-08, ..., 4.6566e-10, + -1.2349e-06, 0.0000e+00], + [ 4.8103e-07, 0.0000e+00, 6.9849e-09, ..., 0.0000e+00, + 1.8347e-07, 0.0000e+00], + ..., + [-3.6843e-06, 0.0000e+00, 1.9092e-08, ..., 4.6566e-10, + 7.2410e-07, 0.0000e+00], + [ 4.9826e-07, 0.0000e+00, 1.5786e-07, ..., 2.7940e-09, + 1.9558e-07, 0.0000e+00], + [ 4.2748e-07, 0.0000e+00, 3.6787e-08, ..., 4.6566e-10, + 2.1420e-08, 0.0000e+00]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0048, 0.0219, 0.0035, -0.0017, 0.0113, 0.0054, 0.0251, -0.0121, + 0.0079, -0.0041], device='cuda:0'), grad: tensor([-3.2969e-07, -1.4976e-06, 1.2200e-06, 1.3344e-05, 3.1153e-07, + -1.0528e-05, -1.2852e-07, -4.8652e-06, 1.2126e-06, 1.2703e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 217.46, cls_loss 0.0018 cls_loss_mapping 0.0045 cls_loss_causal 0.5394 re_mapping 0.0066 re_causal 0.0205 /// teacc 98.95 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0501, -0.0797, -0.0562, ..., -0.0986, -0.0151, -0.0288], + [-0.0502, 0.0317, -0.1499, ..., 0.0564, 0.0662, -0.1251], + [ 0.0577, -0.0035, -0.0547, ..., -0.0518, -0.0352, 0.0165], + ..., + [ 0.0534, -0.0463, -0.0064, ..., 0.0218, -0.0384, 0.0223], + [-0.0513, -0.0275, -0.0775, ..., -0.0175, -0.0314, -0.0616], + [-0.1336, -0.1782, 0.0285, ..., -0.0570, 0.0366, 0.0440]], + device='cuda:0'), grad: tensor([[ 1.4901e-07, 0.0000e+00, 1.4435e-07, ..., 9.3132e-10, + 2.0117e-07, 0.0000e+00], + [ 3.7067e-07, 0.0000e+00, 2.6077e-08, ..., 0.0000e+00, + -4.3772e-08, 0.0000e+00], + [ 3.8277e-07, 0.0000e+00, 2.7940e-08, ..., 0.0000e+00, + 6.6124e-08, 0.0000e+00], + ..., + [-2.6897e-06, 0.0000e+00, 2.2352e-08, ..., 0.0000e+00, + 1.7695e-08, 0.0000e+00], + [ 1.3281e-06, 0.0000e+00, 1.5348e-06, ..., 1.8626e-09, + 2.2277e-06, 0.0000e+00], + [ 1.6894e-06, 0.0000e+00, -2.5593e-06, ..., 3.7253e-09, + -4.1239e-06, 0.0000e+00]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0047, 0.0219, 0.0033, -0.0006, 0.0113, 0.0038, 0.0250, -0.0120, + 0.0077, -0.0040], device='cuda:0'), grad: tensor([ 1.6699e-06, 6.5565e-07, 1.9521e-06, 4.6343e-06, 1.2545e-06, + 3.6787e-07, 1.9372e-07, -4.6156e-06, 1.6019e-05, -2.2128e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 217.71, cls_loss 0.0026 cls_loss_mapping 0.0034 cls_loss_causal 0.5116 re_mapping 0.0066 re_causal 0.0198 /// teacc 98.87 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0506, -0.0815, -0.0566, ..., -0.0989, -0.0152, -0.0288], + [-0.0502, 0.0346, -0.1505, ..., 0.0565, 0.0662, -0.1252], + [ 0.0579, -0.0043, -0.0549, ..., -0.0530, -0.0354, 0.0164], + ..., + [ 0.0535, -0.0482, -0.0068, ..., 0.0217, -0.0384, 0.0225], + [-0.0517, -0.0277, -0.0783, ..., -0.0177, -0.0320, -0.0616], + [-0.1337, -0.1789, 0.0304, ..., -0.0571, 0.0368, 0.0440]], + device='cuda:0'), grad: tensor([[ 1.2387e-07, 0.0000e+00, 4.2655e-07, ..., 8.3819e-09, + 1.3970e-08, 0.0000e+00], + [ 4.2282e-07, 0.0000e+00, 5.6904e-07, ..., 9.3132e-08, + -3.9116e-08, 0.0000e+00], + [ 3.0641e-07, 0.0000e+00, 4.9546e-07, ..., 6.7987e-08, + 9.3132e-09, 0.0000e+00], + ..., + [ 2.0396e-07, 0.0000e+00, 4.9081e-07, ..., 7.4506e-08, + 8.9407e-08, 0.0000e+00], + [ 2.8498e-07, 0.0000e+00, 1.1828e-07, ..., 1.8626e-09, + 2.5146e-08, 0.0000e+00], + [ 2.2072e-07, 0.0000e+00, -1.6734e-05, ..., 2.1420e-08, + -1.3132e-07, 0.0000e+00]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0047, 0.0219, 0.0033, -0.0008, 0.0094, 0.0048, 0.0238, -0.0120, + 0.0073, -0.0023], device='cuda:0'), grad: tensor([ 8.7637e-07, 1.5022e-06, 1.6857e-06, -4.4107e-05, 2.3648e-05, + 4.7028e-05, 1.9073e-06, 1.6000e-06, 5.8115e-07, -3.4839e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 217.51, cls_loss 0.0023 cls_loss_mapping 0.0032 cls_loss_causal 0.5346 re_mapping 0.0065 re_causal 0.0200 /// teacc 98.95 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0510, -0.0844, -0.0569, ..., -0.0993, -0.0153, -0.0289], + [-0.0501, 0.0381, -0.1520, ..., 0.0567, 0.0666, -0.1252], + [ 0.0577, -0.0067, -0.0551, ..., -0.0537, -0.0363, 0.0163], + ..., + [ 0.0536, -0.0498, -0.0068, ..., 0.0217, -0.0386, 0.0224], + [-0.0519, -0.0299, -0.0786, ..., -0.0178, -0.0324, -0.0616], + [-0.1342, -0.1804, 0.0313, ..., -0.0572, 0.0367, 0.0441]], + device='cuda:0'), grad: tensor([[ 2.3358e-06, 0.0000e+00, 1.5367e-06, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 1.4555e-04, 0.0000e+00, 3.4645e-07, ..., 0.0000e+00, + -1.2191e-06, 0.0000e+00], + [ 1.8477e-05, 0.0000e+00, 5.4576e-07, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [-2.3496e-04, 0.0000e+00, 7.2643e-08, ..., 0.0000e+00, + 1.5553e-07, 0.0000e+00], + [ 8.0049e-05, 0.0000e+00, 1.8710e-06, ..., 0.0000e+00, + 5.8580e-07, 0.0000e+00], + [ 1.0774e-05, 0.0000e+00, 1.1653e-05, ..., 0.0000e+00, + 1.1921e-07, 0.0000e+00]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0047, 0.0223, 0.0025, -0.0010, 0.0084, 0.0045, 0.0237, -0.0121, + 0.0073, -0.0015], device='cuda:0'), grad: tensor([ 1.3605e-05, 4.6992e-04, 5.9366e-05, 4.4644e-05, 4.5486e-06, + -8.6725e-05, -4.1515e-05, -7.9012e-04, 2.5797e-04, 6.7592e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 217.40, cls_loss 0.0024 cls_loss_mapping 0.0037 cls_loss_causal 0.5344 re_mapping 0.0067 re_causal 0.0197 /// teacc 98.92 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0513, -0.0846, -0.0570, ..., -0.0994, -0.0154, -0.0289], + [-0.0502, 0.0382, -0.1522, ..., 0.0568, 0.0666, -0.1252], + [ 0.0580, -0.0068, -0.0552, ..., -0.0538, -0.0356, 0.0163], + ..., + [ 0.0537, -0.0498, -0.0069, ..., 0.0215, -0.0387, 0.0224], + [-0.0527, -0.0304, -0.0791, ..., -0.0179, -0.0325, -0.0616], + [-0.1346, -0.1809, 0.0311, ..., -0.0572, 0.0369, 0.0441]], + device='cuda:0'), grad: tensor([[ 8.1956e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 1.9008e-06, 0.0000e+00, 1.0524e-07, ..., 0.0000e+00, + -5.1316e-07, 0.0000e+00], + [ 7.5437e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + ..., + [ 4.9081e-07, 0.0000e+00, -1.0338e-07, ..., 0.0000e+00, + 6.6683e-07, 0.0000e+00], + [ 4.0978e-08, 0.0000e+00, 7.6089e-07, ..., 0.0000e+00, + -4.0978e-08, 0.0000e+00], + [-7.0184e-06, 0.0000e+00, 2.9523e-07, ..., 0.0000e+00, + -5.6066e-07, 0.0000e+00]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0048, 0.0223, 0.0030, -0.0011, 0.0106, 0.0045, 0.0239, -0.0121, + 0.0068, -0.0036], device='cuda:0'), grad: tensor([ 1.0151e-07, 2.7232e-06, 2.9430e-07, 2.7195e-06, 1.2636e-05, + -2.7567e-06, 6.4913e-07, 8.3297e-06, 1.2098e-06, -2.5928e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 217.36, cls_loss 0.0020 cls_loss_mapping 0.0032 cls_loss_causal 0.4922 re_mapping 0.0066 re_causal 0.0197 /// teacc 99.00 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0528, -0.0848, -0.0571, ..., -0.0995, -0.0156, -0.0289], + [-0.0502, 0.0383, -0.1555, ..., 0.0569, 0.0663, -0.1252], + [ 0.0580, -0.0068, -0.0551, ..., -0.0546, -0.0356, 0.0163], + ..., + [ 0.0536, -0.0499, -0.0071, ..., 0.0215, -0.0389, 0.0224], + [-0.0532, -0.0306, -0.0793, ..., -0.0179, -0.0336, -0.0616], + [-0.1353, -0.1813, 0.0312, ..., -0.0567, 0.0378, 0.0441]], + device='cuda:0'), grad: tensor([[-1.0338e-06, 0.0000e+00, 1.1176e-08, ..., 9.3132e-10, + 1.0077e-06, 0.0000e+00], + [ 2.9057e-07, -9.3132e-10, 1.4342e-07, ..., 1.2107e-08, + -5.4296e-07, 0.0000e+00], + [-7.0669e-06, 0.0000e+00, 2.4214e-08, ..., 9.3132e-10, + 1.4687e-06, 0.0000e+00], + ..., + [ 8.2254e-06, 9.3132e-10, 2.7940e-08, ..., 1.8626e-09, + 4.9453e-07, 0.0000e+00], + [ 5.4911e-06, 0.0000e+00, 4.0978e-08, ..., 9.3132e-10, + 7.5996e-07, 0.0000e+00], + [ 2.1681e-06, 0.0000e+00, 1.3784e-07, ..., 1.1176e-08, + 1.1455e-07, 0.0000e+00]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0053, 0.0218, 0.0030, -0.0009, 0.0117, 0.0045, 0.0242, -0.0123, + 0.0061, -0.0035], device='cuda:0'), grad: tensor([-1.0490e-05, -2.0899e-06, -1.4510e-06, -1.3478e-05, 6.2101e-06, + 7.2829e-07, -1.9118e-05, 1.6659e-05, 1.0513e-05, 1.2442e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.36, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5240 re_mapping 0.0070 re_causal 0.0196 /// teacc 98.83 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0541, -0.0853, -0.0572, ..., -0.0995, -0.0157, -0.0289], + [-0.0502, 0.0384, -0.1557, ..., 0.0569, 0.0664, -0.1253], + [ 0.0584, -0.0046, -0.0547, ..., -0.0546, -0.0357, 0.0164], + ..., + [ 0.0535, -0.0516, -0.0072, ..., 0.0215, -0.0390, 0.0223], + [-0.0538, -0.0308, -0.0794, ..., -0.0179, -0.0336, -0.0617], + [-0.1358, -0.1820, 0.0310, ..., -0.0568, 0.0379, 0.0441]], + device='cuda:0'), grad: tensor([[-9.3132e-09, 0.0000e+00, 3.7253e-08, ..., 0.0000e+00, + 7.4863e-05, 0.0000e+00], + [ 6.7800e-07, 0.0000e+00, 8.7917e-07, ..., 0.0000e+00, + -1.2340e-06, 0.0000e+00], + [ 3.1330e-06, 0.0000e+00, 1.5739e-07, ..., 0.0000e+00, + 3.6694e-07, 0.0000e+00], + ..., + [ 4.0978e-06, 0.0000e+00, 2.1141e-07, ..., 0.0000e+00, + 2.4959e-06, 0.0000e+00], + [-7.8380e-06, 0.0000e+00, 9.4995e-08, ..., 0.0000e+00, + 9.9558e-07, 0.0000e+00], + [-5.6904e-07, 0.0000e+00, 3.1032e-06, ..., 0.0000e+00, + -4.0382e-06, 0.0000e+00]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0050, 0.0217, 0.0031, -0.0003, 0.0118, 0.0042, 0.0246, -0.0124, + 0.0060, -0.0036], device='cuda:0'), grad: tensor([ 9.6560e-04, 2.4959e-06, 1.2040e-05, 1.9390e-06, -4.2245e-06, + 3.2391e-06, -9.8038e-04, 4.6074e-05, -4.3392e-05, -4.9397e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 217.26, cls_loss 0.0032 cls_loss_mapping 0.0034 cls_loss_causal 0.5091 re_mapping 0.0062 re_causal 0.0188 /// teacc 98.98 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0544, -0.0860, -0.0573, ..., -0.0995, -0.0160, -0.0322], + [-0.0497, 0.0399, -0.1558, ..., 0.0569, 0.0667, -0.1261], + [ 0.0580, -0.0057, -0.0549, ..., -0.0547, -0.0361, 0.0138], + ..., + [ 0.0532, -0.0525, -0.0076, ..., 0.0215, -0.0393, 0.0215], + [-0.0537, -0.0309, -0.0796, ..., -0.0179, -0.0338, -0.0619], + [-0.1372, -0.1827, 0.0305, ..., -0.0568, 0.0384, 0.0474]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 0.0000e+00, 4.9360e-08, ..., 5.5879e-09, + 1.6112e-07, 9.3132e-10], + [ 7.4226e-07, 0.0000e+00, 8.8476e-08, ..., -1.3569e-06, + -1.0379e-05, 9.3132e-10], + [ 3.1535e-06, 0.0000e+00, 4.8429e-08, ..., 2.7940e-09, + 1.0338e-07, -2.4214e-08], + ..., + [-5.9530e-06, 0.0000e+00, 3.3155e-07, ..., 9.6392e-07, + 8.4266e-06, 2.0489e-08], + [ 7.8231e-08, 0.0000e+00, -3.7253e-09, ..., 1.3970e-08, + -2.3935e-07, 0.0000e+00], + [-2.8908e-06, 0.0000e+00, 6.4075e-07, ..., 2.2165e-07, + -1.3754e-05, 0.0000e+00]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0074, 0.0221, 0.0024, -0.0004, 0.0120, 0.0045, 0.0251, -0.0127, + 0.0062, -0.0027], device='cuda:0'), grad: tensor([-2.8223e-05, -1.3173e-05, 7.3463e-06, 3.0454e-06, 8.0705e-05, + 1.8431e-06, 6.7241e-06, 8.4639e-06, -3.4757e-06, -6.3300e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 217.34, cls_loss 0.0033 cls_loss_mapping 0.0051 cls_loss_causal 0.5516 re_mapping 0.0072 re_causal 0.0200 /// teacc 98.91 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0548, -0.0890, -0.0563, ..., -0.0997, -0.0162, -0.0323], + [-0.0496, 0.0410, -0.1559, ..., 0.0568, 0.0668, -0.1264], + [ 0.0580, -0.0062, -0.0546, ..., -0.0548, -0.0364, 0.0142], + ..., + [ 0.0566, -0.0531, -0.0078, ..., 0.0215, -0.0381, 0.0211], + [-0.0545, -0.0310, -0.0795, ..., -0.0181, -0.0330, -0.0622], + [-0.1408, -0.1838, 0.0300, ..., -0.0565, 0.0356, 0.0475]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, 3.2596e-08, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 2.6077e-08, 0.0000e+00, 2.6077e-08, ..., 0.0000e+00, + -1.5832e-07, 0.0000e+00], + [-4.2655e-07, 0.0000e+00, 3.0734e-08, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + ..., + [ 4.7404e-07, 0.0000e+00, 3.7253e-08, ..., 0.0000e+00, + 1.3318e-07, 0.0000e+00], + [ 4.3772e-08, 0.0000e+00, 3.6135e-07, ..., 0.0000e+00, + 1.9092e-07, 0.0000e+00], + [ 9.0338e-08, 0.0000e+00, 3.1851e-07, ..., 0.0000e+00, + -1.0151e-07, 0.0000e+00]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0066, 0.0222, 0.0023, -0.0042, 0.0120, 0.0041, 0.0255, -0.0102, + 0.0067, -0.0053], device='cuda:0'), grad: tensor([-3.2317e-07, -2.5425e-07, -3.8464e-07, -4.3213e-07, 1.0058e-07, + -1.5080e-05, 1.3612e-05, 9.4157e-07, 1.3318e-06, 4.6473e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 217.48, cls_loss 0.0021 cls_loss_mapping 0.0037 cls_loss_causal 0.5057 re_mapping 0.0069 re_causal 0.0202 /// teacc 99.00 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0550, -0.0909, -0.0564, ..., -0.0999, -0.0163, -0.0323], + [-0.0499, 0.0418, -0.1561, ..., 0.0568, 0.0668, -0.1268], + [ 0.0580, -0.0065, -0.0556, ..., -0.0548, -0.0367, 0.0136], + ..., + [ 0.0564, -0.0537, -0.0085, ..., 0.0216, -0.0380, 0.0213], + [-0.0557, -0.0313, -0.0794, ..., -0.0182, -0.0336, -0.0627], + [-0.1408, -0.1867, 0.0301, ..., -0.0561, 0.0358, 0.0475]], + device='cuda:0'), grad: tensor([[ 2.1979e-07, 9.3132e-09, 3.7253e-08, ..., 0.0000e+00, + 2.3283e-08, 1.6764e-08], + [-1.0796e-05, 6.5193e-09, 1.0245e-08, ..., 0.0000e+00, + -1.1310e-05, 3.7253e-09], + [-2.8443e-06, -2.1327e-07, 1.3039e-08, ..., 0.0000e+00, + 1.7229e-07, 5.5879e-09], + ..., + [ 1.1265e-05, 2.1700e-07, 5.4948e-08, ..., 0.0000e+00, + 1.1042e-05, 2.5146e-08], + [ 9.1176e-07, 5.5879e-09, 6.6590e-07, ..., 0.0000e+00, + -4.8950e-06, 5.5879e-09], + [ 7.1526e-07, 2.2352e-08, 1.1548e-07, ..., 0.0000e+00, + 5.3085e-08, 5.1223e-08]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0065, 0.0221, 0.0021, -0.0034, 0.0118, 0.0036, 0.0256, -0.0103, + 0.0062, -0.0051], device='cuda:0'), grad: tensor([-8.0094e-07, -2.5913e-05, -3.1013e-06, 2.0325e-05, 3.2689e-07, + -1.8045e-05, 1.4335e-05, 2.5362e-05, -1.4544e-05, 2.0210e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 217.24, cls_loss 0.0029 cls_loss_mapping 0.0048 cls_loss_causal 0.5477 re_mapping 0.0070 re_causal 0.0204 /// teacc 98.82 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0544, -0.0920, -0.0566, ..., -0.1026, -0.0160, -0.0323], + [-0.0502, 0.0422, -0.1562, ..., 0.0568, 0.0670, -0.1272], + [ 0.0581, -0.0055, -0.0555, ..., -0.0556, -0.0371, 0.0135], + ..., + [ 0.0566, -0.0549, -0.0094, ..., 0.0219, -0.0381, 0.0214], + [-0.0569, -0.0312, -0.0802, ..., -0.0186, -0.0344, -0.0627], + [-0.1411, -0.1877, 0.0294, ..., -0.0572, 0.0357, 0.0475]], + device='cuda:0'), grad: tensor([[ 2.7847e-07, 0.0000e+00, 4.3027e-07, ..., 0.0000e+00, + 4.3400e-07, 0.0000e+00], + [ 1.3225e-07, 0.0000e+00, 3.3341e-07, ..., 3.7253e-09, + -2.3376e-07, 0.0000e+00], + [-9.7323e-07, 0.0000e+00, 3.4831e-07, ..., 0.0000e+00, + 1.1399e-06, 0.0000e+00], + ..., + [ 7.5251e-07, 0.0000e+00, 1.2005e-06, ..., -5.5879e-09, + 2.8759e-06, 0.0000e+00], + [ 1.5711e-06, 0.0000e+00, 8.8010e-07, ..., 0.0000e+00, + 2.6412e-06, 0.0000e+00], + [-2.1495e-06, 0.0000e+00, -1.5972e-06, ..., 9.3132e-10, + -9.2760e-06, 0.0000e+00]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0061, 0.0220, 0.0020, -0.0034, 0.0120, 0.0038, 0.0250, -0.0102, + 0.0054, -0.0053], device='cuda:0'), grad: tensor([ 3.7812e-06, 1.0710e-07, 1.7788e-06, 1.1902e-06, 5.0850e-07, + -2.4047e-06, 1.5311e-06, 1.0528e-05, 1.1221e-05, -2.8238e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 217.79, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5195 re_mapping 0.0068 re_causal 0.0195 /// teacc 98.89 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0545, -0.0928, -0.0553, ..., -0.1031, -0.0161, -0.0323], + [-0.0503, 0.0431, -0.1564, ..., 0.0567, 0.0671, -0.1274], + [ 0.0588, -0.0058, -0.0556, ..., -0.0570, -0.0373, 0.0136], + ..., + [ 0.0566, -0.0556, -0.0097, ..., 0.0219, -0.0382, 0.0213], + [-0.0583, -0.0305, -0.0802, ..., -0.0189, -0.0337, -0.0628], + [-0.1410, -0.1885, 0.0290, ..., -0.0563, 0.0361, 0.0475]], + device='cuda:0'), grad: tensor([[ 1.4994e-07, 0.0000e+00, -4.2841e-08, ..., 6.6124e-08, + 8.4750e-08, 0.0000e+00], + [ 2.0731e-06, 0.0000e+00, 3.2596e-08, ..., 1.6810e-06, + -1.0934e-06, 0.0000e+00], + [ 6.6776e-07, 0.0000e+00, 2.1420e-08, ..., 3.0734e-08, + 7.6462e-07, 0.0000e+00], + ..., + [-4.1425e-06, 0.0000e+00, 5.1223e-08, ..., -2.5332e-06, + 2.3190e-07, 0.0000e+00], + [ 8.0094e-08, 0.0000e+00, 8.2888e-08, ..., 2.9802e-08, + 6.8918e-08, 0.0000e+00], + [ 5.2620e-07, 0.0000e+00, 8.8196e-07, ..., 4.2468e-07, + -2.0862e-07, 0.0000e+00]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0056, 0.0220, 0.0027, -0.0043, 0.0120, 0.0060, 0.0236, -0.0103, + 0.0055, -0.0054], device='cuda:0'), grad: tensor([ 1.6671e-07, 5.0887e-06, 2.9076e-06, 1.7481e-06, -1.5274e-07, + -2.8033e-07, -1.3877e-06, -1.1422e-05, 4.6659e-07, 2.8554e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 217.34, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5266 re_mapping 0.0068 re_causal 0.0198 /// teacc 98.95 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0548, -0.0931, -0.0579, ..., -0.1046, -0.0160, -0.0323], + [-0.0504, 0.0432, -0.1566, ..., 0.0568, 0.0672, -0.1276], + [ 0.0590, -0.0058, -0.0560, ..., -0.0575, -0.0375, 0.0136], + ..., + [ 0.0565, -0.0556, -0.0118, ..., 0.0215, -0.0383, 0.0212], + [-0.0586, -0.0308, -0.0785, ..., -0.0201, -0.0335, -0.0629], + [-0.1411, -0.1887, 0.0287, ..., -0.0570, 0.0361, 0.0475]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -1.9092e-07, ..., 0.0000e+00, + 2.3097e-07, 0.0000e+00], + [ 6.1467e-08, 0.0000e+00, 1.2480e-07, ..., 0.0000e+00, + -6.1281e-07, 0.0000e+00], + [-8.5682e-08, 0.0000e+00, 5.2154e-08, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00], + ..., + [ 2.1141e-07, 0.0000e+00, 2.0843e-06, ..., 0.0000e+00, + 1.1548e-07, 9.3132e-10], + [-3.3602e-06, 0.0000e+00, 6.7987e-08, ..., 0.0000e+00, + 2.4084e-06, 0.0000e+00], + [ 3.0063e-06, 0.0000e+00, 3.0279e-05, ..., 0.0000e+00, + -2.7940e-08, -3.7253e-09]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0056, 0.0219, 0.0027, -0.0043, 0.0123, 0.0053, 0.0230, -0.0104, + 0.0073, -0.0055], device='cuda:0'), grad: tensor([-1.3756e-06, 5.6140e-06, 1.0943e-06, 7.5251e-07, -5.7906e-05, + 2.9095e-06, -1.0684e-05, 6.6906e-06, -2.1085e-05, 7.3910e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 217.74, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.5201 re_mapping 0.0067 re_causal 0.0185 /// teacc 99.01 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0551, -0.0933, -0.0580, ..., -0.1058, -0.0160, -0.0323], + [-0.0503, 0.0434, -0.1567, ..., 0.0578, 0.0675, -0.1280], + [ 0.0588, -0.0061, -0.0567, ..., -0.0578, -0.0380, 0.0134], + ..., + [ 0.0566, -0.0555, -0.0122, ..., 0.0207, -0.0386, 0.0206], + [-0.0596, -0.0312, -0.0787, ..., -0.0206, -0.0336, -0.0637], + [-0.1412, -0.1891, 0.0284, ..., -0.0587, 0.0364, 0.0476]], + device='cuda:0'), grad: tensor([[ 3.4273e-07, 0.0000e+00, 6.4261e-08, ..., 0.0000e+00, + 1.7043e-07, 0.0000e+00], + [ 6.6496e-07, 0.0000e+00, 1.2107e-07, ..., 0.0000e+00, + -2.0396e-07, 0.0000e+00], + [-1.7926e-05, 0.0000e+00, 1.6112e-07, ..., 0.0000e+00, + 2.8871e-08, 0.0000e+00], + ..., + [ 1.2584e-05, 0.0000e+00, 1.0263e-06, ..., 0.0000e+00, + 1.8226e-06, 4.8429e-08], + [ 4.0829e-06, 0.0000e+00, 1.2945e-07, ..., 0.0000e+00, + 7.9535e-07, 0.0000e+00], + [ 3.6135e-07, 0.0000e+00, -3.2820e-06, ..., 0.0000e+00, + -2.6561e-06, -5.7742e-08]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0055, 0.0221, 0.0024, -0.0041, 0.0121, 0.0050, 0.0241, -0.0106, + 0.0072, -0.0053], device='cuda:0'), grad: tensor([ 2.3600e-06, 1.3169e-06, -2.9624e-05, 9.9465e-07, 1.6585e-05, + 8.2105e-06, -9.0748e-06, 3.7760e-05, 1.7405e-05, -4.5925e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 217.68, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5198 re_mapping 0.0063 re_causal 0.0189 /// teacc 98.92 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0553, -0.0941, -0.0582, ..., -0.1067, -0.0162, -0.0323], + [-0.0504, 0.0444, -0.1573, ..., 0.0578, 0.0676, -0.1281], + [ 0.0593, -0.0071, -0.0552, ..., -0.0579, -0.0369, 0.0134], + ..., + [ 0.0565, -0.0558, -0.0124, ..., 0.0207, -0.0390, 0.0195], + [-0.0602, -0.0324, -0.0787, ..., -0.0209, -0.0323, -0.0639], + [-0.1407, -0.1899, 0.0281, ..., -0.0607, 0.0369, 0.0476]], + device='cuda:0'), grad: tensor([[ 5.5507e-07, 0.0000e+00, 3.4925e-07, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + [ 7.9349e-07, 0.0000e+00, 6.1467e-08, ..., 0.0000e+00, + 5.3085e-08, 0.0000e+00], + [-1.7434e-05, 0.0000e+00, 8.5030e-07, ..., 0.0000e+00, + 1.3225e-07, 0.0000e+00], + ..., + [-4.6659e-07, 0.0000e+00, 5.4948e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.7984e-06, 2.7940e-09, 1.5832e-07, ..., 0.0000e+00, + 8.5682e-08, 0.0000e+00], + [ 2.0638e-06, 0.0000e+00, 5.6997e-06, ..., 0.0000e+00, + -2.4773e-07, 0.0000e+00]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0055, 0.0220, 0.0032, -0.0042, 0.0124, 0.0048, 0.0240, -0.0109, + 0.0081, -0.0052], device='cuda:0'), grad: tensor([ 1.1846e-06, 1.5805e-06, -4.0084e-05, 3.7253e-05, -1.7092e-05, + 2.5891e-07, 1.2200e-07, -7.5344e-07, 3.8780e-06, 1.3664e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 217.32, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5193 re_mapping 0.0064 re_causal 0.0192 /// teacc 98.91 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0554, -0.0953, -0.0581, ..., -0.1068, -0.0164, -0.0323], + [-0.0505, 0.0446, -0.1574, ..., 0.0593, 0.0681, -0.1284], + [ 0.0592, -0.0063, -0.0565, ..., -0.0582, -0.0371, 0.0127], + ..., + [ 0.0565, -0.0565, -0.0119, ..., 0.0196, -0.0394, 0.0195], + [-0.0607, -0.0328, -0.0795, ..., -0.0212, -0.0324, -0.0647], + [-0.1407, -0.1907, 0.0285, ..., -0.0607, 0.0370, 0.0476]], + device='cuda:0'), grad: tensor([[ 1.5339e-06, 9.2201e-08, 4.7497e-08, ..., 1.8626e-09, + 1.5553e-07, 3.7253e-09], + [-5.2340e-07, -3.1143e-06, 3.9861e-07, ..., 4.0978e-08, + -5.1856e-06, 2.0489e-08], + [-2.5079e-05, 6.6031e-07, 5.6811e-08, ..., 3.7253e-09, + 1.0421e-06, 8.3819e-09], + ..., + [ 3.1106e-06, 1.7853e-06, 7.7952e-07, ..., 3.7253e-09, + 2.9616e-06, 2.7940e-08], + [ 1.5395e-06, 1.9558e-07, 7.8231e-08, ..., 0.0000e+00, + 3.1292e-07, 4.9360e-08], + [ 1.0505e-06, 1.3970e-07, 2.2426e-06, ..., 1.1828e-07, + 2.2911e-07, 6.5193e-08]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0055, 0.0223, 0.0030, -0.0042, 0.0123, 0.0047, 0.0245, -0.0111, + 0.0076, -0.0052], device='cuda:0'), grad: tensor([ 1.1865e-06, -8.8587e-06, -4.2409e-05, -5.9485e-05, -9.2685e-06, + 9.1076e-05, 2.1942e-06, 1.2279e-05, 3.5465e-06, 9.6038e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 217.87, cls_loss 0.0019 cls_loss_mapping 0.0027 cls_loss_causal 0.5406 re_mapping 0.0064 re_causal 0.0199 /// teacc 98.99 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0557, -0.0962, -0.0581, ..., -0.1073, -0.0163, -0.0323], + [-0.0505, 0.0448, -0.1578, ..., 0.0610, 0.0685, -0.1292], + [ 0.0596, -0.0063, -0.0563, ..., -0.0582, -0.0371, 0.0127], + ..., + [ 0.0565, -0.0566, -0.0121, ..., 0.0180, -0.0397, 0.0198], + [-0.0611, -0.0326, -0.0796, ..., -0.0214, -0.0332, -0.0649], + [-0.1408, -0.1916, 0.0284, ..., -0.0611, 0.0370, 0.0476]], + device='cuda:0'), grad: tensor([[ 4.8429e-08, 3.4273e-07, 1.2852e-07, ..., 2.7940e-09, + 1.7975e-07, 0.0000e+00], + [-1.0589e-06, -1.9640e-05, 2.3283e-07, ..., 6.5193e-09, + -4.0047e-07, 0.0000e+00], + [ 2.2165e-07, 1.2331e-05, 1.1455e-07, ..., 3.7253e-09, + 3.2783e-07, 0.0000e+00], + ..., + [ 2.0582e-07, 7.9721e-07, 2.6450e-07, ..., 5.1223e-08, + 1.2740e-06, 0.0000e+00], + [ 8.1025e-08, 1.8906e-07, 2.5053e-07, ..., 9.3132e-10, + -2.5108e-06, 0.0000e+00], + [-5.1223e-08, 2.9802e-08, 2.7698e-06, ..., 4.6566e-09, + -2.4009e-06, 0.0000e+00]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0052, 0.0227, 0.0035, -0.0040, 0.0123, 0.0033, 0.0253, -0.0114, + 0.0076, -0.0052], device='cuda:0'), grad: tensor([ 2.0191e-06, -7.1228e-05, 4.4823e-05, 4.6268e-06, 6.1393e-06, + 1.7136e-05, 7.8678e-06, 8.6725e-06, -1.5676e-05, -4.4778e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 217.83, cls_loss 0.0023 cls_loss_mapping 0.0035 cls_loss_causal 0.5231 re_mapping 0.0063 re_causal 0.0190 /// teacc 98.96 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0559, -0.0985, -0.0588, ..., -0.1076, -0.0170, -0.0324], + [-0.0501, 0.0477, -0.1580, ..., 0.0612, 0.0687, -0.1306], + [ 0.0588, -0.0090, -0.0593, ..., -0.0583, -0.0379, 0.0126], + ..., + [ 0.0566, -0.0575, -0.0122, ..., 0.0179, -0.0396, 0.0192], + [-0.0611, -0.0319, -0.0797, ..., -0.0215, -0.0339, -0.0655], + [-0.1412, -0.1933, 0.0307, ..., -0.0612, 0.0371, 0.0478]], + device='cuda:0'), grad: tensor([[ 1.9372e-07, 0.0000e+00, 7.4506e-08, ..., 0.0000e+00, + 1.6019e-07, 1.6578e-07], + [ 4.8056e-07, 0.0000e+00, 4.1071e-07, ..., 0.0000e+00, + -8.9221e-07, 2.7940e-09], + [ 1.8701e-06, 0.0000e+00, 1.8813e-07, ..., 0.0000e+00, + 1.2945e-07, 1.6764e-08], + ..., + [ 4.8243e-06, 0.0000e+00, 7.8976e-07, ..., 0.0000e+00, + 8.7265e-07, 0.0000e+00], + [ 7.1060e-07, 0.0000e+00, 3.9302e-07, ..., 0.0000e+00, + -4.9919e-07, 1.6484e-07], + [ 1.2619e-06, 0.0000e+00, 1.7649e-06, ..., 0.0000e+00, + 4.1351e-07, 7.4506e-09]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0056, 0.0231, 0.0017, -0.0038, 0.0125, 0.0014, 0.0254, -0.0113, + 0.0074, -0.0046], device='cuda:0'), grad: tensor([ 2.2724e-06, -1.3318e-07, 2.1532e-06, -1.0058e-05, -4.9174e-06, + -1.0151e-07, -4.2915e-06, 8.5682e-06, -2.0396e-07, 6.6534e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 190---------------------------------------------------- +epoch 190, time 218.54, cls_loss 0.0026 cls_loss_mapping 0.0034 cls_loss_causal 0.5075 re_mapping 0.0065 re_causal 0.0187 /// teacc 99.13 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0564, -0.1009, -0.0593, ..., -0.1078, -0.0182, -0.0324], + [-0.0500, 0.0478, -0.1583, ..., 0.0612, 0.0689, -0.1308], + [ 0.0549, -0.0103, -0.0633, ..., -0.0583, -0.0407, 0.0126], + ..., + [ 0.0572, -0.0554, -0.0125, ..., 0.0179, -0.0394, 0.0192], + [-0.0620, -0.0329, -0.0800, ..., -0.0216, -0.0335, -0.0655], + [-0.1417, -0.1968, 0.0305, ..., -0.0612, 0.0378, 0.0478]], + device='cuda:0'), grad: tensor([[ 1.4715e-07, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 5.8673e-08, 0.0000e+00], + [-9.4064e-08, 9.3132e-10, 5.5879e-09, ..., 0.0000e+00, + -3.1013e-06, 0.0000e+00], + [ 6.9570e-07, 9.3132e-10, 1.8626e-09, ..., 0.0000e+00, + 8.5682e-07, 0.0000e+00], + ..., + [ 2.1327e-07, -1.7695e-08, 7.4506e-09, ..., 0.0000e+00, + 1.7565e-06, 0.0000e+00], + [-5.7742e-08, 1.8626e-09, 1.6287e-05, ..., 0.0000e+00, + 4.7088e-06, 0.0000e+00], + [ 8.6613e-08, 1.8626e-09, -1.7077e-05, ..., 0.0000e+00, + -5.1893e-06, 0.0000e+00]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0063, 0.0232, -0.0031, -0.0041, 0.0126, 0.0044, 0.0250, -0.0107, + 0.0071, -0.0044], device='cuda:0'), grad: tensor([-8.1584e-07, -7.7188e-06, 2.5351e-06, -2.5053e-06, 2.0508e-06, + 2.9467e-06, 2.2277e-06, 5.1036e-06, 3.5316e-05, -3.9220e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 217.78, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4607 re_mapping 0.0063 re_causal 0.0180 /// teacc 98.80 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0561, -0.1010, -0.0593, ..., -0.1079, -0.0184, -0.0324], + [-0.0504, 0.0478, -0.1585, ..., 0.0612, 0.0689, -0.1308], + [ 0.0550, -0.0103, -0.0634, ..., -0.0583, -0.0411, 0.0126], + ..., + [ 0.0574, -0.0554, -0.0136, ..., 0.0179, -0.0394, 0.0192], + [-0.0623, -0.0326, -0.0802, ..., -0.0217, -0.0336, -0.0656], + [-0.1419, -0.1969, 0.0305, ..., -0.0612, 0.0380, 0.0478]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, 0.0000e+00, 6.2399e-08, ..., 0.0000e+00, + 3.9116e-08, 0.0000e+00], + [ 1.1455e-07, 0.0000e+00, 1.1548e-07, ..., 9.3132e-10, + -2.7940e-07, 0.0000e+00], + [-2.2277e-06, 0.0000e+00, 5.7742e-08, ..., 0.0000e+00, + 3.0734e-08, 0.0000e+00], + ..., + [ 4.8578e-06, -9.3132e-10, 1.5274e-07, ..., -3.7253e-09, + 2.7101e-07, 0.0000e+00], + [ 1.0421e-06, 0.0000e+00, 9.4716e-07, ..., 0.0000e+00, + 6.7335e-07, 0.0000e+00], + [-2.3469e-06, 0.0000e+00, -8.0541e-06, ..., 9.3132e-10, + -6.3926e-06, 0.0000e+00]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0063, 0.0229, -0.0032, -0.0041, 0.0125, 0.0045, 0.0251, -0.0106, + 0.0069, -0.0043], device='cuda:0'), grad: tensor([ 1.2601e-06, -1.0058e-07, -2.3302e-06, 4.3511e-05, -9.2667e-07, + 1.4929e-06, 2.5891e-07, 7.1712e-06, 5.6811e-06, -5.6058e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 217.55, cls_loss 0.0022 cls_loss_mapping 0.0027 cls_loss_causal 0.4963 re_mapping 0.0064 re_causal 0.0185 /// teacc 99.02 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0559, -0.1018, -0.0595, ..., -0.1079, -0.0187, -0.0324], + [-0.0507, 0.0484, -0.1586, ..., 0.0612, 0.0690, -0.1311], + [ 0.0552, -0.0105, -0.0633, ..., -0.0583, -0.0413, 0.0129], + ..., + [ 0.0573, -0.0561, -0.0134, ..., 0.0180, -0.0395, 0.0189], + [-0.0629, -0.0313, -0.0803, ..., -0.0219, -0.0337, -0.0657], + [-0.1417, -0.1972, 0.0308, ..., -0.0612, 0.0386, 0.0478]], + device='cuda:0'), grad: tensor([[ 6.6124e-08, 0.0000e+00, -1.5749e-06, ..., 0.0000e+00, + 3.7253e-08, 2.7940e-09], + [ 4.4666e-06, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 9.4716e-07, 4.6566e-08], + [-1.3016e-05, 0.0000e+00, 1.2107e-08, ..., 0.0000e+00, + -5.3458e-06, 1.8626e-08], + ..., + [-8.2105e-06, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + -1.1446e-06, -2.3935e-07], + [ 1.4435e-07, 0.0000e+00, 8.3819e-08, ..., 0.0000e+00, + 1.4994e-07, 8.3819e-09], + [ 1.2061e-06, 0.0000e+00, 4.7218e-07, ..., 0.0000e+00, + -1.2759e-07, 5.5879e-08]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0071, 0.0228, -0.0029, -0.0041, 0.0121, 0.0047, 0.0248, -0.0108, + 0.0068, -0.0035], device='cuda:0'), grad: tensor([-8.2403e-06, 1.9744e-05, -8.9824e-05, 6.2846e-06, 4.1910e-06, + 1.0291e-06, 9.0480e-05, -2.6152e-05, 2.3749e-07, 2.1402e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 217.64, cls_loss 0.0020 cls_loss_mapping 0.0040 cls_loss_causal 0.5150 re_mapping 0.0060 re_causal 0.0186 /// teacc 98.87 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0560, -0.1023, -0.0593, ..., -0.1092, -0.0187, -0.0324], + [-0.0494, 0.0485, -0.1588, ..., 0.0612, 0.0702, -0.1329], + [ 0.0555, -0.0100, -0.0633, ..., -0.0587, -0.0413, 0.0128], + ..., + [ 0.0564, -0.0569, -0.0131, ..., 0.0187, -0.0407, 0.0198], + [-0.0635, -0.0286, -0.0805, ..., -0.0237, -0.0343, -0.0658], + [-0.1417, -0.1975, 0.0304, ..., -0.0614, 0.0388, 0.0478]], + device='cuda:0'), grad: tensor([[ 3.0734e-08, 0.0000e+00, 4.4238e-08, ..., 1.1176e-08, + 1.5907e-06, 0.0000e+00], + [ 5.3504e-07, 9.3132e-10, 1.5181e-07, ..., 2.6077e-08, + -6.4790e-05, 0.0000e+00], + [ 1.3132e-07, 0.0000e+00, 4.2375e-08, ..., 1.0710e-08, + 3.1441e-05, 0.0000e+00], + ..., + [-1.4473e-06, -1.8626e-09, 2.0768e-07, ..., 5.3551e-08, + 1.1120e-06, 0.0000e+00], + [ 9.9652e-08, 0.0000e+00, 5.8208e-08, ..., 5.1223e-09, + 2.7835e-05, 0.0000e+00], + [ 2.9476e-07, 0.0000e+00, 1.7919e-06, ..., 3.1339e-07, + 8.3353e-08, 0.0000e+00]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0069, 0.0242, -0.0026, -0.0042, 0.0122, 0.0045, 0.0248, -0.0119, + 0.0071, -0.0037], device='cuda:0'), grad: tensor([ 3.6284e-06, -1.4067e-04, 6.9082e-05, 3.7532e-07, -6.2436e-06, + 1.4976e-06, 5.6848e-06, 7.6927e-07, 6.1214e-05, 4.6268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 217.62, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.5276 re_mapping 0.0060 re_causal 0.0191 /// teacc 98.90 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0561, -0.1033, -0.0594, ..., -0.1103, -0.0213, -0.0324], + [-0.0495, 0.0485, -0.1590, ..., 0.0612, 0.0703, -0.1334], + [ 0.0563, -0.0076, -0.0632, ..., -0.0588, -0.0417, 0.0128], + ..., + [ 0.0562, -0.0595, -0.0136, ..., 0.0187, -0.0407, 0.0200], + [-0.0639, -0.0286, -0.0806, ..., -0.0242, -0.0348, -0.0658], + [-0.1419, -0.1980, 0.0303, ..., -0.0613, 0.0401, 0.0478]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, 0.0000e+00, 3.4925e-08, ..., 8.3819e-09, + 1.4529e-07, 0.0000e+00], + [ 3.1758e-07, 0.0000e+00, 9.6858e-08, ..., 1.7695e-08, + -3.7719e-08, 0.0000e+00], + [-5.3868e-06, 0.0000e+00, -8.4145e-07, ..., 1.8626e-08, + 1.7695e-08, 0.0000e+00], + ..., + [-4.7218e-07, 0.0000e+00, 3.5018e-07, ..., 1.0291e-07, + 3.1199e-08, 0.0000e+00], + [ 6.6124e-08, 0.0000e+00, -3.3341e-07, ..., 1.8626e-09, + -3.5390e-08, 0.0000e+00], + [ 5.2154e-08, 0.0000e+00, 5.9092e-07, ..., 1.1642e-08, + -2.6450e-07, 0.0000e+00]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0090, 0.0242, -0.0018, -0.0041, 0.0124, 0.0045, 0.0242, -0.0122, + 0.0069, -0.0023], device='cuda:0'), grad: tensor([ 8.5980e-06, 1.0505e-06, -6.4299e-06, 1.0617e-06, -1.9446e-06, + 8.2701e-06, -2.0280e-05, -5.0105e-07, 8.1286e-06, 2.0191e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 218.00, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.5352 re_mapping 0.0064 re_causal 0.0190 /// teacc 98.94 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0559, -0.1038, -0.0593, ..., -0.1106, -0.0215, -0.0324], + [-0.0495, 0.0485, -0.1592, ..., 0.0611, 0.0705, -0.1337], + [ 0.0567, -0.0072, -0.0629, ..., -0.0589, -0.0419, 0.0133], + ..., + [ 0.0562, -0.0598, -0.0135, ..., 0.0187, -0.0408, 0.0193], + [-0.0646, -0.0285, -0.0807, ..., -0.0245, -0.0362, -0.0660], + [-0.1419, -0.1988, 0.0303, ..., -0.0610, 0.0402, 0.0478]], + device='cuda:0'), grad: tensor([[-1.1688e-07, 0.0000e+00, 3.8883e-07, ..., 1.8626e-09, + 6.7987e-08, 0.0000e+00], + [ 3.2177e-07, 0.0000e+00, 2.2305e-07, ..., 9.7789e-09, + -8.3353e-08, 0.0000e+00], + [ 7.3528e-07, 0.0000e+00, 1.4482e-07, ..., 2.3283e-09, + 1.9558e-07, 0.0000e+00], + ..., + [ 8.3167e-07, 0.0000e+00, 7.4971e-07, ..., 4.7963e-08, + 1.1129e-07, 0.0000e+00], + [ 5.8115e-07, 0.0000e+00, 1.7490e-06, ..., 4.6566e-10, + 5.6811e-08, 0.0000e+00], + [ 1.5358e-06, 0.0000e+00, 4.4517e-06, ..., 1.3970e-08, + 1.0617e-07, 0.0000e+00]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0087, 0.0243, -0.0016, -0.0041, 0.0123, 0.0043, 0.0238, -0.0123, + 0.0063, -0.0023], device='cuda:0'), grad: tensor([-1.9390e-06, 4.6426e-07, 2.5649e-06, 7.7128e-05, 1.9874e-06, + -9.3937e-05, -6.4820e-06, 2.4643e-06, 4.3772e-06, 1.3426e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 217.69, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5102 re_mapping 0.0061 re_causal 0.0183 /// teacc 98.95 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0553, -0.1042, -0.0594, ..., -0.1115, -0.0203, -0.0324], + [-0.0496, 0.0486, -0.1594, ..., 0.0614, 0.0704, -0.1337], + [ 0.0570, -0.0072, -0.0629, ..., -0.0589, -0.0423, 0.0133], + ..., + [ 0.0562, -0.0599, -0.0138, ..., 0.0184, -0.0408, 0.0193], + [-0.0661, -0.0282, -0.0810, ..., -0.0248, -0.0366, -0.0660], + [-0.1421, -0.1994, 0.0307, ..., -0.0607, 0.0405, 0.0478]], + device='cuda:0'), grad: tensor([[-4.8168e-06, 0.0000e+00, 1.0710e-08, ..., 0.0000e+00, + -5.8264e-06, 0.0000e+00], + [ 2.8554e-06, 9.3132e-10, 8.2888e-08, ..., 4.6566e-10, + 1.5646e-06, 0.0000e+00], + [ 3.4273e-06, -4.1910e-09, 1.2573e-08, ..., 0.0000e+00, + 6.5286e-07, 0.0000e+00], + ..., + [-4.3064e-06, 2.7940e-09, 4.4703e-08, ..., 9.3132e-10, + 3.6322e-07, 0.0000e+00], + [ 4.5309e-07, 0.0000e+00, 1.7229e-08, ..., 0.0000e+00, + 1.2247e-07, 4.6566e-10], + [ 2.3469e-07, 0.0000e+00, 2.4401e-07, ..., 2.3283e-09, + 2.3423e-07, 0.0000e+00]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0077, 0.0241, -0.0015, -0.0044, 0.0123, 0.0042, 0.0239, -0.0124, + 0.0050, -0.0023], device='cuda:0'), grad: tensor([-3.3975e-05, 1.3359e-05, 7.4357e-06, 1.5944e-06, 3.2457e-07, + -1.4855e-07, 1.3851e-05, -5.1409e-06, 9.0431e-07, 1.8040e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 217.76, cls_loss 0.0023 cls_loss_mapping 0.0036 cls_loss_causal 0.5214 re_mapping 0.0064 re_causal 0.0179 /// teacc 98.89 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0551, -0.1055, -0.0586, ..., -0.1120, -0.0202, -0.0324], + [-0.0494, 0.0486, -0.1595, ..., 0.0612, 0.0706, -0.1345], + [ 0.0567, -0.0073, -0.0626, ..., -0.0592, -0.0431, 0.0136], + ..., + [ 0.0563, -0.0597, -0.0141, ..., 0.0190, -0.0409, 0.0195], + [-0.0657, -0.0283, -0.0811, ..., -0.0251, -0.0361, -0.0663], + [-0.1426, -0.2014, 0.0302, ..., -0.0626, 0.0406, 0.0478]], + device='cuda:0'), grad: tensor([[ 7.3668e-07, 9.3132e-10, 1.8813e-07, ..., 0.0000e+00, + 2.0210e-07, 1.6764e-08], + [ 3.2410e-06, 2.5146e-08, 7.4226e-07, ..., 9.3132e-10, + 2.8964e-07, 7.4506e-09], + [ 4.4592e-06, 1.7695e-08, 9.4436e-07, ..., 9.3132e-10, + 7.5996e-07, 1.1176e-08], + ..., + [-2.0385e-05, -6.7055e-08, 1.2452e-06, ..., -1.2107e-08, + -4.1537e-06, 2.3842e-07], + [-1.3702e-05, 9.3132e-09, -6.6943e-06, ..., 0.0000e+00, + -1.0058e-05, 9.3132e-09], + [ 1.7971e-05, 7.4506e-09, 8.9128e-07, ..., 0.0000e+00, + 1.1697e-05, -2.9430e-07]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0066, 0.0242, -0.0018, -0.0045, 0.0124, 0.0046, 0.0243, -0.0122, + 0.0054, -0.0036], device='cuda:0'), grad: tensor([ 2.0191e-06, 9.6858e-06, 1.3098e-05, 1.1854e-05, 1.3299e-05, + 8.8010e-07, -1.8626e-08, -5.0873e-05, -9.9599e-05, 9.9659e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 217.76, cls_loss 0.0025 cls_loss_mapping 0.0038 cls_loss_causal 0.5459 re_mapping 0.0061 re_causal 0.0182 /// teacc 98.90 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0557, -0.1071, -0.0590, ..., -0.1147, -0.0204, -0.0324], + [-0.0494, 0.0491, -0.1600, ..., 0.0614, 0.0709, -0.1355], + [ 0.0565, -0.0077, -0.0625, ..., -0.0595, -0.0434, 0.0117], + ..., + [ 0.0565, -0.0596, -0.0155, ..., 0.0187, -0.0410, 0.0189], + [-0.0661, -0.0266, -0.0825, ..., -0.0281, -0.0365, -0.0693], + [-0.1430, -0.2037, 0.0285, ..., -0.0653, 0.0397, 0.0478]], + device='cuda:0'), grad: tensor([[ 5.6811e-08, 9.3132e-10, 3.4459e-08, ..., 1.4622e-07, + 6.3330e-07, 2.7940e-09], + [ 6.2399e-07, 9.3132e-09, 1.8626e-08, ..., -4.4346e-05, + -5.9754e-05, 2.8871e-08], + [-1.0710e-06, 2.7940e-09, 1.8626e-09, ..., 1.0431e-07, + 2.7847e-07, 7.4506e-09], + ..., + [ 5.6531e-07, 7.4506e-09, 2.7940e-08, ..., 3.6180e-05, + 4.7743e-05, 2.3283e-08], + [ 2.6356e-07, 2.0489e-08, 1.4808e-07, ..., 2.7940e-09, + 2.4121e-07, 5.5879e-08], + [ 2.1979e-07, 1.6764e-08, 1.3784e-07, ..., 4.5113e-06, + 5.9046e-06, 5.1223e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0068, 0.0244, -0.0021, -0.0049, 0.0147, 0.0054, 0.0251, -0.0122, + 0.0047, -0.0056], device='cuda:0'), grad: tensor([ 1.7108e-06, -2.0659e-04, -7.8697e-07, -6.5029e-05, 1.6347e-05, + 6.3658e-05, 5.0478e-07, 1.6820e-04, 1.1297e-06, 2.1100e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 217.80, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.5363 re_mapping 0.0064 re_causal 0.0190 /// teacc 98.85 lr 0.00001000 +Epoch 201, weight, value: tensor([[-0.0558, -0.1075, -0.0597, ..., -0.1150, -0.0204, -0.0324], + [-0.0494, 0.0489, -0.1612, ..., 0.0615, 0.0709, -0.1365], + [ 0.0563, -0.0076, -0.0633, ..., -0.0596, -0.0440, 0.0089], + ..., + [ 0.0566, -0.0597, -0.0167, ..., 0.0187, -0.0411, 0.0182], + [-0.0664, -0.0264, -0.0824, ..., -0.0285, -0.0351, -0.0708], + [-0.1432, -0.2039, 0.0286, ..., -0.0658, 0.0400, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.5739e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.0338e-08, 0.0000e+00], + [ 4.1984e-06, 0.0000e+00, 3.4180e-07, ..., 0.0000e+00, + 1.9372e-06, 0.0000e+00], + [ 2.9579e-06, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 2.7195e-06, 0.0000e+00], + ..., + [-4.7237e-05, 0.0000e+00, 3.6992e-06, ..., 0.0000e+00, + -2.3842e-05, -9.3132e-10], + [ 5.6103e-06, 0.0000e+00, 5.3570e-06, ..., 0.0000e+00, + 3.3602e-06, 9.3132e-10], + [ 3.3200e-05, 0.0000e+00, -2.1994e-05, ..., 0.0000e+00, + 1.4156e-05, 0.0000e+00]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0068, 0.0244, -0.0026, -0.0052, 0.0149, 0.0056, 0.0250, -0.0123, + 0.0057, -0.0054], device='cuda:0'), grad: tensor([ 1.4668e-06, 3.3379e-05, 4.2379e-05, 1.5264e-06, 3.7491e-05, + 9.4026e-06, 9.5926e-08, -4.4703e-04, 7.0095e-05, 2.5082e-04], + device='cuda:0') +100 +1e-05 +changing lr +epoch 200, time 217.27, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5173 re_mapping 0.0064 re_causal 0.0187 /// teacc 98.95 lr 0.00001000 +Epoch 202, weight, value: tensor([[-0.0558, -0.1076, -0.0598, ..., -0.1151, -0.0204, -0.0325], + [-0.0495, 0.0490, -0.1613, ..., 0.0615, 0.0709, -0.1365], + [ 0.0563, -0.0076, -0.0633, ..., -0.0596, -0.0440, 0.0089], + ..., + [ 0.0566, -0.0597, -0.0167, ..., 0.0187, -0.0411, 0.0181], + [-0.0664, -0.0263, -0.0824, ..., -0.0285, -0.0350, -0.0709], + [-0.1433, -0.2039, 0.0287, ..., -0.0659, 0.0400, 0.0480]], + device='cuda:0'), grad: tensor([[ 7.7300e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.7695e-08, 0.0000e+00], + [ 1.8533e-07, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + -4.2841e-08, 0.0000e+00], + [-4.6752e-07, 0.0000e+00, -1.2107e-08, ..., 0.0000e+00, + 1.9558e-08, 0.0000e+00], + ..., + [ 1.5115e-06, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 3.6787e-07, 0.0000e+00], + [ 7.6368e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-08, 0.0000e+00], + [ 1.4808e-07, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -7.7114e-07, 0.0000e+00]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0068, 0.0244, -0.0026, -0.0051, 0.0148, 0.0056, 0.0250, -0.0123, + 0.0058, -0.0054], device='cuda:0'), grad: tensor([-2.4475e-06, 6.0629e-07, -1.9027e-06, -1.8049e-06, 1.6913e-06, + 2.6543e-07, -1.1642e-07, 2.5313e-06, 5.5693e-07, 6.2771e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 201, time 217.37, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5187 re_mapping 0.0058 re_causal 0.0183 /// teacc 98.99 lr 0.00001000 +Epoch 203, weight, value: tensor([[-0.0558, -0.1076, -0.0598, ..., -0.1151, -0.0204, -0.0325], + [-0.0496, 0.0490, -0.1613, ..., 0.0615, 0.0709, -0.1366], + [ 0.0563, -0.0076, -0.0633, ..., -0.0596, -0.0441, 0.0089], + ..., + [ 0.0567, -0.0597, -0.0168, ..., 0.0187, -0.0411, 0.0181], + [-0.0664, -0.0263, -0.0824, ..., -0.0285, -0.0350, -0.0709], + [-0.1433, -0.2040, 0.0286, ..., -0.0659, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.7323e-07, 9.3132e-10, 6.5193e-09, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 4.9453e-07, -1.0338e-07, 7.1153e-07, ..., 0.0000e+00, + -8.5682e-08, 0.0000e+00], + [-5.2713e-06, 6.7987e-08, 3.0734e-08, ..., 0.0000e+00, + 5.8673e-08, 0.0000e+00], + ..., + [ 6.1467e-07, 2.0489e-08, 9.0525e-07, ..., 0.0000e+00, + 7.7765e-07, 0.0000e+00], + [ 3.8445e-06, 9.3132e-10, 1.1176e-08, ..., 0.0000e+00, + 3.2596e-08, 0.0000e+00], + [-4.9546e-07, 0.0000e+00, -9.7603e-07, ..., 0.0000e+00, + -1.0189e-06, 0.0000e+00]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0068, 0.0242, -0.0026, -0.0051, 0.0148, 0.0055, 0.0249, -0.0122, + 0.0058, -0.0054], device='cuda:0'), grad: tensor([-4.6939e-07, 1.6829e-06, -1.3433e-05, 6.0815e-07, 1.9409e-06, + 5.5321e-07, 5.7556e-07, 1.0334e-05, 9.3132e-06, -1.1131e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 202, time 217.21, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.5039 re_mapping 0.0056 re_causal 0.0177 /// teacc 99.04 lr 0.00001000 +Epoch 204, weight, value: tensor([[-0.0558, -0.1077, -0.0599, ..., -0.1151, -0.0204, -0.0325], + [-0.0496, 0.0490, -0.1613, ..., 0.0615, 0.0709, -0.1367], + [ 0.0564, -0.0076, -0.0633, ..., -0.0596, -0.0441, 0.0089], + ..., + [ 0.0567, -0.0597, -0.0169, ..., 0.0187, -0.0411, 0.0181], + [-0.0665, -0.0264, -0.0824, ..., -0.0286, -0.0350, -0.0709], + [-0.1433, -0.2040, 0.0286, ..., -0.0659, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 5.1223e-08, ..., 9.3132e-10, + 2.7940e-08, 9.3132e-10], + [ 4.9360e-08, 0.0000e+00, 3.1907e-06, ..., 9.3132e-09, + 7.2643e-08, 0.0000e+00], + [-1.9558e-08, 0.0000e+00, 1.0431e-07, ..., 1.8626e-09, + 3.5390e-08, 0.0000e+00], + ..., + [-1.9558e-08, 0.0000e+00, 2.5444e-06, ..., 4.7497e-08, + 2.9523e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.5087e-07, ..., 0.0000e+00, + 1.4342e-07, 1.8626e-09], + [ 2.7940e-09, 0.0000e+00, 6.3144e-06, ..., 5.5879e-09, + 1.6391e-07, 0.0000e+00]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0068, 0.0242, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0121, + 0.0058, -0.0054], device='cuda:0'), grad: tensor([ 2.1327e-07, 1.0818e-05, 3.7812e-07, 1.4342e-07, -3.7342e-05, + 6.9756e-07, -8.4843e-07, 7.5474e-06, 9.1828e-07, 1.7449e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 203, time 217.32, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4828 re_mapping 0.0055 re_causal 0.0171 /// teacc 99.02 lr 0.00001000 +Epoch 205, weight, value: tensor([[-0.0558, -0.1078, -0.0599, ..., -0.1151, -0.0205, -0.0325], + [-0.0497, 0.0490, -0.1614, ..., 0.0616, 0.0709, -0.1367], + [ 0.0564, -0.0076, -0.0633, ..., -0.0596, -0.0441, 0.0088], + ..., + [ 0.0568, -0.0597, -0.0169, ..., 0.0186, -0.0412, 0.0181], + [-0.0665, -0.0264, -0.0824, ..., -0.0286, -0.0351, -0.0709], + [-0.1433, -0.2040, 0.0286, ..., -0.0659, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.0338e-07, 0.0000e+00, 1.4994e-07, ..., 0.0000e+00, + 5.2154e-08, 0.0000e+00], + [ 2.0210e-07, 0.0000e+00, 3.5390e-08, ..., 9.3132e-10, + 1.6391e-07, 0.0000e+00], + [-5.2713e-07, 0.0000e+00, 3.6322e-08, ..., 0.0000e+00, + 1.4994e-07, 0.0000e+00], + ..., + [ 1.8626e-07, 0.0000e+00, 9.1270e-08, ..., 9.3132e-10, + 1.8720e-07, 0.0000e+00], + [ 4.9919e-07, 0.0000e+00, 1.4333e-06, ..., 0.0000e+00, + 1.2573e-07, 0.0000e+00], + [ 2.1979e-07, 0.0000e+00, -1.7742e-06, ..., 3.7253e-09, + -1.3188e-06, 0.0000e+00]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0068, 0.0241, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0121, + 0.0058, -0.0054], device='cuda:0'), grad: tensor([-2.7567e-07, 1.3702e-05, 5.9698e-07, 2.7511e-06, 1.5814e-06, + -2.4289e-06, 1.5087e-07, 1.3495e-06, -1.0237e-05, -7.2308e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 204, time 217.47, cls_loss 0.0019 cls_loss_mapping 0.0018 cls_loss_causal 0.4965 re_mapping 0.0054 re_causal 0.0175 /// teacc 99.00 lr 0.00001000 +Epoch 206, weight, value: tensor([[-0.0558, -0.1078, -0.0599, ..., -0.1152, -0.0205, -0.0325], + [-0.0499, 0.0491, -0.1614, ..., 0.0618, 0.0710, -0.1369], + [ 0.0564, -0.0076, -0.0633, ..., -0.0596, -0.0441, 0.0088], + ..., + [ 0.0569, -0.0597, -0.0169, ..., 0.0184, -0.0412, 0.0182], + [-0.0665, -0.0264, -0.0823, ..., -0.0287, -0.0351, -0.0709], + [-0.1433, -0.2040, 0.0285, ..., -0.0660, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.4435e-07, 0.0000e+00, -2.6543e-07, ..., 0.0000e+00, + 5.4948e-08, 0.0000e+00], + [ 2.2724e-07, 0.0000e+00, 8.1025e-08, ..., 0.0000e+00, + -5.5879e-08, 0.0000e+00], + [-7.2550e-07, 0.0000e+00, 1.7695e-08, ..., 0.0000e+00, + 1.7695e-08, 0.0000e+00], + ..., + [ 3.0734e-08, 0.0000e+00, 3.8184e-08, ..., 0.0000e+00, + 4.6566e-08, 0.0000e+00], + [ 1.6149e-06, 0.0000e+00, 7.2643e-08, ..., 0.0000e+00, + 3.7253e-08, 0.0000e+00], + [ 9.3132e-08, 0.0000e+00, 6.8545e-07, ..., 0.0000e+00, + -6.2399e-08, 0.0000e+00]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0068, 0.0240, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0120, + 0.0059, -0.0054], device='cuda:0'), grad: tensor([-8.0373e-07, 6.0908e-07, -6.3889e-07, -2.4177e-06, 5.5283e-06, + 5.7369e-07, -7.8827e-06, -1.6950e-07, 3.4273e-06, 1.7742e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 205, time 217.63, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.5079 re_mapping 0.0053 re_causal 0.0177 /// teacc 99.02 lr 0.00001000 +Epoch 207, weight, value: tensor([[-0.0558, -0.1079, -0.0599, ..., -0.1153, -0.0205, -0.0325], + [-0.0500, 0.0491, -0.1614, ..., 0.0618, 0.0710, -0.1369], + [ 0.0564, -0.0076, -0.0634, ..., -0.0596, -0.0441, 0.0088], + ..., + [ 0.0569, -0.0597, -0.0170, ..., 0.0184, -0.0412, 0.0181], + [-0.0665, -0.0264, -0.0823, ..., -0.0288, -0.0351, -0.0709], + [-0.1434, -0.2041, 0.0284, ..., -0.0660, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 9.0338e-08, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 3.1758e-07, 0.0000e+00], + [ 3.8277e-07, 0.0000e+00, 5.4948e-08, ..., 0.0000e+00, + -1.8235e-06, 0.0000e+00], + [ 9.8720e-07, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + 1.7416e-07, 0.0000e+00], + ..., + [-2.5183e-06, 0.0000e+00, 2.4308e-07, ..., 0.0000e+00, + 7.6741e-07, 0.0000e+00], + [ 8.7544e-08, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 1.0710e-07, 0.0000e+00], + [ 1.8813e-07, 0.0000e+00, 2.6729e-07, ..., 1.8626e-09, + 8.2888e-08, 0.0000e+00]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0068, 0.0239, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0059, -0.0054], device='cuda:0'), grad: tensor([-2.4647e-05, -2.9095e-06, 3.4720e-06, 1.2005e-06, -3.2783e-07, + 1.5674e-06, 2.0713e-05, -2.5723e-06, 2.7046e-06, 8.3167e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 206, time 217.72, cls_loss 0.0016 cls_loss_mapping 0.0014 cls_loss_causal 0.5105 re_mapping 0.0053 re_causal 0.0176 /// teacc 99.01 lr 0.00001000 +Epoch 208, weight, value: tensor([[-0.0559, -0.1079, -0.0599, ..., -0.1154, -0.0205, -0.0325], + [-0.0500, 0.0492, -0.1614, ..., 0.0620, 0.0711, -0.1369], + [ 0.0564, -0.0076, -0.0634, ..., -0.0597, -0.0443, 0.0088], + ..., + [ 0.0569, -0.0597, -0.0170, ..., 0.0183, -0.0412, 0.0181], + [-0.0665, -0.0264, -0.0823, ..., -0.0288, -0.0352, -0.0709], + [-0.1434, -0.2041, 0.0284, ..., -0.0660, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 0.0000e+00, -1.3039e-08, ..., 2.7940e-09, + 5.8673e-08, 0.0000e+00], + [ 8.0932e-07, -1.1176e-08, 1.8626e-09, ..., 1.2573e-07, + -8.9407e-07, 0.0000e+00], + [ 6.7055e-08, 1.8626e-09, 9.3132e-10, ..., 5.5879e-09, + 7.6368e-08, 0.0000e+00], + ..., + [-2.0042e-06, 7.4506e-09, 2.7940e-09, ..., -3.2596e-07, + 3.5763e-07, 0.0000e+00], + [ 1.3039e-08, 0.0000e+00, 1.7695e-08, ..., 9.3132e-10, + 6.9849e-08, 0.0000e+00], + [ 9.5740e-07, 0.0000e+00, 4.6566e-09, ..., 1.7416e-07, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0067, 0.0239, -0.0027, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0059, -0.0055], device='cuda:0'), grad: tensor([ 4.1630e-07, -6.7987e-07, 1.2973e-06, 3.9116e-07, 1.0571e-06, + 3.4366e-07, -2.8312e-07, -3.2913e-06, -1.4212e-06, 2.1700e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 207, time 217.40, cls_loss 0.0015 cls_loss_mapping 0.0014 cls_loss_causal 0.4776 re_mapping 0.0052 re_causal 0.0165 /// teacc 99.04 lr 0.00001000 +Epoch 209, weight, value: tensor([[-0.0559, -0.1080, -0.0599, ..., -0.1154, -0.0205, -0.0325], + [-0.0501, 0.0492, -0.1615, ..., 0.0620, 0.0711, -0.1369], + [ 0.0564, -0.0076, -0.0634, ..., -0.0597, -0.0443, 0.0088], + ..., + [ 0.0570, -0.0597, -0.0170, ..., 0.0183, -0.0412, 0.0181], + [-0.0665, -0.0265, -0.0823, ..., -0.0289, -0.0352, -0.0709], + [-0.1434, -0.2041, 0.0284, ..., -0.0661, 0.0401, 0.0480]], + device='cuda:0'), grad: tensor([[ 3.7253e-08, 9.3132e-10, 6.7055e-08, ..., 0.0000e+00, + 8.1956e-08, 0.0000e+00], + [ 9.4716e-07, -3.9116e-08, 1.1083e-07, ..., 1.1176e-08, + -1.5832e-08, 0.0000e+00], + [ 2.2072e-07, 1.7695e-08, 2.9802e-08, ..., 9.3132e-10, + 1.3690e-07, 0.0000e+00], + ..., + [-1.1697e-06, 8.3819e-09, 7.5437e-08, ..., 3.7253e-09, + 1.8068e-07, 0.0000e+00], + [ 6.8918e-08, 0.0000e+00, 2.4680e-07, ..., 9.3132e-10, + 6.1560e-07, 0.0000e+00], + [ 4.3400e-07, 0.0000e+00, 1.0245e-07, ..., 7.4506e-09, + -1.7835e-06, 0.0000e+00]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0068, 0.0239, -0.0027, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 4.2096e-07, 1.8552e-06, 1.0254e-06, 2.9095e-06, 1.0394e-06, + -2.1867e-06, -9.6019e-07, -1.7770e-06, 1.6447e-06, -3.9972e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 208, time 217.22, cls_loss 0.0015 cls_loss_mapping 0.0013 cls_loss_causal 0.4651 re_mapping 0.0050 re_causal 0.0163 /// teacc 99.05 lr 0.00001000 +Epoch 210, weight, value: tensor([[-0.0559, -0.1080, -0.0600, ..., -0.1156, -0.0205, -0.0325], + [-0.0501, 0.0492, -0.1615, ..., 0.0621, 0.0711, -0.1370], + [ 0.0564, -0.0076, -0.0634, ..., -0.0597, -0.0444, 0.0088], + ..., + [ 0.0570, -0.0597, -0.0170, ..., 0.0182, -0.0413, 0.0181], + [-0.0666, -0.0264, -0.0823, ..., -0.0289, -0.0352, -0.0709], + [-0.1434, -0.2041, 0.0284, ..., -0.0660, 0.0402, 0.0480]], + device='cuda:0'), grad: tensor([[ 3.1237e-06, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 6.9197e-07, 0.0000e+00, 5.2154e-08, ..., 0.0000e+00, + -3.7253e-08, 0.0000e+00], + [-9.0972e-06, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + ..., + [-3.3807e-07, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + -8.4750e-08, 0.0000e+00], + [ 2.5462e-06, 0.0000e+00, -3.8091e-07, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 1.1455e-07, 0.0000e+00, 2.3376e-07, ..., 0.0000e+00, + -1.2107e-08, 0.0000e+00]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0068, 0.0239, -0.0027, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0061, -0.0055], device='cuda:0'), grad: tensor([ 1.0259e-05, 1.9968e-06, -3.1561e-05, 6.5044e-06, 1.6205e-07, + 7.1432e-07, 4.2766e-06, -6.2864e-07, 6.6943e-06, 1.5358e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 209, time 217.24, cls_loss 0.0013 cls_loss_mapping 0.0013 cls_loss_causal 0.5117 re_mapping 0.0050 re_causal 0.0171 /// teacc 99.05 lr 0.00001000 +Epoch 211, weight, value: tensor([[-0.0559, -0.1080, -0.0600, ..., -0.1156, -0.0205, -0.0325], + [-0.0502, 0.0492, -0.1615, ..., 0.0621, 0.0712, -0.1370], + [ 0.0564, -0.0076, -0.0634, ..., -0.0597, -0.0444, 0.0088], + ..., + [ 0.0571, -0.0597, -0.0170, ..., 0.0181, -0.0413, 0.0181], + [-0.0666, -0.0265, -0.0823, ..., -0.0290, -0.0353, -0.0709], + [-0.1435, -0.2041, 0.0283, ..., -0.0659, 0.0402, 0.0480]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 5.3085e-08, 0.0000e+00], + [ 2.4680e-07, 0.0000e+00, 1.0058e-07, ..., 0.0000e+00, + -2.2259e-07, 0.0000e+00], + [-8.8476e-08, 0.0000e+00, 1.2107e-08, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + ..., + [-3.6787e-07, 0.0000e+00, 6.8918e-08, ..., 0.0000e+00, + 1.4622e-07, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 4.2934e-07, ..., 0.0000e+00, + 1.3970e-07, 0.0000e+00], + [ 8.4750e-08, 0.0000e+00, 5.0571e-07, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0068, 0.0238, -0.0027, -0.0051, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 1.2852e-07, 4.2189e-07, -2.7940e-08, 9.0338e-07, -1.6801e-06, + -7.8510e-07, -7.4785e-07, -3.5949e-07, 3.5483e-07, 1.7742e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 210, time 217.42, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4885 re_mapping 0.0050 re_causal 0.0169 /// teacc 99.05 lr 0.00001000 +Epoch 212, weight, value: tensor([[-0.0559, -0.1081, -0.0600, ..., -0.1157, -0.0205, -0.0325], + [-0.0502, 0.0492, -0.1615, ..., 0.0624, 0.0713, -0.1371], + [ 0.0565, -0.0076, -0.0634, ..., -0.0597, -0.0444, 0.0089], + ..., + [ 0.0571, -0.0597, -0.0170, ..., 0.0179, -0.0414, 0.0180], + [-0.0667, -0.0265, -0.0823, ..., -0.0291, -0.0353, -0.0709], + [-0.1435, -0.2042, 0.0283, ..., -0.0659, 0.0402, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.9092e-07, 0.0000e+00, 2.9150e-07, ..., 0.0000e+00, + 1.5646e-07, 5.5879e-09], + [ 8.0280e-07, 0.0000e+00, 6.0908e-07, ..., 2.7940e-08, + -8.4098e-07, 0.0000e+00], + [-1.1511e-06, 0.0000e+00, -1.4622e-07, ..., 9.3132e-10, + 1.0803e-07, 0.0000e+00], + ..., + [ 1.0980e-06, 0.0000e+00, 1.1213e-06, ..., 4.6566e-08, + 3.0268e-07, 3.7253e-09], + [ 6.6124e-08, 0.0000e+00, 2.2352e-08, ..., 0.0000e+00, + 1.0123e-06, 0.0000e+00], + [ 4.7591e-07, 0.0000e+00, 2.4363e-06, ..., 2.4214e-08, + -1.5665e-06, -2.9802e-08]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0067, 0.0239, -0.0027, -0.0051, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 1.0114e-06, 7.6648e-07, -1.4771e-06, 8.3260e-07, -3.3956e-06, + -4.9137e-06, 5.3737e-07, 3.7849e-06, 3.8408e-06, -9.9372e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 211, time 217.28, cls_loss 0.0011 cls_loss_mapping 0.0009 cls_loss_causal 0.4808 re_mapping 0.0049 re_causal 0.0169 /// teacc 99.03 lr 0.00001000 +Epoch 213, weight, value: tensor([[-0.0559, -0.1081, -0.0600, ..., -0.1158, -0.0205, -0.0325], + [-0.0502, 0.0493, -0.1616, ..., 0.0624, 0.0713, -0.1371], + [ 0.0565, -0.0076, -0.0634, ..., -0.0597, -0.0444, 0.0089], + ..., + [ 0.0571, -0.0598, -0.0171, ..., 0.0179, -0.0414, 0.0180], + [-0.0667, -0.0265, -0.0823, ..., -0.0292, -0.0354, -0.0709], + [-0.1435, -0.2042, 0.0283, ..., -0.0659, 0.0402, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 6.5193e-08, 0.0000e+00], + [ 4.4517e-07, -9.3132e-10, 5.3085e-08, ..., 0.0000e+00, + -5.6997e-07, 0.0000e+00], + [ 1.8161e-07, 0.0000e+00, 3.5390e-08, ..., 0.0000e+00, + 8.5682e-08, 0.0000e+00], + ..., + [-1.8710e-06, 9.3132e-10, -5.5879e-09, ..., 0.0000e+00, + 3.1292e-07, 0.0000e+00], + [-1.2387e-07, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + -1.5646e-07, 0.0000e+00], + [ 2.3283e-07, 0.0000e+00, -4.0606e-07, ..., 0.0000e+00, + -2.2538e-07, 0.0000e+00]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0067, 0.0239, -0.0027, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 2.2911e-07, -5.0385e-07, 3.4571e-06, 1.9055e-06, 5.1036e-06, + 3.6415e-07, 1.2387e-07, -8.7395e-06, -9.7603e-07, -9.7137e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 212, time 217.45, cls_loss 0.0011 cls_loss_mapping 0.0010 cls_loss_causal 0.5021 re_mapping 0.0049 re_causal 0.0171 /// teacc 99.06 lr 0.00001000 +Epoch 214, weight, value: tensor([[-0.0559, -0.1081, -0.0601, ..., -0.1159, -0.0205, -0.0325], + [-0.0502, 0.0493, -0.1616, ..., 0.0626, 0.0714, -0.1371], + [ 0.0565, -0.0076, -0.0634, ..., -0.0598, -0.0444, 0.0089], + ..., + [ 0.0571, -0.0598, -0.0171, ..., 0.0177, -0.0415, 0.0180], + [-0.0667, -0.0266, -0.0823, ..., -0.0293, -0.0354, -0.0709], + [-0.1435, -0.2042, 0.0283, ..., -0.0660, 0.0402, 0.0480]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 0.0000e+00, 2.7008e-08, ..., 9.3132e-09, + 3.6322e-08, 0.0000e+00], + [ 6.1374e-07, 0.0000e+00, 1.4622e-07, ..., -4.5076e-06, + -7.3947e-06, 0.0000e+00], + [-3.1106e-07, 0.0000e+00, 5.2154e-08, ..., 3.7253e-09, + 2.5146e-08, 0.0000e+00], + ..., + [-6.5286e-07, 0.0000e+00, 7.6275e-07, ..., 2.5872e-06, + 4.2431e-06, 0.0000e+00], + [ 4.5635e-08, 0.0000e+00, 3.5390e-08, ..., 1.3039e-08, + 5.6811e-08, 0.0000e+00], + [ 8.7544e-08, 0.0000e+00, 1.4110e-06, ..., 3.1013e-07, + 4.1258e-07, 0.0000e+00]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0067, 0.0239, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([-2.6915e-07, -1.7598e-05, -5.1875e-07, 3.6042e-07, -1.0915e-06, + -1.1921e-07, 2.6356e-07, 1.2495e-05, 3.7346e-07, 6.0759e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 213, time 217.46, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4840 re_mapping 0.0048 re_causal 0.0169 /// teacc 99.05 lr 0.00001000 +Epoch 215, weight, value: tensor([[-0.0559, -0.1082, -0.0601, ..., -0.1160, -0.0205, -0.0325], + [-0.0502, 0.0493, -0.1616, ..., 0.0627, 0.0714, -0.1371], + [ 0.0565, -0.0076, -0.0634, ..., -0.0598, -0.0445, 0.0088], + ..., + [ 0.0571, -0.0598, -0.0171, ..., 0.0176, -0.0415, 0.0180], + [-0.0668, -0.0267, -0.0823, ..., -0.0294, -0.0354, -0.0709], + [-0.1435, -0.2042, 0.0283, ..., -0.0658, 0.0403, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 0.0000e+00, 2.3935e-07, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.9092e-07, 0.0000e+00, 1.6950e-07, ..., 9.3132e-10, + -2.6729e-07, 0.0000e+00], + [-5.6811e-08, 0.0000e+00, 4.4703e-08, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00], + ..., + [-2.8871e-07, 0.0000e+00, 3.1013e-07, ..., 9.3132e-10, + 1.8999e-07, 0.0000e+00], + [ 8.9407e-08, 0.0000e+00, 1.7202e-06, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 1.1176e-07, 0.0000e+00, 8.6520e-07, ..., 1.2107e-08, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0067, 0.0239, -0.0026, -0.0051, 0.0149, 0.0055, 0.0248, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 6.1560e-07, 2.0023e-07, 8.5682e-08, 3.1114e-05, -1.6857e-06, + -4.1664e-05, 3.9414e-06, 5.8394e-07, 4.5672e-06, 2.3115e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 214, time 217.44, cls_loss 0.0013 cls_loss_mapping 0.0011 cls_loss_causal 0.4966 re_mapping 0.0049 re_causal 0.0167 /// teacc 99.03 lr 0.00001000 +Epoch 216, weight, value: tensor([[-0.0559, -0.1082, -0.0601, ..., -0.1161, -0.0206, -0.0325], + [-0.0502, 0.0493, -0.1616, ..., 0.0627, 0.0715, -0.1372], + [ 0.0566, -0.0076, -0.0634, ..., -0.0598, -0.0445, 0.0088], + ..., + [ 0.0571, -0.0598, -0.0172, ..., 0.0176, -0.0416, 0.0180], + [-0.0668, -0.0267, -0.0823, ..., -0.0295, -0.0355, -0.0709], + [-0.1435, -0.2042, 0.0282, ..., -0.0659, 0.0403, 0.0480]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 0.0000e+00, -2.5146e-08, ..., 0.0000e+00, + 1.0431e-07, 0.0000e+00], + [-1.7695e-07, 0.0000e+00, 6.7055e-08, ..., 1.8626e-09, + -2.3693e-06, 0.0000e+00], + [-8.1025e-08, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 9.8627e-07, 0.0000e+00], + ..., + [ 4.2375e-07, 0.0000e+00, 2.5798e-07, ..., 9.3132e-10, + 1.1530e-06, 0.0000e+00], + [ 4.1910e-08, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 4.5728e-07, 0.0000e+00], + [ 4.0978e-08, 0.0000e+00, 1.0338e-07, ..., 4.6566e-09, + -1.3132e-07, 0.0000e+00]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0067, 0.0239, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([-6.3516e-07, -5.3756e-06, 2.2668e-06, -4.6287e-07, 3.8091e-07, + 1.2405e-06, -2.3022e-06, 3.4161e-06, 1.4408e-06, 6.6124e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 215, time 217.22, cls_loss 0.0013 cls_loss_mapping 0.0011 cls_loss_causal 0.4995 re_mapping 0.0049 re_causal 0.0170 /// teacc 99.06 lr 0.00001000 +Epoch 217, weight, value: tensor([[-0.0559, -0.1082, -0.0601, ..., -0.1162, -0.0206, -0.0325], + [-0.0502, 0.0494, -0.1617, ..., 0.0628, 0.0715, -0.1372], + [ 0.0566, -0.0076, -0.0634, ..., -0.0598, -0.0445, 0.0088], + ..., + [ 0.0571, -0.0598, -0.0173, ..., 0.0175, -0.0416, 0.0180], + [-0.0668, -0.0267, -0.0823, ..., -0.0296, -0.0355, -0.0709], + [-0.1436, -0.2042, 0.0281, ..., -0.0659, 0.0403, 0.0480]], + device='cuda:0'), grad: tensor([[ 7.7300e-08, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 3.2410e-07, 0.0000e+00], + [ 5.3085e-08, 0.0000e+00, 2.4214e-08, ..., 0.0000e+00, + -1.2927e-06, 0.0000e+00], + [-7.5251e-07, 0.0000e+00, 2.7940e-08, ..., 0.0000e+00, + 1.0338e-07, 0.0000e+00], + ..., + [ 4.2375e-07, 0.0000e+00, 3.1665e-08, ..., 0.0000e+00, + 4.8615e-07, 0.0000e+00], + [ 1.0245e-07, 9.3132e-10, -1.0608e-06, ..., 1.8626e-09, + -7.7300e-08, 9.3132e-10], + [ 1.3970e-08, 0.0000e+00, 1.1688e-06, ..., 9.3132e-10, + 2.5239e-07, 0.0000e+00]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0067, 0.0240, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 8.7358e-07, -2.3674e-06, -8.9686e-07, 1.2740e-06, 3.4831e-07, + 1.4622e-07, 2.3004e-07, 1.9260e-06, -7.6666e-06, 6.1281e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 216, time 217.54, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4542 re_mapping 0.0048 re_causal 0.0164 /// teacc 99.00 lr 0.00001000 +Epoch 218, weight, value: tensor([[-0.0559, -0.1083, -0.0601, ..., -0.1163, -0.0206, -0.0325], + [-0.0502, 0.0494, -0.1617, ..., 0.0628, 0.0715, -0.1372], + [ 0.0567, -0.0076, -0.0634, ..., -0.0598, -0.0446, 0.0088], + ..., + [ 0.0571, -0.0598, -0.0173, ..., 0.0175, -0.0416, 0.0180], + [-0.0668, -0.0267, -0.0823, ..., -0.0296, -0.0356, -0.0709], + [-0.1436, -0.2043, 0.0281, ..., -0.0658, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.4678e-06, 0.0000e+00, 9.8720e-08, ..., 4.6566e-09, + 2.3283e-08, 0.0000e+00], + [ 9.6858e-08, 0.0000e+00, 4.3772e-08, ..., 7.4506e-09, + -4.3120e-07, 0.0000e+00], + [-2.9597e-06, 0.0000e+00, -1.2666e-07, ..., 9.3132e-10, + 3.5390e-08, 0.0000e+00], + ..., + [ 5.0664e-07, 0.0000e+00, 2.9802e-08, ..., 3.7253e-09, + 2.6077e-07, 0.0000e+00], + [ 9.0711e-07, 0.0000e+00, 2.5984e-07, ..., 4.1910e-08, + 5.1223e-08, 0.0000e+00], + [ 1.7136e-07, 0.0000e+00, 3.9395e-07, ..., 1.2107e-08, + 4.7497e-08, 0.0000e+00]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0067, 0.0239, -0.0026, -0.0051, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([ 9.0003e-06, -4.7963e-07, -1.5974e-05, 7.8753e-06, -3.7253e-07, + -8.4788e-06, 2.8033e-07, 1.5674e-06, 5.2229e-06, 1.3541e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 217, time 217.58, cls_loss 0.0011 cls_loss_mapping 0.0009 cls_loss_causal 0.4670 re_mapping 0.0047 re_causal 0.0166 /// teacc 99.02 lr 0.00001000 +Epoch 219, weight, value: tensor([[-0.0559, -0.1083, -0.0601, ..., -0.1163, -0.0206, -0.0325], + [-0.0503, 0.0494, -0.1617, ..., 0.0628, 0.0715, -0.1372], + [ 0.0567, -0.0076, -0.0633, ..., -0.0598, -0.0446, 0.0088], + ..., + [ 0.0571, -0.0598, -0.0174, ..., 0.0175, -0.0416, 0.0179], + [-0.0669, -0.0267, -0.0823, ..., -0.0297, -0.0356, -0.0709], + [-0.1437, -0.2043, 0.0280, ..., -0.0659, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.7788e-07, 0.0000e+00, 5.2620e-08, ..., 0.0000e+00, + 4.4238e-08, 4.6566e-10], + [ 4.7497e-07, 0.0000e+00, 4.1910e-09, ..., -2.3283e-09, + -7.1479e-07, 0.0000e+00], + [-6.0163e-07, 0.0000e+00, -1.8720e-07, ..., 0.0000e+00, + 1.3690e-07, 0.0000e+00], + ..., + [-1.0598e-06, 0.0000e+00, 2.5611e-08, ..., 4.6566e-10, + 6.7661e-07, 0.0000e+00], + [ 4.9919e-07, 0.0000e+00, 4.5169e-08, ..., 9.3132e-10, + 1.4110e-07, 9.3132e-10], + [ 2.2305e-07, 0.0000e+00, -2.7893e-07, ..., 4.6566e-10, + -1.4026e-06, 0.0000e+00]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0067, 0.0239, -0.0026, -0.0050, 0.0149, 0.0055, 0.0248, -0.0119, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([-6.8732e-07, -2.9057e-07, -5.4715e-07, 5.2294e-07, 8.1211e-06, + 2.3982e-07, 1.4668e-07, 1.5218e-06, 1.2591e-06, -1.0327e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 218, time 217.43, cls_loss 0.0010 cls_loss_mapping 0.0007 cls_loss_causal 0.4783 re_mapping 0.0048 re_causal 0.0166 /// teacc 99.08 lr 0.00001000 +Epoch 220, weight, value: tensor([[-0.0560, -0.1083, -0.0601, ..., -0.1164, -0.0206, -0.0325], + [-0.0504, 0.0494, -0.1618, ..., 0.0628, 0.0715, -0.1372], + [ 0.0568, -0.0076, -0.0633, ..., -0.0598, -0.0446, 0.0088], + ..., + [ 0.0572, -0.0598, -0.0174, ..., 0.0175, -0.0416, 0.0179], + [-0.0669, -0.0267, -0.0823, ..., -0.0298, -0.0356, -0.0710], + [-0.1437, -0.2043, 0.0280, ..., -0.0660, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [ 5.2620e-08, 0.0000e+00, 3.1199e-08, ..., 0.0000e+00, + -1.4389e-07, 0.0000e+00], + [ 3.2131e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 2.0955e-08, 0.0000e+00], + ..., + [-6.4727e-08, 0.0000e+00, 1.7229e-08, ..., 4.6566e-10, + 1.2619e-07, 0.0000e+00], + [ 2.1886e-08, 0.0000e+00, 3.2131e-08, ..., 4.6566e-10, + 4.1444e-08, 0.0000e+00], + [ 2.6543e-08, 0.0000e+00, 1.3495e-06, ..., 4.6566e-10, + -1.6764e-07, 0.0000e+00]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0067, 0.0239, -0.0025, -0.0050, 0.0149, 0.0055, 0.0249, -0.0119, + 0.0060, -0.0056], device='cuda:0'), grad: tensor([-1.6419e-06, -5.6811e-08, 1.2480e-06, 1.0319e-06, -7.6834e-08, + -3.3490e-06, -6.5193e-09, 2.6636e-07, 2.9337e-07, 2.3078e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 219, time 217.35, cls_loss 0.0012 cls_loss_mapping 0.0009 cls_loss_causal 0.4607 re_mapping 0.0048 re_causal 0.0166 /// teacc 99.08 lr 0.00001000 +Epoch 221, weight, value: tensor([[-0.0560, -0.1083, -0.0601, ..., -0.1164, -0.0206, -0.0325], + [-0.0505, 0.0495, -0.1618, ..., 0.0629, 0.0716, -0.1372], + [ 0.0568, -0.0076, -0.0633, ..., -0.0598, -0.0447, 0.0088], + ..., + [ 0.0573, -0.0598, -0.0175, ..., 0.0174, -0.0416, 0.0179], + [-0.0670, -0.0267, -0.0823, ..., -0.0299, -0.0357, -0.0710], + [-0.1438, -0.2043, 0.0280, ..., -0.0660, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 0.0000e+00, 2.7940e-09, ..., 3.2596e-09, + 1.4901e-08, 3.2596e-09], + [ 1.0049e-06, 0.0000e+00, 8.6147e-08, ..., 2.0443e-07, + -2.0675e-07, 4.6566e-10], + [ 9.4529e-08, 0.0000e+00, 7.9162e-09, ..., 7.4506e-09, + 1.7695e-08, 9.3132e-10], + ..., + [-1.5199e-06, 0.0000e+00, 1.4901e-08, ..., -3.4412e-07, + 1.7369e-07, 0.0000e+00], + [ 1.4063e-07, 0.0000e+00, -3.4459e-08, ..., 1.8626e-09, + 2.4121e-07, 2.3283e-09], + [ 5.0943e-07, 0.0000e+00, -3.6694e-07, ..., 1.0151e-07, + -5.3551e-07, 0.0000e+00]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0067, 0.0238, -0.0026, -0.0050, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0060, -0.0055], device='cuda:0'), grad: tensor([-1.2210e-06, 2.8107e-06, 3.2131e-07, -4.5495e-07, 3.1479e-06, + 8.6799e-07, -6.4448e-07, -4.1798e-06, 5.5414e-08, -7.0315e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 220, time 217.49, cls_loss 0.0009 cls_loss_mapping 0.0009 cls_loss_causal 0.4765 re_mapping 0.0048 re_causal 0.0169 /// teacc 99.09 lr 0.00001000 +Epoch 222, weight, value: tensor([[-0.0560, -0.1084, -0.0601, ..., -0.1165, -0.0206, -0.0325], + [-0.0505, 0.0495, -0.1618, ..., 0.0629, 0.0716, -0.1372], + [ 0.0568, -0.0076, -0.0633, ..., -0.0598, -0.0447, 0.0088], + ..., + [ 0.0573, -0.0598, -0.0175, ..., 0.0174, -0.0417, 0.0179], + [-0.0671, -0.0267, -0.0823, ..., -0.0300, -0.0357, -0.0710], + [-0.1438, -0.2043, 0.0280, ..., -0.0659, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [ 2.6124e-07, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + -7.3574e-08, 0.0000e+00], + [ 4.2375e-08, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 2.3749e-08, 0.0000e+00], + ..., + [-1.9651e-07, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.0990e-07, 0.0000e+00], + [ 2.3283e-08, 0.0000e+00, 6.9384e-08, ..., 0.0000e+00, + -1.8021e-07, 0.0000e+00], + [ 8.6613e-08, 0.0000e+00, 1.6298e-08, ..., 0.0000e+00, + -4.9360e-08, 0.0000e+00]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0066, 0.0238, -0.0025, -0.0050, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0060, -0.0056], device='cuda:0'), grad: tensor([-2.9802e-08, 9.0152e-07, 2.3050e-07, -9.6392e-08, 3.3760e-07, + 1.4082e-06, -9.2573e-07, -3.0873e-07, -1.6680e-06, 1.7788e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 221, time 217.43, cls_loss 0.0009 cls_loss_mapping 0.0009 cls_loss_causal 0.4679 re_mapping 0.0048 re_causal 0.0166 /// teacc 99.06 lr 0.00001000 +Epoch 223, weight, value: tensor([[-0.0560, -0.1084, -0.0601, ..., -0.1165, -0.0206, -0.0325], + [-0.0505, 0.0495, -0.1618, ..., 0.0629, 0.0717, -0.1372], + [ 0.0569, -0.0076, -0.0634, ..., -0.0598, -0.0448, 0.0088], + ..., + [ 0.0572, -0.0599, -0.0175, ..., 0.0174, -0.0417, 0.0179], + [-0.0672, -0.0268, -0.0823, ..., -0.0300, -0.0358, -0.0710], + [-0.1439, -0.2044, 0.0280, ..., -0.0659, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 7.8697e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -2.0489e-08, 0.0000e+00], + [ 2.1141e-07, 4.6566e-10, 3.9581e-08, ..., 0.0000e+00, + -2.4308e-07, 0.0000e+00], + [ 1.1809e-06, 4.6566e-10, 4.6566e-09, ..., 0.0000e+00, + 5.1223e-08, 0.0000e+00], + ..., + [ 1.0272e-06, -3.7253e-09, 2.8405e-08, ..., 0.0000e+00, + 1.8161e-07, 0.0000e+00], + [ 8.6613e-08, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 3.0734e-08, 0.0000e+00], + [ 2.5798e-07, 0.0000e+00, 1.3504e-07, ..., 0.0000e+00, + -1.3178e-07, 0.0000e+00]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0066, 0.0238, -0.0025, -0.0050, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0059, -0.0056], device='cuda:0'), grad: tensor([-2.0564e-06, 1.2107e-08, 2.0638e-06, -3.6675e-06, -4.1304e-07, + 6.1328e-07, 4.9965e-07, 1.7984e-06, 2.7707e-07, 8.4564e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 222, time 217.28, cls_loss 0.0011 cls_loss_mapping 0.0008 cls_loss_causal 0.4832 re_mapping 0.0048 re_causal 0.0167 /// teacc 99.09 lr 0.00001000 +Epoch 224, weight, value: tensor([[-0.0560, -0.1084, -0.0601, ..., -0.1167, -0.0206, -0.0325], + [-0.0505, 0.0496, -0.1619, ..., 0.0629, 0.0717, -0.1372], + [ 0.0569, -0.0076, -0.0633, ..., -0.0599, -0.0448, 0.0088], + ..., + [ 0.0573, -0.0599, -0.0175, ..., 0.0174, -0.0417, 0.0179], + [-0.0673, -0.0268, -0.0823, ..., -0.0301, -0.0358, -0.0710], + [-0.1439, -0.2044, 0.0279, ..., -0.0658, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[ 2.8871e-08, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 3.8650e-08, 4.6566e-10], + [ 5.2154e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -1.2573e-07, 0.0000e+00], + [-4.5029e-07, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 2.1886e-08, 0.0000e+00], + ..., + [-5.6811e-08, 0.0000e+00, 1.7229e-08, ..., 0.0000e+00, + 9.7323e-08, 0.0000e+00], + [ 3.8091e-07, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + 2.7474e-08, 5.5879e-09], + [ 5.4948e-08, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + -7.8697e-08, 0.0000e+00]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0066, 0.0238, -0.0025, -0.0051, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0059, -0.0056], device='cuda:0'), grad: tensor([-3.6601e-07, 4.7963e-08, -3.9814e-07, 2.4494e-07, 2.1666e-05, + 2.1085e-06, -2.4587e-05, 4.5914e-07, 1.1511e-06, -3.2643e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 223, time 217.27, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4846 re_mapping 0.0047 re_causal 0.0167 /// teacc 99.09 lr 0.00001000 +Epoch 225, weight, value: tensor([[-0.0560, -0.1085, -0.0601, ..., -0.1168, -0.0206, -0.0325], + [-0.0505, 0.0497, -0.1619, ..., 0.0629, 0.0718, -0.1372], + [ 0.0569, -0.0076, -0.0633, ..., -0.0599, -0.0448, 0.0088], + ..., + [ 0.0573, -0.0600, -0.0175, ..., 0.0174, -0.0418, 0.0179], + [-0.0674, -0.0268, -0.0823, ..., -0.0302, -0.0359, -0.0710], + [-0.1440, -0.2044, 0.0279, ..., -0.0659, 0.0404, 0.0480]], + device='cuda:0'), grad: tensor([[-1.1967e-07, 0.0000e+00, 1.3970e-08, ..., 1.8626e-09, + 2.7986e-07, 0.0000e+00], + [ 1.5181e-07, 0.0000e+00, 3.9814e-07, ..., 1.0710e-08, + -8.0392e-06, 0.0000e+00], + [-1.5661e-05, 0.0000e+00, 3.9581e-08, ..., 3.7253e-09, + 1.5087e-07, 0.0000e+00], + ..., + [ 1.5557e-05, 0.0000e+00, 1.5693e-07, ..., 3.4459e-08, + 1.1967e-06, 0.0000e+00], + [ 2.4354e-07, 0.0000e+00, 3.7719e-08, ..., 2.3283e-09, + 4.0270e-06, 0.0000e+00], + [ 2.9383e-07, 0.0000e+00, 5.9605e-07, ..., 9.7789e-09, + 2.2724e-07, 0.0000e+00]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0066, 0.0239, -0.0025, -0.0051, 0.0149, 0.0055, 0.0249, -0.0118, + 0.0059, -0.0056], device='cuda:0'), grad: tensor([-8.4564e-06, -3.4094e-05, -1.8910e-05, -6.5891e-07, 7.0930e-06, + 8.5216e-07, 9.5554e-07, 2.4691e-05, 1.8150e-05, 1.0386e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 224, time 217.37, cls_loss 0.0009 cls_loss_mapping 0.0008 cls_loss_causal 0.4598 re_mapping 0.0046 re_causal 0.0162 /// teacc 99.07 lr 0.00001000 +Epoch 226, weight, value: tensor([[-0.0560, -0.1087, -0.0601, ..., -0.1169, -0.0206, -0.0325], + [-0.0505, 0.0498, -0.1619, ..., 0.0629, 0.0718, -0.1372], + [ 0.0570, -0.0076, -0.0633, ..., -0.0599, -0.0448, 0.0088], + ..., + [ 0.0573, -0.0600, -0.0176, ..., 0.0174, -0.0418, 0.0179], + [-0.0675, -0.0269, -0.0824, ..., -0.0303, -0.0360, -0.0710], + [-0.1440, -0.2045, 0.0278, ..., -0.0659, 0.0405, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.3970e-07, 7.9162e-09, 8.1025e-08, ..., 0.0000e+00, + -1.1036e-07, 0.0000e+00], + [ 1.7118e-06, 3.4319e-07, 3.0268e-08, ..., 0.0000e+00, + -2.6543e-08, 0.0000e+00], + [-3.2084e-07, 9.8441e-07, 8.7079e-08, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + ..., + [ 1.5059e-06, 2.4540e-07, 2.5099e-07, ..., 1.8626e-09, + 1.0058e-07, 0.0000e+00], + [-9.4846e-06, -4.2617e-06, -1.8366e-06, ..., 0.0000e+00, + -3.2596e-09, 4.6566e-10], + [ 1.4128e-06, 2.2352e-08, 9.1875e-07, ..., 0.0000e+00, + -1.8021e-07, 0.0000e+00]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0066, 0.0239, -0.0025, -0.0051, 0.0148, 0.0055, 0.0249, -0.0118, + 0.0058, -0.0056], device='cuda:0'), grad: tensor([-1.4342e-06, 5.1707e-06, 7.7859e-06, 5.4725e-06, 1.0990e-06, + 9.1642e-06, 1.3642e-05, 4.9323e-06, -5.1469e-05, 5.7034e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 225, time 217.26, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4718 re_mapping 0.0046 re_causal 0.0162 /// teacc 99.06 lr 0.00001000 +Epoch 227, weight, value: tensor([[-0.0560, -0.1087, -0.0601, ..., -0.1169, -0.0207, -0.0325], + [-0.0505, 0.0499, -0.1620, ..., 0.0629, 0.0718, -0.1373], + [ 0.0570, -0.0077, -0.0633, ..., -0.0599, -0.0448, 0.0088], + ..., + [ 0.0573, -0.0600, -0.0176, ..., 0.0174, -0.0418, 0.0178], + [-0.0676, -0.0269, -0.0824, ..., -0.0303, -0.0360, -0.0710], + [-0.1441, -0.2045, 0.0278, ..., -0.0659, 0.0405, 0.0480]], + device='cuda:0'), grad: tensor([[ 1.0887e-06, 0.0000e+00, -4.2282e-07, ..., 4.6566e-10, + 1.7229e-08, 0.0000e+00], + [ 8.6501e-06, 4.6566e-10, 5.1223e-08, ..., 8.3819e-09, + -1.8300e-07, 0.0000e+00], + [ 1.9930e-07, -3.0268e-08, 8.7079e-08, ..., 4.6566e-10, + 2.9337e-08, 0.0000e+00], + ..., + [-1.4834e-05, 2.7940e-08, 4.0978e-08, ..., 1.8626e-09, + 1.8068e-07, 0.0000e+00], + [ 8.0559e-08, 4.6566e-10, -3.1665e-08, ..., 0.0000e+00, + 4.4238e-08, 0.0000e+00], + [ 7.6182e-07, 0.0000e+00, 2.2445e-07, ..., 5.1223e-09, + -2.0023e-07, 0.0000e+00]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0066, 0.0239, -0.0024, -0.0050, 0.0148, 0.0055, 0.0250, -0.0118, + 0.0058, -0.0056], device='cuda:0'), grad: tensor([ 5.4482e-08, 1.9237e-05, 1.0636e-06, 3.3602e-06, 4.5709e-06, + 2.9020e-06, -2.7474e-08, -3.2455e-05, -4.2375e-08, 1.3420e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 226, time 217.33, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4768 re_mapping 0.0046 re_causal 0.0162 /// teacc 99.06 lr 0.00001000 +Epoch 228, weight, value: tensor([[-0.0560, -0.1089, -0.0601, ..., -0.1170, -0.0207, -0.0325], + [-0.0506, 0.0501, -0.1620, ..., 0.0629, 0.0718, -0.1373], + [ 0.0571, -0.0077, -0.0633, ..., -0.0599, -0.0449, 0.0088], + ..., + [ 0.0573, -0.0602, -0.0177, ..., 0.0174, -0.0418, 0.0178], + [-0.0677, -0.0269, -0.0824, ..., -0.0304, -0.0360, -0.0710], + [-0.1441, -0.2046, 0.0278, ..., -0.0660, 0.0406, 0.0481]], + device='cuda:0'), grad: tensor([[-1.5069e-06, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + 4.1444e-08, 0.0000e+00], + [ 2.5202e-06, 0.0000e+00, 1.4435e-08, ..., 1.2424e-06, + -2.6915e-07, 0.0000e+00], + [-7.4320e-07, 0.0000e+00, -6.9849e-09, ..., 2.3283e-09, + 2.0489e-08, 0.0000e+00], + ..., + [-3.3397e-06, 0.0000e+00, -1.0617e-07, ..., -1.3513e-06, + 2.9290e-07, 0.0000e+00], + [ 4.2468e-07, 0.0000e+00, 1.5832e-08, ..., 0.0000e+00, + 4.4703e-08, 1.3970e-09], + [ 1.7658e-06, 0.0000e+00, 6.2399e-08, ..., 8.1956e-08, + -2.5518e-07, 0.0000e+00]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0065, 0.0238, -0.0024, -0.0051, 0.0148, 0.0055, 0.0249, -0.0118, + 0.0058, -0.0056], device='cuda:0'), grad: tensor([-7.0482e-06, 5.2638e-06, -5.5600e-07, 6.2585e-07, 1.0468e-06, + 4.6752e-07, 1.5907e-06, -7.0408e-06, 1.1679e-06, 4.4964e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 227, time 217.31, cls_loss 0.0011 cls_loss_mapping 0.0008 cls_loss_causal 0.4716 re_mapping 0.0045 re_causal 0.0158 /// teacc 99.06 lr 0.00001000 +Epoch 229, weight, value: tensor([[-0.0560, -0.1089, -0.0602, ..., -0.1171, -0.0207, -0.0325], + [-0.0506, 0.0501, -0.1620, ..., 0.0629, 0.0719, -0.1373], + [ 0.0571, -0.0077, -0.0632, ..., -0.0600, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0178, ..., 0.0174, -0.0418, 0.0178], + [-0.0678, -0.0269, -0.0824, ..., -0.0305, -0.0360, -0.0710], + [-0.1442, -0.2046, 0.0277, ..., -0.0660, 0.0406, 0.0481]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 0.0000e+00, -9.5740e-06, ..., 4.6566e-10, + 5.0291e-08, 0.0000e+00], + [ 4.0382e-06, 0.0000e+00, 7.0781e-08, ..., 0.0000e+00, + -1.7462e-07, 0.0000e+00], + [ 1.1083e-07, 0.0000e+00, 1.4435e-08, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00], + ..., + [-1.2569e-05, 0.0000e+00, 1.5600e-07, ..., 4.6566e-10, + 1.7276e-07, 0.0000e+00], + [ 1.5134e-07, 0.0000e+00, -1.1735e-07, ..., 1.3970e-09, + -7.8231e-08, 0.0000e+00], + [ 5.5693e-06, 0.0000e+00, 1.9604e-07, ..., 1.3970e-09, + -6.3051e-07, 0.0000e+00]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0065, 0.0238, -0.0024, -0.0051, 0.0148, 0.0055, 0.0249, -0.0118, + 0.0058, -0.0056], device='cuda:0'), grad: tensor([-7.5281e-05, 1.2934e-05, 9.1121e-06, 9.0823e-06, 3.4235e-06, + 1.7181e-05, 4.0114e-05, -4.0084e-05, 1.9614e-06, 2.1502e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 228, time 217.25, cls_loss 0.0011 cls_loss_mapping 0.0010 cls_loss_causal 0.4693 re_mapping 0.0045 re_causal 0.0162 /// teacc 99.07 lr 0.00001000 +Epoch 230, weight, value: tensor([[-0.0560, -0.1090, -0.0601, ..., -0.1172, -0.0207, -0.0325], + [-0.0507, 0.0501, -0.1621, ..., 0.0629, 0.0719, -0.1373], + [ 0.0573, -0.0076, -0.0632, ..., -0.0600, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0178, ..., 0.0174, -0.0418, 0.0178], + [-0.0680, -0.0269, -0.0824, ..., -0.0306, -0.0361, -0.0710], + [-0.1442, -0.2046, 0.0277, ..., -0.0660, 0.0406, 0.0481]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 3.9581e-08, 0.0000e+00], + [ 7.5437e-08, 8.3819e-09, 4.2375e-08, ..., 4.6566e-10, + -4.6706e-07, 0.0000e+00], + [-6.9477e-07, -1.3970e-08, 8.3819e-09, ..., 0.0000e+00, + 3.7253e-08, 0.0000e+00], + ..., + [ 4.6426e-07, 5.1223e-09, 3.2596e-08, ..., 0.0000e+00, + 2.7241e-07, 0.0000e+00], + [ 7.3574e-08, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 8.9873e-08, 0.0000e+00], + [ 9.7789e-09, 4.6566e-10, 1.2629e-06, ..., 4.6566e-10, + -1.9791e-07, 0.0000e+00]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0065, 0.0238, -0.0023, -0.0051, 0.0148, 0.0055, 0.0249, -0.0118, + 0.0057, -0.0056], device='cuda:0'), grad: tensor([-2.1653e-07, -7.5204e-07, -7.1526e-07, 1.3178e-07, -2.4363e-06, + 1.5181e-07, -3.0687e-07, 1.3579e-06, 2.9337e-07, 2.4978e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 229, time 217.25, cls_loss 0.0008 cls_loss_mapping 0.0007 cls_loss_causal 0.4313 re_mapping 0.0045 re_causal 0.0159 /// teacc 99.05 lr 0.00001000 +Epoch 231, weight, value: tensor([[-0.0560, -0.1090, -0.0601, ..., -0.1173, -0.0207, -0.0325], + [-0.0507, 0.0501, -0.1621, ..., 0.0629, 0.0719, -0.1374], + [ 0.0574, -0.0076, -0.0631, ..., -0.0600, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0180, ..., 0.0174, -0.0419, 0.0177], + [-0.0681, -0.0269, -0.0825, ..., -0.0306, -0.0362, -0.0710], + [-0.1443, -0.2047, 0.0276, ..., -0.0660, 0.0407, 0.0481]], + device='cuda:0'), grad: tensor([[ 7.7020e-07, 0.0000e+00, 7.1852e-07, ..., 9.3132e-10, + 3.9022e-07, 0.0000e+00], + [ 2.0284e-06, 0.0000e+00, 2.3330e-07, ..., 9.3132e-10, + -5.2787e-06, 0.0000e+00], + [-4.7907e-06, 0.0000e+00, 2.8610e-06, ..., 0.0000e+00, + 5.4017e-07, 0.0000e+00], + ..., + [ 1.9036e-06, 0.0000e+00, 1.4855e-07, ..., 1.4435e-08, + -5.1735e-07, 0.0000e+00], + [-2.0768e-06, 0.0000e+00, -5.4352e-06, ..., 2.3283e-09, + 1.2089e-06, 0.0000e+00], + [ 2.7716e-06, 0.0000e+00, 7.6648e-07, ..., -1.5832e-08, + 8.6613e-07, 0.0000e+00]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0065, 0.0237, -0.0022, -0.0051, 0.0148, 0.0055, 0.0248, -0.0118, + 0.0056, -0.0057], device='cuda:0'), grad: tensor([ 6.5342e-06, -1.7121e-05, 5.6028e-06, -6.1607e-07, 1.5572e-05, + 1.8459e-06, -5.3868e-06, 2.3749e-07, -1.7524e-05, 1.0848e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 230, time 217.45, cls_loss 0.0009 cls_loss_mapping 0.0008 cls_loss_causal 0.4844 re_mapping 0.0045 re_causal 0.0164 /// teacc 99.04 lr 0.00001000 +Epoch 232, weight, value: tensor([[-0.0560, -0.1090, -0.0601, ..., -0.1173, -0.0207, -0.0325], + [-0.0507, 0.0501, -0.1622, ..., 0.0629, 0.0720, -0.1374], + [ 0.0574, -0.0076, -0.0632, ..., -0.0600, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0181, ..., 0.0174, -0.0419, 0.0176], + [-0.0681, -0.0270, -0.0825, ..., -0.0307, -0.0363, -0.0710], + [-0.1443, -0.2047, 0.0276, ..., -0.0660, 0.0407, 0.0481]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 1.8626e-09, 1.2107e-08, ..., 0.0000e+00, + 3.9581e-08, 0.0000e+00], + [-6.3656e-07, -1.3709e-06, 2.4214e-08, ..., 0.0000e+00, + -2.9728e-06, 0.0000e+00], + [ 1.9465e-07, 3.1060e-07, 9.3132e-10, ..., 0.0000e+00, + 7.3761e-07, -4.6566e-10], + ..., + [ 3.9767e-07, 9.4064e-07, 6.9849e-09, ..., 0.0000e+00, + 1.4007e-06, 4.6566e-10], + [-6.9384e-08, 1.3970e-09, 2.7940e-08, ..., 0.0000e+00, + 1.1036e-07, 0.0000e+00], + [ 1.3504e-07, 9.3132e-10, 1.5311e-06, ..., 0.0000e+00, + -6.1467e-08, 0.0000e+00]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0064, 0.0238, -0.0022, -0.0050, 0.0148, 0.0055, 0.0248, -0.0118, + 0.0056, -0.0057], device='cuda:0'), grad: tensor([-2.6077e-07, -1.1854e-05, 3.8091e-06, 5.7835e-07, 1.8813e-06, + -2.9672e-06, 9.7416e-07, 6.3889e-06, -3.2075e-06, 4.6380e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 231, time 217.33, cls_loss 0.0009 cls_loss_mapping 0.0008 cls_loss_causal 0.4745 re_mapping 0.0045 re_causal 0.0162 /// teacc 99.05 lr 0.00001000 +Epoch 233, weight, value: tensor([[-0.0560, -0.1091, -0.0600, ..., -0.1174, -0.0207, -0.0325], + [-0.0508, 0.0502, -0.1622, ..., 0.0629, 0.0720, -0.1374], + [ 0.0575, -0.0076, -0.0631, ..., -0.0601, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0182, ..., 0.0174, -0.0419, 0.0176], + [-0.0682, -0.0271, -0.0825, ..., -0.0308, -0.0364, -0.0710], + [-0.1444, -0.2047, 0.0276, ..., -0.0660, 0.0407, 0.0481]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 9.3132e-10, 3.2596e-09, ..., 0.0000e+00, + 2.4680e-08, 0.0000e+00], + [ 1.0934e-06, -2.8266e-07, 1.5832e-08, ..., 4.6566e-10, + -2.2259e-07, 0.0000e+00], + [-8.6986e-07, 1.7509e-07, 3.7253e-09, ..., 0.0000e+00, + 1.9185e-07, 0.0000e+00], + ..., + [-2.7614e-07, 9.8255e-08, 1.6764e-08, ..., 4.6566e-10, + 1.9511e-07, 0.0000e+00], + [ 1.4482e-07, 4.6566e-10, 1.3039e-08, ..., 0.0000e+00, + 1.3318e-07, 0.0000e+00], + [ 1.7695e-07, 0.0000e+00, -6.0536e-09, ..., 4.6566e-10, + -2.3143e-07, 0.0000e+00]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0064, 0.0237, -0.0022, -0.0051, 0.0148, 0.0055, 0.0248, -0.0118, + 0.0056, -0.0057], device='cuda:0'), grad: tensor([-9.6206e-07, 1.1604e-06, -5.0291e-07, -6.1467e-07, 7.1200e-07, + 5.9977e-07, -1.1763e-06, -5.8673e-08, 6.4075e-07, 2.2445e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 232, time 217.63, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4912 re_mapping 0.0045 re_causal 0.0163 /// teacc 99.05 lr 0.00001000 +Epoch 234, weight, value: tensor([[-0.0560, -0.1091, -0.0600, ..., -0.1175, -0.0208, -0.0325], + [-0.0508, 0.0502, -0.1623, ..., 0.0629, 0.0720, -0.1374], + [ 0.0575, -0.0076, -0.0631, ..., -0.0601, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0182, ..., 0.0174, -0.0419, 0.0176], + [-0.0683, -0.0271, -0.0826, ..., -0.0308, -0.0365, -0.0710], + [-0.1445, -0.2047, 0.0275, ..., -0.0660, 0.0408, 0.0481]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 7.4506e-09, 3.9116e-08, ..., 0.0000e+00, + 4.6566e-08, 1.1176e-08], + [ 1.8841e-06, 3.7998e-07, 2.8126e-07, ..., 0.0000e+00, + 2.3842e-07, 1.8626e-09], + [ 7.9721e-07, 1.8626e-07, 4.1910e-08, ..., 0.0000e+00, + 1.0245e-08, 9.3132e-10], + ..., + [-3.7961e-06, -8.1025e-07, 1.2480e-06, ..., 0.0000e+00, + 1.0896e-07, 0.0000e+00], + [ 4.4703e-08, 4.6566e-09, 1.4901e-08, ..., 0.0000e+00, + -5.3272e-07, 3.4459e-08], + [ 5.1316e-07, 9.3132e-08, 6.3293e-06, ..., 9.3132e-10, + 1.9558e-08, 9.3132e-10]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0064, 0.0237, -0.0021, -0.0051, 0.0148, 0.0056, 0.0248, -0.0117, + 0.0054, -0.0057], device='cuda:0'), grad: tensor([-1.9632e-06, 7.5102e-06, 2.1011e-06, 1.5208e-06, -1.8805e-05, + 1.8440e-06, -2.1793e-07, -5.1335e-06, -3.6955e-06, 1.6794e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 233, time 217.48, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4746 re_mapping 0.0044 re_causal 0.0160 /// teacc 99.04 lr 0.00001000 +Epoch 235, weight, value: tensor([[-0.0561, -0.1092, -0.0600, ..., -0.1175, -0.0208, -0.0325], + [-0.0509, 0.0502, -0.1624, ..., 0.0629, 0.0721, -0.1374], + [ 0.0577, -0.0076, -0.0630, ..., -0.0601, -0.0449, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0183, ..., 0.0174, -0.0419, 0.0176], + [-0.0684, -0.0271, -0.0826, ..., -0.0308, -0.0366, -0.0710], + [-0.1445, -0.2047, 0.0274, ..., -0.0661, 0.0408, 0.0481]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 2.7940e-09, 6.5193e-09, ..., 0.0000e+00, + 3.9116e-08, 0.0000e+00], + [ 6.7521e-07, 1.0058e-07, 1.8626e-08, ..., 0.0000e+00, + -3.9116e-08, 0.0000e+00], + [ 8.5589e-07, 2.2538e-07, -8.3819e-09, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + ..., + [-2.7753e-07, 6.8918e-08, 2.5146e-08, ..., 0.0000e+00, + 7.7300e-08, 4.6566e-09], + [-4.3176e-06, -1.2023e-06, 1.2387e-07, ..., 0.0000e+00, + 7.9162e-08, 0.0000e+00], + [ 6.6776e-07, 7.4506e-09, 8.1025e-08, ..., 0.0000e+00, + -1.7136e-07, -5.5879e-09]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0064, 0.0237, -0.0021, -0.0051, 0.0148, 0.0056, 0.0248, -0.0117, + 0.0053, -0.0058], device='cuda:0'), grad: tensor([ 1.6764e-07, 3.3490e-06, 5.7146e-06, 7.9721e-07, 8.7172e-07, + 8.1435e-06, 8.4713e-06, -1.3318e-07, -2.8715e-05, 1.3076e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 234, time 217.43, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4849 re_mapping 0.0044 re_causal 0.0165 /// teacc 99.03 lr 0.00001000 +Epoch 236, weight, value: tensor([[-0.0560, -0.1092, -0.0600, ..., -0.1176, -0.0208, -0.0325], + [-0.0509, 0.0502, -0.1625, ..., 0.0630, 0.0721, -0.1374], + [ 0.0578, -0.0076, -0.0629, ..., -0.0601, -0.0450, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0184, ..., 0.0173, -0.0420, 0.0176], + [-0.0686, -0.0271, -0.0827, ..., -0.0309, -0.0368, -0.0710], + [-0.1446, -0.2048, 0.0273, ..., -0.0661, 0.0409, 0.0481]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 9.3132e-10, -2.7940e-09, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [-9.7696e-07, -1.7183e-06, 2.7008e-08, ..., 0.0000e+00, + -2.5313e-06, 0.0000e+00], + [ 6.7800e-07, 9.8720e-07, 6.5193e-09, ..., 0.0000e+00, + 1.3402e-06, 0.0000e+00], + ..., + [-1.2135e-06, 6.8825e-07, -7.0781e-08, ..., 0.0000e+00, + 1.0477e-06, 0.0000e+00], + [ 4.3493e-07, 9.3132e-10, 9.6858e-08, ..., 0.0000e+00, + 3.2596e-08, 0.0000e+00], + [ 1.0971e-06, 0.0000e+00, 1.8813e-07, ..., 9.3132e-10, + -5.9605e-08, 0.0000e+00]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0063, 0.0237, -0.0020, -0.0051, 0.0148, 0.0056, 0.0248, -0.0118, + 0.0052, -0.0058], device='cuda:0'), grad: tensor([-2.3283e-07, -4.9025e-06, 2.6934e-06, -2.4587e-07, 5.1595e-07, + 5.3085e-08, 1.3411e-07, -2.5406e-06, 1.4957e-06, 3.0212e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 235, time 217.60, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4786 re_mapping 0.0045 re_causal 0.0162 /// teacc 99.04 lr 0.00001000 +Epoch 237, weight, value: tensor([[-0.0561, -0.1093, -0.0601, ..., -0.1177, -0.0209, -0.0325], + [-0.0509, 0.0502, -0.1626, ..., 0.0629, 0.0722, -0.1374], + [ 0.0578, -0.0077, -0.0629, ..., -0.0601, -0.0451, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0185, ..., 0.0173, -0.0420, 0.0175], + [-0.0687, -0.0271, -0.0827, ..., -0.0310, -0.0369, -0.0711], + [-0.1446, -0.2048, 0.0274, ..., -0.0661, 0.0410, 0.0481]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 1.8720e-07, 0.0000e+00, 1.3970e-08, ..., 4.6566e-09, + -2.3376e-07, 0.0000e+00], + [-2.7474e-07, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 5.1223e-08, 0.0000e+00], + ..., + [-4.1723e-07, 0.0000e+00, 3.7253e-09, ..., -2.3283e-08, + 2.1700e-07, 0.0000e+00], + [ 6.7055e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 4.0047e-08, 0.0000e+00], + [ 7.6368e-08, 0.0000e+00, -1.1176e-08, ..., 3.7253e-09, + -1.0272e-06, 0.0000e+00]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0063, 0.0237, -0.0020, -0.0051, 0.0147, 0.0057, 0.0249, -0.0118, + 0.0052, -0.0057], device='cuda:0'), grad: tensor([-4.1425e-06, 2.0675e-07, -2.4866e-07, 1.0775e-06, 3.2783e-06, + 3.0287e-06, 7.1060e-07, -1.2517e-06, 3.1572e-07, -2.9635e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 236, time 217.65, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4453 re_mapping 0.0043 re_causal 0.0157 /// teacc 99.08 lr 0.00001000 +Epoch 238, weight, value: tensor([[-0.0561, -0.1093, -0.0600, ..., -0.1178, -0.0210, -0.0325], + [-0.0509, 0.0502, -0.1626, ..., 0.0629, 0.0722, -0.1375], + [ 0.0579, -0.0077, -0.0629, ..., -0.0602, -0.0451, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0185, ..., 0.0174, -0.0420, 0.0175], + [-0.0688, -0.0271, -0.0828, ..., -0.0310, -0.0370, -0.0711], + [-0.1447, -0.2048, 0.0273, ..., -0.0661, 0.0411, 0.0481]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, 1.6671e-07, ..., 9.3132e-10, + 1.0245e-08, 0.0000e+00], + [ 1.5553e-07, 0.0000e+00, 9.8720e-08, ..., 1.8626e-09, + 3.6322e-08, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 1.5739e-07, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + ..., + [-9.9652e-08, 0.0000e+00, 4.9081e-07, ..., 1.6671e-07, + 6.8732e-07, 0.0000e+00], + [ 3.7253e-08, 0.0000e+00, 8.0932e-07, ..., 1.8626e-09, + -3.3900e-07, 0.0000e+00], + [-6.0536e-08, 0.0000e+00, -8.2795e-07, ..., -2.1048e-07, + -5.6252e-07, 0.0000e+00]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0063, 0.0237, -0.0020, -0.0051, 0.0147, 0.0057, 0.0249, -0.0118, + 0.0051, -0.0057], device='cuda:0'), grad: tensor([-6.0536e-06, 1.3392e-06, 4.1537e-07, 3.9041e-05, 8.0913e-06, + -4.2588e-05, 1.8114e-06, 5.6326e-06, 1.7956e-06, -9.5442e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 237, time 217.36, cls_loss 0.0008 cls_loss_mapping 0.0008 cls_loss_causal 0.4947 re_mapping 0.0045 re_causal 0.0168 /// teacc 99.07 lr 0.00001000 +Epoch 239, weight, value: tensor([[-0.0562, -0.1093, -0.0601, ..., -0.1178, -0.0210, -0.0325], + [-0.0509, 0.0503, -0.1627, ..., 0.0629, 0.0723, -0.1375], + [ 0.0579, -0.0077, -0.0629, ..., -0.0602, -0.0452, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0186, ..., 0.0174, -0.0421, 0.0175], + [-0.0689, -0.0271, -0.0828, ..., -0.0311, -0.0370, -0.0711], + [-0.1447, -0.2048, 0.0272, ..., -0.0661, 0.0411, 0.0481]], + device='cuda:0'), grad: tensor([[ 7.6368e-08, 0.0000e+00, 8.6613e-08, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00], + [ 8.9407e-08, 0.0000e+00, 2.9802e-08, ..., 0.0000e+00, + -6.5193e-08, 0.0000e+00], + [-2.7940e-08, 0.0000e+00, 1.2107e-08, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [-1.8626e-08, 0.0000e+00, 2.5146e-08, ..., 0.0000e+00, + 4.1910e-08, 0.0000e+00], + [ 1.5832e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 6.0536e-08, 0.0000e+00, -9.1176e-07, ..., 0.0000e+00, + -3.8650e-07, 0.0000e+00]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0063, 0.0237, -0.0020, -0.0051, 0.0147, 0.0057, 0.0249, -0.0118, + 0.0051, -0.0057], device='cuda:0'), grad: tensor([ 2.4214e-06, 1.5926e-07, 8.4750e-07, -1.7695e-08, 4.8876e-06, + 6.4634e-07, -6.2585e-06, 1.3970e-08, 3.7253e-08, -2.7139e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 238, time 217.38, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4748 re_mapping 0.0045 re_causal 0.0163 /// teacc 99.08 lr 0.00001000 +Epoch 240, weight, value: tensor([[-0.0561, -0.1093, -0.0601, ..., -0.1179, -0.0210, -0.0325], + [-0.0509, 0.0503, -0.1627, ..., 0.0632, 0.0725, -0.1375], + [ 0.0580, -0.0077, -0.0629, ..., -0.0602, -0.0453, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0187, ..., 0.0171, -0.0423, 0.0175], + [-0.0690, -0.0271, -0.0828, ..., -0.0312, -0.0371, -0.0711], + [-0.1448, -0.2048, 0.0271, ..., -0.0661, 0.0412, 0.0481]], + device='cuda:0'), grad: tensor([[ 9.2108e-07, 0.0000e+00, 1.4901e-08, ..., 2.0489e-08, + 6.0536e-08, 0.0000e+00], + [ 1.9372e-07, 0.0000e+00, 1.3690e-07, ..., 1.7229e-07, + -9.3132e-10, 0.0000e+00], + [-2.3954e-06, 0.0000e+00, 5.9605e-08, ..., 9.6858e-08, + 8.0094e-08, 0.0000e+00], + ..., + [-3.3453e-06, 0.0000e+00, -6.5006e-06, ..., -1.0706e-05, + 4.2841e-08, 0.0000e+00], + [ 4.1723e-07, 0.0000e+00, 2.4214e-08, ..., 5.5879e-09, + 6.9849e-08, 0.0000e+00], + [ 2.5146e-07, 0.0000e+00, 6.9197e-07, ..., 2.8126e-07, + -6.7987e-08, 0.0000e+00]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0063, 0.0238, -0.0020, -0.0051, 0.0147, 0.0057, 0.0249, -0.0119, + 0.0050, -0.0058], device='cuda:0'), grad: tensor([ 2.0154e-06, 8.3912e-07, -5.2452e-06, 2.4885e-06, 1.5751e-05, + 4.3623e-06, -1.8626e-07, -2.3693e-05, 1.2135e-06, 2.4140e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 239, time 217.94, cls_loss 0.0009 cls_loss_mapping 0.0009 cls_loss_causal 0.4928 re_mapping 0.0045 re_causal 0.0165 /// teacc 99.09 lr 0.00001000 +Epoch 241, weight, value: tensor([[-0.0561, -0.1094, -0.0601, ..., -0.1179, -0.0210, -0.0325], + [-0.0510, 0.0503, -0.1628, ..., 0.0634, 0.0727, -0.1375], + [ 0.0581, -0.0077, -0.0628, ..., -0.0603, -0.0453, 0.0088], + ..., + [ 0.0574, -0.0602, -0.0187, ..., 0.0170, -0.0425, 0.0174], + [-0.0691, -0.0271, -0.0828, ..., -0.0312, -0.0372, -0.0711], + [-0.1448, -0.2048, 0.0270, ..., -0.0661, 0.0412, 0.0481]], + device='cuda:0'), grad: tensor([[ 1.3318e-07, 0.0000e+00, 4.7870e-07, ..., 0.0000e+00, + 8.3167e-07, 0.0000e+00], + [ 3.0641e-07, 0.0000e+00, 3.7439e-07, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00], + [-3.0976e-06, 0.0000e+00, 1.9483e-06, ..., 0.0000e+00, + 3.1944e-07, 0.0000e+00], + ..., + [ 2.5146e-06, 0.0000e+00, 4.3958e-07, ..., 0.0000e+00, + 4.5635e-08, 0.0000e+00], + [-4.4424e-07, 0.0000e+00, -3.5241e-06, ..., 0.0000e+00, + 1.4156e-07, 0.0000e+00], + [ 4.0699e-07, 0.0000e+00, 5.8860e-07, ..., 0.0000e+00, + -8.3819e-08, 0.0000e+00]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0063, 0.0239, -0.0019, -0.0052, 0.0148, 0.0057, 0.0249, -0.0119, + 0.0050, -0.0058], device='cuda:0'), grad: tensor([ 4.8317e-06, 1.8971e-06, -1.1455e-07, 4.8056e-07, -1.3271e-06, + 2.4453e-05, -2.7820e-05, 8.4937e-06, -1.3821e-05, 2.8722e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 240, time 218.12, cls_loss 0.0009 cls_loss_mapping 0.0008 cls_loss_causal 0.4641 re_mapping 0.0044 re_causal 0.0161 /// teacc 99.09 lr 0.00001000 +Epoch 242, weight, value: tensor([[-0.0562, -0.1094, -0.0602, ..., -0.1180, -0.0210, -0.0325], + [-0.0509, 0.0504, -0.1629, ..., 0.0634, 0.0727, -0.1375], + [ 0.0581, -0.0077, -0.0628, ..., -0.0603, -0.0454, 0.0088], + ..., + [ 0.0574, -0.0603, -0.0189, ..., 0.0170, -0.0425, 0.0174], + [-0.0692, -0.0271, -0.0829, ..., -0.0313, -0.0373, -0.0711], + [-0.1449, -0.2048, 0.0270, ..., -0.0661, 0.0413, 0.0481]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 4.2003e-07, 0.0000e+00, 1.4435e-07, ..., 3.7253e-09, + -1.0338e-07, 0.0000e+00], + [-7.1246e-07, 0.0000e+00, -2.4214e-08, ..., 9.3132e-10, + 1.4901e-08, 0.0000e+00], + ..., + [-9.2201e-08, 0.0000e+00, -4.0978e-08, ..., -4.0047e-08, + 6.3889e-07, 0.0000e+00], + [ 4.1910e-08, 0.0000e+00, -2.9802e-08, ..., 0.0000e+00, + 7.0781e-08, 0.0000e+00], + [ 1.8161e-07, 0.0000e+00, 1.7136e-07, ..., 2.4214e-08, + -1.0300e-06, 0.0000e+00]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0063, 0.0239, -0.0019, -0.0051, 0.0147, 0.0057, 0.0248, -0.0120, + 0.0049, -0.0057], device='cuda:0'), grad: tensor([ 1.1269e-07, 1.0245e-06, -9.7975e-07, -6.3796e-07, 4.9639e-07, + 1.2433e-06, 1.0058e-07, 1.3551e-06, 6.0536e-08, -2.7735e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 241, time 218.15, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4478 re_mapping 0.0045 re_causal 0.0161 /// teacc 99.10 lr 0.00001000 +Epoch 243, weight, value: tensor([[-0.0562, -0.1094, -0.0603, ..., -0.1181, -0.0211, -0.0326], + [-0.0509, 0.0505, -0.1630, ..., 0.0634, 0.0728, -0.1376], + [ 0.0582, -0.0077, -0.0628, ..., -0.0603, -0.0455, 0.0087], + ..., + [ 0.0574, -0.0603, -0.0189, ..., 0.0170, -0.0426, 0.0174], + [-0.0694, -0.0271, -0.0830, ..., -0.0314, -0.0374, -0.0711], + [-0.1449, -0.2049, 0.0270, ..., -0.0661, 0.0415, 0.0481]], + device='cuda:0'), grad: tensor([[ 4.8429e-08, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 9.1922e-07, 0.0000e+00, 6.0536e-08, ..., 0.0000e+00, + -5.3085e-08, 0.0000e+00], + [-6.2399e-08, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + -5.5879e-08, 0.0000e+00], + ..., + [-1.2908e-06, 0.0000e+00, 9.4995e-08, ..., 0.0000e+00, + 7.9162e-08, 0.0000e+00], + [ 6.7055e-08, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 4.1910e-08, 0.0000e+00, 2.3022e-05, ..., 0.0000e+00, + -5.5879e-08, 0.0000e+00]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0063, 0.0239, -0.0019, -0.0051, 0.0147, 0.0058, 0.0247, -0.0120, + 0.0048, -0.0057], device='cuda:0'), grad: tensor([ 1.2945e-07, 1.6727e-06, -3.2783e-07, 6.6124e-07, -5.6118e-05, + -1.3923e-06, 1.3961e-06, -1.9539e-06, 6.9849e-08, 5.5909e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 242, time 217.94, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4683 re_mapping 0.0046 re_causal 0.0163 /// teacc 99.09 lr 0.00001000 +Epoch 244, weight, value: tensor([[-0.0562, -0.1095, -0.0603, ..., -0.1181, -0.0211, -0.0326], + [-0.0509, 0.0508, -0.1631, ..., 0.0634, 0.0728, -0.1376], + [ 0.0583, -0.0077, -0.0628, ..., -0.0603, -0.0455, 0.0087], + ..., + [ 0.0574, -0.0605, -0.0191, ..., 0.0170, -0.0426, 0.0174], + [-0.0695, -0.0271, -0.0830, ..., -0.0314, -0.0374, -0.0711], + [-0.1450, -0.2049, 0.0268, ..., -0.0661, 0.0415, 0.0481]], + device='cuda:0'), grad: tensor([[ 5.7742e-08, 0.0000e+00, 1.2107e-07, ..., 0.0000e+00, + 1.6298e-07, 0.0000e+00], + [ 3.0547e-07, 0.0000e+00, 6.8918e-08, ..., 0.0000e+00, + -1.2144e-06, 0.0000e+00], + [ 1.3877e-07, 0.0000e+00, 6.1933e-07, ..., 0.0000e+00, + 1.3905e-06, 0.0000e+00], + ..., + [-2.1011e-06, 0.0000e+00, 3.8184e-08, ..., 0.0000e+00, + 7.4320e-07, 6.5193e-09], + [-9.3132e-09, 0.0000e+00, 4.1164e-07, ..., 0.0000e+00, + 1.4901e-07, 9.3132e-10], + [ 9.5554e-07, 0.0000e+00, 2.2817e-07, ..., 0.0000e+00, + -3.0920e-07, -2.8871e-08]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0062, 0.0239, -0.0018, -0.0051, 0.0148, 0.0058, 0.0247, -0.0120, + 0.0048, -0.0058], device='cuda:0'), grad: tensor([ 1.6363e-06, -1.4044e-06, 9.6411e-06, 2.4773e-06, 6.1169e-06, + -5.4110e-07, -1.8746e-05, -3.8296e-06, 1.9372e-06, 2.7083e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 243, time 217.97, cls_loss 0.0008 cls_loss_mapping 0.0007 cls_loss_causal 0.4854 re_mapping 0.0045 re_causal 0.0166 /// teacc 99.10 lr 0.00001000 +Epoch 245, weight, value: tensor([[-0.0563, -0.1095, -0.0603, ..., -0.1181, -0.0211, -0.0326], + [-0.0509, 0.0509, -0.1632, ..., 0.0634, 0.0729, -0.1376], + [ 0.0584, -0.0077, -0.0628, ..., -0.0604, -0.0455, 0.0088], + ..., + [ 0.0574, -0.0606, -0.0191, ..., 0.0170, -0.0426, 0.0174], + [-0.0698, -0.0271, -0.0830, ..., -0.0314, -0.0375, -0.0711], + [-0.1450, -0.2049, 0.0267, ..., -0.0661, 0.0415, 0.0481]], + device='cuda:0'), grad: tensor([[ 2.0582e-07, 0.0000e+00, 8.7544e-08, ..., 0.0000e+00, + 1.8720e-07, 0.0000e+00], + [ 2.5984e-07, 0.0000e+00, 1.6298e-07, ..., 0.0000e+00, + -5.3309e-06, 0.0000e+00], + [-8.6427e-06, 0.0000e+00, -4.4517e-07, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [ 7.5400e-06, 0.0000e+00, 3.9302e-07, ..., 0.0000e+00, + 3.6508e-06, 0.0000e+00], + [ 4.4610e-07, 0.0000e+00, 3.0082e-07, ..., 0.0000e+00, + 1.5274e-07, 0.0000e+00], + [ 7.1712e-08, 0.0000e+00, 8.5495e-07, ..., 0.0000e+00, + 9.6578e-07, 0.0000e+00]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0062, 0.0239, -0.0018, -0.0050, 0.0148, 0.0057, 0.0248, -0.0120, + 0.0047, -0.0058], device='cuda:0'), grad: tensor([-4.5542e-07, -1.5318e-05, -1.4447e-05, 3.1665e-07, -2.4922e-06, + 6.5193e-09, -4.9360e-08, 2.4512e-05, 1.8803e-06, 6.0573e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 244, time 217.95, cls_loss 0.0008 cls_loss_mapping 0.0008 cls_loss_causal 0.4894 re_mapping 0.0045 re_causal 0.0166 /// teacc 99.10 lr 0.00001000 +Epoch 246, weight, value: tensor([[-0.0563, -0.1096, -0.0603, ..., -0.1181, -0.0211, -0.0326], + [-0.0509, 0.0511, -0.1632, ..., 0.0634, 0.0730, -0.1376], + [ 0.0585, -0.0078, -0.0628, ..., -0.0604, -0.0455, 0.0088], + ..., + [ 0.0573, -0.0608, -0.0193, ..., 0.0169, -0.0427, 0.0173], + [-0.0701, -0.0271, -0.0830, ..., -0.0315, -0.0375, -0.0711], + [-0.1451, -0.2049, 0.0266, ..., -0.0662, 0.0416, 0.0481]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 0.0000e+00, -7.1712e-08, ..., 9.3132e-10, + 1.4901e-08, 8.3819e-09], + [ 1.7975e-07, 0.0000e+00, 6.1616e-06, ..., 9.1735e-07, + 1.8338e-06, 4.6566e-09], + [ 9.0338e-07, 0.0000e+00, 4.8429e-08, ..., 3.7253e-09, + 2.7008e-08, 8.3819e-08], + ..., + [-4.5914e-07, 0.0000e+00, 2.3935e-07, ..., 6.5193e-09, + 1.5926e-07, 2.1420e-08], + [ 3.5390e-08, 0.0000e+00, 7.2643e-08, ..., 0.0000e+00, + 1.1176e-08, 3.7253e-09], + [ 8.1956e-08, 0.0000e+00, 3.5651e-06, ..., 4.9639e-07, + 1.1204e-06, 7.4506e-09]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0062, 0.0240, -0.0017, -0.0050, 0.0148, 0.0057, 0.0247, -0.0121, + 0.0046, -0.0058], device='cuda:0'), grad: tensor([-3.7961e-06, 1.4223e-05, 1.6093e-06, -1.5646e-06, -2.2992e-05, + 1.1595e-06, 1.9930e-06, 2.4866e-07, 2.8033e-07, 8.8140e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 245, time 217.97, cls_loss 0.0010 cls_loss_mapping 0.0007 cls_loss_causal 0.4716 re_mapping 0.0044 re_causal 0.0160 /// teacc 99.06 lr 0.00001000 +Epoch 247, weight, value: tensor([[-0.0564, -0.1097, -0.0602, ..., -0.1182, -0.0212, -0.0326], + [-0.0510, 0.0512, -0.1634, ..., 0.0634, 0.0731, -0.1376], + [ 0.0586, -0.0077, -0.0628, ..., -0.0604, -0.0456, 0.0087], + ..., + [ 0.0574, -0.0609, -0.0194, ..., 0.0169, -0.0428, 0.0172], + [-0.0703, -0.0271, -0.0831, ..., -0.0315, -0.0377, -0.0711], + [-0.1452, -0.2049, 0.0266, ..., -0.0662, 0.0417, 0.0482]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 1.5944e-06, 1.8626e-09], + [ 1.2852e-07, 0.0000e+00, 8.3819e-08, ..., 0.0000e+00, + -1.0617e-07, 2.7940e-09], + [-7.4506e-09, 0.0000e+00, 4.0047e-08, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + ..., + [-6.7055e-08, 0.0000e+00, 7.7300e-08, ..., 5.5879e-09, + 7.9442e-07, 1.2293e-07], + [ 1.4901e-08, 0.0000e+00, 1.3970e-08, ..., 0.0000e+00, + 3.7253e-08, 0.0000e+00], + [ 3.4459e-08, 0.0000e+00, 3.1114e-05, ..., -2.2352e-08, + -2.0117e-06, -3.1479e-07]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0061, 0.0240, -0.0017, -0.0050, 0.0148, 0.0057, 0.0248, -0.0121, + 0.0044, -0.0059], device='cuda:0'), grad: tensor([ 1.9282e-05, 2.9895e-07, 1.5646e-07, -8.5682e-08, -9.7990e-05, + 3.8743e-06, -2.3216e-05, 3.5446e-06, 2.1420e-07, 9.3937e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 246, time 217.96, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4909 re_mapping 0.0044 re_causal 0.0165 /// teacc 99.06 lr 0.00001000 +Epoch 248, weight, value: tensor([[-0.0564, -0.1098, -0.0602, ..., -0.1182, -0.0213, -0.0326], + [-0.0509, 0.0514, -0.1635, ..., 0.0636, 0.0733, -0.1376], + [ 0.0586, -0.0078, -0.0628, ..., -0.0604, -0.0457, 0.0087], + ..., + [ 0.0573, -0.0610, -0.0195, ..., 0.0168, -0.0430, 0.0171], + [-0.0705, -0.0271, -0.0832, ..., -0.0316, -0.0378, -0.0711], + [-0.1452, -0.2049, 0.0264, ..., -0.0662, 0.0417, 0.0482]], + device='cuda:0'), grad: tensor([[ 7.4506e-08, 6.5193e-09, -9.6858e-08, ..., 0.0000e+00, + 1.1455e-07, 0.0000e+00], + [-1.0006e-05, -3.0398e-06, 2.2743e-06, ..., 0.0000e+00, + 2.7474e-07, 0.0000e+00], + [ 8.7768e-06, 2.6524e-06, 4.0047e-08, ..., 0.0000e+00, + 2.4717e-06, 0.0000e+00], + ..., + [ 1.4026e-06, 3.5856e-07, 9.4995e-08, ..., 0.0000e+00, + 3.8743e-07, 0.0000e+00], + [ 3.7160e-07, 1.8626e-09, 1.4693e-05, ..., 0.0000e+00, + 2.1145e-05, 0.0000e+00], + [ 9.4995e-08, 0.0000e+00, 4.7907e-06, ..., 0.0000e+00, + -6.0536e-08, 0.0000e+00]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0061, 0.0241, -0.0017, -0.0049, 0.0148, 0.0056, 0.0248, -0.0122, + 0.0044, -0.0059], device='cuda:0'), grad: tensor([-1.3225e-07, -2.0951e-05, 3.2604e-05, -8.4937e-07, -1.3880e-05, + -1.9765e-04, 7.2956e-05, 5.0738e-06, 1.0872e-04, 1.3739e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 247, time 217.99, cls_loss 0.0012 cls_loss_mapping 0.0010 cls_loss_causal 0.4750 re_mapping 0.0044 re_causal 0.0158 /// teacc 99.03 lr 0.00001000 +Epoch 249, weight, value: tensor([[-0.0564, -0.1098, -0.0603, ..., -0.1183, -0.0214, -0.0326], + [-0.0510, 0.0514, -0.1637, ..., 0.0636, 0.0735, -0.1377], + [ 0.0586, -0.0078, -0.0628, ..., -0.0605, -0.0458, 0.0086], + ..., + [ 0.0574, -0.0611, -0.0197, ..., 0.0167, -0.0431, 0.0171], + [-0.0706, -0.0272, -0.0832, ..., -0.0316, -0.0380, -0.0711], + [-0.1453, -0.2050, 0.0264, ..., -0.0664, 0.0419, 0.0482]], + device='cuda:0'), grad: tensor([[ 4.7777e-07, 9.3132e-10, 1.1176e-08, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00], + [ 3.9767e-07, -4.0699e-07, 1.5832e-08, ..., 0.0000e+00, + -8.4192e-07, 0.0000e+00], + [-1.7695e-06, 1.0245e-08, -1.1642e-07, ..., 0.0000e+00, + 3.4459e-08, 0.0000e+00], + ..., + [ 3.7067e-07, 3.8184e-07, 2.4214e-08, ..., 0.0000e+00, + 7.2829e-07, 0.0000e+00], + [ 3.8221e-06, 5.5879e-09, 1.2107e-08, ..., 0.0000e+00, + 4.0047e-08, 0.0000e+00], + [ 9.4995e-08, 0.0000e+00, 1.5832e-08, ..., 0.0000e+00, + -1.0151e-07, 0.0000e+00]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0061, 0.0242, -0.0017, -0.0048, 0.0148, 0.0056, 0.0247, -0.0123, + 0.0042, -0.0059], device='cuda:0'), grad: tensor([ 5.9325e-07, -1.1474e-06, -7.4543e-06, -6.0275e-06, 5.3085e-07, + 2.7977e-06, -6.1467e-08, 1.8887e-06, 9.1195e-06, -2.4028e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 248, time 217.94, cls_loss 0.0009 cls_loss_mapping 0.0008 cls_loss_causal 0.4825 re_mapping 0.0043 re_causal 0.0161 /// teacc 99.07 lr 0.00001000 +Epoch 250, weight, value: tensor([[-0.0565, -0.1099, -0.0603, ..., -0.1183, -0.0214, -0.0326], + [-0.0510, 0.0515, -0.1638, ..., 0.0636, 0.0735, -0.1377], + [ 0.0587, -0.0078, -0.0628, ..., -0.0605, -0.0459, 0.0086], + ..., + [ 0.0574, -0.0611, -0.0198, ..., 0.0167, -0.0432, 0.0171], + [-0.0707, -0.0272, -0.0833, ..., -0.0317, -0.0379, -0.0711], + [-0.1453, -0.2050, 0.0263, ..., -0.0664, 0.0419, 0.0482]], + device='cuda:0'), grad: tensor([[ 1.6857e-07, 0.0000e+00, 1.2107e-08, ..., 0.0000e+00, + 8.0094e-08, 0.0000e+00], + [ 3.0454e-07, 0.0000e+00, 7.5437e-08, ..., 0.0000e+00, + -2.2631e-07, 0.0000e+00], + [-6.8098e-06, 0.0000e+00, 7.0781e-08, ..., 0.0000e+00, + 3.0734e-08, 0.0000e+00], + ..., + [ 6.0499e-06, 0.0000e+00, 9.1270e-08, ..., 0.0000e+00, + 8.4750e-08, 0.0000e+00], + [ 1.1921e-07, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 8.5682e-08, 0.0000e+00, 4.1910e-08, ..., 0.0000e+00, + 6.8918e-08, 0.0000e+00]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0060, 0.0242, -0.0017, -0.0049, 0.0149, 0.0056, 0.0246, -0.0122, + 0.0042, -0.0059], device='cuda:0'), grad: tensor([ 4.5355e-07, 3.3993e-07, -9.6262e-06, 1.1735e-07, 3.1199e-07, + 3.3528e-07, -9.0897e-07, 8.1509e-06, 2.8033e-07, 5.3365e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 249, time 217.65, cls_loss 0.0009 cls_loss_mapping 0.0008 cls_loss_causal 0.4359 re_mapping 0.0043 re_causal 0.0156 /// teacc 99.02 lr 0.00001000 +---------------------saving last model at epoch 249---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep250_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_epoch250', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep250_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_epoch250/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last +randm: False +stride: 3 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.879997 98.769997 ... 76.482315 71.892851 +ShearY 98.739998 98.629997 ... 76.482315 62.471693 +AutoContrast 98.970001 99.000000 ... 76.482315 58.870535 +Invert 98.750000 97.680000 ... 76.482315 62.949374 +Equalize 98.329994 97.570000 ... 76.482315 69.557647 +Solarize 98.070000 97.029999 ... 76.482315 63.104321 +SolarizeAdd 98.239998 97.389999 ... 76.482315 69.777911 +Posterize 99.010002 98.849998 ... 76.482315 73.372108 +Contrast 99.049995 98.970001 ... 76.482315 68.781693 +Color 99.089996 99.019997 ... 76.482315 63.887585 +Brightness 99.000000 98.979996 ... 76.482315 66.205976 +Sharpness 99.019997 98.979996 ... 76.482315 71.537817 +NoiseSalt 99.119995 98.970001 ... 76.482315 57.677266 +NoiseGaussian 99.110001 99.010002 ... 76.482315 58.686552 +w/o do (original x) 99.020000 0.000000 ... 0.000000 72.226558 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.04 66.122465 78.591268 75.662096 80.568012 75.23596 diff --git a/Meta-causal/code-withStyleAttack/66517.error b/Meta-causal/code-withStyleAttack/66517.error new file mode 100644 index 0000000000000000000000000000000000000000..6d4d04b85103027d163f9cff0fc8075b560ec0bc --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66517.error @@ -0,0 +1,4 @@ +run_my_joint_test.sh: line 28: 14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1: command not found +slurmstepd: error: *** JOB 66517 ON gcp-us-0 CANCELLED AT 2024-07-21T15:11:49 DUE TO TIME LIMIT *** +slurmstepd: error: *** STEP 66517.0 ON gcp-us-0 CANCELLED AT 2024-07-21T15:11:49 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-withStyleAttack/66517.log b/Meta-causal/code-withStyleAttack/66517.log new file mode 100644 index 0000000000000000000000000000000000000000..586302bafcb8386e092a11fbbc32f157ed84ef87 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66517.log @@ -0,0 +1,22817 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0256, 0.0311, 0.0202, ..., -0.0092, 0.0157, 0.0004], + [ 0.0206, 0.0280, -0.0197, ..., -0.0287, -0.0107, 0.0182], + [ 0.0262, 0.0067, -0.0170, ..., -0.0002, -0.0249, 0.0204], + ..., + [ 0.0079, 0.0292, 0.0166, ..., 0.0049, 0.0199, 0.0240], + [ 0.0082, -0.0053, -0.0057, ..., -0.0270, -0.0140, 0.0196], + [ 0.0090, 0.0293, -0.0138, ..., -0.0019, -0.0216, 0.0208]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0081, -0.0085, -0.0046, -0.0105, -0.0169, -0.0071, 0.0164, -0.0019, + 0.0246, 0.0014], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 225.02, cls_loss 1.7388 cls_loss_mapping 2.0137 cls_loss_causal 2.2418 re_mapping 0.0700 re_causal 0.0682 /// teacc 79.37 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0355, 0.0352, 0.0261, ..., -0.0077, 0.0137, -0.0090], + [ 0.0269, 0.0208, -0.0298, ..., -0.0371, -0.0168, 0.0250], + [ 0.0299, 0.0008, -0.0227, ..., -0.0027, -0.0262, 0.0209], + ..., + [ 0.0082, 0.0308, 0.0243, ..., 0.0127, 0.0284, 0.0257], + [ 0.0117, -0.0102, -0.0127, ..., -0.0320, -0.0156, 0.0231], + [ 0.0038, 0.0307, -0.0128, ..., -0.0043, -0.0193, 0.0209]], + device='cuda:0'), grad: tensor([[ 2.6875e-03, 1.0078e-02, 5.2795e-03, ..., 1.4557e-02, + 2.4967e-03, 1.6403e-03], + [-1.4076e-02, 2.6509e-05, 1.7376e-03, ..., -5.1956e-03, + -3.9673e-03, -2.7588e-02], + [ 2.3007e-04, -1.8082e-02, 1.5087e-03, ..., 1.2903e-03, + -8.3876e-04, 1.2703e-02], + ..., + [-1.5198e-02, 1.3838e-03, -3.1250e-02, ..., -2.1011e-02, + -3.7079e-02, -1.4610e-02], + [-1.9958e-02, 1.3199e-02, 9.2392e-03, ..., 1.4076e-02, + 7.1068e-03, -1.3199e-02], + [ 1.6708e-02, -4.0627e-03, 5.7678e-03, ..., 1.8570e-02, + 2.8183e-02, 3.6713e-02]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0083, -0.0080, -0.0044, -0.0110, -0.0160, -0.0072, 0.0160, -0.0009, + 0.0239, 0.0007], device='cuda:0'), grad: tensor([ 0.0194, -0.0116, -0.0517, 0.0079, -0.0030, -0.0464, 0.0831, -0.0077, + -0.0031, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 231.02, cls_loss 0.5377 cls_loss_mapping 0.8528 cls_loss_causal 1.9007 re_mapping 0.2121 re_causal 0.2516 /// teacc 91.60 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0381, 0.0368, 0.0271, ..., -0.0083, 0.0109, -0.0109], + [ 0.0276, 0.0206, -0.0329, ..., -0.0403, -0.0184, 0.0273], + [ 0.0315, -0.0022, -0.0229, ..., -0.0030, -0.0253, 0.0185], + ..., + [ 0.0103, 0.0296, 0.0270, ..., 0.0141, 0.0320, 0.0264], + [ 0.0138, -0.0134, -0.0159, ..., -0.0359, -0.0176, 0.0251], + [ 0.0027, 0.0326, -0.0096, ..., -0.0036, -0.0175, 0.0213]], + device='cuda:0'), grad: tensor([[ 1.2026e-03, -1.1848e-02, -1.3016e-02, ..., -4.4212e-03, + 6.9389e-03, 1.2054e-03], + [ 3.7785e-03, 2.2049e-03, 1.7500e-03, ..., 1.8940e-03, + 9.8896e-04, 5.9509e-03], + [-2.5040e-02, -1.4849e-03, 2.0199e-03, ..., -1.7673e-05, + 8.8739e-04, -2.8656e-02], + ..., + [ 2.3890e-04, 1.3123e-02, 1.4961e-02, ..., 1.1703e-02, + 3.7632e-03, 5.4703e-03], + [-1.3817e-02, 7.8011e-03, 4.6577e-03, ..., -4.8370e-03, + 1.2150e-03, -1.4687e-02], + [ 3.4981e-03, -4.7668e-02, -8.7158e-02, ..., -6.1340e-02, + -4.0375e-02, 4.7150e-03]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0084, -0.0081, -0.0047, -0.0105, -0.0162, -0.0060, 0.0152, -0.0014, + 0.0237, 0.0007], device='cuda:0'), grad: tensor([-0.0068, 0.0059, -0.0256, 0.0390, -0.0054, 0.0425, -0.0387, 0.0103, + -0.0045, -0.0167], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 231.25, cls_loss 0.3231 cls_loss_mapping 0.4890 cls_loss_causal 1.6727 re_mapping 0.1563 re_causal 0.2430 /// teacc 93.14 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0399, 0.0377, 0.0280, ..., -0.0088, 0.0094, -0.0126], + [ 0.0285, 0.0211, -0.0349, ..., -0.0418, -0.0185, 0.0287], + [ 0.0327, -0.0044, -0.0236, ..., -0.0032, -0.0251, 0.0173], + ..., + [ 0.0119, 0.0292, 0.0291, ..., 0.0155, 0.0348, 0.0265], + [ 0.0147, -0.0160, -0.0187, ..., -0.0386, -0.0189, 0.0267], + [ 0.0009, 0.0343, -0.0074, ..., -0.0033, -0.0172, 0.0215]], + device='cuda:0'), grad: tensor([[-6.9427e-03, -2.3895e-02, -2.1408e-02, ..., -2.5192e-02, + 6.3658e-04, -2.8877e-03], + [ 3.5095e-03, 3.6955e-04, 3.7460e-03, ..., 3.1166e-03, + 8.3876e-04, 9.4593e-05], + [-2.3926e-02, 4.6272e-03, -1.8311e-03, ..., -1.1002e-02, + 6.0368e-04, -1.2459e-02], + ..., + [-1.6832e-03, -2.8286e-03, -2.7435e-02, ..., -3.6316e-03, + -3.2745e-02, -1.8036e-02], + [ 1.2085e-02, 1.2169e-02, 1.8158e-02, ..., 1.3412e-02, + 4.6997e-03, 1.4366e-02], + [ 8.0414e-03, -1.5160e-02, 3.3545e-04, ..., 1.1925e-02, + 1.6388e-02, -6.1378e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0083, -0.0077, -0.0046, -0.0106, -0.0162, -0.0057, 0.0149, -0.0011, + 0.0233, 0.0007], device='cuda:0'), grad: tensor([-0.0356, 0.0061, -0.0238, 0.0141, 0.0317, -0.0174, 0.0058, -0.0032, + 0.0307, -0.0085], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 230.96, cls_loss 0.2490 cls_loss_mapping 0.3600 cls_loss_causal 1.5010 re_mapping 0.1211 re_causal 0.2153 /// teacc 94.93 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0417, 0.0385, 0.0286, ..., -0.0085, 0.0079, -0.0141], + [ 0.0292, 0.0212, -0.0366, ..., -0.0427, -0.0190, 0.0299], + [ 0.0345, -0.0061, -0.0237, ..., -0.0029, -0.0248, 0.0163], + ..., + [ 0.0129, 0.0290, 0.0308, ..., 0.0162, 0.0375, 0.0260], + [ 0.0153, -0.0186, -0.0224, ..., -0.0411, -0.0205, 0.0281], + [-0.0012, 0.0359, -0.0057, ..., -0.0040, -0.0174, 0.0220]], + device='cuda:0'), grad: tensor([[ 0.0006, -0.0004, -0.0003, ..., -0.0010, 0.0004, 0.0006], + [-0.0272, -0.0137, 0.0006, ..., -0.0071, -0.0086, -0.0281], + [ 0.0099, 0.0067, 0.0028, ..., 0.0053, 0.0053, 0.0085], + ..., + [-0.0113, -0.0112, -0.0245, ..., -0.0173, -0.0207, -0.0051], + [ 0.0141, 0.0128, 0.0042, ..., 0.0067, 0.0031, 0.0190], + [ 0.0095, 0.0090, 0.0152, ..., 0.0122, 0.0122, 0.0054]], + device='cuda:0') +Epoch 5, bias, value: tensor([-0.0082, -0.0076, -0.0043, -0.0104, -0.0162, -0.0062, 0.0144, -0.0010, + 0.0234, 0.0007], device='cuda:0'), grad: tensor([ 0.0015, -0.0276, 0.0137, 0.0022, 0.0031, -0.0129, -0.0070, -0.0186, + 0.0291, 0.0164], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 230.80, cls_loss 0.1967 cls_loss_mapping 0.2618 cls_loss_causal 1.3837 re_mapping 0.0988 re_causal 0.1978 /// teacc 95.61 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0435, 0.0389, 0.0284, ..., -0.0084, 0.0068, -0.0149], + [ 0.0297, 0.0227, -0.0381, ..., -0.0433, -0.0191, 0.0309], + [ 0.0356, -0.0082, -0.0247, ..., -0.0029, -0.0255, 0.0153], + ..., + [ 0.0142, 0.0284, 0.0324, ..., 0.0169, 0.0394, 0.0261], + [ 0.0164, -0.0208, -0.0246, ..., -0.0438, -0.0212, 0.0292], + [-0.0038, 0.0377, -0.0044, ..., -0.0040, -0.0179, 0.0223]], + device='cuda:0'), grad: tensor([[ 4.8447e-03, -1.4887e-03, -1.7977e-03, ..., -2.4700e-03, + 1.3971e-04, 3.6097e-04], + [-1.0719e-03, -1.5701e-02, -1.3733e-02, ..., -8.9111e-03, + -6.4313e-05, -1.3161e-02], + [ 1.1597e-03, 1.5335e-03, 1.5078e-03, ..., 1.4095e-03, + 1.8072e-04, 1.7118e-03], + ..., + [-6.5136e-04, 4.1733e-03, 9.7656e-04, ..., 3.4618e-04, + -3.2387e-03, 3.8948e-03], + [-1.2444e-02, 5.9891e-03, 5.8174e-03, ..., 4.2038e-03, + 2.2519e-04, -2.7447e-03], + [ 5.8842e-04, 1.0742e-02, 6.9809e-03, ..., 3.3832e-04, + 1.7824e-03, 9.4604e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0081, -0.0071, -0.0044, -0.0106, -0.0164, -0.0062, 0.0141, -0.0006, + 0.0236, 0.0004], device='cuda:0'), grad: tensor([ 0.0181, -0.0152, 0.0045, 0.0120, -0.0160, 0.0135, 0.0078, 0.0042, + -0.0415, 0.0128], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 231.00, cls_loss 0.1648 cls_loss_mapping 0.2202 cls_loss_causal 1.3461 re_mapping 0.0845 re_causal 0.1866 /// teacc 96.50 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0448, 0.0392, 0.0286, ..., -0.0078, 0.0058, -0.0157], + [ 0.0295, 0.0236, -0.0394, ..., -0.0445, -0.0196, 0.0320], + [ 0.0365, -0.0102, -0.0257, ..., -0.0033, -0.0253, 0.0148], + ..., + [ 0.0148, 0.0279, 0.0336, ..., 0.0178, 0.0410, 0.0253], + [ 0.0174, -0.0230, -0.0271, ..., -0.0458, -0.0221, 0.0303], + [-0.0062, 0.0395, -0.0034, ..., -0.0043, -0.0187, 0.0222]], + device='cuda:0'), grad: tensor([[ 0.0008, -0.0011, 0.0002, ..., 0.0009, 0.0003, 0.0003], + [-0.0026, 0.0005, 0.0012, ..., 0.0008, -0.0003, -0.0047], + [-0.0034, 0.0010, -0.0062, ..., -0.0117, 0.0001, 0.0027], + ..., + [-0.0081, -0.0045, -0.0128, ..., -0.0055, -0.0092, -0.0030], + [-0.0015, -0.0029, 0.0020, ..., 0.0019, 0.0009, -0.0036], + [ 0.0040, 0.0009, 0.0016, ..., 0.0020, 0.0028, 0.0036]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0079, -0.0070, -0.0044, -0.0104, -0.0159, -0.0069, 0.0138, -0.0006, + 0.0234, 0.0006], device='cuda:0'), grad: tensor([ 0.0025, -0.0015, -0.0120, 0.0113, 0.0019, 0.0014, 0.0009, -0.0065, + -0.0031, 0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 6, time 214.86, cls_loss 0.1555 cls_loss_mapping 0.1990 cls_loss_causal 1.2760 re_mapping 0.0740 re_causal 0.1701 /// teacc 96.28 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0462, 0.0399, 0.0280, ..., -0.0080, 0.0046, -0.0164], + [ 0.0300, 0.0240, -0.0402, ..., -0.0445, -0.0188, 0.0324], + [ 0.0372, -0.0110, -0.0254, ..., -0.0028, -0.0252, 0.0140], + ..., + [ 0.0157, 0.0273, 0.0343, ..., 0.0181, 0.0423, 0.0252], + [ 0.0183, -0.0244, -0.0287, ..., -0.0469, -0.0229, 0.0315], + [-0.0080, 0.0407, -0.0023, ..., -0.0052, -0.0191, 0.0222]], + device='cuda:0'), grad: tensor([[ 0.0011, 0.0012, 0.0027, ..., 0.0022, 0.0015, 0.0002], + [-0.0056, -0.0027, 0.0010, ..., -0.0013, 0.0011, -0.0086], + [-0.0023, 0.0006, -0.0048, ..., -0.0166, 0.0025, 0.0034], + ..., + [-0.0025, -0.0020, -0.0053, ..., 0.0010, -0.0102, 0.0034], + [-0.0135, 0.0025, 0.0015, ..., -0.0071, 0.0009, -0.0045], + [ 0.0027, -0.0085, -0.0126, ..., -0.0011, -0.0026, -0.0025]], + device='cuda:0') +Epoch 8, bias, value: tensor([-0.0077, -0.0069, -0.0042, -0.0106, -0.0161, -0.0070, 0.0135, -0.0008, + 0.0238, 0.0007], device='cuda:0'), grad: tensor([ 0.0035, -0.0025, -0.0062, 0.0287, 0.0078, 0.0044, 0.0069, -0.0029, + -0.0362, -0.0034], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 231.65, cls_loss 0.1254 cls_loss_mapping 0.1577 cls_loss_causal 1.2108 re_mapping 0.0669 re_causal 0.1563 /// teacc 96.58 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0470, 0.0400, 0.0276, ..., -0.0083, 0.0035, -0.0170], + [ 0.0305, 0.0250, -0.0412, ..., -0.0451, -0.0183, 0.0331], + [ 0.0379, -0.0122, -0.0260, ..., -0.0023, -0.0251, 0.0135], + ..., + [ 0.0167, 0.0267, 0.0354, ..., 0.0189, 0.0437, 0.0247], + [ 0.0190, -0.0264, -0.0298, ..., -0.0482, -0.0240, 0.0325], + [-0.0099, 0.0426, -0.0013, ..., -0.0058, -0.0196, 0.0225]], + device='cuda:0'), grad: tensor([[ 5.0783e-04, -3.1548e-03, 1.0719e-03, ..., -1.7424e-03, + 4.3869e-04, 6.3467e-04], + [-1.5078e-03, -7.1144e-04, 4.3678e-04, ..., 1.5050e-05, + -2.7180e-03, -2.8667e-03], + [ 1.8272e-03, 2.2106e-03, 1.0233e-03, ..., 9.5415e-04, + 1.4057e-03, 2.2430e-03], + ..., + [-4.6272e-03, -1.1772e-02, -1.9196e-02, ..., -4.1428e-03, + -8.1482e-03, -3.4571e-04], + [-5.5122e-03, 3.2735e-04, 1.1435e-03, ..., 1.0223e-03, + 4.0555e-04, -4.4250e-03], + [ 4.7150e-03, 4.5509e-03, 1.2329e-02, ..., 3.3875e-03, + 6.4201e-03, -9.6436e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0074, -0.0067, -0.0041, -0.0111, -0.0161, -0.0073, 0.0133, -0.0008, + 0.0241, 0.0009], device='cuda:0'), grad: tensor([-0.0112, -0.0030, 0.0077, 0.0049, 0.0041, 0.0046, 0.0076, -0.0107, + -0.0057, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 231.29, cls_loss 0.1204 cls_loss_mapping 0.1486 cls_loss_causal 1.1595 re_mapping 0.0601 re_causal 0.1435 /// teacc 97.07 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0482, 0.0401, 0.0272, ..., -0.0080, 0.0025, -0.0177], + [ 0.0310, 0.0252, -0.0426, ..., -0.0458, -0.0184, 0.0337], + [ 0.0383, -0.0147, -0.0269, ..., -0.0029, -0.0250, 0.0128], + ..., + [ 0.0177, 0.0272, 0.0366, ..., 0.0199, 0.0451, 0.0249], + [ 0.0199, -0.0275, -0.0313, ..., -0.0497, -0.0242, 0.0336], + [-0.0117, 0.0436, -0.0004, ..., -0.0065, -0.0201, 0.0219]], + device='cuda:0'), grad: tensor([[ 7.2479e-05, 1.6737e-03, 1.6737e-03, ..., 2.4676e-04, + 1.2600e-04, 8.1182e-05], + [-5.8770e-05, 2.7704e-04, 2.9683e-04, ..., 2.1422e-04, + 1.1049e-05, -1.8883e-04], + [ 4.5204e-04, 1.6556e-03, 1.3800e-03, ..., 3.4046e-04, + 6.3229e-04, 2.7323e-04], + ..., + [-1.4143e-03, -8.2111e-04, -4.1695e-03, ..., -1.6098e-03, + -3.2959e-03, 1.6832e-04], + [-2.5392e-04, 1.6994e-03, 8.0061e-04, ..., 4.7112e-04, + 1.5676e-04, -6.0558e-04], + [ 4.5323e-04, -7.3242e-03, -3.7174e-03, ..., 1.0544e-04, + 1.7042e-03, -1.7319e-03]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0071, -0.0065, -0.0046, -0.0109, -0.0157, -0.0079, 0.0130, -0.0005, + 0.0243, 0.0006], device='cuda:0'), grad: tensor([ 0.0020, 0.0005, 0.0015, 0.0006, -0.0028, 0.0010, 0.0036, -0.0012, + 0.0010, -0.0061], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 230.84, cls_loss 0.0973 cls_loss_mapping 0.1216 cls_loss_causal 1.1107 re_mapping 0.0573 re_causal 0.1388 /// teacc 97.32 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0497, 0.0407, 0.0269, ..., -0.0075, 0.0013, -0.0185], + [ 0.0313, 0.0260, -0.0437, ..., -0.0469, -0.0182, 0.0344], + [ 0.0391, -0.0161, -0.0273, ..., -0.0023, -0.0243, 0.0124], + ..., + [ 0.0183, 0.0273, 0.0372, ..., 0.0203, 0.0460, 0.0247], + [ 0.0208, -0.0291, -0.0327, ..., -0.0513, -0.0251, 0.0345], + [-0.0126, 0.0451, 0.0008, ..., -0.0067, -0.0204, 0.0220]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0006, 0.0009, ..., 0.0006, 0.0002, 0.0001], + [ 0.0005, 0.0008, 0.0016, ..., 0.0014, 0.0002, -0.0009], + [-0.0022, 0.0008, 0.0005, ..., -0.0016, -0.0003, 0.0008], + ..., + [ 0.0019, 0.0077, 0.0087, ..., 0.0037, 0.0040, 0.0027], + [-0.0005, 0.0013, 0.0009, ..., 0.0010, 0.0004, -0.0011], + [-0.0018, -0.0242, -0.0300, ..., -0.0141, -0.0067, -0.0032]], + device='cuda:0') +Epoch 11, bias, value: tensor([-0.0072, -0.0066, -0.0044, -0.0112, -0.0158, -0.0083, 0.0130, -0.0002, + 0.0245, 0.0009], device='cuda:0'), grad: tensor([ 1.2264e-03, 1.3876e-03, -2.6345e-04, 1.5327e-02, 1.1435e-03, + -6.1560e-04, 6.2585e-06, 9.3460e-03, 1.1883e-03, -2.8748e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 214.76, cls_loss 0.1025 cls_loss_mapping 0.1267 cls_loss_causal 1.0994 re_mapping 0.0517 re_causal 0.1271 /// teacc 97.25 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0506, 0.0410, 0.0268, ..., -0.0069, 0.0003, -0.0190], + [ 0.0314, 0.0264, -0.0445, ..., -0.0479, -0.0181, 0.0352], + [ 0.0394, -0.0171, -0.0282, ..., -0.0022, -0.0242, 0.0118], + ..., + [ 0.0191, 0.0271, 0.0380, ..., 0.0208, 0.0472, 0.0246], + [ 0.0214, -0.0304, -0.0338, ..., -0.0526, -0.0251, 0.0351], + [-0.0141, 0.0459, 0.0013, ..., -0.0076, -0.0210, 0.0218]], + device='cuda:0'), grad: tensor([[ 0.0001, 0.0003, -0.0008, ..., -0.0008, 0.0004, 0.0002], + [ 0.0002, 0.0030, 0.0004, ..., 0.0003, 0.0003, 0.0042], + [ 0.0002, 0.0016, 0.0005, ..., 0.0006, 0.0003, 0.0004], + ..., + [-0.0008, -0.0014, -0.0041, ..., -0.0011, -0.0033, 0.0001], + [-0.0010, -0.0093, 0.0003, ..., 0.0006, 0.0001, -0.0193], + [ 0.0004, 0.0052, 0.0024, ..., 0.0021, 0.0013, 0.0009]], + device='cuda:0') +Epoch 12, bias, value: tensor([-0.0069, -0.0068, -0.0044, -0.0111, -0.0158, -0.0084, 0.0129, -0.0001, + 0.0246, 0.0007], device='cuda:0'), grad: tensor([ 0.0011, 0.0072, 0.0021, 0.0070, 0.0148, 0.0092, -0.0173, -0.0011, + -0.0288, 0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 214.93, cls_loss 0.0883 cls_loss_mapping 0.1132 cls_loss_causal 1.0790 re_mapping 0.0483 re_causal 0.1220 /// teacc 97.21 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0524, 0.0410, 0.0263, ..., -0.0069, -0.0006, -0.0197], + [ 0.0314, 0.0276, -0.0446, ..., -0.0483, -0.0179, 0.0358], + [ 0.0399, -0.0178, -0.0293, ..., -0.0023, -0.0247, 0.0115], + ..., + [ 0.0200, 0.0264, 0.0385, ..., 0.0213, 0.0485, 0.0246], + [ 0.0223, -0.0316, -0.0339, ..., -0.0531, -0.0255, 0.0359], + [-0.0153, 0.0468, 0.0020, ..., -0.0086, -0.0214, 0.0218]], + device='cuda:0'), grad: tensor([[ 7.5245e-04, -9.6283e-03, 4.6682e-04, ..., -3.3455e-03, + 1.1218e-04, 1.1814e-04], + [-1.4229e-02, -5.9128e-03, -2.3689e-03, ..., -1.8585e-02, + -4.9973e-04, -1.4820e-03], + [ 9.7733e-03, 4.9896e-03, 3.9215e-03, ..., 1.5129e-02, + -7.0190e-04, 1.3924e-03], + ..., + [ 1.3885e-03, 2.5005e-03, 1.3628e-03, ..., 1.0033e-03, + -9.8586e-05, 8.6641e-04], + [-1.4214e-02, -2.7370e-03, -6.0577e-03, ..., -1.2451e-02, + 1.8382e-04, -6.4659e-04], + [ 1.1415e-03, 1.2169e-02, 6.2828e-03, ..., 2.3041e-03, + 8.1396e-04, 4.0970e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0071, -0.0070, -0.0043, -0.0110, -0.0158, -0.0087, 0.0126, -0.0001, + 0.0252, 0.0007], device='cuda:0'), grad: tensor([-0.0119, -0.0182, 0.0153, 0.0171, -0.0175, 0.0057, 0.0061, 0.0044, + -0.0178, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 230.44, cls_loss 0.0744 cls_loss_mapping 0.0954 cls_loss_causal 1.0119 re_mapping 0.0455 re_causal 0.1126 /// teacc 97.58 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0533, 0.0419, 0.0258, ..., -0.0068, -0.0013, -0.0202], + [ 0.0317, 0.0282, -0.0450, ..., -0.0482, -0.0177, 0.0364], + [ 0.0400, -0.0189, -0.0298, ..., -0.0018, -0.0246, 0.0105], + ..., + [ 0.0204, 0.0264, 0.0389, ..., 0.0212, 0.0491, 0.0247], + [ 0.0230, -0.0328, -0.0350, ..., -0.0540, -0.0263, 0.0367], + [-0.0160, 0.0474, 0.0025, ..., -0.0094, -0.0214, 0.0217]], + device='cuda:0'), grad: tensor([[ 7.3135e-05, -7.8630e-04, -1.0347e-04, ..., -1.8823e-04, + 2.2933e-05, 1.6570e-05], + [-8.0407e-05, 1.1033e-04, 3.6120e-05, ..., 1.1736e-04, + -4.0054e-05, -1.3793e-04], + [-2.5606e-04, 4.8375e-04, 9.1016e-05, ..., -3.5316e-05, + -1.4627e-04, 1.4687e-04], + ..., + [ 1.3506e-04, 3.0541e-04, 1.1182e-04, ..., 1.6987e-04, + 4.2200e-05, 1.1498e-04], + [-2.9206e-05, 1.0691e-03, 3.1042e-04, ..., 5.6458e-04, + 6.8843e-05, 7.4267e-05], + [ 2.0444e-05, 2.6836e-03, 1.8024e-04, ..., 2.3401e-04, + -1.8269e-05, 7.7105e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0070, -0.0068, -0.0042, -0.0110, -0.0159, -0.0088, 0.0126, -0.0005, + 0.0252, 0.0008], device='cuda:0'), grad: tensor([ 1.5869e-03, 7.6532e-05, 2.9492e-04, -1.7052e-03, -3.9368e-03, + 1.1196e-03, -2.7504e-03, 4.6372e-04, 1.4658e-03, 3.3798e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 231.16, cls_loss 0.0683 cls_loss_mapping 0.0915 cls_loss_causal 0.9978 re_mapping 0.0413 re_causal 0.1072 /// teacc 97.60 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0540, 0.0426, 0.0253, ..., -0.0064, -0.0019, -0.0206], + [ 0.0318, 0.0288, -0.0462, ..., -0.0492, -0.0174, 0.0369], + [ 0.0405, -0.0204, -0.0305, ..., -0.0019, -0.0250, 0.0104], + ..., + [ 0.0206, 0.0265, 0.0396, ..., 0.0215, 0.0499, 0.0248], + [ 0.0233, -0.0343, -0.0361, ..., -0.0548, -0.0268, 0.0372], + [-0.0172, 0.0483, 0.0031, ..., -0.0102, -0.0217, 0.0217]], + device='cuda:0'), grad: tensor([[ 1.2565e-04, -1.2379e-03, 2.7037e-04, ..., -4.1246e-04, + 5.3525e-05, 2.1979e-05], + [ 4.0889e-04, 2.0313e-03, 8.0442e-04, ..., 5.1832e-04, + 3.1686e-04, 1.5650e-03], + [-2.6417e-03, 1.0185e-03, 5.3501e-04, ..., -1.2131e-03, + -1.4223e-05, -7.8082e-05], + ..., + [-1.0300e-03, 3.9673e-04, -9.6035e-04, ..., -1.4746e-04, + -1.4067e-03, 1.4699e-04], + [ 1.8425e-03, 2.0390e-03, 1.2455e-03, ..., 2.8286e-03, + 3.1614e-04, 5.8746e-04], + [ 4.7565e-04, -1.2352e-02, -4.1509e-04, ..., 5.9967e-03, + 2.6536e-04, -1.3580e-02]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0067, -0.0066, -0.0042, -0.0109, -0.0162, -0.0089, 0.0127, -0.0004, + 0.0251, 0.0008], device='cuda:0'), grad: tensor([-1.2589e-03, 2.4986e-03, -1.6842e-03, -2.6016e-02, 1.1261e-02, + 1.7929e-02, 1.0891e-03, -8.9407e-06, 4.2648e-03, -8.0795e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 230.77, cls_loss 0.0694 cls_loss_mapping 0.0901 cls_loss_causal 0.9835 re_mapping 0.0401 re_causal 0.0973 /// teacc 97.68 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0557, 0.0433, 0.0249, ..., -0.0059, -0.0029, -0.0214], + [ 0.0328, 0.0289, -0.0474, ..., -0.0494, -0.0171, 0.0373], + [ 0.0407, -0.0212, -0.0309, ..., -0.0011, -0.0252, 0.0098], + ..., + [ 0.0208, 0.0268, 0.0406, ..., 0.0221, 0.0510, 0.0243], + [ 0.0243, -0.0349, -0.0370, ..., -0.0561, -0.0271, 0.0381], + [-0.0181, 0.0490, 0.0037, ..., -0.0109, -0.0217, 0.0217]], + device='cuda:0'), grad: tensor([[ 1.6618e-04, -2.1785e-05, 3.9876e-05, ..., 7.1339e-06, + 4.5657e-05, 1.4472e-04], + [ 6.7043e-04, 1.0163e-04, 1.6773e-04, ..., 4.0936e-04, + 2.3460e-04, 2.0373e-04], + [-5.0545e-04, 4.2963e-04, 3.2592e-04, ..., -5.2786e-04, + 1.1396e-04, 3.3998e-04], + ..., + [-7.0333e-04, -7.5245e-04, -1.3685e-03, ..., -4.3869e-04, + -1.2436e-03, 1.2720e-04], + [-7.4863e-04, 1.6184e-03, 2.1064e-04, ..., 4.7326e-04, + 1.5318e-04, -4.7016e-04], + [ 3.2043e-04, 1.3819e-03, 4.9829e-04, ..., 4.3893e-04, + 5.2500e-04, 9.8133e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0066, -0.0066, -0.0042, -0.0112, -0.0160, -0.0092, 0.0129, -0.0004, + 0.0254, 0.0006], device='cuda:0'), grad: tensor([ 0.0003, 0.0010, -0.0002, 0.0006, -0.0020, -0.0059, 0.0025, -0.0006, + 0.0024, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 214.52, cls_loss 0.0726 cls_loss_mapping 0.0841 cls_loss_causal 0.9889 re_mapping 0.0384 re_causal 0.0922 /// teacc 97.36 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0567, 0.0438, 0.0244, ..., -0.0060, -0.0036, -0.0220], + [ 0.0319, 0.0302, -0.0477, ..., -0.0504, -0.0175, 0.0378], + [ 0.0416, -0.0222, -0.0317, ..., -0.0017, -0.0251, 0.0094], + ..., + [ 0.0219, 0.0267, 0.0413, ..., 0.0231, 0.0525, 0.0244], + [ 0.0250, -0.0364, -0.0384, ..., -0.0571, -0.0275, 0.0389], + [-0.0193, 0.0492, 0.0041, ..., -0.0117, -0.0222, 0.0216]], + device='cuda:0'), grad: tensor([[ 6.0081e-04, -4.0114e-05, 4.6706e-04, ..., 2.2781e-04, + 2.5439e-04, 3.6311e-04], + [ 2.0905e-03, -2.8634e-04, 1.0929e-03, ..., 1.1454e-03, + 1.3280e-04, 7.5865e-04], + [-1.2680e-02, 8.1587e-04, 6.7854e-04, ..., -2.0523e-03, + -1.8263e-03, -6.0387e-03], + ..., + [ 1.1154e-02, 4.1771e-03, 1.2299e-02, ..., 1.0269e-02, + 2.7580e-03, 8.2092e-03], + [ 4.1618e-03, 1.1501e-03, 2.4166e-03, ..., 2.4395e-03, + 1.4820e-03, 2.2926e-03], + [ 8.6021e-04, -3.0098e-03, -3.3550e-03, ..., 4.6182e-04, + -1.5199e-04, -8.7833e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-6.4189e-03, -6.6149e-03, -4.1205e-03, -1.0993e-02, -1.5864e-02, + -9.0205e-03, 1.2469e-02, 3.3429e-07, 2.5280e-02, 9.3300e-05], + device='cuda:0'), grad: tensor([ 0.0005, 0.0015, -0.0091, -0.0127, 0.0009, 0.0003, 0.0004, 0.0164, + 0.0046, -0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 214.84, cls_loss 0.0616 cls_loss_mapping 0.0751 cls_loss_causal 0.9772 re_mapping 0.0374 re_causal 0.0935 /// teacc 97.67 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0576, 0.0438, 0.0241, ..., -0.0059, -0.0045, -0.0224], + [ 0.0315, 0.0304, -0.0481, ..., -0.0503, -0.0180, 0.0379], + [ 0.0422, -0.0228, -0.0325, ..., -0.0017, -0.0248, 0.0091], + ..., + [ 0.0223, 0.0262, 0.0416, ..., 0.0229, 0.0533, 0.0243], + [ 0.0257, -0.0375, -0.0396, ..., -0.0579, -0.0281, 0.0398], + [-0.0199, 0.0500, 0.0047, ..., -0.0121, -0.0220, 0.0214]], + device='cuda:0'), grad: tensor([[ 5.5122e-04, 1.3185e-04, 4.3225e-04, ..., 7.4196e-04, + 2.1279e-04, 1.6320e-04], + [ 4.0174e-04, 1.4865e-04, 5.5933e-04, ..., 5.3883e-04, + 1.7798e-04, 1.1009e-04], + [-1.0710e-03, -6.2287e-06, 4.3035e-04, ..., -1.4610e-03, + 2.5845e-04, 1.7989e-04], + ..., + [-2.1827e-04, 8.3160e-03, 1.0284e-02, ..., 2.5177e-03, + 2.5291e-03, -1.4210e-03], + [-3.0923e-04, -2.1374e-04, 3.4475e-04, ..., 7.7200e-04, + 5.5933e-04, -1.8721e-03], + [-1.1892e-03, -7.8430e-03, -1.4580e-02, ..., -5.0621e-03, + -4.8561e-03, 2.3766e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0068, -0.0068, -0.0042, -0.0109, -0.0155, -0.0094, 0.0129, -0.0003, + 0.0253, 0.0003], device='cuda:0'), grad: tensor([ 0.0014, 0.0008, -0.0028, 0.0030, -0.0028, -0.0002, 0.0006, 0.0062, + -0.0006, -0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 214.73, cls_loss 0.0637 cls_loss_mapping 0.0804 cls_loss_causal 0.9249 re_mapping 0.0351 re_causal 0.0871 /// teacc 97.64 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0584, 0.0441, 0.0236, ..., -0.0055, -0.0056, -0.0231], + [ 0.0320, 0.0305, -0.0487, ..., -0.0510, -0.0174, 0.0386], + [ 0.0423, -0.0231, -0.0330, ..., -0.0015, -0.0245, 0.0083], + ..., + [ 0.0226, 0.0261, 0.0423, ..., 0.0230, 0.0541, 0.0243], + [ 0.0262, -0.0385, -0.0407, ..., -0.0593, -0.0288, 0.0410], + [-0.0209, 0.0509, 0.0053, ..., -0.0125, -0.0224, 0.0211]], + device='cuda:0'), grad: tensor([[ 5.8621e-05, -4.5657e-05, 1.1694e-04, ..., 4.2677e-05, + 9.8467e-05, 4.5121e-05], + [-2.3483e-02, -6.8054e-03, -1.3374e-02, ..., -6.3057e-03, + -1.6922e-02, -1.5793e-02], + [-1.1110e-03, 2.7800e-04, 4.5538e-04, ..., 6.1369e-04, + 1.7941e-04, -9.8133e-04], + ..., + [ 2.2278e-02, 6.1302e-03, 1.1856e-02, ..., 5.8594e-03, + 1.5190e-02, 1.5152e-02], + [-5.2243e-05, 2.6441e-04, 3.9816e-04, ..., 7.2670e-04, + 6.2346e-05, -2.3544e-04], + [ 4.2343e-04, 2.2125e-04, 4.9782e-04, ..., 4.9639e-04, + 5.4312e-04, 2.3580e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([-6.8119e-03, -7.0288e-03, -3.9785e-03, -1.0953e-02, -1.5697e-02, + -8.7681e-03, 1.2069e-02, -5.2090e-05, 2.5437e-02, 2.9523e-04], + device='cuda:0'), grad: tensor([ 3.4094e-05, -2.1469e-02, -2.9678e-03, -1.0614e-03, 6.9475e-04, + 1.4849e-03, 3.3379e-04, 2.0935e-02, 1.0843e-03, 9.2363e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 229.04, cls_loss 0.0529 cls_loss_mapping 0.0677 cls_loss_causal 0.9468 re_mapping 0.0340 re_causal 0.0894 /// teacc 98.05 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0595, 0.0443, 0.0231, ..., -0.0056, -0.0066, -0.0236], + [ 0.0318, 0.0311, -0.0496, ..., -0.0520, -0.0176, 0.0392], + [ 0.0430, -0.0233, -0.0335, ..., -0.0009, -0.0238, 0.0079], + ..., + [ 0.0233, 0.0260, 0.0430, ..., 0.0235, 0.0550, 0.0243], + [ 0.0267, -0.0392, -0.0415, ..., -0.0605, -0.0294, 0.0416], + [-0.0223, 0.0512, 0.0055, ..., -0.0134, -0.0226, 0.0210]], + device='cuda:0'), grad: tensor([[ 2.7275e-04, -3.9649e-04, -3.7462e-05, ..., -5.5283e-05, + 4.6194e-05, 1.8466e-04], + [ 1.2046e-04, 1.6308e-04, 1.5426e-04, ..., 2.1279e-04, + -1.3256e-04, 1.1355e-04], + [-1.1170e-04, -1.1605e-04, -1.1516e-04, ..., -6.2275e-04, + -3.0184e-04, 8.7214e-04], + ..., + [ 3.0947e-04, 1.5295e-04, -2.1052e-04, ..., 1.4770e-04, + -3.2216e-05, 3.4142e-04], + [-9.4070e-03, 1.6975e-03, 5.5361e-04, ..., 1.8990e-04, + 6.0469e-05, -1.0239e-02], + [ 3.1528e-03, 2.6665e-03, 8.5771e-05, ..., 1.3828e-04, + 9.7930e-05, 9.3460e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([-7.0737e-03, -7.1977e-03, -3.4984e-03, -1.1235e-02, -1.5931e-02, + -8.6086e-03, 1.2708e-02, 5.7766e-05, 2.5508e-02, -1.2062e-04], + device='cuda:0'), grad: tensor([-0.0003, 0.0005, 0.0014, 0.0006, -0.0129, 0.0043, 0.0090, 0.0008, + -0.0152, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 231.29, cls_loss 0.0562 cls_loss_mapping 0.0771 cls_loss_causal 0.9101 re_mapping 0.0320 re_causal 0.0845 /// teacc 98.20 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0604, 0.0448, 0.0226, ..., -0.0057, -0.0071, -0.0240], + [ 0.0323, 0.0310, -0.0506, ..., -0.0520, -0.0171, 0.0397], + [ 0.0430, -0.0238, -0.0340, ..., -0.0010, -0.0237, 0.0074], + ..., + [ 0.0233, 0.0261, 0.0436, ..., 0.0237, 0.0557, 0.0242], + [ 0.0280, -0.0399, -0.0420, ..., -0.0610, -0.0291, 0.0425], + [-0.0232, 0.0516, 0.0055, ..., -0.0142, -0.0235, 0.0209]], + device='cuda:0'), grad: tensor([[ 1.3626e-04, 1.2457e-04, 3.1203e-05, ..., 2.7746e-05, + 2.6926e-05, 4.7982e-05], + [ 7.6413e-05, 1.8179e-05, 8.6963e-05, ..., 4.3869e-05, + 1.2573e-07, 2.8476e-05], + [-1.0881e-03, 4.1187e-05, -2.3639e-04, ..., -5.9748e-04, + -7.5483e-04, 6.0529e-05], + ..., + [ 5.7077e-04, 8.2076e-05, 2.2316e-04, ..., 3.5930e-04, + 5.4169e-04, 1.7071e-04], + [ 4.1366e-04, 1.0395e-03, 8.0395e-04, ..., 2.1255e-04, + 1.1539e-04, 9.6464e-04], + [ 3.8986e-03, 9.1400e-03, 8.7204e-03, ..., 4.3654e-04, + 9.4223e-04, 1.2726e-02]], device='cuda:0') +Epoch 21, bias, value: tensor([-7.1765e-03, -7.1599e-03, -3.7100e-03, -1.1132e-02, -1.6029e-02, + -8.8079e-03, 1.2442e-02, 1.2523e-04, 2.5943e-02, -4.6145e-05], + device='cuda:0'), grad: tensor([ 6.2180e-04, 2.8467e-04, -1.1215e-03, -2.9355e-05, -1.9073e-02, + 1.1311e-03, -3.1433e-03, 7.9393e-04, 2.4815e-03, 1.8051e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 230.81, cls_loss 0.0493 cls_loss_mapping 0.0664 cls_loss_causal 0.9091 re_mapping 0.0309 re_causal 0.0811 /// teacc 98.22 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0610, 0.0451, 0.0222, ..., -0.0057, -0.0072, -0.0247], + [ 0.0321, 0.0313, -0.0517, ..., -0.0523, -0.0173, 0.0400], + [ 0.0429, -0.0245, -0.0348, ..., -0.0013, -0.0238, 0.0068], + ..., + [ 0.0241, 0.0260, 0.0441, ..., 0.0236, 0.0565, 0.0244], + [ 0.0287, -0.0407, -0.0428, ..., -0.0614, -0.0292, 0.0430], + [-0.0242, 0.0520, 0.0057, ..., -0.0146, -0.0238, 0.0209]], + device='cuda:0'), grad: tensor([[ 6.6161e-05, -3.5739e-04, 1.5116e-04, ..., 9.6858e-05, + 1.2660e-04, 1.3269e-05], + [ 3.0231e-04, 4.7594e-05, 1.6415e-04, ..., 1.4138e-04, + 3.2520e-04, 4.8161e-05], + [-4.8971e-04, 2.4283e-04, -4.0643e-06, ..., -6.6614e-04, + -3.5858e-04, 1.3697e-04], + ..., + [-7.1287e-04, -3.3188e-04, -7.2908e-04, ..., 2.6658e-05, + -1.1215e-03, -2.2137e-04], + [ 6.4373e-05, 2.3580e-04, 1.9383e-04, ..., 1.5819e-04, + 1.5402e-04, -2.5320e-04], + [ 1.1927e-04, 1.1778e-03, 1.2827e-03, ..., 1.0891e-03, + 1.7524e-04, -2.6003e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-6.8432e-03, -7.1374e-03, -4.0128e-03, -1.0986e-02, -1.5977e-02, + -8.7012e-03, 1.2376e-02, 3.2097e-05, 2.5943e-02, -1.4922e-04], + device='cuda:0'), grad: tensor([-0.0006, 0.0003, -0.0004, 0.0016, 0.0007, -0.0070, 0.0004, -0.0002, + 0.0005, 0.0046], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 214.81, cls_loss 0.0481 cls_loss_mapping 0.0610 cls_loss_causal 0.9001 re_mapping 0.0299 re_causal 0.0779 /// teacc 98.17 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0619, 0.0453, 0.0218, ..., -0.0059, -0.0080, -0.0254], + [ 0.0322, 0.0322, -0.0524, ..., -0.0523, -0.0169, 0.0407], + [ 0.0436, -0.0254, -0.0347, ..., -0.0012, -0.0232, 0.0063], + ..., + [ 0.0245, 0.0263, 0.0450, ..., 0.0241, 0.0577, 0.0242], + [ 0.0294, -0.0414, -0.0435, ..., -0.0622, -0.0300, 0.0435], + [-0.0243, 0.0528, 0.0060, ..., -0.0152, -0.0243, 0.0211]], + device='cuda:0'), grad: tensor([[ 3.0923e-04, 2.0742e-04, 5.5408e-04, ..., 2.8658e-04, + 1.9538e-04, 1.0885e-05], + [-2.9698e-05, -3.3665e-04, 4.4346e-05, ..., 1.0705e-04, + -1.7357e-04, -3.3450e-04], + [ 4.1847e-03, 1.3971e-03, 7.7782e-03, ..., 4.7569e-03, + 6.5498e-03, 6.1393e-05], + ..., + [-5.3406e-03, -1.8063e-03, -9.9258e-03, ..., -6.0349e-03, + -7.4959e-03, 2.2089e-04], + [-1.2958e-04, 7.9393e-05, 1.0169e-04, ..., 6.3121e-05, + -7.5817e-05, -2.0885e-04], + [ 2.2733e-04, 2.3854e-04, 5.9336e-05, ..., 2.5010e-04, + 1.9979e-04, 3.2425e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0069, -0.0070, -0.0038, -0.0115, -0.0162, -0.0091, 0.0125, 0.0001, + 0.0261, 0.0003], device='cuda:0'), grad: tensor([ 5.7507e-04, -4.0817e-04, 7.2021e-03, 7.6532e-04, -1.8239e-04, + 4.1866e-04, -1.4460e-04, -8.8196e-03, -5.9634e-05, 6.5804e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 215.05, cls_loss 0.0522 cls_loss_mapping 0.0684 cls_loss_causal 0.8888 re_mapping 0.0288 re_causal 0.0762 /// teacc 98.10 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0627, 0.0454, 0.0212, ..., -0.0062, -0.0088, -0.0260], + [ 0.0326, 0.0327, -0.0535, ..., -0.0515, -0.0169, 0.0413], + [ 0.0434, -0.0267, -0.0356, ..., -0.0016, -0.0238, 0.0059], + ..., + [ 0.0248, 0.0263, 0.0455, ..., 0.0243, 0.0583, 0.0238], + [ 0.0303, -0.0424, -0.0441, ..., -0.0625, -0.0293, 0.0444], + [-0.0250, 0.0538, 0.0065, ..., -0.0157, -0.0245, 0.0209]], + device='cuda:0'), grad: tensor([[ 4.9442e-05, -1.1921e-04, 1.6665e-04, ..., 4.4137e-05, + 7.5459e-05, 1.0774e-05], + [ 4.2105e-04, 1.1176e-04, 1.4591e-04, ..., 3.2115e-04, + 2.8920e-04, 1.5008e-04], + [-7.2432e-04, 1.7536e-04, 1.6749e-04, ..., -5.9652e-04, + -5.0259e-04, -1.2338e-04], + ..., + [-1.3375e-04, 9.4295e-05, -6.2037e-04, ..., -1.5116e-04, + -3.8743e-04, 1.0544e-04], + [-1.2875e-04, 2.3794e-04, 2.1100e-04, ..., 2.0850e-04, + 1.0431e-04, -2.6059e-04], + [ 1.1539e-04, 5.7745e-04, 3.5572e-04, ..., 1.1551e-04, + 5.2691e-04, 3.7998e-05]], device='cuda:0') +Epoch 24, bias, value: tensor([-7.1215e-03, -6.3597e-03, -4.2000e-03, -1.1755e-02, -1.6257e-02, + -8.9121e-03, 1.2129e-02, 5.5141e-05, 2.6057e-02, 7.3294e-04], + device='cuda:0'), grad: tensor([-9.0122e-05, 6.2180e-04, -7.4482e-04, 8.1396e-04, -1.4801e-03, + -3.3021e-04, 7.3195e-05, 1.2644e-05, 9.2268e-05, 1.0300e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 231.37, cls_loss 0.0410 cls_loss_mapping 0.0539 cls_loss_causal 0.8602 re_mapping 0.0289 re_causal 0.0751 /// teacc 98.40 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0635, 0.0456, 0.0209, ..., -0.0061, -0.0095, -0.0263], + [ 0.0323, 0.0335, -0.0534, ..., -0.0520, -0.0169, 0.0416], + [ 0.0441, -0.0278, -0.0364, ..., -0.0016, -0.0242, 0.0057], + ..., + [ 0.0251, 0.0264, 0.0461, ..., 0.0248, 0.0592, 0.0234], + [ 0.0306, -0.0438, -0.0448, ..., -0.0635, -0.0293, 0.0451], + [-0.0253, 0.0540, 0.0065, ..., -0.0165, -0.0248, 0.0207]], + device='cuda:0'), grad: tensor([[ 2.1890e-05, -2.5463e-04, 1.2660e-04, ..., 5.0962e-06, + 1.3530e-05, 3.2838e-06], + [ 1.5364e-03, 6.7663e-04, 2.3985e-04, ..., 9.4271e-04, + 7.8154e-04, 3.7694e-04], + [-1.4496e-03, -5.2595e-04, -1.6701e-04, ..., -9.9945e-04, + -8.6784e-04, 3.0696e-05], + ..., + [ 1.7262e-04, 6.0129e-04, 7.5102e-04, ..., 3.4976e-04, + 1.0836e-04, 4.7743e-05], + [-1.8859e-04, 8.1491e-04, 6.9618e-04, ..., 3.7408e-04, + 1.1504e-04, -7.3576e-04], + [-2.3890e-04, 1.0139e-02, 1.4694e-02, ..., 8.2092e-03, + -2.2948e-04, 8.7202e-05]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0072, -0.0067, -0.0037, -0.0117, -0.0162, -0.0090, 0.0124, 0.0003, + 0.0258, 0.0004], device='cuda:0'), grad: tensor([-0.0004, 0.0021, -0.0017, -0.0133, 0.0013, -0.0030, 0.0002, 0.0008, + 0.0008, 0.0133], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 214.65, cls_loss 0.0394 cls_loss_mapping 0.0513 cls_loss_causal 0.8296 re_mapping 0.0283 re_causal 0.0723 /// teacc 98.19 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0645, 0.0453, 0.0205, ..., -0.0064, -0.0103, -0.0265], + [ 0.0320, 0.0336, -0.0542, ..., -0.0527, -0.0175, 0.0416], + [ 0.0443, -0.0283, -0.0369, ..., -0.0013, -0.0244, 0.0052], + ..., + [ 0.0257, 0.0264, 0.0466, ..., 0.0251, 0.0601, 0.0235], + [ 0.0316, -0.0443, -0.0450, ..., -0.0640, -0.0287, 0.0460], + [-0.0261, 0.0545, 0.0069, ..., -0.0172, -0.0251, 0.0206]], + device='cuda:0'), grad: tensor([[-2.6083e-04, -3.7408e-04, 1.2624e-04, ..., 5.8353e-05, + 1.1152e-04, -1.2302e-04], + [ 7.9803e-03, 4.4098e-03, 5.7411e-03, ..., 3.2310e-03, + 6.7482e-03, 9.7656e-03], + [-1.1140e-04, 3.2806e-04, 4.7374e-04, ..., 3.1209e-04, + 9.0718e-05, 3.3236e-04], + ..., + [-2.4246e-02, -1.8143e-02, -3.1860e-02, ..., -2.0416e-02, + -2.4261e-02, -2.1194e-02], + [ 4.3716e-03, 3.3264e-03, 3.6850e-03, ..., 2.1420e-03, + 2.9812e-03, 4.9934e-03], + [ 1.3056e-03, 8.2970e-04, 1.5440e-03, ..., 1.2951e-03, + 1.1797e-03, 8.7833e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0076, -0.0070, -0.0038, -0.0122, -0.0160, -0.0085, 0.0124, 0.0007, + 0.0261, 0.0003], device='cuda:0'), grad: tensor([-0.0008, 0.0095, 0.0003, 0.0159, 0.0009, -0.0002, -0.0009, -0.0333, + 0.0066, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 214.61, cls_loss 0.0370 cls_loss_mapping 0.0490 cls_loss_causal 0.7954 re_mapping 0.0276 re_causal 0.0724 /// teacc 98.31 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0652, 0.0455, 0.0201, ..., -0.0065, -0.0108, -0.0269], + [ 0.0319, 0.0343, -0.0543, ..., -0.0527, -0.0174, 0.0420], + [ 0.0449, -0.0295, -0.0378, ..., -0.0014, -0.0245, 0.0050], + ..., + [ 0.0261, 0.0266, 0.0472, ..., 0.0256, 0.0609, 0.0236], + [ 0.0315, -0.0452, -0.0460, ..., -0.0651, -0.0293, 0.0463], + [-0.0266, 0.0546, 0.0072, ..., -0.0178, -0.0252, 0.0203]], + device='cuda:0'), grad: tensor([[ 6.1095e-05, -1.2684e-03, 8.1882e-06, ..., -6.0177e-04, + 4.5687e-05, 1.2428e-05], + [ 2.2113e-04, -3.4511e-05, 2.0540e-04, ..., 3.7289e-04, + 1.7321e-04, -1.4591e-04], + [-4.6616e-03, -9.0361e-04, -3.5667e-03, ..., -6.3057e-03, + -4.2610e-03, 1.4150e-04], + ..., + [ 4.0779e-03, 1.4105e-03, 3.1509e-03, ..., 5.8746e-03, + 3.7022e-03, 1.4913e-04], + [-6.7616e-04, 2.0623e-04, -1.2880e-06, ..., -1.7345e-05, + 3.4332e-05, -1.0176e-03], + [ 7.1049e-05, 6.2990e-04, 3.3426e-04, ..., 4.6206e-04, + 4.5687e-05, 8.4043e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0076, -0.0067, -0.0039, -0.0122, -0.0157, -0.0085, 0.0123, 0.0009, + 0.0258, 0.0001], device='cuda:0'), grad: tensor([-1.5726e-03, 2.5845e-04, -5.2605e-03, 6.0654e-04, -3.4547e-04, + 4.4417e-04, 6.1512e-05, 5.6343e-03, -7.5197e-04, 9.2077e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 215.10, cls_loss 0.0322 cls_loss_mapping 0.0477 cls_loss_causal 0.8245 re_mapping 0.0261 re_causal 0.0701 /// teacc 98.33 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0661, 0.0457, 0.0197, ..., -0.0064, -0.0111, -0.0275], + [ 0.0315, 0.0348, -0.0550, ..., -0.0530, -0.0177, 0.0423], + [ 0.0452, -0.0302, -0.0380, ..., -0.0012, -0.0245, 0.0048], + ..., + [ 0.0265, 0.0264, 0.0476, ..., 0.0257, 0.0616, 0.0239], + [ 0.0321, -0.0462, -0.0468, ..., -0.0657, -0.0295, 0.0470], + [-0.0267, 0.0552, 0.0078, ..., -0.0186, -0.0253, 0.0201]], + device='cuda:0'), grad: tensor([[ 1.6257e-05, -4.2588e-05, 3.6955e-05, ..., -2.0284e-06, + 2.2322e-05, 3.6508e-06], + [ 2.9374e-06, -3.2157e-05, 2.7061e-05, ..., 3.2902e-05, + 1.1928e-05, -4.3213e-05], + [-8.9128e-07, 5.0008e-05, 4.8310e-05, ..., 1.4547e-06, + 4.3511e-05, 1.5251e-05], + ..., + [-1.7500e-04, -1.1057e-04, -3.1281e-04, ..., -1.9705e-04, + -3.1233e-04, 2.0698e-05], + [-1.8299e-05, 1.2141e-04, 9.6083e-05, ..., 1.4448e-04, + 2.6613e-05, -3.7789e-05], + [ 7.1943e-05, 1.0633e-04, 7.7188e-05, ..., 2.1970e-04, + 9.6202e-05, 2.6107e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0076, -0.0069, -0.0040, -0.0120, -0.0158, -0.0088, 0.0124, 0.0009, + 0.0257, 0.0004], device='cuda:0'), grad: tensor([ 3.9041e-05, -1.1899e-05, 7.7784e-05, 1.2512e-03, -9.6440e-05, + -1.4696e-03, -2.6643e-05, -2.3746e-04, 1.8024e-04, 2.9516e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 27---------------------------------------------------- +epoch 27, time 231.11, cls_loss 0.0360 cls_loss_mapping 0.0535 cls_loss_causal 0.8226 re_mapping 0.0262 re_causal 0.0724 /// teacc 98.44 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0668, 0.0459, 0.0193, ..., -0.0065, -0.0116, -0.0278], + [ 0.0313, 0.0348, -0.0556, ..., -0.0536, -0.0177, 0.0427], + [ 0.0455, -0.0307, -0.0387, ..., -0.0009, -0.0248, 0.0045], + ..., + [ 0.0268, 0.0262, 0.0479, ..., 0.0257, 0.0622, 0.0237], + [ 0.0324, -0.0469, -0.0472, ..., -0.0668, -0.0299, 0.0476], + [-0.0277, 0.0559, 0.0081, ..., -0.0191, -0.0257, 0.0198]], + device='cuda:0'), grad: tensor([[ 8.4698e-05, -8.8501e-04, -3.2735e-04, ..., -3.0041e-04, + 2.1327e-06, 5.5760e-05], + [-2.8563e-04, -3.3641e-04, -3.9101e-04, ..., 9.1136e-05, + -3.6049e-04, -4.5013e-04], + [-1.6012e-03, 1.2553e-04, 4.9353e-05, ..., -1.1244e-03, + 1.8969e-05, -2.2876e-04], + ..., + [ 5.2023e-04, 3.7003e-04, 3.3092e-04, ..., 1.0610e-04, + 2.3699e-04, 4.7898e-04], + [ 1.9360e-04, -2.0303e-06, 2.2089e-04, ..., 8.0013e-04, + 4.2886e-05, -9.2602e-04], + [ 2.4140e-04, 5.1165e-04, 7.3835e-06, ..., 1.6809e-04, + -3.3796e-05, 1.6022e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0077, -0.0071, -0.0038, -0.0120, -0.0152, -0.0089, 0.0128, 0.0007, + 0.0255, 0.0003], device='cuda:0'), grad: tensor([-8.5020e-04, -2.5630e-04, -2.3289e-03, 7.7200e-04, 7.9870e-05, + -4.7016e-04, 8.8978e-04, 8.6784e-04, 3.9196e-04, 9.0504e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 214.83, cls_loss 0.0353 cls_loss_mapping 0.0468 cls_loss_causal 0.8024 re_mapping 0.0253 re_causal 0.0663 /// teacc 98.34 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0677, 0.0464, 0.0189, ..., -0.0064, -0.0124, -0.0282], + [ 0.0316, 0.0354, -0.0563, ..., -0.0540, -0.0171, 0.0432], + [ 0.0457, -0.0309, -0.0390, ..., -0.0009, -0.0250, 0.0041], + ..., + [ 0.0270, 0.0266, 0.0484, ..., 0.0260, 0.0630, 0.0234], + [ 0.0330, -0.0477, -0.0475, ..., -0.0671, -0.0297, 0.0484], + [-0.0284, 0.0564, 0.0082, ..., -0.0195, -0.0261, 0.0198]], + device='cuda:0'), grad: tensor([[ 5.0902e-05, -1.0490e-04, 5.7459e-05, ..., -1.3486e-05, + 4.7088e-05, 6.9365e-06], + [ 1.3582e-05, -1.9848e-04, 9.1910e-05, ..., 6.2108e-05, + 5.6595e-05, -2.6536e-04], + [-2.7679e-02, -4.2419e-03, -2.6520e-02, ..., -9.3842e-03, + -2.2354e-02, 4.6998e-05], + ..., + [ 2.7039e-02, 4.2877e-03, 2.5955e-02, ..., 9.1400e-03, + 2.1805e-02, 8.1718e-05], + [ 2.3365e-05, 3.0780e-04, 2.7251e-04, ..., 1.1343e-04, + 5.0247e-05, -1.8701e-05], + [ 8.8453e-05, -6.4659e-04, -6.5136e-04, ..., 4.4703e-05, + 1.8775e-05, -7.3135e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0076, -0.0070, -0.0036, -0.0123, -0.0156, -0.0087, 0.0121, 0.0010, + 0.0257, 0.0003], device='cuda:0'), grad: tensor([-3.1978e-05, -1.6284e-04, -2.5864e-02, 2.2182e-03, 1.0481e-03, + -1.8415e-03, -3.3450e-04, 2.5436e-02, 4.0889e-04, -8.7261e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 214.76, cls_loss 0.0360 cls_loss_mapping 0.0462 cls_loss_causal 0.8216 re_mapping 0.0244 re_causal 0.0637 /// teacc 98.39 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0689, 0.0470, 0.0187, ..., -0.0064, -0.0132, -0.0288], + [ 0.0314, 0.0360, -0.0571, ..., -0.0539, -0.0170, 0.0437], + [ 0.0468, -0.0313, -0.0393, ..., -0.0012, -0.0243, 0.0038], + ..., + [ 0.0272, 0.0266, 0.0491, ..., 0.0261, 0.0637, 0.0232], + [ 0.0330, -0.0489, -0.0484, ..., -0.0679, -0.0300, 0.0490], + [-0.0281, 0.0565, 0.0086, ..., -0.0199, -0.0265, 0.0197]], + device='cuda:0'), grad: tensor([[ 6.2525e-05, 2.5034e-05, 7.7248e-05, ..., 5.7817e-06, + 6.6876e-05, 3.4600e-05], + [ 2.3469e-05, -1.2946e-04, 1.6987e-04, ..., 6.0618e-05, + 8.0287e-05, -3.2401e-04], + [ 1.5554e-03, 1.1473e-03, 2.4166e-03, ..., 8.2588e-04, + 2.3098e-03, 1.3483e-04], + ..., + [-3.7365e-03, -2.5158e-03, -5.5923e-03, ..., -1.6718e-03, + -5.1994e-03, 5.3793e-05], + [-1.5986e-04, 1.9503e-04, 2.8968e-04, ..., 1.2720e-04, + 2.1219e-04, -3.8004e-04], + [ 1.8044e-03, 1.8587e-03, 2.8572e-03, ..., 1.0195e-03, + 2.0752e-03, 4.6396e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([-7.5410e-03, -7.1090e-03, -3.3202e-03, -1.2278e-02, -1.5424e-02, + -8.4791e-03, 1.2319e-02, 1.0175e-03, 2.5229e-02, 9.9798e-05], + device='cuda:0'), grad: tensor([ 0.0002, -0.0001, 0.0028, 0.0004, -0.0008, 0.0023, -0.0027, -0.0055, + -0.0003, 0.0038], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 214.77, cls_loss 0.0350 cls_loss_mapping 0.0442 cls_loss_causal 0.7742 re_mapping 0.0233 re_causal 0.0607 /// teacc 98.38 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0698, 0.0471, 0.0180, ..., -0.0063, -0.0140, -0.0294], + [ 0.0312, 0.0365, -0.0574, ..., -0.0544, -0.0170, 0.0439], + [ 0.0470, -0.0316, -0.0396, ..., -0.0010, -0.0241, 0.0039], + ..., + [ 0.0279, 0.0261, 0.0493, ..., 0.0265, 0.0646, 0.0231], + [ 0.0338, -0.0493, -0.0490, ..., -0.0687, -0.0303, 0.0501], + [-0.0289, 0.0567, 0.0087, ..., -0.0203, -0.0270, 0.0192]], + device='cuda:0'), grad: tensor([[ 6.7174e-05, 6.9439e-05, 1.6212e-04, ..., 7.9989e-05, + 6.5923e-05, 2.4542e-05], + [ 4.0591e-05, -3.8475e-05, 2.2411e-04, ..., 1.3733e-04, + 1.0449e-04, -1.2481e-04], + [ 7.4625e-04, 3.6907e-04, 1.5097e-03, ..., 1.1606e-03, + 3.1352e-04, 1.1253e-04], + ..., + [ 2.9640e-03, 4.4346e-04, 5.8556e-03, ..., 5.1460e-03, + 4.7231e-04, 4.2975e-05], + [ 4.9400e-04, 4.3368e-04, 7.8583e-04, ..., 8.7214e-04, + 4.0233e-05, -2.6393e-04], + [ 7.9572e-05, 3.7384e-03, 2.6093e-03, ..., 9.5701e-04, + 9.1374e-05, 3.2067e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0077, -0.0074, -0.0029, -0.0124, -0.0151, -0.0085, 0.0121, 0.0012, + 0.0254, -0.0003], device='cuda:0'), grad: tensor([ 0.0003, 0.0001, 0.0016, -0.0063, -0.0119, 0.0005, 0.0003, 0.0045, + 0.0012, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 214.71, cls_loss 0.0250 cls_loss_mapping 0.0352 cls_loss_causal 0.8064 re_mapping 0.0222 re_causal 0.0629 /// teacc 98.33 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0699, 0.0477, 0.0177, ..., -0.0065, -0.0144, -0.0296], + [ 0.0305, 0.0366, -0.0582, ..., -0.0548, -0.0170, 0.0441], + [ 0.0467, -0.0323, -0.0402, ..., -0.0011, -0.0248, 0.0033], + ..., + [ 0.0283, 0.0263, 0.0498, ..., 0.0267, 0.0655, 0.0230], + [ 0.0348, -0.0502, -0.0500, ..., -0.0694, -0.0304, 0.0508], + [-0.0289, 0.0572, 0.0091, ..., -0.0206, -0.0270, 0.0191]], + device='cuda:0'), grad: tensor([[ 6.2704e-05, -1.7917e-04, 6.0737e-05, ..., -9.1612e-05, + 4.6223e-05, 3.1084e-05], + [ 2.8777e-04, 3.2216e-05, 1.0037e-04, ..., 5.9694e-05, + 8.0585e-05, 2.3520e-04], + [ 3.9139e-03, 1.2655e-03, 4.5052e-03, ..., 3.4695e-03, + 4.5700e-03, 3.4356e-04], + ..., + [-4.1618e-03, -1.5039e-03, -5.4169e-03, ..., -3.7212e-03, + -5.4474e-03, 9.5963e-05], + [-1.5249e-03, -3.2634e-05, 1.4699e-04, ..., 7.2539e-05, + 4.5419e-05, -1.7605e-03], + [ 4.2319e-04, 1.8346e-04, 4.4370e-04, ..., 2.1029e-04, + 4.2057e-04, 1.4031e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0074, -0.0077, -0.0033, -0.0123, -0.0152, -0.0087, 0.0124, 0.0014, + 0.0255, -0.0002], device='cuda:0'), grad: tensor([-0.0001, 0.0005, 0.0054, 0.0008, 0.0003, 0.0002, 0.0002, -0.0057, + -0.0024, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 214.74, cls_loss 0.0275 cls_loss_mapping 0.0347 cls_loss_causal 0.7906 re_mapping 0.0228 re_causal 0.0609 /// teacc 98.28 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0709, 0.0481, 0.0177, ..., -0.0064, -0.0148, -0.0302], + [ 0.0306, 0.0370, -0.0588, ..., -0.0549, -0.0168, 0.0445], + [ 0.0468, -0.0333, -0.0409, ..., -0.0014, -0.0252, 0.0028], + ..., + [ 0.0289, 0.0266, 0.0504, ..., 0.0267, 0.0662, 0.0232], + [ 0.0354, -0.0512, -0.0508, ..., -0.0701, -0.0307, 0.0514], + [-0.0297, 0.0576, 0.0095, ..., -0.0214, -0.0273, 0.0190]], + device='cuda:0'), grad: tensor([[ 1.1760e-04, -2.8815e-06, 4.7266e-05, ..., 9.4697e-06, + 1.6481e-05, 3.3885e-05], + [ 2.0540e-04, -2.6841e-06, 4.9859e-05, ..., 5.8830e-05, + 8.5384e-06, -2.1487e-05], + [ 1.0481e-03, 5.2273e-05, -6.3926e-06, ..., -8.1420e-05, + -1.2085e-05, 3.4189e-04], + ..., + [ 1.2308e-05, 1.4961e-04, -1.1742e-04, ..., -8.9630e-06, + -1.2934e-04, 1.1003e-04], + [-2.4261e-03, 1.1659e-04, 9.6858e-05, ..., 6.0111e-05, + 1.2480e-05, -6.4850e-04], + [ 5.0932e-05, -2.7943e-04, -1.6146e-03, ..., -6.5470e-04, + 4.1068e-05, 2.6631e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0074, -0.0076, -0.0037, -0.0121, -0.0155, -0.0088, 0.0127, 0.0017, + 0.0257, -0.0005], device='cuda:0'), grad: tensor([ 0.0004, 0.0008, 0.0039, 0.0021, -0.0017, 0.0010, 0.0020, 0.0004, + -0.0085, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 214.67, cls_loss 0.0309 cls_loss_mapping 0.0412 cls_loss_causal 0.8196 re_mapping 0.0216 re_causal 0.0615 /// teacc 98.43 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0719, 0.0486, 0.0175, ..., -0.0064, -0.0152, -0.0309], + [ 0.0312, 0.0369, -0.0593, ..., -0.0551, -0.0163, 0.0449], + [ 0.0470, -0.0330, -0.0410, ..., -0.0009, -0.0248, 0.0024], + ..., + [ 0.0287, 0.0263, 0.0505, ..., 0.0263, 0.0665, 0.0230], + [ 0.0361, -0.0520, -0.0518, ..., -0.0708, -0.0309, 0.0523], + [-0.0303, 0.0579, 0.0097, ..., -0.0219, -0.0275, 0.0185]], + device='cuda:0'), grad: tensor([[ 5.1826e-05, 1.6764e-05, 6.0409e-05, ..., 2.8551e-05, + 4.4227e-05, 3.0726e-05], + [-1.5390e-04, -8.7976e-04, -2.2483e-04, ..., -4.4727e-04, + -3.1996e-04, -4.4012e-04], + [ 4.3011e-04, 7.3147e-04, 4.7565e-04, ..., 4.6015e-04, + 4.3297e-04, 4.7827e-04], + ..., + [-2.1095e-03, -4.2248e-04, -3.4962e-03, ..., -2.0218e-03, + -2.5349e-03, 3.0264e-05], + [-7.8201e-04, -5.9843e-04, 1.0914e-04, ..., 8.2433e-05, + 7.7248e-05, -1.2703e-03], + [ 1.5342e-04, -3.3646e-03, -5.2214e-04, ..., 1.0538e-04, + 1.0419e-04, -1.7605e-03]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0073, -0.0081, -0.0029, -0.0120, -0.0154, -0.0085, 0.0123, 0.0011, + 0.0259, -0.0006], device='cuda:0'), grad: tensor([ 1.0675e-04, -1.6880e-03, 1.5955e-03, 3.0899e-03, 5.2834e-03, + 4.7255e-04, -5.4296e-07, -2.3842e-03, -1.5516e-03, -4.9248e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 231.04, cls_loss 0.0304 cls_loss_mapping 0.0418 cls_loss_causal 0.7979 re_mapping 0.0222 re_causal 0.0616 /// teacc 98.56 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0730, 0.0484, 0.0171, ..., -0.0069, -0.0159, -0.0313], + [ 0.0321, 0.0383, -0.0588, ..., -0.0552, -0.0152, 0.0458], + [ 0.0469, -0.0335, -0.0417, ..., -0.0012, -0.0251, 0.0018], + ..., + [ 0.0287, 0.0257, 0.0506, ..., 0.0264, 0.0669, 0.0224], + [ 0.0365, -0.0532, -0.0524, ..., -0.0715, -0.0314, 0.0530], + [-0.0306, 0.0583, 0.0097, ..., -0.0224, -0.0276, 0.0184]], + device='cuda:0'), grad: tensor([[ 6.3002e-05, 4.0680e-05, 5.7071e-05, ..., 4.7743e-05, + 6.7711e-05, 2.5213e-05], + [ 1.9205e-04, -2.1141e-06, 8.7500e-05, ..., 2.0361e-04, + 2.1350e-04, -2.6822e-05], + [-6.2513e-04, 9.3222e-05, 5.3465e-05, ..., -4.2105e-04, + -4.6396e-04, 3.7074e-05], + ..., + [-2.0754e-04, -2.6846e-04, -5.7697e-04, ..., -2.6846e-04, + -4.4823e-04, -6.9216e-06], + [-4.7803e-04, -3.7700e-05, 7.4863e-05, ..., 2.6256e-05, + 1.9944e-04, -1.1587e-03], + [ 2.2185e-04, 2.1666e-05, -1.5870e-05, ..., 5.5820e-05, + 5.4687e-05, 1.9550e-04]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0076, -0.0072, -0.0029, -0.0118, -0.0155, -0.0083, 0.0120, 0.0008, + 0.0256, -0.0006], device='cuda:0'), grad: tensor([ 0.0003, 0.0003, -0.0009, 0.0011, -0.0001, 0.0005, -0.0002, -0.0003, + -0.0011, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 215.06, cls_loss 0.0228 cls_loss_mapping 0.0312 cls_loss_causal 0.7089 re_mapping 0.0209 re_causal 0.0552 /// teacc 98.38 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0733, 0.0489, 0.0169, ..., -0.0071, -0.0165, -0.0317], + [ 0.0318, 0.0382, -0.0595, ..., -0.0556, -0.0155, 0.0460], + [ 0.0470, -0.0343, -0.0423, ..., -0.0013, -0.0255, 0.0012], + ..., + [ 0.0294, 0.0260, 0.0510, ..., 0.0267, 0.0677, 0.0226], + [ 0.0371, -0.0542, -0.0533, ..., -0.0723, -0.0318, 0.0538], + [-0.0312, 0.0586, 0.0101, ..., -0.0229, -0.0271, 0.0182]], + device='cuda:0'), grad: tensor([[ 9.1642e-06, -2.4486e-04, 1.3269e-05, ..., -2.5868e-04, + 6.7167e-06, 1.7229e-06], + [-1.6969e-06, -2.4468e-05, 2.2501e-05, ..., 2.1994e-05, + 8.4341e-06, -6.8605e-05], + [-2.4825e-05, 1.8418e-04, 3.2854e-04, ..., 1.8382e-04, + 2.0158e-04, -4.2021e-05], + ..., + [-4.8232e-04, -1.6838e-05, -4.9639e-04, ..., -3.9983e-04, + -4.9210e-04, 1.5527e-05], + [ 2.4676e-04, 6.3241e-05, 4.5896e-05, ..., 1.3518e-04, + 9.6917e-05, 5.5909e-05], + [-3.2574e-05, -5.4836e-04, -4.9591e-04, ..., 1.0610e-04, + -6.3419e-05, -1.0327e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0071, -0.0075, -0.0032, -0.0118, -0.0153, -0.0084, 0.0120, 0.0009, + 0.0255, -0.0005], device='cuda:0'), grad: tensor([-5.8031e-04, -2.1681e-05, 5.2035e-05, 4.7255e-04, 6.8855e-04, + -8.4758e-05, -5.0098e-05, -3.1781e-04, 4.3988e-04, -5.9700e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 215.13, cls_loss 0.0253 cls_loss_mapping 0.0362 cls_loss_causal 0.7899 re_mapping 0.0213 re_causal 0.0576 /// teacc 98.35 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0744, 0.0490, 0.0165, ..., -0.0070, -0.0171, -0.0323], + [ 0.0321, 0.0389, -0.0592, ..., -0.0561, -0.0152, 0.0466], + [ 0.0471, -0.0355, -0.0426, ..., -0.0014, -0.0256, 0.0007], + ..., + [ 0.0299, 0.0263, 0.0515, ..., 0.0271, 0.0689, 0.0226], + [ 0.0375, -0.0548, -0.0542, ..., -0.0733, -0.0323, 0.0546], + [-0.0321, 0.0589, 0.0100, ..., -0.0234, -0.0279, 0.0177]], + device='cuda:0'), grad: tensor([[ 1.6379e-04, -5.1212e-04, 8.9109e-06, ..., -5.0497e-04, + 2.3082e-05, 7.5996e-05], + [-1.6159e-02, -1.1612e-02, -6.9847e-03, ..., 3.4499e-04, + -1.4000e-02, -7.3586e-03], + [-7.4053e-04, -6.2525e-05, 1.7691e-04, ..., -8.7643e-04, + -5.4932e-04, 2.3234e-04], + ..., + [ 1.4206e-02, 9.9258e-03, 5.6267e-03, ..., 9.1732e-05, + 1.2054e-02, 6.2943e-03], + [-1.6861e-03, -1.0710e-03, 2.9027e-05, ..., 1.6868e-04, + 9.5367e-05, -2.0237e-03], + [ 1.2951e-03, 1.4420e-03, 5.8031e-04, ..., 3.2592e-04, + 1.1873e-03, 6.1178e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0073, -0.0073, -0.0035, -0.0114, -0.0150, -0.0088, 0.0121, 0.0013, + 0.0254, -0.0008], device='cuda:0'), grad: tensor([-0.0021, -0.0184, -0.0019, 0.0010, 0.0014, 0.0021, 0.0027, 0.0165, + -0.0046, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 215.39, cls_loss 0.0241 cls_loss_mapping 0.0367 cls_loss_causal 0.7490 re_mapping 0.0199 re_causal 0.0560 /// teacc 98.36 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0750, 0.0490, 0.0162, ..., -0.0068, -0.0175, -0.0327], + [ 0.0323, 0.0394, -0.0593, ..., -0.0563, -0.0147, 0.0469], + [ 0.0472, -0.0360, -0.0432, ..., -0.0013, -0.0259, 0.0004], + ..., + [ 0.0300, 0.0258, 0.0517, ..., 0.0272, 0.0694, 0.0222], + [ 0.0379, -0.0554, -0.0547, ..., -0.0740, -0.0326, 0.0556], + [-0.0325, 0.0590, 0.0107, ..., -0.0237, -0.0278, 0.0177]], + device='cuda:0'), grad: tensor([[ 5.4955e-05, -3.7909e-05, 6.2346e-05, ..., 2.9787e-05, + 4.7237e-05, 2.5570e-05], + [ 2.5177e-03, 1.6041e-03, 2.5501e-03, ..., 7.4244e-04, + 2.7542e-03, 7.4863e-04], + [ 9.1267e-04, 6.5994e-04, 1.0662e-03, ..., 7.1192e-04, + 7.1096e-04, 2.3913e-04], + ..., + [-3.6392e-03, -2.2011e-03, -3.5324e-03, ..., -4.3249e-04, + -4.5776e-03, -1.0405e-03], + [ 2.0790e-04, 3.7098e-04, 3.6383e-04, ..., 1.4448e-04, + 3.3402e-04, 9.5189e-05], + [ 3.6168e-04, -4.7493e-04, 3.5048e-04, ..., 1.4150e-04, + 3.8052e-04, 2.4242e-03]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0074, -0.0073, -0.0033, -0.0114, -0.0145, -0.0093, 0.0122, 0.0009, + 0.0255, -0.0009], device='cuda:0'), grad: tensor([-1.5527e-05, 3.6716e-03, 1.6241e-03, -9.7656e-04, -2.1248e-03, + 3.0279e-04, 2.2936e-04, -4.6768e-03, 6.5851e-04, 1.3046e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 214.81, cls_loss 0.0247 cls_loss_mapping 0.0314 cls_loss_causal 0.7198 re_mapping 0.0199 re_causal 0.0542 /// teacc 98.49 lr 0.00010000 +Epoch 40, weight, value: tensor([[-7.5558e-02, 4.8734e-02, 1.5599e-02, ..., -7.1549e-03, + -1.8099e-02, -3.3183e-02], + [ 3.1851e-02, 3.9503e-02, -6.0699e-02, ..., -5.7021e-02, + -1.5786e-02, 4.7270e-02], + [ 4.7919e-02, -3.6365e-02, -4.3390e-02, ..., -1.0510e-03, + -2.5506e-02, -5.8012e-05], + ..., + [ 3.0494e-02, 2.6070e-02, 5.2435e-02, ..., 2.7253e-02, + 7.0422e-02, 2.2360e-02], + [ 3.7992e-02, -5.6698e-02, -5.5239e-02, ..., -7.5137e-02, + -3.2685e-02, 5.6172e-02], + [-3.2974e-02, 5.9861e-02, 1.1100e-02, ..., -2.4012e-02, + -2.8156e-02, 1.7345e-02]], device='cuda:0'), grad: tensor([[ 1.0490e-05, -1.3232e-04, 7.2233e-06, ..., -5.8487e-06, + 7.6033e-06, 6.0275e-06], + [ 9.4414e-05, 2.3052e-05, 6.5625e-05, ..., 6.0558e-05, + 7.7546e-05, 1.5363e-05], + [ 2.1398e-05, 2.2963e-05, 1.0818e-04, ..., -2.8923e-05, + 6.1989e-05, 7.6175e-05], + ..., + [-4.9019e-04, -4.7058e-05, -3.9291e-04, ..., -2.2662e-04, + -4.5180e-04, -1.7464e-04], + [ 3.6538e-05, 2.0325e-04, 1.0800e-04, ..., 6.7472e-05, + 9.2566e-05, -4.2230e-05], + [ 1.3918e-05, -4.0698e-04, -1.3936e-04, ..., 3.7044e-05, + 9.9540e-06, -1.0854e-04]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0080, -0.0077, -0.0030, -0.0118, -0.0149, -0.0085, 0.0125, 0.0011, + 0.0250, -0.0004], device='cuda:0'), grad: tensor([-1.5724e-04, 1.3161e-04, -4.2439e-05, 1.3936e-04, 1.5295e-04, + 3.7885e-04, 1.3280e-04, -3.6764e-04, 3.9530e-04, -7.6246e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 230.86, cls_loss 0.0248 cls_loss_mapping 0.0329 cls_loss_causal 0.7405 re_mapping 0.0202 re_causal 0.0538 /// teacc 98.58 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0764, 0.0492, 0.0153, ..., -0.0071, -0.0188, -0.0339], + [ 0.0321, 0.0401, -0.0606, ..., -0.0572, -0.0154, 0.0476], + [ 0.0480, -0.0371, -0.0439, ..., -0.0009, -0.0253, -0.0009], + ..., + [ 0.0313, 0.0259, 0.0529, ..., 0.0275, 0.0709, 0.0221], + [ 0.0383, -0.0574, -0.0562, ..., -0.0763, -0.0333, 0.0571], + [-0.0334, 0.0602, 0.0113, ..., -0.0248, -0.0285, 0.0173]], + device='cuda:0'), grad: tensor([[ 9.9778e-05, 5.3108e-05, 2.1935e-05, ..., 1.3304e-04, + 1.9208e-05, 3.8326e-05], + [ 2.0134e-04, 1.2434e-04, 1.3518e-04, ..., 9.9063e-05, + 1.3590e-04, 1.1492e-04], + [-8.2731e-04, -4.7892e-05, -1.3447e-03, ..., -1.7967e-03, + -9.0408e-04, 4.2176e-04], + ..., + [ 1.7238e-04, -1.5903e-04, 7.2002e-05, ..., 5.0402e-04, + -6.8426e-05, -8.5831e-05], + [-8.4066e-04, 7.6532e-05, 6.1154e-05, ..., 4.3154e-04, + 4.4644e-05, -7.3195e-04], + [ 1.5211e-04, 9.5487e-05, 8.9288e-05, ..., 1.1963e-04, + 7.3314e-05, 7.5698e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0079, -0.0075, -0.0033, -0.0120, -0.0149, -0.0082, 0.0119, 0.0014, + 0.0251, -0.0002], device='cuda:0'), grad: tensor([ 4.3464e-04, 4.3344e-04, -1.5450e-03, 9.8648e-03, 2.4819e-04, + -1.3222e-02, 2.8954e-03, 4.6110e-04, -2.4676e-05, 4.4894e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 214.65, cls_loss 0.0305 cls_loss_mapping 0.0361 cls_loss_causal 0.7446 re_mapping 0.0195 re_causal 0.0523 /// teacc 98.38 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0776, 0.0496, 0.0148, ..., -0.0073, -0.0194, -0.0346], + [ 0.0327, 0.0410, -0.0609, ..., -0.0574, -0.0147, 0.0481], + [ 0.0481, -0.0381, -0.0448, ..., -0.0014, -0.0255, -0.0015], + ..., + [ 0.0317, 0.0262, 0.0535, ..., 0.0280, 0.0720, 0.0225], + [ 0.0387, -0.0582, -0.0568, ..., -0.0773, -0.0341, 0.0579], + [-0.0344, 0.0607, 0.0117, ..., -0.0256, -0.0281, 0.0170]], + device='cuda:0'), grad: tensor([[ 8.4937e-06, -4.3809e-05, 1.6764e-05, ..., 4.9844e-06, + 1.4760e-05, 2.1290e-06], + [ 4.3958e-05, -2.4319e-05, 1.7077e-05, ..., 2.7329e-05, + 3.0905e-05, -1.5274e-05], + [-2.1294e-05, 2.9296e-05, 6.3181e-05, ..., -1.5959e-05, + 2.7195e-05, 1.8582e-05], + ..., + [-1.3912e-04, -6.8903e-05, -2.2995e-04, ..., -1.2720e-04, + -1.8382e-04, 4.9844e-06], + [-5.5730e-05, 2.9385e-05, 1.4126e-05, ..., 2.2411e-05, + -1.3009e-05, -9.1374e-05], + [ 3.4362e-05, 1.3463e-05, 3.3796e-05, ..., 3.4094e-05, + 3.6895e-05, 4.8056e-06]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0080, -0.0073, -0.0036, -0.0123, -0.0154, -0.0074, 0.0115, 0.0020, + 0.0251, -0.0004], device='cuda:0'), grad: tensor([ 3.1233e-05, 2.8715e-05, -1.1124e-05, -8.8394e-05, 1.4640e-05, + -3.9268e-04, 4.9639e-04, -1.3244e-04, -6.5267e-06, 6.0380e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 214.55, cls_loss 0.0251 cls_loss_mapping 0.0297 cls_loss_causal 0.7324 re_mapping 0.0191 re_causal 0.0524 /// teacc 98.43 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0780, 0.0499, 0.0145, ..., -0.0072, -0.0197, -0.0350], + [ 0.0331, 0.0414, -0.0613, ..., -0.0578, -0.0139, 0.0489], + [ 0.0483, -0.0382, -0.0449, ..., -0.0009, -0.0254, -0.0021], + ..., + [ 0.0318, 0.0263, 0.0539, ..., 0.0278, 0.0726, 0.0224], + [ 0.0390, -0.0591, -0.0573, ..., -0.0779, -0.0344, 0.0584], + [-0.0351, 0.0608, 0.0117, ..., -0.0266, -0.0285, 0.0167]], + device='cuda:0'), grad: tensor([[ 4.6402e-05, 1.2726e-05, 2.7597e-05, ..., 5.2184e-05, + 2.1279e-05, 1.8686e-05], + [ 3.6478e-05, 6.3665e-06, 3.7581e-05, ..., 5.4359e-05, + 3.0234e-05, -9.7305e-06], + [ 1.4133e-03, 5.6505e-04, 7.1764e-04, ..., 1.5917e-03, + 6.1989e-04, 7.0333e-04], + ..., + [ 3.3998e-04, 1.2755e-04, 1.3864e-04, ..., 3.9172e-04, + 1.1224e-04, 1.7643e-04], + [ 2.1362e-03, 9.0933e-04, 1.1168e-03, ..., 2.4967e-03, + 9.6512e-04, 9.7609e-04], + [ 3.3587e-05, -6.2168e-05, -5.9843e-05, ..., 4.2260e-05, + 1.7300e-05, 1.4491e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0079, -0.0072, -0.0029, -0.0126, -0.0156, -0.0072, 0.0118, 0.0020, + 0.0247, -0.0009], device='cuda:0'), grad: tensor([ 9.0063e-05, 5.1528e-05, 2.2335e-03, -6.4735e-03, 2.1756e-05, + 1.0878e-04, -6.3539e-05, 5.3453e-04, 3.5114e-03, -8.6352e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.78, cls_loss 0.0230 cls_loss_mapping 0.0354 cls_loss_causal 0.7725 re_mapping 0.0187 re_causal 0.0532 /// teacc 98.47 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0779, 0.0508, 0.0140, ..., -0.0070, -0.0201, -0.0354], + [ 0.0330, 0.0412, -0.0617, ..., -0.0582, -0.0141, 0.0493], + [ 0.0482, -0.0393, -0.0455, ..., -0.0009, -0.0257, -0.0028], + ..., + [ 0.0320, 0.0262, 0.0541, ..., 0.0281, 0.0731, 0.0219], + [ 0.0401, -0.0594, -0.0566, ..., -0.0780, -0.0331, 0.0595], + [-0.0353, 0.0610, 0.0117, ..., -0.0271, -0.0288, 0.0166]], + device='cuda:0'), grad: tensor([[ 2.5845e-04, -6.9857e-04, 2.4974e-05, ..., 1.3149e-04, + 1.3280e-04, 5.4501e-06], + [-2.4295e-04, -7.0620e-04, 9.8228e-05, ..., 6.3479e-05, + -2.3150e-04, -4.5633e-04], + [-3.8099e-04, 1.3816e-04, 1.0812e-04, ..., -4.3035e-04, + -1.0073e-04, 6.3002e-05], + ..., + [-5.2691e-04, -1.0943e-04, -8.2731e-04, ..., -1.6499e-04, + -5.5647e-04, 3.8266e-05], + [ 2.7752e-04, 2.8324e-04, 1.4508e-04, ..., 7.1168e-05, + 2.0766e-04, 9.6083e-05], + [ 2.5749e-04, 3.2353e-04, 3.5405e-04, ..., 2.1088e-04, + 2.3115e-04, 4.2289e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0073, -0.0074, -0.0032, -0.0125, -0.0152, -0.0075, 0.0114, 0.0017, + 0.0250, -0.0006], device='cuda:0'), grad: tensor([-0.0007, -0.0009, -0.0013, 0.0001, 0.0004, 0.0005, 0.0011, -0.0006, + 0.0008, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 214.55, cls_loss 0.0217 cls_loss_mapping 0.0293 cls_loss_causal 0.7475 re_mapping 0.0185 re_causal 0.0512 /// teacc 98.48 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0786, 0.0509, 0.0138, ..., -0.0070, -0.0206, -0.0359], + [ 0.0328, 0.0416, -0.0619, ..., -0.0589, -0.0145, 0.0497], + [ 0.0483, -0.0404, -0.0463, ..., -0.0010, -0.0268, -0.0031], + ..., + [ 0.0327, 0.0266, 0.0547, ..., 0.0287, 0.0747, 0.0218], + [ 0.0404, -0.0603, -0.0570, ..., -0.0787, -0.0334, 0.0602], + [-0.0362, 0.0610, 0.0116, ..., -0.0278, -0.0294, 0.0163]], + device='cuda:0'), grad: tensor([[ 3.1829e-04, 4.2510e-04, 5.9120e-06, ..., 4.8161e-04, + 7.4580e-06, 2.7493e-06], + [ 1.3649e-05, 1.2442e-05, 1.2279e-05, ..., 3.5048e-05, + 2.1070e-05, -1.9804e-05], + [-2.4343e-04, 3.6091e-05, -7.1049e-05, ..., -2.1183e-04, + -2.1911e-04, 1.6779e-05], + ..., + [ 1.2827e-04, 9.1910e-05, 1.2532e-05, ..., 1.5247e-04, + 4.9800e-05, 1.3113e-05], + [ 1.7154e-04, 2.5010e-04, 2.5854e-05, ..., 3.1400e-04, + 3.5256e-05, -2.5019e-05], + [ 3.2395e-05, 9.0361e-05, -2.9966e-05, ..., 4.6700e-05, + -7.5437e-07, 2.9653e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0079, -0.0076, -0.0035, -0.0121, -0.0156, -0.0075, 0.0118, 0.0025, + 0.0249, -0.0009], device='cuda:0'), grad: tensor([ 1.9970e-03, 8.7321e-05, -2.4378e-04, 1.2884e-03, -9.3579e-05, + -4.1428e-03, -8.9931e-04, 4.8137e-04, 1.2684e-03, 2.5797e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 214.67, cls_loss 0.0242 cls_loss_mapping 0.0346 cls_loss_causal 0.7296 re_mapping 0.0174 re_causal 0.0490 /// teacc 98.28 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0793, 0.0512, 0.0133, ..., -0.0071, -0.0215, -0.0364], + [ 0.0323, 0.0413, -0.0627, ..., -0.0593, -0.0153, 0.0499], + [ 0.0483, -0.0404, -0.0465, ..., -0.0010, -0.0266, -0.0038], + ..., + [ 0.0334, 0.0266, 0.0548, ..., 0.0287, 0.0754, 0.0223], + [ 0.0411, -0.0608, -0.0574, ..., -0.0789, -0.0333, 0.0611], + [-0.0362, 0.0615, 0.0125, ..., -0.0283, -0.0290, 0.0162]], + device='cuda:0'), grad: tensor([[ 3.4720e-05, 1.9222e-05, 2.4945e-05, ..., 4.0084e-05, + 2.4185e-05, 1.5244e-05], + [ 8.4281e-05, -2.8539e-04, 6.4611e-05, ..., -1.2141e-04, + -1.8418e-04, -4.9084e-05], + [ 6.1572e-05, 4.9019e-04, 3.4642e-04, ..., 4.7922e-04, + 1.3626e-04, -2.0170e-04], + ..., + [ 2.0809e-03, 1.8606e-03, 2.0218e-03, ..., 3.0270e-03, + 1.1311e-03, 1.6797e-04], + [ 1.0021e-05, 1.7309e-04, 1.5986e-04, ..., 2.1136e-04, + 9.4712e-05, -1.2016e-04], + [ 4.6343e-05, -1.6367e-04, -2.0659e-04, ..., 3.7372e-05, + -8.5533e-05, 4.1872e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0078, -0.0080, -0.0035, -0.0125, -0.0157, -0.0071, 0.0117, 0.0028, + 0.0250, -0.0008], device='cuda:0'), grad: tensor([ 7.5817e-05, -6.5565e-04, 5.6314e-04, -3.8166e-03, 5.6088e-05, + 2.1076e-04, 4.3303e-05, 3.5591e-03, 1.2922e-04, -1.6224e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 214.80, cls_loss 0.0215 cls_loss_mapping 0.0312 cls_loss_causal 0.7238 re_mapping 0.0181 re_causal 0.0491 /// teacc 98.48 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0802, 0.0516, 0.0132, ..., -0.0071, -0.0222, -0.0372], + [ 0.0328, 0.0422, -0.0629, ..., -0.0592, -0.0151, 0.0509], + [ 0.0485, -0.0409, -0.0469, ..., -0.0009, -0.0268, -0.0040], + ..., + [ 0.0340, 0.0266, 0.0553, ..., 0.0289, 0.0764, 0.0221], + [ 0.0410, -0.0617, -0.0581, ..., -0.0796, -0.0338, 0.0616], + [-0.0373, 0.0620, 0.0131, ..., -0.0289, -0.0294, 0.0159]], + device='cuda:0'), grad: tensor([[ 4.8280e-05, 2.5466e-05, 3.0115e-05, ..., 2.5675e-05, + 2.6032e-05, 6.8955e-06], + [ 6.1131e-04, 2.9516e-04, 2.0361e-04, ..., 1.3888e-04, + 4.4560e-04, 1.2410e-04], + [-3.1567e-03, 3.8177e-05, 2.8536e-05, ..., -2.3142e-05, + -2.1515e-03, -7.4673e-04], + ..., + [ 1.4114e-04, -1.5056e-04, -2.7275e-04, ..., 5.8353e-05, + -7.2122e-05, 5.8919e-05], + [ 2.1877e-03, 1.0765e-04, 5.7310e-05, ..., 1.0645e-04, + 1.4563e-03, 4.9543e-04], + [ 2.1827e-04, 8.6737e-04, 1.2660e-04, ..., 1.4567e-04, + 1.3077e-04, 1.8263e-04]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0077, -0.0073, -0.0034, -0.0128, -0.0155, -0.0066, 0.0110, 0.0028, + 0.0242, -0.0008], device='cuda:0'), grad: tensor([ 0.0003, 0.0010, -0.0026, 0.0044, -0.0014, -0.0034, -0.0026, 0.0002, + 0.0024, 0.0018], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 214.83, cls_loss 0.0205 cls_loss_mapping 0.0262 cls_loss_causal 0.6983 re_mapping 0.0171 re_causal 0.0473 /// teacc 98.55 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0801, 0.0523, 0.0129, ..., -0.0068, -0.0221, -0.0376], + [ 0.0330, 0.0423, -0.0631, ..., -0.0589, -0.0147, 0.0517], + [ 0.0485, -0.0414, -0.0470, ..., -0.0009, -0.0272, -0.0047], + ..., + [ 0.0341, 0.0268, 0.0556, ..., 0.0285, 0.0772, 0.0218], + [ 0.0418, -0.0624, -0.0584, ..., -0.0800, -0.0337, 0.0623], + [-0.0380, 0.0622, 0.0131, ..., -0.0293, -0.0298, 0.0157]], + device='cuda:0'), grad: tensor([[ 2.2054e-05, -2.1681e-05, 3.4541e-05, ..., 2.8908e-05, + 2.1026e-05, 1.4275e-05], + [-1.0805e-03, -1.8368e-03, 3.9548e-05, ..., 2.9492e-04, + 1.8215e-04, -1.9484e-03], + [-1.8132e-04, 9.0182e-05, 9.4831e-05, ..., -1.5008e-04, + -1.3983e-04, 4.9412e-05], + ..., + [-4.6611e-04, -1.3046e-03, -2.4986e-03, ..., -3.2592e-04, + -1.5802e-03, 5.8830e-05], + [ 9.5987e-04, 1.6556e-03, 1.5306e-04, ..., 5.9545e-05, + 8.5890e-05, 1.4820e-03], + [ 5.1308e-04, 1.6270e-03, 2.2430e-03, ..., 6.1941e-04, + 1.4133e-03, 1.5652e-04]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0072, -0.0070, -0.0034, -0.0126, -0.0153, -0.0075, 0.0111, 0.0027, + 0.0245, -0.0009], device='cuda:0'), grad: tensor([ 2.0996e-05, -3.0918e-03, -2.3282e-04, -8.3160e-04, 6.3702e-06, + 3.7408e-04, 7.6473e-05, -2.2411e-03, 2.9831e-03, 2.9354e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 47---------------------------------------------------- +epoch 47, time 231.10, cls_loss 0.0233 cls_loss_mapping 0.0282 cls_loss_causal 0.7449 re_mapping 0.0168 re_causal 0.0472 /// teacc 98.60 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0807, 0.0530, 0.0129, ..., -0.0062, -0.0224, -0.0382], + [ 0.0330, 0.0429, -0.0629, ..., -0.0597, -0.0145, 0.0521], + [ 0.0493, -0.0424, -0.0476, ..., -0.0008, -0.0275, -0.0048], + ..., + [ 0.0341, 0.0269, 0.0560, ..., 0.0291, 0.0779, 0.0213], + [ 0.0418, -0.0633, -0.0590, ..., -0.0811, -0.0343, 0.0631], + [-0.0381, 0.0621, 0.0131, ..., -0.0297, -0.0302, 0.0151]], + device='cuda:0'), grad: tensor([[ 2.5127e-06, -3.2723e-05, 7.4804e-06, ..., -1.1906e-05, + 3.1870e-06, 9.3952e-06], + [ 3.4660e-05, 1.1368e-03, 1.9228e-04, ..., 4.2111e-05, + 3.5554e-05, 3.3736e-04], + [ 6.0022e-05, 7.6115e-05, 5.1767e-05, ..., 3.1203e-05, + 5.1886e-05, 2.6092e-05], + ..., + [-1.2445e-04, 4.1395e-05, -1.0788e-04, ..., -4.7952e-05, + -1.2946e-04, 4.5775e-07], + [-3.8324e-07, 4.9496e-04, 1.6391e-05, ..., 2.2888e-04, + 2.9318e-06, 1.5810e-05], + [ 1.2375e-05, -2.1133e-03, -3.3116e-04, ..., 3.5018e-05, + 2.0787e-05, -6.7806e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0070, -0.0068, -0.0033, -0.0126, -0.0153, -0.0076, 0.0114, 0.0026, + 0.0242, -0.0012], device='cuda:0'), grad: tensor([-9.8646e-05, 1.4915e-03, 1.6713e-04, -5.3673e-03, 1.0939e-03, + 3.8357e-03, 3.3522e-04, 1.8133e-06, 1.1120e-03, -2.5654e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 214.72, cls_loss 0.0182 cls_loss_mapping 0.0259 cls_loss_causal 0.6848 re_mapping 0.0174 re_causal 0.0469 /// teacc 98.60 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0813, 0.0534, 0.0126, ..., -0.0062, -0.0228, -0.0388], + [ 0.0330, 0.0431, -0.0634, ..., -0.0602, -0.0148, 0.0527], + [ 0.0496, -0.0426, -0.0474, ..., -0.0009, -0.0272, -0.0056], + ..., + [ 0.0341, 0.0269, 0.0560, ..., 0.0295, 0.0785, 0.0212], + [ 0.0427, -0.0639, -0.0596, ..., -0.0814, -0.0345, 0.0637], + [-0.0385, 0.0624, 0.0135, ..., -0.0304, -0.0304, 0.0151]], + device='cuda:0'), grad: tensor([[ 1.2085e-05, 2.5600e-05, 3.6091e-05, ..., 2.7508e-05, + 6.6087e-06, 2.7686e-05], + [ 2.8715e-05, 3.8922e-05, 2.3350e-05, ..., 3.1799e-05, + 2.3857e-05, 4.7743e-05], + [-4.3631e-05, -2.0891e-05, 2.4587e-05, ..., -1.3255e-05, + -3.9637e-05, -5.8502e-05], + ..., + [ 3.5834e-04, 1.2712e-03, 1.7319e-03, ..., 6.4373e-04, + 8.3590e-04, 5.7369e-05], + [-1.6201e-04, -5.2750e-05, 4.2945e-05, ..., -8.2791e-05, + 1.1340e-05, -6.3181e-04], + [-3.6693e-04, -9.4032e-04, -1.5163e-03, ..., -3.7408e-04, + -8.7023e-04, 8.4698e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0069, -0.0068, -0.0034, -0.0124, -0.0153, -0.0079, 0.0116, 0.0026, + 0.0245, -0.0013], device='cuda:0'), grad: tensor([ 1.0800e-04, 1.2326e-04, 1.7917e-04, -1.4997e-04, -6.8307e-05, + 1.5426e-04, -8.5533e-05, 1.8368e-03, -8.1730e-04, -1.2789e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 49---------------------------------------------------- +epoch 49, time 230.92, cls_loss 0.0142 cls_loss_mapping 0.0208 cls_loss_causal 0.6810 re_mapping 0.0177 re_causal 0.0478 /// teacc 98.68 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0820, 0.0531, 0.0123, ..., -0.0062, -0.0232, -0.0394], + [ 0.0329, 0.0432, -0.0639, ..., -0.0601, -0.0150, 0.0529], + [ 0.0493, -0.0429, -0.0480, ..., -0.0012, -0.0277, -0.0062], + ..., + [ 0.0349, 0.0269, 0.0564, ..., 0.0298, 0.0792, 0.0214], + [ 0.0431, -0.0646, -0.0601, ..., -0.0819, -0.0346, 0.0644], + [-0.0387, 0.0633, 0.0140, ..., -0.0305, -0.0300, 0.0150]], + device='cuda:0'), grad: tensor([[ 5.3197e-06, -2.3283e-06, 9.4175e-06, ..., 2.6152e-06, + 1.1027e-05, 9.1568e-06], + [ 8.7395e-06, 1.2852e-05, 2.4498e-05, ..., 1.2144e-05, + 2.9802e-05, -1.7388e-06], + [-5.1081e-05, 2.3261e-05, -4.5411e-06, ..., -5.7399e-05, + 2.2780e-06, -2.2333e-06], + ..., + [-2.5183e-05, 4.8780e-04, 2.1565e-04, ..., -1.0759e-05, + 4.5180e-04, 3.7193e-04], + [ 1.8150e-05, 6.6638e-05, 5.1796e-05, ..., 6.4611e-05, + 4.8876e-05, 2.2128e-05], + [ 2.4185e-05, 3.0479e-03, 1.6193e-03, ..., 1.8388e-05, + 2.9831e-03, 2.2697e-03]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0074, -0.0069, -0.0037, -0.0128, -0.0150, -0.0079, 0.0109, 0.0028, + 0.0247, -0.0007], device='cuda:0'), grad: tensor([ 1.5855e-05, 4.8816e-05, -7.3433e-05, -3.3307e-04, -7.4081e-03, + 3.6192e-04, -2.1458e-05, 9.9277e-04, 1.9515e-04, 6.2256e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 215.02, cls_loss 0.0143 cls_loss_mapping 0.0205 cls_loss_causal 0.6864 re_mapping 0.0160 re_causal 0.0453 /// teacc 98.57 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0827, 0.0535, 0.0120, ..., -0.0065, -0.0239, -0.0399], + [ 0.0330, 0.0435, -0.0644, ..., -0.0603, -0.0150, 0.0532], + [ 0.0495, -0.0434, -0.0484, ..., -0.0012, -0.0278, -0.0066], + ..., + [ 0.0353, 0.0271, 0.0569, ..., 0.0301, 0.0799, 0.0217], + [ 0.0432, -0.0656, -0.0606, ..., -0.0827, -0.0349, 0.0647], + [-0.0392, 0.0633, 0.0141, ..., -0.0310, -0.0304, 0.0146]], + device='cuda:0'), grad: tensor([[ 5.0592e-04, 8.2636e-04, 5.5075e-04, ..., 9.2030e-04, + 3.9525e-06, 2.3365e-05], + [ 3.7581e-05, 3.9995e-05, 3.8058e-05, ..., 7.8559e-05, + 1.6898e-05, -4.4793e-05], + [ 3.2615e-06, 1.9979e-04, 1.9622e-04, ..., 4.4155e-04, + -1.7539e-05, 1.0684e-05], + ..., + [ 6.5446e-05, 1.4937e-04, -2.4632e-05, ..., 9.9361e-05, + -6.9976e-05, 6.1810e-05], + [ 3.1424e-04, 9.3651e-04, 4.3893e-04, ..., 6.7854e-04, + 1.5795e-05, 2.0337e-04], + [-1.6141e-04, -1.5917e-03, 7.9989e-05, ..., 3.3951e-04, + 2.0638e-05, -1.5812e-03]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0074, -0.0069, -0.0039, -0.0127, -0.0148, -0.0073, 0.0108, 0.0032, + 0.0241, -0.0010], device='cuda:0'), grad: tensor([ 0.0030, 0.0003, 0.0006, -0.0046, 0.0038, -0.0033, 0.0007, 0.0008, + 0.0030, -0.0043], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 215.02, cls_loss 0.0176 cls_loss_mapping 0.0229 cls_loss_causal 0.6949 re_mapping 0.0161 re_causal 0.0464 /// teacc 98.56 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0835, 0.0539, 0.0117, ..., -0.0067, -0.0243, -0.0404], + [ 0.0331, 0.0440, -0.0642, ..., -0.0607, -0.0153, 0.0538], + [ 0.0496, -0.0439, -0.0491, ..., -0.0010, -0.0283, -0.0070], + ..., + [ 0.0361, 0.0271, 0.0576, ..., 0.0308, 0.0811, 0.0218], + [ 0.0434, -0.0662, -0.0613, ..., -0.0838, -0.0353, 0.0653], + [-0.0398, 0.0632, 0.0139, ..., -0.0317, -0.0308, 0.0143]], + device='cuda:0'), grad: tensor([[-1.1735e-05, 3.4256e-03, 4.7731e-04, ..., 3.8803e-05, + -6.3896e-05, -9.6932e-06], + [ 2.0221e-05, 5.4449e-05, 2.8029e-05, ..., 2.7269e-05, + 2.7463e-05, -1.8075e-05], + [ 1.1644e-03, 1.2465e-03, 1.8425e-03, ..., 2.2507e-03, + -2.6673e-06, 2.8670e-05], + ..., + [ 5.5656e-06, 7.7248e-05, -4.4644e-05, ..., 3.1829e-05, + -3.1255e-06, -2.0123e-04], + [ 2.7561e-04, 2.9039e-04, 4.5657e-04, ..., 5.7745e-04, + 7.0520e-06, -1.7762e-05], + [ 3.1978e-05, -3.8052e-03, -4.3869e-04, ..., 4.9323e-05, + 1.2778e-05, 1.4253e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0074, -0.0067, -0.0040, -0.0132, -0.0146, -0.0072, 0.0112, 0.0038, + 0.0238, -0.0015], device='cuda:0'), grad: tensor([ 0.0038, 0.0002, 0.0046, -0.0052, 0.0058, 0.0004, -0.0018, -0.0054, + 0.0011, -0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 214.82, cls_loss 0.0161 cls_loss_mapping 0.0250 cls_loss_causal 0.6895 re_mapping 0.0163 re_causal 0.0460 /// teacc 98.55 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0840, 0.0536, 0.0114, ..., -0.0066, -0.0245, -0.0411], + [ 0.0329, 0.0444, -0.0647, ..., -0.0608, -0.0157, 0.0538], + [ 0.0496, -0.0444, -0.0496, ..., -0.0010, -0.0282, -0.0074], + ..., + [ 0.0361, 0.0275, 0.0578, ..., 0.0306, 0.0814, 0.0209], + [ 0.0446, -0.0671, -0.0610, ..., -0.0842, -0.0338, 0.0669], + [-0.0407, 0.0634, 0.0139, ..., -0.0321, -0.0316, 0.0139]], + device='cuda:0'), grad: tensor([[ 1.2159e-04, 2.4724e-04, 1.0289e-05, ..., 2.0608e-05, + 9.5367e-06, 1.6582e-04], + [-8.2159e-04, -1.1520e-03, 1.3721e-04, ..., 5.8711e-05, + 1.3852e-04, -1.1740e-03], + [ 6.2764e-05, 5.1439e-05, 2.5943e-05, ..., 1.5572e-05, + 2.3007e-05, 1.0133e-04], + ..., + [-1.5962e-04, -3.1561e-05, -2.1887e-04, ..., -8.4639e-05, + -1.9765e-04, -4.0144e-05], + [ 3.6478e-04, 8.3065e-04, 1.5616e-05, ..., -1.3840e-04, + 9.2089e-06, 4.4203e-04], + [ 4.7147e-05, 1.2243e-04, -1.5050e-05, ..., 2.8476e-05, + 2.0579e-05, 1.6499e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0078, -0.0067, -0.0041, -0.0126, -0.0146, -0.0075, 0.0115, 0.0038, + 0.0240, -0.0017], device='cuda:0'), grad: tensor([ 8.6355e-04, -2.9869e-03, 2.0957e-04, 3.2115e-04, -4.3654e-04, + 4.9210e-04, -5.4687e-05, -1.9327e-05, 1.2550e-03, 3.5501e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 215.21, cls_loss 0.0133 cls_loss_mapping 0.0192 cls_loss_causal 0.6994 re_mapping 0.0158 re_causal 0.0451 /// teacc 98.61 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0845, 0.0539, 0.0110, ..., -0.0069, -0.0249, -0.0416], + [ 0.0330, 0.0444, -0.0650, ..., -0.0608, -0.0155, 0.0540], + [ 0.0498, -0.0454, -0.0501, ..., -0.0011, -0.0286, -0.0075], + ..., + [ 0.0364, 0.0274, 0.0583, ..., 0.0308, 0.0819, 0.0208], + [ 0.0449, -0.0677, -0.0616, ..., -0.0847, -0.0339, 0.0674], + [-0.0412, 0.0636, 0.0145, ..., -0.0325, -0.0318, 0.0136]], + device='cuda:0'), grad: tensor([[ 6.7614e-06, -1.0043e-04, 7.3612e-06, ..., -2.2873e-05, + 4.9919e-06, 4.1574e-06], + [-5.8636e-06, -4.6611e-05, 2.0340e-05, ..., 3.2365e-05, + 2.0206e-05, -9.7990e-05], + [-1.2646e-03, -1.8239e-04, -4.9686e-04, ..., -1.0614e-03, + -8.4686e-04, 3.0398e-05], + ..., + [ 4.7803e-04, -3.1066e-04, -8.7309e-04, ..., 1.0080e-03, + 1.6820e-04, -1.9634e-04], + [-2.7984e-05, 7.7188e-05, 1.3375e-04, ..., 2.1309e-05, + 5.9783e-05, -8.3447e-05], + [ 6.4659e-04, 4.4513e-04, 1.1139e-03, ..., 2.5094e-05, + 5.4073e-04, 1.9932e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0074, -0.0068, -0.0042, -0.0128, -0.0145, -0.0074, 0.0114, 0.0038, + 0.0238, -0.0016], device='cuda:0'), grad: tensor([-1.0079e-04, -1.2827e-04, -1.6117e-03, 9.5904e-05, -1.1615e-05, + 7.0214e-05, 1.4174e-04, 2.5654e-04, 4.3839e-05, 1.2445e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 214.93, cls_loss 0.0145 cls_loss_mapping 0.0236 cls_loss_causal 0.6786 re_mapping 0.0157 re_causal 0.0445 /// teacc 98.56 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0852, 0.0540, 0.0103, ..., -0.0070, -0.0255, -0.0423], + [ 0.0334, 0.0450, -0.0649, ..., -0.0611, -0.0149, 0.0549], + [ 0.0500, -0.0456, -0.0503, ..., -0.0006, -0.0283, -0.0081], + ..., + [ 0.0362, 0.0265, 0.0581, ..., 0.0303, 0.0818, 0.0201], + [ 0.0454, -0.0686, -0.0620, ..., -0.0852, -0.0341, 0.0680], + [-0.0417, 0.0645, 0.0150, ..., -0.0329, -0.0315, 0.0136]], + device='cuda:0'), grad: tensor([[ 6.0536e-06, -4.3735e-06, 1.4983e-05, ..., -1.5013e-05, + 5.6215e-06, 6.6049e-06], + [ 1.7926e-05, -1.6494e-06, 3.0965e-05, ..., 1.2048e-05, + 2.3365e-05, -1.2510e-05], + [ 2.8253e-05, 3.3766e-05, 2.5213e-05, ..., 1.1995e-05, + 1.5467e-05, 4.7833e-05], + ..., + [-9.1493e-05, -9.4712e-05, -1.9979e-04, ..., -1.7285e-05, + -1.8990e-04, 3.4682e-06], + [-8.1539e-05, 2.3022e-05, 6.9857e-05, ..., 7.5161e-05, + 4.6700e-05, -1.8251e-04], + [ 1.1837e-04, -7.5459e-05, 6.5088e-05, ..., 6.4731e-05, + 1.4448e-04, 7.3493e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0077, -0.0065, -0.0040, -0.0125, -0.0139, -0.0077, 0.0107, 0.0030, + 0.0241, -0.0012], device='cuda:0'), grad: tensor([-3.7998e-05, 1.6451e-05, 8.8215e-05, -1.1206e-04, 1.1635e-04, + 2.2084e-05, 3.1203e-05, -1.2231e-04, -5.2214e-05, 5.0306e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 55---------------------------------------------------- +epoch 55, time 225.16, cls_loss 0.0174 cls_loss_mapping 0.0217 cls_loss_causal 0.6880 re_mapping 0.0148 re_causal 0.0411 /// teacc 98.69 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0858, 0.0543, 0.0095, ..., -0.0070, -0.0261, -0.0433], + [ 0.0331, 0.0454, -0.0654, ..., -0.0617, -0.0150, 0.0556], + [ 0.0502, -0.0458, -0.0508, ..., -0.0005, -0.0285, -0.0083], + ..., + [ 0.0368, 0.0266, 0.0587, ..., 0.0308, 0.0827, 0.0200], + [ 0.0458, -0.0694, -0.0625, ..., -0.0859, -0.0343, 0.0688], + [-0.0421, 0.0649, 0.0154, ..., -0.0333, -0.0318, 0.0135]], + device='cuda:0'), grad: tensor([[ 6.9812e-06, 1.1392e-05, 1.0960e-05, ..., 3.9563e-06, + 6.8396e-06, 1.2949e-05], + [-4.7588e-04, -4.1604e-04, 1.1820e-04, ..., 4.8950e-06, + 7.8678e-05, -9.3794e-04], + [ 3.3200e-05, 5.0128e-05, 3.1471e-05, ..., -1.6004e-05, + 1.5333e-05, 6.7890e-05], + ..., + [ 1.6940e-04, 4.4179e-04, 4.4584e-04, ..., 4.2375e-07, + 2.9135e-04, 4.1628e-04], + [ 3.8648e-04, 4.3344e-04, 4.8459e-05, ..., 1.5795e-05, + 2.0862e-05, 7.6246e-04], + [-2.1935e-04, -6.6662e-04, -8.6355e-04, ..., 1.9535e-05, + -5.7602e-04, -5.5647e-04]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0074, -0.0064, -0.0037, -0.0128, -0.0141, -0.0080, 0.0108, 0.0031, + 0.0237, -0.0010], device='cuda:0'), grad: tensor([ 6.8569e-04, -1.2522e-03, 8.6427e-05, 1.3277e-05, 3.5715e-04, + -1.2732e-04, -6.8092e-04, 7.1049e-04, 1.1387e-03, -9.3031e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 214.72, cls_loss 0.0144 cls_loss_mapping 0.0217 cls_loss_causal 0.6888 re_mapping 0.0152 re_causal 0.0425 /// teacc 98.63 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0867, 0.0547, 0.0092, ..., -0.0068, -0.0266, -0.0443], + [ 0.0330, 0.0457, -0.0658, ..., -0.0620, -0.0152, 0.0562], + [ 0.0502, -0.0465, -0.0515, ..., -0.0008, -0.0287, -0.0089], + ..., + [ 0.0371, 0.0267, 0.0589, ..., 0.0304, 0.0835, 0.0198], + [ 0.0461, -0.0702, -0.0632, ..., -0.0869, -0.0346, 0.0696], + [-0.0428, 0.0648, 0.0155, ..., -0.0339, -0.0321, 0.0126]], + device='cuda:0'), grad: tensor([[ 5.8681e-05, -7.4990e-06, 1.2226e-05, ..., 9.6262e-06, + 5.5023e-06, 1.0710e-06], + [ 4.9412e-05, -3.8415e-05, 1.5557e-05, ..., 1.1243e-05, + 6.6385e-06, -8.5473e-05], + [ 3.5038e-03, 5.9992e-05, 4.3780e-05, ..., 1.4361e-06, + 1.2808e-05, 6.9201e-05], + ..., + [-5.2631e-05, -2.6554e-05, -1.3769e-04, ..., -3.7640e-05, + -1.0413e-04, 2.1029e-06], + [-4.3259e-03, 3.3498e-05, 3.5495e-05, ..., 3.2604e-05, + 4.3809e-06, 4.6492e-06], + [ 2.3067e-05, -7.8976e-05, -4.3750e-05, ..., 1.9222e-05, + 2.1398e-05, -1.8179e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0075, -0.0063, -0.0042, -0.0120, -0.0141, -0.0076, 0.0116, 0.0027, + 0.0235, -0.0016], device='cuda:0'), grad: tensor([ 1.9288e-04, 1.2312e-06, 1.1772e-02, 1.9157e-04, 1.5736e-04, + 4.7016e-04, 1.6079e-03, -7.3195e-05, -1.4244e-02, -7.2539e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 57---------------------------------------------------- +epoch 57, time 224.63, cls_loss 0.0151 cls_loss_mapping 0.0209 cls_loss_causal 0.6313 re_mapping 0.0150 re_causal 0.0407 /// teacc 98.73 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0872, 0.0549, 0.0088, ..., -0.0067, -0.0270, -0.0450], + [ 0.0326, 0.0454, -0.0661, ..., -0.0624, -0.0160, 0.0565], + [ 0.0505, -0.0464, -0.0520, ..., -0.0003, -0.0283, -0.0093], + ..., + [ 0.0373, 0.0267, 0.0592, ..., 0.0310, 0.0840, 0.0193], + [ 0.0467, -0.0707, -0.0633, ..., -0.0877, -0.0345, 0.0701], + [-0.0424, 0.0648, 0.0158, ..., -0.0345, -0.0320, 0.0126]], + device='cuda:0'), grad: tensor([[ 8.6874e-06, 2.3320e-05, 2.5019e-05, ..., 8.4937e-06, + 1.3910e-05, 3.0678e-06], + [ 2.2933e-05, 1.0467e-04, 7.7665e-05, ..., 2.2486e-05, + 4.1783e-05, -1.1802e-05], + [ 6.7353e-05, 6.8069e-05, 1.5175e-04, ..., -5.9813e-05, + 1.0592e-04, 1.1154e-05], + ..., + [-2.7323e-04, -1.3089e-04, -7.8964e-04, ..., -2.2554e-04, + -4.9353e-04, 9.6858e-06], + [-6.0499e-06, 1.2720e-04, 1.3971e-04, ..., 1.3478e-05, + 1.8090e-05, -9.5144e-06], + [ 4.0203e-05, -1.2207e-04, -4.2021e-05, ..., 3.7849e-05, + 9.8884e-05, -5.4449e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0071, -0.0068, -0.0039, -0.0120, -0.0139, -0.0077, 0.0110, 0.0026, + 0.0240, -0.0017], device='cuda:0'), grad: tensor([ 1.6117e-04, 3.9220e-04, -3.4761e-04, 4.6468e-04, 4.8161e-04, + 2.1422e-04, -6.6376e-04, -8.6498e-04, 2.4152e-04, -7.7665e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 214.76, cls_loss 0.0121 cls_loss_mapping 0.0176 cls_loss_causal 0.6447 re_mapping 0.0155 re_causal 0.0415 /// teacc 98.70 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0879, 0.0552, 0.0085, ..., -0.0066, -0.0274, -0.0456], + [ 0.0323, 0.0455, -0.0665, ..., -0.0626, -0.0160, 0.0566], + [ 0.0506, -0.0468, -0.0528, ..., -0.0003, -0.0285, -0.0096], + ..., + [ 0.0372, 0.0265, 0.0596, ..., 0.0311, 0.0846, 0.0188], + [ 0.0477, -0.0709, -0.0632, ..., -0.0881, -0.0345, 0.0711], + [-0.0428, 0.0646, 0.0159, ..., -0.0352, -0.0323, 0.0123]], + device='cuda:0'), grad: tensor([[ 2.5984e-07, -1.8373e-05, 1.7928e-06, ..., -3.0920e-06, + 1.8720e-07, 6.7521e-07], + [-2.5973e-05, -2.6271e-05, 9.5740e-06, ..., 1.3553e-05, + -4.0904e-06, -5.0426e-05], + [ 5.3644e-05, 4.6194e-05, 9.6560e-05, ..., 1.4615e-04, + 1.6866e-06, 1.5780e-05], + ..., + [ 1.3500e-05, 1.5497e-05, 4.3586e-06, ..., 1.4015e-05, + -2.6748e-06, 1.8597e-05], + [ 1.0645e-06, 1.6496e-05, 1.2629e-05, ..., 1.7643e-05, + 5.4576e-07, -1.4491e-06], + [ 4.5560e-06, 5.5656e-06, -7.8380e-06, ..., 5.9269e-06, + 2.3823e-06, 9.0003e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0072, -0.0070, -0.0040, -0.0122, -0.0132, -0.0074, 0.0109, 0.0024, + 0.0245, -0.0022], device='cuda:0'), grad: tensor([-2.7090e-05, -7.4327e-05, 1.3757e-04, -1.4317e-04, 5.0515e-06, + -6.3658e-05, 5.1647e-05, 4.2975e-05, 4.6194e-05, 2.4796e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 59---------------------------------------------------- +epoch 59, time 230.72, cls_loss 0.0110 cls_loss_mapping 0.0184 cls_loss_causal 0.6622 re_mapping 0.0146 re_causal 0.0414 /// teacc 98.81 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0885, 0.0550, 0.0079, ..., -0.0068, -0.0277, -0.0461], + [ 0.0324, 0.0456, -0.0669, ..., -0.0630, -0.0160, 0.0570], + [ 0.0508, -0.0471, -0.0533, ..., -0.0002, -0.0287, -0.0100], + ..., + [ 0.0377, 0.0261, 0.0600, ..., 0.0310, 0.0851, 0.0185], + [ 0.0479, -0.0715, -0.0637, ..., -0.0885, -0.0347, 0.0716], + [-0.0434, 0.0651, 0.0157, ..., -0.0357, -0.0326, 0.0121]], + device='cuda:0'), grad: tensor([[ 2.7314e-05, 9.9912e-06, 1.5363e-05, ..., 2.6524e-05, + 1.9129e-06, 1.6978e-06], + [-7.2527e-04, -2.0707e-04, 4.4107e-06, ..., 2.5108e-05, + -2.7609e-04, -3.9172e-04], + [ 3.7014e-05, 9.7573e-05, 5.4576e-06, ..., -4.2057e-04, + 2.2733e-04, 2.6393e-04], + ..., + [ 6.0499e-05, 2.2441e-05, -9.7007e-06, ..., 3.0786e-05, + -3.1535e-06, 2.4706e-05], + [ 3.0351e-04, 2.1374e-04, 1.3137e-04, ..., 2.3842e-04, + 2.7180e-05, 3.8505e-05], + [ 6.6757e-05, -7.8738e-05, -9.3102e-05, ..., 1.2577e-04, + 5.9940e-06, 1.9208e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0075, -0.0071, -0.0041, -0.0120, -0.0131, -0.0066, 0.0105, 0.0023, + 0.0243, -0.0023], device='cuda:0'), grad: tensor([ 9.1910e-05, -1.0405e-03, -3.8004e-04, 1.9932e-03, 5.9366e-05, + -2.1114e-03, 1.2529e-04, 1.5521e-04, 1.0223e-03, 8.7261e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 214.73, cls_loss 0.0159 cls_loss_mapping 0.0203 cls_loss_causal 0.6741 re_mapping 0.0141 re_causal 0.0397 /// teacc 98.74 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0893, 0.0549, 0.0075, ..., -0.0072, -0.0282, -0.0467], + [ 0.0321, 0.0461, -0.0673, ..., -0.0637, -0.0163, 0.0574], + [ 0.0512, -0.0472, -0.0537, ..., 0.0004, -0.0291, -0.0100], + ..., + [ 0.0383, 0.0260, 0.0603, ..., 0.0312, 0.0861, 0.0184], + [ 0.0482, -0.0723, -0.0642, ..., -0.0897, -0.0348, 0.0722], + [-0.0438, 0.0648, 0.0157, ..., -0.0365, -0.0330, 0.0118]], + device='cuda:0'), grad: tensor([[ 1.5807e-04, 5.9557e-04, 2.4843e-04, ..., 5.1641e-04, + 3.1257e-04, 2.5675e-05], + [ 8.0967e-04, 9.8526e-05, 1.4901e-04, ..., 5.6934e-04, + 1.0500e-03, -6.9380e-05], + [-8.0681e-04, 9.9897e-05, 8.8751e-05, ..., -4.8923e-04, + -8.7595e-04, 8.1003e-05], + ..., + [-1.3673e-04, -4.2439e-04, -6.8521e-04, ..., -2.9087e-04, + -7.6914e-04, 1.9699e-05], + [-9.8467e-05, 4.9442e-05, 5.1826e-05, ..., 5.7846e-05, + 4.8995e-05, -1.5175e-04], + [ 2.8819e-05, -4.7016e-04, -7.2289e-04, ..., 5.0277e-05, + -3.4904e-04, 1.3404e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0080, -0.0075, -0.0033, -0.0119, -0.0127, -0.0062, 0.0105, 0.0027, + 0.0239, -0.0030], device='cuda:0'), grad: tensor([ 1.1158e-03, 2.0332e-03, -1.7595e-03, 8.9216e-04, 8.9884e-04, + -1.4858e-03, -3.5256e-05, -7.6437e-04, -1.0419e-04, -7.9298e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 214.72, cls_loss 0.0144 cls_loss_mapping 0.0211 cls_loss_causal 0.6536 re_mapping 0.0148 re_causal 0.0424 /// teacc 98.67 lr 0.00010000 +Epoch 63, weight, value: tensor([[-9.0131e-02, 5.5090e-02, 6.9804e-03, ..., -7.2474e-03, + -2.9012e-02, -4.8310e-02], + [ 3.1593e-02, 4.6702e-02, -6.8191e-02, ..., -6.4467e-02, + -1.7217e-02, 5.7917e-02], + [ 5.0951e-02, -4.7940e-02, -5.5087e-02, ..., -2.7335e-05, + -3.0230e-02, -1.0337e-02], + ..., + [ 3.8590e-02, 2.6356e-02, 6.0410e-02, ..., 3.1009e-02, + 8.7287e-02, 1.8017e-02], + [ 4.7961e-02, -7.3184e-02, -6.5033e-02, ..., -9.0978e-02, + -3.5257e-02, 7.2934e-02], + [-4.3146e-02, 6.5725e-02, 1.6971e-02, ..., -3.6201e-02, + -3.2671e-02, 1.3038e-02]], device='cuda:0'), grad: tensor([[ 2.4870e-05, -2.6599e-05, 1.4096e-05, ..., 1.0443e-04, + 1.1340e-05, -5.0198e-07], + [ 1.2353e-05, 2.3320e-05, 3.4958e-05, ..., 4.5359e-05, + 1.6689e-05, -1.3737e-06], + [-3.5334e-04, 5.3257e-05, -3.6895e-05, ..., -5.0354e-04, + -2.6274e-04, 1.0327e-05], + ..., + [ 1.6832e-04, 2.2084e-05, 8.2850e-05, ..., 2.0289e-04, + 1.6689e-04, 7.1488e-06], + [ 3.3855e-05, 4.4048e-05, 4.2647e-05, ..., 6.0171e-05, + 8.4713e-06, 1.3597e-05], + [ 1.3262e-05, 7.0333e-04, -2.0757e-05, ..., 1.1660e-05, + 2.0847e-05, 7.4434e-04]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0085, -0.0074, -0.0036, -0.0118, -0.0133, -0.0059, 0.0104, 0.0024, + 0.0233, -0.0017], device='cuda:0'), grad: tensor([ 1.4734e-04, 8.0764e-05, -9.2649e-04, 2.5702e-04, -1.7815e-03, + 2.6107e-05, 3.0965e-05, 3.5286e-04, 1.6940e-04, 1.6422e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.97, cls_loss 0.0128 cls_loss_mapping 0.0191 cls_loss_causal 0.6625 re_mapping 0.0145 re_causal 0.0408 /// teacc 98.68 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0912, 0.0551, 0.0066, ..., -0.0071, -0.0297, -0.0492], + [ 0.0314, 0.0470, -0.0684, ..., -0.0650, -0.0174, 0.0584], + [ 0.0521, -0.0485, -0.0546, ..., 0.0004, -0.0294, -0.0111], + ..., + [ 0.0384, 0.0268, 0.0608, ..., 0.0314, 0.0875, 0.0177], + [ 0.0481, -0.0745, -0.0653, ..., -0.0917, -0.0349, 0.0735], + [-0.0430, 0.0662, 0.0172, ..., -0.0370, -0.0326, 0.0133]], + device='cuda:0'), grad: tensor([[ 2.2292e-05, -1.4916e-05, 3.9265e-06, ..., 1.7695e-06, + 2.7493e-06, 3.4064e-05], + [ 8.7404e-04, -4.9174e-05, 6.4895e-06, ..., 4.5970e-06, + 2.9188e-06, 1.5402e-03], + [ 2.4706e-05, 2.4274e-05, 1.1846e-05, ..., -7.1637e-06, + -2.6785e-06, 9.0957e-05], + ..., + [-2.0146e-05, 5.0031e-06, -4.6164e-05, ..., -1.9565e-05, + -3.1590e-05, 4.6670e-05], + [-1.0118e-03, 1.3866e-05, 1.6928e-05, ..., 6.2585e-06, + 4.7125e-06, -1.8845e-03], + [ 1.2577e-05, -5.6475e-06, -3.1944e-07, ..., 1.3590e-05, + 5.7966e-06, 5.8338e-06]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0085, -0.0073, -0.0031, -0.0120, -0.0138, -0.0059, 0.0108, 0.0025, + 0.0227, -0.0014], device='cuda:0'), grad: tensor([ 3.5346e-05, 1.7958e-03, 1.1325e-04, 3.0905e-05, 8.4758e-05, + 9.7811e-05, 4.3184e-05, 1.1034e-05, -2.2221e-03, 9.1717e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 215.05, cls_loss 0.0120 cls_loss_mapping 0.0181 cls_loss_causal 0.6824 re_mapping 0.0143 re_causal 0.0408 /// teacc 98.69 lr 0.00010000 +Epoch 65, weight, value: tensor([[-9.1802e-02, 5.5720e-02, 6.3943e-03, ..., -7.2533e-03, + -3.0065e-02, -4.9806e-02], + [ 3.1856e-02, 4.7580e-02, -6.7240e-02, ..., -6.4421e-02, + -1.6967e-02, 5.9152e-02], + [ 5.1922e-02, -4.8785e-02, -5.5649e-02, ..., 6.0204e-05, + -3.0386e-02, -1.1354e-02], + ..., + [ 3.9207e-02, 2.6878e-02, 6.1477e-02, ..., 3.1801e-02, + 8.8768e-02, 1.7651e-02], + [ 4.8599e-02, -7.4676e-02, -6.5740e-02, ..., -9.2264e-02, + -3.5223e-02, 7.4609e-02], + [-4.4124e-02, 6.5959e-02, 1.6824e-02, ..., -3.7561e-02, + -3.3331e-02, 1.2376e-02]], device='cuda:0'), grad: tensor([[ 2.4557e-05, -1.3094e-06, 1.6481e-05, ..., 1.3359e-05, + 2.5824e-05, 8.9481e-06], + [ 4.0501e-05, 2.6807e-05, 2.1839e-04, ..., 1.6093e-04, + 2.0933e-04, -1.3900e-04], + [-3.8290e-04, 4.4167e-05, -1.8871e-04, ..., -2.4307e-04, + -4.8018e-04, 2.3708e-05], + ..., + [-2.1183e-04, -7.7200e-04, -8.9359e-04, ..., -5.5408e-04, + -6.1178e-04, 1.4663e-05], + [ 2.1279e-04, 2.2602e-04, 2.0528e-04, ..., 1.9884e-04, + 1.1170e-04, 2.1803e-04], + [ 3.2711e-04, 4.3344e-04, 5.9414e-04, ..., 4.5490e-04, + 5.8985e-04, 8.0764e-06]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0084, -0.0067, -0.0030, -0.0120, -0.0140, -0.0066, 0.0108, 0.0028, + 0.0232, -0.0021], device='cuda:0'), grad: tensor([ 9.2506e-05, 2.3997e-04, -7.3338e-04, -1.6046e-04, -3.3355e-04, + 1.7262e-04, 2.4605e-04, -1.2217e-03, 5.3549e-04, 1.1597e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 215.02, cls_loss 0.0138 cls_loss_mapping 0.0204 cls_loss_causal 0.6667 re_mapping 0.0140 re_causal 0.0384 /// teacc 98.74 lr 0.00010000 +Epoch 66, weight, value: tensor([[-9.2671e-02, 5.6513e-02, 6.0798e-03, ..., -6.9074e-03, + -3.1055e-02, -5.0347e-02], + [ 3.1625e-02, 4.7524e-02, -6.7843e-02, ..., -6.5017e-02, + -1.7943e-02, 5.9382e-02], + [ 5.1893e-02, -4.9488e-02, -5.5866e-02, ..., -2.9052e-05, + -3.0425e-02, -1.2488e-02], + ..., + [ 3.8996e-02, 2.6471e-02, 6.0450e-02, ..., 3.1353e-02, + 8.8967e-02, 1.8121e-02], + [ 4.8972e-02, -7.5253e-02, -6.6376e-02, ..., -9.2968e-02, + -3.5481e-02, 7.5510e-02], + [-4.3719e-02, 6.6098e-02, 1.8009e-02, ..., -3.7854e-02, + -3.2249e-02, 1.1444e-02]], device='cuda:0'), grad: tensor([[-2.1942e-06, -3.5584e-05, 2.7083e-06, ..., -7.1749e-06, + 3.5968e-06, 3.1162e-06], + [ 4.3102e-06, -5.4359e-05, 4.4674e-05, ..., 1.1288e-05, + 2.6315e-05, -1.3769e-04], + [ 1.4439e-05, 1.4715e-05, 1.8314e-05, ..., 6.1840e-06, + 1.5393e-05, 1.0341e-05], + ..., + [-1.0681e-03, -2.5415e-04, -1.6079e-03, ..., -3.9673e-04, + -1.2474e-03, 3.2261e-06], + [ 3.1084e-05, 5.8919e-05, 2.8700e-05, ..., 1.0252e-05, + 2.0742e-05, 8.2672e-05], + [ 9.4461e-04, 2.2328e-04, 1.3971e-03, ..., 3.5357e-04, + 1.0967e-03, 2.0370e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0077, -0.0068, -0.0034, -0.0119, -0.0133, -0.0062, 0.0108, 0.0018, + 0.0230, -0.0020], device='cuda:0'), grad: tensor([-3.3855e-05, -1.2422e-04, 4.9472e-05, 1.1826e-04, 3.6925e-05, + 2.0236e-05, -1.0383e-04, -1.5221e-03, 1.8656e-04, 1.3733e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 214.75, cls_loss 0.0121 cls_loss_mapping 0.0171 cls_loss_causal 0.6589 re_mapping 0.0136 re_causal 0.0390 /// teacc 98.64 lr 0.00010000 +Epoch 67, weight, value: tensor([[-9.4084e-02, 5.6880e-02, 5.7173e-03, ..., -6.5814e-03, + -3.1736e-02, -5.2937e-02], + [ 3.1793e-02, 4.7365e-02, -6.8117e-02, ..., -6.5509e-02, + -1.7528e-02, 5.9585e-02], + [ 5.2018e-02, -4.9698e-02, -5.6268e-02, ..., 2.7254e-05, + -3.0388e-02, -1.2801e-02], + ..., + [ 3.9262e-02, 2.5856e-02, 6.0914e-02, ..., 3.1889e-02, + 8.9248e-02, 1.7148e-02], + [ 4.9730e-02, -7.5219e-02, -6.6653e-02, ..., -9.3494e-02, + -3.5677e-02, 7.7292e-02], + [-4.4559e-02, 6.6297e-02, 1.7968e-02, ..., -3.8633e-02, + -3.2572e-02, 1.0745e-02]], device='cuda:0'), grad: tensor([[ 1.0110e-05, 5.8413e-05, 6.1154e-05, ..., 6.3837e-05, + 1.1817e-05, 7.7114e-06], + [ 3.6240e-05, 6.1393e-05, 1.0538e-04, ..., 8.6427e-05, + 4.7892e-05, -1.3351e-05], + [ 2.5138e-05, 7.3910e-05, 8.3506e-05, ..., 8.3268e-05, + 2.9832e-05, 1.3001e-05], + ..., + [-1.5104e-04, 2.5034e-04, -7.2300e-05, ..., 1.2743e-04, + -2.2149e-04, 1.2837e-05], + [-3.8981e-05, 7.5459e-05, 2.8944e-04, ..., 3.9721e-04, + 8.1658e-06, -1.9267e-05], + [ 4.0442e-05, -8.3804e-05, -3.7849e-05, ..., 8.3327e-05, + 6.9618e-05, 5.8301e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0075, -0.0073, -0.0033, -0.0121, -0.0128, -0.0060, 0.0109, 0.0016, + 0.0238, -0.0024], device='cuda:0'), grad: tensor([ 0.0014, 0.0010, 0.0012, -0.0016, 0.0020, -0.0133, 0.0007, 0.0061, + 0.0013, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.81, cls_loss 0.0109 cls_loss_mapping 0.0158 cls_loss_causal 0.6325 re_mapping 0.0141 re_causal 0.0367 /// teacc 98.68 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0948, 0.0571, 0.0054, ..., -0.0066, -0.0322, -0.0537], + [ 0.0315, 0.0475, -0.0687, ..., -0.0662, -0.0179, 0.0595], + [ 0.0517, -0.0507, -0.0574, ..., -0.0002, -0.0313, -0.0135], + ..., + [ 0.0400, 0.0262, 0.0618, ..., 0.0328, 0.0902, 0.0166], + [ 0.0500, -0.0762, -0.0673, ..., -0.0946, -0.0359, 0.0782], + [-0.0453, 0.0667, 0.0180, ..., -0.0396, -0.0330, 0.0107]], + device='cuda:0'), grad: tensor([[ 5.0999e-06, 1.0878e-05, 7.9423e-06, ..., 5.9083e-06, + 5.4576e-06, 1.9476e-05], + [-1.2159e-04, -2.5606e-04, 2.3693e-05, ..., 2.5257e-05, + -1.9765e-04, -7.3528e-04], + [ 1.0765e-04, 8.5115e-05, 1.8334e-04, ..., 2.2244e-04, + 5.8711e-05, 1.5545e-04], + ..., + [ 3.3617e-05, 7.1824e-05, 4.7207e-05, ..., 1.8090e-05, + 5.4508e-05, 1.0985e-04], + [ 3.5107e-05, 6.9499e-05, 5.9783e-05, ..., 6.2406e-05, + 2.9176e-05, 1.2994e-04], + [ 6.3330e-06, -3.3283e-04, -2.9469e-04, ..., 1.0379e-05, + -9.7394e-05, 1.5691e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0076, -0.0077, -0.0036, -0.0123, -0.0127, -0.0055, 0.0110, 0.0020, + 0.0235, -0.0023], device='cuda:0'), grad: tensor([ 5.5939e-05, -1.4257e-03, 5.8746e-04, -4.3416e-04, 5.4789e-04, + 1.9526e-04, 2.7561e-04, 2.7657e-04, 3.4571e-04, -4.2415e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.68, cls_loss 0.0116 cls_loss_mapping 0.0155 cls_loss_causal 0.6524 re_mapping 0.0135 re_causal 0.0361 /// teacc 98.79 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0955, 0.0572, 0.0053, ..., -0.0065, -0.0327, -0.0546], + [ 0.0310, 0.0479, -0.0698, ..., -0.0665, -0.0183, 0.0598], + [ 0.0519, -0.0513, -0.0577, ..., -0.0003, -0.0315, -0.0137], + ..., + [ 0.0410, 0.0269, 0.0631, ..., 0.0333, 0.0917, 0.0165], + [ 0.0502, -0.0773, -0.0681, ..., -0.0956, -0.0363, 0.0790], + [-0.0462, 0.0668, 0.0177, ..., -0.0405, -0.0339, 0.0105]], + device='cuda:0'), grad: tensor([[ 7.1749e-06, -2.4140e-06, 1.0736e-05, ..., 4.6194e-06, + 1.1288e-05, 1.9241e-06], + [-1.1816e-03, -1.8013e-04, -1.1883e-03, ..., -1.9178e-03, + -2.4033e-03, -2.1644e-06], + [ 9.5940e-04, 1.6081e-04, 1.0014e-03, ..., 1.4763e-03, + 1.8845e-03, 1.5117e-05], + ..., + [-2.2566e-04, -8.2910e-05, -4.0340e-04, ..., 3.0547e-05, + -1.4603e-04, -8.6010e-05], + [ 5.2303e-05, 7.7665e-05, 1.3137e-04, ..., 3.1918e-05, + 6.0409e-05, 4.3035e-05], + [ 5.2273e-05, -2.7561e-04, -2.3007e-04, ..., 2.6360e-05, + 7.2896e-05, -1.4162e-04]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0079, -0.0077, -0.0040, -0.0125, -0.0129, -0.0048, 0.0113, 0.0027, + 0.0229, -0.0026], device='cuda:0'), grad: tensor([ 3.6545e-06, -4.7455e-03, 3.6755e-03, 1.1730e-03, 3.4869e-05, + 3.3826e-05, 1.5602e-05, 2.1860e-05, 1.9765e-04, -4.0603e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 214.88, cls_loss 0.0116 cls_loss_mapping 0.0173 cls_loss_causal 0.6437 re_mapping 0.0137 re_causal 0.0377 /// teacc 98.71 lr 0.00010000 +Epoch 70, weight, value: tensor([[-9.5797e-02, 5.7788e-02, 6.6002e-03, ..., -6.4182e-03, + -3.3385e-02, -5.5504e-02], + [ 3.1207e-02, 4.8217e-02, -6.9822e-02, ..., -6.6095e-02, + -1.7887e-02, 6.0344e-02], + [ 5.2696e-02, -5.2102e-02, -5.7815e-02, ..., -3.6206e-05, + -3.1531e-02, -1.2868e-02], + ..., + [ 4.1461e-02, 2.7217e-02, 6.3732e-02, ..., 3.3716e-02, + 9.2492e-02, 1.6250e-02], + [ 4.9684e-02, -7.7900e-02, -6.9060e-02, ..., -9.6234e-02, + -3.7097e-02, 7.8802e-02], + [-4.6658e-02, 6.7132e-02, 1.7915e-02, ..., -4.1135e-02, + -3.3683e-02, 1.0405e-02]], device='cuda:0'), grad: tensor([[ 9.0823e-06, 6.8881e-06, 4.8392e-06, ..., 1.1874e-06, + 5.1335e-06, 1.4916e-05], + [-3.1263e-05, 3.7217e-04, 1.7333e-04, ..., 3.5577e-06, + 1.1241e-04, 1.2107e-05], + [ 1.2457e-05, 2.0906e-05, 6.0610e-06, ..., -4.5411e-06, + 1.1958e-05, 3.3677e-05], + ..., + [-2.7686e-05, 1.4782e-04, -1.1094e-05, ..., -2.5094e-05, + 3.0715e-06, 7.1883e-05], + [-6.1691e-05, -4.5925e-05, 6.2697e-06, ..., 2.7250e-06, + 5.7071e-06, -3.1543e-04], + [ 1.9342e-05, -7.9060e-04, -3.0565e-04, ..., 1.1012e-05, + -1.6642e-04, -1.9252e-04]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0077, -0.0073, -0.0037, -0.0125, -0.0133, -0.0054, 0.0117, 0.0029, + 0.0223, -0.0027], device='cuda:0'), grad: tensor([ 2.9817e-05, 3.5238e-04, 3.7223e-05, 1.5426e-04, 5.4985e-05, + 2.0266e-04, 2.0432e-04, 1.4651e-04, -3.6955e-04, -8.1301e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 214.85, cls_loss 0.0095 cls_loss_mapping 0.0135 cls_loss_causal 0.6541 re_mapping 0.0128 re_causal 0.0376 /// teacc 98.77 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0963, 0.0580, 0.0063, ..., -0.0062, -0.0339, -0.0561], + [ 0.0312, 0.0485, -0.0701, ..., -0.0665, -0.0178, 0.0605], + [ 0.0524, -0.0527, -0.0587, ..., -0.0003, -0.0318, -0.0134], + ..., + [ 0.0421, 0.0278, 0.0646, ..., 0.0343, 0.0935, 0.0161], + [ 0.0499, -0.0783, -0.0696, ..., -0.0967, -0.0374, 0.0794], + [-0.0472, 0.0671, 0.0179, ..., -0.0417, -0.0342, 0.0100]], + device='cuda:0'), grad: tensor([[ 1.6645e-05, 2.9709e-06, 1.5721e-06, ..., -8.8755e-07, + 7.3574e-07, 3.3349e-05], + [ 1.0677e-05, 5.1320e-05, 5.8971e-06, ..., 5.7258e-06, + 3.2261e-06, 5.8021e-07], + [-1.7798e-04, -5.0575e-05, 2.9951e-06, ..., -6.8545e-05, + -5.2810e-05, -7.3574e-07], + ..., + [ 7.9334e-05, 2.2531e-05, 8.6129e-06, ..., 4.7356e-05, + 4.4137e-05, 5.7332e-06], + [-5.2512e-05, -3.9190e-05, 3.7197e-06, ..., 6.9663e-06, + 1.0571e-06, -1.8442e-04], + [ 6.9559e-05, 1.9467e-04, 5.0664e-06, ..., 2.5302e-05, + 3.6228e-07, 4.3422e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0078, -0.0073, -0.0041, -0.0124, -0.0135, -0.0054, 0.0122, 0.0036, + 0.0223, -0.0030], device='cuda:0'), grad: tensor([-2.4676e-05, 2.2268e-04, -8.8453e-04, 3.0088e-04, 1.8954e-04, + -5.7125e-04, 3.5000e-04, 1.7214e-04, -2.6393e-04, 5.0831e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 215.06, cls_loss 0.0093 cls_loss_mapping 0.0134 cls_loss_causal 0.6893 re_mapping 0.0127 re_causal 0.0378 /// teacc 98.68 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0968, 0.0581, 0.0057, ..., -0.0063, -0.0345, -0.0565], + [ 0.0308, 0.0482, -0.0707, ..., -0.0668, -0.0181, 0.0604], + [ 0.0522, -0.0536, -0.0594, ..., -0.0001, -0.0323, -0.0140], + ..., + [ 0.0427, 0.0279, 0.0650, ..., 0.0343, 0.0942, 0.0164], + [ 0.0501, -0.0788, -0.0703, ..., -0.0975, -0.0377, 0.0804], + [-0.0472, 0.0674, 0.0182, ..., -0.0422, -0.0345, 0.0100]], + device='cuda:0'), grad: tensor([[ 4.5300e-06, 3.0920e-06, 8.4564e-06, ..., 4.6119e-06, + 3.2689e-06, 2.2091e-06], + [ 7.3552e-05, 4.6372e-05, 5.6356e-05, ..., 2.6882e-05, + 8.0466e-05, 2.2464e-06], + [ 3.1686e-04, 4.9204e-05, 1.8239e-04, ..., 1.4746e-04, + 2.7514e-04, 1.0066e-05], + ..., + [-4.3964e-04, -1.9789e-04, -2.8872e-04, ..., -1.5759e-04, + -4.6039e-04, -4.3176e-06], + [-1.4566e-05, 2.9057e-05, 4.1217e-05, ..., 1.6481e-05, + 5.4017e-06, -3.8415e-05], + [ 1.3411e-05, 3.8803e-05, -7.8604e-06, ..., 1.0207e-05, + 4.2379e-05, 3.6880e-06]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0079, -0.0077, -0.0043, -0.0125, -0.0136, -0.0050, 0.0122, 0.0038, + 0.0222, -0.0028], device='cuda:0'), grad: tensor([ 1.5110e-05, 1.3900e-04, 2.7275e-04, 1.4752e-05, 5.1647e-05, + 3.8803e-05, 1.1645e-05, -6.4182e-04, 2.0236e-05, 7.8380e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 214.90, cls_loss 0.0088 cls_loss_mapping 0.0123 cls_loss_causal 0.6377 re_mapping 0.0128 re_causal 0.0374 /// teacc 98.69 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0973, 0.0580, 0.0053, ..., -0.0068, -0.0349, -0.0569], + [ 0.0307, 0.0483, -0.0717, ..., -0.0662, -0.0185, 0.0607], + [ 0.0517, -0.0546, -0.0596, ..., -0.0004, -0.0327, -0.0148], + ..., + [ 0.0430, 0.0280, 0.0649, ..., 0.0340, 0.0946, 0.0167], + [ 0.0509, -0.0787, -0.0707, ..., -0.0979, -0.0378, 0.0814], + [-0.0470, 0.0680, 0.0188, ..., -0.0426, -0.0342, 0.0096]], + device='cuda:0'), grad: tensor([[-1.6959e-06, -6.9588e-06, 2.1346e-06, ..., 3.5577e-07, + -2.5574e-06, 4.4629e-06], + [ 8.5175e-05, 1.6046e-04, 2.0564e-06, ..., 1.6978e-06, + 1.9693e-04, -3.7346e-07], + [-7.4506e-05, -1.5306e-04, 2.7083e-06, ..., -1.5898e-06, + -1.9455e-04, 1.4551e-05], + ..., + [-1.3955e-05, 1.3366e-05, -8.7991e-06, ..., -1.2420e-05, + -1.6019e-05, 7.1786e-06], + [-1.2589e-04, 1.0423e-05, 1.9521e-06, ..., 1.8999e-06, + 2.1048e-06, -1.9085e-04], + [ 2.0504e-05, 3.8600e-04, -2.4125e-05, ..., 2.5108e-06, + 7.9814e-07, 2.3639e-04]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0082, -0.0074, -0.0051, -0.0124, -0.0140, -0.0047, 0.0123, 0.0034, + 0.0226, -0.0023], device='cuda:0'), grad: tensor([-2.0303e-06, 4.3654e-04, -3.9864e-04, 1.0622e-04, -6.7091e-04, + -7.3016e-05, 2.6107e-04, 1.6972e-05, -3.4857e-04, 6.7377e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.92, cls_loss 0.0094 cls_loss_mapping 0.0141 cls_loss_causal 0.6511 re_mapping 0.0134 re_causal 0.0380 /// teacc 98.73 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0983, 0.0584, 0.0049, ..., -0.0070, -0.0359, -0.0574], + [ 0.0306, 0.0481, -0.0723, ..., -0.0675, -0.0186, 0.0612], + [ 0.0509, -0.0547, -0.0611, ..., -0.0013, -0.0347, -0.0150], + ..., + [ 0.0438, 0.0286, 0.0657, ..., 0.0355, 0.0960, 0.0159], + [ 0.0521, -0.0795, -0.0705, ..., -0.0981, -0.0370, 0.0824], + [-0.0474, 0.0675, 0.0190, ..., -0.0428, -0.0347, 0.0088]], + device='cuda:0'), grad: tensor([[ 1.7704e-06, -1.7130e-04, 3.4347e-06, ..., 1.8291e-06, + 2.6654e-06, 2.9150e-06], + [-2.5239e-07, 1.7248e-06, 9.3281e-06, ..., 3.4552e-06, + 4.0010e-06, -8.4266e-06], + [ 2.2091e-06, 6.8508e-06, 5.1744e-06, ..., -1.1176e-07, + 3.3639e-06, 3.1777e-06], + ..., + [-5.7042e-05, -3.1382e-05, -6.9022e-05, ..., -2.3693e-05, + -9.4235e-05, -1.2964e-05], + [ 3.2485e-06, 1.9550e-05, 1.4067e-05, ..., 7.2531e-06, + 3.9898e-06, 2.1979e-05], + [ 1.8820e-05, -6.7949e-05, -9.1553e-05, ..., -1.7345e-05, + -2.4408e-05, -1.8068e-07]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0081, -0.0078, -0.0053, -0.0130, -0.0130, -0.0046, 0.0119, 0.0041, + 0.0227, -0.0029], device='cuda:0'), grad: tensor([-2.7108e-04, 1.2726e-05, 1.3225e-05, -2.0251e-05, 1.4985e-04, + -8.6606e-05, 2.7156e-04, -6.7651e-05, 8.4937e-05, -8.7440e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 214.80, cls_loss 0.0112 cls_loss_mapping 0.0172 cls_loss_causal 0.6517 re_mapping 0.0124 re_causal 0.0359 /// teacc 98.63 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0987, 0.0589, 0.0044, ..., -0.0071, -0.0368, -0.0576], + [ 0.0306, 0.0491, -0.0729, ..., -0.0685, -0.0182, 0.0622], + [ 0.0515, -0.0553, -0.0617, ..., -0.0017, -0.0348, -0.0144], + ..., + [ 0.0440, 0.0287, 0.0673, ..., 0.0363, 0.0968, 0.0150], + [ 0.0518, -0.0801, -0.0711, ..., -0.0986, -0.0376, 0.0827], + [-0.0471, 0.0671, 0.0186, ..., -0.0430, -0.0351, 0.0083]], + device='cuda:0'), grad: tensor([[ 6.0499e-06, -2.4334e-05, 1.2964e-06, ..., -6.8992e-06, + 2.0172e-06, 5.1297e-06], + [ 2.2352e-05, 6.6683e-06, 1.9535e-05, ..., 1.0237e-05, + 2.8372e-05, -1.2271e-05], + [-4.9084e-05, 6.4969e-06, 5.5954e-06, ..., -3.8177e-05, + -1.1288e-05, 1.6838e-05], + ..., + [-2.4602e-05, -8.3372e-06, -4.0442e-05, ..., -2.9430e-07, + -5.0396e-05, 1.0468e-05], + [ 1.3985e-05, 1.1049e-05, 4.5970e-06, ..., 3.2067e-05, + 1.1779e-05, -3.5673e-05], + [ 2.5064e-05, 1.4216e-05, -8.9407e-07, ..., 1.3143e-05, + 8.7470e-06, 2.2203e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0077, -0.0081, -0.0051, -0.0132, -0.0128, -0.0047, 0.0120, 0.0046, + 0.0221, -0.0031], device='cuda:0'), grad: tensor([-4.2021e-05, 2.6390e-05, -3.5286e-05, -6.6161e-05, 5.5507e-06, + -2.2814e-05, 3.3081e-05, -1.1891e-05, 2.6137e-05, 8.6904e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 215.15, cls_loss 0.0081 cls_loss_mapping 0.0163 cls_loss_causal 0.6193 re_mapping 0.0127 re_causal 0.0374 /// teacc 98.58 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0992, 0.0593, 0.0043, ..., -0.0068, -0.0376, -0.0582], + [ 0.0302, 0.0488, -0.0734, ..., -0.0687, -0.0186, 0.0620], + [ 0.0516, -0.0554, -0.0619, ..., -0.0014, -0.0348, -0.0150], + ..., + [ 0.0445, 0.0296, 0.0681, ..., 0.0364, 0.0976, 0.0149], + [ 0.0524, -0.0804, -0.0712, ..., -0.0990, -0.0376, 0.0838], + [-0.0479, 0.0670, 0.0183, ..., -0.0439, -0.0359, 0.0081]], + device='cuda:0'), grad: tensor([[ 2.5127e-06, 1.1707e-06, 1.5115e-06, ..., 5.0552e-06, + 8.9221e-07, 6.4261e-07], + [ 5.8338e-06, 8.2105e-06, 9.6336e-06, ..., 1.1720e-05, + 6.0722e-06, -3.1460e-06], + [-8.7768e-06, 8.0094e-06, 3.0752e-06, ..., -9.1419e-06, + -2.7604e-06, 6.5751e-06], + ..., + [-6.0536e-06, 2.2352e-07, -9.2313e-06, ..., 3.9898e-06, + -1.1474e-05, 2.2482e-06], + [-3.7663e-06, 1.5885e-05, 5.0701e-06, ..., 9.6709e-06, + 8.9314e-07, -1.1735e-05], + [ 3.7551e-06, 3.5409e-06, 5.2433e-07, ..., 6.8918e-06, + 2.8927e-06, 2.6561e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0073, -0.0087, -0.0049, -0.0136, -0.0131, -0.0047, 0.0122, 0.0053, + 0.0224, -0.0035], device='cuda:0'), grad: tensor([ 1.2010e-05, 2.3037e-05, 1.3430e-06, -2.4533e-04, -9.5218e-06, + 1.4472e-04, 3.2067e-05, 3.6880e-06, 2.2218e-05, 1.5363e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 215.04, cls_loss 0.0077 cls_loss_mapping 0.0111 cls_loss_causal 0.6284 re_mapping 0.0125 re_causal 0.0356 /// teacc 98.72 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.1000, 0.0596, 0.0041, ..., -0.0067, -0.0378, -0.0591], + [ 0.0302, 0.0487, -0.0736, ..., -0.0685, -0.0185, 0.0622], + [ 0.0519, -0.0555, -0.0617, ..., -0.0007, -0.0348, -0.0151], + ..., + [ 0.0450, 0.0297, 0.0686, ..., 0.0366, 0.0983, 0.0147], + [ 0.0528, -0.0809, -0.0715, ..., -0.0996, -0.0378, 0.0849], + [-0.0486, 0.0670, 0.0182, ..., -0.0444, -0.0362, 0.0077]], + device='cuda:0'), grad: tensor([[ 1.3597e-05, 8.5831e-05, 7.6771e-05, ..., 1.8671e-05, + 1.3225e-06, 5.4538e-05], + [-1.8552e-06, 4.3660e-05, 4.4644e-05, ..., 1.1042e-05, + 2.8778e-07, -1.5333e-05], + [ 3.1352e-05, 6.6817e-05, 4.4495e-05, ..., 5.2810e-05, + 8.5309e-07, 5.9038e-05], + ..., + [ 6.7428e-07, 2.8014e-04, 2.4533e-04, ..., 4.9211e-06, + 2.0266e-04, 2.4870e-05], + [ 9.3400e-05, 2.4962e-04, 1.2136e-04, ..., 3.4285e-04, + 3.3341e-06, 6.4492e-05], + [ 2.3656e-06, -1.0786e-03, -9.6416e-04, ..., 7.4506e-06, + -2.4939e-04, -3.4237e-04]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0071, -0.0087, -0.0045, -0.0136, -0.0131, -0.0050, 0.0121, 0.0054, + 0.0226, -0.0038], device='cuda:0'), grad: tensor([ 2.2173e-04, 7.7367e-05, 1.9455e-04, 1.1677e-04, 1.5485e-04, + 4.3559e-04, 2.8044e-05, 4.7708e-04, 5.8079e-04, -2.2869e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 215.06, cls_loss 0.0086 cls_loss_mapping 0.0131 cls_loss_causal 0.6295 re_mapping 0.0118 re_causal 0.0340 /// teacc 98.71 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.1014, 0.0595, 0.0038, ..., -0.0076, -0.0384, -0.0608], + [ 0.0306, 0.0492, -0.0739, ..., -0.0685, -0.0180, 0.0629], + [ 0.0514, -0.0560, -0.0622, ..., -0.0008, -0.0353, -0.0155], + ..., + [ 0.0450, 0.0290, 0.0685, ..., 0.0361, 0.0984, 0.0140], + [ 0.0534, -0.0817, -0.0719, ..., -0.1004, -0.0381, 0.0857], + [-0.0490, 0.0673, 0.0183, ..., -0.0448, -0.0364, 0.0073]], + device='cuda:0'), grad: tensor([[ 3.3733e-06, 1.6987e-04, 7.4506e-05, ..., 1.9461e-05, + 6.8592e-07, 2.1476e-06], + [-3.0976e-06, -5.6297e-05, -5.1185e-06, ..., 6.4857e-06, + -8.2925e-06, -5.1141e-05], + [ 4.2981e-07, 1.3483e-04, 7.7128e-05, ..., 3.4481e-05, + -4.7274e-06, 4.4294e-06], + ..., + [ 4.0419e-06, 4.8965e-05, 2.5362e-05, ..., 8.0839e-06, + 2.1867e-06, 1.5810e-05], + [ 3.0458e-05, 2.5466e-05, 1.7241e-05, ..., 1.0230e-05, + 2.2314e-06, 2.3097e-06], + [ 5.4762e-06, -9.4509e-04, -4.3654e-04, ..., -4.8399e-05, + 2.1886e-07, -1.1297e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0081, -0.0081, -0.0048, -0.0131, -0.0132, -0.0044, 0.0118, 0.0050, + 0.0223, -0.0036], device='cuda:0'), grad: tensor([ 0.0003, -0.0002, 0.0003, 0.0006, 0.0001, 0.0008, -0.0010, 0.0001, + 0.0003, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 215.07, cls_loss 0.0065 cls_loss_mapping 0.0110 cls_loss_causal 0.5932 re_mapping 0.0124 re_causal 0.0350 /// teacc 98.68 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.1018, 0.0595, 0.0036, ..., -0.0077, -0.0388, -0.0611], + [ 0.0309, 0.0492, -0.0742, ..., -0.0685, -0.0179, 0.0630], + [ 0.0512, -0.0562, -0.0626, ..., -0.0008, -0.0356, -0.0158], + ..., + [ 0.0452, 0.0290, 0.0689, ..., 0.0362, 0.0988, 0.0139], + [ 0.0538, -0.0821, -0.0722, ..., -0.1007, -0.0381, 0.0863], + [-0.0492, 0.0678, 0.0185, ..., -0.0450, -0.0361, 0.0072]], + device='cuda:0'), grad: tensor([[ 1.0908e-05, -2.0973e-06, 5.8375e-06, ..., 5.4650e-06, + 7.1116e-06, 7.4096e-06], + [ 1.3697e-04, 2.5928e-05, 1.1516e-04, ..., 7.7426e-05, + 1.6737e-04, -3.4720e-06], + [-2.2009e-05, 7.7337e-06, 2.3857e-05, ..., -3.0726e-05, + 2.2128e-05, 6.5714e-06], + ..., + [-2.8777e-04, -6.4790e-05, -2.8253e-04, ..., -1.6701e-04, + -3.4499e-04, 1.1161e-05], + [-1.5068e-04, -6.7353e-05, 1.2174e-05, ..., 2.5749e-05, + 1.6883e-05, -3.4428e-04], + [ 1.8820e-05, -3.0905e-05, -2.2635e-05, ..., 1.2644e-05, + 1.3493e-05, 3.8780e-06]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0080, -0.0082, -0.0046, -0.0133, -0.0129, -0.0043, 0.0113, 0.0051, + 0.0220, -0.0034], device='cuda:0'), grad: tensor([ 1.9416e-05, 2.2316e-04, -6.8247e-05, 6.2227e-04, -2.3887e-05, + 2.4423e-05, -3.1441e-05, -4.1318e-04, -3.3879e-04, -1.4596e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 215.06, cls_loss 0.0082 cls_loss_mapping 0.0136 cls_loss_causal 0.6094 re_mapping 0.0122 re_causal 0.0341 /// teacc 98.73 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.1029, 0.0588, 0.0033, ..., -0.0077, -0.0398, -0.0617], + [ 0.0308, 0.0495, -0.0745, ..., -0.0689, -0.0183, 0.0637], + [ 0.0513, -0.0569, -0.0632, ..., -0.0009, -0.0353, -0.0162], + ..., + [ 0.0459, 0.0291, 0.0693, ..., 0.0363, 0.0995, 0.0138], + [ 0.0541, -0.0830, -0.0725, ..., -0.1007, -0.0383, 0.0867], + [-0.0497, 0.0686, 0.0186, ..., -0.0453, -0.0365, 0.0066]], + device='cuda:0'), grad: tensor([[ 3.9518e-05, 2.4036e-05, 1.4640e-06, ..., 9.6411e-06, + 6.2212e-06, 6.0737e-05], + [-3.8934e-04, -5.1737e-04, 3.4347e-06, ..., 2.1234e-06, + -6.8665e-05, -7.3814e-04], + [ 5.3376e-05, 9.6500e-05, 3.1516e-06, ..., -5.4657e-05, + 1.7658e-05, 1.6117e-04], + ..., + [ 1.2904e-05, 2.7478e-05, -8.4490e-06, ..., 1.9819e-06, + -9.1568e-06, 3.7789e-05], + [ 1.9884e-04, 3.0684e-04, 6.5342e-06, ..., 2.3693e-05, + 3.8415e-05, 3.5667e-04], + [ 1.1601e-05, -5.4576e-06, -1.9029e-05, ..., 5.1707e-06, + -2.7902e-06, 1.9297e-05]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0088, -0.0082, -0.0047, -0.0139, -0.0128, -0.0042, 0.0113, 0.0055, + 0.0219, -0.0030], device='cuda:0'), grad: tensor([ 1.0753e-04, -1.1368e-03, 5.2989e-05, 7.7009e-05, 3.6955e-05, + 5.7489e-05, 5.0038e-05, 6.5029e-05, 6.8188e-04, 9.3058e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.88, cls_loss 0.0094 cls_loss_mapping 0.0140 cls_loss_causal 0.6520 re_mapping 0.0118 re_causal 0.0348 /// teacc 98.73 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.1040, 0.0592, 0.0041, ..., -0.0080, -0.0406, -0.0625], + [ 0.0310, 0.0501, -0.0745, ..., -0.0691, -0.0181, 0.0646], + [ 0.0518, -0.0576, -0.0632, ..., -0.0005, -0.0348, -0.0166], + ..., + [ 0.0457, 0.0286, 0.0693, ..., 0.0363, 0.0995, 0.0135], + [ 0.0543, -0.0841, -0.0726, ..., -0.1011, -0.0386, 0.0872], + [-0.0496, 0.0691, 0.0188, ..., -0.0457, -0.0366, 0.0063]], + device='cuda:0'), grad: tensor([[ 1.8388e-05, -2.3901e-04, 1.1906e-05, ..., -2.2739e-05, + 1.0930e-05, 3.7216e-06], + [ 1.7118e-04, 2.5302e-05, 3.5912e-05, ..., 7.3254e-05, + 1.4246e-04, -3.3509e-06], + [-1.8561e-04, 4.7117e-05, 1.8284e-05, ..., -6.1691e-05, + -9.3818e-05, -1.2647e-06], + ..., + [-1.4842e-04, -1.5080e-04, -2.8896e-04, ..., -9.1717e-06, + -3.4404e-04, 3.0193e-06], + [ 8.7619e-06, 1.9863e-05, 1.0245e-05, ..., 7.4469e-06, + 4.5635e-06, 2.1663e-06], + [ 9.6977e-05, 2.1291e-04, 1.5986e-04, ..., 1.6496e-05, + 2.3055e-04, 8.6939e-07]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0085, -0.0078, -0.0045, -0.0132, -0.0124, -0.0051, 0.0114, 0.0050, + 0.0213, -0.0028], device='cuda:0'), grad: tensor([-4.6968e-04, 3.0351e-04, -2.3174e-04, 6.3837e-05, 5.5134e-05, + 3.3706e-05, 5.2840e-05, -3.6907e-04, 5.3585e-05, 5.0783e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 80---------------------------------------------------- +epoch 80, time 231.50, cls_loss 0.0073 cls_loss_mapping 0.0132 cls_loss_causal 0.6114 re_mapping 0.0116 re_causal 0.0325 /// teacc 98.84 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.1048, 0.0603, 0.0038, ..., -0.0067, -0.0416, -0.0629], + [ 0.0310, 0.0507, -0.0746, ..., -0.0695, -0.0178, 0.0646], + [ 0.0519, -0.0581, -0.0635, ..., -0.0003, -0.0346, -0.0173], + ..., + [ 0.0457, 0.0279, 0.0695, ..., 0.0364, 0.0995, 0.0129], + [ 0.0553, -0.0842, -0.0728, ..., -0.1016, -0.0384, 0.0888], + [-0.0501, 0.0693, 0.0190, ..., -0.0461, -0.0367, 0.0059]], + device='cuda:0'), grad: tensor([[ 3.1590e-06, -3.9823e-06, 6.5304e-06, ..., 4.0047e-06, + 7.3668e-07, 1.3625e-06], + [ 1.7695e-08, -2.5347e-05, 1.2204e-05, ..., 7.4841e-06, + 3.0827e-07, -3.9965e-05], + [ 5.2452e-05, 5.2713e-06, 1.0842e-04, ..., 6.7115e-05, + 9.2909e-06, 4.7795e-06], + ..., + [ 1.4257e-04, 1.1273e-05, 2.9922e-04, ..., 1.8549e-04, + 1.8895e-05, 1.2405e-05], + [ 6.8098e-06, 4.8913e-06, 1.5274e-05, ..., 9.0152e-06, + 1.8422e-06, 3.0454e-06], + [ 1.9856e-06, 4.7311e-06, -7.1991e-07, ..., 2.2128e-06, + 3.5996e-07, 7.2643e-06]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0077, -0.0080, -0.0045, -0.0139, -0.0118, -0.0051, 0.0107, 0.0047, + 0.0220, -0.0028], device='cuda:0'), grad: tensor([ 7.5623e-07, -6.3598e-05, 1.2046e-04, -4.4727e-04, 1.1146e-05, + 7.3090e-06, 2.5034e-06, 3.2616e-04, 2.5824e-05, 1.6734e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 215.03, cls_loss 0.0090 cls_loss_mapping 0.0148 cls_loss_causal 0.6320 re_mapping 0.0121 re_causal 0.0328 /// teacc 98.64 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.1059, 0.0606, 0.0035, ..., -0.0072, -0.0426, -0.0637], + [ 0.0301, 0.0507, -0.0759, ..., -0.0701, -0.0189, 0.0645], + [ 0.0523, -0.0585, -0.0638, ..., 0.0003, -0.0342, -0.0175], + ..., + [ 0.0467, 0.0281, 0.0705, ..., 0.0367, 0.1007, 0.0132], + [ 0.0557, -0.0846, -0.0732, ..., -0.1025, -0.0386, 0.0898], + [-0.0506, 0.0698, 0.0193, ..., -0.0468, -0.0371, 0.0057]], + device='cuda:0'), grad: tensor([[ 1.6084e-06, 5.5833e-07, 1.6587e-06, ..., 1.0571e-06, + 9.4017e-07, 2.4252e-06], + [-5.2869e-05, -6.4015e-05, 3.5539e-06, ..., 6.1914e-06, + 3.9786e-06, -9.9599e-05], + [ 4.4155e-04, 1.6719e-05, 4.2176e-04, ..., 3.3236e-04, + 7.4530e-04, 1.1578e-05], + ..., + [-4.5013e-04, 7.9814e-07, -4.3225e-04, ..., -3.4308e-04, + -7.6675e-04, 1.2226e-05], + [ 1.2293e-05, 2.6748e-05, 4.1202e-06, ..., 4.9435e-06, + 1.5134e-06, 2.1458e-05], + [ 1.0692e-05, 1.7643e-05, 2.0981e-05, ..., 1.8179e-05, + 4.2021e-06, 1.5825e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0076, -0.0090, -0.0040, -0.0142, -0.0122, -0.0054, 0.0110, 0.0053, + 0.0221, -0.0027], device='cuda:0'), grad: tensor([ 1.0870e-05, -1.7822e-04, 7.2670e-04, -3.3200e-05, -1.4231e-05, + 1.8582e-05, 5.2392e-05, -7.1335e-04, 6.9857e-05, 6.0618e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 214.93, cls_loss 0.0071 cls_loss_mapping 0.0120 cls_loss_causal 0.5977 re_mapping 0.0121 re_causal 0.0341 /// teacc 98.60 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.1066, 0.0606, 0.0031, ..., -0.0074, -0.0433, -0.0643], + [ 0.0301, 0.0508, -0.0760, ..., -0.0704, -0.0190, 0.0650], + [ 0.0518, -0.0593, -0.0648, ..., -0.0004, -0.0349, -0.0179], + ..., + [ 0.0476, 0.0283, 0.0713, ..., 0.0375, 0.1018, 0.0131], + [ 0.0559, -0.0852, -0.0739, ..., -0.1033, -0.0389, 0.0902], + [-0.0514, 0.0696, 0.0192, ..., -0.0475, -0.0379, 0.0051]], + device='cuda:0'), grad: tensor([[-6.8219e-07, -3.4831e-06, 3.3174e-06, ..., 3.6787e-06, + -7.4832e-07, 1.6550e-06], + [ 4.0978e-06, 2.4572e-05, 4.0904e-06, ..., 4.7423e-06, + 3.0808e-06, 4.7265e-07], + [-4.0621e-05, -3.2216e-05, -6.7234e-05, ..., -2.6727e-04, + -2.8342e-05, 1.2971e-05], + ..., + [ 3.4958e-05, 3.7730e-05, 2.6107e-05, ..., 2.5898e-05, + 2.0906e-05, 1.8030e-05], + [-2.3469e-05, 1.3851e-05, 7.7114e-06, ..., 1.4886e-05, + -2.5295e-06, -2.8104e-05], + [ 2.0340e-05, -5.9545e-05, -6.0260e-05, ..., 7.2718e-06, + -1.0282e-06, 2.8521e-05]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0080, -0.0089, -0.0048, -0.0142, -0.0116, -0.0050, 0.0114, 0.0059, + 0.0218, -0.0032], device='cuda:0'), grad: tensor([ 5.4426e-06, 1.0514e-04, -3.8910e-04, 4.1866e-04, -2.3544e-04, + -1.5521e-04, 8.7619e-05, 1.4770e-04, 3.7938e-05, -2.2978e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 214.94, cls_loss 0.0094 cls_loss_mapping 0.0134 cls_loss_causal 0.6151 re_mapping 0.0117 re_causal 0.0323 /// teacc 98.73 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.1078, 0.0607, 0.0026, ..., -0.0077, -0.0442, -0.0652], + [ 0.0303, 0.0517, -0.0758, ..., -0.0712, -0.0185, 0.0659], + [ 0.0513, -0.0604, -0.0666, ..., -0.0010, -0.0357, -0.0180], + ..., + [ 0.0482, 0.0282, 0.0719, ..., 0.0380, 0.1030, 0.0122], + [ 0.0565, -0.0857, -0.0747, ..., -0.1040, -0.0394, 0.0910], + [-0.0522, 0.0696, 0.0194, ..., -0.0480, -0.0386, 0.0047]], + device='cuda:0'), grad: tensor([[ 5.7757e-05, 1.9491e-05, 3.0413e-05, ..., 2.0251e-05, + 3.5256e-05, 3.5018e-05], + [ 2.1338e-04, 3.8177e-05, 1.3685e-04, ..., 3.2216e-05, + 1.5676e-04, 1.3125e-04], + [ 1.4150e-04, 2.8938e-05, 3.8475e-05, ..., -6.1810e-05, + 6.5446e-05, 1.7953e-04], + ..., + [ 9.8228e-04, 1.5390e-04, 6.8665e-04, ..., 1.4973e-04, + 7.5817e-04, 5.7554e-04], + [-2.8276e-04, -1.9968e-04, 1.3173e-04, ..., -3.3468e-05, + 2.9624e-05, -3.9721e-04], + [-1.7071e-03, -1.9598e-04, -1.2560e-03, ..., -1.8954e-04, + -1.3876e-03, -9.5081e-04]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0083, -0.0087, -0.0055, -0.0141, -0.0108, -0.0045, 0.0107, 0.0058, + 0.0220, -0.0035], device='cuda:0'), grad: tensor([ 1.1903e-04, 4.8351e-04, 1.7512e-04, 4.4727e-04, 4.5013e-04, + 1.0210e-04, 4.0084e-05, 2.4204e-03, -3.5137e-05, -4.2000e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 214.85, cls_loss 0.0105 cls_loss_mapping 0.0155 cls_loss_causal 0.5995 re_mapping 0.0115 re_causal 0.0320 /// teacc 98.80 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.1090, 0.0602, 0.0016, ..., -0.0082, -0.0454, -0.0657], + [ 0.0304, 0.0512, -0.0759, ..., -0.0718, -0.0185, 0.0667], + [ 0.0512, -0.0596, -0.0673, ..., -0.0010, -0.0360, -0.0180], + ..., + [ 0.0483, 0.0282, 0.0720, ..., 0.0373, 0.1035, 0.0121], + [ 0.0575, -0.0858, -0.0746, ..., -0.1044, -0.0397, 0.0920], + [-0.0523, 0.0714, 0.0199, ..., -0.0488, -0.0387, 0.0060]], + device='cuda:0'), grad: tensor([[ 2.9206e-06, -2.2918e-05, 2.4419e-06, ..., -9.3430e-06, + 2.2966e-06, 1.0384e-06], + [ 1.4625e-05, -5.1931e-06, 2.2098e-05, ..., 1.3784e-05, + 2.2471e-05, -3.8028e-05], + [ 4.8423e-04, 1.9407e-04, 3.6526e-04, ..., 2.5272e-04, + 3.6240e-04, 2.3633e-05], + ..., + [-5.7793e-04, -1.8823e-04, -4.3368e-04, ..., -2.9159e-04, + -4.4107e-04, 1.3709e-06], + [ 1.6585e-05, 1.1787e-05, 1.5542e-05, ..., 9.7305e-06, + 1.3366e-05, 2.3004e-06], + [ 8.4192e-06, 4.3869e-05, -2.0452e-06, ..., 5.8748e-06, + 7.3276e-06, 2.8741e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0091, -0.0089, -0.0049, -0.0137, -0.0126, -0.0051, 0.0102, 0.0055, + 0.0224, -0.0015], device='cuda:0'), grad: tensor([-5.1767e-05, -1.0192e-05, 8.5545e-04, 4.3988e-05, -1.6224e-04, + 1.3627e-05, 4.8488e-05, -8.7070e-04, 3.9846e-05, 9.3102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 214.78, cls_loss 0.0070 cls_loss_mapping 0.0097 cls_loss_causal 0.6229 re_mapping 0.0112 re_causal 0.0327 /// teacc 98.78 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.1096, 0.0611, 0.0017, ..., -0.0063, -0.0460, -0.0662], + [ 0.0306, 0.0514, -0.0765, ..., -0.0722, -0.0183, 0.0673], + [ 0.0511, -0.0604, -0.0674, ..., -0.0010, -0.0360, -0.0187], + ..., + [ 0.0485, 0.0281, 0.0722, ..., 0.0371, 0.1039, 0.0121], + [ 0.0576, -0.0864, -0.0753, ..., -0.1052, -0.0400, 0.0926], + [-0.0526, 0.0712, 0.0195, ..., -0.0495, -0.0394, 0.0055]], + device='cuda:0'), grad: tensor([[ 2.5835e-06, -2.2948e-04, 9.7603e-07, ..., 1.4529e-06, + 9.3691e-07, -5.4359e-05], + [ 4.9127e-07, 3.9041e-06, 3.2205e-06, ..., 2.8778e-06, + 1.4752e-06, -9.0823e-06], + [-4.5925e-05, -1.2718e-05, 1.2601e-06, ..., -3.4422e-05, + -1.8448e-05, 3.3267e-06], + ..., + [ 4.0978e-06, 1.7397e-06, -1.1154e-05, ..., 6.2324e-06, + -4.0047e-07, 3.8594e-06], + [ 2.6450e-06, 2.8700e-05, 1.5618e-06, ..., 7.6666e-06, + 3.2764e-06, -5.9940e-06], + [ 6.6571e-06, 1.9744e-05, 8.4415e-06, ..., 4.5523e-06, + 4.9062e-06, 9.8348e-06]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0078, -0.0087, -0.0052, -0.0137, -0.0126, -0.0050, 0.0100, 0.0053, + 0.0220, -0.0018], device='cuda:0'), grad: tensor([-3.7336e-04, 6.9067e-06, -1.7571e-04, 3.3140e-05, 2.7597e-05, + 3.4451e-05, 3.0375e-04, 2.4885e-05, 6.9499e-05, 4.8488e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 215.13, cls_loss 0.0069 cls_loss_mapping 0.0112 cls_loss_causal 0.5931 re_mapping 0.0111 re_causal 0.0319 /// teacc 98.79 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.1110, 0.0609, 0.0012, ..., -0.0070, -0.0467, -0.0677], + [ 0.0304, 0.0517, -0.0770, ..., -0.0725, -0.0183, 0.0678], + [ 0.0513, -0.0608, -0.0678, ..., -0.0010, -0.0362, -0.0190], + ..., + [ 0.0490, 0.0282, 0.0732, ..., 0.0376, 0.1049, 0.0122], + [ 0.0578, -0.0871, -0.0760, ..., -0.1062, -0.0404, 0.0931], + [-0.0529, 0.0712, 0.0191, ..., -0.0513, -0.0401, 0.0053]], + device='cuda:0'), grad: tensor([[ 6.6459e-06, 6.9151e-07, 3.9376e-06, ..., 2.9895e-06, + 1.4147e-06, 3.9637e-06], + [-3.6974e-06, -4.7944e-06, 1.1250e-06, ..., 1.3970e-06, + -1.5227e-06, -8.6203e-06], + [ 5.8621e-05, 3.1888e-06, 3.7640e-05, ..., 1.1116e-05, + -8.9183e-06, 5.1320e-05], + ..., + [ 4.4882e-05, -5.3504e-07, 1.5810e-05, ..., 2.2426e-05, + 1.6838e-05, 1.5289e-05], + [-1.4496e-04, 4.4294e-06, -7.9811e-05, ..., -4.8429e-05, + -1.8120e-05, -8.0109e-05], + [ 2.3674e-06, 1.0204e-04, 1.8075e-05, ..., 8.6203e-06, + 2.4904e-06, 1.2374e-04]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0082, -0.0083, -0.0053, -0.0131, -0.0128, -0.0052, 0.0101, 0.0057, + 0.0217, -0.0021], device='cuda:0'), grad: tensor([ 1.4171e-05, -1.2368e-05, 1.0520e-04, 6.8784e-05, -1.9395e-04, + -3.0071e-05, -1.6600e-05, 7.5579e-05, -2.1935e-04, 2.0790e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 215.02, cls_loss 0.0071 cls_loss_mapping 0.0099 cls_loss_causal 0.6009 re_mapping 0.0112 re_causal 0.0325 /// teacc 98.74 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.1122, 0.0605, 0.0008, ..., -0.0070, -0.0472, -0.0687], + [ 0.0300, 0.0521, -0.0774, ..., -0.0730, -0.0185, 0.0684], + [ 0.0514, -0.0611, -0.0679, ..., -0.0008, -0.0360, -0.0199], + ..., + [ 0.0497, 0.0279, 0.0733, ..., 0.0377, 0.1055, 0.0125], + [ 0.0580, -0.0878, -0.0760, ..., -0.1073, -0.0406, 0.0938], + [-0.0536, 0.0721, 0.0195, ..., -0.0518, -0.0403, 0.0048]], + device='cuda:0'), grad: tensor([[-1.0091e-04, -1.4603e-04, 6.1654e-06, ..., 2.0400e-05, + 6.4634e-07, -6.3241e-05], + [-6.7711e-05, -8.5115e-05, -5.1647e-05, ..., 1.5646e-05, + -6.6340e-05, -1.2910e-04], + [ 1.9744e-05, 3.2157e-05, 3.9369e-05, ..., 9.8169e-05, + 2.1264e-05, 1.0058e-05], + ..., + [-1.3225e-05, 1.1623e-05, -2.1383e-05, ..., -2.3291e-05, + -2.3827e-05, 1.9863e-05], + [ 1.4269e-04, 5.3740e-04, 3.9649e-04, ..., 6.0177e-04, + 3.9972e-06, 4.5776e-04], + [ 5.6028e-05, 9.5248e-05, 4.8161e-05, ..., 1.7807e-05, + 5.2005e-05, 1.0008e-04]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0089, -0.0081, -0.0054, -0.0130, -0.0135, -0.0049, 0.0104, 0.0059, + 0.0212, -0.0015], device='cuda:0'), grad: tensor([-3.4761e-04, -1.8466e-04, 1.4138e-04, -1.5907e-03, 3.0851e-04, + -9.8133e-04, 5.6458e-04, 3.3211e-06, 1.8568e-03, 2.3007e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 214.74, cls_loss 0.0096 cls_loss_mapping 0.0159 cls_loss_causal 0.6375 re_mapping 0.0112 re_causal 0.0314 /// teacc 98.64 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.1131, 0.0606, 0.0001, ..., -0.0071, -0.0489, -0.0697], + [ 0.0307, 0.0518, -0.0776, ..., -0.0732, -0.0181, 0.0690], + [ 0.0517, -0.0613, -0.0675, ..., -0.0002, -0.0365, -0.0200], + ..., + [ 0.0500, 0.0289, 0.0748, ..., 0.0378, 0.1071, 0.0119], + [ 0.0583, -0.0885, -0.0767, ..., -0.1080, -0.0408, 0.0946], + [-0.0546, 0.0717, 0.0185, ..., -0.0524, -0.0412, 0.0044]], + device='cuda:0'), grad: tensor([[ 4.6268e-06, 2.9970e-06, 6.0759e-06, ..., 3.5577e-06, + 6.2920e-06, 2.6431e-06], + [-2.1681e-06, 2.6360e-05, 9.0152e-06, ..., 7.6517e-06, + 7.5027e-06, -1.9427e-06], + [ 4.7356e-05, 2.1979e-05, 3.4362e-05, ..., 8.0988e-06, + 2.9147e-05, 1.3880e-05], + ..., + [-8.4400e-05, 2.0459e-05, -4.8041e-05, ..., -9.5069e-06, + -7.4983e-05, 2.2307e-05], + [ 1.5527e-05, 2.5302e-05, 2.4423e-05, ..., 4.7266e-05, + 6.2473e-06, 2.2456e-05], + [ 9.5814e-06, 6.6876e-05, 2.3529e-05, ..., 1.5691e-05, + 1.3337e-05, 4.9978e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0091, -0.0080, -0.0047, -0.0130, -0.0133, -0.0050, 0.0103, 0.0066, + 0.0209, -0.0025], device='cuda:0'), grad: tensor([ 1.9252e-05, 9.8288e-05, 9.6440e-05, -3.3736e-04, -5.9843e-04, + 3.7432e-04, 1.1325e-05, 8.4192e-06, 9.2387e-05, 2.3401e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 215.06, cls_loss 0.0063 cls_loss_mapping 0.0101 cls_loss_causal 0.5909 re_mapping 0.0110 re_causal 0.0315 /// teacc 98.71 lr 0.00010000 +Epoch 91, weight, value: tensor([[-1.1349e-01, 6.1119e-02, -8.3829e-05, ..., -6.9841e-03, + -4.9288e-02, -7.0300e-02], + [ 3.0311e-02, 5.1560e-02, -7.8058e-02, ..., -7.3620e-02, + -1.8435e-02, 6.8914e-02], + [ 5.1805e-02, -6.1402e-02, -6.8099e-02, ..., -2.2289e-04, + -3.6480e-02, -2.0132e-02], + ..., + [ 5.0269e-02, 2.8600e-02, 7.5160e-02, ..., 3.8152e-02, + 1.0761e-01, 1.1717e-02], + [ 5.8803e-02, -8.8817e-02, -7.6880e-02, ..., -1.0861e-01, + -4.0962e-02, 9.5204e-02], + [-5.4935e-02, 7.1068e-02, 1.8568e-02, ..., -5.2845e-02, + -4.1878e-02, 3.8089e-03]], device='cuda:0'), grad: tensor([[ 3.4869e-06, 1.1558e-06, 1.3234e-06, ..., 1.0999e-06, + 1.2405e-06, 1.8319e-06], + [ 1.8790e-05, -3.5372e-06, 1.9401e-05, ..., 2.7437e-06, + -2.6468e-06, -7.0751e-05], + [-3.4541e-05, -1.2420e-05, 9.8944e-06, ..., 3.2037e-07, + 2.0757e-05, 4.0591e-05], + ..., + [-4.0174e-05, 3.3323e-06, -4.1008e-05, ..., -8.3521e-06, + -4.3660e-05, 1.0036e-05], + [ 2.8595e-05, 1.5408e-05, 5.5544e-06, ..., 2.8573e-06, + 4.0159e-06, 5.6699e-06], + [ 1.2010e-05, 3.2842e-05, -3.6741e-07, ..., 3.8445e-06, + 4.5784e-06, 4.4554e-05]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0087, -0.0086, -0.0046, -0.0134, -0.0122, -0.0044, 0.0107, 0.0065, + 0.0207, -0.0034], device='cuda:0'), grad: tensor([ 4.4167e-05, -1.1420e-04, -4.1533e-04, 1.2651e-05, -6.1452e-05, + 1.1809e-05, 3.7905e-06, -3.2455e-05, 3.5310e-04, 1.9801e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 215.07, cls_loss 0.0068 cls_loss_mapping 0.0091 cls_loss_causal 0.6113 re_mapping 0.0107 re_causal 0.0312 /// teacc 98.74 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.1140, 0.0608, -0.0003, ..., -0.0070, -0.0498, -0.0708], + [ 0.0310, 0.0524, -0.0776, ..., -0.0743, -0.0181, 0.0703], + [ 0.0525, -0.0617, -0.0687, ..., -0.0002, -0.0361, -0.0207], + ..., + [ 0.0496, 0.0279, 0.0752, ..., 0.0386, 0.1075, 0.0105], + [ 0.0595, -0.0892, -0.0773, ..., -0.1091, -0.0412, 0.0962], + [-0.0553, 0.0717, 0.0191, ..., -0.0533, -0.0418, 0.0036]], + device='cuda:0'), grad: tensor([[ 8.6278e-06, -3.7644e-06, 5.3942e-06, ..., 2.9095e-06, + 9.5442e-06, 2.0163e-07], + [ 1.3247e-05, 1.8058e-06, 8.7246e-06, ..., 7.8827e-06, + 1.5378e-05, -5.2974e-06], + [-1.5616e-04, -1.7405e-05, -2.9355e-05, ..., -5.8323e-05, + -1.4293e-04, 1.7276e-06], + ..., + [ 3.8028e-05, 1.4892e-06, -1.7956e-05, ..., 1.9357e-05, + 2.3782e-05, 3.6461e-07], + [ 1.0878e-05, 2.0415e-06, 5.9567e-06, ..., 6.8769e-06, + 1.5408e-05, -2.8387e-06], + [ 2.2486e-05, 1.0312e-05, 2.0593e-05, ..., 6.3367e-06, + 1.9327e-05, 8.4564e-07]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0091, -0.0081, -0.0039, -0.0140, -0.0126, -0.0042, 0.0104, 0.0059, + 0.0213, -0.0030], device='cuda:0'), grad: tensor([ 1.4238e-05, 3.3885e-05, -3.7384e-04, 9.1970e-05, 1.2100e-05, + 2.2724e-05, 4.1574e-06, 9.9957e-05, 3.6895e-05, 5.7667e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 215.05, cls_loss 0.0056 cls_loss_mapping 0.0104 cls_loss_causal 0.5917 re_mapping 0.0102 re_causal 0.0309 /// teacc 98.78 lr 0.00010000 +Epoch 93, weight, value: tensor([[-1.1490e-01, 6.1446e-02, -4.2594e-04, ..., -7.0450e-03, + -5.0205e-02, -7.1270e-02], + [ 3.0594e-02, 5.2816e-02, -7.7803e-02, ..., -7.4437e-02, + -1.8102e-02, 7.0136e-02], + [ 5.3685e-02, -6.2480e-02, -6.8219e-02, ..., 9.7826e-05, + -3.5198e-02, -2.0515e-02], + ..., + [ 4.9382e-02, 2.7617e-02, 7.5143e-02, ..., 3.8585e-02, + 1.0728e-01, 1.0451e-02], + [ 5.9562e-02, -9.0459e-02, -7.7996e-02, ..., -1.0960e-01, + -4.1483e-02, 9.6639e-02], + [-5.5637e-02, 7.1233e-02, 1.9046e-02, ..., -5.3717e-02, + -4.2198e-02, 3.2040e-03]], device='cuda:0'), grad: tensor([[ 5.9232e-07, -2.3985e-04, 4.6939e-07, ..., 5.6345e-08, + 4.8243e-07, 5.3924e-07], + [ 6.1274e-05, 3.0994e-06, 5.2094e-05, ..., 1.4968e-05, + 6.3837e-05, 1.2718e-05], + [ 9.2015e-06, 5.9158e-06, 7.1935e-06, ..., 1.9930e-06, + 8.5086e-06, 2.4308e-06], + ..., + [-7.8797e-05, -2.2165e-06, -6.8724e-05, ..., -1.9431e-05, + -8.3387e-05, -1.6034e-05], + [-1.5693e-07, 1.8682e-06, 2.0862e-06, ..., 7.6182e-07, + 1.8161e-06, -1.5432e-06], + [ 5.8897e-06, 2.3949e-04, 4.0010e-06, ..., 1.5013e-06, + 5.6513e-06, 2.2966e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0088, -0.0086, -0.0030, -0.0142, -0.0123, -0.0036, 0.0105, 0.0057, + 0.0207, -0.0035], device='cuda:0'), grad: tensor([-3.8481e-04, 9.8169e-05, 3.5435e-05, 4.2804e-06, -2.6114e-06, + 3.6895e-05, -1.0693e-04, -1.1885e-04, 2.9609e-05, 4.0841e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 214.53, cls_loss 0.0078 cls_loss_mapping 0.0117 cls_loss_causal 0.6235 re_mapping 0.0104 re_causal 0.0310 /// teacc 98.74 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.1160, 0.0617, -0.0009, ..., -0.0073, -0.0511, -0.0719], + [ 0.0293, 0.0531, -0.0799, ..., -0.0763, -0.0199, 0.0703], + [ 0.0529, -0.0635, -0.0696, ..., -0.0010, -0.0363, -0.0213], + ..., + [ 0.0500, 0.0285, 0.0758, ..., 0.0403, 0.1089, 0.0103], + [ 0.0596, -0.0912, -0.0787, ..., -0.1102, -0.0418, 0.0967], + [-0.0538, 0.0704, 0.0199, ..., -0.0541, -0.0415, 0.0037]], + device='cuda:0'), grad: tensor([[ 9.0292e-07, -2.8968e-05, 1.3113e-05, ..., 1.2806e-07, + 9.2993e-07, 4.4331e-07], + [ 2.0210e-06, 1.2964e-06, 8.5607e-06, ..., 1.9427e-06, + 6.1952e-06, -6.5193e-06], + [ 1.2107e-08, 4.0084e-06, 1.0274e-05, ..., -1.5832e-06, + 1.4780e-06, 3.5111e-06], + ..., + [-1.7926e-05, 1.2830e-05, 4.1351e-06, ..., -8.6129e-06, + -2.5168e-05, -7.1898e-07], + [ 6.0396e-07, 5.7817e-06, 2.1551e-06, ..., 1.4268e-06, + 1.3588e-06, -1.2591e-06], + [ 5.7481e-06, -2.9966e-05, -7.6473e-05, ..., 2.3209e-06, + 7.6443e-06, 2.0433e-06]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0089, -0.0095, -0.0038, -0.0143, -0.0115, -0.0037, 0.0110, 0.0065, + 0.0203, -0.0034], device='cuda:0'), grad: tensor([-5.7101e-05, 3.6694e-06, 9.8571e-06, 4.3094e-05, 1.9148e-05, + -3.8773e-05, 7.8917e-05, 2.1815e-05, 2.0742e-05, -1.0163e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 214.42, cls_loss 0.0058 cls_loss_mapping 0.0085 cls_loss_causal 0.5990 re_mapping 0.0103 re_causal 0.0301 /// teacc 98.83 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.1166, 0.0623, -0.0008, ..., -0.0074, -0.0520, -0.0726], + [ 0.0292, 0.0532, -0.0804, ..., -0.0765, -0.0201, 0.0707], + [ 0.0528, -0.0639, -0.0698, ..., -0.0010, -0.0364, -0.0222], + ..., + [ 0.0502, 0.0282, 0.0762, ..., 0.0403, 0.1094, 0.0102], + [ 0.0599, -0.0913, -0.0791, ..., -0.1107, -0.0419, 0.0981], + [-0.0541, 0.0703, 0.0199, ..., -0.0546, -0.0419, 0.0038]], + device='cuda:0'), grad: tensor([[ 8.2375e-07, -2.7791e-05, 1.6829e-06, ..., 7.0687e-07, + 4.2170e-06, 4.9127e-07], + [-2.6450e-07, -5.4855e-07, 3.1367e-06, ..., 5.8347e-07, + 1.8561e-06, -8.2701e-06], + [-2.7474e-06, 4.6305e-06, 1.5497e-06, ..., -9.8422e-06, + -1.2405e-05, 3.1721e-06], + ..., + [ 2.1625e-04, 1.1575e-04, 4.8208e-04, ..., 1.0096e-06, + 3.1400e-04, 2.3264e-06], + [-1.0118e-05, 1.3500e-05, 9.8571e-06, ..., -2.4494e-06, + 2.7325e-06, -1.9580e-05], + [-2.2566e-04, -1.7285e-04, -5.5313e-04, ..., 2.8461e-06, + -3.2091e-04, -1.3039e-08]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0083, -0.0094, -0.0040, -0.0146, -0.0112, -0.0036, 0.0111, 0.0065, + 0.0207, -0.0040], device='cuda:0'), grad: tensor([-6.1750e-05, -3.3416e-06, -4.4376e-05, 3.5316e-05, 1.3089e-04, + 3.6657e-05, 9.8720e-06, 6.0940e-04, 1.4305e-05, -7.2718e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 214.70, cls_loss 0.0076 cls_loss_mapping 0.0116 cls_loss_causal 0.6043 re_mapping 0.0102 re_causal 0.0302 /// teacc 98.82 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.1179, 0.0638, -0.0005, ..., -0.0075, -0.0539, -0.0747], + [ 0.0296, 0.0535, -0.0802, ..., -0.0765, -0.0195, 0.0720], + [ 0.0526, -0.0646, -0.0704, ..., -0.0011, -0.0368, -0.0243], + ..., + [ 0.0502, 0.0280, 0.0758, ..., 0.0398, 0.1095, 0.0099], + [ 0.0613, -0.0939, -0.0807, ..., -0.1110, -0.0424, 0.0989], + [-0.0543, 0.0706, 0.0203, ..., -0.0551, -0.0420, 0.0043]], + device='cuda:0'), grad: tensor([[ 1.5059e-06, -7.7300e-08, 3.3341e-07, ..., 2.0023e-07, + 3.1758e-07, 5.4995e-07], + [-2.6133e-06, -8.7470e-06, 1.0552e-06, ..., 6.2259e-07, + -1.1243e-05, -2.6643e-05], + [ 5.0897e-07, 1.2890e-06, -4.6566e-09, ..., -2.6189e-06, + -7.1432e-07, 2.7679e-06], + ..., + [-1.1241e-06, 4.5262e-06, -4.5262e-06, ..., -1.1362e-07, + 1.8552e-06, 1.1541e-05], + [ 5.0008e-05, 2.5313e-06, 1.4203e-06, ..., 1.4445e-06, + 1.2647e-06, -2.2864e-07], + [ 3.4738e-06, 1.1928e-05, -1.1334e-06, ..., 1.3364e-06, + 8.6278e-06, 1.4663e-05]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0076, -0.0085, -0.0047, -0.0137, -0.0111, -0.0040, 0.0100, 0.0060, + 0.0202, -0.0036], device='cuda:0'), grad: tensor([ 4.0680e-05, -2.9728e-05, 5.3719e-06, 2.2352e-06, -1.2644e-05, + 1.3514e-03, -2.8839e-03, 1.6481e-05, 1.4477e-03, 6.2048e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 215.15, cls_loss 0.0073 cls_loss_mapping 0.0107 cls_loss_causal 0.6109 re_mapping 0.0099 re_causal 0.0303 /// teacc 98.79 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.1189, 0.0638, -0.0009, ..., -0.0075, -0.0549, -0.0754], + [ 0.0283, 0.0540, -0.0811, ..., -0.0768, -0.0200, 0.0716], + [ 0.0522, -0.0656, -0.0713, ..., -0.0015, -0.0376, -0.0255], + ..., + [ 0.0512, 0.0279, 0.0764, ..., 0.0403, 0.1107, 0.0105], + [ 0.0617, -0.0940, -0.0816, ..., -0.1114, -0.0429, 0.0999], + [-0.0538, 0.0704, 0.0205, ..., -0.0560, -0.0420, 0.0041]], + device='cuda:0'), grad: tensor([[ 9.4771e-06, 3.2652e-06, 5.3160e-06, ..., 1.1781e-06, + 3.2745e-06, 1.1742e-05], + [ 1.9029e-05, 4.8894e-07, 2.8193e-05, ..., 6.8136e-06, + 1.6585e-05, -9.9186e-07], + [ 1.2286e-05, 2.3302e-06, 1.7986e-05, ..., 2.4904e-06, + 7.5288e-06, 4.8429e-06], + ..., + [-1.1826e-03, 5.9716e-06, -1.7252e-03, ..., -4.0889e-04, + -1.0014e-03, -1.0067e-04], + [-1.0431e-07, -1.6734e-05, 4.5389e-05, ..., 1.0647e-05, + 2.6062e-05, -4.9978e-05], + [ 1.0767e-03, 8.3327e-05, 1.5478e-03, ..., 3.7026e-04, + 9.3031e-04, 1.5438e-04]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0076, -0.0092, -0.0055, -0.0137, -0.0111, -0.0035, 0.0102, 0.0065, + 0.0200, -0.0036], device='cuda:0'), grad: tensor([ 3.0264e-05, 3.5495e-05, 2.6271e-05, 9.1255e-05, -1.7846e-04, + 4.5538e-05, -2.5406e-05, -2.1400e-03, -2.5064e-05, 2.1420e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 96---------------------------------------------------- +epoch 96, time 231.73, cls_loss 0.0051 cls_loss_mapping 0.0064 cls_loss_causal 0.5691 re_mapping 0.0100 re_causal 0.0285 /// teacc 98.86 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.1199, 0.0634, -0.0016, ..., -0.0078, -0.0556, -0.0786], + [ 0.0286, 0.0539, -0.0808, ..., -0.0771, -0.0200, 0.0721], + [ 0.0517, -0.0661, -0.0724, ..., -0.0016, -0.0384, -0.0259], + ..., + [ 0.0517, 0.0281, 0.0772, ..., 0.0411, 0.1118, 0.0102], + [ 0.0622, -0.0936, -0.0817, ..., -0.1121, -0.0433, 0.1015], + [-0.0544, 0.0705, 0.0201, ..., -0.0568, -0.0427, 0.0033]], + device='cuda:0'), grad: tensor([[ 2.9989e-07, -8.0884e-05, 1.6764e-07, ..., -2.7955e-05, + 3.6601e-07, 9.1968e-07], + [ 4.7358e-07, 1.5600e-07, 3.7765e-07, ..., 2.9476e-07, + 3.5670e-07, -1.6084e-06], + [-1.7211e-06, 5.5693e-06, 3.9395e-07, ..., -6.9384e-07, + -1.0962e-06, 4.9099e-06], + ..., + [-6.9244e-07, 1.7444e-06, -1.7332e-06, ..., -2.9523e-07, + -1.6205e-06, 1.1493e-06], + [-1.2018e-05, 8.7321e-06, 2.3078e-06, ..., 2.9523e-06, + -4.3400e-06, -2.2024e-05], + [ 1.8002e-06, 8.3819e-06, -5.9791e-07, ..., 2.5406e-06, + 1.0636e-06, 1.1344e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0084, -0.0089, -0.0058, -0.0140, -0.0107, -0.0038, 0.0098, 0.0070, + 0.0208, -0.0039], device='cuda:0'), grad: tensor([-2.1243e-04, 1.3672e-06, 3.7514e-06, 3.1684e-06, 1.7643e-05, + 7.4744e-05, 8.4043e-05, 4.2655e-06, 2.5574e-06, 2.1055e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 214.75, cls_loss 0.0058 cls_loss_mapping 0.0100 cls_loss_causal 0.5821 re_mapping 0.0102 re_causal 0.0297 /// teacc 98.70 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.1207, 0.0631, -0.0021, ..., -0.0079, -0.0567, -0.0793], + [ 0.0291, 0.0548, -0.0806, ..., -0.0774, -0.0189, 0.0730], + [ 0.0516, -0.0668, -0.0725, ..., -0.0013, -0.0389, -0.0266], + ..., + [ 0.0520, 0.0277, 0.0777, ..., 0.0413, 0.1121, 0.0099], + [ 0.0624, -0.0941, -0.0823, ..., -0.1129, -0.0437, 0.1021], + [-0.0549, 0.0706, 0.0197, ..., -0.0577, -0.0433, 0.0028]], + device='cuda:0'), grad: tensor([[ 1.2189e-05, 3.3641e-04, 1.4007e-04, ..., 3.4310e-06, + 8.8736e-06, 1.3389e-05], + [ 1.9073e-03, 1.1425e-03, 8.6880e-04, ..., 5.7779e-06, + 1.6899e-03, 2.3632e-03], + [-2.6627e-03, -1.5144e-03, -1.1663e-03, ..., -7.5340e-05, + -2.2945e-03, -3.1700e-03], + ..., + [ 4.7421e-04, 3.0565e-04, 2.2030e-04, ..., 4.0792e-06, + 4.1890e-04, 5.9843e-04], + [ 1.0848e-04, 8.0764e-05, 5.8681e-05, ..., 1.6496e-05, + 8.5652e-05, 1.2600e-04], + [ 6.9976e-05, -2.3568e-04, -1.5450e-04, ..., 5.6177e-06, + 6.0201e-05, 2.5439e-04]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0088, -0.0080, -0.0058, -0.0137, -0.0108, -0.0039, 0.0098, 0.0071, + 0.0205, -0.0042], device='cuda:0'), grad: tensor([ 5.5790e-04, 3.7460e-03, -5.5008e-03, 4.9591e-04, -4.8041e-04, + -1.8552e-05, -4.2692e-06, 9.9277e-04, 3.2973e-04, -1.1545e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 214.44, cls_loss 0.0064 cls_loss_mapping 0.0094 cls_loss_causal 0.5831 re_mapping 0.0103 re_causal 0.0286 /// teacc 98.77 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.1219, 0.0630, -0.0026, ..., -0.0085, -0.0577, -0.0805], + [ 0.0266, 0.0540, -0.0828, ..., -0.0776, -0.0220, 0.0704], + [ 0.0535, -0.0665, -0.0719, ..., 0.0002, -0.0367, -0.0254], + ..., + [ 0.0528, 0.0281, 0.0784, ..., 0.0400, 0.1133, 0.0119], + [ 0.0627, -0.0944, -0.0826, ..., -0.1134, -0.0432, 0.1026], + [-0.0551, 0.0706, 0.0197, ..., -0.0584, -0.0436, 0.0023]], + device='cuda:0'), grad: tensor([[ 1.0230e-05, 2.5965e-06, 1.0975e-05, ..., 1.7853e-06, + 1.4745e-05, 1.0960e-05], + [-7.4530e-04, -2.6298e-04, 8.9705e-06, ..., 2.6803e-06, + -1.8573e-04, -1.6642e-03], + [ 3.5197e-05, 8.3745e-06, 2.8491e-05, ..., 2.1651e-05, + 4.3303e-05, 2.1189e-05], + ..., + [-5.2750e-05, -3.9162e-07, -4.9621e-05, ..., -5.0455e-05, + -1.3494e-04, 5.2661e-05], + [ 4.9305e-04, 1.8644e-04, 1.9908e-05, ..., 2.8946e-06, + 1.4031e-04, 1.0109e-03], + [ 2.1875e-04, 6.0230e-05, -1.5795e-04, ..., 1.6943e-05, + 1.3113e-04, 4.5395e-04]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0092, -0.0102, -0.0038, -0.0131, -0.0111, -0.0026, 0.0094, 0.0075, + 0.0196, -0.0042], device='cuda:0'), grad: tensor([ 2.4885e-05, -2.3937e-03, 7.7546e-05, 1.9157e-04, -5.2035e-05, + -3.8475e-05, 1.7929e-04, 1.4573e-05, 1.5125e-03, 4.8208e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 214.79, cls_loss 0.0052 cls_loss_mapping 0.0083 cls_loss_causal 0.6080 re_mapping 0.0100 re_causal 0.0307 /// teacc 98.85 lr 0.00010000 +Epoch 101, weight, value: tensor([[-1.2243e-01, 6.3021e-02, -2.6671e-03, ..., -8.7930e-03, + -5.8369e-02, -8.0959e-02], + [ 2.6563e-02, 5.4652e-02, -8.2717e-02, ..., -7.7451e-02, + -2.1994e-02, 7.0961e-02], + [ 5.3388e-02, -6.6882e-02, -7.2237e-02, ..., 1.0859e-04, + -3.6889e-02, -2.6186e-02], + ..., + [ 5.2747e-02, 2.7832e-02, 7.8331e-02, ..., 4.0245e-02, + 1.1347e-01, 1.1596e-02], + [ 6.3279e-02, -9.4487e-02, -8.2951e-02, ..., -1.1371e-01, + -4.3377e-02, 1.0349e-01], + [-5.4881e-02, 7.1502e-02, 2.1022e-02, ..., -5.9062e-02, + -4.2726e-02, 1.6109e-03]], device='cuda:0'), grad: tensor([[ 2.4419e-06, -1.2964e-06, 1.0589e-06, ..., -5.0478e-07, + 3.2736e-07, 1.9018e-06], + [ 8.2282e-07, 9.4529e-07, 2.0824e-06, ..., 3.2829e-07, + 7.3807e-07, -2.4568e-06], + [ 4.1611e-06, 3.2969e-06, 3.9004e-06, ..., 3.5856e-06, + 2.5351e-06, 1.6950e-06], + ..., + [ 1.8224e-05, 1.2979e-05, 4.4107e-05, ..., -9.4064e-07, + 2.5719e-05, 4.1723e-06], + [ 7.9349e-06, 9.7975e-06, 8.3596e-06, ..., 7.1758e-07, + 4.8168e-06, 3.5111e-06], + [ 3.0160e-05, -7.3388e-06, -1.2577e-04, ..., 3.8557e-07, + -4.6462e-05, 5.0902e-05]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0095, -0.0097, -0.0042, -0.0134, -0.0119, -0.0028, 0.0098, 0.0071, + 0.0197, -0.0031], device='cuda:0'), grad: tensor([ 2.8480e-06, 4.9137e-06, 1.8403e-05, 1.7136e-05, -3.8147e-04, + 2.9132e-06, 2.3693e-05, 7.8559e-05, 6.1512e-05, 1.7166e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.83, cls_loss 0.0051 cls_loss_mapping 0.0087 cls_loss_causal 0.5521 re_mapping 0.0103 re_causal 0.0288 /// teacc 98.68 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.1231, 0.0632, -0.0029, ..., -0.0090, -0.0592, -0.0813], + [ 0.0261, 0.0548, -0.0837, ..., -0.0789, -0.0227, 0.0711], + [ 0.0531, -0.0673, -0.0726, ..., 0.0002, -0.0373, -0.0264], + ..., + [ 0.0525, 0.0279, 0.0784, ..., 0.0408, 0.1140, 0.0108], + [ 0.0649, -0.0947, -0.0824, ..., -0.1130, -0.0424, 0.1051], + [-0.0543, 0.0713, 0.0216, ..., -0.0599, -0.0424, 0.0010]], + device='cuda:0'), grad: tensor([[ 3.5902e-07, -2.8759e-05, 4.2515e-07, ..., -2.7064e-06, + 3.6322e-07, 1.1818e-06], + [ 4.5076e-06, 6.1141e-07, 4.9435e-06, ..., 7.7672e-07, + 4.7386e-06, -1.0997e-04], + [-7.0315e-08, 3.8790e-07, -7.5018e-07, ..., -1.7807e-06, + -2.0806e-06, 2.3872e-05], + ..., + [-2.2203e-05, 2.2836e-06, -2.3365e-05, ..., -1.5721e-06, + -2.1830e-05, 2.0817e-05], + [ 3.9898e-06, 1.4296e-06, 6.9328e-06, ..., 1.2089e-06, + 5.8301e-06, 1.5900e-05], + [ 1.1519e-05, 1.8030e-05, 1.0058e-05, ..., 2.6170e-06, + 1.1496e-05, 2.4289e-05]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0093, -0.0100, -0.0046, -0.0124, -0.0116, -0.0037, 0.0096, 0.0069, + 0.0203, -0.0032], device='cuda:0'), grad: tensor([-4.1872e-05, -2.2173e-04, 4.4405e-05, 1.5736e-05, 2.3052e-05, + 1.6004e-05, 1.6853e-05, 4.1962e-05, 3.9011e-05, 6.6638e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 101---------------------------------------------------- +epoch 101, time 231.01, cls_loss 0.0065 cls_loss_mapping 0.0099 cls_loss_causal 0.6120 re_mapping 0.0098 re_causal 0.0283 /// teacc 98.91 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.1239, 0.0637, -0.0034, ..., -0.0095, -0.0605, -0.0808], + [ 0.0261, 0.0543, -0.0841, ..., -0.0791, -0.0225, 0.0712], + [ 0.0534, -0.0679, -0.0727, ..., 0.0010, -0.0373, -0.0262], + ..., + [ 0.0524, 0.0280, 0.0787, ..., 0.0410, 0.1141, 0.0105], + [ 0.0649, -0.0952, -0.0833, ..., -0.1148, -0.0429, 0.1054], + [-0.0538, 0.0710, 0.0217, ..., -0.0611, -0.0423, 0.0008]], + device='cuda:0'), grad: tensor([[ 5.6345e-07, -1.8135e-05, 9.8627e-07, ..., -1.1260e-06, + 4.0233e-07, 2.1048e-07], + [ 5.6066e-07, 0.0000e+00, 1.2061e-06, ..., 6.6776e-07, + 9.0152e-07, -9.5088e-07], + [ 6.3032e-06, 2.9206e-06, 6.4969e-05, ..., 7.5758e-05, + -1.4249e-07, 1.4538e-06], + ..., + [-9.5069e-06, 1.5590e-06, -6.3851e-06, ..., 1.4538e-06, + -1.1191e-05, -3.8557e-06], + [ 7.8157e-06, 6.4671e-06, 9.2536e-06, ..., 4.2580e-06, + 4.7982e-06, 2.4717e-06], + [ 8.7824e-07, -6.5677e-06, -1.5661e-05, ..., 3.3937e-06, + 2.8871e-06, 4.5635e-07]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0090, -0.0104, -0.0040, -0.0130, -0.0101, -0.0031, 0.0081, 0.0066, + 0.0197, -0.0032], device='cuda:0'), grad: tensor([-2.8223e-05, 1.9521e-06, 7.1585e-05, -7.5877e-05, 2.8074e-05, + -1.3679e-05, 9.1121e-06, -5.0776e-06, 3.5703e-05, -2.3678e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 214.92, cls_loss 0.0062 cls_loss_mapping 0.0104 cls_loss_causal 0.5684 re_mapping 0.0103 re_causal 0.0278 /// teacc 98.72 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.1247, 0.0646, -0.0036, ..., -0.0094, -0.0613, -0.0812], + [ 0.0258, 0.0539, -0.0848, ..., -0.0793, -0.0226, 0.0712], + [ 0.0537, -0.0682, -0.0729, ..., 0.0014, -0.0372, -0.0265], + ..., + [ 0.0526, 0.0279, 0.0791, ..., 0.0410, 0.1144, 0.0103], + [ 0.0649, -0.0956, -0.0839, ..., -0.1157, -0.0432, 0.1056], + [-0.0536, 0.0715, 0.0221, ..., -0.0616, -0.0425, 0.0016]], + device='cuda:0'), grad: tensor([[ 9.3505e-07, -5.8800e-05, 4.1388e-06, ..., -2.3082e-05, + 2.4643e-06, 7.6182e-07], + [ 3.0547e-06, -1.7779e-06, 4.3325e-06, ..., 2.4773e-06, + 2.4736e-06, -3.9376e-06], + [ 2.3007e-05, 3.4198e-06, 1.9521e-05, ..., 1.1355e-05, + 1.4558e-05, 9.7230e-06], + ..., + [-7.5877e-05, 2.1994e-05, -1.2314e-04, ..., -3.8803e-05, + -6.5684e-05, 6.9290e-06], + [-6.0797e-05, -3.3155e-07, 7.6070e-06, ..., -4.1611e-06, + -2.6271e-05, -5.8979e-05], + [ 6.7532e-05, 1.3605e-05, 4.7266e-05, ..., 2.9087e-05, + 4.2856e-05, 3.1114e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0082, -0.0106, -0.0037, -0.0133, -0.0104, -0.0034, 0.0082, 0.0064, + 0.0192, -0.0025], device='cuda:0'), grad: tensor([-1.0371e-04, 1.5814e-06, 5.0753e-05, 4.0054e-05, 1.9908e-05, + 2.6360e-05, 2.7478e-05, -1.0484e-04, -1.0526e-04, 1.4746e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 103---------------------------------------------------- +epoch 103, time 232.48, cls_loss 0.0041 cls_loss_mapping 0.0069 cls_loss_causal 0.5946 re_mapping 0.0098 re_causal 0.0282 /// teacc 98.99 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.1264, 0.0646, -0.0039, ..., -0.0097, -0.0619, -0.0820], + [ 0.0256, 0.0542, -0.0854, ..., -0.0799, -0.0227, 0.0715], + [ 0.0532, -0.0686, -0.0737, ..., 0.0012, -0.0375, -0.0275], + ..., + [ 0.0534, 0.0277, 0.0801, ..., 0.0418, 0.1151, 0.0105], + [ 0.0652, -0.0957, -0.0847, ..., -0.1163, -0.0437, 0.1061], + [-0.0540, 0.0713, 0.0221, ..., -0.0626, -0.0431, 0.0008]], + device='cuda:0'), grad: tensor([[ 8.6799e-07, -3.2336e-05, 6.5099e-07, ..., -1.3523e-06, + 8.4471e-07, 8.9314e-07], + [ 3.1274e-06, -7.8231e-07, 2.9225e-06, ..., 5.6215e-06, + 4.9360e-06, -4.7795e-06], + [-4.9174e-05, 2.4773e-06, -1.9804e-05, ..., -5.0485e-05, + -4.4137e-05, 2.3320e-06], + ..., + [ 3.2604e-05, 5.3272e-06, 1.3351e-05, ..., 3.3289e-05, + 3.0547e-05, 5.9195e-06], + [-5.5321e-06, 3.0845e-06, 1.7369e-06, ..., -4.2282e-06, + 2.6301e-06, -1.6287e-05], + [ 1.4352e-06, 7.5400e-05, 2.1860e-05, ..., 2.6692e-06, + 2.1473e-05, 5.4479e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0086, -0.0107, -0.0041, -0.0136, -0.0102, -0.0031, 0.0089, 0.0070, + 0.0190, -0.0031], device='cuda:0'), grad: tensor([-5.8353e-05, 4.6231e-06, -8.2135e-05, 4.6164e-05, -1.4043e-04, + 5.4091e-06, 3.7290e-06, 7.0035e-05, -1.4819e-05, 1.6582e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 214.64, cls_loss 0.0060 cls_loss_mapping 0.0097 cls_loss_causal 0.5615 re_mapping 0.0097 re_causal 0.0274 /// teacc 98.86 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.1281, 0.0645, -0.0045, ..., -0.0096, -0.0628, -0.0826], + [ 0.0260, 0.0549, -0.0852, ..., -0.0802, -0.0226, 0.0725], + [ 0.0529, -0.0690, -0.0742, ..., 0.0011, -0.0375, -0.0285], + ..., + [ 0.0538, 0.0274, 0.0806, ..., 0.0426, 0.1154, 0.0102], + [ 0.0655, -0.0961, -0.0854, ..., -0.1168, -0.0440, 0.1067], + [-0.0546, 0.0710, 0.0221, ..., -0.0642, -0.0437, -0.0003]], + device='cuda:0'), grad: tensor([[ 3.6180e-05, 3.4004e-05, 8.0094e-06, ..., 1.8477e-06, + 6.5342e-06, 8.8811e-05], + [ 3.5971e-05, 1.7273e-04, 1.8096e-04, ..., 3.4850e-06, + 1.3459e-04, 1.1940e-06], + [ 9.0674e-06, 1.4357e-05, 2.0787e-06, ..., -2.9624e-05, + -1.2919e-05, 3.6985e-05], + ..., + [-8.9407e-05, -4.0722e-04, -4.3941e-04, ..., -1.3085e-06, + -3.2282e-04, 5.3179e-07], + [-6.6698e-05, -4.6939e-05, 6.2175e-06, ..., 3.0380e-06, + 5.6662e-06, -1.7607e-04], + [ 4.5121e-05, 1.5008e-04, 1.4317e-04, ..., 3.6433e-06, + 1.0788e-04, 4.3780e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0087, -0.0101, -0.0047, -0.0137, -0.0095, -0.0032, 0.0095, 0.0071, + 0.0189, -0.0038], device='cuda:0'), grad: tensor([ 2.1243e-04, 4.1437e-04, 3.2365e-05, 1.4710e-04, 8.0764e-05, + -5.5507e-06, 8.8811e-06, -9.5701e-04, -3.5667e-04, 4.2272e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 214.34, cls_loss 0.0057 cls_loss_mapping 0.0074 cls_loss_causal 0.5831 re_mapping 0.0099 re_causal 0.0279 /// teacc 98.88 lr 0.00010000 +Epoch 107, weight, value: tensor([[-1.2887e-01, 6.4654e-02, -5.2609e-03, ..., -9.8035e-03, + -6.4007e-02, -8.3582e-02], + [ 2.5826e-02, 5.5312e-02, -8.5839e-02, ..., -8.1350e-02, + -2.2914e-02, 7.3066e-02], + [ 5.2679e-02, -7.0137e-02, -7.5388e-02, ..., 7.3491e-04, + -3.8034e-02, -2.8988e-02], + ..., + [ 5.3936e-02, 2.8049e-02, 8.1363e-02, ..., 4.3794e-02, + 1.1619e-01, 9.7921e-03], + [ 6.5762e-02, -9.6593e-02, -8.5960e-02, ..., -1.1750e-01, + -4.4540e-02, 1.0723e-01], + [-5.3784e-02, 7.1393e-02, 2.2733e-02, ..., -6.4972e-02, + -4.3043e-02, 9.7140e-05]], device='cuda:0'), grad: tensor([[ 1.4901e-07, -1.1064e-05, 2.1700e-07, ..., 3.7439e-07, + 2.4308e-07, -7.7710e-06], + [ 8.5682e-08, 1.4603e-06, 5.5321e-07, ..., 3.9954e-07, + 2.0303e-07, -2.0117e-06], + [-0.0000e+00, 3.7253e-06, 1.2228e-06, ..., 2.5332e-07, + -5.5600e-07, 2.0489e-06], + ..., + [ 4.9360e-07, 2.3283e-06, 4.2804e-06, ..., 4.8690e-06, + 9.8255e-07, 7.2084e-07], + [ 3.0175e-07, 7.0743e-06, 1.8133e-06, ..., 1.7295e-06, + 6.2305e-07, 4.1351e-06], + [ 4.7404e-07, 5.2433e-07, -3.2093e-06, ..., 5.5786e-07, + 4.3679e-07, 1.3923e-06]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0090, -0.0100, -0.0052, -0.0143, -0.0100, -0.0032, 0.0096, 0.0072, + 0.0188, -0.0029], device='cuda:0'), grad: tensor([-5.8621e-05, 8.2403e-06, 1.8075e-05, -7.2643e-07, -4.0263e-05, + -1.6302e-05, 2.5615e-05, 1.4529e-05, 3.9399e-05, 1.0058e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 214.56, cls_loss 0.0040 cls_loss_mapping 0.0072 cls_loss_causal 0.5961 re_mapping 0.0093 re_causal 0.0281 /// teacc 98.92 lr 0.00010000 +Epoch 108, weight, value: tensor([[-1.2956e-01, 6.5283e-02, -5.5275e-03, ..., -9.7910e-03, + -6.4728e-02, -8.3802e-02], + [ 2.5746e-02, 5.5225e-02, -8.5800e-02, ..., -8.1675e-02, + -2.2938e-02, 7.3292e-02], + [ 5.2822e-02, -7.0412e-02, -7.5613e-02, ..., 7.4171e-04, + -3.7971e-02, -2.9272e-02], + ..., + [ 5.3924e-02, 2.8354e-02, 8.1749e-02, ..., 4.4249e-02, + 1.1659e-01, 9.5030e-03], + [ 6.5823e-02, -9.6769e-02, -8.6358e-02, ..., -1.1824e-01, + -4.4775e-02, 1.0756e-01], + [-5.3801e-02, 7.1343e-02, 2.2677e-02, ..., -6.5766e-02, + -4.3392e-02, 3.6780e-05]], device='cuda:0'), grad: tensor([[ 1.5330e-06, -1.8813e-07, 5.0664e-07, ..., 1.0030e-06, + 6.2771e-07, 1.6773e-06], + [ 1.2271e-05, 6.7614e-06, 1.5348e-05, ..., 1.1116e-05, + 2.3037e-05, -2.2724e-05], + [ 2.4866e-07, -4.3064e-05, -1.0669e-05, ..., -6.7353e-05, + -6.1691e-06, 1.4454e-05], + ..., + [-4.6581e-05, -7.9572e-06, -3.6806e-05, ..., -2.3901e-05, + -5.3585e-05, 3.5111e-07], + [-2.3156e-05, -1.0855e-05, 1.8142e-06, ..., 2.4103e-06, + 2.2091e-06, -3.9369e-05], + [ 1.5169e-05, 3.2991e-05, 1.1638e-05, ..., 3.9995e-05, + 1.3225e-05, 8.0988e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0083, -0.0100, -0.0054, -0.0143, -0.0100, -0.0030, 0.0095, 0.0071, + 0.0184, -0.0030], device='cuda:0'), grad: tensor([ 4.6939e-06, 2.8852e-06, -1.4842e-04, 1.1772e-04, -1.1683e-05, + 3.9816e-05, -2.3339e-06, -5.1707e-05, -7.8082e-05, 1.2743e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 214.22, cls_loss 0.0045 cls_loss_mapping 0.0080 cls_loss_causal 0.5918 re_mapping 0.0097 re_causal 0.0281 /// teacc 98.86 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.1307, 0.0654, -0.0058, ..., -0.0100, -0.0654, -0.0842], + [ 0.0260, 0.0559, -0.0860, ..., -0.0816, -0.0227, 0.0738], + [ 0.0528, -0.0709, -0.0757, ..., 0.0013, -0.0381, -0.0298], + ..., + [ 0.0542, 0.0279, 0.0819, ..., 0.0442, 0.1168, 0.0094], + [ 0.0655, -0.0973, -0.0870, ..., -0.1190, -0.0451, 0.1078], + [-0.0544, 0.0714, 0.0225, ..., -0.0673, -0.0439, -0.0008]], + device='cuda:0'), grad: tensor([[ 2.2855e-06, 6.4075e-07, 2.5686e-06, ..., 4.8801e-07, + 1.4817e-06, 2.3097e-07], + [ 2.3339e-06, -1.3597e-07, 2.8852e-06, ..., 1.1204e-06, + 1.7742e-06, -1.7490e-06], + [ 9.5129e-05, 1.4752e-06, 7.7546e-05, ..., 2.3752e-05, + 6.6757e-05, 1.3839e-06], + ..., + [-2.6846e-04, -1.1012e-05, -2.6035e-04, ..., -1.2040e-04, + -2.2101e-04, -4.0531e-06], + [ 4.6521e-05, 7.6964e-06, 4.2558e-05, ..., 2.0042e-05, + 2.3887e-05, 1.0040e-06], + [ 4.2528e-05, 7.4916e-06, 3.4183e-05, ..., 1.5870e-05, + 6.6698e-05, 8.3968e-06]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0083, -0.0094, -0.0054, -0.0142, -0.0097, -0.0027, 0.0100, 0.0071, + 0.0175, -0.0034], device='cuda:0'), grad: tensor([ 5.6550e-06, 2.6524e-06, 9.1732e-05, 9.1910e-05, 6.4820e-06, + -7.7665e-05, 4.5478e-05, -3.2043e-04, 9.2149e-05, 6.1929e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 214.61, cls_loss 0.0066 cls_loss_mapping 0.0074 cls_loss_causal 0.5725 re_mapping 0.0096 re_causal 0.0274 /// teacc 98.72 lr 0.00010000 +Epoch 110, weight, value: tensor([[-1.3215e-01, 6.4564e-02, -6.5148e-03, ..., -1.1245e-02, + -6.6529e-02, -8.5058e-02], + [ 2.5540e-02, 5.4919e-02, -8.6100e-02, ..., -8.2607e-02, + -2.3124e-02, 7.3633e-02], + [ 5.2793e-02, -7.0152e-02, -7.7243e-02, ..., 4.1386e-05, + -3.7448e-02, -2.8473e-02], + ..., + [ 5.4851e-02, 2.8397e-02, 8.3007e-02, ..., 4.4650e-02, + 1.1775e-01, 9.2202e-03], + [ 6.6402e-02, -9.7765e-02, -8.7675e-02, ..., -1.1985e-01, + -4.5636e-02, 1.0880e-01], + [-5.5372e-02, 7.1734e-02, 2.2360e-02, ..., -6.9406e-02, + -4.5352e-02, -1.5588e-03]], device='cuda:0'), grad: tensor([[ 1.4342e-07, 1.9372e-05, 1.3970e-07, ..., -1.4510e-06, + 1.3970e-07, 3.1199e-07], + [ 8.0187e-07, 1.8151e-06, 2.7549e-06, ..., 1.8803e-06, + 1.3104e-06, -3.1199e-07], + [ 3.4831e-07, 7.8324e-07, 2.4494e-07, ..., -1.0030e-06, + -1.0217e-06, 6.0257e-07], + ..., + [-3.6117e-06, 9.2015e-07, -2.6450e-07, ..., 2.4252e-06, + -4.2245e-06, 5.7183e-07], + [-1.1541e-05, 1.0552e-06, 3.1050e-06, ..., 2.4587e-06, + -4.2655e-06, -1.5870e-05], + [ 1.1988e-05, 1.2502e-05, 1.0289e-05, ..., 7.7263e-06, + 8.8587e-06, 1.4290e-05]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0094, -0.0106, -0.0052, -0.0144, -0.0100, -0.0015, 0.0104, 0.0078, + 0.0178, -0.0038], device='cuda:0'), grad: tensor([ 6.9141e-05, 6.2808e-06, 1.4603e-06, -3.7789e-05, 4.2439e-05, + 3.2395e-05, -1.4460e-04, 1.9027e-06, -2.1368e-05, 5.0336e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 214.36, cls_loss 0.0054 cls_loss_mapping 0.0080 cls_loss_causal 0.5582 re_mapping 0.0092 re_causal 0.0266 /// teacc 98.76 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.1331, 0.0642, -0.0069, ..., -0.0106, -0.0687, -0.0857], + [ 0.0256, 0.0546, -0.0864, ..., -0.0826, -0.0229, 0.0739], + [ 0.0528, -0.0707, -0.0776, ..., -0.0003, -0.0375, -0.0287], + ..., + [ 0.0550, 0.0278, 0.0831, ..., 0.0448, 0.1179, 0.0091], + [ 0.0665, -0.0984, -0.0883, ..., -0.1208, -0.0460, 0.1092], + [-0.0555, 0.0719, 0.0226, ..., -0.0703, -0.0458, -0.0023]], + device='cuda:0'), grad: tensor([[ 2.3842e-07, -8.9593e-07, 5.0943e-07, ..., 3.8836e-07, + 5.7090e-07, 1.7043e-07], + [-1.2508e-06, -4.4033e-06, 4.1071e-07, ..., 3.9395e-07, + -4.2878e-06, -1.0870e-05], + [ 2.3395e-06, 3.4943e-06, 1.6978e-06, ..., 1.2107e-06, + 5.3942e-06, 8.6799e-06], + ..., + [-7.7248e-05, -4.1366e-05, -1.8632e-04, ..., -3.6240e-05, + -1.5855e-04, 1.1921e-06], + [-2.6729e-06, -1.9632e-06, 2.5891e-07, ..., -5.1875e-07, + 3.6508e-07, -5.1260e-06], + [ 6.6400e-05, 4.1872e-05, 1.6725e-04, ..., 2.1443e-05, + 1.3733e-04, 8.5682e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0100, -0.0107, -0.0054, -0.0144, -0.0099, -0.0013, 0.0105, 0.0084, + 0.0174, -0.0039], device='cuda:0'), grad: tensor([ 2.4866e-07, -1.6987e-05, 1.6823e-05, 1.7434e-05, 3.0696e-06, + 8.1956e-06, -2.9206e-06, -2.2268e-04, -6.3032e-06, 2.0289e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 214.95, cls_loss 0.0049 cls_loss_mapping 0.0083 cls_loss_causal 0.5674 re_mapping 0.0097 re_causal 0.0276 /// teacc 98.90 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.1340, 0.0644, -0.0079, ..., -0.0113, -0.0694, -0.0863], + [ 0.0256, 0.0552, -0.0867, ..., -0.0828, -0.0227, 0.0743], + [ 0.0535, -0.0712, -0.0779, ..., 0.0002, -0.0372, -0.0288], + ..., + [ 0.0548, 0.0276, 0.0833, ..., 0.0447, 0.1179, 0.0088], + [ 0.0667, -0.0986, -0.0888, ..., -0.1218, -0.0464, 0.1100], + [-0.0555, 0.0716, 0.0228, ..., -0.0716, -0.0458, -0.0027]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, -4.0643e-06, 1.5926e-07, ..., -1.5013e-06, + 5.1223e-08, 2.6077e-08], + [-6.2399e-08, -4.9360e-08, 5.4110e-07, ..., 7.3016e-07, + 6.7055e-08, -9.7603e-07], + [ 9.4064e-08, 9.6206e-07, 9.2108e-07, ..., -3.9116e-08, + -3.4552e-07, 2.0210e-07], + ..., + [-1.8720e-07, 6.0536e-07, 2.6915e-07, ..., 9.4343e-07, + -3.3248e-07, 2.7660e-07], + [ 5.4017e-08, 1.1669e-06, 5.7463e-07, ..., 1.0263e-06, + 3.5390e-08, 4.9360e-08], + [ 3.0827e-07, 2.5164e-06, 9.6206e-07, ..., 1.3663e-06, + 1.7509e-07, 2.2445e-07]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0099, -0.0103, -0.0052, -0.0138, -0.0097, -0.0018, 0.0111, 0.0081, + 0.0174, -0.0043], device='cuda:0'), grad: tensor([-4.2096e-06, 3.6228e-07, 2.2389e-06, -5.0217e-06, 1.6168e-06, + 2.6435e-05, -4.1962e-05, 2.2165e-06, 9.4846e-06, 8.7619e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 214.83, cls_loss 0.0041 cls_loss_mapping 0.0068 cls_loss_causal 0.5715 re_mapping 0.0088 re_causal 0.0273 /// teacc 98.81 lr 0.00010000 +Epoch 113, weight, value: tensor([[-1.3488e-01, 6.4776e-02, -8.3121e-03, ..., -1.0928e-02, + -7.0167e-02, -8.6715e-02], + [ 2.5152e-02, 5.5399e-02, -8.6895e-02, ..., -8.3144e-02, + -2.2846e-02, 7.4103e-02], + [ 5.3771e-02, -7.1466e-02, -7.8507e-02, ..., 7.1163e-05, + -3.7105e-02, -2.8292e-02], + ..., + [ 5.4838e-02, 2.7564e-02, 8.3689e-02, ..., 4.5115e-02, + 1.1820e-01, 8.5024e-03], + [ 6.7641e-02, -9.9024e-02, -8.9214e-02, ..., -1.2225e-01, + -4.6433e-02, 1.1102e-01], + [-5.5626e-02, 7.1637e-02, 2.3037e-02, ..., -7.2606e-02, + -4.5843e-02, -2.9646e-03]], device='cuda:0'), grad: tensor([[ 7.2550e-07, -1.4079e-04, 4.3865e-07, ..., 6.2585e-07, + 5.6345e-07, 2.7604e-06], + [ 2.3559e-05, 4.5914e-07, 1.0245e-06, ..., 5.5321e-07, + 2.5973e-05, -3.2485e-06], + [-3.4571e-05, 1.4063e-06, 9.5367e-07, ..., -1.5637e-06, + -3.5763e-05, 1.2852e-07], + ..., + [-4.6603e-06, 8.6892e-07, -1.3702e-05, ..., -5.5395e-06, + -5.3607e-06, 9.7975e-07], + [ 1.5022e-06, 4.0047e-06, 1.1213e-06, ..., 6.9663e-07, + 1.3039e-06, 1.0710e-06], + [ 5.8040e-06, 1.4222e-04, 1.0118e-05, ..., 4.2282e-06, + 6.3106e-06, 1.5963e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0096, -0.0109, -0.0047, -0.0134, -0.0101, -0.0019, 0.0114, 0.0079, + 0.0177, -0.0044], device='cuda:0'), grad: tensor([-1.9121e-04, 9.1374e-05, -1.3173e-04, 1.6719e-05, -4.0144e-05, + -5.8766e-07, -1.7717e-05, 1.6158e-06, 1.9297e-05, 2.5177e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 214.88, cls_loss 0.0063 cls_loss_mapping 0.0076 cls_loss_causal 0.5927 re_mapping 0.0093 re_causal 0.0269 /// teacc 98.89 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.1357, 0.0648, -0.0085, ..., -0.0112, -0.0713, -0.0869], + [ 0.0282, 0.0548, -0.0873, ..., -0.0790, -0.0197, 0.0743], + [ 0.0503, -0.0712, -0.0791, ..., -0.0021, -0.0407, -0.0287], + ..., + [ 0.0551, 0.0275, 0.0842, ..., 0.0454, 0.1185, 0.0085], + [ 0.0681, -0.0994, -0.0895, ..., -0.1229, -0.0468, 0.1118], + [-0.0560, 0.0717, 0.0228, ..., -0.0736, -0.0461, -0.0034]], + device='cuda:0'), grad: tensor([[ 5.2340e-07, -7.4841e-06, 2.4494e-06, ..., 1.1055e-06, + 2.2817e-07, 5.0850e-07], + [ 2.2221e-06, -1.3784e-07, 1.7164e-06, ..., 1.2154e-06, + 2.7902e-06, -4.8149e-07], + [-4.4145e-06, 9.8255e-07, 4.7050e-06, ..., 7.2643e-08, + -6.9253e-06, 7.3574e-07], + ..., + [ 3.9116e-07, 1.5516e-06, 1.0513e-05, ..., 5.5134e-06, + 9.1083e-07, 7.3481e-07], + [-7.1079e-06, 4.0568e-06, 2.7776e-05, ..., 1.2122e-05, + 5.1316e-07, -1.0759e-05], + [ 2.9132e-06, 2.5824e-05, 2.6631e-04, ..., 1.1253e-04, + 8.9779e-07, 4.4703e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0099, -0.0083, -0.0072, -0.0139, -0.0099, -0.0016, 0.0113, 0.0080, + 0.0177, -0.0045], device='cuda:0'), grad: tensor([-1.3210e-05, 5.0589e-06, -3.7663e-06, -3.5930e-04, 1.0822e-06, + 2.5481e-05, 6.7316e-06, 1.6645e-05, 2.0415e-05, 3.0112e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 214.89, cls_loss 0.0062 cls_loss_mapping 0.0073 cls_loss_causal 0.5676 re_mapping 0.0088 re_causal 0.0270 /// teacc 98.84 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.1388, 0.0652, -0.0088, ..., -0.0113, -0.0721, -0.0897], + [ 0.0309, 0.0545, -0.0859, ..., -0.0748, -0.0177, 0.0764], + [ 0.0474, -0.0710, -0.0815, ..., -0.0054, -0.0428, -0.0317], + ..., + [ 0.0555, 0.0274, 0.0848, ..., 0.0456, 0.1191, 0.0084], + [ 0.0692, -0.0990, -0.0898, ..., -0.1235, -0.0473, 0.1134], + [-0.0564, 0.0719, 0.0227, ..., -0.0743, -0.0465, -0.0037]], + device='cuda:0'), grad: tensor([[ 2.7474e-07, 6.9197e-07, 7.6741e-07, ..., 1.6112e-07, + 3.1292e-07, 1.5926e-07], + [ 3.8370e-07, 3.8091e-07, 6.5099e-07, ..., 4.8243e-07, + 5.1130e-07, 2.2817e-07], + [ 1.3039e-08, 2.9150e-07, 1.8906e-07, ..., -6.6124e-07, + -6.3330e-08, 3.0920e-07], + ..., + [-7.0110e-06, 5.0701e-06, -1.2718e-05, ..., -2.0377e-06, + -1.3858e-05, 6.4224e-06], + [ 7.5903e-07, 1.8738e-06, 2.0266e-06, ..., 1.4799e-06, + 5.5693e-07, 6.4448e-07], + [ 4.8615e-06, -1.2584e-05, -4.8839e-06, ..., 2.5388e-06, + 1.1258e-05, 1.5730e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0101, -0.0056, -0.0097, -0.0145, -0.0099, -0.0016, 0.0108, 0.0082, + 0.0185, -0.0046], device='cuda:0'), grad: tensor([ 1.7257e-06, 1.6615e-06, -8.3353e-07, -2.1420e-07, 3.6173e-06, + 2.4997e-06, 1.2564e-06, -8.1062e-06, 6.3628e-06, -7.9349e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 214.63, cls_loss 0.0044 cls_loss_mapping 0.0074 cls_loss_causal 0.5560 re_mapping 0.0085 re_causal 0.0258 /// teacc 98.91 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.1393, 0.0652, -0.0093, ..., -0.0115, -0.0740, -0.0900], + [ 0.0307, 0.0549, -0.0867, ..., -0.0751, -0.0182, 0.0768], + [ 0.0472, -0.0714, -0.0818, ..., -0.0058, -0.0434, -0.0318], + ..., + [ 0.0569, 0.0274, 0.0856, ..., 0.0472, 0.1210, 0.0085], + [ 0.0699, -0.0991, -0.0907, ..., -0.1244, -0.0480, 0.1144], + [-0.0570, 0.0720, 0.0227, ..., -0.0760, -0.0470, -0.0043]], + device='cuda:0'), grad: tensor([[ 5.3197e-06, -2.9316e-03, 7.3984e-06, ..., -5.3365e-07, + -4.9025e-05, 2.6524e-06], + [ 6.6280e-05, 4.8280e-06, 8.2627e-06, ..., 5.5917e-06, + 1.4931e-05, 1.1736e-04], + [-1.5900e-05, 1.0319e-05, 6.5826e-06, ..., -9.0003e-06, + -3.9846e-05, 7.5847e-06], + ..., + [-2.0221e-05, 5.3905e-06, -2.0266e-06, ..., -1.7896e-05, + -2.7701e-05, -2.2128e-06], + [-9.8765e-05, 2.9728e-06, 1.0937e-05, ..., -3.2037e-06, + 3.5465e-06, -2.1231e-04], + [ 3.9488e-05, 2.8419e-03, 4.0114e-05, ..., 2.1458e-05, + 7.9036e-05, 3.0428e-05]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0103, -0.0058, -0.0098, -0.0142, -0.0101, -0.0020, 0.0106, 0.0092, + 0.0187, -0.0050], device='cuda:0'), grad: tensor([-4.9896e-03, 1.9252e-04, 4.6879e-05, 3.2544e-04, 1.2362e-04, + -1.3437e-03, 7.3016e-05, 4.9973e-04, -8.3029e-05, 5.1575e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 214.72, cls_loss 0.0058 cls_loss_mapping 0.0062 cls_loss_causal 0.5531 re_mapping 0.0089 re_causal 0.0248 /// teacc 98.77 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.1403, 0.0662, -0.0098, ..., -0.0116, -0.0759, -0.0903], + [ 0.0309, 0.0538, -0.0868, ..., -0.0749, -0.0179, 0.0771], + [ 0.0469, -0.0701, -0.0824, ..., -0.0059, -0.0438, -0.0324], + ..., + [ 0.0575, 0.0271, 0.0860, ..., 0.0473, 0.1216, 0.0088], + [ 0.0702, -0.0996, -0.0917, ..., -0.1253, -0.0484, 0.1150], + [-0.0572, 0.0724, 0.0233, ..., -0.0767, -0.0476, -0.0036]], + device='cuda:0'), grad: tensor([[ 6.5286e-07, -1.6436e-05, 5.1223e-07, ..., 2.4028e-07, + 4.0606e-07, -1.7695e-08], + [ 1.5378e-05, 1.0449e-06, 1.0647e-05, ..., 5.6960e-06, + 7.5884e-06, 4.0606e-06], + [ 6.6012e-06, 1.2582e-06, 5.9530e-06, ..., -4.2468e-07, + 1.5823e-06, 1.6866e-06], + ..., + [-1.6421e-05, 1.8151e-06, -2.4468e-05, ..., -1.0036e-05, + -1.6600e-05, 1.2092e-05], + [-2.4080e-05, 2.3898e-06, 2.2780e-06, ..., 1.3085e-06, + 1.1632e-06, -2.9862e-05], + [ 3.5372e-06, 1.7043e-07, -1.9185e-06, ..., 8.8383e-07, + 1.6028e-06, 2.0899e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0096, -0.0059, -0.0094, -0.0146, -0.0113, -0.0016, 0.0100, 0.0093, + 0.0184, -0.0046], device='cuda:0'), grad: tensor([-4.1783e-05, 2.7061e-05, 7.5176e-06, 7.4804e-06, 7.3835e-06, + 1.2204e-05, 2.0504e-05, -2.0280e-05, -2.5511e-05, 5.2787e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 214.80, cls_loss 0.0048 cls_loss_mapping 0.0063 cls_loss_causal 0.5858 re_mapping 0.0089 re_causal 0.0266 /// teacc 98.90 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.1412, 0.0664, -0.0105, ..., -0.0127, -0.0768, -0.0908], + [ 0.0310, 0.0554, -0.0864, ..., -0.0750, -0.0179, 0.0780], + [ 0.0471, -0.0702, -0.0823, ..., -0.0057, -0.0435, -0.0326], + ..., + [ 0.0577, 0.0255, 0.0862, ..., 0.0472, 0.1217, 0.0081], + [ 0.0710, -0.0998, -0.0927, ..., -0.1261, -0.0491, 0.1160], + [-0.0578, 0.0715, 0.0228, ..., -0.0774, -0.0481, -0.0052]], + device='cuda:0'), grad: tensor([[ 3.8296e-06, -3.2969e-07, 1.7723e-06, ..., 1.5832e-07, + 4.0680e-06, 1.6391e-06], + [ 4.1466e-03, 1.0179e-06, 1.8196e-03, ..., 4.6901e-06, + 4.3755e-03, 1.6356e-03], + [ 9.2015e-06, 3.9004e-06, 1.6987e-05, ..., -5.7444e-06, + 1.2912e-05, 2.0012e-05], + ..., + [-4.2801e-03, 9.1493e-06, -1.8911e-03, ..., 5.7649e-07, + -4.5166e-03, -1.6918e-03], + [ 2.0400e-05, 7.1339e-06, 9.4995e-06, ..., 1.3532e-06, + 2.0012e-05, 1.0215e-05], + [ 1.4693e-05, 1.2852e-05, 4.9435e-06, ..., 4.1164e-07, + 1.6332e-05, 2.3976e-05]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0099, -0.0056, -0.0092, -0.0140, -0.0110, -0.0020, 0.0096, 0.0094, + 0.0186, -0.0054], device='cuda:0'), grad: tensor([ 6.7502e-06, 7.0381e-03, 3.3289e-05, 1.5926e-04, -7.4208e-05, + -2.8566e-05, 4.6100e-07, -7.2556e-03, 5.0962e-05, 6.8367e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 214.99, cls_loss 0.0034 cls_loss_mapping 0.0078 cls_loss_causal 0.5768 re_mapping 0.0091 re_causal 0.0257 /// teacc 98.88 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.1416, 0.0666, -0.0108, ..., -0.0129, -0.0771, -0.0911], + [ 0.0304, 0.0554, -0.0874, ..., -0.0751, -0.0184, 0.0777], + [ 0.0475, -0.0702, -0.0819, ..., -0.0053, -0.0431, -0.0327], + ..., + [ 0.0582, 0.0251, 0.0866, ..., 0.0471, 0.1222, 0.0084], + [ 0.0719, -0.0999, -0.0931, ..., -0.1266, -0.0495, 0.1175], + [-0.0580, 0.0712, 0.0231, ..., -0.0778, -0.0482, -0.0058]], + device='cuda:0'), grad: tensor([[ 9.2834e-06, 4.1388e-06, 2.2035e-06, ..., 6.9849e-07, + 1.4771e-06, 1.2524e-05], + [-1.2422e-04, -1.1867e-04, 1.7006e-06, ..., 1.7341e-06, + -4.6521e-05, -2.1970e-04], + [ 4.9993e-06, 3.1777e-06, -7.1432e-07, ..., -6.0722e-07, + -5.4762e-06, 4.7088e-05], + ..., + [ 2.1726e-05, 1.7092e-05, -1.2917e-06, ..., -7.9628e-07, + 8.7544e-06, 5.2631e-05], + [-2.1309e-05, 7.1347e-05, 1.1176e-06, ..., -4.3474e-06, + 3.2514e-05, 5.0038e-05], + [ 1.5721e-05, 8.6799e-06, -6.2846e-06, ..., -1.8254e-07, + 4.9882e-06, 3.5375e-05]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0099, -0.0062, -0.0088, -0.0144, -0.0107, -0.0018, 0.0097, 0.0096, + 0.0193, -0.0058], device='cuda:0'), grad: tensor([ 3.1412e-05, -3.4285e-04, 1.8847e-04, 9.9838e-05, -4.0960e-04, + 2.8938e-05, 1.1617e-04, 1.5664e-04, 4.6968e-05, 8.3923e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 214.55, cls_loss 0.0035 cls_loss_mapping 0.0060 cls_loss_causal 0.5798 re_mapping 0.0087 re_causal 0.0265 /// teacc 98.91 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.1421, 0.0665, -0.0110, ..., -0.0133, -0.0781, -0.0915], + [ 0.0302, 0.0557, -0.0878, ..., -0.0752, -0.0188, 0.0779], + [ 0.0486, -0.0703, -0.0809, ..., -0.0049, -0.0422, -0.0321], + ..., + [ 0.0580, 0.0251, 0.0867, ..., 0.0471, 0.1223, 0.0084], + [ 0.0708, -0.1003, -0.0934, ..., -0.1270, -0.0503, 0.1172], + [-0.0589, 0.0711, 0.0231, ..., -0.0786, -0.0488, -0.0063]], + device='cuda:0'), grad: tensor([[ 1.1986e-06, 3.4608e-06, 4.4852e-06, ..., 2.3469e-07, + 5.0571e-07, 7.2233e-06], + [-3.1203e-05, 3.0136e-04, 3.5310e-04, ..., 1.7006e-06, + -2.2613e-06, 3.8171e-04], + [ 6.6347e-06, 1.4022e-05, 1.4544e-05, ..., -1.5814e-06, + -2.8759e-06, 2.7448e-05], + ..., + [ 1.5739e-07, 1.4372e-05, 6.9179e-06, ..., -2.5854e-06, + -2.6207e-06, 2.2456e-05], + [ 1.7196e-05, 5.1796e-05, 2.6733e-05, ..., -5.2713e-07, + 3.9227e-06, 7.3612e-05], + [ 3.0790e-06, -9.1457e-04, -9.6369e-04, ..., 6.8359e-07, + 1.0375e-06, -1.2274e-03]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0100, -0.0065, -0.0078, -0.0145, -0.0110, -0.0016, 0.0096, 0.0096, + 0.0183, -0.0061], device='cuda:0'), grad: tensor([ 1.6421e-05, 1.1120e-03, 5.6505e-05, 1.6394e-03, 4.1455e-05, + 2.1982e-04, 2.4438e-06, 4.4674e-05, 1.5497e-04, -3.2864e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 214.90, cls_loss 0.0036 cls_loss_mapping 0.0054 cls_loss_causal 0.5874 re_mapping 0.0083 re_causal 0.0261 /// teacc 98.84 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.1424, 0.0664, -0.0113, ..., -0.0130, -0.0791, -0.0917], + [ 0.0304, 0.0558, -0.0875, ..., -0.0753, -0.0186, 0.0784], + [ 0.0488, -0.0706, -0.0809, ..., -0.0049, -0.0421, -0.0323], + ..., + [ 0.0574, 0.0247, 0.0864, ..., 0.0472, 0.1221, 0.0075], + [ 0.0716, -0.1009, -0.0931, ..., -0.1273, -0.0496, 0.1182], + [-0.0593, 0.0717, 0.0234, ..., -0.0790, -0.0493, -0.0063]], + device='cuda:0'), grad: tensor([[ 1.1791e-06, -2.0996e-05, 1.5264e-06, ..., -3.1106e-07, + 1.0580e-06, 4.2561e-07], + [ 1.1817e-05, 5.8673e-07, 1.7300e-05, ..., 2.6394e-06, + 1.2539e-05, 2.9057e-06], + [ 6.3144e-06, 1.4156e-06, 8.1658e-06, ..., 1.0012e-06, + 5.6587e-06, 1.9316e-06], + ..., + [-1.0002e-04, 9.0003e-06, -1.5390e-04, ..., -2.2918e-05, + -8.7440e-05, 3.2246e-05], + [ 3.4925e-06, 1.7947e-06, 8.2776e-06, ..., 1.4817e-06, + 6.2175e-06, -2.7474e-07], + [ 6.7294e-05, 9.1121e-06, 8.4817e-05, ..., 1.3545e-05, + 7.3254e-05, 1.4916e-05]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0101, -0.0062, -0.0078, -0.0148, -0.0110, -0.0022, 0.0104, 0.0089, + 0.0183, -0.0056], device='cuda:0'), grad: tensor([-3.6329e-05, 3.0279e-05, 2.1949e-05, 4.1425e-05, -2.1756e-04, + 2.8592e-07, -1.3672e-05, -5.5507e-06, 2.5049e-05, 1.5402e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 214.78, cls_loss 0.0040 cls_loss_mapping 0.0066 cls_loss_causal 0.5680 re_mapping 0.0084 re_causal 0.0251 /// teacc 98.87 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.1430, 0.0675, -0.0117, ..., -0.0131, -0.0799, -0.0919], + [ 0.0305, 0.0556, -0.0872, ..., -0.0754, -0.0185, 0.0787], + [ 0.0491, -0.0706, -0.0808, ..., -0.0044, -0.0419, -0.0325], + ..., + [ 0.0573, 0.0248, 0.0869, ..., 0.0472, 0.1224, 0.0074], + [ 0.0719, -0.1021, -0.0939, ..., -0.1280, -0.0498, 0.1193], + [-0.0602, 0.0725, 0.0233, ..., -0.0797, -0.0500, -0.0063]], + device='cuda:0'), grad: tensor([[ 8.5123e-07, 5.7369e-07, 1.0617e-06, ..., 6.3330e-07, + 1.0468e-06, 3.0920e-07], + [ 1.3877e-07, -1.5553e-07, 2.0266e-06, ..., 1.6941e-06, + 9.1549e-07, -4.0829e-06], + [-4.9453e-07, 3.1330e-06, 3.1721e-06, ..., -1.0217e-06, + -1.2694e-06, 2.3693e-06], + ..., + [-1.0714e-05, -6.8061e-06, -1.5825e-05, ..., -9.6560e-06, + -1.5363e-05, -1.0254e-06], + [-2.7996e-06, 3.5129e-06, 2.8554e-06, ..., 1.5870e-06, + 2.0396e-06, -5.0478e-06], + [ 5.3309e-06, -3.1721e-06, -3.8929e-07, ..., 4.9621e-06, + 6.8061e-06, 1.3476e-06]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0089, -0.0061, -0.0074, -0.0154, -0.0116, -0.0022, 0.0095, 0.0090, + 0.0180, -0.0054], device='cuda:0'), grad: tensor([ 5.2489e-06, -2.2911e-06, 1.0937e-05, 4.9680e-05, 2.4974e-05, + -4.4852e-05, -6.0461e-06, -2.2307e-05, 7.2047e-06, -2.2575e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 214.61, cls_loss 0.0041 cls_loss_mapping 0.0065 cls_loss_causal 0.5779 re_mapping 0.0091 re_causal 0.0244 /// teacc 98.89 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.1437, 0.0677, -0.0122, ..., -0.0132, -0.0809, -0.0924], + [ 0.0303, 0.0563, -0.0876, ..., -0.0755, -0.0187, 0.0791], + [ 0.0494, -0.0709, -0.0806, ..., -0.0041, -0.0415, -0.0326], + ..., + [ 0.0572, 0.0240, 0.0869, ..., 0.0467, 0.1224, 0.0068], + [ 0.0733, -0.1028, -0.0935, ..., -0.1292, -0.0487, 0.1210], + [-0.0607, 0.0728, 0.0237, ..., -0.0806, -0.0505, -0.0071]], + device='cuda:0'), grad: tensor([[ 1.3784e-07, -2.2165e-06, 1.3318e-07, ..., -1.6158e-06, + 1.9651e-07, 2.0210e-07], + [-5.0291e-07, 1.1567e-06, 3.0641e-07, ..., 3.7067e-07, + 4.4052e-07, -2.4829e-06], + [ 7.9162e-07, 5.7928e-07, 1.5544e-06, ..., 7.6741e-07, + 6.5379e-07, 7.4785e-07], + ..., + [-1.5339e-06, 9.3877e-07, -2.1476e-06, ..., -1.1865e-06, + -2.5313e-06, 2.0545e-06], + [ 1.6019e-07, 1.0496e-06, 1.9278e-07, ..., 8.0187e-07, + 1.4435e-07, 7.9256e-07], + [ 1.2945e-07, 1.0077e-06, -7.4506e-08, ..., 4.6100e-07, + 1.2573e-07, 7.3668e-07]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0091, -0.0063, -0.0071, -0.0155, -0.0118, -0.0022, 0.0103, 0.0086, + 0.0183, -0.0055], device='cuda:0'), grad: tensor([-4.4927e-06, 1.6857e-07, 2.6636e-06, 1.3672e-06, -1.0036e-05, + -1.1828e-06, 1.1194e-06, 8.2422e-07, 6.6161e-06, 2.9169e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 214.63, cls_loss 0.0044 cls_loss_mapping 0.0070 cls_loss_causal 0.5524 re_mapping 0.0088 re_causal 0.0257 /// teacc 98.85 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.1442, 0.0682, -0.0125, ..., -0.0124, -0.0809, -0.0922], + [ 0.0304, 0.0566, -0.0876, ..., -0.0762, -0.0189, 0.0797], + [ 0.0494, -0.0713, -0.0815, ..., -0.0043, -0.0418, -0.0326], + ..., + [ 0.0572, 0.0232, 0.0873, ..., 0.0482, 0.1233, 0.0057], + [ 0.0737, -0.1036, -0.0941, ..., -0.1302, -0.0491, 0.1217], + [-0.0610, 0.0722, 0.0236, ..., -0.0821, -0.0507, -0.0078]], + device='cuda:0'), grad: tensor([[ 4.7497e-07, 6.3106e-06, 1.1735e-06, ..., 7.8138e-07, + 3.8743e-07, 8.8811e-06], + [ 1.2722e-06, 2.5984e-06, 2.8033e-06, ..., 1.2089e-06, + 1.8580e-06, 2.0266e-06], + [ 3.5614e-06, 7.2643e-06, 8.5905e-06, ..., 8.2627e-06, + 2.6617e-06, 1.1161e-05], + ..., + [-2.4959e-06, 3.8482e-06, 5.6624e-07, ..., 3.7719e-06, + -3.8408e-06, 4.7237e-06], + [ 4.1686e-06, 4.7460e-06, 9.3728e-06, ..., 8.3223e-06, + 3.4198e-06, 7.3463e-06], + [ 2.0079e-06, -1.4752e-05, -6.2771e-06, ..., 1.1409e-06, + 1.7928e-06, -6.9030e-06]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0083, -0.0063, -0.0071, -0.0147, -0.0113, -0.0024, 0.0111, 0.0082, + 0.0178, -0.0063], device='cuda:0'), grad: tensor([ 3.8207e-05, 1.1228e-05, 3.9220e-05, -3.5614e-05, -1.0973e-04, + 4.0308e-06, 2.8774e-05, 1.4007e-05, 1.9968e-05, -1.0207e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 214.54, cls_loss 0.0051 cls_loss_mapping 0.0078 cls_loss_causal 0.5790 re_mapping 0.0088 re_causal 0.0261 /// teacc 98.81 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.1448, 0.0668, -0.0127, ..., -0.0142, -0.0814, -0.0930], + [ 0.0301, 0.0566, -0.0881, ..., -0.0765, -0.0192, 0.0795], + [ 0.0497, -0.0718, -0.0815, ..., -0.0043, -0.0415, -0.0325], + ..., + [ 0.0571, 0.0229, 0.0868, ..., 0.0475, 0.1237, 0.0053], + [ 0.0738, -0.1044, -0.0953, ..., -0.1314, -0.0495, 0.1220], + [-0.0617, 0.0726, 0.0224, ..., -0.0808, -0.0515, -0.0090]], + device='cuda:0'), grad: tensor([[ 3.5949e-07, -1.1474e-06, 7.1898e-07, ..., 2.8312e-07, + 7.0129e-07, -4.3772e-08], + [ 3.1088e-06, 4.4797e-07, 5.8636e-06, ..., 3.7756e-06, + 6.0014e-06, -1.0738e-06], + [ 6.7167e-06, 3.8557e-07, 1.3053e-05, ..., 7.9572e-06, + 1.3813e-05, 2.5705e-07], + ..., + [-2.8536e-05, 2.8219e-07, -5.5075e-05, ..., -3.4302e-05, + -5.5045e-05, 2.3376e-07], + [-1.6391e-07, 4.4145e-06, 7.3574e-07, ..., 1.0533e-06, + 5.0757e-07, -1.6298e-07], + [ 1.2899e-06, 5.6159e-07, 2.5947e-06, ..., 1.9129e-06, + 2.2836e-06, 4.6939e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0104, -0.0067, -0.0070, -0.0166, -0.0093, 0.0005, 0.0108, 0.0078, + 0.0173, -0.0058], device='cuda:0'), grad: tensor([-1.6857e-07, 8.6501e-06, 1.6987e-05, 5.6267e-05, 2.3320e-06, + -9.1255e-05, 5.0902e-05, -6.6400e-05, 1.8269e-05, 4.4331e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 214.41, cls_loss 0.0038 cls_loss_mapping 0.0062 cls_loss_causal 0.5496 re_mapping 0.0084 re_causal 0.0243 /// teacc 98.93 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.1453, 0.0663, -0.0137, ..., -0.0149, -0.0820, -0.0931], + [ 0.0300, 0.0568, -0.0889, ..., -0.0766, -0.0194, 0.0796], + [ 0.0496, -0.0722, -0.0819, ..., -0.0043, -0.0415, -0.0328], + ..., + [ 0.0575, 0.0223, 0.0873, ..., 0.0477, 0.1241, 0.0052], + [ 0.0746, -0.1055, -0.0966, ..., -0.1333, -0.0497, 0.1229], + [-0.0617, 0.0739, 0.0232, ..., -0.0816, -0.0517, -0.0083]], + device='cuda:0'), grad: tensor([[ 6.3796e-07, 7.3798e-06, 1.0636e-06, ..., 3.1292e-07, + 6.5099e-07, 1.3504e-06], + [ 3.8892e-06, -6.8806e-06, 4.0792e-06, ..., 2.4810e-06, + 4.1015e-06, -2.8223e-05], + [ 8.8364e-06, 3.4254e-06, 1.1049e-05, ..., 6.3404e-06, + 1.2197e-05, 9.2313e-06], + ..., + [-1.9848e-05, 2.5760e-06, -2.7820e-05, ..., -1.5602e-05, + -3.0905e-05, 2.8778e-06], + [-3.0845e-05, 8.9929e-06, 1.1930e-06, ..., 8.4005e-07, + 7.4878e-07, -3.1829e-05], + [ 1.4575e-06, 2.5183e-05, 5.8860e-07, ..., 1.7462e-06, + 1.7378e-06, 1.2016e-04]], device='cuda:0') +Epoch 126, bias, value: tensor([-1.0898e-02, -6.7702e-03, -7.0332e-03, -1.5952e-02, -9.9472e-03, + -5.1618e-05, 1.0435e-02, 7.8232e-03, 1.6963e-02, -4.7845e-03], + device='cuda:0'), grad: tensor([ 4.7743e-05, -4.7207e-05, 3.4213e-05, 2.7090e-05, -3.4380e-04, + 6.2823e-05, 7.0572e-05, -2.4021e-05, -2.0549e-05, 1.9348e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 214.35, cls_loss 0.0042 cls_loss_mapping 0.0068 cls_loss_causal 0.5487 re_mapping 0.0082 re_causal 0.0238 /// teacc 98.87 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.1461, 0.0661, -0.0139, ..., -0.0148, -0.0828, -0.0941], + [ 0.0303, 0.0576, -0.0890, ..., -0.0766, -0.0191, 0.0811], + [ 0.0493, -0.0730, -0.0827, ..., -0.0047, -0.0419, -0.0342], + ..., + [ 0.0576, 0.0214, 0.0876, ..., 0.0477, 0.1244, 0.0046], + [ 0.0747, -0.1079, -0.0975, ..., -0.1344, -0.0500, 0.1232], + [-0.0620, 0.0749, 0.0234, ..., -0.0826, -0.0519, -0.0085]], + device='cuda:0'), grad: tensor([[ 1.3970e-07, -2.5127e-06, 1.5097e-06, ..., 1.7472e-06, + 2.4866e-07, 4.6566e-08], + [ 5.3830e-07, 8.4098e-07, 2.6524e-06, ..., 2.5835e-06, + 1.0682e-06, -7.9162e-08], + [-1.9670e-06, 7.3798e-06, 2.5570e-05, ..., 3.0160e-05, + -3.5632e-06, 2.1607e-07], + ..., + [-1.7500e-04, -9.0241e-05, -4.7684e-04, ..., -3.9309e-05, + -4.3845e-04, 1.0617e-07], + [-1.0096e-06, 1.6615e-06, 2.0005e-06, ..., 2.3842e-06, + 3.8929e-07, -1.5590e-06], + [ 1.7548e-04, 9.1434e-05, 4.7493e-04, ..., 4.3184e-05, + 4.3821e-04, 3.9954e-07]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0110, -0.0063, -0.0076, -0.0147, -0.0102, -0.0010, 0.0109, 0.0076, + 0.0155, -0.0040], device='cuda:0'), grad: tensor([-2.6990e-06, 7.5586e-06, 5.0902e-05, -7.5042e-05, 3.8892e-06, + 1.6403e-04, -1.6320e-04, -6.8521e-04, 7.7263e-06, 6.9237e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 214.78, cls_loss 0.0037 cls_loss_mapping 0.0065 cls_loss_causal 0.5467 re_mapping 0.0085 re_causal 0.0241 /// teacc 98.86 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.1465, 0.0663, -0.0143, ..., -0.0150, -0.0836, -0.0944], + [ 0.0300, 0.0577, -0.0894, ..., -0.0767, -0.0192, 0.0809], + [ 0.0493, -0.0733, -0.0834, ..., -0.0049, -0.0420, -0.0343], + ..., + [ 0.0571, 0.0213, 0.0875, ..., 0.0475, 0.1241, 0.0034], + [ 0.0775, -0.1084, -0.0948, ..., -0.1327, -0.0474, 0.1256], + [-0.0642, 0.0745, 0.0225, ..., -0.0832, -0.0537, -0.0091]], + device='cuda:0'), grad: tensor([[ 7.7020e-07, -6.4634e-07, 1.2536e-06, ..., 4.8615e-07, + 1.2834e-06, 8.8476e-08], + [ 3.3349e-05, 9.6709e-06, 5.4866e-05, ..., 3.4660e-05, + 6.7115e-05, -1.1772e-06], + [ 6.8583e-06, 2.9504e-06, 1.0230e-05, ..., 1.8338e-06, + 1.1399e-05, 5.2992e-07], + ..., + [-1.7989e-04, -5.6863e-05, -2.7418e-04, ..., -1.3602e-04, + -3.3307e-04, -1.5348e-06], + [ 2.8890e-06, 1.7453e-06, 8.1360e-06, ..., 4.6454e-06, + 9.9540e-06, -2.3302e-06], + [ 8.3923e-05, 1.8075e-05, 1.2010e-04, ..., 5.7787e-05, + 1.5318e-04, 6.0257e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0110, -0.0065, -0.0078, -0.0147, -0.0099, -0.0010, 0.0109, 0.0077, + 0.0175, -0.0052], device='cuda:0'), grad: tensor([ 5.9083e-06, 1.1575e-04, 1.9982e-05, 1.1176e-04, 5.8740e-05, + 1.2882e-05, -1.2353e-05, -5.7125e-04, 1.3717e-05, 2.4533e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 214.54, cls_loss 0.0034 cls_loss_mapping 0.0055 cls_loss_causal 0.5744 re_mapping 0.0080 re_causal 0.0250 /// teacc 98.73 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.1472, 0.0663, -0.0146, ..., -0.0151, -0.0844, -0.0949], + [ 0.0295, 0.0577, -0.0902, ..., -0.0773, -0.0198, 0.0810], + [ 0.0493, -0.0734, -0.0839, ..., -0.0049, -0.0422, -0.0343], + ..., + [ 0.0579, 0.0217, 0.0886, ..., 0.0484, 0.1252, 0.0033], + [ 0.0776, -0.1088, -0.0950, ..., -0.1331, -0.0475, 0.1260], + [-0.0647, 0.0747, 0.0223, ..., -0.0839, -0.0542, -0.0090]], + device='cuda:0'), grad: tensor([[ 2.7008e-07, -1.8343e-05, 1.9092e-07, ..., -8.1137e-06, + 2.0489e-08, 7.7393e-07], + [-1.1353e-06, -1.7919e-06, 3.7439e-07, ..., 4.6846e-07, + -1.8626e-09, -4.1649e-06], + [-6.8918e-08, 4.1723e-06, 2.3004e-07, ..., 1.7025e-06, + -1.4156e-07, 4.2003e-07], + ..., + [ 2.5425e-07, 2.6189e-06, 2.5425e-06, ..., 3.8091e-07, + 1.9744e-07, 8.6799e-07], + [ 6.3330e-08, 2.3916e-06, 1.5534e-06, ..., 8.3819e-07, + 6.5193e-08, 4.9174e-07], + [ 6.9849e-08, 2.9225e-06, -1.2204e-05, ..., 4.9770e-06, + -5.4855e-07, -1.2107e-07]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0113, -0.0069, -0.0078, -0.0150, -0.0098, -0.0009, 0.0114, 0.0084, + 0.0173, -0.0052], device='cuda:0'), grad: tensor([-4.7863e-05, -4.5598e-06, 1.0766e-05, 5.3272e-06, 1.4193e-05, + -5.5656e-06, 3.3565e-06, 6.4671e-06, 6.1207e-06, 1.1623e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 214.67, cls_loss 0.0034 cls_loss_mapping 0.0059 cls_loss_causal 0.5587 re_mapping 0.0076 re_causal 0.0229 /// teacc 98.93 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.1476, 0.0661, -0.0149, ..., -0.0156, -0.0852, -0.0952], + [ 0.0282, 0.0575, -0.0910, ..., -0.0774, -0.0210, 0.0799], + [ 0.0496, -0.0736, -0.0840, ..., -0.0046, -0.0420, -0.0343], + ..., + [ 0.0590, 0.0213, 0.0890, ..., 0.0485, 0.1265, 0.0043], + [ 0.0777, -0.1093, -0.0953, ..., -0.1338, -0.0477, 0.1264], + [-0.0648, 0.0749, 0.0230, ..., -0.0846, -0.0542, -0.0091]], + device='cuda:0'), grad: tensor([[ 1.0151e-07, -2.6766e-06, 1.2107e-07, ..., -8.7917e-07, + 6.0536e-08, 1.7043e-07], + [-4.0419e-07, -1.9614e-06, 9.8720e-08, ..., 9.0338e-08, + -8.4843e-07, -4.8392e-06], + [ 4.7125e-07, 5.2247e-07, 6.2399e-08, ..., -2.5891e-07, + -7.8231e-08, 1.5134e-06], + ..., + [ 4.6473e-07, 5.8487e-07, -1.4994e-07, ..., -5.4017e-08, + 2.5798e-07, 1.3784e-06], + [-2.8666e-06, 1.0934e-06, 5.9418e-07, ..., 3.9022e-07, + -1.6019e-07, -3.1944e-06], + [ 7.1898e-07, 1.1355e-05, 2.3358e-06, ..., 2.9858e-06, + -3.1851e-07, 8.5309e-07]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0115, -0.0081, -0.0075, -0.0149, -0.0099, -0.0008, 0.0121, 0.0093, + 0.0169, -0.0052], device='cuda:0'), grad: tensor([-4.2394e-06, -6.8806e-06, 1.9725e-06, 1.2748e-05, 2.2538e-06, + -7.2241e-05, 4.1127e-05, 2.6040e-06, -1.7900e-06, 2.4483e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 214.55, cls_loss 0.0035 cls_loss_mapping 0.0056 cls_loss_causal 0.5515 re_mapping 0.0080 re_causal 0.0236 /// teacc 98.69 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.1480, 0.0664, -0.0151, ..., -0.0151, -0.0861, -0.0955], + [ 0.0289, 0.0592, -0.0887, ..., -0.0775, -0.0201, 0.0812], + [ 0.0495, -0.0739, -0.0846, ..., -0.0046, -0.0420, -0.0344], + ..., + [ 0.0588, 0.0195, 0.0888, ..., 0.0494, 0.1262, 0.0030], + [ 0.0779, -0.1094, -0.0956, ..., -0.1344, -0.0478, 0.1269], + [-0.0654, 0.0751, 0.0227, ..., -0.0858, -0.0551, -0.0086]], + device='cuda:0'), grad: tensor([[ 7.3835e-06, 2.0582e-07, 2.7772e-06, ..., 2.3749e-07, + 5.6531e-07, 9.1344e-06], + [ 3.7923e-06, 1.6261e-06, 3.1590e-06, ..., 9.4622e-07, + 1.6177e-06, 3.9674e-06], + [ 6.6385e-06, 9.3728e-06, 3.9265e-06, ..., -3.8184e-07, + 1.2619e-06, 1.4998e-05], + ..., + [-5.3570e-06, 3.5930e-06, -7.1451e-06, ..., -2.9132e-06, + -6.2995e-06, 5.0813e-06], + [-4.9919e-05, 1.3197e-06, -1.2077e-05, ..., 2.7008e-07, + 9.8534e-07, -6.1393e-05], + [ 3.0696e-05, -1.4830e-04, -2.0409e-04, ..., 6.3796e-07, + -4.9680e-05, -7.9274e-05]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0111, -0.0070, -0.0075, -0.0148, -0.0104, -0.0011, 0.0116, 0.0085, + 0.0170, -0.0050], device='cuda:0'), grad: tensor([ 2.2277e-05, 1.4901e-05, 4.8280e-05, 9.0599e-06, 3.4547e-04, + 7.6145e-06, -4.3549e-06, 3.3975e-06, -1.2457e-04, -3.2258e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 214.69, cls_loss 0.0032 cls_loss_mapping 0.0055 cls_loss_causal 0.5380 re_mapping 0.0081 re_causal 0.0234 /// teacc 98.90 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.1484, 0.0673, -0.0154, ..., -0.0148, -0.0865, -0.0959], + [ 0.0288, 0.0591, -0.0887, ..., -0.0777, -0.0203, 0.0812], + [ 0.0499, -0.0742, -0.0853, ..., -0.0046, -0.0418, -0.0342], + ..., + [ 0.0589, 0.0194, 0.0891, ..., 0.0496, 0.1264, 0.0029], + [ 0.0777, -0.1102, -0.0959, ..., -0.1353, -0.0478, 0.1273], + [-0.0657, 0.0759, 0.0233, ..., -0.0861, -0.0553, -0.0076]], + device='cuda:0'), grad: tensor([[ 5.3365e-07, -2.6263e-06, 1.7695e-07, ..., 8.2888e-08, + 2.4587e-07, 7.0035e-07], + [ 1.1027e-06, -4.3306e-07, 1.0002e-06, ..., 5.3830e-07, + 4.7218e-07, -1.7220e-06], + [ 3.4403e-06, 1.1567e-06, 7.3109e-07, ..., -9.8348e-07, + -4.9453e-07, 6.1616e-06], + ..., + [-4.1537e-07, 2.6729e-06, -5.4203e-06, ..., -1.2973e-06, + -3.4627e-06, 8.0913e-06], + [-2.3651e-04, 4.5113e-06, 1.5981e-06, ..., 3.1013e-07, + 2.5891e-07, -1.9431e-04], + [ 1.8269e-05, 7.4282e-06, 4.3400e-07, ..., 8.5775e-07, + 1.6410e-06, 2.6569e-05]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0098, -0.0073, -0.0071, -0.0147, -0.0115, -0.0012, 0.0104, 0.0085, + 0.0165, -0.0041], device='cuda:0'), grad: tensor([-3.7029e-06, -1.5991e-06, 9.1717e-06, 4.9099e-06, -4.4048e-05, + 5.8126e-04, 5.2974e-06, 8.5607e-06, -6.1798e-04, 5.7995e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 214.58, cls_loss 0.0033 cls_loss_mapping 0.0060 cls_loss_causal 0.5468 re_mapping 0.0078 re_causal 0.0230 /// teacc 98.89 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.1489, 0.0680, -0.0157, ..., -0.0146, -0.0874, -0.0964], + [ 0.0289, 0.0591, -0.0888, ..., -0.0779, -0.0203, 0.0815], + [ 0.0503, -0.0748, -0.0855, ..., -0.0048, -0.0413, -0.0343], + ..., + [ 0.0586, 0.0189, 0.0888, ..., 0.0496, 0.1261, 0.0028], + [ 0.0778, -0.1110, -0.0961, ..., -0.1360, -0.0479, 0.1275], + [-0.0660, 0.0748, 0.0232, ..., -0.0864, -0.0552, -0.0088]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, -2.1961e-06, 3.1944e-07, ..., -4.4890e-07, + 1.3411e-07, 8.5682e-08], + [ 4.0233e-07, 1.8626e-09, 6.1840e-07, ..., 4.2655e-07, + 7.2829e-07, -6.4354e-07], + [-1.6615e-06, 4.9174e-07, 9.8441e-07, ..., -7.1898e-07, + -1.6419e-06, 3.3062e-07], + ..., + [-1.6587e-06, 4.0941e-06, 9.9242e-06, ..., 2.9188e-06, + 5.2713e-07, 2.6543e-07], + [ 5.1130e-07, 1.4510e-06, 1.2871e-06, ..., 7.6834e-07, + 7.7020e-07, 5.7742e-08], + [ 2.1514e-07, -9.2387e-07, -2.1681e-05, ..., -6.5453e-06, + -4.3660e-06, 2.0675e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0090, -0.0073, -0.0070, -0.0146, -0.0109, -0.0014, 0.0106, 0.0083, + 0.0161, -0.0047], device='cuda:0'), grad: tensor([-6.2212e-07, 1.6056e-06, -4.7684e-06, 2.1279e-05, -1.7941e-05, + -3.2216e-05, 1.6987e-05, 2.0728e-05, 8.7917e-06, -1.3866e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 214.49, cls_loss 0.0041 cls_loss_mapping 0.0058 cls_loss_causal 0.5755 re_mapping 0.0078 re_causal 0.0231 /// teacc 98.98 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.1492, 0.0674, -0.0160, ..., -0.0149, -0.0879, -0.0970], + [ 0.0288, 0.0581, -0.0893, ..., -0.0782, -0.0205, 0.0817], + [ 0.0502, -0.0747, -0.0861, ..., -0.0047, -0.0415, -0.0344], + ..., + [ 0.0590, 0.0178, 0.0892, ..., 0.0500, 0.1266, 0.0027], + [ 0.0782, -0.1110, -0.0963, ..., -0.1366, -0.0480, 0.1284], + [-0.0667, 0.0758, 0.0235, ..., -0.0870, -0.0554, -0.0091]], + device='cuda:0'), grad: tensor([[ 3.0641e-07, -6.4913e-07, 4.8429e-08, ..., -2.1420e-08, + 2.2072e-07, 1.7602e-07], + [-1.9502e-06, -1.2051e-06, 6.6124e-08, ..., 2.4214e-07, + -9.3505e-07, -4.4852e-06], + [-5.4017e-07, 6.3330e-07, 5.7090e-07, ..., -4.5542e-07, + -9.8534e-07, 1.2163e-06], + ..., + [ 1.7444e-06, 1.3625e-06, 5.0291e-08, ..., 4.0699e-07, + 9.2480e-07, 2.7176e-06], + [-6.9197e-07, 6.9942e-07, 1.3318e-07, ..., 2.4494e-07, + 2.3842e-07, -5.9791e-07], + [ 3.5763e-07, 1.5832e-06, 4.5635e-08, ..., 1.9837e-07, + 1.4156e-07, 1.1390e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0097, -0.0080, -0.0067, -0.0149, -0.0109, -0.0011, 0.0101, 0.0083, + 0.0166, -0.0040], device='cuda:0'), grad: tensor([ 2.9299e-06, -8.5682e-06, -1.3504e-07, 7.3686e-06, 1.0610e-05, + -6.0834e-06, -2.3082e-05, 8.7991e-06, 1.9353e-06, 6.2063e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 214.55, cls_loss 0.0039 cls_loss_mapping 0.0058 cls_loss_causal 0.5726 re_mapping 0.0079 re_causal 0.0225 /// teacc 98.85 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.1494, 0.0681, -0.0166, ..., -0.0148, -0.0887, -0.0980], + [ 0.0287, 0.0577, -0.0898, ..., -0.0784, -0.0205, 0.0818], + [ 0.0503, -0.0753, -0.0865, ..., -0.0045, -0.0415, -0.0346], + ..., + [ 0.0594, 0.0175, 0.0900, ..., 0.0503, 0.1271, 0.0026], + [ 0.0785, -0.1115, -0.0965, ..., -0.1371, -0.0481, 0.1291], + [-0.0674, 0.0753, 0.0233, ..., -0.0879, -0.0559, -0.0095]], + device='cuda:0'), grad: tensor([[ 4.6100e-07, -3.5405e-05, 2.6189e-06, ..., 5.4017e-06, + 1.0245e-07, 2.1420e-07], + [ 1.6559e-06, 5.6587e-06, 8.4843e-07, ..., 1.3830e-06, + 2.0117e-07, 4.4852e-06], + [-5.3458e-06, 2.0817e-05, 2.2396e-05, ..., 4.5478e-05, + -6.4913e-07, 8.0373e-07], + ..., + [-1.4715e-07, 3.4962e-06, -4.4145e-07, ..., 1.9111e-06, + -4.7311e-07, 1.8552e-06], + [-1.7807e-06, 3.4776e-06, 3.0994e-06, ..., 6.9067e-06, + 1.6298e-07, -4.2617e-06], + [ 9.3691e-07, 5.3376e-05, 4.3660e-06, ..., 8.4937e-06, + 1.4529e-07, 1.1444e-05]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0096, -0.0082, -0.0067, -0.0151, -0.0104, -0.0012, 0.0108, 0.0086, + 0.0167, -0.0047], device='cuda:0'), grad: tensor([-6.6221e-05, 1.6108e-05, 1.3053e-04, 1.3185e-04, -4.8727e-05, + -3.3903e-04, 1.5989e-05, 1.1109e-05, 1.6659e-05, 1.3173e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 214.84, cls_loss 0.0041 cls_loss_mapping 0.0059 cls_loss_causal 0.5543 re_mapping 0.0079 re_causal 0.0227 /// teacc 98.86 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.1499, 0.0682, -0.0174, ..., -0.0144, -0.0892, -0.0985], + [ 0.0275, 0.0581, -0.0916, ..., -0.0785, -0.0219, 0.0804], + [ 0.0506, -0.0756, -0.0872, ..., -0.0043, -0.0415, -0.0347], + ..., + [ 0.0605, 0.0168, 0.0912, ..., 0.0501, 0.1288, 0.0041], + [ 0.0787, -0.1118, -0.0967, ..., -0.1383, -0.0481, 0.1300], + [-0.0678, 0.0755, 0.0236, ..., -0.0890, -0.0562, -0.0097]], + device='cuda:0'), grad: tensor([[ 1.2144e-06, 1.7695e-08, 5.9232e-07, ..., 8.1304e-07, + 1.0533e-06, 1.3243e-06], + [ 3.9428e-05, -9.4343e-07, 2.6166e-05, ..., 3.9786e-06, + 4.8667e-05, 2.4602e-05], + [-2.0653e-05, 1.2293e-06, 5.4762e-06, ..., -6.0946e-06, + -1.1012e-05, -2.8908e-05], + ..., + [-4.8488e-05, 5.3737e-07, -3.8594e-05, ..., 1.2824e-06, + -6.6519e-05, -2.4945e-05], + [ 1.7405e-05, 2.7753e-07, 5.5395e-06, ..., 8.2552e-06, + 1.4618e-05, 1.9640e-05], + [ 1.0710e-06, -3.0827e-07, 5.0291e-07, ..., 5.3179e-07, + 1.4240e-06, 7.4413e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0096, -0.0095, -0.0065, -0.0154, -0.0103, -0.0010, 0.0104, 0.0099, + 0.0169, -0.0048], device='cuda:0'), grad: tensor([ 3.1665e-06, 7.4148e-05, -3.7640e-05, -5.9605e-08, 2.6710e-06, + 3.2857e-06, 1.0990e-06, -8.8274e-05, 3.9876e-05, 1.8952e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 214.58, cls_loss 0.0034 cls_loss_mapping 0.0057 cls_loss_causal 0.5186 re_mapping 0.0082 re_causal 0.0228 /// teacc 98.89 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.1505, 0.0686, -0.0178, ..., -0.0143, -0.0892, -0.0993], + [ 0.0274, 0.0588, -0.0921, ..., -0.0786, -0.0220, 0.0808], + [ 0.0506, -0.0759, -0.0874, ..., -0.0041, -0.0414, -0.0353], + ..., + [ 0.0607, 0.0160, 0.0920, ..., 0.0503, 0.1290, 0.0039], + [ 0.0791, -0.1130, -0.0971, ..., -0.1388, -0.0482, 0.1309], + [-0.0680, 0.0754, 0.0241, ..., -0.0895, -0.0563, -0.0097]], + device='cuda:0'), grad: tensor([[ 7.4226e-07, -6.6124e-08, 1.0394e-06, ..., 2.4587e-07, + 5.5972e-07, 4.5449e-07], + [ 4.2953e-06, 2.1979e-06, 1.9204e-06, ..., 5.6624e-07, + 1.2107e-07, 2.5164e-06], + [-1.9912e-06, 4.0829e-06, 5.9232e-07, ..., -8.4564e-07, + -1.3970e-06, 5.1335e-06], + ..., + [-5.1595e-06, 8.4341e-06, -7.3016e-06, ..., -3.6117e-06, + -5.5619e-06, 5.7966e-06], + [ 8.3297e-06, 1.4968e-05, 1.5320e-06, ..., 6.4820e-07, + 1.5572e-06, 1.9982e-05], + [ 1.8194e-05, 1.1079e-05, -4.1984e-06, ..., 2.7530e-06, + 2.1234e-06, 3.6120e-05]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0092, -0.0095, -0.0065, -0.0159, -0.0105, -0.0006, 0.0105, 0.0099, + 0.0170, -0.0048], device='cuda:0'), grad: tensor([ 1.5087e-07, 4.2468e-06, 5.0962e-06, 4.0680e-06, -1.0067e-04, + 1.5227e-06, 3.5372e-06, 9.0227e-06, 3.6448e-05, 3.6627e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 214.77, cls_loss 0.0027 cls_loss_mapping 0.0050 cls_loss_causal 0.5337 re_mapping 0.0082 re_causal 0.0244 /// teacc 98.93 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.1511, 0.0685, -0.0182, ..., -0.0145, -0.0903, -0.1000], + [ 0.0274, 0.0590, -0.0921, ..., -0.0788, -0.0220, 0.0810], + [ 0.0504, -0.0764, -0.0894, ..., -0.0044, -0.0419, -0.0355], + ..., + [ 0.0612, 0.0168, 0.0930, ..., 0.0514, 0.1297, 0.0038], + [ 0.0791, -0.1146, -0.0973, ..., -0.1396, -0.0483, 0.1308], + [-0.0681, 0.0756, 0.0241, ..., -0.0900, -0.0565, -0.0095]], + device='cuda:0'), grad: tensor([[ 1.9893e-06, -1.5959e-05, 2.5332e-07, ..., -1.4435e-06, + 9.8720e-08, -4.1090e-06], + [ 1.8999e-05, 1.7043e-06, 2.6207e-06, ..., 1.4547e-06, + 5.0105e-06, 1.7717e-05], + [ 1.9088e-05, 1.3150e-06, 8.8476e-07, ..., -1.2945e-07, + -2.2613e-06, 2.2382e-05], + ..., + [ 2.5362e-05, 1.3009e-05, 1.5646e-07, ..., 2.3451e-06, + 5.5097e-06, 3.5942e-05], + [-1.2636e-04, 3.3043e-06, 2.7996e-06, ..., 2.7213e-06, + 7.5065e-07, -1.2505e-04], + [ 3.5483e-06, 3.2187e-06, 1.3411e-07, ..., 1.0235e-06, + 8.6240e-07, 6.4261e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0095, -0.0093, -0.0068, -0.0164, -0.0107, -0.0003, 0.0104, 0.0106, + 0.0162, -0.0046], device='cuda:0'), grad: tensor([-5.4330e-05, 3.1263e-05, 3.0607e-05, -1.9938e-05, -4.9055e-05, + 8.6427e-05, 5.6207e-05, 6.6876e-05, -1.6463e-04, 1.6570e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 214.63, cls_loss 0.0038 cls_loss_mapping 0.0049 cls_loss_causal 0.5768 re_mapping 0.0081 re_causal 0.0227 /// teacc 98.78 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.1511, 0.0671, -0.0181, ..., -0.0146, -0.0883, -0.0993], + [ 0.0275, 0.0593, -0.0924, ..., -0.0789, -0.0218, 0.0817], + [ 0.0508, -0.0770, -0.0897, ..., -0.0041, -0.0416, -0.0358], + ..., + [ 0.0610, 0.0156, 0.0933, ..., 0.0513, 0.1296, 0.0032], + [ 0.0792, -0.1154, -0.0975, ..., -0.1402, -0.0484, 0.1313], + [-0.0684, 0.0771, 0.0242, ..., -0.0908, -0.0570, -0.0097]], + device='cuda:0'), grad: tensor([[ 2.4587e-07, -2.0675e-07, 5.9325e-07, ..., 2.9057e-07, + 2.5332e-07, 3.3155e-07], + [ 4.8056e-07, -2.8964e-07, 2.5686e-06, ..., 8.9221e-07, + 8.5868e-07, -4.9621e-06], + [ 7.1712e-07, 1.4575e-06, 1.1893e-06, ..., 8.3447e-07, + 6.5472e-07, 2.5816e-06], + ..., + [-2.9001e-06, 7.1246e-07, -4.5337e-06, ..., -1.6401e-06, + -4.3251e-06, 6.6496e-07], + [-1.4612e-06, 1.2722e-06, 1.4463e-06, ..., 7.1805e-07, + 1.9651e-07, -2.5630e-06], + [ 9.8161e-07, -2.4922e-06, -2.7083e-06, ..., 9.4064e-07, + 1.2936e-06, 5.3085e-08]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0106, -0.0091, -0.0065, -0.0163, -0.0105, -0.0003, 0.0098, 0.0102, + 0.0159, -0.0037], device='cuda:0'), grad: tensor([ 3.0994e-06, -7.4692e-07, 1.3731e-05, -2.3410e-05, -7.9155e-05, + 2.7210e-05, 6.1870e-05, -4.3213e-06, 1.8235e-06, -1.8533e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 214.46, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5689 re_mapping 0.0076 re_causal 0.0235 /// teacc 98.91 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1521, 0.0675, -0.0186, ..., -0.0147, -0.0884, -0.1003], + [ 0.0275, 0.0599, -0.0927, ..., -0.0791, -0.0220, 0.0822], + [ 0.0508, -0.0773, -0.0903, ..., -0.0042, -0.0415, -0.0358], + ..., + [ 0.0614, 0.0154, 0.0940, ..., 0.0512, 0.1301, 0.0030], + [ 0.0794, -0.1155, -0.0977, ..., -0.1410, -0.0485, 0.1322], + [-0.0695, 0.0767, 0.0233, ..., -0.0922, -0.0581, -0.0104]], + device='cuda:0'), grad: tensor([[ 3.1106e-07, 4.9360e-07, 2.2724e-07, ..., 4.9360e-08, + 6.9849e-08, 1.7229e-07], + [ 7.6741e-07, -2.6114e-06, 1.7919e-06, ..., 1.4864e-06, + 1.0207e-06, -3.8892e-06], + [ 3.7625e-07, 4.3586e-07, 8.4005e-07, ..., 5.4669e-07, + 3.4459e-07, 6.7987e-08], + ..., + [-1.6913e-06, 2.2110e-06, 3.4552e-07, ..., -7.8510e-07, + -1.3197e-06, 8.7824e-07], + [ 3.1590e-06, 9.8050e-06, 1.4910e-06, ..., 7.8045e-07, + 4.7870e-07, 1.2340e-06], + [ 4.7404e-07, -3.6657e-06, -5.7220e-06, ..., 6.3702e-07, + 2.8312e-07, 3.1292e-07]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0104, -0.0090, -0.0064, -0.0156, -0.0102, -0.0010, 0.0094, 0.0103, + 0.0162, -0.0044], device='cuda:0'), grad: tensor([ 1.8878e-06, -4.7684e-07, 1.9632e-06, -5.0180e-06, 7.6815e-06, + -3.1888e-05, 2.9542e-06, 2.3488e-06, 2.7955e-05, -7.4692e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 214.59, cls_loss 0.0033 cls_loss_mapping 0.0061 cls_loss_causal 0.5438 re_mapping 0.0081 re_causal 0.0239 /// teacc 98.94 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.1529, 0.0686, -0.0192, ..., -0.0140, -0.0888, -0.1011], + [ 0.0271, 0.0593, -0.0932, ..., -0.0797, -0.0225, 0.0819], + [ 0.0517, -0.0764, -0.0897, ..., -0.0030, -0.0405, -0.0357], + ..., + [ 0.0612, 0.0154, 0.0942, ..., 0.0504, 0.1301, 0.0033], + [ 0.0798, -0.1165, -0.0983, ..., -0.1422, -0.0486, 0.1329], + [-0.0698, 0.0768, 0.0232, ..., -0.0935, -0.0583, -0.0107]], + device='cuda:0'), grad: tensor([[ 5.4482e-07, -5.2862e-06, 2.1793e-07, ..., -2.5332e-07, + 3.7253e-09, 1.5199e-06], + [ 2.5719e-05, 9.2294e-07, 2.6915e-07, ..., 2.7474e-07, + 1.3039e-08, 7.5221e-05], + [ 1.3225e-06, 5.8580e-07, 3.0268e-07, ..., 9.1270e-08, + -2.6356e-07, 4.4294e-06], + ..., + [ 7.0781e-07, 5.1223e-07, -1.3970e-07, ..., 1.3877e-07, + 9.9652e-08, 2.0098e-06], + [-3.0234e-05, 1.4063e-07, 1.0626e-06, ..., 7.3574e-07, + -7.7300e-08, -8.9645e-05], + [ 2.2911e-07, 6.6683e-06, 7.1246e-07, ..., 2.0098e-06, + 3.2596e-08, 1.0319e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0098, -0.0098, -0.0051, -0.0157, -0.0101, -0.0008, 0.0081, 0.0102, + 0.0164, -0.0044], device='cuda:0'), grad: tensor([-5.3421e-06, 7.9751e-05, 5.3272e-06, -3.4451e-05, -7.0129e-07, + 4.9055e-05, -2.0996e-05, 3.0175e-06, -8.5890e-05, 1.0148e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 214.70, cls_loss 0.0027 cls_loss_mapping 0.0051 cls_loss_causal 0.5738 re_mapping 0.0077 re_causal 0.0234 /// teacc 98.87 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.1546, 0.0680, -0.0212, ..., -0.0148, -0.0908, -0.1016], + [ 0.0272, 0.0597, -0.0933, ..., -0.0798, -0.0224, 0.0823], + [ 0.0516, -0.0770, -0.0901, ..., -0.0030, -0.0407, -0.0362], + ..., + [ 0.0616, 0.0154, 0.0950, ..., 0.0513, 0.1305, 0.0032], + [ 0.0801, -0.1152, -0.0992, ..., -0.1436, -0.0489, 0.1339], + [-0.0702, 0.0764, 0.0228, ..., -0.0952, -0.0586, -0.0111]], + device='cuda:0'), grad: tensor([[ 9.1270e-08, -5.5842e-06, 3.2596e-08, ..., -1.1548e-07, + 5.6811e-08, 3.6508e-07], + [-1.1353e-06, -5.3085e-06, 1.4435e-07, ..., 1.2200e-07, + -4.4052e-07, -6.3591e-06], + [ 3.1479e-07, 6.5472e-07, 1.9744e-07, ..., -4.5355e-07, + -4.2841e-08, 9.6671e-07], + ..., + [ 2.9337e-07, 1.8468e-06, -1.5460e-07, ..., -2.9802e-08, + -5.4948e-08, 2.1290e-06], + [-1.7621e-06, 5.7276e-07, 1.5087e-07, ..., 4.2841e-08, + 6.5193e-08, -2.2799e-06], + [ 2.7288e-07, 3.6769e-06, -1.6261e-06, ..., 4.2934e-07, + 1.3970e-07, 6.8452e-07]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0106, -0.0096, -0.0054, -0.0153, -0.0101, -0.0019, 0.0098, 0.0105, + 0.0173, -0.0048], device='cuda:0'), grad: tensor([-5.9977e-06, -1.1265e-05, 3.1479e-07, 2.7865e-06, 5.8338e-06, + 1.2908e-06, -2.8778e-07, 3.9265e-06, -2.1756e-06, 5.5432e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 214.55, cls_loss 0.0033 cls_loss_mapping 0.0044 cls_loss_causal 0.5623 re_mapping 0.0075 re_causal 0.0228 /// teacc 98.90 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1552, 0.0684, -0.0218, ..., -0.0137, -0.0916, -0.1019], + [ 0.0273, 0.0609, -0.0933, ..., -0.0799, -0.0222, 0.0832], + [ 0.0510, -0.0778, -0.0926, ..., -0.0039, -0.0415, -0.0365], + ..., + [ 0.0621, 0.0147, 0.0963, ..., 0.0525, 0.1312, 0.0029], + [ 0.0801, -0.1164, -0.0997, ..., -0.1446, -0.0490, 0.1340], + [-0.0706, 0.0763, 0.0233, ..., -0.0959, -0.0589, -0.0113]], + device='cuda:0'), grad: tensor([[ 1.1269e-07, 3.7719e-08, 9.3132e-08, ..., -8.2422e-08, + 1.8161e-07, 1.1502e-07], + [ 1.4249e-07, 2.4438e-06, 3.8650e-07, ..., 1.5367e-07, + 2.3562e-07, -6.3982e-07], + [-7.4692e-07, 1.3988e-06, 1.3504e-07, ..., -7.8883e-07, + -7.4925e-07, 3.1525e-07], + ..., + [-2.2352e-07, 2.6003e-06, -2.9132e-06, ..., -3.4459e-07, + -1.7276e-06, 2.8685e-07], + [-3.8650e-08, 8.6501e-06, 1.5739e-07, ..., 4.6380e-07, + 4.0838e-07, -1.5479e-06], + [ 1.0477e-07, 4.2111e-05, 3.9116e-07, ..., 2.4308e-07, + 2.2911e-07, 6.1234e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0103, -0.0092, -0.0063, -0.0158, -0.0101, -0.0008, 0.0091, 0.0108, + 0.0166, -0.0049], device='cuda:0'), grad: tensor([ 1.9148e-05, 1.0580e-05, 2.0266e-06, 5.7034e-06, -3.1471e-05, + 3.0145e-05, -1.4615e-04, 9.7416e-07, 3.6478e-05, 7.2420e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 214.57, cls_loss 0.0032 cls_loss_mapping 0.0057 cls_loss_causal 0.5412 re_mapping 0.0080 re_causal 0.0228 /// teacc 98.89 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1556, 0.0682, -0.0223, ..., -0.0137, -0.0923, -0.1022], + [ 0.0280, 0.0612, -0.0923, ..., -0.0798, -0.0216, 0.0839], + [ 0.0507, -0.0780, -0.0926, ..., -0.0035, -0.0418, -0.0368], + ..., + [ 0.0621, 0.0145, 0.0960, ..., 0.0521, 0.1314, 0.0026], + [ 0.0803, -0.1162, -0.1000, ..., -0.1454, -0.0491, 0.1355], + [-0.0716, 0.0767, 0.0230, ..., -0.0965, -0.0602, -0.0120]], + device='cuda:0'), grad: tensor([[ 1.7090e-07, 6.3963e-06, 1.8999e-06, ..., 1.4948e-07, + 1.4435e-07, 7.3155e-07], + [ 3.1460e-06, 6.3372e-04, 1.5032e-04, ..., 2.7139e-06, + 4.4554e-06, 3.9041e-05], + [ 5.1782e-07, 4.6007e-06, 1.1474e-06, ..., 2.7288e-07, + -3.4925e-08, 1.3299e-06], + ..., + [-4.7088e-06, 4.4614e-05, 7.2494e-06, ..., -4.0010e-06, + -7.0520e-06, 5.4613e-06], + [-4.6305e-06, 7.3314e-06, 1.5236e-06, ..., -1.7630e-06, + -2.1467e-07, -3.7048e-06], + [ 7.6368e-07, -1.0443e-03, -2.6917e-04, ..., 8.3167e-07, + 9.6206e-07, -4.8429e-05]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0106, -0.0082, -0.0061, -0.0157, -0.0109, -0.0010, 0.0096, 0.0103, + 0.0171, -0.0052], device='cuda:0'), grad: tensor([ 1.6600e-05, 1.4591e-03, 1.4253e-05, 5.3197e-05, 5.8699e-04, + 4.8518e-05, 4.1910e-06, 1.3041e-04, 1.4260e-05, -2.3270e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 214.70, cls_loss 0.0025 cls_loss_mapping 0.0051 cls_loss_causal 0.5730 re_mapping 0.0075 re_causal 0.0229 /// teacc 98.91 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1559, 0.0686, -0.0227, ..., -0.0138, -0.0927, -0.1026], + [ 0.0279, 0.0613, -0.0925, ..., -0.0800, -0.0216, 0.0846], + [ 0.0514, -0.0781, -0.0918, ..., -0.0028, -0.0409, -0.0368], + ..., + [ 0.0618, 0.0144, 0.0961, ..., 0.0517, 0.1312, 0.0024], + [ 0.0804, -0.1170, -0.1002, ..., -0.1462, -0.0492, 0.1358], + [-0.0720, 0.0766, 0.0231, ..., -0.0973, -0.0610, -0.0123]], + device='cuda:0'), grad: tensor([[ 2.7940e-07, -1.6671e-07, 2.8778e-07, ..., 9.4995e-08, + 1.4249e-07, 6.0163e-07], + [ 3.4682e-06, 7.5996e-06, 3.5129e-06, ..., 1.5516e-06, + 2.6524e-06, 9.2536e-06], + [ 9.5442e-06, 3.8650e-07, 2.1085e-06, ..., 8.1956e-07, + 1.3569e-06, 8.1658e-06], + ..., + [-1.3933e-05, -2.3730e-06, -1.8269e-05, ..., -7.9498e-06, + -1.3918e-05, -2.1197e-06], + [-2.2292e-05, 3.2298e-06, 2.6375e-06, ..., 1.0757e-06, + 1.7192e-06, -1.8269e-05], + [ 3.5632e-06, -1.5423e-06, 2.7940e-07, ..., 1.4910e-06, + 2.5462e-06, 2.2780e-06]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0105, -0.0083, -0.0053, -0.0141, -0.0109, -0.0023, 0.0096, 0.0100, + 0.0168, -0.0054], device='cuda:0'), grad: tensor([ 2.6785e-06, 4.1455e-05, 2.0742e-05, 2.1994e-05, 5.3085e-07, + 6.1214e-05, -9.7215e-05, -2.0832e-05, -3.2932e-05, 2.2985e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 214.47, cls_loss 0.0025 cls_loss_mapping 0.0049 cls_loss_causal 0.5270 re_mapping 0.0075 re_causal 0.0223 /// teacc 98.83 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1566, 0.0688, -0.0230, ..., -0.0138, -0.0930, -0.1033], + [ 0.0279, 0.0620, -0.0928, ..., -0.0800, -0.0216, 0.0851], + [ 0.0514, -0.0781, -0.0921, ..., -0.0028, -0.0409, -0.0374], + ..., + [ 0.0622, 0.0139, 0.0966, ..., 0.0519, 0.1315, 0.0026], + [ 0.0810, -0.1179, -0.1010, ..., -0.1467, -0.0495, 0.1368], + [-0.0725, 0.0767, 0.0237, ..., -0.0979, -0.0613, -0.0126]], + device='cuda:0'), grad: tensor([[ 5.0291e-07, -3.5558e-06, 5.4948e-08, ..., -5.8487e-07, + 3.4552e-07, 3.8184e-08], + [ 1.1344e-06, 2.0489e-07, 1.4715e-07, ..., 1.7518e-06, + 1.0990e-06, 1.8068e-07], + [-9.9614e-06, 6.7614e-07, -1.8626e-07, ..., -1.4283e-05, + -5.4613e-06, 5.0291e-08], + ..., + [ 1.1055e-06, 1.3690e-06, 8.8103e-07, ..., 1.8422e-06, + 1.3430e-06, 6.1002e-07], + [ 4.6752e-07, 4.3493e-07, 2.3842e-07, ..., 1.2256e-06, + 4.6659e-07, -9.9465e-07], + [ 3.8184e-07, -4.1053e-06, -8.3596e-06, ..., 8.9314e-07, + 2.3190e-07, 9.8255e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0104, -0.0082, -0.0053, -0.0143, -0.0107, -0.0022, 0.0079, 0.0101, + 0.0174, -0.0052], device='cuda:0'), grad: tensor([-6.4522e-06, 5.5768e-06, -4.0919e-05, 2.4915e-05, 8.5831e-06, + -3.6620e-06, 4.8392e-06, 8.1137e-06, 2.9728e-06, -4.0010e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 214.40, cls_loss 0.0031 cls_loss_mapping 0.0045 cls_loss_causal 0.5260 re_mapping 0.0074 re_causal 0.0220 /// teacc 98.99 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.1569, 0.0689, -0.0237, ..., -0.0139, -0.0934, -0.1037], + [ 0.0273, 0.0621, -0.0939, ..., -0.0813, -0.0225, 0.0852], + [ 0.0512, -0.0790, -0.0934, ..., -0.0038, -0.0417, -0.0378], + ..., + [ 0.0633, 0.0141, 0.0983, ..., 0.0540, 0.1332, 0.0028], + [ 0.0810, -0.1184, -0.1014, ..., -0.1478, -0.0496, 0.1370], + [-0.0732, 0.0767, 0.0232, ..., -0.0996, -0.0621, -0.0128]], + device='cuda:0'), grad: tensor([[ 2.8592e-07, -5.5939e-05, -1.8075e-05, ..., 1.3607e-06, + 1.1157e-06, 2.0582e-07], + [ 1.8887e-06, -3.2317e-06, 4.4610e-07, ..., 2.0228e-06, + 1.9912e-06, -8.3521e-06], + [-3.6895e-05, 2.0172e-06, -1.2875e-05, ..., -5.4777e-05, + -5.8830e-05, 1.8328e-06], + ..., + [ 2.4959e-05, -3.5197e-05, -7.7009e-05, ..., 2.7329e-05, + -8.2433e-05, 6.7335e-07], + [ 4.0680e-06, 1.4622e-06, 8.8010e-07, ..., 3.6247e-06, + 3.3304e-06, -6.7987e-07], + [ 1.0598e-06, 3.5375e-05, 1.1690e-05, ..., 1.8161e-06, + 3.3714e-06, -5.0198e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0105, -0.0088, -0.0060, -0.0143, -0.0107, -0.0012, 0.0079, 0.0114, + 0.0160, -0.0054], device='cuda:0'), grad: tensor([-1.1617e-04, -7.3612e-06, -1.1706e-04, 2.3007e-05, 1.9038e-04, + 1.2785e-05, 1.9237e-05, -9.7990e-05, 1.4000e-05, 7.8857e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 214.70, cls_loss 0.0038 cls_loss_mapping 0.0059 cls_loss_causal 0.5834 re_mapping 0.0071 re_causal 0.0223 /// teacc 98.92 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.1581, 0.0686, -0.0247, ..., -0.0135, -0.0947, -0.1046], + [ 0.0267, 0.0625, -0.0949, ..., -0.0825, -0.0232, 0.0850], + [ 0.0508, -0.0796, -0.0943, ..., -0.0036, -0.0422, -0.0381], + ..., + [ 0.0648, 0.0140, 0.0999, ..., 0.0557, 0.1347, 0.0036], + [ 0.0814, -0.1196, -0.1018, ..., -0.1492, -0.0498, 0.1379], + [-0.0740, 0.0793, 0.0242, ..., -0.1009, -0.0625, -0.0108]], + device='cuda:0'), grad: tensor([[ 6.0257e-07, -4.2655e-06, 2.1905e-06, ..., 6.6776e-07, + 5.4017e-08, 2.8498e-07], + [-1.3169e-06, -2.0325e-05, 3.7625e-07, ..., -3.1944e-07, + 5.0291e-08, -3.7074e-05], + [ 5.9567e-06, 1.1802e-05, 1.1865e-06, ..., 1.6904e-06, + 2.6822e-07, 2.6375e-05], + ..., + [-2.1048e-06, 2.7008e-06, -7.8827e-06, ..., -4.2208e-06, + -3.0175e-07, 3.0566e-06], + [-9.1940e-06, 1.6429e-06, 5.4110e-07, ..., -1.2098e-06, + -3.6135e-07, -9.0301e-06], + [ 2.5183e-06, -2.2262e-05, -5.0783e-05, ..., -1.0453e-05, + 1.0710e-07, 5.6140e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0108, -0.0095, -0.0061, -0.0148, -0.0130, -0.0016, 0.0073, 0.0128, + 0.0169, -0.0035], device='cuda:0'), grad: tensor([-3.6322e-06, -8.4579e-05, 6.3002e-05, 6.6996e-05, 1.0140e-05, + 2.0415e-05, -3.0294e-05, -4.4443e-06, -8.1211e-06, -2.9564e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 147---------------------------------------------------- +epoch 147, time 230.51, cls_loss 0.0028 cls_loss_mapping 0.0051 cls_loss_causal 0.5440 re_mapping 0.0071 re_causal 0.0214 /// teacc 99.02 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1587, 0.0694, -0.0249, ..., -0.0129, -0.0951, -0.1053], + [ 0.0276, 0.0633, -0.0943, ..., -0.0825, -0.0224, 0.0864], + [ 0.0505, -0.0806, -0.0947, ..., -0.0038, -0.0425, -0.0390], + ..., + [ 0.0643, 0.0130, 0.0998, ..., 0.0560, 0.1344, 0.0025], + [ 0.0819, -0.1205, -0.1020, ..., -0.1501, -0.0499, 0.1389], + [-0.0748, 0.0785, 0.0241, ..., -0.1022, -0.0629, -0.0120]], + device='cuda:0'), grad: tensor([[ 1.6578e-07, -4.8019e-06, 1.0990e-07, ..., -2.0135e-06, + 1.6764e-07, 2.7008e-08], + [ 2.8964e-07, 1.9651e-07, 2.1141e-07, ..., 8.9128e-07, + 1.9465e-07, -5.7742e-07], + [-2.3227e-06, 5.4296e-07, -8.4378e-07, ..., -5.1409e-06, + -1.8906e-06, 3.0734e-07], + ..., + [ 3.3993e-07, 4.5355e-07, -6.0443e-07, ..., 1.6056e-06, + 2.1420e-07, 5.0198e-07], + [ 4.6007e-07, 1.1073e-06, 3.2969e-07, ..., 1.5693e-06, + 4.2468e-07, 3.2596e-08], + [ 1.4249e-07, 1.8161e-06, 9.0338e-08, ..., 1.0235e-06, + 1.4063e-07, 1.7136e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0102, -0.0081, -0.0067, -0.0151, -0.0120, -0.0012, 0.0072, 0.0118, + 0.0171, -0.0046], device='cuda:0'), grad: tensor([-1.5289e-05, 1.3290e-06, -6.1616e-06, 3.5726e-06, -1.2042e-06, + 1.4901e-06, 8.6706e-07, 3.6322e-06, 5.6811e-06, 6.0461e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 214.66, cls_loss 0.0031 cls_loss_mapping 0.0047 cls_loss_causal 0.5254 re_mapping 0.0076 re_causal 0.0216 /// teacc 98.97 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1596, 0.0697, -0.0248, ..., -0.0120, -0.0956, -0.1061], + [ 0.0274, 0.0635, -0.0947, ..., -0.0831, -0.0224, 0.0867], + [ 0.0504, -0.0809, -0.0951, ..., -0.0037, -0.0425, -0.0397], + ..., + [ 0.0638, 0.0129, 0.0999, ..., 0.0565, 0.1344, 0.0014], + [ 0.0840, -0.1207, -0.1010, ..., -0.1492, -0.0490, 0.1417], + [-0.0755, 0.0782, 0.0242, ..., -0.1035, -0.0632, -0.0126]], + device='cuda:0'), grad: tensor([[ 5.3272e-07, -1.1218e-04, 1.2089e-06, ..., -2.4602e-05, + 2.0023e-07, 1.1465e-06], + [ 2.6952e-06, 6.9737e-06, 4.5113e-06, ..., 6.0312e-06, + 1.5022e-06, 6.4820e-06], + [-1.8645e-06, 2.8703e-06, 7.3537e-06, ..., 2.6841e-06, + -9.7882e-07, 2.2370e-06], + ..., + [-8.8155e-05, 1.4439e-05, -8.9884e-05, ..., -5.4777e-05, + -3.5256e-05, -3.1572e-06], + [ 4.6313e-05, 1.8299e-05, 5.8174e-05, ..., 3.6120e-05, + 1.8373e-05, 1.5497e-05], + [ 1.5637e-06, 3.5620e-04, -2.2221e-06, ..., 5.7742e-06, + 7.8045e-07, 2.4438e-04]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0104, -0.0083, -0.0068, -0.0149, -0.0118, -0.0019, 0.0085, 0.0115, + 0.0186, -0.0052], device='cuda:0'), grad: tensor([-4.3988e-04, 3.4988e-05, 1.8343e-05, 5.9642e-06, -1.2760e-03, + 2.7150e-05, 4.6182e-04, -1.2481e-04, 1.6856e-04, 1.1234e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 214.43, cls_loss 0.0028 cls_loss_mapping 0.0041 cls_loss_causal 0.5267 re_mapping 0.0075 re_causal 0.0219 /// teacc 99.00 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1605, 0.0698, -0.0253, ..., -0.0118, -0.0961, -0.1066], + [ 0.0257, 0.0637, -0.0951, ..., -0.0836, -0.0236, 0.0852], + [ 0.0498, -0.0812, -0.0955, ..., -0.0037, -0.0429, -0.0401], + ..., + [ 0.0651, 0.0124, 0.1000, ..., 0.0567, 0.1359, 0.0022], + [ 0.0855, -0.1209, -0.1006, ..., -0.1505, -0.0492, 0.1431], + [-0.0759, 0.0781, 0.0243, ..., -0.1040, -0.0635, -0.0130]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, -6.3702e-07, 4.6566e-09, ..., -1.1362e-07, + 8.6613e-08, 1.3039e-08], + [-2.2188e-05, -8.7544e-08, 2.0489e-08, ..., -3.4064e-05, + -2.7433e-05, -2.2165e-07], + [ 2.0191e-05, 4.1910e-08, 1.7695e-08, ..., 3.0026e-05, + 2.4140e-05, 8.3912e-07], + ..., + [ 1.8505e-06, 1.7509e-07, -5.5879e-09, ..., 2.8163e-06, + 2.2426e-06, 1.6019e-07], + [-8.4750e-07, 1.3225e-07, 3.5390e-08, ..., 1.4808e-07, + 6.7055e-08, -1.0272e-06], + [ 1.2387e-07, -2.1420e-08, -1.8161e-07, ..., 1.7695e-07, + 7.6368e-08, 1.0524e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0103, -0.0096, -0.0072, -0.0142, -0.0116, -0.0023, 0.0085, 0.0125, + 0.0193, -0.0054], device='cuda:0'), grad: tensor([-1.0217e-06, -1.1641e-04, 1.0335e-04, 2.8498e-06, 3.5018e-07, + 4.0606e-07, 4.3306e-07, 9.9912e-06, -5.3737e-07, 5.8860e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 214.80, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5567 re_mapping 0.0076 re_causal 0.0223 /// teacc 98.97 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1611, 0.0702, -0.0254, ..., -0.0121, -0.0964, -0.1068], + [ 0.0255, 0.0638, -0.0953, ..., -0.0838, -0.0237, 0.0854], + [ 0.0501, -0.0815, -0.0958, ..., -0.0030, -0.0429, -0.0405], + ..., + [ 0.0653, 0.0117, 0.1003, ..., 0.0567, 0.1361, 0.0021], + [ 0.0854, -0.1211, -0.1008, ..., -0.1512, -0.0494, 0.1435], + [-0.0762, 0.0779, 0.0247, ..., -0.1046, -0.0637, -0.0132]], + device='cuda:0'), grad: tensor([[ 9.3132e-08, -2.3935e-06, 3.3993e-07, ..., -3.0361e-07, + 4.6566e-09, 1.5553e-07], + [-2.0877e-05, -4.1910e-08, 4.8615e-07, ..., 2.6915e-07, + 2.0303e-07, -3.4660e-05], + [ 5.8673e-06, 4.8801e-07, 1.0040e-06, ..., 1.0058e-06, + 1.1269e-07, 9.5218e-06], + ..., + [ 7.6741e-06, 1.9558e-07, -1.2051e-06, ..., -4.1258e-07, + -6.5751e-07, 1.3746e-05], + [ 4.0159e-06, 5.6718e-07, 4.6846e-07, ..., 1.8906e-07, + 2.0489e-08, 6.1654e-06], + [ 1.9241e-06, -3.4180e-07, -7.9721e-07, ..., 6.3889e-07, + 2.8498e-07, 2.6189e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0103, -0.0097, -0.0067, -0.0155, -0.0111, -0.0014, 0.0084, 0.0124, + 0.0191, -0.0057], device='cuda:0'), grad: tensor([-3.3416e-06, -6.3956e-05, 1.9401e-05, 6.7167e-06, 4.6045e-06, + -5.8264e-06, -2.9430e-07, 2.4498e-05, 1.3039e-05, 4.9956e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 214.67, cls_loss 0.0037 cls_loss_mapping 0.0061 cls_loss_causal 0.5193 re_mapping 0.0078 re_causal 0.0219 /// teacc 98.96 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1620, 0.0706, -0.0263, ..., -0.0123, -0.0969, -0.1075], + [ 0.0254, 0.0637, -0.0952, ..., -0.0840, -0.0239, 0.0857], + [ 0.0514, -0.0817, -0.0949, ..., -0.0008, -0.0421, -0.0419], + ..., + [ 0.0636, 0.0134, 0.0990, ..., 0.0542, 0.1360, 0.0019], + [ 0.0862, -0.1218, -0.1006, ..., -0.1515, -0.0489, 0.1446], + [-0.0762, 0.0777, 0.0247, ..., -0.1051, -0.0640, -0.0134]], + device='cuda:0'), grad: tensor([[ 2.2538e-07, -9.0618e-07, 1.4156e-07, ..., 6.5193e-08, + 7.1712e-08, 1.6112e-07], + [ 4.4517e-06, -6.2678e-07, 3.9376e-06, ..., 6.6776e-07, + 1.6894e-06, -2.4270e-06], + [ 4.3958e-07, 3.1758e-07, 2.5034e-06, ..., 1.5739e-07, + 8.3726e-07, -7.7765e-07], + ..., + [-1.2279e-05, 4.0140e-07, -1.1384e-05, ..., -2.1569e-06, + -5.2825e-06, 6.7428e-07], + [ 1.8664e-06, 4.8149e-07, 1.0198e-06, ..., 6.0070e-07, + 5.1502e-07, 8.8569e-07], + [ 3.6582e-06, 1.5926e-07, 1.6913e-06, ..., 4.3400e-07, + 1.6019e-06, 2.1700e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0106, -0.0097, -0.0055, -0.0149, -0.0118, -0.0018, 0.0089, 0.0113, + 0.0201, -0.0060], device='cuda:0'), grad: tensor([-1.8859e-06, 4.4629e-06, -2.9057e-07, 1.1874e-06, -2.6152e-05, + 1.0720e-06, 2.4691e-05, -1.5058e-05, 5.2154e-06, 6.7726e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 214.67, cls_loss 0.0028 cls_loss_mapping 0.0040 cls_loss_causal 0.5416 re_mapping 0.0076 re_causal 0.0219 /// teacc 98.94 lr 0.00010000 +Epoch 154, weight, value: tensor([[-1.6294e-01, 7.1158e-02, -2.6980e-02, ..., -1.1700e-02, + -9.7284e-02, -1.0796e-01], + [ 2.4883e-02, 6.3166e-02, -9.6171e-02, ..., -8.4620e-02, + -2.4299e-02, 8.5606e-02], + [ 5.2276e-02, -8.2091e-02, -9.4249e-02, ..., 1.1465e-04, + -4.1739e-02, -4.2401e-02], + ..., + [ 6.3415e-02, 1.4235e-02, 9.9814e-02, ..., 5.3723e-02, + 1.3648e-01, 2.4348e-03], + [ 8.6179e-02, -1.2237e-01, -1.0100e-01, ..., -1.5249e-01, + -4.9038e-02, 1.4487e-01], + [-7.7673e-02, 7.7732e-02, 2.4149e-02, ..., -1.0675e-01, + -6.5447e-02, -1.3673e-02]], device='cuda:0'), grad: tensor([[ 2.3376e-07, -2.1476e-06, 9.6858e-08, ..., 2.2929e-06, + 4.2561e-07, 1.6112e-07], + [ 1.1493e-06, -3.3993e-07, 1.3411e-07, ..., 1.1967e-06, + 4.5914e-07, -4.4145e-07], + [-2.0549e-05, 1.7136e-07, -9.0338e-07, ..., -2.4915e-05, + -6.7018e-06, 6.1374e-07], + ..., + [ 1.0736e-05, 4.1071e-07, 2.8312e-07, ..., 6.3516e-06, + 2.5686e-06, 1.7695e-07], + [ 4.4964e-06, 1.3784e-07, 3.7346e-07, ..., 4.9509e-06, + 1.8869e-06, -2.2706e-06], + [ 7.6182e-07, 6.8732e-07, -5.6531e-07, ..., 4.0885e-07, + 1.0990e-07, 1.2238e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0103, -0.0102, -0.0048, -0.0157, -0.0119, -0.0012, 0.0090, 0.0116, + 0.0197, -0.0063], device='cuda:0'), grad: tensor([ 1.9800e-06, 2.5667e-06, -6.4731e-05, 2.3365e-05, -2.1141e-07, + 2.3860e-06, 5.3421e-06, 1.6972e-05, 8.8364e-06, 3.4813e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 153---------------------------------------------------- +epoch 153, time 231.30, cls_loss 0.0024 cls_loss_mapping 0.0041 cls_loss_causal 0.5377 re_mapping 0.0075 re_causal 0.0209 /// teacc 99.09 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1637, 0.0717, -0.0276, ..., -0.0107, -0.0975, -0.1085], + [ 0.0247, 0.0637, -0.0966, ..., -0.0853, -0.0245, 0.0860], + [ 0.0524, -0.0827, -0.0946, ..., 0.0005, -0.0419, -0.0426], + ..., + [ 0.0639, 0.0138, 0.1003, ..., 0.0540, 0.1369, 0.0025], + [ 0.0859, -0.1231, -0.1022, ..., -0.1543, -0.0493, 0.1448], + [-0.0782, 0.0774, 0.0239, ..., -0.1080, -0.0658, -0.0139]], + device='cuda:0'), grad: tensor([[ 7.8883e-07, -5.3756e-06, 1.8217e-06, ..., -2.3842e-07, + 7.6834e-07, 2.0899e-06], + [-3.8648e-04, -6.5041e-04, -9.2983e-04, ..., 5.6438e-07, + -3.7599e-04, -1.1387e-03], + [-3.1497e-06, 1.5935e-06, 1.4622e-06, ..., -2.2948e-06, + -3.4999e-06, 2.3637e-06], + ..., + [ 3.6025e-04, 6.0177e-04, 8.5974e-04, ..., 1.4538e-06, + 3.5095e-04, 1.0529e-03], + [ 2.8498e-07, 1.0179e-06, 8.8569e-07, ..., 1.4435e-07, + 3.8184e-07, 4.6752e-07], + [ 1.5348e-05, 2.7597e-05, 3.5644e-05, ..., 3.0641e-07, + 1.4871e-05, 4.4793e-05]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0100, -0.0102, -0.0042, -0.0162, -0.0115, -0.0015, 0.0097, 0.0118, + 0.0188, -0.0067], device='cuda:0'), grad: tensor([-1.2584e-05, -1.8759e-03, -3.7868e-06, 2.2743e-06, 5.4479e-05, + 3.7327e-06, 5.7109e-06, 1.7424e-03, 2.3134e-06, 8.0764e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 214.46, cls_loss 0.0020 cls_loss_mapping 0.0036 cls_loss_causal 0.5628 re_mapping 0.0073 re_causal 0.0221 /// teacc 98.92 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1650, 0.0721, -0.0285, ..., -0.0106, -0.0980, -0.1095], + [ 0.0252, 0.0650, -0.0957, ..., -0.0854, -0.0241, 0.0873], + [ 0.0524, -0.0831, -0.0948, ..., 0.0005, -0.0420, -0.0432], + ..., + [ 0.0638, 0.0131, 0.1002, ..., 0.0542, 0.1369, 0.0016], + [ 0.0860, -0.1232, -0.1025, ..., -0.1548, -0.0495, 0.1450], + [-0.0787, 0.0772, 0.0238, ..., -0.1087, -0.0662, -0.0143]], + device='cuda:0'), grad: tensor([[ 5.1782e-07, 1.0896e-07, 2.6543e-07, ..., 1.2387e-07, + 3.4180e-07, 5.4110e-07], + [ 4.2543e-06, -1.7695e-08, 5.0291e-07, ..., 2.9709e-07, + 3.4813e-06, 3.3341e-06], + [-3.6120e-05, 2.9244e-07, 6.6403e-07, ..., -2.1886e-06, + -2.6926e-05, -3.1322e-05], + ..., + [ 8.9854e-06, 1.8068e-07, -1.0142e-06, ..., 2.8033e-07, + 6.1654e-06, 9.1344e-06], + [ 1.9535e-05, 1.1772e-06, 1.7863e-06, ..., 2.3991e-06, + 1.5602e-05, 1.6227e-05], + [ 9.8161e-07, 3.5428e-06, 6.1877e-06, ..., 3.0864e-06, + 3.0268e-07, 1.5246e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0101, -0.0094, -0.0043, -0.0162, -0.0114, -0.0017, 0.0100, 0.0115, + 0.0188, -0.0071], device='cuda:0'), grad: tensor([ 1.5255e-06, 9.4920e-06, -7.7426e-05, -1.7792e-05, 1.3951e-06, + 1.3430e-06, 8.8941e-07, 1.9982e-05, 4.6879e-05, 1.3679e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 214.74, cls_loss 0.0029 cls_loss_mapping 0.0049 cls_loss_causal 0.5499 re_mapping 0.0078 re_causal 0.0219 /// teacc 98.95 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1660, 0.0723, -0.0291, ..., -0.0109, -0.0985, -0.1100], + [ 0.0258, 0.0673, -0.0957, ..., -0.0862, -0.0243, 0.0889], + [ 0.0520, -0.0838, -0.0954, ..., 0.0002, -0.0429, -0.0435], + ..., + [ 0.0639, 0.0112, 0.1009, ..., 0.0550, 0.1379, 0.0002], + [ 0.0865, -0.1223, -0.1029, ..., -0.1563, -0.0494, 0.1462], + [-0.0797, 0.0772, 0.0237, ..., -0.1093, -0.0668, -0.0145]], + device='cuda:0'), grad: tensor([[ 1.1642e-07, -2.1886e-07, 6.0629e-07, ..., 1.3597e-07, + 8.4750e-08, 6.0536e-08], + [ 9.7230e-07, 1.4165e-06, 1.8217e-06, ..., 6.2399e-07, + 9.8348e-07, 8.0094e-08], + [ 3.2783e-07, 1.4780e-06, 2.4140e-06, ..., 1.7229e-07, + 8.7544e-08, 2.5332e-07], + ..., + [-2.1085e-06, 2.0992e-06, -2.8368e-06, ..., -1.0664e-06, + -2.2929e-06, 8.1025e-07], + [-9.8273e-06, 4.6417e-06, 4.1798e-06, ..., 2.1327e-07, + 2.0023e-07, -4.5411e-06], + [ 5.3644e-07, -1.6019e-05, -1.2830e-05, ..., 4.4424e-07, + 3.7439e-07, -4.7404e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0100, -0.0086, -0.0048, -0.0166, -0.0113, -0.0022, 0.0100, 0.0114, + 0.0197, -0.0073], device='cuda:0'), grad: tensor([-9.1270e-07, 5.6773e-06, 4.2170e-06, 2.0102e-05, -4.1164e-07, + 1.2495e-05, 2.9519e-05, 1.7108e-06, -3.7432e-05, -3.4899e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 214.65, cls_loss 0.0025 cls_loss_mapping 0.0063 cls_loss_causal 0.5276 re_mapping 0.0076 re_causal 0.0218 /// teacc 98.95 lr 0.00010000 +Epoch 158, weight, value: tensor([[-1.6648e-01, 7.2482e-02, -2.9594e-02, ..., -1.1325e-02, + -9.8867e-02, -1.1017e-01], + [ 2.6267e-02, 6.8086e-02, -9.6103e-02, ..., -8.7093e-02, + -2.4636e-02, 9.0681e-02], + [ 5.1938e-02, -8.4105e-02, -9.6005e-02, ..., -5.4659e-06, + -4.3686e-02, -4.3579e-02], + ..., + [ 6.4309e-02, 1.1079e-02, 1.0168e-01, ..., 5.5606e-02, + 1.3898e-01, 1.5304e-04], + [ 8.5047e-02, -1.2398e-01, -1.0392e-01, ..., -1.5903e-01, + -4.9659e-02, 1.4417e-01], + [-8.0093e-02, 7.7007e-02, 2.3471e-02, ..., -1.1011e-01, + -6.7483e-02, -1.4686e-02]], device='cuda:0'), grad: tensor([[ 2.1979e-06, -3.3006e-06, 1.0058e-06, ..., -2.5984e-07, + 1.6093e-06, 1.1362e-06], + [ 1.0881e-03, -3.6601e-07, 7.7391e-04, ..., 3.1944e-07, + 7.8917e-04, 5.6410e-04], + [ 3.0160e-05, 2.3842e-07, 2.2665e-05, ..., 6.8173e-07, + 2.1607e-05, 1.5482e-05], + ..., + [-1.1930e-03, 9.8627e-07, -8.4925e-04, ..., -1.8496e-06, + -8.6498e-04, -6.1750e-04], + [ 5.3197e-05, 3.3062e-07, 3.8654e-05, ..., 1.7975e-07, + 3.9428e-05, 2.6479e-05], + [ 9.1717e-06, 1.4156e-06, 5.8822e-06, ..., 4.5821e-07, + 6.5006e-06, 5.2005e-06]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0100, -0.0076, -0.0050, -0.0167, -0.0111, -0.0015, 0.0102, 0.0119, + 0.0172, -0.0076], device='cuda:0'), grad: tensor([-2.5854e-06, 2.1343e-03, 5.8860e-05, 1.4573e-05, 9.2108e-07, + 1.5022e-06, 2.0601e-06, -2.3365e-03, 1.0425e-04, 2.1592e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 214.28, cls_loss 0.0040 cls_loss_mapping 0.0054 cls_loss_causal 0.5501 re_mapping 0.0078 re_causal 0.0209 /// teacc 98.94 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1674, 0.0713, -0.0301, ..., -0.0108, -0.0996, -0.1110], + [ 0.0260, 0.0686, -0.0966, ..., -0.0876, -0.0250, 0.0918], + [ 0.0522, -0.0849, -0.0959, ..., 0.0003, -0.0433, -0.0441], + ..., + [ 0.0647, 0.0103, 0.1021, ..., 0.0554, 0.1394, -0.0002], + [ 0.0845, -0.1257, -0.1043, ..., -0.1599, -0.0499, 0.1435], + [-0.0810, 0.0782, 0.0233, ..., -0.1116, -0.0680, -0.0146]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, -3.5614e-06, 2.9802e-08, ..., -8.1677e-07, + 1.3970e-08, 1.0710e-07], + [-5.5395e-06, 6.9216e-06, 4.0513e-07, ..., -3.2149e-06, + -4.9248e-06, -2.5854e-06], + [ 4.9472e-06, 4.9733e-07, 1.0524e-07, ..., 2.0918e-06, + 5.0478e-06, 8.3596e-06], + ..., + [-1.0990e-06, 8.0932e-07, -2.6263e-06, ..., -1.8124e-06, + -1.7295e-06, 8.5589e-07], + [ 3.0734e-07, 1.2666e-06, 1.6987e-06, ..., 8.8848e-07, + 4.5635e-08, 3.8836e-07], + [ 8.0094e-08, 2.2426e-06, -1.9595e-06, ..., 3.6694e-07, + 6.4261e-08, 6.5751e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0113, -0.0074, -0.0051, -0.0170, -0.0111, -0.0013, 0.0108, 0.0118, + 0.0167, -0.0067], device='cuda:0'), grad: tensor([-6.6049e-06, -2.6412e-06, 1.8999e-05, 3.5912e-06, -2.5302e-05, + 1.4091e-06, 2.0005e-06, -1.0058e-07, 5.8748e-06, 2.7884e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 214.69, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.5370 re_mapping 0.0075 re_causal 0.0218 /// teacc 98.90 lr 0.00010000 +Epoch 160, weight, value: tensor([[-1.6793e-01, 7.1738e-02, -3.0766e-02, ..., -9.9862e-03, + -9.9903e-02, -1.1138e-01], + [ 2.5850e-02, 6.7588e-02, -9.7046e-02, ..., -8.8755e-02, + -2.5160e-02, 9.1904e-02], + [ 5.2039e-02, -8.4225e-02, -9.6477e-02, ..., 7.0035e-05, + -4.3605e-02, -4.4739e-02], + ..., + [ 6.5010e-02, 1.0158e-02, 1.0269e-01, ..., 5.5871e-02, + 1.3984e-01, -2.4710e-04], + [ 8.4581e-02, -1.2606e-01, -1.0469e-01, ..., -1.6075e-01, + -5.0027e-02, 1.4359e-01], + [-8.1196e-02, 7.7011e-02, 2.3064e-02, ..., -1.1242e-01, + -6.8750e-02, -1.5213e-02]], device='cuda:0'), grad: tensor([[ 3.9488e-07, -3.8557e-07, 6.6124e-08, ..., -1.2293e-07, + 1.3784e-07, 2.8219e-07], + [ 2.0657e-06, 3.3621e-07, 5.8487e-07, ..., 9.9279e-07, + 2.5891e-07, 1.1455e-06], + [-5.1335e-06, 1.0869e-06, 2.8405e-07, ..., -3.3937e-06, + -6.0070e-07, 5.1502e-07], + ..., + [-1.8999e-07, 3.4366e-07, -1.1679e-06, ..., -4.9453e-07, + -6.1002e-07, 2.3562e-07], + [-2.8219e-06, 2.7008e-07, 1.1176e-08, ..., 1.8980e-06, + 4.3213e-07, -5.8301e-06], + [ 7.0035e-07, 3.0901e-06, 7.1712e-08, ..., 5.5507e-07, + 7.4506e-08, 1.5385e-06]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0109, -0.0084, -0.0044, -0.0159, -0.0095, -0.0022, 0.0101, 0.0120, + 0.0166, -0.0077], device='cuda:0'), grad: tensor([ 2.6226e-06, 8.1882e-06, -1.8135e-05, 4.5337e-06, -4.9695e-06, + 2.5764e-05, -2.3544e-05, 8.8941e-07, -5.2452e-06, 9.7901e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 214.89, cls_loss 0.0031 cls_loss_mapping 0.0047 cls_loss_causal 0.5505 re_mapping 0.0074 re_causal 0.0207 /// teacc 98.97 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1689, 0.0718, -0.0316, ..., -0.0105, -0.1006, -0.1120], + [ 0.0271, 0.0698, -0.0970, ..., -0.0893, -0.0240, 0.0928], + [ 0.0519, -0.0845, -0.0966, ..., 0.0002, -0.0438, -0.0453], + ..., + [ 0.0644, 0.0081, 0.1031, ..., 0.0562, 0.1394, -0.0013], + [ 0.0847, -0.1267, -0.1051, ..., -0.1616, -0.0502, 0.1437], + [-0.0821, 0.0749, 0.0219, ..., -0.1142, -0.0704, -0.0163]], + device='cuda:0'), grad: tensor([[ 3.7346e-07, -5.1409e-07, 2.7511e-06, ..., 2.0042e-06, + 4.8522e-07, 1.3039e-08], + [ 1.8431e-06, 3.8557e-06, 4.8019e-06, ..., 1.8664e-06, + 2.5872e-06, -3.0268e-07], + [-1.6764e-07, 3.1013e-06, 5.3309e-06, ..., 3.4552e-06, + -6.4261e-08, 7.7300e-08], + ..., + [-3.9749e-06, -4.1425e-06, -4.8988e-06, ..., -4.5542e-07, + -5.4687e-06, 1.8533e-07], + [ 4.3958e-07, 3.0845e-06, 3.1646e-06, ..., 1.5013e-06, + 7.0781e-08, 2.5611e-07], + [ 1.0384e-06, -3.5428e-06, -4.1910e-06, ..., 8.4285e-07, + 1.4175e-06, -8.1025e-08]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0110, -0.0071, -0.0046, -0.0162, -0.0070, -0.0015, 0.0092, 0.0112, + 0.0162, -0.0098], device='cuda:0'), grad: tensor([ 2.9616e-06, 1.1161e-05, 1.1846e-05, -2.7776e-05, 1.1474e-05, + -4.2059e-06, -4.6864e-06, -8.6650e-06, 1.3985e-05, -6.0946e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 214.56, cls_loss 0.0027 cls_loss_mapping 0.0045 cls_loss_causal 0.5481 re_mapping 0.0073 re_causal 0.0204 /// teacc 98.85 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1700, 0.0718, -0.0323, ..., -0.0107, -0.1011, -0.1126], + [ 0.0267, 0.0700, -0.0973, ..., -0.0898, -0.0246, 0.0930], + [ 0.0517, -0.0853, -0.0970, ..., 0.0004, -0.0442, -0.0463], + ..., + [ 0.0652, 0.0079, 0.1036, ..., 0.0564, 0.1403, -0.0009], + [ 0.0845, -0.1272, -0.1056, ..., -0.1622, -0.0507, 0.1437], + [-0.0826, 0.0749, 0.0223, ..., -0.1146, -0.0706, -0.0165]], + device='cuda:0'), grad: tensor([[ 4.3120e-07, 1.2284e-06, 2.3469e-07, ..., 1.5246e-06, + 5.5693e-07, 4.6566e-09], + [ 8.1304e-07, 4.8429e-08, 6.5193e-07, ..., 9.4716e-07, + 1.1632e-06, -1.1083e-07], + [-5.6438e-06, 9.1270e-08, -2.7828e-06, ..., -6.0089e-06, + -8.4415e-06, 3.6322e-08], + ..., + [ 1.0980e-06, 1.4715e-07, 4.5914e-07, ..., 1.2573e-06, + 1.6931e-06, 6.0536e-08], + [ 1.1511e-06, 4.8149e-07, 7.1060e-07, ..., 1.2293e-06, + 1.5385e-06, -3.2596e-08], + [ 1.3784e-07, 1.3877e-07, 1.6205e-07, ..., 3.4552e-07, + 1.6950e-07, 1.4901e-08]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0111, -0.0071, -0.0050, -0.0165, -0.0069, -0.0015, 0.0092, 0.0118, + 0.0160, -0.0098], device='cuda:0'), grad: tensor([ 7.6964e-06, 3.5949e-06, -2.4378e-05, 1.1265e-05, 4.6287e-07, + -1.0721e-05, -3.6769e-06, 5.7034e-06, 8.7619e-06, 1.2759e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 214.67, cls_loss 0.0016 cls_loss_mapping 0.0037 cls_loss_causal 0.5458 re_mapping 0.0069 re_causal 0.0213 /// teacc 98.97 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1703, 0.0726, -0.0325, ..., -0.0106, -0.1013, -0.1130], + [ 0.0265, 0.0697, -0.0977, ..., -0.0902, -0.0248, 0.0929], + [ 0.0518, -0.0857, -0.0970, ..., 0.0005, -0.0440, -0.0464], + ..., + [ 0.0653, 0.0078, 0.1038, ..., 0.0564, 0.1405, -0.0009], + [ 0.0845, -0.1277, -0.1060, ..., -0.1626, -0.0508, 0.1438], + [-0.0827, 0.0748, 0.0231, ..., -0.1150, -0.0706, -0.0165]], + device='cuda:0'), grad: tensor([[ 3.1386e-07, 1.0617e-07, 2.7288e-07, ..., 2.7657e-05, + 3.3528e-07, 5.3458e-07], + [-1.1347e-05, -5.3383e-06, -5.7071e-06, ..., 6.0722e-07, + -1.2577e-05, -2.8253e-05], + [ 1.7164e-06, 8.8662e-07, 2.2743e-06, ..., 3.0696e-05, + 1.5926e-06, 5.9307e-06], + ..., + [ 7.6294e-06, 4.1276e-06, 2.7604e-06, ..., -3.5111e-07, + 8.0839e-06, 1.9908e-05], + [-1.0990e-07, 5.6718e-07, 7.5344e-07, ..., 2.2631e-06, + 3.6322e-07, -3.6880e-07], + [ 7.2550e-07, -3.1199e-07, -7.0315e-07, ..., 6.1281e-07, + 5.4389e-07, 9.0804e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0104, -0.0074, -0.0048, -0.0167, -0.0066, -0.0012, 0.0084, 0.0117, + 0.0157, -0.0099], device='cuda:0'), grad: tensor([ 4.6164e-05, -4.8578e-05, 5.9366e-05, -1.1116e-04, 1.7779e-06, + 8.5160e-06, 4.4592e-06, 3.4839e-05, 4.2357e-06, 4.6566e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 214.83, cls_loss 0.0031 cls_loss_mapping 0.0052 cls_loss_causal 0.5354 re_mapping 0.0069 re_causal 0.0201 /// teacc 98.75 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1708, 0.0727, -0.0332, ..., -0.0109, -0.1018, -0.1138], + [ 0.0268, 0.0696, -0.0964, ..., -0.0897, -0.0246, 0.0932], + [ 0.0519, -0.0862, -0.0971, ..., 0.0005, -0.0438, -0.0467], + ..., + [ 0.0651, 0.0075, 0.1028, ..., 0.0559, 0.1403, -0.0013], + [ 0.0847, -0.1284, -0.1063, ..., -0.1632, -0.0509, 0.1440], + [-0.0855, 0.0737, 0.0213, ..., -0.1164, -0.0713, -0.0175]], + device='cuda:0'), grad: tensor([[ 3.7905e-07, -3.9022e-07, 4.9733e-07, ..., 2.0117e-07, + 2.8126e-07, 4.5169e-07], + [ 9.2462e-06, 2.0899e-06, 6.6049e-06, ..., 4.1388e-06, + 6.7018e-06, -2.7213e-06], + [ 1.5222e-05, 2.2426e-06, 9.8273e-06, ..., 7.0818e-06, + 1.2025e-05, 6.4224e-06], + ..., + [-6.4969e-05, 2.5537e-06, -3.8832e-05, ..., -2.0102e-05, + -4.1068e-05, -1.5959e-05], + [ 2.8253e-05, 1.1725e-06, 1.5885e-05, ..., 5.4464e-06, + 1.3746e-05, 1.0602e-05], + [ 1.4529e-06, -2.8536e-06, -6.7391e-06, ..., 3.6601e-07, + 7.3668e-07, 4.8317e-06]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0107, -0.0069, -0.0048, -0.0160, -0.0046, -0.0015, 0.0099, 0.0110, + 0.0154, -0.0118], device='cuda:0'), grad: tensor([ 3.8091e-07, 1.3463e-05, 2.8774e-05, -8.8364e-06, -5.1349e-05, + 3.5167e-05, 5.4836e-06, -7.3791e-05, 3.6240e-05, 1.4424e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 214.71, cls_loss 0.0027 cls_loss_mapping 0.0039 cls_loss_causal 0.5417 re_mapping 0.0070 re_causal 0.0199 /// teacc 98.99 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1718, 0.0729, -0.0340, ..., -0.0111, -0.1025, -0.1147], + [ 0.0261, 0.0694, -0.0969, ..., -0.0903, -0.0249, 0.0926], + [ 0.0518, -0.0867, -0.0974, ..., 0.0004, -0.0438, -0.0471], + ..., + [ 0.0655, 0.0068, 0.1029, ..., 0.0564, 0.1408, -0.0012], + [ 0.0849, -0.1289, -0.1068, ..., -0.1639, -0.0511, 0.1443], + [-0.0851, 0.0735, 0.0229, ..., -0.1169, -0.0706, -0.0178]], + device='cuda:0'), grad: tensor([[ 3.4412e-07, -2.7288e-07, 4.1258e-07, ..., 1.1316e-07, + 5.1688e-08, 2.1886e-08], + [ 4.4750e-07, 3.3248e-07, 9.3877e-07, ..., 3.0082e-07, + 2.4028e-07, 5.8673e-08], + [-5.1921e-07, 9.9652e-08, 2.9569e-07, ..., -7.1665e-07, + -1.6764e-07, 6.2399e-08], + ..., + [-5.1819e-06, -6.3702e-07, -8.3521e-06, ..., -2.6133e-06, + -2.6617e-06, 1.3504e-08], + [ 1.4529e-07, 4.2887e-07, 4.6939e-07, ..., 3.9395e-07, + 2.3004e-07, -7.2224e-07], + [ 1.8943e-06, -2.3305e-05, -1.2815e-05, ..., 4.4750e-07, + 1.2703e-06, 9.5926e-08]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0106, -0.0078, -0.0048, -0.0160, -0.0043, -0.0012, 0.0114, 0.0109, + 0.0151, -0.0117], device='cuda:0'), grad: tensor([-7.9162e-09, 1.4910e-06, -2.0191e-06, 2.3507e-06, 1.5058e-05, + 3.1888e-05, -1.0105e-07, -9.6560e-06, 1.5413e-06, -4.0561e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.55, cls_loss 0.0034 cls_loss_mapping 0.0042 cls_loss_causal 0.5508 re_mapping 0.0072 re_causal 0.0200 /// teacc 98.96 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1725, 0.0726, -0.0349, ..., -0.0112, -0.1028, -0.1151], + [ 0.0256, 0.0689, -0.0975, ..., -0.0909, -0.0253, 0.0920], + [ 0.0519, -0.0871, -0.0976, ..., 0.0004, -0.0437, -0.0475], + ..., + [ 0.0659, 0.0060, 0.1030, ..., 0.0566, 0.1413, -0.0010], + [ 0.0855, -0.1290, -0.1076, ..., -0.1645, -0.0512, 0.1449], + [-0.0850, 0.0726, 0.0209, ..., -0.1190, -0.0713, -0.0173]], + device='cuda:0'), grad: tensor([[-2.3488e-06, -7.7114e-06, -8.5682e-08, ..., -2.6776e-07, + -7.4785e-07, 2.6543e-08], + [ 2.5518e-07, 5.0012e-07, 8.5216e-08, ..., 4.7497e-08, + 1.6158e-07, -1.1595e-06], + [-1.4948e-07, 3.3714e-07, 2.2817e-08, ..., 1.0710e-08, + -1.9791e-07, 3.3062e-08], + ..., + [ 8.2422e-08, 3.8510e-07, -9.3132e-10, ..., 4.8894e-08, + 8.2422e-08, 7.7765e-08], + [ 1.8710e-06, 5.9754e-06, 6.7428e-07, ..., 1.1362e-07, + 5.6392e-07, 3.0920e-07], + [ 3.3528e-08, -4.8503e-06, -5.0962e-06, ..., 6.4261e-08, + 4.7497e-08, -7.1106e-07]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0109, -0.0089, -0.0048, -0.0162, -0.0026, -0.0011, 0.0126, 0.0109, + 0.0155, -0.0128], device='cuda:0'), grad: tensor([-2.3246e-05, 1.3644e-07, 7.0455e-07, 1.9781e-06, 1.4499e-05, + -3.0547e-07, 1.6149e-06, 1.3402e-06, 1.8686e-05, -1.5408e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 214.69, cls_loss 0.0024 cls_loss_mapping 0.0037 cls_loss_causal 0.5292 re_mapping 0.0069 re_causal 0.0206 /// teacc 99.05 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1730, 0.0728, -0.0355, ..., -0.0113, -0.1031, -0.1156], + [ 0.0244, 0.0687, -0.0987, ..., -0.0922, -0.0269, 0.0918], + [ 0.0518, -0.0874, -0.0978, ..., 0.0005, -0.0442, -0.0478], + ..., + [ 0.0667, 0.0059, 0.1036, ..., 0.0570, 0.1428, -0.0012], + [ 0.0852, -0.1295, -0.1080, ..., -0.1652, -0.0512, 0.1447], + [-0.0847, 0.0728, 0.0214, ..., -0.1196, -0.0711, -0.0175]], + device='cuda:0'), grad: tensor([[ 8.4285e-08, -7.3910e-06, 1.0571e-07, ..., 8.1025e-08, + 2.4680e-08, 3.7253e-09], + [ 1.5087e-07, 2.6077e-08, 2.7614e-07, ..., 1.3225e-07, + 1.3551e-07, -3.3993e-08], + [-2.6077e-08, 1.9139e-07, 3.0827e-07, ..., 3.5856e-08, + -3.9581e-08, 1.7695e-08], + ..., + [-7.2690e-07, 6.9849e-07, -1.2843e-06, ..., -4.0419e-07, + -6.0862e-07, 1.4435e-08], + [ 2.1886e-08, 2.0601e-06, 1.1548e-07, ..., 8.1956e-08, + 2.7474e-08, -1.5646e-07], + [ 1.4435e-07, 9.0105e-07, -3.0128e-07, ..., 1.1828e-07, + 1.1548e-07, 4.3772e-08]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0105, -0.0099, -0.0050, -0.0157, -0.0029, -0.0016, 0.0142, 0.0113, + 0.0148, -0.0124], device='cuda:0'), grad: tensor([-2.7344e-05, 4.0093e-07, 3.8929e-07, 1.0859e-06, 1.0543e-06, + 1.7077e-05, -5.8115e-06, 1.6438e-06, 8.8513e-06, 2.6915e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 214.81, cls_loss 0.0019 cls_loss_mapping 0.0043 cls_loss_causal 0.5338 re_mapping 0.0071 re_causal 0.0206 /// teacc 98.88 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1739, 0.0729, -0.0362, ..., -0.0115, -0.1036, -0.1160], + [ 0.0240, 0.0683, -0.0992, ..., -0.0927, -0.0271, 0.0915], + [ 0.0518, -0.0880, -0.0981, ..., 0.0005, -0.0442, -0.0479], + ..., + [ 0.0671, 0.0059, 0.1046, ..., 0.0575, 0.1434, -0.0013], + [ 0.0859, -0.1303, -0.1085, ..., -0.1660, -0.0512, 0.1454], + [-0.0851, 0.0734, 0.0218, ..., -0.1204, -0.0717, -0.0174]], + device='cuda:0'), grad: tensor([[ 5.9232e-07, -3.0417e-06, 1.3458e-07, ..., 9.1270e-08, + 3.0268e-08, 1.1697e-06], + [-8.4639e-06, -4.9397e-06, 3.5092e-06, ..., 2.3004e-06, + 1.2107e-06, -1.4305e-05], + [ 3.4105e-06, 2.0117e-06, 5.5460e-07, ..., 2.9802e-07, + 2.8405e-08, 5.5879e-06], + ..., + [-7.9945e-06, 1.2061e-07, -1.4871e-05, ..., -9.6560e-06, + -5.3532e-06, -3.2745e-06], + [ 4.2580e-06, 2.4531e-06, 1.1018e-06, ..., 1.1288e-06, + 2.8964e-07, 2.1979e-05], + [ 7.2690e-07, 6.1095e-07, -1.0012e-07, ..., 2.1607e-07, + 4.2841e-08, 1.0533e-06]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0105, -0.0103, -0.0050, -0.0159, -0.0032, -0.0020, 0.0141, 0.0117, + 0.0147, -0.0118], device='cuda:0'), grad: tensor([-1.3560e-06, -3.1769e-05, 1.5900e-05, 1.2524e-05, 2.1700e-06, + 3.8117e-05, -1.7679e-04, -1.6525e-05, 1.5342e-04, 3.9153e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 214.94, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.5463 re_mapping 0.0064 re_causal 0.0198 /// teacc 98.84 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1753, 0.0741, -0.0375, ..., -0.0118, -0.1045, -0.1167], + [ 0.0249, 0.0700, -0.0989, ..., -0.0930, -0.0263, 0.0927], + [ 0.0522, -0.0885, -0.0980, ..., 0.0008, -0.0438, -0.0483], + ..., + [ 0.0663, 0.0048, 0.1042, ..., 0.0574, 0.1427, -0.0026], + [ 0.0858, -0.1318, -0.1089, ..., -0.1669, -0.0514, 0.1453], + [-0.0848, 0.0734, 0.0223, ..., -0.1213, -0.0720, -0.0174]], + device='cuda:0'), grad: tensor([[ 1.5413e-07, -8.2748e-07, 3.2596e-07, ..., 7.9162e-08, + 1.8161e-07, 3.6787e-08], + [ 2.7642e-06, 7.6881e-07, 4.5225e-06, ..., 7.6788e-07, + 3.1609e-06, 4.7497e-08], + [ 2.3842e-07, 4.6799e-07, 7.3249e-07, ..., 5.1223e-08, + 2.3469e-07, 8.7079e-08], + ..., + [-4.0941e-06, -4.1677e-07, -6.4597e-06, ..., -1.2564e-06, + -4.7609e-06, 3.1712e-07], + [-8.3167e-07, 3.6368e-07, 2.3935e-07, ..., 3.7719e-08, + 6.2864e-08, -1.1222e-06], + [ 5.5367e-07, -4.1276e-06, -2.9355e-06, ..., 1.4622e-07, + 6.1234e-07, 1.3234e-06]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0095, -0.0094, -0.0045, -0.0159, -0.0035, -0.0024, 0.0140, 0.0107, + 0.0141, -0.0116], device='cuda:0'), grad: tensor([ 5.3830e-07, 6.8285e-06, 1.4044e-06, 8.1435e-06, -2.2575e-06, + 2.9914e-06, -1.8850e-06, -7.8529e-06, -9.6764e-07, -6.9924e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 215.11, cls_loss 0.0024 cls_loss_mapping 0.0043 cls_loss_causal 0.5554 re_mapping 0.0068 re_causal 0.0199 /// teacc 98.97 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1766, 0.0753, -0.0390, ..., -0.0119, -0.1056, -0.1178], + [ 0.0248, 0.0700, -0.0991, ..., -0.0933, -0.0264, 0.0929], + [ 0.0520, -0.0886, -0.0983, ..., 0.0007, -0.0438, -0.0492], + ..., + [ 0.0659, 0.0047, 0.1039, ..., 0.0576, 0.1426, -0.0035], + [ 0.0881, -0.1319, -0.1066, ..., -0.1671, -0.0497, 0.1467], + [-0.0855, 0.0734, 0.0225, ..., -0.1225, -0.0723, -0.0177]], + device='cuda:0'), grad: tensor([[ 2.3656e-06, 1.3597e-07, 4.4424e-07, ..., 9.3132e-09, + 4.6566e-09, 3.3490e-06], + [ 3.5092e-06, 6.7055e-08, 3.8929e-07, ..., 3.6322e-08, + 3.3528e-08, 5.5917e-06], + [ 3.4943e-06, 1.2014e-07, 3.2224e-07, ..., 4.0047e-08, + 6.7055e-08, 6.2212e-06], + ..., + [ 5.9605e-07, 8.4750e-08, -8.9407e-08, ..., -5.9605e-08, + -1.2480e-07, 1.2843e-06], + [-3.1173e-05, -1.8636e-06, -4.3362e-06, ..., 2.4214e-08, + -3.3528e-08, -4.7654e-05], + [ 6.3609e-07, 9.3039e-07, -7.4506e-09, ..., 3.9116e-08, + 5.5879e-09, 1.1725e-06]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0086, -0.0098, -0.0042, -0.0148, -0.0037, -0.0043, 0.0139, 0.0103, + 0.0160, -0.0118], device='cuda:0'), grad: tensor([ 9.6038e-06, 1.4924e-05, 1.5721e-05, 2.8446e-05, 2.5202e-06, + 3.4094e-05, 1.7479e-05, 3.1106e-06, -1.3018e-04, 4.1872e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.91, cls_loss 0.0028 cls_loss_mapping 0.0045 cls_loss_causal 0.5394 re_mapping 0.0067 re_causal 0.0200 /// teacc 98.99 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1777, 0.0755, -0.0397, ..., -0.0119, -0.1061, -0.1191], + [ 0.0243, 0.0692, -0.0992, ..., -0.0935, -0.0265, 0.0916], + [ 0.0519, -0.0889, -0.0986, ..., 0.0006, -0.0438, -0.0498], + ..., + [ 0.0660, 0.0043, 0.1041, ..., 0.0577, 0.1428, -0.0036], + [ 0.0889, -0.1319, -0.1070, ..., -0.1679, -0.0497, 0.1475], + [-0.0864, 0.0736, 0.0229, ..., -0.1229, -0.0724, -0.0181]], + device='cuda:0'), grad: tensor([[ 5.4389e-07, -5.7332e-06, 2.0582e-07, ..., 1.8161e-07, + 4.7404e-07, 1.3402e-06], + [-2.3767e-06, -4.3306e-07, 1.0263e-06, ..., 9.8534e-07, + -7.3873e-06, -1.6093e-05], + [ 1.0557e-05, 6.7987e-08, 2.0280e-05, ..., -1.6065e-06, + 8.5980e-06, 6.5006e-06], + ..., + [-2.5019e-05, 1.6857e-07, -3.1412e-05, ..., -5.0776e-06, + -1.1683e-05, 4.1388e-06], + [ 6.7651e-06, 4.6100e-07, 3.7961e-06, ..., 2.6394e-06, + 3.4831e-06, 1.3951e-06], + [ 5.0291e-08, -1.5367e-06, -2.0489e-06, ..., 4.7497e-08, + 2.0862e-07, 5.2899e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0087, -0.0120, -0.0043, -0.0145, -0.0041, -0.0050, 0.0164, 0.0103, + 0.0171, -0.0117], device='cuda:0'), grad: tensor([-1.7628e-05, -4.3720e-05, 1.8775e-05, 1.0528e-05, -5.1372e-06, + 3.2056e-06, 3.0085e-05, -1.0841e-05, 1.6063e-05, -1.3271e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 214.56, cls_loss 0.0025 cls_loss_mapping 0.0031 cls_loss_causal 0.5325 re_mapping 0.0065 re_causal 0.0189 /// teacc 98.98 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1794, 0.0758, -0.0415, ..., -0.0123, -0.1073, -0.1196], + [ 0.0239, 0.0691, -0.1002, ..., -0.0942, -0.0270, 0.0917], + [ 0.0521, -0.0899, -0.0989, ..., 0.0008, -0.0442, -0.0501], + ..., + [ 0.0665, 0.0043, 0.1048, ..., 0.0579, 0.1436, -0.0035], + [ 0.0888, -0.1326, -0.1076, ..., -0.1694, -0.0499, 0.1475], + [-0.0867, 0.0740, 0.0234, ..., -0.1234, -0.0726, -0.0182]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, -8.8196e-07, 7.7300e-08, ..., 3.4459e-08, + 4.2841e-08, 4.8429e-08], + [ 2.6822e-07, -8.3819e-07, 8.2236e-07, ..., 3.1851e-07, + 2.2072e-07, -1.6931e-06], + [-9.7789e-08, 1.5367e-07, 2.1793e-07, ..., 4.3772e-08, + 2.2352e-08, 9.7789e-08], + ..., + [-1.2256e-06, 4.0792e-07, -1.5218e-06, ..., -4.9546e-07, + -1.1977e-06, 6.0443e-07], + [ 2.1327e-07, 4.9826e-07, 3.9022e-07, ..., 2.2352e-07, + 1.4622e-07, 8.4750e-08], + [ 3.9022e-07, 1.2191e-06, -6.1933e-07, ..., 1.8533e-07, + 3.3248e-07, 1.0459e-06]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0085, -0.0121, -0.0038, -0.0155, -0.0045, -0.0040, 0.0162, 0.0105, + 0.0166, -0.0114], device='cuda:0'), grad: tensor([-1.1846e-06, -2.1029e-06, 2.3562e-07, 1.5078e-06, -1.1539e-06, + -2.5183e-06, 9.3132e-07, -4.1816e-07, 1.5525e-06, 3.1460e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 214.84, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5421 re_mapping 0.0069 re_causal 0.0195 /// teacc 98.94 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1826, 0.0755, -0.0431, ..., -0.0126, -0.1083, -0.1201], + [ 0.0239, 0.0692, -0.1002, ..., -0.0945, -0.0270, 0.0920], + [ 0.0524, -0.0905, -0.0993, ..., 0.0008, -0.0442, -0.0500], + ..., + [ 0.0668, 0.0042, 0.1054, ..., 0.0584, 0.1440, -0.0037], + [ 0.0884, -0.1335, -0.1081, ..., -0.1711, -0.0502, 0.1477], + [-0.0873, 0.0742, 0.0234, ..., -0.1245, -0.0733, -0.0187]], + device='cuda:0'), grad: tensor([[ 2.4401e-07, 2.1048e-07, 2.2911e-07, ..., 3.7253e-08, + 7.6368e-08, 2.4214e-08], + [ 1.1221e-05, 3.7160e-07, 5.8040e-06, ..., 2.5257e-06, + 6.7912e-06, 2.7958e-06], + [-1.4529e-07, 6.9756e-07, 9.5833e-07, ..., 2.9057e-07, + 4.4610e-07, 6.3330e-08], + ..., + [-2.0087e-05, 1.0803e-07, -1.0081e-05, ..., -4.6864e-06, + -1.2301e-05, -4.9546e-06], + [ 5.2489e-06, 1.1604e-06, 3.1516e-06, ..., 1.1344e-06, + 3.0827e-06, 1.1306e-06], + [ 1.3178e-06, -5.8487e-06, -2.8573e-06, ..., 3.3714e-07, + 7.4226e-07, 3.4552e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0086, -0.0120, -0.0035, -0.0140, -0.0044, -0.0053, 0.0160, 0.0107, + 0.0162, -0.0115], device='cuda:0'), grad: tensor([ 1.0896e-06, 1.7405e-05, 1.5181e-07, 7.5437e-06, 2.1346e-06, + 1.1083e-06, 7.8231e-08, -2.9013e-05, 1.0766e-05, -1.1310e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 214.96, cls_loss 0.0027 cls_loss_mapping 0.0040 cls_loss_causal 0.5096 re_mapping 0.0065 re_causal 0.0190 /// teacc 98.95 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1850, 0.0757, -0.0439, ..., -0.0132, -0.1091, -0.1208], + [ 0.0230, 0.0690, -0.1012, ..., -0.0953, -0.0283, 0.0921], + [ 0.0514, -0.0935, -0.1001, ..., 0.0004, -0.0460, -0.0502], + ..., + [ 0.0682, 0.0044, 0.1065, ..., 0.0590, 0.1460, -0.0040], + [ 0.0886, -0.1336, -0.1083, ..., -0.1716, -0.0504, 0.1482], + [-0.0876, 0.0746, 0.0239, ..., -0.1250, -0.0730, -0.0180]], + device='cuda:0'), grad: tensor([[ 5.5879e-08, 2.3283e-08, 6.4261e-08, ..., 3.7253e-08, + 3.6322e-08, 1.2107e-08], + [ 4.0513e-07, -6.0536e-08, 3.7905e-07, ..., 1.3411e-07, + 3.7346e-07, -2.8405e-07], + [-1.0990e-07, 3.6322e-08, 8.6613e-08, ..., -6.8918e-08, + -4.7497e-08, 5.2154e-08], + ..., + [-7.5065e-07, 8.3819e-08, -6.2399e-07, ..., -1.6112e-07, + -6.8545e-07, 7.4506e-08], + [-5.6811e-08, 1.1735e-07, 1.2293e-07, ..., 4.2841e-08, + 6.4261e-08, -1.0803e-07], + [ 1.9465e-07, -5.4389e-07, -7.1712e-07, ..., 3.9116e-08, + 1.3039e-07, 8.9407e-08]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0086, -0.0126, -0.0052, -0.0142, -0.0048, -0.0042, 0.0156, 0.0119, + 0.0162, -0.0111], device='cuda:0'), grad: tensor([ 3.5390e-07, 3.1013e-07, -1.3690e-07, 1.0338e-07, 1.0589e-06, + 2.9709e-07, -4.5914e-07, -7.2550e-07, 2.1234e-07, -1.0319e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 214.69, cls_loss 0.0027 cls_loss_mapping 0.0046 cls_loss_causal 0.5881 re_mapping 0.0067 re_causal 0.0209 /// teacc 98.85 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1859, 0.0753, -0.0448, ..., -0.0125, -0.1097, -0.1212], + [ 0.0218, 0.0694, -0.1030, ..., -0.0957, -0.0296, 0.0915], + [ 0.0522, -0.0938, -0.1002, ..., 0.0010, -0.0458, -0.0505], + ..., + [ 0.0692, 0.0040, 0.1080, ..., 0.0593, 0.1471, -0.0029], + [ 0.0877, -0.1342, -0.1091, ..., -0.1750, -0.0508, 0.1483], + [-0.0892, 0.0746, 0.0239, ..., -0.1268, -0.0745, -0.0183]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, -4.0978e-08, 1.6857e-07, ..., 1.3970e-07, + 1.9558e-08, 8.4750e-08], + [-1.5553e-07, -6.6403e-07, 8.2795e-07, ..., 7.6368e-07, + -2.3190e-07, -1.9800e-06], + [ 1.9465e-07, 1.7136e-07, 5.3085e-07, ..., 4.7311e-07, + -8.2888e-08, 2.8592e-07], + ..., + [ 3.7253e-07, 6.6683e-07, 1.8012e-06, ..., 2.9244e-07, + 8.7917e-07, 3.5949e-07], + [-2.7008e-06, 1.7136e-07, 4.1630e-07, ..., 3.8184e-07, + -2.0768e-07, -3.4459e-06], + [ 1.7956e-06, -8.7079e-07, -2.3339e-06, ..., 8.1025e-08, + -8.0466e-07, 2.4494e-06]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0089, -0.0133, -0.0044, -0.0143, -0.0054, -0.0019, 0.0141, 0.0128, + 0.0155, -0.0114], device='cuda:0'), grad: tensor([ 8.5961e-07, -8.0839e-07, 1.5246e-06, -2.0772e-05, 2.7753e-06, + 1.4849e-05, 2.0117e-06, 4.1388e-06, -5.2378e-06, 5.8115e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 214.72, cls_loss 0.0032 cls_loss_mapping 0.0062 cls_loss_causal 0.5662 re_mapping 0.0068 re_causal 0.0188 /// teacc 98.93 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1881, 0.0754, -0.0467, ..., -0.0133, -0.1108, -0.1236], + [ 0.0214, 0.0695, -0.1036, ..., -0.0971, -0.0303, 0.0920], + [ 0.0520, -0.0933, -0.1007, ..., 0.0013, -0.0458, -0.0537], + ..., + [ 0.0700, 0.0039, 0.1091, ..., 0.0602, 0.1480, -0.0024], + [ 0.0901, -0.1363, -0.1097, ..., -0.1761, -0.0513, 0.1506], + [-0.0900, 0.0750, 0.0239, ..., -0.1287, -0.0756, -0.0186]], + device='cuda:0'), grad: tensor([[ 4.6007e-06, -3.5856e-07, 4.0699e-07, ..., -1.3039e-08, + 1.2107e-08, 7.9796e-06], + [ 5.7554e-04, 5.4296e-07, 7.2736e-07, ..., 1.7602e-07, + -1.4994e-07, 9.8801e-04], + [ 1.4789e-05, 1.9465e-07, 8.3819e-08, ..., -2.9430e-07, + -1.1083e-07, 2.6003e-05], + ..., + [ 1.1168e-05, 1.3724e-05, 2.4348e-05, ..., 5.5879e-09, + 7.8231e-08, 2.0429e-05], + [-6.3753e-04, 1.7677e-06, 1.0543e-06, ..., 8.7544e-08, + 6.7055e-08, -1.0939e-03], + [ 1.3569e-06, -1.7929e-03, -3.5262e-04, ..., 4.9360e-08, + 2.3283e-08, -1.5087e-03]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0091, -0.0137, -0.0036, -0.0146, -0.0056, -0.0019, 0.0119, 0.0134, + 0.0180, -0.0113], device='cuda:0'), grad: tensor([ 1.1511e-05, 1.5020e-03, 3.9220e-05, 2.5406e-05, 5.2719e-03, + -7.6070e-06, 6.0052e-05, 7.2002e-05, -1.6565e-03, -5.3215e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 214.91, cls_loss 0.0023 cls_loss_mapping 0.0029 cls_loss_causal 0.4960 re_mapping 0.0068 re_causal 0.0188 /// teacc 98.88 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1898, 0.0758, -0.0474, ..., -0.0138, -0.1122, -0.1242], + [ 0.0213, 0.0697, -0.1037, ..., -0.0973, -0.0302, 0.0919], + [ 0.0522, -0.0937, -0.1008, ..., 0.0016, -0.0456, -0.0539], + ..., + [ 0.0700, 0.0037, 0.1094, ..., 0.0603, 0.1481, -0.0026], + [ 0.0904, -0.1371, -0.1101, ..., -0.1767, -0.0515, 0.1511], + [-0.0903, 0.0758, 0.0242, ..., -0.1294, -0.0759, -0.0170]], + device='cuda:0'), grad: tensor([[ 3.0398e-06, 1.4501e-06, 2.1327e-06, ..., 1.3160e-06, + 4.5039e-06, 1.2852e-07], + [ 8.5980e-06, 9.4716e-07, 7.2643e-07, ..., 1.3057e-06, + 2.6301e-06, 7.4469e-06], + [-9.9689e-06, 5.8766e-07, 6.8825e-07, ..., -7.5549e-06, + -5.0813e-06, 5.1782e-07], + ..., + [-2.3376e-07, -1.7611e-06, -5.2080e-06, ..., 9.5926e-08, + -7.3984e-06, 2.9430e-07], + [ 1.7822e-05, 2.3916e-05, 4.8801e-07, ..., 1.0803e-06, + 1.0254e-06, -1.0662e-05], + [ 1.3160e-06, -7.6834e-07, -2.0489e-06, ..., 6.0163e-07, + 7.1060e-07, 4.8149e-07]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0086, -0.0137, -0.0034, -0.0141, -0.0066, -0.0025, 0.0121, 0.0132, + 0.0179, -0.0104], device='cuda:0'), grad: tensor([ 1.4789e-05, 2.0817e-05, -2.2501e-05, 5.4091e-05, 8.9556e-06, + -2.2578e-04, 9.0599e-06, -1.1362e-05, 1.4997e-04, 2.0452e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 214.89, cls_loss 0.0026 cls_loss_mapping 0.0039 cls_loss_causal 0.5085 re_mapping 0.0064 re_causal 0.0183 /// teacc 99.00 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1906, 0.0761, -0.0484, ..., -0.0140, -0.1134, -0.1247], + [ 0.0220, 0.0705, -0.1028, ..., -0.0961, -0.0293, 0.0932], + [ 0.0521, -0.0940, -0.1013, ..., 0.0017, -0.0457, -0.0549], + ..., + [ 0.0697, 0.0031, 0.1093, ..., 0.0602, 0.1478, -0.0040], + [ 0.0902, -0.1380, -0.1108, ..., -0.1778, -0.0519, 0.1511], + [-0.0904, 0.0754, 0.0247, ..., -0.1300, -0.0762, -0.0175]], + device='cuda:0'), grad: tensor([[ 5.6624e-07, -3.9302e-06, 3.1944e-07, ..., 4.2375e-07, + 3.3062e-07, 1.0896e-07], + [ 2.4557e-05, 8.3353e-07, 3.3975e-05, ..., 1.2994e-05, + 3.1471e-05, 1.2435e-05], + [-6.0499e-06, 3.7719e-07, 2.8294e-06, ..., -8.1807e-06, + 5.8021e-07, 9.5647e-07], + ..., + [-3.9667e-05, 2.5425e-07, -5.8413e-05, ..., -1.8805e-05, + -5.2571e-05, -2.0429e-05], + [ 8.5682e-06, 9.3691e-07, 1.2353e-05, ..., 5.5172e-06, + 1.1154e-05, 3.7365e-06], + [ 4.5076e-06, 2.9728e-06, 8.7246e-06, ..., 3.6489e-06, + 6.1579e-06, 2.4531e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0088, -0.0128, -0.0034, -0.0145, -0.0062, -0.0021, 0.0123, 0.0123, + 0.0174, -0.0110], device='cuda:0'), grad: tensor([-5.9567e-06, 6.0827e-05, -2.5198e-05, 1.8686e-05, -2.0713e-06, + 1.0639e-05, -5.8599e-06, -9.2983e-05, 2.3887e-05, 1.7956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 214.75, cls_loss 0.0023 cls_loss_mapping 0.0041 cls_loss_causal 0.5267 re_mapping 0.0066 re_causal 0.0194 /// teacc 98.95 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1915, 0.0764, -0.0495, ..., -0.0132, -0.1140, -0.1245], + [ 0.0223, 0.0709, -0.1022, ..., -0.0951, -0.0289, 0.0938], + [ 0.0526, -0.0928, -0.1011, ..., 0.0015, -0.0455, -0.0543], + ..., + [ 0.0694, 0.0025, 0.1092, ..., 0.0599, 0.1476, -0.0046], + [ 0.0902, -0.1385, -0.1111, ..., -0.1784, -0.0521, 0.1513], + [-0.0913, 0.0753, 0.0250, ..., -0.1315, -0.0770, -0.0179]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, -9.3132e-07, 2.7101e-07, ..., 2.5891e-07, + 5.8673e-08, 1.2107e-08], + [ 4.4703e-08, 6.4261e-08, 3.3900e-07, ..., 2.8498e-07, + 4.9360e-08, -6.9849e-08], + [-1.6391e-07, 3.9302e-07, 1.9372e-06, ..., 1.6065e-06, + -1.9092e-07, 2.0489e-08], + ..., + [-1.7043e-07, 5.8301e-06, 1.3374e-05, ..., 3.9767e-07, + -1.2293e-07, 3.9116e-08], + [-3.1665e-08, 1.2545e-06, 2.3544e-06, ..., 1.7630e-06, + 2.8871e-08, -1.0524e-07], + [-8.3819e-09, -1.8433e-05, -2.6494e-05, ..., 5.1130e-07, + 2.7008e-08, -3.7719e-07]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0084, -0.0126, -0.0022, -0.0147, -0.0062, -0.0019, 0.0122, 0.0114, + 0.0173, -0.0112], device='cuda:0'), grad: tensor([-9.0431e-07, 5.3737e-07, 2.4680e-06, -7.8380e-06, 2.7448e-05, + 2.9802e-06, 2.9895e-07, 2.3603e-05, 4.6566e-06, -5.3227e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 214.99, cls_loss 0.0021 cls_loss_mapping 0.0036 cls_loss_causal 0.5361 re_mapping 0.0067 re_causal 0.0203 /// teacc 98.90 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.1920, 0.0767, -0.0508, ..., -0.0128, -0.1146, -0.1244], + [ 0.0221, 0.0711, -0.1025, ..., -0.0955, -0.0291, 0.0939], + [ 0.0526, -0.0934, -0.1013, ..., 0.0015, -0.0456, -0.0554], + ..., + [ 0.0695, 0.0021, 0.1092, ..., 0.0603, 0.1479, -0.0047], + [ 0.0904, -0.1390, -0.1116, ..., -0.1795, -0.0519, 0.1515], + [-0.0911, 0.0757, 0.0262, ..., -0.1326, -0.0762, -0.0179]], + device='cuda:0'), grad: tensor([[ 2.0619e-06, -2.3041e-06, 9.2201e-08, ..., 4.6194e-07, + 2.2352e-08, 8.8476e-08], + [ 4.4797e-07, 4.3474e-06, 2.8666e-06, ..., 2.0638e-06, + 8.7824e-07, 2.6599e-06], + [-3.5260e-06, 4.6659e-07, 5.1875e-07, ..., -6.4261e-07, + 6.8918e-08, 3.7998e-07], + ..., + [-1.6103e-06, 4.0382e-06, 7.8045e-07, ..., 3.1386e-07, + -4.5914e-07, 2.6133e-06], + [ 1.6177e-06, 1.4845e-06, 1.2778e-06, ..., 6.2305e-07, + 1.5646e-07, 1.2610e-06], + [ 1.4622e-07, 2.4121e-06, 6.7707e-07, ..., 1.3793e-06, + 4.9919e-07, 8.7079e-07]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0083, -0.0126, -0.0024, -0.0139, -0.0066, -0.0025, 0.0122, 0.0113, + 0.0172, -0.0105], device='cuda:0'), grad: tensor([ 2.5444e-06, 1.5706e-05, -1.2234e-05, 2.8256e-06, -4.1813e-05, + 1.6354e-06, 6.0350e-07, 1.1936e-05, 1.0878e-05, 7.8753e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 214.67, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.5335 re_mapping 0.0062 re_causal 0.0192 /// teacc 98.91 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1929, 0.0767, -0.0516, ..., -0.0129, -0.1151, -0.1256], + [ 0.0221, 0.0716, -0.1028, ..., -0.0958, -0.0292, 0.0946], + [ 0.0525, -0.0940, -0.1015, ..., 0.0017, -0.0457, -0.0565], + ..., + [ 0.0699, 0.0022, 0.1099, ..., 0.0608, 0.1483, -0.0048], + [ 0.0904, -0.1395, -0.1121, ..., -0.1802, -0.0521, 0.1515], + [-0.0919, 0.0757, 0.0262, ..., -0.1344, -0.0770, -0.0182]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 5.3830e-07, 2.5146e-07, ..., 1.6764e-08, + 3.8184e-08, 4.1816e-07], + [-2.5146e-08, -2.8964e-06, 2.7101e-07, ..., 1.4994e-07, + 6.2678e-07, -3.9823e-06], + [-4.1444e-07, 2.2445e-07, 2.1420e-07, ..., -2.8871e-08, + -5.3458e-07, 2.7567e-07], + ..., + [-4.1258e-07, 3.7160e-07, -7.6927e-07, ..., -1.5274e-07, + -5.4017e-07, 4.4983e-07], + [ 8.8476e-08, 6.5099e-07, 1.0673e-06, ..., 4.1910e-08, + 7.0781e-08, 4.7497e-07], + [ 2.8592e-07, -1.5656e-06, -3.0473e-06, ..., 8.9407e-08, + 2.3656e-07, 5.7742e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0086, -0.0124, -0.0025, -0.0127, -0.0067, -0.0036, 0.0123, 0.0117, + 0.0171, -0.0106], device='cuda:0'), grad: tensor([ 1.6280e-06, -6.6981e-06, -1.1707e-06, -1.4715e-07, 5.3272e-06, + 8.3819e-09, 2.8685e-06, 3.9674e-07, 3.6098e-06, -5.8264e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 214.98, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.5126 re_mapping 0.0061 re_causal 0.0178 /// teacc 98.92 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.1942, 0.0770, -0.0538, ..., -0.0135, -0.1169, -0.1252], + [ 0.0219, 0.0719, -0.1033, ..., -0.0965, -0.0294, 0.0949], + [ 0.0527, -0.0944, -0.1019, ..., 0.0017, -0.0457, -0.0573], + ..., + [ 0.0703, 0.0022, 0.1108, ..., 0.0614, 0.1488, -0.0046], + [ 0.0905, -0.1403, -0.1124, ..., -0.1808, -0.0522, 0.1517], + [-0.0926, 0.0756, 0.0262, ..., -0.1355, -0.0777, -0.0188]], + device='cuda:0'), grad: tensor([[ 1.3588e-06, 8.1584e-06, 1.8906e-07, ..., 1.4016e-06, + 4.7497e-08, 2.5984e-06], + [ 2.2929e-06, -3.4552e-07, 5.7463e-07, ..., 1.1064e-06, + 2.3805e-06, -1.9595e-06], + [-7.7859e-06, 2.6170e-07, 3.3621e-07, ..., -6.2473e-06, + -3.3751e-06, 4.8894e-07], + ..., + [ 1.8626e-08, 2.5053e-07, -4.2375e-07, ..., 3.4180e-07, + -3.8370e-07, -1.2480e-07], + [ 4.5635e-07, 1.6699e-06, 1.6680e-06, ..., 4.2096e-07, + 1.3225e-07, 5.6531e-07], + [ 1.8626e-07, -6.6698e-05, -7.2777e-05, ..., 1.7509e-07, + 3.8184e-08, -2.5690e-05]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0086, -0.0125, -0.0023, -0.0130, -0.0066, -0.0035, 0.0122, 0.0122, + 0.0171, -0.0108], device='cuda:0'), grad: tensor([ 5.0724e-05, 4.1239e-06, -2.1622e-05, 8.5235e-06, 2.0504e-04, + -1.6510e-05, -1.7151e-05, 1.7807e-06, 7.0520e-06, -2.2161e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 214.80, cls_loss 0.0024 cls_loss_mapping 0.0035 cls_loss_causal 0.5388 re_mapping 0.0065 re_causal 0.0190 /// teacc 98.97 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1962, 0.0774, -0.0560, ..., -0.0139, -0.1183, -0.1261], + [ 0.0211, 0.0721, -0.1047, ..., -0.0983, -0.0308, 0.0955], + [ 0.0515, -0.0953, -0.1034, ..., 0.0005, -0.0473, -0.0575], + ..., + [ 0.0715, 0.0022, 0.1122, ..., 0.0630, 0.1504, -0.0053], + [ 0.0908, -0.1410, -0.1124, ..., -0.1811, -0.0517, 0.1517], + [-0.0929, 0.0759, 0.0268, ..., -0.1359, -0.0778, -0.0190]], + device='cuda:0'), grad: tensor([[ 1.2666e-07, -3.9078e-06, 1.1362e-07, ..., -1.4901e-07, + 1.0990e-07, 2.9244e-07], + [-1.6214e-06, -2.6543e-06, 5.5972e-07, ..., 3.6415e-07, + 5.5600e-07, -1.3523e-05], + [ 1.1772e-06, 7.3668e-07, 1.4296e-06, ..., 6.5006e-07, + 1.4249e-06, 1.7369e-06], + ..., + [-2.5220e-06, 2.4363e-06, -3.6489e-06, ..., -1.9707e-06, + -3.6918e-06, 6.2212e-07], + [ 8.6054e-07, 1.2768e-06, 1.1455e-07, ..., 5.2154e-08, + 5.9605e-08, 5.4017e-06], + [ 2.5984e-07, 1.6578e-07, 1.5460e-07, ..., 1.7043e-07, + 2.7195e-07, 4.0326e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0084, -0.0130, -0.0034, -0.0133, -0.0068, -0.0032, 0.0121, 0.0132, + 0.0170, -0.0106], device='cuda:0'), grad: tensor([-8.6650e-06, -1.7673e-05, 4.7833e-06, 1.4137e-06, 2.4550e-06, + 1.3500e-05, -1.0267e-05, 2.8964e-06, 9.7156e-06, 1.7779e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 214.63, cls_loss 0.0025 cls_loss_mapping 0.0043 cls_loss_causal 0.5092 re_mapping 0.0065 re_causal 0.0183 /// teacc 98.96 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1985, 0.0777, -0.0594, ..., -0.0147, -0.1211, -0.1272], + [ 0.0201, 0.0715, -0.1060, ..., -0.1005, -0.0325, 0.0961], + [ 0.0526, -0.0967, -0.1028, ..., 0.0013, -0.0470, -0.0579], + ..., + [ 0.0714, 0.0027, 0.1125, ..., 0.0636, 0.1512, -0.0059], + [ 0.0909, -0.1411, -0.1127, ..., -0.1819, -0.0517, 0.1520], + [-0.0932, 0.0759, 0.0272, ..., -0.1368, -0.0783, -0.0193]], + device='cuda:0'), grad: tensor([[ 1.8533e-07, -2.3842e-07, 5.1223e-08, ..., 1.6568e-06, + 1.2582e-06, 6.6124e-08], + [-2.7940e-08, -9.0338e-08, 1.6205e-07, ..., 8.5030e-07, + 7.2923e-07, -1.3728e-06], + [-5.1782e-07, 7.4506e-08, 1.1642e-07, ..., -6.1132e-06, + -4.5374e-06, 2.3004e-07], + ..., + [-3.0641e-07, 1.7602e-07, -7.0501e-07, ..., 7.2643e-07, + 9.6858e-08, 5.2992e-07], + [-1.0617e-07, 3.0641e-07, 1.2759e-07, ..., 2.9802e-07, + 2.4028e-07, 1.4063e-07], + [ 2.8592e-07, 7.4599e-07, -1.9558e-08, ..., 3.0268e-07, + 4.5542e-07, 1.4286e-06]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0083, -0.0138, -0.0030, -0.0142, -0.0068, -0.0022, 0.0121, 0.0134, + 0.0170, -0.0105], device='cuda:0'), grad: tensor([ 3.7365e-06, 2.7288e-07, -1.4573e-05, 5.9344e-06, -4.6082e-06, + -7.8231e-06, 8.1137e-06, 2.1700e-06, 2.1160e-06, 4.5933e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 214.89, cls_loss 0.0026 cls_loss_mapping 0.0032 cls_loss_causal 0.5184 re_mapping 0.0064 re_causal 0.0182 /// teacc 98.87 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.2001, 0.0779, -0.0613, ..., -0.0152, -0.1231, -0.1289], + [ 0.0197, 0.0717, -0.1066, ..., -0.1007, -0.0328, 0.0962], + [ 0.0519, -0.0973, -0.1035, ..., 0.0009, -0.0475, -0.0602], + ..., + [ 0.0721, 0.0028, 0.1132, ..., 0.0637, 0.1518, -0.0054], + [ 0.0911, -0.1419, -0.1133, ..., -0.1824, -0.0520, 0.1525], + [-0.0935, 0.0755, 0.0276, ..., -0.1378, -0.0787, -0.0195]], + device='cuda:0'), grad: tensor([[ 4.7404e-07, 8.3540e-07, 1.6671e-07, ..., 2.3656e-07, + 2.3562e-07, 2.4717e-06], + [ 2.9448e-06, 1.2159e-05, 8.4098e-07, ..., 1.0869e-06, + 3.9302e-07, 1.8656e-05], + [-1.2740e-06, 5.7556e-07, 2.6692e-06, ..., 1.9893e-06, + -1.6177e-06, 1.1176e-06], + ..., + [ 1.0747e-06, 4.0159e-06, 6.0443e-07, ..., 1.0058e-06, + 2.5146e-07, 5.3011e-06], + [-1.5900e-05, -5.5991e-06, -7.9423e-06, ..., -7.1824e-06, + 1.3970e-07, -2.2352e-05], + [ 1.1856e-06, 7.7844e-05, 2.8044e-05, ..., 4.5635e-07, + 2.2445e-07, 3.6180e-05]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0084, -0.0139, -0.0037, -0.0149, -0.0068, -0.0017, 0.0124, 0.0139, + 0.0170, -0.0108], device='cuda:0'), grad: tensor([ 8.4117e-06, 3.2604e-05, 1.5255e-06, 2.8461e-05, -2.0063e-04, + 3.2075e-06, 7.0855e-06, 1.2062e-05, -4.2170e-05, 1.4913e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 214.80, cls_loss 0.0018 cls_loss_mapping 0.0039 cls_loss_causal 0.5091 re_mapping 0.0063 re_causal 0.0183 /// teacc 98.86 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1980, 0.0777, -0.0628, ..., -0.0154, -0.1240, -0.1272], + [ 0.0198, 0.0715, -0.1065, ..., -0.1007, -0.0328, 0.0965], + [ 0.0520, -0.0972, -0.1031, ..., 0.0014, -0.0473, -0.0611], + ..., + [ 0.0721, 0.0024, 0.1131, ..., 0.0637, 0.1518, -0.0060], + [ 0.0911, -0.1432, -0.1136, ..., -0.1829, -0.0522, 0.1528], + [-0.0938, 0.0759, 0.0275, ..., -0.1408, -0.0801, -0.0196]], + device='cuda:0'), grad: tensor([[ 1.9465e-07, -5.3924e-07, 5.0291e-08, ..., 7.9162e-08, + 2.0489e-08, -7.4506e-09], + [-1.5631e-05, -2.3972e-06, 7.3574e-08, ..., 6.6124e-08, + -1.4640e-05, -2.8417e-05], + [-1.4016e-06, 4.2561e-07, 3.2596e-08, ..., -7.0687e-07, + -7.3574e-08, 5.7649e-07], + ..., + [ 1.5013e-05, 2.7400e-06, -8.7544e-08, ..., 8.3819e-09, + 1.3888e-05, 2.7165e-05], + [ 1.0272e-06, 3.8035e-06, 1.9688e-06, ..., 2.8498e-06, + 3.7346e-07, -9.0618e-07], + [ 3.3062e-07, -5.3011e-06, 3.1665e-08, ..., 2.3190e-07, + 2.3935e-07, 4.5635e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0085, -0.0138, -0.0030, -0.0153, -0.0066, -0.0015, 0.0125, 0.0136, + 0.0169, -0.0110], device='cuda:0'), grad: tensor([-8.6986e-07, -4.6879e-05, -2.0042e-06, 1.2808e-05, 8.4639e-06, + -2.4796e-05, 2.0191e-06, 4.6045e-05, 1.5676e-05, -1.0461e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 214.55, cls_loss 0.0018 cls_loss_mapping 0.0028 cls_loss_causal 0.5255 re_mapping 0.0061 re_causal 0.0189 /// teacc 99.01 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1985, 0.0778, -0.0641, ..., -0.0157, -0.1248, -0.1276], + [ 0.0190, 0.0715, -0.1072, ..., -0.1009, -0.0332, 0.0963], + [ 0.0535, -0.0980, -0.1019, ..., 0.0025, -0.0463, -0.0623], + ..., + [ 0.0714, 0.0024, 0.1127, ..., 0.0630, 0.1515, -0.0059], + [ 0.0917, -0.1438, -0.1133, ..., -0.1836, -0.0515, 0.1534], + [-0.0940, 0.0761, 0.0276, ..., -0.1412, -0.0807, -0.0195]], + device='cuda:0'), grad: tensor([[ 2.1234e-07, -5.3179e-07, 2.5891e-07, ..., 1.0151e-07, + 1.8347e-07, 3.0734e-08], + [ 2.5332e-07, 1.9744e-07, 3.6601e-07, ..., 1.2666e-07, + 2.5239e-07, -2.2165e-07], + [-6.7987e-08, 2.9244e-07, 1.9185e-07, ..., -2.5798e-07, + 1.0151e-07, 1.1176e-07], + ..., + [-9.7137e-07, 5.1688e-07, -1.2619e-06, ..., -2.7008e-07, + -9.6112e-07, 1.5087e-07], + [ 1.4622e-07, 2.5984e-06, 1.8161e-07, ..., 2.0023e-07, + 8.7544e-08, 4.2841e-08], + [ 7.0781e-07, 1.0788e-05, 3.0734e-07, ..., 1.3039e-07, + 2.3842e-07, 1.4771e-06]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0086, -0.0142, -0.0023, -0.0154, -0.0066, -0.0014, 0.0126, 0.0133, + 0.0169, -0.0108], device='cuda:0'), grad: tensor([ 1.6078e-05, 2.6673e-06, 2.3264e-06, 3.3490e-06, 3.2127e-05, + -3.4660e-05, -6.1810e-05, 3.7067e-07, 1.1638e-05, 2.7895e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 214.60, cls_loss 0.0019 cls_loss_mapping 0.0031 cls_loss_causal 0.5490 re_mapping 0.0058 re_causal 0.0182 /// teacc 98.91 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1990, 0.0777, -0.0654, ..., -0.0159, -0.1264, -0.1281], + [ 0.0198, 0.0720, -0.1069, ..., -0.1010, -0.0323, 0.0967], + [ 0.0530, -0.0989, -0.1023, ..., 0.0024, -0.0470, -0.0629], + ..., + [ 0.0715, 0.0022, 0.1130, ..., 0.0632, 0.1515, -0.0061], + [ 0.0917, -0.1443, -0.1137, ..., -0.1841, -0.0517, 0.1534], + [-0.0950, 0.0758, 0.0277, ..., -0.1421, -0.0815, -0.0202]], + device='cuda:0'), grad: tensor([[ 1.7509e-07, -1.3132e-07, 1.0245e-07, ..., 2.3842e-07, + 2.2631e-07, 8.3819e-09], + [ 2.0098e-06, -1.4342e-07, 1.7658e-06, ..., 2.0042e-06, + 1.9986e-06, -2.5053e-07], + [-2.7627e-05, 1.6764e-08, -1.2085e-05, ..., -2.4825e-05, + -1.9103e-05, 5.3085e-08], + ..., + [ 1.7792e-05, 2.2817e-07, 7.3723e-06, ..., 1.5900e-05, + 1.1779e-05, 1.3039e-08], + [ 1.6754e-06, 6.7987e-08, 7.9628e-07, ..., 1.4482e-06, + 1.1511e-06, -1.1548e-07], + [ 1.5646e-07, -1.6764e-07, -5.1875e-07, ..., 2.3656e-07, + 1.0338e-07, 1.0990e-07]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0087, -0.0132, -0.0032, -0.0155, -0.0063, -0.0014, 0.0128, 0.0132, + 0.0166, -0.0113], device='cuda:0'), grad: tensor([ 3.3528e-07, 5.1670e-06, -5.0992e-05, 8.9854e-06, 5.7835e-07, + 7.7486e-07, 1.9185e-07, 3.2216e-05, 2.8964e-06, -1.8533e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 214.73, cls_loss 0.0016 cls_loss_mapping 0.0033 cls_loss_causal 0.5142 re_mapping 0.0062 re_causal 0.0185 /// teacc 99.02 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1995, 0.0780, -0.0671, ..., -0.0158, -0.1276, -0.1292], + [ 0.0197, 0.0725, -0.1072, ..., -0.1014, -0.0323, 0.0970], + [ 0.0529, -0.0992, -0.1027, ..., 0.0022, -0.0473, -0.0640], + ..., + [ 0.0718, 0.0021, 0.1136, ..., 0.0636, 0.1518, -0.0059], + [ 0.0919, -0.1449, -0.1141, ..., -0.1847, -0.0520, 0.1537], + [-0.0960, 0.0757, 0.0277, ..., -0.1439, -0.0823, -0.0204]], + device='cuda:0'), grad: tensor([[ 3.7719e-07, -2.2314e-06, 2.2538e-07, ..., 2.6543e-07, + 1.2852e-07, 2.2911e-07], + [ 3.1572e-07, 1.1940e-06, 2.8461e-06, ..., 8.6427e-07, + 5.2992e-07, -1.1716e-06], + [ 2.8666e-06, 1.1856e-06, -1.1884e-06, ..., -3.1665e-06, + -4.6492e-06, 4.7721e-06], + ..., + [ 1.6186e-06, 8.6799e-07, 7.0315e-07, ..., 1.4203e-06, + 4.8336e-07, 6.5658e-07], + [-2.0489e-05, 1.4668e-06, 9.7509e-07, ..., 8.3633e-07, + 3.1665e-07, -1.0431e-05], + [ 2.7493e-06, -1.0673e-06, -2.6170e-07, ..., 2.2687e-06, + 1.5274e-06, 2.5798e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0087, -0.0132, -0.0034, -0.0154, -0.0064, -0.0014, 0.0127, 0.0136, + 0.0166, -0.0115], device='cuda:0'), grad: tensor([-2.0470e-06, 6.6832e-06, 9.3728e-06, -5.2452e-06, 2.2240e-06, + 2.3559e-05, 1.1623e-05, 7.0184e-06, -5.8681e-05, 5.5283e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 214.79, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5119 re_mapping 0.0063 re_causal 0.0184 /// teacc 99.00 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.2001, 0.0783, -0.0692, ..., -0.0174, -0.1284, -0.1301], + [ 0.0197, 0.0728, -0.1071, ..., -0.1015, -0.0324, 0.0974], + [ 0.0526, -0.0997, -0.1036, ..., 0.0010, -0.0476, -0.0650], + ..., + [ 0.0720, 0.0021, 0.1140, ..., 0.0640, 0.1521, -0.0063], + [ 0.0924, -0.1460, -0.1140, ..., -0.1845, -0.0517, 0.1541], + [-0.0971, 0.0754, 0.0276, ..., -0.1466, -0.0834, -0.0212]], + device='cuda:0'), grad: tensor([[ 5.0291e-08, 9.6858e-08, 1.8887e-06, ..., -1.4715e-07, + 7.4506e-09, 5.1223e-07], + [-1.3448e-06, -2.7940e-07, 3.3528e-08, ..., 5.5879e-08, + -4.4703e-08, -9.7305e-06], + [ 1.2349e-06, 2.7195e-07, 1.7323e-07, ..., 7.8231e-08, + -1.6764e-08, 3.6843e-06], + ..., + [ 4.7125e-07, 3.7067e-06, 7.3798e-06, ..., 4.2282e-07, + -2.0489e-08, 2.5947e-06], + [-1.5572e-06, 3.9116e-07, 5.7742e-08, ..., 1.6019e-07, + 4.0978e-08, 1.2852e-07], + [ 8.0094e-08, -4.5560e-06, -1.1086e-05, ..., -5.3458e-07, + 9.3132e-09, 6.9290e-07]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0089, -0.0130, -0.0040, -0.0151, -0.0061, -0.0015, 0.0127, 0.0137, + 0.0166, -0.0117], device='cuda:0'), grad: tensor([ 1.6868e-05, -1.0937e-05, 4.7944e-06, 1.8403e-06, -1.2517e-06, + -1.8626e-09, -1.4797e-05, 1.1250e-05, 1.9874e-06, -9.8050e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 214.60, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5276 re_mapping 0.0064 re_causal 0.0179 /// teacc 98.89 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.2010, 0.0795, -0.0714, ..., -0.0180, -0.1298, -0.1305], + [ 0.0196, 0.0732, -0.1073, ..., -0.1019, -0.0325, 0.0977], + [ 0.0527, -0.1001, -0.1041, ..., 0.0006, -0.0476, -0.0645], + ..., + [ 0.0722, 0.0017, 0.1145, ..., 0.0644, 0.1523, -0.0065], + [ 0.0923, -0.1467, -0.1148, ..., -0.1864, -0.0521, 0.1544], + [-0.0977, 0.0755, 0.0279, ..., -0.1480, -0.0840, -0.0214]], + device='cuda:0'), grad: tensor([[ 7.0781e-08, -2.0489e-08, 4.4703e-08, ..., 2.0489e-08, + 4.8429e-08, 1.8068e-07], + [-6.3367e-06, -9.9987e-06, -1.0714e-05, ..., -2.0023e-06, + 1.9372e-07, -2.3112e-05], + [ 3.0715e-06, 6.3330e-08, 2.1793e-06, ..., 1.6969e-06, + 2.7940e-06, 2.3469e-07], + ..., + [ 1.6205e-06, 8.6054e-06, 6.2250e-06, ..., -6.8918e-07, + -3.9451e-06, 2.1994e-05], + [-1.4901e-07, 2.6077e-07, 5.8115e-07, ..., 1.9930e-07, + 1.5646e-07, 5.2154e-08], + [ 5.7369e-07, 8.9966e-07, -1.2107e-07, ..., 2.1793e-07, + 1.1921e-07, 1.7788e-06]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0066, -0.0131, -0.0040, -0.0153, -0.0062, -0.0013, 0.0117, 0.0138, + 0.0165, -0.0117], device='cuda:0'), grad: tensor([ 1.3616e-06, -1.2733e-05, 4.6566e-06, 8.9779e-07, -1.8440e-06, + 2.1145e-05, -4.8161e-05, 2.4199e-05, 4.6343e-06, 5.7295e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 214.71, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5355 re_mapping 0.0067 re_causal 0.0199 /// teacc 99.02 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.2017, 0.0795, -0.0731, ..., -0.0177, -0.1301, -0.1322], + [ 0.0193, 0.0729, -0.1077, ..., -0.1025, -0.0326, 0.0977], + [ 0.0526, -0.1005, -0.1044, ..., 0.0004, -0.0477, -0.0650], + ..., + [ 0.0726, 0.0019, 0.1150, ..., 0.0650, 0.1526, -0.0065], + [ 0.0922, -0.1481, -0.1159, ..., -0.1869, -0.0523, 0.1545], + [-0.0983, 0.0760, 0.0283, ..., -0.1496, -0.0853, -0.0211]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, 8.3819e-08, 1.1176e-08, ..., 5.5879e-09, + 1.8626e-09, 1.1735e-07], + [-5.6997e-06, -1.2510e-05, 3.5390e-08, ..., 2.0489e-08, + 1.3039e-08, -2.0102e-05], + [ 3.1665e-08, 8.3819e-08, 5.9605e-08, ..., 7.4506e-09, + -7.4506e-09, 9.6858e-08], + ..., + [ 2.1793e-07, 8.4750e-07, -7.2643e-08, ..., -2.9802e-08, + -4.4703e-08, 9.9279e-07], + [ 4.9546e-07, 1.1884e-06, 2.2352e-08, ..., 0.0000e+00, + 1.8626e-09, 1.7919e-06], + [ 3.5763e-06, 1.7792e-05, -2.0675e-07, ..., 4.0978e-08, + 1.6764e-08, 1.2442e-05]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0066, -0.0132, -0.0042, -0.0154, -0.0063, -0.0010, 0.0115, 0.0142, + 0.0158, -0.0112], device='cuda:0'), grad: tensor([ 9.3132e-07, -4.1485e-05, 3.0920e-07, 4.2841e-08, -3.0175e-05, + 6.4597e-06, -8.4192e-06, 2.8703e-06, 4.0904e-06, 6.5327e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 214.72, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.5516 re_mapping 0.0066 re_causal 0.0197 /// teacc 98.92 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.2022, 0.0805, -0.0734, ..., -0.0162, -0.1307, -0.1326], + [ 0.0191, 0.0734, -0.1080, ..., -0.1028, -0.0328, 0.0982], + [ 0.0528, -0.1010, -0.1045, ..., 0.0004, -0.0476, -0.0655], + ..., + [ 0.0728, 0.0017, 0.1155, ..., 0.0653, 0.1528, -0.0064], + [ 0.0923, -0.1484, -0.1164, ..., -0.1878, -0.0526, 0.1546], + [-0.0993, 0.0754, 0.0280, ..., -0.1523, -0.0861, -0.0214]], + device='cuda:0'), grad: tensor([[ 8.0466e-07, -3.3714e-07, 6.3702e-07, ..., 4.8988e-07, + 8.9966e-07, 1.3225e-07], + [ 1.8198e-06, 6.0163e-07, 1.9986e-06, ..., 9.8906e-07, + 1.9222e-06, 4.5076e-07], + [ 1.1325e-05, 1.1604e-06, 1.1526e-05, ..., 6.4559e-06, + 1.3165e-05, 7.5251e-07], + ..., + [-1.6630e-05, 5.0291e-08, -1.7375e-05, ..., -9.3952e-06, + -1.9893e-05, 6.5751e-07], + [ 1.5926e-06, 6.7614e-07, 1.7267e-06, ..., 8.8476e-07, + 1.7285e-06, 5.6252e-07], + [ 2.1737e-06, 5.0627e-06, 5.9232e-07, ..., 1.1548e-06, + 6.1654e-07, 6.1579e-06]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0062, -0.0132, -0.0039, -0.0157, -0.0064, -0.0009, 0.0116, 0.0145, + 0.0158, -0.0118], device='cuda:0'), grad: tensor([ 7.4692e-07, 4.8950e-06, 2.3574e-05, 1.2517e-06, -2.4557e-05, + 1.8254e-07, 1.7714e-06, -3.0249e-05, 4.8317e-06, 1.7524e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 214.81, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5263 re_mapping 0.0064 re_causal 0.0186 /// teacc 99.05 lr 0.00010000 +Epoch 194, weight, value: tensor([[-2.0279e-01, 8.0691e-02, -7.4324e-02, ..., -1.5059e-02, + -1.3116e-01, -1.3309e-01], + [ 1.8299e-02, 7.4346e-02, -1.0934e-01, ..., -1.0290e-01, + -3.3242e-02, 9.8005e-02], + [ 5.2739e-02, -1.0213e-01, -1.0478e-01, ..., 1.9529e-04, + -4.7642e-02, -6.6802e-02], + ..., + [ 7.3336e-02, 6.8327e-04, 1.1626e-01, ..., 6.5286e-02, + 1.5309e-01, -5.7657e-03], + [ 9.2697e-02, -1.4880e-01, -1.1673e-01, ..., -1.8856e-01, + -5.2508e-02, 1.5500e-01], + [-1.0024e-01, 7.5419e-02, 2.8391e-02, ..., -1.5283e-01, + -8.6411e-02, -2.1823e-02]], device='cuda:0'), grad: tensor([[ 9.4995e-08, -1.1034e-05, 9.8720e-08, ..., 7.8231e-08, + 7.2643e-08, 5.2154e-08], + [-1.2591e-06, -2.3376e-06, 1.1548e-07, ..., 1.0803e-07, + 9.1270e-08, -4.9844e-06], + [-1.5646e-07, 7.8231e-08, 3.3155e-07, ..., 0.0000e+00, + -2.4214e-07, 1.5832e-07], + ..., + [ 2.5705e-07, 1.0617e-06, 1.4640e-06, ..., 7.1526e-07, + -3.5390e-08, 9.6671e-07], + [ 2.1607e-07, 8.1770e-07, 2.2724e-07, ..., 2.2352e-07, + 1.3784e-07, 6.7428e-07], + [ 1.9930e-07, -1.0971e-06, -3.3099e-06, ..., -5.7742e-07, + -3.4459e-07, 4.5635e-07]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0061, -0.0135, -0.0045, -0.0159, -0.0063, -0.0004, 0.0114, 0.0148, + 0.0159, -0.0120], device='cuda:0'), grad: tensor([-2.6584e-05, -8.4415e-06, -1.7695e-07, 1.3039e-06, 3.8184e-07, + 1.0058e-06, 2.9668e-05, 3.7309e-06, 1.9763e-06, -2.9206e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.93, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5525 re_mapping 0.0063 re_causal 0.0182 /// teacc 99.01 lr 0.00010000 +Epoch 195, weight, value: tensor([[-2.0313e-01, 8.0924e-02, -7.4921e-02, ..., -1.5367e-02, + -1.3183e-01, -1.3359e-01], + [ 1.8617e-02, 7.4323e-02, -1.0978e-01, ..., -1.0316e-01, + -3.2954e-02, 9.8162e-02], + [ 5.2170e-02, -1.0260e-01, -1.0493e-01, ..., 4.6870e-04, + -4.8094e-02, -6.8206e-02], + ..., + [ 7.3647e-02, 8.3612e-05, 1.1651e-01, ..., 6.5584e-02, + 1.5324e-01, -5.9058e-03], + [ 9.3081e-02, -1.4930e-01, -1.1758e-01, ..., -1.8936e-01, + -5.2513e-02, 1.5587e-01], + [-1.0004e-01, 7.5995e-02, 2.9680e-02, ..., -1.5351e-01, + -8.5192e-02, -2.1985e-02]], device='cuda:0'), grad: tensor([[ 4.2841e-08, -1.8626e-09, 1.3970e-07, ..., 7.6368e-08, + 2.6077e-08, 5.5879e-09], + [ 6.8583e-06, -1.4901e-08, 3.3062e-06, ..., 2.8908e-06, + 4.7311e-06, 6.1654e-07], + [ 2.0608e-05, 4.2841e-07, 8.4192e-06, ..., 9.1568e-06, + 1.4156e-05, 2.5406e-06], + ..., + [-2.8655e-05, 1.1176e-07, -1.1794e-05, ..., -1.1772e-05, + -1.9744e-05, -3.3807e-06], + [ 3.0603e-06, 3.7253e-07, 5.0254e-06, ..., 5.6401e-06, + 4.0606e-07, 0.0000e+00], + [ 4.8615e-07, -4.0792e-07, -4.4703e-08, ..., 2.2724e-07, + 3.6880e-07, 7.6368e-08]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0060, -0.0132, -0.0050, -0.0161, -0.0065, -0.0006, 0.0113, 0.0148, + 0.0163, -0.0113], device='cuda:0'), grad: tensor([ 1.8254e-07, 1.1191e-05, 3.4839e-05, -1.0565e-05, 2.5332e-07, + 5.7556e-07, 4.0978e-08, -4.6551e-05, 1.0103e-05, -1.3039e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 214.85, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5520 re_mapping 0.0067 re_causal 0.0190 /// teacc 98.99 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.2046, 0.0810, -0.0762, ..., -0.0157, -0.1333, -0.1354], + [ 0.0200, 0.0761, -0.1080, ..., -0.1036, -0.0317, 0.1004], + [ 0.0517, -0.1034, -0.1056, ..., 0.0004, -0.0486, -0.0699], + ..., + [ 0.0735, -0.0021, 0.1159, ..., 0.0659, 0.1531, -0.0083], + [ 0.0930, -0.1498, -0.1190, ..., -0.1913, -0.0531, 0.1565], + [-0.1011, 0.0758, 0.0296, ..., -0.1549, -0.0871, -0.0223]], + device='cuda:0'), grad: tensor([[ 9.3319e-07, 4.4703e-08, 2.5518e-07, ..., 4.2655e-07, + 6.1467e-07, 3.7253e-08], + [ 2.1413e-05, 1.3039e-08, 4.2506e-06, ..., 1.0245e-05, + 1.3851e-05, -1.1362e-07], + [ 2.8610e-03, 4.8429e-08, 4.0054e-04, ..., 1.3199e-03, + 1.8511e-03, 2.3097e-07], + ..., + [-2.9888e-03, -8.1956e-08, -4.2439e-04, ..., -1.3800e-03, + -1.9331e-03, -2.3656e-07], + [ 3.8415e-05, 3.4831e-07, 6.1616e-06, ..., 1.8001e-05, + 2.6152e-05, -1.7397e-06], + [ 7.4692e-07, 3.2801e-06, 7.9535e-07, ..., 3.8370e-07, + 4.3213e-07, 1.0669e-05]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0062, -0.0116, -0.0057, -0.0162, -0.0060, -0.0006, 0.0114, 0.0138, + 0.0163, -0.0117], device='cuda:0'), grad: tensor([ 1.2442e-06, 2.3648e-05, 3.0632e-03, 6.9737e-05, -3.1888e-05, + 4.8429e-07, 1.0673e-06, -3.2024e-03, 4.1693e-05, 2.9773e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 214.72, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.4992 re_mapping 0.0064 re_causal 0.0176 /// teacc 99.05 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.2055, 0.0811, -0.0771, ..., -0.0161, -0.1351, -0.1357], + [ 0.0199, 0.0762, -0.1082, ..., -0.1040, -0.0318, 0.1006], + [ 0.0498, -0.1037, -0.1066, ..., -0.0013, -0.0499, -0.0706], + ..., + [ 0.0752, -0.0023, 0.1167, ..., 0.0678, 0.1542, -0.0084], + [ 0.0931, -0.1502, -0.1196, ..., -0.1920, -0.0535, 0.1568], + [-0.1021, 0.0761, 0.0300, ..., -0.1560, -0.0876, -0.0226]], + device='cuda:0'), grad: tensor([[ 1.8440e-07, -1.1548e-07, 9.3132e-09, ..., 1.0803e-07, + 9.3132e-09, 2.7753e-07], + [ 1.0923e-05, -3.9116e-07, 5.2154e-08, ..., 1.8440e-07, + 2.7940e-08, 1.6868e-05], + [ 2.1141e-06, 3.7253e-08, 1.4901e-08, ..., -7.4692e-07, + -1.1362e-07, 3.9712e-06], + ..., + [ 6.5193e-07, 1.4901e-07, -1.5460e-07, ..., 8.1956e-08, + -9.4995e-08, 1.3672e-06], + [-1.5616e-05, 2.0117e-07, 6.1467e-08, ..., -3.7253e-09, + 8.5682e-08, -2.5213e-05], + [ 1.4715e-07, 5.4017e-08, -3.1665e-07, ..., 4.6566e-08, + 6.7055e-08, 4.1723e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0063, -0.0116, -0.0068, -0.0163, -0.0063, -0.0004, 0.0112, 0.0149, + 0.0163, -0.0116], device='cuda:0'), grad: tensor([ 3.9935e-06, 3.0100e-05, 4.1351e-06, 7.3574e-07, -9.6858e-08, + 1.7397e-06, -2.2613e-06, 2.4978e-06, -4.1604e-05, 7.0967e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 214.94, cls_loss 0.0017 cls_loss_mapping 0.0032 cls_loss_causal 0.5326 re_mapping 0.0062 re_causal 0.0180 /// teacc 98.90 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.2060, 0.0814, -0.0784, ..., -0.0170, -0.1362, -0.1357], + [ 0.0194, 0.0766, -0.1083, ..., -0.1051, -0.0327, 0.1010], + [ 0.0504, -0.1041, -0.1067, ..., -0.0007, -0.0490, -0.0710], + ..., + [ 0.0748, -0.0028, 0.1167, ..., 0.0676, 0.1542, -0.0087], + [ 0.0932, -0.1506, -0.1200, ..., -0.1927, -0.0537, 0.1570], + [-0.1029, 0.0759, 0.0306, ..., -0.1566, -0.0878, -0.0230]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, -4.6566e-08, 2.9802e-08, ..., 1.8626e-08, + 9.3132e-09, 2.4214e-08], + [-2.5332e-07, -3.5018e-07, 1.8068e-07, ..., 1.1921e-07, + 6.1467e-08, -8.6240e-07], + [ 3.4273e-07, 8.5682e-08, 2.1793e-07, ..., 1.2293e-07, + 9.6858e-08, 2.7008e-07], + ..., + [-1.8813e-07, 9.4995e-08, -3.6694e-07, ..., -2.1420e-07, + -3.0175e-07, 1.7509e-07], + [-2.6058e-06, 3.5390e-08, -9.2760e-07, ..., -1.0114e-06, + -9.8720e-08, -2.8852e-06], + [ 2.9802e-07, 2.9057e-07, 1.6019e-07, ..., 1.4342e-07, + 9.4995e-08, 4.6007e-07]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0061, -0.0121, -0.0060, -0.0163, -0.0048, -0.0005, 0.0115, 0.0140, + 0.0162, -0.0121], device='cuda:0'), grad: tensor([ 6.3330e-08, -1.5069e-06, 8.4378e-07, -7.5996e-06, -2.6077e-08, + 1.2666e-05, 4.0978e-08, -1.1176e-07, -5.6997e-06, 1.3094e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 214.66, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5187 re_mapping 0.0059 re_causal 0.0184 /// teacc 99.06 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.2064, 0.0817, -0.0788, ..., -0.0176, -0.1369, -0.1363], + [ 0.0193, 0.0775, -0.1086, ..., -0.1054, -0.0329, 0.1015], + [ 0.0505, -0.1046, -0.1070, ..., -0.0007, -0.0489, -0.0713], + ..., + [ 0.0750, -0.0028, 0.1172, ..., 0.0677, 0.1544, -0.0088], + [ 0.0927, -0.1526, -0.1205, ..., -0.1932, -0.0541, 0.1566], + [-0.1036, 0.0758, 0.0307, ..., -0.1576, -0.0883, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.0058e-07, -5.9977e-07, 9.4995e-08, ..., 5.9605e-08, + 9.1270e-08, 1.3039e-08], + [ 1.0617e-07, -1.1995e-06, 5.6624e-07, ..., 2.9802e-07, + 4.3586e-07, -2.3153e-06], + [-4.0507e-04, 7.0594e-07, -2.0373e-04, ..., -1.0914e-04, + -2.5249e-04, 1.1008e-06], + ..., + [ 4.0245e-04, 2.9616e-07, 2.0146e-04, ..., 1.0771e-04, + 2.5105e-04, 2.3842e-07], + [ 7.6182e-07, 2.1234e-07, 6.8918e-07, ..., 4.4703e-07, + 2.6822e-07, 1.0245e-07], + [ 2.7195e-07, -2.4214e-08, -6.6496e-07, ..., 7.6368e-08, + 7.6368e-08, 4.1537e-07]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0064, -0.0119, -0.0060, -0.0163, -0.0048, -0.0002, 0.0117, 0.0141, + 0.0154, -0.0123], device='cuda:0'), grad: tensor([-7.4692e-07, -3.2205e-06, -4.4203e-04, 1.9185e-06, 7.6927e-07, + 9.9652e-07, -4.6194e-07, 4.4060e-04, 2.0135e-06, 2.5146e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 214.99, cls_loss 0.0016 cls_loss_mapping 0.0023 cls_loss_causal 0.5206 re_mapping 0.0058 re_causal 0.0182 /// teacc 98.98 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.2071, 0.0817, -0.0803, ..., -0.0177, -0.1377, -0.1373], + [ 0.0178, 0.0773, -0.1095, ..., -0.1056, -0.0335, 0.1006], + [ 0.0512, -0.1042, -0.1069, ..., -0.0005, -0.0488, -0.0682], + ..., + [ 0.0755, -0.0029, 0.1180, ..., 0.0677, 0.1549, -0.0084], + [ 0.0926, -0.1530, -0.1208, ..., -0.1937, -0.0544, 0.1567], + [-0.1054, 0.0758, 0.0305, ..., -0.1580, -0.0897, -0.0242]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 1.8626e-08, ..., 5.5879e-09, + 0.0000e+00, 3.1665e-08], + [-2.2352e-07, -3.2969e-07, 4.6566e-08, ..., 1.3039e-08, + 1.6764e-08, -2.5239e-06], + [ 1.4901e-08, 5.4017e-08, 2.6077e-08, ..., 9.3132e-09, + 3.7253e-09, 1.4342e-07], + ..., + [-2.6077e-08, 1.5460e-07, -1.1176e-08, ..., -7.4506e-09, + -3.7253e-08, 3.2783e-07], + [ 1.6950e-07, 3.2783e-07, 9.8720e-08, ..., 1.4901e-08, + 1.8626e-09, 1.5274e-06], + [ 2.4214e-08, 2.3134e-06, -1.8254e-07, ..., 1.4901e-08, + 7.4506e-09, 1.3076e-06]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0067, -0.0133, -0.0046, -0.0158, -0.0047, -0.0007, 0.0120, 0.0144, + 0.0153, -0.0125], device='cuda:0'), grad: tensor([ 1.1548e-07, -3.6173e-06, 3.2596e-07, 7.0296e-06, -5.8860e-06, + -1.2249e-05, 3.1535e-06, 7.0035e-07, 2.9225e-06, 7.5325e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 214.87, cls_loss 0.0022 cls_loss_mapping 0.0048 cls_loss_causal 0.5339 re_mapping 0.0059 re_causal 0.0176 /// teacc 99.04 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.2092, 0.0791, -0.0835, ..., -0.0194, -0.1401, -0.1396], + [ 0.0183, 0.0789, -0.1096, ..., -0.1060, -0.0334, 0.1020], + [ 0.0513, -0.1053, -0.1071, ..., -0.0005, -0.0488, -0.0696], + ..., + [ 0.0756, -0.0034, 0.1183, ..., 0.0678, 0.1550, -0.0086], + [ 0.0944, -0.1545, -0.1217, ..., -0.1944, -0.0547, 0.1571], + [-0.1064, 0.0776, 0.0308, ..., -0.1591, -0.0903, -0.0248]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, -5.7742e-08, 9.8720e-08, ..., 6.1467e-08, + 5.5879e-09, 2.6077e-08], + [ 1.2107e-07, -1.7323e-07, 2.2165e-07, ..., 1.8068e-07, + 8.1956e-08, -3.2037e-07], + [ 2.9802e-08, 1.0803e-07, 3.0175e-07, ..., -7.0781e-08, + 7.4506e-09, 1.0803e-07], + ..., + [-2.1048e-07, 1.7136e-07, -4.0978e-08, ..., 1.7509e-07, + -2.3469e-07, 1.0990e-07], + [-1.6540e-06, 1.6391e-07, 1.8068e-07, ..., 5.1409e-07, + 9.3132e-09, -1.7136e-06], + [ 1.4398e-06, 7.8231e-08, -1.3039e-08, ..., 3.7812e-07, + 2.7940e-08, 1.5385e-06]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0094, -0.0124, -0.0050, -0.0155, -0.0049, -0.0024, 0.0104, 0.0145, + 0.0176, -0.0113], device='cuda:0'), grad: tensor([ 2.2724e-07, -9.8720e-08, 2.8498e-07, -2.1774e-06, 1.2480e-07, + 7.7859e-07, -3.1106e-07, 6.6869e-07, -2.8312e-06, 3.3155e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 215.21, cls_loss 0.0019 cls_loss_mapping 0.0027 cls_loss_causal 0.5164 re_mapping 0.0057 re_causal 0.0174 /// teacc 99.00 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.2102, 0.0793, -0.0847, ..., -0.0197, -0.1416, -0.1403], + [ 0.0178, 0.0790, -0.1100, ..., -0.1069, -0.0340, 0.1023], + [ 0.0514, -0.1060, -0.1074, ..., -0.0005, -0.0487, -0.0695], + ..., + [ 0.0759, -0.0034, 0.1187, ..., 0.0682, 0.1554, -0.0087], + [ 0.0945, -0.1552, -0.1222, ..., -0.1955, -0.0549, 0.1570], + [-0.1071, 0.0774, 0.0316, ..., -0.1603, -0.0908, -0.0250]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -2.6822e-07, 1.3039e-08, ..., 7.4506e-09, + 9.3132e-09, 5.5879e-09], + [ 4.8429e-08, 5.7742e-08, 8.1956e-08, ..., 1.3039e-08, + 6.5193e-08, -5.2154e-08], + [-2.9802e-08, 3.1665e-08, 3.3528e-08, ..., -6.8918e-08, + -1.3039e-08, 2.0489e-08], + ..., + [-1.7472e-06, 4.6566e-08, -2.3115e-06, ..., 1.8626e-08, + -2.1756e-06, 6.8918e-08], + [ 7.4506e-08, 2.6077e-08, 1.2293e-07, ..., 3.7253e-09, + 1.0617e-07, -5.5879e-09], + [ 1.5181e-06, 1.1921e-07, 1.7062e-06, ..., 1.6764e-08, + 1.8682e-06, 4.0978e-08]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0096, -0.0128, -0.0046, -0.0159, -0.0047, -0.0022, 0.0115, 0.0147, + 0.0170, -0.0118], device='cuda:0'), grad: tensor([-4.0233e-07, 2.1979e-07, -4.5449e-07, 1.3784e-07, -1.0990e-07, + 1.2480e-07, 1.4715e-07, -3.1292e-06, 2.8685e-07, 3.2093e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.77, cls_loss 0.0027 cls_loss_mapping 0.0036 cls_loss_causal 0.5296 re_mapping 0.0061 re_causal 0.0178 /// teacc 98.90 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.2109, 0.0792, -0.0854, ..., -0.0202, -0.1426, -0.1407], + [ 0.0140, 0.0789, -0.1142, ..., -0.1099, -0.0383, 0.0999], + [ 0.0509, -0.1065, -0.1086, ..., -0.0010, -0.0490, -0.0699], + ..., + [ 0.0801, -0.0033, 0.1238, ..., 0.0708, 0.1596, -0.0054], + [ 0.0936, -0.1540, -0.1252, ..., -0.1986, -0.0580, 0.1580], + [-0.1094, 0.0771, 0.0307, ..., -0.1629, -0.0933, -0.0262]], + device='cuda:0'), grad: tensor([[ 9.3691e-07, 3.2615e-06, 3.7253e-09, ..., 0.0000e+00, + 3.7253e-09, 3.5558e-06], + [-5.4948e-07, -7.9125e-06, 9.3132e-09, ..., 3.7253e-09, + -5.7593e-06, -1.4119e-05], + [ 1.0580e-06, 2.3544e-06, 1.1176e-08, ..., 1.8626e-09, + 5.9232e-07, 4.4741e-06], + ..., + [ 2.6338e-06, 6.1095e-07, -5.5879e-09, ..., -3.7253e-09, + 5.0142e-06, 9.8646e-06], + [-6.9402e-06, -8.1137e-06, -1.9372e-07, ..., 1.8626e-09, + -1.1176e-08, -1.6153e-05], + [ 7.0408e-07, 9.2201e-07, 2.0489e-08, ..., 1.8626e-09, + 2.7940e-08, 1.2536e-06]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0102, -0.0172, -0.0047, -0.0168, -0.0039, -0.0022, 0.0114, 0.0199, + 0.0168, -0.0127], device='cuda:0'), grad: tensor([ 1.2212e-05, -3.4153e-05, 9.6112e-06, 8.9630e-06, 1.3754e-05, + 2.8610e-06, 3.9190e-06, 1.7539e-05, -3.8922e-05, 4.2208e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 214.70, cls_loss 0.0013 cls_loss_mapping 0.0028 cls_loss_causal 0.5071 re_mapping 0.0065 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.2109, 0.0794, -0.0854, ..., -0.0203, -0.1428, -0.1408], + [ 0.0142, 0.0800, -0.1142, ..., -0.1100, -0.0380, 0.1004], + [ 0.0511, -0.1069, -0.1087, ..., -0.0009, -0.0489, -0.0700], + ..., + [ 0.0800, -0.0046, 0.1240, ..., 0.0707, 0.1595, -0.0059], + [ 0.0939, -0.1542, -0.1251, ..., -0.1989, -0.0580, 0.1584], + [-0.1115, 0.0772, 0.0305, ..., -0.1642, -0.0943, -0.0266]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, -2.9299e-06, 2.2352e-08, ..., 2.0489e-08, + 1.4901e-08, 9.3132e-09], + [ 1.3411e-07, -3.3155e-07, 1.6578e-07, ..., 1.5087e-07, + 8.7544e-08, -8.6240e-07], + [ 3.2596e-07, 1.3039e-07, 3.2037e-07, ..., 2.1979e-07, + 3.6135e-07, 2.2352e-07], + ..., + [-7.9162e-07, 5.5879e-08, -7.5065e-07, ..., -6.6310e-07, + -8.4750e-07, 2.2165e-07], + [ 2.2165e-07, 6.4261e-07, 1.7881e-07, ..., 3.7812e-07, + 8.5682e-08, 8.5682e-08], + [ 6.3330e-08, 1.8068e-07, -1.1362e-07, ..., 1.7323e-07, + 7.0781e-08, 1.1176e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0101, -0.0172, -0.0042, -0.0169, -0.0040, -0.0023, 0.0118, 0.0198, + 0.0167, -0.0130], device='cuda:0'), grad: tensor([-5.4650e-06, -1.4622e-06, 1.0133e-06, 2.2054e-06, 8.5682e-07, + -4.2804e-06, 3.8557e-07, -8.4378e-07, 6.5342e-06, 1.0636e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.72, cls_loss 0.0040 cls_loss_mapping 0.0045 cls_loss_causal 0.5233 re_mapping 0.0059 re_causal 0.0172 /// teacc 99.03 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.2115, 0.0798, -0.0860, ..., -0.0201, -0.1432, -0.1408], + [ 0.0112, 0.0807, -0.1158, ..., -0.1101, -0.0402, 0.0976], + [ 0.0510, -0.1077, -0.1090, ..., -0.0009, -0.0490, -0.0713], + ..., + [ 0.0828, -0.0060, 0.1250, ..., 0.0707, 0.1614, -0.0029], + [ 0.0941, -0.1548, -0.1252, ..., -0.1993, -0.0582, 0.1588], + [-0.1127, 0.0787, 0.0325, ..., -0.1656, -0.0938, -0.0273]], + device='cuda:0'), grad: tensor([[ 8.1956e-08, -1.8906e-06, -1.2070e-06, ..., -2.2054e-06, + 5.9605e-08, 3.5204e-07], + [ 3.1292e-07, 5.1223e-07, 6.1467e-08, ..., 1.1176e-07, + 1.0803e-07, 2.4773e-07], + [-3.1218e-06, 3.6322e-07, -5.8673e-07, ..., -3.2634e-06, + -4.2953e-06, 9.4436e-07], + ..., + [ 4.2953e-06, 1.3821e-06, 7.8045e-07, ..., 3.2280e-06, + 4.1649e-06, 1.3188e-06], + [-4.0568e-06, 4.7311e-07, 1.5087e-07, ..., 1.5274e-07, + -6.2212e-07, -4.2953e-06], + [ 1.0990e-07, 2.7511e-06, 6.6310e-07, ..., 1.5795e-06, + 4.0978e-08, 5.1223e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0097, -0.0202, -0.0045, -0.0167, -0.0048, -0.0024, 0.0114, 0.0228, + 0.0167, -0.0119], device='cuda:0'), grad: tensor([-1.0155e-05, 2.5481e-06, -6.6757e-06, 5.3197e-06, -1.3106e-05, + 1.5050e-06, 9.5740e-07, 1.5497e-05, -8.5980e-06, 1.2636e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 214.82, cls_loss 0.0021 cls_loss_mapping 0.0030 cls_loss_causal 0.4850 re_mapping 0.0061 re_causal 0.0167 /// teacc 98.86 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.2120, 0.0831, -0.0854, ..., -0.0202, -0.1439, -0.1407], + [ 0.0113, 0.0813, -0.1149, ..., -0.1090, -0.0404, 0.0979], + [ 0.0508, -0.1082, -0.1096, ..., -0.0010, -0.0492, -0.0720], + ..., + [ 0.0829, -0.0073, 0.1242, ..., 0.0702, 0.1617, -0.0032], + [ 0.0943, -0.1554, -0.1252, ..., -0.2000, -0.0583, 0.1590], + [-0.1142, 0.0767, 0.0334, ..., -0.1664, -0.0940, -0.0275]], + device='cuda:0'), grad: tensor([[-1.8626e-08, -7.5623e-07, 3.7253e-09, ..., 5.5879e-08, + 5.0291e-08, 7.2643e-08], + [ 9.6858e-08, -1.3411e-07, 1.1176e-08, ..., 1.3784e-07, + 1.3784e-07, -4.3586e-07], + [-3.2056e-06, 1.0245e-07, -1.4715e-07, ..., -3.0622e-06, + -3.0156e-06, 2.9802e-08], + ..., + [ 1.6894e-06, 1.5832e-07, 7.8231e-08, ..., 1.5888e-06, + 1.5534e-06, 2.5705e-07], + [ 6.8545e-07, 2.1234e-07, 5.5879e-08, ..., 6.4448e-07, + 6.3702e-07, 3.9116e-08], + [ 9.6858e-08, 6.6496e-07, -3.7253e-08, ..., 5.0291e-08, + 5.7742e-08, 5.9605e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0064, -0.0198, -0.0050, -0.0171, -0.0054, -0.0023, 0.0111, 0.0224, + 0.0167, -0.0139], device='cuda:0'), grad: tensor([-1.2182e-06, -1.7323e-07, -8.4043e-06, 2.5332e-06, -1.4398e-06, + -2.4401e-06, 1.4156e-06, 5.0887e-06, 2.7269e-06, 1.8999e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 214.92, cls_loss 0.0016 cls_loss_mapping 0.0034 cls_loss_causal 0.4859 re_mapping 0.0064 re_causal 0.0187 /// teacc 98.91 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.2121, 0.0811, -0.0857, ..., -0.0202, -0.1441, -0.1409], + [ 0.0114, 0.0823, -0.1148, ..., -0.1091, -0.0403, 0.0981], + [ 0.0509, -0.1092, -0.1097, ..., -0.0011, -0.0491, -0.0722], + ..., + [ 0.0828, -0.0084, 0.1241, ..., 0.0702, 0.1616, -0.0033], + [ 0.0943, -0.1557, -0.1253, ..., -0.2003, -0.0583, 0.1590], + [-0.1145, 0.0784, 0.0338, ..., -0.1672, -0.0943, -0.0279]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -3.5390e-08, 3.7253e-09, ..., 3.7253e-09, + 7.4506e-09, 7.4506e-09], + [-5.9791e-07, -9.0525e-07, 3.3528e-08, ..., 2.7940e-08, + -1.8775e-06, -2.9281e-06], + [ 1.5274e-07, 8.5682e-08, 8.1956e-08, ..., 4.8429e-08, + 2.4028e-07, 2.7753e-07], + ..., + [-4.7870e-07, 7.1526e-07, -7.4320e-07, ..., -5.6252e-07, + 8.0653e-07, 2.3264e-06], + [ 2.0862e-07, 3.9116e-08, 2.0489e-07, ..., 1.6578e-07, + 2.1793e-07, -3.3528e-08], + [ 5.0291e-08, 8.5682e-08, 1.3039e-08, ..., 1.1176e-08, + 1.2480e-07, 1.8626e-07]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0083, -0.0197, -0.0050, -0.0167, -0.0053, -0.0025, 0.0112, 0.0223, + 0.0166, -0.0123], device='cuda:0'), grad: tensor([ 6.3330e-08, -7.5549e-06, 8.2888e-07, 7.7672e-07, 1.2480e-07, + 2.5332e-07, -3.1851e-07, 4.8205e-06, 4.3586e-07, 5.6438e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 214.82, cls_loss 0.0016 cls_loss_mapping 0.0039 cls_loss_causal 0.5377 re_mapping 0.0063 re_causal 0.0178 /// teacc 98.92 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.2128, 0.0815, -0.0863, ..., -0.0232, -0.1447, -0.1403], + [ 0.0114, 0.0831, -0.1150, ..., -0.1093, -0.0405, 0.0982], + [ 0.0528, -0.1090, -0.1077, ..., 0.0005, -0.0476, -0.0729], + ..., + [ 0.0824, -0.0092, 0.1236, ..., 0.0691, 0.1612, -0.0033], + [ 0.0941, -0.1561, -0.1255, ..., -0.2016, -0.0584, 0.1592], + [-0.1164, 0.0781, 0.0337, ..., -0.1701, -0.0952, -0.0288]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 1.3039e-08, 5.7742e-08, ..., 1.4901e-08, + 4.2841e-08, 1.6764e-08], + [ 1.3597e-07, -1.7509e-07, 2.7195e-07, ..., 7.6368e-08, + 1.8440e-07, -4.7684e-07], + [ 2.3842e-07, 7.0781e-08, 2.6636e-07, ..., 9.6858e-08, + 2.1048e-07, 2.8312e-07], + ..., + [-8.0690e-06, 1.8999e-07, -1.7390e-05, ..., -3.7346e-06, + -1.7703e-05, 2.7195e-07], + [-6.8918e-08, -9.5926e-07, 3.2224e-07, ..., 8.0094e-08, + 2.3842e-07, -4.7348e-06], + [ 7.5437e-06, -1.1303e-05, 1.0550e-05, ..., 3.4980e-06, + 8.8662e-06, 3.1162e-06]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0082, -0.0197, -0.0034, -0.0187, -0.0055, -0.0003, 0.0114, 0.0221, + 0.0164, -0.0128], device='cuda:0'), grad: tensor([ 1.6764e-07, -1.8254e-07, 1.1064e-06, -1.1817e-05, 4.5300e-05, + 1.5467e-05, 2.2091e-06, -4.2379e-05, -1.3463e-05, 3.5949e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 214.92, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.5284 re_mapping 0.0058 re_causal 0.0175 /// teacc 98.99 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.2131, 0.0816, -0.0868, ..., -0.0233, -0.1448, -0.1406], + [ 0.0113, 0.0831, -0.1150, ..., -0.1095, -0.0405, 0.0983], + [ 0.0529, -0.1103, -0.1079, ..., 0.0004, -0.0476, -0.0736], + ..., + [ 0.0821, -0.0110, 0.1221, ..., 0.0662, 0.1609, -0.0034], + [ 0.0946, -0.1564, -0.1257, ..., -0.2020, -0.0584, 0.1603], + [-0.1160, 0.0793, 0.0367, ..., -0.1706, -0.0929, -0.0282]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, -7.4506e-08, 1.3039e-08, ..., 9.3132e-09, + 5.5879e-09, 5.5879e-09], + [ 1.6391e-07, 1.6950e-07, 7.6368e-08, ..., 4.8429e-08, + 5.4017e-08, -1.2293e-07], + [ 2.6077e-07, 1.0431e-07, 2.3097e-07, ..., 1.1735e-07, + 2.1793e-07, 5.0291e-08], + ..., + [-9.5740e-07, 1.0245e-07, -7.4878e-07, ..., -4.8429e-07, + -7.5996e-07, 8.7544e-08], + [ 1.1921e-07, 3.0361e-07, 1.0058e-07, ..., 8.3819e-08, + 8.9407e-08, -2.6077e-07], + [ 2.9057e-06, 1.4585e-06, -1.3784e-07, ..., 4.2841e-08, + 9.4995e-08, 2.1048e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0082, -0.0197, -0.0037, -0.0172, -0.0072, -0.0003, 0.0111, 0.0215, + 0.0167, -0.0109], device='cuda:0'), grad: tensor([ 3.7253e-09, 6.5006e-07, 8.0280e-07, 3.8818e-06, 6.2585e-06, + 7.5996e-06, 8.0392e-06, -1.2219e-06, 1.3299e-06, -2.7344e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 215.02, cls_loss 0.0019 cls_loss_mapping 0.0035 cls_loss_causal 0.5111 re_mapping 0.0059 re_causal 0.0165 /// teacc 98.92 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.2140, 0.0817, -0.0878, ..., -0.0234, -0.1459, -0.1414], + [ 0.0113, 0.0857, -0.1151, ..., -0.1098, -0.0404, 0.0987], + [ 0.0531, -0.1108, -0.1082, ..., 0.0003, -0.0475, -0.0741], + ..., + [ 0.0821, -0.0120, 0.1221, ..., 0.0660, 0.1609, -0.0035], + [ 0.0953, -0.1569, -0.1259, ..., -0.2022, -0.0585, 0.1617], + [-0.1165, 0.0797, 0.0375, ..., -0.1711, -0.0931, -0.0274]], + device='cuda:0'), grad: tensor([[ 1.1921e-07, 9.3132e-10, 1.8626e-08, ..., 1.5832e-07, + 8.8476e-08, 6.3330e-08], + [ 2.7083e-06, 2.4773e-07, 2.1476e-06, ..., 1.8533e-06, + 2.2408e-06, -2.2817e-07], + [-1.5711e-06, 1.8440e-07, 5.3737e-07, ..., -2.7400e-06, + -1.0794e-06, 3.0082e-07], + ..., + [-4.1127e-06, -1.2834e-06, -4.1500e-06, ..., -2.2538e-06, + -3.4645e-06, -1.5618e-06], + [ 7.1805e-07, 2.5611e-07, 4.0513e-07, ..., 7.4133e-07, + 5.6624e-07, 2.9430e-07], + [ 3.6974e-07, 1.7416e-07, 2.4214e-07, ..., 3.1292e-07, + 3.1572e-07, 2.4308e-07]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0080, -0.0194, -0.0037, -0.0171, -0.0081, -0.0004, 0.0096, 0.0214, + 0.0172, -0.0103], device='cuda:0'), grad: tensor([ 5.0012e-07, 2.2743e-06, -6.1840e-06, 4.1649e-06, 1.4417e-06, + -9.7789e-07, 1.3737e-06, -6.2175e-06, 2.3190e-06, 1.2722e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 214.96, cls_loss 0.0015 cls_loss_mapping 0.0027 cls_loss_causal 0.5111 re_mapping 0.0061 re_causal 0.0177 /// teacc 99.08 lr 0.00010000 +Epoch 211, weight, value: tensor([[-2.1572e-01, 8.1682e-02, -8.8811e-02, ..., -2.4072e-02, + -1.4668e-01, -1.4369e-01], + [ 1.1355e-02, 8.6367e-02, -1.1516e-01, ..., -1.0995e-01, + -4.0405e-02, 9.8854e-02], + [ 5.3085e-02, -1.1158e-01, -1.0865e-01, ..., -1.0879e-05, + -4.7602e-02, -7.4325e-02], + ..., + [ 8.2124e-02, -1.2542e-02, 1.2229e-01, ..., 6.6095e-02, + 1.6103e-01, -3.5835e-03], + [ 9.5953e-02, -1.5709e-01, -1.2617e-01, ..., -2.0381e-01, + -5.8390e-02, 1.6284e-01], + [-1.1716e-01, 7.9502e-02, 3.7543e-02, ..., -1.7193e-01, + -9.3474e-02, -2.7964e-02]], device='cuda:0'), grad: tensor([[ 3.0734e-08, 1.1623e-06, 6.8918e-08, ..., 8.8476e-08, + 2.4214e-08, 4.9639e-07], + [ 2.8964e-07, 6.5565e-06, 9.2760e-07, ..., 2.7474e-07, + 1.4715e-07, 7.7784e-06], + [-3.2745e-06, 3.0827e-07, 5.9605e-08, ..., -2.8685e-06, + -2.0936e-06, 3.7625e-07], + ..., + [ 1.1250e-06, 1.2023e-06, 3.4925e-07, ..., 9.9093e-07, + 9.0059e-07, 1.5721e-06], + [ 1.2433e-06, 1.5765e-05, 1.9409e-06, ..., 1.1893e-06, + 7.4320e-07, 1.8016e-05], + [ 4.9360e-08, -7.0095e-05, -9.8273e-06, ..., 9.3132e-08, + -8.2888e-08, -8.7738e-05]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0081, -0.0193, -0.0037, -0.0167, -0.0080, -0.0003, 0.0089, 0.0214, + 0.0176, -0.0105], device='cuda:0'), grad: tensor([ 4.0680e-06, 2.2709e-05, -7.7635e-06, 2.7828e-06, 1.5783e-04, + -3.0100e-05, 2.2352e-05, 6.9812e-06, 5.6267e-05, -2.3520e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 215.14, cls_loss 0.0028 cls_loss_mapping 0.0043 cls_loss_causal 0.5010 re_mapping 0.0060 re_causal 0.0169 /// teacc 98.94 lr 0.00010000 +Epoch 212, weight, value: tensor([[-2.1496e-01, 8.0081e-02, -8.9318e-02, ..., -2.4282e-02, + -1.4753e-01, -1.4080e-01], + [ 9.3222e-03, 8.7015e-02, -1.1683e-01, ..., -1.1007e-01, + -4.2170e-02, 9.6724e-02], + [ 5.1880e-02, -1.1267e-01, -1.0964e-01, ..., -7.5343e-05, + -4.8630e-02, -7.8786e-02], + ..., + [ 8.3954e-02, -1.2307e-02, 1.2379e-01, ..., 6.5576e-02, + 1.6281e-01, -1.3846e-03], + [ 9.8334e-02, -1.5848e-01, -1.2403e-01, ..., -2.0133e-01, + -5.5925e-02, 1.6416e-01], + [-1.1890e-01, 8.0957e-02, 3.7209e-02, ..., -1.7360e-01, + -9.4773e-02, -2.8545e-02]], device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.3269e-05, 1.7695e-08, ..., -2.2445e-06, + 1.8626e-09, 5.5879e-09], + [-5.4017e-08, -1.1921e-07, 1.4901e-08, ..., 1.3039e-08, + 1.8626e-09, -6.2026e-07], + [-2.7940e-09, 5.3644e-07, 4.1910e-08, ..., 5.4948e-08, + -2.7008e-08, 4.9360e-08], + ..., + [ 3.8184e-08, 3.0082e-07, -0.0000e+00, ..., 3.0734e-08, + -0.0000e+00, 4.9733e-07], + [-5.9605e-08, 2.5705e-07, 7.1712e-08, ..., 2.1420e-08, + -9.3132e-10, -5.0291e-08], + [ 5.5879e-09, 1.1407e-05, -6.1467e-07, ..., 1.9725e-06, + 1.8626e-09, 2.6450e-07]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0097, -0.0212, -0.0055, -0.0169, -0.0080, -0.0001, 0.0090, 0.0232, + 0.0183, -0.0093], device='cuda:0'), grad: tensor([-1.6922e-06, -5.9884e-07, 1.0878e-06, 1.6335e-06, -1.3970e-07, + 2.7120e-06, -2.4840e-05, 9.0618e-07, 7.4320e-07, 2.0117e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 214.86, cls_loss 0.0026 cls_loss_mapping 0.0042 cls_loss_causal 0.5133 re_mapping 0.0058 re_causal 0.0168 /// teacc 98.99 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.2155, 0.0803, -0.0910, ..., -0.0237, -0.1500, -0.1407], + [ 0.0092, 0.0893, -0.1175, ..., -0.1120, -0.0422, 0.0968], + [ 0.0523, -0.1149, -0.1098, ..., 0.0004, -0.0481, -0.0799], + ..., + [ 0.0839, -0.0161, 0.1233, ..., 0.0644, 0.1624, -0.0014], + [ 0.0983, -0.1598, -0.1245, ..., -0.2020, -0.0561, 0.1645], + [-0.1207, 0.0812, 0.0392, ..., -0.1767, -0.0933, -0.0291]], + device='cuda:0'), grad: tensor([[ 1.7695e-07, 1.3195e-05, 3.2596e-08, ..., 6.5193e-09, + 1.3970e-08, 2.0862e-07], + [ 1.8869e-06, 3.0547e-07, 1.3411e-07, ..., 2.4401e-07, + 8.1025e-08, 1.4734e-06], + [-9.1046e-06, 2.5425e-07, 2.3656e-07, ..., -5.8562e-06, + -5.7630e-06, 1.5246e-06], + ..., + [ 9.5293e-06, 5.8021e-07, -8.3167e-07, ..., 5.0627e-06, + 4.9174e-06, 1.3327e-06], + [-4.8093e-06, 2.9191e-05, 2.2538e-07, ..., 2.6450e-07, + 9.1270e-08, -6.2436e-06], + [ 8.8476e-08, 1.1645e-05, 7.3574e-08, ..., 9.4064e-08, + 4.0978e-08, 2.0023e-07]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0095, -0.0211, -0.0056, -0.0157, -0.0080, -0.0004, 0.0089, 0.0229, + 0.0181, -0.0088], device='cuda:0'), grad: tensor([ 1.2267e-04, 7.6741e-06, -1.0386e-05, 5.1737e-05, 5.1409e-07, + -7.3004e-04, 3.4094e-04, 1.7092e-05, 1.7202e-04, 2.7552e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 214.89, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5215 re_mapping 0.0058 re_causal 0.0178 /// teacc 98.93 lr 0.00010000 +Epoch 214, weight, value: tensor([[-2.1595e-01, 8.0279e-02, -9.1560e-02, ..., -2.5181e-02, + -1.5106e-01, -1.4096e-01], + [ 9.2960e-03, 8.9657e-02, -1.1715e-01, ..., -1.1157e-01, + -4.1954e-02, 9.6957e-02], + [ 5.1385e-02, -1.1609e-01, -1.1120e-01, ..., -8.3849e-05, + -4.9176e-02, -8.0157e-02], + ..., + [ 8.4175e-02, -1.6419e-02, 1.2342e-01, ..., 6.4589e-02, + 1.6275e-01, -1.4949e-03], + [ 9.8528e-02, -1.6101e-01, -1.2455e-01, ..., -2.0221e-01, + -5.6079e-02, 1.6502e-01], + [-1.2168e-01, 8.0904e-02, 3.9426e-02, ..., -1.7768e-01, + -9.3383e-02, -2.9649e-02]], device='cuda:0'), grad: tensor([[ 8.3819e-09, -4.9360e-08, 1.8626e-09, ..., 3.7253e-09, + 9.3132e-10, 3.7253e-09], + [-1.3970e-08, -1.8347e-07, 1.3970e-08, ..., 1.9558e-08, + 7.4506e-09, -3.2783e-07], + [-4.9639e-07, 1.4901e-08, 5.0291e-08, ..., -3.3248e-07, + -3.1386e-07, 2.2352e-08], + ..., + [ 3.8091e-07, 1.2107e-07, -5.1223e-08, ..., 2.4680e-07, + 2.8964e-07, 1.7323e-07], + [ 1.3970e-08, 1.5367e-07, 8.0094e-08, ..., 6.8918e-08, + 1.0245e-08, 1.2107e-08], + [ 8.3819e-09, 9.3132e-08, -7.4506e-08, ..., 7.4506e-09, + -1.4901e-08, 7.2643e-08]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0096, -0.0210, -0.0064, -0.0157, -0.0079, 0.0004, 0.0088, 0.0229, + 0.0179, -0.0091], device='cuda:0'), grad: tensor([-2.4214e-08, -4.2282e-07, -9.4622e-07, 1.3784e-07, -8.2888e-08, + -6.5565e-07, 3.1944e-07, 9.7044e-07, 5.2527e-07, 1.9837e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 214.89, cls_loss 0.0016 cls_loss_mapping 0.0033 cls_loss_causal 0.5079 re_mapping 0.0059 re_causal 0.0176 /// teacc 99.04 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.2164, 0.0808, -0.0922, ..., -0.0252, -0.1520, -0.1398], + [ 0.0094, 0.0900, -0.1166, ..., -0.1106, -0.0417, 0.0972], + [ 0.0512, -0.1165, -0.1117, ..., -0.0002, -0.0494, -0.0805], + ..., + [ 0.0842, -0.0184, 0.1231, ..., 0.0643, 0.1627, -0.0018], + [ 0.0986, -0.1622, -0.1247, ..., -0.2024, -0.0561, 0.1651], + [-0.1222, 0.0807, 0.0397, ..., -0.1783, -0.0934, -0.0288]], + device='cuda:0'), grad: tensor([[-0.0000e+00, -5.1223e-08, 4.6566e-09, ..., -9.3132e-10, + 1.8626e-09, 1.3039e-08], + [ 4.5635e-08, -1.5646e-07, 3.7253e-08, ..., 5.2154e-08, + -6.5193e-09, -3.2969e-07], + [ 4.0047e-08, 8.3819e-09, 1.1455e-07, ..., 1.7881e-07, + 2.4214e-08, 1.8626e-08], + ..., + [ 1.5832e-08, 9.4995e-08, 1.1828e-07, ..., 2.7474e-07, + -1.7695e-08, 2.2631e-07], + [-2.8126e-07, 9.3132e-08, 7.4506e-08, ..., 1.1269e-07, + 1.2107e-08, -4.8336e-07], + [ 2.2352e-08, 2.8871e-08, -4.8429e-08, ..., 6.5193e-09, + 4.6566e-09, 8.2888e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([-9.1212e-03, -2.0824e-02, -6.5486e-03, -1.5617e-02, -7.6305e-03, + 9.7456e-05, 8.8742e-03, 2.2635e-02, 1.7739e-02, -9.2100e-03], + device='cuda:0'), grad: tensor([-1.4901e-08, -3.1572e-07, 2.5705e-07, 4.2990e-06, -1.0617e-07, + -4.5225e-06, -2.2911e-07, 5.9232e-07, -8.3819e-08, 1.2200e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.86, cls_loss 0.0021 cls_loss_mapping 0.0032 cls_loss_causal 0.5470 re_mapping 0.0061 re_causal 0.0177 /// teacc 98.95 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.2181, 0.0809, -0.0936, ..., -0.0256, -0.1548, -0.1410], + [ 0.0104, 0.0903, -0.1167, ..., -0.1098, -0.0400, 0.0973], + [ 0.0506, -0.1172, -0.1109, ..., 0.0019, -0.0504, -0.0811], + ..., + [ 0.0835, -0.0191, 0.1221, ..., 0.0624, 0.1615, -0.0019], + [ 0.0995, -0.1614, -0.1250, ..., -0.2030, -0.0562, 0.1671], + [-0.1213, 0.0806, 0.0414, ..., -0.1795, -0.0912, -0.0289]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -7.5717e-07, 1.0338e-07, ..., 5.4948e-08, + 3.5390e-08, 4.6566e-09], + [ 8.5682e-08, 6.3330e-08, 2.5891e-06, ..., 1.4016e-06, + 8.5961e-07, -2.3283e-07], + [ 2.1420e-08, 1.4901e-08, 6.3330e-08, ..., 5.0291e-08, + -5.7090e-07, 6.5193e-09], + ..., + [ 1.0416e-05, 3.6716e-05, 5.1165e-04, ..., 2.8229e-04, + 9.4414e-05, 9.6858e-08], + [ 1.3039e-08, 1.4063e-07, 5.3458e-07, ..., 3.0175e-07, + 1.0803e-07, 9.1270e-08], + [ 2.1700e-07, 1.4026e-06, 2.3581e-06, ..., 1.3243e-06, + 4.6287e-07, 5.2527e-07]], device='cuda:0') +Epoch 216, bias, value: tensor([-9.1327e-03, -2.0039e-02, -7.6493e-03, -1.5267e-02, -7.6808e-03, + -7.8043e-05, 9.0725e-03, 2.2007e-02, 1.8396e-02, -8.6250e-03], + device='cuda:0'), grad: tensor([-1.0272e-06, 4.0457e-06, -1.3206e-06, -6.9094e-04, -1.4203e-06, + 4.5169e-07, 7.7114e-07, 6.8283e-04, 9.8441e-07, 5.6922e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 215.04, cls_loss 0.0017 cls_loss_mapping 0.0025 cls_loss_causal 0.5197 re_mapping 0.0060 re_causal 0.0173 /// teacc 99.01 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.2187, 0.0811, -0.0943, ..., -0.0257, -0.1555, -0.1411], + [ 0.0107, 0.0905, -0.1168, ..., -0.1095, -0.0394, 0.0973], + [ 0.0502, -0.1181, -0.1108, ..., 0.0024, -0.0509, -0.0818], + ..., + [ 0.0835, -0.0193, 0.1218, ..., 0.0617, 0.1614, -0.0019], + [ 0.1000, -0.1609, -0.1250, ..., -0.2030, -0.0562, 0.1683], + [-0.1221, 0.0803, 0.0415, ..., -0.1806, -0.0914, -0.0297]], + device='cuda:0'), grad: tensor([[ 4.5355e-07, -5.4426e-06, 4.2841e-08, ..., 1.9278e-07, + 2.0675e-07, 8.3819e-09], + [ 2.4401e-07, -3.1106e-07, 2.7567e-07, ..., 2.1607e-07, + 1.2200e-07, -9.5088e-07], + [-9.8199e-06, 7.7300e-08, 3.9022e-07, ..., -4.0382e-06, + -4.2915e-06, -7.0781e-08], + ..., + [ 3.1516e-06, 1.9185e-07, -6.7763e-06, ..., -5.5321e-07, + -9.5181e-07, 5.8860e-07], + [ 1.8207e-06, 8.2608e-07, 3.2689e-07, ..., 7.1432e-07, + 7.5810e-07, 8.4750e-08], + [-2.9802e-08, 1.8887e-06, -3.2317e-07, ..., 1.2200e-07, + 2.2165e-07, 2.7753e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([-9.0201e-03, -1.9812e-02, -8.2540e-03, -1.5311e-02, -7.3845e-03, + 8.2756e-05, 9.5756e-03, 2.1864e-02, 1.8853e-02, -8.9927e-03], + device='cuda:0'), grad: tensor([-1.1317e-05, -7.4971e-07, -1.6227e-05, 8.4043e-06, 1.4044e-06, + -6.6124e-08, 6.8694e-06, 3.4645e-06, 6.4597e-06, 1.7872e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 214.69, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5388 re_mapping 0.0058 re_causal 0.0186 /// teacc 98.89 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.2178, 0.0811, -0.0938, ..., -0.0255, -0.1538, -0.1398], + [ 0.0107, 0.0907, -0.1168, ..., -0.1094, -0.0393, 0.0973], + [ 0.0505, -0.1188, -0.1104, ..., 0.0026, -0.0508, -0.0827], + ..., + [ 0.0833, -0.0195, 0.1218, ..., 0.0615, 0.1613, -0.0020], + [ 0.1002, -0.1616, -0.1250, ..., -0.2032, -0.0563, 0.1686], + [-0.1233, 0.0803, 0.0412, ..., -0.1818, -0.0918, -0.0301]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, -6.7987e-08, 4.5635e-08, ..., 4.5635e-08, + 1.2107e-08, 2.7940e-09], + [ 1.8813e-06, -5.5879e-09, 8.6613e-07, ..., 7.0315e-07, + 1.6298e-06, 1.2945e-07], + [ 4.7218e-07, 2.3283e-07, 1.8924e-06, ..., 1.4808e-06, + 1.9185e-07, 8.1956e-08], + ..., + [-3.0976e-06, 1.3970e-08, -2.2855e-06, ..., -9.7230e-07, + -2.3860e-06, -2.0396e-07], + [ 8.3819e-09, 4.3772e-08, 2.4866e-07, ..., 2.4494e-07, + 6.3330e-08, -8.2888e-08], + [ 2.1514e-07, 6.4261e-08, 1.7788e-07, ..., 1.0617e-07, + 1.7602e-07, 2.2352e-08]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0089, -0.0197, -0.0082, -0.0152, -0.0074, 0.0010, 0.0086, 0.0218, + 0.0187, -0.0092], device='cuda:0'), grad: tensor([ 1.3970e-08, 3.2410e-06, 2.8927e-06, -3.8296e-06, 6.9290e-07, + 7.5903e-07, -8.9314e-07, -3.8818e-06, 5.1595e-07, 4.7963e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 214.89, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.5245 re_mapping 0.0057 re_causal 0.0175 /// teacc 99.00 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.2202, 0.0811, -0.0958, ..., -0.0259, -0.1548, -0.1401], + [ 0.0105, 0.0890, -0.1170, ..., -0.1096, -0.0393, 0.0964], + [ 0.0511, -0.1165, -0.1110, ..., 0.0025, -0.0508, -0.0803], + ..., + [ 0.0825, -0.0203, 0.1212, ..., 0.0616, 0.1606, -0.0024], + [ 0.1029, -0.1622, -0.1223, ..., -0.2032, -0.0535, 0.1712], + [-0.1246, 0.0804, 0.0414, ..., -0.1824, -0.0923, -0.0305]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, -4.0699e-07, 3.6322e-08, ..., 4.6566e-09, + 1.0245e-08, 1.5832e-08], + [-2.7474e-06, -1.6298e-07, 1.2666e-07, ..., 3.9116e-08, + 7.2643e-08, -5.9381e-06], + [ 5.6904e-07, 1.1176e-08, 5.6531e-07, ..., 1.8068e-07, + 5.0850e-07, 2.0396e-07], + ..., + [-9.3132e-10, 1.7136e-07, -1.0636e-06, ..., -3.7253e-07, + -1.0207e-06, 2.0601e-06], + [ 1.6820e-06, 4.9174e-07, 4.2841e-07, ..., 5.4017e-08, + 1.6205e-07, 3.0492e-06], + [ 1.7602e-07, -9.0245e-07, -2.1029e-06, ..., 3.4459e-08, + 8.9407e-08, 1.5087e-07]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0090, -0.0215, -0.0057, -0.0151, -0.0074, 0.0007, 0.0085, 0.0214, + 0.0202, -0.0092], device='cuda:0'), grad: tensor([-4.4424e-07, -8.1956e-06, 1.0878e-06, 3.5353e-06, 3.3639e-06, + -3.8445e-06, -6.7055e-08, 1.5665e-06, 5.8375e-06, -2.8647e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 214.88, cls_loss 0.0020 cls_loss_mapping 0.0026 cls_loss_causal 0.5364 re_mapping 0.0055 re_causal 0.0171 /// teacc 98.95 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.2212, 0.0813, -0.0969, ..., -0.0259, -0.1550, -0.1403], + [ 0.0093, 0.0888, -0.1192, ..., -0.1098, -0.0408, 0.0952], + [ 0.0511, -0.1160, -0.1115, ..., 0.0024, -0.0509, -0.0798], + ..., + [ 0.0838, -0.0209, 0.1236, ..., 0.0617, 0.1625, -0.0012], + [ 0.1029, -0.1630, -0.1225, ..., -0.2033, -0.0536, 0.1712], + [-0.1258, 0.0804, 0.0414, ..., -0.1834, -0.0929, -0.0309]], + device='cuda:0'), grad: tensor([[ 7.7393e-07, 1.4622e-07, 5.5879e-09, ..., 6.2399e-08, + 1.2107e-08, 7.8883e-07], + [-3.1125e-06, -8.4471e-07, 2.7008e-08, ..., 1.0245e-08, + -2.2110e-06, -3.3006e-06], + [ 1.8617e-06, 4.9081e-07, 2.0582e-07, ..., 5.3085e-08, + 1.3141e-06, 1.7863e-06], + ..., + [ 6.9942e-07, 4.5542e-07, -2.8592e-07, ..., -5.4948e-08, + 4.8429e-07, 1.2247e-06], + [-1.2023e-06, -2.2631e-07, 2.4214e-08, ..., -1.0431e-07, + 8.3819e-08, -1.2619e-06], + [ 5.7090e-07, 3.7439e-07, 3.2596e-08, ..., 3.7253e-08, + 2.3749e-07, 7.3295e-07]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0090, -0.0228, -0.0052, -0.0154, -0.0077, 0.0009, 0.0085, 0.0228, + 0.0200, -0.0092], device='cuda:0'), grad: tensor([ 1.8226e-06, -1.2957e-05, 7.2494e-06, 2.0489e-07, -1.4501e-06, + 1.4901e-07, 8.3633e-07, 4.2543e-06, -2.5332e-06, 2.3879e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 214.85, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5089 re_mapping 0.0056 re_causal 0.0162 /// teacc 99.05 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.2217, 0.0813, -0.0978, ..., -0.0261, -0.1555, -0.1405], + [ 0.0099, 0.0891, -0.1175, ..., -0.1078, -0.0401, 0.0954], + [ 0.0511, -0.1161, -0.1118, ..., 0.0025, -0.0509, -0.0799], + ..., + [ 0.0833, -0.0215, 0.1222, ..., 0.0611, 0.1619, -0.0014], + [ 0.1032, -0.1637, -0.1223, ..., -0.2037, -0.0534, 0.1718], + [-0.1278, 0.0803, 0.0414, ..., -0.1845, -0.0935, -0.0321]], + device='cuda:0'), grad: tensor([[ 5.0291e-08, 6.1877e-06, 7.5437e-08, ..., 1.1176e-08, + 3.0734e-08, 1.5926e-07], + [ 5.0571e-07, 3.4478e-06, 1.6931e-06, ..., 1.8906e-07, + 4.3400e-07, 1.2098e-06], + [-1.8468e-06, 6.2492e-07, 2.6263e-07, ..., -7.9721e-07, + -1.1632e-06, 3.3714e-07], + ..., + [-1.5842e-06, -1.6298e-07, -4.1239e-06, ..., 7.3016e-07, + -3.3379e-06, 4.8708e-07], + [ 4.6566e-08, 1.1809e-06, 5.8580e-07, ..., 2.2259e-07, + 6.0536e-08, 3.0175e-07], + [ 3.8669e-06, 9.3654e-06, 8.2031e-06, ..., 5.1409e-07, + 3.7625e-06, 2.4941e-06]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0090, -0.0221, -0.0052, -0.0158, -0.0077, 0.0009, 0.0089, 0.0219, + 0.0200, -0.0094], device='cuda:0'), grad: tensor([ 1.6138e-05, 9.0301e-06, -1.9558e-06, -1.6108e-05, 8.7470e-06, + -2.1793e-07, -4.4733e-05, -2.6301e-06, 3.4086e-06, 2.8327e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.93, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.4853 re_mapping 0.0055 re_causal 0.0170 /// teacc 99.01 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.2220, 0.0816, -0.0984, ..., -0.0266, -0.1558, -0.1408], + [ 0.0100, 0.0894, -0.1175, ..., -0.1079, -0.0402, 0.0955], + [ 0.0511, -0.1161, -0.1122, ..., 0.0023, -0.0509, -0.0799], + ..., + [ 0.0834, -0.0218, 0.1224, ..., 0.0612, 0.1621, -0.0015], + [ 0.1032, -0.1645, -0.1225, ..., -0.2040, -0.0535, 0.1718], + [-0.1287, 0.0800, 0.0415, ..., -0.1850, -0.0939, -0.0329]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, 3.4459e-08, 1.3970e-08, ..., 3.4459e-08, + 3.0734e-08, 2.1420e-08], + [-5.5879e-09, -9.2667e-07, 2.2817e-07, ..., 1.0338e-07, + 1.7602e-07, -2.5686e-06], + [ 2.6915e-07, 4.7497e-08, 3.5483e-07, ..., -7.0781e-08, + 6.3330e-08, 9.8720e-08], + ..., + [-1.7174e-06, 3.4925e-07, -1.3728e-06, ..., -5.6066e-07, + -1.0319e-06, 7.0781e-07], + [ 7.5437e-08, 4.0699e-07, 1.8813e-07, ..., 2.0675e-07, + 2.8033e-07, 7.0315e-07], + [ 3.1944e-07, 1.7108e-06, 1.4435e-07, ..., 4.2841e-08, + 1.7881e-07, 3.2037e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0088, -0.0220, -0.0052, -0.0160, -0.0077, 0.0005, 0.0096, 0.0219, + 0.0198, -0.0097], device='cuda:0'), grad: tensor([ 2.6077e-07, -3.1963e-06, -3.5390e-08, 4.7684e-06, -3.6415e-07, + -7.0743e-06, 6.5379e-07, -1.3476e-06, 1.7155e-06, 4.6268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.61, cls_loss 0.0017 cls_loss_mapping 0.0024 cls_loss_causal 0.5264 re_mapping 0.0056 re_causal 0.0169 /// teacc 98.92 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.2221, 0.0817, -0.0992, ..., -0.0271, -0.1563, -0.1411], + [ 0.0098, 0.0905, -0.1177, ..., -0.1080, -0.0405, 0.0956], + [ 0.0511, -0.1162, -0.1124, ..., 0.0024, -0.0509, -0.0800], + ..., + [ 0.0838, -0.0231, 0.1230, ..., 0.0613, 0.1628, -0.0015], + [ 0.1031, -0.1654, -0.1229, ..., -0.2046, -0.0536, 0.1718], + [-0.1331, 0.0796, 0.0401, ..., -0.1865, -0.0962, -0.0336]], + device='cuda:0'), grad: tensor([[ 1.1083e-07, 9.2201e-08, 8.6613e-08, ..., 1.7695e-08, + 6.7987e-08, 1.2759e-07], + [-2.4045e-04, -2.7275e-04, -3.1978e-05, ..., 3.1665e-08, + -1.6689e-04, -2.5487e-04], + [ 5.5600e-07, 1.5274e-07, 4.0792e-07, ..., 8.0094e-08, + 4.2096e-07, 3.9861e-07], + ..., + [ 1.5306e-04, 1.7488e-04, 1.9267e-05, ..., -3.2783e-07, + 1.0568e-04, 1.6415e-04], + [ 3.6675e-06, 3.4068e-06, 1.1148e-06, ..., 1.0245e-07, + 1.8626e-06, 4.6790e-06], + [ 7.8797e-05, 8.9884e-05, 9.9242e-06, ..., 1.7602e-07, + 5.5909e-05, 8.1658e-05]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0087, -0.0220, -0.0053, -0.0166, -0.0073, 0.0009, 0.0098, 0.0221, + 0.0195, -0.0105], device='cuda:0'), grad: tensor([ 3.9302e-07, -6.8665e-04, 1.2266e-06, 9.3319e-07, 9.3132e-06, + 5.9418e-07, 9.1642e-07, 4.3941e-04, 1.1913e-05, 2.2221e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 214.74, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5709 re_mapping 0.0057 re_causal 0.0172 /// teacc 98.98 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.2226, 0.0817, -0.1009, ..., -0.0272, -0.1571, -0.1421], + [ 0.0100, 0.0914, -0.1177, ..., -0.1081, -0.0405, 0.0960], + [ 0.0511, -0.1163, -0.1128, ..., 0.0022, -0.0510, -0.0800], + ..., + [ 0.0839, -0.0239, 0.1232, ..., 0.0614, 0.1629, -0.0016], + [ 0.1030, -0.1681, -0.1232, ..., -0.2049, -0.0535, 0.1713], + [-0.1345, 0.0780, 0.0400, ..., -0.1874, -0.0968, -0.0374]], + device='cuda:0'), grad: tensor([[ 8.9407e-08, -2.7101e-06, 2.4214e-08, ..., 2.7940e-09, + 1.5832e-08, 2.6636e-07], + [-1.0855e-05, -1.9282e-05, 3.1665e-08, ..., 9.3132e-09, + -1.3849e-06, -3.2455e-05], + [ 7.4599e-07, 9.6858e-07, 1.3690e-07, ..., 9.3132e-10, + 2.4308e-07, 2.2333e-06], + ..., + [ 3.0249e-06, 6.8024e-06, -4.5076e-07, ..., -8.1956e-08, + 1.1921e-07, 1.0654e-05], + [ 2.5965e-06, 2.2016e-06, 1.8626e-07, ..., 9.3132e-09, + 5.1782e-07, 7.5325e-06], + [ 2.9616e-06, 7.2569e-06, -9.7696e-07, ..., 4.7497e-08, + 2.2165e-07, 7.6592e-06]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0089, -0.0218, -0.0053, -0.0179, -0.0044, 0.0022, 0.0094, 0.0221, + 0.0188, -0.0122], device='cuda:0'), grad: tensor([-9.5367e-06, -6.8724e-05, 5.7183e-06, 1.5134e-06, 9.3728e-06, + 1.6345e-06, 2.9989e-06, 2.2694e-05, 1.4454e-05, 1.9848e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 214.79, cls_loss 0.0015 cls_loss_mapping 0.0033 cls_loss_causal 0.5010 re_mapping 0.0059 re_causal 0.0166 /// teacc 98.99 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.2230, 0.0818, -0.1019, ..., -0.0266, -0.1574, -0.1424], + [ 0.0100, 0.0920, -0.1176, ..., -0.1083, -0.0405, 0.0961], + [ 0.0511, -0.1163, -0.1131, ..., 0.0023, -0.0509, -0.0801], + ..., + [ 0.0840, -0.0250, 0.1232, ..., 0.0616, 0.1630, -0.0017], + [ 0.1029, -0.1703, -0.1234, ..., -0.2053, -0.0536, 0.1713], + [-0.1352, 0.0780, 0.0405, ..., -0.1893, -0.0972, -0.0377]], + device='cuda:0'), grad: tensor([[ 8.2888e-08, -2.1048e-07, 1.9930e-07, ..., 2.5146e-08, + 7.5437e-08, 2.5146e-08], + [ 2.3818e-04, 2.5611e-07, 6.0815e-07, ..., 2.8944e-04, + 7.7772e-04, -3.7253e-09], + [-2.3901e-04, 9.3132e-08, 8.3074e-07, ..., -2.9159e-04, + -7.8106e-04, 4.9360e-08], + ..., + [-2.9672e-06, 3.3993e-07, -5.3309e-06, ..., 8.2888e-08, + -1.4454e-06, 1.1176e-08], + [ 9.6951e-07, 3.4347e-06, 4.0084e-06, ..., 2.5891e-07, + 6.6590e-07, 6.7241e-07], + [ 1.4231e-06, -2.1145e-05, -1.4864e-05, ..., 9.4995e-07, + 2.9039e-06, -2.1327e-06]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0087, -0.0217, -0.0053, -0.0179, -0.0046, 0.0028, 0.0096, 0.0220, + 0.0179, -0.0122], device='cuda:0'), grad: tensor([ 3.1013e-07, 2.1572e-03, -2.1706e-03, 7.9256e-07, 4.2081e-05, + 2.9430e-06, 3.1851e-07, 1.9968e-06, 9.8050e-06, -4.7296e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.77, cls_loss 0.0015 cls_loss_mapping 0.0032 cls_loss_causal 0.5142 re_mapping 0.0055 re_causal 0.0168 /// teacc 98.95 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.2233, 0.0819, -0.1030, ..., -0.0265, -0.1581, -0.1425], + [ 0.0099, 0.0921, -0.1177, ..., -0.1095, -0.0409, 0.0962], + [ 0.0513, -0.1163, -0.1131, ..., 0.0031, -0.0503, -0.0801], + ..., + [ 0.0840, -0.0253, 0.1235, ..., 0.0616, 0.1630, -0.0018], + [ 0.1031, -0.1708, -0.1235, ..., -0.2057, -0.0535, 0.1717], + [-0.1360, 0.0788, 0.0410, ..., -0.1909, -0.0975, -0.0368]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, -6.4708e-06, 2.0582e-07, ..., -8.5682e-08, + 1.2107e-08, 8.8476e-08], + [ 1.0338e-07, 6.9011e-07, 4.5355e-07, ..., 4.7497e-08, + 5.2154e-08, 2.7381e-07], + [-9.6671e-07, 9.2387e-07, 1.4435e-07, ..., -1.0766e-06, + -9.2480e-07, 1.4249e-07], + ..., + [ 9.2853e-07, 1.0924e-06, 1.4231e-06, ..., 9.7323e-07, + 8.1025e-07, 4.9174e-07], + [-5.4017e-07, 2.6431e-06, 3.4850e-06, ..., -3.5390e-08, + 4.0978e-08, 1.3225e-07], + [-3.1069e-06, -9.4116e-05, -1.1408e-04, ..., 7.0781e-08, + -1.0710e-07, -4.8190e-05]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0087, -0.0219, -0.0050, -0.0179, -0.0060, 0.0034, 0.0091, 0.0220, + 0.0177, -0.0112], device='cuda:0'), grad: tensor([-1.1533e-05, 2.0973e-06, 9.5740e-06, 1.4370e-06, 3.0971e-04, + 1.9297e-06, 8.2105e-06, 6.3777e-06, 8.2031e-06, -3.3593e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 214.61, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5101 re_mapping 0.0054 re_causal 0.0163 /// teacc 98.96 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.2237, 0.0820, -0.1037, ..., -0.0266, -0.1585, -0.1427], + [ 0.0098, 0.0927, -0.1177, ..., -0.1098, -0.0410, 0.0964], + [ 0.0512, -0.1164, -0.1137, ..., 0.0029, -0.0504, -0.0802], + ..., + [ 0.0842, -0.0266, 0.1237, ..., 0.0619, 0.1634, -0.0019], + [ 0.1032, -0.1712, -0.1237, ..., -0.2059, -0.0536, 0.1718], + [-0.1364, 0.0794, 0.0426, ..., -0.1915, -0.0977, -0.0354]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.2107e-08, 8.0094e-08, ..., 4.7497e-08, + 1.8626e-09, 1.8626e-09], + [ 3.4459e-08, -4.6566e-08, 9.9652e-08, ..., 1.0151e-07, + 3.6322e-08, -1.0431e-07], + [-3.7011e-06, 1.8626e-08, -1.0245e-07, ..., -3.8370e-06, + -3.5334e-06, 1.4901e-08], + ..., + [ 3.6526e-06, 5.6811e-08, 5.4203e-07, ..., 4.2170e-06, + 3.4589e-06, 8.7544e-08], + [-4.5635e-08, 1.8533e-07, 2.5984e-07, ..., 6.7055e-08, + -2.7940e-09, -5.5879e-08], + [ 7.4506e-09, 5.9605e-08, 5.5879e-09, ..., 2.6636e-07, + 6.5193e-09, 1.0245e-08]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0087, -0.0219, -0.0050, -0.0180, -0.0070, 0.0032, 0.0094, 0.0220, + 0.0177, -0.0104], device='cuda:0'), grad: tensor([ 5.2154e-08, 4.1910e-08, -7.8008e-06, -2.2631e-07, 2.7940e-07, + -2.1495e-06, 4.2841e-07, 8.7321e-06, 6.7987e-07, -6.7987e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 214.84, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.5069 re_mapping 0.0054 re_causal 0.0168 /// teacc 98.93 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.2241, 0.0820, -0.1051, ..., -0.0268, -0.1590, -0.1431], + [ 0.0099, 0.0930, -0.1171, ..., -0.1095, -0.0408, 0.0965], + [ 0.0510, -0.1165, -0.1142, ..., 0.0029, -0.0506, -0.0803], + ..., + [ 0.0842, -0.0269, 0.1234, ..., 0.0619, 0.1634, -0.0020], + [ 0.1033, -0.1721, -0.1238, ..., -0.2061, -0.0536, 0.1721], + [-0.1373, 0.0793, 0.0427, ..., -0.1929, -0.0982, -0.0357]], + device='cuda:0'), grad: tensor([[ 1.0803e-07, -1.8161e-07, -1.3970e-08, ..., -1.7695e-08, + 6.7987e-08, 2.0489e-08], + [ 4.2655e-07, 1.1837e-06, 2.0489e-08, ..., 2.8871e-08, + 2.2352e-08, 1.6699e-06], + [-5.2806e-07, 2.6636e-07, -6.1467e-08, ..., -5.5134e-07, + -4.7870e-07, 1.0245e-08], + ..., + [ 5.1036e-07, 7.0501e-07, -7.4506e-09, ..., 3.9395e-07, + 3.0547e-07, 8.4657e-07], + [ 4.6473e-07, 3.0976e-06, 4.0978e-08, ..., 9.0338e-08, + 3.5390e-08, 1.6410e-06], + [ 4.8988e-07, -7.4916e-06, -1.3830e-06, ..., 5.6811e-08, + 1.3970e-08, 2.0284e-06]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0088, -0.0216, -0.0051, -0.0164, -0.0070, 0.0019, 0.0095, 0.0218, + 0.0175, -0.0104], device='cuda:0'), grad: tensor([ 1.1623e-06, 3.7849e-06, -8.7544e-07, 1.9800e-06, 2.1383e-05, + -1.4722e-05, 2.9244e-06, 3.3900e-06, 1.0036e-05, -2.9162e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 214.94, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5029 re_mapping 0.0056 re_causal 0.0157 /// teacc 98.94 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.2250, 0.0821, -0.1060, ..., -0.0277, -0.1596, -0.1435], + [ 0.0100, 0.0937, -0.1172, ..., -0.1097, -0.0409, 0.0968], + [ 0.0511, -0.1166, -0.1147, ..., 0.0030, -0.0506, -0.0803], + ..., + [ 0.0843, -0.0283, 0.1237, ..., 0.0621, 0.1636, -0.0023], + [ 0.1031, -0.1733, -0.1240, ..., -0.2068, -0.0537, 0.1723], + [-0.1384, 0.0792, 0.0427, ..., -0.1942, -0.0986, -0.0362]], + device='cuda:0'), grad: tensor([[ 1.9558e-08, 1.6112e-07, 1.8813e-07, ..., 3.9116e-08, + 3.5390e-08, 2.2352e-08], + [ 3.4459e-08, 1.5367e-07, 1.4529e-07, ..., 2.7940e-08, + 2.0489e-08, 3.6322e-08], + [-8.8476e-08, 4.0978e-08, -3.4459e-08, ..., -2.5891e-07, + -2.5053e-07, 2.1420e-08], + ..., + [-2.0489e-08, 3.1851e-07, 1.7975e-07, ..., 6.0536e-08, + 5.0291e-08, 7.6368e-08], + [-1.9576e-06, -8.7451e-07, 3.4925e-07, ..., 4.0978e-08, + 2.8871e-08, -6.1095e-06], + [-4.6194e-07, -1.8001e-05, -1.5959e-05, ..., 1.3970e-08, + 5.5879e-09, -3.4925e-06]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0089, -0.0215, -0.0051, -0.0162, -0.0069, 0.0019, 0.0096, 0.0218, + 0.0170, -0.0106], device='cuda:0'), grad: tensor([ 6.2771e-07, 5.1036e-07, -7.5344e-07, 1.3068e-05, 4.2230e-05, + -2.6226e-05, 2.7090e-05, 1.0040e-06, -1.3314e-05, -4.4316e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 214.98, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.5078 re_mapping 0.0055 re_causal 0.0166 /// teacc 98.99 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.2257, 0.0822, -0.1034, ..., -0.0270, -0.1599, -0.1445], + [ 0.0101, 0.0943, -0.1169, ..., -0.1097, -0.0408, 0.0971], + [ 0.0512, -0.1168, -0.1153, ..., 0.0028, -0.0506, -0.0804], + ..., + [ 0.0842, -0.0295, 0.1234, ..., 0.0619, 0.1637, -0.0025], + [ 0.1031, -0.1731, -0.1242, ..., -0.2072, -0.0537, 0.1725], + [-0.1392, 0.0789, 0.0423, ..., -0.1959, -0.0989, -0.0368]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -4.4517e-07, 2.7940e-09, ..., 1.2107e-08, + 1.3970e-09, -0.0000e+00], + [ 7.9162e-09, -6.1933e-08, 5.5879e-08, ..., 3.0268e-08, + 3.4459e-08, -2.5099e-07], + [ 4.0513e-08, 7.7300e-08, 6.9849e-08, ..., -9.8255e-08, + 5.1223e-08, 6.5658e-08], + ..., + [-1.7090e-07, 8.1491e-08, -2.0675e-07, ..., -4.4238e-08, + -1.6904e-07, 7.3109e-08], + [ 2.7008e-08, 6.0536e-08, 1.9558e-08, ..., 4.7497e-08, + 1.1176e-08, 5.2154e-08], + [ 5.9139e-08, 2.0443e-07, 4.4703e-08, ..., 4.0978e-08, + 5.1223e-08, 3.0268e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0086, -0.0213, -0.0051, -0.0165, -0.0065, 0.0022, 0.0095, 0.0216, + 0.0170, -0.0111], device='cuda:0'), grad: tensor([-6.8173e-07, -2.3050e-07, -3.2969e-07, -6.5193e-09, 5.8673e-08, + 3.3295e-07, -2.3982e-07, 1.9092e-08, 4.8988e-07, 5.9651e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 214.83, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4981 re_mapping 0.0054 re_causal 0.0160 /// teacc 98.94 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.2256, 0.0829, -0.1025, ..., -0.0266, -0.1601, -0.1438], + [ 0.0105, 0.0943, -0.1168, ..., -0.1098, -0.0405, 0.0972], + [ 0.0506, -0.1168, -0.1165, ..., 0.0030, -0.0512, -0.0804], + ..., + [ 0.0845, -0.0298, 0.1240, ..., 0.0618, 0.1639, -0.0026], + [ 0.1030, -0.1737, -0.1245, ..., -0.2077, -0.0538, 0.1724], + [-0.1419, 0.0782, 0.0412, ..., -0.1967, -0.1003, -0.0372]], + device='cuda:0'), grad: tensor([[ 1.5832e-07, 2.4214e-07, 6.5193e-09, ..., 1.8626e-09, + 0.0000e+00, 3.1572e-07], + [-1.8636e-06, -3.4831e-06, 5.1223e-08, ..., 2.3283e-08, + 2.4214e-08, -4.4629e-06], + [ 1.0990e-07, 1.4156e-07, 2.6077e-08, ..., 1.6764e-08, + 2.7940e-09, 2.1234e-07], + ..., + [ 1.3420e-06, 2.2724e-06, -3.3528e-08, ..., -1.2107e-08, + -3.6322e-08, 3.0566e-06], + [-3.3248e-07, 2.5239e-07, 1.3970e-08, ..., 5.5879e-09, + 0.0000e+00, -1.7509e-07], + [ 3.4925e-07, 3.5111e-07, -3.2596e-08, ..., 2.7940e-09, + 2.7940e-09, 5.9139e-07]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0073, -0.0209, -0.0057, -0.0165, -0.0062, 0.0020, 0.0092, 0.0217, + 0.0167, -0.0119], device='cuda:0'), grad: tensor([ 7.0035e-07, -9.1270e-06, 4.9639e-07, -3.3434e-07, 1.1269e-07, + 3.3900e-07, 5.1595e-07, 6.3553e-06, -4.2375e-07, 1.3439e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 214.63, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.5261 re_mapping 0.0058 re_causal 0.0175 /// teacc 98.91 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.2263, 0.0829, -0.1034, ..., -0.0270, -0.1608, -0.1444], + [ 0.0105, 0.0945, -0.1169, ..., -0.1104, -0.0406, 0.0973], + [ 0.0508, -0.1169, -0.1168, ..., 0.0032, -0.0511, -0.0803], + ..., + [ 0.0834, -0.0298, 0.1237, ..., 0.0616, 0.1634, -0.0033], + [ 0.1053, -0.1757, -0.1227, ..., -0.2065, -0.0516, 0.1746], + [-0.1421, 0.0783, 0.0413, ..., -0.1971, -0.1005, -0.0371]], + device='cuda:0'), grad: tensor([[ 7.9162e-08, 3.5375e-05, 6.8009e-05, ..., 1.0245e-08, + 7.4506e-09, 1.0896e-07], + [-1.9574e-04, -1.8189e-06, -9.6142e-05, ..., 2.0489e-08, + -2.0459e-05, -2.2566e-04], + [ 4.7963e-07, 7.4133e-07, 1.6689e-06, ..., 2.3749e-07, + 5.1223e-08, 6.7614e-07], + ..., + [ 1.9217e-04, 1.4678e-06, 9.5606e-05, ..., -5.5879e-09, + 2.0102e-05, 2.2066e-04], + [ 1.2731e-06, 5.6066e-07, 1.3430e-06, ..., 1.6764e-08, + 1.2945e-07, 1.5879e-06], + [ 5.6438e-07, -3.8326e-05, -7.3671e-05, ..., 1.1176e-08, + 5.7742e-08, 7.7952e-07]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0074, -0.0209, -0.0056, -0.0165, -0.0062, 0.0021, 0.0092, 0.0213, + 0.0178, -0.0118], device='cuda:0'), grad: tensor([ 1.5020e-04, -6.4516e-04, 5.1931e-06, 2.6822e-06, 3.9451e-06, + 1.1371e-06, 2.5909e-06, 6.3324e-04, 6.3814e-06, -1.6069e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 214.41, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.4781 re_mapping 0.0059 re_causal 0.0159 /// teacc 98.93 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.2269, 0.0828, -0.1055, ..., -0.0270, -0.1614, -0.1449], + [ 0.0106, 0.0946, -0.1169, ..., -0.1105, -0.0406, 0.0975], + [ 0.0508, -0.1170, -0.1172, ..., 0.0031, -0.0511, -0.0804], + ..., + [ 0.0835, -0.0296, 0.1242, ..., 0.0616, 0.1637, -0.0034], + [ 0.1053, -0.1765, -0.1229, ..., -0.2068, -0.0517, 0.1749], + [-0.1438, 0.0782, 0.0409, ..., -0.1979, -0.1020, -0.0381]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 1.3234e-06, 1.5395e-06, ..., -9.3132e-10, + 0.0000e+00, 7.0781e-08], + [-6.3330e-07, -1.0571e-06, 1.6112e-07, ..., 1.8626e-09, + 0.0000e+00, -1.5106e-06], + [ 5.6811e-08, 1.6484e-07, 6.9849e-08, ..., 8.3819e-09, + 0.0000e+00, 2.0396e-07], + ..., + [ 8.1956e-08, 4.9639e-07, 2.1979e-07, ..., 0.0000e+00, + 0.0000e+00, 4.6100e-07], + [-1.0431e-07, 2.9523e-07, 2.7288e-07, ..., 3.7253e-09, + 0.0000e+00, -1.6205e-07], + [ 5.5879e-09, -3.6415e-06, -4.6305e-06, ..., 0.0000e+00, + 0.0000e+00, 7.7300e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0077, -0.0208, -0.0056, -0.0167, -0.0061, 0.0025, 0.0097, 0.0213, + 0.0175, -0.0121], device='cuda:0'), grad: tensor([ 3.4198e-06, -4.0755e-06, 6.1467e-07, 1.8971e-06, 4.5486e-06, + -2.2296e-06, 3.2093e-06, 1.5181e-06, 8.6240e-07, -9.8124e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 214.52, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4863 re_mapping 0.0057 re_causal 0.0161 /// teacc 98.97 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.2279, 0.0828, -0.1063, ..., -0.0272, -0.1626, -0.1461], + [ 0.0105, 0.0946, -0.1170, ..., -0.1106, -0.0407, 0.0974], + [ 0.0509, -0.1170, -0.1175, ..., 0.0032, -0.0510, -0.0804], + ..., + [ 0.0837, -0.0296, 0.1246, ..., 0.0618, 0.1640, -0.0034], + [ 0.1055, -0.1776, -0.1232, ..., -0.2071, -0.0518, 0.1754], + [-0.1449, 0.0782, 0.0406, ..., -0.1982, -0.1030, -0.0384]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 2.8219e-07, 2.0210e-07, ..., 2.0489e-08, + 2.8871e-08, 2.7008e-08], + [ 1.9059e-05, 2.7362e-06, 2.9817e-05, ..., 3.2596e-08, + 1.9953e-05, 1.4193e-05], + [ 9.9465e-07, 8.4843e-07, 1.6754e-06, ..., 9.8720e-08, + 1.0198e-06, 7.4971e-07], + ..., + [-2.4766e-05, 1.8161e-07, -3.8534e-05, ..., 8.3819e-09, + -2.5928e-05, -1.8403e-05], + [ 9.6858e-08, 1.8515e-06, 4.0699e-07, ..., 4.2841e-08, + 1.2107e-07, 6.7055e-08], + [ 4.3474e-06, -4.1313e-06, 6.2101e-06, ..., 3.4459e-08, + 4.5411e-06, 3.4589e-06]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0078, -0.0208, -0.0056, -0.0172, -0.0060, 0.0025, 0.0105, 0.0213, + 0.0171, -0.0121], device='cuda:0'), grad: tensor([ 2.7865e-06, 8.7619e-05, 7.2792e-06, -2.1793e-06, 3.2056e-06, + 5.0664e-06, -5.4240e-05, -5.2691e-05, 1.1973e-05, -8.7321e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.82, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.5377 re_mapping 0.0057 re_causal 0.0167 /// teacc 99.01 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.2284, 0.0829, -0.1069, ..., -0.0272, -0.1633, -0.1463], + [ 0.0105, 0.0946, -0.1170, ..., -0.1107, -0.0407, 0.0975], + [ 0.0509, -0.1172, -0.1179, ..., 0.0032, -0.0511, -0.0805], + ..., + [ 0.0837, -0.0299, 0.1247, ..., 0.0618, 0.1642, -0.0035], + [ 0.1056, -0.1779, -0.1233, ..., -0.2073, -0.0518, 0.1756], + [-0.1452, 0.0785, 0.0411, ..., -0.1984, -0.1032, -0.0384]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, -2.9698e-05, 3.5390e-08, ..., 3.8184e-08, + 4.6566e-08, -1.5534e-06], + [-1.6764e-08, 2.0787e-05, 7.1712e-08, ..., 4.0978e-08, + 6.6124e-08, 1.0040e-06], + [-2.3209e-06, 5.9120e-06, -5.7556e-07, ..., -1.4435e-06, + -1.9912e-06, 6.8825e-07], + ..., + [ 9.0618e-07, 3.1199e-06, -6.3423e-07, ..., 1.2387e-06, + 9.2853e-07, 1.9483e-06], + [ 1.0049e-06, 9.9652e-07, 1.0859e-06, ..., 1.6019e-07, + 7.1898e-07, 7.5903e-07], + [ 2.4773e-07, 3.5882e-05, 9.8720e-08, ..., 2.3283e-08, + 1.7509e-07, 2.3827e-05]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0078, -0.0208, -0.0056, -0.0175, -0.0062, 0.0028, 0.0102, 0.0213, + 0.0170, -0.0118], device='cuda:0'), grad: tensor([-6.7234e-05, 4.9978e-05, 9.8273e-06, 2.3842e-06, -2.7800e-04, + 7.9628e-07, 3.4459e-06, 2.4274e-05, 7.6033e-06, 2.4700e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.66, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4843 re_mapping 0.0056 re_causal 0.0162 /// teacc 99.01 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.2291, 0.0831, -0.1075, ..., -0.0274, -0.1639, -0.1470], + [ 0.0109, 0.0952, -0.1166, ..., -0.1108, -0.0407, 0.0980], + [ 0.0510, -0.1172, -0.1182, ..., 0.0033, -0.0510, -0.0806], + ..., + [ 0.0833, -0.0317, 0.1244, ..., 0.0617, 0.1641, -0.0039], + [ 0.1056, -0.1787, -0.1236, ..., -0.2077, -0.0518, 0.1758], + [-0.1449, 0.0788, 0.0417, ..., -0.1990, -0.1033, -0.0388]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -1.0142e-06, -2.7847e-07, ..., -3.9767e-07, + 7.4506e-09, -1.3206e-06], + [ 4.8429e-08, 5.8860e-07, 1.5553e-07, ..., 1.7509e-07, + 8.7544e-08, 4.0513e-07], + [ 3.0734e-08, 1.9465e-07, 1.3504e-07, ..., 5.1223e-08, + 3.7253e-08, 1.3225e-07], + ..., + [-1.4529e-07, 4.3679e-07, -1.6671e-07, ..., 4.6566e-09, + -2.1607e-07, 3.2969e-07], + [ 1.0245e-08, 3.6135e-07, 9.4064e-08, ..., 7.3574e-08, + 1.3039e-08, 2.2445e-07], + [ 3.2596e-08, 1.6298e-07, -5.9605e-08, ..., 5.4017e-08, + 4.4703e-08, 1.4994e-07]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0076, -0.0206, -0.0055, -0.0175, -0.0061, 0.0029, 0.0099, 0.0208, + 0.0169, -0.0116], device='cuda:0'), grad: tensor([ 4.2245e-06, 1.9632e-06, 6.8545e-07, 1.1083e-07, 2.1420e-07, + 3.6228e-07, -1.0267e-05, 8.6054e-07, 1.3253e-06, 5.0385e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 214.93, cls_loss 0.0019 cls_loss_mapping 0.0025 cls_loss_causal 0.5347 re_mapping 0.0055 re_causal 0.0163 /// teacc 99.00 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.2295, 0.0831, -0.1084, ..., -0.0275, -0.1647, -0.1473], + [ 0.0110, 0.0962, -0.1163, ..., -0.1095, -0.0407, 0.0989], + [ 0.0509, -0.1173, -0.1193, ..., 0.0026, -0.0511, -0.0810], + ..., + [ 0.0834, -0.0326, 0.1243, ..., 0.0617, 0.1643, -0.0043], + [ 0.1057, -0.1791, -0.1239, ..., -0.2081, -0.0519, 0.1762], + [-0.1451, 0.0788, 0.0423, ..., -0.1997, -0.1035, -0.0396]], + device='cuda:0'), grad: tensor([[ 2.1420e-07, -9.3132e-10, 8.0187e-07, ..., 5.2340e-07, + 2.8964e-07, 2.5053e-07], + [ 1.2904e-05, 1.0207e-05, 1.7866e-05, ..., 7.9572e-06, + 1.7539e-05, 1.1511e-05], + [ 3.7625e-07, 7.0129e-07, 2.6207e-06, ..., 1.8310e-06, + 6.1188e-07, 6.7987e-07], + ..., + [-1.5751e-05, -1.2271e-05, -2.1592e-05, ..., -9.4175e-06, + -2.1279e-05, -1.3970e-05], + [ 2.8033e-07, 2.2631e-07, 5.0105e-07, ..., 2.9802e-07, + 2.9616e-07, 9.3132e-08], + [ 5.6904e-07, 4.5169e-07, 8.5309e-07, ..., 3.9116e-07, + 7.0222e-07, 5.0943e-07]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0075, -0.0200, -0.0057, -0.0190, -0.0064, 0.0025, 0.0106, 0.0205, + 0.0167, -0.0111], device='cuda:0'), grad: tensor([ 1.1884e-06, 3.9846e-05, 6.1914e-06, -7.5586e-06, 2.0936e-06, + 1.9018e-06, 3.8370e-07, -4.7982e-05, 1.7546e-06, 2.1495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 214.70, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.5087 re_mapping 0.0055 re_causal 0.0163 /// teacc 98.97 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.2301, 0.0832, -0.1088, ..., -0.0268, -0.1656, -0.1467], + [ 0.0109, 0.0962, -0.1163, ..., -0.1098, -0.0408, 0.0990], + [ 0.0510, -0.1174, -0.1196, ..., 0.0027, -0.0511, -0.0811], + ..., + [ 0.0835, -0.0328, 0.1242, ..., 0.0614, 0.1644, -0.0044], + [ 0.1057, -0.1796, -0.1240, ..., -0.2095, -0.0519, 0.1766], + [-0.1455, 0.0786, 0.0421, ..., -0.2007, -0.1037, -0.0395]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -3.2131e-07, 2.6077e-08, ..., 2.8871e-08, + 2.7940e-09, 7.2643e-08], + [ 4.8429e-08, 4.5635e-08, 1.2759e-07, ..., 7.5437e-08, + 2.0489e-08, 7.0781e-08], + [ 1.3970e-07, 3.5390e-08, 2.8033e-07, ..., 1.9837e-07, + 5.6811e-08, -4.2003e-07], + ..., + [-5.2247e-07, 1.4622e-07, -2.0768e-07, ..., -1.5274e-07, + -2.0582e-07, 1.3877e-07], + [ 9.4064e-08, 1.0617e-07, 4.2375e-07, ..., 2.3842e-07, + 3.5390e-08, -2.8126e-07], + [ 5.2154e-08, 6.3982e-07, 2.3376e-07, ..., 4.3772e-08, + 1.8626e-08, 6.5472e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0075, -0.0200, -0.0058, -0.0191, -0.0059, 0.0033, 0.0105, 0.0204, + 0.0168, -0.0116], device='cuda:0'), grad: tensor([-2.4121e-07, 6.3609e-07, -1.1958e-06, 3.4086e-06, -2.1383e-06, + -4.5635e-06, 9.1828e-07, 6.3330e-08, 6.3144e-07, 2.4475e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 214.88, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.5114 re_mapping 0.0055 re_causal 0.0165 /// teacc 99.00 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.2308, 0.0829, -0.1093, ..., -0.0270, -0.1661, -0.1471], + [ 0.0110, 0.0963, -0.1161, ..., -0.1096, -0.0407, 0.0992], + [ 0.0509, -0.1175, -0.1202, ..., 0.0025, -0.0511, -0.0813], + ..., + [ 0.0835, -0.0331, 0.1240, ..., 0.0615, 0.1645, -0.0045], + [ 0.1060, -0.1792, -0.1241, ..., -0.2097, -0.0519, 0.1778], + [-0.1458, 0.0788, 0.0423, ..., -0.2012, -0.1038, -0.0397]], + device='cuda:0'), grad: tensor([[ 2.2352e-07, 1.2107e-08, 3.1665e-08, ..., 4.7497e-08, + 6.5193e-09, 1.3039e-08], + [ 1.0906e-06, 3.4459e-08, 1.2200e-07, ..., 1.2768e-06, + 1.3746e-06, -1.1176e-08], + [-1.6950e-07, 1.5832e-08, 3.4645e-07, ..., -3.1330e-06, + -1.4137e-06, 1.6298e-07], + ..., + [-2.6330e-05, 2.0582e-07, -2.5686e-06, ..., 1.2480e-07, + -4.6566e-09, 9.2201e-08], + [ 2.0061e-06, 2.7008e-08, 2.4214e-08, ..., 1.5851e-06, + 9.3132e-09, -3.9395e-07], + [ 2.4028e-07, -1.0123e-06, -1.5413e-06, ..., 5.5879e-09, + 6.5193e-09, 1.3039e-07]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0079, -0.0198, -0.0059, -0.0182, -0.0058, 0.0023, 0.0094, 0.0202, + 0.0178, -0.0115], device='cuda:0'), grad: tensor([ 8.2236e-07, 5.4054e-06, 4.9733e-07, 1.7053e-06, 8.0764e-05, + 4.6678e-06, 3.8464e-07, -9.8884e-05, 5.8450e-06, -1.1576e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.98, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.4805 re_mapping 0.0056 re_causal 0.0157 /// teacc 98.89 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.2320, 0.0828, -0.1105, ..., -0.0274, -0.1672, -0.1476], + [ 0.0112, 0.0970, -0.1160, ..., -0.1097, -0.0408, 0.0995], + [ 0.0508, -0.1176, -0.1210, ..., 0.0024, -0.0512, -0.0814], + ..., + [ 0.0835, -0.0344, 0.1240, ..., 0.0617, 0.1646, -0.0048], + [ 0.1060, -0.1798, -0.1243, ..., -0.2102, -0.0520, 0.1779], + [-0.1459, 0.0792, 0.0429, ..., -0.2019, -0.1038, -0.0397]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, -1.2871e-06, 1.3970e-08, ..., -6.1840e-07, + 6.5193e-09, 1.8626e-09], + [ 7.6368e-08, -6.5193e-09, 1.0524e-07, ..., 5.3085e-08, + 6.4261e-08, -4.7497e-08], + [-1.3039e-08, 9.3412e-07, 6.7055e-08, ..., 4.8522e-07, + 0.0000e+00, 1.9558e-08], + ..., + [-1.5087e-07, 2.7940e-08, -1.8906e-07, ..., -4.4703e-08, + -1.2852e-07, 1.6764e-08], + [-2.5146e-08, 7.7300e-08, 2.8871e-08, ..., 4.5635e-08, + 1.3039e-08, -2.5146e-08], + [ 2.7940e-08, 2.6170e-07, 3.0734e-08, ..., 1.2666e-07, + 2.1420e-08, 5.0291e-08]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0081, -0.0196, -0.0059, -0.0181, -0.0061, 0.0020, 0.0098, 0.0200, + 0.0176, -0.0111], device='cuda:0'), grad: tensor([-5.3793e-06, 1.6578e-07, 3.7998e-06, -2.6077e-07, 5.7742e-08, + -8.5589e-07, 1.2703e-06, -1.7975e-07, 3.0175e-07, 1.0766e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 214.85, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.5140 re_mapping 0.0056 re_causal 0.0163 /// teacc 98.93 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.2328, 0.0829, -0.1110, ..., -0.0276, -0.1682, -0.1482], + [ 0.0104, 0.0979, -0.1167, ..., -0.1096, -0.0418, 0.0997], + [ 0.0506, -0.1177, -0.1219, ..., 0.0022, -0.0512, -0.0817], + ..., + [ 0.0845, -0.0358, 0.1249, ..., 0.0619, 0.1659, -0.0046], + [ 0.1060, -0.1804, -0.1245, ..., -0.2107, -0.0521, 0.1781], + [-0.1466, 0.0794, 0.0431, ..., -0.2027, -0.1041, -0.0400]], + device='cuda:0'), grad: tensor([[ 6.0350e-07, -1.9558e-08, 1.2107e-08, ..., 1.3448e-06, + 9.6858e-08, 4.6566e-09], + [ 1.8626e-07, -9.2201e-08, 1.1828e-07, ..., 1.1176e-07, + 2.0675e-07, -2.2911e-07], + [-9.0990e-07, 4.0047e-08, 2.7567e-07, ..., -2.0899e-06, + -2.1048e-07, 3.2596e-08], + ..., + [-3.3248e-07, -4.6566e-09, -6.1560e-07, ..., 1.1176e-07, + -3.5390e-07, 4.0978e-08], + [ 6.2399e-08, 1.3877e-07, 1.9185e-07, ..., 5.6811e-08, + 2.9802e-08, 1.4901e-08], + [ 7.4506e-08, -2.8126e-07, -3.4273e-07, ..., 1.6764e-08, + 6.7055e-08, 2.2352e-08]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0081, -0.0200, -0.0061, -0.0181, -0.0064, 0.0017, 0.0099, 0.0208, + 0.0174, -0.0110], device='cuda:0'), grad: tensor([ 5.3160e-06, 2.4680e-07, -7.9423e-06, 1.6233e-06, 6.6683e-07, + -1.1921e-07, 4.2282e-07, -2.8126e-07, 6.7241e-07, -6.4541e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 214.96, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.5177 re_mapping 0.0057 re_causal 0.0163 /// teacc 98.93 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.2341, 0.0829, -0.1117, ..., -0.0279, -0.1692, -0.1494], + [ 0.0104, 0.0980, -0.1168, ..., -0.1095, -0.0418, 0.0998], + [ 0.0505, -0.1178, -0.1224, ..., 0.0021, -0.0514, -0.0821], + ..., + [ 0.0846, -0.0361, 0.1250, ..., 0.0620, 0.1660, -0.0046], + [ 0.1060, -0.1810, -0.1247, ..., -0.2112, -0.0521, 0.1781], + [-0.1467, 0.0795, 0.0435, ..., -0.2034, -0.1041, -0.0402]], + device='cuda:0'), grad: tensor([[ 2.8498e-07, 3.2596e-08, 8.3819e-08, ..., 3.0734e-08, + 1.8626e-08, 4.2282e-07], + [ 7.3574e-08, 7.1712e-08, 9.7789e-08, ..., 5.9605e-08, + 5.5879e-08, 4.5635e-08], + [ 3.1479e-07, 3.0734e-08, 3.2410e-07, ..., 1.3411e-07, + 2.2352e-07, 1.7975e-07], + ..., + [-3.0454e-07, 2.1234e-07, 8.4564e-07, ..., -7.8231e-08, + -3.7253e-07, -1.0245e-08], + [-9.4250e-07, 2.5053e-07, 2.9057e-07, ..., 4.9360e-08, + 2.7940e-08, -1.4370e-06], + [ 2.0862e-07, -2.6356e-07, -1.8608e-06, ..., 2.1420e-08, + 7.4506e-09, 3.1665e-07]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0082, -0.0199, -0.0062, -0.0177, -0.0066, 0.0014, 0.0102, 0.0208, + 0.0171, -0.0109], device='cuda:0'), grad: tensor([ 1.2070e-06, 7.0501e-07, 8.2701e-07, 2.1353e-05, 6.2399e-07, + -2.5377e-05, 1.6773e-06, 1.4324e-06, -2.7474e-07, -2.1569e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 215.05, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.5149 re_mapping 0.0052 re_causal 0.0161 /// teacc 98.88 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.2347, 0.0829, -0.1120, ..., -0.0283, -0.1700, -0.1495], + [ 0.0105, 0.0983, -0.1167, ..., -0.1095, -0.0418, 0.1001], + [ 0.0507, -0.1179, -0.1227, ..., 0.0023, -0.0513, -0.0822], + ..., + [ 0.0845, -0.0366, 0.1251, ..., 0.0621, 0.1661, -0.0048], + [ 0.1061, -0.1816, -0.1249, ..., -0.2114, -0.0522, 0.1782], + [-0.1470, 0.0789, 0.0433, ..., -0.2040, -0.1043, -0.0402]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, -1.4957e-06, 1.4715e-07, ..., 5.4948e-08, + 6.4261e-08, 4.6566e-09], + [ 7.8976e-06, -2.1607e-07, 1.6280e-06, ..., 4.1053e-06, + 4.2170e-06, -2.8126e-07], + [-7.6517e-06, 2.3283e-08, 1.3374e-06, ..., -1.8487e-06, + -3.6377e-06, 1.2107e-08], + ..., + [-8.4400e-05, 1.4808e-07, -2.1040e-04, ..., -1.8847e-04, + -7.5102e-05, 1.6671e-07], + [ 1.8161e-07, 2.3283e-08, 2.9895e-07, ..., 2.7195e-07, + 1.3318e-07, 1.6764e-08], + [ 3.6322e-08, 1.4901e-06, 3.2596e-08, ..., 1.2480e-07, + 5.0291e-08, 4.4703e-08]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0084, -0.0198, -0.0062, -0.0180, -0.0060, 0.0015, 0.0105, 0.0207, + 0.0168, -0.0113], device='cuda:0'), grad: tensor([-2.4214e-06, 2.6241e-05, -2.5764e-05, 2.9635e-04, 5.7742e-08, + 4.6473e-07, 2.2165e-07, -2.9826e-04, 7.2177e-07, 2.9225e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 214.89, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.4928 re_mapping 0.0053 re_causal 0.0154 /// teacc 98.92 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.2357, 0.0831, -0.1124, ..., -0.0288, -0.1705, -0.1500], + [ 0.0105, 0.0982, -0.1167, ..., -0.1098, -0.0418, 0.1003], + [ 0.0507, -0.1182, -0.1230, ..., 0.0024, -0.0513, -0.0825], + ..., + [ 0.0846, -0.0373, 0.1252, ..., 0.0624, 0.1662, -0.0050], + [ 0.1062, -0.1821, -0.1249, ..., -0.2118, -0.0522, 0.1785], + [-0.1474, 0.0790, 0.0437, ..., -0.2055, -0.1044, -0.0403]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 3.2689e-07, 5.2247e-07, ..., 3.7253e-09, + 9.3132e-10, 4.5635e-08], + [-2.8685e-07, -5.5414e-07, 1.9278e-07, ..., 6.5193e-09, + 1.8626e-09, -1.2852e-06], + [ 8.6613e-08, 4.1351e-07, 2.1420e-07, ..., -7.4506e-09, + -3.7253e-09, 2.7660e-07], + ..., + [ 1.2107e-07, 3.7160e-07, 4.6566e-08, ..., 0.0000e+00, + -2.7940e-09, 5.7928e-07], + [-5.4296e-07, 2.1514e-07, 2.1793e-07, ..., 2.1420e-08, + 1.8626e-09, -3.7719e-07], + [ 2.1327e-07, -7.2360e-05, -4.9829e-05, ..., 9.3132e-10, + 9.3132e-10, -6.3702e-06]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0083, -0.0198, -0.0063, -0.0183, -0.0056, 0.0013, 0.0112, 0.0206, + 0.0169, -0.0113], device='cuda:0'), grad: tensor([ 1.1977e-06, -2.8517e-06, 1.5153e-06, 2.9922e-05, 6.6124e-07, + 1.8466e-04, 2.4773e-07, 1.5888e-06, -1.0896e-07, -2.1648e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 214.86, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5033 re_mapping 0.0052 re_causal 0.0151 /// teacc 99.04 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.2361, 0.0833, -0.1131, ..., -0.0290, -0.1711, -0.1503], + [ 0.0108, 0.0992, -0.1166, ..., -0.1101, -0.0417, 0.1007], + [ 0.0506, -0.1184, -0.1235, ..., 0.0024, -0.0514, -0.0828], + ..., + [ 0.0844, -0.0387, 0.1253, ..., 0.0626, 0.1662, -0.0054], + [ 0.1063, -0.1828, -0.1252, ..., -0.2125, -0.0522, 0.1787], + [-0.1478, 0.0791, 0.0441, ..., -0.2069, -0.1046, -0.0405]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, -1.6391e-07, 7.2643e-08, ..., -1.4901e-08, + 0.0000e+00, 4.4703e-08], + [ 2.1420e-08, 3.4459e-08, 1.1548e-07, ..., 5.0291e-08, + 1.1176e-08, -4.1910e-08], + [ 6.5006e-07, 7.0781e-08, 2.1886e-07, ..., 9.4995e-08, + 5.5879e-09, 1.1465e-06], + ..., + [-4.6566e-09, 2.4214e-07, 7.1060e-07, ..., 2.9709e-07, + -3.1665e-08, 3.8184e-08], + [-7.1805e-07, 1.6764e-07, 1.5460e-07, ..., 8.9407e-08, + 3.7253e-09, -1.2843e-06], + [ 2.9802e-08, -3.9116e-07, -1.2759e-06, ..., -3.4459e-08, + 5.5879e-09, 8.2888e-08]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0081, -0.0195, -0.0065, -0.0180, -0.0056, 0.0008, 0.0114, 0.0203, + 0.0165, -0.0111], device='cuda:0'), grad: tensor([-4.0978e-08, 2.7940e-07, 2.7623e-06, -7.9628e-07, 2.5984e-07, + 1.1269e-07, -1.8999e-07, 1.6857e-06, -1.7425e-06, -2.3432e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 215.04, cls_loss 0.0012 cls_loss_mapping 0.0027 cls_loss_causal 0.4946 re_mapping 0.0053 re_causal 0.0158 /// teacc 99.00 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.2366, 0.0827, -0.1137, ..., -0.0289, -0.1720, -0.1502], + [ 0.0106, 0.0994, -0.1168, ..., -0.1104, -0.0418, 0.1009], + [ 0.0506, -0.1185, -0.1239, ..., 0.0024, -0.0514, -0.0830], + ..., + [ 0.0847, -0.0391, 0.1256, ..., 0.0629, 0.1665, -0.0054], + [ 0.1063, -0.1836, -0.1254, ..., -0.2132, -0.0522, 0.1788], + [-0.1482, 0.0797, 0.0442, ..., -0.2084, -0.1049, -0.0406]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.3970e-08, 4.3772e-08, ..., -9.3132e-10, + 9.3132e-09, 1.7323e-07], + [-7.1712e-08, -4.0978e-08, 6.6403e-07, ..., 3.3621e-07, + 6.5193e-09, -2.7940e-07], + [ 2.6077e-08, 3.2317e-07, 3.6322e-08, ..., 1.7695e-08, + 7.4506e-09, 4.3865e-07], + ..., + [-3.8184e-08, 1.5404e-06, 1.9684e-05, ..., 9.8944e-06, + -5.2154e-08, 2.6543e-07], + [ 9.3132e-10, 4.3865e-07, 3.8184e-07, ..., 9.4064e-08, + 9.3132e-10, 1.7975e-07], + [ 2.2352e-08, 1.9949e-06, 3.9302e-07, ..., 4.0885e-07, + 1.8626e-08, 3.3695e-06]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0087, -0.0195, -0.0065, -0.0182, -0.0058, 0.0005, 0.0120, 0.0204, + 0.0162, -0.0108], device='cuda:0'), grad: tensor([ 3.6508e-07, 5.7649e-07, 1.2573e-06, -1.8939e-05, -1.1653e-05, + -1.7703e-05, 9.0431e-07, 3.4600e-05, 1.6810e-06, 8.8811e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 215.00, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4881 re_mapping 0.0055 re_causal 0.0152 /// teacc 98.97 lr 0.00010000 +Epoch 247, weight, value: tensor([[-2.3726e-01, 8.2839e-02, -1.1469e-01, ..., -2.9012e-02, + -1.7283e-01, -1.5050e-01], + [ 1.0028e-02, 9.9441e-02, -1.1733e-01, ..., -1.1276e-01, + -4.2461e-02, 1.0092e-01], + [ 4.8989e-02, -1.1864e-01, -1.2695e-01, ..., 1.2988e-04, + -5.3391e-02, -8.3037e-02], + ..., + [ 8.6123e-02, -3.9210e-02, 1.2717e-01, ..., 6.5527e-02, + 1.6855e-01, -5.4687e-03], + [ 1.0614e-01, -1.8428e-01, -1.2573e-01, ..., -2.1414e-01, + -5.2315e-02, 1.7866e-01], + [-1.4929e-01, 7.9187e-02, 4.3832e-02, ..., -2.1112e-01, + -1.0573e-01, -4.1084e-02]], device='cuda:0'), grad: tensor([[ 4.0047e-08, -5.7742e-08, 5.3085e-08, ..., 7.4506e-09, + 5.4948e-08, 3.7253e-09], + [ 1.5367e-07, -8.0094e-08, 3.6694e-07, ..., 4.9360e-08, + 2.3842e-07, -1.9651e-07], + [ 2.8871e-08, 1.6764e-08, 5.4017e-08, ..., 1.8626e-09, + 4.0047e-08, 1.3039e-08], + ..., + [-3.9488e-06, -1.2293e-07, -7.7039e-06, ..., -8.1863e-07, + -4.8541e-06, 1.3318e-07], + [ 4.0792e-07, 3.6322e-08, 8.4378e-07, ..., 8.2888e-08, + 4.8708e-07, 9.3132e-09], + [ 3.0063e-06, 7.1712e-08, 5.7817e-06, ..., 6.1188e-07, + 3.6545e-06, 5.5879e-09]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0088, -0.0198, -0.0071, -0.0180, -0.0054, 0.0036, 0.0091, 0.0213, + 0.0149, -0.0112], device='cuda:0'), grad: tensor([ 4.4703e-08, 9.7789e-08, 1.0151e-07, 2.5146e-07, 6.9104e-07, + -8.3353e-07, 8.0746e-07, -1.0453e-05, 1.2591e-06, 8.0168e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.92, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5513 re_mapping 0.0050 re_causal 0.0155 /// teacc 99.00 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.2386, 0.0829, -0.1154, ..., -0.0297, -0.1737, -0.1515], + [ 0.0107, 0.0992, -0.1163, ..., -0.1125, -0.0425, 0.1014], + [ 0.0491, -0.1187, -0.1270, ..., 0.0003, -0.0533, -0.0831], + ..., + [ 0.0855, -0.0394, 0.1263, ..., 0.0653, 0.1686, -0.0060], + [ 0.1064, -0.1846, -0.1259, ..., -0.2145, -0.0523, 0.1792], + [-0.1498, 0.0791, 0.0442, ..., -0.2120, -0.1061, -0.0414]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -1.5043e-05, 7.8231e-08, ..., -4.8429e-08, + 1.2107e-08, -1.1874e-06], + [ 1.3560e-06, 1.4976e-05, 5.8636e-06, ..., 1.2852e-07, + 1.3690e-07, 8.1807e-06], + [ 1.8626e-08, 1.1902e-06, 2.1700e-07, ..., 4.1910e-08, + 5.5879e-09, 2.3656e-07], + ..., + [ 2.8312e-07, 2.5313e-06, 1.8226e-06, ..., -4.5635e-08, + -2.3283e-07, 2.7865e-06], + [ 7.6368e-08, 8.2608e-07, 3.0547e-07, ..., 3.9116e-08, + 4.6566e-09, 4.3958e-07], + [-5.3011e-06, -2.9922e-05, -2.4080e-05, ..., 8.8476e-08, + 3.9116e-08, -3.2455e-05]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0089, -0.0194, -0.0071, -0.0181, -0.0052, 0.0038, 0.0090, 0.0206, + 0.0148, -0.0113], device='cuda:0'), grad: tensor([-2.4974e-05, 3.3349e-05, 2.2948e-06, -8.5589e-07, 5.5373e-05, + -9.4622e-07, 8.7619e-06, 6.9067e-06, 2.1420e-06, -8.2076e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 215.16, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4831 re_mapping 0.0051 re_causal 0.0152 /// teacc 99.06 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.2400, 0.0831, -0.1157, ..., -0.0302, -0.1743, -0.1519], + [ 0.0110, 0.0998, -0.1162, ..., -0.1127, -0.0423, 0.1017], + [ 0.0490, -0.1189, -0.1273, ..., 0.0004, -0.0534, -0.0834], + ..., + [ 0.0854, -0.0406, 0.1263, ..., 0.0657, 0.1686, -0.0063], + [ 0.1065, -0.1852, -0.1261, ..., -0.2155, -0.0524, 0.1794], + [-0.1501, 0.0791, 0.0447, ..., -0.2130, -0.1062, -0.0412]], + device='cuda:0'), grad: tensor([[ 1.7136e-07, -1.5140e-05, 2.4773e-07, ..., 2.0489e-08, + 1.3225e-07, 6.4261e-08], + [-2.3283e-08, -2.0415e-06, 8.4937e-07, ..., 6.7055e-08, + 4.7591e-07, -1.9725e-06], + [-8.6613e-08, 1.2107e-07, 6.3330e-08, ..., -1.2293e-07, + -2.2352e-08, 1.4901e-08], + ..., + [-2.7828e-06, 3.2503e-07, -4.6566e-06, ..., -2.7381e-07, + -2.6468e-06, 2.3562e-07], + [ 6.8638e-07, 1.1418e-06, 7.8324e-07, ..., 1.1269e-07, + 4.5076e-07, 6.2026e-07], + [ 1.0822e-06, 3.1665e-07, 1.4426e-06, ..., 9.5926e-08, + 9.6764e-07, 1.2573e-07]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0090, -0.0192, -0.0071, -0.0184, -0.0054, 0.0056, 0.0073, 0.0205, + 0.0145, -0.0112], device='cuda:0'), grad: tensor([-2.5705e-05, -3.3304e-06, 4.6380e-07, 8.6799e-07, 7.6555e-07, + 5.7556e-06, 1.5259e-05, -5.9940e-06, 8.7470e-06, 3.1814e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 215.16, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4550 re_mapping 0.0054 re_causal 0.0148 /// teacc 98.96 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.2409, 0.0832, -0.1162, ..., -0.0302, -0.1757, -0.1521], + [ 0.0103, 0.1000, -0.1170, ..., -0.1135, -0.0430, 0.1017], + [ 0.0489, -0.1191, -0.1276, ..., 0.0004, -0.0535, -0.0836], + ..., + [ 0.0862, -0.0409, 0.1274, ..., 0.0662, 0.1695, -0.0062], + [ 0.1065, -0.1863, -0.1267, ..., -0.2168, -0.0527, 0.1801], + [-0.1509, 0.0791, 0.0448, ..., -0.2152, -0.1066, -0.0414]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -3.7253e-08, 8.3819e-09, ..., 2.1420e-08, + 1.8626e-09, -1.8626e-09], + [ 3.6322e-08, 1.3039e-08, 5.5879e-08, ..., 7.5437e-08, + 3.4459e-08, -2.7008e-08], + [-1.0151e-07, 2.7008e-08, 3.2596e-08, ..., -2.3078e-06, + 1.3970e-08, 5.5879e-09], + ..., + [-9.9652e-08, 1.8626e-08, -1.2480e-07, ..., -4.3772e-08, + -9.0338e-08, 1.6764e-08], + [ 1.0990e-07, 1.0803e-07, 1.7695e-08, ..., 2.1569e-06, + 7.4506e-09, -3.7253e-09], + [ 2.1420e-08, 1.0617e-07, 1.5832e-08, ..., 2.3283e-08, + 1.6764e-08, 5.7742e-08]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0091, -0.0196, -0.0072, -0.0188, -0.0055, 0.0056, 0.0076, 0.0210, + 0.0143, -0.0113], device='cuda:0'), grad: tensor([ 1.5832e-08, 2.4866e-07, -8.7023e-06, -1.3523e-05, -6.0536e-08, + 1.2204e-05, 8.1770e-07, -8.1025e-08, 8.7395e-06, 3.4273e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 215.02, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5108 re_mapping 0.0050 re_causal 0.0152 /// teacc 98.96 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.2420, 0.0833, -0.1173, ..., -0.0305, -0.1769, -0.1525], + [ 0.0107, 0.1001, -0.1149, ..., -0.1106, -0.0440, 0.1015], + [ 0.0491, -0.1192, -0.1278, ..., 0.0009, -0.0534, -0.0837], + ..., + [ 0.0860, -0.0410, 0.1254, ..., 0.0634, 0.1704, -0.0060], + [ 0.1067, -0.1871, -0.1268, ..., -0.2175, -0.0528, 0.1806], + [-0.1518, 0.0792, 0.0450, ..., -0.2166, -0.1069, -0.0417]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, 8.8988e-07, 5.2759e-07, ..., 8.3959e-07, + 1.8626e-09, 8.8941e-08], + [ 6.9803e-07, -7.1712e-08, 9.4296e-07, ..., 6.2026e-07, + 5.5181e-07, -1.2107e-08], + [-6.5193e-09, 8.6613e-08, 1.3970e-07, ..., 1.3271e-07, + 8.8476e-09, 3.5390e-08], + ..., + [-1.3653e-06, 9.9186e-08, -1.3905e-06, ..., -6.9663e-07, + -1.0515e-06, -1.8394e-07], + [-1.8207e-07, 8.4657e-07, 6.8685e-07, ..., 9.4669e-07, + 1.7323e-07, -3.7719e-07], + [ 3.9861e-07, 2.9663e-07, 3.1479e-07, ..., 3.2736e-07, + 5.3085e-08, 3.8883e-07]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0091, -0.0183, -0.0070, -0.0186, -0.0056, 0.0056, 0.0076, 0.0196, + 0.0143, -0.0113], device='cuda:0'), grad: tensor([ 2.7549e-06, 1.7229e-06, 3.8370e-07, -7.3537e-06, 1.1362e-07, + 1.0133e-06, 7.4506e-09, -2.5444e-06, 1.7332e-06, 2.1756e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 214.84, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4859 re_mapping 0.0050 re_causal 0.0145 /// teacc 99.00 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.2422, 0.0837, -0.1177, ..., -0.0307, -0.1782, -0.1524], + [ 0.0106, 0.1002, -0.1150, ..., -0.1106, -0.0442, 0.1018], + [ 0.0492, -0.1193, -0.1280, ..., 0.0010, -0.0533, -0.0839], + ..., + [ 0.0860, -0.0414, 0.1256, ..., 0.0635, 0.1706, -0.0062], + [ 0.1068, -0.1875, -0.1269, ..., -0.2177, -0.0529, 0.1808], + [-0.1525, 0.0791, 0.0451, ..., -0.2176, -0.1072, -0.0420]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 3.3528e-08, 1.8626e-09, ..., 1.8626e-09, + 9.3132e-10, 1.8626e-09], + [-4.1444e-08, -1.2154e-07, 2.1886e-08, ..., 5.5879e-09, + 1.7695e-08, -2.4727e-07], + [-2.0955e-08, 7.4506e-09, 2.3283e-09, ..., -4.5635e-08, + -1.4901e-08, 1.2107e-08], + ..., + [ 9.3132e-10, 1.1036e-07, -5.9139e-08, ..., 6.0536e-09, + -3.8650e-08, 1.9232e-07], + [-4.6566e-09, 2.9057e-07, 3.8184e-08, ..., 9.3132e-09, + 6.0536e-09, -1.2107e-08], + [ 3.1199e-08, -5.0291e-08, -4.3772e-08, ..., 3.2596e-09, + 2.2817e-08, 1.5367e-08]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0091, -0.0183, -0.0071, -0.0186, -0.0054, 0.0055, 0.0077, 0.0196, + 0.0143, -0.0116], device='cuda:0'), grad: tensor([ 5.6578e-07, -3.3295e-07, 1.8626e-09, 1.7649e-07, 4.3958e-06, + -6.6590e-07, -6.0238e-06, 3.3155e-07, 1.5991e-06, -6.5658e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 214.84, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5119 re_mapping 0.0053 re_causal 0.0161 /// teacc 98.99 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.2419, 0.0830, -0.1185, ..., -0.0310, -0.1787, -0.1522], + [ 0.0114, 0.1002, -0.1149, ..., -0.1107, -0.0437, 0.1025], + [ 0.0493, -0.1194, -0.1281, ..., 0.0012, -0.0532, -0.0840], + ..., + [ 0.0854, -0.0418, 0.1256, ..., 0.0635, 0.1704, -0.0069], + [ 0.1070, -0.1878, -0.1270, ..., -0.2180, -0.0529, 0.1810], + [-0.1537, 0.0787, 0.0442, ..., -0.2211, -0.1083, -0.0426]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, -8.5635e-07, 1.8626e-09, ..., 3.7719e-08, + 5.2154e-08, 8.8476e-09], + [ 1.3877e-07, -5.0571e-07, 4.1910e-09, ..., 1.1642e-07, + 1.6345e-07, -1.1120e-06], + [-1.1232e-06, 4.0978e-08, 7.4506e-09, ..., -6.5705e-07, + -9.0944e-07, 6.2864e-08], + ..., + [ 4.5123e-07, 1.6950e-07, -1.7229e-08, ..., 2.5611e-07, + 3.4971e-07, 3.1898e-07], + [ 1.6764e-07, 1.3737e-07, 1.3970e-08, ..., 1.2387e-07, + 1.7043e-07, -9.3132e-10], + [ 3.7253e-08, 4.0513e-08, -4.8894e-08, ..., 1.9558e-08, + 2.7474e-08, 7.2177e-08]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0098, -0.0179, -0.0070, -0.0187, -0.0043, 0.0054, 0.0079, 0.0191, + 0.0143, -0.0124], device='cuda:0'), grad: tensor([-5.1595e-07, -1.4920e-06, -3.0696e-06, 3.1106e-07, 7.1712e-07, + 2.5779e-06, -2.4065e-06, 2.0899e-06, 1.5628e-06, 2.0815e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 215.00, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.5275 re_mapping 0.0053 re_causal 0.0157 /// teacc 99.02 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.2421, 0.0831, -0.1189, ..., -0.0314, -0.1803, -0.1519], + [ 0.0116, 0.1003, -0.1149, ..., -0.1107, -0.0436, 0.1027], + [ 0.0494, -0.1194, -0.1280, ..., 0.0015, -0.0531, -0.0839], + ..., + [ 0.0853, -0.0421, 0.1256, ..., 0.0634, 0.1704, -0.0071], + [ 0.1070, -0.1886, -0.1271, ..., -0.2185, -0.0530, 0.1812], + [-0.1543, 0.0784, 0.0440, ..., -0.2220, -0.1086, -0.0430]], + device='cuda:0'), grad: tensor([[ 2.3469e-07, -1.6764e-08, 1.1176e-07, ..., 6.7055e-08, + 2.0489e-07, 5.5879e-08], + [ 9.0338e-07, -6.8918e-08, 4.8243e-07, ..., 2.7567e-07, + 8.4937e-07, -1.3039e-08], + [ 3.5167e-06, 3.7253e-08, 3.3975e-06, ..., -1.3281e-06, + 4.8913e-06, 1.6503e-06], + ..., + [-8.3372e-06, 3.7253e-08, -5.2974e-06, ..., -2.4028e-07, + -8.9034e-06, -2.2966e-06], + [ 1.1418e-06, 5.0291e-08, 4.9919e-07, ..., 4.1537e-07, + 9.3877e-07, 2.0117e-07], + [ 1.0245e-06, -1.0561e-06, -3.8743e-07, ..., 2.7567e-07, + 9.3319e-07, 2.2911e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0098, -0.0177, -0.0069, -0.0187, -0.0040, 0.0055, 0.0078, 0.0190, + 0.0140, -0.0128], device='cuda:0'), grad: tensor([ 5.1409e-07, 1.7826e-06, 8.9332e-06, 6.0536e-07, 2.7213e-06, + 1.3504e-06, 3.2969e-07, -1.8716e-05, 2.6114e-06, -1.6950e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 215.09, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5052 re_mapping 0.0050 re_causal 0.0150 /// teacc 98.99 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.2419, 0.0834, -0.1195, ..., -0.0318, -0.1823, -0.1517], + [ 0.0143, 0.1003, -0.1124, ..., -0.1107, -0.0428, 0.1040], + [ 0.0496, -0.1195, -0.1285, ..., 0.0017, -0.0531, -0.0840], + ..., + [ 0.0828, -0.0424, 0.1233, ..., 0.0635, 0.1698, -0.0086], + [ 0.1071, -0.1883, -0.1275, ..., -0.2192, -0.0533, 0.1818], + [-0.1555, 0.0785, 0.0440, ..., -0.2232, -0.1094, -0.0432]], + device='cuda:0'), grad: tensor([[ 5.7742e-08, -2.5891e-07, 1.4901e-08, ..., 1.8626e-09, + 1.8626e-09, 1.4156e-07], + [ 7.8231e-08, 7.4506e-09, 5.9605e-08, ..., 2.6077e-08, + 5.7742e-08, 4.6566e-08], + [ 8.5682e-08, 1.8626e-09, 5.0291e-08, ..., -1.8626e-09, + 3.1665e-08, 1.4901e-07], + ..., + [-1.1362e-07, 3.7253e-09, -1.1548e-07, ..., -4.0978e-08, + -1.2293e-07, 2.0489e-08], + [-2.5332e-07, 1.5460e-07, -5.2154e-08, ..., 5.5879e-09, + 9.3132e-09, -6.7055e-07], + [ 2.9802e-08, 2.1793e-07, -0.0000e+00, ..., 1.8626e-09, + 7.4506e-09, 3.9116e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0097, -0.0157, -0.0069, -0.0188, -0.0041, 0.0054, 0.0080, 0.0168, + 0.0143, -0.0130], device='cuda:0'), grad: tensor([-7.7672e-07, 2.5518e-07, 4.0233e-07, 3.4794e-06, 6.1467e-08, + -4.1127e-06, 1.0971e-06, -1.2666e-07, -1.1064e-06, 8.1211e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 215.17, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4893 re_mapping 0.0052 re_causal 0.0149 /// teacc 98.92 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.2431, 0.0835, -0.1203, ..., -0.0323, -0.1831, -0.1519], + [ 0.0138, 0.1001, -0.1128, ..., -0.1108, -0.0435, 0.1038], + [ 0.0497, -0.1196, -0.1287, ..., 0.0019, -0.0531, -0.0840], + ..., + [ 0.0831, -0.0423, 0.1237, ..., 0.0634, 0.1705, -0.0084], + [ 0.1076, -0.1888, -0.1275, ..., -0.2195, -0.0530, 0.1826], + [-0.1561, 0.0787, 0.0444, ..., -0.2240, -0.1096, -0.0435]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -1.1735e-07, 3.7253e-08, ..., 1.8626e-09, + 1.3039e-08, 1.4901e-08], + [ 1.3411e-07, 5.5879e-09, 1.8068e-07, ..., 7.6368e-08, + 1.1176e-08, 1.2666e-07], + [ 8.9407e-08, 1.8626e-08, 3.5390e-08, ..., 0.0000e+00, + 9.3132e-09, 1.0990e-07], + ..., + [-5.1260e-06, 1.1921e-07, -3.6880e-06, ..., 3.2783e-07, + -1.7062e-06, 1.0431e-07], + [-8.7544e-07, 6.5193e-08, 5.0291e-08, ..., 5.5879e-09, + 1.6764e-08, -1.4715e-06], + [ 1.3597e-07, 3.5018e-07, 3.9116e-07, ..., 1.3970e-07, + 4.4703e-08, 1.9744e-07]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0097, -0.0160, -0.0068, -0.0188, -0.0044, 0.0053, 0.0081, 0.0171, + 0.0145, -0.0126], device='cuda:0'), grad: tensor([-1.5087e-07, 6.1467e-07, 3.4086e-07, -2.9430e-07, -1.2778e-06, + 9.0078e-06, 1.7881e-07, -7.0259e-06, -3.0436e-06, 1.6373e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 214.98, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5034 re_mapping 0.0053 re_causal 0.0157 /// teacc 98.92 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.2435, 0.0839, -0.1208, ..., -0.0324, -0.1838, -0.1515], + [ 0.0134, 0.1002, -0.1131, ..., -0.1109, -0.0446, 0.1034], + [ 0.0498, -0.1196, -0.1289, ..., 0.0020, -0.0531, -0.0841], + ..., + [ 0.0836, -0.0424, 0.1241, ..., 0.0634, 0.1717, -0.0080], + [ 0.1080, -0.1892, -0.1276, ..., -0.2197, -0.0530, 0.1832], + [-0.1574, 0.0787, 0.0441, ..., -0.2247, -0.1110, -0.0438]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, -2.9616e-07, 2.7940e-08, ..., 1.6764e-08, + 1.8626e-08, 7.4506e-09], + [ 3.0361e-07, -3.1479e-07, 2.7195e-07, ..., 2.0675e-07, + 2.2724e-07, -9.7416e-07], + [-2.6822e-07, 2.4214e-08, 3.3528e-08, ..., -3.4831e-07, + -7.8231e-08, 3.9116e-08], + ..., + [-6.3144e-07, -6.8918e-08, -1.1120e-06, ..., 3.5390e-08, + -7.3947e-07, 8.1956e-08], + [ 6.3330e-08, 7.8231e-08, 1.0431e-07, ..., 1.4901e-07, + 4.4703e-08, 5.0291e-08], + [ 4.2468e-07, 1.7881e-07, 6.5938e-07, ..., 2.9802e-08, + 4.6194e-07, 2.9802e-08]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0091, -0.0163, -0.0068, -0.0185, -0.0044, 0.0050, 0.0080, 0.0175, + 0.0149, -0.0128], device='cuda:0'), grad: tensor([-6.4261e-07, -9.9652e-07, -9.5367e-07, -4.1164e-07, -9.3132e-09, + 4.4703e-07, 1.7751e-06, -7.9535e-07, 5.7183e-07, 9.9652e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 215.12, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.5248 re_mapping 0.0052 re_causal 0.0152 /// teacc 99.01 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.2447, 0.0843, -0.1217, ..., -0.0365, -0.1850, -0.1525], + [ 0.0132, 0.1002, -0.1133, ..., -0.1109, -0.0452, 0.1035], + [ 0.0498, -0.1197, -0.1292, ..., 0.0020, -0.0531, -0.0847], + ..., + [ 0.0839, -0.0426, 0.1244, ..., 0.0634, 0.1723, -0.0079], + [ 0.1081, -0.1900, -0.1278, ..., -0.2203, -0.0530, 0.1833], + [-0.1582, 0.0784, 0.0442, ..., -0.2254, -0.1116, -0.0439]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.0864e-06, 5.5879e-09, ..., 6.2101e-06, + 0.0000e+00, 1.8626e-09], + [ 3.7253e-09, 3.7253e-08, 2.6077e-08, ..., 3.3528e-08, + 3.7253e-09, -1.0058e-07], + [ 5.5879e-09, 9.8720e-08, 5.5879e-09, ..., 1.9558e-07, + 1.8626e-09, 3.1665e-08], + ..., + [-7.4506e-09, 2.9802e-08, 1.1176e-08, ..., 3.7253e-09, + -7.4506e-09, 5.4017e-08], + [-6.8918e-08, 8.7544e-08, 2.6077e-08, ..., 1.3225e-07, + 1.8626e-09, -1.3411e-07], + [ 2.0489e-08, 5.5879e-09, -6.3330e-07, ..., 5.5879e-09, + 0.0000e+00, 6.7241e-07]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0094, -0.0163, -0.0069, -0.0176, -0.0042, 0.0050, 0.0080, 0.0176, + 0.0146, -0.0132], device='cuda:0'), grad: tensor([ 2.5183e-05, 1.4156e-07, 8.6613e-07, -2.9266e-05, -5.2154e-08, + 1.8068e-06, 3.0920e-07, 1.5646e-07, 3.8929e-07, 4.8056e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 214.96, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4852 re_mapping 0.0050 re_causal 0.0154 /// teacc 98.90 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.2451, 0.0847, -0.1220, ..., -0.0369, -0.1855, -0.1528], + [ 0.0132, 0.1001, -0.1134, ..., -0.1109, -0.0454, 0.1036], + [ 0.0501, -0.1197, -0.1292, ..., 0.0021, -0.0527, -0.0848], + ..., + [ 0.0839, -0.0428, 0.1244, ..., 0.0634, 0.1724, -0.0080], + [ 0.1082, -0.1906, -0.1281, ..., -0.2205, -0.0530, 0.1837], + [-0.1586, 0.0784, 0.0446, ..., -0.2256, -0.1118, -0.0442]], + device='cuda:0'), grad: tensor([[-1.4901e-08, 5.2899e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [-5.5879e-09, 6.8918e-08, 3.7253e-09, ..., 1.8626e-09, + 3.7253e-09, -8.5682e-08], + [ 0.0000e+00, 3.5577e-07, 0.0000e+00, ..., -3.7253e-09, + -1.8626e-09, 4.8429e-08], + ..., + [-5.5879e-09, 5.9605e-08, -5.5879e-09, ..., -1.8626e-09, + -5.5879e-09, 3.7253e-08], + [-3.9116e-08, 3.5390e-08, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, -4.8429e-08], + [ 2.4214e-08, 2.0675e-07, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0090, -0.0164, -0.0067, -0.0171, -0.0041, 0.0049, 0.0078, 0.0175, + 0.0146, -0.0133], device='cuda:0'), grad: tensor([ 3.1069e-06, 6.5006e-07, 1.6298e-06, 5.2154e-08, -7.8008e-06, + 3.5390e-08, 1.2759e-06, 1.8440e-07, 1.6764e-08, 8.3447e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 214.62, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.5221 re_mapping 0.0050 re_causal 0.0159 /// teacc 99.01 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.2457, 0.0848, -0.1224, ..., -0.0369, -0.1862, -0.1535], + [ 0.0132, 0.0996, -0.1134, ..., -0.1110, -0.0454, 0.1039], + [ 0.0496, -0.1199, -0.1301, ..., 0.0019, -0.0531, -0.0850], + ..., + [ 0.0840, -0.0429, 0.1246, ..., 0.0636, 0.1727, -0.0081], + [ 0.1082, -0.1912, -0.1284, ..., -0.2207, -0.0532, 0.1838], + [-0.1571, 0.0811, 0.0483, ..., -0.2260, -0.1120, -0.0412]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -6.9477e-07, 3.7253e-09, ..., 1.8626e-09, + 1.8626e-09, -2.8685e-07], + [-1.5832e-07, -4.8988e-07, 1.1176e-08, ..., 0.0000e+00, + -1.8626e-09, -6.5751e-07], + [ 3.7253e-08, 2.9802e-07, 3.7253e-09, ..., -5.5879e-09, + -1.8626e-09, 2.7940e-07], + ..., + [ 3.1665e-08, 1.9744e-07, 7.4506e-09, ..., 0.0000e+00, + -7.4506e-09, 2.6636e-07], + [-1.6764e-08, 5.0105e-07, 4.2841e-08, ..., 1.8626e-09, + 1.8626e-09, 2.0862e-07], + [ 3.3528e-08, 1.9550e-05, 3.0827e-06, ..., 0.0000e+00, + 3.7253e-09, 9.1046e-06]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0089, -0.0163, -0.0069, -0.0176, -0.0069, 0.0051, 0.0078, 0.0175, + 0.0144, -0.0105], device='cuda:0'), grad: tensor([ 2.7753e-07, -8.9779e-07, 1.3839e-06, 5.3458e-07, -4.9055e-05, + 1.1437e-06, -1.1787e-05, 5.9605e-07, 8.2403e-06, 4.9531e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 214.75, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4975 re_mapping 0.0051 re_causal 0.0146 /// teacc 98.92 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.2455, 0.0874, -0.1217, ..., -0.0370, -0.1879, -0.1514], + [ 0.0144, 0.1018, -0.1129, ..., -0.1111, -0.0441, 0.1055], + [ 0.0494, -0.1202, -0.1311, ..., 0.0018, -0.0534, -0.0853], + ..., + [ 0.0830, -0.0454, 0.1242, ..., 0.0637, 0.1718, -0.0096], + [ 0.1081, -0.1929, -0.1288, ..., -0.2212, -0.0533, 0.1838], + [-0.1579, 0.0818, 0.0505, ..., -0.2265, -0.1126, -0.0397]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.7253e-09, 5.5879e-09, ..., 1.8626e-09, + 1.8626e-09, 1.1176e-08], + [ 8.0094e-08, 1.3039e-08, 1.4901e-08, ..., 8.5682e-08, + 7.8231e-08, 2.4214e-08], + [-8.9407e-08, 1.4901e-08, 7.4506e-09, ..., -1.0058e-07, + -8.7544e-08, 1.4901e-08], + ..., + [-3.7253e-09, 9.8720e-08, 3.7253e-09, ..., 5.5879e-09, + -7.4506e-09, 8.9407e-08], + [-1.3039e-08, 1.4342e-07, 2.2911e-07, ..., 1.8626e-09, + 1.8626e-09, 6.3330e-08], + [ 9.3132e-09, 3.5949e-07, -6.5193e-07, ..., 3.7253e-09, + 9.3132e-09, 4.1537e-07]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0062, -0.0156, -0.0070, -0.0175, -0.0089, 0.0049, 0.0078, 0.0169, + 0.0138, -0.0093], device='cuda:0'), grad: tensor([ 3.5390e-08, 5.0105e-07, -3.8929e-07, 3.2410e-07, -1.8664e-06, + -1.8291e-06, 1.1008e-06, 4.0047e-07, 1.2387e-06, 4.8056e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 214.89, cls_loss 0.0018 cls_loss_mapping 0.0029 cls_loss_causal 0.5069 re_mapping 0.0052 re_causal 0.0148 /// teacc 99.00 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.2473, 0.0877, -0.1237, ..., -0.0374, -0.1915, -0.1524], + [ 0.0154, 0.1030, -0.1123, ..., -0.1111, -0.0429, 0.1066], + [ 0.0495, -0.1200, -0.1314, ..., 0.0020, -0.0533, -0.0854], + ..., + [ 0.0824, -0.0467, 0.1239, ..., 0.0641, 0.1713, -0.0104], + [ 0.1082, -0.1944, -0.1292, ..., -0.2219, -0.0536, 0.1839], + [-0.1606, 0.0816, 0.0506, ..., -0.2274, -0.1145, -0.0401]], + device='cuda:0'), grad: tensor([[ 9.6485e-07, 2.7195e-07, 5.5879e-09, ..., 4.7870e-07, + 4.9360e-07, 3.2969e-07], + [-3.3714e-07, -1.4827e-06, 9.3132e-09, ..., 1.5460e-07, + 3.1665e-08, -1.3616e-06], + [-1.3784e-06, 5.0291e-08, 3.7253e-09, ..., -1.1101e-06, + -7.9907e-07, 4.6566e-08], + ..., + [ 1.1362e-07, 9.8720e-08, -1.4901e-08, ..., 6.7055e-08, + 4.2841e-08, 1.0431e-07], + [ 3.8929e-07, 3.1292e-07, 3.7253e-09, ..., 1.7695e-07, + 1.8068e-07, 2.6822e-07], + [ 3.1665e-08, 1.5087e-07, -2.4214e-08, ..., 5.5879e-09, + 5.5879e-09, 8.1956e-08]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0061, -0.0147, -0.0067, -0.0197, -0.0088, 0.0052, 0.0075, 0.0162, + 0.0133, -0.0096], device='cuda:0'), grad: tensor([ 3.8221e-06, -2.8405e-06, -7.5772e-06, 1.4342e-07, 6.5938e-07, + 3.2037e-07, 2.2147e-06, 7.4878e-07, 1.9968e-06, 4.4890e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.94, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4980 re_mapping 0.0051 re_causal 0.0154 /// teacc 99.03 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.2485, 0.0876, -0.1245, ..., -0.0379, -0.1936, -0.1534], + [ 0.0156, 0.1036, -0.1119, ..., -0.1107, -0.0428, 0.1069], + [ 0.0495, -0.1203, -0.1316, ..., 0.0020, -0.0533, -0.0857], + ..., + [ 0.0822, -0.0468, 0.1236, ..., 0.0638, 0.1714, -0.0105], + [ 0.1082, -0.1957, -0.1295, ..., -0.2225, -0.0537, 0.1839], + [-0.1615, 0.0813, 0.0504, ..., -0.2288, -0.1153, -0.0407]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.8626e-09, 1.8626e-09, ..., 5.5879e-09, + 7.4506e-09, 5.5879e-09], + [ 8.2701e-07, -1.0431e-07, 5.4017e-08, ..., 1.6764e-08, + 3.0361e-07, -4.5635e-07], + [ 1.3225e-07, 3.9116e-08, 2.4214e-08, ..., -2.2352e-08, + 5.0291e-08, 2.1607e-07], + ..., + [-1.5065e-05, 7.2643e-08, -7.8045e-07, ..., 1.4901e-08, + -7.1973e-06, -7.7561e-06], + [ 4.5821e-07, 8.0094e-08, 2.6077e-08, ..., 5.5879e-09, + 2.2538e-07, 3.4831e-07], + [ 2.9802e-08, 5.5879e-09, 5.5879e-09, ..., 1.8626e-09, + 1.6764e-08, 2.2352e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0064, -0.0142, -0.0068, -0.0195, -0.0086, 0.0051, 0.0074, 0.0159, + 0.0129, -0.0099], device='cuda:0'), grad: tensor([ 3.5763e-07, -5.0478e-07, 6.2585e-07, 1.4827e-06, 2.2441e-05, + -5.9307e-06, -4.7088e-06, -1.8701e-05, 4.8429e-06, 8.0094e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 215.14, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5000 re_mapping 0.0051 re_causal 0.0147 /// teacc 98.99 lr 0.00010000 +Epoch 264, weight, value: tensor([[-2.4934e-01, 8.7733e-02, -1.2530e-01, ..., -3.8183e-02, + -1.9506e-01, -1.5312e-01], + [ 1.4994e-02, 1.0367e-01, -1.1237e-01, ..., -1.1099e-01, + -4.4027e-02, 1.0671e-01], + [ 4.9086e-02, -1.2046e-01, -1.3335e-01, ..., -5.3777e-06, + -5.3792e-02, -8.5791e-02], + ..., + [ 8.3102e-02, -4.6851e-02, 1.2423e-01, ..., 6.4068e-02, + 1.7311e-01, -1.0195e-02], + [ 1.0853e-01, -1.9584e-01, -1.3002e-01, ..., -2.2325e-01, + -5.3950e-02, 1.8454e-01], + [-1.6298e-01, 8.1286e-02, 5.0317e-02, ..., -2.3228e-01, + -1.1637e-01, -4.0816e-02]], device='cuda:0'), grad: tensor([[ 1.3039e-08, 5.7742e-08, 3.3528e-08, ..., 0.0000e+00, + 1.8626e-09, 1.4901e-08], + [ 3.3528e-07, 2.4028e-06, 8.1956e-08, ..., 3.7253e-09, + 1.4901e-08, 1.5050e-06], + [ 9.4995e-08, 1.8626e-08, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 9.4995e-08], + ..., + [ 3.7253e-09, 2.9989e-07, 4.4890e-07, ..., -7.4506e-09, + -1.6764e-08, 5.5879e-08], + [-1.4901e-08, 2.1048e-07, 2.4959e-07, ..., 0.0000e+00, + 7.4506e-09, -8.0094e-08], + [-5.4017e-08, -1.2666e-07, -1.1176e-06, ..., 1.8626e-09, + -1.3039e-08, 2.1979e-07]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0064, -0.0145, -0.0077, -0.0170, -0.0086, 0.0046, 0.0074, 0.0165, + 0.0133, -0.0100], device='cuda:0'), grad: tensor([ 1.2666e-06, 6.0350e-06, 4.9733e-07, 2.6636e-07, -8.5086e-06, + -4.9360e-07, 1.0990e-07, 8.8476e-07, 7.1712e-07, -8.1770e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 215.09, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.5175 re_mapping 0.0050 re_causal 0.0152 /// teacc 98.96 lr 0.00010000 +Epoch 265, weight, value: tensor([[-2.5100e-01, 8.7694e-02, -1.2624e-01, ..., -3.8266e-02, + -1.9723e-01, -1.5306e-01], + [ 1.4832e-02, 1.0374e-01, -1.1252e-01, ..., -1.1112e-01, + -4.4380e-02, 1.0675e-01], + [ 4.9198e-02, -1.2054e-01, -1.3353e-01, ..., 1.3101e-04, + -5.3759e-02, -8.5866e-02], + ..., + [ 8.3332e-02, -4.6919e-02, 1.2442e-01, ..., 6.4142e-02, + 1.7354e-01, -1.0191e-02], + [ 1.0858e-01, -1.9649e-01, -1.3041e-01, ..., -2.2369e-01, + -5.4105e-02, 1.8475e-01], + [-1.6355e-01, 8.1277e-02, 5.0412e-02, ..., -2.3367e-01, + -1.1674e-01, -4.0880e-02]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.2841e-08, 3.1665e-08, ..., 5.5879e-09, + 0.0000e+00, 2.6077e-08], + [ 1.6764e-08, 8.3819e-06, 7.3165e-06, ..., 8.5682e-08, + 1.6764e-08, 4.0159e-06], + [ 0.0000e+00, 4.4703e-08, 4.4703e-08, ..., 9.3132e-09, + 0.0000e+00, 1.8626e-08], + ..., + [-5.2154e-08, 1.3132e-06, 1.1921e-06, ..., 1.0915e-06, + -3.9116e-08, 2.7567e-07], + [ 5.5879e-09, 8.6613e-07, 7.6555e-07, ..., 2.4959e-07, + 5.5879e-09, 3.3900e-07], + [ 7.4506e-09, 6.2287e-06, 5.2415e-06, ..., 3.5390e-08, + 1.8626e-09, 3.0864e-06]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0066, -0.0146, -0.0077, -0.0169, -0.0086, 0.0046, 0.0073, 0.0166, + 0.0131, -0.0100], device='cuda:0'), grad: tensor([ 1.6019e-07, 1.8045e-05, 9.6858e-08, -9.3877e-05, 2.1979e-07, + 5.5939e-05, -5.5879e-09, 3.8669e-06, 2.1420e-06, 1.3404e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 214.84, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4862 re_mapping 0.0052 re_causal 0.0149 /// teacc 98.88 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.2520, 0.0872, -0.1269, ..., -0.0384, -0.1986, -0.1532], + [ 0.0145, 0.1036, -0.1128, ..., -0.1113, -0.0450, 0.1067], + [ 0.0496, -0.1206, -0.1340, ..., 0.0007, -0.0535, -0.0859], + ..., + [ 0.0837, -0.0470, 0.1249, ..., 0.0643, 0.1743, -0.0102], + [ 0.1086, -0.1972, -0.1310, ..., -0.2245, -0.0543, 0.1848], + [-0.1635, 0.0827, 0.0515, ..., -0.2349, -0.1170, -0.0399]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, -5.5321e-07, 5.4017e-08, ..., 1.1176e-08, + 2.2352e-08, 1.8626e-09], + [ 7.6182e-07, 1.7509e-07, 9.4809e-07, ..., 1.5832e-07, + 5.7928e-07, -5.5879e-08], + [ 1.6578e-07, 7.6368e-08, 2.5332e-07, ..., 9.6858e-08, + 1.3970e-07, 1.8626e-08], + ..., + [-2.3022e-06, -2.4028e-07, -2.6785e-06, ..., -3.6880e-07, + -1.7378e-06, 3.5390e-08], + [ 5.7742e-08, 2.1793e-07, 1.6019e-07, ..., 7.0781e-08, + 4.0978e-08, 2.2352e-08], + [ 1.0934e-06, 4.9546e-07, 1.3355e-06, ..., 2.1793e-07, + 8.0466e-07, 8.3819e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0072, -0.0148, -0.0075, -0.0174, -0.0099, 0.0047, 0.0076, 0.0168, + 0.0126, -0.0088], device='cuda:0'), grad: tensor([-1.0394e-06, 1.7900e-06, 6.1095e-07, -9.3132e-07, -8.8289e-07, + 6.5193e-07, 1.9185e-07, -4.7423e-06, 7.9535e-07, 3.5372e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 214.71, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.4964 re_mapping 0.0052 re_causal 0.0148 /// teacc 98.91 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.2531, 0.0860, -0.1278, ..., -0.0385, -0.1997, -0.1537], + [ 0.0143, 0.1019, -0.1126, ..., -0.1109, -0.0452, 0.1064], + [ 0.0497, -0.1208, -0.1344, ..., 0.0007, -0.0534, -0.0863], + ..., + [ 0.0837, -0.0472, 0.1246, ..., 0.0640, 0.1746, -0.0106], + [ 0.1129, -0.1957, -0.1281, ..., -0.2253, -0.0544, 0.1896], + [-0.1667, 0.0835, 0.0494, ..., -0.2360, -0.1171, -0.0424]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -4.4703e-08, 1.1176e-08, ..., 1.6764e-08, + 1.1176e-08, 2.9802e-08], + [ 5.6140e-06, -1.5087e-07, 5.1335e-06, ..., 1.3970e-07, + 5.0850e-06, 2.4661e-06], + [ 1.8626e-08, 2.9802e-08, 6.5193e-08, ..., -7.8231e-08, + -6.8918e-08, 1.0990e-07], + ..., + [-5.7854e-06, 1.7509e-07, -5.3197e-06, ..., -9.3132e-08, + -5.2750e-06, -2.8424e-06], + [-2.2352e-08, 6.3330e-08, 4.2841e-08, ..., 1.3039e-08, + 3.1665e-08, 9.3132e-08], + [ 5.3830e-07, 3.6508e-07, 3.1665e-08, ..., 1.1176e-08, + 1.2107e-07, 1.4268e-06]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0089, -0.0151, -0.0076, -0.0179, -0.0098, 0.0047, 0.0076, 0.0164, + 0.0173, -0.0109], device='cuda:0'), grad: tensor([-2.9802e-08, 1.3232e-05, 9.3132e-08, -9.3132e-09, -4.1947e-06, + 4.5896e-06, -9.5963e-06, -9.1121e-06, 1.4305e-06, 3.5465e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 214.96, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5036 re_mapping 0.0051 re_causal 0.0143 /// teacc 98.97 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.2537, 0.0860, -0.1284, ..., -0.0402, -0.2009, -0.1540], + [ 0.0145, 0.1013, -0.1121, ..., -0.1101, -0.0456, 0.1070], + [ 0.0527, -0.1194, -0.1341, ..., 0.0037, -0.0507, -0.0864], + ..., + [ 0.0821, -0.0474, 0.1241, ..., 0.0609, 0.1732, -0.0111], + [ 0.1129, -0.1959, -0.1281, ..., -0.2260, -0.0546, 0.1898], + [-0.1667, 0.0833, 0.0495, ..., -0.2368, -0.1177, -0.0426]], + device='cuda:0'), grad: tensor([[ 5.5879e-08, 6.3330e-08, 1.4901e-08, ..., 6.8918e-08, + 3.3528e-08, 1.0617e-07], + [ 8.1956e-08, -3.4831e-07, 1.9744e-07, ..., 2.4028e-07, + 2.0675e-07, -7.4692e-07], + [-6.8918e-07, 5.7742e-08, 1.6764e-07, ..., -7.3947e-07, + -6.0163e-07, 1.2666e-07], + ..., + [ 2.6636e-07, 6.8918e-08, -3.7067e-07, ..., 3.4831e-07, + 1.8254e-07, 1.4901e-07], + [ 5.4017e-08, 1.6391e-07, 2.2352e-08, ..., 8.3819e-08, + 6.1467e-08, 9.4995e-08], + [ 1.0058e-07, 4.6194e-07, 1.2107e-07, ..., 9.8720e-08, + 5.0291e-08, 7.8976e-07]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0092, -0.0149, -0.0045, -0.0180, -0.0094, 0.0047, 0.0076, 0.0147, + 0.0173, -0.0111], device='cuda:0'), grad: tensor([ 5.3085e-07, -8.7544e-07, -1.7248e-06, -8.2888e-07, -2.4047e-06, + 9.8906e-07, -1.0058e-07, 1.1623e-06, 6.7428e-07, 2.5705e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.90, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4941 re_mapping 0.0052 re_causal 0.0149 /// teacc 98.99 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.2542, 0.0861, -0.1289, ..., -0.0405, -0.2015, -0.1550], + [ 0.0145, 0.1012, -0.1119, ..., -0.1100, -0.0459, 0.1070], + [ 0.0526, -0.1195, -0.1344, ..., 0.0036, -0.0508, -0.0865], + ..., + [ 0.0823, -0.0473, 0.1240, ..., 0.0610, 0.1738, -0.0111], + [ 0.1129, -0.1960, -0.1282, ..., -0.2270, -0.0548, 0.1899], + [-0.1668, 0.0832, 0.0495, ..., -0.2379, -0.1187, -0.0426]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-07, -1.8626e-09, 1.7136e-07, ..., 7.6368e-08, + 1.7136e-07, -1.6764e-08], + [ 5.2154e-08, 0.0000e+00, 4.0978e-08, ..., 3.5390e-08, + 5.0291e-08, 1.8626e-09], + ..., + [-4.5635e-07, 0.0000e+00, -4.1164e-07, ..., -2.0303e-07, + -4.2282e-07, 5.5879e-09], + [ 4.4703e-08, 1.8626e-09, 4.8429e-08, ..., 4.6566e-08, + 5.9605e-08, -1.4901e-08], + [ 1.2852e-07, 1.8626e-09, 1.2666e-07, ..., 2.2352e-08, + 1.0990e-07, 5.5879e-09]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0097, -0.0147, -0.0047, -0.0161, -0.0091, 0.0033, 0.0084, 0.0146, + 0.0172, -0.0112], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.9616e-07, 9.1270e-08, 1.0617e-07, 1.6764e-08, + -1.3039e-08, -2.3097e-07, -7.9721e-07, 2.8685e-07, 2.3842e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 215.00, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.5021 re_mapping 0.0050 re_causal 0.0148 /// teacc 98.97 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.2546, 0.0867, -0.1291, ..., -0.0408, -0.2024, -0.1549], + [ 0.0144, 0.1009, -0.1120, ..., -0.1101, -0.0461, 0.1071], + [ 0.0527, -0.1195, -0.1347, ..., 0.0038, -0.0507, -0.0868], + ..., + [ 0.0823, -0.0474, 0.1242, ..., 0.0610, 0.1740, -0.0111], + [ 0.1129, -0.1960, -0.1282, ..., -0.2277, -0.0549, 0.1900], + [-0.1668, 0.0831, 0.0495, ..., -0.2389, -0.1192, -0.0427]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -8.4005e-07, 1.8626e-09, ..., -3.5390e-08, + 0.0000e+00, 7.4506e-09], + [ 6.3330e-08, -6.1467e-08, 3.3528e-08, ..., 3.3528e-08, + 3.1665e-08, -3.5577e-07], + [ 2.8685e-07, 1.0803e-07, 7.4506e-09, ..., -4.6566e-08, + 1.2480e-07, 4.4703e-08], + ..., + [-7.0594e-07, 3.9116e-08, -5.5879e-08, ..., -1.7136e-07, + -2.5518e-07, 5.5879e-08], + [-5.4017e-08, 3.7253e-08, 5.5879e-09, ..., 2.0489e-08, + 3.7253e-09, -6.3330e-08], + [ 1.1176e-08, 5.4203e-07, -2.2352e-08, ..., 3.3528e-08, + 7.4506e-09, 1.1176e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0093, -0.0148, -0.0046, -0.0162, -0.0089, 0.0032, 0.0083, 0.0146, + 0.0172, -0.0113], device='cuda:0'), grad: tensor([-1.8571e-06, -6.2399e-07, 5.7928e-07, 3.3341e-07, 5.5879e-07, + 2.1234e-07, 6.9290e-07, -1.0990e-06, -1.0245e-07, 1.2610e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 214.94, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.5123 re_mapping 0.0049 re_causal 0.0148 /// teacc 98.82 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.2546, 0.0865, -0.1294, ..., -0.0410, -0.2032, -0.1550], + [ 0.0143, 0.1011, -0.1122, ..., -0.1103, -0.0463, 0.1073], + [ 0.0525, -0.1197, -0.1355, ..., 0.0037, -0.0510, -0.0872], + ..., + [ 0.0825, -0.0475, 0.1245, ..., 0.0613, 0.1744, -0.0113], + [ 0.1130, -0.1962, -0.1282, ..., -0.2278, -0.0548, 0.1901], + [-0.1669, 0.0830, 0.0495, ..., -0.2396, -0.1196, -0.0428]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -3.5577e-07, 1.1176e-08, ..., 5.5879e-09, + 5.5879e-09, -7.0781e-08], + [-7.2084e-07, 4.4703e-08, -6.7055e-08, ..., 2.6077e-08, + -9.3132e-08, -1.1455e-06], + [ 3.1665e-08, 1.6764e-08, 1.1362e-07, ..., -6.5193e-08, + 1.8626e-08, 1.4901e-08], + ..., + [ 5.3458e-07, 1.8626e-08, -1.6391e-07, ..., -2.4214e-08, + -1.8626e-08, 1.1306e-06], + [ 3.7253e-09, 2.5518e-07, 7.6368e-08, ..., 7.4506e-09, + 9.3132e-09, 1.8626e-08], + [ 5.5879e-09, 3.7253e-08, -1.2666e-07, ..., 0.0000e+00, + 1.8626e-09, 2.2352e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0095, -0.0147, -0.0045, -0.0162, -0.0089, 0.0032, 0.0080, 0.0147, + 0.0173, -0.0115], device='cuda:0'), grad: tensor([-8.0280e-07, -3.1032e-06, -2.6077e-08, 3.0175e-07, -1.5646e-07, + -7.6368e-08, 2.0117e-07, 2.9989e-06, 7.1153e-07, -5.2154e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 214.94, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4814 re_mapping 0.0051 re_causal 0.0149 /// teacc 98.90 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.2558, 0.0839, -0.1325, ..., -0.0412, -0.2042, -0.1563], + [ 0.0140, 0.1011, -0.1124, ..., -0.1105, -0.0468, 0.1073], + [ 0.0525, -0.1203, -0.1361, ..., 0.0036, -0.0511, -0.0872], + ..., + [ 0.0829, -0.0474, 0.1249, ..., 0.0614, 0.1750, -0.0111], + [ 0.1130, -0.1964, -0.1283, ..., -0.2287, -0.0551, 0.1901], + [-0.1669, 0.0842, 0.0497, ..., -0.2402, -0.1201, -0.0429]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 7.0781e-08, 1.0803e-07, ..., 7.6368e-08, + 5.5879e-09, 1.8626e-09], + [-1.5274e-07, -2.4028e-07, 1.2293e-07, ..., 4.8429e-08, + 8.0094e-08, -1.2759e-06], + [-1.1940e-06, 1.4342e-07, 2.1048e-07, ..., -3.7812e-07, + -7.9162e-07, 2.0489e-08], + ..., + [ 9.5181e-07, 4.2841e-08, -8.5682e-08, ..., 4.3400e-07, + 6.0163e-07, 7.4506e-08], + [ 3.1292e-07, 2.7195e-07, 1.0226e-06, ..., 3.9116e-08, + 4.2841e-08, 1.1288e-06], + [ 7.4506e-09, -2.5891e-07, -1.3076e-06, ..., 1.4901e-08, + 2.9802e-08, 1.4901e-08]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0125, -0.0149, -0.0046, -0.0162, -0.0090, 0.0031, 0.0082, 0.0151, + 0.0171, -0.0109], device='cuda:0'), grad: tensor([ 3.7998e-07, -1.7695e-06, -1.5758e-06, -1.4994e-06, 6.5193e-07, + 3.4459e-07, 6.1467e-08, 2.1327e-06, 4.3549e-06, -3.0864e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 214.88, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4997 re_mapping 0.0049 re_causal 0.0144 /// teacc 98.97 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.2577, 0.0841, -0.1330, ..., -0.0415, -0.2063, -0.1579], + [ 0.0135, 0.1000, -0.1127, ..., -0.1108, -0.0475, 0.1068], + [ 0.0525, -0.1205, -0.1368, ..., 0.0037, -0.0510, -0.0874], + ..., + [ 0.0833, -0.0474, 0.1253, ..., 0.0615, 0.1755, -0.0110], + [ 0.1132, -0.1966, -0.1282, ..., -0.2296, -0.0546, 0.1904], + [-0.1670, 0.0843, 0.0497, ..., -0.2410, -0.1210, -0.0430]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.0617e-07, 2.6077e-08, ..., 1.3039e-08, + 0.0000e+00, 1.8626e-09], + [ 5.0291e-08, -7.6182e-07, 1.8813e-07, ..., 1.1176e-07, + 4.0978e-08, -1.8626e-08], + [-7.2643e-08, 8.3819e-08, 1.5460e-07, ..., 2.7940e-08, + -1.7323e-07, 3.7253e-09], + ..., + [ 9.1270e-08, 2.1979e-07, 1.9930e-07, ..., 1.5274e-07, + 1.2666e-07, 1.3039e-08], + [ 6.3330e-08, 7.7486e-07, 5.5693e-07, ..., 2.9244e-07, + 1.8626e-09, -2.2352e-08], + [ 1.1176e-08, 4.0606e-07, -2.9802e-08, ..., 9.3132e-09, + 1.8626e-09, 1.3039e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0126, -0.0157, -0.0045, -0.0164, -0.0087, 0.0031, 0.0098, 0.0153, + 0.0172, -0.0110], device='cuda:0'), grad: tensor([-2.4959e-07, -1.6894e-06, 3.8929e-07, 1.6298e-06, 1.5274e-07, + -1.3143e-05, 4.2841e-06, 1.3895e-06, 5.9158e-06, 1.3169e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 214.84, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4672 re_mapping 0.0056 re_causal 0.0147 /// teacc 98.90 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.2585, 0.0841, -0.1332, ..., -0.0417, -0.2074, -0.1579], + [ 0.0131, 0.1000, -0.1129, ..., -0.1110, -0.0481, 0.1067], + [ 0.0523, -0.1206, -0.1369, ..., 0.0041, -0.0513, -0.0875], + ..., + [ 0.0840, -0.0481, 0.1256, ..., 0.0618, 0.1764, -0.0109], + [ 0.1132, -0.1967, -0.1283, ..., -0.2309, -0.0548, 0.1906], + [-0.1669, 0.0846, 0.0500, ..., -0.2416, -0.1214, -0.0430]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, 2.1420e-08, 1.2107e-08, ..., 1.5832e-08, + 4.6566e-09, 7.5437e-08], + [-2.6505e-06, -2.4457e-06, 7.5437e-08, ..., 1.3970e-08, + 5.4017e-08, -1.1802e-05], + [ 7.0222e-07, 3.5483e-07, 6.4261e-08, ..., -3.8277e-07, + 3.1665e-08, 6.6273e-06], + ..., + [-8.0466e-07, -1.6857e-07, -4.5635e-06, ..., -7.4506e-09, + -2.0303e-06, 1.4910e-06], + [ 1.6019e-07, 2.9057e-07, 1.6857e-07, ..., 6.0536e-08, + 8.3819e-09, 4.1910e-07], + [ 9.5926e-08, -8.9407e-08, -7.5437e-08, ..., 2.7940e-09, + 4.6566e-08, 1.5367e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0125, -0.0160, -0.0044, -0.0181, -0.0089, 0.0032, 0.0099, 0.0156, + 0.0172, -0.0108], device='cuda:0'), grad: tensor([ 3.6508e-07, -3.3110e-05, 1.5631e-05, 1.6373e-06, 3.0339e-05, + 1.5097e-06, 1.5423e-05, -3.4541e-05, 2.1011e-06, 5.9232e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 214.73, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.5334 re_mapping 0.0056 re_causal 0.0157 /// teacc 99.00 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.2595, 0.0841, -0.1333, ..., -0.0420, -0.2093, -0.1583], + [ 0.0122, 0.0995, -0.1133, ..., -0.1114, -0.0494, 0.1065], + [ 0.0521, -0.1207, -0.1376, ..., 0.0041, -0.0514, -0.0884], + ..., + [ 0.0852, -0.0469, 0.1262, ..., 0.0621, 0.1777, -0.0100], + [ 0.1130, -0.1970, -0.1284, ..., -0.2328, -0.0560, 0.1904], + [-0.1671, 0.0846, 0.0499, ..., -0.2436, -0.1226, -0.0431]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, -5.2154e-08, 3.7253e-08, ..., 2.5146e-08, + 1.8626e-08, 9.3132e-10], + [ 1.5739e-07, 8.3819e-09, 9.4995e-08, ..., 6.4261e-08, + 8.8476e-08, -3.5390e-08], + [ 1.4342e-07, 2.9802e-08, 9.8720e-08, ..., 2.1420e-08, + 5.6811e-08, 5.5879e-09], + ..., + [-6.6962e-07, 1.5832e-08, -3.2783e-07, ..., -1.7881e-07, + -3.5763e-07, 1.4901e-08], + [ 3.2596e-07, 5.2154e-08, 2.2352e-07, ..., 1.7975e-07, + 1.6391e-07, 1.3039e-08], + [ 3.7253e-09, 5.5879e-09, -1.3970e-08, ..., 1.2107e-08, + 1.8626e-09, 1.7695e-08]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0126, -0.0163, -0.0046, -0.0183, -0.0089, 0.0029, 0.0101, 0.0165, + 0.0171, -0.0109], device='cuda:0'), grad: tensor([-1.0245e-07, 1.3039e-06, -4.2282e-06, -1.1930e-06, 2.8405e-07, + 8.5495e-07, 2.8722e-06, -6.2399e-07, 7.2084e-07, 1.2759e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.95, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4583 re_mapping 0.0052 re_causal 0.0144 /// teacc 98.92 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.2603, 0.0838, -0.1335, ..., -0.0422, -0.2100, -0.1589], + [ 0.0121, 0.0991, -0.1134, ..., -0.1115, -0.0495, 0.1068], + [ 0.0521, -0.1209, -0.1381, ..., 0.0041, -0.0515, -0.0892], + ..., + [ 0.0853, -0.0470, 0.1263, ..., 0.0622, 0.1779, -0.0101], + [ 0.1131, -0.1973, -0.1285, ..., -0.2334, -0.0561, 0.1905], + [-0.1671, 0.0847, 0.0500, ..., -0.2442, -0.1230, -0.0431]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -4.1910e-08, 2.7940e-09, ..., 9.3132e-10, + 3.7253e-09, 9.3132e-10], + [ 5.4389e-07, 3.1758e-07, 5.3365e-07, ..., 9.3132e-09, + 4.1258e-07, -3.9767e-07], + [-2.4840e-05, 7.4506e-09, -4.1686e-06, ..., -5.4501e-06, + -1.4797e-05, 4.6566e-09], + ..., + [ 2.3857e-05, -5.9512e-07, 3.1609e-06, ..., 5.4426e-06, + 1.4037e-05, 2.0489e-08], + [ 7.4506e-09, 1.7695e-08, 5.5879e-09, ..., 1.8626e-09, + 5.5879e-09, 9.7789e-08], + [ 4.5635e-07, 5.1409e-07, 4.6752e-07, ..., 2.7940e-09, + 3.5111e-07, 1.1176e-08]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0130, -0.0163, -0.0047, -0.0185, -0.0087, 0.0033, 0.0100, 0.0165, + 0.0170, -0.0109], device='cuda:0'), grad: tensor([-6.9849e-08, 5.6811e-07, -2.9609e-05, 1.9372e-07, -9.8720e-08, + -9.5926e-07, 4.5355e-07, 2.7552e-05, 2.1234e-07, 1.7537e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 215.19, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4908 re_mapping 0.0052 re_causal 0.0149 /// teacc 98.97 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.2610, 0.0834, -0.1337, ..., -0.0423, -0.2104, -0.1600], + [ 0.0112, 0.0991, -0.1140, ..., -0.1116, -0.0511, 0.1063], + [ 0.0522, -0.1210, -0.1385, ..., 0.0041, -0.0515, -0.0893], + ..., + [ 0.0860, -0.0471, 0.1267, ..., 0.0621, 0.1791, -0.0096], + [ 0.1135, -0.1975, -0.1285, ..., -0.2342, -0.0560, 0.1920], + [-0.1670, 0.0845, 0.0502, ..., -0.2447, -0.1222, -0.0433]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.3970e-08, 2.4214e-08, ..., 9.3132e-09, + 1.8626e-09, -0.0000e+00], + [ 3.7253e-08, 1.4435e-07, 3.6880e-07, ..., 7.8231e-08, + 3.4459e-08, 2.7940e-09], + [-7.4506e-09, 8.7544e-08, 2.8778e-07, ..., 1.8626e-07, + -4.6566e-09, 1.8626e-09], + ..., + [-5.3085e-08, 1.6484e-07, 4.4052e-07, ..., 7.8231e-08, + -5.4017e-08, 0.0000e+00], + [-7.4506e-09, 1.2293e-07, 3.7625e-07, ..., 7.5437e-08, + 9.3132e-09, -2.1420e-08], + [ 1.0245e-08, -2.1476e-06, -3.0641e-06, ..., 7.4506e-08, + 5.5879e-09, 6.5193e-09]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0133, -0.0169, -0.0048, -0.0181, -0.0085, 0.0031, 0.0090, 0.0169, + 0.0176, -0.0109], device='cuda:0'), grad: tensor([ 2.4214e-08, 9.2015e-07, 3.8277e-07, -1.8328e-06, 5.8953e-07, + 7.6517e-06, -2.3283e-08, 9.7230e-07, 8.3540e-07, -9.5516e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 215.09, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4571 re_mapping 0.0053 re_causal 0.0140 /// teacc 98.90 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.2621, 0.0835, -0.1341, ..., -0.0424, -0.2117, -0.1604], + [ 0.0104, 0.0990, -0.1147, ..., -0.1117, -0.0523, 0.1060], + [ 0.0522, -0.1211, -0.1392, ..., 0.0040, -0.0517, -0.0894], + ..., + [ 0.0869, -0.0472, 0.1277, ..., 0.0623, 0.1805, -0.0093], + [ 0.1135, -0.1977, -0.1286, ..., -0.2360, -0.0563, 0.1921], + [-0.1673, 0.0845, 0.0500, ..., -0.2462, -0.1235, -0.0434]], + device='cuda:0'), grad: tensor([[ 3.7067e-07, 7.4506e-09, 5.8673e-08, ..., 2.9709e-07, + 1.8161e-07, 1.6764e-08], + [ 2.5705e-07, 3.0734e-08, 4.6380e-07, ..., 4.0792e-07, + 8.3819e-08, -1.9558e-08], + [-1.9595e-06, 2.6077e-08, 1.2293e-07, ..., -1.5739e-06, + -1.0161e-06, 2.2352e-08], + ..., + [ 9.5889e-06, 5.0664e-07, 2.5123e-05, ..., 2.3052e-05, + 3.2596e-07, 2.1793e-07], + [ 2.0601e-06, 3.6880e-07, 4.2766e-06, ..., 4.1537e-06, + 3.5856e-07, -1.1642e-07], + [-1.9558e-08, -2.0508e-06, -1.5963e-06, ..., 1.0524e-07, + 1.3039e-08, 5.6811e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0132, -0.0175, -0.0047, -0.0180, -0.0083, 0.0029, 0.0088, 0.0178, + 0.0175, -0.0111], device='cuda:0'), grad: tensor([ 1.4696e-06, 1.0477e-06, -6.4857e-06, -4.7654e-05, 2.4457e-06, + 1.3690e-06, 2.5891e-07, 4.7415e-05, 1.0625e-05, -1.0520e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 215.03, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4985 re_mapping 0.0050 re_causal 0.0147 /// teacc 98.91 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.2635, 0.0837, -0.1347, ..., -0.0428, -0.2139, -0.1610], + [ 0.0105, 0.0990, -0.1146, ..., -0.1117, -0.0523, 0.1069], + [ 0.0522, -0.1213, -0.1398, ..., 0.0041, -0.0516, -0.0914], + ..., + [ 0.0869, -0.0474, 0.1277, ..., 0.0622, 0.1805, -0.0096], + [ 0.1135, -0.1979, -0.1286, ..., -0.2373, -0.0565, 0.1922], + [-0.1673, 0.0844, 0.0501, ..., -0.2467, -0.1237, -0.0434]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.3190e-07, 3.7253e-09, ..., -2.8871e-08, + 9.3132e-10, 2.3283e-08], + [-9.2201e-08, -8.1025e-08, 2.9802e-08, ..., 1.0245e-08, + 2.4214e-08, -2.5611e-07], + [ 5.5879e-09, 1.3970e-08, 6.5193e-09, ..., -1.8626e-09, + -3.7253e-09, 1.7695e-08], + ..., + [ 2.1420e-08, 6.4261e-08, -5.5879e-08, ..., -1.6764e-08, + -4.8429e-08, 1.7881e-07], + [ 4.6566e-09, 6.5193e-08, 1.0245e-08, ..., 2.7940e-09, + 2.7940e-09, 1.2880e-06], + [ 4.2841e-08, 1.0338e-07, -2.7008e-08, ..., 2.1420e-08, + 2.0489e-08, 4.1910e-08]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0132, -0.0172, -0.0050, -0.0181, -0.0079, 0.0027, 0.0088, 0.0176, + 0.0175, -0.0111], device='cuda:0'), grad: tensor([-2.6356e-07, -6.1467e-07, 6.6124e-08, 3.0082e-07, 1.5181e-07, + 1.3644e-06, -1.2308e-05, 4.2189e-07, 1.0476e-05, 3.7905e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 214.96, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4749 re_mapping 0.0049 re_causal 0.0135 /// teacc 99.09 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.2646, 0.0839, -0.1349, ..., -0.0446, -0.2146, -0.1628], + [ 0.0090, 0.0990, -0.1155, ..., -0.1118, -0.0542, 0.1064], + [ 0.0521, -0.1214, -0.1403, ..., 0.0041, -0.0519, -0.0918], + ..., + [ 0.0883, -0.0475, 0.1286, ..., 0.0622, 0.1824, -0.0092], + [ 0.1136, -0.1980, -0.1287, ..., -0.2381, -0.0566, 0.1926], + [-0.1673, 0.0844, 0.0502, ..., -0.2473, -0.1240, -0.0435]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, -2.0210e-07, 9.3132e-10, ..., 2.7008e-08, + 5.1223e-08, 2.7940e-09], + [ 3.4459e-08, 9.6858e-08, 7.4506e-08, ..., 6.2399e-08, + 4.0047e-08, -2.2072e-07], + [-3.9116e-07, 3.5390e-08, 2.8871e-08, ..., -3.5670e-07, + -6.0908e-07, 2.4214e-08], + ..., + [ 3.0920e-07, 3.4459e-08, 3.4459e-08, ..., 3.4552e-07, + 4.6007e-07, 1.8440e-07], + [-3.9116e-08, 3.5390e-08, 7.3574e-08, ..., 6.5193e-08, + 3.0734e-08, -1.9558e-08], + [ 5.7742e-08, 7.6462e-07, 4.9081e-07, ..., 9.4995e-08, + 1.2107e-08, 2.0210e-07]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0134, -0.0186, -0.0052, -0.0179, -0.0077, 0.0025, 0.0091, 0.0190, + 0.0176, -0.0112], device='cuda:0'), grad: tensor([-2.8498e-07, 1.0990e-07, -1.3607e-06, -1.9707e-06, -2.0787e-06, + 5.0850e-07, 6.3423e-07, 1.5935e-06, 2.4121e-07, 2.6096e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.88, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.5234 re_mapping 0.0045 re_causal 0.0139 /// teacc 98.89 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.2652, 0.0843, -0.1350, ..., -0.0446, -0.2157, -0.1630], + [ 0.0089, 0.0989, -0.1155, ..., -0.1119, -0.0542, 0.1066], + [ 0.0521, -0.1216, -0.1410, ..., 0.0041, -0.0519, -0.0921], + ..., + [ 0.0884, -0.0477, 0.1287, ..., 0.0623, 0.1825, -0.0093], + [ 0.1136, -0.1982, -0.1287, ..., -0.2402, -0.0568, 0.1927], + [-0.1674, 0.0842, 0.0502, ..., -0.2483, -0.1245, -0.0436]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -1.4715e-07, 9.3132e-10, ..., 4.6566e-09, + 4.6566e-09, 0.0000e+00], + [ 1.3970e-07, 2.1420e-08, 6.0536e-08, ..., 6.2399e-08, + 7.5437e-08, -4.2841e-08], + [-5.6997e-07, 1.4901e-08, 8.3819e-09, ..., -3.0454e-07, + -3.0641e-07, 8.3819e-09], + ..., + [-3.3900e-07, 1.3970e-08, -3.4273e-07, ..., -1.1455e-07, + -1.9465e-07, 2.5146e-08], + [ 6.2212e-07, 4.1537e-07, 2.7474e-07, ..., 2.4680e-07, + 2.9709e-07, -1.7695e-08], + [ 1.1828e-07, 1.1194e-06, 5.0291e-08, ..., 1.1176e-08, + 1.3039e-08, 7.4506e-09]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0129, -0.0186, -0.0051, -0.0181, -0.0076, 0.0025, 0.0090, 0.0190, + 0.0175, -0.0113], device='cuda:0'), grad: tensor([-2.4401e-07, 4.1071e-07, -2.4252e-06, 1.2279e-05, 2.2631e-07, + -1.7405e-05, 6.2678e-07, -2.3097e-07, 3.0808e-06, 3.6638e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 215.05, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.4772 re_mapping 0.0049 re_causal 0.0141 /// teacc 98.94 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.2666, 0.0845, -0.1351, ..., -0.0447, -0.2163, -0.1638], + [ 0.0095, 0.0989, -0.1156, ..., -0.1120, -0.0538, 0.1077], + [ 0.0520, -0.1217, -0.1417, ..., 0.0040, -0.0521, -0.0924], + ..., + [ 0.0878, -0.0479, 0.1283, ..., 0.0625, 0.1822, -0.0104], + [ 0.1137, -0.1982, -0.1288, ..., -0.2410, -0.0575, 0.1934], + [-0.1673, 0.0841, 0.0513, ..., -0.2496, -0.1231, -0.0437]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -6.8508e-06, -1.0245e-08, ..., 1.8626e-09, + 3.7253e-09, 9.3132e-09], + [ 5.0291e-08, -6.3330e-08, 4.1910e-08, ..., 6.5193e-08, + 1.6671e-07, -1.4808e-07], + [-7.0781e-08, 2.8871e-08, -2.9802e-08, ..., -8.1956e-08, + -2.2165e-07, -1.2945e-07], + ..., + [ 3.7253e-09, 9.2201e-08, -3.0734e-08, ..., -6.5193e-09, + 2.7940e-09, 1.9930e-07], + [-2.1420e-08, 4.7497e-08, 4.6566e-09, ..., 3.7253e-09, + 8.3819e-09, -3.9116e-08], + [ 1.2107e-08, 2.3097e-07, 9.3132e-09, ..., 5.5879e-09, + 1.2107e-08, 5.1223e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0129, -0.0179, -0.0053, -0.0185, -0.0074, 0.0023, 0.0073, 0.0176, + 0.0184, -0.0105], device='cuda:0'), grad: tensor([-1.5527e-05, -7.4506e-09, -1.2759e-06, 2.6356e-06, 5.4017e-08, + 1.0483e-05, 1.8002e-06, 9.6392e-07, 1.7975e-07, 7.0222e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 214.69, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.5125 re_mapping 0.0049 re_causal 0.0141 /// teacc 98.98 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.2689, 0.0844, -0.1355, ..., -0.0449, -0.2190, -0.1659], + [ 0.0089, 0.0989, -0.1162, ..., -0.1129, -0.0550, 0.1078], + [ 0.0518, -0.1219, -0.1429, ..., 0.0039, -0.0524, -0.0926], + ..., + [ 0.0888, -0.0478, 0.1295, ..., 0.0640, 0.1843, -0.0105], + [ 0.1137, -0.1987, -0.1289, ..., -0.2438, -0.0587, 0.1936], + [-0.1676, 0.0842, 0.0507, ..., -0.2585, -0.1279, -0.0439]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.8626e-09, 1.0245e-08, ..., 5.5879e-09, + 2.7940e-09, 1.8626e-09], + [-1.1921e-07, 9.3132e-09, 6.5193e-08, ..., 4.3772e-08, + 5.4948e-08, -7.1619e-07], + [ 4.2096e-07, 1.5832e-08, 2.4494e-07, ..., 2.3749e-07, + 3.5018e-07, 8.1025e-08], + ..., + [-5.8208e-07, 9.3132e-09, -3.8929e-07, ..., -3.5018e-07, + -5.2713e-07, 7.0781e-08], + [ 1.7695e-07, 4.6566e-08, 8.9407e-08, ..., 4.2841e-08, + 5.5879e-08, 4.6287e-07], + [ 7.1712e-08, -7.5903e-07, -2.8033e-07, ..., 4.1910e-08, + 5.5879e-08, 2.4214e-08]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0133, -0.0185, -0.0054, -0.0162, -0.0073, 0.0004, 0.0075, 0.0186, + 0.0183, -0.0109], device='cuda:0'), grad: tensor([ 8.2888e-08, -1.5749e-06, 8.9314e-07, -4.7404e-07, 1.6000e-06, + 1.0347e-06, 2.1979e-07, -7.8045e-07, 1.5115e-06, -2.5053e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.74, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5120 re_mapping 0.0049 re_causal 0.0144 /// teacc 98.98 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.2696, 0.0839, -0.1358, ..., -0.0451, -0.2201, -0.1664], + [ 0.0089, 0.0982, -0.1164, ..., -0.1131, -0.0551, 0.1083], + [ 0.0518, -0.1220, -0.1433, ..., 0.0039, -0.0524, -0.0929], + ..., + [ 0.0890, -0.0480, 0.1299, ..., 0.0644, 0.1847, -0.0109], + [ 0.1136, -0.1989, -0.1290, ..., -0.2451, -0.0595, 0.1937], + [-0.1678, 0.0849, 0.0507, ..., -0.2595, -0.1289, -0.0439]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -3.7253e-09, 9.3132e-09, ..., 6.5193e-09, + 4.6566e-09, 1.8626e-09], + [ 1.0338e-07, -1.8626e-09, 1.4529e-07, ..., 9.5926e-08, + 9.0338e-08, -4.0047e-08], + [ 9.3132e-10, 2.7940e-09, 5.4017e-08, ..., -1.6764e-08, + 4.5635e-08, 5.5879e-09], + ..., + [-6.7428e-07, 6.5193e-09, -9.7696e-07, ..., -4.9081e-07, + -6.6962e-07, 1.9558e-08], + [ 2.3656e-07, 1.3970e-08, 2.8498e-07, ..., 1.3411e-07, + 1.8254e-07, 5.5879e-09], + [ 4.0978e-08, 9.3225e-07, 3.3528e-08, ..., 4.0047e-08, + 3.9116e-08, 7.8510e-07]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0139, -0.0183, -0.0055, -0.0166, -0.0072, 0.0004, 0.0074, 0.0186, + 0.0182, -0.0106], device='cuda:0'), grad: tensor([ 1.2293e-07, 2.7288e-07, 2.4214e-08, 6.0070e-07, -3.4682e-06, + 1.9092e-07, -7.5810e-07, -1.4128e-06, 6.0536e-07, 3.8110e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.94, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.4898 re_mapping 0.0051 re_causal 0.0136 /// teacc 99.02 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.2702, 0.0839, -0.1367, ..., -0.0458, -0.2222, -0.1670], + [ 0.0088, 0.0979, -0.1166, ..., -0.1133, -0.0552, 0.1087], + [ 0.0520, -0.1222, -0.1438, ..., 0.0040, -0.0523, -0.0931], + ..., + [ 0.0892, -0.0481, 0.1303, ..., 0.0645, 0.1851, -0.0110], + [ 0.1136, -0.1998, -0.1292, ..., -0.2475, -0.0603, 0.1935], + [-0.1679, 0.0852, 0.0508, ..., -0.2600, -0.1297, -0.0436]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.7940e-09, 2.7940e-09, ..., 9.3132e-10, + 1.8626e-09, 1.8626e-09], + [ 1.1362e-07, -3.2596e-08, 1.1176e-07, ..., 6.8918e-08, + 1.4901e-07, -9.4157e-07], + [-1.7695e-07, 1.5832e-08, 1.6578e-07, ..., 1.7229e-07, + -3.3341e-07, 5.7835e-07], + ..., + [-1.3970e-07, 4.3586e-07, 1.8552e-06, ..., -5.5879e-09, + 1.5460e-07, 1.8626e-08], + [ 1.4808e-07, 5.5879e-09, 1.2480e-07, ..., 3.7253e-08, + 1.9372e-07, 1.4901e-08], + [ 3.1665e-08, -4.6566e-07, -2.3153e-06, ..., 5.5879e-09, + -2.1048e-07, 4.0978e-08]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0142, -0.0184, -0.0056, -0.0165, -0.0074, 0.0009, 0.0072, 0.0188, + 0.0179, -0.0105], device='cuda:0'), grad: tensor([ 6.0536e-08, -7.2680e-06, 4.7013e-06, -5.4855e-07, 2.8405e-07, + 3.1665e-08, 2.2762e-06, 2.5686e-06, 4.9639e-07, -2.6412e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 284---------------------------------------------------- +epoch 284, time 231.45, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4645 re_mapping 0.0048 re_causal 0.0137 /// teacc 99.17 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.2708, 0.0840, -0.1369, ..., -0.0461, -0.2231, -0.1672], + [ 0.0088, 0.0980, -0.1160, ..., -0.1134, -0.0553, 0.1088], + [ 0.0521, -0.1222, -0.1442, ..., 0.0041, -0.0522, -0.0932], + ..., + [ 0.0893, -0.0483, 0.1300, ..., 0.0646, 0.1853, -0.0110], + [ 0.1136, -0.2000, -0.1294, ..., -0.2492, -0.0611, 0.1935], + [-0.1681, 0.0848, 0.0506, ..., -0.2604, -0.1303, -0.0439]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.0419e-07, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 5.1223e-08, 6.5193e-09, ..., 4.6566e-09, + 4.6566e-09, -7.4506e-09], + [-5.8673e-08, 1.9558e-08, 4.6566e-09, ..., -4.1910e-08, + -5.0291e-08, 2.7940e-09], + ..., + [ 1.4901e-08, 2.6077e-08, -7.4506e-09, ..., 1.1176e-08, + 1.1176e-08, 5.5879e-09], + [ 2.2352e-08, 8.1956e-08, 1.8626e-09, ..., 1.8626e-08, + 2.2352e-08, -9.3132e-10], + [ 3.7253e-09, -1.8626e-09, -7.9162e-08, ..., 1.8626e-09, + 2.7940e-09, 3.2596e-08]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0140, -0.0178, -0.0055, -0.0165, -0.0070, 0.0008, 0.0073, 0.0184, + 0.0177, -0.0108], device='cuda:0'), grad: tensor([-1.0822e-06, 1.5087e-07, -2.1420e-07, 6.3330e-08, 1.1828e-07, + 2.9430e-07, 1.3970e-07, 1.5274e-07, 3.3341e-07, 6.2399e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 214.98, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4962 re_mapping 0.0049 re_causal 0.0145 /// teacc 99.04 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.2714, 0.0844, -0.1373, ..., -0.0462, -0.2240, -0.1673], + [ 0.0088, 0.0981, -0.1158, ..., -0.1134, -0.0553, 0.1090], + [ 0.0528, -0.1223, -0.1441, ..., 0.0047, -0.0512, -0.0933], + ..., + [ 0.0891, -0.0485, 0.1299, ..., 0.0642, 0.1849, -0.0111], + [ 0.1136, -0.2006, -0.1294, ..., -0.2500, -0.0613, 0.1936], + [-0.1681, 0.0850, 0.0507, ..., -0.2606, -0.1306, -0.0439]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 3.7253e-09, -7.4506e-09, 1.9558e-08, ..., 1.9558e-08, + 8.3819e-09, -3.2596e-08], + [-1.8626e-08, 9.3132e-10, 7.6648e-07, ..., 1.1437e-06, + -1.7695e-08, 3.7253e-09], + ..., + [ 6.5193e-09, 3.7253e-09, -2.1420e-08, ..., 3.7253e-09, + 2.7940e-09, 7.3574e-08], + [ 1.8626e-09, 1.8626e-09, 5.5879e-09, ..., 3.7253e-09, + 9.3132e-10, 7.4506e-09], + [ 2.7940e-09, -4.6566e-09, -1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 9.3132e-09]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0144, -0.0176, -0.0050, -0.0171, -0.0073, 0.0011, 0.0076, 0.0181, + 0.0177, -0.0107], device='cuda:0'), grad: tensor([ 2.0489e-08, 0.0000e+00, 2.5053e-06, -2.6897e-06, -2.9430e-07, + 1.2480e-07, -1.4901e-08, 2.7195e-07, 3.8184e-08, 3.1665e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 215.06, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4806 re_mapping 0.0050 re_causal 0.0139 /// teacc 98.98 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.2723, 0.0847, -0.1374, ..., -0.0462, -0.2245, -0.1676], + [ 0.0088, 0.0980, -0.1159, ..., -0.1136, -0.0553, 0.1091], + [ 0.0524, -0.1225, -0.1458, ..., 0.0043, -0.0516, -0.0935], + ..., + [ 0.0893, -0.0486, 0.1302, ..., 0.0647, 0.1852, -0.0111], + [ 0.1138, -0.2008, -0.1294, ..., -0.2506, -0.0610, 0.1940], + [-0.1682, 0.0843, 0.0508, ..., -0.2610, -0.1308, -0.0443]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.1176e-08, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-09, 3.9078e-06, 1.8626e-09, ..., 9.3132e-10, + 1.8626e-09, -1.2107e-08], + [-9.3132e-09, 3.8184e-08, 8.3819e-09, ..., -2.1420e-08, + -1.8626e-09, 4.6566e-09], + ..., + [-4.6566e-09, 1.8626e-08, -1.4901e-08, ..., 8.3819e-09, + -1.0245e-08, 1.0245e-08], + [ 1.3039e-08, 1.4696e-06, 1.8626e-09, ..., 2.7940e-09, + 1.8626e-09, 3.7253e-09], + [ 6.7055e-08, 2.1830e-06, -6.5193e-09, ..., 9.3132e-10, + 1.8626e-09, 1.1362e-07]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0147, -0.0176, -0.0055, -0.0170, -0.0072, 0.0013, 0.0076, 0.0182, + 0.0178, -0.0109], device='cuda:0'), grad: tensor([ 1.1828e-07, 4.8697e-05, 4.9546e-07, 4.4525e-05, -3.6322e-08, + -1.1933e-04, 1.3039e-07, 1.0617e-07, 1.7866e-05, 7.4692e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 214.77, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4833 re_mapping 0.0049 re_causal 0.0142 /// teacc 99.03 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.2750, 0.0857, -0.1379, ..., -0.0468, -0.2275, -0.1658], + [ 0.0085, 0.0978, -0.1161, ..., -0.1136, -0.0559, 0.1094], + [ 0.0525, -0.1227, -0.1466, ..., 0.0045, -0.0515, -0.0948], + ..., + [ 0.0896, -0.0488, 0.1305, ..., 0.0647, 0.1857, -0.0110], + [ 0.1138, -0.2012, -0.1295, ..., -0.2515, -0.0612, 0.1941], + [-0.1677, 0.0848, 0.0516, ..., -0.2612, -0.1309, -0.0445]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 4.6566e-09, 6.5193e-09, ..., 5.5879e-09, + 3.7253e-09, 0.0000e+00], + [ 5.5879e-08, -8.3819e-09, 9.2201e-08, ..., 2.5146e-08, + 9.0338e-08, -3.4459e-08], + [-1.1176e-08, 3.7253e-09, -9.4064e-08, ..., -1.0245e-08, + -7.4506e-08, 2.7940e-09], + ..., + [-9.4064e-08, 1.3970e-08, -7.5437e-08, ..., -2.3283e-08, + -1.0058e-07, 2.4214e-08], + [ 5.6811e-08, 4.5635e-08, 1.3039e-08, ..., 7.4506e-09, + 1.0245e-08, 9.3132e-10], + [ 3.7253e-08, -2.7940e-09, 4.9360e-08, ..., 1.0245e-08, + 5.0291e-08, 1.8626e-09]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0136, -0.0178, -0.0055, -0.0171, -0.0082, 0.0019, 0.0074, 0.0183, + 0.0177, -0.0105], device='cuda:0'), grad: tensor([ 8.2888e-08, 3.1572e-07, -5.2806e-07, 5.5879e-09, 6.6124e-08, + -2.2873e-06, 1.1595e-06, 4.4703e-08, 9.4902e-07, 1.9372e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 215.26, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4624 re_mapping 0.0048 re_causal 0.0139 /// teacc 98.97 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.2767, 0.0827, -0.1383, ..., -0.0471, -0.2295, -0.1660], + [ 0.0083, 0.0979, -0.1162, ..., -0.1138, -0.0560, 0.1096], + [ 0.0523, -0.1229, -0.1477, ..., 0.0044, -0.0518, -0.0950], + ..., + [ 0.0901, -0.0489, 0.1308, ..., 0.0650, 0.1862, -0.0110], + [ 0.1138, -0.2015, -0.1296, ..., -0.2538, -0.0616, 0.1941], + [-0.1677, 0.0877, 0.0516, ..., -0.2614, -0.1311, -0.0447]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.2387e-07, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 8.3819e-09, 3.9116e-08, 5.5879e-08, ..., 3.7253e-09, + 8.3819e-09, -8.5682e-08], + [ 3.8184e-08, 1.8626e-09, 2.7008e-08, ..., 1.8626e-08, + 4.2841e-08, 1.3039e-08], + ..., + [-4.2841e-08, 6.4261e-08, -2.7940e-08, ..., -2.5146e-08, + -5.7742e-08, 9.1270e-08], + [-1.9558e-08, 1.8626e-08, 1.1176e-08, ..., 9.3132e-10, + 1.8626e-09, -2.4214e-08], + [ 3.7253e-09, -1.5283e-06, -1.0794e-06, ..., 9.3132e-10, + 3.7253e-09, 1.2107e-08]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0165, -0.0178, -0.0057, -0.0171, -0.0086, 0.0021, 0.0071, 0.0188, + 0.0176, -0.0080], device='cuda:0'), grad: tensor([-2.3004e-07, 9.1270e-08, 9.8720e-08, 1.7695e-08, 4.0792e-06, + 9.2201e-08, -6.9942e-07, 2.0489e-07, 2.1420e-08, -3.6657e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 215.28, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4918 re_mapping 0.0049 re_causal 0.0147 /// teacc 98.99 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.2769, 0.0828, -0.1387, ..., -0.0473, -0.2300, -0.1661], + [ 0.0083, 0.0979, -0.1162, ..., -0.1139, -0.0561, 0.1098], + [ 0.0524, -0.1230, -0.1481, ..., 0.0044, -0.0518, -0.0950], + ..., + [ 0.0901, -0.0490, 0.1309, ..., 0.0650, 0.1863, -0.0111], + [ 0.1138, -0.2018, -0.1296, ..., -0.2546, -0.0618, 0.1942], + [-0.1677, 0.0875, 0.0516, ..., -0.2618, -0.1312, -0.0448]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 4.4703e-08, 1.1176e-08, ..., 1.8626e-09, + 4.6566e-09, 1.8626e-09], + [ 4.7497e-08, 7.5437e-08, 9.7789e-08, ..., 2.5146e-08, + 6.8918e-08, -1.0524e-07], + [ 2.9802e-08, 3.4459e-08, 4.7497e-08, ..., 1.3970e-08, + 3.2596e-08, 6.0536e-08], + ..., + [-2.5146e-07, 4.0978e-08, -5.3085e-07, ..., -1.3597e-07, + -3.8836e-07, 1.1455e-07], + [ 8.3819e-09, 2.7213e-06, 2.7940e-08, ..., 4.6566e-09, + 1.1176e-08, 3.7253e-08], + [ 1.6671e-07, 2.6077e-08, 2.4121e-07, ..., 7.5437e-08, + 2.4773e-07, 9.0338e-08]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0165, -0.0178, -0.0057, -0.0171, -0.0084, 0.0015, 0.0079, 0.0188, + 0.0176, -0.0082], device='cuda:0'), grad: tensor([ 2.0675e-07, 2.2445e-07, 2.0396e-07, 2.2501e-06, -2.3283e-07, + -4.7833e-05, 3.3319e-05, -2.8312e-07, 1.1653e-05, 4.4610e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 215.08, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5037 re_mapping 0.0049 re_causal 0.0141 /// teacc 98.97 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.2767, 0.0829, -0.1396, ..., -0.0481, -0.2307, -0.1658], + [ 0.0083, 0.0978, -0.1157, ..., -0.1137, -0.0561, 0.1103], + [ 0.0527, -0.1231, -0.1483, ..., 0.0046, -0.0514, -0.0951], + ..., + [ 0.0901, -0.0493, 0.1305, ..., 0.0648, 0.1862, -0.0115], + [ 0.1138, -0.2040, -0.1297, ..., -0.2565, -0.0621, 0.1940], + [-0.1678, 0.0874, 0.0516, ..., -0.2622, -0.1314, -0.0451]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.5670e-07, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 1.6764e-08, 2.0489e-08, ..., 1.3970e-08, + 4.6566e-09, -1.3039e-08], + [ 2.7940e-08, 1.0245e-08, 2.9802e-08, ..., 2.9802e-08, + 3.2596e-08, 1.2107e-08], + ..., + [-3.3528e-08, 3.8184e-08, -4.6566e-09, ..., -1.3039e-08, + -4.0978e-08, 2.8871e-08], + [-5.0291e-08, 2.2352e-08, 1.6764e-08, ..., 4.6566e-09, + 9.3132e-10, -1.1828e-07], + [ 4.0047e-08, 1.9558e-07, -7.9162e-08, ..., 1.8626e-09, + 9.3132e-10, 2.5611e-07]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0164, -0.0172, -0.0056, -0.0174, -0.0082, 0.0021, 0.0082, 0.0182, + 0.0173, -0.0083], device='cuda:0'), grad: tensor([-7.0222e-07, 9.0338e-08, 1.0524e-07, -6.2399e-08, -4.7870e-07, + 4.1537e-07, -9.5926e-08, 1.1176e-07, -1.5367e-07, 7.7393e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 215.22, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4478 re_mapping 0.0048 re_causal 0.0134 /// teacc 99.02 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.2779, 0.0829, -0.1411, ..., -0.0484, -0.2320, -0.1669], + [ 0.0082, 0.0976, -0.1158, ..., -0.1139, -0.0563, 0.1107], + [ 0.0531, -0.1233, -0.1492, ..., 0.0051, -0.0510, -0.0956], + ..., + [ 0.0901, -0.0495, 0.1307, ..., 0.0645, 0.1864, -0.0116], + [ 0.1138, -0.2044, -0.1298, ..., -0.2578, -0.0625, 0.1942], + [-0.1678, 0.0880, 0.0523, ..., -0.2626, -0.1315, -0.0438]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -4.9360e-08, 9.3132e-10, ..., 2.7940e-09, + 0.0000e+00, 9.3132e-10], + [ 4.2841e-08, -6.5193e-09, 5.7742e-08, ..., 2.7940e-08, + 3.5390e-08, -1.2107e-08], + [ 9.3132e-09, 9.3132e-09, 2.8871e-08, ..., -2.3283e-08, + 1.3970e-08, 6.5193e-09], + ..., + [-6.7987e-08, 8.3819e-09, -1.0431e-07, ..., -3.7253e-08, + -6.8918e-08, 2.2352e-08], + [-4.0047e-08, 1.1176e-08, 4.6566e-09, ..., 8.3819e-09, + 2.7940e-09, -1.3504e-07], + [ 1.4901e-08, 2.5146e-08, -1.4901e-08, ..., 1.0245e-08, + 1.1176e-08, 2.7940e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0164, -0.0172, -0.0057, -0.0171, -0.0099, 0.0020, 0.0085, 0.0181, + 0.0172, -0.0075], device='cuda:0'), grad: tensor([-4.1910e-08, 1.3132e-07, 4.6566e-09, 2.3283e-08, 2.0768e-07, + 2.1514e-07, -2.5146e-06, -8.4750e-08, 1.9800e-06, 8.8476e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 214.97, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4983 re_mapping 0.0048 re_causal 0.0142 /// teacc 99.02 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.2811, 0.0829, -0.1424, ..., -0.0491, -0.2358, -0.1695], + [ 0.0080, 0.0983, -0.1154, ..., -0.1138, -0.0570, 0.1112], + [ 0.0532, -0.1235, -0.1500, ..., 0.0054, -0.0509, -0.0954], + ..., + [ 0.0906, -0.0495, 0.1312, ..., 0.0655, 0.1877, -0.0117], + [ 0.1138, -0.2049, -0.1299, ..., -0.2602, -0.0637, 0.1941], + [-0.1680, 0.0880, 0.0524, ..., -0.2632, -0.1320, -0.0437]], + device='cuda:0'), grad: tensor([[ 3.4925e-08, 4.6566e-10, 6.5193e-08, ..., 5.1223e-09, + 4.7497e-08, 4.6566e-10], + [ 3.4273e-07, -3.3528e-08, 4.8848e-07, ..., 7.0781e-08, + 3.5670e-07, 1.9092e-08], + [ 1.5274e-07, 9.3132e-10, 2.8778e-07, ..., -2.9337e-08, + 2.0955e-07, 7.9162e-09], + ..., + [-3.4943e-06, 1.7695e-08, -6.2138e-06, ..., -1.7462e-07, + -4.5821e-06, -2.2119e-07], + [ 1.9418e-07, 2.3749e-08, 2.8033e-07, ..., 5.9139e-08, + 2.0536e-07, 8.9407e-08], + [ 2.0247e-06, 1.0664e-07, 3.7551e-06, ..., 4.5635e-08, + 2.7716e-06, 6.8452e-08]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0165, -0.0168, -0.0052, -0.0191, -0.0101, 0.0019, 0.0085, 0.0185, + 0.0171, -0.0075], device='cuda:0'), grad: tensor([ 1.4575e-07, 5.5134e-07, 3.0547e-07, 7.6089e-07, 1.5320e-07, + 1.2163e-06, -1.9278e-07, -1.0133e-05, 6.7940e-07, 6.5342e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 215.12, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4852 re_mapping 0.0050 re_causal 0.0142 /// teacc 99.02 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.2817, 0.0830, -0.1428, ..., -0.0494, -0.2363, -0.1689], + [ 0.0078, 0.0985, -0.1157, ..., -0.1141, -0.0574, 0.1113], + [ 0.0532, -0.1238, -0.1507, ..., 0.0054, -0.0509, -0.0957], + ..., + [ 0.0910, -0.0497, 0.1318, ..., 0.0659, 0.1883, -0.0116], + [ 0.1137, -0.2053, -0.1300, ..., -0.2627, -0.0643, 0.1941], + [-0.1681, 0.0879, 0.0523, ..., -0.2643, -0.1333, -0.0438]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -5.6811e-08, 3.7253e-09, ..., 1.8626e-09, + 3.7253e-09, 1.3970e-09], + [ 4.8894e-08, 4.9360e-08, 5.7742e-08, ..., 3.0734e-08, + 4.7963e-08, 1.9092e-08], + [ 1.0291e-07, 1.0245e-08, 1.5320e-07, ..., 7.3109e-08, + 8.8010e-08, 6.0536e-09], + ..., + [-1.1753e-06, 4.1444e-08, -1.6969e-06, ..., -6.8964e-07, + -1.2554e-06, 3.7719e-08], + [ 4.7032e-08, 3.3062e-08, 8.4285e-08, ..., 4.4703e-08, + 5.5414e-08, -2.7008e-08], + [ 9.3970e-07, 8.2701e-06, 1.3616e-06, ..., 5.2527e-07, + 1.0319e-06, 4.8727e-06]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0165, -0.0169, -0.0051, -0.0190, -0.0099, 0.0018, 0.0085, 0.0188, + 0.0169, -0.0077], device='cuda:0'), grad: tensor([-1.1735e-07, 3.1060e-07, 2.3469e-07, 8.8476e-08, -3.3677e-05, + -4.7032e-08, 7.1246e-08, -2.4699e-06, 1.6391e-07, 3.5495e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 215.31, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4753 re_mapping 0.0050 re_causal 0.0141 /// teacc 99.07 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.2823, 0.0830, -0.1432, ..., -0.0495, -0.2365, -0.1692], + [ 0.0078, 0.0991, -0.1157, ..., -0.1142, -0.0574, 0.1125], + [ 0.0538, -0.1241, -0.1514, ..., 0.0058, -0.0507, -0.0962], + ..., + [ 0.0910, -0.0501, 0.1323, ..., 0.0661, 0.1887, -0.0120], + [ 0.1137, -0.2057, -0.1301, ..., -0.2639, -0.0646, 0.1938], + [-0.1682, 0.0876, 0.0521, ..., -0.2649, -0.1341, -0.0440]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -4.0652e-07, 5.1223e-09, ..., 4.6566e-10, + 4.6566e-10, 3.5856e-08], + [ 1.8086e-06, 2.8405e-08, 3.7253e-09, ..., 8.8476e-09, + 9.7789e-09, 6.8843e-06], + [-3.7253e-09, 9.3132e-09, 7.4506e-09, ..., -1.3504e-08, + -2.7474e-08, 1.5879e-07], + ..., + [ 9.4064e-08, 8.9407e-08, 2.5611e-08, ..., 1.1176e-08, + 1.4435e-08, 3.9022e-07], + [-2.0601e-06, 1.2992e-07, 1.3970e-08, ..., 1.8626e-09, + 9.3132e-10, -7.8529e-06], + [ 2.0955e-08, 1.8686e-05, -8.4750e-08, ..., 4.6566e-10, + 4.6566e-10, 1.6466e-05]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0165, -0.0162, -0.0050, -0.0197, -0.0090, 0.0018, 0.0084, 0.0186, + 0.0167, -0.0081], device='cuda:0'), grad: tensor([-7.6601e-07, 8.2403e-06, 1.3504e-07, 6.7055e-08, -6.8903e-05, + -1.8161e-07, 5.8860e-07, 8.0653e-07, -8.8662e-06, 6.8903e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 214.83, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5254 re_mapping 0.0050 re_causal 0.0146 /// teacc 99.02 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.2825, 0.0830, -0.1443, ..., -0.0500, -0.2395, -0.1692], + [ 0.0077, 0.0990, -0.1158, ..., -0.1145, -0.0576, 0.1126], + [ 0.0546, -0.1245, -0.1519, ..., 0.0063, -0.0504, -0.0964], + ..., + [ 0.0909, -0.0495, 0.1328, ..., 0.0660, 0.1891, -0.0121], + [ 0.1138, -0.2061, -0.1302, ..., -0.2634, -0.0650, 0.1942], + [-0.1685, 0.0874, 0.0518, ..., -0.2661, -0.1353, -0.0443]], + device='cuda:0'), grad: tensor([[ 1.9558e-08, 1.1176e-08, 3.5390e-08, ..., 1.1176e-08, + 3.4459e-08, 0.0000e+00], + [ 1.3597e-07, -5.1223e-08, 2.9150e-07, ..., 9.3132e-08, + 2.7474e-07, -2.4773e-07], + [-1.4901e-08, 9.3132e-09, 2.6077e-08, ..., -9.3132e-10, + -1.9558e-08, 1.8626e-09], + ..., + [-2.4959e-07, -1.3690e-07, -5.0012e-07, ..., -1.5646e-07, + -4.4797e-07, 6.6124e-08], + [ 1.3039e-08, 2.2352e-08, 1.1176e-08, ..., 8.3819e-09, + 2.5146e-08, 1.2107e-08], + [ 6.7987e-08, 1.8440e-07, -5.3085e-08, ..., 3.0734e-08, + 8.6613e-08, 1.8254e-07]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0165, -0.0162, -0.0047, -0.0199, -0.0087, 0.0016, 0.0086, 0.0189, + 0.0167, -0.0085], device='cuda:0'), grad: tensor([ 9.4995e-08, 1.7043e-07, -7.3574e-08, 2.2352e-08, 1.5739e-07, + -5.3085e-07, 1.0617e-07, -1.0049e-06, 1.3411e-07, 9.2853e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 214.87, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.4706 re_mapping 0.0051 re_causal 0.0138 /// teacc 98.97 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.2838, 0.0829, -0.1449, ..., -0.0510, -0.2402, -0.1694], + [ 0.0075, 0.0991, -0.1160, ..., -0.1148, -0.0578, 0.1128], + [ 0.0546, -0.1246, -0.1526, ..., 0.0065, -0.0503, -0.0965], + ..., + [ 0.0911, -0.0494, 0.1330, ..., 0.0661, 0.1894, -0.0123], + [ 0.1142, -0.2064, -0.1302, ..., -0.2644, -0.0652, 0.1948], + [-0.1687, 0.0874, 0.0518, ..., -0.2667, -0.1357, -0.0445]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 2.0489e-08, 2.6077e-08, ..., 6.5193e-09, + 1.1176e-08, 7.4506e-09], + [ 1.7881e-07, 6.5286e-07, 7.4413e-07, ..., 5.4948e-08, + 1.9744e-07, 1.9465e-07], + [ 4.6253e-05, 4.4703e-08, 1.5453e-05, ..., 1.7300e-05, + 3.3557e-05, 8.7172e-07], + ..., + [-4.6670e-05, 7.4506e-08, -1.5542e-05, ..., -1.7464e-05, + -3.3945e-05, -8.4471e-07], + [ 1.1269e-07, 8.9407e-08, 7.6368e-08, ..., 5.2154e-08, + 1.0245e-07, -3.4459e-08], + [ 2.1420e-08, 1.4342e-07, -4.3772e-08, ..., 1.1176e-08, + 2.4214e-08, 1.6764e-08]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0169, -0.0162, -0.0045, -0.0176, -0.0088, -0.0002, 0.0094, 0.0190, + 0.0170, -0.0086], device='cuda:0'), grad: tensor([ 9.4995e-08, 2.2501e-06, 5.6505e-05, -6.8210e-06, -1.0645e-06, + 4.0047e-06, 3.2317e-07, -5.6624e-05, 5.0105e-07, 9.1922e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 214.93, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4977 re_mapping 0.0048 re_causal 0.0134 /// teacc 99.03 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.2853, 0.0829, -0.1461, ..., -0.0520, -0.2417, -0.1697], + [ 0.0074, 0.0985, -0.1162, ..., -0.1151, -0.0581, 0.1128], + [ 0.0542, -0.1249, -0.1537, ..., 0.0063, -0.0507, -0.0969], + ..., + [ 0.0915, -0.0498, 0.1334, ..., 0.0665, 0.1900, -0.0123], + [ 0.1143, -0.2076, -0.1303, ..., -0.2654, -0.0653, 0.1949], + [-0.1688, 0.0871, 0.0517, ..., -0.2675, -0.1365, -0.0446]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 2.7940e-08, 9.3132e-09, ..., 3.7253e-09, + 9.3132e-10, 3.7253e-08], + [-4.7032e-07, -2.5649e-06, -4.4424e-07, ..., 1.1176e-08, + 5.5879e-09, -2.3413e-06], + [ 3.7253e-09, 5.8673e-08, 1.3970e-08, ..., -6.3330e-08, + -6.5193e-09, 5.2154e-08], + ..., + [ 3.6694e-07, 2.0564e-06, 3.4366e-07, ..., -4.6566e-09, + -9.3132e-09, 1.8682e-06], + [-2.8871e-08, 1.5274e-07, 2.5146e-08, ..., 3.9116e-08, + 1.8626e-09, 3.3528e-08], + [ 4.0978e-08, 3.9302e-07, 3.5390e-08, ..., 2.1420e-08, + 5.5879e-09, 1.6391e-07]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0171, -0.0164, -0.0047, -0.0167, -0.0073, -0.0011, 0.0102, 0.0191, + 0.0167, -0.0092], device='cuda:0'), grad: tensor([ 1.8068e-07, -9.8422e-06, -1.1735e-06, -9.3132e-10, 1.2107e-08, + -5.3272e-07, 3.3434e-07, 7.9125e-06, 1.3225e-06, 1.7472e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 214.67, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4835 re_mapping 0.0048 re_causal 0.0136 /// teacc 99.01 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.2867, 0.0830, -0.1467, ..., -0.0524, -0.2421, -0.1713], + [ 0.0076, 0.1003, -0.1155, ..., -0.1153, -0.0582, 0.1138], + [ 0.0541, -0.1254, -0.1541, ..., 0.0063, -0.0506, -0.0978], + ..., + [ 0.0916, -0.0513, 0.1333, ..., 0.0666, 0.1902, -0.0126], + [ 0.1146, -0.2081, -0.1304, ..., -0.2661, -0.0655, 0.1953], + [-0.1693, 0.0867, 0.0512, ..., -0.2690, -0.1377, -0.0453]], + device='cuda:0'), grad: tensor([[ 5.7742e-08, -1.5888e-06, 7.4506e-09, ..., 3.7253e-09, + 1.8626e-09, 2.0396e-07], + [ 7.7300e-08, -1.0245e-08, 9.0338e-08, ..., 6.7987e-08, + 4.0047e-08, -3.6322e-08], + [ 5.4017e-08, 1.1176e-08, 8.1025e-08, ..., 4.9360e-08, + 2.5146e-08, 2.7940e-09], + ..., + [-2.9895e-07, 3.4459e-08, -3.1479e-07, ..., -1.9744e-07, + -1.3970e-07, 3.5390e-08], + [-1.2666e-07, 3.9116e-08, 5.0291e-08, ..., 2.9802e-08, + 8.3819e-09, -5.3644e-07], + [ 1.9837e-07, 2.8405e-07, -8.9407e-08, ..., 6.2399e-08, + 4.7497e-08, 2.8498e-07]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0170, -0.0159, -0.0046, -0.0157, -0.0068, -0.0019, 0.0102, 0.0188, + 0.0168, -0.0097], device='cuda:0'), grad: tensor([-5.6922e-06, 2.3060e-06, 1.1735e-07, -3.4459e-08, 5.8115e-07, + 1.9222e-06, -1.0058e-07, -3.5111e-07, -8.7824e-07, 2.1085e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 214.43, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4911 re_mapping 0.0047 re_causal 0.0139 /// teacc 99.00 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.2877, 0.0832, -0.1471, ..., -0.0534, -0.2424, -0.1715], + [ 0.0070, 0.1012, -0.1161, ..., -0.1160, -0.0594, 0.1144], + [ 0.0540, -0.1256, -0.1557, ..., 0.0067, -0.0510, -0.0980], + ..., + [ 0.0922, -0.0521, 0.1341, ..., 0.0668, 0.1914, -0.0129], + [ 0.1147, -0.2083, -0.1304, ..., -0.2668, -0.0657, 0.1955], + [-0.1695, 0.0865, 0.0513, ..., -0.2693, -0.1378, -0.0457]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.3469e-07, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-9.3132e-10, -1.6764e-08, -2.3283e-08, ..., 9.3132e-10, + -1.1176e-08, -4.1816e-07], + [ 2.4214e-08, 1.0245e-08, 1.3039e-08, ..., 9.3132e-10, + 5.5879e-09, 2.0675e-07], + ..., + [ 8.3819e-09, 1.1176e-08, 1.3970e-08, ..., 0.0000e+00, + 5.5879e-09, 2.1793e-07], + [-1.8626e-07, 3.7253e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -1.2573e-07], + [ 6.5193e-09, 2.1886e-07, -4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0169, -0.0161, -0.0044, -0.0156, -0.0069, -0.0017, 0.0099, 0.0191, + 0.0168, -0.0099], device='cuda:0'), grad: tensor([-4.8708e-07, -5.2992e-07, 3.0454e-07, 7.3574e-08, 7.4506e-09, + 1.1269e-07, 1.6112e-07, 2.9150e-07, -3.9674e-07, 4.7032e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 214.54, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4817 re_mapping 0.0046 re_causal 0.0142 /// teacc 99.00 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.2878, 0.0834, -0.1478, ..., -0.0536, -0.2425, -0.1712], + [ 0.0067, 0.1013, -0.1165, ..., -0.1165, -0.0602, 0.1147], + [ 0.0544, -0.1259, -0.1560, ..., 0.0070, -0.0507, -0.0985], + ..., + [ 0.0924, -0.0530, 0.1345, ..., 0.0668, 0.1919, -0.0132], + [ 0.1147, -0.2086, -0.1305, ..., -0.2671, -0.0657, 0.1956], + [-0.1695, 0.0860, 0.0511, ..., -0.2694, -0.1405, -0.0459]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.4342e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 2.5146e-08, -1.3039e-08, 1.3039e-08, ..., 3.7253e-09, + 9.3132e-09, -7.4506e-09], + [ 2.2352e-08, 5.5879e-09, 9.3132e-10, ..., -9.3132e-10, + 0.0000e+00, 3.1665e-08], + ..., + [-1.0245e-08, 1.6764e-08, -1.3039e-08, ..., -9.3132e-10, + -1.0245e-08, 2.6077e-08], + [-1.4063e-07, 1.3039e-08, 1.2107e-08, ..., 4.6566e-09, + 1.8626e-09, -1.9930e-07], + [ 5.5879e-09, 2.9337e-07, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 1.4249e-07]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0167, -0.0162, -0.0044, -0.0159, -0.0054, -0.0015, 0.0098, 0.0191, + 0.0168, -0.0107], device='cuda:0'), grad: tensor([-2.7008e-07, 1.5832e-08, 6.2399e-08, -1.5832e-08, -1.0999e-06, + 2.0582e-07, 1.3318e-07, 4.9360e-08, -2.9523e-07, 1.2359e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 214.96, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.5097 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.03 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.2886, 0.0834, -0.1486, ..., -0.0537, -0.2427, -0.1717], + [ 0.0036, 0.1016, -0.1194, ..., -0.1166, -0.0633, 0.1141], + [ 0.0542, -0.1263, -0.1566, ..., 0.0070, -0.0508, -0.0990], + ..., + [ 0.0953, -0.0560, 0.1368, ..., 0.0668, 0.1950, -0.0124], + [ 0.1148, -0.2090, -0.1305, ..., -0.2674, -0.0660, 0.1957], + [-0.1694, 0.0860, 0.0520, ..., -0.2688, -0.1418, -0.0460]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -1.6298e-07, 1.7695e-08, ..., 7.4506e-09, + 8.3819e-09, 1.8626e-09], + [-8.1062e-06, -1.3784e-07, 4.6566e-08, ..., 1.9558e-08, + -7.3649e-06, -2.6356e-07], + [ 7.6592e-06, 4.6566e-09, 2.1420e-08, ..., 1.0245e-08, + 6.9365e-06, 1.4901e-08], + ..., + [-6.8918e-08, 8.2888e-08, -7.9069e-07, ..., -3.3714e-07, + 4.9360e-08, 1.5181e-07], + [-2.2911e-07, 1.4901e-08, 2.2352e-08, ..., 9.3132e-09, + 1.3039e-08, -3.0920e-07], + [ 5.0757e-07, 2.5146e-08, 4.9360e-07, ..., 2.1793e-07, + 2.5425e-07, 2.1793e-07]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0167, -0.0189, -0.0046, -0.0164, -0.0045, -0.0012, 0.0099, 0.0214, + 0.0168, -0.0109], device='cuda:0'), grad: tensor([-2.7940e-07, -3.6627e-05, 3.4064e-05, 1.0524e-07, 4.0699e-07, + 2.2817e-07, 3.8184e-07, 9.2201e-07, -7.1526e-07, 1.5236e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 215.26, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4741 re_mapping 0.0044 re_causal 0.0128 /// teacc 98.97 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.2911, 0.0836, -0.1492, ..., -0.0557, -0.2445, -0.1718], + [ 0.0026, 0.1020, -0.1203, ..., -0.1174, -0.0644, 0.1142], + [ 0.0546, -0.1268, -0.1580, ..., 0.0073, -0.0512, -0.0993], + ..., + [ 0.0963, -0.0563, 0.1377, ..., 0.0672, 0.1961, -0.0122], + [ 0.1148, -0.2096, -0.1306, ..., -0.2689, -0.0665, 0.1958], + [-0.1695, 0.0859, 0.0521, ..., -0.2693, -0.1420, -0.0463]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.0536e-08], + [ 5.2154e-08, -2.7940e-09, 4.4703e-08, ..., 4.7497e-08, + 2.7940e-08, 7.4506e-09], + [ 1.0058e-07, 0.0000e+00, 3.6322e-08, ..., -4.5635e-08, + 2.6077e-08, 2.3097e-07], + ..., + [-9.6858e-08, 5.5879e-09, -1.0617e-07, ..., -4.8429e-08, + -7.0781e-08, 2.4214e-08], + [-1.5739e-07, 2.7940e-09, 4.6566e-09, ..., 2.7940e-09, + 1.8626e-09, -4.5728e-07], + [ 1.5832e-08, -6.2399e-08, -5.9605e-08, ..., 6.5193e-09, + -2.0489e-08, 1.5832e-08]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0165, -0.0198, -0.0038, -0.0166, -0.0044, -0.0020, 0.0111, 0.0223, + 0.0166, -0.0111], device='cuda:0'), grad: tensor([ 8.0094e-08, 1.8161e-07, 1.2666e-07, 1.7695e-08, 2.1048e-07, + 1.2107e-08, 2.1886e-07, -1.3411e-07, -5.7928e-07, -1.3132e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 215.10, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.4839 re_mapping 0.0046 re_causal 0.0133 /// teacc 98.98 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.2898, 0.0838, -0.1501, ..., -0.0561, -0.2453, -0.1686], + [ 0.0005, 0.1032, -0.1229, ..., -0.1178, -0.0672, 0.1129], + [ 0.0547, -0.1274, -0.1601, ..., 0.0076, -0.0515, -0.0995], + ..., + [ 0.0984, -0.0578, 0.1405, ..., 0.0673, 0.1990, -0.0107], + [ 0.1148, -0.2103, -0.1308, ..., -0.2696, -0.0670, 0.1957], + [-0.1697, 0.0857, 0.0522, ..., -0.2697, -0.1422, -0.0465]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -2.5239e-07, 4.6566e-09, ..., 1.8626e-09, + 1.8626e-09, 3.7253e-09], + [ 5.1223e-08, 2.0489e-08, 1.0896e-07, ..., 3.0734e-08, + 3.0734e-08, 1.8626e-09], + [-1.1176e-08, 2.7940e-09, 5.5879e-09, ..., -1.3039e-08, + -0.0000e+00, 2.0489e-08], + ..., + [-4.1910e-08, 9.3132e-10, 2.7940e-08, ..., 4.0047e-08, + -4.2841e-08, 2.7008e-08], + [-5.2154e-08, 1.3039e-08, 4.6566e-09, ..., 1.1176e-08, + 9.3132e-10, -1.0710e-07], + [ 3.2596e-08, 1.8626e-08, 3.8184e-08, ..., 1.7695e-08, + 7.4506e-09, 4.0047e-08]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0162, -0.0225, -0.0037, -0.0169, -0.0043, 0.0009, 0.0081, 0.0250, + 0.0164, -0.0113], device='cuda:0'), grad: tensor([-5.4762e-07, 2.1979e-07, -3.3528e-08, 3.4086e-07, 4.5635e-08, + -4.0699e-07, 5.3085e-08, 1.1455e-07, 3.0734e-08, 1.8533e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 215.05, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.4494 re_mapping 0.0049 re_causal 0.0126 /// teacc 98.96 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.2901, 0.0838, -0.1509, ..., -0.0567, -0.2456, -0.1688], + [ 0.0005, 0.1040, -0.1228, ..., -0.1181, -0.0673, 0.1137], + [ 0.0548, -0.1279, -0.1607, ..., 0.0076, -0.0513, -0.0999], + ..., + [ 0.0984, -0.0598, 0.1403, ..., 0.0673, 0.1991, -0.0117], + [ 0.1148, -0.2119, -0.1311, ..., -0.2705, -0.0672, 0.1959], + [-0.1695, 0.0864, 0.0541, ..., -0.2699, -0.1423, -0.0454]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 6.5193e-09, 1.6764e-08, ..., 7.4506e-09, + 7.4506e-09, 6.5193e-09], + [ 5.8673e-07, 1.3728e-06, 2.9616e-06, ..., 4.3306e-07, + 3.9674e-07, 1.0896e-06], + [ 1.8859e-06, 2.7940e-08, 1.4203e-06, ..., 1.3756e-06, + 1.3253e-06, 2.7940e-08], + ..., + [-3.5986e-06, 1.1381e-06, -9.2853e-07, ..., -2.6468e-06, + -2.4755e-06, 9.5926e-07], + [ 7.4506e-08, 1.5181e-07, 2.8592e-07, ..., 5.5879e-08, + 5.1223e-08, 8.1956e-08], + [ 9.2201e-08, -2.8089e-06, -5.5246e-06, ..., 6.1467e-08, + 5.8673e-08, -1.8142e-06]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0162, -0.0224, -0.0040, -0.0174, -0.0060, 0.0012, 0.0083, 0.0248, + 0.0159, -0.0099], device='cuda:0'), grad: tensor([ 5.5879e-08, 6.7428e-06, 4.0643e-06, 2.0824e-06, -5.0943e-07, + -2.8312e-07, 2.0210e-07, -2.8946e-06, 9.0059e-07, -1.0334e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 214.82, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5117 re_mapping 0.0047 re_causal 0.0133 /// teacc 99.01 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.2910, 0.0839, -0.1518, ..., -0.0573, -0.2464, -0.1694], + [ 0.0004, 0.1042, -0.1231, ..., -0.1184, -0.0674, 0.1135], + [ 0.0560, -0.1282, -0.1612, ..., 0.0087, -0.0508, -0.1004], + ..., + [ 0.0984, -0.0602, 0.1407, ..., 0.0669, 0.1992, -0.0114], + [ 0.1148, -0.2123, -0.1312, ..., -0.2730, -0.0675, 0.1963], + [-0.1700, 0.0863, 0.0539, ..., -0.2708, -0.1434, -0.0459]], + device='cuda:0'), grad: tensor([[ 2.2445e-07, 5.5879e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.0454e-07], + [ 2.0489e-08, -0.0000e+00, 5.4017e-08, ..., 9.3132e-09, + 4.6566e-09, -8.0094e-08], + [ 1.8626e-09, 1.0245e-08, 2.3283e-08, ..., 3.7253e-09, + -8.3819e-09, 6.6124e-08], + ..., + [ 4.2841e-08, 9.1270e-08, 2.4214e-08, ..., 1.1176e-08, + -9.3132e-10, 6.7055e-08], + [-3.2224e-07, 6.7055e-08, 8.4750e-08, ..., 7.4506e-09, + 1.8626e-09, -5.1409e-07], + [-2.7940e-08, -2.6543e-07, -3.6228e-07, ..., 9.3132e-10, + 1.8626e-09, 1.3411e-07]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0163, -0.0225, -0.0029, -0.0182, -0.0056, 0.0015, 0.0083, 0.0248, + 0.0158, -0.0101], device='cuda:0'), grad: tensor([ 4.8801e-07, -1.8626e-09, 7.9162e-08, -2.6077e-08, 5.4948e-08, + 6.4261e-08, 6.9849e-08, 3.7625e-07, -3.9581e-07, -7.0408e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 214.93, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4907 re_mapping 0.0048 re_causal 0.0134 /// teacc 98.98 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.2916, 0.0839, -0.1525, ..., -0.0576, -0.2471, -0.1696], + [ 0.0004, 0.1046, -0.1232, ..., -0.1189, -0.0675, 0.1139], + [ 0.0560, -0.1286, -0.1620, ..., 0.0089, -0.0508, -0.1016], + ..., + [ 0.0985, -0.0606, 0.1408, ..., 0.0670, 0.1993, -0.0116], + [ 0.1152, -0.2127, -0.1313, ..., -0.2737, -0.0676, 0.1968], + [-0.1702, 0.0862, 0.0540, ..., -0.2712, -0.1436, -0.0461]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 2.0675e-07, -1.9558e-08, 1.7136e-07, ..., 3.0734e-08, + 7.3574e-08, 1.6764e-08], + [-3.7253e-09, 2.7940e-09, 3.7253e-09, ..., -5.5879e-09, + -4.6566e-09, 4.6566e-09], + ..., + [-3.4180e-07, 2.2352e-08, -2.8498e-07, ..., -4.7497e-08, + -1.1735e-07, -4.7497e-08], + [ 7.8231e-08, 1.1176e-08, 6.7055e-08, ..., 1.3039e-08, + 2.7008e-08, 2.7940e-08], + [ 4.1910e-08, 1.1437e-06, 2.9802e-08, ..., 6.5193e-09, + 1.4901e-08, 1.3039e-06]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0162, -0.0224, -0.0032, -0.0187, -0.0053, 0.0016, 0.0084, 0.0247, + 0.0159, -0.0103], device='cuda:0'), grad: tensor([ 1.8626e-09, 2.9150e-07, -4.6566e-09, 2.0489e-08, -5.8301e-06, + 1.0245e-08, 2.7101e-07, -5.0757e-07, 1.9278e-07, 5.5321e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 215.13, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4949 re_mapping 0.0048 re_causal 0.0137 /// teacc 99.01 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.2922, 0.0839, -0.1535, ..., -0.0584, -0.2473, -0.1698], + [ 0.0005, 0.1084, -0.1226, ..., -0.1191, -0.0675, 0.1175], + [ 0.0543, -0.1321, -0.1657, ..., 0.0077, -0.0520, -0.1052], + ..., + [ 0.0987, -0.0630, 0.1408, ..., 0.0680, 0.1995, -0.0132], + [ 0.1153, -0.2132, -0.1313, ..., -0.2751, -0.0678, 0.1971], + [-0.1705, 0.0863, 0.0541, ..., -0.2723, -0.1439, -0.0467]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.7253e-09, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [-1.5646e-07, -3.9022e-07, -1.0245e-08, ..., 5.5879e-09, + 1.8626e-09, -6.7987e-07], + [-1.2293e-07, 1.1176e-08, 2.7940e-09, ..., -8.7544e-08, + -3.6322e-08, 6.5193e-09], + ..., + [ 9.7789e-08, 1.0896e-07, 3.3528e-08, ..., 6.7987e-08, + 3.1665e-08, 6.4261e-08], + [ 1.3970e-08, 7.4506e-09, 4.6566e-09, ..., 1.1176e-08, + 9.3132e-10, 7.4506e-09], + [ 3.7253e-09, 3.2596e-08, -5.8673e-08, ..., 1.8626e-09, + 1.8626e-09, 2.3283e-08]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0162, -0.0209, -0.0070, -0.0192, -0.0056, 0.0018, 0.0085, 0.0246, + 0.0160, -0.0103], device='cuda:0'), grad: tensor([-9.3132e-10, -1.7025e-06, -1.9372e-07, 4.3772e-08, -1.1176e-07, + 2.1420e-08, 1.2880e-06, 5.7090e-07, 6.1467e-08, 4.5635e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 215.30, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4658 re_mapping 0.0049 re_causal 0.0138 /// teacc 99.03 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.2925, 0.0840, -0.1539, ..., -0.0586, -0.2474, -0.1700], + [ 0.0005, 0.1087, -0.1224, ..., -0.1191, -0.0675, 0.1179], + [ 0.0540, -0.1322, -0.1664, ..., 0.0075, -0.0529, -0.1052], + ..., + [ 0.0987, -0.0633, 0.1407, ..., 0.0682, 0.1996, -0.0137], + [ 0.1153, -0.2138, -0.1315, ..., -0.2759, -0.0679, 0.1971], + [-0.1705, 0.0871, 0.0567, ..., -0.2724, -0.1439, -0.0460]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 1.4901e-08, 9.5926e-08, ..., 1.9558e-08, + 2.7940e-09, 1.9930e-07], + [-6.1244e-06, -5.5246e-06, -2.1562e-05, ..., -3.4738e-06, + -1.3597e-07, -4.8161e-05], + [ 1.1530e-06, 8.9593e-07, 3.9302e-06, ..., 5.7090e-07, + 1.1083e-07, 8.0317e-06], + ..., + [ 2.6245e-06, 2.9244e-06, 1.0878e-05, ..., 1.6522e-06, + -1.3597e-07, 2.5541e-05], + [ 1.4622e-07, 1.0431e-07, 3.4459e-07, ..., 8.5682e-08, + 1.4901e-08, 8.8289e-07], + [ 1.5553e-07, 2.6077e-08, 1.2014e-07, ..., 6.0536e-08, + 4.1910e-08, 1.2759e-07]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0162, -0.0207, -0.0071, -0.0192, -0.0069, 0.0019, 0.0083, 0.0244, + 0.0158, -0.0094], device='cuda:0'), grad: tensor([ 3.5856e-07, -8.4579e-05, 1.4186e-05, 2.2903e-05, 2.6822e-07, + 6.0070e-07, 4.0047e-07, 4.3571e-05, 1.8459e-06, 5.6252e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 214.99, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4816 re_mapping 0.0049 re_causal 0.0130 /// teacc 99.06 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.2933, 0.0841, -0.1548, ..., -0.0590, -0.2477, -0.1706], + [ 0.0008, 0.1090, -0.1223, ..., -0.1197, -0.0673, 0.1190], + [ 0.0533, -0.1324, -0.1674, ..., 0.0075, -0.0541, -0.1057], + ..., + [ 0.0988, -0.0638, 0.1401, ..., 0.0688, 0.1997, -0.0146], + [ 0.1148, -0.2144, -0.1317, ..., -0.2774, -0.0683, 0.1967], + [-0.1710, 0.0871, 0.0594, ..., -0.2731, -0.1456, -0.0462]], + device='cuda:0'), grad: tensor([[ 1.9558e-08, -2.9430e-07, 4.6566e-09, ..., 9.3132e-10, + 9.3132e-10, 3.5390e-08], + [ 2.7940e-08, 1.4901e-08, 7.4506e-08, ..., 1.8626e-08, + 3.0734e-08, -3.2596e-08], + [ 2.1420e-08, 7.4506e-09, 2.7940e-08, ..., 5.5879e-09, + 9.3132e-09, 1.1176e-08], + ..., + [-1.3504e-07, 2.5146e-08, -4.2934e-07, ..., -7.3574e-08, + -1.9465e-07, 3.7253e-08], + [-9.1270e-08, 1.1269e-07, 5.0291e-08, ..., 1.8626e-09, + 2.7940e-09, -1.8161e-07], + [ 1.4529e-07, 1.9129e-06, 2.1141e-07, ..., 4.7497e-08, + 1.4063e-07, 1.3411e-07]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0161, -0.0201, -0.0079, -0.0198, -0.0068, 0.0016, 0.0081, 0.0239, + 0.0150, -0.0086], device='cuda:0'), grad: tensor([-3.7998e-07, 1.3970e-07, 7.3574e-08, 6.8638e-07, -2.9802e-08, + -8.7395e-06, 5.6252e-07, -4.5355e-07, 1.9930e-07, 7.9572e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 215.22, cls_loss 0.0017 cls_loss_mapping 0.0017 cls_loss_causal 0.4823 re_mapping 0.0050 re_causal 0.0131 /// teacc 98.96 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.2936, 0.0809, -0.1561, ..., -0.0616, -0.2482, -0.1701], + [ 0.0008, 0.1091, -0.1225, ..., -0.1201, -0.0670, 0.1192], + [ 0.0529, -0.1325, -0.1681, ..., 0.0072, -0.0552, -0.1058], + ..., + [ 0.0991, -0.0641, 0.1406, ..., 0.0699, 0.1996, -0.0149], + [ 0.1148, -0.2151, -0.1319, ..., -0.2790, -0.0688, 0.1967], + [-0.1713, 0.0868, 0.0593, ..., -0.2737, -0.1459, -0.0466]], + device='cuda:0'), grad: tensor([[-8.4657e-07, -1.5683e-06, 1.3039e-08, ..., 9.3132e-09, + 8.3819e-09, -1.9185e-07], + [ 2.5332e-07, 7.7300e-08, 3.8184e-07, ..., 6.0536e-08, + 1.9465e-07, -2.6077e-08], + [ 1.1828e-07, 1.9744e-07, 1.2293e-07, ..., -1.8626e-09, + 3.0734e-08, 4.6566e-08], + ..., + [-2.1700e-07, 5.1409e-07, -1.0766e-06, ..., -1.8068e-07, + -5.4482e-07, 9.7789e-08], + [ 7.1712e-08, 7.8231e-08, 6.9849e-08, ..., 3.0734e-08, + 3.9116e-08, -9.3132e-10], + [ 4.6194e-07, 4.6846e-07, 4.3772e-07, ..., 7.9162e-08, + 2.3469e-07, 6.7987e-08]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0195, -0.0196, -0.0083, -0.0210, -0.0061, 0.0036, 0.0103, 0.0237, + 0.0147, -0.0091], device='cuda:0'), grad: tensor([-4.9919e-06, 8.4471e-07, 6.3889e-07, -4.0047e-08, 9.4995e-08, + 2.8219e-07, 4.8894e-07, 2.0582e-07, 3.5018e-07, 2.1476e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 215.77, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4722 re_mapping 0.0049 re_causal 0.0133 /// teacc 98.99 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.2941, 0.0809, -0.1571, ..., -0.0622, -0.2487, -0.1696], + [ 0.0008, 0.1090, -0.1225, ..., -0.1195, -0.0669, 0.1192], + [ 0.0530, -0.1326, -0.1683, ..., 0.0071, -0.0553, -0.1058], + ..., + [ 0.0991, -0.0645, 0.1406, ..., 0.0696, 0.1995, -0.0151], + [ 0.1152, -0.2156, -0.1321, ..., -0.2801, -0.0690, 0.1977], + [-0.1714, 0.0870, 0.0595, ..., -0.2745, -0.1462, -0.0468]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, -1.5367e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-08], + [ 1.5832e-08, -5.5134e-07, 1.0245e-08, ..., 7.4506e-09, + 4.6566e-09, -4.8149e-07], + [ 3.9116e-08, 4.8429e-08, 2.6077e-08, ..., 1.8626e-08, + 1.3039e-08, 5.9605e-08], + ..., + [-3.9116e-08, 1.4156e-07, -3.4459e-08, ..., -2.0489e-08, + -2.4214e-08, 1.3132e-07], + [-8.2627e-06, 5.6811e-08, -3.0305e-06, ..., -7.2923e-07, + 9.3132e-10, -8.2627e-06], + [ 7.2643e-08, 4.0047e-07, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 3.6694e-07]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0195, -0.0196, -0.0083, -0.0210, -0.0059, 0.0034, 0.0103, 0.0236, + 0.0149, -0.0091], device='cuda:0'), grad: tensor([-1.0990e-07, -1.7229e-06, 2.4680e-07, 2.0206e-05, 6.7987e-08, + 2.1141e-07, 1.4063e-07, 3.9581e-07, -2.0802e-05, 1.3700e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 217.41, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4671 re_mapping 0.0048 re_causal 0.0134 /// teacc 98.99 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.2946, 0.0810, -0.1576, ..., -0.0627, -0.2490, -0.1698], + [ 0.0009, 0.1095, -0.1224, ..., -0.1195, -0.0669, 0.1196], + [ 0.0532, -0.1326, -0.1683, ..., 0.0076, -0.0552, -0.1059], + ..., + [ 0.0990, -0.0653, 0.1407, ..., 0.0695, 0.1995, -0.0154], + [ 0.1154, -0.2162, -0.1321, ..., -0.2812, -0.0691, 0.1981], + [-0.1717, 0.0866, 0.0595, ..., -0.2749, -0.1463, -0.0475]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 3.3528e-08, 9.3132e-10, 4.6566e-09, ..., 9.3132e-09, + 3.7253e-09, 0.0000e+00], + [-3.3248e-07, 0.0000e+00, 1.2107e-08, ..., -8.1956e-08, + 1.2107e-08, 0.0000e+00], + ..., + [-1.0245e-08, 3.7253e-09, -4.6566e-09, ..., -6.5193e-09, + -1.8626e-08, 0.0000e+00], + [ 1.8254e-07, 2.7940e-09, 6.5193e-09, ..., 4.6566e-08, + 1.8626e-09, 0.0000e+00], + [ 6.5193e-09, -1.0245e-08, -2.8871e-08, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0195, -0.0194, -0.0082, -0.0219, -0.0053, 0.0036, 0.0103, 0.0235, + 0.0149, -0.0095], device='cuda:0'), grad: tensor([ 3.1665e-08, 1.2293e-07, -1.3206e-06, 1.5181e-07, 2.3656e-07, + 2.4214e-08, 2.7008e-08, 5.6811e-08, 7.1246e-07, -4.0047e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 219.60, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4936 re_mapping 0.0045 re_causal 0.0131 /// teacc 98.98 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.2968, 0.0809, -0.1602, ..., -0.0633, -0.2495, -0.1715], + [ 0.0012, 0.1098, -0.1225, ..., -0.1198, -0.0669, 0.1204], + [ 0.0522, -0.1327, -0.1688, ..., 0.0073, -0.0556, -0.1059], + ..., + [ 0.0992, -0.0661, 0.1408, ..., 0.0700, 0.1996, -0.0158], + [ 0.1146, -0.2193, -0.1324, ..., -0.2821, -0.0693, 0.1972], + [-0.1715, 0.0863, 0.0596, ..., -0.2754, -0.1466, -0.0473]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 9.3132e-09, 3.5390e-08, ..., 2.7940e-09, + 9.3132e-10, 2.6077e-08], + [ 2.7940e-08, 2.2352e-08, 5.1223e-08, ..., 6.5193e-09, + 9.3132e-09, -3.6322e-08], + [-2.7940e-09, 3.7253e-09, 2.7940e-09, ..., -1.7695e-08, + -4.6566e-09, 1.1176e-08], + ..., + [-2.3283e-08, 1.2200e-07, 8.1025e-08, ..., -9.3132e-10, + -1.6764e-08, 6.0536e-08], + [-1.1912e-06, -2.3562e-07, 4.6566e-08, ..., 1.8626e-09, + 1.8626e-09, -2.5053e-06], + [ 1.8626e-09, -1.8813e-07, -2.8405e-07, ..., 2.7940e-09, + 6.5193e-09, 1.7695e-08]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0195, -0.0191, -0.0086, -0.0221, -0.0043, 0.0041, 0.0103, 0.0235, + 0.0135, -0.0097], device='cuda:0'), grad: tensor([ 8.2888e-08, 5.3085e-08, -2.2352e-08, 1.3039e-07, -2.9523e-07, + 3.5334e-06, 1.1548e-07, 3.4925e-07, -3.6247e-06, -3.2317e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 215.04, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4892 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.01 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.2976, 0.0807, -0.1625, ..., -0.0637, -0.2498, -0.1718], + [ 0.0014, 0.1094, -0.1224, ..., -0.1184, -0.0663, 0.1205], + [ 0.0506, -0.1329, -0.1696, ..., 0.0064, -0.0572, -0.1060], + ..., + [ 0.0994, -0.0663, 0.1409, ..., 0.0700, 0.1993, -0.0159], + [ 0.1148, -0.2206, -0.1326, ..., -0.2844, -0.0696, 0.1975], + [-0.1718, 0.0871, 0.0597, ..., -0.2767, -0.1468, -0.0477]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.9802e-06, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 9.3132e-10], + [ 2.7940e-09, -1.1176e-08, 2.7940e-09, ..., 1.2107e-08, + 6.5193e-09, -3.8184e-08], + [-7.3574e-08, 2.4214e-08, 9.3132e-10, ..., -7.6368e-08, + -3.2596e-08, 1.8626e-09], + ..., + [ 5.9605e-08, 1.4901e-08, -2.7940e-09, ..., 5.0291e-08, + 2.3283e-08, 4.0978e-08], + [ 7.4506e-09, 1.7695e-08, 9.3132e-10, ..., 6.5193e-09, + 2.7940e-09, 1.8626e-09], + [ 9.3132e-10, -3.2224e-06, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0196, -0.0187, -0.0092, -0.0236, -0.0040, 0.0057, 0.0102, 0.0232, + 0.0133, -0.0094], device='cuda:0'), grad: tensor([ 4.9174e-06, -2.3283e-08, -2.0768e-07, 1.2107e-08, 2.6077e-07, + 7.4506e-09, 6.5193e-09, 2.3562e-07, 6.1467e-08, -5.2825e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 215.95, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4695 re_mapping 0.0047 re_causal 0.0135 /// teacc 99.03 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.2980, 0.0808, -0.1628, ..., -0.0640, -0.2500, -0.1718], + [ 0.0013, 0.1094, -0.1225, ..., -0.1187, -0.0663, 0.1205], + [ 0.0504, -0.1330, -0.1698, ..., 0.0065, -0.0574, -0.1060], + ..., + [ 0.0995, -0.0664, 0.1410, ..., 0.0702, 0.1994, -0.0159], + [ 0.1148, -0.2210, -0.1327, ..., -0.2859, -0.0699, 0.1976], + [-0.1721, 0.0867, 0.0596, ..., -0.2769, -0.1473, -0.0480]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.7695e-08, 1.3970e-09, ..., 3.7253e-09, + 4.6566e-10, 9.3132e-10], + [ 4.7963e-08, -1.1176e-08, 7.3574e-08, ..., 4.4238e-08, + 2.0489e-08, -1.1222e-07], + [ 1.1176e-08, 1.6298e-08, 2.0489e-08, ..., 3.3993e-08, + -2.7940e-09, 1.3039e-08], + ..., + [-1.9651e-07, 8.1956e-08, -2.5379e-07, ..., -1.0058e-07, + -5.9605e-08, 5.7742e-08], + [-9.6858e-08, 5.1223e-09, 1.0710e-08, ..., 2.3283e-09, + 1.3970e-09, -2.1234e-07], + [ 2.3283e-08, 1.0850e-07, 1.5367e-08, ..., 1.1642e-08, + 6.9849e-09, 3.7253e-09]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0195, -0.0187, -0.0091, -0.0224, -0.0038, 0.0054, 0.0101, 0.0233, + 0.0130, -0.0098], device='cuda:0'), grad: tensor([-2.9337e-08, 2.7008e-08, 1.1502e-07, 1.8626e-08, -4.5169e-07, + 1.2526e-07, 1.7276e-07, -1.4296e-07, -2.0349e-07, 3.7672e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 214.79, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4800 re_mapping 0.0046 re_causal 0.0126 /// teacc 99.01 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.2982, 0.0796, -0.1641, ..., -0.0643, -0.2504, -0.1720], + [ 0.0012, 0.1095, -0.1226, ..., -0.1190, -0.0664, 0.1207], + [ 0.0504, -0.1332, -0.1703, ..., 0.0059, -0.0575, -0.1062], + ..., + [ 0.0996, -0.0670, 0.1412, ..., 0.0702, 0.1996, -0.0161], + [ 0.1151, -0.2214, -0.1327, ..., -0.2868, -0.0695, 0.1978], + [-0.1725, 0.0894, 0.0595, ..., -0.2773, -0.1484, -0.0484]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -2.0955e-08, 2.7940e-09, ..., 2.3283e-09, + 4.6566e-10, 9.3132e-10], + [ 8.7079e-08, -6.5193e-09, 7.4506e-08, ..., 7.0315e-08, + 4.0513e-08, -3.0734e-08], + [-5.6345e-08, 1.8626e-09, 1.2107e-08, ..., -6.4261e-08, + -8.8476e-09, 2.3283e-09], + ..., + [-1.9139e-07, 1.6298e-08, -1.6345e-07, ..., -1.1688e-07, + -1.0384e-07, 2.5146e-08], + [-1.5832e-08, 9.7789e-09, 1.5367e-08, ..., 2.5611e-08, + 1.0710e-08, -5.0757e-08], + [ 1.3039e-08, -3.9395e-07, -3.7905e-07, ..., 1.5367e-08, + 3.7253e-09, 3.6322e-08]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0205, -0.0188, -0.0093, -0.0226, -0.0037, 0.0057, 0.0102, 0.0233, + 0.0128, -0.0073], device='cuda:0'), grad: tensor([-2.8871e-08, 2.0629e-07, -2.8918e-07, 7.2177e-08, 1.0841e-06, + 7.6834e-08, -4.0978e-08, -2.5705e-07, 4.7963e-08, -8.7125e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 215.15, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4611 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.00 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.2989, 0.0796, -0.1644, ..., -0.0648, -0.2506, -0.1724], + [ 0.0012, 0.1095, -0.1227, ..., -0.1196, -0.0665, 0.1207], + [ 0.0508, -0.1333, -0.1704, ..., 0.0064, -0.0572, -0.1061], + ..., + [ 0.0997, -0.0673, 0.1413, ..., 0.0703, 0.1996, -0.0163], + [ 0.1153, -0.2213, -0.1328, ..., -0.2893, -0.0698, 0.1987], + [-0.1727, 0.0892, 0.0597, ..., -0.2777, -0.1483, -0.0487]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, -4.0513e-08, 7.4506e-09, ..., 2.2817e-08, + 0.0000e+00, 6.5193e-09], + [ 2.3004e-07, 1.1176e-08, 2.9337e-08, ..., 2.1420e-08, + 1.2573e-08, 1.8347e-07], + [ 2.0787e-06, 4.6566e-09, 2.3283e-08, ..., 2.2817e-08, + 1.9558e-08, 1.7509e-06], + ..., + [-6.2399e-08, 4.1910e-09, -5.3085e-08, ..., -4.4238e-08, + -5.4482e-08, 6.5193e-09], + [-2.9989e-06, 5.9139e-08, 7.2643e-08, ..., 2.0023e-08, + 9.3132e-09, -2.6226e-06], + [ 1.3178e-07, -2.9802e-08, -9.4064e-08, ..., 2.7940e-09, + 2.3283e-09, 1.3551e-07]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0205, -0.0188, -0.0092, -0.0235, -0.0031, 0.0062, 0.0102, 0.0233, + 0.0130, -0.0075], device='cuda:0'), grad: tensor([-1.0664e-07, 7.7626e-07, 6.7316e-06, 2.3283e-09, 1.1595e-07, + 1.4007e-06, 3.8464e-07, 1.4435e-08, -9.5218e-06, 2.0675e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 215.02, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4717 re_mapping 0.0045 re_causal 0.0131 /// teacc 98.95 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.2989, 0.0796, -0.1647, ..., -0.0650, -0.2509, -0.1723], + [ 0.0012, 0.1094, -0.1227, ..., -0.1197, -0.0665, 0.1208], + [ 0.0507, -0.1333, -0.1706, ..., 0.0064, -0.0572, -0.1062], + ..., + [ 0.0996, -0.0677, 0.1413, ..., 0.0702, 0.1996, -0.0168], + [ 0.1156, -0.2216, -0.1329, ..., -0.2898, -0.0699, 0.1993], + [-0.1729, 0.0887, 0.0594, ..., -0.2779, -0.1483, -0.0492]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.2731e-06, 9.3132e-09, ..., 7.4506e-09, + 9.3132e-10, 1.8626e-09], + [ 3.8184e-08, 2.7940e-09, 6.3330e-08, ..., 5.2154e-08, + 2.7940e-08, -4.2841e-08], + [-5.6438e-07, 1.4901e-08, -3.4180e-07, ..., -3.4925e-07, + -5.8860e-07, 9.3132e-09], + ..., + [ 5.4110e-07, 1.6764e-08, 3.7625e-07, ..., 3.7253e-07, + 5.5227e-07, 2.4214e-08], + [-1.5832e-07, 1.6764e-08, 7.1712e-08, ..., 4.7497e-08, + 1.8626e-09, -3.4086e-07], + [ 9.3132e-10, 5.4017e-08, -1.2107e-08, ..., 3.7253e-09, + 9.3132e-10, 3.2596e-08]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0205, -0.0188, -0.0092, -0.0235, -0.0017, 0.0062, 0.0102, 0.0232, + 0.0132, -0.0081], device='cuda:0'), grad: tensor([ 2.7940e-06, 9.7789e-08, -1.2489e-06, -4.4610e-07, -1.3504e-07, + 1.5646e-07, -2.0862e-06, 1.3504e-06, -6.1281e-07, 1.3225e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 214.63, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4845 re_mapping 0.0045 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.2993, 0.0799, -0.1656, ..., -0.0656, -0.2519, -0.1707], + [ 0.0012, 0.1095, -0.1228, ..., -0.1201, -0.0666, 0.1211], + [ 0.0504, -0.1334, -0.1711, ..., 0.0063, -0.0574, -0.1062], + ..., + [ 0.0997, -0.0682, 0.1413, ..., 0.0705, 0.1997, -0.0174], + [ 0.1161, -0.2220, -0.1331, ..., -0.2908, -0.0705, 0.2008], + [-0.1728, 0.0890, 0.0609, ..., -0.2785, -0.1482, -0.0494]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, -0.0000e+00], + [ 3.7532e-07, -3.7253e-08, 8.3819e-09, ..., 3.7253e-09, + 8.0094e-08, -8.2888e-08], + [-6.5751e-07, 1.8626e-09, 2.7940e-09, ..., -9.3132e-10, + -1.3690e-07, 2.7940e-09], + ..., + [ 4.3772e-08, 2.9802e-08, -1.5832e-08, ..., -4.6566e-09, + 1.8626e-09, 4.3772e-08], + [ 7.2643e-08, 1.8813e-07, 1.0245e-08, ..., 0.0000e+00, + 1.4901e-08, 2.7940e-09], + [ 1.3970e-08, 1.0245e-08, -7.4506e-09, ..., 2.7940e-09, + 6.5193e-09, 1.8626e-08]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0203, -0.0187, -0.0095, -0.0237, -0.0025, 0.0063, 0.0099, 0.0231, + 0.0138, -0.0077], device='cuda:0'), grad: tensor([ 8.9407e-08, 1.0571e-06, -2.1290e-06, 2.2352e-08, 2.0862e-07, + -4.0904e-06, 3.4533e-06, 2.7753e-07, 1.0347e-06, 8.7544e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 214.85, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4677 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.00 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.3000, 0.0800, -0.1664, ..., -0.0660, -0.2523, -0.1708], + [ 0.0009, 0.1096, -0.1229, ..., -0.1206, -0.0669, 0.1212], + [ 0.0511, -0.1334, -0.1708, ..., 0.0067, -0.0568, -0.1059], + ..., + [ 0.0999, -0.0685, 0.1414, ..., 0.0706, 0.2000, -0.0181], + [ 0.1165, -0.2223, -0.1333, ..., -0.2927, -0.0710, 0.2014], + [-0.1731, 0.0887, 0.0609, ..., -0.2790, -0.1484, -0.0500]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.1211e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 1.0245e-08, -4.5635e-08, 6.5193e-09, ..., 2.7940e-09, + 6.5193e-09, -8.6613e-08], + [-1.2107e-08, 1.5832e-08, 9.3132e-10, ..., -3.7253e-09, + -6.5193e-09, 2.7940e-09], + ..., + [ 1.8626e-09, 2.4214e-08, -6.5193e-09, ..., 9.3132e-10, + 0.0000e+00, 4.1910e-08], + [ 0.0000e+00, 7.4506e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 1.8626e-09, 8.7265e-07, -1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 8.1956e-08]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0203, -0.0188, -0.0088, -0.0238, -0.0021, 0.0061, 0.0099, 0.0231, + 0.0140, -0.0080], device='cuda:0'), grad: tensor([-2.0433e-06, -1.7881e-07, 1.6764e-08, 2.3283e-08, -2.8033e-07, + 1.8626e-09, -8.8476e-08, 1.2200e-07, 9.2201e-08, 2.3283e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 214.79, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.5133 re_mapping 0.0047 re_causal 0.0139 /// teacc 99.06 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.3004, 0.0800, -0.1669, ..., -0.0662, -0.2527, -0.1711], + [ 0.0005, 0.1094, -0.1234, ..., -0.1207, -0.0674, 0.1210], + [ 0.0523, -0.1335, -0.1707, ..., 0.0078, -0.0555, -0.1059], + ..., + [ 0.1001, -0.0687, 0.1419, ..., 0.0699, 0.2004, -0.0176], + [ 0.1164, -0.2226, -0.1336, ..., -0.2941, -0.0717, 0.2015], + [-0.1733, 0.0887, 0.0609, ..., -0.2792, -0.1488, -0.0504]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.7940e-09, 5.5879e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 2.8871e-08, -8.3819e-09, 5.7742e-08, ..., 4.6566e-09, + 3.1665e-08, -3.3528e-08], + [ 4.6566e-09, 9.3132e-10, 1.7695e-08, ..., -5.5879e-09, + 8.3819e-09, -2.6077e-08], + ..., + [-1.9930e-07, 7.4506e-09, -3.6974e-07, ..., 1.7695e-08, + -2.1607e-07, 2.4214e-08], + [ 6.5193e-09, 4.7497e-08, 7.5437e-08, ..., 9.3132e-10, + 3.7253e-09, 2.0489e-08], + [ 1.5087e-07, -2.3283e-08, 1.7881e-07, ..., 0.0000e+00, + 1.5926e-07, -3.7253e-09]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0202, -0.0190, -0.0085, -0.0238, -0.0020, 0.0058, 0.0100, 0.0232, + 0.0138, -0.0081], device='cuda:0'), grad: tensor([ 3.4459e-08, 7.2643e-08, -2.7940e-06, 1.6764e-08, 2.2631e-07, + -1.8720e-07, -9.9652e-08, -5.4762e-07, 5.0571e-07, 2.7698e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 215.01, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4965 re_mapping 0.0046 re_causal 0.0137 /// teacc 99.07 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.3010, 0.0801, -0.1684, ..., -0.0664, -0.2530, -0.1717], + [ 0.0006, 0.1094, -0.1231, ..., -0.1211, -0.0674, 0.1214], + [ 0.0527, -0.1335, -0.1706, ..., 0.0081, -0.0553, -0.1059], + ..., + [ 0.1000, -0.0689, 0.1418, ..., 0.0700, 0.2004, -0.0184], + [ 0.1169, -0.2236, -0.1346, ..., -0.2947, -0.0722, 0.2029], + [-0.1740, 0.0887, 0.0613, ..., -0.2804, -0.1495, -0.0517]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, 1.3420e-06, 0.0000e+00, ..., 9.3132e-09, + 9.3132e-10, 0.0000e+00], + [ 7.2643e-08, 9.3132e-09, 0.0000e+00, ..., 1.1176e-08, + 5.5879e-09, 4.2841e-08], + [-8.8662e-07, 1.8626e-08, 0.0000e+00, ..., -1.4622e-07, + -1.2107e-08, 1.8626e-09], + ..., + [ 2.0489e-08, 5.5879e-09, 0.0000e+00, ..., 3.7253e-09, + 1.8626e-09, 7.4506e-09], + [ 3.5949e-07, 2.1420e-08, 0.0000e+00, ..., 6.4261e-08, + 2.7940e-09, -6.7055e-08], + [ 5.5879e-09, 3.5390e-08, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0202, -0.0190, -0.0082, -0.0237, -0.0019, 0.0056, 0.0100, 0.0230, + 0.0139, -0.0082], device='cuda:0'), grad: tensor([ 5.0440e-06, 2.2352e-07, -2.5127e-06, 8.6427e-07, 3.9116e-08, + 9.4995e-08, -5.1148e-06, 7.9162e-08, 1.1539e-06, 1.1269e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 214.72, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4979 re_mapping 0.0045 re_causal 0.0128 /// teacc 98.97 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.3014, 0.0801, -0.1697, ..., -0.0665, -0.2531, -0.1718], + [ 0.0006, 0.1095, -0.1232, ..., -0.1214, -0.0674, 0.1216], + [ 0.0528, -0.1335, -0.1706, ..., 0.0083, -0.0552, -0.1060], + ..., + [ 0.1000, -0.0695, 0.1418, ..., 0.0699, 0.2005, -0.0188], + [ 0.1169, -0.2242, -0.1347, ..., -0.2956, -0.0724, 0.2032], + [-0.1743, 0.0900, 0.0636, ..., -0.2809, -0.1497, -0.0522]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.9558e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.9278e-07, 8.1956e-08, 2.6822e-07, ..., 4.2841e-08, + 7.3574e-08, -2.5146e-08], + [-4.3772e-08, 1.3039e-08, 1.0245e-08, ..., -3.4459e-08, + 1.8626e-09, 1.8626e-09], + ..., + [-1.5181e-07, 1.6764e-07, 1.5367e-07, ..., -7.4506e-09, + -6.0536e-08, 2.7008e-08], + [-1.6205e-07, 5.5879e-08, 3.6322e-08, ..., 0.0000e+00, + 0.0000e+00, -3.9954e-07], + [ 1.8626e-09, 4.1239e-06, -5.8208e-07, ..., 0.0000e+00, + -1.5832e-08, 1.2945e-07]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0202, -0.0190, -0.0081, -0.0266, -0.0043, 0.0079, 0.0100, 0.0230, + 0.0135, -0.0070], device='cuda:0'), grad: tensor([-3.4459e-08, 1.0366e-06, -1.3597e-07, 3.2596e-07, -2.1279e-05, + 2.7381e-07, 9.9652e-07, 7.7859e-07, -7.6089e-07, 1.8835e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 214.49, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4498 re_mapping 0.0046 re_causal 0.0127 /// teacc 99.04 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.3020, 0.0801, -0.1703, ..., -0.0670, -0.2538, -0.1718], + [ 0.0007, 0.1095, -0.1230, ..., -0.1218, -0.0675, 0.1217], + [ 0.0528, -0.1336, -0.1709, ..., 0.0083, -0.0553, -0.1058], + ..., + [ 0.1000, -0.0698, 0.1420, ..., 0.0700, 0.2007, -0.0193], + [ 0.1171, -0.2245, -0.1348, ..., -0.2962, -0.0726, 0.2037], + [-0.1752, 0.0899, 0.0630, ..., -0.2815, -0.1506, -0.0525]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.4506e-09, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.7695e-08, 1.8626e-09, 1.7695e-08, ..., 1.2107e-08, + 1.4901e-08, 0.0000e+00], + [ 3.4459e-08, 0.0000e+00, 2.7008e-08, ..., 1.8626e-08, + 3.1665e-08, 9.3132e-09], + ..., + [-9.3132e-08, 1.8626e-09, -4.6566e-08, ..., -3.0734e-08, + -8.7544e-08, -1.3039e-08], + [ 2.7940e-09, -0.0000e+00, 1.3970e-08, ..., 1.0245e-08, + 1.6764e-08, -1.3970e-08], + [ 1.1176e-08, 2.7940e-09, 3.7253e-09, ..., 4.6566e-09, + 8.3819e-09, 5.5879e-09]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0202, -0.0191, -0.0070, -0.0267, -0.0041, 0.0079, 0.0101, 0.0229, + 0.0134, -0.0073], device='cuda:0'), grad: tensor([-1.2107e-08, 4.8429e-08, 9.1270e-08, -6.8918e-08, 3.2596e-08, + 5.5879e-08, -1.5832e-08, -1.7229e-07, 9.3132e-09, 3.3528e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 214.54, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4603 re_mapping 0.0045 re_causal 0.0131 /// teacc 99.08 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.3023, 0.0801, -0.1710, ..., -0.0674, -0.2540, -0.1721], + [ 0.0006, 0.1095, -0.1231, ..., -0.1220, -0.0676, 0.1219], + [ 0.0526, -0.1337, -0.1712, ..., 0.0082, -0.0556, -0.1059], + ..., + [ 0.1001, -0.0701, 0.1420, ..., 0.0693, 0.2008, -0.0196], + [ 0.1176, -0.2248, -0.1349, ..., -0.2968, -0.0729, 0.2040], + [-0.1754, 0.0898, 0.0631, ..., -0.2822, -0.1509, -0.0527]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -1.9185e-07, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-09], + [ 7.4506e-09, 6.5193e-09, 9.3132e-09, ..., 9.3132e-10, + 4.6566e-09, -7.4506e-09], + [ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [-2.5146e-08, 7.4506e-09, -2.4214e-08, ..., -3.7253e-09, + -1.4901e-08, 1.2107e-08], + [-5.4017e-08, 1.6764e-08, 1.3970e-08, ..., 9.3132e-10, + 3.7253e-09, -2.4773e-07], + [ 5.4017e-08, 8.3819e-09, -4.7497e-08, ..., 0.0000e+00, + 2.7940e-09, 2.1420e-07]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0202, -0.0194, -0.0063, -0.0265, -0.0042, 0.0074, 0.0101, 0.0228, + 0.0138, -0.0074], device='cuda:0'), grad: tensor([-5.6811e-07, 8.9407e-08, -2.5146e-08, 2.7940e-08, 1.1921e-07, + 4.1910e-08, 3.4645e-07, 3.7253e-09, -6.8452e-07, 6.4541e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 214.75, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4937 re_mapping 0.0045 re_causal 0.0130 /// teacc 98.94 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.3026, 0.0802, -0.1721, ..., -0.0682, -0.2550, -0.1729], + [ 0.0006, 0.1095, -0.1232, ..., -0.1226, -0.0677, 0.1220], + [ 0.0540, -0.1337, -0.1705, ..., 0.0098, -0.0533, -0.1058], + ..., + [ 0.0999, -0.0706, 0.1419, ..., 0.0687, 0.2008, -0.0199], + [ 0.1178, -0.2255, -0.1351, ..., -0.2983, -0.0734, 0.2044], + [-0.1758, 0.0897, 0.0632, ..., -0.2831, -0.1514, -0.0531]], + device='cuda:0'), grad: tensor([[-3.8184e-08, 4.5635e-08, 8.4750e-08, ..., 0.0000e+00, + 2.7940e-09, 1.8626e-09], + [ 3.7253e-08, 6.1188e-07, 1.9651e-07, ..., 9.3132e-10, + 4.3772e-08, -5.5879e-09], + [ 1.3039e-08, 3.5204e-07, 5.4017e-08, ..., -2.7940e-09, + 9.3132e-09, 6.5193e-09], + ..., + [-5.2154e-08, 2.4103e-06, 5.5600e-07, ..., 9.3132e-10, + -6.9849e-08, -7.4506e-09], + [ 7.4506e-09, 2.7381e-07, 7.5437e-08, ..., 0.0000e+00, + 4.6566e-09, 3.7253e-09], + [ 1.7695e-08, 1.4353e-04, 3.6299e-05, ..., 0.0000e+00, + 3.7253e-09, 3.6322e-08]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0202, -0.0195, -0.0050, -0.0275, -0.0038, 0.0073, 0.0102, 0.0226, + 0.0137, -0.0076], device='cuda:0'), grad: tensor([ 2.7940e-09, 1.6848e-06, 9.5181e-07, 4.7032e-07, -3.8767e-04, + 1.1865e-06, -8.4285e-07, 6.2361e-06, 2.2445e-06, 3.7575e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 214.67, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4706 re_mapping 0.0046 re_causal 0.0127 /// teacc 98.95 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.3021, 0.0803, -0.1733, ..., -0.0686, -0.2559, -0.1720], + [ 0.0006, 0.1096, -0.1232, ..., -0.1229, -0.0677, 0.1224], + [ 0.0543, -0.1340, -0.1723, ..., 0.0078, -0.0532, -0.1059], + ..., + [ 0.0999, -0.0719, 0.1419, ..., 0.0686, 0.2008, -0.0208], + [ 0.1182, -0.2261, -0.1353, ..., -0.2994, -0.0736, 0.2053], + [-0.1764, 0.0892, 0.0629, ..., -0.2835, -0.1513, -0.0536]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 6.5193e-09, -1.3970e-08, 1.8626e-09, ..., 3.7253e-09, + 1.8626e-09, -5.8673e-08], + [ 0.0000e+00, 9.3132e-10, -0.0000e+00, ..., -5.5879e-09, + -5.5879e-09, 1.0245e-08], + ..., + [ 9.3132e-09, 1.2107e-08, 3.7253e-09, ..., 2.7940e-09, + 2.7940e-09, 5.7742e-08], + [-1.1548e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.3784e-07], + [ 8.2888e-08, -1.8626e-09, -1.0245e-08, ..., 0.0000e+00, + 9.3132e-10, 1.0338e-07]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0201, -0.0197, -0.0063, -0.0256, -0.0028, 0.0072, 0.0102, 0.0224, + 0.0139, -0.0082], device='cuda:0'), grad: tensor([-3.7253e-09, -8.2888e-08, -1.8626e-09, -2.7940e-09, 1.9558e-08, + 3.6322e-08, 1.3970e-08, 1.2293e-07, -4.0140e-07, 2.9989e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 214.55, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4845 re_mapping 0.0045 re_causal 0.0130 /// teacc 98.99 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.3029, 0.0803, -0.1746, ..., -0.0691, -0.2565, -0.1723], + [ 0.0006, 0.1095, -0.1233, ..., -0.1231, -0.0677, 0.1225], + [ 0.0552, -0.1341, -0.1733, ..., 0.0067, -0.0530, -0.1058], + ..., + [ 0.0998, -0.0723, 0.1419, ..., 0.0685, 0.2008, -0.0213], + [ 0.1187, -0.2265, -0.1351, ..., -0.3001, -0.0742, 0.2063], + [-0.1766, 0.0900, 0.0647, ..., -0.2839, -0.1511, -0.0540]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, -1.3039e-08, 5.5879e-09, ..., 9.3132e-10, + 2.7940e-09, -2.3283e-08], + [ 2.6077e-08, 6.5193e-09, 2.0489e-08, ..., 1.1176e-08, + 1.5832e-08, 6.5193e-09], + ..., + [-3.2596e-08, 1.6764e-08, -1.5832e-08, ..., -1.3970e-08, + -1.9558e-08, 1.4901e-08], + [-7.4506e-09, 2.7940e-09, 5.5879e-09, ..., 9.3132e-10, + 9.3132e-10, -9.3132e-09], + [ 5.5879e-09, -8.6613e-08, -1.5367e-07, ..., 0.0000e+00, + -1.3039e-08, 5.5879e-09]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0201, -0.0198, -0.0071, -0.0245, -0.0044, 0.0070, 0.0102, 0.0222, + 0.0141, -0.0072], device='cuda:0'), grad: tensor([-2.3283e-08, -4.0978e-08, 6.0536e-08, 7.4506e-09, 2.9616e-07, + 8.3819e-09, -2.2352e-08, 7.4506e-09, -5.5879e-09, -2.6915e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 215.25, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4635 re_mapping 0.0044 re_causal 0.0127 /// teacc 99.04 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.3033, 0.0806, -0.1751, ..., -0.0692, -0.2568, -0.1725], + [ 0.0007, 0.1099, -0.1232, ..., -0.1235, -0.0677, 0.1232], + [ 0.0553, -0.1343, -0.1734, ..., 0.0067, -0.0531, -0.1058], + ..., + [ 0.0996, -0.0745, 0.1417, ..., 0.0683, 0.2008, -0.0226], + [ 0.1190, -0.2273, -0.1354, ..., -0.3011, -0.0745, 0.2068], + [-0.1769, 0.0896, 0.0649, ..., -0.2845, -0.1515, -0.0544]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.4249e-07, 3.7253e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 4.4703e-08, 4.6566e-09, 3.8184e-08, ..., 2.3283e-08, + 2.4214e-08, -8.3819e-09], + [-1.8626e-08, 9.3132e-10, 9.1270e-08, ..., -5.5879e-08, + -4.8429e-08, 9.3132e-09], + ..., + [-4.0978e-08, 5.5879e-09, -1.2107e-07, ..., 1.1176e-08, + -1.2107e-08, 6.5193e-09], + [-2.9802e-08, 1.3970e-08, 1.7695e-08, ..., 1.8626e-09, + 2.7940e-09, -4.0978e-08], + [ 6.5193e-09, 5.9605e-08, -3.3528e-08, ..., 2.7940e-09, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0198, -0.0197, -0.0071, -0.0244, -0.0042, 0.0072, 0.0101, 0.0219, + 0.0140, -0.0078], device='cuda:0'), grad: tensor([-3.3434e-07, 1.4529e-07, -3.0827e-07, 6.7987e-08, 9.0338e-08, + -3.8464e-07, 4.5728e-07, 1.4994e-07, -4.6566e-09, 1.3132e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 215.04, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4651 re_mapping 0.0043 re_causal 0.0126 /// teacc 98.96 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.3035, 0.0808, -0.1763, ..., -0.0694, -0.2570, -0.1723], + [ 0.0006, 0.1100, -0.1233, ..., -0.1241, -0.0679, 0.1232], + [ 0.0549, -0.1344, -0.1737, ..., 0.0067, -0.0536, -0.1059], + ..., + [ 0.0998, -0.0750, 0.1421, ..., 0.0687, 0.2011, -0.0228], + [ 0.1198, -0.2276, -0.1357, ..., -0.3020, -0.0751, 0.2082], + [-0.1774, 0.0893, 0.0649, ..., -0.2856, -0.1522, -0.0546]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 2.7940e-09, 9.3132e-10], + [ 3.3528e-08, 3.5390e-08, 7.0781e-08, ..., 1.0245e-08, + 2.0489e-08, 2.9802e-08], + [ 1.2107e-08, 1.7695e-08, 1.8626e-08, ..., 1.8626e-09, + 6.5193e-09, 1.0245e-08], + ..., + [-9.5926e-08, 6.7055e-08, 5.1223e-08, ..., -3.7253e-09, + -7.4506e-08, 7.7300e-08], + [-3.2596e-08, 1.9558e-08, 4.8429e-08, ..., 2.0489e-08, + 2.7940e-09, -7.9162e-08], + [ 2.9802e-08, 1.5460e-07, -1.1828e-07, ..., 5.5879e-09, + 2.1420e-08, 1.9372e-07]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0197, -0.0199, -0.0072, -0.0244, -0.0042, 0.0072, 0.0101, 0.0221, + 0.0149, -0.0081], device='cuda:0'), grad: tensor([ 1.3039e-08, 3.2410e-07, 9.8720e-08, 2.2911e-07, -8.5961e-07, + -5.7835e-07, 1.1548e-07, 2.6636e-07, -2.2445e-07, 6.2399e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 215.17, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4850 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.04 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.3040, 0.0810, -0.1767, ..., -0.0695, -0.2574, -0.1725], + [ 0.0003, 0.1099, -0.1237, ..., -0.1255, -0.0683, 0.1233], + [ 0.0546, -0.1345, -0.1741, ..., 0.0067, -0.0541, -0.1060], + ..., + [ 0.1002, -0.0755, 0.1426, ..., 0.0694, 0.2016, -0.0230], + [ 0.1203, -0.2282, -0.1358, ..., -0.3027, -0.0755, 0.2089], + [-0.1778, 0.0894, 0.0657, ..., -0.2859, -0.1522, -0.0548]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 3.7253e-09, 2.0489e-08], + [-3.7253e-09, -1.6764e-08, 2.7940e-09, ..., 2.7940e-09, + 2.7940e-09, -8.1025e-08], + [-5.0291e-08, 9.3132e-10, 0.0000e+00, ..., -3.3528e-08, + -4.5635e-08, 2.0489e-08], + ..., + [ 4.0978e-08, 1.0990e-07, -1.2107e-08, ..., 2.2352e-08, + 9.3132e-09, 1.6112e-07], + [-2.5146e-08, 1.8626e-09, 2.7940e-09, ..., 3.7253e-09, + 4.6566e-09, -6.3330e-08], + [ 7.4506e-09, 1.0245e-08, -2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 2.7008e-08]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0195, -0.0200, -0.0072, -0.0244, -0.0046, 0.0070, 0.0100, 0.0222, + 0.0151, -0.0080], device='cuda:0'), grad: tensor([ 6.4261e-08, -1.0990e-07, -1.5274e-07, 6.5193e-09, -2.4028e-07, + 1.2107e-08, 8.3819e-09, 4.5169e-07, -9.5926e-08, 6.1467e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 214.92, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.5080 re_mapping 0.0043 re_causal 0.0132 /// teacc 99.00 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.3015, 0.0813, -0.1762, ..., -0.0698, -0.2578, -0.1703], + [ 0.0011, 0.1100, -0.1220, ..., -0.1259, -0.0683, 0.1242], + [ 0.0546, -0.1346, -0.1742, ..., 0.0067, -0.0543, -0.1060], + ..., + [ 0.0995, -0.0761, 0.1409, ..., 0.0696, 0.2017, -0.0244], + [ 0.1209, -0.2287, -0.1361, ..., -0.3033, -0.0759, 0.2096], + [-0.1778, 0.0891, 0.0658, ..., -0.2862, -0.1522, -0.0548]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 9.3132e-10], + [ 1.2107e-08, -1.5832e-08, 1.3039e-08, ..., 1.1176e-08, + 5.5879e-09, -8.1025e-08], + [-3.5949e-07, 3.7253e-09, 1.8626e-08, ..., -1.4529e-07, + -2.7101e-07, 1.3039e-08], + ..., + [ 3.5390e-07, 1.2107e-08, 8.3819e-09, ..., 1.6019e-07, + 2.6636e-07, 3.1665e-08], + [-2.0489e-08, 1.3039e-08, 4.6566e-09, ..., 4.6566e-09, + 2.7940e-09, -1.4901e-08], + [ 9.3132e-10, 2.7008e-08, 3.7253e-09, ..., 5.5879e-09, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0190, -0.0190, -0.0071, -0.0244, -0.0043, 0.0067, 0.0095, 0.0210, + 0.0155, -0.0083], device='cuda:0'), grad: tensor([-1.3039e-08, -1.5274e-07, -7.9256e-07, -1.7323e-07, -8.6613e-08, + 2.8871e-08, 6.7055e-08, 9.7137e-07, 3.0734e-08, 1.2200e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 215.04, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4911 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.03 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.3022, 0.0815, -0.1766, ..., -0.0701, -0.2583, -0.1709], + [ 0.0008, 0.1097, -0.1222, ..., -0.1274, -0.0687, 0.1242], + [ 0.0542, -0.1347, -0.1750, ..., 0.0066, -0.0558, -0.1059], + ..., + [ 0.0999, -0.0765, 0.1413, ..., 0.0705, 0.2024, -0.0247], + [ 0.1217, -0.2291, -0.1365, ..., -0.3065, -0.0774, 0.2105], + [-0.1782, 0.0888, 0.0657, ..., -0.2868, -0.1528, -0.0550]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-1.8626e-09, 1.8626e-09, 9.3132e-09, ..., 9.3132e-10, + 0.0000e+00, -1.8626e-09], + [ 9.3132e-10, 2.7940e-08, -5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0188, -0.0192, -0.0072, -0.0242, -0.0042, 0.0061, 0.0094, 0.0212, + 0.0163, -0.0087], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.5832e-08, -3.4459e-08, -1.1176e-08, -1.2759e-07, + 1.9558e-08, -2.3283e-08, 1.9558e-08, 2.6077e-08, 1.1735e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 214.90, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4990 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.00 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.3025, 0.0816, -0.1773, ..., -0.0708, -0.2587, -0.1715], + [-0.0004, 0.1102, -0.1218, ..., -0.1278, -0.0690, 0.1236], + [ 0.0540, -0.1348, -0.1751, ..., 0.0066, -0.0560, -0.1063], + ..., + [ 0.1001, -0.0804, 0.1415, ..., 0.0707, 0.2030, -0.0271], + [ 0.1248, -0.2262, -0.1368, ..., -0.3076, -0.0781, 0.2138], + [-0.1798, 0.0885, 0.0648, ..., -0.2878, -0.1557, -0.0570]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -5.9605e-08, -9.3132e-10, ..., 2.7940e-09, + 1.8626e-09, 4.6566e-09], + [-1.8626e-09, 4.7497e-08, -1.0990e-07, ..., 4.6566e-09, + 2.2352e-08, -2.1048e-07], + [-9.3132e-09, 1.5832e-08, 8.3819e-09, ..., -1.6764e-08, + -8.3819e-09, 1.4901e-08], + ..., + [-1.0058e-07, 4.6566e-08, -7.3574e-08, ..., -1.8626e-09, + -1.0058e-07, 1.5553e-07], + [-9.3132e-10, 1.6764e-08, 1.5832e-08, ..., 2.7940e-09, + 2.7940e-09, 1.3039e-08], + [ 1.0151e-07, 1.3318e-06, 1.5926e-07, ..., 7.4506e-09, + 8.0094e-08, 6.9849e-08]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0188, -0.0200, -0.0074, -0.0242, -0.0042, 0.0060, 0.0093, 0.0210, + 0.0194, -0.0092], device='cuda:0'), grad: tensor([-1.5553e-07, -8.3353e-07, -2.0489e-08, 6.9849e-08, -5.1558e-06, + 4.3772e-08, 1.1269e-07, 6.9477e-07, 1.4901e-07, 5.0738e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 214.92, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4852 re_mapping 0.0047 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.3025, 0.0816, -0.1779, ..., -0.0719, -0.2591, -0.1711], + [-0.0009, 0.1101, -0.1224, ..., -0.1285, -0.0696, 0.1235], + [ 0.0539, -0.1351, -0.1755, ..., 0.0065, -0.0566, -0.1065], + ..., + [ 0.1009, -0.0805, 0.1424, ..., 0.0714, 0.2037, -0.0266], + [ 0.1247, -0.2265, -0.1374, ..., -0.3088, -0.0790, 0.2138], + [-0.1808, 0.0882, 0.0643, ..., -0.2893, -0.1568, -0.0579]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.5832e-08, 1.7695e-08, ..., 2.7940e-09, + 9.3132e-10, 2.7008e-08], + [ 2.4214e-08, 5.6811e-08, 4.1910e-08, ..., 1.2107e-08, + 4.6566e-09, 6.5193e-09], + [ 1.1828e-07, 4.6566e-09, 1.9558e-08, ..., 5.5879e-09, + 8.3819e-09, 2.7474e-07], + ..., + [-1.8626e-08, 6.5193e-09, -1.3970e-08, ..., -4.6566e-09, + -1.6764e-08, 3.4459e-08], + [-2.6990e-06, 3.5390e-08, 2.0489e-08, ..., 4.6566e-09, + 9.3132e-10, -8.3596e-06], + [ 2.0303e-06, -1.2573e-07, -1.3225e-07, ..., 1.0245e-08, + 1.8626e-09, 6.4112e-06]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0187, -0.0202, -0.0074, -0.0245, -0.0034, 0.0061, 0.0095, 0.0214, + 0.0192, -0.0098], device='cuda:0'), grad: tensor([ 1.5460e-07, 3.0734e-07, 9.3412e-07, 2.2259e-07, 4.7497e-08, + -4.0978e-07, 5.2899e-06, 7.2643e-08, -2.7180e-05, 2.0608e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 214.83, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4924 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.08 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.3025, 0.0817, -0.1785, ..., -0.0726, -0.2593, -0.1709], + [-0.0014, 0.1100, -0.1231, ..., -0.1287, -0.0702, 0.1230], + [ 0.0543, -0.1352, -0.1757, ..., 0.0065, -0.0565, -0.1066], + ..., + [ 0.1015, -0.0806, 0.1431, ..., 0.0713, 0.2044, -0.0253], + [ 0.1247, -0.2267, -0.1380, ..., -0.3107, -0.0793, 0.2139], + [-0.1812, 0.0882, 0.0645, ..., -0.2898, -0.1570, -0.0588]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.4214e-08, 8.1025e-08, ..., 1.8626e-09, + 1.8626e-09, 9.3132e-10], + [-4.6566e-09, -2.7940e-09, 1.1269e-07, ..., 3.7253e-09, + 8.3819e-09, -3.5390e-08], + [ 9.3132e-09, 7.4506e-09, 2.9802e-08, ..., -3.8184e-08, + -1.3039e-08, 7.4506e-09], + ..., + [-3.1665e-08, 8.9686e-07, 3.1702e-06, ..., 1.8626e-09, + -3.2596e-08, 2.9802e-08], + [-7.4506e-09, 1.4622e-07, 5.3737e-07, ..., 1.8626e-09, + 9.3132e-10, -8.3819e-09], + [ 2.6077e-08, -1.8431e-06, -6.8024e-06, ..., 1.8626e-09, + 2.6077e-08, 2.7940e-09]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0187, -0.0206, -0.0073, -0.0249, -0.0032, 0.0074, 0.0096, 0.0218, + 0.0190, -0.0101], device='cuda:0'), grad: tensor([ 2.7288e-07, 2.5146e-07, -3.7253e-09, 3.5018e-07, 7.3649e-06, + 6.3609e-07, 7.6368e-07, 1.0349e-05, 1.7053e-06, -2.1681e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 214.96, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4844 re_mapping 0.0043 re_causal 0.0127 /// teacc 99.06 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.3026, 0.0817, -0.1791, ..., -0.0730, -0.2596, -0.1711], + [-0.0014, 0.1103, -0.1232, ..., -0.1290, -0.0703, 0.1237], + [ 0.0546, -0.1353, -0.1759, ..., 0.0065, -0.0564, -0.1067], + ..., + [ 0.1016, -0.0809, 0.1433, ..., 0.0715, 0.2045, -0.0256], + [ 0.1245, -0.2272, -0.1384, ..., -0.3119, -0.0799, 0.2137], + [-0.1814, 0.0882, 0.0649, ..., -0.2901, -0.1573, -0.0596]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, -1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 3.1665e-08], + [ 8.4750e-08, -8.4750e-08, 8.1025e-08, ..., 2.5146e-08, + 8.2888e-08, -1.8720e-07], + [ 1.1455e-07, 2.2352e-08, 1.0245e-07, ..., 3.1665e-08, + 1.1083e-07, 4.9360e-08], + ..., + [-2.6822e-07, 5.5879e-08, -3.1199e-07, ..., -9.5926e-08, + -2.7195e-07, 1.2759e-07], + [-4.8429e-08, 1.8626e-09, 1.5832e-08, ..., 4.6566e-09, + 1.2107e-08, -1.0058e-07], + [ 2.7008e-08, 1.8626e-09, 1.0245e-08, ..., 2.7940e-09, + 8.3819e-09, 3.9116e-08]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0187, -0.0205, -0.0073, -0.0248, -0.0034, 0.0075, 0.0096, 0.0217, + 0.0188, -0.0101], device='cuda:0'), grad: tensor([ 8.6613e-08, -3.2317e-07, 3.6042e-07, 1.6112e-07, -9.3132e-09, + 2.0489e-08, 6.8918e-08, -2.3562e-07, -2.5984e-07, 1.2759e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 215.11, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4963 re_mapping 0.0046 re_causal 0.0135 /// teacc 98.93 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.3030, 0.0817, -0.1797, ..., -0.0739, -0.2608, -0.1713], + [-0.0017, 0.1106, -0.1236, ..., -0.1293, -0.0706, 0.1239], + [ 0.0549, -0.1353, -0.1760, ..., 0.0065, -0.0560, -0.1067], + ..., + [ 0.1020, -0.0813, 0.1438, ..., 0.0714, 0.2049, -0.0256], + [ 0.1247, -0.2273, -0.1387, ..., -0.3128, -0.0815, 0.2141], + [-0.1818, 0.0882, 0.0649, ..., -0.2906, -0.1581, -0.0601]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.9465e-07, 3.4459e-08, ..., 1.0245e-08, + 0.0000e+00, 4.6566e-09], + [ 4.9360e-08, -1.8626e-08, 4.8429e-08, ..., 1.7695e-08, + 3.7253e-09, -8.3819e-08], + [ 9.3132e-09, 2.1420e-08, 3.5390e-08, ..., 1.3970e-08, + 9.3132e-10, 2.9802e-08], + ..., + [-1.2107e-08, 4.0047e-08, 1.8626e-09, ..., 1.8626e-09, + -7.4506e-09, 1.3039e-07], + [-2.1048e-07, 3.8184e-08, 9.9652e-08, ..., 3.3528e-08, + 0.0000e+00, -3.9861e-07], + [ 6.5193e-09, 1.3690e-07, 2.2165e-07, ..., 7.4506e-08, + 9.3132e-10, 9.3132e-08]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0187, -0.0206, -0.0072, -0.0248, -0.0035, 0.0076, 0.0095, 0.0219, + 0.0190, -0.0102], device='cuda:0'), grad: tensor([ 4.9174e-07, 2.1420e-08, 1.5926e-07, -1.0515e-06, -2.1327e-07, + -4.2468e-07, 8.5402e-07, 2.8033e-07, -8.7637e-07, 7.4692e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 215.06, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4661 re_mapping 0.0047 re_causal 0.0127 /// teacc 98.99 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.3029, 0.0818, -0.1802, ..., -0.0748, -0.2618, -0.1713], + [-0.0019, 0.1107, -0.1238, ..., -0.1300, -0.0708, 0.1241], + [ 0.0588, -0.1354, -0.1731, ..., 0.0081, -0.0523, -0.1069], + ..., + [ 0.1000, -0.0817, 0.1427, ..., 0.0685, 0.2035, -0.0261], + [ 0.1246, -0.2274, -0.1390, ..., -0.3175, -0.0840, 0.2142], + [-0.1839, 0.0881, 0.0635, ..., -0.2922, -0.1611, -0.0613]], + device='cuda:0'), grad: tensor([[ 3.6322e-08, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 8.7544e-08], + [-9.6858e-08, -5.5879e-09, 9.3132e-10, ..., 0.0000e+00, + -2.7940e-09, -4.4331e-07], + [ 4.6566e-08, 2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 1.7788e-07], + ..., + [ 1.0245e-08, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 4.2841e-08], + [-4.8429e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 3.3528e-08, 1.0245e-08, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 7.7300e-08]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0186, -0.0207, -0.0049, -0.0254, -0.0030, 0.0073, 0.0095, 0.0203, + 0.0189, -0.0109], device='cuda:0'), grad: tensor([ 1.2945e-07, -1.1204e-06, 4.2375e-07, 3.7253e-09, -2.0489e-07, + 2.2352e-08, 1.6391e-07, 1.4715e-07, 1.5739e-07, 2.7474e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 214.94, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5094 re_mapping 0.0045 re_causal 0.0127 /// teacc 99.03 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.3030, 0.0819, -0.1812, ..., -0.0759, -0.2623, -0.1718], + [-0.0020, 0.1105, -0.1240, ..., -0.1303, -0.0709, 0.1240], + [ 0.0587, -0.1358, -0.1732, ..., 0.0080, -0.0524, -0.1070], + ..., + [ 0.1002, -0.0819, 0.1428, ..., 0.0685, 0.2037, -0.0261], + [ 0.1246, -0.2277, -0.1393, ..., -0.3180, -0.0843, 0.2143], + [-0.1844, 0.0880, 0.0634, ..., -0.2928, -0.1618, -0.0618]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -5.5879e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [-1.1083e-07, -3.5763e-07, -1.9930e-07, ..., 1.8626e-09, + 0.0000e+00, -3.7905e-07], + [-9.3132e-10, 3.7253e-09, 1.8626e-09, ..., -1.7695e-08, + 0.0000e+00, 5.5879e-09], + ..., + [ 9.1270e-08, 2.9709e-07, 1.6671e-07, ..., 9.3132e-10, + 0.0000e+00, 3.0361e-07], + [ 2.7940e-09, 4.6566e-09, 2.7940e-09, ..., 1.2107e-08, + 0.0000e+00, 5.5879e-09], + [ 1.3039e-08, 4.5635e-08, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 4.0047e-08]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0186, -0.0209, -0.0050, -0.0247, -0.0029, 0.0058, 0.0097, 0.0204, + 0.0187, -0.0111], device='cuda:0'), grad: tensor([ 0.0000e+00, -1.3290e-06, -3.2783e-07, 2.2352e-08, 8.3819e-08, + -1.5832e-08, 2.7940e-08, 1.1222e-06, 2.6263e-07, 1.5739e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 215.07, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4748 re_mapping 0.0043 re_causal 0.0121 /// teacc 98.93 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.3034, 0.0820, -0.1817, ..., -0.0752, -0.2630, -0.1721], + [-0.0021, 0.1105, -0.1242, ..., -0.1309, -0.0710, 0.1241], + [ 0.0587, -0.1360, -0.1733, ..., 0.0080, -0.0524, -0.1070], + ..., + [ 0.1004, -0.0820, 0.1409, ..., 0.0686, 0.2027, -0.0264], + [ 0.1249, -0.2278, -0.1396, ..., -0.3194, -0.0842, 0.2145], + [-0.1857, 0.0880, 0.0654, ..., -0.2952, -0.1623, -0.0622]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.8626e-09, 4.6566e-09, ..., 9.3132e-10, + 9.3132e-10, 7.4506e-09], + [ 1.8254e-07, 1.4901e-08, 1.8626e-08, ..., 4.6566e-09, + 8.3819e-09, 2.2538e-07], + [ 8.4843e-07, 1.8626e-09, 1.0245e-08, ..., -6.5193e-09, + -9.3132e-09, 9.0245e-07], + ..., + [ 3.9116e-08, 1.1176e-08, -4.3772e-08, ..., -1.3039e-08, + -2.4214e-08, 9.4995e-08], + [-1.2135e-06, 4.8429e-08, 8.9407e-08, ..., 2.7940e-09, + 4.6566e-09, -1.2703e-06], + [ 2.3283e-08, -4.4703e-08, -1.2852e-07, ..., 4.6566e-09, + 1.0245e-08, 1.7602e-07]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0185, -0.0210, -0.0050, -0.0247, -0.0028, 0.0056, 0.0097, 0.0188, + 0.0188, -0.0094], device='cuda:0'), grad: tensor([ 2.5518e-07, 7.7952e-07, 2.4177e-06, 6.0070e-07, -3.3062e-07, + 1.7928e-06, -4.7088e-06, 2.1327e-07, -1.2983e-06, 2.8126e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 215.18, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4569 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.03 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.3026, 0.0820, -0.1826, ..., -0.0754, -0.2618, -0.1708], + [-0.0021, 0.1106, -0.1243, ..., -0.1314, -0.0712, 0.1245], + [ 0.0587, -0.1363, -0.1734, ..., 0.0080, -0.0525, -0.1077], + ..., + [ 0.1005, -0.0824, 0.1411, ..., 0.0689, 0.2029, -0.0267], + [ 0.1251, -0.2280, -0.1401, ..., -0.3208, -0.0851, 0.2147], + [-0.1861, 0.0879, 0.0653, ..., -0.2965, -0.1625, -0.0625]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.5832e-07, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, -9.3132e-10, 9.3132e-10, ..., 3.7253e-09, + 5.5879e-09, -4.6566e-09], + [-3.6322e-08, 9.3132e-10, -1.1176e-08, ..., -6.1467e-08, + -8.1956e-08, 9.3132e-10], + ..., + [ 3.1665e-08, 9.3132e-10, 1.0245e-08, ..., 5.0291e-08, + 6.7055e-08, 3.7253e-09], + [-1.6764e-08, 1.8626e-09, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, -2.4214e-08], + [ 8.3819e-09, 3.2596e-08, -9.3132e-09, ..., 1.8626e-09, + 2.7940e-09, 1.2107e-08]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0184, -0.0210, -0.0052, -0.0247, -0.0027, 0.0055, 0.0097, 0.0189, + 0.0188, -0.0095], device='cuda:0'), grad: tensor([-5.7276e-07, 5.5879e-09, -1.6391e-07, 6.5193e-09, 2.7940e-08, + 2.1420e-08, 4.3306e-07, 1.4715e-07, -4.4703e-08, 1.4156e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 214.90, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4526 re_mapping 0.0045 re_causal 0.0130 /// teacc 99.02 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.3028, 0.0820, -0.1840, ..., -0.0758, -0.2623, -0.1709], + [-0.0027, 0.1067, -0.1260, ..., -0.1316, -0.0721, 0.1243], + [ 0.0585, -0.1364, -0.1735, ..., 0.0079, -0.0527, -0.1079], + ..., + [ 0.1013, -0.0827, 0.1414, ..., 0.0691, 0.2038, -0.0256], + [ 0.1247, -0.2283, -0.1412, ..., -0.3220, -0.0884, 0.2145], + [-0.1865, 0.0904, 0.0660, ..., -0.2969, -0.1626, -0.0628]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.4065e-06, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 3.7253e-09, 3.1665e-08, 3.7253e-09, ..., 1.8626e-09, + 4.6566e-09, -2.0489e-08], + [-2.7940e-08, 8.7544e-08, 9.3132e-10, ..., -1.5832e-08, + -2.8871e-08, 1.8626e-09], + ..., + [ 2.3283e-08, 4.0047e-08, -9.3132e-09, ..., 1.3039e-08, + 2.0489e-08, 2.1420e-08], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.3772e-08], + [ 3.7253e-09, -2.7288e-06, 6.5193e-09, ..., 1.8626e-09, + 5.5879e-09, 4.6566e-09]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0185, -0.0233, -0.0053, -0.0247, -0.0028, 0.0061, 0.0097, 0.0194, + 0.0184, -0.0076], device='cuda:0'), grad: tensor([ 6.2771e-06, 6.0536e-08, 1.8161e-07, 4.7497e-08, 3.3434e-07, + -3.1851e-07, -3.7625e-07, 3.6135e-07, 3.0734e-07, -6.8992e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 214.74, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4513 re_mapping 0.0046 re_causal 0.0131 /// teacc 98.95 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.3031, 0.0821, -0.1854, ..., -0.0763, -0.2631, -0.1710], + [-0.0026, 0.1060, -0.1264, ..., -0.1324, -0.0723, 0.1251], + [ 0.0585, -0.1365, -0.1736, ..., 0.0079, -0.0526, -0.1085], + ..., + [ 0.1014, -0.0828, 0.1416, ..., 0.0694, 0.2040, -0.0261], + [ 0.1244, -0.2285, -0.1422, ..., -0.3237, -0.0901, 0.2144], + [-0.1868, 0.0909, 0.0662, ..., -0.2979, -0.1625, -0.0638]], + device='cuda:0'), grad: tensor([[ 1.3597e-07, -7.4506e-09, 1.3504e-07, ..., 9.3132e-10, + 1.3225e-07, 2.9802e-08], + [ 6.2466e-05, 1.1176e-08, 5.9694e-05, ..., 9.3132e-10, + 6.0618e-05, 1.2323e-05], + [ 5.6438e-07, 6.5193e-09, 5.4762e-07, ..., -3.7253e-09, + 5.4762e-07, 1.1455e-07], + ..., + [-6.6161e-05, 4.8429e-08, -6.3181e-05, ..., -1.8626e-09, + -6.4254e-05, -1.3031e-05], + [ 1.9912e-06, 1.2107e-08, 1.9129e-06, ..., 9.3132e-10, + 1.9334e-06, 3.9767e-07], + [ 4.7218e-07, 1.4603e-05, 7.9796e-06, ..., 9.3132e-10, + 4.5821e-07, 5.4985e-06]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0184, -0.0234, -0.0053, -0.0229, -0.0025, 0.0031, 0.0097, 0.0193, + 0.0180, -0.0072], device='cuda:0'), grad: tensor([ 9.7007e-06, 1.3101e-04, 1.2638e-06, 8.3726e-07, -6.3658e-05, + 2.3842e-07, -1.0438e-05, -1.3816e-04, 4.8093e-06, 6.4254e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 215.02, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4655 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.07 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.3037, 0.0821, -0.1861, ..., -0.0781, -0.2648, -0.1711], + [-0.0035, 0.1060, -0.1273, ..., -0.1336, -0.0732, 0.1247], + [ 0.0586, -0.1366, -0.1737, ..., 0.0079, -0.0525, -0.1079], + ..., + [ 0.1022, -0.0825, 0.1422, ..., 0.0699, 0.2048, -0.0257], + [ 0.1244, -0.2287, -0.1428, ..., -0.3245, -0.0919, 0.2146], + [-0.1876, 0.0908, 0.0661, ..., -0.3014, -0.1627, -0.0644]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.3027e-07, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 9.3132e-10, 6.5193e-09, 1.2107e-08, ..., 3.7253e-09, + 1.8626e-09, -6.5193e-09], + [ 4.6566e-09, 9.3132e-09, 3.7253e-09, ..., 3.7253e-09, + 9.3132e-10, 8.3819e-09], + ..., + [-4.6566e-09, 1.1176e-08, 6.5193e-08, ..., -9.3132e-10, + -1.8626e-09, 5.5879e-09], + [-4.6566e-09, 2.4214e-08, 1.8626e-08, ..., 9.3132e-10, + 0.0000e+00, -9.3132e-09], + [ 4.6566e-09, 3.3900e-07, -1.5181e-07, ..., 1.8626e-09, + -1.8626e-09, 2.7940e-09]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0185, -0.0239, -0.0052, -0.0229, -0.0022, 0.0029, 0.0098, 0.0198, + 0.0180, -0.0074], device='cuda:0'), grad: tensor([-1.2936e-06, 3.2596e-08, -1.8813e-07, 8.2888e-08, 3.6042e-07, + 2.2352e-08, -8.6613e-08, 1.0803e-07, 1.1083e-07, 8.6054e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 215.04, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4804 re_mapping 0.0043 re_causal 0.0127 /// teacc 99.02 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.3043, 0.0822, -0.1865, ..., -0.0792, -0.2662, -0.1718], + [-0.0033, 0.1060, -0.1275, ..., -0.1339, -0.0732, 0.1253], + [ 0.0582, -0.1366, -0.1739, ..., 0.0078, -0.0531, -0.1082], + ..., + [ 0.1025, -0.0829, 0.1423, ..., 0.0700, 0.2052, -0.0258], + [ 0.1242, -0.2289, -0.1431, ..., -0.3251, -0.0922, 0.2143], + [-0.1881, 0.0909, 0.0663, ..., -0.3022, -0.1628, -0.0648]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 2.7940e-09, ..., 2.7940e-09, + 0.0000e+00, 9.3132e-10], + [ 1.2107e-08, -2.7940e-09, 4.0978e-08, ..., 2.7940e-08, + 1.0245e-08, -7.3574e-08], + [ 1.7695e-08, 9.3132e-10, 1.2293e-07, ..., 1.3039e-07, + 2.7940e-09, 3.7253e-09], + ..., + [-3.4459e-08, 4.6566e-09, -1.7695e-08, ..., 2.1420e-08, + -3.6322e-08, 5.6811e-08], + [ 3.2596e-08, 1.2107e-08, 2.5518e-07, ..., 2.3562e-07, + 9.3132e-10, 4.6566e-09], + [ 2.3283e-08, -1.1176e-08, -4.2841e-08, ..., 1.2107e-08, + 1.8626e-08, 1.8626e-09]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0188, -0.0236, -0.0054, -0.0229, -0.0028, 0.0006, 0.0114, 0.0198, + 0.0174, -0.0073], device='cuda:0'), grad: tensor([-1.2107e-08, -2.5146e-08, 3.5018e-07, -1.1353e-06, 4.5635e-08, + 3.7253e-08, 1.4901e-08, 9.6858e-08, 7.4413e-07, -1.1735e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 215.15, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4756 re_mapping 0.0042 re_causal 0.0123 /// teacc 98.97 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.3049, 0.0823, -0.1872, ..., -0.0808, -0.2679, -0.1719], + [-0.0033, 0.1061, -0.1275, ..., -0.1341, -0.0733, 0.1255], + [ 0.0582, -0.1368, -0.1739, ..., 0.0079, -0.0530, -0.1084], + ..., + [ 0.1025, -0.0832, 0.1424, ..., 0.0700, 0.2053, -0.0261], + [ 0.1244, -0.2290, -0.1433, ..., -0.3259, -0.0925, 0.2145], + [-0.1884, 0.0909, 0.0663, ..., -0.3028, -0.1629, -0.0654]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.8626e-09, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 2.4214e-08, -2.2352e-08, 2.8871e-08, ..., 7.4506e-09, + 1.9558e-08, -1.3039e-08], + [ 2.7940e-09, 9.3132e-10, 6.5193e-09, ..., -5.5879e-09, + 5.5879e-09, 3.7253e-09], + ..., + [-3.6322e-08, 3.3528e-08, -7.0781e-08, ..., -1.4901e-08, + -6.0536e-08, 6.1467e-08], + [-6.7055e-08, 6.5193e-09, 7.4506e-09, ..., 9.3132e-10, + 1.8626e-09, -5.8115e-07], + [ 2.7940e-08, 8.1304e-07, 4.3306e-07, ..., 6.5193e-09, + 2.8871e-08, 7.2271e-07]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0188, -0.0235, -0.0055, -0.0229, -0.0028, 0.0006, 0.0113, 0.0197, + 0.0175, -0.0074], device='cuda:0'), grad: tensor([ 2.8871e-08, 7.7300e-08, -3.7253e-09, 1.4994e-07, -4.2021e-06, + -1.6764e-07, 8.2329e-07, 5.5879e-09, -9.6951e-07, 4.2580e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 214.86, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4790 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.03 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.3061, 0.0823, -0.1880, ..., -0.0815, -0.2687, -0.1731], + [-0.0033, 0.1060, -0.1276, ..., -0.1334, -0.0729, 0.1256], + [ 0.0583, -0.1369, -0.1741, ..., 0.0077, -0.0530, -0.1085], + ..., + [ 0.1026, -0.0835, 0.1424, ..., 0.0694, 0.2051, -0.0262], + [ 0.1246, -0.2292, -0.1436, ..., -0.3269, -0.0930, 0.2149], + [-0.1888, 0.0908, 0.0664, ..., -0.3045, -0.1629, -0.0660]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 8.3819e-09, 1.8626e-09, 7.9162e-08, ..., 7.3574e-08, + 1.5832e-08, 1.8626e-08], + [ 3.3528e-08, 9.3132e-10, 1.7323e-07, ..., 1.6391e-07, + 3.5390e-08, 3.9116e-08], + ..., + [-3.7253e-09, 0.0000e+00, 8.1956e-08, ..., 7.9162e-08, + 1.3970e-08, 3.7253e-09], + [-6.9849e-08, 1.0245e-08, 6.5193e-09, ..., 3.7253e-09, + 1.8626e-09, -1.1176e-07], + [ 7.4506e-09, 0.0000e+00, -5.5879e-09, ..., 3.7253e-09, + 9.3132e-10, 2.2352e-08]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0188, -0.0233, -0.0056, -0.0232, -0.0026, 0.0011, 0.0113, 0.0195, + 0.0176, -0.0075], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.4680e-07, 5.6252e-07, -8.9034e-07, -2.7940e-09, + -1.7229e-07, 1.6671e-07, 2.1979e-07, -2.0489e-07, 5.8673e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 214.96, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4694 re_mapping 0.0041 re_causal 0.0123 /// teacc 98.93 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.3063, 0.0824, -0.1886, ..., -0.0823, -0.2689, -0.1735], + [-0.0034, 0.1061, -0.1277, ..., -0.1316, -0.0722, 0.1258], + [ 0.0583, -0.1369, -0.1746, ..., 0.0076, -0.0532, -0.1086], + ..., + [ 0.1027, -0.0838, 0.1426, ..., 0.0684, 0.2047, -0.0262], + [ 0.1248, -0.2295, -0.1439, ..., -0.3278, -0.0934, 0.2151], + [-0.1891, 0.0908, 0.0665, ..., -0.3053, -0.1630, -0.0663]], + device='cuda:0'), grad: tensor([[ 3.6322e-08, -1.2107e-08, 6.5193e-09, ..., 0.0000e+00, + 9.3132e-10, 2.8778e-07], + [-2.4159e-06, -9.8720e-08, -2.9150e-07, ..., 2.7940e-09, + 4.6566e-09, -1.9997e-05], + [ 1.5637e-06, 1.8626e-09, 2.0768e-07, ..., -1.6764e-08, + -7.4506e-09, 1.2740e-05], + ..., + [ 7.0781e-07, 1.9558e-08, 3.1479e-07, ..., 1.3039e-08, + 4.0047e-08, 5.6550e-06], + [ 2.2352e-08, 1.3039e-08, 2.6077e-08, ..., 1.8626e-09, + 3.7253e-09, 2.3842e-07], + [ 8.3819e-09, -6.5193e-08, -3.6508e-07, ..., 9.3132e-10, + -5.3085e-08, 5.4948e-08]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0188, -0.0229, -0.0055, -0.0232, -0.0029, 0.0013, 0.0112, 0.0192, + 0.0176, -0.0075], device='cuda:0'), grad: tensor([ 5.5227e-07, -4.0442e-05, 2.5645e-05, 1.6950e-07, 1.1250e-06, + -1.0803e-07, 1.3076e-06, 1.1928e-05, 5.7835e-07, -7.9721e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 215.09, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4769 re_mapping 0.0041 re_causal 0.0123 /// teacc 98.98 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.3065, 0.0824, -0.1892, ..., -0.0835, -0.2694, -0.1737], + [-0.0026, 0.1065, -0.1274, ..., -0.1296, -0.0707, 0.1274], + [ 0.0580, -0.1370, -0.1751, ..., 0.0073, -0.0538, -0.1092], + ..., + [ 0.1024, -0.0851, 0.1427, ..., 0.0672, 0.2040, -0.0277], + [ 0.1249, -0.2304, -0.1443, ..., -0.3292, -0.0937, 0.2147], + [-0.1894, 0.0908, 0.0665, ..., -0.3061, -0.1630, -0.0667]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.2107e-08, 1.2107e-08, ..., 1.4901e-08, + 2.7940e-09, -1.8626e-09], + [ 1.0338e-07, 3.7253e-09, 1.3597e-07, ..., 7.3574e-08, + 1.0710e-07, 6.2399e-08], + [ 8.6613e-08, 9.3132e-10, 7.3574e-08, ..., 5.0291e-08, + 4.1910e-08, 8.5682e-08], + ..., + [-3.6787e-07, 4.6566e-09, -8.9500e-07, ..., -6.3051e-07, + -5.8953e-07, 1.8626e-08], + [-2.7101e-07, 6.5193e-09, 9.3132e-09, ..., 8.3819e-09, + 3.7253e-09, -3.7532e-07], + [ 4.5635e-08, 9.1270e-08, 8.0094e-08, ..., 3.7253e-09, + 9.2201e-08, 3.4459e-08]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0187, -0.0213, -0.0058, -0.0232, -0.0029, 0.0009, 0.0113, 0.0182, + 0.0172, -0.0076], device='cuda:0'), grad: tensor([ 0.0000e+00, 3.9674e-07, 3.2596e-07, 8.8289e-07, -2.1607e-07, + 1.7695e-07, 2.9057e-07, -1.2862e-06, -9.4436e-07, 3.7905e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 214.49, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4473 re_mapping 0.0042 re_causal 0.0121 /// teacc 98.98 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.3067, 0.0825, -0.1899, ..., -0.0836, -0.2698, -0.1735], + [-0.0024, 0.1065, -0.1276, ..., -0.1290, -0.0702, 0.1278], + [ 0.0580, -0.1375, -0.1752, ..., 0.0072, -0.0540, -0.1099], + ..., + [ 0.1024, -0.0853, 0.1428, ..., 0.0665, 0.2038, -0.0278], + [ 0.1249, -0.2308, -0.1447, ..., -0.3302, -0.0941, 0.2148], + [-0.1898, 0.0908, 0.0666, ..., -0.3075, -0.1631, -0.0671]], + device='cuda:0'), grad: tensor([[-9.3132e-10, -9.9186e-07, 7.4506e-09, ..., 2.0489e-08, + 0.0000e+00, -1.8626e-08], + [ 4.5635e-08, 2.7940e-08, 6.7055e-08, ..., 7.3574e-08, + 5.6811e-08, 1.0245e-08], + [ 1.8626e-09, 2.6356e-07, 2.0768e-07, ..., 5.8673e-07, + 1.2107e-08, 1.8626e-09], + ..., + [-5.1223e-08, 3.7253e-09, -4.3772e-08, ..., 9.3132e-10, + -6.2399e-08, -1.8626e-08], + [ 0.0000e+00, 9.3132e-09, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 2.7940e-09], + [ 2.7940e-09, 9.8255e-07, 4.6566e-09, ..., 1.8626e-09, + 3.7253e-09, 2.2352e-08]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0187, -0.0207, -0.0061, -0.0234, -0.0031, 0.0013, 0.0114, 0.0180, + 0.0170, -0.0076], device='cuda:0'), grad: tensor([-2.3842e-06, 2.4308e-07, 1.3784e-06, -2.3395e-06, -1.3970e-08, + 7.4320e-07, 1.5832e-08, -8.8476e-08, 3.1665e-08, 2.4103e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 214.99, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4644 re_mapping 0.0042 re_causal 0.0122 /// teacc 98.99 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.3069, 0.0824, -0.1913, ..., -0.0851, -0.2714, -0.1731], + [-0.0011, 0.1066, -0.1251, ..., -0.1285, -0.0705, 0.1288], + [ 0.0580, -0.1377, -0.1754, ..., 0.0072, -0.0540, -0.1101], + ..., + [ 0.1015, -0.0850, 0.1412, ..., 0.0662, 0.2041, -0.0293], + [ 0.1249, -0.2313, -0.1451, ..., -0.3310, -0.0943, 0.2147], + [-0.1902, 0.0908, 0.0666, ..., -0.3094, -0.1633, -0.0681]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 1.3411e-07, ..., 9.3132e-10, + 1.8626e-08, 9.3132e-10], + [ 5.0291e-08, 1.4901e-08, 1.0803e-07, ..., 3.4459e-08, + 1.1083e-07, -9.3132e-10], + [ 3.7253e-09, 1.8626e-09, 5.1223e-08, ..., -1.9558e-08, + 4.3772e-08, 9.3132e-10], + ..., + [-8.6613e-08, 2.7940e-09, 6.0536e-08, ..., -4.8429e-08, + -1.7509e-07, 2.7940e-09], + [ 5.5879e-09, 1.1176e-08, 3.5390e-08, ..., 1.1176e-08, + 4.6566e-09, 0.0000e+00], + [ 7.4506e-09, -2.7940e-08, -4.9360e-07, ..., 1.2107e-08, + -4.2841e-08, 1.6764e-08]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0187, -0.0190, -0.0061, -0.0235, -0.0030, 0.0013, 0.0113, 0.0166, + 0.0168, -0.0078], device='cuda:0'), grad: tensor([ 6.5658e-07, 3.2503e-07, 1.5739e-07, 1.5181e-07, 3.2969e-07, + 1.8347e-07, -4.7125e-07, 7.0315e-07, 2.5425e-07, -2.3041e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 214.93, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4767 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.00 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.3066, 0.0826, -0.1917, ..., -0.0841, -0.2717, -0.1729], + [-0.0014, 0.1066, -0.1253, ..., -0.1288, -0.0710, 0.1286], + [ 0.0580, -0.1381, -0.1754, ..., 0.0072, -0.0539, -0.1096], + ..., + [ 0.1018, -0.0852, 0.1415, ..., 0.0663, 0.2047, -0.0291], + [ 0.1250, -0.2317, -0.1455, ..., -0.3326, -0.0948, 0.2148], + [-0.1909, 0.0904, 0.0665, ..., -0.3107, -0.1636, -0.0685]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-08, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.5926e-08, -1.8626e-09, 1.1269e-07, ..., 1.5274e-07, + 1.8626e-09, -8.3819e-09], + [ 5.1223e-08, 0.0000e+00, 6.0536e-08, ..., 8.3819e-08, + -9.3132e-10, -0.0000e+00], + ..., + [ 5.0385e-07, 9.3132e-10, 5.8301e-07, ..., 8.1863e-07, + -1.8626e-09, 9.3132e-10], + [ 8.0094e-08, 9.3132e-10, 9.5926e-08, ..., 1.2945e-07, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -2.7940e-09, -1.6764e-08, ..., 1.8626e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0186, -0.0192, -0.0060, -0.0235, -0.0021, 0.0013, 0.0114, 0.0168, + 0.0166, -0.0083], device='cuda:0'), grad: tensor([ 6.7614e-07, 3.8464e-07, 2.1327e-07, -3.0287e-06, 1.4901e-08, + 5.1223e-08, -6.9756e-07, 2.0843e-06, 3.4086e-07, -2.7940e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 214.89, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4647 re_mapping 0.0041 re_causal 0.0125 /// teacc 99.00 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.3073, 0.0819, -0.1929, ..., -0.0851, -0.2719, -0.1749], + [-0.0017, 0.1066, -0.1258, ..., -0.1286, -0.0715, 0.1282], + [ 0.0581, -0.1382, -0.1755, ..., 0.0072, -0.0540, -0.1097], + ..., + [ 0.1022, -0.0855, 0.1419, ..., 0.0662, 0.2053, -0.0282], + [ 0.1249, -0.2323, -0.1463, ..., -0.3348, -0.0950, 0.2148], + [-0.1923, 0.0906, 0.0662, ..., -0.3115, -0.1640, -0.0694]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 6.5193e-09, 4.6566e-09, 1.8626e-09, ..., 2.7940e-09, + 9.3132e-10, 2.7940e-09], + [-1.0803e-07, 9.3132e-10, 0.0000e+00, ..., -3.2596e-08, + -2.1420e-08, 1.8626e-09], + ..., + [ 2.7940e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 1.1176e-08, 7.4506e-09, 5.5879e-09, ..., 2.7940e-08, + 1.9558e-08, -1.9558e-07], + [ 9.3132e-10, 8.3819e-09, -7.4506e-09, ..., 9.3132e-10, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0192, -0.0194, -0.0059, -0.0236, -0.0008, 0.0013, 0.0117, 0.0170, + 0.0163, -0.0085], device='cuda:0'), grad: tensor([ 1.2107e-08, 4.5635e-08, -2.8498e-07, 6.7707e-07, -5.6811e-08, + -7.4878e-07, 5.7742e-07, 3.4459e-08, -2.8778e-07, 4.1910e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 214.51, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4435 re_mapping 0.0044 re_causal 0.0123 /// teacc 98.95 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.3076, 0.0821, -0.1934, ..., -0.0854, -0.2720, -0.1750], + [-0.0021, 0.1066, -0.1261, ..., -0.1286, -0.0721, 0.1278], + [ 0.0581, -0.1384, -0.1756, ..., 0.0072, -0.0540, -0.1099], + ..., + [ 0.1026, -0.0860, 0.1423, ..., 0.0663, 0.2059, -0.0276], + [ 0.1251, -0.2325, -0.1465, ..., -0.3349, -0.0952, 0.2152], + [-0.1927, 0.0909, 0.0663, ..., -0.3122, -0.1642, -0.0686]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8207e-06, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.9558e-08, 9.6112e-07, 1.0245e-08, ..., 5.5879e-09, + 9.3132e-10, -2.8871e-08], + [ 3.7253e-09, 6.1467e-08, 4.6566e-09, ..., 2.7940e-09, + 9.3132e-10, 3.7253e-09], + ..., + [ 1.4901e-08, 2.2352e-08, -6.5193e-09, ..., -2.7940e-09, + -5.5879e-09, 2.8871e-08], + [-3.1665e-08, 1.1176e-08, 8.3819e-09, ..., 9.3132e-10, + 0.0000e+00, -4.5635e-08], + [ 6.5193e-09, 4.9360e-08, -1.9558e-08, ..., 2.7940e-09, + 1.8626e-09, 8.3819e-09]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0191, -0.0196, -0.0059, -0.0236, -0.0014, 0.0011, 0.0117, 0.0172, + 0.0163, -0.0083], device='cuda:0'), grad: tensor([-5.0068e-06, 2.6226e-06, 1.8626e-07, -1.2405e-06, 2.9802e-08, + 1.3830e-06, 1.8626e-06, 1.0524e-07, -7.8231e-08, 1.2107e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 214.85, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4755 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.01 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.3080, 0.0822, -0.1941, ..., -0.0856, -0.2726, -0.1751], + [-0.0024, 0.1064, -0.1261, ..., -0.1291, -0.0726, 0.1283], + [ 0.0581, -0.1391, -0.1757, ..., 0.0072, -0.0541, -0.1113], + ..., + [ 0.1029, -0.0840, 0.1424, ..., 0.0667, 0.2065, -0.0272], + [ 0.1253, -0.2328, -0.1465, ..., -0.3353, -0.0951, 0.2154], + [-0.1933, 0.0907, 0.0663, ..., -0.3133, -0.1644, -0.0696]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -8.6613e-08, 2.4214e-08, ..., 1.3970e-08, + 9.3132e-10, -3.1665e-08], + [ 5.6811e-08, 2.2352e-08, 5.6811e-08, ..., 5.2154e-08, + 1.5832e-08, 0.0000e+00], + [-2.0396e-07, 2.8871e-08, 6.5193e-09, ..., -4.4703e-08, + -2.7940e-08, -2.1420e-08], + ..., + [ 1.3039e-07, 4.6566e-09, 1.4901e-08, ..., 5.4948e-08, + 2.7940e-09, 3.1665e-08], + [-1.7695e-08, 1.5832e-08, 8.3819e-09, ..., 8.3819e-09, + 1.8626e-09, -1.1642e-07], + [ 1.8626e-08, 6.2399e-08, 9.3132e-09, ..., 4.6566e-09, + 4.6566e-09, 8.9407e-08]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0191, -0.0195, -0.0061, -0.0235, -0.0013, 0.0009, 0.0117, 0.0174, + 0.0164, -0.0085], device='cuda:0'), grad: tensor([-2.6356e-07, 2.1979e-07, -3.0734e-07, -3.2317e-07, -2.4214e-08, + 9.3132e-08, 1.0058e-07, 3.1665e-07, -1.5832e-07, 3.4552e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 214.83, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4556 re_mapping 0.0042 re_causal 0.0121 /// teacc 98.97 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.3080, 0.0825, -0.1945, ..., -0.0861, -0.2729, -0.1748], + [-0.0031, 0.1065, -0.1267, ..., -0.1296, -0.0730, 0.1288], + [ 0.0581, -0.1375, -0.1755, ..., 0.0074, -0.0541, -0.1114], + ..., + [ 0.1035, -0.0847, 0.1429, ..., 0.0672, 0.2069, -0.0279], + [ 0.1256, -0.2332, -0.1466, ..., -0.3354, -0.0941, 0.2156], + [-0.1938, 0.0903, 0.0662, ..., -0.3162, -0.1645, -0.0698]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.5832e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -2.7940e-09, 3.7253e-09, ..., 3.7253e-09, + 1.8626e-09, -9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 4.6566e-09, ..., -9.3132e-10, + -5.5879e-09, 0.0000e+00], + ..., + [-3.7253e-09, 1.8626e-09, -4.6566e-09, ..., -9.3132e-10, + -3.7253e-09, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.7695e-08, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0189, -0.0198, -0.0057, -0.0237, -0.0015, 0.0009, 0.0117, 0.0177, + 0.0165, -0.0090], device='cuda:0'), grad: tensor([-3.5390e-08, -1.0245e-08, -2.5146e-08, -1.7695e-08, 2.7008e-08, + 5.5879e-09, -7.4506e-09, 9.3132e-09, 8.3819e-09, 4.2841e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 214.71, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.5017 re_mapping 0.0044 re_causal 0.0123 /// teacc 99.13 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.3086, 0.0827, -0.1951, ..., -0.0868, -0.2734, -0.1747], + [-0.0006, 0.1066, -0.1240, ..., -0.1298, -0.0731, 0.1302], + [ 0.0582, -0.1375, -0.1756, ..., 0.0074, -0.0540, -0.1115], + ..., + [ 0.1013, -0.0851, 0.1404, ..., 0.0670, 0.2071, -0.0297], + [ 0.1257, -0.2337, -0.1469, ..., -0.3365, -0.0944, 0.2157], + [-0.1944, 0.0902, 0.0662, ..., -0.3177, -0.1646, -0.0698]], + device='cuda:0'), grad: tensor([[-9.3132e-10, -4.2841e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-1.8626e-09, -3.7253e-09, 9.3132e-10, ..., 1.8626e-09, + 3.7253e-09, -1.2107e-07], + [ 2.0955e-08, 7.9162e-09, 0.0000e+00, ..., -2.3283e-09, + -5.1223e-09, 4.9360e-08], + ..., + [ 4.6566e-10, 8.3819e-09, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 1.5832e-08], + [-2.2352e-08, 5.5879e-09, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -2.0955e-08], + [ 9.3132e-10, 2.1886e-08, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 2.0023e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0188, -0.0174, -0.0054, -0.0237, -0.0016, 0.0001, 0.0122, 0.0153, + 0.0161, -0.0091], device='cuda:0'), grad: tensor([-1.0664e-07, -3.9907e-07, 1.6112e-07, 1.4435e-08, -9.5461e-08, + -2.9802e-08, 2.9989e-07, 7.7765e-08, -3.2596e-08, 1.2247e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 214.75, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.5007 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.3092, 0.0828, -0.1967, ..., -0.0876, -0.2742, -0.1746], + [-0.0013, 0.1067, -0.1244, ..., -0.1304, -0.0747, 0.1299], + [ 0.0578, -0.1378, -0.1763, ..., 0.0071, -0.0545, -0.1115], + ..., + [ 0.1022, -0.0855, 0.1410, ..., 0.0681, 0.2088, -0.0293], + [ 0.1259, -0.2341, -0.1474, ..., -0.3374, -0.0948, 0.2160], + [-0.1947, 0.0901, 0.0663, ..., -0.3184, -0.1647, -0.0703]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -1.1642e-08, 2.3283e-09, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 5.1223e-09, ..., 3.7253e-09, + 1.8626e-09, -1.3970e-09], + [-2.4680e-08, 2.7940e-09, -8.8476e-09, ..., -2.5611e-08, + 1.3970e-09, 4.6566e-10], + ..., + [-9.3132e-10, 1.3970e-09, -9.3132e-10, ..., 9.3132e-10, + -1.3970e-09, 1.3970e-09], + [ 1.8626e-09, 2.3283e-09, 9.3132e-10, ..., 2.3283e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 4.6566e-09, -4.1910e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0187, -0.0178, -0.0056, -0.0237, -0.0019, 0.0001, 0.0122, 0.0158, + 0.0159, -0.0091], device='cuda:0'), grad: tensor([-1.1642e-08, 1.5367e-08, -1.2014e-07, 6.7521e-08, 2.0023e-08, + 1.8626e-08, -1.4901e-08, 8.8476e-09, 1.9092e-08, 1.0710e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 214.98, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4641 re_mapping 0.0043 re_causal 0.0115 /// teacc 99.03 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.3101, 0.0829, -0.1988, ..., -0.0890, -0.2754, -0.1747], + [-0.0030, 0.1066, -0.1254, ..., -0.1308, -0.0768, 0.1299], + [ 0.0558, -0.1385, -0.1787, ..., 0.0060, -0.0562, -0.1119], + ..., + [ 0.1047, -0.0858, 0.1424, ..., 0.0700, 0.2114, -0.0294], + [ 0.1261, -0.2345, -0.1483, ..., -0.3388, -0.0953, 0.2163], + [-0.1951, 0.0899, 0.0665, ..., -0.3188, -0.1646, -0.0713]], + device='cuda:0'), grad: tensor([[-9.3132e-10, -4.2841e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 4.6566e-09, 9.3132e-10, 4.1910e-09, ..., 9.3132e-10, + 9.3132e-10, -2.7940e-09], + [ 2.3283e-09, 3.7253e-09, 2.7940e-09, ..., 2.3283e-09, + 1.3970e-09, 1.3970e-09], + ..., + [ 1.8626e-09, 3.7253e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [-5.5879e-09, 6.4727e-08, 5.5414e-08, ..., 4.6566e-10, + 0.0000e+00, -6.5193e-09], + [ 1.8626e-09, -5.4482e-08, -8.5216e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 362, bias, value: tensor([-1.8620e-02, -1.8831e-02, -6.9481e-03, -2.3424e-02, -1.2679e-03, + -4.5993e-06, 1.2204e-02, 1.7227e-02, 1.5779e-02, -9.2256e-03], + device='cuda:0'), grad: tensor([-5.5879e-08, 1.1176e-08, 1.7695e-08, 2.7940e-09, 9.1735e-08, + -1.2573e-08, -2.9337e-08, 2.1886e-08, 2.5937e-07, -2.9989e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 214.90, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4836 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.10 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.3110, 0.0831, -0.2001, ..., -0.0903, -0.2764, -0.1747], + [-0.0033, 0.1066, -0.1255, ..., -0.1315, -0.0773, 0.1299], + [ 0.0549, -0.1386, -0.1798, ..., 0.0058, -0.0577, -0.1120], + ..., + [ 0.1054, -0.0856, 0.1427, ..., 0.0712, 0.2126, -0.0294], + [ 0.1264, -0.2349, -0.1490, ..., -0.3399, -0.0961, 0.2166], + [-0.1953, 0.0900, 0.0666, ..., -0.3191, -0.1647, -0.0716]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.8626e-09, 2.7940e-09, ..., 9.3132e-10, + 2.7940e-09, 7.4506e-09], + [ 9.3132e-09, 3.7253e-09, 2.1420e-08, ..., 3.7253e-09, + 1.2107e-08, 2.7940e-09], + [-3.7253e-09, 2.7940e-09, 1.5832e-08, ..., -8.3819e-09, + -1.4901e-08, 2.6077e-08], + ..., + [-1.1176e-08, 3.9116e-08, 2.0955e-07, ..., 3.7253e-09, + 1.0058e-07, 3.7253e-09], + [-9.6858e-08, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 4.6566e-09, -1.8068e-07], + [ 6.7987e-08, -7.0781e-08, -3.0175e-07, ..., 0.0000e+00, + -1.2293e-07, 1.1921e-07]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0185, -0.0190, -0.0072, -0.0234, -0.0016, 0.0003, 0.0120, 0.0175, + 0.0156, -0.0092], device='cuda:0'), grad: tensor([ 1.0524e-07, 1.3877e-07, -1.1455e-07, -3.0734e-08, 5.9232e-07, + 2.6263e-07, -7.3109e-07, 4.6473e-07, -5.4482e-07, -1.4994e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 214.86, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4754 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.00 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.3117, 0.0829, -0.2034, ..., -0.0910, -0.2796, -0.1748], + [-0.0045, 0.1066, -0.1261, ..., -0.1320, -0.0786, 0.1292], + [ 0.0549, -0.1387, -0.1800, ..., 0.0058, -0.0578, -0.1121], + ..., + [ 0.1066, -0.0873, 0.1432, ..., 0.0717, 0.2139, -0.0285], + [ 0.1265, -0.2351, -0.1497, ..., -0.3412, -0.0975, 0.2174], + [-0.1955, 0.0905, 0.0674, ..., -0.3194, -0.1644, -0.0718]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.8626e-09, 2.7940e-09, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00], + [ 3.3528e-08, 1.3970e-08, 3.1665e-08, ..., 1.7695e-08, + 3.0734e-08, 3.7253e-09], + [ 8.2888e-08, 9.3132e-10, 8.5682e-08, ..., 4.4703e-08, + 8.4750e-08, 1.8626e-09], + ..., + [-1.5739e-07, 1.8626e-08, -1.4901e-07, ..., -8.3819e-08, + -1.5460e-07, 9.3132e-10], + [ 3.7253e-09, 3.2596e-08, 1.7695e-08, ..., 3.7253e-09, + 6.5193e-09, -6.5193e-09], + [ 2.7940e-08, 2.8871e-08, -8.3819e-09, ..., 1.3970e-08, + 2.5146e-08, 1.9558e-08]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0186, -0.0197, -0.0071, -0.0235, -0.0018, 0.0008, 0.0116, 0.0182, + 0.0163, -0.0088], device='cuda:0'), grad: tensor([ 4.6566e-09, 1.2293e-07, 1.8720e-07, 5.4017e-08, -3.2783e-07, + -1.8626e-07, 3.4459e-08, -2.5518e-07, 1.0058e-07, 2.6450e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 214.76, cls_loss 0.0011 cls_loss_mapping 0.0011 cls_loss_causal 0.4675 re_mapping 0.0043 re_causal 0.0119 /// teacc 98.95 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.3124, 0.0829, -0.2035, ..., -0.0918, -0.2798, -0.1751], + [-0.0039, 0.1065, -0.1254, ..., -0.1326, -0.0793, 0.1297], + [ 0.0550, -0.1388, -0.1808, ..., 0.0047, -0.0581, -0.1119], + ..., + [ 0.1063, -0.0876, 0.1427, ..., 0.0723, 0.2147, -0.0292], + [ 0.1263, -0.2356, -0.1513, ..., -0.3458, -0.0997, 0.2180], + [-0.1969, 0.0906, 0.0673, ..., -0.3210, -0.1646, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -8.1956e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-09, 4.6566e-09, 2.7940e-09, ..., 3.7253e-09, + 1.8626e-09, -1.8626e-09], + [-7.4506e-09, 2.7940e-09, 1.8626e-09, ..., -5.5879e-09, + 9.3132e-10, 0.0000e+00], + ..., + [-3.7253e-09, 2.7940e-09, -8.3819e-09, ..., -0.0000e+00, + -3.7253e-09, 1.8626e-09], + [ 1.8626e-09, 2.7940e-09, 2.7940e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 9.8720e-08, -9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0187, -0.0191, -0.0074, -0.0228, -0.0016, 0.0003, 0.0116, 0.0176, + 0.0160, -0.0090], device='cuda:0'), grad: tensor([-1.0431e-07, 3.9116e-08, -2.3283e-08, 9.3132e-09, -3.0920e-07, + 9.3132e-09, 1.5646e-07, 2.7940e-09, 2.3283e-08, 2.1048e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 214.85, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4580 re_mapping 0.0041 re_causal 0.0116 /// teacc 99.00 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.3128, 0.0829, -0.2036, ..., -0.0930, -0.2800, -0.1754], + [-0.0040, 0.1066, -0.1255, ..., -0.1330, -0.0795, 0.1299], + [ 0.0550, -0.1390, -0.1809, ..., 0.0048, -0.0581, -0.1122], + ..., + [ 0.1065, -0.0878, 0.1428, ..., 0.0725, 0.2150, -0.0294], + [ 0.1270, -0.2360, -0.1522, ..., -0.3466, -0.1006, 0.2188], + [-0.1973, 0.0898, 0.0674, ..., -0.3223, -0.1647, -0.0745]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -6.4261e-08, 1.6764e-08, ..., 1.8626e-09, + 4.6566e-09, -3.7253e-09], + [ 7.4506e-09, -1.1921e-07, 2.4214e-08, ..., 2.7940e-09, + 6.5193e-09, -1.0878e-06], + [ 2.7940e-09, 2.0489e-08, 3.7253e-09, ..., 9.3132e-10, + 9.3132e-10, 9.5926e-08], + ..., + [-3.4459e-08, 1.8347e-07, -8.5682e-08, ..., -1.4901e-08, + -3.3528e-08, 1.0077e-06], + [-2.7940e-09, 2.4214e-08, 1.5832e-08, ..., 1.8626e-09, + 3.7253e-09, 6.2399e-08], + [ 1.3039e-08, -6.0443e-07, -9.9838e-07, ..., 5.5879e-09, + 1.2107e-08, 8.8476e-08]], device='cuda:0') +Epoch 366, bias, value: tensor([-1.8616e-02, -1.9143e-02, -7.3684e-03, -2.2830e-02, -5.0967e-06, + 4.5515e-04, 1.1458e-02, 1.7689e-02, 1.6315e-02, -9.9327e-03], + device='cuda:0'), grad: tensor([-1.5646e-07, -1.3178e-06, 1.5181e-07, 2.3283e-08, 2.1011e-06, + 4.2841e-08, 5.9605e-08, 1.2713e-06, 1.3504e-07, -2.3171e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 214.76, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4673 re_mapping 0.0041 re_causal 0.0115 /// teacc 98.96 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.3131, 0.0830, -0.2037, ..., -0.0956, -0.2801, -0.1759], + [-0.0042, 0.1065, -0.1256, ..., -0.1336, -0.0796, 0.1306], + [ 0.0555, -0.1397, -0.1811, ..., 0.0048, -0.0581, -0.1124], + ..., + [ 0.1064, -0.0897, 0.1429, ..., 0.0725, 0.2151, -0.0305], + [ 0.1275, -0.2371, -0.1535, ..., -0.3486, -0.1009, 0.2196], + [-0.2003, 0.0888, 0.0673, ..., -0.3236, -0.1649, -0.0782]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.2396e-06, 4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, 5.5879e-09], + [-8.3819e-09, 1.1735e-07, 2.7008e-08, ..., 1.8626e-09, + 1.8626e-09, 1.3877e-07], + [-1.3039e-08, 4.6566e-09, 2.7940e-09, ..., -8.3819e-09, + 0.0000e+00, 3.7253e-09], + ..., + [-2.7940e-09, 4.0047e-08, 1.4901e-08, ..., -9.3132e-10, + -8.3819e-09, 2.7008e-08], + [ 9.3132e-10, 7.8231e-08, 1.6764e-08, ..., 7.4506e-09, + 0.0000e+00, 3.1665e-08], + [ 7.4506e-09, -1.2014e-07, -2.1327e-07, ..., 9.3132e-10, + 1.8626e-09, 8.0094e-08]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0188, -0.0194, -0.0067, -0.0208, 0.0019, -0.0016, 0.0116, 0.0175, + 0.0165, -0.0110], device='cuda:0'), grad: tensor([ 8.0019e-06, 5.4482e-07, -1.6764e-08, -1.8626e-09, 7.0129e-07, + 1.0151e-07, -9.4920e-06, 1.0617e-07, 3.2224e-07, -2.4121e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 215.24, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4770 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.04 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.3134, 0.0830, -0.2038, ..., -0.0961, -0.2803, -0.1764], + [-0.0038, 0.1064, -0.1253, ..., -0.1317, -0.0786, 0.1304], + [ 0.0552, -0.1401, -0.1817, ..., 0.0046, -0.0586, -0.1126], + ..., + [ 0.1064, -0.0886, 0.1431, ..., 0.0717, 0.2148, -0.0299], + [ 0.1277, -0.2381, -0.1553, ..., -0.3496, -0.1026, 0.2200], + [-0.2026, 0.0896, 0.0673, ..., -0.3262, -0.1654, -0.0785]], + device='cuda:0'), grad: tensor([[-0.0000e+00, -1.0617e-07, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -4.5635e-08], + [-9.3132e-10, 5.4948e-08, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.0245e-08], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 2.5146e-08, 1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-08], + [ 0.0000e+00, 9.3132e-09, 2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, 7.4506e-09], + [-0.0000e+00, 3.1665e-08, -2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-08]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0187, -0.0184, -0.0073, -0.0212, 0.0011, -0.0022, 0.0119, 0.0171, + 0.0165, -0.0107], device='cuda:0'), grad: tensor([-2.1048e-07, 8.6613e-08, 6.5193e-09, 1.2107e-08, 1.3970e-08, + -3.9488e-07, 3.3434e-07, 5.9605e-08, 2.7940e-08, 7.4506e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 214.96, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4542 re_mapping 0.0042 re_causal 0.0120 /// teacc 99.02 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.3143, 0.0831, -0.2041, ..., -0.0987, -0.2812, -0.1771], + [-0.0038, 0.1065, -0.1253, ..., -0.1317, -0.0787, 0.1306], + [ 0.0553, -0.1406, -0.1819, ..., 0.0047, -0.0585, -0.1127], + ..., + [ 0.1065, -0.0889, 0.1430, ..., 0.0713, 0.2149, -0.0299], + [ 0.1275, -0.2394, -0.1565, ..., -0.3507, -0.1038, 0.2199], + [-0.2028, 0.0895, 0.0674, ..., -0.3266, -0.1654, -0.0786]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.8918e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.3039e-08], + [ 7.4506e-09, -1.4529e-07, 9.3132e-09, ..., 4.6566e-09, + 5.5879e-09, -6.0070e-07], + [-1.2107e-08, 4.6566e-09, 3.7253e-09, ..., -4.6566e-09, + -1.0245e-08, 1.3970e-08], + ..., + [-0.0000e+00, 8.3819e-09, -1.3039e-08, ..., 9.3132e-10, + 1.8626e-09, 1.5832e-08], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 9.3132e-10, 3.1665e-08, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 3.5390e-08]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0187, -0.0184, -0.0072, -0.0209, 0.0011, -0.0024, 0.0121, 0.0170, + 0.0159, -0.0108], device='cuda:0'), grad: tensor([-1.5181e-07, -9.4622e-07, 1.0245e-08, 1.1362e-07, -1.2107e-07, + -1.1269e-07, 1.0524e-06, 3.1665e-08, 2.9802e-08, 1.0245e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 214.79, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4729 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.00 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.3145, 0.0826, -0.2048, ..., -0.0993, -0.2814, -0.1769], + [-0.0040, 0.1067, -0.1254, ..., -0.1318, -0.0790, 0.1304], + [ 0.0554, -0.1407, -0.1819, ..., 0.0048, -0.0585, -0.1128], + ..., + [ 0.1066, -0.0895, 0.1431, ..., 0.0712, 0.2152, -0.0295], + [ 0.1279, -0.2413, -0.1579, ..., -0.3512, -0.1039, 0.2200], + [-0.2029, 0.0904, 0.0676, ..., -0.3268, -0.1654, -0.0786]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.0978e-08, 1.8626e-09, ..., 3.7253e-09, + 1.8626e-09, 3.3528e-08], + [ 1.1176e-08, -3.4086e-07, 1.1176e-08, ..., 1.8626e-08, + 7.4506e-09, -1.5460e-07], + [-7.8231e-08, 1.1176e-08, -8.1956e-08, ..., -1.3784e-07, + -5.5879e-08, 3.7253e-09], + ..., + [ 3.9116e-08, 7.4506e-09, 4.2841e-08, ..., 7.0781e-08, + 2.9802e-08, 5.5879e-09], + [ 3.7253e-09, 7.4506e-09, 7.4506e-09, ..., 1.3039e-08, + 5.5879e-09, -1.4901e-08], + [ 7.4506e-09, -3.3528e-08, -4.8429e-08, ..., 5.5879e-09, + 1.8626e-09, 1.4901e-08]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0189, -0.0185, -0.0071, -0.0210, 0.0011, -0.0022, 0.0120, 0.0171, + 0.0157, -0.0104], device='cuda:0'), grad: tensor([ 1.4901e-07, -8.9221e-07, -4.7870e-07, 5.0291e-08, 1.3411e-07, + 4.8429e-08, 6.5193e-07, 2.8871e-07, 3.5390e-08, 1.3039e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 215.19, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4985 re_mapping 0.0040 re_causal 0.0118 /// teacc 98.99 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.3150, 0.0810, -0.2057, ..., -0.1003, -0.2818, -0.1770], + [-0.0042, 0.1063, -0.1255, ..., -0.1321, -0.0791, 0.1304], + [ 0.0495, -0.1408, -0.1853, ..., -0.0012, -0.0646, -0.1182], + ..., + [ 0.1119, -0.0896, 0.1440, ..., 0.0773, 0.2205, -0.0249], + [ 0.1280, -0.2417, -0.1586, ..., -0.3537, -0.1059, 0.2204], + [-0.2040, 0.0920, 0.0678, ..., -0.3280, -0.1656, -0.0797]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.1537e-07, 1.8626e-09, ..., -3.9116e-08, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 1.8626e-09, 3.3528e-08, ..., 1.1176e-08, + 5.5879e-09, -1.8626e-09], + [ 9.3132e-09, 1.1176e-08, 1.1176e-08, ..., 5.5879e-09, + 9.3132e-09, 0.0000e+00], + ..., + [-3.1665e-08, 3.7253e-09, 2.0489e-08, ..., 7.4506e-09, + -2.7940e-08, 1.8626e-09], + [-3.7253e-09, 5.5879e-09, 2.7940e-08, ..., 7.4506e-09, + 1.8626e-09, -1.4901e-08], + [ 7.4506e-09, 2.6822e-07, 7.4506e-09, ..., 4.6566e-08, + 5.5879e-09, 5.5879e-09]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0199, -0.0187, -0.0128, -0.0203, 0.0020, -0.0022, 0.0117, 0.0213, + 0.0158, -0.0096], device='cuda:0'), grad: tensor([-9.0152e-07, 6.8918e-08, 6.3330e-08, -1.2852e-07, 2.8685e-07, + 6.8918e-08, -1.0990e-07, -1.8626e-09, 2.6077e-08, 6.3144e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 214.95, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4548 re_mapping 0.0039 re_causal 0.0114 /// teacc 98.94 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.3151, 0.0813, -0.2058, ..., -0.1003, -0.2820, -0.1776], + [-0.0040, 0.1062, -0.1253, ..., -0.1322, -0.0792, 0.1311], + [ 0.0494, -0.1408, -0.1854, ..., -0.0012, -0.0646, -0.1182], + ..., + [ 0.1119, -0.0900, 0.1436, ..., 0.0773, 0.2206, -0.0253], + [ 0.1281, -0.2421, -0.1591, ..., -0.3542, -0.1062, 0.2207], + [-0.2046, 0.0921, 0.0688, ..., -0.3289, -0.1657, -0.0802]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -6.1467e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 5.1223e-07, 9.8720e-08, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 5.9605e-07], + [ 5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 5.5879e-09, 3.7253e-09, -1.3039e-08, ..., -0.0000e+00, + -1.1176e-08, 1.6764e-08], + [-6.2771e-07, -1.1735e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.3388e-07], + [ 1.3039e-08, 4.6566e-08, -9.3132e-09, ..., 0.0000e+00, + 7.4506e-09, 9.3132e-09]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0197, -0.0184, -0.0128, -0.0203, 0.0012, -0.0022, 0.0117, 0.0211, + 0.0158, -0.0091], device='cuda:0'), grad: tensor([-1.3225e-07, 1.8440e-06, 2.2352e-08, 3.3528e-08, 3.3528e-08, + 2.0489e-07, 5.0291e-08, 3.1665e-08, -2.2538e-06, 1.4901e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 214.97, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4748 re_mapping 0.0042 re_causal 0.0120 /// teacc 98.97 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.3145, 0.0816, -0.2061, ..., -0.1008, -0.2821, -0.1770], + [-0.0071, 0.1064, -0.1282, ..., -0.1322, -0.0823, 0.1281], + [ 0.0494, -0.1412, -0.1856, ..., -0.0012, -0.0647, -0.1183], + ..., + [ 0.1145, -0.0908, 0.1464, ..., 0.0773, 0.2234, -0.0223], + [ 0.1288, -0.2429, -0.1597, ..., -0.3551, -0.1066, 0.2215], + [-0.2053, 0.0919, 0.0693, ..., -0.3301, -0.1655, -0.0811]], + device='cuda:0'), grad: tensor([[-3.7253e-09, -4.8429e-08, 3.5390e-08, ..., 2.9802e-08, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 8.3819e-08, 2.0489e-08, ..., 1.4901e-08, + 0.0000e+00, 9.3132e-09], + [-1.6764e-08, 6.3330e-08, 2.0489e-08, ..., 5.5879e-09, + 0.0000e+00, 1.8626e-09], + ..., + [ 1.4901e-08, 1.4715e-07, 1.1176e-08, ..., 2.2352e-08, + -1.8626e-09, 2.0489e-08], + [ 1.8626e-09, 1.7881e-07, 2.7940e-08, ..., 1.3039e-08, + 0.0000e+00, 3.7253e-09], + [ 1.8626e-09, 1.9260e-06, -1.4342e-07, ..., -5.5879e-09, + 0.0000e+00, 3.2410e-07]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0195, -0.0213, -0.0128, -0.0193, 0.0014, -0.0034, 0.0118, 0.0236, + 0.0158, -0.0090], device='cuda:0'), grad: tensor([-2.9802e-08, 3.4645e-07, 1.2293e-07, -9.6858e-08, -1.0461e-05, + 2.3469e-07, 6.3330e-08, 6.7241e-07, 4.7870e-07, 8.6427e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 215.27, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.5016 re_mapping 0.0041 re_causal 0.0118 /// teacc 98.97 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.3150, 0.0819, -0.2061, ..., -0.1016, -0.2823, -0.1767], + [-0.0072, 0.1065, -0.1280, ..., -0.1322, -0.0822, 0.1283], + [ 0.0497, -0.1412, -0.1854, ..., -0.0011, -0.0645, -0.1183], + ..., + [ 0.1144, -0.0916, 0.1462, ..., 0.0772, 0.2233, -0.0225], + [ 0.1291, -0.2438, -0.1604, ..., -0.3555, -0.1070, 0.2218], + [-0.2063, 0.0916, 0.0693, ..., -0.3308, -0.1656, -0.0814]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0489e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3441e-05, -1.8626e-09, -2.2538e-07, ..., 1.8626e-09, + -5.5879e-06, -5.4203e-07], + [ 1.2010e-05, 3.7253e-09, 2.0862e-07, ..., -0.0000e+00, + 4.9919e-06, 4.7497e-07], + ..., + [ 1.4119e-06, 5.5879e-09, 2.2352e-08, ..., -1.8626e-09, + 5.8487e-07, 6.1467e-08], + [ 5.5879e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 7.4506e-09, 9.3132e-09, -9.3132e-09, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0194, -0.0213, -0.0122, -0.0195, 0.0013, -0.0029, 0.0118, 0.0233, + 0.0156, -0.0093], device='cuda:0'), grad: tensor([-5.2154e-08, -5.7995e-05, 5.1796e-05, 1.8626e-09, 9.3132e-09, + 3.3528e-08, -1.0617e-07, 6.1244e-06, 1.4156e-07, 4.8429e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 215.18, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4820 re_mapping 0.0043 re_causal 0.0119 /// teacc 99.04 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.3171, 0.0818, -0.2062, ..., -0.1025, -0.2828, -0.1785], + [-0.0071, 0.1068, -0.1280, ..., -0.1323, -0.0822, 0.1284], + [ 0.0497, -0.1414, -0.1855, ..., -0.0011, -0.0645, -0.1184], + ..., + [ 0.1144, -0.0927, 0.1458, ..., 0.0768, 0.2232, -0.0226], + [ 0.1291, -0.2448, -0.1616, ..., -0.3564, -0.1074, 0.2219], + [-0.2068, 0.0921, 0.0701, ..., -0.3313, -0.1656, -0.0813]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-2.8517e-06, 6.1467e-08, -1.5333e-05, ..., -5.1036e-07, + -2.4229e-05, 1.8626e-09], + [ 3.7253e-08, -3.9116e-08, 1.0729e-06, ..., 2.6077e-08, + 1.6689e-06, 1.1176e-08], + ..., + [ 2.7698e-06, 3.3528e-08, 1.4298e-05, ..., 4.8615e-07, + 2.2486e-05, 1.6764e-08], + [-1.8626e-08, 1.1176e-08, 1.3039e-08, ..., 1.8626e-09, + 1.8626e-09, -2.6077e-08], + [ 7.4506e-09, -1.0245e-06, -1.2238e-06, ..., 0.0000e+00, + 3.7253e-09, -2.4587e-07]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0195, -0.0212, -0.0123, -0.0167, 0.0006, -0.0045, 0.0118, 0.0231, + 0.0151, -0.0087], device='cuda:0'), grad: tensor([ 3.9116e-08, -8.6129e-05, 5.1670e-06, 8.3819e-08, 3.1553e-06, + 4.4703e-08, 1.2666e-07, 8.0705e-05, 2.4214e-08, -3.3565e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 215.09, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4786 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.03 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.3171, 0.0850, -0.2063, ..., -0.1032, -0.2830, -0.1756], + [-0.0070, 0.1067, -0.1276, ..., -0.1323, -0.0821, 0.1287], + [ 0.0498, -0.1416, -0.1856, ..., -0.0010, -0.0644, -0.1184], + ..., + [ 0.1142, -0.0940, 0.1454, ..., 0.0768, 0.2232, -0.0229], + [ 0.1294, -0.2450, -0.1618, ..., -0.3566, -0.1076, 0.2223], + [-0.2073, 0.0922, 0.0708, ..., -0.3316, -0.1655, -0.0818]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.7253e-09, 7.4506e-09, ..., 3.7253e-09, + 0.0000e+00, 1.8626e-09], + [-2.5451e-05, 1.8626e-08, -3.8475e-05, ..., 1.4901e-08, + 5.5879e-09, -2.5600e-05], + [ 9.8720e-08, 1.3970e-07, 2.1979e-07, ..., 1.5460e-07, + 0.0000e+00, 1.4901e-08], + ..., + [ 2.5243e-05, 3.7253e-09, 3.8207e-05, ..., 5.5879e-09, + -1.3039e-08, 2.5392e-05], + [-1.8626e-09, 1.3039e-08, 3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 5.0291e-08, -8.1956e-08, -1.0245e-07, ..., 0.0000e+00, + 5.5879e-09, 4.0978e-08]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0165, -0.0210, -0.0122, -0.0168, 0.0008, -0.0045, 0.0091, 0.0227, + 0.0153, -0.0086], device='cuda:0'), grad: tensor([ 1.6764e-08, -7.6413e-05, 4.8988e-07, -2.3469e-07, 2.6077e-07, + 0.0000e+00, 7.4506e-09, 7.5936e-05, 6.5193e-08, -2.4773e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 215.07, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4505 re_mapping 0.0041 re_causal 0.0114 /// teacc 98.97 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.3177, 0.0850, -0.2063, ..., -0.1035, -0.2834, -0.1755], + [-0.0070, 0.1067, -0.1275, ..., -0.1324, -0.0821, 0.1289], + [ 0.0498, -0.1419, -0.1858, ..., -0.0010, -0.0644, -0.1184], + ..., + [ 0.1142, -0.0942, 0.1453, ..., 0.0768, 0.2232, -0.0231], + [ 0.1300, -0.2454, -0.1623, ..., -0.3568, -0.1069, 0.2228], + [-0.2066, 0.0924, 0.0715, ..., -0.3318, -0.1653, -0.0820]], + device='cuda:0'), grad: tensor([[-7.4506e-09, -2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [ 3.7253e-09, -1.8626e-09, 1.4901e-08, ..., 3.7253e-09, + 5.5879e-09, -3.1665e-08], + [ 3.7253e-09, 5.5879e-09, 5.5879e-09, ..., 3.7253e-09, + 0.0000e+00, 1.8626e-09], + ..., + [-5.2154e-08, 3.7253e-09, -3.3528e-08, ..., -1.8626e-09, + -1.6764e-08, 5.5879e-09], + [ 3.7253e-09, 7.4506e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 2.9802e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 1.1176e-08]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0164, -0.0209, -0.0122, -0.0169, 0.0004, -0.0043, 0.0090, 0.0226, + 0.0153, -0.0080], device='cuda:0'), grad: tensor([-8.1956e-08, -1.8626e-08, 3.1665e-08, -9.3132e-09, 2.4214e-08, + 3.3528e-08, 5.5879e-09, -7.2643e-08, 2.7940e-08, 5.0291e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 215.18, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4847 re_mapping 0.0038 re_causal 0.0114 /// teacc 98.98 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.3207, 0.0851, -0.2067, ..., -0.1067, -0.2866, -0.1755], + [-0.0070, 0.1062, -0.1276, ..., -0.1326, -0.0822, 0.1289], + [ 0.0498, -0.1422, -0.1859, ..., -0.0009, -0.0643, -0.1185], + ..., + [ 0.1142, -0.0949, 0.1453, ..., 0.0768, 0.2232, -0.0231], + [ 0.1312, -0.2464, -0.1628, ..., -0.3569, -0.1072, 0.2238], + [-0.2069, 0.0928, 0.0720, ..., -0.3322, -0.1653, -0.0823]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.0552e-06, 2.9057e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, -5.5879e-09], + [-0.0000e+00, 4.2841e-08, 1.8626e-09, ..., -1.8626e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-09], + [-5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 1.8626e-09, -5.1931e-06, -2.9989e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-08]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0166, -0.0210, -0.0120, -0.0171, 0.0002, -0.0040, 0.0090, 0.0226, + 0.0157, -0.0075], device='cuda:0'), grad: tensor([ 1.0908e-05, 5.5879e-09, 8.9407e-08, -1.8626e-09, 1.2293e-07, + 4.4703e-08, -3.5390e-08, 4.8429e-08, 0.0000e+00, -1.1176e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 215.12, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4715 re_mapping 0.0040 re_causal 0.0117 /// teacc 98.96 lr 0.00010000 +Epoch 379, weight, value: tensor([[-3.2090e-01, 8.5054e-02, -2.0690e-01, ..., -1.0692e-01, + -2.8684e-01, -1.7552e-01], + [-7.0227e-03, 1.0596e-01, -1.2756e-01, ..., -1.3267e-01, + -8.2172e-02, 1.2905e-01], + [ 5.0169e-02, -1.4331e-01, -1.8508e-01, ..., -3.3537e-04, + -6.3194e-02, -1.1866e-01], + ..., + [ 1.1397e-01, -9.5941e-02, 1.4515e-01, ..., 7.6159e-02, + 2.2270e-01, -2.3162e-02], + [ 1.3167e-01, -2.4664e-01, -1.6301e-01, ..., -3.5707e-01, + -1.0748e-01, 2.2432e-01], + [-2.0639e-01, 9.3577e-02, 7.3160e-02, ..., -3.3247e-01, + -1.6465e-01, -8.1888e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.4506e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 7.4506e-09, 9.3132e-09, 4.0978e-08, ..., 2.2352e-08, + 1.8626e-09, -7.4506e-09], + [-0.0000e+00, 5.5879e-09, 3.7253e-09, ..., 1.8626e-09, + -0.0000e+00, 7.4506e-09], + ..., + [ 3.3528e-08, 9.4995e-08, 1.6391e-07, ..., -0.0000e+00, + -1.8626e-09, 1.4901e-08], + [-5.5879e-08, 3.7253e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, -1.3039e-07], + [-8.0094e-08, -2.3097e-07, -4.3027e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0166, -0.0210, -0.0116, -0.0174, -0.0009, -0.0038, 0.0090, 0.0223, + 0.0158, -0.0063], device='cuda:0'), grad: tensor([ 3.7253e-09, 9.1270e-08, 5.0291e-08, -7.0781e-08, 3.7439e-07, + 5.2154e-08, 3.5390e-08, 3.2969e-07, -1.3411e-07, -7.4133e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 214.97, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4912 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.00 lr 0.00010000 +Epoch 380, weight, value: tensor([[-3.2104e-01, 8.5091e-02, -2.0715e-01, ..., -1.0707e-01, + -2.8699e-01, -1.7549e-01], + [-7.0416e-03, 1.0574e-01, -1.2761e-01, ..., -1.3292e-01, + -8.2193e-02, 1.2908e-01], + [ 5.0165e-02, -1.4372e-01, -1.8527e-01, ..., -3.0770e-04, + -6.3153e-02, -1.1873e-01], + ..., + [ 1.1402e-01, -9.6256e-02, 1.4525e-01, ..., 7.6224e-02, + 2.2275e-01, -2.3183e-02], + [ 1.3200e-01, -2.4709e-01, -1.6363e-01, ..., -3.5737e-01, + -1.0792e-01, 2.2467e-01], + [-2.0747e-01, 9.3518e-02, 7.3351e-02, ..., -3.3430e-01, + -1.6489e-01, -8.3039e-02]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -4.0978e-08, 3.7253e-09, ..., 5.5879e-09, + 1.8626e-09, -5.5879e-09], + [ 1.1176e-08, 2.2352e-08, 1.4901e-08, ..., 1.3039e-08, + 5.5879e-09, 4.0978e-08], + [ 1.4901e-08, 5.5879e-09, 0.0000e+00, ..., -2.0489e-08, + -7.4506e-09, 5.5879e-09], + ..., + [-1.6764e-08, 1.1176e-08, -1.8626e-09, ..., -1.1176e-08, + -1.6764e-08, 5.5879e-09], + [ 2.0489e-08, 4.6566e-08, 1.0058e-07, ..., 3.7253e-09, + 0.0000e+00, -1.4901e-08], + [-3.1665e-08, -0.0000e+00, -1.2852e-07, ..., 0.0000e+00, + 0.0000e+00, 4.2841e-08]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0166, -0.0210, -0.0116, -0.0168, -0.0005, -0.0049, 0.0091, 0.0224, + 0.0157, -0.0064], device='cuda:0'), grad: tensor([-6.3330e-08, 1.5087e-07, -1.5087e-07, 7.6368e-08, -1.5087e-07, + -1.3039e-07, 1.6578e-07, 2.9802e-08, 3.3155e-07, -2.6077e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 215.54, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4862 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.02 lr 0.00010000 +Epoch 381, weight, value: tensor([[-3.2150e-01, 8.5052e-02, -2.0748e-01, ..., -1.0733e-01, + -2.8716e-01, -1.7559e-01], + [-7.0471e-03, 1.0629e-01, -1.2764e-01, ..., -1.3314e-01, + -8.2227e-02, 1.2934e-01], + [ 5.0154e-02, -1.4411e-01, -1.8540e-01, ..., -3.1780e-04, + -6.3162e-02, -1.1933e-01], + ..., + [ 1.1404e-01, -9.6640e-02, 1.4526e-01, ..., 7.6249e-02, + 2.2279e-01, -2.3219e-02], + [ 1.3233e-01, -2.4958e-01, -1.6541e-01, ..., -3.5779e-01, + -1.0825e-01, 2.2400e-01], + [-2.0801e-01, 9.3332e-02, 7.3749e-02, ..., -3.3552e-01, + -1.6478e-01, -8.3964e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.8813e-07, -5.4017e-08, ..., 0.0000e+00, + 0.0000e+00, -1.1399e-06], + [-1.8626e-09, 1.5646e-07, 4.6566e-08, ..., -1.8626e-09, + -1.8626e-09, 9.5740e-07], + ..., + [-1.8626e-09, 2.6077e-08, 3.7253e-09, ..., 1.8626e-09, + -3.7253e-09, 1.5274e-07], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 0.0000e+00, -5.5879e-09], + [ 3.7253e-09, -1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 381, bias, value: tensor([-1.6681e-02, -2.0937e-02, -1.1697e-02, -1.5873e-02, 3.4713e-05, + -6.1113e-03, 9.2421e-03, 2.2333e-02, 1.4960e-02, -6.4731e-03], + device='cuda:0'), grad: tensor([ 0.0000e+00, -1.7956e-06, 1.5032e-06, 1.6950e-07, 1.6764e-08, + -1.3411e-07, 5.5879e-09, 2.3842e-07, -5.5879e-09, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 215.48, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4610 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 382, weight, value: tensor([[-3.2173e-01, 8.5094e-02, -2.0763e-01, ..., -1.0748e-01, + -2.8725e-01, -1.7561e-01], + [-7.0837e-03, 1.0636e-01, -1.2772e-01, ..., -1.3344e-01, + -8.2278e-02, 1.2936e-01], + [ 5.0151e-02, -1.4434e-01, -1.8566e-01, ..., -2.9652e-04, + -6.3185e-02, -1.1940e-01], + ..., + [ 1.1434e-01, -9.6628e-02, 1.4596e-01, ..., 7.6293e-02, + 2.2355e-01, -2.3219e-02], + [ 1.3276e-01, -2.5015e-01, -1.6626e-01, ..., -3.5841e-01, + -1.0843e-01, 2.2462e-01], + [-2.1128e-01, 9.3384e-02, 7.1495e-02, ..., -3.3676e-01, + -1.6789e-01, -8.4024e-02]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.3039e-08, 1.8626e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -1.8626e-09, 3.7253e-09, ..., 5.5879e-09, + 3.7253e-09, -2.2352e-08], + [-1.6764e-08, 1.3039e-08, 1.8626e-08, ..., -7.4506e-09, + -9.3132e-09, 5.5879e-09], + ..., + [-1.8626e-09, 5.5879e-09, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 9.3132e-09], + [-1.8626e-09, 3.7253e-09, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, -5.5879e-09], + [ 1.8626e-09, 8.0094e-08, -9.3132e-09, ..., 3.7253e-09, + 0.0000e+00, 1.2480e-07]], device='cuda:0') +Epoch 382, bias, value: tensor([-1.6638e-02, -2.0981e-02, -1.1682e-02, -1.5964e-02, 7.3005e-05, + -6.2297e-03, 9.2144e-03, 2.2756e-02, 1.5060e-02, -8.4564e-03], + device='cuda:0'), grad: tensor([-3.3528e-08, -1.1176e-08, -2.9802e-08, -2.0489e-08, -3.7998e-07, + 1.3039e-08, 1.8626e-08, 2.9802e-08, 9.3132e-09, 3.7625e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 215.31, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4774 re_mapping 0.0038 re_causal 0.0110 /// teacc 98.98 lr 0.00010000 +Epoch 383, weight, value: tensor([[-3.2185e-01, 8.5087e-02, -2.0790e-01, ..., -1.0767e-01, + -2.8729e-01, -1.7566e-01], + [-7.0202e-03, 1.0707e-01, -1.2753e-01, ..., -1.3364e-01, + -8.2255e-02, 1.2986e-01], + [ 5.0204e-02, -1.4579e-01, -1.8580e-01, ..., -2.2696e-04, + -6.3064e-02, -1.1952e-01], + ..., + [ 1.1427e-01, -9.7734e-02, 1.4579e-01, ..., 7.6208e-02, + 2.2350e-01, -2.3501e-02], + [ 1.3301e-01, -2.5067e-01, -1.6688e-01, ..., -3.5915e-01, + -1.0884e-01, 2.2499e-01], + [-2.1138e-01, 9.3380e-02, 7.1447e-02, ..., -3.3756e-01, + -1.6792e-01, -8.5555e-02]], device='cuda:0'), grad: tensor([[ 1.4901e-08, -9.4995e-08, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 2.0489e-08], + [-8.6613e-07, -1.3895e-06, -2.2165e-07, ..., 0.0000e+00, + 0.0000e+00, -1.2070e-06], + [ 3.7253e-09, 2.0489e-08, 1.8626e-09, ..., -1.8626e-09, + -1.8626e-09, 1.4901e-08], + ..., + [ 6.6310e-07, 1.0692e-06, 1.7136e-07, ..., 0.0000e+00, + -1.8626e-09, 9.2573e-07], + [ 9.3132e-09, 1.8626e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 1.1735e-07, 2.4959e-07, -7.8231e-08, ..., 0.0000e+00, + 0.0000e+00, 1.9185e-07]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0167, -0.0207, -0.0116, -0.0161, 0.0008, -0.0059, 0.0091, 0.0225, + 0.0149, -0.0088], device='cuda:0'), grad: tensor([-8.9407e-08, -5.5693e-06, 6.8918e-08, 1.0990e-07, 2.2165e-07, + 1.3784e-07, -1.4529e-07, 4.2878e-06, 8.7544e-08, 8.6986e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 215.16, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4698 re_mapping 0.0041 re_causal 0.0116 /// teacc 99.08 lr 0.00010000 +Epoch 384, weight, value: tensor([[-3.2198e-01, 8.5156e-02, -2.0801e-01, ..., -1.0773e-01, + -2.8733e-01, -1.7554e-01], + [-7.0267e-03, 1.0723e-01, -1.2756e-01, ..., -1.3396e-01, + -8.2289e-02, 1.2996e-01], + [ 5.0197e-02, -1.4623e-01, -1.8598e-01, ..., -2.0350e-04, + -6.3064e-02, -1.1967e-01], + ..., + [ 1.1428e-01, -9.8728e-02, 1.4582e-01, ..., 7.6209e-02, + 2.2354e-01, -2.3558e-02], + [ 1.3367e-01, -2.5118e-01, -1.6694e-01, ..., -3.5866e-01, + -1.0940e-01, 2.2573e-01], + [-2.1146e-01, 9.2691e-02, 7.1362e-02, ..., -3.3812e-01, + -1.6794e-01, -8.7762e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.2527e-07, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [ 0.0000e+00, 1.3039e-08, 1.6764e-08, ..., 7.4506e-09, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, 2.0489e-08, 2.6077e-08, ..., 1.8626e-08, + 0.0000e+00, 5.5879e-09], + ..., + [ 0.0000e+00, 9.3132e-09, 1.3039e-08, ..., 7.4506e-09, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 6.8918e-08, 2.7940e-08, ..., 5.5879e-09, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, -4.4517e-07, -2.1607e-07, ..., 1.8626e-09, + 0.0000e+00, -6.1467e-08]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0166, -0.0207, -0.0116, -0.0161, 0.0020, -0.0061, 0.0092, 0.0225, + 0.0150, -0.0094], device='cuda:0'), grad: tensor([-1.6969e-06, 5.4017e-08, 1.2107e-07, -1.1548e-07, 1.6056e-06, + 6.3330e-08, 9.0897e-07, 5.0291e-08, 2.2352e-07, -1.2219e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 215.27, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4602 re_mapping 0.0042 re_causal 0.0117 /// teacc 99.02 lr 0.00010000 +Epoch 385, weight, value: tensor([[-3.2223e-01, 8.5241e-02, -2.0814e-01, ..., -1.0794e-01, + -2.8739e-01, -1.7551e-01], + [-6.8556e-03, 1.0541e-01, -1.2719e-01, ..., -1.3416e-01, + -8.2304e-02, 1.3054e-01], + [ 5.0199e-02, -1.4680e-01, -1.8611e-01, ..., -1.9148e-04, + -6.3064e-02, -1.1973e-01], + ..., + [ 1.1415e-01, -9.9216e-02, 1.4543e-01, ..., 7.6219e-02, + 2.2356e-01, -2.4042e-02], + [ 1.3386e-01, -2.5168e-01, -1.6753e-01, ..., -3.5927e-01, + -1.0972e-01, 2.2571e-01], + [-2.1157e-01, 9.3685e-02, 7.1602e-02, ..., -3.3855e-01, + -1.6795e-01, -8.9241e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 1.8626e-08, 7.0781e-08, 8.3819e-08, ..., 1.1176e-08, + 3.7253e-09, 5.5879e-09], + [ 1.8626e-09, 1.8626e-09, 7.4506e-09, ..., -3.7253e-09, + 1.8626e-09, 0.0000e+00], + ..., + [-4.8429e-08, 3.7253e-09, -3.3528e-08, ..., -1.8626e-09, + -2.0489e-08, 0.0000e+00], + [-0.0000e+00, 1.4901e-08, 1.8626e-08, ..., 3.7253e-09, + 0.0000e+00, -1.8626e-09], + [ 1.8626e-08, -1.2293e-07, -1.0617e-07, ..., 3.7253e-09, + 9.3132e-09, -5.5879e-09]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0165, -0.0206, -0.0116, -0.0162, 0.0029, -0.0061, 0.0091, 0.0222, + 0.0147, -0.0089], device='cuda:0'), grad: tensor([-3.5390e-08, 4.2468e-07, -5.5879e-09, -5.7742e-08, 6.8918e-08, + 2.0489e-08, 9.4995e-08, -5.7742e-08, 7.8231e-08, -5.4576e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 215.02, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4278 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.06 lr 0.00010000 +Epoch 386, weight, value: tensor([[-3.2140e-01, 8.5284e-02, -2.0821e-01, ..., -1.0810e-01, + -2.8746e-01, -1.7543e-01], + [-6.8479e-03, 1.0463e-01, -1.2720e-01, ..., -1.3442e-01, + -8.2278e-02, 1.3073e-01], + [ 5.0274e-02, -1.4742e-01, -1.8612e-01, ..., -1.0763e-04, + -6.2974e-02, -1.1990e-01], + ..., + [ 1.1411e-01, -9.9288e-02, 1.4542e-01, ..., 7.6160e-02, + 2.2351e-01, -2.4199e-02], + [ 1.3429e-01, -2.5230e-01, -1.6791e-01, ..., -3.5979e-01, + -1.0993e-01, 2.2599e-01], + [-2.1176e-01, 9.3991e-02, 7.1691e-02, ..., -3.3900e-01, + -1.6797e-01, -9.0394e-02]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.8626e-09, 3.7253e-09, ..., 3.7253e-09, + 0.0000e+00, 1.8626e-09], + [-1.9930e-07, 0.0000e+00, -6.9104e-07, ..., 1.3039e-08, + -9.5740e-07, -1.6484e-06], + [-1.8626e-08, 5.5879e-09, 7.4506e-08, ..., 4.8429e-08, + 7.4506e-09, 1.1176e-08], + ..., + [ 2.3097e-07, 1.8626e-09, 7.0222e-07, ..., 5.5879e-09, + 9.4622e-07, 1.6857e-06], + [-3.5390e-08, 1.8626e-09, 2.0489e-08, ..., 3.1665e-08, + 1.8626e-09, -1.3225e-07], + [ 2.2352e-08, 1.8626e-09, 5.5879e-09, ..., 3.7253e-09, + 3.7253e-09, 3.3528e-08]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0164, -0.0206, -0.0114, -0.0162, 0.0035, -0.0061, 0.0090, 0.0221, + 0.0141, -0.0087], device='cuda:0'), grad: tensor([ 1.6764e-08, -2.8275e-06, 1.0431e-07, -3.0175e-07, -7.4506e-09, + 1.1921e-07, -2.0489e-08, 2.9597e-06, -1.2107e-07, 8.1956e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 215.10, cls_loss 0.0007 cls_loss_mapping 0.0008 cls_loss_causal 0.4783 re_mapping 0.0039 re_causal 0.0124 /// teacc 99.03 lr 0.00010000 +Epoch 387, weight, value: tensor([[-3.2151e-01, 8.5298e-02, -2.0825e-01, ..., -1.0823e-01, + -2.8753e-01, -1.7545e-01], + [-6.9461e-03, 1.0449e-01, -1.2734e-01, ..., -1.3466e-01, + -8.2411e-02, 1.3080e-01], + [ 5.0275e-02, -1.4783e-01, -1.8648e-01, ..., -1.9148e-04, + -6.2983e-02, -1.1996e-01], + ..., + [ 1.1419e-01, -9.9868e-02, 1.4543e-01, ..., 7.5978e-02, + 2.2364e-01, -2.4203e-02], + [ 1.3453e-01, -2.5296e-01, -1.6823e-01, ..., -3.6009e-01, + -1.1003e-01, 2.2595e-01], + [-2.1183e-01, 9.4231e-02, 7.1800e-02, ..., -3.3933e-01, + -1.6798e-01, -9.0938e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-7.4506e-09, 7.4506e-09, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -1.6764e-08], + [ 7.4506e-09, -1.0990e-07, -1.8440e-07, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0164, -0.0207, -0.0115, -0.0157, 0.0038, -0.0061, 0.0090, 0.0221, + 0.0139, -0.0087], device='cuda:0'), grad: tensor([ 5.5879e-09, 9.3132e-09, 5.5879e-09, 9.3132e-09, 2.6822e-07, + -1.8626e-08, 1.6764e-08, 6.8918e-08, -9.3132e-09, -3.5204e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 214.84, cls_loss 0.0007 cls_loss_mapping 0.0009 cls_loss_causal 0.4506 re_mapping 0.0040 re_causal 0.0115 /// teacc 99.07 lr 0.00010000 +Epoch 388, weight, value: tensor([[-3.2127e-01, 8.5353e-02, -2.0822e-01, ..., -1.0829e-01, + -2.8754e-01, -1.7547e-01], + [-6.9585e-03, 1.0458e-01, -1.2733e-01, ..., -1.3500e-01, + -8.2464e-02, 1.3088e-01], + [ 5.0291e-02, -1.4857e-01, -1.8662e-01, ..., -1.7818e-04, + -6.2985e-02, -1.2002e-01], + ..., + [ 1.1420e-01, -9.9980e-02, 1.4543e-01, ..., 7.5999e-02, + 2.2369e-01, -2.4268e-02], + [ 1.3485e-01, -2.5335e-01, -1.6865e-01, ..., -3.6163e-01, + -1.1037e-01, 2.2667e-01], + [-2.1204e-01, 9.4008e-02, 7.1792e-02, ..., -3.4015e-01, + -1.6800e-01, -9.1826e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 5.9605e-08, -1.6764e-08, 8.7544e-08, ..., 4.4703e-08, + 4.0978e-08, -6.3330e-08], + [ 3.1665e-08, 1.8626e-09, 4.2841e-08, ..., 2.0489e-08, + 2.0489e-08, 5.5879e-09], + ..., + [-1.0990e-07, 2.0489e-08, -1.4715e-07, ..., -6.7055e-08, + -7.6368e-08, 4.2841e-08], + [-1.8626e-09, 1.1176e-08, 1.6764e-08, ..., 3.7253e-09, + 1.8626e-09, -7.4506e-09], + [ 9.3132e-09, -2.0489e-08, -1.3039e-08, ..., 5.5879e-09, + 7.4506e-09, 9.3132e-09]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0164, -0.0207, -0.0115, -0.0158, 0.0040, -0.0058, 0.0090, 0.0221, + 0.0140, -0.0090], device='cuda:0'), grad: tensor([ 9.3132e-09, 5.2154e-08, 9.1270e-08, -1.1921e-07, 5.5879e-08, + 1.3225e-07, 1.8626e-09, -1.9185e-07, 4.4703e-08, -6.3330e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 214.84, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4720 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.05 lr 0.00010000 +Epoch 389, weight, value: tensor([[-3.1951e-01, 8.5384e-02, -2.0859e-01, ..., -1.0855e-01, + -2.8762e-01, -1.7579e-01], + [-7.0103e-03, 1.0452e-01, -1.2755e-01, ..., -1.3572e-01, + -8.2519e-02, 1.3087e-01], + [ 5.0256e-02, -1.4904e-01, -1.8703e-01, ..., -1.7403e-04, + -6.2967e-02, -1.2028e-01], + ..., + [ 1.1423e-01, -1.0038e-01, 1.4543e-01, ..., 7.6065e-02, + 2.2373e-01, -2.4355e-02], + [ 1.3716e-01, -2.5459e-01, -1.6936e-01, ..., -3.6219e-01, + -1.0988e-01, 2.2842e-01], + [-2.1236e-01, 9.3834e-02, 7.3113e-02, ..., -3.4131e-01, + -1.6792e-01, -9.3531e-02]], device='cuda:0'), grad: tensor([[ 3.7253e-08, 0.0000e+00, 1.1176e-08, ..., 5.2154e-08, + 2.9802e-08, 7.4506e-09], + [ 7.0781e-08, 1.3039e-08, 4.2841e-08, ..., 1.0617e-07, + 5.5879e-08, -3.3267e-06], + [-2.8871e-07, 9.3132e-09, 2.9802e-08, ..., -3.2596e-07, + -2.3469e-07, 2.8927e-06], + ..., + [ 1.0803e-07, 9.3132e-09, 3.7253e-09, ..., 1.5274e-07, + 9.1270e-08, 3.2037e-07], + [ 1.8626e-08, 9.3132e-09, 2.6077e-08, ..., 5.0291e-08, + 2.0489e-08, 1.3039e-08], + [ 7.4506e-09, 1.6764e-08, 1.1176e-08, ..., 2.2352e-08, + 5.5879e-09, 1.8626e-09]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0167, -0.0209, -0.0115, -0.0158, 0.0050, -0.0061, 0.0096, 0.0220, + 0.0144, -0.0087], device='cuda:0'), grad: tensor([ 2.1607e-07, -6.7353e-06, 4.8950e-06, -2.5146e-07, 9.4995e-08, + -4.0978e-08, 1.8068e-07, 1.2927e-06, 2.1793e-07, 1.0431e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 214.88, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4599 re_mapping 0.0041 re_causal 0.0115 /// teacc 99.01 lr 0.00010000 +Epoch 390, weight, value: tensor([[-3.1966e-01, 8.5417e-02, -2.0882e-01, ..., -1.0883e-01, + -2.8767e-01, -1.7586e-01], + [-7.1106e-03, 1.0464e-01, -1.2756e-01, ..., -1.3652e-01, + -8.2541e-02, 1.3091e-01], + [ 5.0391e-02, -1.4952e-01, -1.8761e-01, ..., -2.2423e-04, + -6.2958e-02, -1.2013e-01], + ..., + [ 1.1427e-01, -1.0022e-01, 1.4548e-01, ..., 7.6097e-02, + 2.2379e-01, -2.4361e-02], + [ 1.3762e-01, -2.5477e-01, -1.7044e-01, ..., -3.6356e-01, + -1.1054e-01, 2.2908e-01], + [-2.1252e-01, 9.3816e-02, 7.3209e-02, ..., -3.4229e-01, + -1.6796e-01, -9.4384e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-09, 2.3283e-07, 1.1176e-08, ..., 2.7940e-09, + 3.7253e-09, 9.4995e-08], + [ 1.0245e-08, 9.3132e-10, 9.3132e-10, ..., -9.3132e-10, + -0.0000e+00, 1.7695e-08], + ..., + [-7.4506e-09, 1.3039e-08, -1.3970e-08, ..., -7.4506e-09, + -8.3819e-09, 1.1176e-08], + [-2.9802e-08, 2.7940e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, -4.4703e-08], + [ 9.3132e-10, -9.3132e-10, -2.1420e-08, ..., 9.3132e-10, + 9.3132e-10, 3.7253e-09]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0167, -0.0212, -0.0109, -0.0156, 0.0048, -0.0061, 0.0096, 0.0220, + 0.0144, -0.0089], device='cuda:0'), grad: tensor([-2.7940e-09, 6.0536e-07, 2.0489e-08, 8.3819e-09, -6.6776e-07, + 1.9558e-08, 7.2643e-08, 1.7695e-08, -3.5390e-08, -3.5390e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 214.91, cls_loss 0.0009 cls_loss_mapping 0.0010 cls_loss_causal 0.4698 re_mapping 0.0041 re_causal 0.0113 /// teacc 99.06 lr 0.00010000 +Epoch 391, weight, value: tensor([[-3.1827e-01, 8.5570e-02, -2.0935e-01, ..., -1.0915e-01, + -2.8779e-01, -1.7577e-01], + [-7.1549e-03, 1.0479e-01, -1.2751e-01, ..., -1.3688e-01, + -8.2548e-02, 1.3108e-01], + [ 5.0438e-02, -1.5051e-01, -1.8783e-01, ..., -2.1166e-04, + -6.2938e-02, -1.2037e-01], + ..., + [ 1.1429e-01, -1.0098e-01, 1.4552e-01, ..., 7.6134e-02, + 2.2383e-01, -2.4480e-02], + [ 1.3830e-01, -2.5533e-01, -1.7153e-01, ..., -3.6347e-01, + -1.1119e-01, 2.2973e-01], + [-2.1300e-01, 9.3569e-02, 7.3345e-02, ..., -3.4356e-01, + -1.6803e-01, -9.6333e-02]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.7940e-09, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 1.0245e-08, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 1.6764e-08], + [-6.5193e-09, 9.3132e-10, 0.0000e+00, ..., -9.3132e-09, + -6.5193e-09, 1.8626e-09], + ..., + [ 7.4506e-09, 2.7940e-09, -0.0000e+00, ..., 3.7253e-09, + 1.8626e-09, 1.3039e-08], + [-1.2107e-07, 9.3132e-10, -1.3039e-08, ..., 1.8626e-09, + 9.3132e-10, -2.1514e-07], + [ 9.9652e-08, 1.8347e-07, -4.6566e-09, ..., 9.3132e-10, + 9.3132e-10, 2.0955e-07]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0165, -0.0214, -0.0105, -0.0156, 0.0049, -0.0060, 0.0096, 0.0220, + 0.0144, -0.0094], device='cuda:0'), grad: tensor([ 2.5146e-08, 3.5390e-08, -3.2596e-08, 9.3132e-09, -8.5123e-07, + 2.7940e-09, -2.1420e-08, 4.0978e-08, -3.8557e-07, 1.1800e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 214.75, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4912 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.01 lr 0.00010000 +Epoch 392, weight, value: tensor([[-3.1837e-01, 8.5508e-02, -2.0998e-01, ..., -1.0934e-01, + -2.8786e-01, -1.7579e-01], + [-7.1612e-03, 1.0487e-01, -1.2753e-01, ..., -1.3714e-01, + -8.2639e-02, 1.3131e-01], + [ 5.0486e-02, -1.5131e-01, -1.8801e-01, ..., -1.1163e-04, + -6.2933e-02, -1.2063e-01], + ..., + [ 1.1431e-01, -1.0183e-01, 1.4556e-01, ..., 7.6137e-02, + 2.2393e-01, -2.4623e-02], + [ 1.3811e-01, -2.5578e-01, -1.7251e-01, ..., -3.6563e-01, + -1.1174e-01, 2.3035e-01], + [-2.1317e-01, 9.3884e-02, 7.3522e-02, ..., -3.4447e-01, + -1.6805e-01, -9.6865e-02]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8626e-09, 2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-09], + [ 9.2201e-08, -2.7940e-08, 1.3411e-07, ..., 2.4214e-08, + 1.2014e-07, -3.5390e-08], + [ 1.3039e-08, 5.5879e-09, 1.5832e-08, ..., 1.1176e-08, + 1.1176e-08, 1.3039e-08], + ..., + [-1.1548e-07, 1.3039e-08, -1.6578e-07, ..., -3.5390e-08, + -1.4994e-07, -3.7253e-09], + [-1.8626e-08, 6.5193e-09, 9.3132e-09, ..., 2.7940e-09, + 0.0000e+00, -1.7695e-08], + [ 2.5146e-08, -2.2352e-08, -1.6764e-08, ..., 3.7253e-09, + 1.5832e-08, 2.1420e-08]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0166, -0.0213, -0.0104, -0.0159, 0.0046, -0.0059, 0.0097, 0.0220, + 0.0143, -0.0092], device='cuda:0'), grad: tensor([ 1.6764e-08, 5.2154e-08, 6.6124e-08, 4.4703e-08, 2.5146e-08, + 7.4506e-09, 1.8626e-08, -1.9185e-07, -7.4506e-09, -3.6322e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 214.82, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4509 re_mapping 0.0039 re_causal 0.0114 /// teacc 98.96 lr 0.00010000 +Epoch 393, weight, value: tensor([[-3.1842e-01, 8.5310e-02, -2.1026e-01, ..., -1.0985e-01, + -2.8795e-01, -1.7623e-01], + [-7.7054e-03, 1.0587e-01, -1.2821e-01, ..., -1.3764e-01, + -8.3463e-02, 1.3148e-01], + [ 5.0429e-02, -1.5453e-01, -1.8836e-01, ..., -1.9264e-04, + -6.2962e-02, -1.2185e-01], + ..., + [ 1.1483e-01, -1.0171e-01, 1.4627e-01, ..., 7.6206e-02, + 2.2472e-01, -2.4285e-02], + [ 1.3800e-01, -2.5657e-01, -1.7511e-01, ..., -3.6661e-01, + -1.1415e-01, 2.3035e-01], + [-2.1343e-01, 9.5009e-02, 7.4804e-02, ..., -3.4556e-01, + -1.6811e-01, -9.7190e-02]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -4.0885e-07, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 4.6566e-09, 1.7695e-08, 6.5193e-09, ..., 9.3132e-10, + 2.7940e-09, 4.8429e-08], + [ 2.7940e-09, 8.3819e-09, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 3.7253e-09], + ..., + [-1.0990e-07, 1.5832e-08, -1.6578e-07, ..., -9.3132e-10, + -1.7043e-07, 2.0489e-08], + [ 9.3132e-10, 1.0245e-08, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 7.4506e-09, 3.8743e-07, -1.8626e-08, ..., 0.0000e+00, + 6.5193e-09, 2.0210e-07]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0169, -0.0215, -0.0109, -0.0153, 0.0018, -0.0063, 0.0100, 0.0225, + 0.0138, -0.0084], device='cuda:0'), grad: tensor([-8.7451e-07, 1.8720e-07, 2.8871e-08, 1.0245e-08, -4.7591e-07, + 4.6566e-09, 2.0489e-08, -3.8277e-07, 3.9116e-08, 1.4268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 214.79, cls_loss 0.0010 cls_loss_mapping 0.0010 cls_loss_causal 0.4538 re_mapping 0.0041 re_causal 0.0112 /// teacc 99.04 lr 0.00010000 +Epoch 394, weight, value: tensor([[-3.1859e-01, 8.5213e-02, -2.1132e-01, ..., -1.1034e-01, + -2.8824e-01, -1.7630e-01], + [-7.8000e-03, 1.0551e-01, -1.2843e-01, ..., -1.3819e-01, + -8.3611e-02, 1.3137e-01], + [ 5.0432e-02, -1.5506e-01, -1.8851e-01, ..., -1.6036e-04, + -6.2941e-02, -1.2204e-01], + ..., + [ 1.1493e-01, -1.0225e-01, 1.4650e-01, ..., 7.6275e-02, + 2.2489e-01, -2.4219e-02], + [ 1.3902e-01, -2.5729e-01, -1.7533e-01, ..., -3.6667e-01, + -1.1460e-01, 2.3135e-01], + [-2.1368e-01, 9.5764e-02, 7.5636e-02, ..., -3.4778e-01, + -1.6819e-01, -1.0165e-01]], device='cuda:0'), grad: tensor([[ 6.5193e-09, -1.7695e-08, 4.6566e-09, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + [ 1.7695e-08, 7.4506e-09, 1.6764e-08, ..., 4.1910e-08, + 1.1176e-08, -1.8626e-09], + [-7.1712e-08, 2.7940e-09, 4.6566e-09, ..., -5.2154e-08, + -1.8626e-08, -0.0000e+00], + ..., + [ 4.6566e-09, 4.6566e-09, 8.3819e-09, ..., 1.3039e-08, + 1.8626e-09, 9.3132e-10], + [ 2.7940e-08, 5.5879e-09, 6.5193e-09, ..., 1.6764e-08, + 1.8626e-09, 9.3132e-10], + [ 1.8626e-09, 8.3819e-09, 5.5879e-09, ..., 6.5193e-09, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0171, -0.0217, -0.0109, -0.0159, 0.0025, -0.0058, 0.0100, 0.0226, + 0.0142, -0.0081], device='cuda:0'), grad: tensor([-4.4703e-08, 1.6391e-07, -4.7963e-07, -3.6508e-07, 3.7253e-09, + 3.2317e-07, 1.1269e-07, 4.9360e-08, 2.0396e-07, 4.0978e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 215.01, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4731 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +Epoch 395, weight, value: tensor([[-3.1851e-01, 8.5286e-02, -2.1150e-01, ..., -1.1049e-01, + -2.8841e-01, -1.7626e-01], + [-7.9529e-03, 1.0539e-01, -1.2859e-01, ..., -1.3847e-01, + -8.3835e-02, 1.3134e-01], + [ 5.0488e-02, -1.5568e-01, -1.8866e-01, ..., -1.5392e-04, + -6.2885e-02, -1.2225e-01], + ..., + [ 1.1505e-01, -1.0263e-01, 1.4671e-01, ..., 7.6320e-02, + 2.2511e-01, -2.4177e-02], + [ 1.3889e-01, -2.5836e-01, -1.7716e-01, ..., -3.6749e-01, + -1.1583e-01, 2.3117e-01], + [-2.1397e-01, 9.5773e-02, 7.5640e-02, ..., -3.5120e-01, + -1.6831e-01, -1.0190e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 1.8626e-09], + [ 2.1420e-08, -6.5193e-08, 2.2352e-08, ..., 1.1176e-08, + 1.9558e-08, -5.4017e-08], + [-4.6566e-09, 1.9558e-08, 6.5193e-09, ..., -9.3132e-09, + -5.5879e-09, 1.2107e-08], + ..., + [-5.0291e-08, 3.6322e-08, -5.4017e-08, ..., -1.6764e-08, + -4.4703e-08, 3.0734e-08], + [ 4.6566e-09, 7.4506e-09, 2.7940e-09, ..., 3.7253e-09, + 4.6566e-09, 3.7253e-09], + [ 1.3970e-08, 5.5879e-09, 1.4901e-08, ..., 6.5193e-09, + 1.2107e-08, 5.5879e-09]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0170, -0.0218, -0.0108, -0.0159, 0.0026, -0.0056, 0.0100, 0.0227, + 0.0133, -0.0082], device='cuda:0'), grad: tensor([ 1.1176e-08, -1.2945e-07, -4.7497e-08, 2.8871e-08, 9.3132e-09, + -5.0291e-08, 2.3283e-08, 4.7497e-08, 5.3085e-08, 5.9605e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 215.10, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4515 re_mapping 0.0039 re_causal 0.0111 /// teacc 98.99 lr 0.00010000 +Epoch 396, weight, value: tensor([[-3.1988e-01, 8.5339e-02, -2.1177e-01, ..., -1.1277e-01, + -2.9076e-01, -1.7627e-01], + [-7.9469e-03, 1.0576e-01, -1.2862e-01, ..., -1.3876e-01, + -8.3839e-02, 1.3153e-01], + [ 5.0514e-02, -1.5625e-01, -1.8890e-01, ..., -6.2472e-05, + -6.2731e-02, -1.2250e-01], + ..., + [ 1.1513e-01, -1.0362e-01, 1.4694e-01, ..., 7.6377e-02, + 2.2526e-01, -2.4261e-02], + [ 1.3886e-01, -2.5922e-01, -1.7826e-01, ..., -3.6872e-01, + -1.1758e-01, 2.3123e-01], + [-2.1476e-01, 9.5815e-02, 7.5207e-02, ..., -3.5395e-01, + -1.6886e-01, -1.0220e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.2107e-08, 1.8626e-09, 1.3039e-08, ..., 1.0245e-08, + 1.0245e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., -0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [-1.7695e-08, -2.7940e-09, -2.0489e-08, ..., -1.3970e-08, + -1.3039e-08, 9.3132e-10], + [-0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0171, -0.0218, -0.0107, -0.0159, 0.0023, -0.0050, 0.0097, 0.0227, + 0.0126, -0.0085], device='cuda:0'), grad: tensor([-8.3819e-09, 3.2596e-08, -7.4506e-09, 3.6322e-08, 9.3132e-10, + -2.6077e-08, -2.7940e-09, -3.9116e-08, 2.7940e-09, 1.4901e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 214.79, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4820 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.04 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.3201, 0.0854, -0.2120, ..., -0.1135, -0.2909, -0.1763], + [-0.0081, 0.1058, -0.1288, ..., -0.1398, -0.0840, 0.1315], + [ 0.0503, -0.1556, -0.1908, ..., -0.0004, -0.0631, -0.1226], + ..., + [ 0.1154, -0.1037, 0.1474, ..., 0.0768, 0.2256, -0.0243], + [ 0.1399, -0.2599, -0.1789, ..., -0.3695, -0.1184, 0.2324], + [-0.2150, 0.0957, 0.0754, ..., -0.3557, -0.1690, -0.1026]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -9.3132e-09, 1.8626e-09, ..., 3.7253e-09, + 0.0000e+00, 4.6566e-09], + [ 3.0734e-08, 6.5193e-09, 2.3283e-08, ..., 2.3283e-08, + 1.2107e-08, -3.1665e-08], + [-2.7940e-07, 7.4506e-09, 6.5193e-09, ..., 6.5193e-09, + 1.8626e-09, -3.5390e-08], + ..., + [ 3.7253e-08, 5.0291e-08, -9.3132e-09, ..., 6.5193e-09, + -1.7695e-08, 7.5437e-08], + [ 1.0245e-07, 4.6566e-09, 4.6566e-09, ..., 5.5879e-09, + 9.3132e-10, 1.9558e-08], + [ 3.7253e-09, 1.8626e-08, 2.7940e-09, ..., 9.3132e-10, + 1.8626e-09, 1.6764e-08]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0171, -0.0219, -0.0109, -0.0161, 0.0019, -0.0036, 0.0094, 0.0230, + 0.0133, -0.0087], device='cuda:0'), grad: tensor([ 1.8626e-09, 6.7055e-08, -5.0012e-07, -2.0582e-07, -1.3784e-07, + 8.0094e-08, 1.1176e-07, 2.7660e-07, 2.0862e-07, 7.5437e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 214.92, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4638 re_mapping 0.0041 re_causal 0.0116 /// teacc 99.08 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.3209, 0.0854, -0.2121, ..., -0.1138, -0.2910, -0.1766], + [-0.0081, 0.1058, -0.1288, ..., -0.1403, -0.0841, 0.1316], + [ 0.0503, -0.1559, -0.1910, ..., -0.0004, -0.0631, -0.1227], + ..., + [ 0.1155, -0.1049, 0.1475, ..., 0.0769, 0.2257, -0.0243], + [ 0.1412, -0.2612, -0.1798, ..., -0.3706, -0.1199, 0.2340], + [-0.2156, 0.0956, 0.0754, ..., -0.3575, -0.1691, -0.1043]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.8231e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-3.7253e-09, 1.8626e-09, -1.8626e-09, ..., -1.8626e-09, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 3.1665e-08, -3.7253e-08, ..., 0.0000e+00, + 1.8626e-09, -0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0173, -0.0219, -0.0109, -0.0161, 0.0025, -0.0036, 0.0097, 0.0230, + 0.0136, -0.0091], device='cuda:0'), grad: tensor([-1.4529e-07, 1.8626e-09, 1.8626e-09, 1.8626e-09, 1.1921e-07, + 1.1176e-08, 1.8626e-09, 0.0000e+00, 1.3039e-08, -7.4506e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 214.71, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4579 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.08 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.3205, 0.0864, -0.2120, ..., -0.1133, -0.2910, -0.1765], + [-0.0081, 0.1055, -0.1289, ..., -0.1407, -0.0841, 0.1316], + [ 0.0503, -0.1562, -0.1911, ..., -0.0004, -0.0631, -0.1227], + ..., + [ 0.1155, -0.1053, 0.1475, ..., 0.0770, 0.2258, -0.0243], + [ 0.1414, -0.2626, -0.1806, ..., -0.3718, -0.1204, 0.2341], + [-0.2157, 0.0953, 0.0755, ..., -0.3585, -0.1691, -0.1046]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-09, 9.3132e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-0.0000e+00, -5.5879e-09, -1.6764e-08, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0155, -0.0220, -0.0109, -0.0182, 0.0024, -0.0018, 0.0082, 0.0230, + 0.0128, -0.0092], device='cuda:0'), grad: tensor([ 1.6764e-08, 7.4506e-09, 2.2352e-08, 7.4506e-09, 3.7998e-07, + 1.1176e-07, -5.9232e-07, 3.1665e-08, 4.0978e-08, -2.2352e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 214.96, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4737 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.02 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.3203, 0.0867, -0.2122, ..., -0.1133, -0.2912, -0.1762], + [-0.0082, 0.1055, -0.1289, ..., -0.1408, -0.0842, 0.1316], + [ 0.0503, -0.1564, -0.1911, ..., -0.0004, -0.0630, -0.1225], + ..., + [ 0.1155, -0.1061, 0.1476, ..., 0.0769, 0.2258, -0.0244], + [ 0.1415, -0.2638, -0.1821, ..., -0.3727, -0.1219, 0.2342], + [-0.2158, 0.0948, 0.0757, ..., -0.3588, -0.1691, -0.1057]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.1176e-08, -1.8626e-09, 1.1176e-08, ..., 1.8626e-09, + 1.3039e-08, -5.5879e-09], + [-1.8626e-09, 0.0000e+00, 5.5879e-09, ..., -3.7253e-09, + 3.7253e-09, 0.0000e+00], + ..., + [-6.1467e-08, -0.0000e+00, -5.2154e-08, ..., -3.7253e-09, + -5.2154e-08, 3.7253e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.3528e-08, 1.4901e-08, 2.4214e-08, ..., 1.8626e-09, + 2.4214e-08, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0152, -0.0220, -0.0108, -0.0185, 0.0025, -0.0016, 0.0082, 0.0230, + 0.0123, -0.0097], device='cuda:0'), grad: tensor([ 2.4214e-08, 1.8626e-08, -9.3132e-09, 1.3039e-08, 1.8626e-08, + -4.4703e-08, -1.1176e-08, -1.2852e-07, 1.1176e-08, 1.0617e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 215.07, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4474 re_mapping 0.0039 re_causal 0.0114 /// teacc 98.96 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.3205, 0.0867, -0.2124, ..., -0.1135, -0.2912, -0.1762], + [-0.0082, 0.1056, -0.1290, ..., -0.1412, -0.0842, 0.1317], + [ 0.0503, -0.1567, -0.1912, ..., -0.0004, -0.0630, -0.1228], + ..., + [ 0.1155, -0.1073, 0.1476, ..., 0.0769, 0.2259, -0.0245], + [ 0.1434, -0.2645, -0.1818, ..., -0.3725, -0.1221, 0.2359], + [-0.2167, 0.0942, 0.0754, ..., -0.3595, -0.1692, -0.1081]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, -1.4901e-08, 1.1176e-08, ..., 3.7253e-09, + 1.1176e-08, -3.7253e-09], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + -3.7253e-09, 0.0000e+00], + ..., + [-1.1176e-08, 1.1176e-08, -1.4901e-08, ..., -3.7253e-09, + -1.4901e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.2154e-08]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0152, -0.0220, -0.0109, -0.0193, 0.0037, -0.0009, 0.0084, 0.0230, + 0.0133, -0.0106], device='cuda:0'), grad: tensor([ 0.0000e+00, -2.6077e-08, -7.4506e-09, -3.7253e-09, -2.0117e-07, + -5.5879e-08, 1.8626e-08, 1.4901e-08, 1.1176e-08, 2.1979e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 215.12, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4432 re_mapping 0.0038 re_causal 0.0108 /// teacc 98.96 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.3206, 0.0867, -0.2124, ..., -0.1135, -0.2912, -0.1762], + [-0.0082, 0.1056, -0.1290, ..., -0.1413, -0.0842, 0.1317], + [ 0.0503, -0.1568, -0.1912, ..., -0.0004, -0.0629, -0.1229], + ..., + [ 0.1155, -0.1075, 0.1476, ..., 0.0769, 0.2259, -0.0245], + [ 0.1436, -0.2646, -0.1818, ..., -0.3724, -0.1221, 0.2361], + [-0.2170, 0.0941, 0.0754, ..., -0.3596, -0.1692, -0.1084]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + -0.0000e+00, 3.7253e-09], + ..., + [-0.0000e+00, 0.0000e+00, -0.0000e+00, ..., -0.0000e+00, + -3.7253e-09, 3.7253e-09], + [-3.7253e-09, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, -0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0152, -0.0221, -0.0108, -0.0195, 0.0039, -0.0006, 0.0084, 0.0230, + 0.0133, -0.0108], device='cuda:0'), grad: tensor([ 4.8056e-07, 2.6077e-08, 1.1176e-08, 7.4506e-09, 3.3900e-07, + 1.2293e-07, -1.0133e-06, 7.4506e-09, 0.0000e+00, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 215.15, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4554 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.03 lr 0.00001000 +Epoch 403, weight, value: tensor([[-3.2057e-01, 8.6725e-02, -2.1239e-01, ..., -1.1348e-01, + -2.9122e-01, -1.7622e-01], + [-8.2493e-03, 1.0558e-01, -1.2905e-01, ..., -1.4141e-01, + -8.4256e-02, 1.3166e-01], + [ 5.0345e-02, -1.5681e-01, -1.9119e-01, ..., -3.6259e-04, + -6.2901e-02, -1.2279e-01], + ..., + [ 1.1555e-01, -1.0753e-01, 1.4768e-01, ..., 7.6932e-02, + 2.2590e-01, -2.4496e-02], + [ 1.4362e-01, -2.6465e-01, -1.8183e-01, ..., -3.7246e-01, + -1.2211e-01, 2.3614e-01], + [-2.1698e-01, 9.4074e-02, 7.5402e-02, ..., -3.5971e-01, + -1.6922e-01, -1.0844e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.5682e-08, 4.4703e-08, ..., 3.7253e-08, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, 1.1176e-08, ..., 1.4901e-08, + 3.7253e-09, -2.2352e-08], + [-0.0000e+00, 4.8429e-08, 7.0781e-08, ..., 8.5682e-08, + 3.7253e-09, 0.0000e+00], + ..., + [-1.1176e-08, 1.1176e-08, -0.0000e+00, ..., 7.4506e-09, + -1.1176e-08, 1.8626e-08], + [-1.8626e-08, 1.1176e-08, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, -2.2352e-08], + [ 3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0152, -0.0221, -0.0108, -0.0195, 0.0039, -0.0005, 0.0084, 0.0230, + 0.0132, -0.0108], device='cuda:0'), grad: tensor([-1.2293e-07, -7.4506e-09, 2.2724e-07, -4.4703e-07, -1.4901e-08, + 2.4959e-07, 2.2352e-08, 5.5879e-08, -2.2352e-08, 3.3528e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 215.04, cls_loss 0.0008 cls_loss_mapping 0.0007 cls_loss_causal 0.4496 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.05 lr 0.00001000 +Epoch 404, weight, value: tensor([[-3.2058e-01, 8.6713e-02, -2.1241e-01, ..., -1.1349e-01, + -2.9123e-01, -1.7622e-01], + [-8.2560e-03, 1.0557e-01, -1.2906e-01, ..., -1.4148e-01, + -8.4264e-02, 1.3166e-01], + [ 5.0353e-02, -1.5683e-01, -1.9120e-01, ..., -3.5259e-04, + -6.2889e-02, -1.2278e-01], + ..., + [ 1.1555e-01, -1.0760e-01, 1.4768e-01, ..., 7.6927e-02, + 2.2590e-01, -2.4508e-02], + [ 1.4368e-01, -2.6470e-01, -1.8185e-01, ..., -3.7249e-01, + -1.2214e-01, 2.3623e-01], + [-2.1702e-01, 9.4105e-02, 7.5404e-02, ..., -3.5983e-01, + -1.6923e-01, -1.0851e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1176e-08, 3.7253e-09, 7.4506e-09, ..., 1.1176e-08, + 0.0000e+00, 1.1176e-08], + ..., + [ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-1.4901e-08, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -1.1176e-08], + [-3.7253e-09, -0.0000e+00, -1.4901e-08, ..., 0.0000e+00, + -0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0152, -0.0221, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0108], device='cuda:0'), grad: tensor([ 3.7253e-09, 2.9802e-08, 5.2154e-08, -4.0978e-08, -4.4703e-08, + 3.7253e-09, -1.4901e-08, 1.4901e-08, -2.2352e-08, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 215.01, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4597 re_mapping 0.0035 re_causal 0.0109 /// teacc 99.04 lr 0.00001000 +Epoch 405, weight, value: tensor([[-3.2060e-01, 8.6720e-02, -2.1242e-01, ..., -1.1350e-01, + -2.9124e-01, -1.7623e-01], + [-8.2655e-03, 1.0557e-01, -1.2907e-01, ..., -1.4154e-01, + -8.4278e-02, 1.3166e-01], + [ 5.0358e-02, -1.5685e-01, -1.9120e-01, ..., -3.4654e-04, + -6.2879e-02, -1.2277e-01], + ..., + [ 1.1556e-01, -1.0765e-01, 1.4770e-01, ..., 7.6925e-02, + 2.2591e-01, -2.4510e-02], + [ 1.4375e-01, -2.6474e-01, -1.8187e-01, ..., -3.7253e-01, + -1.2217e-01, 2.3630e-01], + [-2.1705e-01, 9.4092e-02, 7.5402e-02, ..., -3.5993e-01, + -1.6924e-01, -1.0857e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 3.7253e-09, ..., 3.7253e-09, + 0.0000e+00, -4.0978e-08], + [ 0.0000e+00, 7.4506e-09, 1.1176e-08, ..., 7.4506e-09, + 0.0000e+00, 3.7253e-09], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-0.0000e+00, -0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0152, -0.0221, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0133, -0.0109], device='cuda:0'), grad: tensor([ 7.4506e-09, -5.9605e-08, 2.9802e-08, -5.9605e-08, 4.0978e-08, + 7.4506e-09, 1.1176e-08, 1.4901e-08, 3.7253e-09, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 215.25, cls_loss 0.0007 cls_loss_mapping 0.0006 cls_loss_causal 0.4557 re_mapping 0.0035 re_causal 0.0109 /// teacc 99.12 lr 0.00001000 +Epoch 406, weight, value: tensor([[-3.2062e-01, 8.6722e-02, -2.1244e-01, ..., -1.1352e-01, + -2.9124e-01, -1.7623e-01], + [-8.2704e-03, 1.0557e-01, -1.2908e-01, ..., -1.4162e-01, + -8.4286e-02, 1.3167e-01], + [ 5.0355e-02, -1.5688e-01, -1.9125e-01, ..., -3.4780e-04, + -6.2885e-02, -1.2277e-01], + ..., + [ 1.1557e-01, -1.0768e-01, 1.4771e-01, ..., 7.6936e-02, + 2.2593e-01, -2.4521e-02], + [ 1.4380e-01, -2.6478e-01, -1.8190e-01, ..., -3.7257e-01, + -1.2217e-01, 2.3635e-01], + [-2.1709e-01, 9.4078e-02, 7.5398e-02, ..., -3.6005e-01, + -1.6924e-01, -1.0863e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + [ 0.0000e+00, -1.4901e-08, 1.4901e-08, ..., 7.4506e-09, + 1.1176e-08, -5.9605e-08], + [ 7.4506e-09, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + 7.4506e-09, 0.0000e+00], + ..., + [-2.6077e-08, 1.4901e-08, -3.7253e-08, ..., -1.4901e-08, + -3.3528e-08, 4.4703e-08], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.0978e-08, 1.1176e-08, ..., 7.4506e-09, + 3.7253e-09, 4.8429e-08]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0152, -0.0221, -0.0108, -0.0196, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0133, -0.0109], device='cuda:0'), grad: tensor([-4.8429e-08, -9.6858e-08, 1.8626e-08, 2.7940e-07, -2.3097e-07, + -2.9802e-07, 3.7253e-09, 3.3528e-08, 7.4506e-09, 3.0175e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 214.97, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4370 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.11 lr 0.00001000 +Epoch 407, weight, value: tensor([[-3.2062e-01, 8.6729e-02, -2.1243e-01, ..., -1.1352e-01, + -2.9125e-01, -1.7623e-01], + [-8.2816e-03, 1.0557e-01, -1.2911e-01, ..., -1.4168e-01, + -8.4307e-02, 1.3167e-01], + [ 5.0355e-02, -1.5695e-01, -1.9128e-01, ..., -3.4631e-04, + -6.2885e-02, -1.2276e-01], + ..., + [ 1.1558e-01, -1.0776e-01, 1.4775e-01, ..., 7.6940e-02, + 2.2595e-01, -2.4523e-02], + [ 1.4381e-01, -2.6482e-01, -1.8192e-01, ..., -3.7260e-01, + -1.2219e-01, 2.3637e-01], + [-2.1711e-01, 9.4084e-02, 7.5395e-02, ..., -3.6013e-01, + -1.6925e-01, -1.0865e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.1234e-07, 0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, 3.7253e-09], + [-4.0978e-07, -2.9802e-08, 0.0000e+00, ..., 3.7253e-09, + -5.5879e-08, -5.4762e-07], + [ 2.6077e-08, 3.7253e-09, 0.0000e+00, ..., -3.7253e-09, + 3.7253e-09, 3.3528e-08], + ..., + [ 1.8254e-07, 1.8626e-08, 0.0000e+00, ..., 0.0000e+00, + 2.6077e-08, 2.4214e-07], + [ 1.7881e-07, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 2.6077e-08, 2.3469e-07], + [ 7.4506e-09, 1.1176e-08, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0109], device='cuda:0'), grad: tensor([-7.2643e-07, -1.1735e-06, 7.4506e-08, 2.2352e-08, 0.0000e+00, + 9.6858e-08, 5.8487e-07, 5.3644e-07, 5.2154e-07, 5.2154e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 214.90, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4339 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.11 lr 0.00001000 +Epoch 408, weight, value: tensor([[-3.2064e-01, 8.6738e-02, -2.1243e-01, ..., -1.1354e-01, + -2.9126e-01, -1.7624e-01], + [-8.3011e-03, 1.0554e-01, -1.2913e-01, ..., -1.4169e-01, + -8.4333e-02, 1.3167e-01], + [ 5.0354e-02, -1.5705e-01, -1.9130e-01, ..., -3.4760e-04, + -6.2888e-02, -1.2277e-01], + ..., + [ 1.1560e-01, -1.0780e-01, 1.4777e-01, ..., 7.6940e-02, + 2.2598e-01, -2.4516e-02], + [ 1.4384e-01, -2.6486e-01, -1.8196e-01, ..., -3.7261e-01, + -1.2222e-01, 2.3641e-01], + [-2.1713e-01, 9.4096e-02, 7.5403e-02, ..., -3.6017e-01, + -1.6926e-01, -1.0868e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.4901e-08, 3.7253e-09, 1.4901e-08, ..., 7.4506e-09, + 1.1176e-08, 3.7253e-09], + [-1.1176e-08, 0.0000e+00, 7.4506e-09, ..., 7.4506e-09, + -3.7253e-09, -2.2352e-08], + ..., + [-1.4901e-08, 0.0000e+00, -2.6077e-08, ..., -1.4901e-08, + -1.4901e-08, 1.1176e-08], + [ 7.4506e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 3.7253e-09], + [ 3.7253e-09, -3.7253e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0109], device='cuda:0'), grad: tensor([ 1.1176e-08, 4.8429e-08, -3.7253e-08, -5.2154e-08, 7.4506e-09, + -1.4901e-08, 1.8626e-08, -2.2352e-08, 4.0978e-08, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 214.73, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4486 re_mapping 0.0034 re_causal 0.0111 /// teacc 99.14 lr 0.00001000 +Epoch 409, weight, value: tensor([[-3.2066e-01, 8.6728e-02, -2.1245e-01, ..., -1.1355e-01, + -2.9126e-01, -1.7624e-01], + [-8.3043e-03, 1.0555e-01, -1.2914e-01, ..., -1.4173e-01, + -8.4336e-02, 1.3168e-01], + [ 5.0353e-02, -1.5708e-01, -1.9133e-01, ..., -3.4869e-04, + -6.2891e-02, -1.2278e-01], + ..., + [ 1.1560e-01, -1.0787e-01, 1.4778e-01, ..., 7.6942e-02, + 2.2598e-01, -2.4524e-02], + [ 1.4386e-01, -2.6489e-01, -1.8198e-01, ..., -3.7263e-01, + -1.2223e-01, 2.3643e-01], + [-2.1715e-01, 9.4121e-02, 7.5403e-02, ..., -3.6023e-01, + -1.6926e-01, -1.0870e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.8231e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.6858e-08], + [-3.7253e-09, -2.1234e-07, 2.6077e-08, ..., 1.8626e-08, + 1.1176e-08, -2.6822e-07], + [ 0.0000e+00, 7.4506e-09, 1.1176e-08, ..., 3.7253e-09, + 3.7253e-09, 7.4506e-09], + ..., + [ 2.9802e-08, 6.7055e-08, 1.3411e-07, ..., 9.3132e-08, + 5.2154e-08, 8.9407e-08], + [ 0.0000e+00, 2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0109], device='cuda:0'), grad: tensor([ 3.1665e-07, -7.7486e-07, 4.0978e-08, -5.2527e-07, 7.4506e-09, + 5.5879e-08, 5.9605e-08, 6.8918e-07, 9.6858e-08, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 214.89, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4354 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.10 lr 0.00001000 +Epoch 410, weight, value: tensor([[-3.2067e-01, 8.6732e-02, -2.1246e-01, ..., -1.1356e-01, + -2.9127e-01, -1.7624e-01], + [-8.3042e-03, 1.0558e-01, -1.2915e-01, ..., -1.4174e-01, + -8.4342e-02, 1.3170e-01], + [ 5.0354e-02, -1.5710e-01, -1.9135e-01, ..., -3.4723e-04, + -6.2892e-02, -1.2278e-01], + ..., + [ 1.1560e-01, -1.0800e-01, 1.4779e-01, ..., 7.6936e-02, + 2.2599e-01, -2.4530e-02], + [ 1.4386e-01, -2.6492e-01, -1.8201e-01, ..., -3.7267e-01, + -1.2226e-01, 2.3644e-01], + [-2.1717e-01, 9.4117e-02, 7.5414e-02, ..., -3.6025e-01, + -1.6927e-01, -1.0875e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.4901e-08, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + [-7.0781e-08, 0.0000e+00, 0.0000e+00, ..., -4.4703e-08, + -5.9605e-08, 0.0000e+00], + ..., + [ 6.7055e-08, 3.7253e-09, 0.0000e+00, ..., 4.0978e-08, + 5.5879e-08, 0.0000e+00], + [-2.6077e-08, 7.4506e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -4.4703e-08], + [ 0.0000e+00, 7.0781e-08, -1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0109], device='cuda:0'), grad: tensor([-3.7253e-08, 1.8626e-07, -1.5274e-07, 4.8429e-08, -4.7050e-06, + -5.9605e-08, 7.8231e-08, 3.9488e-07, 2.2352e-08, 4.1984e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 214.76, cls_loss 0.0006 cls_loss_mapping 0.0006 cls_loss_causal 0.4290 re_mapping 0.0032 re_causal 0.0105 /// teacc 99.12 lr 0.00001000 +Epoch 411, weight, value: tensor([[-3.2068e-01, 8.6729e-02, -2.1247e-01, ..., -1.1356e-01, + -2.9127e-01, -1.7624e-01], + [-8.3142e-03, 1.0557e-01, -1.2917e-01, ..., -1.4180e-01, + -8.4357e-02, 1.3170e-01], + [ 5.0351e-02, -1.5714e-01, -1.9141e-01, ..., -3.4868e-04, + -6.2895e-02, -1.2278e-01], + ..., + [ 1.1562e-01, -1.0803e-01, 1.4782e-01, ..., 7.6948e-02, + 2.2601e-01, -2.4528e-02], + [ 1.4388e-01, -2.6495e-01, -1.8205e-01, ..., -3.7271e-01, + -1.2230e-01, 2.3647e-01], + [-2.1720e-01, 9.4139e-02, 7.5410e-02, ..., -3.6035e-01, + -1.6928e-01, -1.0879e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.8545e-07, 6.3330e-08, 2.5332e-07, ..., 3.7253e-09, + 3.5763e-07, 2.3842e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-8.5309e-07, -4.4703e-08, -3.1665e-07, ..., 0.0000e+00, + -4.4703e-07, -2.7940e-07], + [ 5.9605e-08, 1.4901e-08, 2.2352e-08, ..., 3.7253e-09, + 3.3528e-08, 2.2352e-08], + [ 8.9407e-08, 1.3411e-07, 2.6077e-08, ..., 0.0000e+00, + 4.4703e-08, 1.3411e-07]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0109], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.3150e-06, 0.0000e+00, 4.0978e-07, -6.5193e-07, + -5.9977e-07, 1.1921e-07, -1.4901e-06, 1.6391e-07, 7.1526e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 214.44, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4290 re_mapping 0.0032 re_causal 0.0106 /// teacc 99.09 lr 0.00001000 +Epoch 412, weight, value: tensor([[-3.2070e-01, 8.6724e-02, -2.1249e-01, ..., -1.1359e-01, + -2.9128e-01, -1.7624e-01], + [-8.3191e-03, 1.0558e-01, -1.2917e-01, ..., -1.4187e-01, + -8.4360e-02, 1.3171e-01], + [ 5.0344e-02, -1.5715e-01, -1.9147e-01, ..., -3.5128e-04, + -6.2906e-02, -1.2279e-01], + ..., + [ 1.1563e-01, -1.0803e-01, 1.4784e-01, ..., 7.6962e-02, + 2.2602e-01, -2.4539e-02], + [ 1.4392e-01, -2.6499e-01, -1.8207e-01, ..., -3.7277e-01, + -1.2234e-01, 2.3653e-01], + [-2.1723e-01, 9.4159e-02, 7.5406e-02, ..., -3.6042e-01, + -1.6930e-01, -1.0882e-01]], device='cuda:0'), grad: tensor([[ 1.4901e-08, -1.8626e-08, 0.0000e+00, ..., 7.4506e-09, + 3.7253e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-4.8429e-08, 0.0000e+00, 0.0000e+00, ..., -2.6077e-08, + -2.2352e-08, 0.0000e+00], + ..., + [ 2.6077e-08, 3.7253e-09, 7.4506e-09, ..., 1.1176e-08, + 1.8626e-08, 7.4506e-09], + [-5.5879e-08, 3.7253e-09, -1.4901e-08, ..., -7.4506e-09, + 0.0000e+00, -7.8231e-08], + [ 7.4506e-09, -5.2154e-08, -7.0781e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0109], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.8626e-08, -1.6764e-07, 1.6391e-07, 1.4156e-07, + 7.4506e-09, 3.3528e-08, 7.8231e-08, -1.4156e-07, -1.4156e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 214.91, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4431 re_mapping 0.0031 re_causal 0.0107 /// teacc 99.10 lr 0.00001000 +Epoch 413, weight, value: tensor([[-3.2070e-01, 8.6737e-02, -2.1249e-01, ..., -1.1360e-01, + -2.9129e-01, -1.7624e-01], + [-8.3373e-03, 1.0560e-01, -1.2920e-01, ..., -1.4191e-01, + -8.4386e-02, 1.3172e-01], + [ 5.0347e-02, -1.5717e-01, -1.9150e-01, ..., -3.4639e-04, + -6.2902e-02, -1.2278e-01], + ..., + [ 1.1565e-01, -1.0811e-01, 1.4787e-01, ..., 7.6968e-02, + 2.2605e-01, -2.4544e-02], + [ 1.4393e-01, -2.6503e-01, -1.8211e-01, ..., -3.7280e-01, + -1.2238e-01, 2.3655e-01], + [-2.1727e-01, 9.4138e-02, 7.5413e-02, ..., -3.6051e-01, + -1.6931e-01, -1.0887e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3097e-07, -3.7253e-09, 1.6019e-07, ..., 1.6391e-07, + 1.2293e-07, -1.8626e-08], + [ 2.6077e-08, 0.0000e+00, 1.8626e-08, ..., 1.8626e-08, + 1.4901e-08, 0.0000e+00], + ..., + [-2.7195e-07, 7.4506e-09, -1.6019e-07, ..., -1.7509e-07, + -1.4156e-07, 3.7253e-09], + [-3.7253e-09, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 3.7253e-09, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0110], device='cuda:0'), grad: tensor([ 0.0000e+00, 4.7684e-07, 6.3330e-08, -2.6077e-08, 7.4506e-09, + 3.7253e-09, 3.7253e-09, -5.3272e-07, -1.1176e-08, 1.1176e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 214.69, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4021 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.13 lr 0.00001000 +Epoch 414, weight, value: tensor([[-3.2073e-01, 8.6743e-02, -2.1251e-01, ..., -1.1362e-01, + -2.9130e-01, -1.7624e-01], + [-8.3455e-03, 1.0561e-01, -1.2921e-01, ..., -1.4199e-01, + -8.4398e-02, 1.3173e-01], + [ 5.0349e-02, -1.5719e-01, -1.9153e-01, ..., -3.4137e-04, + -6.2902e-02, -1.2279e-01], + ..., + [ 1.1566e-01, -1.0815e-01, 1.4789e-01, ..., 7.6978e-02, + 2.2607e-01, -2.4550e-02], + [ 1.4393e-01, -2.6509e-01, -1.8214e-01, ..., -3.7284e-01, + -1.2242e-01, 2.3655e-01], + [-2.1729e-01, 9.4130e-02, 7.5421e-02, ..., -3.6057e-01, + -1.6932e-01, -1.0890e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 3.7253e-09, -0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 3.7253e-09, -7.4506e-09], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-2.6077e-08, 7.4506e-09, -1.8626e-08, ..., -0.0000e+00, + -1.4901e-08, 7.4506e-09], + [ 0.0000e+00, 2.9802e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 1.4901e-08, 3.7253e-09, 1.1176e-08, ..., 0.0000e+00, + 7.4506e-09, 3.7253e-09]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0230, + 0.0132, -0.0110], device='cuda:0'), grad: tensor([ 1.6019e-07, 3.7253e-09, 3.7253e-09, 1.2293e-07, -2.9802e-08, + -1.0356e-06, 5.2154e-07, -1.4901e-08, 2.0862e-07, 5.2154e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 214.86, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4399 re_mapping 0.0031 re_causal 0.0105 /// teacc 99.12 lr 0.00001000 +Epoch 415, weight, value: tensor([[-3.2074e-01, 8.6748e-02, -2.1252e-01, ..., -1.1363e-01, + -2.9131e-01, -1.7625e-01], + [-8.3957e-03, 1.0568e-01, -1.2928e-01, ..., -1.4206e-01, + -8.4466e-02, 1.3174e-01], + [ 5.0347e-02, -1.5720e-01, -1.9156e-01, ..., -3.3974e-04, + -6.2907e-02, -1.2279e-01], + ..., + [ 1.1571e-01, -1.0830e-01, 1.4797e-01, ..., 7.6988e-02, + 2.2614e-01, -2.4552e-02], + [ 1.4393e-01, -2.6514e-01, -1.8220e-01, ..., -3.7290e-01, + -1.2250e-01, 2.3656e-01], + [-2.1735e-01, 9.4113e-02, 7.5407e-02, ..., -3.6063e-01, + -1.6934e-01, -1.0898e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, -0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 7.4506e-09, -0.0000e+00], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + -7.4506e-09, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 3.7253e-09, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0231, + 0.0131, -0.0110], device='cuda:0'), grad: tensor([ 3.7253e-09, 7.4506e-09, -1.4901e-08, 1.4901e-08, 0.0000e+00, + -1.8626e-08, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 214.95, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4755 re_mapping 0.0032 re_causal 0.0109 /// teacc 99.11 lr 0.00001000 +Epoch 416, weight, value: tensor([[-3.2076e-01, 8.6729e-02, -2.1252e-01, ..., -1.1364e-01, + -2.9132e-01, -1.7625e-01], + [-8.4033e-03, 1.0569e-01, -1.2929e-01, ..., -1.4215e-01, + -8.4476e-02, 1.3175e-01], + [ 5.0345e-02, -1.5724e-01, -1.9161e-01, ..., -3.3854e-04, + -6.2914e-02, -1.2281e-01], + ..., + [ 1.1572e-01, -1.0833e-01, 1.4799e-01, ..., 7.7000e-02, + 2.2616e-01, -2.4555e-02], + [ 1.4396e-01, -2.6520e-01, -1.8226e-01, ..., -3.7293e-01, + -1.2253e-01, 2.3659e-01], + [-2.1739e-01, 9.4172e-02, 7.5405e-02, ..., -3.6069e-01, + -1.6936e-01, -1.0899e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 7.4506e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.9605e-08]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0231, + 0.0131, -0.0110], device='cuda:0'), grad: tensor([-8.9407e-08, 0.0000e+00, 3.7253e-09, -4.4703e-08, -2.5705e-07, + -4.4703e-08, 7.0781e-08, 3.7253e-09, -3.7253e-09, 3.5763e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 215.00, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4405 re_mapping 0.0032 re_causal 0.0105 /// teacc 99.12 lr 0.00001000 +Epoch 417, weight, value: tensor([[-3.2078e-01, 8.6707e-02, -2.1255e-01, ..., -1.1364e-01, + -2.9132e-01, -1.7626e-01], + [-8.4017e-03, 1.0569e-01, -1.2928e-01, ..., -1.4220e-01, + -8.4477e-02, 1.3177e-01], + [ 5.0339e-02, -1.5737e-01, -1.9167e-01, ..., -3.4518e-04, + -6.2923e-02, -1.2282e-01], + ..., + [ 1.1572e-01, -1.0845e-01, 1.4799e-01, ..., 7.7016e-02, + 2.2617e-01, -2.4567e-02], + [ 1.4400e-01, -2.6525e-01, -1.8230e-01, ..., -3.7297e-01, + -1.2257e-01, 2.3664e-01], + [-2.1740e-01, 9.4253e-02, 7.5414e-02, ..., -3.6076e-01, + -1.6937e-01, -1.0902e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 3.5018e-07, -5.9605e-08, 1.8254e-07, ..., 7.4506e-09, + 2.9430e-07, 4.4703e-08], + [ 1.8626e-08, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 1.4901e-08, 7.4506e-09], + ..., + [-4.1723e-07, 4.8429e-08, -2.1607e-07, ..., -7.4506e-09, + -3.4645e-07, -9.6858e-08], + [ 2.6077e-08, 3.7253e-09, 1.1176e-08, ..., 0.0000e+00, + 1.8626e-08, 1.4901e-08], + [ 1.8626e-08, 1.8626e-08, 7.4506e-09, ..., 0.0000e+00, + 1.1176e-08, 2.2352e-08]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0195, 0.0040, -0.0005, 0.0085, 0.0231, + 0.0131, -0.0109], device='cuda:0'), grad: tensor([ 3.7253e-09, 3.7253e-07, 3.7253e-08, 0.0000e+00, 0.0000e+00, + -5.9605e-08, 5.5879e-08, -5.5879e-07, 5.5879e-08, 9.3132e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 214.81, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4385 re_mapping 0.0031 re_causal 0.0106 /// teacc 99.11 lr 0.00001000 +Epoch 418, weight, value: tensor([[-3.2078e-01, 8.6695e-02, -2.1264e-01, ..., -1.1365e-01, + -2.9133e-01, -1.7627e-01], + [-8.3893e-03, 1.0577e-01, -1.2927e-01, ..., -1.4224e-01, + -8.4467e-02, 1.3180e-01], + [ 5.0345e-02, -1.5741e-01, -1.9169e-01, ..., -3.3714e-04, + -6.2917e-02, -1.2284e-01], + ..., + [ 1.1572e-01, -1.0875e-01, 1.4799e-01, ..., 7.7015e-02, + 2.2616e-01, -2.4597e-02], + [ 1.4403e-01, -2.6530e-01, -1.8235e-01, ..., -3.7302e-01, + -1.2260e-01, 2.3669e-01], + [-2.1744e-01, 9.4288e-02, 7.5427e-02, ..., -3.6081e-01, + -1.6938e-01, -1.0905e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [-3.7253e-09, 1.1176e-08, 7.4506e-09, ..., -0.0000e+00, + -3.7253e-09, 0.0000e+00], + [-1.1176e-08, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, -7.0781e-08, -7.0781e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0086, 0.0231, + 0.0131, -0.0109], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.8626e-08, 3.3528e-08, 5.2154e-08, 1.0803e-07, + 2.6077e-08, 7.4506e-09, 4.4703e-08, -2.6077e-08, -2.7567e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 214.85, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4392 re_mapping 0.0031 re_causal 0.0106 /// teacc 99.12 lr 0.00001000 +Epoch 419, weight, value: tensor([[-3.2080e-01, 8.6699e-02, -2.1265e-01, ..., -1.1366e-01, + -2.9134e-01, -1.7627e-01], + [-8.4144e-03, 1.0578e-01, -1.2931e-01, ..., -1.4227e-01, + -8.4508e-02, 1.3181e-01], + [ 5.0342e-02, -1.5747e-01, -1.9172e-01, ..., -3.3734e-04, + -6.2921e-02, -1.2285e-01], + ..., + [ 1.1574e-01, -1.0889e-01, 1.4804e-01, ..., 7.7026e-02, + 2.2621e-01, -2.4595e-02], + [ 1.4406e-01, -2.6534e-01, -1.8239e-01, ..., -3.7306e-01, + -1.2263e-01, 2.3671e-01], + [-2.1746e-01, 9.4303e-02, 7.5425e-02, ..., -3.6087e-01, + -1.6940e-01, -1.0908e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 3.3155e-07, 1.0803e-07, 2.9802e-07, ..., 1.5646e-07, + 2.5705e-07, 5.5879e-08], + [ 2.2352e-08, 0.0000e+00, 3.7253e-08, ..., -1.1176e-08, + 2.2352e-08, 3.7253e-09], + ..., + [-4.0606e-07, 1.4901e-08, -3.7253e-07, ..., -1.7509e-07, + -3.2410e-07, 1.1176e-08], + [ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, -2.0117e-07, -2.4214e-07, ..., 3.7253e-09, + 7.4506e-09, 1.1176e-08]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0086, 0.0231, + 0.0131, -0.0109], device='cuda:0'), grad: tensor([ 7.0781e-08, 1.0543e-06, 3.7253e-09, 6.3330e-08, 3.4273e-07, + 6.3330e-08, -4.4703e-07, -7.2643e-07, 3.3528e-07, -7.7859e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 214.62, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4388 re_mapping 0.0031 re_causal 0.0104 /// teacc 99.12 lr 0.00001000 +Epoch 420, weight, value: tensor([[-3.2081e-01, 8.6719e-02, -2.1268e-01, ..., -1.1366e-01, + -2.9135e-01, -1.7627e-01], + [-8.4120e-03, 1.0579e-01, -1.2931e-01, ..., -1.4230e-01, + -8.4502e-02, 1.3183e-01], + [ 5.0348e-02, -1.5752e-01, -1.9173e-01, ..., -3.2662e-04, + -6.2908e-02, -1.2280e-01], + ..., + [ 1.1575e-01, -1.0903e-01, 1.4804e-01, ..., 7.7027e-02, + 2.2621e-01, -2.4614e-02], + [ 1.4405e-01, -2.6540e-01, -1.8243e-01, ..., -3.7313e-01, + -1.2272e-01, 2.3672e-01], + [-2.1751e-01, 9.4296e-02, 7.5434e-02, ..., -3.6098e-01, + -1.6941e-01, -1.0915e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.4901e-08, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, -3.7253e-09], + [-1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0086, 0.0231, + 0.0131, -0.0109], device='cuda:0'), grad: tensor([ 7.4506e-09, 7.8231e-08, -7.4506e-08, 1.8626e-08, 1.8626e-08, + 4.8429e-08, -1.0803e-07, 3.7253e-09, 1.1176e-08, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 214.63, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4420 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.13 lr 0.00001000 +Epoch 421, weight, value: tensor([[-3.2082e-01, 8.6735e-02, -2.1268e-01, ..., -1.1366e-01, + -2.9136e-01, -1.7627e-01], + [-8.4147e-03, 1.0578e-01, -1.2931e-01, ..., -1.4234e-01, + -8.4505e-02, 1.3184e-01], + [ 5.0346e-02, -1.5757e-01, -1.9177e-01, ..., -3.2358e-04, + -6.2912e-02, -1.2281e-01], + ..., + [ 1.1575e-01, -1.0911e-01, 1.4805e-01, ..., 7.7031e-02, + 2.2621e-01, -2.4626e-02], + [ 1.4407e-01, -2.6547e-01, -1.8246e-01, ..., -3.7319e-01, + -1.2276e-01, 2.3675e-01], + [-2.1754e-01, 9.4306e-02, 7.5447e-02, ..., -3.6101e-01, + -1.6942e-01, -1.0919e-01]], device='cuda:0'), grad: tensor([[ 5.5879e-09, -1.8626e-09, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.0489e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 4.0978e-08], + [ 5.0291e-08, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -1.8626e-09, 5.7742e-08], + ..., + [ 3.7253e-09, 9.3132e-09, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 2.0489e-08], + [-6.7055e-08, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, + 0.0000e+00, -6.7055e-08], + [-0.0000e+00, 1.6764e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-08]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0086, 0.0231, + 0.0130, -0.0109], device='cuda:0'), grad: tensor([ 3.7253e-08, 7.8231e-08, 7.6368e-08, 1.6764e-08, -1.7509e-07, + 1.8626e-09, 5.5879e-09, 4.4703e-08, -1.5460e-07, 6.3330e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 214.44, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4524 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.12 lr 0.00001000 +Epoch 422, weight, value: tensor([[-3.2083e-01, 8.6741e-02, -2.1268e-01, ..., -1.1365e-01, + -2.9136e-01, -1.7628e-01], + [-8.4174e-03, 1.0579e-01, -1.2932e-01, ..., -1.4241e-01, + -8.4512e-02, 1.3186e-01], + [ 5.0342e-02, -1.5760e-01, -1.9181e-01, ..., -3.2418e-04, + -6.2921e-02, -1.2283e-01], + ..., + [ 1.1576e-01, -1.0917e-01, 1.4807e-01, ..., 7.7040e-02, + 2.2623e-01, -2.4634e-02], + [ 1.4403e-01, -2.6553e-01, -1.8251e-01, ..., -3.7322e-01, + -1.2282e-01, 2.3673e-01], + [-2.1758e-01, 9.4310e-02, 7.5461e-02, ..., -3.6107e-01, + -1.6944e-01, -1.0925e-01]], device='cuda:0'), grad: tensor([[ 5.5879e-09, -3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 2.6077e-08, 7.4506e-09, 3.5390e-08, ..., 1.8626e-08, + 2.4214e-08, -1.8626e-09], + [ 9.3132e-09, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 5.5879e-09, 1.8626e-09], + ..., + [-8.9407e-08, -1.8626e-08, -1.0245e-07, ..., -4.8429e-08, + -8.1956e-08, 3.7253e-09], + [-3.7253e-09, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 2.2352e-08, -3.5390e-08, -3.1665e-08, ..., 9.3132e-09, + 1.6764e-08, 5.5879e-09]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0086, 0.0231, + 0.0130, -0.0109], device='cuda:0'), grad: tensor([ 3.7253e-09, 6.7055e-08, 2.2352e-08, 6.8918e-08, 9.1270e-08, + 1.8626e-08, 1.8626e-09, -2.2165e-07, -7.4506e-09, -5.7742e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 214.76, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4378 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.12 lr 0.00001000 +Epoch 423, weight, value: tensor([[-3.2085e-01, 8.6736e-02, -2.1270e-01, ..., -1.1366e-01, + -2.9137e-01, -1.7628e-01], + [-8.4228e-03, 1.0579e-01, -1.2934e-01, ..., -1.4247e-01, + -8.4525e-02, 1.3187e-01], + [ 5.0342e-02, -1.5765e-01, -1.9183e-01, ..., -3.1882e-04, + -6.2918e-02, -1.2286e-01], + ..., + [ 1.1577e-01, -1.0917e-01, 1.4809e-01, ..., 7.7043e-02, + 2.2624e-01, -2.4641e-02], + [ 1.4403e-01, -2.6558e-01, -1.8258e-01, ..., -3.7329e-01, + -1.2287e-01, 2.3675e-01], + [-2.1760e-01, 9.4353e-02, 7.5482e-02, ..., -3.6111e-01, + -1.6944e-01, -1.0928e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.9802e-07, 7.4506e-09, 1.2480e-07, ..., 5.5879e-09, + 1.7136e-07, 1.1362e-07], + [-2.6822e-07, 1.8626e-09, 1.1176e-08, ..., -4.0978e-08, + -2.1793e-07, 1.8626e-09], + ..., + [-1.2480e-07, 2.4214e-08, -1.2666e-07, ..., 3.1665e-08, + -1.3039e-08, -1.4342e-07], + [ 3.3528e-08, 1.8626e-09, 1.6764e-08, ..., 0.0000e+00, + 2.2352e-08, 9.3132e-09], + [ 4.0978e-08, -1.8626e-07, -2.7195e-07, ..., 9.3132e-09, + 2.4214e-08, 1.4901e-08]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0040, -0.0005, 0.0086, 0.0231, + 0.0129, -0.0109], device='cuda:0'), grad: tensor([ 1.3039e-08, 5.7556e-07, -3.1665e-07, -2.2352e-08, 5.8673e-07, + 2.0489e-08, -3.9116e-08, -2.9244e-07, 7.4506e-08, -6.1654e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 214.75, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4124 re_mapping 0.0030 re_causal 0.0099 /// teacc 99.13 lr 0.00001000 +Epoch 424, weight, value: tensor([[-3.2086e-01, 8.6750e-02, -2.1271e-01, ..., -1.1367e-01, + -2.9138e-01, -1.7628e-01], + [-8.4125e-03, 1.0585e-01, -1.2933e-01, ..., -1.4250e-01, + -8.4512e-02, 1.3194e-01], + [ 5.0329e-02, -1.5769e-01, -1.9193e-01, ..., -3.2987e-04, + -6.2943e-02, -1.2287e-01], + ..., + [ 1.1578e-01, -1.0944e-01, 1.4811e-01, ..., 7.7074e-02, + 2.2626e-01, -2.4696e-02], + [ 1.4403e-01, -2.6565e-01, -1.8265e-01, ..., -3.7332e-01, + -1.2292e-01, 2.3676e-01], + [-2.1764e-01, 9.4354e-02, 7.5492e-02, ..., -3.6119e-01, + -1.6947e-01, -1.0934e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 4.0978e-08, 7.4506e-09, 4.0978e-08, ..., 1.3039e-08, + 5.0291e-08, 2.9802e-08], + [ 3.7253e-08, 1.8626e-09, 3.9116e-08, ..., 1.8626e-08, + 5.0291e-08, 7.4506e-09], + ..., + [-2.1607e-07, 3.7253e-09, -2.3097e-07, ..., -1.1176e-07, + -2.9616e-07, -3.1665e-08], + [ 9.3132e-09, 3.7253e-09, 9.3132e-09, ..., 1.8626e-09, + 1.1176e-08, 5.5879e-09], + [ 9.3132e-09, 1.8626e-08, 1.1176e-08, ..., 3.7253e-09, + 1.1176e-08, 4.2841e-08]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0196, 0.0039, -0.0005, 0.0086, 0.0231, + 0.0129, -0.0109], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.3039e-07, 9.1270e-08, 3.9861e-07, -1.7136e-07, + -2.7753e-07, 1.3784e-07, -4.8243e-07, 3.1665e-08, 1.3784e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 214.86, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4554 re_mapping 0.0029 re_causal 0.0106 /// teacc 99.12 lr 0.00001000 +Epoch 425, weight, value: tensor([[-3.2087e-01, 8.6757e-02, -2.1272e-01, ..., -1.1368e-01, + -2.9139e-01, -1.7628e-01], + [-8.4127e-03, 1.0587e-01, -1.2933e-01, ..., -1.4256e-01, + -8.4511e-02, 1.3197e-01], + [ 5.0326e-02, -1.5772e-01, -1.9197e-01, ..., -3.3272e-04, + -6.2950e-02, -1.2287e-01], + ..., + [ 1.1579e-01, -1.0953e-01, 1.4812e-01, ..., 7.7099e-02, + 2.2627e-01, -2.4719e-02], + [ 1.4404e-01, -2.6575e-01, -1.8273e-01, ..., -3.7338e-01, + -1.2297e-01, 2.3678e-01], + [-2.1766e-01, 9.4357e-02, 7.5512e-02, ..., -3.6126e-01, + -1.6947e-01, -1.0938e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -5.5879e-09, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 1.8626e-09, ..., 3.7253e-09, + 3.7253e-09, 1.8626e-09], + [ 1.4901e-08, 1.8626e-09, 2.0489e-08, ..., -5.5879e-09, + 6.3330e-08, 0.0000e+00], + ..., + [-4.6566e-08, 1.8626e-09, -2.0489e-08, ..., -1.1176e-08, + -8.3819e-08, 0.0000e+00], + [ 1.1176e-08, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 1.8626e-09, 1.8626e-09], + [ 1.8626e-09, 5.5879e-09, -3.1665e-08, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-08]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0197, 0.0039, -0.0005, 0.0086, 0.0231, + 0.0128, -0.0109], device='cuda:0'), grad: tensor([-9.3132e-09, 1.8626e-08, -2.4214e-08, 2.4214e-08, -1.4529e-07, + -3.7253e-09, 1.1176e-08, -4.0978e-08, 3.5390e-08, 1.2852e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 214.87, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.3997 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.12 lr 0.00001000 +Epoch 426, weight, value: tensor([[-3.2089e-01, 8.6754e-02, -2.1279e-01, ..., -1.1369e-01, + -2.9140e-01, -1.7630e-01], + [-8.4204e-03, 1.0584e-01, -1.2934e-01, ..., -1.4266e-01, + -8.4519e-02, 1.3198e-01], + [ 5.0313e-02, -1.5785e-01, -1.9208e-01, ..., -3.4245e-04, + -6.2970e-02, -1.2291e-01], + ..., + [ 1.1581e-01, -1.0957e-01, 1.4815e-01, ..., 7.7127e-02, + 2.2630e-01, -2.4730e-02], + [ 1.4406e-01, -2.6583e-01, -1.8277e-01, ..., -3.7344e-01, + -1.2301e-01, 2.3683e-01], + [-2.1774e-01, 9.4384e-02, 7.5529e-02, ..., -3.6132e-01, + -1.6948e-01, -1.0952e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-08, 1.8626e-09, 1.8626e-09, ..., 3.7253e-09, + 1.8626e-09, 2.4214e-08], + [ 1.8626e-08, 0.0000e+00, 7.4506e-09, ..., 7.4506e-09, + 1.3039e-08, -3.7253e-09], + [ 5.9605e-08, 1.8626e-09, 5.5879e-09, ..., -1.4901e-08, + -1.1176e-08, 1.1362e-07], + ..., + [-5.5879e-09, 3.7253e-09, -9.3132e-09, ..., 1.3039e-08, + -7.4506e-09, 3.7253e-09], + [-1.1921e-07, 2.2352e-08, -1.8626e-09, ..., -1.8626e-09, + 0.0000e+00, -1.7695e-07], + [ 1.8626e-09, 2.7940e-08, 5.5879e-09, ..., 1.3039e-08, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0197, 0.0040, -0.0004, 0.0086, 0.0231, + 0.0128, -0.0109], device='cuda:0'), grad: tensor([ 7.2643e-08, 4.8429e-08, 2.1048e-07, 4.2394e-06, 1.8626e-09, + -4.6566e-06, 1.5460e-07, 2.9802e-08, -2.8312e-07, 1.8254e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 214.51, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4471 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.14 lr 0.00001000 +Epoch 427, weight, value: tensor([[-3.2092e-01, 8.6757e-02, -2.1297e-01, ..., -1.1370e-01, + -2.9141e-01, -1.7631e-01], + [-8.4149e-03, 1.0589e-01, -1.2934e-01, ..., -1.4270e-01, + -8.4519e-02, 1.3202e-01], + [ 5.0312e-02, -1.5791e-01, -1.9211e-01, ..., -3.3915e-04, + -6.2971e-02, -1.2293e-01], + ..., + [ 1.1581e-01, -1.0969e-01, 1.4816e-01, ..., 7.7133e-02, + 2.2630e-01, -2.4753e-02], + [ 1.4402e-01, -2.6593e-01, -1.8283e-01, ..., -3.7348e-01, + -1.2309e-01, 2.3680e-01], + [-2.1778e-01, 9.4414e-02, 7.5567e-02, ..., -3.6141e-01, + -1.6950e-01, -1.0956e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.5390e-08, 9.3132e-09, ..., 0.0000e+00, + 1.8626e-09, -1.1176e-08], + [ 2.4214e-08, 1.8626e-08, 2.2352e-08, ..., 7.4506e-09, + 1.4901e-08, -3.7253e-09], + [-0.0000e+00, 1.8626e-09, 1.8626e-09, ..., -1.8626e-09, + -1.8626e-09, 0.0000e+00], + ..., + [-3.1665e-08, 9.3132e-09, -2.7940e-08, ..., -7.4506e-09, + -1.8626e-08, 9.3132e-09], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.6764e-08, -1.3039e-08, ..., 0.0000e+00, + -0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0198, 0.0040, -0.0004, 0.0086, 0.0231, + 0.0127, -0.0109], device='cuda:0'), grad: tensor([ 4.1723e-07, 9.1270e-08, -1.8626e-09, 5.5879e-09, 3.5949e-07, + 1.5087e-07, -1.2573e-06, -1.8626e-08, 5.4017e-08, 1.8626e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 214.30, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4207 re_mapping 0.0030 re_causal 0.0101 /// teacc 99.15 lr 0.00001000 +Epoch 428, weight, value: tensor([[-3.2095e-01, 8.6758e-02, -2.1298e-01, ..., -1.1372e-01, + -2.9142e-01, -1.7631e-01], + [-8.4134e-03, 1.0588e-01, -1.2935e-01, ..., -1.4274e-01, + -8.4519e-02, 1.3204e-01], + [ 5.0317e-02, -1.5798e-01, -1.9214e-01, ..., -3.3103e-04, + -6.2964e-02, -1.2293e-01], + ..., + [ 1.1581e-01, -1.0978e-01, 1.4817e-01, ..., 7.7135e-02, + 2.2630e-01, -2.4768e-02], + [ 1.4400e-01, -2.6600e-01, -1.8288e-01, ..., -3.7354e-01, + -1.2316e-01, 2.3680e-01], + [-2.1780e-01, 9.4457e-02, 7.5581e-02, ..., -3.6148e-01, + -1.6951e-01, -1.0960e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, -3.7253e-09], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-5.0291e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, -9.3132e-09, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0198, 0.0040, -0.0004, 0.0086, 0.0231, + 0.0127, -0.0109], device='cuda:0'), grad: tensor([ 5.5879e-09, -5.5879e-09, 1.6764e-08, 7.4506e-09, 1.7136e-07, + 1.1176e-08, 1.8626e-09, -1.7695e-07, -7.4506e-09, -2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 214.58, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4539 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.15 lr 0.00001000 +Epoch 429, weight, value: tensor([[-3.2096e-01, 8.6777e-02, -2.1298e-01, ..., -1.1373e-01, + -2.9143e-01, -1.7632e-01], + [-8.4292e-03, 1.0591e-01, -1.2938e-01, ..., -1.4277e-01, + -8.4549e-02, 1.3207e-01], + [ 5.0317e-02, -1.5802e-01, -1.9218e-01, ..., -3.2489e-04, + -6.2966e-02, -1.2295e-01], + ..., + [ 1.1583e-01, -1.0990e-01, 1.4820e-01, ..., 7.7134e-02, + 2.2634e-01, -2.4780e-02], + [ 1.4393e-01, -2.6611e-01, -1.8295e-01, ..., -3.7361e-01, + -1.2322e-01, 2.3674e-01], + [-2.1785e-01, 9.4441e-02, 7.5592e-02, ..., -3.6155e-01, + -1.6952e-01, -1.0968e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 5.5879e-09, ..., -1.8626e-09, + 3.7253e-09, 0.0000e+00], + ..., + [-5.5879e-09, 1.8626e-09, -9.3132e-09, ..., -1.8626e-09, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 7.4506e-09, -9.3132e-09, ..., 7.4506e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0153, -0.0222, -0.0108, -0.0198, 0.0041, -0.0004, 0.0086, 0.0231, + 0.0125, -0.0109], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.8626e-08, -1.8626e-09, 1.9018e-06, 3.3528e-08, + -2.2631e-06, 1.2107e-07, -3.7253e-09, 8.3819e-08, 1.0990e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 214.80, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4492 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.11 lr 0.00001000 +Epoch 430, weight, value: tensor([[-3.2098e-01, 8.6787e-02, -2.1299e-01, ..., -1.1374e-01, + -2.9144e-01, -1.7632e-01], + [-8.4141e-03, 1.0597e-01, -1.2937e-01, ..., -1.4278e-01, + -8.4529e-02, 1.3213e-01], + [ 5.0311e-02, -1.5804e-01, -1.9224e-01, ..., -3.2547e-04, + -6.2976e-02, -1.2296e-01], + ..., + [ 1.1583e-01, -1.1010e-01, 1.4822e-01, ..., 7.7158e-02, + 2.2634e-01, -2.4825e-02], + [ 1.4396e-01, -2.6618e-01, -1.8299e-01, ..., -3.7369e-01, + -1.2326e-01, 2.3678e-01], + [-2.1787e-01, 9.4426e-02, 7.5609e-02, ..., -3.6162e-01, + -1.6953e-01, -1.0974e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 8.5682e-08, -5.5879e-09, 5.0291e-08, ..., 2.4214e-08, + 6.5193e-08, 1.8626e-09], + [ 5.5879e-09, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 3.7253e-09, 1.8626e-09], + ..., + [-1.0245e-07, 1.8626e-09, -6.5193e-08, ..., -2.7940e-08, + -8.3819e-08, -9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 7.4506e-09, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 1.1176e-08, 1.8626e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0153, -0.0221, -0.0108, -0.0198, 0.0041, -0.0005, 0.0086, 0.0231, + 0.0125, -0.0110], device='cuda:0'), grad: tensor([ 5.0291e-08, 1.6391e-07, 1.4901e-08, 0.0000e+00, 3.7253e-09, + 1.1176e-08, -9.1270e-08, -2.2165e-07, 2.9802e-08, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 214.88, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4358 re_mapping 0.0029 re_causal 0.0102 /// teacc 99.11 lr 0.00001000 +Epoch 431, weight, value: tensor([[-3.2102e-01, 8.6801e-02, -2.1305e-01, ..., -1.1376e-01, + -2.9146e-01, -1.7633e-01], + [-8.4268e-03, 1.0599e-01, -1.2938e-01, ..., -1.4287e-01, + -8.4548e-02, 1.3215e-01], + [ 5.0307e-02, -1.5812e-01, -1.9231e-01, ..., -3.2491e-04, + -6.2986e-02, -1.2298e-01], + ..., + [ 1.1585e-01, -1.1020e-01, 1.4825e-01, ..., 7.7182e-02, + 2.2637e-01, -2.4840e-02], + [ 1.4399e-01, -2.6625e-01, -1.8305e-01, ..., -3.7375e-01, + -1.2330e-01, 2.3681e-01], + [-2.1793e-01, 9.4417e-02, 7.5613e-02, ..., -3.6177e-01, + -1.6955e-01, -1.0983e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0198, 0.0041, -0.0005, 0.0086, 0.0231, + 0.0125, -0.0110], device='cuda:0'), grad: tensor([-3.7253e-09, 0.0000e+00, 0.0000e+00, 1.8626e-09, -3.7253e-09, + -3.1665e-08, 1.4901e-08, 1.8626e-09, 1.8626e-09, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 214.80, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4436 re_mapping 0.0029 re_causal 0.0104 /// teacc 99.11 lr 0.00001000 +Epoch 432, weight, value: tensor([[-3.2105e-01, 8.6822e-02, -2.1306e-01, ..., -1.1379e-01, + -2.9147e-01, -1.7633e-01], + [-8.4455e-03, 1.0598e-01, -1.2941e-01, ..., -1.4278e-01, + -8.4576e-02, 1.3216e-01], + [ 5.0307e-02, -1.5822e-01, -1.9236e-01, ..., -3.1784e-04, + -6.2990e-02, -1.2300e-01], + ..., + [ 1.1588e-01, -1.1026e-01, 1.4829e-01, ..., 7.7173e-02, + 2.2641e-01, -2.4840e-02], + [ 1.4401e-01, -2.6632e-01, -1.8314e-01, ..., -3.7384e-01, + -1.2338e-01, 2.3689e-01], + [-2.1799e-01, 9.4395e-02, 7.5620e-02, ..., -3.6187e-01, + -1.6957e-01, -1.0995e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.9116e-08], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [-1.8626e-09, 2.2352e-08, -1.8626e-09, ..., -0.0000e+00, + -1.8626e-09, 1.4901e-08], + [ 0.0000e+00, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 2.2352e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0199, 0.0042, -0.0005, 0.0086, 0.0231, + 0.0125, -0.0111], device='cuda:0'), grad: tensor([-9.3132e-09, -1.6764e-07, 9.3132e-09, 6.1467e-08, -2.6077e-08, + -9.4995e-08, 2.6077e-08, 8.3819e-08, 1.3039e-08, 8.7544e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 214.49, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4271 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.13 lr 0.00001000 +Epoch 433, weight, value: tensor([[-3.2105e-01, 8.6849e-02, -2.1308e-01, ..., -1.1381e-01, + -2.9148e-01, -1.7633e-01], + [-8.4702e-03, 1.0600e-01, -1.2946e-01, ..., -1.4287e-01, + -8.4614e-02, 1.3216e-01], + [ 5.0311e-02, -1.5827e-01, -1.9249e-01, ..., -3.0153e-04, + -6.2998e-02, -1.2301e-01], + ..., + [ 1.1590e-01, -1.1034e-01, 1.4836e-01, ..., 7.7177e-02, + 2.2646e-01, -2.4838e-02], + [ 1.4402e-01, -2.6639e-01, -1.8321e-01, ..., -3.7394e-01, + -1.2344e-01, 2.3694e-01], + [-2.1804e-01, 9.4350e-02, 7.5619e-02, ..., -3.6197e-01, + -1.6960e-01, -1.1002e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 3.9116e-08, -5.5879e-09, 3.3528e-08, ..., 3.7253e-09, + 4.0978e-08, 7.4506e-09], + [ 2.0489e-08, 1.8626e-09, 1.4901e-08, ..., 1.8626e-09, + 1.8626e-08, 9.3132e-09], + ..., + [-1.1548e-07, -1.8626e-09, -8.7544e-08, ..., -1.3039e-08, + -1.0617e-07, -3.9116e-08], + [ 1.3039e-08, 7.4506e-09, 1.1176e-08, ..., 0.0000e+00, + 1.4901e-08, 9.3132e-09], + [ 2.4214e-08, 1.8626e-09, 1.4901e-08, ..., 3.7253e-09, + 1.8626e-08, 7.4506e-09]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0199, 0.0042, -0.0005, 0.0087, 0.0231, + 0.0124, -0.0111], device='cuda:0'), grad: tensor([ 2.9802e-08, 8.1956e-08, 5.2154e-08, 9.8720e-08, 1.6764e-08, + -8.1956e-07, 4.8615e-07, -1.8254e-07, 1.7695e-07, 5.2154e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 214.70, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4637 re_mapping 0.0029 re_causal 0.0100 /// teacc 99.11 lr 0.00001000 +Epoch 434, weight, value: tensor([[-3.2109e-01, 8.6872e-02, -2.1308e-01, ..., -1.1383e-01, + -2.9149e-01, -1.7633e-01], + [-8.5038e-03, 1.0603e-01, -1.2951e-01, ..., -1.4300e-01, + -8.4663e-02, 1.3217e-01], + [ 5.0309e-02, -1.5834e-01, -1.9262e-01, ..., -2.9112e-04, + -6.3007e-02, -1.2303e-01], + ..., + [ 1.1594e-01, -1.1041e-01, 1.4843e-01, ..., 7.7194e-02, + 2.2652e-01, -2.4838e-02], + [ 1.4404e-01, -2.6652e-01, -1.8329e-01, ..., -3.7403e-01, + -1.2354e-01, 2.3701e-01], + [-2.1813e-01, 9.4302e-02, 7.5609e-02, ..., -3.6215e-01, + -1.6963e-01, -1.1013e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 3.7253e-09, -1.8626e-09], + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -0.0000e+00, 0.0000e+00], + ..., + [-1.1176e-08, 1.8626e-09, -1.1176e-08, ..., -0.0000e+00, + -7.4506e-09, 1.8626e-09], + [ 0.0000e+00, 7.4506e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, -1.4901e-07, -8.3819e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0152, -0.0222, -0.0108, -0.0200, 0.0042, -0.0004, 0.0086, 0.0231, + 0.0124, -0.0112], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.1176e-08, -1.8626e-09, 2.0489e-08, 3.5763e-07, + 1.1176e-08, 0.0000e+00, -1.4901e-08, 2.4214e-08, -4.2282e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 214.44, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4153 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 435, weight, value: tensor([[-3.2112e-01, 8.6913e-02, -2.1309e-01, ..., -1.1383e-01, + -2.9150e-01, -1.7634e-01], + [-8.5018e-03, 1.0608e-01, -1.2951e-01, ..., -1.4308e-01, + -8.4675e-02, 1.3222e-01], + [ 5.0300e-02, -1.5839e-01, -1.9272e-01, ..., -2.8777e-04, + -6.3024e-02, -1.2305e-01], + ..., + [ 1.1595e-01, -1.1059e-01, 1.4845e-01, ..., 7.7212e-02, + 2.2655e-01, -2.4865e-02], + [ 1.4403e-01, -2.6666e-01, -1.8337e-01, ..., -3.7413e-01, + -1.2362e-01, 2.3702e-01], + [-2.1817e-01, 9.4274e-02, 7.5640e-02, ..., -3.6227e-01, + -1.6965e-01, -1.1019e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -3.5390e-08, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, -6.7055e-08], + [-8.3819e-08, 7.4506e-09, -9.3132e-09, ..., -2.7940e-08, + -6.1467e-08, 1.3039e-08], + ..., + [ 7.2643e-08, 2.4214e-08, 7.4506e-09, ..., 2.4214e-08, + 5.7742e-08, 4.2841e-08], + [ 7.4506e-09, 5.5879e-09, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0151, -0.0222, -0.0108, -0.0200, 0.0042, -0.0004, 0.0086, 0.0231, + 0.0123, -0.0113], device='cuda:0'), grad: tensor([-9.3132e-09, -2.7753e-07, -1.2666e-07, 1.2480e-07, 1.8626e-09, + -1.6019e-07, 2.0489e-08, 3.2410e-07, 6.1467e-08, 4.8429e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 214.85, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4355 re_mapping 0.0029 re_causal 0.0100 /// teacc 99.08 lr 0.00001000 +Epoch 436, weight, value: tensor([[-3.2114e-01, 8.6931e-02, -2.1318e-01, ..., -1.1385e-01, + -2.9151e-01, -1.7634e-01], + [-8.5252e-03, 1.0610e-01, -1.2955e-01, ..., -1.4320e-01, + -8.4710e-02, 1.3224e-01], + [ 5.0292e-02, -1.5845e-01, -1.9280e-01, ..., -2.8520e-04, + -6.3042e-02, -1.2307e-01], + ..., + [ 1.1598e-01, -1.1070e-01, 1.4850e-01, ..., 7.7231e-02, + 2.2660e-01, -2.4866e-02], + [ 1.4409e-01, -2.6676e-01, -1.8344e-01, ..., -3.7420e-01, + -1.2366e-01, 2.3707e-01], + [-2.1824e-01, 9.4278e-02, 7.5658e-02, ..., -3.6241e-01, + -1.6968e-01, -1.1031e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09], + [ 0.0000e+00, 3.9116e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-08]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0151, -0.0222, -0.0108, -0.0200, 0.0042, -0.0004, 0.0086, 0.0232, + 0.0123, -0.0113], device='cuda:0'), grad: tensor([-1.1176e-08, 7.4506e-09, 3.7253e-09, 5.4017e-08, -1.3784e-07, + -6.1467e-08, 5.5879e-09, 3.7253e-09, -9.3132e-09, 1.4901e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 215.04, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4310 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.11 lr 0.00001000 +Epoch 437, weight, value: tensor([[-3.2116e-01, 8.6951e-02, -2.1320e-01, ..., -1.1387e-01, + -2.9152e-01, -1.7635e-01], + [-8.5403e-03, 1.0609e-01, -1.2958e-01, ..., -1.4331e-01, + -8.4730e-02, 1.3225e-01], + [ 5.0301e-02, -1.5850e-01, -1.9283e-01, ..., -2.6481e-04, + -6.3025e-02, -1.2307e-01], + ..., + [ 1.1600e-01, -1.1081e-01, 1.4854e-01, ..., 7.7236e-02, + 2.2662e-01, -2.4871e-02], + [ 1.4412e-01, -2.6686e-01, -1.8352e-01, ..., -3.7433e-01, + -1.2375e-01, 2.3714e-01], + [-2.1827e-01, 9.4284e-02, 7.5693e-02, ..., -3.6249e-01, + -1.6969e-01, -1.1038e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.0489e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 5.5879e-08, -1.8626e-09, 6.3330e-08, ..., 2.4214e-08, + 5.2154e-08, -7.4506e-09], + [ 2.7753e-07, 0.0000e+00, 3.2783e-07, ..., 1.2293e-07, + 2.6450e-07, 1.1176e-08], + ..., + [-3.7067e-07, 1.8626e-09, -4.6194e-07, ..., -1.7509e-07, + -3.7439e-07, 1.3039e-08], + [-5.0291e-08, 1.8626e-09, 5.5879e-09, ..., 1.8626e-09, + 3.7253e-09, -5.9605e-08], + [ 5.4017e-08, -7.4506e-09, 3.7253e-08, ..., 2.2352e-08, + 4.8429e-08, 7.4506e-09]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0151, -0.0222, -0.0108, -0.0200, 0.0042, -0.0004, 0.0086, 0.0232, + 0.0122, -0.0113], device='cuda:0'), grad: tensor([-4.2841e-08, 1.1362e-07, 6.9290e-07, 4.6566e-08, 4.0978e-08, + 3.3528e-08, 4.2841e-08, -8.9966e-07, -1.3597e-07, 1.0803e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 215.19, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4881 re_mapping 0.0029 re_causal 0.0105 /// teacc 99.10 lr 0.00001000 +Epoch 438, weight, value: tensor([[-3.2120e-01, 8.6944e-02, -2.1322e-01, ..., -1.1390e-01, + -2.9153e-01, -1.7636e-01], + [-8.5154e-03, 1.0622e-01, -1.2956e-01, ..., -1.4337e-01, + -8.4708e-02, 1.3233e-01], + [ 5.0314e-02, -1.5860e-01, -1.9291e-01, ..., -2.2675e-04, + -6.3017e-02, -1.2310e-01], + ..., + [ 1.1598e-01, -1.1109e-01, 1.4854e-01, ..., 7.7212e-02, + 2.2661e-01, -2.4935e-02], + [ 1.4412e-01, -2.6699e-01, -1.8360e-01, ..., -3.7443e-01, + -1.2381e-01, 2.3715e-01], + [-2.1834e-01, 9.4309e-02, 7.5709e-02, ..., -3.6261e-01, + -1.6970e-01, -1.1054e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 2.4959e-07, 1.6764e-08, 3.1851e-07, ..., 3.7253e-09, + 1.8626e-07, 4.4703e-08], + [-5.5879e-09, 1.8626e-09, 1.8626e-09, ..., -9.3132e-09, + -1.8626e-09, 0.0000e+00], + ..., + [-2.5518e-07, -1.3039e-08, -3.1851e-07, ..., 9.3132e-09, + -1.8999e-07, -3.9116e-08], + [-1.4901e-08, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 1.8626e-09, -3.1665e-08], + [ 1.3039e-08, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 1.4901e-08]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0151, -0.0222, -0.0107, -0.0200, 0.0042, -0.0004, 0.0086, 0.0231, + 0.0121, -0.0113], device='cuda:0'), grad: tensor([ 1.4901e-08, 5.6624e-07, -2.7940e-08, -7.4506e-09, 3.5390e-08, + 1.0990e-07, -1.6391e-07, -5.3085e-07, -4.8429e-08, 4.6566e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 215.00, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4585 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.08 lr 0.00001000 +Epoch 439, weight, value: tensor([[-3.2123e-01, 8.6959e-02, -2.1324e-01, ..., -1.1392e-01, + -2.9155e-01, -1.7637e-01], + [-8.4822e-03, 1.0631e-01, -1.2954e-01, ..., -1.4347e-01, + -8.4682e-02, 1.3241e-01], + [ 5.0318e-02, -1.5866e-01, -1.9296e-01, ..., -2.2024e-04, + -6.3013e-02, -1.2310e-01], + ..., + [ 1.1596e-01, -1.1132e-01, 1.4854e-01, ..., 7.7233e-02, + 2.2659e-01, -2.5004e-02], + [ 1.4411e-01, -2.6708e-01, -1.8369e-01, ..., -3.7454e-01, + -1.2388e-01, 2.3715e-01], + [-2.1843e-01, 9.4271e-02, 7.5718e-02, ..., -3.6276e-01, + -1.6973e-01, -1.1073e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 5.5879e-09, 0.0000e+00], + [ 5.5879e-09, -1.8626e-09, 1.8626e-09, ..., 3.7253e-09, + 3.7253e-09, 5.5879e-09], + [-1.1176e-08, 0.0000e+00, -3.7253e-09, ..., -1.6764e-08, + -2.0489e-08, 1.8626e-09], + ..., + [ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 5.5879e-09, + 9.3132e-09, 1.8626e-09], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.5390e-08], + [ 5.5879e-09, -3.7253e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0151, -0.0221, -0.0107, -0.0200, 0.0042, -0.0004, 0.0087, 0.0231, + 0.0120, -0.0114], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.6764e-08, -6.5193e-08, 1.8626e-09, 9.3132e-09, + 1.8626e-09, 1.8626e-09, 3.1665e-08, -5.5879e-08, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 215.09, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4169 re_mapping 0.0029 re_causal 0.0097 /// teacc 99.12 lr 0.00001000 +Epoch 440, weight, value: tensor([[-3.2126e-01, 8.6995e-02, -2.1324e-01, ..., -1.1392e-01, + -2.9156e-01, -1.7637e-01], + [-8.4846e-03, 1.0633e-01, -1.2955e-01, ..., -1.4359e-01, + -8.4692e-02, 1.3246e-01], + [ 5.0342e-02, -1.5872e-01, -1.9303e-01, ..., -1.8446e-04, + -6.3002e-02, -1.2312e-01], + ..., + [ 1.1596e-01, -1.1147e-01, 1.4857e-01, ..., 7.7222e-02, + 2.2661e-01, -2.5057e-02], + [ 1.4413e-01, -2.6717e-01, -1.8377e-01, ..., -3.7465e-01, + -1.2393e-01, 2.3723e-01], + [-2.1850e-01, 9.4170e-02, 7.5685e-02, ..., -3.6282e-01, + -1.6976e-01, -1.1085e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-09, -1.6764e-08, 9.3132e-09, ..., 7.4506e-09, + 0.0000e+00, -5.9605e-08], + [-0.0000e+00, 1.8626e-09, 0.0000e+00, ..., -1.8626e-09, + -0.0000e+00, 3.7253e-09], + ..., + [-9.3132e-09, 1.8626e-08, -1.3039e-08, ..., -9.3132e-09, + -1.8626e-09, 4.4703e-08], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.5832e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0151, -0.0221, -0.0107, -0.0200, 0.0044, -0.0005, 0.0087, 0.0231, + 0.0120, -0.0116], device='cuda:0'), grad: tensor([-3.7253e-09, 7.2643e-08, 1.3039e-07, 7.4506e-09, -3.3919e-06, + 5.5879e-09, -9.5554e-07, 2.2911e-07, 6.5193e-08, 3.8520e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 215.00, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4557 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.11 lr 0.00001000 +Epoch 441, weight, value: tensor([[-3.2128e-01, 8.7023e-02, -2.1334e-01, ..., -1.1395e-01, + -2.9158e-01, -1.7637e-01], + [-8.5103e-03, 1.0631e-01, -1.2957e-01, ..., -1.4371e-01, + -8.4718e-02, 1.3248e-01], + [ 5.0348e-02, -1.5880e-01, -1.9312e-01, ..., -1.6774e-04, + -6.2995e-02, -1.2310e-01], + ..., + [ 1.1598e-01, -1.1157e-01, 1.4861e-01, ..., 7.7229e-02, + 2.2664e-01, -2.5085e-02], + [ 1.4416e-01, -2.6724e-01, -1.8384e-01, ..., -3.7477e-01, + -1.2405e-01, 2.3728e-01], + [-2.1856e-01, 9.4148e-02, 7.5694e-02, ..., -3.6289e-01, + -1.6978e-01, -1.1091e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.7940e-09, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 6.5193e-09, 9.3132e-09, ..., 7.4506e-09, + 9.3132e-10, 3.7253e-09], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 2.7940e-09, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 4.6566e-09], + [-1.5739e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.2759e-07], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0150, -0.0221, -0.0107, -0.0200, 0.0045, -0.0005, 0.0087, 0.0231, + 0.0120, -0.0116], device='cuda:0'), grad: tensor([-2.7940e-09, 2.7008e-08, -5.5879e-09, -1.3039e-08, -2.1420e-08, + 2.4401e-07, 1.2852e-07, 1.6764e-08, -3.7253e-07, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 214.97, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4085 re_mapping 0.0028 re_causal 0.0095 /// teacc 99.11 lr 0.00001000 +Epoch 442, weight, value: tensor([[-3.2132e-01, 8.7040e-02, -2.1338e-01, ..., -1.1401e-01, + -2.9160e-01, -1.7638e-01], + [-8.5171e-03, 1.0634e-01, -1.2959e-01, ..., -1.4384e-01, + -8.4727e-02, 1.3253e-01], + [ 5.0392e-02, -1.5889e-01, -1.9317e-01, ..., -9.0360e-05, + -6.2929e-02, -1.2314e-01], + ..., + [ 1.1597e-01, -1.1166e-01, 1.4864e-01, ..., 7.7187e-02, + 2.2664e-01, -2.5119e-02], + [ 1.4422e-01, -2.6740e-01, -1.8397e-01, ..., -3.7487e-01, + -1.2410e-01, 2.3732e-01], + [-2.1861e-01, 9.4118e-02, 7.5716e-02, ..., -3.6302e-01, + -1.6980e-01, -1.1103e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.5832e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 2.7940e-09, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, -3.7253e-09], + [ 9.3132e-10, 4.6566e-09, -4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0150, -0.0221, -0.0106, -0.0200, 0.0045, -0.0005, 0.0087, 0.0230, + 0.0119, -0.0117], device='cuda:0'), grad: tensor([-3.9116e-08, 9.3132e-09, 0.0000e+00, -1.4901e-08, 1.8626e-08, + 0.0000e+00, 3.7253e-09, 4.6566e-09, -2.7940e-09, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 215.02, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4399 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.12 lr 0.00001000 +Epoch 443, weight, value: tensor([[-3.2135e-01, 8.7073e-02, -2.1341e-01, ..., -1.1405e-01, + -2.9162e-01, -1.7637e-01], + [-8.5098e-03, 1.0639e-01, -1.2956e-01, ..., -1.4392e-01, + -8.4727e-02, 1.3257e-01], + [ 5.0422e-02, -1.5897e-01, -1.9324e-01, ..., -5.0995e-05, + -6.2897e-02, -1.2316e-01], + ..., + [ 1.1595e-01, -1.1178e-01, 1.4863e-01, ..., 7.7149e-02, + 2.2663e-01, -2.5162e-02], + [ 1.4425e-01, -2.6749e-01, -1.8403e-01, ..., -3.7495e-01, + -1.2415e-01, 2.3740e-01], + [-2.1866e-01, 9.4076e-02, 7.5737e-02, ..., -3.6310e-01, + -1.6980e-01, -1.1113e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.5832e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, -2.7940e-09], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [-5.5879e-09, 4.6566e-09, -6.5193e-09, ..., -1.8626e-09, + -4.6566e-09, 8.3819e-09], + [-4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 1.8626e-09, 1.0245e-08, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 8.3819e-09]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0150, -0.0221, -0.0106, -0.0200, 0.0045, -0.0005, 0.0087, 0.0230, + 0.0119, -0.0118], device='cuda:0'), grad: tensor([-4.0978e-08, 1.2107e-08, 4.6566e-09, 5.5879e-09, -4.6566e-08, + 1.2107e-08, 2.7008e-08, 8.3819e-09, -1.8626e-08, 4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 215.04, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4153 re_mapping 0.0029 re_causal 0.0098 /// teacc 99.12 lr 0.00001000 +Epoch 444, weight, value: tensor([[-3.2142e-01, 8.7069e-02, -2.1351e-01, ..., -1.1409e-01, + -2.9167e-01, -1.7639e-01], + [-8.5239e-03, 1.0641e-01, -1.2958e-01, ..., -1.4411e-01, + -8.4745e-02, 1.3260e-01], + [ 5.0447e-02, -1.5901e-01, -1.9330e-01, ..., -1.0292e-05, + -6.2870e-02, -1.2312e-01], + ..., + [ 1.1596e-01, -1.1183e-01, 1.4866e-01, ..., 7.7153e-02, + 2.2665e-01, -2.5185e-02], + [ 1.4430e-01, -2.6762e-01, -1.8410e-01, ..., -3.7512e-01, + -1.2427e-01, 2.3748e-01], + [-2.1873e-01, 9.4106e-02, 7.5769e-02, ..., -3.6324e-01, + -1.6982e-01, -1.1122e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.6613e-08, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 5.0291e-08, 1.6764e-08, 6.5193e-08, ..., 6.5193e-09, + 4.0978e-08, -2.7940e-09], + [ 5.5879e-09, 9.3132e-09, 1.4901e-08, ..., 1.1176e-08, + 1.8626e-09, 9.3132e-10], + ..., + [-1.1362e-07, -2.6077e-08, -1.3132e-07, ..., 9.3132e-10, + -9.7789e-08, 4.6566e-09], + [ 1.8626e-09, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 4.7497e-08, 5.8673e-08, 5.2154e-08, ..., 0.0000e+00, + 4.0978e-08, 1.3970e-08]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0150, -0.0221, -0.0105, -0.0200, 0.0046, -0.0005, 0.0087, 0.0230, + 0.0118, -0.0118], device='cuda:0'), grad: tensor([-1.7788e-07, 1.3504e-07, 3.6322e-08, -8.8476e-08, -9.8720e-08, + 6.2399e-08, 1.0617e-07, -2.6356e-07, 9.3132e-09, 2.9150e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 214.68, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4373 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.10 lr 0.00001000 +Epoch 445, weight, value: tensor([[-3.2143e-01, 8.7107e-02, -2.1358e-01, ..., -1.1409e-01, + -2.9170e-01, -1.7639e-01], + [-8.4982e-03, 1.0651e-01, -1.2955e-01, ..., -1.4422e-01, + -8.4736e-02, 1.3267e-01], + [ 5.0517e-02, -1.5909e-01, -1.9340e-01, ..., 8.7677e-05, + -6.2789e-02, -1.2309e-01], + ..., + [ 1.1590e-01, -1.1202e-01, 1.4864e-01, ..., 7.7074e-02, + 2.2661e-01, -2.5257e-02], + [ 1.4436e-01, -2.6775e-01, -1.8416e-01, ..., -3.7520e-01, + -1.2431e-01, 2.3752e-01], + [-2.1876e-01, 9.4091e-02, 7.5816e-02, ..., -3.6331e-01, + -1.6984e-01, -1.1125e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, -5.5879e-09], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., -0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [-9.3132e-10, 4.6566e-09, -1.8626e-09, ..., -0.0000e+00, + -9.3132e-10, 6.5193e-09], + [-4.6566e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -6.4261e-08], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0150, -0.0220, -0.0104, -0.0200, 0.0046, -0.0005, 0.0087, 0.0229, + 0.0118, -0.0118], device='cuda:0'), grad: tensor([ 6.5193e-09, -9.3132e-10, 2.7940e-09, -9.3132e-09, -1.7695e-08, + -3.7253e-09, 8.6613e-08, 1.6764e-08, -8.3819e-08, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 214.72, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4291 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 446, weight, value: tensor([[-3.2147e-01, 8.7118e-02, -2.1366e-01, ..., -1.1412e-01, + -2.9171e-01, -1.7640e-01], + [-8.4646e-03, 1.0663e-01, -1.2951e-01, ..., -1.4440e-01, + -8.4726e-02, 1.3277e-01], + [ 5.0585e-02, -1.5913e-01, -1.9348e-01, ..., 1.8265e-04, + -6.2715e-02, -1.2306e-01], + ..., + [ 1.1585e-01, -1.1227e-01, 1.4863e-01, ..., 7.7017e-02, + 2.2659e-01, -2.5363e-02], + [ 1.4439e-01, -2.6784e-01, -1.8423e-01, ..., -3.7530e-01, + -1.2439e-01, 2.3759e-01], + [-2.1881e-01, 9.4123e-02, 7.5860e-02, ..., -3.6338e-01, + -1.6986e-01, -1.1131e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-09, 1.8626e-09, 2.7940e-09, ..., 1.8626e-09, + 3.7253e-09, 9.3132e-10], + ..., + [-4.6566e-09, 0.0000e+00, -1.8626e-09, ..., -1.8626e-09, + -3.7253e-09, 0.0000e+00], + [-3.7253e-09, 2.0489e-08, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 3.7253e-09, -3.4459e-08, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0150, -0.0220, -0.0103, -0.0201, 0.0045, -0.0005, 0.0087, 0.0228, + 0.0117, -0.0118], device='cuda:0'), grad: tensor([-3.0734e-08, 9.3132e-09, 1.4901e-08, 1.7695e-08, 1.6764e-08, + 1.6764e-08, 1.1176e-08, -5.5879e-09, 6.0536e-08, -1.0990e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 214.81, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4334 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.08 lr 0.00001000 +Epoch 447, weight, value: tensor([[-3.2152e-01, 8.7139e-02, -2.1369e-01, ..., -1.1419e-01, + -2.9175e-01, -1.7642e-01], + [-8.4503e-03, 1.0674e-01, -1.2949e-01, ..., -1.4453e-01, + -8.4714e-02, 1.3286e-01], + [ 5.0608e-02, -1.5921e-01, -1.9357e-01, ..., 2.1627e-04, + -6.2683e-02, -1.2304e-01], + ..., + [ 1.1583e-01, -1.1246e-01, 1.4865e-01, ..., 7.7019e-02, + 2.2658e-01, -2.5442e-02], + [ 1.4438e-01, -2.6795e-01, -1.8429e-01, ..., -3.7538e-01, + -1.2447e-01, 2.3762e-01], + [-2.1887e-01, 9.4146e-02, 7.5888e-02, ..., -3.6346e-01, + -1.6989e-01, -1.1138e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.8626e-09, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [-8.3819e-09, 0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0150, -0.0219, -0.0103, -0.0201, 0.0045, -0.0005, 0.0087, 0.0228, + 0.0116, -0.0118], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.0245e-08, -4.4703e-08, 8.3819e-09, 1.8626e-09, + -2.7940e-09, 1.3039e-08, 4.6566e-09, 3.7253e-09, 9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 214.66, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4485 re_mapping 0.0029 re_causal 0.0102 /// teacc 99.11 lr 0.00001000 +Epoch 448, weight, value: tensor([[-3.2156e-01, 8.7184e-02, -2.1371e-01, ..., -1.1422e-01, + -2.9177e-01, -1.7642e-01], + [-8.4533e-03, 1.0676e-01, -1.2948e-01, ..., -1.4471e-01, + -8.4728e-02, 1.3291e-01], + [ 5.0642e-02, -1.5927e-01, -1.9366e-01, ..., 2.5576e-04, + -6.2654e-02, -1.2298e-01], + ..., + [ 1.1583e-01, -1.1257e-01, 1.4864e-01, ..., 7.7020e-02, + 2.2659e-01, -2.5492e-02], + [ 1.4437e-01, -2.6807e-01, -1.8441e-01, ..., -3.7546e-01, + -1.2454e-01, 2.3766e-01], + [-2.1890e-01, 9.4147e-02, 7.5973e-02, ..., -3.6352e-01, + -1.6991e-01, -1.1143e-01]], device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 2.7940e-09, 1.8626e-09], + [ 3.7253e-09, -1.8626e-09, 1.8626e-09, ..., 9.3132e-10, + 1.8626e-09, -9.3132e-10], + [-1.9558e-08, 0.0000e+00, 9.3132e-10, ..., -1.5832e-08, + -4.6566e-09, 1.8626e-09], + ..., + [-2.7940e-09, 9.3132e-10, -1.8626e-09, ..., -1.8626e-09, + -4.6566e-09, 1.8626e-09], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 2.7940e-09, -1.6764e-08], + [ 2.7940e-09, -1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 3.7253e-09]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0149, -0.0219, -0.0102, -0.0201, 0.0045, -0.0005, 0.0087, 0.0227, + 0.0115, -0.0118], device='cuda:0'), grad: tensor([ 5.2154e-08, 2.7940e-09, -1.0617e-07, 4.2841e-08, 1.3039e-08, + -4.3772e-08, 9.3132e-09, 1.8626e-09, 2.7940e-08, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 214.70, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4007 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.12 lr 0.00001000 +Epoch 449, weight, value: tensor([[-3.2163e-01, 8.7166e-02, -2.1381e-01, ..., -1.1426e-01, + -2.9181e-01, -1.7644e-01], + [-8.4302e-03, 1.0681e-01, -1.2947e-01, ..., -1.4482e-01, + -8.4709e-02, 1.3299e-01], + [ 5.0651e-02, -1.5932e-01, -1.9373e-01, ..., 2.8155e-04, + -6.2646e-02, -1.2299e-01], + ..., + [ 1.1582e-01, -1.1272e-01, 1.4864e-01, ..., 7.7034e-02, + 2.2659e-01, -2.5557e-02], + [ 1.4429e-01, -2.6820e-01, -1.8456e-01, ..., -3.7560e-01, + -1.2467e-01, 2.3763e-01], + [-2.1892e-01, 9.4267e-02, 7.6056e-02, ..., -3.6358e-01, + -1.6993e-01, -1.1145e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, 5.5879e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, -4.6566e-09, 8.3819e-09, ..., 4.6566e-09, + 5.5879e-09, -7.1712e-08], + [ 1.3970e-08, 1.8626e-09, 1.3039e-08, ..., 8.3819e-09, + 1.0245e-08, 8.3819e-09], + ..., + [-2.7008e-08, 4.6566e-09, -2.1420e-08, ..., -1.2107e-08, + -2.2352e-08, 5.1223e-08], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 5.5879e-09], + [ 2.7940e-09, 1.1176e-08, 1.5832e-08, ..., 7.4506e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0150, -0.0219, -0.0102, -0.0201, 0.0045, -0.0005, 0.0087, 0.0227, + 0.0114, -0.0117], device='cuda:0'), grad: tensor([ 1.2107e-08, -7.6368e-08, 3.8184e-08, -9.4064e-08, 7.4506e-09, + 5.0291e-08, 9.3132e-10, 1.7695e-08, 9.3132e-09, 3.4459e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 214.83, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4423 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.11 lr 0.00001000 +Epoch 450, weight, value: tensor([[-3.2167e-01, 8.7179e-02, -2.1386e-01, ..., -1.1428e-01, + -2.9183e-01, -1.7645e-01], + [-8.4034e-03, 1.0692e-01, -1.2941e-01, ..., -1.4490e-01, + -8.4705e-02, 1.3310e-01], + [ 5.0666e-02, -1.5939e-01, -1.9379e-01, ..., 2.9645e-04, + -6.2621e-02, -1.2302e-01], + ..., + [ 1.1580e-01, -1.1292e-01, 1.4860e-01, ..., 7.7028e-02, + 2.2658e-01, -2.5653e-02], + [ 1.4433e-01, -2.6828e-01, -1.8463e-01, ..., -3.7566e-01, + -1.2472e-01, 2.3772e-01], + [-2.1900e-01, 9.4276e-02, 7.6106e-02, ..., -3.6365e-01, + -1.6996e-01, -1.1158e-01]], device='cuda:0'), grad: tensor([[ 5.5879e-09, -1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08], + [-2.5146e-08, -3.1665e-08, 3.7253e-09, ..., 1.8626e-09, + 9.3132e-10, -1.4063e-07], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, 5.5879e-09], + ..., + [ 2.5146e-08, 3.2596e-08, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, 1.3132e-07], + [-1.6764e-08, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.3574e-08], + [-9.3132e-10, -2.1420e-08, -3.2596e-08, ..., 0.0000e+00, + 0.0000e+00, 2.8871e-08]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0150, -0.0218, -0.0101, -0.0201, 0.0045, -0.0005, 0.0087, 0.0226, + 0.0114, -0.0118], device='cuda:0'), grad: tensor([ 5.7742e-08, -3.3155e-07, 1.2107e-08, 1.4901e-08, 6.1467e-08, + 1.3039e-08, 4.3772e-08, 3.2876e-07, -2.0303e-07, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 214.87, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4176 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 451, weight, value: tensor([[-3.2173e-01, 8.7190e-02, -2.1395e-01, ..., -1.1432e-01, + -2.9185e-01, -1.7647e-01], + [-8.3951e-03, 1.0708e-01, -1.2943e-01, ..., -1.4502e-01, + -8.4713e-02, 1.3321e-01], + [ 5.0699e-02, -1.5950e-01, -1.9387e-01, ..., 3.4302e-04, + -6.2581e-02, -1.2301e-01], + ..., + [ 1.1578e-01, -1.1314e-01, 1.4862e-01, ..., 7.7003e-02, + 2.2659e-01, -2.5740e-02], + [ 1.4433e-01, -2.6843e-01, -1.8478e-01, ..., -3.7580e-01, + -1.2485e-01, 2.3778e-01], + [-2.1906e-01, 9.4314e-02, 7.6228e-02, ..., -3.6374e-01, + -1.6998e-01, -1.1168e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + -9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -3.1665e-08, -3.8184e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0150, -0.0218, -0.0101, -0.0201, 0.0045, -0.0006, 0.0087, 0.0226, + 0.0113, -0.0118], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.7253e-09, -1.0245e-08, 1.8626e-09, 6.9849e-08, + 9.3132e-10, 0.0000e+00, 2.7940e-09, 3.7253e-09, -6.8918e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 215.30, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4185 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.12 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.3218, 0.0872, -0.2141, ..., -0.1144, -0.2919, -0.1765], + [-0.0084, 0.1072, -0.1294, ..., -0.1452, -0.0847, 0.1333], + [ 0.0507, -0.1596, -0.1939, ..., 0.0004, -0.0625, -0.1230], + ..., + [ 0.1158, -0.1133, 0.1486, ..., 0.0770, 0.2266, -0.0258], + [ 0.1444, -0.2685, -0.1849, ..., -0.3760, -0.1250, 0.2379], + [-0.2191, 0.0943, 0.0763, ..., -0.3638, -0.1700, -0.1117]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, -3.7253e-09, 3.7253e-09, ..., 9.3132e-10, + 2.7940e-09, -1.5832e-08], + [ 3.7253e-09, 9.3132e-10, 2.7940e-09, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00], + ..., + [-8.3819e-09, 3.7253e-09, -8.3819e-09, ..., -3.7253e-09, + -7.4506e-09, 1.2107e-08], + [-0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 1.3970e-08, 3.8184e-08, 2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-08]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0150, -0.0218, -0.0100, -0.0201, 0.0044, -0.0007, 0.0087, 0.0225, + 0.0112, -0.0118], device='cuda:0'), grad: tensor([-3.2596e-08, -2.4214e-08, 1.0245e-08, 5.5879e-09, -1.5646e-07, + 1.8626e-09, 1.3970e-08, 1.3970e-08, 0.0000e+00, 1.6578e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 214.98, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4444 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.16 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.3219, 0.0872, -0.2143, ..., -0.1145, -0.2919, -0.1765], + [-0.0084, 0.1073, -0.1294, ..., -0.1454, -0.0847, 0.1334], + [ 0.0507, -0.1596, -0.1940, ..., 0.0004, -0.0625, -0.1230], + ..., + [ 0.1158, -0.1135, 0.1487, ..., 0.0770, 0.2266, -0.0259], + [ 0.1444, -0.2686, -0.1850, ..., -0.3761, -0.1251, 0.2380], + [-0.2191, 0.0944, 0.0765, ..., -0.3639, -0.1700, -0.1117]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 3.7253e-09, ..., 2.7940e-09, + 9.3132e-10, -7.4506e-09], + [-9.3132e-10, 9.3132e-10, 1.8626e-09, ..., -0.0000e+00, + -0.0000e+00, 0.0000e+00], + ..., + [-1.8626e-09, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + -9.3132e-10, 4.6566e-09], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, -9.3132e-10, -3.7253e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0150, -0.0217, -0.0100, -0.0201, 0.0044, -0.0007, 0.0087, 0.0225, + 0.0112, -0.0117], device='cuda:0'), grad: tensor([ 9.3132e-10, -4.6566e-09, -1.8626e-09, -8.0094e-08, 9.3132e-09, + 6.0536e-08, 9.3132e-09, 6.5193e-09, 6.5193e-09, -4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 214.83, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4656 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.12 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.3219, 0.0872, -0.2143, ..., -0.1145, -0.2920, -0.1765], + [-0.0084, 0.1073, -0.1295, ..., -0.1457, -0.0848, 0.1334], + [ 0.0507, -0.1597, -0.1941, ..., 0.0004, -0.0625, -0.1231], + ..., + [ 0.1158, -0.1135, 0.1487, ..., 0.0771, 0.2266, -0.0259], + [ 0.1445, -0.2688, -0.1851, ..., -0.3763, -0.1252, 0.2381], + [-0.2193, 0.0944, 0.0765, ..., -0.3640, -0.1701, -0.1119]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.1514e-07, 0.0000e+00, ..., -6.4261e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-08, 2.7940e-09, 0.0000e+00, ..., 9.0338e-08, + 1.4901e-08, 4.7497e-08], + [-1.0710e-07, 2.7940e-09, 0.0000e+00, ..., -1.0058e-07, + -1.7695e-08, -5.5879e-08], + ..., + [ 1.0245e-08, 1.8626e-09, 9.3132e-10, ..., 1.0245e-08, + 1.8626e-09, 7.4506e-09], + [ 0.0000e+00, 1.4901e-08, 9.3132e-10, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -5.5879e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0149, -0.0217, -0.0100, -0.0201, 0.0044, -0.0007, 0.0087, 0.0225, + 0.0113, -0.0118], device='cuda:0'), grad: tensor([-5.0850e-07, 5.2433e-07, -5.8115e-07, 4.5169e-07, 7.4506e-09, + 3.7253e-09, 7.4506e-09, 6.1467e-08, 3.7253e-08, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 214.61, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4228 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.14 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.3220, 0.0873, -0.2143, ..., -0.1145, -0.2920, -0.1765], + [-0.0084, 0.1073, -0.1295, ..., -0.1459, -0.0848, 0.1334], + [ 0.0507, -0.1599, -0.1942, ..., 0.0005, -0.0625, -0.1231], + ..., + [ 0.1159, -0.1135, 0.1488, ..., 0.0771, 0.2268, -0.0259], + [ 0.1445, -0.2689, -0.1852, ..., -0.3764, -0.1253, 0.2382], + [-0.2194, 0.0944, 0.0765, ..., -0.3642, -0.1701, -0.1120]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.3039e-08, 1.2107e-08, ..., 4.6566e-09, + 2.7940e-09, 2.7940e-09], + [ 0.0000e+00, 9.3132e-10, -5.5879e-09, ..., -5.5879e-09, + -4.6566e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-09, 5.5879e-09, ..., 1.8626e-09, + 9.3132e-10, 4.6566e-09], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.5635e-08, 1.1176e-08, ..., 9.3132e-10, + 1.8626e-09, 2.0489e-08]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0149, -0.0218, -0.0100, -0.0202, 0.0045, -0.0008, 0.0087, 0.0226, + 0.0112, -0.0119], device='cuda:0'), grad: tensor([ 1.0245e-08, 7.0781e-08, -7.9162e-08, 0.0000e+00, -1.8999e-07, + -1.9558e-08, 1.8626e-08, 3.9116e-08, 7.4506e-09, 1.5367e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 214.78, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4406 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.14 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.3220, 0.0873, -0.2143, ..., -0.1146, -0.2920, -0.1765], + [-0.0085, 0.1073, -0.1296, ..., -0.1460, -0.0849, 0.1335], + [ 0.0507, -0.1600, -0.1943, ..., 0.0005, -0.0625, -0.1232], + ..., + [ 0.1159, -0.1136, 0.1489, ..., 0.0771, 0.2268, -0.0260], + [ 0.1445, -0.2690, -0.1853, ..., -0.3765, -0.1253, 0.2383], + [-0.2194, 0.0944, 0.0765, ..., -0.3643, -0.1702, -0.1121]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.7055e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + [-8.3819e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.7300e-08], + [ 2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.9558e-08], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-08], + [ 0.0000e+00, 2.7008e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0149, -0.0218, -0.0100, -0.0202, 0.0045, -0.0009, 0.0088, 0.0226, + 0.0112, -0.0119], device='cuda:0'), grad: tensor([-1.3784e-07, -9.4064e-08, 2.7008e-08, 2.7940e-09, 2.7940e-09, + 1.0245e-08, 8.3819e-08, 1.9558e-08, 3.7253e-08, 5.8673e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 214.57, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4282 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.12 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.3220, 0.0874, -0.2144, ..., -0.1146, -0.2920, -0.1766], + [-0.0085, 0.1075, -0.1296, ..., -0.1462, -0.0849, 0.1336], + [ 0.0508, -0.1602, -0.1944, ..., 0.0005, -0.0625, -0.1232], + ..., + [ 0.1159, -0.1138, 0.1490, ..., 0.0772, 0.2269, -0.0260], + [ 0.1446, -0.2691, -0.1854, ..., -0.3766, -0.1254, 0.2384], + [-0.2196, 0.0944, 0.0766, ..., -0.3644, -0.1702, -0.1123]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 2.6077e-08, 9.3132e-10, 2.0489e-08, ..., 2.7940e-08, + 2.4214e-08, 0.0000e+00], + [-3.2596e-08, 0.0000e+00, 1.0245e-08, ..., -3.5390e-08, + -2.7940e-09, 9.3132e-10], + ..., + [-1.5832e-08, 0.0000e+00, -4.6566e-08, ..., -6.5193e-09, + -4.3772e-08, 9.3132e-10], + [ 2.7940e-09, 9.3132e-10, 5.5879e-09, ..., 2.7940e-09, + 6.5193e-09, -2.7940e-09], + [ 3.7253e-09, -0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0148, -0.0218, -0.0100, -0.0201, 0.0044, -0.0009, 0.0088, 0.0226, + 0.0112, -0.0120], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.6578e-07, -2.1420e-07, 8.1956e-08, 2.5146e-08, + -5.0291e-08, 5.5879e-09, -3.5390e-08, 7.4506e-09, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 215.10, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4305 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.13 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.3221, 0.0874, -0.2144, ..., -0.1146, -0.2921, -0.1766], + [-0.0085, 0.1075, -0.1296, ..., -0.1463, -0.0849, 0.1336], + [ 0.0508, -0.1603, -0.1945, ..., 0.0006, -0.0624, -0.1232], + ..., + [ 0.1159, -0.1138, 0.1490, ..., 0.0771, 0.2268, -0.0261], + [ 0.1447, -0.2692, -0.1855, ..., -0.3767, -0.1255, 0.2385], + [-0.2196, 0.0943, 0.0767, ..., -0.3645, -0.1702, -0.1124]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1176e-08, -1.8626e-09, 6.5193e-09, ..., 2.7940e-09, + 3.7253e-09, -7.4506e-09], + [-5.5879e-08, 9.3132e-10, 9.3132e-10, ..., -3.7253e-09, + 9.3132e-10, 9.3132e-10], + ..., + [-4.6566e-09, 3.7253e-09, -6.5193e-09, ..., -1.8626e-09, + -4.6566e-09, 5.5879e-09], + [ 3.3528e-08, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, -1.5832e-08], + [ 3.7253e-09, 5.5879e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0148, -0.0218, -0.0099, -0.0201, 0.0044, -0.0009, 0.0088, 0.0225, + 0.0112, -0.0121], device='cuda:0'), grad: tensor([ 0.0000e+00, 2.1420e-08, -1.7881e-07, -1.3970e-08, -1.7695e-08, + 2.1420e-08, 3.4459e-08, 1.0245e-08, 1.0617e-07, 2.4214e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 214.68, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4558 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.10 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.3221, 0.0875, -0.2145, ..., -0.1147, -0.2921, -0.1766], + [-0.0085, 0.1075, -0.1297, ..., -0.1464, -0.0849, 0.1337], + [ 0.0508, -0.1604, -0.1948, ..., 0.0006, -0.0625, -0.1233], + ..., + [ 0.1159, -0.1139, 0.1491, ..., 0.0772, 0.2269, -0.0261], + [ 0.1448, -0.2693, -0.1856, ..., -0.3768, -0.1255, 0.2387], + [-0.2197, 0.0943, 0.0768, ..., -0.3646, -0.1702, -0.1125]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 5.5879e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-09, 3.7253e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [-1.8626e-09, 9.3132e-10, -9.3132e-10, ..., -0.0000e+00, + -1.8626e-09, 1.8626e-09], + [-1.5926e-07, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -2.1607e-07], + [ 9.3132e-10, -1.4901e-08, -1.9558e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0147, -0.0218, -0.0100, -0.0202, 0.0045, -0.0009, 0.0088, 0.0226, + 0.0111, -0.0121], device='cuda:0'), grad: tensor([ 1.7695e-08, 6.2399e-08, 5.5879e-09, 4.2841e-08, 7.4506e-09, + -9.3132e-10, 7.3854e-07, 1.8626e-09, -8.3540e-07, -3.5390e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 215.19, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4086 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.09 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.3221, 0.0875, -0.2145, ..., -0.1147, -0.2921, -0.1766], + [-0.0086, 0.1075, -0.1297, ..., -0.1466, -0.0850, 0.1337], + [ 0.0508, -0.1605, -0.1949, ..., 0.0006, -0.0625, -0.1233], + ..., + [ 0.1160, -0.1140, 0.1492, ..., 0.0772, 0.2270, -0.0261], + [ 0.1449, -0.2695, -0.1857, ..., -0.3769, -0.1256, 0.2388], + [-0.2198, 0.0943, 0.0768, ..., -0.3647, -0.1703, -0.1126]], + device='cuda:0'), grad: tensor([[-1.8626e-09, -8.3819e-09, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, -2.7940e-09], + [ 9.3132e-10, -0.0000e+00, 1.8626e-08, ..., 1.3970e-08, + 1.8626e-09, -1.8626e-08], + [-8.3819e-09, 1.8626e-09, 4.6566e-09, ..., -3.7253e-09, + -4.6566e-09, 9.3132e-10], + ..., + [ 4.6566e-09, 5.5879e-09, 9.3132e-10, ..., 6.5193e-09, + 9.3132e-10, 9.3132e-09], + [ 1.8626e-09, 4.6566e-09, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, 8.3819e-09], + [ 9.3132e-10, -9.3132e-10, -4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0147, -0.0218, -0.0099, -0.0203, 0.0045, -0.0009, 0.0088, 0.0226, + 0.0112, -0.0122], device='cuda:0'), grad: tensor([-1.4901e-08, 1.4901e-08, -1.3039e-08, -6.7987e-08, 8.3819e-09, + 7.4506e-09, -1.8626e-09, 4.3772e-08, 2.6077e-08, 9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 215.51, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4545 re_mapping 0.0026 re_causal 0.0100 /// teacc 99.12 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.3222, 0.0875, -0.2145, ..., -0.1148, -0.2922, -0.1766], + [-0.0086, 0.1076, -0.1298, ..., -0.1466, -0.0851, 0.1338], + [ 0.0508, -0.1606, -0.1951, ..., 0.0006, -0.0625, -0.1233], + ..., + [ 0.1160, -0.1141, 0.1493, ..., 0.0772, 0.2271, -0.0261], + [ 0.1449, -0.2696, -0.1858, ..., -0.3769, -0.1257, 0.2389], + [-0.2200, 0.0944, 0.0768, ..., -0.3648, -0.1704, -0.1128]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 4.0978e-08, 0.0000e+00, ..., 2.7940e-09, + 3.7253e-09, 6.2399e-08], + [ 2.5146e-08, -9.0804e-08, 1.2107e-08, ..., 1.6764e-08, + 2.5611e-08, -1.3644e-07], + [-5.8673e-08, 2.3283e-09, 1.4435e-08, ..., -4.6566e-08, + -5.2154e-08, 5.5879e-09], + ..., + [ 1.2573e-08, 2.1886e-08, -3.3062e-08, ..., 1.4901e-08, + 5.1223e-09, 2.7940e-08], + [ 4.1910e-09, 9.7789e-09, 0.0000e+00, ..., 2.7940e-09, + 4.1910e-09, 1.4901e-08], + [ 3.2596e-09, 3.2596e-09, 1.8626e-09, ..., 1.8626e-09, + 3.2596e-09, 5.1223e-09]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0147, -0.0218, -0.0099, -0.0203, 0.0046, -0.0008, 0.0087, 0.0226, + 0.0111, -0.0122], device='cuda:0'), grad: tensor([ 1.8487e-07, -2.9150e-07, -2.5425e-07, 1.6298e-08, 1.4901e-08, + 2.8871e-08, 4.5169e-08, 1.9278e-07, 5.5414e-08, 2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 215.11, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4339 re_mapping 0.0027 re_causal 0.0100 /// teacc 99.14 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.3223, 0.0875, -0.2146, ..., -0.1148, -0.2922, -0.1766], + [-0.0085, 0.1078, -0.1298, ..., -0.1467, -0.0850, 0.1339], + [ 0.0508, -0.1607, -0.1952, ..., 0.0007, -0.0625, -0.1234], + ..., + [ 0.1160, -0.1143, 0.1493, ..., 0.0772, 0.2270, -0.0262], + [ 0.1448, -0.2698, -0.1859, ..., -0.3770, -0.1258, 0.2388], + [-0.2201, 0.0944, 0.0769, ..., -0.3649, -0.1704, -0.1129]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 9.3132e-10, 1.3970e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.8626e-09, -8.3819e-09, -4.6566e-10, ..., 1.3970e-09, + 1.8626e-09, -2.0489e-08], + [-0.0000e+00, 4.6566e-10, 2.3283e-09, ..., -2.7940e-09, + -2.7940e-09, 4.1910e-09], + ..., + [ 2.7940e-09, 1.1642e-08, 2.3283e-09, ..., 2.7940e-09, + 4.6566e-10, 2.1420e-08], + [-3.3528e-08, 1.8626e-09, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -6.1933e-08], + [-1.3970e-09, -2.3283e-09, -1.4435e-08, ..., 0.0000e+00, + -0.0000e+00, 5.1223e-09]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0147, -0.0217, -0.0099, -0.0203, 0.0046, -0.0008, 0.0087, 0.0225, + 0.0109, -0.0123], device='cuda:0'), grad: tensor([ 4.6566e-09, -3.6787e-08, 0.0000e+00, 6.9384e-08, -1.2573e-08, + -6.5193e-08, 7.0781e-08, 6.1467e-08, -8.2888e-08, 4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 214.54, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4392 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.14 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.3223, 0.0876, -0.2146, ..., -0.1149, -0.2922, -0.1767], + [-0.0086, 0.1078, -0.1299, ..., -0.1469, -0.0851, 0.1339], + [ 0.0508, -0.1608, -0.1953, ..., 0.0007, -0.0624, -0.1234], + ..., + [ 0.1161, -0.1143, 0.1494, ..., 0.0772, 0.2271, -0.0262], + [ 0.1448, -0.2698, -0.1860, ..., -0.3771, -0.1259, 0.2388], + [-0.2201, 0.0943, 0.0769, ..., -0.3651, -0.1705, -0.1130]], + device='cuda:0'), grad: tensor([[-0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, -1.3970e-09], + [ 4.6566e-09, 4.6566e-10, 9.3132e-10, ..., 2.7940e-09, + 0.0000e+00, 3.7253e-09], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-09], + [-6.5193e-09, 4.6566e-10, 4.6566e-10, ..., -2.7940e-09, + 0.0000e+00, -4.6566e-09], + [ 0.0000e+00, 1.1642e-08, 3.2596e-09, ..., 0.0000e+00, + -0.0000e+00, 3.2596e-09]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0147, -0.0217, -0.0098, -0.0202, 0.0046, -0.0009, 0.0087, 0.0226, + 0.0109, -0.0124], device='cuda:0'), grad: tensor([-3.2596e-09, 4.6566e-09, 1.4901e-08, -3.2596e-09, -3.5390e-08, + 1.5832e-08, -5.1223e-08, 7.9162e-09, 2.4680e-08, 3.4459e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 214.91, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4106 re_mapping 0.0027 re_causal 0.0095 /// teacc 99.16 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.3223, 0.0876, -0.2146, ..., -0.1149, -0.2922, -0.1767], + [-0.0086, 0.1080, -0.1300, ..., -0.1471, -0.0852, 0.1341], + [ 0.0509, -0.1609, -0.1954, ..., 0.0008, -0.0624, -0.1234], + ..., + [ 0.1161, -0.1145, 0.1495, ..., 0.0771, 0.2272, -0.0263], + [ 0.1448, -0.2700, -0.1862, ..., -0.3772, -0.1260, 0.2389], + [-0.2203, 0.0943, 0.0770, ..., -0.3652, -0.1705, -0.1132]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.0734e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.9162e-08], + [ 8.3819e-08, -1.0151e-07, 3.9116e-08, ..., 9.3132e-09, + 5.1223e-08, -1.2759e-07], + [-9.3132e-10, 2.7940e-09, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, 5.5879e-09], + ..., + [-9.8720e-08, 2.5146e-08, -4.6566e-08, ..., -1.0245e-08, + -5.9605e-08, -9.3132e-10], + [ 4.6566e-09, 7.4506e-09, 2.7940e-09, ..., 9.3132e-10, + 3.7253e-09, 8.3819e-09], + [ 8.3819e-09, 3.0734e-08, 5.5879e-09, ..., 9.3132e-10, + 4.6566e-09, 1.5832e-08]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0147, -0.0217, -0.0098, -0.0203, 0.0047, -0.0009, 0.0087, 0.0225, + 0.0108, -0.0125], device='cuda:0'), grad: tensor([ 1.7416e-07, -2.8312e-07, 1.3039e-08, 3.7253e-09, -1.9558e-08, + 2.7940e-09, 3.1665e-08, -7.1712e-08, 4.7497e-08, 9.9652e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 214.89, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4100 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.13 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.3223, 0.0876, -0.2147, ..., -0.1149, -0.2923, -0.1767], + [-0.0087, 0.1081, -0.1300, ..., -0.1471, -0.0852, 0.1342], + [ 0.0509, -0.1610, -0.1954, ..., 0.0008, -0.0624, -0.1235], + ..., + [ 0.1161, -0.1146, 0.1496, ..., 0.0771, 0.2272, -0.0263], + [ 0.1447, -0.2702, -0.1863, ..., -0.3773, -0.1261, 0.2389], + [-0.2204, 0.0943, 0.0770, ..., -0.3653, -0.1706, -0.1133]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -6.5193e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-3.3528e-08, -1.3970e-08, 5.5879e-09, ..., 2.7940e-09, + 5.5879e-09, -8.1025e-08], + [ 9.3132e-09, 3.7253e-09, 3.7253e-09, ..., 2.7940e-09, + 9.3132e-10, 1.8626e-08], + ..., + [-2.7940e-09, 2.7940e-09, -7.4506e-09, ..., -2.7940e-09, + -9.3132e-09, 1.3970e-08], + [ 1.4901e-08, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.5390e-08], + [ 1.8626e-09, 1.8626e-09, -2.7940e-09, ..., 9.3132e-10, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0147, -0.0217, -0.0098, -0.0203, 0.0047, -0.0009, 0.0087, 0.0225, + 0.0106, -0.0125], device='cuda:0'), grad: tensor([-1.3039e-08, -2.2724e-07, 5.9605e-08, -9.3132e-09, 6.5193e-09, + 9.3132e-09, 3.2596e-08, 2.6077e-08, 1.1269e-07, 1.0245e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 214.63, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4240 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.13 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.3223, 0.0877, -0.2147, ..., -0.1150, -0.2923, -0.1767], + [-0.0087, 0.1081, -0.1301, ..., -0.1472, -0.0853, 0.1342], + [ 0.0509, -0.1612, -0.1955, ..., 0.0008, -0.0624, -0.1235], + ..., + [ 0.1162, -0.1147, 0.1497, ..., 0.0772, 0.2273, -0.0263], + [ 0.1448, -0.2703, -0.1864, ..., -0.3774, -0.1262, 0.2389], + [-0.2205, 0.0943, 0.0771, ..., -0.3653, -0.1706, -0.1134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, -1.6764e-08, 7.4506e-09, ..., 3.7253e-09, + 5.5879e-09, -2.7940e-09], + [-1.8626e-08, 9.3132e-10, 4.6566e-09, ..., -1.2107e-08, + -1.3039e-08, 0.0000e+00], + ..., + [-3.7253e-09, 1.2107e-08, -3.3528e-08, ..., -9.3132e-09, + -4.6566e-09, 2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 1.8626e-09, -2.7940e-09, -1.2107e-08, ..., 1.8626e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0146, -0.0217, -0.0098, -0.0203, 0.0047, -0.0009, 0.0087, 0.0226, + 0.0106, -0.0126], device='cuda:0'), grad: tensor([-3.7253e-09, -5.1223e-08, -3.7253e-08, 3.7253e-08, 3.6322e-08, + 6.9849e-08, -7.0781e-08, 2.4214e-08, 6.5193e-09, -9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 214.65, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4345 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.17 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.3224, 0.0876, -0.2149, ..., -0.1150, -0.2923, -0.1767], + [-0.0087, 0.1082, -0.1301, ..., -0.1473, -0.0853, 0.1342], + [ 0.0510, -0.1613, -0.1956, ..., 0.0009, -0.0623, -0.1235], + ..., + [ 0.1162, -0.1148, 0.1497, ..., 0.0771, 0.2274, -0.0264], + [ 0.1448, -0.2705, -0.1865, ..., -0.3774, -0.1262, 0.2390], + [-0.2206, 0.0944, 0.0771, ..., -0.3654, -0.1707, -0.1135]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, -9.3132e-10, 5.5879e-09, ..., 3.7253e-09, + 5.5879e-09, -5.5879e-09], + [ 1.0245e-08, 0.0000e+00, 1.4901e-08, ..., 3.7253e-09, + 1.3039e-08, 9.3132e-10], + ..., + [-1.8626e-08, 6.5193e-09, -1.7695e-08, ..., -1.2107e-08, + -2.3283e-08, 7.4506e-09], + [ 9.3132e-10, 3.7253e-09, 6.5193e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, -5.1223e-08, -7.9162e-08, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0146, -0.0217, -0.0097, -0.0203, 0.0047, -0.0009, 0.0087, 0.0226, + 0.0105, -0.0126], device='cuda:0'), grad: tensor([ 2.7940e-09, 5.5879e-09, 1.0245e-08, 1.2107e-08, 1.7043e-07, + 2.7940e-09, -3.7253e-09, -5.5879e-09, 2.0489e-08, -2.1141e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 215.11, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4220 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.12 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.3224, 0.0877, -0.2150, ..., -0.1150, -0.2923, -0.1767], + [-0.0086, 0.1084, -0.1300, ..., -0.1475, -0.0852, 0.1345], + [ 0.0510, -0.1613, -0.1957, ..., 0.0010, -0.0623, -0.1235], + ..., + [ 0.1161, -0.1151, 0.1496, ..., 0.0772, 0.2273, -0.0267], + [ 0.1449, -0.2706, -0.1866, ..., -0.3775, -0.1263, 0.2391], + [-0.2207, 0.0944, 0.0772, ..., -0.3655, -0.1708, -0.1135]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -9.3132e-10, 9.3132e-10, ..., 3.7253e-09, + 2.7940e-09, 0.0000e+00], + [ 1.5832e-08, 9.3132e-10, 1.3039e-08, ..., 1.1176e-08, + 1.5832e-08, 9.3132e-10], + [ 2.3283e-08, 0.0000e+00, 4.7497e-08, ..., -1.4901e-08, + 2.5146e-08, 9.3132e-10], + ..., + [-5.8673e-08, 2.7940e-09, -6.7987e-08, ..., -1.5832e-08, + -6.1467e-08, 1.8626e-09], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, -2.7940e-09], + [ 8.3819e-09, 6.5193e-09, 3.7253e-09, ..., 4.6566e-09, + 9.3132e-09, 3.7253e-09]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0146, -0.0215, -0.0096, -0.0203, 0.0047, -0.0009, 0.0087, 0.0224, + 0.0105, -0.0127], device='cuda:0'), grad: tensor([ 9.3132e-09, 4.4703e-08, 5.5879e-09, 2.1420e-08, -2.7940e-08, + 9.3132e-10, 2.7940e-09, -9.3132e-08, -9.3132e-10, 4.1910e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 215.20, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4285 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.13 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.3224, 0.0877, -0.2150, ..., -0.1150, -0.2924, -0.1767], + [-0.0085, 0.1085, -0.1299, ..., -0.1477, -0.0851, 0.1347], + [ 0.0510, -0.1614, -0.1960, ..., 0.0010, -0.0623, -0.1235], + ..., + [ 0.1161, -0.1153, 0.1497, ..., 0.0773, 0.2273, -0.0268], + [ 0.1450, -0.2707, -0.1867, ..., -0.3776, -0.1263, 0.2392], + [-0.2209, 0.0944, 0.0772, ..., -0.3657, -0.1709, -0.1136]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.3970e-08, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, -3.7253e-09, 4.6566e-09, ..., 2.7940e-09, + 4.6566e-09, -1.7695e-08], + [-3.7253e-09, 0.0000e+00, 7.4506e-09, ..., -1.4901e-08, + -1.2107e-08, 9.3132e-10], + ..., + [-1.2107e-08, 1.8626e-09, -2.8871e-08, ..., 3.7253e-09, + -4.6566e-09, 1.0245e-08], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 1.8626e-09, 7.4506e-09, ..., 2.7940e-09, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0146, -0.0214, -0.0096, -0.0205, 0.0047, -0.0008, 0.0086, 0.0223, + 0.0105, -0.0127], device='cuda:0'), grad: tensor([-3.1665e-08, -2.4214e-08, -2.8871e-08, 1.1176e-08, 1.4901e-08, + 1.6764e-08, -1.8626e-09, 1.1176e-08, 8.3819e-09, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 215.25, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4227 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.11 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.3225, 0.0878, -0.2150, ..., -0.1151, -0.2924, -0.1767], + [-0.0086, 0.1085, -0.1299, ..., -0.1479, -0.0852, 0.1347], + [ 0.0510, -0.1615, -0.1961, ..., 0.0010, -0.0623, -0.1235], + ..., + [ 0.1161, -0.1154, 0.1497, ..., 0.0774, 0.2274, -0.0269], + [ 0.1451, -0.2708, -0.1868, ..., -0.3777, -0.1264, 0.2393], + [-0.2210, 0.0944, 0.0773, ..., -0.3658, -0.1710, -0.1137]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 8.3819e-09, ..., 3.7253e-09, + 3.7253e-09, -3.7253e-09], + [ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, + 3.7253e-09, 0.0000e+00], + ..., + [-9.3132e-09, 1.8626e-09, -1.7695e-08, ..., -9.3132e-09, + -1.2107e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -5.5879e-09, -6.5193e-09, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0146, -0.0214, -0.0096, -0.0206, 0.0047, -0.0009, 0.0086, 0.0223, + 0.0105, -0.0127], device='cuda:0'), grad: tensor([ 2.7940e-09, 6.5193e-09, 9.3132e-09, -1.4901e-08, 2.4214e-08, + 3.7253e-09, 9.3132e-10, -2.1420e-08, 1.8626e-09, -1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 214.93, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4278 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.11 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.3225, 0.0878, -0.2150, ..., -0.1151, -0.2924, -0.1768], + [-0.0086, 0.1085, -0.1300, ..., -0.1481, -0.0852, 0.1348], + [ 0.0510, -0.1615, -0.1962, ..., 0.0010, -0.0623, -0.1236], + ..., + [ 0.1162, -0.1154, 0.1498, ..., 0.0774, 0.2275, -0.0269], + [ 0.1453, -0.2708, -0.1868, ..., -0.3777, -0.1263, 0.2396], + [-0.2211, 0.0944, 0.0772, ..., -0.3659, -0.1711, -0.1137]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.7940e-09, 9.3132e-10, ..., -0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, -2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, -5.5879e-09], + [-9.3132e-09, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + -5.5879e-09, -7.4506e-09], + ..., + [ 4.6566e-09, 1.3970e-08, 2.0489e-08, ..., 0.0000e+00, + 5.5879e-09, 8.3819e-09], + [ 5.5879e-09, 7.4506e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09], + [-0.0000e+00, -9.3132e-09, -4.4703e-08, ..., 0.0000e+00, + -5.5879e-09, 3.7253e-09]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0146, -0.0215, -0.0095, -0.0206, 0.0047, -0.0010, 0.0087, 0.0224, + 0.0107, -0.0127], device='cuda:0'), grad: tensor([-1.8626e-09, -5.5879e-09, -2.6077e-08, 8.3819e-09, 6.1467e-08, + -8.9407e-08, 1.1176e-08, 7.4506e-08, 4.5635e-08, -6.7987e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 215.20, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4095 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.10 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.3226, 0.0878, -0.2150, ..., -0.1151, -0.2925, -0.1768], + [-0.0086, 0.1086, -0.1300, ..., -0.1483, -0.0852, 0.1349], + [ 0.0511, -0.1617, -0.1962, ..., 0.0013, -0.0621, -0.1237], + ..., + [ 0.1161, -0.1155, 0.1499, ..., 0.0772, 0.2274, -0.0270], + [ 0.1456, -0.2709, -0.1868, ..., -0.3778, -0.1263, 0.2399], + [-0.2212, 0.0945, 0.0773, ..., -0.3660, -0.1711, -0.1138]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 5.1223e-08, 5.5879e-09, 6.3330e-08, ..., 2.7940e-09, + 5.6811e-08, 1.3970e-08], + [-4.6566e-09, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, + -9.3132e-10, 0.0000e+00], + ..., + [-6.7055e-08, 1.8626e-09, -8.2888e-08, ..., -0.0000e+00, + -7.4506e-08, -1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-08, -1.8626e-09, 1.3039e-08, ..., 0.0000e+00, + 1.6764e-08, 4.6566e-09]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0146, -0.0215, -0.0094, -0.0206, 0.0046, -0.0010, 0.0087, 0.0223, + 0.0109, -0.0127], device='cuda:0'), grad: tensor([ 2.8871e-08, 1.2200e-07, -5.5879e-08, 1.6764e-08, -1.5832e-08, + 0.0000e+00, 9.3132e-10, -1.2759e-07, 9.3132e-10, 2.4214e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 214.85, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4433 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.13 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.3226, 0.0878, -0.2151, ..., -0.1151, -0.2925, -0.1768], + [-0.0086, 0.1086, -0.1301, ..., -0.1484, -0.0853, 0.1350], + [ 0.0511, -0.1618, -0.1964, ..., 0.0013, -0.0621, -0.1238], + ..., + [ 0.1162, -0.1156, 0.1500, ..., 0.0772, 0.2274, -0.0271], + [ 0.1458, -0.2711, -0.1869, ..., -0.3779, -0.1263, 0.2400], + [-0.2213, 0.0946, 0.0774, ..., -0.3661, -0.1712, -0.1138]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.2352e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 4.6566e-09, ..., 1.8626e-09, + 1.8626e-09, -1.8626e-09], + [-2.7008e-08, 0.0000e+00, 2.7940e-09, ..., -1.3970e-08, + -2.1420e-08, 0.0000e+00], + ..., + [ 2.4214e-08, 9.3132e-10, -2.7940e-09, ..., 1.3970e-08, + 1.8626e-08, 1.8626e-09], + [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 9.3132e-10, 2.4214e-08, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0146, -0.0215, -0.0094, -0.0206, 0.0045, -0.0010, 0.0088, 0.0223, + 0.0110, -0.0127], device='cuda:0'), grad: tensor([-4.7497e-08, 5.5879e-09, -4.3772e-08, -1.3970e-08, -3.7253e-09, + 4.6566e-09, 4.6566e-09, 4.5635e-08, -2.7940e-09, 5.5879e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 214.89, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4316 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.12 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.3226, 0.0878, -0.2152, ..., -0.1152, -0.2925, -0.1768], + [-0.0087, 0.1087, -0.1302, ..., -0.1486, -0.0854, 0.1350], + [ 0.0512, -0.1619, -0.1964, ..., 0.0014, -0.0620, -0.1238], + ..., + [ 0.1162, -0.1157, 0.1501, ..., 0.0771, 0.2275, -0.0271], + [ 0.1458, -0.2712, -0.1870, ..., -0.3780, -0.1264, 0.2402], + [-0.2215, 0.0946, 0.0773, ..., -0.3661, -0.1714, -0.1139]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 1.0245e-08, ..., 1.3039e-08, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 9.3132e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0146, -0.0215, -0.0093, -0.0205, 0.0046, -0.0011, 0.0088, 0.0223, + 0.0110, -0.0128], device='cuda:0'), grad: tensor([-1.4901e-08, 2.6077e-08, 8.3819e-09, -5.4948e-08, 1.8626e-09, + 0.0000e+00, 6.5193e-09, 1.9558e-08, 6.5193e-09, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 215.25, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4442 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.12 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.3227, 0.0879, -0.2153, ..., -0.1153, -0.2926, -0.1769], + [-0.0087, 0.1087, -0.1301, ..., -0.1487, -0.0854, 0.1351], + [ 0.0513, -0.1619, -0.1964, ..., 0.0016, -0.0618, -0.1239], + ..., + [ 0.1162, -0.1158, 0.1501, ..., 0.0770, 0.2275, -0.0272], + [ 0.1459, -0.2713, -0.1872, ..., -0.3781, -0.1265, 0.2402], + [-0.2215, 0.0945, 0.0774, ..., -0.3662, -0.1714, -0.1141]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0489e-08, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -9.3132e-10, ..., -1.8626e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.6077e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0145, -0.0215, -0.0091, -0.0205, 0.0046, -0.0012, 0.0088, 0.0222, + 0.0109, -0.0129], device='cuda:0'), grad: tensor([-3.4459e-08, 3.7253e-09, -7.4506e-09, 6.7987e-08, -2.3283e-08, + -7.3574e-08, 7.4506e-09, 5.5879e-09, 1.8626e-09, 5.4948e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 215.01, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4096 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.11 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.3228, 0.0880, -0.2153, ..., -0.1154, -0.2927, -0.1769], + [-0.0086, 0.1088, -0.1300, ..., -0.1489, -0.0854, 0.1353], + [ 0.0513, -0.1621, -0.1965, ..., 0.0016, -0.0618, -0.1239], + ..., + [ 0.1161, -0.1159, 0.1500, ..., 0.0770, 0.2274, -0.0274], + [ 0.1460, -0.2714, -0.1873, ..., -0.3783, -0.1265, 0.2404], + [-0.2216, 0.0945, 0.0775, ..., -0.3663, -0.1714, -0.1143]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 2.7940e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 9.3132e-10, ..., -1.8626e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -7.4506e-09, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0145, -0.0214, -0.0090, -0.0205, 0.0046, -0.0012, 0.0088, 0.0220, + 0.0109, -0.0129], device='cuda:0'), grad: tensor([ 1.2107e-08, 6.5193e-09, -1.2107e-08, 2.7940e-09, 1.4901e-08, + 9.3132e-10, -4.6566e-09, 3.7253e-09, 4.6566e-09, -1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 214.76, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4095 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.13 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.3229, 0.0881, -0.2153, ..., -0.1155, -0.2928, -0.1769], + [-0.0087, 0.1088, -0.1301, ..., -0.1491, -0.0854, 0.1354], + [ 0.0513, -0.1621, -0.1966, ..., 0.0017, -0.0618, -0.1240], + ..., + [ 0.1161, -0.1160, 0.1501, ..., 0.0770, 0.2275, -0.0274], + [ 0.1460, -0.2716, -0.1874, ..., -0.3785, -0.1267, 0.2405], + [-0.2218, 0.0945, 0.0775, ..., -0.3665, -0.1715, -0.1145]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 4.1723e-07, -5.5879e-09, 3.7346e-07, ..., 9.3132e-10, + 4.3586e-07, 2.3935e-07], + [ 2.7940e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 2.7940e-09, 2.7940e-09], + ..., + [-4.4145e-07, 4.6566e-09, -3.9954e-07, ..., 0.0000e+00, + -4.6846e-07, -2.6636e-07], + [-7.4506e-09, 4.6566e-09, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, -1.3970e-08], + [ 2.3283e-08, 6.7987e-08, 1.3039e-08, ..., 0.0000e+00, + 2.4214e-08, 1.8626e-08]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0144, -0.0214, -0.0089, -0.0206, 0.0046, -0.0011, 0.0087, 0.0220, + 0.0108, -0.0130], device='cuda:0'), grad: tensor([-4.0047e-08, 7.2923e-07, 2.1420e-08, -9.3132e-10, -1.3607e-06, + 6.5193e-09, 9.4064e-08, -7.1991e-07, 2.7940e-09, 1.2834e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 214.92, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4290 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.13 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.3229, 0.0881, -0.2155, ..., -0.1156, -0.2929, -0.1769], + [-0.0087, 0.1090, -0.1302, ..., -0.1493, -0.0855, 0.1355], + [ 0.0514, -0.1623, -0.1967, ..., 0.0018, -0.0617, -0.1240], + ..., + [ 0.1162, -0.1162, 0.1503, ..., 0.0770, 0.2276, -0.0275], + [ 0.1460, -0.2717, -0.1875, ..., -0.3785, -0.1268, 0.2405], + [-0.2220, 0.0945, 0.0775, ..., -0.3666, -0.1717, -0.1145]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.8626e-09, -0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00], + [ 1.0245e-08, -4.6566e-09, 0.0000e+00, ..., 9.3132e-09, + 7.4506e-09, -4.2841e-08], + [-1.3039e-07, 0.0000e+00, -0.0000e+00, ..., -9.1270e-08, + -1.0058e-07, 0.0000e+00], + ..., + [ 1.1455e-07, 9.3132e-10, 9.3132e-10, ..., 7.5437e-08, + 8.9407e-08, 1.8626e-09], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0144, -0.0214, -0.0088, -0.0207, 0.0046, -0.0011, 0.0087, 0.0221, + 0.0107, -0.0131], device='cuda:0'), grad: tensor([ 6.5193e-09, -4.4703e-08, -3.4273e-07, 2.7940e-09, 6.5193e-09, + -1.8626e-08, 9.4995e-08, 2.8498e-07, 8.3819e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 214.97, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4339 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.13 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.3230, 0.0882, -0.2155, ..., -0.1156, -0.2929, -0.1770], + [-0.0088, 0.1092, -0.1303, ..., -0.1494, -0.0856, 0.1356], + [ 0.0514, -0.1624, -0.1968, ..., 0.0019, -0.0617, -0.1241], + ..., + [ 0.1163, -0.1164, 0.1505, ..., 0.0771, 0.2278, -0.0276], + [ 0.1460, -0.2719, -0.1877, ..., -0.3787, -0.1268, 0.2406], + [-0.2221, 0.0945, 0.0776, ..., -0.3667, -0.1718, -0.1147]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 6.5193e-09, 2.7940e-08, 2.7940e-09, ..., 9.3132e-10, + 9.3132e-10, 3.8184e-08], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 3.0734e-08, 3.7253e-09, 1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, 3.5390e-08], + [-5.7742e-08, -9.3132e-10, -1.8626e-08, ..., -9.3132e-10, + 0.0000e+00, -6.6124e-08], + [ 9.3132e-10, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0144, -0.0213, -0.0088, -0.0207, 0.0046, -0.0011, 0.0086, 0.0220, + 0.0106, -0.0132], device='cuda:0'), grad: tensor([ 4.6566e-09, 1.0896e-07, 3.7253e-09, 3.7253e-09, -9.1270e-08, + 5.5879e-08, 1.0245e-08, 1.1642e-07, -2.0303e-07, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 214.79, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3908 re_mapping 0.0026 re_causal 0.0092 /// teacc 99.12 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.3231, 0.0882, -0.2156, ..., -0.1157, -0.2929, -0.1770], + [-0.0088, 0.1092, -0.1304, ..., -0.1495, -0.0857, 0.1357], + [ 0.0515, -0.1624, -0.1970, ..., 0.0020, -0.0616, -0.1242], + ..., + [ 0.1163, -0.1165, 0.1506, ..., 0.0770, 0.2278, -0.0276], + [ 0.1461, -0.2722, -0.1878, ..., -0.3788, -0.1269, 0.2407], + [-0.2223, 0.0945, 0.0777, ..., -0.3668, -0.1718, -0.1148]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.2498e-06, -3.8743e-07, -2.3842e-06, ..., 0.0000e+00, + -2.2724e-07, -1.4836e-06], + [ 9.3132e-10, 1.8626e-09, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 1.2275e-06, 3.8277e-07, 2.3395e-06, ..., -2.7940e-09, + 2.1979e-07, 1.4585e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 8.3819e-09, 5.5879e-09, 1.0245e-08, ..., 9.3132e-10, + 2.7940e-09, 1.1176e-08]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0144, -0.0213, -0.0087, -0.0207, 0.0047, -0.0011, 0.0086, 0.0220, + 0.0105, -0.0133], device='cuda:0'), grad: tensor([-6.5193e-09, -6.1207e-06, 1.0245e-08, 8.1956e-08, 1.2107e-08, + -3.3528e-08, 7.4506e-09, 6.0126e-06, 9.3132e-10, 4.5635e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 214.79, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4137 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.13 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.3231, 0.0883, -0.2156, ..., -0.1157, -0.2930, -0.1769], + [-0.0088, 0.1093, -0.1304, ..., -0.1497, -0.0857, 0.1358], + [ 0.0515, -0.1626, -0.1971, ..., 0.0020, -0.0616, -0.1242], + ..., + [ 0.1163, -0.1166, 0.1506, ..., 0.0770, 0.2278, -0.0277], + [ 0.1461, -0.2724, -0.1880, ..., -0.3789, -0.1270, 0.2407], + [-0.2224, 0.0945, 0.0778, ..., -0.3670, -0.1719, -0.1149]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 2.7940e-09, 5.5879e-09, ..., 3.7253e-09, + 5.5879e-09, 3.7253e-09], + [ 2.1420e-08, 0.0000e+00, 9.3132e-10, ..., -9.3132e-10, + 0.0000e+00, 3.0734e-08], + ..., + [-3.7253e-09, 9.3132e-10, -6.5193e-09, ..., -3.7253e-09, + -6.5193e-09, 5.5879e-09], + [-4.0047e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, -5.5879e-08], + [ 0.0000e+00, -2.7940e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0143, -0.0212, -0.0087, -0.0207, 0.0046, -0.0011, 0.0085, 0.0219, + 0.0104, -0.0133], device='cuda:0'), grad: tensor([ 0.0000e+00, 2.9802e-08, 2.7008e-08, 4.6566e-09, 2.7940e-09, + -5.5879e-09, 1.6764e-08, -5.5879e-09, -6.1467e-08, -6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 215.00, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4177 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.12 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.3232, 0.0884, -0.2157, ..., -0.1158, -0.2930, -0.1770], + [-0.0088, 0.1093, -0.1305, ..., -0.1500, -0.0858, 0.1358], + [ 0.0515, -0.1627, -0.1973, ..., 0.0020, -0.0616, -0.1243], + ..., + [ 0.1164, -0.1166, 0.1507, ..., 0.0771, 0.2279, -0.0277], + [ 0.1461, -0.2725, -0.1881, ..., -0.3790, -0.1270, 0.2408], + [-0.2225, 0.0946, 0.0779, ..., -0.3670, -0.1719, -0.1150]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [-6.0536e-08, 0.0000e+00, 0.0000e+00, ..., -1.8626e-08, + -2.7940e-08, -5.5879e-09], + ..., + [ 2.3283e-08, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 1.1176e-08, 2.7940e-09], + [ 2.2352e-08, 9.3132e-10, 0.0000e+00, ..., 6.5193e-09, + 9.3132e-09, 1.8626e-09], + [ 1.3039e-08, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, + 5.5879e-09, 9.3132e-10]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0143, -0.0212, -0.0087, -0.0206, 0.0045, -0.0009, 0.0084, 0.0220, + 0.0102, -0.0133], device='cuda:0'), grad: tensor([ 2.5146e-08, 5.5879e-09, -1.3504e-07, -4.6566e-09, -6.2399e-08, + -1.7695e-08, 2.0489e-08, 5.8673e-08, 5.8673e-08, 4.8429e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 214.78, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4061 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.12 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.3233, 0.0884, -0.2157, ..., -0.1158, -0.2931, -0.1770], + [-0.0090, 0.1094, -0.1308, ..., -0.1501, -0.0861, 0.1357], + [ 0.0515, -0.1628, -0.1974, ..., 0.0021, -0.0616, -0.1244], + ..., + [ 0.1166, -0.1167, 0.1510, ..., 0.0771, 0.2283, -0.0276], + [ 0.1460, -0.2726, -0.1882, ..., -0.3791, -0.1271, 0.2408], + [-0.2227, 0.0947, 0.0780, ..., -0.3671, -0.1721, -0.1151]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 2.7940e-09, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 4.6566e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [-8.3819e-09, -4.6566e-09, -1.7695e-08, ..., -1.1176e-08, + -8.3819e-09, 9.3132e-10], + [-0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 2.7940e-09, -3.7253e-09, -1.8626e-09, ..., 3.7253e-09, + 2.7940e-09, 9.3132e-10]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0143, -0.0214, -0.0086, -0.0206, 0.0044, -0.0010, 0.0085, 0.0222, + 0.0101, -0.0132], device='cuda:0'), grad: tensor([ 8.3819e-09, 8.3819e-09, 1.8626e-09, 1.8626e-09, 1.9558e-08, + -1.6764e-08, 1.3970e-08, -2.7940e-08, 3.7253e-09, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 214.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4221 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.10 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.3233, 0.0885, -0.2158, ..., -0.1159, -0.2931, -0.1770], + [-0.0090, 0.1095, -0.1308, ..., -0.1502, -0.0861, 0.1359], + [ 0.0515, -0.1630, -0.1975, ..., 0.0021, -0.0616, -0.1245], + ..., + [ 0.1166, -0.1169, 0.1511, ..., 0.0771, 0.2283, -0.0277], + [ 0.1460, -0.2728, -0.1885, ..., -0.3792, -0.1272, 0.2408], + [-0.2228, 0.0947, 0.0781, ..., -0.3672, -0.1722, -0.1152]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 0.0000e+00, -1.6764e-08], + [ 4.6566e-10, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09], + ..., + [ 4.6566e-10, 1.4901e-08, 1.6298e-08, ..., 0.0000e+00, + 0.0000e+00, 1.0710e-08], + [-1.3690e-07, 6.0536e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -1.4715e-07], + [ 0.0000e+00, -6.9849e-08, -8.7544e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0142, -0.0214, -0.0086, -0.0205, 0.0043, -0.0010, 0.0085, 0.0221, + 0.0100, -0.0133], device='cuda:0'), grad: tensor([ 7.9162e-09, -1.2107e-08, 7.4506e-09, 4.3772e-08, 6.5658e-08, + 2.7893e-07, 1.3504e-07, 7.4506e-08, -3.1665e-07, -2.6124e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 214.71, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4298 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.07 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.3234, 0.0886, -0.2159, ..., -0.1160, -0.2931, -0.1771], + [-0.0091, 0.1095, -0.1308, ..., -0.1505, -0.0862, 0.1360], + [ 0.0515, -0.1631, -0.1977, ..., 0.0021, -0.0616, -0.1245], + ..., + [ 0.1167, -0.1170, 0.1512, ..., 0.0771, 0.2284, -0.0278], + [ 0.1461, -0.2729, -0.1886, ..., -0.3793, -0.1272, 0.2409], + [-0.2230, 0.0948, 0.0782, ..., -0.3673, -0.1723, -0.1155]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 2.6543e-08, 1.8626e-09, ..., 1.8626e-09, + 1.3970e-09, 9.3132e-10], + [-1.3039e-08, 0.0000e+00, 5.1223e-09, ..., -6.0536e-09, + 4.1910e-09, 0.0000e+00], + ..., + [-3.7253e-09, 1.3970e-09, -8.8476e-09, ..., -2.3283e-09, + -7.4506e-09, 0.0000e+00], + [ 2.3283e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.3504e-08, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0142, -0.0214, -0.0086, -0.0206, 0.0044, -0.0011, 0.0086, 0.0221, + 0.0100, -0.0133], device='cuda:0'), grad: tensor([ 6.5193e-09, 7.9628e-08, -3.9116e-08, 2.2352e-08, -5.2620e-08, + -7.4506e-09, 2.2817e-08, 5.1223e-09, 7.4506e-09, -3.6322e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 214.53, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4032 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.3234, 0.0886, -0.2159, ..., -0.1160, -0.2932, -0.1771], + [-0.0091, 0.1096, -0.1308, ..., -0.1507, -0.0862, 0.1361], + [ 0.0515, -0.1632, -0.1978, ..., 0.0022, -0.0615, -0.1245], + ..., + [ 0.1167, -0.1171, 0.1512, ..., 0.0772, 0.2284, -0.0279], + [ 0.1463, -0.2730, -0.1887, ..., -0.3794, -0.1273, 0.2412], + [-0.2232, 0.0948, 0.0781, ..., -0.3674, -0.1725, -0.1156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [-9.3132e-09, 9.3132e-10, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.9558e-08], + [ 5.1223e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0142, -0.0213, -0.0086, -0.0206, 0.0045, -0.0011, 0.0086, 0.0220, + 0.0101, -0.0134], device='cuda:0'), grad: tensor([-2.0955e-08, 6.9849e-09, 4.6566e-09, 1.3970e-08, -1.8626e-09, + 3.7253e-09, 1.0245e-08, 6.0536e-09, -3.3993e-08, 2.5611e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 214.70, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4503 re_mapping 0.0026 re_causal 0.0097 /// teacc 99.13 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.3234, 0.0884, -0.2160, ..., -0.1160, -0.2932, -0.1771], + [-0.0090, 0.1097, -0.1308, ..., -0.1508, -0.0862, 0.1362], + [ 0.0515, -0.1633, -0.1979, ..., 0.0022, -0.0616, -0.1246], + ..., + [ 0.1167, -0.1173, 0.1512, ..., 0.0772, 0.2285, -0.0280], + [ 0.1464, -0.2732, -0.1888, ..., -0.3794, -0.1273, 0.2413], + [-0.2233, 0.0951, 0.0782, ..., -0.3675, -0.1725, -0.1157]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-1.9092e-08, -5.1223e-09, -3.3528e-08, ..., 2.3283e-09, + -1.2107e-08, -5.9605e-08], + [-1.0710e-08, 9.3132e-10, -3.2596e-09, ..., -1.3504e-08, + -7.9162e-09, 2.3283e-09], + ..., + [ 2.7474e-08, 1.6764e-08, 3.1199e-08, ..., 1.2573e-08, + 1.6764e-08, 6.6590e-08], + [ 2.3283e-09, 1.4435e-08, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, -1.3970e-09], + [ 5.1223e-09, 3.7253e-09, 7.4506e-09, ..., 4.6566e-10, + 3.2596e-09, 7.9162e-09]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0144, -0.0212, -0.0086, -0.0207, 0.0045, -0.0011, 0.0086, 0.0219, + 0.0101, -0.0132], device='cuda:0'), grad: tensor([-5.5879e-09, -1.3178e-07, -2.6077e-08, -4.6566e-10, -3.7253e-08, + -3.9767e-07, 2.7101e-07, 2.0117e-07, 1.1176e-07, 3.2596e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 214.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4232 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.10 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.3235, 0.0885, -0.2160, ..., -0.1160, -0.2932, -0.1771], + [-0.0091, 0.1097, -0.1309, ..., -0.1511, -0.0863, 0.1362], + [ 0.0515, -0.1634, -0.1981, ..., 0.0022, -0.0616, -0.1247], + ..., + [ 0.1168, -0.1173, 0.1513, ..., 0.0773, 0.2286, -0.0280], + [ 0.1465, -0.2733, -0.1889, ..., -0.3795, -0.1274, 0.2415], + [-0.2235, 0.0951, 0.0782, ..., -0.3676, -0.1727, -0.1157]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0431e-07, 1.8626e-09, ..., 4.6566e-10, + 0.0000e+00, -8.7544e-08], + [ 2.3283e-09, 8.8476e-09, 7.4506e-09, ..., 5.1223e-09, + 7.4506e-09, 1.8626e-09], + [-3.2596e-09, 5.1223e-09, 5.1223e-09, ..., -4.1910e-09, + -1.0710e-08, 0.0000e+00], + ..., + [-4.6566e-10, 2.7940e-09, 6.5193e-09, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + [-9.3132e-10, 2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 4.6566e-10, -2.4680e-08, -3.2596e-08, ..., 0.0000e+00, + -1.8626e-09, 3.7253e-09]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0144, -0.0213, -0.0085, -0.0208, 0.0046, -0.0012, 0.0086, 0.0220, + 0.0100, -0.0133], device='cuda:0'), grad: tensor([-3.8836e-07, 4.5635e-08, -1.4435e-08, 6.0536e-09, 8.8476e-09, + 6.6124e-08, 3.2037e-07, 2.1886e-08, 4.1910e-09, -6.6124e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 214.74, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4124 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.11 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.3235, 0.0886, -0.2161, ..., -0.1161, -0.2933, -0.1771], + [-0.0094, 0.1098, -0.1313, ..., -0.1512, -0.0867, 0.1360], + [ 0.0515, -0.1635, -0.1983, ..., 0.0023, -0.0616, -0.1248], + ..., + [ 0.1171, -0.1174, 0.1518, ..., 0.0773, 0.2290, -0.0278], + [ 0.1466, -0.2734, -0.1889, ..., -0.3796, -0.1274, 0.2416], + [-0.2237, 0.0950, 0.0781, ..., -0.3677, -0.1730, -0.1160]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 1.3970e-09, 1.3970e-09], + [ 1.7136e-07, -1.3970e-09, 3.5856e-08, ..., 1.2061e-07, + 1.5367e-07, 1.8813e-07], + [-2.1141e-07, 0.0000e+00, -3.5390e-08, ..., -1.5460e-07, + -1.8254e-07, -2.5844e-07], + ..., + [ 2.4680e-08, 2.3283e-09, -3.2596e-09, ..., 2.2817e-08, + 1.4435e-08, 5.8208e-08], + [ 2.3283e-09, 4.6566e-10, 4.6566e-10, ..., 1.8626e-09, + 1.8626e-09, 2.3283e-09], + [ 7.4506e-09, 2.3283e-09, -0.0000e+00, ..., 5.1223e-09, + 6.5193e-09, 8.3819e-09]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0143, -0.0215, -0.0086, -0.0208, 0.0045, -0.0012, 0.0086, 0.0223, + 0.0101, -0.0134], device='cuda:0'), grad: tensor([ 7.4506e-09, 6.2538e-07, -8.1724e-07, 6.5193e-09, -4.6566e-10, + -1.4901e-08, 9.3132e-09, 1.4110e-07, 1.0245e-08, 3.2131e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 215.28, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4284 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.09 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.3236, 0.0887, -0.2161, ..., -0.1161, -0.2933, -0.1771], + [-0.0094, 0.1099, -0.1312, ..., -0.1513, -0.0867, 0.1361], + [ 0.0516, -0.1636, -0.1984, ..., 0.0024, -0.0615, -0.1248], + ..., + [ 0.1171, -0.1175, 0.1518, ..., 0.0773, 0.2290, -0.0278], + [ 0.1467, -0.2736, -0.1890, ..., -0.3797, -0.1275, 0.2418], + [-0.2239, 0.0950, 0.0780, ..., -0.3678, -0.1731, -0.1161]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.9849e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.5367e-08, -1.3970e-09, 1.9558e-08, ..., 0.0000e+00, + 1.5367e-08, -3.2596e-09], + [ 4.6566e-10, 4.6566e-10, 9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [-1.9092e-08, 1.8626e-09, -2.4214e-08, ..., 0.0000e+00, + -2.0023e-08, 2.3283e-09], + [ 4.6566e-10, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 3.7253e-09, 2.3283e-09, 2.3283e-09, ..., 0.0000e+00, + 3.2596e-09, 9.3132e-10]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0142, -0.0214, -0.0084, -0.0208, 0.0044, -0.0012, 0.0086, 0.0222, + 0.0101, -0.0135], device='cuda:0'), grad: tensor([-2.9011e-07, 2.7008e-08, 4.1910e-09, 1.0245e-08, 3.2596e-09, + 4.1444e-08, 2.2026e-07, -3.2596e-08, 5.5879e-09, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 215.10, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4150 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.12 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.3236, 0.0888, -0.2162, ..., -0.1161, -0.2934, -0.1772], + [-0.0094, 0.1098, -0.1313, ..., -0.1515, -0.0868, 0.1362], + [ 0.0515, -0.1638, -0.1987, ..., 0.0023, -0.0616, -0.1249], + ..., + [ 0.1172, -0.1175, 0.1519, ..., 0.0775, 0.2291, -0.0279], + [ 0.1468, -0.2738, -0.1891, ..., -0.3798, -0.1275, 0.2420], + [-0.2241, 0.0951, 0.0782, ..., -0.3679, -0.1732, -0.1163]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-09, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, -0.0000e+00], + [ 4.6566e-09, 4.6566e-10, 1.3970e-09, ..., 2.7940e-09, + 3.7253e-09, 0.0000e+00], + [-4.3306e-08, 0.0000e+00, 0.0000e+00, ..., -2.5146e-08, + -3.3993e-08, 0.0000e+00], + ..., + [ 3.7719e-08, 9.3132e-10, 9.3132e-10, ..., 2.1886e-08, + 2.9802e-08, 4.6566e-10], + [ 9.3132e-10, 1.3970e-09, 1.3970e-09, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + [ 9.3132e-10, 2.3283e-09, -2.7940e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0142, -0.0215, -0.0086, -0.0210, 0.0043, -0.0012, 0.0087, 0.0223, + 0.0101, -0.0136], device='cuda:0'), grad: tensor([-1.9092e-08, 1.3504e-08, -1.0431e-07, 6.9849e-09, 4.6566e-09, + -1.0245e-08, 1.2107e-08, 9.4529e-08, 8.8476e-09, 9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 214.97, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4210 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.09 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.3237, 0.0888, -0.2163, ..., -0.1161, -0.2934, -0.1772], + [-0.0094, 0.1098, -0.1313, ..., -0.1516, -0.0868, 0.1362], + [ 0.0514, -0.1639, -0.1990, ..., 0.0023, -0.0617, -0.1250], + ..., + [ 0.1172, -0.1175, 0.1520, ..., 0.0775, 0.2292, -0.0279], + [ 0.1469, -0.2739, -0.1893, ..., -0.3798, -0.1276, 0.2421], + [-0.2242, 0.0952, 0.0783, ..., -0.3680, -0.1733, -0.1164]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 8.1025e-08, 1.2806e-07, ..., 7.1246e-08, + 3.7253e-09, 1.3970e-09], + [ 4.6566e-10, 3.2596e-09, 5.1223e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + ..., + [-2.7940e-09, 1.4901e-08, 1.0245e-08, ..., 3.7253e-09, + -4.1910e-09, 1.2107e-08], + [ 0.0000e+00, 2.3283e-09, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 1.8626e-09, 3.7253e-08, 4.9360e-08, ..., 3.2596e-08, + 1.8626e-09, 1.3970e-09]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0142, -0.0214, -0.0086, -0.0211, 0.0043, -0.0013, 0.0086, 0.0223, + 0.0101, -0.0135], device='cuda:0'), grad: tensor([-4.6566e-09, 2.3842e-07, 1.0245e-08, -2.6869e-07, -2.5611e-08, + -9.2201e-08, -2.3283e-09, 4.1444e-08, 7.9162e-09, 1.0803e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 214.86, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4149 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.07 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.3238, 0.0889, -0.2164, ..., -0.1162, -0.2934, -0.1772], + [-0.0095, 0.1098, -0.1314, ..., -0.1518, -0.0868, 0.1363], + [ 0.0514, -0.1640, -0.1992, ..., 0.0023, -0.0617, -0.1251], + ..., + [ 0.1173, -0.1176, 0.1521, ..., 0.0776, 0.2293, -0.0279], + [ 0.1470, -0.2741, -0.1894, ..., -0.3799, -0.1276, 0.2422], + [-0.2243, 0.0952, 0.0783, ..., -0.3681, -0.1734, -0.1165]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-09, -2.3283e-09, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, -3.7253e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, 2.7940e-09], + ..., + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 2.3283e-09], + [-1.2107e-08, 4.6566e-10, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -2.5146e-08], + [ 6.5193e-09, 9.3132e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3504e-08]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0141, -0.0215, -0.0087, -0.0211, 0.0043, -0.0013, 0.0086, 0.0223, + 0.0100, -0.0135], device='cuda:0'), grad: tensor([ 2.7940e-09, -2.3283e-09, 2.3283e-09, 6.9849e-09, 2.3283e-09, + -2.5146e-08, 1.9092e-08, 1.0710e-08, -4.4703e-08, 2.8405e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 214.88, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4163 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.08 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.3238, 0.0890, -0.2164, ..., -0.1162, -0.2935, -0.1772], + [-0.0095, 0.1099, -0.1314, ..., -0.1521, -0.0869, 0.1364], + [ 0.0514, -0.1642, -0.1993, ..., 0.0023, -0.0618, -0.1251], + ..., + [ 0.1174, -0.1178, 0.1521, ..., 0.0777, 0.2294, -0.0281], + [ 0.1472, -0.2741, -0.1895, ..., -0.3800, -0.1277, 0.2424], + [-0.2244, 0.0952, 0.0784, ..., -0.3682, -0.1735, -0.1167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, -1.8626e-09, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.0245e-08, -1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0140, -0.0215, -0.0087, -0.0212, 0.0042, -0.0013, 0.0085, 0.0223, + 0.0101, -0.0136], device='cuda:0'), grad: tensor([ 3.7253e-09, -8.8476e-09, 9.3132e-10, 0.0000e+00, 2.6543e-08, + 2.7940e-09, 1.8626e-09, 5.1223e-09, 9.3132e-10, -2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 214.86, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4189 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.3239, 0.0891, -0.2164, ..., -0.1163, -0.2935, -0.1772], + [-0.0096, 0.1098, -0.1315, ..., -0.1522, -0.0870, 0.1365], + [ 0.0514, -0.1643, -0.1994, ..., 0.0023, -0.0618, -0.1252], + ..., + [ 0.1175, -0.1178, 0.1522, ..., 0.0778, 0.2295, -0.0281], + [ 0.1473, -0.2743, -0.1896, ..., -0.3800, -0.1277, 0.2425], + [-0.2245, 0.0953, 0.0786, ..., -0.3683, -0.1735, -0.1168]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -2.7940e-09, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3598e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + ..., + [ 0.0000e+00, 4.1910e-09, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 7.9162e-08], + [-5.5879e-09, 4.6566e-10, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -1.0710e-08], + [ 4.6566e-09, -7.9162e-09, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 1.0245e-08]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0140, -0.0215, -0.0087, -0.0211, 0.0041, -0.0015, 0.0086, 0.0223, + 0.0100, -0.0135], device='cuda:0'), grad: tensor([ 1.3970e-09, -1.2992e-07, 6.9849e-09, 1.3970e-09, 3.2131e-08, + 1.8626e-09, 2.7940e-09, 1.1502e-07, -2.5146e-08, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 214.56, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4229 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.10 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.3239, 0.0892, -0.2165, ..., -0.1163, -0.2935, -0.1772], + [-0.0095, 0.1099, -0.1314, ..., -0.1523, -0.0870, 0.1366], + [ 0.0513, -0.1644, -0.1996, ..., 0.0023, -0.0618, -0.1253], + ..., + [ 0.1175, -0.1180, 0.1522, ..., 0.0778, 0.2295, -0.0282], + [ 0.1474, -0.2745, -0.1898, ..., -0.3801, -0.1278, 0.2426], + [-0.2246, 0.0952, 0.0787, ..., -0.3684, -0.1736, -0.1169]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 2.2352e-08, 4.6566e-10, 6.5193e-09, ..., 9.7789e-09, + 1.2107e-08, 4.6566e-10], + [ 5.5972e-07, 1.8626e-09, 1.8487e-07, ..., 2.4308e-07, + 3.0361e-07, 4.6566e-09], + ..., + [-6.3283e-07, 0.0000e+00, -1.9977e-07, ..., -2.7195e-07, + -3.4133e-07, -6.9849e-09], + [ 2.0955e-08, 2.7940e-09, 2.7940e-09, ..., 7.9162e-09, + 1.0245e-08, 1.3970e-09], + [ 9.7789e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + 4.1910e-09, 9.3132e-10]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0139, -0.0214, -0.0087, -0.0209, 0.0040, -0.0016, 0.0087, 0.0223, + 0.0100, -0.0135], device='cuda:0'), grad: tensor([ 1.7695e-08, 3.5390e-08, 6.8359e-07, 2.8871e-08, 4.4703e-08, + -9.9186e-08, 0.0000e+00, -7.9861e-07, 6.9849e-08, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 214.77, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4213 re_mapping 0.0025 re_causal 0.0091 /// teacc 99.10 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.3239, 0.0892, -0.2166, ..., -0.1164, -0.2935, -0.1771], + [-0.0096, 0.1099, -0.1315, ..., -0.1525, -0.0870, 0.1367], + [ 0.0512, -0.1645, -0.1999, ..., 0.0023, -0.0620, -0.1254], + ..., + [ 0.1176, -0.1181, 0.1523, ..., 0.0780, 0.2296, -0.0283], + [ 0.1474, -0.2746, -0.1899, ..., -0.3802, -0.1278, 0.2428], + [-0.2248, 0.0954, 0.0788, ..., -0.3685, -0.1737, -0.1170]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, -4.6566e-10, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, -1.8626e-09], + [-9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-9.3132e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 4.6566e-10, -4.6566e-09, -7.9162e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0139, -0.0215, -0.0089, -0.0210, 0.0038, -0.0016, 0.0087, 0.0223, + 0.0100, -0.0135], device='cuda:0'), grad: tensor([ 4.6566e-10, -2.7940e-09, -2.7940e-09, -4.1910e-09, 1.4901e-08, + 4.6566e-09, 0.0000e+00, 7.4506e-09, 0.0000e+00, -1.2573e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 214.80, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4248 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.09 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.3240, 0.0892, -0.2166, ..., -0.1164, -0.2936, -0.1771], + [-0.0097, 0.1098, -0.1316, ..., -0.1528, -0.0871, 0.1367], + [ 0.0512, -0.1647, -0.2000, ..., 0.0023, -0.0620, -0.1254], + ..., + [ 0.1177, -0.1181, 0.1525, ..., 0.0781, 0.2298, -0.0283], + [ 0.1476, -0.2747, -0.1900, ..., -0.3803, -0.1279, 0.2430], + [-0.2249, 0.0955, 0.0789, ..., -0.3686, -0.1738, -0.1171]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.7695e-08, 2.3283e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 8.8476e-09, 1.3970e-09, 1.0245e-08, ..., 6.0536e-09, + 6.5193e-09, 1.3970e-09], + [ 5.5879e-09, 9.3132e-10, 1.2573e-08, ..., 1.2107e-08, + 2.3283e-09, 9.3132e-10], + ..., + [-9.7789e-09, 9.3132e-10, 4.6566e-10, ..., 9.3132e-09, + -1.0245e-08, -1.8626e-09], + [ 2.3283e-09, 1.3970e-09, 5.1223e-09, ..., 4.1910e-09, + 1.3970e-09, 4.6566e-10], + [ 4.6566e-10, 9.7789e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0139, -0.0216, -0.0088, -0.0210, 0.0037, -0.0016, 0.0087, 0.0224, + 0.0100, -0.0133], device='cuda:0'), grad: tensor([-3.5856e-08, 2.9802e-08, 4.3306e-08, -7.1712e-08, 0.0000e+00, + -3.5390e-08, 1.0245e-08, 1.9558e-08, 1.8161e-08, 2.5611e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 214.81, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3965 re_mapping 0.0026 re_causal 0.0091 /// teacc 99.08 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.3242, 0.0893, -0.2167, ..., -0.1165, -0.2937, -0.1770], + [-0.0099, 0.1098, -0.1319, ..., -0.1531, -0.0874, 0.1366], + [ 0.0512, -0.1648, -0.2002, ..., 0.0023, -0.0620, -0.1253], + ..., + [ 0.1180, -0.1181, 0.1528, ..., 0.0783, 0.2300, -0.0282], + [ 0.1476, -0.2749, -0.1901, ..., -0.3804, -0.1280, 0.2430], + [-0.2251, 0.0956, 0.0789, ..., -0.3688, -0.1740, -0.1173]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [-1.1642e-08, -3.6322e-08, 5.5879e-09, ..., 3.2596e-09, + 7.9162e-09, -7.8231e-08], + [-9.3132e-10, 9.3132e-10, 4.6566e-10, ..., -2.3283e-09, + -9.3132e-10, 2.3283e-09], + ..., + [ 7.9162e-09, 2.8871e-08, -7.4506e-09, ..., -2.7940e-09, + -8.8476e-09, 6.3330e-08], + [-4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + [ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-09]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0138, -0.0218, -0.0087, -0.0211, 0.0036, -0.0016, 0.0086, 0.0226, + 0.0099, -0.0134], device='cuda:0'), grad: tensor([ 4.6566e-10, -1.8906e-07, -3.7253e-09, 9.3132e-10, 4.6566e-09, + -2.6543e-08, 3.4925e-08, 1.5181e-07, 7.9162e-09, 2.1886e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 214.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4403 re_mapping 0.0025 re_causal 0.0096 /// teacc 99.10 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.3242, 0.0894, -0.2167, ..., -0.1165, -0.2938, -0.1770], + [-0.0100, 0.1099, -0.1319, ..., -0.1533, -0.0874, 0.1367], + [ 0.0511, -0.1649, -0.2006, ..., 0.0023, -0.0621, -0.1255], + ..., + [ 0.1181, -0.1182, 0.1529, ..., 0.0785, 0.2302, -0.0283], + [ 0.1477, -0.2750, -0.1902, ..., -0.3805, -0.1280, 0.2432], + [-0.2254, 0.0956, 0.0790, ..., -0.3689, -0.1741, -0.1175]], + device='cuda:0'), grad: tensor([[-2.7940e-09, -1.1176e-08, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, -3.2596e-09], + [ 3.2596e-09, 3.2596e-09, 1.3970e-09, ..., 2.3283e-09, + 9.3132e-10, 4.6566e-09], + [-9.3132e-10, 4.6566e-10, 0.0000e+00, ..., -3.7253e-09, + -1.3970e-09, 2.7940e-09], + ..., + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 1.3970e-09, + 4.6566e-10, 1.3970e-09], + [-7.9162e-09, 1.8626e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, -1.2573e-08], + [ 4.1910e-09, -3.2596e-09, -1.4901e-08, ..., 4.6566e-10, + 0.0000e+00, 1.3504e-08]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0138, -0.0218, -0.0089, -0.0211, 0.0034, -0.0016, 0.0087, 0.0227, + 0.0099, -0.0135], device='cuda:0'), grad: tensor([-2.4680e-08, 2.0955e-08, -9.3132e-09, 4.6566e-10, 1.2107e-08, + 1.8626e-09, 4.1910e-09, 1.0710e-08, -1.6298e-08, 8.8476e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 214.93, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4227 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.05 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0100, -0.0114, -0.0008, ..., -0.0300, 0.0253, 0.0120], + [ 0.0002, 0.0172, -0.0180, ..., 0.0283, 0.0296, -0.0199], + [ 0.0140, 0.0142, 0.0171, ..., 0.0041, 0.0127, 0.0219], + ..., + [-0.0155, -0.0106, 0.0301, ..., -0.0249, -0.0211, -0.0042], + [-0.0038, 0.0173, -0.0014, ..., -0.0064, -0.0125, -0.0106], + [ 0.0016, 0.0114, 0.0042, ..., -0.0289, -0.0007, 0.0232]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0146, -0.0216, -0.0005, 0.0296, -0.0112, 0.0282, 0.0049, 0.0115, + 0.0094, -0.0021], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 231.85, cls_loss 1.3938 cls_loss_mapping 1.8860 cls_loss_causal 2.2245 re_mapping 0.1345 re_causal 0.1397 /// teacc 85.45 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0093, -0.0118, -0.0030, ..., -0.0260, 0.0210, 0.0120], + [ 0.0008, 0.0176, -0.0176, ..., 0.0259, 0.0305, -0.0199], + [ 0.0131, 0.0155, 0.0158, ..., 0.0021, 0.0173, 0.0219], + ..., + [-0.0163, -0.0116, 0.0276, ..., -0.0288, -0.0281, -0.0042], + [-0.0041, 0.0172, -0.0042, ..., -0.0102, -0.0097, -0.0106], + [ 0.0007, 0.0102, 0.0018, ..., -0.0340, -0.0060, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., 0.0107, 0.0344, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0045, -0.0017, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0224, -0.0208, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0038, 0.0081, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0157, 0.0263, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0050, 0.0150, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0124, -0.0185, -0.0008, 0.0294, -0.0118, 0.0289, 0.0053, 0.0118, + 0.0082, -0.0029], device='cuda:0'), grad: tensor([ 0.0261, 0.0023, -0.0052, -0.0114, 0.0083, -0.0551, -0.0673, -0.0037, + 0.0613, 0.0446], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 231.40, cls_loss 0.4578 cls_loss_mapping 0.7867 cls_loss_causal 1.9281 re_mapping 0.2059 re_causal 0.2627 /// teacc 92.16 lr 0.00010000 +Epoch 3, weight, value: tensor([[-9.2938e-03, -1.1764e-02, -3.0247e-03, ..., -2.3329e-02, + 1.9973e-02, 1.1960e-02], + [ 8.7744e-04, 1.7649e-02, -1.7532e-02, ..., 2.5183e-02, + 2.9312e-02, -1.9872e-02], + [ 1.3047e-02, 1.5507e-02, 1.5782e-02, ..., -1.7187e-05, + 1.9124e-02, 2.1870e-02], + ..., + [-1.6276e-02, -1.1638e-02, 2.7572e-02, ..., -3.1240e-02, + -2.9613e-02, -4.2206e-03], + [-4.3214e-03, 1.7225e-02, -4.1866e-03, ..., -1.3714e-02, + -7.7291e-03, -1.0556e-02], + [ 7.1773e-04, 1.0208e-02, 1.7482e-03, ..., -3.5895e-02, + -9.1795e-03, 2.3166e-02]], device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., 0.0271, 0.0068, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., -0.0011, 0.0105, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0045, 0.0060, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0009, 0.0037, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0041, -0.0287, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0044, 0.0039, 0.0000]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0119, -0.0182, -0.0007, 0.0295, -0.0117, 0.0299, 0.0050, 0.0113, + 0.0077, -0.0028], device='cuda:0'), grad: tensor([ 0.0381, 0.0062, 0.0194, 0.0093, -0.0139, -0.0243, -0.0286, 0.0022, + -0.0200, 0.0116], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 231.11, cls_loss 0.2856 cls_loss_mapping 0.4688 cls_loss_causal 1.7267 re_mapping 0.1550 re_causal 0.2505 /// teacc 93.81 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0093, -0.0128, -0.0030, ..., -0.0220, 0.0196, 0.0120], + [ 0.0009, 0.0173, -0.0175, ..., 0.0256, 0.0289, -0.0199], + [ 0.0130, 0.0186, 0.0158, ..., -0.0018, 0.0201, 0.0219], + ..., + [-0.0163, -0.0126, 0.0276, ..., -0.0327, -0.0304, -0.0042], + [-0.0043, 0.0149, -0.0042, ..., -0.0160, -0.0058, -0.0106], + [ 0.0007, 0.0098, 0.0017, ..., -0.0372, -0.0114, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.2532e-05, 0.0000e+00, ..., 5.3024e-04, + 1.9407e-03, 0.0000e+00], + [ 0.0000e+00, 1.2681e-05, 0.0000e+00, ..., 6.7711e-04, + 1.2560e-03, 0.0000e+00], + [ 0.0000e+00, 7.0953e-03, 0.0000e+00, ..., 7.2708e-03, + 5.3497e-02, 0.0000e+00], + ..., + [ 0.0000e+00, -7.3242e-03, 0.0000e+00, ..., -4.5929e-03, + -4.4159e-02, 0.0000e+00], + [ 0.0000e+00, 2.4602e-05, 0.0000e+00, ..., 1.2825e-02, + 3.9795e-02, 0.0000e+00], + [ 0.0000e+00, 2.2147e-06, 0.0000e+00, ..., 5.0497e-04, + -1.8873e-03, 0.0000e+00]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0117, -0.0178, -0.0004, 0.0294, -0.0120, 0.0301, 0.0047, 0.0113, + 0.0080, -0.0031], device='cuda:0'), grad: tensor([ 0.0018, -0.0016, 0.0365, -0.0023, 0.0026, -0.0017, -0.0288, -0.0201, + 0.0322, -0.0188], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 231.39, cls_loss 0.2267 cls_loss_mapping 0.3438 cls_loss_causal 1.5339 re_mapping 0.1193 re_causal 0.2169 /// teacc 95.53 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0093, -0.0134, -0.0030, ..., -0.0209, 0.0196, 0.0120], + [ 0.0009, 0.0170, -0.0175, ..., 0.0259, 0.0284, -0.0199], + [ 0.0130, 0.0178, 0.0158, ..., -0.0030, 0.0208, 0.0219], + ..., + [-0.0163, -0.0112, 0.0276, ..., -0.0341, -0.0316, -0.0042], + [-0.0043, 0.0129, -0.0042, ..., -0.0181, -0.0045, -0.0106], + [ 0.0007, 0.0095, 0.0017, ..., -0.0379, -0.0128, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., -0.0019, 0.0194, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0052, 0.0103, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0043, -0.0147, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0006, 0.0041, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0022, 0.0119, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0009, -0.0220, 0.0000]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0121, -0.0173, -0.0003, 0.0291, -0.0116, 0.0298, 0.0046, 0.0111, + 0.0078, -0.0031], device='cuda:0'), grad: tensor([ 0.0193, 0.0073, -0.0134, 0.0260, -0.0079, -0.0156, -0.0226, 0.0017, + 0.0151, -0.0099], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 231.05, cls_loss 0.1704 cls_loss_mapping 0.2512 cls_loss_causal 1.4027 re_mapping 0.1000 re_causal 0.2003 /// teacc 95.83 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0093, -0.0134, -0.0030, ..., -0.0199, 0.0197, 0.0120], + [ 0.0009, 0.0170, -0.0175, ..., 0.0268, 0.0285, -0.0199], + [ 0.0130, 0.0178, 0.0158, ..., -0.0044, 0.0214, 0.0219], + ..., + [-0.0163, -0.0112, 0.0276, ..., -0.0352, -0.0331, -0.0042], + [-0.0043, 0.0129, -0.0042, ..., -0.0194, -0.0029, -0.0106], + [ 0.0007, 0.0095, 0.0017, ..., -0.0394, -0.0143, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., -0.0091, -0.0091, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0023, 0.0042, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., -0.0006, -0.0194, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0006, 0.0017, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0075, 0.0224, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0022, 0.0031, 0.0000]], + device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0123, -0.0167, -0.0004, 0.0290, -0.0117, 0.0294, 0.0044, 0.0109, + 0.0080, -0.0031], device='cuda:0'), grad: tensor([-0.0095, 0.0058, -0.0125, -0.0062, 0.0070, -0.0328, 0.0072, 0.0090, + 0.0296, 0.0024], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 232.19, cls_loss 0.1403 cls_loss_mapping 0.2009 cls_loss_causal 1.2733 re_mapping 0.0831 re_causal 0.1784 /// teacc 96.87 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0093, -0.0136, -0.0030, ..., -0.0197, 0.0192, 0.0120], + [ 0.0009, 0.0170, -0.0175, ..., 0.0262, 0.0275, -0.0199], + [ 0.0130, 0.0179, 0.0158, ..., -0.0053, 0.0223, 0.0219], + ..., + [-0.0163, -0.0112, 0.0276, ..., -0.0362, -0.0343, -0.0042], + [-0.0042, 0.0127, -0.0042, ..., -0.0206, -0.0015, -0.0106], + [ 0.0007, 0.0094, 0.0017, ..., -0.0400, -0.0157, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., 0.0017, -0.0036, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0003, 0.0010, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0015, 0.0025, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0009, 0.0027, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0030, -0.0009, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0012, 0.0028, 0.0000]], + device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0123, -0.0168, -0.0003, 0.0288, -0.0116, 0.0292, 0.0043, 0.0110, + 0.0082, -0.0030], device='cuda:0'), grad: tensor([-0.0120, 0.0004, 0.0056, -0.0004, -0.0185, 0.0036, 0.0042, 0.0066, + 0.0020, 0.0084], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 230.80, cls_loss 0.1203 cls_loss_mapping 0.1726 cls_loss_causal 1.2307 re_mapping 0.0705 re_causal 0.1611 /// teacc 97.13 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0096, -0.0137, -0.0030, ..., -0.0192, 0.0188, 0.0120], + [ 0.0008, 0.0168, -0.0175, ..., 0.0262, 0.0267, -0.0199], + [ 0.0128, 0.0175, 0.0158, ..., -0.0064, 0.0228, 0.0219], + ..., + [-0.0163, -0.0113, 0.0276, ..., -0.0370, -0.0348, -0.0042], + [-0.0058, 0.0124, -0.0042, ..., -0.0213, -0.0007, -0.0106], + [ 0.0004, 0.0092, 0.0017, ..., -0.0414, -0.0169, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., 0.0011, 0.0014, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0022, 0.0009, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0007, -0.0015, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0002, 0.0015, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0027, 0.0021, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0005, 0.0008, 0.0000]], + device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0124, -0.0168, -0.0003, 0.0288, -0.0113, 0.0291, 0.0041, 0.0110, + 0.0081, -0.0030], device='cuda:0'), grad: tensor([ 0.0021, 0.0057, 0.0009, 0.0005, -0.0181, -0.0112, 0.0013, 0.0044, + 0.0089, 0.0056], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 230.55, cls_loss 0.1114 cls_loss_mapping 0.1632 cls_loss_causal 1.2534 re_mapping 0.0612 re_causal 0.1490 /// teacc 97.19 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0149, -0.0137, -0.0030, ..., -0.0183, 0.0184, 0.0170], + [ 0.0049, 0.0168, -0.0175, ..., 0.0262, 0.0259, -0.0219], + [ 0.0073, 0.0175, 0.0158, ..., -0.0075, 0.0233, 0.0159], + ..., + [-0.0237, -0.0113, 0.0276, ..., -0.0374, -0.0353, -0.0039], + [-0.0072, 0.0124, -0.0042, ..., -0.0228, 0.0006, -0.0187], + [-0.0051, 0.0092, 0.0017, ..., -0.0423, -0.0177, 0.0145]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., 0.0011, 0.0022, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0055, 0.0025, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0028, 0.0036, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0003, 0.0018, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0016, -0.0039, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0005, 0.0033, 0.0000]], + device='cuda:0') +Epoch 9, bias, value: tensor([ 1.2328e-02, -1.6829e-02, -5.0162e-05, 2.8744e-02, -1.1248e-02, + 2.8744e-02, 4.0316e-03, 1.1183e-02, 8.4184e-03, -3.2370e-03], + device='cuda:0'), grad: tensor([ 0.0042, -0.0015, 0.0082, 0.0193, -0.0048, -0.0291, -0.0100, 0.0054, + -0.0002, 0.0085], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 232.00, cls_loss 0.1004 cls_loss_mapping 0.1419 cls_loss_causal 1.1903 re_mapping 0.0560 re_causal 0.1379 /// teacc 97.23 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0149, -0.0137, -0.0030, ..., -0.0175, 0.0180, 0.0174], + [ 0.0049, 0.0168, -0.0175, ..., 0.0260, 0.0251, -0.0219], + [ 0.0073, 0.0175, 0.0158, ..., -0.0083, 0.0239, 0.0155], + ..., + [-0.0237, -0.0113, 0.0276, ..., -0.0383, -0.0365, -0.0040], + [-0.0072, 0.0124, -0.0042, ..., -0.0232, 0.0016, -0.0194], + [-0.0051, 0.0092, 0.0017, ..., -0.0439, -0.0188, 0.0140]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0000, 0.0000, ..., 0.0096, 0.0079, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., -0.0002, 0.0002, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0007, -0.0005, 0.0000], + ..., + [ 0.0000, 0.0000, 0.0000, ..., 0.0004, 0.0006, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., 0.0006, -0.0003, 0.0000], + [ 0.0000, 0.0000, 0.0000, ..., -0.0100, -0.0075, 0.0000]], + device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0123, -0.0170, 0.0004, 0.0291, -0.0111, 0.0287, 0.0038, 0.0108, + 0.0085, -0.0033], device='cuda:0'), grad: tensor([ 0.0214, -0.0003, -0.0020, 0.0076, 0.0012, -0.0104, 0.0014, 0.0019, + 0.0011, -0.0219], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 230.65, cls_loss 0.0951 cls_loss_mapping 0.1368 cls_loss_causal 1.1031 re_mapping 0.0531 re_causal 0.1253 /// teacc 97.40 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0150, -0.0140, -0.0030, ..., -0.0168, 0.0176, 0.0131], + [ 0.0049, 0.0166, -0.0175, ..., 0.0265, 0.0244, -0.0321], + [ 0.0073, 0.0177, 0.0158, ..., -0.0095, 0.0243, 0.0079], + ..., + [-0.0237, -0.0113, 0.0276, ..., -0.0391, -0.0372, 0.0032], + [-0.0072, 0.0123, -0.0042, ..., -0.0241, 0.0027, -0.0333], + [-0.0051, 0.0092, 0.0017, ..., -0.0450, -0.0197, 0.0091]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8158e-03, + -7.8964e-04, -3.3259e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.2731e-05, + 2.9397e-04, 8.9929e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.5735e-04, + 4.5586e-04, 6.1631e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6463e-04, + 8.9347e-05, 7.0147e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2043e-03, + 2.1801e-03, 2.3991e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.0640e-04, + 3.2663e-04, 3.4243e-05]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0121, -0.0168, 0.0003, 0.0294, -0.0109, 0.0285, 0.0034, 0.0108, + 0.0088, -0.0036], device='cuda:0'), grad: tensor([-0.0008, 0.0261, -0.0030, 0.0094, 0.0074, 0.0049, -0.0079, -0.0114, + 0.0078, -0.0325], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 230.64, cls_loss 0.0825 cls_loss_mapping 0.1230 cls_loss_causal 1.0997 re_mapping 0.0469 re_causal 0.1231 /// teacc 97.57 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0152, -0.0143, -0.0030, ..., -0.0165, 0.0168, 0.0107], + [ 0.0049, 0.0165, -0.0175, ..., 0.0258, 0.0235, -0.0470], + [ 0.0072, 0.0178, 0.0158, ..., -0.0103, 0.0246, 0.0043], + ..., + [-0.0225, -0.0113, 0.0276, ..., -0.0395, -0.0379, 0.0064], + [-0.0073, 0.0121, -0.0042, ..., -0.0248, 0.0034, -0.0384], + [-0.0058, 0.0091, 0.0017, ..., -0.0466, -0.0204, 0.0024]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8204e-02, + -5.3787e-03, 1.9092e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1319e-04, + 1.3006e-04, 6.3516e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7861e-04, + -6.9857e-04, 3.1712e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7202e-04, + 3.9220e-04, 7.2550e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1387e-03, + 4.3607e-04, 1.2433e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.2908e-04, + 3.3855e-04, 4.4703e-07]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0121, -0.0174, 0.0004, 0.0294, -0.0111, 0.0285, 0.0036, 0.0109, + 0.0090, -0.0036], device='cuda:0'), grad: tensor([-0.0137, 0.0009, 0.0013, 0.0034, -0.0055, 0.0028, 0.0064, 0.0062, + 0.0020, -0.0038], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 230.26, cls_loss 0.0615 cls_loss_mapping 0.0927 cls_loss_causal 1.0239 re_mapping 0.0452 re_causal 0.1190 /// teacc 97.78 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0145, -0.0146, -0.0030, ..., -0.0158, 0.0164, 0.0100], + [ 0.0051, 0.0164, -0.0175, ..., 0.0255, 0.0227, -0.0515], + [ 0.0070, 0.0180, 0.0158, ..., -0.0108, 0.0252, 0.0054], + ..., + [-0.0224, -0.0114, 0.0276, ..., -0.0398, -0.0386, 0.0070], + [-0.0074, 0.0117, -0.0042, ..., -0.0254, 0.0043, -0.0394], + [-0.0060, 0.0089, 0.0017, ..., -0.0476, -0.0213, -0.0005]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.7261e-04, + -4.3154e-04, 9.2201e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.9414e-04, + -8.1539e-04, 2.4568e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0366e-03, + -2.7485e-03, 1.7695e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.1016e-05, + -9.9659e-04, 2.3078e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7800e-04, + 3.0537e-03, 3.8603e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.6560e-05, + 2.4188e-04, 4.7721e-06]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0121, -0.0175, 0.0008, 0.0293, -0.0113, 0.0284, 0.0036, 0.0110, + 0.0093, -0.0038], device='cuda:0'), grad: tensor([-0.0006, -0.0059, 0.0042, -0.0104, -0.0002, 0.0028, 0.0028, -0.0044, + 0.0102, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 230.74, cls_loss 0.0686 cls_loss_mapping 0.1031 cls_loss_causal 1.0073 re_mapping 0.0424 re_causal 0.1082 /// teacc 97.85 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0148, -0.0146, -0.0030, ..., -0.0152, 0.0160, 0.0095], + [ 0.0059, 0.0163, -0.0175, ..., 0.0255, 0.0225, -0.0550], + [ 0.0038, 0.0181, 0.0158, ..., -0.0114, 0.0256, 0.0038], + ..., + [-0.0152, -0.0114, 0.0276, ..., -0.0403, -0.0393, 0.0106], + [-0.0099, 0.0117, -0.0042, ..., -0.0262, 0.0050, -0.0397], + [-0.0131, 0.0089, 0.0017, ..., -0.0484, -0.0215, -0.0025]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0826e-02, + -7.7782e-03, 1.4298e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6454e-06, + 6.1512e-04, 5.3346e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.4925e-04, + -3.9577e-04, 1.0669e-04], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8079e-04, + 6.5708e-04, 7.2212e-03], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.0926e-03, + 4.6844e-03, 4.2796e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.0265e-04, + 6.3801e-04, 3.1209e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0119, -0.0175, 0.0007, 0.0294, -0.0112, 0.0282, 0.0034, 0.0113, + 0.0090, -0.0036], device='cuda:0'), grad: tensor([-1.3512e-02, -1.9684e-03, 9.2602e-04, 3.2139e-03, -1.0506e-02, + 3.3170e-05, 4.4942e-04, 1.2093e-02, 6.9580e-03, 2.3136e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 230.57, cls_loss 0.0578 cls_loss_mapping 0.0852 cls_loss_causal 1.0313 re_mapping 0.0390 re_causal 0.1042 /// teacc 98.07 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0148, -0.0146, -0.0030, ..., -0.0142, 0.0158, 0.0082], + [ 0.0059, 0.0163, -0.0175, ..., 0.0256, 0.0219, -0.0609], + [ 0.0038, 0.0181, 0.0158, ..., -0.0122, 0.0260, 0.0017], + ..., + [-0.0152, -0.0114, 0.0276, ..., -0.0409, -0.0401, 0.0116], + [-0.0099, 0.0116, -0.0042, ..., -0.0270, 0.0054, -0.0407], + [-0.0132, 0.0089, 0.0017, ..., -0.0493, -0.0218, -0.0043]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4402e-04, + 4.0054e-04, 3.0470e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0452e-03, + 2.4738e-03, 1.8096e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.4169e-04, + -1.4760e-05, 6.0052e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.9778e-05, + 2.2471e-04, -8.1158e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6060e-03, + -3.9101e-03, 9.1732e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7297e-04, + 2.4331e-04, 7.6389e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0119, -0.0175, 0.0006, 0.0296, -0.0110, 0.0285, 0.0035, 0.0113, + 0.0089, -0.0039], device='cuda:0'), grad: tensor([ 0.0025, 0.0056, 0.0004, 0.0017, 0.0025, 0.0006, 0.0002, -0.0039, + -0.0037, -0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 214.62, cls_loss 0.0541 cls_loss_mapping 0.0854 cls_loss_causal 0.9871 re_mapping 0.0361 re_causal 0.1006 /// teacc 97.99 lr 0.00010000 +Epoch 16, weight, value: tensor([[-1.4800e-02, -1.4654e-02, -2.5723e-03, ..., -1.3739e-02, + 1.5383e-02, 7.1815e-03], + [ 5.8514e-03, 1.6344e-02, -1.7533e-02, ..., 2.5465e-02, + 2.1403e-02, -6.8740e-02], + [ 3.7659e-03, 1.8055e-02, 1.5764e-02, ..., -1.2773e-02, + 2.6297e-02, 1.5657e-05], + ..., + [-1.5164e-02, -1.1371e-02, 2.7566e-02, ..., -4.1242e-02, + -4.0861e-02, 1.2808e-02], + [-9.8966e-03, 1.1638e-02, -4.2019e-03, ..., -2.7897e-02, + 6.1218e-03, -4.2522e-02], + [-1.3214e-02, 8.9298e-03, 1.7273e-03, ..., -5.0079e-02, + -2.2540e-02, -7.0787e-03]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.8130e-03, + -1.4706e-03, 4.6417e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4923e-05, + 3.3569e-04, 2.9221e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.2602e-04, + 7.8058e-04, 1.1578e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9898e-04, + 4.7731e-04, -1.6183e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3590e-04, + -1.1940e-03, 2.1651e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8920e-04, + 1.8609e-04, 1.7262e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0116, -0.0173, 0.0004, 0.0295, -0.0107, 0.0280, 0.0037, 0.0114, + 0.0093, -0.0041], device='cuda:0'), grad: tensor([-8.1329e-03, -7.8201e-05, 5.3253e-03, -5.2376e-03, 8.2397e-04, + 2.2011e-03, 3.3875e-03, -6.5088e-04, 1.2894e-03, 1.0595e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 230.66, cls_loss 0.0487 cls_loss_mapping 0.0756 cls_loss_causal 0.9825 re_mapping 0.0351 re_causal 0.0983 /// teacc 98.23 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0148, -0.0147, -0.0026, ..., -0.0132, 0.0151, 0.0056], + [ 0.0059, 0.0163, -0.0175, ..., 0.0253, 0.0210, -0.0787], + [ 0.0038, 0.0181, 0.0158, ..., -0.0132, 0.0268, -0.0010], + ..., + [-0.0152, -0.0114, 0.0276, ..., -0.0416, -0.0417, 0.0141], + [-0.0099, 0.0116, -0.0042, ..., -0.0283, 0.0066, -0.0446], + [-0.0132, 0.0089, 0.0017, ..., -0.0508, -0.0226, -0.0104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.3408e-05, + 1.0014e-04, 3.0845e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.0134e-04, + -6.1214e-05, 1.3518e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3065e-04, + -4.8733e-04, 4.7565e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8433e-05, + 1.0643e-03, -5.9754e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.5882e-04, + 8.5163e-04, 1.2374e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9594e-05, + 2.7633e-04, 1.4906e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0117, -0.0176, 0.0009, 0.0301, -0.0109, 0.0279, 0.0036, 0.0113, + 0.0093, -0.0045], device='cuda:0'), grad: tensor([ 0.0003, 0.0006, -0.0201, -0.0054, -0.0043, 0.0015, -0.0002, 0.0190, + 0.0055, 0.0030], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 214.19, cls_loss 0.0481 cls_loss_mapping 0.0727 cls_loss_causal 0.9366 re_mapping 0.0328 re_causal 0.0910 /// teacc 98.08 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0153, -0.0147, -0.0026, ..., -0.0126, 0.0149, 0.0041], + [ 0.0057, 0.0163, -0.0175, ..., 0.0250, 0.0204, -0.0875], + [ 0.0032, 0.0181, 0.0158, ..., -0.0136, 0.0272, -0.0025], + ..., + [-0.0153, -0.0114, 0.0276, ..., -0.0420, -0.0422, 0.0147], + [-0.0105, 0.0116, -0.0042, ..., -0.0291, 0.0074, -0.0473], + [-0.0141, 0.0089, 0.0017, ..., -0.0510, -0.0231, -0.0134]], + device='cuda:0'), grad: tensor([[ 1.2228e-06, 0.0000e+00, 0.0000e+00, ..., 8.4114e-04, + 7.8201e-04, 8.5905e-06], + [ 2.6394e-06, 0.0000e+00, 0.0000e+00, ..., -2.0429e-05, + 1.6928e-04, 5.3167e-05], + [ 1.0274e-05, 0.0000e+00, 0.0000e+00, ..., 2.4414e-04, + -7.9155e-04, 4.0442e-05], + ..., + [ 1.0006e-05, 0.0000e+00, 0.0000e+00, ..., 8.4758e-05, + 1.9610e-04, 7.1406e-05], + [ 3.0845e-06, 0.0000e+00, 0.0000e+00, ..., 4.5776e-03, + 5.8937e-03, 5.5522e-05], + [ 4.4741e-06, 0.0000e+00, 0.0000e+00, ..., 5.0402e-04, + 6.7949e-04, -5.5462e-05]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0118, -0.0179, 0.0010, 0.0300, -0.0110, 0.0282, 0.0034, 0.0113, + 0.0093, -0.0043], device='cuda:0'), grad: tensor([ 0.0003, 0.0031, 0.0005, -0.0010, 0.0003, -0.0134, 0.0054, -0.0045, + 0.0085, 0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 214.40, cls_loss 0.0520 cls_loss_mapping 0.0774 cls_loss_causal 0.9198 re_mapping 0.0313 re_causal 0.0860 /// teacc 98.12 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0181, -0.0147, -0.0026, ..., -0.0120, 0.0143, 0.0019], + [ 0.0057, 0.0163, -0.0175, ..., 0.0252, 0.0199, -0.0981], + [ 0.0009, 0.0181, 0.0158, ..., -0.0142, 0.0278, -0.0041], + ..., + [-0.0155, -0.0114, 0.0276, ..., -0.0423, -0.0431, 0.0135], + [-0.0083, 0.0116, -0.0042, ..., -0.0297, 0.0080, -0.0503], + [-0.0158, 0.0089, 0.0017, ..., -0.0518, -0.0236, -0.0152]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2798e-03, + 6.4707e-04, 2.9281e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8919e-04, + 2.7776e-04, 3.0041e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8729e-04, + -9.0256e-03, 3.8557e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6372e-05, + 9.8705e-04, 9.8765e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1212e-04, + 3.9177e-03, 3.7074e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.3479e-05, + 4.3035e-04, -8.4019e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0114, -0.0181, 0.0014, 0.0301, -0.0108, 0.0282, 0.0033, 0.0113, + 0.0095, -0.0046], device='cuda:0'), grad: tensor([ 0.0026, 0.0003, -0.0143, 0.0010, 0.0064, 0.0028, 0.0012, 0.0022, + 0.0096, -0.0118], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 230.67, cls_loss 0.0371 cls_loss_mapping 0.0563 cls_loss_causal 0.8916 re_mapping 0.0310 re_causal 0.0903 /// teacc 98.28 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0181, -0.0147, -0.0026, ..., -0.0115, 0.0141, 0.0022], + [ 0.0059, 0.0163, -0.0175, ..., 0.0256, 0.0196, -0.1037], + [-0.0004, 0.0181, 0.0158, ..., -0.0148, 0.0281, -0.0025], + ..., + [-0.0144, -0.0114, 0.0276, ..., -0.0426, -0.0435, 0.0127], + [-0.0085, 0.0116, -0.0042, ..., -0.0302, 0.0087, -0.0530], + [-0.0168, 0.0089, 0.0017, ..., -0.0521, -0.0238, -0.0153]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8944e-04, + 5.4073e-04, 2.0042e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.2953e-06, + 8.1897e-05, 2.0519e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.2525e-05, + 2.0885e-04, 1.7583e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5661e-05, + 6.3956e-05, 1.9819e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8253e-04, + -5.5504e-04, 1.0565e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5267e-05, + 1.6761e-04, 1.5736e-05]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0117, -0.0179, 0.0013, 0.0300, -0.0109, 0.0280, 0.0031, 0.0115, + 0.0095, -0.0046], device='cuda:0'), grad: tensor([ 6.0558e-04, -4.3839e-05, 2.4462e-04, 3.6192e-04, 3.1567e-04, + 5.4312e-04, -1.0729e-03, 2.1935e-04, -6.9237e-04, -4.8065e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 230.99, cls_loss 0.0412 cls_loss_mapping 0.0590 cls_loss_causal 0.9101 re_mapping 0.0303 re_causal 0.0833 /// teacc 98.36 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0180, -0.0147, -0.0026, ..., -0.0108, 0.0138, 0.0023], + [ 0.0059, 0.0163, -0.0175, ..., 0.0256, 0.0192, -0.1072], + [-0.0013, 0.0181, 0.0158, ..., -0.0154, 0.0284, -0.0055], + ..., + [-0.0126, -0.0114, 0.0276, ..., -0.0429, -0.0437, 0.0133], + [-0.0086, 0.0116, -0.0042, ..., -0.0309, 0.0090, -0.0558], + [-0.0172, 0.0089, 0.0017, ..., -0.0527, -0.0240, -0.0173]], + device='cuda:0'), grad: tensor([[ 8.8662e-07, 0.0000e+00, 0.0000e+00, ..., -1.8525e-04, + 8.6355e-04, 1.0170e-05], + [-4.0978e-08, 0.0000e+00, 0.0000e+00, ..., -8.8692e-04, + 3.9488e-05, 2.8417e-05], + [ 4.4219e-06, 0.0000e+00, 0.0000e+00, ..., 2.1553e-04, + 6.7770e-05, 3.2037e-05], + ..., + [-2.0236e-05, 0.0000e+00, 0.0000e+00, ..., 8.9705e-05, + 5.6028e-05, 4.8220e-05], + [ 3.2261e-06, 0.0000e+00, 0.0000e+00, ..., -8.3847e-03, + -2.0218e-02, 1.1645e-05], + [ 6.2622e-06, 0.0000e+00, 0.0000e+00, ..., 4.5872e-04, + 7.5293e-04, 3.5667e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0121, -0.0179, 0.0012, 0.0302, -0.0111, 0.0281, 0.0031, 0.0117, + 0.0091, -0.0047], device='cuda:0'), grad: tensor([ 7.8726e-04, -2.0885e-03, 9.2077e-04, 1.6678e-02, 9.8038e-04, + 5.8708e-03, 9.3174e-04, 6.3658e-05, -2.6581e-02, 2.4242e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 214.38, cls_loss 0.0477 cls_loss_mapping 0.0689 cls_loss_causal 0.9220 re_mapping 0.0285 re_causal 0.0796 /// teacc 98.32 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0177, -0.0147, -0.0026, ..., -0.0108, 0.0133, 0.0011], + [ 0.0062, 0.0163, -0.0175, ..., 0.0258, 0.0187, -0.1115], + [-0.0020, 0.0181, 0.0158, ..., -0.0158, 0.0288, -0.0074], + ..., + [-0.0125, -0.0114, 0.0276, ..., -0.0432, -0.0446, 0.0142], + [-0.0086, 0.0116, -0.0042, ..., -0.0312, 0.0097, -0.0591], + [-0.0178, 0.0089, 0.0017, ..., -0.0522, -0.0246, -0.0192]], + device='cuda:0'), grad: tensor([[ 9.9000e-07, 0.0000e+00, 0.0000e+00, ..., 2.9030e-03, + 2.4738e-03, 1.5028e-05], + [-2.8126e-06, 0.0000e+00, 0.0000e+00, ..., 1.2457e-04, + 5.1022e-04, 4.9382e-05], + [ 1.6419e-06, 0.0000e+00, 0.0000e+00, ..., 7.2098e-04, + 2.2221e-03, 9.8586e-05], + ..., + [-6.4857e-06, 0.0000e+00, 0.0000e+00, ..., 7.7128e-05, + 6.8426e-04, 7.7772e-04], + [ 9.9186e-07, 0.0000e+00, 0.0000e+00, ..., 1.5700e-04, + -7.1526e-03, 2.1785e-05], + [ 2.4047e-06, 0.0000e+00, 0.0000e+00, ..., 5.9891e-04, + 1.4353e-03, 3.0327e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0113, -0.0178, 0.0014, 0.0302, -0.0112, 0.0282, 0.0031, 0.0116, + 0.0094, -0.0046], device='cuda:0'), grad: tensor([ 0.0087, 0.0009, 0.0059, -0.0170, -0.0002, 0.0022, 0.0070, 0.0010, + -0.0114, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 230.78, cls_loss 0.0349 cls_loss_mapping 0.0525 cls_loss_causal 0.8437 re_mapping 0.0280 re_causal 0.0774 /// teacc 98.43 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0177, -0.0147, -0.0026, ..., -0.0104, 0.0128, 0.0008], + [ 0.0064, 0.0163, -0.0175, ..., 0.0258, 0.0184, -0.1137], + [-0.0012, 0.0181, 0.0158, ..., -0.0163, 0.0291, -0.0078], + ..., + [-0.0124, -0.0114, 0.0276, ..., -0.0434, -0.0455, 0.0141], + [-0.0087, 0.0116, -0.0042, ..., -0.0321, 0.0102, -0.0600], + [-0.0180, 0.0089, 0.0017, ..., -0.0528, -0.0255, -0.0204]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.6839e-04, + -1.9157e-04, 1.8962e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3708e-05, + 1.1772e-04, 4.4219e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7834e-04, + -3.9597e-03, -2.7537e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4909e-05, + 8.5640e-04, 7.3276e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0933e-04, + 2.2087e-03, 2.6673e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4094e-05, + 6.5982e-05, 7.2360e-05]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0109, -0.0178, 0.0013, 0.0303, -0.0107, 0.0283, 0.0033, 0.0117, + 0.0092, -0.0049], device='cuda:0'), grad: tensor([-6.4039e-04, 9.0742e-04, -4.2763e-03, 2.2522e-02, 5.5838e-04, + -2.7508e-05, -4.0627e-04, -2.5940e-02, 4.3526e-03, 2.9659e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 231.05, cls_loss 0.0335 cls_loss_mapping 0.0572 cls_loss_causal 0.8764 re_mapping 0.0265 re_causal 0.0784 /// teacc 98.45 lr 0.00010000 +Epoch 24, weight, value: tensor([[-1.7729e-02, -1.4681e-02, -2.5723e-03, ..., -9.6231e-03, + 1.2939e-02, 1.8176e-03], + [ 6.1973e-03, 1.6330e-02, -1.7533e-02, ..., 2.5636e-02, + 1.7671e-02, -1.1878e-01], + [-1.1816e-04, 1.8066e-02, 1.5764e-02, ..., -1.6775e-02, + 2.9279e-02, -8.5139e-03], + ..., + [-1.2394e-02, -1.1373e-02, 2.7566e-02, ..., -4.3673e-02, + -4.6214e-02, 1.3608e-02], + [-8.7420e-03, 1.1618e-02, -4.2019e-03, ..., -3.2935e-02, + 1.0413e-02, -6.1694e-02], + [-1.8145e-02, 8.9221e-03, 1.7273e-03, ..., -5.3003e-02, + -2.5866e-02, -2.2892e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.8290e-04, + -1.4572e-03, 7.3109e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.5632e-06, + 7.9513e-05, 4.1164e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7059e-04, + -1.6654e-04, 7.1973e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.9945e-06, + 2.1958e-04, 5.9381e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6654e-04, + 1.0519e-03, 1.4855e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.9561e-05, + -1.5879e-03, 1.6168e-05]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0114, -0.0182, 0.0012, 0.0301, -0.0105, 0.0288, 0.0031, 0.0118, + 0.0091, -0.0051], device='cuda:0'), grad: tensor([-1.8387e-03, 6.2346e-05, -3.7265e-04, 6.0177e-04, 2.2554e-04, + 3.4237e-03, -1.8239e-05, 4.1008e-04, 2.1820e-03, -4.6768e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 214.54, cls_loss 0.0327 cls_loss_mapping 0.0481 cls_loss_causal 0.8450 re_mapping 0.0254 re_causal 0.0711 /// teacc 98.33 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0181, -0.0147, -0.0026, ..., -0.0093, 0.0125, 0.0003], + [ 0.0065, 0.0163, -0.0175, ..., 0.0258, 0.0173, -0.1198], + [-0.0007, 0.0181, 0.0158, ..., -0.0172, 0.0295, -0.0094], + ..., + [-0.0119, -0.0114, 0.0276, ..., -0.0438, -0.0470, 0.0152], + [-0.0089, 0.0116, -0.0042, ..., -0.0333, 0.0110, -0.0635], + [-0.0188, 0.0089, 0.0017, ..., -0.0532, -0.0260, -0.0247]], + device='cuda:0'), grad: tensor([[ 2.0862e-07, 0.0000e+00, 0.0000e+00, ..., -2.6464e-04, + 7.4911e-04, 9.2208e-05], + [-6.6422e-06, 0.0000e+00, 0.0000e+00, ..., -2.0489e-05, + 4.2868e-04, 8.4519e-05], + [ 4.6194e-07, 0.0000e+00, 0.0000e+00, ..., 3.9518e-05, + -3.8929e-03, 1.9503e-04], + ..., + [ 2.2091e-06, 0.0000e+00, 0.0000e+00, ..., 1.6168e-05, + 1.8244e-03, 1.9407e-04], + [ 1.3560e-06, 0.0000e+00, 0.0000e+00, ..., 7.4625e-05, + 8.6367e-05, 8.1837e-05], + [ 5.8208e-07, 0.0000e+00, 0.0000e+00, ..., 5.0992e-05, + 1.6654e-04, 1.0500e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0112, -0.0180, 0.0008, 0.0300, -0.0108, 0.0287, 0.0034, 0.0122, + 0.0090, -0.0047], device='cuda:0'), grad: tensor([ 0.0009, 0.0008, -0.0030, 0.0014, -0.0045, 0.0008, 0.0021, -0.0019, + 0.0006, 0.0028], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 230.84, cls_loss 0.0232 cls_loss_mapping 0.0399 cls_loss_causal 0.8480 re_mapping 0.0257 re_causal 0.0771 /// teacc 98.48 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0193, -0.0147, -0.0026, ..., -0.0090, 0.0120, -0.0004], + [ 0.0080, 0.0163, -0.0175, ..., 0.0265, 0.0172, -0.1201], + [-0.0026, 0.0181, 0.0158, ..., -0.0179, 0.0298, -0.0093], + ..., + [-0.0118, -0.0114, 0.0276, ..., -0.0439, -0.0479, 0.0160], + [-0.0073, 0.0116, -0.0042, ..., -0.0340, 0.0113, -0.0654], + [-0.0191, 0.0089, 0.0017, ..., -0.0539, -0.0265, -0.0269]], + device='cuda:0'), grad: tensor([[-5.6118e-05, 0.0000e+00, 0.0000e+00, ..., 4.7035e-03, + 4.8294e-03, 8.7246e-06], + [-4.7445e-05, 0.0000e+00, 0.0000e+00, ..., 9.3520e-05, + 5.4777e-05, 2.2620e-05], + [ 5.2482e-05, 0.0000e+00, 0.0000e+00, ..., 2.9802e-04, + 3.5095e-04, 1.9625e-05], + ..., + [ 9.5516e-06, 0.0000e+00, 0.0000e+00, ..., 6.6519e-05, + 5.8711e-05, 5.9247e-05], + [ 2.8223e-05, 0.0000e+00, 0.0000e+00, ..., 2.9516e-04, + -2.7871e-04, 4.1366e-05], + [ 2.7809e-06, 0.0000e+00, 0.0000e+00, ..., 1.2505e-04, + 7.7009e-05, 4.1693e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0113, -0.0178, 0.0010, 0.0301, -0.0107, 0.0286, 0.0032, 0.0127, + 0.0087, -0.0053], device='cuda:0'), grad: tensor([ 6.4087e-03, 2.5535e-04, 1.0805e-03, 2.1725e-03, -2.3675e-04, + -1.0967e-03, -8.3923e-03, 7.6592e-05, 8.3733e-04, -1.1005e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 214.53, cls_loss 0.0337 cls_loss_mapping 0.0495 cls_loss_causal 0.8459 re_mapping 0.0244 re_causal 0.0703 /// teacc 98.22 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0182, -0.0147, -0.0026, ..., -0.0085, 0.0117, 0.0016], + [ 0.0107, 0.0163, -0.0175, ..., 0.0264, 0.0167, -0.1221], + [-0.0045, 0.0181, 0.0158, ..., -0.0183, 0.0306, -0.0108], + ..., + [-0.0114, -0.0114, 0.0276, ..., -0.0442, -0.0490, 0.0152], + [-0.0080, 0.0116, -0.0042, ..., -0.0348, 0.0114, -0.0674], + [-0.0225, 0.0089, 0.0017, ..., -0.0539, -0.0271, -0.0282]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 0.0000e+00, 0.0000e+00, ..., -1.3142e-03, + -6.8712e-04, 9.2462e-06], + [-1.1977e-06, 0.0000e+00, 0.0000e+00, ..., -5.3376e-05, + 1.5008e-04, 2.9504e-05], + [ 8.8476e-08, 0.0000e+00, 0.0000e+00, ..., 1.0347e-04, + 5.8413e-04, 4.9829e-05], + ..., + [ 2.2445e-07, 0.0000e+00, 0.0000e+00, ..., 2.6256e-05, + -4.1500e-06, 1.7035e-04], + [ 2.2072e-07, 0.0000e+00, 0.0000e+00, ..., 1.7190e-04, + -2.0351e-03, 3.1859e-05], + [ 4.0047e-08, 0.0000e+00, 0.0000e+00, ..., 8.9598e-04, + 7.5769e-04, 5.0211e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0115, -0.0180, 0.0010, 0.0311, -0.0104, 0.0284, 0.0031, 0.0122, + 0.0083, -0.0054], device='cuda:0'), grad: tensor([-2.1648e-03, 9.5129e-05, 3.2234e-03, 1.1438e-04, -2.7800e-04, + 1.6861e-03, -7.7724e-05, -5.2986e-03, -1.6890e-03, 4.3907e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 214.57, cls_loss 0.0309 cls_loss_mapping 0.0514 cls_loss_causal 0.8547 re_mapping 0.0235 re_causal 0.0698 /// teacc 98.46 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0175, -0.0147, -0.0026, ..., -0.0081, 0.0110, 0.0016], + [ 0.0116, 0.0163, -0.0175, ..., 0.0269, 0.0161, -0.1246], + [-0.0078, 0.0181, 0.0158, ..., -0.0188, 0.0314, -0.0114], + ..., + [-0.0097, -0.0114, 0.0276, ..., -0.0446, -0.0495, 0.0148], + [-0.0085, 0.0116, -0.0042, ..., -0.0354, 0.0118, -0.0691], + [-0.0245, 0.0089, 0.0017, ..., -0.0542, -0.0268, -0.0294]], + device='cuda:0'), grad: tensor([[-7.8738e-05, 0.0000e+00, 0.0000e+00, ..., -3.7789e-04, + 2.0313e-04, 8.2981e-07], + [ 5.1856e-06, 0.0000e+00, 0.0000e+00, ..., 4.3005e-05, + 1.2922e-04, 3.7123e-06], + [ 7.1190e-06, 0.0000e+00, 0.0000e+00, ..., 8.2254e-05, + -7.0000e-04, 2.2389e-06], + ..., + [ 3.1628e-06, 0.0000e+00, 0.0000e+00, ..., 1.9833e-05, + 1.5318e-04, -3.0175e-05], + [ 4.2543e-06, 0.0000e+00, 0.0000e+00, ..., 5.0724e-05, + -2.3575e-03, 9.7789e-07], + [ 3.0145e-05, 0.0000e+00, 0.0000e+00, ..., 1.5533e-04, + 1.5421e-03, 1.2390e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0111, -0.0173, 0.0013, 0.0311, -0.0105, 0.0283, 0.0030, 0.0122, + 0.0082, -0.0055], device='cuda:0'), grad: tensor([-6.8712e-04, 3.4833e-04, -9.3603e-04, 7.9393e-04, 2.2125e-04, + 4.5228e-04, 1.9240e-04, -4.6909e-05, -2.8343e-03, 2.4967e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 214.47, cls_loss 0.0317 cls_loss_mapping 0.0504 cls_loss_causal 0.8206 re_mapping 0.0229 re_causal 0.0671 /// teacc 98.47 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0174, -0.0147, -0.0026, ..., -0.0077, 0.0109, 0.0006], + [ 0.0119, 0.0163, -0.0175, ..., 0.0270, 0.0155, -0.1283], + [-0.0107, 0.0181, 0.0158, ..., -0.0193, 0.0316, -0.0125], + ..., + [-0.0061, -0.0114, 0.0276, ..., -0.0448, -0.0499, 0.0159], + [-0.0087, 0.0116, -0.0042, ..., -0.0359, 0.0126, -0.0708], + [-0.0253, 0.0089, 0.0017, ..., -0.0546, -0.0276, -0.0297]], + device='cuda:0'), grad: tensor([[ 1.1856e-06, 0.0000e+00, 0.0000e+00, ..., 1.3262e-05, + 5.5885e-04, 1.4365e-05], + [-2.3663e-05, 0.0000e+00, 0.0000e+00, ..., 2.8625e-05, + 2.3329e-04, 3.4422e-05], + [ 3.4142e-06, 0.0000e+00, 0.0000e+00, ..., 2.2560e-05, + -5.3215e-03, 5.4985e-05], + ..., + [ 5.8562e-06, 0.0000e+00, 0.0000e+00, ..., 1.7345e-05, + 5.8556e-04, 7.1168e-05], + [ 3.7383e-06, 0.0000e+00, 0.0000e+00, ..., 1.8120e-04, + -1.5812e-03, -6.0081e-04], + [ 1.4910e-06, 0.0000e+00, 0.0000e+00, ..., 2.2694e-05, + 9.1887e-04, 2.8777e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0113, -0.0174, 0.0010, 0.0309, -0.0106, 0.0284, 0.0028, 0.0127, + 0.0086, -0.0058], device='cuda:0'), grad: tensor([ 0.0009, 0.0003, -0.0076, 0.0050, 0.0013, 0.0037, -0.0008, 0.0017, + -0.0067, 0.0021], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 230.56, cls_loss 0.0244 cls_loss_mapping 0.0386 cls_loss_causal 0.8062 re_mapping 0.0232 re_causal 0.0684 /// teacc 98.62 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0169, -0.0147, -0.0026, ..., -0.0072, 0.0108, -0.0005], + [ 0.0133, 0.0163, -0.0175, ..., 0.0268, 0.0152, -0.1302], + [-0.0119, 0.0181, 0.0158, ..., -0.0196, 0.0320, -0.0131], + ..., + [-0.0059, -0.0114, 0.0276, ..., -0.0450, -0.0508, 0.0167], + [-0.0091, 0.0116, -0.0042, ..., -0.0363, 0.0130, -0.0732], + [-0.0262, 0.0089, 0.0017, ..., -0.0549, -0.0278, -0.0312]], + device='cuda:0'), grad: tensor([[ 7.0408e-07, 0.0000e+00, 0.0000e+00, ..., 1.0878e-04, + 1.6844e-04, 1.5542e-05], + [-7.3165e-06, 0.0000e+00, 0.0000e+00, ..., -1.5807e-04, + 6.6161e-05, 7.4387e-05], + [ 4.2189e-07, 0.0000e+00, 0.0000e+00, ..., 1.2210e-06, + -4.3607e-04, 1.5453e-05], + ..., + [ 1.4408e-06, 0.0000e+00, 0.0000e+00, ..., 4.9025e-05, + 4.7177e-05, 2.6524e-06], + [ 8.5775e-07, 0.0000e+00, 0.0000e+00, ..., 1.8263e-04, + 3.3474e-04, 3.5465e-05], + [ 5.6997e-07, 0.0000e+00, 0.0000e+00, ..., 8.6784e-05, + 8.8513e-05, 7.2956e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0114, -0.0178, 0.0008, 0.0313, -0.0105, 0.0285, 0.0025, 0.0125, + 0.0085, -0.0055], device='cuda:0'), grad: tensor([ 0.0003, -0.0003, -0.0004, 0.0003, -0.0015, -0.0008, 0.0005, -0.0001, + 0.0007, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 214.80, cls_loss 0.0233 cls_loss_mapping 0.0403 cls_loss_causal 0.8009 re_mapping 0.0230 re_causal 0.0651 /// teacc 98.52 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0168, -0.0147, -0.0026, ..., -0.0070, 0.0105, -0.0004], + [ 0.0147, 0.0163, -0.0175, ..., 0.0264, 0.0144, -0.1324], + [-0.0139, 0.0181, 0.0158, ..., -0.0199, 0.0322, -0.0137], + ..., + [-0.0051, -0.0114, 0.0276, ..., -0.0452, -0.0511, 0.0173], + [-0.0093, 0.0116, -0.0042, ..., -0.0364, 0.0137, -0.0744], + [-0.0270, 0.0089, 0.0017, ..., -0.0553, -0.0283, -0.0324]], + device='cuda:0'), grad: tensor([[-5.9791e-07, 0.0000e+00, 0.0000e+00, ..., 2.1141e-06, + 1.1855e-04, 4.6752e-06], + [ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., -2.6032e-05, + 1.6975e-04, 9.5516e-06], + [ 5.1223e-08, 0.0000e+00, 0.0000e+00, ..., 1.1034e-05, + -1.5229e-05, 2.1383e-05], + ..., + [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 1.4886e-05, + 2.5826e-03, 1.8641e-05], + [ 3.9116e-08, 0.0000e+00, 0.0000e+00, ..., 1.2720e-04, + -4.2839e-03, 1.3441e-05], + [ 3.1665e-08, 0.0000e+00, 0.0000e+00, ..., 1.6361e-05, + 3.3140e-04, 6.1893e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0113, -0.0181, 0.0010, 0.0311, -0.0103, 0.0281, 0.0030, 0.0130, + 0.0085, -0.0058], device='cuda:0'), grad: tensor([ 0.0004, 0.0002, 0.0003, 0.0100, -0.0010, 0.0023, 0.0006, 0.0103, + -0.0162, -0.0069], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 215.20, cls_loss 0.0236 cls_loss_mapping 0.0428 cls_loss_causal 0.8644 re_mapping 0.0202 re_causal 0.0655 /// teacc 98.48 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0170, -0.0147, -0.0026, ..., -0.0068, 0.0101, -0.0016], + [ 0.0140, 0.0163, -0.0175, ..., 0.0265, 0.0139, -0.1338], + [-0.0121, 0.0181, 0.0158, ..., -0.0204, 0.0328, -0.0148], + ..., + [-0.0052, -0.0114, 0.0276, ..., -0.0454, -0.0517, 0.0167], + [-0.0092, 0.0116, -0.0042, ..., -0.0365, 0.0143, -0.0764], + [-0.0272, 0.0089, 0.0017, ..., -0.0557, -0.0290, -0.0325]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4891e-04, + 4.5991e-04, 1.9714e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9011e-05, + 1.5485e-04, 9.2089e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8323e-05, + -1.5526e-03, 3.9548e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6345e-05, + 1.0452e-03, 1.0097e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9014e-04, + -9.6035e-04, 5.7995e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.8862e-05, + 5.8174e-04, 9.0182e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0111, -0.0183, 0.0013, 0.0306, -0.0100, 0.0285, 0.0024, 0.0129, + 0.0087, -0.0057], device='cuda:0'), grad: tensor([ 0.0009, 0.0004, -0.0029, 0.0004, -0.0004, 0.0005, -0.0009, 0.0029, + 0.0006, -0.0016], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 214.77, cls_loss 0.0225 cls_loss_mapping 0.0373 cls_loss_causal 0.7768 re_mapping 0.0213 re_causal 0.0622 /// teacc 98.59 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0167, -0.0147, -0.0026, ..., -0.0062, 0.0100, -0.0010], + [ 0.0140, 0.0163, -0.0175, ..., 0.0264, 0.0132, -0.1350], + [-0.0111, 0.0181, 0.0158, ..., -0.0208, 0.0333, -0.0155], + ..., + [-0.0052, -0.0114, 0.0276, ..., -0.0457, -0.0518, 0.0161], + [-0.0095, 0.0116, -0.0042, ..., -0.0368, 0.0147, -0.0782], + [-0.0280, 0.0089, 0.0017, ..., -0.0558, -0.0295, -0.0344]], + device='cuda:0'), grad: tensor([[ 5.1409e-07, 0.0000e+00, 0.0000e+00, ..., 1.8823e-04, + 2.1338e-04, 1.9655e-05], + [-1.7509e-05, 0.0000e+00, 0.0000e+00, ..., 8.9109e-06, + 5.0366e-05, 2.2531e-05], + [ 1.8692e-06, 0.0000e+00, 0.0000e+00, ..., 2.2978e-05, + -4.0382e-05, 1.8924e-05], + ..., + [ 2.5369e-06, 0.0000e+00, 0.0000e+00, ..., 7.4469e-06, + 6.5207e-05, 7.1339e-06], + [ 7.4692e-06, 0.0000e+00, 0.0000e+00, ..., 1.0586e-04, + -6.6221e-05, 7.7784e-06], + [ 4.3120e-07, 0.0000e+00, 0.0000e+00, ..., 3.0130e-05, + 6.3598e-05, 5.1916e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0111, -0.0187, 0.0014, 0.0306, -0.0096, 0.0285, 0.0025, 0.0130, + 0.0084, -0.0056], device='cuda:0'), grad: tensor([ 3.2163e-04, 6.1512e-05, 2.7156e-04, -6.8951e-04, -4.6563e-04, + 2.7418e-04, -4.9067e-04, 3.4094e-04, 1.0163e-04, 2.7394e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 214.66, cls_loss 0.0201 cls_loss_mapping 0.0359 cls_loss_causal 0.7668 re_mapping 0.0206 re_causal 0.0623 /// teacc 98.57 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0167, -0.0147, -0.0026, ..., -0.0063, 0.0094, -0.0013], + [ 0.0143, 0.0163, -0.0175, ..., 0.0266, 0.0129, -0.1367], + [-0.0111, 0.0181, 0.0158, ..., -0.0212, 0.0334, -0.0156], + ..., + [-0.0052, -0.0114, 0.0276, ..., -0.0459, -0.0524, 0.0159], + [-0.0096, 0.0116, -0.0042, ..., -0.0373, 0.0148, -0.0792], + [-0.0284, 0.0089, 0.0017, ..., -0.0556, -0.0300, -0.0342]], + device='cuda:0'), grad: tensor([[-3.6005e-06, 0.0000e+00, 0.0000e+00, ..., -2.2411e-03, + -1.1320e-03, 6.2212e-06], + [-5.0478e-06, 0.0000e+00, 0.0000e+00, ..., 1.0198e-04, + 8.6963e-05, 8.4579e-05], + [ 3.3416e-06, 0.0000e+00, 0.0000e+00, ..., 3.2991e-05, + 1.1158e-04, 3.8177e-05], + ..., + [ 7.3481e-07, 0.0000e+00, 0.0000e+00, ..., 8.7395e-06, + 2.5541e-05, 5.6595e-05], + [ 1.8170e-06, 0.0000e+00, 0.0000e+00, ..., 1.7011e-04, + -9.3162e-05, 5.6684e-05], + [ 9.9931e-07, 0.0000e+00, 0.0000e+00, ..., 4.0025e-05, + 2.7728e-04, 3.7026e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0104, -0.0185, 0.0017, 0.0308, -0.0099, 0.0287, 0.0030, 0.0126, + 0.0082, -0.0054], device='cuda:0'), grad: tensor([-2.0905e-03, 4.2939e-04, 2.8014e-04, -3.1624e-03, -1.2884e-03, + 2.4319e-03, 2.1572e-03, 9.9540e-05, 1.2600e-04, 1.0176e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 230.82, cls_loss 0.0192 cls_loss_mapping 0.0362 cls_loss_causal 0.8033 re_mapping 0.0204 re_causal 0.0595 /// teacc 98.71 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0158, -0.0147, -0.0026, ..., -0.0060, 0.0091, -0.0002], + [ 0.0154, 0.0163, -0.0175, ..., 0.0270, 0.0123, -0.1384], + [-0.0127, 0.0181, 0.0158, ..., -0.0218, 0.0338, -0.0175], + ..., + [-0.0048, -0.0114, 0.0276, ..., -0.0461, -0.0527, 0.0156], + [-0.0102, 0.0116, -0.0042, ..., -0.0381, 0.0151, -0.0793], + [-0.0302, 0.0089, 0.0017, ..., -0.0563, -0.0303, -0.0345]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.0545e-04, + -6.0654e-04, 3.0082e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.4052e-07, + 4.8041e-05, 4.9949e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0204e-04, + 2.8276e-04, 1.3351e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0756e-05, + 5.5403e-05, 4.4972e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2331e-03, + -1.8978e-03, 2.1741e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9683e-04, + 3.9816e-04, 1.3196e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0107, -0.0188, 0.0016, 0.0309, -0.0096, 0.0286, 0.0028, 0.0129, + 0.0082, -0.0056], device='cuda:0'), grad: tensor([-3.2444e-03, -5.7364e-04, 1.1730e-03, 1.9655e-05, -5.4479e-05, + 1.7986e-03, 6.2084e-04, 3.9244e-04, -2.1267e-03, 1.9913e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 214.78, cls_loss 0.0223 cls_loss_mapping 0.0392 cls_loss_causal 0.7438 re_mapping 0.0190 re_causal 0.0526 /// teacc 98.68 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0158, -0.0147, -0.0026, ..., -0.0055, 0.0088, -0.0007], + [ 0.0156, 0.0163, -0.0175, ..., 0.0268, 0.0116, -0.1388], + [-0.0127, 0.0181, 0.0158, ..., -0.0224, 0.0344, -0.0184], + ..., + [-0.0048, -0.0114, 0.0276, ..., -0.0463, -0.0536, 0.0157], + [-0.0100, 0.0116, -0.0042, ..., -0.0377, 0.0162, -0.0799], + [-0.0304, 0.0089, 0.0017, ..., -0.0568, -0.0309, -0.0358]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.9838e-05, + 9.5129e-05, 4.2561e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.2978e-06, + -4.2076e-03, 3.0268e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3826e-05, + 3.2520e-03, 2.3097e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.2436e-06, + 2.6584e-04, 1.0170e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2395e-05, + -1.9991e-04, 1.5572e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4633e-05, + -3.3498e-05, 2.7269e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0107, -0.0190, 0.0016, 0.0309, -0.0097, 0.0280, 0.0025, 0.0133, + 0.0090, -0.0059], device='cuda:0'), grad: tensor([ 0.0001, -0.0078, 0.0061, 0.0006, 0.0003, 0.0004, 0.0003, 0.0006, + 0.0004, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 215.24, cls_loss 0.0198 cls_loss_mapping 0.0357 cls_loss_causal 0.7668 re_mapping 0.0182 re_causal 0.0573 /// teacc 98.65 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0166, -0.0147, -0.0026, ..., -0.0055, 0.0084, -0.0009], + [ 0.0163, 0.0163, -0.0175, ..., 0.0269, 0.0115, -0.1397], + [-0.0139, 0.0181, 0.0158, ..., -0.0232, 0.0343, -0.0183], + ..., + [-0.0043, -0.0114, 0.0276, ..., -0.0464, -0.0544, 0.0156], + [-0.0094, 0.0116, -0.0042, ..., -0.0381, 0.0168, -0.0805], + [-0.0311, 0.0089, 0.0017, ..., -0.0568, -0.0314, -0.0365]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -3.9029e-04, + -1.7226e-04, 3.8184e-06], + [ 9.2201e-08, 0.0000e+00, 0.0000e+00, ..., 1.7321e-04, + 1.5295e-04, 2.3693e-05], + [ 2.6077e-08, 0.0000e+00, 0.0000e+00, ..., 7.5459e-05, + -6.8069e-05, 1.4327e-05], + ..., + [-3.0641e-07, 0.0000e+00, 0.0000e+00, ..., 1.2420e-05, + 3.9667e-05, 1.8477e-05], + [ 3.6322e-08, 0.0000e+00, 0.0000e+00, ..., 7.3957e-04, + 6.2847e-04, 9.2834e-06], + [ 6.2399e-08, 0.0000e+00, 0.0000e+00, ..., 2.3305e-04, + 1.0568e-04, 1.0335e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0100, -0.0186, 0.0015, 0.0306, -0.0092, 0.0283, 0.0027, 0.0132, + 0.0089, -0.0059], device='cuda:0'), grad: tensor([-6.4993e-04, 3.8958e-04, 6.8247e-05, -3.3903e-04, -1.4031e-04, + 7.3957e-04, -1.7347e-03, -1.3566e-04, 1.2007e-03, 6.0225e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 214.82, cls_loss 0.0172 cls_loss_mapping 0.0316 cls_loss_causal 0.7487 re_mapping 0.0187 re_causal 0.0568 /// teacc 98.54 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0167, -0.0147, -0.0026, ..., -0.0049, 0.0084, -0.0024], + [ 0.0164, 0.0163, -0.0175, ..., 0.0269, 0.0108, -0.1410], + [-0.0144, 0.0181, 0.0158, ..., -0.0237, 0.0350, -0.0187], + ..., + [-0.0040, -0.0114, 0.0276, ..., -0.0469, -0.0548, 0.0156], + [-0.0094, 0.0116, -0.0042, ..., -0.0386, 0.0170, -0.0813], + [-0.0319, 0.0089, 0.0017, ..., -0.0573, -0.0322, -0.0376]], + device='cuda:0'), grad: tensor([[-1.1712e-05, 0.0000e+00, 0.0000e+00, ..., -2.3823e-03, + 8.2672e-05, 5.8636e-06], + [ 3.1386e-06, 0.0000e+00, 0.0000e+00, ..., 6.8092e-04, + 1.0138e-03, 2.0936e-05], + [-7.7295e-04, 0.0000e+00, 0.0000e+00, ..., 3.3259e-04, + -2.0103e-03, 5.9843e-05], + ..., + [ 7.5436e-04, 0.0000e+00, 0.0000e+00, ..., 1.4968e-05, + 2.4815e-03, 6.3255e-06], + [ 6.8210e-06, 0.0000e+00, 0.0000e+00, ..., -2.1315e-04, + -4.0779e-03, 5.7407e-06], + [ 1.2498e-06, 0.0000e+00, 0.0000e+00, ..., 1.1301e-04, + 8.6880e-04, 1.6615e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0103, -0.0186, 0.0015, 0.0309, -0.0093, 0.0282, 0.0028, 0.0134, + 0.0085, -0.0061], device='cuda:0'), grad: tensor([-0.0022, 0.0022, -0.0013, 0.0014, 0.0021, 0.0006, 0.0012, 0.0026, + -0.0078, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 230.79, cls_loss 0.0154 cls_loss_mapping 0.0268 cls_loss_causal 0.7396 re_mapping 0.0177 re_causal 0.0550 /// teacc 98.78 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0170, -0.0147, -0.0026, ..., -0.0045, 0.0081, -0.0027], + [ 0.0162, 0.0163, -0.0175, ..., 0.0272, 0.0100, -0.1436], + [-0.0143, 0.0181, 0.0158, ..., -0.0241, 0.0359, -0.0191], + ..., + [-0.0035, -0.0114, 0.0276, ..., -0.0471, -0.0555, 0.0156], + [-0.0095, 0.0116, -0.0042, ..., -0.0386, 0.0174, -0.0834], + [-0.0334, 0.0089, 0.0017, ..., -0.0577, -0.0327, -0.0377]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3256e-04, + 2.7597e-05, 2.4065e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.4107e-06, + 8.5905e-06, 8.1003e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1336e-05, + 1.5244e-05, 8.6784e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3314e-05, + 7.5102e-06, 1.1647e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0171e-05, + 2.7090e-05, 1.4752e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9439e-05, + 3.4899e-05, 2.1386e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0104, -0.0189, 0.0019, 0.0307, -0.0090, 0.0287, 0.0024, 0.0135, + 0.0084, -0.0065], device='cuda:0'), grad: tensor([-3.9983e-04, 8.6486e-05, 3.2544e-04, -6.9714e-04, -9.5987e-04, + 4.1103e-04, -1.0848e-04, 3.8695e-04, 2.3746e-04, 7.1859e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 214.28, cls_loss 0.0195 cls_loss_mapping 0.0328 cls_loss_causal 0.7374 re_mapping 0.0171 re_causal 0.0493 /// teacc 98.66 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0171, -0.0147, -0.0026, ..., -0.0042, 0.0076, -0.0035], + [ 0.0160, 0.0163, -0.0175, ..., 0.0273, 0.0095, -0.1455], + [-0.0148, 0.0181, 0.0158, ..., -0.0246, 0.0368, -0.0196], + ..., + [-0.0030, -0.0114, 0.0276, ..., -0.0472, -0.0563, 0.0144], + [-0.0096, 0.0116, -0.0042, ..., -0.0389, 0.0175, -0.0850], + [-0.0340, 0.0089, 0.0017, ..., -0.0580, -0.0334, -0.0380]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.6359e-05, + 2.7820e-05, 8.7544e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.9744e-05, + 2.0400e-05, 5.8953e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4393e-05, + -4.5449e-05, 1.0040e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.4964e-06, + 1.7658e-05, 6.7428e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4481e-05, + 4.8786e-05, 6.7055e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2457e-05, + -9.4235e-05, 4.7032e-07]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0103, -0.0187, 0.0022, 0.0308, -0.0090, 0.0288, 0.0023, 0.0134, + 0.0080, -0.0065], device='cuda:0'), grad: tensor([ 1.2755e-04, 8.7404e-04, 8.3160e-04, 4.3130e-04, 2.4009e-04, + 9.4295e-05, 2.3320e-05, 2.6108e-02, 6.1560e-04, -2.9343e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 214.56, cls_loss 0.0145 cls_loss_mapping 0.0238 cls_loss_causal 0.7078 re_mapping 0.0176 re_causal 0.0496 /// teacc 98.60 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0169, -0.0147, -0.0026, ..., -0.0040, 0.0074, -0.0039], + [ 0.0172, 0.0163, -0.0175, ..., 0.0274, 0.0091, -0.1470], + [-0.0153, 0.0181, 0.0158, ..., -0.0249, 0.0372, -0.0200], + ..., + [-0.0033, -0.0114, 0.0276, ..., -0.0472, -0.0566, 0.0144], + [-0.0098, 0.0116, -0.0042, ..., -0.0398, 0.0178, -0.0853], + [-0.0356, 0.0089, 0.0017, ..., -0.0584, -0.0340, -0.0387]], + device='cuda:0'), grad: tensor([[ 6.5379e-07, 0.0000e+00, 0.0000e+00, ..., -4.7274e-06, + 1.3791e-05, 9.7137e-07], + [-4.0412e-05, 0.0000e+00, 0.0000e+00, ..., -4.4182e-06, + 5.6416e-05, 5.7928e-07], + [ 6.1579e-06, 0.0000e+00, 0.0000e+00, ..., 3.5185e-06, + -9.0718e-05, 4.6268e-06], + ..., + [ 1.1519e-05, 0.0000e+00, 0.0000e+00, ..., 1.2526e-06, + 4.0650e-05, 8.0019e-06], + [ 4.6566e-06, 0.0000e+00, 0.0000e+00, ..., 4.1351e-06, + -3.5405e-05, 3.1684e-06], + [ 4.3735e-06, 0.0000e+00, 0.0000e+00, ..., 2.5071e-06, + 1.6376e-05, 1.9923e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0100, -0.0189, 0.0017, 0.0309, -0.0089, 0.0286, 0.0029, 0.0138, + 0.0078, -0.0064], device='cuda:0'), grad: tensor([ 3.9726e-05, 5.6171e-04, 5.0020e-04, -1.2960e-03, 7.1287e-04, + 1.1581e-04, 2.7373e-05, 7.7486e-04, 1.2070e-04, -1.5574e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 214.33, cls_loss 0.0160 cls_loss_mapping 0.0296 cls_loss_causal 0.7057 re_mapping 0.0170 re_causal 0.0513 /// teacc 98.52 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0168, -0.0147, -0.0026, ..., -0.0032, 0.0077, -0.0044], + [ 0.0173, 0.0163, -0.0175, ..., 0.0273, 0.0089, -0.1488], + [-0.0154, 0.0181, 0.0158, ..., -0.0253, 0.0376, -0.0208], + ..., + [-0.0028, -0.0114, 0.0276, ..., -0.0476, -0.0575, 0.0153], + [-0.0099, 0.0116, -0.0042, ..., -0.0401, 0.0182, -0.0871], + [-0.0359, 0.0089, 0.0017, ..., -0.0586, -0.0345, -0.0390]], + device='cuda:0'), grad: tensor([[ 6.5658e-08, 0.0000e+00, 0.0000e+00, ..., -4.4078e-05, + 3.3379e-05, 3.2280e-06], + [ 4.6939e-07, 0.0000e+00, 0.0000e+00, ..., -9.2760e-06, + 3.5558e-06, 1.3612e-05], + [ 8.8383e-07, 0.0000e+00, 0.0000e+00, ..., 1.0952e-05, + -1.7738e-04, 1.2934e-05], + ..., + [ 9.4809e-07, 0.0000e+00, 0.0000e+00, ..., 8.1286e-06, + 9.3102e-05, 9.5487e-05], + [ 7.5903e-08, 0.0000e+00, 0.0000e+00, ..., 5.8979e-05, + 1.2207e-04, 9.3505e-06], + [ 4.3772e-07, 0.0000e+00, 0.0000e+00, ..., 3.4690e-05, + 2.1845e-05, 8.1658e-06]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0104, -0.0187, 0.0013, 0.0312, -0.0095, 0.0284, 0.0027, 0.0142, + 0.0078, -0.0064], device='cuda:0'), grad: tensor([-7.9870e-06, -3.6287e-04, 8.4713e-06, 1.6677e-04, -1.6797e-04, + -7.2908e-04, 2.6631e-04, 7.0763e-04, 3.0470e-04, -1.8537e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 214.66, cls_loss 0.0161 cls_loss_mapping 0.0258 cls_loss_causal 0.7043 re_mapping 0.0172 re_causal 0.0500 /// teacc 98.62 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0167, -0.0147, -0.0026, ..., -0.0029, 0.0074, -0.0048], + [ 0.0180, 0.0163, -0.0175, ..., 0.0273, 0.0086, -0.1496], + [-0.0154, 0.0181, 0.0158, ..., -0.0257, 0.0376, -0.0215], + ..., + [-0.0018, -0.0114, 0.0276, ..., -0.0478, -0.0579, 0.0150], + [-0.0103, 0.0116, -0.0042, ..., -0.0404, 0.0190, -0.0890], + [-0.0373, 0.0089, 0.0017, ..., -0.0592, -0.0355, -0.0408]], + device='cuda:0'), grad: tensor([[ 1.9232e-07, 0.0000e+00, 0.0000e+00, ..., 5.3525e-05, + 1.5569e-04, 2.1365e-06], + [ 4.2329e-07, 0.0000e+00, 0.0000e+00, ..., 3.6713e-06, + 2.2128e-05, 9.3803e-06], + [ 4.2003e-07, 0.0000e+00, 0.0000e+00, ..., 2.6487e-06, + -4.1890e-04, 7.9945e-06], + ..., + [ 1.0058e-07, 0.0000e+00, 0.0000e+00, ..., 5.2527e-06, + 5.7787e-05, 1.9088e-05], + [ 2.6496e-07, 0.0000e+00, 0.0000e+00, ..., 5.0396e-05, + 2.0492e-04, 3.5558e-06], + [ 1.6112e-06, 0.0000e+00, 0.0000e+00, ..., -2.7633e-04, + -2.6107e-04, 2.4259e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0104, -0.0184, 0.0009, 0.0310, -0.0094, 0.0288, 0.0026, 0.0141, + 0.0079, -0.0064], device='cuda:0'), grad: tensor([ 0.0003, 0.0001, -0.0004, -0.0011, 0.0004, 0.0008, 0.0001, -0.0003, + 0.0006, -0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.53, cls_loss 0.0172 cls_loss_mapping 0.0278 cls_loss_causal 0.7190 re_mapping 0.0171 re_causal 0.0510 /// teacc 98.50 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0169, -0.0147, -0.0026, ..., -0.0025, 0.0071, -0.0056], + [ 0.0186, 0.0163, -0.0175, ..., 0.0272, 0.0079, -0.1500], + [-0.0146, 0.0181, 0.0158, ..., -0.0260, 0.0384, -0.0227], + ..., + [-0.0012, -0.0114, 0.0276, ..., -0.0480, -0.0581, 0.0169], + [-0.0106, 0.0116, -0.0042, ..., -0.0411, 0.0186, -0.0899], + [-0.0382, 0.0089, 0.0017, ..., -0.0596, -0.0358, -0.0421]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3030e-04, + 4.3535e-04, 1.7788e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3006e-06, + 8.1778e-05, 1.2405e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.9754e-06, + -1.2159e-03, 7.1013e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4650e-06, + 1.3857e-03, 3.3025e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8907e-04, + -1.1396e-03, 4.1397e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2338e-05, + 2.2161e-04, 9.1195e-06]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0104, -0.0178, 0.0017, 0.0303, -0.0095, 0.0295, 0.0027, 0.0142, + 0.0072, -0.0068], device='cuda:0'), grad: tensor([ 5.4455e-04, -1.4648e-05, -2.8763e-03, 6.4313e-05, 1.1981e-04, + 1.5032e-04, 7.7903e-05, 3.2520e-03, -1.6375e-03, 3.1662e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 214.61, cls_loss 0.0167 cls_loss_mapping 0.0320 cls_loss_causal 0.6996 re_mapping 0.0165 re_causal 0.0486 /// teacc 98.67 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0158, -0.0147, -0.0026, ..., -0.0026, 0.0060, -0.0045], + [ 0.0186, 0.0163, -0.0175, ..., 0.0270, 0.0073, -0.1509], + [-0.0130, 0.0181, 0.0158, ..., -0.0264, 0.0388, -0.0238], + ..., + [-0.0019, -0.0114, 0.0276, ..., -0.0483, -0.0588, 0.0159], + [-0.0109, 0.0116, -0.0042, ..., -0.0417, 0.0186, -0.0910], + [-0.0396, 0.0089, 0.0017, ..., -0.0598, -0.0360, -0.0420]], + device='cuda:0'), grad: tensor([[ 2.0172e-06, 0.0000e+00, 0.0000e+00, ..., -1.0735e-02, + -4.7989e-03, 2.8079e-07], + [ 8.1286e-06, 0.0000e+00, 0.0000e+00, ..., 5.1260e-06, + 2.1160e-05, 1.2433e-07], + [ 5.8934e-06, 0.0000e+00, 0.0000e+00, ..., 7.5661e-06, + -2.5344e-04, 1.7975e-07], + ..., + [-6.2108e-05, 0.0000e+00, 0.0000e+00, ..., 1.3299e-06, + 6.2346e-05, 7.8231e-08], + [ 1.1459e-05, 0.0000e+00, 0.0000e+00, ..., 2.5705e-05, + 1.4377e-04, 5.2620e-08], + [ 2.1905e-05, 0.0000e+00, 0.0000e+00, ..., 1.4208e-05, + 8.7693e-06, 3.3341e-07]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0100, -0.0181, 0.0014, 0.0313, -0.0093, 0.0293, 0.0033, 0.0137, + 0.0067, -0.0065], device='cuda:0'), grad: tensor([-9.7809e-03, 1.0949e-04, -2.4986e-04, 1.3940e-05, 1.0097e-04, + 1.4269e-04, 9.6588e-03, -5.0926e-04, 3.0828e-04, 2.0289e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 214.71, cls_loss 0.0136 cls_loss_mapping 0.0248 cls_loss_causal 0.7131 re_mapping 0.0156 re_causal 0.0492 /// teacc 98.69 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0150, -0.0147, -0.0026, ..., -0.0022, 0.0056, -0.0047], + [ 0.0183, 0.0163, -0.0175, ..., 0.0275, 0.0068, -0.1516], + [-0.0132, 0.0181, 0.0158, ..., -0.0269, 0.0395, -0.0240], + ..., + [-0.0013, -0.0114, 0.0276, ..., -0.0486, -0.0591, 0.0160], + [-0.0112, 0.0116, -0.0042, ..., -0.0423, 0.0188, -0.0914], + [-0.0408, 0.0089, 0.0017, ..., -0.0598, -0.0362, -0.0430]], + device='cuda:0'), grad: tensor([[-5.0990e-07, 0.0000e+00, 0.0000e+00, ..., -1.6734e-05, + 8.8615e-07, -1.6298e-03], + [ 6.1933e-08, 0.0000e+00, 0.0000e+00, ..., -1.9163e-05, + 6.5006e-06, 1.0714e-05], + [ 5.0897e-07, 0.0000e+00, 0.0000e+00, ..., 7.8142e-05, + 1.4532e-04, 2.0027e-04], + ..., + [-5.2620e-07, 0.0000e+00, 0.0000e+00, ..., 3.5409e-06, + 5.5209e-06, 1.8001e-05], + [ 3.5390e-08, 0.0000e+00, 0.0000e+00, ..., -6.3300e-05, + -3.3450e-04, 3.1620e-05], + [ 5.9605e-08, 0.0000e+00, 0.0000e+00, ..., 5.7109e-06, + 1.2323e-05, 2.1732e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0102, -0.0180, 0.0014, 0.0308, -0.0090, 0.0299, 0.0024, 0.0138, + 0.0068, -0.0067], device='cuda:0'), grad: tensor([-4.2915e-03, -7.2598e-05, 6.8855e-04, 3.2234e-04, 2.2144e-03, + 2.6846e-04, 2.8944e-04, 2.1011e-05, -3.4600e-05, 5.9652e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 214.77, cls_loss 0.0131 cls_loss_mapping 0.0234 cls_loss_causal 0.7104 re_mapping 0.0164 re_causal 0.0479 /// teacc 98.74 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0150, -0.0147, -0.0026, ..., -0.0022, 0.0055, -0.0045], + [ 0.0179, 0.0163, -0.0175, ..., 0.0275, 0.0063, -0.1531], + [-0.0129, 0.0181, 0.0158, ..., -0.0273, 0.0400, -0.0246], + ..., + [-0.0008, -0.0114, 0.0276, ..., -0.0488, -0.0597, 0.0170], + [-0.0112, 0.0116, -0.0042, ..., -0.0426, 0.0191, -0.0926], + [-0.0412, 0.0089, 0.0017, ..., -0.0597, -0.0368, -0.0433]], + device='cuda:0'), grad: tensor([[ 6.8452e-08, 0.0000e+00, 0.0000e+00, ..., -1.2898e-04, + -2.5809e-05, 2.0452e-06], + [ 1.7509e-07, 0.0000e+00, 0.0000e+00, ..., 4.0159e-06, + 3.1638e-04, 3.8370e-06], + [ 8.9407e-08, 0.0000e+00, 0.0000e+00, ..., 1.0923e-05, + -1.4400e-03, -3.9563e-06], + ..., + [-1.8608e-06, 0.0000e+00, 0.0000e+00, ..., 3.2745e-06, + 3.6788e-04, 7.5763e-07], + [ 2.7008e-07, 0.0000e+00, 0.0000e+00, ..., 1.1995e-05, + 3.6657e-05, 2.1998e-06], + [ 9.0711e-07, 0.0000e+00, 0.0000e+00, ..., 5.2154e-05, + 2.7955e-05, 2.0230e-04]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0101, -0.0178, 0.0010, 0.0307, -0.0093, 0.0297, 0.0025, 0.0144, + 0.0067, -0.0065], device='cuda:0'), grad: tensor([-0.0004, 0.0006, -0.0026, 0.0009, -0.0005, 0.0003, 0.0002, 0.0005, + 0.0001, 0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 214.44, cls_loss 0.0165 cls_loss_mapping 0.0267 cls_loss_causal 0.6946 re_mapping 0.0150 re_causal 0.0441 /// teacc 98.55 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0160, -0.0147, -0.0026, ..., -0.0018, 0.0056, -0.0048], + [ 0.0192, 0.0163, -0.0175, ..., 0.0275, 0.0063, -0.1545], + [-0.0131, 0.0181, 0.0158, ..., -0.0276, 0.0402, -0.0250], + ..., + [-0.0009, -0.0114, 0.0276, ..., -0.0489, -0.0603, 0.0170], + [-0.0106, 0.0116, -0.0042, ..., -0.0435, 0.0190, -0.0926], + [-0.0423, 0.0089, 0.0017, ..., -0.0600, -0.0378, -0.0440]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -2.0313e-04, + 1.1355e-04, 8.9547e-07], + [ 4.5169e-08, 0.0000e+00, 0.0000e+00, ..., 6.1750e-05, + 8.4639e-05, 6.2212e-06], + [ 3.3993e-08, 0.0000e+00, 0.0000e+00, ..., 3.5465e-05, + 1.3006e-04, 3.4347e-06], + ..., + [ 1.6810e-07, 0.0000e+00, 0.0000e+00, ..., 1.2606e-05, + 9.4116e-05, 4.4554e-05], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 5.5313e-05, + -9.3222e-04, 1.8906e-06], + [ 2.0023e-08, 0.0000e+00, 0.0000e+00, ..., 4.2059e-06, + 2.7871e-04, 2.2173e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0102, -0.0175, 0.0007, 0.0311, -0.0085, 0.0291, 0.0029, 0.0146, + 0.0064, -0.0071], device='cuda:0'), grad: tensor([-0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0012, + -0.0015, -0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 214.85, cls_loss 0.0145 cls_loss_mapping 0.0280 cls_loss_causal 0.6932 re_mapping 0.0152 re_causal 0.0441 /// teacc 98.76 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0161, -0.0146, -0.0026, ..., -0.0013, 0.0048, -0.0051], + [ 0.0199, 0.0163, -0.0175, ..., 0.0278, 0.0058, -0.1559], + [-0.0125, 0.0181, 0.0158, ..., -0.0280, 0.0406, -0.0254], + ..., + [-0.0011, -0.0114, 0.0276, ..., -0.0491, -0.0610, 0.0164], + [-0.0109, 0.0116, -0.0042, ..., -0.0439, 0.0196, -0.0932], + [-0.0431, 0.0088, 0.0017, ..., -0.0602, -0.0380, -0.0441]], + device='cuda:0'), grad: tensor([[-1.1269e-07, 0.0000e+00, 0.0000e+00, ..., 1.0407e-04, + 1.4997e-04, 1.7639e-06], + [ 7.3835e-06, 0.0000e+00, 0.0000e+00, ..., 6.7353e-05, + 1.4365e-04, 2.4185e-05], + [-9.7156e-06, 0.0000e+00, 0.0000e+00, ..., 1.8513e-04, + 5.1886e-05, 1.4409e-05], + ..., + [ 2.0750e-06, 0.0000e+00, 0.0000e+00, ..., 7.1973e-06, + 5.1677e-05, 7.2829e-06], + [ 1.0906e-06, 0.0000e+00, 0.0000e+00, ..., 1.4961e-04, + 2.1768e-04, 2.6021e-06], + [ 4.2934e-07, 0.0000e+00, 0.0000e+00, ..., 2.3052e-05, + 3.3110e-05, 1.4544e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0102, -0.0180, 0.0007, 0.0314, -0.0084, 0.0288, 0.0028, 0.0144, + 0.0065, -0.0068], device='cuda:0'), grad: tensor([ 0.0002, 0.0002, 0.0001, -0.0002, 0.0001, -0.0004, -0.0009, 0.0001, + 0.0005, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 214.93, cls_loss 0.0119 cls_loss_mapping 0.0219 cls_loss_causal 0.6532 re_mapping 0.0153 re_causal 0.0476 /// teacc 98.75 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0153, 0.0003, -0.0026, ..., -0.0012, 0.0041, -0.0054], + [ 0.0198, 0.0156, -0.0175, ..., 0.0275, 0.0051, -0.1569], + [-0.0124, 0.0180, 0.0159, ..., -0.0281, 0.0409, -0.0252], + ..., + [-0.0004, -0.0114, 0.0275, ..., -0.0492, -0.0618, 0.0161], + [-0.0112, 0.0104, -0.0042, ..., -0.0442, 0.0202, -0.0933], + [-0.0434, -0.0008, 0.0017, ..., -0.0604, -0.0380, -0.0451]], + device='cuda:0'), grad: tensor([[ 1.2107e-07, -9.9465e-07, 0.0000e+00, ..., -1.7826e-06, + 1.6704e-05, 2.1514e-07], + [ 1.0710e-07, 1.2107e-08, 0.0000e+00, ..., 2.1979e-06, + 1.4864e-05, 6.8359e-07], + [-1.6317e-06, 8.7544e-08, 0.0000e+00, ..., 2.7381e-06, + -4.6313e-05, 4.7963e-07], + ..., + [-1.3690e-07, 2.4214e-08, 0.0000e+00, ..., 3.6880e-07, + 1.4201e-05, -4.0606e-07], + [ 3.6974e-07, 3.0734e-08, 0.0000e+00, ..., 3.3583e-06, + -5.2571e-05, 4.9360e-07], + [ 1.5926e-07, 4.9733e-07, 0.0000e+00, ..., 5.0440e-06, + 1.2361e-05, 6.3360e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0099, -0.0184, 0.0005, 0.0319, -0.0080, 0.0285, 0.0027, 0.0144, + 0.0069, -0.0068], device='cuda:0'), grad: tensor([ 3.8505e-05, 2.0579e-05, 4.5508e-05, -1.3423e-04, -3.8266e-05, + 4.3720e-05, 3.2298e-06, -6.6638e-05, -4.4465e-05, 1.3185e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 214.82, cls_loss 0.0128 cls_loss_mapping 0.0227 cls_loss_causal 0.7097 re_mapping 0.0144 re_causal 0.0449 /// teacc 98.74 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0156, -0.0019, -0.0026, ..., -0.0010, 0.0037, -0.0058], + [ 0.0203, 0.0153, -0.0175, ..., 0.0268, 0.0044, -0.1563], + [-0.0100, 0.0179, 0.0159, ..., -0.0283, 0.0413, -0.0259], + ..., + [-0.0019, -0.0115, 0.0275, ..., -0.0494, -0.0622, 0.0160], + [-0.0120, 0.0101, -0.0042, ..., -0.0442, 0.0206, -0.0951], + [-0.0459, 0.0004, 0.0017, ..., -0.0603, -0.0380, -0.0463]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, 0.0000e+00, 0.0000e+00, ..., -1.3292e-04, + -2.4110e-05, 8.1863e-07], + [-6.6124e-07, 0.0000e+00, 0.0000e+00, ..., 6.1616e-06, + 5.4240e-06, 6.2644e-05], + [ 2.4214e-07, 0.0000e+00, 0.0000e+00, ..., 1.9148e-05, + 8.7768e-06, 2.4587e-06], + ..., + [ 1.5739e-07, 0.0000e+00, 0.0000e+00, ..., 2.8517e-06, + 2.8722e-06, 9.6112e-06], + [ 2.2445e-07, 0.0000e+00, 0.0000e+00, ..., 3.2157e-05, + 2.7880e-05, 1.1072e-05], + [ 1.3225e-07, 0.0000e+00, 0.0000e+00, ..., 9.8050e-05, + 3.1382e-05, 5.8985e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0092, -0.0179, 0.0008, 0.0319, -0.0078, 0.0283, 0.0027, 0.0139, + 0.0067, -0.0065], device='cuda:0'), grad: tensor([-2.0766e-04, 1.3101e-04, 3.7849e-05, 7.9274e-06, -1.4982e-03, + 3.0294e-05, -1.1522e-04, 1.8075e-05, 9.1374e-05, 1.5039e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 214.73, cls_loss 0.0102 cls_loss_mapping 0.0173 cls_loss_causal 0.6816 re_mapping 0.0142 re_causal 0.0440 /// teacc 98.65 lr 0.00010000 +Epoch 52, weight, value: tensor([[-1.5604e-02, -1.7682e-03, -2.6036e-03, ..., -1.3666e-04, + 3.8299e-03, -5.8003e-03], + [ 2.0632e-02, 1.5280e-02, -1.7541e-02, ..., 2.6685e-02, + 3.7885e-03, -1.5824e-01], + [-8.6888e-03, 1.7895e-02, 1.5938e-02, ..., -2.8979e-02, + 4.1930e-02, -2.5700e-02], + ..., + [-2.5019e-03, -1.1479e-02, 2.7466e-02, ..., -4.9577e-02, + -6.2743e-02, 1.6270e-02], + [-1.2240e-02, 1.0001e-02, -4.2081e-03, ..., -4.4389e-02, + 2.0932e-02, -9.4686e-02], + [-4.6414e-02, 2.6877e-04, 1.7246e-03, ..., -6.0675e-02, + -3.8472e-02, -4.6642e-02]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -3.9840e-04, + -7.3731e-05, -1.6198e-05], + [ 2.9802e-08, 0.0000e+00, 0.0000e+00, ..., 1.1744e-06, + 2.8670e-05, 3.1423e-06], + [ 5.8673e-08, 0.0000e+00, 0.0000e+00, ..., 4.8429e-05, + 5.2422e-05, 1.9014e-05], + ..., + [ 1.0338e-07, 0.0000e+00, 0.0000e+00, ..., 3.4682e-06, + 1.5235e-04, 5.2191e-06], + [ 1.5832e-08, 0.0000e+00, 0.0000e+00, ..., 5.6684e-05, + -2.9206e-04, 1.1370e-05], + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 2.1982e-04, + 8.8811e-05, 5.1670e-06]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0099, -0.0178, 0.0011, 0.0315, -0.0086, 0.0286, 0.0024, 0.0140, + 0.0066, -0.0064], device='cuda:0'), grad: tensor([-5.2166e-04, 1.7822e-04, 8.4496e-04, -1.1120e-03, -7.9334e-05, + 4.4316e-05, 1.1754e-04, 7.2145e-04, -7.5531e-04, 5.6171e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 214.91, cls_loss 0.0123 cls_loss_mapping 0.0209 cls_loss_causal 0.6756 re_mapping 0.0142 re_causal 0.0441 /// teacc 98.76 lr 0.00010000 +Epoch 53, weight, value: tensor([[-1.5591e-02, -1.8932e-03, -2.6036e-03, ..., 5.4966e-05, + 3.7565e-03, -6.2338e-03], + [ 2.0548e-02, 1.5252e-02, -1.7541e-02, ..., 2.6717e-02, + 3.6637e-03, -1.6061e-01], + [-8.2082e-03, 1.7889e-02, 1.5938e-02, ..., -2.9483e-02, + 4.3062e-02, -2.6643e-02], + ..., + [-1.9734e-03, -1.1480e-02, 2.7466e-02, ..., -4.9709e-02, + -6.3530e-02, 1.6824e-02], + [-1.2314e-02, 8.5633e-03, -4.2081e-03, ..., -4.4756e-02, + 2.0514e-02, -9.6491e-02], + [-4.6583e-02, 2.6242e-04, 1.7246e-03, ..., -6.0903e-02, + -3.8885e-02, -4.8476e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.1601e-04, + -3.9577e-04, 9.1922e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.3772e-07, + 3.7760e-05, 1.0297e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7688e-05, + 2.3961e-04, 9.7811e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3066e-06, + 3.7044e-05, 9.2909e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6499e-04, + 9.9540e-05, 4.2357e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3648e-05, + 1.0341e-04, 3.1255e-06]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0096, -0.0181, 0.0012, 0.0315, -0.0069, 0.0290, 0.0022, 0.0144, + 0.0060, -0.0075], device='cuda:0'), grad: tensor([-5.3453e-04, 1.5545e-04, 4.5943e-04, -7.9203e-04, -3.3855e-04, + 5.0455e-05, 6.4731e-05, 7.3051e-04, 1.3804e-04, 6.5982e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 214.86, cls_loss 0.0113 cls_loss_mapping 0.0197 cls_loss_causal 0.6856 re_mapping 0.0140 re_causal 0.0426 /// teacc 98.77 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0157, -0.0019, -0.0026, ..., 0.0002, 0.0032, -0.0066], + [ 0.0205, 0.0152, -0.0175, ..., 0.0267, 0.0034, -0.1622], + [-0.0084, 0.0179, 0.0159, ..., -0.0300, 0.0435, -0.0274], + ..., + [-0.0017, -0.0115, 0.0275, ..., -0.0499, -0.0648, 0.0166], + [-0.0124, 0.0085, -0.0042, ..., -0.0451, 0.0209, -0.0975], + [-0.0470, 0.0003, 0.0017, ..., -0.0611, -0.0393, -0.0489]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-08, + 1.1727e-05, 7.6368e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.1548e-06, + 2.1681e-05, 2.9616e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2755e-05, + 7.5474e-06, -1.5637e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5646e-06, + 5.5641e-05, 3.4552e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8522e-05, + -7.8440e-04, 2.9057e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.2867e-06, + 6.5088e-04, 6.1691e-06]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0094, -0.0178, 0.0009, 0.0309, -0.0073, 0.0299, 0.0023, 0.0142, + 0.0060, -0.0070], device='cuda:0'), grad: tensor([ 1.9699e-05, 9.0778e-05, 1.2529e-04, -9.0933e-04, 3.2640e-04, + 1.3685e-03, -3.6860e-04, 3.5810e-04, -4.5166e-03, 3.5076e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 53---------------------------------------------------- +epoch 53, time 230.77, cls_loss 0.0094 cls_loss_mapping 0.0181 cls_loss_causal 0.6865 re_mapping 0.0135 re_causal 0.0416 /// teacc 98.81 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0158, -0.0019, -0.0026, ..., 0.0007, 0.0029, -0.0066], + [ 0.0208, 0.0152, -0.0175, ..., 0.0266, 0.0026, -0.1636], + [-0.0084, 0.0179, 0.0159, ..., -0.0304, 0.0440, -0.0280], + ..., + [-0.0016, -0.0115, 0.0275, ..., -0.0499, -0.0654, 0.0165], + [-0.0125, 0.0085, -0.0042, ..., -0.0455, 0.0214, -0.0984], + [-0.0479, 0.0003, 0.0017, ..., -0.0618, -0.0398, -0.0492]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 3.4459e-07, + 3.3408e-05, 5.6773e-06], + [-8.1956e-08, 0.0000e+00, 0.0000e+00, ..., -3.0361e-06, + 1.0170e-05, 9.5144e-06], + [ 1.9558e-08, 0.0000e+00, 0.0000e+00, ..., 1.5333e-05, + 2.1592e-05, 1.3411e-05], + ..., + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 6.5900e-06, + 2.6584e-05, 1.5557e-05], + [ 1.6764e-08, 0.0000e+00, 0.0000e+00, ..., 1.6555e-05, + -1.3404e-05, 1.3016e-05], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 2.4259e-05, + 3.5286e-05, 3.3557e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0095, -0.0176, 0.0012, 0.0308, -0.0078, 0.0299, 0.0024, 0.0140, + 0.0060, -0.0069], device='cuda:0'), grad: tensor([ 4.7326e-05, -6.0499e-05, 1.6332e-04, -2.3508e-04, -1.1063e-04, + -7.4923e-05, 1.0645e-04, 1.0455e-04, 1.0890e-04, -4.9621e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 54---------------------------------------------------- +epoch 54, time 230.29, cls_loss 0.0092 cls_loss_mapping 0.0180 cls_loss_causal 0.6663 re_mapping 0.0132 re_causal 0.0411 /// teacc 98.82 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0156, -0.0019, -0.0026, ..., 0.0014, 0.0025, -0.0068], + [ 0.0208, 0.0152, -0.0175, ..., 0.0266, 0.0023, -0.1638], + [-0.0087, 0.0179, 0.0159, ..., -0.0307, 0.0444, -0.0280], + ..., + [-0.0008, -0.0115, 0.0275, ..., -0.0501, -0.0656, 0.0178], + [-0.0129, 0.0085, -0.0042, ..., -0.0460, 0.0216, -0.0996], + [-0.0483, 0.0003, 0.0017, ..., -0.0627, -0.0402, -0.0497]], + device='cuda:0'), grad: tensor([[-1.8962e-06, 0.0000e+00, 0.0000e+00, ..., -2.9609e-05, + 4.7505e-05, -2.3283e-08], + [ 1.0245e-07, 0.0000e+00, 0.0000e+00, ..., -2.9244e-07, + 3.9554e-04, 1.9774e-05], + [ 1.6764e-07, 0.0000e+00, 0.0000e+00, ..., 7.9647e-06, + -1.3094e-03, 1.2545e-06], + ..., + [ 8.9407e-08, 0.0000e+00, 0.0000e+00, ..., 3.4217e-06, + 3.7503e-04, 6.9141e-05], + [ 9.9652e-08, 0.0000e+00, 0.0000e+00, ..., 1.9461e-05, + 3.8648e-04, -1.5485e-04], + [ 2.7381e-07, 0.0000e+00, 0.0000e+00, ..., 8.6129e-06, + 2.4676e-05, 3.1918e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0100, -0.0179, 0.0010, 0.0308, -0.0083, 0.0299, 0.0024, 0.0148, + 0.0059, -0.0071], device='cuda:0'), grad: tensor([ 1.4037e-05, 8.4352e-04, -1.9970e-03, 1.0860e-04, 2.8300e-04, + -7.2360e-05, 1.3161e-04, 1.1282e-03, -6.0797e-04, 1.6749e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 214.82, cls_loss 0.0106 cls_loss_mapping 0.0208 cls_loss_causal 0.6701 re_mapping 0.0135 re_causal 0.0410 /// teacc 98.77 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0156, -0.0019, -0.0026, ..., 0.0014, 0.0018, -0.0069], + [ 0.0210, 0.0152, -0.0175, ..., 0.0270, 0.0020, -0.1660], + [-0.0087, 0.0179, 0.0159, ..., -0.0313, 0.0447, -0.0281], + ..., + [-0.0008, -0.0115, 0.0275, ..., -0.0504, -0.0657, 0.0175], + [-0.0130, 0.0085, -0.0042, ..., -0.0463, 0.0221, -0.1011], + [-0.0485, 0.0003, 0.0017, ..., -0.0630, -0.0406, -0.0501]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3202e-05, + 3.5375e-05, 1.6326e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.1791e-06, + 4.2677e-05, 2.5053e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0975e-05, + -1.3900e-04, 1.2945e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1202e-06, + 3.5197e-05, 1.1101e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3344e-05, + 2.9832e-05, 2.1718e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.7784e-06, + -4.6730e-05, -2.7359e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0093, -0.0177, 0.0004, 0.0307, -0.0086, 0.0300, 0.0027, 0.0160, + 0.0058, -0.0075], device='cuda:0'), grad: tensor([ 5.0627e-06, 4.1276e-05, -9.1553e-05, -3.9077e-04, 7.6354e-05, + -5.7268e-04, 8.6451e-04, 1.5783e-04, 5.5790e-04, -6.4850e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 214.20, cls_loss 0.0112 cls_loss_mapping 0.0199 cls_loss_causal 0.6509 re_mapping 0.0140 re_causal 0.0403 /// teacc 98.53 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0155, -0.0019, -0.0026, ..., 0.0020, 0.0010, -0.0058], + [ 0.0210, 0.0152, -0.0175, ..., 0.0267, 0.0014, -0.1669], + [-0.0088, 0.0179, 0.0159, ..., -0.0317, 0.0453, -0.0283], + ..., + [-0.0008, -0.0115, 0.0275, ..., -0.0506, -0.0666, 0.0171], + [-0.0130, 0.0080, -0.0042, ..., -0.0459, 0.0232, -0.1024], + [-0.0487, 0.0003, 0.0017, ..., -0.0635, -0.0409, -0.0508]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.4597e-06, + 2.2367e-05, 2.5276e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0641e-06, + 1.0237e-05, 3.9898e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3786e-06, + -1.8752e-04, -4.1574e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.5554e-07, + 5.6252e-06, -3.5409e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3135e-05, + 2.1875e-05, 2.7381e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8573e-06, + 5.5805e-06, 1.4156e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0096, -0.0177, 0.0004, 0.0304, -0.0081, 0.0298, 0.0022, 0.0162, + 0.0062, -0.0079], device='cuda:0'), grad: tensor([ 4.4852e-05, -2.3305e-05, -5.4646e-04, 4.7350e-04, 4.6074e-05, + 8.9884e-05, -1.7118e-04, -8.7768e-06, 5.3704e-05, 4.1246e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 214.31, cls_loss 0.0078 cls_loss_mapping 0.0157 cls_loss_causal 0.6950 re_mapping 0.0130 re_causal 0.0409 /// teacc 98.75 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0156, -0.0019, -0.0026, ..., 0.0024, 0.0010, -0.0063], + [ 0.0210, 0.0152, -0.0175, ..., 0.0264, 0.0004, -0.1685], + [-0.0088, 0.0179, 0.0159, ..., -0.0321, 0.0461, -0.0285], + ..., + [-0.0007, -0.0115, 0.0275, ..., -0.0507, -0.0674, 0.0174], + [-0.0131, 0.0080, -0.0042, ..., -0.0463, 0.0232, -0.1038], + [-0.0488, 0.0003, 0.0017, ..., -0.0639, -0.0414, -0.0513]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0647e-05, + 4.1425e-06, 4.7609e-06], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., -1.2554e-06, + 5.8532e-05, 3.2168e-06], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.5744e-06, + 1.4573e-05, 3.4682e-06], + ..., + [-2.1420e-08, 0.0000e+00, 0.0000e+00, ..., 1.4314e-06, + 1.3202e-05, 6.5677e-06], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -5.2899e-06, + -3.2473e-04, 3.2559e-06], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 2.1961e-06, + 1.1611e-04, 1.5211e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0101, -0.0179, 0.0007, 0.0306, -0.0084, 0.0297, 0.0027, 0.0162, + 0.0057, -0.0079], device='cuda:0'), grad: tensor([ 3.8370e-06, 6.4254e-05, 7.6532e-05, -7.6108e-06, -2.6703e-04, + 1.1069e-04, 3.3379e-05, 7.8022e-05, -5.3024e-04, 4.3797e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 214.38, cls_loss 0.0085 cls_loss_mapping 0.0153 cls_loss_causal 0.6456 re_mapping 0.0130 re_causal 0.0394 /// teacc 98.79 lr 0.00010000 +Epoch 60, weight, value: tensor([[-1.5569e-02, -1.9140e-03, -1.1108e-02, ..., 2.6590e-03, + 6.1448e-04, -6.8986e-03], + [ 2.1066e-02, 1.5187e-02, -1.8161e-02, ..., 2.6644e-02, + -1.0456e-04, -1.7079e-01], + [-8.7496e-03, 1.7886e-02, 1.8631e-02, ..., -3.2501e-02, + 4.7178e-02, -2.8823e-02], + ..., + [-7.1947e-04, -1.1485e-02, 1.9661e-02, ..., -5.0877e-02, + -6.8379e-02, 1.8281e-02], + [-1.3076e-02, 8.0391e-03, -8.1184e-03, ..., -4.6442e-02, + 2.3376e-02, -1.0460e-01], + [-4.8779e-02, 2.5053e-04, 1.6376e-03, ..., -6.4115e-02, + -4.1854e-02, -5.1713e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9856e-06, + 8.6874e-06, 2.4214e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3197e-06, + 1.3027e-03, 8.3167e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8722e-06, + -1.5697e-03, 1.0720e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9523e-07, + 8.2433e-05, 1.5246e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.3474e-06, + 3.5614e-05, 5.3830e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8953e-07, + 1.3731e-05, 3.9637e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0101, -0.0180, 0.0014, 0.0309, -0.0086, 0.0295, 0.0025, 0.0160, + 0.0055, -0.0079], device='cuda:0'), grad: tensor([ 1.2331e-05, 2.1782e-03, -2.6588e-03, 1.5354e-04, 8.0645e-05, + 8.2374e-05, -8.3089e-05, 1.3280e-04, 8.2076e-05, 1.8418e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 214.23, cls_loss 0.0082 cls_loss_mapping 0.0154 cls_loss_causal 0.6605 re_mapping 0.0127 re_causal 0.0379 /// teacc 98.66 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0164, -0.0019, -0.0131, ..., 0.0028, 0.0004, -0.0073], + [ 0.0210, 0.0152, -0.0186, ..., 0.0267, -0.0007, -0.1718], + [-0.0088, 0.0179, 0.0183, ..., -0.0328, 0.0478, -0.0310], + ..., + [-0.0007, -0.0115, 0.0213, ..., -0.0510, -0.0694, 0.0183], + [-0.0120, 0.0080, -0.0082, ..., -0.0468, 0.0236, -0.1054], + [-0.0490, 0.0002, 0.0014, ..., -0.0644, -0.0421, -0.0527]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.1897e-05, + 8.4996e-05, -2.3190e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.3048e-06, + 4.9844e-06, 1.7583e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.5607e-06, + 1.2651e-05, 2.8387e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6152e-06, + 2.0433e-06, 1.0096e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7761e-05, + 4.3474e-06, -2.9486e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.0988e-06, + 5.1670e-06, 7.6145e-06]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0097, -0.0179, 0.0009, 0.0311, -0.0084, 0.0295, 0.0026, 0.0165, + 0.0052, -0.0079], device='cuda:0'), grad: tensor([ 6.3777e-05, 3.5942e-05, 3.1233e-05, 3.9577e-04, 1.1787e-05, + -1.2374e-04, -1.1510e-04, -6.9189e-04, 8.1182e-05, 3.1114e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 214.22, cls_loss 0.0107 cls_loss_mapping 0.0181 cls_loss_causal 0.6738 re_mapping 0.0129 re_causal 0.0364 /// teacc 98.80 lr 0.00010000 +Epoch 62, weight, value: tensor([[-1.5928e-02, -1.9056e-03, -1.3076e-02, ..., 3.0189e-03, + 7.8417e-05, -4.5376e-03], + [ 2.1049e-02, 1.5181e-02, -1.8643e-02, ..., 2.6420e-02, + -1.8745e-03, -1.7288e-01], + [-8.7342e-03, 1.7886e-02, 1.8267e-02, ..., -3.3239e-02, + 4.8803e-02, -3.2231e-02], + ..., + [-1.4462e-05, -1.1486e-02, 2.1325e-02, ..., -5.1294e-02, + -7.0027e-02, 1.9072e-02], + [-1.2270e-02, 8.0165e-03, -8.2102e-03, ..., -4.7296e-02, + 2.3842e-02, -1.0756e-01], + [-5.0247e-02, 2.4880e-04, 1.4279e-03, ..., -6.4406e-02, + -4.2770e-02, -5.3086e-02]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -1.9372e-05, + 1.1683e-05, -2.9355e-06], + [ 2.7008e-08, 0.0000e+00, 0.0000e+00, ..., 1.1614e-06, + 8.2031e-06, 6.4969e-06], + [ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 7.6592e-06, + -1.7810e-04, -8.7172e-06], + ..., + [-1.3225e-07, 0.0000e+00, 0.0000e+00, ..., 7.3574e-07, + 2.1964e-05, 1.8716e-05], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., -1.0216e-04, + -2.3890e-04, 2.7522e-05], + [ 5.0291e-08, 0.0000e+00, 0.0000e+00, ..., 8.6352e-06, + 2.1279e-05, 5.8651e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0100, -0.0184, 0.0013, 0.0310, -0.0092, 0.0302, 0.0024, 0.0158, + 0.0052, -0.0071], device='cuda:0'), grad: tensor([-1.4767e-05, 1.3849e-06, -2.7394e-04, 2.1458e-04, -7.7188e-05, + -8.3160e-04, 1.1168e-03, 6.3419e-05, -1.3614e-04, -6.3002e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 214.42, cls_loss 0.0069 cls_loss_mapping 0.0137 cls_loss_causal 0.6144 re_mapping 0.0126 re_causal 0.0386 /// teacc 98.82 lr 0.00010000 +Epoch 63, weight, value: tensor([[-1.5847e-02, -1.9056e-03, -1.3076e-02, ..., 3.0183e-03, + -5.2993e-04, -4.4439e-03], + [ 2.1025e-02, 1.5181e-02, -1.8643e-02, ..., 2.6407e-02, + -2.6354e-03, -1.7319e-01], + [-8.5288e-03, 1.7886e-02, 1.8267e-02, ..., -3.3514e-02, + 4.9340e-02, -3.2394e-02], + ..., + [-7.6745e-05, -1.1486e-02, 2.1325e-02, ..., -5.1409e-02, + -7.0417e-02, 1.9167e-02], + [-1.1997e-02, 8.0165e-03, -8.2102e-03, ..., -4.7746e-02, + 2.4225e-02, -1.0810e-01], + [-5.1168e-02, 2.4880e-04, 1.4279e-03, ..., -6.4660e-02, + -4.3769e-02, -5.4183e-02]], device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, 0.0000e+00, ..., -8.7595e-04, + -1.1140e-04, 1.2424e-06], + [ 1.8626e-08, 0.0000e+00, 0.0000e+00, ..., 3.6620e-06, + 3.4664e-06, 1.3784e-06], + [ 1.3970e-08, 0.0000e+00, 0.0000e+00, ..., 1.9222e-05, + 2.2855e-06, 1.2117e-06], + ..., + [-2.3562e-07, 0.0000e+00, 0.0000e+00, ..., 5.3383e-06, + 5.3570e-06, 4.9882e-06], + [ 1.3970e-08, 0.0000e+00, 0.0000e+00, ..., 1.4715e-05, + -1.2182e-05, -3.2336e-06], + [ 9.4064e-08, 0.0000e+00, 0.0000e+00, ..., 6.3419e-05, + 1.8239e-05, 6.5863e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0097, -0.0182, 0.0010, 0.0311, -0.0091, 0.0303, 0.0027, 0.0156, + 0.0053, -0.0072], device='cuda:0'), grad: tensor([-5.2948e-03, -9.4175e-05, 1.3673e-04, 7.2598e-05, 7.8142e-05, + 3.5992e-03, 8.4496e-04, 3.9369e-05, 8.3327e-05, 5.3358e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.98, cls_loss 0.0067 cls_loss_mapping 0.0132 cls_loss_causal 0.6132 re_mapping 0.0120 re_causal 0.0386 /// teacc 98.78 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0166, -0.0019, -0.0131, ..., 0.0034, -0.0010, -0.0046], + [ 0.0210, 0.0152, -0.0186, ..., 0.0260, -0.0028, -0.1742], + [-0.0088, 0.0179, 0.0183, ..., -0.0338, 0.0499, -0.0317], + ..., + [ 0.0007, -0.0115, 0.0213, ..., -0.0515, -0.0712, 0.0191], + [-0.0117, 0.0080, -0.0082, ..., -0.0479, 0.0245, -0.1084], + [-0.0516, 0.0002, 0.0014, ..., -0.0649, -0.0443, -0.0556]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 0.0000e+00, 0.0000e+00, ..., 3.0577e-05, + 5.3555e-05, 4.1351e-06], + [ 1.1362e-07, 0.0000e+00, 0.0000e+00, ..., 6.0350e-07, + 3.7942e-06, 1.0796e-05], + [ 6.0536e-08, 0.0000e+00, 0.0000e+00, ..., 2.6654e-06, + -3.7670e-05, 8.8066e-06], + ..., + [-5.7928e-07, 0.0000e+00, 0.0000e+00, ..., 3.6694e-07, + 2.6017e-05, 1.2666e-05], + [ 9.4995e-08, 0.0000e+00, 0.0000e+00, ..., 5.4352e-06, + 1.1569e-04, 1.0557e-05], + [ 1.0058e-07, 0.0000e+00, 0.0000e+00, ..., 2.8238e-06, + 1.0091e-04, -1.1724e-04]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0098, -0.0178, 0.0008, 0.0312, -0.0083, 0.0300, 0.0029, 0.0156, + 0.0055, -0.0081], device='cuda:0'), grad: tensor([ 7.2837e-05, 3.1739e-05, -3.0503e-05, 1.1945e-04, 4.2224e-04, + -5.8794e-04, -3.3021e-05, 1.9300e-04, 4.5753e-04, -6.4564e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 214.81, cls_loss 0.0087 cls_loss_mapping 0.0161 cls_loss_causal 0.6450 re_mapping 0.0121 re_causal 0.0380 /// teacc 98.64 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0166, -0.0019, -0.0131, ..., 0.0037, -0.0011, -0.0040], + [ 0.0210, 0.0152, -0.0186, ..., 0.0259, -0.0035, -0.1759], + [-0.0088, 0.0179, 0.0183, ..., -0.0340, 0.0506, -0.0317], + ..., + [ 0.0009, -0.0115, 0.0213, ..., -0.0516, -0.0719, 0.0190], + [-0.0118, 0.0080, -0.0082, ..., -0.0481, 0.0247, -0.1086], + [-0.0517, 0.0002, 0.0014, ..., -0.0647, -0.0445, -0.0557]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 8.7544e-06, + 1.1779e-05, 6.2678e-07], + [ 6.1467e-08, 0.0000e+00, 0.0000e+00, ..., 1.2862e-06, + 5.6103e-06, 8.0839e-07], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 2.5984e-06, + -6.4485e-06, 1.1288e-06], + ..., + [-1.4249e-07, 0.0000e+00, 0.0000e+00, ..., 1.3486e-06, + 6.5155e-06, 3.7458e-06], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 8.2925e-06, + 1.2740e-05, 5.5041e-07], + [ 3.4459e-08, 0.0000e+00, 0.0000e+00, ..., 2.7455e-06, + 8.8662e-06, 1.7136e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0097, -0.0175, 0.0011, 0.0313, -0.0079, 0.0301, 0.0028, 0.0153, + 0.0052, -0.0084], device='cuda:0'), grad: tensor([ 2.2292e-05, -1.6078e-05, -1.3625e-06, 3.4869e-05, -3.6955e-05, + -1.2565e-04, 3.2276e-05, 5.4725e-06, 3.0637e-05, 5.4181e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 214.33, cls_loss 0.0097 cls_loss_mapping 0.0168 cls_loss_causal 0.6550 re_mapping 0.0125 re_causal 0.0360 /// teacc 98.76 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0169, -0.0019, -0.0131, ..., 0.0036, -0.0015, -0.0039], + [ 0.0210, 0.0152, -0.0186, ..., 0.0258, -0.0043, -0.1765], + [-0.0087, 0.0179, 0.0183, ..., -0.0345, 0.0510, -0.0325], + ..., + [ 0.0009, -0.0115, 0.0213, ..., -0.0518, -0.0731, 0.0183], + [-0.0116, 0.0080, -0.0082, ..., -0.0483, 0.0254, -0.1089], + [-0.0519, 0.0002, 0.0014, ..., -0.0650, -0.0453, -0.0566]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1906e-05, + 2.3589e-05, 3.3211e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8720e-07, + 2.9691e-06, -8.3804e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.6392e-07, + 4.3400e-06, 2.6897e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.5914e-07, + 4.8354e-06, 4.3720e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.4373e-06, + -1.6320e-04, 1.7285e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7772e-06, + 4.6045e-05, -2.9787e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0096, -0.0179, 0.0008, 0.0317, -0.0076, 0.0295, 0.0037, 0.0155, + 0.0052, -0.0085], device='cuda:0'), grad: tensor([ 6.4075e-05, 6.1417e-03, 2.1124e-04, 3.9482e-04, 2.1291e-04, + -3.6389e-05, -7.7367e-05, -6.7406e-03, -6.6161e-05, -1.0014e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 214.52, cls_loss 0.0085 cls_loss_mapping 0.0138 cls_loss_causal 0.6618 re_mapping 0.0119 re_causal 0.0374 /// teacc 98.66 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0170, -0.0019, -0.0132, ..., 0.0043, -0.0016, -0.0036], + [ 0.0221, 0.0152, -0.0186, ..., 0.0260, -0.0040, -0.1767], + [-0.0087, 0.0179, 0.0183, ..., -0.0348, 0.0511, -0.0319], + ..., + [ 0.0003, -0.0115, 0.0213, ..., -0.0520, -0.0742, 0.0184], + [-0.0118, 0.0080, -0.0082, ..., -0.0490, 0.0256, -0.1107], + [-0.0549, 0.0002, 0.0014, ..., -0.0654, -0.0459, -0.0563]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0973e-04, + 1.4089e-05, 5.7276e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.2627e-06, + 5.9567e-06, 1.2880e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7896e-05, + 4.9397e-06, 8.0559e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3882e-06, + 3.1758e-06, -7.2159e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7819e-05, + -1.6421e-05, 5.3458e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0759e-05, + 8.7768e-06, 5.7161e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0097, -0.0166, 0.0001, 0.0318, -0.0082, 0.0292, 0.0040, 0.0157, + 0.0048, -0.0087], device='cuda:0'), grad: tensor([ 5.8126e-04, 4.6402e-05, 8.3923e-05, -9.0599e-04, -2.1413e-05, + -1.2740e-05, 2.4691e-05, 2.1383e-05, 1.6344e-04, 2.0131e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.37, cls_loss 0.0113 cls_loss_mapping 0.0179 cls_loss_causal 0.6203 re_mapping 0.0124 re_causal 0.0347 /// teacc 98.64 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0173, -0.0019, -0.0138, ..., 0.0040, -0.0025, -0.0039], + [ 0.0232, 0.0152, -0.0189, ..., 0.0257, -0.0047, -0.1768], + [-0.0094, 0.0179, 0.0183, ..., -0.0353, 0.0517, -0.0322], + ..., + [ 0.0010, -0.0115, 0.0208, ..., -0.0523, -0.0755, 0.0191], + [-0.0128, 0.0080, -0.0084, ..., -0.0494, 0.0258, -0.1123], + [-0.0565, 0.0002, 0.0014, ..., -0.0648, -0.0466, -0.0583]], + device='cuda:0'), grad: tensor([[ 5.2154e-08, 0.0000e+00, 0.0000e+00, ..., -6.9812e-06, + 1.9297e-05, 1.0833e-05], + [ 2.3097e-06, 0.0000e+00, 0.0000e+00, ..., 2.0284e-06, + 3.1501e-05, 3.0369e-05], + [ 1.9278e-07, 0.0000e+00, 0.0000e+00, ..., 1.2219e-06, + 1.5073e-05, 1.9521e-05], + ..., + [-3.9898e-06, 0.0000e+00, 0.0000e+00, ..., 9.3132e-07, + 2.1532e-05, 4.3623e-06], + [ 5.6438e-07, 0.0000e+00, 0.0000e+00, ..., 4.5188e-06, + 3.5465e-05, 2.6405e-05], + [ 2.4121e-07, 0.0000e+00, 0.0000e+00, ..., 1.5702e-06, + 7.1764e-04, 5.2071e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0085, -0.0171, 0.0002, 0.0326, -0.0077, 0.0302, 0.0033, 0.0159, + 0.0047, -0.0092], device='cuda:0'), grad: tensor([ 1.1826e-04, 2.5702e-04, 2.3556e-04, -7.1526e-03, 3.8290e-04, + 3.7491e-05, 1.9312e-05, -7.7859e-06, 3.3593e-04, 5.7716e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.15, cls_loss 0.0079 cls_loss_mapping 0.0164 cls_loss_causal 0.6224 re_mapping 0.0115 re_causal 0.0349 /// teacc 98.71 lr 0.00010000 +Epoch 69, weight, value: tensor([[-1.7238e-02, -1.9006e-03, -1.4748e-02, ..., 4.7513e-03, + -2.5865e-03, -3.9697e-03], + [ 2.4412e-02, 1.5174e-02, -1.8955e-02, ..., 2.5404e-02, + -5.6061e-03, -1.7799e-01], + [-9.5482e-03, 1.7884e-02, 1.8164e-02, ..., -3.6062e-02, + 5.2127e-02, -3.2323e-02], + ..., + [-1.1510e-04, -1.1487e-02, 1.9757e-02, ..., -5.2650e-02, + -7.6401e-02, 2.0071e-02], + [-1.2333e-02, 8.0033e-03, -8.5257e-03, ..., -4.9144e-02, + 2.6646e-02, -1.1298e-01], + [-5.7329e-02, 2.4841e-04, 1.3499e-03, ..., -6.4649e-02, + -4.6971e-02, -5.8469e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.2861e-04, + 4.7016e-04, 9.5926e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8077e-06, + 8.3447e-06, 1.8859e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9087e-05, + 1.1668e-05, 3.9674e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2936e-06, + 4.5076e-06, -2.4159e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.1322e-05, + 2.5883e-05, 1.6484e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.5049e-05, + 1.7300e-05, 2.4326e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0086, -0.0168, 0.0002, 0.0327, -0.0079, 0.0305, 0.0025, 0.0153, + 0.0055, -0.0092], device='cuda:0'), grad: tensor([ 7.4959e-04, 6.4075e-05, 4.2707e-05, -8.2314e-05, 5.7489e-05, + 2.1458e-03, -3.0479e-03, -7.9393e-05, 9.0659e-05, 5.9545e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 68---------------------------------------------------- +epoch 68, time 230.60, cls_loss 0.0070 cls_loss_mapping 0.0136 cls_loss_causal 0.6409 re_mapping 0.0110 re_causal 0.0334 /// teacc 98.83 lr 0.00010000 +Epoch 70, weight, value: tensor([[-1.7206e-02, -1.9006e-03, -1.4760e-02, ..., 5.4426e-03, + -3.0568e-03, -3.9407e-03], + [ 2.5376e-02, 1.5174e-02, -1.8956e-02, ..., 2.5440e-02, + -5.9514e-03, -1.7985e-01], + [-9.5825e-03, 1.7884e-02, 1.8163e-02, ..., -3.6424e-02, + 5.2919e-02, -3.2722e-02], + ..., + [-1.7789e-04, -1.1487e-02, 1.9754e-02, ..., -5.2725e-02, + -7.7182e-02, 1.9831e-02], + [-1.2731e-02, 8.0033e-03, -8.5261e-03, ..., -4.9374e-02, + 2.6793e-02, -1.1333e-01], + [-5.8321e-02, 2.4841e-04, 1.3470e-03, ..., -6.5284e-02, + -4.7594e-02, -5.8522e-02]], device='cuda:0'), grad: tensor([[ 1.7574e-06, 0.0000e+00, 0.0000e+00, ..., -4.3884e-06, + 2.6494e-05, 9.6858e-08], + [-3.9339e-05, 0.0000e+00, 0.0000e+00, ..., -7.4469e-06, + 3.4928e-04, 3.9814e-07], + [-2.4855e-05, 0.0000e+00, 0.0000e+00, ..., 1.1045e-06, + -3.3498e-04, 3.5670e-07], + ..., + [ 1.8299e-05, 0.0000e+00, 0.0000e+00, ..., 1.6419e-06, + 5.5462e-05, 3.3434e-07], + [ 5.3905e-06, 0.0000e+00, 0.0000e+00, ..., 7.6666e-06, + -3.3617e-04, 3.0035e-07], + [ 2.8498e-06, 0.0000e+00, 0.0000e+00, ..., 2.4773e-06, + 4.8786e-05, 1.4938e-06]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0091, -0.0169, 0.0005, 0.0322, -0.0080, 0.0307, 0.0026, 0.0153, + 0.0055, -0.0092], device='cuda:0'), grad: tensor([ 5.4955e-05, -4.9710e-05, -3.9625e-04, 3.2377e-04, 3.1734e-04, + 7.3910e-05, 5.0366e-05, 1.7250e-04, -6.4516e-04, 9.9421e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 214.62, cls_loss 0.0067 cls_loss_mapping 0.0120 cls_loss_causal 0.6414 re_mapping 0.0110 re_causal 0.0353 /// teacc 98.79 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0187, -0.0019, -0.0154, ..., 0.0064, -0.0040, -0.0042], + [ 0.0257, 0.0152, -0.0194, ..., 0.0253, -0.0063, -0.1812], + [-0.0098, 0.0179, 0.0179, ..., -0.0366, 0.0535, -0.0331], + ..., + [ 0.0007, -0.0115, 0.0196, ..., -0.0529, -0.0779, 0.0201], + [-0.0127, 0.0080, -0.0076, ..., -0.0496, 0.0273, -0.1131], + [-0.0589, 0.0002, 0.0013, ..., -0.0656, -0.0479, -0.0591]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., -5.3123e-06, + 2.2560e-05, 1.4175e-06], + [-3.9395e-07, 0.0000e+00, 0.0000e+00, ..., 1.2228e-06, + -1.7300e-05, 1.1781e-06], + [ 1.8626e-08, 0.0000e+00, 0.0000e+00, ..., 6.7763e-06, + -8.1599e-05, 3.0100e-06], + ..., + [ 5.4948e-08, 0.0000e+00, 0.0000e+00, ..., 1.1586e-06, + 6.8486e-05, 2.7213e-06], + [ 1.1921e-07, 0.0000e+00, 0.0000e+00, ..., 5.8003e-06, + 2.6479e-05, 6.3218e-06], + [ 5.2154e-08, 0.0000e+00, 0.0000e+00, ..., 7.2680e-06, + 1.6376e-05, -1.0920e-04]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0093, -0.0167, 0.0005, 0.0318, -0.0076, 0.0304, 0.0022, 0.0155, + 0.0055, -0.0093], device='cuda:0'), grad: tensor([ 8.6486e-05, -3.0971e-04, -5.1796e-05, 3.5357e-04, 4.9829e-04, + -2.9826e-04, 9.1612e-05, 5.0217e-05, 2.4152e-04, -6.6280e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 214.50, cls_loss 0.0079 cls_loss_mapping 0.0144 cls_loss_causal 0.6221 re_mapping 0.0113 re_causal 0.0331 /// teacc 98.65 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0173, -0.0019, -0.0156, ..., 0.0063, -0.0054, -0.0041], + [ 0.0270, 0.0152, -0.0186, ..., 0.0249, -0.0069, -0.1826], + [-0.0103, 0.0179, 0.0178, ..., -0.0374, 0.0536, -0.0335], + ..., + [ 0.0004, -0.0115, 0.0194, ..., -0.0532, -0.0786, 0.0209], + [-0.0133, 0.0080, -0.0076, ..., -0.0497, 0.0284, -0.1141], + [-0.0596, 0.0002, 0.0013, ..., -0.0662, -0.0490, -0.0585]], + device='cuda:0'), grad: tensor([[ 5.4482e-07, 0.0000e+00, 0.0000e+00, ..., -4.2230e-05, + 6.3851e-06, 5.9418e-06], + [-5.3085e-06, 0.0000e+00, 0.0000e+00, ..., -4.5709e-06, + 1.2562e-05, 1.3754e-05], + [ 7.8045e-07, 0.0000e+00, 0.0000e+00, ..., 5.9307e-06, + -2.1458e-05, 4.4480e-06], + ..., + [ 1.2331e-06, 0.0000e+00, 0.0000e+00, ..., 5.5544e-06, + 1.5102e-05, -1.2018e-05], + [ 1.0431e-06, 0.0000e+00, 0.0000e+00, ..., 1.4871e-05, + -8.2180e-06, 3.0119e-06], + [ 1.7043e-07, 0.0000e+00, 0.0000e+00, ..., 6.5006e-06, + 5.5134e-06, 3.4779e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0082, -0.0170, 0.0002, 0.0322, -0.0079, 0.0305, 0.0030, 0.0153, + 0.0062, -0.0092], device='cuda:0'), grad: tensor([-4.7863e-05, 1.4153e-03, 1.9908e-04, 1.8072e-04, -1.3018e-04, + -1.7341e-06, 4.1693e-05, -2.5635e-03, 1.0681e-04, 7.9918e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 214.60, cls_loss 0.0065 cls_loss_mapping 0.0113 cls_loss_causal 0.6243 re_mapping 0.0111 re_causal 0.0334 /// teacc 98.80 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0187, -0.0019, -0.0163, ..., 0.0064, -0.0060, -0.0041], + [ 0.0274, 0.0152, -0.0193, ..., 0.0246, -0.0080, -0.1835], + [-0.0106, 0.0179, 0.0177, ..., -0.0379, 0.0539, -0.0336], + ..., + [ 0.0008, -0.0115, 0.0192, ..., -0.0534, -0.0791, 0.0210], + [-0.0125, 0.0080, -0.0067, ..., -0.0499, 0.0290, -0.1140], + [-0.0604, 0.0002, 0.0012, ..., -0.0665, -0.0497, -0.0587]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.6418e-05, + 3.3230e-05, 4.6985e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.5241e-06, + 4.5039e-06, 1.1390e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.6981e-06, + -3.9965e-05, 2.4345e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.4005e-07, + 1.5348e-05, 4.7088e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.5526e-05, + 1.9446e-05, 1.5888e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4755e-06, + 7.5884e-06, -4.2282e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0081, -0.0177, 0.0002, 0.0321, -0.0078, 0.0304, 0.0031, 0.0149, + 0.0068, -0.0088], device='cuda:0'), grad: tensor([ 7.2479e-05, 1.4782e-05, -4.5508e-05, 5.0992e-05, 2.9057e-05, + 2.0206e-04, -1.8179e-04, -1.6546e-04, 1.4806e-04, -1.2469e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.54, cls_loss 0.0077 cls_loss_mapping 0.0128 cls_loss_causal 0.6603 re_mapping 0.0108 re_causal 0.0328 /// teacc 98.82 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0191, -0.0019, -0.0165, ..., 0.0067, -0.0062, -0.0026], + [ 0.0295, 0.0152, -0.0194, ..., 0.0253, -0.0088, -0.1845], + [-0.0113, 0.0179, 0.0178, ..., -0.0381, 0.0551, -0.0341], + ..., + [ 0.0005, -0.0115, 0.0192, ..., -0.0536, -0.0789, 0.0203], + [-0.0136, 0.0080, -0.0067, ..., -0.0504, 0.0287, -0.1143], + [-0.0614, 0.0002, 0.0012, ..., -0.0670, -0.0501, -0.0588]], + device='cuda:0'), grad: tensor([[-2.2817e-08, 0.0000e+00, 0.0000e+00, ..., -8.9705e-06, + 4.0941e-06, 5.7137e-07], + [-2.7893e-07, 0.0000e+00, 0.0000e+00, ..., -1.2871e-06, + 8.8476e-07, 2.4438e-06], + [ 2.8405e-08, 0.0000e+00, 0.0000e+00, ..., 2.6580e-06, + 1.7760e-06, 3.1432e-07], + ..., + [ 3.7719e-08, 0.0000e+00, 0.0000e+00, ..., 6.7195e-07, + 1.2703e-06, 3.1721e-06], + [ 1.3877e-07, 0.0000e+00, 0.0000e+00, ..., 4.1425e-06, + -2.8824e-07, 2.1681e-06], + [ 2.5146e-08, 0.0000e+00, 0.0000e+00, ..., 3.2056e-06, + -4.2766e-06, -3.0342e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0088, -0.0173, 0.0006, 0.0323, -0.0084, 0.0297, 0.0039, 0.0154, + 0.0053, -0.0087], device='cuda:0'), grad: tensor([ 1.6302e-05, -1.5814e-06, 2.2352e-05, 9.6083e-05, 8.4877e-05, + -2.6658e-05, 5.5701e-05, 1.0394e-05, 2.6584e-04, -5.2357e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 73---------------------------------------------------- +epoch 73, time 231.86, cls_loss 0.0098 cls_loss_mapping 0.0185 cls_loss_causal 0.6327 re_mapping 0.0107 re_causal 0.0314 /// teacc 98.92 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0202, -0.0019, -0.0172, ..., 0.0063, -0.0060, -0.0027], + [ 0.0326, 0.0152, -0.0176, ..., 0.0250, -0.0092, -0.1850], + [-0.0116, 0.0179, 0.0178, ..., -0.0386, 0.0557, -0.0345], + ..., + [-0.0007, -0.0115, 0.0186, ..., -0.0540, -0.0792, 0.0200], + [-0.0162, 0.0080, -0.0068, ..., -0.0510, 0.0284, -0.1149], + [-0.0652, 0.0002, 0.0012, ..., -0.0675, -0.0506, -0.0594]], + device='cuda:0'), grad: tensor([[-2.9569e-07, 0.0000e+00, 0.0000e+00, ..., -3.3919e-06, + 5.0843e-05, 9.3319e-07], + [ 4.4927e-06, 0.0000e+00, 0.0000e+00, ..., 7.3612e-06, + 1.0389e-04, 2.9504e-06], + [ 1.6782e-06, 0.0000e+00, 0.0000e+00, ..., 3.6024e-06, + 2.5630e-06, 1.4110e-06], + ..., + [ 7.9395e-07, 0.0000e+00, 0.0000e+00, ..., 1.3821e-06, + 2.5913e-05, 3.1646e-06], + [ 4.2235e-07, 0.0000e+00, 0.0000e+00, ..., -4.4179e-04, + -4.4937e-03, 1.3299e-05], + [ 1.3849e-06, 0.0000e+00, 0.0000e+00, ..., 3.7774e-06, + 4.4674e-05, 2.7752e-04]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0079, -0.0177, 0.0002, 0.0327, -0.0087, 0.0295, 0.0050, 0.0162, + 0.0042, -0.0083], device='cuda:0'), grad: tensor([ 0.0001, 0.0004, -0.0004, 0.0013, -0.0007, 0.0007, 0.0049, 0.0002, + -0.0072, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 214.23, cls_loss 0.0096 cls_loss_mapping 0.0169 cls_loss_causal 0.6218 re_mapping 0.0111 re_causal 0.0320 /// teacc 98.68 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0207, -0.0019, -0.0172, ..., 0.0070, -0.0059, -0.0028], + [ 0.0312, 0.0152, -0.0176, ..., 0.0252, -0.0105, -0.1831], + [-0.0099, 0.0179, 0.0178, ..., -0.0395, 0.0564, -0.0362], + ..., + [-0.0009, -0.0115, 0.0185, ..., -0.0542, -0.0800, 0.0201], + [-0.0167, 0.0080, -0.0068, ..., -0.0512, 0.0295, -0.1147], + [-0.0668, 0.0002, 0.0012, ..., -0.0665, -0.0511, -0.0602]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5870e-05, + 2.9102e-05, 5.6982e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.2750e-05, + 5.6028e-04, 2.1175e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.9651e-05, + 2.6792e-05, 8.9169e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0446e-04, + 1.9515e-04, 6.8665e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.1940e-06, + -7.6723e-04, 9.0659e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.5936e-05, + 8.0347e-05, 1.4997e-04]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0079, -0.0172, 0.0003, 0.0322, -0.0077, 0.0298, 0.0038, 0.0159, + 0.0048, -0.0086], device='cuda:0'), grad: tensor([ 0.0005, 0.0015, 0.0008, 0.0013, 0.0008, -0.0123, 0.0003, 0.0065, + -0.0009, 0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 214.30, cls_loss 0.0062 cls_loss_mapping 0.0110 cls_loss_causal 0.6040 re_mapping 0.0106 re_causal 0.0333 /// teacc 98.76 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0214, -0.0019, -0.0173, ..., 0.0070, -0.0065, -0.0030], + [ 0.0314, 0.0152, -0.0177, ..., 0.0247, -0.0113, -0.1842], + [-0.0100, 0.0179, 0.0178, ..., -0.0398, 0.0565, -0.0366], + ..., + [-0.0009, -0.0115, 0.0185, ..., -0.0547, -0.0806, 0.0213], + [-0.0160, 0.0080, -0.0068, ..., -0.0516, 0.0300, -0.1163], + [-0.0673, 0.0002, 0.0012, ..., -0.0667, -0.0515, -0.0610]], + device='cuda:0'), grad: tensor([[ 5.0291e-08, 0.0000e+00, 0.0000e+00, ..., -3.5930e-06, + 8.2776e-06, 2.7800e-07], + [-8.4331e-07, 0.0000e+00, 0.0000e+00, ..., 5.8673e-08, + 2.1428e-05, 5.0012e-07], + [ 1.7323e-07, 0.0000e+00, 0.0000e+00, ..., 6.0117e-07, + -1.2536e-06, 1.1034e-05], + ..., + [ 1.2293e-07, 0.0000e+00, 0.0000e+00, ..., 1.3132e-07, + 2.2680e-05, 3.1991e-07], + [ 1.8487e-07, 0.0000e+00, 0.0000e+00, ..., 5.4343e-07, + -1.5926e-04, -1.6198e-05], + [ 5.2154e-08, 0.0000e+00, 0.0000e+00, ..., 6.5984e-07, + 1.4015e-05, 1.5637e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0076, -0.0174, 0.0001, 0.0319, -0.0077, 0.0309, 0.0042, 0.0159, + 0.0048, -0.0091], device='cuda:0'), grad: tensor([ 7.4625e-05, 3.5465e-05, 2.0111e-04, 2.8276e-04, 2.3112e-05, + 2.4438e-05, 5.5790e-05, -5.0592e-04, -2.5320e-04, 6.2466e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 214.29, cls_loss 0.0068 cls_loss_mapping 0.0114 cls_loss_causal 0.5680 re_mapping 0.0109 re_causal 0.0317 /// teacc 98.76 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0215, -0.0019, -0.0180, ..., 0.0082, -0.0069, -0.0031], + [ 0.0313, 0.0152, -0.0185, ..., 0.0237, -0.0122, -0.1852], + [-0.0103, 0.0179, 0.0176, ..., -0.0405, 0.0567, -0.0364], + ..., + [ 0.0004, -0.0115, 0.0181, ..., -0.0544, -0.0809, 0.0209], + [-0.0167, 0.0080, -0.0056, ..., -0.0516, 0.0306, -0.1165], + [-0.0698, 0.0002, 0.0011, ..., -0.0669, -0.0519, -0.0605]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, 0.0000e+00, 0.0000e+00, ..., 1.4715e-05, + 9.7230e-06, 5.5367e-07], + [-6.4913e-07, 0.0000e+00, 0.0000e+00, ..., -2.0593e-05, + 2.5667e-06, 3.4226e-07], + [ 2.5192e-07, 0.0000e+00, 0.0000e+00, ..., 7.2382e-06, + 1.3895e-06, 3.6042e-07], + ..., + [ 1.1129e-07, 0.0000e+00, 0.0000e+00, ..., 2.3544e-06, + 4.8382e-07, 1.0347e-06], + [ 1.2014e-07, 0.0000e+00, 0.0000e+00, ..., 1.7017e-05, + -6.7241e-06, 6.2259e-07], + [ 4.3306e-08, 0.0000e+00, 0.0000e+00, ..., 3.2242e-06, + 1.5935e-06, 3.4213e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0084, -0.0181, -0.0002, 0.0318, -0.0077, 0.0306, 0.0037, 0.0165, + 0.0049, -0.0088], device='cuda:0'), grad: tensor([ 5.8591e-05, -1.5962e-04, 3.9369e-05, 3.4750e-05, 2.2575e-06, + -1.2413e-05, -2.6315e-05, 1.3947e-05, 3.8862e-05, 1.0744e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 214.63, cls_loss 0.0065 cls_loss_mapping 0.0098 cls_loss_causal 0.6058 re_mapping 0.0105 re_causal 0.0323 /// teacc 98.83 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0226, -0.0019, -0.0189, ..., 0.0081, -0.0075, -0.0033], + [ 0.0315, 0.0152, -0.0188, ..., 0.0238, -0.0131, -0.1856], + [-0.0098, 0.0179, 0.0178, ..., -0.0410, 0.0576, -0.0366], + ..., + [-0.0010, -0.0115, 0.0177, ..., -0.0547, -0.0819, 0.0203], + [-0.0173, 0.0080, -0.0050, ..., -0.0523, 0.0308, -0.1177], + [-0.0718, 0.0002, 0.0011, ..., -0.0672, -0.0524, -0.0603]], + device='cuda:0'), grad: tensor([[ 2.1413e-05, 0.0000e+00, 0.0000e+00, ..., -1.4551e-05, + 3.7014e-05, 3.6180e-05], + [ 9.4920e-06, 0.0000e+00, 0.0000e+00, ..., -1.6754e-06, + 8.6129e-06, 1.8910e-05], + [ 6.0983e-06, 0.0000e+00, 0.0000e+00, ..., -5.4687e-06, + -2.0194e-04, 6.4634e-06], + ..., + [-5.5462e-05, 0.0000e+00, 0.0000e+00, ..., -2.8498e-06, + 3.8326e-05, -9.9599e-05], + [ 2.4270e-06, 0.0000e+00, 0.0000e+00, ..., 6.8955e-06, + 6.4731e-05, 6.8396e-06], + [ 3.5651e-06, 0.0000e+00, 0.0000e+00, ..., 6.0052e-06, + 2.4084e-06, 7.8306e-06]], device='cuda:0') +Epoch 79, bias, value: tensor([ 7.5956e-03, -1.7898e-02, -8.5355e-05, 3.1815e-02, -7.6049e-03, + 3.0820e-02, 3.8487e-03, 1.5939e-02, 4.2507e-03, -7.9647e-03], + device='cuda:0'), grad: tensor([ 1.3864e-04, 5.8383e-05, -2.2185e-04, 1.5140e-04, 6.2525e-05, + 3.6538e-05, 4.2707e-05, -4.9639e-04, 1.1688e-04, 1.1140e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 214.13, cls_loss 0.0071 cls_loss_mapping 0.0143 cls_loss_causal 0.6171 re_mapping 0.0113 re_causal 0.0327 /// teacc 98.74 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0231, -0.0019, -0.0250, ..., 0.0081, -0.0079, -0.0031], + [ 0.0321, 0.0152, -0.0194, ..., 0.0260, -0.0134, -0.1861], + [-0.0098, 0.0179, 0.0162, ..., -0.0416, 0.0583, -0.0368], + ..., + [-0.0013, -0.0115, 0.0123, ..., -0.0551, -0.0831, 0.0203], + [-0.0182, 0.0080, -0.0077, ..., -0.0539, 0.0307, -0.1187], + [-0.0731, 0.0002, 0.0023, ..., -0.0678, -0.0531, -0.0605]], + device='cuda:0'), grad: tensor([[-5.6252e-07, 0.0000e+00, 0.0000e+00, ..., -8.2403e-06, + 4.8757e-05, 2.5630e-05], + [ 6.4727e-08, 0.0000e+00, 0.0000e+00, ..., 1.2862e-06, + 1.0881e-03, 9.3207e-06], + [ 9.7323e-08, 0.0000e+00, 0.0000e+00, ..., 2.2165e-06, + -3.2520e-04, -3.1972e-04], + ..., + [ 2.6543e-08, 0.0000e+00, 0.0000e+00, ..., 3.9600e-06, + 1.3411e-04, 1.1015e-04], + [ 2.1420e-08, 0.0000e+00, 0.0000e+00, ..., 4.9733e-06, + -1.9817e-03, 1.6376e-05], + [ 6.3330e-08, 0.0000e+00, 0.0000e+00, ..., 5.5023e-06, + 7.1955e-04, 8.7991e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0074, -0.0175, 0.0005, 0.0316, -0.0071, 0.0314, 0.0033, 0.0157, + 0.0036, -0.0083], device='cuda:0'), grad: tensor([ 1.2815e-04, 4.7760e-03, -6.1750e-04, 2.6077e-05, 3.0684e-04, + 1.8227e-04, 2.1207e-04, 9.0361e-05, -7.1449e-03, 2.0409e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.27, cls_loss 0.0061 cls_loss_mapping 0.0119 cls_loss_causal 0.6139 re_mapping 0.0106 re_causal 0.0329 /// teacc 98.81 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0232, -0.0019, -0.0266, ..., 0.0083, -0.0083, -0.0032], + [ 0.0326, 0.0152, -0.0219, ..., 0.0255, -0.0142, -0.1862], + [-0.0098, 0.0179, 0.0153, ..., -0.0419, 0.0590, -0.0368], + ..., + [-0.0014, -0.0115, 0.0096, ..., -0.0554, -0.0843, 0.0203], + [-0.0188, 0.0079, -0.0086, ..., -0.0545, 0.0323, -0.1210], + [-0.0747, 0.0002, 0.0020, ..., -0.0684, -0.0539, -0.0606]], + device='cuda:0'), grad: tensor([[-6.6590e-08, 0.0000e+00, 0.0000e+00, ..., -3.0845e-06, + 4.1313e-06, 1.8952e-07], + [-2.7474e-08, 0.0000e+00, 0.0000e+00, ..., 1.2377e-06, + 3.6228e-06, 6.4261e-08], + [ 2.0489e-08, 0.0000e+00, 0.0000e+00, ..., 1.1241e-06, + -2.1055e-05, -1.3812e-06], + ..., + [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 1.5693e-07, + 1.3774e-06, 2.7288e-07], + [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 2.0087e-05, + 6.1989e-05, 4.3064e-06], + [ 6.9849e-09, 0.0000e+00, 0.0000e+00, ..., 9.7509e-07, + 2.7455e-06, -9.0199e-07]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0072, -0.0171, 0.0008, 0.0316, -0.0075, 0.0315, 0.0036, 0.0152, + 0.0044, -0.0086], device='cuda:0'), grad: tensor([ 2.6934e-06, 1.5289e-05, -9.3430e-06, 8.7738e-05, 1.7211e-05, + -3.6180e-05, -8.6308e-05, -5.8293e-05, 1.1563e-04, -4.8637e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 214.52, cls_loss 0.0065 cls_loss_mapping 0.0126 cls_loss_causal 0.5807 re_mapping 0.0107 re_causal 0.0314 /// teacc 98.74 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0232, -0.0019, -0.0285, ..., 0.0085, -0.0087, -0.0029], + [ 0.0326, 0.0152, -0.0248, ..., 0.0262, -0.0130, -0.1866], + [-0.0098, 0.0179, 0.0146, ..., -0.0427, 0.0595, -0.0368], + ..., + [-0.0013, -0.0115, 0.0085, ..., -0.0557, -0.0863, 0.0203], + [-0.0188, 0.0079, -0.0078, ..., -0.0547, 0.0323, -0.1214], + [-0.0749, 0.0002, 0.0020, ..., -0.0686, -0.0547, -0.0604]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., -1.4566e-05, + -2.5079e-05, 5.8673e-08], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -2.0266e-05, + -1.8075e-05, 1.7835e-07], + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 1.7256e-05, + 2.7731e-05, 1.4110e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6876e-06, + 2.3004e-06, 1.4435e-07], + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 4.3549e-06, + 6.3144e-06, 5.9605e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.4047e-06, + 3.4571e-06, 2.4065e-06]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0073, -0.0162, 0.0010, 0.0315, -0.0072, 0.0311, 0.0041, 0.0152, + 0.0034, -0.0090], device='cuda:0'), grad: tensor([-1.1420e-04, -2.1100e-04, 8.2874e-04, 6.2132e-04, 1.3518e-04, + 3.2365e-05, 6.1989e-05, -1.4362e-03, 7.0572e-05, 1.1928e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 214.36, cls_loss 0.0067 cls_loss_mapping 0.0123 cls_loss_causal 0.6098 re_mapping 0.0106 re_causal 0.0306 /// teacc 98.84 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0237, -0.0019, -0.0314, ..., 0.0088, -0.0088, -0.0031], + [ 0.0325, 0.0151, -0.0251, ..., 0.0262, -0.0135, -0.1881], + [-0.0100, 0.0179, 0.0141, ..., -0.0431, 0.0600, -0.0371], + ..., + [-0.0014, -0.0115, 0.0070, ..., -0.0561, -0.0872, 0.0209], + [-0.0174, 0.0079, -0.0086, ..., -0.0556, 0.0322, -0.1211], + [-0.0751, 0.0002, 0.0013, ..., -0.0686, -0.0553, -0.0607]], + device='cuda:0'), grad: tensor([[ 9.7323e-08, 0.0000e+00, 4.6566e-09, ..., -9.5665e-05, + 9.8813e-07, 7.2690e-07], + [ 2.4354e-07, 0.0000e+00, 1.5832e-08, ..., 1.4603e-04, + 5.0592e-04, 2.3618e-06], + [ 2.3376e-06, 0.0000e+00, 4.1910e-09, ..., -4.2245e-06, + -1.0449e-04, -1.3113e-06], + ..., + [ 3.1246e-07, 0.0000e+00, 5.4482e-08, ..., 5.3383e-06, + 1.6943e-05, 1.9781e-06], + [ 9.2201e-08, 0.0000e+00, 7.3109e-08, ..., 2.5272e-03, + 8.8120e-03, 1.9064e-06], + [ 1.4435e-07, 0.0000e+00, -2.6310e-07, ..., 6.3002e-05, + 4.1544e-05, 8.3566e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0072, -0.0165, 0.0007, 0.0322, -0.0076, 0.0311, 0.0041, 0.0158, + 0.0028, -0.0089], device='cuda:0'), grad: tensor([-1.8215e-04, 7.2670e-04, -1.6189e-04, 6.3062e-05, -2.5794e-05, + -1.5282e-02, 1.8253e-03, -1.0467e-04, 1.2741e-02, 3.9482e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 214.46, cls_loss 0.0065 cls_loss_mapping 0.0105 cls_loss_causal 0.5919 re_mapping 0.0096 re_causal 0.0287 /// teacc 98.91 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0238, -0.0019, -0.0336, ..., 0.0099, -0.0094, -0.0032], + [ 0.0325, 0.0151, -0.0227, ..., 0.0244, -0.0138, -0.1888], + [-0.0103, 0.0179, 0.0137, ..., -0.0435, 0.0606, -0.0366], + ..., + [-0.0009, -0.0115, 0.0052, ..., -0.0564, -0.0875, 0.0223], + [-0.0176, 0.0079, -0.0094, ..., -0.0563, 0.0318, -0.1217], + [-0.0754, 0.0002, 0.0018, ..., -0.0690, -0.0560, -0.0617]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.8818e-06, + 7.2755e-06, 2.8089e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.3726e-07, + 4.1649e-06, 5.6485e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0356e-06, + 2.7925e-05, 2.1601e-04], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.7323e-08, + 6.9797e-05, 4.6846e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9840e-06, + 1.0356e-05, 1.7118e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6822e-07, + 9.1344e-06, 1.2867e-05]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0079, -0.0168, 0.0007, 0.0320, -0.0067, 0.0315, 0.0041, 0.0155, + 0.0026, -0.0096], device='cuda:0'), grad: tensor([ 2.4438e-05, -1.2410e-04, 2.7037e-04, 4.6921e-04, -5.5695e-04, + -4.5395e-04, 1.7717e-05, 2.7752e-04, 9.5129e-05, -1.8597e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 214.30, cls_loss 0.0053 cls_loss_mapping 0.0099 cls_loss_causal 0.5955 re_mapping 0.0099 re_causal 0.0298 /// teacc 98.90 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0240, -0.0018, -0.0365, ..., 0.0104, -0.0098, -0.0026], + [ 0.0325, 0.0151, -0.0188, ..., 0.0242, -0.0140, -0.1892], + [-0.0103, 0.0179, 0.0135, ..., -0.0439, 0.0614, -0.0371], + ..., + [-0.0009, -0.0115, 0.0024, ..., -0.0566, -0.0885, 0.0230], + [-0.0176, 0.0076, -0.0117, ..., -0.0561, 0.0319, -0.1230], + [-0.0756, 0.0002, 0.0018, ..., -0.0694, -0.0565, -0.0617]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.7907e-06, + -3.9399e-05, -3.3116e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.6310e-07, + 2.7660e-06, 2.1718e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.8406e-07, + 1.1519e-05, 3.7640e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.3272e-07, + 3.5409e-06, 1.3638e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1383e-06, + -2.3317e-04, 4.9204e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5274e-06, + 2.2471e-04, -9.0897e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0083, -0.0175, 0.0008, 0.0326, -0.0071, 0.0313, 0.0038, 0.0159, + 0.0025, -0.0094], device='cuda:0'), grad: tensor([-6.7472e-04, 1.3679e-05, 1.1408e-04, 6.3360e-05, 2.2423e-04, + 4.0382e-05, 1.8847e-04, 1.3132e-03, -6.9571e-04, -5.8746e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 214.44, cls_loss 0.0045 cls_loss_mapping 0.0092 cls_loss_causal 0.5685 re_mapping 0.0098 re_causal 0.0296 /// teacc 98.85 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0240, -0.0018, -0.0380, ..., 0.0104, -0.0101, -0.0020], + [ 0.0325, 0.0150, -0.0179, ..., 0.0241, -0.0143, -0.1895], + [-0.0104, 0.0179, 0.0143, ..., -0.0444, 0.0618, -0.0376], + ..., + [-0.0004, -0.0115, 0.0016, ..., -0.0568, -0.0890, 0.0230], + [-0.0177, 0.0075, -0.0125, ..., -0.0565, 0.0321, -0.1237], + [-0.0756, 0.0002, 0.0015, ..., -0.0697, -0.0574, -0.0621]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0505e-05, + 9.5833e-07, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2526e-07, + 9.4436e-07, 1.1269e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.7707e-07, + 2.0210e-06, 9.8720e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6391e-07, + 7.1526e-07, 2.2864e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.0199e-07, + -1.0329e-04, 3.6787e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3581e-06, + 1.2495e-05, 6.5845e-07]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0083, -0.0176, 0.0010, 0.0327, -0.0072, 0.0315, 0.0041, 0.0158, + 0.0025, -0.0097], device='cuda:0'), grad: tensor([-2.1607e-05, 6.1691e-06, 6.4559e-06, 1.2600e-04, 1.5117e-05, + 1.2815e-05, 3.8177e-05, -8.3596e-06, -1.6272e-04, -1.1981e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 85---------------------------------------------------- +epoch 85, time 230.99, cls_loss 0.0048 cls_loss_mapping 0.0106 cls_loss_causal 0.6130 re_mapping 0.0100 re_causal 0.0307 /// teacc 98.98 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0240, -0.0018, -0.0381, ..., 0.0106, -0.0102, -0.0021], + [ 0.0325, 0.0150, -0.0175, ..., 0.0239, -0.0149, -0.1899], + [-0.0104, 0.0179, 0.0142, ..., -0.0448, 0.0623, -0.0377], + ..., + [-0.0005, -0.0115, 0.0015, ..., -0.0570, -0.0896, 0.0229], + [-0.0177, 0.0074, -0.0127, ..., -0.0563, 0.0324, -0.1240], + [-0.0756, 0.0002, 0.0015, ..., -0.0699, -0.0580, -0.0625]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -2.3339e-06, + 2.8059e-05, 1.7695e-07], + [ 0.0000e+00, 0.0000e+00, 1.8161e-08, ..., 1.1727e-05, + 1.9193e-05, 3.5297e-07], + [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 5.6922e-06, + 1.9714e-05, 4.0559e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 7.7952e-07, ..., 5.5274e-07, + 2.8405e-06, 2.5192e-07], + [ 0.0000e+00, 0.0000e+00, 5.1223e-08, ..., 3.8356e-05, + -7.3254e-05, 2.0908e-07], + [ 0.0000e+00, 0.0000e+00, -9.7603e-07, ..., 4.8541e-06, + 1.0990e-05, 9.7416e-07]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0083, -0.0182, 0.0010, 0.0325, -0.0070, 0.0318, 0.0043, 0.0162, + 0.0026, -0.0100], device='cuda:0'), grad: tensor([ 2.6494e-05, 9.2447e-05, 1.0347e-04, -8.8215e-04, -1.4998e-05, + 2.1482e-04, -1.7250e-04, 5.1498e-04, 1.8224e-05, 9.8765e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 214.63, cls_loss 0.0050 cls_loss_mapping 0.0096 cls_loss_causal 0.5969 re_mapping 0.0094 re_causal 0.0293 /// teacc 98.93 lr 0.00010000 +Epoch 88, weight, value: tensor([[-2.4045e-02, 6.0286e-05, -3.8161e-02, ..., 1.0878e-02, + -1.0894e-02, -2.1233e-03], + [ 3.2484e-02, 1.3429e-02, -1.7501e-02, ..., 2.4416e-02, + -1.5517e-02, -1.9029e-01], + [-1.0455e-02, 1.7681e-02, 1.4187e-02, ..., -4.5611e-02, + 6.2936e-02, -3.6689e-02], + ..., + [-1.0260e-04, -1.1694e-02, 1.4155e-03, ..., -5.7610e-02, + -9.0238e-02, 2.2545e-02], + [-1.7547e-02, 4.5833e-03, -1.2814e-02, ..., -5.6170e-02, + 3.3264e-02, -1.2426e-01], + [-7.5678e-02, -6.5520e-04, 1.5733e-03, ..., -7.0434e-02, + -5.8796e-02, -6.2800e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0736e-05, + 2.7381e-06, 1.3504e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.5949e-07, + 6.6683e-07, 1.5507e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1194e-06, + 1.0274e-05, 8.9873e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0932e-07, + 1.2822e-05, 1.8254e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9384e-07, + -2.8223e-05, 1.3504e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1090e-06, + 1.0226e-06, 1.9046e-06]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0083, -0.0192, 0.0009, 0.0325, -0.0072, 0.0319, 0.0041, 0.0167, + 0.0032, -0.0098], device='cuda:0'), grad: tensor([-1.0781e-05, -9.4902e-07, 1.9267e-05, 3.5893e-06, 2.0489e-06, + -3.1088e-06, 1.0513e-05, 2.3603e-05, -3.1352e-05, -1.2837e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 214.50, cls_loss 0.0069 cls_loss_mapping 0.0123 cls_loss_causal 0.5881 re_mapping 0.0095 re_causal 0.0287 /// teacc 98.80 lr 0.00010000 +Epoch 89, weight, value: tensor([[-2.6043e-02, 3.7143e-04, -3.8220e-02, ..., 1.1324e-02, + -1.1646e-02, -2.1896e-03], + [ 3.2594e-02, 1.2937e-02, -1.7353e-02, ..., 2.4447e-02, + -1.6228e-02, -1.9058e-01], + [-1.0277e-02, 1.7776e-02, 1.4158e-02, ..., -4.6358e-02, + 6.3708e-02, -3.6303e-02], + ..., + [-1.6251e-04, -1.1774e-02, 1.3448e-03, ..., -5.7997e-02, + -9.0856e-02, 2.2416e-02], + [-1.5837e-02, 3.1783e-03, -1.2859e-02, ..., -5.6959e-02, + 3.4906e-02, -1.2475e-01], + [-7.5843e-02, -7.7885e-04, 1.5771e-03, ..., -7.1726e-02, + -6.0967e-02, -6.2492e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4982e-03, + 9.5558e-04, 1.0272e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.8587e-06, + 2.9907e-05, 6.5519e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.2724e-05, + -2.2367e-05, 6.6683e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9488e-06, + 1.3532e-06, -2.0657e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1439e-05, + 2.8964e-06, 2.3679e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5187e-04, + 6.7830e-05, 2.0936e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0085, -0.0187, 0.0014, 0.0325, -0.0077, 0.0307, 0.0044, 0.0161, + 0.0055, -0.0107], device='cuda:0'), grad: tensor([ 2.7504e-03, -4.0150e-04, 1.6916e-04, 2.9206e-04, 2.2197e-04, + 8.1778e-05, -3.4676e-03, -2.9191e-05, 6.8843e-05, 3.1090e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 214.34, cls_loss 0.0048 cls_loss_mapping 0.0110 cls_loss_causal 0.5940 re_mapping 0.0098 re_causal 0.0306 /// teacc 98.91 lr 0.00010000 +Epoch 90, weight, value: tensor([[-2.5820e-02, 3.8353e-04, -3.8229e-02, ..., 1.2556e-02, + -1.2130e-02, -1.8754e-03], + [ 3.2646e-02, 1.2911e-02, -1.7361e-02, ..., 2.6358e-02, + -1.5865e-02, -1.9105e-01], + [-1.0334e-02, 1.7818e-02, 1.4152e-02, ..., -4.6718e-02, + 6.4342e-02, -3.6447e-02], + ..., + [-8.8909e-05, -1.1808e-02, 1.3427e-03, ..., -5.8261e-02, + -9.1204e-02, 2.2884e-02], + [-1.5873e-02, 3.0592e-03, -1.2856e-02, ..., -5.8163e-02, + 3.4494e-02, -1.2520e-01], + [-7.5976e-02, -7.8098e-04, 1.5904e-03, ..., -7.2031e-02, + -6.1460e-02, -6.2914e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.2806e-07, ..., -5.1633e-06, + 4.5657e-05, 1.1059e-07], + [ 0.0000e+00, 0.0000e+00, 2.5146e-06, ..., 1.4454e-06, + 1.7896e-05, 5.4948e-06], + [ 0.0000e+00, 0.0000e+00, 2.1998e-06, ..., 1.6978e-06, + -3.9196e-04, 2.4564e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.2678e-07, ..., 4.6287e-07, + 1.2793e-05, 3.9376e-06], + [ 0.0000e+00, 0.0000e+00, -1.2167e-05, ..., 1.3493e-05, + 6.6340e-05, 4.3996e-06], + [ 0.0000e+00, 0.0000e+00, 6.0163e-07, ..., 1.8328e-06, + 6.7987e-06, 4.1537e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0094, -0.0179, 0.0016, 0.0322, -0.0077, 0.0311, 0.0033, 0.0163, + 0.0046, -0.0108], device='cuda:0'), grad: tensor([ 7.2539e-05, 6.0976e-05, -7.7248e-04, 4.1509e-04, -1.0175e-04, + 1.1718e-04, -4.6045e-05, 5.6684e-05, 1.7238e-04, 2.4617e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 89---------------------------------------------------- +epoch 89, time 230.47, cls_loss 0.0040 cls_loss_mapping 0.0085 cls_loss_causal 0.6025 re_mapping 0.0089 re_causal 0.0291 /// teacc 98.99 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0256, 0.0004, -0.0384, ..., 0.0127, -0.0126, -0.0008], + [ 0.0327, 0.0129, -0.0175, ..., 0.0261, -0.0169, -0.1915], + [-0.0104, 0.0178, 0.0140, ..., -0.0473, 0.0647, -0.0363], + ..., + [ 0.0003, -0.0118, 0.0013, ..., -0.0583, -0.0916, 0.0227], + [-0.0160, 0.0030, -0.0124, ..., -0.0583, 0.0349, -0.1262], + [-0.0764, -0.0008, 0.0015, ..., -0.0722, -0.0617, -0.0628]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0841e-06, + 1.4268e-06, 4.5612e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5111e-07, + -2.4751e-05, 2.2158e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.3889e-07, + -3.0082e-06, 1.6401e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.0094e-08, + 3.6284e-06, 6.6720e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.9381e-06, + 5.0306e-05, 7.5512e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9116e-07, + 2.8238e-06, 1.9193e-05]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0096, -0.0178, 0.0013, 0.0319, -0.0077, 0.0314, 0.0026, 0.0162, + 0.0046, -0.0102], device='cuda:0'), grad: tensor([ 1.1049e-05, -9.5591e-06, 3.5167e-05, 3.4451e-04, -2.4486e-04, + -3.7044e-05, 1.0476e-05, -8.2827e-04, 2.0218e-04, 5.1641e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 214.47, cls_loss 0.0048 cls_loss_mapping 0.0091 cls_loss_causal 0.5793 re_mapping 0.0088 re_causal 0.0258 /// teacc 98.99 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0256, 0.0004, -0.0384, ..., 0.0126, -0.0132, -0.0009], + [ 0.0327, 0.0128, -0.0175, ..., 0.0259, -0.0174, -0.1919], + [-0.0104, 0.0179, 0.0140, ..., -0.0483, 0.0652, -0.0368], + ..., + [ 0.0002, -0.0118, 0.0013, ..., -0.0587, -0.0923, 0.0234], + [-0.0160, 0.0029, -0.0124, ..., -0.0587, 0.0351, -0.1267], + [-0.0765, -0.0008, 0.0015, ..., -0.0725, -0.0621, -0.0633]], + device='cuda:0'), grad: tensor([[ 3.6554e-08, 0.0000e+00, 0.0000e+00, ..., 2.4080e-05, + 1.9386e-05, 2.2585e-08], + [-7.1293e-07, 0.0000e+00, 0.0000e+00, ..., 4.9779e-07, + 5.0012e-07, 1.0966e-07], + [ 1.5250e-07, 0.0000e+00, 0.0000e+00, ..., 5.8766e-07, + -1.4435e-07, 5.7276e-08], + ..., + [ 2.3260e-07, 0.0000e+00, 0.0000e+00, ..., 1.4110e-07, + 1.4529e-07, 2.1351e-07], + [ 1.0082e-07, 0.0000e+00, 0.0000e+00, ..., 1.4104e-05, + 1.2830e-05, 5.2853e-08], + [ 3.0501e-08, 0.0000e+00, 0.0000e+00, ..., 1.4147e-06, + 1.0561e-06, 5.4063e-07]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0094, -0.0179, 0.0013, 0.0319, -0.0076, 0.0317, 0.0027, 0.0166, + 0.0041, -0.0105], device='cuda:0'), grad: tensor([ 3.1084e-05, -4.2111e-05, 1.1362e-05, -3.1982e-06, 1.4842e-05, + 1.5676e-05, -6.9559e-05, -4.5672e-06, 4.7535e-05, -1.1884e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 214.10, cls_loss 0.0044 cls_loss_mapping 0.0092 cls_loss_causal 0.5948 re_mapping 0.0092 re_causal 0.0285 /// teacc 98.85 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0254, 0.0004, -0.0388, ..., 0.0127, -0.0142, -0.0010], + [ 0.0326, 0.0128, -0.0175, ..., 0.0255, -0.0179, -0.1923], + [-0.0100, 0.0179, 0.0139, ..., -0.0490, 0.0657, -0.0364], + ..., + [-0.0002, -0.0118, 0.0009, ..., -0.0591, -0.0945, 0.0233], + [-0.0165, 0.0029, -0.0131, ..., -0.0598, 0.0352, -0.1276], + [-0.0771, -0.0008, 0.0014, ..., -0.0728, -0.0625, -0.0652]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3458e-06, + 4.1910e-06, 4.0559e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6287e-07, + 2.9765e-06, 2.9872e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.2305e-07, + 8.5458e-06, 9.5554e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.1491e-09, + 2.6859e-06, -2.8163e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4557e-06, + -6.2585e-06, 3.6391e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.3796e-08, + 1.5497e-06, 1.0347e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0094, -0.0179, 0.0012, 0.0317, -0.0067, 0.0315, 0.0041, 0.0157, + 0.0041, -0.0109], device='cuda:0'), grad: tensor([ 1.6779e-05, 1.9774e-05, 6.0856e-05, -4.5061e-04, 2.0996e-05, + 3.4451e-04, -1.6894e-06, -2.7969e-05, 9.8944e-06, 6.9141e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 214.49, cls_loss 0.0046 cls_loss_mapping 0.0083 cls_loss_causal 0.5993 re_mapping 0.0088 re_causal 0.0282 /// teacc 98.88 lr 0.00010000 +Epoch 94, weight, value: tensor([[-2.5531e-02, 3.7669e-04, -3.8857e-02, ..., 1.2643e-02, + -1.4657e-02, -1.1499e-03], + [ 3.2577e-02, 1.2690e-02, -1.7246e-02, ..., 2.5422e-02, + -1.8753e-02, -1.9236e-01], + [-1.0200e-02, 1.7993e-02, 1.3831e-02, ..., -4.9359e-02, + 6.6858e-02, -3.6859e-02], + ..., + [ 5.0469e-05, -1.1882e-02, 7.6131e-04, ..., -5.9456e-02, + -9.4878e-02, 2.5345e-02], + [-1.6533e-02, 2.7472e-03, -1.3100e-02, ..., -6.0227e-02, + 3.5041e-02, -1.2816e-01], + [-7.7844e-02, -7.8318e-04, 1.4251e-03, ..., -7.1927e-02, + -6.2247e-02, -6.5308e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.4661e-05, + 3.2708e-06, 1.2228e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.3889e-07, + 3.5763e-06, 5.6438e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.5560e-06, + -7.3433e-05, 5.8254e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0117e-06, + 7.1600e-06, 4.9034e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.8855e-06, + 5.0180e-06, 2.0470e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0997e-05, + 7.0810e-05, 3.6597e-04]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0094, -0.0178, 0.0009, 0.0315, -0.0075, 0.0316, 0.0041, 0.0167, + 0.0037, -0.0106], device='cuda:0'), grad: tensor([-4.1544e-05, 1.4879e-05, -1.2082e-04, 3.8177e-05, -1.5125e-03, + 2.5943e-05, 1.3649e-05, -1.0997e-04, 4.6253e-05, 1.6451e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 214.43, cls_loss 0.0055 cls_loss_mapping 0.0096 cls_loss_causal 0.5908 re_mapping 0.0085 re_causal 0.0256 /// teacc 98.97 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0266, 0.0004, -0.0389, ..., 0.0127, -0.0153, -0.0010], + [ 0.0327, 0.0122, -0.0172, ..., 0.0258, -0.0189, -0.1928], + [-0.0105, 0.0181, 0.0138, ..., -0.0499, 0.0675, -0.0376], + ..., + [-0.0003, -0.0120, 0.0008, ..., -0.0596, -0.0954, 0.0256], + [-0.0154, 0.0013, -0.0131, ..., -0.0610, 0.0345, -0.1295], + [-0.0786, -0.0008, 0.0014, ..., -0.0721, -0.0626, -0.0662]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1064e-06, + 1.0818e-05, 5.3905e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0775e-06, + 1.0245e-05, 2.0102e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3066e-06, + 5.0187e-05, 2.8729e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.6846e-08, + 2.6152e-06, -7.1466e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0536e-06, + -1.4031e-04, 3.6005e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.1863e-07, + 1.5795e-05, 4.1723e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0089, -0.0173, 0.0008, 0.0318, -0.0072, 0.0316, 0.0043, 0.0165, + 0.0031, -0.0107], device='cuda:0'), grad: tensor([ 6.2525e-05, 1.5593e-04, 2.2459e-04, 7.9095e-05, -6.2466e-05, + 2.1720e-04, -2.2322e-05, -5.0020e-04, -2.6083e-04, 1.0705e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 214.54, cls_loss 0.0041 cls_loss_mapping 0.0084 cls_loss_causal 0.6257 re_mapping 0.0089 re_causal 0.0278 /// teacc 98.91 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0263, 0.0004, -0.0389, ..., 0.0129, -0.0151, -0.0008], + [ 0.0327, 0.0120, -0.0172, ..., 0.0256, -0.0193, -0.1927], + [-0.0105, 0.0184, 0.0138, ..., -0.0502, 0.0683, -0.0381], + ..., + [-0.0004, -0.0121, 0.0008, ..., -0.0599, -0.0958, 0.0259], + [-0.0154, 0.0004, -0.0131, ..., -0.0613, 0.0344, -0.1301], + [-0.0787, -0.0008, 0.0014, ..., -0.0724, -0.0639, -0.0664]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.9912e-06, + 4.1313e-06, 2.3306e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.4389e-07, + 2.8554e-06, 3.9139e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.0012e-07, + -5.7340e-05, 3.2783e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8603e-07, + 3.3498e-05, 1.3029e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3916e-06, + 8.0541e-06, 9.3831e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.1945e-07, + 3.7570e-06, 2.8294e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0093, -0.0170, 0.0009, 0.0321, -0.0074, 0.0314, 0.0047, 0.0164, + 0.0026, -0.0107], device='cuda:0'), grad: tensor([ 3.4180e-06, 1.3765e-06, -5.4449e-05, -4.7207e-05, -4.3772e-06, + 1.7971e-05, 1.4482e-06, 3.3379e-05, 1.8477e-05, 2.9966e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 214.28, cls_loss 0.0048 cls_loss_mapping 0.0107 cls_loss_causal 0.6157 re_mapping 0.0088 re_causal 0.0275 /// teacc 98.96 lr 0.00010000 +Epoch 97, weight, value: tensor([[-2.6379e-02, 3.7092e-04, -3.8925e-02, ..., 1.2854e-02, + -1.5872e-02, -1.1195e-03], + [ 3.2735e-02, 1.2013e-02, -1.7248e-02, ..., 2.5669e-02, + -1.9368e-02, -1.9303e-01], + [-1.0594e-02, 1.8418e-02, 1.3818e-02, ..., -5.0999e-02, + 6.8655e-02, -3.8432e-02], + ..., + [-1.1431e-04, -1.2071e-02, 7.5122e-04, ..., -6.0458e-02, + -9.5947e-02, 2.5837e-02], + [-1.5513e-02, 3.2639e-04, -1.3118e-02, ..., -6.1755e-02, + 3.4323e-02, -1.3221e-01], + [-7.8796e-02, -7.9271e-04, 1.3922e-03, ..., -7.2737e-02, + -6.4374e-02, -6.4827e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0924e-06, + 2.0210e-06, 2.8638e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3232e-05, + 3.1870e-06, 2.1420e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4273e-06, + -1.1468e-04, 6.7521e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.4680e-07, + 9.0480e-05, 8.8476e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0375e-06, + -2.3004e-06, 1.7462e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.8836e-07, + 1.1986e-06, -7.3062e-07]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0087, -0.0167, 0.0002, 0.0314, -0.0080, 0.0322, 0.0052, 0.0164, + 0.0017, -0.0097], device='cuda:0'), grad: tensor([ 8.2552e-06, -2.0847e-05, -2.1350e-04, 4.6641e-05, 1.3486e-05, + 6.2399e-06, 1.3329e-05, 1.3590e-04, -3.9265e-06, 1.4082e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 214.44, cls_loss 0.0043 cls_loss_mapping 0.0077 cls_loss_causal 0.5913 re_mapping 0.0087 re_causal 0.0264 /// teacc 98.91 lr 0.00010000 +Epoch 98, weight, value: tensor([[-2.6397e-02, 3.8426e-04, -3.8951e-02, ..., 1.2849e-02, + -1.6458e-02, -1.2790e-03], + [ 3.2766e-02, 1.1911e-02, -1.7105e-02, ..., 2.5664e-02, + -2.0407e-02, -1.9286e-01], + [-1.0606e-02, 1.8494e-02, 1.3802e-02, ..., -5.1515e-02, + 6.8232e-02, -4.0259e-02], + ..., + [-1.0498e-04, -1.2086e-02, 6.9432e-04, ..., -6.1278e-02, + -9.6316e-02, 2.5541e-02], + [-1.5532e-02, 1.4026e-04, -1.3128e-02, ..., -6.1373e-02, + 3.5862e-02, -1.3175e-01], + [-7.8917e-02, -7.9638e-04, 1.3444e-03, ..., -7.2830e-02, + -6.4986e-02, -6.5550e-02]], device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -5.4240e-06, + 9.4846e-06, 4.2585e-07], + [-3.3062e-08, 0.0000e+00, 0.0000e+00, ..., 3.8631e-06, + 1.3143e-05, 3.3807e-06], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.2817e-06, + -3.5763e-05, 4.0978e-06], + ..., + [ 7.6834e-09, 0.0000e+00, 0.0000e+00, ..., 6.6496e-07, + 7.5102e-06, -2.5511e-05], + [ 6.7521e-09, 0.0000e+00, 0.0000e+00, ..., -4.3549e-06, + -3.0205e-05, -5.5768e-06], + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 2.1830e-06, + 3.3993e-06, 9.7230e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0079, -0.0160, -0.0007, 0.0313, -0.0074, 0.0322, 0.0049, 0.0153, + 0.0031, -0.0097], device='cuda:0'), grad: tensor([ 3.9600e-06, 4.5270e-05, 3.3408e-05, -5.2881e-04, 1.9610e-05, + 5.4419e-05, -1.7090e-06, 3.3975e-04, -8.9109e-06, 4.2677e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 214.49, cls_loss 0.0041 cls_loss_mapping 0.0080 cls_loss_causal 0.5886 re_mapping 0.0084 re_causal 0.0267 /// teacc 98.91 lr 0.00010000 +Epoch 99, weight, value: tensor([[-2.6402e-02, 4.0008e-04, -4.0351e-02, ..., 1.3207e-02, + -1.6888e-02, -1.3616e-03], + [ 3.2800e-02, 1.1888e-02, -1.7244e-02, ..., 2.5518e-02, + -2.0885e-02, -1.9345e-01], + [-1.0613e-02, 1.8486e-02, 1.3137e-02, ..., -5.2154e-02, + 6.8454e-02, -4.0836e-02], + ..., + [-1.0977e-04, -1.2088e-02, 2.3358e-04, ..., -6.1502e-02, + -9.6831e-02, 2.5934e-02], + [-1.5576e-02, 9.5116e-05, -1.3085e-02, ..., -6.1595e-02, + 3.6393e-02, -1.3070e-01], + [-7.8929e-02, -7.9806e-04, 1.1398e-03, ..., -7.3033e-02, + -6.4873e-02, -6.6700e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3919e-06, + 1.0446e-05, 2.8638e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4249e-07, + -1.5080e-05, 3.5623e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.9314e-07, + 7.8976e-06, 1.3504e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.9884e-08, + 2.1160e-06, -1.0943e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1686e-06, + 6.1020e-06, 5.0757e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7604e-06, + 2.5891e-06, -1.6969e-06]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0084, -0.0156, -0.0009, 0.0313, -0.0067, 0.0317, 0.0048, 0.0157, + 0.0030, -0.0105], device='cuda:0'), grad: tensor([ 9.3520e-05, -8.3733e-04, 5.2023e-04, 1.9383e-04, 7.6950e-05, + -9.2387e-05, 3.7670e-05, 8.7857e-05, 8.7619e-05, -1.6892e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 98---------------------------------------------------- +epoch 98, time 230.54, cls_loss 0.0047 cls_loss_mapping 0.0097 cls_loss_causal 0.5927 re_mapping 0.0083 re_causal 0.0254 /// teacc 99.01 lr 0.00010000 +Epoch 100, weight, value: tensor([[-2.6388e-02, 4.2707e-04, -4.0352e-02, ..., 1.3332e-02, + -1.7394e-02, -1.1085e-03], + [ 3.2792e-02, 1.1559e-02, -1.7243e-02, ..., 2.5655e-02, + -2.1047e-02, -1.9408e-01], + [-1.0621e-02, 1.8473e-02, 1.3136e-02, ..., -5.3293e-02, + 6.8738e-02, -4.0730e-02], + ..., + [-1.3178e-04, -1.2143e-02, 2.3299e-04, ..., -6.1695e-02, + -9.7114e-02, 2.5818e-02], + [-1.5587e-02, 2.0205e-03, -1.3082e-02, ..., -6.2121e-02, + 3.6567e-02, -1.3099e-01], + [-7.9046e-02, -8.6052e-04, 1.1395e-03, ..., -7.3249e-02, + -6.5678e-02, -6.7092e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.8348e-06, + 3.9130e-05, 7.4739e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8124e-06, + 4.0841e-04, 2.5146e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.4482e-07, + -3.2455e-05, 2.0815e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3108e-07, + -4.2820e-04, 1.0142e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9208e-05, + -1.1161e-05, 2.2654e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0990e-06, + 3.1501e-05, -2.6766e-06]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0083, -0.0148, -0.0017, 0.0308, -0.0060, 0.0318, 0.0041, 0.0159, + 0.0030, -0.0107], device='cuda:0'), grad: tensor([ 6.0827e-05, 8.1024e-03, 1.5712e-04, 1.0347e-04, 2.1949e-05, + 2.3007e-04, -2.5153e-04, -8.5831e-03, 1.0532e-04, 4.7922e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 99---------------------------------------------------- +epoch 99, time 230.73, cls_loss 0.0043 cls_loss_mapping 0.0096 cls_loss_causal 0.5791 re_mapping 0.0089 re_causal 0.0260 /// teacc 99.02 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0263, 0.0004, -0.0404, ..., 0.0135, -0.0179, -0.0008], + [ 0.0331, 0.0115, -0.0172, ..., 0.0257, -0.0216, -0.1945], + [-0.0107, 0.0185, 0.0131, ..., -0.0534, 0.0697, -0.0408], + ..., + [-0.0003, -0.0122, 0.0002, ..., -0.0619, -0.0980, 0.0256], + [-0.0158, 0.0024, -0.0131, ..., -0.0627, 0.0365, -0.1311], + [-0.0794, -0.0009, 0.0011, ..., -0.0723, -0.0656, -0.0673]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0863e-05, + 5.9372e-07, 3.4599e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5637e-06, + 4.1872e-06, 2.0638e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2582e-06, + -3.1367e-06, 5.1036e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6973e-07, + 1.2470e-06, 2.2322e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6412e-06, + -7.0669e-06, 1.2675e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4445e-06, + 1.6680e-06, -5.1081e-05]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0085, -0.0159, -0.0011, 0.0308, -0.0063, 0.0317, 0.0035, 0.0168, + 0.0026, -0.0100], device='cuda:0'), grad: tensor([-3.3975e-05, 3.3259e-05, 1.2346e-05, 3.5584e-05, 1.8597e-04, + 1.4879e-05, -5.1893e-06, 1.5211e-04, -2.7329e-05, -3.6740e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.82, cls_loss 0.0041 cls_loss_mapping 0.0085 cls_loss_causal 0.5466 re_mapping 0.0093 re_causal 0.0260 /// teacc 98.84 lr 0.00010000 +Epoch 102, weight, value: tensor([[-2.6309e-02, 4.2359e-04, -4.0785e-02, ..., 1.3267e-02, + -1.9116e-02, -8.9251e-04], + [ 3.3348e-02, 1.1521e-02, -1.7255e-02, ..., 2.5214e-02, + -2.2571e-02, -1.9499e-01], + [-1.0648e-02, 1.8491e-02, 1.3106e-02, ..., -5.3552e-02, + 7.1194e-02, -4.0439e-02], + ..., + [-2.6704e-04, -1.2165e-02, -5.4076e-05, ..., -6.2110e-02, + -9.8864e-02, 2.5528e-02], + [-1.6081e-02, 2.4466e-03, -1.3277e-02, ..., -6.3302e-02, + 3.6582e-02, -1.3212e-01], + [-7.9658e-02, -8.7668e-04, 9.5353e-04, ..., -7.2517e-02, + -6.7374e-02, -6.7209e-02]], device='cuda:0'), grad: tensor([[ 5.7742e-08, 0.0000e+00, 0.0000e+00, ..., -1.0123e-06, + 6.2548e-06, 7.3528e-07], + [-1.3681e-06, 0.0000e+00, 0.0000e+00, ..., 3.4692e-08, + 3.4980e-06, -2.1560e-07], + [ 4.6124e-07, 0.0000e+00, 0.0000e+00, ..., 1.1483e-06, + -1.7369e-04, 1.2713e-06], + ..., + [ 1.5087e-07, 0.0000e+00, 0.0000e+00, ..., 2.6869e-07, + 3.6597e-04, -1.7677e-06], + [ 5.5600e-07, 0.0000e+00, 0.0000e+00, ..., 1.2508e-06, + 1.0677e-05, 2.3525e-06], + [ 8.6147e-09, 0.0000e+00, 0.0000e+00, ..., 8.2422e-07, + 2.8178e-05, 1.7434e-05]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0080, -0.0159, -0.0003, 0.0313, -0.0064, 0.0315, 0.0038, 0.0169, + 0.0023, -0.0102], device='cuda:0'), grad: tensor([ 4.9882e-06, -4.7870e-06, 1.2779e-03, -2.2182e-03, -3.8534e-05, + -8.9645e-05, 1.0066e-05, 9.1743e-04, 5.0068e-05, 9.1076e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 214.67, cls_loss 0.0043 cls_loss_mapping 0.0092 cls_loss_causal 0.5705 re_mapping 0.0089 re_causal 0.0268 /// teacc 98.92 lr 0.00010000 +Epoch 103, weight, value: tensor([[-2.5175e-02, 4.2347e-04, -4.0793e-02, ..., 1.3201e-02, + -1.9744e-02, -9.5481e-04], + [ 3.3280e-02, 1.1520e-02, -1.7255e-02, ..., 2.5228e-02, + -2.3098e-02, -1.9377e-01], + [-1.0462e-02, 1.8493e-02, 1.3105e-02, ..., -5.3766e-02, + 7.2100e-02, -4.0374e-02], + ..., + [-7.1211e-04, -1.2165e-02, -5.9020e-05, ..., -6.2323e-02, + -1.0024e-01, 2.5066e-02], + [-1.5839e-02, 2.4413e-03, -1.3280e-02, ..., -6.3879e-02, + 3.6644e-02, -1.3338e-01], + [-8.0247e-02, -8.7671e-04, 9.5004e-04, ..., -7.2652e-02, + -6.8075e-02, -7.0035e-02]], device='cuda:0'), grad: tensor([[ 1.9115e-07, 0.0000e+00, 0.0000e+00, ..., -2.4363e-06, + -8.6753e-07, 1.0780e-07], + [ 1.5013e-05, 0.0000e+00, 0.0000e+00, ..., 4.2375e-08, + 2.2620e-05, 1.1344e-06], + [ 2.0936e-06, 0.0000e+00, 0.0000e+00, ..., 8.4518e-08, + -1.6615e-05, -1.1228e-05], + ..., + [ 1.0198e-06, 0.0000e+00, 0.0000e+00, ..., 2.5146e-08, + 1.8664e-06, 1.9986e-06], + [-2.6688e-05, 0.0000e+00, 0.0000e+00, ..., 3.9977e-07, + -3.5733e-05, 1.8720e-07], + [ 6.1886e-07, 0.0000e+00, 0.0000e+00, ..., 4.8755e-07, + 3.5334e-06, 1.1260e-06]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0074, -0.0152, -0.0001, 0.0314, -0.0049, 0.0320, 0.0038, 0.0165, + 0.0019, -0.0120], device='cuda:0'), grad: tensor([-4.2021e-06, 5.6028e-05, -1.0006e-05, 2.9281e-06, 1.4558e-05, + 8.5756e-06, 1.5944e-05, 1.1146e-05, -1.0180e-04, 6.7838e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 214.77, cls_loss 0.0047 cls_loss_mapping 0.0083 cls_loss_causal 0.5855 re_mapping 0.0082 re_causal 0.0251 /// teacc 98.92 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0252, 0.0004, -0.0412, ..., 0.0129, -0.0209, -0.0011], + [ 0.0336, 0.0115, -0.0173, ..., 0.0250, -0.0237, -0.1949], + [-0.0105, 0.0186, 0.0127, ..., -0.0542, 0.0726, -0.0397], + ..., + [-0.0008, -0.0122, -0.0011, ..., -0.0625, -0.1011, 0.0231], + [-0.0160, 0.0021, -0.0139, ..., -0.0638, 0.0376, -0.1332], + [-0.0804, -0.0009, 0.0008, ..., -0.0729, -0.0685, -0.0694]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.5523e-06, + 1.0379e-05, 6.1328e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.9652e-07, + 3.5409e-06, 1.1986e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-06, + 3.7272e-06, 1.1502e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.2888e-08, + 1.1094e-05, 9.0897e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8275e-06, + -2.2039e-05, 2.6654e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0606e-07, + 1.1124e-05, -2.4557e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0066, -0.0158, -0.0008, 0.0313, -0.0052, 0.0318, 0.0040, 0.0162, + 0.0025, -0.0108], device='cuda:0'), grad: tensor([ 2.8074e-05, 2.0325e-05, 8.0109e-05, -8.2433e-05, 1.0937e-04, + 3.0965e-05, -9.6858e-05, 1.7032e-05, -1.1003e-04, 3.5800e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 214.61, cls_loss 0.0041 cls_loss_mapping 0.0082 cls_loss_causal 0.5757 re_mapping 0.0090 re_causal 0.0262 /// teacc 98.93 lr 0.00010000 +Epoch 105, weight, value: tensor([[-2.5252e-02, 4.4206e-04, -4.1214e-02, ..., 1.3731e-02, + -2.1497e-02, -1.5622e-03], + [ 3.3872e-02, 1.1480e-02, -1.7279e-02, ..., 2.4055e-02, + -2.4620e-02, -1.9610e-01], + [-1.0746e-02, 1.8579e-02, 1.2720e-02, ..., -5.4612e-02, + 7.3308e-02, -4.1202e-02], + ..., + [-9.7954e-05, -1.2181e-02, -1.0788e-03, ..., -6.2654e-02, + -1.0159e-01, 2.3523e-02], + [-1.6004e-02, 2.0461e-03, -1.3936e-02, ..., -6.4151e-02, + 3.7887e-02, -1.3433e-01], + [-8.0779e-02, -8.8479e-04, 8.1276e-04, ..., -7.3101e-02, + -6.9073e-02, -6.9520e-02]], device='cuda:0'), grad: tensor([[ 3.5856e-08, 0.0000e+00, 0.0000e+00, ..., -5.5507e-07, + 3.0756e-05, 8.8662e-06], + [-8.8988e-07, 0.0000e+00, 0.0000e+00, ..., -2.6785e-06, + 2.3562e-06, 1.4175e-06], + [ 7.9628e-08, 0.0000e+00, 0.0000e+00, ..., 5.2573e-07, + -1.1659e-04, -2.5973e-05], + ..., + [ 9.1735e-08, 0.0000e+00, 0.0000e+00, ..., 3.9488e-07, + 6.3293e-06, 1.9241e-06], + [ 1.3877e-07, 0.0000e+00, 0.0000e+00, ..., 2.1905e-06, + 5.8323e-05, 1.4611e-05], + [ 1.8161e-08, 0.0000e+00, 0.0000e+00, ..., 7.9814e-07, + 1.4521e-05, 1.5318e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0077, -0.0159, -0.0009, 0.0311, -0.0050, 0.0317, 0.0041, 0.0167, + 0.0022, -0.0113], device='cuda:0'), grad: tensor([ 4.7565e-05, 2.4766e-05, -1.7190e-04, 2.4930e-05, -2.2143e-05, + -2.3991e-05, 2.2277e-05, -1.0693e-04, 9.6619e-05, 1.0860e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 214.31, cls_loss 0.0040 cls_loss_mapping 0.0064 cls_loss_causal 0.5788 re_mapping 0.0090 re_causal 0.0268 /// teacc 98.64 lr 0.00010000 +Epoch 106, weight, value: tensor([[-2.5079e-02, 5.5103e-04, -4.2571e-02, ..., 1.4015e-02, + -2.2079e-02, -1.5906e-03], + [ 3.3907e-02, 1.1277e-02, -1.7400e-02, ..., 2.3865e-02, + -2.5469e-02, -1.9664e-01], + [-1.0604e-02, 1.8600e-02, 1.2314e-02, ..., -5.5951e-02, + 7.3833e-02, -4.1283e-02], + ..., + [-1.4211e-04, -1.2197e-02, -1.2024e-03, ..., -6.2836e-02, + -1.0222e-01, 2.4059e-02], + [-1.5984e-02, 1.5456e-03, -1.5684e-02, ..., -6.4458e-02, + 3.8411e-02, -1.3534e-01], + [-8.1021e-02, -9.0084e-04, 4.4046e-03, ..., -7.3310e-02, + -6.9933e-02, -6.9465e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.3819e-09, 0.0000e+00, ..., -8.4713e-06, + 3.5223e-06, 1.9651e-07], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.5635e-08, + 8.5160e-06, 4.7637e-07], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 3.1991e-07, + 4.1699e-04, 1.1943e-05], + ..., + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 6.0908e-06, 3.2037e-07], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 3.3015e-07, + 3.4451e-04, -5.9865e-06], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.1874e-07, + 1.0878e-05, -4.9509e-06]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0079, -0.0162, -0.0014, 0.0307, -0.0056, 0.0318, 0.0041, 0.0172, + 0.0024, -0.0110], device='cuda:0'), grad: tensor([ 3.6210e-05, 2.6166e-05, 1.3266e-03, -1.2665e-03, 3.3522e-04, + -7.8011e-03, 7.4005e-03, 2.6822e-05, 4.2844e-04, -5.0879e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 214.52, cls_loss 0.0054 cls_loss_mapping 0.0087 cls_loss_causal 0.5805 re_mapping 0.0090 re_causal 0.0266 /// teacc 98.89 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0251, 0.0007, -0.0448, ..., 0.0135, -0.0251, -0.0017], + [ 0.0342, 0.0110, -0.0185, ..., 0.0220, -0.0252, -0.1974], + [-0.0107, 0.0169, 0.0103, ..., -0.0566, 0.0742, -0.0426], + ..., + [-0.0003, -0.0123, -0.0020, ..., -0.0628, -0.1038, 0.0245], + [-0.0161, 0.0005, -0.0149, ..., -0.0646, 0.0386, -0.1360], + [-0.0812, -0.0009, 0.0041, ..., -0.0735, -0.0708, -0.0695]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3270e-04, + 1.7798e-06, 5.4017e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.2519e-06, + 2.0236e-05, 2.6040e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0682e-06, + -4.5180e-05, 3.3528e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1269e-07, + 9.6038e-06, 1.2383e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6550e-06, + -1.9118e-05, -1.4400e-04], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5640e-06, + 2.1040e-05, 1.3161e-04]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0072, -0.0160, -0.0023, 0.0327, -0.0057, 0.0331, 0.0030, 0.0168, + 0.0016, -0.0106], device='cuda:0'), grad: tensor([-4.9448e-04, 3.3438e-05, -6.7294e-05, 3.5703e-05, 6.5565e-06, + 7.7263e-06, 4.6253e-04, 5.6535e-05, -5.5933e-04, 5.1785e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 214.70, cls_loss 0.0033 cls_loss_mapping 0.0068 cls_loss_causal 0.5594 re_mapping 0.0088 re_causal 0.0267 /// teacc 98.87 lr 0.00010000 +Epoch 108, weight, value: tensor([[-2.5069e-02, 6.7492e-04, -4.4982e-02, ..., 1.2651e-02, + -2.7238e-02, -1.8463e-03], + [ 3.3570e-02, 1.0880e-02, -1.8550e-02, ..., 2.1666e-02, + -2.5546e-02, -1.9796e-01], + [-1.0710e-02, 1.7071e-02, 1.0493e-02, ..., -5.6933e-02, + 7.4878e-02, -4.2716e-02], + ..., + [-4.2557e-04, -1.2446e-02, -2.2267e-03, ..., -6.2856e-02, + -1.0518e-01, 2.4449e-02], + [-1.4771e-02, 1.0695e-04, -1.4874e-02, ..., -6.5176e-02, + 3.8106e-02, -1.3612e-01], + [-8.1334e-02, -9.2197e-04, 4.0446e-03, ..., -7.3927e-02, + -7.1795e-02, -6.9576e-02]], device='cuda:0'), grad: tensor([[ 5.1223e-09, 0.0000e+00, 1.8626e-09, ..., -7.9572e-06, + -3.5530e-07, 4.5635e-08], + [-1.2061e-07, 0.0000e+00, 2.3283e-09, ..., 1.0431e-07, + 1.3877e-06, -2.2352e-08], + [ 1.6764e-08, 0.0000e+00, -5.2620e-08, ..., 3.4180e-07, + -1.0289e-05, 1.0058e-07], + ..., + [ 1.1176e-08, 0.0000e+00, 8.8476e-09, ..., 1.0896e-07, + 5.8524e-06, 5.3458e-07], + [ 2.0489e-08, 0.0000e+00, 9.3132e-09, ..., 1.3132e-06, + 3.3081e-05, 5.3179e-07], + [ 4.6566e-09, 0.0000e+00, 4.6566e-10, ..., 6.5565e-07, + -3.8624e-05, 1.5795e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0069, -0.0162, -0.0018, 0.0315, -0.0056, 0.0343, 0.0035, 0.0168, + 0.0013, -0.0112], device='cuda:0'), grad: tensor([-8.1360e-06, -6.9812e-06, -1.0073e-05, 2.4617e-05, 4.1425e-06, + 9.6262e-06, 9.7528e-06, -6.5006e-06, 1.6391e-04, -1.8013e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 214.63, cls_loss 0.0040 cls_loss_mapping 0.0090 cls_loss_causal 0.5833 re_mapping 0.0082 re_causal 0.0254 /// teacc 98.91 lr 0.00010000 +Epoch 109, weight, value: tensor([[-2.5073e-02, 7.4129e-04, -4.5403e-02, ..., 1.3788e-02, + -2.7527e-02, -1.8825e-03], + [ 3.3599e-02, 1.0822e-02, -1.7457e-02, ..., 2.1449e-02, + -2.5714e-02, -1.9908e-01], + [-1.0718e-02, 1.7069e-02, 1.0269e-02, ..., -5.7128e-02, + 7.5361e-02, -4.2705e-02], + ..., + [-4.4699e-04, -1.2452e-02, -2.4399e-03, ..., -6.3208e-02, + -1.0603e-01, 2.6023e-02], + [-1.4842e-02, 1.1402e-05, -1.4924e-02, ..., -6.6743e-02, + 3.7627e-02, -1.3680e-01], + [-8.1438e-02, -9.2831e-04, 3.6368e-03, ..., -7.5008e-02, + -7.2738e-02, -6.9386e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.0892e-06, + -3.7719e-06, 6.3796e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.1204e-06, + 8.1398e-07, -6.7018e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.1898e-07, + -3.1322e-05, 8.3912e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4948e-07, + 7.9796e-06, 5.4110e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.7696e-07, + 1.1154e-05, 6.5658e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1409e-06, + 4.1798e-06, 2.2352e-07]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0077, -0.0156, -0.0016, 0.0307, -0.0062, 0.0342, 0.0040, 0.0168, + 0.0003, -0.0108], device='cuda:0'), grad: tensor([-1.6332e-05, -7.3254e-05, -2.8506e-05, 1.7822e-05, 5.1588e-05, + 5.5619e-06, 7.8008e-06, -1.0133e-05, 2.6941e-05, 1.8537e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 214.66, cls_loss 0.0044 cls_loss_mapping 0.0092 cls_loss_causal 0.5710 re_mapping 0.0083 re_causal 0.0249 /// teacc 98.90 lr 0.00010000 +Epoch 110, weight, value: tensor([[-2.5049e-02, 8.3833e-04, -4.5989e-02, ..., 1.3796e-02, + -2.8239e-02, -2.1544e-03], + [ 3.3710e-02, 1.0774e-02, -1.8599e-02, ..., 2.1232e-02, + -2.6346e-02, -2.0078e-01], + [-1.0807e-02, 1.7056e-02, 1.2040e-02, ..., -5.7559e-02, + 7.5872e-02, -4.2708e-02], + ..., + [-5.2193e-04, -1.2455e-02, -2.6422e-03, ..., -6.3510e-02, + -1.0695e-01, 2.5565e-02], + [-1.4706e-02, -1.2762e-04, -1.5102e-02, ..., -6.7118e-02, + 3.8047e-02, -1.3674e-01], + [-8.1574e-02, -9.3937e-04, 3.9050e-03, ..., -7.5589e-02, + -7.3453e-02, -6.9350e-02]], device='cuda:0'), grad: tensor([[ 2.5192e-07, 0.0000e+00, 0.0000e+00, ..., -1.9949e-06, + 5.9279e-07, 5.0478e-07], + [-4.1388e-06, 0.0000e+00, 0.0000e+00, ..., 6.4727e-08, + 7.6042e-07, -7.1675e-06], + [ 2.2501e-06, 0.0000e+00, 0.0000e+00, ..., 2.1048e-07, + -4.0829e-06, 3.9935e-06], + ..., + [ 5.4995e-07, 0.0000e+00, 0.0000e+00, ..., 2.7474e-08, + 2.7996e-06, 1.0645e-06], + [ 3.0966e-07, 0.0000e+00, 0.0000e+00, ..., 3.7020e-07, + -1.0565e-05, 7.0687e-07], + [ 1.5367e-08, 0.0000e+00, 0.0000e+00, ..., 5.1688e-07, + 4.2953e-06, 5.9092e-07]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0069, -0.0164, -0.0016, 0.0300, -0.0058, 0.0344, 0.0040, 0.0174, + 0.0004, -0.0106], device='cuda:0'), grad: tensor([-1.0459e-06, -2.4796e-05, 7.4394e-06, 5.2787e-06, 1.3195e-05, + 4.0829e-06, 8.0168e-06, 4.7684e-06, -2.0519e-05, 3.5875e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 214.86, cls_loss 0.0034 cls_loss_mapping 0.0062 cls_loss_causal 0.5652 re_mapping 0.0082 re_causal 0.0245 /// teacc 98.85 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0251, 0.0012, -0.0464, ..., 0.0139, -0.0290, -0.0023], + [ 0.0337, 0.0106, -0.0189, ..., 0.0211, -0.0266, -0.2010], + [-0.0108, 0.0171, 0.0118, ..., -0.0579, 0.0770, -0.0420], + ..., + [-0.0006, -0.0125, -0.0029, ..., -0.0636, -0.1091, 0.0252], + [-0.0147, -0.0007, -0.0141, ..., -0.0671, 0.0383, -0.1371], + [-0.0816, -0.0010, 0.0038, ..., -0.0760, -0.0739, -0.0696]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8375e-06, ..., -1.3880e-05, + -2.8573e-06, -2.3749e-08], + [ 0.0000e+00, 0.0000e+00, 3.8743e-06, ..., 1.1493e-06, + 1.4655e-05, 3.9535e-07], + [ 0.0000e+00, 0.0000e+00, 4.1090e-06, ..., 7.1973e-06, + -2.8059e-05, 1.8766e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6570e-05, ..., 3.7961e-06, + 6.0707e-05, 3.6415e-07], + [ 0.0000e+00, 0.0000e+00, 1.3113e-05, ..., 3.9302e-06, + 3.4243e-05, 5.6811e-08], + [ 0.0000e+00, 0.0000e+00, 9.0944e-07, ..., 2.9579e-06, + 1.0118e-05, 9.9279e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0068, -0.0162, -0.0009, 0.0299, -0.0053, 0.0338, 0.0042, 0.0166, + 0.0003, -0.0107], device='cuda:0'), grad: tensor([-1.0699e-04, 3.8534e-05, -3.9577e-05, -1.8158e-03, 1.4029e-05, + 1.5745e-03, 5.3167e-05, 1.4198e-04, 7.8559e-05, 6.1005e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 214.54, cls_loss 0.0046 cls_loss_mapping 0.0073 cls_loss_causal 0.5659 re_mapping 0.0081 re_causal 0.0246 /// teacc 99.01 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0243, 0.0016, -0.0517, ..., 0.0144, -0.0296, -0.0022], + [ 0.0342, 0.0103, -0.0185, ..., 0.0210, -0.0284, -0.2016], + [-0.0109, 0.0170, 0.0103, ..., -0.0585, 0.0784, -0.0424], + ..., + [-0.0011, -0.0128, -0.0055, ..., -0.0641, -0.1099, 0.0252], + [-0.0149, -0.0021, -0.0156, ..., -0.0676, 0.0386, -0.1373], + [-0.0822, -0.0010, 0.0027, ..., -0.0764, -0.0746, -0.0697]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -6.1002e-08, 0.0000e+00, ..., 5.8394e-07, + 3.1851e-06, 1.3970e-09], + [-1.3551e-07, 1.3970e-09, 0.0000e+00, ..., 5.8208e-07, + 2.4494e-06, 3.7253e-09], + [ 1.3970e-08, 1.0245e-08, 0.0000e+00, ..., 3.7011e-06, + -2.1998e-06, 5.5879e-09], + ..., + [ 2.6077e-08, 4.6566e-10, 0.0000e+00, ..., 6.3796e-08, + 2.5425e-06, 3.7253e-08], + [ 4.2375e-08, 7.9162e-09, 0.0000e+00, ..., 1.2368e-06, + 9.8813e-07, 4.1910e-09], + [ 3.2596e-09, 8.8476e-09, 0.0000e+00, ..., 3.9348e-07, + 1.2014e-06, 2.3749e-08]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0067, -0.0163, -0.0004, 0.0302, -0.0053, 0.0339, 0.0039, 0.0160, + 0.0014, -0.0110], device='cuda:0'), grad: tensor([ 4.3893e-04, -1.2379e-03, 1.1230e-04, 8.2552e-05, 8.8334e-05, + 1.9324e-04, 6.1035e-05, 1.1814e-04, 5.4598e-04, -4.0197e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 214.56, cls_loss 0.0034 cls_loss_mapping 0.0054 cls_loss_causal 0.5623 re_mapping 0.0080 re_causal 0.0245 /// teacc 98.92 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0256, 0.0034, -0.0529, ..., 0.0142, -0.0306, -0.0023], + [ 0.0357, 0.0086, -0.0199, ..., 0.0209, -0.0286, -0.2018], + [-0.0111, 0.0177, 0.0115, ..., -0.0592, 0.0787, -0.0423], + ..., + [-0.0021, -0.0141, -0.0057, ..., -0.0646, -0.1107, 0.0251], + [-0.0137, -0.0071, -0.0148, ..., -0.0679, 0.0391, -0.1378], + [-0.0839, -0.0014, 0.0020, ..., -0.0766, -0.0757, -0.0697]], + device='cuda:0'), grad: tensor([[ 1.7602e-07, 0.0000e+00, 9.3132e-10, ..., 1.3290e-06, + 2.2911e-06, 3.4459e-08], + [ 1.9044e-05, 0.0000e+00, 4.6566e-10, ..., 3.6806e-06, + 9.3728e-06, 2.5705e-07], + [ 9.7975e-07, 0.0000e+00, 0.0000e+00, ..., 2.7776e-05, + 3.2306e-05, 6.0536e-08], + ..., + [-2.5526e-05, 0.0000e+00, 4.6566e-10, ..., 1.2619e-07, + 5.6252e-06, 7.9907e-07], + [ 2.1067e-06, 0.0000e+00, 2.7940e-09, ..., 4.6119e-06, + 4.8727e-06, 1.0617e-07], + [ 1.3318e-06, 0.0000e+00, 2.3283e-09, ..., 6.0024e-07, + 1.3411e-06, 9.6783e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0062, -0.0161, -0.0007, 0.0311, -0.0053, 0.0332, 0.0039, 0.0161, + 0.0014, -0.0110], device='cuda:0'), grad: tensor([ 9.0376e-06, 3.5620e-04, 6.4909e-05, 2.8461e-05, 8.8215e-05, + 1.3661e-04, -3.0351e-04, -4.1509e-04, 4.9293e-05, -1.3962e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 214.32, cls_loss 0.0030 cls_loss_mapping 0.0065 cls_loss_causal 0.5427 re_mapping 0.0081 re_causal 0.0240 /// teacc 98.89 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0257, 0.0041, -0.0554, ..., 0.0142, -0.0311, -0.0023], + [ 0.0360, 0.0079, -0.0197, ..., 0.0207, -0.0286, -0.2020], + [-0.0111, 0.0180, 0.0108, ..., -0.0598, 0.0789, -0.0427], + ..., + [-0.0019, -0.0143, -0.0091, ..., -0.0647, -0.1115, 0.0255], + [-0.0138, -0.0085, -0.0167, ..., -0.0683, 0.0391, -0.1379], + [-0.0851, -0.0016, 0.0009, ..., -0.0767, -0.0762, -0.0697]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., -1.6168e-06, + 7.5717e-07, 3.9581e-08], + [ 6.6590e-08, 0.0000e+00, 0.0000e+00, ..., 8.7637e-07, + 1.7602e-06, 2.9849e-07], + [ 3.5390e-08, 0.0000e+00, 0.0000e+00, ..., 2.6263e-07, + -1.0394e-06, 1.1176e-07], + ..., + [ 7.6834e-08, 0.0000e+00, 0.0000e+00, ..., 4.9826e-08, + 4.3400e-07, 2.5379e-07], + [ 2.8405e-08, 0.0000e+00, 0.0000e+00, ..., 2.9784e-06, + -2.5164e-06, 1.6345e-07], + [ 9.5461e-08, 0.0000e+00, 0.0000e+00, ..., 1.1809e-06, + 2.7157e-06, 1.1409e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0059, -0.0156, -0.0010, 0.0316, -0.0055, 0.0332, 0.0040, 0.0161, + 0.0010, -0.0110], device='cuda:0'), grad: tensor([ 1.5152e-04, 3.7923e-06, 2.4930e-05, 3.1255e-06, 6.1810e-05, + 7.1712e-07, -3.3230e-06, 1.0699e-05, -3.0585e-06, -2.5010e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 214.38, cls_loss 0.0036 cls_loss_mapping 0.0075 cls_loss_causal 0.5540 re_mapping 0.0073 re_causal 0.0225 /// teacc 98.84 lr 0.00010000 +Epoch 115, weight, value: tensor([[-2.5900e-02, 4.0682e-03, -5.6205e-02, ..., 1.4995e-02, + -3.0818e-02, -2.4174e-03], + [ 3.7785e-02, 7.2656e-03, -1.9819e-02, ..., 2.0514e-02, + -2.9193e-02, -2.0231e-01], + [-1.1750e-02, 1.9344e-02, 1.0618e-02, ..., -6.0270e-02, + 7.9521e-02, -4.2877e-02], + ..., + [-1.5224e-03, -1.5982e-02, -9.0951e-03, ..., -6.6428e-02, + -1.1237e-01, 2.5568e-02], + [-1.5179e-02, -9.6954e-03, -1.7284e-02, ..., -6.8071e-02, + 3.9919e-02, -1.3824e-01], + [-8.7812e-02, -1.6461e-03, -8.8172e-05, ..., -7.6994e-02, + -7.6684e-02, -6.9680e-02]], device='cuda:0'), grad: tensor([[ 4.0978e-08, 6.1048e-07, 1.5832e-08, ..., -1.0714e-05, + 1.0483e-05, 8.8941e-08], + [-1.5553e-06, 6.2864e-08, 1.0245e-08, ..., 5.6205e-07, + 1.9282e-05, 1.6997e-07], + [ 2.0070e-07, -4.1015e-06, 3.7253e-09, ..., 6.6450e-07, + -3.4523e-04, 4.7497e-08], + ..., + [ 3.1106e-07, 2.6673e-06, 1.3504e-08, ..., 3.5390e-07, + 2.6202e-04, 5.3737e-07], + [ 2.4121e-07, 3.4878e-07, 3.3528e-08, ..., 9.1316e-07, + 1.0572e-05, 1.2852e-07], + [ 1.1222e-07, 3.1199e-08, 6.2399e-08, ..., 3.2503e-06, + 8.4713e-06, -2.6617e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0065, -0.0161, -0.0010, 0.0310, -0.0056, 0.0327, 0.0040, 0.0166, + 0.0015, -0.0108], device='cuda:0'), grad: tensor([ 1.1064e-05, 4.6134e-05, -1.0090e-03, 6.3479e-05, 3.6299e-05, + 9.3654e-06, 1.4737e-05, 7.7963e-04, 3.0577e-05, 1.6555e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 214.27, cls_loss 0.0029 cls_loss_mapping 0.0052 cls_loss_causal 0.5182 re_mapping 0.0073 re_causal 0.0226 /// teacc 98.85 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0257, 0.0047, -0.0569, ..., 0.0169, -0.0309, -0.0026], + [ 0.0389, 0.0049, -0.0195, ..., 0.0200, -0.0295, -0.2026], + [-0.0118, 0.0172, 0.0105, ..., -0.0609, 0.0798, -0.0433], + ..., + [-0.0020, -0.0201, -0.0091, ..., -0.0669, -0.1132, 0.0254], + [-0.0153, -0.0134, -0.0149, ..., -0.0682, 0.0403, -0.1389], + [-0.0899, -0.0018, -0.0008, ..., -0.0779, -0.0771, -0.0699]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.2690e-07, + 3.6173e-06, 1.5646e-07], + [-1.9092e-08, 0.0000e+00, 0.0000e+00, ..., 5.1828e-07, + 1.4557e-06, 8.1351e-07], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.7439e-07, + -5.2080e-06, 6.3609e-07], + ..., + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 1.4761e-07, + 6.2250e-06, 7.6042e-07], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 1.0774e-05, + 1.9461e-05, 2.7288e-07], + [ 9.3132e-10, 0.0000e+00, 4.6566e-10, ..., 3.5111e-06, + 7.3984e-06, 3.2764e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0081, -0.0160, -0.0015, 0.0317, -0.0056, 0.0325, 0.0035, 0.0164, + 0.0015, -0.0109], device='cuda:0'), grad: tensor([ 5.7444e-06, 5.5917e-06, -3.5446e-06, 2.8467e-04, -5.8487e-06, + -1.6093e-04, -1.7917e-04, -4.2878e-06, 3.9428e-05, 1.7807e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 214.36, cls_loss 0.0041 cls_loss_mapping 0.0063 cls_loss_causal 0.5766 re_mapping 0.0074 re_causal 0.0238 /// teacc 99.02 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0259, 0.0048, -0.0573, ..., 0.0167, -0.0321, -0.0028], + [ 0.0395, 0.0048, -0.0193, ..., 0.0198, -0.0299, -0.2051], + [-0.0121, 0.0171, 0.0105, ..., -0.0622, 0.0804, -0.0436], + ..., + [-0.0025, -0.0201, -0.0091, ..., -0.0671, -0.1144, 0.0256], + [-0.0152, -0.0136, -0.0150, ..., -0.0683, 0.0407, -0.1410], + [-0.0919, -0.0018, -0.0015, ..., -0.0781, -0.0763, -0.0699]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., -3.9442e-07, + 8.1677e-07, 1.2945e-07], + [-1.1129e-07, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, + 1.7900e-06, 5.0524e-07], + [ 2.5611e-08, 0.0000e+00, 0.0000e+00, ..., 6.3330e-08, + -5.7518e-06, 8.7079e-08], + ..., + [ 2.6077e-08, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 1.2768e-06, 8.0420e-07], + [ 1.7229e-08, 0.0000e+00, 0.0000e+00, ..., 1.0198e-07, + 5.0999e-06, 2.0582e-07], + [ 2.4214e-08, 0.0000e+00, 0.0000e+00, ..., 9.2201e-08, + 1.4216e-05, -1.5691e-05]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0072, -0.0165, -0.0003, 0.0300, -0.0055, 0.0323, 0.0035, 0.0162, + 0.0006, -0.0093], device='cuda:0'), grad: tensor([ 2.0750e-06, 2.0728e-05, -3.9302e-07, 3.1263e-05, 9.9957e-05, + -6.7770e-05, 1.0423e-05, -2.8640e-05, 1.2375e-05, -7.9930e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 116---------------------------------------------------- +epoch 116, time 230.55, cls_loss 0.0025 cls_loss_mapping 0.0053 cls_loss_causal 0.5874 re_mapping 0.0077 re_causal 0.0243 /// teacc 99.07 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0260, 0.0048, -0.0576, ..., 0.0171, -0.0322, -0.0025], + [ 0.0391, 0.0048, -0.0200, ..., 0.0197, -0.0301, -0.2060], + [-0.0109, 0.0171, 0.0101, ..., -0.0626, 0.0812, -0.0438], + ..., + [-0.0028, -0.0201, -0.0092, ..., -0.0673, -0.1153, 0.0255], + [-0.0156, -0.0136, -0.0143, ..., -0.0682, 0.0411, -0.1421], + [-0.0939, -0.0018, -0.0020, ..., -0.0783, -0.0770, -0.0703]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -5.2661e-05, + -1.6168e-05, 1.8626e-08], + [-9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 1.1595e-07, + 1.6345e-07, 3.8650e-08], + [ 4.6566e-10, -1.3970e-09, 0.0000e+00, ..., 1.6004e-05, + 2.5239e-06, 1.7695e-08], + ..., + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 1.1167e-06, + 2.8824e-07, 7.7300e-08], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 2.8815e-06, + 7.1060e-07, 2.1886e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 6.8918e-06, + 2.0787e-06, 6.1840e-07]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0074, -0.0171, -0.0002, 0.0301, -0.0049, 0.0319, 0.0035, 0.0170, + 0.0007, -0.0099], device='cuda:0'), grad: tensor([-1.0800e-04, -5.0664e-07, 3.6120e-05, 8.5160e-06, 1.1511e-06, + 1.5989e-05, 3.2127e-05, -7.6056e-05, 8.5458e-06, 8.2076e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 214.11, cls_loss 0.0029 cls_loss_mapping 0.0061 cls_loss_causal 0.5667 re_mapping 0.0073 re_causal 0.0234 /// teacc 98.97 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0261, 0.0048, -0.0577, ..., 0.0171, -0.0327, -0.0026], + [ 0.0388, 0.0047, -0.0198, ..., 0.0194, -0.0308, -0.2065], + [-0.0108, 0.0171, 0.0100, ..., -0.0631, 0.0819, -0.0437], + ..., + [-0.0029, -0.0201, -0.0092, ..., -0.0676, -0.1154, 0.0255], + [-0.0152, -0.0137, -0.0143, ..., -0.0685, 0.0412, -0.1423], + [-0.0951, -0.0018, -0.0022, ..., -0.0785, -0.0773, -0.0705]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -4.9695e-06, + 5.6587e-06, 3.2652e-06], + [-2.0955e-08, 0.0000e+00, 0.0000e+00, ..., 1.6866e-06, + 1.0477e-06, 7.7114e-06], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.3802e-06, + -4.0494e-06, 8.6194e-07], + ..., + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 5.9139e-07, + 2.8834e-06, -2.1338e-05], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.2596e-06, + 7.6890e-06, 2.0536e-07], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 6.2445e-07, + 1.1183e-05, 1.4059e-05]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0072, -0.0170, 0.0001, 0.0297, -0.0048, 0.0317, 0.0038, 0.0171, + 0.0005, -0.0102], device='cuda:0'), grad: tensor([ 3.2693e-05, 1.1086e-04, 8.1807e-06, -2.8819e-05, -5.1200e-05, + 5.2750e-05, -1.2562e-05, -2.9111e-04, 5.9277e-05, 1.1992e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 214.49, cls_loss 0.0029 cls_loss_mapping 0.0055 cls_loss_causal 0.5789 re_mapping 0.0077 re_causal 0.0238 /// teacc 98.92 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0262, 0.0048, -0.0579, ..., 0.0180, -0.0326, -0.0027], + [ 0.0389, 0.0047, -0.0196, ..., 0.0188, -0.0313, -0.2071], + [-0.0108, 0.0171, 0.0098, ..., -0.0642, 0.0822, -0.0436], + ..., + [-0.0031, -0.0202, -0.0091, ..., -0.0680, -0.1160, 0.0255], + [-0.0154, -0.0139, -0.0145, ..., -0.0694, 0.0413, -0.1428], + [-0.0952, -0.0019, -0.0016, ..., -0.0789, -0.0782, -0.0713]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 4.7207e-04, + 3.3951e-04, 2.1094e-07], + [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., -6.3609e-07, + 3.2336e-06, 1.6792e-06], + [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 8.1807e-06, + 3.0864e-06, 2.4959e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 7.8231e-07, + 1.7043e-06, 2.1961e-06], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 1.0595e-05, + 8.0317e-06, 2.5146e-07], + [ 0.0000e+00, 0.0000e+00, 8.9873e-08, ..., 3.7793e-06, + 3.3546e-06, 1.4141e-05]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0080, -0.0171, 0.0003, 0.0291, -0.0035, 0.0319, 0.0033, 0.0173, + 0.0002, -0.0111], device='cuda:0'), grad: tensor([ 5.2214e-04, -3.6210e-06, 1.3359e-05, -9.3997e-05, 5.8919e-05, + 4.0627e-04, -1.0681e-03, 2.5332e-05, 2.8938e-05, 1.0985e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 214.38, cls_loss 0.0032 cls_loss_mapping 0.0072 cls_loss_causal 0.5833 re_mapping 0.0077 re_causal 0.0234 /// teacc 98.82 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0260, 0.0048, -0.0580, ..., 0.0173, -0.0341, -0.0030], + [ 0.0392, 0.0046, -0.0196, ..., 0.0186, -0.0316, -0.2076], + [-0.0108, 0.0172, 0.0097, ..., -0.0652, 0.0826, -0.0437], + ..., + [-0.0032, -0.0202, -0.0091, ..., -0.0684, -0.1165, 0.0253], + [-0.0155, -0.0139, -0.0145, ..., -0.0701, 0.0417, -0.1432], + [-0.0955, -0.0019, -0.0021, ..., -0.0791, -0.0788, -0.0714]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.8673e-08, 0.0000e+00, ..., 7.1190e-06, + 7.1004e-06, 3.6322e-08], + [ 0.0000e+00, 1.0105e-07, 0.0000e+00, ..., 5.8860e-07, + 1.1595e-06, 1.8906e-07], + [ 0.0000e+00, -9.6206e-07, 0.0000e+00, ..., 1.6410e-06, + -3.3733e-06, 6.9384e-08], + ..., + [ 0.0000e+00, 2.9849e-07, 0.0000e+00, ..., 1.1735e-07, + 1.6736e-06, 3.0594e-07], + [ 0.0000e+00, 1.2340e-07, 0.0000e+00, ..., 1.2945e-06, + 2.2762e-06, 9.9186e-08], + [ 0.0000e+00, 2.6543e-08, 0.0000e+00, ..., 6.7009e-07, + 1.5777e-06, 8.7544e-07]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0074, -0.0172, 0.0003, 0.0292, -0.0035, 0.0316, 0.0039, 0.0172, + 0.0003, -0.0111], device='cuda:0'), grad: tensor([ 1.0736e-05, -2.5444e-06, -6.3442e-06, 1.2629e-05, 8.6203e-06, + -2.1696e-05, -1.8954e-05, 5.1670e-06, 7.2978e-06, 5.0366e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 214.22, cls_loss 0.0031 cls_loss_mapping 0.0060 cls_loss_causal 0.5597 re_mapping 0.0079 re_causal 0.0238 /// teacc 99.06 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0258, 0.0049, -0.0582, ..., 0.0167, -0.0349, -0.0032], + [ 0.0392, 0.0043, -0.0197, ..., 0.0186, -0.0320, -0.2080], + [-0.0107, 0.0172, 0.0095, ..., -0.0651, 0.0832, -0.0433], + ..., + [-0.0033, -0.0202, -0.0093, ..., -0.0684, -0.1168, 0.0254], + [-0.0155, -0.0140, -0.0150, ..., -0.0705, 0.0426, -0.1432], + [-0.0957, -0.0019, -0.0020, ..., -0.0796, -0.0806, -0.0714]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.7032e-08, 0.0000e+00, ..., 1.2553e-04, + 3.8743e-05, 5.0142e-06], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 7.8045e-07, + 1.2917e-06, 4.5635e-08], + [ 0.0000e+00, -7.5437e-08, 0.0000e+00, ..., 2.3004e-07, + -3.7774e-06, -1.7323e-07], + ..., + [ 0.0000e+00, 6.0536e-09, 0.0000e+00, ..., 6.0536e-08, + 1.1371e-06, 1.2433e-07], + [ 0.0000e+00, 1.1642e-08, 0.0000e+00, ..., 3.5223e-06, + 5.6028e-06, 8.0559e-08], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.5385e-06, + 2.0973e-06, 1.8720e-07]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0065, -0.0167, 0.0002, 0.0299, -0.0042, 0.0317, 0.0043, 0.0167, + 0.0011, -0.0114], device='cuda:0'), grad: tensor([ 2.3592e-04, 4.2915e-06, -6.6869e-06, 4.1947e-06, 5.2392e-05, + -8.1658e-05, -2.4045e-04, 2.0880e-06, 2.9489e-05, 2.2585e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 214.15, cls_loss 0.0036 cls_loss_mapping 0.0063 cls_loss_causal 0.5443 re_mapping 0.0079 re_causal 0.0224 /// teacc 99.03 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0259, 0.0039, -0.0583, ..., 0.0163, -0.0362, -0.0032], + [ 0.0395, 0.0035, -0.0197, ..., 0.0184, -0.0324, -0.2089], + [-0.0099, 0.0173, 0.0095, ..., -0.0644, 0.0825, -0.0430], + ..., + [-0.0035, -0.0203, -0.0093, ..., -0.0693, -0.1177, 0.0253], + [-0.0158, -0.0158, -0.0150, ..., -0.0727, 0.0437, -0.1437], + [-0.0959, -0.0019, -0.0023, ..., -0.0796, -0.0813, -0.0715]], + device='cuda:0'), grad: tensor([[ 2.2212e-07, 0.0000e+00, 0.0000e+00, ..., 1.3433e-05, + 1.6436e-05, 4.0792e-06], + [-2.3395e-05, 0.0000e+00, 0.0000e+00, ..., 8.0019e-06, + 3.2112e-06, -6.2995e-06], + [ 1.2383e-05, 0.0000e+00, 0.0000e+00, ..., 1.4149e-05, + 5.6356e-05, 9.4324e-06], + ..., + [ 3.8743e-07, 0.0000e+00, 0.0000e+00, ..., 1.4566e-06, + 4.9062e-06, 1.5507e-07], + [ 6.9439e-06, 0.0000e+00, 0.0000e+00, ..., -1.1408e-04, + -3.7432e-04, -3.4332e-05], + [ 4.9826e-08, 0.0000e+00, 0.0000e+00, ..., 2.1197e-06, + 3.1125e-06, 1.1548e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0056, -0.0160, -0.0009, 0.0295, -0.0042, 0.0324, 0.0049, 0.0170, + 0.0010, -0.0116], device='cuda:0'), grad: tensor([ 3.3647e-05, -1.5843e-04, 1.8728e-04, 4.9062e-06, 6.3539e-05, + 3.5286e-04, 6.2466e-05, 2.9523e-07, -5.5885e-04, 1.1683e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 214.19, cls_loss 0.0049 cls_loss_mapping 0.0085 cls_loss_causal 0.5761 re_mapping 0.0083 re_causal 0.0234 /// teacc 98.88 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0258, 0.0039, -0.0587, ..., 0.0169, -0.0342, -0.0028], + [ 0.0403, 0.0032, -0.0197, ..., 0.0181, -0.0326, -0.2102], + [-0.0095, 0.0174, 0.0093, ..., -0.0666, 0.0833, -0.0437], + ..., + [-0.0042, -0.0204, -0.0094, ..., -0.0701, -0.1198, 0.0269], + [-0.0162, -0.0160, -0.0145, ..., -0.0737, 0.0435, -0.1439], + [-0.0973, -0.0019, -0.0026, ..., -0.0774, -0.0824, -0.0715]], + device='cuda:0'), grad: tensor([[ 8.7311e-08, 2.3283e-09, 2.3283e-08, ..., -1.8150e-05, + 1.0058e-06, -1.6242e-05], + [-6.4773e-07, 5.5879e-09, 5.0990e-08, ..., 1.1437e-06, + 1.4631e-06, 6.6422e-06], + [ 7.0548e-08, -3.3062e-08, 2.3982e-08, ..., 1.2629e-06, + 4.3819e-07, 2.3488e-06], + ..., + [ 2.9523e-07, 3.9581e-09, 6.9849e-09, ..., 3.0338e-07, + 2.9965e-07, 2.8647e-06], + [ 1.6601e-07, 7.2177e-09, 3.6485e-07, ..., 9.1866e-06, + 6.2399e-06, 1.0133e-05], + [ 1.8487e-07, 6.9849e-10, 4.8894e-08, ..., 5.3979e-06, + 2.4997e-06, 3.3647e-05]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0054, -0.0162, -0.0007, 0.0299, -0.0051, 0.0323, 0.0041, 0.0176, + 0.0002, -0.0104], device='cuda:0'), grad: tensor([-6.5744e-05, 2.6494e-05, 1.4573e-05, 2.4170e-05, -9.9421e-05, + 1.8775e-05, -4.1187e-05, -4.1306e-05, 5.3287e-05, 1.1009e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 214.18, cls_loss 0.0048 cls_loss_mapping 0.0080 cls_loss_causal 0.5402 re_mapping 0.0083 re_causal 0.0226 /// teacc 98.92 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0261, 0.0035, -0.0590, ..., 0.0168, -0.0352, -0.0029], + [ 0.0420, 0.0003, -0.0200, ..., 0.0168, -0.0330, -0.2115], + [-0.0089, 0.0159, 0.0089, ..., -0.0669, 0.0842, -0.0411], + ..., + [-0.0022, -0.0208, -0.0095, ..., -0.0715, -0.1214, 0.0270], + [-0.0175, -0.0182, -0.0142, ..., -0.0746, 0.0435, -0.1444], + [-0.1003, -0.0020, -0.0029, ..., -0.0776, -0.0837, -0.0710]], + device='cuda:0'), grad: tensor([[ 3.6089e-08, 0.0000e+00, 0.0000e+00, ..., 1.4454e-05, + 1.3024e-05, 1.1665e-07], + [-1.6959e-06, 0.0000e+00, 0.0000e+00, ..., -4.7572e-06, + 6.0424e-06, -5.5097e-06], + [ 7.1479e-08, 0.0000e+00, 0.0000e+00, ..., 1.2917e-06, + -1.9109e-04, 1.4743e-06], + ..., + [ 5.7183e-07, 0.0000e+00, 0.0000e+00, ..., 2.3888e-07, + 2.6450e-06, 1.6997e-07], + [ 6.3330e-08, 0.0000e+00, 0.0000e+00, ..., 5.3365e-07, + 1.7452e-04, 1.1688e-07], + [ 1.7346e-07, 0.0000e+00, 0.0000e+00, ..., 4.3027e-06, + 3.3388e-07, 4.7055e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0055, -0.0157, -0.0004, 0.0305, -0.0061, 0.0321, 0.0048, 0.0173, + -0.0004, -0.0110], device='cuda:0'), grad: tensor([ 1.4938e-05, -5.9694e-05, -2.6798e-04, 1.0811e-05, 2.2218e-05, + 6.5714e-06, -5.8822e-06, 3.9265e-06, 2.6274e-04, 1.2212e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 214.21, cls_loss 0.0033 cls_loss_mapping 0.0057 cls_loss_causal 0.5346 re_mapping 0.0076 re_causal 0.0232 /// teacc 99.02 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0260, 0.0032, -0.0598, ..., 0.0170, -0.0354, -0.0030], + [ 0.0428, -0.0014, -0.0198, ..., 0.0169, -0.0331, -0.2134], + [-0.0090, 0.0163, 0.0083, ..., -0.0667, 0.0845, -0.0410], + ..., + [-0.0022, -0.0217, -0.0097, ..., -0.0722, -0.1223, 0.0265], + [-0.0169, -0.0212, -0.0142, ..., -0.0748, 0.0440, -0.1444], + [-0.1021, -0.0034, -0.0036, ..., -0.0778, -0.0850, -0.0714]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6196e-06, + -7.8231e-08, 2.3283e-09], + [-2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 5.3085e-08, + 5.8673e-08, 3.0268e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.5181e-07, + 7.4971e-08, 5.5879e-09], + ..., + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 3.3528e-08, + 7.1712e-08, 2.3749e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.2480e-07, + 2.6869e-07, 3.0268e-08], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 6.8313e-07, + 1.0431e-07, 8.7544e-08]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0055, -0.0147, -0.0013, 0.0301, -0.0056, 0.0319, 0.0048, 0.0170, + -0.0002, -0.0112], device='cuda:0'), grad: tensor([-1.1455e-06, 5.2489e-06, 4.3847e-06, -3.4213e-05, 9.4716e-07, + 3.0369e-05, 8.0932e-07, -2.0131e-05, 7.6443e-06, 6.0610e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 214.50, cls_loss 0.0044 cls_loss_mapping 0.0067 cls_loss_causal 0.5498 re_mapping 0.0074 re_causal 0.0210 /// teacc 98.99 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0263, 0.0032, -0.0613, ..., 0.0172, -0.0360, -0.0033], + [ 0.0415, -0.0016, -0.0205, ..., 0.0164, -0.0335, -0.2144], + [-0.0090, 0.0163, 0.0072, ..., -0.0672, 0.0849, -0.0411], + ..., + [-0.0025, -0.0219, -0.0098, ..., -0.0725, -0.1234, 0.0262], + [-0.0147, -0.0216, -0.0132, ..., -0.0749, 0.0443, -0.1448], + [-0.1044, -0.0035, -0.0033, ..., -0.0780, -0.0835, -0.0711]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 1.3039e-08, 3.7253e-08, ..., -4.7795e-06, + 1.9427e-06, -1.0975e-05], + [-9.3132e-08, 1.1642e-08, -3.1050e-06, ..., 1.9372e-07, + 4.3735e-06, 5.8813e-07], + [ 1.6764e-08, -1.7090e-07, 1.5972e-07, ..., 3.7765e-07, + -5.2415e-06, -1.3746e-05], + ..., + [ 4.0513e-08, 4.4238e-08, 9.3412e-07, ..., 1.3039e-07, + 1.4499e-05, 9.0748e-06], + [ 1.2573e-08, 2.9337e-08, 1.3784e-07, ..., 3.3583e-06, + -3.6597e-05, 1.3802e-06], + [ 1.0245e-08, 3.7253e-09, 7.6834e-07, ..., 7.8827e-06, + 2.0355e-05, 6.1989e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0053, -0.0147, -0.0013, 0.0301, -0.0066, 0.0320, 0.0036, 0.0171, + -0.0006, -0.0093], device='cuda:0'), grad: tensor([-2.5138e-05, -6.7912e-06, -3.6713e-06, 1.2234e-05, 1.0461e-05, + -7.2181e-05, 5.4657e-05, 3.4511e-05, -6.7174e-05, 6.2943e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 214.37, cls_loss 0.0026 cls_loss_mapping 0.0053 cls_loss_causal 0.5637 re_mapping 0.0075 re_causal 0.0235 /// teacc 98.92 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0264, 0.0028, -0.0622, ..., 0.0173, -0.0366, -0.0032], + [ 0.0424, -0.0020, -0.0204, ..., 0.0162, -0.0338, -0.2150], + [-0.0090, 0.0164, 0.0071, ..., -0.0675, 0.0851, -0.0415], + ..., + [-0.0036, -0.0220, -0.0099, ..., -0.0727, -0.1239, 0.0263], + [-0.0147, -0.0221, -0.0137, ..., -0.0752, 0.0444, -0.1451], + [-0.1071, -0.0035, -0.0036, ..., -0.0782, -0.0840, -0.0723]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 8.1882e-06, + 1.2785e-05, 5.6550e-06], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.6764e-07, + 8.9174e-07, 8.4750e-07], + [ 0.0000e+00, -6.0536e-09, 0.0000e+00, ..., 2.0117e-06, + -1.4111e-05, 1.3094e-06], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 9.7789e-09, + 1.6121e-06, -3.9581e-08], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 2.0508e-06, + 1.6004e-05, 1.5618e-06], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.3388e-07, + 5.3411e-07, -5.9092e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0053, -0.0145, -0.0012, 0.0303, -0.0047, 0.0318, 0.0038, 0.0168, + -0.0007, -0.0109], device='cuda:0'), grad: tensor([ 2.5272e-05, 1.3757e-04, -1.8492e-05, -5.7854e-06, 2.6762e-05, + 7.6219e-06, -4.7714e-05, 4.8876e-06, 5.2929e-05, -1.8299e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 214.65, cls_loss 0.0038 cls_loss_mapping 0.0070 cls_loss_causal 0.5325 re_mapping 0.0074 re_causal 0.0209 /// teacc 98.99 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0273, 0.0028, -0.0660, ..., 0.0175, -0.0371, -0.0034], + [ 0.0424, -0.0025, -0.0179, ..., 0.0161, -0.0341, -0.2164], + [-0.0094, 0.0165, 0.0047, ..., -0.0683, 0.0851, -0.0437], + ..., + [-0.0016, -0.0221, -0.0103, ..., -0.0729, -0.1244, 0.0267], + [-0.0144, -0.0222, -0.0157, ..., -0.0756, 0.0446, -0.1458], + [-0.1093, -0.0035, -0.0069, ..., -0.0784, -0.0844, -0.0724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.2864e-08, 2.1886e-08, ..., -5.8822e-06, + 1.0412e-06, 4.6566e-09], + [ 0.0000e+00, 9.0804e-08, 2.4904e-06, ..., 5.3085e-08, + 9.4622e-06, 3.5390e-08], + [ 0.0000e+00, -6.0014e-06, -2.9709e-06, ..., 1.5041e-07, + -3.3408e-05, 1.4435e-08], + ..., + [ 4.6566e-10, 1.0617e-07, 1.7416e-07, ..., 4.7497e-08, + 1.4817e-06, -1.0813e-06], + [ 0.0000e+00, 5.7276e-08, 3.3062e-08, ..., 2.5984e-07, + -2.7288e-07, 8.8476e-09], + [ 4.6566e-10, 8.4285e-08, 5.8673e-08, ..., 4.2841e-06, + 4.8243e-07, 1.3364e-07]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0055, -0.0146, -0.0018, 0.0300, -0.0038, 0.0320, 0.0041, 0.0178, + -0.0008, -0.0119], device='cuda:0'), grad: tensor([-8.3670e-06, 3.0145e-05, -1.0550e-04, 6.6996e-05, 5.0440e-06, + 9.8161e-07, 1.5795e-06, -1.8673e-07, -2.3888e-07, 9.6411e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 128---------------------------------------------------- +epoch 128, time 230.51, cls_loss 0.0021 cls_loss_mapping 0.0039 cls_loss_causal 0.5381 re_mapping 0.0075 re_causal 0.0224 /// teacc 99.10 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0273, 0.0025, -0.0672, ..., 0.0175, -0.0377, -0.0036], + [ 0.0424, -0.0029, -0.0179, ..., 0.0160, -0.0342, -0.2168], + [-0.0094, 0.0166, 0.0049, ..., -0.0686, 0.0856, -0.0434], + ..., + [-0.0015, -0.0222, -0.0101, ..., -0.0730, -0.1256, 0.0266], + [-0.0143, -0.0229, -0.0164, ..., -0.0759, 0.0444, -0.1462], + [-0.1094, -0.0035, -0.0071, ..., -0.0785, -0.0847, -0.0723]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, -4.3772e-08, 1.8626e-09, ..., 3.0408e-07, + 3.3043e-06, 6.1002e-08], + [-4.0606e-07, 4.6566e-10, 4.6566e-10, ..., 2.4363e-06, + 5.0254e-06, 1.7509e-07], + [-5.8673e-08, 3.7253e-09, 1.3970e-09, ..., 4.9531e-05, + 8.0705e-05, 1.0896e-07], + ..., + [ 2.4587e-07, 9.3132e-10, 0.0000e+00, ..., 7.5437e-08, + 6.4122e-07, 5.5367e-07], + [ 4.3306e-08, 9.3132e-10, 9.3132e-10, ..., 6.2995e-06, + -1.1869e-05, 4.5169e-08], + [ 1.2573e-08, 2.7008e-08, 0.0000e+00, ..., 5.9605e-07, + 2.6599e-06, 4.4517e-06]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0054, -0.0149, -0.0016, 0.0300, -0.0041, 0.0320, 0.0043, 0.0182, + -0.0007, -0.0120], device='cuda:0'), grad: tensor([ 3.6843e-06, 8.0094e-06, 1.8907e-04, 1.3791e-05, 9.7826e-06, + 6.5416e-06, -1.3936e-04, -8.9228e-05, -2.1756e-05, 1.9416e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 214.32, cls_loss 0.0028 cls_loss_mapping 0.0068 cls_loss_causal 0.5137 re_mapping 0.0073 re_causal 0.0209 /// teacc 99.04 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0274, 0.0030, -0.0683, ..., 0.0197, -0.0365, -0.0037], + [ 0.0416, -0.0035, -0.0181, ..., 0.0155, -0.0343, -0.2175], + [-0.0096, 0.0166, 0.0062, ..., -0.0704, 0.0858, -0.0435], + ..., + [-0.0016, -0.0221, -0.0101, ..., -0.0735, -0.1269, 0.0256], + [-0.0129, -0.0233, -0.0172, ..., -0.0763, 0.0431, -0.1437], + [-0.1100, -0.0040, -0.0075, ..., -0.0793, -0.0853, -0.0724]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 1.3970e-08, ..., -1.5525e-06, + 5.9791e-07, -6.4261e-08], + [-9.5926e-08, 1.8626e-09, 1.8626e-09, ..., 2.6450e-07, + 1.8533e-07, 1.3411e-07], + [ 1.2107e-08, -4.6566e-09, 9.3132e-10, ..., 1.8161e-07, + -3.6135e-07, 3.3528e-08], + ..., + [ 4.1910e-08, 9.3132e-10, 0.0000e+00, ..., 8.1025e-08, + 1.7881e-07, 3.6415e-07], + [ 9.3132e-09, 0.0000e+00, 8.3819e-09, ..., 2.5705e-07, + 2.1048e-07, 7.1712e-08], + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.0440e-06, + 1.4063e-07, -4.4703e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0071, -0.0146, -0.0018, 0.0301, -0.0039, 0.0320, 0.0045, 0.0179, + -0.0020, -0.0124], device='cuda:0'), grad: tensor([-6.1058e-06, 9.9838e-07, 2.1514e-07, 1.6233e-06, 7.6741e-07, + 1.1632e-06, -1.6196e-06, -1.2498e-06, 3.1106e-06, 1.0654e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 214.29, cls_loss 0.0027 cls_loss_mapping 0.0046 cls_loss_causal 0.5347 re_mapping 0.0073 re_causal 0.0220 /// teacc 98.93 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0274, 0.0030, -0.0686, ..., 0.0198, -0.0366, -0.0036], + [ 0.0420, -0.0037, -0.0179, ..., 0.0152, -0.0346, -0.2181], + [-0.0098, 0.0165, 0.0057, ..., -0.0707, 0.0866, -0.0431], + ..., + [-0.0019, -0.0222, -0.0099, ..., -0.0737, -0.1291, 0.0258], + [-0.0129, -0.0235, -0.0175, ..., -0.0767, 0.0432, -0.1439], + [-0.1103, -0.0040, -0.0076, ..., -0.0791, -0.0859, -0.0725]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.0489e-08, 0.0000e+00, ..., 1.9595e-06, + 1.2275e-06, 3.4273e-07], + [-1.8626e-09, 6.5193e-09, 0.0000e+00, ..., 7.9162e-08, + 5.0385e-07, 2.7567e-07], + [ 9.3132e-10, -4.4610e-07, 0.0000e+00, ..., 2.8871e-08, + -4.8988e-06, -3.6322e-08], + ..., + [-5.5879e-09, 1.2852e-07, 0.0000e+00, ..., 4.6566e-09, + 2.0731e-06, -1.0960e-05], + [ 1.8626e-09, 1.2852e-07, 0.0000e+00, ..., 1.3728e-06, + 4.9584e-06, 1.2666e-07], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 2.1141e-07, + 3.8650e-07, 3.4515e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0069, -0.0146, -0.0014, 0.0300, -0.0057, 0.0319, 0.0046, 0.0177, + -0.0021, -0.0109], device='cuda:0'), grad: tensor([ 6.3851e-06, 1.6108e-05, -1.6261e-06, 2.8536e-05, 2.4382e-06, + 1.3694e-05, -9.8124e-06, -1.3399e-04, 2.0698e-05, 5.7578e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 214.43, cls_loss 0.0033 cls_loss_mapping 0.0065 cls_loss_causal 0.5632 re_mapping 0.0070 re_causal 0.0209 /// teacc 98.96 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0277, 0.0030, -0.0692, ..., 0.0205, -0.0372, -0.0038], + [ 0.0419, -0.0046, -0.0179, ..., 0.0122, -0.0346, -0.2187], + [-0.0098, 0.0166, 0.0051, ..., -0.0715, 0.0882, -0.0419], + ..., + [-0.0016, -0.0222, -0.0101, ..., -0.0740, -0.1302, 0.0258], + [-0.0128, -0.0248, -0.0177, ..., -0.0770, 0.0420, -0.1460], + [-0.1108, -0.0043, -0.0076, ..., -0.0792, -0.0866, -0.0733]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 6.6310e-07, + 1.3327e-06, 1.3970e-08], + [-1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 3.6322e-08, + 2.8033e-07, 3.3528e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.4622e-07, + -1.8608e-06, 7.4506e-09], + ..., + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 5.3365e-07, 6.6124e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -4.9453e-07, + -3.2838e-06, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1420e-07, + 1.6633e-06, -1.2778e-05]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0080, -0.0148, -0.0010, 0.0308, -0.0054, 0.0316, 0.0046, 0.0178, + -0.0034, -0.0116], device='cuda:0'), grad: tensor([ 4.9561e-05, -6.5193e-07, 2.1964e-05, 1.8761e-05, 1.1161e-05, + 7.3493e-05, 1.1697e-06, 7.4953e-06, -1.0625e-05, -1.7262e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 214.05, cls_loss 0.0026 cls_loss_mapping 0.0055 cls_loss_causal 0.5338 re_mapping 0.0074 re_causal 0.0216 /// teacc 98.81 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0278, 0.0031, -0.0697, ..., 0.0202, -0.0382, -0.0040], + [ 0.0420, -0.0048, -0.0179, ..., 0.0116, -0.0348, -0.2194], + [-0.0097, 0.0166, 0.0050, ..., -0.0731, 0.0884, -0.0416], + ..., + [-0.0019, -0.0222, -0.0104, ..., -0.0746, -0.1313, 0.0250], + [-0.0127, -0.0249, -0.0179, ..., -0.0766, 0.0425, -0.1458], + [-0.1110, -0.0043, -0.0077, ..., -0.0792, -0.0867, -0.0734]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 7.2643e-08, + 4.3679e-07, 1.0757e-06], + [ 9.3132e-10, 7.4506e-09, 0.0000e+00, ..., 2.6077e-08, + 1.6578e-07, 6.5658e-07], + [ 9.3132e-10, -1.5832e-08, 0.0000e+00, ..., -1.8440e-07, + -3.2410e-06, 3.5856e-07], + ..., + [ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 1.7881e-07, + 2.4419e-06, 4.3865e-07], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 7.9162e-08, + 3.5204e-07, 2.1793e-07], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 1.0990e-07, + 1.7788e-07, 7.0371e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0075, -0.0146, -0.0010, 0.0294, -0.0048, 0.0318, 0.0045, 0.0183, + -0.0034, -0.0115], device='cuda:0'), grad: tensor([ 2.8461e-06, 2.8480e-06, -5.4389e-06, -1.7956e-06, -1.9938e-05, + -1.4193e-06, 2.2855e-06, 7.7039e-06, 3.9488e-06, 8.9183e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 214.16, cls_loss 0.0026 cls_loss_mapping 0.0048 cls_loss_causal 0.5719 re_mapping 0.0068 re_causal 0.0214 /// teacc 98.89 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0279, 0.0027, -0.0704, ..., 0.0203, -0.0391, -0.0045], + [ 0.0418, -0.0060, -0.0179, ..., 0.0112, -0.0349, -0.2228], + [-0.0098, 0.0168, 0.0046, ..., -0.0736, 0.0887, -0.0415], + ..., + [-0.0019, -0.0224, -0.0106, ..., -0.0753, -0.1323, 0.0253], + [-0.0134, -0.0257, -0.0176, ..., -0.0774, 0.0427, -0.1448], + [-0.1113, -0.0045, -0.0078, ..., -0.0797, -0.0876, -0.0733]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -8.5682e-08, 0.0000e+00, ..., 3.8091e-07, + 2.2165e-06, 2.6263e-07], + [ 2.2352e-08, 9.3132e-10, 0.0000e+00, ..., -1.9744e-07, + 2.4959e-06, 2.7288e-07], + [ 5.5879e-09, 2.7940e-09, 0.0000e+00, ..., 5.4762e-07, + -3.6806e-05, 9.8720e-08], + ..., + [ 8.3819e-09, 9.3132e-10, 0.0000e+00, ..., 1.1548e-07, + 2.1346e-06, 2.4401e-07], + [ 9.3132e-10, 8.3819e-09, 0.0000e+00, ..., 2.8647e-06, + -1.1265e-05, -2.7101e-06], + [ 1.4901e-08, 1.0245e-08, 0.0000e+00, ..., 1.9372e-07, + 3.6776e-05, 1.9670e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0073, -0.0147, -0.0010, 0.0295, -0.0048, 0.0318, 0.0049, 0.0181, + -0.0034, -0.0115], device='cuda:0'), grad: tensor([ 1.0096e-05, 1.0133e-05, -9.7811e-05, 2.8446e-05, 1.2726e-05, + 1.9774e-05, -2.4706e-05, -5.6326e-05, -1.4670e-05, 1.1241e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 214.53, cls_loss 0.0029 cls_loss_mapping 0.0058 cls_loss_causal 0.5158 re_mapping 0.0070 re_causal 0.0212 /// teacc 98.99 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0283, 0.0023, -0.0708, ..., 0.0203, -0.0400, -0.0064], + [ 0.0439, -0.0065, -0.0180, ..., 0.0111, -0.0353, -0.2219], + [-0.0100, 0.0169, 0.0044, ..., -0.0741, 0.0890, -0.0416], + ..., + [-0.0013, -0.0225, -0.0107, ..., -0.0752, -0.1317, 0.0262], + [-0.0158, -0.0267, -0.0173, ..., -0.0779, 0.0427, -0.1458], + [-0.1117, -0.0046, -0.0078, ..., -0.0792, -0.0870, -0.0740]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -8.5980e-06, + 1.6205e-07, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.5635e-08, + 2.2445e-06, 1.1548e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.0978e-08, + -2.5909e-06, -1.0803e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 7.4506e-09, + 4.1630e-07, 7.5717e-07], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 6.1467e-08, + 7.9162e-08, 2.7940e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 1.6298e-07, + 2.1886e-07, 9.6858e-08]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0068, -0.0149, -0.0011, 0.0296, -0.0047, 0.0313, 0.0051, 0.0189, + -0.0043, -0.0112], device='cuda:0'), grad: tensor([-1.7315e-05, 1.0139e-04, 2.1718e-06, 7.8306e-06, -2.0806e-06, + -3.0994e-06, 1.7166e-05, -1.1039e-04, 1.8133e-06, 2.3730e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 214.22, cls_loss 0.0030 cls_loss_mapping 0.0065 cls_loss_causal 0.5390 re_mapping 0.0071 re_causal 0.0210 /// teacc 98.89 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0284, 0.0024, -0.0716, ..., 0.0215, -0.0394, -0.0080], + [ 0.0442, -0.0069, -0.0186, ..., 0.0108, -0.0355, -0.2222], + [-0.0102, 0.0169, 0.0030, ..., -0.0742, 0.0895, -0.0418], + ..., + [-0.0014, -0.0226, -0.0109, ..., -0.0762, -0.1325, 0.0260], + [-0.0158, -0.0278, -0.0149, ..., -0.0781, 0.0433, -0.1460], + [-0.1123, -0.0047, -0.0080, ..., -0.0801, -0.0868, -0.0744]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.5751e-06, + 5.0515e-06, 5.8766e-07], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.6578e-07, + 3.6415e-07, 2.6915e-07], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 8.9873e-07, + -1.6848e-06, 1.9185e-07], + ..., + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-08, + 1.5777e-06, 7.7300e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.2189e-07, + -1.7695e-08, 3.9209e-07], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 3.2410e-07, + 9.5926e-07, 3.6545e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0078, -0.0155, -0.0009, 0.0302, -0.0060, 0.0303, 0.0047, 0.0196, + -0.0043, -0.0101], device='cuda:0'), grad: tensor([ 1.5825e-05, 1.0030e-06, 1.7677e-06, -5.0336e-05, 1.7703e-05, + 8.4750e-08, -3.0935e-05, -3.6545e-06, 2.5965e-06, 4.5925e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 214.38, cls_loss 0.0029 cls_loss_mapping 0.0060 cls_loss_causal 0.5327 re_mapping 0.0072 re_causal 0.0204 /// teacc 98.80 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0286, 0.0022, -0.0719, ..., 0.0216, -0.0397, -0.0083], + [ 0.0439, -0.0071, -0.0187, ..., 0.0101, -0.0359, -0.2228], + [-0.0106, 0.0170, 0.0027, ..., -0.0749, 0.0901, -0.0417], + ..., + [-0.0014, -0.0227, -0.0117, ..., -0.0763, -0.1338, 0.0260], + [-0.0161, -0.0284, -0.0147, ..., -0.0787, 0.0436, -0.1460], + [-0.1131, -0.0047, -0.0080, ..., -0.0802, -0.0875, -0.0750]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.1526e-07, + 1.3106e-05, 4.9286e-06], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.0654e-05, + 8.1539e-05, 2.4810e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3736e-05, + 2.5773e-04, 7.7486e-05], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.2929e-06, + 1.9073e-05, 6.2250e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.5938e-04, + -1.2264e-03, -3.6502e-04], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.8859e-06, + 7.4580e-06, -2.4676e-05]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0080, -0.0158, -0.0005, 0.0300, -0.0061, 0.0306, 0.0048, 0.0209, + -0.0041, -0.0113], device='cuda:0'), grad: tensor([ 2.0891e-05, 1.4663e-04, 4.5872e-04, 5.2929e-05, 1.8597e-04, + 1.0643e-03, 3.6335e-04, 4.9204e-05, -2.1420e-03, -2.0027e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 214.76, cls_loss 0.0029 cls_loss_mapping 0.0052 cls_loss_causal 0.5549 re_mapping 0.0066 re_causal 0.0204 /// teacc 98.93 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0288, 0.0020, -0.0731, ..., 0.0213, -0.0411, -0.0087], + [ 0.0442, -0.0077, -0.0188, ..., 0.0096, -0.0362, -0.2236], + [-0.0107, 0.0172, 0.0029, ..., -0.0751, 0.0904, -0.0421], + ..., + [-0.0016, -0.0231, -0.0128, ..., -0.0770, -0.1344, 0.0258], + [-0.0171, -0.0288, -0.0148, ..., -0.0765, 0.0470, -0.1449], + [-0.1168, -0.0048, -0.0083, ..., -0.0811, -0.0889, -0.0739]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1064e-04, + 5.4955e-05, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.2201e-07, + -1.5274e-07, 3.1944e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8664e-06, + 3.7253e-09, 2.6077e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2666e-07, + 1.8533e-07, 4.4703e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.6054e-07, + -5.7407e-06, 5.1223e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-06, + 7.2345e-06, 4.9993e-06]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0074, -0.0160, -0.0004, 0.0299, -0.0066, 0.0286, 0.0047, 0.0206, + -0.0012, -0.0111], device='cuda:0'), grad: tensor([ 2.5630e-04, -1.9863e-05, 9.4622e-06, 7.1861e-06, -1.7554e-05, + -9.1270e-06, -2.7132e-04, -4.4443e-06, 2.9281e-06, 4.6462e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 214.25, cls_loss 0.0022 cls_loss_mapping 0.0045 cls_loss_causal 0.5412 re_mapping 0.0067 re_causal 0.0206 /// teacc 98.84 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0289, -0.0006, -0.0732, ..., 0.0211, -0.0421, -0.0089], + [ 0.0442, -0.0110, -0.0188, ..., 0.0095, -0.0363, -0.2238], + [-0.0091, 0.0179, 0.0029, ..., -0.0757, 0.0908, -0.0418], + ..., + [-0.0020, -0.0243, -0.0129, ..., -0.0773, -0.1366, 0.0258], + [-0.0177, -0.0312, -0.0147, ..., -0.0767, 0.0470, -0.1451], + [-0.1179, -0.0050, -0.0084, ..., -0.0814, -0.0894, -0.0747]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., -1.2897e-05, + 4.1444e-07, 7.8231e-08], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 2.4680e-07, + 6.7987e-07, 2.2631e-07], + [ 3.3528e-08, 0.0000e+00, 0.0000e+00, ..., 4.0140e-07, + 2.2631e-07, 1.2200e-07], + ..., + [ 2.7940e-08, 0.0000e+00, 0.0000e+00, ..., 2.1420e-07, + 1.2852e-07, 7.4599e-07], + [ 1.9558e-08, 0.0000e+00, 0.0000e+00, ..., 2.4289e-06, + -1.1727e-05, 1.2387e-07], + [ 5.0291e-08, 0.0000e+00, 0.0000e+00, ..., 5.2005e-06, + -8.7917e-07, 3.7532e-06]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0071, -0.0158, -0.0002, 0.0301, -0.0062, 0.0289, 0.0048, 0.0200, + -0.0012, -0.0115], device='cuda:0'), grad: tensor([-2.1607e-05, 7.5810e-07, 2.5816e-06, 5.7258e-06, 3.7160e-07, + 4.2021e-05, -4.5300e-06, 4.6343e-06, -2.1413e-05, -8.6054e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 214.20, cls_loss 0.0024 cls_loss_mapping 0.0055 cls_loss_causal 0.5458 re_mapping 0.0066 re_causal 0.0202 /// teacc 98.90 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0293, -0.0008, -0.0732, ..., 0.0212, -0.0425, -0.0089], + [ 0.0445, -0.0118, -0.0188, ..., 0.0091, -0.0365, -0.2240], + [-0.0078, 0.0181, 0.0030, ..., -0.0761, 0.0911, -0.0417], + ..., + [-0.0025, -0.0245, -0.0129, ..., -0.0775, -0.1374, 0.0256], + [-0.0185, -0.0319, -0.0147, ..., -0.0762, 0.0478, -0.1456], + [-0.1200, -0.0053, -0.0084, ..., -0.0816, -0.0899, -0.0745]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 7.5996e-07, + 1.8794e-06, 2.2054e-06], + [ 4.2841e-08, 6.5193e-09, 0.0000e+00, ..., 2.1979e-07, + 1.0971e-06, 9.3412e-07], + [ 1.8626e-08, -1.3039e-08, 0.0000e+00, ..., -1.2582e-06, + -4.2468e-06, 6.9104e-07], + ..., + [ 1.9558e-08, 7.4506e-08, 0.0000e+00, ..., 2.6915e-07, + 1.0086e-06, 6.2678e-07], + [ 4.6566e-09, 2.3283e-08, 0.0000e+00, ..., 8.4750e-07, + -5.2527e-06, 1.5553e-07], + [ 1.3970e-08, 1.6764e-08, 0.0000e+00, ..., 1.9185e-07, + 1.2098e-06, 2.8927e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0070, -0.0157, -0.0002, 0.0301, -0.0066, 0.0283, 0.0048, 0.0196, + -0.0007, -0.0110], device='cuda:0'), grad: tensor([ 1.1824e-05, 5.7109e-06, -2.9877e-06, -2.4125e-05, -2.0951e-05, + 1.4290e-05, -2.8238e-06, 2.2091e-06, -9.0972e-06, 2.5943e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 214.46, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.5299 re_mapping 0.0065 re_causal 0.0202 /// teacc 99.02 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0296, -0.0018, -0.0732, ..., 0.0210, -0.0433, -0.0090], + [ 0.0443, -0.0139, -0.0188, ..., 0.0088, -0.0375, -0.2243], + [-0.0076, 0.0185, 0.0030, ..., -0.0762, 0.0923, -0.0416], + ..., + [-0.0021, -0.0252, -0.0129, ..., -0.0779, -0.1383, 0.0254], + [-0.0184, -0.0337, -0.0147, ..., -0.0765, 0.0479, -0.1459], + [-0.1214, -0.0055, -0.0084, ..., -0.0816, -0.0907, -0.0746]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.5991e-06, + -3.7253e-08, 1.0245e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 4.8149e-07, + 2.1532e-06, 2.7940e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.3947e-07, + -1.1601e-05, 2.4214e-08], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.5460e-07, + 5.8375e-06, 3.6322e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.6054e-07, + 5.2489e-06, 1.5832e-08], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 1.7639e-06, + 1.1381e-06, 3.3528e-07]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0068, -0.0163, 0.0004, 0.0298, -0.0064, 0.0285, 0.0047, 0.0198, + -0.0006, -0.0113], device='cuda:0'), grad: tensor([-7.1898e-06, 6.2771e-06, -3.2693e-05, 7.2122e-06, 6.0409e-05, + -5.9232e-06, -6.5386e-05, 1.6823e-05, 1.5102e-05, 5.2713e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 214.13, cls_loss 0.0027 cls_loss_mapping 0.0051 cls_loss_causal 0.5361 re_mapping 0.0068 re_causal 0.0206 /// teacc 98.97 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0297, -0.0022, -0.0732, ..., 0.0210, -0.0440, -0.0101], + [ 0.0444, -0.0172, -0.0188, ..., 0.0086, -0.0378, -0.2245], + [-0.0076, 0.0192, 0.0029, ..., -0.0767, 0.0925, -0.0417], + ..., + [-0.0022, -0.0272, -0.0129, ..., -0.0783, -0.1390, 0.0250], + [-0.0184, -0.0368, -0.0147, ..., -0.0781, 0.0478, -0.1456], + [-0.1221, -0.0059, -0.0084, ..., -0.0817, -0.0913, -0.0748]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 0.0000e+00, 0.0000e+00, ..., -3.0771e-06, + 1.5376e-06, -6.2212e-07], + [ 5.6066e-07, 0.0000e+00, 0.0000e+00, ..., 1.3877e-07, + 4.9919e-07, 1.2191e-06], + [ 3.1665e-08, 0.0000e+00, 0.0000e+00, ..., 6.0163e-07, + 1.1101e-06, 9.3598e-07], + ..., + [ 8.6613e-08, 0.0000e+00, 0.0000e+00, ..., 2.3283e-08, + 7.5158e-07, 1.0394e-06], + [ 3.7253e-08, 0.0000e+00, 0.0000e+00, ..., 2.9802e-06, + 4.1015e-06, 5.5693e-07], + [ 8.7544e-08, 0.0000e+00, 0.0000e+00, ..., 1.8068e-07, + 2.7455e-06, 5.0753e-05]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0064, -0.0161, 0.0003, 0.0298, -0.0061, 0.0289, 0.0048, 0.0196, + -0.0007, -0.0116], device='cuda:0'), grad: tensor([ 3.0659e-06, 1.9908e-05, 1.1988e-05, 3.0965e-05, -2.0325e-04, + -9.2387e-06, -2.0228e-06, 1.9684e-05, 7.2420e-05, 5.6416e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 214.24, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5601 re_mapping 0.0066 re_causal 0.0213 /// teacc 99.01 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0297, -0.0021, -0.0732, ..., 0.0210, -0.0448, -0.0097], + [ 0.0444, -0.0182, -0.0188, ..., 0.0085, -0.0380, -0.2248], + [-0.0073, 0.0195, 0.0030, ..., -0.0766, 0.0934, -0.0416], + ..., + [-0.0023, -0.0277, -0.0129, ..., -0.0789, -0.1412, 0.0251], + [-0.0184, -0.0379, -0.0147, ..., -0.0787, 0.0477, -0.1456], + [-0.1223, -0.0062, -0.0084, ..., -0.0818, -0.0921, -0.0749]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.9116e-08, 0.0000e+00, ..., -4.7795e-06, + 8.7544e-07, 7.9721e-07], + [-0.0000e+00, 2.2352e-08, 0.0000e+00, ..., 1.3411e-07, + 2.2873e-06, 1.1399e-05], + [ 0.0000e+00, -1.5013e-06, 0.0000e+00, ..., 4.6380e-07, + -3.1888e-05, 1.3970e-07], + ..., + [ 0.0000e+00, 1.3281e-06, 0.0000e+00, ..., 7.2643e-08, + 2.0221e-05, 2.7232e-06], + [ 0.0000e+00, 3.7253e-08, 0.0000e+00, ..., 1.8626e-07, + 6.9365e-06, 1.4137e-06], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 8.1398e-07, + 2.8685e-07, 1.0125e-05]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0063, -0.0163, 0.0009, 0.0299, -0.0062, 0.0288, 0.0052, 0.0195, + -0.0006, -0.0119], device='cuda:0'), grad: tensor([-6.3255e-06, 2.8819e-05, -6.2764e-05, 1.7554e-05, -5.1767e-05, + -1.1437e-05, 1.1384e-05, 3.1054e-05, 2.0280e-05, 2.3246e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 214.21, cls_loss 0.0023 cls_loss_mapping 0.0059 cls_loss_causal 0.5612 re_mapping 0.0064 re_causal 0.0206 /// teacc 98.87 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0298, -0.0019, -0.0732, ..., 0.0215, -0.0452, -0.0100], + [ 0.0439, -0.0188, -0.0188, ..., 0.0083, -0.0385, -0.2253], + [-0.0068, 0.0196, 0.0030, ..., -0.0772, 0.0941, -0.0415], + ..., + [-0.0020, -0.0283, -0.0129, ..., -0.0793, -0.1430, 0.0251], + [-0.0185, -0.0381, -0.0147, ..., -0.0787, 0.0481, -0.1458], + [-0.1228, -0.0063, -0.0084, ..., -0.0822, -0.0930, -0.0751]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 5.0291e-08, 0.0000e+00, ..., 7.4506e-09, + 5.0105e-07, 7.2829e-07], + [ 3.7253e-09, 6.1467e-08, 0.0000e+00, ..., 1.3039e-08, + 5.0291e-07, 3.3900e-07], + [ 5.5879e-09, -1.2238e-06, 0.0000e+00, ..., 4.8429e-08, + -9.2089e-06, -1.1846e-06], + ..., + [ 1.8626e-09, 3.7812e-07, 0.0000e+00, ..., 1.8626e-09, + 2.8796e-06, 6.9477e-07], + [ 0.0000e+00, 2.0303e-07, 0.0000e+00, ..., 1.8999e-07, + 2.5537e-06, 4.6007e-07], + [ 3.7253e-09, 4.8429e-08, 0.0000e+00, ..., 1.8626e-08, + 4.9174e-07, 2.4214e-07]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0066, -0.0168, 0.0012, 0.0287, -0.0053, 0.0294, 0.0051, 0.0186, + -0.0001, -0.0124], device='cuda:0'), grad: tensor([ 6.4038e-06, 2.1774e-06, -2.3171e-05, 4.8205e-06, -2.3082e-05, + -2.8871e-07, 1.8761e-05, 7.2718e-06, 9.3430e-06, -2.2389e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 214.04, cls_loss 0.0024 cls_loss_mapping 0.0051 cls_loss_causal 0.5355 re_mapping 0.0065 re_causal 0.0200 /// teacc 98.91 lr 0.00010000 +Epoch 146, weight, value: tensor([[-2.9972e-02, -1.6859e-03, -7.3313e-02, ..., 2.2002e-02, + -4.5504e-02, -1.0078e-02], + [ 4.3854e-02, -1.9340e-02, -1.8838e-02, ..., 8.2824e-03, + -3.8671e-02, -2.2606e-01], + [-8.2891e-03, 1.9488e-02, 2.9622e-03, ..., -7.7830e-02, + 9.4235e-02, -4.1875e-02], + ..., + [ 1.7593e-04, -2.9214e-02, -1.3004e-02, ..., -7.9702e-02, + -1.4291e-01, 2.5086e-02], + [-1.8966e-02, -3.8517e-02, -1.4766e-02, ..., -7.9359e-02, + 4.8101e-02, -1.4603e-01], + [-1.2302e-01, -6.4641e-03, -8.3773e-03, ..., -8.2751e-02, + -9.3163e-02, -7.5578e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.1234e-07, + 4.7125e-07, 5.8115e-07], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 3.0734e-07, 7.7859e-07], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.6858e-08, + -6.3702e-06, 1.2740e-06], + ..., + [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 5.6066e-07, -1.3459e-04], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-07, + -1.1645e-05, 1.1250e-06], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 3.9116e-08, + 1.4275e-05, 3.5223e-06]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0071, -0.0168, 0.0003, 0.0288, -0.0049, 0.0297, 0.0052, 0.0194, + -0.0006, -0.0127], device='cuda:0'), grad: tensor([ 7.1302e-06, 1.6093e-05, -7.4863e-05, 1.3721e-04, 1.9705e-04, + 1.5366e-04, -9.3132e-09, -6.3801e-04, -2.7806e-05, 2.2888e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 214.30, cls_loss 0.0019 cls_loss_mapping 0.0039 cls_loss_causal 0.5207 re_mapping 0.0065 re_causal 0.0200 /// teacc 98.99 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0300, -0.0016, -0.0733, ..., 0.0226, -0.0451, -0.0106], + [ 0.0439, -0.0199, -0.0188, ..., 0.0081, -0.0386, -0.2264], + [-0.0085, 0.0196, 0.0030, ..., -0.0782, 0.0948, -0.0419], + ..., + [ 0.0007, -0.0294, -0.0130, ..., -0.0801, -0.1447, 0.0255], + [-0.0193, -0.0390, -0.0148, ..., -0.0796, 0.0481, -0.1463], + [-0.1232, -0.0066, -0.0084, ..., -0.0828, -0.0934, -0.0752]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1418e-06, + 1.6652e-06, 7.5065e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5087e-07, + 2.7008e-07, 9.1270e-08], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 4.8243e-07, + 2.7008e-07, 2.9057e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-08, + 4.3958e-07, 2.8778e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9057e-07, + 6.3330e-08, 1.5460e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.6858e-08, + 1.3411e-07, 3.1665e-08]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0075, -0.0164, 0.0002, 0.0289, -0.0054, 0.0296, 0.0053, 0.0194, + -0.0008, -0.0125], device='cuda:0'), grad: tensor([ 4.8280e-06, 3.7942e-06, 4.4197e-05, -6.1333e-05, 6.5342e-06, + 4.9993e-06, -2.0459e-05, 5.2750e-06, 1.1206e-05, 9.4064e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 214.31, cls_loss 0.0027 cls_loss_mapping 0.0065 cls_loss_causal 0.5587 re_mapping 0.0066 re_causal 0.0206 /// teacc 98.99 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0301, -0.0020, -0.0733, ..., 0.0225, -0.0460, -0.0111], + [ 0.0439, -0.0226, -0.0188, ..., 0.0080, -0.0390, -0.2269], + [-0.0085, 0.0207, 0.0030, ..., -0.0785, 0.0954, -0.0414], + ..., + [ 0.0008, -0.0317, -0.0130, ..., -0.0805, -0.1453, 0.0249], + [-0.0192, -0.0416, -0.0148, ..., -0.0800, 0.0479, -0.1467], + [-0.1236, -0.0078, -0.0084, ..., -0.0829, -0.0939, -0.0766]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -5.4017e-08, + 5.2527e-07, 4.4703e-08], + [ 3.9116e-08, 0.0000e+00, 0.0000e+00, ..., 1.2107e-07, + 5.0105e-07, 4.3027e-07], + [ 7.4506e-09, -0.0000e+00, 0.0000e+00, ..., 1.0245e-07, + -6.7614e-07, 8.7544e-08], + ..., + [-2.4028e-07, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + 2.6636e-07, -2.4457e-06], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -1.8477e-06, + -2.8133e-05, 9.1270e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.6263e-07, + 4.6194e-07, -5.4762e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0069, -0.0161, 0.0007, 0.0291, -0.0045, 0.0296, 0.0058, 0.0187, + -0.0012, -0.0131], device='cuda:0'), grad: tensor([ 9.6858e-07, 1.0274e-05, -6.3330e-08, 1.3828e-05, 1.8775e-05, + 7.8231e-06, 3.2514e-05, -2.8849e-05, -4.9740e-05, -5.6103e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 214.61, cls_loss 0.0028 cls_loss_mapping 0.0053 cls_loss_causal 0.5317 re_mapping 0.0067 re_causal 0.0204 /// teacc 98.92 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0311, -0.0041, -0.0733, ..., 0.0233, -0.0467, -0.0107], + [ 0.0436, -0.0243, -0.0189, ..., 0.0061, -0.0392, -0.2272], + [-0.0085, 0.0221, 0.0029, ..., -0.0800, 0.0962, -0.0412], + ..., + [ 0.0008, -0.0336, -0.0130, ..., -0.0811, -0.1468, 0.0247], + [-0.0191, -0.0460, -0.0147, ..., -0.0803, 0.0479, -0.1470], + [-0.1248, -0.0080, -0.0084, ..., -0.0823, -0.0948, -0.0760]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.3388e-07, + 9.4064e-07, 8.5682e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 6.7055e-08, + -1.4529e-06, 3.6135e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.0781e-08, + 3.9302e-07, 3.1479e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 3.5577e-07, 1.3784e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9744e-07, + -5.0336e-05, 1.6950e-07], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-08, + 5.1171e-05, 1.3579e-06]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0079, -0.0165, 0.0010, 0.0295, -0.0056, 0.0295, 0.0051, 0.0180, + -0.0013, -0.0117], device='cuda:0'), grad: tensor([ 2.8163e-06, -6.2063e-06, 5.0217e-06, -7.2271e-07, -6.3702e-06, + -1.6373e-06, 1.6298e-06, -3.7625e-07, -3.9983e-04, 4.0579e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 214.19, cls_loss 0.0019 cls_loss_mapping 0.0050 cls_loss_causal 0.5470 re_mapping 0.0066 re_causal 0.0208 /// teacc 99.00 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0313, -0.0050, -0.0733, ..., 0.0226, -0.0485, -0.0118], + [ 0.0431, -0.0250, -0.0189, ..., 0.0059, -0.0399, -0.2276], + [-0.0083, 0.0222, 0.0029, ..., -0.0808, 0.0964, -0.0412], + ..., + [ 0.0008, -0.0335, -0.0130, ..., -0.0815, -0.1470, 0.0247], + [-0.0194, -0.0463, -0.0146, ..., -0.0807, 0.0480, -0.1473], + [-0.1264, -0.0061, -0.0084, ..., -0.0824, -0.0961, -0.0766]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.3681e-06, + -1.6484e-06, 2.2352e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.6950e-07, + 1.6950e-07, 1.6578e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0361e-07, + 2.4214e-07, 3.7253e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + 2.0489e-08, 9.8720e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5491e-06, + 1.3642e-05, 2.3656e-07], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.8720e-08, + -5.8860e-07, -1.4603e-06]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0073, -0.0166, 0.0007, 0.0294, -0.0052, 0.0296, 0.0062, 0.0182, + -0.0012, -0.0123], device='cuda:0'), grad: tensor([-1.1437e-05, 3.0193e-06, 3.7588e-06, 3.9265e-06, 2.6137e-05, + 4.9919e-05, -6.4015e-05, -5.0440e-06, 2.9504e-05, -3.5852e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 214.29, cls_loss 0.0024 cls_loss_mapping 0.0044 cls_loss_causal 0.5414 re_mapping 0.0067 re_causal 0.0205 /// teacc 98.95 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0315, -0.0051, -0.0733, ..., 0.0228, -0.0489, -0.0123], + [ 0.0431, -0.0279, -0.0189, ..., 0.0061, -0.0399, -0.2280], + [-0.0058, 0.0225, 0.0029, ..., -0.0810, 0.0974, -0.0397], + ..., + [ 0.0006, -0.0326, -0.0130, ..., -0.0817, -0.1480, 0.0251], + [-0.0220, -0.0468, -0.0146, ..., -0.0807, 0.0478, -0.1495], + [-0.1267, -0.0062, -0.0084, ..., -0.0826, -0.0967, -0.0768]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1176e-08, 0.0000e+00, ..., -3.1628e-06, + 1.3933e-06, 7.4506e-09], + [ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 1.2107e-07, + 6.5044e-06, 1.6764e-08], + [ 0.0000e+00, -1.6950e-07, 0.0000e+00, ..., 1.0245e-07, + 1.2770e-05, 1.1176e-08], + ..., + [ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 1.6764e-08, + 1.1958e-06, 1.1176e-08], + [ 0.0000e+00, 6.7055e-08, 0.0000e+00, ..., 1.3225e-07, + -1.1110e-04, 1.1176e-08], + [ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 1.3225e-07, + 1.4082e-06, 2.0489e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0073, -0.0167, 0.0007, 0.0310, -0.0051, 0.0296, 0.0057, 0.0173, + -0.0015, -0.0123], device='cuda:0'), grad: tensor([-4.0680e-06, 7.9423e-06, 2.5377e-05, 7.9274e-06, 2.1756e-06, + 1.2589e-04, 3.0726e-05, 3.1441e-06, -2.0468e-04, 5.3756e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 214.27, cls_loss 0.0025 cls_loss_mapping 0.0051 cls_loss_causal 0.5644 re_mapping 0.0063 re_causal 0.0195 /// teacc 98.93 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0316, -0.0058, -0.0734, ..., 0.0224, -0.0503, -0.0125], + [ 0.0431, -0.0288, -0.0190, ..., 0.0059, -0.0400, -0.2282], + [-0.0055, 0.0229, 0.0030, ..., -0.0818, 0.0984, -0.0395], + ..., + [ 0.0007, -0.0329, -0.0131, ..., -0.0818, -0.1486, 0.0256], + [-0.0223, -0.0481, -0.0143, ..., -0.0809, 0.0475, -0.1499], + [-0.1280, -0.0062, -0.0084, ..., -0.0822, -0.0971, -0.0772]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 3.9302e-07, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.3958e-07, 5.0850e-07], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -7.0147e-06, 8.3819e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.6782e-06, 1.5087e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 3.3118e-06, 5.1409e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 3.6694e-07, 1.1630e-05]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0062, -0.0162, 0.0011, 0.0311, -0.0049, 0.0300, 0.0054, 0.0171, + -0.0023, -0.0124], device='cuda:0'), grad: tensor([ 1.9614e-06, 3.8557e-06, -5.5879e-08, 1.0216e-04, -3.4481e-05, + 8.9779e-07, 1.5516e-06, -1.2851e-04, 1.3530e-05, 3.8803e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 214.28, cls_loss 0.0019 cls_loss_mapping 0.0041 cls_loss_causal 0.5370 re_mapping 0.0060 re_causal 0.0197 /// teacc 98.98 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0316, -0.0058, -0.0740, ..., 0.0218, -0.0513, -0.0120], + [ 0.0432, -0.0292, -0.0189, ..., 0.0058, -0.0402, -0.2283], + [-0.0056, 0.0230, 0.0027, ..., -0.0824, 0.0986, -0.0395], + ..., + [ 0.0007, -0.0330, -0.0136, ..., -0.0822, -0.1488, 0.0256], + [-0.0223, -0.0482, -0.0144, ..., -0.0811, 0.0475, -0.1500], + [-0.1289, -0.0063, -0.0085, ..., -0.0817, -0.0967, -0.0774]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2852e-07, 0.0000e+00, ..., -2.6971e-05, + -7.8306e-06, 2.9802e-08], + [ 0.0000e+00, 2.0489e-08, 0.0000e+00, ..., 1.4883e-06, + 1.4212e-06, 2.0452e-06], + [-1.8626e-09, -9.3132e-09, 0.0000e+00, ..., 6.9365e-06, + 8.6427e-07, 1.6578e-07], + ..., + [ 0.0000e+00, 2.2352e-08, 0.0000e+00, ..., 9.6858e-07, + 2.2445e-06, 1.8440e-07], + [ 0.0000e+00, 3.5390e-08, 0.0000e+00, ..., 4.6253e-05, + 5.2214e-05, 1.0990e-07], + [ 0.0000e+00, 3.5390e-08, 0.0000e+00, ..., 1.0528e-05, + 6.3963e-06, 2.4233e-06]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0055, -0.0153, 0.0009, 0.0308, -0.0045, 0.0299, 0.0049, 0.0166, + -0.0024, -0.0120], device='cuda:0'), grad: tensor([-8.2016e-05, 1.9938e-05, 2.6599e-05, 1.7598e-05, -2.2829e-05, + 4.9448e-04, -6.0034e-04, -9.7156e-06, 1.1307e-04, 4.2826e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 214.72, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5333 re_mapping 0.0060 re_causal 0.0188 /// teacc 98.99 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0319, -0.0022, -0.0741, ..., 0.0226, -0.0515, -0.0119], + [ 0.0431, -0.0370, -0.0189, ..., 0.0051, -0.0405, -0.2287], + [-0.0056, 0.0237, 0.0026, ..., -0.0831, 0.0990, -0.0397], + ..., + [ 0.0009, -0.0346, -0.0136, ..., -0.0825, -0.1492, 0.0244], + [-0.0223, -0.0490, -0.0144, ..., -0.0816, 0.0476, -0.1500], + [-0.1299, -0.0074, -0.0085, ..., -0.0822, -0.0976, -0.0775]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., 4.9733e-07, + 6.7987e-07, 5.9605e-08], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 2.6077e-08, + 1.2666e-07, 3.5018e-07], + [ 0.0000e+00, -2.6077e-08, 0.0000e+00, ..., 7.0781e-08, + 8.8662e-07, 1.3597e-07], + ..., + [ 0.0000e+00, -8.1956e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1735e-07, 1.4342e-07], + [ 0.0000e+00, 9.3132e-09, 0.0000e+00, ..., 2.5891e-07, + 1.3672e-06, 3.7067e-07], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 2.4214e-08, + 4.4331e-07, 2.6785e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0064, -0.0154, 0.0007, 0.0310, -0.0044, 0.0294, 0.0053, 0.0164, + -0.0025, -0.0118], device='cuda:0'), grad: tensor([ 1.4808e-06, 9.5740e-07, 2.3171e-06, 7.3761e-06, -1.5795e-05, + -1.3538e-05, 1.8049e-06, -1.1176e-07, 6.0201e-06, 9.4846e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 214.79, cls_loss 0.0022 cls_loss_mapping 0.0043 cls_loss_causal 0.5464 re_mapping 0.0063 re_causal 0.0206 /// teacc 99.00 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0322, -0.0021, -0.0745, ..., 0.0224, -0.0517, -0.0126], + [ 0.0432, -0.0390, -0.0190, ..., 0.0046, -0.0406, -0.2290], + [-0.0056, 0.0239, 0.0019, ..., -0.0838, 0.0993, -0.0397], + ..., + [ 0.0032, -0.0342, -0.0139, ..., -0.0827, -0.1495, 0.0264], + [-0.0223, -0.0493, -0.0139, ..., -0.0830, 0.0474, -0.1501], + [-0.1319, -0.0077, -0.0086, ..., -0.0824, -0.0983, -0.0775]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.4640e-06, + 7.5437e-07, 9.3132e-09], + [ 1.1176e-08, 3.7253e-09, 0.0000e+00, ..., 1.6764e-08, + 1.0818e-05, 3.5390e-08], + [ 3.7253e-09, -1.1176e-08, 0.0000e+00, ..., 1.4901e-08, + 1.1265e-04, 5.5879e-08], + ..., + [ 4.4703e-08, -1.1176e-08, 0.0000e+00, ..., 9.3132e-09, + 3.3081e-06, 9.4995e-08], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., -1.0058e-07, + -2.9707e-04, 1.5087e-07], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 6.5193e-08, + 8.2254e-06, 2.7753e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0062, -0.0157, 0.0008, 0.0301, -0.0039, 0.0301, 0.0052, 0.0167, + -0.0031, -0.0118], device='cuda:0'), grad: tensor([ 4.6901e-06, 1.7196e-05, 1.6797e-04, 6.0081e-04, 2.2396e-05, + -4.1747e-04, 4.2677e-05, 1.5855e-05, -5.0402e-04, 5.0724e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 214.62, cls_loss 0.0022 cls_loss_mapping 0.0046 cls_loss_causal 0.5244 re_mapping 0.0063 re_causal 0.0191 /// teacc 98.99 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0327, -0.0023, -0.0746, ..., 0.0224, -0.0527, -0.0128], + [ 0.0462, -0.0447, -0.0188, ..., 0.0039, -0.0411, -0.2292], + [-0.0072, 0.0244, 0.0016, ..., -0.0865, 0.0994, -0.0397], + ..., + [ 0.0030, -0.0345, -0.0140, ..., -0.0830, -0.1500, 0.0264], + [-0.0226, -0.0499, -0.0140, ..., -0.0850, 0.0459, -0.1501], + [-0.1334, -0.0067, -0.0086, ..., -0.0822, -0.0987, -0.0775]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -1.6242e-05, + 3.1665e-08, 2.0489e-08], + [ 9.3132e-08, 0.0000e+00, 0.0000e+00, ..., 5.2154e-07, + 1.0803e-07, 5.7742e-08], + [ 5.9605e-08, 3.7253e-09, 0.0000e+00, ..., 5.0664e-07, + 7.6368e-08, 2.6077e-08], + ..., + [ 4.2841e-08, 0.0000e+00, 0.0000e+00, ..., 3.9861e-07, + 5.2154e-08, 3.1665e-08], + [ 3.7253e-09, -1.6764e-08, 0.0000e+00, ..., 1.2740e-06, + -2.7940e-07, 2.7940e-08], + [ 2.4214e-08, 0.0000e+00, 0.0000e+00, ..., 4.7609e-06, + 2.2538e-07, 9.1344e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0059, -0.0160, 0.0004, 0.0298, -0.0032, 0.0317, 0.0034, 0.0168, + -0.0042, -0.0115], device='cuda:0'), grad: tensor([-3.0696e-05, -1.8626e-09, 3.4627e-06, -7.2084e-06, -1.2434e-04, + 2.8554e-06, 2.1800e-05, 2.3358e-06, 4.9956e-06, 1.2672e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 214.90, cls_loss 0.0025 cls_loss_mapping 0.0041 cls_loss_causal 0.5232 re_mapping 0.0065 re_causal 0.0190 /// teacc 98.90 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0330, -0.0022, -0.0753, ..., 0.0207, -0.0534, -0.0121], + [ 0.0467, -0.0497, -0.0188, ..., 0.0033, -0.0412, -0.2293], + [-0.0069, 0.0262, 0.0018, ..., -0.0867, 0.1005, -0.0393], + ..., + [ 0.0030, -0.0374, -0.0142, ..., -0.0837, -0.1507, 0.0263], + [-0.0226, -0.0508, -0.0140, ..., -0.0851, 0.0460, -0.1502], + [-0.1344, -0.0075, -0.0087, ..., -0.0830, -0.0992, -0.0777]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.2585e-07, + 4.9174e-07, -1.1921e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.0781e-08, + 9.1270e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 9.3691e-07, + 1.0505e-06, -1.8626e-09], + ..., + [-0.0000e+00, -3.7253e-09, 0.0000e+00, ..., 1.8626e-08, + 4.2841e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0547e-07, + 3.5204e-07, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.5949e-07, + 1.1548e-07, 3.5390e-08]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0043, -0.0164, 0.0010, 0.0293, -0.0030, 0.0316, 0.0052, 0.0178, + -0.0042, -0.0124], device='cuda:0'), grad: tensor([-5.3085e-07, 8.2701e-06, 3.7014e-05, -2.6971e-06, -5.0724e-05, + 1.1809e-06, -4.8354e-06, 8.8960e-06, 4.4703e-06, -1.0375e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 214.54, cls_loss 0.0024 cls_loss_mapping 0.0040 cls_loss_causal 0.5499 re_mapping 0.0062 re_causal 0.0196 /// teacc 98.93 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0333, -0.0024, -0.0756, ..., 0.0208, -0.0536, -0.0122], + [ 0.0469, -0.0537, -0.0189, ..., 0.0033, -0.0415, -0.2294], + [-0.0070, 0.0287, 0.0039, ..., -0.0875, 0.1007, -0.0393], + ..., + [ 0.0032, -0.0374, -0.0147, ..., -0.0845, -0.1526, 0.0264], + [-0.0227, -0.0512, -0.0140, ..., -0.0853, 0.0464, -0.1502], + [-0.1352, -0.0079, -0.0087, ..., -0.0832, -0.0997, -0.0778]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -1.5702e-06, + -5.4948e-07, 9.3132e-09], + [ 0.0000e+00, 1.8626e-09, -6.3330e-07, ..., 1.3039e-08, + 2.4214e-08, 3.3528e-08], + [ 0.0000e+00, -1.3597e-07, 3.0734e-07, ..., 9.3132e-09, + -9.4995e-07, -1.1362e-07], + ..., + [ 0.0000e+00, 3.1665e-08, 1.8254e-07, ..., 9.3132e-09, + 2.2352e-07, 1.0990e-07], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.0058e-07, + 1.8254e-07, 3.7253e-08], + [ 0.0000e+00, -1.8626e-09, 8.0094e-08, ..., 2.1979e-07, + 1.1176e-07, 2.5705e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0043, -0.0169, 0.0007, 0.0291, -0.0009, 0.0314, 0.0057, 0.0182, + -0.0036, -0.0144], device='cuda:0'), grad: tensor([-2.3134e-06, 2.1148e-04, 3.9816e-05, 5.1618e-05, -2.2531e-05, + 1.3299e-06, 5.0589e-06, -2.9898e-04, 3.5260e-06, 1.0565e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 214.63, cls_loss 0.0018 cls_loss_mapping 0.0052 cls_loss_causal 0.5475 re_mapping 0.0063 re_causal 0.0192 /// teacc 99.09 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0340, -0.0023, -0.0766, ..., 0.0212, -0.0542, -0.0122], + [ 0.0468, -0.0545, -0.0188, ..., 0.0030, -0.0417, -0.2296], + [-0.0068, 0.0299, 0.0011, ..., -0.0884, 0.1014, -0.0392], + ..., + [ 0.0035, -0.0379, -0.0137, ..., -0.0851, -0.1532, 0.0264], + [-0.0226, -0.0515, -0.0126, ..., -0.0857, 0.0465, -0.1503], + [-0.1359, -0.0083, -0.0085, ..., -0.0827, -0.0999, -0.0779]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3730e-06, + 1.3039e-07, -2.4643e-06], + [ 1.6764e-08, 0.0000e+00, 0.0000e+00, ..., -1.0118e-05, + -4.0494e-06, 5.7742e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 5.8487e-07, + -1.2144e-06, 1.3039e-07], + ..., + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.4156e-07, + 4.2841e-07, 6.1467e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.5779e-06, + -2.2724e-07, 2.6450e-07], + [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 9.1642e-07, + 2.0862e-07, 5.4762e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0042, -0.0169, 0.0010, 0.0288, -0.0010, 0.0313, 0.0054, 0.0184, + -0.0034, -0.0144], device='cuda:0'), grad: tensor([-2.2680e-05, -1.3137e-04, 2.8554e-06, 2.5276e-06, 8.7768e-06, + 7.8306e-06, 9.0659e-05, 2.2203e-06, 2.9266e-05, 9.8944e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 214.79, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5177 re_mapping 0.0057 re_causal 0.0183 /// teacc 98.91 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0349, -0.0023, -0.0767, ..., 0.0216, -0.0543, -0.0116], + [ 0.0466, -0.0552, -0.0188, ..., 0.0028, -0.0419, -0.2299], + [-0.0071, 0.0303, 0.0012, ..., -0.0896, 0.1015, -0.0394], + ..., + [ 0.0041, -0.0379, -0.0138, ..., -0.0858, -0.1536, 0.0271], + [-0.0211, -0.0517, -0.0126, ..., -0.0858, 0.0467, -0.1499], + [-0.1389, -0.0086, -0.0085, ..., -0.0828, -0.1003, -0.0777]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., -0.0000e+00, + 7.4990e-06, 3.9116e-08], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 2.2352e-08, + 1.9241e-06, 1.5274e-07], + [-1.5646e-07, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 6.7800e-06, -3.6694e-07], + ..., + [-3.7253e-09, -0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 1.6019e-07, 1.5087e-07], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 3.8557e-07, + -2.6479e-05, 2.7381e-07], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 8.9407e-08, + 1.9781e-06, -7.2159e-06]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0046, -0.0169, 0.0007, 0.0290, -0.0019, 0.0312, 0.0064, 0.0186, + -0.0032, -0.0142], device='cuda:0'), grad: tensor([ 1.4320e-05, 4.9639e-04, 5.1320e-05, 1.6123e-05, 1.1164e-04, + 4.5449e-06, 1.3426e-05, -6.4325e-04, -1.7837e-05, -4.6283e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 214.83, cls_loss 0.0019 cls_loss_mapping 0.0030 cls_loss_causal 0.5154 re_mapping 0.0057 re_causal 0.0190 /// teacc 98.91 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0355, -0.0024, -0.0768, ..., 0.0217, -0.0545, -0.0114], + [ 0.0466, -0.0562, -0.0188, ..., 0.0026, -0.0421, -0.2304], + [-0.0074, 0.0305, 0.0013, ..., -0.0898, 0.1015, -0.0395], + ..., + [ 0.0050, -0.0371, -0.0139, ..., -0.0864, -0.1553, 0.0269], + [-0.0210, -0.0518, -0.0126, ..., -0.0858, 0.0469, -0.1501], + [-0.1412, -0.0098, -0.0085, ..., -0.0826, -0.1010, -0.0763]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., -1.3597e-07, + 3.2596e-07, 2.7940e-08], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 1.8626e-08, + 1.9316e-06, 6.3330e-08], + [ 0.0000e+00, 1.4342e-07, 0.0000e+00, ..., 2.6077e-08, + -7.6666e-06, -4.7311e-07], + ..., + [ 0.0000e+00, -1.6205e-07, 0.0000e+00, ..., 0.0000e+00, + 8.0839e-07, 1.1548e-07], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 6.8918e-08, + 7.7672e-07, 6.5193e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + 1.1548e-07, -1.1362e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0045, -0.0169, -0.0003, 0.0294, -0.0028, 0.0320, 0.0060, 0.0185, + -0.0031, -0.0134], device='cuda:0'), grad: tensor([ 2.9430e-06, 2.9862e-05, 8.9183e-06, 8.7321e-06, 8.0243e-06, + 1.8924e-06, 2.1327e-06, -1.1617e-04, 6.1542e-06, 4.7714e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 214.69, cls_loss 0.0018 cls_loss_mapping 0.0028 cls_loss_causal 0.5208 re_mapping 0.0065 re_causal 0.0188 /// teacc 98.91 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0359, -0.0038, -0.0769, ..., 0.0217, -0.0547, -0.0116], + [ 0.0466, -0.0590, -0.0183, ..., 0.0021, -0.0421, -0.2309], + [-0.0077, 0.0317, 0.0012, ..., -0.0900, 0.1020, -0.0396], + ..., + [ 0.0058, -0.0342, -0.0142, ..., -0.0870, -0.1558, 0.0270], + [-0.0210, -0.0548, -0.0126, ..., -0.0861, 0.0468, -0.1502], + [-0.1431, -0.0102, -0.0087, ..., -0.0828, -0.1016, -0.0766]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-08, + 4.1537e-07, 5.5879e-09], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.0990e-07, + 5.9418e-07, 2.6077e-08], + [ 5.5879e-09, -1.8626e-09, 0.0000e+00, ..., 5.2154e-08, + -1.2498e-06, 3.7253e-08], + ..., + [ 2.2352e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 1.2740e-06, 1.1362e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4715e-07, + -1.5013e-05, 7.4506e-09], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-08, + 1.1638e-05, -1.6764e-08]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0045, -0.0167, -0.0004, 0.0290, -0.0025, 0.0318, 0.0063, 0.0190, + -0.0033, -0.0139], device='cuda:0'), grad: tensor([ 1.0282e-06, 1.2647e-06, -1.9185e-06, 1.1791e-06, 5.1819e-06, + 2.9020e-06, -2.3674e-06, 3.6545e-06, -3.3081e-05, 2.2098e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 214.48, cls_loss 0.0018 cls_loss_mapping 0.0029 cls_loss_causal 0.5143 re_mapping 0.0056 re_causal 0.0187 /// teacc 98.96 lr 0.00010000 +Epoch 163, weight, value: tensor([[-3.6175e-02, -4.2249e-03, -7.7493e-02, ..., 2.1844e-02, + -5.5218e-02, -1.2040e-02], + [ 4.5909e-02, -5.9068e-02, -1.8211e-02, ..., -3.1705e-04, + -4.3087e-02, -2.3168e-01], + [-7.7440e-03, 3.2182e-02, 8.4007e-05, ..., -8.7916e-02, + 1.0384e-01, -3.9890e-02], + ..., + [ 5.8335e-03, -3.4449e-02, -1.6251e-02, ..., -8.7635e-02, + -1.5711e-01, 2.6894e-02], + [-2.0657e-02, -5.5914e-02, -1.2688e-02, ..., -8.6360e-02, + 4.7059e-02, -1.5017e-01], + [-1.4551e-01, -1.0856e-02, -8.7226e-03, ..., -8.2605e-02, + -1.0214e-01, -7.6828e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4587e-07, + 3.1423e-06, 7.8231e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.2841e-08, + 1.6857e-06, 4.0047e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3597e-07, + -1.0319e-06, 6.4448e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 2.7195e-06, 7.4506e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0361e-07, + 1.7357e-04, 3.2410e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-08, + 4.4331e-06, 8.6799e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0045, -0.0171, 0.0005, 0.0293, -0.0023, 0.0317, 0.0058, 0.0185, + -0.0029, -0.0140], device='cuda:0'), grad: tensor([ 8.4937e-06, 4.8764e-06, 5.5805e-06, 1.7095e-04, -5.5492e-05, + -6.8140e-04, 5.6833e-05, 1.4909e-05, 4.5347e-04, 2.1085e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 214.38, cls_loss 0.0021 cls_loss_mapping 0.0033 cls_loss_causal 0.5394 re_mapping 0.0059 re_causal 0.0187 /// teacc 98.97 lr 0.00010000 +Epoch 164, weight, value: tensor([[-3.6239e-02, -4.2950e-03, -7.7639e-02, ..., 2.1355e-02, + -5.6479e-02, -1.2102e-02], + [ 4.6074e-02, -5.9127e-02, -1.8193e-02, ..., -7.4355e-04, + -4.3711e-02, -2.3206e-01], + [-7.8192e-03, 3.2255e-02, -2.2185e-04, ..., -8.8292e-02, + 1.0390e-01, -4.0020e-02], + ..., + [ 5.7885e-03, -3.4404e-02, -1.6731e-02, ..., -8.8033e-02, + -1.5773e-01, 2.6760e-02], + [-2.0701e-02, -5.6200e-02, -1.2584e-02, ..., -8.6329e-02, + 4.8093e-02, -1.5018e-01], + [-1.4594e-01, -1.0719e-02, -8.7326e-03, ..., -8.2350e-02, + -1.0287e-01, -7.7015e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.4703e-07, + 6.6683e-07, 8.3819e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.2480e-07, + 1.5181e-06, 2.6077e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8440e-07, + 2.5332e-07, -3.9116e-08], + ..., + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 1.1548e-07, 2.8871e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.8592e-07, + 8.2925e-06, 7.4506e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 6.0536e-08, + 8.6520e-07, 8.1956e-08]], device='cuda:0') +Epoch 164, bias, value: tensor([ 3.2093e-03, -1.6165e-02, -2.8765e-05, 2.9432e-02, -2.1408e-03, + 3.1646e-02, 5.7994e-03, 1.7964e-02, -2.8449e-03, -1.3877e-02], + device='cuda:0'), grad: tensor([ 1.6848e-06, -3.2663e-05, 1.0151e-06, 1.3053e-05, 6.6683e-06, + -4.2737e-05, 7.5251e-06, 1.1288e-06, 4.1485e-05, 2.8349e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 163---------------------------------------------------- +epoch 163, time 230.81, cls_loss 0.0023 cls_loss_mapping 0.0043 cls_loss_causal 0.5119 re_mapping 0.0058 re_causal 0.0177 /// teacc 99.13 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0377, -0.0043, -0.0777, ..., 0.0225, -0.0562, -0.0123], + [ 0.0461, -0.0593, -0.0181, ..., -0.0018, -0.0442, -0.2327], + [-0.0079, 0.0323, -0.0004, ..., -0.0910, 0.1039, -0.0403], + ..., + [ 0.0060, -0.0342, -0.0169, ..., -0.0881, -0.1577, 0.0262], + [-0.0207, -0.0566, -0.0126, ..., -0.0868, 0.0481, -0.1502], + [-0.1474, -0.0109, -0.0087, ..., -0.0833, -0.1036, -0.0773]], + device='cuda:0'), grad: tensor([[ 1.0524e-07, 9.3132e-10, 0.0000e+00, ..., -4.4424e-07, + 1.8040e-06, 1.4808e-07], + [ 1.1455e-07, 9.3132e-09, 0.0000e+00, ..., 3.3528e-08, + 1.4091e-06, 1.6671e-07], + [-1.8522e-05, 5.5879e-09, 0.0000e+00, ..., 5.7742e-08, + -3.9250e-05, -2.5287e-05], + ..., + [ 1.4836e-06, -6.7055e-08, 0.0000e+00, ..., 4.6566e-09, + 3.7607e-06, 2.0936e-06], + [ 1.2547e-05, 3.7253e-09, 0.0000e+00, ..., 9.4064e-08, + 4.7421e-04, 1.7151e-05], + [ 2.7940e-09, 3.8184e-08, 0.0000e+00, ..., 3.0454e-07, + -5.7173e-04, 1.0245e-08]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0040, -0.0154, -0.0006, 0.0292, -0.0023, 0.0321, 0.0061, 0.0151, + -0.0008, -0.0138], device='cuda:0'), grad: tensor([ 6.1840e-06, 5.4464e-06, -2.8834e-05, 4.0196e-06, 3.9911e-04, + 4.3225e-04, 7.9274e-06, 6.0052e-06, 1.8330e-03, -2.6665e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.59, cls_loss 0.0017 cls_loss_mapping 0.0041 cls_loss_causal 0.5276 re_mapping 0.0059 re_causal 0.0179 /// teacc 99.10 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0380, -0.0043, -0.0782, ..., 0.0234, -0.0563, -0.0123], + [ 0.0461, -0.0594, -0.0178, ..., -0.0023, -0.0447, -0.2332], + [-0.0075, 0.0324, -0.0004, ..., -0.0914, 0.1043, -0.0401], + ..., + [ 0.0061, -0.0342, -0.0171, ..., -0.0880, -0.1580, 0.0262], + [-0.0211, -0.0571, -0.0129, ..., -0.0870, 0.0479, -0.1505], + [-0.1478, -0.0110, -0.0088, ..., -0.0836, -0.1022, -0.0778]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.8394e-07, + 7.8231e-08, 1.8626e-09], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + 2.7008e-06, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + -2.8964e-06, 2.7940e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 1.0803e-07, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.5146e-08, + 6.6124e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.8918e-08, + -6.7241e-07, 7.2643e-08]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0045, -0.0147, -0.0007, 0.0290, -0.0028, 0.0321, 0.0061, 0.0148, + -0.0014, -0.0127], device='cuda:0'), grad: tensor([-6.9849e-08, 3.6508e-06, 2.0787e-05, -3.0175e-05, 1.5981e-06, + 2.3730e-06, 5.8208e-07, 5.1185e-06, 9.9614e-06, -1.3925e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 214.44, cls_loss 0.0019 cls_loss_mapping 0.0044 cls_loss_causal 0.5231 re_mapping 0.0061 re_causal 0.0183 /// teacc 99.08 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0384, -0.0043, -0.0782, ..., 0.0235, -0.0569, -0.0127], + [ 0.0461, -0.0594, -0.0173, ..., -0.0033, -0.0451, -0.2339], + [-0.0075, 0.0324, -0.0006, ..., -0.0915, 0.1049, -0.0404], + ..., + [ 0.0062, -0.0341, -0.0171, ..., -0.0882, -0.1586, 0.0259], + [-0.0210, -0.0571, -0.0129, ..., -0.0871, 0.0480, -0.1505], + [-0.1486, -0.0110, -0.0089, ..., -0.0833, -0.1027, -0.0785]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.7265e-07, + 2.7958e-06, 2.7940e-09], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 3.9116e-08, + 1.1455e-07, 3.7253e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -1.6820e-06, + -5.1782e-06, 1.7695e-08], + ..., + [ 0.0000e+00, -2.6077e-08, 0.0000e+00, ..., 1.3970e-08, + 3.3155e-07, 1.8626e-08], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 5.9605e-08, + 2.5798e-07, 6.5193e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 4.6566e-08, + 7.8231e-08, -4.5635e-08]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0043, -0.0154, -0.0006, 0.0290, -0.0022, 0.0319, 0.0062, 0.0148, + -0.0015, -0.0123], device='cuda:0'), grad: tensor([ 7.7188e-06, -4.2655e-07, -1.4022e-05, 7.8306e-06, 1.3737e-06, + -3.7737e-06, 7.5251e-07, 6.3609e-07, 1.2172e-06, -1.3104e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 214.54, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.4984 re_mapping 0.0058 re_causal 0.0181 /// teacc 99.08 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0385, -0.0043, -0.0793, ..., 0.0250, -0.0562, -0.0122], + [ 0.0462, -0.0594, -0.0168, ..., -0.0038, -0.0453, -0.2347], + [-0.0074, 0.0324, -0.0011, ..., -0.0924, 0.1053, -0.0408], + ..., + [ 0.0063, -0.0342, -0.0176, ..., -0.0895, -0.1601, 0.0258], + [-0.0208, -0.0573, -0.0145, ..., -0.0864, 0.0487, -0.1499], + [-0.1493, -0.0110, -0.0091, ..., -0.0841, -0.1032, -0.0785]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.4100e-06, + 2.9802e-08, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.1420e-08, + 1.2573e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8871e-08, + -1.2396e-06, -1.6764e-08], + ..., + [-6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 5.2713e-07, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.2072e-07, + -2.4401e-07, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.0094e-08, + 3.2969e-07, -3.8184e-08]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0054, -0.0150, -0.0006, 0.0288, -0.0023, 0.0318, 0.0055, 0.0144, + -0.0014, -0.0122], device='cuda:0'), grad: tensor([-1.3886e-06, 4.4145e-07, -1.8841e-06, -3.2634e-05, 8.6706e-07, + 5.1558e-06, 2.1532e-06, 1.4165e-06, 2.2441e-05, 3.3937e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 214.60, cls_loss 0.0024 cls_loss_mapping 0.0045 cls_loss_causal 0.5148 re_mapping 0.0054 re_causal 0.0168 /// teacc 98.89 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0389, -0.0043, -0.0824, ..., 0.0271, -0.0568, -0.0119], + [ 0.0460, -0.0594, -0.0189, ..., -0.0044, -0.0461, -0.2377], + [-0.0075, 0.0325, -0.0042, ..., -0.0930, 0.1062, -0.0407], + ..., + [ 0.0070, -0.0342, -0.0147, ..., -0.0902, -0.1613, 0.0258], + [-0.0209, -0.0574, -0.0175, ..., -0.0869, 0.0487, -0.1499], + [-0.1505, -0.0110, -0.0106, ..., -0.0882, -0.1046, -0.0780]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 4.5262e-07, + 1.1427e-06, 1.9744e-07], + [ 1.3970e-08, 9.3132e-10, 0.0000e+00, ..., 5.0887e-06, + 1.3195e-05, 2.8778e-07], + [ 6.5193e-09, 1.8626e-09, 0.0000e+00, ..., -6.6042e-05, + -1.7333e-04, 2.8871e-07], + ..., + [ 1.2107e-08, -8.3819e-09, 0.0000e+00, ..., 9.8720e-08, + 2.1271e-06, 1.4994e-07], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.9209e-07, + 9.3319e-07, 7.2177e-07], + [ 2.7008e-08, 9.3132e-10, 0.0000e+00, ..., 8.4750e-08, + 1.8720e-07, 1.4889e-04]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0072, -0.0157, -0.0004, 0.0283, -0.0029, 0.0322, 0.0056, 0.0152, + -0.0014, -0.0126], device='cuda:0'), grad: tensor([ 2.6524e-06, 1.9863e-05, -3.2854e-04, 1.1921e-06, -4.0936e-04, + -9.6858e-08, 2.9421e-04, 4.2468e-06, 4.4294e-06, 4.1032e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 214.68, cls_loss 0.0018 cls_loss_mapping 0.0036 cls_loss_causal 0.5260 re_mapping 0.0058 re_causal 0.0181 /// teacc 99.05 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0391, -0.0045, -0.0825, ..., 0.0255, -0.0596, -0.0120], + [ 0.0462, -0.0594, -0.0188, ..., -0.0039, -0.0466, -0.2388], + [-0.0075, 0.0325, -0.0041, ..., -0.0926, 0.1072, -0.0407], + ..., + [ 0.0071, -0.0341, -0.0147, ..., -0.0905, -0.1625, 0.0276], + [-0.0210, -0.0575, -0.0178, ..., -0.0879, 0.0486, -0.1500], + [-0.1515, -0.0111, -0.0107, ..., -0.0885, -0.1051, -0.0782]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.4261e-08, + 1.8161e-07, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.2154e-08, + 8.7544e-08, 1.6764e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.1665e-08, + 8.3819e-09, 2.8871e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.0978e-08, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.2399e-07, + -3.1386e-07, 6.5193e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7008e-08, + 1.7788e-07, 9.5926e-08]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0054, -0.0152, 0.0003, 0.0283, -0.0026, 0.0318, 0.0057, 0.0155, + -0.0020, -0.0127], device='cuda:0'), grad: tensor([ 9.8441e-07, 1.1019e-05, 4.0680e-06, 1.7256e-05, 6.2212e-07, + 4.0978e-06, -2.3413e-06, -7.3671e-05, 1.6823e-05, 2.1085e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.61, cls_loss 0.0023 cls_loss_mapping 0.0046 cls_loss_causal 0.5360 re_mapping 0.0055 re_causal 0.0168 /// teacc 99.05 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0392, -0.0047, -0.0825, ..., 0.0251, -0.0607, -0.0116], + [ 0.0459, -0.0595, -0.0188, ..., -0.0051, -0.0474, -0.2392], + [-0.0075, 0.0331, -0.0042, ..., -0.0927, 0.1080, -0.0409], + ..., + [ 0.0076, -0.0344, -0.0146, ..., -0.0909, -0.1628, 0.0276], + [-0.0211, -0.0593, -0.0178, ..., -0.0861, 0.0491, -0.1502], + [-0.1531, -0.0112, -0.0107, ..., -0.0890, -0.1064, -0.0788]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -8.0094e-08, + 4.5635e-08, 2.7940e-09], + [ 1.3132e-07, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 2.6822e-07, 1.9558e-07], + [ 2.6077e-08, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + -1.3039e-07, 4.0978e-08], + ..., + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.2014e-07, 1.3970e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + -2.5574e-06, -2.8871e-08], + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 2.4214e-08, + 1.0012e-06, 4.0978e-08]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0048, -0.0149, 0.0002, 0.0285, -0.0025, 0.0318, 0.0060, 0.0161, + -0.0023, -0.0134], device='cuda:0'), grad: tensor([ 6.3330e-08, 1.2647e-06, 1.4137e-06, -2.9318e-06, -1.6512e-06, + 1.0449e-06, 8.5030e-07, 1.7546e-06, -4.4145e-06, 2.5891e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 214.54, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.5321 re_mapping 0.0054 re_causal 0.0175 /// teacc 99.02 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0397, -0.0048, -0.0827, ..., 0.0251, -0.0614, -0.0111], + [ 0.0465, -0.0596, -0.0188, ..., -0.0081, -0.0484, -0.2398], + [-0.0077, 0.0333, -0.0023, ..., -0.0941, 0.1091, -0.0413], + ..., + [ 0.0076, -0.0345, -0.0154, ..., -0.0921, -0.1639, 0.0276], + [-0.0214, -0.0599, -0.0179, ..., -0.0862, 0.0489, -0.1503], + [-0.1542, -0.0112, -0.0107, ..., -0.0891, -0.1055, -0.0793]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.5832e-07, + 9.7416e-07, 1.8626e-09], + [-3.0268e-07, 0.0000e+00, 0.0000e+00, ..., 5.7649e-07, + 4.2580e-06, 1.1083e-07], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 2.4568e-06, + 9.7528e-06, 1.0245e-08], + ..., + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 2.8312e-07, + 1.2387e-06, 1.2107e-08], + [ 3.3528e-08, 0.0000e+00, 0.0000e+00, ..., -2.9951e-05, + -1.3101e-04, 6.5193e-09], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.1548e-07, + 2.6543e-07, 5.2154e-08]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0046, -0.0147, 0.0006, 0.0295, -0.0023, 0.0300, 0.0069, 0.0158, + -0.0032, -0.0124], device='cuda:0'), grad: tensor([ 1.5758e-06, 4.2543e-06, 1.8179e-05, -6.3777e-06, 3.0454e-06, + 1.7893e-04, 2.7508e-05, 5.4836e-06, -2.3234e-04, -2.3749e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 214.83, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5325 re_mapping 0.0054 re_causal 0.0180 /// teacc 98.92 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0400, -0.0048, -0.0827, ..., 0.0244, -0.0622, -0.0098], + [ 0.0465, -0.0596, -0.0188, ..., -0.0085, -0.0489, -0.2400], + [-0.0077, 0.0333, -0.0024, ..., -0.0942, 0.1069, -0.0407], + ..., + [ 0.0077, -0.0345, -0.0153, ..., -0.0925, -0.1614, 0.0276], + [-0.0214, -0.0600, -0.0179, ..., -0.0858, 0.0490, -0.1504], + [-0.1548, -0.0112, -0.0108, ..., -0.0894, -0.1057, -0.0795]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + 8.7544e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.2352e-08, + 1.3690e-07, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 2.1420e-08, + -3.6787e-07, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 9.3132e-10, + 1.5739e-07, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 3.9116e-08, + 6.1095e-07, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.1176e-08, + 5.5600e-07, 2.0489e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0039, -0.0149, -0.0019, 0.0293, -0.0025, 0.0301, 0.0073, 0.0181, + -0.0033, -0.0120], device='cuda:0'), grad: tensor([ 3.2876e-07, -9.5516e-06, 2.0638e-06, 3.8445e-06, -1.8165e-05, + -3.6377e-06, 1.2852e-07, 1.2301e-05, 2.0321e-06, 1.0610e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 214.55, cls_loss 0.0032 cls_loss_mapping 0.0049 cls_loss_causal 0.5043 re_mapping 0.0057 re_causal 0.0163 /// teacc 98.98 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0425, -0.0048, -0.0828, ..., 0.0247, -0.0624, -0.0100], + [ 0.0461, -0.0597, -0.0188, ..., -0.0110, -0.0519, -0.2403], + [-0.0059, 0.0334, -0.0040, ..., -0.0943, 0.1080, -0.0402], + ..., + [ 0.0077, -0.0345, -0.0143, ..., -0.0937, -0.1615, 0.0276], + [-0.0215, -0.0601, -0.0179, ..., -0.0844, 0.0505, -0.1505], + [-0.1591, -0.0113, -0.0108, ..., -0.0890, -0.1060, -0.0797]], + device='cuda:0'), grad: tensor([[ 9.8720e-08, 0.0000e+00, 0.0000e+00, ..., 1.2536e-06, + 2.1793e-06, 3.7253e-09], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 6.7987e-08, + 2.1327e-07, 2.0489e-08], + [ 1.8626e-08, 0.0000e+00, 0.0000e+00, ..., 1.4435e-07, + -4.9360e-07, -2.3283e-08], + ..., + [ 1.6764e-08, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 3.7160e-07, 1.6764e-08], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 5.0291e-08, + -4.1761e-06, 2.7940e-09], + [ 1.0058e-07, 0.0000e+00, 0.0000e+00, ..., 9.2201e-08, + 3.5837e-06, 9.2201e-08]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0039, -0.0168, -0.0011, 0.0308, -0.0048, 0.0276, 0.0057, 0.0181, + -0.0022, -0.0091], device='cuda:0'), grad: tensor([ 6.4336e-06, 5.1502e-07, -4.9919e-07, 1.2684e-04, 1.9744e-06, + -1.2982e-04, -1.1072e-05, 1.6494e-06, -8.7172e-06, 1.2584e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 214.64, cls_loss 0.0026 cls_loss_mapping 0.0051 cls_loss_causal 0.4920 re_mapping 0.0058 re_causal 0.0167 /// teacc 98.89 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0440, -0.0049, -0.0829, ..., 0.0242, -0.0635, -0.0098], + [ 0.0459, -0.0597, -0.0189, ..., -0.0112, -0.0533, -0.2406], + [-0.0056, 0.0334, -0.0059, ..., -0.0945, 0.1089, -0.0399], + ..., + [ 0.0076, -0.0344, -0.0130, ..., -0.0954, -0.1616, 0.0274], + [-0.0215, -0.0603, -0.0181, ..., -0.0847, 0.0506, -0.1508], + [-0.1616, -0.0116, -0.0108, ..., -0.0898, -0.1068, -0.0828]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., -5.5581e-05, + -3.2127e-05, 1.5832e-08], + [-7.4506e-09, 2.7940e-09, 0.0000e+00, ..., 7.5437e-08, + 1.1735e-07, 1.7881e-07], + [ 1.8626e-09, -4.3027e-07, 0.0000e+00, ..., 2.9895e-07, + -1.3057e-06, 8.1025e-08], + ..., + [ 9.3132e-10, 3.4738e-07, 0.0000e+00, ..., 5.8673e-08, + 1.2834e-06, 3.1106e-07], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 1.1763e-06, + 6.7521e-07, 3.6322e-08], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 6.6496e-07, + 3.9395e-07, 9.7509e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0032, -0.0177, -0.0005, 0.0310, -0.0007, 0.0280, 0.0059, 0.0180, + -0.0025, -0.0129], device='cuda:0'), grad: tensor([-9.9480e-05, 3.9302e-06, -4.1053e-06, -4.0233e-06, 1.7196e-05, + 5.1081e-05, 9.0897e-05, -1.5432e-06, 8.5980e-06, -6.2704e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 214.61, cls_loss 0.0026 cls_loss_mapping 0.0047 cls_loss_causal 0.5372 re_mapping 0.0057 re_causal 0.0180 /// teacc 99.05 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0437, -0.0049, -0.0829, ..., 0.0243, -0.0640, -0.0099], + [ 0.0461, -0.0597, -0.0187, ..., -0.0114, -0.0537, -0.2407], + [-0.0057, 0.0335, -0.0058, ..., -0.0946, 0.1093, -0.0400], + ..., + [ 0.0075, -0.0344, -0.0131, ..., -0.0963, -0.1617, 0.0274], + [-0.0215, -0.0603, -0.0180, ..., -0.0852, 0.0506, -0.1509], + [-0.1625, -0.0116, -0.0108, ..., -0.0900, -0.1073, -0.0822]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.7695e-08, + 1.3439e-06, 1.8626e-09], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.3132e-07, 2.3283e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -8.1025e-08, 6.5193e-09], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.6042e-07, 1.6764e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.5635e-08, + -5.2154e-08, 1.2107e-08], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 9.3691e-07, 4.1910e-08]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0030, -0.0180, -0.0012, 0.0313, -0.0022, 0.0281, 0.0052, 0.0184, + -0.0027, -0.0108], device='cuda:0'), grad: tensor([ 8.7470e-06, 6.2305e-07, 9.3691e-07, 1.7628e-05, 9.6187e-06, + -2.4647e-05, 2.5742e-06, 4.5687e-05, 2.6412e-06, -6.3717e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 214.15, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5475 re_mapping 0.0052 re_causal 0.0172 /// teacc 98.93 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0460, -0.0050, -0.0829, ..., 0.0245, -0.0646, -0.0104], + [ 0.0457, -0.0602, -0.0188, ..., -0.0119, -0.0539, -0.2412], + [-0.0057, 0.0334, -0.0062, ..., -0.0950, 0.1095, -0.0401], + ..., + [ 0.0109, -0.0338, -0.0130, ..., -0.0964, -0.1618, 0.0278], + [-0.0216, -0.0605, -0.0180, ..., -0.0856, 0.0506, -0.1510], + [-0.1660, -0.0117, -0.0109, ..., -0.0902, -0.1080, -0.0825]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.0757e-07, + 6.0536e-08, 1.8626e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.0245e-08, + 2.3842e-07, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + -5.5544e-06, -2.9802e-08], + ..., + [ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., 3.7253e-09, + 5.1148e-06, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8673e-08, + 1.5274e-07, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.2154e-08, + 1.1548e-07, -4.8429e-08]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0029, -0.0183, -0.0013, 0.0315, -0.0023, 0.0281, 0.0051, 0.0195, + -0.0028, -0.0115], device='cuda:0'), grad: tensor([-1.4501e-06, 1.9148e-06, -2.6733e-05, 1.2862e-06, 1.6280e-06, + -1.7677e-06, 4.5355e-07, 2.5243e-05, 7.6182e-07, -1.3085e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 214.34, cls_loss 0.0027 cls_loss_mapping 0.0037 cls_loss_causal 0.5293 re_mapping 0.0054 re_causal 0.0158 /// teacc 99.00 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0486, -0.0050, -0.0830, ..., 0.0247, -0.0663, -0.0106], + [ 0.0449, -0.0602, -0.0188, ..., -0.0124, -0.0543, -0.2421], + [-0.0043, 0.0335, -0.0071, ..., -0.0957, 0.1100, -0.0390], + ..., + [ 0.0129, -0.0337, -0.0133, ..., -0.0970, -0.1622, 0.0291], + [-0.0214, -0.0608, -0.0181, ..., -0.0862, 0.0511, -0.1512], + [-0.1689, -0.0120, -0.0109, ..., -0.0904, -0.1089, -0.0830]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.7229e-07, + 1.7323e-07, 3.3528e-08], + [ 1.5832e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.4529e-07, 3.5297e-07], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 1.2107e-07, + 5.2154e-08, 4.3772e-08], + ..., + [ 9.4995e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.2573e-07, 4.7125e-07], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -1.5367e-07, + -5.2247e-07, 1.0245e-07], + [ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 6.3237e-07, -9.5740e-07]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0025, -0.0181, -0.0008, 0.0310, -0.0026, 0.0283, 0.0055, 0.0195, + -0.0025, -0.0117], device='cuda:0'), grad: tensor([ 2.6021e-06, 1.1884e-05, 2.2594e-06, -5.2974e-06, 3.9116e-06, + -5.0776e-06, 6.8434e-06, -9.4026e-06, 1.7807e-06, -9.5814e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 214.21, cls_loss 0.0019 cls_loss_mapping 0.0026 cls_loss_causal 0.5177 re_mapping 0.0056 re_causal 0.0179 /// teacc 99.03 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0495, -0.0052, -0.0832, ..., 0.0246, -0.0670, -0.0113], + [ 0.0448, -0.0612, -0.0187, ..., -0.0127, -0.0545, -0.2427], + [-0.0044, 0.0336, -0.0062, ..., -0.0959, 0.1103, -0.0389], + ..., + [ 0.0129, -0.0332, -0.0136, ..., -0.0972, -0.1623, 0.0290], + [-0.0214, -0.0612, -0.0182, ..., -0.0866, 0.0511, -0.1512], + [-0.1695, -0.0113, -0.0109, ..., -0.0902, -0.1093, -0.0832]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0431e-07, 1.6484e-07, + 3.8184e-08], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.4064e-08, 1.3225e-07, + 1.1791e-06], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9802e-08, 1.4901e-08, + 8.4750e-08], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 1.2107e-08, + 3.8370e-07], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7975e-07, 2.3376e-07, + 8.2888e-08], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-08, 3.5390e-08, + 3.5577e-07]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0017, -0.0189, -0.0008, 0.0308, -0.0027, 0.0289, 0.0054, 0.0203, + -0.0030, -0.0117], device='cuda:0'), grad: tensor([ 4.4703e-07, 5.6364e-06, 5.6252e-07, 1.8068e-07, -1.1139e-05, + 1.8105e-06, -1.9968e-06, 1.2172e-06, 9.3132e-07, 2.3469e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 214.37, cls_loss 0.0021 cls_loss_mapping 0.0037 cls_loss_causal 0.5079 re_mapping 0.0054 re_causal 0.0165 /// teacc 98.91 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0495, -0.0052, -0.0833, ..., 0.0245, -0.0678, -0.0115], + [ 0.0449, -0.0614, -0.0187, ..., -0.0131, -0.0547, -0.2432], + [-0.0044, 0.0336, -0.0062, ..., -0.0961, 0.1129, -0.0391], + ..., + [ 0.0129, -0.0327, -0.0139, ..., -0.0980, -0.1650, 0.0289], + [-0.0214, -0.0613, -0.0182, ..., -0.0864, 0.0513, -0.1513], + [-0.1697, -0.0114, -0.0110, ..., -0.0904, -0.1097, -0.0837]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.9092e-07, + 6.3330e-08, 1.8626e-09], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 3.2596e-08, 2.2352e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + -1.2480e-07, 3.7253e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 6.5193e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6764e-08, + 2.5425e-07, 1.1176e-08], + [ 2.3283e-08, 0.0000e+00, 0.0000e+00, ..., 5.6811e-08, + -1.0459e-06, 1.3039e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0012, -0.0187, 0.0013, 0.0310, -0.0027, 0.0289, 0.0050, 0.0184, + -0.0028, -0.0116], device='cuda:0'), grad: tensor([ 1.4622e-06, -1.3523e-06, 1.0496e-06, -1.7695e-08, 5.2378e-06, + 1.7166e-05, 7.3388e-07, 1.2172e-06, 8.9854e-06, -3.4511e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 214.02, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.5163 re_mapping 0.0055 re_causal 0.0167 /// teacc 99.01 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0496, -0.0053, -0.0834, ..., 0.0246, -0.0681, -0.0117], + [ 0.0447, -0.0615, -0.0187, ..., -0.0136, -0.0552, -0.2466], + [-0.0044, 0.0345, -0.0060, ..., -0.0960, 0.1133, -0.0393], + ..., + [ 0.0129, -0.0333, -0.0136, ..., -0.0985, -0.1650, 0.0302], + [-0.0215, -0.0616, -0.0182, ..., -0.0865, 0.0515, -0.1516], + [-0.1709, -0.0115, -0.0110, ..., -0.0905, -0.1120, -0.0838]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 8.3223e-06, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4214e-08, + 2.1979e-07, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-08, + 3.1199e-07, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 9.0338e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.9325e-07, + -1.6004e-05, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3842e-07, + 5.3830e-06, 3.3528e-08]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0011, -0.0189, 0.0014, 0.0311, -0.0025, 0.0294, 0.0049, 0.0192, + -0.0024, -0.0132], device='cuda:0'), grad: tensor([ 2.6837e-05, -8.0243e-06, 1.9297e-06, 4.0047e-07, -6.4634e-06, + 2.9281e-06, 4.2766e-06, 2.0992e-06, -4.3780e-05, 1.9833e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 214.11, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.5222 re_mapping 0.0053 re_causal 0.0163 /// teacc 98.98 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0496, -0.0055, -0.0834, ..., 0.0247, -0.0688, -0.0114], + [ 0.0445, -0.0615, -0.0187, ..., -0.0148, -0.0558, -0.2468], + [-0.0043, 0.0347, -0.0061, ..., -0.0963, 0.1134, -0.0394], + ..., + [ 0.0129, -0.0335, -0.0136, ..., -0.0989, -0.1651, 0.0302], + [-0.0215, -0.0618, -0.0182, ..., -0.0869, 0.0519, -0.1516], + [-0.1716, -0.0116, -0.0110, ..., -0.0909, -0.1125, -0.0843]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 2.0489e-07, 1.8626e-09], + [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 2.9802e-07, 2.0489e-08], + [ 4.6566e-09, -9.3132e-10, 0.0000e+00, ..., 1.6764e-08, + 3.6322e-08, 4.6566e-09], + ..., + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.2536e-06, 8.3819e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 5.3085e-08, + 2.8592e-07, 1.8626e-09], + [ 6.4261e-08, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 2.0210e-07, 7.8231e-08]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0010, -0.0191, 0.0015, 0.0311, -0.0024, 0.0293, 0.0054, 0.0193, + -0.0023, -0.0134], device='cuda:0'), grad: tensor([ 2.7493e-06, 7.4022e-06, 1.1414e-05, -4.6670e-05, -7.5245e-04, + 1.9833e-05, 2.2352e-06, 9.1314e-05, 1.1474e-05, 6.5184e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 214.73, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.4966 re_mapping 0.0055 re_causal 0.0162 /// teacc 98.92 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0497, -0.0055, -0.0839, ..., 0.0249, -0.0696, -0.0114], + [ 0.0445, -0.0616, -0.0188, ..., -0.0153, -0.0563, -0.2469], + [-0.0050, 0.0347, -0.0061, ..., -0.0963, 0.1139, -0.0393], + ..., + [ 0.0137, -0.0335, -0.0147, ..., -0.0989, -0.1653, 0.0301], + [-0.0216, -0.0618, -0.0184, ..., -0.0877, 0.0523, -0.1518], + [-0.1722, -0.0116, -0.0111, ..., -0.0912, -0.1133, -0.0844]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-08, + 1.5181e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1420e-08, + 2.1420e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, + 6.1467e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 8.3819e-09, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.9162e-08, + 1.4249e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.1025e-08, + -1.5646e-07, 1.3970e-08]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0005, -0.0192, 0.0017, 0.0312, -0.0023, 0.0288, 0.0060, 0.0194, + -0.0022, -0.0137], device='cuda:0'), grad: tensor([ 5.6066e-06, -4.8995e-05, 1.2726e-05, 4.7609e-06, 3.8815e-04, + 1.6674e-05, 1.9982e-05, 1.0386e-05, 7.7665e-05, -4.8661e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 214.21, cls_loss 0.0017 cls_loss_mapping 0.0040 cls_loss_causal 0.5260 re_mapping 0.0054 re_causal 0.0172 /// teacc 98.93 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0498, -0.0055, -0.0839, ..., 0.0249, -0.0702, -0.0104], + [ 0.0445, -0.0616, -0.0189, ..., -0.0154, -0.0555, -0.2470], + [-0.0050, 0.0349, -0.0062, ..., -0.0964, 0.1137, -0.0393], + ..., + [ 0.0137, -0.0336, -0.0147, ..., -0.0992, -0.1653, 0.0300], + [-0.0216, -0.0619, -0.0182, ..., -0.0878, 0.0526, -0.1519], + [-0.1723, -0.0116, -0.0111, ..., -0.0917, -0.1147, -0.0849]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.1025e-08, 0.0000e+00, ..., -2.4736e-06, + -1.1614e-06, 0.0000e+00], + [ 0.0000e+00, 1.1502e-07, 0.0000e+00, ..., 6.1467e-08, + 3.0920e-07, 4.6566e-10], + [ 0.0000e+00, -5.6997e-06, 0.0000e+00, ..., 3.6601e-07, + -8.7023e-06, 4.6566e-10], + ..., + [ 0.0000e+00, 2.6263e-06, 0.0000e+00, ..., 1.1176e-07, + 4.0904e-06, 9.3132e-10], + [ 0.0000e+00, 1.0850e-07, 0.0000e+00, ..., 1.5907e-06, + 5.7295e-06, 4.6566e-10], + [ 0.0000e+00, 1.5367e-08, 0.0000e+00, ..., 1.6019e-07, + 3.3621e-07, 4.6566e-10]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0004, -0.0184, 0.0012, 0.0321, -0.0021, 0.0289, 0.0059, 0.0193, + -0.0023, -0.0140], device='cuda:0'), grad: tensor([-7.1377e-06, 7.4459e-07, -2.8968e-05, 1.8790e-05, 7.7719e-07, + -1.3612e-05, 1.5516e-06, 1.3568e-05, 1.3806e-05, 5.0757e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 214.68, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5428 re_mapping 0.0052 re_causal 0.0168 /// teacc 98.99 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0499, -0.0055, -0.0840, ..., 0.0245, -0.0711, -0.0101], + [ 0.0434, -0.0616, -0.0189, ..., -0.0157, -0.0558, -0.2471], + [-0.0056, 0.0352, -0.0063, ..., -0.0972, 0.1137, -0.0394], + ..., + [ 0.0141, -0.0338, -0.0148, ..., -0.0997, -0.1653, 0.0300], + [-0.0204, -0.0620, -0.0182, ..., -0.0899, 0.0522, -0.1520], + [-0.1727, -0.0116, -0.0111, ..., -0.0917, -0.1152, -0.0848]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.3446e-07, + 5.1688e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + 3.1991e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.2399e-08, + 5.1223e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 1.2852e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + -1.5264e-06, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2154e-07, + 1.4296e-07, 8.8476e-09]], device='cuda:0') +Epoch 185, bias, value: tensor([ 5.3367e-05, -1.9743e-02, 9.6193e-04, 3.1588e-02, -2.1789e-03, + 2.9116e-02, 7.1058e-03, 2.0214e-02, -2.5643e-03, -1.3988e-02], + device='cuda:0'), grad: tensor([-3.1153e-07, 3.6247e-06, 5.5647e-07, 8.0839e-06, 3.6135e-07, + 2.7157e-06, 1.0002e-06, 1.5553e-06, -1.9670e-05, 2.0768e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 214.42, cls_loss 0.0017 cls_loss_mapping 0.0038 cls_loss_causal 0.5424 re_mapping 0.0055 re_causal 0.0167 /// teacc 98.97 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0499, -0.0055, -0.0840, ..., 0.0250, -0.0714, -0.0102], + [ 0.0432, -0.0616, -0.0189, ..., -0.0156, -0.0559, -0.2473], + [-0.0051, 0.0353, -0.0063, ..., -0.0973, 0.1139, -0.0389], + ..., + [ 0.0140, -0.0339, -0.0149, ..., -0.1002, -0.1655, 0.0294], + [-0.0206, -0.0621, -0.0183, ..., -0.0900, 0.0519, -0.1527], + [-0.1730, -0.0116, -0.0111, ..., -0.0919, -0.1147, -0.0850]], + device='cuda:0'), grad: tensor([[ 1.3085e-07, 0.0000e+00, 0.0000e+00, ..., -3.9954e-07, + 9.3132e-08, 4.5262e-07], + [ 1.2293e-07, 0.0000e+00, 0.0000e+00, ..., 3.5390e-08, + 5.3830e-07, 4.9360e-07], + [ 2.9337e-07, 0.0000e+00, 0.0000e+00, ..., 2.4680e-08, + -6.1877e-06, 4.6939e-07], + ..., + [ 1.2107e-07, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + 4.5970e-06, 8.4145e-07], + [ 4.3306e-08, 0.0000e+00, 0.0000e+00, ..., 1.0477e-07, + 2.4401e-07, 2.2771e-07], + [ 3.2131e-08, 0.0000e+00, 0.0000e+00, ..., 1.0105e-07, + 7.6834e-08, 4.8289e-07]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0002, -0.0193, 0.0010, 0.0314, -0.0020, 0.0294, 0.0071, 0.0200, + -0.0033, -0.0138], device='cuda:0'), grad: tensor([ 5.8394e-07, -2.8964e-06, -2.4050e-05, -1.0297e-05, -1.6555e-05, + 4.3809e-06, 1.5683e-06, 2.7940e-05, 6.9290e-06, 1.2331e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 214.33, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.5018 re_mapping 0.0056 re_causal 0.0164 /// teacc 98.95 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0502, -0.0055, -0.0840, ..., 0.0254, -0.0717, -0.0073], + [ 0.0439, -0.0617, -0.0189, ..., -0.0161, -0.0563, -0.2475], + [-0.0052, 0.0353, -0.0063, ..., -0.0972, 0.1141, -0.0395], + ..., + [ 0.0138, -0.0339, -0.0149, ..., -0.1002, -0.1656, 0.0298], + [-0.0206, -0.0622, -0.0183, ..., -0.0918, 0.0512, -0.1525], + [-0.1735, -0.0116, -0.0111, ..., -0.0919, -0.1152, -0.0854]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.8394e-07, + -4.2841e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 6.6124e-08, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + -5.8534e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 7.4971e-08, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0268e-08, + -8.9360e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5553e-07, + 1.1874e-06, 1.0757e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0005, -0.0194, 0.0011, 0.0314, -0.0019, 0.0291, 0.0075, 0.0196, + -0.0032, -0.0136], device='cuda:0'), grad: tensor([-1.1763e-06, 3.2503e-07, -8.9826e-07, 4.8289e-07, 8.6008e-07, + 4.0559e-07, 4.9500e-07, -3.3304e-06, -1.2498e-06, 4.0792e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 214.29, cls_loss 0.0016 cls_loss_mapping 0.0031 cls_loss_causal 0.5159 re_mapping 0.0055 re_causal 0.0167 /// teacc 98.90 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0502, -0.0055, -0.0840, ..., 0.0255, -0.0725, -0.0074], + [ 0.0442, -0.0617, -0.0189, ..., -0.0165, -0.0573, -0.2477], + [-0.0052, 0.0354, -0.0063, ..., -0.0977, 0.1148, -0.0394], + ..., + [ 0.0137, -0.0339, -0.0149, ..., -0.1013, -0.1657, 0.0299], + [-0.0206, -0.0622, -0.0183, ..., -0.0914, 0.0519, -0.1525], + [-0.1738, -0.0116, -0.0111, ..., -0.0931, -0.1163, -0.0855]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.3160e-06, + -3.6061e-06, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3877e-07, + 2.3982e-07, 4.9826e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.1572e-07, + 1.1353e-06, 2.4214e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.8429e-08, + 1.0710e-07, 5.4017e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.1609e-06, + 3.1352e-05, 9.6392e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9281e-06, + 2.8200e-06, -7.4506e-09]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0003, -0.0198, 0.0016, 0.0313, -0.0018, 0.0292, 0.0067, 0.0195, + -0.0026, -0.0138], device='cuda:0'), grad: tensor([-1.1230e-04, 2.6762e-05, 5.9828e-06, 2.2560e-05, -6.9916e-05, + -4.4316e-05, -1.9357e-05, 8.3596e-06, 9.6500e-05, 8.5652e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 214.09, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5275 re_mapping 0.0057 re_causal 0.0169 /// teacc 98.85 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0503, -0.0056, -0.0840, ..., 0.0242, -0.0751, -0.0074], + [ 0.0442, -0.0617, -0.0189, ..., -0.0172, -0.0577, -0.2478], + [-0.0050, 0.0356, -0.0063, ..., -0.0982, 0.1149, -0.0389], + ..., + [ 0.0137, -0.0340, -0.0149, ..., -0.1017, -0.1658, 0.0306], + [-0.0209, -0.0625, -0.0183, ..., -0.0919, 0.0515, -0.1532], + [-0.1739, -0.0120, -0.0111, ..., -0.0930, -0.1169, -0.0862]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.8452e-08, + 1.8487e-07, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.2919e-05, 6.3796e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -5.1945e-05, 8.0559e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 3.3993e-07, 2.3749e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0536e-09, + 3.1805e-07, 1.2573e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.8476e-09, + 6.7055e-08, 4.0978e-08]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0016, -0.0199, 0.0016, 0.0322, -0.0012, 0.0294, 0.0069, 0.0197, + -0.0031, -0.0144], device='cuda:0'), grad: tensor([ 2.5565e-07, 2.0206e-05, -7.1406e-05, -1.6332e-05, 4.3176e-06, + 8.3372e-06, 5.3525e-05, 1.9446e-06, 9.3179e-07, -1.6708e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 214.19, cls_loss 0.0017 cls_loss_mapping 0.0028 cls_loss_causal 0.5340 re_mapping 0.0056 re_causal 0.0171 /// teacc 99.07 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0503, -0.0057, -0.0844, ..., 0.0250, -0.0749, -0.0080], + [ 0.0443, -0.0617, -0.0190, ..., -0.0176, -0.0584, -0.2480], + [-0.0050, 0.0356, -0.0070, ..., -0.0983, 0.1155, -0.0390], + ..., + [ 0.0136, -0.0339, -0.0161, ..., -0.1023, -0.1662, 0.0283], + [-0.0209, -0.0626, -0.0193, ..., -0.0920, 0.0514, -0.1533], + [-0.1740, -0.0120, -0.0113, ..., -0.0932, -0.1173, -0.0864]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.7807e-06, + -4.8429e-08, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.5996e-07, + -2.3376e-07, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1548e-07, + 6.0769e-07, 3.3993e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + -3.9814e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1781e-07, + -5.5227e-07, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0559e-07, + 3.1153e-07, 3.2596e-09]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0015, -0.0197, 0.0019, 0.0326, -0.0010, 0.0298, 0.0065, 0.0189, + -0.0034, -0.0142], device='cuda:0'), grad: tensor([-4.8205e-06, 1.7229e-07, 7.1637e-06, -2.2221e-06, 6.7288e-07, + 1.6540e-06, 1.1064e-06, -6.5565e-06, 5.8999e-07, 2.2240e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 214.16, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5309 re_mapping 0.0055 re_causal 0.0172 /// teacc 98.94 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0505, -0.0057, -0.0846, ..., 0.0243, -0.0760, -0.0084], + [ 0.0445, -0.0618, -0.0190, ..., -0.0177, -0.0585, -0.2487], + [-0.0050, 0.0355, -0.0073, ..., -0.0986, 0.1156, -0.0392], + ..., + [ 0.0136, -0.0335, -0.0165, ..., -0.1025, -0.1662, 0.0283], + [-0.0210, -0.0629, -0.0196, ..., -0.0920, 0.0516, -0.1534], + [-0.1741, -0.0124, -0.0114, ..., -0.0934, -0.1181, -0.0863]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.8697e-08, + 5.6811e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 4.1444e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + -1.4603e-06, 4.1910e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 5.7928e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9337e-08, + 7.0548e-07, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4435e-08, + 3.0873e-07, 2.4680e-08]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0021, -0.0196, 0.0019, 0.0323, -0.0011, 0.0299, 0.0068, 0.0199, + -0.0033, -0.0151], device='cuda:0'), grad: tensor([-5.2620e-08, -8.6380e-07, -4.6119e-06, 1.3597e-06, 3.4971e-07, + -1.6969e-06, 6.3144e-07, 2.0564e-06, 1.4892e-06, 1.3299e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 214.21, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5182 re_mapping 0.0051 re_causal 0.0162 /// teacc 99.08 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0509, -0.0056, -0.0846, ..., 0.0255, -0.0758, -0.0084], + [ 0.0445, -0.0618, -0.0191, ..., -0.0176, -0.0587, -0.2489], + [-0.0050, 0.0355, -0.0068, ..., -0.0988, 0.1158, -0.0398], + ..., + [ 0.0137, -0.0336, -0.0140, ..., -0.1032, -0.1662, 0.0286], + [-0.0210, -0.0630, -0.0195, ..., -0.0930, 0.0509, -0.1536], + [-0.1743, -0.0126, -0.0114, ..., -0.0939, -0.1194, -0.0865]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 3.4925e-08, + -9.4855e-07, 6.3051e-07], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 2.0722e-07, + 1.8161e-07, 2.1048e-07], + [ 0.0000e+00, 0.0000e+00, -4.6100e-08, ..., 8.8476e-08, + -5.4250e-07, 4.8382e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 4.0513e-08, + 2.2259e-07, 2.0210e-07], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 6.4541e-07, + 5.8766e-07, 4.1910e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.8079e-07, + 1.6158e-07, 4.4098e-07]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0012, -0.0187, 0.0018, 0.0316, -0.0008, 0.0299, 0.0074, 0.0198, + -0.0039, -0.0157], device='cuda:0'), grad: tensor([ 1.1697e-06, 1.1493e-06, 3.9376e-06, 1.0170e-06, -1.5609e-06, + 8.0094e-08, -5.7817e-06, -4.2915e-06, 2.7604e-06, 1.4752e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 214.45, cls_loss 0.0014 cls_loss_mapping 0.0026 cls_loss_causal 0.4971 re_mapping 0.0050 re_causal 0.0157 /// teacc 98.92 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0510, -0.0056, -0.0846, ..., 0.0255, -0.0758, -0.0086], + [ 0.0446, -0.0618, -0.0191, ..., -0.0175, -0.0586, -0.2491], + [-0.0050, 0.0355, -0.0068, ..., -0.0990, 0.1159, -0.0412], + ..., + [ 0.0137, -0.0336, -0.0139, ..., -0.1036, -0.1663, 0.0285], + [-0.0210, -0.0630, -0.0195, ..., -0.0933, 0.0509, -0.1537], + [-0.1745, -0.0126, -0.0114, ..., -0.0929, -0.1199, -0.0868]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3271e-07, + -1.9558e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -7.1153e-07, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 2.6915e-07, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-08, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + 1.6950e-07, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 9.3598e-08, -4.5169e-08]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0019, -0.0180, 0.0017, 0.0315, -0.0007, 0.0301, 0.0072, 0.0197, + -0.0041, -0.0157], device='cuda:0'), grad: tensor([-9.4529e-08, -5.2676e-06, 4.8354e-06, 1.8459e-06, 6.4820e-06, + -6.5193e-07, 1.7388e-06, 1.2569e-05, 1.1839e-05, -3.3289e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 214.55, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5178 re_mapping 0.0051 re_causal 0.0162 /// teacc 99.12 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0510, -0.0057, -0.0846, ..., 0.0253, -0.0764, -0.0088], + [ 0.0447, -0.0618, -0.0189, ..., -0.0176, -0.0587, -0.2492], + [-0.0051, 0.0355, -0.0068, ..., -0.0992, 0.1161, -0.0414], + ..., + [ 0.0137, -0.0336, -0.0139, ..., -0.1039, -0.1664, 0.0289], + [-0.0210, -0.0631, -0.0194, ..., -0.0936, 0.0509, -0.1536], + [-0.1745, -0.0126, -0.0114, ..., -0.0929, -0.1209, -0.0865]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6699e-06, + 5.5879e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-07, + 1.1036e-07, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1036e-07, + 1.5832e-08, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9092e-08, + 7.1246e-08, -1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2340e-07, + -2.6543e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8720e-07, + 8.0327e-07, 5.6345e-08]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0022, -0.0177, 0.0015, 0.0334, -0.0015, 0.0301, 0.0074, 0.0190, + -0.0044, -0.0150], device='cuda:0'), grad: tensor([-5.7369e-06, 2.3078e-06, 4.2357e-06, 1.6138e-05, 8.4285e-07, + -3.8743e-06, 3.4887e-06, -2.2978e-05, 5.1223e-07, 5.0366e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.23, cls_loss 0.0024 cls_loss_mapping 0.0040 cls_loss_causal 0.5253 re_mapping 0.0059 re_causal 0.0162 /// teacc 99.01 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0512, -0.0057, -0.0847, ..., 0.0255, -0.0770, -0.0094], + [ 0.0446, -0.0619, -0.0189, ..., -0.0179, -0.0588, -0.2494], + [-0.0052, 0.0355, -0.0069, ..., -0.1008, 0.1168, -0.0408], + ..., + [ 0.0137, -0.0336, -0.0139, ..., -0.1044, -0.1667, 0.0304], + [-0.0210, -0.0631, -0.0195, ..., -0.0941, 0.0507, -0.1537], + [-0.1750, -0.0126, -0.0115, ..., -0.0932, -0.1216, -0.0874]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.4680e-08, + 3.6322e-08, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + 9.0338e-08, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-08, + -1.1791e-06, 1.2573e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.2117e-06, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.0781e-08, + 2.6263e-07, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + 1.1176e-07, 2.7940e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0024, -0.0158, 0.0038, 0.0327, -0.0017, 0.0297, 0.0084, 0.0166, + -0.0030, -0.0162], device='cuda:0'), grad: tensor([ 1.0151e-07, 7.6322e-07, -2.5295e-06, -1.2696e-05, 8.5756e-06, + -4.2096e-06, -5.3411e-07, 1.0230e-05, 1.0163e-05, -9.8422e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 214.26, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4979 re_mapping 0.0056 re_causal 0.0163 /// teacc 98.98 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0512, -0.0057, -0.0847, ..., 0.0254, -0.0776, -0.0095], + [ 0.0447, -0.0619, -0.0189, ..., -0.0180, -0.0597, -0.2498], + [-0.0053, 0.0355, -0.0070, ..., -0.1013, 0.1175, -0.0414], + ..., + [ 0.0137, -0.0336, -0.0140, ..., -0.1065, -0.1669, 0.0304], + [-0.0211, -0.0631, -0.0195, ..., -0.0943, 0.0505, -0.1538], + [-0.1751, -0.0126, -0.0115, ..., -0.0931, -0.1223, -0.0873]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -5.6438e-06, + -1.8859e-07, 3.4459e-08], + [ 2.8405e-08, 0.0000e+00, 0.0000e+00, ..., 1.9139e-07, + 4.2329e-07, 2.0023e-08], + [ 6.2399e-08, 0.0000e+00, 0.0000e+00, ..., 3.3202e-07, + -6.2631e-07, 7.8231e-08], + ..., + [-1.7136e-07, 0.0000e+00, 0.0000e+00, ..., 2.2817e-08, + 2.2817e-07, 9.7789e-09], + [ 2.5611e-08, 0.0000e+00, 0.0000e+00, ..., 1.5311e-06, + 2.8610e-05, 9.3132e-09], + [ 1.3504e-08, 0.0000e+00, 0.0000e+00, ..., 1.7229e-07, + 1.5264e-06, 2.7241e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0028, -0.0162, 0.0039, 0.0327, -0.0020, 0.0299, 0.0088, 0.0167, + -0.0028, -0.0161], device='cuda:0'), grad: tensor([-2.2531e-05, 8.2562e-07, 5.9456e-06, -1.3039e-07, 2.7940e-09, + -1.3387e-04, 9.4295e-05, -7.4226e-07, 5.1260e-05, 4.7088e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 214.49, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5079 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.02 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0515, -0.0057, -0.0847, ..., 0.0260, -0.0773, -0.0097], + [ 0.0446, -0.0619, -0.0189, ..., -0.0183, -0.0566, -0.2505], + [-0.0054, 0.0355, -0.0070, ..., -0.1016, 0.1172, -0.0420], + ..., + [ 0.0139, -0.0336, -0.0140, ..., -0.1069, -0.1673, 0.0303], + [-0.0212, -0.0631, -0.0195, ..., -0.0955, 0.0472, -0.1552], + [-0.1757, -0.0126, -0.0115, ..., -0.0935, -0.1230, -0.0870]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8161e-08, + 4.1910e-09, 6.5193e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.0757e-07, 1.9558e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.5274e-07, 1.9558e-08], + ..., + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-08, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -2.1309e-06, 1.8626e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 1.3057e-06, 4.2608e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0022, -0.0134, 0.0031, 0.0321, -0.0022, 0.0301, 0.0093, 0.0165, + -0.0053, -0.0158], device='cuda:0'), grad: tensor([ 2.8405e-08, 6.5565e-07, 7.3295e-07, 1.5898e-06, -1.8580e-06, + 2.0256e-07, 4.2329e-07, -1.6652e-06, -6.2101e-06, 6.1020e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 214.14, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.5284 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.00 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0518, -0.0057, -0.0848, ..., 0.0269, -0.0764, -0.0097], + [ 0.0446, -0.0619, -0.0189, ..., -0.0184, -0.0570, -0.2511], + [-0.0069, 0.0355, -0.0079, ..., -0.1020, 0.1178, -0.0423], + ..., + [ 0.0154, -0.0336, -0.0141, ..., -0.1076, -0.1674, 0.0298], + [-0.0215, -0.0632, -0.0200, ..., -0.0969, 0.0461, -0.1554], + [-0.1764, -0.0126, -0.0116, ..., -0.0935, -0.1237, -0.0871]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.4296e-07, + 9.5461e-07, 0.0000e+00], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + 1.9521e-06, 5.1223e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 8.8476e-09, + 6.8694e-06, 1.3970e-09], + ..., + [-1.7695e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.7311e-06, 9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 4.7963e-08, + 2.7999e-05, 2.7940e-09], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 2.5146e-08, + 1.6475e-06, 1.4855e-07]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0016, -0.0136, 0.0033, 0.0319, -0.0020, 0.0307, 0.0094, 0.0164, + -0.0063, -0.0157], device='cuda:0'), grad: tensor([ 7.3351e-06, 1.5587e-05, 6.6102e-05, -4.1747e-04, 1.6298e-06, + 4.2975e-05, -7.4506e-08, 3.6418e-05, 2.3377e-04, 1.3731e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 214.23, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5238 re_mapping 0.0050 re_causal 0.0161 /// teacc 99.06 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0521, -0.0057, -0.0851, ..., 0.0275, -0.0762, -0.0097], + [ 0.0437, -0.0619, -0.0190, ..., -0.0187, -0.0573, -0.2513], + [-0.0073, 0.0355, -0.0090, ..., -0.1024, 0.1182, -0.0425], + ..., + [ 0.0158, -0.0336, -0.0139, ..., -0.1084, -0.1676, 0.0298], + [-0.0210, -0.0632, -0.0205, ..., -0.0971, 0.0459, -0.1555], + [-0.1786, -0.0126, -0.0117, ..., -0.0927, -0.1241, -0.0875]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -3.5856e-08, + 5.2713e-07, 1.8626e-09], + [ 3.3993e-08, 0.0000e+00, 0.0000e+00, ..., 8.3353e-08, + -5.1968e-06, 6.5193e-09], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 3.8184e-08, + 6.5602e-06, 2.1420e-08], + ..., + [-5.4948e-08, 0.0000e+00, 0.0000e+00, ..., 1.5832e-08, + 1.3765e-06, 6.0536e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -2.3171e-06, + -1.8179e-05, 2.3283e-09], + [ 1.8626e-08, 0.0000e+00, 0.0000e+00, ..., 3.4412e-07, + 9.1344e-06, 1.7229e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0012, -0.0135, 0.0036, 0.0321, -0.0017, 0.0308, 0.0089, 0.0161, + -0.0067, -0.0154], device='cuda:0'), grad: tensor([ 1.5311e-06, -5.4538e-05, 4.8608e-05, 3.6787e-06, 1.4231e-06, + 2.5958e-05, -9.2015e-06, -9.1910e-05, -4.8369e-05, 1.2290e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 214.39, cls_loss 0.0017 cls_loss_mapping 0.0024 cls_loss_causal 0.5140 re_mapping 0.0051 re_causal 0.0154 /// teacc 99.02 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0521, -0.0057, -0.0851, ..., 0.0261, -0.0767, -0.0097], + [ 0.0437, -0.0619, -0.0190, ..., -0.0191, -0.0574, -0.2514], + [-0.0072, 0.0356, -0.0090, ..., -0.1023, 0.1185, -0.0426], + ..., + [ 0.0158, -0.0336, -0.0139, ..., -0.1093, -0.1681, 0.0299], + [-0.0210, -0.0632, -0.0206, ..., -0.0972, 0.0463, -0.1555], + [-0.1791, -0.0127, -0.0117, ..., -0.0904, -0.1253, -0.0877]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.5837e-06, + -2.1309e-06, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 4.1444e-08, 1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.8476e-09, + -5.6345e-08, -2.4214e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 3.1199e-08, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + -1.4249e-07, 9.7789e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9581e-08, + 2.1001e-07, -2.5705e-07]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0031, -0.0135, 0.0037, 0.0323, -0.0016, 0.0303, 0.0092, 0.0159, + -0.0066, -0.0152], device='cuda:0'), grad: tensor([-7.3612e-06, 1.2014e-06, 2.4447e-07, 1.2629e-05, 3.6955e-06, + -7.4431e-06, 7.6666e-06, 9.7603e-07, 8.2841e-07, -1.2450e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 214.33, cls_loss 0.0019 cls_loss_mapping 0.0029 cls_loss_causal 0.5327 re_mapping 0.0053 re_causal 0.0158 /// teacc 98.92 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0522, -0.0057, -0.0853, ..., 0.0266, -0.0767, -0.0098], + [ 0.0416, -0.0639, -0.0193, ..., -0.0193, -0.0575, -0.2526], + [-0.0074, 0.0347, -0.0104, ..., -0.1031, 0.1189, -0.0431], + ..., + [ 0.0173, -0.0310, -0.0131, ..., -0.1097, -0.1686, 0.0306], + [-0.0210, -0.0640, -0.0212, ..., -0.0973, 0.0462, -0.1561], + [-0.1801, -0.0129, -0.0119, ..., -0.0905, -0.1258, -0.0863]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0338e-07, + 2.5611e-08, 1.3970e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.2678e-07, 1.1921e-07], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + -2.2240e-06, 7.9162e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.0850e-07, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8161e-08, + 9.5367e-07, 1.4435e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 1.4901e-08, 1.0263e-06]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0025, -0.0137, 0.0038, 0.0317, -0.0025, 0.0300, 0.0085, 0.0162, + -0.0065, -0.0142], device='cuda:0'), grad: tensor([ 2.5379e-07, -2.6375e-06, 1.0990e-06, 2.2464e-06, -4.5113e-06, + 3.0790e-06, 4.8522e-07, -1.0572e-05, 2.9672e-06, 7.5698e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 214.29, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.4973 re_mapping 0.0048 re_causal 0.0151 /// teacc 98.99 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0522, -0.0057, -0.0856, ..., 0.0265, -0.0778, -0.0099], + [ 0.0415, -0.0646, -0.0194, ..., -0.0194, -0.0575, -0.2531], + [-0.0073, 0.0345, -0.0131, ..., -0.1032, 0.1190, -0.0431], + ..., + [ 0.0173, -0.0303, -0.0103, ..., -0.1100, -0.1687, 0.0305], + [-0.0211, -0.0645, -0.0214, ..., -0.0976, 0.0462, -0.1564], + [-0.1803, -0.0128, -0.0120, ..., -0.0905, -0.1262, -0.0864]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.9977e-06, + -2.4736e-06, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4901e-07, + 2.0936e-06, 6.3796e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7136e-07, + -2.1517e-05, 1.6298e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1316e-07, + 1.4171e-05, 6.5193e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9525e-06, + 2.4773e-06, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9663e-07, + 5.9977e-07, 6.3004e-07]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0029, -0.0137, 0.0034, 0.0317, -0.0022, 0.0296, 0.0086, 0.0168, + -0.0066, -0.0144], device='cuda:0'), grad: tensor([-2.5094e-05, 1.0140e-05, -7.2598e-05, 1.3486e-05, -1.0459e-06, + 1.7062e-06, 4.7311e-06, 3.8832e-05, 2.0131e-05, 9.7454e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.21, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5354 re_mapping 0.0052 re_causal 0.0164 /// teacc 98.98 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0525, -0.0058, -0.0865, ..., 0.0268, -0.0778, -0.0099], + [ 0.0436, -0.0646, -0.0197, ..., -0.0196, -0.0574, -0.2534], + [-0.0068, 0.0346, -0.0141, ..., -0.1031, 0.1190, -0.0436], + ..., + [ 0.0161, -0.0303, -0.0103, ..., -0.1110, -0.1689, 0.0322], + [-0.0216, -0.0647, -0.0230, ..., -0.0978, 0.0468, -0.1554], + [-0.1821, -0.0127, -0.0125, ..., -0.0906, -0.1266, -0.0862]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.2352e-08, ..., -5.5879e-09, + 7.3481e-07, 2.4214e-08], + [ 3.7253e-09, 0.0000e+00, 8.6613e-08, ..., 2.7940e-09, + -7.7579e-07, 1.0617e-07], + [ 6.5193e-09, 0.0000e+00, 5.4948e-08, ..., 3.7253e-09, + -1.7239e-06, 3.1665e-08], + ..., + [ 9.3132e-10, 0.0000e+00, -1.4249e-06, ..., 0.0000e+00, + 2.1234e-07, -1.2051e-06], + [ 4.6566e-09, 0.0000e+00, 3.0734e-08, ..., 4.6566e-09, + 1.1036e-06, 2.9802e-08], + [ 9.3132e-09, 0.0000e+00, 8.8755e-07, ..., 3.7253e-09, + 1.8906e-07, 7.8045e-07]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0025, -0.0132, 0.0031, 0.0308, -0.0027, 0.0303, 0.0079, 0.0169, + -0.0065, -0.0140], device='cuda:0'), grad: tensor([ 1.6103e-06, -2.3153e-06, -2.9895e-07, 1.0375e-06, 3.6471e-06, + 1.9297e-06, 1.5646e-07, -1.4707e-05, 2.8126e-06, 6.0722e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 214.48, cls_loss 0.0019 cls_loss_mapping 0.0033 cls_loss_causal 0.5163 re_mapping 0.0053 re_causal 0.0159 /// teacc 99.07 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0533, -0.0063, -0.0866, ..., 0.0246, -0.0781, -0.0101], + [ 0.0400, -0.0647, -0.0198, ..., -0.0220, -0.0587, -0.2568], + [-0.0075, 0.0351, -0.0142, ..., -0.1037, 0.1197, -0.0412], + ..., + [ 0.0186, -0.0304, -0.0105, ..., -0.1122, -0.1691, 0.0322], + [-0.0238, -0.0651, -0.0230, ..., -0.0980, 0.0468, -0.1562], + [-0.1862, -0.0127, -0.0129, ..., -0.0878, -0.1265, -0.0857]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -5.5879e-09, + 1.0245e-08, 9.3132e-10], + [-1.9558e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.3039e-08, 5.5879e-09], + [-6.2399e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.1816e-07, 4.6566e-09], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.1002e-07, 2.7940e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-08, + -6.9849e-08, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 8.6613e-08, 6.7987e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0050, -0.0143, 0.0036, 0.0303, -0.0030, 0.0320, 0.0066, 0.0172, + -0.0068, -0.0128], device='cuda:0'), grad: tensor([ 5.7742e-08, 2.2259e-07, 4.0203e-05, 1.1921e-07, -9.8255e-07, + 1.4622e-07, 9.0338e-08, -4.2737e-05, 1.0915e-06, 1.8431e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.48, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.4798 re_mapping 0.0053 re_causal 0.0150 /// teacc 99.09 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0536, -0.0082, -0.0872, ..., 0.0250, -0.0782, -0.0098], + [ 0.0400, -0.0648, -0.0198, ..., -0.0226, -0.0587, -0.2571], + [-0.0082, 0.0363, -0.0146, ..., -0.1041, 0.1198, -0.0416], + ..., + [ 0.0195, -0.0304, -0.0106, ..., -0.1125, -0.1694, 0.0326], + [-0.0236, -0.0662, -0.0235, ..., -0.0981, 0.0471, -0.1567], + [-0.1898, -0.0156, -0.0130, ..., -0.0881, -0.1275, -0.0881]], + device='cuda:0'), grad: tensor([[ 1.7788e-07, 0.0000e+00, 0.0000e+00, ..., -1.2107e-08, + 4.3772e-08, 1.0710e-07], + [ 1.9651e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.3085e-08, 1.4808e-07], + [ 7.7859e-06, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.0291e-06, 4.5076e-06], + ..., + [-1.6317e-05, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.1961e-06, -9.0003e-06], + [ 2.0955e-07, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.0117e-07, 8.8383e-07], + [ 1.2107e-07, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.0803e-07, -1.2247e-06]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0046, -0.0131, 0.0035, 0.0305, -0.0007, 0.0318, 0.0063, 0.0167, + -0.0067, -0.0151], device='cuda:0'), grad: tensor([ 8.0001e-07, 7.1004e-06, 3.5912e-05, 6.0573e-06, 8.4043e-06, + 2.1860e-05, 7.5065e-07, -8.1062e-05, 9.9987e-06, -9.9763e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 214.02, cls_loss 0.0018 cls_loss_mapping 0.0030 cls_loss_causal 0.5108 re_mapping 0.0050 re_causal 0.0149 /// teacc 99.03 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0545, -0.0084, -0.0872, ..., 0.0249, -0.0793, -0.0100], + [ 0.0407, -0.0648, -0.0198, ..., -0.0239, -0.0590, -0.2572], + [-0.0077, 0.0365, -0.0146, ..., -0.1021, 0.1202, -0.0418], + ..., + [ 0.0194, -0.0300, -0.0106, ..., -0.1130, -0.1699, 0.0327], + [-0.0239, -0.0664, -0.0235, ..., -0.0986, 0.0471, -0.1569], + [-0.1913, -0.0171, -0.0130, ..., -0.0881, -0.1284, -0.0882]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + 2.1514e-07, 6.1467e-08], + [-1.1083e-07, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.2817e-07, 4.5635e-08], + [ 1.6764e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -1.3513e-06, 9.8720e-08], + ..., + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.5763e-07, 5.7742e-08], + [ 3.7253e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.6380e-07, 7.9162e-08], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.0978e-08, 5.4725e-06]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0047, -0.0132, 0.0040, 0.0300, -0.0007, 0.0323, 0.0062, 0.0167, + -0.0069, -0.0153], device='cuda:0'), grad: tensor([ 5.5879e-07, -1.0431e-07, -8.8848e-07, -6.8471e-06, -1.0327e-05, + -3.6322e-08, 7.9349e-07, 3.1125e-06, 3.8818e-06, 9.8050e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 214.35, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4974 re_mapping 0.0049 re_causal 0.0156 /// teacc 99.10 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0566, -0.0084, -0.0874, ..., 0.0250, -0.0796, -0.0101], + [ 0.0412, -0.0648, -0.0200, ..., -0.0242, -0.0591, -0.2573], + [-0.0081, 0.0366, -0.0147, ..., -0.1022, 0.1202, -0.0420], + ..., + [ 0.0192, -0.0299, -0.0106, ..., -0.1131, -0.1699, 0.0331], + [-0.0237, -0.0665, -0.0233, ..., -0.0987, 0.0469, -0.1565], + [-0.1935, -0.0175, -0.0131, ..., -0.0882, -0.1288, -0.0882]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3132e-06, + 1.7695e-08, -1.4016e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-08, + 2.0489e-08, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.6811e-08, + -2.7940e-09, 1.7229e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 2.7940e-09, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.7497e-08, + -6.4261e-08, 1.2945e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.8348e-07, + 2.8871e-08, 3.5856e-07]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0048, -0.0134, 0.0034, 0.0305, -0.0005, 0.0327, 0.0053, 0.0174, + -0.0072, -0.0154], device='cuda:0'), grad: tensor([-9.8497e-06, 2.0072e-05, 3.8184e-06, 2.1365e-06, 1.4920e-06, + 4.1537e-07, 1.1595e-06, -2.4557e-05, 6.4448e-07, 4.6268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 214.18, cls_loss 0.0017 cls_loss_mapping 0.0034 cls_loss_causal 0.5093 re_mapping 0.0050 re_causal 0.0150 /// teacc 98.99 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0569, -0.0086, -0.0874, ..., 0.0248, -0.0806, -0.0101], + [ 0.0417, -0.0649, -0.0200, ..., -0.0266, -0.0598, -0.2590], + [-0.0082, 0.0370, -0.0148, ..., -0.1022, 0.1206, -0.0403], + ..., + [ 0.0192, -0.0301, -0.0107, ..., -0.1144, -0.1700, 0.0331], + [-0.0238, -0.0666, -0.0233, ..., -0.0978, 0.0489, -0.1567], + [-0.1940, -0.0175, -0.0131, ..., -0.0882, -0.1293, -0.0883]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + 2.9802e-08, 1.8626e-09], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-09, 9.4995e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 5.2154e-08, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + -1.7136e-07, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.4156e-07, 3.3714e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0052, -0.0142, 0.0038, 0.0305, -0.0009, 0.0317, 0.0059, 0.0173, + -0.0062, -0.0149], device='cuda:0'), grad: tensor([ 1.4529e-07, -1.8897e-06, 6.6217e-07, -1.9539e-06, -7.7672e-07, + -9.1270e-07, 3.6880e-07, 4.4145e-07, 2.1979e-06, 1.7174e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 214.49, cls_loss 0.0017 cls_loss_mapping 0.0036 cls_loss_causal 0.4927 re_mapping 0.0051 re_causal 0.0159 /// teacc 98.86 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.0570, -0.0087, -0.0875, ..., 0.0249, -0.0820, -0.0103], + [ 0.0421, -0.0650, -0.0203, ..., -0.0268, -0.0597, -0.2591], + [-0.0083, 0.0378, -0.0150, ..., -0.1025, 0.1207, -0.0403], + ..., + [ 0.0191, -0.0305, -0.0108, ..., -0.1146, -0.1699, 0.0328], + [-0.0241, -0.0663, -0.0227, ..., -0.0983, 0.0485, -0.1573], + [-0.1943, -0.0181, -0.0132, ..., -0.0885, -0.1302, -0.0880]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-07, + 6.9849e-08, 2.4214e-08], + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4214e-08, + 1.1176e-07, 2.9150e-07], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 1.6764e-08, + 4.9639e-07, 6.5193e-08], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 4.6566e-09, + 2.8964e-07, 1.2387e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2596e-08, + 1.1034e-05, 5.9884e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.8184e-08, + 1.5367e-07, 7.8976e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0057, -0.0139, 0.0037, 0.0300, -0.0012, 0.0331, 0.0066, 0.0171, + -0.0069, -0.0147], device='cuda:0'), grad: tensor([-2.9989e-07, 3.0011e-05, 2.9523e-06, -9.2936e-04, -6.2734e-06, + 8.6784e-04, 2.0377e-06, -2.9996e-05, 6.7651e-05, -3.8520e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 214.07, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.5079 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.01 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.0571, -0.0088, -0.0875, ..., 0.0249, -0.0824, -0.0109], + [ 0.0423, -0.0650, -0.0203, ..., -0.0269, -0.0608, -0.2596], + [-0.0082, 0.0381, -0.0150, ..., -0.1028, 0.1210, -0.0393], + ..., + [ 0.0191, -0.0307, -0.0108, ..., -0.1149, -0.1700, 0.0327], + [-0.0243, -0.0665, -0.0228, ..., -0.0985, 0.0495, -0.1579], + [-0.1945, -0.0182, -0.0132, ..., -0.0886, -0.1308, -0.0889]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -8.8476e-08, + 3.7253e-09, 2.6077e-08], + [-1.0151e-07, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -4.1910e-08, 1.5087e-07], + [ 1.7695e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -2.1048e-07, 4.0978e-08], + ..., + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.3039e-07, 4.4517e-07], + [ 1.3970e-08, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + 5.0291e-08, 1.5460e-07], + [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 1.3039e-08, 2.4792e-06]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0058, -0.0145, 0.0038, 0.0306, -0.0007, 0.0326, 0.0066, 0.0170, + -0.0062, -0.0152], device='cuda:0'), grad: tensor([-1.1083e-07, -6.9197e-07, 1.7788e-07, -7.1712e-08, -8.9332e-06, + 4.2375e-07, 4.9546e-07, 1.7565e-06, 9.1270e-07, 6.0722e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 214.60, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5093 re_mapping 0.0047 re_causal 0.0149 /// teacc 99.00 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0576, -0.0088, -0.0875, ..., 0.0248, -0.0830, -0.0110], + [ 0.0424, -0.0650, -0.0203, ..., -0.0271, -0.0609, -0.2600], + [-0.0079, 0.0383, -0.0150, ..., -0.1029, 0.1212, -0.0397], + ..., + [ 0.0193, -0.0308, -0.0108, ..., -0.1152, -0.1702, 0.0337], + [-0.0243, -0.0666, -0.0228, ..., -0.0988, 0.0494, -0.1583], + [-0.1955, -0.0182, -0.0132, ..., -0.0887, -0.1316, -0.0895]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.2107e-08, 0.0000e+00, ..., -1.3765e-06, + 3.2596e-08, 2.7940e-09], + [ 0.0000e+00, 1.4901e-08, 0.0000e+00, ..., 5.7742e-08, + 5.5879e-08, 3.7253e-08], + [ 0.0000e+00, -5.1316e-07, 0.0000e+00, ..., 3.0454e-07, + -3.2663e-05, -5.8264e-06], + ..., + [ 0.0000e+00, 3.4273e-07, 0.0000e+00, ..., 1.6764e-08, + 3.2127e-05, 5.8711e-06], + [ 0.0000e+00, 2.8871e-08, 0.0000e+00, ..., 1.3225e-07, + 5.9605e-08, 6.5193e-09], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 5.7276e-07, + 1.5832e-08, 9.8720e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0062, -0.0146, 0.0039, 0.0303, -0.0001, 0.0329, 0.0068, 0.0176, + -0.0063, -0.0160], device='cuda:0'), grad: tensor([-1.1280e-05, 4.5039e-06, -1.2827e-04, 1.8775e-06, 7.2643e-07, + 4.1444e-07, 1.2824e-06, 1.2136e-04, 2.4959e-06, 7.0147e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 214.22, cls_loss 0.0018 cls_loss_mapping 0.0036 cls_loss_causal 0.5147 re_mapping 0.0051 re_causal 0.0156 /// teacc 98.84 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0580, -0.0078, -0.0875, ..., 0.0259, -0.0826, -0.0111], + [ 0.0430, -0.0654, -0.0203, ..., -0.0288, -0.0610, -0.2601], + [-0.0079, 0.0387, -0.0150, ..., -0.1031, 0.1241, -0.0399], + ..., + [ 0.0192, -0.0309, -0.0108, ..., -0.1158, -0.1705, 0.0351], + [-0.0246, -0.0669, -0.0228, ..., -0.0990, 0.0469, -0.1559], + [-0.1959, -0.0183, -0.0132, ..., -0.0890, -0.1322, -0.0895]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 1.0431e-07, 3.7253e-09], + [ 1.8813e-07, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + -7.0967e-06, 1.0245e-08], + [ 1.4901e-07, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 6.3051e-07, -2.5146e-08], + ..., + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.4645e-07, 1.3039e-08], + [ 4.6566e-08, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 6.0350e-07, 2.7940e-09], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 9.5926e-08, 1.2107e-08]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0053, -0.0156, 0.0065, 0.0301, -0.0008, 0.0330, 0.0059, 0.0191, + -0.0087, -0.0160], device='cuda:0'), grad: tensor([ 9.2573e-07, -8.2552e-05, 7.5623e-06, -2.2631e-07, 1.6272e-05, + 1.1893e-06, 5.8144e-05, 7.4804e-05, 6.3814e-06, -8.2374e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 214.17, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5163 re_mapping 0.0053 re_causal 0.0153 /// teacc 98.99 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0580, -0.0077, -0.0878, ..., 0.0266, -0.0833, -0.0104], + [ 0.0429, -0.0664, -0.0222, ..., -0.0291, -0.0612, -0.2602], + [-0.0075, 0.0392, -0.0161, ..., -0.1031, 0.1233, -0.0393], + ..., + [ 0.0191, -0.0313, -0.0111, ..., -0.1176, -0.1707, 0.0352], + [-0.0251, -0.0643, -0.0204, ..., -0.0989, 0.0479, -0.1570], + [-0.1960, -0.0185, -0.0134, ..., -0.0891, -0.1342, -0.0895]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.8673e-08, + 8.0094e-08, 1.9651e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5832e-08, + 6.7614e-07, 2.4121e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.8871e-08, + -6.9849e-07, 1.1455e-07], + ..., + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 2.7940e-09, + 1.1362e-07, 2.9523e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -6.3051e-07, 4.5449e-06], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 8.8476e-08, + 4.0885e-07, -3.3975e-05]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0052, -0.0159, 0.0059, 0.0301, -0.0009, 0.0334, 0.0050, 0.0193, + -0.0079, -0.0161], device='cuda:0'), grad: tensor([ 1.0310e-06, 2.1346e-06, -4.7125e-07, 1.6242e-06, 1.4949e-04, + 2.5481e-06, 1.5907e-06, 1.8803e-06, 2.2948e-05, -1.8251e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 214.40, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.4856 re_mapping 0.0051 re_causal 0.0149 /// teacc 98.97 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0581, -0.0077, -0.0888, ..., 0.0261, -0.0844, -0.0104], + [ 0.0430, -0.0668, -0.0222, ..., -0.0294, -0.0620, -0.2603], + [-0.0073, 0.0399, -0.0164, ..., -0.1034, 0.1233, -0.0394], + ..., + [ 0.0191, -0.0312, -0.0114, ..., -0.1181, -0.1709, 0.0351], + [-0.0251, -0.0649, -0.0208, ..., -0.1003, 0.0479, -0.1579], + [-0.1963, -0.0174, -0.0136, ..., -0.0891, -0.1354, -0.0896]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.7404e-07, + 5.5879e-09, 0.0000e+00], + [-9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-08, + 2.6077e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.2201e-08, + 3.2596e-08, 1.8626e-09], + ..., + [ 9.3132e-10, -5.5879e-09, 0.0000e+00, ..., 6.5193e-09, + 2.7940e-09, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 5.8673e-08, + -4.3772e-08, 9.3132e-10], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 2.2072e-07, + 2.2352e-08, 1.4901e-08]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0056, -0.0161, 0.0059, 0.0304, -0.0010, 0.0333, 0.0061, 0.0193, + -0.0078, -0.0162], device='cuda:0'), grad: tensor([-1.5646e-06, 4.7311e-07, 8.0094e-07, -9.7416e-07, 5.4017e-08, + 9.6858e-08, -7.6368e-08, -2.0899e-06, 9.9000e-07, 2.2724e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 214.19, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.4950 re_mapping 0.0049 re_causal 0.0152 /// teacc 99.03 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0582, -0.0098, -0.0889, ..., 0.0267, -0.0853, -0.0104], + [ 0.0443, -0.0670, -0.0222, ..., -0.0295, -0.0620, -0.2604], + [-0.0073, 0.0442, -0.0154, ..., -0.1040, 0.1233, -0.0393], + ..., + [ 0.0190, -0.0335, -0.0116, ..., -0.1216, -0.1711, 0.0351], + [-0.0264, -0.0658, -0.0208, ..., -0.0996, 0.0480, -0.1583], + [-0.1967, -0.0169, -0.0136, ..., -0.0892, -0.1363, -0.0896]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-08, + 5.8673e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.0734e-08, + 9.1270e-08, 9.3132e-10], + [ 9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 2.7940e-08, + 1.7695e-08, 9.3132e-10], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 5.3085e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0210e-07, + 1.8813e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.2596e-08, + 9.6858e-08, 0.0000e+00]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0053, -0.0160, 0.0059, 0.0302, -0.0013, 0.0331, 0.0062, 0.0193, + -0.0078, -0.0161], device='cuda:0'), grad: tensor([ 4.6566e-09, 9.5740e-07, 1.6019e-07, 7.1712e-08, 2.1279e-05, + 2.0266e-06, -3.0212e-06, -2.2113e-05, 3.4552e-07, 3.3993e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.37, cls_loss 0.0016 cls_loss_mapping 0.0036 cls_loss_causal 0.5251 re_mapping 0.0053 re_causal 0.0154 /// teacc 99.01 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.0583, -0.0099, -0.0907, ..., 0.0252, -0.0883, -0.0103], + [ 0.0443, -0.0671, -0.0222, ..., -0.0297, -0.0619, -0.2605], + [-0.0073, 0.0457, -0.0176, ..., -0.1017, 0.1234, -0.0394], + ..., + [ 0.0189, -0.0348, -0.0122, ..., -0.1224, -0.1717, 0.0346], + [-0.0265, -0.0660, -0.0228, ..., -0.0998, 0.0481, -0.1584], + [-0.1970, -0.0170, -0.0137, ..., -0.0893, -0.1391, -0.0898]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + 2.3283e-08, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 1.4901e-08, 1.2666e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0862e-07, + 0.0000e+00, 7.0781e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8533e-07, + 7.0781e-08, 1.4938e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0803e-07, + -9.8720e-08, 9.1270e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-08, + 1.0710e-07, 3.9525e-06]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0078, -0.0158, 0.0059, 0.0320, -0.0014, 0.0331, 0.0063, 0.0190, + -0.0075, -0.0164], device='cuda:0'), grad: tensor([ 2.9895e-07, 9.8906e-07, 2.5537e-06, -6.0238e-06, -3.3587e-05, + 1.2107e-06, 1.9092e-07, 1.0699e-05, 1.2536e-06, 2.2441e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 214.60, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.5039 re_mapping 0.0051 re_causal 0.0150 /// teacc 98.96 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0585, -0.0099, -0.0908, ..., 0.0258, -0.0884, -0.0103], + [ 0.0440, -0.0672, -0.0223, ..., -0.0310, -0.0622, -0.2607], + [-0.0070, 0.0471, -0.0177, ..., -0.1019, 0.1235, -0.0396], + ..., + [ 0.0188, -0.0360, -0.0124, ..., -0.1233, -0.1723, 0.0344], + [-0.0258, -0.0663, -0.0229, ..., -0.0992, 0.0482, -0.1587], + [-0.1972, -0.0170, -0.0137, ..., -0.0894, -0.1408, -0.0899]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -5.1223e-08, + 1.0990e-07, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-08, + 7.0874e-07, 1.1362e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0151e-07, + -5.6624e-07, -1.5553e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.1886e-07, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0536e-08, + 9.6019e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + 2.5872e-06, -6.5193e-09]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0073, -0.0158, 0.0058, 0.0320, -0.0013, 0.0330, 0.0055, 0.0191, + -0.0074, -0.0166], device='cuda:0'), grad: tensor([ 2.5984e-07, -1.1581e-04, -1.4435e-07, 3.1203e-05, 7.2867e-06, + -3.6895e-05, -6.2473e-06, 1.0562e-04, 3.1963e-06, 1.1712e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 214.17, cls_loss 0.0016 cls_loss_mapping 0.0023 cls_loss_causal 0.4589 re_mapping 0.0050 re_causal 0.0138 /// teacc 98.91 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0588, -0.0130, -0.0909, ..., 0.0263, -0.0887, -0.0101], + [ 0.0441, -0.0674, -0.0224, ..., -0.0314, -0.0621, -0.2614], + [-0.0071, 0.0504, -0.0177, ..., -0.1005, 0.1235, -0.0399], + ..., + [ 0.0187, -0.0380, -0.0125, ..., -0.1266, -0.1728, 0.0337], + [-0.0256, -0.0668, -0.0229, ..., -0.0997, 0.0483, -0.1565], + [-0.1975, -0.0175, -0.0138, ..., -0.0896, -0.1427, -0.0899]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7008e-08, 0.0000e+00, ..., -5.4948e-08, + 2.7101e-07, 1.2107e-08], + [ 0.0000e+00, 1.1176e-08, 0.0000e+00, ..., 1.1176e-08, + 3.7905e-07, 3.7253e-08], + [ 0.0000e+00, -4.2934e-07, 0.0000e+00, ..., 1.4901e-08, + -2.8927e-06, 6.6124e-08], + ..., + [ 0.0000e+00, 3.1851e-07, 0.0000e+00, ..., 9.3132e-10, + 2.7008e-06, 8.9407e-08], + [ 0.0000e+00, 4.8429e-08, 0.0000e+00, ..., 8.1025e-08, + -4.9800e-05, 8.4750e-08], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 5.4948e-08, + 3.5405e-05, 3.4757e-06]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0070, -0.0155, 0.0056, 0.0325, -0.0011, 0.0329, 0.0054, 0.0190, + -0.0075, -0.0165], device='cuda:0'), grad: tensor([ 5.2899e-07, 6.8452e-07, -6.1318e-06, 2.4214e-06, -1.1556e-05, + 1.9997e-05, 8.5607e-06, 6.3144e-06, -1.0818e-04, 8.7321e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 214.40, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5078 re_mapping 0.0047 re_causal 0.0145 /// teacc 98.96 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0591, -0.0131, -0.0912, ..., 0.0266, -0.0894, -0.0102], + [ 0.0436, -0.0678, -0.0224, ..., -0.0317, -0.0629, -0.2639], + [-0.0072, 0.0508, -0.0180, ..., -0.1010, 0.1237, -0.0377], + ..., + [ 0.0189, -0.0370, -0.0136, ..., -0.1269, -0.1734, 0.0339], + [-0.0246, -0.0688, -0.0233, ..., -0.1001, 0.0484, -0.1566], + [-0.1982, -0.0176, -0.0138, ..., -0.0900, -0.1444, -0.0902]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.8312e-07, 0.0000e+00, ..., -5.5879e-09, + 8.3819e-09, 9.3132e-10], + [ 0.0000e+00, 6.0536e-08, 0.0000e+00, ..., 3.7253e-09, + 3.5390e-08, 1.0245e-08], + [ 0.0000e+00, 3.5297e-07, 0.0000e+00, ..., 4.6566e-09, + -7.8231e-08, 8.3819e-09], + ..., + [ 0.0000e+00, -1.6652e-06, 0.0000e+00, ..., 9.3132e-10, + 2.4214e-08, 1.5832e-08], + [ 0.0000e+00, 1.5087e-07, 0.0000e+00, ..., -1.5832e-08, + -3.0920e-07, 1.3970e-08], + [ 0.0000e+00, 4.0606e-07, 0.0000e+00, ..., 3.7253e-09, + 1.1176e-08, 9.0338e-08]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0068, -0.0156, 0.0056, 0.0324, -0.0009, 0.0325, 0.0057, 0.0188, + -0.0073, -0.0169], device='cuda:0'), grad: tensor([ 5.1185e-06, -2.8554e-06, 6.5416e-06, 3.2894e-06, 4.9733e-07, + 3.6620e-06, 4.3493e-07, -2.6584e-05, 2.1458e-06, 7.7188e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 214.44, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.4754 re_mapping 0.0050 re_causal 0.0145 /// teacc 99.00 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.0596, -0.0131, -0.0915, ..., 0.0268, -0.0899, -0.0100], + [ 0.0439, -0.0672, -0.0215, ..., -0.0318, -0.0626, -0.2639], + [-0.0068, 0.0509, -0.0184, ..., -0.1013, 0.1251, -0.0378], + ..., + [ 0.0188, -0.0364, -0.0144, ..., -0.1258, -0.1764, 0.0338], + [-0.0246, -0.0697, -0.0239, ..., -0.0996, 0.0485, -0.1567], + [-0.1995, -0.0178, -0.0139, ..., -0.0902, -0.1475, -0.0903]], + device='cuda:0'), grad: tensor([[ 1.0245e-07, 0.0000e+00, 0.0000e+00, ..., -1.8701e-06, + -2.0675e-07, 0.0000e+00], + [ 7.6368e-08, 0.0000e+00, 0.0000e+00, ..., 4.0326e-07, + 2.7008e-08, 1.8626e-09], + [ 1.3784e-07, 0.0000e+00, 0.0000e+00, ..., 4.0792e-07, + 2.8871e-08, 9.3132e-10], + ..., + [-4.7963e-07, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + 2.0489e-08, 1.8626e-09], + [ 4.4703e-08, 0.0000e+00, 0.0000e+00, ..., 2.9057e-07, + 2.9337e-07, 2.8871e-08], + [ 4.7497e-08, 0.0000e+00, 0.0000e+00, ..., 4.0885e-07, + 5.0571e-07, -1.5106e-06]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0069, -0.0137, 0.0069, 0.0321, -0.0008, 0.0327, 0.0052, 0.0164, + -0.0073, -0.0184], device='cuda:0'), grad: tensor([-5.7518e-06, 2.6748e-06, 3.4999e-06, 1.1157e-06, 1.5840e-05, + -3.7812e-06, 3.3937e-06, -6.4783e-06, 2.4736e-06, -1.3039e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 214.29, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4923 re_mapping 0.0052 re_causal 0.0149 /// teacc 99.00 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.0598, -0.0131, -0.0915, ..., 0.0279, -0.0901, -0.0100], + [ 0.0431, -0.0683, -0.0217, ..., -0.0327, -0.0634, -0.2640], + [-0.0069, 0.0508, -0.0185, ..., -0.1018, 0.1252, -0.0378], + ..., + [ 0.0189, -0.0356, -0.0144, ..., -0.1249, -0.1764, 0.0337], + [-0.0236, -0.0700, -0.0237, ..., -0.1004, 0.0485, -0.1568], + [-0.2000, -0.0183, -0.0139, ..., -0.0910, -0.1480, -0.0906]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-07, + 1.0617e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6019e-07, + 2.2445e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.6322e-08, + -4.6473e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 1.3597e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8440e-07, + 3.0454e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9092e-07, + 1.1176e-08, 3.7253e-09]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0055, -0.0140, 0.0069, 0.0321, -0.0005, 0.0323, 0.0054, 0.0166, + -0.0073, -0.0188], device='cuda:0'), grad: tensor([-5.6438e-07, 1.7464e-05, 2.3860e-06, 5.8860e-07, 2.1532e-06, + 1.1483e-06, -1.1250e-06, 2.0117e-05, 2.3246e-06, -4.4465e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.47, cls_loss 0.0012 cls_loss_mapping 0.0024 cls_loss_causal 0.4978 re_mapping 0.0044 re_causal 0.0143 /// teacc 98.92 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0599, -0.0130, -0.0915, ..., 0.0283, -0.0901, -0.0099], + [ 0.0430, -0.0683, -0.0218, ..., -0.0333, -0.0639, -0.2646], + [-0.0069, 0.0508, -0.0185, ..., -0.1019, 0.1252, -0.0375], + ..., + [ 0.0190, -0.0354, -0.0144, ..., -0.1251, -0.1765, 0.0336], + [-0.0234, -0.0701, -0.0236, ..., -0.1035, 0.0483, -0.1568], + [-0.2003, -0.0185, -0.0139, ..., -0.0912, -0.1483, -0.0906]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.6764e-07, 0.0000e+00, ..., 2.5537e-06, + 3.1628e-06, 0.0000e+00], + [-2.1420e-08, 2.7940e-09, 0.0000e+00, ..., 8.1025e-08, + 1.0524e-07, 9.3132e-10], + [ 9.3132e-09, -4.3772e-07, 0.0000e+00, ..., -6.5230e-06, + -7.9796e-06, 1.8626e-09], + ..., + [ 2.7940e-09, 1.0151e-07, 0.0000e+00, ..., 1.5227e-06, + 1.8841e-06, -1.1176e-08], + [ 1.8626e-09, 3.4459e-08, 0.0000e+00, ..., 1.2238e-06, + 1.5227e-06, 9.3132e-10], + [ 1.8626e-09, 1.3039e-08, 0.0000e+00, ..., 2.1048e-07, + 2.6636e-07, 5.5879e-09]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0052, -0.0140, 0.0069, 0.0319, -0.0006, 0.0324, 0.0073, 0.0165, + -0.0077, -0.0186], device='cuda:0'), grad: tensor([ 8.2925e-06, 3.7588e-06, -1.9133e-05, 3.4980e-06, 2.4319e-05, + 2.3060e-06, -1.4156e-06, 1.8636e-06, 3.0145e-05, -5.3674e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.52, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.5200 re_mapping 0.0047 re_causal 0.0151 /// teacc 98.98 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0605, -0.0130, -0.0915, ..., 0.0292, -0.0898, -0.0103], + [ 0.0428, -0.0684, -0.0218, ..., -0.0354, -0.0640, -0.2648], + [-0.0074, 0.0509, -0.0185, ..., -0.1020, 0.1253, -0.0377], + ..., + [ 0.0190, -0.0354, -0.0144, ..., -0.1252, -0.1766, 0.0334], + [-0.0234, -0.0701, -0.0236, ..., -0.1036, 0.0483, -0.1567], + [-0.2006, -0.0186, -0.0139, ..., -0.0914, -0.1487, -0.0907]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -6.0536e-09, + 1.1642e-08, 0.0000e+00], + [-5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 8.1025e-08, 1.7695e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + -2.4447e-07, 9.3132e-10], + ..., + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 7.8697e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 7.9628e-08, 7.4506e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 5.0757e-08, 1.2573e-08]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0047, -0.0143, 0.0069, 0.0333, 0.0001, 0.0312, 0.0071, 0.0167, + -0.0078, -0.0190], device='cuda:0'), grad: tensor([ 2.5146e-08, 6.3982e-07, -1.9930e-07, 5.8534e-07, -1.9800e-06, + -7.7346e-07, 2.7847e-07, 2.3935e-07, 9.2108e-07, 2.8266e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 214.26, cls_loss 0.0018 cls_loss_mapping 0.0029 cls_loss_causal 0.5053 re_mapping 0.0047 re_causal 0.0137 /// teacc 99.03 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0610, -0.0131, -0.0915, ..., 0.0272, -0.0901, -0.0108], + [ 0.0434, -0.0685, -0.0218, ..., -0.0364, -0.0647, -0.2650], + [-0.0090, 0.0510, -0.0185, ..., -0.1027, 0.1253, -0.0384], + ..., + [ 0.0187, -0.0353, -0.0144, ..., -0.1256, -0.1765, 0.0341], + [-0.0215, -0.0703, -0.0236, ..., -0.1046, 0.0483, -0.1565], + [-0.2012, -0.0190, -0.0139, ..., -0.0888, -0.1492, -0.0909]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -1.0245e-08, + 4.2375e-08, 1.8626e-09], + [-2.2352e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 8.5682e-08, 7.2177e-08], + [ 1.3504e-08, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + -2.7148e-07, 1.3039e-08], + ..., + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.1781e-07, 3.2131e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 1.5739e-07, 2.6543e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 1.5134e-07, 9.8255e-08]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0070, -0.0141, 0.0065, 0.0330, 0.0001, 0.0305, 0.0083, 0.0171, + -0.0079, -0.0178], device='cuda:0'), grad: tensor([ 3.9209e-07, 5.2759e-07, 7.1302e-06, 3.1609e-06, -5.7593e-06, + 1.0077e-06, 1.1753e-06, -1.2860e-05, 3.0566e-06, 2.1663e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 214.69, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5359 re_mapping 0.0045 re_causal 0.0141 /// teacc 98.99 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.0624, -0.0131, -0.0915, ..., 0.0272, -0.0904, -0.0110], + [ 0.0431, -0.0686, -0.0218, ..., -0.0380, -0.0653, -0.2652], + [-0.0100, 0.0511, -0.0185, ..., -0.1029, 0.1254, -0.0386], + ..., + [ 0.0188, -0.0353, -0.0144, ..., -0.1265, -0.1766, 0.0341], + [-0.0210, -0.0703, -0.0236, ..., -0.1017, 0.0500, -0.1558], + [-0.2028, -0.0191, -0.0139, ..., -0.0888, -0.1499, -0.0910]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3097e-06, + 2.9337e-08, 0.0000e+00], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-08, + 2.8405e-08, 9.3132e-10], + [ 9.3132e-10, -0.0000e+00, 0.0000e+00, ..., 1.1409e-07, + -7.6788e-07, 4.6566e-10], + ..., + [-1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 5.5740e-07, + 1.1781e-07, 2.3283e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.3667e-07, + 1.1595e-07, 2.3283e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 5.5833e-07, + 1.5832e-08, -1.6391e-07]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0070, -0.0142, 0.0064, 0.0331, 0.0001, 0.0295, 0.0054, 0.0172, + -0.0062, -0.0179], device='cuda:0'), grad: tensor([-5.3644e-06, 9.5926e-08, -9.0664e-07, 2.1979e-06, 1.5879e-06, + 6.0583e-07, -2.4633e-07, 1.5376e-06, 1.0617e-06, -5.6205e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.29, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.5375 re_mapping 0.0045 re_causal 0.0145 /// teacc 98.94 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.0627, -0.0131, -0.0915, ..., 0.0274, -0.0906, -0.0111], + [ 0.0429, -0.0686, -0.0218, ..., -0.0398, -0.0661, -0.2654], + [-0.0104, 0.0511, -0.0185, ..., -0.1068, 0.1249, -0.0386], + ..., + [ 0.0189, -0.0353, -0.0144, ..., -0.1271, -0.1767, 0.0341], + [-0.0200, -0.0704, -0.0236, ..., -0.1010, 0.0506, -0.1556], + [-0.2036, -0.0190, -0.0139, ..., -0.0889, -0.1509, -0.0911]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2247e-06, + -2.5192e-07, -2.3935e-07], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 8.5682e-08, 2.3283e-09], + [ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 1.4575e-07, + -5.0254e-06, -1.0710e-08], + ..., + [ 4.6566e-10, 9.3132e-09, 0.0000e+00, ..., 6.9849e-09, + 5.9120e-06, 3.1199e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.8545e-07, + -1.1809e-06, 5.7276e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0210e-07, + 4.8708e-07, 1.1548e-07]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0070, -0.0141, 0.0059, 0.0322, -0.0001, 0.0296, 0.0056, 0.0173, + -0.0056, -0.0179], device='cuda:0'), grad: tensor([-7.1228e-06, 2.3935e-07, -8.9332e-06, 5.1828e-07, 3.8557e-07, + -6.6636e-07, 9.3458e-07, 1.1496e-05, -1.0096e-06, 4.1388e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 214.32, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.5035 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.05 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0628, -0.0149, -0.0923, ..., 0.0274, -0.0910, -0.0112], + [ 0.0428, -0.0693, -0.0218, ..., -0.0421, -0.0668, -0.2654], + [-0.0107, 0.0513, -0.0201, ..., -0.1070, 0.1249, -0.0387], + ..., + [ 0.0192, -0.0347, -0.0153, ..., -0.1277, -0.1767, 0.0341], + [-0.0200, -0.0710, -0.0245, ..., -0.1011, 0.0506, -0.1557], + [-0.2041, -0.0164, -0.0140, ..., -0.0886, -0.1518, -0.0911]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -1.5730e-06, + -4.9779e-07, 3.0268e-08], + [-3.3854e-07, 9.3132e-10, 0.0000e+00, ..., -8.8708e-07, + -5.4808e-07, 1.6624e-07], + [ 9.5461e-08, 9.3132e-10, 0.0000e+00, ..., 1.1511e-06, + 6.7893e-07, 5.1968e-07], + ..., + [ 1.0710e-08, -8.8476e-09, 0.0000e+00, ..., 3.5390e-08, + 9.7789e-08, 2.8405e-08], + [ 2.2817e-08, 9.3132e-10, 0.0000e+00, ..., 1.5367e-07, + 3.5763e-07, 1.2945e-07], + [ 2.3283e-09, 1.8626e-09, 0.0000e+00, ..., 1.1921e-07, + 1.2293e-07, 1.2899e-07]], device='cuda:0') +Epoch 227, bias, value: tensor([-7.2020e-03, -1.4230e-02, 5.3840e-03, 3.2020e-02, -6.8047e-05, + 3.1221e-02, 5.3288e-03, 1.7816e-02, -5.6898e-03, -1.7882e-02], + device='cuda:0'), grad: tensor([-3.0622e-06, -2.1365e-06, 5.3085e-06, -6.8769e-06, 5.6811e-08, + 1.2955e-06, 2.5444e-06, 4.5169e-07, 1.5683e-06, 8.5216e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 214.41, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.5043 re_mapping 0.0046 re_causal 0.0146 /// teacc 98.98 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0629, -0.0201, -0.0923, ..., 0.0273, -0.0909, -0.0088], + [ 0.0426, -0.0718, -0.0218, ..., -0.0419, -0.0667, -0.2656], + [-0.0110, 0.0547, -0.0201, ..., -0.1071, 0.1249, -0.0393], + ..., + [ 0.0192, -0.0373, -0.0153, ..., -0.1280, -0.1767, 0.0332], + [-0.0193, -0.0732, -0.0245, ..., -0.1012, 0.0507, -0.1558], + [-0.2043, -0.0115, -0.0140, ..., -0.0882, -0.1545, -0.0912]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.0245e-08, 0.0000e+00, ..., -2.7716e-06, + -6.3889e-07, 0.0000e+00], + [-9.3132e-10, 6.0536e-09, 0.0000e+00, ..., 1.2433e-07, + 4.5635e-08, 9.3132e-10], + [ 0.0000e+00, 5.1223e-09, 0.0000e+00, ..., 3.9116e-07, + 9.3598e-08, 4.6566e-10], + ..., + [ 4.6566e-10, 6.3330e-08, 0.0000e+00, ..., 4.7963e-08, + 6.3796e-08, 1.8626e-09], + [ 0.0000e+00, 1.7695e-08, 0.0000e+00, ..., 1.1129e-07, + -4.4703e-07, 4.6566e-10], + [ 0.0000e+00, -1.5832e-07, 0.0000e+00, ..., 1.2713e-06, + 5.6904e-07, -1.1176e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([-7.2335e-03, -1.4186e-02, 5.3442e-03, 2.9328e-02, 1.8841e-05, + 3.3840e-02, 5.3064e-03, 1.7720e-02, -5.5352e-03, -1.8148e-02], + device='cuda:0'), grad: tensor([-7.7337e-06, 5.5647e-07, 1.7490e-06, 3.0966e-07, 3.5554e-05, + 1.0021e-06, 8.6986e-07, 5.4576e-07, -2.9989e-07, -3.2544e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 214.25, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.5340 re_mapping 0.0045 re_causal 0.0143 /// teacc 99.05 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0632, -0.0204, -0.0951, ..., 0.0273, -0.0911, -0.0091], + [ 0.0430, -0.0752, -0.0222, ..., -0.0421, -0.0669, -0.2658], + [-0.0115, 0.0550, -0.0223, ..., -0.1071, 0.1249, -0.0387], + ..., + [ 0.0193, -0.0359, -0.0177, ..., -0.1282, -0.1768, 0.0331], + [-0.0193, -0.0741, -0.0274, ..., -0.1012, 0.0507, -0.1560], + [-0.2052, -0.0110, -0.0148, ..., -0.0882, -0.1551, -0.0917]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.5832e-08, + -3.7253e-09, 1.8626e-09], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 5.1223e-09, 1.3039e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.8871e-08, 5.1223e-09], + ..., + [ 2.7940e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 2.4214e-08], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 1.3970e-09, + -4.3772e-08, 4.1910e-09], + [ 9.3132e-10, -1.2107e-08, 0.0000e+00, ..., 5.5879e-09, + 6.0536e-09, -1.1642e-08]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0072, -0.0143, 0.0053, 0.0307, 0.0006, 0.0333, 0.0053, 0.0174, + -0.0055, -0.0180], device='cuda:0'), grad: tensor([ 1.6764e-08, 4.2059e-06, 1.2055e-05, -9.9186e-07, 8.9966e-07, + 4.2235e-07, 5.7276e-08, -1.8254e-05, 3.1255e-06, -1.5246e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 214.35, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.5010 re_mapping 0.0045 re_causal 0.0137 /// teacc 99.02 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0632, -0.0214, -0.0952, ..., 0.0274, -0.0913, -0.0082], + [ 0.0430, -0.0768, -0.0224, ..., -0.0418, -0.0671, -0.2662], + [-0.0116, 0.0549, -0.0226, ..., -0.1073, 0.1250, -0.0388], + ..., + [ 0.0193, -0.0331, -0.0182, ..., -0.1283, -0.1768, 0.0331], + [-0.0194, -0.0756, -0.0278, ..., -0.1011, 0.0509, -0.1560], + [-0.2053, -0.0115, -0.0149, ..., -0.0883, -0.1559, -0.0919]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.5193e-08, 0.0000e+00, ..., -1.8440e-07, + 6.3796e-08, 1.3504e-08], + [-3.1199e-08, 1.8626e-09, 0.0000e+00, ..., 7.8231e-08, + 9.6392e-08, 1.3039e-07], + [ 1.8161e-08, 5.5879e-09, 0.0000e+00, ..., 6.3190e-07, + 6.8359e-07, 2.9383e-07], + ..., + [ 6.5193e-09, 4.6566e-10, 0.0000e+00, ..., 2.0489e-08, + 3.8184e-08, 6.4727e-08], + [ 4.6566e-09, 1.8626e-09, 0.0000e+00, ..., 1.0338e-07, + 3.0734e-08, 2.9337e-07], + [ 0.0000e+00, 4.2375e-08, 0.0000e+00, ..., 1.8859e-07, + 5.3551e-08, 5.8673e-07]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0072, -0.0143, 0.0052, 0.0308, 0.0009, 0.0326, 0.0053, 0.0175, + -0.0055, -0.0182], device='cuda:0'), grad: tensor([-4.6156e-06, 7.9861e-07, 4.3698e-06, 5.4482e-07, 8.8066e-06, + 1.3448e-06, -1.9789e-05, -1.6578e-07, 2.2743e-06, 6.3814e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 214.39, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.5202 re_mapping 0.0045 re_causal 0.0140 /// teacc 98.92 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0647, -0.0215, -0.0952, ..., 0.0276, -0.0913, -0.0084], + [ 0.0436, -0.0773, -0.0224, ..., -0.0427, -0.0673, -0.2664], + [-0.0111, 0.0568, -0.0226, ..., -0.1073, 0.1251, -0.0387], + ..., + [ 0.0189, -0.0342, -0.0182, ..., -0.1286, -0.1770, 0.0330], + [-0.0198, -0.0773, -0.0278, ..., -0.1012, 0.0509, -0.1562], + [-0.2062, -0.0112, -0.0149, ..., -0.0884, -0.1569, -0.0920]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.4435e-08, 0.0000e+00, ..., 2.2026e-07, + 2.9523e-07, 2.3283e-09], + [-2.0303e-07, 4.1910e-09, 0.0000e+00, ..., -4.0047e-08, + 4.5588e-07, 4.6566e-10], + [ 4.3306e-08, 1.3970e-09, 0.0000e+00, ..., 1.4668e-07, + 1.6391e-06, 0.0000e+00], + ..., + [ 1.5832e-08, -8.3819e-09, 0.0000e+00, ..., 2.0489e-08, + 1.9427e-06, 1.3970e-09], + [ 5.6811e-08, 6.9849e-09, 0.0000e+00, ..., 1.2573e-07, + -5.5730e-06, 4.6566e-10], + [ 8.8476e-09, -3.5856e-08, 0.0000e+00, ..., 4.6566e-09, + 2.0768e-07, -2.7940e-09]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0072, -0.0135, 0.0053, 0.0307, 0.0007, 0.0321, 0.0055, 0.0169, + -0.0055, -0.0183], device='cuda:0'), grad: tensor([ 9.6671e-07, 1.8198e-06, 5.2229e-06, 3.2410e-06, 1.7891e-06, + 8.8215e-06, -7.7412e-06, 1.3418e-05, -2.6926e-05, -6.3283e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 214.17, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5263 re_mapping 0.0047 re_causal 0.0142 /// teacc 98.94 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.0655, -0.0216, -0.0952, ..., 0.0255, -0.0945, -0.0087], + [ 0.0442, -0.0775, -0.0224, ..., -0.0440, -0.0699, -0.2666], + [-0.0117, 0.0571, -0.0230, ..., -0.1074, 0.1252, -0.0388], + ..., + [ 0.0198, -0.0342, -0.0180, ..., -0.1287, -0.1772, 0.0331], + [-0.0201, -0.0779, -0.0279, ..., -0.1013, 0.0513, -0.1562], + [-0.2120, -0.0113, -0.0149, ..., -0.0884, -0.1581, -0.0920]], + device='cuda:0'), grad: tensor([[ 7.3109e-08, 0.0000e+00, 0.0000e+00, ..., -1.0729e-06, + -2.7940e-09, 4.1910e-09], + [ 1.2573e-08, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + 1.7416e-07, 1.3039e-08], + [ 8.0559e-08, 0.0000e+00, 0.0000e+00, ..., 1.0896e-07, + 2.3186e-05, 4.6566e-09], + ..., + [ 1.8626e-08, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 1.4808e-07, 7.9162e-09], + [-2.6682e-07, 0.0000e+00, 0.0000e+00, ..., 8.5216e-08, + -2.5034e-05, 1.1176e-08], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 4.9919e-07, + 5.3504e-07, -6.9384e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0087, -0.0126, 0.0052, 0.0306, -0.0009, 0.0311, 0.0068, 0.0159, + -0.0051, -0.0176], device='cuda:0'), grad: tensor([-2.7642e-06, 6.9756e-07, 5.6207e-05, 1.3653e-06, 2.1160e-06, + 8.2515e-07, 1.7267e-06, 5.7276e-07, -5.9485e-05, -1.1930e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 214.17, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4880 re_mapping 0.0046 re_causal 0.0146 /// teacc 98.98 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.0660, -0.0216, -0.0952, ..., 0.0255, -0.0947, -0.0088], + [ 0.0447, -0.0780, -0.0225, ..., -0.0448, -0.0699, -0.2667], + [-0.0121, 0.0571, -0.0230, ..., -0.1076, 0.1252, -0.0389], + ..., + [ 0.0199, -0.0337, -0.0180, ..., -0.1294, -0.1772, 0.0331], + [-0.0203, -0.0780, -0.0279, ..., -0.1013, 0.0513, -0.1563], + [-0.2134, -0.0117, -0.0151, ..., -0.0882, -0.1589, -0.0925]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.3085e-08, + 1.2573e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 1.9092e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -9.6392e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 1.0012e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0536e-09, + -3.7299e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2573e-08, + 1.6438e-07, 3.2596e-09]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0087, -0.0127, 0.0051, 0.0304, -0.0004, 0.0313, 0.0068, 0.0161, + -0.0050, -0.0183], device='cuda:0'), grad: tensor([-1.0105e-07, 1.4668e-07, -3.2131e-08, 1.5413e-07, 4.4238e-08, + 1.8068e-07, 2.0349e-07, -2.1886e-07, -8.7637e-07, 4.8429e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 214.31, cls_loss 0.0017 cls_loss_mapping 0.0039 cls_loss_causal 0.4971 re_mapping 0.0047 re_causal 0.0143 /// teacc 98.93 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.0660, -0.0216, -0.0967, ..., 0.0258, -0.0948, -0.0091], + [ 0.0449, -0.0805, -0.0230, ..., -0.0449, -0.0701, -0.2669], + [-0.0124, 0.0572, -0.0243, ..., -0.1077, 0.1252, -0.0387], + ..., + [ 0.0199, -0.0329, -0.0182, ..., -0.1297, -0.1773, 0.0335], + [-0.0203, -0.0782, -0.0283, ..., -0.0999, 0.0529, -0.1565], + [-0.2138, -0.0117, -0.0160, ..., -0.0883, -0.1600, -0.0926]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 8.6147e-08, 4.6566e-10], + [-9.7789e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.0047e-08, 9.7789e-09], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -2.8824e-07, 2.4680e-08], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.2352e-08, 2.7940e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 4.7032e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.8161e-08, 1.4435e-08]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0086, -0.0127, 0.0050, 0.0312, -0.0005, 0.0304, 0.0047, 0.0163, + -0.0036, -0.0183], device='cuda:0'), grad: tensor([ 1.8859e-07, 1.7518e-06, 1.3877e-07, 5.4240e-06, -2.0443e-07, + -5.2620e-07, 1.1781e-07, -8.9034e-06, 1.6093e-06, 3.7719e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.22, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.5276 re_mapping 0.0044 re_causal 0.0140 /// teacc 99.01 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.0655, -0.0216, -0.0968, ..., 0.0259, -0.0950, -0.0069], + [ 0.0449, -0.0808, -0.0230, ..., -0.0450, -0.0702, -0.2669], + [-0.0122, 0.0572, -0.0243, ..., -0.1078, 0.1258, -0.0386], + ..., + [ 0.0207, -0.0327, -0.0182, ..., -0.1303, -0.1784, 0.0335], + [-0.0207, -0.0783, -0.0283, ..., -0.1004, 0.0527, -0.1566], + [-0.2144, -0.0119, -0.0161, ..., -0.0884, -0.1607, -0.0927]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.1246e-08, + 2.7381e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.1420e-08, + 5.6345e-08, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.0047e-08, + 7.7486e-07, 0.0000e+00], + ..., + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 2.5611e-08, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -1.3411e-07, + -1.0328e-06, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 3.7719e-08, + 7.5437e-08, 2.3283e-09]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0083, -0.0129, 0.0057, 0.0312, -0.0006, 0.0302, 0.0051, 0.0160, + -0.0039, -0.0186], device='cuda:0'), grad: tensor([-4.7171e-07, 1.8850e-06, 2.5909e-06, -1.0431e-06, -6.5088e-05, + 2.0526e-06, 1.3188e-06, 4.0270e-06, -2.5369e-06, 5.7250e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.17, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5052 re_mapping 0.0046 re_causal 0.0146 /// teacc 98.88 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.0681, -0.0216, -0.0968, ..., 0.0262, -0.0949, -0.0079], + [ 0.0457, -0.0808, -0.0231, ..., -0.0451, -0.0704, -0.2671], + [-0.0138, 0.0572, -0.0244, ..., -0.1078, 0.1259, -0.0386], + ..., + [ 0.0217, -0.0327, -0.0183, ..., -0.1305, -0.1785, 0.0336], + [-0.0212, -0.0783, -0.0283, ..., -0.1006, 0.0527, -0.1567], + [-0.2157, -0.0119, -0.0161, ..., -0.0883, -0.1612, -0.0928]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -3.7719e-08, + 8.3819e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + 4.7963e-08, 0.0000e+00], + [-2.3749e-08, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + -4.0280e-07, 0.0000e+00], + ..., + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 6.5658e-08, + 1.9278e-07, 0.0000e+00], + [ 8.8476e-09, 0.0000e+00, 0.0000e+00, ..., 3.4459e-07, + 1.1176e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -4.7171e-07, + 1.7229e-08, 0.0000e+00]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0081, -0.0130, 0.0057, 0.0314, -0.0002, 0.0302, 0.0050, 0.0160, + -0.0040, -0.0190], device='cuda:0'), grad: tensor([ 6.4867e-07, 6.8592e-07, -4.9965e-07, 1.9763e-06, -1.5013e-06, + 3.1060e-07, 1.3690e-07, 2.4904e-06, 1.0669e-05, -1.4916e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 214.22, cls_loss 0.0017 cls_loss_mapping 0.0021 cls_loss_causal 0.4881 re_mapping 0.0046 re_causal 0.0134 /// teacc 99.05 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.0673, -0.0216, -0.0968, ..., 0.0275, -0.0937, -0.0083], + [ 0.0405, -0.0808, -0.0232, ..., -0.0455, -0.0715, -0.2674], + [-0.0131, 0.0572, -0.0245, ..., -0.1094, 0.1258, -0.0390], + ..., + [ 0.0238, -0.0327, -0.0184, ..., -0.1310, -0.1786, 0.0334], + [-0.0197, -0.0783, -0.0286, ..., -0.1006, 0.0527, -0.1561], + [-0.2192, -0.0118, -0.0161, ..., -0.0885, -0.1617, -0.0928]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3411e-07, + -1.1595e-07, 9.3132e-10], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.7032e-08, 6.0536e-09], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -7.3016e-07, 1.8626e-09], + ..., + [-1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.2177e-08, 6.5193e-09], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 4.9360e-08, + 5.3551e-07, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 1.1642e-08, 2.0955e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0066, -0.0133, 0.0053, 0.0316, -0.0011, 0.0302, 0.0049, 0.0162, + -0.0040, -0.0180], device='cuda:0'), grad: tensor([-2.7847e-07, -1.2629e-06, -6.9384e-07, 3.5809e-07, 9.3132e-10, + -2.3330e-07, 2.3283e-07, 2.0443e-07, 9.4064e-07, 7.3202e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 214.27, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4963 re_mapping 0.0045 re_causal 0.0138 /// teacc 99.07 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.0668, -0.0216, -0.0968, ..., 0.0283, -0.0937, -0.0085], + [ 0.0408, -0.0813, -0.0232, ..., -0.0460, -0.0709, -0.2676], + [-0.0138, 0.0571, -0.0245, ..., -0.1096, 0.1258, -0.0412], + ..., + [ 0.0244, -0.0322, -0.0184, ..., -0.1313, -0.1787, 0.0327], + [-0.0198, -0.0783, -0.0286, ..., -0.1007, 0.0527, -0.1564], + [-0.2201, -0.0123, -0.0161, ..., -0.0893, -0.1630, -0.0929]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6186e-06, + -1.3085e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7206e-07, + 3.3211e-06, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.4482e-08, + -3.7644e-06, -1.5367e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.1246e-08, + 1.7742e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.1479e-07, + 3.6461e-07, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.2387e-07, + 4.9826e-08, 2.3283e-09]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0056, -0.0125, 0.0051, 0.0316, -0.0006, 0.0302, 0.0049, 0.0163, + -0.0048, -0.0185], device='cuda:0'), grad: tensor([-5.3048e-06, 1.7002e-05, -5.8971e-06, -6.9803e-07, 1.2359e-06, + 1.6615e-06, -3.4133e-07, -1.0572e-05, 3.0231e-06, -1.0477e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 214.24, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5074 re_mapping 0.0044 re_causal 0.0137 /// teacc 98.92 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.0670, -0.0217, -0.0969, ..., 0.0292, -0.0934, -0.0097], + [ 0.0407, -0.0815, -0.0233, ..., -0.0462, -0.0711, -0.2678], + [-0.0140, 0.0571, -0.0246, ..., -0.1097, 0.1258, -0.0417], + ..., + [ 0.0246, -0.0320, -0.0187, ..., -0.1323, -0.1786, 0.0327], + [-0.0197, -0.0783, -0.0289, ..., -0.1006, 0.0528, -0.1562], + [-0.2216, -0.0124, -0.0162, ..., -0.0895, -0.1650, -0.0935]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -2.4680e-08, + 6.1933e-08, 0.0000e+00], + [ 2.6543e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 9.7323e-08, 1.8626e-09], + [ 8.8476e-09, 0.0000e+00, 0.0000e+00, ..., -5.7276e-08, + -8.8802e-07, 0.0000e+00], + ..., + [-8.4285e-08, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 3.6089e-07, 0.0000e+00], + [ 1.3504e-08, 0.0000e+00, 0.0000e+00, ..., 1.3504e-08, + -1.3411e-06, 0.0000e+00], + [ 2.0955e-08, 0.0000e+00, 0.0000e+00, ..., 2.2817e-08, + 1.4212e-06, 9.3132e-10]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0049, -0.0126, 0.0047, 0.0315, 0.0002, 0.0303, 0.0045, 0.0167, + -0.0046, -0.0191], device='cuda:0'), grad: tensor([ 8.8941e-08, 4.1584e-07, -1.0021e-06, 1.0338e-06, 4.5588e-07, + -1.3234e-06, 6.7055e-07, 7.0967e-07, -3.0622e-06, 2.0023e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.50, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.4912 re_mapping 0.0044 re_causal 0.0136 /// teacc 98.98 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.0672, -0.0217, -0.0969, ..., 0.0298, -0.0933, -0.0093], + [ 0.0395, -0.0815, -0.0233, ..., -0.0464, -0.0703, -0.2679], + [-0.0124, 0.0571, -0.0246, ..., -0.1097, 0.1256, -0.0417], + ..., + [ 0.0247, -0.0320, -0.0187, ..., -0.1334, -0.1787, 0.0326], + [-0.0197, -0.0783, -0.0289, ..., -0.1007, 0.0529, -0.1562], + [-0.2227, -0.0124, -0.0162, ..., -0.0906, -0.1666, -0.0936]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -3.8277e-07, + 1.7313e-06, 4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 4.3400e-07, + 1.3066e-06, 9.3132e-10], + [-1.5832e-08, 0.0000e+00, 0.0000e+00, ..., 4.4657e-07, + -1.9697e-07, -7.4506e-09], + ..., + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 1.0198e-07, + 4.3446e-07, 5.1223e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -3.0939e-06, + -8.3521e-06, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.7218e-07, + 1.2852e-06, 2.3283e-09]], device='cuda:0') +Epoch 240, bias, value: tensor([-4.4404e-03, -1.2073e-02, 4.1750e-03, 3.1554e-02, 4.0330e-05, + 3.0267e-02, 4.5086e-03, 1.6711e-02, -4.4796e-03, -1.9461e-02], + device='cuda:0'), grad: tensor([ 2.5928e-06, 3.7104e-06, 2.7791e-06, 3.0249e-06, 1.0552e-06, + 4.7423e-06, 3.3490e-06, -2.7958e-06, -2.3648e-05, 5.1931e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 214.48, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4738 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.0668, -0.0217, -0.0969, ..., 0.0301, -0.0934, -0.0093], + [ 0.0378, -0.0815, -0.0233, ..., -0.0441, -0.0706, -0.2681], + [-0.0147, 0.0571, -0.0246, ..., -0.1098, 0.1256, -0.0416], + ..., + [ 0.0272, -0.0320, -0.0187, ..., -0.1343, -0.1787, 0.0325], + [-0.0172, -0.0783, -0.0289, ..., -0.1006, 0.0530, -0.1563], + [-0.2237, -0.0124, -0.0162, ..., -0.0910, -0.1675, -0.0936]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, 0.0000e+00, 0.0000e+00, ..., -5.8208e-08, + -1.7229e-08, 2.7940e-09], + [ 6.7055e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 3.4925e-08, 2.1420e-08], + [ 2.9802e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -1.7788e-07, 1.8626e-08], + ..., + [ 4.2375e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.0070e-08, 2.0955e-08], + [ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + 1.7695e-08, 3.7253e-09], + [ 1.2200e-07, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 3.2596e-09, 4.4703e-08]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0043, -0.0119, 0.0038, 0.0317, -0.0008, 0.0302, 0.0042, 0.0170, + -0.0044, -0.0187], device='cuda:0'), grad: tensor([-6.7521e-08, 1.4529e-06, 2.2464e-06, 1.1437e-06, -5.3551e-07, + 5.4482e-08, 3.7486e-07, -5.3607e-06, 1.3364e-07, 5.4995e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 214.83, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5257 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.00 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.0672, -0.0217, -0.0969, ..., 0.0302, -0.0937, -0.0093], + [ 0.0379, -0.0815, -0.0233, ..., -0.0442, -0.0709, -0.2682], + [-0.0150, 0.0571, -0.0246, ..., -0.1099, 0.1257, -0.0419], + ..., + [ 0.0275, -0.0320, -0.0187, ..., -0.1342, -0.1788, 0.0322], + [-0.0172, -0.0783, -0.0289, ..., -0.1007, 0.0531, -0.1558], + [-0.2256, -0.0124, -0.0163, ..., -0.0914, -0.1689, -0.0936]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.0641e-07, + 1.7323e-07, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9802e-08, + 1.2247e-07, 2.1886e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.8184e-08, + -4.0559e-07, -1.4994e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7229e-08, + 3.1479e-07, 4.8894e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0803e-07, + -4.9267e-07, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4948e-07, + -2.0117e-07, 1.8626e-09]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0040, -0.0123, 0.0036, 0.0333, -0.0007, 0.0284, 0.0042, 0.0177, + -0.0044, -0.0192], device='cuda:0'), grad: tensor([-5.4622e-07, 2.8033e-07, -6.0583e-07, 5.1558e-06, 5.3970e-07, + 2.4270e-06, -8.3260e-07, 9.0711e-07, -1.8906e-07, -7.1637e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 214.65, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.5003 re_mapping 0.0043 re_causal 0.0138 /// teacc 99.03 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.0674, -0.0217, -0.0969, ..., 0.0307, -0.0941, -0.0095], + [ 0.0378, -0.0815, -0.0233, ..., -0.0449, -0.0710, -0.2687], + [-0.0150, 0.0574, -0.0246, ..., -0.1102, 0.1258, -0.0427], + ..., + [ 0.0280, -0.0322, -0.0187, ..., -0.1345, -0.1789, 0.0316], + [-0.0173, -0.0785, -0.0289, ..., -0.1007, 0.0531, -0.1561], + [-0.2284, -0.0124, -0.0163, ..., -0.0914, -0.1696, -0.0937]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.5367e-08, + 8.8476e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 7.9162e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -9.3132e-09, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.6298e-08, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -1.1548e-07, + -1.4408e-06, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 2.8871e-08, 1.7229e-08]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0035, -0.0124, 0.0036, 0.0346, 0.0001, 0.0271, 0.0042, 0.0176, + -0.0044, -0.0196], device='cuda:0'), grad: tensor([ 1.4994e-07, -4.4294e-06, 3.5875e-06, 2.0750e-06, -2.1607e-07, + 8.3586e-07, 2.6487e-06, -2.0005e-06, -2.8983e-06, 2.2026e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 214.88, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4919 re_mapping 0.0045 re_causal 0.0136 /// teacc 99.02 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.0676, -0.0217, -0.0969, ..., 0.0308, -0.0942, -0.0096], + [ 0.0375, -0.0815, -0.0233, ..., -0.0450, -0.0711, -0.2688], + [-0.0146, 0.0574, -0.0247, ..., -0.1102, 0.1259, -0.0427], + ..., + [ 0.0282, -0.0322, -0.0189, ..., -0.1346, -0.1791, 0.0315], + [-0.0174, -0.0786, -0.0289, ..., -0.1008, 0.0531, -0.1562], + [-0.2302, -0.0124, -0.0163, ..., -0.0915, -0.1704, -0.0939]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.9384e-08, + 9.7789e-09, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.2573e-08, + 9.0804e-08, 2.1420e-08], + [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 2.6077e-07, + 2.2706e-06, 2.1420e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 2.4214e-08, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.5239e-07, + -2.4755e-06, 4.6566e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.3504e-08, + 1.3504e-08, 2.3283e-08]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0035, -0.0127, 0.0034, 0.0344, 0.0005, 0.0272, 0.0042, 0.0176, + -0.0040, -0.0195], device='cuda:0'), grad: tensor([-1.7835e-07, 1.0664e-07, 4.2208e-06, 1.2619e-07, -2.4214e-07, + -1.2619e-07, 1.2247e-07, 2.4680e-08, -4.2282e-06, 1.9092e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 214.88, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4654 re_mapping 0.0046 re_causal 0.0140 /// teacc 98.94 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.0714, -0.0217, -0.0969, ..., 0.0304, -0.0961, -0.0097], + [ 0.0375, -0.0816, -0.0233, ..., -0.0452, -0.0717, -0.2689], + [-0.0147, 0.0589, -0.0247, ..., -0.1102, 0.1261, -0.0428], + ..., + [ 0.0288, -0.0337, -0.0189, ..., -0.1348, -0.1793, 0.0315], + [-0.0176, -0.0790, -0.0289, ..., -0.1008, 0.0534, -0.1562], + [-0.2338, -0.0124, -0.0163, ..., -0.0918, -0.1722, -0.0940]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 5.8208e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 8.3819e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2573e-08, + -4.0047e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 2.8871e-07, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3749e-08, + 9.2201e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-09, 2.7940e-09]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0050, -0.0128, 0.0035, 0.0344, 0.0005, 0.0273, 0.0042, 0.0177, + -0.0034, -0.0201], device='cuda:0'), grad: tensor([ 1.8347e-07, -4.1723e-06, 8.0559e-08, 3.4040e-07, 3.3062e-07, + -1.2573e-08, -1.0664e-07, 3.0231e-06, 2.8405e-07, 5.6811e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 214.62, cls_loss 0.0014 cls_loss_mapping 0.0026 cls_loss_causal 0.4923 re_mapping 0.0048 re_causal 0.0138 /// teacc 98.97 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.0716, -0.0217, -0.0970, ..., 0.0300, -0.0967, -0.0098], + [ 0.0375, -0.0816, -0.0250, ..., -0.0453, -0.0734, -0.2692], + [-0.0136, 0.0593, -0.0266, ..., -0.1104, 0.1265, -0.0421], + ..., + [ 0.0282, -0.0340, -0.0190, ..., -0.1350, -0.1795, 0.0306], + [-0.0178, -0.0791, -0.0270, ..., -0.1009, 0.0534, -0.1572], + [-0.2341, -0.0124, -0.0167, ..., -0.0920, -0.1729, -0.0942]], + device='cuda:0'), grad: tensor([[6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 2.2817e-08, 1.7509e-07, + 2.7940e-09], + [9.9186e-08, 0.0000e+00, 0.0000e+00, ..., 9.7323e-08, 2.3236e-07, + 4.9360e-08], + [9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 6.2399e-08, 1.7229e-07, + 4.6566e-09], + ..., + [1.7509e-07, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, 1.3364e-07, + 8.7079e-08], + [1.0710e-08, 0.0000e+00, 0.0000e+00, ..., 2.6356e-07, 2.5313e-06, + 5.5879e-09], + [8.1025e-08, 0.0000e+00, 0.0000e+00, ..., 3.4925e-08, 3.7393e-07, + 4.0047e-08]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0051, -0.0131, 0.0038, 0.0343, 0.0012, 0.0273, 0.0043, 0.0177, + -0.0036, -0.0199], device='cuda:0'), grad: tensor([ 6.5332e-07, 2.8275e-06, 1.0859e-06, 2.7772e-06, -1.1306e-06, + -1.3970e-05, 3.1032e-06, -2.2575e-06, 8.0615e-06, -1.1381e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 214.82, cls_loss 0.0012 cls_loss_mapping 0.0024 cls_loss_causal 0.4712 re_mapping 0.0047 re_causal 0.0139 /// teacc 98.97 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.0716, -0.0217, -0.0970, ..., 0.0310, -0.0960, -0.0095], + [ 0.0375, -0.0818, -0.0251, ..., -0.0454, -0.0736, -0.2695], + [-0.0132, 0.0592, -0.0262, ..., -0.1106, 0.1266, -0.0417], + ..., + [ 0.0280, -0.0338, -0.0193, ..., -0.1353, -0.1796, 0.0305], + [-0.0179, -0.0794, -0.0270, ..., -0.1010, 0.0534, -0.1578], + [-0.2349, -0.0124, -0.0169, ..., -0.0932, -0.1754, -0.0943]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2387e-07, + 7.4506e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 2.7791e-06, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.8836e-07, + 2.3246e-06, -6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + 1.1083e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.3306e-08, + -3.6918e-06, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0664e-07, + 4.6706e-07, 0.0000e+00]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0038, -0.0132, 0.0038, 0.0342, 0.0011, 0.0274, 0.0043, 0.0178, + -0.0038, -0.0203], device='cuda:0'), grad: tensor([-3.4645e-07, 1.2681e-05, 6.5304e-06, 2.0787e-06, 1.0626e-06, + 7.5763e-07, -5.6811e-06, -4.3362e-05, 2.3350e-05, 2.8424e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.69, cls_loss 0.0017 cls_loss_mapping 0.0021 cls_loss_causal 0.5021 re_mapping 0.0044 re_causal 0.0134 /// teacc 98.97 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.0718, -0.0217, -0.0970, ..., 0.0313, -0.0963, -0.0096], + [ 0.0376, -0.0818, -0.0252, ..., -0.0458, -0.0737, -0.2697], + [-0.0134, 0.0592, -0.0262, ..., -0.1108, 0.1267, -0.0407], + ..., + [ 0.0279, -0.0338, -0.0194, ..., -0.1355, -0.1799, 0.0297], + [-0.0177, -0.0794, -0.0270, ..., -0.1010, 0.0534, -0.1578], + [-0.2355, -0.0124, -0.0169, ..., -0.0930, -0.1764, -0.0949]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.6834e-08, + 1.0943e-07, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7008e-08, + 7.6368e-08, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + -1.5879e-07, -8.6147e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.1444e-08, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.5681e-07, + 5.8301e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + 1.0571e-07, 9.2853e-07]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0036, -0.0152, 0.0029, 0.0341, 0.0017, 0.0275, 0.0042, 0.0205, + -0.0041, -0.0211], device='cuda:0'), grad: tensor([ 3.3062e-07, -7.4226e-07, -4.6566e-08, 1.5721e-06, -2.0713e-06, + -1.1455e-06, -2.7288e-06, 1.4305e-06, 1.5832e-06, 1.8179e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 214.73, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4905 re_mapping 0.0041 re_causal 0.0133 /// teacc 98.99 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.0718, -0.0217, -0.0970, ..., 0.0317, -0.0963, -0.0078], + [ 0.0376, -0.0819, -0.0252, ..., -0.0460, -0.0743, -0.2709], + [-0.0103, 0.0609, -0.0262, ..., -0.1106, 0.1274, -0.0401], + ..., + [ 0.0248, -0.0356, -0.0194, ..., -0.1372, -0.1811, 0.0297], + [-0.0177, -0.0796, -0.0270, ..., -0.1011, 0.0533, -0.1581], + [-0.2356, -0.0123, -0.0169, ..., -0.0931, -0.1771, -0.0957]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -1.9558e-08, + 1.3970e-09, 1.8626e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-09, 4.6566e-09], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -7.9162e-09, 9.3132e-10], + ..., + [-4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 5.5879e-09, 1.1176e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 8.8476e-09, + 1.5367e-08, 5.5879e-09], + [ 0.0000e+00, -4.6566e-09, 0.0000e+00, ..., 1.3970e-09, + 1.3970e-09, -6.0070e-08]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0032, -0.0152, 0.0038, 0.0341, 0.0020, 0.0275, 0.0042, 0.0200, + -0.0042, -0.0213], device='cuda:0'), grad: tensor([-7.4506e-09, -1.2226e-05, 4.0121e-06, 4.3167e-07, 5.2853e-07, + 6.8452e-08, 1.8626e-08, 7.6964e-06, 1.9418e-07, -7.3062e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 214.44, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4999 re_mapping 0.0045 re_causal 0.0138 /// teacc 98.94 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.0722, -0.0217, -0.0970, ..., 0.0317, -0.0964, -0.0079], + [ 0.0376, -0.0819, -0.0252, ..., -0.0460, -0.0745, -0.2719], + [-0.0103, 0.0609, -0.0262, ..., -0.1106, 0.1276, -0.0396], + ..., + [ 0.0249, -0.0356, -0.0194, ..., -0.1382, -0.1812, 0.0296], + [-0.0177, -0.0797, -0.0271, ..., -0.1012, 0.0533, -0.1584], + [-0.2364, -0.0122, -0.0170, ..., -0.0928, -0.1779, -0.0957]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 5.6205e-07, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.8429e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.1886e-08, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + -7.1200e-07, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.1176e-08, -6.5193e-09]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0034, -0.0150, 0.0038, 0.0341, 0.0017, 0.0276, 0.0041, 0.0199, + -0.0043, -0.0210], device='cuda:0'), grad: tensor([ 5.8208e-08, 7.3649e-06, 8.2282e-07, 3.0641e-07, 3.7812e-07, + 8.3819e-09, 1.9418e-07, -5.6103e-06, -2.0172e-06, -1.4827e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 214.49, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5072 re_mapping 0.0043 re_causal 0.0136 /// teacc 98.99 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.0724, -0.0217, -0.0970, ..., 0.0313, -0.0969, -0.0080], + [ 0.0374, -0.0819, -0.0252, ..., -0.0465, -0.0750, -0.2720], + [-0.0103, 0.0609, -0.0263, ..., -0.1105, 0.1276, -0.0396], + ..., + [ 0.0249, -0.0356, -0.0194, ..., -0.1402, -0.1813, 0.0296], + [-0.0174, -0.0797, -0.0271, ..., -0.1013, 0.0533, -0.1585], + [-0.2366, -0.0122, -0.0170, ..., -0.0927, -0.1788, -0.0959]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 7.9162e-09, + 1.5367e-08, 0.0000e+00], + [ 0.0000e+00, 7.9162e-09, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-09, 2.3283e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + -7.7300e-08, -9.3132e-10], + ..., + [ 0.0000e+00, 1.0896e-07, 0.0000e+00, ..., 0.0000e+00, + 2.0023e-08, 1.3970e-09], + [ 0.0000e+00, 1.2573e-08, 0.0000e+00, ..., 1.8161e-08, + 2.3749e-08, 1.8626e-09], + [ 0.0000e+00, -1.7090e-07, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 1.1176e-08]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0036, -0.0152, 0.0036, 0.0340, 0.0019, 0.0277, 0.0043, 0.0199, + -0.0042, -0.0209], device='cuda:0'), grad: tensor([ 1.1222e-07, 1.4491e-06, 1.4529e-05, -1.2247e-07, 5.3504e-07, + 1.6345e-07, -3.3062e-08, -1.6257e-05, 2.0284e-06, -2.4065e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 214.41, cls_loss 0.0019 cls_loss_mapping 0.0031 cls_loss_causal 0.5016 re_mapping 0.0046 re_causal 0.0134 /// teacc 98.99 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.0726, -0.0218, -0.0972, ..., 0.0314, -0.0976, -0.0077], + [ 0.0371, -0.0826, -0.0259, ..., -0.0474, -0.0783, -0.2751], + [-0.0104, 0.0612, -0.0293, ..., -0.1091, 0.1279, -0.0398], + ..., + [ 0.0252, -0.0355, -0.0166, ..., -0.1418, -0.1815, 0.0313], + [-0.0175, -0.0798, -0.0275, ..., -0.1015, 0.0563, -0.1556], + [-0.2370, -0.0123, -0.0174, ..., -0.0927, -0.1798, -0.0970]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.2061e-07, + -1.1874e-07, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.9337e-08, + 1.9558e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.1665e-08, + -9.1502e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 9.1689e-07, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.3958e-07, + 3.0082e-07, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.9139e-07, + 9.4064e-08, 3.2596e-09]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0039, -0.0182, 0.0042, 0.0339, 0.0023, 0.0278, 0.0036, 0.0198, + -0.0011, -0.0210], device='cuda:0'), grad: tensor([-1.6941e-06, -3.2317e-07, -1.5125e-06, 1.9092e-08, 9.3132e-09, + -8.4471e-07, 4.6706e-07, 1.7844e-06, 1.4445e-06, 6.5332e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 214.18, cls_loss 0.0017 cls_loss_mapping 0.0028 cls_loss_causal 0.5023 re_mapping 0.0047 re_causal 0.0140 /// teacc 98.99 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.0727, -0.0218, -0.0973, ..., 0.0298, -0.1003, -0.0073], + [ 0.0372, -0.0828, -0.0261, ..., -0.0496, -0.0784, -0.2754], + [-0.0103, 0.0620, -0.0294, ..., -0.1095, 0.1281, -0.0396], + ..., + [ 0.0251, -0.0362, -0.0165, ..., -0.1422, -0.1820, 0.0308], + [-0.0180, -0.0802, -0.0275, ..., -0.1017, 0.0563, -0.1556], + [-0.2372, -0.0122, -0.0177, ..., -0.0929, -0.1810, -0.0964]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 1.3970e-09, ..., -5.5879e-09, + 6.1002e-08, 0.0000e+00], + [-1.8999e-07, 0.0000e+00, 3.7253e-09, ..., 1.3970e-09, + 1.1642e-08, 0.0000e+00], + [ 2.0489e-08, 0.0000e+00, 1.2806e-07, ..., 8.8476e-09, + 7.0781e-08, -4.1910e-09], + ..., + [ 1.8626e-08, 0.0000e+00, 2.4680e-08, ..., 4.6566e-10, + 1.3970e-09, 0.0000e+00], + [ 6.7055e-08, 0.0000e+00, 2.9756e-07, ..., 2.6543e-08, + -3.1479e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + 2.5146e-08, -1.3970e-09]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0066, -0.0183, 0.0046, 0.0339, 0.0025, 0.0281, 0.0041, 0.0192, + -0.0011, -0.0195], device='cuda:0'), grad: tensor([ 2.1141e-07, -3.8091e-06, 1.3076e-06, -1.5507e-06, 1.4938e-06, + 3.6834e-07, 5.3272e-07, -9.6858e-07, 2.2072e-06, 1.9232e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 214.23, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4952 re_mapping 0.0043 re_causal 0.0132 /// teacc 99.04 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.0728, -0.0218, -0.0974, ..., 0.0304, -0.1000, -0.0064], + [ 0.0377, -0.0828, -0.0261, ..., -0.0507, -0.0784, -0.2754], + [-0.0103, 0.0620, -0.0294, ..., -0.1122, 0.1272, -0.0394], + ..., + [ 0.0251, -0.0362, -0.0165, ..., -0.1425, -0.1821, 0.0308], + [-0.0181, -0.0802, -0.0279, ..., -0.1019, 0.0563, -0.1557], + [-0.2377, -0.0122, -0.0178, ..., -0.0935, -0.1814, -0.0966]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6345e-07, + 4.4703e-08, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 1.7416e-07, 4.6566e-10], + [-1.0710e-08, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + -1.6093e-06, 0.0000e+00], + ..., + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 8.0327e-07, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.2375e-08, + 1.3271e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + 3.7253e-08, -8.3819e-09]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0057, -0.0183, 0.0037, 0.0339, 0.0018, 0.0280, 0.0061, 0.0191, + -0.0011, -0.0196], device='cuda:0'), grad: tensor([ 1.5367e-08, -6.5342e-06, -2.9914e-06, 6.5472e-07, 1.1623e-06, + 5.9651e-07, 5.8673e-07, -1.0468e-06, 3.1460e-06, 4.4145e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 214.32, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4795 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.03 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.0728, -0.0221, -0.0975, ..., 0.0308, -0.1002, -0.0065], + [ 0.0378, -0.0828, -0.0262, ..., -0.0508, -0.0784, -0.2757], + [-0.0103, 0.0639, -0.0295, ..., -0.1123, 0.1274, -0.0384], + ..., + [ 0.0251, -0.0375, -0.0166, ..., -0.1432, -0.1824, 0.0305], + [-0.0182, -0.0812, -0.0281, ..., -0.1020, 0.0563, -0.1558], + [-0.2379, -0.0129, -0.0178, ..., -0.0939, -0.1824, -0.0966]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 8.3819e-09, + 4.4703e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 3.7253e-09, + 1.3830e-07, 7.9162e-09], + [-3.7253e-09, -7.9162e-09, 8.9873e-08, ..., 6.0536e-09, + -5.8021e-07, 7.9162e-09], + ..., + [ 2.7940e-09, 5.5879e-09, 2.6077e-08, ..., 0.0000e+00, + 1.1735e-07, -2.3283e-09], + [ 0.0000e+00, 4.6566e-10, -3.1060e-07, ..., 3.4459e-08, + -3.0594e-07, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.3970e-09, + 1.8161e-08, -4.6566e-09]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0055, -0.0180, 0.0035, 0.0349, 0.0011, 0.0271, 0.0061, 0.0188, + -0.0011, -0.0197], device='cuda:0'), grad: tensor([ 1.3039e-07, -3.5390e-08, -8.6147e-07, 2.2128e-06, 5.0291e-08, + -1.0505e-06, -1.3039e-08, 2.3982e-07, -6.5658e-07, -2.7008e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 214.47, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.5109 re_mapping 0.0044 re_causal 0.0137 /// teacc 99.07 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.0728, -0.0221, -0.0975, ..., 0.0309, -0.1004, -0.0067], + [ 0.0382, -0.0829, -0.0263, ..., -0.0504, -0.0784, -0.2758], + [-0.0103, 0.0641, -0.0296, ..., -0.1122, 0.1277, -0.0384], + ..., + [ 0.0250, -0.0376, -0.0166, ..., -0.1436, -0.1828, 0.0303], + [-0.0183, -0.0814, -0.0280, ..., -0.1023, 0.0563, -0.1559], + [-0.2380, -0.0129, -0.0179, ..., -0.0939, -0.1833, -0.0972]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-08, + 8.8476e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + 1.5367e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + -4.9360e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 5.3085e-08, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1828e-07, + 1.9465e-07, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0617e-07, + 2.3283e-09, 1.8626e-08]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0053, -0.0177, 0.0036, 0.0350, 0.0008, 0.0271, 0.0061, 0.0184, + -0.0012, -0.0204], device='cuda:0'), grad: tensor([ 9.3412e-07, -8.2999e-06, 6.1840e-07, -1.6600e-05, 1.8114e-07, + 2.4345e-06, -5.8860e-07, -3.7579e-07, 1.0796e-05, 1.0923e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 214.41, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4939 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.0728, -0.0222, -0.0975, ..., 0.0313, -0.1005, -0.0052], + [ 0.0384, -0.0832, -0.0263, ..., -0.0505, -0.0784, -0.2759], + [-0.0103, 0.0647, -0.0296, ..., -0.1122, 0.1279, -0.0383], + ..., + [ 0.0250, -0.0381, -0.0166, ..., -0.1438, -0.1830, 0.0316], + [-0.0183, -0.0816, -0.0280, ..., -0.1024, 0.0563, -0.1561], + [-0.2380, -0.0129, -0.0179, ..., -0.0943, -0.1844, -0.0965]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + 1.3504e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 4.7032e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -1.2033e-06, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.0403e-06, 3.8650e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -1.4063e-07, 3.0734e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.0710e-08, -8.7079e-08]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0048, -0.0177, 0.0037, 0.0350, -0.0008, 0.0272, 0.0064, 0.0183, + -0.0013, -0.0201], device='cuda:0'), grad: tensor([ 5.8208e-08, -3.6554e-07, -3.0845e-06, 1.5693e-07, 7.7533e-07, + 2.8871e-08, -2.3143e-07, 2.9560e-06, 2.7940e-09, -2.7288e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 214.26, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4914 re_mapping 0.0045 re_causal 0.0138 /// teacc 99.00 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.0728, -0.0228, -0.0976, ..., 0.0318, -0.1008, -0.0053], + [ 0.0383, -0.0848, -0.0264, ..., -0.0510, -0.0784, -0.2759], + [-0.0103, 0.0655, -0.0296, ..., -0.1122, 0.1280, -0.0380], + ..., + [ 0.0251, -0.0379, -0.0166, ..., -0.1439, -0.1830, 0.0328], + [-0.0182, -0.0829, -0.0282, ..., -0.1025, 0.0562, -0.1561], + [-0.2381, -0.0130, -0.0179, ..., -0.0948, -0.1829, -0.0967]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.8476e-09, + 1.0757e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3656e-07, + -2.8741e-06, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -1.4240e-06, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 9.0804e-07, 2.3749e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3516e-07, + 3.0864e-06, 2.3283e-09], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 5.1223e-09, + 3.9581e-08, 3.7719e-08]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0046, -0.0176, 0.0036, 0.0349, -0.0020, 0.0272, 0.0068, 0.0186, + -0.0014, -0.0200], device='cuda:0'), grad: tensor([ 6.1467e-07, -2.7135e-05, -3.6582e-06, 5.9279e-07, -1.0431e-07, + -3.3528e-08, 1.3830e-07, 2.8424e-06, 2.6792e-05, -7.1712e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 214.39, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4607 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.08 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.0728, -0.0229, -0.0976, ..., 0.0320, -0.1010, -0.0063], + [ 0.0383, -0.0850, -0.0264, ..., -0.0514, -0.0784, -0.2761], + [-0.0102, 0.0655, -0.0296, ..., -0.1122, 0.1281, -0.0381], + ..., + [ 0.0251, -0.0381, -0.0166, ..., -0.1442, -0.1831, 0.0315], + [-0.0183, -0.0823, -0.0282, ..., -0.1026, 0.0563, -0.1562], + [-0.2381, -0.0126, -0.0179, ..., -0.0947, -0.1834, -0.0937]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.5879e-09, + 5.0291e-08, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 3.3062e-08, 3.2596e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -6.5705e-07, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2480e-07, -4.4703e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5832e-08, + 3.3714e-07, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 5.1223e-09, 1.6717e-07]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0041, -0.0176, 0.0035, 0.0349, -0.0043, 0.0272, 0.0065, 0.0186, + -0.0014, -0.0175], device='cuda:0'), grad: tensor([ 1.2806e-07, 1.9353e-06, -4.7125e-07, 3.2410e-07, -1.6391e-07, + 1.7136e-07, 5.0291e-08, -3.8967e-06, 8.3493e-07, 1.0785e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 214.54, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4899 re_mapping 0.0041 re_causal 0.0131 /// teacc 99.05 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.0728, -0.0229, -0.0976, ..., 0.0319, -0.1021, -0.0077], + [ 0.0387, -0.0850, -0.0264, ..., -0.0515, -0.0785, -0.2763], + [-0.0102, 0.0655, -0.0296, ..., -0.1123, 0.1284, -0.0383], + ..., + [ 0.0250, -0.0381, -0.0167, ..., -0.1444, -0.1833, 0.0318], + [-0.0184, -0.0822, -0.0283, ..., -0.1027, 0.0562, -0.1563], + [-0.2383, -0.0125, -0.0179, ..., -0.0947, -0.1843, -0.0934]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-09, + 5.4017e-08, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 1.4901e-08, 2.6543e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 5.5879e-09, 6.9849e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 7.4506e-09, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.5157e-07, + 5.5274e-07, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2573e-07, + 2.1374e-07, 1.3383e-06]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0047, -0.0177, 0.0033, 0.0349, -0.0046, 0.0273, 0.0065, 0.0190, + -0.0015, -0.0172], device='cuda:0'), grad: tensor([ 2.9709e-07, -1.3039e-05, 6.8499e-07, 7.9256e-07, -4.2617e-05, + -5.2303e-06, 2.9616e-06, 8.4192e-06, 2.1048e-06, 4.5717e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 214.47, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4920 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.08 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.0728, -0.0229, -0.0976, ..., 0.0321, -0.1023, -0.0076], + [ 0.0387, -0.0851, -0.0264, ..., -0.0516, -0.0785, -0.2763], + [-0.0102, 0.0655, -0.0296, ..., -0.1123, 0.1284, -0.0382], + ..., + [ 0.0250, -0.0381, -0.0167, ..., -0.1460, -0.1837, 0.0318], + [-0.0184, -0.0822, -0.0282, ..., -0.1024, 0.0565, -0.1564], + [-0.2384, -0.0125, -0.0179, ..., -0.0949, -0.1868, -0.0935]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + 3.7532e-07, 1.7229e-08], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.6787e-08, + 7.6834e-07, 1.0757e-07], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -9.7789e-08, + -2.1774e-06, 4.6566e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-08, + 1.9092e-07, 2.3749e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.5635e-08, + -1.2526e-07, 5.2620e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.7765e-08, + 9.7789e-08, 9.7044e-07]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0046, -0.0175, 0.0031, 0.0348, -0.0045, 0.0272, 0.0066, 0.0188, + -0.0013, -0.0176], device='cuda:0'), grad: tensor([ 6.6124e-07, 2.0117e-06, -3.4403e-06, 9.2015e-07, -9.0525e-06, + 3.1292e-07, 6.8080e-07, 1.4994e-06, -1.0198e-07, 6.4820e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 214.41, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4759 re_mapping 0.0043 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.0729, -0.0229, -0.0976, ..., 0.0323, -0.1027, -0.0077], + [ 0.0385, -0.0851, -0.0264, ..., -0.0525, -0.0785, -0.2764], + [-0.0102, 0.0656, -0.0296, ..., -0.1123, 0.1286, -0.0381], + ..., + [ 0.0250, -0.0381, -0.0167, ..., -0.1464, -0.1839, 0.0319], + [-0.0186, -0.0823, -0.0283, ..., -0.1027, 0.0564, -0.1564], + [-0.2384, -0.0126, -0.0179, ..., -0.0953, -0.1883, -0.0935]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0023e-08, + 2.4214e-08, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 2.7940e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + -3.7253e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3528e-08, + 1.4435e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 2.7940e-09, 8.8941e-08]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0047, -0.0175, 0.0031, 0.0351, -0.0046, 0.0270, 0.0067, 0.0189, + -0.0013, -0.0178], device='cuda:0'), grad: tensor([ 1.2806e-07, 1.4994e-07, 2.6729e-07, -7.1619e-07, -9.8068e-07, + 4.8429e-08, 2.0023e-08, 1.1176e-08, 1.7602e-07, 8.9314e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.30, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.5034 re_mapping 0.0040 re_causal 0.0130 /// teacc 99.00 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.0729, -0.0230, -0.0977, ..., 0.0321, -0.1034, -0.0079], + [ 0.0387, -0.0852, -0.0264, ..., -0.0535, -0.0785, -0.2770], + [-0.0102, 0.0656, -0.0296, ..., -0.1124, 0.1287, -0.0380], + ..., + [ 0.0250, -0.0381, -0.0167, ..., -0.1466, -0.1839, 0.0320], + [-0.0187, -0.0823, -0.0283, ..., -0.1031, 0.0564, -0.1565], + [-0.2385, -0.0124, -0.0180, ..., -0.0954, -0.1893, -0.0962]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0245e-08, + 3.4925e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 3.7253e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -1.5367e-08, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.4238e-09, 3.9581e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3271e-08, + -2.6310e-08, 1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 3.0268e-09, 1.5064e-07]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0052, -0.0176, 0.0030, 0.0352, -0.0020, 0.0270, 0.0070, 0.0190, + -0.0014, -0.0203], device='cuda:0'), grad: tensor([ 3.2131e-08, -2.4572e-05, 1.8394e-07, 1.3225e-07, 1.5181e-07, + 2.3097e-07, 6.2631e-08, 2.1622e-05, 7.2364e-07, 1.4603e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 214.56, cls_loss 0.0015 cls_loss_mapping 0.0035 cls_loss_causal 0.5160 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.08 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.0729, -0.0230, -0.0977, ..., 0.0291, -0.1057, -0.0081], + [ 0.0389, -0.0856, -0.0264, ..., -0.0535, -0.0785, -0.2772], + [-0.0102, 0.0659, -0.0296, ..., -0.1125, 0.1289, -0.0383], + ..., + [ 0.0250, -0.0379, -0.0167, ..., -0.1472, -0.1843, 0.0317], + [-0.0189, -0.0839, -0.0283, ..., -0.1032, 0.0564, -0.1567], + [-0.2386, -0.0113, -0.0180, ..., -0.0931, -0.1897, -0.0952]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.2340e-08, + 3.3993e-08, 9.7789e-09], + [-2.3283e-10, 3.0268e-09, 0.0000e+00, ..., 6.5193e-09, + 1.7695e-08, 1.2200e-07], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 7.4506e-09, + 1.8789e-07, 3.3528e-08], + ..., + [ 0.0000e+00, -1.1874e-08, 0.0000e+00, ..., 2.3283e-10, + 3.6089e-08, -7.9721e-07], + [ 0.0000e+00, 3.4925e-09, 0.0000e+00, ..., 1.0454e-07, + -1.8231e-07, 2.4913e-08], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 1.1642e-09, + 5.6112e-08, 4.3656e-07]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0072, -0.0173, 0.0030, 0.0354, -0.0034, 0.0267, 0.0071, 0.0189, + -0.0015, -0.0188], device='cuda:0'), grad: tensor([ 1.6857e-07, 1.5041e-06, 9.6671e-07, 9.7603e-07, 1.6252e-06, + 5.2620e-08, -2.1793e-07, -8.6203e-06, -9.7323e-08, 3.6359e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 214.77, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4683 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.11 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.0730, -0.0246, -0.0977, ..., 0.0279, -0.1063, -0.0083], + [ 0.0402, -0.0886, -0.0265, ..., -0.0538, -0.0785, -0.2773], + [-0.0102, 0.0667, -0.0296, ..., -0.1124, 0.1293, -0.0391], + ..., + [ 0.0249, -0.0364, -0.0167, ..., -0.1472, -0.1845, 0.0319], + [-0.0193, -0.0861, -0.0284, ..., -0.1033, 0.0563, -0.1567], + [-0.2396, -0.0115, -0.0180, ..., -0.0921, -0.1908, -0.0952]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 0.0000e+00, 0.0000e+00, ..., 1.9092e-08, + 4.9360e-08, 1.6298e-09], + [-7.6788e-07, 1.1642e-09, 0.0000e+00, ..., 6.2864e-09, + 6.8685e-08, 3.4692e-08], + [ 4.1211e-08, -2.3283e-09, 0.0000e+00, ..., -1.6089e-07, + -2.2002e-07, 2.0023e-08], + ..., + [ 2.6310e-08, 9.3132e-10, 0.0000e+00, ..., 8.1724e-08, + 1.4692e-07, -8.4052e-08], + [ 2.0489e-08, 0.0000e+00, 0.0000e+00, ..., 4.6100e-08, + -1.5497e-06, 4.4238e-09], + [ 1.2573e-08, 0.0000e+00, 0.0000e+00, ..., 1.2573e-08, + 3.4226e-08, 1.5832e-08]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0088, -0.0170, 0.0033, 0.0354, -0.0035, 0.0270, 0.0072, 0.0182, + -0.0016, -0.0186], device='cuda:0'), grad: tensor([ 2.5122e-07, -4.1835e-06, -1.6787e-07, 2.0470e-06, 1.6596e-06, + 3.6787e-06, 1.6987e-06, -2.6054e-07, -5.1185e-06, 3.7951e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 214.76, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.5048 re_mapping 0.0044 re_causal 0.0132 /// teacc 99.06 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.0731, -0.0248, -0.0977, ..., 0.0290, -0.1082, -0.0084], + [ 0.0427, -0.0893, -0.0265, ..., -0.0539, -0.0786, -0.2774], + [-0.0103, 0.0670, -0.0296, ..., -0.1117, 0.1296, -0.0389], + ..., + [ 0.0248, -0.0361, -0.0167, ..., -0.1508, -0.1848, 0.0316], + [-0.0204, -0.0870, -0.0284, ..., -0.1039, 0.0575, -0.1568], + [-0.2416, -0.0117, -0.0180, ..., -0.0927, -0.1916, -0.0952]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.8173e-08, + 7.4506e-09, 4.6566e-10], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 1.6065e-08, + 1.9325e-08, 7.4506e-09], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., -2.1886e-08, + -5.0524e-08, 5.5879e-09], + ..., + [ 0.0000e+00, -1.0245e-08, 0.0000e+00, ..., 6.7521e-09, + 1.5367e-08, 7.2177e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.1642e-08, + 6.3330e-08, 2.0955e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 9.0804e-09, + 3.7253e-09, 8.3819e-09]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0108, -0.0166, 0.0034, 0.0351, -0.0037, 0.0265, 0.0085, 0.0177, + -0.0011, -0.0186], device='cuda:0'), grad: tensor([-5.7509e-08, 1.0873e-07, 5.7742e-08, 9.9186e-08, -8.4192e-07, + -2.0047e-07, 7.6601e-08, -2.2422e-07, 2.6356e-07, 7.2038e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 214.56, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.4663 re_mapping 0.0044 re_causal 0.0134 /// teacc 98.94 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.0731, -0.0248, -0.0977, ..., 0.0294, -0.1087, -0.0088], + [ 0.0428, -0.0894, -0.0265, ..., -0.0546, -0.0786, -0.2775], + [-0.0103, 0.0671, -0.0296, ..., -0.1117, 0.1301, -0.0389], + ..., + [ 0.0248, -0.0360, -0.0167, ..., -0.1509, -0.1852, 0.0315], + [-0.0204, -0.0871, -0.0284, ..., -0.1040, 0.0574, -0.1569], + [-0.2417, -0.0117, -0.0180, ..., -0.0927, -0.1919, -0.0953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 4.3772e-08, + 7.2876e-08, 4.4238e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 6.1933e-08, + 6.5891e-08, 9.8906e-07], + [ 0.0000e+00, -1.7928e-08, 0.0000e+00, ..., 1.2806e-08, + -1.5367e-08, 2.4680e-08], + ..., + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., -4.8894e-09, + 3.4226e-08, 2.5006e-07], + [ 0.0000e+00, 8.8476e-09, 0.0000e+00, ..., 6.7241e-06, + 5.3979e-06, 1.9092e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, + 5.8673e-08, -1.4435e-06]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0115, -0.0138, 0.0036, 0.0347, -0.0038, 0.0270, 0.0083, 0.0147, + -0.0012, -0.0185], device='cuda:0'), grad: tensor([ 2.1025e-07, 8.2850e-06, 4.1747e-07, 4.7777e-07, 1.2927e-06, + -1.0245e-07, -1.9372e-05, 1.4603e-06, 1.8716e-05, -1.1377e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 214.52, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4889 re_mapping 0.0041 re_causal 0.0129 /// teacc 99.00 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.0731, -0.0249, -0.0977, ..., 0.0292, -0.1089, -0.0090], + [ 0.0428, -0.0926, -0.0265, ..., -0.0544, -0.0786, -0.2776], + [-0.0102, 0.0666, -0.0296, ..., -0.1119, 0.1307, -0.0388], + ..., + [ 0.0248, -0.0329, -0.0167, ..., -0.1510, -0.1853, 0.0320], + [-0.0205, -0.0874, -0.0284, ..., -0.1033, 0.0575, -0.1571], + [-0.2418, -0.0147, -0.0180, ..., -0.0927, -0.1934, -0.0953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.0559e-08, + -3.2596e-09, 1.8626e-09], + [ 0.0000e+00, 3.2596e-09, 0.0000e+00, ..., 3.7253e-09, + 1.0431e-07, 3.3993e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.5832e-08, + 7.1013e-07, 4.6566e-10], + ..., + [ 0.0000e+00, -6.5193e-09, 0.0000e+00, ..., 2.3283e-09, + -1.0300e-06, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.8894e-08, + 1.5786e-07, 9.3132e-10], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 6.0536e-09, + 4.8894e-08, -4.8894e-08]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0117, -0.0138, 0.0038, 0.0346, -0.0038, 0.0272, 0.0074, 0.0149, + -0.0011, -0.0189], device='cuda:0'), grad: tensor([ 1.4100e-06, 8.5652e-05, 1.0973e-04, 1.3091e-05, 4.8056e-07, + 8.5589e-07, -1.4668e-07, -2.3675e-04, 5.8115e-06, 1.9968e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.19, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.5017 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.06 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.0731, -0.0251, -0.0977, ..., 0.0291, -0.1093, -0.0093], + [ 0.0430, -0.0930, -0.0265, ..., -0.0545, -0.0787, -0.2778], + [-0.0102, 0.0686, -0.0297, ..., -0.1120, 0.1309, -0.0420], + ..., + [ 0.0248, -0.0334, -0.0168, ..., -0.1512, -0.1855, 0.0318], + [-0.0203, -0.0901, -0.0285, ..., -0.1035, 0.0574, -0.1574], + [-0.2419, -0.0147, -0.0180, ..., -0.0929, -0.1956, -0.0953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 1.1520e-06, + 1.2051e-06, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.6764e-08, + 3.7253e-08, 3.2596e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 1.2107e-08, + -1.1781e-07, 4.6566e-10], + ..., + [ 0.0000e+00, -2.2817e-08, 0.0000e+00, ..., 1.3970e-09, + 6.7055e-08, 9.3132e-10], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 6.0815e-07, + 7.9582e-07, 9.3132e-10], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 7.6834e-08, + 8.1956e-08, 1.7276e-07]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0118, -0.0141, 0.0035, 0.0345, -0.0037, 0.0275, 0.0076, 0.0153, + -0.0012, -0.0196], device='cuda:0'), grad: tensor([ 2.8089e-06, -9.7789e-09, -7.7300e-08, -1.1548e-07, -5.6485e-07, + 1.6857e-06, -6.1169e-06, -3.0734e-08, 1.6056e-06, 8.1398e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 214.53, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4621 re_mapping 0.0041 re_causal 0.0129 /// teacc 99.05 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.0731, -0.0252, -0.0977, ..., 0.0291, -0.1097, -0.0096], + [ 0.0430, -0.0934, -0.0265, ..., -0.0545, -0.0787, -0.2779], + [-0.0102, 0.0694, -0.0297, ..., -0.1120, 0.1310, -0.0424], + ..., + [ 0.0248, -0.0334, -0.0168, ..., -0.1513, -0.1858, 0.0320], + [-0.0202, -0.0910, -0.0285, ..., -0.1033, 0.0575, -0.1577], + [-0.2419, -0.0146, -0.0181, ..., -0.0932, -0.1966, -0.0953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3993e-08, + 1.4296e-07, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.3993e-08, + -2.5049e-05, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.7789e-09, + 3.5077e-05, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.0850e-06, 7.9628e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.9858e-06, + -2.0057e-05, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.3597e-06, 7.9628e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0119, -0.0141, 0.0034, 0.0344, -0.0038, 0.0276, 0.0075, 0.0153, + -0.0011, -0.0196], device='cuda:0'), grad: tensor([ 1.3504e-07, -1.4710e-04, 1.6761e-04, 7.2978e-06, -6.8452e-08, + 7.1675e-06, 2.2575e-06, 4.7944e-06, -4.5329e-05, 3.4422e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 214.43, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4958 re_mapping 0.0039 re_causal 0.0131 /// teacc 99.10 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.0731, -0.0252, -0.0977, ..., 0.0289, -0.1102, -0.0091], + [ 0.0434, -0.0935, -0.0266, ..., -0.0547, -0.0787, -0.2780], + [-0.0104, 0.0699, -0.0297, ..., -0.1122, 0.1312, -0.0427], + ..., + [ 0.0248, -0.0335, -0.0168, ..., -0.1519, -0.1862, 0.0351], + [-0.0202, -0.0912, -0.0285, ..., -0.1038, 0.0576, -0.1579], + [-0.2420, -0.0146, -0.0181, ..., -0.0931, -0.1979, -0.0954]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.8476e-09, + 5.8208e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6764e-08, + 3.5390e-08, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -2.9244e-07, -2.9337e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6764e-08, + 2.8871e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.2667e-08, + 2.6729e-07, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-09, 2.1886e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0122, -0.0141, 0.0033, 0.0344, -0.0060, 0.0275, 0.0078, 0.0158, + -0.0011, -0.0197], device='cuda:0'), grad: tensor([ 1.2480e-07, 5.9092e-07, -4.5029e-07, 1.0896e-07, -2.7753e-07, + 1.0896e-07, -2.2305e-07, -5.9325e-07, 5.8115e-07, 3.9116e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 214.43, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.5114 re_mapping 0.0040 re_causal 0.0132 /// teacc 99.04 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.0731, -0.0253, -0.0978, ..., 0.0289, -0.1103, -0.0096], + [ 0.0434, -0.0935, -0.0266, ..., -0.0547, -0.0787, -0.2781], + [-0.0104, 0.0700, -0.0298, ..., -0.1122, 0.1313, -0.0442], + ..., + [ 0.0249, -0.0335, -0.0168, ..., -0.1522, -0.1864, 0.0351], + [-0.0202, -0.0914, -0.0285, ..., -0.1039, 0.0576, -0.1579], + [-0.2422, -0.0146, -0.0181, ..., -0.0931, -0.1982, -0.0954]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-08, + 1.9511e-07, 3.1712e-07], + [-1.5367e-08, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 3.5856e-08, 3.7719e-08], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.4064e-08, + 2.5705e-07, 2.1607e-07], + ..., + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 8.8476e-09, 6.0536e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 5.8208e-08, + 1.1036e-07, 9.0804e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 6.0536e-09, + 4.3772e-08, 3.6787e-08]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0124, -0.0143, 0.0031, 0.0344, -0.0059, 0.0276, 0.0079, 0.0159, + -0.0011, -0.0193], device='cuda:0'), grad: tensor([ 1.0906e-06, 1.3672e-06, 9.9987e-06, -1.0766e-05, 3.5930e-06, + 1.6242e-06, -7.3835e-06, -1.4063e-06, 1.4342e-06, 4.6100e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 214.37, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.4933 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.06 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.0732, -0.0254, -0.0978, ..., 0.0287, -0.1106, -0.0104], + [ 0.0433, -0.0935, -0.0266, ..., -0.0548, -0.0787, -0.2782], + [-0.0106, 0.0703, -0.0298, ..., -0.1122, 0.1319, -0.0425], + ..., + [ 0.0249, -0.0335, -0.0168, ..., -0.1526, -0.1866, 0.0354], + [-0.0203, -0.0920, -0.0285, ..., -0.1043, 0.0575, -0.1580], + [-0.2424, -0.0146, -0.0181, ..., -0.0925, -0.1986, -0.0955]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, + 1.0245e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 7.4506e-09, 4.6566e-10], + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., 9.3132e-10, + -8.3819e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.1269e-07, 2.7940e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 5.5879e-09, + -8.8476e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 6.0536e-09, 4.6566e-10]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0131, -0.0143, 0.0032, 0.0341, -0.0060, 0.0281, 0.0078, 0.0160, + -0.0013, -0.0191], device='cuda:0'), grad: tensor([ 4.6100e-08, -1.8582e-05, 9.2201e-08, 7.4096e-06, 1.8766e-07, + 3.0361e-07, 5.5879e-09, 1.7881e-05, 3.5157e-07, -7.6517e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 214.55, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4684 re_mapping 0.0043 re_causal 0.0122 /// teacc 99.11 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.0733, -0.0257, -0.0978, ..., 0.0294, -0.1107, -0.0106], + [ 0.0461, -0.0936, -0.0266, ..., -0.0553, -0.0788, -0.2787], + [-0.0106, 0.0704, -0.0298, ..., -0.1124, 0.1325, -0.0397], + ..., + [ 0.0248, -0.0335, -0.0168, ..., -0.1529, -0.1868, 0.0361], + [-0.0232, -0.0902, -0.0285, ..., -0.1047, 0.0575, -0.1601], + [-0.2427, -0.0150, -0.0181, ..., -0.0923, -0.1996, -0.0955]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + 4.8894e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-08, + 2.7008e-08, 1.5413e-07], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., 3.2596e-08, + -1.2899e-07, -7.4040e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 1.2200e-07, 5.4529e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1130e-07, + 4.5262e-07, 9.7789e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-08, + 9.7789e-09, -7.4739e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0129, -0.0143, 0.0035, 0.0339, -0.0061, 0.0282, 0.0075, 0.0161, + -0.0015, -0.0192], device='cuda:0'), grad: tensor([ 4.4703e-08, 1.0449e-06, -3.2596e-09, 2.5192e-07, 1.3560e-06, + 3.7393e-07, -1.4696e-06, 6.2063e-06, 1.1120e-06, -8.9407e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 214.48, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4863 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.00 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.0734, -0.0265, -0.0978, ..., 0.0292, -0.1110, -0.0107], + [ 0.0482, -0.0950, -0.0266, ..., -0.0554, -0.0788, -0.2790], + [-0.0113, 0.0708, -0.0298, ..., -0.1124, 0.1327, -0.0403], + ..., + [ 0.0253, -0.0325, -0.0168, ..., -0.1541, -0.1870, 0.0361], + [-0.0252, -0.0902, -0.0285, ..., -0.1054, 0.0573, -0.1602], + [-0.2430, -0.0152, -0.0181, ..., -0.0920, -0.2003, -0.0953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + 1.5367e-08, 9.3132e-10], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 4.8894e-08, + 2.6962e-07, 1.5832e-08], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 2.7474e-08, + 1.3737e-07, 6.9849e-09], + ..., + [ 1.3970e-09, -4.6566e-10, 0.0000e+00, ..., 1.0710e-08, + 6.4727e-08, -8.6147e-08], + [-1.6764e-08, 0.0000e+00, 0.0000e+00, ..., -9.4995e-08, + -3.4971e-07, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 2.7474e-08, 3.0268e-08]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0133, -0.0143, 0.0034, 0.0338, -0.0070, 0.0284, 0.0076, 0.0162, + -0.0018, -0.0177], device='cuda:0'), grad: tensor([-1.0245e-08, 9.7882e-07, 5.2340e-07, 7.8464e-07, 3.6322e-08, + -1.2787e-06, 4.5775e-07, -6.9151e-07, -1.1977e-06, 4.0000e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.54, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4849 re_mapping 0.0041 re_causal 0.0126 /// teacc 99.08 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.0734, -0.0268, -0.0978, ..., 0.0288, -0.1114, -0.0107], + [ 0.0490, -0.0974, -0.0266, ..., -0.0555, -0.0788, -0.2792], + [-0.0114, 0.0705, -0.0298, ..., -0.1126, 0.1318, -0.0403], + ..., + [ 0.0252, -0.0305, -0.0168, ..., -0.1547, -0.1875, 0.0362], + [-0.0258, -0.0895, -0.0285, ..., -0.1056, 0.0577, -0.1604], + [-0.2431, -0.0159, -0.0181, ..., -0.0925, -0.2018, -0.0953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 6.9849e-09, 4.1910e-09], + [-1.3970e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-08, 2.8405e-08], + [ 0.0000e+00, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + -6.1933e-08, 6.5193e-09], + ..., + [ 1.3970e-09, 4.1910e-09, 0.0000e+00, ..., 4.6566e-10, + 6.5193e-08, 2.2352e-08], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 9.3132e-09, + 2.5472e-07, 4.1910e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 1.1781e-07, 1.3821e-06]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0138, -0.0143, 0.0018, 0.0340, -0.0070, 0.0282, 0.0082, 0.0163, + -0.0015, -0.0178], device='cuda:0'), grad: tensor([ 3.1665e-08, 1.9325e-07, 5.5879e-09, 4.3679e-07, -4.7386e-06, + -1.6037e-06, 1.8440e-07, 7.8231e-08, 6.3889e-07, 4.7721e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 214.52, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4610 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.11 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.0735, -0.0271, -0.0978, ..., 0.0276, -0.1130, -0.0109], + [ 0.0491, -0.0975, -0.0267, ..., -0.0552, -0.0797, -0.2818], + [-0.0114, 0.0706, -0.0298, ..., -0.1126, 0.1341, -0.0377], + ..., + [ 0.0252, -0.0301, -0.0168, ..., -0.1558, -0.1880, 0.0362], + [-0.0258, -0.0892, -0.0285, ..., -0.1058, 0.0577, -0.1605], + [-0.2442, -0.0167, -0.0181, ..., -0.0926, -0.2026, -0.0954]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.3504e-08, + 2.0117e-07, 4.6566e-10], + [-2.1886e-08, 4.5169e-08, 0.0000e+00, ..., -1.9278e-07, + 2.8592e-07, 1.5832e-08], + [ 4.6566e-09, 4.6566e-10, 0.0000e+00, ..., 6.0536e-09, + -5.0180e-06, 3.2596e-09], + ..., + [ 1.0710e-08, -6.0070e-08, 0.0000e+00, ..., 1.4901e-08, + 2.5779e-06, 4.1910e-09], + [ 6.0536e-09, 2.7940e-09, 0.0000e+00, ..., 1.0943e-07, + 9.4622e-07, 2.3283e-09], + [ 2.7940e-09, 7.9162e-09, 0.0000e+00, ..., 4.1910e-09, + 1.1129e-07, 5.0291e-08]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0145, -0.0147, 0.0045, 0.0338, -0.0071, 0.0284, 0.0085, 0.0162, + -0.0016, -0.0179], device='cuda:0'), grad: tensor([ 8.7731e-07, -3.0994e-05, -1.4082e-05, -7.2494e-06, 1.7527e-06, + 5.8301e-06, 8.8429e-07, 2.2486e-05, 6.5751e-06, 1.3947e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 214.45, cls_loss 0.0016 cls_loss_mapping 0.0020 cls_loss_causal 0.4938 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.08 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.0736, -0.0272, -0.0978, ..., 0.0304, -0.1132, -0.0077], + [ 0.0489, -0.0976, -0.0267, ..., -0.0541, -0.0797, -0.2820], + [-0.0112, 0.0706, -0.0298, ..., -0.1125, 0.1349, -0.0381], + ..., + [ 0.0253, -0.0296, -0.0168, ..., -0.1564, -0.1886, 0.0362], + [-0.0261, -0.0917, -0.0285, ..., -0.1064, 0.0575, -0.1606], + [-0.2452, -0.0168, -0.0181, ..., -0.0955, -0.2061, -0.0956]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -1.3458e-07, + 1.9092e-08, -6.4708e-06], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 1.4547e-06, 1.2107e-08], + [-4.3772e-08, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + -1.8002e-06, 9.2713e-07], + ..., + [ 2.8405e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 2.0536e-07, 8.3819e-09], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 1.5041e-07, + 1.5227e-07, 2.9663e-07], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 5.5879e-09, 4.3819e-07]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0127, -0.0140, 0.0047, 0.0337, -0.0077, 0.0286, 0.0075, 0.0158, + -0.0019, -0.0189], device='cuda:0'), grad: tensor([-1.1206e-05, 3.0939e-06, -1.7425e-06, 1.0477e-06, 6.1281e-06, + 4.3772e-08, 5.8673e-07, -1.5926e-07, 1.4333e-06, 7.7533e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 214.39, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4793 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.08 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.0737, -0.0272, -0.0978, ..., 0.0315, -0.1130, -0.0067], + [ 0.0476, -0.0981, -0.0267, ..., -0.0542, -0.0798, -0.2821], + [-0.0107, 0.0727, -0.0298, ..., -0.1127, 0.1357, -0.0381], + ..., + [ 0.0257, -0.0287, -0.0168, ..., -0.1566, -0.1904, 0.0358], + [-0.0262, -0.0952, -0.0285, ..., -0.1069, 0.0574, -0.1606], + [-0.2484, -0.0169, -0.0181, ..., -0.0956, -0.2080, -0.0958]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2945e-07, 1.2219e-06, + 4.6566e-09], + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.5832e-08, 1.3318e-07, + 2.8871e-08], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7951e-07, 2.9318e-06, + 3.7253e-09], + ..., + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, 4.4052e-07, + 5.2154e-08], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5646e-07, 4.0652e-07, + 2.3283e-09], + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.8161e-08, 3.1246e-07, + 9.5554e-07]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0121, -0.0140, 0.0055, 0.0337, -0.0075, 0.0289, 0.0068, 0.0157, + -0.0021, -0.0193], device='cuda:0'), grad: tensor([ 2.6245e-06, -2.8405e-07, 6.4895e-06, 1.9893e-06, -2.7120e-06, + -1.7047e-05, 3.7812e-06, 1.2908e-06, 5.4203e-07, 3.3230e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 214.38, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4491 re_mapping 0.0042 re_causal 0.0121 /// teacc 99.11 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.0738, -0.0273, -0.0978, ..., 0.0302, -0.1133, -0.0070], + [ 0.0472, -0.0986, -0.0267, ..., -0.0542, -0.0798, -0.2822], + [-0.0110, 0.0719, -0.0298, ..., -0.1129, 0.1355, -0.0381], + ..., + [ 0.0264, -0.0278, -0.0168, ..., -0.1574, -0.1905, 0.0357], + [-0.0262, -0.0953, -0.0285, ..., -0.1071, 0.0576, -0.1607], + [-0.2493, -0.0169, -0.0181, ..., -0.0936, -0.2088, -0.0956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.5856e-08, + 3.7253e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 6.9849e-09, 2.3283e-09], + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -7.4506e-09, 9.3132e-10], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 5.1223e-09, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + -2.8871e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-08, + 2.0955e-08, 1.4808e-07]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0139, -0.0141, 0.0049, 0.0338, -0.0080, 0.0288, 0.0066, 0.0159, + -0.0020, -0.0178], device='cuda:0'), grad: tensor([-1.5227e-07, 1.9977e-07, 3.4459e-08, 1.5181e-07, -3.4552e-07, + -7.5437e-08, 7.5903e-08, -4.4471e-07, 1.5832e-08, 5.4622e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.38, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4847 re_mapping 0.0038 re_causal 0.0120 /// teacc 99.04 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.0738, -0.0273, -0.0978, ..., 0.0297, -0.1135, -0.0073], + [ 0.0476, -0.0987, -0.0268, ..., -0.0544, -0.0799, -0.2824], + [-0.0109, 0.0719, -0.0298, ..., -0.1129, 0.1351, -0.0379], + ..., + [ 0.0262, -0.0278, -0.0172, ..., -0.1580, -0.1908, 0.0358], + [-0.0261, -0.0952, -0.0286, ..., -0.1072, 0.0580, -0.1611], + [-0.2499, -0.0170, -0.0182, ..., -0.0929, -0.2103, -0.0957]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8161e-08, + 1.8626e-08, 0.0000e+00], + [-4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 3.5856e-08, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 3.2596e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 3.7253e-09, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7229e-08, + -7.2177e-08, 9.3132e-10], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 3.2596e-09, 2.3283e-08]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0145, -0.0142, 0.0045, 0.0337, -0.0079, 0.0289, 0.0064, 0.0160, + -0.0017, -0.0178], device='cuda:0'), grad: tensor([ 5.6345e-08, 2.7232e-06, 1.3877e-07, -2.7474e-08, -1.7881e-07, + -1.5786e-07, 3.4459e-08, -3.1423e-06, -7.2177e-08, 6.2445e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 214.65, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4904 re_mapping 0.0039 re_causal 0.0125 /// teacc 99.07 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.0739, -0.0273, -0.0978, ..., 0.0299, -0.1136, -0.0074], + [ 0.0475, -0.0991, -0.0268, ..., -0.0544, -0.0799, -0.2826], + [-0.0109, 0.0719, -0.0298, ..., -0.1129, 0.1353, -0.0375], + ..., + [ 0.0262, -0.0275, -0.0173, ..., -0.1585, -0.1912, 0.0360], + [-0.0255, -0.0953, -0.0287, ..., -0.1073, 0.0581, -0.1611], + [-0.2500, -0.0170, -0.0182, ..., -0.0929, -0.2107, -0.0958]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.0955e-08, + 3.1665e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 9.9652e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 2.3562e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.5635e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + -2.4457e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + 8.9873e-08, 4.6566e-10]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0144, -0.0141, 0.0042, 0.0336, -0.0078, 0.0290, 0.0063, 0.0161, + -0.0017, -0.0181], device='cuda:0'), grad: tensor([-1.1642e-08, 2.6496e-07, 5.7407e-06, 1.2862e-06, 4.6566e-08, + -2.2948e-06, 3.9628e-07, 8.6147e-08, -5.8413e-06, 3.0454e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 214.56, cls_loss 0.0012 cls_loss_mapping 0.0024 cls_loss_causal 0.4950 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.05 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.0739, -0.0274, -0.0978, ..., 0.0298, -0.1139, -0.0073], + [ 0.0479, -0.0991, -0.0272, ..., -0.0545, -0.0829, -0.2826], + [-0.0113, 0.0721, -0.0300, ..., -0.1130, 0.1391, -0.0370], + ..., + [ 0.0263, -0.0276, -0.0176, ..., -0.1589, -0.1916, 0.0362], + [-0.0254, -0.0954, -0.0283, ..., -0.1074, 0.0580, -0.1612], + [-0.2500, -0.0170, -0.0183, ..., -0.0929, -0.2113, -0.0960]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6298e-08, + 5.1223e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8161e-08, + 4.3772e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 1.2573e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 1.1176e-08, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.1548e-07, + -2.8871e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0145, -0.0160, 0.0075, 0.0340, -0.0077, 0.0287, 0.0062, 0.0163, + -0.0017, -0.0185], device='cuda:0'), grad: tensor([-1.4901e-08, 2.3702e-07, 3.8324e-07, 1.2219e-06, 7.8697e-08, + -1.2526e-07, 4.4238e-08, -6.2166e-07, -1.1511e-06, -7.4040e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.68, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.5055 re_mapping 0.0038 re_causal 0.0123 /// teacc 98.94 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.0740, -0.0274, -0.0978, ..., 0.0299, -0.1144, -0.0073], + [ 0.0481, -0.0991, -0.0274, ..., -0.0547, -0.0829, -0.2828], + [-0.0115, 0.0721, -0.0300, ..., -0.1131, 0.1391, -0.0366], + ..., + [ 0.0262, -0.0275, -0.0181, ..., -0.1592, -0.1919, 0.0362], + [-0.0254, -0.0954, -0.0285, ..., -0.1077, 0.0580, -0.1614], + [-0.2501, -0.0170, -0.0185, ..., -0.0929, -0.2117, -0.0972]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, + 1.1176e-08, 0.0000e+00], + [-4.6566e-09, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.9558e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-08, + -7.9162e-08, -4.6566e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 7.2643e-08, 8.3819e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 6.2399e-08, + 9.8720e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-09, 7.4506e-09]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0145, -0.0160, 0.0073, 0.0337, -0.0045, 0.0288, 0.0064, 0.0165, + -0.0018, -0.0222], device='cuda:0'), grad: tensor([-2.7940e-09, 3.9581e-07, 2.5146e-08, 1.4514e-05, -4.4703e-08, + 9.4995e-08, -3.2224e-07, -1.5378e-05, 4.7125e-07, 2.2072e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.73, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.4791 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.01 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.0741, -0.0274, -0.0979, ..., 0.0298, -0.1147, -0.0075], + [ 0.0499, -0.0992, -0.0309, ..., -0.0549, -0.0830, -0.2830], + [-0.0117, 0.0721, -0.0297, ..., -0.1135, 0.1392, -0.0387], + ..., + [ 0.0259, -0.0275, -0.0179, ..., -0.1611, -0.1929, 0.0361], + [-0.0261, -0.0954, -0.0286, ..., -0.1082, 0.0580, -0.1618], + [-0.2502, -0.0167, -0.0186, ..., -0.0929, -0.2112, -0.0980]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 8.3819e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -2.5239e-07, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.2573e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 2.2352e-08, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0146, -0.0162, 0.0072, 0.0336, -0.0028, 0.0290, 0.0062, 0.0167, + -0.0021, -0.0231], device='cuda:0'), grad: tensor([ 1.9558e-08, 1.4901e-07, -6.2305e-07, 1.2759e-07, 6.0536e-08, + -1.1362e-07, 1.6764e-08, 3.2503e-07, 9.8720e-08, -5.8673e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 214.93, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4970 re_mapping 0.0040 re_causal 0.0132 /// teacc 98.92 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.0741, -0.0274, -0.0979, ..., 0.0299, -0.1150, -0.0075], + [ 0.0499, -0.0992, -0.0301, ..., -0.0548, -0.0830, -0.2831], + [-0.0116, 0.0722, -0.0297, ..., -0.1134, 0.1394, -0.0390], + ..., + [ 0.0258, -0.0275, -0.0184, ..., -0.1622, -0.1934, 0.0364], + [-0.0262, -0.0955, -0.0292, ..., -0.1086, 0.0578, -0.1618], + [-0.2502, -0.0166, -0.0188, ..., -0.0929, -0.2120, -0.0981]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -1.2387e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1362e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9116e-08, + 2.9802e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0147, -0.0161, 0.0073, 0.0335, -0.0029, 0.0292, 0.0062, 0.0166, + -0.0023, -0.0233], device='cuda:0'), grad: tensor([ 2.7008e-08, 1.1409e-06, 9.0338e-08, -3.5390e-08, 6.6031e-07, + 2.7195e-07, -2.7195e-07, -2.7400e-06, 5.6811e-07, 2.7101e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 214.74, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4736 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.03 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.0741, -0.0274, -0.0980, ..., 0.0302, -0.1152, -0.0068], + [ 0.0498, -0.0992, -0.0299, ..., -0.0565, -0.0831, -0.2832], + [-0.0117, 0.0722, -0.0297, ..., -0.1139, 0.1395, -0.0398], + ..., + [ 0.0259, -0.0275, -0.0190, ..., -0.1627, -0.1938, 0.0365], + [-0.0255, -0.0955, -0.0294, ..., -0.1088, 0.0580, -0.1619], + [-0.2504, -0.0166, -0.0193, ..., -0.0930, -0.2137, -0.0981]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9092e-07, + 1.2293e-07, 0.0000e+00], + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-08, + 4.9081e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7008e-08, + -9.2667e-07, -1.8626e-09], + ..., + [-9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 2.7940e-09, + 2.1141e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.6601e-07, + 1.6298e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.5367e-07, + 6.5193e-08, 1.0245e-08]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0146, -0.0163, 0.0072, 0.0334, -0.0030, 0.0292, 0.0064, 0.0168, + -0.0021, -0.0231], device='cuda:0'), grad: tensor([ 5.6252e-07, 2.6748e-06, -4.7311e-06, 9.4064e-07, 6.8732e-07, + 2.8592e-07, -2.5220e-06, 7.9162e-07, 8.9314e-07, 3.9581e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 214.56, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4974 re_mapping 0.0041 re_causal 0.0124 /// teacc 98.99 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.0742, -0.0275, -0.0980, ..., 0.0349, -0.1155, -0.0064], + [ 0.0499, -0.0995, -0.0332, ..., -0.0543, -0.0831, -0.2833], + [-0.0117, 0.0725, -0.0295, ..., -0.1145, 0.1397, -0.0393], + ..., + [ 0.0261, -0.0273, -0.0183, ..., -0.1636, -0.1945, 0.0366], + [-0.0256, -0.0956, -0.0294, ..., -0.1095, 0.0578, -0.1624], + [-0.2522, -0.0168, -0.0214, ..., -0.0941, -0.2146, -0.0989]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-08, + 7.4506e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 9.3132e-10, + 9.3132e-10, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 1.8626e-09, + -1.3039e-08, 7.4506e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.6508e-07, ..., 0.0000e+00, + 3.7253e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 4.6566e-09, + 1.3039e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-08, + 0.0000e+00, 4.0978e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0101, -0.0165, 0.0073, 0.0337, -0.0026, 0.0290, 0.0032, 0.0171, + -0.0023, -0.0254], device='cuda:0'), grad: tensor([-4.1910e-08, 1.5926e-07, 1.8161e-07, -7.5325e-06, -1.7416e-07, + 6.0536e-08, 1.7695e-08, 6.9290e-06, 1.6298e-07, 2.3097e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 214.87, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4850 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.0742, -0.0275, -0.0980, ..., 0.0354, -0.1157, -0.0072], + [ 0.0502, -0.0995, -0.0328, ..., -0.0542, -0.0830, -0.2834], + [-0.0117, 0.0725, -0.0295, ..., -0.1147, 0.1396, -0.0391], + ..., + [ 0.0260, -0.0273, -0.0185, ..., -0.1644, -0.1957, 0.0367], + [-0.0257, -0.0957, -0.0294, ..., -0.1100, 0.0579, -0.1625], + [-0.2523, -0.0168, -0.0217, ..., -0.0945, -0.2155, -0.0991]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + 8.2888e-08, 6.6124e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.6368e-08, + 1.5367e-07, 2.5146e-06], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1420e-07, + 2.4587e-07, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2945e-07, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 8.9407e-08, + 3.3248e-07, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.3039e-08, -7.0669e-06]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0096, -0.0165, 0.0069, 0.0335, -0.0018, 0.0295, 0.0027, 0.0174, + -0.0023, -0.0263], device='cuda:0'), grad: tensor([ 5.5134e-07, 1.4186e-05, 5.2713e-07, 9.3784e-07, 2.6256e-05, + -6.2957e-07, -4.1761e-06, 3.9302e-07, 8.8383e-07, -3.8981e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 214.55, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4786 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.0743, -0.0275, -0.0980, ..., 0.0356, -0.1155, -0.0075], + [ 0.0503, -0.0996, -0.0331, ..., -0.0548, -0.0830, -0.2836], + [-0.0119, 0.0725, -0.0295, ..., -0.1161, 0.1396, -0.0392], + ..., + [ 0.0261, -0.0273, -0.0184, ..., -0.1655, -0.1962, 0.0367], + [-0.0256, -0.0957, -0.0294, ..., -0.1115, 0.0577, -0.1625], + [-0.2524, -0.0168, -0.0218, ..., -0.0946, -0.2160, -0.0992]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.7295e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + -2.6822e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.4809e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, + -1.4994e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0095, -0.0164, 0.0067, 0.0336, -0.0017, 0.0296, 0.0030, 0.0174, + -0.0026, -0.0263], device='cuda:0'), grad: tensor([ 2.3283e-08, 1.1325e-06, -5.8971e-06, 9.1270e-08, 8.8476e-08, + 2.5891e-07, 2.7940e-09, 4.4852e-06, -2.8312e-07, 6.5193e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 214.22, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4615 re_mapping 0.0038 re_causal 0.0116 /// teacc 99.06 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.0743, -0.0275, -0.0983, ..., 0.0356, -0.1156, -0.0079], + [ 0.0503, -0.0996, -0.0343, ..., -0.0549, -0.0830, -0.2837], + [-0.0119, 0.0725, -0.0294, ..., -0.1165, 0.1395, -0.0393], + ..., + [ 0.0261, -0.0273, -0.0188, ..., -0.1657, -0.1971, 0.0367], + [-0.0255, -0.0957, -0.0296, ..., -0.1123, 0.0580, -0.1625], + [-0.2525, -0.0168, -0.0224, ..., -0.0946, -0.2167, -0.0992]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, 0.0000e+00, ..., 2.2445e-07, + 3.5949e-07, 6.6124e-08], + [-1.4994e-07, 0.0000e+00, 0.0000e+00, ..., 2.5146e-08, + 6.5193e-08, 0.0000e+00], + [ 7.7300e-08, 0.0000e+00, 0.0000e+00, ..., 1.3970e-08, + -5.5879e-09, 1.5832e-08], + ..., + [ 8.3819e-09, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, + 2.8871e-08, 0.0000e+00], + [ 4.0047e-08, 0.0000e+00, 0.0000e+00, ..., 1.2675e-06, + 1.6084e-06, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-08, + 3.8184e-08, 9.3132e-10]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0094, -0.0164, 0.0065, 0.0337, -0.0018, 0.0295, 0.0030, 0.0174, + -0.0024, -0.0262], device='cuda:0'), grad: tensor([ 7.9349e-07, 1.6633e-06, 2.9895e-07, 3.1292e-07, 1.0431e-07, + -8.7172e-07, -3.8184e-06, -2.4512e-06, 3.5688e-06, 4.1071e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 214.49, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4817 re_mapping 0.0038 re_causal 0.0122 /// teacc 99.08 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.0745, -0.0275, -0.0984, ..., 0.0356, -0.1160, -0.0082], + [ 0.0508, -0.0996, -0.0381, ..., -0.0549, -0.0832, -0.2839], + [-0.0120, 0.0725, -0.0288, ..., -0.1168, 0.1397, -0.0393], + ..., + [ 0.0261, -0.0273, -0.0181, ..., -0.1660, -0.1973, 0.0363], + [-0.0259, -0.0957, -0.0272, ..., -0.1129, 0.0579, -0.1625], + [-0.2528, -0.0168, -0.0250, ..., -0.0945, -0.2171, -0.0990]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 1.0245e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -5.1223e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.5635e-08, 0.0000e+00]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0095, -0.0164, 0.0065, 0.0337, -0.0018, 0.0296, 0.0031, 0.0173, + -0.0024, -0.0259], device='cuda:0'), grad: tensor([-1.5832e-08, -5.5227e-07, -8.5682e-08, 0.0000e+00, 1.1176e-07, + 8.1956e-08, -1.7695e-07, 5.4948e-08, 4.6194e-07, 1.2014e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 214.94, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4692 re_mapping 0.0037 re_causal 0.0120 /// teacc 99.02 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.0745, -0.0275, -0.0984, ..., 0.0356, -0.1162, -0.0083], + [ 0.0505, -0.0997, -0.0413, ..., -0.0556, -0.0832, -0.2839], + [-0.0120, 0.0724, -0.0282, ..., -0.1167, 0.1398, -0.0400], + ..., + [ 0.0264, -0.0271, -0.0169, ..., -0.1665, -0.1976, 0.0363], + [-0.0260, -0.0957, -0.0271, ..., -0.1134, 0.0578, -0.1626], + [-0.2530, -0.0168, -0.0254, ..., -0.0945, -0.2173, -0.0990]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1420e-08, + 3.7253e-09, 0.0000e+00], + [ 5.5879e-08, 0.0000e+00, 0.0000e+00, ..., -1.1455e-07, + 2.6356e-07, 0.0000e+00], + [ 3.2596e-08, 0.0000e+00, 0.0000e+00, ..., 2.3283e-08, + 5.4017e-08, 0.0000e+00], + ..., + [-1.0990e-07, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 7.4506e-09, 2.7940e-09], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 3.8184e-08, + -1.8999e-07, 0.0000e+00], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 4.5635e-08, + 9.3132e-09, -8.3819e-09]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0095, -0.0165, 0.0064, 0.0334, -0.0018, 0.0302, 0.0027, 0.0176, + -0.0026, -0.0261], device='cuda:0'), grad: tensor([ 2.9616e-07, 3.6545e-06, 2.8741e-06, 2.9430e-07, 5.4482e-07, + -2.6263e-07, 9.3132e-10, -8.5458e-06, -5.0291e-08, 1.2042e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 214.96, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4835 re_mapping 0.0038 re_causal 0.0122 /// teacc 99.01 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.0745, -0.0275, -0.0984, ..., 0.0356, -0.1169, -0.0083], + [ 0.0506, -0.0997, -0.0412, ..., -0.0560, -0.0832, -0.2841], + [-0.0118, 0.0725, -0.0284, ..., -0.1162, 0.1399, -0.0398], + ..., + [ 0.0262, -0.0271, -0.0168, ..., -0.1679, -0.1985, 0.0364], + [-0.0261, -0.0957, -0.0270, ..., -0.1135, 0.0580, -0.1626], + [-0.2532, -0.0168, -0.0256, ..., -0.0945, -0.2178, -0.0991]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 4.6566e-09, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, -1.2387e-07, ..., 1.8626e-09, + 4.6566e-09, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 2.7940e-09, + -5.5879e-09, 8.3819e-09], + ..., + [ 0.0000e+00, 3.7253e-09, 8.1025e-08, ..., 0.0000e+00, + 1.1176e-08, -5.6811e-08], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 1.8626e-08, + 1.0803e-07, 2.3283e-08], + [ 0.0000e+00, -1.0245e-08, 5.5879e-09, ..., 1.8626e-09, + 7.4506e-09, -1.8626e-09]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0096, -0.0165, 0.0064, 0.0333, -0.0018, 0.0303, 0.0028, 0.0176, + -0.0024, -0.0261], device='cuda:0'), grad: tensor([ 2.2352e-08, -9.6112e-07, 2.0955e-07, -1.6950e-07, 1.7881e-07, + -2.4121e-07, 4.1537e-07, 1.9558e-08, 5.4576e-07, -2.8871e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 214.91, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4873 re_mapping 0.0037 re_causal 0.0123 /// teacc 99.11 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.0746, -0.0275, -0.0984, ..., 0.0357, -0.1174, -0.0084], + [ 0.0507, -0.0998, -0.0411, ..., -0.0564, -0.0832, -0.2843], + [-0.0120, 0.0724, -0.0285, ..., -0.1162, 0.1395, -0.0397], + ..., + [ 0.0262, -0.0271, -0.0169, ..., -0.1689, -0.1991, 0.0363], + [-0.0261, -0.0958, -0.0270, ..., -0.1143, 0.0588, -0.1626], + [-0.2535, -0.0167, -0.0258, ..., -0.0947, -0.2194, -0.0993]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9744e-07, + 9.3225e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4342e-07, + 6.7893e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.9162e-08, + 3.6880e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + 6.3330e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.2678e-07, + -2.9653e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + 8.5682e-08, -0.0000e+00]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0093, -0.0165, 0.0061, 0.0330, -0.0017, 0.0305, 0.0031, 0.0175, + -0.0017, -0.0266], device='cuda:0'), grad: tensor([ 3.6284e-06, -4.3988e-05, 1.5190e-06, 1.7807e-06, 4.4703e-08, + 4.6846e-07, 1.0021e-06, 4.6492e-05, -1.1414e-05, 4.4797e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 214.67, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4773 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.11 lr 0.00010000 diff --git a/Meta-causal/code-withStyleAttack/66524.error b/Meta-causal/code-withStyleAttack/66524.error new file mode 100644 index 0000000000000000000000000000000000000000..aa2c7d390039dd22ff394c4ca54b97e4505a2c80 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66524.error @@ -0,0 +1,4 @@ +run_my_joint_test.sh: line 28: actor_num}fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_RA: command not found +slurmstepd: error: *** STEP 66524.0 ON gcp-us-0 CANCELLED AT 2024-07-21T15:23:20 DUE TO TIME LIMIT *** +slurmstepd: error: *** JOB 66524 ON gcp-us-0 CANCELLED AT 2024-07-21T15:23:20 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-withStyleAttack/66524.log b/Meta-causal/code-withStyleAttack/66524.log new file mode 100644 index 0000000000000000000000000000000000000000..49e2766242b2515ca4a26112f48804dd69ce2a39 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66524.log @@ -0,0 +1,21785 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0052, 0.0130, -0.0176, ..., 0.0223, 0.0266, -0.0208], + [-0.0176, -0.0280, 0.0186, ..., 0.0191, -0.0301, 0.0311], + [ 0.0254, -0.0046, 0.0059, ..., 0.0028, -0.0158, 0.0082], + ..., + [ 0.0110, -0.0042, 0.0268, ..., 0.0126, 0.0060, -0.0196], + [ 0.0099, 0.0191, -0.0010, ..., -0.0097, -0.0281, -0.0062], + [-0.0233, 0.0243, 0.0245, ..., -0.0151, 0.0074, -0.0035]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0078, 0.0122, -0.0029, -0.0130, 0.0187, -0.0291, 0.0140, -0.0049, + 0.0100, -0.0102], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 231.48, cls_loss 2.2991 cls_loss_mapping 2.2972 cls_loss_causal 2.3020 re_mapping 0.0018 re_causal 0.0018 /// teacc 51.05 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0057, 0.0138, -0.0174, ..., 0.0207, 0.0243, -0.0223], + [-0.0171, -0.0256, 0.0198, ..., 0.0240, -0.0312, 0.0361], + [ 0.0257, -0.0039, 0.0054, ..., 0.0023, -0.0149, 0.0072], + ..., + [ 0.0108, -0.0039, 0.0281, ..., 0.0136, 0.0058, -0.0165], + [ 0.0086, 0.0168, -0.0015, ..., -0.0120, -0.0299, -0.0087], + [-0.0234, 0.0253, 0.0256, ..., -0.0143, 0.0068, -0.0027]], + device='cuda:0'), grad: tensor([[ 2.0131e-05, -4.5509e-03, -3.9043e-03, ..., -3.8662e-03, + 0.0000e+00, -5.1994e-03], + [ 2.1353e-05, -3.3188e-03, -2.7256e-03, ..., -3.5000e-03, + 0.0000e+00, -8.1558e-03], + [ 2.0415e-05, 4.5815e-03, 4.3068e-03, ..., 4.3297e-03, + 0.0000e+00, 8.0948e-03], + ..., + [ 2.0906e-05, -1.2436e-03, -3.0708e-03, ..., -4.7803e-04, + 0.0000e+00, -4.5052e-03], + [ 2.0370e-05, -1.9798e-03, -5.7125e-04, ..., 8.0884e-05, + 0.0000e+00, -2.6474e-03], + [ 2.0131e-05, 4.6387e-03, 3.8185e-03, ..., 3.9520e-03, + 0.0000e+00, 7.1259e-03]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0085, 0.0149, -0.0026, -0.0132, 0.0174, -0.0308, 0.0131, -0.0040, + 0.0083, -0.0096], device='cuda:0'), grad: tensor([-0.0554, -0.0491, 0.0742, 0.0446, -0.0300, 0.0557, -0.0311, -0.0500, + -0.0261, 0.0673], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 231.13, cls_loss 2.2235 cls_loss_mapping 1.8040 cls_loss_causal 2.1830 re_mapping 0.0486 re_causal 0.0262 /// teacc 70.41 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0057, 0.0134, -0.0203, ..., 0.0168, 0.0210, -0.0283], + [-0.0170, -0.0229, 0.0160, ..., 0.0305, -0.0278, 0.0433], + [ 0.0257, -0.0042, 0.0025, ..., 0.0016, -0.0156, 0.0025], + ..., + [ 0.0108, -0.0039, 0.0327, ..., 0.0135, 0.0005, -0.0107], + [ 0.0085, 0.0171, -0.0018, ..., -0.0124, -0.0313, -0.0093], + [-0.0234, 0.0231, 0.0270, ..., -0.0171, 0.0069, -0.0036]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0002, 0.0002, ..., -0.0026, -0.0003, -0.0005], + [ 0.0000, 0.0095, 0.0058, ..., 0.0019, 0.0033, 0.0050], + [ 0.0000, 0.0026, 0.0005, ..., 0.0007, 0.0013, 0.0014], + ..., + [ 0.0000, -0.0065, -0.0026, ..., 0.0020, 0.0046, -0.0061], + [ 0.0000, -0.0006, -0.0065, ..., -0.0021, -0.0051, -0.0043], + [ 0.0000, 0.0024, -0.0024, ..., 0.0006, -0.0031, -0.0012]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0083, 0.0164, -0.0031, -0.0141, 0.0177, -0.0318, 0.0141, -0.0039, + 0.0079, -0.0107], device='cuda:0'), grad: tensor([-0.1131, 0.1017, 0.0032, -0.0595, 0.0090, 0.0259, -0.0312, -0.0031, + 0.0331, 0.0340], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 230.77, cls_loss 2.0055 cls_loss_mapping 0.9298 cls_loss_causal 1.8686 re_mapping 0.1309 re_causal 0.1116 /// teacc 82.97 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0057, 0.0108, -0.0246, ..., 0.0131, 0.0245, -0.0328], + [-0.0170, -0.0256, 0.0107, ..., 0.0358, -0.0283, 0.0501], + [ 0.0257, -0.0046, -0.0009, ..., 0.0018, -0.0145, -0.0048], + ..., + [ 0.0108, -0.0064, 0.0380, ..., 0.0102, -0.0014, -0.0064], + [ 0.0085, 0.0187, -0.0044, ..., -0.0124, -0.0324, -0.0098], + [-0.0234, 0.0227, 0.0314, ..., -0.0204, 0.0048, -0.0017]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0023, -0.0015, ..., 0.0057, 0.0006, -0.0002], + [ 0.0000, 0.0026, 0.0052, ..., -0.0006, 0.0002, -0.0010], + [ 0.0000, 0.0002, -0.0013, ..., 0.0012, -0.0009, 0.0003], + ..., + [ 0.0000, 0.0013, -0.0043, ..., -0.0007, -0.0028, 0.0030], + [ 0.0000, -0.0004, 0.0012, ..., -0.0003, 0.0006, -0.0058], + [ 0.0000, 0.0017, 0.0057, ..., 0.0017, 0.0006, 0.0050]], + device='cuda:0') +Epoch 4, bias, value: tensor([-0.0082, 0.0170, -0.0044, -0.0138, 0.0166, -0.0314, 0.0139, -0.0038, + 0.0083, -0.0098], device='cuda:0'), grad: tensor([ 0.0600, 0.0519, 0.0250, 0.0266, -0.0198, -0.0127, 0.0191, -0.1173, + -0.0431, 0.0104], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 230.30, cls_loss 1.7935 cls_loss_mapping 0.5301 cls_loss_causal 1.6149 re_mapping 0.1324 re_causal 0.1553 /// teacc 87.69 lr 0.00010000 +Epoch 5, weight, value: tensor([[-5.6976e-03, 6.1139e-03, -2.8751e-02, ..., 8.1039e-03, + 2.8980e-02, -3.6270e-02], + [-1.7020e-02, -2.9472e-02, 6.6490e-03, ..., 4.3678e-02, + -3.0254e-02, 5.7222e-02], + [ 2.5681e-02, -6.2469e-03, 9.9714e-06, ..., -2.0465e-03, + -1.4558e-02, -1.1056e-02], + ..., + [ 1.0841e-02, -1.0236e-02, 4.3665e-02, ..., 7.8569e-03, + -3.1083e-03, -2.3670e-03], + [ 8.4641e-03, 2.2885e-02, -7.8319e-03, ..., -1.3490e-02, + -3.3058e-02, -1.1133e-02], + [-2.3407e-02, 2.3156e-02, 3.1968e-02, ..., -2.2863e-02, + -2.9661e-04, -2.2200e-03]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.6041e-03, -2.8348e-04, ..., -5.1117e-03, + 7.3166e-03, 3.3016e-03], + [ 0.0000e+00, -6.5565e-04, -1.3733e-04, ..., -9.2621e-03, + 3.2196e-03, -8.3466e-03], + [ 0.0000e+00, 1.2693e-03, 4.9210e-03, ..., 1.3763e-02, + -6.2108e-05, 3.8967e-03], + ..., + [ 0.0000e+00, -2.7218e-03, -1.1337e-02, ..., -1.0216e-02, + -6.6853e-04, -2.3590e-02], + [ 0.0000e+00, 1.1978e-02, 5.9509e-03, ..., 1.5991e-02, + 4.9973e-03, 1.1780e-02], + [ 0.0000e+00, -5.0020e-04, -9.7275e-04, ..., -7.8049e-03, + 4.7302e-03, 9.9335e-03]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0096, 0.0179, -0.0054, -0.0133, 0.0172, -0.0323, 0.0138, -0.0028, + 0.0084, -0.0098], device='cuda:0'), grad: tensor([ 0.0082, 0.0043, 0.0320, 0.0408, 0.0610, -0.0357, -0.0361, -0.1022, + 0.0490, -0.0213], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 230.85, cls_loss 1.6301 cls_loss_mapping 0.3783 cls_loss_causal 1.4143 re_mapping 0.1115 re_causal 0.1606 /// teacc 91.28 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0057, 0.0019, -0.0315, ..., 0.0045, 0.0333, -0.0413], + [-0.0170, -0.0331, 0.0023, ..., 0.0490, -0.0319, 0.0636], + [ 0.0257, -0.0082, 0.0010, ..., -0.0025, -0.0134, -0.0130], + ..., + [ 0.0108, -0.0146, 0.0499, ..., 0.0067, -0.0061, 0.0004], + [ 0.0085, 0.0252, -0.0136, ..., -0.0162, -0.0331, -0.0149], + [-0.0234, 0.0241, 0.0327, ..., -0.0250, -0.0071, -0.0007]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0007, 0.0009, ..., 0.0052, -0.0066, 0.0089], + [ 0.0000, 0.0091, 0.0186, ..., 0.0093, 0.0028, 0.0051], + [ 0.0000, 0.0049, 0.0037, ..., -0.0011, 0.0065, -0.0066], + ..., + [ 0.0000, 0.0086, -0.0149, ..., -0.0098, 0.0006, -0.0166], + [ 0.0000, -0.0359, -0.0265, ..., -0.0090, 0.0027, -0.0201], + [ 0.0000, 0.0060, 0.0044, ..., 0.0035, -0.0018, 0.0087]], + device='cuda:0') +Epoch 6, bias, value: tensor([-0.0101, 0.0178, -0.0046, -0.0132, 0.0171, -0.0323, 0.0143, -0.0033, + 0.0076, -0.0092], device='cuda:0'), grad: tensor([ 0.0208, 0.0356, -0.0256, -0.0134, -0.0276, 0.0387, 0.0295, 0.0223, + -0.0927, 0.0123], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 230.59, cls_loss 1.5362 cls_loss_mapping 0.2963 cls_loss_causal 1.3447 re_mapping 0.0951 re_causal 0.1638 /// teacc 91.60 lr 0.00010000 +Epoch 7, weight, value: tensor([[-5.6976e-03, -1.3057e-03, -3.3664e-02, ..., 1.4445e-03, + 3.5436e-02, -4.2144e-02], + [-1.7020e-02, -3.5231e-02, 1.2087e-05, ..., 5.3517e-02, + -3.3874e-02, 6.8826e-02], + [ 2.5681e-02, -9.4805e-03, 1.5294e-03, ..., -2.3634e-03, + -1.3073e-02, -1.6458e-02], + ..., + [ 1.0841e-02, -1.8503e-02, 5.4470e-02, ..., 4.9900e-03, + -6.5514e-03, 2.6829e-03], + [ 8.4641e-03, 2.6405e-02, -1.8146e-02, ..., -1.7512e-02, + -3.1838e-02, -1.6849e-02], + [-2.3407e-02, 2.4911e-02, 3.3542e-02, ..., -2.6230e-02, + -8.8147e-03, -4.4122e-04]], device='cuda:0'), grad: tensor([[ 0.0000, -0.0097, 0.0028, ..., -0.0003, -0.0027, 0.0009], + [ 0.0000, -0.0025, -0.0077, ..., 0.0042, 0.0006, -0.0096], + [ 0.0000, -0.0136, -0.0192, ..., 0.0147, -0.0019, 0.0023], + ..., + [ 0.0000, 0.0026, -0.0014, ..., 0.0043, 0.0004, -0.0055], + [ 0.0000, 0.0216, 0.0242, ..., -0.0037, 0.0005, 0.0202], + [ 0.0000, 0.0005, 0.0087, ..., 0.0090, 0.0006, 0.0055]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0097, 0.0184, -0.0047, -0.0136, 0.0160, -0.0319, 0.0143, -0.0028, + 0.0067, -0.0088], device='cuda:0'), grad: tensor([-0.0353, -0.0064, -0.0164, 0.0045, 0.0006, -0.0163, -0.0499, 0.0174, + 0.0787, 0.0231], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 230.55, cls_loss 1.4292 cls_loss_mapping 0.2519 cls_loss_causal 1.2321 re_mapping 0.0855 re_causal 0.1583 /// teacc 94.39 lr 0.00010000 +Epoch 8, weight, value: tensor([[-5.7063e-03, -4.2141e-03, -3.4838e-02, ..., -4.4554e-06, + 3.6260e-02, -4.2675e-02], + [-1.7024e-02, -3.9776e-02, -3.1388e-03, ..., 5.6652e-02, + -3.5487e-02, 7.2111e-02], + [ 2.5821e-02, -9.4525e-03, 1.0607e-03, ..., -4.5522e-03, + -1.1909e-02, -2.0593e-02], + ..., + [ 1.0823e-02, -2.1375e-02, 5.7177e-02, ..., 4.1580e-03, + -7.7836e-03, 5.6238e-03], + [ 8.4577e-03, 2.8233e-02, -2.1511e-02, ..., -1.7527e-02, + -3.1161e-02, -1.8683e-02], + [-2.3411e-02, 2.5982e-02, 3.5122e-02, ..., -2.8937e-02, + -1.0780e-02, -5.7312e-04]], device='cuda:0'), grad: tensor([[ 0.0000, 0.0054, -0.0019, ..., -0.0006, 0.0081, 0.0052], + [ 0.0000, -0.0052, 0.0056, ..., -0.0134, 0.0007, -0.0234], + [ 0.0000, 0.0087, 0.0022, ..., 0.0008, 0.0017, 0.0071], + ..., + [ 0.0000, 0.0027, -0.0172, ..., 0.0022, -0.0006, 0.0037], + [ 0.0000, 0.0408, 0.0140, ..., 0.0141, -0.0161, 0.0293], + [ 0.0000, -0.0354, -0.0096, ..., -0.0059, 0.0006, -0.0403]], + device='cuda:0') +Epoch 8, bias, value: tensor([-0.0094, 0.0180, -0.0048, -0.0121, 0.0155, -0.0324, 0.0143, -0.0031, + 0.0063, -0.0084], device='cuda:0'), grad: tensor([-0.0265, -0.0152, 0.0107, -0.0571, 0.0038, 0.0730, -0.0297, 0.0327, + 0.0764, -0.0680], device='cuda:0') +100 +0.0001 +changing lr +epoch 7, time 214.16, cls_loss 1.3617 cls_loss_mapping 0.2373 cls_loss_causal 1.1436 re_mapping 0.0755 re_causal 0.1420 /// teacc 94.14 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0057, -0.0069, -0.0361, ..., -0.0004, 0.0379, -0.0449], + [-0.0170, -0.0426, -0.0061, ..., 0.0594, -0.0368, 0.0751], + [ 0.0258, -0.0093, 0.0014, ..., -0.0063, -0.0099, -0.0228], + ..., + [ 0.0108, -0.0231, 0.0598, ..., 0.0040, -0.0073, 0.0077], + [ 0.0085, 0.0279, -0.0245, ..., -0.0187, -0.0310, -0.0200], + [-0.0234, 0.0290, 0.0370, ..., -0.0294, -0.0122, -0.0004]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0095, 0.0045, ..., 0.0101, -0.0065, -0.0095], + [ 0.0000, 0.0068, 0.0005, ..., -0.0016, 0.0022, -0.0028], + [ 0.0000, 0.0129, 0.0077, ..., 0.0043, 0.0029, 0.0068], + ..., + [ 0.0000, 0.0034, 0.0057, ..., -0.0059, 0.0024, 0.0022], + [ 0.0000, 0.0078, 0.0044, ..., 0.0035, 0.0037, 0.0053], + [ 0.0000, -0.0394, -0.0240, ..., -0.0087, -0.0117, -0.0271]], + device='cuda:0') +Epoch 9, bias, value: tensor([-0.0094, 0.0181, -0.0042, -0.0121, 0.0149, -0.0325, 0.0148, -0.0026, + 0.0050, -0.0081], device='cuda:0'), grad: tensor([ 0.0015, 0.0416, 0.0329, -0.0115, 0.0080, -0.0279, 0.0376, -0.0254, + 0.0072, -0.0640], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 230.28, cls_loss 1.3292 cls_loss_mapping 0.2062 cls_loss_causal 1.1216 re_mapping 0.0680 re_causal 0.1325 /// teacc 95.28 lr 0.00010000 +Epoch 10, weight, value: tensor([[-5.7063e-03, -8.2545e-03, -3.7036e-02, ..., -1.3193e-03, + 3.7539e-02, -4.7933e-02], + [-1.7024e-02, -4.4170e-02, -7.0778e-03, ..., 6.1542e-02, + -3.6655e-02, 7.8375e-02], + [ 2.5821e-02, -1.0798e-02, 1.8739e-03, ..., -6.4659e-03, + -9.6074e-03, -2.4274e-02], + ..., + [ 1.0823e-02, -2.6266e-02, 6.1551e-02, ..., 4.3617e-03, + -8.3729e-03, 9.8277e-03], + [ 8.4577e-03, 2.7872e-02, -2.7418e-02, ..., -1.9583e-02, + -3.2037e-02, -2.1523e-02], + [-2.3411e-02, 3.0944e-02, 3.8484e-02, ..., -3.2093e-02, + -1.1417e-02, 8.9331e-05]], device='cuda:0'), grad: tensor([[ 0.0000, -0.0504, -0.0321, ..., -0.0421, -0.0050, -0.0135], + [ 0.0000, 0.0081, 0.0075, ..., 0.0014, 0.0003, -0.0011], + [ 0.0000, 0.0166, 0.0157, ..., 0.0053, 0.0010, 0.0043], + ..., + [ 0.0000, -0.0191, -0.0439, ..., -0.0104, 0.0009, -0.0150], + [ 0.0000, -0.0040, 0.0118, ..., 0.0146, -0.0019, 0.0102], + [ 0.0000, -0.0053, 0.0004, ..., 0.0002, 0.0008, -0.0031]], + device='cuda:0') +Epoch 10, bias, value: tensor([-0.0095, 0.0181, -0.0042, -0.0114, 0.0148, -0.0323, 0.0143, -0.0021, + 0.0048, -0.0084], device='cuda:0'), grad: tensor([-0.1255, 0.0139, 0.0131, 0.0331, 0.0308, 0.0127, 0.0201, -0.0258, + 0.0405, -0.0129], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 214.28, cls_loss 1.2755 cls_loss_mapping 0.1872 cls_loss_causal 1.0805 re_mapping 0.0634 re_causal 0.1248 /// teacc 95.16 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0057, -0.0093, -0.0371, ..., -0.0008, 0.0370, -0.0489], + [-0.0170, -0.0447, -0.0082, ..., 0.0632, -0.0366, 0.0801], + [ 0.0258, -0.0121, 0.0004, ..., -0.0077, -0.0086, -0.0264], + ..., + [ 0.0108, -0.0282, 0.0630, ..., 0.0029, -0.0088, 0.0114], + [ 0.0085, 0.0290, -0.0282, ..., -0.0199, -0.0320, -0.0226], + [-0.0234, 0.0314, 0.0386, ..., -0.0328, -0.0118, 0.0001]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.2154e-02, 1.1292e-02, ..., 1.3756e-02, + 3.3379e-03, 1.3878e-02], + [ 0.0000e+00, -7.3090e-03, -5.6952e-05, ..., -1.3878e-02, + -1.3943e-03, -1.5961e-02], + [ 0.0000e+00, -1.2596e-02, -6.9580e-03, ..., 7.9060e-04, + -1.4992e-02, -1.3374e-02], + ..., + [ 0.0000e+00, 1.2138e-02, 9.0179e-03, ..., 1.0658e-02, + 1.7605e-03, 7.7477e-03], + [ 0.0000e+00, 2.4048e-02, 1.2833e-02, ..., 1.8066e-02, + 1.9608e-02, 2.7283e-02], + [ 0.0000e+00, 6.2218e-03, 6.2752e-03, ..., 1.3855e-02, + 2.1381e-03, 9.5596e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0088, 0.0184, -0.0046, -0.0107, 0.0157, -0.0324, 0.0136, -0.0027, + 0.0047, -0.0090], device='cuda:0'), grad: tensor([ 0.0291, -0.0144, -0.0583, -0.0068, -0.0561, 0.0073, -0.0119, 0.0357, + 0.0603, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 214.45, cls_loss 1.2490 cls_loss_mapping 0.1857 cls_loss_causal 1.0529 re_mapping 0.0588 re_causal 0.1205 /// teacc 95.15 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0057, -0.0115, -0.0390, ..., -0.0027, 0.0383, -0.0507], + [-0.0170, -0.0458, -0.0089, ..., 0.0651, -0.0361, 0.0820], + [ 0.0258, -0.0134, 0.0001, ..., -0.0081, -0.0109, -0.0275], + ..., + [ 0.0108, -0.0293, 0.0650, ..., 0.0024, -0.0091, 0.0114], + [ 0.0085, 0.0303, -0.0309, ..., -0.0206, -0.0304, -0.0225], + [-0.0234, 0.0323, 0.0402, ..., -0.0338, -0.0132, 0.0004]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 7.8506e-03, 1.2169e-02, ..., 1.0902e-02, + 1.3313e-03, 1.3725e-02], + [ 3.4459e-08, 7.7515e-03, 9.5367e-03, ..., -1.4992e-03, + 8.0204e-04, 1.4496e-03], + [ 1.2200e-07, 7.7629e-03, 1.3344e-02, ..., 9.8953e-03, + -1.5001e-03, 1.0300e-02], + ..., + [ 2.2957e-07, 5.7983e-03, 5.1079e-03, ..., 6.0616e-03, + 7.7772e-04, 1.0544e-02], + [ 1.8626e-08, 4.6005e-03, 1.1604e-02, ..., 8.9722e-03, + 3.9177e-03, -2.4319e-03], + [ 3.5446e-06, 6.3400e-03, 2.8095e-03, ..., 3.8433e-03, + 2.1019e-03, 2.3270e-02]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0088, 0.0186, -0.0045, -0.0105, 0.0156, -0.0327, 0.0134, -0.0026, + 0.0045, -0.0091], device='cuda:0'), grad: tensor([ 0.0159, 0.0198, 0.0463, -0.0345, -0.0121, 0.0160, -0.0397, -0.0027, + -0.0250, 0.0160], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 230.58, cls_loss 1.2282 cls_loss_mapping 0.1739 cls_loss_causal 1.0582 re_mapping 0.0539 re_causal 0.1088 /// teacc 95.63 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0057, -0.0129, -0.0408, ..., -0.0041, 0.0390, -0.0515], + [-0.0170, -0.0458, -0.0095, ..., 0.0659, -0.0375, 0.0833], + [ 0.0258, -0.0141, -0.0004, ..., -0.0081, -0.0100, -0.0288], + ..., + [ 0.0108, -0.0304, 0.0665, ..., 0.0019, -0.0101, 0.0122], + [ 0.0085, 0.0306, -0.0331, ..., -0.0210, -0.0285, -0.0234], + [-0.0234, 0.0333, 0.0402, ..., -0.0335, -0.0142, -0.0002]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0033, 0.0039, ..., -0.0014, -0.0049, -0.0007], + [ 0.0000, 0.0078, 0.0257, ..., 0.0158, 0.0012, 0.0152], + [ 0.0000, 0.0077, 0.0300, ..., 0.0060, 0.0025, 0.0079], + ..., + [ 0.0000, -0.0022, -0.0633, ..., -0.0396, -0.0046, -0.0187], + [ 0.0000, 0.0030, 0.0149, ..., 0.0127, 0.0014, 0.0171], + [ 0.0000, -0.0025, -0.0137, ..., 0.0024, 0.0015, -0.0148]], + device='cuda:0') +Epoch 13, bias, value: tensor([-0.0088, 0.0184, -0.0041, -0.0111, 0.0161, -0.0322, 0.0133, -0.0025, + 0.0044, -0.0093], device='cuda:0'), grad: tensor([-0.0602, 0.0477, 0.0446, 0.0307, 0.0518, -0.0677, 0.0133, -0.0789, + 0.0342, -0.0156], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 230.53, cls_loss 1.1920 cls_loss_mapping 0.1679 cls_loss_causal 1.0173 re_mapping 0.0519 re_causal 0.1088 /// teacc 96.06 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0057, -0.0138, -0.0416, ..., -0.0056, 0.0381, -0.0526], + [-0.0170, -0.0460, -0.0109, ..., 0.0668, -0.0372, 0.0850], + [ 0.0258, -0.0144, -0.0007, ..., -0.0082, -0.0096, -0.0303], + ..., + [ 0.0108, -0.0328, 0.0678, ..., 0.0019, -0.0110, 0.0136], + [ 0.0085, 0.0313, -0.0343, ..., -0.0217, -0.0274, -0.0247], + [-0.0234, 0.0348, 0.0406, ..., -0.0343, -0.0148, -0.0010]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0042, 0.0042, ..., -0.0118, 0.0015, -0.0078], + [ 0.0000, 0.0026, 0.0036, ..., 0.0144, 0.0029, 0.0218], + [ 0.0000, 0.0082, 0.0061, ..., 0.0040, 0.0016, 0.0087], + ..., + [ 0.0000, -0.0115, -0.0224, ..., -0.0204, 0.0011, -0.0213], + [ 0.0000, -0.0094, -0.0034, ..., -0.0094, -0.0033, 0.0101], + [ 0.0000, 0.0012, -0.0025, ..., 0.0078, 0.0017, -0.0030]], + device='cuda:0') +Epoch 14, bias, value: tensor([-0.0095, 0.0185, -0.0039, -0.0108, 0.0162, -0.0323, 0.0142, -0.0030, + 0.0044, -0.0098], device='cuda:0'), grad: tensor([-0.0181, 0.0347, 0.0071, 0.0473, -0.0150, -0.0280, 0.0194, -0.0493, + 0.0052, -0.0033], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 230.63, cls_loss 1.1858 cls_loss_mapping 0.1477 cls_loss_causal 1.0152 re_mapping 0.0511 re_causal 0.1113 /// teacc 96.25 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0057, -0.0155, -0.0430, ..., -0.0066, 0.0385, -0.0534], + [-0.0170, -0.0459, -0.0108, ..., 0.0676, -0.0366, 0.0863], + [ 0.0258, -0.0156, -0.0010, ..., -0.0085, -0.0096, -0.0312], + ..., + [ 0.0108, -0.0330, 0.0696, ..., 0.0022, -0.0112, 0.0150], + [ 0.0085, 0.0314, -0.0368, ..., -0.0222, -0.0269, -0.0263], + [-0.0234, 0.0360, 0.0413, ..., -0.0347, -0.0149, -0.0010]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.5727e-03, -1.9043e-02, ..., -1.8387e-02, + -2.7065e-03, -2.5467e-02], + [ 0.0000e+00, -2.5833e-02, -3.7048e-02, ..., -2.6443e-02, + 1.2922e-03, -1.9821e-02], + [ 0.0000e+00, -4.6577e-03, -1.8711e-03, ..., -6.5804e-03, + 4.2439e-04, -1.4397e-02], + ..., + [ 0.0000e+00, -3.8681e-03, 2.8992e-03, ..., 3.1616e-02, + 5.3167e-04, 1.9501e-02], + [ 0.0000e+00, 1.0986e-02, 2.5925e-02, ..., 2.3300e-02, + 1.7667e-04, 2.0370e-02], + [ 0.0000e+00, 2.9640e-03, 1.1894e-02, ..., 7.7095e-03, + 8.2433e-05, 1.8091e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0094, 0.0183, -0.0040, -0.0110, 0.0161, -0.0323, 0.0146, -0.0031, + 0.0047, -0.0099], device='cuda:0'), grad: tensor([-0.0427, -0.0552, -0.0231, -0.0127, 0.0011, 0.0015, 0.0099, 0.0356, + 0.0691, 0.0164], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 214.51, cls_loss 1.1437 cls_loss_mapping 0.1443 cls_loss_causal 0.9690 re_mapping 0.0482 re_causal 0.0995 /// teacc 95.87 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0057, -0.0166, -0.0445, ..., -0.0059, 0.0382, -0.0527], + [-0.0170, -0.0460, -0.0104, ..., 0.0684, -0.0368, 0.0880], + [ 0.0258, -0.0156, -0.0003, ..., -0.0087, -0.0087, -0.0315], + ..., + [ 0.0108, -0.0350, 0.0702, ..., 0.0021, -0.0116, 0.0152], + [ 0.0085, 0.0316, -0.0387, ..., -0.0227, -0.0274, -0.0276], + [-0.0234, 0.0376, 0.0421, ..., -0.0365, -0.0150, -0.0019]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4120e-03, 5.3749e-03, ..., 5.5981e-04, + 9.0551e-04, -1.1273e-05], + [ 0.0000e+00, -9.0456e-04, -3.5858e-03, ..., -4.3716e-03, + 1.2474e-03, 7.6256e-03], + [ 0.0000e+00, 1.6785e-02, 2.5543e-02, ..., 1.3115e-02, + 4.5662e-03, 1.0201e-02], + ..., + [ 0.0000e+00, -8.8274e-05, -2.0072e-05, ..., -1.6464e-02, + 1.8263e-03, -4.7836e-03], + [ 0.0000e+00, -2.5635e-02, -3.0624e-02, ..., -1.7853e-02, + -4.3221e-03, -2.0847e-03], + [ 0.0000e+00, 1.1971e-02, 8.2169e-03, ..., 6.9389e-03, + 1.3828e-03, 4.3526e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0098, 0.0189, -0.0036, -0.0106, 0.0163, -0.0324, 0.0145, -0.0034, + 0.0043, -0.0102], device='cuda:0'), grad: tensor([ 0.0199, 0.0127, 0.0422, 0.0016, 0.0079, -0.0252, 0.0194, -0.0428, + -0.0649, 0.0292], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 230.86, cls_loss 1.0933 cls_loss_mapping 0.1370 cls_loss_causal 0.9343 re_mapping 0.0453 re_causal 0.0968 /// teacc 96.33 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0057, -0.0175, -0.0456, ..., -0.0074, 0.0383, -0.0527], + [-0.0170, -0.0457, -0.0113, ..., 0.0692, -0.0375, 0.0893], + [ 0.0258, -0.0159, -0.0008, ..., -0.0096, -0.0079, -0.0328], + ..., + [ 0.0108, -0.0371, 0.0709, ..., 0.0021, -0.0132, 0.0162], + [ 0.0085, 0.0316, -0.0405, ..., -0.0228, -0.0274, -0.0289], + [-0.0234, 0.0388, 0.0421, ..., -0.0369, -0.0138, -0.0026]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0085, 0.0079, ..., 0.0090, -0.0102, 0.0067], + [ 0.0000, 0.0074, 0.0085, ..., 0.0076, 0.0026, 0.0045], + [ 0.0000, 0.0162, 0.0376, ..., 0.0025, 0.0021, 0.0046], + ..., + [ 0.0000, -0.0462, -0.1125, ..., -0.0179, -0.0059, 0.0003], + [ 0.0000, 0.0057, -0.0008, ..., 0.0112, -0.0086, -0.0024], + [ 0.0000, 0.0005, 0.0062, ..., -0.0068, 0.0044, -0.0423]], + device='cuda:0') +Epoch 17, bias, value: tensor([-0.0098, 0.0189, -0.0037, -0.0098, 0.0165, -0.0326, 0.0142, -0.0038, + 0.0041, -0.0100], device='cuda:0'), grad: tensor([ 0.0194, 0.0386, 0.0330, -0.0646, 0.0542, 0.0101, 0.0228, -0.0846, + 0.0230, -0.0520], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 214.63, cls_loss 1.0821 cls_loss_mapping 0.1310 cls_loss_causal 0.9099 re_mapping 0.0450 re_causal 0.0989 /// teacc 96.17 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0057, -0.0190, -0.0462, ..., -0.0079, 0.0392, -0.0534], + [-0.0170, -0.0461, -0.0110, ..., 0.0694, -0.0382, 0.0905], + [ 0.0258, -0.0165, -0.0017, ..., -0.0089, -0.0077, -0.0332], + ..., + [ 0.0108, -0.0370, 0.0720, ..., 0.0032, -0.0132, 0.0168], + [ 0.0085, 0.0322, -0.0408, ..., -0.0226, -0.0282, -0.0296], + [-0.0234, 0.0386, 0.0421, ..., -0.0376, -0.0144, -0.0027]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0029, 0.0007, ..., -0.0165, 0.0024, 0.0107], + [ 0.0000, 0.0082, 0.0187, ..., 0.0074, 0.0010, 0.0063], + [ 0.0000, -0.0102, -0.0046, ..., -0.0076, -0.0107, -0.0089], + ..., + [ 0.0000, -0.0078, -0.0259, ..., -0.0158, 0.0003, -0.0314], + [ 0.0000, 0.0151, 0.0052, ..., 0.0062, 0.0057, 0.0047], + [ 0.0000, 0.0112, -0.0048, ..., 0.0020, 0.0008, 0.0064]], + device='cuda:0') +Epoch 18, bias, value: tensor([-0.0101, 0.0190, -0.0040, -0.0098, 0.0165, -0.0323, 0.0143, -0.0038, + 0.0043, -0.0101], device='cuda:0'), grad: tensor([-0.0100, 0.0122, -0.0394, -0.0069, 0.0207, 0.0092, 0.0261, -0.0874, + 0.0461, 0.0295], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 230.64, cls_loss 1.0623 cls_loss_mapping 0.1226 cls_loss_causal 0.9100 re_mapping 0.0436 re_causal 0.0967 /// teacc 96.38 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0062, -0.0209, -0.0462, ..., -0.0081, 0.0395, -0.0536], + [-0.0182, -0.0464, -0.0110, ..., 0.0698, -0.0375, 0.0915], + [ 0.0274, -0.0155, -0.0006, ..., -0.0096, -0.0070, -0.0336], + ..., + [ 0.0112, -0.0378, 0.0727, ..., 0.0043, -0.0146, 0.0181], + [ 0.0074, 0.0322, -0.0423, ..., -0.0235, -0.0272, -0.0309], + [-0.0239, 0.0394, 0.0419, ..., -0.0380, -0.0142, -0.0032]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.5343e-03, -5.5008e-03, ..., -9.9869e-03, + 2.7447e-03, 5.4693e-04], + [ 0.0000e+00, 1.3298e-02, 7.5188e-03, ..., 5.1498e-03, + 1.0850e-07, -3.5334e-04], + [ 0.0000e+00, 3.6316e-03, 1.0252e-03, ..., 3.5114e-03, + 7.2177e-08, 3.8204e-03], + ..., + [ 0.0000e+00, 4.3869e-03, 8.7070e-04, ..., 8.8196e-03, + 1.8626e-09, 1.6876e-02], + [ 0.0000e+00, 4.7607e-03, -6.4325e-04, ..., 1.8148e-03, + 6.9580e-03, -3.2349e-03], + [ 0.0000e+00, 1.3054e-02, 2.0004e-02, ..., 5.4703e-03, + 4.0792e-06, -1.2764e-02]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0100, 0.0188, -0.0035, -0.0103, 0.0163, -0.0321, 0.0137, -0.0030, + 0.0045, -0.0104], device='cuda:0'), grad: tensor([-0.0400, 0.0367, 0.0092, 0.0061, -0.0376, 0.0064, 0.0059, 0.0210, + -0.0187, 0.0112], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 230.69, cls_loss 1.0420 cls_loss_mapping 0.1086 cls_loss_causal 0.8787 re_mapping 0.0426 re_causal 0.0909 /// teacc 96.77 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0088, -0.0225, -0.0456, ..., -0.0079, 0.0388, -0.0535], + [-0.0195, -0.0476, -0.0122, ..., 0.0712, -0.0369, 0.0927], + [ 0.0291, -0.0159, -0.0015, ..., -0.0107, -0.0078, -0.0344], + ..., + [ 0.0118, -0.0388, 0.0747, ..., 0.0049, -0.0148, 0.0190], + [ 0.0052, 0.0329, -0.0426, ..., -0.0252, -0.0270, -0.0314], + [-0.0235, 0.0399, 0.0411, ..., -0.0374, -0.0132, -0.0037]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.4325e-04, -3.8671e-04, ..., -1.1082e-03, + 1.7858e-04, 1.0538e-03], + [ 0.0000e+00, -1.3847e-02, -9.8190e-03, ..., -1.5945e-02, + 2.0385e-05, -8.1711e-03], + [ 0.0000e+00, -1.8555e-02, -1.5583e-03, ..., 6.3095e-03, + -4.2707e-05, 2.5063e-03], + ..., + [ 0.0000e+00, -9.4528e-03, -1.6235e-02, ..., 2.6264e-03, + -3.6597e-04, 1.9102e-03], + [ 0.0000e+00, 1.4061e-02, 1.5316e-03, ..., 7.0152e-03, + -7.3147e-04, 4.2725e-03], + [ 0.0000e+00, 7.2060e-03, -1.3485e-03, ..., 1.1053e-03, + 7.0333e-05, 4.8180e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0104, 0.0189, -0.0034, -0.0098, 0.0160, -0.0315, 0.0133, -0.0026, + 0.0043, -0.0108], device='cuda:0'), grad: tensor([-0.0082, -0.0535, 0.0033, 0.0564, 0.0372, -0.0007, -0.0421, -0.0018, + -0.0070, 0.0166], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 230.74, cls_loss 1.0194 cls_loss_mapping 0.1015 cls_loss_causal 0.8567 re_mapping 0.0432 re_causal 0.0896 /// teacc 97.10 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0094, -0.0237, -0.0466, ..., -0.0081, 0.0378, -0.0540], + [-0.0200, -0.0471, -0.0128, ..., 0.0715, -0.0363, 0.0935], + [ 0.0284, -0.0169, -0.0014, ..., -0.0108, -0.0076, -0.0347], + ..., + [ 0.0110, -0.0395, 0.0752, ..., 0.0051, -0.0159, 0.0188], + [ 0.0053, 0.0333, -0.0436, ..., -0.0259, -0.0258, -0.0316], + [-0.0246, 0.0416, 0.0414, ..., -0.0381, -0.0122, -0.0035]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0097, -0.0069, ..., -0.0188, -0.0145, 0.0012], + [ 0.0000, 0.0280, 0.0170, ..., 0.0364, 0.0003, 0.0051], + [ 0.0000, 0.0032, 0.0022, ..., -0.0003, 0.0002, 0.0007], + ..., + [ 0.0000, -0.0147, -0.0228, ..., -0.0197, -0.0029, -0.0023], + [ 0.0000, 0.0015, 0.0004, ..., 0.0051, 0.0008, 0.0020], + [ 0.0000, 0.0178, 0.0230, ..., 0.0195, 0.0012, 0.0032]], + device='cuda:0') +Epoch 21, bias, value: tensor([-0.0105, 0.0191, -0.0038, -0.0098, 0.0168, -0.0312, 0.0135, -0.0026, + 0.0038, -0.0112], device='cuda:0'), grad: tensor([-0.0828, 0.0558, 0.0151, 0.0014, -0.0040, 0.0023, -0.0119, -0.0461, + 0.0113, 0.0590], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 214.59, cls_loss 1.0290 cls_loss_mapping 0.1070 cls_loss_causal 0.8829 re_mapping 0.0406 re_causal 0.0875 /// teacc 96.63 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0103, -0.0236, -0.0462, ..., -0.0069, 0.0383, -0.0528], + [-0.0197, -0.0470, -0.0131, ..., 0.0723, -0.0357, 0.0946], + [ 0.0265, -0.0165, -0.0020, ..., -0.0110, -0.0074, -0.0355], + ..., + [ 0.0069, -0.0404, 0.0750, ..., 0.0043, -0.0159, 0.0187], + [ 0.0037, 0.0346, -0.0435, ..., -0.0268, -0.0253, -0.0320], + [-0.0240, 0.0414, 0.0420, ..., -0.0386, -0.0150, -0.0037]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0089, 0.0079, ..., 0.0039, 0.0263, 0.0096], + [ 0.0000, -0.0002, -0.0003, ..., -0.0027, 0.0025, 0.0010], + [ 0.0000, -0.0006, -0.0010, ..., -0.0009, -0.0009, -0.0001], + ..., + [ 0.0000, -0.0136, -0.0176, ..., -0.0044, -0.0296, -0.0146], + [ 0.0000, -0.0081, -0.0061, ..., -0.0002, -0.0052, -0.0060], + [ 0.0000, 0.0163, 0.0134, ..., 0.0033, 0.0091, 0.0064]], + device='cuda:0') +Epoch 22, bias, value: tensor([-0.0098, 0.0195, -0.0036, -0.0094, 0.0165, -0.0313, 0.0126, -0.0033, + 0.0035, -0.0108], device='cuda:0'), grad: tensor([ 0.0441, -0.0061, -0.0102, -0.0086, 0.0300, 0.0080, 0.0011, -0.0500, + -0.0302, 0.0219], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 214.43, cls_loss 0.9815 cls_loss_mapping 0.0940 cls_loss_causal 0.8416 re_mapping 0.0407 re_causal 0.0898 /// teacc 96.77 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0111, -0.0248, -0.0471, ..., -0.0072, 0.0382, -0.0530], + [-0.0194, -0.0476, -0.0141, ..., 0.0722, -0.0358, 0.0946], + [ 0.0249, -0.0171, -0.0028, ..., -0.0103, -0.0082, -0.0359], + ..., + [ 0.0081, -0.0398, 0.0752, ..., 0.0044, -0.0165, 0.0197], + [ 0.0013, 0.0350, -0.0447, ..., -0.0273, -0.0252, -0.0324], + [-0.0256, 0.0412, 0.0427, ..., -0.0393, -0.0148, -0.0042]], + device='cuda:0'), grad: tensor([[ 1.1809e-05, 4.4554e-05, 2.0004e-02, ..., 1.1124e-02, + 1.0506e-02, 1.3714e-03], + [ 5.4315e-06, 2.6608e-03, 8.5144e-03, ..., 9.2773e-03, + 3.6297e-03, 7.7515e-03], + [ 3.1859e-05, -7.5760e-03, -1.9531e-02, ..., -1.0475e-02, + 1.7792e-02, -5.7526e-03], + ..., + [-2.3127e-04, 1.9665e-03, -4.6234e-03, ..., 2.4910e-03, + 1.8568e-03, -2.3327e-03], + [ 2.6926e-05, 8.2588e-04, 5.3101e-03, ..., 2.5654e-03, + -7.1564e-03, 3.8376e-03], + [ 1.1361e-04, 3.3321e-03, 1.8311e-02, ..., 1.1139e-02, + 3.1815e-03, 6.0844e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0098, 0.0193, -0.0028, -0.0092, 0.0165, -0.0310, 0.0127, -0.0034, + 0.0031, -0.0113], device='cuda:0'), grad: tensor([ 0.0338, 0.0195, -0.0182, 0.0045, -0.0589, -0.0193, -0.0021, 0.0158, + -0.0052, 0.0301], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 214.11, cls_loss 0.9876 cls_loss_mapping 0.0873 cls_loss_causal 0.8426 re_mapping 0.0389 re_causal 0.0871 /// teacc 97.00 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0128, -0.0252, -0.0478, ..., -0.0073, 0.0380, -0.0538], + [-0.0204, -0.0483, -0.0143, ..., 0.0728, -0.0356, 0.0957], + [ 0.0205, -0.0175, -0.0032, ..., -0.0096, -0.0072, -0.0362], + ..., + [ 0.0105, -0.0406, 0.0764, ..., 0.0045, -0.0171, 0.0201], + [ 0.0021, 0.0350, -0.0465, ..., -0.0277, -0.0249, -0.0335], + [-0.0266, 0.0416, 0.0425, ..., -0.0391, -0.0146, -0.0043]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0057, 0.0129, ..., -0.0012, -0.0036, 0.0022], + [ 0.0000, -0.0018, -0.0063, ..., -0.0128, -0.0008, -0.0021], + [ 0.0000, 0.0104, 0.0149, ..., 0.0092, 0.0043, 0.0062], + ..., + [ 0.0000, 0.0009, 0.0012, ..., 0.0077, 0.0004, 0.0057], + [ 0.0000, 0.0155, 0.0144, ..., 0.0105, 0.0013, 0.0058], + [ 0.0000, 0.0011, -0.0094, ..., 0.0015, -0.0017, -0.0030]], + device='cuda:0') +Epoch 24, bias, value: tensor([-0.0102, 0.0195, -0.0025, -0.0089, 0.0160, -0.0307, 0.0122, -0.0030, + 0.0027, -0.0111], device='cuda:0'), grad: tensor([ 0.0138, -0.0347, 0.0408, -0.0825, 0.0094, 0.0108, 0.0262, 0.0062, + 0.0193, -0.0093], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 214.21, cls_loss 0.9664 cls_loss_mapping 0.0840 cls_loss_causal 0.8152 re_mapping 0.0386 re_causal 0.0818 /// teacc 96.81 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0125, -0.0258, -0.0487, ..., -0.0075, 0.0382, -0.0546], + [-0.0192, -0.0492, -0.0147, ..., 0.0734, -0.0366, 0.0971], + [ 0.0207, -0.0187, -0.0035, ..., -0.0095, -0.0074, -0.0374], + ..., + [ 0.0102, -0.0408, 0.0771, ..., 0.0044, -0.0170, 0.0204], + [ 0.0020, 0.0352, -0.0467, ..., -0.0278, -0.0248, -0.0334], + [-0.0270, 0.0424, 0.0427, ..., -0.0394, -0.0145, -0.0043]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0036, 0.0054, ..., 0.0078, 0.0008, 0.0013], + [ 0.0000, 0.0062, -0.0013, ..., 0.0048, 0.0004, 0.0017], + [ 0.0000, 0.0063, 0.0092, ..., 0.0187, -0.0002, 0.0025], + ..., + [ 0.0000, 0.0107, 0.0070, ..., 0.0115, -0.0070, 0.0020], + [ 0.0000, -0.0223, -0.0242, ..., -0.0467, 0.0014, -0.0077], + [ 0.0000, -0.0095, -0.0074, ..., 0.0117, 0.0013, 0.0013]], + device='cuda:0') +Epoch 25, bias, value: tensor([-0.0106, 0.0193, -0.0025, -0.0087, 0.0162, -0.0308, 0.0121, -0.0029, + 0.0032, -0.0115], device='cuda:0'), grad: tensor([-0.0025, 0.0294, 0.0336, 0.0500, -0.0493, -0.0283, 0.0104, 0.0058, + -0.0548, 0.0058], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 230.49, cls_loss 0.9551 cls_loss_mapping 0.0811 cls_loss_causal 0.8176 re_mapping 0.0376 re_causal 0.0824 /// teacc 97.27 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0135, -0.0259, -0.0490, ..., -0.0086, 0.0394, -0.0538], + [-0.0195, -0.0493, -0.0151, ..., 0.0735, -0.0366, 0.0975], + [ 0.0216, -0.0185, -0.0039, ..., -0.0092, -0.0052, -0.0372], + ..., + [ 0.0106, -0.0423, 0.0774, ..., 0.0040, -0.0182, 0.0211], + [ 0.0014, 0.0363, -0.0477, ..., -0.0277, -0.0251, -0.0349], + [-0.0273, 0.0431, 0.0440, ..., -0.0383, -0.0152, -0.0041]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0109, -0.0156, ..., -0.0074, -0.0019, 0.0018], + [ 0.0000, -0.0055, -0.0195, ..., -0.0180, 0.0010, -0.0117], + [ 0.0000, 0.0020, 0.0095, ..., 0.0056, -0.0117, 0.0025], + ..., + [ 0.0000, -0.0006, 0.0079, ..., 0.0131, -0.0071, -0.0054], + [ 0.0000, 0.0091, 0.0055, ..., -0.0089, 0.0030, 0.0025], + [ 0.0000, 0.0074, 0.0182, ..., 0.0065, 0.0089, 0.0090]], + device='cuda:0') +Epoch 26, bias, value: tensor([-0.0108, 0.0193, -0.0024, -0.0092, 0.0165, -0.0310, 0.0124, -0.0028, + 0.0032, -0.0111], device='cuda:0'), grad: tensor([-0.0511, -0.0464, 0.0230, 0.0404, -0.0549, 0.0044, 0.0345, 0.0228, + -0.0020, 0.0293], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 230.98, cls_loss 0.9691 cls_loss_mapping 0.0824 cls_loss_causal 0.8262 re_mapping 0.0366 re_causal 0.0807 /// teacc 97.41 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0138, -0.0268, -0.0492, ..., -0.0089, 0.0395, -0.0531], + [-0.0221, -0.0488, -0.0151, ..., 0.0742, -0.0374, 0.0985], + [ 0.0192, -0.0191, -0.0037, ..., -0.0085, -0.0069, -0.0369], + ..., + [ 0.0075, -0.0433, 0.0773, ..., 0.0030, -0.0186, 0.0209], + [ 0.0019, 0.0366, -0.0469, ..., -0.0285, -0.0236, -0.0352], + [-0.0290, 0.0439, 0.0442, ..., -0.0383, -0.0156, -0.0046]], + device='cuda:0'), grad: tensor([[ 8.4543e-04, 1.4610e-03, -8.3923e-03, ..., -6.9313e-03, + 5.7487e-03, -7.3624e-04], + [ 1.0437e-04, 1.1818e-02, -2.7714e-03, ..., 3.0243e-02, + 4.5204e-04, 7.5340e-03], + [-3.1796e-03, -2.3407e-02, -5.7907e-03, ..., -3.7659e-02, + -1.0120e-01, 1.0395e-03], + ..., + [ 1.4496e-04, 5.3215e-04, -7.8659e-03, ..., -6.7673e-03, + 9.4652e-05, -1.0729e-03], + [ 2.8872e-04, 2.0477e-02, 7.7896e-03, ..., -3.1490e-03, + -2.9068e-03, -1.0712e-02], + [ 1.3089e-04, -1.6968e-02, -1.0582e-02, ..., -2.0866e-03, + 2.9469e-03, -3.0613e-03]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0106, 0.0193, -0.0019, -0.0088, 0.0168, -0.0307, 0.0114, -0.0031, + 0.0030, -0.0114], device='cuda:0'), grad: tensor([-0.0053, -0.0015, -0.0857, 0.0587, 0.0111, 0.0262, 0.0207, -0.0121, + -0.0034, -0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 214.49, cls_loss 0.9388 cls_loss_mapping 0.0804 cls_loss_causal 0.8093 re_mapping 0.0360 re_causal 0.0814 /// teacc 96.99 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0120, -0.0272, -0.0486, ..., -0.0088, 0.0386, -0.0528], + [-0.0232, -0.0489, -0.0150, ..., 0.0744, -0.0373, 0.0995], + [ 0.0182, -0.0191, -0.0045, ..., -0.0090, -0.0063, -0.0376], + ..., + [ 0.0059, -0.0443, 0.0777, ..., 0.0035, -0.0185, 0.0218], + [-0.0003, 0.0360, -0.0481, ..., -0.0285, -0.0216, -0.0356], + [-0.0291, 0.0448, 0.0448, ..., -0.0387, -0.0166, -0.0042]], + device='cuda:0'), grad: tensor([[ 3.8892e-05, -4.3654e-04, -1.0918e-02, ..., -1.8234e-02, + -8.0643e-03, -1.1902e-02], + [ 2.7150e-05, 4.4632e-03, 1.0307e-02, ..., 1.4847e-02, + -2.0428e-03, -2.9392e-03], + [-4.9782e-04, 2.3174e-03, 4.5090e-03, ..., 3.7308e-03, + -4.1122e-03, 6.1378e-03], + ..., + [-5.3883e-04, 1.2331e-03, 3.6983e-03, ..., 1.4854e-02, + 7.9727e-03, 3.9215e-03], + [ 2.3139e-04, -3.4733e-03, -1.6388e-02, ..., -1.6174e-02, + -6.5918e-03, -1.0490e-02], + [ 1.0127e-04, 4.6616e-03, -2.7275e-03, ..., -6.8512e-03, + 1.9894e-03, 3.7746e-03]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0104, 0.0194, -0.0026, -0.0089, 0.0170, -0.0305, 0.0117, -0.0027, + 0.0026, -0.0115], device='cuda:0'), grad: tensor([-0.0611, 0.0518, 0.0084, 0.0106, 0.0463, 0.0008, 0.0115, 0.0267, + -0.0682, -0.0268], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 214.45, cls_loss 0.9572 cls_loss_mapping 0.0949 cls_loss_causal 0.8101 re_mapping 0.0345 re_causal 0.0707 /// teacc 96.91 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0130, -0.0281, -0.0485, ..., -0.0091, 0.0388, -0.0527], + [-0.0237, -0.0496, -0.0159, ..., 0.0742, -0.0398, 0.1004], + [ 0.0183, -0.0196, -0.0050, ..., -0.0090, -0.0064, -0.0387], + ..., + [ 0.0039, -0.0440, 0.0783, ..., 0.0039, -0.0201, 0.0217], + [-0.0022, 0.0364, -0.0490, ..., -0.0286, -0.0202, -0.0362], + [-0.0309, 0.0450, 0.0442, ..., -0.0384, -0.0159, -0.0045]], + device='cuda:0'), grad: tensor([[ 2.2464e-06, 1.1988e-03, -1.1492e-03, ..., -3.0918e-03, + 5.3650e-02, -8.2254e-05], + [ 2.5868e-04, 3.5248e-03, 4.2000e-03, ..., 1.2131e-02, + 2.2984e-03, 1.1978e-02], + [-2.6751e-04, 2.2068e-03, 2.7752e-03, ..., 3.3016e-03, + -1.3428e-02, 2.5120e-03], + ..., + [-5.2166e-04, 3.1300e-03, 2.6817e-03, ..., -1.2817e-03, + 2.2186e-02, -5.1079e-03], + [ 6.2466e-05, 2.5024e-02, 1.6571e-02, ..., 6.2485e-03, + 2.8610e-02, 4.6005e-03], + [ 2.1446e-04, -4.6204e-02, -5.7983e-02, ..., -2.7847e-02, + -5.5115e-02, -2.4673e-02]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0099, 0.0187, -0.0022, -0.0087, 0.0177, -0.0309, 0.0112, -0.0033, + 0.0027, -0.0113], device='cuda:0'), grad: tensor([ 0.0046, 0.0255, -0.0058, 0.0426, 0.0173, -0.0009, -0.0399, 0.0182, + 0.0657, -0.1273], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 230.65, cls_loss 0.9483 cls_loss_mapping 0.0837 cls_loss_causal 0.8046 re_mapping 0.0352 re_causal 0.0739 /// teacc 97.43 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0156, -0.0292, -0.0495, ..., -0.0102, 0.0387, -0.0537], + [-0.0220, -0.0501, -0.0152, ..., 0.0749, -0.0393, 0.1017], + [ 0.0187, -0.0202, -0.0059, ..., -0.0094, -0.0055, -0.0399], + ..., + [ 0.0012, -0.0446, 0.0790, ..., 0.0042, -0.0205, 0.0221], + [-0.0029, 0.0366, -0.0498, ..., -0.0290, -0.0215, -0.0363], + [-0.0344, 0.0452, 0.0450, ..., -0.0385, -0.0174, -0.0043]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.2278e-03, -1.3008e-03, ..., -1.6146e-03, + 4.8846e-05, 1.7300e-03], + [ 0.0000e+00, 2.1992e-03, -3.6373e-03, ..., -1.1569e-04, + -4.0531e-06, 4.4212e-03], + [ 0.0000e+00, 1.3590e-03, 1.9634e-04, ..., 3.4428e-04, + 6.7465e-06, -1.2840e-02], + ..., + [ 0.0000e+00, 2.2430e-03, 1.7014e-03, ..., 4.1084e-03, + 7.9051e-06, 2.7809e-03], + [ 0.0000e+00, -4.2648e-03, 3.0022e-03, ..., 2.1286e-03, + 2.2721e-04, 5.1270e-03], + [ 0.0000e+00, 4.3602e-03, 3.1433e-03, ..., -2.2297e-03, + 3.5912e-05, 6.7444e-03]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0102, 0.0186, -0.0017, -0.0083, 0.0172, -0.0303, 0.0114, -0.0034, + 0.0022, -0.0113], device='cuda:0'), grad: tensor([-0.0360, 0.0242, -0.0216, -0.0219, 0.0151, -0.0102, 0.0446, -0.0141, + -0.0143, 0.0341], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 230.37, cls_loss 0.9240 cls_loss_mapping 0.0825 cls_loss_causal 0.7768 re_mapping 0.0336 re_causal 0.0721 /// teacc 97.51 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0176, -0.0295, -0.0493, ..., -0.0102, 0.0394, -0.0544], + [-0.0230, -0.0508, -0.0158, ..., 0.0746, -0.0389, 0.1014], + [ 0.0176, -0.0209, -0.0071, ..., -0.0094, -0.0059, -0.0396], + ..., + [ 0.0020, -0.0451, 0.0801, ..., 0.0034, -0.0198, 0.0235], + [-0.0033, 0.0368, -0.0509, ..., -0.0286, -0.0217, -0.0381], + [-0.0318, 0.0456, 0.0459, ..., -0.0377, -0.0182, -0.0042]], + device='cuda:0'), grad: tensor([[ 1.3396e-05, 4.5395e-03, 4.1618e-03, ..., 5.3177e-03, + 1.4114e-03, 1.8158e-03], + [ 3.3677e-06, 7.4310e-03, 9.7580e-03, ..., 1.3855e-02, + 2.4300e-03, 6.7596e-03], + [ 1.7226e-05, -3.7136e-03, -8.7357e-03, ..., -1.3023e-02, + -1.4496e-03, -8.0032e-03], + ..., + [ 4.6231e-06, -1.2032e-02, -2.1118e-02, ..., -2.0432e-02, + -3.5343e-03, -4.3297e-03], + [-4.6325e-04, 3.2082e-03, 1.8044e-03, ..., -2.8629e-03, + 3.4809e-03, 1.7157e-03], + [ 2.3019e-04, 2.2690e-02, 2.5589e-02, ..., 1.7487e-02, + 6.7139e-03, 1.4244e-02]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0103, 0.0186, -0.0022, -0.0081, 0.0168, -0.0307, 0.0114, -0.0036, + 0.0029, -0.0108], device='cuda:0'), grad: tensor([ 0.0209, 0.0385, -0.0432, 0.0525, 0.0174, -0.0623, -0.0567, -0.0388, + -0.0026, 0.0744], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 230.43, cls_loss 0.9209 cls_loss_mapping 0.0861 cls_loss_causal 0.7888 re_mapping 0.0324 re_causal 0.0695 /// teacc 97.55 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0196, -0.0298, -0.0501, ..., -0.0092, 0.0398, -0.0551], + [-0.0212, -0.0508, -0.0156, ..., 0.0752, -0.0384, 0.1018], + [ 0.0176, -0.0222, -0.0063, ..., -0.0101, -0.0060, -0.0398], + ..., + [ 0.0017, -0.0454, 0.0795, ..., 0.0021, -0.0213, 0.0228], + [-0.0044, 0.0375, -0.0509, ..., -0.0286, -0.0216, -0.0377], + [-0.0340, 0.0459, 0.0465, ..., -0.0367, -0.0186, -0.0038]], + device='cuda:0'), grad: tensor([[ 8.0094e-08, -2.4853e-03, 2.7008e-03, ..., -8.0795e-03, + 3.4618e-03, 1.3933e-03], + [ 1.2107e-07, -1.0399e-02, -4.7951e-03, ..., -2.4796e-02, + -2.0630e-02, -1.8219e-02], + [-8.8476e-07, 6.8378e-04, 6.1722e-03, ..., 7.1030e-03, + 5.0697e-03, 1.3618e-03], + ..., + [ 1.2733e-05, 6.2866e-03, -8.1329e-03, ..., 9.3031e-04, + 7.2861e-04, 5.4245e-03], + [ 4.6100e-07, 2.9526e-03, 2.3937e-03, ..., 1.2894e-03, + 1.9550e-03, -4.8218e-03], + [ 1.8673e-06, 3.1638e-04, 8.4763e-03, ..., 1.6434e-02, + -4.4417e-04, 1.0506e-02]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0098, 0.0187, -0.0019, -0.0084, 0.0165, -0.0303, 0.0113, -0.0041, + 0.0030, -0.0110], device='cuda:0'), grad: tensor([-0.0110, -0.0691, 0.0233, -0.0019, 0.0732, -0.0158, -0.0031, 0.0237, + 0.0036, -0.0229], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 214.29, cls_loss 0.8799 cls_loss_mapping 0.0645 cls_loss_causal 0.7468 re_mapping 0.0331 re_causal 0.0723 /// teacc 97.42 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0191, -0.0307, -0.0506, ..., -0.0097, 0.0387, -0.0554], + [-0.0206, -0.0505, -0.0158, ..., 0.0754, -0.0371, 0.1025], + [ 0.0174, -0.0219, -0.0062, ..., -0.0101, -0.0058, -0.0394], + ..., + [ 0.0021, -0.0448, 0.0790, ..., 0.0023, -0.0224, 0.0225], + [-0.0042, 0.0379, -0.0507, ..., -0.0287, -0.0226, -0.0376], + [-0.0346, 0.0463, 0.0469, ..., -0.0370, -0.0169, -0.0041]], + device='cuda:0'), grad: tensor([[ 9.3132e-08, 2.8305e-03, 3.9864e-03, ..., 6.4964e-03, + 8.0585e-04, 1.3905e-03], + [ 3.3434e-07, 1.5187e-04, 8.9931e-04, ..., 9.6560e-04, + 6.8283e-04, -2.9392e-03], + [ 2.4475e-06, 5.0497e-04, -7.8087e-03, ..., 2.5158e-03, + 4.9543e-04, 1.4973e-03], + ..., + [-2.3201e-05, 1.8568e-03, -7.4768e-03, ..., 5.4300e-05, + 4.6968e-05, 4.5633e-04], + [ 3.1181e-06, -2.5139e-03, -9.5901e-03, ..., -9.0256e-03, + 2.1133e-03, 4.0092e-03], + [ 4.2468e-06, 4.7913e-03, 9.3613e-03, ..., 9.8190e-03, + 2.6226e-04, 1.0118e-03]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0101, 0.0186, -0.0021, -0.0080, 0.0168, -0.0302, 0.0115, -0.0043, + 0.0028, -0.0110], device='cuda:0'), grad: tensor([ 0.0227, 0.0072, 0.0034, 0.0209, -0.0114, 0.0382, -0.0957, -0.0016, + -0.0165, 0.0329], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 214.22, cls_loss 0.8847 cls_loss_mapping 0.0652 cls_loss_causal 0.7558 re_mapping 0.0327 re_causal 0.0737 /// teacc 97.47 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0193, -0.0317, -0.0502, ..., -0.0098, 0.0387, -0.0561], + [-0.0208, -0.0504, -0.0157, ..., 0.0771, -0.0347, 0.1038], + [ 0.0173, -0.0223, -0.0066, ..., -0.0108, -0.0082, -0.0396], + ..., + [ 0.0019, -0.0447, 0.0801, ..., 0.0014, -0.0238, 0.0232], + [-0.0041, 0.0389, -0.0505, ..., -0.0283, -0.0217, -0.0380], + [-0.0346, 0.0459, 0.0450, ..., -0.0382, -0.0175, -0.0053]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0032, 0.0069, ..., 0.0065, -0.0008, 0.0112], + [ 0.0000, 0.0042, -0.0024, ..., 0.0110, -0.0016, -0.0075], + [ 0.0000, 0.0050, 0.0073, ..., 0.0034, 0.0020, 0.0024], + ..., + [ 0.0000, 0.0036, -0.0076, ..., 0.0025, -0.0120, 0.0016], + [ 0.0000, 0.0132, 0.0122, ..., 0.0072, 0.0031, 0.0023], + [ 0.0000, -0.0276, -0.0126, ..., -0.0014, 0.0026, 0.0002]], + device='cuda:0') +Epoch 34, bias, value: tensor([-0.0097, 0.0192, -0.0026, -0.0074, 0.0170, -0.0301, 0.0115, -0.0048, + 0.0027, -0.0118], device='cuda:0'), grad: tensor([ 0.0225, 0.0112, 0.0286, 0.0406, -0.0466, 0.0123, -0.0445, -0.0400, + 0.0551, -0.0393], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 214.50, cls_loss 0.8869 cls_loss_mapping 0.0616 cls_loss_causal 0.7513 re_mapping 0.0318 re_causal 0.0704 /// teacc 97.37 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0197, -0.0324, -0.0503, ..., -0.0098, 0.0383, -0.0551], + [-0.0212, -0.0505, -0.0157, ..., 0.0774, -0.0356, 0.1055], + [ 0.0161, -0.0223, -0.0066, ..., -0.0107, -0.0075, -0.0404], + ..., + [ 0.0024, -0.0452, 0.0808, ..., 0.0018, -0.0240, 0.0240], + [-0.0041, 0.0385, -0.0509, ..., -0.0285, -0.0222, -0.0389], + [-0.0349, 0.0475, 0.0457, ..., -0.0389, -0.0164, -0.0063]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0017, 0.0076, ..., 0.0055, 0.0009, 0.0026], + [ 0.0000, -0.0055, -0.0098, ..., -0.0089, -0.0004, -0.0075], + [ 0.0000, -0.0033, -0.0153, ..., -0.0137, 0.0001, -0.0041], + ..., + [ 0.0000, 0.0030, -0.0046, ..., 0.0032, 0.0118, 0.0028], + [ 0.0000, 0.0045, 0.0049, ..., 0.0004, 0.0015, 0.0036], + [ 0.0000, 0.0011, 0.0063, ..., 0.0089, 0.0009, 0.0033]], + device='cuda:0') +Epoch 35, bias, value: tensor([-0.0101, 0.0189, -0.0027, -0.0071, 0.0165, -0.0306, 0.0115, -0.0038, + 0.0027, -0.0113], device='cuda:0'), grad: tensor([ 0.0250, -0.0105, -0.0544, -0.0125, -0.0432, 0.0295, 0.0173, -0.0130, + 0.0077, 0.0541], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 231.00, cls_loss 0.8966 cls_loss_mapping 0.0667 cls_loss_causal 0.7646 re_mapping 0.0308 re_causal 0.0685 /// teacc 97.62 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0203, -0.0330, -0.0513, ..., -0.0109, 0.0384, -0.0551], + [-0.0212, -0.0495, -0.0154, ..., 0.0778, -0.0356, 0.1061], + [ 0.0163, -0.0224, -0.0066, ..., -0.0114, -0.0068, -0.0408], + ..., + [ 0.0022, -0.0470, 0.0811, ..., 0.0024, -0.0239, 0.0238], + [-0.0035, 0.0379, -0.0517, ..., -0.0285, -0.0219, -0.0397], + [-0.0347, 0.0487, 0.0459, ..., -0.0389, -0.0169, -0.0063]], + device='cuda:0'), grad: tensor([[ 9.2983e-06, -1.7883e-02, -3.5439e-03, ..., -2.6321e-02, + -9.7809e-03, -5.3940e-03], + [ 1.7270e-05, 2.8248e-03, -2.8114e-03, ..., -1.0338e-02, + 2.1420e-03, -1.6403e-03], + [-8.8310e-04, 7.2975e-03, -1.8559e-03, ..., -7.8506e-03, + 1.0628e-02, 2.7523e-03], + ..., + [ 3.0017e-04, 8.6746e-03, 1.1230e-02, ..., 9.0485e-03, + 1.3285e-03, 7.3738e-03], + [ 1.1104e-04, 1.1368e-02, 5.6343e-03, ..., 1.7746e-02, + 4.1084e-03, 1.9646e-03], + [ 7.5214e-06, -1.2123e-02, -1.2764e-02, ..., -5.2910e-03, + 3.0308e-03, -8.0948e-03]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0113, 0.0192, -0.0025, -0.0070, 0.0166, -0.0303, 0.0115, -0.0031, + 0.0024, -0.0116], device='cuda:0'), grad: tensor([-0.0520, -0.0199, -0.0167, -0.0229, 0.0099, 0.0288, 0.0127, 0.0289, + 0.0496, -0.0183], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 230.67, cls_loss 0.8979 cls_loss_mapping 0.0703 cls_loss_causal 0.7711 re_mapping 0.0299 re_causal 0.0665 /// teacc 97.81 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0214, -0.0338, -0.0509, ..., -0.0100, 0.0382, -0.0546], + [-0.0210, -0.0489, -0.0156, ..., 0.0780, -0.0358, 0.1064], + [ 0.0173, -0.0221, -0.0072, ..., -0.0121, -0.0071, -0.0422], + ..., + [ 0.0021, -0.0475, 0.0819, ..., 0.0027, -0.0233, 0.0242], + [-0.0048, 0.0377, -0.0518, ..., -0.0294, -0.0221, -0.0393], + [-0.0347, 0.0491, 0.0450, ..., -0.0389, -0.0174, -0.0066]], + device='cuda:0'), grad: tensor([[ 1.5271e-04, 5.3139e-03, 7.6218e-03, ..., -5.7793e-03, + -1.8711e-03, -7.7477e-03], + [ 7.6592e-05, 2.6016e-03, 3.0613e-03, ..., 1.9775e-02, + 3.8981e-04, 9.1171e-03], + [ 1.2791e-04, 8.2111e-04, 1.7631e-04, ..., -3.5591e-03, + 1.0738e-03, -1.0691e-03], + ..., + [ 4.5329e-05, 6.9427e-04, 3.5400e-03, ..., 8.4839e-03, + 9.2888e-04, 1.3390e-03], + [-8.8978e-04, -7.7858e-03, -5.1727e-03, ..., 6.6566e-03, + -1.3218e-03, 1.8654e-03], + [ 2.1398e-04, 1.3638e-03, -1.0689e-02, ..., -6.5079e-03, + -1.3494e-04, 3.9711e-03]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0105, 0.0188, -0.0022, -0.0072, 0.0167, -0.0302, 0.0107, -0.0033, + 0.0027, -0.0115], device='cuda:0'), grad: tensor([-0.0070, 0.0391, 0.0070, 0.0205, -0.0211, 0.0307, -0.0658, 0.0286, + -0.0067, -0.0253], device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 214.49, cls_loss 0.8868 cls_loss_mapping 0.0695 cls_loss_causal 0.7595 re_mapping 0.0309 re_causal 0.0675 /// teacc 97.49 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0213, -0.0335, -0.0510, ..., -0.0104, 0.0389, -0.0557], + [-0.0212, -0.0497, -0.0159, ..., 0.0778, -0.0366, 0.1074], + [ 0.0173, -0.0210, -0.0071, ..., -0.0124, -0.0076, -0.0430], + ..., + [ 0.0027, -0.0479, 0.0827, ..., 0.0046, -0.0233, 0.0248], + [-0.0032, 0.0376, -0.0528, ..., -0.0293, -0.0211, -0.0387], + [-0.0371, 0.0498, 0.0457, ..., -0.0397, -0.0177, -0.0068]], + device='cuda:0'), grad: tensor([[ 1.0962e-06, 3.0880e-03, 6.4049e-03, ..., 2.0355e-02, + 3.8204e-03, 2.0790e-03], + [ 4.8764e-06, -8.3771e-03, -3.3073e-03, ..., 8.1329e-03, + -2.2507e-02, -7.1030e-03], + [ 2.4962e-04, 7.9727e-04, -4.1046e-03, ..., -3.8376e-03, + 1.2743e-04, 1.2770e-03], + ..., + [ 2.4572e-05, 6.3248e-03, 5.5008e-03, ..., 3.1414e-03, + 1.2192e-02, 3.6049e-03], + [ 2.7746e-05, 6.3133e-04, 3.9558e-03, ..., 8.9188e-03, + 1.6069e-03, 1.8940e-03], + [ 5.8375e-06, -5.1355e-04, 7.9651e-03, ..., 1.2093e-02, + 3.8700e-03, 2.0294e-03]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0103, 0.0187, -0.0022, -0.0069, 0.0166, -0.0309, 0.0103, -0.0027, + 0.0028, -0.0115], device='cuda:0'), grad: tensor([ 0.0613, -0.0061, -0.0187, -0.0197, -0.0136, -0.0352, -0.0471, 0.0176, + 0.0265, 0.0349], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 214.46, cls_loss 0.8399 cls_loss_mapping 0.0594 cls_loss_causal 0.7121 re_mapping 0.0301 re_causal 0.0660 /// teacc 97.61 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0219, -0.0344, -0.0514, ..., -0.0116, 0.0396, -0.0573], + [-0.0215, -0.0502, -0.0160, ..., 0.0779, -0.0361, 0.1080], + [ 0.0160, -0.0210, -0.0059, ..., -0.0117, -0.0078, -0.0440], + ..., + [ 0.0035, -0.0483, 0.0821, ..., 0.0039, -0.0239, 0.0252], + [-0.0027, 0.0390, -0.0539, ..., -0.0303, -0.0204, -0.0393], + [-0.0375, 0.0494, 0.0466, ..., -0.0392, -0.0182, -0.0070]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.2888e-03, -3.5629e-03, ..., 4.6082e-03, + 8.1110e-04, -1.0605e-03], + [ 3.5390e-08, 2.4567e-03, 7.3776e-03, ..., 2.8515e-03, + -6.3479e-05, -9.7885e-03], + [ 1.8068e-07, 2.7905e-03, -2.2945e-03, ..., -2.3773e-02, + -3.3474e-04, 7.0381e-04], + ..., + [ 1.7695e-08, -3.2330e-03, -1.2138e-02, ..., -4.0092e-03, + -1.8191e-04, -3.2330e-03], + [ 2.4438e-06, -3.5906e-04, 1.5373e-03, ..., 2.6016e-03, + 7.1716e-04, -4.3793e-03], + [ 5.5879e-09, 7.1983e-03, 4.2572e-03, ..., 9.1410e-04, + 4.2176e-04, 1.1940e-02]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0104, 0.0192, -0.0010, -0.0069, 0.0156, -0.0309, 0.0107, -0.0033, + 0.0021, -0.0110], device='cuda:0'), grad: tensor([-0.0080, 0.0214, -0.0464, 0.0231, 0.0078, -0.0651, 0.0531, -0.0052, + 0.0060, 0.0133], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 232.24, cls_loss 0.8235 cls_loss_mapping 0.0562 cls_loss_causal 0.6951 re_mapping 0.0313 re_causal 0.0668 /// teacc 97.86 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0232, -0.0353, -0.0522, ..., -0.0119, 0.0396, -0.0579], + [-0.0223, -0.0513, -0.0173, ..., 0.0778, -0.0347, 0.1081], + [ 0.0162, -0.0213, -0.0054, ..., -0.0124, -0.0080, -0.0447], + ..., + [ 0.0035, -0.0474, 0.0828, ..., 0.0043, -0.0255, 0.0265], + [-0.0036, 0.0396, -0.0541, ..., -0.0300, -0.0202, -0.0385], + [-0.0388, 0.0498, 0.0466, ..., -0.0393, -0.0182, -0.0076]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.8893e-03, 4.6164e-05, ..., -4.1924e-03, + -3.8552e-04, -1.7309e-03], + [ 0.0000e+00, 2.0826e-04, 1.2989e-03, ..., -6.9275e-03, + 8.7500e-04, -1.0010e-02], + [ 0.0000e+00, -2.1195e-04, 7.0763e-03, ..., 1.3145e-02, + -4.4479e-03, 1.5831e-03], + ..., + [ 3.7253e-07, 5.7697e-04, 6.4812e-03, ..., 1.2619e-02, + 4.1008e-04, 1.6708e-03], + [ 1.9651e-07, -2.8629e-03, -3.3531e-03, ..., 9.7198e-03, + 2.9945e-03, 2.7122e-03], + [-7.8231e-07, -2.3289e-03, -1.4664e-02, ..., -1.4641e-02, + -6.2103e-03, 3.3855e-03]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0107, 0.0186, -0.0012, -0.0070, 0.0165, -0.0311, 0.0110, -0.0034, + 0.0019, -0.0107], device='cuda:0'), grad: tensor([-0.0117, -0.0254, 0.0353, 0.0096, -0.1215, 0.0702, 0.0079, 0.0341, + 0.0174, -0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 214.37, cls_loss 0.8404 cls_loss_mapping 0.0599 cls_loss_causal 0.7230 re_mapping 0.0281 re_causal 0.0608 /// teacc 97.67 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0242, -0.0349, -0.0527, ..., -0.0125, 0.0394, -0.0578], + [-0.0215, -0.0518, -0.0172, ..., 0.0781, -0.0361, 0.1100], + [ 0.0154, -0.0219, -0.0056, ..., -0.0122, -0.0069, -0.0449], + ..., + [ 0.0030, -0.0476, 0.0833, ..., 0.0050, -0.0263, 0.0256], + [-0.0042, 0.0402, -0.0543, ..., -0.0298, -0.0202, -0.0383], + [-0.0385, 0.0500, 0.0467, ..., -0.0391, -0.0177, -0.0080]], + device='cuda:0'), grad: tensor([[ 3.9376e-06, -3.7918e-03, -4.1122e-03, ..., -3.0785e-03, + 2.0351e-03, 2.2621e-03], + [ 1.5320e-06, 1.2569e-03, 1.0624e-03, ..., -9.4128e-04, + 6.5327e-04, -2.0142e-03], + [ 5.2527e-07, 7.3481e-04, -3.4084e-03, ..., -9.0256e-03, + -2.0599e-03, 6.9857e-04], + ..., + [ 3.4887e-06, 4.3869e-03, 5.2376e-03, ..., 5.4665e-03, + 2.1572e-03, 3.5992e-03], + [-2.6494e-05, -3.9337e-02, -1.6846e-02, ..., 2.0542e-03, + -3.4607e-02, 2.0542e-03], + [ 2.2240e-06, 2.6733e-02, 7.9269e-03, ..., -8.3542e-03, + 2.3590e-02, -1.2383e-02]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0107, 0.0192, -0.0012, -0.0070, 0.0164, -0.0315, 0.0106, -0.0034, + 0.0022, -0.0107], device='cuda:0'), grad: tensor([-0.0532, 0.0065, -0.0217, 0.0141, 0.0393, 0.0370, -0.0051, 0.0211, + -0.0089, -0.0292], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 230.61, cls_loss 0.8366 cls_loss_mapping 0.0535 cls_loss_causal 0.7018 re_mapping 0.0299 re_causal 0.0630 /// teacc 98.00 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0247, -0.0359, -0.0540, ..., -0.0134, 0.0387, -0.0594], + [-0.0214, -0.0534, -0.0181, ..., 0.0777, -0.0359, 0.1100], + [ 0.0154, -0.0225, -0.0056, ..., -0.0117, -0.0067, -0.0444], + ..., + [ 0.0017, -0.0487, 0.0838, ..., 0.0053, -0.0268, 0.0265], + [-0.0057, 0.0392, -0.0554, ..., -0.0297, -0.0203, -0.0385], + [-0.0379, 0.0497, 0.0467, ..., -0.0386, -0.0191, -0.0089]], + device='cuda:0'), grad: tensor([[ 2.4587e-05, 7.0095e-05, 2.5406e-03, ..., 6.4888e-03, + 8.6117e-04, 2.3022e-03], + [ 3.7905e-06, 2.1350e-04, -5.5695e-03, ..., -4.8447e-03, + 5.2643e-03, -7.2098e-03], + [-9.7580e-03, -8.0719e-03, -7.3792e-02, ..., -2.8397e-02, + -2.1912e-02, -2.8961e-02], + ..., + [ 6.9847e-03, 7.0610e-03, 3.2990e-02, ..., 2.2400e-02, + 1.0139e-02, 2.0859e-02], + [-3.2158e-03, 2.5201e-04, 4.3716e-03, ..., 5.4359e-03, + 1.1997e-03, 2.3441e-03], + [ 4.0948e-05, -4.3526e-03, -2.2144e-03, ..., 1.7796e-03, + -1.9312e-03, 7.9966e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0112, 0.0189, -0.0012, -0.0072, 0.0172, -0.0308, 0.0105, -0.0036, + 0.0027, -0.0113], device='cuda:0'), grad: tensor([ 2.6901e-02, 7.4911e-04, -1.1530e-01, 5.2155e-02, -5.7012e-05, + -2.2003e-02, -9.4223e-03, 6.0272e-02, 8.9798e-03, -2.2106e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 214.46, cls_loss 0.8315 cls_loss_mapping 0.0527 cls_loss_causal 0.7021 re_mapping 0.0285 re_causal 0.0613 /// teacc 97.82 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0258, -0.0356, -0.0537, ..., -0.0131, 0.0393, -0.0585], + [-0.0229, -0.0541, -0.0179, ..., 0.0780, -0.0367, 0.1108], + [ 0.0163, -0.0231, -0.0052, ..., -0.0122, -0.0064, -0.0452], + ..., + [ 0.0010, -0.0485, 0.0835, ..., 0.0052, -0.0281, 0.0275], + [-0.0056, 0.0395, -0.0567, ..., -0.0292, -0.0200, -0.0392], + [-0.0376, 0.0507, 0.0472, ..., -0.0396, -0.0181, -0.0104]], + device='cuda:0'), grad: tensor([[ 8.6352e-06, -8.6021e-04, 1.6575e-03, ..., 1.9789e-04, + 6.4468e-03, -2.5082e-03], + [ 1.4426e-06, -3.9792e-04, -7.7629e-03, ..., -5.9929e-03, + -5.9662e-03, -7.4348e-03], + [ 2.0824e-06, 1.4639e-03, 1.8492e-03, ..., 4.7760e-03, + 1.4191e-03, 2.4548e-03], + ..., + [ 8.9109e-06, 7.3814e-04, -1.7061e-03, ..., -4.4479e-03, + -8.2855e-03, -5.9090e-03], + [ 3.3110e-05, 9.6436e-03, 5.7526e-03, ..., 8.7051e-03, + -1.0338e-02, 4.0550e-03], + [ 2.0951e-05, 1.7033e-03, 3.3531e-03, ..., 4.6730e-03, + 6.4812e-03, 4.9553e-03]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0111, 0.0189, -0.0016, -0.0065, 0.0169, -0.0315, 0.0103, -0.0037, + 0.0030, -0.0110], device='cuda:0'), grad: tensor([ 0.0116, -0.0250, 0.0182, -0.0217, 0.0144, -0.0065, 0.0105, -0.0472, + 0.0113, 0.0345], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.13, cls_loss 0.8412 cls_loss_mapping 0.0525 cls_loss_causal 0.7236 re_mapping 0.0277 re_causal 0.0601 /// teacc 97.83 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0266, -0.0364, -0.0533, ..., -0.0139, 0.0397, -0.0579], + [-0.0228, -0.0546, -0.0179, ..., 0.0775, -0.0369, 0.1109], + [ 0.0169, -0.0229, -0.0058, ..., -0.0121, -0.0070, -0.0470], + ..., + [ 0.0010, -0.0499, 0.0838, ..., 0.0056, -0.0290, 0.0279], + [-0.0068, 0.0392, -0.0563, ..., -0.0286, -0.0204, -0.0395], + [-0.0386, 0.0511, 0.0463, ..., -0.0400, -0.0173, -0.0105]], + device='cuda:0'), grad: tensor([[ 2.2545e-05, 3.1929e-03, 2.9945e-03, ..., -4.0054e-03, + -1.1391e-02, -9.6512e-03], + [ 4.6206e-04, 1.9331e-03, 2.5883e-03, ..., 5.8441e-03, + 5.9242e-03, 7.4844e-03], + [ 2.7359e-05, 5.6458e-04, 9.0714e-03, ..., 7.5865e-04, + 2.2602e-03, 1.0052e-03], + ..., + [-1.2302e-03, 2.4533e-04, 4.7684e-03, ..., 9.0790e-04, + 1.0262e-02, -8.8272e-03], + [ 4.4394e-04, 6.4278e-03, 6.1684e-03, ..., 4.4060e-03, + 4.0741e-03, 6.6528e-03], + [ 8.2612e-05, -8.5449e-03, -2.5635e-02, ..., 1.9855e-03, + -1.9180e-02, -1.7929e-03]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0118, 0.0184, -0.0016, -0.0068, 0.0172, -0.0310, 0.0108, -0.0042, + 0.0037, -0.0111], device='cuda:0'), grad: tensor([-0.0061, 0.0322, 0.0255, 0.0339, 0.0120, -0.0120, -0.0852, -0.0016, + 0.0267, -0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 214.28, cls_loss 0.8535 cls_loss_mapping 0.0513 cls_loss_causal 0.7193 re_mapping 0.0275 re_causal 0.0591 /// teacc 97.80 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0269, -0.0368, -0.0535, ..., -0.0133, 0.0403, -0.0577], + [-0.0232, -0.0534, -0.0171, ..., 0.0778, -0.0359, 0.1114], + [ 0.0170, -0.0240, -0.0063, ..., -0.0116, -0.0078, -0.0469], + ..., + [ 0.0019, -0.0497, 0.0852, ..., 0.0055, -0.0287, 0.0278], + [-0.0083, 0.0391, -0.0571, ..., -0.0287, -0.0193, -0.0408], + [-0.0386, 0.0511, 0.0475, ..., -0.0404, -0.0181, -0.0099]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0035, 0.0029, ..., 0.0034, -0.0003, 0.0030], + [ 0.0000, -0.0058, 0.0009, ..., -0.0060, 0.0005, -0.0031], + [ 0.0000, 0.0058, 0.0068, ..., 0.0003, -0.0017, 0.0025], + ..., + [ 0.0000, -0.0235, -0.0384, ..., 0.0007, -0.0003, -0.0023], + [ 0.0000, 0.0278, 0.0376, ..., 0.0070, 0.0064, 0.0043], + [ 0.0000, -0.0258, 0.0038, ..., 0.0023, 0.0011, 0.0009]], + device='cuda:0') +Epoch 45, bias, value: tensor([-0.0112, 0.0187, -0.0015, -0.0073, 0.0166, -0.0308, 0.0110, -0.0043, + 0.0042, -0.0115], device='cuda:0'), grad: tensor([ 0.0165, -0.0094, 0.0056, 0.0151, 0.0002, -0.0140, -0.0166, -0.0218, + 0.0394, -0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 214.16, cls_loss 0.7860 cls_loss_mapping 0.0463 cls_loss_causal 0.6671 re_mapping 0.0279 re_causal 0.0599 /// teacc 97.91 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0265, -0.0362, -0.0530, ..., -0.0141, 0.0403, -0.0572], + [-0.0233, -0.0544, -0.0175, ..., 0.0787, -0.0359, 0.1127], + [ 0.0172, -0.0239, -0.0063, ..., -0.0117, -0.0078, -0.0469], + ..., + [ 0.0008, -0.0494, 0.0853, ..., 0.0055, -0.0285, 0.0283], + [-0.0064, 0.0384, -0.0582, ..., -0.0291, -0.0187, -0.0421], + [-0.0400, 0.0522, 0.0479, ..., -0.0413, -0.0173, -0.0098]], + device='cuda:0'), grad: tensor([[ 2.3842e-06, 2.9011e-03, 7.9803e-03, ..., 1.0704e-02, + 0.0000e+00, 3.8757e-03], + [ 5.5954e-06, -7.1907e-03, -6.5613e-03, ..., -1.4473e-02, + 3.7253e-09, -5.2109e-03], + [ 1.7256e-05, -1.6289e-03, -8.6288e-03, ..., 9.3231e-03, + -1.5691e-05, 4.6158e-03], + ..., + [-1.2803e-04, 1.6356e-03, 7.8735e-03, ..., 1.2817e-02, + 5.7630e-06, 8.8577e-03], + [ 3.5673e-05, 2.8172e-03, 7.3128e-03, ..., 9.1629e-03, + 5.0385e-07, 6.5079e-03], + [ 1.0127e-04, -3.5667e-03, -6.1188e-03, ..., -1.1826e-02, + 0.0000e+00, -8.1406e-03]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0118, 0.0187, -0.0010, -0.0075, 0.0171, -0.0307, 0.0106, -0.0039, + 0.0035, -0.0112], device='cuda:0'), grad: tensor([ 0.0303, -0.0233, 0.0107, -0.0123, -0.0010, 0.0173, -0.0536, 0.0394, + 0.0294, -0.0369], device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 214.15, cls_loss 0.8077 cls_loss_mapping 0.0431 cls_loss_causal 0.6804 re_mapping 0.0275 re_causal 0.0606 /// teacc 97.75 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0266, -0.0369, -0.0545, ..., -0.0157, 0.0402, -0.0584], + [-0.0245, -0.0549, -0.0185, ..., 0.0784, -0.0354, 0.1143], + [ 0.0178, -0.0237, -0.0060, ..., -0.0128, -0.0078, -0.0451], + ..., + [ 0.0006, -0.0493, 0.0860, ..., 0.0066, -0.0288, 0.0279], + [-0.0064, 0.0391, -0.0579, ..., -0.0291, -0.0190, -0.0424], + [-0.0396, 0.0510, 0.0470, ..., -0.0416, -0.0171, -0.0110]], + device='cuda:0'), grad: tensor([[ 8.8394e-05, -2.6360e-03, 2.5177e-03, ..., 3.4523e-03, + -9.5062e-03, 7.6818e-04], + [ 3.8922e-05, 6.7616e-04, 6.0892e-04, ..., -8.0566e-03, + 5.8889e-04, 1.2195e-04], + [-1.1063e-04, -1.4893e-02, 3.1021e-02, ..., 1.4191e-02, + -5.3823e-05, 3.0766e-03], + ..., + [-2.4624e-03, -4.1389e-04, -5.9204e-02, ..., -3.5675e-02, + 5.2786e-04, -3.4599e-03], + [-2.1072e-02, -1.1787e-02, 1.5574e-03, ..., 8.1635e-03, + 2.4452e-03, -2.6464e-05], + [ 2.0645e-02, 1.3275e-02, -1.0099e-03, ..., 6.5422e-03, + 9.8133e-04, -1.7529e-03]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0122, 0.0188, -0.0008, -0.0070, 0.0179, -0.0313, 0.0108, -0.0037, + 0.0032, -0.0118], device='cuda:0'), grad: tensor([-0.0014, -0.0146, 0.0055, 0.0422, -0.0097, 0.0379, -0.0322, -0.0515, + -0.0092, 0.0328], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 214.19, cls_loss 0.7863 cls_loss_mapping 0.0411 cls_loss_causal 0.6665 re_mapping 0.0271 re_causal 0.0585 /// teacc 97.63 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0253, -0.0380, -0.0553, ..., -0.0157, 0.0411, -0.0588], + [-0.0249, -0.0556, -0.0182, ..., 0.0796, -0.0358, 0.1152], + [ 0.0178, -0.0237, -0.0064, ..., -0.0136, -0.0072, -0.0459], + ..., + [ 0.0003, -0.0495, 0.0873, ..., 0.0066, -0.0283, 0.0281], + [-0.0044, 0.0392, -0.0587, ..., -0.0287, -0.0194, -0.0431], + [-0.0416, 0.0516, 0.0463, ..., -0.0425, -0.0181, -0.0108]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0014, -0.0050, ..., -0.0029, 0.0008, -0.0017], + [ 0.0000, 0.0085, 0.0052, ..., 0.0086, 0.0043, 0.0013], + [ 0.0000, 0.0004, -0.0135, ..., -0.0191, -0.0093, -0.0071], + ..., + [ 0.0000, 0.0024, 0.0115, ..., 0.0154, 0.0011, 0.0073], + [ 0.0000, -0.0082, -0.0032, ..., -0.0034, -0.0106, -0.0022], + [ 0.0000, -0.0030, 0.0038, ..., 0.0057, 0.0009, 0.0024]], + device='cuda:0') +Epoch 48, bias, value: tensor([-0.0125, 0.0200, -0.0012, -0.0066, 0.0175, -0.0311, 0.0102, -0.0040, + 0.0037, -0.0121], device='cuda:0'), grad: tensor([-0.0083, 0.0297, -0.0665, -0.0049, 0.0281, -0.0093, 0.0003, 0.0418, + -0.0057, -0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 214.51, cls_loss 0.8393 cls_loss_mapping 0.0487 cls_loss_causal 0.7269 re_mapping 0.0271 re_causal 0.0585 /// teacc 97.90 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0266, -0.0391, -0.0557, ..., -0.0162, 0.0411, -0.0594], + [-0.0249, -0.0560, -0.0191, ..., 0.0792, -0.0370, 0.1154], + [ 0.0175, -0.0236, -0.0069, ..., -0.0134, -0.0065, -0.0461], + ..., + [ 0.0002, -0.0500, 0.0877, ..., 0.0076, -0.0297, 0.0286], + [-0.0049, 0.0392, -0.0581, ..., -0.0300, -0.0193, -0.0435], + [-0.0412, 0.0524, 0.0463, ..., -0.0431, -0.0174, -0.0112]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0019, -0.0104, ..., -0.0130, 0.0040, -0.0121], + [ 0.0000, 0.0019, 0.0043, ..., 0.0066, 0.0036, 0.0040], + [ 0.0000, 0.0015, 0.0037, ..., 0.0072, 0.0028, 0.0037], + ..., + [ 0.0000, 0.0019, -0.0011, ..., 0.0106, 0.0035, -0.0009], + [ 0.0000, 0.0031, 0.0024, ..., 0.0076, 0.0053, 0.0077], + [ 0.0000, -0.0044, -0.0052, ..., -0.0353, -0.0100, -0.0035]], + device='cuda:0') +Epoch 49, bias, value: tensor([-0.0130, 0.0194, -0.0012, -0.0069, 0.0181, -0.0308, 0.0110, -0.0039, + 0.0034, -0.0123], device='cuda:0'), grad: tensor([-0.0370, 0.0163, 0.0215, 0.0089, -0.0383, 0.0341, -0.0009, 0.0178, + 0.0581, -0.0804], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 214.35, cls_loss 0.8274 cls_loss_mapping 0.0519 cls_loss_causal 0.7216 re_mapping 0.0261 re_causal 0.0562 /// teacc 97.90 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0273, -0.0390, -0.0560, ..., -0.0154, 0.0410, -0.0593], + [-0.0251, -0.0572, -0.0198, ..., 0.0793, -0.0365, 0.1161], + [ 0.0178, -0.0234, -0.0062, ..., -0.0134, -0.0071, -0.0453], + ..., + [ 0.0007, -0.0503, 0.0878, ..., 0.0075, -0.0297, 0.0291], + [-0.0055, 0.0393, -0.0576, ..., -0.0303, -0.0191, -0.0456], + [-0.0407, 0.0527, 0.0462, ..., -0.0429, -0.0175, -0.0120]], + device='cuda:0'), grad: tensor([[ 4.6082e-06, 3.2663e-04, 4.4465e-04, ..., -1.6022e-04, + 9.1171e-04, 9.1410e-04], + [ 1.0714e-05, 1.8692e-03, 1.5697e-03, ..., 7.6981e-03, + 5.7640e-03, 1.9932e-03], + [-5.0402e-04, 1.1301e-03, -1.5518e-02, ..., -6.5231e-03, + 1.2665e-03, -1.7633e-03], + ..., + [ 3.8004e-04, 7.6437e-04, 9.4528e-03, ..., -9.5129e-04, + -2.4170e-02, 1.9093e-03], + [ 1.0267e-05, -1.7975e-02, -6.0797e-04, ..., -2.1343e-03, + 3.9406e-03, -1.1225e-03], + [ 9.0227e-06, 1.7691e-03, 4.9057e-03, ..., 9.3231e-03, + 6.0310e-03, 4.9400e-03]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0127, 0.0189, -0.0014, -0.0070, 0.0179, -0.0315, 0.0110, -0.0039, + 0.0040, -0.0116], device='cuda:0'), grad: tensor([-0.0119, 0.0499, -0.0160, 0.0387, -0.0546, 0.0052, 0.0005, -0.0764, + 0.0156, 0.0490], device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 214.61, cls_loss 0.7766 cls_loss_mapping 0.0506 cls_loss_causal 0.6567 re_mapping 0.0265 re_causal 0.0555 /// teacc 97.78 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0280, -0.0398, -0.0561, ..., -0.0159, 0.0412, -0.0589], + [-0.0236, -0.0579, -0.0202, ..., 0.0794, -0.0367, 0.1173], + [ 0.0189, -0.0229, -0.0056, ..., -0.0130, -0.0065, -0.0448], + ..., + [ 0.0014, -0.0511, 0.0877, ..., 0.0074, -0.0297, 0.0287], + [-0.0054, 0.0405, -0.0583, ..., -0.0307, -0.0194, -0.0465], + [-0.0424, 0.0530, 0.0462, ..., -0.0429, -0.0191, -0.0122]], + device='cuda:0'), grad: tensor([[ 6.1393e-06, 3.2330e-04, 1.0433e-03, ..., 5.2185e-03, + 1.5116e-03, 3.6850e-03], + [-3.0088e-04, -9.0885e-04, -6.3324e-03, ..., -1.8402e-02, + -8.2321e-03, -1.6571e-02], + [ 2.5541e-05, 7.1239e-04, -1.2581e-02, ..., -1.9684e-02, + -1.2159e-03, -9.7046e-03], + ..., + [ 4.7469e-04, 4.2200e-04, 2.8992e-02, ..., 1.7029e-02, + 1.2379e-03, 2.0401e-02], + [ 1.1081e-04, -4.3373e-03, -3.7594e-03, ..., -8.1635e-03, + 4.6730e-04, -4.9591e-03], + [ 9.7561e-04, 2.9125e-03, -1.5869e-02, ..., 4.5624e-03, + 9.5129e-05, -3.1338e-03]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0128, 0.0190, -0.0006, -0.0072, 0.0179, -0.0310, 0.0118, -0.0039, + 0.0030, -0.0121], device='cuda:0'), grad: tensor([ 0.0217, -0.0477, -0.0093, 0.0204, -0.0313, 0.0171, 0.0219, 0.0746, + -0.0645, -0.0030], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 50---------------------------------------------------- +epoch 50, time 231.21, cls_loss 0.7822 cls_loss_mapping 0.0408 cls_loss_causal 0.6600 re_mapping 0.0263 re_causal 0.0578 /// teacc 98.04 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0297, -0.0397, -0.0564, ..., -0.0170, 0.0408, -0.0593], + [-0.0215, -0.0588, -0.0217, ..., 0.0805, -0.0365, 0.1182], + [ 0.0194, -0.0240, -0.0057, ..., -0.0130, -0.0063, -0.0445], + ..., + [-0.0006, -0.0517, 0.0879, ..., 0.0069, -0.0306, 0.0294], + [-0.0053, 0.0405, -0.0587, ..., -0.0307, -0.0185, -0.0477], + [-0.0434, 0.0540, 0.0470, ..., -0.0432, -0.0197, -0.0127]], + device='cuda:0'), grad: tensor([[ 1.3923e-07, 7.1049e-04, 7.1955e-04, ..., -1.3535e-02, + 1.7548e-02, -1.3336e-02], + [ 3.4738e-07, -8.0776e-04, -1.0624e-03, ..., -2.0161e-03, + -2.7878e-02, -3.7041e-03], + [ 2.0787e-06, 1.5688e-03, -7.2956e-04, ..., -6.3477e-03, + 3.5439e-03, -5.0211e-04], + ..., + [ 2.8452e-07, 6.3133e-03, 6.3553e-03, ..., 4.1733e-03, + -4.0550e-03, 1.7151e-02], + [-4.8243e-06, -4.5419e-04, 8.8577e-03, ..., 1.0109e-02, + -1.5736e-03, 5.9624e-03], + [ 3.0873e-07, -1.2032e-02, -1.7151e-02, ..., 3.1548e-03, + 1.7567e-03, -1.4557e-02]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0132, 0.0192, -0.0005, -0.0067, 0.0181, -0.0316, 0.0113, -0.0039, + 0.0035, -0.0124], device='cuda:0'), grad: tensor([-0.0156, -0.0148, -0.0235, 0.0211, -0.0049, 0.0145, -0.0173, 0.0565, + 0.0051, -0.0214], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 214.54, cls_loss 0.7783 cls_loss_mapping 0.0382 cls_loss_causal 0.6703 re_mapping 0.0244 re_causal 0.0536 /// teacc 97.94 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0277, -0.0394, -0.0567, ..., -0.0176, 0.0394, -0.0594], + [-0.0216, -0.0584, -0.0228, ..., 0.0801, -0.0360, 0.1187], + [ 0.0196, -0.0261, -0.0066, ..., -0.0140, -0.0063, -0.0454], + ..., + [-0.0010, -0.0527, 0.0885, ..., 0.0074, -0.0300, 0.0288], + [-0.0058, 0.0413, -0.0588, ..., -0.0311, -0.0185, -0.0478], + [-0.0436, 0.0536, 0.0475, ..., -0.0426, -0.0213, -0.0120]], + device='cuda:0'), grad: tensor([[ 1.0366e-06, 2.3997e-04, 8.7738e-04, ..., -4.0512e-03, + -5.0278e-03, -2.8496e-03], + [ 2.0210e-06, 1.7464e-05, -1.7538e-03, ..., 3.3131e-03, + 1.1883e-03, -8.7214e-04], + [ 1.0341e-05, 8.1599e-05, 4.2076e-03, ..., 8.3847e-03, + 5.9271e-04, 1.4267e-03], + ..., + [ 5.9642e-06, 6.2943e-04, 3.0785e-03, ..., 4.0894e-03, + 7.3433e-04, 1.8616e-03], + [ 7.4208e-05, 1.3466e-03, 4.1389e-03, ..., 7.3166e-03, + 8.9979e-04, 2.6627e-03], + [ 3.3453e-06, -2.6131e-03, -6.2132e-04, ..., 2.2354e-03, + 5.2023e-04, 7.2575e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0131, 0.0187, -0.0014, -0.0058, 0.0180, -0.0322, 0.0117, -0.0038, + 0.0037, -0.0121], device='cuda:0'), grad: tensor([-0.0162, 0.0025, 0.0170, -0.0415, -0.0117, -0.0176, 0.0053, 0.0078, + 0.0278, 0.0266], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 52---------------------------------------------------- +epoch 52, time 225.86, cls_loss 0.8073 cls_loss_mapping 0.0421 cls_loss_causal 0.6886 re_mapping 0.0252 re_causal 0.0560 /// teacc 98.05 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0281, -0.0392, -0.0567, ..., -0.0176, 0.0395, -0.0594], + [-0.0218, -0.0575, -0.0233, ..., 0.0798, -0.0356, 0.1201], + [ 0.0196, -0.0264, -0.0068, ..., -0.0138, -0.0056, -0.0445], + ..., + [-0.0014, -0.0528, 0.0890, ..., 0.0077, -0.0312, 0.0297], + [-0.0065, 0.0426, -0.0597, ..., -0.0308, -0.0177, -0.0493], + [-0.0434, 0.0544, 0.0477, ..., -0.0433, -0.0213, -0.0119]], + device='cuda:0'), grad: tensor([[ 7.4133e-06, 3.2365e-05, -3.4027e-03, ..., -2.8114e-03, + 2.4033e-03, 1.1330e-03], + [ 8.7768e-06, -4.8065e-04, -1.9341e-03, ..., -7.5684e-03, + 7.3910e-04, -2.7275e-03], + [ 1.1295e-05, 8.2374e-05, -1.5516e-03, ..., -3.3207e-03, + 4.0913e-04, -6.9008e-03], + ..., + [ 8.3894e-06, -3.4124e-05, -9.5901e-03, ..., 2.9984e-03, + -1.9577e-02, -3.8338e-03], + [-1.8616e-03, -9.6416e-04, 3.0651e-03, ..., -4.7150e-03, + 1.4715e-03, -3.3817e-03], + [ 1.2696e-05, 5.8591e-05, 2.7275e-03, ..., -6.1646e-03, + 1.5574e-03, 2.0199e-03]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0135, 0.0190, -0.0014, -0.0054, 0.0179, -0.0318, 0.0111, -0.0035, + 0.0034, -0.0123], device='cuda:0'), grad: tensor([-0.0037, -0.0090, -0.0682, 0.0423, 0.0419, 0.0076, 0.0186, -0.0336, + 0.0029, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 214.44, cls_loss 0.8212 cls_loss_mapping 0.0434 cls_loss_causal 0.7036 re_mapping 0.0250 re_causal 0.0556 /// teacc 98.01 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0287, -0.0406, -0.0567, ..., -0.0183, 0.0381, -0.0600], + [-0.0223, -0.0577, -0.0231, ..., 0.0804, -0.0359, 0.1208], + [ 0.0194, -0.0270, -0.0077, ..., -0.0143, -0.0064, -0.0449], + ..., + [-0.0018, -0.0532, 0.0891, ..., 0.0069, -0.0310, 0.0296], + [-0.0063, 0.0417, -0.0602, ..., -0.0312, -0.0167, -0.0481], + [-0.0431, 0.0542, 0.0482, ..., -0.0409, -0.0214, -0.0112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.1948e-04, -2.6631e-04, ..., -5.7144e-03, + -1.9722e-03, -8.1396e-04], + [ 0.0000e+00, 1.2815e-04, 1.4429e-03, ..., 1.8921e-02, + 5.6297e-05, 2.9316e-03], + [ 0.0000e+00, 6.8245e-03, 1.6890e-03, ..., 4.2458e-03, + 8.3208e-05, 4.3273e-04], + ..., + [ 0.0000e+00, 2.6913e-03, -6.9313e-03, ..., -4.3068e-03, + 1.6570e-05, -3.6926e-03], + [ 0.0000e+00, 2.9736e-03, 1.3962e-03, ..., -3.3779e-03, + 1.0920e-03, 6.6328e-04], + [ 0.0000e+00, -3.1223e-03, 8.0204e-04, ..., 5.0507e-03, + 6.8843e-05, 1.8816e-03]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0132, 0.0183, -0.0014, -0.0057, 0.0177, -0.0321, 0.0108, -0.0041, + 0.0038, -0.0106], device='cuda:0'), grad: tensor([-0.0047, 0.0443, -0.0064, -0.0505, 0.0266, 0.0004, -0.0220, -0.0387, + 0.0181, 0.0328], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 54---------------------------------------------------- +epoch 54, time 225.54, cls_loss 0.7806 cls_loss_mapping 0.0391 cls_loss_causal 0.6577 re_mapping 0.0252 re_causal 0.0520 /// teacc 98.07 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0294, -0.0408, -0.0560, ..., -0.0185, 0.0391, -0.0595], + [-0.0217, -0.0582, -0.0228, ..., 0.0811, -0.0338, 0.1209], + [ 0.0205, -0.0273, -0.0082, ..., -0.0143, -0.0068, -0.0450], + ..., + [-0.0016, -0.0535, 0.0899, ..., 0.0068, -0.0320, 0.0296], + [-0.0073, 0.0405, -0.0613, ..., -0.0312, -0.0173, -0.0482], + [-0.0430, 0.0554, 0.0482, ..., -0.0415, -0.0213, -0.0120]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0028, 0.0024, ..., 0.0088, 0.0022, 0.0006], + [ 0.0000, 0.0012, -0.0077, ..., 0.0049, 0.0007, -0.0081], + [ 0.0000, -0.0104, -0.0026, ..., -0.0283, -0.0184, 0.0015], + ..., + [ 0.0000, 0.0023, 0.0058, ..., -0.0049, 0.0012, 0.0017], + [ 0.0000, 0.0058, -0.0004, ..., 0.0025, 0.0016, 0.0010], + [ 0.0000, -0.0023, -0.0026, ..., -0.0049, 0.0012, 0.0023]], + device='cuda:0') +Epoch 56, bias, value: tensor([-0.0128, 0.0191, -0.0011, -0.0063, 0.0173, -0.0320, 0.0117, -0.0046, + 0.0032, -0.0109], device='cuda:0'), grad: tensor([ 0.0180, -0.0137, -0.0435, 0.0233, 0.0139, -0.0347, 0.0319, 0.0141, + -0.0040, -0.0053], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 55---------------------------------------------------- +epoch 55, time 230.52, cls_loss 0.7956 cls_loss_mapping 0.0358 cls_loss_causal 0.6772 re_mapping 0.0240 re_causal 0.0522 /// teacc 98.14 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0287, -0.0414, -0.0566, ..., -0.0188, 0.0385, -0.0596], + [-0.0225, -0.0580, -0.0231, ..., 0.0812, -0.0343, 0.1225], + [ 0.0199, -0.0282, -0.0085, ..., -0.0151, -0.0064, -0.0457], + ..., + [-0.0019, -0.0535, 0.0900, ..., 0.0075, -0.0339, 0.0300], + [-0.0079, 0.0407, -0.0614, ..., -0.0307, -0.0167, -0.0481], + [-0.0424, 0.0553, 0.0477, ..., -0.0428, -0.0199, -0.0134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.7285e-06, -3.2532e-02, ..., -5.9700e-03, + -1.4107e-02, -3.7622e-04], + [ 0.0000e+00, -2.8872e-04, -8.3008e-03, ..., -1.1055e-02, + 2.3632e-03, -8.0585e-04], + [ 0.0000e+00, 1.4944e-03, 6.0005e-03, ..., 3.7861e-03, + 3.6678e-03, 9.0647e-04], + ..., + [ 0.0000e+00, 4.1103e-04, 1.7120e-02, ..., -3.1700e-03, + 6.7978e-03, -1.6479e-02], + [ 0.0000e+00, -6.2103e-03, 9.0408e-04, ..., -1.0384e-02, + -1.3420e-02, -3.3264e-03], + [ 0.0000e+00, 2.3117e-03, 4.7646e-03, ..., 1.5459e-03, + 4.7607e-03, 3.7270e-03]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0127, 0.0192, -0.0008, -0.0067, 0.0177, -0.0309, 0.0109, -0.0048, + 0.0034, -0.0115], device='cuda:0'), grad: tensor([-0.0558, -0.0441, 0.0076, 0.0169, 0.0490, -0.0013, 0.0370, 0.0051, + -0.0312, 0.0169], device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 214.70, cls_loss 0.7937 cls_loss_mapping 0.0359 cls_loss_causal 0.6811 re_mapping 0.0238 re_causal 0.0520 /// teacc 97.78 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0274, -0.0410, -0.0570, ..., -0.0190, 0.0396, -0.0601], + [-0.0225, -0.0583, -0.0235, ..., 0.0818, -0.0351, 0.1233], + [ 0.0200, -0.0282, -0.0090, ..., -0.0149, -0.0065, -0.0461], + ..., + [-0.0019, -0.0543, 0.0905, ..., 0.0070, -0.0338, 0.0298], + [-0.0086, 0.0411, -0.0628, ..., -0.0318, -0.0161, -0.0481], + [-0.0424, 0.0542, 0.0480, ..., -0.0424, -0.0206, -0.0133]], + device='cuda:0'), grad: tensor([[ 6.7614e-06, -2.1577e-04, 3.1395e-03, ..., 7.5607e-03, + -4.6272e-03, 1.2121e-03], + [ 9.0152e-06, 5.4359e-05, 3.2692e-03, ..., 1.8723e-02, + 3.4447e-03, 6.3229e-04], + [ 2.6369e-04, 9.8038e-04, -1.4343e-03, ..., -1.3405e-02, + 2.4128e-03, 1.4963e-03], + ..., + [-5.3692e-04, 1.5345e-03, -8.4381e-03, ..., -6.0387e-03, + 3.2425e-03, 1.6775e-03], + [ 9.3207e-06, 8.1599e-05, 2.5406e-03, ..., -9.1095e-03, + -2.4071e-03, -6.6817e-05], + [ 1.9953e-05, -2.5158e-03, 4.8399e-04, ..., 9.0637e-03, + 1.9264e-03, -2.2106e-03]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0125, 0.0190, -0.0008, -0.0062, 0.0171, -0.0302, 0.0113, -0.0053, + 0.0025, -0.0109], device='cuda:0'), grad: tensor([ 0.0277, 0.0346, -0.0205, 0.0072, -0.0176, -0.0400, 0.0084, -0.0038, + -0.0107, 0.0147], device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 214.58, cls_loss 0.7865 cls_loss_mapping 0.0434 cls_loss_causal 0.6664 re_mapping 0.0235 re_causal 0.0518 /// teacc 97.83 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0280, -0.0411, -0.0578, ..., -0.0205, 0.0392, -0.0620], + [-0.0227, -0.0586, -0.0253, ..., 0.0817, -0.0333, 0.1234], + [ 0.0205, -0.0285, -0.0089, ..., -0.0151, -0.0071, -0.0470], + ..., + [-0.0006, -0.0542, 0.0918, ..., 0.0077, -0.0357, 0.0295], + [-0.0074, 0.0412, -0.0629, ..., -0.0312, -0.0159, -0.0484], + [-0.0448, 0.0547, 0.0471, ..., -0.0434, -0.0196, -0.0135]], + device='cuda:0'), grad: tensor([[ 5.2601e-06, 2.4796e-03, -2.1291e-04, ..., 1.4076e-03, + 1.9646e-03, 4.1068e-05], + [ 2.7493e-06, 8.6355e-04, -9.3079e-03, ..., -2.1057e-02, + -2.1863e-04, -7.4100e-04], + [-5.3681e-06, 4.0550e-03, 6.1655e-04, ..., 8.0299e-04, + 3.8376e-03, 6.7115e-05], + ..., + [ 7.4618e-06, 1.1139e-02, -2.5558e-03, ..., 7.1106e-03, + -5.7650e-04, -3.4833e-04], + [ 2.3901e-05, -1.0681e-03, 1.1223e-02, ..., 1.2222e-02, + 6.2408e-03, 9.2268e-05], + [ 7.5810e-06, 1.4664e-02, 2.7817e-02, ..., 1.2016e-02, + 6.0730e-03, 2.6894e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0133, 0.0187, -0.0007, -0.0060, 0.0178, -0.0307, 0.0117, -0.0051, + 0.0030, -0.0117], device='cuda:0'), grad: tensor([ 0.0029, -0.0468, 0.0190, -0.0065, -0.0021, -0.0289, -0.0163, 0.0143, + 0.0153, 0.0491], device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 214.87, cls_loss 0.7794 cls_loss_mapping 0.0452 cls_loss_causal 0.6511 re_mapping 0.0243 re_causal 0.0499 /// teacc 98.04 lr 0.00010000 +Epoch 60, weight, value: tensor([[-2.8222e-02, -4.2292e-02, -5.7959e-02, ..., -1.9913e-02, + 3.9309e-02, -6.2043e-02], + [-2.1760e-02, -5.8907e-02, -2.5017e-02, ..., 8.2219e-02, + -3.3826e-02, 1.2416e-01], + [ 1.9955e-02, -2.8185e-02, -8.8367e-03, ..., -1.4431e-02, + -7.0914e-03, -4.7690e-02], + ..., + [-1.7030e-05, -5.5246e-02, 9.2046e-02, ..., 7.7005e-03, + -3.6705e-02, 2.9819e-02], + [-7.8637e-03, 4.1395e-02, -6.3567e-02, ..., -3.2979e-02, + -1.6744e-02, -4.8675e-02], + [-4.6252e-02, 5.4961e-02, 4.8190e-02, ..., -4.2455e-02, + -1.8769e-02, -1.3074e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3804e-04, 2.0580e-03, ..., 1.0277e-02, + -3.0022e-03, 5.0592e-04], + [ 0.0000e+00, 2.2469e-03, 1.9503e-03, ..., -1.2650e-02, + -7.1259e-03, -1.7643e-03], + [ 0.0000e+00, 3.9291e-04, -8.6136e-03, ..., -7.4806e-03, + 1.6880e-03, 7.6342e-04], + ..., + [ 0.0000e+00, -8.2970e-04, 9.5291e-03, ..., -3.5934e-03, + 5.0478e-06, -2.0180e-03], + [ 0.0000e+00, 2.0790e-03, -1.4658e-03, ..., 1.1091e-03, + 5.5599e-04, 1.7147e-03], + [ 0.0000e+00, -2.4834e-03, -2.0370e-03, ..., -2.8717e-02, + 4.3106e-04, -1.0042e-03]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0131, 0.0192, -0.0005, -0.0060, 0.0171, -0.0305, 0.0116, -0.0053, + 0.0022, -0.0111], device='cuda:0'), grad: tensor([ 0.0210, -0.0219, -0.0134, 0.0221, 0.0072, -0.0016, 0.0487, 0.0094, + -0.0053, -0.0661], device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 214.61, cls_loss 0.7733 cls_loss_mapping 0.0368 cls_loss_causal 0.6547 re_mapping 0.0247 re_causal 0.0509 /// teacc 97.97 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0295, -0.0430, -0.0592, ..., -0.0191, 0.0398, -0.0631], + [-0.0226, -0.0584, -0.0250, ..., 0.0823, -0.0348, 0.1254], + [ 0.0179, -0.0287, -0.0090, ..., -0.0141, -0.0063, -0.0483], + ..., + [-0.0005, -0.0557, 0.0929, ..., 0.0078, -0.0376, 0.0291], + [-0.0080, 0.0413, -0.0632, ..., -0.0337, -0.0167, -0.0483], + [-0.0467, 0.0550, 0.0483, ..., -0.0429, -0.0183, -0.0127]], + device='cuda:0'), grad: tensor([[ 2.2594e-06, -9.8133e-04, -1.9608e-03, ..., -8.5449e-03, + -8.3466e-03, -4.6110e-04], + [ 7.6815e-06, -7.4615e-03, 4.2305e-03, ..., 1.4324e-03, + 1.8680e-04, -4.1428e-03], + [ 7.8529e-06, -1.3489e-02, -1.1612e-02, ..., -1.6724e-02, + 1.8587e-03, -4.8561e-03], + ..., + [ 3.5707e-06, -8.3017e-04, -3.4618e-03, ..., -4.7455e-03, + 2.3758e-04, -7.1030e-03], + [-1.8135e-05, 5.5656e-03, 5.0125e-03, ..., 1.2085e-02, + -8.2850e-05, 4.9324e-03], + [ 1.2353e-05, 4.0092e-03, -1.1444e-03, ..., -3.1185e-03, + 5.5790e-04, 9.6273e-04]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0132, 0.0194, -0.0007, -0.0061, 0.0170, -0.0307, 0.0124, -0.0050, + 0.0021, -0.0112], device='cuda:0'), grad: tensor([-0.0207, 0.0097, -0.0433, -0.0087, -0.0021, 0.0204, 0.0264, -0.0080, + 0.0335, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 214.27, cls_loss 0.8048 cls_loss_mapping 0.0422 cls_loss_causal 0.6955 re_mapping 0.0233 re_causal 0.0530 /// teacc 98.07 lr 0.00010000 +Epoch 62, weight, value: tensor([[-2.9973e-02, -4.3493e-02, -5.9757e-02, ..., -1.9587e-02, + 3.9180e-02, -6.2672e-02], + [-2.3133e-02, -5.8316e-02, -2.5087e-02, ..., 8.2396e-02, + -3.4091e-02, 1.2474e-01], + [ 1.9940e-02, -2.7896e-02, -9.1051e-03, ..., -1.3478e-02, + -5.6789e-03, -4.8153e-02], + ..., + [-1.1881e-04, -5.6700e-02, 9.3152e-02, ..., 8.2287e-03, + -3.6931e-02, 2.8929e-02], + [-9.3391e-03, 4.2025e-02, -6.3434e-02, ..., -3.2429e-02, + -1.5159e-02, -4.8453e-02], + [-4.6007e-02, 5.4671e-02, 4.8128e-02, ..., -4.2771e-02, + -1.8978e-02, -1.2381e-02]], device='cuda:0'), grad: tensor([[ 9.9745e-07, 7.1001e-04, 2.6588e-03, ..., -3.6392e-03, + 3.1700e-03, 9.7752e-04], + [ 1.6379e-04, 5.4693e-04, 3.5057e-03, ..., -6.7043e-04, + -3.6831e-03, 3.7408e-04], + [ 1.1817e-05, 5.5923e-03, 9.0790e-03, ..., 2.0828e-02, + 3.3684e-03, 3.2024e-03], + ..., + [ 2.5257e-06, -8.8196e-03, -1.1375e-02, ..., -9.3079e-03, + 6.8331e-04, 3.5458e-03], + [-2.1708e-04, 8.4915e-03, 9.8648e-03, ..., 1.3138e-02, + 1.8120e-03, 2.5387e-03], + [ 1.5423e-05, 2.9182e-03, 7.6294e-04, ..., -1.4832e-02, + 1.1921e-03, 1.5345e-03]], device='cuda:0') +Epoch 62, bias, value: tensor([-1.3690e-02, 1.8916e-02, 6.9420e-05, -6.1791e-03, 1.7115e-02, + -3.1179e-02, 1.2333e-02, -5.3703e-03, 2.9190e-03, -1.1188e-02], + device='cuda:0'), grad: tensor([-0.0018, 0.0013, 0.0528, 0.0201, 0.0096, -0.0649, 0.0056, -0.0304, + 0.0374, -0.0298], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 61---------------------------------------------------- +epoch 61, time 230.86, cls_loss 0.7794 cls_loss_mapping 0.0370 cls_loss_causal 0.6688 re_mapping 0.0229 re_causal 0.0494 /// teacc 98.21 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0306, -0.0429, -0.0598, ..., -0.0199, 0.0400, -0.0637], + [-0.0237, -0.0591, -0.0257, ..., 0.0822, -0.0341, 0.1263], + [ 0.0189, -0.0276, -0.0105, ..., -0.0144, -0.0048, -0.0481], + ..., + [ 0.0009, -0.0571, 0.0941, ..., 0.0098, -0.0384, 0.0288], + [-0.0102, 0.0431, -0.0630, ..., -0.0326, -0.0153, -0.0489], + [-0.0470, 0.0550, 0.0481, ..., -0.0430, -0.0194, -0.0124]], + device='cuda:0'), grad: tensor([[ 1.6745e-06, 1.6606e-04, 1.4896e-03, ..., 5.7831e-03, + -1.3514e-03, 7.6866e-04], + [ 6.3324e-04, 9.5844e-04, -3.8662e-03, ..., -9.5444e-03, + 1.6749e-04, 2.6283e-03], + [ 2.1350e-04, 2.4128e-03, -1.5869e-02, ..., 1.3237e-02, + 9.1410e-04, -2.2247e-02], + ..., + [ 1.4293e-04, 3.7994e-03, 1.8265e-02, ..., -3.3245e-03, + 1.0771e-04, 2.3544e-02], + [-1.5526e-03, -1.6113e-02, -2.0981e-03, ..., -2.9800e-02, + -4.9019e-03, 2.9969e-04], + [ 1.4830e-04, -2.9049e-03, -3.0537e-03, ..., -5.4359e-03, + 9.3031e-04, -6.5918e-03]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0140, 0.0186, -0.0003, -0.0063, 0.0182, -0.0308, 0.0124, -0.0050, + 0.0027, -0.0117], device='cuda:0'), grad: tensor([ 0.0152, -0.0266, 0.0030, 0.0113, -0.0172, 0.0199, 0.0359, 0.0392, + -0.0588, -0.0219], device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.68, cls_loss 0.7705 cls_loss_mapping 0.0356 cls_loss_causal 0.6557 re_mapping 0.0230 re_causal 0.0515 /// teacc 98.09 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0307, -0.0440, -0.0601, ..., -0.0199, 0.0393, -0.0639], + [-0.0239, -0.0598, -0.0267, ..., 0.0826, -0.0345, 0.1269], + [ 0.0181, -0.0283, -0.0098, ..., -0.0143, -0.0049, -0.0472], + ..., + [-0.0003, -0.0567, 0.0939, ..., 0.0091, -0.0388, 0.0282], + [-0.0091, 0.0426, -0.0630, ..., -0.0327, -0.0155, -0.0478], + [-0.0466, 0.0547, 0.0480, ..., -0.0428, -0.0189, -0.0120]], + device='cuda:0'), grad: tensor([[ 2.6673e-06, -1.7529e-03, -5.4264e-04, ..., -1.7071e-03, + 4.7064e-04, 1.6320e-04], + [ 5.7705e-06, 7.3957e-04, 3.2425e-03, ..., 6.0005e-03, + 5.0116e-04, 6.4039e-04], + [ 1.0222e-04, 9.0551e-04, 3.0098e-03, ..., 3.3379e-03, + 9.6512e-04, 1.6677e-04], + ..., + [-3.0117e-03, -1.7365e-02, -9.9487e-03, ..., -7.9012e-04, + 9.9659e-04, -3.5076e-03], + [ 8.6451e-04, 5.0020e-04, 3.5820e-03, ..., -1.3208e-03, + -1.3435e-02, 1.6317e-03], + [ 2.0373e-04, -4.4937e-03, -3.9902e-03, ..., 1.3695e-03, + -3.7537e-03, -1.3443e-02]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0142, 0.0186, -0.0005, -0.0051, 0.0183, -0.0309, 0.0116, -0.0053, + 0.0030, -0.0119], device='cuda:0'), grad: tensor([-0.0104, 0.0216, 0.0129, -0.0236, 0.0143, 0.0364, 0.0154, -0.0318, + -0.0057, -0.0292], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 63---------------------------------------------------- +epoch 63, time 230.83, cls_loss 0.7837 cls_loss_mapping 0.0378 cls_loss_causal 0.6728 re_mapping 0.0227 re_causal 0.0491 /// teacc 98.27 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0326, -0.0453, -0.0611, ..., -0.0192, 0.0391, -0.0640], + [-0.0230, -0.0587, -0.0270, ..., 0.0826, -0.0354, 0.1274], + [ 0.0170, -0.0268, -0.0098, ..., -0.0138, -0.0057, -0.0483], + ..., + [-0.0009, -0.0574, 0.0942, ..., 0.0080, -0.0408, 0.0287], + [-0.0091, 0.0419, -0.0634, ..., -0.0326, -0.0146, -0.0484], + [-0.0448, 0.0554, 0.0484, ..., -0.0424, -0.0174, -0.0113]], + device='cuda:0'), grad: tensor([[ 4.3912e-07, 1.4811e-03, 2.3937e-03, ..., 8.1253e-03, + 2.9400e-05, 9.4128e-04], + [ 4.7199e-06, 1.0109e-03, -1.5612e-03, ..., -8.7814e-03, + 6.8685e-07, 4.5705e-04], + [ 1.0282e-06, 2.6016e-03, 3.0613e-03, ..., 9.9030e-03, + 5.4240e-05, 1.4029e-03], + ..., + [ 2.6217e-07, 1.5478e-03, -1.1444e-03, ..., 6.8779e-03, + 1.1949e-06, -9.5081e-04], + [-1.1124e-05, 1.3397e-02, 6.6681e-03, ..., -8.6136e-03, + 2.8476e-05, -1.5984e-03], + [ 3.6368e-07, 2.7275e-03, 7.0038e-03, ..., 4.1351e-03, + 7.7263e-06, -4.7994e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0139, 0.0188, -0.0002, -0.0055, 0.0183, -0.0315, 0.0111, -0.0058, + 0.0030, -0.0106], device='cuda:0'), grad: tensor([ 0.0195, -0.0193, 0.0237, -0.0790, 0.0264, 0.0042, 0.0154, 0.0081, + -0.0133, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 214.88, cls_loss 0.7728 cls_loss_mapping 0.0391 cls_loss_causal 0.6462 re_mapping 0.0222 re_causal 0.0479 /// teacc 97.98 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0344, -0.0472, -0.0617, ..., -0.0193, 0.0401, -0.0652], + [-0.0217, -0.0594, -0.0277, ..., 0.0833, -0.0355, 0.1281], + [ 0.0155, -0.0271, -0.0095, ..., -0.0139, -0.0054, -0.0483], + ..., + [-0.0014, -0.0578, 0.0946, ..., 0.0081, -0.0403, 0.0296], + [-0.0095, 0.0419, -0.0644, ..., -0.0323, -0.0147, -0.0486], + [-0.0451, 0.0566, 0.0486, ..., -0.0433, -0.0183, -0.0119]], + device='cuda:0'), grad: tensor([[ 1.6677e-04, -4.5433e-03, -8.5068e-03, ..., 3.1967e-03, + 1.6190e-02, 6.0940e-04], + [ 1.1854e-03, 4.2498e-05, -1.0643e-03, ..., -1.1740e-03, + 3.9339e-04, 1.8749e-03], + [ 3.7241e-04, 6.5470e-04, 4.0855e-03, ..., 3.0403e-03, + 7.3280e-03, 5.0157e-05], + ..., + [ 1.7130e-04, 1.5440e-03, -4.8709e-04, ..., 8.1003e-05, + 6.8550e-03, 1.7166e-03], + [ 7.4244e-04, 2.2936e-04, 2.4891e-03, ..., 7.2670e-03, + 8.3466e-03, 1.8167e-03], + [-8.8577e-03, -2.0111e-02, -2.2583e-02, ..., -1.7807e-02, + -5.4016e-02, -6.9122e-03]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0139, 0.0191, -0.0003, -0.0053, 0.0183, -0.0308, 0.0105, -0.0057, + 0.0027, -0.0110], device='cuda:0'), grad: tensor([-0.0168, -0.0016, 0.0166, 0.0699, 0.0384, -0.0055, -0.0444, 0.0156, + 0.0272, -0.0992], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 65---------------------------------------------------- +epoch 65, time 232.11, cls_loss 0.7753 cls_loss_mapping 0.0314 cls_loss_causal 0.6593 re_mapping 0.0221 re_causal 0.0490 /// teacc 98.29 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0355, -0.0474, -0.0616, ..., -0.0192, 0.0389, -0.0653], + [-0.0225, -0.0593, -0.0276, ..., 0.0830, -0.0346, 0.1294], + [ 0.0157, -0.0263, -0.0092, ..., -0.0141, -0.0050, -0.0467], + ..., + [-0.0007, -0.0582, 0.0955, ..., 0.0091, -0.0400, 0.0288], + [-0.0113, 0.0424, -0.0649, ..., -0.0326, -0.0143, -0.0493], + [-0.0434, 0.0561, 0.0489, ..., -0.0430, -0.0177, -0.0123]], + device='cuda:0'), grad: tensor([[ 9.2611e-06, 8.1003e-05, -1.5717e-03, ..., -4.5929e-03, + -2.4700e-03, 1.6861e-03], + [ 8.3596e-06, 8.6737e-04, -4.8780e-04, ..., 2.8439e-03, + 4.3321e-04, 1.6193e-03], + [-7.5638e-05, 2.4199e-04, -2.0847e-03, ..., -1.1909e-02, + 4.4136e-03, -3.3975e-04], + ..., + [ 1.3068e-05, -2.9030e-03, 6.7997e-04, ..., 2.2163e-03, + 1.1883e-03, 8.9979e-04], + [ 9.9745e-07, -1.2178e-03, -1.9341e-03, ..., 2.4071e-03, + -1.0475e-02, -2.4662e-03], + [ 3.2276e-05, 6.7596e-03, -5.2166e-04, ..., 2.5272e-03, + -2.4738e-03, -2.0046e-03]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0135, 0.0191, -0.0001, -0.0059, 0.0181, -0.0304, 0.0099, -0.0052, + 0.0026, -0.0108], device='cuda:0'), grad: tensor([-0.0199, -0.0025, -0.0133, 0.0071, 0.0314, 0.0170, -0.0057, 0.0226, + -0.0180, -0.0188], device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.16, cls_loss 0.7631 cls_loss_mapping 0.0317 cls_loss_causal 0.6550 re_mapping 0.0221 re_causal 0.0480 /// teacc 98.26 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0370, -0.0488, -0.0625, ..., -0.0198, 0.0393, -0.0666], + [-0.0237, -0.0591, -0.0289, ..., 0.0828, -0.0340, 0.1305], + [ 0.0175, -0.0263, -0.0096, ..., -0.0140, -0.0050, -0.0478], + ..., + [-0.0009, -0.0578, 0.0963, ..., 0.0099, -0.0402, 0.0301], + [-0.0094, 0.0427, -0.0651, ..., -0.0329, -0.0143, -0.0486], + [-0.0449, 0.0568, 0.0498, ..., -0.0435, -0.0176, -0.0139]], + device='cuda:0'), grad: tensor([[ 3.9814e-08, 7.4720e-04, 1.2674e-03, ..., 3.8013e-03, + 1.8063e-03, -6.5267e-05], + [ 4.0513e-08, 4.4680e-04, 7.0906e-04, ..., 1.9102e-03, + 5.2881e-04, -1.4588e-05], + [-9.9558e-07, 1.8463e-03, 9.9850e-04, ..., -8.0185e-03, + -1.4658e-03, 6.7949e-05], + ..., + [ 3.9600e-06, 2.0771e-03, -3.0003e-03, ..., 3.6201e-03, + 2.7752e-03, 6.9439e-05], + [ 1.1539e-06, -5.8594e-03, -5.6343e-03, ..., -1.8158e-02, + -1.3412e-02, 2.2755e-03], + [-5.5134e-06, -1.6724e-02, -5.6763e-03, ..., -3.6564e-03, + -7.8888e-03, -4.2114e-03]], device='cuda:0') +Epoch 68, bias, value: tensor([-1.4135e-02, 1.9191e-02, -8.1638e-05, -5.9446e-03, 1.8357e-02, + -3.0554e-02, 9.8137e-03, -4.4962e-03, 2.6353e-03, -1.1173e-02], + device='cuda:0'), grad: tensor([-0.0096, 0.0060, -0.0197, 0.0282, 0.0213, 0.0330, -0.0038, 0.0100, + -0.0257, -0.0398], device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.45, cls_loss 0.7545 cls_loss_mapping 0.0340 cls_loss_causal 0.6415 re_mapping 0.0215 re_causal 0.0458 /// teacc 98.15 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0382, -0.0507, -0.0631, ..., -0.0204, 0.0395, -0.0672], + [-0.0236, -0.0589, -0.0285, ..., 0.0831, -0.0355, 0.1321], + [ 0.0170, -0.0256, -0.0092, ..., -0.0131, -0.0042, -0.0481], + ..., + [ 0.0018, -0.0601, 0.0967, ..., 0.0095, -0.0413, 0.0300], + [-0.0115, 0.0424, -0.0655, ..., -0.0331, -0.0132, -0.0492], + [-0.0474, 0.0582, 0.0494, ..., -0.0444, -0.0164, -0.0135]], + device='cuda:0'), grad: tensor([[ 1.3001e-06, -9.3079e-04, 7.4768e-04, ..., 1.9894e-03, + -4.0627e-03, 1.2627e-03], + [ 1.2413e-05, 2.6035e-04, 1.2035e-03, ..., -4.9639e-04, + 8.6355e-04, 1.2512e-03], + [ 7.5661e-06, -1.8768e-03, 1.5240e-03, ..., -2.8381e-03, + -5.9967e-03, 1.1606e-03], + ..., + [ 8.4257e-04, 1.0033e-03, -7.0095e-04, ..., 4.4365e-03, + 4.0126e-04, 2.7637e-03], + [ 3.0369e-05, 1.2465e-03, 2.1286e-03, ..., 4.9057e-03, + 1.6603e-03, 2.9831e-03], + [-9.8133e-04, 7.2956e-05, -1.4508e-04, ..., 5.0240e-03, + 1.3380e-03, -4.1351e-03]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0140, 0.0200, 0.0008, -0.0058, 0.0184, -0.0304, 0.0098, -0.0049, + 0.0020, -0.0120], device='cuda:0'), grad: tensor([ 0.0073, -0.0008, 0.0006, 0.0038, -0.0111, -0.0410, 0.0114, 0.0206, + 0.0278, -0.0186], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 68---------------------------------------------------- +epoch 68, time 231.11, cls_loss 0.7596 cls_loss_mapping 0.0393 cls_loss_causal 0.6566 re_mapping 0.0215 re_causal 0.0465 /// teacc 98.34 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0387, -0.0513, -0.0632, ..., -0.0212, 0.0392, -0.0667], + [-0.0227, -0.0596, -0.0282, ..., 0.0829, -0.0359, 0.1326], + [ 0.0165, -0.0255, -0.0092, ..., -0.0124, -0.0042, -0.0482], + ..., + [ 0.0024, -0.0595, 0.0977, ..., 0.0102, -0.0397, 0.0302], + [-0.0111, 0.0430, -0.0656, ..., -0.0341, -0.0128, -0.0498], + [-0.0482, 0.0571, 0.0490, ..., -0.0434, -0.0184, -0.0141]], + device='cuda:0'), grad: tensor([[ 1.7812e-07, 2.5058e-04, 1.9197e-03, ..., -9.7427e-03, + 4.6945e-04, 1.2751e-03], + [ 3.1502e-07, 4.9114e-04, -3.2978e-03, ..., -1.8585e-02, + -2.6989e-03, -1.4404e-02], + [ 5.7183e-06, 2.2113e-04, 6.9237e-04, ..., 3.5973e-03, + 1.5049e-03, 6.8331e-04], + ..., + [ 5.6624e-07, 5.2404e-04, -1.0323e-02, ..., -5.0545e-03, + -3.5524e-04, -1.0700e-03], + [-1.8790e-05, -1.5488e-03, 2.6474e-03, ..., 5.9128e-03, + -4.5240e-05, 1.1377e-03], + [ 1.8273e-06, 8.4400e-04, -7.8964e-03, ..., -5.5122e-03, + 1.1225e-03, 1.2789e-03]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0141, 0.0200, 0.0006, -0.0059, 0.0189, -0.0293, 0.0095, -0.0045, + 0.0007, -0.0120], device='cuda:0'), grad: tensor([-0.0359, -0.0536, 0.0173, 0.0609, 0.0135, -0.0020, 0.0056, -0.0097, + 0.0153, -0.0113], device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 214.81, cls_loss 0.7298 cls_loss_mapping 0.0259 cls_loss_causal 0.6206 re_mapping 0.0213 re_causal 0.0466 /// teacc 98.22 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0387, -0.0529, -0.0630, ..., -0.0210, 0.0387, -0.0673], + [-0.0227, -0.0585, -0.0291, ..., 0.0827, -0.0351, 0.1338], + [ 0.0168, -0.0251, -0.0088, ..., -0.0118, -0.0045, -0.0469], + ..., + [ 0.0023, -0.0604, 0.0975, ..., 0.0097, -0.0414, 0.0299], + [-0.0109, 0.0418, -0.0665, ..., -0.0339, -0.0138, -0.0492], + [-0.0480, 0.0564, 0.0491, ..., -0.0439, -0.0184, -0.0145]], + device='cuda:0'), grad: tensor([[ 1.8984e-05, -6.7596e-03, 1.3733e-03, ..., 2.1553e-03, + -1.1292e-02, -3.6072e-02], + [ 1.0364e-05, 1.9054e-03, -3.5673e-05, ..., 1.5554e-03, + 2.5597e-03, 1.9623e-02], + [ 5.0116e-04, 5.1384e-03, 4.1618e-03, ..., 6.6605e-03, + 3.6454e-04, 3.0670e-03], + ..., + [ 1.4603e-04, 2.5768e-03, 2.0695e-03, ..., 9.5215e-03, + 2.2144e-03, 1.9531e-03], + [ 2.2471e-05, 7.8125e-03, -6.0234e-03, ..., -7.0534e-03, + 1.0567e-02, 2.5158e-03], + [ 4.9204e-05, 4.9019e-03, 3.5038e-03, ..., 1.0025e-02, + 3.0479e-03, 3.5610e-03]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0138, 0.0199, 0.0015, -0.0056, 0.0190, -0.0303, 0.0099, -0.0056, + 0.0007, -0.0119], device='cuda:0'), grad: tensor([-0.0588, -0.0026, 0.0392, -0.0612, -0.0289, 0.0340, 0.0022, 0.0341, + -0.0041, 0.0461], device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 214.67, cls_loss 0.7333 cls_loss_mapping 0.0246 cls_loss_causal 0.6204 re_mapping 0.0207 re_causal 0.0462 /// teacc 98.19 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0396, -0.0548, -0.0628, ..., -0.0207, 0.0388, -0.0675], + [-0.0229, -0.0586, -0.0292, ..., 0.0828, -0.0355, 0.1342], + [ 0.0162, -0.0225, -0.0080, ..., -0.0118, -0.0037, -0.0474], + ..., + [ 0.0031, -0.0605, 0.0979, ..., 0.0103, -0.0411, 0.0302], + [-0.0112, 0.0413, -0.0666, ..., -0.0337, -0.0139, -0.0483], + [-0.0478, 0.0558, 0.0486, ..., -0.0448, -0.0185, -0.0151]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.0996e-03, -1.1139e-02, ..., 2.9964e-03, + 1.1757e-02, -5.5647e-04], + [ 2.5611e-09, 3.4595e-04, 2.7084e-03, ..., 7.3314e-06, + -1.6006e-02, 7.6962e-04], + [ 1.2945e-07, 1.6069e-03, 1.5068e-03, ..., -3.8414e-03, + -2.3413e-04, -1.4067e-03], + ..., + [ 4.8429e-08, 6.7949e-04, 1.0963e-02, ..., -4.9438e-03, + 1.9658e-04, -1.9531e-03], + [ 7.7998e-08, 9.8610e-04, 2.7065e-03, ..., 1.0231e-02, + 5.2166e-04, 3.7155e-03], + [ 9.3132e-10, -2.9869e-03, -4.7493e-03, ..., -6.9847e-03, + 9.4712e-05, -2.7466e-03]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0132, 0.0198, 0.0019, -0.0067, 0.0180, -0.0307, 0.0115, -0.0049, + 0.0006, -0.0123], device='cuda:0'), grad: tensor([-6.2485e-03, 1.3493e-05, -9.1705e-03, -2.8381e-03, -1.4420e-02, + 6.3820e-03, 2.0615e-02, -2.2352e-07, 2.0615e-02, -1.4954e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 214.65, cls_loss 0.7500 cls_loss_mapping 0.0294 cls_loss_causal 0.6426 re_mapping 0.0207 re_causal 0.0457 /// teacc 98.27 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0404, -0.0552, -0.0627, ..., -0.0207, 0.0399, -0.0676], + [-0.0233, -0.0594, -0.0299, ..., 0.0829, -0.0352, 0.1347], + [ 0.0159, -0.0227, -0.0082, ..., -0.0115, -0.0052, -0.0473], + ..., + [ 0.0040, -0.0605, 0.0977, ..., 0.0108, -0.0415, 0.0315], + [-0.0112, 0.0434, -0.0660, ..., -0.0342, -0.0119, -0.0495], + [-0.0474, 0.0549, 0.0495, ..., -0.0450, -0.0194, -0.0159]], + device='cuda:0'), grad: tensor([[ 3.9348e-07, 1.4868e-03, -4.0512e-03, ..., -6.4163e-03, + 6.1417e-03, -2.2907e-03], + [ 1.8068e-06, 1.5652e-04, 3.1548e-03, ..., 1.8982e-02, + 3.1090e-04, 7.9269e-03], + [ 2.7083e-06, 6.1035e-04, 3.1261e-03, ..., 1.4297e-02, + 1.3409e-03, 4.5395e-03], + ..., + [ 9.3281e-06, 6.8426e-05, -1.0681e-02, ..., -6.9656e-03, + 6.6102e-05, -8.5144e-03], + [-1.1988e-05, -1.4282e-02, -1.3733e-03, ..., -2.1942e-02, + -5.2277e-02, -9.9869e-03], + [-5.4419e-05, 5.9366e-04, 4.1270e-04, ..., 2.5082e-03, + 4.4560e-04, 1.2455e-03]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0131, 0.0196, 0.0016, -0.0070, 0.0187, -0.0303, 0.0106, -0.0046, + 0.0006, -0.0121], device='cuda:0'), grad: tensor([-0.0124, 0.0369, 0.0306, 0.0413, 0.0770, -0.0125, -0.0566, -0.0253, + -0.0781, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.69, cls_loss 0.7289 cls_loss_mapping 0.0266 cls_loss_causal 0.6123 re_mapping 0.0213 re_causal 0.0458 /// teacc 98.08 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0409, -0.0567, -0.0632, ..., -0.0215, 0.0397, -0.0678], + [-0.0232, -0.0598, -0.0301, ..., 0.0834, -0.0358, 0.1352], + [ 0.0155, -0.0227, -0.0084, ..., -0.0120, -0.0056, -0.0483], + ..., + [ 0.0046, -0.0608, 0.0989, ..., 0.0113, -0.0414, 0.0327], + [-0.0113, 0.0425, -0.0673, ..., -0.0338, -0.0114, -0.0492], + [-0.0475, 0.0549, 0.0493, ..., -0.0448, -0.0194, -0.0167]], + device='cuda:0'), grad: tensor([[ 2.3972e-06, 4.1151e-04, 6.4421e-04, ..., 5.3520e-03, + -3.7270e-03, 1.5326e-03], + [ 9.8571e-06, 2.0230e-04, 1.8082e-03, ..., 3.4351e-03, + 2.0218e-03, -3.0956e-03], + [ 2.6875e-03, 1.9255e-03, 1.3435e-04, ..., 5.3711e-03, + 2.7599e-03, -5.6148e-05], + ..., + [ 2.6464e-04, 5.4436e-03, -5.8289e-03, ..., -2.2869e-03, + -2.5406e-03, 6.9580e-03], + [ 9.4414e-05, 3.6896e-02, 2.5272e-03, ..., 2.0935e-02, + 9.0256e-03, 3.1586e-03], + [ 5.2392e-05, -3.8033e-03, 1.1360e-02, ..., 1.3924e-02, + 4.1504e-03, 1.6260e-03]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0137, 0.0202, 0.0017, -0.0076, 0.0181, -0.0302, 0.0112, -0.0043, + 0.0012, -0.0126], device='cuda:0'), grad: tensor([ 0.0058, 0.0084, 0.0143, -0.0567, -0.0392, 0.0264, -0.0383, -0.0016, + 0.0536, 0.0274], device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 214.84, cls_loss 0.7446 cls_loss_mapping 0.0360 cls_loss_causal 0.6316 re_mapping 0.0200 re_causal 0.0447 /// teacc 98.16 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0413, -0.0580, -0.0633, ..., -0.0205, 0.0394, -0.0662], + [-0.0243, -0.0593, -0.0299, ..., 0.0839, -0.0334, 0.1357], + [ 0.0150, -0.0226, -0.0086, ..., -0.0116, -0.0049, -0.0477], + ..., + [ 0.0048, -0.0611, 0.0984, ..., 0.0113, -0.0411, 0.0308], + [-0.0102, 0.0423, -0.0674, ..., -0.0332, -0.0110, -0.0497], + [-0.0474, 0.0556, 0.0501, ..., -0.0450, -0.0202, -0.0153]], + device='cuda:0'), grad: tensor([[ 7.1805e-07, -1.3342e-03, -2.5139e-03, ..., -1.4526e-02, + -1.3847e-02, -2.3899e-03], + [ 2.1422e-04, 2.1133e-03, 2.5201e-04, ..., -4.2953e-03, + 3.5725e-03, 8.9824e-05], + [ 5.3868e-06, -5.6854e-02, 1.3790e-03, ..., -1.0767e-03, + -1.4696e-03, -4.1656e-03], + ..., + [ 3.0327e-04, 4.4594e-03, -1.7967e-03, ..., 5.5351e-03, + -2.1515e-02, 3.0384e-03], + [ 2.7135e-05, 5.6953e-03, 4.0603e-04, ..., -2.2659e-03, + -1.1368e-03, 1.5860e-03], + [-7.7629e-04, 4.1580e-03, 4.7135e-04, ..., -1.5213e-02, + 2.9640e-03, -1.0841e-02]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0127, 0.0207, 0.0017, -0.0079, 0.0177, -0.0309, 0.0102, -0.0045, + 0.0018, -0.0121], device='cuda:0'), grad: tensor([-0.0404, -0.0019, -0.0294, 0.0311, 0.0255, 0.0341, 0.0291, -0.0131, + 0.0048, -0.0398], device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 214.34, cls_loss 0.7553 cls_loss_mapping 0.0307 cls_loss_causal 0.6321 re_mapping 0.0198 re_causal 0.0435 /// teacc 98.16 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0427, -0.0578, -0.0639, ..., -0.0209, 0.0396, -0.0642], + [-0.0252, -0.0588, -0.0298, ..., 0.0842, -0.0326, 0.1370], + [ 0.0147, -0.0221, -0.0086, ..., -0.0114, -0.0044, -0.0481], + ..., + [ 0.0072, -0.0615, 0.0995, ..., 0.0113, -0.0419, 0.0319], + [-0.0100, 0.0421, -0.0676, ..., -0.0331, -0.0107, -0.0502], + [-0.0469, 0.0557, 0.0490, ..., -0.0464, -0.0209, -0.0169]], + device='cuda:0'), grad: tensor([[ 9.7677e-06, 1.6317e-03, 8.9741e-04, ..., 1.2993e-02, + 1.5488e-03, 6.4926e-03], + [ 3.4750e-05, 4.3526e-03, 2.5425e-03, ..., 1.2001e-02, + 1.8816e-03, 2.5692e-03], + [ 4.2468e-05, -5.0011e-03, 2.3727e-03, ..., 9.6817e-03, + -1.1765e-02, 6.7940e-03], + ..., + [-5.7268e-04, 2.5253e-03, -5.4665e-03, ..., -1.1299e-02, + 1.2197e-03, -3.1328e-04], + [ 9.5844e-05, 9.9945e-03, 6.6490e-03, ..., 1.7319e-02, + 4.5013e-03, 1.7366e-03], + [ 1.9932e-04, 1.1921e-03, 2.3232e-03, ..., 8.4457e-03, + 1.1730e-03, 4.3259e-03]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0128, 0.0205, 0.0024, -0.0083, 0.0176, -0.0314, 0.0111, -0.0044, + 0.0023, -0.0130], device='cuda:0'), grad: tensor([ 0.0309, 0.0181, 0.0038, -0.0706, -0.0091, 0.0090, -0.0172, -0.0278, + 0.0403, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 214.44, cls_loss 0.7261 cls_loss_mapping 0.0275 cls_loss_causal 0.6276 re_mapping 0.0201 re_causal 0.0451 /// teacc 98.22 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0423, -0.0584, -0.0645, ..., -0.0221, 0.0404, -0.0659], + [-0.0254, -0.0599, -0.0298, ..., 0.0836, -0.0331, 0.1357], + [ 0.0150, -0.0216, -0.0075, ..., -0.0109, -0.0039, -0.0475], + ..., + [ 0.0067, -0.0620, 0.0989, ..., 0.0114, -0.0426, 0.0321], + [-0.0102, 0.0421, -0.0692, ..., -0.0329, -0.0108, -0.0497], + [-0.0466, 0.0557, 0.0493, ..., -0.0468, -0.0213, -0.0174]], + device='cuda:0'), grad: tensor([[ 8.9686e-07, 4.3225e-04, 1.6451e-04, ..., -1.9913e-02, + -1.0986e-02, -1.1950e-03], + [ 3.5744e-06, 3.3689e-04, 7.0667e-04, ..., 3.9215e-03, + 8.7595e-04, -8.0185e-03], + [ 6.8247e-05, 1.0994e-02, 7.0286e-04, ..., 1.9318e-02, + 2.2324e-02, 3.5667e-03], + ..., + [ 3.4302e-05, 5.1641e-04, 1.0214e-03, ..., -3.1223e-03, + 2.4414e-03, 4.2319e-04], + [ 9.7036e-05, -2.9556e-02, 1.3056e-03, ..., 1.1200e-02, + -4.4220e-02, 9.4299e-03], + [ 4.7386e-06, 5.5075e-04, 1.4734e-03, ..., 6.3782e-03, + 1.9150e-03, 2.2697e-03]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0128, 0.0192, 0.0035, -0.0079, 0.0185, -0.0311, 0.0105, -0.0047, + 0.0021, -0.0133], device='cuda:0'), grad: tensor([-0.0467, 0.0068, 0.0537, -0.0875, 0.0252, 0.0047, 0.0100, -0.0129, + 0.0226, 0.0241], device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 214.18, cls_loss 0.7315 cls_loss_mapping 0.0290 cls_loss_causal 0.6198 re_mapping 0.0196 re_causal 0.0433 /// teacc 98.16 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0438, -0.0580, -0.0643, ..., -0.0222, 0.0408, -0.0651], + [-0.0264, -0.0610, -0.0301, ..., 0.0837, -0.0332, 0.1366], + [ 0.0157, -0.0204, -0.0082, ..., -0.0114, -0.0041, -0.0482], + ..., + [ 0.0060, -0.0622, 0.0993, ..., 0.0116, -0.0434, 0.0313], + [-0.0107, 0.0412, -0.0698, ..., -0.0332, -0.0107, -0.0504], + [-0.0472, 0.0570, 0.0496, ..., -0.0468, -0.0213, -0.0169]], + device='cuda:0'), grad: tensor([[ 3.8624e-04, -3.2902e-03, -1.3816e-04, ..., -8.7662e-03, + -3.5691e-04, -1.7405e-03], + [-1.7633e-03, 2.0370e-03, 1.0004e-03, ..., -3.0212e-03, + 9.1457e-04, -2.2411e-03], + [ 2.3365e-03, 1.2960e-03, 1.4067e-03, ..., -3.0937e-03, + 5.2357e-04, 1.8501e-03], + ..., + [ 1.1520e-03, 6.6137e-04, 5.7364e-04, ..., 3.6449e-03, + 2.0659e-04, 2.7256e-03], + [-4.2953e-03, -4.7493e-03, -1.3018e-03, ..., -6.8665e-03, + -3.4504e-03, -7.4615e-03], + [ 9.6202e-05, 6.5422e-04, 6.8665e-04, ..., 3.8509e-03, + 2.2995e-04, 1.7843e-03]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0123, 0.0194, 0.0022, -0.0072, 0.0185, -0.0316, 0.0102, -0.0046, + 0.0028, -0.0135], device='cuda:0'), grad: tensor([-0.0218, -0.0436, -0.0061, 0.0522, -0.0095, 0.0045, 0.0126, 0.0147, + -0.0184, 0.0155], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 77---------------------------------------------------- +epoch 77, time 230.68, cls_loss 0.7053 cls_loss_mapping 0.0269 cls_loss_causal 0.5878 re_mapping 0.0199 re_causal 0.0439 /// teacc 98.47 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0446, -0.0581, -0.0653, ..., -0.0223, 0.0411, -0.0657], + [-0.0272, -0.0609, -0.0303, ..., 0.0836, -0.0333, 0.1370], + [ 0.0164, -0.0210, -0.0086, ..., -0.0117, -0.0038, -0.0477], + ..., + [ 0.0065, -0.0627, 0.0998, ..., 0.0122, -0.0433, 0.0318], + [-0.0105, 0.0412, -0.0710, ..., -0.0340, -0.0114, -0.0526], + [-0.0477, 0.0564, 0.0489, ..., -0.0469, -0.0211, -0.0174]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, 7.2527e-04, 1.3275e-03, ..., -7.3433e-03, + 2.2430e-03, 1.2082e-04], + [-4.4227e-05, -1.5736e-03, -5.0116e-04, ..., -3.1281e-03, + -1.7273e-02, -1.7426e-02], + [ 7.1041e-06, 2.4366e-04, 1.6260e-03, ..., -2.5864e-03, + 3.9978e-03, 5.3215e-03], + ..., + [ 2.1961e-06, -5.8098e-03, -4.4174e-03, ..., 2.0542e-03, + 4.0889e-04, -7.1907e-04], + [ 1.8835e-05, 4.6158e-03, 9.3222e-04, ..., 3.2959e-03, + 4.5242e-03, 3.4599e-03], + [ 1.8207e-06, 2.0847e-03, 2.4967e-03, ..., 4.8103e-03, + 1.5984e-03, 4.7302e-03]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0126, 0.0196, 0.0021, -0.0069, 0.0188, -0.0317, 0.0108, -0.0042, + 0.0016, -0.0134], device='cuda:0'), grad: tensor([-0.0130, -0.0419, 0.0103, -0.0080, 0.0265, -0.0212, -0.0053, 0.0064, + 0.0287, 0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 214.18, cls_loss 0.7602 cls_loss_mapping 0.0275 cls_loss_causal 0.6495 re_mapping 0.0194 re_causal 0.0444 /// teacc 98.12 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0446, -0.0590, -0.0658, ..., -0.0215, 0.0409, -0.0649], + [-0.0262, -0.0611, -0.0303, ..., 0.0845, -0.0326, 0.1382], + [ 0.0167, -0.0212, -0.0097, ..., -0.0111, -0.0041, -0.0486], + ..., + [ 0.0074, -0.0632, 0.1001, ..., 0.0114, -0.0444, 0.0312], + [-0.0114, 0.0416, -0.0714, ..., -0.0348, -0.0111, -0.0527], + [-0.0483, 0.0578, 0.0495, ..., -0.0473, -0.0200, -0.0174]], + device='cuda:0'), grad: tensor([[ 1.8422e-06, -2.7714e-03, -6.5956e-03, ..., -6.0730e-03, + -2.3060e-03, -4.9896e-03], + [ 2.7633e-04, 1.1559e-03, 1.6937e-03, ..., 1.0681e-02, + 1.3132e-03, 3.0441e-03], + [ 5.5075e-05, 8.9598e-04, 1.9531e-03, ..., 6.1913e-03, + 7.2718e-04, 2.0332e-03], + ..., + [ 3.1531e-05, 5.7373e-03, 8.8959e-03, ..., 7.4463e-03, + 1.6232e-03, 8.8425e-03], + [-4.1246e-04, 1.5659e-03, -7.7343e-04, ..., 1.7061e-03, + 3.5496e-03, -2.7580e-03], + [ 1.6540e-05, -4.9362e-03, -3.9864e-03, ..., -2.1915e-03, + 1.1429e-02, -7.1640e-03]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0126, 0.0198, 0.0019, -0.0067, 0.0186, -0.0312, 0.0107, -0.0045, + 0.0012, -0.0132], device='cuda:0'), grad: tensor([-0.0165, 0.0303, 0.0198, -0.0184, -0.0507, 0.0205, -0.0080, 0.0316, + 0.0025, -0.0111], device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.39, cls_loss 0.7230 cls_loss_mapping 0.0254 cls_loss_causal 0.6186 re_mapping 0.0194 re_causal 0.0438 /// teacc 98.31 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0458, -0.0603, -0.0659, ..., -0.0225, 0.0405, -0.0649], + [-0.0276, -0.0600, -0.0304, ..., 0.0843, -0.0309, 0.1386], + [ 0.0159, -0.0208, -0.0105, ..., -0.0110, -0.0045, -0.0496], + ..., + [ 0.0080, -0.0630, 0.1004, ..., 0.0119, -0.0446, 0.0318], + [-0.0111, 0.0415, -0.0712, ..., -0.0350, -0.0109, -0.0528], + [-0.0488, 0.0575, 0.0497, ..., -0.0469, -0.0203, -0.0176]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3113e-03, 7.7784e-05, ..., 2.9545e-03, + 4.6730e-04, 5.7077e-04], + [ 0.0000e+00, 1.1093e-02, 6.6071e-03, ..., 5.3749e-03, + 2.3098e-03, 3.8147e-03], + [ 0.0000e+00, 3.9253e-03, 8.8882e-04, ..., 1.0368e-02, + 9.4604e-04, 9.6416e-04], + ..., + [ 3.2596e-09, -5.0049e-03, -1.5114e-02, ..., -1.1925e-02, + -1.1383e-02, 4.8971e-04], + [ 1.5832e-07, 9.2316e-03, 7.6752e-03, ..., 1.3397e-02, + 4.6692e-03, 5.3329e-03], + [-9.1270e-07, -1.4175e-02, -6.3133e-03, ..., -6.1569e-03, + 1.6346e-03, -1.3313e-02]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0138, 0.0191, 0.0023, -0.0066, 0.0191, -0.0308, 0.0112, -0.0043, + 0.0012, -0.0134], device='cuda:0'), grad: tensor([ 0.0145, 0.0214, 0.0332, 0.0003, -0.0131, 0.0251, -0.0114, -0.0977, + 0.0464, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 214.35, cls_loss 0.7162 cls_loss_mapping 0.0221 cls_loss_causal 0.6147 re_mapping 0.0191 re_causal 0.0417 /// teacc 98.27 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0454, -0.0601, -0.0667, ..., -0.0224, 0.0411, -0.0646], + [-0.0291, -0.0605, -0.0316, ..., 0.0847, -0.0314, 0.1385], + [ 0.0163, -0.0202, -0.0098, ..., -0.0100, -0.0041, -0.0489], + ..., + [ 0.0085, -0.0647, 0.1013, ..., 0.0112, -0.0451, 0.0326], + [-0.0114, 0.0422, -0.0715, ..., -0.0361, -0.0103, -0.0534], + [-0.0490, 0.0571, 0.0499, ..., -0.0460, -0.0202, -0.0176]], + device='cuda:0'), grad: tensor([[ 9.6299e-07, -4.2496e-03, -1.5521e-04, ..., -8.4610e-03, + -1.2779e-03, -2.9736e-03], + [ 8.4490e-06, 7.2250e-03, 1.1986e-02, ..., 3.9978e-02, + 9.1248e-03, 2.6108e-02], + [-7.1764e-05, -1.5106e-03, -4.8561e-03, ..., -4.7731e-04, + 1.2326e-04, -1.7309e-03], + ..., + [ 6.5714e-06, 2.7733e-03, 4.3259e-03, ..., 1.2932e-02, + 1.0371e-04, 7.0915e-03], + [ 2.9877e-05, 3.0613e-03, 1.8082e-03, ..., 6.5422e-04, + 6.2370e-04, 3.5000e-03], + [ 1.8654e-06, 2.6970e-03, -4.5586e-03, ..., -7.4806e-03, + 3.3593e-04, -8.8654e-03]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0133, 0.0196, 0.0033, -0.0075, 0.0188, -0.0307, 0.0104, -0.0049, + 0.0011, -0.0127], device='cuda:0'), grad: tensor([-0.0240, 0.0909, -0.0295, -0.0858, -0.0014, 0.0192, 0.0246, 0.0409, + 0.0118, -0.0466], device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 214.51, cls_loss 0.6804 cls_loss_mapping 0.0220 cls_loss_causal 0.5738 re_mapping 0.0189 re_causal 0.0397 /// teacc 98.37 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0440, -0.0597, -0.0668, ..., -0.0223, 0.0409, -0.0640], + [-0.0309, -0.0634, -0.0327, ..., 0.0843, -0.0327, 0.1379], + [ 0.0143, -0.0196, -0.0104, ..., -0.0103, -0.0026, -0.0484], + ..., + [ 0.0085, -0.0645, 0.1022, ..., 0.0115, -0.0442, 0.0323], + [-0.0122, 0.0417, -0.0719, ..., -0.0368, -0.0100, -0.0528], + [-0.0486, 0.0569, 0.0499, ..., -0.0456, -0.0210, -0.0166]], + device='cuda:0'), grad: tensor([[ 1.1921e-07, 5.0850e-03, 5.8556e-04, ..., 5.2376e-03, + 1.3382e-02, 6.5498e-03], + [ 3.4459e-08, -8.8501e-03, 8.2445e-04, ..., -1.0132e-02, + 1.5030e-03, -2.2376e-04], + [ 1.3318e-07, 4.6234e-03, -2.7561e-03, ..., -7.8964e-03, + -2.6035e-03, 1.6379e-04], + ..., + [ 1.2040e-05, 9.7847e-04, 1.5032e-04, ..., -1.3895e-03, + -1.2207e-03, -6.0730e-03], + [ 4.6268e-06, 9.7656e-03, 6.3086e-04, ..., 1.6296e-02, + -6.4354e-03, -8.4686e-03], + [-2.7820e-05, 3.5439e-03, 2.9469e-03, ..., -1.1703e-02, + 1.5516e-03, 9.9792e-03]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0133, 0.0186, 0.0031, -0.0076, 0.0184, -0.0304, 0.0107, -0.0050, + 0.0014, -0.0120], device='cuda:0'), grad: tensor([ 0.0234, -0.0043, -0.0282, -0.0134, 0.0163, 0.0126, 0.0014, 0.0162, + 0.0238, -0.0477], device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 214.62, cls_loss 0.7092 cls_loss_mapping 0.0244 cls_loss_causal 0.6076 re_mapping 0.0195 re_causal 0.0412 /// teacc 98.22 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0511, -0.0622, -0.0680, ..., -0.0228, 0.0395, -0.0647], + [-0.0305, -0.0617, -0.0318, ..., 0.0847, -0.0323, 0.1391], + [ 0.0144, -0.0201, -0.0097, ..., -0.0102, -0.0020, -0.0485], + ..., + [ 0.0094, -0.0648, 0.1034, ..., 0.0116, -0.0444, 0.0333], + [-0.0143, 0.0421, -0.0731, ..., -0.0369, -0.0093, -0.0530], + [-0.0476, 0.0571, 0.0492, ..., -0.0462, -0.0214, -0.0182]], + device='cuda:0'), grad: tensor([[-5.4970e-03, 3.7223e-05, -2.0142e-03, ..., -1.8021e-02, + -2.3289e-03, -1.0080e-03], + [ 5.7077e-04, -1.0233e-03, -2.9588e-04, ..., -2.4853e-03, + 2.6846e-04, -4.8561e-03], + [ 3.3970e-03, 7.9498e-03, 8.1015e-04, ..., 1.2268e-02, + 4.1542e-03, 1.2970e-03], + ..., + [ 6.4468e-04, -5.9814e-03, -5.3635e-03, ..., -6.1264e-03, + 3.6907e-04, -3.2215e-03], + [ 7.7248e-04, -2.0142e-02, 1.1263e-03, ..., -4.3068e-03, + -1.1940e-02, 2.7752e-03], + [ 1.0862e-03, 3.2940e-03, 6.5088e-04, ..., -2.1420e-03, + 3.3212e-04, -1.7376e-03]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0133, 0.0187, 0.0030, -0.0076, 0.0187, -0.0308, 0.0109, -0.0043, + 0.0018, -0.0131], device='cuda:0'), grad: tensor([-0.0670, 0.0031, 0.0473, -0.0142, 0.0094, -0.0138, 0.0230, -0.0069, + 0.0165, 0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 214.18, cls_loss 0.6981 cls_loss_mapping 0.0195 cls_loss_causal 0.5917 re_mapping 0.0202 re_causal 0.0436 /// teacc 98.25 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0526, -0.0616, -0.0686, ..., -0.0222, 0.0395, -0.0647], + [-0.0306, -0.0618, -0.0321, ..., 0.0845, -0.0334, 0.1403], + [ 0.0121, -0.0200, -0.0107, ..., -0.0097, -0.0025, -0.0487], + ..., + [ 0.0091, -0.0657, 0.1042, ..., 0.0111, -0.0450, 0.0330], + [-0.0133, 0.0422, -0.0728, ..., -0.0372, -0.0075, -0.0542], + [-0.0490, 0.0577, 0.0489, ..., -0.0466, -0.0222, -0.0181]], + device='cuda:0'), grad: tensor([[ 1.0625e-05, 1.2264e-03, 5.1355e-04, ..., 4.7607e-03, + 5.6744e-04, 2.7809e-03], + [ 6.8583e-06, 5.1117e-04, 5.9223e-04, ..., 4.8866e-03, + 5.3740e-04, 2.2087e-03], + [ 8.5294e-05, 5.0783e-04, -3.4218e-03, ..., -7.8125e-03, + -6.4430e-03, 5.8842e-04], + ..., + [ 2.9355e-05, 3.6469e-03, 5.6877e-03, ..., 6.0940e-04, + 4.0894e-03, 4.3640e-03], + [-7.6115e-05, 2.7374e-02, 2.0599e-03, ..., 3.2387e-03, + 4.0375e-02, 1.1044e-03], + [ 4.9084e-05, -3.1395e-03, -6.2981e-03, ..., -4.9210e-03, + 2.0943e-03, -5.2643e-03]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0132, 0.0193, 0.0031, -0.0073, 0.0188, -0.0303, 0.0102, -0.0050, + 0.0017, -0.0132], device='cuda:0'), grad: tensor([ 0.0177, 0.0154, -0.0143, -0.0387, 0.0206, -0.0255, 0.0406, -0.0046, + -0.0044, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 214.15, cls_loss 0.6978 cls_loss_mapping 0.0237 cls_loss_causal 0.5955 re_mapping 0.0191 re_causal 0.0410 /// teacc 98.20 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0550, -0.0622, -0.0679, ..., -0.0219, 0.0402, -0.0657], + [-0.0304, -0.0627, -0.0328, ..., 0.0842, -0.0331, 0.1408], + [ 0.0116, -0.0194, -0.0112, ..., -0.0097, -0.0042, -0.0497], + ..., + [ 0.0082, -0.0665, 0.1048, ..., 0.0106, -0.0450, 0.0332], + [-0.0144, 0.0430, -0.0726, ..., -0.0370, -0.0080, -0.0540], + [-0.0495, 0.0567, 0.0487, ..., -0.0466, -0.0213, -0.0187]], + device='cuda:0'), grad: tensor([[ 0.0001, 0.0013, 0.0011, ..., 0.0126, 0.0006, 0.0035], + [ 0.0011, 0.0051, 0.0038, ..., 0.0018, 0.0007, 0.0078], + [ 0.0006, 0.0014, 0.0017, ..., 0.0082, 0.0008, 0.0030], + ..., + [ 0.0032, 0.0129, 0.0196, ..., 0.0275, 0.0008, 0.0280], + [-0.0042, -0.0091, -0.0061, ..., -0.0163, -0.0030, -0.0088], + [-0.0041, -0.0007, -0.0078, ..., -0.0063, 0.0015, -0.0071]], + device='cuda:0') +Epoch 86, bias, value: tensor([-0.0130, 0.0190, 0.0030, -0.0074, 0.0186, -0.0306, 0.0109, -0.0052, + 0.0020, -0.0132], device='cuda:0'), grad: tensor([ 0.0358, 0.0248, 0.0213, -0.0330, -0.0149, -0.0277, -0.0162, 0.0671, + -0.0508, -0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 214.48, cls_loss 0.7108 cls_loss_mapping 0.0221 cls_loss_causal 0.6040 re_mapping 0.0188 re_causal 0.0399 /// teacc 98.25 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0550, -0.0621, -0.0682, ..., -0.0223, 0.0392, -0.0658], + [-0.0308, -0.0622, -0.0343, ..., 0.0844, -0.0336, 0.1410], + [ 0.0119, -0.0193, -0.0109, ..., -0.0094, -0.0040, -0.0499], + ..., + [ 0.0073, -0.0668, 0.1051, ..., 0.0102, -0.0448, 0.0326], + [-0.0142, 0.0418, -0.0729, ..., -0.0375, -0.0072, -0.0539], + [-0.0482, 0.0567, 0.0479, ..., -0.0466, -0.0214, -0.0181]], + device='cuda:0'), grad: tensor([[ 2.2456e-05, -5.8842e-04, 4.1342e-04, ..., 9.2554e-04, + -1.3905e-03, 2.1732e-04], + [-3.3712e-04, -1.0508e-04, 1.7309e-04, ..., -9.1248e-03, + 5.3167e-05, -1.7004e-03], + [-3.9053e-04, 5.5218e-04, 7.8487e-04, ..., 1.2646e-03, + -6.5684e-05, 4.1723e-04], + ..., + [ 1.0997e-04, -4.3030e-03, -1.0574e-02, ..., -6.9389e-03, + 7.9095e-05, -6.1569e-03], + [ 1.0246e-04, -4.5586e-03, -9.6917e-05, ..., -4.5013e-03, + 4.3321e-04, 1.4915e-03], + [ 1.8105e-05, 8.8882e-03, 7.3471e-03, ..., 7.5264e-03, + 1.5755e-03, 5.9891e-03]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0136, 0.0192, 0.0036, -0.0075, 0.0186, -0.0306, 0.0107, -0.0049, + 0.0018, -0.0134], device='cuda:0'), grad: tensor([ 0.0024, -0.0148, 0.0079, 0.0001, -0.0187, 0.0253, -0.0047, -0.0134, + -0.0170, 0.0327], device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 214.96, cls_loss 0.7184 cls_loss_mapping 0.0240 cls_loss_causal 0.6118 re_mapping 0.0182 re_causal 0.0395 /// teacc 98.25 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0558, -0.0616, -0.0691, ..., -0.0236, 0.0395, -0.0664], + [-0.0311, -0.0636, -0.0355, ..., 0.0853, -0.0325, 0.1411], + [ 0.0108, -0.0182, -0.0106, ..., -0.0090, -0.0039, -0.0501], + ..., + [ 0.0076, -0.0669, 0.1052, ..., 0.0088, -0.0453, 0.0338], + [-0.0146, 0.0424, -0.0745, ..., -0.0374, -0.0068, -0.0546], + [-0.0486, 0.0570, 0.0482, ..., -0.0463, -0.0220, -0.0185]], + device='cuda:0'), grad: tensor([[ 7.0870e-05, -8.5220e-03, 1.1784e-04, ..., 2.2542e-04, + -1.1383e-02, 1.6463e-04], + [ 1.1129e-06, -7.7152e-04, 6.5863e-05, ..., -7.3509e-03, + 2.0428e-03, -3.2520e-03], + [ 3.4988e-05, 5.4264e-04, -1.1082e-03, ..., -5.5809e-03, + -1.2608e-03, 1.4365e-04], + ..., + [ 9.1344e-06, -6.3133e-04, -1.9741e-03, ..., 1.3056e-03, + 1.8559e-03, -1.4753e-03], + [ 1.8656e-05, 1.3023e-02, 2.1601e-04, ..., 3.4313e-03, + 1.5612e-03, 9.4509e-04], + [-9.4116e-05, 1.7989e-04, 1.5306e-03, ..., 3.0937e-03, + 3.0155e-03, 1.2379e-03]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0137, 0.0196, 0.0043, -0.0079, 0.0191, -0.0308, 0.0101, -0.0059, + 0.0021, -0.0130], device='cuda:0'), grad: tensor([-0.0209, -0.0217, -0.0169, 0.0271, -0.0033, 0.0277, -0.0364, 0.0102, + 0.0205, 0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 214.88, cls_loss 0.7241 cls_loss_mapping 0.0221 cls_loss_causal 0.6280 re_mapping 0.0190 re_causal 0.0420 /// teacc 98.22 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0567, -0.0619, -0.0694, ..., -0.0237, 0.0400, -0.0656], + [-0.0316, -0.0639, -0.0361, ..., 0.0846, -0.0317, 0.1425], + [ 0.0138, -0.0190, -0.0100, ..., -0.0093, -0.0045, -0.0503], + ..., + [ 0.0062, -0.0663, 0.1059, ..., 0.0099, -0.0468, 0.0345], + [-0.0164, 0.0422, -0.0754, ..., -0.0376, -0.0065, -0.0557], + [-0.0500, 0.0579, 0.0486, ..., -0.0460, -0.0214, -0.0191]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, 1.7481e-03, 4.5371e-04, ..., 8.3771e-03, + 1.1569e-04, 5.7077e-04], + [ 1.8626e-08, 1.9112e-03, 1.1301e-03, ..., 9.7885e-03, + 9.5606e-05, -6.2132e-04], + [ 4.0233e-07, 6.4888e-03, 2.8992e-03, ..., 9.2316e-03, + 6.2828e-03, 1.1253e-03], + ..., + [ 3.0734e-08, 2.7256e-03, 4.5471e-03, ..., -8.0566e-03, + 4.6849e-05, 3.4275e-03], + [ 1.3903e-05, 4.9629e-03, 1.1187e-03, ..., 1.3611e-02, + 4.9019e-04, 1.9665e-03], + [ 2.4252e-06, -3.6278e-03, 6.8808e-04, ..., -3.9368e-03, + 3.9673e-04, 7.4530e-04]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0141, 0.0193, 0.0041, -0.0072, 0.0185, -0.0303, 0.0099, -0.0050, + 0.0012, -0.0124], device='cuda:0'), grad: tensor([ 0.0228, 0.0318, 0.0291, -0.0038, 0.0032, -0.0437, -0.0395, -0.0028, + 0.0376, -0.0347], device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 214.71, cls_loss 0.7223 cls_loss_mapping 0.0180 cls_loss_causal 0.6036 re_mapping 0.0181 re_causal 0.0409 /// teacc 98.20 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0554, -0.0621, -0.0688, ..., -0.0234, 0.0408, -0.0654], + [-0.0321, -0.0646, -0.0359, ..., 0.0847, -0.0315, 0.1424], + [ 0.0134, -0.0184, -0.0105, ..., -0.0099, -0.0031, -0.0495], + ..., + [ 0.0070, -0.0665, 0.1066, ..., 0.0106, -0.0469, 0.0344], + [-0.0163, 0.0418, -0.0764, ..., -0.0376, -0.0072, -0.0555], + [-0.0508, 0.0592, 0.0488, ..., -0.0454, -0.0216, -0.0194]], + device='cuda:0'), grad: tensor([[ 2.6092e-05, 6.7215e-03, 2.5940e-03, ..., 1.0689e-02, + 7.3128e-03, 1.1797e-03], + [ 6.1095e-05, -1.2884e-03, 1.4954e-03, ..., 7.8888e-03, + 9.5272e-04, 2.0920e-02], + [-1.1005e-03, -2.7695e-03, -1.3037e-03, ..., -5.2567e-03, + 4.0016e-03, 4.3988e-04], + ..., + [ 7.0870e-05, 3.9864e-03, 4.0932e-03, ..., 6.5575e-03, + 2.3708e-03, 1.4353e-03], + [ 5.2547e-04, -3.9864e-03, -5.9204e-03, ..., -4.9362e-03, + -3.0518e-03, 1.0948e-03], + [ 1.5545e-04, 6.9351e-03, 2.3067e-04, ..., 5.2109e-03, + 4.3640e-03, -2.4843e-04]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0136, 0.0189, 0.0034, -0.0081, 0.0187, -0.0294, 0.0101, -0.0056, + 0.0012, -0.0118], device='cuda:0'), grad: tensor([ 0.0402, 0.0054, -0.0067, 0.0273, 0.0073, -0.0865, 0.0045, 0.0257, + -0.0325, 0.0154], device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 214.92, cls_loss 0.7175 cls_loss_mapping 0.0242 cls_loss_causal 0.6094 re_mapping 0.0177 re_causal 0.0395 /// teacc 98.30 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0560, -0.0627, -0.0701, ..., -0.0241, 0.0405, -0.0661], + [-0.0326, -0.0636, -0.0368, ..., 0.0846, -0.0320, 0.1433], + [ 0.0125, -0.0179, -0.0110, ..., -0.0107, -0.0029, -0.0494], + ..., + [ 0.0083, -0.0680, 0.1064, ..., 0.0115, -0.0472, 0.0339], + [-0.0158, 0.0420, -0.0769, ..., -0.0381, -0.0069, -0.0562], + [-0.0522, 0.0592, 0.0483, ..., -0.0451, -0.0228, -0.0190]], + device='cuda:0'), grad: tensor([[ 0.0006, -0.0063, -0.0035, ..., -0.0109, -0.0096, -0.0038], + [-0.0033, -0.0006, 0.0004, ..., -0.0264, 0.0014, -0.0114], + [ 0.0002, 0.0020, 0.0009, ..., -0.0062, -0.0210, 0.0012], + ..., + [ 0.0006, 0.0050, 0.0008, ..., 0.0116, 0.0144, 0.0017], + [ 0.0009, 0.0064, 0.0013, ..., 0.0267, 0.0018, 0.0108], + [ 0.0002, 0.0027, 0.0008, ..., 0.0086, 0.0043, 0.0011]], + device='cuda:0') +Epoch 91, bias, value: tensor([-0.0145, 0.0191, 0.0030, -0.0073, 0.0186, -0.0296, 0.0100, -0.0049, + 0.0017, -0.0122], device='cuda:0'), grad: tensor([-0.0341, -0.0588, -0.0352, 0.0253, -0.0040, -0.0306, -0.0205, 0.0293, + 0.0837, 0.0450], device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 214.94, cls_loss 0.6774 cls_loss_mapping 0.0213 cls_loss_causal 0.5704 re_mapping 0.0179 re_causal 0.0384 /// teacc 98.28 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0569, -0.0637, -0.0702, ..., -0.0239, 0.0407, -0.0660], + [-0.0326, -0.0622, -0.0367, ..., 0.0852, -0.0320, 0.1437], + [ 0.0115, -0.0194, -0.0109, ..., -0.0119, -0.0044, -0.0496], + ..., + [ 0.0084, -0.0673, 0.1065, ..., 0.0136, -0.0468, 0.0342], + [-0.0151, 0.0422, -0.0762, ..., -0.0391, -0.0069, -0.0568], + [-0.0515, 0.0592, 0.0480, ..., -0.0456, -0.0227, -0.0195]], + device='cuda:0'), grad: tensor([[ 1.0230e-05, -9.5701e-04, 1.7262e-04, ..., -1.7868e-02, + 2.1076e-04, 1.2003e-05], + [ 2.0146e-05, 5.3444e-03, 9.4473e-05, ..., 2.6154e-02, + 6.5565e-05, 5.5313e-05], + [ 1.8299e-05, -7.8735e-03, 2.4199e-04, ..., -9.0103e-03, + 4.0627e-04, 2.2024e-05], + ..., + [-6.3121e-05, -2.3613e-03, 2.4128e-03, ..., -1.5030e-02, + 1.4067e-04, 1.2130e-04], + [ 1.8060e-05, 2.2507e-03, 7.5579e-04, ..., 1.0124e-02, + 9.0456e-04, 1.5211e-04], + [ 3.3170e-05, 1.4791e-03, -2.6035e-03, ..., 8.9569e-03, + 5.1880e-04, -3.9816e-04]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0146, 0.0191, 0.0023, -0.0071, 0.0187, -0.0295, 0.0103, -0.0040, + 0.0012, -0.0125], device='cuda:0'), grad: tensor([-0.0435, 0.0421, -0.0288, 0.0143, 0.0229, 0.0064, -0.0333, 0.0187, + 0.0050, -0.0038], device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 214.79, cls_loss 0.7244 cls_loss_mapping 0.0244 cls_loss_causal 0.6202 re_mapping 0.0176 re_causal 0.0388 /// teacc 98.17 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0547, -0.0639, -0.0705, ..., -0.0234, 0.0419, -0.0663], + [-0.0338, -0.0618, -0.0381, ..., 0.0850, -0.0317, 0.1438], + [ 0.0120, -0.0187, -0.0099, ..., -0.0117, -0.0053, -0.0493], + ..., + [ 0.0103, -0.0685, 0.1068, ..., 0.0149, -0.0471, 0.0358], + [-0.0158, 0.0421, -0.0766, ..., -0.0404, -0.0053, -0.0574], + [-0.0527, 0.0604, 0.0487, ..., -0.0451, -0.0242, -0.0198]], + device='cuda:0'), grad: tensor([[ 2.8089e-05, -1.2398e-02, 6.5327e-04, ..., -4.9667e-03, + -1.4366e-02, 1.3125e-04], + [ 6.0499e-05, 1.1597e-03, 4.6325e-04, ..., 3.1662e-03, + 1.1139e-03, 2.4307e-04], + [ 1.3857e-03, 7.5989e-03, 1.5917e-03, ..., 5.0049e-03, + -4.7035e-03, 6.7377e-04], + ..., + [ 1.3208e-04, -7.0877e-03, 2.7905e-03, ..., -9.9869e-03, + 1.3008e-03, 1.9588e-03], + [ 3.7336e-04, 1.6724e-02, 1.6918e-03, ..., -3.1986e-03, + 7.2594e-03, 2.7523e-03], + [-1.2417e-03, -3.5248e-03, -8.1635e-03, ..., -1.0422e-02, + -4.0245e-03, -9.2773e-03]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0144, 0.0189, 0.0025, -0.0072, 0.0177, -0.0297, 0.0104, -0.0035, + 0.0006, -0.0115], device='cuda:0'), grad: tensor([-0.0583, 0.0129, 0.0244, 0.0215, -0.0200, 0.0444, 0.0383, -0.0414, + 0.0150, -0.0367], device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 214.92, cls_loss 0.7294 cls_loss_mapping 0.0224 cls_loss_causal 0.6338 re_mapping 0.0182 re_causal 0.0408 /// teacc 98.12 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0527, -0.0648, -0.0717, ..., -0.0242, 0.0426, -0.0673], + [-0.0328, -0.0630, -0.0381, ..., 0.0851, -0.0317, 0.1440], + [ 0.0122, -0.0192, -0.0103, ..., -0.0124, -0.0048, -0.0503], + ..., + [ 0.0095, -0.0669, 0.1077, ..., 0.0149, -0.0465, 0.0359], + [-0.0156, 0.0422, -0.0769, ..., -0.0403, -0.0055, -0.0576], + [-0.0527, 0.0600, 0.0482, ..., -0.0450, -0.0243, -0.0185]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0003, 0.0006, ..., 0.0097, 0.0013, 0.0005], + [-0.0018, -0.0001, -0.0033, ..., -0.0219, 0.0004, -0.0046], + [ 0.0018, 0.0003, 0.0015, ..., 0.0139, 0.0051, 0.0017], + ..., + [ 0.0004, -0.0162, -0.0017, ..., 0.0073, 0.0006, -0.0024], + [ 0.0007, 0.0055, 0.0005, ..., -0.0043, 0.0009, 0.0011], + [-0.0037, 0.0061, 0.0039, ..., -0.0062, -0.0040, 0.0009]], + device='cuda:0') +Epoch 94, bias, value: tensor([-0.0151, 0.0191, 0.0016, -0.0068, 0.0180, -0.0295, 0.0105, -0.0029, + 0.0005, -0.0117], device='cuda:0'), grad: tensor([ 0.0388, -0.0574, 0.0411, -0.0054, -0.0684, 0.0256, 0.0176, 0.0193, + -0.0167, 0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 215.00, cls_loss 0.7074 cls_loss_mapping 0.0173 cls_loss_causal 0.6167 re_mapping 0.0179 re_causal 0.0423 /// teacc 98.03 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0538, -0.0651, -0.0718, ..., -0.0246, 0.0420, -0.0671], + [-0.0331, -0.0631, -0.0398, ..., 0.0847, -0.0315, 0.1449], + [ 0.0129, -0.0197, -0.0099, ..., -0.0118, -0.0049, -0.0499], + ..., + [ 0.0091, -0.0665, 0.1091, ..., 0.0149, -0.0467, 0.0354], + [-0.0145, 0.0417, -0.0776, ..., -0.0401, -0.0062, -0.0570], + [-0.0540, 0.0601, 0.0476, ..., -0.0455, -0.0249, -0.0194]], + device='cuda:0'), grad: tensor([[ 5.2261e-04, 6.7759e-04, 4.9782e-04, ..., 2.5272e-03, + 6.4802e-04, 2.5463e-03], + [ 5.9456e-05, -2.4261e-03, 1.9729e-04, ..., -9.6817e-03, + -2.5845e-03, -2.1763e-03], + [ 2.8324e-04, 3.5667e-03, -5.9319e-04, ..., 2.1000e-03, + 6.0081e-04, 7.9298e-04], + ..., + [ 4.9639e-04, 4.0092e-03, 2.6054e-03, ..., 3.3150e-03, + 1.0080e-03, 4.1389e-03], + [ 1.9550e-04, 3.0689e-03, 1.8806e-03, ..., 2.9316e-03, + 1.1244e-03, -3.6335e-03], + [-1.5898e-03, 1.4677e-03, -4.3144e-03, ..., -9.1324e-03, + 7.9823e-04, -3.5667e-03]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0150, 0.0179, 0.0027, -0.0061, 0.0182, -0.0295, 0.0102, -0.0034, + 0.0012, -0.0124], device='cuda:0'), grad: tensor([ 0.0296, -0.0254, 0.0225, -0.0404, 0.0130, -0.0270, 0.0191, 0.0267, + 0.0191, -0.0374], device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 214.48, cls_loss 0.7119 cls_loss_mapping 0.0227 cls_loss_causal 0.6049 re_mapping 0.0177 re_causal 0.0384 /// teacc 98.26 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0530, -0.0648, -0.0707, ..., -0.0233, 0.0420, -0.0664], + [-0.0324, -0.0639, -0.0409, ..., 0.0849, -0.0309, 0.1465], + [ 0.0136, -0.0200, -0.0110, ..., -0.0114, -0.0050, -0.0496], + ..., + [ 0.0078, -0.0673, 0.1093, ..., 0.0156, -0.0472, 0.0345], + [-0.0143, 0.0412, -0.0778, ..., -0.0402, -0.0061, -0.0584], + [-0.0502, 0.0614, 0.0485, ..., -0.0461, -0.0255, -0.0191]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0006, -0.0018, ..., -0.0100, 0.0002, -0.0050], + [ 0.0002, -0.0162, -0.0141, ..., -0.0132, 0.0004, -0.0091], + [ 0.0018, -0.0116, -0.0058, ..., -0.0200, -0.0041, -0.0028], + ..., + [ 0.0009, 0.0020, 0.0018, ..., 0.0068, 0.0003, -0.0003], + [-0.0015, 0.0074, 0.0066, ..., 0.0070, 0.0007, 0.0029], + [ 0.0002, 0.0056, 0.0061, ..., 0.0094, 0.0010, 0.0051]], + device='cuda:0') +Epoch 96, bias, value: tensor([-0.0142, 0.0189, 0.0024, -0.0065, 0.0180, -0.0300, 0.0100, -0.0030, + 0.0006, -0.0125], device='cuda:0'), grad: tensor([-0.0113, -0.0378, -0.0428, -0.0018, 0.0170, 0.0272, -0.0091, 0.0219, + 0.0006, 0.0360], device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 214.47, cls_loss 0.6691 cls_loss_mapping 0.0170 cls_loss_causal 0.5616 re_mapping 0.0173 re_causal 0.0393 /// teacc 98.24 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0538, -0.0651, -0.0709, ..., -0.0237, 0.0428, -0.0675], + [-0.0327, -0.0640, -0.0405, ..., 0.0853, -0.0309, 0.1484], + [ 0.0131, -0.0200, -0.0114, ..., -0.0107, -0.0051, -0.0505], + ..., + [ 0.0080, -0.0668, 0.1095, ..., 0.0146, -0.0469, 0.0348], + [-0.0149, 0.0419, -0.0785, ..., -0.0398, -0.0063, -0.0589], + [-0.0495, 0.0615, 0.0491, ..., -0.0470, -0.0265, -0.0187]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8930e-04, -6.5374e-04, ..., -3.1872e-03, + 4.7565e-04, 3.9673e-04], + [ 4.3586e-07, -1.8328e-05, -4.8399e-04, ..., -3.9749e-03, + 7.3433e-05, -1.7490e-03], + [ 6.3777e-06, -1.7004e-03, 4.4346e-04, ..., -1.4236e-02, + -2.5606e-04, -5.5027e-04], + ..., + [-5.7407e-06, 4.6134e-05, 1.4424e-04, ..., 2.9621e-03, + 1.6391e-04, -5.9319e-03], + [ 7.2364e-07, -1.5306e-03, 2.8872e-04, ..., 5.0316e-03, + -7.9422e-03, 1.5602e-03], + [-2.5760e-06, 9.1076e-04, -6.7592e-05, ..., 2.1229e-03, + 4.1428e-03, 3.4599e-03]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0144, 0.0196, 0.0032, -0.0066, 0.0184, -0.0301, 0.0094, -0.0037, + 0.0009, -0.0131], device='cuda:0'), grad: tensor([-0.0143, -0.0032, -0.0343, -0.0144, 0.0230, 0.0084, 0.0076, 0.0008, + 0.0042, 0.0223], device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 214.75, cls_loss 0.6912 cls_loss_mapping 0.0195 cls_loss_causal 0.6044 re_mapping 0.0169 re_causal 0.0373 /// teacc 98.19 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0541, -0.0651, -0.0704, ..., -0.0239, 0.0419, -0.0684], + [-0.0338, -0.0640, -0.0412, ..., 0.0857, -0.0317, 0.1498], + [ 0.0123, -0.0206, -0.0117, ..., -0.0106, -0.0053, -0.0514], + ..., + [ 0.0098, -0.0660, 0.1092, ..., 0.0147, -0.0472, 0.0363], + [-0.0153, 0.0420, -0.0771, ..., -0.0398, -0.0058, -0.0604], + [-0.0511, 0.0612, 0.0504, ..., -0.0469, -0.0253, -0.0195]], + device='cuda:0'), grad: tensor([[ 1.1641e-04, 8.4152e-03, -5.7220e-04, ..., -3.5286e-03, + 2.2659e-03, -2.3708e-03], + [ 1.2481e-04, -9.7573e-05, 7.0906e-04, ..., -7.3509e-03, + 1.4770e-04, -3.7823e-03], + [ 7.2956e-04, 1.7563e-02, 1.1482e-03, ..., 3.1605e-03, + 4.4289e-03, 1.9665e-03], + ..., + [ 2.0516e-04, 6.8045e-04, -5.5542e-03, ..., -4.4899e-03, + 1.3790e-03, -2.8019e-03], + [ 2.5606e-04, -1.0633e-03, -2.5501e-03, ..., -2.5959e-03, + -4.8332e-03, -4.4556e-03], + [-8.0566e-03, 3.3169e-03, 6.5384e-03, ..., 1.7583e-04, + 2.7256e-03, 4.8141e-03]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0154, 0.0200, 0.0033, -0.0075, 0.0183, -0.0304, 0.0101, -0.0031, + 0.0005, -0.0123], device='cuda:0'), grad: tensor([-0.0095, -0.0190, 0.0275, -0.0154, 0.0339, -0.0072, 0.0234, -0.0282, + -0.0113, 0.0059], device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 214.71, cls_loss 0.6714 cls_loss_mapping 0.0201 cls_loss_causal 0.5691 re_mapping 0.0166 re_causal 0.0360 /// teacc 98.36 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0548, -0.0643, -0.0715, ..., -0.0237, 0.0422, -0.0684], + [-0.0327, -0.0649, -0.0428, ..., 0.0857, -0.0318, 0.1500], + [ 0.0117, -0.0198, -0.0114, ..., -0.0103, -0.0052, -0.0520], + ..., + [ 0.0091, -0.0661, 0.1106, ..., 0.0149, -0.0471, 0.0368], + [-0.0162, 0.0425, -0.0769, ..., -0.0392, -0.0051, -0.0595], + [-0.0502, 0.0610, 0.0494, ..., -0.0474, -0.0262, -0.0202]], + device='cuda:0'), grad: tensor([[ 2.8029e-05, 2.4796e-05, 1.9443e-04, ..., 2.7695e-03, + -1.5678e-03, 6.8855e-04], + [ 6.3241e-05, 2.7609e-04, 1.0090e-03, ..., -4.6349e-03, + 1.1511e-03, 1.2026e-03], + [ 9.4223e-04, 6.5327e-05, 6.0558e-04, ..., 1.3794e-02, + 7.3862e-04, 2.3041e-03], + ..., + [ 1.1790e-04, 4.2558e-04, 1.1406e-03, ..., 3.9215e-03, + 1.0061e-03, 2.5921e-03], + [ 3.5119e-04, 3.1185e-03, 3.5553e-03, ..., 3.3932e-03, + 3.4599e-03, 1.9665e-03], + [-9.8991e-04, -4.3640e-03, -1.2657e-02, ..., -1.6174e-02, + -1.2123e-02, -1.9806e-02]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0155, 0.0194, 0.0046, -0.0078, 0.0180, -0.0299, 0.0093, -0.0030, + 0.0016, -0.0132], device='cuda:0'), grad: tensor([ 0.0065, -0.0177, 0.0357, -0.0200, -0.0095, 0.0267, 0.0072, 0.0144, + 0.0156, -0.0589], device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 214.72, cls_loss 0.6958 cls_loss_mapping 0.0179 cls_loss_causal 0.5973 re_mapping 0.0169 re_causal 0.0388 /// teacc 98.23 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0546, -0.0646, -0.0722, ..., -0.0236, 0.0422, -0.0688], + [-0.0326, -0.0653, -0.0434, ..., 0.0850, -0.0317, 0.1495], + [ 0.0108, -0.0196, -0.0103, ..., -0.0102, -0.0048, -0.0516], + ..., + [ 0.0089, -0.0669, 0.1100, ..., 0.0143, -0.0469, 0.0367], + [-0.0164, 0.0424, -0.0779, ..., -0.0389, -0.0054, -0.0592], + [-0.0499, 0.0610, 0.0513, ..., -0.0458, -0.0266, -0.0189]], + device='cuda:0'), grad: tensor([[ 1.4231e-05, 3.4690e-04, 1.9264e-03, ..., 5.2261e-03, + 4.9591e-04, 3.2139e-03], + [ 1.8520e-03, 3.1319e-03, 5.7888e-04, ..., 2.3556e-03, + 9.4748e-04, 3.3855e-03], + [ 7.6485e-04, -2.4490e-03, 1.5936e-03, ..., 9.5901e-03, + 3.7217e-04, 5.7449e-03], + ..., + [ 5.5981e-04, 1.1330e-03, -6.9351e-03, ..., 8.5211e-04, + 4.0150e-04, -9.8267e-03], + [ 1.4770e-04, 4.0507e-04, 1.0624e-03, ..., -4.7646e-03, + -6.6757e-03, -8.5068e-03], + [-2.6531e-03, -3.6602e-03, -4.9248e-03, ..., -1.1017e-02, + 7.9632e-04, -2.2621e-03]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0154, 0.0192, 0.0048, -0.0082, 0.0183, -0.0305, 0.0097, -0.0039, + 0.0009, -0.0113], device='cuda:0'), grad: tensor([ 2.3102e-02, -1.5732e-02, 5.9634e-05, 1.8494e-02, 1.4145e-02, + 8.7433e-03, -6.3744e-03, -7.1220e-03, -1.8082e-02, -1.7242e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 214.38, cls_loss 0.6968 cls_loss_mapping 0.0184 cls_loss_causal 0.5988 re_mapping 0.0170 re_causal 0.0377 /// teacc 98.13 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0546, -0.0650, -0.0727, ..., -0.0236, 0.0419, -0.0691], + [-0.0333, -0.0652, -0.0437, ..., 0.0847, -0.0315, 0.1495], + [ 0.0110, -0.0195, -0.0100, ..., -0.0093, -0.0047, -0.0527], + ..., + [ 0.0089, -0.0670, 0.1104, ..., 0.0141, -0.0475, 0.0362], + [-0.0173, 0.0429, -0.0799, ..., -0.0380, -0.0054, -0.0576], + [-0.0510, 0.0605, 0.0516, ..., -0.0462, -0.0271, -0.0182]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0011, 0.0002, ..., 0.0023, 0.0011, 0.0001], + [ 0.0005, 0.0004, 0.0003, ..., -0.0070, 0.0007, -0.0001], + [ 0.0010, 0.0102, 0.0003, ..., -0.0025, 0.0023, 0.0001], + ..., + [ 0.0015, 0.0006, 0.0011, ..., 0.0023, 0.0012, 0.0003], + [ 0.0018, -0.0135, -0.0047, ..., -0.0041, -0.0157, -0.0006], + [ 0.0024, 0.0019, 0.0013, ..., 0.0049, 0.0030, 0.0002]], + device='cuda:0') +Epoch 101, bias, value: tensor([-0.0162, 0.0195, 0.0054, -0.0078, 0.0179, -0.0309, 0.0099, -0.0041, + 0.0018, -0.0120], device='cuda:0'), grad: tensor([ 0.0178, -0.0387, 0.0031, 0.0343, 0.0168, -0.0199, -0.0162, 0.0194, + -0.0452, 0.0287], device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.57, cls_loss 0.6910 cls_loss_mapping 0.0218 cls_loss_causal 0.5988 re_mapping 0.0171 re_causal 0.0372 /// teacc 98.23 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0546, -0.0657, -0.0725, ..., -0.0230, 0.0412, -0.0691], + [-0.0340, -0.0639, -0.0421, ..., 0.0857, -0.0307, 0.1502], + [ 0.0095, -0.0205, -0.0110, ..., -0.0089, -0.0051, -0.0525], + ..., + [ 0.0087, -0.0672, 0.1097, ..., 0.0138, -0.0483, 0.0367], + [-0.0171, 0.0434, -0.0813, ..., -0.0381, -0.0048, -0.0580], + [-0.0510, 0.0614, 0.0523, ..., -0.0458, -0.0270, -0.0181]], + device='cuda:0'), grad: tensor([[ 1.0775e-06, 6.0415e-04, 2.4915e-04, ..., 8.8882e-04, + -1.0233e-03, 7.9727e-04], + [ 1.2733e-05, -2.6751e-04, 6.9475e-04, ..., -3.4103e-03, + 7.1716e-04, -1.5335e-03], + [ 3.3474e-04, 2.8877e-03, 6.3210e-03, ..., 5.5695e-03, + 1.6165e-03, 4.4479e-03], + ..., + [-4.3082e-04, -1.8585e-02, -4.9248e-03, ..., -1.9608e-03, + 3.8910e-04, -2.1606e-02], + [ 1.4476e-05, 1.8129e-03, 2.3174e-04, ..., 2.1133e-03, + 9.7656e-04, 1.3037e-03], + [ 8.7172e-06, 1.0887e-02, -4.6730e-03, ..., -5.8746e-03, + 2.6822e-04, 1.3916e-02]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0159, 0.0196, 0.0056, -0.0078, 0.0178, -0.0302, 0.0093, -0.0040, + 0.0014, -0.0122], device='cuda:0'), grad: tensor([ 0.0073, -0.0026, 0.0153, -0.0031, 0.0271, -0.0270, 0.0089, -0.0222, + -0.0083, 0.0047], device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 214.31, cls_loss 0.6757 cls_loss_mapping 0.0189 cls_loss_causal 0.5726 re_mapping 0.0182 re_causal 0.0390 /// teacc 98.03 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0543, -0.0655, -0.0731, ..., -0.0226, 0.0412, -0.0697], + [-0.0358, -0.0630, -0.0408, ..., 0.0855, -0.0291, 0.1514], + [ 0.0089, -0.0217, -0.0108, ..., -0.0091, -0.0063, -0.0521], + ..., + [ 0.0108, -0.0667, 0.1096, ..., 0.0137, -0.0479, 0.0365], + [-0.0177, 0.0427, -0.0824, ..., -0.0386, -0.0043, -0.0582], + [-0.0520, 0.0612, 0.0532, ..., -0.0458, -0.0265, -0.0179]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, 2.1839e-04, -3.9458e-04, ..., 2.8968e-04, + 1.5497e-04, 5.4884e-04], + [ 1.0291e-07, 1.7241e-05, 8.6355e-04, ..., 2.4033e-04, + 1.6546e-04, 3.6564e-03], + [ 5.4017e-07, 2.1648e-03, 9.5081e-04, ..., 2.0771e-03, + 1.5771e-04, 3.5648e-03], + ..., + [ 3.8706e-06, -1.9550e-03, -1.4102e-04, ..., -4.1656e-03, + -1.3227e-03, -6.0272e-03], + [ 3.7774e-06, 1.6756e-03, -2.8276e-04, ..., 3.1071e-03, + 1.2314e-04, 1.1787e-03], + [-6.3591e-06, 3.4466e-03, 9.6178e-04, ..., 7.7019e-03, + 9.2089e-05, 1.1276e-02]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0156, 0.0198, 0.0048, -0.0076, 0.0179, -0.0305, 0.0100, -0.0036, + 0.0009, -0.0124], device='cuda:0'), grad: tensor([-0.0006, 0.0117, -0.0004, 0.0293, -0.0522, -0.0675, 0.0420, -0.0060, + 0.0097, 0.0340], device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 214.84, cls_loss 0.6613 cls_loss_mapping 0.0201 cls_loss_causal 0.5661 re_mapping 0.0167 re_causal 0.0376 /// teacc 98.19 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0546, -0.0658, -0.0733, ..., -0.0225, 0.0409, -0.0699], + [-0.0359, -0.0631, -0.0395, ..., 0.0855, -0.0290, 0.1524], + [ 0.0084, -0.0220, -0.0117, ..., -0.0081, -0.0069, -0.0531], + ..., + [ 0.0108, -0.0661, 0.1103, ..., 0.0131, -0.0486, 0.0360], + [-0.0174, 0.0432, -0.0826, ..., -0.0390, -0.0040, -0.0592], + [-0.0521, 0.0606, 0.0530, ..., -0.0460, -0.0265, -0.0173]], + device='cuda:0'), grad: tensor([[ 1.1243e-05, -3.1257e-04, 3.8290e-04, ..., 7.6628e-04, + -2.5597e-03, -1.3769e-04], + [ 1.2346e-05, 9.9087e-04, 1.1759e-03, ..., -8.5449e-03, + -4.0531e-04, -1.9236e-03], + [ 3.3498e-05, -8.7204e-03, -3.7193e-03, ..., 1.2306e-02, + 1.1635e-03, 2.5024e-03], + ..., + [ 1.1718e-04, -8.8348e-03, -1.1467e-02, ..., -1.2047e-02, + -3.5648e-03, 2.3899e-03], + [ 3.7104e-05, 5.6725e-03, 2.3632e-03, ..., 2.3746e-03, + 1.0815e-03, 2.2507e-03], + [-4.2629e-04, -2.8931e-02, -8.8043e-03, ..., -2.1454e-02, + -4.2419e-03, -6.0158e-03]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0161, 0.0198, 0.0054, -0.0076, 0.0175, -0.0299, 0.0095, -0.0038, + 0.0009, -0.0123], device='cuda:0'), grad: tensor([ 0.0018, -0.0016, 0.0105, 0.0735, -0.0173, 0.0293, 0.0138, -0.0338, + 0.0020, -0.0782], device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 214.94, cls_loss 0.6783 cls_loss_mapping 0.0171 cls_loss_causal 0.5742 re_mapping 0.0171 re_causal 0.0375 /// teacc 98.47 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0552, -0.0668, -0.0730, ..., -0.0228, 0.0408, -0.0692], + [-0.0343, -0.0629, -0.0396, ..., 0.0853, -0.0295, 0.1538], + [ 0.0080, -0.0216, -0.0117, ..., -0.0083, -0.0064, -0.0535], + ..., + [ 0.0105, -0.0666, 0.1102, ..., 0.0131, -0.0488, 0.0346], + [-0.0177, 0.0443, -0.0831, ..., -0.0386, -0.0051, -0.0605], + [-0.0520, 0.0610, 0.0533, ..., -0.0455, -0.0268, -0.0162]], + device='cuda:0'), grad: tensor([[ 3.4925e-08, 1.0996e-03, 5.2929e-04, ..., 3.8319e-03, + 2.6345e-04, 7.7903e-05], + [ 7.0781e-08, 5.5170e-04, 1.8644e-04, ..., 1.9331e-03, + 1.5664e-04, -7.3016e-05], + [-2.5537e-06, 3.3779e-03, 1.0872e-03, ..., -1.7328e-03, + -1.7834e-03, 2.5678e-04], + ..., + [ 2.0396e-07, 1.8044e-03, -5.6343e-03, ..., -2.2011e-03, + -1.7996e-03, 3.9577e-05], + [ 1.2415e-06, 1.5503e-02, 1.9312e-03, ..., 1.5640e-03, + 6.8617e-04, -7.2098e-04], + [ 3.4878e-07, 3.0251e-03, 1.5755e-03, ..., 6.8474e-03, + 7.1621e-04, 1.6346e-03]], device='cuda:0') +Epoch 105, bias, value: tensor([-1.5807e-02, 1.9879e-02, 5.4242e-03, -7.1421e-03, 1.7674e-02, + -2.9805e-02, 9.5737e-03, -4.2043e-03, 1.9889e-05, -1.1983e-02], + device='cuda:0'), grad: tensor([ 0.0144, 0.0075, 0.0028, -0.0922, 0.0170, 0.0244, 0.0184, -0.0105, + -0.0089, 0.0270], device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 214.69, cls_loss 0.6877 cls_loss_mapping 0.0148 cls_loss_causal 0.5926 re_mapping 0.0165 re_causal 0.0375 /// teacc 98.28 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0556, -0.0677, -0.0727, ..., -0.0236, 0.0411, -0.0693], + [-0.0349, -0.0611, -0.0389, ..., 0.0861, -0.0300, 0.1546], + [ 0.0073, -0.0205, -0.0115, ..., -0.0081, -0.0059, -0.0538], + ..., + [ 0.0109, -0.0672, 0.1114, ..., 0.0138, -0.0499, 0.0347], + [-0.0174, 0.0444, -0.0838, ..., -0.0388, -0.0046, -0.0611], + [-0.0515, 0.0606, 0.0513, ..., -0.0463, -0.0256, -0.0163]], + device='cuda:0'), grad: tensor([[ 6.3330e-08, 9.3460e-04, 4.0078e-04, ..., 6.0730e-03, + 4.0894e-03, 6.7253e-03], + [ 2.1327e-07, 1.3885e-03, 5.8794e-04, ..., 6.5002e-03, + 1.4057e-03, 4.9706e-03], + [ 4.2748e-07, 7.1478e-04, 1.3027e-03, ..., 2.1305e-03, + 3.3989e-03, 7.1030e-03], + ..., + [ 1.4994e-07, 3.6931e-04, 7.5960e-04, ..., -3.3321e-03, + 7.0667e-04, 1.4791e-03], + [-3.9153e-06, 1.6966e-03, -3.1853e-04, ..., -5.3787e-03, + -1.9852e-02, -1.3107e-02], + [ 1.6838e-06, 4.0197e-04, -1.7385e-03, ..., 4.6730e-03, + 2.5196e-03, -1.5318e-04]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0170, 0.0205, 0.0058, -0.0073, 0.0183, -0.0299, 0.0097, -0.0035, + -0.0001, -0.0129], device='cuda:0'), grad: tensor([ 0.0049, 0.0275, -0.0005, -0.0051, 0.0411, -0.0818, 0.0229, -0.0031, + -0.0079, 0.0021], device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 214.79, cls_loss 0.6213 cls_loss_mapping 0.0150 cls_loss_causal 0.5284 re_mapping 0.0170 re_causal 0.0369 /// teacc 98.43 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0553, -0.0685, -0.0725, ..., -0.0237, 0.0402, -0.0700], + [-0.0350, -0.0617, -0.0390, ..., 0.0866, -0.0294, 0.1550], + [ 0.0072, -0.0199, -0.0129, ..., -0.0080, -0.0052, -0.0550], + ..., + [ 0.0110, -0.0684, 0.1116, ..., 0.0143, -0.0500, 0.0353], + [-0.0196, 0.0438, -0.0849, ..., -0.0388, -0.0044, -0.0601], + [-0.0502, 0.0616, 0.0515, ..., -0.0467, -0.0262, -0.0165]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.1778e-05, 1.1183e-05, ..., -1.5732e-02, + 1.3180e-03, -4.9438e-03], + [ 0.0000e+00, -6.7520e-04, 3.5667e-04, ..., 8.1301e-04, + 3.8028e-04, -7.5417e-03], + [ 0.0000e+00, 1.4181e-03, 4.9859e-05, ..., -4.2801e-03, + -2.0580e-03, 3.0651e-03], + ..., + [ 0.0000e+00, 9.6226e-04, 8.4043e-05, ..., 6.7863e-03, + 1.0271e-03, 6.4812e-03], + [ 0.0000e+00, 3.3989e-03, 1.0288e-04, ..., 1.5402e-03, + -8.6746e-03, 4.6730e-03], + [ 0.0000e+00, -9.3689e-03, 2.8992e-04, ..., 9.0561e-03, + 1.4448e-03, 8.6136e-03]], device='cuda:0') +Epoch 107, bias, value: tensor([-1.7105e-02, 2.0296e-02, 5.6789e-03, -7.4903e-03, 1.8478e-02, + -2.9656e-02, 9.3917e-03, -3.1784e-03, -2.1515e-05, -1.2826e-02], + device='cuda:0'), grad: tensor([-0.0399, -0.0222, -0.0042, -0.0082, 0.0499, -0.0350, -0.0115, 0.0347, + 0.0187, 0.0179], device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 214.95, cls_loss 0.7025 cls_loss_mapping 0.0184 cls_loss_causal 0.6107 re_mapping 0.0167 re_causal 0.0373 /// teacc 98.42 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0556, -0.0695, -0.0731, ..., -0.0238, 0.0406, -0.0702], + [-0.0350, -0.0625, -0.0384, ..., 0.0862, -0.0290, 0.1548], + [ 0.0075, -0.0209, -0.0138, ..., -0.0084, -0.0063, -0.0554], + ..., + [ 0.0107, -0.0692, 0.1120, ..., 0.0148, -0.0530, 0.0354], + [-0.0197, 0.0439, -0.0854, ..., -0.0390, -0.0035, -0.0613], + [-0.0498, 0.0623, 0.0512, ..., -0.0466, -0.0262, -0.0163]], + device='cuda:0'), grad: tensor([[ 2.0433e-06, 4.8971e-04, 6.8784e-05, ..., 2.3232e-03, + 8.5020e-04, 2.2836e-06], + [ 1.6224e-06, 5.6654e-05, 5.1355e-04, ..., 3.9597e-03, + 1.6842e-03, 3.0566e-06], + [ 6.2864e-07, 1.9714e-02, -2.0361e-04, ..., 1.7410e-02, + 3.4943e-02, 6.2525e-05], + ..., + [ 5.4622e-07, 1.9157e-04, -2.1636e-04, ..., -3.8052e-03, + -1.8215e-03, -1.4758e-04], + [ 3.7439e-06, 6.4011e-03, 2.6965e-04, ..., -1.4572e-03, + 1.9255e-03, 7.7635e-06], + [ 2.7586e-06, -4.8488e-05, 9.2387e-05, ..., -8.1711e-03, + -3.1128e-03, 3.5111e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0170, 0.0203, 0.0051, -0.0076, 0.0189, -0.0296, 0.0096, -0.0037, + 0.0001, -0.0126], device='cuda:0'), grad: tensor([ 0.0127, 0.0148, 0.0244, -0.0311, 0.0247, -0.0108, 0.0223, -0.0191, + -0.0066, -0.0314], device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 214.89, cls_loss 0.6554 cls_loss_mapping 0.0183 cls_loss_causal 0.5746 re_mapping 0.0161 re_causal 0.0348 /// teacc 98.41 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0558, -0.0690, -0.0746, ..., -0.0239, 0.0427, -0.0689], + [-0.0357, -0.0616, -0.0395, ..., 0.0859, -0.0303, 0.1550], + [ 0.0081, -0.0221, -0.0146, ..., -0.0088, -0.0064, -0.0557], + ..., + [ 0.0097, -0.0695, 0.1133, ..., 0.0140, -0.0530, 0.0351], + [-0.0197, 0.0443, -0.0853, ..., -0.0396, -0.0015, -0.0620], + [-0.0503, 0.0617, 0.0506, ..., -0.0459, -0.0281, -0.0174]], + device='cuda:0'), grad: tensor([[ 7.1432e-07, -4.6883e-03, -1.1911e-03, ..., -1.0521e-02, + 7.8506e-03, 1.1673e-02], + [ 9.6112e-07, 1.8084e-04, 3.4404e-04, ..., -1.5488e-02, + -1.4099e-02, -1.3100e-02], + [-8.4543e-04, 5.7030e-04, -1.6718e-03, ..., 1.1887e-02, + 7.7171e-03, 1.6108e-03], + ..., + [ 3.4004e-05, 3.8862e-04, -1.0484e-04, ..., -4.6959e-03, + -2.7943e-04, 1.4057e-03], + [ 5.7173e-04, 1.3094e-03, 6.8617e-04, ..., -3.3264e-03, + -3.6469e-03, -4.2305e-03], + [ 1.2323e-05, 6.5088e-04, 6.1893e-04, ..., 5.1804e-03, + 2.4853e-03, 1.5869e-03]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0173, 0.0194, 0.0044, -0.0068, 0.0191, -0.0297, 0.0094, -0.0034, + 0.0003, -0.0118], device='cuda:0'), grad: tensor([ 0.0087, -0.0741, 0.0279, 0.0338, 0.0131, -0.0327, 0.0285, -0.0044, + -0.0254, 0.0246], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 108---------------------------------------------------- +epoch 108, time 230.94, cls_loss 0.6631 cls_loss_mapping 0.0174 cls_loss_causal 0.5759 re_mapping 0.0167 re_causal 0.0371 /// teacc 98.48 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0558, -0.0680, -0.0749, ..., -0.0235, 0.0424, -0.0689], + [-0.0362, -0.0620, -0.0401, ..., 0.0860, -0.0286, 0.1555], + [ 0.0080, -0.0228, -0.0151, ..., -0.0089, -0.0092, -0.0562], + ..., + [ 0.0092, -0.0691, 0.1138, ..., 0.0146, -0.0517, 0.0362], + [-0.0205, 0.0445, -0.0857, ..., -0.0400, -0.0006, -0.0616], + [-0.0497, 0.0613, 0.0524, ..., -0.0457, -0.0283, -0.0181]], + device='cuda:0'), grad: tensor([[ 6.9523e-04, 7.1526e-05, 7.1096e-04, ..., 3.1986e-03, + 5.2691e-04, 1.3437e-03], + [ 3.7837e-04, 2.0885e-03, 1.6713e-04, ..., -9.7656e-04, + -5.2595e-04, -2.0237e-03], + [ 3.5620e-04, 5.7399e-05, -3.3617e-04, ..., 3.4180e-03, + 5.6219e-04, 1.3695e-03], + ..., + [ 4.1389e-04, 1.7233e-03, -6.3562e-04, ..., 1.3647e-03, + 2.1851e-04, 1.2312e-03], + [ 5.3215e-04, 1.4641e-02, 1.8511e-03, ..., -8.6823e-03, + 9.2850e-03, 2.5539e-03], + [ 7.7724e-04, -5.4131e-03, -2.3994e-03, ..., -4.6921e-03, + 3.2723e-05, -3.2177e-03]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0169, 0.0198, 0.0039, -0.0075, 0.0196, -0.0296, 0.0087, -0.0035, + 0.0007, -0.0117], device='cuda:0'), grad: tensor([ 0.0231, -0.0198, 0.0182, -0.0550, 0.0197, 0.0297, -0.0097, 0.0168, + 0.0021, -0.0250], device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 214.79, cls_loss 0.6736 cls_loss_mapping 0.0133 cls_loss_causal 0.5793 re_mapping 0.0166 re_causal 0.0372 /// teacc 98.35 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0558, -0.0680, -0.0741, ..., -0.0234, 0.0429, -0.0682], + [-0.0365, -0.0617, -0.0394, ..., 0.0859, -0.0294, 0.1556], + [ 0.0073, -0.0231, -0.0154, ..., -0.0095, -0.0101, -0.0555], + ..., + [ 0.0096, -0.0689, 0.1150, ..., 0.0153, -0.0511, 0.0360], + [-0.0202, 0.0436, -0.0873, ..., -0.0409, -0.0012, -0.0616], + [-0.0500, 0.0615, 0.0509, ..., -0.0455, -0.0290, -0.0188]], + device='cuda:0'), grad: tensor([[ 6.7241e-06, 1.2865e-03, 7.3910e-05, ..., 4.7913e-03, + 3.2336e-05, 3.8886e-04], + [ 2.8744e-05, 7.5531e-04, -4.7827e-04, ..., -1.5205e-02, + -1.0535e-05, -1.1463e-03], + [ 4.7356e-05, 4.1046e-03, 1.9760e-03, ..., 2.2629e-02, + 3.2872e-05, 1.4868e-03], + ..., + [-2.0194e-04, -1.0857e-02, -3.3245e-03, ..., -2.1423e-02, + 5.1036e-06, -8.1253e-03], + [ 9.5144e-06, 3.1891e-02, 2.0909e-04, ..., -2.9850e-03, + -1.4913e-04, 4.5204e-03], + [ 2.0787e-05, -5.1697e-02, -5.1975e-05, ..., 2.8839e-03, + 1.0490e-05, -1.7042e-03]], device='cuda:0') +Epoch 111, bias, value: tensor([-1.7131e-02, 2.0179e-02, 4.1354e-03, -6.8802e-03, 1.8627e-02, + -2.9886e-02, 9.6293e-03, -3.6894e-03, 9.0893e-05, -1.1622e-02], + device='cuda:0'), grad: tensor([ 0.0164, -0.0503, 0.0421, 0.0308, 0.0161, -0.0165, 0.0166, -0.0582, + 0.0222, -0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 214.82, cls_loss 0.6663 cls_loss_mapping 0.0164 cls_loss_causal 0.5709 re_mapping 0.0153 re_causal 0.0351 /// teacc 98.31 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0559, -0.0667, -0.0734, ..., -0.0240, 0.0428, -0.0682], + [-0.0368, -0.0616, -0.0391, ..., 0.0862, -0.0292, 0.1553], + [ 0.0077, -0.0230, -0.0151, ..., -0.0094, -0.0108, -0.0555], + ..., + [ 0.0094, -0.0680, 0.1154, ..., 0.0164, -0.0510, 0.0356], + [-0.0208, 0.0438, -0.0877, ..., -0.0419, -0.0009, -0.0614], + [-0.0501, 0.0611, 0.0508, ..., -0.0455, -0.0296, -0.0183]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.0146e-05, 5.2005e-05, ..., -7.1602e-03, + 1.2712e-03, -1.0223e-03], + [ 0.0000e+00, 8.4221e-05, 2.4128e-04, ..., 6.6032e-03, + 1.9300e-04, 1.8606e-03], + [ 0.0000e+00, 4.1885e-03, 1.4629e-03, ..., 1.1353e-02, + 4.6005e-03, 2.5730e-03], + ..., + [ 0.0000e+00, -1.9970e-03, -3.0003e-03, ..., -2.0809e-03, + -4.8141e-03, -5.5847e-03], + [ 0.0000e+00, -1.2993e-02, 3.1071e-03, ..., -1.6842e-03, + -8.9874e-03, 3.1528e-03], + [ 0.0000e+00, 1.0824e-03, -4.5514e-04, ..., -7.6065e-03, + 2.4738e-03, -2.4433e-03]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0171, 0.0202, 0.0048, -0.0071, 0.0193, -0.0295, 0.0088, -0.0036, + -0.0005, -0.0118], device='cuda:0'), grad: tensor([-0.0292, 0.0335, 0.0439, -0.0127, 0.0146, -0.0106, 0.0211, 0.0047, + 0.0025, -0.0678], device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 214.65, cls_loss 0.6733 cls_loss_mapping 0.0192 cls_loss_causal 0.5739 re_mapping 0.0159 re_causal 0.0347 /// teacc 98.48 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0562, -0.0680, -0.0741, ..., -0.0238, 0.0427, -0.0692], + [-0.0369, -0.0624, -0.0397, ..., 0.0866, -0.0282, 0.1552], + [ 0.0067, -0.0227, -0.0154, ..., -0.0104, -0.0104, -0.0556], + ..., + [ 0.0110, -0.0684, 0.1168, ..., 0.0174, -0.0510, 0.0355], + [-0.0207, 0.0446, -0.0890, ..., -0.0419, -0.0016, -0.0604], + [-0.0505, 0.0618, 0.0504, ..., -0.0464, -0.0279, -0.0182]], + device='cuda:0'), grad: tensor([[ 6.7521e-08, 2.9683e-04, 4.3607e-04, ..., 7.2899e-03, + 1.8501e-03, 2.3117e-03], + [ 5.9884e-07, 7.2861e-04, 5.1117e-04, ..., -1.5015e-02, + 1.2388e-03, -6.8398e-03], + [-2.7977e-06, 1.8682e-03, 8.4162e-04, ..., 3.2177e-03, + 3.2539e-03, 2.9182e-03], + ..., + [ 1.7602e-06, 1.7428e-04, -3.0861e-03, ..., 9.2392e-03, + 3.9177e-03, -1.0118e-03], + [-1.5497e-06, -3.0426e-02, 8.0585e-04, ..., -1.0178e-02, + -2.1420e-03, -1.1129e-03], + [ 8.3540e-07, 2.6459e-02, 5.5771e-03, ..., 9.3155e-03, + 8.5144e-03, 5.2338e-03]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0174, 0.0199, 0.0053, -0.0083, 0.0198, -0.0293, 0.0097, -0.0039, + -0.0006, -0.0116], device='cuda:0'), grad: tensor([ 0.0304, -0.0311, 0.0283, -0.0158, -0.0546, -0.0037, -0.0117, 0.0060, + -0.0236, 0.0759], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 112---------------------------------------------------- +epoch 112, time 230.90, cls_loss 0.6772 cls_loss_mapping 0.0181 cls_loss_causal 0.5860 re_mapping 0.0162 re_causal 0.0365 /// teacc 98.49 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0565, -0.0674, -0.0746, ..., -0.0256, 0.0437, -0.0704], + [-0.0372, -0.0632, -0.0408, ..., 0.0866, -0.0282, 0.1564], + [ 0.0067, -0.0211, -0.0146, ..., -0.0093, -0.0117, -0.0560], + ..., + [ 0.0111, -0.0693, 0.1168, ..., 0.0190, -0.0541, 0.0353], + [-0.0202, 0.0452, -0.0886, ..., -0.0416, -0.0007, -0.0596], + [-0.0512, 0.0613, 0.0511, ..., -0.0463, -0.0266, -0.0173]], + device='cuda:0'), grad: tensor([[ 1.5367e-07, -1.8044e-03, 5.3978e-04, ..., 8.2874e-04, + -5.3253e-03, 4.1127e-04], + [ 6.1514e-07, 9.3079e-04, 1.7548e-03, ..., 1.3695e-02, + 1.1152e-04, 5.2595e-04], + [-3.1851e-06, 1.4315e-03, 2.3136e-03, ..., -2.3327e-03, + 1.6584e-03, 4.3297e-04], + ..., + [ 5.5321e-06, 2.4486e-04, -4.3106e-03, ..., -8.3084e-03, + 9.5427e-05, -4.8866e-03], + [-8.2608e-07, -1.8663e-03, -7.9269e-03, ..., -2.5578e-03, + 2.6855e-03, 2.6875e-03], + [-8.0913e-06, 9.9480e-05, 3.2043e-03, ..., -1.1894e-02, + 3.6907e-04, 5.1880e-04]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0189, 0.0202, 0.0052, -0.0082, 0.0200, -0.0292, 0.0094, -0.0026, + -0.0004, -0.0120], device='cuda:0'), grad: tensor([ 0.0034, 0.0343, -0.0029, 0.0478, -0.0086, -0.0281, 0.0005, -0.0319, + 0.0150, -0.0294], device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 214.52, cls_loss 0.6698 cls_loss_mapping 0.0130 cls_loss_causal 0.5678 re_mapping 0.0160 re_causal 0.0360 /// teacc 98.41 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0571, -0.0688, -0.0745, ..., -0.0266, 0.0431, -0.0701], + [-0.0374, -0.0634, -0.0416, ..., 0.0870, -0.0288, 0.1567], + [ 0.0061, -0.0212, -0.0143, ..., -0.0085, -0.0122, -0.0563], + ..., + [ 0.0118, -0.0694, 0.1175, ..., 0.0179, -0.0537, 0.0357], + [-0.0199, 0.0450, -0.0895, ..., -0.0408, 0.0017, -0.0588], + [-0.0517, 0.0619, 0.0501, ..., -0.0464, -0.0271, -0.0174]], + device='cuda:0'), grad: tensor([[ 2.6505e-06, 5.8830e-05, 6.4731e-05, ..., 8.5526e-03, + -2.6264e-03, 1.1835e-03], + [ 1.0677e-05, 1.4687e-04, 1.7095e-04, ..., 1.6663e-02, + 2.4006e-05, -3.7193e-05], + [ 1.4484e-05, 1.4675e-04, -8.3260e-07, ..., 5.3520e-03, + 6.3848e-04, 9.0456e-04], + ..., + [ 1.6232e-03, 1.3590e-03, 3.3236e-04, ..., -1.4191e-02, + 1.0061e-04, 5.6190e-03], + [-2.0838e-04, 7.6389e-04, 2.8157e-04, ..., 1.0277e-02, + 4.3654e-04, 1.7958e-03], + [-1.5211e-03, -1.0195e-03, 3.8266e-04, ..., 7.0992e-03, + 1.8919e-04, -7.8812e-03]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0199, 0.0201, 0.0060, -0.0083, 0.0202, -0.0284, 0.0089, -0.0027, + -0.0005, -0.0121], device='cuda:0'), grad: tensor([ 0.0246, 0.0320, 0.0071, 0.0306, -0.0676, 0.0351, -0.0634, -0.0377, + 0.0331, 0.0060], device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 214.84, cls_loss 0.6693 cls_loss_mapping 0.0166 cls_loss_causal 0.5774 re_mapping 0.0159 re_causal 0.0363 /// teacc 98.36 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0577, -0.0684, -0.0755, ..., -0.0277, 0.0435, -0.0699], + [-0.0352, -0.0627, -0.0431, ..., 0.0871, -0.0286, 0.1561], + [ 0.0056, -0.0216, -0.0144, ..., -0.0091, -0.0116, -0.0569], + ..., + [ 0.0116, -0.0704, 0.1174, ..., 0.0171, -0.0546, 0.0357], + [-0.0199, 0.0452, -0.0914, ..., -0.0409, 0.0019, -0.0581], + [-0.0534, 0.0619, 0.0507, ..., -0.0456, -0.0270, -0.0172]], + device='cuda:0'), grad: tensor([[ 1.7062e-05, 7.7057e-04, 1.8191e-04, ..., 4.9934e-03, + 5.8889e-04, 1.4620e-03], + [ 1.0335e-04, -1.3809e-02, 6.4945e-04, ..., -1.4076e-02, + 1.7434e-05, -9.9411e-03], + [-7.8499e-05, -1.0723e-04, -5.7077e-04, ..., 4.0550e-03, + -7.9989e-05, 7.9107e-04], + ..., + [-3.5834e-04, 5.8746e-04, -3.2654e-03, ..., -7.3547e-03, + 1.1578e-05, 1.3065e-03], + [ 4.6849e-05, 4.9591e-03, 3.8934e-04, ..., 1.5030e-03, + 1.2469e-04, -1.0843e-03], + [ 3.4153e-05, 1.3142e-03, 2.8801e-04, ..., 7.5836e-03, + 3.3975e-05, 1.8282e-03]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0212, 0.0201, 0.0059, -0.0088, 0.0214, -0.0278, 0.0093, -0.0039, + -0.0004, -0.0112], device='cuda:0'), grad: tensor([ 0.0159, -0.0298, 0.0149, 0.0319, 0.0228, -0.0360, -0.0292, -0.0202, + -0.0003, 0.0300], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 115---------------------------------------------------- +epoch 115, time 231.17, cls_loss 0.6448 cls_loss_mapping 0.0100 cls_loss_causal 0.5449 re_mapping 0.0160 re_causal 0.0373 /// teacc 98.50 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0581, -0.0689, -0.0750, ..., -0.0269, 0.0435, -0.0701], + [-0.0367, -0.0629, -0.0430, ..., 0.0872, -0.0274, 0.1577], + [ 0.0064, -0.0228, -0.0157, ..., -0.0099, -0.0111, -0.0569], + ..., + [ 0.0105, -0.0714, 0.1177, ..., 0.0169, -0.0546, 0.0354], + [-0.0208, 0.0454, -0.0921, ..., -0.0411, 0.0016, -0.0585], + [-0.0535, 0.0619, 0.0524, ..., -0.0456, -0.0266, -0.0182]], + device='cuda:0'), grad: tensor([[ 3.0752e-06, 1.4174e-04, 2.3282e-04, ..., 2.1820e-03, + -1.2569e-05, 8.8978e-04], + [ 1.8656e-05, 3.0732e-04, 8.9073e-04, ..., -9.4223e-03, + 8.8103e-07, -1.9855e-03], + [ 1.3165e-05, -3.5572e-03, -4.3602e-03, ..., -4.2381e-03, + -4.6641e-06, -3.6640e-03], + ..., + [ 1.5646e-05, 1.8024e-03, 1.8778e-03, ..., 5.4703e-03, + 1.2461e-06, 3.1891e-03], + [ 2.0280e-05, 5.9814e-03, 1.2951e-03, ..., 5.0049e-03, + 2.7809e-06, 2.2793e-03], + [ 5.3607e-06, -8.4076e-03, 1.3075e-03, ..., 4.4403e-03, + 9.0431e-07, 4.1885e-03]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0217, 0.0207, 0.0056, -0.0085, 0.0205, -0.0267, 0.0087, -0.0039, + -0.0004, -0.0110], device='cuda:0'), grad: tensor([ 0.0312, -0.0386, -0.0126, 0.0030, -0.0291, 0.0076, -0.0149, 0.0222, + 0.0278, 0.0035], device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 214.27, cls_loss 0.6873 cls_loss_mapping 0.0164 cls_loss_causal 0.5862 re_mapping 0.0151 re_causal 0.0330 /// teacc 98.33 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0586, -0.0692, -0.0761, ..., -0.0264, 0.0442, -0.0698], + [-0.0366, -0.0620, -0.0415, ..., 0.0878, -0.0266, 0.1586], + [ 0.0062, -0.0230, -0.0167, ..., -0.0106, -0.0114, -0.0572], + ..., + [ 0.0109, -0.0720, 0.1180, ..., 0.0165, -0.0549, 0.0363], + [-0.0215, 0.0461, -0.0928, ..., -0.0412, 0.0019, -0.0583], + [-0.0538, 0.0612, 0.0529, ..., -0.0457, -0.0269, -0.0188]], + device='cuda:0'), grad: tensor([[ 2.4402e-04, 6.4774e-03, 2.0962e-03, ..., 4.0779e-03, + 4.5929e-03, 9.1171e-04], + [ 1.4104e-05, 1.8778e-03, -1.7197e-02, ..., 8.3160e-03, + -3.3600e-02, 1.6279e-03], + [ 1.0133e-04, 4.1924e-03, 4.8714e-03, ..., 7.1449e-03, + 8.1558e-03, 8.8501e-04], + ..., + [ 1.0014e-05, -1.7532e-02, -1.1339e-03, ..., 8.7128e-03, + 1.8244e-03, -1.2531e-03], + [ 4.6182e-04, 7.8583e-03, 7.7400e-03, ..., -4.8981e-03, + 1.4214e-02, 1.6193e-03], + [ 9.2536e-06, -1.7807e-05, -3.5343e-03, ..., -2.8706e-03, + 9.6703e-04, -6.2332e-03]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0208, 0.0216, 0.0043, -0.0089, 0.0203, -0.0270, 0.0090, -0.0039, + -0.0002, -0.0111], device='cuda:0'), grad: tensor([ 0.0355, 0.0001, 0.0386, -0.0031, -0.0111, -0.0138, -0.0309, -0.0211, + 0.0197, -0.0141], device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 214.23, cls_loss 0.6480 cls_loss_mapping 0.0132 cls_loss_causal 0.5558 re_mapping 0.0153 re_causal 0.0345 /// teacc 98.34 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0596, -0.0692, -0.0749, ..., -0.0260, 0.0446, -0.0692], + [-0.0370, -0.0629, -0.0417, ..., 0.0874, -0.0254, 0.1583], + [ 0.0063, -0.0225, -0.0175, ..., -0.0103, -0.0122, -0.0564], + ..., + [ 0.0119, -0.0704, 0.1188, ..., 0.0178, -0.0559, 0.0366], + [-0.0215, 0.0452, -0.0943, ..., -0.0417, 0.0025, -0.0592], + [-0.0543, 0.0621, 0.0534, ..., -0.0461, -0.0275, -0.0194]], + device='cuda:0'), grad: tensor([[ 1.1660e-05, 5.5122e-04, -2.9011e-03, ..., -2.4891e-03, + 1.2350e-03, -1.7881e-03], + [ 1.8990e-06, 7.2479e-04, 1.4238e-03, ..., 2.5833e-02, + 4.6182e-04, 2.5177e-02], + [ 1.1332e-05, 1.9989e-03, 1.3039e-02, ..., 1.4763e-02, + 1.5488e-03, 3.7060e-03], + ..., + [ 2.8110e-04, 1.4553e-03, -6.4201e-03, ..., -4.5357e-03, + 1.6260e-04, -8.6670e-03], + [ 7.4059e-06, -2.3060e-03, 1.7977e-03, ..., 2.0170e-04, + 1.4019e-03, 8.9645e-04], + [ 2.0180e-03, -5.4092e-03, -7.2002e-04, ..., 1.4923e-02, + 5.6553e-04, 1.3741e-02]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0206, 0.0214, 0.0046, -0.0091, 0.0204, -0.0280, 0.0099, -0.0029, + -0.0010, -0.0112], device='cuda:0'), grad: tensor([-0.0062, 0.0181, 0.0445, -0.0038, 0.0209, -0.0723, -0.0256, -0.0215, + 0.0059, 0.0402], device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 214.35, cls_loss 0.6700 cls_loss_mapping 0.0173 cls_loss_causal 0.5746 re_mapping 0.0157 re_causal 0.0359 /// teacc 98.46 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0591, -0.0709, -0.0751, ..., -0.0257, 0.0450, -0.0695], + [-0.0379, -0.0639, -0.0423, ..., 0.0867, -0.0245, 0.1588], + [ 0.0069, -0.0221, -0.0180, ..., -0.0107, -0.0126, -0.0581], + ..., + [ 0.0090, -0.0710, 0.1183, ..., 0.0186, -0.0564, 0.0370], + [-0.0222, 0.0452, -0.0950, ..., -0.0413, 0.0020, -0.0606], + [-0.0541, 0.0622, 0.0540, ..., -0.0467, -0.0280, -0.0187]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0027, 0.0025, ..., 0.0060, 0.0004, 0.0065], + [-0.0025, -0.0004, -0.0016, ..., -0.0122, 0.0003, -0.0113], + [ 0.0001, 0.0011, 0.0015, ..., 0.0029, 0.0003, 0.0029], + ..., + [ 0.0004, 0.0026, 0.0041, ..., 0.0090, 0.0005, 0.0076], + [ 0.0002, 0.0023, 0.0027, ..., 0.0065, 0.0008, 0.0061], + [ 0.0004, 0.0016, 0.0003, ..., 0.0022, 0.0006, 0.0025]], + device='cuda:0') +Epoch 120, bias, value: tensor([-0.0198, 0.0212, 0.0034, -0.0086, 0.0203, -0.0275, 0.0092, -0.0025, + -0.0009, -0.0113], device='cuda:0'), grad: tensor([ 0.0265, -0.0368, 0.0135, -0.0378, 0.0070, -0.0322, -0.0132, 0.0374, + 0.0278, 0.0077], device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 214.81, cls_loss 0.6545 cls_loss_mapping 0.0149 cls_loss_causal 0.5676 re_mapping 0.0161 re_causal 0.0361 /// teacc 98.46 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0551, -0.0712, -0.0750, ..., -0.0250, 0.0452, -0.0697], + [-0.0396, -0.0647, -0.0437, ..., 0.0865, -0.0272, 0.1591], + [ 0.0095, -0.0219, -0.0156, ..., -0.0100, -0.0106, -0.0566], + ..., + [ 0.0097, -0.0712, 0.1181, ..., 0.0175, -0.0570, 0.0354], + [-0.0236, 0.0454, -0.0949, ..., -0.0411, 0.0012, -0.0614], + [-0.0529, 0.0624, 0.0531, ..., -0.0468, -0.0277, -0.0178]], + device='cuda:0'), grad: tensor([[ 1.5683e-06, 1.0693e-04, 1.4555e-04, ..., 2.1172e-03, + 8.9467e-05, 5.5045e-05], + [ 1.4715e-07, 1.7452e-04, 5.9175e-04, ..., 4.4174e-03, + 2.0397e-04, 2.6417e-04], + [ 2.3037e-05, 9.3412e-04, 3.7460e-03, ..., 1.2863e-02, + 4.3130e-04, 3.3398e-03], + ..., + [ 1.9744e-07, -8.3780e-04, -5.4741e-03, ..., -2.1225e-02, + -1.0204e-03, -4.2114e-03], + [ 3.0939e-06, 1.1520e-03, 1.0939e-03, ..., 3.8700e-03, + 2.2423e-04, 5.2071e-04], + [ 1.2852e-06, -5.1514e-02, 1.0004e-03, ..., -3.3069e-04, + -2.6001e-02, 2.3174e-03]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0194, 0.0214, 0.0047, -0.0089, 0.0205, -0.0285, 0.0091, -0.0035, + -0.0003, -0.0118], device='cuda:0'), grad: tensor([ 0.0092, 0.0230, 0.0346, 0.0089, -0.0326, 0.0040, 0.0097, -0.0600, + -0.0087, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 214.79, cls_loss 0.6545 cls_loss_mapping 0.0160 cls_loss_causal 0.5590 re_mapping 0.0154 re_causal 0.0339 /// teacc 98.46 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0555, -0.0720, -0.0750, ..., -0.0258, 0.0436, -0.0714], + [-0.0396, -0.0655, -0.0431, ..., 0.0871, -0.0275, 0.1608], + [ 0.0090, -0.0239, -0.0173, ..., -0.0111, -0.0108, -0.0568], + ..., + [ 0.0100, -0.0713, 0.1190, ..., 0.0184, -0.0575, 0.0356], + [-0.0236, 0.0459, -0.0938, ..., -0.0412, 0.0024, -0.0621], + [-0.0537, 0.0618, 0.0519, ..., -0.0473, -0.0280, -0.0167]], + device='cuda:0'), grad: tensor([[-2.1279e-04, -1.3475e-03, 4.4513e-04, ..., -4.7035e-03, + 1.2851e-04, -8.3399e-04], + [ 1.1809e-05, 9.2164e-06, 4.1604e-04, ..., 5.5084e-03, + 3.0017e-04, 2.8419e-03], + [ 1.8110e-03, -1.5778e-02, 3.9482e-03, ..., -4.3716e-03, + 1.2136e-04, -1.6174e-03], + ..., + [ 1.4029e-03, -6.7902e-03, -8.9073e-04, ..., -5.5981e-04, + 1.7190e-04, 1.4658e-03], + [ 8.6963e-05, 5.8861e-03, 8.1873e-04, ..., 5.7297e-03, + 8.3780e-04, 4.3793e-03], + [ 3.3051e-05, 9.1324e-03, 7.5912e-04, ..., 3.9330e-03, + 2.5582e-04, -1.4067e-03]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0200, 0.0217, 0.0042, -0.0086, 0.0213, -0.0285, 0.0089, -0.0033, + -0.0003, -0.0120], device='cuda:0'), grad: tensor([ 0.0044, 0.0122, -0.0046, -0.0050, 0.0244, -0.0371, -0.0015, -0.0112, + 0.0347, -0.0163], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 121---------------------------------------------------- +epoch 121, time 231.02, cls_loss 0.6448 cls_loss_mapping 0.0115 cls_loss_causal 0.5644 re_mapping 0.0153 re_causal 0.0348 /// teacc 98.54 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0558, -0.0721, -0.0766, ..., -0.0265, 0.0430, -0.0718], + [-0.0391, -0.0659, -0.0429, ..., 0.0871, -0.0291, 0.1599], + [ 0.0100, -0.0230, -0.0158, ..., -0.0103, -0.0113, -0.0553], + ..., + [ 0.0090, -0.0710, 0.1190, ..., 0.0176, -0.0584, 0.0360], + [-0.0234, 0.0460, -0.0949, ..., -0.0409, 0.0034, -0.0630], + [-0.0541, 0.0612, 0.0514, ..., -0.0472, -0.0278, -0.0178]], + device='cuda:0'), grad: tensor([[ 1.7476e-04, -6.4697e-03, -1.0162e-02, ..., -3.5278e-02, + -1.5268e-03, 3.8218e-04], + [-3.8266e-04, -4.1389e-03, 8.7023e-05, ..., -7.2784e-03, + 1.1454e-03, -7.1869e-03], + [ 2.3115e-04, 3.8910e-03, 5.9052e-03, ..., 9.9182e-03, + 1.3647e-03, 4.4990e-04], + ..., + [ 1.3542e-04, 2.2948e-04, 3.9749e-03, ..., 8.2703e-03, + 7.3147e-04, 1.0033e-02], + [ 5.9414e-04, 5.2261e-03, 1.0767e-03, ..., 8.5373e-03, + -4.4327e-03, 7.6370e-03], + [ 1.3053e-04, 5.7411e-04, -4.6463e-03, ..., -1.3888e-04, + 9.3365e-04, -1.1665e-02]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0207, 0.0213, 0.0046, -0.0082, 0.0204, -0.0288, 0.0098, -0.0032, + 0.0004, -0.0123], device='cuda:0'), grad: tensor([-0.0918, -0.0172, -0.0010, 0.0340, -0.0060, -0.0007, 0.0168, 0.0410, + 0.0317, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 214.58, cls_loss 0.6333 cls_loss_mapping 0.0143 cls_loss_causal 0.5399 re_mapping 0.0149 re_causal 0.0333 /// teacc 98.33 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0559, -0.0714, -0.0743, ..., -0.0252, 0.0426, -0.0711], + [-0.0389, -0.0659, -0.0439, ..., 0.0870, -0.0284, 0.1606], + [ 0.0101, -0.0231, -0.0182, ..., -0.0115, -0.0116, -0.0554], + ..., + [ 0.0087, -0.0713, 0.1188, ..., 0.0181, -0.0588, 0.0362], + [-0.0228, 0.0459, -0.0959, ..., -0.0421, 0.0046, -0.0636], + [-0.0543, 0.0618, 0.0512, ..., -0.0477, -0.0286, -0.0186]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7336e-04, -5.1409e-05, ..., -2.3697e-02, + -4.7150e-03, -1.9531e-03], + [ 0.0000e+00, 3.8567e-03, 3.0918e-03, ..., 2.3026e-02, + 2.0161e-03, 1.4282e-02], + [ 0.0000e+00, 1.3542e-03, -3.9368e-03, ..., -1.7653e-03, + -4.0359e-03, -2.9182e-03], + ..., + [ 0.0000e+00, 6.5327e-04, 4.0512e-03, ..., 1.1887e-02, + 7.4768e-04, 7.4463e-03], + [ 0.0000e+00, -4.1008e-03, 4.5919e-04, ..., -5.0774e-03, + 6.2180e-04, -1.1787e-02], + [ 0.0000e+00, 6.4659e-04, 8.8644e-04, ..., 5.7106e-03, + 6.7616e-04, 3.4466e-03]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0194, 0.0219, 0.0035, -0.0087, 0.0210, -0.0289, 0.0098, -0.0036, + 0.0005, -0.0125], device='cuda:0'), grad: tensor([-0.0768, 0.0848, 0.0168, 0.0234, -0.0169, -0.0417, -0.0034, 0.0410, + -0.0537, 0.0264], device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 214.49, cls_loss 0.6319 cls_loss_mapping 0.0106 cls_loss_causal 0.5370 re_mapping 0.0156 re_causal 0.0350 /// teacc 98.50 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0569, -0.0722, -0.0734, ..., -0.0254, 0.0438, -0.0721], + [-0.0399, -0.0655, -0.0434, ..., 0.0873, -0.0283, 0.1601], + [ 0.0108, -0.0238, -0.0188, ..., -0.0110, -0.0113, -0.0568], + ..., + [ 0.0088, -0.0708, 0.1196, ..., 0.0174, -0.0591, 0.0374], + [-0.0220, 0.0461, -0.0966, ..., -0.0423, 0.0047, -0.0627], + [-0.0543, 0.0626, 0.0515, ..., -0.0475, -0.0289, -0.0187]], + device='cuda:0'), grad: tensor([[ 1.5235e-04, -5.4121e-05, 1.9217e-04, ..., 1.0614e-03, + 1.5030e-02, 5.0449e-04], + [ 2.3437e-04, 3.0398e-05, 1.8752e-04, ..., 4.4899e-03, + 9.4128e-04, 4.4227e-04], + [ 3.6049e-04, -4.4250e-03, 4.5824e-04, ..., -2.1194e-02, + -1.4969e-02, 8.9312e-04], + ..., + [-1.3819e-03, 1.5402e-03, -3.1853e-03, ..., -3.9220e-05, + 4.6134e-04, -7.6599e-03], + [ 8.0538e-04, 4.7636e-04, 1.5478e-03, ..., 5.9509e-03, + 8.5497e-04, 1.3971e-03], + [ 1.4114e-03, 6.2287e-05, 1.8559e-03, ..., 4.2381e-03, + 3.2282e-04, 3.6430e-03]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0195, 0.0217, 0.0034, -0.0081, 0.0210, -0.0284, 0.0094, -0.0040, + 0.0003, -0.0125], device='cuda:0'), grad: tensor([ 0.0102, 0.0215, -0.0624, 0.0425, -0.0003, -0.0420, 0.0200, -0.0530, + 0.0311, 0.0327], device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 214.50, cls_loss 0.6436 cls_loss_mapping 0.0100 cls_loss_causal 0.5437 re_mapping 0.0151 re_causal 0.0347 /// teacc 98.36 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0578, -0.0730, -0.0734, ..., -0.0252, 0.0433, -0.0724], + [-0.0412, -0.0650, -0.0434, ..., 0.0873, -0.0279, 0.1609], + [ 0.0125, -0.0242, -0.0196, ..., -0.0104, -0.0107, -0.0577], + ..., + [ 0.0100, -0.0717, 0.1196, ..., 0.0173, -0.0606, 0.0375], + [-0.0217, 0.0454, -0.0970, ..., -0.0426, 0.0044, -0.0632], + [-0.0543, 0.0635, 0.0524, ..., -0.0472, -0.0296, -0.0170]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.4643e-03, 2.9683e-04, ..., -2.9793e-03, + -2.6989e-03, -4.5747e-05], + [ 0.0000e+00, -7.3051e-03, 1.7726e-04, ..., -9.5654e-04, + 5.3674e-05, -2.3997e-04], + [ 0.0000e+00, -1.2589e-03, -1.6727e-03, ..., -5.1384e-03, + 5.7936e-04, 5.7161e-05], + ..., + [ 0.0000e+00, -4.4739e-02, -4.0703e-03, ..., -2.6016e-03, + 3.2812e-05, -6.0577e-03], + [ 0.0000e+00, 2.0733e-03, 2.2650e-04, ..., 2.1915e-03, + 5.9938e-04, 1.5986e-04], + [ 0.0000e+00, 5.1056e-02, 4.4518e-03, ..., 4.0398e-03, + 1.8334e-04, 6.0043e-03]], device='cuda:0') +Epoch 126, bias, value: tensor([-1.8590e-02, 2.1252e-02, 4.5445e-03, -7.9737e-03, 2.0034e-02, + -2.9461e-02, 9.0704e-03, -3.5808e-03, -4.5916e-05, -1.1839e-02], + device='cuda:0'), grad: tensor([-0.0197, 0.0083, -0.0076, -0.0020, 0.0154, 0.0175, 0.0012, -0.0823, + 0.0200, 0.0493], device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 214.54, cls_loss 0.6505 cls_loss_mapping 0.0134 cls_loss_causal 0.5602 re_mapping 0.0150 re_causal 0.0335 /// teacc 98.53 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0585, -0.0738, -0.0738, ..., -0.0262, 0.0432, -0.0746], + [-0.0410, -0.0644, -0.0442, ..., 0.0869, -0.0271, 0.1617], + [ 0.0120, -0.0248, -0.0193, ..., -0.0098, -0.0108, -0.0574], + ..., + [ 0.0124, -0.0702, 0.1196, ..., 0.0185, -0.0611, 0.0375], + [-0.0213, 0.0464, -0.0945, ..., -0.0424, 0.0053, -0.0624], + [-0.0571, 0.0621, 0.0523, ..., -0.0483, -0.0310, -0.0169]], + device='cuda:0'), grad: tensor([[ 8.1025e-08, 3.2163e-04, -1.5007e-02, ..., -1.5015e-02, + -5.8289e-03, -5.4131e-03], + [ 1.7613e-05, 5.2452e-05, 2.5749e-04, ..., -5.5199e-03, + 6.4278e-04, -3.2768e-03], + [ 3.3248e-06, 1.7481e-03, 1.1950e-03, ..., 2.7142e-03, + 6.0606e-04, 7.1192e-04], + ..., + [-2.8864e-05, 2.1782e-03, 9.2468e-03, ..., 9.3536e-03, + 1.2856e-03, 3.6793e-03], + [-1.1921e-06, -2.4338e-03, 4.3654e-04, ..., 2.6455e-03, + 1.0395e-03, 7.8535e-04], + [ 4.4964e-06, -2.2659e-03, 2.4948e-03, ..., 5.0278e-03, + 9.9754e-04, 1.6680e-03]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0204, 0.0221, 0.0046, -0.0092, 0.0196, -0.0294, 0.0103, -0.0024, + 0.0008, -0.0128], device='cuda:0'), grad: tensor([-0.0571, -0.0101, 0.0160, 0.0158, -0.0036, 0.0021, 0.0232, 0.0435, + -0.0167, -0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 214.78, cls_loss 0.6559 cls_loss_mapping 0.0111 cls_loss_causal 0.5633 re_mapping 0.0148 re_causal 0.0337 /// teacc 98.36 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0583, -0.0748, -0.0733, ..., -0.0244, 0.0431, -0.0742], + [-0.0409, -0.0647, -0.0453, ..., 0.0868, -0.0278, 0.1622], + [ 0.0112, -0.0245, -0.0185, ..., -0.0103, -0.0097, -0.0575], + ..., + [ 0.0130, -0.0711, 0.1189, ..., 0.0183, -0.0639, 0.0362], + [-0.0225, 0.0459, -0.0951, ..., -0.0426, 0.0049, -0.0624], + [-0.0560, 0.0623, 0.0531, ..., -0.0487, -0.0315, -0.0160]], + device='cuda:0'), grad: tensor([[ 1.2137e-05, 2.0027e-03, 3.1710e-04, ..., -3.5248e-03, + 3.5419e-03, -3.9554e-04], + [ 2.4962e-04, 1.5030e-03, 7.3338e-04, ..., 3.3932e-03, + -1.6678e-02, 8.8263e-04], + [ 1.1663e-03, 2.4662e-03, 1.7662e-03, ..., -1.6012e-03, + -9.1982e-04, 2.9640e-03], + ..., + [-2.2717e-03, -6.7406e-03, -5.6229e-03, ..., -1.8173e-02, + -8.3466e-03, -7.5569e-03], + [ 3.0994e-04, -1.0391e-02, 7.7343e-04, ..., 3.9215e-03, + 1.2085e-02, 8.8072e-04], + [ 2.3890e-04, -3.7659e-02, -4.1199e-03, ..., 3.8414e-03, + -5.2986e-03, 1.2693e-03]], device='cuda:0') +Epoch 128, bias, value: tensor([-1.8855e-02, 2.1831e-02, 5.1873e-03, -8.4500e-03, 1.9203e-02, + -2.9468e-02, 9.8211e-03, -3.1128e-03, -2.0713e-05, -1.2689e-02], + device='cuda:0'), grad: tensor([-0.0109, -0.0083, -0.0086, 0.0563, 0.0306, 0.0204, -0.0054, -0.0776, + 0.0239, -0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 214.52, cls_loss 0.6372 cls_loss_mapping 0.0103 cls_loss_causal 0.5333 re_mapping 0.0146 re_causal 0.0333 /// teacc 98.35 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0577, -0.0747, -0.0735, ..., -0.0245, 0.0434, -0.0746], + [-0.0427, -0.0671, -0.0451, ..., 0.0872, -0.0269, 0.1622], + [ 0.0113, -0.0238, -0.0190, ..., -0.0106, -0.0102, -0.0571], + ..., + [ 0.0136, -0.0711, 0.1194, ..., 0.0178, -0.0640, 0.0364], + [-0.0220, 0.0455, -0.0961, ..., -0.0422, 0.0039, -0.0621], + [-0.0570, 0.0636, 0.0528, ..., -0.0494, -0.0316, -0.0165]], + device='cuda:0'), grad: tensor([[ 1.9324e-04, -2.4681e-03, 6.6900e-04, ..., 9.5797e-04, + 2.7485e-03, 4.2686e-03], + [ 2.7156e-04, 2.4486e-04, -2.4891e-04, ..., -7.7095e-03, + 4.7569e-03, -7.1259e-03], + [ 1.6093e-04, 2.9488e-03, 1.4138e-04, ..., -5.5343e-05, + 3.8948e-03, -1.8158e-03], + ..., + [ 6.7353e-05, -1.6899e-03, 1.5278e-03, ..., 8.6260e-04, + 2.6798e-03, -5.9748e-04], + [ 2.0933e-04, -2.5539e-03, 5.4026e-04, ..., 9.8896e-04, + 4.6501e-03, 5.0354e-03], + [ 1.7905e-04, 7.5769e-04, -4.6349e-03, ..., 3.4542e-03, + 2.5597e-03, 1.8272e-03]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0187, 0.0220, 0.0049, -0.0084, 0.0203, -0.0287, 0.0091, -0.0038, + 0.0003, -0.0135], device='cuda:0'), grad: tensor([-0.0124, -0.0218, 0.0058, 0.0223, -0.0561, 0.0056, 0.0078, 0.0065, + 0.0219, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 214.81, cls_loss 0.6622 cls_loss_mapping 0.0132 cls_loss_causal 0.5774 re_mapping 0.0141 re_causal 0.0318 /// teacc 98.37 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0575, -0.0752, -0.0738, ..., -0.0254, 0.0436, -0.0758], + [-0.0454, -0.0670, -0.0448, ..., 0.0877, -0.0259, 0.1636], + [ 0.0115, -0.0248, -0.0201, ..., -0.0106, -0.0114, -0.0569], + ..., + [ 0.0142, -0.0722, 0.1202, ..., 0.0178, -0.0638, 0.0362], + [-0.0229, 0.0464, -0.0967, ..., -0.0414, 0.0032, -0.0624], + [-0.0567, 0.0637, 0.0530, ..., -0.0485, -0.0311, -0.0162]], + device='cuda:0'), grad: tensor([[-8.1301e-04, -7.4911e-04, -2.3899e-03, ..., -7.0953e-03, + -1.6689e-03, -2.9755e-03], + [ 8.2684e-04, 1.9085e-04, 6.0196e-03, ..., 1.0612e-02, + 5.8289e-03, -1.0071e-02], + [ 4.5371e-04, -2.5082e-03, -3.7308e-03, ..., -7.1945e-03, + 2.6722e-03, -7.5455e-03], + ..., + [ 1.3494e-04, 5.8746e-03, 1.1986e-02, ..., 4.0550e-03, + -6.1274e-05, 1.5762e-02], + [ 9.6560e-04, 1.5480e-02, 8.4000e-03, ..., 1.1139e-02, + 1.5640e-02, 6.2599e-03], + [ 3.5000e-04, 1.2566e-02, -8.5297e-03, ..., 6.8283e-03, + 4.8294e-03, 4.8447e-03]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0188, 0.0229, 0.0048, -0.0087, 0.0207, -0.0295, 0.0085, -0.0041, + 0.0009, -0.0132], device='cuda:0'), grad: tensor([-0.0190, 0.0321, -0.0117, -0.0955, 0.0134, -0.0087, 0.0030, 0.0190, + 0.0475, 0.0199], device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 214.45, cls_loss 0.6417 cls_loss_mapping 0.0149 cls_loss_causal 0.5528 re_mapping 0.0149 re_causal 0.0329 /// teacc 98.40 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0575, -0.0759, -0.0730, ..., -0.0261, 0.0437, -0.0757], + [-0.0441, -0.0687, -0.0459, ..., 0.0887, -0.0277, 0.1638], + [ 0.0104, -0.0232, -0.0186, ..., -0.0107, -0.0107, -0.0548], + ..., + [ 0.0136, -0.0725, 0.1194, ..., 0.0174, -0.0632, 0.0362], + [-0.0215, 0.0472, -0.0982, ..., -0.0418, 0.0028, -0.0641], + [-0.0573, 0.0636, 0.0536, ..., -0.0472, -0.0296, -0.0156]], + device='cuda:0'), grad: tensor([[ 3.9196e-04, 2.9964e-03, 6.2609e-04, ..., 9.8705e-04, + 8.7509e-03, 2.0161e-03], + [-2.3918e-03, -3.7422e-03, 2.3544e-05, ..., -1.2283e-02, + -2.0294e-03, -1.2751e-03], + [ 1.9159e-03, 2.1172e-03, 1.7471e-03, ..., 2.2564e-03, + -1.9894e-03, 3.5725e-03], + ..., + [-2.5997e-03, 8.5020e-04, -3.7422e-03, ..., 3.2234e-03, + 1.2165e-04, -3.2215e-03], + [ 1.4162e-03, 3.1490e-03, 1.7939e-03, ..., 8.4000e-03, + 1.0710e-03, 1.3571e-03], + [ 2.5368e-03, 3.8357e-03, 4.4594e-03, ..., 9.7656e-03, + 4.0650e-04, 5.9509e-03]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0194, 0.0234, 0.0050, -0.0082, 0.0194, -0.0300, 0.0090, -0.0043, + 0.0010, -0.0125], device='cuda:0'), grad: tensor([ 0.0003, -0.0346, -0.0005, 0.0262, -0.0157, 0.0271, -0.0869, 0.0098, + 0.0333, 0.0409], device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 214.42, cls_loss 0.6261 cls_loss_mapping 0.0111 cls_loss_causal 0.5351 re_mapping 0.0151 re_causal 0.0346 /// teacc 98.30 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0587, -0.0767, -0.0733, ..., -0.0265, 0.0436, -0.0770], + [-0.0433, -0.0679, -0.0462, ..., 0.0889, -0.0280, 0.1639], + [ 0.0103, -0.0218, -0.0188, ..., -0.0106, -0.0110, -0.0551], + ..., + [ 0.0135, -0.0722, 0.1193, ..., 0.0174, -0.0626, 0.0367], + [-0.0225, 0.0476, -0.0991, ..., -0.0438, 0.0036, -0.0648], + [-0.0553, 0.0637, 0.0535, ..., -0.0471, -0.0284, -0.0150]], + device='cuda:0'), grad: tensor([[ 7.7933e-06, 8.1825e-04, 2.7204e-04, ..., 8.5526e-03, + 2.1875e-04, 2.7790e-03], + [ 1.8835e-04, 8.8573e-05, 7.7486e-04, ..., -9.7752e-04, + -6.5994e-04, -1.0881e-03], + [ 2.7180e-04, 3.0303e-04, 1.3247e-03, ..., 1.5564e-02, + 6.5565e-04, 3.9635e-03], + ..., + [ 3.2425e-04, 2.9302e-04, -5.5027e-04, ..., -6.2027e-03, + 5.6362e-04, -3.6469e-03], + [-6.0111e-05, 7.0763e-04, 4.5347e-04, ..., -1.2505e-02, + -1.8520e-03, -2.8229e-03], + [ 2.6122e-05, -5.7518e-05, 8.5688e-04, ..., -5.2147e-03, + 1.1259e-04, 7.6246e-04]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0191, 0.0240, 0.0045, -0.0087, 0.0195, -0.0307, 0.0104, -0.0044, + -0.0003, -0.0118], device='cuda:0'), grad: tensor([ 0.0205, -0.0033, 0.0345, -0.0237, 0.0091, 0.0029, 0.0168, -0.0257, + -0.0155, -0.0156], device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 214.67, cls_loss 0.6577 cls_loss_mapping 0.0176 cls_loss_causal 0.5748 re_mapping 0.0144 re_causal 0.0314 /// teacc 98.38 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0591, -0.0774, -0.0733, ..., -0.0264, 0.0445, -0.0776], + [-0.0449, -0.0685, -0.0457, ..., 0.0896, -0.0268, 0.1647], + [ 0.0101, -0.0223, -0.0203, ..., -0.0119, -0.0114, -0.0574], + ..., + [ 0.0151, -0.0720, 0.1198, ..., 0.0175, -0.0625, 0.0363], + [-0.0222, 0.0479, -0.0996, ..., -0.0431, 0.0037, -0.0639], + [-0.0551, 0.0642, 0.0544, ..., -0.0469, -0.0286, -0.0142]], + device='cuda:0'), grad: tensor([[ 5.7650e-04, 1.9207e-03, 8.2135e-05, ..., -2.4757e-03, + 1.3647e-03, 3.3879e-04], + [ 8.1539e-04, 3.0422e-04, 3.6329e-05, ..., 1.9882e-02, + 1.9479e-04, 1.2722e-03], + [ 6.7558e-03, 1.2798e-03, 1.9336e-04, ..., -9.3231e-03, + 4.3416e-04, 8.8978e-04], + ..., + [ 1.0567e-03, 4.3392e-04, 9.1791e-05, ..., -1.4668e-03, + 5.9271e-04, 1.0395e-03], + [-1.1772e-02, -9.5062e-03, -1.1271e-04, ..., -1.4420e-02, + -6.7787e-03, -1.1906e-05], + [ 4.3869e-04, -6.6795e-03, -2.1114e-03, ..., 4.0131e-03, + -1.1539e-03, 1.7624e-03]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0189, 0.0239, 0.0041, -0.0091, 0.0194, -0.0297, 0.0096, -0.0047, + 0.0004, -0.0116], device='cuda:0'), grad: tensor([-0.0056, 0.0471, 0.0191, 0.0196, -0.0171, 0.0249, -0.0070, -0.0037, + -0.0782, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 214.41, cls_loss 0.6281 cls_loss_mapping 0.0107 cls_loss_causal 0.5414 re_mapping 0.0148 re_causal 0.0331 /// teacc 98.36 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0587, -0.0783, -0.0736, ..., -0.0267, 0.0444, -0.0786], + [-0.0452, -0.0691, -0.0474, ..., 0.0888, -0.0260, 0.1649], + [ 0.0085, -0.0219, -0.0198, ..., -0.0109, -0.0130, -0.0565], + ..., + [ 0.0155, -0.0731, 0.1208, ..., 0.0177, -0.0626, 0.0363], + [-0.0222, 0.0477, -0.0996, ..., -0.0428, 0.0040, -0.0646], + [-0.0558, 0.0636, 0.0547, ..., -0.0471, -0.0284, -0.0137]], + device='cuda:0'), grad: tensor([[ 3.1734e-04, -3.9902e-03, -8.8453e-04, ..., -4.5395e-03, + 4.9067e-04, 7.2861e-04], + [ 2.7344e-05, -2.0004e-02, -2.2605e-05, ..., -3.1464e-02, + -1.0471e-03, 3.2485e-05], + [-4.3640e-03, -1.2032e-02, 1.6203e-03, ..., 7.3204e-03, + -1.4351e-02, -8.3923e-04], + ..., + [-3.7384e-03, 7.4043e-03, -4.5662e-03, ..., 3.8280e-03, + 5.5885e-03, -1.5726e-03], + [ 1.8418e-04, 1.1833e-02, 1.4315e-03, ..., 1.4130e-02, + 6.5880e-03, 8.1110e-04], + [ 5.0888e-03, 6.2218e-03, 5.8937e-03, ..., 2.6226e-03, + 7.1793e-03, 5.8670e-03]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0194, 0.0238, 0.0042, -0.0090, 0.0193, -0.0301, 0.0102, -0.0049, + 0.0008, -0.0117], device='cuda:0'), grad: tensor([-0.0088, -0.0664, 0.0169, 0.0451, -0.0775, 0.0250, -0.0115, 0.0081, + 0.0356, 0.0334], device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 214.52, cls_loss 0.6659 cls_loss_mapping 0.0102 cls_loss_causal 0.5677 re_mapping 0.0141 re_causal 0.0318 /// teacc 98.29 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0596, -0.0790, -0.0738, ..., -0.0273, 0.0447, -0.0790], + [-0.0462, -0.0695, -0.0469, ..., 0.0890, -0.0268, 0.1652], + [ 0.0090, -0.0227, -0.0190, ..., -0.0112, -0.0123, -0.0560], + ..., + [ 0.0160, -0.0738, 0.1214, ..., 0.0173, -0.0646, 0.0365], + [-0.0220, 0.0478, -0.0999, ..., -0.0428, 0.0028, -0.0650], + [-0.0553, 0.0643, 0.0547, ..., -0.0468, -0.0277, -0.0144]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0003, 0.0001, ..., 0.0024, 0.0004, 0.0009], + [-0.0030, -0.0362, 0.0002, ..., -0.0144, -0.0033, -0.0165], + [ 0.0002, 0.0009, 0.0004, ..., 0.0020, 0.0003, -0.0092], + ..., + [ 0.0005, 0.0025, 0.0005, ..., 0.0071, 0.0012, 0.0031], + [ 0.0003, -0.0032, -0.0018, ..., -0.0071, -0.0005, 0.0006], + [ 0.0008, 0.0300, 0.0039, ..., 0.0136, 0.0044, 0.0183]], + device='cuda:0') +Epoch 135, bias, value: tensor([-0.0200, 0.0244, 0.0034, -0.0086, 0.0203, -0.0307, 0.0107, -0.0055, + 0.0006, -0.0113], device='cuda:0'), grad: tensor([ 0.0040, -0.0779, -0.0180, -0.0176, -0.0074, 0.0182, 0.0170, 0.0224, + -0.0059, 0.0652], device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 214.57, cls_loss 0.6480 cls_loss_mapping 0.0119 cls_loss_causal 0.5529 re_mapping 0.0144 re_causal 0.0313 /// teacc 98.39 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0599, -0.0811, -0.0745, ..., -0.0275, 0.0440, -0.0788], + [-0.0467, -0.0690, -0.0480, ..., 0.0889, -0.0278, 0.1654], + [ 0.0086, -0.0213, -0.0192, ..., -0.0106, -0.0126, -0.0554], + ..., + [ 0.0165, -0.0731, 0.1224, ..., 0.0169, -0.0629, 0.0373], + [-0.0235, 0.0481, -0.1006, ..., -0.0436, 0.0033, -0.0647], + [-0.0547, 0.0637, 0.0546, ..., -0.0459, -0.0280, -0.0150]], + device='cuda:0'), grad: tensor([[ 6.2622e-06, 1.1911e-03, 2.3198e-04, ..., 5.0774e-03, + 1.0502e-04, 3.6430e-04], + [-6.9618e-05, -2.7542e-03, 2.9612e-04, ..., 3.0708e-03, + 6.6459e-05, 3.2082e-03], + [ 3.1590e-05, 3.4542e-03, 8.0013e-04, ..., 4.0016e-03, + 3.2473e-04, 8.4639e-04], + ..., + [ 4.8339e-05, -3.3073e-03, -1.1339e-03, ..., -8.0872e-04, + 7.3373e-05, -2.5558e-03], + [ 2.0576e-04, -2.7409e-03, -1.0292e-02, ..., 7.0038e-03, + 1.3876e-04, 4.5657e-04], + [ 1.2815e-04, 6.3591e-03, 1.1391e-02, ..., 3.9825e-03, + 1.3709e-04, 1.0529e-03]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0200, 0.0242, 0.0039, -0.0086, 0.0197, -0.0302, 0.0102, -0.0055, + 0.0002, -0.0107], device='cuda:0'), grad: tensor([ 0.0208, 0.0370, 0.0205, -0.0831, 0.0219, -0.0465, 0.0006, -0.0213, + -0.0015, 0.0517], device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 214.34, cls_loss 0.6488 cls_loss_mapping 0.0135 cls_loss_causal 0.5572 re_mapping 0.0149 re_causal 0.0326 /// teacc 98.46 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0606, -0.0804, -0.0753, ..., -0.0279, 0.0440, -0.0783], + [-0.0491, -0.0694, -0.0492, ..., 0.0896, -0.0281, 0.1660], + [ 0.0080, -0.0205, -0.0200, ..., -0.0100, -0.0129, -0.0550], + ..., + [ 0.0188, -0.0737, 0.1238, ..., 0.0185, -0.0629, 0.0363], + [-0.0230, 0.0481, -0.0993, ..., -0.0446, 0.0032, -0.0652], + [-0.0554, 0.0643, 0.0535, ..., -0.0458, -0.0279, -0.0134]], + device='cuda:0'), grad: tensor([[ 6.9141e-05, 1.7190e-04, -4.9829e-05, ..., -6.7368e-03, + 1.7598e-05, -5.2166e-04], + [ 6.2704e-05, 2.5730e-03, 2.6588e-03, ..., 2.4204e-03, + 6.4433e-05, 1.5574e-03], + [ 6.0368e-04, 8.9598e-04, 1.0920e-03, ..., 2.5177e-03, + -1.3602e-04, 7.6771e-04], + ..., + [ 1.4935e-03, -2.9602e-03, -3.3512e-03, ..., -7.2594e-03, + 4.4751e-04, -1.8396e-03], + [ 9.0790e-04, 3.3455e-03, 2.4319e-03, ..., 9.6512e-03, + 5.1826e-05, 1.4610e-03], + [ 2.7180e-04, -4.1313e-03, 2.6155e-04, ..., -2.8443e-04, + 1.5461e-04, -1.0548e-03]], device='cuda:0') +Epoch 137, bias, value: tensor([-2.0258e-02, 2.4275e-02, 3.7209e-03, -7.8909e-03, 1.9090e-02, + -3.0254e-02, 9.7566e-03, -4.5812e-03, -9.1325e-05, -1.0589e-02], + device='cuda:0'), grad: tensor([-0.0208, -0.0051, 0.0082, -0.0413, 0.0138, 0.0172, 0.0098, -0.0008, + 0.0258, -0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 214.63, cls_loss 0.6265 cls_loss_mapping 0.0141 cls_loss_causal 0.5398 re_mapping 0.0150 re_causal 0.0331 /// teacc 98.28 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0585, -0.0816, -0.0751, ..., -0.0273, 0.0439, -0.0782], + [-0.0495, -0.0696, -0.0500, ..., 0.0905, -0.0282, 0.1665], + [ 0.0079, -0.0210, -0.0215, ..., -0.0093, -0.0135, -0.0559], + ..., + [ 0.0197, -0.0757, 0.1244, ..., 0.0179, -0.0627, 0.0367], + [-0.0231, 0.0480, -0.1004, ..., -0.0456, 0.0025, -0.0648], + [-0.0560, 0.0647, 0.0535, ..., -0.0460, -0.0280, -0.0133]], + device='cuda:0'), grad: tensor([[ 1.8859e-04, -7.5798e-03, 8.4352e-04, ..., -1.1322e-02, + 4.6277e-04, 2.5797e-04], + [-1.4931e-02, -1.8053e-03, -2.1637e-02, ..., -1.8311e-02, + 7.4863e-05, -1.6586e-02], + [ 1.3371e-03, 6.2847e-04, 2.6207e-03, ..., -9.3613e-03, + 4.0936e-04, 1.1806e-03], + ..., + [ 6.1607e-03, 1.1950e-03, 8.7051e-03, ..., 7.6904e-03, + 5.9634e-05, 7.7744e-03], + [ 4.9400e-04, -4.6635e-04, -2.0027e-03, ..., -3.0651e-03, + -2.1038e-03, -6.6376e-04], + [ 6.1989e-04, 2.8801e-03, 2.4676e-04, ..., 9.0408e-03, + 3.1495e-04, 1.5764e-03]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0195, 0.0248, 0.0040, -0.0079, 0.0194, -0.0301, 0.0094, -0.0052, + -0.0014, -0.0101], device='cuda:0'), grad: tensor([-0.0219, -0.0665, -0.0270, 0.0667, -0.0090, -0.0097, 0.0172, 0.0374, + -0.0127, 0.0256], device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 215.02, cls_loss 0.6287 cls_loss_mapping 0.0105 cls_loss_causal 0.5338 re_mapping 0.0144 re_causal 0.0314 /// teacc 98.32 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0592, -0.0824, -0.0758, ..., -0.0285, 0.0457, -0.0795], + [-0.0494, -0.0701, -0.0503, ..., 0.0904, -0.0282, 0.1663], + [ 0.0088, -0.0214, -0.0223, ..., -0.0092, -0.0134, -0.0560], + ..., + [ 0.0188, -0.0745, 0.1242, ..., 0.0185, -0.0637, 0.0378], + [-0.0234, 0.0477, -0.0999, ..., -0.0452, 0.0025, -0.0641], + [-0.0552, 0.0647, 0.0529, ..., -0.0470, -0.0273, -0.0140]], + device='cuda:0'), grad: tensor([[ 3.0661e-04, 3.2520e-04, 2.0742e-04, ..., -2.6608e-03, + -2.9945e-03, 1.2712e-03], + [ 2.7728e-04, -1.2074e-03, 7.5293e-04, ..., -1.5182e-02, + -1.2141e-04, -3.7251e-03], + [ 2.1152e-03, 6.0844e-04, 1.1902e-03, ..., 1.1986e-02, + 1.1911e-03, 2.8954e-03], + ..., + [ 9.9754e-04, 3.4142e-04, -2.8763e-03, ..., -4.0436e-03, + 6.5756e-04, 2.3117e-03], + [ 1.0462e-03, 1.0595e-03, 4.0501e-05, ..., 4.6577e-03, + 1.5163e-03, 1.0246e-04], + [ 8.6784e-04, -1.7099e-03, 6.4039e-04, ..., -2.0676e-03, + 8.9455e-04, -2.8801e-04]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0206, 0.0247, 0.0040, -0.0088, 0.0197, -0.0298, 0.0100, -0.0051, + -0.0004, -0.0107], device='cuda:0'), grad: tensor([-0.0139, -0.0360, 0.0358, 0.0254, -0.0299, 0.0327, 0.0251, -0.0224, + 0.0171, -0.0341], device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 214.63, cls_loss 0.6173 cls_loss_mapping 0.0103 cls_loss_causal 0.5352 re_mapping 0.0143 re_causal 0.0321 /// teacc 98.24 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0591, -0.0825, -0.0746, ..., -0.0274, 0.0455, -0.0801], + [-0.0490, -0.0706, -0.0483, ..., 0.0905, -0.0268, 0.1668], + [ 0.0104, -0.0210, -0.0228, ..., -0.0098, -0.0137, -0.0552], + ..., + [ 0.0178, -0.0756, 0.1252, ..., 0.0178, -0.0635, 0.0372], + [-0.0236, 0.0479, -0.1019, ..., -0.0451, 0.0029, -0.0648], + [-0.0553, 0.0646, 0.0524, ..., -0.0470, -0.0278, -0.0140]], + device='cuda:0'), grad: tensor([[ 6.5923e-05, 4.2486e-04, 7.5722e-04, ..., -2.8458e-03, + 5.5838e-04, 7.7391e-04], + [ 1.9193e-04, 4.5133e-04, 3.3970e-03, ..., 7.4654e-03, + 4.8423e-04, 5.5275e-03], + [ 3.0637e-04, 3.4313e-03, -5.1575e-03, ..., 7.2336e-04, + 3.0289e-03, -1.2009e-02], + ..., + [ 1.9181e-04, 7.3004e-04, 2.1095e-03, ..., 3.4046e-03, + -6.1760e-03, 4.4250e-03], + [-1.4668e-03, -8.0643e-03, -1.1597e-02, ..., -1.6144e-02, + -5.9128e-03, -5.6534e-03], + [ 3.2544e-04, 7.8869e-04, 2.1343e-03, ..., 8.6670e-03, + 5.8556e-03, 2.8629e-03]], device='cuda:0') +Epoch 140, bias, value: tensor([-2.0416e-02, 2.4332e-02, 4.1285e-03, -8.3246e-03, 1.9837e-02, + -3.0711e-02, 1.0390e-02, -5.5880e-03, 2.9677e-05, -1.0502e-02], + device='cuda:0'), grad: tensor([-0.0166, 0.0269, 0.0138, 0.0399, -0.0131, -0.0105, -0.0101, 0.0121, + -0.0728, 0.0304], device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 214.61, cls_loss 0.6596 cls_loss_mapping 0.0104 cls_loss_causal 0.5693 re_mapping 0.0138 re_causal 0.0321 /// teacc 98.53 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0594, -0.0834, -0.0729, ..., -0.0269, 0.0453, -0.0785], + [-0.0498, -0.0694, -0.0486, ..., 0.0911, -0.0273, 0.1676], + [ 0.0099, -0.0208, -0.0234, ..., -0.0102, -0.0135, -0.0549], + ..., + [ 0.0182, -0.0756, 0.1256, ..., 0.0180, -0.0636, 0.0370], + [-0.0237, 0.0465, -0.1025, ..., -0.0441, 0.0030, -0.0655], + [-0.0554, 0.0651, 0.0519, ..., -0.0468, -0.0264, -0.0144]], + device='cuda:0'), grad: tensor([[ 1.0710e-07, 4.9686e-04, 8.5545e-04, ..., 7.8964e-03, + -8.2445e-04, 2.2507e-03], + [ 6.6077e-07, 9.5177e-04, 9.3317e-04, ..., 3.2120e-03, + 2.3544e-04, -7.8821e-04], + [ 5.6103e-06, 1.1005e-03, 1.6193e-03, ..., 8.7662e-03, + 7.5531e-04, 1.6327e-03], + ..., + [ 5.4985e-06, 1.7118e-03, 2.3403e-03, ..., 9.3002e-03, + 1.3149e-04, 1.6508e-03], + [-5.9605e-05, -6.9389e-03, -5.8250e-03, ..., -3.2898e-02, + -4.9133e-03, -9.5444e-03], + [ 4.0889e-05, 5.1384e-03, 2.3174e-03, ..., 4.9800e-05, + 1.3781e-03, 3.2883e-03]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0204, 0.0247, 0.0043, -0.0088, 0.0195, -0.0316, 0.0106, -0.0054, + 0.0004, -0.0102], device='cuda:0'), grad: tensor([ 0.0182, 0.0093, 0.0216, 0.0196, 0.0052, 0.0121, -0.0216, 0.0228, + -0.0911, 0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 214.80, cls_loss 0.6338 cls_loss_mapping 0.0117 cls_loss_causal 0.5492 re_mapping 0.0139 re_causal 0.0308 /// teacc 98.39 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0595, -0.0837, -0.0718, ..., -0.0269, 0.0454, -0.0778], + [-0.0504, -0.0700, -0.0493, ..., 0.0906, -0.0281, 0.1674], + [ 0.0102, -0.0191, -0.0232, ..., -0.0109, -0.0124, -0.0561], + ..., + [ 0.0185, -0.0739, 0.1255, ..., 0.0193, -0.0655, 0.0378], + [-0.0238, 0.0468, -0.1037, ..., -0.0431, 0.0038, -0.0651], + [-0.0559, 0.0649, 0.0518, ..., -0.0480, -0.0264, -0.0162]], + device='cuda:0'), grad: tensor([[ 0.0001, 0.0026, 0.0005, ..., 0.0157, 0.0075, 0.0003], + [ 0.0002, 0.0004, 0.0007, ..., 0.0108, 0.0013, 0.0003], + [ 0.0002, 0.0017, 0.0013, ..., 0.0101, 0.0179, 0.0007], + ..., + [ 0.0005, 0.0034, 0.0035, ..., -0.0111, 0.0009, 0.0021], + [ 0.0002, 0.0052, 0.0032, ..., 0.0093, -0.0647, 0.0018], + [-0.0003, -0.0135, -0.0107, ..., -0.0057, -0.0062, -0.0016]], + device='cuda:0') +Epoch 142, bias, value: tensor([-0.0206, 0.0242, 0.0040, -0.0082, 0.0199, -0.0313, 0.0101, -0.0047, + 0.0009, -0.0113], device='cuda:0'), grad: tensor([ 0.0510, 0.0247, 0.0488, 0.0113, -0.0088, -0.0009, -0.0357, -0.0319, + -0.0197, -0.0386], device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 214.50, cls_loss 0.6261 cls_loss_mapping 0.0120 cls_loss_causal 0.5338 re_mapping 0.0133 re_causal 0.0288 /// teacc 98.47 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0595, -0.0837, -0.0724, ..., -0.0269, 0.0458, -0.0782], + [-0.0497, -0.0689, -0.0490, ..., 0.0910, -0.0284, 0.1688], + [ 0.0112, -0.0183, -0.0229, ..., -0.0106, -0.0114, -0.0567], + ..., + [ 0.0188, -0.0737, 0.1263, ..., 0.0196, -0.0651, 0.0376], + [-0.0247, 0.0460, -0.1054, ..., -0.0442, 0.0035, -0.0669], + [-0.0566, 0.0646, 0.0513, ..., -0.0475, -0.0267, -0.0160]], + device='cuda:0'), grad: tensor([[ 8.4788e-06, 2.2566e-04, 2.0397e-04, ..., 3.0575e-03, + -3.1084e-05, 1.6499e-03], + [ 7.4692e-06, 4.6039e-04, 4.1151e-04, ..., 3.0823e-03, + 5.1670e-06, 2.0866e-03], + [ 6.6161e-05, 6.0177e-04, 5.4264e-04, ..., -8.3313e-03, + 5.8174e-05, -3.1204e-03], + ..., + [ 8.0317e-06, 2.0256e-03, -5.3358e-04, ..., 1.9331e-03, + 1.9997e-05, 4.3564e-03], + [ 4.8757e-05, 1.2369e-03, 7.0000e-04, ..., -6.2752e-03, + 3.0851e-04, -2.0294e-03], + [ 9.7677e-06, -4.4556e-03, -3.1261e-03, ..., -9.2411e-04, + 3.8028e-05, -6.8741e-03]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0197, 0.0247, 0.0037, -0.0083, 0.0195, -0.0314, 0.0102, -0.0045, + 0.0001, -0.0113], device='cuda:0'), grad: tensor([ 0.0126, 0.0138, -0.0230, 0.0024, 0.0327, 0.0007, 0.0115, -0.0050, + -0.0458, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 142---------------------------------------------------- +epoch 142, time 231.13, cls_loss 0.6230 cls_loss_mapping 0.0147 cls_loss_causal 0.5340 re_mapping 0.0145 re_causal 0.0318 /// teacc 98.58 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0602, -0.0847, -0.0736, ..., -0.0279, 0.0450, -0.0806], + [-0.0511, -0.0687, -0.0487, ..., 0.0917, -0.0290, 0.1696], + [ 0.0135, -0.0184, -0.0234, ..., -0.0105, -0.0113, -0.0562], + ..., + [ 0.0179, -0.0739, 0.1258, ..., 0.0195, -0.0658, 0.0362], + [-0.0251, 0.0464, -0.1053, ..., -0.0452, 0.0037, -0.0671], + [-0.0559, 0.0645, 0.0520, ..., -0.0468, -0.0268, -0.0141]], + device='cuda:0'), grad: tensor([[-2.6092e-05, -2.7122e-03, 4.4078e-05, ..., 6.2943e-03, + 1.9968e-06, 3.8886e-04], + [ 3.0696e-05, -2.0373e-04, -6.4325e-04, ..., 4.5395e-03, + 9.0804e-08, -3.0670e-03], + [ 4.9695e-06, 1.2989e-03, 1.0270e-04, ..., 5.7411e-03, + 1.9651e-07, 4.7660e-04], + ..., + [-2.2030e-04, 4.2915e-04, 2.9507e-03, ..., 3.4733e-03, + 3.0734e-08, 4.0321e-03], + [ 2.4468e-05, -5.9357e-03, 6.5804e-04, ..., -5.6229e-03, + 1.8507e-05, 5.5046e-03], + [ 1.8549e-04, 5.1403e-04, -3.0174e-03, ..., 5.0011e-03, + 5.8161e-07, -1.8120e-03]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0209, 0.0249, 0.0045, -0.0079, 0.0196, -0.0315, 0.0105, -0.0048, + -0.0003, -0.0110], device='cuda:0'), grad: tensor([-0.0026, 0.0009, 0.0161, -0.0133, -0.0186, -0.0132, -0.0027, 0.0206, + 0.0101, 0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 214.47, cls_loss 0.6241 cls_loss_mapping 0.0122 cls_loss_causal 0.5456 re_mapping 0.0135 re_causal 0.0303 /// teacc 98.51 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0598, -0.0848, -0.0740, ..., -0.0275, 0.0451, -0.0808], + [-0.0504, -0.0681, -0.0490, ..., 0.0921, -0.0289, 0.1701], + [ 0.0128, -0.0186, -0.0241, ..., -0.0108, -0.0112, -0.0555], + ..., + [ 0.0161, -0.0746, 0.1261, ..., 0.0200, -0.0653, 0.0362], + [-0.0250, 0.0462, -0.1057, ..., -0.0452, 0.0035, -0.0670], + [-0.0528, 0.0656, 0.0535, ..., -0.0462, -0.0271, -0.0143]], + device='cuda:0'), grad: tensor([[ 1.8406e-04, 7.1430e-04, -1.3552e-03, ..., 1.2636e-03, + 8.7967e-03, 1.1253e-03], + [ 5.5599e-04, 1.2894e-03, 7.6532e-05, ..., 3.4103e-03, + 1.6189e-04, -2.6531e-03], + [ 9.0599e-04, 1.1330e-03, 3.1543e-04, ..., 2.0676e-02, + 1.8444e-03, 1.2253e-02], + ..., + [ 3.1257e-04, -5.7449e-03, -3.1814e-06, ..., -3.3447e-02, + 3.3331e-04, -9.8495e-03], + [-8.9884e-04, 2.6493e-03, 5.9634e-05, ..., 8.6441e-03, + 5.1117e-04, 6.3598e-05], + [ 2.5415e-04, -1.1377e-03, 3.6645e-04, ..., 3.0880e-03, + 8.0204e-04, -8.6451e-04]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0208, 0.0251, 0.0039, -0.0087, 0.0198, -0.0319, 0.0109, -0.0044, + -0.0005, -0.0104], device='cuda:0'), grad: tensor([ 0.0163, 0.0047, 0.0134, -0.0067, -0.0034, -0.0365, 0.0147, -0.0269, + 0.0126, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 214.33, cls_loss 0.6626 cls_loss_mapping 0.0088 cls_loss_causal 0.5725 re_mapping 0.0134 re_causal 0.0307 /// teacc 98.51 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0588, -0.0857, -0.0751, ..., -0.0262, 0.0454, -0.0813], + [-0.0493, -0.0694, -0.0492, ..., 0.0916, -0.0285, 0.1705], + [ 0.0107, -0.0193, -0.0250, ..., -0.0122, -0.0105, -0.0567], + ..., + [ 0.0172, -0.0748, 0.1274, ..., 0.0200, -0.0669, 0.0364], + [-0.0253, 0.0459, -0.1057, ..., -0.0452, 0.0030, -0.0671], + [-0.0534, 0.0662, 0.0527, ..., -0.0454, -0.0274, -0.0145]], + device='cuda:0'), grad: tensor([[ 1.2204e-05, -3.2043e-03, -3.6449e-03, ..., -1.3756e-02, + 1.8686e-05, -5.7526e-03], + [ 1.9610e-05, 1.3673e-04, 1.4648e-03, ..., -2.8820e-03, + 5.5254e-05, 9.9087e-04], + [ 5.9269e-06, -7.2575e-04, -4.0507e-04, ..., -9.8953e-03, + -3.9911e-04, 3.6287e-04], + ..., + [-4.2038e-03, 1.5545e-03, 5.5466e-03, ..., 2.3804e-02, + 7.6234e-05, 2.4246e-02], + [ 3.8356e-05, 7.1335e-03, 4.0817e-03, ..., 1.1948e-02, + 1.5306e-04, 2.6474e-03], + [ 3.9597e-03, -1.3046e-03, 3.5648e-03, ..., -2.0905e-03, + 6.6049e-06, 2.6302e-03]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0201, 0.0243, 0.0032, -0.0078, 0.0191, -0.0311, 0.0107, -0.0045, + -0.0005, -0.0103], device='cuda:0'), grad: tensor([-0.0502, -0.0053, -0.0178, -0.0490, 0.0018, 0.0057, -0.0023, 0.0880, + 0.0350, -0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 214.43, cls_loss 0.6596 cls_loss_mapping 0.0113 cls_loss_causal 0.5645 re_mapping 0.0139 re_causal 0.0297 /// teacc 98.39 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0598, -0.0866, -0.0751, ..., -0.0264, 0.0452, -0.0811], + [-0.0486, -0.0698, -0.0478, ..., 0.0929, -0.0278, 0.1722], + [ 0.0108, -0.0182, -0.0258, ..., -0.0114, -0.0103, -0.0571], + ..., + [ 0.0179, -0.0761, 0.1283, ..., 0.0190, -0.0669, 0.0352], + [-0.0244, 0.0460, -0.1064, ..., -0.0452, 0.0026, -0.0663], + [-0.0551, 0.0649, 0.0517, ..., -0.0447, -0.0275, -0.0153]], + device='cuda:0'), grad: tensor([[ 8.7976e-05, 1.1292e-02, 9.5081e-04, ..., -5.1727e-03, + 3.3234e-02, -4.3654e-04], + [ 2.6441e-04, 1.6298e-03, 2.2564e-03, ..., 1.5182e-02, + 5.6982e-04, 2.1347e-02], + [ 4.2772e-04, 8.2779e-04, 1.7033e-03, ..., 5.5275e-03, + 2.0218e-03, 2.9926e-03], + ..., + [-3.3607e-03, -4.6959e-03, -2.1713e-02, ..., -3.9215e-02, + 8.2791e-05, -1.4099e-02], + [-9.1374e-05, 2.0943e-03, 1.6251e-03, ..., -7.4425e-03, + -5.0259e-04, -2.5208e-02], + [ 1.0948e-03, -3.7670e-03, 7.1602e-03, ..., 1.4008e-02, + 3.0971e-04, 6.5498e-03]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0203, 0.0247, 0.0039, -0.0082, 0.0186, -0.0302, 0.0103, -0.0051, + -0.0002, -0.0103], device='cuda:0'), grad: tensor([ 0.0150, 0.0502, 0.0196, 0.0342, 0.0406, -0.0718, 0.0057, -0.0895, + -0.0301, 0.0261], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 146---------------------------------------------------- +epoch 146, time 230.51, cls_loss 0.6111 cls_loss_mapping 0.0105 cls_loss_causal 0.5259 re_mapping 0.0140 re_causal 0.0301 /// teacc 98.66 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0611, -0.0873, -0.0763, ..., -0.0269, 0.0457, -0.0812], + [-0.0492, -0.0705, -0.0468, ..., 0.0937, -0.0280, 0.1727], + [ 0.0107, -0.0171, -0.0245, ..., -0.0115, -0.0103, -0.0571], + ..., + [ 0.0182, -0.0770, 0.1289, ..., 0.0197, -0.0673, 0.0351], + [-0.0236, 0.0458, -0.1066, ..., -0.0442, 0.0038, -0.0655], + [-0.0550, 0.0655, 0.0495, ..., -0.0450, -0.0276, -0.0157]], + device='cuda:0'), grad: tensor([[ 8.0705e-05, 4.7994e-04, 2.9135e-04, ..., 5.8403e-03, + 1.8368e-03, 1.0872e-03], + [ 1.7238e-04, 9.8705e-04, -2.9011e-03, ..., -4.4327e-03, + 1.3580e-03, -2.8877e-03], + [-6.8045e-04, 1.3828e-03, 1.1759e-03, ..., 4.3755e-03, + 8.2111e-04, 1.2341e-03], + ..., + [-3.5286e-03, -8.0490e-03, -3.9062e-03, ..., -4.6158e-03, + 1.2159e-03, -4.9257e-04], + [ 1.8129e-03, 5.3101e-03, 3.6755e-03, ..., -6.6566e-03, + 1.9121e-03, -4.8637e-03], + [ 6.8712e-04, -4.9072e-02, -3.1013e-03, ..., 5.4054e-03, + -9.4299e-03, 8.6670e-03]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0202, 0.0253, 0.0038, -0.0078, 0.0188, -0.0318, 0.0097, -0.0049, + 0.0005, -0.0104], device='cuda:0'), grad: tensor([ 0.0150, -0.0116, 0.0132, 0.0087, -0.0143, -0.0082, 0.0149, -0.0151, + 0.0082, -0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 214.59, cls_loss 0.6284 cls_loss_mapping 0.0089 cls_loss_causal 0.5393 re_mapping 0.0141 re_causal 0.0327 /// teacc 98.48 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0620, -0.0887, -0.0789, ..., -0.0269, 0.0458, -0.0820], + [-0.0491, -0.0707, -0.0462, ..., 0.0945, -0.0287, 0.1731], + [ 0.0114, -0.0177, -0.0223, ..., -0.0111, -0.0106, -0.0567], + ..., + [ 0.0174, -0.0770, 0.1291, ..., 0.0198, -0.0672, 0.0355], + [-0.0239, 0.0465, -0.1079, ..., -0.0439, 0.0049, -0.0668], + [-0.0541, 0.0654, 0.0481, ..., -0.0456, -0.0282, -0.0164]], + device='cuda:0'), grad: tensor([[ 7.3202e-07, -1.0061e-03, -4.0665e-03, ..., -4.7188e-03, + 4.5657e-05, -2.8572e-03], + [ 7.0110e-06, 1.0090e-03, 1.0786e-03, ..., 6.8970e-03, + 1.8165e-05, 4.8280e-04], + [ 2.5094e-05, 2.3270e-03, 1.7309e-03, ..., -7.1716e-03, + 3.1877e-04, 6.4325e-04], + ..., + [ 7.9393e-04, 8.7357e-03, -2.0790e-03, ..., -3.7212e-03, + 2.7132e-04, 1.7014e-03], + [ 6.6102e-05, -1.8799e-04, 2.6245e-03, ..., -6.0806e-03, + 5.8365e-04, -1.9569e-03], + [-1.0319e-03, -1.0246e-02, 1.6642e-03, ..., 4.6310e-03, + 3.9959e-04, -2.8825e-04]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0206, 0.0254, 0.0049, -0.0078, 0.0186, -0.0323, 0.0087, -0.0043, + 0.0011, -0.0108], device='cuda:0'), grad: tensor([-1.1955e-02, 2.2125e-02, 3.9637e-05, 5.3329e-03, -4.2686e-03, + 2.5711e-02, -1.5068e-02, 2.3918e-03, -2.8900e-02, 4.6196e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 214.39, cls_loss 0.6221 cls_loss_mapping 0.0115 cls_loss_causal 0.5204 re_mapping 0.0141 re_causal 0.0302 /// teacc 98.51 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0636, -0.0899, -0.0788, ..., -0.0276, 0.0466, -0.0829], + [-0.0481, -0.0693, -0.0476, ..., 0.0947, -0.0294, 0.1726], + [ 0.0120, -0.0177, -0.0223, ..., -0.0115, -0.0109, -0.0573], + ..., + [ 0.0173, -0.0778, 0.1300, ..., 0.0217, -0.0669, 0.0364], + [-0.0260, 0.0468, -0.1091, ..., -0.0448, 0.0058, -0.0664], + [-0.0538, 0.0667, 0.0491, ..., -0.0451, -0.0286, -0.0164]], + device='cuda:0'), grad: tensor([[ 1.1003e-04, 1.7691e-04, 1.0338e-03, ..., 4.1389e-03, + 9.8133e-04, 2.0170e-04], + [-3.3302e-03, 2.7442e-04, -3.3417e-03, ..., -1.3901e-02, + 3.5733e-05, 2.0361e-04], + [ 7.4387e-04, -8.9312e-04, -2.5501e-03, ..., -7.3051e-03, + -4.4136e-03, 4.9543e-04], + ..., + [ 4.3511e-04, -8.9645e-04, -1.0462e-03, ..., -6.7673e-03, + 4.4465e-05, -1.8341e-02], + [ 3.9697e-04, -2.4292e-02, 1.1557e-04, ..., 3.9444e-03, + 1.4753e-03, 1.2941e-03], + [ 2.4283e-04, -3.3450e-04, -1.1873e-03, ..., -6.9771e-03, + 3.7432e-05, -1.5373e-03]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0210, 0.0252, 0.0045, -0.0094, 0.0192, -0.0319, 0.0092, -0.0033, + 0.0010, -0.0106], device='cuda:0'), grad: tensor([ 0.0100, -0.0409, -0.0126, 0.0282, 0.0185, 0.0207, 0.0179, -0.0195, + -0.0152, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 214.42, cls_loss 0.6181 cls_loss_mapping 0.0116 cls_loss_causal 0.5344 re_mapping 0.0140 re_causal 0.0309 /// teacc 98.65 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0639, -0.0883, -0.0800, ..., -0.0276, 0.0463, -0.0835], + [-0.0501, -0.0713, -0.0482, ..., 0.0935, -0.0301, 0.1735], + [ 0.0117, -0.0174, -0.0226, ..., -0.0115, -0.0100, -0.0568], + ..., + [ 0.0178, -0.0771, 0.1304, ..., 0.0220, -0.0672, 0.0365], + [-0.0256, 0.0464, -0.1095, ..., -0.0444, 0.0046, -0.0666], + [-0.0527, 0.0669, 0.0492, ..., -0.0459, -0.0276, -0.0163]], + device='cuda:0'), grad: tensor([[ 3.6787e-07, 4.8327e-04, -4.5970e-06, ..., -2.9488e-03, + 2.8248e-03, 9.2840e-04], + [ 3.3583e-06, 3.1662e-04, 1.8339e-03, ..., -1.2569e-03, + 5.9509e-04, 9.1324e-03], + [-7.7546e-05, 3.2759e-04, 2.3842e-03, ..., 2.7893e-02, + 8.9169e-04, 4.7565e-04], + ..., + [ 1.9204e-06, 8.6212e-04, 1.4257e-03, ..., 1.1702e-03, + 3.9215e-03, 7.8201e-04], + [ 6.6698e-05, 1.0405e-03, -1.3361e-03, ..., -3.6804e-02, + 6.4039e-04, -1.2619e-02], + [-1.4529e-06, 7.3242e-04, 3.4580e-03, ..., 5.5656e-03, + 2.8839e-03, 4.8757e-04]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0216, 0.0242, 0.0052, -0.0093, 0.0189, -0.0309, 0.0096, -0.0033, + 0.0005, -0.0105], device='cuda:0'), grad: tensor([-0.0003, 0.0181, 0.0461, -0.0093, -0.0227, 0.0244, -0.0309, 0.0064, + -0.0590, 0.0272], device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 214.44, cls_loss 0.6042 cls_loss_mapping 0.0092 cls_loss_causal 0.5135 re_mapping 0.0141 re_causal 0.0309 /// teacc 98.64 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0646, -0.0878, -0.0790, ..., -0.0266, 0.0465, -0.0849], + [-0.0507, -0.0712, -0.0484, ..., 0.0938, -0.0313, 0.1743], + [ 0.0110, -0.0176, -0.0239, ..., -0.0112, -0.0098, -0.0571], + ..., + [ 0.0177, -0.0766, 0.1321, ..., 0.0214, -0.0670, 0.0368], + [-0.0239, 0.0462, -0.1106, ..., -0.0446, 0.0042, -0.0662], + [-0.0528, 0.0667, 0.0494, ..., -0.0458, -0.0281, -0.0165]], + device='cuda:0'), grad: tensor([[-1.1017e-02, -1.3046e-02, -2.1515e-02, ..., 1.7347e-03, + 1.7095e-04, 9.5367e-05], + [-2.3201e-05, -4.5586e-03, 9.5749e-04, ..., -2.6035e-03, + 5.5790e-04, -5.2490e-03], + [ 1.2070e-04, 1.3885e-03, 2.2278e-03, ..., 4.8714e-03, + 1.4782e-03, 2.9135e-04], + ..., + [ 2.9411e-06, -3.2940e-03, -7.9422e-03, ..., -1.8356e-02, + -3.7518e-03, 2.2296e-06], + [ 1.6570e-04, 2.0275e-03, 8.7261e-04, ..., 3.6144e-03, + 3.0565e-04, 1.7023e-03], + [ 1.8671e-05, 1.0424e-03, 1.1597e-03, ..., 3.7231e-03, + 4.7636e-04, 4.5252e-04]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0203, 0.0240, 0.0049, -0.0096, 0.0192, -0.0305, 0.0098, -0.0042, + 0.0003, -0.0105], device='cuda:0'), grad: tensor([-0.0497, -0.0045, 0.0223, 0.0462, -0.0137, 0.0181, 0.0233, -0.0765, + 0.0177, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 214.38, cls_loss 0.5993 cls_loss_mapping 0.0111 cls_loss_causal 0.5268 re_mapping 0.0134 re_causal 0.0293 /// teacc 98.44 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0631, -0.0885, -0.0782, ..., -0.0269, 0.0461, -0.0850], + [-0.0504, -0.0711, -0.0482, ..., 0.0938, -0.0319, 0.1745], + [ 0.0112, -0.0175, -0.0248, ..., -0.0112, -0.0099, -0.0565], + ..., + [ 0.0183, -0.0779, 0.1291, ..., 0.0211, -0.0671, 0.0366], + [-0.0244, 0.0466, -0.1130, ..., -0.0448, 0.0051, -0.0666], + [-0.0536, 0.0660, 0.0501, ..., -0.0457, -0.0286, -0.0163]], + device='cuda:0'), grad: tensor([[-2.1992e-03, -1.7338e-03, -3.2291e-03, ..., -4.4785e-03, + -4.9286e-03, -7.3195e-04], + [ 4.3082e-04, -4.4975e-03, -6.1655e-04, ..., -5.6763e-03, + -4.8470e-04, -2.5253e-03], + [ 1.5297e-03, 5.8327e-03, 5.5313e-03, ..., 9.2316e-03, + 2.7313e-03, 1.3506e-04], + ..., + [ 6.2466e-04, 1.2827e-03, 8.4829e-04, ..., -6.8016e-03, + 5.9223e-04, -2.9907e-05], + [ 1.1425e-03, 1.0170e-02, 5.8136e-03, ..., 9.0179e-03, + 3.9787e-03, 7.4911e-04], + [ 8.4782e-04, -1.6129e-02, -1.5533e-02, ..., -8.3008e-03, + 6.5804e-04, 1.2445e-04]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0207, 0.0247, 0.0048, -0.0095, 0.0190, -0.0304, 0.0101, -0.0045, + -0.0001, -0.0104], device='cuda:0'), grad: tensor([-0.0041, -0.0244, 0.0205, -0.0129, -0.0095, 0.0654, -0.0376, -0.0310, + 0.0493, -0.0156], device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 214.38, cls_loss 0.6190 cls_loss_mapping 0.0108 cls_loss_causal 0.5289 re_mapping 0.0130 re_causal 0.0298 /// teacc 98.58 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0640, -0.0881, -0.0791, ..., -0.0273, 0.0459, -0.0855], + [-0.0489, -0.0720, -0.0497, ..., 0.0937, -0.0304, 0.1756], + [ 0.0117, -0.0179, -0.0253, ..., -0.0111, -0.0100, -0.0562], + ..., + [ 0.0181, -0.0776, 0.1298, ..., 0.0216, -0.0667, 0.0353], + [-0.0256, 0.0457, -0.1133, ..., -0.0453, 0.0062, -0.0681], + [-0.0548, 0.0663, 0.0520, ..., -0.0457, -0.0287, -0.0149]], + device='cuda:0'), grad: tensor([[ 9.6178e-04, 1.6069e-03, 1.9197e-03, ..., 4.2229e-03, + 3.1441e-05, -1.0902e-04], + [ 4.2939e-04, -7.0000e-04, -4.4250e-02, ..., -1.2718e-02, + -2.1410e-04, -1.2306e-02], + [ 1.0300e-03, 9.6655e-04, 3.7766e-03, ..., 4.6272e-03, + 3.4928e-05, 1.6851e-03], + ..., + [ 1.3676e-03, 1.6298e-03, 3.8330e-02, ..., 1.3298e-02, + 1.6183e-05, 9.6817e-03], + [-1.2226e-03, -4.1008e-05, -5.5933e-04, ..., -1.6708e-03, + 2.3112e-05, -1.1978e-03], + [-3.1614e-04, -4.1656e-03, 2.2087e-03, ..., -1.0620e-02, + 4.5955e-05, -3.4103e-03]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0212, 0.0254, 0.0046, -0.0098, 0.0198, -0.0306, 0.0105, -0.0048, + -0.0002, -0.0107], device='cuda:0'), grad: tensor([ 0.0115, -0.0601, 0.0181, -0.0156, 0.0337, 0.0079, -0.0107, 0.0635, + -0.0033, -0.0449], device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 214.22, cls_loss 0.6143 cls_loss_mapping 0.0089 cls_loss_causal 0.5258 re_mapping 0.0130 re_causal 0.0283 /// teacc 98.61 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0633, -0.0879, -0.0792, ..., -0.0276, 0.0471, -0.0846], + [-0.0496, -0.0729, -0.0495, ..., 0.0946, -0.0310, 0.1761], + [ 0.0114, -0.0182, -0.0246, ..., -0.0109, -0.0104, -0.0557], + ..., + [ 0.0170, -0.0782, 0.1294, ..., 0.0210, -0.0667, 0.0359], + [-0.0255, 0.0473, -0.1142, ..., -0.0457, 0.0059, -0.0689], + [-0.0528, 0.0662, 0.0514, ..., -0.0456, -0.0292, -0.0162]], + device='cuda:0'), grad: tensor([[ 9.3162e-05, 4.8161e-04, 4.6611e-04, ..., -1.1833e-02, + 5.9223e-04, -8.3256e-04], + [-2.9335e-03, -3.5019e-03, -7.9041e-03, ..., -1.0101e-02, + 2.6584e-05, -5.8250e-03], + [ 7.6818e-04, 2.6054e-03, 2.0599e-03, ..., 5.6343e-03, + 1.9395e-04, 5.0306e-04], + ..., + [ 7.7581e-04, 5.7697e-05, -5.9395e-03, ..., 9.5272e-04, + -2.5201e-04, 1.6718e-03], + [ 5.0974e-04, 1.4925e-03, 2.8439e-03, ..., -5.4893e-03, + 5.2643e-04, 9.0599e-04], + [ 1.2026e-03, 3.7994e-02, 1.1072e-03, ..., 1.8978e-03, + 1.0538e-04, 1.1387e-03]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0205, 0.0259, 0.0047, -0.0099, 0.0206, -0.0300, 0.0088, -0.0055, + -0.0007, -0.0104], device='cuda:0'), grad: tensor([-0.0369, -0.0377, 0.0279, 0.0250, -0.0301, 0.0221, 0.0327, -0.0116, + -0.0347, 0.0433], device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 214.21, cls_loss 0.6219 cls_loss_mapping 0.0070 cls_loss_causal 0.5372 re_mapping 0.0130 re_causal 0.0309 /// teacc 98.56 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0628, -0.0884, -0.0780, ..., -0.0277, 0.0475, -0.0833], + [-0.0494, -0.0734, -0.0503, ..., 0.0952, -0.0313, 0.1763], + [ 0.0104, -0.0179, -0.0251, ..., -0.0104, -0.0102, -0.0558], + ..., + [ 0.0173, -0.0792, 0.1297, ..., 0.0206, -0.0663, 0.0357], + [-0.0253, 0.0466, -0.1147, ..., -0.0462, 0.0060, -0.0679], + [-0.0525, 0.0663, 0.0527, ..., -0.0443, -0.0305, -0.0158]], + device='cuda:0'), grad: tensor([[ 2.6643e-05, 7.8011e-04, 6.5184e-04, ..., 6.4201e-03, + -1.1238e-02, 5.6305e-03], + [ 1.1235e-05, -4.9362e-03, 1.1215e-03, ..., -1.4862e-02, + -1.5144e-03, -5.2547e-04], + [ 5.8621e-05, 4.3793e-03, 1.1406e-03, ..., 7.9193e-03, + 2.7523e-03, 2.3746e-03], + ..., + [ 8.9049e-05, 6.2180e-04, 1.6317e-03, ..., 7.8506e-03, + 2.8181e-04, -1.4191e-02], + [ 4.2748e-04, 2.4624e-03, -4.7340e-03, ..., -6.1607e-03, + 4.6959e-03, 1.5936e-03], + [-1.9703e-03, -1.2732e-03, -3.5839e-03, ..., 4.9553e-03, + 1.7900e-03, 1.4839e-03]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0208, 0.0270, 0.0046, -0.0101, 0.0190, -0.0302, 0.0092, -0.0055, + -0.0009, -0.0093], device='cuda:0'), grad: tensor([ 0.0171, -0.0280, 0.0342, -0.0280, -0.0060, 0.0141, -0.0020, -0.0106, + -0.0030, 0.0124], device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 214.48, cls_loss 0.6112 cls_loss_mapping 0.0106 cls_loss_causal 0.5283 re_mapping 0.0127 re_causal 0.0289 /// teacc 98.59 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0636, -0.0891, -0.0773, ..., -0.0278, 0.0488, -0.0836], + [-0.0501, -0.0743, -0.0499, ..., 0.0946, -0.0313, 0.1767], + [ 0.0107, -0.0165, -0.0244, ..., -0.0104, -0.0089, -0.0560], + ..., + [ 0.0185, -0.0795, 0.1299, ..., 0.0201, -0.0666, 0.0370], + [-0.0251, 0.0468, -0.1151, ..., -0.0460, 0.0061, -0.0688], + [-0.0526, 0.0654, 0.0519, ..., -0.0455, -0.0311, -0.0166]], + device='cuda:0'), grad: tensor([[ 6.4895e-06, 7.6294e-04, 7.2813e-04, ..., 1.3672e-02, + 1.2026e-03, 1.5676e-04], + [ 1.3813e-05, 3.5744e-03, 5.0879e-04, ..., 3.3112e-02, + 3.7169e-04, 1.1200e-04], + [ 8.4829e-04, 7.3957e-04, 7.5226e-03, ..., -6.6681e-03, + 5.1498e-03, 1.5821e-03], + ..., + [-9.3365e-04, 2.1133e-03, 6.6185e-03, ..., -3.9902e-03, + 1.1322e-02, -1.5182e-03], + [ 2.3112e-05, -1.3702e-02, 2.1362e-03, ..., -2.3575e-02, + 5.8479e-03, 2.1350e-04], + [-3.3587e-05, 1.8139e-03, 7.6027e-03, ..., -3.9825e-03, + 2.8801e-03, -7.6008e-04]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0208, 0.0265, 0.0055, -0.0110, 0.0191, -0.0300, 0.0100, -0.0056, + -0.0003, -0.0102], device='cuda:0'), grad: tensor([ 0.0307, 0.0305, -0.0034, -0.0126, -0.0163, 0.0244, -0.0342, 0.0010, + -0.0265, 0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 214.29, cls_loss 0.6337 cls_loss_mapping 0.0116 cls_loss_causal 0.5443 re_mapping 0.0129 re_causal 0.0281 /// teacc 98.39 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0644, -0.0909, -0.0776, ..., -0.0283, 0.0486, -0.0831], + [-0.0503, -0.0736, -0.0499, ..., 0.0947, -0.0314, 0.1776], + [ 0.0109, -0.0172, -0.0238, ..., -0.0103, -0.0090, -0.0566], + ..., + [ 0.0193, -0.0805, 0.1301, ..., 0.0194, -0.0667, 0.0368], + [-0.0251, 0.0485, -0.1167, ..., -0.0453, 0.0057, -0.0693], + [-0.0536, 0.0645, 0.0518, ..., -0.0465, -0.0313, -0.0163]], + device='cuda:0'), grad: tensor([[ 1.1735e-06, 1.3285e-03, 4.6825e-04, ..., 6.6147e-03, + -2.7313e-03, 2.5558e-03], + [ 2.6375e-06, 3.2687e-04, 6.2883e-05, ..., -3.1891e-02, + 3.6168e-04, -1.4442e-02], + [ 2.2739e-05, 2.7895e-04, -7.0906e-04, ..., 5.6152e-03, + 7.9060e-04, 2.0008e-03], + ..., + [ 3.7402e-06, 4.6492e-04, -1.9968e-05, ..., 8.4000e-03, + 2.7132e-04, 1.2960e-03], + [-7.0520e-06, -2.5139e-03, 1.6565e-03, ..., 1.0300e-02, + 8.8263e-04, 1.6689e-03], + [ 2.9981e-05, -5.9052e-03, 7.7772e-04, ..., 5.4016e-03, + 8.9073e-04, -1.2367e-02]], device='cuda:0') +Epoch 158, bias, value: tensor([-2.0801e-02, 2.6678e-02, 5.1584e-03, -1.0649e-02, 1.9775e-02, + -2.9717e-02, 9.4442e-03, -6.0179e-03, 5.5559e-05, -1.0768e-02], + device='cuda:0'), grad: tensor([ 0.0139, -0.0572, 0.0056, -0.0779, 0.0005, 0.0248, 0.0351, 0.0262, + 0.0268, 0.0020], device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 214.54, cls_loss 0.5840 cls_loss_mapping 0.0110 cls_loss_causal 0.4949 re_mapping 0.0128 re_causal 0.0282 /// teacc 98.53 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0661, -0.0914, -0.0774, ..., -0.0278, 0.0497, -0.0842], + [-0.0508, -0.0742, -0.0501, ..., 0.0958, -0.0321, 0.1781], + [ 0.0123, -0.0174, -0.0237, ..., -0.0112, -0.0093, -0.0573], + ..., + [ 0.0189, -0.0796, 0.1303, ..., 0.0198, -0.0654, 0.0370], + [-0.0264, 0.0471, -0.1184, ..., -0.0469, 0.0057, -0.0702], + [-0.0529, 0.0654, 0.0515, ..., -0.0464, -0.0320, -0.0155]], + device='cuda:0'), grad: tensor([[ 1.0710e-03, 1.7633e-03, 7.7534e-04, ..., -1.1734e-02, + 3.2978e-03, 3.8052e-04], + [ 1.7071e-03, 6.6221e-05, 2.8400e-03, ..., 4.7150e-03, + 4.7073e-03, -1.1282e-03], + [ 1.2169e-03, 3.9940e-03, 4.4899e-03, ..., 2.0828e-02, + 2.8400e-03, 7.9155e-04], + ..., + [ 8.0252e-04, -2.2171e-02, 1.0033e-03, ..., 9.4795e-04, + 1.2236e-03, -4.9934e-03], + [ 1.8702e-03, -2.1439e-02, -1.4229e-02, ..., -1.0498e-02, + -1.4458e-02, -4.8351e-04], + [ 1.8396e-03, 1.3145e-02, 8.3351e-04, ..., 7.3433e-03, + 1.0328e-03, 2.8915e-03]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0205, 0.0270, 0.0045, -0.0107, 0.0203, -0.0294, 0.0092, -0.0056, + -0.0005, -0.0112], device='cuda:0'), grad: tensor([-0.0116, 0.0122, 0.0538, 0.0035, -0.0646, -0.0087, 0.0361, -0.0247, + -0.0195, 0.0234], device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 214.21, cls_loss 0.5987 cls_loss_mapping 0.0075 cls_loss_causal 0.5201 re_mapping 0.0127 re_causal 0.0287 /// teacc 98.47 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0651, -0.0922, -0.0767, ..., -0.0267, 0.0489, -0.0847], + [-0.0504, -0.0748, -0.0512, ..., 0.0949, -0.0329, 0.1782], + [ 0.0118, -0.0160, -0.0235, ..., -0.0107, -0.0111, -0.0577], + ..., + [ 0.0185, -0.0804, 0.1308, ..., 0.0191, -0.0662, 0.0374], + [-0.0273, 0.0467, -0.1182, ..., -0.0468, 0.0067, -0.0697], + [-0.0526, 0.0671, 0.0511, ..., -0.0472, -0.0321, -0.0158]], + device='cuda:0'), grad: tensor([[ 5.6922e-05, -1.7662e-03, 1.7357e-04, ..., -6.5117e-03, + -6.1178e-04, -1.1749e-02], + [ 6.0022e-05, 2.2054e-04, -3.5596e-04, ..., -8.3313e-03, + 3.6001e-05, -4.3755e-03], + [ 8.5771e-05, 2.2972e-04, 5.3482e-03, ..., 4.1161e-03, + 3.0589e-04, 4.3755e-03], + ..., + [-1.3247e-03, -5.8937e-03, -1.3969e-02, ..., 1.8816e-03, + 1.9725e-06, 1.2312e-03], + [ 1.5056e-04, 1.1997e-03, 8.0919e-04, ..., -1.0689e-02, + 1.0586e-03, -2.0046e-03], + [ 1.3762e-03, 5.4550e-03, 6.0730e-03, ..., 6.7062e-03, + 4.0913e-04, 2.8267e-03]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0200, 0.0269, 0.0037, -0.0098, 0.0211, -0.0295, 0.0084, -0.0058, + -0.0001, -0.0117], device='cuda:0'), grad: tensor([-0.0484, -0.0172, 0.0200, -0.0115, -0.0046, 0.0266, 0.0244, -0.0315, + 0.0002, 0.0420], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 159---------------------------------------------------- +epoch 159, time 230.93, cls_loss 0.6082 cls_loss_mapping 0.0101 cls_loss_causal 0.5282 re_mapping 0.0128 re_causal 0.0286 /// teacc 98.68 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0653, -0.0938, -0.0785, ..., -0.0267, 0.0478, -0.0849], + [-0.0514, -0.0750, -0.0491, ..., 0.0953, -0.0322, 0.1784], + [ 0.0125, -0.0167, -0.0257, ..., -0.0111, -0.0120, -0.0574], + ..., + [ 0.0197, -0.0783, 0.1331, ..., 0.0207, -0.0648, 0.0378], + [-0.0282, 0.0470, -0.1178, ..., -0.0482, 0.0078, -0.0701], + [-0.0522, 0.0656, 0.0505, ..., -0.0464, -0.0329, -0.0158]], + device='cuda:0'), grad: tensor([[ 3.2878e-04, -8.8310e-04, 1.3933e-05, ..., 3.1447e-04, + -6.8247e-05, -7.7295e-04], + [ 7.4029e-05, 6.9475e-04, 6.8951e-04, ..., -4.5319e-03, + 7.0315e-07, -1.1510e-04], + [-4.6883e-03, -1.9928e-02, -1.5335e-02, ..., -1.1375e-02, + 1.5721e-05, 4.8375e-04], + ..., + [-1.5125e-03, 2.0275e-03, 4.4584e-04, ..., 3.5686e-03, + 5.4389e-07, 1.2505e-04], + [ 4.7159e-04, 1.1490e-02, -2.7905e-03, ..., -1.1377e-03, + 3.9756e-05, 9.3579e-05], + [ 1.2245e-03, -2.0256e-03, 5.0507e-03, ..., 8.1015e-04, + 5.0999e-06, 3.3355e-04]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0208, 0.0266, 0.0044, -0.0096, 0.0205, -0.0295, 0.0084, -0.0052, + -0.0006, -0.0112], device='cuda:0'), grad: tensor([-0.0044, -0.0191, -0.0610, 0.0449, -0.0272, 0.0237, 0.0127, 0.0086, + -0.0089, 0.0308], device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 214.32, cls_loss 0.5948 cls_loss_mapping 0.0084 cls_loss_causal 0.5122 re_mapping 0.0136 re_causal 0.0299 /// teacc 98.54 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0640, -0.0911, -0.0782, ..., -0.0272, 0.0481, -0.0856], + [-0.0512, -0.0743, -0.0490, ..., 0.0956, -0.0324, 0.1792], + [ 0.0124, -0.0177, -0.0258, ..., -0.0116, -0.0119, -0.0575], + ..., + [ 0.0196, -0.0786, 0.1329, ..., 0.0207, -0.0646, 0.0379], + [-0.0289, 0.0478, -0.1183, ..., -0.0482, 0.0084, -0.0715], + [-0.0529, 0.0653, 0.0509, ..., -0.0464, -0.0326, -0.0155]], + device='cuda:0'), grad: tensor([[ 1.0180e-04, 1.5802e-03, -9.6035e-04, ..., 4.7836e-03, + 2.3317e-04, 2.0218e-04], + [ 6.7234e-05, 5.8508e-04, 3.6883e-04, ..., 5.2910e-03, + 2.3675e-04, 9.3102e-05], + [-7.8487e-04, -3.8795e-03, -4.9896e-03, ..., -1.4206e-02, + -8.8692e-04, 3.7479e-04], + ..., + [ 2.2578e-04, 1.1606e-03, 1.1311e-03, ..., 9.2163e-03, + 5.8222e-04, 8.6403e-04], + [-3.2663e-04, 4.6234e-03, -2.3098e-03, ..., 1.2108e-02, + 1.7891e-03, 1.1911e-03], + [ 1.1545e-04, 2.1219e-04, 1.5640e-03, ..., -1.3695e-02, + -7.4005e-04, -4.8447e-03]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0203, 0.0269, 0.0043, -0.0102, 0.0199, -0.0301, 0.0096, -0.0051, + -0.0008, -0.0109], device='cuda:0'), grad: tensor([ 0.0055, 0.0172, -0.0334, -0.0037, -0.0142, -0.0329, 0.0283, 0.0288, + 0.0383, -0.0339], device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 214.11, cls_loss 0.5875 cls_loss_mapping 0.0095 cls_loss_causal 0.5011 re_mapping 0.0131 re_causal 0.0286 /// teacc 98.65 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0633, -0.0917, -0.0780, ..., -0.0276, 0.0466, -0.0859], + [-0.0514, -0.0744, -0.0490, ..., 0.0965, -0.0325, 0.1796], + [ 0.0124, -0.0176, -0.0268, ..., -0.0119, -0.0112, -0.0574], + ..., + [ 0.0202, -0.0799, 0.1336, ..., 0.0205, -0.0650, 0.0370], + [-0.0288, 0.0474, -0.1180, ..., -0.0484, 0.0088, -0.0730], + [-0.0532, 0.0662, 0.0506, ..., -0.0471, -0.0327, -0.0143]], + device='cuda:0'), grad: tensor([[ 4.7833e-05, 2.2163e-03, 1.5032e-04, ..., 5.1651e-03, + -4.2349e-05, 1.6391e-04], + [ 8.7619e-05, 2.8968e-04, 4.2343e-04, ..., 3.9215e-03, + 6.2399e-08, 5.5933e-04], + [-4.3654e-04, 5.9242e-03, -8.7214e-04, ..., -4.2419e-03, + 1.4693e-05, -1.0948e-03], + ..., + [ 1.5724e-04, 1.4725e-03, -2.6608e-04, ..., 3.9558e-03, + 4.7963e-07, -4.3893e-04], + [ 4.1902e-05, 5.2109e-03, 2.9969e-04, ..., 7.4577e-03, + 2.1160e-06, 1.8966e-04], + [ 9.4324e-06, 1.4138e-04, 4.5013e-04, ..., -3.7022e-03, + 3.8557e-06, 2.5892e-04]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0207, 0.0279, 0.0038, -0.0111, 0.0209, -0.0291, 0.0095, -0.0060, + -0.0008, -0.0112], device='cuda:0'), grad: tensor([ 0.0193, 0.0158, -0.0186, -0.0292, -0.0089, 0.0292, -0.0253, 0.0128, + 0.0269, -0.0219], device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 214.24, cls_loss 0.6252 cls_loss_mapping 0.0091 cls_loss_causal 0.5403 re_mapping 0.0130 re_causal 0.0287 /// teacc 98.60 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0642, -0.0921, -0.0789, ..., -0.0285, 0.0470, -0.0858], + [-0.0523, -0.0740, -0.0496, ..., 0.0961, -0.0327, 0.1791], + [ 0.0144, -0.0184, -0.0276, ..., -0.0119, -0.0103, -0.0554], + ..., + [ 0.0199, -0.0800, 0.1342, ..., 0.0203, -0.0653, 0.0368], + [-0.0295, 0.0477, -0.1186, ..., -0.0481, 0.0089, -0.0737], + [-0.0528, 0.0665, 0.0512, ..., -0.0467, -0.0333, -0.0144]], + device='cuda:0'), grad: tensor([[ 1.4954e-03, 1.1539e-03, 1.6708e-03, ..., 5.6572e-03, + 1.7738e-04, 8.4400e-04], + [ 1.4460e-04, 4.7684e-04, 1.6556e-03, ..., -3.7708e-03, + -8.2970e-04, 1.4238e-03], + [ 6.5136e-04, 1.6727e-03, 1.7805e-03, ..., 5.4016e-03, + 6.2180e-04, 1.2522e-03], + ..., + [ 7.9036e-05, -5.5771e-03, -1.2810e-02, ..., -1.2886e-02, + -4.3464e-04, -6.5041e-03], + [-4.3602e-03, -7.7534e-04, -9.8896e-04, ..., -1.8415e-03, + 4.2892e-04, 1.7195e-03], + [ 5.7268e-04, -1.7138e-03, -1.8272e-03, ..., -1.2634e-02, + 1.5175e-04, -1.4648e-03]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0216, 0.0277, 0.0042, -0.0113, 0.0214, -0.0283, 0.0092, -0.0058, + -0.0010, -0.0111], device='cuda:0'), grad: tensor([ 0.0159, -0.0126, 0.0169, 0.0253, 0.0176, -0.0008, 0.0184, -0.0368, + -0.0047, -0.0391], device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 214.06, cls_loss 0.6337 cls_loss_mapping 0.0084 cls_loss_causal 0.5519 re_mapping 0.0125 re_causal 0.0282 /// teacc 98.66 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0657, -0.0936, -0.0803, ..., -0.0290, 0.0470, -0.0876], + [-0.0531, -0.0739, -0.0509, ..., 0.0967, -0.0334, 0.1796], + [ 0.0142, -0.0202, -0.0278, ..., -0.0119, -0.0105, -0.0572], + ..., + [ 0.0201, -0.0785, 0.1345, ..., 0.0207, -0.0649, 0.0368], + [-0.0289, 0.0479, -0.1195, ..., -0.0488, 0.0096, -0.0724], + [-0.0532, 0.0667, 0.0522, ..., -0.0467, -0.0327, -0.0136]], + device='cuda:0'), grad: tensor([[ 1.2074e-03, 3.9268e-04, 1.0653e-03, ..., 6.6566e-04, + -1.9705e-04, 8.5926e-04], + [ 5.6267e-04, 7.1108e-05, -1.3895e-03, ..., -3.7212e-03, + 5.0449e-04, -5.5838e-04], + [ 1.7910e-03, -1.2035e-03, -2.0866e-03, ..., -1.3489e-02, + -1.7786e-03, 1.6422e-03], + ..., + [-2.1820e-03, 3.6311e-04, -3.3131e-03, ..., -7.3290e-04, + -9.5654e-04, 7.8306e-06], + [ 1.4973e-03, -3.3784e-04, -8.9169e-04, ..., 2.5444e-03, + -1.4839e-02, 1.0567e-03], + [ 1.9503e-03, 7.6914e-04, 1.6546e-03, ..., 1.0658e-02, + 5.4665e-03, 2.0065e-03]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0224, 0.0274, 0.0042, -0.0107, 0.0207, -0.0284, 0.0100, -0.0061, + -0.0005, -0.0111], device='cuda:0'), grad: tensor([-0.0090, -0.0142, -0.0334, 0.0280, -0.0081, -0.0282, 0.0143, -0.0043, + -0.0081, 0.0630], device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.25, cls_loss 0.6098 cls_loss_mapping 0.0081 cls_loss_causal 0.5236 re_mapping 0.0125 re_causal 0.0278 /// teacc 98.66 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0663, -0.0932, -0.0816, ..., -0.0304, 0.0465, -0.0892], + [-0.0551, -0.0757, -0.0506, ..., 0.0969, -0.0338, 0.1793], + [ 0.0138, -0.0205, -0.0283, ..., -0.0121, -0.0105, -0.0579], + ..., + [ 0.0214, -0.0790, 0.1347, ..., 0.0203, -0.0648, 0.0373], + [-0.0279, 0.0486, -0.1191, ..., -0.0481, 0.0095, -0.0726], + [-0.0534, 0.0661, 0.0516, ..., -0.0462, -0.0338, -0.0146]], + device='cuda:0'), grad: tensor([[ 5.7518e-06, -1.0468e-02, -1.2123e-02, ..., -1.1383e-02, + 2.1648e-04, -5.9128e-04], + [ 4.6372e-05, -1.1735e-05, 8.0156e-04, ..., 2.9907e-03, + 2.7919e-04, -4.9591e-03], + [-4.9114e-04, 1.4806e-04, 2.2435e-04, ..., -3.8033e-03, + 2.7132e-04, 4.1318e-04], + ..., + [ 3.0965e-05, 1.0842e-04, 1.9121e-04, ..., 6.0425e-03, + 2.8968e-04, 5.6887e-04], + [ 2.3878e-04, 1.0433e-03, 2.0618e-03, ..., 4.0802e-02, + 2.5368e-04, 2.1553e-03], + [ 6.9067e-06, 1.3125e-04, -3.1109e-03, ..., -6.1750e-04, + 3.6430e-04, -6.0606e-04]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0234, 0.0277, 0.0037, -0.0098, 0.0209, -0.0288, 0.0097, -0.0063, + 0.0003, -0.0111], device='cuda:0'), grad: tensor([-0.0220, 0.0021, -0.0091, 0.0350, 0.0222, -0.0729, -0.0136, 0.0165, + 0.0416, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 213.91, cls_loss 0.6014 cls_loss_mapping 0.0072 cls_loss_causal 0.5102 re_mapping 0.0126 re_causal 0.0298 /// teacc 98.58 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0660, -0.0931, -0.0803, ..., -0.0289, 0.0477, -0.0897], + [-0.0566, -0.0754, -0.0513, ..., 0.0956, -0.0331, 0.1813], + [ 0.0146, -0.0209, -0.0287, ..., -0.0116, -0.0105, -0.0589], + ..., + [ 0.0212, -0.0786, 0.1354, ..., 0.0204, -0.0640, 0.0368], + [-0.0292, 0.0483, -0.1206, ..., -0.0486, 0.0092, -0.0729], + [-0.0540, 0.0658, 0.0519, ..., -0.0463, -0.0342, -0.0144]], + device='cuda:0'), grad: tensor([[-0.0004, -0.0012, 0.0010, ..., -0.0035, -0.0020, 0.0018], + [ 0.0002, 0.0005, 0.0056, ..., 0.0186, 0.0019, 0.0100], + [-0.0010, 0.0012, 0.0029, ..., 0.0004, 0.0018, 0.0007], + ..., + [ 0.0003, 0.0005, -0.0132, ..., -0.0031, -0.0025, -0.0227], + [ 0.0004, 0.0016, 0.0016, ..., 0.0073, 0.0021, 0.0003], + [ 0.0002, 0.0009, 0.0051, ..., -0.0439, 0.0004, 0.0163]], + device='cuda:0') +Epoch 167, bias, value: tensor([-0.0220, 0.0269, 0.0039, -0.0109, 0.0205, -0.0280, 0.0096, -0.0053, + -0.0003, -0.0114], device='cuda:0'), grad: tensor([-0.0186, 0.0483, 0.0039, -0.0298, 0.0054, -0.0036, 0.0434, -0.0222, + 0.0219, -0.0486], device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 213.82, cls_loss 0.6109 cls_loss_mapping 0.0077 cls_loss_causal 0.5288 re_mapping 0.0119 re_causal 0.0279 /// teacc 98.50 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0660, -0.0927, -0.0807, ..., -0.0305, 0.0491, -0.0898], + [-0.0567, -0.0760, -0.0514, ..., 0.0952, -0.0315, 0.1815], + [ 0.0148, -0.0210, -0.0300, ..., -0.0117, -0.0101, -0.0599], + ..., + [ 0.0209, -0.0783, 0.1370, ..., 0.0209, -0.0651, 0.0377], + [-0.0305, 0.0485, -0.1224, ..., -0.0487, 0.0087, -0.0729], + [-0.0534, 0.0651, 0.0512, ..., -0.0448, -0.0339, -0.0148]], + device='cuda:0'), grad: tensor([[ 2.0966e-05, 8.9931e-04, 2.4548e-03, ..., 2.7943e-03, + 5.4016e-02, 1.6487e-04], + [ 4.9919e-05, -1.8656e-04, -6.9332e-04, ..., -5.6038e-03, + 3.3188e-04, -5.2261e-04], + [ 7.9155e-05, 3.3355e-04, 2.9907e-03, ..., -7.0572e-04, + 1.2150e-03, 1.3006e-04], + ..., + [-4.1175e-04, 1.4782e-04, -4.8485e-03, ..., -1.2436e-02, + -2.5487e-04, -5.2643e-04], + [ 3.7760e-05, -5.7268e-04, 1.6518e-03, ..., 6.3591e-03, + -5.4596e-02, 2.0802e-05], + [ 1.6963e-04, 4.0698e-04, 9.0742e-04, ..., 7.4539e-03, + 9.5654e-04, -4.4823e-04]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0229, 0.0264, 0.0038, -0.0096, 0.0194, -0.0282, 0.0098, -0.0047, + -0.0007, -0.0104], device='cuda:0'), grad: tensor([ 0.0197, -0.0059, -0.0009, 0.0081, -0.0474, 0.0148, 0.0284, -0.0289, + -0.0112, 0.0233], device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 214.24, cls_loss 0.6114 cls_loss_mapping 0.0090 cls_loss_causal 0.5300 re_mapping 0.0124 re_causal 0.0287 /// teacc 98.53 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0670, -0.0929, -0.0808, ..., -0.0309, 0.0500, -0.0901], + [-0.0574, -0.0760, -0.0512, ..., 0.0959, -0.0288, 0.1825], + [ 0.0156, -0.0217, -0.0294, ..., -0.0115, -0.0109, -0.0608], + ..., + [ 0.0207, -0.0788, 0.1360, ..., 0.0203, -0.0670, 0.0379], + [-0.0305, 0.0484, -0.1221, ..., -0.0481, 0.0097, -0.0735], + [-0.0522, 0.0654, 0.0523, ..., -0.0448, -0.0358, -0.0141]], + device='cuda:0'), grad: tensor([[ 9.1940e-06, 1.7500e-04, 5.1212e-04, ..., 3.8338e-03, + 2.3651e-04, 9.7454e-05], + [ 1.9133e-05, -1.5545e-03, -7.7438e-04, ..., -1.4351e-02, + -3.6392e-03, -1.0328e-03], + [ 9.0361e-05, 9.4461e-04, 4.6768e-03, ..., -6.7329e-03, + 1.1806e-03, 3.3641e-04], + ..., + [ 1.5771e-04, 7.7629e-04, -6.3133e-03, ..., -5.2948e-03, + 2.9039e-04, -6.5379e-06], + [-1.0195e-03, -1.6678e-02, -1.1765e-02, ..., 3.8223e-03, + 1.4477e-03, -1.5962e-04], + [ 6.3753e-04, 1.5976e-02, 1.0452e-02, ..., -3.4733e-03, + -1.2360e-03, 9.0933e-04]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0229, 0.0270, 0.0033, -0.0097, 0.0191, -0.0281, 0.0096, -0.0052, + 0.0004, -0.0106], device='cuda:0'), grad: tensor([ 0.0104, -0.0273, -0.0358, 0.0291, 0.0105, 0.0062, 0.0138, -0.0081, + -0.0174, 0.0184], device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 213.98, cls_loss 0.5814 cls_loss_mapping 0.0077 cls_loss_causal 0.5032 re_mapping 0.0121 re_causal 0.0280 /// teacc 98.60 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0686, -0.0937, -0.0815, ..., -0.0306, 0.0494, -0.0896], + [-0.0581, -0.0757, -0.0495, ..., 0.0967, -0.0296, 0.1845], + [ 0.0147, -0.0216, -0.0305, ..., -0.0120, -0.0117, -0.0618], + ..., + [ 0.0196, -0.0788, 0.1368, ..., 0.0202, -0.0655, 0.0374], + [-0.0299, 0.0481, -0.1241, ..., -0.0487, 0.0095, -0.0731], + [-0.0536, 0.0653, 0.0519, ..., -0.0445, -0.0369, -0.0150]], + device='cuda:0'), grad: tensor([[ 9.6709e-06, 2.4652e-04, 6.8331e-04, ..., 3.6354e-03, + 1.6713e-04, 2.3246e-04], + [ 4.4584e-05, 4.9162e-04, 1.0462e-03, ..., -2.7313e-03, + 3.8171e-04, -1.4496e-03], + [-7.2360e-05, -1.6451e-03, -4.4365e-03, ..., -4.7188e-03, + -7.1049e-04, 5.3501e-04], + ..., + [-3.9518e-05, -6.3229e-04, 5.3310e-04, ..., -2.8973e-03, + -8.1062e-04, -6.5207e-05], + [ 5.1212e-04, 4.4703e-04, -1.8501e-03, ..., 1.6241e-03, + 3.1796e-03, -1.2207e-03], + [ 1.5652e-04, 2.0933e-04, -1.6508e-03, ..., -8.6498e-04, + -2.3193e-03, 4.5538e-04]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0225, 0.0277, 0.0029, -0.0085, 0.0194, -0.0283, 0.0089, -0.0063, + 0.0003, -0.0108], device='cuda:0'), grad: tensor([ 0.0111, -0.0102, -0.0098, 0.0017, 0.0182, -0.0168, 0.0152, -0.0060, + 0.0023, -0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.24, cls_loss 0.5749 cls_loss_mapping 0.0078 cls_loss_causal 0.4880 re_mapping 0.0125 re_causal 0.0270 /// teacc 98.61 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0690, -0.0943, -0.0810, ..., -0.0305, 0.0496, -0.0899], + [-0.0581, -0.0759, -0.0500, ..., 0.0965, -0.0296, 0.1842], + [ 0.0153, -0.0196, -0.0307, ..., -0.0123, -0.0118, -0.0618], + ..., + [ 0.0201, -0.0795, 0.1377, ..., 0.0207, -0.0654, 0.0383], + [-0.0314, 0.0476, -0.1253, ..., -0.0488, 0.0083, -0.0731], + [-0.0524, 0.0660, 0.0507, ..., -0.0459, -0.0370, -0.0156]], + device='cuda:0'), grad: tensor([[ 2.5500e-06, -1.4753e-03, -2.2049e-03, ..., -1.5312e-02, + 9.8133e-04, 3.4785e-04], + [ 8.0988e-06, 1.6677e-04, 1.3742e-03, ..., 3.9062e-03, + 1.2279e-04, -1.4477e-03], + [ 5.8711e-05, 7.9155e-04, 2.9449e-03, ..., 6.2447e-03, + 5.4216e-04, 5.2357e-04], + ..., + [-5.7280e-05, 1.2455e-03, 3.1796e-03, ..., 6.2828e-03, + 1.7774e-04, 3.8624e-04], + [ 2.6226e-05, 3.7289e-03, 2.5578e-03, ..., 3.7136e-03, + 6.4564e-04, 4.4107e-04], + [ 7.4096e-06, -7.0648e-03, -1.7395e-02, ..., -1.9257e-02, + -5.9891e-03, 5.6267e-04]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0221, 0.0283, 0.0029, -0.0083, 0.0197, -0.0281, 0.0082, -0.0058, + -0.0002, -0.0116], device='cuda:0'), grad: tensor([-0.0429, 0.0146, 0.0209, 0.0289, -0.0174, -0.0113, 0.0257, 0.0247, + 0.0213, -0.0645], device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 214.05, cls_loss 0.6002 cls_loss_mapping 0.0081 cls_loss_causal 0.5238 re_mapping 0.0129 re_causal 0.0278 /// teacc 98.63 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0703, -0.0945, -0.0797, ..., -0.0316, 0.0494, -0.0887], + [-0.0581, -0.0764, -0.0509, ..., 0.0963, -0.0298, 0.1846], + [ 0.0149, -0.0189, -0.0306, ..., -0.0110, -0.0116, -0.0633], + ..., + [ 0.0197, -0.0799, 0.1366, ..., 0.0208, -0.0657, 0.0387], + [-0.0326, 0.0482, -0.1244, ..., -0.0489, 0.0086, -0.0733], + [-0.0515, 0.0659, 0.0511, ..., -0.0457, -0.0373, -0.0167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3602e-04, -1.9550e-03, ..., 6.9046e-04, + -4.0197e-04, -1.5497e-03], + [ 0.0000e+00, 2.2936e-04, 1.1454e-03, ..., 1.0170e-02, + 1.4315e-03, 1.5535e-03], + [ 2.7940e-09, 1.3332e-03, 1.8654e-03, ..., -3.5095e-03, + 9.6321e-04, 2.5821e-04], + ..., + [ 0.0000e+00, 3.7980e-04, 1.8101e-03, ..., 1.0704e-02, + 9.8896e-04, 1.6050e-03], + [ 3.7253e-09, 1.5129e-02, 5.2500e-04, ..., -8.5068e-03, + 3.7975e-03, -5.4779e-03], + [ 0.0000e+00, 4.9448e-04, 1.6117e-03, ..., 1.1284e-02, + 1.4257e-03, 2.3251e-03]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0227, 0.0280, 0.0041, -0.0086, 0.0195, -0.0283, 0.0084, -0.0058, + -0.0005, -0.0112], device='cuda:0'), grad: tensor([ 0.0069, 0.0344, -0.0157, 0.0006, 0.0018, 0.0052, -0.0663, 0.0125, + -0.0163, 0.0368], device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 214.07, cls_loss 0.5918 cls_loss_mapping 0.0100 cls_loss_causal 0.4966 re_mapping 0.0129 re_causal 0.0276 /// teacc 98.59 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0692, -0.0940, -0.0798, ..., -0.0313, 0.0493, -0.0898], + [-0.0581, -0.0764, -0.0515, ..., 0.0967, -0.0308, 0.1843], + [ 0.0142, -0.0196, -0.0294, ..., -0.0111, -0.0122, -0.0619], + ..., + [ 0.0196, -0.0800, 0.1374, ..., 0.0224, -0.0656, 0.0395], + [-0.0317, 0.0477, -0.1243, ..., -0.0496, 0.0095, -0.0733], + [-0.0516, 0.0659, 0.0504, ..., -0.0450, -0.0372, -0.0160]], + device='cuda:0'), grad: tensor([[-5.9090e-03, 1.1247e-04, 2.1706e-03, ..., -1.1757e-02, + 1.1797e-03, 6.1846e-04], + [ 2.0103e-03, 5.0753e-05, 2.1038e-03, ..., 1.1871e-02, + 6.6566e-04, 1.6594e-03], + [ 1.4801e-03, 2.9755e-04, 1.2388e-03, ..., 7.0152e-03, + -4.9896e-03, 1.7166e-03], + ..., + [-8.5413e-05, -8.4829e-04, -3.3321e-03, ..., 4.7112e-03, + 1.1587e-03, -2.5654e-03], + [-1.9324e-04, 2.2554e-04, -1.8559e-03, ..., -5.0163e-03, + 1.4944e-03, -7.0477e-04], + [ 7.3147e-04, -6.9714e-04, -4.0770e-04, ..., 1.6222e-03, + 7.2527e-04, 1.8339e-03]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0224, 0.0281, 0.0036, -0.0089, 0.0192, -0.0284, 0.0081, -0.0041, + -0.0011, -0.0112], device='cuda:0'), grad: tensor([-0.0119, 0.0305, 0.0139, -0.0007, -0.0499, 0.0225, -0.0170, 0.0192, + -0.0113, 0.0046], device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 214.46, cls_loss 0.5987 cls_loss_mapping 0.0086 cls_loss_causal 0.5133 re_mapping 0.0127 re_causal 0.0278 /// teacc 98.64 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0689, -0.0946, -0.0805, ..., -0.0299, 0.0492, -0.0895], + [-0.0601, -0.0773, -0.0513, ..., 0.0964, -0.0303, 0.1838], + [ 0.0143, -0.0197, -0.0293, ..., -0.0114, -0.0127, -0.0622], + ..., + [ 0.0211, -0.0797, 0.1376, ..., 0.0228, -0.0664, 0.0396], + [-0.0317, 0.0480, -0.1254, ..., -0.0501, 0.0096, -0.0725], + [-0.0507, 0.0654, 0.0504, ..., -0.0454, -0.0382, -0.0168]], + device='cuda:0'), grad: tensor([[ 6.0469e-05, 3.0541e-04, 1.4944e-03, ..., 4.1847e-03, + -8.6880e-04, 2.3887e-05], + [ 2.7835e-05, 3.1501e-05, -1.0567e-02, ..., -6.0387e-03, + -2.2078e-04, -2.1038e-03], + [ 2.0771e-03, 2.6512e-03, 8.3389e-03, ..., 5.6305e-03, + 2.0542e-03, 9.0837e-04], + ..., + [-1.9875e-03, 1.9217e-04, -1.9064e-03, ..., -6.4735e-03, + 1.6856e-04, -8.3542e-04], + [ 7.8499e-05, 7.6580e-04, 1.0653e-03, ..., -2.6131e-03, + 2.3758e-04, 1.3173e-04], + [ 2.0540e-04, 3.1447e-04, -2.4261e-03, ..., -3.4924e-03, + -6.4898e-04, 8.9550e-04]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0218, 0.0277, 0.0039, -0.0090, 0.0199, -0.0286, 0.0083, -0.0047, + -0.0016, -0.0111], device='cuda:0'), grad: tensor([ 0.0125, -0.0151, 0.0239, -0.0104, 0.0075, 0.0099, 0.0169, -0.0201, + -0.0160, -0.0090], device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 214.18, cls_loss 0.6239 cls_loss_mapping 0.0089 cls_loss_causal 0.5474 re_mapping 0.0127 re_causal 0.0287 /// teacc 98.57 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0698, -0.0948, -0.0820, ..., -0.0298, 0.0490, -0.0893], + [-0.0606, -0.0765, -0.0512, ..., 0.0968, -0.0304, 0.1832], + [ 0.0160, -0.0190, -0.0293, ..., -0.0127, -0.0119, -0.0637], + ..., + [ 0.0205, -0.0798, 0.1379, ..., 0.0229, -0.0665, 0.0399], + [-0.0320, 0.0480, -0.1259, ..., -0.0509, 0.0106, -0.0738], + [-0.0524, 0.0650, 0.0509, ..., -0.0463, -0.0390, -0.0161]], + device='cuda:0'), grad: tensor([[ 1.3806e-05, 2.5824e-05, 1.9097e-04, ..., 5.0697e-03, + 1.7405e-05, 3.0994e-04], + [-5.6219e-04, -1.6510e-04, 1.1927e-04, ..., -1.3397e-02, + -1.5512e-05, -4.1389e-03], + [ 2.6166e-05, 6.3241e-05, 3.0470e-04, ..., 5.5809e-03, + 6.5565e-05, 3.2735e-04], + ..., + [ 1.9372e-05, 8.9645e-05, -2.6340e-03, ..., -3.9291e-03, + 2.3946e-05, 3.2104e-02], + [ 3.0494e-04, 1.5306e-03, 3.6526e-04, ..., 8.4839e-03, + 3.4027e-03, 4.5204e-03], + [-2.4259e-05, -1.8477e-05, 1.2474e-03, ..., -2.9125e-03, + 3.8385e-05, 6.2637e-03]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0223, 0.0284, 0.0029, -0.0079, 0.0203, -0.0284, 0.0086, -0.0052, + -0.0017, -0.0117], device='cuda:0'), grad: tensor([ 0.0184, -0.0220, 0.0179, 0.0106, -0.0635, 0.0084, 0.0073, 0.0008, + 0.0231, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 174---------------------------------------------------- +epoch 174, time 230.55, cls_loss 0.5745 cls_loss_mapping 0.0084 cls_loss_causal 0.4976 re_mapping 0.0124 re_causal 0.0282 /// teacc 98.69 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0709, -0.0953, -0.0832, ..., -0.0307, 0.0490, -0.0903], + [-0.0611, -0.0767, -0.0515, ..., 0.0966, -0.0306, 0.1833], + [ 0.0172, -0.0183, -0.0285, ..., -0.0127, -0.0109, -0.0618], + ..., + [ 0.0202, -0.0804, 0.1377, ..., 0.0234, -0.0674, 0.0383], + [-0.0323, 0.0475, -0.1272, ..., -0.0514, 0.0106, -0.0756], + [-0.0529, 0.0653, 0.0519, ..., -0.0468, -0.0392, -0.0146]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 1.1530e-03, 1.0052e-03, ..., 2.8934e-03, + 3.7122e-04, 7.4387e-04], + [ 8.0094e-08, -2.8491e-04, 5.3549e-04, ..., 1.3943e-03, + 1.6856e-04, -1.9836e-03], + [ 4.3865e-07, 1.4365e-04, 7.2718e-04, ..., 2.5101e-03, + 2.6107e-05, 3.9148e-04], + ..., + [-1.3094e-06, -1.4553e-03, -1.6373e-02, ..., -9.9945e-04, + 3.0184e-04, -1.0544e-02], + [-1.5851e-06, 1.4000e-02, 1.0662e-03, ..., 2.5177e-03, + 1.7440e-04, 2.1291e-04], + [ 8.3819e-08, 3.8548e-03, 1.7319e-02, ..., 6.1150e-03, + 2.1613e-04, 9.8801e-03]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0231, 0.0288, 0.0028, -0.0076, 0.0206, -0.0292, 0.0103, -0.0047, + -0.0027, -0.0121], device='cuda:0'), grad: tensor([ 0.0127, 0.0045, -0.0170, -0.0107, -0.0547, -0.0070, -0.0024, 0.0125, + 0.0329, 0.0293], device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 214.51, cls_loss 0.5774 cls_loss_mapping 0.0070 cls_loss_causal 0.5013 re_mapping 0.0125 re_causal 0.0277 /// teacc 98.59 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0713, -0.0955, -0.0830, ..., -0.0299, 0.0485, -0.0889], + [-0.0605, -0.0771, -0.0525, ..., 0.0966, -0.0295, 0.1833], + [ 0.0171, -0.0185, -0.0274, ..., -0.0128, -0.0104, -0.0624], + ..., + [ 0.0199, -0.0807, 0.1376, ..., 0.0228, -0.0680, 0.0382], + [-0.0327, 0.0481, -0.1273, ..., -0.0516, 0.0108, -0.0729], + [-0.0541, 0.0652, 0.0517, ..., -0.0462, -0.0391, -0.0154]], + device='cuda:0'), grad: tensor([[ 6.6471e-04, 4.9019e-04, 2.5425e-03, ..., 6.3171e-03, + 8.3256e-04, 1.6508e-03], + [ 8.9109e-05, 5.1856e-05, -3.3647e-05, ..., -8.8882e-04, + 1.8978e-04, -2.4390e-04], + [ 1.4806e-04, 1.3666e-03, -7.5674e-04, ..., -1.0124e-02, + -2.1350e-04, -8.4972e-04], + ..., + [ 1.7715e-04, 1.5688e-04, 1.2026e-03, ..., 5.4474e-03, + 2.9397e-04, 5.2881e-04], + [ 4.8470e-04, 1.5173e-03, 7.9918e-04, ..., 5.3978e-03, + 6.1893e-04, 1.0128e-03], + [-9.1076e-04, -9.0885e-04, -2.5463e-03, ..., 4.2229e-03, + -6.0129e-04, -2.8820e-03]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0225, 0.0290, 0.0035, -0.0089, 0.0206, -0.0291, 0.0097, -0.0056, + -0.0028, -0.0109], device='cuda:0'), grad: tensor([ 0.0303, -0.0086, -0.0311, 0.0073, -0.0333, -0.0421, 0.0175, 0.0269, + 0.0267, 0.0065], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 176---------------------------------------------------- +epoch 176, time 230.63, cls_loss 0.5973 cls_loss_mapping 0.0063 cls_loss_causal 0.5171 re_mapping 0.0122 re_causal 0.0276 /// teacc 98.70 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0725, -0.0949, -0.0824, ..., -0.0298, 0.0499, -0.0904], + [-0.0607, -0.0775, -0.0523, ..., 0.0968, -0.0318, 0.1832], + [ 0.0168, -0.0187, -0.0279, ..., -0.0133, -0.0099, -0.0621], + ..., + [ 0.0202, -0.0807, 0.1381, ..., 0.0232, -0.0686, 0.0397], + [-0.0334, 0.0478, -0.1262, ..., -0.0517, 0.0113, -0.0734], + [-0.0549, 0.0661, 0.0520, ..., -0.0469, -0.0399, -0.0156]], + device='cuda:0'), grad: tensor([[ 3.1963e-06, 1.6379e-04, 4.9882e-06, ..., 2.9716e-03, + 2.7485e-03, 3.3788e-06], + [ 5.0571e-07, 5.0843e-05, 1.0721e-05, ..., 2.7122e-03, + 2.7294e-03, -1.4853e-04], + [ 4.3660e-05, -1.4007e-04, 7.6652e-05, ..., -4.1275e-03, + -3.6163e-03, 2.5451e-05], + ..., + [ 3.3509e-06, 2.6941e-04, 1.4877e-04, ..., -4.9934e-03, + -2.1782e-03, 1.8632e-04], + [ 8.8513e-05, -1.3056e-03, -1.6956e-03, ..., 1.3676e-03, + 3.6812e-03, -6.7115e-05], + [-1.1331e-04, 8.3590e-04, 5.4979e-04, ..., -2.9049e-03, + 1.1438e-04, 7.5877e-05]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0219, 0.0291, 0.0026, -0.0092, 0.0215, -0.0283, 0.0094, -0.0058, + -0.0028, -0.0114], device='cuda:0'), grad: tensor([ 0.0219, 0.0177, -0.0103, -0.0095, 0.0220, -0.0303, 0.0185, -0.0251, + 0.0355, -0.0404], device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 214.50, cls_loss 0.6163 cls_loss_mapping 0.0062 cls_loss_causal 0.5338 re_mapping 0.0118 re_causal 0.0272 /// teacc 98.57 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0726, -0.0941, -0.0818, ..., -0.0301, 0.0488, -0.0890], + [-0.0602, -0.0780, -0.0527, ..., 0.0957, -0.0309, 0.1831], + [ 0.0155, -0.0182, -0.0279, ..., -0.0121, -0.0093, -0.0617], + ..., + [ 0.0228, -0.0813, 0.1381, ..., 0.0235, -0.0684, 0.0390], + [-0.0349, 0.0476, -0.1270, ..., -0.0523, 0.0093, -0.0732], + [-0.0556, 0.0666, 0.0518, ..., -0.0474, -0.0395, -0.0157]], + device='cuda:0'), grad: tensor([[ 1.9073e-05, 6.8173e-07, 6.9998e-06, ..., 2.3689e-03, + 1.4820e-03, 1.5154e-03], + [ 1.1843e-04, 3.2842e-05, 1.0931e-04, ..., -2.1019e-03, + 4.1866e-04, 1.8435e-03], + [-2.1994e-04, 1.2779e-04, -1.1027e-04, ..., -8.9569e-03, + -9.5224e-04, 8.7214e-04], + ..., + [ 1.5810e-05, 3.4124e-06, -2.5916e-04, ..., -9.2621e-03, + 6.9320e-05, -1.5930e-02], + [-2.6837e-05, -4.1080e-04, 3.0547e-05, ..., 2.6474e-03, + 1.1075e-04, 2.5139e-03], + [ 5.5581e-05, 9.6858e-05, 1.0186e-04, ..., 3.1586e-03, + 1.5283e-04, 2.1324e-03]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0222, 0.0290, 0.0034, -0.0096, 0.0219, -0.0285, 0.0095, -0.0053, + -0.0038, -0.0111], device='cuda:0'), grad: tensor([ 0.0166, -0.0125, -0.0198, 0.0060, -0.0159, 0.0150, 0.0354, -0.0244, + 0.0064, -0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 214.32, cls_loss 0.5902 cls_loss_mapping 0.0070 cls_loss_causal 0.5095 re_mapping 0.0124 re_causal 0.0283 /// teacc 98.58 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0748, -0.0947, -0.0823, ..., -0.0302, 0.0489, -0.0890], + [-0.0604, -0.0770, -0.0534, ..., 0.0964, -0.0318, 0.1827], + [ 0.0157, -0.0189, -0.0291, ..., -0.0118, -0.0104, -0.0617], + ..., + [ 0.0228, -0.0818, 0.1388, ..., 0.0239, -0.0679, 0.0400], + [-0.0349, 0.0479, -0.1276, ..., -0.0522, 0.0091, -0.0727], + [-0.0544, 0.0662, 0.0519, ..., -0.0474, -0.0383, -0.0159]], + device='cuda:0'), grad: tensor([[ 1.7095e-04, 3.4618e-04, 2.9182e-04, ..., -5.7030e-03, + 2.8872e-04, 7.9679e-04], + [ 2.3341e-04, -2.9621e-03, 3.0112e-04, ..., -1.0628e-02, + -5.2786e-04, -2.3861e-03], + [ 7.1621e-04, -1.0836e-04, 1.1816e-03, ..., 5.9967e-03, + -4.5323e-04, 2.5883e-03], + ..., + [ 1.8158e-03, 8.3065e-04, 3.5934e-03, ..., 8.4305e-03, + 7.3671e-05, 3.1776e-03], + [ 3.0499e-03, 3.8891e-03, 3.0403e-03, ..., 9.3994e-03, + 4.3440e-04, 5.3253e-03], + [-3.2997e-03, 1.6823e-03, -8.7967e-03, ..., -7.8201e-04, + 7.7963e-05, -2.8000e-03]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0227, 0.0287, 0.0042, -0.0113, 0.0205, -0.0284, 0.0102, -0.0039, + -0.0030, -0.0112], device='cuda:0'), grad: tensor([-0.0177, -0.0317, 0.0226, -0.0227, 0.0111, 0.0121, -0.0232, 0.0303, + 0.0264, -0.0073], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 179---------------------------------------------------- +epoch 179, time 230.73, cls_loss 0.5837 cls_loss_mapping 0.0056 cls_loss_causal 0.5102 re_mapping 0.0119 re_causal 0.0281 /// teacc 98.73 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0765, -0.0962, -0.0816, ..., -0.0295, 0.0479, -0.0876], + [-0.0613, -0.0756, -0.0535, ..., 0.0972, -0.0307, 0.1828], + [ 0.0163, -0.0199, -0.0296, ..., -0.0128, -0.0098, -0.0627], + ..., + [ 0.0225, -0.0827, 0.1389, ..., 0.0231, -0.0692, 0.0406], + [-0.0348, 0.0485, -0.1287, ..., -0.0531, 0.0082, -0.0740], + [-0.0523, 0.0657, 0.0524, ..., -0.0461, -0.0393, -0.0161]], + device='cuda:0'), grad: tensor([[-2.8133e-03, -2.1210e-03, -2.6093e-03, ..., -7.0915e-03, + 7.6115e-05, -5.5361e-04], + [ 1.2982e-04, 1.8525e-04, 2.8563e-04, ..., 7.2861e-03, + 4.1798e-06, 5.8975e-03], + [ 5.5599e-04, 5.7983e-04, 6.7043e-04, ..., -1.3132e-03, + 6.0871e-06, -8.0347e-04], + ..., + [-4.8816e-05, 4.0207e-03, 7.1716e-04, ..., 4.2801e-03, + 8.9034e-06, 4.2877e-03], + [ 2.5153e-04, 1.4458e-03, 7.3004e-04, ..., -6.8741e-03, + 5.9754e-05, -4.1733e-03], + [ 2.9850e-04, -7.7972e-03, -2.1286e-03, ..., -2.0866e-03, + 1.7807e-05, -6.8626e-03]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0221, 0.0295, 0.0038, -0.0112, 0.0203, -0.0285, 0.0101, -0.0048, + -0.0038, -0.0101], device='cuda:0'), grad: tensor([-0.0230, 0.0051, -0.0117, -0.0127, 0.0126, 0.0120, 0.0205, 0.0180, + -0.0142, -0.0065], device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 214.48, cls_loss 0.5651 cls_loss_mapping 0.0068 cls_loss_causal 0.4787 re_mapping 0.0121 re_causal 0.0259 /// teacc 98.71 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0794, -0.0949, -0.0819, ..., -0.0311, 0.0492, -0.0875], + [-0.0633, -0.0762, -0.0537, ..., 0.0971, -0.0308, 0.1835], + [ 0.0156, -0.0195, -0.0287, ..., -0.0135, -0.0112, -0.0628], + ..., + [ 0.0248, -0.0835, 0.1388, ..., 0.0243, -0.0695, 0.0406], + [-0.0353, 0.0495, -0.1302, ..., -0.0526, 0.0069, -0.0754], + [-0.0531, 0.0648, 0.0529, ..., -0.0459, -0.0395, -0.0164]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0003, -0.0010, ..., 0.0007, -0.0028, -0.0014], + [ 0.0004, -0.0032, 0.0004, ..., -0.0042, 0.0002, -0.0040], + [-0.0059, -0.0010, 0.0017, ..., -0.0042, 0.0007, -0.0004], + ..., + [ 0.0002, 0.0007, 0.0039, ..., 0.0060, 0.0006, 0.0038], + [ 0.0007, 0.0016, 0.0015, ..., 0.0144, 0.0003, 0.0047], + [ 0.0001, 0.0025, -0.0022, ..., -0.0136, -0.0025, 0.0014]], + device='cuda:0') +Epoch 182, bias, value: tensor([-0.0235, 0.0287, 0.0028, -0.0109, 0.0201, -0.0283, 0.0105, -0.0032, + -0.0033, -0.0101], device='cuda:0'), grad: tensor([ 0.0094, -0.0029, -0.0002, 0.0320, -0.0309, 0.0162, -0.0259, 0.0312, + 0.0414, -0.0703], device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 214.08, cls_loss 0.6250 cls_loss_mapping 0.0082 cls_loss_causal 0.5457 re_mapping 0.0115 re_causal 0.0255 /// teacc 98.64 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0799, -0.0950, -0.0820, ..., -0.0322, 0.0498, -0.0887], + [-0.0621, -0.0757, -0.0549, ..., 0.0969, -0.0296, 0.1845], + [ 0.0148, -0.0193, -0.0290, ..., -0.0135, -0.0117, -0.0633], + ..., + [ 0.0239, -0.0840, 0.1388, ..., 0.0247, -0.0695, 0.0404], + [-0.0354, 0.0492, -0.1306, ..., -0.0524, 0.0066, -0.0753], + [-0.0531, 0.0654, 0.0532, ..., -0.0459, -0.0395, -0.0164]], + device='cuda:0'), grad: tensor([[ 2.6035e-04, 6.7091e-04, 1.2140e-03, ..., 1.1765e-02, + 4.9925e-04, 2.9397e-04], + [ 5.3024e-04, -9.4986e-04, 1.6861e-03, ..., 7.3013e-03, + -6.0558e-04, -2.5436e-05], + [ 6.5470e-04, -1.3151e-03, 6.5079e-03, ..., 4.8218e-03, + -1.3351e-03, 9.4986e-04], + ..., + [ 3.7074e-04, -2.3537e-03, -1.8524e-02, ..., -1.5762e-02, + -4.4417e-04, -1.5926e-03], + [-1.7185e-03, 2.1591e-03, 3.0155e-03, ..., 5.6763e-03, + 1.0853e-03, -1.8561e-04], + [ 4.0078e-04, -2.8515e-03, 1.4944e-03, ..., 9.7942e-04, + -2.9335e-03, 6.6376e-04]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0244, 0.0284, 0.0026, -0.0105, 0.0194, -0.0276, 0.0104, -0.0031, + -0.0031, -0.0093], device='cuda:0'), grad: tensor([ 0.0335, 0.0277, -0.0092, 0.0063, -0.0063, 0.0396, -0.0715, -0.0408, + 0.0171, 0.0035], device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 214.13, cls_loss 0.5779 cls_loss_mapping 0.0060 cls_loss_causal 0.4966 re_mapping 0.0120 re_causal 0.0266 /// teacc 98.66 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0784, -0.0950, -0.0817, ..., -0.0314, 0.0495, -0.0883], + [-0.0631, -0.0752, -0.0553, ..., 0.0954, -0.0287, 0.1850], + [ 0.0139, -0.0202, -0.0302, ..., -0.0133, -0.0111, -0.0639], + ..., + [ 0.0251, -0.0847, 0.1386, ..., 0.0257, -0.0685, 0.0411], + [-0.0358, 0.0489, -0.1311, ..., -0.0529, 0.0066, -0.0766], + [-0.0533, 0.0659, 0.0539, ..., -0.0453, -0.0403, -0.0168]], + device='cuda:0'), grad: tensor([[ 4.1425e-05, 8.4066e-04, 2.4855e-05, ..., -2.8419e-03, + 1.5423e-05, 5.8651e-05], + [ 3.3706e-05, 7.1764e-04, 1.9193e-05, ..., 6.4163e-03, + 6.7726e-06, 4.5076e-06], + [ 7.9803e-03, -1.0490e-03, 3.8958e-04, ..., -1.8219e-02, + 4.1437e-04, -7.0524e-04], + ..., + [ 8.0287e-05, 4.6110e-04, 1.2171e-04, ..., -2.2697e-03, + 8.3372e-06, 9.5725e-05], + [-1.0590e-02, -1.9112e-03, -1.3275e-03, ..., 6.0730e-03, + 6.4850e-05, 8.3566e-05], + [ 4.9543e-04, 1.4801e-03, 4.5514e-04, ..., 7.6752e-03, + 7.0155e-05, 9.3699e-05]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0238, 0.0281, 0.0026, -0.0109, 0.0211, -0.0286, 0.0106, -0.0033, + -0.0036, -0.0092], device='cuda:0'), grad: tensor([-1.6602e-02, 2.5787e-02, -5.0476e-02, 7.9651e-03, 4.2023e-02, + -7.8726e-04, -7.5579e-05, -6.3049e-02, -3.6411e-03, 5.8838e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 183---------------------------------------------------- +epoch 183, time 231.12, cls_loss 0.6019 cls_loss_mapping 0.0071 cls_loss_causal 0.5192 re_mapping 0.0124 re_causal 0.0285 /// teacc 98.84 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0764, -0.0957, -0.0819, ..., -0.0318, 0.0489, -0.0890], + [-0.0634, -0.0762, -0.0565, ..., 0.0956, -0.0292, 0.1856], + [ 0.0137, -0.0186, -0.0295, ..., -0.0131, -0.0108, -0.0641], + ..., + [ 0.0266, -0.0860, 0.1386, ..., 0.0259, -0.0689, 0.0400], + [-0.0362, 0.0500, -0.1309, ..., -0.0528, 0.0064, -0.0761], + [-0.0534, 0.0653, 0.0533, ..., -0.0462, -0.0405, -0.0168]], + device='cuda:0'), grad: tensor([[ 1.3745e-04, 1.2493e-04, 1.4555e-04, ..., 2.5806e-03, + -6.6459e-05, 3.7718e-04], + [ 2.4307e-04, 2.3022e-03, -2.5826e-03, ..., 1.2962e-02, + 4.9807e-06, -9.6273e-04], + [ 2.4581e-04, 1.0815e-03, 1.2484e-03, ..., 6.5155e-03, + 1.2308e-05, 1.2426e-03], + ..., + [ 5.9223e-04, 9.1028e-04, -6.5804e-04, ..., 8.4381e-03, + 4.9956e-06, -1.3828e-04], + [ 1.0389e-04, -1.3573e-02, 7.0286e-04, ..., -2.6169e-02, + -2.6166e-05, -5.2691e-04], + [ 5.2357e-04, 8.8692e-04, 3.8872e-03, ..., 7.3767e-04, + 8.9183e-06, 7.4310e-03]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0239, 0.0284, 0.0044, -0.0105, 0.0205, -0.0282, 0.0095, -0.0027, + -0.0043, -0.0102], device='cuda:0'), grad: tensor([ 0.0058, 0.0177, 0.0137, -0.0023, -0.0010, 0.0078, -0.0195, 0.0198, + -0.0494, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 214.84, cls_loss 0.5941 cls_loss_mapping 0.0061 cls_loss_causal 0.5142 re_mapping 0.0116 re_causal 0.0280 /// teacc 98.73 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0762, -0.0968, -0.0830, ..., -0.0308, 0.0485, -0.0904], + [-0.0626, -0.0751, -0.0553, ..., 0.0975, -0.0305, 0.1865], + [ 0.0135, -0.0198, -0.0304, ..., -0.0136, -0.0106, -0.0633], + ..., + [ 0.0265, -0.0871, 0.1385, ..., 0.0253, -0.0696, 0.0405], + [-0.0359, 0.0500, -0.1306, ..., -0.0519, 0.0067, -0.0766], + [-0.0532, 0.0651, 0.0534, ..., -0.0470, -0.0409, -0.0174]], + device='cuda:0'), grad: tensor([[ 2.6560e-04, 3.3882e-06, 2.2316e-04, ..., 4.6844e-03, + -2.3305e-05, 1.3914e-03], + [-6.0606e-04, 4.7714e-05, 4.7237e-05, ..., 3.1996e-04, + 9.9182e-05, 1.2712e-03], + [ 1.4019e-03, 1.2960e-03, 1.4448e-03, ..., -9.2087e-03, + -1.3971e-04, 2.2030e-03], + ..., + [ 4.5242e-03, -2.7428e-03, 5.5046e-03, ..., 6.8130e-03, + 3.4064e-05, 7.2098e-03], + [ 5.1595e-06, -4.5943e-04, 2.2268e-04, ..., 4.0741e-03, + 4.7237e-05, -1.5488e-03], + [ 2.1946e-04, 2.9206e-04, 8.5640e-04, ..., -7.6065e-03, + 1.1891e-05, -6.5880e-03]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0238, 0.0299, 0.0040, -0.0109, 0.0207, -0.0283, 0.0085, -0.0036, + -0.0035, -0.0102], device='cuda:0'), grad: tensor([ 0.0157, -0.0115, -0.0131, -0.0436, 0.0119, 0.0240, 0.0146, 0.0327, + -0.0222, -0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 214.83, cls_loss 0.6060 cls_loss_mapping 0.0070 cls_loss_causal 0.5332 re_mapping 0.0120 re_causal 0.0271 /// teacc 98.58 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0749, -0.0965, -0.0827, ..., -0.0302, 0.0485, -0.0903], + [-0.0620, -0.0728, -0.0549, ..., 0.0983, -0.0304, 0.1870], + [ 0.0130, -0.0204, -0.0308, ..., -0.0133, -0.0108, -0.0620], + ..., + [ 0.0263, -0.0920, 0.1371, ..., 0.0245, -0.0702, 0.0394], + [-0.0371, 0.0515, -0.1301, ..., -0.0515, 0.0076, -0.0755], + [-0.0523, 0.0674, 0.0540, ..., -0.0468, -0.0419, -0.0170]], + device='cuda:0'), grad: tensor([[ 2.9989e-07, -1.5745e-03, 2.2554e-04, ..., 3.9330e-03, + -2.8515e-03, 1.1749e-03], + [-1.8254e-06, 1.1568e-03, 4.2844e-04, ..., 1.1398e-02, + 1.7770e-06, 3.3150e-03], + [ 4.2506e-06, 9.8419e-04, 1.0290e-03, ..., 6.1760e-03, + 1.1873e-04, 1.7710e-03], + ..., + [ 3.5465e-06, -2.2106e-03, -3.3760e-03, ..., -2.1362e-03, + 4.5872e-04, -1.9293e-03], + [-1.4551e-05, 1.3132e-03, 3.9840e-04, ..., 5.1765e-03, + 1.2994e-04, 1.2455e-03], + [ 1.2908e-06, -3.2768e-03, 1.2054e-03, ..., -3.7098e-03, + 1.8620e-04, -4.4763e-05]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0232, 0.0310, 0.0038, -0.0113, 0.0190, -0.0279, 0.0094, -0.0044, + -0.0034, -0.0099], device='cuda:0'), grad: tensor([ 0.0051, 0.0322, 0.0191, -0.0137, 0.0163, -0.0195, -0.0435, 0.0018, + 0.0160, -0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 215.19, cls_loss 0.5923 cls_loss_mapping 0.0070 cls_loss_causal 0.5163 re_mapping 0.0122 re_causal 0.0268 /// teacc 98.50 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0757, -0.0978, -0.0833, ..., -0.0299, 0.0487, -0.0903], + [-0.0621, -0.0740, -0.0563, ..., 0.0990, -0.0306, 0.1863], + [ 0.0128, -0.0199, -0.0309, ..., -0.0132, -0.0118, -0.0623], + ..., + [ 0.0262, -0.0904, 0.1379, ..., 0.0246, -0.0702, 0.0398], + [-0.0374, 0.0519, -0.1291, ..., -0.0519, 0.0077, -0.0745], + [-0.0513, 0.0665, 0.0527, ..., -0.0475, -0.0420, -0.0160]], + device='cuda:0'), grad: tensor([[ 8.4460e-05, -6.0463e-04, -1.6966e-03, ..., -5.6534e-03, + 2.0135e-06, -2.8229e-03], + [ 7.0095e-04, 1.2875e-03, 1.2846e-03, ..., 7.5531e-03, + 2.2352e-07, 1.1663e-03], + [ 5.6362e-04, 1.8883e-03, 2.9302e-04, ..., 3.9825e-03, + -1.3602e-04, 7.1716e-04], + ..., + [-1.2598e-03, 3.5362e-03, -1.0691e-03, ..., -1.0208e-02, + 1.1139e-06, 1.0481e-03], + [ 1.7319e-03, -1.1948e-02, -5.7411e-04, ..., -1.4122e-02, + 2.4512e-06, -2.6550e-03], + [ 1.6177e-04, -5.1918e-03, 5.8126e-04, ..., 4.2381e-03, + 5.1409e-07, 1.8203e-04]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0230, 0.0315, 0.0034, -0.0124, 0.0189, -0.0273, 0.0087, -0.0042, + -0.0023, -0.0103], device='cuda:0'), grad: tensor([-0.0173, 0.0199, 0.0173, -0.0139, -0.0011, -0.0410, 0.0676, -0.0142, + -0.0228, 0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 214.22, cls_loss 0.6344 cls_loss_mapping 0.0087 cls_loss_causal 0.5603 re_mapping 0.0117 re_causal 0.0258 /// teacc 98.53 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0754, -0.0983, -0.0827, ..., -0.0294, 0.0486, -0.0906], + [-0.0621, -0.0742, -0.0579, ..., 0.0999, -0.0309, 0.1877], + [ 0.0127, -0.0198, -0.0310, ..., -0.0131, -0.0111, -0.0636], + ..., + [ 0.0260, -0.0907, 0.1385, ..., 0.0234, -0.0715, 0.0401], + [-0.0374, 0.0513, -0.1295, ..., -0.0519, 0.0076, -0.0755], + [-0.0517, 0.0669, 0.0525, ..., -0.0479, -0.0408, -0.0160]], + device='cuda:0'), grad: tensor([[ 6.9022e-05, 6.4039e-04, 1.9102e-03, ..., -4.8351e-04, + 4.3586e-07, -2.4395e-03], + [ 1.0281e-03, 1.6466e-05, 1.4410e-03, ..., -1.6220e-02, + -1.7986e-03, 6.1226e-03], + [ 2.7156e-04, 3.3641e-04, -4.9686e-04, ..., 1.5480e-02, + 1.7586e-03, 4.5547e-03], + ..., + [-1.5373e-03, 5.6219e-04, 9.3603e-04, ..., -4.2915e-04, + 3.7730e-05, -4.2000e-03], + [-3.4046e-04, 4.9973e-04, 9.5034e-04, ..., 8.5373e-03, + 1.2897e-05, 3.1700e-03], + [ 2.5272e-04, 2.5120e-03, 9.1095e-03, ..., -1.0437e-02, + 1.6317e-06, -7.3433e-04]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0231, 0.0311, 0.0042, -0.0124, 0.0195, -0.0270, 0.0083, -0.0042, + -0.0027, -0.0106], device='cuda:0'), grad: tensor([-0.0373, -0.0473, 0.0389, 0.0157, -0.0068, -0.0056, -0.0161, 0.0146, + 0.0273, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 214.87, cls_loss 0.5974 cls_loss_mapping 0.0081 cls_loss_causal 0.5161 re_mapping 0.0122 re_causal 0.0269 /// teacc 98.74 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0744, -0.0975, -0.0812, ..., -0.0294, 0.0483, -0.0909], + [-0.0623, -0.0746, -0.0589, ..., 0.0985, -0.0314, 0.1869], + [ 0.0129, -0.0203, -0.0294, ..., -0.0130, -0.0113, -0.0625], + ..., + [ 0.0258, -0.0902, 0.1390, ..., 0.0241, -0.0714, 0.0419], + [-0.0373, 0.0516, -0.1304, ..., -0.0516, 0.0074, -0.0763], + [-0.0516, 0.0680, 0.0513, ..., -0.0471, -0.0403, -0.0166]], + device='cuda:0'), grad: tensor([[ 4.8652e-06, 1.6336e-03, 1.7080e-03, ..., 6.1913e-03, + -3.8184e-07, 1.0443e-03], + [ 7.6413e-05, 3.1161e-04, 1.2894e-03, ..., 2.1152e-03, + 0.0000e+00, -1.9875e-03], + [ 2.8610e-05, 5.9557e-04, 1.3990e-03, ..., 6.5613e-03, + 2.5332e-07, 7.3242e-04], + ..., + [ 1.9491e-05, 1.8387e-03, 1.1120e-03, ..., 6.9695e-03, + 0.0000e+00, 5.8699e-04], + [ 1.4327e-05, 1.8120e-03, 1.6022e-03, ..., 6.4507e-03, + 7.4506e-09, 1.3876e-03], + [ 5.9068e-05, -4.1542e-03, -1.6584e-03, ..., -4.8981e-03, + 8.5682e-08, 1.4048e-03]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0228, 0.0294, 0.0046, -0.0122, 0.0203, -0.0282, 0.0092, -0.0039, + -0.0026, -0.0106], device='cuda:0'), grad: tensor([ 0.0177, 0.0124, 0.0197, -0.0054, 0.0110, -0.0634, -0.0002, 0.0219, + 0.0195, -0.0332], device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 215.01, cls_loss 0.6058 cls_loss_mapping 0.0064 cls_loss_causal 0.5231 re_mapping 0.0121 re_causal 0.0275 /// teacc 98.69 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0755, -0.0977, -0.0810, ..., -0.0304, 0.0496, -0.0903], + [-0.0619, -0.0762, -0.0581, ..., 0.0999, -0.0317, 0.1887], + [ 0.0130, -0.0199, -0.0298, ..., -0.0146, -0.0107, -0.0627], + ..., + [ 0.0257, -0.0908, 0.1392, ..., 0.0246, -0.0731, 0.0406], + [-0.0363, 0.0519, -0.1294, ..., -0.0501, 0.0076, -0.0767], + [-0.0514, 0.0681, 0.0504, ..., -0.0473, -0.0414, -0.0165]], + device='cuda:0'), grad: tensor([[ 3.4738e-04, 7.5769e-04, 6.7425e-04, ..., 7.2899e-03, + 1.8346e-04, 1.2636e-03], + [-2.7084e-04, 8.5592e-05, -8.0204e-04, ..., -4.9019e-03, + 4.8727e-05, -5.6219e-04], + [ 6.1560e-04, -1.3745e-04, 2.1381e-03, ..., -2.4967e-03, + 2.0191e-05, -2.0657e-03], + ..., + [-2.7466e-03, -6.0272e-03, -1.1749e-03, ..., -1.9592e-02, + 4.2804e-06, -9.3937e-04], + [ 1.1101e-03, 3.6001e-04, 3.4809e-04, ..., -8.6117e-04, + -9.8896e-04, -2.6727e-04], + [ 1.0996e-03, 1.6012e-03, 2.2373e-03, ..., 1.2199e-02, + 4.9174e-06, 3.0041e-03]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0237, 0.0304, 0.0033, -0.0128, 0.0212, -0.0282, 0.0097, -0.0044, + -0.0021, -0.0104], device='cuda:0'), grad: tensor([ 0.0265, -0.0137, -0.0111, 0.0142, 0.0386, 0.0018, -0.0340, -0.0567, + -0.0079, 0.0422], device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 214.72, cls_loss 0.5986 cls_loss_mapping 0.0057 cls_loss_causal 0.5182 re_mapping 0.0120 re_causal 0.0271 /// teacc 98.76 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0763, -0.0974, -0.0808, ..., -0.0308, 0.0499, -0.0896], + [-0.0629, -0.0759, -0.0591, ..., 0.0997, -0.0320, 0.1881], + [ 0.0136, -0.0202, -0.0319, ..., -0.0138, -0.0114, -0.0627], + ..., + [ 0.0269, -0.0914, 0.1405, ..., 0.0245, -0.0729, 0.0409], + [-0.0364, 0.0523, -0.1297, ..., -0.0507, 0.0097, -0.0762], + [-0.0525, 0.0679, 0.0500, ..., -0.0466, -0.0423, -0.0169]], + device='cuda:0'), grad: tensor([[ 2.5234e-03, -1.1168e-03, 1.3046e-03, ..., 1.5541e-02, + -2.9049e-03, 8.3637e-04], + [ 1.3857e-03, 6.8210e-06, 3.8242e-04, ..., 8.6212e-03, + 1.7869e-04, 5.0354e-04], + [ 1.6670e-03, 4.2140e-05, 3.7575e-04, ..., 1.3065e-03, + 5.6839e-04, 3.0923e-04], + ..., + [ 1.1654e-03, 1.2271e-05, 3.8314e-04, ..., 2.4624e-03, + 1.9670e-04, 1.1158e-03], + [ 1.1206e-03, 2.3212e-03, 6.3705e-04, ..., -8.8739e-04, + 4.7226e-03, 1.0185e-03], + [ 1.1940e-03, 4.6611e-05, 5.8794e-04, ..., 7.8583e-03, + 8.4162e-04, 6.4135e-04]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0235, 0.0303, 0.0032, -0.0128, 0.0210, -0.0279, 0.0091, -0.0044, + -0.0022, -0.0099], device='cuda:0'), grad: tensor([ 0.0451, 0.0278, 0.0157, -0.0543, 0.0185, -0.0660, -0.0187, 0.0002, + 0.0022, 0.0294], device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 214.60, cls_loss 0.5947 cls_loss_mapping 0.0047 cls_loss_causal 0.5228 re_mapping 0.0119 re_causal 0.0267 /// teacc 98.70 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0771, -0.0985, -0.0805, ..., -0.0313, 0.0494, -0.0902], + [-0.0629, -0.0764, -0.0586, ..., 0.1000, -0.0309, 0.1878], + [ 0.0128, -0.0177, -0.0315, ..., -0.0144, -0.0112, -0.0627], + ..., + [ 0.0278, -0.0917, 0.1407, ..., 0.0248, -0.0734, 0.0410], + [-0.0362, 0.0522, -0.1316, ..., -0.0507, 0.0098, -0.0768], + [-0.0530, 0.0679, 0.0497, ..., -0.0465, -0.0413, -0.0171]], + device='cuda:0'), grad: tensor([[ 0.0011, 0.0034, 0.0127, ..., 0.0018, 0.0109, 0.0007], + [ 0.0004, 0.0003, -0.0029, ..., -0.0033, 0.0008, -0.0038], + [ 0.0009, 0.0014, 0.0019, ..., -0.0071, 0.0085, -0.0017], + ..., + [ 0.0013, 0.0003, 0.0043, ..., 0.0063, 0.0021, 0.0010], + [ 0.0003, -0.0002, -0.0171, ..., 0.0063, -0.0282, 0.0011], + [-0.0045, 0.0016, -0.0030, ..., -0.0167, 0.0005, 0.0013]], + device='cuda:0') +Epoch 193, bias, value: tensor([-0.0245, 0.0312, 0.0030, -0.0136, 0.0217, -0.0271, 0.0085, -0.0037, + -0.0025, -0.0100], device='cuda:0'), grad: tensor([ 0.0062, 0.0010, -0.0162, 0.0259, 0.0101, -0.0145, 0.0149, 0.0221, + -0.0084, -0.0412], device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 214.81, cls_loss 0.5940 cls_loss_mapping 0.0082 cls_loss_causal 0.5169 re_mapping 0.0118 re_causal 0.0262 /// teacc 98.65 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0765, -0.0991, -0.0807, ..., -0.0320, 0.0500, -0.0902], + [-0.0639, -0.0754, -0.0567, ..., 0.1003, -0.0316, 0.1884], + [ 0.0130, -0.0180, -0.0320, ..., -0.0134, -0.0108, -0.0626], + ..., + [ 0.0275, -0.0929, 0.1406, ..., 0.0249, -0.0731, 0.0405], + [-0.0353, 0.0526, -0.1319, ..., -0.0517, 0.0105, -0.0772], + [-0.0535, 0.0689, 0.0505, ..., -0.0452, -0.0413, -0.0165]], + device='cuda:0'), grad: tensor([[ 4.0293e-05, 9.2125e-04, 5.1117e-04, ..., 5.6992e-03, + 8.4281e-05, 4.5013e-04], + [ 3.1255e-06, 9.8801e-04, 1.2529e-04, ..., -1.1375e-02, + 9.1970e-05, -3.5458e-03], + [ 4.2409e-05, -4.0207e-03, 5.1308e-04, ..., -6.0081e-03, + -1.8489e-04, 1.3030e-04], + ..., + [ 1.5736e-05, 5.2166e-04, 1.5664e-04, ..., 3.2711e-03, + 7.9095e-05, 1.3089e-04], + [ 8.3029e-05, 2.4078e-02, 5.5809e-03, ..., -2.0866e-03, + 1.6708e-03, 5.3596e-04], + [ 5.3585e-05, 1.1520e-03, 4.7565e-04, ..., -4.7386e-05, + 2.1660e-04, 5.5695e-04]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0249, 0.0310, 0.0036, -0.0123, 0.0211, -0.0274, 0.0081, -0.0042, + -0.0027, -0.0094], device='cuda:0'), grad: tensor([ 0.0174, -0.0395, -0.0147, -0.0326, 0.0148, 0.0301, 0.0200, 0.0122, + 0.0067, -0.0144], device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.70, cls_loss 0.6051 cls_loss_mapping 0.0073 cls_loss_causal 0.5268 re_mapping 0.0113 re_causal 0.0258 /// teacc 98.65 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0763, -0.0992, -0.0805, ..., -0.0323, 0.0501, -0.0898], + [-0.0648, -0.0758, -0.0569, ..., 0.1006, -0.0319, 0.1880], + [ 0.0128, -0.0188, -0.0329, ..., -0.0137, -0.0106, -0.0629], + ..., + [ 0.0263, -0.0927, 0.1420, ..., 0.0253, -0.0726, 0.0403], + [-0.0340, 0.0532, -0.1295, ..., -0.0509, 0.0114, -0.0755], + [-0.0544, 0.0683, 0.0493, ..., -0.0464, -0.0426, -0.0166]], + device='cuda:0'), grad: tensor([[ 6.7234e-05, 8.7261e-05, -1.4000e-03, ..., -2.3212e-03, + 3.3522e-04, 1.0176e-03], + [ 3.7730e-05, 6.2883e-05, 1.8644e-03, ..., -6.8970e-03, + -2.9907e-05, 4.7803e-04], + [-6.1083e-04, -1.2245e-03, -4.0550e-03, ..., -5.9738e-03, + 2.9540e-04, 5.0879e-04], + ..., + [ 2.4259e-05, 2.6122e-05, 8.2703e-03, ..., 5.6801e-03, + 1.6670e-03, -1.4973e-03], + [-1.6584e-03, -2.1610e-03, -8.9722e-03, ..., -7.2250e-03, + -3.6983e-03, 8.5211e-04], + [ 5.6177e-05, 6.5565e-05, -8.4152e-03, ..., -4.7226e-03, + -1.0328e-03, -6.2656e-04]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0247, 0.0309, 0.0032, -0.0113, 0.0214, -0.0280, 0.0078, -0.0032, + -0.0022, -0.0110], device='cuda:0'), grad: tensor([-0.0157, -0.0148, -0.0049, 0.0463, 0.0321, -0.0129, -0.0112, 0.0315, + -0.0176, -0.0328], device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 214.86, cls_loss 0.5630 cls_loss_mapping 0.0058 cls_loss_causal 0.4843 re_mapping 0.0121 re_causal 0.0268 /// teacc 98.70 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0765, -0.0990, -0.0810, ..., -0.0323, 0.0501, -0.0910], + [-0.0648, -0.0756, -0.0583, ..., 0.1007, -0.0333, 0.1884], + [ 0.0111, -0.0184, -0.0314, ..., -0.0135, -0.0102, -0.0633], + ..., + [ 0.0264, -0.0929, 0.1416, ..., 0.0259, -0.0746, 0.0411], + [-0.0341, 0.0530, -0.1275, ..., -0.0514, 0.0117, -0.0759], + [-0.0540, 0.0687, 0.0488, ..., -0.0471, -0.0405, -0.0170]], + device='cuda:0'), grad: tensor([[-4.4823e-03, 4.2886e-05, 5.7077e-04, ..., -9.0027e-03, + 3.8333e-06, -6.3019e-03], + [ 3.1233e-04, 6.7838e-06, 6.1321e-04, ..., 3.1261e-03, + 4.0792e-06, 5.7888e-04], + [ 4.5705e-04, 2.3377e-04, -1.2112e-04, ..., -5.7335e-03, + 2.5928e-06, 7.3290e-04], + ..., + [ 1.9550e-03, 3.4642e-04, 1.5745e-03, ..., 5.0812e-03, + 5.3316e-05, 1.7810e-04], + [-6.6280e-04, -4.8339e-05, 1.0471e-03, ..., 3.2005e-03, + 1.3506e-04, 3.7766e-03], + [-1.9760e-03, -6.2418e-04, -4.4174e-03, ..., -6.1455e-03, + -2.3592e-04, -5.3787e-03]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0247, 0.0308, 0.0038, -0.0117, 0.0206, -0.0281, 0.0088, -0.0033, + -0.0023, -0.0109], device='cuda:0'), grad: tensor([-0.0079, 0.0103, -0.0205, 0.0092, 0.0154, -0.0033, -0.0125, 0.0180, + 0.0389, -0.0476], device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 214.72, cls_loss 0.5883 cls_loss_mapping 0.0055 cls_loss_causal 0.5160 re_mapping 0.0112 re_causal 0.0256 /// teacc 98.55 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0781, -0.0995, -0.0820, ..., -0.0320, 0.0511, -0.0904], + [-0.0656, -0.0762, -0.0568, ..., 0.1015, -0.0336, 0.1892], + [ 0.0104, -0.0190, -0.0320, ..., -0.0142, -0.0113, -0.0638], + ..., + [ 0.0254, -0.0939, 0.1408, ..., 0.0259, -0.0756, 0.0414], + [-0.0332, 0.0538, -0.1283, ..., -0.0514, 0.0111, -0.0761], + [-0.0545, 0.0692, 0.0477, ..., -0.0477, -0.0413, -0.0177]], + device='cuda:0'), grad: tensor([[-3.0828e-04, 1.1206e-04, -3.0518e-04, ..., 2.6722e-03, + -7.1831e-03, 2.2888e-04], + [-1.3466e-03, -8.8739e-04, -5.9280e-03, ..., -8.0795e-03, + -1.5411e-03, -2.3594e-03], + [ 2.8992e-04, 1.1426e-04, -9.0170e-04, ..., -1.2825e-02, + 7.1859e-04, 5.7364e-04], + ..., + [ 5.0020e-04, 1.6141e-04, 1.8673e-03, ..., 6.9695e-03, + 1.0862e-03, -2.6684e-03], + [ 6.3598e-05, 7.4327e-05, 1.4868e-03, ..., 5.2299e-03, + 1.4648e-03, 1.5364e-03], + [ 2.0421e-04, 4.4256e-05, 1.5297e-03, ..., -3.2158e-03, + 9.9754e-04, 1.8129e-03]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0240, 0.0315, 0.0029, -0.0120, 0.0206, -0.0281, 0.0089, -0.0032, + -0.0018, -0.0120], device='cuda:0'), grad: tensor([-0.0118, -0.0182, -0.0383, 0.0258, 0.0022, -0.0062, 0.0185, 0.0003, + 0.0298, -0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 214.88, cls_loss 0.6026 cls_loss_mapping 0.0044 cls_loss_causal 0.5242 re_mapping 0.0119 re_causal 0.0274 /// teacc 98.58 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0781, -0.1003, -0.0827, ..., -0.0322, 0.0506, -0.0904], + [-0.0659, -0.0752, -0.0568, ..., 0.1014, -0.0335, 0.1899], + [ 0.0098, -0.0192, -0.0321, ..., -0.0139, -0.0113, -0.0646], + ..., + [ 0.0250, -0.0934, 0.1410, ..., 0.0254, -0.0754, 0.0414], + [-0.0343, 0.0532, -0.1278, ..., -0.0520, 0.0114, -0.0771], + [-0.0541, 0.0695, 0.0478, ..., -0.0485, -0.0426, -0.0173]], + device='cuda:0'), grad: tensor([[ 8.9288e-05, 2.7001e-05, 8.6784e-04, ..., -8.2731e-04, + 5.9223e-04, 9.6178e-04], + [ 1.2708e-04, -6.1035e-04, -2.0638e-03, ..., -5.7373e-03, + -9.1400e-03, 1.0166e-03], + [ 1.2231e-04, 4.1991e-05, 1.2398e-03, ..., 9.3231e-03, + 7.2021e-03, 1.1778e-03], + ..., + [ 9.2745e-05, 4.5002e-05, 7.8678e-04, ..., 6.9351e-03, + 3.4285e-04, 1.1034e-03], + [ 1.0371e-04, 2.9397e-04, -1.1435e-03, ..., -4.8447e-03, + 3.8099e-04, 1.4210e-04], + [-9.6130e-04, 4.2677e-05, 6.3753e-04, ..., -4.5090e-03, + 2.4414e-04, 1.0071e-03]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0238, 0.0313, 0.0031, -0.0116, 0.0220, -0.0278, 0.0084, -0.0034, + -0.0018, -0.0134], device='cuda:0'), grad: tensor([-0.0056, -0.0483, 0.0606, -0.0031, -0.0047, 0.0021, 0.0328, 0.0241, + -0.0458, -0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 214.70, cls_loss 0.5718 cls_loss_mapping 0.0060 cls_loss_causal 0.4911 re_mapping 0.0118 re_causal 0.0263 /// teacc 98.72 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0773, -0.1009, -0.0829, ..., -0.0315, 0.0499, -0.0912], + [-0.0657, -0.0746, -0.0571, ..., 0.1004, -0.0327, 0.1904], + [ 0.0092, -0.0200, -0.0322, ..., -0.0130, -0.0118, -0.0634], + ..., + [ 0.0263, -0.0940, 0.1408, ..., 0.0255, -0.0757, 0.0416], + [-0.0342, 0.0534, -0.1274, ..., -0.0531, 0.0114, -0.0777], + [-0.0554, 0.0710, 0.0480, ..., -0.0474, -0.0426, -0.0178]], + device='cuda:0'), grad: tensor([[ 1.7452e-04, 3.9148e-04, -3.3116e-04, ..., -4.0436e-03, + -1.3323e-03, 6.7282e-04], + [ 1.2010e-04, 2.3472e-04, 7.0763e-04, ..., 2.1343e-03, + 5.0426e-05, 5.4216e-04], + [ 7.3791e-05, 7.5626e-04, 8.6832e-04, ..., 2.4147e-03, + 5.7650e-04, 3.7718e-04], + ..., + [ 3.5048e-04, 1.1730e-03, -3.4370e-03, ..., -1.1375e-02, + 1.0175e-04, -3.4561e-03], + [ 3.1567e-04, -1.6193e-03, 1.1005e-03, ..., 2.2316e-03, + -1.1206e-05, -1.9989e-03], + [ 2.2840e-04, 6.1750e-04, 1.1120e-03, ..., 4.9400e-03, + -3.4666e-04, 7.7515e-03]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0232, 0.0304, 0.0033, -0.0115, 0.0208, -0.0270, 0.0088, -0.0033, + -0.0027, -0.0125], device='cuda:0'), grad: tensor([-0.0240, 0.0076, 0.0107, 0.0102, -0.0082, 0.0064, 0.0121, -0.0282, + 0.0013, 0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 214.35, cls_loss 0.5603 cls_loss_mapping 0.0063 cls_loss_causal 0.4761 re_mapping 0.0119 re_causal 0.0268 /// teacc 98.53 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0765, -0.1026, -0.0835, ..., -0.0326, 0.0506, -0.0905], + [-0.0656, -0.0747, -0.0567, ..., 0.1002, -0.0334, 0.1905], + [ 0.0078, -0.0196, -0.0313, ..., -0.0136, -0.0116, -0.0622], + ..., + [ 0.0252, -0.0940, 0.1406, ..., 0.0272, -0.0755, 0.0415], + [-0.0331, 0.0545, -0.1273, ..., -0.0534, 0.0117, -0.0790], + [-0.0553, 0.0706, 0.0476, ..., -0.0468, -0.0435, -0.0187]], + device='cuda:0'), grad: tensor([[ 1.0058e-07, 3.7670e-05, 5.8031e-04, ..., 2.0771e-03, + 1.3947e-04, 6.8760e-04], + [ 1.5367e-07, 1.4889e-04, 1.1425e-03, ..., 4.1962e-03, + 2.0862e-04, 1.1406e-03], + [ 7.3947e-07, 3.0708e-04, 1.3533e-03, ..., 2.5444e-03, + 1.5628e-04, 9.7275e-04], + ..., + [ 2.6636e-07, 9.5463e-04, 1.6403e-02, ..., 1.1261e-02, + 3.7074e-04, 7.3395e-03], + [ 6.6590e-07, 2.3117e-03, 2.0065e-03, ..., 2.0809e-03, + 1.2046e-04, 1.4467e-03], + [ 1.6494e-06, -8.6308e-04, -1.6464e-02, ..., 2.2640e-03, + 2.6608e-04, -3.6068e-03]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0234, 0.0300, 0.0025, -0.0112, 0.0221, -0.0281, 0.0089, -0.0018, + -0.0035, -0.0125], device='cuda:0'), grad: tensor([ 0.0158, 0.0258, 0.0127, -0.0200, -0.0500, 0.0110, -0.0157, 0.0565, + -0.0107, -0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 214.28, cls_loss 0.5897 cls_loss_mapping 0.0053 cls_loss_causal 0.5166 re_mapping 0.0117 re_causal 0.0274 /// teacc 98.56 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0751, -0.1030, -0.0840, ..., -0.0318, 0.0506, -0.0899], + [-0.0660, -0.0758, -0.0564, ..., 0.1008, -0.0335, 0.1908], + [ 0.0074, -0.0189, -0.0316, ..., -0.0136, -0.0102, -0.0615], + ..., + [ 0.0257, -0.0942, 0.1403, ..., 0.0270, -0.0770, 0.0422], + [-0.0337, 0.0547, -0.1275, ..., -0.0547, 0.0115, -0.0785], + [-0.0559, 0.0700, 0.0476, ..., -0.0471, -0.0436, -0.0192]], + device='cuda:0'), grad: tensor([[ 0.0001, 0.0005, -0.0004, ..., 0.0026, 0.0008, 0.0024], + [ 0.0007, 0.0004, -0.0031, ..., -0.0013, 0.0002, -0.0071], + [-0.0005, -0.0011, -0.0025, ..., 0.0044, -0.0024, 0.0022], + ..., + [ 0.0024, 0.0017, 0.0064, ..., 0.0201, 0.0007, 0.0072], + [ 0.0007, 0.0009, 0.0006, ..., -0.0316, 0.0009, -0.0024], + [-0.0054, -0.0035, -0.0076, ..., -0.0101, -0.0014, -0.0101]], + device='cuda:0') +Epoch 201, bias, value: tensor([-0.0228, 0.0306, 0.0025, -0.0118, 0.0215, -0.0267, 0.0084, -0.0024, + -0.0038, -0.0123], device='cuda:0'), grad: tensor([ 0.0174, -0.0272, 0.0086, 0.0007, 0.0043, 0.0008, 0.0355, 0.0552, + -0.0561, -0.0392], device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 213.90, cls_loss 0.5929 cls_loss_mapping 0.0074 cls_loss_causal 0.5219 re_mapping 0.0119 re_causal 0.0269 /// teacc 98.45 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0744, -0.1027, -0.0850, ..., -0.0322, 0.0507, -0.0903], + [-0.0663, -0.0767, -0.0577, ..., 0.1013, -0.0358, 0.1913], + [ 0.0072, -0.0190, -0.0319, ..., -0.0129, -0.0099, -0.0625], + ..., + [ 0.0248, -0.0940, 0.1400, ..., 0.0264, -0.0781, 0.0414], + [-0.0330, 0.0547, -0.1275, ..., -0.0540, 0.0110, -0.0777], + [-0.0547, 0.0702, 0.0479, ..., -0.0475, -0.0445, -0.0195]], + device='cuda:0'), grad: tensor([[ 1.6214e-06, -1.7524e-04, 5.2243e-05, ..., 3.3398e-03, + 6.0701e-04, 6.0129e-04], + [ 6.0290e-05, 5.4203e-07, 2.4092e-04, ..., -1.6479e-02, + -3.1376e-03, -4.2076e-03], + [ 1.3523e-05, 8.1539e-05, 9.5606e-05, ..., 3.8662e-03, + 4.2653e-04, 8.8835e-04], + ..., + [-1.1940e-03, 1.2694e-06, -4.5357e-03, ..., -1.1040e-02, + 4.6897e-04, -7.3776e-03], + [ 1.2442e-05, -5.4110e-07, 7.1347e-05, ..., 6.7215e-03, + 2.8057e-03, 1.4915e-03], + [ 8.8024e-04, 1.5676e-05, 3.3588e-03, ..., 7.9575e-03, + 2.3956e-03, 6.1569e-03]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0226, 0.0311, 0.0022, -0.0109, 0.0220, -0.0271, 0.0074, -0.0030, + -0.0030, -0.0129], device='cuda:0'), grad: tensor([ 0.0127, -0.0362, 0.0131, -0.0207, 0.0181, -0.0076, 0.0139, -0.0471, + 0.0220, 0.0318], device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.42, cls_loss 0.5849 cls_loss_mapping 0.0060 cls_loss_causal 0.5129 re_mapping 0.0116 re_causal 0.0269 /// teacc 98.54 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0742, -0.1026, -0.0845, ..., -0.0317, 0.0505, -0.0913], + [-0.0669, -0.0771, -0.0582, ..., 0.1010, -0.0367, 0.1904], + [ 0.0055, -0.0185, -0.0324, ..., -0.0126, -0.0098, -0.0646], + ..., + [ 0.0266, -0.0931, 0.1420, ..., 0.0264, -0.0758, 0.0428], + [-0.0333, 0.0545, -0.1272, ..., -0.0531, 0.0109, -0.0763], + [-0.0554, 0.0694, 0.0471, ..., -0.0476, -0.0439, -0.0182]], + device='cuda:0'), grad: tensor([[ 5.6885e-06, 4.5925e-05, 1.9467e-04, ..., 2.4471e-03, + 8.5592e-04, 3.5310e-04], + [-3.5024e-04, 1.8060e-05, 4.7264e-03, ..., 1.3504e-03, + 1.9062e-04, -1.7977e-03], + [ 3.1292e-05, 8.8871e-05, 3.2883e-03, ..., 6.5880e-03, + 2.1613e-04, 5.1498e-04], + ..., + [ 5.0455e-05, 1.3466e-03, -6.9008e-03, ..., -9.8419e-03, + 5.1451e-04, 2.3651e-03], + [ 2.8253e-04, -2.9099e-02, 1.2274e-03, ..., -1.4544e-03, + -1.5326e-03, 1.1768e-03], + [ 2.5064e-05, 2.6794e-02, -6.9284e-04, ..., -3.0842e-03, + 5.0259e-04, 9.4891e-04]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0229, 0.0308, 0.0027, -0.0109, 0.0215, -0.0277, 0.0069, -0.0025, + -0.0028, -0.0122], device='cuda:0'), grad: tensor([ 0.0183, -0.0063, 0.0189, 0.0147, -0.0334, 0.0099, 0.0068, 0.0090, + -0.0630, 0.0251], device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 214.77, cls_loss 0.5840 cls_loss_mapping 0.0067 cls_loss_causal 0.5131 re_mapping 0.0118 re_causal 0.0267 /// teacc 98.67 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0744, -0.1036, -0.0845, ..., -0.0317, 0.0505, -0.0912], + [-0.0676, -0.0773, -0.0590, ..., 0.1013, -0.0373, 0.1902], + [ 0.0061, -0.0183, -0.0316, ..., -0.0140, -0.0088, -0.0646], + ..., + [ 0.0247, -0.0937, 0.1419, ..., 0.0267, -0.0748, 0.0428], + [-0.0323, 0.0538, -0.1287, ..., -0.0545, 0.0100, -0.0772], + [-0.0563, 0.0691, 0.0473, ..., -0.0473, -0.0437, -0.0182]], + device='cuda:0'), grad: tensor([[-1.5039e-03, 8.5950e-05, 1.0967e-03, ..., -8.9722e-03, + 4.0131e-03, -1.9970e-03], + [ 5.4926e-05, 2.9117e-05, 8.3685e-04, ..., 1.8940e-03, + 1.2436e-03, 6.7294e-05], + [ 2.6298e-04, 1.1140e-04, 8.9359e-04, ..., 3.2005e-03, + 1.2083e-03, 2.8944e-04], + ..., + [ 2.5725e-04, 1.1891e-04, 1.6279e-03, ..., 3.6278e-03, + 2.3880e-03, 4.5705e-04], + [ 2.7418e-04, 1.2560e-03, 3.2120e-03, ..., 3.8128e-03, + 3.6888e-03, 5.8317e-04], + [ 3.5405e-05, 2.6846e-04, -6.5880e-03, ..., 3.5782e-03, + -2.0676e-02, -3.4547e-04]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0230, 0.0304, 0.0018, -0.0109, 0.0216, -0.0271, 0.0080, -0.0022, + -0.0036, -0.0121], device='cuda:0'), grad: tensor([-0.0125, 0.0167, 0.0217, -0.0207, -0.0407, 0.0055, -0.0113, 0.0238, + 0.0274, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.45, cls_loss 0.5750 cls_loss_mapping 0.0052 cls_loss_causal 0.4986 re_mapping 0.0116 re_causal 0.0266 /// teacc 98.48 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0734, -0.1043, -0.0832, ..., -0.0314, 0.0496, -0.0913], + [-0.0673, -0.0770, -0.0589, ..., 0.1017, -0.0361, 0.1912], + [ 0.0069, -0.0184, -0.0319, ..., -0.0141, -0.0086, -0.0656], + ..., + [ 0.0247, -0.0938, 0.1411, ..., 0.0264, -0.0736, 0.0425], + [-0.0325, 0.0540, -0.1291, ..., -0.0537, 0.0099, -0.0765], + [-0.0570, 0.0688, 0.0482, ..., -0.0468, -0.0427, -0.0183]], + device='cuda:0'), grad: tensor([[ 2.4006e-05, 4.0197e-04, 1.6737e-04, ..., 2.2411e-03, + 1.0455e-04, 2.3782e-05], + [ 1.7537e-06, -2.7466e-04, 7.0035e-05, ..., -1.0765e-02, + -1.0252e-03, -2.8610e-04], + [ 9.6560e-05, 5.4073e-04, 5.6410e-04, ..., -1.5402e-03, + 7.7248e-05, 1.0175e-04], + ..., + [ 8.3297e-06, 6.4564e-04, 6.1464e-04, ..., 3.5248e-03, + 6.5506e-05, 6.2823e-05], + [ 3.0309e-05, 9.4223e-04, 7.4863e-04, ..., 3.7212e-03, + 2.0123e-04, 1.8585e-04], + [ 9.3058e-06, 3.5119e-04, 1.4334e-03, ..., -7.2300e-05, + -5.6314e-04, 2.7704e-04]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0240, 0.0306, 0.0026, -0.0110, 0.0214, -0.0279, 0.0080, -0.0026, + -0.0026, -0.0116], device='cuda:0'), grad: tensor([ 0.0122, -0.0504, -0.0175, 0.0094, 0.0135, 0.0145, 0.0016, 0.0192, + -0.0119, 0.0094], device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 214.45, cls_loss 0.5822 cls_loss_mapping 0.0054 cls_loss_causal 0.5048 re_mapping 0.0114 re_causal 0.0264 /// teacc 98.57 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0747, -0.1044, -0.0828, ..., -0.0320, 0.0497, -0.0923], + [-0.0664, -0.0772, -0.0591, ..., 0.1020, -0.0359, 0.1914], + [ 0.0058, -0.0181, -0.0309, ..., -0.0139, -0.0087, -0.0667], + ..., + [ 0.0252, -0.0929, 0.1399, ..., 0.0265, -0.0745, 0.0434], + [-0.0322, 0.0543, -0.1292, ..., -0.0531, 0.0097, -0.0774], + [-0.0571, 0.0687, 0.0489, ..., -0.0484, -0.0424, -0.0187]], + device='cuda:0'), grad: tensor([[ 2.6202e-04, 4.7423e-06, 6.1750e-04, ..., 2.8496e-03, + 3.6407e-04, 2.5010e-04], + [-3.6125e-03, 5.0478e-07, -6.8779e-03, ..., -9.0485e-03, + -4.3221e-03, -2.2449e-03], + [ 6.1560e-04, 2.5725e-04, 9.5034e-04, ..., 2.6302e-03, + 4.6253e-04, 2.6250e-04], + ..., + [ 1.3647e-03, 2.4498e-05, 2.4986e-03, ..., 4.2381e-03, + 1.5697e-03, 1.1587e-03], + [ 4.4441e-03, 3.4302e-02, 2.5139e-03, ..., 4.8523e-03, + 2.1637e-02, 2.7390e-03], + [ 7.2241e-04, -3.2898e-02, 1.2960e-03, ..., 2.7866e-03, + -2.0584e-02, 6.2847e-04]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0230, 0.0315, 0.0027, -0.0106, 0.0215, -0.0289, 0.0086, -0.0039, + -0.0024, -0.0125], device='cuda:0'), grad: tensor([ 0.0119, -0.0352, 0.0118, -0.0016, -0.0209, -0.0333, 0.0065, 0.0182, + 0.0556, -0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 214.31, cls_loss 0.5920 cls_loss_mapping 0.0062 cls_loss_causal 0.5251 re_mapping 0.0116 re_causal 0.0270 /// teacc 98.38 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0740, -0.1039, -0.0827, ..., -0.0325, 0.0494, -0.0932], + [-0.0650, -0.0785, -0.0581, ..., 0.1014, -0.0329, 0.1929], + [ 0.0057, -0.0180, -0.0296, ..., -0.0128, -0.0085, -0.0650], + ..., + [ 0.0259, -0.0932, 0.1398, ..., 0.0260, -0.0751, 0.0434], + [-0.0320, 0.0538, -0.1305, ..., -0.0529, 0.0102, -0.0777], + [-0.0581, 0.0689, 0.0488, ..., -0.0487, -0.0425, -0.0186]], + device='cuda:0'), grad: tensor([[ 1.2279e-05, 5.4181e-05, 4.8317e-06, ..., -3.4561e-03, + -4.1428e-03, -8.6355e-04], + [-4.2498e-05, 1.7986e-05, -1.5236e-05, ..., -2.6741e-03, + 6.1417e-04, 1.2960e-03], + [ 4.7505e-05, 9.2328e-05, 1.7357e-04, ..., 1.0848e-05, + 2.0332e-03, 5.0259e-04], + ..., + [-1.2243e-04, 3.0994e-05, -4.9925e-04, ..., 3.0651e-03, + 7.0143e-04, -5.4741e-03], + [ 2.3544e-05, 2.0850e-04, 8.4043e-05, ..., 2.9106e-03, + 4.7326e-04, 7.2289e-04], + [ 5.0575e-05, 2.1780e-04, 8.6904e-05, ..., -2.3499e-03, + 6.1178e-04, 1.1015e-03]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0231, 0.0311, 0.0024, -0.0099, 0.0209, -0.0283, 0.0084, -0.0037, + -0.0022, -0.0127], device='cuda:0'), grad: tensor([-0.0097, -0.0114, 0.0094, -0.0426, 0.0252, 0.0376, 0.0184, -0.0074, + 0.0177, -0.0372], device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 214.72, cls_loss 0.5697 cls_loss_mapping 0.0050 cls_loss_causal 0.4924 re_mapping 0.0113 re_causal 0.0252 /// teacc 98.50 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0744, -0.1043, -0.0822, ..., -0.0317, 0.0491, -0.0924], + [-0.0648, -0.0788, -0.0570, ..., 0.1016, -0.0334, 0.1910], + [ 0.0061, -0.0180, -0.0301, ..., -0.0139, -0.0081, -0.0647], + ..., + [ 0.0270, -0.0937, 0.1406, ..., 0.0261, -0.0761, 0.0429], + [-0.0339, 0.0548, -0.1310, ..., -0.0532, 0.0096, -0.0791], + [-0.0577, 0.0688, 0.0494, ..., -0.0484, -0.0428, -0.0179]], + device='cuda:0'), grad: tensor([[ 3.4809e-04, 1.8036e-04, 1.8663e-03, ..., 2.1725e-03, + -1.1816e-03, 1.8721e-03], + [ 5.0306e-04, 1.3053e-04, 1.1911e-03, ..., 3.0785e-03, + 3.7408e-04, -7.8506e-03], + [-1.8110e-03, 1.1787e-02, -2.6493e-03, ..., -2.6321e-04, + 1.9875e-03, -8.4341e-06], + ..., + [-3.8071e-03, -1.2274e-03, -9.5901e-03, ..., -5.5084e-03, + -4.3449e-03, -1.3380e-03], + [-7.3910e-04, 1.9043e-02, 4.6396e-04, ..., 9.7466e-04, + 1.2553e-04, 2.7752e-04], + [ 3.3150e-03, 9.8324e-04, 4.4250e-03, ..., -7.2212e-03, + 3.1066e-04, -1.3006e-04]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0227, 0.0316, 0.0023, -0.0095, 0.0220, -0.0297, 0.0083, -0.0041, + -0.0026, -0.0126], device='cuda:0'), grad: tensor([ 0.0245, 0.0055, 0.0083, -0.0018, 0.0043, 0.0127, -0.0097, -0.0229, + -0.0085, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 214.48, cls_loss 0.5369 cls_loss_mapping 0.0058 cls_loss_causal 0.4714 re_mapping 0.0113 re_causal 0.0255 /// teacc 98.52 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.0743, -0.1049, -0.0834, ..., -0.0322, 0.0490, -0.0934], + [-0.0647, -0.0793, -0.0578, ..., 0.1010, -0.0334, 0.1918], + [ 0.0066, -0.0184, -0.0296, ..., -0.0132, -0.0078, -0.0660], + ..., + [ 0.0269, -0.0932, 0.1401, ..., 0.0266, -0.0764, 0.0440], + [-0.0338, 0.0545, -0.1309, ..., -0.0532, 0.0096, -0.0795], + [-0.0581, 0.0690, 0.0491, ..., -0.0496, -0.0431, -0.0187]], + device='cuda:0'), grad: tensor([[ 4.5896e-04, 7.6473e-05, -3.3140e-04, ..., -6.0539e-03, + 8.0729e-04, 9.2220e-04], + [-2.2221e-04, 7.2002e-05, 1.2932e-03, ..., -7.3814e-03, + -1.0307e-02, 3.0155e-03], + [ 1.0376e-03, 1.9646e-03, 6.5041e-04, ..., 4.6844e-03, + 1.6899e-03, 1.0853e-03], + ..., + [-3.5000e-04, 3.3408e-05, -7.3814e-04, ..., 1.1454e-03, + 6.6662e-04, 1.3771e-03], + [ 2.8968e-04, 9.0122e-04, 1.3554e-04, ..., 6.0387e-03, + 1.1072e-03, 5.2214e-04], + [ 1.2293e-03, -3.9548e-05, 7.4291e-04, ..., 6.5269e-03, + 5.6505e-04, 4.9324e-03]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0238, 0.0310, 0.0032, -0.0088, 0.0217, -0.0286, 0.0081, -0.0039, + -0.0025, -0.0133], device='cuda:0'), grad: tensor([-0.0382, -0.0286, 0.0015, 0.0038, -0.0443, 0.0020, 0.0435, -0.0013, + 0.0304, 0.0311], device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 214.81, cls_loss 0.5768 cls_loss_mapping 0.0072 cls_loss_causal 0.4982 re_mapping 0.0111 re_causal 0.0241 /// teacc 98.63 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.0737, -0.1051, -0.0835, ..., -0.0328, 0.0489, -0.0930], + [-0.0656, -0.0790, -0.0582, ..., 0.1012, -0.0320, 0.1915], + [ 0.0064, -0.0181, -0.0292, ..., -0.0145, -0.0097, -0.0662], + ..., + [ 0.0268, -0.0928, 0.1400, ..., 0.0265, -0.0768, 0.0438], + [-0.0341, 0.0545, -0.1312, ..., -0.0529, 0.0087, -0.0793], + [-0.0577, 0.0697, 0.0493, ..., -0.0495, -0.0427, -0.0190]], + device='cuda:0'), grad: tensor([[ 7.6115e-05, 1.9813e-04, 2.2435e-04, ..., 2.1839e-03, + 3.5739e-04, 6.6400e-05], + [ 2.7752e-04, 3.9864e-04, 5.7745e-04, ..., -1.1110e-03, + 5.5790e-04, -4.9973e-04], + [ 1.1814e-04, -2.5826e-03, 8.1444e-04, ..., 1.8148e-03, + 9.6283e-03, 7.1573e-04], + ..., + [ 2.2304e-04, 1.3323e-03, 7.4434e-04, ..., 5.3329e-03, + 2.6107e-04, 1.7309e-04], + [ 1.7953e-04, 1.6129e-04, -2.8706e-03, ..., -1.2642e-02, + -3.8971e-02, 1.3065e-04], + [-2.1400e-03, -6.1035e-04, -2.2030e-03, ..., -6.3477e-03, + 2.0754e-04, -2.0084e-03]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0235, 0.0308, 0.0025, -0.0092, 0.0219, -0.0273, 0.0069, -0.0043, + -0.0016, -0.0132], device='cuda:0'), grad: tensor([ 0.0172, -0.0247, 0.0184, 0.0362, 0.0249, 0.0246, -0.0527, 0.0337, + -0.0596, -0.0180], device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 214.34, cls_loss 0.6107 cls_loss_mapping 0.0060 cls_loss_causal 0.5219 re_mapping 0.0114 re_causal 0.0255 /// teacc 98.54 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0724, -0.1056, -0.0836, ..., -0.0329, 0.0493, -0.0936], + [-0.0658, -0.0799, -0.0577, ..., 0.1006, -0.0319, 0.1922], + [ 0.0061, -0.0176, -0.0295, ..., -0.0150, -0.0094, -0.0669], + ..., + [ 0.0267, -0.0929, 0.1411, ..., 0.0268, -0.0770, 0.0449], + [-0.0353, 0.0540, -0.1322, ..., -0.0520, 0.0106, -0.0799], + [-0.0571, 0.0697, 0.0497, ..., -0.0493, -0.0436, -0.0196]], + device='cuda:0'), grad: tensor([[-5.6410e-04, 2.7254e-05, 3.3617e-05, ..., 1.1835e-03, + 1.2834e-06, 5.1260e-05], + [ 4.3482e-05, 2.4170e-05, 3.6627e-05, ..., -3.0518e-03, + 4.3139e-06, -2.4533e-04], + [-4.9162e-04, 2.5916e-04, 1.0246e-04, ..., -3.9339e-04, + 1.0049e-04, 9.5606e-05], + ..., + [ 2.0146e-04, 9.5427e-05, 1.2083e-03, ..., 2.1553e-03, + 1.8626e-07, 9.0981e-04], + [ 1.5819e-04, -4.5967e-04, 1.4567e-04, ..., -6.7139e-03, + 1.7174e-06, 1.7560e-04], + [ 4.3273e-05, -5.9992e-05, -3.7518e-03, ..., -9.7656e-04, + 8.8140e-06, -3.7041e-03]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0241, 0.0308, 0.0025, -0.0101, 0.0219, -0.0275, 0.0079, -0.0049, + -0.0010, -0.0127], device='cuda:0'), grad: tensor([-0.0123, -0.0259, 0.0157, -0.0126, 0.0193, 0.0190, 0.0167, 0.0147, + -0.0281, -0.0065], device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 214.20, cls_loss 0.5693 cls_loss_mapping 0.0049 cls_loss_causal 0.4964 re_mapping 0.0112 re_causal 0.0255 /// teacc 98.57 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0728, -0.1057, -0.0831, ..., -0.0335, 0.0495, -0.0938], + [-0.0658, -0.0800, -0.0575, ..., 0.1011, -0.0322, 0.1933], + [ 0.0056, -0.0182, -0.0290, ..., -0.0144, -0.0096, -0.0671], + ..., + [ 0.0270, -0.0936, 0.1415, ..., 0.0261, -0.0772, 0.0445], + [-0.0355, 0.0543, -0.1341, ..., -0.0517, 0.0121, -0.0788], + [-0.0574, 0.0686, 0.0487, ..., -0.0485, -0.0450, -0.0198]], + device='cuda:0'), grad: tensor([[ 1.7881e-04, 3.5381e-04, 4.0030e-04, ..., -4.5891e-03, + 5.9986e-04, 3.7342e-05], + [ 9.3281e-05, 1.2591e-05, 1.8585e-04, ..., -2.8381e-03, + -2.5959e-03, 1.9398e-03], + [ 1.2803e-04, -8.5211e-04, -1.2903e-03, ..., 1.8597e-03, + -4.0859e-05, 3.4404e-04], + ..., + [-3.1328e-04, 8.1837e-05, -4.8494e-04, ..., -3.9825e-03, + 2.3043e-04, -9.0256e-03], + [ 1.4806e-04, 1.3399e-04, 5.1832e-04, ..., 3.1185e-03, + 4.1866e-04, 3.8433e-03], + [ 2.1446e-04, 3.0503e-05, 3.0708e-04, ..., 3.8204e-03, + 2.3937e-04, 9.2077e-04]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0240, 0.0324, 0.0030, -0.0108, 0.0219, -0.0285, 0.0083, -0.0062, + -0.0009, -0.0123], device='cuda:0'), grad: tensor([-0.0358, 0.0043, 0.0173, 0.0018, -0.0034, 0.0294, -0.0375, -0.0066, + 0.0038, 0.0267], device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 214.07, cls_loss 0.5560 cls_loss_mapping 0.0067 cls_loss_causal 0.4791 re_mapping 0.0117 re_causal 0.0258 /// teacc 98.53 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0739, -0.1072, -0.0843, ..., -0.0335, 0.0500, -0.0945], + [-0.0656, -0.0813, -0.0578, ..., 0.1010, -0.0319, 0.1936], + [ 0.0053, -0.0187, -0.0276, ..., -0.0135, -0.0100, -0.0679], + ..., + [ 0.0262, -0.0939, 0.1411, ..., 0.0259, -0.0769, 0.0448], + [-0.0343, 0.0551, -0.1329, ..., -0.0511, 0.0115, -0.0774], + [-0.0583, 0.0682, 0.0493, ..., -0.0494, -0.0459, -0.0201]], + device='cuda:0'), grad: tensor([[ 2.8461e-06, 4.2021e-05, 4.3154e-05, ..., 2.9469e-03, + 4.3988e-04, 7.9918e-04], + [ 3.8557e-06, 7.7844e-05, 1.4055e-04, ..., -1.3390e-02, + -3.4866e-03, -6.4850e-03], + [-4.4964e-06, 1.6022e-04, 1.3285e-03, ..., 2.1477e-03, + 5.3787e-04, 1.2484e-03], + ..., + [ 2.8089e-05, 3.6567e-05, -1.7338e-03, ..., 7.6199e-04, + 2.4378e-04, -6.9380e-04], + [-1.0252e-03, -2.2566e-04, 1.6749e-04, ..., 2.8095e-03, + -4.5586e-04, 1.1101e-03], + [ 2.2531e-05, 3.5310e-04, 4.2915e-05, ..., 1.5097e-03, + 3.5334e-04, 2.6264e-03]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0242, 0.0316, 0.0032, -0.0109, 0.0229, -0.0282, 0.0088, -0.0065, + -0.0008, -0.0128], device='cuda:0'), grad: tensor([ 0.0157, -0.0582, 0.0180, 0.0091, -0.0141, 0.0170, 0.0035, -0.0028, + 0.0094, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 213.99, cls_loss 0.5731 cls_loss_mapping 0.0058 cls_loss_causal 0.4952 re_mapping 0.0111 re_causal 0.0249 /// teacc 98.55 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0739, -0.1070, -0.0843, ..., -0.0338, 0.0502, -0.0954], + [-0.0652, -0.0808, -0.0585, ..., 0.1014, -0.0325, 0.1955], + [ 0.0054, -0.0191, -0.0274, ..., -0.0133, -0.0100, -0.0684], + ..., + [ 0.0262, -0.0942, 0.1408, ..., 0.0258, -0.0773, 0.0431], + [-0.0349, 0.0553, -0.1328, ..., -0.0512, 0.0113, -0.0785], + [-0.0576, 0.0686, 0.0515, ..., -0.0493, -0.0456, -0.0182]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0006, 0.0011, ..., 0.0039, 0.0019, 0.0002], + [ 0.0004, 0.0004, -0.0003, ..., -0.0005, 0.0004, -0.0011], + [ 0.0006, 0.0006, 0.0004, ..., -0.0107, -0.0016, -0.0010], + ..., + [-0.0019, -0.0004, -0.0045, ..., -0.0005, 0.0002, 0.0001], + [ 0.0007, 0.0009, 0.0017, ..., 0.0067, 0.0009, 0.0006], + [ 0.0006, 0.0007, 0.0013, ..., 0.0027, 0.0002, 0.0002]], + device='cuda:0') +Epoch 214, bias, value: tensor([-0.0248, 0.0319, 0.0033, -0.0112, 0.0235, -0.0281, 0.0075, -0.0064, + -0.0010, -0.0118], device='cuda:0'), grad: tensor([ 0.0069, -0.0064, -0.0331, 0.0267, -0.0627, 0.0283, -0.0279, 0.0061, + 0.0406, 0.0214], device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 214.09, cls_loss 0.5702 cls_loss_mapping 0.0061 cls_loss_causal 0.4836 re_mapping 0.0113 re_causal 0.0259 /// teacc 98.59 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0742, -0.1078, -0.0847, ..., -0.0338, 0.0498, -0.0958], + [-0.0648, -0.0807, -0.0583, ..., 0.1019, -0.0322, 0.1956], + [ 0.0045, -0.0200, -0.0292, ..., -0.0139, -0.0099, -0.0679], + ..., + [ 0.0273, -0.0955, 0.1404, ..., 0.0261, -0.0774, 0.0435], + [-0.0355, 0.0561, -0.1347, ..., -0.0518, 0.0113, -0.0791], + [-0.0581, 0.0677, 0.0539, ..., -0.0494, -0.0458, -0.0187]], + device='cuda:0'), grad: tensor([[ 1.0860e-04, -1.0872e-03, -1.4715e-03, ..., 1.0023e-03, + -1.7109e-03, 1.2410e-04], + [ 3.2514e-05, 6.8657e-06, -1.6870e-03, ..., -6.4583e-03, + 7.2420e-06, -1.7004e-03], + [ 4.5635e-06, 8.9169e-05, 5.0259e-04, ..., 3.0155e-03, + 1.2553e-04, 3.6430e-04], + ..., + [-1.2612e-04, 6.8545e-05, 2.0802e-04, ..., 2.5578e-03, + 6.7174e-05, 9.2149e-05], + [ 2.4125e-05, 2.0325e-04, 4.4131e-04, ..., 1.8349e-03, + 1.0383e-04, 2.8801e-04], + [ 7.8022e-05, -3.6216e-04, -5.5164e-05, ..., 1.7405e-03, + 2.2709e-04, 8.7023e-05]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0239, 0.0312, 0.0027, -0.0112, 0.0234, -0.0270, 0.0074, -0.0055, + -0.0013, -0.0128], device='cuda:0'), grad: tensor([ 0.0021, -0.0485, 0.0186, 0.0160, -0.0656, 0.0120, 0.0176, 0.0187, + 0.0125, 0.0167], device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.07, cls_loss 0.5990 cls_loss_mapping 0.0072 cls_loss_causal 0.5236 re_mapping 0.0103 re_causal 0.0235 /// teacc 98.73 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.0747, -0.1082, -0.0842, ..., -0.0340, 0.0499, -0.0962], + [-0.0648, -0.0788, -0.0580, ..., 0.1016, -0.0317, 0.1951], + [ 0.0048, -0.0186, -0.0306, ..., -0.0141, -0.0099, -0.0674], + ..., + [ 0.0275, -0.0961, 0.1411, ..., 0.0260, -0.0773, 0.0444], + [-0.0334, 0.0571, -0.1345, ..., -0.0515, 0.0111, -0.0779], + [-0.0584, 0.0665, 0.0531, ..., -0.0503, -0.0450, -0.0196]], + device='cuda:0'), grad: tensor([[-1.6766e-03, -2.1706e-03, -2.3804e-03, ..., -8.8196e-03, + 8.6963e-05, -5.6839e-03], + [ 1.8954e-04, 1.7989e-04, -4.7264e-03, ..., -4.7913e-03, + -1.4114e-03, 1.5240e-03], + [ 1.3876e-04, 3.3379e-04, 1.0920e-03, ..., -1.7757e-03, + 2.8896e-04, -3.4943e-03], + ..., + [ 3.8862e-05, 4.7636e-04, 9.1553e-04, ..., 4.9667e-03, + 1.0306e-04, 1.6413e-03], + [-1.4055e-04, 9.5785e-05, 6.0511e-04, ..., 2.2757e-04, + 1.0997e-04, 7.9393e-04], + [ 9.3985e-04, -5.0873e-05, 8.7261e-04, ..., -9.9564e-03, + 6.9737e-05, 9.5654e-04]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0239, 0.0308, 0.0041, -0.0113, 0.0233, -0.0277, 0.0075, -0.0055, + -0.0009, -0.0134], device='cuda:0'), grad: tensor([-0.0517, -0.0034, -0.0014, 0.0219, 0.0385, 0.0168, 0.0247, 0.0160, + -0.0020, -0.0595], device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 214.16, cls_loss 0.5983 cls_loss_mapping 0.0069 cls_loss_causal 0.5249 re_mapping 0.0112 re_causal 0.0248 /// teacc 98.72 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0756, -0.1082, -0.0847, ..., -0.0337, 0.0499, -0.0950], + [-0.0666, -0.0787, -0.0587, ..., 0.1014, -0.0321, 0.1952], + [ 0.0045, -0.0171, -0.0307, ..., -0.0138, -0.0100, -0.0674], + ..., + [ 0.0270, -0.0970, 0.1413, ..., 0.0264, -0.0778, 0.0436], + [-0.0322, 0.0565, -0.1333, ..., -0.0520, 0.0118, -0.0778], + [-0.0577, 0.0672, 0.0535, ..., -0.0506, -0.0452, -0.0190]], + device='cuda:0'), grad: tensor([[-2.2469e-03, -7.9584e-04, 9.6202e-05, ..., -1.5839e-02, + -2.4819e-04, -1.3809e-03], + [ 1.1177e-03, 3.7718e-04, 6.3324e-04, ..., 1.5930e-02, + 7.2829e-07, 5.3215e-03], + [ 2.1815e-04, 8.2731e-05, 3.3879e-04, ..., -8.6288e-03, + 1.9997e-05, -5.6610e-03], + ..., + [ 4.8608e-05, 1.8585e-04, -4.1924e-03, ..., -1.5793e-02, + 1.8645e-06, -1.3351e-03], + [ 9.0480e-05, 1.9610e-04, 1.9703e-03, ..., 5.5656e-03, + 1.9699e-05, -2.4164e-04], + [ 7.8321e-05, -2.9111e-04, 4.4394e-04, ..., 3.0766e-03, + 6.2883e-05, 5.3644e-04]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0233, 0.0317, 0.0043, -0.0111, 0.0224, -0.0280, 0.0065, -0.0054, + -0.0005, -0.0136], device='cuda:0'), grad: tensor([-0.0470, 0.0547, -0.0432, 0.0159, 0.0076, 0.0119, 0.0261, -0.0571, + 0.0127, 0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 214.43, cls_loss 0.5651 cls_loss_mapping 0.0055 cls_loss_causal 0.4823 re_mapping 0.0119 re_causal 0.0260 /// teacc 98.69 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0759, -0.1082, -0.0852, ..., -0.0336, 0.0494, -0.0965], + [-0.0657, -0.0808, -0.0582, ..., 0.1013, -0.0320, 0.1956], + [ 0.0072, -0.0174, -0.0304, ..., -0.0137, -0.0095, -0.0670], + ..., + [ 0.0269, -0.0987, 0.1409, ..., 0.0278, -0.0774, 0.0433], + [-0.0357, 0.0562, -0.1339, ..., -0.0522, 0.0114, -0.0779], + [-0.0563, 0.0667, 0.0532, ..., -0.0495, -0.0453, -0.0184]], + device='cuda:0'), grad: tensor([[ 3.6621e-04, -9.9945e-03, 2.0218e-04, ..., -8.1558e-03, + -1.3723e-03, 2.1625e-04], + [ 2.6665e-03, 5.5361e-04, 1.2169e-03, ..., 8.1863e-03, + 1.3754e-05, 1.8330e-03], + [ 2.2430e-03, 1.9646e-03, 3.4103e-03, ..., 7.9269e-03, + 2.7466e-04, 1.3304e-03], + ..., + [-1.8768e-03, -9.1839e-04, -4.3297e-03, ..., 5.2786e-04, + 4.5002e-06, -1.5898e-03], + [-4.4098e-03, 8.9264e-03, -8.0252e-04, ..., -2.1648e-03, + 1.4296e-03, -3.3054e-03], + [ 3.5715e-04, -5.0812e-03, 2.7966e-04, ..., 2.3766e-03, + 3.7283e-05, 2.0742e-04]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0230, 0.0310, 0.0051, -0.0109, 0.0212, -0.0287, 0.0066, -0.0044, + -0.0009, -0.0131], device='cuda:0'), grad: tensor([-0.0450, 0.0313, 0.0322, -0.0626, 0.0303, 0.0267, -0.0163, 0.0027, + 0.0120, -0.0114], device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 214.12, cls_loss 0.5768 cls_loss_mapping 0.0051 cls_loss_causal 0.5084 re_mapping 0.0115 re_causal 0.0269 /// teacc 98.68 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0764, -0.1083, -0.0858, ..., -0.0336, 0.0489, -0.0982], + [-0.0663, -0.0794, -0.0579, ..., 0.1012, -0.0322, 0.1953], + [ 0.0062, -0.0179, -0.0315, ..., -0.0142, -0.0095, -0.0670], + ..., + [ 0.0262, -0.0978, 0.1417, ..., 0.0279, -0.0773, 0.0435], + [-0.0353, 0.0575, -0.1316, ..., -0.0516, 0.0120, -0.0762], + [-0.0548, 0.0653, 0.0519, ..., -0.0500, -0.0455, -0.0194]], + device='cuda:0'), grad: tensor([[ 3.5357e-04, 1.6719e-05, 5.6791e-04, ..., 7.2527e-04, + 4.7922e-05, 4.7183e-04], + [ 5.3644e-04, 1.7837e-05, 4.9114e-04, ..., 9.4070e-03, + 7.8231e-08, 3.0231e-04], + [ 2.9159e-04, 2.5332e-05, -9.1612e-05, ..., 6.5269e-03, + 6.7130e-06, 4.6182e-04], + ..., + [ 8.9836e-04, 1.4096e-05, 2.4962e-04, ..., 6.6795e-03, + 1.4156e-07, 5.0402e-04], + [-9.9564e-04, 2.5940e-04, -4.3144e-03, ..., -1.4029e-03, + 4.1686e-06, 7.5769e-04], + [ 7.6056e-04, 2.1979e-05, 9.6464e-04, ..., 8.2092e-03, + 1.1101e-06, 6.6137e-04]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0230, 0.0310, 0.0048, -0.0109, 0.0211, -0.0278, 0.0064, -0.0045, + -0.0003, -0.0139], device='cuda:0'), grad: tensor([ 0.0064, 0.0476, 0.0331, -0.1183, -0.0215, -0.0126, -0.0029, 0.0177, + 0.0117, 0.0390], device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 214.01, cls_loss 0.6002 cls_loss_mapping 0.0049 cls_loss_causal 0.5166 re_mapping 0.0112 re_causal 0.0253 /// teacc 98.77 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.0781, -0.1081, -0.0863, ..., -0.0330, 0.0483, -0.0990], + [-0.0651, -0.0794, -0.0585, ..., 0.1023, -0.0323, 0.1960], + [ 0.0059, -0.0171, -0.0306, ..., -0.0143, -0.0096, -0.0665], + ..., + [ 0.0277, -0.0982, 0.1423, ..., 0.0288, -0.0775, 0.0443], + [-0.0378, 0.0586, -0.1315, ..., -0.0525, 0.0120, -0.0762], + [-0.0558, 0.0649, 0.0515, ..., -0.0502, -0.0448, -0.0199]], + device='cuda:0'), grad: tensor([[ 4.1962e-04, 8.2111e-04, 7.3314e-05, ..., 6.5804e-03, + 1.1654e-03, 1.2045e-03], + [ 1.4935e-03, 4.7147e-05, 1.2660e-04, ..., 8.8196e-03, + 9.6783e-06, 5.5084e-03], + [ 1.1892e-03, 5.4598e-04, 6.1095e-05, ..., 6.2103e-03, + 8.8811e-05, 4.5300e-04], + ..., + [ 7.2384e-04, 1.6851e-03, 2.5692e-03, ..., -1.4515e-03, + 1.2517e-05, 3.5973e-03], + [-1.5802e-03, -8.4758e-05, -3.3712e-04, ..., -1.6327e-02, + 7.3576e-04, -7.8354e-03], + [ 3.7885e-04, 4.5784e-06, -3.1605e-03, ..., 5.3291e-03, + 1.1110e-04, -1.1520e-03]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0236, 0.0315, 0.0051, -0.0112, 0.0211, -0.0274, 0.0056, -0.0043, + -0.0012, -0.0128], device='cuda:0'), grad: tensor([ 0.0312, 0.0277, 0.0145, -0.0169, -0.0230, 0.0320, -0.0288, -0.0087, + -0.0453, 0.0171], device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 214.35, cls_loss 0.5992 cls_loss_mapping 0.0048 cls_loss_causal 0.5243 re_mapping 0.0110 re_causal 0.0261 /// teacc 98.82 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.0754, -0.1086, -0.0858, ..., -0.0336, 0.0482, -0.0967], + [-0.0686, -0.0799, -0.0592, ..., 0.1023, -0.0326, 0.1958], + [ 0.0047, -0.0178, -0.0311, ..., -0.0145, -0.0099, -0.0652], + ..., + [ 0.0285, -0.0990, 0.1421, ..., 0.0287, -0.0779, 0.0445], + [-0.0369, 0.0600, -0.1317, ..., -0.0522, 0.0116, -0.0775], + [-0.0551, 0.0650, 0.0514, ..., -0.0504, -0.0432, -0.0204]], + device='cuda:0'), grad: tensor([[ 7.4720e-04, 5.3072e-04, 6.2704e-04, ..., 5.0354e-03, + 3.8207e-05, 6.2370e-04], + [ 8.5688e-04, -2.0862e-04, -3.9406e-03, ..., -7.6523e-03, + -1.3428e-03, -1.3962e-03], + [-4.7302e-03, -4.4823e-03, -1.2617e-03, ..., -2.2369e-02, + 9.5069e-05, -2.1706e-03], + ..., + [ 8.4114e-04, 1.1663e-03, 2.1744e-03, ..., 5.8098e-03, + 2.2268e-04, 7.0047e-04], + [ 1.0900e-03, 1.4763e-03, 1.9102e-03, ..., 2.6913e-03, + 4.2915e-04, 1.3380e-03], + [ 1.1796e-04, 1.2660e-04, 1.1032e-02, ..., 4.7569e-03, + 4.5955e-05, 3.9506e-04]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0240, 0.0309, 0.0052, -0.0101, 0.0222, -0.0276, 0.0051, -0.0044, + -0.0013, -0.0132], device='cuda:0'), grad: tensor([ 0.0250, -0.0569, -0.0581, 0.0018, 0.0286, -0.0322, 0.0130, 0.0335, + 0.0160, 0.0293], device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.16, cls_loss 0.5475 cls_loss_mapping 0.0048 cls_loss_causal 0.4777 re_mapping 0.0115 re_causal 0.0249 /// teacc 98.76 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0746, -0.1093, -0.0852, ..., -0.0337, 0.0482, -0.0965], + [-0.0685, -0.0799, -0.0605, ..., 0.1022, -0.0326, 0.1962], + [ 0.0066, -0.0176, -0.0313, ..., -0.0129, -0.0100, -0.0659], + ..., + [ 0.0278, -0.0997, 0.1428, ..., 0.0289, -0.0776, 0.0448], + [-0.0365, 0.0598, -0.1324, ..., -0.0533, 0.0119, -0.0779], + [-0.0556, 0.0656, 0.0511, ..., -0.0501, -0.0431, -0.0203]], + device='cuda:0'), grad: tensor([[-2.8248e-03, 2.0850e-04, 2.3901e-04, ..., -1.1261e-02, + -8.6355e-04, -9.3460e-03], + [ 2.6970e-03, -2.1374e-04, 5.8842e-04, ..., 4.2191e-03, + 8.4400e-05, 3.5286e-03], + [ 2.0180e-03, 1.2569e-03, 1.0595e-03, ..., 3.3684e-03, + 6.5684e-05, 2.2984e-03], + ..., + [-9.2468e-03, -1.9302e-03, -5.9166e-03, ..., -1.4053e-02, + 2.2304e-04, -5.2528e-03], + [ 8.0872e-04, 6.8569e-04, 2.8992e-04, ..., -2.9964e-03, + 9.0301e-05, 1.0881e-03], + [ 2.5063e-03, 6.2370e-04, 1.3590e-03, ..., 8.4915e-03, + 2.2161e-04, 3.2387e-03]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0246, 0.0320, 0.0060, -0.0097, 0.0226, -0.0285, 0.0049, -0.0039, + -0.0028, -0.0133], device='cuda:0'), grad: tensor([-0.0427, 0.0091, 0.0032, 0.0105, 0.0315, 0.0234, -0.0151, -0.0388, + -0.0350, 0.0539], device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.17, cls_loss 0.5763 cls_loss_mapping 0.0080 cls_loss_causal 0.5041 re_mapping 0.0115 re_causal 0.0249 /// teacc 98.64 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0731, -0.1094, -0.0852, ..., -0.0318, 0.0487, -0.0963], + [-0.0687, -0.0786, -0.0603, ..., 0.1026, -0.0329, 0.1970], + [ 0.0056, -0.0172, -0.0314, ..., -0.0141, -0.0096, -0.0650], + ..., + [ 0.0272, -0.0996, 0.1431, ..., 0.0278, -0.0780, 0.0436], + [-0.0352, 0.0591, -0.1322, ..., -0.0525, 0.0120, -0.0773], + [-0.0545, 0.0648, 0.0502, ..., -0.0498, -0.0434, -0.0206]], + device='cuda:0'), grad: tensor([[ 1.0643e-03, -1.1587e-03, 7.6485e-04, ..., -9.4604e-03, + -5.8508e-04, -3.4618e-03], + [ 7.0477e-04, -9.4700e-04, 4.5419e-04, ..., 4.3259e-03, + 5.0336e-05, 2.5964e-04], + [ 8.4209e-04, 4.1127e-04, 1.3371e-03, ..., -3.4881e-04, + 7.2360e-05, 1.5354e-03], + ..., + [ 6.4039e-04, -9.1648e-04, -7.8049e-03, ..., 2.6073e-03, + 6.5088e-05, -6.1512e-04], + [ 8.4448e-04, 9.6893e-04, 3.3283e-03, ..., 8.2245e-03, + 8.2672e-05, 2.3403e-03], + [ 1.0414e-03, 6.4945e-04, 1.5247e-04, ..., 5.5656e-03, + 5.3453e-04, 2.7370e-04]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0224, 0.0324, 0.0054, -0.0111, 0.0216, -0.0290, 0.0048, -0.0044, + -0.0018, -0.0127], device='cuda:0'), grad: tensor([-0.0044, 0.0173, 0.0046, 0.0491, -0.0539, -0.0088, -0.0236, -0.0196, + 0.0489, -0.0095], device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 214.32, cls_loss 0.5663 cls_loss_mapping 0.0045 cls_loss_causal 0.4862 re_mapping 0.0116 re_causal 0.0252 /// teacc 98.54 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0735, -0.1106, -0.0860, ..., -0.0322, 0.0490, -0.0977], + [-0.0700, -0.0798, -0.0616, ..., 0.1030, -0.0329, 0.1982], + [ 0.0052, -0.0171, -0.0323, ..., -0.0141, -0.0096, -0.0650], + ..., + [ 0.0273, -0.0992, 0.1436, ..., 0.0265, -0.0783, 0.0446], + [-0.0351, 0.0589, -0.1320, ..., -0.0527, 0.0119, -0.0776], + [-0.0559, 0.0655, 0.0499, ..., -0.0505, -0.0438, -0.0223]], + device='cuda:0'), grad: tensor([[ 0.0006, 0.0002, 0.0001, ..., -0.0014, 0.0000, 0.0001], + [ 0.0066, 0.0004, 0.0003, ..., 0.0067, 0.0000, 0.0030], + [ 0.0011, 0.0007, 0.0003, ..., 0.0046, 0.0000, 0.0019], + ..., + [-0.0056, -0.0047, -0.0063, ..., -0.0095, 0.0000, -0.0092], + [-0.0002, -0.0018, -0.0001, ..., -0.0047, 0.0000, 0.0017], + [ 0.0016, 0.0009, 0.0011, ..., 0.0044, 0.0000, 0.0031]], + device='cuda:0') +Epoch 224, bias, value: tensor([-0.0231, 0.0330, 0.0048, -0.0104, 0.0220, -0.0287, 0.0057, -0.0047, + -0.0019, -0.0138], device='cuda:0'), grad: tensor([-0.0156, 0.0368, 0.0228, 0.0197, 0.0014, -0.0449, 0.0172, -0.0463, + -0.0126, 0.0214], device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 213.97, cls_loss 0.5663 cls_loss_mapping 0.0071 cls_loss_causal 0.5022 re_mapping 0.0114 re_causal 0.0251 /// teacc 98.83 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.0731, -0.1099, -0.0862, ..., -0.0325, 0.0493, -0.0972], + [-0.0709, -0.0822, -0.0620, ..., 0.1024, -0.0332, 0.1987], + [ 0.0047, -0.0149, -0.0302, ..., -0.0135, -0.0102, -0.0658], + ..., + [ 0.0278, -0.0980, 0.1438, ..., 0.0267, -0.0784, 0.0430], + [-0.0345, 0.0598, -0.1324, ..., -0.0527, 0.0114, -0.0786], + [-0.0553, 0.0646, 0.0494, ..., -0.0505, -0.0437, -0.0212]], + device='cuda:0'), grad: tensor([[ 2.8461e-06, 2.2831e-03, 3.6508e-05, ..., 6.0158e-03, + 3.4809e-05, 6.7651e-05], + [ 1.8108e-04, 3.0541e-04, -6.8998e-04, ..., -5.6992e-03, + 1.3039e-06, 1.8954e-04], + [ 2.4176e-04, -1.1034e-03, 1.5755e-03, ..., -1.6754e-02, + 7.4327e-05, -6.0081e-05], + ..., + [ 7.0877e-03, 5.5695e-04, 4.9019e-03, ..., 2.5116e-02, + 3.5707e-06, 5.8556e-03], + [ 6.0081e-05, -3.9253e-03, -2.9683e-04, ..., -4.1046e-03, + -5.9462e-04, -1.8370e-04], + [ 7.0632e-05, -7.6103e-04, -2.5978e-03, ..., 2.3766e-03, + 4.7684e-06, -2.7885e-03]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0231, 0.0323, 0.0046, -0.0103, 0.0221, -0.0284, 0.0060, -0.0051, + -0.0020, -0.0134], device='cuda:0'), grad: tensor([ 3.7628e-02, -7.4997e-03, -3.1494e-02, -7.7248e-03, 3.2043e-02, + -4.3427e-02, -1.8875e-02, 5.7281e-02, -1.7914e-02, -1.9684e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.78, cls_loss 0.5742 cls_loss_mapping 0.0047 cls_loss_causal 0.4994 re_mapping 0.0109 re_causal 0.0249 /// teacc 98.71 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.0722, -0.1093, -0.0850, ..., -0.0332, 0.0495, -0.0977], + [-0.0717, -0.0824, -0.0617, ..., 0.1025, -0.0334, 0.1998], + [ 0.0031, -0.0158, -0.0301, ..., -0.0144, -0.0100, -0.0664], + ..., + [ 0.0287, -0.0974, 0.1435, ..., 0.0276, -0.0793, 0.0425], + [-0.0341, 0.0593, -0.1324, ..., -0.0523, 0.0128, -0.0786], + [-0.0547, 0.0644, 0.0493, ..., -0.0500, -0.0438, -0.0202]], + device='cuda:0'), grad: tensor([[ 5.2786e-04, 8.1718e-05, 4.0793e-04, ..., 3.3360e-03, + 3.3450e-04, 1.3947e-04], + [ 9.3746e-04, 4.4376e-05, 2.8014e-04, ..., -5.2910e-03, + 2.3520e-04, -1.7614e-03], + [ 1.2846e-03, 1.2589e-04, 1.0080e-03, ..., 3.7193e-03, + 3.9816e-04, 4.1628e-04], + ..., + [ 8.4782e-04, 8.3148e-05, 3.0208e-04, ..., -1.9970e-03, + 1.4424e-04, -2.3384e-03], + [ 8.4448e-04, 2.3890e-04, 1.0691e-03, ..., 4.7951e-03, + 2.8586e-04, -1.2684e-04], + [ 7.0572e-04, 7.4565e-05, 2.8324e-04, ..., 3.5191e-03, + 1.2362e-04, 6.8188e-04]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0241, 0.0325, 0.0036, -0.0095, 0.0217, -0.0282, 0.0061, -0.0046, + -0.0016, -0.0132], device='cuda:0'), grad: tensor([ 0.0205, -0.0203, 0.0284, -0.0641, -0.0561, 0.0060, 0.0310, -0.0166, + 0.0341, 0.0371], device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 214.43, cls_loss 0.5869 cls_loss_mapping 0.0046 cls_loss_causal 0.5136 re_mapping 0.0115 re_causal 0.0262 /// teacc 98.68 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0711, -0.1111, -0.0851, ..., -0.0326, 0.0497, -0.0980], + [-0.0719, -0.0826, -0.0610, ..., 0.1026, -0.0338, 0.2002], + [ 0.0030, -0.0158, -0.0309, ..., -0.0148, -0.0099, -0.0671], + ..., + [ 0.0297, -0.0968, 0.1437, ..., 0.0277, -0.0791, 0.0433], + [-0.0352, 0.0601, -0.1330, ..., -0.0530, 0.0137, -0.0788], + [-0.0554, 0.0637, 0.0488, ..., -0.0490, -0.0434, -0.0199]], + device='cuda:0'), grad: tensor([[-2.5005e-03, 1.7715e-04, -8.5831e-03, ..., -1.4473e-02, + -3.4161e-03, -1.4687e-03], + [ 1.1892e-03, 2.4891e-04, 4.1389e-04, ..., 3.2196e-03, + 6.1655e-04, 5.3740e-04], + [-2.5520e-03, 7.3814e-04, 1.7700e-03, ..., -3.4695e-03, + -9.1324e-03, 7.1144e-04], + ..., + [ 5.8460e-04, 1.7624e-03, 3.5114e-03, ..., 5.6610e-03, + 9.4986e-03, 1.0586e-03], + [ 3.5262e-04, -6.5470e-04, -1.2217e-03, ..., -3.3951e-03, + 4.2176e-04, 9.1121e-06], + [ 5.6934e-04, 1.6189e-04, 8.1205e-04, ..., 4.4289e-03, + 8.1015e-04, 5.3644e-04]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0239, 0.0325, 0.0042, -0.0092, 0.0219, -0.0296, 0.0054, -0.0047, + -0.0013, -0.0128], device='cuda:0'), grad: tensor([-0.0480, 0.0073, -0.0360, -0.0316, 0.0310, 0.0374, 0.0148, 0.0127, + -0.0152, 0.0276], device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 214.64, cls_loss 0.5610 cls_loss_mapping 0.0046 cls_loss_causal 0.4914 re_mapping 0.0112 re_causal 0.0247 /// teacc 98.69 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0704, -0.1111, -0.0857, ..., -0.0326, 0.0507, -0.0977], + [-0.0719, -0.0809, -0.0613, ..., 0.1030, -0.0347, 0.2002], + [ 0.0016, -0.0163, -0.0322, ..., -0.0157, -0.0102, -0.0684], + ..., + [ 0.0298, -0.0976, 0.1451, ..., 0.0285, -0.0770, 0.0433], + [-0.0354, 0.0600, -0.1332, ..., -0.0524, 0.0134, -0.0789], + [-0.0546, 0.0636, 0.0481, ..., -0.0504, -0.0444, -0.0198]], + device='cuda:0'), grad: tensor([[-2.8553e-03, -2.4071e-03, -3.1495e-04, ..., -5.1155e-03, + -3.2539e-03, 7.4387e-05], + [ 1.1921e-04, 1.9062e-04, 1.4997e-04, ..., 2.4853e-03, + 1.4198e-04, 1.5450e-04], + [ 6.7186e-04, 1.7548e-03, 3.6740e-04, ..., 8.5526e-03, + 8.2207e-04, 1.0949e-04], + ..., + [ 1.6689e-04, 5.9509e-04, -1.9741e-03, ..., -1.3649e-02, + 2.0218e-04, -1.7357e-03], + [ 2.1684e-04, 2.0084e-03, 1.1826e-04, ..., -5.1003e-03, + 2.7442e-04, 1.8567e-05], + [ 1.5056e-04, 4.7970e-04, 1.1988e-03, ..., 1.8883e-03, + 2.4343e-04, 1.2550e-03]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0240, 0.0330, 0.0026, -0.0089, 0.0224, -0.0295, 0.0056, -0.0042, + -0.0008, -0.0136], device='cuda:0'), grad: tensor([-0.0011, -0.0101, 0.0345, -0.0046, 0.0180, 0.0070, 0.0244, -0.0599, + -0.0038, -0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 214.34, cls_loss 0.5399 cls_loss_mapping 0.0051 cls_loss_causal 0.4694 re_mapping 0.0108 re_causal 0.0234 /// teacc 98.65 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0706, -0.1106, -0.0863, ..., -0.0339, 0.0485, -0.0980], + [-0.0701, -0.0812, -0.0621, ..., 0.1019, -0.0345, 0.2030], + [ 0.0034, -0.0164, -0.0329, ..., -0.0146, -0.0102, -0.0684], + ..., + [ 0.0287, -0.0966, 0.1455, ..., 0.0290, -0.0763, 0.0420], + [-0.0361, 0.0589, -0.1334, ..., -0.0522, 0.0136, -0.0798], + [-0.0540, 0.0633, 0.0486, ..., -0.0503, -0.0451, -0.0184]], + device='cuda:0'), grad: tensor([[ 5.6159e-07, 2.9802e-04, 3.5906e-04, ..., 2.8267e-03, + -2.3251e-03, 2.0485e-03], + [ 7.1526e-07, 2.2382e-05, -2.7585e-04, ..., -1.1169e-02, + 1.1958e-05, -1.0422e-02], + [-3.3993e-06, -1.1711e-03, -9.2208e-05, ..., -5.0011e-03, + 1.0979e-04, 8.1539e-04], + ..., + [ 1.9938e-05, 1.1569e-04, -7.0906e-04, ..., -2.2745e-04, + 4.0245e-04, -2.9793e-03], + [ 1.5870e-05, -3.5620e-04, 1.4293e-04, ..., 2.1343e-03, + 2.8110e-04, 1.8787e-03], + [-6.1452e-05, 4.0233e-05, 1.3518e-04, ..., 3.1624e-03, + 1.5128e-04, 2.0161e-03]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0243, 0.0332, 0.0024, -0.0092, 0.0227, -0.0296, 0.0059, -0.0037, + -0.0010, -0.0138], device='cuda:0'), grad: tensor([-0.0116, -0.0450, -0.0069, 0.0139, -0.0077, 0.0103, 0.0267, -0.0065, + 0.0127, 0.0143], device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 214.56, cls_loss 0.5712 cls_loss_mapping 0.0068 cls_loss_causal 0.5101 re_mapping 0.0107 re_causal 0.0242 /// teacc 98.61 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0709, -0.1119, -0.0876, ..., -0.0350, 0.0492, -0.0993], + [-0.0702, -0.0806, -0.0622, ..., 0.1019, -0.0342, 0.2021], + [ 0.0043, -0.0158, -0.0332, ..., -0.0147, -0.0108, -0.0697], + ..., + [ 0.0284, -0.0968, 0.1461, ..., 0.0289, -0.0756, 0.0429], + [-0.0361, 0.0586, -0.1340, ..., -0.0530, 0.0137, -0.0796], + [-0.0535, 0.0617, 0.0483, ..., -0.0494, -0.0459, -0.0191]], + device='cuda:0'), grad: tensor([[ 2.9150e-07, 1.0672e-03, 6.2622e-06, ..., -4.0817e-03, + 5.4240e-05, 1.9538e-04], + [ 4.0047e-07, 1.9634e-04, 3.0845e-06, ..., 4.3907e-03, + 8.6948e-06, -6.1572e-05], + [ 9.0674e-06, 6.2037e-04, 3.1412e-05, ..., 2.7390e-03, + 1.4491e-05, 1.5473e-04], + ..., + [ 3.5495e-05, 2.6155e-04, 1.9670e-04, ..., 3.4332e-03, + 1.6153e-04, 5.6124e-04], + [ 9.9242e-06, 4.2000e-03, 6.9618e-05, ..., 3.5801e-03, + 1.0210e-04, 3.9744e-04], + [-3.2932e-05, 1.2121e-03, -4.2796e-04, ..., -1.4172e-03, + -4.5228e-04, -1.1530e-03]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0250, 0.0329, 0.0009, -0.0087, 0.0223, -0.0282, 0.0067, -0.0032, + -0.0013, -0.0137], device='cuda:0'), grad: tensor([-0.0169, 0.0139, 0.0095, -0.0095, -0.0542, 0.0131, 0.0189, 0.0124, + 0.0180, -0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 214.54, cls_loss 0.5767 cls_loss_mapping 0.0053 cls_loss_causal 0.4908 re_mapping 0.0109 re_causal 0.0233 /// teacc 98.72 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0705, -0.1122, -0.0894, ..., -0.0346, 0.0504, -0.0989], + [-0.0705, -0.0804, -0.0631, ..., 0.1005, -0.0345, 0.2028], + [ 0.0051, -0.0165, -0.0319, ..., -0.0144, -0.0111, -0.0726], + ..., + [ 0.0283, -0.0967, 0.1468, ..., 0.0290, -0.0762, 0.0430], + [-0.0363, 0.0577, -0.1340, ..., -0.0528, 0.0128, -0.0800], + [-0.0518, 0.0632, 0.0474, ..., -0.0494, -0.0469, -0.0197]], + device='cuda:0'), grad: tensor([[ 1.3900e-04, 7.6354e-05, 7.6914e-04, ..., 1.9569e-03, + -9.5129e-04, 8.3971e-04], + [-2.3613e-03, -1.5015e-02, -5.2299e-03, ..., -4.5593e-02, + 3.9995e-05, -7.2708e-03], + [ 9.7752e-05, 4.7188e-03, 2.7313e-03, ..., 1.2878e-02, + 2.5034e-04, -7.3528e-04], + ..., + [ 4.4435e-05, 2.4700e-03, 2.2233e-04, ..., 1.0353e-02, + 2.6464e-05, 1.8654e-03], + [ 2.5678e-04, 2.1191e-03, 1.4353e-03, ..., 4.0894e-03, + 9.5129e-04, 1.4772e-03], + [ 4.7731e-04, -1.3447e-03, 1.3971e-03, ..., 8.3313e-03, + 2.9683e-04, 1.4420e-03]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0251, 0.0317, 0.0008, -0.0093, 0.0238, -0.0279, 0.0055, -0.0022, + -0.0008, -0.0139], device='cuda:0'), grad: tensor([ 0.0060, -0.0731, 0.0141, 0.0069, -0.0076, 0.0104, -0.0202, 0.0261, + 0.0183, 0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 214.63, cls_loss 0.5712 cls_loss_mapping 0.0051 cls_loss_causal 0.4958 re_mapping 0.0109 re_causal 0.0246 /// teacc 98.74 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.0701, -0.1123, -0.0887, ..., -0.0354, 0.0504, -0.1002], + [-0.0707, -0.0797, -0.0628, ..., 0.1006, -0.0355, 0.2027], + [ 0.0054, -0.0166, -0.0329, ..., -0.0133, -0.0137, -0.0706], + ..., + [ 0.0291, -0.0977, 0.1482, ..., 0.0295, -0.0770, 0.0437], + [-0.0363, 0.0589, -0.1311, ..., -0.0530, 0.0147, -0.0796], + [-0.0522, 0.0640, 0.0473, ..., -0.0496, -0.0464, -0.0215]], + device='cuda:0'), grad: tensor([[ 2.4866e-06, 2.6250e-04, 8.8036e-05, ..., 8.2302e-04, + 4.0627e-04, 1.4424e-04], + [ 5.4419e-05, 3.1996e-04, 6.2704e-04, ..., 1.2245e-03, + 4.5627e-05, -1.2474e-03], + [-5.5695e-04, 1.3208e-04, -6.8188e-04, ..., -9.0790e-04, + 6.5029e-05, -4.5300e-04], + ..., + [ 4.2605e-04, 1.7910e-03, 5.0240e-03, ..., 1.9670e-04, + 2.6211e-05, 2.3632e-03], + [ 8.5309e-06, 6.2828e-03, 4.8542e-04, ..., 1.1730e-03, + 1.1528e-02, 4.5061e-04], + [-3.9153e-06, -1.0872e-03, -3.6030e-03, ..., 1.0967e-03, + -5.8144e-05, -2.3346e-03]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0257, 0.0315, 0.0013, -0.0092, 0.0235, -0.0292, 0.0061, -0.0015, + -0.0011, -0.0133], device='cuda:0'), grad: tensor([ 0.0132, 0.0147, 0.0053, 0.0095, 0.0179, 0.0032, -0.0523, -0.0313, + 0.0186, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 214.66, cls_loss 0.5765 cls_loss_mapping 0.0047 cls_loss_causal 0.5064 re_mapping 0.0111 re_causal 0.0256 /// teacc 98.73 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.0697, -0.1135, -0.0889, ..., -0.0355, 0.0492, -0.1017], + [-0.0704, -0.0799, -0.0638, ..., 0.1006, -0.0369, 0.2032], + [ 0.0051, -0.0167, -0.0324, ..., -0.0127, -0.0140, -0.0709], + ..., + [ 0.0288, -0.0982, 0.1481, ..., 0.0295, -0.0765, 0.0432], + [-0.0363, 0.0590, -0.1305, ..., -0.0528, 0.0156, -0.0793], + [-0.0525, 0.0643, 0.0474, ..., -0.0491, -0.0447, -0.0208]], + device='cuda:0'), grad: tensor([[ 2.8387e-06, 2.2042e-04, 5.7817e-06, ..., -2.5940e-03, + 2.5201e-04, 2.4700e-03], + [ 1.4806e-04, 3.3474e-04, 1.8764e-04, ..., 3.1853e-03, + 3.8266e-04, 3.0303e-04], + [ 2.4331e-04, 5.3883e-04, 3.1638e-04, ..., -3.7918e-03, + 5.9938e-04, 6.6853e-04], + ..., + [-2.1515e-03, 5.2691e-04, -2.8057e-03, ..., -2.0847e-03, + 4.8971e-04, -2.2335e-03], + [ 5.9307e-05, 5.3358e-04, 8.2135e-05, ..., 2.4242e-03, + 6.2084e-04, 2.2531e-04], + [ 2.7344e-05, -5.8060e-03, 1.4567e-04, ..., 1.9608e-03, + -6.2447e-03, 3.3054e-03]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0252, 0.0317, 0.0010, -0.0096, 0.0228, -0.0293, 0.0059, -0.0016, + -0.0008, -0.0125], device='cuda:0'), grad: tensor([-0.0121, 0.0139, -0.0163, -0.0060, 0.0126, -0.0164, 0.0141, 0.0078, + 0.0121, -0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 214.26, cls_loss 0.5644 cls_loss_mapping 0.0045 cls_loss_causal 0.4830 re_mapping 0.0109 re_causal 0.0238 /// teacc 98.63 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.0702, -0.1138, -0.0882, ..., -0.0352, 0.0492, -0.1022], + [-0.0707, -0.0792, -0.0640, ..., 0.1016, -0.0383, 0.2039], + [ 0.0055, -0.0167, -0.0323, ..., -0.0134, -0.0146, -0.0721], + ..., + [ 0.0278, -0.0979, 0.1483, ..., 0.0298, -0.0763, 0.0428], + [-0.0364, 0.0588, -0.1311, ..., -0.0533, 0.0138, -0.0805], + [-0.0513, 0.0642, 0.0488, ..., -0.0489, -0.0437, -0.0198]], + device='cuda:0'), grad: tensor([[-1.1975e-04, 2.3153e-06, 5.5805e-06, ..., -1.0490e-03, + -1.3843e-05, 3.4642e-04], + [ 1.7136e-06, 3.3192e-06, -3.0816e-05, ..., 2.3289e-03, + 1.5330e-06, 6.4774e-03], + [ 7.6771e-04, 8.3148e-05, 1.4350e-05, ..., 2.8667e-03, + 6.1281e-06, 4.8733e-04], + ..., + [ 7.6199e-04, 1.0139e-04, 3.2568e-04, ..., 2.4014e-03, + 2.1029e-06, 4.8065e-04], + [-2.1954e-03, 1.1092e-04, 7.3004e-04, ..., 1.9588e-03, + 3.8123e-04, 1.1520e-03], + [ 1.7905e-04, -2.9325e-05, -3.8743e-04, ..., 3.3779e-03, + 2.7288e-06, 4.2486e-04]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0255, 0.0324, 0.0010, -0.0092, 0.0221, -0.0289, 0.0063, -0.0027, + -0.0010, -0.0120], device='cuda:0'), grad: tensor([-0.0234, 0.0062, 0.0137, 0.0073, -0.0247, -0.0024, -0.0121, 0.0110, + 0.0032, 0.0209], device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.07, cls_loss 0.5947 cls_loss_mapping 0.0067 cls_loss_causal 0.5170 re_mapping 0.0111 re_causal 0.0253 /// teacc 98.75 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.0699, -0.1143, -0.0880, ..., -0.0348, 0.0489, -0.1033], + [-0.0707, -0.0783, -0.0645, ..., 0.1018, -0.0362, 0.2055], + [ 0.0045, -0.0160, -0.0324, ..., -0.0142, -0.0155, -0.0729], + ..., + [ 0.0272, -0.0984, 0.1466, ..., 0.0303, -0.0762, 0.0422], + [-0.0369, 0.0583, -0.1314, ..., -0.0531, 0.0127, -0.0808], + [-0.0509, 0.0649, 0.0491, ..., -0.0487, -0.0426, -0.0187]], + device='cuda:0'), grad: tensor([[ 8.3074e-06, 1.1742e-04, 9.1016e-05, ..., 6.1703e-04, + 1.9252e-04, 6.1846e-04], + [ 8.8289e-07, 1.2741e-03, 2.3019e-04, ..., 1.9932e-03, + 1.7869e-04, -1.2608e-03], + [-6.3121e-05, -2.6822e-04, 1.0884e-04, ..., -1.8063e-03, + 1.1981e-04, 1.3602e-04], + ..., + [ 2.6092e-05, 3.2401e-04, 6.5506e-05, ..., 1.9817e-03, + 2.9612e-04, 1.5855e-04], + [ 4.4405e-05, 1.0509e-03, 2.0254e-04, ..., 3.6297e-03, + 1.8847e-04, 8.9943e-05], + [ 1.2733e-05, -3.0518e-03, 5.2124e-05, ..., 1.2131e-03, + 3.6049e-04, 2.8076e-03]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0255, 0.0328, 0.0009, -0.0097, 0.0238, -0.0287, 0.0045, -0.0025, + -0.0013, -0.0118], device='cuda:0'), grad: tensor([ 0.0101, 0.0037, -0.0083, -0.0040, -0.0082, -0.0322, -0.0377, 0.0342, + 0.0347, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.28, cls_loss 0.5594 cls_loss_mapping 0.0052 cls_loss_causal 0.4838 re_mapping 0.0110 re_causal 0.0245 /// teacc 98.79 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.0696, -0.1144, -0.0875, ..., -0.0345, 0.0503, -0.1021], + [-0.0713, -0.0791, -0.0659, ..., 0.1028, -0.0345, 0.2041], + [ 0.0049, -0.0151, -0.0322, ..., -0.0141, -0.0157, -0.0732], + ..., + [ 0.0279, -0.0995, 0.1470, ..., 0.0301, -0.0767, 0.0436], + [-0.0370, 0.0574, -0.1318, ..., -0.0540, 0.0133, -0.0820], + [-0.0512, 0.0647, 0.0491, ..., -0.0493, -0.0426, -0.0197]], + device='cuda:0'), grad: tensor([[ 6.6310e-07, 5.5656e-06, 3.1471e-04, ..., 2.4071e-03, + -2.3469e-07, 1.2016e-04], + [ 1.4976e-06, 1.1861e-05, 1.5330e-04, ..., 2.9663e-02, + 7.8231e-08, 1.9104e-02], + [ 7.7412e-06, 4.3482e-05, 1.1454e-03, ..., -1.7593e-02, + 4.9174e-07, -1.8692e-02], + ..., + [ 6.7987e-07, 5.7071e-06, 1.5135e-03, ..., 2.9831e-03, + 3.7253e-08, -4.2415e-04], + [-1.0349e-05, 6.7800e-06, 1.1616e-03, ..., 5.1804e-03, + 2.7940e-08, 6.0225e-04], + [ 2.3749e-06, -6.0387e-03, 3.4332e-04, ..., 1.9875e-03, + 1.4901e-07, 2.4462e-04]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0256, 0.0333, 0.0010, -0.0094, 0.0243, -0.0289, 0.0042, -0.0021, + -0.0024, -0.0120], device='cuda:0'), grad: tensor([-0.0199, 0.0538, -0.0140, 0.0453, -0.0123, -0.0746, -0.0145, 0.0121, + 0.0191, 0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 214.52, cls_loss 0.5479 cls_loss_mapping 0.0064 cls_loss_causal 0.4875 re_mapping 0.0110 re_causal 0.0243 /// teacc 98.68 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.0702, -0.1147, -0.0879, ..., -0.0353, 0.0512, -0.1030], + [-0.0718, -0.0796, -0.0668, ..., 0.1016, -0.0350, 0.2047], + [ 0.0047, -0.0154, -0.0316, ..., -0.0133, -0.0160, -0.0734], + ..., + [ 0.0276, -0.0996, 0.1481, ..., 0.0292, -0.0773, 0.0430], + [-0.0358, 0.0576, -0.1326, ..., -0.0520, 0.0143, -0.0817], + [-0.0511, 0.0645, 0.0472, ..., -0.0488, -0.0427, -0.0193]], + device='cuda:0'), grad: tensor([[ 2.8110e-04, 7.5698e-05, -7.0419e-03, ..., -5.1498e-03, + 6.0380e-05, -7.2403e-03], + [ 2.4462e-04, 3.5435e-05, 1.6775e-03, ..., 3.6716e-03, + 5.1737e-05, 1.8911e-03], + [ 6.2275e-04, 5.4359e-04, -4.5824e-04, ..., 2.1820e-03, + -2.2995e-04, -1.8797e-03], + ..., + [ 1.8370e-04, 4.8876e-05, 1.3533e-03, ..., 5.2414e-03, + 1.2457e-04, 2.1420e-03], + [ 2.5201e-04, 3.5787e-04, 1.7033e-03, ..., -5.4779e-03, + 3.6693e-04, 2.0828e-03], + [ 2.3329e-04, 1.0848e-04, 1.4591e-03, ..., -3.0117e-03, + 5.4836e-04, 1.9760e-03]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0269, 0.0322, 0.0009, -0.0088, 0.0236, -0.0288, 0.0045, -0.0017, + -0.0008, -0.0122], device='cuda:0'), grad: tensor([-0.0104, 0.0109, 0.0174, 0.0257, -0.0159, 0.0159, -0.0121, 0.0004, + 0.0036, -0.0355], device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 213.92, cls_loss 0.5707 cls_loss_mapping 0.0036 cls_loss_causal 0.4901 re_mapping 0.0106 re_causal 0.0250 /// teacc 98.69 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.0700, -0.1144, -0.0879, ..., -0.0344, 0.0511, -0.1041], + [-0.0704, -0.0802, -0.0659, ..., 0.1016, -0.0356, 0.2056], + [ 0.0039, -0.0159, -0.0308, ..., -0.0128, -0.0157, -0.0731], + ..., + [ 0.0285, -0.0999, 0.1485, ..., 0.0290, -0.0778, 0.0429], + [-0.0362, 0.0585, -0.1337, ..., -0.0522, 0.0141, -0.0815], + [-0.0517, 0.0645, 0.0473, ..., -0.0487, -0.0426, -0.0198]], + device='cuda:0'), grad: tensor([[ 1.4894e-05, 2.0891e-05, 4.9859e-05, ..., 2.0332e-03, + 1.2791e-04, 1.0548e-03], + [ 5.3197e-06, 3.1948e-05, 1.6344e-04, ..., -3.0899e-03, + 2.8744e-05, 1.7939e-03], + [ 1.2450e-05, 1.3161e-04, 3.0541e-04, ..., 1.9245e-03, + -6.1132e-06, 1.1034e-03], + ..., + [-1.0335e-04, 3.2485e-05, -1.5736e-03, ..., 1.0071e-03, + 2.3589e-05, -3.8099e-04], + [ 4.0717e-06, 3.6359e-04, 2.3174e-04, ..., -7.8888e-03, + 1.2007e-03, -7.0915e-03], + [ 6.5506e-05, 2.3976e-05, 6.7234e-04, ..., 1.8244e-03, + 4.2528e-05, 1.5345e-03]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0262, 0.0319, 0.0002, -0.0089, 0.0232, -0.0298, 0.0063, -0.0013, + -0.0006, -0.0127], device='cuda:0'), grad: tensor([ 0.0100, -0.0126, 0.0099, 0.0061, 0.0063, 0.0071, 0.0026, 0.0060, + -0.0460, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 214.15, cls_loss 0.5794 cls_loss_mapping 0.0046 cls_loss_causal 0.5053 re_mapping 0.0105 re_causal 0.0253 /// teacc 98.65 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.0706, -0.1147, -0.0887, ..., -0.0346, 0.0510, -0.1051], + [-0.0719, -0.0810, -0.0666, ..., 0.1016, -0.0356, 0.2066], + [ 0.0048, -0.0157, -0.0309, ..., -0.0126, -0.0166, -0.0738], + ..., + [ 0.0291, -0.1021, 0.1493, ..., 0.0287, -0.0786, 0.0425], + [-0.0362, 0.0589, -0.1345, ..., -0.0527, 0.0150, -0.0814], + [-0.0518, 0.0656, 0.0467, ..., -0.0487, -0.0427, -0.0196]], + device='cuda:0'), grad: tensor([[-7.9930e-05, 1.6773e-04, 1.1188e-04, ..., -2.7351e-03, + 8.7440e-05, 8.8811e-05], + [ 1.1295e-05, 5.1641e-04, 7.1669e-04, ..., 6.7043e-04, + 1.2684e-04, 8.7166e-04], + [ 2.1294e-05, 4.6682e-04, -8.6844e-05, ..., -3.2959e-03, + 1.0335e-04, -3.1662e-04], + ..., + [ 6.4746e-06, -8.2350e-04, -2.0351e-03, ..., 1.3351e-04, + 1.2970e-04, 1.0498e-02], + [-3.5614e-05, -9.6130e-03, 5.6791e-04, ..., -7.2861e-04, + 1.2290e-04, 8.5175e-05], + [ 1.6391e-05, 1.6708e-03, 6.0834e-06, ..., -2.0046e-03, + 8.7559e-05, 5.9471e-03]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0274, 0.0316, 0.0005, -0.0082, 0.0236, -0.0291, 0.0069, -0.0020, + -0.0011, -0.0124], device='cuda:0'), grad: tensor([ 0.0129, -0.0031, -0.0277, 0.0425, -0.0438, 0.0210, 0.0203, 0.0035, + -0.0292, 0.0035], device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.82, cls_loss 0.6129 cls_loss_mapping 0.0051 cls_loss_causal 0.5329 re_mapping 0.0104 re_causal 0.0245 /// teacc 98.58 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.0712, -0.1159, -0.0902, ..., -0.0359, 0.0514, -0.1059], + [-0.0719, -0.0820, -0.0674, ..., 0.1003, -0.0356, 0.2055], + [ 0.0051, -0.0147, -0.0312, ..., -0.0117, -0.0171, -0.0726], + ..., + [ 0.0296, -0.1006, 0.1500, ..., 0.0284, -0.0789, 0.0430], + [-0.0366, 0.0591, -0.1347, ..., -0.0516, 0.0160, -0.0821], + [-0.0524, 0.0662, 0.0473, ..., -0.0492, -0.0433, -0.0201]], + device='cuda:0'), grad: tensor([[ 4.9859e-05, 1.5497e-04, -2.6245e-03, ..., 1.5574e-03, + -1.6623e-03, -2.7332e-03], + [ 2.0355e-05, 9.0599e-05, 3.8624e-04, ..., 4.4327e-03, + 7.1943e-05, 1.9722e-03], + [ 2.8014e-04, 3.4595e-04, 2.2831e-03, ..., 6.6109e-03, + 2.8086e-04, 8.5688e-04], + ..., + [ 2.0355e-05, 2.8029e-05, 4.7255e-04, ..., -1.1168e-03, + 7.4387e-05, 4.0092e-03], + [ 2.8725e-03, 3.5515e-03, 9.8038e-04, ..., 2.8610e-03, + 3.3188e-04, -4.1542e-03], + [ 4.1938e-04, 3.8242e-03, -2.7161e-03, ..., -1.8753e-02, + 1.9634e-04, -2.0638e-03]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0286, 0.0311, 0.0006, -0.0079, 0.0235, -0.0284, 0.0070, -0.0018, + -0.0008, -0.0126], device='cuda:0'), grad: tensor([ 0.0045, 0.0224, 0.0296, -0.0446, 0.0248, 0.0092, -0.0345, 0.0165, + 0.0366, -0.0645], device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 215.05, cls_loss 0.5754 cls_loss_mapping 0.0051 cls_loss_causal 0.4988 re_mapping 0.0109 re_causal 0.0245 /// teacc 98.74 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.0715, -0.1157, -0.0901, ..., -0.0382, 0.0517, -0.1045], + [-0.0730, -0.0830, -0.0679, ..., 0.1010, -0.0360, 0.2054], + [ 0.0048, -0.0149, -0.0319, ..., -0.0120, -0.0144, -0.0727], + ..., + [ 0.0307, -0.0998, 0.1500, ..., 0.0294, -0.0796, 0.0424], + [-0.0355, 0.0586, -0.1352, ..., -0.0529, 0.0149, -0.0818], + [-0.0531, 0.0658, 0.0477, ..., -0.0476, -0.0430, -0.0200]], + device='cuda:0'), grad: tensor([[ 4.3601e-05, 7.3493e-05, 5.2977e-04, ..., -4.5280e-03, + 1.2901e-02, 7.5221e-05], + [ 2.1672e-04, -9.5367e-03, 5.2595e-04, ..., -3.0708e-03, + 1.5450e-04, -1.3008e-03], + [ 1.0042e-03, 1.1057e-04, 1.7834e-03, ..., 6.1226e-03, + 5.7888e-04, 1.1292e-03], + ..., + [-5.9967e-03, 1.7703e-04, 1.2083e-03, ..., 6.3934e-03, + 2.2519e-04, -8.6737e-04], + [ 5.2977e-04, 9.6512e-03, 9.0647e-04, ..., 7.3738e-03, + 5.0259e-04, 8.7404e-04], + [ 7.6370e-03, 1.7653e-03, -1.6508e-03, ..., -8.7738e-03, + 9.1028e-04, 3.3340e-03]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0296, 0.0311, 0.0017, -0.0084, 0.0230, -0.0284, 0.0069, -0.0010, + -0.0021, -0.0112], device='cuda:0'), grad: tensor([-0.0099, -0.0255, 0.0215, -0.0442, -0.0216, 0.0067, -0.0120, 0.0331, + 0.0423, 0.0095], device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 214.88, cls_loss 0.5958 cls_loss_mapping 0.0047 cls_loss_causal 0.5138 re_mapping 0.0110 re_causal 0.0250 /// teacc 98.69 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.0702, -0.1157, -0.0889, ..., -0.0391, 0.0525, -0.1033], + [-0.0725, -0.0829, -0.0666, ..., 0.1013, -0.0367, 0.2054], + [ 0.0034, -0.0159, -0.0307, ..., -0.0122, -0.0138, -0.0711], + ..., + [ 0.0306, -0.0984, 0.1494, ..., 0.0312, -0.0807, 0.0415], + [-0.0357, 0.0598, -0.1358, ..., -0.0523, 0.0154, -0.0812], + [-0.0542, 0.0657, 0.0464, ..., -0.0493, -0.0439, -0.0210]], + device='cuda:0'), grad: tensor([[-2.8181e-04, 1.9276e-04, 3.5256e-05, ..., -7.6790e-03, + 1.8060e-04, 7.4506e-09], + [ 9.1434e-05, 2.6867e-05, 2.5079e-05, ..., 5.6534e-03, + 6.5804e-05, 5.5879e-08], + [ 8.2588e-04, 4.9639e-04, 7.2622e-04, ..., 1.0353e-02, + 5.0879e-04, 2.6077e-08], + ..., + [ 8.5115e-05, 6.4909e-05, 7.4983e-05, ..., -3.3798e-03, + -3.9577e-04, 1.3784e-07], + [-9.9373e-04, 6.6161e-05, 7.6890e-05, ..., 7.0906e-04, + 8.9645e-05, 9.1270e-08], + [ 3.7694e-04, 2.3520e-04, 2.1601e-04, ..., -3.2673e-03, + 1.4579e-04, 7.4506e-07]], device='cuda:0') +Epoch 242, bias, value: tensor([-3.0170e-02, 3.1824e-02, 1.8377e-03, -9.3172e-03, 2.3220e-02, + -2.9180e-02, 7.8249e-03, 8.9777e-05, -1.8461e-03, -1.2339e-02], + device='cuda:0'), grad: tensor([-0.0230, 0.0270, 0.0348, 0.0250, 0.0120, -0.0424, 0.0211, -0.0144, + -0.0231, -0.0170], device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 214.49, cls_loss 0.5373 cls_loss_mapping 0.0046 cls_loss_causal 0.4704 re_mapping 0.0109 re_causal 0.0243 /// teacc 98.79 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.0717, -0.1167, -0.0898, ..., -0.0391, 0.0524, -0.1040], + [-0.0694, -0.0825, -0.0659, ..., 0.1015, -0.0388, 0.2054], + [ 0.0045, -0.0165, -0.0297, ..., -0.0116, -0.0124, -0.0696], + ..., + [ 0.0304, -0.0993, 0.1499, ..., 0.0312, -0.0812, 0.0420], + [-0.0358, 0.0590, -0.1363, ..., -0.0526, 0.0162, -0.0820], + [-0.0543, 0.0649, 0.0458, ..., -0.0506, -0.0446, -0.0212]], + device='cuda:0'), grad: tensor([[ 8.2627e-06, 1.0833e-05, 2.4080e-05, ..., 3.3131e-03, + 2.5578e-03, 3.0689e-03], + [ 4.9584e-06, 1.8275e-04, 3.0756e-05, ..., -5.4207e-03, + -4.6425e-03, -3.7537e-03], + [-9.0504e-04, 3.8719e-03, -8.6689e-04, ..., 6.4583e-03, + 5.7945e-03, -3.5381e-03], + ..., + [ 7.0429e-04, 2.2721e-04, 8.0347e-04, ..., 4.0321e-03, + 1.4286e-03, 1.2913e-03], + [ 3.7241e-04, -3.6297e-03, 1.7157e-03, ..., -9.2697e-03, + 1.5545e-03, 1.5831e-03], + [ 7.0512e-05, 6.6137e-04, 2.3060e-06, ..., -5.2719e-03, + -7.4654e-03, 5.6028e-04]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0299, 0.0318, 0.0018, -0.0094, 0.0230, -0.0277, 0.0066, 0.0002, + -0.0009, -0.0136], device='cuda:0'), grad: tensor([ 0.0132, -0.0222, 0.0095, 0.0056, 0.0196, 0.0320, -0.0234, 0.0246, + -0.0311, -0.0278], device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 214.67, cls_loss 0.5616 cls_loss_mapping 0.0051 cls_loss_causal 0.5045 re_mapping 0.0103 re_causal 0.0239 /// teacc 98.68 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.0718, -0.1132, -0.0903, ..., -0.0389, 0.0530, -0.1038], + [-0.0692, -0.0835, -0.0664, ..., 0.1010, -0.0362, 0.2054], + [ 0.0048, -0.0172, -0.0303, ..., -0.0119, -0.0124, -0.0699], + ..., + [ 0.0301, -0.0995, 0.1512, ..., 0.0312, -0.0806, 0.0429], + [-0.0359, 0.0586, -0.1362, ..., -0.0527, 0.0152, -0.0832], + [-0.0541, 0.0659, 0.0451, ..., -0.0501, -0.0409, -0.0205]], + device='cuda:0'), grad: tensor([[ 3.7253e-08, 3.8052e-04, 3.4504e-03, ..., 1.3247e-03, + 6.3360e-05, -1.2978e-02], + [ 8.5682e-08, 6.2132e-04, 7.6103e-04, ..., 2.2717e-03, + 1.8060e-04, 1.3990e-03], + [ 4.3027e-07, 7.0095e-04, -1.4252e-02, ..., -6.4659e-03, + 2.7204e-04, -3.7575e-03], + ..., + [ 6.1654e-07, -3.4475e-04, 2.1839e-03, ..., 1.3428e-03, + -1.9801e-04, -3.9315e-04], + [ 4.4703e-07, -4.0779e-03, -9.8801e-04, ..., -9.9869e-03, + -1.0023e-03, 2.4662e-03], + [ 8.6986e-07, 1.4267e-03, 2.6360e-03, ..., 3.4637e-03, + 3.9053e-04, 3.6030e-03]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0295, 0.0323, 0.0014, -0.0097, 0.0233, -0.0288, 0.0066, 0.0010, + -0.0008, -0.0139], device='cuda:0'), grad: tensor([-0.0242, 0.0115, -0.0415, 0.0107, -0.0053, 0.0259, 0.0285, 0.0122, + -0.0367, 0.0190], device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 214.57, cls_loss 0.5822 cls_loss_mapping 0.0049 cls_loss_causal 0.5070 re_mapping 0.0106 re_causal 0.0253 /// teacc 98.72 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.0708, -0.1147, -0.0911, ..., -0.0393, 0.0526, -0.1047], + [-0.0685, -0.0822, -0.0662, ..., 0.1007, -0.0345, 0.2062], + [ 0.0048, -0.0179, -0.0303, ..., -0.0128, -0.0130, -0.0692], + ..., + [ 0.0304, -0.0998, 0.1513, ..., 0.0305, -0.0804, 0.0426], + [-0.0371, 0.0589, -0.1364, ..., -0.0525, 0.0138, -0.0845], + [-0.0548, 0.0654, 0.0454, ..., -0.0515, -0.0405, -0.0205]], + device='cuda:0'), grad: tensor([[ 1.5676e-05, 1.6809e-04, 5.2547e-04, ..., 3.1223e-03, + 6.4707e-04, 9.2888e-04], + [ 4.6678e-06, 6.2084e-04, 6.1321e-04, ..., -6.5689e-03, + -1.1843e-04, -6.7177e-03], + [ 6.2361e-06, 3.1781e-04, -2.4872e-03, ..., -1.1505e-02, + -2.3384e-03, 1.2960e-03], + ..., + [ 2.1961e-06, 1.7347e-03, 3.3402e-04, ..., 2.9850e-03, + 2.0733e-03, 2.0313e-03], + [-5.9664e-05, -1.5154e-03, 6.0987e-04, ..., 3.5038e-03, + 4.3011e-04, 1.3494e-03], + [ 1.3724e-05, -4.5929e-03, -3.0346e-03, ..., 2.6302e-03, + 9.4056e-05, -3.5501e-04]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0293, 0.0331, 0.0001, -0.0079, 0.0225, -0.0287, 0.0063, 0.0007, + -0.0004, -0.0145], device='cuda:0'), grad: tensor([ 0.0155, -0.0155, -0.0496, 0.0151, 0.0066, -0.0101, -0.0109, 0.0181, + 0.0186, 0.0121], device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 214.42, cls_loss 0.5706 cls_loss_mapping 0.0048 cls_loss_causal 0.4838 re_mapping 0.0103 re_causal 0.0229 /// teacc 98.72 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.0717, -0.1140, -0.0915, ..., -0.0400, 0.0536, -0.1049], + [-0.0688, -0.0840, -0.0664, ..., 0.1006, -0.0348, 0.2075], + [ 0.0049, -0.0172, -0.0308, ..., -0.0119, -0.0118, -0.0695], + ..., + [ 0.0304, -0.1001, 0.1522, ..., 0.0299, -0.0815, 0.0425], + [-0.0369, 0.0597, -0.1362, ..., -0.0521, 0.0134, -0.0852], + [-0.0544, 0.0646, 0.0452, ..., -0.0517, -0.0422, -0.0203]], + device='cuda:0'), grad: tensor([[ 1.6654e-04, 2.1136e-04, 3.0637e-05, ..., 5.2185e-03, + 4.1699e-04, 2.3580e-04], + [ 4.0746e-04, 1.3959e-04, 1.5080e-04, ..., -1.4915e-02, + 1.5423e-05, 3.3140e-05], + [ 3.6573e-04, 4.3941e-04, 4.0221e-04, ..., 1.0727e-02, + 2.2352e-04, 2.8706e-04], + ..., + [ 2.1768e-04, 3.5143e-04, 2.4939e-04, ..., 3.9101e-03, + 6.1572e-05, 3.4785e-04], + [ 2.3365e-04, 6.8378e-04, 5.8365e-04, ..., -2.8419e-03, + 3.5906e-04, 4.9877e-04], + [-1.5326e-03, 2.0332e-03, 2.9583e-03, ..., -1.2215e-02, + -1.5554e-03, 2.2984e-03]], device='cuda:0') +Epoch 246, bias, value: tensor([-3.0795e-02, 3.3426e-02, 1.6408e-03, -6.8361e-03, 2.2200e-02, + -2.9178e-02, 6.2983e-03, -1.6305e-05, 9.1407e-05, -1.5115e-02], + device='cuda:0'), grad: tensor([ 0.0163, -0.0177, 0.0255, 0.0109, 0.0010, -0.0187, 0.0160, 0.0134, + -0.0126, -0.0342], device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 214.82, cls_loss 0.5477 cls_loss_mapping 0.0047 cls_loss_causal 0.4680 re_mapping 0.0106 re_causal 0.0226 /// teacc 98.69 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.0715, -0.1138, -0.0923, ..., -0.0407, 0.0547, -0.1051], + [-0.0694, -0.0843, -0.0668, ..., 0.0998, -0.0354, 0.2075], + [ 0.0059, -0.0173, -0.0309, ..., -0.0117, -0.0118, -0.0704], + ..., + [ 0.0316, -0.0991, 0.1535, ..., 0.0315, -0.0818, 0.0432], + [-0.0370, 0.0591, -0.1369, ..., -0.0515, 0.0135, -0.0847], + [-0.0556, 0.0665, 0.0449, ..., -0.0532, -0.0418, -0.0207]], + device='cuda:0'), grad: tensor([[ 1.6201e-04, 8.3828e-04, 1.9002e-04, ..., 2.8515e-03, + 5.6401e-06, 3.0613e-03], + [ 2.6718e-05, 9.0301e-05, 4.8757e-04, ..., -8.9798e-03, + 3.3155e-07, -1.3313e-03], + [ 1.8334e-04, 4.8280e-04, 3.0875e-04, ..., -2.3670e-03, + 7.8440e-05, 1.0443e-03], + ..., + [ 4.2826e-05, 2.9969e-04, -2.0905e-03, ..., -2.0599e-03, + 8.7619e-06, 1.8330e-03], + [ 2.4033e-04, 6.4135e-04, 3.4451e-04, ..., 4.2381e-03, + 2.3097e-05, 1.4753e-03], + [ 1.2326e-04, -6.6662e-04, 6.0606e-04, ..., -8.1825e-04, + 1.6224e-06, -8.9035e-03]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0312, 0.0325, 0.0012, -0.0069, 0.0225, -0.0286, 0.0061, 0.0011, + 0.0003, -0.0154], device='cuda:0'), grad: tensor([ 0.0175, -0.0448, -0.0108, 0.0078, -0.0017, 0.0206, 0.0219, 0.0025, + 0.0219, -0.0349], device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.98, cls_loss 0.5609 cls_loss_mapping 0.0045 cls_loss_causal 0.4780 re_mapping 0.0104 re_causal 0.0243 /// teacc 98.74 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.0706, -0.1145, -0.0934, ..., -0.0397, 0.0546, -0.1060], + [-0.0695, -0.0851, -0.0671, ..., 0.1000, -0.0360, 0.2074], + [ 0.0050, -0.0161, -0.0307, ..., -0.0120, -0.0115, -0.0705], + ..., + [ 0.0337, -0.0990, 0.1544, ..., 0.0309, -0.0831, 0.0433], + [-0.0375, 0.0588, -0.1374, ..., -0.0514, 0.0126, -0.0833], + [-0.0567, 0.0661, 0.0447, ..., -0.0549, -0.0414, -0.0220]], + device='cuda:0'), grad: tensor([[ 1.4174e-04, 1.1724e-04, 1.4372e-05, ..., 5.1498e-03, + 1.9133e-04, 1.3220e-04], + [ 7.9393e-04, 1.1265e-04, 1.3955e-05, ..., 6.0310e-03, + 1.4710e-04, 1.3981e-03], + [ 9.8896e-04, 1.8716e-04, 5.6922e-05, ..., -1.3741e-02, + 2.0027e-04, 4.8804e-04], + ..., + [-6.7949e-05, 1.0049e-04, -2.0504e-04, ..., -2.9964e-03, + -1.6727e-03, -1.3103e-03], + [ 5.8508e-04, 7.0429e-04, 6.2168e-05, ..., 6.2904e-03, + 1.5831e-04, 6.4898e-04], + [ 2.3627e-04, 1.3094e-03, 3.4738e-04, ..., 6.6719e-03, + 1.6963e-04, 9.8515e-04]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0314, 0.0333, 0.0011, -0.0060, 0.0232, -0.0295, 0.0056, 0.0009, + 0.0005, -0.0161], device='cuda:0'), grad: tensor([ 0.0220, 0.0007, -0.0276, -0.0007, -0.0294, 0.0136, 0.0053, -0.0142, + 0.0021, 0.0282], device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 214.85, cls_loss 0.5726 cls_loss_mapping 0.0052 cls_loss_causal 0.5064 re_mapping 0.0101 re_causal 0.0236 /// teacc 98.73 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.0693, -0.1133, -0.0931, ..., -0.0400, 0.0555, -0.1049], + [-0.0695, -0.0851, -0.0680, ..., 0.0993, -0.0360, 0.2075], + [ 0.0036, -0.0171, -0.0314, ..., -0.0111, -0.0127, -0.0703], + ..., + [ 0.0337, -0.0976, 0.1548, ..., 0.0312, -0.0835, 0.0445], + [-0.0375, 0.0587, -0.1378, ..., -0.0523, 0.0134, -0.0830], + [-0.0571, 0.0678, 0.0458, ..., -0.0545, -0.0418, -0.0236]], + device='cuda:0'), grad: tensor([[ 2.5593e-06, 1.4591e-03, 1.6127e-03, ..., 5.0497e-04, + 2.9707e-04, 2.3711e-04], + [ 1.3039e-08, 2.6393e-04, 1.2949e-05, ..., -2.9507e-03, + 2.2464e-06, -2.8774e-05], + [ 5.2154e-08, 4.0126e-04, 1.8454e-04, ..., 2.9259e-03, + 4.4852e-05, 7.8261e-05], + ..., + [ 2.9802e-08, -8.7738e-03, 8.1110e-04, ..., -3.0689e-03, + 2.0397e-04, 2.0218e-04], + [ 3.5204e-07, 5.1231e-03, 1.4753e-03, ..., 4.4746e-03, + 4.0364e-04, 2.5511e-04], + [ 1.6019e-07, -2.4414e-03, -8.5602e-03, ..., -1.2444e-02, + -2.0809e-03, -1.0309e-03]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0310, 0.0329, 0.0012, -0.0062, 0.0237, -0.0293, 0.0052, 0.0005, + 0.0008, -0.0161], device='cuda:0'), grad: tensor([-0.0041, -0.0243, 0.0176, 0.0201, 0.0203, 0.0162, 0.0075, -0.0045, + 0.0165, -0.0653], device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 214.38, cls_loss 0.5590 cls_loss_mapping 0.0037 cls_loss_causal 0.4835 re_mapping 0.0104 re_causal 0.0244 /// teacc 98.69 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.0694, -0.1139, -0.0919, ..., -0.0394, 0.0558, -0.1044], + [-0.0699, -0.0855, -0.0680, ..., 0.0993, -0.0362, 0.2069], + [ 0.0037, -0.0167, -0.0309, ..., -0.0112, -0.0123, -0.0706], + ..., + [ 0.0336, -0.0974, 0.1559, ..., 0.0303, -0.0841, 0.0449], + [-0.0373, 0.0578, -0.1395, ..., -0.0521, 0.0122, -0.0817], + [-0.0567, 0.0675, 0.0454, ..., -0.0534, -0.0411, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.1809e-06, 1.7142e-04, 8.2135e-05, ..., -9.1171e-04, + 6.9104e-07, 2.3305e-04], + [ 6.0797e-05, 2.4071e-03, 4.6659e-04, ..., 1.5656e-02, + 1.4659e-06, 2.1610e-03], + [-1.3448e-06, 3.0351e-04, 3.9625e-04, ..., 3.8319e-03, + 2.1644e-06, 2.8372e-04], + ..., + [ 6.1318e-06, 3.3450e-04, 1.0395e-03, ..., 4.8943e-03, + 4.1686e-06, 1.7090e-03], + [ 6.8545e-05, 1.7014e-03, 5.0640e-04, ..., 7.3385e-04, + 1.9502e-06, -6.2561e-04], + [-1.0610e-04, -3.2101e-03, 2.1439e-03, ..., 5.6877e-03, + 1.0028e-05, 3.0079e-03]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0301, 0.0332, 0.0009, -0.0062, 0.0237, -0.0295, 0.0045, -0.0003, + 0.0004, -0.0149], device='cuda:0'), grad: tensor([-0.0194, 0.0559, 0.0179, -0.0793, -0.0168, 0.0098, -0.0101, 0.0236, + -0.0027, 0.0209], device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 214.21, cls_loss 0.5649 cls_loss_mapping 0.0054 cls_loss_causal 0.4907 re_mapping 0.0101 re_causal 0.0234 /// teacc 98.57 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.0685, -0.1149, -0.0926, ..., -0.0382, 0.0551, -0.1050], + [-0.0702, -0.0848, -0.0678, ..., 0.1000, -0.0368, 0.2077], + [ 0.0047, -0.0179, -0.0300, ..., -0.0113, -0.0130, -0.0705], + ..., + [ 0.0326, -0.0970, 0.1552, ..., 0.0298, -0.0851, 0.0453], + [-0.0375, 0.0581, -0.1392, ..., -0.0529, 0.0119, -0.0813], + [-0.0560, 0.0666, 0.0472, ..., -0.0537, -0.0416, -0.0236]], + device='cuda:0'), grad: tensor([[ 3.5167e-04, 1.4651e-04, 1.4651e-04, ..., 4.3983e-03, + 5.1069e-04, 3.3766e-05], + [ 2.2292e-04, 1.1498e-04, -8.3771e-03, ..., -4.8256e-03, + 8.3399e-04, -5.5161e-03], + [ 3.5882e-04, 7.4565e-05, 3.3116e-04, ..., 2.9240e-03, + 2.9469e-04, 1.4699e-04], + ..., + [-4.8065e-03, 4.3583e-04, 8.9188e-03, ..., 1.6785e-03, + 3.6359e-04, 5.5695e-03], + [-2.4796e-04, 2.8086e-04, -2.5311e-03, ..., 1.1768e-03, + -3.7804e-03, 1.2040e-04], + [ 1.7252e-03, 1.0862e-03, 3.1033e-03, ..., 8.8501e-03, + 1.1091e-03, 1.7824e-03]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0296, 0.0344, 0.0007, -0.0059, 0.0242, -0.0302, 0.0049, -0.0012, + -0.0004, -0.0151], device='cuda:0'), grad: tensor([ 0.0144, 0.0031, 0.0057, -0.0361, -0.0151, -0.0326, 0.0427, 0.0030, + -0.0111, 0.0261], device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 214.20, cls_loss 0.5404 cls_loss_mapping 0.0051 cls_loss_causal 0.4746 re_mapping 0.0099 re_causal 0.0232 /// teacc 98.59 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.0684, -0.1143, -0.0912, ..., -0.0386, 0.0549, -0.1048], + [-0.0698, -0.0848, -0.0679, ..., 0.1005, -0.0383, 0.2083], + [ 0.0046, -0.0177, -0.0317, ..., -0.0104, -0.0137, -0.0703], + ..., + [ 0.0330, -0.0959, 0.1540, ..., 0.0285, -0.0851, 0.0447], + [-0.0388, 0.0578, -0.1387, ..., -0.0525, 0.0122, -0.0819], + [-0.0558, 0.0667, 0.0485, ..., -0.0529, -0.0403, -0.0235]], + device='cuda:0'), grad: tensor([[ 7.6368e-08, 8.7991e-06, 1.2510e-05, ..., -5.3711e-03, + 8.1587e-04, 4.0263e-05], + [ 6.9477e-07, 6.1417e-04, 4.5228e-04, ..., 5.6725e-03, + 1.2052e-04, 4.9353e-04], + [ 1.1735e-06, -7.2384e-04, -4.6825e-04, ..., -3.3302e-03, + -2.5482e-03, 1.0484e-04], + ..., + [ 1.2293e-06, 3.0175e-05, 3.4776e-06, ..., 1.4610e-03, + 3.7074e-04, -1.1005e-03], + [ 2.7940e-08, 2.2724e-05, 3.4302e-05, ..., 1.1978e-03, + 6.9761e-04, 1.3590e-04], + [ 1.7375e-05, 2.0695e-04, 3.1281e-04, ..., 1.0509e-03, + 1.8144e-04, 1.5392e-03]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0294, 0.0345, 0.0012, -0.0070, 0.0238, -0.0297, 0.0052, -0.0024, + -0.0001, -0.0144], device='cuda:0'), grad: tensor([-0.0191, 0.0200, -0.0117, 0.0083, -0.0303, 0.0431, -0.0146, 0.0086, + 0.0096, -0.0138], device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 214.72, cls_loss 0.5577 cls_loss_mapping 0.0043 cls_loss_causal 0.4949 re_mapping 0.0100 re_causal 0.0236 /// teacc 98.73 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.0690, -0.1149, -0.0917, ..., -0.0388, 0.0564, -0.1054], + [-0.0704, -0.0846, -0.0681, ..., 0.1012, -0.0392, 0.2091], + [ 0.0040, -0.0177, -0.0326, ..., -0.0107, -0.0141, -0.0708], + ..., + [ 0.0313, -0.0970, 0.1540, ..., 0.0285, -0.0861, 0.0452], + [-0.0376, 0.0580, -0.1389, ..., -0.0532, 0.0132, -0.0826], + [-0.0543, 0.0673, 0.0492, ..., -0.0528, -0.0408, -0.0226]], + device='cuda:0'), grad: tensor([[ 7.3493e-05, -1.3220e-04, -2.5463e-03, ..., 4.2653e-04, + 7.3671e-05, -1.1730e-03], + [ 5.7966e-05, 1.2964e-06, 9.9361e-05, ..., 1.3290e-02, + 3.0422e-04, 1.8811e-04], + [ 8.9109e-05, 3.8862e-05, 9.4128e-04, ..., -1.0582e-02, + 1.5430e-03, 8.4829e-04], + ..., + [ 4.6635e-04, 3.7879e-05, 1.1911e-03, ..., 3.4771e-03, + 4.9543e-04, 7.8583e-04], + [ 4.8113e-04, 2.4676e-05, 5.9557e-04, ..., 5.0888e-03, + 8.2445e-04, 9.8515e-04], + [ 8.0395e-04, -2.5868e-05, 1.3628e-03, ..., -6.9275e-03, + 4.0531e-04, 1.3351e-03]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0288, 0.0348, 0.0010, -0.0070, 0.0238, -0.0295, 0.0054, -0.0028, + -0.0010, -0.0141], device='cuda:0'), grad: tensor([ 0.0071, 0.0384, -0.0383, 0.0184, -0.0419, -0.0133, -0.0042, 0.0169, + 0.0252, -0.0082], device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 214.56, cls_loss 0.5469 cls_loss_mapping 0.0033 cls_loss_causal 0.4645 re_mapping 0.0104 re_causal 0.0233 /// teacc 98.70 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.0690, -0.1157, -0.0922, ..., -0.0390, 0.0561, -0.1061], + [-0.0701, -0.0857, -0.0690, ..., 0.1006, -0.0399, 0.2096], + [ 0.0039, -0.0174, -0.0332, ..., -0.0103, -0.0132, -0.0715], + ..., + [ 0.0291, -0.0966, 0.1549, ..., 0.0292, -0.0859, 0.0453], + [-0.0384, 0.0586, -0.1391, ..., -0.0532, 0.0128, -0.0823], + [-0.0526, 0.0675, 0.0487, ..., -0.0513, -0.0415, -0.0228]], + device='cuda:0'), grad: tensor([[-2.3746e-04, 2.4468e-05, 1.0334e-05, ..., 9.2793e-04, + -8.7118e-04, 3.8862e-05], + [ 7.0110e-06, 9.5367e-05, 2.7347e-04, ..., 6.9199e-03, + 2.2459e-04, 2.9430e-03], + [ 4.4608e-04, 6.6853e-04, 2.1255e-04, ..., 5.5275e-03, + 4.8423e-04, 1.8244e-03], + ..., + [ 7.1973e-06, 9.0718e-05, -6.3229e-04, ..., 3.1681e-03, + 7.4327e-05, -1.0933e-02], + [-7.5626e-04, -5.1308e-04, -2.8133e-04, ..., -1.2360e-03, + 4.6992e-04, 7.6256e-03], + [-3.4899e-05, -4.3011e-04, 2.1517e-04, ..., 1.2267e-04, + -1.7405e-04, 3.6659e-03]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0293, 0.0339, 0.0011, -0.0069, 0.0237, -0.0297, 0.0051, -0.0028, + -0.0005, -0.0130], device='cuda:0'), grad: tensor([-0.0025, 0.0237, 0.0248, 0.0179, -0.0491, 0.0002, 0.0114, -0.0116, + -0.0092, -0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 214.35, cls_loss 0.5553 cls_loss_mapping 0.0038 cls_loss_causal 0.4791 re_mapping 0.0102 re_causal 0.0234 /// teacc 98.77 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.0695, -0.1161, -0.0941, ..., -0.0386, 0.0555, -0.1067], + [-0.0699, -0.0863, -0.0694, ..., 0.1002, -0.0403, 0.2105], + [ 0.0046, -0.0170, -0.0331, ..., -0.0106, -0.0130, -0.0731], + ..., + [ 0.0299, -0.0968, 0.1548, ..., 0.0299, -0.0867, 0.0469], + [-0.0395, 0.0593, -0.1388, ..., -0.0541, 0.0137, -0.0828], + [-0.0526, 0.0681, 0.0499, ..., -0.0513, -0.0408, -0.0240]], + device='cuda:0'), grad: tensor([[ 1.6415e-04, 1.8620e-04, -4.3297e-04, ..., -3.1872e-03, + 2.0707e-04, -6.9618e-04], + [ 4.6158e-04, 2.6417e-04, 1.1215e-03, ..., 4.4060e-03, + 1.4687e-04, 1.8396e-03], + [ 3.5133e-03, -1.6975e-04, -1.4000e-03, ..., 7.6256e-03, + -2.5787e-03, 2.2602e-03], + ..., + [-4.9286e-03, 1.0319e-03, 3.5076e-03, ..., -3.2482e-03, + 4.1246e-04, 4.5357e-03], + [ 8.8692e-05, -1.4591e-04, -2.4624e-03, ..., -4.1809e-03, + 8.6927e-04, -4.8981e-03], + [-4.2096e-06, 7.6532e-05, 1.3189e-03, ..., 4.7646e-03, + 3.7122e-04, 2.5730e-03]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0297, 0.0334, 0.0012, -0.0066, 0.0233, -0.0286, 0.0052, -0.0023, + -0.0009, -0.0133], device='cuda:0'), grad: tensor([-0.0085, -0.0064, 0.0247, 0.0304, -0.0204, 0.0009, -0.0375, -0.0121, + -0.0019, 0.0310], device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 214.83, cls_loss 0.5811 cls_loss_mapping 0.0057 cls_loss_causal 0.5104 re_mapping 0.0100 re_causal 0.0236 /// teacc 98.80 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.0697, -0.1180, -0.0943, ..., -0.0382, 0.0553, -0.1078], + [-0.0708, -0.0864, -0.0700, ..., 0.1000, -0.0407, 0.2111], + [ 0.0030, -0.0174, -0.0334, ..., -0.0113, -0.0131, -0.0729], + ..., + [ 0.0326, -0.0973, 0.1545, ..., 0.0304, -0.0866, 0.0474], + [-0.0406, 0.0601, -0.1368, ..., -0.0545, 0.0141, -0.0835], + [-0.0536, 0.0680, 0.0495, ..., -0.0509, -0.0416, -0.0231]], + device='cuda:0'), grad: tensor([[ 5.6531e-07, 2.6729e-07, 3.8475e-05, ..., 1.9302e-03, + 9.1642e-07, 6.7174e-05], + [ 1.8878e-06, 1.0151e-06, 6.2823e-05, ..., 2.9125e-03, + 2.0489e-07, 8.4579e-05], + [-8.2329e-06, 3.0899e-04, 1.2550e-03, ..., 2.0905e-03, + 4.8950e-06, 5.5122e-04], + ..., + [ 5.5507e-06, -2.9826e-04, -1.4000e-03, ..., 4.4136e-03, + -1.7300e-05, 1.9035e-03], + [-6.3423e-07, 6.2466e-05, 2.6870e-04, ..., 2.3746e-03, + 4.5728e-07, -3.5882e-04], + [ 7.4506e-07, -1.6224e-04, -1.0467e-04, ..., -6.7978e-03, + 5.4613e-06, 3.8052e-04]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0303, 0.0334, 0.0019, -0.0059, 0.0234, -0.0294, 0.0054, -0.0023, + -0.0013, -0.0132], device='cuda:0'), grad: tensor([ 0.0161, 0.0205, 0.0183, -0.0444, 0.0166, -0.0071, -0.0076, 0.0257, + 0.0032, -0.0414], device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 214.95, cls_loss 0.5336 cls_loss_mapping 0.0043 cls_loss_causal 0.4542 re_mapping 0.0099 re_causal 0.0216 /// teacc 98.79 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.0700, -0.1179, -0.0939, ..., -0.0391, 0.0557, -0.1077], + [-0.0714, -0.0859, -0.0698, ..., 0.0991, -0.0407, 0.2112], + [ 0.0021, -0.0163, -0.0323, ..., -0.0107, -0.0123, -0.0747], + ..., + [ 0.0326, -0.0986, 0.1533, ..., 0.0302, -0.0888, 0.0476], + [-0.0401, 0.0602, -0.1370, ..., -0.0554, 0.0140, -0.0841], + [-0.0545, 0.0677, 0.0499, ..., -0.0514, -0.0424, -0.0222]], + device='cuda:0'), grad: tensor([[ 4.9956e-06, 1.2755e-04, 1.2922e-04, ..., 1.1122e-04, + 1.5604e-04, 3.2067e-05], + [-1.9569e-03, -1.3676e-03, 8.5533e-05, ..., -4.1771e-03, + 1.6168e-06, -4.1122e-03], + [ 1.5005e-05, -1.1367e-04, 3.7842e-03, ..., -5.8174e-03, + 6.0272e-04, 9.2447e-05], + ..., + [ 2.7370e-04, 1.8442e-04, -4.8866e-03, ..., 4.8447e-03, + 2.2113e-04, 7.7581e-04], + [ 1.3864e-04, 2.4819e-04, 2.0790e-04, ..., 3.2253e-03, + 1.0920e-04, 3.9291e-04], + [ 1.0214e-03, 1.1597e-03, -1.4420e-03, ..., 1.7004e-03, + 9.2268e-05, 2.1820e-03]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0306, 0.0339, 0.0022, -0.0055, 0.0233, -0.0300, 0.0060, -0.0025, + -0.0020, -0.0133], device='cuda:0'), grad: tensor([-0.0139, -0.0070, -0.0050, 0.0252, 0.0223, -0.0363, -0.0256, 0.0227, + 0.0202, -0.0025], device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 214.87, cls_loss 0.5722 cls_loss_mapping 0.0056 cls_loss_causal 0.5064 re_mapping 0.0102 re_causal 0.0229 /// teacc 98.79 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.0716, -0.1182, -0.0931, ..., -0.0401, 0.0574, -0.1072], + [-0.0712, -0.0854, -0.0707, ..., 0.0993, -0.0414, 0.2117], + [ 0.0015, -0.0156, -0.0325, ..., -0.0112, -0.0116, -0.0753], + ..., + [ 0.0330, -0.0982, 0.1547, ..., 0.0306, -0.0894, 0.0478], + [-0.0405, 0.0603, -0.1371, ..., -0.0544, 0.0135, -0.0845], + [-0.0536, 0.0667, 0.0495, ..., -0.0514, -0.0404, -0.0218]], + device='cuda:0'), grad: tensor([[ 5.1588e-05, 1.1051e-04, 1.5509e-04, ..., 1.1644e-03, + 4.5228e-04, 1.7655e-04], + [ 2.2447e-04, 2.7633e-04, 6.8235e-04, ..., 3.0670e-03, + 5.1212e-04, -1.3599e-03], + [ 1.1158e-04, 1.6057e-04, 5.6934e-04, ..., 2.2297e-03, + 2.7418e-04, 1.7494e-05], + ..., + [ 4.1276e-05, 5.9038e-05, -3.0670e-03, ..., 2.8038e-04, + -2.3708e-03, 1.9479e-04], + [-9.8991e-04, -2.1458e-03, -2.7313e-03, ..., -8.3466e-03, + -1.2960e-03, 1.0319e-03], + [ 5.4806e-05, 8.9943e-05, 6.5327e-04, ..., 6.8474e-04, + 8.9836e-04, 1.7479e-05]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0296, 0.0342, 0.0014, -0.0057, 0.0243, -0.0309, 0.0060, -0.0023, + -0.0019, -0.0139], device='cuda:0'), grad: tensor([ 0.0204, -0.0178, 0.0224, 0.0347, 0.0191, 0.0089, -0.0105, -0.0301, + -0.0379, -0.0092], device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 214.45, cls_loss 0.5553 cls_loss_mapping 0.0043 cls_loss_causal 0.4856 re_mapping 0.0104 re_causal 0.0246 /// teacc 98.79 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.0716, -0.1184, -0.0915, ..., -0.0395, 0.0563, -0.1059], + [-0.0715, -0.0854, -0.0713, ..., 0.0985, -0.0414, 0.2118], + [ 0.0007, -0.0160, -0.0333, ..., -0.0120, -0.0111, -0.0760], + ..., + [ 0.0345, -0.0980, 0.1545, ..., 0.0307, -0.0910, 0.0481], + [-0.0410, 0.0603, -0.1379, ..., -0.0531, 0.0115, -0.0842], + [-0.0547, 0.0668, 0.0508, ..., -0.0508, -0.0407, -0.0219]], + device='cuda:0'), grad: tensor([[ 1.5795e-05, 2.4819e-04, 7.1339e-06, ..., 3.6693e-04, + 8.4457e-03, 4.4298e-04], + [ 9.0003e-06, 5.5820e-05, -2.7288e-07, ..., 3.4142e-04, + 7.9298e-04, 1.1696e-02], + [ 6.0797e-05, 4.5967e-04, 1.8075e-05, ..., 7.1859e-04, + 2.0657e-03, -1.5350e-02], + ..., + [ 6.1274e-05, 3.3975e-04, 3.2544e-05, ..., 1.0357e-03, + 5.3072e-04, 1.9970e-03], + [-7.6056e-04, -3.3417e-03, -1.2022e-04, ..., -6.3019e-03, + 1.8875e-02, 6.2346e-05], + [ 1.1295e-04, 2.0134e-04, 6.8963e-05, ..., 1.3981e-03, + 1.3962e-03, 6.2799e-04]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0280, 0.0328, -0.0004, -0.0059, 0.0248, -0.0311, 0.0057, -0.0014, + -0.0011, -0.0136], device='cuda:0'), grad: tensor([-0.0523, 0.0323, -0.0156, -0.0257, -0.0167, 0.0336, 0.0339, -0.0082, + 0.0062, 0.0125], device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 214.53, cls_loss 0.5592 cls_loss_mapping 0.0043 cls_loss_causal 0.4812 re_mapping 0.0105 re_causal 0.0235 /// teacc 98.77 lr 0.00010000 +Epoch 260, weight, value: tensor([[-6.9996e-02, -1.1878e-01, -9.2139e-02, ..., -3.9649e-02, + 5.6356e-02, -1.0356e-01], + [-7.1794e-02, -8.5669e-02, -7.0546e-02, ..., 9.9003e-02, + -3.9067e-02, 2.1214e-01], + [ 5.9977e-05, -1.6503e-02, -3.2933e-02, ..., -1.1230e-02, + -1.2390e-02, -7.5796e-02], + ..., + [ 3.5977e-02, -9.7729e-02, 1.5377e-01, ..., 2.9928e-02, + -9.0320e-02, 4.8165e-02], + [-3.9652e-02, 5.9831e-02, -1.3623e-01, ..., -5.3923e-02, + 1.1384e-02, -8.3473e-02], + [-5.4399e-02, 6.6018e-02, 5.1118e-02, ..., -5.1551e-02, + -4.0772e-02, -2.1985e-02]], device='cuda:0'), grad: tensor([[ 2.9616e-07, 6.8474e-04, 1.7948e-03, ..., 1.9989e-03, + 1.1005e-03, 1.0371e-04], + [ 1.6734e-05, 7.0393e-05, 1.7309e-04, ..., -1.2493e-03, + 1.8585e-04, -3.4595e-04], + [ 3.4850e-06, 5.8632e-03, 3.6502e-04, ..., 8.9417e-03, + 3.7169e-04, 6.0350e-05], + ..., + [-5.6362e-04, 2.3127e-04, -3.1891e-03, ..., -6.2990e-04, + 5.7077e-04, -3.6640e-03], + [ 9.8944e-06, 1.9550e-03, 4.9896e-03, ..., 4.9133e-03, + 6.2180e-03, 3.8886e-04], + [ 9.6142e-05, -1.5440e-03, 1.5488e-03, ..., 1.8015e-03, + 1.0815e-03, 6.6042e-04]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0279, 0.0326, -0.0003, -0.0052, 0.0253, -0.0311, 0.0063, -0.0016, + -0.0016, -0.0146], device='cuda:0'), grad: tensor([ 0.0158, -0.0185, -0.0010, -0.0432, 0.0026, 0.0161, -0.0142, 0.0015, + 0.0435, -0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 214.75, cls_loss 0.5284 cls_loss_mapping 0.0051 cls_loss_causal 0.4593 re_mapping 0.0101 re_causal 0.0224 /// teacc 98.70 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.0696, -0.1185, -0.0920, ..., -0.0396, 0.0571, -0.1037], + [-0.0722, -0.0870, -0.0703, ..., 0.0980, -0.0403, 0.2140], + [ 0.0003, -0.0166, -0.0327, ..., -0.0103, -0.0133, -0.0744], + ..., + [ 0.0379, -0.0974, 0.1535, ..., 0.0310, -0.0902, 0.0474], + [-0.0408, 0.0585, -0.1356, ..., -0.0536, 0.0119, -0.0845], + [-0.0552, 0.0658, 0.0503, ..., -0.0524, -0.0414, -0.0210]], + device='cuda:0'), grad: tensor([[-1.9217e-03, 1.8346e-04, 3.2163e-04, ..., -2.9411e-03, + 8.3685e-04, 2.3067e-04], + [ 1.0484e-04, 3.7640e-05, 6.8784e-05, ..., -1.0246e-02, + 4.8733e-04, -7.0930e-05], + [-5.3167e-05, -5.9545e-05, -4.2033e-04, ..., 5.0659e-03, + 4.7541e-04, 2.7809e-06], + ..., + [ 1.8013e-04, 3.7622e-04, 7.6771e-04, ..., 2.5482e-03, + 1.0309e-03, 3.8719e-04], + [ 2.5678e-04, 1.2457e-04, 3.8886e-04, ..., 2.6913e-03, + 9.2363e-04, 3.2687e-04], + [ 8.3506e-05, -1.6298e-03, -2.8076e-03, ..., -6.0692e-03, + -4.6463e-03, -2.0027e-03]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0289, 0.0337, 0.0005, -0.0050, 0.0249, -0.0298, 0.0063, -0.0021, + -0.0023, -0.0154], device='cuda:0'), grad: tensor([ 0.0015, -0.0428, 0.0362, 0.0112, 0.0232, 0.0135, -0.0390, 0.0199, + 0.0207, -0.0443], device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 214.67, cls_loss 0.5555 cls_loss_mapping 0.0042 cls_loss_causal 0.4889 re_mapping 0.0105 re_causal 0.0238 /// teacc 98.75 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.0693, -0.1188, -0.0909, ..., -0.0390, 0.0577, -0.1046], + [-0.0710, -0.0877, -0.0716, ..., 0.0982, -0.0417, 0.2143], + [ 0.0002, -0.0167, -0.0324, ..., -0.0107, -0.0139, -0.0748], + ..., + [ 0.0378, -0.0973, 0.1540, ..., 0.0305, -0.0909, 0.0482], + [-0.0407, 0.0600, -0.1361, ..., -0.0535, 0.0123, -0.0842], + [-0.0554, 0.0653, 0.0497, ..., -0.0518, -0.0411, -0.0210]], + device='cuda:0'), grad: tensor([[ 3.0361e-07, 2.1145e-05, 1.7202e-04, ..., -5.5199e-03, + 1.9491e-04, -8.5735e-04], + [ 7.4878e-07, 2.9877e-05, -7.5293e-04, ..., -2.5845e-03, + 1.3754e-05, 3.7074e-04], + [ 3.4962e-06, 7.2289e-04, 3.2978e-03, ..., 3.5915e-03, + 2.2173e-05, 2.9259e-03], + ..., + [ 1.7881e-06, 7.8261e-05, -1.6003e-03, ..., -2.2888e-03, + 1.1243e-05, -2.1172e-03], + [-1.6347e-05, 2.1958e-04, 9.4748e-04, ..., 1.3361e-03, + 3.2377e-04, 6.2370e-04], + [ 6.2399e-07, -9.8705e-04, -4.0894e-03, ..., -5.1880e-04, + 9.7811e-05, -2.4204e-03]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0291, 0.0336, 0.0003, -0.0061, 0.0238, -0.0296, 0.0069, -0.0019, + -0.0022, -0.0139], device='cuda:0'), grad: tensor([-0.0118, -0.0252, 0.0139, 0.0059, -0.0027, 0.0055, 0.0110, 0.0078, + 0.0088, -0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.67, cls_loss 0.5388 cls_loss_mapping 0.0040 cls_loss_causal 0.4759 re_mapping 0.0106 re_causal 0.0233 /// teacc 98.63 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.0688, -0.1205, -0.0901, ..., -0.0397, 0.0553, -0.1044], + [-0.0711, -0.0874, -0.0726, ..., 0.0982, -0.0409, 0.2144], + [-0.0004, -0.0172, -0.0331, ..., -0.0106, -0.0132, -0.0752], + ..., + [ 0.0370, -0.0974, 0.1541, ..., 0.0307, -0.0923, 0.0489], + [-0.0399, 0.0594, -0.1362, ..., -0.0535, 0.0126, -0.0843], + [-0.0558, 0.0655, 0.0500, ..., -0.0515, -0.0405, -0.0223]], + device='cuda:0'), grad: tensor([[ 4.1127e-05, 8.7261e-05, 1.9491e-05, ..., 2.5311e-03, + 1.3232e-04, 3.3379e-04], + [ 4.3011e-04, 6.9328e-06, -4.3225e-04, ..., -6.1798e-03, + 7.2680e-06, -1.9197e-03], + [ 7.9691e-05, 6.6385e-06, 1.9467e-04, ..., -3.3054e-03, + 8.9407e-06, 1.6117e-03], + ..., + [-1.0939e-03, 5.9962e-05, 1.2684e-04, ..., -2.2049e-03, + 9.8199e-06, -3.1013e-03], + [ 1.1164e-04, 6.6042e-05, 2.0176e-05, ..., 4.2343e-03, + 5.2869e-05, 4.9973e-04], + [ 2.9445e-04, 4.2105e-04, -9.1255e-05, ..., 3.7117e-03, + 5.7608e-05, 1.2150e-03]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0289, 0.0340, -0.0006, -0.0062, 0.0232, -0.0290, 0.0071, -0.0012, + -0.0015, -0.0149], device='cuda:0'), grad: tensor([ 0.0165, -0.0225, -0.0386, 0.0152, -0.0762, 0.0141, 0.0067, 0.0080, + 0.0277, 0.0490], device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 214.64, cls_loss 0.5624 cls_loss_mapping 0.0039 cls_loss_causal 0.5019 re_mapping 0.0100 re_causal 0.0238 /// teacc 98.63 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.0683, -0.1209, -0.0899, ..., -0.0396, 0.0551, -0.1022], + [-0.0715, -0.0871, -0.0728, ..., 0.0980, -0.0416, 0.2142], + [-0.0012, -0.0189, -0.0332, ..., -0.0101, -0.0132, -0.0766], + ..., + [ 0.0368, -0.0991, 0.1542, ..., 0.0297, -0.0938, 0.0504], + [-0.0395, 0.0609, -0.1368, ..., -0.0535, 0.0124, -0.0854], + [-0.0561, 0.0667, 0.0502, ..., -0.0512, -0.0407, -0.0225]], + device='cuda:0'), grad: tensor([[ 3.6448e-05, 2.1875e-04, 5.7697e-04, ..., -1.7633e-03, + 7.8857e-05, 3.0780e-04], + [ 5.1081e-05, -5.9891e-04, -2.1000e-03, ..., -1.8982e-02, + 1.5832e-06, 2.0158e-04], + [ 7.1824e-05, 2.2709e-04, 3.0766e-03, ..., 1.1604e-02, + -8.2092e-03, 4.4656e-04], + ..., + [-7.0691e-05, 1.6606e-04, -4.4894e-04, ..., 2.9907e-03, + 3.6478e-04, -1.4734e-03], + [ 2.8342e-05, 2.2650e-04, 4.9019e-04, ..., 4.7913e-03, + 1.0471e-03, 3.3855e-04], + [ 7.0512e-05, 1.8275e-04, 5.0545e-04, ..., 4.6730e-03, + 5.6152e-03, 7.2527e-04]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0291, 0.0342, -0.0002, -0.0066, 0.0238, -0.0290, 0.0060, -0.0023, + -0.0010, -0.0140], device='cuda:0'), grad: tensor([-0.0140, -0.0431, -0.0039, -0.0065, -0.0078, -0.0120, 0.0192, 0.0140, + 0.0234, 0.0307], device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 215.01, cls_loss 0.5374 cls_loss_mapping 0.0044 cls_loss_causal 0.4645 re_mapping 0.0107 re_causal 0.0244 /// teacc 98.79 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.0680, -0.1207, -0.0904, ..., -0.0406, 0.0561, -0.1031], + [-0.0705, -0.0878, -0.0721, ..., 0.0995, -0.0418, 0.2157], + [-0.0017, -0.0198, -0.0326, ..., -0.0106, -0.0134, -0.0771], + ..., + [ 0.0372, -0.0989, 0.1551, ..., 0.0295, -0.0944, 0.0508], + [-0.0402, 0.0607, -0.1369, ..., -0.0536, 0.0120, -0.0863], + [-0.0570, 0.0676, 0.0480, ..., -0.0499, -0.0415, -0.0219]], + device='cuda:0'), grad: tensor([[ 1.4819e-05, 6.4552e-05, 6.1810e-05, ..., 4.0092e-03, + 9.7275e-05, 4.0889e-04], + [ 4.5419e-05, 3.2932e-05, 2.7704e-04, ..., 1.5625e-02, + 8.5890e-05, 1.9217e-03], + [-2.2495e-04, 6.6996e-04, 5.2005e-05, ..., -5.8937e-03, + 3.5763e-05, 4.0960e-04], + ..., + [ 1.2383e-05, 1.3363e-04, 1.0270e-04, ..., 3.1376e-03, + 5.3048e-06, 3.7718e-04], + [ 4.2439e-05, -3.2210e-04, 8.2016e-05, ..., 5.2185e-03, + 5.2422e-05, 4.2391e-04], + [ 3.1233e-05, 3.8445e-05, -6.8545e-05, ..., -1.7441e-02, + 4.9062e-06, -3.7975e-03]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0292, 0.0350, -0.0005, -0.0060, 0.0236, -0.0290, 0.0058, -0.0027, + -0.0009, -0.0144], device='cuda:0'), grad: tensor([ 0.0192, 0.0599, -0.0107, -0.0480, -0.0051, -0.0134, 0.0003, 0.0180, + 0.0238, -0.0438], device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 214.97, cls_loss 0.5086 cls_loss_mapping 0.0032 cls_loss_causal 0.4510 re_mapping 0.0102 re_causal 0.0245 /// teacc 98.69 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.0672, -0.1202, -0.0909, ..., -0.0404, 0.0553, -0.1037], + [-0.0709, -0.0884, -0.0725, ..., 0.0993, -0.0427, 0.2159], + [-0.0020, -0.0185, -0.0327, ..., -0.0099, -0.0137, -0.0768], + ..., + [ 0.0373, -0.0996, 0.1550, ..., 0.0296, -0.0950, 0.0516], + [-0.0397, 0.0611, -0.1367, ..., -0.0550, 0.0120, -0.0872], + [-0.0570, 0.0680, 0.0472, ..., -0.0494, -0.0412, -0.0222]], + device='cuda:0'), grad: tensor([[ 5.2340e-07, 1.9217e-04, 2.9492e-04, ..., -1.4984e-02, + -9.2936e-04, 1.2577e-04], + [ 2.9616e-07, 4.1842e-04, 1.2112e-03, ..., 5.3749e-03, + 1.7732e-05, 6.1989e-04], + [ 7.4506e-08, 2.8920e-04, 3.8123e-04, ..., 1.1530e-03, + -2.9373e-04, 1.8764e-04], + ..., + [ 2.7940e-08, 1.6153e-04, 1.4372e-03, ..., 5.0402e-04, + 9.9540e-05, 1.0176e-03], + [-1.6540e-06, -1.8063e-03, 4.8971e-04, ..., -7.2002e-04, + 2.1482e-04, 2.0766e-04], + [ 2.3283e-07, -3.7201e-02, 9.2387e-05, ..., 2.4738e-03, + 1.1021e-04, 4.3201e-04]], device='cuda:0') +Epoch 266, bias, value: tensor([-2.9052e-02, 3.4495e-02, -3.9107e-05, -6.0491e-03, 2.3473e-02, + -2.8890e-02, 5.6849e-03, -1.7799e-03, -2.2408e-03, -1.3759e-02], + device='cuda:0'), grad: tensor([-0.0217, 0.0183, 0.0073, 0.0166, 0.0100, 0.0025, 0.0147, -0.0467, + 0.0188, -0.0199], device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 214.43, cls_loss 0.5622 cls_loss_mapping 0.0047 cls_loss_causal 0.4946 re_mapping 0.0094 re_causal 0.0230 /// teacc 98.61 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.0667, -0.1208, -0.0914, ..., -0.0397, 0.0566, -0.1041], + [-0.0711, -0.0889, -0.0743, ..., 0.0984, -0.0426, 0.2160], + [-0.0031, -0.0196, -0.0327, ..., -0.0107, -0.0124, -0.0778], + ..., + [ 0.0371, -0.0995, 0.1560, ..., 0.0301, -0.0955, 0.0529], + [-0.0397, 0.0614, -0.1376, ..., -0.0543, 0.0122, -0.0866], + [-0.0568, 0.0692, 0.0477, ..., -0.0502, -0.0417, -0.0231]], + device='cuda:0'), grad: tensor([[ 2.2724e-07, -6.1073e-03, -5.1403e-04, ..., -4.2200e-04, + 8.8811e-05, 1.0147e-03], + [ 1.4156e-07, 3.7408e-04, 2.2757e-04, ..., 3.6449e-03, + 7.4148e-04, 8.8835e-04], + [-6.6459e-06, 8.2779e-04, -1.2436e-03, ..., -1.2367e-02, + -2.5826e-03, -2.5711e-03], + ..., + [ 2.1413e-05, 1.4734e-04, 8.0538e-04, ..., 2.8706e-03, + 1.7834e-04, 3.0565e-04], + [ 9.0674e-06, 2.7580e-03, 4.2462e-04, ..., -3.5973e-03, + 2.0218e-04, -2.5654e-03], + [-4.4852e-05, 1.1387e-03, 6.7949e-04, ..., 1.2560e-03, + 2.9135e-04, 1.6050e-03]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0285, 0.0347, -0.0003, -0.0064, 0.0244, -0.0287, 0.0055, -0.0022, + -0.0029, -0.0135], device='cuda:0'), grad: tensor([-0.0375, 0.0223, -0.0444, 0.0095, 0.0216, 0.0031, 0.0225, 0.0169, + -0.0068, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 214.43, cls_loss 0.5450 cls_loss_mapping 0.0048 cls_loss_causal 0.4722 re_mapping 0.0098 re_causal 0.0230 /// teacc 98.60 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.0656, -0.1195, -0.0912, ..., -0.0390, 0.0562, -0.1056], + [-0.0707, -0.0890, -0.0734, ..., 0.0983, -0.0436, 0.2167], + [-0.0023, -0.0182, -0.0319, ..., -0.0101, -0.0122, -0.0781], + ..., + [ 0.0366, -0.0996, 0.1549, ..., 0.0300, -0.0921, 0.0522], + [-0.0402, 0.0603, -0.1377, ..., -0.0552, 0.0109, -0.0872], + [-0.0573, 0.0691, 0.0473, ..., -0.0521, -0.0430, -0.0223]], + device='cuda:0'), grad: tensor([[ 6.5789e-06, 1.1995e-05, 9.2089e-05, ..., 2.8763e-03, + 2.8324e-04, 2.9731e-04], + [ 2.4624e-06, 2.0921e-05, 5.4628e-05, ..., -2.4567e-03, + 1.3137e-04, 1.4648e-03], + [ 9.4891e-05, 1.5640e-04, 2.4772e-04, ..., -3.8967e-03, + 1.5759e-04, 8.6606e-05], + ..., + [ 1.0721e-05, 7.9274e-05, 7.2718e-05, ..., 9.4986e-04, + -2.1267e-03, -3.2024e-03], + [-2.9469e-04, -1.4315e-03, -4.6015e-04, ..., 4.8218e-03, + 9.4473e-05, 5.1346e-03], + [ 9.1568e-06, 1.8997e-03, 2.4319e-03, ..., 1.3494e-03, + 5.6934e-04, 3.4351e-03]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0270, 0.0349, -0.0005, -0.0063, 0.0240, -0.0292, 0.0059, -0.0021, + -0.0031, -0.0144], device='cuda:0'), grad: tensor([ 0.0249, -0.0052, -0.0374, -0.0087, -0.0021, -0.0244, 0.0248, -0.0316, + 0.0141, 0.0454], device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.48, cls_loss 0.5431 cls_loss_mapping 0.0033 cls_loss_causal 0.4823 re_mapping 0.0095 re_causal 0.0233 /// teacc 98.63 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.0656, -0.1197, -0.0918, ..., -0.0408, 0.0563, -0.1060], + [-0.0708, -0.0885, -0.0730, ..., 0.0993, -0.0428, 0.2169], + [-0.0027, -0.0188, -0.0324, ..., -0.0103, -0.0131, -0.0779], + ..., + [ 0.0368, -0.1011, 0.1540, ..., 0.0299, -0.0924, 0.0504], + [-0.0399, 0.0610, -0.1383, ..., -0.0551, 0.0110, -0.0890], + [-0.0572, 0.0693, 0.0486, ..., -0.0506, -0.0439, -0.0197]], + device='cuda:0'), grad: tensor([[-1.8196e-03, 1.4725e-03, 5.4073e-04, ..., -4.8294e-03, + 4.2868e-04, 7.5722e-04], + [ 3.2949e-04, 6.1214e-05, -4.2229e-03, ..., -5.4016e-03, + 1.1045e-06, -5.4932e-03], + [ 1.3638e-03, 3.5629e-03, 1.0815e-03, ..., 6.1150e-03, + 2.8774e-05, 1.2712e-03], + ..., + [ 9.6607e-04, 2.4796e-04, 3.8033e-03, ..., 3.3302e-03, + 1.2234e-05, 2.7370e-03], + [-1.4973e-03, -4.5738e-03, 6.8378e-04, ..., -7.9956e-03, + 9.4697e-06, 8.0824e-04], + [-5.2786e-04, -1.8826e-03, -4.4441e-03, ..., -3.9291e-04, + 2.2203e-05, -3.3627e-03]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0284, 0.0348, -0.0004, -0.0076, 0.0252, -0.0296, 0.0063, -0.0024, + -0.0022, -0.0139], device='cuda:0'), grad: tensor([ 0.0112, -0.0366, 0.0327, -0.0091, 0.0284, 0.0188, 0.0182, 0.0393, + -0.0304, -0.0725], device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 214.52, cls_loss 0.5500 cls_loss_mapping 0.0046 cls_loss_causal 0.4756 re_mapping 0.0098 re_causal 0.0233 /// teacc 98.79 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.0667, -0.1204, -0.0931, ..., -0.0411, 0.0564, -0.1066], + [-0.0721, -0.0892, -0.0727, ..., 0.0992, -0.0412, 0.2166], + [-0.0021, -0.0202, -0.0318, ..., -0.0113, -0.0132, -0.0766], + ..., + [ 0.0362, -0.1010, 0.1545, ..., 0.0290, -0.0946, 0.0501], + [-0.0386, 0.0610, -0.1392, ..., -0.0547, 0.0105, -0.0869], + [-0.0585, 0.0696, 0.0483, ..., -0.0510, -0.0440, -0.0197]], + device='cuda:0'), grad: tensor([[-2.6189e-06, -4.2820e-04, -3.4034e-05, ..., -3.2330e-03, + -3.4906e-06, -1.5938e-04], + [ 7.0408e-07, 1.5914e-05, 7.1764e-05, ..., 5.3453e-04, + 4.5262e-07, 2.9206e-05], + [ 8.4937e-07, 8.7380e-05, 2.2471e-04, ..., 4.6158e-03, + 2.4792e-06, 9.6440e-05], + ..., + [-1.8962e-06, -1.3161e-04, -9.2554e-04, ..., -1.2741e-02, + 3.3900e-07, -3.6240e-04], + [ 3.0734e-07, 4.5538e-04, 2.7752e-04, ..., 4.8518e-04, + 3.1710e-05, 9.8467e-05], + [ 1.9036e-06, 2.8086e-04, 1.6677e-04, ..., 3.1414e-03, + 1.2871e-06, 2.0301e-04]], device='cuda:0') +Epoch 270, bias, value: tensor([-2.8578e-02, 3.5703e-02, 2.1092e-05, -7.2605e-03, 2.4996e-02, + -2.9846e-02, 6.7894e-03, -3.1578e-03, -1.9086e-03, -1.4832e-02], + device='cuda:0'), grad: tensor([-1.3535e-02, -7.1106e-03, 2.0279e-02, -2.9697e-03, 2.2430e-02, + -4.9829e-05, -8.8043e-03, -1.5488e-02, -6.6757e-03, 1.1925e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 214.95, cls_loss 0.5654 cls_loss_mapping 0.0044 cls_loss_causal 0.4948 re_mapping 0.0102 re_causal 0.0239 /// teacc 98.68 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.0675, -0.1231, -0.0932, ..., -0.0416, 0.0559, -0.1059], + [-0.0717, -0.0899, -0.0732, ..., 0.0995, -0.0407, 0.2166], + [-0.0026, -0.0194, -0.0325, ..., -0.0119, -0.0138, -0.0770], + ..., + [ 0.0358, -0.0977, 0.1556, ..., 0.0303, -0.0945, 0.0506], + [-0.0372, 0.0594, -0.1385, ..., -0.0541, 0.0103, -0.0864], + [-0.0576, 0.0699, 0.0475, ..., -0.0508, -0.0441, -0.0203]], + device='cuda:0'), grad: tensor([[ 2.6356e-06, 1.7047e-04, 7.6592e-05, ..., -2.2335e-03, + 4.1202e-06, 4.8757e-04], + [ 2.7008e-07, 5.3316e-05, 2.1088e-04, ..., 3.7212e-03, + 1.7229e-06, 1.0529e-03], + [ 2.1700e-06, 2.2197e-04, 1.6809e-04, ..., -7.6942e-03, + -1.2472e-05, -6.1691e-05], + ..., + [ 1.3504e-06, 1.2767e-04, 8.5831e-05, ..., 3.4409e-03, + 2.6021e-06, 1.1692e-03], + [-1.7476e-04, 6.1178e-04, 5.5647e-04, ..., 5.2261e-03, + 1.3202e-05, 2.7695e-03], + [ 1.4659e-06, 1.0643e-03, 7.9441e-04, ..., 5.5199e-03, + 1.7285e-05, 7.7286e-03]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0289, 0.0357, -0.0005, -0.0076, 0.0235, -0.0292, 0.0074, -0.0018, + -0.0021, -0.0145], device='cuda:0'), grad: tensor([-0.0187, 0.0187, -0.0149, -0.0168, -0.0585, 0.0206, -0.0162, 0.0183, + 0.0269, 0.0405], device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 215.08, cls_loss 0.5718 cls_loss_mapping 0.0037 cls_loss_causal 0.4850 re_mapping 0.0106 re_causal 0.0238 /// teacc 98.75 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.0674, -0.1235, -0.0923, ..., -0.0423, 0.0554, -0.1062], + [-0.0699, -0.0904, -0.0727, ..., 0.0997, -0.0420, 0.2184], + [-0.0035, -0.0193, -0.0335, ..., -0.0123, -0.0135, -0.0777], + ..., + [ 0.0349, -0.0966, 0.1564, ..., 0.0297, -0.0943, 0.0510], + [-0.0384, 0.0601, -0.1396, ..., -0.0530, 0.0096, -0.0876], + [-0.0573, 0.0697, 0.0467, ..., -0.0510, -0.0424, -0.0206]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0005, 0.0002, ..., -0.0036, 0.0009, -0.0038], + [ 0.0006, 0.0002, 0.0007, ..., 0.0033, 0.0003, 0.0013], + [ 0.0008, 0.0010, 0.0006, ..., -0.0014, 0.0021, 0.0006], + ..., + [ 0.0004, 0.0006, 0.0007, ..., 0.0086, 0.0011, 0.0021], + [ 0.0008, -0.0122, 0.0004, ..., 0.0040, -0.0172, 0.0025], + [ 0.0005, 0.0064, -0.0020, ..., -0.0142, 0.0098, -0.0125]], + device='cuda:0') +Epoch 272, bias, value: tensor([-0.0296, 0.0363, -0.0006, -0.0071, 0.0231, -0.0293, 0.0077, -0.0020, + -0.0017, -0.0150], device='cuda:0'), grad: tensor([-0.0038, -0.0035, 0.0215, 0.0087, -0.0066, 0.0134, 0.0084, 0.0358, + -0.0078, -0.0661], device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 214.69, cls_loss 0.5244 cls_loss_mapping 0.0048 cls_loss_causal 0.4580 re_mapping 0.0102 re_causal 0.0232 /// teacc 98.69 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.0683, -0.1246, -0.0918, ..., -0.0424, 0.0558, -0.1062], + [-0.0693, -0.0914, -0.0731, ..., 0.0992, -0.0417, 0.2188], + [-0.0045, -0.0190, -0.0340, ..., -0.0122, -0.0135, -0.0771], + ..., + [ 0.0359, -0.0971, 0.1563, ..., 0.0310, -0.0952, 0.0513], + [-0.0380, 0.0599, -0.1394, ..., -0.0530, 0.0104, -0.0875], + [-0.0589, 0.0709, 0.0480, ..., -0.0510, -0.0428, -0.0207]], + device='cuda:0'), grad: tensor([[ 3.0589e-04, 5.8126e-04, 2.0301e-04, ..., 2.2202e-03, + 4.7398e-04, 3.4714e-04], + [ 5.0497e-04, -1.3496e-02, 3.9965e-05, ..., 4.1428e-03, + 1.0138e-03, -3.2444e-03], + [-3.0088e-04, -1.9665e-03, 2.1400e-03, ..., 4.7112e-03, + 3.3832e-04, -1.2541e-03], + ..., + [ 3.5906e-04, 4.5633e-04, -1.4496e-04, ..., 2.8458e-03, + 8.1873e-04, -1.8978e-03], + [ 3.5501e-04, 9.1171e-03, 2.7394e-04, ..., 3.8128e-03, + 8.3780e-04, 2.4834e-03], + [-2.6073e-03, 1.2684e-03, 1.8883e-04, ..., -5.7335e-03, + -3.5496e-03, 2.1305e-03]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0300, 0.0360, -0.0001, -0.0074, 0.0241, -0.0295, 0.0076, -0.0016, + -0.0019, -0.0153], device='cuda:0'), grad: tensor([ 0.0085, -0.0104, 0.0265, -0.0464, 0.0096, -0.0222, 0.0098, 0.0070, + 0.0327, -0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 214.91, cls_loss 0.5693 cls_loss_mapping 0.0053 cls_loss_causal 0.4939 re_mapping 0.0108 re_causal 0.0237 /// teacc 98.72 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.0694, -0.1247, -0.0928, ..., -0.0436, 0.0558, -0.1056], + [-0.0699, -0.0913, -0.0743, ..., 0.0999, -0.0417, 0.2187], + [-0.0043, -0.0205, -0.0342, ..., -0.0127, -0.0127, -0.0754], + ..., + [ 0.0371, -0.0962, 0.1580, ..., 0.0314, -0.0957, 0.0513], + [-0.0379, 0.0600, -0.1398, ..., -0.0529, 0.0109, -0.0873], + [-0.0588, 0.0731, 0.0470, ..., -0.0503, -0.0417, -0.0213]], + device='cuda:0'), grad: tensor([[ 1.0915e-06, 1.8179e-05, 2.9850e-04, ..., 1.0557e-03, + 6.2990e-04, 5.2404e-04], + [ 2.6170e-06, 2.2066e-04, 4.4763e-05, ..., 2.0046e-03, + 1.2417e-03, -4.2140e-05], + [ 7.7546e-05, 1.0281e-03, 1.7080e-03, ..., 2.9984e-03, + 1.1234e-03, 1.1139e-03], + ..., + [-3.2276e-05, 1.4174e-04, -3.0556e-03, ..., -6.1264e-03, + -5.1155e-03, -2.9240e-03], + [-1.5602e-03, -4.8447e-04, -1.5915e-02, ..., 1.2207e-03, + 5.0735e-04, 1.3304e-04], + [ 1.3599e-03, 8.6129e-05, 1.4664e-02, ..., 1.1053e-03, + 6.3133e-04, 9.3174e-04]], device='cuda:0') +Epoch 274, bias, value: tensor([-3.0559e-02, 3.6354e-02, -6.7239e-05, -7.7747e-03, 2.3536e-02, + -2.9986e-02, 6.9054e-03, -4.4771e-04, -1.0688e-03, -1.5261e-02], + device='cuda:0'), grad: tensor([ 0.0159, 0.0217, 0.0299, 0.0120, -0.0089, -0.0410, 0.0072, -0.0390, + -0.0393, 0.0415], device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 214.62, cls_loss 0.5341 cls_loss_mapping 0.0038 cls_loss_causal 0.4657 re_mapping 0.0104 re_causal 0.0237 /// teacc 98.66 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.0683, -0.1266, -0.0925, ..., -0.0437, 0.0557, -0.1059], + [-0.0695, -0.0934, -0.0739, ..., 0.1003, -0.0425, 0.2206], + [-0.0046, -0.0198, -0.0353, ..., -0.0125, -0.0125, -0.0766], + ..., + [ 0.0372, -0.0970, 0.1571, ..., 0.0307, -0.0923, 0.0515], + [-0.0382, 0.0600, -0.1401, ..., -0.0538, 0.0112, -0.0883], + [-0.0596, 0.0731, 0.0476, ..., -0.0492, -0.0434, -0.0216]], + device='cuda:0'), grad: tensor([[ 6.3330e-07, -1.0741e-04, 3.8832e-05, ..., 2.3699e-04, + -1.0147e-03, 2.1148e-04], + [ 1.4305e-06, 9.3412e-04, -4.1175e-04, ..., -1.3676e-03, + -1.5569e-04, -1.1139e-03], + [ 3.1620e-05, 2.1696e-04, 1.2434e-04, ..., -3.2463e-03, + -4.6959e-03, 3.3903e-04], + ..., + [ 4.1611e-06, -1.2207e-02, -1.2112e-03, ..., -6.7024e-03, + 1.4234e-04, -3.9368e-03], + [-1.1814e-04, -2.3899e-03, -4.9019e-04, ..., 8.0299e-04, + 4.9162e-04, 1.2255e-03], + [ 2.9150e-06, 4.3221e-03, 3.0637e-04, ..., 2.8839e-03, + 5.4693e-04, 1.3657e-03]], device='cuda:0') +Epoch 275, bias, value: tensor([-3.0433e-02, 3.6129e-02, -7.8578e-04, -8.1034e-03, 2.3436e-02, + -2.8976e-02, 7.1996e-03, 9.9257e-05, -1.6243e-03, -1.5196e-02], + device='cuda:0'), grad: tensor([ 0.0020, 0.0004, -0.0132, 0.0117, -0.0052, 0.0036, 0.0131, -0.0249, + 0.0114, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.80, cls_loss 0.5424 cls_loss_mapping 0.0032 cls_loss_causal 0.4754 re_mapping 0.0098 re_causal 0.0231 /// teacc 98.52 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.0684, -0.1262, -0.0933, ..., -0.0432, 0.0574, -0.1052], + [-0.0695, -0.0921, -0.0728, ..., 0.0996, -0.0424, 0.2216], + [-0.0046, -0.0194, -0.0364, ..., -0.0121, -0.0139, -0.0773], + ..., + [ 0.0374, -0.0976, 0.1571, ..., 0.0315, -0.0927, 0.0521], + [-0.0384, 0.0610, -0.1404, ..., -0.0543, 0.0116, -0.0887], + [-0.0600, 0.0735, 0.0477, ..., -0.0504, -0.0428, -0.0220]], + device='cuda:0'), grad: tensor([[ 2.0742e-04, -2.1458e-03, -1.8320e-03, ..., -2.8248e-03, + 9.4986e-04, -2.5673e-03], + [ 2.0015e-04, 7.0477e-04, 3.7885e-04, ..., 1.0357e-03, + 8.9502e-04, 6.9141e-04], + [ 6.2823e-05, 8.1015e-04, 4.5443e-04, ..., 2.1515e-03, + 2.2888e-03, 9.7942e-04], + ..., + [ 1.4009e-03, -6.5899e-04, -3.6335e-04, ..., 9.7609e-04, + -9.5291e-03, -1.7965e-04], + [-1.8177e-03, 3.2902e-04, 6.0272e-04, ..., 3.9520e-03, + 1.6308e-03, -1.2808e-03], + [-9.1028e-04, 9.1648e-04, 2.7800e-04, ..., 1.2875e-03, + 9.1076e-04, 1.2112e-03]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0296, 0.0371, -0.0006, -0.0081, 0.0229, -0.0294, 0.0074, 0.0002, + -0.0019, -0.0162], device='cuda:0'), grad: tensor([-0.0221, 0.0129, 0.0185, 0.0323, -0.0182, 0.0176, -0.0135, -0.0391, + 0.0218, -0.0101], device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 214.36, cls_loss 0.5225 cls_loss_mapping 0.0046 cls_loss_causal 0.4499 re_mapping 0.0102 re_causal 0.0225 /// teacc 98.66 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.0694, -0.1260, -0.0930, ..., -0.0432, 0.0579, -0.1072], + [-0.0720, -0.0916, -0.0735, ..., 0.0987, -0.0424, 0.2201], + [-0.0039, -0.0195, -0.0358, ..., -0.0114, -0.0143, -0.0772], + ..., + [ 0.0388, -0.0995, 0.1571, ..., 0.0318, -0.0936, 0.0530], + [-0.0392, 0.0608, -0.1420, ..., -0.0537, 0.0114, -0.0883], + [-0.0591, 0.0759, 0.0486, ..., -0.0508, -0.0419, -0.0223]], + device='cuda:0'), grad: tensor([[ 4.3452e-05, -2.3878e-04, 1.8084e-04, ..., 3.1948e-04, + -2.4319e-03, 1.2565e-04], + [-1.7424e-03, -2.5978e-03, -3.3417e-03, ..., -6.1951e-03, + -1.5736e-03, -9.3174e-04], + [ 3.8815e-04, 2.1553e-03, 1.0195e-03, ..., 2.7390e-03, + 1.9836e-03, 2.4188e-04], + ..., + [ 1.5497e-04, -6.8283e-04, -9.7847e-04, ..., 3.8910e-04, + -9.6703e-04, -7.7105e-04], + [ 1.3399e-04, 1.2541e-03, 3.7479e-04, ..., 1.0014e-03, + 1.0242e-03, 3.4976e-04], + [ 1.1760e-04, 1.4172e-03, 4.2105e-04, ..., 1.2197e-03, + 1.2007e-03, 1.8144e-04]], device='cuda:0') +Epoch 277, bias, value: tensor([-2.9843e-02, 3.6935e-02, -2.1415e-06, -7.3384e-03, 2.2992e-02, + -2.9982e-02, 8.3405e-03, -3.5220e-04, -3.0945e-03, -1.5737e-02], + device='cuda:0'), grad: tensor([-0.0204, -0.0323, 0.0253, 0.0271, -0.0326, -0.0151, 0.0292, -0.0109, + 0.0136, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 214.78, cls_loss 0.5474 cls_loss_mapping 0.0047 cls_loss_causal 0.4718 re_mapping 0.0104 re_causal 0.0235 /// teacc 98.65 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.0706, -0.1271, -0.0930, ..., -0.0433, 0.0585, -0.1071], + [-0.0725, -0.0906, -0.0710, ..., 0.0991, -0.0430, 0.2202], + [-0.0029, -0.0202, -0.0361, ..., -0.0120, -0.0138, -0.0761], + ..., + [ 0.0402, -0.0977, 0.1563, ..., 0.0326, -0.0935, 0.0537], + [-0.0400, 0.0621, -0.1426, ..., -0.0539, 0.0107, -0.0891], + [-0.0594, 0.0740, 0.0482, ..., -0.0504, -0.0415, -0.0231]], + device='cuda:0'), grad: tensor([[ 9.9599e-05, 5.5885e-04, 3.0732e-04, ..., 2.0146e-04, + -2.2352e-08, 1.4651e-04], + [-2.1152e-03, -2.7924e-03, -1.7500e-03, ..., -4.2648e-03, + -4.1294e-04, -7.2975e-03], + [ 4.1389e-04, -2.2268e-04, -1.1711e-03, ..., -8.6355e-04, + 7.4171e-06, 4.1318e-04], + ..., + [ 9.1791e-04, 3.1414e-03, 1.7042e-03, ..., 7.0333e-04, + 1.4417e-06, 5.1308e-04], + [ 5.0259e-04, -1.8105e-05, 5.1212e-04, ..., 6.9237e-04, + 2.4214e-05, 4.0078e-04], + [-3.3283e-03, 5.6496e-03, 3.3741e-03, ..., 6.1512e-04, + 4.6156e-06, 1.5173e-03]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0300, 0.0371, -0.0007, -0.0079, 0.0239, -0.0290, 0.0068, 0.0003, + -0.0027, -0.0159], device='cuda:0'), grad: tensor([ 0.0184, -0.0141, -0.0188, 0.0061, 0.0282, -0.0235, 0.0239, 0.0033, + -0.0077, -0.0158], device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 214.76, cls_loss 0.5395 cls_loss_mapping 0.0058 cls_loss_causal 0.4751 re_mapping 0.0101 re_causal 0.0227 /// teacc 98.56 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.0706, -0.1265, -0.0926, ..., -0.0425, 0.0580, -0.1087], + [-0.0725, -0.0908, -0.0721, ..., 0.0980, -0.0443, 0.2201], + [-0.0041, -0.0192, -0.0354, ..., -0.0112, -0.0129, -0.0769], + ..., + [ 0.0415, -0.0980, 0.1562, ..., 0.0331, -0.0938, 0.0550], + [-0.0385, 0.0612, -0.1433, ..., -0.0539, 0.0115, -0.0907], + [-0.0603, 0.0739, 0.0476, ..., -0.0509, -0.0418, -0.0240]], + device='cuda:0'), grad: tensor([[ 1.4186e-04, 3.1352e-04, 1.6141e-04, ..., -6.9275e-03, + -1.5383e-03, 4.6402e-05], + [ 2.7910e-05, 2.2972e-04, 1.8567e-05, ..., 1.4095e-03, + 2.4939e-04, -8.3923e-05], + [-1.8568e-03, -4.6883e-03, -3.8185e-03, ..., -4.4975e-03, + 5.0426e-05, -1.6356e-04], + ..., + [ 3.6865e-05, 1.6761e-04, -2.8920e-04, ..., 7.6065e-03, + 2.3782e-04, 1.1959e-03], + [ 1.9336e-04, 4.2419e-03, 7.1955e-04, ..., 1.7185e-03, + 2.3460e-04, 4.6670e-05], + [ 1.1086e-04, -5.7983e-03, -2.6798e-04, ..., 2.1076e-03, + 2.3508e-04, 5.8699e-04]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0287, 0.0357, 0.0004, -0.0082, 0.0238, -0.0285, 0.0075, 0.0006, + -0.0029, -0.0175], device='cuda:0'), grad: tensor([-0.0163, 0.0122, -0.0068, -0.0975, 0.0155, 0.0133, 0.0206, 0.0349, + 0.0165, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 214.24, cls_loss 0.5724 cls_loss_mapping 0.0039 cls_loss_causal 0.5079 re_mapping 0.0100 re_causal 0.0239 /// teacc 98.78 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.0712, -0.1257, -0.0933, ..., -0.0432, 0.0595, -0.1093], + [-0.0732, -0.0903, -0.0724, ..., 0.0984, -0.0433, 0.2188], + [-0.0025, -0.0178, -0.0349, ..., -0.0106, -0.0112, -0.0767], + ..., + [ 0.0407, -0.0987, 0.1566, ..., 0.0324, -0.0939, 0.0552], + [-0.0392, 0.0629, -0.1431, ..., -0.0525, 0.0108, -0.0898], + [-0.0591, 0.0743, 0.0479, ..., -0.0510, -0.0407, -0.0241]], + device='cuda:0'), grad: tensor([[ 3.6645e-04, 6.3419e-04, 2.3880e-03, ..., 4.0741e-03, + 7.3195e-04, 6.2799e-04], + [ 7.2002e-05, 3.2759e-04, 5.2643e-04, ..., 8.5640e-04, + 1.4198e-04, 2.3317e-04], + [-8.9264e-04, 3.7689e-03, -2.0561e-03, ..., -8.1329e-03, + -4.9973e-04, -1.1482e-03], + ..., + [ 4.0710e-05, 9.6512e-04, 5.2691e-04, ..., 7.6008e-04, + 1.6379e-04, 7.0572e-04], + [ 4.3005e-05, 2.4109e-03, 4.8599e-03, ..., -4.4656e-04, + 1.1902e-03, -2.9621e-03], + [ 7.4148e-05, 7.7820e-03, 7.6065e-03, ..., 1.1797e-03, + 1.4722e-04, 9.1095e-03]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0292, 0.0359, 0.0002, -0.0076, 0.0232, -0.0294, 0.0071, 0.0014, + -0.0023, -0.0173], device='cuda:0'), grad: tensor([ 0.0211, -0.0198, -0.0155, -0.0236, -0.0075, 0.0321, 0.0089, 0.0134, + -0.0471, 0.0381], device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.35, cls_loss 0.5252 cls_loss_mapping 0.0034 cls_loss_causal 0.4514 re_mapping 0.0101 re_causal 0.0230 /// teacc 98.77 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.0689, -0.1274, -0.0933, ..., -0.0424, 0.0593, -0.1100], + [-0.0736, -0.0898, -0.0726, ..., 0.0980, -0.0440, 0.2193], + [-0.0021, -0.0194, -0.0344, ..., -0.0104, -0.0115, -0.0772], + ..., + [ 0.0396, -0.0979, 0.1576, ..., 0.0325, -0.0928, 0.0561], + [-0.0400, 0.0618, -0.1439, ..., -0.0529, 0.0111, -0.0906], + [-0.0584, 0.0745, 0.0471, ..., -0.0508, -0.0414, -0.0251]], + device='cuda:0'), grad: tensor([[-3.9911e-04, -4.4614e-05, 9.9242e-06, ..., 1.6079e-03, + -1.1474e-04, 1.5271e-04], + [ 1.1466e-05, -1.8373e-05, -9.6917e-05, ..., 5.0926e-03, + 1.4612e-06, 7.1831e-03], + [ 1.1510e-04, 3.3975e-04, 2.7433e-05, ..., -5.8899e-03, + 3.5971e-05, 3.2377e-04], + ..., + [ 7.2084e-06, 2.8458e-03, 1.3466e-03, ..., 1.0109e-03, + 6.9261e-05, -2.3785e-03], + [ 4.0054e-05, -1.6146e-03, 1.9357e-05, ..., -6.4240e-03, + 5.5432e-06, -7.1869e-03], + [ 1.7360e-05, -2.5311e-03, -1.4086e-03, ..., 5.8985e-04, + -8.6892e-07, 1.4820e-03]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0292, 0.0360, 0.0004, -0.0078, 0.0233, -0.0290, 0.0072, 0.0011, + -0.0032, -0.0168], device='cuda:0'), grad: tensor([ 0.0131, 0.0080, -0.0245, 0.0159, -0.0213, 0.0100, -0.0155, 0.0119, + -0.0136, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 214.62, cls_loss 0.5371 cls_loss_mapping 0.0053 cls_loss_causal 0.4812 re_mapping 0.0102 re_causal 0.0223 /// teacc 98.68 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.0662, -0.1301, -0.0948, ..., -0.0426, 0.0584, -0.1104], + [-0.0741, -0.0895, -0.0734, ..., 0.0969, -0.0445, 0.2194], + [-0.0038, -0.0196, -0.0344, ..., -0.0100, -0.0104, -0.0757], + ..., + [ 0.0381, -0.0982, 0.1573, ..., 0.0328, -0.0920, 0.0559], + [-0.0398, 0.0624, -0.1430, ..., -0.0521, 0.0118, -0.0913], + [-0.0560, 0.0739, 0.0473, ..., -0.0508, -0.0423, -0.0245]], + device='cuda:0'), grad: tensor([[ 3.9339e-04, 2.0778e-04, 2.8294e-06, ..., 1.0052e-03, + 8.0280e-07, -4.3106e-03], + [ 4.8876e-04, 4.2725e-04, 8.0280e-07, ..., 5.1022e-04, + 6.6124e-08, 4.4107e-04], + [ 3.4165e-04, 8.1635e-04, 2.0280e-05, ..., 1.0862e-03, + 4.0345e-06, -2.1229e-03], + ..., + [ 1.5807e-04, 4.5037e-04, 1.1452e-05, ..., 6.1274e-04, + 1.4538e-06, 9.6178e-04], + [ 2.6703e-04, -9.6273e-04, 1.2435e-05, ..., 1.0042e-03, + 1.8375e-06, 1.1597e-03], + [ 2.4772e-04, -2.1420e-03, 1.9455e-04, ..., 9.5463e-04, + 1.0030e-06, 9.5177e-04]], device='cuda:0') +Epoch 282, bias, value: tensor([-3.0393e-02, 3.7481e-02, 5.0758e-05, -7.5064e-03, 2.4061e-02, + -2.9940e-02, 6.4498e-03, 1.5631e-03, -3.0915e-03, -1.6913e-02], + device='cuda:0'), grad: tensor([-0.0341, 0.0500, -0.0676, 0.0072, 0.0288, -0.0330, 0.0287, -0.0032, + 0.0286, -0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 215.06, cls_loss 0.5469 cls_loss_mapping 0.0053 cls_loss_causal 0.4715 re_mapping 0.0101 re_causal 0.0231 /// teacc 98.69 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.0665, -0.1286, -0.0940, ..., -0.0422, 0.0587, -0.1096], + [-0.0727, -0.0900, -0.0754, ..., 0.0969, -0.0445, 0.2193], + [-0.0039, -0.0200, -0.0339, ..., -0.0101, -0.0103, -0.0747], + ..., + [ 0.0383, -0.0984, 0.1570, ..., 0.0332, -0.0917, 0.0538], + [-0.0394, 0.0622, -0.1432, ..., -0.0526, 0.0116, -0.0926], + [-0.0576, 0.0748, 0.0479, ..., -0.0507, -0.0429, -0.0232]], + device='cuda:0'), grad: tensor([[ 4.1509e-04, -1.9316e-06, 4.5031e-05, ..., -5.7936e-04, + 2.7618e-03, 5.3596e-04], + [ 8.7619e-05, 1.2092e-05, 8.2612e-05, ..., 5.7173e-04, + 2.1279e-05, 8.9765e-05], + [-1.3614e-04, -2.2233e-05, -6.3133e-04, ..., -1.1003e-04, + 2.0489e-05, -2.5725e-04], + ..., + [ 7.3791e-05, 2.7552e-05, 9.5308e-05, ..., 6.4516e-04, + 5.1379e-05, 1.1456e-04], + [-9.6512e-04, 2.7275e-03, 1.4982e-03, ..., -2.4071e-03, + 4.1574e-05, -1.2197e-03], + [ 1.1313e-04, 4.4131e-04, 9.2745e-05, ..., 6.9952e-04, + 3.4124e-05, 1.4818e-04]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0294, 0.0364, 0.0007, -0.0090, 0.0241, -0.0285, 0.0061, 0.0016, + -0.0034, -0.0166], device='cuda:0'), grad: tensor([-0.0041, 0.0166, 0.0108, -0.0162, -0.0018, 0.0094, -0.0135, -0.0146, + -0.0049, 0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.44, cls_loss 0.5402 cls_loss_mapping 0.0043 cls_loss_causal 0.4731 re_mapping 0.0099 re_causal 0.0224 /// teacc 98.74 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.0661, -0.1295, -0.0935, ..., -0.0418, 0.0587, -0.1098], + [-0.0725, -0.0899, -0.0755, ..., 0.0968, -0.0440, 0.2190], + [-0.0028, -0.0206, -0.0342, ..., -0.0104, -0.0106, -0.0746], + ..., + [ 0.0375, -0.0981, 0.1572, ..., 0.0323, -0.0919, 0.0539], + [-0.0406, 0.0624, -0.1441, ..., -0.0533, 0.0119, -0.0940], + [-0.0574, 0.0753, 0.0470, ..., -0.0513, -0.0428, -0.0231]], + device='cuda:0'), grad: tensor([[-1.4249e-07, 2.8592e-06, -1.6127e-03, ..., -1.2627e-03, + 6.2943e-05, -1.7757e-03], + [ 1.0431e-07, 7.6648e-07, 9.7275e-04, ..., 4.0680e-05, + 3.1590e-05, 2.0714e-03], + [-4.9733e-07, -1.3089e-04, 5.2261e-04, ..., 2.9111e-04, + 8.1241e-05, 7.0095e-04], + ..., + [ 3.1572e-07, 1.5163e-04, 2.7943e-03, ..., 1.9588e-03, + 2.9278e-04, 4.5013e-03], + [-2.3749e-06, 3.6359e-06, 8.0013e-04, ..., 8.4734e-04, + 3.6269e-05, 1.2712e-03], + [ 1.6093e-06, -3.4899e-05, -2.7776e-04, ..., 8.2731e-04, + 7.5281e-05, -8.3542e-03]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0292, 0.0353, -0.0006, -0.0065, 0.0246, -0.0291, 0.0066, 0.0019, + -0.0044, -0.0167], device='cuda:0'), grad: tensor([-0.0354, 0.0124, 0.0130, 0.0188, -0.0049, 0.0110, -0.0241, 0.0033, + 0.0137, -0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.24, cls_loss 0.5435 cls_loss_mapping 0.0043 cls_loss_causal 0.4847 re_mapping 0.0100 re_causal 0.0232 /// teacc 98.67 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.0664, -0.1306, -0.0950, ..., -0.0418, 0.0580, -0.1100], + [-0.0735, -0.0898, -0.0753, ..., 0.0965, -0.0435, 0.2194], + [-0.0021, -0.0204, -0.0350, ..., -0.0098, -0.0109, -0.0753], + ..., + [ 0.0364, -0.0985, 0.1580, ..., 0.0321, -0.0916, 0.0527], + [-0.0403, 0.0617, -0.1438, ..., -0.0536, 0.0117, -0.0938], + [-0.0567, 0.0753, 0.0458, ..., -0.0509, -0.0428, -0.0221]], + device='cuda:0'), grad: tensor([[ 3.3712e-04, 9.9018e-06, 1.2553e-04, ..., 9.6369e-04, + 5.8860e-06, 4.1652e-04], + [ 2.7966e-04, 6.9104e-06, 1.0169e-04, ..., 1.0796e-03, + 7.7784e-06, 6.9094e-04], + [-2.6989e-03, 2.9588e-04, 2.2244e-04, ..., -2.7409e-03, + 6.8545e-06, 3.0732e-04], + ..., + [ 4.7040e-04, 3.2902e-04, 3.8952e-05, ..., 1.1005e-03, + 6.0014e-06, -1.4572e-02], + [-9.0504e-04, 4.8218e-03, -1.6785e-03, ..., 1.8024e-04, + 5.9381e-06, -4.9019e-03], + [ 4.2653e-04, 4.6349e-04, -2.5660e-05, ..., 9.3746e-04, + 5.5507e-06, 2.5272e-03]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0296, 0.0345, -0.0005, -0.0073, 0.0243, -0.0282, 0.0069, 0.0022, + -0.0036, -0.0168], device='cuda:0'), grad: tensor([ 0.0172, 0.0179, -0.0085, -0.0050, 0.0059, -0.0086, 0.0419, -0.0593, + 0.0066, -0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 214.22, cls_loss 0.5170 cls_loss_mapping 0.0039 cls_loss_causal 0.4479 re_mapping 0.0100 re_causal 0.0223 /// teacc 98.70 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.0657, -0.1296, -0.0944, ..., -0.0423, 0.0592, -0.1122], + [-0.0737, -0.0900, -0.0746, ..., 0.0965, -0.0436, 0.2197], + [-0.0027, -0.0196, -0.0336, ..., -0.0084, -0.0107, -0.0749], + ..., + [ 0.0362, -0.0997, 0.1564, ..., 0.0326, -0.0910, 0.0539], + [-0.0405, 0.0611, -0.1438, ..., -0.0540, 0.0109, -0.0937], + [-0.0566, 0.0759, 0.0470, ..., -0.0518, -0.0432, -0.0226]], + device='cuda:0'), grad: tensor([[ 1.7472e-06, 3.0041e-05, 1.9145e-04, ..., 8.7976e-04, + 3.5095e-04, 8.7881e-04], + [ 1.5028e-05, -3.7014e-05, -2.2352e-04, ..., -1.7252e-03, + 3.2455e-05, -2.0256e-03], + [ 7.8261e-05, 7.5626e-04, 1.8330e-03, ..., 3.5172e-03, + 2.9135e-04, 1.4124e-03], + ..., + [-2.2620e-05, -3.9029e-04, -2.7790e-03, ..., 1.0157e-03, + 7.8201e-05, -2.2293e-02], + [ 8.3894e-06, 1.6153e-04, 3.1233e-04, ..., -4.7264e-03, + 3.9309e-05, 6.2466e-04], + [ 1.5177e-05, 2.6417e-04, 3.3398e-03, ..., 9.9468e-04, + 1.5795e-04, 2.2659e-02]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0301, 0.0355, 0.0007, -0.0070, 0.0233, -0.0290, 0.0067, 0.0023, + -0.0039, -0.0167], device='cuda:0'), grad: tensor([ 0.0100, -0.0209, 0.0260, -0.0219, 0.0133, 0.0048, 0.0130, -0.0289, + -0.0200, 0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 214.48, cls_loss 0.5701 cls_loss_mapping 0.0043 cls_loss_causal 0.5061 re_mapping 0.0098 re_causal 0.0232 /// teacc 98.80 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.0654, -0.1295, -0.0948, ..., -0.0424, 0.0588, -0.1118], + [-0.0742, -0.0903, -0.0753, ..., 0.0966, -0.0447, 0.2181], + [-0.0025, -0.0207, -0.0333, ..., -0.0072, -0.0109, -0.0743], + ..., + [ 0.0361, -0.0983, 0.1565, ..., 0.0323, -0.0901, 0.0559], + [-0.0404, 0.0606, -0.1434, ..., -0.0546, 0.0135, -0.0937], + [-0.0555, 0.0748, 0.0463, ..., -0.0512, -0.0430, -0.0227]], + device='cuda:0'), grad: tensor([[ 2.2337e-05, 3.8576e-04, 2.0623e-05, ..., -1.8473e-03, + -4.1962e-03, 6.9439e-05], + [ 1.0565e-05, -1.3361e-03, -2.1801e-03, ..., -1.2993e-02, + 6.5804e-04, -1.1963e-02], + [-2.2709e-04, -4.6501e-03, 1.2512e-03, ..., 7.4565e-05, + -1.0204e-03, 7.3004e-04], + ..., + [ 3.2187e-05, 7.7677e-04, -2.8706e-03, ..., 2.2049e-03, + 3.7837e-04, -1.9863e-05], + [ 2.0355e-05, -1.4160e-02, 1.4153e-03, ..., 1.0977e-03, + -3.2990e-02, 1.3008e-03], + [ 4.9062e-06, 5.1498e-04, 3.3855e-04, ..., 1.7757e-03, + 5.0831e-04, 1.4973e-03]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0309, 0.0345, 0.0002, -0.0074, 0.0234, -0.0280, 0.0068, 0.0034, + -0.0036, -0.0166], device='cuda:0'), grad: tensor([-0.0309, -0.0534, -0.0082, 0.0181, 0.0454, 0.0094, 0.0160, -0.0225, + 0.0068, 0.0193], device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 214.33, cls_loss 0.5146 cls_loss_mapping 0.0034 cls_loss_causal 0.4438 re_mapping 0.0104 re_causal 0.0229 /// teacc 98.81 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.0647, -0.1297, -0.0935, ..., -0.0408, 0.0594, -0.1097], + [-0.0741, -0.0898, -0.0754, ..., 0.0974, -0.0450, 0.2173], + [-0.0040, -0.0207, -0.0342, ..., -0.0067, -0.0109, -0.0741], + ..., + [ 0.0375, -0.0987, 0.1572, ..., 0.0317, -0.0903, 0.0553], + [-0.0393, 0.0622, -0.1428, ..., -0.0551, 0.0145, -0.0952], + [-0.0551, 0.0740, 0.0465, ..., -0.0510, -0.0434, -0.0222]], + device='cuda:0'), grad: tensor([[ 1.3695e-03, 1.9622e-04, 3.0082e-07, ..., 4.4394e-04, + 9.3132e-08, 5.7463e-07], + [ 1.6475e-04, 6.4015e-05, 3.8967e-06, ..., 1.1854e-03, + 6.9849e-08, -2.1636e-05], + [ 6.2275e-04, 2.8038e-03, 3.4478e-06, ..., 2.4490e-03, + 1.6296e-04, 4.0308e-06], + ..., + [-2.7714e-03, 3.6030e-03, 1.9836e-04, ..., -3.5343e-03, + 9.6858e-08, 1.9717e-04], + [-2.7966e-04, 9.1612e-05, 4.1068e-05, ..., -5.5618e-03, + 1.1466e-05, -3.0935e-05], + [ 1.2165e-04, -8.7051e-03, 8.8573e-05, ..., 1.6584e-03, + 9.1363e-07, 2.3346e-03]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0297, 0.0351, 0.0004, -0.0069, 0.0233, -0.0279, 0.0053, 0.0038, + -0.0040, -0.0174], device='cuda:0'), grad: tensor([-0.0070, 0.0142, 0.0268, 0.0230, 0.0008, -0.0087, -0.0173, -0.0012, + -0.0420, 0.0115], device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 214.41, cls_loss 0.5588 cls_loss_mapping 0.0049 cls_loss_causal 0.4882 re_mapping 0.0101 re_causal 0.0229 /// teacc 98.79 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.0647, -0.1313, -0.0939, ..., -0.0418, 0.0588, -0.1098], + [-0.0760, -0.0883, -0.0759, ..., 0.0974, -0.0446, 0.2170], + [-0.0043, -0.0201, -0.0342, ..., -0.0070, -0.0110, -0.0750], + ..., + [ 0.0361, -0.1003, 0.1573, ..., 0.0327, -0.0904, 0.0554], + [-0.0374, 0.0617, -0.1435, ..., -0.0558, 0.0151, -0.0962], + [-0.0544, 0.0746, 0.0474, ..., -0.0507, -0.0434, -0.0222]], + device='cuda:0'), grad: tensor([[ 3.1781e-04, 1.9693e-04, 2.5123e-05, ..., 1.1549e-03, + -3.4630e-05, 2.7132e-04], + [-4.8141e-03, -1.6141e-04, 9.5224e-04, ..., -1.4896e-03, + 2.6077e-08, 9.5725e-05], + [ 1.4162e-03, -4.2176e-04, -3.0041e-03, ..., -4.3941e-04, + 2.4121e-07, 2.8634e-04], + ..., + [ 4.8828e-04, 4.1556e-04, 2.5463e-04, ..., -6.1569e-03, + 6.4820e-07, 1.9884e-04], + [ 5.6839e-04, 1.0223e-03, 1.1909e-04, ..., 3.4714e-03, + 4.5262e-06, 1.1797e-03], + [ 3.9697e-04, 6.1178e-04, -3.9387e-04, ..., 1.9798e-03, + 7.2420e-05, 2.2161e-04]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0303, 0.0354, -0.0001, -0.0071, 0.0239, -0.0272, 0.0063, 0.0036, + -0.0049, -0.0175], device='cuda:0'), grad: tensor([ 0.0090, -0.0425, 0.0007, 0.0150, -0.0246, -0.0104, 0.0217, -0.0117, + 0.0255, 0.0172], device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 214.22, cls_loss 0.5497 cls_loss_mapping 0.0049 cls_loss_causal 0.4850 re_mapping 0.0098 re_causal 0.0219 /// teacc 98.70 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.0642, -0.1329, -0.0942, ..., -0.0410, 0.0587, -0.1096], + [-0.0760, -0.0880, -0.0766, ..., 0.0976, -0.0447, 0.2174], + [-0.0056, -0.0217, -0.0346, ..., -0.0077, -0.0101, -0.0751], + ..., + [ 0.0356, -0.1000, 0.1583, ..., 0.0324, -0.0904, 0.0560], + [-0.0366, 0.0621, -0.1422, ..., -0.0547, 0.0154, -0.0958], + [-0.0528, 0.0742, 0.0470, ..., -0.0510, -0.0436, -0.0223]], + device='cuda:0'), grad: tensor([[ 7.8022e-05, 2.4509e-04, 1.9383e-04, ..., 5.7650e-04, + -1.5831e-03, 2.5058e-04], + [ 3.1781e-04, 5.4884e-04, 4.5204e-04, ..., -2.3899e-03, + 1.7092e-05, 7.5912e-04], + [ 2.0194e-04, 1.1387e-03, 4.8828e-03, ..., 2.6398e-03, + 6.2180e-04, 9.3174e-04], + ..., + [-1.5593e-04, 2.8419e-04, 1.5202e-03, ..., -4.4131e-04, + 1.7032e-05, -8.3160e-04], + [-5.0589e-06, -1.2865e-03, 8.1682e-04, ..., 7.3290e-04, + 4.8310e-05, 5.2452e-04], + [ 1.6081e-04, 3.4142e-04, 3.2949e-04, ..., 7.6389e-04, + 7.5996e-05, 5.0259e-04]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0296, 0.0359, -0.0004, -0.0068, 0.0244, -0.0278, 0.0055, 0.0027, + -0.0043, -0.0176], device='cuda:0'), grad: tensor([ 0.0178, -0.0291, 0.0264, -0.0349, -0.0343, -0.0113, 0.0271, -0.0038, + 0.0152, 0.0269], device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 214.34, cls_loss 0.5731 cls_loss_mapping 0.0037 cls_loss_causal 0.4949 re_mapping 0.0096 re_causal 0.0216 /// teacc 98.81 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.0641, -0.1324, -0.0948, ..., -0.0421, 0.0588, -0.1090], + [-0.0772, -0.0890, -0.0771, ..., 0.0989, -0.0450, 0.2174], + [-0.0057, -0.0210, -0.0353, ..., -0.0078, -0.0093, -0.0758], + ..., + [ 0.0356, -0.1013, 0.1585, ..., 0.0322, -0.0916, 0.0564], + [-0.0349, 0.0628, -0.1433, ..., -0.0543, 0.0154, -0.0958], + [-0.0532, 0.0737, 0.0468, ..., -0.0514, -0.0445, -0.0218]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0004, 0.0008, ..., 0.0010, 0.0003, -0.0001], + [ 0.0009, 0.0004, -0.0017, ..., -0.0017, 0.0003, -0.0025], + [-0.0020, 0.0016, 0.0008, ..., -0.0088, 0.0006, 0.0012], + ..., + [ 0.0016, 0.0007, 0.0028, ..., 0.0044, 0.0003, 0.0043], + [ 0.0004, 0.0012, 0.0028, ..., 0.0015, 0.0002, -0.0001], + [ 0.0005, 0.0002, 0.0010, ..., 0.0014, 0.0001, 0.0009]], + device='cuda:0') +Epoch 291, bias, value: tensor([-0.0305, 0.0371, -0.0008, -0.0078, 0.0251, -0.0277, 0.0062, 0.0024, + -0.0044, -0.0177], device='cuda:0'), grad: tensor([ 0.0246, -0.0260, -0.0124, 0.0007, 0.0019, 0.0068, -0.0438, 0.0209, + 0.0119, 0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 214.36, cls_loss 0.5533 cls_loss_mapping 0.0033 cls_loss_causal 0.4761 re_mapping 0.0098 re_causal 0.0225 /// teacc 98.69 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.0642, -0.1329, -0.0961, ..., -0.0418, 0.0595, -0.1092], + [-0.0768, -0.0889, -0.0778, ..., 0.0984, -0.0448, 0.2174], + [-0.0050, -0.0217, -0.0363, ..., -0.0074, -0.0093, -0.0761], + ..., + [ 0.0350, -0.1016, 0.1592, ..., 0.0311, -0.0912, 0.0568], + [-0.0349, 0.0613, -0.1443, ..., -0.0538, 0.0148, -0.0955], + [-0.0530, 0.0735, 0.0464, ..., -0.0512, -0.0416, -0.0222]], + device='cuda:0'), grad: tensor([[-1.9608e-03, 3.1400e-04, -3.3054e-03, ..., -1.8036e-02, + 2.4261e-03, -7.5865e-04], + [ 1.3542e-04, -4.6110e-04, -2.0027e-04, ..., 9.4986e-04, + 1.0544e-04, -6.7472e-05], + [ 2.1589e-04, 1.3292e-04, -4.3464e-04, ..., 8.9264e-04, + 3.0537e-03, 9.5081e-04], + ..., + [ 4.9353e-04, -5.1832e-04, 7.1049e-04, ..., 2.8706e-03, + -1.3039e-02, -2.9392e-03], + [ 5.3787e-04, 3.5334e-04, 9.8896e-04, ..., 3.4466e-03, + 1.0663e-04, 3.8338e-04], + [ 1.2779e-04, 1.8187e-03, 3.6788e-04, ..., 1.7300e-03, + 6.7139e-03, 1.7891e-03]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0314, 0.0369, 0.0003, -0.0062, 0.0247, -0.0290, 0.0051, 0.0023, + -0.0032, -0.0178], device='cuda:0'), grad: tensor([-0.0473, -0.0231, 0.0142, 0.0179, -0.0050, 0.0066, 0.0059, -0.0104, + 0.0018, 0.0394], device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 214.58, cls_loss 0.5847 cls_loss_mapping 0.0039 cls_loss_causal 0.5144 re_mapping 0.0097 re_causal 0.0225 /// teacc 98.80 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.0653, -0.1327, -0.0948, ..., -0.0410, 0.0586, -0.1085], + [-0.0771, -0.0886, -0.0782, ..., 0.0981, -0.0447, 0.2178], + [-0.0064, -0.0212, -0.0369, ..., -0.0082, -0.0097, -0.0770], + ..., + [ 0.0359, -0.1021, 0.1593, ..., 0.0320, -0.0904, 0.0568], + [-0.0348, 0.0614, -0.1460, ..., -0.0540, 0.0147, -0.0950], + [-0.0541, 0.0741, 0.0449, ..., -0.0518, -0.0429, -0.0227]], + device='cuda:0'), grad: tensor([[ 1.4305e-03, 3.1185e-04, 1.0891e-03, ..., 1.7738e-03, + 1.6677e-04, 4.3869e-03], + [ 2.8496e-03, 1.6451e-03, 1.1759e-03, ..., 3.2673e-03, + 1.0896e-04, 6.9542e-03], + [-6.3276e-04, -2.4681e-03, -4.3869e-03, ..., -3.2749e-03, + 5.8353e-05, -1.3816e-04], + ..., + [ 9.0647e-04, 2.8515e-04, 1.2161e-02, ..., 2.4796e-03, + 9.0897e-05, -1.6403e-03], + [-3.4275e-03, 7.6962e-04, 8.0776e-04, ..., -3.9864e-03, + -9.7513e-04, -1.4944e-03], + [ 2.0428e-03, 6.7091e-04, 5.2109e-03, ..., 2.4605e-03, + 9.1612e-05, 3.0518e-03]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0304, 0.0366, -0.0005, -0.0060, 0.0246, -0.0289, 0.0063, 0.0019, + -0.0041, -0.0178], device='cuda:0'), grad: tensor([-0.0245, 0.0509, 0.0011, 0.0066, -0.0429, -0.0214, -0.0021, 0.0343, + -0.0344, 0.0323], device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 214.37, cls_loss 0.5437 cls_loss_mapping 0.0031 cls_loss_causal 0.4769 re_mapping 0.0102 re_causal 0.0238 /// teacc 98.72 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.0666, -0.1334, -0.0956, ..., -0.0420, 0.0584, -0.1078], + [-0.0779, -0.0887, -0.0782, ..., 0.0985, -0.0450, 0.2184], + [-0.0066, -0.0222, -0.0358, ..., -0.0080, -0.0109, -0.0769], + ..., + [ 0.0367, -0.1026, 0.1595, ..., 0.0323, -0.0906, 0.0568], + [-0.0361, 0.0615, -0.1480, ..., -0.0539, 0.0146, -0.0930], + [-0.0547, 0.0743, 0.0454, ..., -0.0521, -0.0435, -0.0234]], + device='cuda:0'), grad: tensor([[ 1.3590e-04, 1.5860e-03, -3.5496e-03, ..., -4.1237e-03, + 1.0407e-04, 2.2297e-03], + [ 2.1264e-05, 5.0879e-04, 2.4986e-03, ..., 3.4122e-03, + 3.5465e-05, 4.9438e-03], + [-5.2881e-04, -1.4210e-03, 1.3180e-03, ..., -5.7268e-04, + 3.4809e-05, 2.2507e-03], + ..., + [ 8.0538e-04, -1.7204e-03, -3.5267e-03, ..., -9.4604e-03, + 1.3553e-05, -7.5340e-03], + [ 1.1748e-04, -4.5357e-03, -9.5444e-03, ..., -1.7881e-03, + -6.8140e-04, -1.0284e-02], + [ 1.4439e-03, -2.6226e-03, 6.0196e-03, ..., 4.0550e-03, + 6.0827e-05, -1.3565e-02]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0304, 0.0359, 0.0001, -0.0060, 0.0242, -0.0292, 0.0064, 0.0024, + -0.0046, -0.0171], device='cuda:0'), grad: tensor([-0.0342, 0.0324, 0.0016, 0.0157, 0.0294, 0.0392, 0.0244, -0.0449, + -0.0443, -0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 214.23, cls_loss 0.5758 cls_loss_mapping 0.0031 cls_loss_causal 0.5006 re_mapping 0.0101 re_causal 0.0238 /// teacc 98.74 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.0656, -0.1340, -0.0950, ..., -0.0415, 0.0606, -0.1063], + [-0.0763, -0.0889, -0.0791, ..., 0.0996, -0.0438, 0.2178], + [-0.0060, -0.0224, -0.0376, ..., -0.0085, -0.0098, -0.0773], + ..., + [ 0.0370, -0.1017, 0.1582, ..., 0.0332, -0.0925, 0.0557], + [-0.0373, 0.0610, -0.1467, ..., -0.0540, 0.0134, -0.0921], + [-0.0538, 0.0742, 0.0475, ..., -0.0518, -0.0438, -0.0226]], + device='cuda:0'), grad: tensor([[-3.8548e-03, -6.8378e-04, -1.4582e-03, ..., -4.6883e-03, + -1.8835e-03, -3.6945e-03], + [ 3.6564e-03, 5.8740e-05, 6.8951e-04, ..., 5.7220e-03, + 1.0986e-03, 2.9850e-03], + [-1.2798e-03, -4.1695e-03, -1.8814e-02, ..., -1.0818e-02, + 8.2016e-05, -4.0245e-03], + ..., + [-1.8539e-03, 1.0767e-03, 1.2848e-02, ..., 1.7128e-03, + 4.7952e-05, 2.2106e-03], + [ 4.8113e-04, 1.2045e-03, 1.8301e-03, ..., 1.8539e-03, + 1.8275e-04, 4.9305e-04], + [ 6.7234e-04, 1.7433e-03, 5.3291e-03, ..., 3.4943e-03, + 5.1928e-04, 1.1454e-03]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0307, 0.0352, 0.0004, -0.0071, 0.0239, -0.0286, 0.0069, 0.0021, + -0.0039, -0.0166], device='cuda:0'), grad: tensor([-0.0233, 0.0333, -0.0634, -0.0028, 0.0170, 0.0134, -0.0162, 0.0175, + -0.0104, 0.0348], device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 214.20, cls_loss 0.5424 cls_loss_mapping 0.0035 cls_loss_causal 0.4673 re_mapping 0.0099 re_causal 0.0227 /// teacc 98.75 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.0661, -0.1335, -0.0954, ..., -0.0426, 0.0611, -0.1064], + [-0.0763, -0.0893, -0.0797, ..., 0.1003, -0.0435, 0.2182], + [-0.0067, -0.0227, -0.0377, ..., -0.0073, -0.0109, -0.0774], + ..., + [ 0.0378, -0.1011, 0.1589, ..., 0.0326, -0.0929, 0.0548], + [-0.0370, 0.0612, -0.1469, ..., -0.0548, 0.0135, -0.0914], + [-0.0535, 0.0736, 0.0453, ..., -0.0519, -0.0432, -0.0232]], + device='cuda:0'), grad: tensor([[ 6.1572e-05, 5.6684e-05, 4.4060e-04, ..., 1.2579e-03, + 3.7968e-05, 3.9363e-04], + [ 4.7237e-05, 9.5427e-05, 5.3406e-04, ..., 2.4147e-03, + 2.1905e-05, 4.9210e-04], + [ 1.2660e-04, 8.6665e-05, 1.4381e-03, ..., -4.8103e-03, + 3.0637e-05, -2.4319e-03], + ..., + [-4.1389e-03, 9.6512e-04, -7.6962e-04, ..., 1.8692e-03, + 5.3263e-04, -2.1038e-03], + [ 1.0099e-03, -2.2831e-03, 5.8222e-04, ..., -8.0681e-04, + -4.0507e-04, 1.8387e-03], + [ 2.4834e-03, -2.4986e-03, -5.5389e-03, ..., -8.7357e-03, + -1.4534e-03, -4.0102e-04]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0314, 0.0360, 0.0005, -0.0076, 0.0234, -0.0271, 0.0073, 0.0023, + -0.0041, -0.0175], device='cuda:0'), grad: tensor([ 0.0147, -0.0072, -0.0110, 0.0329, 0.0171, 0.0185, -0.0432, 0.0039, + 0.0046, -0.0304], device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 214.15, cls_loss 0.5791 cls_loss_mapping 0.0037 cls_loss_causal 0.5114 re_mapping 0.0094 re_causal 0.0231 /// teacc 98.69 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.0676, -0.1347, -0.0957, ..., -0.0434, 0.0606, -0.1049], + [-0.0759, -0.0895, -0.0781, ..., 0.1001, -0.0430, 0.2188], + [-0.0061, -0.0222, -0.0395, ..., -0.0048, -0.0113, -0.0772], + ..., + [ 0.0368, -0.1024, 0.1601, ..., 0.0327, -0.0937, 0.0547], + [-0.0376, 0.0611, -0.1459, ..., -0.0547, 0.0130, -0.0920], + [-0.0526, 0.0737, 0.0453, ..., -0.0515, -0.0430, -0.0232]], + device='cuda:0'), grad: tensor([[ 1.4342e-07, 1.3065e-04, 3.8433e-04, ..., 6.6948e-04, + 2.4140e-04, 2.9278e-04], + [ 1.6950e-07, 8.0943e-05, 2.6679e-04, ..., 5.2357e-04, + 1.8144e-04, 1.4853e-04], + [ 1.7229e-06, 1.7157e-03, 1.5154e-03, ..., 2.7227e-04, + 3.8738e-03, 4.6277e-04], + ..., + [ 5.4203e-06, 2.1005e-04, 7.0477e-04, ..., 9.8610e-04, + 4.4179e-04, -2.1561e-02], + [ 1.3709e-06, -9.4223e-03, -5.4131e-03, ..., 6.6471e-04, + -2.0676e-02, 3.5429e-04], + [ 3.6601e-06, 1.2445e-04, -2.1725e-03, ..., -2.4662e-03, + 2.6965e-04, 1.9547e-02]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0307, 0.0370, 0.0015, -0.0076, 0.0225, -0.0277, 0.0071, 0.0014, + -0.0047, -0.0170], device='cuda:0'), grad: tensor([ 0.0166, 0.0206, 0.0221, 0.0360, -0.0445, 0.0189, -0.0116, -0.0388, + -0.0411, 0.0218], device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 214.29, cls_loss 0.5360 cls_loss_mapping 0.0048 cls_loss_causal 0.4685 re_mapping 0.0099 re_causal 0.0216 /// teacc 98.65 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.0692, -0.1351, -0.0956, ..., -0.0432, 0.0596, -0.1064], + [-0.0764, -0.0897, -0.0793, ..., 0.0997, -0.0430, 0.2197], + [-0.0066, -0.0221, -0.0387, ..., -0.0048, -0.0107, -0.0767], + ..., + [ 0.0373, -0.1020, 0.1615, ..., 0.0343, -0.0948, 0.0552], + [-0.0370, 0.0608, -0.1465, ..., -0.0547, 0.0142, -0.0926], + [-0.0529, 0.0741, 0.0443, ..., -0.0522, -0.0429, -0.0239]], + device='cuda:0'), grad: tensor([[ 6.3086e-04, 3.5977e-04, 1.2617e-03, ..., 2.6302e-03, + 1.1981e-04, 7.8535e-04], + [ 5.5742e-04, 6.4087e-04, 1.4229e-03, ..., 4.0932e-03, + 2.7210e-05, 8.2445e-04], + [ 2.1744e-03, 1.4563e-03, -1.2236e-03, ..., 5.5542e-03, + 8.7976e-05, 2.1534e-03], + ..., + [-5.7220e-04, 6.4039e-04, 2.4872e-03, ..., 4.2458e-03, + 1.7333e-04, 2.9163e-03], + [ 7.8869e-04, 5.6648e-03, 3.1853e-03, ..., 6.0654e-03, + 1.3137e-04, 1.8616e-03], + [ 8.8196e-03, 5.4741e-04, 1.9474e-03, ..., 2.7599e-03, + 8.3208e-05, -7.6828e-03]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0306, 0.0369, 0.0018, -0.0091, 0.0230, -0.0278, 0.0069, 0.0029, + -0.0045, -0.0175], device='cuda:0'), grad: tensor([ 0.0251, 0.0313, 0.0102, -0.0199, -0.0341, -0.0719, -0.0156, 0.0434, + 0.0148, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 214.31, cls_loss 0.5190 cls_loss_mapping 0.0044 cls_loss_causal 0.4552 re_mapping 0.0098 re_causal 0.0217 /// teacc 98.76 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.0695, -0.1357, -0.0967, ..., -0.0437, 0.0601, -0.1070], + [-0.0774, -0.0898, -0.0794, ..., 0.0996, -0.0432, 0.2202], + [-0.0073, -0.0224, -0.0405, ..., -0.0064, -0.0107, -0.0777], + ..., + [ 0.0380, -0.1013, 0.1628, ..., 0.0348, -0.0957, 0.0550], + [-0.0364, 0.0601, -0.1465, ..., -0.0543, 0.0148, -0.0921], + [-0.0524, 0.0751, 0.0448, ..., -0.0513, -0.0423, -0.0254]], + device='cuda:0'), grad: tensor([[ 7.6914e-04, -5.2500e-04, 1.1826e-03, ..., 4.0779e-03, + 1.5318e-04, 5.9128e-04], + [ 2.0733e-03, 2.3559e-05, 2.0676e-03, ..., 2.4109e-03, + 1.9312e-04, 1.8406e-04], + [ 2.3460e-03, 5.2547e-04, 2.9964e-03, ..., 3.1891e-03, + 6.7616e-04, 1.1379e-04], + ..., + [-5.8479e-03, 9.4891e-05, -3.7098e-03, ..., -3.3169e-03, + 4.8518e-04, 1.2362e-04], + [ 1.2913e-03, -1.6613e-03, 2.7227e-04, ..., 5.4407e-04, + -1.3578e-04, 2.3632e-03], + [-3.3283e-03, 1.2541e-04, 1.0042e-03, ..., -5.1498e-03, + 3.6335e-04, 4.9496e-04]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0315, 0.0366, 0.0006, -0.0088, 0.0235, -0.0286, 0.0075, 0.0027, + -0.0043, -0.0159], device='cuda:0'), grad: tensor([-0.0246, 0.0220, 0.0183, -0.0324, 0.0135, 0.0132, 0.0133, -0.0393, + 0.0180, -0.0019], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 298---------------------------------------------------- +epoch 298, time 230.97, cls_loss 0.5381 cls_loss_mapping 0.0030 cls_loss_causal 0.4792 re_mapping 0.0095 re_causal 0.0224 /// teacc 98.87 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.0704, -0.1357, -0.0968, ..., -0.0438, 0.0610, -0.1074], + [-0.0774, -0.0896, -0.0789, ..., 0.0995, -0.0435, 0.2203], + [-0.0085, -0.0215, -0.0402, ..., -0.0059, -0.0104, -0.0776], + ..., + [ 0.0397, -0.1006, 0.1640, ..., 0.0360, -0.0945, 0.0560], + [-0.0362, 0.0598, -0.1481, ..., -0.0544, 0.0132, -0.0927], + [-0.0532, 0.0756, 0.0442, ..., -0.0530, -0.0443, -0.0264]], + device='cuda:0'), grad: tensor([[-5.5432e-06, 2.4164e-04, -3.1700e-03, ..., -8.4763e-03, + 2.5034e-04, -4.7112e-03], + [ 2.2733e-04, -3.6316e-03, -4.6005e-03, ..., -1.1040e-02, + 3.5691e-04, -2.2678e-03], + [ 1.7471e-03, 1.5945e-03, 4.1580e-03, ..., 6.9542e-03, + 2.7847e-04, 1.3123e-03], + ..., + [ 4.5800e-04, 2.9755e-04, -2.0966e-05, ..., 2.1038e-03, + 1.6952e-04, 7.7820e-04], + [ 4.6730e-03, -1.0662e-03, 2.1660e-04, ..., 8.3618e-03, + -5.1498e-04, 1.0185e-03], + [ 2.0826e-04, 7.4100e-04, -1.6010e-04, ..., 8.0824e-04, + 2.8706e-04, 1.0424e-03]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0310, 0.0374, 0.0008, -0.0094, 0.0236, -0.0280, 0.0082, 0.0031, + -0.0053, -0.0175], device='cuda:0'), grad: tensor([-0.0764, -0.0228, 0.0478, -0.0054, 0.0251, -0.0125, 0.0294, 0.0100, + 0.0160, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 214.49, cls_loss 0.5649 cls_loss_mapping 0.0033 cls_loss_causal 0.4883 re_mapping 0.0092 re_causal 0.0228 /// teacc 98.80 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.0713, -0.1356, -0.0978, ..., -0.0446, 0.0604, -0.1089], + [-0.0780, -0.0895, -0.0797, ..., 0.0990, -0.0439, 0.2210], + [-0.0065, -0.0212, -0.0405, ..., -0.0060, -0.0102, -0.0777], + ..., + [ 0.0396, -0.1007, 0.1636, ..., 0.0356, -0.0948, 0.0549], + [-0.0366, 0.0600, -0.1469, ..., -0.0544, 0.0127, -0.0929], + [-0.0530, 0.0755, 0.0447, ..., -0.0523, -0.0442, -0.0259]], + device='cuda:0'), grad: tensor([[ 4.5002e-05, 8.7395e-06, 6.1798e-04, ..., -3.9330e-03, + 1.5473e-04, 7.9012e-04], + [ 2.2873e-05, 1.5646e-07, -7.4863e-05, ..., -2.1327e-04, + -8.4877e-04, -5.2571e-05], + [ 4.8786e-05, 1.4398e-06, -5.7411e-04, ..., 9.6703e-04, + 2.2256e-04, 1.9240e-04], + ..., + [ 1.4275e-05, 8.3819e-08, -1.8692e-03, ..., -7.6246e-04, + 1.0896e-04, 2.1636e-04], + [ 2.3559e-05, -2.1420e-03, -2.3537e-03, ..., 3.3360e-03, + 2.1958e-04, 5.9462e-04], + [ 3.0547e-05, 1.0476e-05, 7.0286e-04, ..., -4.1656e-03, + 1.3697e-04, -2.8057e-03]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0309, 0.0368, 0.0008, -0.0091, 0.0233, -0.0278, 0.0073, 0.0032, + -0.0044, -0.0174], device='cuda:0'), grad: tensor([ 0.0010, -0.0039, 0.0189, -0.0459, 0.0221, -0.0128, 0.0434, -0.0388, + 0.0253, -0.0093], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 300---------------------------------------------------- +epoch 300, time 230.28, cls_loss 0.5614 cls_loss_mapping 0.0030 cls_loss_causal 0.4895 re_mapping 0.0091 re_causal 0.0222 /// teacc 98.90 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.0722, -0.1370, -0.0979, ..., -0.0448, 0.0601, -0.1097], + [-0.0775, -0.0891, -0.0791, ..., 0.1001, -0.0436, 0.2210], + [-0.0064, -0.0219, -0.0407, ..., -0.0076, -0.0089, -0.0779], + ..., + [ 0.0400, -0.1011, 0.1643, ..., 0.0363, -0.0952, 0.0548], + [-0.0364, 0.0607, -0.1467, ..., -0.0538, 0.0127, -0.0930], + [-0.0530, 0.0758, 0.0450, ..., -0.0524, -0.0441, -0.0257]], + device='cuda:0'), grad: tensor([[ 1.0262e-03, 7.7534e-04, 2.5673e-03, ..., 2.8419e-03, + 9.8825e-05, 2.0579e-05], + [ 4.9973e-04, 1.5223e-04, 2.2068e-03, ..., 2.1420e-03, + 4.4815e-06, -1.3292e-05], + [-2.0580e-03, 3.4409e-03, 5.2643e-03, ..., 6.5422e-04, + 2.8014e-04, 1.3366e-05], + ..., + [ 2.2161e-04, 1.1307e-04, -8.0299e-04, ..., -3.0880e-03, + 2.4810e-06, 2.3019e-04], + [ 9.0170e-04, 1.6689e-04, 2.2297e-03, ..., 2.1820e-03, + 8.1584e-06, 3.2395e-05], + [ 5.4646e-04, 7.0810e-05, 1.0500e-03, ..., 9.3269e-04, + 4.2953e-06, 5.0688e-04]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0308, 0.0372, -0.0005, -0.0086, 0.0226, -0.0279, 0.0072, 0.0037, + -0.0042, -0.0169], device='cuda:0'), grad: tensor([ 0.0181, 0.0198, 0.0106, -0.0269, -0.0126, 0.0151, -0.0123, -0.0466, + 0.0190, 0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 214.32, cls_loss 0.5257 cls_loss_mapping 0.0032 cls_loss_causal 0.4599 re_mapping 0.0101 re_causal 0.0234 /// teacc 98.73 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.0709, -0.1369, -0.0960, ..., -0.0463, 0.0597, -0.1101], + [-0.0777, -0.0895, -0.0791, ..., 0.1001, -0.0442, 0.2217], + [-0.0043, -0.0219, -0.0392, ..., -0.0074, -0.0076, -0.0780], + ..., + [ 0.0394, -0.1019, 0.1639, ..., 0.0367, -0.0948, 0.0546], + [-0.0366, 0.0613, -0.1460, ..., -0.0530, 0.0118, -0.0921], + [-0.0529, 0.0757, 0.0457, ..., -0.0526, -0.0441, -0.0263]], + device='cuda:0'), grad: tensor([[ 9.8050e-05, 1.1086e-04, 3.7670e-05, ..., -7.8278e-03, + 4.9472e-05, 1.1064e-05], + [ 6.0022e-05, 3.2693e-05, 2.5019e-05, ..., 5.8317e-04, + 9.7230e-06, -2.0981e-05], + [ 3.0088e-04, 2.5821e-04, 7.5519e-05, ..., 2.5101e-03, + 4.5687e-05, 6.9514e-06], + ..., + [-8.6451e-04, 3.7551e-05, 1.1539e-04, ..., -3.1509e-03, + 9.9018e-06, -1.3053e-04], + [-2.4721e-05, -3.7079e-03, -1.3666e-03, ..., -1.1911e-03, + 6.2943e-05, 7.7710e-06], + [ 1.0026e-04, 1.0478e-04, 9.2685e-06, ..., 6.5994e-04, + 6.9514e-06, 1.2755e-04]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0303, 0.0369, 0.0002, -0.0083, 0.0220, -0.0291, 0.0077, 0.0035, + -0.0045, -0.0165], device='cuda:0'), grad: tensor([-0.0247, 0.0104, 0.0153, 0.0104, 0.0020, 0.0077, 0.0214, -0.0269, + -0.0193, 0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 214.23, cls_loss 0.5430 cls_loss_mapping 0.0051 cls_loss_causal 0.4648 re_mapping 0.0100 re_causal 0.0227 /// teacc 98.75 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.0706, -0.1361, -0.0968, ..., -0.0463, 0.0600, -0.1106], + [-0.0776, -0.0899, -0.0789, ..., 0.1006, -0.0437, 0.2231], + [-0.0040, -0.0234, -0.0398, ..., -0.0078, -0.0079, -0.0791], + ..., + [ 0.0381, -0.1023, 0.1644, ..., 0.0364, -0.0961, 0.0548], + [-0.0366, 0.0626, -0.1455, ..., -0.0538, 0.0125, -0.0928], + [-0.0514, 0.0753, 0.0461, ..., -0.0531, -0.0448, -0.0268]], + device='cuda:0'), grad: tensor([[ 1.8346e-04, -6.0439e-05, -6.4468e-04, ..., -1.1768e-03, + 1.3590e-04, -7.2908e-04], + [ 7.7605e-05, 7.1339e-07, 7.1144e-04, ..., 8.1921e-04, + 1.4722e-04, 3.8952e-05], + [ 3.4952e-04, 3.7812e-06, 1.5593e-03, ..., 1.6985e-03, + 2.3937e-04, 6.0177e-04], + ..., + [ 2.5496e-05, 6.9141e-06, -5.5504e-03, ..., 5.2547e-04, + 1.2743e-04, -2.7027e-03], + [-8.2626e-03, 1.0997e-05, -4.0016e-03, ..., -1.4114e-03, + 2.3317e-04, 1.2751e-03], + [ 4.9978e-05, 1.5509e-04, 4.0412e-04, ..., 8.7309e-04, + 1.3208e-04, 3.7819e-05]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0303, 0.0370, 0.0007, -0.0084, 0.0232, -0.0293, 0.0071, 0.0035, + -0.0047, -0.0172], device='cuda:0'), grad: tensor([-0.0282, 0.0161, 0.0475, 0.0076, 0.0116, -0.0019, -0.0282, -0.0215, + -0.0018, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 214.05, cls_loss 0.5533 cls_loss_mapping 0.0034 cls_loss_causal 0.4823 re_mapping 0.0095 re_causal 0.0217 /// teacc 98.78 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.0707, -0.1361, -0.0954, ..., -0.0438, 0.0579, -0.1085], + [-0.0764, -0.0903, -0.0790, ..., 0.0993, -0.0437, 0.2221], + [-0.0039, -0.0234, -0.0402, ..., -0.0074, -0.0078, -0.0791], + ..., + [ 0.0367, -0.1018, 0.1655, ..., 0.0371, -0.0968, 0.0552], + [-0.0363, 0.0625, -0.1455, ..., -0.0539, 0.0114, -0.0930], + [-0.0511, 0.0755, 0.0462, ..., -0.0518, -0.0453, -0.0274]], + device='cuda:0'), grad: tensor([[ 1.2791e-04, 1.2314e-04, -4.8409e-03, ..., 1.9112e-03, + 6.1631e-05, -3.1708e-02], + [ 4.3392e-05, 2.5311e-03, -3.4882e-02, ..., -6.3400e-03, + -2.3060e-03, -1.0735e-02], + [-8.1587e-04, 3.5346e-05, 2.5845e-03, ..., 1.1148e-03, + 1.3990e-03, 1.1950e-03], + ..., + [ 2.6058e-06, 2.4872e-03, 3.6102e-02, ..., 1.0099e-03, + 7.3433e-05, 2.8885e-02], + [ 5.8937e-04, 1.8711e-03, 4.7836e-03, ..., 2.9297e-03, + 1.5700e-04, 1.3336e-02], + [ 2.2709e-05, 4.4727e-04, 2.3289e-03, ..., 3.9940e-03, + 2.1383e-05, 4.9734e-04]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0299, 0.0370, 0.0004, -0.0091, 0.0244, -0.0302, 0.0071, 0.0041, + -0.0055, -0.0165], device='cuda:0'), grad: tensor([-0.0042, -0.0242, 0.0021, -0.0253, -0.0110, -0.0126, 0.0229, 0.0517, + -0.0226, 0.0234], device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 214.36, cls_loss 0.5151 cls_loss_mapping 0.0036 cls_loss_causal 0.4422 re_mapping 0.0099 re_causal 0.0220 /// teacc 98.81 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.0719, -0.1359, -0.0951, ..., -0.0438, 0.0582, -0.1070], + [-0.0749, -0.0902, -0.0778, ..., 0.0992, -0.0414, 0.2223], + [-0.0026, -0.0239, -0.0401, ..., -0.0073, -0.0087, -0.0795], + ..., + [ 0.0361, -0.1019, 0.1643, ..., 0.0380, -0.0974, 0.0556], + [-0.0370, 0.0629, -0.1461, ..., -0.0548, 0.0119, -0.0927], + [-0.0511, 0.0759, 0.0471, ..., -0.0530, -0.0445, -0.0289]], + device='cuda:0'), grad: tensor([[ 2.6655e-04, 2.3308e-03, 6.8092e-04, ..., 1.7595e-04, + 8.4579e-05, 1.6487e-04], + [-1.8320e-03, 1.0419e-04, 6.2132e-04, ..., -4.8027e-03, + -1.9989e-03, 1.4460e-04], + [ 2.2564e-03, -4.5776e-03, 3.0518e-03, ..., 4.9744e-03, + 1.1597e-03, 1.1247e-04], + ..., + [ 2.4629e-04, 8.9455e-04, 3.1158e-02, ..., 5.7640e-03, + 1.9097e-04, 1.7929e-02], + [ 7.6008e-04, 2.4185e-03, 1.5202e-03, ..., -1.8444e-03, + 3.5375e-05, -1.3857e-03], + [ 3.4189e-04, -1.2875e-03, -3.3447e-02, ..., -3.6259e-03, + 2.3568e-04, -1.7975e-02]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0300, 0.0363, -0.0002, -0.0084, 0.0242, -0.0294, 0.0069, 0.0048, + -0.0052, -0.0172], device='cuda:0'), grad: tensor([ 0.0171, -0.0234, 0.0033, -0.0207, 0.0063, -0.0204, 0.0031, 0.0418, + 0.0242, -0.0313], device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 214.37, cls_loss 0.5548 cls_loss_mapping 0.0033 cls_loss_causal 0.4946 re_mapping 0.0096 re_causal 0.0224 /// teacc 98.79 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.0707, -0.1365, -0.0957, ..., -0.0433, 0.0584, -0.1074], + [-0.0735, -0.0883, -0.0762, ..., 0.0992, -0.0399, 0.2235], + [-0.0028, -0.0224, -0.0395, ..., -0.0083, -0.0083, -0.0797], + ..., + [ 0.0371, -0.1025, 0.1646, ..., 0.0383, -0.0977, 0.0553], + [-0.0383, 0.0622, -0.1463, ..., -0.0556, 0.0125, -0.0943], + [-0.0521, 0.0760, 0.0462, ..., -0.0540, -0.0448, -0.0280]], + device='cuda:0'), grad: tensor([[ 7.4841e-06, 1.7488e-04, 4.2725e-04, ..., -1.6232e-03, + 1.3947e-04, 9.6798e-04], + [ 4.8131e-06, 9.2685e-05, -1.6046e-04, ..., -8.3771e-03, + 1.3649e-04, -2.6512e-03], + [ 3.0190e-05, 3.8147e-04, 2.0638e-03, ..., 2.4948e-03, + 4.8256e-04, 5.1727e-03], + ..., + [ 1.6719e-05, 2.4009e-04, -5.8060e-03, ..., -1.0557e-03, + 2.8825e-04, -8.0948e-03], + [ 3.4541e-05, 5.3749e-03, 7.3290e-04, ..., 1.8673e-03, + 8.8882e-03, 6.2132e-04], + [ 4.0948e-05, -5.0621e-03, 9.0837e-04, ..., 1.4849e-03, + -1.0109e-02, 6.3229e-04]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0296, 0.0370, -0.0005, -0.0089, 0.0244, -0.0295, 0.0071, 0.0051, + -0.0056, -0.0177], device='cuda:0'), grad: tensor([-0.0350, -0.0245, 0.0220, -0.0031, 0.0219, 0.0394, -0.0023, -0.0230, + 0.0312, -0.0266], device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 214.46, cls_loss 0.5533 cls_loss_mapping 0.0033 cls_loss_causal 0.4785 re_mapping 0.0096 re_causal 0.0237 /// teacc 98.80 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.0716, -0.1366, -0.0967, ..., -0.0427, 0.0587, -0.1086], + [-0.0720, -0.0880, -0.0768, ..., 0.0997, -0.0390, 0.2245], + [-0.0029, -0.0227, -0.0395, ..., -0.0082, -0.0083, -0.0809], + ..., + [ 0.0368, -0.1034, 0.1647, ..., 0.0366, -0.0999, 0.0555], + [-0.0385, 0.0620, -0.1458, ..., -0.0531, 0.0130, -0.0921], + [-0.0523, 0.0766, 0.0465, ..., -0.0555, -0.0458, -0.0276]], + device='cuda:0'), grad: tensor([[-3.2592e-04, -4.1103e-04, 6.3705e-04, ..., -6.3442e-06, + -8.7917e-05, 1.3952e-03], + [ 3.0007e-06, 3.2973e-04, -3.5572e-03, ..., -5.1460e-03, + 8.2731e-05, -3.3245e-03], + [ 1.1319e-04, -2.4986e-03, -5.3101e-03, ..., -2.4166e-03, + -9.7132e-04, -3.7174e-03], + ..., + [ 1.6853e-05, -1.7154e-04, -2.5585e-05, ..., -8.4591e-04, + 1.0252e-04, -5.0163e-03], + [ 3.6269e-05, 6.8235e-04, 2.2182e-03, ..., 3.5515e-03, + 1.9836e-04, 1.7462e-03], + [ 4.9710e-05, 6.5506e-05, -7.5102e-04, ..., 6.9809e-04, + 1.3578e-04, 4.1122e-03]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0289, 0.0372, -0.0022, -0.0077, 0.0236, -0.0280, 0.0075, 0.0041, + -0.0050, -0.0187], device='cuda:0'), grad: tensor([ 0.0031, -0.0302, -0.0126, 0.0150, 0.0094, 0.0217, -0.0030, -0.0051, + 0.0169, -0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 214.44, cls_loss 0.5827 cls_loss_mapping 0.0042 cls_loss_causal 0.5089 re_mapping 0.0092 re_causal 0.0220 /// teacc 98.85 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.0728, -0.1374, -0.0974, ..., -0.0410, 0.0583, -0.1101], + [-0.0725, -0.0879, -0.0761, ..., 0.1001, -0.0402, 0.2252], + [-0.0019, -0.0229, -0.0397, ..., -0.0078, -0.0077, -0.0814], + ..., + [ 0.0369, -0.1040, 0.1647, ..., 0.0359, -0.1008, 0.0548], + [-0.0389, 0.0624, -0.1463, ..., -0.0525, 0.0142, -0.0918], + [-0.0529, 0.0763, 0.0465, ..., -0.0571, -0.0461, -0.0279]], + device='cuda:0'), grad: tensor([[ 6.2585e-05, 1.4037e-05, 1.3351e-04, ..., 6.8569e-04, + -5.2595e-04, 8.7082e-05], + [ 2.2992e-05, 5.4277e-06, 5.3972e-05, ..., 2.0466e-03, + 1.3649e-04, 5.9158e-05], + [-3.7766e-04, 7.9691e-05, -6.9141e-04, ..., -2.7771e-03, + -8.4579e-05, 2.7966e-04], + ..., + [ 1.7178e-04, 2.8896e-04, 4.4489e-04, ..., -1.1349e-03, + -9.6283e-03, -5.4016e-03], + [-6.2847e-04, -2.2335e-03, -2.2106e-03, ..., -9.8419e-04, + 2.9707e-04, -5.0211e-04], + [ 4.2033e-04, 1.4086e-03, 1.4019e-03, ..., 2.3115e-04, + 1.9016e-03, 1.0843e-03]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0284, 0.0375, -0.0033, -0.0078, 0.0233, -0.0286, 0.0086, 0.0038, + -0.0040, -0.0193], device='cuda:0'), grad: tensor([ 0.0120, 0.0165, 0.0062, -0.0152, 0.0310, 0.0173, 0.0145, -0.0760, + 0.0041, -0.0104], device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 216.03, cls_loss 0.5669 cls_loss_mapping 0.0036 cls_loss_causal 0.4994 re_mapping 0.0096 re_causal 0.0233 /// teacc 98.87 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.0730, -0.1376, -0.0976, ..., -0.0418, 0.0565, -0.1103], + [-0.0730, -0.0889, -0.0771, ..., 0.0999, -0.0409, 0.2239], + [-0.0022, -0.0239, -0.0405, ..., -0.0073, -0.0074, -0.0806], + ..., + [ 0.0368, -0.1044, 0.1654, ..., 0.0356, -0.1019, 0.0545], + [-0.0397, 0.0616, -0.1474, ..., -0.0536, 0.0131, -0.0912], + [-0.0529, 0.0761, 0.0475, ..., -0.0553, -0.0445, -0.0270]], + device='cuda:0'), grad: tensor([[ 2.7753e-06, 2.0862e-04, 7.8869e-04, ..., 1.4353e-03, + 1.7667e-04, 2.7919e-04], + [ 3.6191e-06, 1.1593e-05, 1.9002e-04, ..., 5.5361e-04, + 6.5565e-05, 3.2091e-04], + [ 5.1767e-05, -2.4092e-04, -3.0346e-03, ..., -1.0672e-03, + -1.2350e-04, 2.8324e-04], + ..., + [ 1.0774e-05, 1.4305e-04, -7.5065e-06, ..., 1.2712e-03, + 1.0145e-04, 2.1362e-04], + [-1.5903e-04, -1.3056e-03, 4.5705e-04, ..., 1.8015e-03, + 7.9989e-05, 2.6488e-04], + [ 3.9078e-06, 1.3685e-04, 6.1131e-04, ..., 1.4839e-03, + 1.4639e-04, 3.9959e-04]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0283, 0.0370, -0.0025, -0.0081, 0.0229, -0.0292, 0.0080, 0.0041, + -0.0039, -0.0181], device='cuda:0'), grad: tensor([ 0.0194, -0.0128, -0.0103, -0.0103, -0.0104, -0.0090, 0.0179, -0.0051, + -0.0032, 0.0238], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 309---------------------------------------------------- +epoch 309, time 233.29, cls_loss 0.5324 cls_loss_mapping 0.0058 cls_loss_causal 0.4597 re_mapping 0.0094 re_causal 0.0220 /// teacc 98.97 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.0729, -0.1373, -0.0982, ..., -0.0423, 0.0567, -0.1102], + [-0.0739, -0.0906, -0.0764, ..., 0.1011, -0.0403, 0.2242], + [-0.0015, -0.0234, -0.0404, ..., -0.0078, -0.0080, -0.0804], + ..., + [ 0.0374, -0.1049, 0.1662, ..., 0.0358, -0.1030, 0.0547], + [-0.0401, 0.0618, -0.1480, ..., -0.0543, 0.0131, -0.0919], + [-0.0525, 0.0775, 0.0477, ..., -0.0558, -0.0439, -0.0278]], + device='cuda:0'), grad: tensor([[ 1.0356e-06, 9.4771e-06, 4.8256e-04, ..., 2.8431e-05, + 0.0000e+00, -6.8128e-05], + [ 1.3039e-08, -6.6496e-07, -4.3654e-04, ..., 1.0056e-02, + 0.0000e+00, -4.7013e-06], + [ 8.4378e-07, 7.9423e-06, -7.1192e-04, ..., 1.1082e-03, + 3.7253e-09, 1.9157e-04], + ..., + [ 2.7381e-07, 4.2230e-05, 9.3079e-04, ..., 9.2850e-03, + 1.8626e-09, -1.7476e-04], + [ 2.3190e-06, 2.6181e-05, -6.1131e-04, ..., 2.7027e-03, + 1.1735e-07, 4.0919e-05], + [ 2.0728e-05, 1.1259e-04, -1.6327e-03, ..., -1.9882e-02, + 0.0000e+00, -4.5276e-04]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0293, 0.0382, -0.0020, -0.0090, 0.0236, -0.0297, 0.0078, 0.0040, + -0.0040, -0.0178], device='cuda:0'), grad: tensor([ 0.0107, 0.0085, -0.0146, 0.0176, -0.0042, -0.0106, 0.0167, 0.0325, + -0.0095, -0.0472], device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 217.34, cls_loss 0.5278 cls_loss_mapping 0.0024 cls_loss_causal 0.4609 re_mapping 0.0094 re_causal 0.0221 /// teacc 98.86 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.0716, -0.1369, -0.0983, ..., -0.0419, 0.0576, -0.1108], + [-0.0726, -0.0915, -0.0767, ..., 0.1001, -0.0391, 0.2244], + [-0.0028, -0.0244, -0.0404, ..., -0.0089, -0.0079, -0.0804], + ..., + [ 0.0374, -0.1049, 0.1663, ..., 0.0369, -0.1029, 0.0542], + [-0.0393, 0.0622, -0.1492, ..., -0.0558, 0.0131, -0.0926], + [-0.0533, 0.0785, 0.0474, ..., -0.0562, -0.0456, -0.0275]], + device='cuda:0'), grad: tensor([[ 3.2596e-07, 1.5891e-04, 2.2328e-04, ..., 4.0359e-03, + 4.6444e-04, 1.1349e-04], + [ 2.2165e-06, 6.5446e-05, 4.2748e-04, ..., -4.7226e-03, + 1.0127e-04, -1.4529e-07], + [ 3.1948e-05, 6.9022e-05, -3.2425e-03, ..., 4.8733e-04, + 9.6083e-05, 1.8373e-05], + ..., + [ 1.2361e-05, 1.6046e-04, -5.3329e-03, ..., -1.0653e-03, + 1.6129e-04, -6.4634e-07], + [ 1.1295e-05, 3.4828e-03, 4.4203e-04, ..., 2.8706e-03, + 2.0123e-04, 2.3782e-05], + [ 1.2480e-06, -1.8549e-04, 2.6493e-03, ..., 2.2888e-03, + 3.3045e-04, 5.5760e-05]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0284, 0.0376, -0.0031, -0.0072, 0.0248, -0.0304, 0.0074, 0.0036, + -0.0042, -0.0183], device='cuda:0'), grad: tensor([ 0.0194, -0.0198, -0.0478, 0.0243, -0.0481, 0.0016, -0.0094, 0.0059, + 0.0274, 0.0464], device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 214.83, cls_loss 0.5729 cls_loss_mapping 0.0038 cls_loss_causal 0.5081 re_mapping 0.0088 re_causal 0.0212 /// teacc 98.88 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.0725, -0.1364, -0.0988, ..., -0.0429, 0.0585, -0.1103], + [-0.0720, -0.0916, -0.0764, ..., 0.1006, -0.0377, 0.2245], + [-0.0022, -0.0247, -0.0419, ..., -0.0085, -0.0090, -0.0802], + ..., + [ 0.0365, -0.1044, 0.1675, ..., 0.0374, -0.1036, 0.0542], + [-0.0403, 0.0610, -0.1494, ..., -0.0555, 0.0120, -0.0924], + [-0.0527, 0.0795, 0.0472, ..., -0.0576, -0.0454, -0.0278]], + device='cuda:0'), grad: tensor([[ 2.9877e-06, 3.8290e-04, -2.6131e-04, ..., 1.2712e-03, + 7.6408e-03, 2.6727e-04], + [ 4.4629e-06, 2.7791e-05, 2.6493e-03, ..., 6.0806e-03, + 1.4343e-03, -4.3541e-05], + [ 9.0659e-05, 1.8864e-03, 5.4855e-03, ..., 5.2071e-03, + -9.2392e-03, 6.8140e-04], + ..., + [ 8.8140e-06, 8.8274e-05, 1.8320e-03, ..., 2.3308e-03, + 8.7070e-04, -1.1139e-03], + [-2.2144e-03, -2.0962e-03, -6.1188e-03, ..., -1.5926e-03, + 1.0920e-03, -7.2908e-04], + [ 1.2361e-05, 1.8740e-04, 1.3752e-03, ..., 1.4210e-03, + 2.3651e-03, 1.1206e-03]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0292, 0.0381, -0.0024, -0.0078, 0.0249, -0.0302, 0.0065, 0.0040, + -0.0039, -0.0185], device='cuda:0'), grad: tensor([ 0.0036, 0.0241, 0.0246, -0.0681, -0.0493, 0.0256, 0.0402, 0.0197, + -0.0419, 0.0216], device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 214.93, cls_loss 0.5334 cls_loss_mapping 0.0038 cls_loss_causal 0.4632 re_mapping 0.0096 re_causal 0.0225 /// teacc 98.83 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.0737, -0.1367, -0.0990, ..., -0.0428, 0.0585, -0.1095], + [-0.0715, -0.0932, -0.0775, ..., 0.1001, -0.0372, 0.2242], + [-0.0012, -0.0240, -0.0410, ..., -0.0080, -0.0075, -0.0803], + ..., + [ 0.0365, -0.1052, 0.1670, ..., 0.0368, -0.1027, 0.0543], + [-0.0410, 0.0611, -0.1495, ..., -0.0550, 0.0112, -0.0925], + [-0.0532, 0.0788, 0.0467, ..., -0.0562, -0.0456, -0.0280]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.4703e-08, 1.9930e-07, ..., 9.4995e-07, + 3.0428e-05, 3.7365e-03], + [ 0.0000e+00, 2.7940e-08, 1.8831e-06, ..., 1.0140e-05, + 3.1114e-05, 9.1362e-04], + [ 0.0000e+00, 5.0105e-07, 2.6263e-07, ..., 4.4890e-07, + 3.4511e-05, 3.4928e-04], + ..., + [ 0.0000e+00, 6.5938e-07, 6.8918e-07, ..., 2.7400e-06, + -3.0828e-04, -1.3481e-02], + [ 0.0000e+00, -1.0379e-05, 1.7621e-06, ..., 5.4203e-07, + 4.2230e-05, 1.0204e-03], + [ 0.0000e+00, -7.4506e-06, -8.9966e-07, ..., 1.2875e-05, + 3.3736e-05, 1.6575e-03]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0287, 0.0374, -0.0024, -0.0080, 0.0249, -0.0293, 0.0054, 0.0028, + -0.0038, -0.0167], device='cuda:0'), grad: tensor([ 0.0169, 0.0199, 0.0145, -0.0149, -0.0087, -0.0464, 0.0177, -0.0341, + 0.0177, 0.0174], device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 214.51, cls_loss 0.5423 cls_loss_mapping 0.0028 cls_loss_causal 0.4743 re_mapping 0.0093 re_causal 0.0220 /// teacc 98.87 lr 0.00010000 +Epoch 315, weight, value: tensor([[-7.3882e-02, -1.3584e-01, -9.9880e-02, ..., -4.2358e-02, + 5.8917e-02, -1.0793e-01], + [-7.2051e-02, -9.0594e-02, -7.8172e-02, ..., 1.0080e-01, + -3.4654e-02, 2.2368e-01], + [-1.0736e-04, -2.4511e-02, -4.0650e-02, ..., -7.5537e-03, + -7.2541e-03, -8.0188e-02], + ..., + [ 3.7215e-02, -1.0462e-01, 1.6823e-01, ..., 3.6740e-02, + -1.0393e-01, 5.4984e-02], + [-4.1482e-02, 6.0425e-02, -1.5096e-01, ..., -5.5579e-02, + 1.2466e-02, -9.3170e-02], + [-5.3299e-02, 7.9450e-02, 4.6068e-02, ..., -5.5345e-02, + -4.5439e-02, -2.7753e-02]], device='cuda:0'), grad: tensor([[ 4.9067e-04, 6.7282e-04, 1.6661e-03, ..., 7.2441e-03, + 1.0383e-04, 1.1988e-03], + [ 9.3222e-05, 5.0873e-05, 2.4204e-03, ..., 1.2770e-03, + 7.8753e-06, 8.1253e-04], + [-1.1045e-04, 3.1996e-04, 1.9464e-03, ..., 3.3588e-03, + 4.0710e-05, 1.3037e-03], + ..., + [-9.6703e-04, 5.3048e-05, 1.3990e-03, ..., 1.2150e-03, + 6.6496e-06, 1.0386e-03], + [ 4.1461e-04, 7.0381e-04, 1.5831e-03, ..., 3.8967e-03, + 1.0109e-04, 1.1749e-03], + [ 2.6727e-04, -8.7357e-04, -6.5651e-03, ..., -8.0261e-03, + 3.8326e-05, -1.7967e-03]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0286, 0.0387, -0.0020, -0.0075, 0.0240, -0.0289, 0.0041, 0.0025, + -0.0047, -0.0161], device='cuda:0'), grad: tensor([ 0.0427, -0.0065, 0.0210, -0.0019, 0.0167, -0.0047, -0.0084, -0.0083, + 0.0241, -0.0747], device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 215.10, cls_loss 0.5412 cls_loss_mapping 0.0038 cls_loss_causal 0.4709 re_mapping 0.0093 re_causal 0.0217 /// teacc 98.93 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.0739, -0.1364, -0.1015, ..., -0.0416, 0.0588, -0.1088], + [-0.0733, -0.0904, -0.0786, ..., 0.1007, -0.0344, 0.2232], + [ 0.0010, -0.0241, -0.0397, ..., -0.0081, -0.0073, -0.0802], + ..., + [ 0.0366, -0.1047, 0.1676, ..., 0.0358, -0.1046, 0.0550], + [-0.0398, 0.0611, -0.1505, ..., -0.0549, 0.0120, -0.0936], + [-0.0539, 0.0794, 0.0469, ..., -0.0555, -0.0451, -0.0278]], + device='cuda:0'), grad: tensor([[-9.0647e-04, 6.9104e-07, 2.9296e-05, ..., -3.0556e-03, + 1.0282e-05, 3.4213e-05], + [ 4.2105e-04, 3.0398e-05, 2.3353e-04, ..., -1.6565e-03, + -6.9857e-05, 8.8587e-06], + [ 3.2330e-04, 1.1355e-05, 2.3830e-04, ..., 9.3555e-04, + 2.1726e-05, -2.2769e-04], + ..., + [ 3.5119e-04, 8.3912e-07, -6.5613e-03, ..., 1.2274e-03, + 1.9222e-05, -1.2085e-05], + [ 3.6287e-04, 1.1623e-04, 1.4341e-04, ..., 1.2245e-03, + 1.6391e-05, 4.9084e-05], + [ 3.6454e-04, 3.4273e-06, 3.4409e-03, ..., 1.7118e-03, + 9.4950e-05, 3.3051e-05]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0284, 0.0388, -0.0023, -0.0083, 0.0244, -0.0289, 0.0043, 0.0020, + -0.0041, -0.0160], device='cuda:0'), grad: tensor([-0.1007, -0.0087, 0.0252, 0.0025, -0.0083, 0.0104, 0.0166, 0.0057, + 0.0208, 0.0366], device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 214.63, cls_loss 0.5262 cls_loss_mapping 0.0034 cls_loss_causal 0.4462 re_mapping 0.0095 re_causal 0.0224 /// teacc 98.88 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.0733, -0.1368, -0.1009, ..., -0.0415, 0.0582, -0.1093], + [-0.0738, -0.0899, -0.0778, ..., 0.1004, -0.0343, 0.2230], + [ 0.0010, -0.0238, -0.0394, ..., -0.0084, -0.0083, -0.0809], + ..., + [ 0.0361, -0.1052, 0.1677, ..., 0.0351, -0.1049, 0.0550], + [-0.0403, 0.0604, -0.1508, ..., -0.0545, 0.0129, -0.0942], + [-0.0528, 0.0799, 0.0466, ..., -0.0557, -0.0451, -0.0282]], + device='cuda:0'), grad: tensor([[-3.3989e-03, 1.7360e-06, 7.1764e-05, ..., -3.5305e-03, + -2.8286e-03, -1.9302e-03], + [ 2.1386e-04, 2.9746e-06, 1.5843e-04, ..., 3.3836e-03, + 1.2803e-04, 8.1599e-05], + [ 1.7285e-04, -3.0622e-06, -6.8760e-04, ..., -8.2684e-04, + 1.2207e-04, -5.9462e-04], + ..., + [ 2.5635e-03, 3.9376e-06, 9.3430e-06, ..., 1.0319e-03, + 7.4625e-05, 1.2445e-03], + [ 7.5817e-04, 1.1340e-05, 8.7142e-05, ..., -2.1839e-03, + 2.8086e-04, 4.1604e-04], + [-3.4084e-03, 6.3144e-07, 8.2076e-05, ..., 3.3331e-04, + 6.7472e-05, -1.3361e-03]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0284, 0.0390, -0.0030, -0.0070, 0.0239, -0.0292, 0.0046, 0.0021, + -0.0041, -0.0165], device='cuda:0'), grad: tensor([-0.0054, 0.0350, -0.0134, -0.0079, 0.0022, -0.0437, 0.0271, 0.0115, + -0.0065, 0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 214.66, cls_loss 0.5303 cls_loss_mapping 0.0038 cls_loss_causal 0.4633 re_mapping 0.0091 re_causal 0.0205 /// teacc 98.83 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.0722, -0.1372, -0.1011, ..., -0.0405, 0.0584, -0.1085], + [-0.0742, -0.0884, -0.0778, ..., 0.1002, -0.0327, 0.2218], + [ 0.0007, -0.0245, -0.0401, ..., -0.0086, -0.0088, -0.0810], + ..., + [ 0.0365, -0.1061, 0.1669, ..., 0.0344, -0.1060, 0.0556], + [-0.0401, 0.0606, -0.1506, ..., -0.0546, 0.0133, -0.0947], + [-0.0531, 0.0795, 0.0475, ..., -0.0553, -0.0457, -0.0281]], + device='cuda:0'), grad: tensor([[ 1.9208e-05, 8.5878e-04, 9.5844e-04, ..., 1.9646e-03, + 5.2500e-04, 1.8431e-06], + [ 2.7418e-05, 3.8648e-04, 6.1655e-04, ..., 2.5253e-03, + 2.6464e-04, 7.8440e-05], + [-3.7551e-05, 4.7851e-04, 4.5943e-04, ..., 1.0872e-03, + 1.3220e-04, 5.7101e-05], + ..., + [ 1.9278e-06, 7.4208e-05, -3.0175e-05, ..., -3.0575e-03, + 3.7730e-05, 2.5496e-05], + [ 3.9697e-05, -6.6519e-04, 9.2554e-04, ..., 1.3857e-03, + 5.1689e-04, 7.7784e-06], + [ 2.4904e-06, 3.0518e-04, -1.5621e-03, ..., 8.0681e-04, + 6.8724e-05, -1.5664e-04]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0276, 0.0388, -0.0042, -0.0067, 0.0240, -0.0299, 0.0048, 0.0032, + -0.0041, -0.0167], device='cuda:0'), grad: tensor([-0.0098, -0.0036, 0.0033, 0.0043, 0.0197, 0.0123, -0.0093, -0.0071, + -0.0181, 0.0082], device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 214.66, cls_loss 0.5291 cls_loss_mapping 0.0027 cls_loss_causal 0.4678 re_mapping 0.0094 re_causal 0.0219 /// teacc 98.85 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.0723, -0.1371, -0.0998, ..., -0.0421, 0.0581, -0.1091], + [-0.0739, -0.0888, -0.0787, ..., 0.1006, -0.0335, 0.2221], + [ 0.0007, -0.0255, -0.0402, ..., -0.0083, -0.0084, -0.0815], + ..., + [ 0.0362, -0.1067, 0.1670, ..., 0.0341, -0.1076, 0.0556], + [-0.0401, 0.0611, -0.1507, ..., -0.0555, 0.0140, -0.0952], + [-0.0530, 0.0782, 0.0478, ..., -0.0557, -0.0463, -0.0275]], + device='cuda:0'), grad: tensor([[-4.9591e-04, -2.6035e-04, 1.3925e-05, ..., 5.0850e-03, + -2.9430e-03, 1.2839e-04], + [ 1.9625e-05, 6.8605e-05, 1.4353e-04, ..., 1.6260e-03, + -7.4625e-04, -3.8099e-04], + [ 1.7536e-04, 4.6396e-04, 9.4533e-05, ..., 1.3218e-03, + 1.1206e-03, 4.0740e-05], + ..., + [ 2.9668e-05, 4.8310e-05, -2.3961e-04, ..., -9.2163e-03, + -7.8678e-04, 1.1402e-04], + [-2.2165e-07, -1.5554e-03, 8.2672e-05, ..., -4.6349e-03, + 5.2691e-04, -3.9563e-06], + [ 9.7334e-05, 1.7083e-04, 6.8903e-05, ..., 1.2789e-03, + 6.3848e-04, 6.7353e-05]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0279, 0.0382, -0.0043, -0.0054, 0.0244, -0.0293, 0.0048, 0.0031, + -0.0046, -0.0174], device='cuda:0'), grad: tensor([ 0.0305, -0.0092, 0.0214, 0.0222, 0.0034, 0.0074, -0.0105, -0.0749, + -0.0273, 0.0370], device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 214.67, cls_loss 0.5367 cls_loss_mapping 0.0027 cls_loss_causal 0.4680 re_mapping 0.0096 re_causal 0.0217 /// teacc 98.84 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.0713, -0.1383, -0.0995, ..., -0.0432, 0.0578, -0.1096], + [-0.0745, -0.0876, -0.0796, ..., 0.1013, -0.0346, 0.2227], + [ 0.0007, -0.0273, -0.0409, ..., -0.0063, -0.0082, -0.0817], + ..., + [ 0.0367, -0.1076, 0.1668, ..., 0.0333, -0.1094, 0.0558], + [-0.0404, 0.0613, -0.1510, ..., -0.0542, 0.0136, -0.0955], + [-0.0531, 0.0789, 0.0485, ..., -0.0568, -0.0460, -0.0286]], + device='cuda:0'), grad: tensor([[ 2.8417e-05, 1.1486e-04, -1.7405e-03, ..., -4.1246e-04, + -4.5276e-04, -1.7750e-04], + [-2.7731e-05, 6.9499e-05, 9.2030e-04, ..., 2.4261e-03, + 1.3742e-03, 4.7541e-04], + [ 1.9634e-04, 4.7255e-04, 1.1721e-03, ..., -7.4387e-03, + -1.5965e-03, 6.6662e-04], + ..., + [ 1.1422e-05, 3.8872e-03, -9.2077e-04, ..., 2.9163e-03, + 6.1703e-04, 4.8637e-03], + [-1.6356e-04, -4.2992e-03, 3.3736e-04, ..., 1.4305e-03, + 9.6416e-04, 4.1699e-04], + [ 1.3344e-05, 1.9968e-04, 5.7983e-04, ..., 1.0595e-03, + 8.3780e-04, 4.9305e-04]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0292, 0.0390, -0.0035, -0.0050, 0.0235, -0.0301, 0.0053, 0.0027, + -0.0035, -0.0177], device='cuda:0'), grad: tensor([ 0.0073, 0.0232, -0.0161, -0.0335, 0.0197, 0.0321, -0.0373, 0.0239, + -0.0027, -0.0165], device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 214.57, cls_loss 0.5736 cls_loss_mapping 0.0027 cls_loss_causal 0.5084 re_mapping 0.0095 re_causal 0.0220 /// teacc 98.79 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.0718, -0.1382, -0.0996, ..., -0.0443, 0.0583, -0.1090], + [-0.0763, -0.0871, -0.0801, ..., 0.1003, -0.0336, 0.2236], + [ 0.0017, -0.0276, -0.0397, ..., -0.0034, -0.0083, -0.0819], + ..., + [ 0.0376, -0.1077, 0.1669, ..., 0.0324, -0.1095, 0.0549], + [-0.0393, 0.0603, -0.1506, ..., -0.0539, 0.0130, -0.0959], + [-0.0535, 0.0791, 0.0482, ..., -0.0577, -0.0473, -0.0275]], + device='cuda:0'), grad: tensor([[ 1.6868e-05, 2.3556e-04, 2.1470e-04, ..., 4.4250e-04, + -6.6109e-03, 1.5104e-04], + [ 1.4520e-04, 4.3344e-04, 1.8942e-04, ..., 1.3697e-04, + 3.7217e-04, -7.5340e-04], + [ 1.1313e-04, 1.0805e-03, 4.4870e-04, ..., 3.8261e-03, + 5.6534e-03, 1.8322e-04], + ..., + [ 2.3067e-04, 5.5170e-04, -7.5769e-04, ..., -2.4915e-04, + 1.9932e-04, 5.0545e-04], + [ 5.6839e-04, 1.5621e-03, 8.0919e-04, ..., 1.9703e-03, + 2.8682e-04, 3.1734e-04], + [-8.4686e-04, -2.3327e-03, 5.5170e-04, ..., -4.5991e-04, + 2.5702e-04, 1.2994e-04]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0298, 0.0389, -0.0019, -0.0056, 0.0233, -0.0297, 0.0053, 0.0031, + -0.0041, -0.0182], device='cuda:0'), grad: tensor([-0.0119, 0.0091, 0.0319, -0.0358, -0.0158, 0.0162, 0.0174, -0.0199, + 0.0150, -0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 214.25, cls_loss 0.5546 cls_loss_mapping 0.0037 cls_loss_causal 0.4972 re_mapping 0.0096 re_causal 0.0219 /// teacc 98.84 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.0714, -0.1397, -0.1003, ..., -0.0417, 0.0588, -0.1092], + [-0.0760, -0.0872, -0.0806, ..., 0.1001, -0.0341, 0.2249], + [ 0.0010, -0.0279, -0.0400, ..., -0.0041, -0.0068, -0.0835], + ..., + [ 0.0384, -0.1081, 0.1680, ..., 0.0325, -0.1083, 0.0546], + [-0.0384, 0.0604, -0.1501, ..., -0.0546, 0.0121, -0.0967], + [-0.0550, 0.0797, 0.0478, ..., -0.0579, -0.0471, -0.0275]], + device='cuda:0'), grad: tensor([[ 3.3714e-06, 9.2164e-06, 1.4983e-05, ..., 6.6280e-04, + 2.4214e-08, 7.9107e-04], + [ 1.8077e-06, 4.5039e-06, 1.0543e-05, ..., -1.3794e-02, + 5.5879e-09, -6.9084e-03], + [ 5.6207e-05, 6.2764e-05, 1.2493e-04, ..., 1.3380e-03, + 4.6566e-08, 1.1005e-03], + ..., + [ 3.8128e-06, 6.5006e-06, -3.6087e-03, ..., 9.4461e-04, + 0.0000e+00, -1.5945e-02], + [-2.4331e-04, -4.6968e-04, -1.8191e-04, ..., 3.6831e-03, + 1.7881e-07, 1.3142e-03], + [ 5.3942e-06, 2.0593e-05, 3.5954e-03, ..., -8.0967e-04, + 7.4506e-09, 1.7593e-02]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0282, 0.0387, -0.0018, -0.0055, 0.0233, -0.0306, 0.0049, 0.0022, + -0.0045, -0.0171], device='cuda:0'), grad: tensor([-0.0105, -0.0434, 0.0177, 0.0166, 0.0162, -0.0153, 0.0279, -0.0116, + 0.0242, -0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 214.87, cls_loss 0.5378 cls_loss_mapping 0.0035 cls_loss_causal 0.4724 re_mapping 0.0092 re_causal 0.0216 /// teacc 98.86 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.0703, -0.1402, -0.1006, ..., -0.0414, 0.0593, -0.1097], + [-0.0756, -0.0885, -0.0819, ..., 0.0999, -0.0364, 0.2254], + [ 0.0005, -0.0271, -0.0404, ..., -0.0046, -0.0072, -0.0833], + ..., + [ 0.0385, -0.1073, 0.1689, ..., 0.0331, -0.1071, 0.0552], + [-0.0385, 0.0603, -0.1504, ..., -0.0547, 0.0126, -0.0972], + [-0.0558, 0.0791, 0.0478, ..., -0.0577, -0.0477, -0.0280]], + device='cuda:0'), grad: tensor([[ 7.8344e-04, -3.9172e-04, 5.2643e-04, ..., 7.9727e-04, + 2.1191e-03, 5.0277e-05], + [ 6.1131e-04, 5.4855e-07, 5.1403e-04, ..., 2.1378e-02, + 2.0790e-04, 1.0341e-04], + [-5.2986e-03, 2.1711e-05, 9.5224e-04, ..., -2.3315e-02, + -3.9276e-02, 3.0756e-04], + ..., + [ 3.2139e-04, -1.7047e-05, 5.2071e-04, ..., 2.0003e-04, + 4.0936e-04, -1.0529e-03], + [ 1.0328e-03, 4.2647e-05, -7.9117e-03, ..., -4.0932e-03, + 2.1973e-02, 8.1182e-05], + [ 8.2970e-04, 3.5024e-04, 1.6365e-03, ..., 1.1187e-03, + 1.5011e-03, 1.1320e-03]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0277, 0.0386, -0.0022, -0.0064, 0.0230, -0.0307, 0.0057, 0.0023, + -0.0034, -0.0177], device='cuda:0'), grad: tensor([-0.0076, 0.0458, -0.0632, -0.0058, -0.0288, 0.0220, 0.0331, 0.0235, + -0.0501, 0.0313], device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 214.48, cls_loss 0.5654 cls_loss_mapping 0.0034 cls_loss_causal 0.4951 re_mapping 0.0090 re_causal 0.0213 /// teacc 98.83 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.0723, -0.1364, -0.0995, ..., -0.0400, 0.0589, -0.1095], + [-0.0768, -0.0869, -0.0818, ..., 0.1000, -0.0352, 0.2258], + [ 0.0020, -0.0261, -0.0403, ..., -0.0051, -0.0065, -0.0837], + ..., + [ 0.0395, -0.1086, 0.1690, ..., 0.0337, -0.1071, 0.0554], + [-0.0371, 0.0593, -0.1509, ..., -0.0551, 0.0130, -0.0980], + [-0.0576, 0.0786, 0.0482, ..., -0.0584, -0.0482, -0.0276]], + device='cuda:0'), grad: tensor([[-3.5858e-03, 2.6112e-03, 1.1466e-05, ..., 2.6836e-03, + 1.5764e-03, 7.2084e-07], + [ 3.3474e-04, 4.6778e-04, 1.4789e-05, ..., 7.7295e-04, + 1.5221e-03, -2.4531e-06], + [-2.2163e-03, -1.6403e-03, -1.0052e-03, ..., 3.6831e-03, + 1.5821e-03, 4.8801e-06], + ..., + [ 1.4267e-03, 7.7772e-04, 8.7142e-05, ..., 2.1572e-03, + 6.4230e-04, -4.3797e-04], + [ 7.2622e-04, 2.8753e-04, 1.4842e-04, ..., 6.6805e-04, + 3.7789e-04, 8.1677e-07], + [ 4.8161e-04, 1.1122e-04, 7.8380e-05, ..., 1.9729e-04, + 9.0456e-04, 1.4961e-04]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0273, 0.0385, -0.0022, -0.0066, 0.0227, -0.0302, 0.0056, 0.0028, + -0.0039, -0.0178], device='cuda:0'), grad: tensor([-0.0067, 0.0095, 0.0565, -0.0278, -0.0186, -0.0215, -0.0148, 0.0108, + 0.0074, 0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 214.60, cls_loss 0.5348 cls_loss_mapping 0.0025 cls_loss_causal 0.4616 re_mapping 0.0095 re_causal 0.0211 /// teacc 98.80 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.0722, -0.1371, -0.0993, ..., -0.0414, 0.0594, -0.1091], + [-0.0792, -0.0878, -0.0823, ..., 0.1006, -0.0357, 0.2263], + [ 0.0018, -0.0257, -0.0403, ..., -0.0052, -0.0079, -0.0831], + ..., + [ 0.0396, -0.1080, 0.1696, ..., 0.0335, -0.1081, 0.0553], + [-0.0368, 0.0602, -0.1515, ..., -0.0556, 0.0130, -0.0978], + [-0.0557, 0.0782, 0.0479, ..., -0.0573, -0.0470, -0.0277]], + device='cuda:0'), grad: tensor([[ 1.6892e-04, 9.1970e-05, 1.6868e-04, ..., 3.9792e-04, + -2.4509e-04, 3.0234e-05], + [ 2.9063e-04, 2.1303e-04, 4.0894e-03, ..., 5.9843e-04, + 3.2902e-05, 3.3021e-05], + [ 1.1406e-03, 1.1978e-03, 2.4757e-03, ..., 2.1839e-03, + 5.1826e-05, 2.8682e-04], + ..., + [ 3.6287e-04, 5.1641e-04, -5.6190e-03, ..., 1.0471e-03, + 2.7716e-05, 7.0453e-05], + [ 9.5177e-04, 6.3419e-04, 2.8210e-03, ..., 1.2579e-03, + 4.3243e-05, 6.7830e-05], + [ 1.4663e-04, -1.8701e-05, 1.6117e-04, ..., 2.3699e-04, + 2.4617e-05, -5.2899e-05]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0277, 0.0386, -0.0013, -0.0064, 0.0222, -0.0304, 0.0056, 0.0014, + -0.0045, -0.0161], device='cuda:0'), grad: tensor([-0.0134, -0.0049, 0.0234, -0.0177, 0.0126, 0.0030, 0.0245, -0.0281, + 0.0190, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 214.48, cls_loss 0.5519 cls_loss_mapping 0.0027 cls_loss_causal 0.4844 re_mapping 0.0091 re_causal 0.0213 /// teacc 98.68 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.0696, -0.1379, -0.1001, ..., -0.0410, 0.0599, -0.1086], + [-0.0797, -0.0877, -0.0827, ..., 0.0999, -0.0340, 0.2267], + [ 0.0013, -0.0263, -0.0424, ..., -0.0064, -0.0083, -0.0838], + ..., + [ 0.0376, -0.1081, 0.1709, ..., 0.0341, -0.1085, 0.0548], + [-0.0373, 0.0593, -0.1513, ..., -0.0559, 0.0124, -0.0990], + [-0.0547, 0.0777, 0.0468, ..., -0.0570, -0.0467, -0.0289]], + device='cuda:0'), grad: tensor([[ 6.0424e-06, 1.7777e-05, 6.2525e-05, ..., 5.5027e-04, + 3.3522e-04, 1.2195e-04], + [ 4.9546e-06, 2.7344e-05, 4.9174e-05, ..., -1.6756e-03, + -6.1417e-04, -1.5378e-04], + [ 1.2219e-04, 2.2495e-04, 2.9230e-04, ..., -1.9765e-04, + -2.2354e-03, 2.2018e-04], + ..., + [ 3.9172e-04, 5.3215e-04, 2.6684e-03, ..., 1.6584e-03, + 4.7445e-04, 3.5095e-04], + [-1.3609e-03, -1.6870e-03, -2.1591e-03, ..., 3.5596e-04, + 2.9540e-04, -3.9887e-04], + [ 7.2539e-05, 1.0860e-04, -1.1120e-03, ..., -3.8643e-03, + 3.1924e-04, 9.4748e-04]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0269, 0.0390, -0.0023, -0.0071, 0.0211, -0.0288, 0.0054, 0.0025, + -0.0054, -0.0159], device='cuda:0'), grad: tensor([ 0.0275, -0.0425, -0.0341, -0.0009, 0.0162, 0.0037, 0.0127, 0.0449, + 0.0141, -0.0416], device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 214.77, cls_loss 0.5389 cls_loss_mapping 0.0029 cls_loss_causal 0.4814 re_mapping 0.0093 re_causal 0.0226 /// teacc 98.79 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.0689, -0.1384, -0.1004, ..., -0.0395, 0.0590, -0.1086], + [-0.0794, -0.0876, -0.0819, ..., 0.1009, -0.0345, 0.2271], + [ 0.0009, -0.0269, -0.0422, ..., -0.0064, -0.0079, -0.0823], + ..., + [ 0.0380, -0.1092, 0.1698, ..., 0.0329, -0.1093, 0.0552], + [-0.0363, 0.0590, -0.1512, ..., -0.0548, 0.0118, -0.0986], + [-0.0543, 0.0782, 0.0466, ..., -0.0570, -0.0469, -0.0303]], + device='cuda:0'), grad: tensor([[ 3.0065e-04, 1.6451e-04, 3.4571e-04, ..., -7.5722e-04, + 4.4912e-05, 8.2779e-04], + [ 3.1109e-03, 2.6774e-04, -3.5954e-04, ..., 1.4000e-02, + 2.3648e-05, 8.6927e-04], + [ 3.5357e-04, 6.6566e-04, 2.0905e-03, ..., 1.1797e-03, + 1.2529e-04, 2.5272e-03], + ..., + [ 1.4229e-03, 1.5755e-03, -3.9876e-05, ..., -3.2735e-04, + 7.6675e-04, -2.9540e-04], + [-6.2294e-03, -7.8917e-04, 6.4611e-04, ..., -1.1406e-02, + 9.8825e-05, 9.8896e-04], + [ 8.0633e-04, -2.3441e-03, -1.2341e-03, ..., 2.3155e-03, + 1.9681e-04, -2.6330e-05]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0272, 0.0406, -0.0013, -0.0075, 0.0222, -0.0299, 0.0045, 0.0015, + -0.0055, -0.0158], device='cuda:0'), grad: tensor([-0.0079, 0.0475, 0.0215, 0.0343, -0.0223, 0.0051, -0.0388, 0.0027, + -0.0565, 0.0144], device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 214.46, cls_loss 0.5098 cls_loss_mapping 0.0027 cls_loss_causal 0.4442 re_mapping 0.0095 re_causal 0.0218 /// teacc 98.78 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.0682, -0.1390, -0.1004, ..., -0.0394, 0.0596, -0.1081], + [-0.0799, -0.0878, -0.0821, ..., 0.1007, -0.0343, 0.2265], + [ 0.0013, -0.0263, -0.0423, ..., -0.0069, -0.0081, -0.0829], + ..., + [ 0.0373, -0.1099, 0.1699, ..., 0.0322, -0.1093, 0.0557], + [-0.0359, 0.0583, -0.1512, ..., -0.0535, 0.0107, -0.0988], + [-0.0539, 0.0794, 0.0468, ..., -0.0560, -0.0462, -0.0310]], + device='cuda:0'), grad: tensor([[ 3.2163e-04, 2.3283e-08, 3.1088e-06, ..., 9.6464e-04, + 8.6498e-04, 1.0389e-04], + [ 9.5034e-04, 3.8184e-08, 1.5207e-05, ..., 8.1730e-04, + 7.3481e-04, 2.8920e-04], + [ 3.2377e-04, 8.6334e-07, 7.0095e-05, ..., 7.6628e-04, + 7.6103e-04, 1.3947e-04], + ..., + [-1.0101e-02, 8.3819e-09, -2.4605e-04, ..., 4.7302e-04, + 5.8937e-04, -3.2959e-03], + [ 9.2506e-05, 7.7300e-08, 8.0541e-06, ..., -3.0098e-03, + -1.9531e-03, 3.4988e-05], + [ 3.9940e-03, 4.7497e-08, 1.4178e-05, ..., 1.0719e-03, + 1.2865e-03, 1.2484e-03]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0276, 0.0408, -0.0013, -0.0070, 0.0216, -0.0301, 0.0043, 0.0018, + -0.0054, -0.0159], device='cuda:0'), grad: tensor([ 0.0230, -0.0054, 0.0227, -0.0416, -0.0211, -0.0041, 0.0276, 0.0049, + -0.0438, 0.0379], device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 214.24, cls_loss 0.5267 cls_loss_mapping 0.0026 cls_loss_causal 0.4683 re_mapping 0.0091 re_causal 0.0213 /// teacc 98.81 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.0677, -0.1397, -0.0997, ..., -0.0389, 0.0584, -0.1093], + [-0.0798, -0.0883, -0.0837, ..., 0.1006, -0.0346, 0.2261], + [ 0.0016, -0.0264, -0.0434, ..., -0.0068, -0.0075, -0.0838], + ..., + [ 0.0369, -0.1101, 0.1692, ..., 0.0324, -0.1105, 0.0565], + [-0.0353, 0.0582, -0.1506, ..., -0.0531, 0.0098, -0.0993], + [-0.0534, 0.0797, 0.0482, ..., -0.0547, -0.0474, -0.0295]], + device='cuda:0'), grad: tensor([[ 1.3387e-04, 4.8113e-04, 1.1049e-05, ..., 8.7643e-04, + 7.4327e-05, 2.0862e-04], + [-1.5221e-03, -2.3842e-03, -9.8610e-04, ..., -8.1711e-03, + -6.4373e-04, 9.0897e-05], + [ 2.0027e-03, 3.1700e-03, 9.2840e-04, ..., 9.4681e-03, + 8.2159e-04, 2.5177e-04], + ..., + [ 1.6201e-04, -2.2755e-03, -8.4281e-05, ..., -4.4861e-03, + -1.1432e-04, -3.2063e-03], + [ 3.9077e-04, 3.7432e-04, 8.5216e-07, ..., 1.3885e-03, + 1.3828e-04, 4.1604e-04], + [-4.9591e-03, 4.6206e-04, 7.1585e-05, ..., -3.5076e-03, + -1.1263e-03, 7.4768e-04]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0278, 0.0412, -0.0023, -0.0067, 0.0214, -0.0296, 0.0045, 0.0017, + -0.0053, -0.0159], device='cuda:0'), grad: tensor([ 0.0084, -0.0204, 0.0372, 0.0158, 0.0213, -0.0232, 0.0086, -0.0512, + 0.0102, -0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 214.46, cls_loss 0.5344 cls_loss_mapping 0.0032 cls_loss_causal 0.4654 re_mapping 0.0091 re_causal 0.0218 /// teacc 98.81 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.0691, -0.1397, -0.1000, ..., -0.0411, 0.0587, -0.1099], + [-0.0788, -0.0884, -0.0842, ..., 0.1011, -0.0346, 0.2268], + [ 0.0029, -0.0260, -0.0447, ..., -0.0055, -0.0074, -0.0843], + ..., + [ 0.0379, -0.1101, 0.1705, ..., 0.0342, -0.1112, 0.0576], + [-0.0351, 0.0596, -0.1509, ..., -0.0532, 0.0105, -0.1011], + [-0.0538, 0.0798, 0.0484, ..., -0.0550, -0.0483, -0.0309]], + device='cuda:0'), grad: tensor([[ 4.1932e-05, 2.9016e-04, 6.2132e-04, ..., 2.5005e-03, + 2.0504e-04, 3.2157e-05], + [ 2.6971e-05, 1.3876e-04, 5.4646e-04, ..., 1.4496e-03, + 1.1766e-04, -6.1607e-04], + [-1.0319e-03, -3.9887e-04, -2.7442e-04, ..., -3.9935e-04, + 1.2106e-04, -5.2023e-04], + ..., + [ 1.9646e-04, 2.3103e-04, -1.2695e-02, ..., 2.0161e-03, + 1.8430e-04, -6.5498e-03], + [ 5.9366e-04, -2.9087e-03, -3.6297e-03, ..., -1.2253e-02, + -1.6212e-03, 4.8447e-04], + [ 3.6925e-05, 3.9601e-04, 1.3367e-02, ..., 4.1428e-03, + 2.1935e-04, 6.7558e-03]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0291, 0.0411, -0.0020, -0.0071, 0.0217, -0.0292, 0.0043, 0.0032, + -0.0051, -0.0164], device='cuda:0'), grad: tensor([ 0.0229, 0.0243, 0.0140, -0.0035, -0.0084, 0.0229, -0.0011, -0.0235, + -0.0911, 0.0436], device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 214.44, cls_loss 0.5505 cls_loss_mapping 0.0034 cls_loss_causal 0.4836 re_mapping 0.0088 re_causal 0.0208 /// teacc 98.57 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.0699, -0.1401, -0.1006, ..., -0.0417, 0.0586, -0.1102], + [-0.0790, -0.0886, -0.0841, ..., 0.1016, -0.0346, 0.2278], + [ 0.0022, -0.0261, -0.0450, ..., -0.0044, -0.0077, -0.0832], + ..., + [ 0.0377, -0.1109, 0.1709, ..., 0.0345, -0.1122, 0.0564], + [-0.0349, 0.0600, -0.1514, ..., -0.0541, 0.0116, -0.1013], + [-0.0534, 0.0796, 0.0473, ..., -0.0559, -0.0488, -0.0305]], + device='cuda:0'), grad: tensor([[ 7.3481e-07, 4.0412e-04, 4.2248e-04, ..., 1.3227e-03, + 8.6367e-05, 1.7548e-04], + [ 1.0645e-06, 6.0368e-04, 8.0252e-04, ..., -6.9666e-04, + 1.8394e-04, -1.3809e-02], + [-1.5467e-05, -1.8473e-03, -2.4414e-03, ..., 2.6932e-03, + 1.8620e-04, 2.7061e-04], + ..., + [ 7.5363e-06, 1.5926e-03, 1.7891e-03, ..., 3.0117e-03, + 2.1780e-04, 4.6873e-04], + [ 3.8333e-06, 1.4868e-03, 1.5049e-03, ..., 3.0956e-03, + 2.3162e-04, 6.9284e-04], + [ 4.0047e-08, -1.2253e-02, -1.7233e-03, ..., -5.0316e-03, + -1.4410e-03, -1.3762e-03]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0284, 0.0410, -0.0023, -0.0081, 0.0218, -0.0292, 0.0044, 0.0034, + -0.0049, -0.0164], device='cuda:0'), grad: tensor([ 0.0084, -0.0080, -0.0023, 0.0145, 0.0205, -0.0456, -0.0004, 0.0190, + 0.0185, -0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 214.92, cls_loss 0.5206 cls_loss_mapping 0.0031 cls_loss_causal 0.4554 re_mapping 0.0092 re_causal 0.0215 /// teacc 98.84 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.0702, -0.1402, -0.1002, ..., -0.0422, 0.0599, -0.1098], + [-0.0781, -0.0890, -0.0845, ..., 0.1021, -0.0350, 0.2289], + [ 0.0029, -0.0243, -0.0433, ..., -0.0037, -0.0076, -0.0832], + ..., + [ 0.0392, -0.1117, 0.1714, ..., 0.0365, -0.1122, 0.0561], + [-0.0356, 0.0599, -0.1530, ..., -0.0553, 0.0112, -0.1010], + [-0.0550, 0.0819, 0.0471, ..., -0.0592, -0.0498, -0.0306]], + device='cuda:0'), grad: tensor([[-4.3793e-03, 7.5960e-04, -2.7771e-03, ..., -1.1215e-02, + 0.0000e+00, 1.8442e-04], + [ 1.4842e-04, 5.2738e-04, 1.2010e-04, ..., 2.5749e-03, + 0.0000e+00, 7.4267e-05], + [ 1.1625e-03, 1.6546e-03, 8.1253e-04, ..., 2.2202e-03, + 0.0000e+00, -1.0786e-03], + ..., + [ 6.2037e-04, -7.3767e-04, 1.9388e-03, ..., 1.1358e-03, + 0.0000e+00, -1.5318e-04], + [ 3.3092e-04, 3.7217e-04, 5.5981e-04, ..., 2.1839e-03, + 0.0000e+00, 2.3520e-04], + [ 3.4022e-04, 5.2595e-04, 1.1482e-03, ..., 1.8950e-03, + 0.0000e+00, 3.5930e-04]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0276, 0.0412, -0.0012, -0.0086, 0.0221, -0.0291, 0.0046, 0.0033, + -0.0063, -0.0169], device='cuda:0'), grad: tensor([-0.0450, 0.0156, 0.0101, -0.0200, 0.0176, 0.0179, -0.0099, -0.0138, + 0.0130, 0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 214.78, cls_loss 0.5624 cls_loss_mapping 0.0033 cls_loss_causal 0.4938 re_mapping 0.0094 re_causal 0.0225 /// teacc 98.82 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.0707, -0.1414, -0.1013, ..., -0.0435, 0.0606, -0.1093], + [-0.0780, -0.0880, -0.0855, ..., 0.1014, -0.0352, 0.2295], + [ 0.0037, -0.0252, -0.0444, ..., -0.0028, -0.0082, -0.0815], + ..., + [ 0.0380, -0.1118, 0.1720, ..., 0.0365, -0.1122, 0.0558], + [-0.0373, 0.0588, -0.1534, ..., -0.0552, 0.0120, -0.1038], + [-0.0530, 0.0831, 0.0478, ..., -0.0594, -0.0493, -0.0308]], + device='cuda:0'), grad: tensor([[ 5.5600e-07, 1.0777e-03, 3.9387e-04, ..., 4.0932e-03, + 1.2779e-03, 1.9894e-03], + [ 7.8380e-06, 5.9557e-04, 1.9658e-04, ..., 1.9474e-03, + 2.1553e-04, 8.1587e-04], + [-5.7191e-05, 8.1253e-04, 2.7776e-04, ..., 2.9583e-03, + -3.1173e-05, 2.8954e-03], + ..., + [ 3.9458e-05, -2.7370e-03, -1.1587e-03, ..., -7.3509e-03, + -2.0079e-06, -8.9874e-03], + [ 3.7011e-06, -1.4229e-03, 1.0508e-04, ..., -6.7101e-03, + 1.3280e-04, 8.1778e-04], + [ 4.0326e-07, 9.7942e-04, 7.1573e-04, ..., 2.5215e-03, + 2.2304e-04, 1.5297e-03]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0286, 0.0413, -0.0014, -0.0089, 0.0218, -0.0291, 0.0049, 0.0031, + -0.0057, -0.0162], device='cuda:0'), grad: tensor([ 0.0215, 0.0115, -0.0117, 0.0040, 0.0129, 0.0329, -0.0030, -0.0371, + -0.0472, 0.0162], device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 214.96, cls_loss 0.5500 cls_loss_mapping 0.0031 cls_loss_causal 0.4859 re_mapping 0.0094 re_causal 0.0224 /// teacc 98.80 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.0716, -0.1432, -0.1001, ..., -0.0421, 0.0615, -0.1093], + [-0.0790, -0.0881, -0.0849, ..., 0.1019, -0.0346, 0.2303], + [ 0.0039, -0.0241, -0.0451, ..., -0.0006, -0.0084, -0.0800], + ..., + [ 0.0394, -0.1130, 0.1711, ..., 0.0355, -0.1139, 0.0555], + [-0.0379, 0.0586, -0.1543, ..., -0.0554, 0.0116, -0.1052], + [-0.0520, 0.0845, 0.0491, ..., -0.0597, -0.0499, -0.0303]], + device='cuda:0'), grad: tensor([[ 3.4153e-05, 3.4122e-03, 1.0121e-04, ..., 5.6696e-04, + 8.4639e-05, 4.6768e-03], + [ 3.8910e-03, 6.5613e-04, -1.3387e-04, ..., -5.7793e-04, + 1.5251e-05, 3.7327e-03], + [ 1.2326e-04, 1.8196e-03, 4.3344e-04, ..., 2.3251e-03, + 3.2687e-04, 2.0466e-03], + ..., + [ 2.7823e-04, 1.7481e-03, 7.2539e-05, ..., 1.3742e-03, + 3.2306e-05, 2.8362e-03], + [ 1.0258e-04, -5.6419e-03, 1.0449e-04, ..., -1.2598e-03, + 7.6950e-05, -8.0566e-03], + [ 5.7697e-04, -2.2278e-03, 2.2743e-06, ..., -1.1748e-04, + 1.3769e-04, -1.7738e-03]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0274, 0.0418, -0.0007, -0.0090, 0.0221, -0.0293, 0.0041, 0.0032, + -0.0067, -0.0165], device='cuda:0'), grad: tensor([ 0.0305, -0.0086, 0.0124, 0.0070, -0.0076, 0.0071, -0.0199, 0.0193, + -0.0200, -0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 214.46, cls_loss 0.5674 cls_loss_mapping 0.0028 cls_loss_causal 0.5045 re_mapping 0.0094 re_causal 0.0232 /// teacc 98.71 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.0718, -0.1441, -0.0992, ..., -0.0410, 0.0613, -0.1110], + [-0.0791, -0.0884, -0.0846, ..., 0.1018, -0.0350, 0.2302], + [ 0.0032, -0.0260, -0.0462, ..., -0.0025, -0.0087, -0.0817], + ..., + [ 0.0391, -0.1133, 0.1710, ..., 0.0362, -0.1161, 0.0558], + [-0.0374, 0.0602, -0.1536, ..., -0.0552, 0.0130, -0.1056], + [-0.0518, 0.0845, 0.0481, ..., -0.0608, -0.0506, -0.0286]], + device='cuda:0'), grad: tensor([[ 2.1648e-04, 1.8322e-04, 9.6977e-05, ..., 1.1015e-03, + 5.0259e-04, 3.8218e-04], + [ 3.4714e-04, 6.0463e-04, 1.0121e-04, ..., 2.5463e-03, + 1.1358e-03, 6.0606e-04], + [ 3.5858e-04, 3.7313e-04, 2.2852e-04, ..., -8.6823e-03, + 1.5383e-03, 6.4373e-04], + ..., + [-2.9049e-03, 5.9414e-04, -6.8169e-03, ..., -9.1553e-03, + 1.1444e-03, -3.7098e-03], + [-1.1854e-03, -4.5776e-03, 1.9159e-03, ..., 8.2321e-03, + 1.1854e-03, -5.3167e-04], + [ 1.5650e-03, -6.3705e-04, 2.5978e-03, ..., 2.4643e-03, + -1.1299e-02, 1.4515e-03]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0275, 0.0417, -0.0024, -0.0077, 0.0223, -0.0290, 0.0041, 0.0030, + -0.0059, -0.0171], device='cuda:0'), grad: tensor([ 0.0100, 0.0166, 0.0082, 0.0229, -0.0256, 0.0004, -0.0060, -0.0370, + 0.0155, -0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 214.91, cls_loss 0.5139 cls_loss_mapping 0.0028 cls_loss_causal 0.4494 re_mapping 0.0097 re_causal 0.0226 /// teacc 98.81 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.0728, -0.1446, -0.0996, ..., -0.0408, 0.0614, -0.1121], + [-0.0798, -0.0868, -0.0844, ..., 0.1026, -0.0337, 0.2311], + [ 0.0033, -0.0272, -0.0469, ..., -0.0043, -0.0089, -0.0821], + ..., + [ 0.0402, -0.1147, 0.1723, ..., 0.0350, -0.1142, 0.0558], + [-0.0369, 0.0598, -0.1537, ..., -0.0544, 0.0118, -0.1061], + [-0.0517, 0.0840, 0.0479, ..., -0.0597, -0.0500, -0.0273]], + device='cuda:0'), grad: tensor([[ 1.5326e-03, -1.2177e-04, 2.5773e-04, ..., -9.2850e-03, + -1.5068e-03, -7.7896e-03], + [ 1.0185e-03, 2.1970e-04, 7.5531e-04, ..., 2.6855e-03, + -5.1117e-03, 1.2617e-03], + [ 7.0715e-04, 9.1374e-05, 1.8895e-04, ..., 2.2049e-03, + 2.6941e-04, 3.2854e-04], + ..., + [ 5.0402e-04, -3.8576e-04, -1.3733e-03, ..., 2.7542e-03, + 2.6512e-04, -1.4000e-03], + [ 8.9788e-04, -7.1096e-04, -2.2564e-03, ..., -1.4496e-02, + 4.6754e-04, -8.1682e-04], + [ 6.9809e-04, 4.8208e-04, 1.5593e-03, ..., 3.0937e-03, + 5.6791e-04, 2.0504e-03]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0273, 0.0423, -0.0038, -0.0085, 0.0223, -0.0292, 0.0047, 0.0023, + -0.0050, -0.0166], device='cuda:0'), grad: tensor([-0.0011, -0.0078, 0.0145, -0.0145, 0.0198, -0.0179, 0.0406, -0.0142, + -0.0435, 0.0241], device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 214.91, cls_loss 0.5407 cls_loss_mapping 0.0037 cls_loss_causal 0.4757 re_mapping 0.0093 re_causal 0.0213 /// teacc 98.73 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.0735, -0.1435, -0.0995, ..., -0.0393, 0.0617, -0.1120], + [-0.0798, -0.0873, -0.0836, ..., 0.1023, -0.0342, 0.2305], + [ 0.0029, -0.0272, -0.0458, ..., -0.0041, -0.0094, -0.0816], + ..., + [ 0.0397, -0.1144, 0.1711, ..., 0.0343, -0.1162, 0.0561], + [-0.0366, 0.0597, -0.1544, ..., -0.0543, 0.0118, -0.1061], + [-0.0522, 0.0826, 0.0475, ..., -0.0603, -0.0488, -0.0281]], + device='cuda:0'), grad: tensor([[ 1.9205e-04, 1.0529e-03, 2.8133e-04, ..., 1.8406e-03, + 0.0000e+00, 1.0509e-03], + [-2.1038e-03, -2.9774e-03, -1.0252e-03, ..., -4.7379e-03, + 0.0000e+00, -7.4120e-03], + [ 1.6406e-05, 8.4400e-04, -1.5402e-04, ..., 6.4230e-04, + 0.0000e+00, 8.8310e-04], + ..., + [ 2.5988e-04, 2.6631e-04, -2.4235e-04, ..., 1.4772e-03, + 0.0000e+00, 3.3808e-04], + [ 7.1573e-04, 6.7329e-04, 4.5276e-04, ..., 1.9693e-04, + 0.0000e+00, 1.7481e-03], + [ 5.0211e-04, -4.1656e-03, -7.4625e-04, ..., -7.4291e-04, + 0.0000e+00, 1.1559e-03]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0266, 0.0420, -0.0031, -0.0079, 0.0229, -0.0292, 0.0045, 0.0014, + -0.0050, -0.0175], device='cuda:0'), grad: tensor([ 0.0271, -0.0557, 0.0183, 0.0192, 0.0188, 0.0231, -0.0425, 0.0221, + -0.0057, -0.0247], device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 215.02, cls_loss 0.5653 cls_loss_mapping 0.0032 cls_loss_causal 0.4932 re_mapping 0.0093 re_causal 0.0217 /// teacc 98.83 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.0737, -0.1444, -0.0995, ..., -0.0374, 0.0610, -0.1110], + [-0.0804, -0.0888, -0.0838, ..., 0.1021, -0.0340, 0.2310], + [ 0.0022, -0.0277, -0.0466, ..., -0.0038, -0.0099, -0.0814], + ..., + [ 0.0397, -0.1138, 0.1706, ..., 0.0340, -0.1173, 0.0562], + [-0.0354, 0.0589, -0.1547, ..., -0.0560, 0.0130, -0.1064], + [-0.0534, 0.0823, 0.0488, ..., -0.0600, -0.0483, -0.0291]], + device='cuda:0'), grad: tensor([[ 6.9384e-07, 5.8556e-04, 5.1141e-05, ..., -2.9564e-03, + -2.8934e-03, 4.6566e-06], + [ 4.2655e-06, 4.2510e-04, 7.0095e-05, ..., 8.1396e-04, + 3.5405e-05, 1.6916e-04], + [ 4.1351e-06, 2.9221e-03, 4.2272e-04, ..., 2.3460e-03, + 4.4048e-05, 3.2753e-05], + ..., + [-5.2899e-05, 2.5225e-04, -2.2972e-04, ..., 6.6566e-04, + 1.2018e-05, -6.1893e-04], + [ 4.2170e-06, 6.7902e-04, 5.3704e-05, ..., 9.6512e-04, + 4.9531e-05, 1.5631e-05], + [ 3.2902e-05, 6.5565e-04, 1.8167e-04, ..., 1.1969e-03, + 8.6427e-05, 3.6097e-04]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0256, 0.0414, -0.0035, -0.0075, 0.0235, -0.0301, 0.0046, 0.0010, + -0.0057, -0.0166], device='cuda:0'), grad: tensor([-0.0230, 0.0062, 0.0118, 0.0023, 0.0122, -0.0285, -0.0005, 0.0049, + 0.0069, 0.0078], device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 214.78, cls_loss 0.5079 cls_loss_mapping 0.0039 cls_loss_causal 0.4341 re_mapping 0.0097 re_causal 0.0211 /// teacc 98.84 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.0734, -0.1434, -0.1004, ..., -0.0378, 0.0611, -0.1102], + [-0.0803, -0.0875, -0.0832, ..., 0.1015, -0.0347, 0.2303], + [ 0.0028, -0.0275, -0.0468, ..., -0.0038, -0.0096, -0.0825], + ..., + [ 0.0387, -0.1161, 0.1703, ..., 0.0348, -0.1183, 0.0562], + [-0.0353, 0.0598, -0.1546, ..., -0.0549, 0.0133, -0.1041], + [-0.0540, 0.0810, 0.0484, ..., -0.0597, -0.0483, -0.0293]], + device='cuda:0'), grad: tensor([[ 6.6376e-04, 3.3545e-04, -1.9670e-04, ..., -3.6955e-04, + 5.6553e-04, -1.3733e-04], + [-2.0409e-03, -2.8419e-04, 3.0756e-04, ..., -4.3793e-03, + -3.8300e-03, 2.1343e-03], + [-6.4316e-03, 2.1279e-04, -2.3410e-05, ..., -9.3651e-04, + 1.5521e-04, 2.5916e-04], + ..., + [ 1.1988e-03, -1.4067e-03, -3.3379e-03, ..., -2.0142e-03, + 2.4080e-04, -2.1935e-03], + [ 1.6041e-03, 4.7493e-03, 4.8661e-04, ..., -2.0838e-04, + 3.3450e-04, -3.0174e-03], + [ 1.2131e-03, 1.0214e-03, 9.2745e-04, ..., 2.5654e-03, + 8.8596e-04, 9.2363e-04]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0254, 0.0413, -0.0035, -0.0071, 0.0235, -0.0311, 0.0051, 0.0008, + -0.0052, -0.0170], device='cuda:0'), grad: tensor([-0.0072, -0.0313, -0.0401, 0.0317, 0.0219, 0.0154, -0.0018, -0.0119, + -0.0133, 0.0367], device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 214.28, cls_loss 0.5148 cls_loss_mapping 0.0028 cls_loss_causal 0.4615 re_mapping 0.0093 re_causal 0.0208 /// teacc 98.71 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.0739, -0.1426, -0.1004, ..., -0.0370, 0.0600, -0.1093], + [-0.0799, -0.0880, -0.0839, ..., 0.1009, -0.0352, 0.2298], + [ 0.0037, -0.0272, -0.0467, ..., -0.0032, -0.0094, -0.0832], + ..., + [ 0.0390, -0.1165, 0.1705, ..., 0.0360, -0.1188, 0.0563], + [-0.0363, 0.0600, -0.1554, ..., -0.0552, 0.0126, -0.1051], + [-0.0546, 0.0815, 0.0484, ..., -0.0600, -0.0484, -0.0289]], + device='cuda:0'), grad: tensor([[ 1.4687e-04, 2.5272e-04, 1.8626e-06, ..., 9.9957e-05, + -9.9945e-04, 3.9840e-04], + [ 2.4110e-05, -3.7346e-03, 6.0529e-05, ..., 5.7173e-04, + 1.5259e-04, -4.2319e-04], + [-1.9875e-03, 8.8739e-04, 1.8910e-05, ..., 1.0357e-03, + 1.0026e-04, -3.2883e-03], + ..., + [ 1.3304e-03, -3.7766e-04, -3.7231e-03, ..., 3.2854e-04, + 1.1718e-04, 5.5552e-04], + [ 7.8157e-06, 3.8910e-03, 1.3292e-05, ..., -8.9502e-04, + 1.1039e-04, -5.7602e-04], + [ 3.7050e-04, -4.6387e-03, 3.3092e-03, ..., -2.7466e-03, + 1.1444e-04, 1.6069e-03]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0256, 0.0405, -0.0033, -0.0062, 0.0241, -0.0311, 0.0046, 0.0011, + -0.0050, -0.0176], device='cuda:0'), grad: tensor([-0.0163, -0.0209, -0.0117, -0.0120, 0.0013, 0.0147, 0.0193, 0.0304, + -0.0041, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 214.98, cls_loss 0.5295 cls_loss_mapping 0.0027 cls_loss_causal 0.4647 re_mapping 0.0093 re_causal 0.0212 /// teacc 98.74 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.0749, -0.1456, -0.1015, ..., -0.0374, 0.0605, -0.1082], + [-0.0799, -0.0879, -0.0852, ..., 0.1005, -0.0350, 0.2295], + [ 0.0040, -0.0267, -0.0468, ..., -0.0036, -0.0087, -0.0839], + ..., + [ 0.0394, -0.1168, 0.1718, ..., 0.0371, -0.1184, 0.0572], + [-0.0376, 0.0598, -0.1560, ..., -0.0554, 0.0119, -0.1050], + [-0.0546, 0.0816, 0.0485, ..., -0.0597, -0.0479, -0.0290]], + device='cuda:0'), grad: tensor([[ 1.3552e-03, -1.9002e-04, 2.3139e-04, ..., -9.0075e-04, + 1.4210e-04, -1.2531e-03], + [ 4.6897e-04, 5.7650e-04, 3.6031e-05, ..., 1.1082e-03, + 5.3011e-06, 3.9601e-04], + [-3.4103e-03, -5.6152e-03, 9.6917e-05, ..., -3.3627e-03, + -4.6692e-03, 3.1281e-04], + ..., + [-1.6174e-03, 4.3941e-04, 1.4746e-04, ..., -2.1591e-03, + 4.6015e-05, 4.6086e-04], + [ 2.7237e-03, -3.7932e-04, 2.2340e-04, ..., 1.9217e-03, + 3.3836e-03, -1.2560e-03], + [ 1.8244e-03, 1.1072e-03, 1.7524e-04, ..., 2.2621e-03, + 1.4126e-04, 2.6321e-04]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0251, 0.0400, -0.0021, -0.0071, 0.0230, -0.0307, 0.0037, 0.0019, + -0.0055, -0.0168], device='cuda:0'), grad: tensor([-0.0198, 0.0145, -0.0195, 0.0180, -0.0187, 0.0183, 0.0139, -0.0192, + -0.0035, 0.0161], device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 214.81, cls_loss 0.5180 cls_loss_mapping 0.0023 cls_loss_causal 0.4561 re_mapping 0.0094 re_causal 0.0217 /// teacc 98.78 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.0776, -0.1459, -0.1018, ..., -0.0380, 0.0606, -0.1085], + [-0.0790, -0.0888, -0.0857, ..., 0.0994, -0.0346, 0.2292], + [ 0.0049, -0.0260, -0.0476, ..., -0.0038, -0.0091, -0.0849], + ..., + [ 0.0395, -0.1176, 0.1723, ..., 0.0380, -0.1191, 0.0579], + [-0.0369, 0.0586, -0.1562, ..., -0.0555, 0.0113, -0.1054], + [-0.0550, 0.0820, 0.0493, ..., -0.0591, -0.0481, -0.0295]], + device='cuda:0'), grad: tensor([[ 6.7987e-07, 3.4356e-04, 1.1599e-04, ..., 1.6584e-03, + 1.3439e-06, 6.2943e-04], + [ 2.4736e-06, 6.5279e-04, 3.4142e-04, ..., 1.4772e-03, + 2.4159e-06, 4.0603e-04], + [ 1.9878e-05, -3.8624e-03, -3.1776e-03, ..., -2.1000e-03, + 1.6659e-05, -1.4172e-03], + ..., + [-2.1055e-05, 6.3562e-04, 3.5787e-04, ..., 2.6646e-03, + 6.0238e-06, 8.7738e-04], + [ 3.5111e-06, -1.1311e-03, 5.7077e-04, ..., 2.5272e-03, + 2.8685e-05, 9.6369e-04], + [ 7.4469e-06, 4.2582e-04, 1.8740e-04, ..., 2.8439e-03, + 2.8074e-05, 2.1000e-03]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0258, 0.0397, -0.0017, -0.0068, 0.0231, -0.0313, 0.0043, 0.0028, + -0.0060, -0.0169], device='cuda:0'), grad: tensor([ 0.0127, 0.0161, -0.0118, -0.0114, -0.0429, 0.0187, -0.0069, 0.0199, + -0.0111, 0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 214.97, cls_loss 0.5165 cls_loss_mapping 0.0042 cls_loss_causal 0.4608 re_mapping 0.0092 re_causal 0.0207 /// teacc 98.85 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.0782, -0.1465, -0.1015, ..., -0.0382, 0.0605, -0.1103], + [-0.0778, -0.0885, -0.0863, ..., 0.1006, -0.0338, 0.2305], + [ 0.0035, -0.0253, -0.0481, ..., -0.0034, -0.0079, -0.0844], + ..., + [ 0.0399, -0.1180, 0.1712, ..., 0.0377, -0.1193, 0.0564], + [-0.0361, 0.0585, -0.1549, ..., -0.0562, 0.0126, -0.1050], + [-0.0552, 0.0818, 0.0501, ..., -0.0582, -0.0483, -0.0289]], + device='cuda:0'), grad: tensor([[ 3.3355e-04, 8.3726e-07, 1.4462e-05, ..., 2.2068e-03, + 3.0947e-04, 1.5378e-04], + [ 7.7581e-04, 2.9132e-06, 1.0538e-03, ..., -9.2220e-04, + 6.4325e-04, 6.4945e-04], + [-4.5252e-04, -4.5824e-04, -4.7684e-04, ..., 1.6699e-03, + 4.3082e-04, -2.7895e-04], + ..., + [ 8.2874e-04, 5.7705e-06, -8.0967e-04, ..., -6.0129e-04, + 4.3559e-04, -1.3266e-03], + [ 4.4942e-04, 6.5207e-05, 3.2604e-05, ..., -1.0138e-03, + 3.9339e-04, 1.0669e-04], + [ 5.0354e-04, 2.0079e-06, 9.7752e-05, ..., -3.6964e-03, + 4.8018e-04, 2.8896e-04]], device='cuda:0') +Epoch 343, bias, value: tensor([-2.6358e-02, 4.0076e-02, -5.1674e-05, -7.1641e-03, 2.4269e-02, + -3.0877e-02, 2.8620e-03, 1.1799e-03, -5.6565e-03, -1.6877e-02], + device='cuda:0'), grad: tensor([ 0.0172, -0.0038, 0.0163, 0.0171, 0.0179, 0.0142, -0.0155, -0.0039, + -0.0157, -0.0439], device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 215.22, cls_loss 0.5221 cls_loss_mapping 0.0037 cls_loss_causal 0.4497 re_mapping 0.0095 re_causal 0.0214 /// teacc 98.82 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.0754, -0.1475, -0.1003, ..., -0.0369, 0.0602, -0.1095], + [-0.0769, -0.0889, -0.0870, ..., 0.0998, -0.0339, 0.2302], + [ 0.0018, -0.0252, -0.0485, ..., -0.0026, -0.0096, -0.0827], + ..., + [ 0.0398, -0.1179, 0.1711, ..., 0.0375, -0.1175, 0.0561], + [-0.0349, 0.0590, -0.1556, ..., -0.0570, 0.0122, -0.1058], + [-0.0558, 0.0820, 0.0502, ..., -0.0587, -0.0493, -0.0294]], + device='cuda:0'), grad: tensor([[ 5.4568e-05, 1.5688e-04, 4.2349e-05, ..., 3.5610e-03, + 2.6112e-03, 1.1406e-03], + [ 1.8282e-03, 5.0688e-04, 8.6355e-04, ..., 5.7793e-03, + 4.5681e-04, 2.9964e-03], + [-2.6779e-03, 3.6025e-04, 5.9414e-04, ..., -1.7807e-02, + 1.4400e-04, -5.0163e-03], + ..., + [ 3.2783e-04, 2.8110e-04, 8.6641e-04, ..., 5.4779e-03, + 8.8644e-04, 2.1992e-03], + [-8.7857e-05, 1.6737e-03, -1.1444e-03, ..., 4.0436e-03, + 1.1826e-03, 7.2813e-04], + [ 1.7440e-04, 2.6298e-04, -1.7099e-03, ..., 2.6321e-03, + -6.6223e-03, -7.4196e-04]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0258, 0.0401, 0.0003, -0.0071, 0.0240, -0.0313, 0.0032, 0.0018, + -0.0060, -0.0178], device='cuda:0'), grad: tensor([ 0.0231, 0.0243, -0.0743, 0.0176, -0.0284, 0.0248, -0.0210, 0.0326, + 0.0109, -0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 214.68, cls_loss 0.5410 cls_loss_mapping 0.0029 cls_loss_causal 0.4837 re_mapping 0.0094 re_causal 0.0215 /// teacc 98.63 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.0760, -0.1476, -0.1014, ..., -0.0361, 0.0612, -0.1094], + [-0.0767, -0.0892, -0.0870, ..., 0.1003, -0.0323, 0.2312], + [ 0.0020, -0.0252, -0.0485, ..., -0.0027, -0.0111, -0.0816], + ..., + [ 0.0396, -0.1179, 0.1714, ..., 0.0369, -0.1182, 0.0552], + [-0.0344, 0.0591, -0.1539, ..., -0.0560, 0.0129, -0.1066], + [-0.0558, 0.0802, 0.0499, ..., -0.0599, -0.0495, -0.0299]], + device='cuda:0'), grad: tensor([[ 3.3993e-06, 2.3693e-05, 4.4179e-04, ..., -1.0004e-03, + 2.3861e-03, 5.2681e-03], + [ 7.5758e-05, 2.9162e-05, 2.4068e-04, ..., 5.1074e-06, + -1.7378e-06, -4.6706e-04], + [ 1.5235e-04, 4.4316e-05, 2.5153e-04, ..., -4.6760e-05, + 7.3862e-04, -1.1444e-03], + ..., + [ 5.2810e-05, -3.4761e-04, -2.4223e-03, ..., -2.5826e-03, + -8.0490e-03, -1.1589e-02], + [-4.9770e-06, 3.4660e-05, 1.8883e-04, ..., 8.6355e-04, + -2.7433e-05, 1.1778e-03], + [ 2.3440e-05, -1.1317e-05, 2.5535e-04, ..., 1.1015e-03, + 9.5987e-04, 1.5554e-03]], device='cuda:0') +Epoch 345, bias, value: tensor([-2.6289e-02, 4.0450e-02, -4.1657e-05, -6.4780e-03, 2.4713e-02, + -3.1340e-02, 2.4803e-03, 8.1632e-04, -5.6647e-03, -1.7279e-02], + device='cuda:0'), grad: tensor([-0.0060, -0.0065, -0.0106, 0.0184, 0.0300, 0.0157, 0.0164, -0.0336, + -0.0457, 0.0220], device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 214.98, cls_loss 0.5749 cls_loss_mapping 0.0030 cls_loss_causal 0.5074 re_mapping 0.0090 re_causal 0.0211 /// teacc 98.75 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.0757, -0.1462, -0.1025, ..., -0.0359, 0.0586, -0.1098], + [-0.0771, -0.0894, -0.0867, ..., 0.0993, -0.0323, 0.2324], + [ 0.0028, -0.0253, -0.0489, ..., -0.0031, -0.0135, -0.0820], + ..., + [ 0.0397, -0.1181, 0.1720, ..., 0.0372, -0.1198, 0.0568], + [-0.0351, 0.0584, -0.1550, ..., -0.0557, 0.0129, -0.1087], + [-0.0558, 0.0801, 0.0482, ..., -0.0594, -0.0495, -0.0307]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0008, 0.0004, ..., -0.0016, 0.0005, 0.0003], + [ 0.0015, -0.0014, -0.0017, ..., -0.0014, -0.0011, -0.0015], + [ 0.0013, 0.0015, 0.0006, ..., 0.0020, 0.0008, 0.0005], + ..., + [-0.0042, -0.0001, -0.0025, ..., -0.0074, -0.0026, -0.0021], + [ 0.0010, -0.0372, 0.0003, ..., 0.0014, 0.0011, 0.0004], + [-0.0050, 0.0167, 0.0005, ..., -0.0002, -0.0013, 0.0005]], + device='cuda:0') +Epoch 346, bias, value: tensor([-0.0259, 0.0406, -0.0004, -0.0059, 0.0250, -0.0315, 0.0021, 0.0001, + -0.0059, -0.0168], device='cuda:0'), grad: tensor([-0.0168, -0.0181, 0.0196, 0.0246, 0.0189, 0.0274, 0.0097, -0.0374, + -0.0357, 0.0078], device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 215.05, cls_loss 0.5015 cls_loss_mapping 0.0017 cls_loss_causal 0.4328 re_mapping 0.0095 re_causal 0.0222 /// teacc 98.82 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.0757, -0.1465, -0.1019, ..., -0.0356, 0.0591, -0.1087], + [-0.0756, -0.0893, -0.0854, ..., 0.0997, -0.0320, 0.2346], + [ 0.0016, -0.0236, -0.0500, ..., -0.0036, -0.0140, -0.0832], + ..., + [ 0.0380, -0.1184, 0.1726, ..., 0.0377, -0.1198, 0.0565], + [-0.0345, 0.0583, -0.1552, ..., -0.0560, 0.0125, -0.1102], + [-0.0562, 0.0804, 0.0478, ..., -0.0605, -0.0501, -0.0317]], + device='cuda:0'), grad: tensor([[ 1.2422e-04, 2.2316e-04, 1.8254e-07, ..., 1.0233e-03, + 1.5807e-04, 6.8331e-04], + [ 3.2187e-05, 5.2214e-05, 1.0425e-04, ..., 1.8530e-03, + 2.1189e-05, 1.5974e-03], + [ 1.5879e-04, 7.4005e-04, 6.0424e-06, ..., 1.2856e-03, + 2.1815e-04, 6.3515e-04], + ..., + [-9.1887e-04, -1.6117e-03, -2.0242e-04, ..., 1.3285e-03, + -1.1663e-03, 4.9829e-04], + [ 8.3566e-05, 1.5891e-04, 3.8259e-06, ..., 9.8705e-04, + 9.7394e-05, 7.6771e-04], + [ 1.8513e-04, 7.1907e-03, 7.4685e-05, ..., 8.7881e-04, + 2.4188e-04, 2.5249e-04]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0257, 0.0417, -0.0012, -0.0060, 0.0238, -0.0311, 0.0036, -0.0003, + -0.0062, -0.0170], device='cuda:0'), grad: tensor([ 0.0151, -0.0068, -0.0115, -0.0154, -0.0117, -0.0181, -0.0137, 0.0044, + 0.0152, 0.0425], device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 214.51, cls_loss 0.5516 cls_loss_mapping 0.0035 cls_loss_causal 0.4983 re_mapping 0.0089 re_causal 0.0209 /// teacc 98.88 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.0754, -0.1462, -0.1021, ..., -0.0358, 0.0594, -0.1076], + [-0.0765, -0.0892, -0.0857, ..., 0.0993, -0.0322, 0.2338], + [ 0.0027, -0.0228, -0.0486, ..., -0.0041, -0.0157, -0.0833], + ..., + [ 0.0393, -0.1183, 0.1726, ..., 0.0387, -0.1209, 0.0575], + [-0.0336, 0.0581, -0.1542, ..., -0.0571, 0.0146, -0.1103], + [-0.0566, 0.0804, 0.0482, ..., -0.0602, -0.0503, -0.0323]], + device='cuda:0'), grad: tensor([[ 1.5986e-04, 8.3633e-07, 1.8731e-05, ..., 1.6189e-04, + 1.9386e-05, 5.0992e-05], + [ 1.1230e-04, 8.9686e-07, 1.4044e-05, ..., -7.3433e-05, + 6.6012e-06, -3.6144e-04], + [ 2.5272e-04, 4.4823e-05, 8.8930e-05, ..., 1.9169e-04, + 1.8934e-06, 1.5008e-04], + ..., + [ 1.7822e-04, 2.9922e-04, -8.4925e-04, ..., 7.3624e-04, + 1.8016e-05, -3.1815e-03], + [-2.9507e-03, 5.2214e-05, -2.4915e-04, ..., -1.6060e-03, + 4.3921e-06, 9.0599e-05], + [ 1.8919e-04, -5.1880e-04, 9.6607e-04, ..., -1.2360e-03, + -9.2387e-06, 3.1376e-03]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0254, 0.0412, -0.0021, -0.0052, 0.0236, -0.0305, 0.0025, 0.0008, + -0.0061, -0.0174], device='cuda:0'), grad: tensor([ 0.0359, 0.0067, 0.0109, 0.0063, -0.0179, 0.0176, 0.0125, -0.0194, + -0.0234, -0.0293], device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 214.64, cls_loss 0.5238 cls_loss_mapping 0.0039 cls_loss_causal 0.4550 re_mapping 0.0089 re_causal 0.0200 /// teacc 98.91 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.0745, -0.1460, -0.1033, ..., -0.0367, 0.0567, -0.1079], + [-0.0772, -0.0893, -0.0863, ..., 0.0997, -0.0320, 0.2337], + [ 0.0026, -0.0227, -0.0483, ..., -0.0037, -0.0125, -0.0841], + ..., + [ 0.0402, -0.1181, 0.1728, ..., 0.0385, -0.1200, 0.0586], + [-0.0344, 0.0583, -0.1541, ..., -0.0572, 0.0142, -0.1099], + [-0.0566, 0.0804, 0.0481, ..., -0.0595, -0.0505, -0.0323]], + device='cuda:0'), grad: tensor([[-1.2970e-03, 3.8743e-04, 4.2653e-04, ..., -3.2578e-03, + 1.1635e-04, -1.0166e-03], + [ 1.2517e-04, 2.5558e-04, 1.5557e-04, ..., 8.3590e-04, + 4.1798e-06, 6.4802e-04], + [-1.0471e-03, -1.1215e-03, -1.7366e-03, ..., -1.6603e-03, + -5.1928e-04, -5.1165e-04], + ..., + [ 1.7560e-04, 2.5702e-04, 7.0453e-05, ..., -6.8331e-04, + 5.9903e-05, -1.6041e-03], + [ 8.9407e-04, -6.4945e-04, 5.7650e-04, ..., 1.5144e-03, + 1.3717e-05, 1.1511e-03], + [ 2.4021e-04, -9.2149e-05, 1.4365e-04, ..., 7.6199e-04, + 7.5638e-05, 7.9966e-04]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0255, 0.0404, -0.0014, -0.0050, 0.0232, -0.0309, 0.0029, -0.0002, + -0.0060, -0.0162], device='cuda:0'), grad: tensor([-0.0388, 0.0069, -0.0031, 0.0048, 0.0028, 0.0052, 0.0034, 0.0038, + 0.0098, 0.0052], device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 214.61, cls_loss 0.5689 cls_loss_mapping 0.0050 cls_loss_causal 0.5020 re_mapping 0.0092 re_causal 0.0220 /// teacc 98.90 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.0739, -0.1461, -0.1028, ..., -0.0371, 0.0571, -0.1076], + [-0.0774, -0.0887, -0.0858, ..., 0.1004, -0.0324, 0.2334], + [ 0.0008, -0.0238, -0.0493, ..., -0.0042, -0.0132, -0.0861], + ..., + [ 0.0399, -0.1180, 0.1740, ..., 0.0387, -0.1191, 0.0591], + [-0.0340, 0.0582, -0.1555, ..., -0.0569, 0.0145, -0.1087], + [-0.0569, 0.0809, 0.0481, ..., -0.0597, -0.0506, -0.0323]], + device='cuda:0'), grad: tensor([[-9.5987e-04, 9.1255e-05, 2.4295e-04, ..., -1.3351e-05, + 2.7204e-04, -9.9599e-05], + [ 6.6683e-06, 2.9534e-05, 3.5954e-04, ..., 3.6545e-03, + 2.7680e-04, 1.0118e-03], + [ 2.3162e-04, 2.5344e-04, 1.8797e-03, ..., 4.5280e-03, + 3.8171e-04, 1.7157e-03], + ..., + [-1.4794e-04, 4.2289e-05, 1.6260e-03, ..., -1.1200e-02, + 2.8348e-04, -1.4277e-03], + [-4.1723e-05, 4.1342e-04, 6.6710e-04, ..., 2.6569e-03, + 2.7370e-04, 7.5388e-04], + [ 2.5821e-04, 1.5569e-04, 6.9284e-04, ..., 5.8842e-04, + 4.4703e-04, 2.1231e-04]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0258, 0.0409, -0.0011, -0.0058, 0.0241, -0.0302, 0.0034, -0.0007, + -0.0068, -0.0166], device='cuda:0'), grad: tensor([-0.0124, 0.0334, 0.0285, 0.0085, -0.0039, -0.0285, -0.0031, -0.0391, + 0.0234, -0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 214.74, cls_loss 0.5081 cls_loss_mapping 0.0031 cls_loss_causal 0.4535 re_mapping 0.0090 re_causal 0.0212 /// teacc 98.95 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.0750, -0.1465, -0.1031, ..., -0.0388, 0.0571, -0.1095], + [-0.0763, -0.0882, -0.0860, ..., 0.1018, -0.0332, 0.2328], + [-0.0007, -0.0246, -0.0505, ..., -0.0050, -0.0147, -0.0854], + ..., + [ 0.0407, -0.1188, 0.1763, ..., 0.0396, -0.1169, 0.0599], + [-0.0339, 0.0593, -0.1551, ..., -0.0569, 0.0161, -0.1089], + [-0.0575, 0.0807, 0.0474, ..., -0.0593, -0.0521, -0.0328]], + device='cuda:0'), grad: tensor([[ 5.0354e-04, 7.6473e-05, 2.0540e-04, ..., 7.5340e-04, + 3.3975e-06, 9.8884e-05], + [ 2.0103e-03, 1.3418e-03, 3.3712e-04, ..., 4.1962e-03, + 1.5562e-06, 2.7633e-04], + [ 1.5383e-03, 1.2659e-05, 3.8576e-04, ..., 1.1759e-03, + -1.5950e-04, 4.5943e-04], + ..., + [-2.5272e-04, 2.4848e-06, -1.4076e-02, ..., 1.2417e-03, + 1.3733e-04, -3.1376e-04], + [-1.1200e-02, -1.8635e-03, 1.5297e-03, ..., -3.2635e-03, + 1.3694e-05, 1.9574e-04], + [ 1.4963e-03, 6.0201e-06, 1.7376e-03, ..., -5.3520e-03, + 2.3469e-06, -3.7408e-04]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0266, 0.0409, -0.0019, -0.0062, 0.0243, -0.0303, 0.0028, 0.0004, + -0.0059, -0.0161], device='cuda:0'), grad: tensor([ 0.0072, 0.0229, 0.0084, 0.0608, -0.0155, 0.0210, 0.0130, -0.0392, + -0.0278, -0.0508], device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 214.79, cls_loss 0.4999 cls_loss_mapping 0.0038 cls_loss_causal 0.4463 re_mapping 0.0087 re_causal 0.0197 /// teacc 98.80 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.0754, -0.1466, -0.1032, ..., -0.0379, 0.0566, -0.1074], + [-0.0762, -0.0884, -0.0855, ..., 0.1022, -0.0331, 0.2317], + [-0.0004, -0.0259, -0.0507, ..., -0.0054, -0.0132, -0.0844], + ..., + [ 0.0412, -0.1205, 0.1756, ..., 0.0389, -0.1177, 0.0609], + [-0.0331, 0.0597, -0.1556, ..., -0.0570, 0.0165, -0.1096], + [-0.0575, 0.0811, 0.0481, ..., -0.0594, -0.0528, -0.0334]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0007, 0.0010, ..., 0.0020, 0.0004, 0.0008], + [ 0.0006, -0.0012, -0.0033, ..., -0.0013, 0.0010, -0.0022], + [ 0.0015, 0.0072, 0.0064, ..., 0.0102, 0.0035, 0.0043], + ..., + [ 0.0008, 0.0017, -0.0058, ..., 0.0009, 0.0023, 0.0035], + [-0.0014, 0.0004, 0.0014, ..., -0.0023, 0.0006, 0.0010], + [ 0.0003, 0.0013, 0.0025, ..., 0.0032, 0.0014, 0.0021]], + device='cuda:0') +Epoch 352, bias, value: tensor([-0.0266, 0.0400, -0.0009, -0.0063, 0.0233, -0.0289, 0.0018, 0.0017, + -0.0063, -0.0164], device='cuda:0'), grad: tensor([ 0.0157, -0.0096, 0.0570, -0.0846, -0.0362, 0.0421, 0.0139, -0.0097, + -0.0127, 0.0241], device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 214.56, cls_loss 0.5453 cls_loss_mapping 0.0031 cls_loss_causal 0.4725 re_mapping 0.0089 re_causal 0.0204 /// teacc 98.80 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.0756, -0.1456, -0.1027, ..., -0.0369, 0.0579, -0.1064], + [-0.0763, -0.0876, -0.0867, ..., 0.1023, -0.0334, 0.2328], + [ 0.0008, -0.0261, -0.0518, ..., -0.0054, -0.0142, -0.0852], + ..., + [ 0.0417, -0.1208, 0.1759, ..., 0.0372, -0.1195, 0.0608], + [-0.0319, 0.0610, -0.1560, ..., -0.0577, 0.0174, -0.1109], + [-0.0574, 0.0808, 0.0483, ..., -0.0602, -0.0541, -0.0337]], + device='cuda:0'), grad: tensor([[ 5.3942e-05, -5.5641e-05, -4.0698e-04, ..., -1.6296e-02, + -9.7132e-04, -1.9470e-02], + [ 1.0490e-04, 1.6257e-05, -1.5459e-03, ..., 3.4428e-03, + -1.8044e-03, 9.6283e-03], + [ 2.8849e-05, 4.3297e-04, 3.4046e-04, ..., 1.0614e-03, + 3.9220e-04, 1.6212e-03], + ..., + [-4.4560e-04, -2.5368e-03, 2.8300e-04, ..., 2.0447e-03, + 6.4421e-04, -5.5695e-03], + [ 1.2481e-04, 6.8188e-05, 1.8239e-04, ..., 8.1587e-04, + 2.1625e-04, 8.6403e-04], + [ 5.1737e-05, 1.7681e-03, 4.9019e-04, ..., 1.2426e-03, + 3.3212e-04, 5.8289e-03]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0259, 0.0409, -0.0012, -0.0066, 0.0224, -0.0289, 0.0027, 0.0016, + -0.0067, -0.0169], device='cuda:0'), grad: tensor([-0.0426, -0.0475, -0.0047, 0.0222, 0.0257, 0.0235, -0.0072, -0.0298, + 0.0207, 0.0397], device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 214.38, cls_loss 0.5440 cls_loss_mapping 0.0033 cls_loss_causal 0.4797 re_mapping 0.0086 re_causal 0.0207 /// teacc 98.93 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.0740, -0.1461, -0.1028, ..., -0.0371, 0.0587, -0.1057], + [-0.0775, -0.0875, -0.0867, ..., 0.1020, -0.0323, 0.2317], + [ 0.0020, -0.0254, -0.0522, ..., -0.0049, -0.0142, -0.0830], + ..., + [ 0.0408, -0.1219, 0.1758, ..., 0.0375, -0.1216, 0.0612], + [-0.0329, 0.0613, -0.1544, ..., -0.0577, 0.0167, -0.1130], + [-0.0562, 0.0796, 0.0481, ..., -0.0594, -0.0536, -0.0333]], + device='cuda:0'), grad: tensor([[-1.1053e-03, -1.9684e-03, 1.6853e-05, ..., -1.6584e-03, + 5.5122e-04, -6.8569e-04], + [ 3.2501e-03, 5.7727e-05, 4.9286e-06, ..., -1.8597e-03, + 7.7772e-04, 1.8692e-04], + [ 1.6375e-03, 4.7952e-05, 3.3116e-04, ..., 4.9896e-03, + 2.0676e-03, -2.8229e-04], + ..., + [ 4.9829e-04, 5.0116e-04, -5.5790e-04, ..., 2.4204e-03, + 3.9101e-04, 2.3346e-03], + [-7.0305e-03, 8.2970e-04, 2.8878e-05, ..., -5.9509e-03, + 2.9111e-04, 1.1644e-03], + [ 4.8137e-04, -1.0557e-03, 1.0186e-04, ..., -2.4490e-03, + -6.0539e-03, -6.8092e-03]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0266, 0.0416, -0.0010, -0.0078, 0.0225, -0.0297, 0.0033, 0.0014, + -0.0078, -0.0145], device='cuda:0'), grad: tensor([ 0.0016, 0.0317, 0.0341, 0.0168, 0.0184, 0.0108, -0.0134, -0.0090, + -0.0568, -0.0342], device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 215.01, cls_loss 0.5350 cls_loss_mapping 0.0036 cls_loss_causal 0.4688 re_mapping 0.0085 re_causal 0.0205 /// teacc 98.87 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.0723, -0.1460, -0.1024, ..., -0.0363, 0.0589, -0.1057], + [-0.0780, -0.0869, -0.0873, ..., 0.1017, -0.0354, 0.2330], + [ 0.0027, -0.0257, -0.0523, ..., -0.0054, -0.0132, -0.0820], + ..., + [ 0.0415, -0.1221, 0.1752, ..., 0.0378, -0.1235, 0.0618], + [-0.0335, 0.0620, -0.1531, ..., -0.0584, 0.0172, -0.1132], + [-0.0576, 0.0795, 0.0472, ..., -0.0593, -0.0528, -0.0342]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, 8.6799e-06, 2.9188e-06, ..., -4.7569e-03, + -1.5221e-03, -4.3259e-03], + [ 1.3635e-06, 5.3346e-06, 4.9435e-06, ..., 7.2861e-04, + 4.2915e-04, 6.9332e-04], + [ 1.5553e-06, -4.8220e-05, 2.3887e-05, ..., 1.1101e-03, + 4.7493e-04, 1.2741e-03], + ..., + [-1.1280e-05, 7.5735e-06, -1.0654e-06, ..., 8.1015e-04, + 3.6192e-04, 7.9679e-04], + [ 1.3411e-07, -1.5132e-05, 4.6998e-05, ..., 3.8195e-04, + 2.1780e-04, 3.9220e-04], + [ 7.5586e-06, 4.8168e-06, 1.0297e-05, ..., 3.9601e-04, + 1.8740e-04, 4.0698e-04]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0246, 0.0406, -0.0008, -0.0083, 0.0229, -0.0286, 0.0017, 0.0014, + -0.0085, -0.0144], device='cuda:0'), grad: tensor([-0.0261, 0.0041, 0.0066, 0.0019, 0.0073, 0.0021, -0.0048, 0.0044, + 0.0022, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 354---------------------------------------------------- +epoch 354, time 232.28, cls_loss 0.5171 cls_loss_mapping 0.0021 cls_loss_causal 0.4497 re_mapping 0.0087 re_causal 0.0208 /// teacc 98.98 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.0722, -0.1458, -0.1033, ..., -0.0372, 0.0590, -0.1047], + [-0.0774, -0.0869, -0.0873, ..., 0.1017, -0.0358, 0.2314], + [ 0.0022, -0.0259, -0.0528, ..., -0.0064, -0.0131, -0.0828], + ..., + [ 0.0413, -0.1209, 0.1753, ..., 0.0385, -0.1232, 0.0614], + [-0.0321, 0.0610, -0.1533, ..., -0.0569, 0.0166, -0.1122], + [-0.0585, 0.0797, 0.0468, ..., -0.0610, -0.0519, -0.0351]], + device='cuda:0'), grad: tensor([[-8.8358e-04, -2.2335e-03, -1.3542e-04, ..., -4.2953e-03, + 3.2330e-04, -6.8436e-03], + [-1.8015e-03, 1.6749e-05, -1.5044e-04, ..., -3.0861e-03, + -2.8877e-03, 2.1191e-03], + [ 2.1040e-04, 1.3340e-04, 8.7082e-05, ..., -6.7091e-04, + 2.7490e-04, -2.4586e-03], + ..., + [ 1.2243e-04, 2.2352e-05, 1.8373e-05, ..., 1.6623e-03, + 1.8489e-04, 3.1223e-03], + [ 3.8743e-04, 1.1368e-03, 5.0843e-05, ..., -4.8599e-03, + 2.3139e-04, -5.4169e-03], + [ 1.9014e-04, 3.4571e-04, 2.9162e-05, ..., 2.2221e-03, + 2.3973e-04, 3.1185e-03]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0258, 0.0412, -0.0011, -0.0089, 0.0228, -0.0268, 0.0016, 0.0020, + -0.0081, -0.0152], device='cuda:0'), grad: tensor([-0.0616, -0.0110, -0.0168, 0.0440, 0.0217, 0.0033, -0.0137, 0.0194, + -0.0078, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 214.90, cls_loss 0.5472 cls_loss_mapping 0.0026 cls_loss_causal 0.4828 re_mapping 0.0087 re_causal 0.0198 /// teacc 98.84 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.0730, -0.1457, -0.1031, ..., -0.0379, 0.0590, -0.1051], + [-0.0769, -0.0878, -0.0868, ..., 0.1005, -0.0344, 0.2315], + [ 0.0027, -0.0260, -0.0508, ..., -0.0072, -0.0134, -0.0824], + ..., + [ 0.0416, -0.1195, 0.1753, ..., 0.0396, -0.1223, 0.0616], + [-0.0319, 0.0608, -0.1537, ..., -0.0552, 0.0165, -0.1111], + [-0.0585, 0.0796, 0.0467, ..., -0.0598, -0.0520, -0.0347]], + device='cuda:0'), grad: tensor([[ 8.2493e-04, 1.8757e-06, 1.1548e-05, ..., 1.8301e-03, + 1.1299e-02, 9.2459e-04], + [ 6.0701e-04, 2.1458e-06, 1.4745e-05, ..., 3.2597e-03, + 2.7627e-05, 1.6079e-03], + [-4.0436e-03, 9.4414e-05, 5.8365e-04, ..., -8.0872e-03, + 4.3464e-04, -9.3536e-03], + ..., + [-2.4929e-03, 7.2300e-05, 4.0340e-04, ..., 8.6737e-04, + 3.0279e-04, -2.7776e-04], + [ 2.7561e-04, 9.4175e-05, 5.7316e-04, ..., 1.4172e-03, + 5.2691e-04, 6.4087e-04], + [ 3.1052e-03, -4.4033e-06, 4.4614e-05, ..., 1.0328e-03, + 1.2350e-04, 1.9550e-03]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0257, 0.0410, 0.0002, -0.0089, 0.0226, -0.0277, 0.0007, 0.0023, + -0.0080, -0.0150], device='cuda:0'), grad: tensor([ 0.0298, 0.0161, -0.0495, 0.0023, -0.0414, 0.0112, 0.0087, 0.0026, + -0.0086, 0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 215.05, cls_loss 0.5552 cls_loss_mapping 0.0025 cls_loss_causal 0.4793 re_mapping 0.0090 re_causal 0.0218 /// teacc 98.90 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.0721, -0.1463, -0.1027, ..., -0.0372, 0.0584, -0.1057], + [-0.0778, -0.0885, -0.0872, ..., 0.0996, -0.0344, 0.2312], + [ 0.0030, -0.0260, -0.0508, ..., -0.0055, -0.0138, -0.0823], + ..., + [ 0.0421, -0.1191, 0.1753, ..., 0.0397, -0.1225, 0.0618], + [-0.0317, 0.0610, -0.1535, ..., -0.0549, 0.0169, -0.1101], + [-0.0594, 0.0800, 0.0468, ..., -0.0607, -0.0519, -0.0360]], + device='cuda:0'), grad: tensor([[ 3.3092e-04, 1.2629e-06, 1.4696e-06, ..., 1.4782e-03, + 3.3112e-02, -2.0218e-04], + [ 1.3399e-04, 7.5772e-06, 4.3102e-06, ..., 1.3447e-03, + 1.0748e-03, 1.1690e-05], + [-1.4772e-03, 2.3913e-04, 1.1265e-04, ..., -4.7455e-03, + 2.4185e-03, 2.7016e-05], + ..., + [ 2.8086e-04, 4.0025e-05, -4.8369e-05, ..., 9.4604e-04, + 6.7663e-04, -6.8951e-04], + [ 3.2973e-04, 1.8463e-05, 5.0366e-05, ..., -6.8207e-03, + -8.7357e-03, 1.0896e-04], + [ 2.3437e-04, -1.2672e-04, -1.1250e-05, ..., 1.1053e-03, + 2.6779e-03, 5.8889e-05]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0253, 0.0408, 0.0002, -0.0089, 0.0223, -0.0280, 0.0005, 0.0026, + -0.0068, -0.0160], device='cuda:0'), grad: tensor([ 0.0446, 0.0153, -0.0433, 0.0170, -0.0271, -0.0161, 0.0326, 0.0075, + -0.0467, 0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 214.89, cls_loss 0.5153 cls_loss_mapping 0.0042 cls_loss_causal 0.4497 re_mapping 0.0086 re_causal 0.0204 /// teacc 98.77 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.0711, -0.1470, -0.1032, ..., -0.0361, 0.0584, -0.1056], + [-0.0782, -0.0887, -0.0859, ..., 0.0997, -0.0339, 0.2298], + [ 0.0026, -0.0254, -0.0502, ..., -0.0061, -0.0143, -0.0804], + ..., + [ 0.0418, -0.1200, 0.1756, ..., 0.0404, -0.1225, 0.0622], + [-0.0325, 0.0618, -0.1545, ..., -0.0545, 0.0180, -0.1102], + [-0.0582, 0.0800, 0.0469, ..., -0.0603, -0.0522, -0.0361]], + device='cuda:0'), grad: tensor([[ 1.2517e-04, 2.0075e-04, 5.0247e-05, ..., 1.4706e-03, + -1.1459e-05, 4.7874e-04], + [ 1.7174e-06, 9.1887e-04, 3.9011e-05, ..., -2.5768e-03, + 9.3132e-09, 3.0136e-04], + [ 1.8578e-03, 1.3762e-03, 7.3767e-04, ..., 3.9024e-03, + 6.2771e-07, -4.0603e-04], + ..., + [ 2.8357e-05, 4.7517e-04, -1.6642e-04, ..., 1.6422e-03, + 7.4506e-09, -9.8825e-05], + [ 2.7120e-05, 1.8415e-03, 2.1279e-05, ..., 2.6646e-03, + 1.1958e-06, 1.3962e-03], + [-5.3644e-07, 3.0637e-04, 3.7640e-05, ..., 7.7057e-04, + 2.0992e-06, 4.7760e-03]], device='cuda:0') +Epoch 359, bias, value: tensor([-2.5404e-02, 4.1678e-02, 6.0835e-05, -1.0551e-02, 2.1157e-02, + -2.7930e-02, 7.6998e-04, 2.5739e-03, -5.7277e-03, -1.5184e-02], + device='cuda:0'), grad: tensor([ 0.0117, 0.0091, -0.0345, -0.0109, -0.0031, -0.0370, -0.0220, 0.0204, + 0.0464, 0.0199], device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 215.04, cls_loss 0.5407 cls_loss_mapping 0.0027 cls_loss_causal 0.4708 re_mapping 0.0087 re_causal 0.0218 /// teacc 98.81 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.0723, -0.1477, -0.1046, ..., -0.0370, 0.0584, -0.1062], + [-0.0789, -0.0900, -0.0868, ..., 0.0981, -0.0347, 0.2303], + [ 0.0035, -0.0261, -0.0508, ..., -0.0063, -0.0150, -0.0810], + ..., + [ 0.0411, -0.1178, 0.1764, ..., 0.0433, -0.1220, 0.0619], + [-0.0337, 0.0621, -0.1534, ..., -0.0543, 0.0175, -0.1103], + [-0.0568, 0.0803, 0.0462, ..., -0.0599, -0.0526, -0.0351]], + device='cuda:0'), grad: tensor([[ 3.9520e-03, 2.3830e-04, 1.9646e-04, ..., -2.6016e-03, + 7.2360e-05, -3.7217e-04], + [-5.3692e-04, 2.1052e-04, -4.2498e-05, ..., -1.1330e-02, + 1.7118e-04, -3.0994e-03], + [ 1.2197e-03, 8.0252e-04, 6.7234e-04, ..., 4.7073e-03, + 1.7905e-04, 1.6747e-03], + ..., + [ 4.8409e-03, 3.6411e-03, 2.1458e-04, ..., 1.1688e-02, + 1.0616e-04, 2.5597e-03], + [ 1.1683e-03, 6.3610e-04, 5.1785e-04, ..., 3.6545e-03, + 4.1753e-05, 2.3251e-03], + [ 3.5706e-03, 1.6766e-03, 2.9540e-04, ..., 3.4618e-03, + 6.5207e-05, 3.4504e-03]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0265, 0.0421, -0.0003, -0.0104, 0.0207, -0.0273, 0.0011, 0.0036, + -0.0068, -0.0149], device='cuda:0'), grad: tensor([-0.0045, -0.0057, 0.0231, 0.0044, -0.0430, -0.0277, -0.0371, 0.0424, + 0.0203, 0.0278], device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 215.05, cls_loss 0.5354 cls_loss_mapping 0.0025 cls_loss_causal 0.4849 re_mapping 0.0087 re_causal 0.0204 /// teacc 98.93 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.0741, -0.1485, -0.1055, ..., -0.0372, 0.0584, -0.1053], + [-0.0791, -0.0912, -0.0866, ..., 0.0983, -0.0348, 0.2302], + [ 0.0016, -0.0265, -0.0534, ..., -0.0076, -0.0147, -0.0822], + ..., + [ 0.0420, -0.1181, 0.1781, ..., 0.0438, -0.1226, 0.0618], + [-0.0328, 0.0627, -0.1538, ..., -0.0531, 0.0157, -0.1087], + [-0.0568, 0.0789, 0.0447, ..., -0.0608, -0.0530, -0.0356]], + device='cuda:0'), grad: tensor([[-0.0061, 0.0003, 0.0004, ..., -0.0069, -0.0013, -0.0020], + [ 0.0026, 0.0006, 0.0008, ..., 0.0025, 0.0007, -0.0006], + [-0.0001, -0.0047, 0.0006, ..., -0.0003, 0.0004, 0.0014], + ..., + [ 0.0006, 0.0005, -0.0064, ..., 0.0022, 0.0004, -0.0074], + [-0.0039, 0.0017, 0.0012, ..., -0.0057, -0.0006, 0.0017], + [ 0.0023, 0.0006, 0.0070, ..., 0.0026, 0.0004, 0.0032]], + device='cuda:0') +Epoch 361, bias, value: tensor([-0.0268, 0.0418, -0.0017, -0.0101, 0.0223, -0.0283, 0.0012, 0.0034, + -0.0059, -0.0147], device='cuda:0'), grad: tensor([-0.0112, 0.0078, -0.0345, 0.0062, 0.0254, 0.0319, -0.0018, -0.0134, + -0.0331, 0.0227], device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 214.78, cls_loss 0.5041 cls_loss_mapping 0.0024 cls_loss_causal 0.4445 re_mapping 0.0084 re_causal 0.0199 /// teacc 98.92 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.0735, -0.1487, -0.1054, ..., -0.0371, 0.0576, -0.1058], + [-0.0788, -0.0911, -0.0866, ..., 0.0983, -0.0357, 0.2301], + [ 0.0034, -0.0273, -0.0525, ..., -0.0068, -0.0144, -0.0821], + ..., + [ 0.0407, -0.1191, 0.1784, ..., 0.0442, -0.1207, 0.0631], + [-0.0332, 0.0636, -0.1544, ..., -0.0540, 0.0152, -0.1097], + [-0.0575, 0.0793, 0.0448, ..., -0.0605, -0.0523, -0.0349]], + device='cuda:0'), grad: tensor([[ 2.0489e-07, 2.6077e-07, 1.1846e-06, ..., -4.1733e-03, + 1.5271e-04, -1.2283e-03], + [ 1.4063e-06, 1.6578e-07, 3.8743e-06, ..., 1.1883e-03, + 2.7418e-04, 1.1766e-04], + [ 9.6083e-05, 4.7348e-06, 2.5988e-04, ..., 2.2106e-03, + 6.7902e-04, 1.4937e-04], + ..., + [ 2.8849e-04, 1.0431e-07, 4.9782e-04, ..., 2.0180e-03, + 6.7186e-04, 2.6560e-04], + [ 4.9546e-06, 3.2008e-05, 1.0043e-05, ..., -7.7009e-04, + 5.2166e-04, 1.0872e-04], + [ 7.7486e-05, 4.4517e-06, 1.5485e-04, ..., 1.1635e-03, + 1.0719e-03, 1.4389e-04]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0263, 0.0425, -0.0010, -0.0112, 0.0228, -0.0296, 0.0011, 0.0044, + -0.0059, -0.0154], device='cuda:0'), grad: tensor([-0.0441, 0.0245, -0.0353, 0.0098, 0.0177, -0.0090, -0.0136, 0.0322, + -0.0056, 0.0234], device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 214.94, cls_loss 0.5597 cls_loss_mapping 0.0028 cls_loss_causal 0.4838 re_mapping 0.0093 re_causal 0.0215 /// teacc 98.83 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.0738, -0.1494, -0.1058, ..., -0.0374, 0.0574, -0.1053], + [-0.0794, -0.0922, -0.0862, ..., 0.0993, -0.0361, 0.2300], + [ 0.0029, -0.0258, -0.0539, ..., -0.0071, -0.0150, -0.0831], + ..., + [ 0.0405, -0.1184, 0.1777, ..., 0.0429, -0.1220, 0.0633], + [-0.0326, 0.0631, -0.1531, ..., -0.0535, 0.0156, -0.1086], + [-0.0571, 0.0795, 0.0446, ..., -0.0612, -0.0515, -0.0361]], + device='cuda:0'), grad: tensor([[ 1.9699e-05, 4.3440e-04, 4.3958e-07, ..., 5.4264e-04, + 3.7886e-06, 1.1873e-04], + [ 3.0965e-05, -2.9583e-03, 6.3069e-06, ..., -2.8801e-03, + 1.7174e-06, -1.0281e-03], + [ 1.0985e-04, 1.0452e-03, 2.9311e-05, ..., 1.6708e-03, + 5.2378e-06, 2.1183e-04], + ..., + [-1.0505e-05, 5.2738e-04, -5.2780e-05, ..., 3.7718e-04, + 5.8681e-05, 1.0496e-04], + [ 3.9428e-05, 1.8291e-03, 2.4159e-06, ..., 6.5660e-04, + -8.7585e-03, -5.3520e-03], + [ 1.5974e-05, 8.7128e-03, 1.5929e-05, ..., 3.9768e-04, + 6.7406e-03, 4.1847e-03]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0261, 0.0426, -0.0010, -0.0105, 0.0234, -0.0304, 0.0005, 0.0034, + -0.0047, -0.0159], device='cuda:0'), grad: tensor([-0.0171, -0.0159, 0.0201, -0.0428, 0.0172, 0.0056, -0.0022, -0.0195, + -0.0135, 0.0683], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 362---------------------------------------------------- +epoch 362, time 231.09, cls_loss 0.5709 cls_loss_mapping 0.0042 cls_loss_causal 0.5029 re_mapping 0.0091 re_causal 0.0209 /// teacc 98.99 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.0721, -0.1493, -0.1056, ..., -0.0373, 0.0580, -0.1053], + [-0.0804, -0.0924, -0.0866, ..., 0.0982, -0.0362, 0.2305], + [ 0.0031, -0.0265, -0.0548, ..., -0.0065, -0.0151, -0.0815], + ..., + [ 0.0415, -0.1189, 0.1767, ..., 0.0423, -0.1191, 0.0631], + [-0.0321, 0.0633, -0.1535, ..., -0.0529, 0.0164, -0.1092], + [-0.0592, 0.0800, 0.0452, ..., -0.0619, -0.0513, -0.0354]], + device='cuda:0'), grad: tensor([[ 8.8882e-04, 7.4339e-04, 4.1223e-04, ..., 2.2240e-03, + 1.7920e-03, 7.7724e-04], + [ 2.3496e-04, 2.6643e-05, 1.0574e-04, ..., -7.1144e-03, + 4.9400e-04, -4.8256e-03], + [ 4.9496e-04, 1.7929e-02, 5.0087e-03, ..., 3.9711e-03, + -1.0777e-03, 5.1355e-04], + ..., + [-2.6870e-04, -2.0203e-02, -7.8106e-04, ..., 1.0526e-04, + 7.0620e-04, -7.8964e-04], + [-2.3727e-03, 1.5569e-04, 3.5596e-04, ..., -7.2145e-04, + -6.4735e-03, -9.9945e-04], + [ 6.1560e-04, 4.6110e-04, 6.6090e-04, ..., 1.3151e-03, + 7.4768e-04, 7.4768e-04]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0267, 0.0407, 0.0003, -0.0111, 0.0229, -0.0300, 0.0018, 0.0034, + -0.0042, -0.0157], device='cuda:0'), grad: tensor([ 0.0255, -0.0139, 0.0483, 0.0106, -0.0128, -0.0466, 0.0271, -0.0119, + -0.0463, 0.0200], device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 214.79, cls_loss 0.5515 cls_loss_mapping 0.0026 cls_loss_causal 0.4865 re_mapping 0.0087 re_causal 0.0208 /// teacc 98.91 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.0719, -0.1497, -0.1066, ..., -0.0372, 0.0587, -0.1065], + [-0.0800, -0.0928, -0.0865, ..., 0.0984, -0.0366, 0.2290], + [ 0.0024, -0.0276, -0.0551, ..., -0.0082, -0.0152, -0.0802], + ..., + [ 0.0416, -0.1174, 0.1768, ..., 0.0429, -0.1198, 0.0639], + [-0.0314, 0.0634, -0.1534, ..., -0.0539, 0.0167, -0.1086], + [-0.0596, 0.0798, 0.0455, ..., -0.0610, -0.0517, -0.0355]], + device='cuda:0'), grad: tensor([[ 1.1748e-04, 3.8818e-06, 7.0669e-06, ..., 8.3113e-04, + 5.4455e-04, 1.6678e-02], + [ 8.6486e-05, 6.7428e-07, 8.1025e-07, ..., 2.0874e-04, + 1.6952e-04, 3.1686e-04], + [ 7.8392e-04, -5.1886e-05, 2.3991e-06, ..., 5.0497e-04, + 2.6655e-04, 3.2272e-03], + ..., + [ 3.8195e-04, 6.6757e-06, 2.0191e-05, ..., 1.9777e-04, + 5.6803e-05, 1.2960e-03], + [ 2.5539e-03, -5.0887e-06, 8.5086e-06, ..., 1.0080e-03, + 2.6917e-04, 8.0776e-04], + [ 1.4286e-03, 5.5954e-06, 6.2138e-06, ..., 4.9639e-04, + 1.0878e-04, 5.2299e-03]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0257, 0.0405, -0.0012, -0.0100, 0.0222, -0.0296, 0.0007, 0.0037, + -0.0044, -0.0149], device='cuda:0'), grad: tensor([ 0.0009, 0.0048, 0.0076, -0.0142, -0.0452, 0.0414, -0.0057, 0.0046, + -0.0203, 0.0259], device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 214.82, cls_loss 0.5162 cls_loss_mapping 0.0033 cls_loss_causal 0.4604 re_mapping 0.0086 re_causal 0.0201 /// teacc 98.79 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.0698, -0.1494, -0.1057, ..., -0.0367, 0.0592, -0.1071], + [-0.0799, -0.0929, -0.0866, ..., 0.0988, -0.0370, 0.2300], + [ 0.0039, -0.0281, -0.0536, ..., -0.0078, -0.0165, -0.0798], + ..., + [ 0.0411, -0.1170, 0.1764, ..., 0.0429, -0.1195, 0.0636], + [-0.0328, 0.0638, -0.1522, ..., -0.0538, 0.0180, -0.1092], + [-0.0598, 0.0795, 0.0452, ..., -0.0611, -0.0522, -0.0363]], + device='cuda:0'), grad: tensor([[ 7.5251e-07, 1.0805e-03, 3.5787e-04, ..., 1.2522e-03, + -1.1548e-05, 7.1144e-04], + [ 2.3339e-06, 2.1863e-04, -2.1291e-04, ..., -1.5125e-03, + -3.7581e-05, -2.7866e-03], + [ 1.8731e-05, 7.2098e-03, 4.9820e-03, ..., 1.1883e-03, + 2.2011e-03, 1.2817e-03], + ..., + [-2.9355e-05, 1.2455e-03, -1.3456e-03, ..., 9.8324e-04, + 3.2693e-05, 9.9659e-04], + [ 4.5635e-07, -4.3488e-03, -4.1733e-03, ..., 1.9894e-03, + -2.7351e-03, 2.1706e-03], + [ 5.4464e-06, -7.6752e-03, -8.0109e-04, ..., -2.2621e-03, + 1.4566e-05, -2.9316e-03]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0264, 0.0395, -0.0011, -0.0109, 0.0231, -0.0299, 0.0015, 0.0044, + -0.0040, -0.0151], device='cuda:0'), grad: tensor([-0.0117, 0.0084, 0.0491, -0.0031, -0.0041, -0.0071, -0.0133, 0.0103, + 0.0047, -0.0332], device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 214.70, cls_loss 0.5041 cls_loss_mapping 0.0022 cls_loss_causal 0.4473 re_mapping 0.0087 re_causal 0.0208 /// teacc 98.81 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.0683, -0.1500, -0.1059, ..., -0.0368, 0.0600, -0.1074], + [-0.0810, -0.0919, -0.0858, ..., 0.0986, -0.0375, 0.2301], + [ 0.0035, -0.0290, -0.0536, ..., -0.0078, -0.0168, -0.0796], + ..., + [ 0.0415, -0.1194, 0.1766, ..., 0.0430, -0.1198, 0.0642], + [-0.0320, 0.0661, -0.1514, ..., -0.0527, 0.0173, -0.1100], + [-0.0617, 0.0801, 0.0448, ..., -0.0609, -0.0526, -0.0361]], + device='cuda:0'), grad: tensor([[ 6.2037e-04, 2.8000e-03, 2.7204e-04, ..., -1.1307e-04, + -7.2060e-03, 6.8331e-04], + [ 8.0299e-04, 5.1916e-05, 4.1187e-05, ..., 1.3316e-04, + 9.4748e-04, -1.7624e-03], + [ 7.6532e-04, 7.0214e-05, 9.6977e-05, ..., 1.4582e-03, + 1.0805e-03, 8.8882e-04], + ..., + [ 6.3372e-04, 5.7578e-05, 2.6455e-03, ..., 4.3259e-03, + 1.5860e-03, 4.5433e-03], + [ 6.1131e-04, 1.9760e-03, 5.8794e-04, ..., -2.8877e-03, + 7.8726e-04, -2.7428e-03], + [ 1.0118e-03, -4.1747e-04, 8.2397e-04, ..., 2.1114e-03, + 1.3857e-03, 1.7653e-03]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0268, 0.0402, -0.0016, -0.0107, 0.0235, -0.0306, 0.0017, 0.0042, + -0.0034, -0.0154], device='cuda:0'), grad: tensor([-0.0006, 0.0099, 0.0173, 0.0116, -0.0635, -0.0340, 0.0167, 0.0334, + -0.0113, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 214.78, cls_loss 0.5072 cls_loss_mapping 0.0022 cls_loss_causal 0.4381 re_mapping 0.0096 re_causal 0.0219 /// teacc 98.74 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.0679, -0.1515, -0.1063, ..., -0.0375, 0.0607, -0.1081], + [-0.0805, -0.0923, -0.0862, ..., 0.0988, -0.0380, 0.2311], + [ 0.0028, -0.0286, -0.0535, ..., -0.0064, -0.0159, -0.0797], + ..., + [ 0.0409, -0.1204, 0.1766, ..., 0.0434, -0.1200, 0.0638], + [-0.0318, 0.0654, -0.1508, ..., -0.0531, 0.0173, -0.1105], + [-0.0618, 0.0799, 0.0440, ..., -0.0612, -0.0529, -0.0366]], + device='cuda:0'), grad: tensor([[ 4.2391e-04, 4.0627e-04, 1.9759e-05, ..., 6.2847e-04, + 3.2520e-04, 9.6738e-05], + [ 2.9087e-04, 2.2697e-04, 1.6704e-05, ..., -4.0855e-03, + 1.8847e-04, 4.6343e-05], + [-3.2349e-03, 2.8372e-04, 1.1611e-04, ..., -4.6062e-04, + -4.0865e-04, 4.0007e-04], + ..., + [ 4.2105e-04, 1.3864e-04, -2.2805e-04, ..., -1.7130e-04, + 1.8239e-04, -4.0970e-03], + [ 5.8502e-05, -5.1041e-03, -6.2823e-05, ..., 2.2042e-04, + -3.1452e-03, -1.5879e-04], + [ 6.2609e-04, 4.6110e-04, 5.2065e-05, ..., 1.0386e-03, + 3.2258e-04, 3.2196e-03]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0277, 0.0413, -0.0016, -0.0115, 0.0229, -0.0301, 0.0019, 0.0042, + -0.0037, -0.0146], device='cuda:0'), grad: tensor([ 0.0229, -0.0322, -0.0617, -0.0137, 0.0347, 0.0394, 0.0079, 0.0021, + -0.0631, 0.0638], device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 214.88, cls_loss 0.5229 cls_loss_mapping 0.0030 cls_loss_causal 0.4625 re_mapping 0.0085 re_causal 0.0195 /// teacc 98.84 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.0683, -0.1515, -0.1067, ..., -0.0370, 0.0620, -0.1087], + [-0.0803, -0.0932, -0.0863, ..., 0.0983, -0.0371, 0.2317], + [ 0.0021, -0.0277, -0.0538, ..., -0.0073, -0.0161, -0.0799], + ..., + [ 0.0408, -0.1199, 0.1765, ..., 0.0433, -0.1205, 0.0641], + [-0.0324, 0.0664, -0.1517, ..., -0.0534, 0.0173, -0.1121], + [-0.0616, 0.0788, 0.0436, ..., -0.0616, -0.0526, -0.0357]], + device='cuda:0'), grad: tensor([[-2.0599e-03, 1.2434e-04, 5.8003e-06, ..., -7.3814e-03, + 8.3029e-05, -4.1084e-03], + [ 1.6248e-04, 3.4589e-06, 7.7412e-06, ..., 1.6680e-03, + 1.4156e-06, -6.4392e-03], + [ 1.8406e-03, 3.4362e-05, -8.0490e-03, ..., -7.1411e-03, + 4.9062e-06, -5.3291e-03], + ..., + [ 1.8749e-03, 1.1072e-05, 8.0032e-03, ..., 8.7128e-03, + 5.3272e-07, 4.4746e-03], + [ 5.8651e-04, 1.1003e-04, 4.2558e-05, ..., 1.9016e-03, + 4.2975e-05, 9.6941e-04], + [-7.3051e-04, -5.9724e-05, 3.1352e-05, ..., 2.0618e-03, + 1.9930e-06, 1.3132e-03]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0272, 0.0414, -0.0023, -0.0102, 0.0229, -0.0308, 0.0015, 0.0045, + -0.0039, -0.0149], device='cuda:0'), grad: tensor([-0.0459, -0.0159, -0.0390, 0.0071, 0.0386, -0.0382, 0.0154, 0.0570, + 0.0054, 0.0154], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 368---------------------------------------------------- +epoch 368, time 231.14, cls_loss 0.5215 cls_loss_mapping 0.0030 cls_loss_causal 0.4525 re_mapping 0.0090 re_causal 0.0211 /// teacc 99.01 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.0678, -0.1514, -0.1062, ..., -0.0373, 0.0621, -0.1098], + [-0.0804, -0.0935, -0.0858, ..., 0.1006, -0.0380, 0.2337], + [ 0.0019, -0.0275, -0.0526, ..., -0.0071, -0.0158, -0.0798], + ..., + [ 0.0406, -0.1198, 0.1763, ..., 0.0426, -0.1195, 0.0649], + [-0.0327, 0.0662, -0.1526, ..., -0.0537, 0.0177, -0.1130], + [-0.0614, 0.0781, 0.0448, ..., -0.0621, -0.0522, -0.0354]], + device='cuda:0'), grad: tensor([[ 1.1120e-03, 5.5879e-09, 6.6683e-07, ..., -1.3101e-04, + -2.9244e-06, -1.0918e-02], + [ 3.7980e-04, 9.3132e-09, 3.0510e-06, ..., -6.0892e-04, + 6.6683e-07, -2.7299e-04], + [-6.1836e-03, 8.1956e-08, 7.7561e-06, ..., -2.9221e-03, + 2.6636e-07, 1.0997e-04], + ..., + [ 1.4830e-04, 4.4703e-08, -8.5056e-05, ..., 1.2302e-03, + 3.3528e-08, 3.2043e-03], + [ 2.1112e-04, -4.2841e-08, 5.4426e-06, ..., 4.9263e-05, + 3.1106e-07, 1.2426e-03], + [ 9.0551e-04, 1.9372e-07, 4.2886e-05, ..., 1.6775e-03, + 3.3900e-07, 4.6806e-03]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0276, 0.0410, -0.0018, -0.0095, 0.0232, -0.0311, 0.0015, 0.0043, + -0.0039, -0.0151], device='cuda:0'), grad: tensor([ 0.0015, -0.0161, -0.0146, 0.0195, -0.0179, -0.0175, 0.0147, 0.0218, + -0.0159, 0.0245], device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 214.69, cls_loss 0.5130 cls_loss_mapping 0.0022 cls_loss_causal 0.4464 re_mapping 0.0088 re_causal 0.0209 /// teacc 98.88 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.0675, -0.1530, -0.1048, ..., -0.0363, 0.0622, -0.1070], + [-0.0814, -0.0939, -0.0861, ..., 0.1006, -0.0380, 0.2331], + [ 0.0023, -0.0274, -0.0536, ..., -0.0076, -0.0158, -0.0802], + ..., + [ 0.0398, -0.1198, 0.1769, ..., 0.0421, -0.1197, 0.0642], + [-0.0319, 0.0663, -0.1535, ..., -0.0535, 0.0187, -0.1138], + [-0.0614, 0.0789, 0.0452, ..., -0.0620, -0.0518, -0.0356]], + device='cuda:0'), grad: tensor([[ 9.0837e-04, 1.9050e-04, 3.2115e-04, ..., -1.6994e-03, + 2.6774e-04, 3.0231e-04], + [ 9.5272e-04, 1.2624e-04, 2.8467e-04, ..., 4.9305e-04, + 2.4028e-07, 2.9421e-04], + [ 1.5268e-03, 6.2418e-04, -1.1971e-02, ..., 2.9039e-04, + 1.0526e-04, 8.3733e-04], + ..., + [ 8.8263e-04, 3.2330e-04, 2.4414e-03, ..., 2.1038e-03, + 8.7544e-08, 2.1338e-04], + [-3.0766e-03, 1.4086e-03, 4.2915e-04, ..., -7.0763e-03, + 9.1553e-03, -2.2144e-03], + [ 1.6174e-03, -1.9989e-03, 2.5702e-04, ..., 1.9646e-03, + 1.6794e-05, 3.4785e-04]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0275, 0.0410, -0.0010, -0.0091, 0.0230, -0.0303, 0.0016, 0.0039, + -0.0045, -0.0158], device='cuda:0'), grad: tensor([-0.0215, 0.0120, -0.0027, 0.0064, 0.0125, 0.0117, -0.0195, 0.0161, + -0.0166, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 214.86, cls_loss 0.5178 cls_loss_mapping 0.0025 cls_loss_causal 0.4575 re_mapping 0.0090 re_causal 0.0214 /// teacc 98.91 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.0690, -0.1537, -0.1052, ..., -0.0359, 0.0620, -0.1074], + [-0.0811, -0.0941, -0.0863, ..., 0.1004, -0.0376, 0.2336], + [ 0.0018, -0.0287, -0.0543, ..., -0.0065, -0.0155, -0.0805], + ..., + [ 0.0417, -0.1203, 0.1767, ..., 0.0422, -0.1201, 0.0648], + [-0.0308, 0.0657, -0.1544, ..., -0.0530, 0.0177, -0.1146], + [-0.0623, 0.0802, 0.0456, ..., -0.0627, -0.0507, -0.0347]], + device='cuda:0'), grad: tensor([[ 3.3528e-05, 2.6751e-04, 1.8442e-04, ..., -3.6383e-04, + -8.6427e-06, 9.1314e-05], + [ 1.1787e-03, 8.9109e-05, 6.3801e-04, ..., 1.1589e-02, + 1.0990e-07, 6.1874e-03], + [ 7.5161e-05, 8.3625e-05, 4.6806e-03, ..., 2.7676e-03, + 4.4890e-07, 2.6932e-03], + ..., + [-1.9083e-03, 7.4148e-05, -2.2755e-03, ..., -5.6725e-03, + 5.0478e-07, 4.3793e-03], + [ 4.4972e-05, -1.5974e-03, 1.3363e-04, ..., -6.2943e-03, + 4.4890e-07, -3.8853e-03], + [ 5.8699e-04, 1.5545e-04, 3.9148e-04, ..., 2.0866e-03, + 6.2026e-06, -5.7030e-03]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0271, 0.0416, -0.0012, -0.0101, 0.0236, -0.0292, 0.0012, 0.0034, + -0.0055, -0.0154], device='cuda:0'), grad: tensor([-0.0026, 0.0424, -0.0493, 0.0265, -0.0138, 0.0317, 0.0496, -0.0291, + -0.0478, -0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 214.70, cls_loss 0.5205 cls_loss_mapping 0.0025 cls_loss_causal 0.4585 re_mapping 0.0085 re_causal 0.0211 /// teacc 98.81 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.0711, -0.1536, -0.1072, ..., -0.0364, 0.0608, -0.1071], + [-0.0803, -0.0930, -0.0867, ..., 0.0999, -0.0376, 0.2344], + [ 0.0028, -0.0296, -0.0533, ..., -0.0061, -0.0144, -0.0807], + ..., + [ 0.0417, -0.1202, 0.1758, ..., 0.0420, -0.1201, 0.0642], + [-0.0306, 0.0655, -0.1539, ..., -0.0521, 0.0178, -0.1144], + [-0.0624, 0.0805, 0.0462, ..., -0.0627, -0.0494, -0.0354]], + device='cuda:0'), grad: tensor([[ 3.2258e-04, 3.5733e-05, 6.4354e-07, ..., 1.5459e-03, + 8.1003e-05, 1.8775e-04], + [ 6.0844e-04, 3.0547e-05, 7.1339e-07, ..., 1.8072e-03, + 3.9071e-05, 8.4341e-05], + [ 4.3631e-04, 6.9916e-05, 2.3153e-06, ..., 5.9366e-04, + 2.4033e-04, 2.5296e-04], + ..., + [ 3.6764e-04, 9.2268e-05, 6.9700e-06, ..., 1.9979e-04, + 5.1588e-05, -1.0786e-03], + [ 1.0169e-04, -1.7614e-03, 1.1384e-05, ..., 1.9369e-03, + -2.1763e-03, -4.1604e-05], + [-3.0613e-03, -3.5584e-05, -9.0599e-05, ..., -2.3708e-03, + 6.8307e-05, 1.6570e-04]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0280, 0.0414, -0.0012, -0.0102, 0.0236, -0.0300, 0.0012, 0.0045, + -0.0048, -0.0154], device='cuda:0'), grad: tensor([ 0.0106, 0.0125, 0.0059, -0.0130, -0.0181, 0.0148, 0.0104, 0.0084, + -0.0122, -0.0191], device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 214.82, cls_loss 0.4915 cls_loss_mapping 0.0015 cls_loss_causal 0.4403 re_mapping 0.0089 re_causal 0.0211 /// teacc 98.86 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.0718, -0.1529, -0.1076, ..., -0.0361, 0.0606, -0.1076], + [-0.0797, -0.0925, -0.0870, ..., 0.0992, -0.0378, 0.2343], + [ 0.0027, -0.0298, -0.0543, ..., -0.0061, -0.0146, -0.0814], + ..., + [ 0.0407, -0.1208, 0.1768, ..., 0.0418, -0.1203, 0.0651], + [-0.0308, 0.0659, -0.1526, ..., -0.0526, 0.0191, -0.1139], + [-0.0622, 0.0805, 0.0454, ..., -0.0624, -0.0496, -0.0365]], + device='cuda:0'), grad: tensor([[-1.3940e-05, 4.8816e-05, 8.5890e-05, ..., 3.4642e-04, + -3.7575e-04, 7.6628e-04], + [-1.7989e-04, 1.7512e-04, 4.4847e-04, ..., -9.7466e-04, + 3.7961e-06, -2.1019e-03], + [ 1.1241e-04, 4.6206e-04, 8.1158e-04, ..., 2.9049e-03, + 1.7893e-04, 3.5715e-04], + ..., + [-2.9802e-04, 8.4019e-04, 1.1263e-03, ..., 3.7270e-03, + 6.7838e-06, 1.1015e-03], + [ 1.2374e-04, 2.9087e-05, 1.2290e-04, ..., 9.1696e-04, + 1.5259e-05, 2.8748e-02], + [-1.9252e-04, 1.6236e-04, 3.4142e-04, ..., -3.1605e-03, + 1.7107e-05, -3.3600e-02]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0275, 0.0414, -0.0013, -0.0103, 0.0230, -0.0295, 0.0013, 0.0038, + -0.0042, -0.0156], device='cuda:0'), grad: tensor([ 0.0131, -0.0102, 0.0265, 0.0005, -0.0069, -0.0158, 0.0224, -0.0054, + 0.0174, -0.0416], device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 214.79, cls_loss 0.5109 cls_loss_mapping 0.0025 cls_loss_causal 0.4517 re_mapping 0.0084 re_causal 0.0199 /// teacc 98.89 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.0720, -0.1528, -0.1077, ..., -0.0362, 0.0606, -0.1078], + [-0.0790, -0.0919, -0.0863, ..., 0.1001, -0.0385, 0.2344], + [ 0.0032, -0.0297, -0.0547, ..., -0.0059, -0.0150, -0.0811], + ..., + [ 0.0400, -0.1207, 0.1766, ..., 0.0408, -0.1208, 0.0648], + [-0.0293, 0.0662, -0.1531, ..., -0.0523, 0.0193, -0.1164], + [-0.0616, 0.0798, 0.0450, ..., -0.0621, -0.0497, -0.0354]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0007, 0.0002, ..., 0.0016, 0.0038, 0.0007], + [ 0.0015, 0.0009, 0.0003, ..., 0.0021, 0.0004, 0.0009], + [ 0.0021, 0.0014, 0.0003, ..., 0.0029, 0.0233, 0.0061], + ..., + [-0.0080, -0.0048, -0.0024, ..., -0.0075, -0.0019, -0.0075], + [ 0.0045, 0.0027, 0.0001, ..., 0.0047, 0.0029, 0.0011], + [ 0.0032, 0.0025, 0.0006, ..., 0.0042, 0.0005, 0.0027]], + device='cuda:0') +Epoch 375, bias, value: tensor([-0.0275, 0.0411, -0.0014, -0.0107, 0.0239, -0.0309, 0.0021, 0.0032, + -0.0035, -0.0152], device='cuda:0'), grad: tensor([ 0.0192, 0.0189, 0.0365, 0.0174, -0.0167, 0.0115, -0.0753, -0.0502, + 0.0040, 0.0346], device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 214.53, cls_loss 0.5199 cls_loss_mapping 0.0023 cls_loss_causal 0.4486 re_mapping 0.0086 re_causal 0.0213 /// teacc 98.76 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.0724, -0.1538, -0.1081, ..., -0.0363, 0.0607, -0.1059], + [-0.0799, -0.0924, -0.0872, ..., 0.0996, -0.0386, 0.2343], + [ 0.0030, -0.0298, -0.0550, ..., -0.0053, -0.0162, -0.0818], + ..., + [ 0.0399, -0.1209, 0.1767, ..., 0.0409, -0.1214, 0.0644], + [-0.0287, 0.0667, -0.1524, ..., -0.0528, 0.0192, -0.1180], + [-0.0629, 0.0813, 0.0446, ..., -0.0634, -0.0495, -0.0367]], + device='cuda:0'), grad: tensor([[-1.5526e-03, -5.1594e-04, -3.4285e-04, ..., -5.3787e-04, + -1.5569e-04, -2.3961e-04], + [ 2.0117e-05, 2.1088e-04, 3.6269e-05, ..., 5.5790e-04, + 2.9691e-06, 3.7766e-04], + [ 1.7834e-04, 2.3305e-04, 1.8322e-04, ..., 4.9305e-04, + 4.6372e-05, 2.8777e-04], + ..., + [-2.8953e-05, 7.0035e-05, -6.0320e-04, ..., 7.2956e-05, + 6.1095e-06, -1.0896e-04], + [ 2.3246e-04, 1.5430e-03, 3.0518e-04, ..., 3.5324e-03, + 2.5123e-05, 2.5234e-03], + [ 8.5473e-05, -4.1733e-03, 7.7486e-05, ..., 6.1846e-04, + 1.5095e-05, 4.1747e-04]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0278, 0.0417, -0.0012, -0.0106, 0.0230, -0.0312, 0.0032, 0.0034, + -0.0033, -0.0160], device='cuda:0'), grad: tensor([ 0.0006, 0.0036, 0.0173, -0.0281, 0.0215, 0.0140, -0.0131, -0.0165, + 0.0133, -0.0127], device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 214.66, cls_loss 0.5050 cls_loss_mapping 0.0046 cls_loss_causal 0.4531 re_mapping 0.0086 re_causal 0.0205 /// teacc 98.88 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.0723, -0.1537, -0.1084, ..., -0.0364, 0.0597, -0.1055], + [-0.0798, -0.0917, -0.0881, ..., 0.0998, -0.0397, 0.2336], + [ 0.0034, -0.0295, -0.0557, ..., -0.0059, -0.0128, -0.0817], + ..., + [ 0.0397, -0.1208, 0.1784, ..., 0.0413, -0.1218, 0.0646], + [-0.0292, 0.0663, -0.1521, ..., -0.0527, 0.0183, -0.1187], + [-0.0622, 0.0814, 0.0436, ..., -0.0639, -0.0499, -0.0358]], + device='cuda:0'), grad: tensor([[ 1.9586e-04, 1.1480e-04, 1.5163e-04, ..., 1.0996e-03, + 3.8385e-05, 4.6563e-04], + [ 8.7619e-05, 3.5793e-05, 4.6194e-05, ..., -3.7041e-03, + 4.0770e-05, -2.1782e-03], + [-1.2798e-03, 2.4271e-04, 1.2290e-04, ..., -1.4048e-03, + -4.4584e-04, 7.1108e-05], + ..., + [ 1.0490e-04, 2.0730e-04, -7.1526e-04, ..., 1.8911e-03, + 2.9534e-05, -3.4642e-04], + [ 6.7174e-05, -4.6015e-04, 1.8859e-04, ..., 1.4391e-03, + 4.0121e-06, 5.0497e-04], + [ 2.1362e-04, 2.3186e-04, 4.7016e-04, ..., 1.1759e-03, + 4.0323e-05, 3.0828e-04]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0279, 0.0425, -0.0016, -0.0104, 0.0229, -0.0307, 0.0025, 0.0041, + -0.0043, -0.0160], device='cuda:0'), grad: tensor([ 0.0077, 0.0046, -0.0399, -0.0366, 0.0226, 0.0217, -0.0183, 0.0157, + 0.0092, 0.0133], device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 214.64, cls_loss 0.5240 cls_loss_mapping 0.0024 cls_loss_causal 0.4583 re_mapping 0.0088 re_causal 0.0217 /// teacc 98.79 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.0728, -0.1535, -0.1098, ..., -0.0375, 0.0587, -0.1059], + [-0.0810, -0.0920, -0.0899, ..., 0.1000, -0.0405, 0.2344], + [ 0.0035, -0.0304, -0.0566, ..., -0.0065, -0.0128, -0.0823], + ..., + [ 0.0395, -0.1202, 0.1782, ..., 0.0404, -0.1215, 0.0648], + [-0.0292, 0.0661, -0.1527, ..., -0.0523, 0.0193, -0.1182], + [-0.0625, 0.0813, 0.0432, ..., -0.0623, -0.0501, -0.0366]], + device='cuda:0'), grad: tensor([[-1.5411e-03, -3.9139e-03, 2.4006e-05, ..., -5.8899e-03, + 5.2404e-04, -1.2482e-02], + [ 2.3041e-03, 3.2131e-06, -4.6760e-05, ..., -3.4485e-03, + 9.1732e-05, 8.2550e-03], + [-4.2877e-03, 9.5904e-05, -6.8665e-04, ..., 6.9618e-03, + 5.5701e-05, -4.5180e-04], + ..., + [ 5.7316e-04, 5.1200e-05, 4.7493e-04, ..., -6.0501e-03, + -1.9398e-03, 1.0490e-03], + [ 2.5630e-04, 7.3195e-04, -1.2174e-05, ..., 1.7376e-03, + 7.2181e-05, 8.2445e-04], + [ 4.8470e-04, 2.9850e-03, 3.2753e-05, ..., -5.0964e-03, + 5.4806e-05, 2.0981e-05]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0283, 0.0420, -0.0024, -0.0110, 0.0226, -0.0297, 0.0022, 0.0048, + -0.0032, -0.0159], device='cuda:0'), grad: tensor([-0.0920, 0.0159, 0.0355, 0.0162, 0.0305, -0.0076, 0.0400, -0.0852, + 0.0317, 0.0149], device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 214.75, cls_loss 0.5187 cls_loss_mapping 0.0028 cls_loss_causal 0.4612 re_mapping 0.0085 re_causal 0.0202 /// teacc 98.87 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.0722, -0.1529, -0.1113, ..., -0.0378, 0.0593, -0.1056], + [-0.0809, -0.0934, -0.0911, ..., 0.1000, -0.0411, 0.2346], + [ 0.0045, -0.0303, -0.0577, ..., -0.0064, -0.0132, -0.0812], + ..., + [ 0.0398, -0.1214, 0.1790, ..., 0.0415, -0.1226, 0.0641], + [-0.0291, 0.0650, -0.1515, ..., -0.0524, 0.0191, -0.1184], + [-0.0630, 0.0820, 0.0422, ..., -0.0628, -0.0503, -0.0356]], + device='cuda:0'), grad: tensor([[ 7.2360e-05, 2.6077e-07, 3.8184e-08, ..., 4.5896e-04, + -1.5097e-03, 1.6842e-03], + [ 5.3024e-04, -9.6917e-05, -7.1190e-06, ..., 3.1643e-03, + 8.8751e-05, 1.5202e-03], + [ 2.4962e-04, 2.8722e-06, 4.8764e-06, ..., 2.4567e-03, + 9.7811e-05, 1.2064e-03], + ..., + [ 3.9649e-04, 5.5023e-06, -1.0139e-04, ..., 3.3073e-03, + 1.7488e-04, 1.5850e-03], + [-4.3259e-03, 6.5327e-05, 4.6045e-05, ..., 1.3428e-03, + 8.2791e-05, 1.1616e-03], + [ 4.6825e-04, 5.3942e-05, 8.9481e-06, ..., 2.8133e-03, + 2.1899e-04, 1.3990e-03]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0289, 0.0423, -0.0020, -0.0105, 0.0218, -0.0294, 0.0022, 0.0062, + -0.0048, -0.0160], device='cuda:0'), grad: tensor([-0.0108, 0.0199, 0.0170, -0.0114, 0.0015, -0.0082, -0.0018, 0.0206, + -0.0376, 0.0108], device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 214.79, cls_loss 0.5364 cls_loss_mapping 0.0031 cls_loss_causal 0.4723 re_mapping 0.0087 re_causal 0.0212 /// teacc 98.83 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.0733, -0.1540, -0.1118, ..., -0.0391, 0.0600, -0.1057], + [-0.0799, -0.0930, -0.0898, ..., 0.1000, -0.0419, 0.2344], + [ 0.0047, -0.0300, -0.0574, ..., -0.0066, -0.0133, -0.0821], + ..., + [ 0.0398, -0.1222, 0.1795, ..., 0.0413, -0.1225, 0.0643], + [-0.0277, 0.0654, -0.1507, ..., -0.0511, 0.0194, -0.1188], + [-0.0634, 0.0814, 0.0413, ..., -0.0624, -0.0513, -0.0353]], + device='cuda:0'), grad: tensor([[ 2.8229e-04, 1.7309e-04, -4.1428e-03, ..., -4.7455e-03, + 6.0511e-04, 4.8208e-04], + [-4.5013e-03, 2.3329e-04, -5.2929e-04, ..., 3.5095e-03, + 1.1355e-05, 4.0398e-03], + [ 3.8934e-04, -2.1130e-05, 6.6423e-04, ..., 1.4296e-03, + 2.5436e-05, 5.9986e-04], + ..., + [ 1.7033e-03, 2.3115e-04, 1.0033e-03, ..., -2.5578e-03, + 3.5048e-05, 5.2977e-04], + [ 5.8365e-04, -3.8838e-04, 5.1165e-04, ..., 6.4516e-04, + 4.9382e-05, 1.6749e-04], + [ 9.9754e-04, 2.0897e-04, 1.0900e-03, ..., 2.0447e-03, + 4.4155e-04, 1.2503e-03]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0295, 0.0419, -0.0021, -0.0094, 0.0218, -0.0298, 0.0024, 0.0057, + -0.0040, -0.0162], device='cuda:0'), grad: tensor([ 0.0126, 0.0095, 0.0173, 0.0285, -0.0129, 0.0189, -0.0698, 0.0070, + -0.0043, -0.0066], device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 214.58, cls_loss 0.5259 cls_loss_mapping 0.0027 cls_loss_causal 0.4594 re_mapping 0.0083 re_causal 0.0202 /// teacc 98.85 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.0733, -0.1544, -0.1123, ..., -0.0385, 0.0604, -0.1058], + [-0.0805, -0.0932, -0.0908, ..., 0.1001, -0.0421, 0.2338], + [ 0.0053, -0.0303, -0.0562, ..., -0.0063, -0.0128, -0.0813], + ..., + [ 0.0391, -0.1233, 0.1793, ..., 0.0411, -0.1224, 0.0635], + [-0.0280, 0.0657, -0.1498, ..., -0.0499, 0.0195, -0.1181], + [-0.0616, 0.0822, 0.0396, ..., -0.0630, -0.0512, -0.0353]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.4404e-04, 1.2014e-06, ..., -2.1954e-03, + 0.0000e+00, 3.4068e-06], + [ 0.0000e+00, 3.7527e-04, 2.0154e-06, ..., 6.6233e-04, + 0.0000e+00, -3.8385e-05], + [ 2.0489e-08, 5.0592e-04, 5.7891e-06, ..., 7.1192e-04, + 0.0000e+00, 3.3882e-06], + ..., + [ 9.3132e-10, 3.2568e-04, -6.2305e-07, ..., 4.2295e-04, + 0.0000e+00, -7.8604e-06], + [-9.4064e-08, -1.3733e-03, 2.2985e-06, ..., -1.6661e-03, + 0.0000e+00, 2.5313e-06], + [ 2.7940e-09, -2.9230e-04, 1.0924e-06, ..., 5.6410e-04, + 0.0000e+00, 4.8988e-06]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0290, 0.0412, -0.0016, -0.0102, 0.0219, -0.0311, 0.0030, 0.0056, + -0.0032, -0.0158], device='cuda:0'), grad: tensor([-0.0596, -0.0102, 0.0238, 0.0227, 0.0220, -0.0589, 0.0340, 0.0220, + -0.0072, 0.0113], device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 214.88, cls_loss 0.5105 cls_loss_mapping 0.0027 cls_loss_causal 0.4426 re_mapping 0.0088 re_causal 0.0210 /// teacc 98.86 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.0730, -0.1551, -0.1115, ..., -0.0385, 0.0592, -0.1058], + [-0.0791, -0.0913, -0.0899, ..., 0.1016, -0.0415, 0.2346], + [ 0.0048, -0.0303, -0.0570, ..., -0.0061, -0.0129, -0.0817], + ..., + [ 0.0381, -0.1223, 0.1789, ..., 0.0407, -0.1225, 0.0616], + [-0.0276, 0.0652, -0.1502, ..., -0.0508, 0.0196, -0.1189], + [-0.0611, 0.0833, 0.0402, ..., -0.0628, -0.0517, -0.0344]], + device='cuda:0'), grad: tensor([[ 1.0118e-03, 9.9912e-06, 1.0155e-05, ..., 2.5959e-03, + 9.7137e-07, 1.5283e-04], + [ 9.5415e-04, 5.7787e-05, 6.0052e-05, ..., 2.5349e-03, + 3.3528e-08, 2.8205e-04], + [ 1.5211e-03, 2.8517e-06, 3.0100e-05, ..., 2.9774e-03, + 2.3190e-07, 3.4308e-04], + ..., + [ 1.5163e-03, 1.9461e-05, -3.7730e-05, ..., 1.4710e-04, + 8.0094e-08, -1.4293e-04], + [ 8.9550e-04, 1.0132e-02, 1.5512e-05, ..., 2.0351e-03, + 2.4643e-06, 2.0707e-04], + [ 7.0429e-04, 4.3780e-05, 2.9826e-04, ..., -8.6689e-04, + 4.1537e-07, 8.8453e-04]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0293, 0.0419, -0.0008, -0.0101, 0.0212, -0.0318, 0.0032, 0.0062, + -0.0047, -0.0149], device='cuda:0'), grad: tensor([ 0.0157, 0.0145, -0.0024, -0.0122, -0.0237, 0.0124, 0.0061, -0.0142, + 0.0172, -0.0134], device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 214.92, cls_loss 0.4762 cls_loss_mapping 0.0020 cls_loss_causal 0.4151 re_mapping 0.0085 re_causal 0.0202 /// teacc 98.83 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.0728, -0.1547, -0.1127, ..., -0.0374, 0.0590, -0.1060], + [-0.0801, -0.0930, -0.0886, ..., 0.1005, -0.0406, 0.2356], + [ 0.0041, -0.0282, -0.0594, ..., -0.0065, -0.0130, -0.0828], + ..., + [ 0.0368, -0.1214, 0.1794, ..., 0.0405, -0.1229, 0.0615], + [-0.0285, 0.0643, -0.1491, ..., -0.0508, 0.0196, -0.1194], + [-0.0616, 0.0833, 0.0409, ..., -0.0626, -0.0520, -0.0343]], + device='cuda:0'), grad: tensor([[-2.7180e-04, 3.1376e-04, 3.2131e-06, ..., 1.3332e-03, + 8.3148e-05, 3.9029e-04], + [ 2.5487e-04, 1.9431e-05, -3.3021e-05, ..., -2.2926e-03, + 2.3633e-05, -3.1757e-03], + [ 6.0892e-04, 2.5964e-04, 4.9211e-06, ..., 2.2850e-03, + 1.1164e-04, 1.5986e-04], + ..., + [-2.0099e-04, 2.4843e-04, -4.6939e-05, ..., -2.9850e-03, + -4.1437e-04, 3.8123e-04], + [ 3.4428e-04, 7.8630e-04, 6.3106e-06, ..., 2.0943e-03, + 1.7852e-05, 4.8876e-04], + [ 3.5858e-04, 3.5763e-04, 3.9160e-05, ..., 1.1425e-03, + 2.7478e-05, 4.0340e-04]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0287, 0.0421, -0.0013, -0.0102, 0.0217, -0.0317, 0.0029, 0.0060, + -0.0049, -0.0150], device='cuda:0'), grad: tensor([ 0.0200, -0.0384, 0.0301, -0.0416, -0.0053, -0.0039, 0.0294, -0.0097, + 0.0278, -0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 214.86, cls_loss 0.5252 cls_loss_mapping 0.0023 cls_loss_causal 0.4561 re_mapping 0.0083 re_causal 0.0204 /// teacc 98.79 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.0732, -0.1555, -0.1117, ..., -0.0381, 0.0592, -0.1061], + [-0.0816, -0.0943, -0.0891, ..., 0.0993, -0.0422, 0.2355], + [ 0.0038, -0.0287, -0.0598, ..., -0.0081, -0.0127, -0.0836], + ..., + [ 0.0370, -0.1214, 0.1799, ..., 0.0407, -0.1225, 0.0619], + [-0.0296, 0.0640, -0.1492, ..., -0.0512, 0.0194, -0.1195], + [-0.0604, 0.0836, 0.0409, ..., -0.0609, -0.0519, -0.0343]], + device='cuda:0'), grad: tensor([[ 1.2589e-04, 1.9148e-05, 4.1395e-05, ..., -1.5926e-03, + 2.4307e-04, 8.6904e-05], + [ 4.6325e-04, -2.9251e-05, 2.0123e-04, ..., -1.8597e-03, + 8.1635e-04, -3.6836e-04], + [ 8.5115e-04, -2.1160e-04, 3.1739e-05, ..., 3.0918e-03, + 3.0947e-04, 3.8433e-04], + ..., + [ 7.6103e-04, 1.3977e-05, 5.2065e-05, ..., -3.3302e-03, + 1.5581e-04, -3.2997e-03], + [-5.4512e-03, 1.1501e-03, -9.6798e-04, ..., -6.8588e-03, + -2.9316e-03, 9.3985e-04], + [ 4.8184e-04, 7.8455e-06, 1.6272e-04, ..., 3.4695e-03, + 3.4332e-04, -1.1861e-04]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0291, 0.0418, -0.0018, -0.0101, 0.0211, -0.0317, 0.0030, 0.0071, + -0.0051, -0.0144], device='cuda:0'), grad: tensor([-0.0166, -0.0278, 0.0153, 0.0170, 0.0227, -0.0145, 0.0179, 0.0011, + -0.0304, 0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 214.54, cls_loss 0.5078 cls_loss_mapping 0.0026 cls_loss_causal 0.4481 re_mapping 0.0083 re_causal 0.0195 /// teacc 98.73 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.0734, -0.1572, -0.1121, ..., -0.0378, 0.0599, -0.1060], + [-0.0819, -0.0939, -0.0889, ..., 0.0968, -0.0422, 0.2343], + [ 0.0038, -0.0282, -0.0599, ..., -0.0071, -0.0131, -0.0850], + ..., + [ 0.0370, -0.1209, 0.1813, ..., 0.0412, -0.1222, 0.0631], + [-0.0292, 0.0643, -0.1511, ..., -0.0516, 0.0194, -0.1184], + [-0.0613, 0.0836, 0.0410, ..., -0.0618, -0.0523, -0.0355]], + device='cuda:0'), grad: tensor([[ 2.5463e-04, 2.3797e-05, 3.7730e-05, ..., 2.9011e-03, + 2.0218e-04, 1.9398e-03], + [ 1.8179e-04, 6.8665e-05, 1.0580e-04, ..., 5.9700e-03, + -2.2812e-03, 5.8899e-03], + [ 2.8920e-04, 5.3018e-05, 8.2552e-05, ..., -1.8988e-03, + 5.8937e-04, -2.8286e-03], + ..., + [ 2.7442e-04, 1.1415e-03, 1.7834e-03, ..., 3.7117e-03, + 2.6417e-04, 5.2528e-03], + [-8.8120e-04, 8.0049e-05, 1.2326e-04, ..., -1.6220e-02, + 2.0468e-04, -9.5749e-03], + [ 3.0303e-04, -2.3365e-03, -3.6316e-03, ..., 2.0180e-03, + 3.4642e-04, 3.8815e-03]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0298, 0.0412, -0.0012, -0.0113, 0.0217, -0.0303, 0.0029, 0.0072, + -0.0047, -0.0149], device='cuda:0'), grad: tensor([ 0.0171, 0.0233, -0.0124, 0.0172, -0.0085, -0.0093, 0.0331, 0.0303, + -0.0461, -0.0446], device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 214.82, cls_loss 0.5528 cls_loss_mapping 0.0026 cls_loss_causal 0.4898 re_mapping 0.0085 re_causal 0.0208 /// teacc 98.83 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.0730, -0.1568, -0.1126, ..., -0.0389, 0.0591, -0.1062], + [-0.0829, -0.0928, -0.0889, ..., 0.0963, -0.0418, 0.2346], + [ 0.0033, -0.0284, -0.0597, ..., -0.0061, -0.0141, -0.0845], + ..., + [ 0.0383, -0.1200, 0.1812, ..., 0.0408, -0.1218, 0.0618], + [-0.0278, 0.0634, -0.1515, ..., -0.0508, 0.0197, -0.1191], + [-0.0620, 0.0832, 0.0412, ..., -0.0622, -0.0527, -0.0335]], + device='cuda:0'), grad: tensor([[ 2.9683e-04, 4.6730e-04, 5.0402e-04, ..., 3.2139e-03, + 5.8699e-04, 1.2751e-03], + [-2.5082e-03, -1.3456e-03, 6.6936e-05, ..., -2.9716e-03, + -1.6642e-03, -7.6180e-03], + [ 2.4021e-04, 1.2711e-02, -8.8120e-03, ..., -1.3962e-02, + 5.6410e-04, -1.6220e-02], + ..., + [ 1.6403e-04, 1.4277e-03, 8.6899e-03, ..., 2.2430e-02, + 3.0351e-04, 1.7212e-02], + [ 9.1553e-04, 4.7455e-03, 1.5771e-04, ..., 8.4305e-03, + 4.4250e-03, 1.1620e-02], + [-4.6654e-03, -7.1945e-03, -1.8997e-03, ..., -1.8646e-02, + 3.6573e-04, -1.4465e-02]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0290, 0.0411, -0.0017, -0.0118, 0.0229, -0.0307, 0.0017, 0.0076, + -0.0042, -0.0148], device='cuda:0'), grad: tensor([ 0.0303, -0.0287, -0.0479, -0.0113, 0.0495, 0.0281, -0.0261, 0.0388, + 0.0815, -0.1141], device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 215.05, cls_loss 0.5121 cls_loss_mapping 0.0015 cls_loss_causal 0.4435 re_mapping 0.0085 re_causal 0.0202 /// teacc 98.87 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.0721, -0.1563, -0.1133, ..., -0.0391, 0.0606, -0.1070], + [-0.0818, -0.0940, -0.0893, ..., 0.0962, -0.0422, 0.2348], + [ 0.0033, -0.0287, -0.0580, ..., -0.0062, -0.0145, -0.0838], + ..., + [ 0.0379, -0.1213, 0.1809, ..., 0.0408, -0.1216, 0.0621], + [-0.0280, 0.0637, -0.1518, ..., -0.0513, 0.0201, -0.1211], + [-0.0618, 0.0841, 0.0408, ..., -0.0608, -0.0523, -0.0327]], + device='cuda:0'), grad: tensor([[-1.3208e-03, 1.2803e-04, 3.5691e-04, ..., -3.1109e-03, + -6.8724e-05, -1.7881e-07], + [ 2.7905e-03, 4.9543e-04, 3.4630e-05, ..., 6.1035e-03, + 2.5928e-06, -5.3085e-06], + [ 8.6737e-04, 1.9276e-04, 8.6975e-04, ..., 1.0490e-02, + -1.7853e-06, 3.0965e-05], + ..., + [ 2.4486e-04, 1.0502e-04, 1.6904e-04, ..., 2.1057e-03, + 2.9709e-07, -2.1681e-05], + [ 8.2245e-03, 1.1486e-04, 1.6057e-04, ..., -1.0399e-02, + 6.1654e-07, 1.5562e-06], + [ 2.9874e-04, 1.4186e-04, 1.7047e-04, ..., 1.8492e-03, + 1.5059e-06, -1.0002e-04]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0288, 0.0410, -0.0011, -0.0105, 0.0206, -0.0314, 0.0030, 0.0075, + -0.0046, -0.0149], device='cuda:0'), grad: tensor([-0.0127, 0.0293, 0.0232, -0.0243, -0.0408, -0.0006, -0.0126, 0.0119, + 0.0160, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 214.32, cls_loss 0.5403 cls_loss_mapping 0.0026 cls_loss_causal 0.4889 re_mapping 0.0078 re_causal 0.0194 /// teacc 98.88 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.0737, -0.1569, -0.1121, ..., -0.0403, 0.0619, -0.1073], + [-0.0813, -0.0937, -0.0895, ..., 0.0973, -0.0409, 0.2347], + [ 0.0047, -0.0283, -0.0570, ..., -0.0056, -0.0128, -0.0838], + ..., + [ 0.0374, -0.1217, 0.1798, ..., 0.0395, -0.1217, 0.0613], + [-0.0272, 0.0650, -0.1512, ..., -0.0493, 0.0202, -0.1204], + [-0.0630, 0.0834, 0.0422, ..., -0.0619, -0.0550, -0.0320]], + device='cuda:0'), grad: tensor([[ 2.5487e-04, 8.0061e-04, 2.6539e-05, ..., 2.0981e-03, + -1.1435e-03, 1.0717e-04], + [ 1.1225e-03, 2.0135e-06, 1.9044e-05, ..., 1.2703e-03, + 1.4508e-04, -1.5316e-03], + [ 4.7493e-04, 1.3277e-05, 1.1241e-04, ..., 2.5711e-03, + 1.3626e-04, 1.8597e-04], + ..., + [-5.2261e-03, 4.3541e-05, -4.7607e-03, ..., -1.6876e-02, + -1.2884e-03, -1.2712e-03], + [ 5.3787e-04, 5.1212e-04, 1.3900e-04, ..., 2.2583e-03, + 1.8764e-04, 2.7585e-04], + [ 1.0042e-03, 1.5564e-03, 4.3640e-03, ..., 2.4548e-03, + 2.6560e-04, 1.4849e-03]], device='cuda:0') +Epoch 388, bias, value: tensor([-2.8841e-02, 4.2452e-02, -6.8343e-05, -1.0974e-02, 1.9718e-02, + -3.1884e-02, 2.9449e-03, 6.6885e-03, -3.3665e-03, -1.5919e-02], + device='cuda:0'), grad: tensor([ 1.0574e-02, 1.4023e-02, 1.5518e-02, -4.5410e-02, -1.6968e-02, + 3.4668e-02, 1.9180e-02, -5.1361e-02, 1.9669e-02, 7.4625e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 214.32, cls_loss 0.5148 cls_loss_mapping 0.0017 cls_loss_causal 0.4388 re_mapping 0.0082 re_causal 0.0203 /// teacc 98.84 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.0727, -0.1576, -0.1103, ..., -0.0398, 0.0631, -0.1074], + [-0.0814, -0.0930, -0.0894, ..., 0.0984, -0.0416, 0.2357], + [ 0.0036, -0.0289, -0.0580, ..., -0.0065, -0.0121, -0.0854], + ..., + [ 0.0374, -0.1226, 0.1804, ..., 0.0402, -0.1226, 0.0617], + [-0.0268, 0.0655, -0.1521, ..., -0.0498, 0.0197, -0.1208], + [-0.0633, 0.0840, 0.0408, ..., -0.0627, -0.0552, -0.0331]], + device='cuda:0'), grad: tensor([[ 2.3752e-05, 1.2884e-03, 9.7573e-05, ..., 6.9427e-03, + -1.2531e-03, 2.2805e-04], + [ 3.6597e-05, 4.4495e-05, 2.0752e-03, ..., 2.0733e-03, + 1.9693e-04, -1.4048e-03], + [-8.2374e-05, 9.3365e-04, 3.5152e-03, ..., 9.8877e-03, + 7.0047e-04, 4.3678e-03], + ..., + [ 3.7104e-05, 3.6489e-06, -6.0806e-03, ..., -1.2405e-02, + 1.6189e-04, -5.8823e-03], + [ 5.7995e-05, 1.7703e-04, 1.4973e-04, ..., 8.1730e-04, + 2.1458e-04, 8.8930e-05], + [ 1.0110e-05, 4.5002e-05, 2.5272e-04, ..., 4.7874e-04, + 1.6117e-04, 4.6182e-04]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0294, 0.0427, -0.0007, -0.0100, 0.0200, -0.0306, 0.0022, 0.0062, + -0.0028, -0.0167], device='cuda:0'), grad: tensor([-0.0064, 0.0088, 0.0182, -0.0493, 0.0061, 0.0066, 0.0082, -0.0080, + 0.0076, 0.0082], device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 214.21, cls_loss 0.5304 cls_loss_mapping 0.0043 cls_loss_causal 0.4642 re_mapping 0.0081 re_causal 0.0198 /// teacc 98.79 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.0729, -0.1578, -0.1091, ..., -0.0390, 0.0640, -0.1069], + [-0.0812, -0.0939, -0.0897, ..., 0.0985, -0.0425, 0.2366], + [ 0.0031, -0.0286, -0.0581, ..., -0.0064, -0.0122, -0.0854], + ..., + [ 0.0374, -0.1224, 0.1811, ..., 0.0395, -0.1230, 0.0607], + [-0.0271, 0.0641, -0.1528, ..., -0.0503, 0.0182, -0.1209], + [-0.0630, 0.0846, 0.0408, ..., -0.0630, -0.0543, -0.0326]], + device='cuda:0'), grad: tensor([[ 1.7767e-03, 4.1686e-06, 8.4686e-04, ..., 4.2496e-03, + 2.6431e-06, 4.9820e-03], + [ 2.5539e-03, 5.1677e-05, 7.5436e-04, ..., 6.8054e-03, + 1.0245e-08, 3.5629e-03], + [ 1.3924e-03, 8.3494e-04, 3.5248e-03, ..., 9.5596e-03, + 5.2154e-08, 1.0414e-03], + ..., + [-3.5381e-03, 4.7612e-04, 1.5736e-04, ..., -1.1703e-02, + 9.3132e-10, -1.0071e-02], + [ 8.9884e-04, 3.1799e-05, 5.5790e-04, ..., 1.7319e-03, + 5.7835e-07, 7.7343e-04], + [ 1.6193e-03, 1.4208e-05, 6.9618e-05, ..., 5.5456e-04, + 5.4017e-08, 3.7551e-04]], device='cuda:0') +Epoch 390, bias, value: tensor([-3.0347e-02, 4.2891e-02, 4.1757e-05, -1.0039e-02, 2.0394e-02, + -3.1797e-02, 2.1579e-03, 6.5624e-03, -2.6145e-03, -1.6630e-02], + device='cuda:0'), grad: tensor([ 0.0360, 0.0287, 0.0436, -0.0231, -0.0665, -0.0083, -0.0407, -0.0272, + 0.0313, 0.0261], device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 214.26, cls_loss 0.5398 cls_loss_mapping 0.0030 cls_loss_causal 0.4760 re_mapping 0.0083 re_causal 0.0198 /// teacc 98.75 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.0725, -0.1575, -0.1101, ..., -0.0385, 0.0645, -0.1068], + [-0.0805, -0.0939, -0.0888, ..., 0.0989, -0.0415, 0.2369], + [ 0.0032, -0.0296, -0.0591, ..., -0.0064, -0.0129, -0.0855], + ..., + [ 0.0369, -0.1195, 0.1821, ..., 0.0397, -0.1226, 0.0601], + [-0.0283, 0.0639, -0.1529, ..., -0.0509, 0.0196, -0.1204], + [-0.0630, 0.0833, 0.0409, ..., -0.0623, -0.0560, -0.0324]], + device='cuda:0'), grad: tensor([[ 7.8535e-04, 1.0476e-05, 2.5487e-04, ..., 2.6512e-03, + 6.7377e-04, 2.4376e-03], + [ 7.3624e-04, 2.0847e-05, 4.3958e-05, ..., -1.8950e-03, + 1.6189e-04, -1.1911e-03], + [ 2.6417e-03, 1.3418e-05, 3.9744e-04, ..., 5.7640e-03, + 3.9062e-03, 2.0385e-04], + ..., + [-1.1759e-03, 1.5938e-04, 1.1005e-03, ..., -9.3842e-03, + -5.4588e-03, -4.0855e-03], + [-6.0501e-03, 5.0583e-03, 3.0365e-03, ..., -2.1267e-03, + -2.6093e-03, 3.9554e-04], + [-1.9741e-03, 1.1599e-04, -2.0027e-03, ..., -7.8344e-04, + 2.6441e-04, 1.7971e-05]], device='cuda:0') +Epoch 391, bias, value: tensor([-3.0113e-02, 4.2971e-02, -3.3338e-05, -1.0680e-02, 2.1779e-02, + -3.2027e-02, 1.9074e-03, 6.7073e-03, -3.0853e-03, -1.6813e-02], + device='cuda:0'), grad: tensor([ 0.0139, -0.0182, 0.0246, 0.0246, 0.0182, 0.0389, 0.0191, -0.0541, + -0.0280, -0.0390], device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 214.59, cls_loss 0.5103 cls_loss_mapping 0.0029 cls_loss_causal 0.4399 re_mapping 0.0087 re_causal 0.0198 /// teacc 98.72 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.0733, -0.1569, -0.1105, ..., -0.0382, 0.0648, -0.1081], + [-0.0806, -0.0943, -0.0896, ..., 0.0986, -0.0400, 0.2370], + [ 0.0028, -0.0306, -0.0596, ..., -0.0068, -0.0137, -0.0847], + ..., + [ 0.0360, -0.1191, 0.1821, ..., 0.0396, -0.1222, 0.0607], + [-0.0284, 0.0652, -0.1525, ..., -0.0514, 0.0197, -0.1213], + [-0.0614, 0.0832, 0.0413, ..., -0.0619, -0.0567, -0.0324]], + device='cuda:0'), grad: tensor([[ 1.2083e-03, 1.7118e-04, 8.7440e-05, ..., 1.0834e-03, + 2.6112e-03, 6.7711e-04], + [ 4.1103e-04, 3.2711e-04, 1.0855e-05, ..., 4.8351e-04, + 1.1224e-04, -1.3673e-04], + [ 1.2839e-04, 1.5235e-04, 2.6211e-05, ..., -4.0932e-03, + 1.2922e-04, -1.4715e-03], + ..., + [ 2.3413e-04, 2.5797e-04, 1.8358e-04, ..., 5.3072e-04, + 9.2924e-05, 5.5122e-04], + [-1.7014e-03, 2.9083e-02, 2.7299e-04, ..., -2.2449e-03, + 2.3091e-04, -3.1338e-03], + [ 1.6727e-03, 2.2173e-04, 2.5845e-03, ..., 5.8594e-03, + 2.4247e-04, 2.3937e-03]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0310, 0.0420, -0.0009, -0.0109, 0.0226, -0.0316, 0.0014, 0.0077, + -0.0026, -0.0162], device='cuda:0'), grad: tensor([ 0.0248, -0.0138, -0.0434, 0.0195, -0.0097, 0.0142, -0.0314, 0.0166, + -0.0145, 0.0377], device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 214.58, cls_loss 0.5168 cls_loss_mapping 0.0030 cls_loss_causal 0.4605 re_mapping 0.0085 re_causal 0.0205 /// teacc 98.84 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.0744, -0.1585, -0.1113, ..., -0.0384, 0.0645, -0.1084], + [-0.0812, -0.0949, -0.0900, ..., 0.0979, -0.0388, 0.2365], + [ 0.0021, -0.0303, -0.0591, ..., -0.0069, -0.0131, -0.0858], + ..., + [ 0.0385, -0.1189, 0.1817, ..., 0.0399, -0.1237, 0.0628], + [-0.0287, 0.0655, -0.1528, ..., -0.0502, 0.0206, -0.1213], + [-0.0618, 0.0835, 0.0419, ..., -0.0623, -0.0560, -0.0336]], + device='cuda:0'), grad: tensor([[ 5.1297e-06, 4.0203e-05, 4.6968e-05, ..., 3.2997e-04, + 1.3161e-03, 2.3508e-04], + [ 6.8620e-06, 1.0246e-04, 3.9756e-05, ..., 4.1223e-04, + 1.2560e-03, 2.5749e-04], + [-3.6269e-05, 4.3094e-05, -4.8522e-07, ..., 4.2295e-04, + 1.9197e-03, 3.2306e-04], + ..., + [ 5.5939e-05, 2.4462e-04, 4.1676e-04, ..., -2.1255e-04, + 1.1301e-03, 1.3962e-03], + [ 3.8743e-06, 1.1975e-04, 5.1320e-05, ..., 4.9067e-04, + -1.3100e-02, 2.0885e-03], + [ 1.8418e-05, 3.7670e-04, -8.0299e-04, ..., -1.4000e-03, + 1.5068e-03, -1.7710e-03]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0324, 0.0410, -0.0007, -0.0110, 0.0226, -0.0313, 0.0021, 0.0081, + -0.0029, -0.0151], device='cuda:0'), grad: tensor([ 0.0144, 0.0197, 0.0244, 0.0210, -0.0351, 0.0249, 0.0215, -0.0322, + -0.0299, -0.0287], device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 214.49, cls_loss 0.5043 cls_loss_mapping 0.0026 cls_loss_causal 0.4454 re_mapping 0.0081 re_causal 0.0201 /// teacc 98.78 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.0754, -0.1583, -0.1127, ..., -0.0380, 0.0641, -0.1082], + [-0.0814, -0.0949, -0.0905, ..., 0.0976, -0.0386, 0.2375], + [ 0.0033, -0.0308, -0.0590, ..., -0.0051, -0.0120, -0.0837], + ..., + [ 0.0398, -0.1200, 0.1823, ..., 0.0401, -0.1222, 0.0626], + [-0.0286, 0.0666, -0.1533, ..., -0.0512, 0.0205, -0.1217], + [-0.0630, 0.0835, 0.0421, ..., -0.0627, -0.0566, -0.0346]], + device='cuda:0'), grad: tensor([[ 1.4658e-03, 1.1915e-04, 3.7050e-04, ..., 3.5019e-03, + 9.1362e-04, 1.3380e-03], + [-1.0223e-02, 2.8253e-05, -2.8992e-03, ..., -1.3588e-02, + -2.0332e-03, -9.1324e-03], + [ 1.0891e-03, 3.8385e-04, 5.9366e-04, ..., -3.2787e-03, + 4.5180e-04, -2.3746e-03], + ..., + [ 2.2602e-03, 1.0949e-04, 1.0633e-03, ..., 3.7098e-03, + 2.4724e-04, 3.9177e-03], + [-4.5662e-03, 1.8322e-04, 8.7452e-04, ..., 2.5597e-03, + 2.6703e-04, 3.5267e-03], + [ 3.2654e-03, 3.0804e-04, 1.7996e-03, ..., 3.7498e-03, + 2.7561e-04, 5.8823e-03]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0314, 0.0406, -0.0001, -0.0108, 0.0229, -0.0317, 0.0010, 0.0081, + -0.0027, -0.0154], device='cuda:0'), grad: tensor([ 0.0278, -0.0955, -0.0167, 0.0114, -0.0061, 0.0135, 0.0191, 0.0465, + 0.0014, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 214.53, cls_loss 0.5105 cls_loss_mapping 0.0029 cls_loss_causal 0.4473 re_mapping 0.0083 re_causal 0.0201 /// teacc 98.83 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.0749, -0.1586, -0.1112, ..., -0.0364, 0.0637, -0.1076], + [-0.0817, -0.0950, -0.0909, ..., 0.0982, -0.0383, 0.2386], + [ 0.0030, -0.0313, -0.0587, ..., -0.0061, -0.0103, -0.0832], + ..., + [ 0.0402, -0.1202, 0.1813, ..., 0.0394, -0.1222, 0.0618], + [-0.0280, 0.0669, -0.1540, ..., -0.0518, 0.0195, -0.1225], + [-0.0639, 0.0835, 0.0434, ..., -0.0628, -0.0570, -0.0341]], + device='cuda:0'), grad: tensor([[ 5.8670e-03, 8.0967e-04, -7.1449e-03, ..., 1.0231e-02, + -7.2861e-03, 8.4610e-03], + [-4.7607e-03, 4.8327e-04, -2.6817e-03, ..., 3.1796e-03, + 1.7810e-04, -3.7880e-03], + [ 1.6146e-03, 4.7660e-04, 8.6498e-04, ..., 1.5236e-02, + 2.3460e-04, 1.7920e-03], + ..., + [ 1.1206e-03, 9.3997e-05, 3.9744e-04, ..., -9.7809e-03, + 3.0518e-04, 1.4896e-03], + [ 1.5869e-03, 5.4884e-04, 9.3174e-04, ..., 2.4185e-03, + 4.2009e-04, 1.0519e-03], + [ 1.0290e-03, 1.0878e-04, 7.4005e-04, ..., -1.8148e-03, + 2.9159e-04, -8.1778e-04]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0299, 0.0409, -0.0004, -0.0098, 0.0214, -0.0326, 0.0011, 0.0075, + -0.0027, -0.0148], device='cuda:0'), grad: tensor([ 0.0129, -0.0012, 0.0304, 0.0248, -0.0414, -0.0318, 0.0329, 0.0063, + -0.0159, -0.0171], device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 214.41, cls_loss 0.4976 cls_loss_mapping 0.0038 cls_loss_causal 0.4470 re_mapping 0.0085 re_causal 0.0203 /// teacc 98.86 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.0748, -0.1595, -0.1103, ..., -0.0366, 0.0631, -0.1095], + [-0.0821, -0.0938, -0.0924, ..., 0.0982, -0.0390, 0.2400], + [ 0.0027, -0.0335, -0.0591, ..., -0.0078, -0.0107, -0.0842], + ..., + [ 0.0401, -0.1208, 0.1819, ..., 0.0395, -0.1234, 0.0625], + [-0.0281, 0.0672, -0.1532, ..., -0.0507, 0.0203, -0.1219], + [-0.0644, 0.0827, 0.0437, ..., -0.0624, -0.0565, -0.0349]], + device='cuda:0'), grad: tensor([[-1.4845e-06, 7.2062e-05, 3.3879e-04, ..., 1.4877e-03, + -1.0282e-06, 6.2227e-04], + [ 2.4557e-05, 6.7234e-05, 4.2582e-04, ..., 9.8991e-04, + 3.5390e-08, 2.0766e-04], + [-1.2836e-03, -4.8828e-03, -1.9798e-03, ..., -3.6259e-03, + 3.0827e-07, -1.8158e-03], + ..., + [ 1.5759e-04, 9.0265e-04, 4.2009e-04, ..., 1.0567e-03, + 6.7614e-07, 5.4169e-04], + [ 6.1321e-04, -1.3893e-02, 1.3227e-03, ..., 3.4924e-03, + 2.3153e-06, 1.6298e-03], + [ 4.5925e-05, 2.9602e-02, 1.0681e-02, ..., 6.1846e-04, + 3.3900e-07, 3.5405e-04]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0293, 0.0407, -0.0010, -0.0101, 0.0210, -0.0331, 0.0025, 0.0076, + -0.0027, -0.0150], device='cuda:0'), grad: tensor([-0.0154, 0.0187, -0.0448, -0.0229, -0.0044, 0.0232, -0.0007, 0.0171, + 0.0068, 0.0226], device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 214.33, cls_loss 0.5284 cls_loss_mapping 0.0018 cls_loss_causal 0.4681 re_mapping 0.0078 re_causal 0.0197 /// teacc 98.80 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.0753, -0.1611, -0.1110, ..., -0.0376, 0.0643, -0.1094], + [-0.0830, -0.0945, -0.0912, ..., 0.0985, -0.0394, 0.2390], + [ 0.0025, -0.0330, -0.0614, ..., -0.0072, -0.0111, -0.0836], + ..., + [ 0.0394, -0.1202, 0.1832, ..., 0.0387, -0.1230, 0.0627], + [-0.0286, 0.0670, -0.1529, ..., -0.0501, 0.0183, -0.1221], + [-0.0642, 0.0805, 0.0424, ..., -0.0622, -0.0540, -0.0344]], + device='cuda:0'), grad: tensor([[ 6.2406e-05, 3.5137e-05, 3.7289e-04, ..., 1.7118e-03, + 1.5891e-04, 1.2074e-03], + [ 8.1587e-04, 2.1231e-04, 2.8954e-03, ..., 7.5073e-03, + 1.0496e-04, 1.2352e-02], + [ 1.2121e-03, 2.3270e-03, 1.4668e-03, ..., 3.6106e-03, + 2.7752e-04, 1.9588e-03], + ..., + [ 9.1314e-05, 1.8072e-04, 4.8876e-04, ..., -3.2253e-03, + 1.7643e-04, 3.7441e-03], + [ 6.8045e-04, 1.1082e-03, 1.1559e-03, ..., 3.3054e-03, + 1.4663e-04, -1.6602e-02], + [ 1.0139e-04, 7.7248e-05, 4.8923e-04, ..., 1.6899e-03, + 1.1659e-04, 2.5940e-03]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0304, 0.0419, -0.0013, -0.0109, 0.0213, -0.0333, 0.0034, 0.0068, + -0.0030, -0.0139], device='cuda:0'), grad: tensor([-0.0399, 0.0595, 0.0312, -0.0471, -0.0258, 0.0166, -0.0089, -0.0069, + -0.0029, 0.0242], device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 214.17, cls_loss 0.5032 cls_loss_mapping 0.0017 cls_loss_causal 0.4429 re_mapping 0.0082 re_causal 0.0203 /// teacc 98.79 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.0759, -0.1623, -0.1120, ..., -0.0380, 0.0648, -0.1102], + [-0.0826, -0.0952, -0.0904, ..., 0.0997, -0.0390, 0.2393], + [ 0.0036, -0.0334, -0.0615, ..., -0.0070, -0.0110, -0.0834], + ..., + [ 0.0391, -0.1194, 0.1830, ..., 0.0384, -0.1233, 0.0622], + [-0.0289, 0.0684, -0.1529, ..., -0.0493, 0.0187, -0.1212], + [-0.0647, 0.0809, 0.0431, ..., -0.0631, -0.0541, -0.0345]], + device='cuda:0'), grad: tensor([[ 5.7316e-04, 1.8412e-06, 1.2934e-04, ..., 1.8349e-03, + -2.6271e-05, 1.5032e-04], + [-4.9305e-04, 6.5863e-05, -1.6165e-03, ..., -9.6664e-03, + 3.1758e-07, -2.4915e-04], + [ 3.5930e-04, 2.7671e-05, 2.6608e-04, ..., 3.1853e-03, + 2.6505e-06, 2.0337e-04], + ..., + [ 2.7680e-04, 2.8998e-05, 2.6369e-04, ..., 2.4681e-03, + 3.4738e-07, 1.0526e-04], + [ 6.8069e-05, 3.1382e-05, 1.5104e-04, ..., -4.7455e-03, + 1.1893e-06, 1.2875e-04], + [ 3.0351e-04, 2.5168e-05, 1.9038e-04, ..., 1.7843e-03, + 2.5369e-06, 7.0632e-05]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0311, 0.0408, -0.0012, -0.0101, 0.0202, -0.0344, 0.0043, 0.0076, + -0.0025, -0.0132], device='cuda:0'), grad: tensor([ 0.0119, -0.0140, 0.0161, -0.0161, -0.0172, 0.0129, 0.0269, 0.0148, + -0.0180, -0.0173], device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 214.27, cls_loss 0.5143 cls_loss_mapping 0.0023 cls_loss_causal 0.4465 re_mapping 0.0082 re_causal 0.0196 /// teacc 98.88 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.0743, -0.1616, -0.1109, ..., -0.0369, 0.0655, -0.1109], + [-0.0823, -0.0961, -0.0898, ..., 0.1005, -0.0390, 0.2407], + [ 0.0019, -0.0330, -0.0628, ..., -0.0073, -0.0108, -0.0844], + ..., + [ 0.0385, -0.1205, 0.1838, ..., 0.0381, -0.1252, 0.0626], + [-0.0275, 0.0684, -0.1536, ..., -0.0487, 0.0179, -0.1208], + [-0.0640, 0.0811, 0.0422, ..., -0.0633, -0.0537, -0.0353]], + device='cuda:0'), grad: tensor([[ 5.2357e-04, 1.7881e-06, 2.7714e-03, ..., 1.9484e-03, + 2.6435e-05, 2.9945e-03], + [ 3.7527e-04, 1.6894e-06, -1.9407e-04, ..., 7.2050e-04, + 2.2399e-04, 3.0732e-04], + [ 4.9734e-04, 3.2723e-05, 9.0361e-04, ..., 2.8419e-04, + -1.2045e-03, 1.4362e-03], + ..., + [ 9.1505e-04, 3.8791e-04, -6.0616e-03, ..., 4.0588e-03, + 1.3880e-05, 7.7477e-03], + [ 6.3667e-03, -3.7134e-05, 4.5323e-04, ..., 3.9062e-03, + 2.1443e-05, 1.5764e-03], + [-8.1329e-03, -4.5705e-04, -9.5320e-04, ..., -8.7662e-03, + 1.8869e-06, -1.2733e-02]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0304, 0.0409, -0.0021, -0.0110, 0.0214, -0.0342, 0.0038, 0.0069, + -0.0014, -0.0135], device='cuda:0'), grad: tensor([ 0.0154, 0.0092, 0.0099, 0.0116, -0.0178, 0.0156, -0.0166, 0.0103, + 0.0252, -0.0627], device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 214.27, cls_loss 0.5264 cls_loss_mapping 0.0028 cls_loss_causal 0.4678 re_mapping 0.0084 re_causal 0.0207 /// teacc 98.86 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.0745, -0.1635, -0.1112, ..., -0.0370, 0.0653, -0.1112], + [-0.0824, -0.0958, -0.0886, ..., 0.1000, -0.0383, 0.2406], + [ 0.0005, -0.0341, -0.0626, ..., -0.0080, -0.0092, -0.0849], + ..., + [ 0.0382, -0.1190, 0.1843, ..., 0.0386, -0.1242, 0.0621], + [-0.0251, 0.0676, -0.1526, ..., -0.0481, 0.0179, -0.1204], + [-0.0638, 0.0825, 0.0416, ..., -0.0644, -0.0536, -0.0353]], + device='cuda:0'), grad: tensor([[ 2.9540e-04, -2.5070e-02, -2.3403e-03, ..., 1.1110e-03, + 3.4666e-04, 2.0802e-04], + [ 1.8902e-03, 2.3899e-03, 4.4513e-04, ..., 2.9869e-03, + 5.2071e-04, 1.0290e-03], + [ 4.1127e-04, 1.8234e-02, 4.2801e-03, ..., 1.2121e-03, + 1.7130e-04, 2.0134e-04], + ..., + [ 1.0262e-03, 1.2140e-03, -5.4207e-03, ..., 1.9932e-03, + 1.8466e-04, 3.7146e-04], + [-6.3629e-03, -3.1834e-03, 3.6502e-04, ..., -1.1871e-02, + -1.3582e-05, -2.6436e-03], + [ 8.2970e-04, 1.6003e-03, 5.0163e-04, ..., 1.7719e-03, + 3.6025e-04, 4.7350e-04]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0300, 0.0405, -0.0021, -0.0107, 0.0206, -0.0341, 0.0043, 0.0074, + -0.0014, -0.0139], device='cuda:0'), grad: tensor([-0.0091, 0.0196, 0.0299, 0.0313, 0.0119, -0.0767, 0.0103, 0.0092, + -0.0412, 0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 214.19, cls_loss 0.5172 cls_loss_mapping 0.0021 cls_loss_causal 0.4596 re_mapping 0.0083 re_causal 0.0195 /// teacc 98.88 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.0748, -0.1627, -0.1106, ..., -0.0387, 0.0654, -0.1112], + [-0.0838, -0.0954, -0.0888, ..., 0.0987, -0.0383, 0.2397], + [ 0.0020, -0.0323, -0.0629, ..., -0.0071, -0.0098, -0.0825], + ..., + [ 0.0372, -0.1194, 0.1832, ..., 0.0389, -0.1247, 0.0618], + [-0.0249, 0.0659, -0.1532, ..., -0.0471, 0.0176, -0.1185], + [-0.0641, 0.0816, 0.0415, ..., -0.0651, -0.0534, -0.0362]], + device='cuda:0'), grad: tensor([[ 2.2578e-04, 7.8297e-04, 2.3556e-04, ..., 1.6766e-03, + 2.9254e-04, 2.4533e-04], + [ 2.2278e-03, -1.4864e-05, 2.8324e-04, ..., 8.5068e-04, + -1.4410e-03, -5.2595e-04], + [ 3.6716e-04, 2.8496e-03, 1.0252e-03, ..., 3.4847e-03, + 3.6836e-04, 1.1091e-03], + ..., + [-3.2940e-03, 1.6883e-05, -2.4166e-03, ..., -1.2566e-02, + -1.4410e-03, -3.0136e-03], + [ 1.1587e-03, 1.5316e-03, 1.2655e-03, ..., 4.4403e-03, + 3.9458e-04, 2.3136e-03], + [-7.7248e-04, -3.7785e-03, 3.6335e-04, ..., -1.0689e-02, + 3.2663e-04, -1.5736e-03]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0302, 0.0405, -0.0013, -0.0106, 0.0208, -0.0345, 0.0042, 0.0066, + -0.0001, -0.0148], device='cuda:0'), grad: tensor([ 0.0209, -0.0217, 0.0315, 0.0170, 0.0200, 0.0098, 0.0372, -0.0793, + 0.0083, -0.0438], device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 214.46, cls_loss 0.5432 cls_loss_mapping 0.0018 cls_loss_causal 0.4772 re_mapping 0.0084 re_causal 0.0202 /// teacc 98.90 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.0750, -0.1629, -0.1107, ..., -0.0388, 0.0654, -0.1112], + [-0.0837, -0.0955, -0.0888, ..., 0.0986, -0.0382, 0.2397], + [ 0.0023, -0.0323, -0.0629, ..., -0.0071, -0.0099, -0.0825], + ..., + [ 0.0371, -0.1196, 0.1832, ..., 0.0391, -0.1248, 0.0618], + [-0.0251, 0.0658, -0.1534, ..., -0.0474, 0.0176, -0.1188], + [-0.0640, 0.0818, 0.0416, ..., -0.0650, -0.0534, -0.0360]], + device='cuda:0'), grad: tensor([[ 3.0324e-05, 2.0102e-05, -3.8052e-04, ..., -9.5062e-03, + -6.6698e-05, -7.5493e-03], + [ 5.5619e-06, 2.0728e-05, 9.4604e-04, ..., 2.5539e-03, + 2.3320e-05, 5.4216e-04], + [ 4.9353e-05, 2.3174e-04, -3.0956e-03, ..., -1.9627e-03, + -9.1791e-04, 1.9855e-03], + ..., + [ 6.9022e-05, 1.5795e-04, 1.3008e-03, ..., 5.4359e-03, + 3.7551e-05, 3.6736e-03], + [ 3.0303e-04, 1.4944e-03, 1.0242e-03, ..., -1.2264e-03, + 2.1711e-05, 2.4242e-03], + [-1.4758e-04, -1.1969e-04, -2.8563e-04, ..., 1.0538e-03, + 2.2262e-05, 2.2924e-04]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0303, 0.0405, -0.0013, -0.0106, 0.0210, -0.0344, 0.0042, 0.0066, + -0.0004, -0.0147], device='cuda:0'), grad: tensor([-0.0431, 0.0197, 0.0058, 0.0082, -0.0221, -0.0053, -0.0232, 0.0309, + 0.0173, 0.0119], device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 214.31, cls_loss 0.5123 cls_loss_mapping 0.0013 cls_loss_causal 0.4572 re_mapping 0.0079 re_causal 0.0199 /// teacc 98.89 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.0749, -0.1629, -0.1107, ..., -0.0387, 0.0653, -0.1113], + [-0.0837, -0.0955, -0.0888, ..., 0.0985, -0.0382, 0.2397], + [ 0.0024, -0.0323, -0.0630, ..., -0.0070, -0.0100, -0.0826], + ..., + [ 0.0372, -0.1195, 0.1832, ..., 0.0391, -0.1246, 0.0618], + [-0.0253, 0.0659, -0.1534, ..., -0.0475, 0.0176, -0.1188], + [-0.0642, 0.0817, 0.0416, ..., -0.0651, -0.0534, -0.0358]], + device='cuda:0'), grad: tensor([[-8.2403e-06, -6.2752e-04, 3.7163e-05, ..., 2.8062e-04, + 5.2154e-08, 2.5058e-04], + [ 1.4519e-06, 6.4754e-04, 1.6034e-04, ..., 7.0477e-04, + 2.1420e-08, 1.7071e-04], + [ 6.2957e-07, 1.4865e-04, 1.1635e-03, ..., -1.7576e-03, + -2.7679e-06, -2.4986e-04], + ..., + [ 1.4836e-06, 3.9458e-04, 2.0828e-02, ..., 1.1234e-03, + 3.4459e-08, 1.0735e-04], + [ 1.1861e-05, -3.6640e-03, 8.6308e-04, ..., -3.5610e-03, + 1.7425e-06, -1.4296e-03], + [ 1.3784e-05, 8.3017e-04, 8.4066e-04, ..., 8.5688e-04, + 1.2759e-07, 2.2531e-04]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0302, 0.0404, -0.0012, -0.0108, 0.0212, -0.0344, 0.0042, 0.0066, + -0.0006, -0.0147], device='cuda:0'), grad: tensor([-0.0379, 0.0217, -0.0343, -0.0342, 0.0237, -0.0257, 0.0313, 0.0459, + -0.0112, 0.0206], device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 214.48, cls_loss 0.5362 cls_loss_mapping 0.0015 cls_loss_causal 0.4704 re_mapping 0.0079 re_causal 0.0200 /// teacc 98.93 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.0748, -0.1630, -0.1108, ..., -0.0386, 0.0654, -0.1115], + [-0.0836, -0.0955, -0.0886, ..., 0.0987, -0.0383, 0.2400], + [ 0.0024, -0.0324, -0.0630, ..., -0.0069, -0.0100, -0.0827], + ..., + [ 0.0372, -0.1195, 0.1831, ..., 0.0389, -0.1247, 0.0618], + [-0.0254, 0.0658, -0.1534, ..., -0.0475, 0.0178, -0.1189], + [-0.0642, 0.0817, 0.0417, ..., -0.0651, -0.0535, -0.0359]], + device='cuda:0'), grad: tensor([[ 2.6202e-04, 7.8008e-06, 1.3316e-04, ..., -6.5079e-03, + 2.5582e-04, 2.7418e-04], + [ 1.6512e-06, 3.5524e-05, 5.3883e-05, ..., -6.2847e-04, + 5.9128e-05, -1.0624e-03], + [ 4.4513e-04, 2.7752e-04, 3.4904e-04, ..., 3.2387e-03, + 1.4424e-04, 1.5926e-04], + ..., + [-5.7983e-03, 3.2425e-05, -1.1721e-03, ..., -1.0536e-02, + 3.7253e-05, -2.8839e-03], + [ 1.1082e-03, 2.6155e-04, 1.5850e-03, ..., 3.9787e-03, + 1.9538e-04, 5.0354e-04], + [ 2.5063e-03, -1.0443e-03, 1.0855e-05, ..., 1.8682e-03, + 6.8605e-05, 1.3247e-03]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0302, 0.0405, -0.0013, -0.0107, 0.0214, -0.0344, 0.0043, 0.0065, + -0.0007, -0.0148], device='cuda:0'), grad: tensor([ 0.0016, 0.0130, -0.0007, -0.0182, 0.0310, -0.0019, 0.0244, -0.0168, + 0.0072, -0.0396], device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 214.93, cls_loss 0.4982 cls_loss_mapping 0.0012 cls_loss_causal 0.4382 re_mapping 0.0078 re_causal 0.0188 /// teacc 98.91 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.0748, -0.1631, -0.1108, ..., -0.0388, 0.0656, -0.1117], + [-0.0835, -0.0953, -0.0887, ..., 0.0987, -0.0383, 0.2400], + [ 0.0024, -0.0322, -0.0630, ..., -0.0070, -0.0101, -0.0827], + ..., + [ 0.0373, -0.1197, 0.1830, ..., 0.0390, -0.1248, 0.0619], + [-0.0255, 0.0658, -0.1535, ..., -0.0476, 0.0179, -0.1191], + [-0.0643, 0.0819, 0.0417, ..., -0.0651, -0.0535, -0.0358]], + device='cuda:0'), grad: tensor([[ 3.0422e-03, 4.9515e-03, 6.5279e-04, ..., 7.2212e-03, + 1.2362e-04, 1.7262e-03], + [-1.8120e-04, 1.9357e-05, -1.7042e-03, ..., -4.4250e-03, + 9.4712e-05, -1.3876e-03], + [ 8.6427e-05, 3.8028e-05, 1.5163e-04, ..., 6.1560e-04, + 7.8201e-05, 3.6311e-04], + ..., + [ 1.1218e-04, 2.2590e-05, 1.2350e-04, ..., 1.0214e-03, + 7.7724e-05, 3.8242e-04], + [-6.1941e-04, 1.4007e-04, 2.2912e-04, ..., -1.5497e-03, + -8.8835e-04, 9.1076e-04], + [ 1.0401e-04, 4.3899e-05, 6.7949e-05, ..., 4.5681e-04, + 9.6977e-05, 5.0211e-04]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0303, 0.0405, -0.0014, -0.0107, 0.0214, -0.0345, 0.0045, 0.0066, + -0.0008, -0.0148], device='cuda:0'), grad: tensor([ 0.0480, 0.0043, 0.0113, 0.0131, 0.0122, -0.0299, -0.0397, -0.0183, + -0.0143, 0.0132], device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 214.82, cls_loss 0.4975 cls_loss_mapping 0.0010 cls_loss_causal 0.4284 re_mapping 0.0076 re_causal 0.0190 /// teacc 98.90 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.0747, -0.1632, -0.1108, ..., -0.0387, 0.0656, -0.1118], + [-0.0835, -0.0950, -0.0887, ..., 0.0988, -0.0383, 0.2401], + [ 0.0024, -0.0323, -0.0631, ..., -0.0070, -0.0101, -0.0827], + ..., + [ 0.0372, -0.1195, 0.1830, ..., 0.0389, -0.1248, 0.0618], + [-0.0255, 0.0657, -0.1534, ..., -0.0477, 0.0180, -0.1189], + [-0.0643, 0.0820, 0.0418, ..., -0.0652, -0.0536, -0.0358]], + device='cuda:0'), grad: tensor([[-6.3210e-03, 2.7323e-04, -3.2120e-03, ..., -4.8866e-03, + -3.9768e-04, 6.2609e-04], + [ 6.8665e-04, 1.7881e-04, 4.6182e-04, ..., 1.2865e-03, + 4.3184e-05, 9.8515e-04], + [ 3.1900e-04, 1.1641e-04, 2.1386e-04, ..., 1.1673e-03, + 2.0087e-05, 4.8280e-04], + ..., + [ 6.1321e-04, 1.1377e-03, 5.0497e-04, ..., 1.2856e-03, + 3.8564e-05, 8.9931e-04], + [ 1.4889e-04, -2.9736e-03, -9.4128e-04, ..., -7.8297e-04, + 9.3654e-06, -1.5049e-03], + [ 2.5139e-03, 4.6005e-03, 1.7328e-03, ..., 3.9005e-03, + 1.5807e-04, 2.3727e-03]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0304, 0.0404, -0.0014, -0.0107, 0.0216, -0.0345, 0.0046, 0.0066, + -0.0008, -0.0149], device='cuda:0'), grad: tensor([-0.0157, 0.0104, 0.0092, 0.0130, -0.0584, 0.0091, -0.0068, 0.0151, + -0.0216, 0.0457], device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 214.71, cls_loss 0.5050 cls_loss_mapping 0.0010 cls_loss_causal 0.4480 re_mapping 0.0075 re_causal 0.0186 /// teacc 98.92 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.0747, -0.1633, -0.1107, ..., -0.0387, 0.0656, -0.1119], + [-0.0835, -0.0950, -0.0886, ..., 0.0989, -0.0382, 0.2401], + [ 0.0024, -0.0323, -0.0631, ..., -0.0071, -0.0102, -0.0827], + ..., + [ 0.0371, -0.1195, 0.1830, ..., 0.0388, -0.1249, 0.0619], + [-0.0255, 0.0657, -0.1534, ..., -0.0476, 0.0180, -0.1189], + [-0.0643, 0.0821, 0.0418, ..., -0.0654, -0.0537, -0.0360]], + device='cuda:0'), grad: tensor([[ 3.0041e-04, 2.8051e-06, 1.8403e-05, ..., 3.1357e-03, + 3.5620e-04, 5.6624e-06], + [ 1.6010e-04, 1.1949e-06, -7.8142e-05, ..., 3.1586e-03, + 1.4286e-03, -3.7241e-04], + [ 3.8934e-04, 2.0847e-05, 1.7157e-03, ..., 1.8253e-03, + 4.5300e-04, 3.3379e-04], + ..., + [-9.6858e-05, -2.2069e-05, -2.4092e-04, ..., 7.1564e-03, + 4.5624e-03, 6.8893e-03], + [ 1.3554e-04, -2.5997e-03, 3.2449e-04, ..., -1.0719e-02, + -7.7896e-03, 1.3316e-04], + [ 1.1796e-04, 1.1168e-05, 5.2786e-04, ..., -5.2376e-03, + 2.0161e-03, 1.2360e-03]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0305, 0.0405, -0.0014, -0.0106, 0.0216, -0.0345, 0.0045, 0.0066, + -0.0009, -0.0149], device='cuda:0'), grad: tensor([ 0.0120, 0.0070, 0.0090, 0.0138, -0.0227, 0.0034, -0.0094, -0.0202, + 0.0251, -0.0181], device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 214.83, cls_loss 0.4905 cls_loss_mapping 0.0010 cls_loss_causal 0.4271 re_mapping 0.0069 re_causal 0.0175 /// teacc 98.91 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.0748, -0.1632, -0.1106, ..., -0.0388, 0.0656, -0.1120], + [-0.0834, -0.0950, -0.0887, ..., 0.0990, -0.0382, 0.2402], + [ 0.0024, -0.0323, -0.0631, ..., -0.0071, -0.0103, -0.0827], + ..., + [ 0.0370, -0.1195, 0.1831, ..., 0.0388, -0.1251, 0.0619], + [-0.0255, 0.0658, -0.1532, ..., -0.0475, 0.0181, -0.1188], + [-0.0643, 0.0822, 0.0417, ..., -0.0655, -0.0537, -0.0359]], + device='cuda:0'), grad: tensor([[-6.8998e-04, -1.3409e-03, 3.2224e-07, ..., 2.3234e-04, + 0.0000e+00, 1.8044e-03], + [-1.3475e-03, -6.7689e-06, -6.2585e-05, ..., -4.5395e-03, + 0.0000e+00, -5.3101e-03], + [ 3.4547e-04, 1.3435e-04, 3.4552e-07, ..., 3.8600e-04, + 0.0000e+00, -5.8556e-03], + ..., + [ 6.5136e-04, 1.4842e-04, 2.5421e-05, ..., 6.7425e-04, + 0.0000e+00, 1.6499e-03], + [ 6.8235e-04, -4.1938e-04, 7.5325e-06, ..., 2.7514e-04, + 0.0000e+00, 1.0900e-03], + [-1.6241e-03, 7.5769e-04, 9.8571e-06, ..., 7.5626e-04, + 0.0000e+00, 1.1101e-03]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0306, 0.0408, -0.0014, -0.0106, 0.0215, -0.0345, 0.0045, 0.0065, + -0.0008, -0.0149], device='cuda:0'), grad: tensor([-0.0088, -0.0051, -0.0046, 0.0213, 0.0204, 0.0207, 0.0134, -0.0079, + -0.0523, 0.0029], device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 214.63, cls_loss 0.5277 cls_loss_mapping 0.0011 cls_loss_causal 0.4663 re_mapping 0.0071 re_causal 0.0184 /// teacc 98.90 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.0746, -0.1632, -0.1106, ..., -0.0386, 0.0657, -0.1119], + [-0.0835, -0.0950, -0.0886, ..., 0.0990, -0.0380, 0.2401], + [ 0.0025, -0.0324, -0.0630, ..., -0.0070, -0.0103, -0.0827], + ..., + [ 0.0369, -0.1196, 0.1832, ..., 0.0387, -0.1251, 0.0619], + [-0.0255, 0.0657, -0.1533, ..., -0.0476, 0.0181, -0.1190], + [-0.0643, 0.0820, 0.0417, ..., -0.0654, -0.0538, -0.0357]], + device='cuda:0'), grad: tensor([[ 7.7534e-04, 1.0997e-05, 4.6253e-05, ..., 3.0403e-03, + 2.6356e-07, 5.9414e-04], + [-3.1567e-03, 7.8201e-05, -3.1710e-04, ..., -1.6724e-02, + 5.5879e-09, -5.1651e-03], + [ 1.4105e-03, 2.8998e-05, 9.3639e-05, ..., 2.8667e-03, + 4.0889e-05, 7.5197e-04], + ..., + [ 7.9041e-03, 1.6761e-04, 1.2711e-02, ..., 7.8659e-03, + 0.0000e+00, 3.9902e-03], + [-5.0087e-03, 5.0217e-05, 1.8799e-04, ..., 3.5362e-03, + 2.5611e-07, 2.2850e-03], + [ 9.0265e-04, 1.7345e-04, 1.0777e-03, ..., 3.7498e-03, + 1.8626e-09, 3.3321e-03]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0305, 0.0408, -0.0014, -0.0106, 0.0216, -0.0346, 0.0045, 0.0065, + -0.0009, -0.0150], device='cuda:0'), grad: tensor([ 0.0171, -0.0327, 0.0231, -0.0067, 0.0093, 0.0183, -0.0120, 0.0326, + -0.0700, 0.0211], device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 214.54, cls_loss 0.4993 cls_loss_mapping 0.0010 cls_loss_causal 0.4333 re_mapping 0.0069 re_causal 0.0178 /// teacc 98.88 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.0746, -0.1632, -0.1106, ..., -0.0386, 0.0658, -0.1120], + [-0.0835, -0.0950, -0.0886, ..., 0.0992, -0.0381, 0.2402], + [ 0.0026, -0.0324, -0.0630, ..., -0.0071, -0.0104, -0.0825], + ..., + [ 0.0371, -0.1196, 0.1832, ..., 0.0387, -0.1251, 0.0620], + [-0.0255, 0.0658, -0.1533, ..., -0.0477, 0.0181, -0.1192], + [-0.0644, 0.0820, 0.0416, ..., -0.0654, -0.0538, -0.0358]], + device='cuda:0'), grad: tensor([[ 1.3560e-05, 2.9996e-05, 1.2569e-05, ..., 5.0545e-04, + 1.0747e-06, 2.1625e-04], + [-8.3017e-04, -3.6311e-04, -4.1103e-04, ..., -1.1367e-04, + 3.8147e-05, -1.1463e-03], + [-5.2080e-06, 4.7874e-04, -2.1800e-05, ..., -2.0008e-03, + 3.8624e-05, 1.9884e-04], + ..., + [ 3.5316e-05, 2.4867e-04, 6.2990e-04, ..., -1.4629e-03, + 3.2187e-06, -1.5831e-03], + [ 5.5504e-04, -2.5201e-04, 6.5088e-04, ..., 7.7295e-04, + 8.1122e-05, 1.1110e-03], + [ 3.2753e-05, 4.8714e-03, 8.0109e-03, ..., 4.7565e-04, + 2.4308e-06, 1.7824e-03]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0306, 0.0409, -0.0014, -0.0106, 0.0216, -0.0346, 0.0046, 0.0065, + -0.0009, -0.0150], device='cuda:0'), grad: tensor([ 0.0164, 0.0122, -0.0442, 0.0126, 0.0115, -0.0145, 0.0071, -0.0481, + 0.0112, 0.0357], device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 215.00, cls_loss 0.4799 cls_loss_mapping 0.0010 cls_loss_causal 0.4181 re_mapping 0.0071 re_causal 0.0177 /// teacc 98.92 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.0745, -0.1633, -0.1105, ..., -0.0386, 0.0657, -0.1121], + [-0.0835, -0.0950, -0.0887, ..., 0.0992, -0.0381, 0.2400], + [ 0.0026, -0.0324, -0.0629, ..., -0.0071, -0.0103, -0.0823], + ..., + [ 0.0371, -0.1195, 0.1833, ..., 0.0388, -0.1252, 0.0620], + [-0.0256, 0.0657, -0.1533, ..., -0.0478, 0.0182, -0.1192], + [-0.0644, 0.0820, 0.0416, ..., -0.0653, -0.0538, -0.0358]], + device='cuda:0'), grad: tensor([[ 6.0463e-04, 2.7871e-04, 4.6706e-04, ..., 6.9761e-04, + 9.7081e-06, -8.6260e-04], + [ 1.4019e-03, 9.1505e-04, 1.3456e-03, ..., 4.8981e-03, + 8.2254e-05, 1.5373e-03], + [-1.5850e-03, 3.1281e-04, 4.8804e-04, ..., 1.9464e-03, + 1.4462e-05, -4.5815e-03], + ..., + [ 2.2602e-03, 1.0519e-03, 1.6870e-03, ..., -9.1476e-03, + 5.6863e-05, 3.6125e-03], + [ 1.2398e-03, 3.9339e-04, 8.0013e-04, ..., 4.2763e-03, + 9.0897e-05, 2.7046e-03], + [ 1.4296e-03, 3.0565e-04, -1.5354e-04, ..., 3.5324e-03, + 3.9220e-05, 2.2945e-03]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0305, 0.0409, -0.0014, -0.0105, 0.0217, -0.0345, 0.0044, 0.0066, + -0.0011, -0.0150], device='cuda:0'), grad: tensor([-0.0446, 0.0252, -0.0442, -0.0253, 0.0051, 0.0252, -0.0169, 0.0035, + 0.0297, 0.0424], device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 214.68, cls_loss 0.5196 cls_loss_mapping 0.0011 cls_loss_causal 0.4605 re_mapping 0.0068 re_causal 0.0185 /// teacc 98.95 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.0746, -0.1634, -0.1104, ..., -0.0387, 0.0657, -0.1122], + [-0.0834, -0.0949, -0.0888, ..., 0.0992, -0.0381, 0.2400], + [ 0.0026, -0.0325, -0.0629, ..., -0.0071, -0.0104, -0.0823], + ..., + [ 0.0372, -0.1197, 0.1832, ..., 0.0388, -0.1252, 0.0621], + [-0.0257, 0.0657, -0.1534, ..., -0.0478, 0.0182, -0.1193], + [-0.0644, 0.0820, 0.0417, ..., -0.0652, -0.0540, -0.0358]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0003, 0.0002, ..., 0.0009, 0.0001, 0.0002], + [ 0.0006, 0.0005, 0.0003, ..., 0.0006, 0.0002, 0.0003], + [ 0.0009, 0.0002, 0.0013, ..., 0.0003, -0.0009, -0.0002], + ..., + [ 0.0018, 0.0050, 0.0082, ..., 0.0090, 0.0003, 0.0056], + [-0.0006, -0.0003, 0.0003, ..., 0.0004, 0.0003, 0.0002], + [ 0.0007, -0.0051, -0.0043, ..., -0.0013, -0.0002, -0.0037]], + device='cuda:0') +Epoch 412, bias, value: tensor([-0.0306, 0.0409, -0.0014, -0.0106, 0.0217, -0.0345, 0.0047, 0.0066, + -0.0013, -0.0150], device='cuda:0'), grad: tensor([ 0.0101, 0.0121, 0.0076, -0.0186, -0.0197, -0.0666, 0.0116, 0.0430, + 0.0232, -0.0027], device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 214.62, cls_loss 0.5089 cls_loss_mapping 0.0010 cls_loss_causal 0.4414 re_mapping 0.0069 re_causal 0.0179 /// teacc 98.95 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.0746, -0.1635, -0.1103, ..., -0.0388, 0.0657, -0.1122], + [-0.0834, -0.0950, -0.0888, ..., 0.0993, -0.0381, 0.2400], + [ 0.0025, -0.0326, -0.0630, ..., -0.0072, -0.0105, -0.0822], + ..., + [ 0.0371, -0.1197, 0.1832, ..., 0.0387, -0.1253, 0.0621], + [-0.0257, 0.0657, -0.1533, ..., -0.0478, 0.0182, -0.1195], + [-0.0645, 0.0819, 0.0418, ..., -0.0651, -0.0539, -0.0358]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5885e-04, 1.1456e-04, ..., 4.3106e-04, + 4.2617e-05, 1.4985e-04], + [ 0.0000e+00, -1.4102e-04, 1.0979e-04, ..., -2.9254e-04, + 4.4286e-05, 3.6502e-04], + [ 1.8626e-09, -1.3614e-04, 2.8515e-04, ..., -2.0552e-04, + 8.6248e-05, 2.4819e-04], + ..., + [ 0.0000e+00, 1.7080e-03, 5.5170e-04, ..., 1.8716e-04, + 6.7651e-05, 2.4395e-03], + [-9.3132e-10, -5.9853e-03, 3.7980e-04, ..., 7.3671e-04, + 1.1462e-04, 1.4057e-03], + [ 0.0000e+00, 2.4529e-03, -1.4982e-03, ..., 1.3196e-04, + 1.1384e-04, -5.1651e-03]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0307, 0.0410, -0.0015, -0.0106, 0.0218, -0.0344, 0.0046, 0.0066, + -0.0012, -0.0150], device='cuda:0'), grad: tensor([ 0.0147, -0.0094, -0.0128, 0.0242, 0.0140, 0.0204, -0.0444, 0.0278, + -0.0076, -0.0269], device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 214.73, cls_loss 0.5004 cls_loss_mapping 0.0011 cls_loss_causal 0.4348 re_mapping 0.0065 re_causal 0.0170 /// teacc 98.91 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.0745, -0.1635, -0.1104, ..., -0.0388, 0.0657, -0.1121], + [-0.0832, -0.0951, -0.0888, ..., 0.0994, -0.0382, 0.2400], + [ 0.0023, -0.0326, -0.0630, ..., -0.0072, -0.0105, -0.0823], + ..., + [ 0.0371, -0.1199, 0.1831, ..., 0.0387, -0.1253, 0.0621], + [-0.0257, 0.0657, -0.1532, ..., -0.0480, 0.0183, -0.1196], + [-0.0644, 0.0819, 0.0418, ..., -0.0651, -0.0539, -0.0358]], + device='cuda:0'), grad: tensor([[-4.5586e-03, 1.9908e-04, 4.2230e-05, ..., -6.1340e-03, + -3.2735e-04, -8.7585e-03], + [ 1.7424e-03, 2.4796e-04, 2.8992e-04, ..., 3.9711e-03, + 1.1927e-04, 2.1839e-03], + [ 8.7786e-04, 1.8752e-04, 1.4102e-04, ..., 2.4242e-03, + -3.2306e-04, 9.7752e-04], + ..., + [ 1.1263e-03, 1.9968e-04, 5.0926e-04, ..., -5.2872e-03, + 1.9097e-04, -1.1730e-03], + [-5.7411e-03, -7.9155e-04, 9.1612e-05, ..., -7.2975e-03, + 1.1063e-04, 1.2827e-03], + [-2.0123e-03, -1.3103e-03, -2.7695e-03, ..., -6.2370e-03, + -8.9645e-04, -1.7395e-03]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0307, 0.0410, -0.0015, -0.0104, 0.0217, -0.0345, 0.0047, 0.0065, + -0.0014, -0.0150], device='cuda:0'), grad: tensor([-0.0119, 0.0269, 0.0059, 0.0205, 0.0309, 0.0268, 0.0335, -0.0508, + -0.0382, -0.0436], device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 214.83, cls_loss 0.4833 cls_loss_mapping 0.0011 cls_loss_causal 0.4192 re_mapping 0.0069 re_causal 0.0172 /// teacc 98.94 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.0745, -0.1635, -0.1104, ..., -0.0389, 0.0656, -0.1122], + [-0.0831, -0.0949, -0.0889, ..., 0.0995, -0.0381, 0.2402], + [ 0.0024, -0.0326, -0.0629, ..., -0.0071, -0.0105, -0.0824], + ..., + [ 0.0368, -0.1200, 0.1830, ..., 0.0385, -0.1253, 0.0621], + [-0.0258, 0.0657, -0.1533, ..., -0.0480, 0.0183, -0.1197], + [-0.0643, 0.0819, 0.0418, ..., -0.0651, -0.0541, -0.0358]], + device='cuda:0'), grad: tensor([[ 8.6880e-04, 6.0177e-04, 1.6820e-06, ..., 1.0643e-03, + 5.6267e-04, -6.6853e-04], + [-3.7231e-03, -1.2522e-03, 2.4274e-05, ..., -4.1504e-03, + 1.5366e-04, 8.7547e-04], + [ 9.2363e-04, 8.5211e-04, 9.1612e-05, ..., 1.5316e-03, + 9.1910e-05, 7.4577e-04], + ..., + [ 1.1368e-03, 1.1625e-03, -4.2462e-04, ..., 1.4963e-03, + 1.1802e-04, 4.1270e-04], + [ 1.4668e-03, 1.6375e-03, 7.3127e-06, ..., 1.7281e-03, + 3.4809e-04, 9.9277e-04], + [-1.7776e-03, -6.9046e-03, 2.7347e-04, ..., -2.2876e-04, + -5.9223e-04, -3.4952e-04]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0309, 0.0410, -0.0013, -0.0104, 0.0218, -0.0344, 0.0048, 0.0065, + -0.0014, -0.0152], device='cuda:0'), grad: tensor([-0.0077, -0.0605, 0.0235, 0.0261, -0.0448, 0.0210, -0.0016, 0.0253, + 0.0289, -0.0102], device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 214.43, cls_loss 0.5010 cls_loss_mapping 0.0011 cls_loss_causal 0.4382 re_mapping 0.0066 re_causal 0.0171 /// teacc 98.97 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.0744, -0.1636, -0.1104, ..., -0.0389, 0.0656, -0.1122], + [-0.0831, -0.0949, -0.0890, ..., 0.0996, -0.0382, 0.2403], + [ 0.0023, -0.0325, -0.0628, ..., -0.0070, -0.0104, -0.0822], + ..., + [ 0.0368, -0.1201, 0.1828, ..., 0.0384, -0.1254, 0.0621], + [-0.0257, 0.0657, -0.1532, ..., -0.0481, 0.0183, -0.1197], + [-0.0640, 0.0819, 0.0421, ..., -0.0650, -0.0540, -0.0358]], + device='cuda:0'), grad: tensor([[ 2.7561e-04, 2.9922e-04, 8.0228e-05, ..., 1.3695e-03, + -5.9414e-04, 1.0214e-03], + [ 2.0027e-03, -5.6982e-04, 6.0654e-04, ..., 6.5002e-03, + 2.2948e-06, 6.2675e-03], + [ 3.4189e-04, 4.2701e-04, -4.1199e-04, ..., 9.5797e-04, + 2.4036e-05, 6.9580e-03], + ..., + [ 4.9543e-04, 8.9693e-04, 4.6682e-04, ..., 2.4719e-03, + 1.1407e-05, 1.9350e-03], + [ 7.3862e-04, 1.0805e-03, 2.2304e-04, ..., 6.8741e-03, + 2.2575e-05, 3.9368e-03], + [ 3.0613e-04, -7.7963e-04, 2.7370e-04, ..., 1.4286e-03, + 5.8889e-04, 9.0885e-04]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0308, 0.0411, -0.0011, -0.0104, 0.0219, -0.0345, 0.0046, 0.0064, + -0.0014, -0.0151], device='cuda:0'), grad: tensor([-0.0174, 0.0110, 0.0215, -0.0432, 0.0019, 0.0220, -0.0504, 0.0167, + 0.0250, 0.0129], device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 214.47, cls_loss 0.5117 cls_loss_mapping 0.0009 cls_loss_causal 0.4466 re_mapping 0.0066 re_causal 0.0179 /// teacc 99.00 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.0744, -0.1634, -0.1104, ..., -0.0388, 0.0656, -0.1123], + [-0.0829, -0.0950, -0.0889, ..., 0.0997, -0.0382, 0.2404], + [ 0.0022, -0.0327, -0.0629, ..., -0.0070, -0.0104, -0.0823], + ..., + [ 0.0368, -0.1202, 0.1828, ..., 0.0384, -0.1255, 0.0621], + [-0.0259, 0.0658, -0.1533, ..., -0.0481, 0.0185, -0.1199], + [-0.0640, 0.0820, 0.0420, ..., -0.0650, -0.0541, -0.0358]], + device='cuda:0'), grad: tensor([[ 7.8869e-04, 7.5493e-03, 6.7830e-05, ..., 2.4185e-03, + 3.1680e-05, 1.1663e-03], + [ 8.8835e-04, 8.5497e-04, 4.6015e-04, ..., -2.1160e-04, + 1.8322e-04, -3.4504e-03], + [-3.5973e-03, 3.3712e-04, -7.4244e-04, ..., 1.3638e-03, + 2.4617e-05, 8.4925e-04], + ..., + [ 9.1457e-04, 5.6696e-04, 1.6320e-04, ..., 2.6169e-03, + 9.1791e-05, 1.6584e-03], + [ 1.9379e-03, -9.8228e-04, 8.9169e-04, ..., -6.2466e-04, + 3.0375e-04, -1.1272e-03], + [ 1.1292e-03, 3.7441e-03, 2.8744e-03, ..., -2.8400e-03, + 1.3714e-03, -2.5673e-03]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0307, 0.0413, -0.0012, -0.0103, 0.0218, -0.0347, 0.0044, 0.0063, + -0.0013, -0.0150], device='cuda:0'), grad: tensor([ 0.0317, 0.0037, -0.0210, 0.0061, -0.0185, -0.0160, -0.0076, 0.0132, + 0.0119, -0.0035], device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 214.63, cls_loss 0.4796 cls_loss_mapping 0.0009 cls_loss_causal 0.4269 re_mapping 0.0067 re_causal 0.0175 /// teacc 98.98 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.0744, -0.1634, -0.1104, ..., -0.0387, 0.0656, -0.1122], + [-0.0829, -0.0949, -0.0892, ..., 0.0997, -0.0379, 0.2404], + [ 0.0023, -0.0325, -0.0628, ..., -0.0071, -0.0102, -0.0823], + ..., + [ 0.0369, -0.1203, 0.1828, ..., 0.0385, -0.1255, 0.0622], + [-0.0261, 0.0658, -0.1534, ..., -0.0482, 0.0184, -0.1199], + [-0.0641, 0.0821, 0.0421, ..., -0.0649, -0.0543, -0.0359]], + device='cuda:0'), grad: tensor([[-3.0861e-03, -4.6272e-03, 4.1664e-05, ..., -7.8278e-03, + -1.2302e-03, 7.0190e-04], + [ 2.0635e-04, 2.9182e-04, 1.5759e-04, ..., -4.6959e-03, + -1.4429e-03, -5.5170e-04], + [ 1.9145e-04, 1.9064e-03, -1.8990e-04, ..., -2.5387e-03, + 2.1160e-04, -2.9831e-03], + ..., + [ 5.2595e-04, 2.4462e-04, 1.1909e-04, ..., 1.4534e-03, + 4.3333e-05, 1.3409e-03], + [ 1.6823e-03, 1.5755e-03, -3.0518e-04, ..., 7.7248e-04, + 6.7854e-04, -1.2708e-04], + [ 1.0717e-04, -1.9875e-03, -8.3971e-04, ..., 5.9557e-04, + 1.5664e-04, -8.4229e-03]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0305, 0.0413, -0.0011, -0.0105, 0.0218, -0.0347, 0.0044, 0.0063, + -0.0015, -0.0150], device='cuda:0'), grad: tensor([-0.0398, -0.0067, -0.0042, -0.0406, 0.0114, 0.0607, 0.0202, 0.0163, + -0.0044, -0.0131], device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 215.12, cls_loss 0.5054 cls_loss_mapping 0.0010 cls_loss_causal 0.4313 re_mapping 0.0068 re_causal 0.0175 /// teacc 99.00 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.0744, -0.1635, -0.1105, ..., -0.0387, 0.0657, -0.1123], + [-0.0830, -0.0949, -0.0893, ..., 0.0996, -0.0380, 0.2404], + [ 0.0024, -0.0324, -0.0629, ..., -0.0070, -0.0102, -0.0823], + ..., + [ 0.0368, -0.1205, 0.1826, ..., 0.0384, -0.1258, 0.0620], + [-0.0262, 0.0658, -0.1534, ..., -0.0484, 0.0183, -0.1200], + [-0.0641, 0.0823, 0.0423, ..., -0.0649, -0.0543, -0.0358]], + device='cuda:0'), grad: tensor([[ 9.1267e-04, 1.1234e-03, 1.8454e-04, ..., 1.7653e-03, + 3.8862e-04, 2.7966e-04], + [ 1.8778e-03, -4.2267e-03, -6.0387e-03, ..., -2.9678e-03, + 4.8876e-05, -1.5755e-03], + [ 1.1711e-03, 1.6251e-03, -5.0783e-04, ..., -1.6384e-03, + 4.1342e-04, 5.7459e-04], + ..., + [ 6.5994e-03, 2.1305e-03, 3.0670e-03, ..., 8.8120e-03, + 1.7810e-04, 3.1967e-03], + [ 1.0595e-03, -2.6608e-03, -8.2397e-04, ..., -8.6117e-04, + 2.1732e-04, -3.1929e-03], + [-1.1551e-02, 1.7262e-03, 1.7805e-03, ..., -6.5422e-03, + -2.5964e-04, -1.1358e-03]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0306, 0.0412, -0.0010, -0.0105, 0.0218, -0.0346, 0.0044, 0.0063, + -0.0016, -0.0150], device='cuda:0'), grad: tensor([ 0.0142, -0.0114, 0.0037, 0.0420, 0.0173, -0.0139, -0.0185, 0.0164, + -0.0022, -0.0475], device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 214.77, cls_loss 0.4977 cls_loss_mapping 0.0009 cls_loss_causal 0.4244 re_mapping 0.0070 re_causal 0.0177 /// teacc 98.96 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.0743, -0.1635, -0.1105, ..., -0.0386, 0.0655, -0.1124], + [-0.0830, -0.0949, -0.0894, ..., 0.0997, -0.0380, 0.2404], + [ 0.0024, -0.0326, -0.0630, ..., -0.0071, -0.0104, -0.0823], + ..., + [ 0.0366, -0.1205, 0.1826, ..., 0.0385, -0.1255, 0.0620], + [-0.0261, 0.0659, -0.1532, ..., -0.0484, 0.0185, -0.1200], + [-0.0641, 0.0823, 0.0423, ..., -0.0650, -0.0545, -0.0356]], + device='cuda:0'), grad: tensor([[ 0.0004, -0.0004, 0.0003, ..., -0.0139, 0.0000, 0.0006], + [ 0.0002, 0.0004, 0.0003, ..., 0.0019, 0.0000, -0.0020], + [-0.0048, -0.0035, -0.0014, ..., -0.0047, 0.0000, -0.0023], + ..., + [ 0.0005, 0.0022, -0.0002, ..., 0.0021, 0.0000, 0.0020], + [ 0.0012, 0.0002, 0.0002, ..., -0.0017, 0.0000, -0.0002], + [ 0.0008, 0.0008, 0.0004, ..., 0.0036, 0.0000, 0.0010]], + device='cuda:0') +Epoch 420, bias, value: tensor([-0.0305, 0.0412, -0.0010, -0.0106, 0.0217, -0.0345, 0.0046, 0.0062, + -0.0015, -0.0149], device='cuda:0'), grad: tensor([-0.0219, 0.0174, -0.0076, 0.0338, -0.0146, 0.0215, -0.0304, 0.0240, + -0.0446, 0.0225], device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 214.80, cls_loss 0.5004 cls_loss_mapping 0.0009 cls_loss_causal 0.4443 re_mapping 0.0067 re_causal 0.0175 /// teacc 98.96 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.0743, -0.1633, -0.1104, ..., -0.0386, 0.0655, -0.1125], + [-0.0829, -0.0949, -0.0894, ..., 0.0998, -0.0379, 0.2406], + [ 0.0025, -0.0327, -0.0630, ..., -0.0071, -0.0104, -0.0821], + ..., + [ 0.0367, -0.1205, 0.1828, ..., 0.0385, -0.1255, 0.0620], + [-0.0260, 0.0658, -0.1532, ..., -0.0485, 0.0184, -0.1201], + [-0.0641, 0.0823, 0.0423, ..., -0.0650, -0.0544, -0.0357]], + device='cuda:0'), grad: tensor([[ 7.4959e-04, 4.7946e-04, 1.5497e-05, ..., 2.1095e-03, + 1.9774e-05, 2.1720e-04], + [ 1.0939e-03, 3.3426e-04, 4.2856e-05, ..., 2.9755e-03, + 6.0201e-05, 2.1803e-04], + [ 1.3266e-03, -4.2915e-03, -2.3289e-03, ..., 7.7295e-04, + -3.1433e-03, 2.4652e-04], + ..., + [ 1.1740e-03, 9.1076e-04, 5.2452e-05, ..., 2.4223e-03, + 6.8188e-05, 2.9993e-04], + [ 1.9646e-03, 6.8665e-03, 1.7500e-03, ..., 6.3133e-03, + 2.3518e-03, 1.8740e-04], + [-6.1913e-03, -9.5139e-03, 1.4758e-04, ..., -8.0795e-03, + 2.0504e-04, 1.6594e-04]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0305, 0.0412, -0.0008, -0.0108, 0.0218, -0.0346, 0.0047, 0.0063, + -0.0017, -0.0149], device='cuda:0'), grad: tensor([ 0.0102, 0.0124, -0.0074, 0.0265, 0.0112, -0.0215, -0.0478, 0.0128, + 0.0361, -0.0326], device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 214.90, cls_loss 0.5166 cls_loss_mapping 0.0009 cls_loss_causal 0.4452 re_mapping 0.0067 re_causal 0.0181 /// teacc 98.98 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.0742, -0.1634, -0.1105, ..., -0.0386, 0.0656, -0.1125], + [-0.0830, -0.0949, -0.0895, ..., 0.0997, -0.0381, 0.2405], + [ 0.0024, -0.0326, -0.0632, ..., -0.0070, -0.0104, -0.0823], + ..., + [ 0.0367, -0.1206, 0.1829, ..., 0.0385, -0.1256, 0.0620], + [-0.0261, 0.0658, -0.1531, ..., -0.0485, 0.0186, -0.1201], + [-0.0641, 0.0824, 0.0423, ..., -0.0650, -0.0545, -0.0357]], + device='cuda:0'), grad: tensor([[ 1.3936e-04, 1.1772e-04, 5.9992e-05, ..., 4.8280e-04, + -3.0007e-06, 2.4414e-04], + [ 1.4462e-05, -4.7183e-04, 3.7646e-04, ..., 2.6059e-04, + 2.8126e-07, 7.7295e-04], + [-8.7768e-06, -4.9019e-04, 1.5459e-03, ..., 4.9973e-03, + 1.0747e-06, 5.1460e-03], + ..., + [-5.2214e-05, 1.2839e-04, 7.2908e-04, ..., -4.4174e-03, + 3.7253e-09, -2.6188e-03], + [ 1.5783e-04, 1.3363e-04, -4.8180e-03, ..., -4.5662e-03, + 4.8019e-06, -7.2403e-03], + [ 7.2420e-05, 1.1963e-04, 1.3733e-03, ..., 1.4935e-03, + 2.9020e-06, 2.0828e-03]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0306, 0.0412, -0.0008, -0.0106, 0.0218, -0.0345, 0.0045, 0.0062, + -0.0017, -0.0149], device='cuda:0'), grad: tensor([ 0.0072, -0.0224, -0.0156, 0.0084, 0.0085, -0.0218, 0.0322, 0.0121, + -0.0228, 0.0141], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 421---------------------------------------------------- +epoch 421, time 231.11, cls_loss 0.5121 cls_loss_mapping 0.0009 cls_loss_causal 0.4512 re_mapping 0.0068 re_causal 0.0176 /// teacc 99.02 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.0742, -0.1635, -0.1105, ..., -0.0386, 0.0657, -0.1124], + [-0.0831, -0.0948, -0.0895, ..., 0.0996, -0.0382, 0.2405], + [ 0.0025, -0.0326, -0.0631, ..., -0.0069, -0.0101, -0.0823], + ..., + [ 0.0366, -0.1206, 0.1828, ..., 0.0383, -0.1257, 0.0619], + [-0.0261, 0.0658, -0.1531, ..., -0.0485, 0.0186, -0.1201], + [-0.0642, 0.0824, 0.0424, ..., -0.0648, -0.0545, -0.0355]], + device='cuda:0'), grad: tensor([[ 3.0790e-06, 1.7524e-05, 1.7333e-04, ..., 1.0805e-03, + 2.1577e-05, 4.6659e-04], + [ 1.6868e-05, 2.3082e-05, 1.7333e-04, ..., 8.3876e-04, + 1.2137e-05, 4.0865e-04], + [ 2.5719e-05, 1.4019e-04, 3.5167e-04, ..., -1.1501e-03, + -1.4603e-04, 2.8968e-04], + ..., + [-1.4432e-05, -4.1164e-07, -2.3975e-03, ..., -1.0979e-02, + 2.7806e-05, -3.8948e-03], + [ 4.8250e-05, 1.7233e-03, 2.3098e-03, ..., 4.9448e-04, + 5.8794e-04, -1.5612e-03], + [-2.0921e-04, -1.8568e-03, 9.8133e-04, ..., 2.9259e-03, + 5.7817e-05, 1.7347e-03]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0305, 0.0412, -0.0007, -0.0106, 0.0219, -0.0346, 0.0044, 0.0061, + -0.0017, -0.0148], device='cuda:0'), grad: tensor([ 0.0075, 0.0072, -0.0142, 0.0383, 0.0103, -0.0211, 0.0072, -0.0330, + -0.0160, 0.0138], device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 214.56, cls_loss 0.4926 cls_loss_mapping 0.0009 cls_loss_causal 0.4374 re_mapping 0.0066 re_causal 0.0176 /// teacc 98.99 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.0739, -0.1634, -0.1105, ..., -0.0386, 0.0658, -0.1125], + [-0.0831, -0.0947, -0.0894, ..., 0.0997, -0.0382, 0.2405], + [ 0.0026, -0.0325, -0.0630, ..., -0.0068, -0.0102, -0.0821], + ..., + [ 0.0366, -0.1205, 0.1828, ..., 0.0383, -0.1259, 0.0618], + [-0.0262, 0.0658, -0.1531, ..., -0.0486, 0.0186, -0.1202], + [-0.0642, 0.0826, 0.0424, ..., -0.0648, -0.0545, -0.0356]], + device='cuda:0'), grad: tensor([[ 6.1005e-05, 2.1422e-04, 1.3602e-04, ..., 9.2793e-04, + -1.0338e-06, 2.5582e-04], + [ 6.5446e-05, -4.0627e-04, 2.3103e-04, ..., 1.3037e-03, + 3.7253e-09, 3.7026e-04], + [ 1.6975e-03, 8.2397e-04, 1.4248e-03, ..., 8.8348e-03, + 5.9605e-08, 1.9369e-03], + ..., + [-2.3022e-03, 3.1185e-04, -1.5879e-03, ..., -1.0750e-02, + 4.8429e-08, -2.1725e-03], + [ 1.4901e-04, -8.7261e-04, 3.3283e-04, ..., 2.0218e-03, + 1.5832e-07, 6.0940e-04], + [ 1.7285e-05, -6.7234e-04, 1.5676e-04, ..., 1.5144e-03, + 2.6450e-07, 4.8447e-04]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0305, 0.0411, -0.0006, -0.0106, 0.0218, -0.0345, 0.0043, 0.0060, + -0.0017, -0.0148], device='cuda:0'), grad: tensor([ 0.0165, -0.0154, 0.0349, 0.0219, -0.0470, 0.0219, -0.0155, -0.0072, + 0.0050, -0.0151], device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 214.85, cls_loss 0.5143 cls_loss_mapping 0.0009 cls_loss_causal 0.4456 re_mapping 0.0066 re_causal 0.0178 /// teacc 98.97 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.0740, -0.1635, -0.1106, ..., -0.0387, 0.0658, -0.1124], + [-0.0830, -0.0948, -0.0894, ..., 0.1000, -0.0382, 0.2405], + [ 0.0023, -0.0325, -0.0630, ..., -0.0070, -0.0102, -0.0822], + ..., + [ 0.0367, -0.1206, 0.1830, ..., 0.0383, -0.1259, 0.0620], + [-0.0260, 0.0659, -0.1530, ..., -0.0484, 0.0185, -0.1202], + [-0.0644, 0.0826, 0.0422, ..., -0.0648, -0.0544, -0.0357]], + device='cuda:0'), grad: tensor([[ 0.0011, 0.0003, 0.0005, ..., 0.0032, 0.0004, 0.0020], + [ 0.0021, -0.0002, 0.0011, ..., -0.0015, 0.0008, -0.0030], + [ 0.0013, 0.0008, 0.0030, ..., 0.0039, 0.0259, 0.0027], + ..., + [ 0.0017, 0.0007, -0.0019, ..., 0.0013, 0.0004, -0.0053], + [-0.0052, -0.0022, -0.0042, ..., -0.0090, -0.0027, -0.0029], + [ 0.0004, -0.0006, -0.0013, ..., 0.0026, 0.0003, 0.0007]], + device='cuda:0') +Epoch 425, bias, value: tensor([-0.0307, 0.0413, -0.0009, -0.0106, 0.0220, -0.0346, 0.0043, 0.0061, + -0.0015, -0.0148], device='cuda:0'), grad: tensor([ 0.0217, -0.0266, 0.0535, -0.0322, -0.0013, 0.0198, 0.0185, -0.0043, + -0.0692, 0.0201], device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 214.79, cls_loss 0.4620 cls_loss_mapping 0.0008 cls_loss_causal 0.3966 re_mapping 0.0067 re_causal 0.0171 /// teacc 99.01 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.0740, -0.1636, -0.1106, ..., -0.0386, 0.0662, -0.1123], + [-0.0829, -0.0948, -0.0895, ..., 0.1000, -0.0382, 0.2405], + [ 0.0025, -0.0325, -0.0630, ..., -0.0069, -0.0103, -0.0821], + ..., + [ 0.0367, -0.1206, 0.1829, ..., 0.0382, -0.1259, 0.0618], + [-0.0261, 0.0661, -0.1529, ..., -0.0484, 0.0186, -0.1203], + [-0.0643, 0.0825, 0.0423, ..., -0.0648, -0.0546, -0.0357]], + device='cuda:0'), grad: tensor([[ 3.3647e-05, -7.7486e-04, 1.2100e-05, ..., 1.1578e-05, + 6.2275e-04, 2.7180e-04], + [-3.5071e-04, 1.2871e-06, -7.3075e-05, ..., -4.9095e-03, + -3.7556e-03, 9.0313e-04], + [ 9.4235e-05, 4.7398e-04, 2.0266e-04, ..., -6.0043e-03, + 8.6355e-04, -4.2992e-03], + ..., + [ 6.0439e-05, 2.2852e-04, -1.1057e-04, ..., 5.8022e-03, + 7.1430e-04, 3.3188e-04], + [ 3.3021e-05, 2.9349e-04, 1.6794e-05, ..., 1.4734e-03, + 7.2670e-04, 4.6182e-04], + [ 6.4254e-05, 2.5272e-04, 4.6402e-05, ..., 1.5354e-03, + 6.0701e-04, 3.9244e-04]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0306, 0.0413, -0.0009, -0.0106, 0.0220, -0.0348, 0.0043, 0.0060, + -0.0015, -0.0148], device='cuda:0'), grad: tensor([-0.0165, -0.0399, -0.0094, 0.0239, 0.0221, -0.0408, -0.0125, 0.0346, + 0.0182, 0.0203], device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 214.96, cls_loss 0.4991 cls_loss_mapping 0.0008 cls_loss_causal 0.4288 re_mapping 0.0065 re_causal 0.0176 /// teacc 99.02 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.0741, -0.1636, -0.1105, ..., -0.0386, 0.0662, -0.1124], + [-0.0829, -0.0949, -0.0896, ..., 0.0998, -0.0382, 0.2404], + [ 0.0026, -0.0326, -0.0630, ..., -0.0068, -0.0104, -0.0820], + ..., + [ 0.0366, -0.1207, 0.1828, ..., 0.0382, -0.1260, 0.0617], + [-0.0263, 0.0660, -0.1531, ..., -0.0484, 0.0185, -0.1204], + [-0.0642, 0.0825, 0.0423, ..., -0.0648, -0.0546, -0.0357]], + device='cuda:0'), grad: tensor([[ 8.4448e-04, 7.4673e-04, 5.5730e-06, ..., -1.2535e-02, + 0.0000e+00, 1.3566e-04], + [-7.2517e-03, 4.3678e-04, 3.6925e-05, ..., 2.5673e-03, + 0.0000e+00, -1.9753e-04], + [ 6.0368e-04, 4.4394e-04, 5.7667e-05, ..., 2.3670e-03, + 0.0000e+00, 1.7405e-04], + ..., + [ 7.3624e-04, 7.0095e-04, 6.0701e-04, ..., 6.7940e-03, + 0.0000e+00, 2.2488e-03], + [ 7.3051e-04, -5.5462e-05, -2.9969e-04, ..., -5.0316e-03, + 0.0000e+00, -2.7599e-03], + [ 9.3985e-04, -2.5425e-03, -4.0779e-03, ..., 2.6417e-03, + 0.0000e+00, -3.7050e-04]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0305, 0.0413, -0.0009, -0.0107, 0.0222, -0.0348, 0.0044, 0.0059, + -0.0016, -0.0148], device='cuda:0'), grad: tensor([-0.0060, -0.0025, 0.0245, 0.0280, 0.0357, -0.0720, -0.0078, 0.0137, + -0.0026, -0.0110], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 426---------------------------------------------------- +epoch 426, time 231.28, cls_loss 0.4901 cls_loss_mapping 0.0008 cls_loss_causal 0.4202 re_mapping 0.0067 re_causal 0.0175 /// teacc 99.03 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.0740, -0.1636, -0.1105, ..., -0.0384, 0.0661, -0.1123], + [-0.0830, -0.0947, -0.0897, ..., 0.0999, -0.0383, 0.2404], + [ 0.0028, -0.0326, -0.0629, ..., -0.0068, -0.0102, -0.0820], + ..., + [ 0.0366, -0.1209, 0.1828, ..., 0.0382, -0.1261, 0.0618], + [-0.0262, 0.0661, -0.1531, ..., -0.0482, 0.0186, -0.1204], + [-0.0644, 0.0826, 0.0424, ..., -0.0649, -0.0548, -0.0357]], + device='cuda:0'), grad: tensor([[-1.5631e-03, -9.2268e-04, 3.7408e-04, ..., 1.0633e-03, + 2.1434e-04, 8.9407e-04], + [-6.2180e-03, 2.4557e-04, -4.9744e-03, ..., -7.2327e-03, + -5.1641e-04, -6.5565e-05], + [ 2.3937e-03, 5.5408e-04, 6.2084e-04, ..., 3.1929e-03, + 9.8419e-04, 1.4124e-03], + ..., + [-5.8708e-03, -1.5564e-03, 8.0109e-04, ..., -6.5460e-03, + -3.6411e-03, -2.4853e-03], + [ 3.6144e-03, 8.8882e-04, 3.5286e-04, ..., 1.7576e-03, + 1.0958e-03, 2.4929e-03], + [ 9.7418e-04, 4.2528e-05, -2.3818e-04, ..., -1.5602e-03, + 2.3580e-04, -8.5220e-03]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0306, 0.0414, -0.0009, -0.0109, 0.0222, -0.0346, 0.0045, 0.0058, + -0.0016, -0.0148], device='cuda:0'), grad: tensor([ 0.0088, -0.0132, 0.0216, 0.0245, 0.0198, -0.0105, -0.0140, -0.0449, + 0.0247, -0.0167], device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 214.82, cls_loss 0.4723 cls_loss_mapping 0.0009 cls_loss_causal 0.4176 re_mapping 0.0066 re_causal 0.0174 /// teacc 99.01 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.0739, -0.1635, -0.1105, ..., -0.0382, 0.0662, -0.1123], + [-0.0830, -0.0948, -0.0898, ..., 0.0999, -0.0383, 0.2404], + [ 0.0027, -0.0327, -0.0628, ..., -0.0068, -0.0102, -0.0821], + ..., + [ 0.0369, -0.1207, 0.1827, ..., 0.0382, -0.1260, 0.0619], + [-0.0261, 0.0661, -0.1530, ..., -0.0483, 0.0185, -0.1205], + [-0.0645, 0.0826, 0.0425, ..., -0.0649, -0.0549, -0.0356]], + device='cuda:0'), grad: tensor([[ 1.8582e-05, 3.1543e-04, 4.2379e-05, ..., 6.6376e-04, + 5.8338e-06, 5.4836e-04], + [ 5.8937e-04, 1.2302e-03, 8.0109e-04, ..., 1.0529e-03, + 1.7309e-04, 1.7042e-03], + [-2.3727e-03, 7.3290e-04, 1.5211e-04, ..., 1.4009e-03, + 6.6385e-06, 1.5628e-04], + ..., + [ 4.1819e-04, -1.5326e-03, 7.5912e-04, ..., -4.4365e-03, + 1.9968e-05, -2.3403e-03], + [ 2.0180e-03, 1.4114e-03, -3.7456e-04, ..., -2.1610e-03, + 1.1466e-05, 1.4200e-03], + [-2.5964e-04, 3.3379e-04, -8.6260e-04, ..., 3.1590e-04, + 1.3316e-04, -1.6842e-03]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0306, 0.0414, -0.0009, -0.0109, 0.0221, -0.0346, 0.0043, 0.0059, + -0.0016, -0.0147], device='cuda:0'), grad: tensor([ 0.0068, 0.0169, 0.0087, 0.0052, 0.0079, 0.0067, -0.0251, -0.0319, + 0.0057, -0.0009], device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 214.56, cls_loss 0.4734 cls_loss_mapping 0.0007 cls_loss_causal 0.4129 re_mapping 0.0066 re_causal 0.0174 /// teacc 99.01 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.0739, -0.1636, -0.1107, ..., -0.0381, 0.0662, -0.1122], + [-0.0831, -0.0948, -0.0897, ..., 0.0999, -0.0382, 0.2403], + [ 0.0028, -0.0326, -0.0628, ..., -0.0068, -0.0103, -0.0821], + ..., + [ 0.0370, -0.1207, 0.1828, ..., 0.0383, -0.1261, 0.0620], + [-0.0261, 0.0661, -0.1530, ..., -0.0482, 0.0186, -0.1204], + [-0.0647, 0.0826, 0.0424, ..., -0.0650, -0.0551, -0.0358]], + device='cuda:0'), grad: tensor([[ 1.6578e-07, -6.6459e-06, 5.8487e-06, ..., 2.3949e-04, + -1.6773e-04, 3.7491e-05], + [-5.5470e-06, 1.3337e-06, 3.8171e-04, ..., 5.6992e-03, + 7.0781e-08, 5.5885e-03], + [ 2.8312e-07, 3.2391e-06, 4.1509e-04, ..., 7.7820e-04, + 1.1176e-07, 1.1677e-04], + ..., + [ 4.3027e-07, 7.4692e-07, -2.7447e-03, ..., -6.4430e-03, + 6.3330e-08, -5.8022e-03], + [-8.7991e-06, 1.2472e-05, 2.9778e-04, ..., 5.6839e-04, + 1.3057e-06, 1.0264e-04], + [ 4.1723e-07, 2.1383e-06, 1.6046e-04, ..., 3.5305e-03, + 2.2613e-06, 2.7084e-03]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0306, 0.0415, -0.0009, -0.0111, 0.0220, -0.0344, 0.0044, 0.0059, + -0.0015, -0.0149], device='cuda:0'), grad: tensor([ 0.0086, 0.0157, 0.0083, 0.0055, -0.0364, 0.0101, 0.0228, -0.0352, + 0.0057, -0.0051], device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 214.81, cls_loss 0.4506 cls_loss_mapping 0.0009 cls_loss_causal 0.3870 re_mapping 0.0065 re_causal 0.0162 /// teacc 98.97 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.0740, -0.1635, -0.1107, ..., -0.0382, 0.0661, -0.1123], + [-0.0830, -0.0949, -0.0897, ..., 0.1000, -0.0383, 0.2404], + [ 0.0029, -0.0326, -0.0628, ..., -0.0067, -0.0103, -0.0820], + ..., + [ 0.0371, -0.1207, 0.1829, ..., 0.0383, -0.1259, 0.0621], + [-0.0262, 0.0660, -0.1530, ..., -0.0483, 0.0187, -0.1205], + [-0.0648, 0.0827, 0.0425, ..., -0.0649, -0.0550, -0.0359]], + device='cuda:0'), grad: tensor([[ 8.6832e-04, 3.0454e-06, 9.0152e-07, ..., 5.0688e-04, + 0.0000e+00, 1.2051e-06], + [ 1.7905e-04, 1.4007e-06, 3.9451e-06, ..., 3.0971e-04, + 0.0000e+00, -2.8163e-06], + [ 1.2970e-03, 3.6228e-06, 4.3847e-06, ..., -4.0665e-03, + 0.0000e+00, 4.1202e-06], + ..., + [ 2.7585e-04, 2.9989e-07, -6.4373e-05, ..., 1.1997e-03, + 0.0000e+00, -6.9797e-05], + [ 1.3990e-03, 5.6028e-05, 4.4703e-06, ..., 6.5804e-04, + 0.0000e+00, -6.9141e-06], + [-7.7400e-03, 1.6898e-05, 6.3896e-05, ..., 5.0879e-04, + 0.0000e+00, 6.5088e-05]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0306, 0.0417, -0.0008, -0.0111, 0.0219, -0.0345, 0.0045, 0.0061, + -0.0016, -0.0150], device='cuda:0'), grad: tensor([ 0.0102, -0.0188, -0.0168, 0.0088, 0.0139, -0.0149, 0.0125, 0.0090, + 0.0148, -0.0186], device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 214.65, cls_loss 0.4913 cls_loss_mapping 0.0008 cls_loss_causal 0.4241 re_mapping 0.0064 re_causal 0.0172 /// teacc 98.96 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.0740, -0.1636, -0.1108, ..., -0.0382, 0.0661, -0.1122], + [-0.0830, -0.0949, -0.0898, ..., 0.1000, -0.0382, 0.2403], + [ 0.0029, -0.0325, -0.0629, ..., -0.0068, -0.0102, -0.0821], + ..., + [ 0.0371, -0.1209, 0.1829, ..., 0.0384, -0.1260, 0.0622], + [-0.0262, 0.0661, -0.1529, ..., -0.0483, 0.0187, -0.1205], + [-0.0647, 0.0828, 0.0426, ..., -0.0649, -0.0552, -0.0358]], + device='cuda:0'), grad: tensor([[ 8.1956e-05, 1.2589e-04, 1.7452e-04, ..., 2.6345e-04, + 1.1873e-04, 2.4021e-04], + [ 3.8952e-05, 4.7594e-05, 7.7665e-05, ..., -2.4509e-04, + 4.8667e-05, -7.5388e-04], + [-5.7650e-04, 9.1732e-05, 1.2267e-04, ..., -1.2321e-03, + 8.3089e-05, 4.4405e-05], + ..., + [-2.2203e-06, 3.3498e-05, -9.0711e-07, ..., 4.1032e-04, + 3.5048e-05, -9.9277e-04], + [ 1.1611e-04, 3.6001e-04, 1.7905e-04, ..., 3.1066e-04, + 1.2136e-04, 3.6979e-04], + [ 3.2902e-05, 4.0859e-05, 5.3465e-05, ..., 1.8024e-04, + 2.6062e-05, 5.5170e-04]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0307, 0.0415, -0.0008, -0.0111, 0.0220, -0.0345, 0.0045, 0.0061, + -0.0016, -0.0149], device='cuda:0'), grad: tensor([ 0.0073, -0.0224, -0.0213, -0.0130, 0.0122, 0.0306, 0.0070, -0.0246, + 0.0100, 0.0142], device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 214.83, cls_loss 0.4837 cls_loss_mapping 0.0010 cls_loss_causal 0.4295 re_mapping 0.0065 re_causal 0.0171 /// teacc 98.94 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.0741, -0.1637, -0.1108, ..., -0.0383, 0.0661, -0.1122], + [-0.0830, -0.0950, -0.0898, ..., 0.1000, -0.0383, 0.2403], + [ 0.0028, -0.0327, -0.0629, ..., -0.0068, -0.0102, -0.0821], + ..., + [ 0.0370, -0.1209, 0.1829, ..., 0.0383, -0.1259, 0.0621], + [-0.0261, 0.0663, -0.1529, ..., -0.0483, 0.0187, -0.1205], + [-0.0647, 0.0828, 0.0426, ..., -0.0649, -0.0552, -0.0359]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, 1.5879e-04, 1.2182e-06, ..., 9.6464e-04, + 8.2207e-04, 1.7939e-03], + [ 2.4065e-06, 1.4998e-05, 6.5155e-06, ..., 1.0557e-03, + 7.3290e-04, 8.5115e-05], + [ 2.6152e-05, 1.4341e-04, 5.9158e-05, ..., 8.5592e-04, + 5.3978e-04, 1.6224e-04], + ..., + [ 2.0582e-06, 1.9819e-05, 3.8780e-06, ..., 6.1655e-04, + 3.6216e-04, -5.0392e-03], + [ 2.4363e-06, -3.5152e-03, 6.8173e-06, ..., 1.1559e-03, + -3.0289e-03, 1.2684e-04], + [ 1.9930e-07, 1.7333e-04, -5.9336e-05, ..., 7.4005e-04, + 7.1669e-04, 2.6665e-03]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0308, 0.0414, -0.0008, -0.0109, 0.0218, -0.0345, 0.0046, 0.0061, + -0.0016, -0.0148], device='cuda:0'), grad: tensor([-0.0175, 0.0060, 0.0172, -0.0200, 0.0126, -0.0129, 0.0210, -0.0246, + 0.0042, 0.0139], device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 214.60, cls_loss 0.5035 cls_loss_mapping 0.0010 cls_loss_causal 0.4349 re_mapping 0.0063 re_causal 0.0169 /// teacc 98.97 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.0741, -0.1636, -0.1110, ..., -0.0384, 0.0661, -0.1122], + [-0.0828, -0.0950, -0.0899, ..., 0.1001, -0.0382, 0.2404], + [ 0.0027, -0.0327, -0.0629, ..., -0.0067, -0.0102, -0.0821], + ..., + [ 0.0370, -0.1210, 0.1828, ..., 0.0382, -0.1259, 0.0619], + [-0.0262, 0.0663, -0.1528, ..., -0.0484, 0.0186, -0.1205], + [-0.0646, 0.0828, 0.0426, ..., -0.0648, -0.0553, -0.0356]], + device='cuda:0'), grad: tensor([[ 1.2153e-04, 3.1829e-04, 1.0854e-04, ..., 1.1148e-03, + 8.5354e-04, 5.6601e-04], + [ 1.2267e-04, 2.6059e-04, 8.5711e-05, ..., 7.8659e-03, + 3.7909e-04, 3.3684e-03], + [ 7.8249e-04, -7.5226e-03, -1.1730e-03, ..., -6.0415e-04, + 1.2951e-03, 5.5075e-04], + ..., + [-2.7065e-03, -2.6970e-03, 3.5024e-04, ..., -4.2648e-03, + -1.4992e-03, 2.6250e-04], + [ 2.9874e-04, -1.1787e-03, 2.5082e-04, ..., 1.9760e-03, + -3.3550e-03, 7.2908e-04], + [ 5.5075e-04, 3.9444e-03, 1.9550e-03, ..., 2.1477e-03, + 9.5892e-04, 1.5125e-03]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0309, 0.0416, -0.0008, -0.0109, 0.0218, -0.0346, 0.0045, 0.0061, + -0.0017, -0.0147], device='cuda:0'), grad: tensor([ 0.0175, 0.0229, -0.0153, 0.0249, -0.0296, 0.0374, -0.0242, -0.0489, + 0.0223, -0.0071], device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 214.69, cls_loss 0.4694 cls_loss_mapping 0.0010 cls_loss_causal 0.4016 re_mapping 0.0064 re_causal 0.0169 /// teacc 98.97 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.0742, -0.1637, -0.1108, ..., -0.0384, 0.0661, -0.1123], + [-0.0829, -0.0951, -0.0901, ..., 0.1000, -0.0384, 0.2404], + [ 0.0027, -0.0327, -0.0629, ..., -0.0067, -0.0102, -0.0822], + ..., + [ 0.0370, -0.1209, 0.1829, ..., 0.0382, -0.1259, 0.0619], + [-0.0260, 0.0662, -0.1528, ..., -0.0483, 0.0186, -0.1207], + [-0.0646, 0.0828, 0.0425, ..., -0.0649, -0.0554, -0.0356]], + device='cuda:0'), grad: tensor([[ 4.8256e-03, 5.1212e-04, 8.8274e-05, ..., 5.6190e-03, + 5.9175e-04, 7.3099e-04], + [ 1.2999e-03, 4.9067e-04, 7.3099e-04, ..., -6.9962e-03, + -1.4668e-03, 3.5620e-04], + [ 5.4979e-04, 4.8876e-04, 1.1665e-04, ..., 2.5940e-03, + 2.2924e-04, 5.1260e-04], + ..., + [ 1.8005e-03, -1.1492e-03, 7.1335e-03, ..., -4.9877e-04, + 2.2948e-04, 7.2060e-03], + [-9.3508e-04, -1.1187e-03, 1.4389e-04, ..., -4.0588e-03, + 7.3552e-05, -2.3251e-03], + [-8.9121e-04, 3.7551e-04, -8.7051e-03, ..., -2.1541e-04, + 4.3917e-04, -7.7972e-03]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0309, 0.0417, -0.0009, -0.0109, 0.0216, -0.0346, 0.0045, 0.0062, + -0.0017, -0.0147], device='cuda:0'), grad: tensor([ 0.0289, -0.0079, 0.0146, 0.0188, 0.0199, -0.0463, -0.0022, 0.0164, + -0.0036, -0.0386], device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 214.09, cls_loss 0.4927 cls_loss_mapping 0.0008 cls_loss_causal 0.4379 re_mapping 0.0065 re_causal 0.0175 /// teacc 98.91 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.0744, -0.1638, -0.1109, ..., -0.0385, 0.0660, -0.1122], + [-0.0828, -0.0952, -0.0900, ..., 0.1000, -0.0384, 0.2404], + [ 0.0026, -0.0328, -0.0630, ..., -0.0068, -0.0103, -0.0822], + ..., + [ 0.0371, -0.1208, 0.1829, ..., 0.0382, -0.1259, 0.0619], + [-0.0260, 0.0663, -0.1529, ..., -0.0483, 0.0187, -0.1207], + [-0.0646, 0.0830, 0.0425, ..., -0.0648, -0.0554, -0.0356]], + device='cuda:0'), grad: tensor([[ 5.5695e-04, 2.3931e-05, 4.3321e-04, ..., 5.4550e-04, + 3.5286e-04, 3.8415e-05], + [-4.4365e-03, -2.0866e-03, 6.5956e-03, ..., 4.2877e-03, + -4.2458e-03, 1.0269e-02], + [ 6.1989e-04, 3.5143e-04, 1.2004e-04, ..., 1.7462e-03, + 5.0259e-04, 1.7703e-04], + ..., + [ 4.2305e-03, 1.8883e-03, 2.6474e-03, ..., -6.4964e-03, + 4.6563e-04, -1.1040e-02], + [ 7.4482e-04, 1.6463e-04, 1.1797e-03, ..., 2.0008e-03, + 3.9530e-04, 1.5116e-03], + [ 7.2479e-04, 2.3508e-04, 6.0120e-03, ..., 1.1597e-03, + 5.4693e-04, 7.8487e-04]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0309, 0.0416, -0.0010, -0.0108, 0.0218, -0.0346, 0.0044, 0.0062, + -0.0017, -0.0146], device='cuda:0'), grad: tensor([ 0.0269, -0.0501, 0.0117, -0.0123, -0.0022, -0.0229, 0.0115, 0.0274, + 0.0149, -0.0050], device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 214.33, cls_loss 0.4858 cls_loss_mapping 0.0010 cls_loss_causal 0.4282 re_mapping 0.0063 re_causal 0.0166 /// teacc 98.99 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.0743, -0.1638, -0.1110, ..., -0.0385, 0.0660, -0.1121], + [-0.0827, -0.0952, -0.0900, ..., 0.1000, -0.0383, 0.2403], + [ 0.0027, -0.0327, -0.0628, ..., -0.0068, -0.0102, -0.0822], + ..., + [ 0.0370, -0.1210, 0.1828, ..., 0.0382, -0.1259, 0.0619], + [-0.0259, 0.0665, -0.1530, ..., -0.0483, 0.0187, -0.1206], + [-0.0646, 0.0830, 0.0427, ..., -0.0648, -0.0556, -0.0357]], + device='cuda:0'), grad: tensor([[ 1.6904e-04, -6.0081e-04, 1.3816e-04, ..., 8.9502e-04, + 8.3685e-05, 4.2105e-04], + [ 8.0228e-05, -1.6174e-03, -1.1530e-03, ..., 3.5400e-03, + 1.1034e-05, 1.4172e-03], + [ 1.7202e-04, 4.1938e-04, -2.9404e-02, ..., 1.5860e-03, + 4.2409e-05, 1.0223e-03], + ..., + [ 7.9393e-04, 9.7418e-04, 2.0020e-02, ..., 3.5057e-03, + 1.0854e-04, 8.9264e-04], + [ 1.4639e-03, 5.5981e-04, 7.1907e-04, ..., 5.8784e-03, + 1.7241e-05, 2.0733e-03], + [ 6.2466e-04, -1.4706e-03, -1.0166e-03, ..., 2.7065e-03, + -3.6383e-04, -3.7241e-04]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0308, 0.0416, -0.0009, -0.0109, 0.0217, -0.0346, 0.0045, 0.0062, + -0.0016, -0.0147], device='cuda:0'), grad: tensor([-0.0109, -0.0518, -0.0143, 0.0308, 0.0229, -0.0108, -0.0578, 0.0412, + 0.0341, 0.0165], device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 214.33, cls_loss 0.5079 cls_loss_mapping 0.0009 cls_loss_causal 0.4379 re_mapping 0.0062 re_causal 0.0169 /// teacc 99.00 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.0744, -0.1639, -0.1111, ..., -0.0386, 0.0659, -0.1120], + [-0.0828, -0.0952, -0.0900, ..., 0.1000, -0.0382, 0.2403], + [ 0.0026, -0.0327, -0.0628, ..., -0.0066, -0.0102, -0.0821], + ..., + [ 0.0370, -0.1212, 0.1828, ..., 0.0383, -0.1260, 0.0619], + [-0.0259, 0.0665, -0.1531, ..., -0.0485, 0.0188, -0.1207], + [-0.0646, 0.0831, 0.0427, ..., -0.0648, -0.0557, -0.0355]], + device='cuda:0'), grad: tensor([[ 1.3709e-04, -4.7231e-04, 9.9480e-05, ..., 1.7090e-03, + 7.3016e-05, 1.8382e-04], + [ 5.5742e-04, -1.0386e-03, 2.5868e-04, ..., 2.2068e-03, + 3.9768e-04, 6.6042e-04], + [-2.4185e-03, 4.2677e-04, 3.9043e-03, ..., 5.5618e-03, + 6.6221e-05, -3.6182e-03], + ..., + [ 2.5883e-03, 7.6342e-04, -3.9330e-03, ..., 2.1496e-03, + 7.4208e-05, 3.8013e-03], + [-1.6928e-03, -1.7519e-03, -9.4986e-04, ..., 3.8891e-03, + 7.7784e-05, -7.9155e-04], + [-1.3924e-03, 2.6741e-03, 1.2360e-03, ..., -2.7065e-03, + -1.5764e-03, -2.0523e-03]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0310, 0.0416, -0.0008, -0.0109, 0.0216, -0.0347, 0.0046, 0.0063, + -0.0017, -0.0146], device='cuda:0'), grad: tensor([-0.0107, -0.0558, 0.0386, 0.0235, 0.0248, -0.0169, -0.0367, 0.0119, + 0.0248, -0.0035], device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 214.26, cls_loss 0.4852 cls_loss_mapping 0.0010 cls_loss_causal 0.4187 re_mapping 0.0061 re_causal 0.0165 /// teacc 98.95 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.0743, -0.1640, -0.1112, ..., -0.0386, 0.0659, -0.1121], + [-0.0828, -0.0951, -0.0899, ..., 0.1000, -0.0381, 0.2403], + [ 0.0026, -0.0326, -0.0627, ..., -0.0067, -0.0102, -0.0822], + ..., + [ 0.0371, -0.1213, 0.1828, ..., 0.0385, -0.1262, 0.0620], + [-0.0260, 0.0665, -0.1532, ..., -0.0487, 0.0189, -0.1208], + [-0.0648, 0.0831, 0.0426, ..., -0.0649, -0.0558, -0.0357]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.6393e-04, 9.8050e-05, ..., 1.1578e-03, + 4.9591e-04, 2.6894e-04], + [ 1.8626e-09, -5.3024e-04, 7.7367e-05, ..., 4.1795e-04, + 3.2902e-05, -1.6556e-03], + [ 7.5065e-07, 2.6751e-04, 7.1406e-05, ..., 7.3528e-04, + -2.5196e-03, 2.1875e-04], + ..., + [ 1.3039e-08, 3.5214e-04, -8.1682e-04, ..., -8.4152e-03, + 2.1785e-05, -6.1274e-04], + [-2.8927e-06, 3.6526e-04, 8.4400e-05, ..., 1.3695e-03, + 3.3498e-04, 2.7442e-04], + [ 1.6764e-08, -2.3537e-03, 1.5604e-04, ..., -1.9951e-03, + 1.0157e-04, 3.4380e-04]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0310, 0.0416, -0.0009, -0.0108, 0.0218, -0.0346, 0.0046, 0.0063, + -0.0018, -0.0147], device='cuda:0'), grad: tensor([ 0.0213, -0.0433, -0.0133, 0.0230, 0.0160, 0.0247, 0.0291, -0.0432, + 0.0214, -0.0357], device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 214.47, cls_loss 0.5243 cls_loss_mapping 0.0009 cls_loss_causal 0.4552 re_mapping 0.0065 re_causal 0.0177 /// teacc 98.96 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.0745, -0.1639, -0.1113, ..., -0.0388, 0.0659, -0.1123], + [-0.0829, -0.0951, -0.0898, ..., 0.1001, -0.0382, 0.2404], + [ 0.0027, -0.0326, -0.0626, ..., -0.0068, -0.0101, -0.0822], + ..., + [ 0.0370, -0.1214, 0.1828, ..., 0.0385, -0.1263, 0.0620], + [-0.0261, 0.0665, -0.1533, ..., -0.0488, 0.0189, -0.1208], + [-0.0647, 0.0832, 0.0425, ..., -0.0649, -0.0559, -0.0359]], + device='cuda:0'), grad: tensor([[-4.1237e-03, -6.2883e-05, 6.9104e-06, ..., -3.8986e-03, + 8.5115e-05, -1.1663e-03], + [ 8.5497e-04, 3.6945e-03, 3.0119e-06, ..., -2.6627e-03, + 9.0748e-06, -4.4250e-03], + [-7.8278e-03, -1.8845e-03, 3.5667e-03, ..., -1.0597e-02, + 8.5980e-06, 1.2312e-03], + ..., + [ 1.1139e-03, -2.5368e-04, -4.3983e-03, ..., 1.2579e-03, + 2.7902e-06, 5.7125e-04], + [-1.5430e-03, -3.0746e-03, 3.1662e-04, ..., -1.4351e-02, + 3.1680e-05, -4.5815e-03], + [ 9.9564e-04, -8.6975e-03, -5.6744e-04, ..., 4.6425e-03, + 1.6317e-05, 8.5878e-04]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0310, 0.0417, -0.0009, -0.0108, 0.0219, -0.0347, 0.0047, 0.0063, + -0.0019, -0.0148], device='cuda:0'), grad: tensor([-0.0358, 0.0142, -0.0044, 0.0542, 0.0147, -0.0225, -0.0083, 0.0054, + -0.0296, 0.0122], device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 214.51, cls_loss 0.4752 cls_loss_mapping 0.0009 cls_loss_causal 0.4080 re_mapping 0.0062 re_causal 0.0172 /// teacc 98.95 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.0746, -0.1639, -0.1113, ..., -0.0387, 0.0659, -0.1122], + [-0.0829, -0.0949, -0.0899, ..., 0.1002, -0.0384, 0.2402], + [ 0.0027, -0.0325, -0.0627, ..., -0.0067, -0.0101, -0.0819], + ..., + [ 0.0371, -0.1214, 0.1829, ..., 0.0384, -0.1263, 0.0620], + [-0.0261, 0.0664, -0.1533, ..., -0.0490, 0.0189, -0.1208], + [-0.0648, 0.0833, 0.0424, ..., -0.0651, -0.0560, -0.0359]], + device='cuda:0'), grad: tensor([[-4.5471e-03, -6.2714e-03, -1.9350e-03, ..., -1.2756e-02, + 9.2149e-05, -3.6373e-03], + [-1.4610e-03, 6.6853e-04, 3.0184e-04, ..., 5.3167e-04, + 6.6943e-06, 1.3733e-03], + [ 6.1703e-04, 2.5082e-04, 4.3178e-04, ..., 3.2215e-03, + 3.5316e-05, 2.1172e-03], + ..., + [ 4.4441e-04, 2.0587e-04, 5.4026e-04, ..., 3.2902e-03, + 5.8651e-05, 1.4391e-03], + [ 9.6798e-04, 1.6727e-03, 3.2812e-05, ..., 2.8038e-03, + 1.8682e-03, -8.4221e-05], + [ 6.2656e-04, 5.1975e-04, 1.2398e-03, ..., 3.1509e-03, + 2.4188e-04, 1.2941e-03]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0309, 0.0416, -0.0008, -0.0108, 0.0220, -0.0346, 0.0047, 0.0063, + -0.0020, -0.0150], device='cuda:0'), grad: tensor([-0.0951, 0.0078, 0.0152, 0.0219, -0.0260, 0.0035, 0.0185, 0.0166, + 0.0176, 0.0200], device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 214.35, cls_loss 0.4906 cls_loss_mapping 0.0009 cls_loss_causal 0.4196 re_mapping 0.0064 re_causal 0.0174 /// teacc 98.99 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.0746, -0.1637, -0.1113, ..., -0.0386, 0.0660, -0.1123], + [-0.0830, -0.0951, -0.0900, ..., 0.1001, -0.0384, 0.2403], + [ 0.0028, -0.0322, -0.0626, ..., -0.0067, -0.0100, -0.0819], + ..., + [ 0.0372, -0.1215, 0.1828, ..., 0.0385, -0.1263, 0.0619], + [-0.0261, 0.0665, -0.1534, ..., -0.0491, 0.0189, -0.1208], + [-0.0649, 0.0831, 0.0423, ..., -0.0651, -0.0561, -0.0359]], + device='cuda:0'), grad: tensor([[ 9.4366e-04, 1.0514e-04, 3.5446e-06, ..., 1.0519e-03, + -3.8259e-06, 8.2552e-05], + [ 1.0765e-04, 4.5508e-05, 7.3910e-06, ..., 9.9659e-04, + 0.0000e+00, -2.7400e-06], + [ 4.7469e-04, 1.0639e-04, 1.0681e-04, ..., 7.3814e-04, + 5.9698e-07, 1.4043e-04], + ..., + [ 1.0037e-04, -3.9792e-04, -1.5259e-04, ..., -1.5087e-05, + 0.0000e+00, 6.6757e-05], + [ 8.9073e-04, 1.0794e-04, 6.1877e-06, ..., 1.4486e-03, + 0.0000e+00, 2.5773e-04], + [ 2.2292e-04, -4.0889e-04, 1.1027e-05, ..., 5.4026e-04, + 2.8461e-06, 1.3609e-03]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0308, 0.0417, -0.0007, -0.0109, 0.0219, -0.0346, 0.0047, 0.0062, + -0.0021, -0.0150], device='cuda:0'), grad: tensor([ 0.0186, 0.0195, -0.0146, 0.0188, 0.0116, -0.0354, 0.0190, -0.0471, + 0.0196, -0.0100], device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 214.32, cls_loss 0.4853 cls_loss_mapping 0.0009 cls_loss_causal 0.4109 re_mapping 0.0064 re_causal 0.0173 /// teacc 98.98 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.0746, -0.1637, -0.1114, ..., -0.0387, 0.0661, -0.1125], + [-0.0830, -0.0953, -0.0900, ..., 0.1001, -0.0384, 0.2403], + [ 0.0027, -0.0320, -0.0627, ..., -0.0068, -0.0103, -0.0820], + ..., + [ 0.0373, -0.1216, 0.1829, ..., 0.0386, -0.1263, 0.0620], + [-0.0261, 0.0664, -0.1531, ..., -0.0489, 0.0191, -0.1206], + [-0.0650, 0.0831, 0.0423, ..., -0.0652, -0.0559, -0.0359]], + device='cuda:0'), grad: tensor([[ 3.4857e-04, -1.0414e-03, 1.6177e-04, ..., -3.5858e-04, + 2.5005e-03, 1.5509e-04], + [ 2.3365e-03, 2.5487e-04, 9.4223e-04, ..., 2.8038e-03, + 4.6992e-04, 3.5882e-04], + [ 6.4421e-04, 5.9175e-04, 2.9635e-04, ..., 7.7200e-04, + 9.0694e-04, 2.0897e-04], + ..., + [ 2.4605e-04, 2.2173e-04, -8.7881e-04, ..., -5.6839e-04, + 2.5630e-04, 1.0484e-04], + [ 4.6802e-04, -2.5868e-05, 4.4131e-04, ..., 1.2703e-03, + 1.3485e-03, 1.1688e-04], + [ 3.7766e-04, 5.6362e-04, 4.5538e-04, ..., 1.1663e-03, + 7.7963e-04, 1.2267e-04]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0310, 0.0417, -0.0007, -0.0109, 0.0220, -0.0347, 0.0048, 0.0063, + -0.0019, -0.0150], device='cuda:0'), grad: tensor([-0.0418, 0.0227, 0.0172, -0.0250, -0.0100, -0.0214, 0.0140, 0.0092, + 0.0182, 0.0170], device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 215.03, cls_loss 0.4797 cls_loss_mapping 0.0008 cls_loss_causal 0.4180 re_mapping 0.0063 re_causal 0.0167 /// teacc 98.97 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.0743, -0.1637, -0.1114, ..., -0.0387, 0.0661, -0.1126], + [-0.0831, -0.0953, -0.0902, ..., 0.1000, -0.0385, 0.2403], + [ 0.0027, -0.0320, -0.0628, ..., -0.0067, -0.0104, -0.0819], + ..., + [ 0.0373, -0.1215, 0.1830, ..., 0.0387, -0.1261, 0.0620], + [-0.0262, 0.0663, -0.1532, ..., -0.0490, 0.0191, -0.1206], + [-0.0649, 0.0831, 0.0425, ..., -0.0649, -0.0557, -0.0358]], + device='cuda:0'), grad: tensor([[-5.9204e-03, -7.1716e-03, -3.5744e-03, ..., -1.0399e-02, + -2.1152e-03, -4.6883e-03], + [ 2.7504e-03, -3.0136e-04, 1.6212e-03, ..., 1.6266e-02, + 1.9717e-04, 2.5826e-03], + [ 8.3065e-04, 5.1689e-04, 4.6206e-04, ..., 4.2953e-03, + 1.3053e-04, 1.8044e-03], + ..., + [-1.1206e-03, 5.9652e-04, 9.8953e-03, ..., 3.4046e-04, + 1.4591e-04, -4.6043e-03], + [-6.2895e-04, 1.9608e-03, 1.3676e-03, ..., 9.5665e-06, + 6.0415e-04, 4.8981e-03], + [-3.9744e-04, 8.8167e-04, -1.3008e-02, ..., -7.8430e-03, + 3.2353e-04, -1.9102e-03]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0311, 0.0416, -0.0007, -0.0110, 0.0220, -0.0348, 0.0048, 0.0064, + -0.0020, -0.0149], device='cuda:0'), grad: tensor([-0.0685, 0.0093, 0.0219, 0.0435, -0.0101, 0.0009, -0.0027, 0.0079, + 0.0113, -0.0133], device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 214.98, cls_loss 0.5121 cls_loss_mapping 0.0009 cls_loss_causal 0.4403 re_mapping 0.0062 re_causal 0.0171 /// teacc 99.00 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.0744, -0.1637, -0.1114, ..., -0.0388, 0.0662, -0.1127], + [-0.0829, -0.0954, -0.0902, ..., 0.1001, -0.0386, 0.2404], + [ 0.0026, -0.0321, -0.0630, ..., -0.0066, -0.0104, -0.0819], + ..., + [ 0.0373, -0.1216, 0.1830, ..., 0.0386, -0.1261, 0.0621], + [-0.0263, 0.0664, -0.1533, ..., -0.0491, 0.0192, -0.1207], + [-0.0649, 0.0831, 0.0425, ..., -0.0648, -0.0556, -0.0356]], + device='cuda:0'), grad: tensor([[ 5.4948e-07, 1.3151e-03, 3.1900e-04, ..., 2.1667e-03, + 2.1759e-02, 7.4148e-05], + [ 5.7071e-06, 2.0778e-04, 3.8218e-04, ..., 1.5144e-03, + 3.6657e-05, -7.0453e-05], + [ 5.8971e-06, -6.8665e-05, -2.1648e-03, ..., -5.9662e-03, + 1.9372e-04, 2.5678e-04], + ..., + [ 7.6723e-04, 8.0299e-04, 4.8375e-04, ..., 1.4277e-03, + 6.2406e-05, -2.2829e-04], + [ 1.1176e-04, 1.8454e-03, 1.0526e-04, ..., 1.1129e-03, + 1.1940e-03, -4.7898e-04], + [-1.4696e-03, -8.5163e-04, 1.6868e-04, ..., 9.6178e-04, + 7.5102e-05, 1.5771e-04]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0312, 0.0418, -0.0007, -0.0109, 0.0220, -0.0348, 0.0049, 0.0064, + -0.0020, -0.0149], device='cuda:0'), grad: tensor([ 0.0453, 0.0239, -0.0329, -0.0362, 0.0169, -0.0063, -0.0112, -0.0085, + -0.0074, 0.0164], device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 214.89, cls_loss 0.4655 cls_loss_mapping 0.0009 cls_loss_causal 0.4020 re_mapping 0.0061 re_causal 0.0166 /// teacc 98.99 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.0744, -0.1633, -0.1114, ..., -0.0386, 0.0662, -0.1126], + [-0.0830, -0.0954, -0.0902, ..., 0.0998, -0.0386, 0.2403], + [ 0.0026, -0.0321, -0.0630, ..., -0.0066, -0.0105, -0.0819], + ..., + [ 0.0374, -0.1215, 0.1832, ..., 0.0387, -0.1261, 0.0622], + [-0.0262, 0.0663, -0.1533, ..., -0.0491, 0.0190, -0.1208], + [-0.0650, 0.0831, 0.0425, ..., -0.0650, -0.0557, -0.0356]], + device='cuda:0'), grad: tensor([[ 8.4281e-05, 1.2171e-04, 4.5262e-06, ..., 1.9150e-03, + 8.1301e-05, 3.3766e-05], + [-1.1396e-03, -2.8968e-04, 3.1203e-05, ..., -7.6332e-03, + -1.2903e-03, -3.5596e-04], + [ 2.8515e-04, -1.1196e-03, -9.5520e-03, ..., 2.3937e-03, + 3.7837e-04, 2.4244e-05], + ..., + [ 1.1832e-04, -2.6178e-04, -6.8188e-05, ..., 1.2712e-03, + 1.3256e-04, 6.8069e-05], + [ 1.8239e-04, -3.1161e-04, 2.7329e-05, ..., 1.8578e-03, + 1.9312e-04, 6.5565e-05], + [ 4.1187e-05, -8.5640e-04, -1.4982e-03, ..., 1.5898e-03, + 4.5925e-05, 2.7791e-05]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0312, 0.0416, -0.0007, -0.0109, 0.0221, -0.0347, 0.0049, 0.0065, + -0.0020, -0.0150], device='cuda:0'), grad: tensor([ 0.0144, -0.0333, 0.0107, -0.0048, 0.0139, 0.0154, 0.0153, -0.0348, + -0.0135, 0.0168], device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 215.03, cls_loss 0.5093 cls_loss_mapping 0.0009 cls_loss_causal 0.4464 re_mapping 0.0062 re_causal 0.0170 /// teacc 98.99 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.0743, -0.1633, -0.1112, ..., -0.0384, 0.0663, -0.1126], + [-0.0830, -0.0954, -0.0902, ..., 0.0999, -0.0384, 0.2402], + [ 0.0026, -0.0322, -0.0629, ..., -0.0066, -0.0107, -0.0818], + ..., + [ 0.0374, -0.1216, 0.1831, ..., 0.0387, -0.1262, 0.0621], + [-0.0261, 0.0664, -0.1534, ..., -0.0491, 0.0189, -0.1206], + [-0.0649, 0.0832, 0.0426, ..., -0.0651, -0.0557, -0.0356]], + device='cuda:0'), grad: tensor([[ 1.6727e-03, 2.0409e-03, 1.5593e-04, ..., -7.1602e-03, + -6.4993e-04, -2.4796e-03], + [ 4.9019e-04, 1.9800e-06, 7.4692e-07, ..., 2.3842e-03, + 5.2094e-05, 2.8920e-04], + [ 9.0313e-04, 4.7565e-04, 4.1306e-05, ..., 4.6692e-03, + 2.3156e-05, 5.9223e-04], + ..., + [-4.2000e-03, 2.8744e-05, 1.2644e-05, ..., -5.7220e-03, + 1.7628e-05, 8.4991e-03], + [ 1.9875e-03, 1.8787e-03, 1.6415e-04, ..., 7.8278e-03, + 3.4571e-05, 4.5300e-04], + [-2.1992e-03, -5.2185e-03, -5.3978e-04, ..., 1.5993e-03, + 2.7108e-04, -8.1482e-03]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0312, 0.0418, -0.0006, -0.0109, 0.0220, -0.0347, 0.0049, 0.0064, + -0.0021, -0.0150], device='cuda:0'), grad: tensor([-0.0397, 0.0136, 0.0017, -0.0181, -0.0042, 0.0177, 0.0158, 0.0056, + 0.0301, -0.0224], device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 215.23, cls_loss 0.4908 cls_loss_mapping 0.0008 cls_loss_causal 0.4181 re_mapping 0.0060 re_causal 0.0167 /// teacc 98.97 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.0743, -0.1634, -0.1112, ..., -0.0384, 0.0665, -0.1126], + [-0.0831, -0.0955, -0.0901, ..., 0.1000, -0.0384, 0.2403], + [ 0.0025, -0.0320, -0.0630, ..., -0.0066, -0.0108, -0.0819], + ..., + [ 0.0374, -0.1216, 0.1830, ..., 0.0387, -0.1262, 0.0622], + [-0.0263, 0.0663, -0.1535, ..., -0.0491, 0.0188, -0.1207], + [-0.0649, 0.0832, 0.0426, ..., -0.0650, -0.0556, -0.0356]], + device='cuda:0'), grad: tensor([[ 2.9087e-04, 3.8910e-04, 1.1673e-03, ..., 7.4730e-03, + 4.2081e-04, 2.9469e-03], + [ 7.5519e-05, 9.4795e-04, 1.2243e-04, ..., 8.0185e-03, + 5.3018e-05, -7.4768e-04], + [ 1.3411e-04, 3.0208e-04, 4.0197e-04, ..., -4.5395e-03, + 1.8930e-04, 9.4080e-04], + ..., + [-3.7384e-04, -2.5702e-04, 1.3552e-03, ..., -1.4748e-02, + -3.2330e-04, 2.0325e-04], + [-5.8317e-04, -2.6093e-02, -2.7885e-03, ..., -4.8752e-03, + -4.7493e-03, -5.9128e-03], + [ 9.3758e-05, 8.2779e-04, -1.1559e-03, ..., -2.0027e-04, + 2.4402e-04, 1.5364e-03]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0312, 0.0418, -0.0007, -0.0108, 0.0219, -0.0347, 0.0049, 0.0063, + -0.0022, -0.0149], device='cuda:0'), grad: tensor([ 0.0328, 0.0036, -0.0091, 0.0215, 0.0084, 0.0262, -0.0115, -0.0314, + -0.0343, -0.0064], device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 215.29, cls_loss 0.5015 cls_loss_mapping 0.0009 cls_loss_causal 0.4251 re_mapping 0.0061 re_causal 0.0167 /// teacc 99.02 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.0741, -0.1632, -0.1113, ..., -0.0384, 0.0665, -0.1127], + [-0.0831, -0.0955, -0.0900, ..., 0.1000, -0.0385, 0.2404], + [ 0.0024, -0.0320, -0.0629, ..., -0.0068, -0.0106, -0.0819], + ..., + [ 0.0374, -0.1218, 0.1830, ..., 0.0388, -0.1260, 0.0620], + [-0.0262, 0.0663, -0.1536, ..., -0.0490, 0.0188, -0.1207], + [-0.0650, 0.0832, 0.0427, ..., -0.0651, -0.0557, -0.0355]], + device='cuda:0'), grad: tensor([[ 4.2367e-04, 5.7173e-04, 2.7275e-04, ..., 1.7509e-03, + 5.0974e-04, 2.4390e-04], + [-1.5440e-03, 3.7003e-04, -3.6899e-06, ..., -6.5880e-03, + 9.0957e-05, -1.5144e-03], + [-5.3406e-03, -1.0391e-02, 1.2708e-04, ..., -3.4161e-03, + -1.4435e-02, 1.0592e-04], + ..., + [-2.2202e-03, 6.1226e-04, -4.8327e-04, ..., 1.8501e-04, + 1.4341e-04, -1.2817e-03], + [ 4.4746e-03, 3.1452e-03, 1.3724e-05, ..., 2.3575e-03, + 6.5269e-03, 2.1362e-03], + [ 7.8344e-04, 1.5554e-03, 1.0443e-04, ..., 1.2493e-03, + 1.6081e-04, 1.6940e-04]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0311, 0.0418, -0.0007, -0.0109, 0.0220, -0.0346, 0.0047, 0.0063, + -0.0020, -0.0150], device='cuda:0'), grad: tensor([ 0.0142, -0.0534, -0.0268, 0.0143, 0.0112, 0.0108, -0.0168, 0.0072, + 0.0250, 0.0144], device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 215.00, cls_loss 0.4637 cls_loss_mapping 0.0009 cls_loss_causal 0.3959 re_mapping 0.0062 re_causal 0.0165 /// teacc 98.99 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.0741, -0.1631, -0.1112, ..., -0.0383, 0.0665, -0.1127], + [-0.0831, -0.0958, -0.0901, ..., 0.1000, -0.0383, 0.2404], + [ 0.0025, -0.0318, -0.0630, ..., -0.0066, -0.0106, -0.0821], + ..., + [ 0.0374, -0.1221, 0.1830, ..., 0.0388, -0.1260, 0.0618], + [-0.0262, 0.0662, -0.1537, ..., -0.0490, 0.0187, -0.1206], + [-0.0649, 0.0833, 0.0429, ..., -0.0650, -0.0556, -0.0354]], + device='cuda:0'), grad: tensor([[ 1.0556e-04, -3.9792e-04, 1.1571e-05, ..., -5.1451e-04, + 1.4119e-05, 5.0306e-04], + [ 5.3681e-06, 9.6440e-05, 1.8507e-05, ..., -3.4790e-03, + 1.3802e-06, 5.0507e-03], + [ 1.2755e-04, 2.5773e-04, -4.0030e-04, ..., 4.0936e-04, + 1.2204e-05, 4.0746e-04], + ..., + [ 9.4343e-07, 1.6570e-04, -3.3689e-04, ..., 3.5834e-04, + 1.1213e-06, 7.3099e-04], + [-5.4550e-04, 1.5116e-04, 6.4552e-05, ..., -3.1185e-04, + 2.8276e-04, -1.4565e-02], + [ 1.9491e-05, -4.6223e-05, 6.1560e-04, ..., 4.1246e-04, + 8.6427e-06, 5.9319e-04]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0311, 0.0418, -0.0006, -0.0109, 0.0219, -0.0346, 0.0046, 0.0063, + -0.0020, -0.0149], device='cuda:0'), grad: tensor([-0.0258, 0.0112, 0.0068, -0.0049, 0.0151, 0.0030, 0.0082, -0.0204, + -0.0254, 0.0321], device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 214.86, cls_loss 0.4897 cls_loss_mapping 0.0007 cls_loss_causal 0.4290 re_mapping 0.0063 re_causal 0.0171 /// teacc 99.01 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.0740, -0.1631, -0.1111, ..., -0.0382, 0.0663, -0.1127], + [-0.0833, -0.0958, -0.0899, ..., 0.1000, -0.0383, 0.2403], + [ 0.0025, -0.0318, -0.0631, ..., -0.0067, -0.0105, -0.0821], + ..., + [ 0.0375, -0.1220, 0.1831, ..., 0.0388, -0.1260, 0.0620], + [-0.0262, 0.0661, -0.1538, ..., -0.0490, 0.0186, -0.1205], + [-0.0650, 0.0833, 0.0428, ..., -0.0651, -0.0557, -0.0356]], + device='cuda:0'), grad: tensor([[ 1.3885e-03, 4.5896e-04, 4.7374e-04, ..., 4.8218e-03, + 7.9691e-05, 9.2411e-04], + [ 8.2445e-04, 2.4402e-04, 1.2465e-03, ..., 3.0766e-03, + 1.5557e-05, 8.4972e-04], + [ 7.8201e-04, 2.7370e-04, 6.4754e-04, ..., 2.8782e-03, + 3.6091e-05, 5.8794e-04], + ..., + [ 5.9128e-04, 3.5000e-04, -4.2152e-04, ..., -9.8190e-03, + 3.8505e-05, -2.5368e-03], + [ 9.1743e-04, 7.2002e-04, 3.4642e-04, ..., 3.5915e-03, + 6.3598e-05, 6.6757e-04], + [ 1.8711e-03, -1.8823e-04, 5.4646e-04, ..., 4.7798e-03, + 1.0592e-04, 9.5797e-04]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0311, 0.0417, -0.0007, -0.0108, 0.0220, -0.0345, 0.0047, 0.0064, + -0.0022, -0.0150], device='cuda:0'), grad: tensor([ 0.0304, 0.0303, 0.0237, -0.0665, 0.0098, -0.0104, -0.0063, -0.0368, + 0.0295, -0.0037], device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 214.89, cls_loss 0.4779 cls_loss_mapping 0.0008 cls_loss_causal 0.4158 re_mapping 0.0062 re_causal 0.0170 /// teacc 98.99 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.0741, -0.1631, -0.1112, ..., -0.0381, 0.0664, -0.1128], + [-0.0833, -0.0959, -0.0900, ..., 0.0998, -0.0383, 0.2403], + [ 0.0024, -0.0317, -0.0631, ..., -0.0067, -0.0105, -0.0821], + ..., + [ 0.0374, -0.1221, 0.1830, ..., 0.0389, -0.1260, 0.0620], + [-0.0262, 0.0660, -0.1539, ..., -0.0492, 0.0185, -0.1206], + [-0.0650, 0.0833, 0.0429, ..., -0.0651, -0.0560, -0.0355]], + device='cuda:0'), grad: tensor([[ 3.1519e-04, 4.7654e-05, 5.2738e-04, ..., 1.4191e-03, + 4.9639e-04, 3.6860e-04], + [ 3.7026e-04, 4.5538e-05, 5.0354e-04, ..., 8.3256e-04, + 2.4116e-04, 3.3045e-04], + [ 3.7837e-04, 5.2929e-04, 5.9986e-04, ..., -1.8864e-03, + -1.2898e-04, 2.9087e-04], + ..., + [ 3.7408e-04, 1.6892e-04, 9.1410e-04, ..., 1.3666e-03, + 2.2686e-04, 7.1573e-04], + [ 1.0576e-03, 6.2866e-03, 1.1520e-03, ..., 7.2441e-03, + 6.8605e-05, 1.0805e-03], + [ 6.0225e-04, 1.0240e-04, 2.1114e-03, ..., 2.5463e-03, + 3.9339e-04, 1.6298e-03]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0310, 0.0417, -0.0005, -0.0109, 0.0221, -0.0347, 0.0046, 0.0064, + -0.0022, -0.0150], device='cuda:0'), grad: tensor([ 0.0143, 0.0083, -0.0581, 0.0094, -0.0274, 0.0117, -0.0223, 0.0127, + 0.0295, 0.0220], device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 214.68, cls_loss 0.4959 cls_loss_mapping 0.0008 cls_loss_causal 0.4313 re_mapping 0.0062 re_causal 0.0166 /// teacc 98.98 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.0739, -0.1631, -0.1111, ..., -0.0381, 0.0663, -0.1129], + [-0.0834, -0.0958, -0.0900, ..., 0.0998, -0.0383, 0.2403], + [ 0.0025, -0.0317, -0.0631, ..., -0.0068, -0.0104, -0.0819], + ..., + [ 0.0374, -0.1220, 0.1831, ..., 0.0391, -0.1260, 0.0621], + [-0.0262, 0.0662, -0.1535, ..., -0.0491, 0.0186, -0.1206], + [-0.0653, 0.0833, 0.0429, ..., -0.0652, -0.0560, -0.0356]], + device='cuda:0'), grad: tensor([[ 1.5521e-04, 2.8515e-04, 9.9540e-05, ..., -1.1101e-02, + 1.6661e-06, 9.3222e-05], + [-1.5650e-03, 1.7488e-04, 8.7500e-05, ..., -3.7980e-04, + 2.7865e-06, 8.5294e-05], + [ 2.7061e-04, 6.5231e-04, 8.1301e-04, ..., 3.7155e-03, + 3.0905e-05, 3.3712e-04], + ..., + [ 5.1200e-05, 1.8251e-04, 1.7975e-02, ..., 1.1627e-02, + 1.4812e-05, -4.0388e-04], + [ 2.4724e-04, 1.0872e-03, 3.8767e-04, ..., 2.4929e-03, + 9.0525e-06, -1.2522e-03], + [ 8.6188e-05, -2.5391e-02, -1.6708e-03, ..., 1.6575e-03, + 6.7770e-05, 8.1110e-04]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0310, 0.0416, -0.0006, -0.0109, 0.0221, -0.0346, 0.0046, 0.0065, + -0.0022, -0.0150], device='cuda:0'), grad: tensor([-0.0420, -0.0187, 0.0292, -0.0196, 0.0137, 0.0167, 0.0226, -0.0145, + 0.0124, 0.0001], device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 214.83, cls_loss 0.4911 cls_loss_mapping 0.0010 cls_loss_causal 0.4322 re_mapping 0.0061 re_causal 0.0164 /// teacc 98.99 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.0738, -0.1631, -0.1111, ..., -0.0380, 0.0664, -0.1127], + [-0.0836, -0.0958, -0.0901, ..., 0.0997, -0.0383, 0.2402], + [ 0.0024, -0.0318, -0.0631, ..., -0.0069, -0.0105, -0.0821], + ..., + [ 0.0376, -0.1219, 0.1832, ..., 0.0392, -0.1260, 0.0620], + [-0.0261, 0.0662, -0.1534, ..., -0.0493, 0.0187, -0.1206], + [-0.0653, 0.0834, 0.0428, ..., -0.0652, -0.0559, -0.0356]], + device='cuda:0'), grad: tensor([[ 8.3971e-04, 4.0364e-04, 2.1845e-05, ..., 1.7319e-03, + 5.4932e-04, 6.6710e-04], + [ 4.8351e-04, 1.4877e-04, 1.6913e-05, ..., 1.2569e-03, + 3.7241e-04, 3.7456e-04], + [ 3.3116e-04, 5.3376e-05, 4.9257e-04, ..., 1.4563e-03, + 2.3901e-04, 4.3941e-04], + ..., + [ 5.8842e-04, 3.5882e-05, -5.5790e-04, ..., -6.6233e-04, + 5.7364e-04, 1.7250e-04], + [ 7.8249e-04, 6.9761e-04, 5.5522e-05, ..., -9.9373e-04, + 3.4165e-04, 6.2180e-04], + [-3.3569e-03, -3.0003e-03, -2.2182e-03, ..., -4.0970e-03, + -3.7117e-03, -2.1248e-03]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0308, 0.0414, -0.0006, -0.0109, 0.0220, -0.0346, 0.0046, 0.0067, + -0.0023, -0.0152], device='cuda:0'), grad: tensor([ 0.0157, 0.0161, 0.0129, -0.0499, -0.0198, 0.0325, 0.0430, -0.0161, + -0.0144, -0.0200], device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 214.71, cls_loss 0.4797 cls_loss_mapping 0.0009 cls_loss_causal 0.4175 re_mapping 0.0059 re_causal 0.0162 /// teacc 98.99 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.0738, -0.1631, -0.1110, ..., -0.0381, 0.0665, -0.1128], + [-0.0836, -0.0958, -0.0899, ..., 0.0997, -0.0383, 0.2403], + [ 0.0024, -0.0317, -0.0632, ..., -0.0069, -0.0106, -0.0820], + ..., + [ 0.0376, -0.1216, 0.1834, ..., 0.0393, -0.1260, 0.0620], + [-0.0261, 0.0662, -0.1534, ..., -0.0492, 0.0188, -0.1205], + [-0.0653, 0.0834, 0.0427, ..., -0.0653, -0.0561, -0.0358]], + device='cuda:0'), grad: tensor([[ 1.6198e-05, 9.1672e-05, 5.6531e-07, ..., -5.1022e-04, + -3.7178e-06, -2.2106e-03], + [-6.9663e-06, 7.2643e-07, -8.9630e-06, ..., -4.0054e-04, + 2.2352e-08, 4.8709e-04], + [ 9.5740e-07, 3.0119e-06, -6.1765e-06, ..., 8.3876e-04, + -3.0577e-05, -2.6321e-03], + ..., + [ 5.7146e-06, 1.0449e-06, 1.5154e-05, ..., -1.6766e-03, + 2.0877e-05, 3.6311e-04], + [ 2.2626e-04, 1.2856e-03, 5.5246e-06, ..., 1.0796e-03, + 1.1129e-06, 1.7319e-03], + [ 2.0564e-06, 6.8992e-06, 4.3586e-06, ..., 6.8283e-04, + 1.2666e-07, 3.2330e-04]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0308, 0.0415, -0.0007, -0.0109, 0.0219, -0.0345, 0.0046, 0.0067, + -0.0022, -0.0153], device='cuda:0'), grad: tensor([-0.0098, -0.0160, 0.0106, -0.0105, 0.0050, 0.0138, 0.0127, -0.0467, + 0.0193, 0.0216], device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 214.43, cls_loss 0.4938 cls_loss_mapping 0.0010 cls_loss_causal 0.4257 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.98 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.0740, -0.1632, -0.1111, ..., -0.0382, 0.0665, -0.1128], + [-0.0838, -0.0958, -0.0900, ..., 0.0997, -0.0383, 0.2401], + [ 0.0024, -0.0319, -0.0631, ..., -0.0071, -0.0105, -0.0821], + ..., + [ 0.0377, -0.1217, 0.1834, ..., 0.0393, -0.1262, 0.0621], + [-0.0260, 0.0662, -0.1535, ..., -0.0490, 0.0188, -0.1204], + [-0.0652, 0.0834, 0.0428, ..., -0.0653, -0.0562, -0.0359]], + device='cuda:0'), grad: tensor([[-2.7103e-03, -5.1231e-03, 1.9681e-04, ..., -2.6169e-03, + 2.3174e-04, -1.3447e-03], + [ 1.3218e-03, 1.9951e-03, 2.5302e-05, ..., 1.1797e-03, + 1.3781e-04, 5.1737e-04], + [ 7.4196e-04, 8.5402e-04, 1.4424e-04, ..., 2.1591e-03, + -5.6953e-03, 2.2578e-04], + ..., + [ 3.2177e-03, 1.7071e-04, -1.6336e-03, ..., 5.7297e-03, + 1.6594e-04, 1.6630e-04], + [ 4.7827e-04, 4.5943e-04, 7.0512e-05, ..., 2.1706e-03, + 9.4354e-05, 1.2243e-04], + [ 1.6298e-03, 1.5116e-03, 3.0351e-04, ..., -2.2736e-03, + 2.8896e-04, 6.0749e-04]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0310, 0.0415, -0.0008, -0.0108, 0.0218, -0.0343, 0.0045, 0.0068, + -0.0021, -0.0153], device='cuda:0'), grad: tensor([-0.0043, -0.0957, -0.0015, 0.0254, 0.0458, -0.0638, 0.0271, 0.0398, + 0.0242, 0.0031], device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 214.31, cls_loss 0.4966 cls_loss_mapping 0.0009 cls_loss_causal 0.4306 re_mapping 0.0061 re_causal 0.0164 /// teacc 98.94 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.0740, -0.1633, -0.1113, ..., -0.0381, 0.0665, -0.1129], + [-0.0837, -0.0959, -0.0900, ..., 0.0998, -0.0381, 0.2401], + [ 0.0023, -0.0316, -0.0630, ..., -0.0070, -0.0105, -0.0820], + ..., + [ 0.0378, -0.1217, 0.1834, ..., 0.0393, -0.1262, 0.0622], + [-0.0261, 0.0662, -0.1534, ..., -0.0492, 0.0187, -0.1205], + [-0.0651, 0.0833, 0.0428, ..., -0.0651, -0.0563, -0.0360]], + device='cuda:0'), grad: tensor([[ 1.0766e-05, -4.4513e-04, 4.3213e-06, ..., -1.9503e-03, + 8.7358e-07, 4.8018e-04], + [ 2.3637e-06, 2.3901e-04, -2.9936e-05, ..., 9.0637e-03, + -1.1325e-05, 7.8583e-03], + [ 1.1362e-06, 4.0221e-04, 2.7433e-05, ..., -1.1658e-02, + 9.6951e-07, -1.2955e-02], + ..., + [ 5.1372e-06, -2.8110e-04, 6.3717e-05, ..., 8.4448e-04, + 1.1303e-05, 1.0195e-03], + [ 7.5519e-05, 5.4884e-04, 7.4446e-05, ..., 2.8229e-03, + 9.6262e-06, 1.1740e-03], + [ 1.8585e-04, 1.1911e-03, 1.9989e-03, ..., 1.3123e-03, + 3.6931e-04, 1.8549e-03]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0310, 0.0415, -0.0007, -0.0108, 0.0218, -0.0342, 0.0045, 0.0067, + -0.0021, -0.0152], device='cuda:0'), grad: tensor([-0.0456, 0.0371, -0.0130, -0.0152, -0.0160, 0.0145, 0.0292, -0.0083, + 0.0221, -0.0049], device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 214.42, cls_loss 0.4948 cls_loss_mapping 0.0009 cls_loss_causal 0.4291 re_mapping 0.0062 re_causal 0.0167 /// teacc 98.95 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.0740, -0.1634, -0.1115, ..., -0.0382, 0.0665, -0.1130], + [-0.0837, -0.0960, -0.0900, ..., 0.0997, -0.0382, 0.2403], + [ 0.0023, -0.0316, -0.0630, ..., -0.0068, -0.0104, -0.0819], + ..., + [ 0.0379, -0.1217, 0.1833, ..., 0.0392, -0.1263, 0.0622], + [-0.0262, 0.0663, -0.1533, ..., -0.0491, 0.0188, -0.1204], + [-0.0651, 0.0832, 0.0429, ..., -0.0652, -0.0561, -0.0361]], + device='cuda:0'), grad: tensor([[-2.8076e-03, 5.7518e-05, 1.8525e-04, ..., 2.5272e-04, + 5.5015e-05, 1.1091e-03], + [ 2.2392e-03, -6.9571e-04, -1.9026e-03, ..., -4.9782e-03, + -1.1200e-04, -1.7452e-04], + [ 3.9749e-03, 6.7055e-05, 4.4203e-04, ..., 3.5839e-03, + 1.7241e-05, 1.0099e-03], + ..., + [ 1.2703e-03, 7.5161e-05, 8.4448e-04, ..., 4.1924e-03, + 2.0400e-05, 1.0996e-03], + [ 2.5444e-03, 9.5963e-05, -1.7757e-03, ..., 4.0436e-03, + 8.2016e-05, 1.0309e-03], + [-5.6038e-03, 5.7697e-05, 1.2007e-03, ..., -1.2665e-02, + -2.4176e-04, -3.2787e-03]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0310, 0.0414, -0.0004, -0.0109, 0.0218, -0.0341, 0.0044, 0.0066, + -0.0022, -0.0152], device='cuda:0'), grad: tensor([-0.0003, -0.0688, 0.0280, 0.0269, -0.0064, -0.0101, 0.0327, 0.0229, + -0.0053, -0.0194], device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 214.34, cls_loss 0.4946 cls_loss_mapping 0.0010 cls_loss_causal 0.4292 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.96 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.0741, -0.1633, -0.1114, ..., -0.0382, 0.0664, -0.1130], + [-0.0836, -0.0960, -0.0900, ..., 0.0997, -0.0382, 0.2404], + [ 0.0023, -0.0317, -0.0630, ..., -0.0069, -0.0104, -0.0820], + ..., + [ 0.0378, -0.1217, 0.1831, ..., 0.0392, -0.1264, 0.0621], + [-0.0262, 0.0662, -0.1534, ..., -0.0492, 0.0187, -0.1204], + [-0.0652, 0.0833, 0.0432, ..., -0.0652, -0.0561, -0.0360]], + device='cuda:0'), grad: tensor([[ 4.0698e-04, 1.8864e-03, 3.9043e-03, ..., 9.0694e-04, + 5.1117e-03, 7.3612e-05], + [ 2.2054e-04, -8.4102e-05, -1.8668e-04, ..., 1.3552e-03, + 3.8743e-04, -6.0844e-04], + [ 4.6444e-04, 9.9003e-05, 1.5187e-04, ..., 1.2274e-03, + 3.5238e-04, 1.5020e-04], + ..., + [ 7.7915e-04, -4.3130e-04, -1.0090e-03, ..., 4.0984e-04, + 3.6478e-04, -1.0118e-03], + [ 8.2779e-04, 1.9634e-04, 3.1137e-04, ..., 1.8435e-03, + 2.7585e-04, 6.4516e-04], + [ 5.8794e-04, 2.3603e-04, 3.5095e-04, ..., 1.5745e-03, + 4.0770e-04, 3.3569e-04]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0310, 0.0414, -0.0005, -0.0109, 0.0218, -0.0339, 0.0045, 0.0064, + -0.0023, -0.0151], device='cuda:0'), grad: tensor([ 0.0252, 0.0104, 0.0091, -0.0167, 0.0089, -0.0219, -0.0422, 0.0039, + 0.0114, 0.0117], device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 214.41, cls_loss 0.4704 cls_loss_mapping 0.0009 cls_loss_causal 0.4042 re_mapping 0.0058 re_causal 0.0158 /// teacc 98.96 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.0741, -0.1633, -0.1115, ..., -0.0382, 0.0666, -0.1130], + [-0.0837, -0.0962, -0.0901, ..., 0.0996, -0.0382, 0.2404], + [ 0.0024, -0.0316, -0.0629, ..., -0.0068, -0.0103, -0.0819], + ..., + [ 0.0378, -0.1217, 0.1832, ..., 0.0393, -0.1264, 0.0621], + [-0.0263, 0.0662, -0.1534, ..., -0.0492, 0.0187, -0.1205], + [-0.0652, 0.0833, 0.0432, ..., -0.0653, -0.0562, -0.0359]], + device='cuda:0'), grad: tensor([[-9.6035e-04, 2.7027e-03, 2.8563e-04, ..., 2.0409e-03, + 1.6031e-03, 2.3353e-04], + [-4.9858e-03, -4.9438e-03, -7.3738e-03, ..., 1.7761e-02, + 1.9580e-05, -3.1204e-03], + [-1.5144e-03, -1.1665e-02, -9.5558e-04, ..., -6.3972e-03, + -6.9084e-03, 1.0908e-04], + ..., + [ 2.2068e-03, 2.3232e-03, 2.0275e-03, ..., 1.6689e-03, + 4.1425e-05, 1.8959e-03], + [ 4.0078e-04, 6.5857e-02, 8.3313e-03, ..., 4.8923e-04, + 1.8158e-03, 3.0613e-04], + [ 7.4005e-04, 9.2793e-04, 8.7786e-04, ..., 1.4400e-03, + 1.1897e-04, 6.9952e-04]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0310, 0.0413, -0.0004, -0.0109, 0.0217, -0.0340, 0.0046, 0.0065, + -0.0023, -0.0151], device='cuda:0'), grad: tensor([-0.0171, 0.0193, -0.0194, 0.0096, -0.0209, 0.0154, -0.0182, 0.0157, + 0.0019, 0.0137], device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 214.92, cls_loss 0.4810 cls_loss_mapping 0.0009 cls_loss_causal 0.4272 re_mapping 0.0063 re_causal 0.0171 /// teacc 98.96 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.0741, -0.1633, -0.1115, ..., -0.0382, 0.0665, -0.1130], + [-0.0838, -0.0962, -0.0901, ..., 0.0995, -0.0383, 0.2404], + [ 0.0023, -0.0316, -0.0629, ..., -0.0069, -0.0103, -0.0819], + ..., + [ 0.0377, -0.1218, 0.1832, ..., 0.0393, -0.1265, 0.0622], + [-0.0263, 0.0662, -0.1535, ..., -0.0495, 0.0187, -0.1207], + [-0.0651, 0.0834, 0.0432, ..., -0.0652, -0.0562, -0.0358]], + device='cuda:0'), grad: tensor([[ 7.7438e-04, 4.6879e-05, 3.2592e-04, ..., 6.1005e-05, + 3.2520e-04, 7.6914e-04], + [-1.8096e-04, -4.6692e-03, -2.7156e-04, ..., -9.2239e-03, + 3.0828e-04, -6.0539e-03], + [-1.9054e-03, 2.1994e-04, -1.4257e-03, ..., -3.5954e-03, + -4.7798e-03, -1.0595e-03], + ..., + [ 6.3658e-04, -7.2908e-04, -3.4733e-03, ..., -3.3779e-03, + 2.2709e-04, -4.9744e-03], + [ 6.3324e-04, 4.2229e-03, 2.3770e-04, ..., 8.8730e-03, + 7.3576e-04, 6.0654e-03], + [ 5.3215e-04, 4.6372e-04, 1.6060e-03, ..., 3.0975e-03, + 2.8324e-04, 9.3307e-03]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0310, 0.0414, -0.0004, -0.0108, 0.0215, -0.0339, 0.0047, 0.0065, + -0.0024, -0.0150], device='cuda:0'), grad: tensor([-0.0200, -0.0600, -0.0213, -0.0085, -0.0074, 0.0200, 0.0150, 0.0010, + 0.0437, 0.0374], device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 214.71, cls_loss 0.5072 cls_loss_mapping 0.0010 cls_loss_causal 0.4433 re_mapping 0.0061 re_causal 0.0166 /// teacc 98.96 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.0742, -0.1632, -0.1115, ..., -0.0381, 0.0665, -0.1130], + [-0.0836, -0.0961, -0.0901, ..., 0.0996, -0.0383, 0.2405], + [ 0.0022, -0.0317, -0.0630, ..., -0.0070, -0.0103, -0.0820], + ..., + [ 0.0376, -0.1219, 0.1832, ..., 0.0393, -0.1266, 0.0621], + [-0.0263, 0.0663, -0.1535, ..., -0.0495, 0.0187, -0.1206], + [-0.0648, 0.0833, 0.0433, ..., -0.0650, -0.0561, -0.0357]], + device='cuda:0'), grad: tensor([[ 3.9434e-04, 1.7977e-03, 3.7283e-05, ..., 4.2114e-03, + 2.2912e-04, 1.8835e-03], + [-5.0020e-04, 8.7547e-04, 2.7013e-04, ..., 4.6005e-03, + 7.6342e-04, 7.6675e-03], + [ 9.4604e-04, -1.0986e-02, -1.6565e-03, ..., -2.5940e-03, + 2.5606e-04, -1.6220e-02], + ..., + [-4.7455e-03, 8.8310e-04, -6.3658e-04, ..., -4.2458e-03, + 1.8167e-04, 9.6226e-04], + [ 7.6246e-04, 7.5674e-04, 7.2181e-05, ..., 1.2083e-03, + 3.1567e-04, 4.7493e-04], + [ 1.1492e-03, 1.4057e-03, 1.1277e-04, ..., 3.2043e-03, + 2.4676e-04, 1.7738e-03]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0310, 0.0415, -0.0004, -0.0109, 0.0215, -0.0340, 0.0047, 0.0064, + -0.0025, -0.0149], device='cuda:0'), grad: tensor([ 0.0267, 0.0179, -0.1306, 0.0014, 0.0720, 0.0232, -0.0459, -0.0035, + 0.0112, 0.0277], device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 214.73, cls_loss 0.5031 cls_loss_mapping 0.0010 cls_loss_causal 0.4264 re_mapping 0.0060 re_causal 0.0167 /// teacc 98.96 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.0742, -0.1631, -0.1115, ..., -0.0382, 0.0666, -0.1130], + [-0.0837, -0.0962, -0.0901, ..., 0.0997, -0.0383, 0.2406], + [ 0.0022, -0.0316, -0.0630, ..., -0.0070, -0.0103, -0.0820], + ..., + [ 0.0375, -0.1220, 0.1831, ..., 0.0393, -0.1266, 0.0621], + [-0.0261, 0.0664, -0.1535, ..., -0.0496, 0.0188, -0.1205], + [-0.0647, 0.0833, 0.0434, ..., -0.0648, -0.0560, -0.0357]], + device='cuda:0'), grad: tensor([[ 4.4346e-04, -3.9330e-03, 2.1592e-05, ..., -3.5019e-03, + 6.5231e-04, 3.3164e-04], + [-2.9731e-04, -1.6289e-03, -6.6519e-04, ..., 1.0568e-04, + 7.1466e-05, -5.6114e-03], + [ 2.0075e-04, 8.6498e-04, 6.2406e-05, ..., -7.7477e-03, + 2.6751e-04, 4.0388e-04], + ..., + [ 1.1235e-04, 1.4076e-03, 9.3579e-05, ..., 4.5090e-03, + 9.1970e-05, 6.8283e-04], + [ 1.2791e-04, -7.7844e-05, 1.4269e-04, ..., -9.7046e-03, + 9.4473e-05, 7.2193e-04], + [ 1.1069e-04, 1.9372e-05, 7.0095e-05, ..., 3.7079e-03, + 1.0562e-04, 1.1091e-03]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0310, 0.0416, -0.0003, -0.0108, 0.0215, -0.0340, 0.0047, 0.0063, + -0.0025, -0.0149], device='cuda:0'), grad: tensor([-0.0591, -0.0398, -0.0127, 0.0114, 0.0289, 0.0344, 0.0294, 0.0339, + -0.0580, 0.0315], device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 215.05, cls_loss 0.4848 cls_loss_mapping 0.0009 cls_loss_causal 0.4168 re_mapping 0.0062 re_causal 0.0168 /// teacc 98.95 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.0742, -0.1630, -0.1112, ..., -0.0382, 0.0666, -0.1130], + [-0.0836, -0.0960, -0.0900, ..., 0.0997, -0.0384, 0.2407], + [ 0.0024, -0.0318, -0.0630, ..., -0.0071, -0.0102, -0.0821], + ..., + [ 0.0375, -0.1220, 0.1833, ..., 0.0392, -0.1267, 0.0621], + [-0.0260, 0.0663, -0.1536, ..., -0.0495, 0.0188, -0.1206], + [-0.0647, 0.0833, 0.0432, ..., -0.0650, -0.0559, -0.0358]], + device='cuda:0'), grad: tensor([[-9.2554e-04, 1.6415e-04, 6.5136e-04, ..., -7.8499e-05, + 1.9045e-03, 7.9572e-05], + [-2.5578e-03, 1.7583e-04, 2.4319e-03, ..., 9.4843e-04, + 4.3144e-03, 6.2406e-05], + [ 2.3613e-03, 2.2554e-04, 2.5330e-03, ..., 1.1665e-02, + 4.2534e-03, 1.0550e-04], + ..., + [ 2.5344e-04, 1.6890e-03, -5.1308e-03, ..., 1.3742e-03, + 1.6844e-04, -6.0616e-03], + [-1.2207e-03, 1.6600e-05, -1.1063e-02, ..., -1.4313e-02, + -9.3231e-03, 1.4579e-04], + [ 1.8489e-04, -3.1853e-03, 5.7459e-04, ..., -1.1797e-03, + 2.5511e-04, -1.8864e-03]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0311, 0.0417, -0.0003, -0.0109, 0.0214, -0.0339, 0.0047, 0.0063, + -0.0024, -0.0150], device='cuda:0'), grad: tensor([ 0.0082, 0.0026, 0.0174, 0.0390, 0.0106, -0.0197, 0.0365, -0.0265, + -0.0460, -0.0221], device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 214.87, cls_loss 0.4994 cls_loss_mapping 0.0009 cls_loss_causal 0.4403 re_mapping 0.0061 re_causal 0.0168 /// teacc 99.01 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.0741, -0.1630, -0.1112, ..., -0.0383, 0.0664, -0.1130], + [-0.0837, -0.0961, -0.0901, ..., 0.0997, -0.0386, 0.2408], + [ 0.0023, -0.0317, -0.0631, ..., -0.0070, -0.0105, -0.0821], + ..., + [ 0.0375, -0.1221, 0.1834, ..., 0.0392, -0.1266, 0.0622], + [-0.0260, 0.0663, -0.1537, ..., -0.0495, 0.0190, -0.1206], + [-0.0647, 0.0833, 0.0432, ..., -0.0649, -0.0559, -0.0358]], + device='cuda:0'), grad: tensor([[ 2.1446e-04, 2.1529e-04, 2.5129e-04, ..., 1.0281e-03, + -7.6675e-04, 4.2033e-04], + [-1.4982e-03, 4.0531e-04, 6.5565e-04, ..., -1.8940e-03, + 3.2997e-04, -4.8065e-03], + [ 1.9491e-04, 2.8300e-04, 4.2248e-04, ..., 1.1683e-03, + 4.9877e-04, 4.2272e-04], + ..., + [ 1.9574e-04, -4.0340e-04, 6.9523e-04, ..., 2.6971e-05, + 2.4140e-04, 5.7650e-04], + [ 7.4482e-04, -1.2648e-04, 2.7132e-04, ..., 1.7738e-03, + 4.2081e-04, 1.2341e-03], + [ 1.0195e-03, 8.4543e-04, -9.5654e-04, ..., 3.4142e-03, + 7.8392e-04, 8.7166e-04]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0312, 0.0415, -0.0004, -0.0109, 0.0213, -0.0339, 0.0048, 0.0064, + -0.0024, -0.0148], device='cuda:0'), grad: tensor([ 0.0144, -0.0085, 0.0160, -0.0103, 0.0147, -0.0118, -0.0094, -0.0429, + 0.0184, 0.0194], device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 214.93, cls_loss 0.4829 cls_loss_mapping 0.0010 cls_loss_causal 0.4080 re_mapping 0.0061 re_causal 0.0163 /// teacc 99.01 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.0741, -0.1631, -0.1113, ..., -0.0383, 0.0665, -0.1130], + [-0.0837, -0.0960, -0.0902, ..., 0.0997, -0.0386, 0.2407], + [ 0.0023, -0.0317, -0.0631, ..., -0.0070, -0.0104, -0.0822], + ..., + [ 0.0375, -0.1222, 0.1835, ..., 0.0392, -0.1265, 0.0623], + [-0.0259, 0.0664, -0.1536, ..., -0.0496, 0.0190, -0.1206], + [-0.0648, 0.0834, 0.0432, ..., -0.0649, -0.0561, -0.0359]], + device='cuda:0'), grad: tensor([[ 7.7963e-04, 3.8791e-04, 7.2050e-04, ..., 1.5011e-03, + 2.4259e-05, 2.1130e-05], + [ 1.6701e-04, 7.9012e-04, 4.1217e-05, ..., -3.3665e-04, + 1.8263e-06, 3.1143e-05], + [ 3.1853e-03, 7.3929e-03, -1.9765e-04, ..., 4.0817e-03, + 1.9300e-04, 1.0169e-04], + ..., + [ 2.5158e-03, 5.1422e-03, -1.7256e-05, ..., 2.1534e-03, + 3.6322e-06, -6.3992e-04], + [-1.0948e-03, -8.8882e-04, -1.5078e-03, ..., -3.3722e-03, + 1.8859e-06, -1.6958e-05], + [ 8.9264e-04, 1.7834e-03, 3.7169e-04, ..., 1.5907e-03, + 1.7002e-05, 2.7585e-04]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0312, 0.0416, -0.0003, -0.0109, 0.0212, -0.0340, 0.0049, 0.0064, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([-0.0095, 0.0090, 0.0238, -0.0100, -0.0212, 0.0129, -0.0222, 0.0186, + -0.0197, 0.0184], device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 214.61, cls_loss 0.4920 cls_loss_mapping 0.0009 cls_loss_causal 0.4334 re_mapping 0.0062 re_causal 0.0167 /// teacc 99.01 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.0740, -0.1631, -0.1114, ..., -0.0382, 0.0665, -0.1131], + [-0.0835, -0.0961, -0.0902, ..., 0.0998, -0.0386, 0.2409], + [ 0.0022, -0.0317, -0.0630, ..., -0.0068, -0.0104, -0.0823], + ..., + [ 0.0375, -0.1223, 0.1835, ..., 0.0392, -0.1265, 0.0623], + [-0.0260, 0.0664, -0.1537, ..., -0.0496, 0.0189, -0.1206], + [-0.0648, 0.0833, 0.0432, ..., -0.0649, -0.0561, -0.0360]], + device='cuda:0'), grad: tensor([[ 2.2960e-04, 8.0466e-05, 9.9838e-06, ..., 7.9441e-04, + 4.0233e-05, 3.1471e-04], + [ 6.2132e-04, 4.0710e-05, 1.7017e-05, ..., -1.2611e-02, + 1.3495e-06, 6.6471e-04], + [ 2.4772e-04, -3.8296e-05, 5.4181e-05, ..., 1.1988e-03, + 6.2399e-07, 3.9744e-04], + ..., + [ 7.4625e-04, 2.5296e-04, -2.0874e-04, ..., 1.5421e-03, + 8.6613e-08, 3.4122e-03], + [-3.2558e-03, 2.8682e-04, 2.9612e-04, ..., 6.9466e-03, + 1.0990e-05, -2.9736e-03], + [ 1.8847e-04, -3.2330e-04, -5.9366e-04, ..., 1.0252e-03, + 1.2619e-06, 3.1471e-03]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0311, 0.0417, -0.0002, -0.0110, 0.0213, -0.0342, 0.0049, 0.0064, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0110, -0.0016, 0.0117, -0.0189, -0.0022, -0.0199, 0.0148, 0.0201, + -0.0314, 0.0165], device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 214.75, cls_loss 0.4995 cls_loss_mapping 0.0009 cls_loss_causal 0.4323 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.98 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.0738, -0.1631, -0.1112, ..., -0.0380, 0.0666, -0.1132], + [-0.0835, -0.0960, -0.0902, ..., 0.0997, -0.0387, 0.2409], + [ 0.0021, -0.0317, -0.0630, ..., -0.0070, -0.0103, -0.0823], + ..., + [ 0.0376, -0.1222, 0.1834, ..., 0.0392, -0.1266, 0.0624], + [-0.0261, 0.0665, -0.1538, ..., -0.0496, 0.0189, -0.1205], + [-0.0648, 0.0831, 0.0433, ..., -0.0648, -0.0561, -0.0360]], + device='cuda:0'), grad: tensor([[ 3.7789e-04, 6.7353e-05, 1.2163e-06, ..., 1.3418e-03, + 3.9548e-05, 4.3750e-04], + [ 1.6117e-03, 9.6500e-05, 1.4448e-03, ..., -3.6354e-03, + 2.4366e-04, 1.1301e-03], + [ 3.3212e-04, -6.2513e-04, 6.7241e-06, ..., -4.6754e-04, + 5.4032e-05, -1.6861e-03], + ..., + [ 3.6740e-04, 1.5807e-04, 2.0182e-04, ..., -3.0255e-04, + 3.5584e-05, -5.9700e-04], + [ 6.6757e-04, 1.6603e-03, 6.4754e-04, ..., 2.9526e-03, + 5.3704e-05, 6.8545e-05], + [-5.9662e-03, -1.7128e-03, -7.2718e-04, ..., -7.4234e-03, + -6.9761e-04, -3.0918e-03]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0310, 0.0417, -0.0003, -0.0109, 0.0213, -0.0341, 0.0049, 0.0064, + -0.0026, -0.0149], device='cuda:0'), grad: tensor([ 0.0109, -0.0257, -0.0234, 0.0211, 0.0138, -0.0006, 0.0220, 0.0082, + -0.0089, -0.0173], device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 214.58, cls_loss 0.4702 cls_loss_mapping 0.0008 cls_loss_causal 0.4158 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.99 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.0739, -0.1630, -0.1113, ..., -0.0381, 0.0668, -0.1131], + [-0.0833, -0.0960, -0.0903, ..., 0.0997, -0.0388, 0.2409], + [ 0.0022, -0.0316, -0.0630, ..., -0.0070, -0.0103, -0.0825], + ..., + [ 0.0376, -0.1223, 0.1835, ..., 0.0393, -0.1265, 0.0625], + [-0.0261, 0.0664, -0.1540, ..., -0.0496, 0.0189, -0.1206], + [-0.0647, 0.0832, 0.0434, ..., -0.0647, -0.0560, -0.0360]], + device='cuda:0'), grad: tensor([[ 3.1509e-03, -3.6192e-04, -6.5470e-04, ..., -7.7963e-04, + -9.4604e-04, -9.7942e-04], + [ 9.2936e-04, 3.1447e-04, 7.2308e-06, ..., -1.5621e-03, + 1.2946e-04, -1.7524e-04], + [ 6.9809e-03, 2.7905e-03, 1.0872e-04, ..., 9.6989e-04, + 7.6532e-04, 1.7047e-04], + ..., + [-1.4503e-02, -5.7182e-03, 3.1441e-05, ..., -9.7322e-04, + 1.4663e-04, 4.5717e-05], + [ 1.1177e-03, -5.0163e-04, 3.3200e-05, ..., 2.7251e-04, + 5.7507e-04, 8.0645e-05], + [ 5.8126e-04, 2.9802e-04, 1.8501e-04, ..., 5.4121e-04, + 4.9686e-04, 1.9348e-04]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0310, 0.0417, -0.0003, -0.0109, 0.0212, -0.0341, 0.0049, 0.0064, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([ 0.0046, -0.0155, 0.0201, -0.0391, 0.0145, 0.0068, -0.0089, -0.0075, + 0.0109, 0.0140], device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 214.74, cls_loss 0.4823 cls_loss_mapping 0.0008 cls_loss_causal 0.4159 re_mapping 0.0060 re_causal 0.0166 /// teacc 98.98 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.0739, -0.1632, -0.1113, ..., -0.0381, 0.0667, -0.1133], + [-0.0833, -0.0960, -0.0903, ..., 0.0998, -0.0387, 0.2409], + [ 0.0022, -0.0316, -0.0631, ..., -0.0069, -0.0102, -0.0826], + ..., + [ 0.0375, -0.1223, 0.1834, ..., 0.0391, -0.1265, 0.0626], + [-0.0262, 0.0664, -0.1541, ..., -0.0496, 0.0191, -0.1206], + [-0.0646, 0.0833, 0.0435, ..., -0.0648, -0.0558, -0.0359]], + device='cuda:0'), grad: tensor([[ 0.0008, 0.0008, 0.0009, ..., 0.0037, 0.0006, 0.0016], + [ 0.0006, 0.0009, 0.0006, ..., 0.0045, 0.0006, 0.0009], + [-0.0024, -0.0034, -0.0030, ..., -0.0041, -0.0028, -0.0012], + ..., + [ 0.0009, 0.0010, -0.0062, ..., -0.0175, 0.0006, -0.0029], + [ 0.0039, 0.0042, 0.0042, ..., 0.0058, 0.0006, 0.0023], + [ 0.0023, 0.0021, 0.0026, ..., 0.0030, -0.0002, 0.0016]], + device='cuda:0') +Epoch 470, bias, value: tensor([-0.0310, 0.0418, -0.0002, -0.0109, 0.0212, -0.0341, 0.0049, 0.0063, + -0.0027, -0.0148], device='cuda:0'), grad: tensor([ 0.0071, 0.0159, -0.0352, 0.0031, -0.0248, 0.0206, 0.0058, -0.0277, + 0.0327, 0.0024], device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 214.55, cls_loss 0.4720 cls_loss_mapping 0.0008 cls_loss_causal 0.4059 re_mapping 0.0061 re_causal 0.0165 /// teacc 98.98 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.0739, -0.1632, -0.1113, ..., -0.0381, 0.0666, -0.1133], + [-0.0832, -0.0960, -0.0903, ..., 0.0996, -0.0386, 0.2409], + [ 0.0023, -0.0316, -0.0630, ..., -0.0070, -0.0104, -0.0825], + ..., + [ 0.0374, -0.1223, 0.1835, ..., 0.0393, -0.1266, 0.0624], + [-0.0261, 0.0662, -0.1541, ..., -0.0496, 0.0191, -0.1207], + [-0.0646, 0.0833, 0.0434, ..., -0.0649, -0.0559, -0.0360]], + device='cuda:0'), grad: tensor([[ 2.9802e-04, 2.2892e-06, -1.9140e-03, ..., -1.7090e-02, + 3.5310e-04, -1.5221e-03], + [ 3.4261e-04, 2.1771e-05, 6.7329e-04, ..., -3.0727e-03, + 1.5497e-04, 3.2353e-04], + [ 2.9421e-04, -2.3234e-04, -2.4009e-04, ..., 2.0008e-03, + 9.9421e-05, -5.0497e-04], + ..., + [ 3.7551e-04, -1.2189e-04, -2.9087e-04, ..., 6.7101e-03, + 8.7440e-05, 4.9782e-04], + [ 6.3658e-04, 1.8388e-05, 2.1362e-04, ..., 4.3449e-03, + 1.0496e-04, 1.2808e-03], + [ 7.8917e-04, 1.7691e-04, 1.2684e-03, ..., 5.8823e-03, + 7.9155e-05, 1.2083e-03]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0311, 0.0418, -0.0002, -0.0109, 0.0212, -0.0340, 0.0050, 0.0063, + -0.0027, -0.0149], device='cuda:0'), grad: tensor([-0.0375, -0.0240, -0.0279, 0.0026, 0.0070, -0.0371, 0.0414, 0.0127, + 0.0262, 0.0366], device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 214.38, cls_loss 0.4676 cls_loss_mapping 0.0007 cls_loss_causal 0.4103 re_mapping 0.0061 re_causal 0.0165 /// teacc 98.99 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.0738, -0.1633, -0.1113, ..., -0.0379, 0.0665, -0.1132], + [-0.0832, -0.0961, -0.0903, ..., 0.0995, -0.0387, 0.2409], + [ 0.0022, -0.0315, -0.0627, ..., -0.0070, -0.0104, -0.0826], + ..., + [ 0.0373, -0.1224, 0.1834, ..., 0.0394, -0.1265, 0.0624], + [-0.0260, 0.0661, -0.1539, ..., -0.0495, 0.0193, -0.1206], + [-0.0644, 0.0834, 0.0434, ..., -0.0649, -0.0559, -0.0358]], + device='cuda:0'), grad: tensor([[ 3.8445e-06, -1.7047e-04, 7.0930e-05, ..., 5.9128e-04, + -9.6464e-04, 3.9387e-04], + [-7.1585e-05, 3.6478e-05, -1.6499e-04, ..., -5.7888e-04, + 2.2445e-07, -6.6137e-04], + [-3.8773e-05, 8.6188e-05, -5.6028e-05, ..., 2.1911e-04, + 2.3976e-05, 3.1781e-04], + ..., + [ 7.5400e-05, 9.5510e-04, -1.2651e-05, ..., 6.7234e-04, + 1.5739e-07, -3.1614e-04], + [-3.3472e-06, 3.6659e-03, 3.5018e-05, ..., 4.2725e-03, + 2.7820e-05, 2.0191e-05], + [ 8.9929e-06, -6.6719e-03, 4.1515e-05, ..., -7.4959e-03, + 1.4231e-05, 1.0359e-04]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0311, 0.0417, -0.0004, -0.0111, 0.0211, -0.0339, 0.0050, 0.0065, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([ 0.0024, 0.0025, 0.0036, 0.0094, 0.0032, -0.0280, 0.0029, 0.0077, + 0.0166, -0.0203], device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 214.45, cls_loss 0.4893 cls_loss_mapping 0.0008 cls_loss_causal 0.4213 re_mapping 0.0061 re_causal 0.0168 /// teacc 99.00 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.0737, -0.1633, -0.1115, ..., -0.0379, 0.0665, -0.1131], + [-0.0832, -0.0961, -0.0903, ..., 0.0995, -0.0388, 0.2409], + [ 0.0020, -0.0315, -0.0627, ..., -0.0069, -0.0104, -0.0827], + ..., + [ 0.0377, -0.1223, 0.1832, ..., 0.0394, -0.1266, 0.0623], + [-0.0260, 0.0661, -0.1539, ..., -0.0494, 0.0194, -0.1206], + [-0.0645, 0.0834, 0.0437, ..., -0.0648, -0.0559, -0.0358]], + device='cuda:0'), grad: tensor([[ 8.1122e-05, 2.9609e-05, 1.2331e-05, ..., 1.9703e-03, + 4.5359e-05, 6.6605e-03], + [ 5.6219e-04, 1.4460e-04, 1.1042e-05, ..., 4.9210e-04, + 3.4273e-05, 1.8244e-03], + [-6.4278e-03, -1.2674e-03, 1.9282e-05, ..., -3.1948e-05, + 1.2150e-03, -8.6975e-03], + ..., + [-3.0965e-05, 2.3559e-05, -4.1056e-04, ..., 5.2065e-05, + 2.7016e-05, -9.9480e-05], + [ 5.9967e-03, 1.0071e-03, 2.6733e-05, ..., 8.0645e-05, + 6.7759e-04, 6.7635e-03], + [-1.1663e-03, 3.9177e-03, 3.8934e-04, ..., 7.2002e-05, + 1.5855e-05, 5.1651e-03]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0310, 0.0417, -0.0005, -0.0111, 0.0212, -0.0340, 0.0048, 0.0064, + -0.0025, -0.0146], device='cuda:0'), grad: tensor([ 0.0059, 0.0034, -0.0067, 0.0021, -0.0030, -0.0275, -0.0153, -0.0002, + 0.0268, 0.0146], device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 214.85, cls_loss 0.4848 cls_loss_mapping 0.0009 cls_loss_causal 0.4227 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.98 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.0737, -0.1634, -0.1115, ..., -0.0379, 0.0664, -0.1132], + [-0.0832, -0.0960, -0.0904, ..., 0.0995, -0.0388, 0.2409], + [ 0.0021, -0.0316, -0.0627, ..., -0.0067, -0.0104, -0.0825], + ..., + [ 0.0377, -0.1223, 0.1833, ..., 0.0394, -0.1266, 0.0623], + [-0.0263, 0.0662, -0.1537, ..., -0.0495, 0.0196, -0.1204], + [-0.0646, 0.0834, 0.0436, ..., -0.0648, -0.0557, -0.0360]], + device='cuda:0'), grad: tensor([[ 7.4339e-04, -2.9206e-05, 7.0691e-05, ..., 1.7529e-03, + 9.3341e-05, 1.3456e-03], + [ 7.2145e-04, 2.1234e-06, 3.4332e-05, ..., 7.4053e-04, + 9.3952e-06, 5.4550e-04], + [ 1.1034e-03, 9.4995e-06, 2.1994e-04, ..., -6.5517e-04, + 3.7241e-04, 1.4420e-03], + ..., + [ 1.9913e-03, 2.8405e-07, -5.3024e-04, ..., 6.3782e-03, + 5.7173e-04, 6.0768e-03], + [ 9.7609e-04, 1.2815e-04, -1.1063e-04, ..., 2.4185e-03, + 4.2868e-04, 9.2077e-04], + [ 2.1038e-03, 1.4111e-05, 1.6105e-04, ..., -3.2921e-03, + 4.4870e-04, -1.0483e-02]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0309, 0.0417, -0.0004, -0.0112, 0.0213, -0.0339, 0.0048, 0.0064, + -0.0026, -0.0146], device='cuda:0'), grad: tensor([ 0.0136, 0.0136, -0.0125, -0.0400, -0.0091, 0.0140, 0.0098, 0.0383, + 0.0033, -0.0310], device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 214.51, cls_loss 0.4639 cls_loss_mapping 0.0008 cls_loss_causal 0.4044 re_mapping 0.0060 re_causal 0.0161 /// teacc 98.99 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.0737, -0.1635, -0.1113, ..., -0.0379, 0.0665, -0.1134], + [-0.0833, -0.0961, -0.0904, ..., 0.0994, -0.0388, 0.2410], + [ 0.0021, -0.0317, -0.0627, ..., -0.0067, -0.0105, -0.0826], + ..., + [ 0.0377, -0.1222, 0.1833, ..., 0.0395, -0.1267, 0.0624], + [-0.0263, 0.0663, -0.1537, ..., -0.0496, 0.0196, -0.1205], + [-0.0645, 0.0836, 0.0436, ..., -0.0648, -0.0556, -0.0359]], + device='cuda:0'), grad: tensor([[ 2.5253e-03, -9.0027e-04, 2.5839e-05, ..., 6.3782e-03, + -2.5129e-04, 2.9716e-03], + [ 1.9372e-04, -3.0589e-04, 3.5819e-06, ..., -4.6563e-04, + 4.4632e-04, 3.0541e-04], + [ 4.8923e-04, -5.2392e-05, 2.5511e-04, ..., 1.6522e-04, + 4.5562e-04, 6.9666e-04], + ..., + [ 1.9016e-03, 3.0994e-03, 1.5306e-03, ..., 6.7253e-03, + 5.4407e-04, 2.8934e-03], + [-2.3689e-03, -3.1281e-03, -2.5196e-03, ..., -5.7487e-03, + 6.1798e-04, -3.9635e-03], + [-3.7460e-03, 9.9754e-04, 1.1951e-04, ..., -5.4626e-03, + -4.5815e-03, -4.2419e-03]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0309, 0.0417, -0.0004, -0.0113, 0.0212, -0.0339, 0.0048, 0.0064, + -0.0026, -0.0145], device='cuda:0'), grad: tensor([ 0.0425, 0.0009, -0.0009, -0.0331, -0.0029, 0.0056, 0.0354, 0.0452, + -0.0331, -0.0596], device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 214.64, cls_loss 0.4544 cls_loss_mapping 0.0009 cls_loss_causal 0.3868 re_mapping 0.0061 re_causal 0.0160 /// teacc 98.98 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.0738, -0.1634, -0.1114, ..., -0.0380, 0.0665, -0.1134], + [-0.0833, -0.0961, -0.0904, ..., 0.0995, -0.0386, 0.2410], + [ 0.0022, -0.0317, -0.0627, ..., -0.0066, -0.0104, -0.0828], + ..., + [ 0.0379, -0.1221, 0.1834, ..., 0.0395, -0.1267, 0.0624], + [-0.0263, 0.0662, -0.1538, ..., -0.0496, 0.0197, -0.1204], + [-0.0647, 0.0835, 0.0435, ..., -0.0650, -0.0558, -0.0360]], + device='cuda:0'), grad: tensor([[ 6.1941e-04, 2.1338e-04, 4.6825e-04, ..., 2.1629e-03, + 6.4611e-05, 7.6103e-04], + [-1.2283e-03, 3.3641e-04, 6.0940e-04, ..., -1.5316e-03, + 7.6294e-05, 9.6083e-04], + [ 1.2751e-03, 6.4087e-04, 5.3310e-04, ..., -2.9099e-02, + 1.6356e-04, -1.5381e-02], + ..., + [-2.4834e-03, 1.1549e-03, 1.5097e-03, ..., 3.0533e-02, + -1.7083e-04, 1.3985e-02], + [ 4.2458e-03, -7.5960e-04, 1.9121e-03, ..., 6.5460e-03, + -4.4465e-04, 4.1122e-03], + [-2.5196e-03, 3.0494e-04, 5.5552e-04, ..., -3.7384e-03, + 4.8494e-04, 1.0366e-03]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0310, 0.0417, -0.0003, -0.0112, 0.0211, -0.0339, 0.0047, 0.0065, + -0.0025, -0.0147], device='cuda:0'), grad: tensor([ 0.0105, 0.0048, -0.0211, -0.0026, -0.0217, 0.0139, -0.0238, 0.0351, + 0.0191, -0.0141], device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 215.01, cls_loss 0.4877 cls_loss_mapping 0.0008 cls_loss_causal 0.4208 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.95 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.0738, -0.1633, -0.1113, ..., -0.0379, 0.0667, -0.1134], + [-0.0833, -0.0961, -0.0903, ..., 0.0996, -0.0387, 0.2411], + [ 0.0020, -0.0318, -0.0629, ..., -0.0069, -0.0105, -0.0829], + ..., + [ 0.0378, -0.1220, 0.1835, ..., 0.0395, -0.1267, 0.0623], + [-0.0263, 0.0662, -0.1538, ..., -0.0496, 0.0197, -0.1204], + [-0.0646, 0.0834, 0.0434, ..., -0.0650, -0.0558, -0.0360]], + device='cuda:0'), grad: tensor([[ 4.1127e-04, -7.4005e-04, 4.9919e-06, ..., -1.4315e-03, + 1.7297e-04, 3.3903e-04], + [-4.3602e-03, 9.9838e-05, 7.2792e-06, ..., -1.2875e-03, + 2.5868e-04, 4.4680e-04], + [ 1.2636e-03, 3.6669e-04, 3.5495e-05, ..., 1.1873e-03, + 1.8764e-04, 3.6430e-04], + ..., + [ 4.6825e-04, 1.0443e-04, 1.4670e-05, ..., 9.8610e-04, + 2.1219e-04, 4.1509e-04], + [ 1.8644e-03, -5.2452e-05, 1.2243e-04, ..., 1.7891e-03, + -3.2711e-04, 5.5027e-04], + [ 4.7135e-04, 1.4496e-04, 1.1086e-05, ..., 9.0837e-04, + 1.8740e-04, 3.6263e-04]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0309, 0.0417, -0.0006, -0.0113, 0.0211, -0.0338, 0.0048, 0.0066, + -0.0024, -0.0147], device='cuda:0'), grad: tensor([-0.0257, -0.0120, 0.0150, -0.0060, -0.0215, 0.0080, 0.0152, 0.0052, + 0.0140, 0.0078], device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 214.85, cls_loss 0.4968 cls_loss_mapping 0.0009 cls_loss_causal 0.4354 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.94 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.0737, -0.1633, -0.1112, ..., -0.0378, 0.0667, -0.1132], + [-0.0831, -0.0961, -0.0902, ..., 0.0996, -0.0390, 0.2413], + [ 0.0019, -0.0317, -0.0629, ..., -0.0068, -0.0105, -0.0828], + ..., + [ 0.0378, -0.1220, 0.1834, ..., 0.0394, -0.1267, 0.0621], + [-0.0263, 0.0662, -0.1540, ..., -0.0496, 0.0196, -0.1205], + [-0.0646, 0.0835, 0.0434, ..., -0.0649, -0.0557, -0.0360]], + device='cuda:0'), grad: tensor([[ 5.2869e-05, 4.6563e-04, -3.5381e-04, ..., 1.0195e-03, + 3.7217e-04, 7.7295e-04], + [ 3.8892e-05, -6.6710e-04, 9.7096e-05, ..., 2.1124e-04, + 2.2426e-05, 2.0790e-04], + [ 4.8965e-05, 3.9220e-04, 1.9610e-04, ..., 1.6525e-02, + 4.4942e-05, 8.4162e-04], + ..., + [ 4.6581e-05, 3.4070e-04, -8.5533e-05, ..., -1.1307e-02, + 6.0126e-06, -3.8681e-03], + [-3.6573e-04, 8.2350e-04, 8.1968e-04, ..., -8.1940e-03, + 2.9862e-05, -5.0888e-03], + [ 2.5824e-05, 1.0538e-03, 1.6105e-04, ..., 2.2411e-03, + 3.4064e-05, 4.1533e-04]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0309, 0.0417, -0.0005, -0.0113, 0.0211, -0.0338, 0.0047, 0.0064, + -0.0025, -0.0146], device='cuda:0'), grad: tensor([-0.0075, -0.0167, 0.0295, -0.0177, 0.0490, 0.0186, -0.0326, -0.0289, + -0.0139, 0.0203], device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 214.65, cls_loss 0.4678 cls_loss_mapping 0.0009 cls_loss_causal 0.4048 re_mapping 0.0059 re_causal 0.0160 /// teacc 98.96 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.0738, -0.1634, -0.1111, ..., -0.0378, 0.0667, -0.1133], + [-0.0830, -0.0961, -0.0901, ..., 0.0996, -0.0390, 0.2413], + [ 0.0018, -0.0317, -0.0630, ..., -0.0067, -0.0105, -0.0827], + ..., + [ 0.0378, -0.1221, 0.1835, ..., 0.0394, -0.1268, 0.0622], + [-0.0264, 0.0663, -0.1540, ..., -0.0496, 0.0197, -0.1206], + [-0.0647, 0.0835, 0.0434, ..., -0.0650, -0.0556, -0.0363]], + device='cuda:0'), grad: tensor([[ 8.2552e-05, 2.3627e-04, 2.6751e-04, ..., 7.1144e-04, + 1.4938e-05, 2.1827e-04], + [ 5.7667e-06, 1.9054e-03, 3.2091e-04, ..., 8.1635e-03, + 2.0131e-05, 1.9608e-03], + [ 3.1620e-05, 1.1568e-03, 1.3924e-03, ..., 2.4948e-03, + 1.0622e-04, 3.3331e-04], + ..., + [ 2.2972e-04, 8.1682e-04, 9.7084e-04, ..., -7.3471e-03, + 5.7548e-05, -3.0918e-03], + [ 2.1601e-04, -1.5669e-03, 3.3784e-04, ..., -2.8152e-03, + 1.6317e-05, 8.1420e-05], + [-8.7786e-04, -2.0409e-04, -1.1748e-04, ..., -2.4438e-04, + 1.5795e-05, -5.5361e-04]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0310, 0.0418, -0.0005, -0.0114, 0.0212, -0.0338, 0.0049, 0.0064, + -0.0025, -0.0147], device='cuda:0'), grad: tensor([ 0.0082, 0.0338, 0.0151, -0.0066, 0.0097, 0.0086, 0.0078, -0.0175, + -0.0019, -0.0572], device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 214.73, cls_loss 0.4791 cls_loss_mapping 0.0008 cls_loss_causal 0.4171 re_mapping 0.0062 re_causal 0.0166 /// teacc 98.91 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.0737, -0.1635, -0.1112, ..., -0.0378, 0.0667, -0.1133], + [-0.0830, -0.0961, -0.0902, ..., 0.0997, -0.0391, 0.2412], + [ 0.0017, -0.0316, -0.0630, ..., -0.0066, -0.0106, -0.0827], + ..., + [ 0.0380, -0.1220, 0.1837, ..., 0.0393, -0.1266, 0.0624], + [-0.0264, 0.0663, -0.1542, ..., -0.0498, 0.0198, -0.1207], + [-0.0648, 0.0834, 0.0433, ..., -0.0650, -0.0557, -0.0363]], + device='cuda:0'), grad: tensor([[ 9.0957e-05, 1.9777e-04, 1.9801e-04, ..., 7.3862e-04, + 4.0054e-04, 2.8419e-04], + [ 1.3056e-03, 1.1702e-03, -6.1810e-05, ..., 1.6918e-03, + 3.6907e-04, 1.8358e-03], + [ 1.0520e-04, 3.2210e-04, 2.4366e-04, ..., 1.1415e-03, + 5.3120e-04, 6.1226e-04], + ..., + [ 2.0707e-04, -5.7487e-03, 1.1891e-04, ..., 1.2255e-03, + 2.1601e-04, -6.9695e-03], + [-2.9697e-03, 2.9716e-03, -1.0127e-04, ..., -4.4594e-03, + 4.7636e-04, -4.8332e-03], + [ 5.1308e-04, 2.9349e-04, 1.4591e-04, ..., 1.6098e-03, + -2.8706e-03, 5.2834e-03]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0310, 0.0418, -0.0005, -0.0113, 0.0213, -0.0338, 0.0049, 0.0065, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([ 0.0088, 0.0124, 0.0117, 0.0125, -0.0109, 0.0438, -0.0202, -0.0135, + -0.0327, -0.0119], device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 215.04, cls_loss 0.4659 cls_loss_mapping 0.0007 cls_loss_causal 0.4034 re_mapping 0.0061 re_causal 0.0166 /// teacc 98.96 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.0738, -0.1636, -0.1113, ..., -0.0379, 0.0666, -0.1134], + [-0.0831, -0.0962, -0.0904, ..., 0.0997, -0.0393, 0.2412], + [ 0.0017, -0.0316, -0.0630, ..., -0.0067, -0.0103, -0.0828], + ..., + [ 0.0380, -0.1220, 0.1838, ..., 0.0395, -0.1265, 0.0625], + [-0.0263, 0.0664, -0.1542, ..., -0.0499, 0.0196, -0.1209], + [-0.0648, 0.0836, 0.0435, ..., -0.0650, -0.0557, -0.0362]], + device='cuda:0'), grad: tensor([[ 3.2640e-04, -1.2743e-04, 1.2279e-04, ..., 7.6628e-04, + 1.0318e-04, 1.1241e-04], + [ 3.4022e-04, 2.6536e-04, 1.8740e-04, ..., 1.1740e-03, + 1.7476e-04, -8.3685e-05], + [ 8.3971e-04, 3.0208e-04, 1.7214e-04, ..., 1.9026e-03, + 4.1938e-04, 4.2886e-05], + ..., + [ 4.6730e-04, 2.7871e-04, 8.7380e-05, ..., 1.0033e-03, + 1.8966e-04, -3.4165e-04], + [ 5.1594e-04, 1.3077e-04, 9.3579e-05, ..., 1.3084e-03, + 2.6345e-04, 7.4446e-05], + [ 5.0688e-04, 4.1151e-04, 3.0112e-04, ..., 1.2293e-03, + 2.2602e-04, 8.7917e-06]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0311, 0.0417, -0.0006, -0.0113, 0.0213, -0.0337, 0.0049, 0.0066, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([ 0.0082, 0.0148, 0.0179, -0.0696, 0.0147, 0.0197, -0.0149, -0.0210, + 0.0137, 0.0167], device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 215.03, cls_loss 0.5139 cls_loss_mapping 0.0008 cls_loss_causal 0.4440 re_mapping 0.0062 re_causal 0.0170 /// teacc 98.99 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.0738, -0.1636, -0.1113, ..., -0.0377, 0.0666, -0.1134], + [-0.0831, -0.0962, -0.0904, ..., 0.0997, -0.0393, 0.2410], + [ 0.0018, -0.0317, -0.0630, ..., -0.0067, -0.0103, -0.0825], + ..., + [ 0.0380, -0.1220, 0.1838, ..., 0.0395, -0.1265, 0.0626], + [-0.0263, 0.0665, -0.1542, ..., -0.0499, 0.0198, -0.1208], + [-0.0648, 0.0836, 0.0435, ..., -0.0650, -0.0557, -0.0363]], + device='cuda:0'), grad: tensor([[ 1.1367e-04, 1.2083e-03, 4.4394e-04, ..., 2.0733e-03, + 1.0824e-03, 6.1369e-04], + [-7.7307e-05, -1.9588e-03, 3.1090e-04, ..., -4.7073e-03, + -1.7433e-03, 1.3266e-03], + [-1.4627e-04, 7.3471e-03, 2.3186e-04, ..., -1.2144e-05, + 1.0231e-02, 1.7679e-04], + ..., + [ 2.9111e-04, 3.2663e-04, -5.4970e-03, ..., -8.4763e-03, + 7.6711e-05, -1.8549e-03], + [-1.8864e-03, -8.5354e-04, 2.9588e-04, ..., -1.2484e-03, + -2.5070e-02, 8.8882e-04], + [ 4.0555e-04, -3.8195e-04, 1.7147e-03, ..., 2.8973e-03, + 2.3735e-04, 1.9913e-03]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0310, 0.0417, -0.0006, -0.0113, 0.0212, -0.0338, 0.0049, 0.0066, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([ 0.0212, -0.0109, 0.0322, 0.0036, 0.0009, -0.0031, 0.0396, -0.0795, + -0.0264, 0.0224], device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 214.80, cls_loss 0.4861 cls_loss_mapping 0.0006 cls_loss_causal 0.4272 re_mapping 0.0062 re_causal 0.0171 /// teacc 99.02 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.0738, -0.1635, -0.1112, ..., -0.0377, 0.0666, -0.1133], + [-0.0833, -0.0961, -0.0905, ..., 0.0998, -0.0394, 0.2410], + [ 0.0018, -0.0317, -0.0630, ..., -0.0067, -0.0104, -0.0825], + ..., + [ 0.0380, -0.1220, 0.1839, ..., 0.0395, -0.1266, 0.0625], + [-0.0263, 0.0664, -0.1542, ..., -0.0500, 0.0200, -0.1209], + [-0.0648, 0.0835, 0.0435, ..., -0.0650, -0.0558, -0.0364]], + device='cuda:0'), grad: tensor([[ 6.0827e-05, 6.4969e-05, 2.7347e-04, ..., 1.6060e-03, + 4.1842e-05, 1.1164e-04], + [ 1.0794e-04, -1.9181e-04, 3.2854e-04, ..., -1.3089e-04, + 4.0829e-05, 1.8990e-04], + [ 3.0780e-04, 5.2881e-04, 6.1941e-04, ..., 1.4925e-03, + 3.5524e-05, 7.1239e-04], + ..., + [-1.0214e-03, -1.2102e-03, -1.9321e-03, ..., -9.4376e-03, + 2.7609e-04, -2.0046e-03], + [ 2.1064e-04, 3.0112e-04, 3.4380e-04, ..., 9.4557e-04, + 1.5020e-03, 3.9911e-04], + [ 2.0957e-04, 2.2602e-04, 8.8024e-04, ..., 1.9989e-03, + 1.8501e-04, 6.9714e-04]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0309, 0.0416, -0.0006, -0.0114, 0.0213, -0.0338, 0.0050, 0.0066, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0084, -0.0225, 0.0082, 0.0057, 0.0061, -0.0154, 0.0240, -0.0298, + 0.0049, 0.0105], device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 214.97, cls_loss 0.5004 cls_loss_mapping 0.0008 cls_loss_causal 0.4306 re_mapping 0.0060 re_causal 0.0164 /// teacc 98.99 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.0739, -0.1636, -0.1112, ..., -0.0379, 0.0666, -0.1133], + [-0.0831, -0.0962, -0.0904, ..., 0.0998, -0.0393, 0.2410], + [ 0.0018, -0.0318, -0.0630, ..., -0.0068, -0.0105, -0.0824], + ..., + [ 0.0381, -0.1220, 0.1838, ..., 0.0395, -0.1266, 0.0623], + [-0.0263, 0.0664, -0.1543, ..., -0.0498, 0.0199, -0.1208], + [-0.0648, 0.0835, 0.0434, ..., -0.0650, -0.0559, -0.0365]], + device='cuda:0'), grad: tensor([[ 2.7428e-03, -1.9741e-03, 6.9189e-04, ..., 6.7282e-04, + 1.6272e-05, 2.6798e-03], + [ 5.0049e-03, 2.2621e-03, 2.4624e-03, ..., 6.0654e-03, + 2.1964e-05, 4.7340e-03], + [-8.0414e-03, -2.2717e-03, 4.3988e-04, ..., -2.0981e-03, + 3.2522e-06, -6.2370e-03], + ..., + [ 5.7983e-04, 5.5838e-04, 9.1362e-04, ..., 3.3436e-03, + 1.4931e-05, 4.4584e-04], + [ 2.1610e-03, 1.8902e-03, 1.8673e-03, ..., 4.7417e-03, + 1.5318e-04, 1.9264e-03], + [ 1.2608e-03, 1.1654e-03, 1.0309e-03, ..., 2.5158e-03, + 1.5056e-04, 1.1072e-03]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0309, 0.0418, -0.0007, -0.0114, 0.0214, -0.0338, 0.0050, 0.0065, + -0.0025, -0.0150], device='cuda:0'), grad: tensor([-0.0094, 0.0287, -0.0115, -0.0361, 0.0179, -0.0106, -0.0214, 0.0107, + 0.0205, 0.0111], device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 214.83, cls_loss 0.4764 cls_loss_mapping 0.0007 cls_loss_causal 0.4096 re_mapping 0.0060 re_causal 0.0164 /// teacc 98.97 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.0738, -0.1637, -0.1111, ..., -0.0380, 0.0667, -0.1132], + [-0.0831, -0.0962, -0.0905, ..., 0.0998, -0.0394, 0.2412], + [ 0.0018, -0.0319, -0.0631, ..., -0.0068, -0.0105, -0.0824], + ..., + [ 0.0380, -0.1222, 0.1836, ..., 0.0393, -0.1267, 0.0621], + [-0.0263, 0.0665, -0.1544, ..., -0.0498, 0.0199, -0.1208], + [-0.0647, 0.0836, 0.0434, ..., -0.0650, -0.0560, -0.0365]], + device='cuda:0'), grad: tensor([[-0.0017, -0.0007, 0.0004, ..., -0.0073, 0.0001, 0.0009], + [ 0.0005, 0.0004, 0.0006, ..., 0.0039, 0.0002, 0.0003], + [-0.0016, -0.0001, -0.0029, ..., -0.0076, -0.0011, 0.0003], + ..., + [-0.0025, -0.0003, -0.0054, ..., -0.0022, 0.0002, -0.0055], + [ 0.0009, 0.0007, 0.0008, ..., 0.0040, 0.0001, 0.0007], + [ 0.0014, 0.0035, 0.0037, ..., 0.0064, 0.0002, 0.0022]], + device='cuda:0') +Epoch 485, bias, value: tensor([-0.0310, 0.0419, -0.0008, -0.0114, 0.0213, -0.0338, 0.0052, 0.0065, + -0.0024, -0.0149], device='cuda:0'), grad: tensor([-0.0242, 0.0184, -0.0453, 0.0048, -0.0176, 0.0340, -0.0132, -0.0066, + 0.0186, 0.0312], device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 214.97, cls_loss 0.4846 cls_loss_mapping 0.0008 cls_loss_causal 0.4156 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.97 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.0738, -0.1636, -0.1112, ..., -0.0379, 0.0667, -0.1132], + [-0.0831, -0.0963, -0.0904, ..., 0.0998, -0.0393, 0.2411], + [ 0.0019, -0.0320, -0.0633, ..., -0.0069, -0.0106, -0.0823], + ..., + [ 0.0381, -0.1221, 0.1836, ..., 0.0394, -0.1268, 0.0621], + [-0.0265, 0.0663, -0.1545, ..., -0.0498, 0.0198, -0.1210], + [-0.0647, 0.0837, 0.0436, ..., -0.0650, -0.0559, -0.0365]], + device='cuda:0'), grad: tensor([[ 6.8617e-04, -1.0654e-05, 5.1588e-05, ..., 1.4114e-03, + 2.0075e-04, 4.5156e-04], + [-4.5509e-03, 8.4698e-05, 4.6700e-05, ..., 1.0290e-03, + 1.2791e-04, 5.4789e-04], + [ 7.6818e-04, 1.4877e-04, 2.1095e-03, ..., 3.6583e-03, + 6.6757e-05, 3.0875e-04], + ..., + [ 4.8828e-04, -2.0242e-04, 5.6922e-05, ..., -4.9973e-04, + 3.2693e-05, 1.9383e-04], + [ 5.4216e-04, 3.9053e-04, 2.7037e-04, ..., 1.7099e-03, + 5.0211e-04, 3.7813e-04], + [ 6.9761e-04, 1.0985e-04, 3.0056e-05, ..., 1.1921e-03, + 7.0155e-05, 3.4523e-04]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0310, 0.0418, -0.0008, -0.0114, 0.0213, -0.0340, 0.0052, 0.0066, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0163, -0.0113, 0.0201, -0.0191, -0.0139, -0.0264, 0.0075, -0.0129, + 0.0201, 0.0196], device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 215.08, cls_loss 0.4710 cls_loss_mapping 0.0008 cls_loss_causal 0.4079 re_mapping 0.0062 re_causal 0.0165 /// teacc 99.00 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.0737, -0.1636, -0.1112, ..., -0.0380, 0.0667, -0.1133], + [-0.0832, -0.0962, -0.0903, ..., 0.0999, -0.0393, 0.2411], + [ 0.0019, -0.0320, -0.0632, ..., -0.0069, -0.0106, -0.0823], + ..., + [ 0.0382, -0.1221, 0.1836, ..., 0.0394, -0.1267, 0.0621], + [-0.0264, 0.0663, -0.1546, ..., -0.0497, 0.0197, -0.1209], + [-0.0648, 0.0836, 0.0435, ..., -0.0650, -0.0560, -0.0365]], + device='cuda:0'), grad: tensor([[-8.1635e-04, 1.3185e-04, 8.3923e-05, ..., -8.9417e-03, + 5.5879e-08, 2.4259e-05], + [-1.6890e-03, 1.9431e-05, 1.3745e-04, ..., -5.7983e-03, + 1.7695e-06, 5.5343e-05], + [ 3.0308e-03, 2.6345e-05, 2.9397e-04, ..., 6.2637e-03, + 1.4249e-07, 1.6198e-05], + ..., + [ 1.5030e-03, 1.5616e-05, 5.8842e-04, ..., 3.4294e-03, + 7.3314e-06, 3.3426e-04], + [ 9.0551e-04, -6.1333e-05, 3.8356e-05, ..., 2.7161e-03, + 1.2387e-07, -4.7591e-07], + [-2.0561e-03, 1.0794e-04, 3.1322e-05, ..., -4.9248e-03, + 2.0102e-05, 1.0061e-03]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0310, 0.0419, -0.0008, -0.0115, 0.0211, -0.0340, 0.0051, 0.0066, + -0.0023, -0.0148], device='cuda:0'), grad: tensor([-0.0449, -0.0218, 0.0348, -0.0141, 0.0063, 0.0085, 0.0137, 0.0158, + 0.0158, -0.0141], device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 214.95, cls_loss 0.4836 cls_loss_mapping 0.0008 cls_loss_causal 0.4272 re_mapping 0.0060 re_causal 0.0164 /// teacc 99.02 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.0736, -0.1637, -0.1114, ..., -0.0380, 0.0667, -0.1133], + [-0.0832, -0.0962, -0.0904, ..., 0.1000, -0.0394, 0.2412], + [ 0.0018, -0.0318, -0.0632, ..., -0.0069, -0.0106, -0.0824], + ..., + [ 0.0382, -0.1221, 0.1836, ..., 0.0395, -0.1266, 0.0622], + [-0.0265, 0.0664, -0.1544, ..., -0.0497, 0.0197, -0.1209], + [-0.0648, 0.0836, 0.0434, ..., -0.0649, -0.0560, -0.0367]], + device='cuda:0'), grad: tensor([[ 6.1750e-04, 5.3453e-04, 3.1972e-04, ..., 1.3485e-03, + 2.9397e-04, 2.5415e-04], + [ 2.6178e-04, 1.8156e-04, 4.5836e-05, ..., 1.1835e-03, + 3.9279e-05, 1.9944e-04], + [ 1.0529e-03, 1.0490e-03, 6.9237e-04, ..., 2.5768e-03, + 5.6744e-04, 7.3338e-04], + ..., + [ 3.6335e-04, 2.3282e-04, 8.4877e-05, ..., -1.8129e-03, + 7.4208e-05, -1.0672e-03], + [ 1.8101e-03, 3.2926e-04, 2.5392e-04, ..., 1.3990e-03, + 2.5177e-04, 1.9562e-04], + [ 2.3232e-03, -3.8586e-03, 2.5702e-04, ..., 1.9760e-03, + 2.0707e-04, 3.7789e-04]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0311, 0.0418, -0.0007, -0.0113, 0.0212, -0.0341, 0.0050, 0.0066, + -0.0024, -0.0147], device='cuda:0'), grad: tensor([ 0.0117, 0.0131, 0.0180, -0.0495, 0.0225, -0.0069, -0.0160, -0.0190, + 0.0164, 0.0095], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 487---------------------------------------------------- +epoch 487, time 231.08, cls_loss 0.5005 cls_loss_mapping 0.0008 cls_loss_causal 0.4342 re_mapping 0.0061 re_causal 0.0168 /// teacc 99.05 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.0736, -0.1638, -0.1113, ..., -0.0381, 0.0665, -0.1133], + [-0.0832, -0.0963, -0.0904, ..., 0.0999, -0.0394, 0.2412], + [ 0.0018, -0.0318, -0.0632, ..., -0.0069, -0.0105, -0.0823], + ..., + [ 0.0382, -0.1221, 0.1837, ..., 0.0396, -0.1267, 0.0622], + [-0.0265, 0.0665, -0.1544, ..., -0.0497, 0.0197, -0.1210], + [-0.0648, 0.0836, 0.0434, ..., -0.0651, -0.0561, -0.0366]], + device='cuda:0'), grad: tensor([[ 8.2850e-05, -6.6662e-04, 9.0338e-08, ..., -1.4467e-03, + 6.2399e-08, -3.6340e-06], + [ 1.0312e-04, 3.2997e-04, 9.3970e-07, ..., 1.8044e-03, + 2.2352e-08, -1.7345e-05], + [ 1.1402e-04, 2.5392e-04, -3.1199e-06, ..., 1.4477e-03, + 5.9046e-07, -5.7548e-05], + ..., + [ 8.9943e-05, 3.0112e-04, 2.2545e-05, ..., 1.7157e-03, + 1.0245e-08, 6.9380e-05], + [ 2.2674e-04, 3.1495e-04, 3.5334e-06, ..., 2.0561e-03, + 1.4296e-06, 1.9986e-06], + [-1.0319e-03, -3.2806e-04, -3.7253e-05, ..., 1.5497e-03, + 5.8860e-07, -8.8066e-06]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0311, 0.0418, -0.0006, -0.0114, 0.0214, -0.0341, 0.0050, 0.0067, + -0.0024, -0.0148], device='cuda:0'), grad: tensor([-0.0209, 0.0145, 0.0120, 0.0122, 0.0146, -0.0210, -0.0486, 0.0133, + 0.0142, 0.0098], device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 214.82, cls_loss 0.4755 cls_loss_mapping 0.0008 cls_loss_causal 0.4105 re_mapping 0.0060 re_causal 0.0161 /// teacc 99.03 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.0737, -0.1638, -0.1113, ..., -0.0381, 0.0667, -0.1133], + [-0.0831, -0.0963, -0.0903, ..., 0.0999, -0.0394, 0.2413], + [ 0.0017, -0.0318, -0.0631, ..., -0.0068, -0.0105, -0.0822], + ..., + [ 0.0384, -0.1221, 0.1837, ..., 0.0398, -0.1268, 0.0622], + [-0.0266, 0.0664, -0.1543, ..., -0.0498, 0.0196, -0.1211], + [-0.0649, 0.0838, 0.0434, ..., -0.0651, -0.0561, -0.0366]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 1.8287e-04, 2.8554e-06, ..., 1.9665e-03, + 3.2640e-04, 1.0431e-04], + [ 6.4790e-05, 5.1022e-04, 3.9482e-04, ..., 3.1452e-03, + 5.3495e-05, 1.0796e-03], + [ 4.1164e-07, 9.4593e-05, 2.2665e-05, ..., 1.6861e-03, + -8.4352e-04, 1.9503e-04], + ..., + [ 1.3256e-03, 4.1652e-04, 7.8201e-03, ..., 8.6517e-03, + 1.7095e-04, 1.5511e-02], + [ 1.3374e-05, 1.6289e-03, 2.4378e-04, ..., 5.4169e-03, + 7.1812e-04, 1.0767e-03], + [-1.5316e-03, 3.8948e-03, -8.6594e-03, ..., -1.4893e-02, + 8.1253e-04, -1.6342e-02]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0312, 0.0417, -0.0006, -0.0113, 0.0213, -0.0340, 0.0050, 0.0068, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0128, 0.0182, 0.0088, -0.0123, -0.0675, 0.0153, 0.0210, 0.0034, + 0.0258, -0.0253], device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 214.72, cls_loss 0.4683 cls_loss_mapping 0.0009 cls_loss_causal 0.4003 re_mapping 0.0060 re_causal 0.0157 /// teacc 99.01 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.0736, -0.1638, -0.1114, ..., -0.0379, 0.0665, -0.1132], + [-0.0832, -0.0964, -0.0903, ..., 0.0998, -0.0394, 0.2411], + [ 0.0017, -0.0319, -0.0630, ..., -0.0068, -0.0107, -0.0822], + ..., + [ 0.0384, -0.1222, 0.1835, ..., 0.0397, -0.1269, 0.0621], + [-0.0265, 0.0664, -0.1542, ..., -0.0497, 0.0196, -0.1210], + [-0.0650, 0.0839, 0.0434, ..., -0.0651, -0.0562, -0.0366]], + device='cuda:0'), grad: tensor([[-2.1896e-02, -3.5667e-03, 9.7975e-07, ..., -1.2650e-02, + -2.0866e-03, -1.8463e-03], + [ 1.7910e-03, 1.8644e-04, 2.7567e-07, ..., 3.4676e-03, + 5.7173e-04, 1.4439e-03], + [ 2.4490e-03, 8.9455e-04, 4.8459e-05, ..., 2.4452e-03, + 2.7809e-03, 5.9843e-04], + ..., + [ 1.5326e-03, 1.6248e-04, 1.7866e-05, ..., 3.5877e-03, + 2.2662e-04, 1.6031e-03], + [ 1.4687e-03, 1.6422e-03, 5.8651e-05, ..., 2.5387e-03, + 1.1482e-03, 8.9550e-04], + [ 1.2255e-03, 1.1569e-04, -5.1379e-05, ..., 1.7786e-03, + 3.5954e-04, 6.8235e-04]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0310, 0.0416, -0.0007, -0.0113, 0.0213, -0.0340, 0.0050, 0.0067, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([-0.0361, 0.0136, 0.0181, -0.0140, 0.0085, -0.0202, -0.0037, 0.0126, + 0.0129, 0.0084], device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 214.74, cls_loss 0.4458 cls_loss_mapping 0.0007 cls_loss_causal 0.3882 re_mapping 0.0059 re_causal 0.0158 /// teacc 99.01 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.0735, -0.1637, -0.1115, ..., -0.0380, 0.0666, -0.1131], + [-0.0833, -0.0963, -0.0903, ..., 0.0998, -0.0394, 0.2411], + [ 0.0018, -0.0318, -0.0631, ..., -0.0069, -0.0107, -0.0822], + ..., + [ 0.0384, -0.1221, 0.1836, ..., 0.0398, -0.1268, 0.0623], + [-0.0265, 0.0664, -0.1543, ..., -0.0497, 0.0196, -0.1211], + [-0.0650, 0.0840, 0.0435, ..., -0.0649, -0.0560, -0.0366]], + device='cuda:0'), grad: tensor([[-1.1241e-04, 6.3241e-05, 1.1528e-04, ..., 1.0004e-03, + 2.3854e-04, 9.8720e-06], + [ 3.2768e-03, 3.6311e-04, 2.4090e-03, ..., 4.9706e-03, + -2.4395e-03, 2.2030e-03], + [-3.5973e-03, -8.0185e-03, 3.3712e-04, ..., 6.1941e-04, + 3.8862e-04, 1.6522e-04], + ..., + [-2.8439e-03, 2.7990e-04, -2.2335e-03, ..., -7.8659e-03, + 8.7500e-05, -2.5539e-03], + [ 5.1975e-04, 6.9141e-04, 2.9540e-04, ..., -1.1101e-03, + 3.0231e-04, 1.8254e-05], + [-3.7117e-03, 1.7252e-03, -3.4027e-03, ..., -9.3384e-03, + -3.3045e-04, -5.3329e-03]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0311, 0.0415, -0.0006, -0.0114, 0.0212, -0.0341, 0.0050, 0.0069, + -0.0024, -0.0147], device='cuda:0'), grad: tensor([ 0.0061, -0.0057, -0.0138, 0.0073, 0.0358, -0.0006, 0.0100, -0.0009, + -0.0212, -0.0170], device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 214.83, cls_loss 0.5111 cls_loss_mapping 0.0009 cls_loss_causal 0.4448 re_mapping 0.0060 re_causal 0.0166 /// teacc 99.01 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.0735, -0.1636, -0.1115, ..., -0.0381, 0.0667, -0.1132], + [-0.0834, -0.0964, -0.0903, ..., 0.0999, -0.0393, 0.2411], + [ 0.0018, -0.0318, -0.0632, ..., -0.0069, -0.0107, -0.0823], + ..., + [ 0.0387, -0.1222, 0.1835, ..., 0.0398, -0.1268, 0.0622], + [-0.0263, 0.0665, -0.1544, ..., -0.0496, 0.0196, -0.1209], + [-0.0653, 0.0841, 0.0434, ..., -0.0651, -0.0561, -0.0366]], + device='cuda:0'), grad: tensor([[ 1.6857e-07, -4.0889e-05, 2.8778e-07, ..., 2.4395e-03, + -9.0170e-04, 7.9870e-04], + [ 1.1083e-07, -1.3285e-03, 1.6894e-06, ..., -1.1696e-02, + 2.7008e-08, 1.4257e-04], + [-4.3865e-07, 2.2399e-04, 2.1085e-05, ..., -3.5229e-03, + 1.7598e-05, -6.5079e-03], + ..., + [ 3.0268e-07, 2.5916e-04, 3.4589e-06, ..., 2.4948e-03, + 2.6803e-06, 7.3290e-04], + [ 5.2806e-07, 3.0875e-04, 3.4899e-05, ..., 2.9469e-03, + 6.7770e-05, 1.0080e-03], + [ 2.0582e-06, 4.2343e-04, 8.7917e-07, ..., 2.6302e-03, + 5.6076e-04, 9.1171e-04]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0311, 0.0416, -0.0006, -0.0115, 0.0211, -0.0340, 0.0049, 0.0069, + -0.0023, -0.0147], device='cuda:0'), grad: tensor([ 0.0137, -0.0385, -0.0142, -0.0372, 0.0161, 0.0148, 0.0180, 0.0199, + 0.0191, -0.0117], device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 214.69, cls_loss 0.4722 cls_loss_mapping 0.0007 cls_loss_causal 0.4119 re_mapping 0.0061 re_causal 0.0166 /// teacc 99.01 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.0736, -0.1637, -0.1117, ..., -0.0382, 0.0665, -0.1132], + [-0.0834, -0.0965, -0.0902, ..., 0.0998, -0.0393, 0.2411], + [ 0.0020, -0.0318, -0.0631, ..., -0.0067, -0.0106, -0.0823], + ..., + [ 0.0386, -0.1221, 0.1836, ..., 0.0399, -0.1267, 0.0622], + [-0.0264, 0.0664, -0.1545, ..., -0.0497, 0.0195, -0.1209], + [-0.0652, 0.0840, 0.0433, ..., -0.0652, -0.0563, -0.0369]], + device='cuda:0'), grad: tensor([[ 6.7186e-04, 1.6665e-04, 2.2244e-04, ..., 1.4048e-03, + -1.3006e-04, 6.0469e-05], + [ 5.6314e-04, -6.2418e-04, 2.6393e-04, ..., -1.4877e-03, + 9.6709e-06, 2.5415e-04], + [ 1.4668e-03, 1.8406e-04, 9.9754e-04, ..., 2.4166e-03, + 6.2168e-05, 1.8060e-04], + ..., + [-3.8534e-05, 6.4516e-04, 2.0623e-04, ..., 1.6031e-03, + 2.3637e-06, 1.1139e-03], + [ 8.1015e-04, 3.4189e-04, 1.9920e-04, ..., 1.6184e-03, + 1.3924e-04, 3.7408e-04], + [ 1.0538e-03, -4.9019e-04, 2.4283e-04, ..., -3.5133e-03, + 3.8058e-05, 7.7581e-04]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0312, 0.0416, -0.0004, -0.0115, 0.0212, -0.0339, 0.0049, 0.0071, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0154, -0.0176, 0.0279, -0.0142, -0.0084, 0.0229, 0.0005, -0.0069, + 0.0152, -0.0347], device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 214.73, cls_loss 0.4974 cls_loss_mapping 0.0008 cls_loss_causal 0.4407 re_mapping 0.0059 re_causal 0.0168 /// teacc 99.03 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.0734, -0.1638, -0.1114, ..., -0.0381, 0.0666, -0.1131], + [-0.0835, -0.0965, -0.0902, ..., 0.0996, -0.0393, 0.2413], + [ 0.0020, -0.0317, -0.0631, ..., -0.0067, -0.0105, -0.0823], + ..., + [ 0.0387, -0.1221, 0.1837, ..., 0.0401, -0.1268, 0.0621], + [-0.0265, 0.0664, -0.1547, ..., -0.0498, 0.0196, -0.1210], + [-0.0653, 0.0840, 0.0432, ..., -0.0652, -0.0564, -0.0370]], + device='cuda:0'), grad: tensor([[ 3.0708e-03, 5.5170e-04, 1.2636e-04, ..., 1.0887e-02, + 4.3559e-04, 3.6983e-03], + [-4.0483e-04, -6.2323e-04, -1.7471e-03, ..., 3.9148e-04, + 3.4630e-05, 1.0929e-03], + [ 3.1872e-03, 1.3971e-03, 9.7752e-04, ..., -3.5954e-03, + 2.8157e-04, -5.9166e-03], + ..., + [-1.4143e-03, -4.5919e-04, 8.2588e-04, ..., 3.2806e-03, + 4.5002e-05, 2.0447e-03], + [ 1.3971e-03, -2.4843e-04, -1.6630e-05, ..., 3.5286e-03, + 2.6107e-04, 1.5295e-04], + [ 1.3390e-03, 1.1616e-03, 4.6206e-04, ..., 4.4975e-03, + 8.2552e-05, 6.2704e-04]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0312, 0.0414, -0.0004, -0.0116, 0.0212, -0.0338, 0.0050, 0.0072, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0424, 0.0125, -0.0009, -0.0082, -0.0480, -0.0145, -0.0280, 0.0025, + 0.0212, 0.0211], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 494---------------------------------------------------- +epoch 494, time 231.02, cls_loss 0.4693 cls_loss_mapping 0.0007 cls_loss_causal 0.4061 re_mapping 0.0059 re_causal 0.0164 /// teacc 99.06 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.0735, -0.1639, -0.1114, ..., -0.0384, 0.0666, -0.1132], + [-0.0834, -0.0964, -0.0900, ..., 0.0997, -0.0393, 0.2414], + [ 0.0019, -0.0317, -0.0631, ..., -0.0066, -0.0105, -0.0823], + ..., + [ 0.0387, -0.1220, 0.1835, ..., 0.0401, -0.1270, 0.0620], + [-0.0267, 0.0663, -0.1549, ..., -0.0498, 0.0194, -0.1210], + [-0.0654, 0.0840, 0.0433, ..., -0.0653, -0.0564, -0.0370]], + device='cuda:0'), grad: tensor([[ 4.1270e-04, -1.2598e-03, 4.0293e-05, ..., -3.0708e-03, + 3.6120e-04, 1.9035e-03], + [ 6.8808e-04, 9.3079e-04, 2.0310e-05, ..., 4.2038e-03, + 7.4577e-04, -8.3466e-03], + [ 1.5366e-04, 1.2267e-04, 2.2307e-05, ..., -2.1305e-03, + 2.4056e-04, -7.8249e-04], + ..., + [ 2.4796e-04, 5.3978e-04, 1.1474e-05, ..., 3.0270e-03, + 4.0102e-04, 2.4939e-04], + [ 7.3552e-05, -1.9283e-03, 1.1921e-04, ..., 2.9063e-04, + 2.8610e-04, 2.1744e-03], + [-1.0357e-03, 6.0463e-04, -1.7204e-03, ..., -5.7526e-03, + 4.2534e-04, -6.3820e-03]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0313, 0.0415, -0.0004, -0.0115, 0.0212, -0.0338, 0.0050, 0.0071, + -0.0026, -0.0148], device='cuda:0'), grad: tensor([-0.0184, 0.0168, 0.0070, 0.0196, -0.0113, -0.0144, 0.0221, -0.0108, + -0.0176, 0.0070], device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 214.52, cls_loss 0.4661 cls_loss_mapping 0.0008 cls_loss_causal 0.4009 re_mapping 0.0059 re_causal 0.0159 /// teacc 99.05 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.0733, -0.1638, -0.1113, ..., -0.0383, 0.0666, -0.1132], + [-0.0832, -0.0965, -0.0900, ..., 0.0997, -0.0392, 0.2415], + [ 0.0019, -0.0315, -0.0630, ..., -0.0064, -0.0104, -0.0824], + ..., + [ 0.0389, -0.1218, 0.1837, ..., 0.0403, -0.1270, 0.0622], + [-0.0269, 0.0662, -0.1549, ..., -0.0500, 0.0194, -0.1211], + [-0.0656, 0.0840, 0.0432, ..., -0.0654, -0.0565, -0.0371]], + device='cuda:0'), grad: tensor([[-3.9011e-05, -1.0669e-04, 3.1088e-06, ..., 2.6178e-04, + 9.0431e-07, 2.2262e-05], + [ 3.1650e-05, 2.5749e-04, 1.2465e-05, ..., 9.7370e-04, + 1.3588e-06, 1.4286e-03], + [ 5.3674e-05, 1.5819e-04, 1.6063e-05, ..., 1.2989e-03, + 1.3318e-07, 1.2887e-04], + ..., + [-1.7869e-04, -4.6425e-03, -1.3864e-04, ..., -7.5684e-03, + 1.9409e-06, -3.2532e-02], + [ 1.8597e-05, 1.7900e-03, 2.2247e-05, ..., 8.1158e-04, + 2.8778e-07, 1.3275e-02], + [ 2.3693e-05, 7.7057e-04, 2.3946e-05, ..., 8.5974e-04, + 4.4703e-05, 4.1809e-03]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0312, 0.0416, -0.0003, -0.0115, 0.0212, -0.0339, 0.0048, 0.0072, + -0.0027, -0.0149], device='cuda:0'), grad: tensor([ 0.0032, 0.0084, 0.0080, 0.0089, 0.0126, 0.0043, 0.0056, -0.0746, + 0.0146, 0.0090], device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 215.24, cls_loss 0.4860 cls_loss_mapping 0.0009 cls_loss_causal 0.4350 re_mapping 0.0059 re_causal 0.0163 /// teacc 99.06 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.0733, -0.1639, -0.1114, ..., -0.0383, 0.0667, -0.1132], + [-0.0832, -0.0965, -0.0901, ..., 0.0996, -0.0392, 0.2413], + [ 0.0019, -0.0315, -0.0629, ..., -0.0063, -0.0106, -0.0824], + ..., + [ 0.0389, -0.1218, 0.1838, ..., 0.0403, -0.1268, 0.0624], + [-0.0269, 0.0660, -0.1550, ..., -0.0500, 0.0194, -0.1211], + [-0.0656, 0.0841, 0.0432, ..., -0.0654, -0.0564, -0.0371]], + device='cuda:0'), grad: tensor([[ 8.3618e-03, 1.1396e-03, 1.2612e-04, ..., 8.2779e-03, + -2.7940e-09, 1.4677e-03], + [ 6.8808e-04, 5.5408e-04, 6.3562e-04, ..., 2.2774e-03, + 0.0000e+00, 5.3215e-04], + [ 8.6308e-04, 6.5279e-04, 7.3099e-04, ..., 2.5024e-03, + 7.1712e-08, 7.9727e-04], + ..., + [ 8.8406e-04, 5.5265e-04, 1.1282e-03, ..., 2.9583e-03, + 0.0000e+00, 2.9039e-04], + [-3.5114e-03, -1.6623e-03, -4.6844e-03, ..., -9.8038e-03, + 5.5879e-09, -2.5711e-03], + [ 4.2534e-04, -1.8060e-04, 4.8470e-04, ..., -1.1635e-03, + 1.8626e-09, 1.5986e-04]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0313, 0.0414, -0.0004, -0.0115, 0.0213, -0.0339, 0.0049, 0.0072, + -0.0027, -0.0148], device='cuda:0'), grad: tensor([ 0.0327, 0.0132, 0.0122, 0.0109, -0.0185, 0.0346, -0.0185, 0.0165, + -0.0631, -0.0201], device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 214.87, cls_loss 0.4995 cls_loss_mapping 0.0010 cls_loss_causal 0.4303 re_mapping 0.0059 re_causal 0.0159 /// teacc 99.05 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.0734, -0.1639, -0.1115, ..., -0.0383, 0.0668, -0.1134], + [-0.0831, -0.0965, -0.0901, ..., 0.0996, -0.0392, 0.2415], + [ 0.0018, -0.0316, -0.0629, ..., -0.0064, -0.0107, -0.0824], + ..., + [ 0.0389, -0.1219, 0.1838, ..., 0.0403, -0.1268, 0.0624], + [-0.0268, 0.0659, -0.1548, ..., -0.0498, 0.0195, -0.1210], + [-0.0657, 0.0842, 0.0431, ..., -0.0655, -0.0564, -0.0372]], + device='cuda:0'), grad: tensor([[ 2.0526e-06, 1.1599e-04, 3.9674e-06, ..., 8.0156e-04, + 1.7323e-06, 1.2457e-04], + [ 2.1718e-06, 1.2767e-04, 4.7646e-06, ..., 6.6090e-04, + 4.9882e-06, 4.7326e-04], + [ 1.5050e-05, -1.7071e-04, 2.1607e-05, ..., -1.9150e-03, + 3.6713e-06, 2.8563e-04], + ..., + [-2.6107e-05, 1.5056e-04, -1.7658e-06, ..., 9.3079e-04, + 1.7136e-07, 3.2177e-03], + [ 1.1568e-03, 4.6730e-03, 1.8492e-03, ..., 2.7008e-03, + 1.3486e-05, 1.2293e-03], + [ 3.5942e-05, 2.2948e-04, 6.0320e-05, ..., 7.8106e-04, + 2.6822e-07, -6.7596e-03]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0314, 0.0415, -0.0004, -0.0115, 0.0213, -0.0339, 0.0048, 0.0071, + -0.0025, -0.0148], device='cuda:0'), grad: tensor([ 0.0097, 0.0105, -0.0022, -0.0169, -0.0512, -0.0065, 0.0089, 0.0208, + 0.0300, -0.0032], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 498---------------------------------------------------- +epoch 498, time 230.80, cls_loss 0.4923 cls_loss_mapping 0.0010 cls_loss_causal 0.4357 re_mapping 0.0058 re_causal 0.0159 /// teacc 99.08 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.0734, -0.1639, -0.1115, ..., -0.0385, 0.0668, -0.1135], + [-0.0831, -0.0966, -0.0902, ..., 0.0995, -0.0393, 0.2414], + [ 0.0020, -0.0316, -0.0629, ..., -0.0063, -0.0107, -0.0824], + ..., + [ 0.0387, -0.1219, 0.1838, ..., 0.0402, -0.1268, 0.0623], + [-0.0267, 0.0662, -0.1548, ..., -0.0497, 0.0195, -0.1210], + [-0.0658, 0.0839, 0.0430, ..., -0.0656, -0.0565, -0.0373]], + device='cuda:0'), grad: tensor([[ 4.6196e-03, 2.1381e-03, 4.7266e-05, ..., 2.9850e-04, + 2.4567e-03, -4.0245e-04], + [ 1.4277e-03, -2.5868e-05, 3.7044e-05, ..., 1.8721e-03, + 5.4264e-04, 2.8030e-02], + [-3.7270e-03, -1.0548e-03, 1.8179e-04, ..., 7.0906e-04, + -3.5362e-03, 1.0023e-03], + ..., + [ 4.8637e-04, 1.2010e-05, -6.0129e-04, ..., 4.3917e-04, + 5.8317e-04, 1.5450e-04], + [-5.5695e-03, 6.5536e-03, 3.1519e-04, ..., -1.4067e-03, + 7.5150e-04, -1.1492e-03], + [ 7.2289e-04, 4.7743e-05, 1.2660e-04, ..., -9.4128e-04, + 5.6744e-04, 3.6168e-04]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0315, 0.0414, -0.0003, -0.0116, 0.0216, -0.0339, 0.0049, 0.0072, + -0.0025, -0.0150], device='cuda:0'), grad: tensor([-0.0128, 0.0381, -0.0125, 0.0065, -0.0036, -0.0352, 0.0231, 0.0067, + 0.0058, -0.0160], device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 214.32, cls_loss 0.4767 cls_loss_mapping 0.0009 cls_loss_causal 0.4114 re_mapping 0.0058 re_causal 0.0154 /// teacc 99.07 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0298, 0.0183, 0.0140, ..., -0.0305, -0.0006, 0.0097], + [ 0.0207, 0.0108, -0.0263, ..., -0.0054, -0.0157, -0.0142], + [ 0.0241, -0.0026, 0.0090, ..., 0.0294, -0.0064, -0.0062], + ..., + [ 0.0203, -0.0303, 0.0013, ..., 0.0076, -0.0020, 0.0291], + [ 0.0259, -0.0132, -0.0208, ..., -0.0018, 0.0170, 0.0220], + [-0.0296, -0.0233, 0.0248, ..., 0.0041, 0.0078, -0.0206]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0219, -0.0165, -0.0295, 0.0027, 0.0002, 0.0287, 0.0149, -0.0226, + 0.0137, -0.0017], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 231.12, cls_loss 1.3557 cls_loss_mapping 1.8773 cls_loss_causal 2.2124 re_mapping 0.1405 re_causal 0.1486 /// teacc 85.36 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0261, 0.0146, 0.0117, ..., -0.0395, -0.0012, 0.0054], + [ 0.0187, 0.0061, -0.0362, ..., -0.0016, -0.0163, -0.0201], + [ 0.0299, -0.0099, 0.0060, ..., 0.0290, -0.0070, -0.0025], + ..., + [ 0.0142, -0.0289, -0.0028, ..., 0.0013, -0.0014, 0.0339], + [ 0.0233, -0.0159, -0.0253, ..., 0.0008, 0.0164, 0.0185], + [-0.0380, -0.0188, 0.0301, ..., 0.0066, 0.0072, -0.0158]], + device='cuda:0'), grad: tensor([[-0.0638, -0.0174, -0.0414, ..., -0.0077, 0.0000, -0.0010], + [-0.0094, 0.0015, 0.0069, ..., -0.0097, 0.0000, 0.0014], + [ 0.0204, 0.0081, 0.0418, ..., 0.0165, 0.0000, 0.0032], + ..., + [ 0.0043, -0.0229, -0.0016, ..., -0.0025, 0.0000, -0.0250], + [-0.0118, 0.0077, 0.0156, ..., 0.0009, 0.0000, 0.0015], + [ 0.0181, 0.0073, -0.0224, ..., -0.0145, 0.0000, 0.0116]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0234, -0.0160, -0.0298, 0.0025, -0.0013, 0.0293, 0.0163, -0.0233, + 0.0133, -0.0012], device='cuda:0'), grad: tensor([-0.0615, -0.0135, 0.0319, 0.0369, 0.0296, -0.0230, 0.0191, -0.0185, + -0.0079, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 229.55, cls_loss 0.4405 cls_loss_mapping 0.8122 cls_loss_causal 1.9444 re_mapping 0.2013 re_causal 0.2671 /// teacc 91.82 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0239, 0.0132, 0.0099, ..., -0.0434, -0.0012, 0.0033], + [ 0.0189, 0.0043, -0.0411, ..., -0.0020, -0.0163, -0.0240], + [ 0.0320, -0.0129, 0.0052, ..., 0.0293, -0.0070, -0.0014], + ..., + [ 0.0124, -0.0285, -0.0051, ..., -0.0017, -0.0014, 0.0374], + [ 0.0227, -0.0193, -0.0271, ..., -0.0005, 0.0164, 0.0152], + [-0.0425, -0.0181, 0.0322, ..., 0.0075, 0.0072, -0.0140]], + device='cuda:0'), grad: tensor([[ 0.0221, 0.0141, 0.0029, ..., 0.0018, 0.0000, 0.0072], + [ 0.0323, 0.0042, 0.0040, ..., 0.0042, 0.0000, 0.0014], + [-0.0214, 0.0074, -0.0037, ..., -0.0054, 0.0000, -0.0093], + ..., + [ 0.0075, 0.0217, 0.0161, ..., 0.0089, 0.0000, 0.0104], + [-0.0015, 0.0184, 0.0100, ..., 0.0114, 0.0000, 0.0055], + [ 0.0095, -0.0368, -0.0193, ..., -0.0115, 0.0000, -0.0136]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0234, -0.0166, -0.0296, 0.0026, -0.0012, 0.0303, 0.0163, -0.0239, + 0.0132, -0.0011], device='cuda:0'), grad: tensor([ 0.0244, 0.0311, -0.0232, -0.0091, 0.0268, -0.0454, -0.0161, 0.0274, + 0.0113, -0.0271], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 230.25, cls_loss 0.2939 cls_loss_mapping 0.4845 cls_loss_causal 1.7351 re_mapping 0.1476 re_causal 0.2378 /// teacc 93.83 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0227, 0.0125, 0.0080, ..., -0.0451, -0.0049, 0.0026], + [ 0.0192, 0.0024, -0.0440, ..., -0.0019, -0.0213, -0.0262], + [ 0.0324, -0.0149, 0.0040, ..., 0.0288, -0.0193, -0.0004], + ..., + [ 0.0105, -0.0291, -0.0068, ..., -0.0032, -0.0007, 0.0388], + [ 0.0242, -0.0217, -0.0272, ..., -0.0016, 0.0086, 0.0134], + [-0.0463, -0.0165, 0.0337, ..., 0.0084, 0.0020, -0.0131]], + device='cuda:0'), grad: tensor([[-1.2619e-02, -1.0042e-03, -1.2321e-03, ..., 1.0138e-03, + 1.1414e-04, 1.4186e-04], + [-1.0429e-02, 8.6069e-04, 1.0233e-03, ..., 3.0971e-04, + 1.5545e-04, 1.4687e-04], + [ 5.7869e-03, 8.2254e-04, 1.1120e-03, ..., 1.0223e-03, + 7.3552e-05, 9.5901e-03], + ..., + [ 2.6722e-03, 1.5764e-03, 6.5079e-03, ..., 3.7060e-03, + 7.7128e-05, -2.9993e-04], + [-1.7023e-03, 3.3112e-03, 7.1411e-03, ..., 5.9052e-03, + 5.0497e-04, -1.4984e-02], + [ 3.5820e-03, -1.8415e-03, -5.1842e-03, ..., -1.1663e-03, + 1.5199e-04, 1.8768e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0234, -0.0165, -0.0296, 0.0025, -0.0012, 0.0309, 0.0162, -0.0248, + 0.0133, -0.0009], device='cuda:0'), grad: tensor([-0.0110, -0.0221, 0.0187, 0.0164, -0.0054, -0.0089, 0.0042, 0.0079, + -0.0017, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 229.98, cls_loss 0.2224 cls_loss_mapping 0.3528 cls_loss_causal 1.5172 re_mapping 0.1235 re_causal 0.2194 /// teacc 95.76 lr 0.00010000 +Epoch 5, weight, value: tensor([[-2.1751e-02, 1.2036e-02, 6.7400e-03, ..., -4.6559e-02, + -6.1361e-03, 2.1058e-03], + [ 1.9783e-02, -1.0993e-04, -4.6375e-02, ..., -2.2998e-03, + -2.1551e-02, -2.8241e-02], + [ 3.2679e-02, -1.6947e-02, 3.4368e-03, ..., 2.8977e-02, + -2.2000e-02, -8.3496e-05], + ..., + [ 1.0788e-02, -2.9527e-02, -8.5121e-03, ..., -4.8042e-03, + -8.0969e-04, 4.0829e-02], + [ 2.4870e-02, -2.3275e-02, -2.7209e-02, ..., -2.2898e-03, + 6.7635e-03, 1.1713e-02], + [-4.9579e-02, -1.5481e-02, 3.5391e-02, ..., 9.6285e-03, + 1.6631e-03, -1.3594e-02]], device='cuda:0'), grad: tensor([[ 0.0008, 0.0016, 0.0005, ..., 0.0012, 0.0000, 0.0002], + [ 0.0009, 0.0022, 0.0016, ..., 0.0002, 0.0000, 0.0005], + [ 0.0022, 0.0021, 0.0010, ..., 0.0014, 0.0000, 0.0015], + ..., + [-0.0014, -0.0178, 0.0004, ..., 0.0016, 0.0000, -0.0096], + [-0.0079, 0.0035, 0.0039, ..., 0.0047, 0.0000, 0.0008], + [ 0.0016, -0.0167, -0.0118, ..., -0.0067, 0.0000, -0.0010]], + device='cuda:0') +Epoch 5, bias, value: tensor([-0.0234, -0.0165, -0.0292, 0.0023, -0.0013, 0.0308, 0.0163, -0.0247, + 0.0131, -0.0008], device='cuda:0'), grad: tensor([ 0.0023, 0.0007, 0.0045, 0.0472, -0.0135, 0.0151, -0.0314, -0.0096, + -0.0012, -0.0140], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 229.39, cls_loss 0.1693 cls_loss_mapping 0.2619 cls_loss_causal 1.3809 re_mapping 0.1012 re_causal 0.1942 /// teacc 96.28 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0203, 0.0108, 0.0056, ..., -0.0474, -0.0075, 0.0013], + [ 0.0199, -0.0015, -0.0480, ..., -0.0026, -0.0207, -0.0296], + [ 0.0326, -0.0182, 0.0025, ..., 0.0289, -0.0240, 0.0007], + ..., + [ 0.0103, -0.0300, -0.0105, ..., -0.0061, -0.0010, 0.0420], + [ 0.0258, -0.0248, -0.0278, ..., -0.0031, 0.0045, 0.0107], + [-0.0528, -0.0146, 0.0371, ..., 0.0106, 0.0013, -0.0143]], + device='cuda:0'), grad: tensor([[-9.3079e-03, -4.4674e-05, 2.3627e-04, ..., 1.3292e-04, + 2.8357e-05, -9.1362e-04], + [ 9.8648e-03, 3.4161e-03, 5.0116e-04, ..., 2.8563e-04, + 6.2548e-06, 5.2643e-04], + [ 6.6681e-03, 2.3823e-03, 9.9087e-04, ..., 6.3944e-04, + 3.4943e-06, 4.5166e-03], + ..., + [-1.5869e-03, 2.6970e-03, 5.4932e-03, ..., 2.9278e-03, + 2.0228e-06, -7.3051e-03], + [-2.8172e-03, 1.6413e-03, -5.4932e-04, ..., -9.3877e-05, + 2.7657e-04, 4.8399e-04], + [ 2.7962e-03, -6.1073e-03, -1.4839e-02, ..., -1.0231e-02, + 2.0545e-06, -9.7370e-04]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0233, -0.0166, -0.0294, 0.0028, -0.0009, 0.0305, 0.0161, -0.0249, + 0.0132, -0.0009], device='cuda:0'), grad: tensor([-9.0485e-03, 1.2978e-02, 1.3687e-02, -1.6129e-02, 1.1322e-02, + 1.0696e-02, -1.3939e-02, 7.9453e-05, -2.0046e-03, -7.6408e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 229.26, cls_loss 0.1371 cls_loss_mapping 0.2095 cls_loss_causal 1.3417 re_mapping 0.0842 re_causal 0.1816 /// teacc 96.33 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0190, 0.0101, 0.0050, ..., -0.0484, -0.0093, 0.0006], + [ 0.0194, -0.0033, -0.0493, ..., -0.0029, -0.0233, -0.0311], + [ 0.0326, -0.0189, 0.0018, ..., 0.0290, -0.0262, 0.0016], + ..., + [ 0.0110, -0.0306, -0.0121, ..., -0.0072, -0.0007, 0.0428], + [ 0.0269, -0.0258, -0.0279, ..., -0.0039, 0.0031, 0.0100], + [-0.0550, -0.0139, 0.0380, ..., 0.0115, 0.0007, -0.0148]], + device='cuda:0'), grad: tensor([[-1.3304e-03, 4.1080e-04, 2.6417e-04, ..., 1.5914e-04, + 1.3542e-04, 4.3464e-04], + [ 2.2388e-04, 9.0933e-04, 2.8872e-04, ..., 3.0088e-04, + 2.3097e-07, 8.1825e-04], + [-1.2674e-03, 1.2207e-03, 4.2844e-04, ..., 5.4169e-04, + 9.9912e-06, -2.3842e-03], + ..., + [ 9.9754e-04, 6.2561e-04, 1.2989e-03, ..., 1.0405e-03, + 8.1420e-05, 7.0524e-04], + [ 1.0895e-02, 3.4733e-03, 3.8757e-03, ..., 6.2599e-03, + 3.4165e-04, 8.8978e-04], + [-4.1223e-04, -5.8517e-03, -9.8419e-03, ..., -8.1406e-03, + 1.7971e-05, 6.8283e-04]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0230, -0.0171, -0.0292, 0.0029, -0.0010, 0.0302, 0.0159, -0.0250, + 0.0135, -0.0010], device='cuda:0'), grad: tensor([-0.0002, 0.0012, -0.0019, -0.0040, 0.0019, -0.0210, 0.0172, 0.0026, + 0.0118, -0.0076], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 227.95, cls_loss 0.1238 cls_loss_mapping 0.1801 cls_loss_causal 1.2398 re_mapping 0.0719 re_causal 0.1588 /// teacc 97.04 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0177, 0.0093, 0.0043, ..., -0.0494, -0.0089, -0.0008], + [ 0.0193, -0.0044, -0.0499, ..., -0.0027, -0.0240, -0.0321], + [ 0.0330, -0.0208, 0.0012, ..., 0.0293, -0.0275, 0.0027], + ..., + [ 0.0112, -0.0304, -0.0132, ..., -0.0080, -0.0012, 0.0438], + [ 0.0273, -0.0270, -0.0282, ..., -0.0047, 0.0028, 0.0090], + [-0.0575, -0.0134, 0.0390, ..., 0.0124, 0.0003, -0.0157]], + device='cuda:0'), grad: tensor([[-9.5139e-03, -8.9598e-04, 5.3024e-04, ..., 2.4915e-04, + 6.3062e-05, 3.6597e-04], + [ 3.6907e-03, 1.6050e-03, 5.9843e-04, ..., 2.2495e-04, + 3.6061e-05, 1.1377e-03], + [ 1.4582e-03, 1.4229e-03, 1.4877e-03, ..., 3.6836e-04, + 2.1368e-05, -8.6641e-04], + ..., + [-1.0841e-02, -3.9520e-03, -1.6336e-03, ..., 5.4538e-05, + 2.0236e-05, -4.8332e-03], + [ 2.5673e-03, 1.9207e-03, 1.4076e-03, ..., 8.5974e-04, + 1.6618e-04, 7.6962e-04], + [ 3.9482e-03, -2.0203e-02, -3.1982e-02, ..., -5.3635e-03, + 3.4958e-05, 7.7295e-04]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0228, -0.0169, -0.0289, 0.0029, -0.0011, 0.0300, 0.0156, -0.0250, + 0.0136, -0.0011], device='cuda:0'), grad: tensor([-0.0052, 0.0053, 0.0026, 0.0071, 0.0268, 0.0028, -0.0016, -0.0183, + 0.0023, -0.0217], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 226.51, cls_loss 0.1205 cls_loss_mapping 0.1655 cls_loss_causal 1.2069 re_mapping 0.0650 re_causal 0.1466 /// teacc 97.29 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0169, 0.0085, 0.0050, ..., -0.0497, -0.0097, -0.0020], + [ 0.0195, -0.0052, -0.0508, ..., -0.0029, -0.0236, -0.0329], + [ 0.0326, -0.0226, 0.0002, ..., 0.0293, -0.0283, 0.0030], + ..., + [ 0.0115, -0.0310, -0.0144, ..., -0.0088, -0.0016, 0.0447], + [ 0.0279, -0.0281, -0.0284, ..., -0.0059, 0.0019, 0.0083], + [-0.0588, -0.0129, 0.0396, ..., 0.0134, 0.0002, -0.0160]], + device='cuda:0'), grad: tensor([[-1.2993e-02, -7.6675e-04, 2.0587e-04, ..., -3.3054e-03, + 7.5512e-06, 2.2948e-04], + [ 9.7427e-03, 2.9259e-03, 1.0710e-03, ..., 3.1300e-03, + 6.7614e-07, 3.3331e-04], + [-9.1476e-03, 8.1635e-04, 6.5565e-04, ..., 1.2550e-03, + 8.9034e-06, 9.5749e-04], + ..., + [ 1.1358e-03, -7.6828e-03, -4.4370e-04, ..., -6.7024e-03, + 2.3954e-06, -1.4977e-02], + [-3.4275e-03, -1.9817e-03, 3.3073e-03, ..., -1.5068e-03, + 2.3860e-06, 5.8699e-04], + [ 2.4643e-03, 5.6915e-03, 2.5845e-03, ..., 4.0779e-03, + 6.2259e-07, 5.4588e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0227, -0.0169, -0.0294, 0.0026, -0.0008, 0.0302, 0.0153, -0.0247, + 0.0137, -0.0012], device='cuda:0'), grad: tensor([-0.0173, 0.0191, -0.0114, 0.0114, -0.0068, 0.0127, 0.0030, -0.0151, + -0.0110, 0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 8, time 214.06, cls_loss 0.0877 cls_loss_mapping 0.1275 cls_loss_causal 1.1551 re_mapping 0.0587 re_causal 0.1401 /// teacc 97.24 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0163, 0.0073, 0.0044, ..., -0.0507, -0.0098, -0.0030], + [ 0.0193, -0.0057, -0.0516, ..., -0.0034, -0.0237, -0.0343], + [ 0.0330, -0.0234, -0.0004, ..., 0.0296, -0.0285, 0.0042], + ..., + [ 0.0113, -0.0314, -0.0156, ..., -0.0094, -0.0016, 0.0449], + [ 0.0287, -0.0288, -0.0290, ..., -0.0070, 0.0019, 0.0072], + [-0.0600, -0.0120, 0.0406, ..., 0.0150, 0.0001, -0.0165]], + device='cuda:0'), grad: tensor([[-8.5449e-04, 4.1962e-04, 1.3614e-04, ..., 1.9026e-04, + 5.7276e-07, 1.0288e-04], + [ 1.6403e-02, 2.2850e-03, 2.9016e-04, ..., 1.6460e-03, + 2.0722e-07, 3.2163e-04], + [-2.2717e-03, 3.5706e-03, 3.5787e-04, ..., 6.5804e-04, + 1.2433e-07, 7.1764e-05], + ..., + [-7.1907e-04, 2.7442e-04, 5.1832e-04, ..., 5.8889e-04, + 2.6124e-07, -9.9182e-04], + [-1.4778e-02, 1.6365e-03, 8.8120e-04, ..., 3.4308e-04, + 9.5367e-06, 6.2895e-04], + [ 1.8778e-03, 2.4490e-03, 3.7270e-03, ..., 4.3182e-03, + 4.8382e-07, 5.4979e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0227, -0.0172, -0.0287, 0.0032, -0.0011, 0.0295, 0.0154, -0.0251, + 0.0138, -0.0009], device='cuda:0'), grad: tensor([ 0.0002, 0.0244, 0.0056, -0.0192, 0.0008, -0.0053, 0.0047, 0.0006, + -0.0174, 0.0056], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 228.28, cls_loss 0.0827 cls_loss_mapping 0.1256 cls_loss_causal 1.1255 re_mapping 0.0537 re_causal 0.1329 /// teacc 97.62 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0160, 0.0066, 0.0041, ..., -0.0514, -0.0099, -0.0038], + [ 0.0189, -0.0069, -0.0531, ..., -0.0035, -0.0234, -0.0354], + [ 0.0332, -0.0244, -0.0014, ..., 0.0294, -0.0294, 0.0051], + ..., + [ 0.0118, -0.0316, -0.0166, ..., -0.0104, -0.0015, 0.0458], + [ 0.0292, -0.0294, -0.0292, ..., -0.0075, 0.0016, 0.0064], + [-0.0617, -0.0117, 0.0415, ..., 0.0156, -0.0003, -0.0173]], + device='cuda:0'), grad: tensor([[-3.2730e-03, 1.9193e-04, 2.2900e-04, ..., 1.0026e-04, + 9.5516e-06, 1.4171e-05], + [-3.3360e-03, -1.1635e-04, 3.3587e-05, ..., 2.5779e-05, + 4.3213e-06, -5.2154e-05], + [ 1.1377e-03, 4.5681e-04, 6.5851e-04, ..., 2.4700e-04, + 1.1269e-06, -6.7902e-04], + ..., + [ 8.3876e-04, 1.0519e-03, 1.5650e-03, ..., 7.5197e-04, + 4.7460e-06, 5.6362e-04], + [-3.8185e-03, -6.9923e-03, -1.2810e-02, ..., -4.8752e-03, + 4.2617e-05, 5.6475e-05], + [ 1.3056e-03, -2.2542e-04, 5.5075e-04, ..., -1.4186e-04, + 2.9933e-06, -4.8056e-07]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0229, -0.0173, -0.0288, 0.0032, -0.0008, 0.0295, 0.0153, -0.0249, + 0.0140, -0.0012], device='cuda:0'), grad: tensor([-0.0019, -0.0039, 0.0014, 0.0010, 0.0079, 0.0023, 0.0056, 0.0022, + -0.0167, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 228.40, cls_loss 0.0788 cls_loss_mapping 0.1163 cls_loss_causal 1.0973 re_mapping 0.0492 re_causal 0.1238 /// teacc 97.65 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0156, 0.0062, 0.0042, ..., -0.0520, -0.0097, -0.0042], + [ 0.0188, -0.0081, -0.0537, ..., -0.0037, -0.0234, -0.0363], + [ 0.0332, -0.0258, -0.0019, ..., 0.0295, -0.0300, 0.0055], + ..., + [ 0.0127, -0.0316, -0.0173, ..., -0.0108, -0.0015, 0.0463], + [ 0.0296, -0.0301, -0.0295, ..., -0.0079, 0.0015, 0.0058], + [-0.0631, -0.0111, 0.0422, ..., 0.0165, -0.0004, -0.0175]], + device='cuda:0'), grad: tensor([[-4.7517e-04, 4.4298e-04, 4.9114e-04, ..., 9.2328e-05, + 1.7229e-07, -3.3426e-04], + [ 6.4278e-04, 3.6373e-03, 5.4550e-03, ..., 1.7233e-03, + 1.3188e-06, 3.3379e-04], + [ 2.3425e-04, 2.7847e-04, 2.7084e-04, ..., 8.5652e-05, + 5.9651e-07, -5.6648e-04], + ..., + [-3.1543e-04, -4.2963e-04, 1.8911e-03, ..., 1.1129e-03, + 2.9840e-06, -4.9305e-04], + [ 6.6948e-03, 4.0283e-03, 2.6016e-03, ..., 6.8760e-04, + 7.1432e-07, -9.7632e-05], + [-1.5312e-02, -6.5041e-03, -4.0779e-03, ..., -1.9989e-03, + 2.5127e-06, 4.4179e-04]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0226, -0.0176, -0.0289, 0.0032, -0.0009, 0.0296, 0.0150, -0.0248, + 0.0142, -0.0012], device='cuda:0'), grad: tensor([ 0.0010, 0.0083, 0.0003, 0.0048, -0.0098, 0.0090, 0.0021, -0.0002, + 0.0136, -0.0291], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 230.74, cls_loss 0.0682 cls_loss_mapping 0.1035 cls_loss_causal 1.0708 re_mapping 0.0445 re_causal 0.1164 /// teacc 97.93 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0151, 0.0054, 0.0039, ..., -0.0525, -0.0100, -0.0051], + [ 0.0195, -0.0086, -0.0535, ..., -0.0040, -0.0239, -0.0375], + [ 0.0332, -0.0269, -0.0026, ..., 0.0292, -0.0313, 0.0064], + ..., + [ 0.0128, -0.0316, -0.0184, ..., -0.0114, -0.0007, 0.0470], + [ 0.0297, -0.0310, -0.0300, ..., -0.0089, 0.0008, 0.0051], + [-0.0639, -0.0109, 0.0428, ..., 0.0171, -0.0007, -0.0182]], + device='cuda:0'), grad: tensor([[-1.7252e-03, -6.0034e-04, 1.0163e-04, ..., 1.4281e-04, + 7.7114e-06, -3.5620e-04], + [ 1.4377e-04, 4.0960e-04, 1.5879e-04, ..., -4.3124e-05, + 2.2471e-05, 3.4285e-04], + [ 9.3699e-05, 1.6999e-04, 2.0099e-04, ..., 2.4724e-04, + 7.6815e-06, -2.4676e-04], + ..., + [ 7.0572e-04, 1.5469e-03, 2.6016e-03, ..., 1.2617e-03, + 2.1368e-05, 1.2636e-03], + [-5.6076e-04, 5.2452e-04, 2.1005e-04, ..., 8.7690e-04, + 1.1392e-05, 1.2999e-03], + [ 9.9277e-04, -1.4696e-03, -3.0918e-03, ..., -1.6451e-03, + 1.8939e-05, -5.0306e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0226, -0.0171, -0.0290, 0.0033, -0.0010, 0.0292, 0.0151, -0.0247, + 0.0140, -0.0013], device='cuda:0'), grad: tensor([-0.0019, 0.0002, 0.0007, -0.0012, -0.0008, 0.0007, -0.0001, 0.0036, + 0.0002, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 225.36, cls_loss 0.0699 cls_loss_mapping 0.0996 cls_loss_causal 1.0461 re_mapping 0.0432 re_causal 0.1136 /// teacc 98.09 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0146, 0.0052, 0.0044, ..., -0.0521, -0.0086, -0.0057], + [ 0.0189, -0.0098, -0.0539, ..., -0.0041, -0.0250, -0.0391], + [ 0.0330, -0.0285, -0.0030, ..., 0.0284, -0.0379, 0.0071], + ..., + [ 0.0126, -0.0315, -0.0194, ..., -0.0119, 0.0008, 0.0475], + [ 0.0305, -0.0319, -0.0309, ..., -0.0098, 0.0006, 0.0043], + [-0.0645, -0.0106, 0.0436, ..., 0.0179, -0.0025, -0.0188]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0007, 0.0006, ..., 0.0002, 0.0000, 0.0003], + [ 0.0007, 0.0010, 0.0002, ..., 0.0002, 0.0000, 0.0006], + [ 0.0002, 0.0006, 0.0002, ..., -0.0003, 0.0000, -0.0002], + ..., + [ 0.0007, 0.0021, 0.0025, ..., 0.0019, 0.0000, -0.0015], + [-0.0075, -0.0067, -0.0019, ..., -0.0019, 0.0000, -0.0013], + [ 0.0014, -0.0060, -0.0070, ..., -0.0052, 0.0000, -0.0023]], + device='cuda:0') +Epoch 14, bias, value: tensor([-0.0223, -0.0178, -0.0292, 0.0036, -0.0009, 0.0296, 0.0151, -0.0249, + 0.0141, -0.0014], device='cuda:0'), grad: tensor([ 0.0012, 0.0019, 0.0005, 0.0072, 0.0030, 0.0040, 0.0020, 0.0027, + -0.0172, -0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 214.36, cls_loss 0.0602 cls_loss_mapping 0.0899 cls_loss_causal 1.0118 re_mapping 0.0411 re_causal 0.1074 /// teacc 98.00 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0140, 0.0046, 0.0045, ..., -0.0526, -0.0088, -0.0069], + [ 0.0189, -0.0106, -0.0546, ..., -0.0042, -0.0247, -0.0400], + [ 0.0329, -0.0292, -0.0036, ..., 0.0280, -0.0389, 0.0077], + ..., + [ 0.0126, -0.0317, -0.0201, ..., -0.0125, 0.0011, 0.0481], + [ 0.0311, -0.0327, -0.0309, ..., -0.0105, 0.0002, 0.0036], + [-0.0659, -0.0103, 0.0442, ..., 0.0185, -0.0030, -0.0197]], + device='cuda:0'), grad: tensor([[ 5.3644e-04, 1.4699e-04, 2.1636e-04, ..., 6.3300e-05, + 5.8413e-05, 1.7047e-04], + [ 3.4332e-03, 2.4056e-04, 1.4524e-03, ..., 3.8886e-04, + 5.2005e-05, 4.8113e-04], + [ 6.9962e-03, 2.6727e-04, 2.2831e-03, ..., 6.8378e-04, + 2.6673e-05, 1.8883e-03], + ..., + [-1.4429e-03, 2.2244e-04, 7.7820e-04, ..., 3.8671e-04, + -1.4246e-04, -1.8263e-03], + [ 2.2411e-05, 2.3603e-04, 1.4811e-03, ..., 5.6553e-04, + 3.5733e-05, -1.1883e-03], + [ 9.0599e-05, -1.1272e-03, -1.7929e-03, ..., -1.1463e-03, + 3.4332e-05, 6.4671e-05]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0223, -0.0181, -0.0292, 0.0037, -0.0008, 0.0298, 0.0146, -0.0246, + 0.0144, -0.0017], device='cuda:0'), grad: tensor([ 0.0007, 0.0037, 0.0079, 0.0024, -0.0009, 0.0009, -0.0122, -0.0019, + 0.0005, -0.0013], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 230.45, cls_loss 0.0540 cls_loss_mapping 0.0815 cls_loss_causal 0.9902 re_mapping 0.0366 re_causal 0.1008 /// teacc 98.24 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0134, 0.0039, 0.0042, ..., -0.0532, -0.0089, -0.0079], + [ 0.0187, -0.0115, -0.0547, ..., -0.0039, -0.0243, -0.0419], + [ 0.0330, -0.0297, -0.0043, ..., 0.0278, -0.0391, 0.0082], + ..., + [ 0.0124, -0.0315, -0.0208, ..., -0.0131, 0.0005, 0.0488], + [ 0.0316, -0.0330, -0.0311, ..., -0.0111, -0.0005, 0.0032], + [-0.0668, -0.0100, 0.0449, ..., 0.0195, -0.0031, -0.0203]], + device='cuda:0'), grad: tensor([[-6.6853e-04, 4.0603e-04, 4.3249e-04, ..., 1.3995e-04, + 7.2131e-07, 6.6400e-05], + [-2.4414e-03, -3.2830e-04, -1.2226e-03, ..., -6.6900e-04, + -4.3720e-05, 1.6069e-04], + [ 1.6129e-04, 1.7667e-04, 4.1342e-04, ..., 1.4889e-04, + 4.4983e-07, -1.1225e-03], + ..., + [ 2.4748e-04, -5.4091e-05, 5.5647e-04, ..., 1.8501e-04, + 2.2389e-06, -2.7442e-04], + [ 1.1072e-03, 2.8782e-03, 3.7689e-03, ..., 1.3380e-03, + 1.4953e-05, 2.3091e-04], + [ 3.0351e-04, 1.1986e-02, 1.4351e-02, ..., 3.8776e-03, + 7.0035e-06, 1.4853e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0221, -0.0183, -0.0291, 0.0035, -0.0008, 0.0298, 0.0144, -0.0246, + 0.0147, -0.0017], device='cuda:0'), grad: tensor([ 0.0001, -0.0082, 0.0014, -0.0244, 0.0020, 0.0008, 0.0006, 0.0010, + 0.0070, 0.0197], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 214.07, cls_loss 0.0537 cls_loss_mapping 0.0797 cls_loss_causal 0.9844 re_mapping 0.0358 re_causal 0.0970 /// teacc 97.98 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0130, 0.0036, 0.0044, ..., -0.0536, -0.0092, -0.0095], + [ 0.0186, -0.0123, -0.0549, ..., -0.0038, -0.0242, -0.0436], + [ 0.0328, -0.0307, -0.0050, ..., 0.0279, -0.0403, 0.0087], + ..., + [ 0.0128, -0.0313, -0.0216, ..., -0.0137, 0.0025, 0.0496], + [ 0.0322, -0.0341, -0.0319, ..., -0.0120, -0.0008, 0.0029], + [-0.0678, -0.0095, 0.0457, ..., 0.0205, -0.0037, -0.0208]], + device='cuda:0'), grad: tensor([[ 2.5821e-04, 1.8942e-04, 7.4327e-05, ..., 9.7096e-05, + 0.0000e+00, 1.4806e-04], + [ 3.3784e-04, 2.1422e-04, 1.8132e-04, ..., 1.1498e-04, + 0.0000e+00, 1.9014e-04], + [ 1.6665e-04, 1.9026e-04, 1.1617e-04, ..., 5.0187e-05, + 0.0000e+00, 1.6913e-05], + ..., + [-1.5926e-04, 3.9825e-03, 1.6527e-03, ..., 1.0881e-03, + 0.0000e+00, 3.7251e-03], + [-8.9931e-04, 1.2236e-03, 5.1165e-04, ..., 8.9025e-04, + 0.0000e+00, 4.3154e-04], + [ 6.6996e-04, -9.6262e-05, -4.8714e-03, ..., -3.4428e-03, + 0.0000e+00, 1.6413e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0220, -0.0185, -0.0294, 0.0036, -0.0011, 0.0295, 0.0146, -0.0241, + 0.0148, -0.0017], device='cuda:0'), grad: tensor([ 4.7207e-04, 7.0286e-04, 4.2629e-04, -1.5080e-04, 2.2449e-03, + -6.3515e-03, -1.9014e-04, 3.9005e-03, 1.1884e-05, -1.0662e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 229.83, cls_loss 0.0562 cls_loss_mapping 0.0891 cls_loss_causal 0.9423 re_mapping 0.0330 re_causal 0.0891 /// teacc 98.27 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0130, 0.0031, 0.0039, ..., -0.0546, -0.0097, -0.0105], + [ 0.0185, -0.0134, -0.0557, ..., -0.0041, -0.0247, -0.0443], + [ 0.0331, -0.0319, -0.0057, ..., 0.0272, -0.0412, 0.0091], + ..., + [ 0.0127, -0.0310, -0.0227, ..., -0.0150, 0.0028, 0.0501], + [ 0.0327, -0.0348, -0.0318, ..., -0.0123, -0.0019, 0.0026], + [-0.0690, -0.0093, 0.0465, ..., 0.0213, -0.0041, -0.0215]], + device='cuda:0'), grad: tensor([[ 5.2404e-04, 2.5105e-04, 3.7241e-04, ..., 2.2829e-04, + 9.2667e-08, 2.3162e-04], + [ 6.4611e-04, 1.9813e-04, 2.2602e-04, ..., 1.4269e-04, + -1.6997e-06, 4.3780e-05], + [-8.1062e-04, 4.8661e-04, 7.3242e-04, ..., 5.1498e-04, + 2.7195e-07, -6.4421e-04], + ..., + [-1.9455e-04, -2.3890e-04, 4.1723e-04, ..., 6.8843e-05, + 3.2829e-07, -2.7442e-04], + [-1.4219e-03, 3.0994e-04, -4.2319e-04, ..., 2.1350e-04, + 2.3609e-07, 2.3448e-04], + [ 6.9571e-04, 3.0851e-04, -8.2910e-05, ..., -5.6237e-05, + 1.2247e-07, 1.4174e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0222, -0.0188, -0.0295, 0.0036, -0.0009, 0.0301, 0.0144, -0.0241, + 0.0150, -0.0021], device='cuda:0'), grad: tensor([ 1.0881e-03, 1.0080e-03, -1.0815e-03, 1.0185e-03, -2.9430e-03, + -8.8334e-05, 1.7996e-03, -3.5810e-04, -1.3514e-03, 9.1314e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 230.06, cls_loss 0.0506 cls_loss_mapping 0.0746 cls_loss_causal 0.9436 re_mapping 0.0325 re_causal 0.0885 /// teacc 98.46 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0126, 0.0024, 0.0038, ..., -0.0553, -0.0098, -0.0118], + [ 0.0180, -0.0134, -0.0556, ..., -0.0038, -0.0247, -0.0448], + [ 0.0333, -0.0326, -0.0061, ..., 0.0271, -0.0421, 0.0097], + ..., + [ 0.0127, -0.0309, -0.0237, ..., -0.0159, 0.0034, 0.0510], + [ 0.0333, -0.0358, -0.0321, ..., -0.0131, -0.0022, 0.0015], + [-0.0702, -0.0087, 0.0474, ..., 0.0225, -0.0043, -0.0219]], + device='cuda:0'), grad: tensor([[ 4.5824e-04, 2.1720e-04, 8.6248e-05, ..., 1.9932e-04, + 3.4971e-07, 4.5371e-04], + [-6.9559e-05, 1.9479e-04, -1.3709e-04, ..., 1.8299e-05, + 9.0338e-07, 3.9220e-04], + [-1.3456e-03, 4.1986e-04, 2.6846e-04, ..., 2.2840e-04, + 5.0664e-07, -2.1648e-03], + ..., + [-9.9850e-04, -1.8320e-03, 1.7777e-05, ..., -7.7724e-05, + -6.4112e-06, -4.0321e-03], + [ 1.3399e-03, 4.4155e-04, 1.2732e-04, ..., 3.9077e-04, + 7.0874e-07, 9.7513e-04], + [ 9.2316e-04, 7.4387e-04, 1.5783e-04, ..., 3.1948e-04, + 1.2489e-06, 8.5449e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0222, -0.0190, -0.0291, 0.0035, -0.0011, 0.0302, 0.0143, -0.0240, + 0.0152, -0.0022], device='cuda:0'), grad: tensor([ 8.5640e-04, -4.0323e-05, -2.2717e-03, 4.0932e-03, -1.1892e-03, + -1.7862e-03, 7.9441e-04, -5.2185e-03, 2.3727e-03, 2.3880e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 225.04, cls_loss 0.0359 cls_loss_mapping 0.0588 cls_loss_causal 0.9127 re_mapping 0.0303 re_causal 0.0877 /// teacc 98.54 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0120, 0.0020, 0.0036, ..., -0.0558, -0.0098, -0.0119], + [ 0.0182, -0.0141, -0.0560, ..., -0.0036, -0.0247, -0.0460], + [ 0.0334, -0.0335, -0.0068, ..., 0.0269, -0.0421, 0.0104], + ..., + [ 0.0128, -0.0303, -0.0244, ..., -0.0164, 0.0035, 0.0518], + [ 0.0337, -0.0364, -0.0323, ..., -0.0132, -0.0022, 0.0003], + [-0.0711, -0.0085, 0.0479, ..., 0.0230, -0.0043, -0.0226]], + device='cuda:0'), grad: tensor([[ 1.6582e-04, 1.2290e-04, 3.6478e-04, ..., 1.5450e-04, + 6.2175e-06, 1.5795e-05], + [ 4.2200e-05, 5.7220e-05, 4.2887e-07, ..., -9.8348e-06, + 1.9316e-06, 7.0393e-05], + [ 1.1361e-04, 6.1631e-05, 8.8334e-05, ..., 1.5274e-05, + 1.3597e-06, -2.4959e-05], + ..., + [-2.3437e-04, 3.9840e-04, 7.2658e-05, ..., -1.5199e-04, + -1.2386e-04, 7.3242e-04], + [-4.7982e-05, 1.0288e-04, 3.5614e-05, ..., 2.7180e-04, + 3.4384e-06, 3.1173e-05], + [ 1.1425e-03, 6.0272e-04, 7.1526e-04, ..., 3.8624e-04, + 7.1824e-05, 4.1032e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0220, -0.0188, -0.0292, 0.0035, -0.0011, 0.0301, 0.0141, -0.0237, + 0.0153, -0.0025], device='cuda:0'), grad: tensor([ 2.2018e-04, 5.4359e-05, 1.6773e-04, -2.0891e-05, 1.2589e-04, + -5.0430e-03, 2.3975e-03, 1.9622e-04, 8.9705e-05, 1.8187e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 213.92, cls_loss 0.0404 cls_loss_mapping 0.0628 cls_loss_causal 0.8779 re_mapping 0.0286 re_causal 0.0809 /// teacc 98.41 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0123, 0.0014, 0.0030, ..., -0.0564, -0.0099, -0.0126], + [ 0.0183, -0.0143, -0.0562, ..., -0.0041, -0.0248, -0.0463], + [ 0.0332, -0.0344, -0.0072, ..., 0.0265, -0.0421, 0.0111], + ..., + [ 0.0126, -0.0301, -0.0253, ..., -0.0170, 0.0035, 0.0523], + [ 0.0341, -0.0373, -0.0324, ..., -0.0139, -0.0024, -0.0004], + [-0.0718, -0.0081, 0.0487, ..., 0.0241, -0.0044, -0.0231]], + device='cuda:0'), grad: tensor([[ 5.2094e-05, 9.6977e-05, 8.6069e-05, ..., 4.5806e-05, + 4.2887e-07, 2.2605e-05], + [ 1.8752e-04, 7.2241e-05, -3.1328e-04, ..., 2.0635e-04, + 3.0268e-08, 4.6039e-04], + [-4.3178e-04, 4.0859e-05, 3.8207e-05, ..., -5.8126e-04, + 4.6566e-09, -5.1928e-04], + ..., + [-6.2943e-05, -2.1935e-04, 4.9353e-05, ..., 1.3185e-04, + 5.4948e-08, -1.8489e-04], + [-1.0931e-04, 2.8300e-04, 4.1127e-04, ..., 3.5191e-04, + 3.3453e-06, 5.1647e-05], + [ 2.0874e-04, -2.6250e-04, -8.0204e-04, ..., -5.0020e-04, + 1.2759e-07, 9.2864e-05]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0225, -0.0188, -0.0292, 0.0033, -0.0010, 0.0305, 0.0141, -0.0240, + 0.0152, -0.0024], device='cuda:0'), grad: tensor([ 1.0568e-04, 9.1982e-04, -1.5783e-03, 2.7919e-04, 5.5504e-04, + -7.9679e-04, 8.3590e-04, -2.5344e-04, 5.4687e-05, -1.1921e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 213.81, cls_loss 0.0358 cls_loss_mapping 0.0563 cls_loss_causal 0.9041 re_mapping 0.0274 re_causal 0.0809 /// teacc 98.24 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0119, 0.0009, 0.0024, ..., -0.0567, -0.0100, -0.0139], + [ 0.0185, -0.0149, -0.0565, ..., -0.0050, -0.0248, -0.0467], + [ 0.0331, -0.0352, -0.0078, ..., 0.0271, -0.0422, 0.0111], + ..., + [ 0.0125, -0.0303, -0.0263, ..., -0.0181, 0.0036, 0.0526], + [ 0.0341, -0.0382, -0.0328, ..., -0.0145, -0.0026, -0.0008], + [-0.0722, -0.0077, 0.0498, ..., 0.0252, -0.0045, -0.0236]], + device='cuda:0'), grad: tensor([[-1.3094e-03, -1.0318e-04, -4.2558e-05, ..., 5.1260e-05, + 2.0955e-08, 3.7909e-05], + [-4.1068e-05, 7.4983e-05, -1.1724e-04, ..., -2.7642e-05, + -7.0734e-07, 1.1754e-04], + [-9.3162e-05, 1.6236e-04, 2.3678e-05, ..., 4.2886e-05, + 6.2864e-08, -1.4520e-04], + ..., + [-5.0724e-05, 3.7849e-05, 1.9312e-04, ..., 2.5797e-04, + 1.8673e-07, -2.1565e-04], + [ 3.5465e-05, 4.9400e-04, 2.5225e-04, ..., 2.7871e-04, + 1.2526e-07, 1.8609e-04], + [ 1.0735e-04, -1.5764e-03, -1.8425e-03, ..., -3.3054e-03, + 1.0151e-07, 2.3961e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0227, -0.0187, -0.0294, 0.0035, -0.0008, 0.0304, 0.0144, -0.0244, + 0.0151, -0.0019], device='cuda:0'), grad: tensor([-1.2884e-03, -1.4067e-04, -3.4511e-05, -7.1621e-04, 2.3746e-04, + -1.1658e-02, 1.4793e-02, 1.6356e-04, 6.3992e-04, -1.9932e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 214.16, cls_loss 0.0321 cls_loss_mapping 0.0554 cls_loss_causal 0.8826 re_mapping 0.0266 re_causal 0.0792 /// teacc 98.34 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0115, 0.0005, 0.0020, ..., -0.0574, -0.0101, -0.0147], + [ 0.0184, -0.0151, -0.0567, ..., -0.0050, -0.0249, -0.0475], + [ 0.0332, -0.0360, -0.0080, ..., 0.0270, -0.0424, 0.0118], + ..., + [ 0.0126, -0.0301, -0.0269, ..., -0.0190, 0.0039, 0.0531], + [ 0.0346, -0.0388, -0.0332, ..., -0.0151, -0.0029, -0.0007], + [-0.0732, -0.0075, 0.0504, ..., 0.0261, -0.0048, -0.0243]], + device='cuda:0'), grad: tensor([[-3.4618e-03, -2.9892e-05, 4.9651e-05, ..., -8.9264e-04, + 0.0000e+00, -1.9073e-04], + [ 2.6083e-04, 2.1839e-04, 5.4777e-05, ..., 1.4925e-04, + 0.0000e+00, 5.5647e-04], + [-2.2829e-04, 3.0428e-05, 4.9949e-05, ..., 2.8074e-05, + 0.0000e+00, -8.5735e-04], + ..., + [ 1.6344e-04, 1.0633e-03, 1.1864e-03, ..., 1.2245e-03, + 0.0000e+00, 8.3923e-04], + [ 4.3631e-04, 1.2267e-04, 8.1778e-05, ..., 1.9896e-04, + 0.0000e+00, 2.7251e-04], + [ 3.5882e-04, -1.6041e-03, -1.7691e-03, ..., -1.7052e-03, + 0.0000e+00, -9.8896e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0225, -0.0186, -0.0291, 0.0033, -0.0008, 0.0305, 0.0138, -0.0243, + 0.0152, -0.0021], device='cuda:0'), grad: tensor([-4.5166e-03, 6.9952e-04, -6.4945e-04, 1.1486e-04, 1.2999e-03, + 2.1858e-03, 8.9586e-05, 2.4357e-03, 8.3494e-04, -2.4967e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 214.09, cls_loss 0.0318 cls_loss_mapping 0.0542 cls_loss_causal 0.8527 re_mapping 0.0259 re_causal 0.0762 /// teacc 98.50 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0115, 0.0001, 0.0014, ..., -0.0577, -0.0101, -0.0150], + [ 0.0184, -0.0157, -0.0567, ..., -0.0049, -0.0253, -0.0487], + [ 0.0334, -0.0363, -0.0086, ..., 0.0267, -0.0427, 0.0132], + ..., + [ 0.0129, -0.0299, -0.0278, ..., -0.0198, 0.0042, 0.0536], + [ 0.0346, -0.0394, -0.0336, ..., -0.0158, -0.0035, -0.0019], + [-0.0737, -0.0072, 0.0510, ..., 0.0266, -0.0052, -0.0245]], + device='cuda:0'), grad: tensor([[-1.5106e-03, 8.3160e-04, 1.8740e-04, ..., 4.2224e-04, + 1.8040e-06, 3.2568e-04], + [ 7.1526e-05, 8.7857e-05, -1.3053e-05, ..., 5.3078e-05, + -6.7893e-07, 1.2684e-04], + [-6.2585e-05, 2.0289e-04, 4.6343e-05, ..., 8.9586e-05, + 7.1805e-07, -6.4564e-04], + ..., + [ 9.0718e-05, 1.0109e-03, 1.1700e-04, ..., 3.7789e-04, + -8.5309e-07, 2.0256e-03], + [ 2.3842e-04, 5.4836e-04, 7.7307e-05, ..., 3.6931e-04, + 1.2293e-05, 6.3801e-04], + [ 1.1539e-04, 5.5504e-04, -2.0504e-05, ..., 2.1005e-04, + 3.8594e-06, 4.7874e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0227, -0.0186, -0.0288, 0.0033, -0.0006, 0.0301, 0.0140, -0.0243, + 0.0149, -0.0020], device='cuda:0'), grad: tensor([-0.0005, 0.0002, -0.0003, -0.0063, 0.0002, 0.0012, 0.0017, 0.0020, + 0.0009, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 230.20, cls_loss 0.0314 cls_loss_mapping 0.0542 cls_loss_causal 0.8720 re_mapping 0.0250 re_causal 0.0765 /// teacc 98.57 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0113, -0.0005, 0.0008, ..., -0.0591, -0.0102, -0.0156], + [ 0.0182, -0.0167, -0.0567, ..., -0.0058, -0.0243, -0.0495], + [ 0.0337, -0.0367, -0.0092, ..., 0.0275, -0.0434, 0.0139], + ..., + [ 0.0127, -0.0297, -0.0286, ..., -0.0206, 0.0052, 0.0538], + [ 0.0349, -0.0403, -0.0339, ..., -0.0160, -0.0038, -0.0023], + [-0.0743, -0.0071, 0.0517, ..., 0.0274, -0.0069, -0.0251]], + device='cuda:0'), grad: tensor([[-2.4819e-04, -5.8979e-05, 2.3529e-05, ..., 3.3140e-05, + 6.0201e-06, 6.5565e-05], + [ 3.7026e-04, 2.1410e-04, -1.2241e-05, ..., -1.2659e-05, + 1.2768e-06, 2.7680e-04], + [ 8.9121e-04, 1.6201e-04, 4.5598e-05, ..., 2.2978e-05, + 1.5274e-07, 2.3746e-03], + ..., + [-3.3798e-03, -1.7519e-03, 1.2183e-04, ..., 1.1247e-04, + 7.8827e-06, -3.4676e-03], + [ 9.8133e-04, 5.8413e-04, 2.5058e-04, ..., 3.0398e-04, + 4.4674e-05, 2.9349e-04], + [ 2.3437e-04, -2.9516e-04, -6.6042e-04, ..., -4.8757e-04, + 1.1446e-06, 2.3708e-05]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0226, -0.0187, -0.0285, 0.0035, -0.0005, 0.0298, 0.0139, -0.0244, + 0.0151, -0.0021], device='cuda:0'), grad: tensor([-2.3723e-04, 5.9700e-04, 2.2488e-03, 4.9973e-04, 1.8702e-03, + -2.0824e-06, 1.5497e-04, -6.7596e-03, 1.7366e-03, -1.1307e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 214.10, cls_loss 0.0270 cls_loss_mapping 0.0493 cls_loss_causal 0.8440 re_mapping 0.0243 re_causal 0.0721 /// teacc 98.37 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0112, -0.0008, 0.0009, ..., -0.0602, -0.0103, -0.0164], + [ 0.0176, -0.0177, -0.0567, ..., -0.0064, -0.0240, -0.0511], + [ 0.0342, -0.0373, -0.0097, ..., 0.0278, -0.0440, 0.0150], + ..., + [ 0.0126, -0.0298, -0.0295, ..., -0.0216, 0.0054, 0.0542], + [ 0.0353, -0.0410, -0.0346, ..., -0.0164, -0.0042, -0.0032], + [-0.0750, -0.0066, 0.0522, ..., 0.0280, -0.0073, -0.0254]], + device='cuda:0'), grad: tensor([[ 5.7727e-05, 4.4322e-04, 1.3793e-04, ..., 2.8038e-04, + 0.0000e+00, 4.9829e-04], + [-3.8934e-04, 3.0565e-04, 1.1700e-04, ..., -9.4032e-04, + 0.0000e+00, 3.1376e-04], + [-1.3189e-03, -4.0483e-04, 1.0961e-04, ..., -2.0516e-04, + 0.0000e+00, -1.5345e-03], + ..., + [-4.1962e-03, -8.7814e-03, -4.1046e-03, ..., -7.9203e-04, + 0.0000e+00, -7.4959e-03], + [ 3.5048e-04, 6.2180e-04, 1.9109e-04, ..., 4.4942e-04, + 0.0000e+00, 6.2084e-04], + [ 4.1389e-03, -2.8825e-04, -1.6617e-02, ..., -9.6588e-03, + 0.0000e+00, 8.6823e-03]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0227, -0.0191, -0.0283, 0.0033, -0.0005, 0.0301, 0.0140, -0.0247, + 0.0152, -0.0020], device='cuda:0'), grad: tensor([ 0.0008, -0.0010, -0.0029, -0.0002, 0.0139, -0.0019, 0.0013, -0.0173, + 0.0014, 0.0060], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 229.97, cls_loss 0.0239 cls_loss_mapping 0.0413 cls_loss_causal 0.8294 re_mapping 0.0230 re_causal 0.0687 /// teacc 98.62 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0110, -0.0012, 0.0004, ..., -0.0606, -0.0105, -0.0171], + [ 0.0176, -0.0185, -0.0569, ..., -0.0065, -0.0220, -0.0518], + [ 0.0346, -0.0378, -0.0102, ..., 0.0281, -0.0454, 0.0155], + ..., + [ 0.0129, -0.0296, -0.0300, ..., -0.0216, 0.0058, 0.0544], + [ 0.0352, -0.0417, -0.0352, ..., -0.0172, -0.0050, -0.0034], + [-0.0759, -0.0064, 0.0529, ..., 0.0287, -0.0087, -0.0263]], + device='cuda:0'), grad: tensor([[ 6.8247e-05, 2.5392e-05, 5.2303e-05, ..., 2.3142e-05, + 2.7753e-07, 7.0594e-06], + [ 4.2617e-05, 2.9698e-05, 4.5091e-05, ..., 2.0772e-05, + 8.4937e-07, 3.3051e-05], + [ 1.1724e-04, 5.2840e-05, 7.7069e-05, ..., 2.8402e-05, + 8.2515e-07, -1.0617e-05], + ..., + [ 2.5421e-05, 5.9977e-06, 1.9026e-04, ..., 9.7692e-05, + -8.5905e-06, -8.1182e-05], + [-1.4658e-03, 1.4842e-04, -2.6226e-04, ..., -5.6887e-04, + 6.0722e-07, -2.2188e-05], + [ 3.6645e-04, -6.2943e-04, -1.0672e-03, ..., -4.0627e-04, + 2.4028e-06, 4.4465e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0228, -0.0188, -0.0283, 0.0031, -0.0006, 0.0302, 0.0142, -0.0246, + 0.0151, -0.0020], device='cuda:0'), grad: tensor([ 1.0806e-04, 7.7844e-05, 2.0182e-04, 1.3578e-04, 6.2561e-04, + 8.8501e-04, -7.5936e-05, 1.3232e-04, -1.6441e-03, -4.4703e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 214.14, cls_loss 0.0234 cls_loss_mapping 0.0431 cls_loss_causal 0.8243 re_mapping 0.0232 re_causal 0.0705 /// teacc 98.39 lr 0.00010000 +Epoch 28, weight, value: tensor([[-1.1142e-02, -1.3557e-03, 9.4968e-05, ..., -6.1634e-02, + -1.0396e-02, -1.7993e-02], + [ 1.7261e-02, -1.9552e-02, -5.7030e-02, ..., -6.8516e-03, + -2.2105e-02, -5.3031e-02], + [ 3.4955e-02, -3.8593e-02, -1.0770e-02, ..., 2.8062e-02, + -4.5719e-02, 1.6113e-02], + ..., + [ 1.3258e-02, -2.9211e-02, -3.0561e-02, ..., -2.2106e-02, + 6.9902e-03, 5.5054e-02], + [ 3.5638e-02, -4.2237e-02, -3.5133e-02, ..., -1.7420e-02, + -5.2915e-03, -3.8666e-03], + [-7.7052e-02, -6.0905e-03, 5.3500e-02, ..., 2.9434e-02, + -1.0803e-02, -2.6800e-02]], device='cuda:0'), grad: tensor([[-7.5674e-04, 5.0992e-05, 1.0872e-04, ..., 1.1042e-05, + 6.9384e-07, 4.1395e-05], + [ 5.3197e-05, 4.2737e-05, -6.0707e-05, ..., -6.7830e-05, + 1.8068e-06, 1.3661e-04], + [ 5.7411e-04, 1.4234e-04, 1.3220e-04, ..., 5.3227e-05, + 1.9651e-07, 1.3208e-03], + ..., + [ 3.7432e-05, -4.6134e-05, 2.2650e-04, ..., 1.1957e-04, + -6.5118e-06, -1.9121e-04], + [ 3.3140e-04, 1.0359e-04, 1.3840e-04, ..., 6.3658e-05, + 4.4964e-06, 4.8089e-04], + [ 7.0632e-05, -1.0818e-04, -4.1318e-04, ..., -2.9469e-04, + 3.3993e-06, 1.3161e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0230, -0.0193, -0.0279, 0.0031, -0.0008, 0.0302, 0.0143, -0.0244, + 0.0150, -0.0020], device='cuda:0'), grad: tensor([-5.7697e-04, 4.0680e-05, 1.5526e-03, -1.7042e-03, 6.1464e-04, + 1.9205e-04, -6.4564e-04, -1.0645e-04, 6.5756e-04, -2.3812e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 27---------------------------------------------------- +epoch 27, time 230.04, cls_loss 0.0261 cls_loss_mapping 0.0451 cls_loss_causal 0.8714 re_mapping 0.0229 re_causal 0.0678 /// teacc 98.69 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0110, -0.0020, -0.0006, ..., -0.0625, -0.0103, -0.0196], + [ 0.0172, -0.0201, -0.0572, ..., -0.0064, -0.0222, -0.0542], + [ 0.0350, -0.0397, -0.0115, ..., 0.0278, -0.0458, 0.0165], + ..., + [ 0.0133, -0.0293, -0.0313, ..., -0.0237, 0.0070, 0.0556], + [ 0.0357, -0.0427, -0.0350, ..., -0.0178, -0.0055, -0.0043], + [-0.0779, -0.0056, 0.0540, ..., 0.0304, -0.0109, -0.0269]], + device='cuda:0'), grad: tensor([[ 2.8825e-04, 2.5243e-05, 8.1360e-05, ..., 4.1097e-05, + 1.5274e-07, 5.5313e-04], + [ 1.2732e-04, 7.6652e-05, -8.3819e-06, ..., -5.3197e-05, + 2.7008e-08, 1.4257e-04], + [-7.1669e-04, 6.0201e-05, 6.5148e-05, ..., 2.4214e-05, + 1.1176e-08, -9.9087e-04], + ..., + [ 2.6298e-04, 9.2924e-05, 3.5191e-04, ..., 1.7571e-04, + 1.6019e-07, -4.2510e-04], + [ 6.9976e-05, 5.2989e-05, 5.7220e-05, ..., 4.9680e-05, + 7.7579e-07, 1.9479e-04], + [ 2.2745e-04, 1.6141e-04, 1.1629e-04, ..., 7.3552e-05, + 8.4750e-08, 5.4032e-05]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0233, -0.0196, -0.0281, 0.0034, -0.0006, 0.0301, 0.0143, -0.0244, + 0.0153, -0.0020], device='cuda:0'), grad: tensor([ 8.8215e-04, 6.9678e-05, -1.6632e-03, 3.1734e-04, -9.1982e-04, + -1.8507e-05, 5.1022e-04, 1.2732e-04, 3.5071e-04, 3.4475e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 213.92, cls_loss 0.0217 cls_loss_mapping 0.0385 cls_loss_causal 0.7690 re_mapping 0.0222 re_causal 0.0630 /// teacc 98.44 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0103, -0.0024, -0.0003, ..., -0.0626, -0.0104, -0.0199], + [ 0.0177, -0.0206, -0.0565, ..., -0.0058, -0.0214, -0.0546], + [ 0.0348, -0.0406, -0.0120, ..., 0.0273, -0.0477, 0.0169], + ..., + [ 0.0132, -0.0292, -0.0320, ..., -0.0244, 0.0070, 0.0560], + [ 0.0359, -0.0433, -0.0359, ..., -0.0184, -0.0059, -0.0045], + [-0.0788, -0.0054, 0.0542, ..., 0.0308, -0.0111, -0.0273]], + device='cuda:0'), grad: tensor([[ 4.4554e-05, 1.7181e-05, 5.4240e-05, ..., 1.1928e-05, + 5.5786e-07, 4.3243e-05], + [ 4.5389e-05, 2.0161e-05, -3.7879e-05, ..., -9.0778e-05, + -2.2873e-06, 9.7156e-05], + [ 4.7833e-06, 2.6762e-05, 4.3154e-05, ..., 1.6615e-05, + 6.5379e-07, -3.8862e-04], + ..., + [ 3.7879e-05, -7.2047e-06, 9.0301e-05, ..., 7.1526e-05, + -7.0706e-06, -5.4896e-05], + [-6.2799e-04, -1.0461e-04, -5.1498e-04, ..., -2.4188e-04, + 4.4629e-06, -1.0061e-04], + [ 3.6430e-04, -8.3506e-05, 2.1681e-05, ..., -5.8919e-05, + 3.1237e-06, 2.0516e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0229, -0.0189, -0.0283, 0.0033, -0.0005, 0.0301, 0.0144, -0.0246, + 0.0151, -0.0022], device='cuda:0'), grad: tensor([ 8.9407e-05, 5.1111e-05, -2.3866e-04, 2.1601e-04, 2.0611e-04, + 2.9588e-04, -8.5592e-05, 9.9897e-05, -1.2150e-03, 5.8174e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 230.24, cls_loss 0.0240 cls_loss_mapping 0.0394 cls_loss_causal 0.8085 re_mapping 0.0212 re_causal 0.0631 /// teacc 98.70 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0101, -0.0027, -0.0009, ..., -0.0633, -0.0105, -0.0207], + [ 0.0176, -0.0211, -0.0567, ..., -0.0058, -0.0184, -0.0553], + [ 0.0352, -0.0413, -0.0126, ..., 0.0272, -0.0488, 0.0173], + ..., + [ 0.0131, -0.0290, -0.0327, ..., -0.0252, 0.0080, 0.0565], + [ 0.0363, -0.0443, -0.0362, ..., -0.0190, -0.0063, -0.0052], + [-0.0798, -0.0050, 0.0553, ..., 0.0318, -0.0125, -0.0279]], + device='cuda:0'), grad: tensor([[-2.5702e-04, 1.5900e-05, -1.4579e-04, ..., 2.5496e-05, + 5.1595e-07, 2.0087e-04], + [ 2.3568e-04, 1.4508e-04, 7.5996e-05, ..., 6.9916e-05, + -8.2031e-06, 3.4881e-04], + [-4.6349e-03, -2.5635e-03, 2.4629e-04, ..., -4.5180e-04, + 8.1211e-07, -8.6517e-03], + ..., + [ 2.0390e-03, 1.0071e-03, 4.6349e-04, ..., 5.6505e-04, + 3.5856e-06, 3.3932e-03], + [ 2.2144e-03, 1.4429e-03, 1.1997e-03, ..., 8.5068e-04, + 5.0701e-06, 2.6245e-03], + [ 5.2452e-04, 2.9922e-04, -5.5027e-04, ..., -3.3975e-04, + 2.1271e-06, 8.9979e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0231, -0.0189, -0.0284, 0.0035, -0.0011, 0.0300, 0.0146, -0.0242, + 0.0151, -0.0021], device='cuda:0'), grad: tensor([-0.0004, 0.0006, -0.0109, 0.0013, -0.0013, 0.0003, -0.0006, 0.0049, + 0.0051, 0.0010], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 214.02, cls_loss 0.0189 cls_loss_mapping 0.0342 cls_loss_causal 0.7963 re_mapping 0.0204 re_causal 0.0629 /// teacc 98.54 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0098, -0.0024, -0.0004, ..., -0.0638, -0.0106, -0.0213], + [ 0.0175, -0.0214, -0.0569, ..., -0.0057, -0.0182, -0.0554], + [ 0.0354, -0.0424, -0.0131, ..., 0.0269, -0.0493, 0.0173], + ..., + [ 0.0129, -0.0287, -0.0333, ..., -0.0261, 0.0095, 0.0573], + [ 0.0364, -0.0453, -0.0366, ..., -0.0199, -0.0065, -0.0058], + [-0.0805, -0.0049, 0.0557, ..., 0.0322, -0.0145, -0.0284]], + device='cuda:0'), grad: tensor([[-1.6153e-04, 2.3395e-05, -7.0453e-05, ..., 3.9756e-05, + 7.7020e-07, 3.4362e-05], + [ 2.0131e-05, 4.0352e-05, -1.7536e-04, ..., -1.3435e-04, + 5.8115e-07, 3.9816e-05], + [-5.0217e-05, 1.5289e-05, 7.1824e-05, ..., 5.2124e-05, + 4.4610e-07, -1.2863e-04], + ..., + [ 1.2182e-05, 7.0000e-04, 1.0796e-03, ..., 7.8154e-04, + -9.4026e-06, 8.4400e-05], + [-4.2558e-04, 3.3349e-05, -3.5644e-04, ..., -3.8385e-04, + 4.1910e-07, 5.0873e-05], + [ 4.5776e-04, -6.4850e-04, -8.1873e-04, ..., -3.3069e-04, + 2.9076e-06, -3.6359e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0228, -0.0187, -0.0287, 0.0034, -0.0008, 0.0302, 0.0143, -0.0241, + 0.0149, -0.0023], device='cuda:0'), grad: tensor([-1.9884e-04, -2.1458e-04, 3.2216e-05, -3.4857e-04, 2.1935e-04, + 1.3006e-04, 1.6534e-04, 1.0920e-03, -1.0090e-03, 1.3089e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 213.93, cls_loss 0.0201 cls_loss_mapping 0.0357 cls_loss_causal 0.7376 re_mapping 0.0205 re_causal 0.0592 /// teacc 98.54 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0096, -0.0031, -0.0008, ..., -0.0645, -0.0113, -0.0219], + [ 0.0174, -0.0209, -0.0565, ..., -0.0052, -0.0150, -0.0564], + [ 0.0352, -0.0433, -0.0135, ..., 0.0261, -0.0496, 0.0174], + ..., + [ 0.0128, -0.0287, -0.0342, ..., -0.0265, 0.0092, 0.0579], + [ 0.0368, -0.0456, -0.0365, ..., -0.0203, -0.0068, -0.0062], + [-0.0814, -0.0047, 0.0564, ..., 0.0329, -0.0151, -0.0291]], + device='cuda:0'), grad: tensor([[ 2.0191e-05, 7.8976e-05, 1.2517e-05, ..., 9.8109e-05, + 5.3085e-08, 7.4685e-05], + [ 2.4304e-05, 7.7426e-05, -3.2723e-05, ..., -2.5123e-05, + -1.9390e-06, 9.3460e-05], + [-5.4264e-04, 9.4056e-05, -2.8324e-04, ..., 6.1631e-05, + 1.5274e-07, 2.4632e-05], + ..., + [-8.9312e-04, -1.3971e-03, 2.3931e-05, ..., -2.5330e-03, + 7.4692e-07, -1.2165e-04], + [ 2.8753e-04, 1.9646e-04, 3.0971e-04, ..., 1.7917e-04, + 2.7381e-07, 2.6393e-04], + [ 5.2929e-04, 6.8760e-04, -5.0336e-05, ..., 1.1549e-03, + 2.9244e-07, 2.2602e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0229, -0.0184, -0.0292, 0.0036, -0.0006, 0.0299, 0.0146, -0.0240, + 0.0151, -0.0025], device='cuda:0'), grad: tensor([ 1.5366e-04, 1.9148e-05, -6.3896e-04, -8.4591e-04, 4.7660e-04, + 1.6584e-03, 7.7724e-05, -4.3716e-03, 9.9277e-04, 2.4815e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 213.73, cls_loss 0.0162 cls_loss_mapping 0.0351 cls_loss_causal 0.7262 re_mapping 0.0207 re_causal 0.0591 /// teacc 98.56 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0094, -0.0036, -0.0014, ..., -0.0652, -0.0113, -0.0225], + [ 0.0165, -0.0214, -0.0564, ..., -0.0051, -0.0149, -0.0577], + [ 0.0354, -0.0442, -0.0140, ..., 0.0258, -0.0497, 0.0178], + ..., + [ 0.0128, -0.0280, -0.0348, ..., -0.0268, 0.0092, 0.0586], + [ 0.0374, -0.0462, -0.0367, ..., -0.0207, -0.0069, -0.0065], + [-0.0818, -0.0046, 0.0569, ..., 0.0336, -0.0151, -0.0296]], + device='cuda:0'), grad: tensor([[-4.0665e-03, -1.3344e-02, -1.4183e-02, ..., 1.3933e-05, + 0.0000e+00, 4.5151e-05], + [ 8.7142e-05, 2.8062e-04, 4.0889e-04, ..., 2.9397e-04, + 0.0000e+00, 5.1171e-05], + [ 7.4530e-04, 1.6432e-03, 1.8578e-03, ..., 6.8069e-05, + 0.0000e+00, -6.2418e-04], + ..., + [ 8.6606e-05, -2.3580e-04, 2.1785e-05, ..., -1.5628e-04, + 0.0000e+00, 5.2869e-05], + [ 1.2541e-04, 2.8253e-04, 3.2663e-04, ..., 8.6963e-05, + 0.0000e+00, 1.1897e-04], + [ 1.3030e-04, 4.4179e-04, 3.1859e-05, ..., -3.7861e-04, + 0.0000e+00, 2.1362e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0229, -0.0192, -0.0293, 0.0035, -0.0004, 0.0300, 0.0145, -0.0236, + 0.0154, -0.0026], device='cuda:0'), grad: tensor([-2.2766e-02, 5.3024e-04, 2.6207e-03, 5.7745e-04, 1.4488e-02, + 7.4673e-04, 2.4929e-03, -8.0466e-05, 6.0034e-04, 7.9298e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 213.93, cls_loss 0.0193 cls_loss_mapping 0.0346 cls_loss_causal 0.7724 re_mapping 0.0191 re_causal 0.0585 /// teacc 98.47 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0094, -0.0036, -0.0012, ..., -0.0659, -0.0115, -0.0235], + [ 0.0165, -0.0220, -0.0564, ..., -0.0045, -0.0147, -0.0586], + [ 0.0358, -0.0447, -0.0146, ..., 0.0251, -0.0500, 0.0187], + ..., + [ 0.0128, -0.0282, -0.0353, ..., -0.0274, 0.0095, 0.0585], + [ 0.0376, -0.0467, -0.0372, ..., -0.0210, -0.0071, -0.0063], + [-0.0821, -0.0042, 0.0575, ..., 0.0338, -0.0155, -0.0300]], + device='cuda:0'), grad: tensor([[-3.3188e-04, 3.6657e-05, 3.5018e-05, ..., 2.5541e-05, + 1.2293e-07, 3.9607e-05], + [ 7.7963e-05, 2.0936e-05, 2.4498e-05, ..., -5.8055e-05, + -4.2804e-06, 3.1054e-05], + [ 1.0437e-04, 1.0437e-04, 1.0383e-04, ..., 6.2466e-05, + 9.0618e-07, -6.1452e-05], + ..., + [ 1.1188e-04, 3.8815e-04, 5.8889e-04, ..., 5.2309e-04, + 1.1623e-06, 2.9278e-04], + [-5.9652e-04, 7.7307e-05, 9.6500e-05, ..., 8.3327e-05, + 1.0179e-06, -4.5061e-04], + [ 3.5584e-05, -6.8855e-04, -1.5043e-05, ..., -4.8208e-04, + 2.0023e-07, -3.0208e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0228, -0.0192, -0.0290, 0.0031, -0.0002, 0.0301, 0.0142, -0.0238, + 0.0155, -0.0026], device='cuda:0'), grad: tensor([-0.0004, -0.0005, 0.0004, 0.0001, -0.0009, 0.0021, 0.0002, 0.0013, + -0.0021, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 214.01, cls_loss 0.0180 cls_loss_mapping 0.0329 cls_loss_causal 0.7456 re_mapping 0.0191 re_causal 0.0553 /// teacc 98.24 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0091, -0.0040, -0.0020, ..., -0.0669, -0.0119, -0.0243], + [ 0.0167, -0.0225, -0.0568, ..., -0.0044, -0.0142, -0.0586], + [ 0.0358, -0.0456, -0.0151, ..., 0.0249, -0.0502, 0.0188], + ..., + [ 0.0127, -0.0285, -0.0361, ..., -0.0282, 0.0097, 0.0589], + [ 0.0375, -0.0473, -0.0375, ..., -0.0215, -0.0074, -0.0070], + [-0.0823, -0.0036, 0.0582, ..., 0.0349, -0.0160, -0.0304]], + device='cuda:0'), grad: tensor([[-2.9812e-03, -8.4639e-04, -2.4885e-05, ..., -7.7820e-04, + 3.2596e-08, 2.0787e-05], + [ 9.4986e-04, 5.8383e-05, 8.6308e-05, ..., 2.7966e-04, + 6.1467e-08, 4.8250e-05], + [ 3.1805e-04, 8.4281e-05, 1.3435e-04, ..., 1.4198e-04, + 2.7940e-08, -1.1396e-04], + ..., + [ 9.2030e-05, -9.4235e-05, 1.3933e-05, ..., 3.7611e-05, + -1.0710e-06, -2.3007e-04], + [-3.7837e-04, -1.3125e-04, -6.2847e-04, ..., -3.6621e-04, + 4.4703e-08, -1.6344e-04], + [ 1.1272e-03, 7.7367e-05, 1.2088e-04, ..., 9.1791e-05, + 6.6590e-07, 9.5725e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([-2.3136e-02, -1.9189e-02, -2.9389e-02, 3.4711e-03, -4.6271e-05, + 2.9720e-02, 1.4597e-02, -2.3982e-02, 1.5105e-02, -1.8812e-03], + device='cuda:0'), grad: tensor([-7.8430e-03, 1.7233e-03, 7.1955e-04, 1.3294e-03, 5.7554e-04, + 1.9150e-03, -4.5967e-04, 7.9870e-05, -1.0090e-03, 2.9659e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 214.11, cls_loss 0.0126 cls_loss_mapping 0.0262 cls_loss_causal 0.7348 re_mapping 0.0188 re_causal 0.0590 /// teacc 98.50 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0086, -0.0041, -0.0022, ..., -0.0672, -0.0120, -0.0251], + [ 0.0165, -0.0230, -0.0568, ..., -0.0041, -0.0142, -0.0584], + [ 0.0357, -0.0457, -0.0150, ..., 0.0254, -0.0503, 0.0192], + ..., + [ 0.0128, -0.0284, -0.0368, ..., -0.0294, 0.0096, 0.0594], + [ 0.0380, -0.0477, -0.0374, ..., -0.0217, -0.0078, -0.0075], + [-0.0833, -0.0037, 0.0586, ..., 0.0353, -0.0162, -0.0313]], + device='cuda:0'), grad: tensor([[ 2.9415e-05, 3.1948e-05, 4.8757e-05, ..., 1.6600e-05, + 0.0000e+00, 1.7628e-05], + [ 1.3702e-05, 4.6492e-05, -4.3362e-06, ..., -2.4348e-05, + 0.0000e+00, 6.6400e-05], + [-2.8491e-05, 2.2978e-05, 2.5213e-05, ..., 1.5393e-05, + 0.0000e+00, -2.3186e-04], + ..., + [ 8.7321e-05, 8.7857e-05, 1.5521e-04, ..., 1.2827e-04, + 0.0000e+00, 1.4925e-04], + [-2.4319e-04, 5.4121e-05, 1.1790e-04, ..., 1.3661e-04, + 0.0000e+00, -1.2970e-04], + [ 4.2766e-05, 2.7537e-05, -1.2577e-04, ..., -1.4055e-04, + 0.0000e+00, -6.0201e-06]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0226, -0.0189, -0.0292, 0.0034, -0.0004, 0.0299, 0.0146, -0.0241, + 0.0152, -0.0024], device='cuda:0'), grad: tensor([ 1.0306e-04, 8.8692e-05, -7.5638e-05, 3.9506e-04, -1.2982e-04, + -5.7489e-05, -2.1970e-04, 4.0650e-04, -5.1785e-04, 6.6422e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 214.05, cls_loss 0.0198 cls_loss_mapping 0.0337 cls_loss_causal 0.7589 re_mapping 0.0182 re_causal 0.0539 /// teacc 98.70 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0087, -0.0042, -0.0021, ..., -0.0677, -0.0121, -0.0261], + [ 0.0160, -0.0232, -0.0570, ..., -0.0042, -0.0138, -0.0588], + [ 0.0359, -0.0460, -0.0154, ..., 0.0254, -0.0508, 0.0195], + ..., + [ 0.0127, -0.0286, -0.0376, ..., -0.0305, 0.0098, 0.0601], + [ 0.0386, -0.0481, -0.0377, ..., -0.0226, -0.0083, -0.0078], + [-0.0843, -0.0036, 0.0590, ..., 0.0358, -0.0168, -0.0319]], + device='cuda:0'), grad: tensor([[ 1.8060e-04, 9.3877e-05, 3.1948e-05, ..., 2.7061e-04, + 5.6028e-06, 1.3900e-04], + [ 8.7082e-05, 1.0127e-04, 7.6532e-05, ..., 8.1360e-05, + 3.3062e-07, 2.0194e-04], + [-1.4687e-04, 7.4053e-04, 4.5925e-05, ..., 5.6386e-05, + 4.9546e-07, 1.1988e-03], + ..., + [-1.2386e-04, -5.8794e-04, 9.4771e-05, ..., 2.4945e-05, + 1.8338e-06, -6.0225e-04], + [ 7.8201e-04, 1.9300e-04, 2.8515e-04, ..., 6.1321e-04, + 1.8239e-05, 4.7708e-04], + [ 2.9898e-04, 1.2106e-04, 1.6534e-04, ..., 5.4693e-04, + 1.3404e-05, 1.7667e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([-2.2031e-02, -1.9155e-02, -2.8648e-02, 3.0503e-03, 2.1248e-06, + 2.9922e-02, 1.4117e-02, -2.4113e-02, 1.5294e-02, -2.7980e-03], + device='cuda:0'), grad: tensor([ 0.0006, 0.0005, 0.0022, -0.0032, -0.0007, -0.0022, 0.0001, -0.0007, + 0.0022, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 231.54, cls_loss 0.0162 cls_loss_mapping 0.0308 cls_loss_causal 0.7490 re_mapping 0.0184 re_causal 0.0547 /// teacc 98.71 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0086, -0.0046, -0.0026, ..., -0.0691, -0.0123, -0.0270], + [ 0.0156, -0.0234, -0.0571, ..., -0.0045, -0.0136, -0.0593], + [ 0.0360, -0.0468, -0.0160, ..., 0.0251, -0.0512, 0.0200], + ..., + [ 0.0130, -0.0281, -0.0381, ..., -0.0307, 0.0106, 0.0607], + [ 0.0387, -0.0488, -0.0383, ..., -0.0231, -0.0090, -0.0083], + [-0.0851, -0.0033, 0.0598, ..., 0.0366, -0.0178, -0.0324]], + device='cuda:0'), grad: tensor([[ 9.4831e-05, 1.0818e-04, 5.4359e-05, ..., 1.6034e-04, + 4.9360e-08, 2.9296e-05], + [-6.6519e-05, -5.9843e-04, 8.8871e-05, ..., -3.4142e-04, + -2.4792e-06, -1.1692e-03], + [ 1.0252e-04, 6.5994e-04, 6.9499e-05, ..., 3.8671e-04, + 2.1793e-07, 8.8549e-04], + ..., + [ 4.2915e-05, 1.3895e-03, 1.5926e-03, ..., 9.4175e-04, + 7.4320e-07, 7.5865e-04], + [-3.1710e-05, 1.4067e-04, 9.1612e-05, ..., 1.5306e-04, + 2.6263e-07, 6.8069e-05], + [ 6.6578e-05, -8.8751e-05, -2.0707e-04, ..., 1.1578e-03, + 1.3877e-07, -6.5374e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0223, -0.0193, -0.0287, 0.0030, -0.0001, 0.0301, 0.0142, -0.0237, + 0.0151, -0.0028], device='cuda:0'), grad: tensor([ 4.5300e-04, -3.1300e-03, 2.6703e-03, 2.6855e-03, -1.1663e-03, + -4.4708e-03, -6.0573e-06, 3.0441e-03, 1.2290e-04, -2.0337e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 214.16, cls_loss 0.0159 cls_loss_mapping 0.0280 cls_loss_causal 0.7469 re_mapping 0.0171 re_causal 0.0541 /// teacc 98.65 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0085, -0.0048, -0.0027, ..., -0.0703, -0.0119, -0.0284], + [ 0.0154, -0.0242, -0.0571, ..., -0.0044, -0.0137, -0.0605], + [ 0.0361, -0.0467, -0.0163, ..., 0.0254, -0.0515, 0.0209], + ..., + [ 0.0129, -0.0283, -0.0390, ..., -0.0320, 0.0106, 0.0610], + [ 0.0386, -0.0494, -0.0388, ..., -0.0238, -0.0093, -0.0092], + [-0.0851, -0.0032, 0.0605, ..., 0.0372, -0.0175, -0.0327]], + device='cuda:0'), grad: tensor([[-1.4476e-05, 2.7135e-05, 1.3225e-06, ..., -2.3007e-05, + 1.2573e-07, 5.2929e-05], + [ 1.7965e-04, 6.4611e-05, 4.1932e-05, ..., 1.2118e-04, + -4.9472e-06, 3.7813e-04], + [-9.5606e-04, -4.0960e-04, 1.0926e-04, ..., -6.5148e-05, + 5.5414e-07, -2.4834e-03], + ..., + [ 6.5804e-04, 3.1161e-04, 5.9605e-05, ..., 1.2681e-05, + 2.0601e-06, 1.7309e-03], + [-1.0389e-04, 3.6567e-05, -1.0383e-04, ..., -4.6641e-05, + 3.9302e-07, 6.1929e-05], + [ 6.6578e-05, 4.5180e-05, -1.6883e-05, ..., -3.4243e-05, + 4.2841e-07, 9.5606e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0225, -0.0193, -0.0283, 0.0031, -0.0002, 0.0306, 0.0144, -0.0241, + 0.0146, -0.0026], device='cuda:0'), grad: tensor([ 6.1810e-05, 6.3934e-03, -1.0025e-02, 1.0943e-04, -8.7690e-04, + 6.4492e-05, 1.0653e-03, 2.9449e-03, 2.9936e-05, 2.3139e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 213.85, cls_loss 0.0136 cls_loss_mapping 0.0258 cls_loss_causal 0.7419 re_mapping 0.0171 re_causal 0.0535 /// teacc 98.66 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0079, -0.0050, -0.0034, ..., -0.0707, -0.0119, -0.0287], + [ 0.0153, -0.0252, -0.0578, ..., -0.0047, -0.0136, -0.0610], + [ 0.0361, -0.0478, -0.0169, ..., 0.0253, -0.0517, 0.0211], + ..., + [ 0.0128, -0.0278, -0.0392, ..., -0.0322, 0.0106, 0.0613], + [ 0.0391, -0.0498, -0.0388, ..., -0.0239, -0.0095, -0.0094], + [-0.0856, -0.0032, 0.0612, ..., 0.0377, -0.0175, -0.0334]], + device='cuda:0'), grad: tensor([[ 9.5010e-05, 3.4064e-05, 2.5660e-05, ..., 2.7239e-05, + 0.0000e+00, 1.1988e-05], + [-5.0163e-04, 1.5162e-05, 6.8247e-05, ..., 8.8990e-05, + 0.0000e+00, 8.0392e-06], + [ 7.4506e-05, 3.0503e-05, 1.3866e-05, ..., 1.2882e-05, + 0.0000e+00, 6.9253e-06], + ..., + [-9.7752e-06, 4.8757e-05, 9.8288e-05, ..., 1.1498e-04, + 0.0000e+00, -1.1653e-05], + [ 2.1815e-04, 9.5904e-05, 7.9393e-05, ..., 9.0599e-05, + 0.0000e+00, -3.0249e-05], + [ 1.3009e-05, -1.0328e-03, -1.2617e-03, ..., -1.7185e-03, + 0.0000e+00, 8.3521e-06]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0223, -0.0195, -0.0284, 0.0031, -0.0001, 0.0304, 0.0143, -0.0239, + 0.0149, -0.0028], device='cuda:0'), grad: tensor([ 2.8729e-04, -1.6346e-03, 2.4891e-04, 2.0492e-04, 1.0166e-03, + 2.6584e-04, 6.4194e-05, 1.3328e-04, 1.0672e-03, -1.6537e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 214.29, cls_loss 0.0153 cls_loss_mapping 0.0290 cls_loss_causal 0.6821 re_mapping 0.0179 re_causal 0.0501 /// teacc 98.64 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0077, -0.0048, -0.0037, ..., -0.0714, -0.0119, -0.0292], + [ 0.0153, -0.0259, -0.0577, ..., -0.0042, -0.0135, -0.0613], + [ 0.0362, -0.0488, -0.0173, ..., 0.0254, -0.0519, 0.0211], + ..., + [ 0.0129, -0.0273, -0.0397, ..., -0.0327, 0.0107, 0.0619], + [ 0.0398, -0.0505, -0.0393, ..., -0.0245, -0.0096, -0.0097], + [-0.0858, -0.0032, 0.0616, ..., 0.0380, -0.0176, -0.0339]], + device='cuda:0'), grad: tensor([[-3.2455e-05, 2.2337e-05, 1.9804e-05, ..., 1.6049e-05, + 0.0000e+00, 7.2606e-06], + [ 3.9525e-06, 2.5183e-05, -9.2089e-06, ..., -4.1425e-06, + 0.0000e+00, 1.7852e-05], + [ 5.0753e-05, 4.3273e-05, 4.7028e-05, ..., 1.6764e-05, + 0.0000e+00, 2.5809e-05], + ..., + [-1.5795e-05, 7.2479e-05, 6.1929e-05, ..., 3.6001e-05, + 0.0000e+00, 2.3842e-05], + [ 6.1035e-04, 2.5225e-04, 1.6725e-04, ..., 4.3988e-04, + 0.0000e+00, 1.6677e-04], + [ 5.0813e-05, 5.4169e-04, 2.2049e-03, ..., 6.0177e-04, + 0.0000e+00, 4.6968e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([-2.2394e-02, -1.8976e-02, -2.8686e-02, 3.1184e-03, -9.4069e-05, + 3.0615e-02, 1.3077e-02, -2.3668e-02, 1.5094e-02, -2.7784e-03], + device='cuda:0'), grad: tensor([-6.1810e-05, -1.3605e-05, 1.7381e-04, -6.3419e-04, -2.3460e-03, + -3.1233e-04, -3.5143e-04, 1.0681e-04, 1.2264e-03, 2.2125e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 214.16, cls_loss 0.0117 cls_loss_mapping 0.0244 cls_loss_causal 0.7297 re_mapping 0.0166 re_causal 0.0510 /// teacc 98.69 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0077, -0.0051, -0.0041, ..., -0.0722, -0.0120, -0.0296], + [ 0.0151, -0.0269, -0.0578, ..., -0.0047, -0.0133, -0.0626], + [ 0.0358, -0.0488, -0.0179, ..., 0.0260, -0.0522, 0.0215], + ..., + [ 0.0131, -0.0279, -0.0403, ..., -0.0333, 0.0109, 0.0621], + [ 0.0395, -0.0513, -0.0401, ..., -0.0258, -0.0098, -0.0099], + [-0.0859, -0.0030, 0.0621, ..., 0.0385, -0.0181, -0.0344]], + device='cuda:0'), grad: tensor([[-5.6356e-05, 1.1198e-05, 2.2426e-05, ..., 1.0706e-05, + 4.3772e-08, 2.6315e-05], + [ 6.3956e-05, 1.0014e-05, -1.8820e-05, ..., -2.1532e-05, + 1.1269e-07, 2.1264e-05], + [-1.3137e-04, 1.6168e-05, 9.7007e-06, ..., 1.1936e-05, + 2.6077e-08, -1.1820e-04], + ..., + [ 1.0067e-04, 9.4846e-06, 3.1620e-05, ..., 4.4078e-05, + -5.7649e-07, 5.4806e-05], + [ 2.3019e-04, 3.8207e-05, 1.2040e-04, ..., 2.4402e-04, + 2.9802e-08, 1.3903e-05], + [ 2.4274e-05, -1.8620e-04, -8.3351e-04, ..., -4.0340e-04, + 1.1828e-07, 7.3873e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0225, -0.0194, -0.0288, 0.0035, 0.0006, 0.0307, 0.0135, -0.0237, + 0.0144, -0.0029], device='cuda:0'), grad: tensor([-5.0366e-05, 7.9155e-05, -1.8239e-04, 1.2469e-04, 3.2020e-04, + -9.2840e-04, 7.2479e-05, 2.1577e-04, 7.5817e-04, -4.0913e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.13, cls_loss 0.0152 cls_loss_mapping 0.0314 cls_loss_causal 0.6925 re_mapping 0.0162 re_causal 0.0483 /// teacc 98.66 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0079, -0.0056, -0.0048, ..., -0.0739, -0.0122, -0.0308], + [ 0.0147, -0.0273, -0.0574, ..., -0.0043, -0.0131, -0.0631], + [ 0.0358, -0.0491, -0.0184, ..., 0.0258, -0.0524, 0.0217], + ..., + [ 0.0137, -0.0279, -0.0415, ..., -0.0345, 0.0111, 0.0628], + [ 0.0396, -0.0520, -0.0403, ..., -0.0263, -0.0099, -0.0102], + [-0.0861, -0.0027, 0.0628, ..., 0.0392, -0.0184, -0.0353]], + device='cuda:0'), grad: tensor([[-8.7690e-04, -7.0751e-05, 1.0008e-04, ..., -4.5359e-05, + 0.0000e+00, 8.0168e-05], + [-3.0100e-05, -3.6640e-03, -5.1594e-04, ..., -1.1513e-02, + 0.0000e+00, -4.8523e-03], + [ 7.8857e-05, 4.1634e-05, 2.5973e-05, ..., 5.3614e-05, + 0.0000e+00, -1.0309e-03], + ..., + [ 1.1426e-04, 3.6926e-03, 6.5231e-04, ..., 1.1200e-02, + 0.0000e+00, 5.3558e-03], + [ 1.8859e-04, 2.1827e-04, 3.4666e-04, ..., 3.5977e-04, + 0.0000e+00, 1.0061e-04], + [ 9.8825e-05, -5.3740e-04, -9.5701e-04, ..., -9.5320e-04, + 0.0000e+00, 6.4671e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0232, -0.0197, -0.0292, 0.0038, 0.0002, 0.0306, 0.0139, -0.0230, + 0.0146, -0.0029], device='cuda:0'), grad: tensor([-0.0007, -0.0267, -0.0010, 0.0005, 0.0006, 0.0011, -0.0006, 0.0267, + 0.0009, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 43---------------------------------------------------- +epoch 43, time 230.58, cls_loss 0.0144 cls_loss_mapping 0.0287 cls_loss_causal 0.7013 re_mapping 0.0171 re_causal 0.0477 /// teacc 98.72 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0076, -0.0061, -0.0043, ..., -0.0747, -0.0125, -0.0312], + [ 0.0143, -0.0277, -0.0579, ..., -0.0040, -0.0130, -0.0634], + [ 0.0359, -0.0498, -0.0189, ..., 0.0254, -0.0528, 0.0224], + ..., + [ 0.0140, -0.0283, -0.0423, ..., -0.0361, 0.0113, 0.0631], + [ 0.0398, -0.0528, -0.0406, ..., -0.0269, -0.0106, -0.0111], + [-0.0871, -0.0025, 0.0632, ..., 0.0398, -0.0186, -0.0358]], + device='cuda:0'), grad: tensor([[ 2.2709e-05, 3.4813e-06, 6.6124e-06, ..., 2.0757e-05, + 7.9162e-09, 9.6336e-06], + [-1.3518e-04, 7.1786e-06, -3.6299e-05, ..., -2.7990e-04, + -1.9697e-07, 3.5739e-04], + [-1.2636e-04, 8.3297e-06, 9.6411e-06, ..., 7.6175e-05, + 1.7229e-08, -4.8208e-04], + ..., + [ 9.0659e-05, -1.4842e-05, 1.1854e-05, ..., 7.8082e-05, + 3.9116e-08, 3.9339e-05], + [-3.2830e-04, 3.4869e-05, 3.2008e-05, ..., -1.1134e-04, + 4.2841e-08, 5.9128e-05], + [ 4.8041e-05, 1.2815e-04, 1.7062e-06, ..., -4.7311e-06, + 1.8626e-08, 1.1104e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0233, -0.0196, -0.0291, 0.0038, 0.0002, 0.0311, 0.0145, -0.0235, + 0.0144, -0.0030], device='cuda:0'), grad: tensor([ 1.0145e-04, -3.9124e-04, -7.8869e-04, -2.1017e-04, 1.5438e-04, + 4.9162e-04, 6.4313e-05, 4.3631e-04, -2.0707e-04, 3.4928e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 44---------------------------------------------------- +epoch 44, time 230.17, cls_loss 0.0106 cls_loss_mapping 0.0179 cls_loss_causal 0.6893 re_mapping 0.0156 re_causal 0.0468 /// teacc 98.75 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0073, -0.0065, -0.0044, ..., -0.0751, -0.0125, -0.0314], + [ 0.0144, -0.0281, -0.0576, ..., -0.0037, -0.0128, -0.0636], + [ 0.0360, -0.0502, -0.0193, ..., 0.0266, -0.0530, 0.0230], + ..., + [ 0.0141, -0.0284, -0.0429, ..., -0.0369, 0.0113, 0.0633], + [ 0.0400, -0.0532, -0.0411, ..., -0.0280, -0.0107, -0.0117], + [-0.0878, -0.0021, 0.0636, ..., 0.0404, -0.0187, -0.0364]], + device='cuda:0'), grad: tensor([[-5.2834e-04, 3.4630e-05, 6.2823e-05, ..., 3.6269e-05, + 1.2040e-05, -1.2314e-04], + [-2.6211e-05, -1.4410e-03, -1.4553e-03, ..., -1.9112e-03, + -8.9502e-04, 1.0782e-04], + [ 1.5569e-04, 7.1704e-05, 1.0139e-04, ..., 2.3782e-05, + 8.1286e-06, -4.7714e-05], + ..., + [ 1.2070e-04, 1.2553e-04, 1.8394e-04, ..., 1.6248e-04, + 1.3314e-05, 6.5565e-05], + [ 6.3992e-04, 4.0650e-05, 2.0671e-04, ..., 5.7667e-05, + 6.1989e-06, 2.7150e-05], + [ 4.5389e-05, -2.9302e-04, -3.9768e-04, ..., -3.4046e-04, + 7.0870e-05, 1.9148e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0233, -0.0189, -0.0290, 0.0040, 0.0002, 0.0308, 0.0143, -0.0237, + 0.0141, -0.0031], device='cuda:0'), grad: tensor([-4.6778e-04, -6.8550e-03, 3.7909e-04, 5.9280e-03, 8.7357e-04, + 1.1063e-03, -2.0905e-03, 2.7865e-05, 1.2207e-03, -1.1665e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 213.99, cls_loss 0.0122 cls_loss_mapping 0.0238 cls_loss_causal 0.6697 re_mapping 0.0168 re_causal 0.0460 /// teacc 98.75 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0072, -0.0067, -0.0052, ..., -0.0758, -0.0126, -0.0318], + [ 0.0144, -0.0282, -0.0571, ..., -0.0027, -0.0123, -0.0646], + [ 0.0361, -0.0506, -0.0200, ..., 0.0261, -0.0531, 0.0237], + ..., + [ 0.0139, -0.0289, -0.0435, ..., -0.0378, 0.0113, 0.0631], + [ 0.0402, -0.0538, -0.0417, ..., -0.0288, -0.0107, -0.0123], + [-0.0878, -0.0018, 0.0646, ..., 0.0413, -0.0188, -0.0368]], + device='cuda:0'), grad: tensor([[ 9.7528e-06, 3.7253e-06, 3.7760e-05, ..., 5.4464e-06, + 0.0000e+00, 6.5528e-06], + [ 2.6274e-04, 1.0133e-04, 5.8534e-07, ..., 9.6083e-05, + 0.0000e+00, 1.1295e-04], + [ 3.7372e-05, 1.6600e-05, 1.3396e-05, ..., 6.7502e-06, + 0.0000e+00, 3.1948e-05], + ..., + [-1.2803e-04, -2.5988e-04, 5.5954e-06, ..., -1.1164e-04, + 0.0000e+00, -4.3559e-04], + [ 8.0168e-05, 1.8537e-05, 2.3112e-05, ..., 4.2558e-05, + 0.0000e+00, 2.3589e-05], + [ 2.8953e-05, 1.7047e-05, -1.7732e-06, ..., 1.5207e-05, + 0.0000e+00, 3.2693e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-2.3800e-02, -1.8677e-02, -2.9266e-02, 4.2510e-03, -5.9817e-06, + 3.0865e-02, 1.4344e-02, -2.3651e-02, 1.3706e-02, -2.5278e-03], + device='cuda:0'), grad: tensor([ 1.0431e-05, 9.7847e-04, 1.1820e-04, 2.0504e-03, 1.9050e-04, + -2.5692e-03, 5.2631e-05, -1.1597e-03, 2.0504e-04, 1.2350e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 214.10, cls_loss 0.0105 cls_loss_mapping 0.0225 cls_loss_causal 0.6822 re_mapping 0.0154 re_causal 0.0463 /// teacc 98.70 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0069, -0.0069, -0.0050, ..., -0.0763, -0.0127, -0.0317], + [ 0.0139, -0.0287, -0.0576, ..., -0.0029, -0.0123, -0.0657], + [ 0.0361, -0.0510, -0.0204, ..., 0.0260, -0.0531, 0.0244], + ..., + [ 0.0139, -0.0290, -0.0443, ..., -0.0388, 0.0113, 0.0633], + [ 0.0410, -0.0535, -0.0401, ..., -0.0275, -0.0113, -0.0122], + [-0.0887, -0.0019, 0.0644, ..., 0.0410, -0.0188, -0.0377]], + device='cuda:0'), grad: tensor([[ 3.0637e-05, 3.1203e-05, 9.5963e-06, ..., 5.7071e-06, + 1.1455e-07, 6.5029e-05], + [ 2.4304e-05, 1.5807e-04, -1.6272e-05, ..., -2.0057e-05, + 4.7963e-08, 4.2129e-04], + [ 5.0306e-05, 7.6950e-05, 1.0841e-05, ..., 6.1914e-06, + 1.8626e-08, 1.4997e-04], + ..., + [-1.4079e-04, -3.7456e-04, 2.1443e-05, ..., 2.2680e-05, + 1.0245e-07, -1.0290e-03], + [ 5.0259e-04, 1.1623e-04, 2.0012e-05, ..., 3.0422e-04, + 7.7626e-07, 2.8157e-04], + [ 2.2963e-05, 2.6822e-05, 1.8314e-05, ..., -1.6198e-05, + 1.9604e-07, 9.1434e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0236, -0.0193, -0.0291, 0.0041, 0.0003, 0.0309, 0.0141, -0.0239, + 0.0154, -0.0036], device='cuda:0'), grad: tensor([ 1.4925e-04, 8.4066e-04, 3.5715e-04, 5.6744e-05, 5.9336e-05, + -5.0879e-04, 2.1085e-05, -2.2564e-03, 1.0815e-03, 2.0051e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 47---------------------------------------------------- +epoch 47, time 230.32, cls_loss 0.0105 cls_loss_mapping 0.0213 cls_loss_causal 0.6823 re_mapping 0.0146 re_causal 0.0434 /// teacc 98.77 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0072, -0.0073, -0.0051, ..., -0.0772, -0.0127, -0.0319], + [ 0.0132, -0.0290, -0.0568, ..., -0.0016, -0.0124, -0.0668], + [ 0.0364, -0.0515, -0.0211, ..., 0.0257, -0.0533, 0.0254], + ..., + [ 0.0141, -0.0287, -0.0449, ..., -0.0397, 0.0120, 0.0635], + [ 0.0411, -0.0539, -0.0408, ..., -0.0288, -0.0114, -0.0126], + [-0.0896, -0.0021, 0.0646, ..., 0.0412, -0.0196, -0.0381]], + device='cuda:0'), grad: tensor([[ 8.8364e-06, 6.6161e-06, 1.3612e-05, ..., 1.9640e-05, + 3.3528e-08, 1.8463e-05], + [ 2.6181e-05, 3.2902e-05, 6.1952e-06, ..., 9.3877e-06, + 2.6543e-08, 4.4554e-05], + [-6.2168e-05, 2.9549e-05, 1.2130e-05, ..., -7.6652e-05, + 2.9802e-08, -8.4341e-05], + ..., + [ 7.8678e-06, 1.2808e-05, 1.8090e-05, ..., 3.3170e-05, + -6.5239e-07, 2.1368e-05], + [-1.1337e-04, 2.5958e-05, -9.2924e-05, ..., -6.1929e-05, + 2.7940e-08, -2.4736e-05], + [ 9.5963e-05, -1.7002e-05, 3.4124e-05, ..., 1.3679e-05, + 8.7544e-08, 5.5879e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0239, -0.0192, -0.0289, 0.0041, 0.0004, 0.0311, 0.0148, -0.0239, + 0.0150, -0.0041], device='cuda:0'), grad: tensor([ 4.5329e-05, 8.3148e-05, -9.0480e-05, -8.1301e-04, 1.5903e-04, + 7.0047e-04, -8.1599e-05, 7.3671e-05, -2.8253e-04, 2.0599e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 48---------------------------------------------------- +epoch 48, time 230.15, cls_loss 0.0099 cls_loss_mapping 0.0186 cls_loss_causal 0.6774 re_mapping 0.0141 re_causal 0.0444 /// teacc 98.78 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0067, -0.0075, -0.0051, ..., -0.0778, -0.0125, -0.0325], + [ 0.0128, -0.0295, -0.0574, ..., -0.0020, -0.0124, -0.0673], + [ 0.0368, -0.0516, -0.0217, ..., 0.0259, -0.0535, 0.0261], + ..., + [ 0.0142, -0.0285, -0.0452, ..., -0.0401, 0.0118, 0.0640], + [ 0.0411, -0.0545, -0.0414, ..., -0.0297, -0.0117, -0.0136], + [-0.0896, -0.0019, 0.0651, ..., 0.0421, -0.0197, -0.0385]], + device='cuda:0'), grad: tensor([[ 1.0139e-04, 3.1620e-05, 3.1829e-05, ..., 6.9141e-05, + 7.4506e-09, 2.5198e-05], + [ 1.5593e-04, 3.6091e-05, -1.6987e-05, ..., 2.4378e-05, + 3.9581e-08, 2.7016e-05], + [ 5.6624e-05, 1.0812e-04, 2.6971e-05, ..., 2.0072e-05, + 2.4214e-08, 4.5538e-05], + ..., + [ 1.2493e-04, 1.0431e-04, 6.4731e-05, ..., 7.3075e-05, + -2.3423e-07, 8.3864e-05], + [ 3.6030e-03, 1.0341e-04, 1.4296e-03, ..., 2.5597e-03, + 8.8476e-09, 2.8878e-05], + [ 1.2362e-04, 2.6196e-05, -9.1791e-05, ..., -1.4257e-04, + 9.6858e-08, 7.5400e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0235, -0.0196, -0.0286, 0.0039, 0.0005, 0.0309, 0.0149, -0.0235, + 0.0146, -0.0040], device='cuda:0'), grad: tensor([ 0.0002, 0.0002, 0.0003, -0.0007, 0.0002, -0.0105, 0.0047, 0.0005, + 0.0049, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 49---------------------------------------------------- +epoch 49, time 229.22, cls_loss 0.0111 cls_loss_mapping 0.0212 cls_loss_causal 0.7217 re_mapping 0.0143 re_causal 0.0444 /// teacc 98.85 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0066, -0.0080, -0.0057, ..., -0.0789, -0.0126, -0.0336], + [ 0.0127, -0.0295, -0.0574, ..., -0.0015, -0.0124, -0.0664], + [ 0.0370, -0.0525, -0.0220, ..., 0.0258, -0.0544, 0.0261], + ..., + [ 0.0141, -0.0284, -0.0459, ..., -0.0413, 0.0122, 0.0647], + [ 0.0409, -0.0551, -0.0420, ..., -0.0304, -0.0118, -0.0143], + [-0.0903, -0.0013, 0.0659, ..., 0.0429, -0.0198, -0.0392]], + device='cuda:0'), grad: tensor([[ 3.1054e-05, 3.5673e-05, 1.1586e-05, ..., 1.6108e-05, + 0.0000e+00, 7.8440e-05], + [ 5.5343e-05, 1.3399e-04, -1.4000e-05, ..., -1.6823e-05, + 0.0000e+00, 2.4176e-04], + [-1.4201e-05, 1.7107e-04, 2.3380e-05, ..., 3.7074e-05, + 0.0000e+00, -6.0380e-05], + ..., + [-1.6868e-04, -2.0254e-04, 6.0707e-05, ..., 6.3479e-05, + 0.0000e+00, -4.5609e-04], + [ 1.9348e-04, 5.6601e-04, -2.9504e-05, ..., 9.0837e-05, + 0.0000e+00, 1.2894e-03], + [ 8.2552e-05, 9.9361e-05, 5.8442e-05, ..., 9.1314e-05, + 0.0000e+00, 1.1617e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0236, -0.0189, -0.0290, 0.0038, 0.0005, 0.0309, 0.0153, -0.0234, + 0.0140, -0.0037], device='cuda:0'), grad: tensor([ 0.0002, 0.0003, 0.0002, -0.0028, -0.0002, 0.0004, 0.0002, -0.0006, + 0.0020, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 214.56, cls_loss 0.0101 cls_loss_mapping 0.0195 cls_loss_causal 0.6924 re_mapping 0.0143 re_causal 0.0439 /// teacc 98.79 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0060, -0.0083, -0.0063, ..., -0.0805, -0.0129, -0.0341], + [ 0.0123, -0.0306, -0.0581, ..., -0.0023, -0.0122, -0.0672], + [ 0.0370, -0.0533, -0.0226, ..., 0.0257, -0.0553, 0.0267], + ..., + [ 0.0138, -0.0286, -0.0466, ..., -0.0419, 0.0121, 0.0649], + [ 0.0411, -0.0557, -0.0425, ..., -0.0310, -0.0122, -0.0150], + [-0.0907, -0.0011, 0.0665, ..., 0.0434, -0.0203, -0.0397]], + device='cuda:0'), grad: tensor([[ 4.7183e-04, 1.1927e-04, 4.1151e-04, ..., 2.6733e-05, + 1.0006e-05, 1.3983e-04], + [ 3.5954e-04, 4.5866e-05, 3.9291e-04, ..., -1.9634e-04, + -2.3380e-05, 1.7881e-04], + [ 4.0317e-04, 4.6110e-04, 4.5109e-04, ..., 3.8236e-05, + 6.1810e-05, 9.7084e-04], + ..., + [-1.8001e-05, -1.2884e-03, 4.1366e-05, ..., 3.5912e-05, + -2.0742e-04, -2.7199e-03], + [-4.4614e-05, -2.0102e-05, 9.4235e-05, ..., 8.3268e-05, + 4.1313e-06, 3.3826e-05], + [ 6.9380e-05, 2.8586e-04, 2.1234e-05, ..., 9.7752e-05, + 4.8637e-05, 5.8079e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0236, -0.0196, -0.0289, 0.0041, 0.0011, 0.0308, 0.0151, -0.0237, + 0.0137, -0.0033], device='cuda:0'), grad: tensor([ 9.8324e-04, 9.8419e-04, 2.4910e-03, 7.7772e-04, 4.0579e-04, + 6.2048e-05, -3.4218e-03, -3.3455e-03, 1.2481e-04, 9.3603e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 51---------------------------------------------------- +epoch 51, time 230.52, cls_loss 0.0104 cls_loss_mapping 0.0165 cls_loss_causal 0.6852 re_mapping 0.0141 re_causal 0.0411 /// teacc 98.88 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0061, -0.0090, -0.0071, ..., -0.0816, -0.0132, -0.0348], + [ 0.0120, -0.0310, -0.0581, ..., -0.0019, -0.0128, -0.0680], + [ 0.0371, -0.0537, -0.0234, ..., 0.0258, -0.0566, 0.0273], + ..., + [ 0.0137, -0.0282, -0.0469, ..., -0.0429, 0.0132, 0.0658], + [ 0.0417, -0.0564, -0.0432, ..., -0.0318, -0.0128, -0.0158], + [-0.0915, -0.0010, 0.0671, ..., 0.0437, -0.0206, -0.0404]], + device='cuda:0'), grad: tensor([[-3.6097e-04, -9.4414e-05, 3.7123e-06, ..., 3.0249e-06, + 0.0000e+00, -1.6773e-04], + [ 1.7211e-05, -7.1861e-06, -6.5506e-05, ..., -7.0572e-05, + 0.0000e+00, 1.2927e-05], + [ 9.2089e-05, 3.4630e-05, 1.2942e-05, ..., 1.0446e-05, + 0.0000e+00, -1.3420e-06], + ..., + [ 4.2111e-05, 1.8358e-05, 3.6836e-05, ..., 3.6240e-05, + 0.0000e+00, 2.8938e-05], + [ 1.7434e-05, 9.8124e-06, 2.6122e-05, ..., 3.1233e-05, + 0.0000e+00, 1.2510e-05], + [ 7.1049e-05, 2.6608e-04, 6.1798e-04, ..., 4.4394e-04, + 0.0000e+00, 3.6538e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0240, -0.0200, -0.0289, 0.0037, 0.0014, 0.0309, 0.0154, -0.0230, + 0.0139, -0.0037], device='cuda:0'), grad: tensor([-5.9557e-04, -1.1724e-04, 1.4031e-04, 1.0079e-04, -6.2466e-04, + 5.3495e-05, 6.6996e-05, 1.4436e-04, 7.0035e-05, 7.6103e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 214.05, cls_loss 0.0112 cls_loss_mapping 0.0222 cls_loss_causal 0.6638 re_mapping 0.0136 re_causal 0.0424 /// teacc 98.76 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0062, -0.0090, -0.0079, ..., -0.0823, -0.0133, -0.0348], + [ 0.0118, -0.0310, -0.0582, ..., -0.0018, -0.0121, -0.0677], + [ 0.0371, -0.0542, -0.0238, ..., 0.0257, -0.0568, 0.0278], + ..., + [ 0.0136, -0.0281, -0.0475, ..., -0.0427, 0.0131, 0.0656], + [ 0.0419, -0.0569, -0.0437, ..., -0.0327, -0.0131, -0.0158], + [-0.0916, -0.0008, 0.0674, ..., 0.0438, -0.0210, -0.0408]], + device='cuda:0'), grad: tensor([[-3.4958e-05, 2.3931e-05, 5.8040e-06, ..., 7.1935e-06, + 3.0734e-08, 3.8713e-05], + [ 3.2842e-05, 5.8502e-05, -4.9919e-05, ..., -6.9678e-05, + -2.6897e-06, 1.3340e-04], + [-6.5804e-05, -1.9386e-05, 1.4193e-05, ..., 1.5087e-05, + 3.5018e-07, -1.9753e-04], + ..., + [ 1.2361e-05, -1.7309e-04, 2.1458e-05, ..., 2.0832e-05, + 8.4564e-07, -2.2995e-04], + [ 2.2620e-05, 3.7104e-05, 1.3322e-05, ..., 1.5974e-05, + 2.2165e-07, 6.3181e-05], + [ 1.9580e-05, 1.8775e-04, 8.1837e-05, ..., 1.6534e-04, + 6.7428e-07, 1.3709e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0243, -0.0200, -0.0290, 0.0041, 0.0016, 0.0306, 0.0152, -0.0229, + 0.0139, -0.0037], device='cuda:0'), grad: tensor([-1.0359e-04, 7.5758e-05, -1.4889e-04, 4.0025e-05, 1.6510e-05, + -1.4293e-04, 2.4468e-05, -2.8229e-04, 1.1539e-04, 4.0531e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 213.79, cls_loss 0.0089 cls_loss_mapping 0.0185 cls_loss_causal 0.6885 re_mapping 0.0144 re_causal 0.0439 /// teacc 98.77 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0060, -0.0087, -0.0079, ..., -0.0829, -0.0133, -0.0348], + [ 0.0113, -0.0310, -0.0579, ..., -0.0014, -0.0116, -0.0682], + [ 0.0374, -0.0545, -0.0242, ..., 0.0256, -0.0570, 0.0281], + ..., + [ 0.0138, -0.0278, -0.0480, ..., -0.0432, 0.0128, 0.0663], + [ 0.0421, -0.0573, -0.0436, ..., -0.0327, -0.0131, -0.0163], + [-0.0922, -0.0013, 0.0674, ..., 0.0437, -0.0211, -0.0420]], + device='cuda:0'), grad: tensor([[-4.1723e-06, 2.6859e-06, 5.7854e-06, ..., 2.0303e-06, + 0.0000e+00, 3.1888e-06], + [-1.3553e-05, -7.6234e-05, -1.8930e-04, ..., -1.9479e-04, + 0.0000e+00, -2.0429e-05], + [ 2.8506e-05, 1.2122e-05, 2.3276e-05, ..., 1.1660e-05, + 0.0000e+00, 3.3844e-06], + ..., + [ 1.7164e-06, 3.2604e-05, 5.8234e-05, ..., 6.1393e-05, + 0.0000e+00, 8.0243e-06], + [-2.9698e-05, 1.7107e-05, 1.7449e-05, ..., 1.6540e-05, + 0.0000e+00, 1.1332e-05], + [ 7.0520e-06, 1.3244e-04, -1.3880e-05, ..., 3.2991e-05, + 0.0000e+00, 1.3971e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0239, -0.0199, -0.0290, 0.0045, 0.0018, 0.0302, 0.0154, -0.0227, + 0.0141, -0.0045], device='cuda:0'), grad: tensor([ 1.3486e-06, -3.7694e-04, 7.8917e-05, -3.1948e-04, 4.8971e-04, + 5.4896e-05, -3.1424e-04, 1.1718e-04, 1.1995e-05, 2.5558e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 214.22, cls_loss 0.0094 cls_loss_mapping 0.0185 cls_loss_causal 0.6649 re_mapping 0.0130 re_causal 0.0411 /// teacc 98.80 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0055, -0.0095, -0.0080, ..., -0.0839, -0.0134, -0.0358], + [ 0.0111, -0.0317, -0.0575, ..., -0.0017, -0.0111, -0.0690], + [ 0.0376, -0.0544, -0.0248, ..., 0.0262, -0.0582, 0.0291], + ..., + [ 0.0136, -0.0282, -0.0498, ..., -0.0445, 0.0129, 0.0665], + [ 0.0418, -0.0582, -0.0444, ..., -0.0337, -0.0135, -0.0173], + [-0.0924, -0.0003, 0.0686, ..., 0.0451, -0.0213, -0.0419]], + device='cuda:0'), grad: tensor([[ 6.8665e-04, 6.3610e-04, 7.7629e-04, ..., 6.0606e-04, + 9.2201e-08, 7.7426e-05], + [-4.7863e-05, 1.8090e-05, -9.1910e-05, ..., -5.5015e-05, + -3.5316e-06, 4.2319e-06], + [ 1.2887e-04, 5.5015e-05, 1.3006e-04, ..., 1.0967e-04, + 5.1036e-07, 1.2234e-05], + ..., + [ 9.7156e-05, 6.5386e-05, 6.2704e-05, ..., 1.0538e-04, + 8.9779e-07, 3.1237e-06], + [ 8.7214e-04, 1.1740e-03, 1.4534e-03, ..., 9.0170e-04, + 7.0734e-07, 1.4293e-04], + [-2.4395e-03, -3.5496e-03, -4.8141e-03, ..., -2.4872e-03, + 4.1770e-07, -4.6945e-04]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0235, -0.0194, -0.0289, 0.0039, 0.0017, 0.0309, 0.0150, -0.0236, + 0.0132, -0.0035], device='cuda:0'), grad: tensor([ 0.0016, -0.0005, 0.0004, 0.0015, 0.0003, 0.0014, -0.0006, 0.0002, + 0.0024, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 214.27, cls_loss 0.0088 cls_loss_mapping 0.0170 cls_loss_causal 0.6533 re_mapping 0.0132 re_causal 0.0391 /// teacc 98.85 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0052, -0.0099, -0.0083, ..., -0.0846, -0.0135, -0.0367], + [ 0.0113, -0.0318, -0.0569, ..., -0.0011, -0.0109, -0.0691], + [ 0.0376, -0.0551, -0.0255, ..., 0.0259, -0.0592, 0.0296], + ..., + [ 0.0134, -0.0280, -0.0507, ..., -0.0448, 0.0136, 0.0669], + [ 0.0422, -0.0588, -0.0444, ..., -0.0341, -0.0141, -0.0178], + [-0.0929, 0.0003, 0.0694, ..., 0.0458, -0.0223, -0.0424]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 7.5437e-06, 1.0312e-05, ..., 5.2303e-06, + 3.0128e-07, 8.8140e-06], + [-6.3367e-06, 1.5244e-05, -2.5630e-04, ..., -1.2386e-04, + -1.5162e-05, 3.0965e-05], + [-3.2306e-05, 1.1869e-05, 4.0561e-05, ..., 2.3380e-05, + 1.6736e-06, -9.1434e-05], + ..., + [-3.0264e-05, -1.0794e-04, 6.5625e-05, ..., 4.4346e-05, + 6.7055e-06, -5.4955e-05], + [ 1.2077e-05, 2.1160e-05, 1.2195e-04, ..., 5.7250e-05, + 1.4547e-06, 2.4080e-05], + [ 3.0786e-05, 4.6968e-05, 3.2306e-05, ..., 1.3247e-05, + 1.9111e-06, 5.1558e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0235, -0.0190, -0.0292, 0.0038, 0.0011, 0.0306, 0.0149, -0.0232, + 0.0133, -0.0033], device='cuda:0'), grad: tensor([ 3.1501e-05, -5.8126e-04, 1.7751e-06, -1.7154e-04, 8.6576e-06, + 2.5678e-04, 5.7548e-05, -1.6963e-04, 3.2401e-04, 2.4199e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 214.01, cls_loss 0.0100 cls_loss_mapping 0.0169 cls_loss_causal 0.6736 re_mapping 0.0134 re_causal 0.0392 /// teacc 98.78 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0055, -0.0102, -0.0085, ..., -0.0853, -0.0137, -0.0372], + [ 0.0112, -0.0322, -0.0569, ..., -0.0013, -0.0106, -0.0692], + [ 0.0375, -0.0556, -0.0260, ..., 0.0258, -0.0599, 0.0302], + ..., + [ 0.0137, -0.0279, -0.0515, ..., -0.0456, 0.0137, 0.0674], + [ 0.0422, -0.0597, -0.0450, ..., -0.0351, -0.0144, -0.0182], + [-0.0934, 0.0005, 0.0697, ..., 0.0462, -0.0225, -0.0431]], + device='cuda:0'), grad: tensor([[ 1.6522e-04, 6.2659e-06, 1.5497e-04, ..., 8.9258e-06, + 4.1910e-09, 7.5176e-06], + [ 2.7582e-05, 8.2776e-06, 2.3216e-05, ..., -6.9737e-06, + -2.9197e-07, 9.7752e-06], + [ 6.7520e-04, 1.8328e-05, 6.3038e-04, ..., 3.5435e-05, + 5.3085e-08, 5.7787e-05], + ..., + [ 1.2346e-05, -1.4976e-05, 2.6837e-05, ..., 8.0317e-06, + 1.1129e-07, -6.0380e-05], + [ 1.4973e-04, 1.4812e-05, 1.6701e-04, ..., 1.4886e-05, + 3.5390e-08, 1.3411e-05], + [ 6.6757e-05, -3.7163e-05, -2.0117e-05, ..., -1.3387e-04, + 2.3749e-08, 1.9386e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0234, -0.0190, -0.0291, 0.0039, 0.0017, 0.0308, 0.0146, -0.0234, + 0.0129, -0.0033], device='cuda:0'), grad: tensor([ 3.4475e-04, 4.3452e-05, 1.4648e-03, -1.4555e-04, 9.8705e-05, + 7.5388e-04, -2.9812e-03, -3.7476e-06, 3.8242e-04, 4.3571e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 213.99, cls_loss 0.0069 cls_loss_mapping 0.0158 cls_loss_causal 0.6528 re_mapping 0.0128 re_causal 0.0393 /// teacc 98.73 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0055, -0.0106, -0.0089, ..., -0.0859, -0.0144, -0.0379], + [ 0.0103, -0.0323, -0.0563, ..., -0.0002, -0.0077, -0.0698], + [ 0.0381, -0.0566, -0.0267, ..., 0.0252, -0.0630, 0.0306], + ..., + [ 0.0138, -0.0275, -0.0519, ..., -0.0461, 0.0138, 0.0679], + [ 0.0426, -0.0600, -0.0450, ..., -0.0353, -0.0154, -0.0187], + [-0.0943, 0.0005, 0.0700, ..., 0.0465, -0.0236, -0.0436]], + device='cuda:0'), grad: tensor([[-8.6725e-06, 2.8014e-06, 1.5900e-05, ..., 1.1146e-05, + 2.1886e-08, 1.2321e-06], + [ 2.2352e-05, 9.1642e-06, 1.7196e-05, ..., 7.3723e-06, + -1.8477e-06, 2.0862e-05], + [ 7.0706e-06, 3.6150e-05, 3.0726e-05, ..., 2.4408e-05, + 4.5449e-07, -1.0312e-05], + ..., + [ 1.7285e-05, 2.9206e-05, 3.8505e-05, ..., 3.8087e-05, + 8.4331e-07, 3.4004e-05], + [-3.8338e-04, 2.0921e-05, -6.4707e-04, ..., -3.6836e-04, + 1.2433e-07, -2.6894e-04], + [ 3.5453e-04, -2.9042e-05, 4.5109e-04, ..., 2.6512e-04, + 1.0198e-07, 2.3437e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0237, -0.0191, -0.0288, 0.0036, 0.0016, 0.0309, 0.0150, -0.0233, + 0.0132, -0.0037], device='cuda:0'), grad: tensor([-2.1949e-05, 5.8293e-05, 9.8288e-05, 1.0568e-04, -9.5546e-05, + 6.6102e-05, -1.4138e-04, 1.2767e-04, -1.2016e-03, 1.0023e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 213.96, cls_loss 0.0077 cls_loss_mapping 0.0130 cls_loss_causal 0.6669 re_mapping 0.0123 re_causal 0.0387 /// teacc 98.77 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0054, -0.0109, -0.0094, ..., -0.0866, -0.0148, -0.0385], + [ 0.0099, -0.0326, -0.0569, ..., -0.0005, -0.0073, -0.0704], + [ 0.0385, -0.0570, -0.0272, ..., 0.0250, -0.0643, 0.0318], + ..., + [ 0.0127, -0.0276, -0.0523, ..., -0.0466, 0.0139, 0.0671], + [ 0.0427, -0.0604, -0.0453, ..., -0.0360, -0.0156, -0.0190], + [-0.0948, 0.0008, 0.0705, ..., 0.0469, -0.0237, -0.0439]], + device='cuda:0'), grad: tensor([[ 1.0088e-05, 1.0440e-06, 2.0042e-05, ..., 2.5686e-06, + 6.9849e-08, 5.2452e-06], + [ 9.0897e-06, 2.8741e-06, -2.9281e-05, ..., -3.1859e-05, + -3.9041e-06, -1.3247e-03], + [ 1.4745e-05, 7.7859e-06, 1.6019e-05, ..., 1.0453e-05, + 2.6589e-07, 1.1139e-03], + ..., + [-3.0190e-05, -7.0818e-06, 1.8314e-05, ..., 1.5005e-05, + 2.1588e-06, 9.5129e-05], + [ 4.7207e-04, 1.0431e-05, 5.0163e-04, ..., 4.0710e-05, + 3.1851e-07, 1.7047e-05], + [ 8.8066e-06, -2.0549e-05, -6.4850e-05, ..., -4.9531e-05, + 4.2189e-07, 3.3155e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0239, -0.0196, -0.0284, 0.0035, 0.0018, 0.0317, 0.0151, -0.0239, + 0.0132, -0.0036], device='cuda:0'), grad: tensor([ 3.4899e-05, -2.4948e-03, 2.0733e-03, 9.6917e-05, 9.7632e-05, + 1.2827e-04, -1.0805e-03, 1.9670e-04, 9.8324e-04, -3.6865e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 214.11, cls_loss 0.0070 cls_loss_mapping 0.0160 cls_loss_causal 0.6448 re_mapping 0.0131 re_causal 0.0390 /// teacc 98.75 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0049, -0.0110, -0.0091, ..., -0.0873, -0.0149, -0.0382], + [ 0.0097, -0.0328, -0.0571, ..., -0.0003, -0.0071, -0.0702], + [ 0.0385, -0.0577, -0.0277, ..., 0.0250, -0.0646, 0.0315], + ..., + [ 0.0132, -0.0278, -0.0530, ..., -0.0474, 0.0138, 0.0678], + [ 0.0425, -0.0605, -0.0461, ..., -0.0366, -0.0158, -0.0193], + [-0.0953, 0.0012, 0.0715, ..., 0.0478, -0.0239, -0.0450]], + device='cuda:0'), grad: tensor([[-2.0042e-05, 2.3600e-06, 1.8822e-06, ..., 6.5975e-06, + 1.0710e-08, 3.2149e-06], + [ 3.7141e-06, 9.5367e-06, -2.4289e-06, ..., 1.5832e-08, + 2.7427e-07, 1.5676e-05], + [ 6.4969e-06, 4.9919e-06, 7.8082e-06, ..., 5.3421e-06, + 4.0978e-08, -2.5988e-05], + ..., + [ 8.3894e-06, -6.7472e-05, 7.7188e-06, ..., -5.7518e-06, + -1.2629e-06, -5.8711e-05], + [-1.0140e-05, 8.7693e-06, 5.9977e-06, ..., 1.1988e-05, + 1.5367e-08, 1.0252e-05], + [ 1.4327e-05, 3.3438e-05, -5.2691e-05, ..., -4.1187e-05, + 1.5134e-07, 4.2170e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0229, -0.0193, -0.0289, 0.0038, 0.0011, 0.0315, 0.0149, -0.0239, + 0.0126, -0.0031], device='cuda:0'), grad: tensor([-2.8580e-05, 1.6749e-05, 4.0829e-06, 1.4579e-04, 7.6517e-06, + -1.6081e-04, 1.2457e-05, -9.5606e-05, 9.8497e-06, 8.8394e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 213.95, cls_loss 0.0087 cls_loss_mapping 0.0138 cls_loss_causal 0.6371 re_mapping 0.0126 re_causal 0.0359 /// teacc 98.68 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0054, -0.0113, -0.0102, ..., -0.0879, -0.0150, -0.0390], + [ 0.0092, -0.0331, -0.0570, ..., 0.0003, -0.0066, -0.0706], + [ 0.0384, -0.0582, -0.0280, ..., 0.0249, -0.0649, 0.0317], + ..., + [ 0.0133, -0.0281, -0.0535, ..., -0.0480, 0.0138, 0.0680], + [ 0.0430, -0.0608, -0.0463, ..., -0.0370, -0.0159, -0.0199], + [-0.0948, 0.0012, 0.0722, ..., 0.0479, -0.0242, -0.0458]], + device='cuda:0'), grad: tensor([[ 8.1420e-05, 3.4925e-06, 5.0426e-05, ..., 1.2606e-05, + 3.3528e-08, 4.8566e-04], + [ 8.6308e-05, 1.0118e-05, 2.2858e-05, ..., 2.9400e-05, + -5.1782e-07, 2.3329e-04], + [-7.6723e-04, 3.7104e-06, 1.6168e-05, ..., 5.1558e-06, + 5.6345e-08, -2.2717e-03], + ..., + [ 4.1199e-04, 2.1964e-05, 4.0293e-05, ..., 6.5804e-05, + 1.7090e-07, 1.0633e-03], + [-1.9088e-05, 3.7581e-05, -3.7942e-06, ..., 9.1553e-05, + 9.7789e-08, 2.0981e-04], + [ 9.9301e-05, -1.0324e-04, -1.2541e-04, ..., -2.3162e-04, + 8.1956e-08, 7.0930e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0235, -0.0193, -0.0293, 0.0044, 0.0013, 0.0314, 0.0151, -0.0241, + 0.0127, -0.0029], device='cuda:0'), grad: tensor([ 6.2990e-04, 3.8171e-04, -3.4828e-03, 2.6965e-04, -8.8513e-06, + 4.0054e-04, -2.7680e-04, 1.8320e-03, 2.8419e-04, -2.9355e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 214.05, cls_loss 0.0076 cls_loss_mapping 0.0142 cls_loss_causal 0.6571 re_mapping 0.0123 re_causal 0.0371 /// teacc 98.80 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0051, -0.0115, -0.0105, ..., -0.0886, -0.0154, -0.0397], + [ 0.0090, -0.0336, -0.0569, ..., 0.0005, -0.0062, -0.0703], + [ 0.0387, -0.0583, -0.0285, ..., 0.0247, -0.0662, 0.0324], + ..., + [ 0.0130, -0.0280, -0.0545, ..., -0.0493, 0.0136, 0.0682], + [ 0.0435, -0.0611, -0.0466, ..., -0.0374, -0.0161, -0.0205], + [-0.0955, 0.0015, 0.0728, ..., 0.0488, -0.0243, -0.0462]], + device='cuda:0'), grad: tensor([[-6.0588e-05, 7.3463e-06, 2.8417e-05, ..., 2.1160e-06, + 4.1910e-08, -3.2604e-05], + [ 4.9680e-05, 7.4096e-06, 5.4911e-06, ..., -3.9078e-06, + -6.5565e-07, 3.1680e-05], + [ 1.0848e-04, 3.5763e-05, 1.0699e-04, ..., 1.5542e-05, + 1.7742e-07, 8.9586e-05], + ..., + [-1.4687e-04, -9.1672e-05, 2.9176e-05, ..., 1.7837e-05, + -4.0233e-07, -3.7551e-04], + [-3.8087e-05, 1.2711e-05, 4.9382e-05, ..., 7.0594e-06, + 8.8476e-08, 7.8857e-05], + [ 4.3422e-05, -4.2729e-06, -2.3231e-05, ..., -7.7903e-05, + 2.3935e-07, 8.5652e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0234, -0.0192, -0.0293, 0.0041, 0.0016, 0.0313, 0.0148, -0.0244, + 0.0133, -0.0029], device='cuda:0'), grad: tensor([-4.2176e-04, 1.1176e-04, 4.4298e-04, 2.7370e-04, -2.2435e-04, + 5.6356e-05, 4.3511e-05, -5.5742e-04, 4.8667e-05, 2.2626e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.17, cls_loss 0.0078 cls_loss_mapping 0.0149 cls_loss_causal 0.6220 re_mapping 0.0129 re_causal 0.0379 /// teacc 98.86 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0048, -0.0116, -0.0104, ..., -0.0891, -0.0156, -0.0400], + [ 0.0086, -0.0337, -0.0564, ..., 0.0013, -0.0054, -0.0703], + [ 0.0391, -0.0584, -0.0289, ..., 0.0246, -0.0669, 0.0327], + ..., + [ 0.0128, -0.0279, -0.0554, ..., -0.0505, 0.0130, 0.0689], + [ 0.0442, -0.0615, -0.0468, ..., -0.0373, -0.0163, -0.0211], + [-0.0962, 0.0016, 0.0729, ..., 0.0490, -0.0246, -0.0471]], + device='cuda:0'), grad: tensor([[-1.7214e-04, 3.5837e-06, -5.1968e-06, ..., 3.7998e-06, + 0.0000e+00, 9.5367e-06], + [ 1.2957e-05, 1.8999e-05, 8.4750e-08, ..., 1.0848e-05, + 0.0000e+00, 2.7433e-05], + [-6.6161e-06, 1.0617e-05, 3.6918e-06, ..., 8.0466e-06, + 0.0000e+00, -5.4479e-05], + ..., + [ 9.0227e-06, 1.4283e-05, 3.9302e-06, ..., 2.2367e-05, + 0.0000e+00, 2.4721e-05], + [ 6.0409e-05, 3.6359e-05, 2.4617e-05, ..., 3.8385e-05, + 0.0000e+00, 4.8965e-05], + [ 5.8651e-05, 9.7975e-06, -2.8417e-05, ..., -1.0416e-05, + 0.0000e+00, 3.7491e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0230, -0.0189, -0.0297, 0.0031, 0.0018, 0.0322, 0.0141, -0.0241, + 0.0135, -0.0034], device='cuda:0'), grad: tensor([-3.9101e-04, 5.9545e-05, -2.2456e-05, -1.8339e-03, 2.7329e-05, + 1.8711e-03, -1.6928e-04, 7.2539e-05, 2.2662e-04, 1.5986e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 213.72, cls_loss 0.0082 cls_loss_mapping 0.0155 cls_loss_causal 0.6341 re_mapping 0.0123 re_causal 0.0364 /// teacc 98.84 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0045, -0.0120, -0.0108, ..., -0.0900, -0.0156, -0.0408], + [ 0.0085, -0.0345, -0.0560, ..., 0.0018, -0.0053, -0.0713], + [ 0.0394, -0.0590, -0.0293, ..., 0.0246, -0.0670, 0.0337], + ..., + [ 0.0126, -0.0274, -0.0558, ..., -0.0508, 0.0130, 0.0692], + [ 0.0442, -0.0620, -0.0477, ..., -0.0385, -0.0163, -0.0218], + [-0.0969, 0.0014, 0.0732, ..., 0.0492, -0.0246, -0.0479]], + device='cuda:0'), grad: tensor([[-1.1876e-05, 1.8835e-05, 1.8343e-05, ..., 1.5557e-05, + 0.0000e+00, 2.6494e-05], + [ 1.7695e-06, 1.9163e-05, -5.5544e-06, ..., -2.5004e-05, + 0.0000e+00, 2.2143e-05], + [-2.2233e-05, 2.5824e-05, 3.8266e-05, ..., 4.2856e-05, + 0.0000e+00, -1.3709e-04], + ..., + [ 2.2072e-06, -2.4829e-06, 5.4479e-05, ..., 5.4955e-05, + 0.0000e+00, -2.5406e-05], + [-6.5006e-06, 7.4625e-05, 1.1396e-04, ..., 1.4377e-04, + 0.0000e+00, 6.4969e-05], + [ 1.2457e-05, 7.5436e-04, 1.3914e-03, ..., 1.4553e-03, + 0.0000e+00, 6.2346e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0229, -0.0192, -0.0292, 0.0037, 0.0017, 0.0323, 0.0145, -0.0238, + 0.0126, -0.0038], device='cuda:0'), grad: tensor([ 3.8445e-05, -1.3456e-05, -5.4359e-05, -1.0222e-04, -3.1624e-03, + 6.8724e-05, 2.7388e-05, 9.2924e-05, 2.8682e-04, 2.8191e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 213.98, cls_loss 0.0070 cls_loss_mapping 0.0123 cls_loss_causal 0.6522 re_mapping 0.0121 re_causal 0.0359 /// teacc 98.78 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0040, -0.0120, -0.0105, ..., -0.0903, -0.0157, -0.0414], + [ 0.0080, -0.0349, -0.0558, ..., 0.0022, -0.0053, -0.0716], + [ 0.0396, -0.0597, -0.0299, ..., 0.0245, -0.0670, 0.0339], + ..., + [ 0.0123, -0.0279, -0.0564, ..., -0.0514, 0.0130, 0.0694], + [ 0.0445, -0.0626, -0.0483, ..., -0.0394, -0.0164, -0.0219], + [-0.0974, 0.0023, 0.0736, ..., 0.0497, -0.0246, -0.0476]], + device='cuda:0'), grad: tensor([[-1.2023e-06, 1.0254e-06, 6.8657e-06, ..., 4.3586e-06, + 3.0245e-07, 1.9856e-06], + [ 7.0445e-06, -1.4473e-06, -5.8442e-05, ..., -9.5189e-05, + -8.3745e-06, 1.9027e-06], + [-2.3711e-06, -7.2876e-08, 7.2680e-06, ..., -4.9965e-07, + 2.6217e-07, -2.7150e-05], + ..., + [-2.2613e-06, -4.0047e-06, 3.3021e-05, ..., 4.5925e-05, + 4.6678e-06, -9.7975e-06], + [-2.5794e-05, 1.6168e-06, -1.0319e-05, ..., 1.1384e-05, + 7.2131e-07, 3.5800e-06], + [ 1.2897e-05, -3.3416e-06, -2.6729e-06, ..., -3.9637e-06, + 1.1139e-06, 3.8967e-06]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0220, -0.0191, -0.0295, 0.0034, 0.0014, 0.0320, 0.0149, -0.0244, + 0.0128, -0.0035], device='cuda:0'), grad: tensor([ 4.7795e-06, -1.4114e-04, -1.9461e-05, 5.9485e-05, 3.9190e-05, + 5.5671e-05, -4.9204e-05, 6.2883e-05, -4.8578e-05, 3.6240e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 65---------------------------------------------------- +epoch 65, time 231.41, cls_loss 0.0065 cls_loss_mapping 0.0140 cls_loss_causal 0.6513 re_mapping 0.0122 re_causal 0.0371 /// teacc 98.91 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0042, -0.0125, -0.0107, ..., -0.0915, -0.0158, -0.0424], + [ 0.0078, -0.0343, -0.0552, ..., 0.0032, -0.0053, -0.0717], + [ 0.0393, -0.0612, -0.0310, ..., 0.0233, -0.0672, 0.0339], + ..., + [ 0.0123, -0.0273, -0.0573, ..., -0.0524, 0.0131, 0.0702], + [ 0.0441, -0.0631, -0.0486, ..., -0.0407, -0.0164, -0.0221], + [-0.0977, 0.0021, 0.0741, ..., 0.0503, -0.0248, -0.0486]], + device='cuda:0'), grad: tensor([[-1.1188e-04, -1.6049e-05, -8.2731e-05, ..., 1.9640e-05, + 1.3015e-07, 1.3575e-05], + [ 1.1891e-05, 1.3150e-05, 1.1427e-06, ..., 1.4738e-07, + 1.1688e-07, 3.8475e-05], + [-7.5459e-05, -6.4969e-05, 5.2638e-06, ..., -5.1767e-05, + 3.8883e-08, -3.9816e-04], + ..., + [-1.2212e-05, -5.4598e-05, 1.1414e-05, ..., 1.6645e-05, + -2.3190e-06, 7.2159e-06], + [ 2.7329e-05, 6.7294e-05, 8.2254e-05, ..., 7.4685e-05, + 4.1188e-07, 1.1516e-04], + [ 8.5652e-05, -2.8586e-04, -5.4598e-04, ..., -4.9782e-04, + 9.2899e-07, 2.7135e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0224, -0.0188, -0.0301, 0.0037, 0.0018, 0.0326, 0.0147, -0.0243, + 0.0123, -0.0036], device='cuda:0'), grad: tensor([-2.8467e-04, 6.2525e-05, -6.1798e-04, 3.1066e-04, 5.4932e-04, + 7.2837e-05, 6.6876e-05, -8.7023e-05, 2.7919e-04, -3.5143e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.04, cls_loss 0.0064 cls_loss_mapping 0.0132 cls_loss_causal 0.6351 re_mapping 0.0120 re_causal 0.0347 /// teacc 98.82 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0044, -0.0127, -0.0112, ..., -0.0917, -0.0163, -0.0429], + [ 0.0077, -0.0346, -0.0553, ..., 0.0034, -0.0052, -0.0721], + [ 0.0391, -0.0621, -0.0312, ..., 0.0229, -0.0673, 0.0346], + ..., + [ 0.0127, -0.0267, -0.0573, ..., -0.0524, 0.0130, 0.0707], + [ 0.0442, -0.0637, -0.0487, ..., -0.0410, -0.0168, -0.0232], + [-0.0977, 0.0020, 0.0745, ..., 0.0505, -0.0250, -0.0492]], + device='cuda:0'), grad: tensor([[ 3.9558e-07, 7.7188e-06, 5.9754e-06, ..., 1.9986e-06, + 0.0000e+00, 8.6501e-06], + [ 1.4350e-05, 6.8694e-06, 1.6376e-05, ..., -3.1907e-06, + 0.0000e+00, 7.7188e-06], + [ 2.5347e-05, 1.1422e-05, 1.6376e-05, ..., 2.6841e-06, + 0.0000e+00, 1.0036e-05], + ..., + [ 1.2759e-06, 1.5888e-06, 3.9227e-06, ..., 4.8503e-06, + 0.0000e+00, 4.3735e-06], + [-3.1710e-05, 5.2750e-06, 6.0312e-06, ..., 1.1042e-05, + 0.0000e+00, 6.7912e-06], + [ 7.6294e-06, 7.2010e-06, 8.1025e-07, ..., 6.3069e-06, + 0.0000e+00, 8.0466e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0229, -0.0188, -0.0299, 0.0038, 0.0015, 0.0325, 0.0151, -0.0238, + 0.0121, -0.0035], device='cuda:0'), grad: tensor([-5.7146e-06, 2.6837e-05, 8.4043e-05, -4.6670e-05, 9.9987e-06, + -5.1260e-05, -3.1561e-05, 2.2352e-05, -4.1425e-05, 3.3289e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.11, cls_loss 0.0080 cls_loss_mapping 0.0164 cls_loss_causal 0.6825 re_mapping 0.0116 re_causal 0.0348 /// teacc 98.69 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0040, -0.0131, -0.0115, ..., -0.0925, -0.0164, -0.0435], + [ 0.0075, -0.0354, -0.0547, ..., 0.0038, -0.0052, -0.0724], + [ 0.0392, -0.0626, -0.0317, ..., 0.0228, -0.0673, 0.0348], + ..., + [ 0.0128, -0.0263, -0.0582, ..., -0.0531, 0.0130, 0.0712], + [ 0.0443, -0.0642, -0.0489, ..., -0.0418, -0.0168, -0.0238], + [-0.0983, 0.0022, 0.0752, ..., 0.0513, -0.0250, -0.0496]], + device='cuda:0'), grad: tensor([[-2.8402e-05, 3.0369e-05, 1.8403e-05, ..., 1.9431e-05, + 0.0000e+00, 4.5657e-05], + [ 2.5593e-06, 2.1353e-05, -1.2942e-05, ..., -1.3188e-05, + 0.0000e+00, 3.1143e-05], + [ 1.0051e-05, 2.3648e-05, 1.2547e-05, ..., 1.4827e-05, + 0.0000e+00, 3.2037e-06], + ..., + [-8.8587e-06, 2.6096e-06, 1.0669e-05, ..., 1.2480e-05, + 0.0000e+00, 1.0751e-05], + [-3.1054e-05, 2.4116e-04, 6.3777e-05, ..., 5.3078e-05, + 0.0000e+00, 3.4118e-04], + [ 1.1437e-05, 1.3329e-05, -3.8505e-05, ..., -3.5495e-05, + 0.0000e+00, 3.1501e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0226, -0.0187, -0.0300, 0.0031, 0.0013, 0.0331, 0.0148, -0.0236, + 0.0119, -0.0034], device='cuda:0'), grad: tensor([ 2.7493e-05, 4.2439e-05, 1.0890e-04, -1.5688e-03, 2.1502e-05, + 3.8338e-04, 5.5760e-05, 3.8534e-05, 8.1301e-04, 7.8797e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 213.97, cls_loss 0.0072 cls_loss_mapping 0.0149 cls_loss_causal 0.6004 re_mapping 0.0119 re_causal 0.0338 /// teacc 98.69 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0034, -0.0133, -0.0118, ..., -0.0932, -0.0164, -0.0440], + [ 0.0072, -0.0361, -0.0540, ..., 0.0044, -0.0051, -0.0730], + [ 0.0397, -0.0631, -0.0321, ..., 0.0227, -0.0674, 0.0355], + ..., + [ 0.0118, -0.0253, -0.0590, ..., -0.0536, 0.0132, 0.0716], + [ 0.0453, -0.0647, -0.0493, ..., -0.0418, -0.0168, -0.0240], + [-0.0988, 0.0021, 0.0757, ..., 0.0519, -0.0254, -0.0504]], + device='cuda:0'), grad: tensor([[-5.1856e-05, 3.1143e-06, 1.2837e-05, ..., 6.1207e-06, + 0.0000e+00, -5.5181e-08], + [ 3.2540e-06, 9.8720e-06, -3.4153e-05, ..., 4.9546e-06, + 0.0000e+00, -5.9217e-05], + [ 9.6038e-06, 4.7013e-06, 4.0025e-05, ..., 3.7476e-06, + 0.0000e+00, 5.9634e-05], + ..., + [ 2.8890e-06, 5.6103e-06, 3.3855e-05, ..., 3.6240e-05, + 0.0000e+00, -7.4431e-06], + [-2.7433e-05, 2.1547e-05, 1.5823e-06, ..., 3.0503e-05, + 0.0000e+00, 7.2382e-06], + [ 3.0965e-05, -2.4557e-05, -6.4969e-05, ..., -9.0003e-05, + 0.0000e+00, 1.3895e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0220, -0.0190, -0.0297, 0.0041, 0.0010, 0.0316, 0.0144, -0.0239, + 0.0131, -0.0036], device='cuda:0'), grad: tensor([-1.4234e-04, -2.6703e-04, 2.8992e-04, -7.4133e-06, 3.3468e-05, + -1.0872e-04, 1.5759e-04, 5.3644e-05, 1.9163e-05, -2.8953e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 214.16, cls_loss 0.0073 cls_loss_mapping 0.0139 cls_loss_causal 0.6482 re_mapping 0.0118 re_causal 0.0336 /// teacc 98.75 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0036, -0.0140, -0.0126, ..., -0.0946, -0.0165, -0.0449], + [ 0.0070, -0.0352, -0.0528, ..., 0.0059, -0.0053, -0.0732], + [ 0.0398, -0.0637, -0.0327, ..., 0.0231, -0.0675, 0.0359], + ..., + [ 0.0118, -0.0252, -0.0599, ..., -0.0545, 0.0136, 0.0720], + [ 0.0456, -0.0657, -0.0495, ..., -0.0438, -0.0169, -0.0247], + [-0.0995, 0.0019, 0.0759, ..., 0.0520, -0.0255, -0.0512]], + device='cuda:0'), grad: tensor([[ 1.0449e-06, 1.1716e-06, 1.1191e-05, ..., 9.5591e-06, + 0.0000e+00, 1.7583e-06], + [ 6.0871e-06, 5.5842e-06, -3.7938e-05, ..., -9.7871e-05, + 0.0000e+00, 5.6922e-06], + [ 4.9062e-06, 7.9572e-06, 2.7105e-05, ..., 3.6031e-05, + 0.0000e+00, 1.9521e-05], + ..., + [-1.6075e-06, 6.8955e-06, 3.5971e-05, ..., 5.5373e-05, + 0.0000e+00, -2.4706e-05], + [-2.9325e-04, 5.0738e-06, -1.6764e-05, ..., -1.0121e-04, + 0.0000e+00, 6.2361e-06], + [ 6.5044e-06, 2.1443e-05, 6.0469e-05, ..., 5.8770e-05, + 0.0000e+00, 1.4633e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0226, -0.0186, -0.0296, 0.0037, 0.0010, 0.0326, 0.0141, -0.0235, + 0.0128, -0.0043], device='cuda:0'), grad: tensor([ 1.3739e-05, -1.5450e-04, 9.9301e-05, -2.6762e-05, -1.7345e-04, + 5.8937e-04, -1.8060e-04, 6.9439e-05, -3.7265e-04, 1.3602e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 214.11, cls_loss 0.0071 cls_loss_mapping 0.0138 cls_loss_causal 0.6422 re_mapping 0.0113 re_causal 0.0346 /// teacc 98.80 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0038, -0.0142, -0.0133, ..., -0.0954, -0.0166, -0.0453], + [ 0.0065, -0.0351, -0.0535, ..., 0.0058, -0.0053, -0.0741], + [ 0.0398, -0.0641, -0.0331, ..., 0.0229, -0.0677, 0.0355], + ..., + [ 0.0115, -0.0246, -0.0604, ..., -0.0550, 0.0136, 0.0735], + [ 0.0460, -0.0662, -0.0498, ..., -0.0443, -0.0170, -0.0252], + [-0.0995, 0.0020, 0.0773, ..., 0.0532, -0.0255, -0.0519]], + device='cuda:0'), grad: tensor([[ 1.5140e-05, 1.2573e-06, 1.3769e-05, ..., 2.2277e-06, + 0.0000e+00, 3.5726e-06], + [ 1.9521e-06, 7.2410e-07, 2.6170e-06, ..., 1.4352e-06, + 0.0000e+00, 1.5954e-06], + [-4.5262e-06, 1.1446e-06, 3.7402e-06, ..., -1.1645e-05, + 0.0000e+00, -2.5123e-05], + ..., + [-1.8310e-06, -2.6450e-06, 1.9111e-06, ..., 2.8349e-06, + 0.0000e+00, -4.2915e-06], + [-1.2182e-05, 1.5358e-06, 5.4874e-06, ..., 6.0014e-06, + 0.0000e+00, -3.5693e-07], + [ 3.5074e-06, -2.5611e-06, 1.5497e-06, ..., 1.5153e-06, + 0.0000e+00, 1.6270e-06]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0234, -0.0189, -0.0305, 0.0035, 0.0006, 0.0326, 0.0142, -0.0226, + 0.0128, -0.0034], device='cuda:0'), grad: tensor([ 3.4243e-05, 4.3139e-06, -3.7909e-05, 2.4721e-05, 3.5018e-05, + -3.6545e-06, -4.7922e-05, -2.3022e-06, -1.6853e-05, 1.0319e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 213.85, cls_loss 0.0074 cls_loss_mapping 0.0130 cls_loss_causal 0.6313 re_mapping 0.0109 re_causal 0.0332 /// teacc 98.76 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0037, -0.0145, -0.0136, ..., -0.0962, -0.0167, -0.0461], + [ 0.0061, -0.0346, -0.0526, ..., 0.0055, -0.0052, -0.0733], + [ 0.0409, -0.0646, -0.0334, ..., 0.0246, -0.0679, 0.0346], + ..., + [ 0.0115, -0.0242, -0.0615, ..., -0.0560, 0.0135, 0.0758], + [ 0.0461, -0.0666, -0.0501, ..., -0.0446, -0.0171, -0.0266], + [-0.1003, 0.0018, 0.0772, ..., 0.0531, -0.0256, -0.0526]], + device='cuda:0'), grad: tensor([[-2.6926e-05, 2.0072e-05, -1.3411e-05, ..., 1.7852e-05, + 9.6858e-08, 9.5248e-05], + [ 1.3486e-05, 3.0732e-04, 2.8820e-03, ..., 2.0218e-03, + -2.5658e-07, 9.2149e-05], + [-9.5785e-05, -4.7475e-05, 2.9787e-05, ..., 1.5691e-05, + 6.0303e-08, -8.7547e-04], + ..., + [ 3.6806e-05, 5.1856e-05, 1.8191e-04, ..., 1.2875e-04, + 1.4086e-07, 2.8253e-04], + [ 1.8060e-05, 8.7500e-05, 1.0484e-04, ..., 6.3837e-05, + 1.1222e-07, 1.5008e-04], + [ 9.9242e-06, -2.8038e-04, 1.3602e-04, ..., -5.1677e-05, + 2.5611e-08, 3.6389e-05]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0233, -0.0179, -0.0304, 0.0026, 0.0008, 0.0324, 0.0146, -0.0221, + 0.0125, -0.0040], device='cuda:0'), grad: tensor([-1.1891e-04, 3.8986e-03, -1.0281e-03, -3.8296e-05, -4.4518e-03, + 3.3545e-04, 1.9109e-04, 6.2513e-04, 4.4823e-04, 1.4150e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.04, cls_loss 0.0062 cls_loss_mapping 0.0121 cls_loss_causal 0.6364 re_mapping 0.0109 re_causal 0.0324 /// teacc 98.81 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0038, -0.0148, -0.0136, ..., -0.0971, -0.0169, -0.0464], + [ 0.0054, -0.0352, -0.0534, ..., 0.0052, -0.0051, -0.0735], + [ 0.0408, -0.0654, -0.0339, ..., 0.0244, -0.0682, 0.0349], + ..., + [ 0.0117, -0.0241, -0.0623, ..., -0.0566, 0.0135, 0.0761], + [ 0.0467, -0.0671, -0.0503, ..., -0.0446, -0.0173, -0.0274], + [-0.1008, 0.0026, 0.0780, ..., 0.0539, -0.0256, -0.0532]], + device='cuda:0'), grad: tensor([[ 9.7677e-06, 1.6708e-06, 8.7321e-06, ..., 4.3809e-06, + 3.7439e-07, 3.9190e-06], + [ 6.6124e-06, 9.3356e-06, -1.2465e-05, ..., -3.0268e-06, + -1.0552e-06, 1.5162e-05], + [ 4.5784e-06, 2.2314e-06, 4.5486e-06, ..., 5.3048e-06, + 1.0640e-07, -7.2658e-05], + ..., + [-4.6206e-04, -2.4652e-04, -1.3185e-04, ..., -5.1546e-04, + 3.7672e-07, -2.5773e-04], + [ 7.1898e-06, 3.1032e-06, 5.1074e-06, ..., 1.1854e-05, + 3.5577e-07, 9.4995e-06], + [ 4.3297e-04, 2.2483e-04, 1.2839e-04, ..., 4.9067e-04, + 4.7963e-08, 2.8300e-04]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0234, -0.0180, -0.0307, 0.0025, 0.0008, 0.0321, 0.0147, -0.0223, + 0.0131, -0.0037], device='cuda:0'), grad: tensor([ 2.6405e-05, -3.7104e-05, -2.9698e-05, 3.2663e-05, 2.9951e-05, + 1.7092e-05, -1.4283e-05, -2.0828e-03, 4.9978e-05, 2.0084e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 213.91, cls_loss 0.0049 cls_loss_mapping 0.0105 cls_loss_causal 0.5973 re_mapping 0.0106 re_causal 0.0321 /// teacc 98.80 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0037, -0.0150, -0.0139, ..., -0.0980, -0.0170, -0.0472], + [ 0.0053, -0.0358, -0.0535, ..., 0.0052, -0.0050, -0.0740], + [ 0.0410, -0.0653, -0.0343, ..., 0.0246, -0.0688, 0.0360], + ..., + [ 0.0114, -0.0242, -0.0630, ..., -0.0577, 0.0133, 0.0761], + [ 0.0472, -0.0676, -0.0507, ..., -0.0452, -0.0180, -0.0278], + [-0.1011, 0.0031, 0.0785, ..., 0.0545, -0.0258, -0.0535]], + device='cuda:0'), grad: tensor([[ 2.9281e-05, 1.1912e-06, 4.5449e-05, ..., 3.4999e-06, + 1.4086e-07, 2.9534e-05], + [ 3.1758e-06, 4.6268e-06, -7.9302e-07, ..., -2.7958e-06, + 4.2608e-08, 1.1966e-05], + [ 2.8834e-05, 9.0972e-06, 2.3738e-05, ..., 3.5763e-06, + 4.3074e-08, -4.6380e-06], + ..., + [ 1.5320e-06, -2.5146e-08, 8.1584e-06, ..., 1.3381e-05, + 1.3667e-07, 1.5646e-05], + [-1.5223e-04, 5.4203e-06, -1.3137e-04, ..., 5.7042e-05, + 5.4110e-07, -1.9228e-04], + [ 5.2989e-05, -2.1458e-05, -9.9093e-06, ..., -5.5879e-05, + 1.7532e-07, 3.8624e-05]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0236, -0.0183, -0.0299, 0.0022, 0.0011, 0.0320, 0.0146, -0.0227, + 0.0132, -0.0034], device='cuda:0'), grad: tensor([ 5.7369e-05, 1.5900e-05, 6.4373e-05, 2.6560e-04, 1.5095e-05, + -6.7949e-05, 8.1956e-05, 3.2783e-05, -5.1546e-04, 5.0128e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 214.11, cls_loss 0.0064 cls_loss_mapping 0.0122 cls_loss_causal 0.6505 re_mapping 0.0106 re_causal 0.0332 /// teacc 98.82 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0047, -0.0152, -0.0149, ..., -0.0992, -0.0172, -0.0479], + [ 0.0034, -0.0367, -0.0537, ..., 0.0042, -0.0050, -0.0744], + [ 0.0415, -0.0653, -0.0348, ..., 0.0264, -0.0692, 0.0364], + ..., + [ 0.0115, -0.0242, -0.0637, ..., -0.0588, 0.0133, 0.0761], + [ 0.0479, -0.0680, -0.0508, ..., -0.0457, -0.0183, -0.0282], + [-0.1007, 0.0028, 0.0784, ..., 0.0539, -0.0260, -0.0541]], + device='cuda:0'), grad: tensor([[ 1.4447e-05, 2.6766e-06, 1.4983e-05, ..., 1.3381e-05, + 3.6322e-08, 2.6375e-06], + [ 8.5905e-06, 6.5304e-06, -4.2329e-07, ..., -5.8599e-06, + 1.0477e-08, 5.9977e-06], + [ 4.1544e-05, 1.1198e-05, 2.7478e-05, ..., 1.6093e-05, + 4.5402e-08, 1.2666e-06], + ..., + [ 4.3213e-06, 4.2208e-06, 2.2709e-05, ..., 3.1024e-05, + 4.4238e-09, -4.7013e-06], + [-9.0897e-05, 9.3058e-06, -1.8075e-05, ..., 5.3525e-05, + 5.7742e-08, -8.8066e-06], + [ 4.3124e-05, -1.8284e-05, -2.1756e-05, ..., 4.5508e-05, + 1.0710e-08, 1.2055e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0247, -0.0192, -0.0293, 0.0025, 0.0024, 0.0319, 0.0147, -0.0229, + 0.0137, -0.0036], device='cuda:0'), grad: tensor([ 3.9965e-05, 4.2953e-06, 1.1629e-04, 4.5091e-05, -3.9130e-05, + -2.8181e-04, 1.0043e-04, 4.8548e-05, -1.4567e-04, 1.1200e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 214.26, cls_loss 0.0053 cls_loss_mapping 0.0126 cls_loss_causal 0.6224 re_mapping 0.0111 re_causal 0.0337 /// teacc 98.76 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0045, -0.0153, -0.0153, ..., -0.0996, -0.0174, -0.0488], + [ 0.0034, -0.0371, -0.0535, ..., 0.0047, -0.0050, -0.0751], + [ 0.0414, -0.0657, -0.0353, ..., 0.0263, -0.0700, 0.0365], + ..., + [ 0.0121, -0.0244, -0.0643, ..., -0.0592, 0.0133, 0.0771], + [ 0.0485, -0.0686, -0.0511, ..., -0.0459, -0.0183, -0.0291], + [-0.1011, 0.0031, 0.0791, ..., 0.0542, -0.0260, -0.0546]], + device='cuda:0'), grad: tensor([[-2.7847e-04, 9.2620e-07, -1.3578e-04, ..., -2.3127e-05, + 0.0000e+00, -6.3121e-05], + [ 1.3068e-05, 2.7493e-06, -3.8952e-05, ..., -6.4969e-05, + 0.0000e+00, 1.6585e-05], + [ 3.7819e-05, -1.0198e-04, 8.0645e-05, ..., 5.0873e-05, + 0.0000e+00, -6.5184e-04], + ..., + [ 4.0457e-06, 9.1136e-05, 1.0081e-05, ..., 9.3579e-06, + 0.0000e+00, 5.8222e-04], + [ 2.5094e-05, 4.9844e-06, 3.0413e-05, ..., 8.3074e-06, + 0.0000e+00, 2.8715e-05], + [ 1.7285e-04, -8.3633e-07, 7.9155e-05, ..., 2.7176e-06, + 0.0000e+00, 5.4836e-05]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0247, -0.0192, -0.0295, 0.0026, 0.0020, 0.0313, 0.0147, -0.0223, + 0.0137, -0.0035], device='cuda:0'), grad: tensor([-1.0424e-03, -6.9857e-05, -6.6710e-04, 1.5211e-04, 8.8573e-05, + 3.2336e-05, -1.6129e-04, 8.5878e-04, 1.3924e-04, 6.6948e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 213.78, cls_loss 0.0071 cls_loss_mapping 0.0114 cls_loss_causal 0.6075 re_mapping 0.0105 re_causal 0.0298 /// teacc 98.79 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0041, -0.0155, -0.0162, ..., -0.1015, -0.0175, -0.0491], + [ 0.0031, -0.0371, -0.0537, ..., 0.0049, -0.0050, -0.0753], + [ 0.0414, -0.0661, -0.0356, ..., 0.0261, -0.0701, 0.0369], + ..., + [ 0.0122, -0.0251, -0.0650, ..., -0.0593, 0.0133, 0.0768], + [ 0.0488, -0.0688, -0.0514, ..., -0.0468, -0.0184, -0.0296], + [-0.1014, 0.0034, 0.0801, ..., 0.0554, -0.0260, -0.0549]], + device='cuda:0'), grad: tensor([[-3.6329e-05, -2.2631e-06, -9.1270e-06, ..., 4.7535e-06, + 6.0536e-09, 6.0126e-06], + [ 9.2685e-06, 1.5283e-06, -1.3135e-05, ..., -1.8731e-05, + -2.2119e-07, 9.1568e-06], + [-8.2791e-05, 1.2349e-06, 5.0627e-06, ..., -1.5870e-05, + 1.2573e-08, -1.2034e-04], + ..., + [ 4.4741e-06, 1.7202e-06, 1.3180e-05, ..., 1.3962e-05, + 5.2620e-08, 1.1753e-06], + [ 4.6909e-05, 2.0806e-06, 1.9521e-05, ..., 1.8314e-05, + 4.1444e-08, 2.9698e-05], + [ 2.6435e-05, 5.6848e-06, 1.7390e-05, ..., 9.1717e-06, + 1.9092e-08, 1.7742e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0247, -0.0191, -0.0296, 0.0028, 0.0012, 0.0315, 0.0144, -0.0224, + 0.0135, -0.0028], device='cuda:0'), grad: tensor([-5.3555e-05, -2.8595e-05, -5.5122e-04, 2.6751e-04, -1.8114e-06, + 2.6569e-05, 1.3104e-06, 6.8009e-05, 1.9526e-04, 7.6950e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 214.34, cls_loss 0.0068 cls_loss_mapping 0.0119 cls_loss_causal 0.6067 re_mapping 0.0102 re_causal 0.0299 /// teacc 98.79 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0041, -0.0158, -0.0172, ..., -0.1046, -0.0177, -0.0492], + [ 0.0028, -0.0357, -0.0535, ..., 0.0053, -0.0048, -0.0743], + [ 0.0413, -0.0668, -0.0364, ..., 0.0259, -0.0707, 0.0368], + ..., + [ 0.0123, -0.0261, -0.0662, ..., -0.0601, 0.0135, 0.0762], + [ 0.0494, -0.0692, -0.0517, ..., -0.0475, -0.0188, -0.0301], + [-0.1023, 0.0037, 0.0811, ..., 0.0565, -0.0262, -0.0552]], + device='cuda:0'), grad: tensor([[-1.0602e-05, 5.0478e-07, -5.3942e-06, ..., 1.7211e-06, + 1.2573e-08, 7.4832e-07], + [ 1.1865e-06, 9.1037e-07, -3.1590e-06, ..., -6.8322e-06, + -5.0664e-07, 2.7530e-06], + [ 3.9451e-06, 2.0303e-06, 2.6412e-06, ..., 1.9465e-06, + 1.0338e-07, 1.1109e-05], + ..., + [ 6.5193e-09, 8.9779e-06, 2.2613e-06, ..., 3.9674e-06, + 1.3271e-07, 8.6939e-07], + [-5.8748e-06, 1.0300e-06, 4.1686e-06, ..., 3.3397e-06, + 4.0513e-08, 4.3362e-06], + [ 5.5507e-06, -1.0720e-06, -1.5954e-06, ..., -6.3097e-07, + 4.3306e-08, 2.5406e-06]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0257, -0.0196, -0.0299, 0.0031, 0.0012, 0.0312, 0.0149, -0.0220, + 0.0139, -0.0023], device='cuda:0'), grad: tensor([-6.2108e-05, -7.9945e-06, 2.6584e-05, 5.4464e-06, 7.0706e-06, + 3.9935e-06, -2.6636e-07, 8.9854e-06, 1.5832e-08, 1.8224e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 214.07, cls_loss 0.0059 cls_loss_mapping 0.0125 cls_loss_causal 0.6306 re_mapping 0.0107 re_causal 0.0316 /// teacc 98.71 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0029, -0.0160, -0.0169, ..., -0.1053, -0.0179, -0.0491], + [ 0.0026, -0.0360, -0.0538, ..., 0.0054, -0.0046, -0.0745], + [ 0.0412, -0.0676, -0.0376, ..., 0.0256, -0.0709, 0.0371], + ..., + [ 0.0122, -0.0269, -0.0668, ..., -0.0607, 0.0136, 0.0759], + [ 0.0493, -0.0696, -0.0524, ..., -0.0488, -0.0189, -0.0306], + [-0.1032, 0.0034, 0.0811, ..., 0.0565, -0.0263, -0.0560]], + device='cuda:0'), grad: tensor([[ 5.2974e-06, 1.0859e-06, 2.4214e-06, ..., 2.4378e-05, + 0.0000e+00, 3.5949e-07], + [ 2.6580e-06, 1.3532e-06, -6.6981e-06, ..., -1.2200e-06, + 0.0000e+00, 1.0598e-06], + [ 1.1615e-05, 8.7358e-07, 3.0920e-06, ..., 3.2157e-05, + 0.0000e+00, -2.5406e-06], + ..., + [ 1.9390e-06, 5.0329e-06, 1.4499e-05, ..., 2.3559e-05, + 0.0000e+00, -1.3281e-06], + [ 4.7572e-06, 1.0982e-05, 3.5197e-05, ..., 6.2168e-05, + 0.0000e+00, 8.1537e-07], + [ 6.3106e-06, -2.9430e-05, -6.4850e-05, ..., -4.8608e-05, + 0.0000e+00, -1.5330e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0248, -0.0196, -0.0302, 0.0037, 0.0021, 0.0312, 0.0145, -0.0223, + 0.0131, -0.0026], device='cuda:0'), grad: tensor([ 7.9155e-05, -9.1717e-06, 1.1224e-04, 2.2912e-04, 1.5542e-05, + -6.8521e-04, 8.8811e-05, 6.3598e-05, 1.5545e-04, -4.9829e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.13, cls_loss 0.0078 cls_loss_mapping 0.0123 cls_loss_causal 0.6312 re_mapping 0.0107 re_causal 0.0313 /// teacc 98.67 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0037, -0.0162, -0.0173, ..., -0.1063, -0.0181, -0.0495], + [ 0.0024, -0.0364, -0.0535, ..., 0.0045, -0.0038, -0.0750], + [ 0.0412, -0.0681, -0.0383, ..., 0.0268, -0.0714, 0.0382], + ..., + [ 0.0123, -0.0267, -0.0674, ..., -0.0613, 0.0136, 0.0761], + [ 0.0489, -0.0701, -0.0532, ..., -0.0500, -0.0190, -0.0318], + [-0.1039, 0.0034, 0.0810, ..., 0.0564, -0.0265, -0.0568]], + device='cuda:0'), grad: tensor([[-1.5235e-04, 2.7679e-06, -1.0192e-04, ..., 9.5516e-06, + 0.0000e+00, 3.6694e-06], + [ 6.5863e-06, 4.8615e-06, 8.7768e-06, ..., 3.7514e-06, + 0.0000e+00, 1.4879e-05], + [ 1.0163e-04, 7.6830e-05, 1.4231e-05, ..., 9.8348e-06, + 0.0000e+00, 2.2602e-04], + ..., + [-1.5795e-04, -3.1328e-04, -3.6812e-04, ..., 4.2826e-05, + 0.0000e+00, -5.5695e-04], + [ 3.1441e-05, 6.2026e-06, 2.7597e-05, ..., 4.7773e-05, + 0.0000e+00, 3.2157e-05], + [ 1.7881e-04, 5.9795e-04, 4.7989e-03, ..., 3.7384e-03, + 0.0000e+00, 2.6464e-04]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0251, -0.0199, -0.0295, 0.0034, 0.0027, 0.0318, 0.0155, -0.0223, + 0.0120, -0.0031], device='cuda:0'), grad: tensor([-4.0102e-04, 3.9369e-05, 3.1781e-04, 9.2149e-05, -6.5002e-03, + -1.1109e-05, 3.3379e-05, -1.5364e-03, 1.3912e-04, 7.8278e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 214.31, cls_loss 0.0055 cls_loss_mapping 0.0096 cls_loss_causal 0.6093 re_mapping 0.0106 re_causal 0.0313 /// teacc 98.83 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0034, -0.0163, -0.0179, ..., -0.1070, -0.0181, -0.0500], + [ 0.0022, -0.0366, -0.0534, ..., 0.0047, -0.0037, -0.0751], + [ 0.0409, -0.0689, -0.0392, ..., 0.0265, -0.0716, 0.0380], + ..., + [ 0.0118, -0.0279, -0.0685, ..., -0.0621, 0.0136, 0.0756], + [ 0.0490, -0.0703, -0.0534, ..., -0.0503, -0.0190, -0.0321], + [-0.1045, 0.0038, 0.0825, ..., 0.0568, -0.0265, -0.0585]], + device='cuda:0'), grad: tensor([[ 4.3362e-05, 9.1493e-06, 5.3704e-05, ..., 1.3709e-06, + 0.0000e+00, 3.7581e-05], + [ 1.8373e-05, 1.4529e-05, 2.1487e-05, ..., 6.1579e-06, + 0.0000e+00, 4.2021e-05], + [ 6.8426e-05, 8.4043e-06, 1.0633e-04, ..., 5.7183e-06, + 0.0000e+00, -2.1577e-05], + ..., + [-3.9816e-05, -6.3121e-05, 4.0919e-05, ..., 2.6971e-05, + 0.0000e+00, -1.5271e-04], + [ 8.6725e-06, 7.1414e-06, 2.6733e-05, ..., 1.4834e-05, + 0.0000e+00, 1.3463e-05], + [ 1.0654e-05, -6.5304e-06, -3.5346e-05, ..., -4.4018e-05, + 0.0000e+00, 2.5406e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0251, -0.0198, -0.0301, 0.0045, 0.0029, 0.0318, 0.0160, -0.0229, + 0.0121, -0.0031], device='cuda:0'), grad: tensor([ 1.3697e-04, 9.0957e-05, 1.7393e-04, 1.2058e-04, 7.9775e-04, + 1.5199e-04, -1.3399e-03, -1.7691e-04, 5.3376e-05, -8.8066e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 214.16, cls_loss 0.0054 cls_loss_mapping 0.0096 cls_loss_causal 0.5908 re_mapping 0.0106 re_causal 0.0305 /// teacc 98.81 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0044, -0.0166, -0.0174, ..., -0.1079, -0.0181, -0.0500], + [ 0.0020, -0.0373, -0.0536, ..., 0.0044, -0.0038, -0.0760], + [ 0.0407, -0.0694, -0.0399, ..., 0.0265, -0.0720, 0.0375], + ..., + [ 0.0121, -0.0273, -0.0694, ..., -0.0627, 0.0138, 0.0773], + [ 0.0489, -0.0707, -0.0527, ..., -0.0503, -0.0190, -0.0327], + [-0.1057, 0.0041, 0.0824, ..., 0.0571, -0.0266, -0.0592]], + device='cuda:0'), grad: tensor([[-1.2882e-05, 6.2399e-07, 2.1607e-06, ..., 3.6471e-06, + 0.0000e+00, 1.8263e-06], + [ 6.8620e-06, 1.3262e-06, -3.9600e-06, ..., -8.5980e-06, + 0.0000e+00, 6.1654e-06], + [ 1.8813e-06, 1.4165e-06, 2.5667e-06, ..., 3.2112e-06, + 0.0000e+00, -1.2569e-05], + ..., + [ 4.2096e-07, 1.5117e-05, 4.3899e-05, ..., 1.2505e-04, + 0.0000e+00, 5.0478e-06], + [ 6.7428e-06, 5.2750e-06, 7.2382e-06, ..., 5.3197e-06, + 0.0000e+00, 6.2808e-06], + [ 4.0457e-06, -3.6299e-05, -1.0151e-04, ..., -2.8300e-04, + 0.0000e+00, -3.0212e-06]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0256, -0.0203, -0.0306, 0.0037, 0.0029, 0.0324, 0.0173, -0.0220, + 0.0123, -0.0035], device='cuda:0'), grad: tensor([-2.4438e-05, -3.3993e-06, -6.0908e-06, -2.2411e-05, 6.7830e-05, + 1.1861e-04, -2.6718e-05, 1.3804e-04, 3.4302e-05, -2.7561e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 214.28, cls_loss 0.0054 cls_loss_mapping 0.0113 cls_loss_causal 0.5978 re_mapping 0.0104 re_causal 0.0309 /// teacc 98.80 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0040, -0.0167, -0.0177, ..., -0.1084, -0.0182, -0.0507], + [ 0.0014, -0.0365, -0.0511, ..., 0.0061, -0.0035, -0.0762], + [ 0.0408, -0.0699, -0.0403, ..., 0.0266, -0.0723, 0.0378], + ..., + [ 0.0127, -0.0269, -0.0724, ..., -0.0654, 0.0138, 0.0784], + [ 0.0492, -0.0715, -0.0540, ..., -0.0508, -0.0190, -0.0338], + [-0.1064, 0.0043, 0.0835, ..., 0.0585, -0.0266, -0.0607]], + device='cuda:0'), grad: tensor([[-2.6133e-06, 3.9823e-06, 7.6070e-06, ..., 4.6529e-06, + 0.0000e+00, 1.7257e-06], + [ 4.3288e-06, 3.2037e-05, 1.5631e-05, ..., 1.7017e-05, + 0.0000e+00, 1.8954e-05], + [ 1.6307e-06, 6.3144e-06, 5.3868e-06, ..., 1.6112e-06, + 0.0000e+00, -1.1705e-05], + ..., + [ 2.9318e-06, -5.5641e-05, 9.4101e-06, ..., 1.1489e-05, + 0.0000e+00, -4.5776e-05], + [ 5.9277e-05, 3.9637e-05, 5.5820e-05, ..., 2.5406e-05, + 0.0000e+00, 1.4238e-05], + [ 2.4047e-06, -4.5925e-05, -6.4671e-05, ..., -1.0151e-04, + 0.0000e+00, -4.2692e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0253, -0.0195, -0.0305, 0.0034, 0.0018, 0.0320, 0.0173, -0.0220, + 0.0118, -0.0030], device='cuda:0'), grad: tensor([ 5.5740e-07, 7.3910e-05, -3.8631e-06, 9.3043e-05, 1.3389e-05, + 3.8773e-05, -2.5320e-04, -9.1612e-05, 2.2280e-04, -9.3520e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 213.97, cls_loss 0.0062 cls_loss_mapping 0.0123 cls_loss_causal 0.6255 re_mapping 0.0098 re_causal 0.0302 /// teacc 98.84 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0036, -0.0172, -0.0181, ..., -0.1089, -0.0182, -0.0516], + [ 0.0012, -0.0372, -0.0516, ..., 0.0056, -0.0036, -0.0764], + [ 0.0408, -0.0683, -0.0407, ..., 0.0266, -0.0723, 0.0395], + ..., + [ 0.0129, -0.0274, -0.0744, ..., -0.0663, 0.0139, 0.0776], + [ 0.0494, -0.0725, -0.0540, ..., -0.0514, -0.0191, -0.0348], + [-0.1068, 0.0048, 0.0849, ..., 0.0595, -0.0267, -0.0608]], + device='cuda:0'), grad: tensor([[ 9.5963e-06, 1.5143e-06, 1.9863e-05, ..., 6.2864e-07, + 3.9116e-08, 3.2447e-06], + [ 1.8999e-06, 7.2531e-06, 3.0035e-07, ..., -5.1409e-06, + 3.3993e-07, 1.8075e-05], + [ 8.5533e-06, 9.6336e-06, 1.3903e-05, ..., 1.5236e-06, + 1.1129e-07, 1.3828e-05], + ..., + [ 3.6974e-07, -2.5645e-05, 3.7830e-06, ..., 1.9930e-06, + -1.2899e-06, -6.2943e-05], + [ 4.2878e-06, 4.3847e-06, 8.3670e-06, ..., 3.1982e-06, + 1.0058e-07, 1.0014e-05], + [ 3.9190e-06, 8.2552e-06, 1.0759e-05, ..., 3.8557e-06, + 1.0012e-07, 1.5222e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0255, -0.0198, -0.0289, 0.0039, 0.0018, 0.0314, 0.0167, -0.0230, + 0.0115, -0.0023], device='cuda:0'), grad: tensor([-1.2565e-04, 3.3855e-05, 1.1003e-04, 5.1744e-06, -5.2005e-05, + 1.5885e-05, -9.0063e-05, -8.5235e-05, 5.2452e-05, 1.3554e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 214.15, cls_loss 0.0062 cls_loss_mapping 0.0117 cls_loss_causal 0.6338 re_mapping 0.0098 re_causal 0.0290 /// teacc 98.84 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0034, -0.0173, -0.0186, ..., -0.1094, -0.0184, -0.0523], + [ 0.0007, -0.0370, -0.0517, ..., 0.0058, -0.0037, -0.0767], + [ 0.0406, -0.0686, -0.0415, ..., 0.0265, -0.0731, 0.0393], + ..., + [ 0.0127, -0.0274, -0.0748, ..., -0.0655, 0.0143, 0.0784], + [ 0.0504, -0.0727, -0.0541, ..., -0.0523, -0.0194, -0.0348], + [-0.1074, 0.0044, 0.0839, ..., 0.0588, -0.0269, -0.0615]], + device='cuda:0'), grad: tensor([[-2.8349e-06, 6.7474e-07, 5.8375e-06, ..., 2.0489e-06, + 0.0000e+00, 1.5935e-06], + [ 8.7777e-07, 2.3581e-06, 8.4713e-06, ..., 1.0237e-05, + 0.0000e+00, 3.1833e-06], + [ 9.4697e-06, 5.5209e-06, 6.3665e-06, ..., 3.4496e-06, + 0.0000e+00, 5.5209e-06], + ..., + [ 5.2853e-07, 1.1344e-06, 2.1935e-05, ..., 2.6941e-05, + 0.0000e+00, -8.6799e-07], + [-4.4882e-05, 1.0438e-05, 1.4432e-05, ..., 4.5955e-05, + 0.0000e+00, 3.8259e-06], + [ 8.6799e-06, -2.2337e-05, -9.9659e-05, ..., -1.3113e-04, + 0.0000e+00, 2.3153e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0253, -0.0200, -0.0292, 0.0030, 0.0031, 0.0324, 0.0163, -0.0223, + 0.0121, -0.0040], device='cuda:0'), grad: tensor([ 1.0259e-05, 2.1145e-05, 4.2945e-05, 2.7433e-05, 3.4183e-05, + 3.8087e-05, 1.5438e-05, 3.8624e-05, -8.0764e-05, -1.4722e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 214.30, cls_loss 0.0042 cls_loss_mapping 0.0093 cls_loss_causal 0.5841 re_mapping 0.0100 re_causal 0.0294 /// teacc 98.85 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0031, -0.0175, -0.0188, ..., -0.1098, -0.0185, -0.0536], + [ 0.0003, -0.0372, -0.0514, ..., 0.0060, -0.0037, -0.0770], + [ 0.0411, -0.0688, -0.0422, ..., 0.0264, -0.0734, 0.0401], + ..., + [ 0.0128, -0.0273, -0.0754, ..., -0.0660, 0.0143, 0.0787], + [ 0.0500, -0.0733, -0.0548, ..., -0.0532, -0.0195, -0.0362], + [-0.1079, 0.0046, 0.0840, ..., 0.0594, -0.0269, -0.0618]], + device='cuda:0'), grad: tensor([[-2.5153e-04, -2.0802e-05, -1.0216e-04, ..., 2.2771e-07, + 1.5367e-08, -7.2300e-05], + [ 1.0110e-05, 2.4270e-06, -2.4531e-06, ..., -2.2538e-06, + -4.4331e-07, 3.5781e-06], + [ 1.3568e-05, 1.3664e-05, 6.3814e-06, ..., 4.0466e-07, + 5.1688e-08, 2.0698e-05], + ..., + [ 1.4842e-05, -4.4823e-05, 1.0587e-05, ..., 2.8238e-06, + 1.3970e-07, -6.2227e-05], + [-7.5847e-06, -6.2697e-06, -1.0923e-05, ..., 7.6368e-07, + 3.8650e-08, 9.3579e-06], + [ 1.2314e-04, 1.6913e-05, 6.3717e-05, ..., -5.7779e-06, + 8.8476e-08, 4.5389e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0251, -0.0200, -0.0287, 0.0024, 0.0030, 0.0332, 0.0163, -0.0223, + 0.0111, -0.0041], device='cuda:0'), grad: tensor([-5.3072e-04, 1.6183e-05, 4.7505e-05, 4.5985e-05, 3.6865e-05, + 4.5389e-05, 1.1313e-04, -3.1680e-05, -2.8417e-05, 2.8539e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 214.21, cls_loss 0.0056 cls_loss_mapping 0.0105 cls_loss_causal 0.6345 re_mapping 0.0103 re_causal 0.0303 /// teacc 98.85 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0028, -0.0182, -0.0191, ..., -0.1097, -0.0190, -0.0543], + [ 0.0001, -0.0393, -0.0516, ..., 0.0056, -0.0038, -0.0773], + [ 0.0408, -0.0695, -0.0435, ..., 0.0262, -0.0748, 0.0407], + ..., + [ 0.0118, -0.0266, -0.0756, ..., -0.0662, 0.0169, 0.0785], + [ 0.0508, -0.0727, -0.0552, ..., -0.0537, -0.0198, -0.0353], + [-0.1087, 0.0052, 0.0844, ..., 0.0598, -0.0299, -0.0625]], + device='cuda:0'), grad: tensor([[ 8.8811e-05, 1.7639e-06, 1.2359e-06, ..., 1.6261e-06, + 0.0000e+00, 1.4277e-03], + [ 3.3956e-06, 4.1910e-06, -2.9076e-06, ..., -4.2506e-06, + 0.0000e+00, 4.0770e-05], + [-1.2469e-04, 2.1104e-06, 2.8498e-06, ..., 4.0866e-06, + 0.0000e+00, -1.8978e-03], + ..., + [ 7.2159e-06, 1.7241e-05, 3.6478e-05, ..., 7.2420e-05, + 0.0000e+00, -3.5446e-06], + [ 1.8328e-05, 3.6117e-06, 3.2727e-06, ..., 6.6608e-06, + 0.0000e+00, 2.7204e-04], + [ 1.7568e-05, 5.5939e-05, 5.9307e-05, ..., 1.2600e-04, + 0.0000e+00, 8.9347e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0248, -0.0204, -0.0290, 0.0022, 0.0029, 0.0334, 0.0162, -0.0223, + 0.0123, -0.0042], device='cuda:0'), grad: tensor([ 2.4242e-03, 3.8370e-06, -3.1929e-03, 8.4639e-05, -3.1185e-04, + 1.1638e-05, 2.0325e-04, 7.4744e-05, 4.6873e-04, 2.3413e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 87---------------------------------------------------- +epoch 87, time 230.48, cls_loss 0.0044 cls_loss_mapping 0.0088 cls_loss_causal 0.6012 re_mapping 0.0094 re_causal 0.0286 /// teacc 98.94 lr 0.00010000 +Epoch 89, weight, value: tensor([[-2.8128e-03, -1.8939e-02, -2.0102e-02, ..., -1.1003e-01, + -1.9154e-02, -5.6805e-02], + [-4.8598e-05, -3.9678e-02, -5.1523e-02, ..., 5.9077e-03, + -3.6896e-03, -7.7730e-02], + [ 4.0551e-02, -6.9865e-02, -4.4864e-02, ..., 2.6145e-02, + -7.5475e-02, 4.1167e-02], + ..., + [ 1.1943e-02, -2.6143e-02, -7.6031e-02, ..., -6.6541e-02, + 1.7846e-02, 7.9364e-02], + [ 5.0706e-02, -7.3334e-02, -5.6043e-02, ..., -5.4623e-02, + -2.0180e-02, -3.5804e-02], + [-1.0912e-01, 5.3183e-03, 8.4716e-02, ..., 6.0002e-02, + -3.0892e-02, -6.3285e-02]], device='cuda:0'), grad: tensor([[-7.6771e-05, 1.6699e-06, 3.2447e-06, ..., 2.2352e-07, + 1.5367e-08, 1.6317e-06], + [ 2.3097e-06, 7.0892e-06, -1.4873e-06, ..., -2.7586e-06, + -6.9384e-08, 7.1377e-06], + [ 3.8147e-05, 8.5458e-06, 9.3281e-06, ..., 7.7626e-07, + 7.9628e-08, -9.8348e-06], + ..., + [ 4.6268e-06, -2.1005e-04, 5.5954e-06, ..., 3.0193e-06, + -1.8161e-07, -1.7548e-04], + [-7.2382e-06, 1.0550e-05, 1.7077e-05, ..., 1.2182e-06, + 3.6322e-08, 1.4670e-05], + [ 1.1757e-05, 8.6069e-05, 2.1979e-05, ..., -3.6471e-06, + 8.1956e-08, 7.0214e-05]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0255, -0.0204, -0.0291, 0.0017, 0.0027, 0.0333, 0.0174, -0.0218, + 0.0119, -0.0042], device='cuda:0'), grad: tensor([-1.7643e-04, 9.1195e-06, 8.6069e-05, 1.5640e-04, -8.4639e-05, + 3.0518e-05, 6.2943e-05, -2.9635e-04, 3.7193e-05, 1.7524e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 213.96, cls_loss 0.0040 cls_loss_mapping 0.0095 cls_loss_causal 0.6171 re_mapping 0.0099 re_causal 0.0307 /// teacc 98.85 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0023, -0.0187, -0.0205, ..., -0.1102, -0.0198, -0.0572], + [-0.0002, -0.0395, -0.0503, ..., 0.0064, -0.0035, -0.0774], + [ 0.0405, -0.0703, -0.0455, ..., 0.0261, -0.0768, 0.0408], + ..., + [ 0.0120, -0.0261, -0.0775, ..., -0.0674, 0.0184, 0.0802], + [ 0.0508, -0.0740, -0.0564, ..., -0.0552, -0.0206, -0.0367], + [-0.1095, 0.0055, 0.0850, ..., 0.0602, -0.0317, -0.0640]], + device='cuda:0'), grad: tensor([[-2.8431e-05, -1.4730e-05, 2.1029e-06, ..., 3.2000e-06, + 0.0000e+00, 9.6578e-07], + [ 1.2312e-06, 5.3532e-06, 1.2750e-06, ..., 1.1241e-06, + 0.0000e+00, 5.7071e-06], + [ 2.7381e-06, 7.1153e-06, 5.5507e-06, ..., 7.8604e-06, + 0.0000e+00, 4.7386e-06], + ..., + [-5.3504e-07, 1.2255e-04, 9.7513e-05, ..., 7.1466e-05, + 0.0000e+00, 8.7738e-05], + [-2.2594e-06, 1.0669e-05, 7.3798e-06, ..., 1.5318e-05, + 0.0000e+00, 6.2846e-06], + [ 6.3106e-06, -9.3162e-05, -1.4532e-04, ..., -1.2815e-04, + 0.0000e+00, 9.0292e-07]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0253, -0.0196, -0.0296, 0.0015, 0.0027, 0.0331, 0.0170, -0.0218, + 0.0118, -0.0041], device='cuda:0'), grad: tensor([-5.2214e-05, 1.1742e-05, 2.2337e-05, -1.2112e-04, 2.5973e-05, + 4.5896e-06, 3.3230e-05, 2.4700e-04, 1.5616e-05, -1.8775e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 213.90, cls_loss 0.0051 cls_loss_mapping 0.0113 cls_loss_causal 0.5659 re_mapping 0.0100 re_causal 0.0280 /// teacc 98.86 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0013, -0.0178, -0.0190, ..., -0.1111, -0.0201, -0.0580], + [-0.0005, -0.0398, -0.0504, ..., 0.0064, -0.0035, -0.0777], + [ 0.0408, -0.0708, -0.0462, ..., 0.0259, -0.0773, 0.0408], + ..., + [ 0.0121, -0.0259, -0.0781, ..., -0.0677, 0.0185, 0.0812], + [ 0.0507, -0.0747, -0.0569, ..., -0.0561, -0.0208, -0.0383], + [-0.1101, 0.0058, 0.0855, ..., 0.0611, -0.0320, -0.0648]], + device='cuda:0'), grad: tensor([[-2.4661e-05, 3.2829e-07, 2.6543e-06, ..., 3.3714e-06, + 7.4506e-09, 1.0710e-08], + [ 6.9244e-07, 2.9616e-07, -2.4796e-05, ..., -2.5317e-05, + -2.4121e-07, 2.4009e-06], + [ 2.0936e-05, 2.5518e-06, 7.8380e-06, ..., 6.3218e-06, + 2.4214e-08, -1.8299e-05], + ..., + [ 6.1933e-07, -7.6964e-06, 3.8035e-06, ..., 4.5709e-06, + 6.6590e-08, -3.8054e-06], + [-5.8301e-06, 1.2498e-06, 5.0999e-06, ..., 7.9870e-06, + 2.4680e-08, 1.9558e-06], + [ 9.3132e-06, 7.0706e-06, 3.3844e-06, ..., 1.7658e-05, + 3.8650e-08, 6.2101e-06]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0238, -0.0197, -0.0295, 0.0022, 0.0026, 0.0324, 0.0161, -0.0212, + 0.0109, -0.0042], device='cuda:0'), grad: tensor([-7.6294e-05, -1.9383e-04, 7.9632e-05, 3.0428e-05, 7.6257e-06, + -1.7077e-05, 3.9041e-06, 1.0777e-04, 3.2894e-06, 5.4449e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 214.13, cls_loss 0.0042 cls_loss_mapping 0.0101 cls_loss_causal 0.5891 re_mapping 0.0092 re_causal 0.0283 /// teacc 98.84 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0012, -0.0180, -0.0193, ..., -0.1118, -0.0202, -0.0584], + [-0.0006, -0.0399, -0.0506, ..., 0.0064, -0.0033, -0.0780], + [ 0.0407, -0.0714, -0.0467, ..., 0.0260, -0.0779, 0.0413], + ..., + [ 0.0119, -0.0254, -0.0783, ..., -0.0679, 0.0190, 0.0816], + [ 0.0510, -0.0754, -0.0574, ..., -0.0571, -0.0210, -0.0394], + [-0.1106, 0.0058, 0.0860, ..., 0.0616, -0.0326, -0.0659]], + device='cuda:0'), grad: tensor([[-7.2196e-06, 1.0431e-06, 5.9903e-06, ..., 5.5209e-06, + 1.2107e-08, 5.6028e-06], + [ 4.5309e-07, 9.8869e-06, 1.2207e-03, ..., 1.0347e-03, + -4.2375e-07, 9.6083e-04], + [ 1.3700e-06, 3.7365e-06, -1.4162e-03, ..., -1.2197e-03, + 5.0757e-08, -1.0662e-03], + ..., + [ 4.7917e-07, -6.6578e-05, 1.2383e-05, ..., 1.3016e-05, + 8.1025e-08, -4.1306e-05], + [ 9.7044e-07, 6.9626e-06, 2.1994e-05, ..., -1.5289e-05, + 7.1246e-08, 1.6525e-05], + [ 1.2433e-06, -5.0589e-06, -1.4615e-04, ..., -1.4818e-04, + 4.8894e-08, 2.9236e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0239, -0.0198, -0.0291, 0.0021, 0.0024, 0.0320, 0.0164, -0.0210, + 0.0107, -0.0044], device='cuda:0'), grad: tensor([ 1.9297e-05, 7.1106e-03, -7.8278e-03, 4.0501e-05, 3.0971e-04, + 5.4717e-05, 5.8746e-04, -5.5879e-05, -1.3387e-04, -1.0026e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 214.08, cls_loss 0.0038 cls_loss_mapping 0.0084 cls_loss_causal 0.5794 re_mapping 0.0093 re_causal 0.0285 /// teacc 98.94 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0007, -0.0182, -0.0195, ..., -0.1124, -0.0203, -0.0588], + [-0.0008, -0.0401, -0.0502, ..., 0.0066, -0.0033, -0.0784], + [ 0.0404, -0.0717, -0.0474, ..., 0.0259, -0.0785, 0.0416], + ..., + [ 0.0115, -0.0252, -0.0787, ..., -0.0682, 0.0192, 0.0821], + [ 0.0511, -0.0759, -0.0585, ..., -0.0588, -0.0212, -0.0396], + [-0.1109, 0.0061, 0.0865, ..., 0.0624, -0.0326, -0.0664]], + device='cuda:0'), grad: tensor([[-2.2665e-05, 1.8766e-07, 1.1608e-05, ..., -4.2245e-06, + 0.0000e+00, 1.9232e-07], + [ 1.8300e-06, 2.1625e-06, -2.0601e-06, ..., -3.6061e-06, + 0.0000e+00, 3.2224e-06], + [ 1.7760e-06, 3.5623e-07, 1.7304e-06, ..., 6.7661e-07, + 0.0000e+00, 3.7067e-07], + ..., + [ 2.0657e-06, -7.5623e-06, 5.6485e-07, ..., 1.2880e-06, + 0.0000e+00, -1.1660e-05], + [ 1.4715e-05, 1.0431e-06, 1.1034e-05, ..., 2.3022e-06, + 0.0000e+00, 5.3458e-07], + [ 6.7018e-06, 4.1574e-06, 5.1670e-06, ..., 1.1073e-06, + 0.0000e+00, 6.5155e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0234, -0.0195, -0.0294, 0.0023, 0.0020, 0.0321, 0.0159, -0.0208, + 0.0101, -0.0040], device='cuda:0'), grad: tensor([-4.3690e-05, 1.9614e-06, 5.7109e-06, 3.1084e-05, 4.7386e-06, + 5.5027e-04, -5.9557e-04, -1.4506e-05, 3.3408e-05, 2.7210e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 213.84, cls_loss 0.0038 cls_loss_mapping 0.0096 cls_loss_causal 0.5762 re_mapping 0.0093 re_causal 0.0290 /// teacc 98.86 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0004, -0.0184, -0.0194, ..., -0.1126, -0.0204, -0.0595], + [-0.0028, -0.0406, -0.0492, ..., 0.0072, -0.0033, -0.0788], + [ 0.0421, -0.0720, -0.0483, ..., 0.0257, -0.0786, 0.0415], + ..., + [ 0.0117, -0.0248, -0.0794, ..., -0.0686, 0.0192, 0.0828], + [ 0.0513, -0.0763, -0.0582, ..., -0.0594, -0.0212, -0.0398], + [-0.1116, 0.0052, 0.0860, ..., 0.0618, -0.0326, -0.0671]], + device='cuda:0'), grad: tensor([[ 7.7635e-06, 5.7463e-07, 1.7598e-05, ..., 1.7956e-05, + 0.0000e+00, 5.3234e-06], + [-3.2067e-05, 1.6913e-06, -5.9277e-05, ..., -5.7161e-05, + 0.0000e+00, 6.5975e-06], + [-2.2680e-05, 1.2651e-05, 4.4554e-06, ..., -5.7518e-06, + 0.0000e+00, -4.8608e-05], + ..., + [ 1.2442e-05, -3.7365e-06, 8.3447e-06, ..., 1.1250e-05, + 0.0000e+00, 4.1336e-05], + [-1.7986e-05, -4.5225e-06, -1.4357e-05, ..., -2.5347e-05, + 0.0000e+00, -2.3330e-07], + [ 2.9311e-05, 1.9092e-06, 2.4080e-05, ..., 3.3677e-05, + 0.0000e+00, 7.1013e-07]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0231, -0.0202, -0.0285, 0.0013, 0.0028, 0.0328, 0.0153, -0.0205, + 0.0105, -0.0051], device='cuda:0'), grad: tensor([ 5.6237e-05, -2.1052e-04, -3.7521e-05, -1.5751e-05, 1.8790e-05, + 3.3379e-05, 3.4809e-05, 6.3539e-05, -5.1141e-05, 1.0842e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 214.00, cls_loss 0.0033 cls_loss_mapping 0.0081 cls_loss_causal 0.5752 re_mapping 0.0088 re_causal 0.0282 /// teacc 98.94 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0004, -0.0187, -0.0195, ..., -0.1129, -0.0205, -0.0607], + [-0.0028, -0.0409, -0.0476, ..., 0.0077, -0.0033, -0.0790], + [ 0.0419, -0.0725, -0.0490, ..., 0.0257, -0.0787, 0.0410], + ..., + [ 0.0124, -0.0241, -0.0808, ..., -0.0685, 0.0192, 0.0842], + [ 0.0516, -0.0767, -0.0588, ..., -0.0605, -0.0213, -0.0400], + [-0.1121, 0.0051, 0.0862, ..., 0.0617, -0.0327, -0.0680]], + device='cuda:0'), grad: tensor([[-1.2843e-06, 7.5344e-07, 1.7993e-06, ..., 9.1940e-06, + 0.0000e+00, 1.2606e-05], + [ 8.0978e-07, 1.0859e-06, -3.6806e-06, ..., -3.7551e-06, + 0.0000e+00, 4.5858e-06], + [-3.9876e-05, -4.4815e-06, -4.3094e-05, ..., -1.4544e-04, + 0.0000e+00, -1.1814e-04], + ..., + [ 4.8727e-06, -7.9647e-06, 5.4855e-07, ..., 7.4282e-06, + 0.0000e+00, -1.9502e-06], + [ 4.2431e-06, 3.6675e-06, 1.2182e-05, ..., 1.6317e-05, + 0.0000e+00, 1.0945e-05], + [ 2.9311e-05, 2.7139e-06, 8.4415e-06, ..., 1.1182e-04, + 0.0000e+00, 1.2636e-04]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0232, -0.0198, -0.0289, 0.0012, 0.0030, 0.0326, 0.0152, -0.0197, + 0.0105, -0.0056], device='cuda:0'), grad: tensor([ 1.2733e-05, 1.6680e-06, -2.9683e-04, -1.6057e-04, 2.8849e-05, + 9.7692e-05, -2.6301e-06, -2.3972e-06, 4.0472e-05, 2.8157e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 214.09, cls_loss 0.0057 cls_loss_mapping 0.0108 cls_loss_causal 0.5485 re_mapping 0.0091 re_causal 0.0264 /// teacc 98.78 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0006, -0.0192, -0.0200, ..., -0.1143, -0.0209, -0.0611], + [-0.0027, -0.0392, -0.0482, ..., 0.0078, 0.0017, -0.0777], + [ 0.0417, -0.0729, -0.0496, ..., 0.0256, -0.0817, 0.0410], + ..., + [ 0.0127, -0.0249, -0.0796, ..., -0.0680, 0.0145, 0.0838], + [ 0.0517, -0.0773, -0.0589, ..., -0.0614, -0.0217, -0.0402], + [-0.1130, 0.0049, 0.0863, ..., 0.0618, -0.0330, -0.0688]], + device='cuda:0'), grad: tensor([[ 4.0093e-07, 8.5821e-07, 1.2545e-06, ..., 4.3679e-07, + 0.0000e+00, 3.4869e-06], + [ 1.2154e-06, 3.7756e-06, -8.6948e-06, ..., -5.4762e-06, + 0.0000e+00, 6.1691e-06], + [-3.3289e-05, -1.1154e-05, 2.0601e-06, ..., 1.0571e-06, + 0.0000e+00, -1.1092e-04], + ..., + [ 2.0877e-05, -2.3603e-05, 4.2357e-06, ..., 2.0880e-06, + 0.0000e+00, 3.3110e-05], + [ 8.3297e-06, 4.8429e-06, 4.6380e-06, ..., 2.3600e-06, + 0.0000e+00, 2.5421e-05], + [ 1.2303e-06, 2.6524e-06, 6.3330e-08, ..., -2.1840e-07, + 0.0000e+00, 2.7139e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0237, -0.0191, -0.0293, 0.0009, 0.0024, 0.0332, 0.0151, -0.0194, + 0.0105, -0.0061], device='cuda:0'), grad: tensor([ 6.9477e-06, -1.2398e-05, -1.9443e-04, 7.1585e-05, -4.6566e-06, + -1.5028e-05, 1.8954e-05, 6.2704e-05, 5.5104e-05, 1.1139e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 214.09, cls_loss 0.0050 cls_loss_mapping 0.0080 cls_loss_causal 0.6068 re_mapping 0.0091 re_causal 0.0270 /// teacc 98.78 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 2.7008e-05, -1.9431e-02, -2.0019e-02, ..., -1.1500e-01, + -2.1879e-02, -6.1779e-02], + [-3.1136e-03, -3.9202e-02, -4.7500e-02, ..., 8.5775e-03, + 2.0387e-03, -7.8032e-02], + [ 4.1612e-02, -7.3134e-02, -5.0606e-02, ..., 2.5769e-02, + -8.6027e-02, 4.1686e-02], + ..., + [ 1.3096e-02, -2.4886e-02, -8.0500e-02, ..., -6.8003e-02, + 1.4154e-02, 8.3845e-02], + [ 5.2241e-02, -7.7765e-02, -5.8625e-02, ..., -6.2104e-02, + -2.3918e-02, -4.0583e-02], + [-1.1401e-01, 4.3534e-03, 8.5946e-02, ..., 6.1329e-02, + -3.3267e-02, -6.9433e-02]], device='cuda:0'), grad: tensor([[-1.0200e-05, 4.8662e-07, 1.0869e-06, ..., 1.4855e-06, + 6.0536e-09, 1.2340e-07], + [ 8.8708e-07, 1.8943e-06, -3.4627e-06, ..., -4.5709e-06, + -3.0641e-07, 1.6037e-06], + [ 1.4603e-06, 1.0077e-06, 1.5991e-06, ..., 1.8366e-06, + 5.6345e-08, 9.2061e-07], + ..., + [ 6.0769e-07, 3.1292e-06, 1.0200e-05, ..., 1.3500e-05, + 1.2713e-07, -7.4133e-07], + [-4.9584e-06, 1.4529e-06, 2.7772e-06, ..., 4.9137e-06, + 2.3749e-08, 7.7207e-07], + [ 2.6096e-06, -1.7509e-05, -3.3200e-05, ..., -4.4286e-05, + 2.2352e-08, -1.8021e-07]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0227, -0.0189, -0.0290, 0.0010, 0.0024, 0.0328, 0.0147, -0.0191, + 0.0108, -0.0073], device='cuda:0'), grad: tensor([-2.0981e-05, -5.7742e-06, 8.2627e-06, 3.6024e-06, 2.1800e-05, + 1.7524e-05, 7.2010e-06, 1.9059e-05, -6.3591e-06, -4.4495e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 214.13, cls_loss 0.0041 cls_loss_mapping 0.0087 cls_loss_causal 0.5779 re_mapping 0.0092 re_causal 0.0273 /// teacc 98.82 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0003, -0.0200, -0.0212, ..., -0.1172, -0.0222, -0.0625], + [-0.0039, -0.0393, -0.0478, ..., 0.0084, 0.0026, -0.0781], + [ 0.0415, -0.0737, -0.0513, ..., 0.0258, -0.0867, 0.0416], + ..., + [ 0.0135, -0.0256, -0.0819, ..., -0.0694, 0.0136, 0.0841], + [ 0.0529, -0.0783, -0.0590, ..., -0.0629, -0.0240, -0.0408], + [-0.1144, 0.0064, 0.0873, ..., 0.0632, -0.0333, -0.0699]], + device='cuda:0'), grad: tensor([[ 1.0571e-06, 2.1197e-06, 3.8054e-06, ..., 3.6377e-06, + 0.0000e+00, 2.1085e-06], + [ 1.9483e-06, 7.8753e-06, -2.1756e-06, ..., 9.1083e-07, + 1.1642e-09, 7.0296e-06], + [ 2.6096e-06, 3.3192e-06, 3.7234e-06, ..., 3.4682e-06, + 2.3283e-10, -4.9531e-05], + ..., + [-9.5427e-05, -1.4818e-04, 2.9821e-06, ..., -1.8942e-04, + -3.7253e-09, -9.9987e-06], + [ 1.3737e-06, 6.6459e-06, 1.4137e-06, ..., 1.1481e-05, + 2.3283e-10, 1.0841e-05], + [ 9.9361e-05, 1.5223e-04, -2.1644e-06, ..., 1.9944e-04, + 9.3132e-10, 3.4779e-05]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0231, -0.0188, -0.0297, 0.0017, 0.0022, 0.0320, 0.0145, -0.0196, + 0.0111, -0.0058], device='cuda:0'), grad: tensor([ 8.7246e-06, 2.2322e-05, -4.8816e-05, 9.6416e-04, 5.1335e-06, + -1.0462e-03, 1.0826e-05, -7.1669e-04, 3.7253e-05, 7.6246e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 214.14, cls_loss 0.0037 cls_loss_mapping 0.0078 cls_loss_causal 0.5835 re_mapping 0.0091 re_causal 0.0262 /// teacc 98.83 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0006, -0.0202, -0.0213, ..., -0.1173, -0.0226, -0.0629], + [-0.0041, -0.0396, -0.0468, ..., 0.0097, 0.0025, -0.0783], + [ 0.0415, -0.0748, -0.0521, ..., 0.0257, -0.0871, 0.0419], + ..., + [ 0.0133, -0.0257, -0.0830, ..., -0.0704, 0.0137, 0.0842], + [ 0.0532, -0.0783, -0.0591, ..., -0.0637, -0.0242, -0.0411], + [-0.1150, 0.0065, 0.0875, ..., 0.0634, -0.0336, -0.0706]], + device='cuda:0'), grad: tensor([[-3.2842e-05, 1.1744e-06, 1.9446e-06, ..., -7.9721e-06, + 1.7229e-08, -3.9227e-06], + [ 2.9076e-06, 8.7684e-07, -1.0274e-05, ..., -1.4357e-05, + -6.5146e-07, 4.6082e-06], + [ 1.4149e-05, 7.6508e-07, 3.6042e-06, ..., 7.7188e-06, + 1.4156e-07, -4.1462e-06], + ..., + [ 2.1067e-06, 4.7684e-06, 1.4365e-05, ..., 2.8148e-05, + 2.6636e-07, 1.4920e-06], + [-2.0638e-05, 2.1793e-06, 4.1984e-06, ..., 1.3195e-05, + 7.1712e-08, 7.5363e-06], + [ 2.2784e-05, -9.6738e-05, -1.4400e-04, ..., -3.0160e-04, + 3.6787e-08, 5.9642e-06]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0228, -0.0184, -0.0299, 0.0018, 0.0023, 0.0326, 0.0139, -0.0203, + 0.0118, -0.0062], device='cuda:0'), grad: tensor([-1.1879e-04, -2.3246e-05, 8.7857e-05, 1.1772e-05, 2.0897e-04, + 9.0301e-05, 2.4259e-05, 5.1349e-05, -1.1146e-04, -2.2042e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 214.05, cls_loss 0.0037 cls_loss_mapping 0.0084 cls_loss_causal 0.5821 re_mapping 0.0084 re_causal 0.0262 /// teacc 98.85 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0006, -0.0208, -0.0222, ..., -0.1184, -0.0228, -0.0644], + [-0.0045, -0.0397, -0.0466, ..., 0.0102, 0.0025, -0.0786], + [ 0.0419, -0.0748, -0.0526, ..., 0.0255, -0.0875, 0.0425], + ..., + [ 0.0130, -0.0258, -0.0838, ..., -0.0713, 0.0137, 0.0842], + [ 0.0530, -0.0790, -0.0596, ..., -0.0649, -0.0250, -0.0415], + [-0.1158, 0.0071, 0.0882, ..., 0.0645, -0.0336, -0.0713]], + device='cuda:0'), grad: tensor([[ 6.7055e-06, 7.5363e-06, 7.0155e-05, ..., 5.5134e-05, + 1.2573e-08, 2.6543e-07], + [ 2.4699e-06, 6.8592e-07, 1.4281e-04, ..., 6.9022e-05, + -6.0536e-08, 7.5921e-06], + [-1.6065e-07, 1.0515e-06, 6.1877e-06, ..., 4.4480e-06, + 8.9873e-08, -6.8285e-06], + ..., + [ 6.5565e-07, -9.2527e-07, 1.0967e-05, ..., 8.0392e-06, + -3.1665e-07, -1.7555e-06], + [ 8.9109e-06, 9.3738e-07, 2.8849e-05, ..., 1.3687e-05, + 2.8871e-08, 3.0827e-07], + [ 1.9725e-06, -9.9689e-06, -5.2035e-05, ..., -5.1200e-05, + 5.3551e-08, 3.6461e-07]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0234, -0.0183, -0.0297, 0.0018, 0.0020, 0.0333, 0.0130, -0.0205, + 0.0115, -0.0055], device='cuda:0'), grad: tensor([ 1.5783e-04, 1.5676e-04, 3.8184e-08, 6.6534e-06, -1.6844e-04, + -4.0323e-05, -1.1243e-05, 1.6466e-05, 3.0994e-05, -1.4853e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 214.11, cls_loss 0.0047 cls_loss_mapping 0.0106 cls_loss_causal 0.5720 re_mapping 0.0089 re_causal 0.0263 /// teacc 98.92 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0006, -0.0211, -0.0240, ..., -0.1188, -0.0229, -0.0651], + [-0.0051, -0.0399, -0.0466, ..., 0.0106, 0.0025, -0.0787], + [ 0.0415, -0.0755, -0.0541, ..., 0.0251, -0.0884, 0.0426], + ..., + [ 0.0129, -0.0256, -0.0842, ..., -0.0717, 0.0137, 0.0845], + [ 0.0530, -0.0781, -0.0597, ..., -0.0658, -0.0256, -0.0413], + [-0.1146, 0.0071, 0.0887, ..., 0.0648, -0.0337, -0.0719]], + device='cuda:0'), grad: tensor([[ 5.5917e-06, 1.5795e-06, 6.9588e-06, ..., 2.9337e-06, + 7.2597e-07, 3.3118e-06], + [ 4.5188e-06, 5.6513e-06, 2.8685e-06, ..., 1.7677e-06, + 2.5574e-06, 1.5959e-05], + [-5.4508e-05, -4.9353e-05, 5.1595e-06, ..., 3.1423e-06, + 2.7493e-06, -2.8014e-04], + ..., + [ 5.1796e-05, 3.9577e-05, 7.0892e-06, ..., 7.1637e-06, + -8.5607e-06, 2.3437e-04], + [ 1.7548e-04, 5.0926e-04, 5.0485e-05, ..., 2.7820e-05, + 9.2480e-07, 3.3379e-04], + [ 9.0823e-06, 4.3921e-06, 6.1274e-05, ..., 6.3598e-05, + 1.8207e-06, 5.2713e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0246, -0.0183, -0.0302, 0.0012, 0.0021, 0.0336, 0.0128, -0.0212, + 0.0129, -0.0045], device='cuda:0'), grad: tensor([ 2.0325e-05, 3.8952e-05, -5.8603e-04, -1.6050e-03, -1.1790e-04, + 1.1808e-04, -3.3975e-04, 5.2738e-04, 1.8167e-03, 1.2863e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.17, cls_loss 0.0037 cls_loss_mapping 0.0087 cls_loss_causal 0.5837 re_mapping 0.0089 re_causal 0.0260 /// teacc 98.93 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0011, -0.0213, -0.0244, ..., -0.1194, -0.0236, -0.0656], + [-0.0053, -0.0402, -0.0464, ..., 0.0109, 0.0027, -0.0793], + [ 0.0414, -0.0760, -0.0549, ..., 0.0247, -0.0900, 0.0424], + ..., + [ 0.0129, -0.0252, -0.0847, ..., -0.0722, 0.0137, 0.0854], + [ 0.0524, -0.0787, -0.0607, ..., -0.0669, -0.0265, -0.0420], + [-0.1153, 0.0075, 0.0892, ..., 0.0653, -0.0342, -0.0726]], + device='cuda:0'), grad: tensor([[-1.7677e-06, 1.1688e-07, 5.5367e-07, ..., 6.6217e-07, + 2.3283e-09, 3.7719e-07], + [ 3.4692e-07, 3.1246e-07, -7.8324e-07, ..., -1.0068e-06, + 4.6566e-10, 2.4065e-06], + [-1.8664e-06, -8.8336e-07, 6.4541e-07, ..., 3.8883e-07, + 0.0000e+00, -8.6799e-06], + ..., + [ 1.7509e-07, -6.6031e-07, 6.3330e-07, ..., 7.8278e-07, + 1.3970e-09, 2.4848e-06], + [ 3.8054e-06, 2.8312e-07, 3.3975e-05, ..., 3.9339e-05, + 5.5879e-09, 4.1202e-06], + [-7.2643e-06, 1.3830e-07, -4.9740e-05, ..., -5.7638e-05, + 5.5879e-09, -4.4778e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0255, -0.0184, -0.0304, 0.0013, 0.0019, 0.0336, 0.0152, -0.0208, + 0.0121, -0.0046], device='cuda:0'), grad: tensor([-2.1122e-06, 1.0114e-06, -1.2495e-05, 6.2697e-06, 1.8567e-05, + 7.0594e-06, 3.9861e-06, 3.0044e-06, 5.0217e-05, -7.5400e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 214.43, cls_loss 0.0038 cls_loss_mapping 0.0068 cls_loss_causal 0.5816 re_mapping 0.0086 re_causal 0.0245 /// teacc 98.72 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0004, -0.0216, -0.0246, ..., -0.1194, -0.0243, -0.0667], + [-0.0056, -0.0409, -0.0463, ..., 0.0111, 0.0028, -0.0797], + [ 0.0412, -0.0762, -0.0559, ..., 0.0246, -0.0913, 0.0428], + ..., + [ 0.0126, -0.0245, -0.0847, ..., -0.0724, 0.0137, 0.0859], + [ 0.0525, -0.0792, -0.0612, ..., -0.0675, -0.0281, -0.0426], + [-0.1162, 0.0075, 0.0894, ..., 0.0655, -0.0347, -0.0736]], + device='cuda:0'), grad: tensor([[-2.8744e-05, 3.5064e-07, -1.3337e-05, ..., 9.5321e-07, + 1.9092e-08, 6.4308e-07], + [ 3.0510e-06, 1.6997e-06, 1.4063e-07, ..., -2.1346e-06, + -8.2189e-07, 3.9376e-06], + [ 2.9057e-06, 7.8278e-07, 2.7847e-06, ..., 9.0105e-07, + 9.9652e-08, -1.7524e-05], + ..., + [ 2.0768e-06, 7.6070e-06, 1.3098e-05, ..., 9.2760e-06, + 3.5064e-07, 8.3074e-06], + [ 3.0816e-05, 1.6578e-06, -9.1735e-08, ..., 3.8564e-05, + 1.0896e-07, -2.9877e-06], + [ 1.1340e-05, 1.3672e-05, 1.7062e-05, ..., 1.3016e-05, + 6.0536e-08, 2.2706e-06]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0250, -0.0192, -0.0302, 0.0011, 0.0016, 0.0330, 0.0159, -0.0200, + 0.0122, -0.0050], device='cuda:0'), grad: tensor([-1.1557e-04, 9.6634e-06, -5.9828e-06, 7.5758e-05, -5.4777e-05, + -1.0335e-04, 2.6003e-05, 3.9488e-05, 5.9634e-05, 6.9201e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 214.27, cls_loss 0.0039 cls_loss_mapping 0.0069 cls_loss_causal 0.5809 re_mapping 0.0086 re_causal 0.0256 /// teacc 98.76 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0002, -0.0219, -0.0249, ..., -0.1196, -0.0247, -0.0674], + [-0.0060, -0.0413, -0.0470, ..., 0.0108, 0.0029, -0.0798], + [ 0.0413, -0.0767, -0.0560, ..., 0.0249, -0.0920, 0.0429], + ..., + [ 0.0124, -0.0237, -0.0853, ..., -0.0729, 0.0136, 0.0868], + [ 0.0533, -0.0795, -0.0616, ..., -0.0680, -0.0284, -0.0430], + [-0.1167, 0.0074, 0.0895, ..., 0.0652, -0.0348, -0.0744]], + device='cuda:0'), grad: tensor([[-1.8515e-06, 1.1828e-07, 3.1851e-07, ..., 9.0804e-07, + 4.1910e-09, 1.5460e-07], + [-1.6810e-06, 5.9791e-07, -7.1302e-06, ..., -3.9697e-05, + -1.9651e-07, 8.1584e-07], + [ 8.4424e-07, 5.0152e-07, 7.6881e-07, ..., 2.8461e-06, + 5.4948e-08, -8.0513e-07], + ..., + [ 2.2771e-07, 6.3004e-07, 5.7332e-06, ..., 1.0580e-05, + 8.2422e-08, -2.1122e-06], + [ 6.0629e-07, 5.4343e-07, 1.9968e-06, ..., 8.5160e-06, + 7.4506e-09, 2.6682e-07], + [ 9.6112e-07, -1.1139e-05, -2.2650e-05, ..., -2.8431e-05, + 7.4506e-09, 4.1770e-07]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0250, -0.0194, -0.0300, 0.0005, 0.0018, 0.0331, 0.0165, -0.0199, + 0.0123, -0.0055], device='cuda:0'), grad: tensor([-3.1181e-06, -8.8811e-05, 8.3894e-06, 7.0482e-06, 4.5180e-05, + 2.4065e-05, 9.9316e-06, 1.2964e-05, 1.9390e-06, -1.7583e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 103---------------------------------------------------- +epoch 103, time 230.17, cls_loss 0.0033 cls_loss_mapping 0.0063 cls_loss_causal 0.5784 re_mapping 0.0083 re_causal 0.0253 /// teacc 98.98 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0002, -0.0226, -0.0253, ..., -0.1204, -0.0253, -0.0677], + [-0.0061, -0.0414, -0.0465, ..., 0.0112, 0.0033, -0.0801], + [ 0.0412, -0.0773, -0.0565, ..., 0.0254, -0.0939, 0.0432], + ..., + [ 0.0124, -0.0237, -0.0861, ..., -0.0728, 0.0135, 0.0871], + [ 0.0536, -0.0797, -0.0617, ..., -0.0686, -0.0298, -0.0433], + [-0.1170, 0.0080, 0.0897, ..., 0.0655, -0.0355, -0.0748]], + device='cuda:0'), grad: tensor([[-4.6194e-06, 1.6345e-07, 6.7800e-07, ..., 1.8897e-06, + 1.3970e-08, 1.7369e-07], + [ 1.5553e-06, 1.2675e-06, -2.4997e-06, ..., -1.4482e-06, + 1.0058e-07, 1.2526e-06], + [ 1.2591e-06, 7.6974e-07, 8.4192e-07, ..., 2.0489e-06, + 5.9605e-08, 2.3469e-07], + ..., + [ 9.3179e-07, -2.5611e-06, 2.2799e-06, ..., 6.5416e-06, + -8.3027e-07, -2.9635e-06], + [ 5.0515e-06, 1.1176e-06, 1.3160e-06, ..., 9.5889e-06, + 1.0990e-07, 7.6508e-07], + [ 2.3302e-06, 3.4785e-07, -1.0831e-06, ..., 4.2981e-07, + 3.8370e-07, 1.2424e-06]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0252, -0.0190, -0.0300, 0.0002, 0.0014, 0.0327, 0.0165, -0.0199, + 0.0126, -0.0054], device='cuda:0'), grad: tensor([-5.7891e-06, -1.1459e-05, 8.3223e-06, 2.4378e-04, 7.5903e-07, + -7.8154e-04, 5.0020e-04, 1.4126e-05, 2.2233e-05, 9.2760e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 214.09, cls_loss 0.0031 cls_loss_mapping 0.0069 cls_loss_causal 0.5849 re_mapping 0.0083 re_causal 0.0262 /// teacc 98.88 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 0.0002, -0.0228, -0.0260, ..., -0.1208, -0.0262, -0.0682], + [-0.0063, -0.0414, -0.0460, ..., 0.0117, 0.0034, -0.0802], + [ 0.0411, -0.0774, -0.0569, ..., 0.0252, -0.0970, 0.0434], + ..., + [ 0.0119, -0.0237, -0.0872, ..., -0.0740, 0.0141, 0.0873], + [ 0.0544, -0.0799, -0.0625, ..., -0.0701, -0.0321, -0.0436], + [-0.1174, 0.0078, 0.0889, ..., 0.0648, -0.0383, -0.0761]], + device='cuda:0'), grad: tensor([[ 2.0023e-07, 1.0245e-07, 1.0887e-06, ..., 8.3121e-07, + 3.7253e-09, 1.8021e-07], + [ 5.1036e-07, 1.6289e-06, -2.3600e-06, ..., -1.8338e-06, + -1.9092e-08, 2.4550e-06], + [ 7.7765e-08, 2.9430e-07, 1.4845e-06, ..., 9.8813e-07, + 9.7789e-09, -9.5461e-07], + ..., + [ 6.5146e-07, -3.5204e-06, 1.9502e-06, ..., 2.1607e-06, + -8.9407e-08, -4.5411e-06], + [ 6.3553e-06, 4.2049e-07, 4.4517e-06, ..., 1.0826e-05, + 1.5832e-08, 7.3761e-07], + [ 1.3314e-05, 6.8685e-07, 4.1537e-06, ..., 1.9431e-05, + 5.3551e-08, 1.9055e-06]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0250, -0.0186, -0.0302, 0.0002, 0.0028, 0.0323, 0.0163, -0.0203, + 0.0129, -0.0062], device='cuda:0'), grad: tensor([ 1.7518e-06, -5.9605e-06, 4.9323e-06, 1.3355e-06, 5.8934e-06, + -4.6819e-05, -6.8918e-06, -3.1274e-06, 1.7658e-05, 3.1203e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 214.09, cls_loss 0.0039 cls_loss_mapping 0.0062 cls_loss_causal 0.5789 re_mapping 0.0086 re_causal 0.0249 /// teacc 98.93 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0004, -0.0231, -0.0266, ..., -0.1218, -0.0285, -0.0686], + [-0.0069, -0.0416, -0.0466, ..., 0.0115, 0.0035, -0.0807], + [ 0.0412, -0.0778, -0.0574, ..., 0.0251, -0.0999, 0.0438], + ..., + [ 0.0123, -0.0236, -0.0876, ..., -0.0743, 0.0149, 0.0881], + [ 0.0543, -0.0802, -0.0633, ..., -0.0710, -0.0330, -0.0443], + [-0.1177, 0.0082, 0.0900, ..., 0.0654, -0.0401, -0.0773]], + device='cuda:0'), grad: tensor([[-2.1249e-05, 6.2911e-07, 5.0245e-07, ..., 5.6066e-07, + 2.3283e-08, 3.3434e-07], + [ 1.6978e-06, 8.0764e-06, 3.6862e-06, ..., 2.8089e-06, + 3.4459e-08, 4.1574e-06], + [ 1.3784e-06, 1.5609e-06, 5.8347e-07, ..., 6.6496e-07, + 1.2573e-08, 1.4976e-06], + ..., + [ 9.2201e-07, -3.0957e-06, 8.6054e-07, ..., 1.1297e-06, + 3.5390e-08, -7.1637e-06], + [ 7.0967e-06, 1.0140e-05, 5.7295e-06, ..., 7.9721e-06, + 1.6065e-07, 4.3064e-06], + [ 6.0312e-06, 1.7941e-05, 7.7710e-06, ..., 5.1185e-06, + 1.9558e-08, 8.3223e-06]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0250, -0.0191, -0.0299, 0.0001, 0.0021, 0.0329, 0.0163, -0.0199, + 0.0124, -0.0057], device='cuda:0'), grad: tensor([-3.9577e-05, 3.0756e-05, 7.4580e-06, -1.3208e-04, 2.0787e-06, + 3.7085e-06, 7.6108e-06, -5.1633e-06, 4.9800e-05, 7.5459e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 214.25, cls_loss 0.0033 cls_loss_mapping 0.0072 cls_loss_causal 0.5680 re_mapping 0.0081 re_causal 0.0241 /// teacc 98.97 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0008, -0.0232, -0.0270, ..., -0.1224, -0.0292, -0.0688], + [-0.0074, -0.0416, -0.0468, ..., 0.0111, 0.0035, -0.0811], + [ 0.0414, -0.0783, -0.0579, ..., 0.0259, -0.1003, 0.0445], + ..., + [ 0.0123, -0.0238, -0.0882, ..., -0.0743, 0.0149, 0.0881], + [ 0.0541, -0.0805, -0.0641, ..., -0.0719, -0.0336, -0.0450], + [-0.1178, 0.0083, 0.0905, ..., 0.0659, -0.0403, -0.0781]], + device='cuda:0'), grad: tensor([[ 6.1989e-06, 1.4808e-06, 5.1968e-06, ..., 2.2026e-07, + 0.0000e+00, 1.5721e-06], + [ 7.0333e-06, 1.9148e-06, 1.3104e-06, ..., -2.3982e-07, + 0.0000e+00, 1.1101e-05], + [-1.1034e-05, 7.9349e-07, 1.1185e-06, ..., 1.7369e-07, + 0.0000e+00, -2.1711e-05], + ..., + [-6.3190e-07, -7.4394e-06, 1.4035e-06, ..., 1.2536e-06, + 0.0000e+00, -7.0892e-06], + [ 6.4522e-06, 8.5123e-07, 6.4932e-06, ..., 8.9826e-07, + 0.0000e+00, 5.2676e-06], + [ 2.2296e-06, 1.7323e-07, -2.5332e-07, ..., -1.5711e-06, + 0.0000e+00, 9.4576e-07]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0248, -0.0192, -0.0294, 0.0007, 0.0021, 0.0324, 0.0160, -0.0199, + 0.0119, -0.0056], device='cuda:0'), grad: tensor([ 1.2405e-05, 3.9369e-05, -6.5327e-05, 9.2462e-06, -5.7697e-05, + 6.9439e-06, 2.8804e-05, -9.5963e-06, 2.9087e-05, 6.6645e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 213.81, cls_loss 0.0027 cls_loss_mapping 0.0065 cls_loss_causal 0.5698 re_mapping 0.0078 re_causal 0.0237 /// teacc 98.91 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0011, -0.0234, -0.0273, ..., -0.1229, -0.0293, -0.0691], + [-0.0078, -0.0418, -0.0466, ..., 0.0114, 0.0033, -0.0814], + [ 0.0412, -0.0789, -0.0584, ..., 0.0256, -0.1005, 0.0443], + ..., + [ 0.0125, -0.0233, -0.0888, ..., -0.0748, 0.0151, 0.0887], + [ 0.0541, -0.0809, -0.0652, ..., -0.0729, -0.0341, -0.0456], + [-0.1182, 0.0083, 0.0909, ..., 0.0664, -0.0403, -0.0788]], + device='cuda:0'), grad: tensor([[ 3.0119e-06, 6.6590e-07, 1.6153e-04, ..., 1.1593e-04, + 0.0000e+00, 3.9525e-06], + [ 2.3935e-06, 9.8273e-06, 5.3681e-06, ..., -1.9372e-07, + 0.0000e+00, 1.9401e-05], + [-1.6081e-04, 2.9476e-07, 4.0792e-06, ..., 3.1870e-06, + 0.0000e+00, -7.5638e-05], + ..., + [ 2.0619e-06, -9.8869e-06, -2.3711e-06, ..., 4.0568e-06, + 0.0000e+00, -1.7449e-05], + [ 1.1820e-04, 3.4180e-06, 3.3807e-07, ..., 6.7540e-06, + 0.0000e+00, 7.5340e-05], + [ 2.4773e-06, 1.1791e-06, -1.6165e-04, ..., -1.0830e-04, + 0.0000e+00, 3.4943e-06]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0246, -0.0191, -0.0300, 0.0004, 0.0021, 0.0321, 0.0164, -0.0197, + 0.0117, -0.0054], device='cuda:0'), grad: tensor([ 3.5572e-04, 5.1379e-05, -3.7527e-04, 9.5591e-06, -7.3537e-06, + -3.0875e-05, 6.3658e-05, -3.5971e-05, 3.1257e-04, -3.4428e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 214.10, cls_loss 0.0030 cls_loss_mapping 0.0073 cls_loss_causal 0.5551 re_mapping 0.0080 re_causal 0.0242 /// teacc 98.78 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 0.0016, -0.0234, -0.0279, ..., -0.1242, -0.0296, -0.0696], + [-0.0082, -0.0419, -0.0465, ..., 0.0119, 0.0033, -0.0817], + [ 0.0416, -0.0790, -0.0591, ..., 0.0251, -0.1008, 0.0448], + ..., + [ 0.0124, -0.0231, -0.0891, ..., -0.0755, 0.0151, 0.0890], + [ 0.0543, -0.0810, -0.0654, ..., -0.0734, -0.0342, -0.0460], + [-0.1184, 0.0081, 0.0901, ..., 0.0656, -0.0403, -0.0795]], + device='cuda:0'), grad: tensor([[-2.3525e-06, 7.0641e-07, 7.2177e-07, ..., 2.1122e-06, + 0.0000e+00, 2.5984e-06], + [ 3.3015e-07, 8.4285e-07, -1.9595e-05, ..., -2.2441e-05, + 0.0000e+00, 1.1167e-06], + [ 6.4028e-07, 2.3115e-06, 5.8226e-06, ..., 8.3074e-06, + 0.0000e+00, 5.8636e-06], + ..., + [ 4.5775e-07, 1.8617e-06, 7.5363e-06, ..., 1.0744e-05, + 0.0000e+00, 6.3609e-07], + [ 1.8897e-06, 1.9409e-06, 2.3842e-06, ..., 1.6227e-05, + 0.0000e+00, 3.3360e-06], + [ 5.6736e-06, -1.4409e-05, -4.0144e-05, ..., -2.5496e-05, + 0.0000e+00, 4.5076e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0247, -0.0191, -0.0298, 0.0009, 0.0035, 0.0316, 0.0159, -0.0198, + 0.0118, -0.0062], device='cuda:0'), grad: tensor([ 4.5914e-07, -4.5121e-05, 2.6137e-05, 2.9027e-05, 6.5267e-05, + -1.0717e-04, 5.8338e-06, 1.9684e-05, 2.5809e-05, -1.9908e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 213.91, cls_loss 0.0041 cls_loss_mapping 0.0080 cls_loss_causal 0.6073 re_mapping 0.0081 re_causal 0.0249 /// teacc 98.80 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0018, -0.0238, -0.0283, ..., -0.1251, -0.0296, -0.0702], + [-0.0084, -0.0421, -0.0459, ..., 0.0128, 0.0033, -0.0821], + [ 0.0413, -0.0797, -0.0601, ..., 0.0241, -0.1008, 0.0446], + ..., + [ 0.0129, -0.0228, -0.0898, ..., -0.0760, 0.0151, 0.0905], + [ 0.0542, -0.0814, -0.0655, ..., -0.0739, -0.0342, -0.0477], + [-0.1190, 0.0087, 0.0903, ..., 0.0661, -0.0403, -0.0806]], + device='cuda:0'), grad: tensor([[ 2.4274e-05, 3.4183e-05, 7.3433e-05, ..., 7.3910e-06, + 4.6566e-10, 1.1874e-06], + [-4.8965e-05, 7.5102e-05, 4.7207e-05, ..., -1.7896e-05, + 0.0000e+00, 2.9113e-06], + [ 6.3293e-06, 7.9647e-06, 1.0952e-05, ..., 7.9200e-06, + 0.0000e+00, -3.1907e-06], + ..., + [ 2.3786e-06, 3.1924e-04, 2.7013e-04, ..., 2.7537e-04, + 9.3132e-10, 5.0515e-06], + [-3.2604e-05, -3.5226e-05, -5.0783e-05, ..., 1.2212e-05, + 5.1223e-09, -1.2726e-05], + [ 1.1295e-05, -4.4298e-04, -3.7003e-04, ..., -3.7336e-04, + 4.6566e-10, -6.6794e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0246, -0.0186, -0.0308, 0.0008, 0.0029, 0.0319, 0.0157, -0.0191, + 0.0115, -0.0062], device='cuda:0'), grad: tensor([ 1.3471e-04, -8.6334e-07, 3.2723e-05, 1.8048e-04, 1.1235e-04, + 5.7101e-05, -2.1473e-05, 8.3733e-04, -2.1839e-04, -1.1139e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 214.08, cls_loss 0.0029 cls_loss_mapping 0.0058 cls_loss_causal 0.5760 re_mapping 0.0081 re_causal 0.0249 /// teacc 98.86 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0019, -0.0244, -0.0285, ..., -0.1263, -0.0296, -0.0708], + [-0.0083, -0.0422, -0.0457, ..., 0.0131, 0.0033, -0.0819], + [ 0.0414, -0.0804, -0.0607, ..., 0.0241, -0.1008, 0.0447], + ..., + [ 0.0127, -0.0226, -0.0904, ..., -0.0763, 0.0151, 0.0908], + [ 0.0542, -0.0818, -0.0654, ..., -0.0741, -0.0342, -0.0482], + [-0.1198, 0.0085, 0.0901, ..., 0.0659, -0.0403, -0.0814]], + device='cuda:0'), grad: tensor([[-3.8706e-06, 7.4971e-08, 8.6613e-08, ..., 2.0163e-07, + 5.1223e-09, 2.0489e-07], + [ 2.4233e-06, 2.2817e-06, -3.3453e-06, ..., -1.6559e-06, + 1.3970e-09, 6.3702e-06], + [ 1.2226e-05, 1.1466e-05, 1.5264e-06, ..., 7.1013e-07, + 1.3970e-09, 3.0756e-05], + ..., + [-1.4961e-05, -1.5318e-05, 8.1910e-07, ..., 8.2888e-07, + 5.5879e-09, -4.2468e-05], + [ 5.0515e-06, 1.4929e-06, 1.5348e-06, ..., 4.5784e-06, + 2.5146e-08, 3.7029e-06], + [ 1.2200e-06, -1.4789e-06, -3.6955e-06, ..., -2.3190e-06, + 1.7229e-08, 2.1514e-07]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0246, -0.0182, -0.0311, 0.0011, 0.0033, 0.0316, 0.0158, -0.0191, + 0.0115, -0.0067], device='cuda:0'), grad: tensor([-6.5267e-06, -1.7621e-06, 6.0111e-05, 2.1011e-06, 5.4315e-06, + -8.8885e-06, 1.8366e-06, -6.9797e-05, 1.8224e-05, -8.2096e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 214.12, cls_loss 0.0036 cls_loss_mapping 0.0070 cls_loss_causal 0.5455 re_mapping 0.0082 re_causal 0.0240 /// teacc 98.86 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 0.0020, -0.0246, -0.0287, ..., -0.1271, -0.0302, -0.0714], + [-0.0088, -0.0424, -0.0456, ..., 0.0132, 0.0034, -0.0823], + [ 0.0415, -0.0786, -0.0593, ..., 0.0253, -0.1012, 0.0460], + ..., + [ 0.0127, -0.0230, -0.0922, ..., -0.0777, 0.0151, 0.0909], + [ 0.0546, -0.0821, -0.0654, ..., -0.0750, -0.0344, -0.0486], + [-0.1207, 0.0092, 0.0911, ..., 0.0668, -0.0403, -0.0828]], + device='cuda:0'), grad: tensor([[-1.1288e-06, 7.7765e-08, 1.8850e-05, ..., 6.7465e-06, + 0.0000e+00, 3.1991e-07], + [ 9.2993e-07, 2.7165e-05, 4.0419e-06, ..., 2.7996e-06, + 0.0000e+00, 3.2514e-05], + [ 5.6904e-07, 3.8603e-07, 5.9940e-06, ..., 2.5164e-06, + 0.0000e+00, -2.5686e-06], + ..., + [-2.8266e-07, -3.4124e-05, 4.6641e-06, ..., 3.0063e-06, + 0.0000e+00, -4.1068e-05], + [-2.2613e-06, 3.7625e-06, 2.6934e-06, ..., 3.9265e-06, + 0.0000e+00, 2.0154e-06], + [ 5.2527e-06, -1.9763e-06, 2.4527e-05, ..., 1.2498e-06, + 0.0000e+00, 1.7313e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0247, -0.0184, -0.0300, 0.0010, 0.0021, 0.0329, 0.0153, -0.0196, + 0.0118, -0.0061], device='cuda:0'), grad: tensor([ 2.8417e-05, 8.4937e-05, 8.9109e-06, 1.7136e-05, -1.2231e-04, + -2.7101e-06, 1.8671e-05, -8.8990e-05, -1.3098e-05, 6.8784e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 214.19, cls_loss 0.0030 cls_loss_mapping 0.0062 cls_loss_causal 0.5819 re_mapping 0.0083 re_causal 0.0247 /// teacc 98.92 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0023, -0.0253, -0.0293, ..., -0.1276, -0.0308, -0.0722], + [-0.0084, -0.0425, -0.0449, ..., 0.0139, 0.0034, -0.0823], + [ 0.0412, -0.0790, -0.0598, ..., 0.0252, -0.1013, 0.0459], + ..., + [ 0.0125, -0.0227, -0.0927, ..., -0.0781, 0.0151, 0.0916], + [ 0.0544, -0.0824, -0.0659, ..., -0.0761, -0.0345, -0.0491], + [-0.1213, 0.0091, 0.0909, ..., 0.0665, -0.0403, -0.0833]], + device='cuda:0'), grad: tensor([[ 1.0226e-06, 2.3982e-07, 1.2610e-06, ..., 1.2685e-06, + 0.0000e+00, 3.1758e-07], + [ 7.1013e-07, 6.9570e-07, 1.6578e-07, ..., -4.3120e-07, + 0.0000e+00, 1.1772e-06], + [ 5.8822e-06, 5.3570e-06, 3.1106e-06, ..., 2.3600e-06, + 0.0000e+00, 1.0423e-05], + ..., + [-5.4315e-06, -1.0103e-05, 1.8775e-06, ..., 1.7956e-06, + 0.0000e+00, -2.0415e-05], + [-1.2860e-05, -1.7695e-06, -2.1219e-05, ..., -1.8269e-05, + 0.0000e+00, 2.1085e-06], + [ 8.8438e-06, 3.2037e-06, 2.3693e-05, ..., 1.8239e-05, + 0.0000e+00, 2.2799e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0245, -0.0181, -0.0306, 0.0006, 0.0024, 0.0338, 0.0149, -0.0193, + 0.0115, -0.0066], device='cuda:0'), grad: tensor([ 4.7609e-06, 2.2203e-06, 2.6926e-05, 2.2665e-05, -2.8849e-05, + 2.7314e-05, 1.2368e-05, -2.6017e-05, -1.4639e-04, 1.0520e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 214.11, cls_loss 0.0029 cls_loss_mapping 0.0064 cls_loss_causal 0.5467 re_mapping 0.0080 re_causal 0.0241 /// teacc 98.82 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 0.0034, -0.0256, -0.0286, ..., -0.1272, -0.0308, -0.0725], + [-0.0085, -0.0426, -0.0437, ..., 0.0139, 0.0034, -0.0826], + [ 0.0422, -0.0793, -0.0605, ..., 0.0251, -0.1014, 0.0468], + ..., + [ 0.0109, -0.0228, -0.0931, ..., -0.0783, 0.0151, 0.0920], + [ 0.0547, -0.0824, -0.0661, ..., -0.0765, -0.0346, -0.0498], + [-0.1220, 0.0093, 0.0908, ..., 0.0664, -0.0404, -0.0840]], + device='cuda:0'), grad: tensor([[-5.6997e-07, 7.5810e-07, 5.5879e-08, ..., 1.6615e-06, + 0.0000e+00, 2.6897e-06], + [ 6.2119e-07, 7.5204e-07, -9.6206e-07, ..., -6.2538e-07, + 0.0000e+00, 1.2536e-06], + [ 7.9488e-07, 3.9600e-06, 4.6194e-07, ..., 9.3132e-06, + 0.0000e+00, -8.5607e-06], + ..., + [ 5.2415e-06, 5.5470e-06, 9.7789e-07, ..., 1.3448e-05, + 0.0000e+00, 1.2904e-05], + [-6.7465e-06, 1.6298e-06, 2.2398e-07, ..., 4.4890e-06, + 0.0000e+00, 6.2995e-06], + [ 1.5665e-06, 1.7472e-06, 4.1872e-06, ..., 4.2208e-06, + 0.0000e+00, 1.3988e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0231, -0.0177, -0.0296, 0.0007, 0.0025, 0.0339, 0.0131, -0.0194, + 0.0111, -0.0069], device='cuda:0'), grad: tensor([ 1.4268e-06, 5.8813e-07, 6.1542e-06, 2.4939e-04, 2.1625e-04, + -8.6069e-04, 3.3784e-04, 3.4660e-05, 2.6878e-06, 1.1735e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 214.17, cls_loss 0.0031 cls_loss_mapping 0.0072 cls_loss_causal 0.5638 re_mapping 0.0075 re_causal 0.0235 /// teacc 98.98 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0034, -0.0261, -0.0296, ..., -0.1279, -0.0312, -0.0732], + [-0.0083, -0.0428, -0.0436, ..., 0.0139, 0.0034, -0.0835], + [ 0.0420, -0.0795, -0.0609, ..., 0.0248, -0.1016, 0.0471], + ..., + [ 0.0107, -0.0223, -0.0932, ..., -0.0776, 0.0153, 0.0934], + [ 0.0533, -0.0827, -0.0673, ..., -0.0773, -0.0348, -0.0503], + [-0.1225, 0.0098, 0.0912, ..., 0.0667, -0.0409, -0.0857]], + device='cuda:0'), grad: tensor([[-3.4124e-05, 4.5029e-07, -1.0151e-07, ..., 6.0908e-07, + 1.8626e-09, 3.9339e-06], + [ 1.0431e-06, 8.5775e-07, -3.1665e-08, ..., -1.9670e-06, + 4.6566e-10, 5.2005e-06], + [-2.2985e-06, -5.9605e-06, 3.8976e-07, ..., 1.3160e-06, + 4.6566e-10, -7.1466e-05], + ..., + [ 5.8580e-07, -5.4622e-07, 1.0654e-06, ..., 1.8561e-06, + 0.0000e+00, 2.2262e-05], + [ 3.7868e-06, 1.0608e-06, 7.2410e-07, ..., 4.1649e-06, + 1.3970e-09, 1.4000e-05], + [ 2.0359e-06, 5.4669e-07, -1.8515e-06, ..., -1.9670e-06, + 0.0000e+00, 2.7139e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0234, -0.0179, -0.0296, -0.0008, 0.0019, 0.0343, 0.0144, -0.0185, + 0.0101, -0.0066], device='cuda:0'), grad: tensor([-4.7207e-05, 2.3954e-06, -8.3745e-05, 2.0313e-04, 2.4661e-06, + -1.9717e-04, 5.4747e-05, 2.6435e-05, 3.2157e-05, 6.9477e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 214.14, cls_loss 0.0029 cls_loss_mapping 0.0062 cls_loss_causal 0.5632 re_mapping 0.0076 re_causal 0.0230 /// teacc 98.98 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 0.0037, -0.0265, -0.0300, ..., -0.1285, -0.0314, -0.0743], + [-0.0086, -0.0430, -0.0436, ..., 0.0139, 0.0034, -0.0838], + [ 0.0419, -0.0802, -0.0615, ..., 0.0250, -0.1020, 0.0472], + ..., + [ 0.0105, -0.0219, -0.0934, ..., -0.0778, 0.0154, 0.0943], + [ 0.0532, -0.0832, -0.0678, ..., -0.0778, -0.0349, -0.0510], + [-0.1230, 0.0098, 0.0913, ..., 0.0667, -0.0411, -0.0872]], + device='cuda:0'), grad: tensor([[ 3.1013e-06, 1.0710e-07, 4.3586e-06, ..., 5.4017e-07, + 4.6566e-10, 1.2154e-06], + [ 1.0151e-06, 6.3470e-07, 1.8608e-06, ..., 8.9873e-07, + 4.6566e-09, 2.2817e-06], + [ 3.1665e-07, 8.8941e-07, 2.7120e-06, ..., 9.5088e-07, + 3.2596e-09, -2.3916e-05], + ..., + [ 1.3411e-06, 1.9949e-06, 3.9563e-06, ..., 4.3511e-06, + -2.7474e-08, 1.7375e-05], + [ 1.3635e-05, 1.5097e-06, 1.4715e-05, ..., 5.7369e-06, + 1.8626e-09, 1.7909e-06], + [ 9.2713e-07, 9.3728e-06, 4.9204e-05, ..., 5.2392e-05, + 1.0245e-08, 2.1812e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0235, -0.0177, -0.0299, -0.0028, 0.0019, 0.0361, 0.0147, -0.0182, + 0.0097, -0.0068], device='cuda:0'), grad: tensor([ 1.1325e-05, 6.9886e-06, -1.9073e-05, -3.2168e-06, -7.9036e-05, + 2.3935e-06, -7.7426e-05, 2.7463e-05, 4.5896e-05, 8.4639e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 214.19, cls_loss 0.0036 cls_loss_mapping 0.0067 cls_loss_causal 0.5687 re_mapping 0.0076 re_causal 0.0227 /// teacc 98.96 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 0.0031, -0.0267, -0.0304, ..., -0.1291, -0.0317, -0.0751], + [-0.0087, -0.0428, -0.0433, ..., 0.0147, 0.0036, -0.0841], + [ 0.0419, -0.0818, -0.0616, ..., 0.0251, -0.1025, 0.0470], + ..., + [ 0.0097, -0.0222, -0.0947, ..., -0.0793, 0.0152, 0.0950], + [ 0.0538, -0.0833, -0.0681, ..., -0.0783, -0.0350, -0.0514], + [-0.1234, 0.0100, 0.0914, ..., 0.0668, -0.0412, -0.0878]], + device='cuda:0'), grad: tensor([[-6.6459e-06, 3.2131e-08, 2.9150e-06, ..., 7.2131e-07, + 9.3132e-10, 2.0675e-07], + [ 1.8971e-06, 1.8999e-07, 2.8219e-07, ..., -1.6643e-06, + -4.1444e-08, 8.2515e-07], + [ 4.3660e-06, 9.8255e-08, 2.7269e-06, ..., 1.4650e-06, + 3.2596e-09, -2.7064e-06], + ..., + [ 1.0161e-06, 3.3528e-07, 1.7388e-06, ..., 2.0582e-06, + 1.0710e-08, 1.1735e-06], + [-3.2067e-05, 2.2631e-07, -3.1870e-06, ..., -5.8204e-05, + 6.0536e-09, 2.8126e-07], + [ 3.1263e-05, -7.5111e-07, 7.0073e-06, ..., 4.8906e-05, + 3.7253e-09, 2.4633e-07]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0239, -0.0172, -0.0304, -0.0022, 0.0020, 0.0357, 0.0152, -0.0187, + 0.0099, -0.0070], device='cuda:0'), grad: tensor([-1.5587e-05, 2.2613e-06, 1.0632e-05, 3.2596e-06, -1.1548e-05, + 2.1920e-05, -1.2875e-05, 8.7544e-06, -1.3316e-04, 1.2612e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 214.04, cls_loss 0.0026 cls_loss_mapping 0.0054 cls_loss_causal 0.5765 re_mapping 0.0071 re_causal 0.0219 /// teacc 98.91 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0040, -0.0269, -0.0305, ..., -0.1297, -0.0319, -0.0760], + [-0.0090, -0.0429, -0.0433, ..., 0.0147, 0.0037, -0.0848], + [ 0.0418, -0.0824, -0.0621, ..., 0.0251, -0.1033, 0.0469], + ..., + [ 0.0097, -0.0222, -0.0955, ..., -0.0803, 0.0151, 0.0956], + [ 0.0536, -0.0835, -0.0682, ..., -0.0786, -0.0355, -0.0517], + [-0.1241, 0.0101, 0.0915, ..., 0.0670, -0.0411, -0.0883]], + device='cuda:0'), grad: tensor([[-1.5944e-05, 6.9151e-08, 2.1537e-07, ..., 2.5034e-06, + 0.0000e+00, 1.7490e-06], + [ 2.8498e-06, 1.8161e-07, 1.6461e-07, ..., -2.7716e-06, + 0.0000e+00, 5.2080e-06], + [-8.4862e-06, 8.1724e-07, 1.4864e-06, ..., 3.9935e-06, + 0.0000e+00, -4.0621e-05], + ..., + [ 6.0908e-06, -1.4808e-07, 1.3383e-06, ..., 6.1542e-06, + 0.0000e+00, 1.3314e-05], + [ 3.8981e-05, 7.5903e-07, 6.7847e-07, ..., 2.7806e-05, + 0.0000e+00, 1.5303e-05], + [ 8.3670e-06, -1.7695e-06, -6.7018e-06, ..., 5.5544e-06, + 0.0000e+00, 8.0559e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0234, -0.0175, -0.0306, -0.0017, 0.0022, 0.0352, 0.0153, -0.0186, + 0.0099, -0.0071], device='cuda:0'), grad: tensor([-4.1813e-05, -1.4424e-05, -8.7380e-05, 9.7632e-05, 5.4948e-06, + -1.7166e-04, 9.3430e-06, 6.9022e-05, 1.0234e-04, 3.1382e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 213.99, cls_loss 0.0024 cls_loss_mapping 0.0055 cls_loss_causal 0.5724 re_mapping 0.0073 re_causal 0.0232 /// teacc 98.91 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0040, -0.0269, -0.0309, ..., -0.1302, -0.0324, -0.0762], + [-0.0092, -0.0432, -0.0433, ..., 0.0143, 0.0036, -0.0850], + [ 0.0418, -0.0827, -0.0628, ..., 0.0258, -0.1042, 0.0474], + ..., + [ 0.0095, -0.0220, -0.0959, ..., -0.0808, 0.0151, 0.0958], + [ 0.0538, -0.0838, -0.0688, ..., -0.0797, -0.0358, -0.0522], + [-0.1242, 0.0093, 0.0915, ..., 0.0670, -0.0411, -0.0895]], + device='cuda:0'), grad: tensor([[-2.4796e-07, -4.3958e-07, 9.3132e-10, ..., 3.1060e-07, + 0.0000e+00, 1.1232e-06], + [ 5.4985e-06, 1.3430e-06, 2.2966e-06, ..., 5.2154e-06, + 0.0000e+00, 4.0941e-06], + [ 1.5303e-05, 1.1921e-07, 5.3555e-05, ..., 1.0282e-06, + 0.0000e+00, -2.2471e-05], + ..., + [ 1.2368e-06, -4.8988e-06, -1.5907e-06, ..., 9.0385e-07, + 0.0000e+00, -2.3912e-07], + [-7.1675e-06, -9.1502e-08, -3.5744e-06, ..., -9.0674e-06, + 0.0000e+00, 5.7109e-06], + [ 4.9956e-06, 2.6263e-06, 6.9197e-07, ..., -1.6754e-06, + 0.0000e+00, 6.2399e-06]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0235, -0.0176, -0.0304, -0.0011, 0.0031, 0.0347, 0.0147, -0.0186, + 0.0096, -0.0074], device='cuda:0'), grad: tensor([-6.9104e-07, 2.6837e-05, 7.0155e-05, 5.7071e-06, 3.9458e-05, + 3.6433e-06, -1.5116e-04, 1.9521e-06, -1.9431e-05, 2.3484e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 214.01, cls_loss 0.0024 cls_loss_mapping 0.0064 cls_loss_causal 0.5443 re_mapping 0.0079 re_causal 0.0237 /// teacc 98.77 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0038, -0.0271, -0.0311, ..., -0.1306, -0.0324, -0.0768], + [-0.0092, -0.0433, -0.0429, ..., 0.0153, 0.0036, -0.0853], + [ 0.0425, -0.0829, -0.0633, ..., 0.0257, -0.1042, 0.0480], + ..., + [ 0.0092, -0.0219, -0.0972, ..., -0.0824, 0.0151, 0.0960], + [ 0.0538, -0.0838, -0.0709, ..., -0.0814, -0.0358, -0.0531], + [-0.1245, 0.0095, 0.0918, ..., 0.0674, -0.0411, -0.0893]], + device='cuda:0'), grad: tensor([[ 4.8093e-06, 1.1828e-07, 1.0088e-05, ..., 8.4788e-06, + 0.0000e+00, 1.8254e-07], + [ 1.0002e-04, 9.3412e-07, 2.6655e-04, ..., 3.0971e-04, + 0.0000e+00, 3.2596e-07], + [ 2.5984e-06, 6.7055e-08, 6.2250e-06, ..., 8.2478e-06, + 0.0000e+00, -1.9949e-06], + ..., + [ 4.6613e-07, 4.0382e-06, 2.0072e-05, ..., 1.8016e-05, + 0.0000e+00, 2.2966e-06], + [-1.2410e-04, 1.7378e-06, -3.1972e-04, ..., -3.7837e-04, + 0.0000e+00, 6.5565e-07], + [ 2.8759e-06, -1.0341e-05, -5.0008e-05, ..., -4.0591e-05, + 0.0000e+00, -2.9691e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0237, -0.0174, -0.0298, -0.0002, 0.0037, 0.0334, 0.0150, -0.0190, + 0.0086, -0.0073], device='cuda:0'), grad: tensor([ 2.6658e-05, 1.3762e-03, 3.4392e-05, 3.3677e-05, 4.0293e-05, + 5.4777e-05, 1.6141e-04, 5.5611e-05, -1.7290e-03, -5.4210e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 213.94, cls_loss 0.0040 cls_loss_mapping 0.0072 cls_loss_causal 0.5551 re_mapping 0.0078 re_causal 0.0220 /// teacc 98.84 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0041, -0.0282, -0.0314, ..., -0.1329, -0.0325, -0.0760], + [-0.0096, -0.0443, -0.0431, ..., 0.0153, 0.0028, -0.0872], + [ 0.0424, -0.0833, -0.0644, ..., 0.0251, -0.1044, 0.0485], + ..., + [ 0.0083, -0.0213, -0.0988, ..., -0.0836, 0.0160, 0.0974], + [ 0.0543, -0.0840, -0.0705, ..., -0.0817, -0.0360, -0.0538], + [-0.1255, 0.0104, 0.0930, ..., 0.0684, -0.0412, -0.0892]], + device='cuda:0'), grad: tensor([[-1.0971e-06, 3.2131e-08, 1.6214e-06, ..., 1.2890e-06, + 0.0000e+00, 8.4750e-08], + [-2.5611e-06, 3.5763e-07, -7.0095e-05, ..., -5.6297e-05, + 0.0000e+00, 1.2992e-06], + [ 5.9884e-07, 6.3470e-07, 3.7532e-06, ..., 9.4855e-07, + 0.0000e+00, 5.6764e-07], + ..., + [ 4.2608e-07, -9.7696e-07, 5.4017e-06, ..., 4.3623e-06, + 0.0000e+00, -2.1402e-06], + [ 1.9297e-06, 2.6077e-07, 4.6253e-05, ..., 3.7223e-05, + 0.0000e+00, 5.8068e-07], + [ 5.3504e-07, -1.8999e-07, -4.1025e-07, ..., -2.1374e-07, + 0.0000e+00, 1.2992e-07]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0237, -0.0184, -0.0302, -0.0005, 0.0028, 0.0350, 0.0143, -0.0185, + 0.0090, -0.0065], device='cuda:0'), grad: tensor([ 3.5809e-07, -1.6320e-04, 7.1041e-06, -1.2644e-05, -1.2720e-04, + 1.2226e-05, 1.6081e-04, 9.8199e-06, 1.0979e-04, 3.0268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 213.88, cls_loss 0.0025 cls_loss_mapping 0.0051 cls_loss_causal 0.5645 re_mapping 0.0074 re_causal 0.0228 /// teacc 98.81 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0049, -0.0285, -0.0330, ..., -0.1338, -0.0326, -0.0757], + [-0.0098, -0.0444, -0.0424, ..., 0.0161, 0.0028, -0.0876], + [ 0.0423, -0.0837, -0.0646, ..., 0.0255, -0.1044, 0.0486], + ..., + [ 0.0081, -0.0215, -0.0999, ..., -0.0846, 0.0160, 0.0976], + [ 0.0550, -0.0842, -0.0706, ..., -0.0818, -0.0362, -0.0541], + [-0.1254, 0.0105, 0.0934, ..., 0.0686, -0.0412, -0.0897]], + device='cuda:0'), grad: tensor([[-3.5968e-06, 2.6589e-07, 1.3895e-06, ..., 1.8766e-07, + 0.0000e+00, 5.7090e-07], + [ 1.2303e-06, 3.7439e-07, -7.3714e-07, ..., -1.3895e-06, + 0.0000e+00, 8.7311e-07], + [ 1.8328e-06, 1.4547e-06, 7.3900e-07, ..., 6.6077e-07, + 0.0000e+00, 6.7148e-07], + ..., + [ 1.7602e-07, -1.6410e-06, 6.1840e-07, ..., 5.5414e-07, + 0.0000e+00, -3.8967e-06], + [-5.3793e-06, 3.8790e-07, 9.8627e-07, ..., -1.4547e-06, + 0.0000e+00, 5.9232e-07], + [ 5.7928e-07, -4.8056e-07, -2.1048e-06, ..., -2.5760e-06, + 0.0000e+00, 1.5786e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0238, -0.0181, -0.0304, -0.0003, 0.0025, 0.0346, 0.0141, -0.0187, + 0.0093, -0.0062], device='cuda:0'), grad: tensor([-4.0941e-06, -6.2864e-08, 7.2382e-06, -2.4401e-06, 1.9018e-06, + 2.1726e-05, -5.8189e-06, -3.2000e-06, -1.6138e-05, 8.9779e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 214.05, cls_loss 0.0025 cls_loss_mapping 0.0050 cls_loss_causal 0.5623 re_mapping 0.0070 re_causal 0.0214 /// teacc 98.85 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0047, -0.0290, -0.0336, ..., -0.1343, -0.0326, -0.0763], + [-0.0099, -0.0446, -0.0416, ..., 0.0170, 0.0028, -0.0880], + [ 0.0423, -0.0840, -0.0658, ..., 0.0253, -0.1044, 0.0489], + ..., + [ 0.0079, -0.0212, -0.1005, ..., -0.0851, 0.0160, 0.0982], + [ 0.0562, -0.0843, -0.0704, ..., -0.0820, -0.0362, -0.0545], + [-0.1267, 0.0103, 0.0932, ..., 0.0684, -0.0412, -0.0910]], + device='cuda:0'), grad: tensor([[ 1.3486e-06, 9.4017e-07, 7.6881e-07, ..., 9.2946e-07, + 2.5705e-07, 2.4512e-06], + [ 1.8571e-06, 1.7834e-04, 6.5826e-06, ..., 1.3895e-06, + 5.4855e-07, 2.6870e-04], + [ 3.4459e-07, 2.0847e-05, 1.5134e-07, ..., 8.8476e-08, + 1.3039e-08, 1.0717e-04], + ..., + [ 1.6745e-06, -1.7834e-04, -6.6198e-06, ..., 2.5015e-06, + 5.8860e-07, -2.5153e-04], + [ 1.5488e-06, 7.5325e-06, 2.4680e-07, ..., 1.8561e-06, + 4.9593e-07, 3.1382e-05], + [ 1.5246e-06, 1.9372e-06, -5.5647e-07, ..., 8.6054e-07, + 4.2375e-07, 3.8221e-06]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0241, -0.0177, -0.0307, -0.0002, 0.0024, 0.0345, 0.0146, -0.0185, + 0.0100, -0.0069], device='cuda:0'), grad: tensor([ 6.7912e-06, 7.5960e-04, 1.8704e-04, -3.1543e-04, 5.6578e-07, + 2.4274e-05, -9.4390e-07, -7.3338e-04, 6.0052e-05, 1.1139e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 214.23, cls_loss 0.0027 cls_loss_mapping 0.0060 cls_loss_causal 0.5614 re_mapping 0.0072 re_causal 0.0216 /// teacc 98.86 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0053, -0.0293, -0.0332, ..., -0.1347, -0.0326, -0.0768], + [-0.0101, -0.0449, -0.0414, ..., 0.0171, 0.0028, -0.0883], + [ 0.0423, -0.0843, -0.0667, ..., 0.0252, -0.1045, 0.0488], + ..., + [ 0.0076, -0.0212, -0.1022, ..., -0.0862, 0.0160, 0.0987], + [ 0.0562, -0.0846, -0.0711, ..., -0.0828, -0.0363, -0.0549], + [-0.1278, 0.0105, 0.0931, ..., 0.0685, -0.0412, -0.0911]], + device='cuda:0'), grad: tensor([[ 5.1036e-07, 3.6275e-07, 1.8608e-06, ..., 1.6904e-06, + 0.0000e+00, 5.1176e-07], + [ 1.2619e-06, 1.0207e-06, -3.4012e-06, ..., -2.6189e-06, + 0.0000e+00, 2.7418e-06], + [ 1.0077e-06, 1.1278e-06, 1.5479e-06, ..., 1.2154e-06, + 0.0000e+00, 2.7847e-06], + ..., + [ 3.3621e-07, -4.2804e-06, 3.6284e-06, ..., 2.9150e-06, + 0.0000e+00, -9.6634e-06], + [-2.0474e-05, 1.0785e-06, 2.3358e-06, ..., 9.1866e-06, + 0.0000e+00, 1.6764e-06], + [ 1.1837e-06, 2.3562e-06, -1.9725e-06, ..., -2.0087e-05, + 0.0000e+00, 3.5334e-06]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0229, -0.0177, -0.0312, -0.0002, 0.0025, 0.0341, 0.0151, -0.0185, + 0.0097, -0.0075], device='cuda:0'), grad: tensor([ 6.4522e-06, -4.0531e-06, 1.1690e-05, 6.6012e-06, -1.8522e-05, + 5.5462e-05, 1.0550e-05, -3.9116e-06, -6.5863e-05, 1.5888e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 214.23, cls_loss 0.0032 cls_loss_mapping 0.0052 cls_loss_causal 0.5463 re_mapping 0.0070 re_causal 0.0214 /// teacc 98.87 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0058, -0.0296, -0.0337, ..., -0.1355, -0.0332, -0.0768], + [-0.0109, -0.0450, -0.0415, ..., 0.0170, 0.0028, -0.0888], + [ 0.0427, -0.0846, -0.0667, ..., 0.0249, -0.1052, 0.0498], + ..., + [ 0.0069, -0.0211, -0.1029, ..., -0.0868, 0.0159, 0.0987], + [ 0.0570, -0.0848, -0.0713, ..., -0.0833, -0.0382, -0.0555], + [-0.1286, 0.0102, 0.0930, ..., 0.0684, -0.0412, -0.0921]], + device='cuda:0'), grad: tensor([[ 6.6962e-07, 4.4703e-08, 5.1856e-06, ..., 1.2675e-06, + 0.0000e+00, 3.4552e-07], + [ 7.4133e-07, 2.5984e-07, 3.5942e-05, ..., 1.2554e-05, + 0.0000e+00, 3.6489e-06], + [ 1.9791e-07, 1.7742e-07, 1.3091e-05, ..., 4.6231e-06, + 0.0000e+00, -1.0338e-07], + ..., + [ 1.9884e-07, -5.4063e-07, 5.9828e-06, ..., 2.3041e-06, + 0.0000e+00, -4.8149e-07], + [-5.8338e-06, -9.3132e-09, 5.7742e-06, ..., 7.4087e-07, + 2.7940e-09, 8.5775e-07], + [ 6.3889e-07, 1.5553e-07, 5.3570e-06, ..., 1.9893e-06, + 4.6566e-10, 5.5460e-07]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0225, -0.0181, -0.0305, -0.0001, 0.0028, 0.0342, 0.0151, -0.0191, + 0.0103, -0.0079], device='cuda:0'), grad: tensor([ 7.5623e-06, 7.0512e-05, 2.4095e-05, 1.8626e-06, -2.1636e-04, + 1.3553e-05, 7.8261e-05, 1.0692e-05, -2.5891e-06, 1.2375e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 213.97, cls_loss 0.0033 cls_loss_mapping 0.0077 cls_loss_causal 0.5647 re_mapping 0.0075 re_causal 0.0222 /// teacc 98.95 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0059, -0.0298, -0.0364, ..., -0.1371, -0.0333, -0.0774], + [-0.0110, -0.0448, -0.0406, ..., 0.0179, 0.0028, -0.0888], + [ 0.0427, -0.0853, -0.0678, ..., 0.0245, -0.1053, 0.0496], + ..., + [ 0.0066, -0.0216, -0.1051, ..., -0.0883, 0.0164, 0.0986], + [ 0.0555, -0.0853, -0.0743, ..., -0.0868, -0.0383, -0.0567], + [-0.1266, 0.0106, 0.0952, ..., 0.0697, -0.0426, -0.0930]], + device='cuda:0'), grad: tensor([[ 1.7524e-05, 1.7732e-06, 1.8105e-05, ..., 4.8578e-06, + 4.6566e-10, 5.4650e-06], + [ 2.6613e-05, 4.9993e-06, 2.5406e-05, ..., 4.2051e-05, + 5.5879e-09, 1.3225e-05], + [ 1.7462e-06, 9.2015e-07, 8.7963e-07, ..., 1.6242e-06, + 1.3970e-09, -4.6283e-05], + ..., + [ 1.0887e-06, -5.1856e-05, 2.0005e-06, ..., 3.3509e-06, + -2.5146e-08, -6.7532e-05], + [ 7.1704e-05, 1.0848e-05, 8.6069e-05, ..., -3.5793e-05, + 2.7940e-09, 3.9786e-06], + [ 3.5465e-06, 2.8938e-05, -1.4111e-05, ..., -1.3597e-05, + 1.8626e-09, 5.2691e-05]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0243, -0.0177, -0.0309, 0.0003, 0.0032, 0.0344, 0.0151, -0.0198, + 0.0079, -0.0056], device='cuda:0'), grad: tensor([ 9.3460e-05, 1.9336e-04, -6.7532e-05, 6.1095e-05, 6.3419e-05, + 7.3731e-05, -7.0095e-04, -1.6499e-04, 3.2210e-04, 1.2612e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 214.17, cls_loss 0.0035 cls_loss_mapping 0.0068 cls_loss_causal 0.5348 re_mapping 0.0071 re_causal 0.0211 /// teacc 98.94 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0066, -0.0301, -0.0364, ..., -0.1375, -0.0333, -0.0779], + [-0.0118, -0.0456, -0.0419, ..., 0.0165, 0.0028, -0.0896], + [ 0.0414, -0.0854, -0.0690, ..., 0.0240, -0.1054, 0.0498], + ..., + [ 0.0066, -0.0213, -0.1063, ..., -0.0890, 0.0166, 0.0998], + [ 0.0530, -0.0858, -0.0776, ..., -0.0895, -0.0384, -0.0571], + [-0.1243, 0.0111, 0.0974, ..., 0.0713, -0.0433, -0.0949]], + device='cuda:0'), grad: tensor([[ 3.0845e-06, 2.4028e-07, 1.5553e-06, ..., 6.0862e-07, + 0.0000e+00, 5.6066e-07], + [ 1.4175e-06, 3.0696e-05, 2.2557e-06, ..., 1.1986e-06, + 0.0000e+00, 4.8846e-05], + [-1.8543e-06, 3.9954e-07, 1.0207e-06, ..., 6.6869e-07, + 0.0000e+00, -1.6928e-05], + ..., + [ 3.2652e-06, -3.3349e-05, 1.3830e-06, ..., 1.0980e-06, + 0.0000e+00, -3.9160e-05], + [ 1.4398e-06, 1.4724e-06, 3.1330e-06, ..., -3.1199e-07, + 0.0000e+00, 1.9092e-06], + [ 1.4519e-06, 1.2675e-06, 1.3700e-06, ..., 1.4585e-06, + 0.0000e+00, 2.7716e-06]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0238, -0.0190, -0.0314, -0.0001, 0.0037, 0.0347, 0.0151, -0.0193, + 0.0051, -0.0034], device='cuda:0'), grad: tensor([ 9.7081e-06, 1.7536e-04, -1.2182e-05, 6.6608e-06, -1.8433e-05, + 1.7248e-06, -1.8224e-05, -1.6451e-04, 5.5619e-06, 1.4104e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 214.29, cls_loss 0.0022 cls_loss_mapping 0.0049 cls_loss_causal 0.5456 re_mapping 0.0074 re_causal 0.0219 /// teacc 98.87 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0065, -0.0304, -0.0365, ..., -0.1380, -0.0334, -0.0788], + [-0.0121, -0.0459, -0.0419, ..., 0.0166, 0.0028, -0.0899], + [ 0.0416, -0.0857, -0.0691, ..., 0.0240, -0.1069, 0.0502], + ..., + [ 0.0064, -0.0212, -0.1067, ..., -0.0895, 0.0165, 0.1000], + [ 0.0535, -0.0861, -0.0776, ..., -0.0896, -0.0385, -0.0574], + [-0.1245, 0.0111, 0.0974, ..., 0.0713, -0.0434, -0.0957]], + device='cuda:0'), grad: tensor([[-1.3178e-07, 1.9819e-06, 2.4084e-06, ..., 3.6824e-06, + 1.0198e-07, 7.0501e-07], + [ 9.5926e-08, 4.2655e-07, -3.9749e-06, ..., -2.9787e-05, + 3.5390e-08, 3.5856e-07], + [-3.2596e-08, 9.4995e-08, 1.2005e-06, ..., 2.7008e-06, + 9.3132e-09, -2.5798e-06], + ..., + [ 2.9150e-07, 6.2995e-06, 6.5751e-06, ..., 1.6034e-05, + 2.1253e-06, 2.3395e-06], + [ 1.9697e-07, 4.3511e-06, 6.9439e-06, ..., 1.6734e-05, + 5.8627e-07, 1.6578e-06], + [ 8.1398e-07, -8.2105e-06, -8.4490e-06, ..., -1.0632e-05, + 1.0608e-06, 1.6093e-06]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0240, -0.0192, -0.0313, 0.0003, 0.0040, 0.0341, 0.0151, -0.0194, + 0.0055, -0.0036], device='cuda:0'), grad: tensor([ 8.8736e-06, -7.4863e-05, 3.3230e-06, 3.6545e-06, -2.7269e-05, + 2.6766e-06, 3.4794e-06, 4.9233e-05, 4.4107e-05, -1.3307e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 214.22, cls_loss 0.0027 cls_loss_mapping 0.0065 cls_loss_causal 0.5350 re_mapping 0.0078 re_causal 0.0224 /// teacc 98.96 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0063, -0.0306, -0.0375, ..., -0.1418, -0.0335, -0.0799], + [-0.0131, -0.0460, -0.0418, ..., 0.0169, 0.0029, -0.0892], + [ 0.0422, -0.0860, -0.0691, ..., 0.0253, -0.1073, 0.0512], + ..., + [ 0.0059, -0.0209, -0.1073, ..., -0.0904, 0.0166, 0.1000], + [ 0.0544, -0.0864, -0.0776, ..., -0.0899, -0.0389, -0.0582], + [-0.1246, 0.0111, 0.0977, ..., 0.0715, -0.0439, -0.0972]], + device='cuda:0'), grad: tensor([[-2.5015e-06, 1.3039e-08, 3.6927e-07, ..., 1.2666e-07, + 0.0000e+00, 2.7940e-09], + [ 7.9069e-07, 2.5611e-08, -7.1898e-07, ..., -2.9616e-06, + 0.0000e+00, 3.3062e-08], + [ 2.0768e-07, 1.2573e-08, 3.2643e-07, ..., 5.8766e-07, + 0.0000e+00, -3.4925e-08], + ..., + [ 9.1735e-08, 1.3132e-07, 7.7533e-07, ..., 1.3262e-06, + 0.0000e+00, 2.7474e-08], + [ 1.4380e-06, 9.1735e-08, 1.6363e-06, ..., 1.1222e-06, + 0.0000e+00, 7.9162e-09], + [ 7.5949e-07, -9.5041e-07, -1.7090e-06, ..., -2.3060e-06, + 0.0000e+00, 1.2573e-08]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0246, -0.0187, -0.0304, 0.0001, 0.0039, 0.0338, 0.0152, -0.0200, + 0.0059, -0.0036], device='cuda:0'), grad: tensor([-7.3798e-06, -4.2468e-06, 1.6950e-06, 8.7172e-07, 3.4682e-06, + 3.2093e-06, -4.6827e-06, 2.6561e-06, 4.8801e-06, -4.8894e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 129---------------------------------------------------- +epoch 129, time 230.15, cls_loss 0.0027 cls_loss_mapping 0.0063 cls_loss_causal 0.5592 re_mapping 0.0072 re_causal 0.0208 /// teacc 99.02 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0056, -0.0309, -0.0376, ..., -0.1420, -0.0338, -0.0803], + [-0.0138, -0.0461, -0.0416, ..., 0.0172, 0.0033, -0.0898], + [ 0.0413, -0.0862, -0.0696, ..., 0.0250, -0.1117, 0.0500], + ..., + [ 0.0057, -0.0214, -0.1084, ..., -0.0922, 0.0168, 0.1014], + [ 0.0552, -0.0867, -0.0777, ..., -0.0899, -0.0396, -0.0586], + [-0.1247, 0.0119, 0.0980, ..., 0.0723, -0.0444, -0.0970]], + device='cuda:0'), grad: tensor([[ 1.0235e-06, 1.3318e-07, 1.2713e-06, ..., 2.4214e-07, + 1.5367e-08, 2.8033e-07], + [ 9.4995e-07, 3.4971e-07, 8.1733e-06, ..., 6.4149e-06, + -2.8638e-07, 1.5134e-06], + [ 5.9977e-07, 3.5856e-07, 1.0198e-06, ..., 9.4762e-07, + 7.7300e-08, 1.0170e-06], + ..., + [ 8.5682e-08, -8.4098e-07, 2.6356e-06, ..., 2.3060e-06, + 9.3132e-09, -3.8967e-06], + [ 4.4703e-06, 3.7532e-07, 5.1036e-06, ..., 1.3644e-06, + 3.7253e-08, 1.8133e-06], + [ 4.9360e-07, 3.0873e-07, 2.2054e-06, ..., 2.1495e-06, + 8.8941e-08, 1.6699e-06]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0250, -0.0188, -0.0323, -0.0001, 0.0034, 0.0334, 0.0161, -0.0197, + 0.0066, -0.0034], device='cuda:0'), grad: tensor([ 4.7423e-06, 1.8641e-05, 6.0908e-06, -6.8918e-06, -2.6464e-05, + 1.9744e-06, -2.7105e-05, 5.5693e-07, 1.9982e-05, 8.4266e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 213.94, cls_loss 0.0023 cls_loss_mapping 0.0044 cls_loss_causal 0.5660 re_mapping 0.0069 re_causal 0.0211 /// teacc 99.00 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0062, -0.0315, -0.0379, ..., -0.1421, -0.0343, -0.0808], + [-0.0140, -0.0462, -0.0416, ..., 0.0170, 0.0033, -0.0902], + [ 0.0411, -0.0864, -0.0704, ..., 0.0256, -0.1133, 0.0505], + ..., + [ 0.0053, -0.0205, -0.1087, ..., -0.0926, 0.0172, 0.1023], + [ 0.0553, -0.0870, -0.0777, ..., -0.0901, -0.0426, -0.0591], + [-0.1249, 0.0115, 0.0979, ..., 0.0719, -0.0446, -0.0979]], + device='cuda:0'), grad: tensor([[-5.3756e-06, 1.0952e-06, 1.0962e-06, ..., 4.0494e-06, + 1.2899e-07, 1.6047e-06], + [ 1.1232e-06, 3.2093e-06, 1.7136e-07, ..., 7.9125e-06, + 2.8824e-07, 6.3777e-06], + [ 2.1365e-06, 1.1669e-06, 5.1549e-07, ..., -9.4697e-06, + 4.2375e-08, -3.2961e-05], + ..., + [-9.8497e-06, -1.6108e-05, 3.8087e-05, ..., 2.8580e-05, + 2.6356e-06, -2.6613e-05], + [-2.3972e-06, 1.1967e-06, 1.7378e-06, ..., 2.6878e-06, + 8.9407e-08, 2.8871e-06], + [ 1.0632e-05, 2.0191e-06, -4.6849e-05, ..., -6.4135e-05, + -4.2841e-06, 4.1544e-05]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0246, -0.0189, -0.0322, -0.0007, 0.0037, 0.0329, 0.0167, -0.0191, + 0.0067, -0.0038], device='cuda:0'), grad: tensor([-2.3603e-05, 2.3261e-05, -2.7239e-05, 3.9905e-05, 2.4438e-06, + 1.7509e-05, 6.2250e-06, -5.6267e-05, 2.8778e-07, 1.7330e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 131---------------------------------------------------- +epoch 131, time 230.09, cls_loss 0.0022 cls_loss_mapping 0.0049 cls_loss_causal 0.5415 re_mapping 0.0070 re_causal 0.0211 /// teacc 99.12 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0064, -0.0321, -0.0381, ..., -0.1426, -0.0349, -0.0811], + [-0.0142, -0.0464, -0.0415, ..., 0.0171, 0.0032, -0.0907], + [ 0.0401, -0.0867, -0.0714, ..., 0.0257, -0.1143, 0.0505], + ..., + [ 0.0054, -0.0202, -0.1094, ..., -0.0930, 0.0173, 0.1031], + [ 0.0554, -0.0874, -0.0777, ..., -0.0904, -0.0434, -0.0598], + [-0.1251, 0.0114, 0.0979, ..., 0.0716, -0.0451, -0.0988]], + device='cuda:0'), grad: tensor([[-1.9833e-05, -6.0536e-08, 6.6590e-08, ..., 1.0757e-07, + 9.3132e-10, -3.1777e-06], + [ 5.8813e-07, 4.6752e-07, -2.9150e-07, ..., -1.9697e-07, + 1.3039e-08, 7.1479e-07], + [ 3.1590e-06, 6.5705e-07, 1.1967e-07, ..., 5.6392e-07, + 1.4901e-08, -4.1090e-06], + ..., + [ 1.6699e-06, 3.4785e-07, 2.2911e-07, ..., 4.7451e-07, + -6.1933e-08, 3.6210e-06], + [ 6.0396e-07, -7.4506e-09, -1.9418e-07, ..., 5.4808e-07, + 6.9849e-09, 1.7872e-06], + [ 5.2862e-06, 1.4920e-06, -5.5879e-07, ..., 1.1949e-06, + -6.5193e-09, 2.6841e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0246, -0.0190, -0.0328, -0.0005, 0.0039, 0.0330, 0.0169, -0.0187, + 0.0066, -0.0041], device='cuda:0'), grad: tensor([-6.9499e-05, 4.3139e-06, 8.8587e-06, -3.7462e-05, 3.5223e-06, + 4.7207e-05, 1.4402e-05, 1.0036e-05, -4.1239e-06, 2.2769e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 213.99, cls_loss 0.0023 cls_loss_mapping 0.0051 cls_loss_causal 0.5631 re_mapping 0.0064 re_causal 0.0208 /// teacc 98.99 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0066, -0.0326, -0.0382, ..., -0.1431, -0.0352, -0.0818], + [-0.0143, -0.0464, -0.0412, ..., 0.0173, 0.0042, -0.0913], + [ 0.0402, -0.0871, -0.0722, ..., 0.0254, -0.1151, 0.0509], + ..., + [ 0.0051, -0.0201, -0.1104, ..., -0.0938, 0.0163, 0.1036], + [ 0.0554, -0.0878, -0.0777, ..., -0.0904, -0.0437, -0.0606], + [-0.1252, 0.0114, 0.0979, ..., 0.0717, -0.0451, -0.0998]], + device='cuda:0'), grad: tensor([[-2.9337e-08, 3.4412e-07, 1.0328e-06, ..., 1.0179e-06, + 1.3970e-09, 1.4706e-06], + [ 1.9092e-07, 2.7334e-07, -9.7323e-07, ..., -7.9069e-07, + 2.3283e-09, 7.0501e-07], + [-4.9919e-07, 1.2107e-07, 1.6671e-07, ..., -1.2806e-07, + 3.2596e-09, -8.7321e-06], + ..., + [ 1.7695e-07, 4.3027e-07, 1.4780e-06, ..., 1.4268e-06, + 1.8626e-09, 5.0478e-07], + [ 6.0070e-08, 6.1886e-07, 1.1465e-06, ..., 1.7462e-06, + 3.7253e-09, 2.7474e-06], + [ 6.4587e-07, -2.9206e-06, -7.9125e-06, ..., -5.9344e-06, + 3.2596e-09, 2.5472e-07]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0245, -0.0191, -0.0327, 0.0006, 0.0041, 0.0321, 0.0167, -0.0187, + 0.0067, -0.0043], device='cuda:0'), grad: tensor([ 5.3085e-06, -1.7062e-06, -1.9297e-05, 2.0146e-05, 3.5278e-06, + -1.4961e-05, 3.7253e-06, 3.8054e-06, 8.0913e-06, -8.6948e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 214.19, cls_loss 0.0026 cls_loss_mapping 0.0051 cls_loss_causal 0.5574 re_mapping 0.0063 re_causal 0.0205 /// teacc 98.86 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0068, -0.0331, -0.0382, ..., -0.1437, -0.0355, -0.0826], + [-0.0147, -0.0466, -0.0414, ..., 0.0172, 0.0042, -0.0916], + [ 0.0401, -0.0867, -0.0730, ..., 0.0253, -0.1155, 0.0529], + ..., + [ 0.0050, -0.0201, -0.1109, ..., -0.0941, 0.0163, 0.1030], + [ 0.0561, -0.0881, -0.0773, ..., -0.0902, -0.0444, -0.0614], + [-0.1254, 0.0116, 0.0980, ..., 0.0717, -0.0452, -0.1003]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 2.6077e-08, 4.0978e-08, ..., 9.9652e-08, + 9.3132e-10, 2.1514e-07], + [ 4.4983e-07, 1.9744e-07, -5.2992e-07, ..., -3.7253e-08, + 0.0000e+00, 1.9446e-06], + [-6.5099e-07, 8.1956e-08, 1.1455e-07, ..., -1.1306e-06, + 0.0000e+00, -5.5060e-06], + ..., + [ 4.9360e-08, 2.1514e-07, 6.9384e-07, ..., 6.9663e-07, + 0.0000e+00, 1.9558e-07], + [-5.0571e-07, 1.3877e-07, 3.2317e-07, ..., 4.5076e-07, + 1.8626e-09, 3.4738e-07], + [ 6.7987e-08, -4.2841e-08, -6.7055e-08, ..., 4.3400e-07, + 9.3132e-10, 2.0023e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([-2.4294e-02, -1.9520e-02, -3.1836e-02, -9.0919e-05, 3.8505e-03, + 3.2389e-02, 1.6331e-02, -1.9080e-02, 7.6257e-03, -4.4534e-03], + device='cuda:0'), grad: tensor([ 4.8801e-07, 3.1069e-06, -1.0476e-05, 1.9092e-07, -7.6368e-08, + -1.6764e-07, 4.2953e-06, 1.7360e-06, -2.0862e-07, 1.1045e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 214.13, cls_loss 0.0024 cls_loss_mapping 0.0054 cls_loss_causal 0.5367 re_mapping 0.0065 re_causal 0.0207 /// teacc 99.02 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0069, -0.0338, -0.0383, ..., -0.1446, -0.0358, -0.0840], + [-0.0148, -0.0467, -0.0413, ..., 0.0173, 0.0044, -0.0918], + [ 0.0400, -0.0874, -0.0737, ..., 0.0252, -0.1163, 0.0534], + ..., + [ 0.0049, -0.0198, -0.1117, ..., -0.0948, 0.0161, 0.1033], + [ 0.0563, -0.0885, -0.0773, ..., -0.0905, -0.0446, -0.0623], + [-0.1255, 0.0127, 0.0995, ..., 0.0732, -0.0452, -0.1016]], + device='cuda:0'), grad: tensor([[ 3.0268e-07, 1.0338e-07, 2.1607e-07, ..., 7.7300e-08, + 0.0000e+00, 2.1793e-07], + [ 3.1665e-08, 2.4308e-07, -1.2377e-06, ..., -1.1502e-06, + 0.0000e+00, 4.3772e-07], + [ 8.8476e-08, 5.6811e-07, 2.4308e-07, ..., 1.9372e-07, + 0.0000e+00, 9.8161e-07], + ..., + [ 2.2352e-08, -4.3102e-06, 3.9116e-07, ..., 3.7439e-07, + 0.0000e+00, -8.1137e-06], + [-1.9260e-06, 8.6334e-07, -4.6473e-07, ..., 5.0664e-07, + 0.0000e+00, 1.8356e-06], + [ 1.2768e-06, 4.9733e-07, 4.9546e-07, ..., 3.1665e-08, + 0.0000e+00, 8.8196e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0245, -0.0194, -0.0318, 0.0002, 0.0026, 0.0327, 0.0160, -0.0191, + 0.0077, -0.0039], device='cuda:0'), grad: tensor([ 1.4380e-06, -1.8599e-06, 2.3115e-06, 6.6683e-06, -1.5553e-07, + -1.7043e-07, 5.5041e-07, -1.1809e-05, -2.6450e-06, 5.6364e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 213.97, cls_loss 0.0023 cls_loss_mapping 0.0045 cls_loss_causal 0.5100 re_mapping 0.0069 re_causal 0.0191 /// teacc 98.97 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0073, -0.0342, -0.0383, ..., -0.1453, -0.0360, -0.0847], + [-0.0150, -0.0469, -0.0411, ..., 0.0174, 0.0044, -0.0921], + [ 0.0401, -0.0879, -0.0742, ..., 0.0251, -0.1169, 0.0536], + ..., + [ 0.0050, -0.0193, -0.1126, ..., -0.0958, 0.0164, 0.1042], + [ 0.0562, -0.0890, -0.0773, ..., -0.0906, -0.0451, -0.0635], + [-0.1257, 0.0122, 0.0991, ..., 0.0728, -0.0459, -0.1031]], + device='cuda:0'), grad: tensor([[-3.1944e-07, 2.0768e-07, 1.5087e-07, ..., 3.3528e-08, + 0.0000e+00, 6.0350e-07], + [ 1.1548e-07, 8.3074e-07, -1.6578e-07, ..., -1.6671e-07, + 0.0000e+00, 2.0899e-06], + [ 1.6671e-07, 2.1551e-06, 2.1979e-07, ..., 6.7987e-08, + 0.0000e+00, -5.2005e-06], + ..., + [-9.1642e-07, -7.0557e-06, 3.1572e-07, ..., 3.9581e-07, + 0.0000e+00, -4.0308e-06], + [ 1.6298e-07, 1.1707e-06, 5.3551e-07, ..., 3.9954e-07, + 0.0000e+00, 3.0380e-06], + [ 7.8231e-07, 2.5034e-06, -2.0433e-06, ..., -2.1681e-06, + 0.0000e+00, 3.1758e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0243, -0.0194, -0.0320, 0.0007, 0.0035, 0.0323, 0.0156, -0.0186, + 0.0076, -0.0045], device='cuda:0'), grad: tensor([-5.9418e-07, 3.4440e-06, -8.5011e-06, -1.9614e-06, 2.6822e-06, + 1.5935e-06, -4.1816e-07, -8.8811e-06, 6.9775e-06, 5.6475e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 214.03, cls_loss 0.0023 cls_loss_mapping 0.0050 cls_loss_causal 0.5661 re_mapping 0.0067 re_causal 0.0207 /// teacc 98.94 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0089, -0.0337, -0.0381, ..., -0.1451, -0.0363, -0.0846], + [-0.0154, -0.0467, -0.0404, ..., 0.0177, 0.0047, -0.0925], + [ 0.0401, -0.0880, -0.0749, ..., 0.0252, -0.1172, 0.0542], + ..., + [ 0.0049, -0.0186, -0.1137, ..., -0.0964, 0.0164, 0.1052], + [ 0.0562, -0.0894, -0.0774, ..., -0.0909, -0.0447, -0.0644], + [-0.1264, 0.0117, 0.0990, ..., 0.0725, -0.0469, -0.1048]], + device='cuda:0'), grad: tensor([[-3.5912e-05, 3.4831e-07, -5.4240e-06, ..., -2.4457e-06, + 9.3132e-10, 2.7269e-06], + [ 2.2143e-05, 4.0419e-07, 4.2729e-06, ..., 2.2948e-06, + 0.0000e+00, 6.8359e-07], + [ 2.9113e-06, 5.6904e-07, 7.8883e-07, ..., 4.9546e-07, + 0.0000e+00, -2.7582e-05], + ..., + [ 3.8277e-07, 2.3190e-07, 1.1474e-06, ..., 1.3858e-06, + 9.3132e-10, 1.4946e-05], + [-5.0440e-06, -7.8082e-06, -8.2329e-06, ..., -5.3197e-06, + 2.7940e-09, 8.6948e-06], + [ 2.0787e-06, -2.3432e-06, -1.1571e-05, ..., -1.2442e-05, + 3.7253e-09, 4.5262e-07]], device='cuda:0') +Epoch 138, bias, value: tensor([-2.2879e-02, -1.9259e-02, -3.1616e-02, 2.9890e-05, 3.8349e-03, + 3.2057e-02, 1.5197e-02, -1.8279e-02, 7.4792e-03, -5.1000e-03], + device='cuda:0'), grad: tensor([-1.0347e-04, 7.1824e-05, -3.0473e-05, 7.2241e-05, 7.9200e-06, + 1.3821e-05, 1.8150e-05, 2.6017e-05, -7.2539e-05, -3.3714e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 214.18, cls_loss 0.0023 cls_loss_mapping 0.0049 cls_loss_causal 0.5353 re_mapping 0.0072 re_causal 0.0203 /// teacc 99.00 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 0.0087, -0.0343, -0.0389, ..., -0.1456, -0.0364, -0.0849], + [-0.0159, -0.0473, -0.0406, ..., 0.0176, 0.0047, -0.0932], + [ 0.0397, -0.0881, -0.0765, ..., 0.0254, -0.1175, 0.0544], + ..., + [ 0.0052, -0.0175, -0.1140, ..., -0.0962, 0.0166, 0.1064], + [ 0.0564, -0.0897, -0.0775, ..., -0.0911, -0.0450, -0.0646], + [-0.1264, 0.0120, 0.0992, ..., 0.0727, -0.0473, -0.1053]], + device='cuda:0'), grad: tensor([[ 4.7963e-07, 2.1011e-06, 3.1125e-06, ..., 2.2464e-06, + 0.0000e+00, 7.0035e-07], + [-2.5518e-07, 4.0457e-06, 2.4661e-06, ..., 1.5572e-06, + 0.0000e+00, 2.3730e-06], + [ 1.0524e-07, 6.1654e-06, 1.7034e-06, ..., 3.4105e-06, + 0.0000e+00, 6.8396e-06], + ..., + [ 7.2643e-08, 1.3714e-03, 2.2602e-03, ..., 1.6298e-03, + 0.0000e+00, -7.0482e-06], + [ 9.0338e-08, 9.0823e-06, 6.4187e-06, ..., 5.5246e-06, + 0.0000e+00, 6.1654e-06], + [ 1.8720e-07, -1.4582e-03, -2.3956e-03, ..., -1.7223e-03, + 0.0000e+00, 8.3894e-06]], device='cuda:0') +Epoch 139, bias, value: tensor([-2.3504e-02, -1.9865e-02, -3.2235e-02, -9.1996e-05, 3.5770e-03, + 3.2123e-02, 1.6044e-02, -1.7537e-02, 7.6777e-03, -4.9101e-03], + device='cuda:0'), grad: tensor([ 8.9258e-06, 1.1906e-05, 1.9804e-05, -4.5002e-05, 2.7776e-04, + 1.2144e-05, -1.6652e-06, 5.2032e-03, 2.9892e-05, -5.5161e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 214.04, cls_loss 0.0024 cls_loss_mapping 0.0039 cls_loss_causal 0.5259 re_mapping 0.0072 re_causal 0.0205 /// teacc 98.97 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 0.0092, -0.0354, -0.0389, ..., -0.1460, -0.0366, -0.0859], + [-0.0172, -0.0474, -0.0408, ..., 0.0179, 0.0047, -0.0937], + [ 0.0394, -0.0887, -0.0772, ..., 0.0253, -0.1177, 0.0543], + ..., + [ 0.0023, -0.0192, -0.1157, ..., -0.0983, 0.0165, 0.1060], + [ 0.0579, -0.0879, -0.0776, ..., -0.0913, -0.0457, -0.0627], + [-0.1271, 0.0128, 0.0994, ..., 0.0730, -0.0473, -0.1077]], + device='cuda:0'), grad: tensor([[-2.8722e-06, 5.4482e-07, 1.0328e-06, ..., 1.1781e-06, + 2.9057e-07, 1.1921e-07], + [ 3.2689e-07, -2.2128e-06, -3.4403e-06, ..., -1.4119e-06, + 1.5832e-08, 7.8976e-07], + [ 3.3993e-07, 8.0466e-07, 5.5879e-07, ..., 4.5821e-07, + 6.5193e-09, 8.5868e-07], + ..., + [ 1.3784e-07, -3.3621e-07, 1.6857e-06, ..., 1.1735e-06, + 3.9116e-08, -3.4627e-06], + [ 1.0990e-07, 9.5740e-07, 1.3001e-06, ..., 1.2862e-06, + 1.2480e-07, 6.8825e-07], + [ 8.7544e-07, -1.8135e-05, -3.4213e-05, ..., -4.2200e-05, + 4.8429e-08, 4.6100e-07]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0234, -0.0208, -0.0327, 0.0003, 0.0034, 0.0321, 0.0174, -0.0191, + 0.0092, -0.0050], device='cuda:0'), grad: tensor([-7.4804e-06, -1.0699e-05, 3.8221e-06, 9.4026e-06, 8.5011e-06, + 3.3051e-05, 6.9663e-06, 1.2442e-06, 3.6173e-06, -4.8459e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 214.08, cls_loss 0.0027 cls_loss_mapping 0.0058 cls_loss_causal 0.5564 re_mapping 0.0067 re_causal 0.0194 /// teacc 98.94 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 0.0095, -0.0363, -0.0395, ..., -0.1467, -0.0372, -0.0865], + [-0.0177, -0.0473, -0.0406, ..., 0.0181, 0.0047, -0.0942], + [ 0.0392, -0.0890, -0.0763, ..., 0.0252, -0.1184, 0.0531], + ..., + [ 0.0020, -0.0194, -0.1161, ..., -0.0992, 0.0164, 0.1079], + [ 0.0584, -0.0882, -0.0776, ..., -0.0917, -0.0463, -0.0630], + [-0.1274, 0.0131, 0.0996, ..., 0.0734, -0.0471, -0.1081]], + device='cuda:0'), grad: tensor([[-7.7784e-06, 8.4750e-08, 1.7975e-07, ..., 2.7567e-07, + 2.6077e-08, 1.6298e-07], + [ 1.4557e-06, 3.7160e-07, 2.9802e-08, ..., -5.8673e-08, + 3.9116e-08, 1.0151e-06], + [ 1.0626e-06, 1.3104e-06, 1.8720e-07, ..., 2.7474e-07, + 1.0338e-07, 4.0494e-06], + ..., + [ 3.7160e-07, -1.1781e-06, 5.2340e-06, ..., 4.5374e-06, + 3.1665e-08, -5.5954e-06], + [-6.3144e-07, 7.8883e-07, 1.8300e-06, ..., 2.2743e-06, + 6.5193e-08, 1.6661e-06], + [ 7.0967e-07, -9.8906e-07, -7.5586e-06, ..., -6.2846e-06, + 3.1665e-08, -1.4482e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0235, -0.0209, -0.0337, -0.0014, 0.0032, 0.0338, 0.0168, -0.0183, + 0.0096, -0.0050], device='cuda:0'), grad: tensor([-2.0102e-05, 4.9546e-06, 9.9018e-06, 1.5497e-06, -1.4650e-06, + 6.4857e-06, 5.5879e-06, 4.4703e-06, 2.1569e-06, -1.3605e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 214.08, cls_loss 0.0025 cls_loss_mapping 0.0056 cls_loss_causal 0.5952 re_mapping 0.0066 re_causal 0.0210 /// teacc 98.94 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 0.0094, -0.0387, -0.0398, ..., -0.1489, -0.0376, -0.0893], + [-0.0176, -0.0473, -0.0401, ..., 0.0183, 0.0047, -0.0942], + [ 0.0395, -0.0892, -0.0769, ..., 0.0258, -0.1190, 0.0531], + ..., + [ 0.0019, -0.0191, -0.1163, ..., -0.0991, 0.0184, 0.1084], + [ 0.0584, -0.0886, -0.0778, ..., -0.0923, -0.0466, -0.0636], + [-0.1276, 0.0127, 0.0998, ..., 0.0735, -0.0500, -0.1099]], + device='cuda:0'), grad: tensor([[ 6.5267e-06, 2.4401e-07, 1.2264e-05, ..., 1.0960e-05, + 2.9802e-08, 1.8626e-09], + [ 2.9579e-06, 3.7774e-06, 6.3963e-06, ..., 4.7572e-06, + 1.0245e-08, 4.4517e-06], + [ 4.1947e-06, 1.1986e-06, 7.2271e-06, ..., 5.9158e-06, + 4.6566e-09, 1.2601e-06], + ..., + [ 2.2240e-06, -3.0790e-06, 5.6066e-06, ..., 4.7460e-06, + 4.6566e-09, -5.1335e-06], + [ 2.1383e-05, -7.0706e-06, 2.4527e-05, ..., 3.8683e-05, + 3.8184e-08, 7.7114e-07], + [-5.3972e-05, 3.0641e-06, -1.0198e-04, ..., -1.0931e-04, + 1.4901e-08, 4.3958e-07]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0243, -0.0204, -0.0339, 0.0011, 0.0031, 0.0319, 0.0170, -0.0181, + 0.0094, -0.0052], device='cuda:0'), grad: tensor([ 2.5675e-05, 3.6061e-05, 2.3559e-05, 1.8835e-05, 1.1343e-04, + 6.1691e-05, -9.9301e-05, -1.4588e-05, 3.2216e-05, -1.9753e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 214.14, cls_loss 0.0024 cls_loss_mapping 0.0056 cls_loss_causal 0.5682 re_mapping 0.0068 re_causal 0.0208 /// teacc 99.02 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 0.0098, -0.0387, -0.0397, ..., -0.1494, -0.0379, -0.0892], + [-0.0177, -0.0475, -0.0400, ..., 0.0184, 0.0047, -0.0939], + [ 0.0393, -0.0889, -0.0776, ..., 0.0255, -0.1192, 0.0536], + ..., + [ 0.0019, -0.0191, -0.1166, ..., -0.0992, 0.0185, 0.1083], + [ 0.0585, -0.0886, -0.0778, ..., -0.0926, -0.0468, -0.0639], + [-0.1280, 0.0123, 0.0995, ..., 0.0731, -0.0502, -0.1107]], + device='cuda:0'), grad: tensor([[-1.2722e-06, 4.7591e-07, 3.1572e-07, ..., 3.4459e-07, + 0.0000e+00, 4.6007e-07], + [ 3.0082e-07, 6.6403e-07, -8.9873e-07, ..., -6.3423e-07, + 7.4506e-09, 1.4519e-06], + [-6.3360e-05, -1.2076e-04, 2.6543e-07, ..., 2.4028e-07, + 9.3132e-10, -2.5201e-04], + ..., + [ 4.5598e-05, 8.4996e-05, 1.8226e-06, ..., 2.0266e-06, + -2.2352e-08, 1.7703e-04], + [ 1.5646e-05, 2.9370e-05, 6.9570e-07, ..., 1.1520e-06, + 5.5879e-09, 6.1214e-05], + [ 6.2957e-07, -7.9814e-07, -5.0887e-06, ..., -3.6266e-06, + 7.4506e-09, 5.3737e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0240, -0.0199, -0.0337, 0.0001, 0.0039, 0.0331, 0.0156, -0.0183, + 0.0095, -0.0057], device='cuda:0'), grad: tensor([-2.0675e-06, -1.1250e-06, -3.5501e-04, 1.5855e-05, 3.5763e-06, + -2.3954e-06, 1.1297e-06, 2.5535e-04, 8.9586e-05, -5.1074e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 214.06, cls_loss 0.0021 cls_loss_mapping 0.0049 cls_loss_causal 0.5501 re_mapping 0.0068 re_causal 0.0204 /// teacc 98.93 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0104, -0.0388, -0.0398, ..., -0.1495, -0.0379, -0.0887], + [-0.0178, -0.0472, -0.0400, ..., 0.0184, 0.0047, -0.0938], + [ 0.0405, -0.0879, -0.0780, ..., 0.0255, -0.1192, 0.0546], + ..., + [ 0.0018, -0.0191, -0.1170, ..., -0.0990, 0.0185, 0.1085], + [ 0.0582, -0.0896, -0.0778, ..., -0.0928, -0.0468, -0.0661], + [-0.1282, 0.0106, 0.0981, ..., 0.0713, -0.0502, -0.1114]], + device='cuda:0'), grad: tensor([[-3.5390e-08, 5.2154e-08, 1.6484e-07, ..., 1.9744e-07, + 0.0000e+00, 1.8999e-07], + [ 1.2573e-07, 1.7509e-07, 8.6613e-08, ..., -1.2107e-07, + 0.0000e+00, 7.9349e-07], + [-9.1270e-08, 1.0151e-07, 1.1921e-07, ..., 1.2014e-07, + 0.0000e+00, -2.8759e-06], + ..., + [ 4.4703e-08, -3.1199e-07, 1.7360e-06, ..., 1.5283e-06, + 0.0000e+00, 1.4342e-07], + [-2.2352e-08, 7.0594e-07, 5.9512e-07, ..., 1.3644e-06, + 0.0000e+00, 8.3912e-07], + [ 1.2200e-07, -1.4603e-06, -5.3570e-06, ..., -4.9844e-06, + 0.0000e+00, 3.2596e-07]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0237, -0.0195, -0.0328, 0.0009, 0.0059, 0.0321, 0.0153, -0.0182, + 0.0088, -0.0074], device='cuda:0'), grad: tensor([ 6.2119e-07, 2.6077e-06, -4.7199e-06, 1.9614e-06, 3.3285e-06, + -6.0815e-07, 8.7824e-07, 3.2373e-06, 2.2352e-08, -7.3537e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 214.14, cls_loss 0.0026 cls_loss_mapping 0.0042 cls_loss_causal 0.5118 re_mapping 0.0066 re_causal 0.0188 /// teacc 98.94 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0107, -0.0392, -0.0398, ..., -0.1502, -0.0384, -0.0894], + [-0.0179, -0.0482, -0.0411, ..., 0.0174, 0.0048, -0.0945], + [ 0.0406, -0.0882, -0.0788, ..., 0.0252, -0.1194, 0.0553], + ..., + [ 0.0018, -0.0190, -0.1180, ..., -0.1000, 0.0185, 0.1088], + [ 0.0580, -0.0903, -0.0781, ..., -0.0934, -0.0473, -0.0671], + [-0.1284, 0.0113, 0.0992, ..., 0.0726, -0.0502, -0.1119]], + device='cuda:0'), grad: tensor([[-7.2271e-07, 6.8918e-08, 1.8254e-07, ..., 1.2293e-07, + 7.4506e-09, 1.0245e-07], + [ 3.3528e-08, -3.3434e-07, -3.0566e-06, ..., -5.1111e-06, + -1.0347e-06, 1.1222e-06], + [ 2.7381e-07, 6.1281e-07, 4.1258e-07, ..., 5.2806e-07, + 2.6077e-08, 1.0915e-06], + ..., + [ 2.9802e-08, -4.4256e-06, 1.2675e-06, ..., 1.8012e-06, + 3.2689e-07, -4.1351e-06], + [ 2.7940e-08, 3.9395e-07, 5.9232e-07, ..., 7.1339e-07, + 8.6613e-08, 4.3958e-07], + [ 1.4901e-07, 1.3355e-06, -2.1700e-07, ..., 1.1157e-06, + 4.6939e-07, 7.9535e-07]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0237, -0.0203, -0.0323, 0.0010, 0.0052, 0.0320, 0.0157, -0.0183, + 0.0083, -0.0064], device='cuda:0'), grad: tensor([-2.1663e-06, -1.0893e-05, 3.9712e-06, 1.5602e-05, 8.1304e-07, + -9.9540e-06, 5.5879e-09, -5.8115e-06, 2.5425e-06, 5.8934e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 214.02, cls_loss 0.0038 cls_loss_mapping 0.0057 cls_loss_causal 0.5343 re_mapping 0.0072 re_causal 0.0188 /// teacc 98.97 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0125, -0.0395, -0.0368, ..., -0.1481, -0.0384, -0.0907], + [-0.0183, -0.0458, -0.0402, ..., 0.0182, 0.0048, -0.0933], + [ 0.0405, -0.0889, -0.0818, ..., 0.0239, -0.1191, 0.0539], + ..., + [ 0.0019, -0.0199, -0.1186, ..., -0.1011, 0.0185, 0.1110], + [ 0.0582, -0.0907, -0.0782, ..., -0.0940, -0.0474, -0.0678], + [-0.1287, 0.0110, 0.0979, ..., 0.0723, -0.0502, -0.1139]], + device='cuda:0'), grad: tensor([[ 8.0317e-06, 3.7085e-06, 1.4957e-06, ..., 1.6987e-06, + 0.0000e+00, 1.1921e-07], + [ 3.3025e-06, 1.4976e-06, 3.7905e-07, ..., 4.2655e-07, + 0.0000e+00, 1.3839e-06], + [ 2.3507e-06, 5.4762e-07, 3.4552e-07, ..., -4.5486e-06, + 0.0000e+00, -1.1407e-05], + ..., + [ 1.8347e-07, -5.5414e-07, 1.6298e-07, ..., 1.2452e-06, + 0.0000e+00, 1.6978e-06], + [ 3.6240e-05, 1.4335e-05, 4.9472e-06, ..., 3.9339e-06, + 0.0000e+00, 1.4761e-06], + [ 1.3644e-06, 6.5099e-07, 1.6987e-06, ..., 1.9781e-06, + 0.0000e+00, 1.0962e-06]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0207, -0.0191, -0.0337, 0.0005, 0.0053, 0.0325, 0.0138, -0.0172, + 0.0080, -0.0082], device='cuda:0'), grad: tensor([ 2.5302e-05, 1.2696e-05, -9.7156e-06, 9.7007e-06, 8.3074e-06, + 4.7624e-05, -2.0516e-04, 3.1032e-06, 9.0957e-05, 1.7196e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 214.10, cls_loss 0.0019 cls_loss_mapping 0.0038 cls_loss_causal 0.5206 re_mapping 0.0069 re_causal 0.0205 /// teacc 98.92 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 0.0113, -0.0421, -0.0373, ..., -0.1487, -0.0384, -0.0912], + [-0.0185, -0.0457, -0.0400, ..., 0.0185, 0.0048, -0.0936], + [ 0.0401, -0.0893, -0.0825, ..., 0.0238, -0.1191, 0.0541], + ..., + [ 0.0016, -0.0202, -0.1193, ..., -0.1018, 0.0185, 0.1110], + [ 0.0584, -0.0916, -0.0785, ..., -0.0945, -0.0475, -0.0683], + [-0.1288, 0.0112, 0.0980, ..., 0.0726, -0.0502, -0.1146]], + device='cuda:0'), grad: tensor([[ 2.0396e-07, 1.6950e-07, 9.4436e-07, ..., 1.1008e-06, + 0.0000e+00, 2.8778e-07], + [ 6.6124e-08, 1.2396e-06, 9.2834e-06, ..., 9.7230e-06, + 0.0000e+00, 7.7952e-07], + [ 4.6752e-07, 3.2689e-07, 1.8533e-06, ..., 1.5087e-06, + 0.0000e+00, -2.9188e-06], + ..., + [ 1.0431e-07, -6.3330e-08, 7.6741e-06, ..., 8.0243e-06, + 0.0000e+00, -5.9791e-07], + [-3.3155e-07, 5.4762e-07, 1.5991e-06, ..., 2.0415e-06, + 0.0000e+00, 8.9407e-07], + [ 4.4703e-07, 2.8443e-06, 3.1948e-05, ..., 3.2276e-05, + 0.0000e+00, 4.0233e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0214, -0.0187, -0.0339, 0.0005, 0.0056, 0.0327, 0.0154, -0.0177, + 0.0078, -0.0083], device='cuda:0'), grad: tensor([ 2.9132e-06, 1.9610e-05, 3.5204e-07, 2.7195e-06, -1.3447e-04, + 8.4341e-06, 1.9401e-05, 1.3664e-05, 3.3043e-06, 6.4194e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 214.14, cls_loss 0.0021 cls_loss_mapping 0.0049 cls_loss_causal 0.5289 re_mapping 0.0064 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0108, -0.0426, -0.0374, ..., -0.1490, -0.0385, -0.0919], + [-0.0185, -0.0462, -0.0409, ..., 0.0174, 0.0048, -0.0942], + [ 0.0398, -0.0899, -0.0839, ..., 0.0234, -0.1191, 0.0543], + ..., + [ 0.0016, -0.0197, -0.1197, ..., -0.1024, 0.0185, 0.1112], + [ 0.0585, -0.0921, -0.0787, ..., -0.0952, -0.0476, -0.0686], + [-0.1290, 0.0113, 0.0984, ..., 0.0731, -0.0502, -0.1154]], + device='cuda:0'), grad: tensor([[ 1.1846e-05, 7.4022e-06, -5.5879e-08, ..., 2.0176e-05, + 2.7940e-09, 1.1541e-05], + [ 3.9935e-05, 5.1707e-06, -3.7253e-09, ..., 1.5190e-06, + 9.3132e-10, 8.1718e-05], + [ 7.2978e-06, 3.2652e-06, 2.4121e-07, ..., 1.2904e-05, + 0.0000e+00, -1.8454e-04], + ..., + [ 4.5821e-06, 2.8778e-06, 9.8348e-07, ..., 7.1563e-06, + 2.7940e-09, 3.6240e-05], + [-8.3819e-06, 1.8284e-05, 2.8592e-07, ..., 6.6280e-05, + 5.5879e-09, 2.8387e-05], + [ 1.9930e-06, 1.2675e-06, -1.9222e-06, ..., 3.4384e-06, + 9.3132e-09, 6.7055e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0218, -0.0193, -0.0341, 0.0008, 0.0062, 0.0326, 0.0155, -0.0177, + 0.0075, -0.0081], device='cuda:0'), grad: tensor([ 6.2048e-05, 3.0279e-04, -3.7241e-04, 1.2648e-04, 5.8040e-06, + -3.5048e-04, 4.4584e-05, 1.0467e-04, 6.7532e-05, 8.8364e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 213.82, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5555 re_mapping 0.0064 re_causal 0.0194 /// teacc 98.98 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0111, -0.0434, -0.0375, ..., -0.1494, -0.0386, -0.0931], + [-0.0189, -0.0465, -0.0408, ..., 0.0177, 0.0048, -0.0946], + [ 0.0395, -0.0901, -0.0846, ..., 0.0233, -0.1192, 0.0547], + ..., + [ 0.0016, -0.0194, -0.1203, ..., -0.1034, 0.0185, 0.1115], + [ 0.0584, -0.0926, -0.0790, ..., -0.0958, -0.0476, -0.0691], + [-0.1292, 0.0113, 0.0986, ..., 0.0733, -0.0502, -0.1162]], + device='cuda:0'), grad: tensor([[-2.1681e-05, 3.5390e-08, 4.2375e-07, ..., 3.8091e-07, + 0.0000e+00, 3.4738e-07], + [ 5.8115e-07, 1.9372e-07, -2.0154e-06, ..., -2.6803e-06, + 0.0000e+00, 4.8336e-07], + [ 1.6065e-06, 1.6671e-07, 4.5262e-07, ..., -1.1688e-06, + 0.0000e+00, -5.0887e-06], + ..., + [ 3.1851e-07, 2.6077e-07, 3.2634e-06, ..., 2.9691e-06, + 0.0000e+00, 2.3656e-07], + [ 1.3681e-06, 8.1584e-07, 1.2787e-06, ..., 1.7155e-06, + 0.0000e+00, 1.3737e-06], + [ 6.8080e-07, -2.5127e-06, -3.0082e-06, ..., -5.0366e-06, + 0.0000e+00, 1.8813e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0218, -0.0194, -0.0340, 0.0009, 0.0067, 0.0324, 0.0158, -0.0179, + 0.0071, -0.0081], device='cuda:0'), grad: tensor([-5.2571e-05, -6.2250e-06, -3.5726e-06, 1.2085e-05, -1.2428e-05, + 1.1876e-05, 2.7776e-05, 1.5125e-05, 1.0014e-05, -2.2426e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 213.65, cls_loss 0.0021 cls_loss_mapping 0.0036 cls_loss_causal 0.5311 re_mapping 0.0064 re_causal 0.0190 /// teacc 99.03 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0109, -0.0439, -0.0387, ..., -0.1503, -0.0386, -0.0939], + [-0.0177, -0.0466, -0.0409, ..., 0.0181, 0.0048, -0.0951], + [ 0.0393, -0.0904, -0.0851, ..., 0.0230, -0.1192, 0.0549], + ..., + [ 0.0015, -0.0192, -0.1206, ..., -0.1037, 0.0185, 0.1119], + [ 0.0579, -0.0929, -0.0790, ..., -0.0970, -0.0477, -0.0694], + [-0.1292, 0.0121, 0.0995, ..., 0.0739, -0.0503, -0.1167]], + device='cuda:0'), grad: tensor([[ 4.7125e-07, 1.2014e-07, 4.9546e-07, ..., 3.5111e-07, + 0.0000e+00, 1.2200e-07], + [ 1.0684e-05, 2.2762e-06, -7.9814e-07, ..., -4.3735e-06, + 0.0000e+00, 3.0361e-07], + [ 7.8753e-06, 1.6373e-06, 5.9418e-07, ..., 4.0866e-06, + 0.0000e+00, 3.4459e-08], + ..., + [ 3.1423e-06, 7.3481e-07, 1.8720e-07, ..., 7.5530e-07, + 0.0000e+00, 9.5926e-08], + [-2.8402e-05, -5.7258e-06, 4.4424e-07, ..., -3.1199e-06, + 0.0000e+00, 5.3085e-07], + [ 9.1456e-07, 1.3411e-07, 3.3248e-07, ..., 5.0478e-07, + 0.0000e+00, 2.9057e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0235, -0.0191, -0.0340, 0.0005, 0.0052, 0.0327, 0.0179, -0.0176, + 0.0066, -0.0069], device='cuda:0'), grad: tensor([ 2.1961e-06, 2.0131e-05, 3.1978e-05, -1.4696e-06, -1.1362e-07, + 5.3674e-05, -3.7879e-05, 9.7603e-06, -8.2910e-05, 4.5784e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 214.05, cls_loss 0.0019 cls_loss_mapping 0.0032 cls_loss_causal 0.5083 re_mapping 0.0061 re_causal 0.0185 /// teacc 98.93 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0110, -0.0440, -0.0388, ..., -0.1505, -0.0386, -0.0946], + [-0.0179, -0.0463, -0.0404, ..., 0.0187, 0.0049, -0.0954], + [ 0.0393, -0.0906, -0.0855, ..., 0.0231, -0.1192, 0.0557], + ..., + [ 0.0015, -0.0194, -0.1219, ..., -0.1052, 0.0185, 0.1118], + [ 0.0581, -0.0932, -0.0789, ..., -0.0972, -0.0477, -0.0700], + [-0.1294, 0.0122, 0.0995, ..., 0.0739, -0.0503, -0.1179]], + device='cuda:0'), grad: tensor([[ 1.1409e-06, 3.3807e-07, 1.3011e-06, ..., 2.3227e-06, + 4.2841e-08, 1.5246e-06], + [ 1.2266e-06, 4.5821e-07, 2.0154e-06, ..., 3.8445e-06, + 1.0245e-08, 8.7731e-07], + [ 4.4983e-07, 2.0862e-07, 6.0722e-07, ..., 1.1111e-06, + 1.5832e-08, 1.2051e-06], + ..., + [ 3.3341e-07, 9.5926e-08, 6.3889e-07, ..., 5.7310e-05, + 3.8184e-08, -1.7993e-06], + [ 5.2638e-06, 1.5134e-06, 6.6943e-06, ..., 7.9349e-06, + 7.9162e-08, 2.6003e-06], + [ 2.0713e-06, -7.1991e-07, -2.0247e-06, ..., 6.5938e-06, + 1.9465e-07, 4.6287e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0236, -0.0185, -0.0336, 0.0006, 0.0053, 0.0324, 0.0182, -0.0182, + 0.0066, -0.0070], device='cuda:0'), grad: tensor([ 1.9699e-05, 1.0407e-04, 2.4185e-05, 2.2054e-05, -1.9140e-03, + -6.9961e-06, 5.6028e-06, 1.5411e-03, 1.0419e-04, 1.0216e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 214.00, cls_loss 0.0017 cls_loss_mapping 0.0034 cls_loss_causal 0.5472 re_mapping 0.0061 re_causal 0.0200 /// teacc 99.00 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0112, -0.0441, -0.0388, ..., -0.1506, -0.0387, -0.0946], + [-0.0180, -0.0466, -0.0408, ..., 0.0184, 0.0049, -0.0957], + [ 0.0391, -0.0905, -0.0863, ..., 0.0231, -0.1199, 0.0561], + ..., + [ 0.0014, -0.0198, -0.1225, ..., -0.1058, 0.0188, 0.1119], + [ 0.0583, -0.0928, -0.0788, ..., -0.0972, -0.0478, -0.0701], + [-0.1296, 0.0127, 0.0996, ..., 0.0740, -0.0508, -0.1187]], + device='cuda:0'), grad: tensor([[-3.4049e-06, 1.3439e-06, 2.3860e-06, ..., 1.4994e-06, + 9.3132e-10, 4.4703e-08], + [ 1.1642e-07, 1.6317e-06, -1.2554e-06, ..., -3.0175e-06, + 7.4506e-08, 2.1700e-06], + [ 3.2410e-07, 8.9407e-08, 1.1399e-06, ..., 2.0694e-06, + 9.3132e-10, -3.1665e-07], + ..., + [ 3.3528e-08, -2.2911e-06, 1.9278e-06, ..., 1.4361e-06, + -1.4901e-07, -4.5039e-06], + [ 2.6729e-07, 1.0245e-06, 7.4413e-07, ..., 7.1526e-07, + 4.0047e-08, 1.1660e-06], + [ 8.1584e-07, -4.4294e-06, -8.4713e-06, ..., -4.2096e-06, + 1.8626e-09, 4.3120e-07]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0234, -0.0191, -0.0335, 0.0002, 0.0057, 0.0325, 0.0180, -0.0188, + 0.0072, -0.0069], device='cuda:0'), grad: tensor([-7.2680e-06, -5.9046e-06, 8.4043e-06, 3.8743e-06, 3.3379e-06, + 1.2405e-06, 4.8988e-06, -4.6529e-06, 4.9993e-06, -8.9109e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 214.14, cls_loss 0.0021 cls_loss_mapping 0.0042 cls_loss_causal 0.5083 re_mapping 0.0061 re_causal 0.0182 /// teacc 99.00 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 0.0116, -0.0443, -0.0389, ..., -0.1508, -0.0388, -0.0951], + [-0.0182, -0.0468, -0.0405, ..., 0.0188, 0.0049, -0.0969], + [ 0.0389, -0.0908, -0.0866, ..., 0.0230, -0.1202, 0.0565], + ..., + [ 0.0013, -0.0195, -0.1227, ..., -0.1060, 0.0191, 0.1127], + [ 0.0583, -0.0929, -0.0788, ..., -0.0974, -0.0482, -0.0708], + [-0.1299, 0.0128, 0.0997, ..., 0.0741, -0.0512, -0.1199]], + device='cuda:0'), grad: tensor([[-5.7090e-07, 1.5832e-08, 3.8277e-07, ..., -4.0978e-08, + 0.0000e+00, 4.0513e-07], + [ 7.9162e-08, 7.5437e-08, 1.3504e-07, ..., -6.1840e-07, + 4.6566e-09, 8.4378e-07], + [ 2.0489e-08, 7.7300e-08, 4.3958e-07, ..., 4.0885e-07, + 9.3132e-10, -6.8903e-05], + ..., + [ 2.4214e-08, -3.0641e-07, 4.5449e-07, ..., 5.6718e-07, + -1.3970e-08, 6.1750e-05], + [-1.1437e-06, -1.0338e-07, -1.6108e-05, ..., -8.8811e-06, + 1.8626e-09, 4.4815e-06], + [ 4.7404e-07, -5.9605e-08, 1.2353e-05, ..., 7.2047e-06, + 9.3132e-10, 9.3132e-08]], device='cuda:0') +Epoch 153, bias, value: tensor([-2.3356e-02, -1.9349e-02, -3.3384e-02, -5.8688e-05, 5.5851e-03, + 3.1564e-02, 1.9279e-02, -1.8170e-02, 7.1672e-03, -7.1442e-03], + device='cuda:0'), grad: tensor([ 1.4529e-07, 4.7591e-07, -1.0240e-04, 4.9137e-06, 1.6596e-06, + 2.3302e-06, 2.0340e-06, 9.4414e-05, -2.6986e-05, 2.3603e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 214.17, cls_loss 0.0020 cls_loss_mapping 0.0041 cls_loss_causal 0.5485 re_mapping 0.0063 re_causal 0.0196 /// teacc 98.99 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0126, -0.0445, -0.0387, ..., -0.1505, -0.0390, -0.0951], + [-0.0187, -0.0466, -0.0403, ..., 0.0193, 0.0059, -0.0979], + [ 0.0384, -0.0910, -0.0874, ..., 0.0225, -0.1212, 0.0568], + ..., + [ 0.0012, -0.0196, -0.1235, ..., -0.1067, 0.0185, 0.1134], + [ 0.0581, -0.0931, -0.0789, ..., -0.0975, -0.0484, -0.0713], + [-0.1303, 0.0130, 0.0997, ..., 0.0741, -0.0513, -0.1208]], + device='cuda:0'), grad: tensor([[ 7.9162e-08, 5.1223e-08, 2.1886e-07, ..., 3.4831e-07, + 0.0000e+00, 7.4506e-09], + [ 3.3528e-08, 1.5832e-08, -5.1782e-07, ..., -6.0536e-07, + 0.0000e+00, 1.8626e-08], + [ 2.6077e-08, 1.4901e-08, 1.2666e-07, ..., 1.5832e-07, + 0.0000e+00, -5.8673e-08], + ..., + [ 8.3819e-09, 2.5332e-07, 1.2321e-06, ..., 2.3767e-06, + 0.0000e+00, -1.6764e-08], + [ 1.0338e-07, 3.0827e-07, 1.4827e-06, ..., 1.9893e-06, + 1.0245e-08, 3.7253e-08], + [ 8.9407e-08, -4.1723e-07, -3.0734e-06, ..., 6.1952e-06, + 1.8626e-09, 2.7008e-08]], device='cuda:0') +Epoch 154, bias, value: tensor([-2.2624e-02, -1.9688e-02, -3.3614e-02, -1.0908e-05, 5.4765e-03, + 3.1340e-02, 1.9519e-02, -1.7817e-02, 6.9986e-03, -7.3701e-03], + device='cuda:0'), grad: tensor([ 8.1398e-07, -3.2187e-06, 5.8115e-07, 3.5428e-06, 1.9949e-06, + -1.7956e-05, -6.5472e-07, 5.2154e-06, 4.7162e-06, 4.9509e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 214.23, cls_loss 0.0019 cls_loss_mapping 0.0047 cls_loss_causal 0.5166 re_mapping 0.0062 re_causal 0.0183 /// teacc 98.86 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 0.0127, -0.0448, -0.0388, ..., -0.1507, -0.0392, -0.0955], + [-0.0189, -0.0469, -0.0403, ..., 0.0195, 0.0059, -0.0988], + [ 0.0383, -0.0914, -0.0881, ..., 0.0222, -0.1215, 0.0570], + ..., + [ 0.0008, -0.0191, -0.1238, ..., -0.1067, 0.0187, 0.1141], + [ 0.0583, -0.0932, -0.0791, ..., -0.0986, -0.0486, -0.0720], + [-0.1305, 0.0135, 0.1002, ..., 0.0744, -0.0516, -0.1216]], + device='cuda:0'), grad: tensor([[-2.3842e-07, 2.9337e-08, 3.6275e-07, ..., 4.9174e-07, + 0.0000e+00, 1.2573e-08], + [ 4.5169e-08, 8.6613e-08, -1.1949e-06, ..., -3.7290e-06, + 0.0000e+00, 2.8824e-07], + [ 4.7497e-08, 3.5856e-08, 3.3481e-07, ..., 7.1432e-07, + 0.0000e+00, 5.7276e-08], + ..., + [ 2.4680e-08, -6.8918e-08, 8.6706e-07, ..., 2.1551e-06, + 0.0000e+00, -6.4354e-07], + [ 4.8662e-07, 4.8848e-07, 1.3458e-06, ..., 5.5879e-06, + 4.6566e-10, -1.3830e-07], + [ 2.6962e-07, -2.4028e-06, -9.6858e-06, ..., -1.0937e-05, + 0.0000e+00, 1.8440e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0226, -0.0200, -0.0337, -0.0005, 0.0050, 0.0321, 0.0195, -0.0173, + 0.0068, -0.0072], device='cuda:0'), grad: tensor([-5.7369e-06, -8.3596e-06, 6.3404e-06, 3.7942e-06, 1.4275e-05, + -8.6650e-06, 3.0771e-06, 3.9265e-06, 7.2680e-06, -1.5959e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 214.41, cls_loss 0.0020 cls_loss_mapping 0.0044 cls_loss_causal 0.5239 re_mapping 0.0060 re_causal 0.0184 /// teacc 98.87 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 0.0128, -0.0450, -0.0388, ..., -0.1510, -0.0393, -0.0955], + [-0.0190, -0.0470, -0.0399, ..., 0.0205, 0.0059, -0.0982], + [ 0.0379, -0.0918, -0.0888, ..., 0.0216, -0.1216, 0.0566], + ..., + [ 0.0007, -0.0207, -0.1249, ..., -0.1078, 0.0187, 0.1132], + [ 0.0584, -0.0935, -0.0792, ..., -0.0995, -0.0489, -0.0721], + [-0.1307, 0.0145, 0.1009, ..., 0.0753, -0.0516, -0.1216]], + device='cuda:0'), grad: tensor([[ 1.1101e-05, 1.5274e-05, 1.5348e-05, ..., 1.3458e-07, + 4.6566e-10, 4.2375e-08], + [ 2.5937e-07, 4.1071e-07, 3.0920e-07, ..., -1.5786e-07, + -1.4435e-08, 9.3132e-08], + [ 2.5658e-07, 2.1327e-07, 2.5099e-07, ..., 8.8010e-08, + 9.3132e-10, 1.3923e-07], + ..., + [-3.0734e-08, 5.0664e-07, 1.3281e-06, ..., 1.3327e-06, + 2.3283e-09, -2.5425e-07], + [ 1.4221e-06, 2.4270e-06, 2.4177e-06, ..., 1.8999e-07, + 1.3970e-09, 2.7008e-08], + [ 1.6671e-07, -1.5972e-06, -3.6359e-06, ..., -3.8520e-06, + 6.0536e-09, 1.3690e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0226, -0.0189, -0.0352, 0.0006, 0.0042, 0.0320, 0.0199, -0.0182, + 0.0067, -0.0066], device='cuda:0'), grad: tensor([ 6.3598e-05, 1.3979e-06, 1.2983e-06, 3.7979e-06, 2.1160e-06, + 1.1936e-05, -8.9109e-05, 2.2519e-06, 9.3728e-06, -6.7428e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 214.43, cls_loss 0.0021 cls_loss_mapping 0.0040 cls_loss_causal 0.4858 re_mapping 0.0065 re_causal 0.0178 /// teacc 99.02 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 0.0136, -0.0461, -0.0394, ..., -0.1513, -0.0394, -0.0974], + [-0.0191, -0.0472, -0.0398, ..., 0.0208, 0.0059, -0.0990], + [ 0.0379, -0.0919, -0.0893, ..., 0.0217, -0.1218, 0.0575], + ..., + [ 0.0004, -0.0208, -0.1254, ..., -0.1083, 0.0187, 0.1132], + [ 0.0584, -0.0939, -0.0790, ..., -0.0998, -0.0491, -0.0727], + [-0.1310, 0.0148, 0.1014, ..., 0.0755, -0.0516, -0.1225]], + device='cuda:0'), grad: tensor([[-1.4198e-04, 6.6124e-08, -2.0936e-05, ..., 3.6787e-08, + 5.1223e-09, 1.4342e-07], + [ 3.2708e-06, 3.3947e-07, -1.3318e-07, ..., -9.0059e-07, + -1.8300e-07, 7.5158e-07], + [ 2.9895e-06, 2.1374e-07, 4.1770e-07, ..., 9.3132e-08, + 1.6298e-08, -7.1852e-07], + ..., + [ 2.7362e-06, -1.7285e-06, 5.0804e-07, ..., 3.2550e-07, + -6.0536e-09, -2.5555e-06], + [ 1.2694e-06, 4.1025e-07, 5.1549e-07, ..., 5.4156e-07, + 4.0978e-08, 1.4082e-06], + [ 1.0759e-05, 3.9162e-07, 9.1502e-07, ..., -8.3819e-08, + 5.4482e-08, 1.0729e-06]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0228, -0.0190, -0.0346, 0.0013, 0.0041, 0.0313, 0.0197, -0.0183, + 0.0065, -0.0063], device='cuda:0'), grad: tensor([-3.5954e-04, 9.0078e-06, 9.2164e-06, 1.4745e-05, 3.8631e-06, + 2.7195e-05, 2.5463e-04, 4.5002e-06, 5.7481e-06, 3.0175e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 214.16, cls_loss 0.0016 cls_loss_mapping 0.0032 cls_loss_causal 0.5248 re_mapping 0.0061 re_causal 0.0186 /// teacc 99.09 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 0.0149, -0.0465, -0.0395, ..., -0.1517, -0.0394, -0.0978], + [-0.0199, -0.0481, -0.0403, ..., 0.0206, 0.0059, -0.0997], + [ 0.0377, -0.0924, -0.0905, ..., 0.0215, -0.1221, 0.0579], + ..., + [ 0.0003, -0.0200, -0.1253, ..., -0.1084, 0.0188, 0.1136], + [ 0.0582, -0.0945, -0.0792, ..., -0.1003, -0.0511, -0.0732], + [-0.1313, 0.0149, 0.1014, ..., 0.0755, -0.0517, -0.1231]], + device='cuda:0'), grad: tensor([[-1.1222e-06, 2.2817e-08, 6.7987e-08, ..., 1.1269e-07, + 2.7940e-09, 6.7521e-08], + [ 6.8452e-08, -6.0070e-08, -1.7546e-06, ..., -2.5146e-06, + -5.6345e-08, 2.6310e-07], + [ 6.5193e-08, 4.7497e-08, 1.2433e-07, ..., 7.7765e-08, + 2.7940e-09, -7.4599e-07], + ..., + [ 5.0291e-08, 1.7416e-07, 1.1856e-06, ..., 1.2144e-06, + 1.1642e-08, 2.3283e-07], + [ 8.1491e-08, 2.0396e-07, 1.2554e-06, ..., 1.0589e-06, + 2.9802e-08, 3.0547e-07], + [ 5.2107e-07, -1.6345e-07, -3.0138e-06, ..., -9.4250e-07, + 1.7695e-08, 1.8487e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0221, -0.0201, -0.0345, 0.0008, 0.0043, 0.0315, 0.0199, -0.0175, + 0.0059, -0.0066], device='cuda:0'), grad: tensor([-2.8815e-06, -6.8247e-06, -1.3877e-07, -2.7334e-07, 1.9353e-06, + 4.4191e-07, 1.6876e-06, 4.9397e-06, 2.9104e-07, 8.1630e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 214.31, cls_loss 0.0024 cls_loss_mapping 0.0046 cls_loss_causal 0.5525 re_mapping 0.0061 re_causal 0.0186 /// teacc 99.03 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 1.4998e-02, -4.7427e-02, -3.9757e-02, ..., -1.5228e-01, + -3.9689e-02, -9.8258e-02], + [-2.0227e-02, -4.8230e-02, -4.0027e-02, ..., 2.0955e-02, + 6.0889e-03, -1.0024e-01], + [ 3.7767e-02, -9.3293e-02, -9.1949e-02, ..., 2.1716e-02, + -1.2260e-01, 5.7279e-02], + ..., + [-7.5509e-05, -2.1655e-02, -1.2873e-01, ..., -1.1154e-01, + 1.8671e-02, 1.1481e-01], + [ 5.8247e-02, -9.4839e-02, -7.9280e-02, ..., -1.0141e-01, + -5.1910e-02, -7.3777e-02], + [-1.3170e-01, 1.5335e-02, 1.0065e-01, ..., 7.6024e-02, + -5.1766e-02, -1.2262e-01]], device='cuda:0'), grad: tensor([[ 1.4435e-07, 5.1223e-08, 2.7660e-07, ..., 1.9185e-07, + 1.7695e-08, 1.9558e-08], + [ 3.0734e-08, 1.8999e-07, 3.1572e-07, ..., 2.4680e-07, + -8.3819e-09, 2.2352e-07], + [ 2.8871e-08, 3.8184e-08, 1.1269e-07, ..., 1.1828e-07, + 1.3039e-08, 6.7987e-08], + ..., + [ 1.8626e-08, -1.2107e-08, 6.1281e-07, ..., 6.9104e-07, + -6.4261e-08, -6.3237e-07], + [ 3.6601e-07, 1.9372e-07, 3.8464e-07, ..., 1.0338e-06, + 1.1176e-08, 9.2201e-08], + [ 2.1327e-07, -1.7229e-06, -6.4671e-06, ..., -5.9977e-06, + 2.8871e-08, 1.7509e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0222, -0.0200, -0.0354, 0.0005, 0.0061, 0.0321, 0.0201, -0.0188, + 0.0057, -0.0070], device='cuda:0'), grad: tensor([ 6.1933e-07, 8.2050e-07, 4.2748e-07, 1.1250e-06, 7.0743e-06, + -1.2852e-06, -9.6112e-07, 1.4249e-07, 1.7630e-06, -9.7081e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 214.25, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5793 re_mapping 0.0060 re_causal 0.0189 /// teacc 99.07 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0150, -0.0478, -0.0400, ..., -0.1532, -0.0407, -0.0994], + [-0.0215, -0.0484, -0.0405, ..., 0.0206, 0.0059, -0.1005], + [ 0.0384, -0.0937, -0.0926, ..., 0.0210, -0.1235, 0.0574], + ..., + [-0.0003, -0.0216, -0.1289, ..., -0.1117, 0.0186, 0.1151], + [ 0.0575, -0.0942, -0.0795, ..., -0.1034, -0.0547, -0.0740], + [-0.1328, 0.0156, 0.1012, ..., 0.0767, -0.0519, -0.1233]], + device='cuda:0'), grad: tensor([[-8.1956e-06, 1.7416e-07, -4.0010e-06, ..., 1.2107e-08, + 0.0000e+00, 2.3935e-07], + [ 8.2981e-07, 1.2163e-06, 1.4221e-06, ..., -2.3935e-07, + -4.6566e-09, 3.2317e-06], + [ 1.1204e-06, 2.9951e-06, 8.3353e-07, ..., 6.0536e-08, + 9.3132e-10, 7.9647e-06], + ..., + [ 1.8533e-07, -6.4299e-06, 1.4622e-07, ..., 8.4750e-08, + 1.8626e-09, -2.1085e-05], + [-1.5348e-05, -3.9823e-06, -3.5912e-05, ..., 7.0781e-08, + 1.8626e-09, 1.0775e-06], + [ 7.4431e-06, 1.2806e-06, 4.6417e-06, ..., 2.7474e-07, + 9.3132e-10, 2.8498e-06]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0223, -0.0206, -0.0353, 0.0002, 0.0058, 0.0348, 0.0194, -0.0187, + 0.0042, -0.0069], device='cuda:0'), grad: tensor([-3.3140e-05, 1.2688e-05, 2.0072e-05, 6.3539e-05, 1.3314e-05, + 1.1064e-05, 7.2420e-05, -3.3379e-05, -1.6594e-04, 3.9518e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 214.35, cls_loss 0.0016 cls_loss_mapping 0.0032 cls_loss_causal 0.4942 re_mapping 0.0059 re_causal 0.0181 /// teacc 99.05 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0148, -0.0488, -0.0401, ..., -0.1546, -0.0410, -0.0999], + [-0.0222, -0.0487, -0.0407, ..., 0.0205, 0.0060, -0.1009], + [ 0.0374, -0.0953, -0.0933, ..., 0.0200, -0.1238, 0.0571], + ..., + [-0.0004, -0.0211, -0.1290, ..., -0.1118, 0.0186, 0.1156], + [ 0.0579, -0.0944, -0.0794, ..., -0.1031, -0.0552, -0.0743], + [-0.1332, 0.0156, 0.1014, ..., 0.0768, -0.0519, -0.1243]], + device='cuda:0'), grad: tensor([[ 9.1270e-08, 3.2783e-07, 3.6508e-07, ..., 2.8126e-07, + 5.5879e-09, 3.5763e-06], + [ 1.4529e-07, 1.2927e-06, -3.6322e-07, ..., 4.3865e-07, + -1.8626e-09, 2.2445e-06], + [ 1.8161e-07, 2.3730e-06, 3.1292e-07, ..., 1.7108e-06, + 1.8626e-09, -9.3132e-06], + ..., + [ 9.0338e-08, -6.7055e-05, 4.3493e-07, ..., -5.2750e-05, + 3.7253e-09, -1.0437e-04], + [-1.3206e-06, 6.1765e-06, -5.6773e-06, ..., 3.5670e-06, + 4.0978e-08, 1.1541e-05], + [ 1.2405e-06, 5.7966e-05, 4.5151e-06, ..., 4.6670e-05, + 1.4901e-08, 8.8751e-05]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0226, -0.0210, -0.0360, 0.0005, 0.0057, 0.0348, 0.0196, -0.0183, + 0.0047, -0.0070], device='cuda:0'), grad: tensor([ 9.5293e-06, 5.5358e-06, -1.9699e-05, -3.4064e-05, 3.7253e-06, + 4.7356e-05, -2.3227e-06, -2.9659e-04, 2.1964e-05, 2.6441e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 214.10, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5275 re_mapping 0.0057 re_causal 0.0174 /// teacc 98.94 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0152, -0.0493, -0.0402, ..., -0.1556, -0.0413, -0.1018], + [-0.0226, -0.0485, -0.0400, ..., 0.0212, 0.0060, -0.1013], + [ 0.0382, -0.0957, -0.0941, ..., 0.0199, -0.1243, 0.0578], + ..., + [-0.0007, -0.0205, -0.1291, ..., -0.1115, 0.0187, 0.1164], + [ 0.0582, -0.0940, -0.0796, ..., -0.1035, -0.0557, -0.0750], + [-0.1335, 0.0155, 0.1017, ..., 0.0766, -0.0520, -0.1277]], + device='cuda:0'), grad: tensor([[ 8.9873e-07, 3.8184e-08, 8.4471e-07, ..., 1.3132e-07, + 0.0000e+00, 9.2201e-08], + [ 2.5146e-06, 1.0058e-06, 5.1707e-06, ..., 4.7423e-06, + 0.0000e+00, 5.7556e-07], + [ 8.7917e-07, 2.7474e-07, 7.4413e-07, ..., 2.2259e-07, + 0.0000e+00, 1.5097e-06], + ..., + [ 6.7055e-08, 9.2201e-08, 4.6846e-07, ..., 6.6683e-07, + 0.0000e+00, -9.1270e-08], + [-2.6077e-08, 2.3749e-07, 2.4587e-07, ..., 3.0082e-07, + 0.0000e+00, 1.1669e-06], + [ 1.9092e-07, 6.0536e-07, 2.0843e-06, ..., 2.9840e-06, + 0.0000e+00, 3.8091e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0229, -0.0203, -0.0351, 0.0001, 0.0055, 0.0349, 0.0187, -0.0177, + 0.0048, -0.0074], device='cuda:0'), grad: tensor([ 2.6654e-06, 1.5289e-05, 5.3719e-06, -9.1195e-06, -1.4976e-05, + -1.7723e-06, -7.8827e-06, 1.1604e-06, 2.4661e-06, 6.7391e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 214.15, cls_loss 0.0021 cls_loss_mapping 0.0037 cls_loss_causal 0.5365 re_mapping 0.0060 re_causal 0.0182 /// teacc 98.99 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0157, -0.0496, -0.0403, ..., -0.1560, -0.0416, -0.1023], + [-0.0232, -0.0489, -0.0398, ..., 0.0217, 0.0060, -0.1019], + [ 0.0379, -0.0960, -0.0958, ..., 0.0191, -0.1248, 0.0585], + ..., + [-0.0012, -0.0202, -0.1293, ..., -0.1116, 0.0188, 0.1164], + [ 0.0591, -0.0942, -0.0800, ..., -0.1041, -0.0560, -0.0754], + [-0.1338, 0.0155, 0.1028, ..., 0.0777, -0.0522, -0.1285]], + device='cuda:0'), grad: tensor([[ 1.8375e-06, 7.3854e-07, 6.1467e-08, ..., 3.2503e-06, + 5.4017e-08, 2.0489e-08], + [ 3.0454e-07, 1.4994e-07, -2.7940e-09, ..., -1.8746e-05, + 1.9558e-08, 7.3574e-08], + [ 1.6674e-05, 6.7577e-06, 4.7497e-08, ..., 3.4541e-05, + 1.0245e-08, -1.0431e-07], + ..., + [ 4.2003e-07, 6.4261e-08, 2.4587e-07, ..., 4.8093e-06, + 1.4249e-07, -8.9407e-08], + [ 3.2753e-05, 1.3337e-05, 2.8405e-07, ..., 4.5955e-05, + 2.7101e-07, 5.2154e-08], + [ 1.6959e-06, -1.0710e-07, -7.3668e-07, ..., 1.6224e-06, + 5.1223e-07, 3.8184e-08]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0228, -0.0205, -0.0351, 0.0005, 0.0047, 0.0345, 0.0186, -0.0176, + 0.0052, -0.0069], device='cuda:0'), grad: tensor([ 6.2436e-06, -4.6670e-05, 7.0989e-05, 2.2709e-05, 3.3677e-06, + -1.7369e-04, 1.7777e-05, 1.1489e-05, 8.3566e-05, 4.2506e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 213.89, cls_loss 0.0029 cls_loss_mapping 0.0038 cls_loss_causal 0.5287 re_mapping 0.0067 re_causal 0.0185 /// teacc 98.98 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0167, -0.0502, -0.0411, ..., -0.1570, -0.0425, -0.1028], + [-0.0236, -0.0490, -0.0396, ..., 0.0219, 0.0065, -0.1022], + [ 0.0377, -0.0968, -0.0967, ..., 0.0186, -0.1268, 0.0590], + ..., + [-0.0013, -0.0204, -0.1307, ..., -0.1124, 0.0191, 0.1164], + [ 0.0593, -0.0949, -0.0804, ..., -0.1048, -0.0571, -0.0758], + [-0.1341, 0.0174, 0.1038, ..., 0.0789, -0.0524, -0.1290]], + device='cuda:0'), grad: tensor([[-3.1590e-06, 2.7008e-08, 8.7544e-08, ..., 5.4017e-08, + 4.6566e-09, 1.5646e-07], + [ 6.3330e-07, 4.9360e-08, -1.4808e-07, ..., -4.7591e-07, + 0.0000e+00, 1.0058e-07], + [ 5.3924e-07, 4.4703e-08, 1.9465e-07, ..., 6.5193e-08, + 0.0000e+00, -9.4902e-07], + ..., + [ 1.8626e-07, -1.3039e-08, 2.5239e-07, ..., 2.8778e-07, + 0.0000e+00, 3.0268e-07], + [ 5.5879e-07, 3.8184e-08, 1.0245e-07, ..., 1.4529e-07, + 0.0000e+00, 1.3597e-07], + [ 6.1188e-07, 3.9116e-08, 7.2084e-07, ..., 7.5623e-07, + 0.0000e+00, 1.1828e-07]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0234, -0.0206, -0.0350, 0.0005, 0.0043, 0.0345, 0.0180, -0.0182, + 0.0052, -0.0057], device='cuda:0'), grad: tensor([-7.2941e-06, -3.4552e-07, 7.7765e-07, -1.6950e-07, -2.6729e-06, + 1.4752e-06, 1.4408e-06, 1.4221e-06, 1.9539e-06, 3.4161e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 213.78, cls_loss 0.0018 cls_loss_mapping 0.0036 cls_loss_causal 0.5553 re_mapping 0.0062 re_causal 0.0188 /// teacc 99.04 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0176, -0.0519, -0.0410, ..., -0.1567, -0.0434, -0.1031], + [-0.0240, -0.0493, -0.0397, ..., 0.0217, 0.0064, -0.1024], + [ 0.0376, -0.0988, -0.0979, ..., 0.0185, -0.1274, 0.0588], + ..., + [-0.0013, -0.0203, -0.1315, ..., -0.1127, 0.0197, 0.1166], + [ 0.0593, -0.0952, -0.0810, ..., -0.1056, -0.0575, -0.0761], + [-0.1349, 0.0175, 0.1039, ..., 0.0791, -0.0532, -0.1296]], + device='cuda:0'), grad: tensor([[-2.9430e-07, 4.6752e-07, 6.8732e-07, ..., 9.9186e-07, + 0.0000e+00, 4.1537e-07], + [ 4.1910e-08, 2.1532e-06, 4.1537e-07, ..., 3.1441e-06, + 0.0000e+00, 3.7700e-06], + [ 7.5437e-08, 2.9534e-05, 7.8883e-07, ..., 5.1856e-05, + 0.0000e+00, 5.9336e-05], + ..., + [ 2.0489e-08, -4.3839e-05, 6.9849e-07, ..., -7.7426e-05, + 0.0000e+00, -8.9884e-05], + [-2.4028e-07, 6.7726e-06, 3.2578e-06, ..., 1.1280e-05, + 0.0000e+00, 2.2873e-06], + [ 3.8464e-07, -4.8336e-07, -1.1615e-05, ..., -1.6475e-06, + 0.0000e+00, 1.8120e-05]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0230, -0.0205, -0.0355, 0.0007, 0.0046, 0.0344, 0.0181, -0.0181, + 0.0050, -0.0060], device='cuda:0'), grad: tensor([ 4.9267e-07, 1.0662e-05, 1.5199e-04, -1.4408e-06, 2.4840e-05, + 3.2149e-06, 4.3958e-06, -2.2435e-04, 2.0966e-05, 9.4548e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.16, cls_loss 0.0019 cls_loss_mapping 0.0030 cls_loss_causal 0.5221 re_mapping 0.0061 re_causal 0.0176 /// teacc 98.92 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0178, -0.0518, -0.0419, ..., -0.1571, -0.0441, -0.1035], + [-0.0243, -0.0491, -0.0394, ..., 0.0222, 0.0064, -0.1022], + [ 0.0366, -0.0992, -0.0988, ..., 0.0181, -0.1279, 0.0570], + ..., + [-0.0007, -0.0203, -0.1319, ..., -0.1130, 0.0199, 0.1184], + [ 0.0594, -0.0958, -0.0812, ..., -0.1060, -0.0577, -0.0770], + [-0.1352, 0.0174, 0.1043, ..., 0.0790, -0.0533, -0.1307]], + device='cuda:0'), grad: tensor([[ 4.7963e-07, -7.0781e-07, 3.2503e-07, ..., 1.9744e-07, + 0.0000e+00, 7.2680e-06], + [ 2.3562e-07, -3.7104e-05, -1.9872e-04, ..., -1.4198e-04, + 0.0000e+00, 2.7996e-06], + [-4.8243e-06, -2.0023e-07, 3.4180e-07, ..., -1.1446e-06, + 0.0000e+00, -1.8179e-05], + ..., + [ 1.6857e-07, 5.7369e-06, 3.0696e-05, ..., 2.2262e-05, + 0.0000e+00, 2.4959e-07], + [ 2.0023e-07, 3.3807e-07, 1.1362e-06, ..., 2.1216e-06, + 0.0000e+00, 6.7689e-06], + [ 2.3358e-06, 2.9728e-05, 1.5914e-04, ..., 1.1283e-04, + 0.0000e+00, 6.3330e-08]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0235, -0.0201, -0.0373, 0.0004, 0.0049, 0.0345, 0.0178, -0.0169, + 0.0048, -0.0059], device='cuda:0'), grad: tensor([ 6.7726e-06, -3.6383e-04, -5.1320e-05, 3.5651e-06, 1.3344e-05, + 1.6186e-06, 3.0510e-06, 5.8949e-05, 2.1607e-05, 3.0637e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 214.24, cls_loss 0.0019 cls_loss_mapping 0.0042 cls_loss_causal 0.5270 re_mapping 0.0059 re_causal 0.0178 /// teacc 98.99 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0180, -0.0531, -0.0422, ..., -0.1576, -0.0446, -0.1050], + [-0.0252, -0.0493, -0.0391, ..., 0.0226, 0.0064, -0.1032], + [ 0.0367, -0.0992, -0.0992, ..., 0.0180, -0.1282, 0.0571], + ..., + [-0.0006, -0.0201, -0.1322, ..., -0.1134, 0.0200, 0.1188], + [ 0.0596, -0.0964, -0.0813, ..., -0.1065, -0.0577, -0.0777], + [-0.1355, 0.0173, 0.1045, ..., 0.0792, -0.0534, -0.1321]], + device='cuda:0'), grad: tensor([[-5.2936e-06, 1.6764e-08, 5.6159e-07, ..., 8.8569e-07, + 0.0000e+00, 6.9849e-08], + [ 4.6380e-07, 9.4995e-08, 9.3691e-07, ..., 2.0936e-06, + 0.0000e+00, 2.5239e-07], + [ 3.4738e-07, 1.6298e-07, 1.7826e-06, ..., 4.8801e-06, + 0.0000e+00, 9.7789e-08], + ..., + [ 1.3132e-07, 1.2666e-07, 2.7996e-06, ..., 7.6592e-06, + 0.0000e+00, 3.3621e-07], + [-6.3330e-08, 1.8533e-07, 1.3495e-06, ..., 3.4813e-06, + 0.0000e+00, 6.9570e-07], + [ 8.3912e-07, -7.1712e-08, 1.9111e-06, ..., 5.6438e-06, + 0.0000e+00, 8.1025e-08]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0238, -0.0205, -0.0372, 0.0010, 0.0049, 0.0346, 0.0179, -0.0164, + 0.0045, -0.0061], device='cuda:0'), grad: tensor([-2.1175e-05, 8.7246e-06, 1.6958e-05, -4.0792e-06, -6.7830e-05, + 6.7689e-06, 4.1388e-06, 2.4498e-05, 1.0923e-05, 2.1070e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 214.17, cls_loss 0.0022 cls_loss_mapping 0.0046 cls_loss_causal 0.5218 re_mapping 0.0059 re_causal 0.0172 /// teacc 98.89 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0183, -0.0539, -0.0423, ..., -0.1582, -0.0448, -0.1069], + [-0.0257, -0.0506, -0.0390, ..., 0.0229, 0.0064, -0.1043], + [ 0.0365, -0.0997, -0.1003, ..., 0.0177, -0.1283, 0.0575], + ..., + [-0.0008, -0.0194, -0.1327, ..., -0.1140, 0.0201, 0.1192], + [ 0.0595, -0.0966, -0.0828, ..., -0.1080, -0.0580, -0.0780], + [-0.1352, 0.0173, 0.1050, ..., 0.0797, -0.0535, -0.1332]], + device='cuda:0'), grad: tensor([[ 1.2219e-06, 6.5193e-09, 1.0766e-06, ..., 9.2201e-08, + 3.7253e-09, 1.0245e-08], + [ 6.3330e-08, 1.0990e-07, 4.9081e-07, ..., 3.4086e-07, + 9.3132e-10, 3.3993e-07], + [ 8.5123e-07, 1.4901e-08, 1.4519e-06, ..., 1.3411e-07, + 0.0000e+00, 2.0489e-08], + ..., + [ 1.2107e-08, -1.9558e-08, 7.0222e-07, ..., 7.3668e-07, + 1.8626e-09, -5.3924e-07], + [-3.5390e-08, 4.5635e-08, -9.3132e-09, ..., 1.4901e-07, + 1.9558e-08, 1.2666e-07], + [ 1.4622e-07, 5.3085e-07, 9.3728e-06, ..., 8.2031e-06, + 1.6764e-08, 8.7544e-08]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0238, -0.0214, -0.0371, 0.0012, 0.0050, 0.0344, 0.0180, -0.0161, + 0.0040, -0.0058], device='cuda:0'), grad: tensor([ 3.6657e-06, -5.1130e-07, 5.2415e-06, 1.6680e-06, -2.0131e-05, + 1.1176e-06, -9.9093e-06, 8.4657e-07, 2.2631e-07, 1.7792e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 214.40, cls_loss 0.0037 cls_loss_mapping 0.0062 cls_loss_causal 0.5055 re_mapping 0.0061 re_causal 0.0172 /// teacc 98.92 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0194, -0.0543, -0.0424, ..., -0.1585, -0.0451, -0.1079], + [-0.0262, -0.0540, -0.0365, ..., 0.0222, 0.0065, -0.1071], + [ 0.0364, -0.1000, -0.1013, ..., 0.0172, -0.1290, 0.0583], + ..., + [-0.0010, -0.0184, -0.1331, ..., -0.1145, 0.0200, 0.1204], + [ 0.0599, -0.0971, -0.0849, ..., -0.1110, -0.0581, -0.0786], + [-0.1357, 0.0173, 0.1050, ..., 0.0795, -0.0536, -0.1341]], + device='cuda:0'), grad: tensor([[-1.0617e-07, 2.0489e-07, 5.5879e-08, ..., 1.3039e-08, + 0.0000e+00, 5.0478e-07], + [ 5.4948e-08, 1.1317e-05, 6.3218e-06, ..., -5.9977e-07, + 0.0000e+00, 1.5888e-06], + [-1.5944e-06, 4.7963e-07, 1.4808e-07, ..., 4.0047e-08, + 0.0000e+00, -7.5512e-06], + ..., + [ 1.0058e-07, -2.1026e-05, -1.2495e-05, ..., 2.3190e-07, + 0.0000e+00, -1.9651e-06], + [ 1.2470e-06, 2.0862e-07, 2.2072e-07, ..., 6.9849e-08, + 0.0000e+00, 6.2883e-06], + [ 8.4750e-08, 1.1176e-05, 1.6868e-05, ..., 9.7603e-06, + 0.0000e+00, 1.2685e-06]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0235, -0.0223, -0.0367, -0.0023, 0.0052, 0.0380, 0.0176, -0.0145, + 0.0014, -0.0062], device='cuda:0'), grad: tensor([ 3.9581e-07, 4.6134e-05, -1.4365e-05, -2.5816e-06, -1.4611e-05, + 2.0768e-06, -2.4214e-08, -8.6546e-05, 1.3381e-05, 5.6267e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 214.21, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5158 re_mapping 0.0063 re_causal 0.0183 /// teacc 98.94 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0198, -0.0546, -0.0424, ..., -0.1590, -0.0455, -0.1098], + [-0.0271, -0.0546, -0.0363, ..., 0.0220, 0.0065, -0.1086], + [ 0.0362, -0.1007, -0.1016, ..., 0.0172, -0.1290, 0.0576], + ..., + [-0.0012, -0.0183, -0.1339, ..., -0.1150, 0.0200, 0.1223], + [ 0.0598, -0.0976, -0.0846, ..., -0.1115, -0.0626, -0.0790], + [-0.1359, 0.0173, 0.1053, ..., 0.0799, -0.0536, -0.1353]], + device='cuda:0'), grad: tensor([[ 5.1223e-08, 9.3132e-10, 2.5146e-08, ..., 1.6764e-08, + 0.0000e+00, 1.2852e-07], + [ 3.7253e-08, 1.5832e-08, -2.4401e-07, ..., -4.0606e-07, + 0.0000e+00, 1.6671e-07], + [-1.2713e-06, 4.0513e-07, 3.9116e-08, ..., 1.5646e-07, + 0.0000e+00, -1.8906e-07], + ..., + [ 2.6170e-07, -4.7777e-07, 1.6298e-07, ..., -5.2154e-08, + 0.0000e+00, -2.9579e-06], + [ 6.5099e-07, 8.3819e-09, 1.1176e-08, ..., 5.2154e-08, + 0.0000e+00, 2.2352e-06], + [ 1.3039e-08, 2.5146e-08, 9.3132e-10, ..., 1.2852e-07, + -4.6566e-09, 3.1944e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0235, -0.0231, -0.0380, -0.0031, 0.0051, 0.0386, 0.0176, -0.0127, + 0.0018, -0.0068], device='cuda:0'), grad: tensor([ 3.1665e-07, -5.9605e-07, -2.7791e-06, 2.9523e-07, 3.2876e-07, + 8.6240e-07, -2.4494e-07, -2.6263e-06, 3.5763e-06, 8.6427e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.20, cls_loss 0.0023 cls_loss_mapping 0.0052 cls_loss_causal 0.5366 re_mapping 0.0062 re_causal 0.0180 /// teacc 98.99 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0198, -0.0550, -0.0426, ..., -0.1601, -0.0471, -0.1106], + [-0.0274, -0.0546, -0.0363, ..., 0.0220, 0.0069, -0.1085], + [ 0.0365, -0.1010, -0.1024, ..., 0.0173, -0.1329, 0.0578], + ..., + [-0.0017, -0.0186, -0.1351, ..., -0.1157, 0.0209, 0.1222], + [ 0.0596, -0.0981, -0.0846, ..., -0.1119, -0.0630, -0.0803], + [-0.1361, 0.0178, 0.1057, ..., 0.0805, -0.0543, -0.1358]], + device='cuda:0'), grad: tensor([[-8.7544e-08, 9.1568e-06, 3.8743e-07, ..., 2.0280e-05, + 0.0000e+00, 1.8984e-05], + [ 3.2596e-08, 2.6543e-06, 6.6217e-07, ..., 5.9083e-06, + 0.0000e+00, 4.2841e-06], + [ 1.0896e-07, 1.1986e-06, 1.3309e-06, ..., 2.2352e-06, + 0.0000e+00, -1.8537e-05], + ..., + [-1.9558e-08, 1.9083e-03, 4.2003e-07, ..., 4.2419e-03, + -2.7940e-09, 2.4567e-03], + [-3.1665e-07, 6.9756e-07, 7.9349e-07, ..., 1.3234e-06, + 0.0000e+00, 7.8827e-06], + [ 1.2480e-07, -1.9312e-03, 3.3528e-08, ..., -4.2915e-03, + 9.3132e-10, -2.4853e-03]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0236, -0.0231, -0.0381, -0.0025, 0.0048, 0.0387, 0.0179, -0.0132, + 0.0015, -0.0063], device='cuda:0'), grad: tensor([ 6.2823e-05, 1.6809e-05, -3.1650e-05, 3.1888e-05, 5.9903e-06, + 1.2815e-05, 7.7207e-07, 1.0109e-02, 1.7375e-05, -1.0231e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 214.07, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5417 re_mapping 0.0058 re_causal 0.0185 /// teacc 98.94 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0200, -0.0553, -0.0429, ..., -0.1607, -0.0490, -0.1113], + [-0.0277, -0.0551, -0.0361, ..., 0.0210, 0.0069, -0.1085], + [ 0.0365, -0.1013, -0.1030, ..., 0.0167, -0.1337, 0.0579], + ..., + [-0.0018, -0.0205, -0.1350, ..., -0.1185, 0.0218, 0.1208], + [ 0.0599, -0.0983, -0.0843, ..., -0.1121, -0.0631, -0.0805], + [-0.1365, 0.0194, 0.1057, ..., 0.0824, -0.0552, -0.1335]], + device='cuda:0'), grad: tensor([[ 9.5181e-07, 1.1176e-08, 2.0582e-07, ..., 1.5832e-07, + 0.0000e+00, 4.9174e-07], + [ 2.1793e-07, 4.7497e-08, 1.2014e-07, ..., -3.5409e-06, + 0.0000e+00, -4.8131e-06], + [ 2.2333e-06, 4.3772e-08, 2.0862e-07, ..., 2.9206e-06, + 0.0000e+00, 4.9882e-06], + ..., + [ 9.2946e-07, -3.9395e-07, 3.6787e-07, ..., 7.8697e-07, + 0.0000e+00, 3.1106e-07], + [-5.7667e-06, 2.2352e-08, -8.5011e-06, ..., -6.5118e-06, + 0.0000e+00, -2.6152e-06], + [ 9.1270e-08, 4.1910e-07, 5.4240e-06, ..., 4.2915e-06, + 9.3132e-10, 8.3074e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0236, -0.0234, -0.0383, -0.0027, 0.0049, 0.0392, 0.0175, -0.0148, + 0.0017, -0.0050], device='cuda:0'), grad: tensor([ 6.2622e-06, -3.2801e-06, 3.0994e-05, 7.4096e-06, 2.6338e-06, + 2.5690e-05, 5.8375e-06, 5.2378e-06, -1.4400e-04, 6.3241e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 214.01, cls_loss 0.0018 cls_loss_mapping 0.0039 cls_loss_causal 0.5290 re_mapping 0.0056 re_causal 0.0175 /// teacc 99.03 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0200, -0.0556, -0.0430, ..., -0.1610, -0.0495, -0.1118], + [-0.0286, -0.0551, -0.0363, ..., 0.0210, 0.0069, -0.1083], + [ 0.0363, -0.1016, -0.1034, ..., 0.0167, -0.1338, 0.0578], + ..., + [-0.0019, -0.0197, -0.1337, ..., -0.1178, 0.0230, 0.1211], + [ 0.0607, -0.0981, -0.0840, ..., -0.1120, -0.0631, -0.0810], + [-0.1369, 0.0182, 0.1051, ..., 0.0815, -0.0569, -0.1342]], + device='cuda:0'), grad: tensor([[ 1.9558e-08, 1.0245e-08, 1.7788e-07, ..., 6.8918e-08, + 3.7253e-09, 1.5181e-07], + [ 2.1420e-08, 4.0978e-08, 0.0000e+00, ..., -7.4506e-09, + 5.5879e-09, 1.3225e-07], + [-1.7416e-07, 3.7253e-08, 2.2352e-08, ..., 1.5832e-08, + 2.7940e-09, -7.1712e-07], + ..., + [ 1.9558e-08, -2.3469e-07, 9.5926e-08, ..., 5.6811e-08, + -2.8871e-08, -5.5507e-07], + [ 1.6857e-07, 4.5635e-08, 5.7742e-08, ..., 3.9954e-07, + 3.6322e-08, 2.9244e-07], + [ 1.0896e-07, -3.2596e-08, -1.5488e-06, ..., -7.9535e-07, + 1.8626e-08, 2.9709e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0237, -0.0234, -0.0387, -0.0026, 0.0052, 0.0391, 0.0177, -0.0143, + 0.0021, -0.0061], device='cuda:0'), grad: tensor([ 4.0419e-07, 2.9709e-07, -1.4640e-06, 1.1427e-06, 2.2408e-06, + -1.0887e-06, 1.8440e-07, -9.2946e-07, 1.1148e-06, -1.8971e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 213.83, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5183 re_mapping 0.0056 re_causal 0.0171 /// teacc 98.86 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0198, -0.0559, -0.0435, ..., -0.1613, -0.0498, -0.1124], + [-0.0291, -0.0553, -0.0362, ..., 0.0211, 0.0067, -0.1085], + [ 0.0366, -0.1018, -0.1038, ..., 0.0168, -0.1337, 0.0582], + ..., + [-0.0020, -0.0192, -0.1336, ..., -0.1177, 0.0238, 0.1214], + [ 0.0599, -0.0982, -0.0839, ..., -0.1125, -0.0632, -0.0822], + [-0.1374, 0.0179, 0.1048, ..., 0.0811, -0.0574, -0.1344]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 2.4214e-08, 5.0012e-07, ..., 5.6811e-07, + 0.0000e+00, 8.0094e-08], + [ 1.8626e-09, 2.5146e-08, 1.3970e-08, ..., 2.4214e-08, + 0.0000e+00, 8.9407e-08], + [ 1.8626e-09, 8.3819e-09, 1.9744e-07, ..., 2.0955e-07, + 0.0000e+00, -1.5358e-06], + ..., + [ 9.3132e-10, 6.8918e-08, 5.9791e-07, ..., 9.4250e-07, + 0.0000e+00, 6.2585e-07], + [ 2.2352e-08, 3.8184e-08, 9.5274e-07, ..., 1.1232e-06, + 5.5879e-09, 2.6729e-07], + [ 3.7253e-09, -4.5821e-07, -3.7272e-06, ..., -5.1111e-06, + 9.3132e-10, 5.4948e-08]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0241, -0.0234, -0.0380, -0.0043, 0.0056, 0.0399, 0.0175, -0.0140, + 0.0018, -0.0066], device='cuda:0'), grad: tensor([ 1.5069e-06, 1.1362e-07, -1.3746e-06, -1.8841e-06, 2.4512e-06, + 4.2170e-06, 1.4901e-07, 2.4363e-06, 3.1516e-06, -1.0766e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 214.19, cls_loss 0.0028 cls_loss_mapping 0.0046 cls_loss_causal 0.5263 re_mapping 0.0057 re_causal 0.0168 /// teacc 98.90 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0206, -0.0564, -0.0411, ..., -0.1617, -0.0512, -0.1136], + [-0.0297, -0.0554, -0.0363, ..., 0.0211, 0.0065, -0.1092], + [ 0.0364, -0.1043, -0.1053, ..., 0.0171, -0.1342, 0.0589], + ..., + [-0.0021, -0.0191, -0.1340, ..., -0.1178, 0.0241, 0.1215], + [ 0.0602, -0.0986, -0.0837, ..., -0.1127, -0.0638, -0.0834], + [-0.1384, 0.0178, 0.1059, ..., 0.0832, -0.0575, -0.1347]], + device='cuda:0'), grad: tensor([[-2.9802e-08, 3.2596e-08, 1.5087e-07, ..., 1.7975e-07, + 0.0000e+00, 3.7253e-09], + [ 1.7975e-07, 5.6997e-07, 1.9763e-06, ..., 2.9616e-06, + 0.0000e+00, 2.2352e-08], + [ 1.3970e-08, 9.3132e-09, 2.4214e-08, ..., 3.5390e-08, + 0.0000e+00, -2.0489e-08], + ..., + [ 1.0245e-08, 8.7544e-08, 4.7125e-07, ..., 8.8289e-07, + 0.0000e+00, 1.9558e-08], + [-7.8231e-08, 1.0245e-07, 3.3993e-07, ..., 4.8336e-07, + 0.0000e+00, 1.9558e-08], + [ 1.3225e-07, -9.9931e-07, -3.6955e-06, ..., -5.2005e-06, + 0.0000e+00, 1.3970e-08]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0214, -0.0235, -0.0374, -0.0040, 0.0040, 0.0397, 0.0168, -0.0135, + 0.0018, -0.0073], device='cuda:0'), grad: tensor([ 1.3132e-07, 5.3719e-06, 1.1828e-07, 3.9767e-07, -8.3540e-07, + 1.3504e-07, -5.0943e-07, 2.2091e-06, -6.0536e-07, -6.4000e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 214.01, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.4932 re_mapping 0.0058 re_causal 0.0173 /// teacc 99.04 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0202, -0.0569, -0.0412, ..., -0.1622, -0.0513, -0.1144], + [-0.0301, -0.0556, -0.0363, ..., 0.0212, 0.0064, -0.1100], + [ 0.0336, -0.1057, -0.1056, ..., 0.0177, -0.1342, 0.0587], + ..., + [-0.0019, -0.0188, -0.1342, ..., -0.1179, 0.0241, 0.1223], + [ 0.0630, -0.0978, -0.0836, ..., -0.1124, -0.0640, -0.0842], + [-0.1390, 0.0177, 0.1059, ..., 0.0832, -0.0576, -0.1348]], + device='cuda:0'), grad: tensor([[-1.5367e-07, 1.0245e-08, 2.8871e-08, ..., 1.6857e-07, + 0.0000e+00, 3.0734e-08], + [ 1.3970e-08, 5.3085e-08, 2.7269e-05, ..., 1.7777e-05, + 0.0000e+00, 1.3970e-07], + [ 5.5879e-08, 5.9605e-08, 4.3400e-07, ..., 2.7474e-07, + 0.0000e+00, 6.1467e-08], + ..., + [ 5.5879e-09, -2.2259e-07, 3.1814e-06, ..., 2.1104e-06, + 0.0000e+00, -6.9290e-07], + [-1.0617e-07, 9.3132e-08, 1.4286e-06, ..., 1.0626e-06, + 0.0000e+00, 1.9744e-07], + [ 1.3597e-07, -6.7987e-08, 1.7993e-06, ..., 1.1083e-06, + 0.0000e+00, 8.8476e-08]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0214, -0.0237, -0.0386, -0.0038, 0.0040, 0.0395, 0.0172, -0.0131, + 0.0028, -0.0075], device='cuda:0'), grad: tensor([-5.3924e-07, 5.3793e-05, 1.1120e-06, 3.6601e-07, -7.0333e-05, + -6.3609e-07, 3.8408e-06, 5.1185e-06, 2.7735e-06, 4.3809e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 214.24, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5381 re_mapping 0.0056 re_causal 0.0180 /// teacc 98.99 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0201, -0.0572, -0.0414, ..., -0.1629, -0.0514, -0.1152], + [-0.0309, -0.0557, -0.0364, ..., 0.0212, 0.0066, -0.1102], + [ 0.0335, -0.1057, -0.1056, ..., 0.0179, -0.1344, 0.0592], + ..., + [-0.0022, -0.0190, -0.1348, ..., -0.1180, 0.0240, 0.1223], + [ 0.0634, -0.0980, -0.0839, ..., -0.1128, -0.0641, -0.0849], + [-0.1393, 0.0178, 0.1058, ..., 0.0829, -0.0576, -0.1349]], + device='cuda:0'), grad: tensor([[ 6.0350e-07, 4.9919e-07, 1.8664e-06, ..., 3.1665e-08, + 0.0000e+00, 2.5798e-06], + [ 1.8999e-07, 1.9930e-07, -6.0536e-08, ..., -3.3714e-07, + 9.3132e-10, 4.0047e-07], + [ 1.0524e-07, -6.1929e-05, -1.1033e-04, ..., 1.0524e-07, + 9.3132e-10, -3.8481e-04], + ..., + [ 1.7695e-08, 8.5235e-06, 1.2945e-07, ..., 1.3225e-07, + 1.3504e-07, 5.2005e-06], + [ 8.0559e-07, 6.4261e-07, 4.8839e-06, ..., 5.7742e-08, + 9.3132e-10, 1.7416e-06], + [ 2.2631e-07, 1.1548e-07, -3.4459e-08, ..., -5.0012e-07, + 9.3132e-10, 7.0594e-07]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0216, -0.0237, -0.0385, -0.0037, 0.0043, 0.0395, 0.0174, -0.0132, + 0.0028, -0.0077], device='cuda:0'), grad: tensor([ 8.0615e-06, 7.9349e-07, -7.3719e-04, -9.9018e-06, 7.1144e-04, + 3.4366e-06, -1.1194e-06, 1.1683e-05, 1.1094e-05, 1.5264e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 214.20, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.5071 re_mapping 0.0054 re_causal 0.0163 /// teacc 98.94 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0201, -0.0588, -0.0415, ..., -0.1648, -0.0536, -0.1158], + [-0.0316, -0.0558, -0.0365, ..., 0.0213, 0.0093, -0.1105], + [ 0.0339, -0.1051, -0.1049, ..., 0.0188, -0.1376, 0.0606], + ..., + [-0.0026, -0.0191, -0.1351, ..., -0.1182, 0.0233, 0.1222], + [ 0.0635, -0.0985, -0.0838, ..., -0.1130, -0.0651, -0.0859], + [-0.1398, 0.0179, 0.1058, ..., 0.0830, -0.0577, -0.1352]], + device='cuda:0'), grad: tensor([[-1.1977e-06, 4.2841e-08, 1.4156e-07, ..., 1.0431e-07, + 0.0000e+00, 7.2643e-08], + [ 1.0990e-07, 1.5832e-07, 1.0710e-07, ..., 1.0151e-07, + 3.7253e-09, 1.8347e-07], + [ 5.9605e-08, 9.6858e-08, 4.7497e-08, ..., 2.4214e-08, + 9.3132e-09, 7.0781e-08], + ..., + [ 2.8871e-08, -9.9652e-07, 1.2014e-07, ..., 1.3690e-07, + -1.9558e-08, -1.4938e-06], + [ 3.9116e-07, 1.4622e-07, 2.1309e-06, ..., 4.6939e-06, + 2.7940e-09, 1.7509e-07], + [ 3.7625e-07, 5.2527e-07, -5.5693e-07, ..., -3.6415e-07, + -9.3132e-10, 1.1502e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0218, -0.0237, -0.0374, -0.0036, 0.0041, 0.0393, 0.0184, -0.0134, + 0.0028, -0.0078], device='cuda:0'), grad: tensor([-4.1425e-06, 1.1427e-06, 5.1316e-07, 1.0123e-06, 8.0746e-07, + -4.8503e-06, -9.3877e-07, -6.0946e-06, 5.7742e-06, 6.7577e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 214.20, cls_loss 0.0017 cls_loss_mapping 0.0037 cls_loss_causal 0.5075 re_mapping 0.0056 re_causal 0.0166 /// teacc 98.97 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0207, -0.0608, -0.0422, ..., -0.1662, -0.0542, -0.1164], + [-0.0321, -0.0560, -0.0360, ..., 0.0216, 0.0096, -0.1108], + [ 0.0337, -0.1054, -0.1061, ..., 0.0198, -0.1381, 0.0610], + ..., + [-0.0028, -0.0190, -0.1354, ..., -0.1184, 0.0232, 0.1222], + [ 0.0642, -0.0996, -0.0845, ..., -0.1133, -0.0654, -0.0869], + [-0.1402, 0.0181, 0.1061, ..., 0.0830, -0.0578, -0.1352]], + device='cuda:0'), grad: tensor([[ 1.1176e-07, 1.5832e-08, 1.4715e-07, ..., 2.4494e-07, + 0.0000e+00, 3.6322e-08], + [ 4.7032e-08, 1.1548e-07, -6.7987e-08, ..., -9.6858e-08, + 0.0000e+00, 2.0815e-07], + [-6.2445e-07, 3.9116e-08, 6.3330e-08, ..., 5.4017e-08, + 0.0000e+00, -9.5833e-07], + ..., + [ 5.0291e-08, 4.0978e-08, 1.0990e-07, ..., 1.6671e-07, + -0.0000e+00, 7.4506e-08], + [ 4.6333e-07, 2.7940e-08, 1.8906e-07, ..., 3.2596e-07, + 4.6566e-10, 1.0189e-06], + [ 8.3819e-08, -3.2596e-09, 7.7765e-08, ..., 1.7043e-07, + 4.6566e-10, 3.8184e-08]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0220, -0.0236, -0.0374, -0.0031, 0.0042, 0.0392, 0.0183, -0.0135, + 0.0027, -0.0076], device='cuda:0'), grad: tensor([ 7.5810e-07, 3.7486e-07, -1.8571e-06, -1.3001e-05, -4.1584e-07, + 1.2495e-05, -2.0722e-07, 7.0920e-07, 1.9744e-07, 9.3039e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 214.02, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.5210 re_mapping 0.0054 re_causal 0.0166 /// teacc 99.01 lr 0.00010000 +Epoch 180, weight, value: tensor([[ 0.0214, -0.0613, -0.0427, ..., -0.1662, -0.0550, -0.1169], + [-0.0353, -0.0555, -0.0351, ..., 0.0219, 0.0096, -0.1086], + [ 0.0336, -0.1058, -0.1067, ..., 0.0199, -0.1382, 0.0611], + ..., + [-0.0030, -0.0196, -0.1374, ..., -0.1190, 0.0232, 0.1207], + [ 0.0648, -0.0998, -0.0846, ..., -0.1134, -0.0656, -0.0872], + [-0.1404, 0.0184, 0.1066, ..., 0.0833, -0.0577, -0.1353]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 6.8452e-08, 9.6392e-08, ..., 1.6810e-07, + 0.0000e+00, 2.2165e-07], + [ 7.2783e-07, -1.5479e-06, -6.0163e-06, ..., 1.0552e-06, + 0.0000e+00, -1.1154e-05], + [ 7.2177e-08, 1.4808e-07, 4.5635e-08, ..., 8.9873e-08, + 0.0000e+00, 4.2142e-07], + ..., + [ 1.4435e-08, 1.4026e-06, 5.1633e-06, ..., 6.0070e-08, + 0.0000e+00, 9.4399e-06], + [-5.9940e-06, 2.0023e-07, -9.8534e-07, ..., -8.7619e-06, + 0.0000e+00, 1.0459e-06], + [ 3.9116e-07, 2.9337e-08, 4.3306e-08, ..., 4.7125e-07, + 0.0000e+00, 2.0163e-07]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0224, -0.0227, -0.0375, -0.0028, 0.0040, 0.0390, 0.0189, -0.0152, + 0.0029, -0.0071], device='cuda:0'), grad: tensor([ 1.6801e-06, -1.0002e-04, 1.6205e-06, 2.8089e-06, 6.8918e-08, + 2.6509e-05, 6.7940e-07, 8.6129e-05, -2.2516e-05, 2.8759e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 214.09, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.5163 re_mapping 0.0054 re_causal 0.0169 /// teacc 99.01 lr 0.00010000 +Epoch 181, weight, value: tensor([[ 0.0214, -0.0616, -0.0429, ..., -0.1674, -0.0553, -0.1177], + [-0.0351, -0.0555, -0.0346, ..., 0.0224, 0.0098, -0.1088], + [ 0.0338, -0.1061, -0.1072, ..., 0.0196, -0.1385, 0.0608], + ..., + [-0.0032, -0.0197, -0.1382, ..., -0.1192, 0.0230, 0.1212], + [ 0.0647, -0.1000, -0.0850, ..., -0.1142, -0.0657, -0.0875], + [-0.1404, 0.0187, 0.1067, ..., 0.0834, -0.0578, -0.1354]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 1.5553e-07, 6.2352e-07, ..., 5.5367e-07, + 0.0000e+00, 8.3819e-09], + [ 3.8184e-08, 2.1495e-06, 9.3207e-06, ..., 8.9854e-06, + 0.0000e+00, 1.3970e-08], + [ 2.8405e-08, 8.4285e-08, 3.5996e-07, ..., 3.5483e-07, + 0.0000e+00, -2.2817e-08], + ..., + [ 1.6298e-08, 2.9262e-06, 1.2137e-05, ..., 1.1742e-05, + 0.0000e+00, 1.2759e-07], + [-4.3586e-07, 8.3819e-08, 3.3341e-07, ..., 3.2736e-07, + 0.0000e+00, 9.7789e-09], + [ 1.1222e-07, -7.5661e-06, -3.5673e-05, ..., -3.7044e-05, + 0.0000e+00, 3.7719e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0225, -0.0224, -0.0378, -0.0027, 0.0041, 0.0391, 0.0180, -0.0153, + 0.0027, -0.0071], device='cuda:0'), grad: tensor([ 1.1157e-06, 1.5840e-05, 7.3249e-07, -7.3109e-07, 1.7554e-05, + 1.4398e-06, 2.4680e-08, 2.1115e-05, -1.6689e-06, -5.5522e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 213.80, cls_loss 0.0022 cls_loss_mapping 0.0029 cls_loss_causal 0.5353 re_mapping 0.0055 re_causal 0.0160 /// teacc 98.97 lr 0.00010000 +Epoch 182, weight, value: tensor([[ 0.0239, -0.0622, -0.0431, ..., -0.1664, -0.0555, -0.1196], + [-0.0358, -0.0556, -0.0336, ..., 0.0229, 0.0098, -0.1092], + [ 0.0338, -0.1064, -0.1079, ..., 0.0213, -0.1388, 0.0612], + ..., + [-0.0035, -0.0196, -0.1389, ..., -0.1195, 0.0239, 0.1217], + [ 0.0651, -0.1008, -0.0850, ..., -0.1147, -0.0658, -0.0890], + [-0.1409, 0.0167, 0.1051, ..., 0.0813, -0.0580, -0.1359]], + device='cuda:0'), grad: tensor([[ 2.5099e-07, 6.9849e-09, 2.4680e-07, ..., 8.8941e-08, + 0.0000e+00, 6.6124e-08], + [ 8.0559e-08, 4.1910e-08, -2.2398e-07, ..., -2.1979e-07, + -3.7253e-09, 1.1735e-07], + [-1.3411e-07, 9.9186e-08, 1.2387e-07, ..., -1.5739e-07, + 0.0000e+00, 6.2864e-08], + ..., + [ 4.2841e-08, -3.1106e-07, 1.5646e-07, ..., 1.6252e-07, + 1.3970e-09, -7.9582e-07], + [-3.8594e-06, 6.6590e-08, -4.9314e-07, ..., 2.8312e-07, + 1.3970e-09, 1.2852e-07], + [ 2.9337e-07, -4.1910e-08, -9.9093e-07, ..., -9.4902e-07, + 4.6566e-10, 2.8498e-07]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0218, -0.0223, -0.0376, -0.0017, 0.0059, 0.0390, 0.0162, -0.0153, + 0.0026, -0.0088], device='cuda:0'), grad: tensor([ 9.5088e-07, -2.5192e-07, 1.5553e-07, 2.1607e-07, 1.7453e-06, + 9.1968e-07, 5.1185e-06, -1.3374e-06, -7.5102e-06, -5.3085e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 213.66, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5292 re_mapping 0.0055 re_causal 0.0170 /// teacc 98.96 lr 0.00010000 +Epoch 183, weight, value: tensor([[ 0.0206, -0.0627, -0.0432, ..., -0.1669, -0.0561, -0.1207], + [-0.0361, -0.0557, -0.0335, ..., 0.0230, 0.0102, -0.1097], + [ 0.0338, -0.1066, -0.1085, ..., 0.0213, -0.1393, 0.0610], + ..., + [-0.0037, -0.0194, -0.1392, ..., -0.1195, 0.0240, 0.1225], + [ 0.0653, -0.1011, -0.0852, ..., -0.1150, -0.0659, -0.0895], + [-0.1413, 0.0166, 0.1051, ..., 0.0814, -0.0581, -0.1361]], + device='cuda:0'), grad: tensor([[ 2.8405e-07, 8.9407e-08, 8.5076e-07, ..., 6.7707e-07, + 0.0000e+00, 1.4668e-07], + [ 1.9064e-06, 1.0245e-07, 4.0196e-06, ..., 2.6785e-06, + 0.0000e+00, 2.1094e-07], + [ 9.3039e-07, 1.2480e-07, 2.5947e-06, ..., 1.8440e-06, + 0.0000e+00, 1.9046e-07], + ..., + [ 8.3819e-09, 1.1316e-07, 4.5542e-07, ..., 5.7928e-07, + 0.0000e+00, -2.0768e-07], + [-4.5449e-07, 2.3888e-07, 8.2189e-07, ..., 1.0310e-06, + 0.0000e+00, 2.3888e-07], + [ 6.7521e-08, 3.3388e-07, 1.4538e-06, ..., 1.4435e-06, + 0.0000e+00, 5.1875e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0247, -0.0223, -0.0380, -0.0017, 0.0057, 0.0390, 0.0193, -0.0150, + 0.0025, -0.0088], device='cuda:0'), grad: tensor([ 2.3469e-06, 1.0073e-05, 6.1281e-06, -2.7893e-07, 5.1916e-05, + 3.4720e-06, -8.0526e-05, 1.0245e-06, 7.4832e-07, 5.2378e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 213.95, cls_loss 0.0020 cls_loss_mapping 0.0035 cls_loss_causal 0.5297 re_mapping 0.0055 re_causal 0.0167 /// teacc 99.06 lr 0.00010000 +Epoch 184, weight, value: tensor([[ 0.0208, -0.0627, -0.0435, ..., -0.1671, -0.0590, -0.1214], + [-0.0370, -0.0558, -0.0333, ..., 0.0230, 0.0110, -0.1099], + [ 0.0334, -0.1070, -0.1098, ..., 0.0224, -0.1410, 0.0616], + ..., + [-0.0039, -0.0195, -0.1399, ..., -0.1197, 0.0238, 0.1228], + [ 0.0653, -0.1013, -0.0856, ..., -0.1159, -0.0665, -0.0898], + [-0.1412, 0.0167, 0.1054, ..., 0.0816, -0.0582, -0.1361]], + device='cuda:0'), grad: tensor([[-2.5425e-03, 4.7451e-07, -5.4926e-05, ..., -1.8752e-04, + 0.0000e+00, 3.0156e-06], + [ 8.2493e-04, 2.7940e-08, 1.7956e-05, ..., 6.0856e-05, + 0.0000e+00, 3.2177e-07], + [ 8.9332e-06, 3.2596e-08, 2.2072e-07, ..., 7.0641e-07, + 0.0000e+00, -7.9395e-07], + ..., + [ 2.0638e-06, -6.3377e-07, 8.0094e-08, ..., 2.3795e-07, + 0.0000e+00, -9.1493e-06], + [ 1.4663e-05, 5.4017e-08, 3.6927e-07, ..., 1.2880e-06, + 0.0000e+00, 8.9454e-07], + [ 1.4529e-06, 3.1851e-07, 9.1735e-08, ..., 2.2491e-07, + 0.0000e+00, 4.4294e-06]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0245, -0.0223, -0.0378, -0.0022, 0.0056, 0.0391, 0.0193, -0.0151, + 0.0023, -0.0085], device='cuda:0'), grad: tensor([-8.5220e-03, 2.7676e-03, 2.9355e-05, 1.9416e-05, 7.1600e-06, + 5.8264e-06, 5.6381e-03, -8.6874e-06, 5.0217e-05, 1.2875e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 214.05, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.4989 re_mapping 0.0053 re_causal 0.0153 /// teacc 99.05 lr 0.00010000 +Epoch 185, weight, value: tensor([[ 0.0215, -0.0634, -0.0435, ..., -0.1661, -0.0618, -0.1250], + [-0.0397, -0.0552, -0.0331, ..., 0.0233, 0.0146, -0.1088], + [ 0.0332, -0.1070, -0.1103, ..., 0.0224, -0.1471, 0.0647], + ..., + [-0.0057, -0.0207, -0.1406, ..., -0.1199, 0.0211, 0.1208], + [ 0.0676, -0.1012, -0.0856, ..., -0.1165, -0.0674, -0.0901], + [-0.1422, 0.0169, 0.1056, ..., 0.0818, -0.0600, -0.1365]], + device='cuda:0'), grad: tensor([[-8.8587e-06, -1.1055e-06, 5.8534e-07, ..., 4.6473e-07, + 6.9849e-09, -1.6037e-06], + [ 1.1446e-06, 6.3516e-07, 1.5981e-06, ..., 1.3383e-06, + 3.7253e-09, 5.3598e-07], + [ 1.7518e-06, 7.5204e-07, 6.4727e-07, ..., 3.6601e-07, + 1.7928e-07, 1.2992e-06], + ..., + [ 1.0645e-06, 7.1293e-07, 5.3458e-06, ..., 4.6417e-06, + -3.6182e-07, -1.9837e-06], + [ 1.7779e-06, 3.1181e-06, 1.2159e-05, ..., 1.0461e-05, + 2.4214e-08, 9.1363e-07], + [ 1.8673e-06, -3.8177e-05, -1.8108e-04, ..., -1.5628e-04, + 2.1420e-08, 1.0114e-06]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0236, -0.0220, -0.0351, -0.0021, 0.0054, 0.0391, 0.0183, -0.0172, + 0.0029, -0.0085], device='cuda:0'), grad: tensor([-2.4542e-05, 6.3367e-06, 6.9700e-06, 4.2701e-07, 3.3832e-04, + 4.8578e-06, 5.6066e-07, 1.0461e-05, 3.0786e-05, -3.7384e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 214.46, cls_loss 0.0018 cls_loss_mapping 0.0025 cls_loss_causal 0.4940 re_mapping 0.0058 re_causal 0.0159 /// teacc 98.87 lr 0.00010000 +Epoch 186, weight, value: tensor([[ 0.0215, -0.0636, -0.0436, ..., -0.1663, -0.0639, -0.1257], + [-0.0399, -0.0553, -0.0332, ..., 0.0234, 0.0158, -0.1087], + [ 0.0330, -0.1074, -0.1112, ..., 0.0220, -0.1501, 0.0650], + ..., + [-0.0058, -0.0207, -0.1410, ..., -0.1200, 0.0212, 0.1209], + [ 0.0676, -0.1019, -0.0860, ..., -0.1175, -0.0684, -0.0909], + [-0.1430, 0.0170, 0.1055, ..., 0.0816, -0.0598, -0.1369]], + device='cuda:0'), grad: tensor([[ 3.2689e-07, 1.6764e-08, 8.3167e-07, ..., 6.0070e-08, + 4.6566e-10, 3.4459e-08], + [ 3.0315e-07, 7.9162e-08, 6.5006e-07, ..., -1.0245e-08, + -4.6566e-10, 1.6578e-07], + [ 5.1549e-07, 9.3132e-08, 1.1474e-06, ..., 7.9162e-08, + 0.0000e+00, 1.4855e-07], + ..., + [ 2.5611e-08, -1.1502e-07, 1.6950e-07, ..., 1.3737e-07, + 2.3283e-09, -3.9116e-07], + [ 1.1269e-07, 4.6566e-08, 4.5262e-07, ..., 1.4901e-07, + 8.8476e-09, 7.3109e-08], + [ 8.9873e-08, -5.2154e-08, -5.2154e-08, ..., -7.3109e-08, + 4.4703e-08, 7.6834e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0236, -0.0219, -0.0350, -0.0020, 0.0057, 0.0391, 0.0184, -0.0174, + 0.0025, -0.0087], device='cuda:0'), grad: tensor([ 1.8198e-06, 1.7360e-06, 2.9542e-06, -1.7267e-06, 2.4840e-05, + 5.0701e-06, -3.5882e-05, -2.7614e-07, 1.0012e-06, 4.8149e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 213.95, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5162 re_mapping 0.0055 re_causal 0.0170 /// teacc 98.97 lr 0.00010000 +Epoch 187, weight, value: tensor([[ 0.0215, -0.0666, -0.0442, ..., -0.1681, -0.0644, -0.1261], + [-0.0400, -0.0554, -0.0333, ..., 0.0233, 0.0158, -0.1089], + [ 0.0332, -0.1076, -0.1120, ..., 0.0221, -0.1493, 0.0654], + ..., + [-0.0060, -0.0207, -0.1414, ..., -0.1202, 0.0219, 0.1210], + [ 0.0675, -0.1024, -0.0863, ..., -0.1181, -0.0690, -0.0913], + [-0.1430, 0.0172, 0.1056, ..., 0.0817, -0.0598, -0.1370]], + device='cuda:0'), grad: tensor([[ 3.0315e-07, 2.7660e-07, 3.6089e-07, ..., 1.9418e-07, + 1.3970e-09, 4.7497e-08], + [ 1.2573e-08, 5.4017e-08, 1.9092e-08, ..., -1.3504e-08, + 0.0000e+00, 8.2422e-08], + [ 1.2573e-08, 1.1316e-07, 4.0513e-08, ..., 4.1910e-09, + 4.6566e-10, 1.1362e-07], + ..., + [ 2.7940e-09, -3.8184e-08, 1.5367e-08, ..., 2.3749e-08, + 0.0000e+00, -7.4040e-08], + [ 2.8405e-08, 3.0734e-08, 1.9092e-08, ..., 7.1712e-08, + 4.6566e-10, 7.6834e-08], + [ 4.8429e-08, 3.2596e-09, 3.3993e-08, ..., 4.6566e-08, + 0.0000e+00, 1.8161e-08]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0237, -0.0219, -0.0351, -0.0020, 0.0059, 0.0390, 0.0185, -0.0174, + 0.0026, -0.0087], device='cuda:0'), grad: tensor([ 1.2722e-06, 1.7649e-07, 2.1420e-07, -3.8231e-07, 1.5134e-07, + 7.2597e-07, -2.6841e-06, -8.2422e-08, 2.7940e-07, 3.2829e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 213.97, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.4896 re_mapping 0.0057 re_causal 0.0159 /// teacc 98.95 lr 0.00010000 +Epoch 188, weight, value: tensor([[ 0.0215, -0.0666, -0.0444, ..., -0.1687, -0.0687, -0.1264], + [-0.0401, -0.0555, -0.0332, ..., 0.0235, 0.0175, -0.1096], + [ 0.0331, -0.1083, -0.1135, ..., 0.0223, -0.1522, 0.0652], + ..., + [-0.0061, -0.0205, -0.1429, ..., -0.1206, 0.0215, 0.1216], + [ 0.0674, -0.1028, -0.0868, ..., -0.1189, -0.0697, -0.0916], + [-0.1432, 0.0172, 0.1056, ..., 0.0816, -0.0592, -0.1371]], + device='cuda:0'), grad: tensor([[-3.2000e-06, -1.1213e-06, 2.3283e-08, ..., 1.9092e-08, + 1.5832e-08, 1.6298e-07], + [ 1.3504e-07, 3.0687e-07, 6.3330e-08, ..., 3.3528e-08, + 3.8184e-08, 4.7544e-07], + [ 3.0827e-07, 2.3842e-07, 2.5611e-08, ..., 1.5832e-08, + 1.4435e-08, -4.9500e-07], + ..., + [ 6.2864e-08, -7.4692e-07, 1.0384e-07, ..., 6.4727e-08, + -3.9116e-08, -1.0543e-06], + [ 8.9360e-07, -9.8627e-07, 1.9092e-08, ..., 2.2445e-07, + -2.2771e-07, -1.5153e-06], + [ 4.4098e-07, 3.5670e-07, 2.0023e-07, ..., 2.7241e-07, + 9.0338e-08, 3.0547e-07]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0238, -0.0219, -0.0355, -0.0020, 0.0066, 0.0389, 0.0186, -0.0176, + 0.0023, -0.0086], device='cuda:0'), grad: tensor([-9.8050e-06, 1.6224e-06, 6.0117e-07, 5.8636e-06, -1.3504e-06, + 5.1744e-06, 1.3579e-06, -1.6373e-06, -4.6194e-06, 2.8014e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 214.01, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5346 re_mapping 0.0055 re_causal 0.0159 /// teacc 99.09 lr 0.00010000 +Epoch 189, weight, value: tensor([[ 0.0215, -0.0678, -0.0446, ..., -0.1691, -0.0733, -0.1272], + [-0.0400, -0.0556, -0.0328, ..., 0.0237, 0.0180, -0.1101], + [ 0.0330, -0.1095, -0.1143, ..., 0.0233, -0.1494, 0.0660], + ..., + [-0.0063, -0.0205, -0.1438, ..., -0.1210, 0.0207, 0.1219], + [ 0.0673, -0.1043, -0.0871, ..., -0.1197, -0.0705, -0.0934], + [-0.1442, 0.0172, 0.1053, ..., 0.0813, -0.0608, -0.1374]], + device='cuda:0'), grad: tensor([[-5.1670e-06, -3.9265e-06, -1.5236e-06, ..., 9.8255e-08, + 3.2596e-09, 8.3353e-08], + [ 2.9430e-07, 2.4959e-07, 2.9057e-07, ..., -7.4040e-08, + 1.5367e-08, 1.9046e-07], + [ 1.7136e-07, 1.3644e-07, 2.4587e-07, ..., 3.6322e-08, + 1.3970e-08, -4.8196e-07], + ..., + [ 6.4727e-08, 1.5879e-07, 8.7358e-07, ..., 6.1467e-07, + -3.6787e-08, -3.6322e-08], + [ 2.3283e-06, 1.3905e-06, 1.9409e-06, ..., 4.5495e-07, + 3.2596e-09, 1.0291e-07], + [ 2.5071e-06, -6.4634e-07, -9.3505e-06, ..., -6.7428e-06, + 6.5193e-09, 5.4017e-08]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0238, -0.0219, -0.0350, -0.0027, 0.0071, 0.0393, 0.0186, -0.0175, + 0.0018, -0.0091], device='cuda:0'), grad: tensor([-1.8805e-05, 1.3877e-06, -2.5285e-07, 1.5888e-06, 1.6883e-05, + -9.8627e-07, -1.1139e-06, 1.8561e-06, 8.9407e-06, -9.4846e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 214.13, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.5275 re_mapping 0.0054 re_causal 0.0163 /// teacc 99.06 lr 0.00010000 +Epoch 190, weight, value: tensor([[ 0.0215, -0.0679, -0.0446, ..., -0.1695, -0.0741, -0.1280], + [-0.0400, -0.0557, -0.0324, ..., 0.0240, 0.0180, -0.1103], + [ 0.0330, -0.1101, -0.1151, ..., 0.0231, -0.1493, 0.0664], + ..., + [-0.0065, -0.0208, -0.1444, ..., -0.1212, 0.0206, 0.1220], + [ 0.0674, -0.1044, -0.0872, ..., -0.1202, -0.0712, -0.0939], + [-0.1448, 0.0178, 0.1057, ..., 0.0818, -0.0612, -0.1375]], + device='cuda:0'), grad: tensor([[ 2.0582e-07, 1.5832e-07, 5.4110e-07, ..., 6.0070e-07, + 1.0245e-08, 9.5926e-08], + [ 8.0559e-08, -4.8382e-07, -1.9148e-06, ..., -1.6615e-06, + 1.8813e-07, 9.3225e-07], + [ 7.1246e-08, 5.6997e-07, 7.4971e-08, ..., 1.4855e-07, + 3.1665e-08, 3.0492e-06], + ..., + [-3.0268e-08, -3.8976e-07, 1.3029e-06, ..., 1.2880e-06, + -5.1502e-07, -4.6268e-06], + [ 5.9884e-07, 4.3726e-07, 8.3121e-07, ..., 4.2468e-06, + 1.7695e-08, 4.6380e-07], + [ 6.5193e-08, 2.9849e-07, -2.0899e-06, ..., -1.4883e-06, + 9.1735e-08, 1.4249e-06]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0239, -0.0218, -0.0347, -0.0024, 0.0067, 0.0393, 0.0185, -0.0177, + 0.0017, -0.0088], device='cuda:0'), grad: tensor([ 1.9614e-06, -3.4552e-06, 5.9716e-06, 1.8075e-05, 2.0973e-06, + -2.0221e-05, -8.5309e-06, -4.8913e-06, 9.4548e-06, -4.9686e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 214.04, cls_loss 0.0020 cls_loss_mapping 0.0040 cls_loss_causal 0.5079 re_mapping 0.0055 re_causal 0.0154 /// teacc 98.84 lr 0.00010000 +Epoch 191, weight, value: tensor([[ 0.0215, -0.0677, -0.0447, ..., -0.1703, -0.0773, -0.1283], + [-0.0412, -0.0547, -0.0307, ..., 0.0252, 0.0180, -0.1104], + [ 0.0323, -0.1106, -0.1168, ..., 0.0225, -0.1496, 0.0663], + ..., + [-0.0066, -0.0210, -0.1467, ..., -0.1226, 0.0214, 0.1224], + [ 0.0665, -0.1047, -0.0890, ..., -0.1212, -0.0714, -0.0943], + [-0.1459, 0.0183, 0.1062, ..., 0.0823, -0.0616, -0.1380]], + device='cuda:0'), grad: tensor([[ 6.4354e-07, 6.5193e-09, 5.4948e-07, ..., 1.0245e-08, + 3.8184e-08, 4.5449e-07], + [ 9.9652e-08, 8.6613e-08, -6.7055e-08, ..., -1.4342e-07, + 2.5891e-07, 1.3718e-06], + [ 8.9221e-07, 2.6077e-08, 7.3854e-07, ..., 9.3132e-10, + 7.7579e-07, 1.0887e-06], + ..., + [ 7.2643e-08, -8.0280e-07, 1.1548e-07, ..., 8.2888e-08, + -1.1921e-06, -3.5297e-06], + [-2.9355e-06, 5.4017e-08, -2.3562e-06, ..., 3.0734e-08, + 4.5635e-08, -1.0217e-06], + [ 1.3225e-07, 6.0070e-07, 1.8626e-08, ..., -9.2201e-08, + 3.2596e-08, 9.0525e-07]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0238, -0.0218, -0.0355, -0.0025, 0.0060, 0.0391, 0.0196, -0.0179, + 0.0015, -0.0088], device='cuda:0'), grad: tensor([ 4.7646e-06, 3.3751e-06, 7.7710e-06, 3.6135e-06, 3.5837e-06, + 4.4424e-07, 6.3237e-07, -8.6650e-06, -1.9357e-05, 3.8128e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 213.74, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.5065 re_mapping 0.0055 re_causal 0.0162 /// teacc 98.98 lr 0.00010000 +Epoch 192, weight, value: tensor([[ 0.0215, -0.0678, -0.0448, ..., -0.1708, -0.0784, -0.1296], + [-0.0415, -0.0541, -0.0298, ..., 0.0261, 0.0179, -0.1104], + [ 0.0323, -0.1108, -0.1174, ..., 0.0229, -0.1497, 0.0667], + ..., + [-0.0067, -0.0214, -0.1480, ..., -0.1234, 0.0215, 0.1224], + [ 0.0673, -0.1048, -0.0891, ..., -0.1215, -0.0713, -0.0951], + [-0.1465, 0.0184, 0.1059, ..., 0.0817, -0.0638, -0.1382]], + device='cuda:0'), grad: tensor([[-2.1607e-05, -3.6601e-06, 2.4401e-07, ..., 1.7136e-07, + 2.8685e-07, 5.1409e-07], + [ 5.1502e-07, 6.9104e-07, 1.1744e-06, ..., 7.8604e-07, + 6.7521e-07, 7.5251e-07], + [ 2.3078e-06, 4.5542e-07, 2.3656e-07, ..., 1.9372e-07, + 2.0489e-08, -2.5854e-06], + ..., + [-9.5740e-07, -3.5241e-06, 8.7544e-08, ..., -5.7835e-07, + -4.0755e-06, -3.6396e-06], + [ 8.4285e-07, 3.1292e-07, 2.0117e-07, ..., 2.7474e-07, + 1.8440e-07, 4.6380e-07], + [ 6.1654e-06, 3.2932e-06, 3.4180e-07, ..., 7.8324e-07, + 2.7213e-06, 2.4512e-06]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0238, -0.0215, -0.0355, -0.0036, 0.0065, 0.0390, 0.0197, -0.0183, + 0.0023, -0.0093], device='cuda:0'), grad: tensor([-5.6982e-05, 5.9493e-06, 2.6114e-06, 3.7402e-06, -6.3665e-06, + 2.6479e-05, 9.2015e-06, -1.3612e-05, 3.8520e-06, 2.5153e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 214.20, cls_loss 0.0016 cls_loss_mapping 0.0037 cls_loss_causal 0.4969 re_mapping 0.0061 re_causal 0.0162 /// teacc 98.98 lr 0.00010000 +Epoch 193, weight, value: tensor([[ 0.0213, -0.0679, -0.0450, ..., -0.1712, -0.0818, -0.1304], + [-0.0418, -0.0543, -0.0299, ..., 0.0262, 0.0175, -0.1106], + [ 0.0321, -0.1109, -0.1179, ..., 0.0228, -0.1495, 0.0680], + ..., + [-0.0067, -0.0210, -0.1482, ..., -0.1234, 0.0240, 0.1227], + [ 0.0673, -0.1052, -0.0893, ..., -0.1217, -0.0711, -0.0955], + [-0.1479, 0.0184, 0.1059, ..., 0.0816, -0.0669, -0.1391]], + device='cuda:0'), grad: tensor([[-4.0978e-07, 1.5832e-08, 4.7404e-07, ..., 1.6391e-07, + 2.1420e-08, 4.7497e-08], + [ 1.4901e-08, 6.7055e-08, -6.4075e-07, ..., -1.1250e-06, + -4.1630e-07, 1.8906e-07], + [ 6.8918e-08, 3.5390e-08, 2.5053e-07, ..., 9.1270e-08, + 1.7695e-08, 7.7300e-08], + ..., + [ 5.5879e-09, -7.8231e-08, 4.2282e-07, ..., 6.5938e-07, + 2.3376e-07, -1.6019e-07], + [ 2.7940e-08, 4.7497e-08, 1.7509e-07, ..., 2.5146e-07, + 5.7742e-08, 1.1642e-07], + [ 4.0885e-07, 7.4506e-09, 1.8440e-07, ..., 1.4622e-07, + 5.6811e-08, 1.1921e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0241, -0.0215, -0.0350, -0.0040, 0.0064, 0.0390, 0.0203, -0.0181, + 0.0023, -0.0097], device='cuda:0'), grad: tensor([-7.7114e-07, -2.4568e-06, 9.2201e-07, -1.3066e-06, 2.5146e-07, + 1.9334e-06, -3.3975e-06, 1.6131e-06, 7.7393e-07, 2.4177e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 213.82, cls_loss 0.0020 cls_loss_mapping 0.0050 cls_loss_causal 0.5192 re_mapping 0.0054 re_causal 0.0162 /// teacc 98.95 lr 0.00010000 +Epoch 194, weight, value: tensor([[ 0.0213, -0.0681, -0.0453, ..., -0.1716, -0.0822, -0.1310], + [-0.0419, -0.0546, -0.0322, ..., 0.0262, 0.0160, -0.1154], + [ 0.0321, -0.1115, -0.1191, ..., 0.0226, -0.1492, 0.0682], + ..., + [-0.0067, -0.0208, -0.1459, ..., -0.1238, 0.0257, 0.1274], + [ 0.0671, -0.1059, -0.0897, ..., -0.1225, -0.0724, -0.0969], + [-0.1483, 0.0187, 0.1063, ..., 0.0819, -0.0670, -0.1393]], + device='cuda:0'), grad: tensor([[-9.3132e-09, 3.7253e-09, 1.0710e-07, ..., 4.0978e-08, + 9.3132e-09, 4.6566e-08], + [ 1.3970e-08, 1.3039e-08, -2.3544e-06, ..., -6.9477e-07, + 2.1979e-07, 7.3109e-07], + [ 1.1176e-08, 2.4214e-08, 1.5553e-07, ..., 6.1467e-08, + 5.4948e-08, 2.7847e-07], + ..., + [ 1.8626e-09, -2.6077e-08, 3.1386e-07, ..., 4.9733e-07, + -5.0850e-06, -1.5870e-05], + [ 8.3819e-09, 2.7940e-08, 5.2527e-07, ..., 1.7229e-07, + 1.8347e-07, 6.7800e-07], + [ 1.8626e-08, 9.3132e-10, 1.6950e-07, ..., -3.5390e-08, + 2.7940e-08, 1.3504e-07]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0241, -0.0251, -0.0350, -0.0042, 0.0062, 0.0389, 0.0204, -0.0142, + 0.0021, -0.0093], device='cuda:0'), grad: tensor([ 3.4086e-07, -5.4538e-06, 1.2424e-06, -9.1922e-07, 4.6015e-05, + 1.3243e-06, 5.4948e-07, -4.7475e-05, 3.3118e-06, 1.0179e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.02, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.4919 re_mapping 0.0053 re_causal 0.0160 /// teacc 98.95 lr 0.00010000 +Epoch 195, weight, value: tensor([[ 0.0213, -0.0684, -0.0456, ..., -0.1721, -0.0823, -0.1317], + [-0.0420, -0.0547, -0.0322, ..., 0.0263, 0.0161, -0.1155], + [ 0.0320, -0.1117, -0.1231, ..., 0.0202, -0.1493, 0.0680], + ..., + [-0.0069, -0.0207, -0.1460, ..., -0.1239, 0.0264, 0.1275], + [ 0.0681, -0.1062, -0.0902, ..., -0.1227, -0.0728, -0.0974], + [-0.1485, 0.0190, 0.1068, ..., 0.0822, -0.0676, -0.1396]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 8.3819e-09, 1.1362e-07, ..., 2.2352e-08, + 9.3132e-10, 1.5274e-07], + [ 5.5879e-09, 1.4622e-07, -4.3437e-06, ..., -5.0291e-08, + -3.7253e-09, 2.3711e-06], + [-2.0489e-08, 4.7963e-07, 1.8347e-07, ..., 3.8184e-08, + 9.3132e-10, 7.6368e-06], + ..., + [ 1.3039e-08, -8.0839e-07, 2.7604e-06, ..., 1.0151e-07, + -1.8626e-09, -1.3202e-05], + [-2.7008e-08, 8.4750e-08, 6.0815e-07, ..., 8.3819e-08, + 9.3132e-10, 1.5022e-06], + [ 1.9558e-08, 6.1467e-08, 2.5053e-07, ..., -6.5193e-09, + 9.3132e-10, 1.0561e-06]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0241, -0.0251, -0.0356, -0.0040, 0.0063, 0.0387, 0.0204, -0.0141, + 0.0022, -0.0091], device='cuda:0'), grad: tensor([ 6.0908e-07, -1.0274e-05, 1.4514e-05, 6.1933e-07, 1.4370e-06, + 4.9453e-07, 3.8277e-07, -1.5184e-05, 4.5262e-06, 2.8349e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 214.20, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4883 re_mapping 0.0053 re_causal 0.0153 /// teacc 99.06 lr 0.00010000 +Epoch 196, weight, value: tensor([[ 0.0213, -0.0687, -0.0458, ..., -0.1725, -0.0826, -0.1324], + [-0.0422, -0.0548, -0.0321, ..., 0.0264, 0.0170, -0.1155], + [ 0.0323, -0.1122, -0.1241, ..., 0.0210, -0.1500, 0.0682], + ..., + [-0.0071, -0.0223, -0.1470, ..., -0.1247, 0.0265, 0.1276], + [ 0.0681, -0.1067, -0.0904, ..., -0.1232, -0.0733, -0.0986], + [-0.1490, 0.0204, 0.1074, ..., 0.0826, -0.0679, -0.1399]], + device='cuda:0'), grad: tensor([[-6.9477e-07, 5.7742e-07, 3.1758e-07, ..., 5.8860e-07, + 1.8626e-09, 1.4715e-07], + [ 4.5635e-08, 1.5460e-07, -1.6391e-07, ..., -3.9767e-07, + -1.4249e-07, 4.1910e-08], + [ 1.1642e-07, 3.3528e-08, 1.8347e-07, ..., 6.8918e-08, + 8.3819e-09, -8.5682e-08], + ..., + [ 2.4214e-08, 5.9456e-06, 5.7407e-06, ..., 6.0312e-06, + 8.2888e-08, 8.3353e-07], + [-2.7008e-07, 1.7975e-07, 3.9395e-07, ..., 4.8708e-07, + 1.2107e-08, 1.2759e-07], + [ 2.2072e-07, -7.9721e-06, -7.1973e-06, ..., -7.1824e-06, + 2.0489e-08, -1.2033e-06]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0242, -0.0250, -0.0366, -0.0032, 0.0063, 0.0385, 0.0205, -0.0146, + 0.0019, -0.0080], device='cuda:0'), grad: tensor([-1.3653e-06, -6.9663e-07, 9.3691e-07, 4.5523e-06, 2.7288e-07, + -6.8452e-07, 1.6093e-06, 2.4900e-05, 4.1723e-07, -2.9922e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 213.83, cls_loss 0.0020 cls_loss_mapping 0.0039 cls_loss_causal 0.5144 re_mapping 0.0052 re_causal 0.0150 /// teacc 99.05 lr 0.00010000 +Epoch 197, weight, value: tensor([[ 0.0214, -0.0695, -0.0459, ..., -0.1729, -0.0828, -0.1330], + [-0.0424, -0.0541, -0.0317, ..., 0.0265, 0.0201, -0.1153], + [ 0.0321, -0.1138, -0.1260, ..., 0.0206, -0.1501, 0.0675], + ..., + [-0.0072, -0.0227, -0.1473, ..., -0.1248, 0.0236, 0.1277], + [ 0.0667, -0.1070, -0.0905, ..., -0.1237, -0.0741, -0.0981], + [-0.1504, 0.0205, 0.1077, ..., 0.0826, -0.0681, -0.1402]], + device='cuda:0'), grad: tensor([[ 7.3388e-06, 1.1131e-05, 1.8328e-05, ..., 1.2107e-08, + 5.5879e-09, 2.4494e-07], + [ 7.2643e-08, 1.4249e-07, 7.7300e-08, ..., -2.7008e-08, + 9.3132e-10, 2.1048e-07], + [ 4.4703e-08, 8.1956e-08, 4.2841e-08, ..., -2.7940e-08, + 9.3132e-10, -2.8480e-06], + ..., + [ 8.4750e-08, 1.1083e-07, 5.0291e-08, ..., 5.4948e-08, + -0.0000e+00, 2.5369e-06], + [-1.5739e-07, 2.3283e-07, 1.9744e-07, ..., 5.0291e-08, + 9.3132e-10, 3.6974e-07], + [ 9.6858e-08, 1.0710e-07, -1.8626e-09, ..., 9.3132e-09, + 9.3132e-10, 1.6857e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0242, -0.0244, -0.0378, -0.0037, 0.0059, 0.0390, 0.0205, -0.0149, + 0.0019, -0.0082], device='cuda:0'), grad: tensor([ 3.9756e-05, 7.0222e-07, -3.8594e-06, -8.4996e-05, 1.4957e-06, + 8.3685e-05, -4.2200e-05, 4.3139e-06, 3.6508e-07, 7.3109e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 213.96, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.5296 re_mapping 0.0054 re_causal 0.0164 /// teacc 99.02 lr 0.00010000 +Epoch 198, weight, value: tensor([[ 0.0211, -0.0704, -0.0464, ..., -0.1734, -0.0835, -0.1343], + [-0.0425, -0.0541, -0.0317, ..., 0.0266, 0.0202, -0.1154], + [ 0.0322, -0.1138, -0.1271, ..., 0.0202, -0.1502, 0.0681], + ..., + [-0.0074, -0.0227, -0.1475, ..., -0.1251, 0.0236, 0.1277], + [ 0.0672, -0.1073, -0.0908, ..., -0.1238, -0.0743, -0.0986], + [-0.1506, 0.0206, 0.1083, ..., 0.0830, -0.0682, -0.1405]], + device='cuda:0'), grad: tensor([[-8.3819e-08, 6.6124e-08, 3.1665e-08, ..., 2.8871e-08, + 4.6566e-09, 1.9185e-07], + [ 1.5832e-08, 2.3097e-07, -3.4552e-07, ..., -4.3586e-07, + -1.3039e-07, 4.4703e-07], + [ 1.6764e-08, 3.7905e-07, 1.2107e-08, ..., 1.5832e-08, + 1.8626e-09, -6.7614e-07], + ..., + [ 1.5832e-08, 3.1386e-07, 2.3842e-07, ..., 3.0547e-07, + 9.8720e-08, 1.8151e-06], + [ 2.2538e-07, 2.4121e-07, 3.1665e-08, ..., 4.7963e-07, + 9.3132e-09, 5.5134e-07], + [ 5.5879e-08, 4.7497e-08, 2.1420e-08, ..., 8.1025e-08, + 6.5193e-09, 1.0617e-07]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0245, -0.0244, -0.0375, -0.0036, 0.0054, 0.0389, 0.0207, -0.0150, + 0.0020, -0.0079], device='cuda:0'), grad: tensor([ 1.5646e-07, -3.5577e-07, -7.1339e-07, -6.7465e-06, 6.7987e-08, + 5.8860e-07, 2.0862e-07, 4.3735e-06, 1.9334e-06, 4.9733e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 213.95, cls_loss 0.0027 cls_loss_mapping 0.0034 cls_loss_causal 0.5373 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.06 lr 0.00010000 +Epoch 199, weight, value: tensor([[ 0.0212, -0.0706, -0.0475, ..., -0.1759, -0.0836, -0.1351], + [-0.0425, -0.0542, -0.0308, ..., 0.0287, 0.0237, -0.1173], + [ 0.0321, -0.1137, -0.1293, ..., 0.0177, -0.1498, 0.0682], + ..., + [-0.0075, -0.0227, -0.1486, ..., -0.1272, 0.0204, 0.1292], + [ 0.0671, -0.1081, -0.0916, ..., -0.1247, -0.0759, -0.0963], + [-0.1510, 0.0206, 0.1092, ..., 0.0837, -0.0687, -0.1403]], + device='cuda:0'), grad: tensor([[-2.6077e-08, 6.5193e-09, 1.0431e-07, ..., 7.3574e-08, + 4.6566e-09, 8.6613e-08], + [ 5.4948e-08, 7.4506e-09, -5.0012e-07, ..., -5.2061e-07, + -1.5739e-07, 4.4703e-08], + [ 1.4901e-08, 2.0489e-08, 2.0489e-08, ..., 1.2107e-08, + 1.8626e-09, 5.1223e-08], + ..., + [ 1.4901e-08, -3.9116e-08, 3.4552e-07, ..., 4.3400e-07, + 8.9407e-08, 1.5087e-07], + [-5.4669e-07, 1.0151e-07, 2.2817e-07, ..., 5.8860e-07, + 1.0990e-07, 1.3020e-06], + [ 1.2107e-07, 1.2107e-08, 2.6915e-07, ..., 6.3796e-07, + 2.8871e-08, 6.5193e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0246, -0.0262, -0.0380, -0.0026, 0.0052, 0.0385, 0.0206, -0.0135, + 0.0030, -0.0070], device='cuda:0'), grad: tensor([-5.7928e-07, -1.0869e-06, 3.1013e-07, -5.0962e-06, -2.9616e-07, + 4.2841e-07, 9.8813e-07, 1.2843e-06, 1.6680e-06, 2.3581e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 214.35, cls_loss 0.0016 cls_loss_mapping 0.0032 cls_loss_causal 0.5104 re_mapping 0.0051 re_causal 0.0151 /// teacc 99.04 lr 0.00010000 +Epoch 200, weight, value: tensor([[ 0.0213, -0.0719, -0.0479, ..., -0.1766, -0.0836, -0.1358], + [-0.0426, -0.0542, -0.0306, ..., 0.0289, 0.0245, -0.1171], + [ 0.0319, -0.1138, -0.1301, ..., 0.0179, -0.1503, 0.0690], + ..., + [-0.0076, -0.0225, -0.1489, ..., -0.1275, 0.0199, 0.1290], + [ 0.0673, -0.1084, -0.0921, ..., -0.1252, -0.0767, -0.0973], + [-0.1517, 0.0206, 0.1093, ..., 0.0833, -0.0696, -0.1408]], + device='cuda:0'), grad: tensor([[-6.5565e-07, 2.2352e-08, 3.6694e-07, ..., 3.2317e-07, + 1.1176e-08, 1.0151e-07], + [ 5.0757e-07, 5.9605e-08, -1.3132e-07, ..., 1.5609e-06, + 9.3132e-09, 1.0757e-06], + [-6.2771e-07, 2.3283e-08, 6.5006e-07, ..., -2.2259e-06, + 1.8626e-08, -1.4156e-06], + ..., + [ 2.7008e-08, 3.1665e-08, 1.8505e-06, ..., 1.0794e-06, + 3.9116e-08, 5.5879e-08], + [ 4.2003e-07, 1.1548e-07, 1.0990e-06, ..., 1.2983e-06, + 3.2596e-08, 2.3004e-07], + [ 1.7509e-07, 5.6811e-08, -1.0125e-05, ..., -5.0627e-06, + -2.9895e-07, 9.4064e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0245, -0.0256, -0.0375, -0.0029, 0.0053, 0.0387, 0.0205, -0.0141, + 0.0026, -0.0074], device='cuda:0'), grad: tensor([-1.4119e-06, 4.4890e-06, -5.5730e-06, -1.2387e-07, 2.7083e-06, + -5.3868e-06, 1.4357e-05, 4.0829e-06, 4.5188e-06, -1.7688e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 214.07, cls_loss 0.0016 cls_loss_mapping 0.0020 cls_loss_causal 0.5074 re_mapping 0.0052 re_causal 0.0154 /// teacc 99.10 lr 0.00010000 +Epoch 201, weight, value: tensor([[ 0.0215, -0.0700, -0.0480, ..., -0.1772, -0.0837, -0.1368], + [-0.0427, -0.0540, -0.0303, ..., 0.0292, 0.0246, -0.1174], + [ 0.0319, -0.1139, -0.1300, ..., 0.0187, -0.1504, 0.0699], + ..., + [-0.0078, -0.0227, -0.1493, ..., -0.1280, 0.0198, 0.1292], + [ 0.0675, -0.1087, -0.0925, ..., -0.1257, -0.0771, -0.0976], + [-0.1520, 0.0207, 0.1095, ..., 0.0834, -0.0702, -0.1410]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 3.7253e-09, 3.1665e-08, ..., 4.3772e-08, + 0.0000e+00, 5.2154e-08], + [ 6.5193e-09, 8.3819e-09, -8.1956e-08, ..., -2.0768e-07, + 9.3132e-10, 6.9849e-08], + [-7.6368e-08, 9.3132e-09, 2.5146e-08, ..., 2.7940e-08, + 0.0000e+00, -1.4845e-06], + ..., + [ 3.6322e-08, 1.7695e-08, 2.5705e-07, ..., 3.4645e-07, + -9.3132e-10, 8.6147e-07], + [ 4.9360e-08, 3.4459e-08, 1.0803e-07, ..., 1.3504e-07, + 1.8626e-09, 3.7905e-07], + [ 2.4214e-08, -2.6431e-06, -2.4617e-05, ..., -1.7554e-05, + 3.7253e-09, 2.8871e-08]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0242, -0.0257, -0.0370, -0.0029, 0.0053, 0.0388, 0.0201, -0.0140, + 0.0025, -0.0074], device='cuda:0'), grad: tensor([ 1.6764e-07, -5.0943e-07, -1.7853e-06, -5.6345e-07, 2.8744e-05, + -2.9840e-06, 3.3602e-06, 1.8440e-06, 8.4378e-07, -2.9117e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 214.06, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5030 re_mapping 0.0054 re_causal 0.0155 /// teacc 99.06 lr 0.00010000 +Epoch 202, weight, value: tensor([[ 0.0215, -0.0706, -0.0481, ..., -0.1777, -0.0838, -0.1379], + [-0.0435, -0.0541, -0.0299, ..., 0.0295, 0.0247, -0.1176], + [ 0.0331, -0.1160, -0.1308, ..., 0.0189, -0.1506, 0.0701], + ..., + [-0.0082, -0.0230, -0.1496, ..., -0.1282, 0.0202, 0.1294], + [ 0.0676, -0.1095, -0.0930, ..., -0.1264, -0.0778, -0.0983], + [-0.1528, 0.0208, 0.1092, ..., 0.0831, -0.0730, -0.1416]], + device='cuda:0'), grad: tensor([[-4.4703e-08, 1.1176e-08, 9.6858e-08, ..., 6.4261e-08, + 1.8626e-09, 2.7008e-08], + [ 1.2107e-08, 2.1420e-08, -2.2724e-07, ..., -3.1199e-07, + 1.8626e-09, 2.0005e-06], + [-1.0245e-07, 2.3283e-08, 3.5390e-08, ..., -1.8626e-08, + 9.3132e-10, 1.8813e-07], + ..., + [ 8.3819e-09, -5.4017e-08, 3.6322e-07, ..., 3.8277e-07, + -5.5879e-09, -2.9393e-06], + [ 9.0338e-08, 8.8476e-08, 5.9791e-07, ..., 4.9919e-07, + 3.7253e-09, 6.0257e-07], + [ 3.3528e-08, -1.0245e-07, -5.3924e-07, ..., -5.5414e-07, + 9.3132e-10, 7.3574e-08]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0241, -0.0258, -0.0367, -0.0021, 0.0056, 0.0387, 0.0201, -0.0139, + 0.0022, -0.0080], device='cuda:0'), grad: tensor([-9.3132e-10, 4.4331e-06, 6.1002e-07, 2.2352e-07, -3.7439e-07, + 7.5437e-08, -6.6124e-08, -6.8508e-06, 2.5928e-06, -6.3051e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.23, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.4941 re_mapping 0.0052 re_causal 0.0157 /// teacc 99.02 lr 0.00010000 +Epoch 203, weight, value: tensor([[ 0.0215, -0.0707, -0.0483, ..., -0.1787, -0.0838, -0.1384], + [-0.0436, -0.0533, -0.0297, ..., 0.0293, 0.0247, -0.1176], + [ 0.0338, -0.1165, -0.1312, ..., 0.0214, -0.1507, 0.0711], + ..., + [-0.0084, -0.0236, -0.1498, ..., -0.1284, 0.0202, 0.1294], + [ 0.0674, -0.1102, -0.0932, ..., -0.1268, -0.0809, -0.0987], + [-0.1534, 0.0209, 0.1094, ..., 0.0832, -0.0730, -0.1417]], + device='cuda:0'), grad: tensor([[-1.2852e-07, 2.7940e-09, 2.0489e-08, ..., 1.8626e-08, + 0.0000e+00, 1.8626e-09], + [ 5.5879e-09, 1.3970e-08, -5.1688e-07, ..., -3.0827e-07, + 0.0000e+00, 1.6764e-08], + [ 8.3819e-09, 5.5879e-09, 2.7008e-08, ..., 4.4703e-08, + 0.0000e+00, 6.5193e-09], + ..., + [ 3.7253e-09, 9.3132e-10, 3.9861e-07, ..., 3.3248e-07, + -9.3132e-10, -3.9116e-08], + [ 2.2352e-08, 5.0291e-08, 2.9802e-07, ..., 4.2096e-07, + 0.0000e+00, 1.5832e-08], + [ 9.5926e-08, -9.7789e-08, -2.3842e-07, ..., -9.9652e-08, + 9.3132e-10, 2.1420e-08]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0242, -0.0256, -0.0357, -0.0008, 0.0056, 0.0379, 0.0202, -0.0141, + 0.0020, -0.0079], device='cuda:0'), grad: tensor([-3.5204e-07, -1.5656e-06, 1.5460e-07, 3.2276e-05, -9.2201e-08, + -3.3140e-05, 5.5414e-07, 1.0543e-06, 1.0552e-06, 3.5390e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 213.86, cls_loss 0.0016 cls_loss_mapping 0.0034 cls_loss_causal 0.5092 re_mapping 0.0051 re_causal 0.0150 /// teacc 99.02 lr 0.00010000 +Epoch 204, weight, value: tensor([[ 0.0216, -0.0709, -0.0485, ..., -0.1791, -0.0838, -0.1390], + [-0.0436, -0.0534, -0.0295, ..., 0.0297, 0.0248, -0.1177], + [ 0.0338, -0.1171, -0.1319, ..., 0.0212, -0.1508, 0.0713], + ..., + [-0.0083, -0.0236, -0.1500, ..., -0.1286, 0.0201, 0.1295], + [ 0.0672, -0.1113, -0.0933, ..., -0.1274, -0.0815, -0.0992], + [-0.1539, 0.0208, 0.1094, ..., 0.0831, -0.0732, -0.1420]], + device='cuda:0'), grad: tensor([[-3.1665e-08, 7.4506e-09, 6.0536e-08, ..., 1.3690e-07, + 3.7253e-09, 1.6112e-07], + [ 1.9930e-07, 1.0151e-07, -3.9674e-07, ..., 5.3085e-08, + 4.6566e-08, 4.5728e-07], + [ 1.3039e-08, 9.3132e-08, 6.8918e-08, ..., -4.1444e-07, + 9.3132e-10, -1.8058e-06], + ..., + [ 2.0489e-08, -3.9861e-07, 2.0489e-08, ..., 4.3586e-07, + 9.3132e-10, -8.2795e-07], + [ 2.3749e-07, 5.1223e-08, 1.3690e-07, ..., 1.4845e-06, + 6.4261e-08, 1.4715e-07], + [ 6.8918e-08, 1.3132e-07, 1.6671e-07, ..., 2.5705e-07, + 7.4506e-09, 3.6228e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0241, -0.0256, -0.0359, 0.0020, 0.0055, 0.0358, 0.0201, -0.0141, + 0.0018, -0.0080], device='cuda:0'), grad: tensor([ 5.2620e-07, 5.9325e-07, -4.6454e-06, 2.4252e-06, 1.0794e-06, + -3.3900e-06, 7.9721e-07, -1.2843e-06, 2.3469e-06, 1.5395e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.23, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4801 re_mapping 0.0052 re_causal 0.0155 /// teacc 98.93 lr 0.00010000 +Epoch 205, weight, value: tensor([[ 0.0216, -0.0712, -0.0487, ..., -0.1793, -0.0839, -0.1397], + [-0.0439, -0.0536, -0.0292, ..., 0.0301, 0.0248, -0.1177], + [ 0.0339, -0.1178, -0.1332, ..., 0.0207, -0.1508, 0.0709], + ..., + [-0.0085, -0.0235, -0.1502, ..., -0.1289, 0.0201, 0.1296], + [ 0.0672, -0.1121, -0.0936, ..., -0.1278, -0.0816, -0.0995], + [-0.1548, 0.0206, 0.1094, ..., 0.0829, -0.0733, -0.1426]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 9.4064e-08, 1.2573e-06, ..., 1.1502e-06, + 0.0000e+00, 7.7300e-08], + [ 3.0734e-08, 5.4948e-08, 2.6636e-07, ..., -3.7067e-07, + 1.8626e-09, 2.8219e-07], + [ 3.1665e-08, 2.2352e-08, 2.4773e-07, ..., 2.3749e-07, + 9.3132e-10, -9.5088e-07], + ..., + [ 5.5879e-09, 5.6811e-08, 9.8627e-07, ..., 1.1036e-06, + -7.4506e-09, -5.0291e-08], + [-3.7625e-07, 7.9162e-08, 6.9104e-07, ..., 8.1584e-07, + 9.3132e-10, 6.5472e-07], + [ 1.2945e-07, -5.0571e-07, -4.8503e-06, ..., -5.2117e-06, + 9.3132e-10, 9.7789e-08]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0241, -0.0255, -0.0363, 0.0021, 0.0056, 0.0359, 0.0202, -0.0141, + 0.0018, -0.0085], device='cuda:0'), grad: tensor([ 3.3639e-06, -8.8383e-07, -1.0105e-06, 7.6182e-07, -1.4901e-07, + 1.2275e-06, 3.2652e-06, 3.0790e-06, 1.6671e-06, -1.1340e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 213.88, cls_loss 0.0020 cls_loss_mapping 0.0031 cls_loss_causal 0.5215 re_mapping 0.0052 re_causal 0.0148 /// teacc 99.07 lr 0.00010000 +Epoch 206, weight, value: tensor([[ 0.0218, -0.0722, -0.0493, ..., -0.1804, -0.0839, -0.1425], + [-0.0445, -0.0535, -0.0291, ..., 0.0303, 0.0248, -0.1177], + [ 0.0342, -0.1180, -0.1355, ..., 0.0190, -0.1509, 0.0714], + ..., + [-0.0088, -0.0237, -0.1506, ..., -0.1295, 0.0202, 0.1296], + [ 0.0677, -0.1126, -0.0937, ..., -0.1283, -0.0818, -0.0998], + [-0.1554, 0.0201, 0.1092, ..., 0.0826, -0.0733, -0.1428]], + device='cuda:0'), grad: tensor([[ 1.0245e-07, 2.1420e-08, 1.1735e-07, ..., 6.0536e-08, + 0.0000e+00, 3.0734e-08], + [ 1.7323e-07, 2.9802e-08, -5.2433e-07, ..., -8.0559e-07, + 0.0000e+00, 4.0047e-08], + [ 7.6368e-08, 3.4459e-08, 8.5682e-08, ..., 3.6322e-08, + 0.0000e+00, -2.0303e-07], + ..., + [ 2.2445e-07, 7.9162e-08, 4.4890e-07, ..., 4.9453e-07, + -9.3132e-10, -3.9116e-08], + [-9.5274e-07, 4.6100e-07, -3.7253e-08, ..., 2.6356e-07, + 0.0000e+00, 1.7583e-06], + [ 2.1886e-07, 7.7300e-08, -2.4401e-07, ..., -2.1420e-07, + 0.0000e+00, 1.3039e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0241, -0.0255, -0.0366, 0.0019, 0.0068, 0.0361, 0.0199, -0.0142, + 0.0021, -0.0089], device='cuda:0'), grad: tensor([ 8.3819e-07, -1.0831e-06, 2.7753e-07, -6.1579e-06, 6.0815e-07, + 4.6100e-07, -1.8626e-09, 2.7306e-06, 1.1595e-06, 1.1446e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 213.79, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.5117 re_mapping 0.0050 re_causal 0.0151 /// teacc 99.00 lr 0.00010000 +Epoch 207, weight, value: tensor([[ 0.0219, -0.0726, -0.0493, ..., -0.1807, -0.0840, -0.1430], + [-0.0455, -0.0536, -0.0295, ..., 0.0303, 0.0248, -0.1178], + [ 0.0339, -0.1182, -0.1373, ..., 0.0189, -0.1510, 0.0712], + ..., + [-0.0090, -0.0239, -0.1510, ..., -0.1299, 0.0203, 0.1298], + [ 0.0681, -0.1132, -0.0936, ..., -0.1287, -0.0819, -0.1001], + [-0.1562, 0.0190, 0.1082, ..., 0.0816, -0.0737, -0.1430]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, 1.3039e-08, 8.0094e-08, ..., 1.8533e-07, + 0.0000e+00, 2.8871e-08], + [ 3.9414e-06, 1.3970e-08, 1.0148e-05, ..., -1.2759e-07, + 0.0000e+00, 4.0047e-08], + [ 1.9278e-07, 2.4214e-08, 2.6636e-07, ..., 1.1642e-07, + 0.0000e+00, 1.8813e-07], + ..., + [ 2.7940e-07, 6.6124e-08, 2.3749e-07, ..., 4.4983e-07, + 0.0000e+00, -6.8918e-08], + [-5.3570e-06, 6.3330e-08, -1.1533e-05, ..., 4.3586e-07, + 0.0000e+00, 2.3004e-07], + [ 1.8347e-07, 1.5832e-08, 2.2165e-07, ..., 2.1234e-07, + 0.0000e+00, 4.3772e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0240, -0.0256, -0.0375, 0.0017, 0.0070, 0.0361, 0.0221, -0.0142, + 0.0023, -0.0105], device='cuda:0'), grad: tensor([ 1.8347e-07, 5.0306e-05, 1.8338e-06, -1.4715e-06, 1.3802e-06, + -3.2336e-05, 3.5048e-05, 1.4203e-06, -5.7727e-05, 1.3830e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 214.29, cls_loss 0.0021 cls_loss_mapping 0.0037 cls_loss_causal 0.4789 re_mapping 0.0055 re_causal 0.0150 /// teacc 99.03 lr 0.00010000 +Epoch 208, weight, value: tensor([[ 0.0222, -0.0730, -0.0495, ..., -0.1812, -0.0841, -0.1435], + [-0.0467, -0.0508, -0.0270, ..., 0.0312, 0.0247, -0.1169], + [ 0.0336, -0.1188, -0.1385, ..., 0.0184, -0.1510, 0.0697], + ..., + [-0.0101, -0.0267, -0.1535, ..., -0.1306, 0.0204, 0.1291], + [ 0.0697, -0.1136, -0.0932, ..., -0.1290, -0.0820, -0.1001], + [-0.1596, 0.0193, 0.1085, ..., 0.0818, -0.0744, -0.1438]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 2.7940e-09, 1.5832e-08, ..., 2.2352e-08, + 0.0000e+00, 7.8231e-08], + [ 2.3283e-08, 6.5193e-09, -3.2783e-07, ..., -3.3341e-07, + 0.0000e+00, 6.7987e-08], + [ 4.4703e-08, 9.3132e-09, 6.5193e-09, ..., 1.1176e-08, + 0.0000e+00, -4.0326e-07], + ..., + [ 1.0245e-08, 1.6764e-08, 1.5367e-07, ..., 1.6391e-07, + 0.0000e+00, 6.1467e-08], + [-2.1793e-07, 3.0734e-08, 6.0536e-08, ..., 1.8347e-07, + 0.0000e+00, 1.2759e-07], + [ 6.7987e-08, 9.3132e-10, -2.8871e-08, ..., -1.3970e-08, + 0.0000e+00, 1.9558e-08]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0239, -0.0247, -0.0391, 0.0011, 0.0068, 0.0366, 0.0219, -0.0150, + 0.0035, -0.0107], device='cuda:0'), grad: tensor([ 1.8999e-07, -8.4843e-07, -5.1036e-07, -9.1568e-06, 3.4180e-07, + -7.3910e-06, 1.6749e-05, 6.0350e-07, -1.1642e-07, 2.1234e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 214.23, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.4950 re_mapping 0.0055 re_causal 0.0157 /// teacc 99.00 lr 0.00010000 +Epoch 209, weight, value: tensor([[ 0.0192, -0.0730, -0.0498, ..., -0.1823, -0.0842, -0.1439], + [-0.0469, -0.0508, -0.0267, ..., 0.0316, 0.0247, -0.1169], + [ 0.0306, -0.1190, -0.1399, ..., 0.0155, -0.1511, 0.0702], + ..., + [-0.0108, -0.0267, -0.1535, ..., -0.1307, 0.0204, 0.1291], + [ 0.0739, -0.1138, -0.0934, ..., -0.1273, -0.0837, -0.0995], + [-0.1605, 0.0194, 0.1068, ..., 0.0800, -0.0744, -0.1442]], + device='cuda:0'), grad: tensor([[-3.3714e-07, 0.0000e+00, 8.3819e-09, ..., 4.4703e-08, + 0.0000e+00, 2.1420e-07], + [ 1.1176e-08, 2.7940e-09, -7.3574e-08, ..., 3.1292e-07, + 0.0000e+00, 4.3865e-07], + [ 1.1176e-08, 1.8626e-09, 5.5879e-09, ..., -2.3283e-08, + 0.0000e+00, -2.7910e-05], + ..., + [ 9.3132e-09, -2.7940e-09, 2.9802e-08, ..., 3.8277e-07, + 0.0000e+00, 2.3633e-05], + [ 4.0047e-08, 2.7940e-09, 2.0489e-08, ..., 2.3097e-07, + 0.0000e+00, 1.1455e-06], + [ 1.4994e-07, 9.3132e-10, -7.4506e-09, ..., 4.4890e-07, + 0.0000e+00, 2.9337e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0268, -0.0246, -0.0407, 0.0034, 0.0084, 0.0347, 0.0233, -0.0150, + 0.0064, -0.0123], device='cuda:0'), grad: tensor([-6.8825e-07, 1.7611e-06, -3.8803e-05, -1.4985e-04, 1.5460e-07, + 1.4853e-04, 4.7684e-07, 3.3677e-05, 2.4773e-06, 2.5202e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 214.20, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.4903 re_mapping 0.0053 re_causal 0.0155 /// teacc 99.05 lr 0.00010000 +Epoch 210, weight, value: tensor([[ 0.0195, -0.0737, -0.0499, ..., -0.1830, -0.0842, -0.1449], + [-0.0469, -0.0508, -0.0260, ..., 0.0327, 0.0247, -0.1177], + [ 0.0307, -0.1191, -0.1402, ..., 0.0157, -0.1511, 0.0713], + ..., + [-0.0110, -0.0268, -0.1542, ..., -0.1318, 0.0204, 0.1298], + [ 0.0738, -0.1141, -0.0938, ..., -0.1282, -0.0837, -0.0999], + [-0.1622, 0.0194, 0.1063, ..., 0.0795, -0.0744, -0.1448]], + device='cuda:0'), grad: tensor([[-2.2247e-05, 6.5193e-09, 5.1223e-08, ..., 5.6811e-08, + 1.8626e-09, 1.3970e-08], + [ 8.6240e-07, 1.3039e-08, 2.5239e-07, ..., 2.9337e-07, + 1.4901e-08, 4.1910e-08], + [ 7.8138e-07, 1.0245e-08, 1.1083e-07, ..., 1.2759e-07, + 0.0000e+00, 1.8626e-08], + ..., + [ 1.8161e-07, 1.1176e-08, 4.7684e-07, ..., 5.9232e-07, + 1.4901e-08, -4.3772e-08], + [ 1.8897e-06, 1.1176e-08, 3.4552e-07, ..., 5.8766e-07, + 7.8231e-08, 4.1910e-08], + [ 3.3155e-06, -5.5879e-09, -4.2692e-06, ..., -4.9509e-06, + 1.6764e-08, 4.7497e-08]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0266, -0.0251, -0.0403, 0.0023, 0.0087, 0.0358, 0.0230, -0.0145, + 0.0060, -0.0128], device='cuda:0'), grad: tensor([-7.2658e-05, 3.2596e-06, 2.8014e-06, 4.4517e-07, 5.3309e-06, + 9.0525e-07, 4.8339e-05, 1.4370e-06, 6.7912e-06, 3.3453e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 214.17, cls_loss 0.0020 cls_loss_mapping 0.0037 cls_loss_causal 0.4852 re_mapping 0.0050 re_causal 0.0147 /// teacc 99.08 lr 0.00010000 +Epoch 211, weight, value: tensor([[ 0.0196, -0.0739, -0.0505, ..., -0.1839, -0.0842, -0.1458], + [-0.0474, -0.0508, -0.0259, ..., 0.0329, 0.0247, -0.1178], + [ 0.0310, -0.1194, -0.1407, ..., 0.0158, -0.1510, 0.0716], + ..., + [-0.0116, -0.0270, -0.1552, ..., -0.1324, 0.0204, 0.1300], + [ 0.0736, -0.1146, -0.0945, ..., -0.1290, -0.0838, -0.1003], + [-0.1634, 0.0197, 0.1068, ..., 0.0797, -0.0745, -0.1455]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 2.1160e-06, ..., 6.3842e-07, + 0.0000e+00, 2.1420e-08], + [ 1.9558e-08, 5.1223e-09, 6.7428e-07, ..., 1.3085e-07, + 0.0000e+00, 5.1223e-08], + [-3.2596e-08, 1.3970e-09, 1.0291e-07, ..., -7.0781e-08, + 0.0000e+00, -3.6601e-07], + ..., + [ 2.7940e-08, 1.4901e-08, 6.9104e-07, ..., 3.6461e-07, + 0.0000e+00, 1.7229e-07], + [ 5.7975e-07, 3.2131e-08, -8.6147e-08, ..., 2.1923e-06, + 0.0000e+00, 7.2177e-08], + [ 1.2433e-07, -3.6322e-08, -2.0768e-07, ..., -1.3132e-07, + 0.0000e+00, 4.8894e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0265, -0.0250, -0.0399, 0.0046, 0.0086, 0.0337, 0.0229, -0.0147, + 0.0056, -0.0124], device='cuda:0'), grad: tensor([ 6.9775e-06, 2.3134e-06, -4.2422e-07, 7.6229e-07, -1.8224e-05, + -4.2506e-06, 6.6645e-06, 2.5108e-06, 3.2187e-06, 4.2468e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 214.27, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5131 re_mapping 0.0049 re_causal 0.0155 /// teacc 99.09 lr 0.00010000 +Epoch 212, weight, value: tensor([[ 0.0197, -0.0740, -0.0505, ..., -0.1842, -0.0842, -0.1462], + [-0.0476, -0.0508, -0.0258, ..., 0.0330, 0.0247, -0.1178], + [ 0.0310, -0.1196, -0.1414, ..., 0.0156, -0.1510, 0.0710], + ..., + [-0.0118, -0.0270, -0.1555, ..., -0.1328, 0.0204, 0.1301], + [ 0.0736, -0.1149, -0.0953, ..., -0.1298, -0.0838, -0.1006], + [-0.1640, 0.0197, 0.1070, ..., 0.0799, -0.0745, -0.1458]], + device='cuda:0'), grad: tensor([[ 1.9884e-07, 2.3283e-09, 1.9260e-06, ..., 2.3190e-07, + 0.0000e+00, 1.2666e-07], + [ 4.3306e-08, 4.1910e-09, 4.1910e-07, ..., 3.3062e-08, + 0.0000e+00, 5.1223e-08], + [ 9.8720e-08, 5.5879e-09, 1.6009e-06, ..., 4.0792e-07, + 0.0000e+00, -1.1362e-07], + ..., + [ 9.7789e-09, 1.6764e-08, 3.9628e-07, ..., 3.6415e-07, + 0.0000e+00, 5.3551e-08], + [ 9.8720e-08, 1.1176e-08, 3.7113e-07, ..., 3.9767e-07, + 0.0000e+00, 6.0536e-08], + [ 5.9605e-08, -2.7008e-08, 6.3749e-07, ..., 1.8347e-07, + 0.0000e+00, 4.2375e-08]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0264, -0.0250, -0.0403, 0.0046, 0.0085, 0.0338, 0.0227, -0.0147, + 0.0053, -0.0121], device='cuda:0'), grad: tensor([ 4.6827e-06, -1.1316e-07, 3.0920e-06, -2.5537e-06, -3.4302e-05, + 1.9670e-06, 2.2545e-05, 1.2852e-06, 1.8664e-06, 1.4622e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 214.15, cls_loss 0.0011 cls_loss_mapping 0.0026 cls_loss_causal 0.5099 re_mapping 0.0051 re_causal 0.0154 /// teacc 99.00 lr 0.00010000 +Epoch 213, weight, value: tensor([[ 0.0197, -0.0741, -0.0507, ..., -0.1847, -0.0842, -0.1467], + [-0.0477, -0.0509, -0.0254, ..., 0.0339, 0.0247, -0.1178], + [ 0.0312, -0.1199, -0.1426, ..., 0.0158, -0.1510, 0.0714], + ..., + [-0.0121, -0.0270, -0.1556, ..., -0.1331, 0.0204, 0.1302], + [ 0.0734, -0.1155, -0.0959, ..., -0.1306, -0.0838, -0.1012], + [-0.1644, 0.0199, 0.1071, ..., 0.0800, -0.0745, -0.1460]], + device='cuda:0'), grad: tensor([[-2.3283e-09, 4.1910e-09, 1.7649e-07, ..., 4.6566e-07, + 0.0000e+00, 2.7008e-08], + [ 1.4901e-08, 2.3283e-08, -5.1223e-08, ..., 1.1362e-07, + 0.0000e+00, 5.4482e-08], + [ 1.7695e-08, 3.2596e-09, 1.6764e-08, ..., 3.0268e-08, + 0.0000e+00, -4.7497e-07], + ..., + [ 2.1886e-08, -3.9581e-08, 1.0384e-07, ..., 2.2445e-07, + 0.0000e+00, 2.9476e-07], + [-1.9697e-07, 2.1886e-08, 9.3412e-07, ..., 2.5406e-06, + 0.0000e+00, -9.3132e-10], + [ 5.9139e-08, -2.4214e-08, -6.0648e-06, ..., -1.7151e-05, + 0.0000e+00, 1.5832e-08]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0264, -0.0249, -0.0403, 0.0046, 0.0084, 0.0338, 0.0228, -0.0147, + 0.0048, -0.0121], device='cuda:0'), grad: tensor([ 1.1977e-06, -7.1079e-06, 1.2107e-08, 6.9365e-06, 1.3728e-06, + 1.2159e-05, 6.1560e-07, 3.0622e-06, 5.5730e-06, -2.3812e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 214.27, cls_loss 0.0015 cls_loss_mapping 0.0036 cls_loss_causal 0.5337 re_mapping 0.0050 re_causal 0.0149 /// teacc 99.02 lr 0.00010000 +Epoch 214, weight, value: tensor([[ 0.0197, -0.0743, -0.0511, ..., -0.1861, -0.0842, -0.1474], + [-0.0477, -0.0510, -0.0250, ..., 0.0347, 0.0247, -0.1179], + [ 0.0312, -0.1198, -0.1438, ..., 0.0159, -0.1511, 0.0725], + ..., + [-0.0124, -0.0269, -0.1557, ..., -0.1335, 0.0204, 0.1303], + [ 0.0733, -0.1171, -0.0973, ..., -0.1319, -0.0839, -0.1018], + [-0.1648, 0.0200, 0.1073, ..., 0.0802, -0.0746, -0.1468]], + device='cuda:0'), grad: tensor([[ 1.0403e-06, 9.3132e-09, 8.3866e-07, ..., 3.9116e-08, + 9.3132e-10, 1.8626e-08], + [ 7.4040e-08, 1.6764e-08, -8.7079e-08, ..., -2.8405e-07, + 5.1223e-09, 3.6787e-08], + [ 5.9139e-08, 2.6077e-08, 6.3330e-08, ..., 2.4214e-08, + 1.5832e-08, 3.4459e-08], + ..., + [ 7.4506e-09, -6.3330e-08, 1.1409e-07, ..., 2.0443e-07, + -5.7742e-08, -1.9697e-07], + [ 5.7789e-07, 2.5611e-08, 6.1048e-07, ..., 1.6531e-07, + 3.2596e-09, 4.7497e-08], + [ 5.7742e-08, -1.7229e-08, -1.4389e-07, ..., -2.7381e-07, + 1.0710e-08, 6.5658e-08]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0264, -0.0248, -0.0397, 0.0046, 0.0083, 0.0339, 0.0227, -0.0147, + 0.0042, -0.0120], device='cuda:0'), grad: tensor([ 5.5730e-06, -1.0291e-07, 4.1351e-07, 1.1409e-07, 1.4435e-08, + 6.4000e-06, -1.5870e-05, 1.0151e-07, 3.3900e-06, -3.7719e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 214.26, cls_loss 0.0018 cls_loss_mapping 0.0025 cls_loss_causal 0.4798 re_mapping 0.0052 re_causal 0.0148 /// teacc 98.97 lr 0.00010000 +Epoch 215, weight, value: tensor([[ 0.0198, -0.0751, -0.0516, ..., -0.1886, -0.0843, -0.1492], + [-0.0479, -0.0512, -0.0249, ..., 0.0331, 0.0248, -0.1180], + [ 0.0312, -0.1204, -0.1435, ..., 0.0164, -0.1510, 0.0727], + ..., + [-0.0130, -0.0259, -0.1558, ..., -0.1314, 0.0205, 0.1313], + [ 0.0737, -0.1173, -0.0977, ..., -0.1325, -0.0843, -0.1020], + [-0.1660, 0.0194, 0.1073, ..., 0.0799, -0.0752, -0.1497]], + device='cuda:0'), grad: tensor([[-4.4331e-06, -2.9104e-07, -1.1083e-07, ..., 5.8673e-08, + 0.0000e+00, 2.4680e-08], + [ 1.4296e-07, 8.8476e-09, 1.4110e-07, ..., 1.5507e-07, + 2.7940e-09, 1.0291e-07], + [ 2.2212e-07, 2.8405e-08, 1.2340e-07, ..., 1.2480e-07, + 4.6566e-10, 9.9186e-08], + ..., + [ 3.3062e-08, -1.5832e-08, 3.0501e-07, ..., 4.3772e-07, + 1.3970e-09, 2.1420e-08], + [-2.8778e-07, 1.1176e-08, 1.6717e-07, ..., 2.0070e-07, + 4.6566e-10, -3.5157e-07], + [ 1.3513e-06, 4.8429e-08, 7.0175e-07, ..., 9.2620e-07, + 4.6566e-09, 8.1491e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0264, -0.0251, -0.0395, 0.0041, 0.0080, 0.0345, 0.0227, -0.0141, + 0.0043, -0.0126], device='cuda:0'), grad: tensor([-2.3320e-05, 9.8161e-07, 1.6829e-06, 8.0764e-06, -4.1351e-06, + 4.4331e-07, 7.0743e-06, 9.3738e-07, -1.0896e-06, 9.2834e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.03, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.4906 re_mapping 0.0051 re_causal 0.0143 /// teacc 99.01 lr 0.00010000 +Epoch 216, weight, value: tensor([[ 0.0198, -0.0752, -0.0518, ..., -0.1895, -0.0844, -0.1499], + [-0.0481, -0.0508, -0.0238, ..., 0.0344, 0.0248, -0.1180], + [ 0.0311, -0.1208, -0.1451, ..., 0.0162, -0.1514, 0.0726], + ..., + [-0.0142, -0.0263, -0.1568, ..., -0.1330, 0.0205, 0.1314], + [ 0.0736, -0.1181, -0.0984, ..., -0.1335, -0.0846, -0.1019], + [-0.1672, 0.0195, 0.1076, ..., 0.0804, -0.0752, -0.1501]], + device='cuda:0'), grad: tensor([[ 6.0024e-07, 4.8522e-07, 3.1712e-07, ..., 6.6590e-08, + 0.0000e+00, 1.3970e-09], + [ 2.3423e-07, 1.5134e-07, -1.7183e-07, ..., -1.7323e-07, + 9.3132e-10, 1.3039e-08], + [ 1.3690e-07, 7.0781e-08, 1.0850e-07, ..., 5.3085e-08, + 0.0000e+00, 9.3132e-09], + ..., + [ 1.5879e-07, 1.0617e-07, 3.6787e-07, ..., 2.9802e-07, + -3.7253e-09, -2.4214e-08], + [-6.3851e-06, -4.5262e-06, -1.6894e-06, ..., 2.9057e-07, + 4.6566e-10, 2.9802e-08], + [ 4.0419e-06, 2.8666e-06, 3.0594e-07, ..., -1.1763e-06, + 4.6566e-10, 1.0245e-08]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0263, -0.0249, -0.0398, 0.0038, 0.0079, 0.0347, 0.0226, -0.0144, + 0.0042, -0.0122], device='cuda:0'), grad: tensor([ 5.4874e-06, 6.7987e-07, 1.0971e-06, 6.6832e-06, 7.5856e-07, + 2.6543e-06, 9.8813e-07, 2.2482e-06, -5.0962e-05, 3.0428e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 214.04, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5174 re_mapping 0.0049 re_causal 0.0146 /// teacc 99.10 lr 0.00010000 +Epoch 217, weight, value: tensor([[ 0.0198, -0.0766, -0.0522, ..., -0.1913, -0.0844, -0.1514], + [-0.0483, -0.0510, -0.0235, ..., 0.0348, 0.0249, -0.1180], + [ 0.0311, -0.1213, -0.1462, ..., 0.0162, -0.1514, 0.0730], + ..., + [-0.0129, -0.0261, -0.1570, ..., -0.1329, 0.0205, 0.1317], + [ 0.0738, -0.1200, -0.0982, ..., -0.1343, -0.0847, -0.1028], + [-0.1686, 0.0197, 0.1078, ..., 0.0804, -0.0752, -0.1507]], + device='cuda:0'), grad: tensor([[-8.9407e-08, 4.3306e-08, 7.6834e-08, ..., 1.3364e-07, + 0.0000e+00, 2.5690e-05], + [ 2.8405e-08, 5.2620e-08, 1.0505e-06, ..., 1.2983e-06, + 0.0000e+00, 1.5674e-06], + [ 2.6543e-08, 5.5879e-09, 8.3819e-08, ..., 8.4285e-08, + 0.0000e+00, -1.9968e-04], + ..., + [ 1.7229e-08, 4.2394e-06, 2.7400e-06, ..., 9.5963e-06, + 0.0000e+00, 9.9540e-05], + [ 1.7760e-06, 5.1223e-08, 1.2025e-05, ..., 4.0792e-06, + 0.0000e+00, 6.4112e-06], + [ 7.3109e-08, -4.9807e-06, -1.5348e-06, ..., -9.4846e-06, + 0.0000e+00, 5.4166e-06]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0264, -0.0248, -0.0398, 0.0037, 0.0076, 0.0349, 0.0224, -0.0142, + 0.0042, -0.0122], device='cuda:0'), grad: tensor([ 6.0648e-05, 5.7667e-06, -4.7231e-04, 1.3077e-04, -2.4468e-05, + -6.4261e-07, 8.0913e-06, 2.5344e-04, 4.3839e-05, -5.5395e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 214.09, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5175 re_mapping 0.0051 re_causal 0.0156 /// teacc 99.07 lr 0.00010000 +Epoch 218, weight, value: tensor([[ 0.0199, -0.0769, -0.0522, ..., -0.1910, -0.0844, -0.1530], + [-0.0490, -0.0511, -0.0234, ..., 0.0350, 0.0249, -0.1182], + [ 0.0315, -0.1215, -0.1468, ..., 0.0163, -0.1511, 0.0752], + ..., + [-0.0134, -0.0261, -0.1576, ..., -0.1338, 0.0205, 0.1317], + [ 0.0742, -0.1199, -0.1011, ..., -0.1363, -0.0850, -0.1051], + [-0.1698, 0.0198, 0.1097, ..., 0.0822, -0.0752, -0.1480]], + device='cuda:0'), grad: tensor([[-2.7940e-09, 1.9092e-08, 2.6543e-08, ..., 3.6322e-08, + 0.0000e+00, 4.8894e-08], + [-2.2817e-07, 6.3796e-08, -6.9384e-08, ..., -1.7043e-07, + 0.0000e+00, -6.3237e-07], + [ 1.3225e-07, 2.1420e-08, 3.0734e-08, ..., 2.9337e-08, + 0.0000e+00, 2.6263e-07], + ..., + [ 1.5087e-07, -4.1462e-06, 6.9384e-08, ..., 9.9652e-08, + 0.0000e+00, -9.2248e-07], + [-8.6613e-08, 1.4156e-07, -3.4925e-08, ..., 2.8452e-07, + 0.0000e+00, 3.4925e-08], + [ 6.7055e-08, 3.9488e-06, -1.1642e-07, ..., -8.6613e-08, + 0.0000e+00, 1.2405e-06]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0263, -0.0249, -0.0381, 0.0036, 0.0054, 0.0352, 0.0219, -0.0143, + 0.0030, -0.0097], device='cuda:0'), grad: tensor([ 2.0908e-07, -3.9488e-06, 1.5870e-06, 2.5257e-06, 6.2073e-07, + -3.2596e-06, 7.5856e-07, -9.6709e-06, -2.5611e-07, 1.1444e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 217---------------------------------------------------- +epoch 217, time 230.24, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4936 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.13 lr 0.00010000 +Epoch 219, weight, value: tensor([[ 0.0199, -0.0772, -0.0526, ..., -0.1903, -0.0844, -0.1533], + [-0.0494, -0.0511, -0.0219, ..., 0.0366, 0.0249, -0.1183], + [ 0.0317, -0.1219, -0.1473, ..., 0.0166, -0.1511, 0.0753], + ..., + [-0.0122, -0.0259, -0.1590, ..., -0.1348, 0.0205, 0.1321], + [ 0.0744, -0.1197, -0.1009, ..., -0.1366, -0.0850, -0.1056], + [-0.1721, 0.0195, 0.1095, ..., 0.0820, -0.0752, -0.1483]], + device='cuda:0'), grad: tensor([[ 4.8755e-07, 1.1269e-07, 3.9116e-08, ..., 9.7789e-09, + 0.0000e+00, 3.3639e-06], + [ 1.2992e-07, 1.3039e-07, 1.6298e-08, ..., 2.0443e-07, + 0.0000e+00, 7.6974e-07], + [ 1.3178e-07, 4.5169e-08, 1.2480e-07, ..., 8.8476e-09, + 0.0000e+00, 6.3097e-07], + ..., + [-1.9390e-06, -3.3574e-07, 2.4680e-07, ..., 1.3132e-07, + 0.0000e+00, -1.0781e-05], + [ 1.4715e-07, 9.4064e-08, 7.2177e-08, ..., 4.3772e-08, + 0.0000e+00, 8.4285e-07], + [ 6.9058e-07, -1.1642e-08, 4.1351e-06, ..., 2.5406e-06, + 0.0000e+00, 3.6340e-06]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0263, -0.0245, -0.0381, 0.0037, 0.0054, 0.0352, 0.0219, -0.0144, + 0.0030, -0.0101], device='cuda:0'), grad: tensor([ 7.7412e-06, 1.4873e-06, 2.0042e-06, 2.2817e-07, -5.4836e-06, + 2.0415e-06, 3.3295e-07, -2.5034e-05, 2.3320e-06, 1.4357e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 214.28, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.4895 re_mapping 0.0048 re_causal 0.0146 /// teacc 99.08 lr 0.00010000 +Epoch 220, weight, value: tensor([[ 0.0199, -0.0776, -0.0528, ..., -0.1911, -0.0844, -0.1540], + [-0.0500, -0.0513, -0.0217, ..., 0.0372, 0.0249, -0.1184], + [ 0.0321, -0.1223, -0.1479, ..., 0.0169, -0.1511, 0.0755], + ..., + [-0.0100, -0.0252, -0.1585, ..., -0.1341, 0.0205, 0.1327], + [ 0.0746, -0.1197, -0.1016, ..., -0.1376, -0.0850, -0.1061], + [-0.1733, 0.0189, 0.1092, ..., 0.0819, -0.0753, -0.1487]], + device='cuda:0'), grad: tensor([[-1.3690e-07, 3.1665e-08, 2.7474e-08, ..., 5.0757e-08, + 4.6566e-09, 8.2422e-08], + [ 3.5856e-08, 1.0850e-07, 3.7719e-08, ..., 1.5367e-08, + 1.4435e-08, 1.8487e-07], + [-2.5425e-07, 4.8894e-08, 4.8894e-08, ..., 3.7719e-08, + 8.3819e-09, -1.0757e-07], + ..., + [ 4.5169e-08, 4.7356e-05, 1.5358e-06, ..., 2.0359e-06, + 7.3947e-06, 6.5684e-05], + [-4.9360e-07, 2.4680e-07, -6.2250e-06, ..., -1.0384e-07, + 3.6322e-08, 3.6508e-07], + [ 2.8126e-07, 4.1956e-07, -1.6764e-06, ..., -2.2184e-06, + 4.6566e-08, 5.6345e-07]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0263, -0.0245, -0.0380, 0.0035, 0.0053, 0.0352, 0.0220, -0.0139, + 0.0026, -0.0105], device='cuda:0'), grad: tensor([ 2.9756e-07, 5.6345e-07, -1.5851e-06, -1.3494e-04, 2.4121e-07, + 1.7518e-06, 1.7643e-05, 1.3745e-04, -2.0340e-05, -1.0403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 214.00, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4987 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.00 lr 0.00010000 +Epoch 221, weight, value: tensor([[ 0.0200, -0.0777, -0.0529, ..., -0.1916, -0.0844, -0.1545], + [-0.0502, -0.0514, -0.0219, ..., 0.0369, 0.0248, -0.1184], + [ 0.0319, -0.1227, -0.1488, ..., 0.0167, -0.1511, 0.0756], + ..., + [-0.0097, -0.0252, -0.1585, ..., -0.1342, 0.0205, 0.1328], + [ 0.0747, -0.1201, -0.1019, ..., -0.1381, -0.0855, -0.1064], + [-0.1738, 0.0188, 0.1095, ..., 0.0820, -0.0756, -0.1489]], + device='cuda:0'), grad: tensor([[-3.3434e-06, 6.5193e-09, 3.4925e-08, ..., 9.2667e-08, + 0.0000e+00, 1.0245e-08], + [ 2.4214e-08, 1.4435e-08, 6.7987e-08, ..., 6.1467e-08, + 0.0000e+00, 2.1886e-08], + [ 6.7521e-08, 1.8626e-08, 7.0781e-08, ..., 1.2200e-07, + 0.0000e+00, -1.7229e-08], + ..., + [ 5.4482e-08, 1.5367e-08, 5.1223e-08, ..., 8.5682e-08, + 0.0000e+00, 3.8184e-08], + [ 5.0757e-08, 2.5146e-08, 9.6858e-08, ..., 2.0256e-07, + 0.0000e+00, 3.1199e-08], + [ 1.1921e-07, -6.3330e-08, -4.8429e-08, ..., 1.6298e-08, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0262, -0.0246, -0.0383, 0.0035, 0.0053, 0.0353, 0.0220, -0.0139, + 0.0024, -0.0105], device='cuda:0'), grad: tensor([-9.1791e-06, 7.3574e-07, 6.6683e-07, 1.8347e-07, -9.7416e-07, + 1.3579e-06, 8.4862e-06, 5.8906e-07, -3.5595e-06, 1.6876e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.04, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.5030 re_mapping 0.0051 re_causal 0.0146 /// teacc 99.10 lr 0.00010000 +Epoch 222, weight, value: tensor([[ 0.0200, -0.0779, -0.0533, ..., -0.1921, -0.0845, -0.1550], + [-0.0504, -0.0516, -0.0218, ..., 0.0360, 0.0248, -0.1184], + [ 0.0319, -0.1229, -0.1512, ..., 0.0166, -0.1511, 0.0743], + ..., + [-0.0101, -0.0252, -0.1587, ..., -0.1344, 0.0208, 0.1329], + [ 0.0747, -0.1213, -0.1021, ..., -0.1387, -0.0856, -0.1047], + [-0.1745, 0.0188, 0.1094, ..., 0.0819, -0.0770, -0.1492]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 4.1910e-09, 6.2399e-08, ..., 2.1420e-08, + 0.0000e+00, 1.5367e-08], + [ 1.6764e-08, 2.1188e-07, -2.0117e-07, ..., 3.3854e-07, + 0.0000e+00, 1.8114e-06], + [ 2.0023e-08, 1.7881e-07, 2.4214e-08, ..., -8.0699e-07, + 0.0000e+00, -7.3481e-07], + ..., + [-1.8161e-08, -5.6159e-07, 4.0838e-07, ..., 4.4052e-07, + 0.0000e+00, -1.8664e-06], + [-2.8592e-07, 1.9185e-07, -1.7416e-07, ..., 8.3819e-08, + 0.0000e+00, 6.5425e-07], + [ 7.6368e-08, 1.5786e-07, 1.8626e-07, ..., 7.7765e-08, + 0.0000e+00, 2.3982e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0262, -0.0247, -0.0397, 0.0036, 0.0055, 0.0354, 0.0218, -0.0139, + 0.0031, -0.0106], device='cuda:0'), grad: tensor([ 3.1292e-07, 3.3285e-06, -2.1793e-06, -6.3237e-07, -1.4156e-06, + 2.4913e-07, 7.4971e-07, -2.1979e-06, 4.6985e-07, 1.3132e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.07, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4886 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.02 lr 0.00010000 +Epoch 223, weight, value: tensor([[ 0.0200, -0.0782, -0.0535, ..., -0.1926, -0.0845, -0.1555], + [-0.0508, -0.0517, -0.0219, ..., 0.0387, 0.0247, -0.1176], + [ 0.0318, -0.1233, -0.1521, ..., 0.0135, -0.1511, 0.0716], + ..., + [-0.0101, -0.0253, -0.1589, ..., -0.1346, 0.0209, 0.1329], + [ 0.0749, -0.1216, -0.1023, ..., -0.1391, -0.0857, -0.1048], + [-0.1749, 0.0192, 0.1095, ..., 0.0819, -0.0771, -0.1492]], + device='cuda:0'), grad: tensor([[ 9.6392e-08, 3.7998e-07, 3.3434e-07, ..., 8.0653e-07, + 0.0000e+00, 8.9966e-07], + [ 2.7008e-08, 1.8347e-07, -3.1758e-06, ..., -5.6103e-06, + 0.0000e+00, 7.7393e-07], + [ 1.7229e-08, 3.2503e-07, 1.1548e-07, ..., -4.0978e-08, + 0.0000e+00, -7.0315e-07], + ..., + [ 1.1176e-08, 3.4645e-07, 1.3700e-06, ..., 1.8803e-06, + 0.0000e+00, 3.6228e-07], + [ 3.9116e-08, 1.6810e-07, 2.8592e-07, ..., 6.0396e-07, + 0.0000e+00, 6.7055e-07], + [ 1.8161e-08, -4.9919e-07, -5.5181e-07, ..., -2.9057e-07, + 0.0000e+00, 1.5879e-07]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0262, -0.0233, -0.0428, 0.0033, 0.0056, 0.0357, 0.0218, -0.0139, + 0.0032, -0.0106], device='cuda:0'), grad: tensor([ 4.0084e-06, -1.1452e-05, -2.1104e-06, -4.6417e-06, 6.8359e-07, + 3.6526e-06, 1.3504e-06, 4.8205e-06, 2.8890e-06, 7.4599e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 214.22, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5121 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.10 lr 0.00010000 +Epoch 224, weight, value: tensor([[ 0.0200, -0.0798, -0.0541, ..., -0.1937, -0.0845, -0.1569], + [-0.0516, -0.0520, -0.0208, ..., 0.0392, 0.0247, -0.1178], + [ 0.0315, -0.1238, -0.1527, ..., 0.0136, -0.1511, 0.0725], + ..., + [-0.0097, -0.0248, -0.1602, ..., -0.1355, 0.0210, 0.1335], + [ 0.0761, -0.1204, -0.1012, ..., -0.1396, -0.0857, -0.1051], + [-0.1775, 0.0189, 0.1093, ..., 0.0819, -0.0771, -0.1495]], + device='cuda:0'), grad: tensor([[-2.1607e-07, 5.1223e-09, 7.3574e-08, ..., 7.3574e-08, + 0.0000e+00, 5.1223e-08], + [ 2.1886e-08, 2.4680e-08, 6.4727e-08, ..., 7.6368e-08, + 4.6566e-10, 9.4064e-08], + [ 8.8476e-09, 7.9162e-09, -3.2922e-07, ..., 2.2817e-08, + 0.0000e+00, -1.2089e-06], + ..., + [ 1.0245e-08, -1.5832e-08, 7.2923e-07, ..., 8.6892e-07, + -4.6566e-10, 2.0955e-08], + [ 2.5611e-08, 7.9162e-09, 1.5767e-06, ..., 2.0191e-06, + 0.0000e+00, 1.0431e-07], + [ 1.4249e-07, 1.5367e-08, -3.5334e-06, ..., -4.2319e-06, + 0.0000e+00, -4.2375e-08]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0264, -0.0230, -0.0425, 0.0031, 0.0057, 0.0355, 0.0218, -0.0141, + 0.0044, -0.0110], device='cuda:0'), grad: tensor([-4.9639e-07, 5.3924e-07, -2.0862e-06, -1.0524e-06, 3.2857e-06, + -2.0396e-07, 2.9877e-06, 2.4326e-06, 5.4240e-06, -1.0826e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 214.08, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.5000 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.02 lr 0.00010000 +Epoch 225, weight, value: tensor([[ 0.0201, -0.0800, -0.0542, ..., -0.1945, -0.0845, -0.1565], + [-0.0518, -0.0521, -0.0207, ..., 0.0392, 0.0247, -0.1180], + [ 0.0314, -0.1242, -0.1531, ..., 0.0136, -0.1512, 0.0725], + ..., + [-0.0102, -0.0248, -0.1604, ..., -0.1354, 0.0210, 0.1338], + [ 0.0761, -0.1209, -0.1015, ..., -0.1406, -0.0857, -0.1055], + [-0.1788, 0.0190, 0.1093, ..., 0.0818, -0.0771, -0.1497]], + device='cuda:0'), grad: tensor([[ 6.5658e-08, 2.7940e-09, 6.5193e-08, ..., 5.8208e-08, + 0.0000e+00, 5.9605e-08], + [ 1.8161e-08, 8.2422e-08, 2.3749e-06, ..., 1.8338e-06, + 0.0000e+00, 2.1700e-07], + [-5.1875e-07, 5.5879e-09, 4.0513e-08, ..., -1.8114e-07, + 0.0000e+00, -6.7288e-07], + ..., + [ 1.4901e-08, -5.3551e-08, 2.0396e-07, ..., 2.4354e-07, + 0.0000e+00, -2.1514e-07], + [ 1.0096e-06, 8.1025e-08, 1.5767e-06, ..., 4.3362e-06, + 0.0000e+00, 5.1828e-07], + [ 2.0256e-07, 1.2340e-07, 8.6650e-06, ..., 6.9737e-06, + 0.0000e+00, 1.7136e-07]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0262, -0.0230, -0.0426, 0.0031, 0.0057, 0.0356, 0.0217, -0.0140, + 0.0041, -0.0111], device='cuda:0'), grad: tensor([ 3.9721e-07, 5.6364e-06, -2.3544e-06, 2.3358e-06, -2.7567e-05, + -9.1046e-06, 2.0722e-07, 2.7334e-07, 9.7901e-06, 2.0355e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.05, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.4997 re_mapping 0.0047 re_causal 0.0143 /// teacc 99.02 lr 0.00010000 +Epoch 226, weight, value: tensor([[ 0.0202, -0.0803, -0.0549, ..., -0.1945, -0.0845, -0.1565], + [-0.0521, -0.0521, -0.0204, ..., 0.0395, 0.0247, -0.1181], + [ 0.0311, -0.1247, -0.1543, ..., 0.0136, -0.1512, 0.0719], + ..., + [-0.0108, -0.0250, -0.1607, ..., -0.1360, 0.0211, 0.1342], + [ 0.0752, -0.1212, -0.1024, ..., -0.1432, -0.0858, -0.1057], + [-0.1804, 0.0194, 0.1092, ..., 0.0817, -0.0771, -0.1501]], + device='cuda:0'), grad: tensor([[-1.5777e-06, 9.3132e-10, 1.8626e-09, ..., 1.3039e-08, + 0.0000e+00, 5.1223e-09], + [ 9.4064e-08, 1.3039e-08, -1.7416e-07, ..., -2.3190e-07, + 0.0000e+00, 7.1246e-08], + [ 1.3597e-07, 5.1223e-09, 8.3819e-09, ..., 7.9162e-09, + 0.0000e+00, -5.1223e-09], + ..., + [ 5.9139e-08, 8.9407e-08, 1.1642e-07, ..., 2.1514e-07, + 0.0000e+00, 5.8068e-07], + [-1.3039e-08, 5.0291e-08, 7.5437e-08, ..., 7.4506e-08, + 0.0000e+00, 3.3341e-07], + [ 4.0932e-07, -1.9092e-08, -7.0315e-08, ..., -6.0070e-08, + 0.0000e+00, 2.9802e-08]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0261, -0.0229, -0.0427, 0.0031, 0.0056, 0.0360, 0.0218, -0.0140, + 0.0032, -0.0114], device='cuda:0'), grad: tensor([-5.9418e-06, -2.7474e-07, 5.9605e-07, -1.6214e-06, 1.6158e-07, + 1.1614e-06, 1.8822e-06, 1.7211e-06, 5.4110e-07, 1.7807e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 214.19, cls_loss 0.0013 cls_loss_mapping 0.0033 cls_loss_causal 0.4750 re_mapping 0.0051 re_causal 0.0147 /// teacc 99.01 lr 0.00010000 +Epoch 227, weight, value: tensor([[ 0.0204, -0.0806, -0.0551, ..., -0.1949, -0.0849, -0.1565], + [-0.0524, -0.0523, -0.0206, ..., 0.0394, 0.0247, -0.1184], + [ 0.0310, -0.1260, -0.1575, ..., 0.0136, -0.1515, 0.0719], + ..., + [-0.0110, -0.0243, -0.1603, ..., -0.1359, 0.0215, 0.1348], + [ 0.0749, -0.1215, -0.1026, ..., -0.1438, -0.0863, -0.1063], + [-0.1815, 0.0185, 0.1095, ..., 0.0820, -0.0788, -0.1510]], + device='cuda:0'), grad: tensor([[-5.5041e-07, 3.9116e-08, 3.9041e-05, ..., 3.0547e-06, + 0.0000e+00, 1.1176e-08], + [ 3.9116e-08, 3.7253e-08, -5.1782e-07, ..., -5.3234e-06, + 0.0000e+00, 7.0781e-08], + [ 7.6368e-08, 5.1223e-08, 5.0012e-07, ..., 7.0501e-07, + 0.0000e+00, 6.5193e-09], + ..., + [ 1.9558e-08, 8.8476e-08, 9.0338e-07, ..., 1.5227e-06, + 0.0000e+00, 1.5367e-07], + [-1.4156e-07, 9.4995e-08, 3.9209e-07, ..., 1.8720e-06, + 0.0000e+00, 5.4017e-08], + [ 4.8541e-06, 1.5587e-05, -2.6345e-05, ..., 6.9499e-05, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0258, -0.0230, -0.0430, 0.0031, 0.0054, 0.0361, 0.0216, -0.0136, + 0.0028, -0.0115], device='cuda:0'), grad: tensor([ 9.2983e-05, -9.3132e-06, 2.5984e-06, 2.5973e-05, 4.5747e-06, + -1.1778e-04, 1.5525e-06, 4.7684e-06, 5.2154e-08, -5.6177e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 213.96, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5015 re_mapping 0.0046 re_causal 0.0143 /// teacc 99.13 lr 0.00010000 +Epoch 228, weight, value: tensor([[ 0.0205, -0.0807, -0.0556, ..., -0.1956, -0.0849, -0.1561], + [-0.0530, -0.0524, -0.0207, ..., 0.0394, 0.0247, -0.1184], + [ 0.0309, -0.1265, -0.1586, ..., 0.0136, -0.1515, 0.0719], + ..., + [-0.0113, -0.0242, -0.1605, ..., -0.1362, 0.0216, 0.1349], + [ 0.0748, -0.1221, -0.1030, ..., -0.1443, -0.0864, -0.1067], + [-0.1821, 0.0182, 0.1096, ..., 0.0820, -0.0791, -0.1513]], + device='cuda:0'), grad: tensor([[-3.4459e-08, 6.0536e-08, 3.0734e-08, ..., 9.1270e-08, + 0.0000e+00, 1.4342e-07], + [ 4.6566e-09, 4.4145e-07, -5.6438e-07, ..., -3.7532e-07, + 9.3132e-10, 7.0967e-07], + [ 1.1176e-08, 6.3330e-08, 3.4459e-08, ..., 5.5879e-08, + 9.3132e-10, -8.3074e-07], + ..., + [ 2.7940e-09, -6.1374e-07, 2.9150e-07, ..., 2.7288e-07, + -2.7940e-09, -4.9639e-07], + [-3.6322e-08, 5.1502e-07, 1.3225e-07, ..., 5.6531e-07, + 9.3132e-10, 1.1791e-06], + [ 2.7940e-08, 2.1439e-06, -3.9116e-08, ..., 3.1888e-06, + 9.3132e-10, 4.9584e-06]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0256, -0.0230, -0.0431, 0.0032, 0.0054, 0.0361, 0.0215, -0.0136, + 0.0024, -0.0116], device='cuda:0'), grad: tensor([ 4.5635e-07, -2.0713e-06, -7.0874e-07, -2.1547e-05, 9.9745e-07, + 4.0010e-06, 1.6671e-07, -4.7497e-08, 3.5688e-06, 1.5184e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 214.13, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.5021 re_mapping 0.0046 re_causal 0.0141 /// teacc 99.08 lr 0.00010000 +Epoch 229, weight, value: tensor([[ 0.0206, -0.0809, -0.0556, ..., -0.1961, -0.0849, -0.1562], + [-0.0538, -0.0527, -0.0202, ..., 0.0396, 0.0247, -0.1184], + [ 0.0306, -0.1268, -0.1600, ..., 0.0135, -0.1524, 0.0718], + ..., + [-0.0116, -0.0243, -0.1616, ..., -0.1382, 0.0217, 0.1349], + [ 0.0745, -0.1230, -0.1040, ..., -0.1448, -0.0865, -0.1073], + [-0.1827, 0.0176, 0.1096, ..., 0.0819, -0.0791, -0.1517]], + device='cuda:0'), grad: tensor([[-8.3540e-07, 1.8626e-09, 3.1665e-08, ..., 5.5879e-09, + 0.0000e+00, 4.6566e-09], + [ 7.6368e-08, 7.9162e-08, -1.9558e-08, ..., -5.0291e-08, + 0.0000e+00, 3.3714e-07], + [ 7.4506e-08, 1.5832e-08, 4.0978e-08, ..., 9.3132e-10, + 0.0000e+00, 6.3330e-08], + ..., + [ 1.2573e-07, -8.7544e-08, 5.4017e-08, ..., 7.7300e-08, + 0.0000e+00, -4.2934e-07], + [ 2.0396e-07, 1.2107e-08, 5.5879e-08, ..., 1.1083e-07, + 0.0000e+00, 2.6077e-08], + [ 2.2538e-07, -1.0245e-08, -7.7300e-08, ..., 4.5635e-08, + 0.0000e+00, 4.8429e-08]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0254, -0.0227, -0.0433, 0.0035, 0.0061, 0.0362, 0.0217, -0.0144, + 0.0017, -0.0121], device='cuda:0'), grad: tensor([-3.3639e-06, 1.0533e-06, 5.2992e-07, 1.0300e-06, 4.1816e-07, + -8.0094e-07, -3.8929e-07, -4.2375e-07, 9.7975e-07, 9.5926e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 213.92, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5019 re_mapping 0.0051 re_causal 0.0142 /// teacc 99.05 lr 0.00010000 +Epoch 230, weight, value: tensor([[ 0.0206, -0.0845, -0.0565, ..., -0.1976, -0.0849, -0.1600], + [-0.0547, -0.0530, -0.0206, ..., 0.0395, 0.0247, -0.1185], + [ 0.0306, -0.1278, -0.1615, ..., 0.0135, -0.1554, 0.0706], + ..., + [-0.0116, -0.0242, -0.1627, ..., -0.1395, 0.0224, 0.1361], + [ 0.0745, -0.1257, -0.1049, ..., -0.1457, -0.0870, -0.1082], + [-0.1829, 0.0198, 0.1099, ..., 0.0824, -0.0795, -0.1521]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 2.4214e-08, 3.6322e-08, ..., 1.6764e-08, + 0.0000e+00, 6.4261e-08], + [ 2.4214e-08, 3.4459e-08, 8.3819e-09, ..., 8.3819e-09, + 0.0000e+00, 7.8231e-08], + [ 8.3819e-09, 1.1176e-08, 1.8626e-09, ..., -4.6566e-09, + 0.0000e+00, -4.5635e-08], + ..., + [-4.2841e-08, -1.8626e-09, 4.6566e-08, ..., 7.4506e-08, + 0.0000e+00, -8.1956e-08], + [-5.0291e-08, 2.7940e-08, 8.8476e-08, ..., 1.4901e-07, + 0.0000e+00, 5.6811e-08], + [ 1.5832e-08, -3.4459e-08, -6.6124e-07, ..., -7.9162e-07, + 0.0000e+00, 2.4214e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0256, -0.0228, -0.0439, 0.0036, 0.0063, 0.0360, 0.0219, -0.0141, + 0.0010, -0.0117], device='cuda:0'), grad: tensor([ 2.6263e-07, 2.8685e-07, -4.1910e-08, -2.7008e-06, 8.2236e-07, + 2.6189e-06, 3.4459e-08, -1.1828e-07, 9.3132e-08, -1.2666e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 214.16, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.4752 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.06 lr 0.00010000 +Epoch 231, weight, value: tensor([[ 0.0206, -0.0847, -0.0570, ..., -0.1984, -0.0850, -0.1602], + [-0.0551, -0.0532, -0.0206, ..., 0.0396, 0.0247, -0.1190], + [ 0.0306, -0.1287, -0.1631, ..., 0.0135, -0.1554, 0.0707], + ..., + [-0.0118, -0.0243, -0.1632, ..., -0.1404, 0.0225, 0.1365], + [ 0.0746, -0.1258, -0.1045, ..., -0.1458, -0.0871, -0.1088], + [-0.1833, 0.0189, 0.1096, ..., 0.0823, -0.0796, -0.1522]], + device='cuda:0'), grad: tensor([[-1.3970e-07, 3.6322e-08, 9.3132e-08, ..., 7.4506e-08, + 0.0000e+00, 6.5193e-09], + [ 2.2352e-08, 3.0734e-08, 2.7940e-08, ..., -3.2596e-08, + 0.0000e+00, 5.6811e-08], + [ 6.5193e-09, 2.5146e-08, 1.8626e-08, ..., 1.5832e-08, + 0.0000e+00, -1.4901e-08], + ..., + [ 9.3132e-09, -5.6811e-08, 6.2399e-08, ..., 8.3819e-08, + 0.0000e+00, -2.2724e-07], + [-1.8710e-06, 1.9558e-08, -2.2631e-06, ..., 4.1630e-07, + 0.0000e+00, 9.1270e-08], + [ 5.7742e-08, -2.0862e-07, -4.5355e-07, ..., -2.4773e-07, + 0.0000e+00, 1.6764e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0257, -0.0230, -0.0438, 0.0037, 0.0069, 0.0360, 0.0219, -0.0141, + 0.0012, -0.0121], device='cuda:0'), grad: tensor([-2.4494e-07, 1.7229e-07, 7.0781e-08, 1.5646e-07, 4.6752e-07, + 1.4305e-06, 9.3207e-06, -3.0641e-07, -1.0595e-05, -5.1875e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 214.00, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5040 re_mapping 0.0047 re_causal 0.0137 /// teacc 99.02 lr 0.00010000 +Epoch 232, weight, value: tensor([[ 0.0209, -0.0850, -0.0573, ..., -0.2000, -0.0852, -0.1609], + [-0.0555, -0.0533, -0.0206, ..., 0.0399, 0.0246, -0.1191], + [ 0.0304, -0.1305, -0.1658, ..., 0.0135, -0.1554, 0.0709], + ..., + [-0.0119, -0.0242, -0.1636, ..., -0.1410, 0.0226, 0.1366], + [ 0.0746, -0.1271, -0.1053, ..., -0.1469, -0.0857, -0.1093], + [-0.1837, 0.0189, 0.1095, ..., 0.0819, -0.0801, -0.1524]], + device='cuda:0'), grad: tensor([[-9.7677e-06, 3.7253e-09, 4.3772e-08, ..., -1.2657e-06, + 0.0000e+00, 7.4506e-09], + [ 1.0245e-07, 5.3085e-08, 3.8184e-08, ..., 3.3528e-08, + 0.0000e+00, 2.5239e-07], + [ 4.3493e-07, 2.1048e-07, 5.1223e-08, ..., 7.8231e-08, + 0.0000e+00, 1.1344e-06], + ..., + [ 2.7008e-08, -2.8405e-07, 1.0803e-07, ..., 9.2201e-08, + 0.0000e+00, -1.5218e-06], + [ 1.6671e-07, 4.7497e-08, 1.6578e-07, ..., 1.4156e-07, + 0.0000e+00, 1.7881e-07], + [ 2.6822e-07, -5.5879e-09, 4.8522e-07, ..., 2.6170e-07, + 0.0000e+00, 2.7008e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0253, -0.0230, -0.0438, 0.0038, 0.0074, 0.0356, 0.0219, -0.0142, + 0.0011, -0.0125], device='cuda:0'), grad: tensor([-2.7806e-05, 7.7579e-07, 3.1497e-06, -3.2969e-07, -6.8173e-06, + 1.4640e-06, 2.8536e-05, -2.0824e-06, 1.1390e-06, 1.9334e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 213.92, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4833 re_mapping 0.0047 re_causal 0.0139 /// teacc 99.05 lr 0.00010000 +Epoch 233, weight, value: tensor([[ 0.0211, -0.0851, -0.0576, ..., -0.2002, -0.0855, -0.1610], + [-0.0559, -0.0526, -0.0204, ..., 0.0405, 0.0246, -0.1189], + [ 0.0291, -0.1337, -0.1684, ..., 0.0135, -0.1554, 0.0708], + ..., + [-0.0121, -0.0250, -0.1639, ..., -0.1424, 0.0226, 0.1365], + [ 0.0750, -0.1246, -0.1054, ..., -0.1471, -0.0858, -0.1090], + [-0.1843, 0.0205, 0.1109, ..., 0.0831, -0.0802, -0.1525]], + device='cuda:0'), grad: tensor([[-5.4576e-07, 1.5832e-08, 1.0058e-07, ..., 8.9407e-08, + 0.0000e+00, 6.5193e-09], + [ 3.9116e-08, 1.6205e-06, 9.1195e-06, ..., 8.8513e-06, + 0.0000e+00, 1.2200e-07], + [ 2.3283e-08, 1.4901e-08, 5.8673e-08, ..., 5.6811e-08, + 0.0000e+00, -4.8429e-08], + ..., + [ 1.3039e-08, 1.7183e-06, 1.2055e-05, ..., 1.1571e-05, + -9.3132e-10, -1.7229e-07], + [-1.0990e-07, 4.3772e-08, 1.6205e-07, ..., 1.5739e-07, + 0.0000e+00, 4.5635e-08], + [ 1.1176e-07, -9.5814e-06, -5.9873e-05, ..., -5.7697e-05, + 0.0000e+00, 2.8871e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0250, -0.0226, -0.0441, 0.0037, 0.0067, 0.0354, 0.0220, -0.0145, + 0.0013, -0.0118], device='cuda:0'), grad: tensor([-1.6764e-06, 1.6376e-05, 1.1828e-07, 2.5705e-07, 6.6578e-05, + 2.2445e-07, 1.4473e-06, 2.0489e-05, -3.6880e-07, -1.0353e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 213.95, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.5007 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.01 lr 0.00010000 +Epoch 234, weight, value: tensor([[ 0.0211, -0.0851, -0.0578, ..., -0.2008, -0.0855, -0.1611], + [-0.0562, -0.0526, -0.0200, ..., 0.0377, 0.0246, -0.1212], + [ 0.0294, -0.1339, -0.1716, ..., 0.0165, -0.1555, 0.0740], + ..., + [-0.0125, -0.0249, -0.1644, ..., -0.1432, 0.0227, 0.1369], + [ 0.0747, -0.1248, -0.1057, ..., -0.1478, -0.0858, -0.1098], + [-0.1847, 0.0209, 0.1111, ..., 0.0833, -0.0806, -0.1527]], + device='cuda:0'), grad: tensor([[-1.4715e-07, 6.5193e-08, 1.6484e-07, ..., 1.8813e-07, + 0.0000e+00, 5.6811e-08], + [ 2.0489e-08, 1.3318e-07, -2.7008e-08, ..., 8.9407e-08, + 9.3132e-10, 1.8347e-07], + [-2.7940e-09, 3.4459e-08, 4.4703e-08, ..., 5.6811e-08, + 9.3132e-10, -4.6473e-07], + ..., + [ 9.3132e-09, 1.7323e-07, 9.1828e-07, ..., 9.9372e-07, + -9.3132e-10, 6.3330e-08], + [ 1.3504e-07, 9.0338e-08, 4.8243e-07, ..., 6.7893e-07, + 2.7940e-09, 1.1548e-07], + [ 3.1665e-08, -7.3500e-06, -3.7313e-05, ..., -3.0756e-05, + 0.0000e+00, 5.4948e-08]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0249, -0.0252, -0.0410, 0.0036, 0.0066, 0.0359, 0.0216, -0.0145, + 0.0010, -0.0118], device='cuda:0'), grad: tensor([-4.0047e-08, -8.8196e-07, -7.4692e-07, 1.5777e-06, 4.9293e-05, + -5.2433e-07, -4.7497e-08, 2.9244e-06, 2.1346e-06, -5.3793e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.07, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4976 re_mapping 0.0045 re_causal 0.0139 /// teacc 98.99 lr 0.00010000 +Epoch 235, weight, value: tensor([[ 0.0213, -0.0865, -0.0571, ..., -0.2017, -0.0855, -0.1630], + [-0.0564, -0.0526, -0.0200, ..., 0.0377, 0.0247, -0.1213], + [ 0.0297, -0.1341, -0.1719, ..., 0.0165, -0.1551, 0.0745], + ..., + [-0.0130, -0.0250, -0.1648, ..., -0.1437, 0.0226, 0.1367], + [ 0.0744, -0.1250, -0.1067, ..., -0.1485, -0.0864, -0.1107], + [-0.1862, 0.0210, 0.1113, ..., 0.0834, -0.0807, -0.1528]], + device='cuda:0'), grad: tensor([[-1.3616e-06, 9.3132e-10, -3.4925e-07, ..., 1.2107e-08, + 0.0000e+00, 5.5879e-09], + [ 8.2329e-07, 2.7940e-09, 2.5798e-07, ..., 2.7940e-08, + 0.0000e+00, 2.6077e-08], + [ 4.1910e-08, 1.8626e-09, 1.7695e-08, ..., 1.8626e-09, + 0.0000e+00, -4.6566e-09], + ..., + [ 1.3970e-08, 9.3132e-09, 2.0582e-07, ..., 1.8626e-07, + 0.0000e+00, 2.7008e-08], + [ 7.4506e-08, 1.0245e-08, 1.3504e-07, ..., 9.9652e-08, + 0.0000e+00, 8.6613e-08], + [ 5.5879e-08, -6.2399e-08, -1.7881e-06, ..., -1.7192e-06, + 0.0000e+00, 2.0489e-08]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0249, -0.0252, -0.0409, 0.0046, 0.0066, 0.0353, 0.0214, -0.0148, + 0.0002, -0.0118], device='cuda:0'), grad: tensor([-4.5635e-06, 2.8946e-06, 1.5460e-07, -7.1432e-07, 2.3711e-06, + 6.6310e-07, 1.0086e-06, 4.7497e-07, 3.8557e-07, -2.6785e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.35, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5038 re_mapping 0.0044 re_causal 0.0142 /// teacc 98.99 lr 0.00010000 +Epoch 236, weight, value: tensor([[ 0.0213, -0.0867, -0.0590, ..., -0.2022, -0.0855, -0.1633], + [-0.0569, -0.0505, -0.0195, ..., 0.0378, 0.0273, -0.1208], + [ 0.0297, -0.1343, -0.1724, ..., 0.0165, -0.1551, 0.0747], + ..., + [-0.0135, -0.0266, -0.1656, ..., -0.1449, 0.0201, 0.1361], + [ 0.0744, -0.1253, -0.1072, ..., -0.1488, -0.0867, -0.1110], + [-0.1871, 0.0208, 0.1115, ..., 0.0835, -0.0807, -0.1532]], + device='cuda:0'), grad: tensor([[-2.6077e-08, 1.8626e-09, 4.1910e-08, ..., 8.3819e-09, + 0.0000e+00, 1.1176e-08], + [ 2.7940e-09, 5.0291e-08, -3.2596e-08, ..., -6.4261e-08, + -7.4506e-09, 1.6019e-07], + [ 3.7253e-09, 1.5832e-08, 1.0245e-08, ..., 2.7940e-09, + 0.0000e+00, 6.7987e-08], + ..., + [ 1.8626e-09, 2.2072e-07, 6.3330e-08, ..., 8.1025e-08, + 4.6566e-09, 4.3772e-08], + [ 1.8626e-08, 1.1176e-08, 6.2399e-08, ..., 2.3283e-08, + 0.0000e+00, 4.3772e-08], + [ 2.7008e-08, -1.4901e-08, 7.8231e-08, ..., 3.5390e-08, + 9.3132e-10, 3.6322e-08]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0251, -0.0247, -0.0408, 0.0045, 0.0064, 0.0353, 0.0220, -0.0158, + -0.0002, -0.0117], device='cuda:0'), grad: tensor([-1.5832e-08, 1.4994e-07, 1.4249e-07, -5.8580e-07, -2.4959e-07, + -5.2154e-08, -2.1607e-07, 3.2224e-07, 2.2724e-07, 2.7753e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 213.93, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4988 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.07 lr 0.00010000 +Epoch 237, weight, value: tensor([[ 0.0213, -0.0868, -0.0590, ..., -0.2027, -0.0855, -0.1636], + [-0.0572, -0.0505, -0.0191, ..., 0.0379, 0.0274, -0.1209], + [ 0.0315, -0.1349, -0.1727, ..., 0.0165, -0.1536, 0.0750], + ..., + [-0.0141, -0.0264, -0.1661, ..., -0.1454, 0.0201, 0.1364], + [ 0.0739, -0.1254, -0.1074, ..., -0.1493, -0.0891, -0.1127], + [-0.1887, 0.0208, 0.1114, ..., 0.0834, -0.0808, -0.1533]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 9.3132e-10, 4.6566e-08, ..., 5.5879e-09, + 0.0000e+00, 5.5879e-09], + [ 1.2107e-08, 1.4901e-08, -9.3132e-10, ..., -6.5193e-09, + -0.0000e+00, 4.6566e-08], + [ 7.4506e-09, 3.7253e-09, 8.3819e-09, ..., 1.8626e-09, + 0.0000e+00, -1.5181e-07], + ..., + [ 7.4506e-09, -1.8626e-09, 7.6368e-08, ..., 3.2596e-08, + 0.0000e+00, 7.5437e-08], + [-3.4459e-08, 1.0245e-08, 5.7742e-08, ..., 9.3132e-09, + 0.0000e+00, 3.9116e-08], + [ 1.8626e-08, -1.5832e-08, 1.5832e-08, ..., 5.5879e-08, + 0.0000e+00, 8.3819e-09]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0250, -0.0247, -0.0407, 0.0044, 0.0066, 0.0352, 0.0221, -0.0159, + -0.0009, -0.0120], device='cuda:0'), grad: tensor([ 1.1921e-07, 1.0245e-07, -1.4808e-07, -7.5437e-08, -9.3132e-08, + 2.2724e-07, -2.7008e-07, 3.0454e-07, -4.1630e-07, 2.4308e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 214.02, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.5086 re_mapping 0.0044 re_causal 0.0138 /// teacc 99.02 lr 0.00010000 +Epoch 238, weight, value: tensor([[ 0.0213, -0.0873, -0.0611, ..., -0.2038, -0.0864, -0.1638], + [-0.0577, -0.0506, -0.0190, ..., 0.0379, 0.0273, -0.1206], + [ 0.0313, -0.1352, -0.1730, ..., 0.0165, -0.1536, 0.0753], + ..., + [-0.0145, -0.0264, -0.1663, ..., -0.1456, 0.0201, 0.1359], + [ 0.0739, -0.1253, -0.1077, ..., -0.1499, -0.0894, -0.1132], + [-0.1898, 0.0210, 0.1115, ..., 0.0834, -0.0810, -0.1535]], + device='cuda:0'), grad: tensor([[ 5.0291e-08, 1.3970e-08, 4.6566e-08, ..., 2.4214e-08, + 2.7940e-09, 4.0047e-08], + [ 2.5146e-08, 2.7008e-08, -1.0068e-06, ..., -8.4005e-07, + 0.0000e+00, 4.6566e-08], + [ 5.2247e-07, 3.7253e-09, 2.0489e-08, ..., -1.4435e-07, + 0.0000e+00, -3.7439e-07], + ..., + [ 1.3039e-08, 6.3330e-08, 1.4063e-07, ..., 1.4994e-07, + 0.0000e+00, 8.4750e-08], + [-1.7136e-06, 6.2399e-08, -4.2841e-08, ..., 8.1956e-08, + 9.3132e-10, 7.4506e-08], + [ 1.4901e-08, 4.6566e-09, 1.1176e-08, ..., 7.7300e-08, + 0.0000e+00, 2.4121e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0253, -0.0245, -0.0407, 0.0044, 0.0066, 0.0357, 0.0218, -0.0163, + -0.0008, -0.0121], device='cuda:0'), grad: tensor([ 3.3807e-07, -1.3821e-06, 1.2545e-06, -1.8664e-06, 1.3746e-06, + 6.3926e-06, 1.5115e-06, 5.5786e-07, -8.9034e-06, 6.8638e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 213.91, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.4891 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.03 lr 0.00010000 +Epoch 239, weight, value: tensor([[ 0.0213, -0.0878, -0.0619, ..., -0.2056, -0.0866, -0.1641], + [-0.0583, -0.0507, -0.0191, ..., 0.0379, 0.0273, -0.1207], + [ 0.0312, -0.1353, -0.1734, ..., 0.0165, -0.1536, 0.0763], + ..., + [-0.0147, -0.0264, -0.1664, ..., -0.1458, 0.0201, 0.1353], + [ 0.0739, -0.1254, -0.1077, ..., -0.1501, -0.0896, -0.1137], + [-0.1906, 0.0217, 0.1119, ..., 0.0838, -0.0810, -0.1536]], + device='cuda:0'), grad: tensor([[-1.0971e-06, 0.0000e+00, 6.5193e-09, ..., -2.2911e-07, + 0.0000e+00, 9.3132e-10], + [ 2.7847e-07, 4.6566e-09, -1.9595e-06, ..., 8.6613e-08, + 0.0000e+00, 1.1176e-08], + [ 6.2399e-08, 0.0000e+00, 1.5423e-06, ..., 3.3528e-08, + 0.0000e+00, -5.0850e-06], + ..., + [ 9.2201e-08, 9.3132e-10, 3.4925e-07, ..., 3.0734e-08, + 0.0000e+00, 5.0552e-06], + [ 1.3970e-08, 2.7940e-09, 2.6077e-08, ..., 2.1886e-07, + 0.0000e+00, 4.3772e-08], + [ 1.8254e-07, -4.6566e-09, 1.7695e-08, ..., 6.9849e-08, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0254, -0.0245, -0.0405, 0.0043, 0.0063, 0.0358, 0.0221, -0.0166, + -0.0006, -0.0119], device='cuda:0'), grad: tensor([-2.8461e-06, -2.1607e-05, 1.4015e-05, 1.2955e-06, 3.4086e-07, + 4.3586e-07, 3.0566e-06, 1.2435e-05, -9.3132e-06, 2.2184e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.01, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4944 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.03 lr 0.00010000 +Epoch 240, weight, value: tensor([[ 0.0214, -0.0880, -0.0629, ..., -0.2066, -0.0866, -0.1642], + [-0.0585, -0.0509, -0.0202, ..., 0.0377, 0.0273, -0.1209], + [ 0.0311, -0.1357, -0.1735, ..., 0.0165, -0.1536, 0.0764], + ..., + [-0.0152, -0.0260, -0.1666, ..., -0.1461, 0.0201, 0.1356], + [ 0.0737, -0.1258, -0.1086, ..., -0.1511, -0.0896, -0.1141], + [-0.1914, 0.0217, 0.1124, ..., 0.0843, -0.0810, -0.1539]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 0.0000e+00, 3.6322e-08, ..., 1.5832e-08, + 0.0000e+00, 1.2014e-07], + [ 3.7253e-09, 1.8626e-09, -1.1455e-07, ..., -1.1362e-07, + 0.0000e+00, 1.1642e-07], + [-9.5926e-08, 0.0000e+00, 1.8626e-09, ..., -1.8626e-08, + 0.0000e+00, -7.4357e-06], + ..., + [ 9.4064e-08, 0.0000e+00, 1.8626e-08, ..., 2.4214e-08, + 0.0000e+00, 7.0184e-06], + [ 1.2107e-08, 9.3132e-10, 2.9802e-08, ..., 8.5682e-08, + 0.0000e+00, 1.0710e-07], + [ 1.3970e-08, -1.8626e-08, -7.7300e-08, ..., -8.3819e-08, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0254, -0.0247, -0.0405, 0.0039, 0.0062, 0.0362, 0.0222, -0.0167, + -0.0008, -0.0114], device='cuda:0'), grad: tensor([ 3.4552e-07, -4.3772e-08, -1.4380e-05, 1.3504e-07, 9.4995e-08, + 1.1176e-08, 6.2399e-08, 1.4946e-05, -1.1902e-06, 4.0047e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 214.28, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.5017 re_mapping 0.0046 re_causal 0.0143 /// teacc 99.10 lr 0.00010000 +Epoch 241, weight, value: tensor([[ 0.0214, -0.0883, -0.0638, ..., -0.2073, -0.0867, -0.1647], + [-0.0589, -0.0502, -0.0180, ..., 0.0382, 0.0274, -0.1210], + [ 0.0309, -0.1367, -0.1741, ..., 0.0165, -0.1538, 0.0767], + ..., + [-0.0156, -0.0267, -0.1690, ..., -0.1488, 0.0201, 0.1359], + [ 0.0739, -0.1259, -0.1082, ..., -0.1512, -0.0898, -0.1144], + [-0.1930, 0.0218, 0.1124, ..., 0.0843, -0.0810, -0.1541]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 7.4506e-09, 3.4459e-08, ..., 1.2107e-08, + 0.0000e+00, 2.7940e-08], + [ 8.3819e-09, 2.6170e-07, 6.6124e-08, ..., -6.7055e-08, + 0.0000e+00, 6.1933e-07], + [ 1.8626e-09, 4.0978e-08, 3.0734e-08, ..., 8.3819e-09, + 0.0000e+00, 1.5292e-06], + ..., + [ 2.7940e-09, -5.0291e-07, -8.3819e-08, ..., 1.0058e-07, + 0.0000e+00, -3.1367e-06], + [-6.1467e-08, 2.5146e-08, 2.6636e-07, ..., 2.9802e-07, + 0.0000e+00, 6.1188e-07], + [ 5.5879e-09, 1.7788e-07, -1.4529e-07, ..., -7.1712e-08, + 0.0000e+00, 3.0361e-07]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0256, -0.0243, -0.0404, 0.0036, 0.0063, 0.0362, 0.0223, -0.0173, + -0.0005, -0.0116], device='cuda:0'), grad: tensor([ 1.3318e-07, 1.4212e-06, 2.5816e-06, 6.4261e-08, -1.5022e-06, + 4.3306e-07, 1.0571e-06, -5.8711e-06, 8.1118e-07, 8.4843e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 214.26, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.4761 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.01 lr 0.00010000 +Epoch 242, weight, value: tensor([[ 0.0216, -0.0891, -0.0644, ..., -0.2088, -0.0864, -0.1652], + [-0.0593, -0.0502, -0.0178, ..., 0.0382, 0.0275, -0.1211], + [ 0.0311, -0.1373, -0.1749, ..., 0.0166, -0.1540, 0.0769], + ..., + [-0.0175, -0.0267, -0.1696, ..., -0.1493, 0.0201, 0.1361], + [ 0.0736, -0.1250, -0.1101, ..., -0.1521, -0.0917, -0.1149], + [-0.1960, 0.0219, 0.1135, ..., 0.0853, -0.0816, -0.1544]], + device='cuda:0'), grad: tensor([[ 2.0396e-07, 4.6566e-09, 1.4808e-07, ..., 2.4214e-08, + 9.3132e-10, 1.8626e-08], + [ 5.7742e-08, 1.6484e-07, -2.8685e-07, ..., -5.0198e-07, + -7.2643e-08, 5.7183e-07], + [ 2.3283e-08, 6.6124e-08, 3.3528e-08, ..., 1.5646e-07, + 5.5879e-09, -1.8626e-08], + ..., + [ 1.8626e-09, -4.8801e-07, 3.1292e-07, ..., -6.0629e-07, + 1.3970e-08, -1.8505e-06], + [ 1.9278e-07, 8.3819e-09, 2.5984e-07, ..., 8.8476e-08, + 4.6566e-09, 1.6019e-07], + [ 1.2107e-08, 1.4901e-07, -3.6880e-07, ..., 3.2131e-07, + 7.4506e-09, 1.1148e-06]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0254, -0.0242, -0.0404, 0.0035, 0.0053, 0.0360, 0.0227, -0.0175, + -0.0008, -0.0107], device='cuda:0'), grad: tensor([ 5.6252e-07, 8.1956e-08, 1.7043e-07, -1.4249e-07, 5.3365e-07, + 1.0636e-06, -2.0452e-06, -3.2187e-06, 1.1520e-06, 1.8515e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 214.21, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4993 re_mapping 0.0046 re_causal 0.0141 /// teacc 99.08 lr 0.00010000 +Epoch 243, weight, value: tensor([[ 0.0217, -0.0901, -0.0648, ..., -0.2092, -0.0864, -0.1642], + [-0.0601, -0.0508, -0.0176, ..., 0.0383, 0.0275, -0.1213], + [ 0.0308, -0.1380, -0.1761, ..., 0.0165, -0.1541, 0.0767], + ..., + [-0.0177, -0.0257, -0.1695, ..., -0.1494, 0.0201, 0.1369], + [ 0.0739, -0.1256, -0.1100, ..., -0.1522, -0.0917, -0.1157], + [-0.1969, 0.0214, 0.1133, ..., 0.0851, -0.0816, -0.1554]], + device='cuda:0'), grad: tensor([[-5.3737e-07, -5.9605e-08, 3.2037e-07, ..., 1.5274e-07, + 2.4214e-08, 4.8429e-08], + [ 3.6322e-08, 1.3039e-08, -3.8184e-07, ..., -2.4401e-07, + 0.0000e+00, 2.2352e-08], + [ 1.0179e-06, 3.8184e-08, 5.7444e-06, ..., 5.2415e-06, + 0.0000e+00, 1.1092e-06], + ..., + [ 1.6764e-08, 2.7940e-09, 2.1327e-07, ..., 2.7940e-07, + 0.0000e+00, -2.1420e-08], + [-5.3085e-07, -9.2201e-08, 3.0734e-08, ..., 4.7497e-08, + 9.3132e-10, -1.3970e-08], + [ 2.1700e-07, 2.2165e-07, 1.0757e-06, ..., 8.0094e-07, + 0.0000e+00, 2.2165e-07]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0252, -0.0239, -0.0409, 0.0034, 0.0055, 0.0361, 0.0223, -0.0170, + -0.0007, -0.0111], device='cuda:0'), grad: tensor([-4.5169e-07, -1.0550e-05, 1.8567e-05, 1.7434e-06, -1.9372e-05, + 8.5216e-07, 5.5581e-06, 1.0254e-06, -3.5018e-07, 2.9095e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 214.12, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4966 re_mapping 0.0045 re_causal 0.0135 /// teacc 98.94 lr 0.00010000 +Epoch 244, weight, value: tensor([[ 0.0217, -0.0903, -0.0655, ..., -0.2091, -0.0858, -0.1643], + [-0.0612, -0.0509, -0.0174, ..., 0.0383, 0.0274, -0.1214], + [ 0.0307, -0.1383, -0.1770, ..., 0.0165, -0.1543, 0.0770], + ..., + [-0.0185, -0.0248, -0.1698, ..., -0.1496, 0.0200, 0.1378], + [ 0.0738, -0.1261, -0.1106, ..., -0.1529, -0.0947, -0.1164], + [-0.1981, 0.0215, 0.1134, ..., 0.0852, -0.0818, -0.1560]], + device='cuda:0'), grad: tensor([[-4.6566e-09, 4.6566e-09, 3.6322e-08, ..., 3.1665e-08, + 2.7940e-09, 1.1362e-07], + [ 9.3132e-10, 2.6543e-07, 8.3819e-08, ..., 3.0734e-08, + 5.5879e-09, 5.6252e-07], + [ 9.3132e-10, 5.4948e-08, 9.3132e-09, ..., 1.1176e-08, + 3.0734e-08, -1.2904e-05], + ..., + [ 0.0000e+00, -6.7241e-07, 1.3504e-07, ..., 1.8906e-07, + -6.1467e-08, 1.0051e-05], + [ 4.6566e-09, 5.4017e-08, 1.8347e-07, ..., 2.2352e-07, + 2.0489e-08, 1.0673e-06], + [ 2.7940e-09, 2.4308e-07, -4.4238e-07, ..., -6.1002e-07, + 9.3132e-10, 5.1223e-07]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0250, -0.0239, -0.0409, 0.0034, 0.0053, 0.0362, 0.0223, -0.0168, + -0.0012, -0.0112], device='cuda:0'), grad: tensor([ 2.6450e-07, 1.5302e-06, -2.3335e-05, 9.8255e-07, -1.6391e-07, + 2.1607e-07, 1.3597e-07, 1.7509e-05, 2.3134e-06, 5.3179e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 214.06, cls_loss 0.0016 cls_loss_mapping 0.0017 cls_loss_causal 0.5116 re_mapping 0.0047 re_causal 0.0141 /// teacc 98.97 lr 0.00010000 +Epoch 245, weight, value: tensor([[ 0.0216, -0.0904, -0.0659, ..., -0.2098, -0.0859, -0.1649], + [-0.0618, -0.0510, -0.0175, ..., 0.0383, 0.0273, -0.1216], + [ 0.0309, -0.1387, -0.1772, ..., 0.0165, -0.1540, 0.0780], + ..., + [-0.0188, -0.0245, -0.1701, ..., -0.1496, 0.0200, 0.1383], + [ 0.0737, -0.1265, -0.1109, ..., -0.1536, -0.0964, -0.1186], + [-0.2014, 0.0226, 0.1139, ..., 0.0856, -0.0845, -0.1574]], + device='cuda:0'), grad: tensor([[ 2.1607e-07, 2.7940e-09, 4.2841e-08, ..., 6.4820e-07, + 0.0000e+00, 2.8033e-07], + [ 1.0431e-07, 2.6077e-08, 1.1362e-07, ..., 3.5670e-07, + 0.0000e+00, 2.0023e-07], + [ 2.2445e-07, 6.5193e-09, 8.1956e-08, ..., 6.7893e-07, + 0.0000e+00, 1.2051e-06], + ..., + [ 4.9453e-07, 1.4901e-07, 8.0746e-07, ..., 1.9949e-06, + 0.0000e+00, -3.0827e-07], + [ 3.1665e-07, 1.3039e-08, 5.6811e-08, ..., 8.7079e-07, + 0.0000e+00, 4.0606e-07], + [ 3.1386e-07, 1.0617e-07, 1.8924e-05, ..., 1.6600e-05, + 0.0000e+00, 5.3644e-07]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0253, -0.0241, -0.0406, 0.0032, 0.0054, 0.0361, 0.0229, -0.0164, + -0.0018, -0.0118], device='cuda:0'), grad: tensor([ 1.4650e-06, 9.4995e-07, 2.7418e-06, 6.7115e-05, -3.6657e-05, + -7.8440e-05, 1.3532e-06, 3.4459e-06, 2.0713e-06, 3.5912e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 214.22, cls_loss 0.0011 cls_loss_mapping 0.0024 cls_loss_causal 0.4967 re_mapping 0.0044 re_causal 0.0138 /// teacc 98.99 lr 0.00010000 +Epoch 246, weight, value: tensor([[ 0.0216, -0.0904, -0.0664, ..., -0.2103, -0.0861, -0.1651], + [-0.0621, -0.0514, -0.0181, ..., 0.0382, 0.0273, -0.1219], + [ 0.0308, -0.1390, -0.1776, ..., 0.0165, -0.1540, 0.0778], + ..., + [-0.0190, -0.0242, -0.1703, ..., -0.1494, 0.0200, 0.1392], + [ 0.0735, -0.1268, -0.1111, ..., -0.1541, -0.0965, -0.1196], + [-0.2018, 0.0231, 0.1142, ..., 0.0860, -0.0845, -0.1577]], + device='cuda:0'), grad: tensor([[-6.3330e-08, 1.8626e-09, 8.2888e-08, ..., 8.0094e-08, + 0.0000e+00, 2.7940e-09], + [ 2.7940e-09, 1.7695e-08, -1.0235e-06, ..., -9.6578e-07, + 0.0000e+00, 1.9558e-08], + [ 9.3132e-10, 1.0245e-08, 1.7695e-08, ..., 1.8626e-08, + 0.0000e+00, -9.3132e-10], + ..., + [ 9.3132e-10, 2.6077e-08, 2.0768e-07, ..., 2.4028e-07, + 0.0000e+00, 3.0734e-08], + [-6.5193e-09, 5.5879e-09, 4.8336e-07, ..., 4.8988e-07, + 0.0000e+00, 1.3970e-08], + [ 1.1176e-08, 5.8208e-07, 3.8743e-06, ..., 3.4459e-06, + 0.0000e+00, 1.5832e-08]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0253, -0.0242, -0.0407, 0.0032, 0.0052, 0.0361, 0.0231, -0.0160, + -0.0023, -0.0116], device='cuda:0'), grad: tensor([ 4.4703e-08, -2.6934e-06, 4.2841e-08, -3.5297e-07, -6.4597e-06, + 2.7753e-07, 6.6403e-07, 6.0722e-07, 1.2713e-06, 6.6049e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 214.15, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.5008 re_mapping 0.0044 re_causal 0.0142 /// teacc 99.00 lr 0.00010000 +Epoch 247, weight, value: tensor([[ 0.0217, -0.0905, -0.0665, ..., -0.2085, -0.0862, -0.1653], + [-0.0638, -0.0514, -0.0183, ..., 0.0381, 0.0273, -0.1220], + [ 0.0316, -0.1395, -0.1779, ..., 0.0165, -0.1541, 0.0780], + ..., + [-0.0211, -0.0244, -0.1705, ..., -0.1497, 0.0201, 0.1394], + [ 0.0736, -0.1270, -0.1111, ..., -0.1545, -0.0965, -0.1210], + [-0.2023, 0.0233, 0.1146, ..., 0.0864, -0.0852, -0.1581]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 0.0000e+00, 1.3458e-07, ..., 1.2945e-07, + 8.8476e-09, 9.3132e-09], + [ 7.0594e-07, 6.9849e-09, -3.1525e-07, ..., -3.9162e-07, + -5.1688e-08, 3.3062e-07], + [ 1.4063e-07, 3.7253e-09, 6.5193e-08, ..., 9.4995e-08, + 6.9849e-09, -8.7544e-08], + ..., + [ 3.8650e-08, -3.3528e-08, 2.6915e-07, ..., 2.6776e-07, + 9.3132e-09, -7.4646e-07], + [-1.2163e-06, 3.7253e-09, 1.5972e-07, ..., 1.0710e-07, + 1.7229e-08, 1.0850e-07], + [ 2.0489e-08, 2.2817e-08, -5.1362e-07, ..., -4.4983e-07, + -1.0245e-08, 3.1618e-07]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0251, -0.0243, -0.0404, 0.0030, 0.0049, 0.0363, 0.0230, -0.0161, + -0.0029, -0.0114], device='cuda:0'), grad: tensor([ 3.7020e-07, 1.6056e-06, 3.0873e-07, 1.2899e-07, 4.6007e-07, + 8.0373e-07, 9.2667e-08, -1.5404e-06, -2.2016e-06, -2.9802e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.18, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.5058 re_mapping 0.0041 re_causal 0.0134 /// teacc 98.99 lr 0.00010000 +Epoch 248, weight, value: tensor([[ 0.0216, -0.0906, -0.0680, ..., -0.2080, -0.0868, -0.1654], + [-0.0649, -0.0517, -0.0183, ..., 0.0381, 0.0273, -0.1223], + [ 0.0314, -0.1402, -0.1785, ..., 0.0165, -0.1543, 0.0781], + ..., + [-0.0214, -0.0243, -0.1705, ..., -0.1497, 0.0194, 0.1401], + [ 0.0737, -0.1275, -0.1116, ..., -0.1550, -0.0971, -0.1224], + [-0.2028, 0.0236, 0.1153, ..., 0.0870, -0.0855, -0.1584]], + device='cuda:0'), grad: tensor([[ 1.8254e-06, 0.0000e+00, 1.3253e-06, ..., 3.2596e-09, + 0.0000e+00, 4.6566e-09], + [ 1.8161e-08, 2.7940e-09, 5.1688e-08, ..., -2.4214e-08, + -9.3132e-10, 7.5903e-08], + [ 2.1420e-08, 4.6566e-10, 4.0047e-08, ..., 3.2596e-09, + 4.6566e-10, -1.1222e-07], + ..., + [ 3.7253e-09, 2.3283e-09, 7.4040e-08, ..., 4.2375e-08, + 4.6566e-10, 2.2817e-08], + [-2.3860e-06, 1.3970e-09, -3.1367e-06, ..., 5.0757e-08, + 0.0000e+00, -4.6566e-08], + [ 2.7101e-07, -2.7940e-09, 1.3700e-06, ..., 3.2131e-08, + 0.0000e+00, 4.3772e-08]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0252, -0.0245, -0.0404, 0.0031, 0.0042, 0.0362, 0.0235, -0.0158, + -0.0035, -0.0109], device='cuda:0'), grad: tensor([ 1.0513e-05, 3.3341e-07, -2.9337e-08, 1.8254e-07, 2.8731e-07, + 3.5856e-07, 9.6485e-07, 2.3050e-07, -1.8075e-05, 5.2154e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 213.88, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5106 re_mapping 0.0043 re_causal 0.0133 /// teacc 99.12 lr 0.00010000 +Epoch 249, weight, value: tensor([[ 0.0214, -0.0908, -0.0657, ..., -0.2088, -0.0899, -0.1656], + [-0.0655, -0.0518, -0.0187, ..., 0.0381, 0.0270, -0.1225], + [ 0.0310, -0.1407, -0.1796, ..., 0.0165, -0.1538, 0.0781], + ..., + [-0.0218, -0.0243, -0.1707, ..., -0.1502, 0.0196, 0.1402], + [ 0.0744, -0.1279, -0.1118, ..., -0.1556, -0.0972, -0.1224], + [-0.2039, 0.0233, 0.1150, ..., 0.0868, -0.0859, -0.1588]], + device='cuda:0'), grad: tensor([[ 1.4435e-08, 2.0489e-08, 1.3737e-07, ..., 8.0559e-08, + 3.7253e-09, 1.8626e-09], + [ 2.3283e-09, 1.1977e-06, 5.1968e-06, ..., 3.9786e-06, + -9.3132e-10, 2.1420e-08], + [ 7.9162e-09, 3.3993e-08, 1.3411e-07, ..., 9.2201e-08, + 4.6566e-10, 2.1420e-08], + ..., + [ 4.6566e-10, 4.7497e-08, 3.1944e-07, ..., 2.3982e-07, + -9.3132e-10, -3.1199e-08], + [ 5.1223e-09, 8.8010e-08, 1.0030e-06, ..., 3.4180e-07, + 2.3283e-09, 1.5367e-08], + [ 1.8626e-09, -2.0489e-08, 2.5425e-07, ..., 5.7975e-07, + -9.3132e-10, 8.8476e-09]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0248, -0.0246, -0.0404, 0.0031, 0.0044, 0.0363, 0.0237, -0.0159, + -0.0027, -0.0115], device='cuda:0'), grad: tensor([ 3.2876e-07, 1.4298e-05, 4.0606e-07, -3.6322e-08, -2.2903e-05, + 9.3132e-08, 4.6231e-06, 7.3621e-07, 1.8738e-06, 5.8627e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 214.26, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.5035 re_mapping 0.0043 re_causal 0.0131 /// teacc 99.07 lr 0.00010000 +Epoch 250, weight, value: tensor([[ 0.0215, -0.0929, -0.0660, ..., -0.2116, -0.0899, -0.1671], + [-0.0660, -0.0522, -0.0191, ..., 0.0382, 0.0269, -0.1227], + [ 0.0303, -0.1419, -0.1805, ..., 0.0165, -0.1540, 0.0780], + ..., + [-0.0221, -0.0244, -0.1712, ..., -0.1510, 0.0197, 0.1404], + [ 0.0752, -0.1285, -0.1117, ..., -0.1565, -0.0971, -0.1232], + [-0.2046, 0.0234, 0.1146, ..., 0.0865, -0.0854, -0.1590]], + device='cuda:0'), grad: tensor([[-6.1002e-08, 0.0000e+00, 1.6298e-08, ..., 1.8626e-09, + 0.0000e+00, 1.7229e-08], + [ 3.5390e-08, 1.8626e-09, 5.0291e-08, ..., -1.4901e-08, + 0.0000e+00, 9.9186e-08], + [ 3.7719e-08, 4.6566e-10, 7.4506e-09, ..., 4.6566e-10, + 0.0000e+00, -1.0421e-06], + ..., + [ 8.3819e-09, 6.9849e-09, 5.5879e-08, ..., 1.3039e-08, + 0.0000e+00, 3.6322e-08], + [-3.5390e-08, 9.3132e-10, 3.6601e-07, ..., 6.5193e-09, + 0.0000e+00, 8.3912e-07], + [ 1.5832e-08, 9.3132e-10, 9.6392e-08, ..., 2.1886e-08, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0252, -0.0247, -0.0405, 0.0036, 0.0049, 0.0371, 0.0233, -0.0161, + -0.0026, -0.0119], device='cuda:0'), grad: tensor([-1.2293e-07, 5.1362e-07, -1.8105e-06, 1.3923e-07, -1.2992e-06, + 3.0315e-07, -6.4867e-07, 3.6554e-07, 2.1849e-06, 3.8510e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 214.12, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.5059 re_mapping 0.0043 re_causal 0.0133 /// teacc 99.04 lr 0.00010000 +Epoch 251, weight, value: tensor([[ 0.0215, -0.0938, -0.0653, ..., -0.2089, -0.0899, -0.1673], + [-0.0663, -0.0526, -0.0196, ..., 0.0381, 0.0269, -0.1229], + [ 0.0301, -0.1421, -0.1810, ..., 0.0165, -0.1537, 0.0785], + ..., + [-0.0225, -0.0250, -0.1724, ..., -0.1520, 0.0198, 0.1404], + [ 0.0752, -0.1287, -0.1121, ..., -0.1572, -0.0973, -0.1238], + [-0.2050, 0.0246, 0.1149, ..., 0.0867, -0.0855, -0.1592]], + device='cuda:0'), grad: tensor([[-6.8918e-08, 3.2596e-09, 3.7253e-08, ..., 3.5390e-08, + 1.3970e-09, 2.0070e-07], + [ 4.6100e-08, 7.2177e-08, -2.3082e-05, ..., -7.1704e-05, + 1.2107e-08, 4.6380e-07], + [-2.0489e-07, 6.0536e-09, 1.0245e-08, ..., -1.4063e-07, + 4.1910e-09, -1.5805e-06], + ..., + [ 6.9849e-09, -1.1129e-07, 2.1532e-05, ..., 6.6817e-05, + 1.8626e-09, -6.4261e-07], + [ 1.8813e-07, 1.0710e-08, 4.9826e-08, ..., 3.3295e-07, + 6.0536e-09, 1.0235e-06], + [ 7.5903e-08, 1.6298e-08, 1.3392e-06, ..., 4.3213e-06, + 3.7253e-09, 5.6671e-07]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0247, -0.0249, -0.0404, 0.0032, 0.0049, 0.0373, 0.0235, -0.0166, + -0.0028, -0.0116], device='cuda:0'), grad: tensor([ 7.1852e-07, -1.8251e-04, -7.9796e-06, -2.3330e-07, 1.0980e-06, + -4.7963e-07, 4.2049e-07, 1.6963e-04, 5.2042e-06, 1.3836e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 213.97, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4848 re_mapping 0.0044 re_causal 0.0133 /// teacc 99.02 lr 0.00010000 +Epoch 252, weight, value: tensor([[ 0.0216, -0.0940, -0.0656, ..., -0.2093, -0.0901, -0.1675], + [-0.0666, -0.0526, -0.0192, ..., 0.0383, 0.0271, -0.1230], + [ 0.0299, -0.1426, -0.1816, ..., 0.0165, -0.1538, 0.0781], + ..., + [-0.0227, -0.0250, -0.1731, ..., -0.1534, 0.0197, 0.1409], + [ 0.0752, -0.1295, -0.1125, ..., -0.1580, -0.0971, -0.1248], + [-0.2052, 0.0250, 0.1132, ..., 0.0855, -0.0859, -0.1594]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, 3.4459e-08, 2.7008e-08, ..., 2.3283e-08, + 2.7940e-09, 6.7055e-08], + [ 2.7008e-08, 2.9337e-07, -4.6566e-10, ..., 2.5099e-07, + -6.5193e-08, 7.7952e-07], + [ 1.3970e-09, 5.7276e-08, 4.6566e-09, ..., 9.3132e-09, + 9.3132e-10, 7.1712e-08], + ..., + [ 2.7940e-09, -1.2619e-07, 1.2340e-07, ..., -2.7334e-07, + 5.2620e-08, -6.6217e-07], + [ 1.3970e-09, 1.5227e-07, 2.8871e-08, ..., 2.4354e-07, + 9.3132e-09, 2.6543e-07], + [ 2.3749e-08, 1.4901e-07, -4.5355e-07, ..., -3.8743e-07, + 9.3132e-10, 4.9500e-07]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0247, -0.0246, -0.0405, 0.0034, 0.0068, 0.0373, 0.0234, -0.0170, + -0.0033, -0.0132], device='cuda:0'), grad: tensor([ 2.8079e-07, 2.5500e-06, 2.5937e-07, -4.7162e-06, 4.7125e-07, + 2.2408e-06, 6.9384e-08, -2.2110e-06, 1.5041e-07, 9.0757e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 214.27, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4801 re_mapping 0.0044 re_causal 0.0129 /// teacc 99.03 lr 0.00010000 +Epoch 253, weight, value: tensor([[ 0.0219, -0.0938, -0.0655, ..., -0.2096, -0.0903, -0.1677], + [-0.0672, -0.0529, -0.0202, ..., 0.0381, 0.0270, -0.1232], + [ 0.0297, -0.1433, -0.1822, ..., 0.0165, -0.1541, 0.0780], + ..., + [-0.0229, -0.0250, -0.1737, ..., -0.1540, 0.0198, 0.1414], + [ 0.0752, -0.1302, -0.1113, ..., -0.1577, -0.0972, -0.1254], + [-0.2057, 0.0251, 0.1132, ..., 0.0860, -0.0852, -0.1598]], + device='cuda:0'), grad: tensor([[-3.7253e-09, 2.7940e-09, 2.8871e-08, ..., 2.7008e-08, + 3.7253e-09, 1.5646e-07], + [ 1.8626e-09, 2.0489e-08, 7.7300e-08, ..., 4.1910e-08, + -1.8626e-09, 9.9652e-08], + [ 2.7940e-09, 6.5193e-09, 9.3132e-09, ..., 8.3819e-09, + -2.9802e-08, -1.1837e-06], + ..., + [ 1.8626e-09, -1.3039e-08, 5.9605e-08, ..., 5.7742e-08, + 3.7253e-09, -1.8720e-07], + [-2.6077e-08, 9.3132e-09, 6.7987e-08, ..., 7.3574e-08, + 9.3132e-10, 5.3085e-08], + [ 1.1176e-08, 1.5832e-08, 1.2107e-07, ..., 6.7055e-08, + 0.0000e+00, 2.9802e-08]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0243, -0.0250, -0.0405, 0.0033, 0.0071, 0.0368, 0.0233, -0.0170, + -0.0025, -0.0127], device='cuda:0'), grad: tensor([ 3.8091e-07, 3.4366e-07, -2.2687e-06, 1.5078e-06, -4.2282e-07, + -2.4214e-08, 2.8592e-07, -2.0210e-07, -2.4214e-08, 4.2282e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 214.30, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4771 re_mapping 0.0041 re_causal 0.0127 /// teacc 99.07 lr 0.00010000 +Epoch 254, weight, value: tensor([[ 0.0221, -0.0938, -0.0656, ..., -0.2098, -0.0903, -0.1683], + [-0.0677, -0.0531, -0.0205, ..., 0.0380, 0.0269, -0.1233], + [ 0.0296, -0.1436, -0.1827, ..., 0.0165, -0.1541, 0.0800], + ..., + [-0.0232, -0.0253, -0.1743, ..., -0.1547, 0.0198, 0.1404], + [ 0.0751, -0.1304, -0.1117, ..., -0.1582, -0.0973, -0.1261], + [-0.2064, 0.0254, 0.1134, ..., 0.0862, -0.0852, -0.1598]], + device='cuda:0'), grad: tensor([[ 2.6077e-07, 2.7940e-09, 1.1278e-06, ..., 4.4331e-07, + 6.8452e-07, 2.1420e-08], + [ 1.6764e-08, 1.0245e-08, -1.5153e-06, ..., -1.5423e-06, + 3.0734e-08, 1.0524e-07], + [ 4.3772e-08, 2.7940e-09, 1.1269e-07, ..., 2.5146e-08, + 8.9407e-08, -5.6811e-08], + ..., + [ 3.7253e-09, 5.5879e-09, 2.3842e-07, ..., 1.4529e-07, + 5.5879e-09, -1.2759e-07], + [-5.0291e-08, 8.3819e-09, 6.2026e-07, ..., 5.9139e-07, + 1.3970e-08, 5.6811e-08], + [ 6.5193e-09, -1.3039e-08, 1.4901e-08, ..., -3.7253e-08, + 7.4506e-09, 1.8161e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0241, -0.0243, -0.0410, 0.0029, 0.0072, 0.0368, 0.0234, -0.0177, + -0.0028, -0.0125], device='cuda:0'), grad: tensor([ 3.9414e-06, -4.3474e-06, 3.2410e-07, 2.8685e-07, 1.3843e-05, + 8.4471e-07, -1.7256e-05, 3.0268e-07, 1.6196e-06, 3.8836e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 214.50, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4937 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.03 lr 0.00010000 +Epoch 255, weight, value: tensor([[ 0.0220, -0.0956, -0.0662, ..., -0.2099, -0.0907, -0.1688], + [-0.0680, -0.0528, -0.0197, ..., 0.0384, 0.0269, -0.1235], + [ 0.0296, -0.1433, -0.1830, ..., 0.0165, -0.1540, 0.0805], + ..., + [-0.0233, -0.0260, -0.1757, ..., -0.1564, 0.0202, 0.1404], + [ 0.0751, -0.1308, -0.1120, ..., -0.1587, -0.0974, -0.1266], + [-0.2072, 0.0262, 0.1138, ..., 0.0866, -0.0865, -0.1600]], + device='cuda:0'), grad: tensor([[-4.6566e-09, -9.3132e-10, 6.5193e-09, ..., 7.4506e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 3.7253e-09, -1.7695e-08, ..., -4.0047e-08, + 0.0000e+00, 2.2352e-08], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 4.6566e-09], + ..., + [ 0.0000e+00, -1.8626e-09, 2.5146e-08, ..., 3.6322e-08, + 0.0000e+00, -5.5879e-08], + [ 4.6566e-09, 1.8626e-09, 6.5193e-08, ..., 7.8231e-08, + 0.0000e+00, 4.6566e-09], + [ 4.6566e-09, -8.3819e-09, -1.4435e-07, ..., -1.1083e-07, + 0.0000e+00, 1.9558e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0244, -0.0240, -0.0409, 0.0026, 0.0070, 0.0370, 0.0235, -0.0182, + -0.0031, -0.0122], device='cuda:0'), grad: tensor([-1.0245e-08, -3.7253e-08, 1.9558e-08, 6.3330e-08, 1.0896e-07, + -7.9162e-08, 1.6764e-08, -6.6124e-08, 1.7323e-07, -1.8720e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 214.09, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.5008 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.09 lr 0.00010000 +Epoch 256, weight, value: tensor([[ 0.0219, -0.0960, -0.0664, ..., -0.2106, -0.0909, -0.1694], + [-0.0684, -0.0527, -0.0202, ..., 0.0384, 0.0267, -0.1236], + [ 0.0294, -0.1439, -0.1834, ..., 0.0165, -0.1542, 0.0804], + ..., + [-0.0227, -0.0261, -0.1759, ..., -0.1568, 0.0207, 0.1409], + [ 0.0752, -0.1311, -0.1110, ..., -0.1591, -0.0977, -0.1273], + [-0.2079, 0.0261, 0.1139, ..., 0.0866, -0.0873, -0.1608]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 1.7695e-08, 2.4214e-08, ..., 7.8231e-08, + 0.0000e+00, 2.7940e-09], + [ 3.7253e-09, 1.1176e-08, -4.0047e-08, ..., -4.3772e-08, + 0.0000e+00, 3.5390e-08], + [ 3.7253e-09, 4.6566e-09, 7.4506e-09, ..., 1.7695e-08, + 0.0000e+00, 1.4901e-08], + ..., + [ 1.2107e-08, 2.4214e-08, 5.9605e-08, ..., 1.8626e-07, + 0.0000e+00, -9.4995e-08], + [ 8.3819e-09, 9.3132e-09, 2.9802e-08, ..., 1.0058e-07, + 0.0000e+00, 1.3970e-08], + [ 2.4866e-07, -2.0396e-07, 4.2561e-07, ..., 1.6410e-06, + 0.0000e+00, 2.2352e-08]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0246, -0.0240, -0.0410, 0.0028, 0.0071, 0.0370, 0.0236, -0.0181, + -0.0026, -0.0124], device='cuda:0'), grad: tensor([ 1.1642e-07, -4.5635e-08, 7.9162e-08, 2.6226e-06, -8.5495e-07, + -5.5954e-06, 5.3272e-07, 1.9278e-07, 2.0396e-07, 2.7325e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 214.08, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4913 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.03 lr 0.00010000 +Epoch 257, weight, value: tensor([[ 0.0220, -0.0960, -0.0666, ..., -0.2115, -0.0909, -0.1698], + [-0.0687, -0.0526, -0.0199, ..., 0.0387, 0.0266, -0.1236], + [ 0.0293, -0.1449, -0.1840, ..., 0.0164, -0.1542, 0.0806], + ..., + [-0.0230, -0.0260, -0.1764, ..., -0.1578, 0.0209, 0.1412], + [ 0.0752, -0.1314, -0.1111, ..., -0.1597, -0.0978, -0.1279], + [-0.2085, 0.0240, 0.1129, ..., 0.0850, -0.0909, -0.1614]], + device='cuda:0'), grad: tensor([[-2.2352e-08, -5.8748e-06, 1.1176e-08, ..., 1.8626e-08, + 0.0000e+00, 1.3039e-08], + [ 2.7940e-09, 1.5926e-07, 1.0151e-07, ..., 2.0675e-07, + 0.0000e+00, 1.4994e-07], + [ 3.7253e-09, 6.8080e-07, 5.5879e-09, ..., 1.1176e-08, + 0.0000e+00, 7.0781e-08], + ..., + [ 9.3132e-10, 1.5069e-06, 9.7416e-07, ..., 2.1346e-06, + 0.0000e+00, 8.3540e-07], + [-9.3132e-10, 8.5495e-07, 1.9558e-08, ..., 6.2399e-08, + 0.0000e+00, 5.9605e-08], + [ 7.4506e-09, -8.0280e-07, -1.2424e-06, ..., -2.5705e-06, + 0.0000e+00, 2.4214e-08]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0246, -0.0236, -0.0413, 0.0027, 0.0087, 0.0373, 0.0232, -0.0184, + -0.0026, -0.0138], device='cuda:0'), grad: tensor([-3.5316e-05, 7.3202e-07, 4.0568e-06, -9.6764e-07, 1.8135e-05, + 4.1164e-07, 5.9456e-06, 5.9083e-06, 5.1409e-06, -4.0717e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 214.33, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.5119 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 258, weight, value: tensor([[ 0.0221, -0.0957, -0.0668, ..., -0.2122, -0.0910, -0.1704], + [-0.0692, -0.0528, -0.0200, ..., 0.0388, 0.0267, -0.1223], + [ 0.0290, -0.1456, -0.1845, ..., 0.0163, -0.1543, 0.0792], + ..., + [-0.0231, -0.0259, -0.1767, ..., -0.1585, 0.0209, 0.1415], + [ 0.0753, -0.1314, -0.1115, ..., -0.1606, -0.0978, -0.1285], + [-0.2091, 0.0239, 0.1133, ..., 0.0856, -0.0909, -0.1621]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 5.4017e-08, 3.4645e-07, ..., 2.0023e-07, + 9.3132e-09, 2.7940e-09], + [ 3.0734e-08, 1.1921e-07, -1.9278e-07, ..., -5.4762e-07, + -1.1176e-07, 1.9558e-07], + [ 8.9407e-08, 1.7695e-08, 1.3597e-07, ..., 1.1642e-07, + 9.3132e-10, 2.7940e-08], + ..., + [ 5.5879e-09, -3.2224e-07, 2.1700e-07, ..., 2.5611e-07, + -1.7695e-08, -5.0291e-07], + [-5.5879e-08, 5.8673e-08, 1.0161e-06, ..., 7.8045e-07, + 4.6566e-09, 1.6764e-08], + [ 2.4214e-08, 3.8184e-08, -5.1737e-05, ..., -3.8207e-05, + 5.5879e-09, 1.2107e-07]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0245, -0.0227, -0.0421, 0.0026, 0.0085, 0.0373, 0.0230, -0.0185, + -0.0029, -0.0136], device='cuda:0'), grad: tensor([ 3.8091e-07, -1.5348e-06, 6.6031e-07, 6.5472e-07, 8.2970e-05, + 1.4529e-07, 8.9407e-08, -3.1665e-07, 1.9297e-06, -8.4996e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 257---------------------------------------------------- +epoch 257, time 230.40, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4764 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.15 lr 0.00010000 +Epoch 259, weight, value: tensor([[ 0.0220, -0.0972, -0.0672, ..., -0.2144, -0.0910, -0.1706], + [-0.0694, -0.0527, -0.0202, ..., 0.0388, 0.0266, -0.1222], + [ 0.0289, -0.1466, -0.1856, ..., 0.0163, -0.1544, 0.0788], + ..., + [-0.0232, -0.0260, -0.1769, ..., -0.1589, 0.0210, 0.1418], + [ 0.0753, -0.1316, -0.1130, ..., -0.1622, -0.0985, -0.1293], + [-0.2089, 0.0238, 0.1165, ..., 0.0888, -0.0901, -0.1626]], + device='cuda:0'), grad: tensor([[-2.2352e-08, 5.5879e-09, 4.6566e-09, ..., 2.7940e-09, + 9.3132e-10, 1.4901e-08], + [ 3.7253e-09, 3.3528e-08, -5.5879e-08, ..., -8.2888e-08, + 9.3132e-10, 7.6368e-08], + [ 2.7940e-09, 3.4459e-08, 7.4506e-09, ..., 9.3132e-09, + 9.3132e-10, 7.2643e-08], + ..., + [ 0.0000e+00, 3.6322e-08, 3.8184e-08, ..., 5.2154e-08, + 3.7253e-09, 8.4750e-08], + [ 8.3819e-09, 1.0151e-07, 8.3819e-09, ..., 1.3970e-08, + 1.8626e-09, 2.0117e-07], + [ 2.7940e-09, 3.7253e-09, -1.2107e-08, ..., -1.4901e-08, + 1.8626e-09, 4.1910e-08]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0247, -0.0227, -0.0423, 0.0029, 0.0055, 0.0373, 0.0229, -0.0184, + -0.0037, -0.0105], device='cuda:0'), grad: tensor([-1.3970e-08, -7.2643e-08, 2.1793e-07, -1.1148e-06, 2.5146e-08, + -4.5635e-08, 1.8626e-08, 3.2224e-07, 5.7183e-07, 8.5682e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 214.49, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5049 re_mapping 0.0043 re_causal 0.0131 /// teacc 99.02 lr 0.00010000 +Epoch 260, weight, value: tensor([[ 0.0224, -0.0972, -0.0684, ..., -0.2141, -0.0910, -0.1710], + [-0.0704, -0.0525, -0.0201, ..., 0.0389, 0.0266, -0.1223], + [ 0.0288, -0.1475, -0.1862, ..., 0.0164, -0.1545, 0.0791], + ..., + [-0.0235, -0.0267, -0.1772, ..., -0.1595, 0.0209, 0.1418], + [ 0.0725, -0.1319, -0.1132, ..., -0.1657, -0.0986, -0.1308], + [-0.2106, 0.0237, 0.1166, ..., 0.0888, -0.0902, -0.1630]], + device='cuda:0'), grad: tensor([[-3.4459e-08, 2.7940e-09, 1.1269e-07, ..., 6.0536e-08, + 0.0000e+00, 1.3039e-08], + [ 5.5879e-09, 4.2841e-08, 1.0245e-08, ..., 2.1420e-08, + 0.0000e+00, 1.1735e-07], + [ 1.8626e-09, 2.6077e-08, 7.4506e-09, ..., 6.5193e-09, + 0.0000e+00, 1.6671e-07], + ..., + [ 1.8626e-09, 4.0047e-08, 2.2352e-08, ..., 2.3283e-08, + 0.0000e+00, 9.2201e-08], + [-8.3819e-09, 3.5390e-08, 1.1176e-08, ..., 1.6764e-08, + -0.0000e+00, 8.1025e-08], + [ 1.8626e-08, 5.5879e-09, 4.1164e-07, ..., 2.3004e-07, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0248, -0.0224, -0.0425, 0.0027, 0.0056, 0.0391, 0.0225, -0.0186, + -0.0065, -0.0104], device='cuda:0'), grad: tensor([ 1.3411e-07, 3.8650e-07, 6.7707e-07, -6.0163e-06, -9.2573e-07, + 4.3213e-06, 9.9652e-08, 3.1106e-07, 2.1234e-07, 8.1304e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 214.08, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4799 re_mapping 0.0046 re_causal 0.0130 /// teacc 99.05 lr 0.00010000 +Epoch 261, weight, value: tensor([[ 0.0228, -0.0972, -0.0687, ..., -0.2134, -0.0911, -0.1715], + [-0.0723, -0.0527, -0.0195, ..., 0.0392, 0.0265, -0.1226], + [ 0.0286, -0.1478, -0.1868, ..., 0.0163, -0.1547, 0.0788], + ..., + [-0.0238, -0.0270, -0.1773, ..., -0.1599, 0.0232, 0.1434], + [ 0.0714, -0.1334, -0.1147, ..., -0.1666, -0.0988, -0.1334], + [-0.2114, 0.0224, 0.1165, ..., 0.0885, -0.0932, -0.1666]], + device='cuda:0'), grad: tensor([[ 6.3330e-08, 9.3132e-09, 3.9116e-08, ..., 1.0245e-08, + 0.0000e+00, 1.5832e-08], + [ 7.4506e-09, 3.3528e-08, -2.4214e-08, ..., -2.3283e-08, + 0.0000e+00, 6.1467e-08], + [ 9.3132e-09, 6.1467e-08, 1.3039e-08, ..., -8.3819e-09, + 0.0000e+00, 8.5682e-08], + ..., + [ 9.3132e-10, 4.1910e-08, 1.1083e-07, ..., 1.2666e-07, + 0.0000e+00, 4.3772e-08], + [ 6.3330e-08, 1.1362e-07, 3.4459e-08, ..., 7.7300e-08, + 0.0000e+00, 2.3283e-07], + [ 7.4506e-09, -7.3574e-08, -4.6566e-07, ..., -4.5449e-07, + 0.0000e+00, 1.3970e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0244, -0.0223, -0.0425, 0.0066, 0.0056, 0.0365, 0.0233, -0.0177, + -0.0084, -0.0113], device='cuda:0'), grad: tensor([ 2.1886e-07, 1.5832e-08, 1.9651e-07, -8.2143e-07, 4.4424e-07, + 3.3341e-07, -5.6066e-07, 3.5577e-07, 6.0163e-07, -8.0932e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 214.04, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.5038 re_mapping 0.0045 re_causal 0.0138 /// teacc 99.07 lr 0.00010000 +Epoch 262, weight, value: tensor([[ 0.0231, -0.0963, -0.0690, ..., -0.2134, -0.0910, -0.1717], + [-0.0767, -0.0519, -0.0192, ..., 0.0394, 0.0266, -0.1216], + [ 0.0282, -0.1480, -0.1882, ..., 0.0164, -0.1547, 0.0778], + ..., + [-0.0249, -0.0279, -0.1776, ..., -0.1607, 0.0231, 0.1433], + [ 0.0709, -0.1338, -0.1149, ..., -0.1670, -0.0988, -0.1336], + [-0.2124, 0.0223, 0.1164, ..., 0.0885, -0.0932, -0.1668]], + device='cuda:0'), grad: tensor([[ 2.2445e-07, 0.0000e+00, 1.8347e-07, ..., 1.0245e-08, + 0.0000e+00, 3.7253e-09], + [ 3.3900e-07, 8.3819e-09, -5.3085e-08, ..., -2.3469e-07, + -6.5193e-09, 9.8869e-06], + [ 1.2107e-07, 7.4506e-09, 1.2759e-07, ..., 4.5635e-08, + 0.0000e+00, -9.9167e-06], + ..., + [ 2.7940e-08, -2.0489e-08, 1.1548e-07, ..., 1.0431e-07, + 1.8626e-09, -5.4017e-08], + [ 1.5264e-06, 1.8626e-09, 8.6520e-07, ..., 3.4459e-08, + 9.3132e-10, 1.3970e-08], + [ 3.0734e-08, 9.3132e-10, 5.5879e-08, ..., 4.1910e-08, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0240, -0.0211, -0.0436, 0.0066, 0.0057, 0.0364, 0.0240, -0.0184, + -0.0085, -0.0115], device='cuda:0'), grad: tensor([ 1.1604e-06, 2.4781e-05, -2.3589e-05, 8.7731e-07, -5.6848e-06, + 6.0797e-06, -1.2264e-05, 6.3051e-07, 7.6964e-06, 2.8498e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.37, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4700 re_mapping 0.0045 re_causal 0.0125 /// teacc 99.12 lr 0.00010000 +Epoch 263, weight, value: tensor([[ 0.0232, -0.0969, -0.0693, ..., -0.2138, -0.0910, -0.1724], + [-0.0771, -0.0520, -0.0193, ..., 0.0394, 0.0266, -0.1222], + [ 0.0280, -0.1488, -0.1894, ..., 0.0163, -0.1547, 0.0778], + ..., + [-0.0255, -0.0277, -0.1775, ..., -0.1601, 0.0231, 0.1448], + [ 0.0706, -0.1340, -0.1160, ..., -0.1674, -0.0991, -0.1338], + [-0.2130, 0.0223, 0.1161, ..., 0.0881, -0.0932, -0.1685]], + device='cuda:0'), grad: tensor([[-2.8871e-08, 9.3132e-10, 2.7940e-09, ..., 2.7940e-09, + 0.0000e+00, 9.3132e-09], + [ 2.7940e-09, 8.3819e-09, -1.9558e-08, ..., -2.7008e-08, + 4.6566e-09, 2.0489e-07], + [ 9.3132e-10, 1.7695e-08, 1.8626e-09, ..., 2.7940e-09, + 3.0734e-08, 1.3346e-06], + ..., + [ 9.3132e-10, -1.5832e-08, 1.2759e-07, ..., 2.2911e-07, + -4.8429e-08, -1.7593e-06], + [ 6.5193e-09, 3.7253e-09, 6.5193e-09, ..., 1.0245e-08, + 1.8626e-09, 5.1223e-08], + [ 8.3819e-09, 1.8626e-09, 1.8720e-07, ..., 3.5390e-07, + 1.8626e-09, 5.0291e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0241, -0.0211, -0.0438, 0.0064, 0.0061, 0.0365, 0.0242, -0.0178, + -0.0089, -0.0120], device='cuda:0'), grad: tensor([-1.0803e-07, 1.8068e-07, 1.9064e-06, -9.6858e-08, -1.7425e-06, + 2.6356e-07, 1.1269e-07, -1.7416e-06, 1.4342e-07, 1.0766e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 214.47, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4603 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.08 lr 0.00010000 +Epoch 264, weight, value: tensor([[ 0.0236, -0.0954, -0.0692, ..., -0.2137, -0.0910, -0.1728], + [-0.0780, -0.0523, -0.0180, ..., 0.0401, 0.0267, -0.1224], + [ 0.0278, -0.1514, -0.1897, ..., 0.0164, -0.1548, 0.0780], + ..., + [-0.0263, -0.0281, -0.1794, ..., -0.1624, 0.0233, 0.1448], + [ 0.0711, -0.1343, -0.1158, ..., -0.1675, -0.0994, -0.1342], + [-0.2146, 0.0222, 0.1161, ..., 0.0881, -0.0935, -0.1690]], + device='cuda:0'), grad: tensor([[-1.6764e-08, 0.0000e+00, 1.7695e-08, ..., 2.7940e-08, + 2.6077e-08, 9.3132e-10], + [ 1.3039e-08, 1.3970e-08, -1.4817e-06, ..., -2.4289e-06, + -2.5202e-06, 4.3772e-08], + [ 4.6566e-09, 2.7940e-09, 7.0781e-08, ..., 1.1642e-07, + 1.2014e-07, 1.1176e-08], + ..., + [ 2.7940e-09, -3.3528e-08, 1.4342e-07, ..., 2.3097e-07, + 2.1886e-07, -1.1269e-07], + [-6.2492e-07, 1.8626e-09, 5.6531e-07, ..., 9.2667e-07, + 9.6112e-07, 7.4506e-09], + [ 6.5193e-09, 9.3132e-09, 3.5390e-08, ..., 7.0781e-08, + 7.2643e-08, 2.6077e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0233, -0.0207, -0.0436, 0.0065, 0.0061, 0.0365, 0.0239, -0.0189, + -0.0086, -0.0122], device='cuda:0'), grad: tensor([ 5.7742e-08, -9.8124e-06, 5.2620e-07, 1.7136e-07, 3.8035e-06, + 1.5208e-06, 2.9430e-07, 6.3144e-07, 2.3898e-06, 3.8743e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 214.46, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4884 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.09 lr 0.00010000 +Epoch 265, weight, value: tensor([[ 0.0238, -0.0961, -0.0703, ..., -0.2136, -0.0912, -0.1730], + [-0.0801, -0.0526, -0.0172, ..., 0.0405, 0.0268, -0.1230], + [ 0.0278, -0.1516, -0.1902, ..., 0.0164, -0.1550, 0.0782], + ..., + [-0.0247, -0.0279, -0.1802, ..., -0.1634, 0.0234, 0.1458], + [ 0.0710, -0.1348, -0.1158, ..., -0.1679, -0.0997, -0.1364], + [-0.2160, 0.0222, 0.1160, ..., 0.0880, -0.0936, -0.1694]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 4.6566e-09, -2.0489e-08, ..., -8.3819e-08, + 1.4901e-08, 3.9116e-08], + [ 1.1176e-08, 9.3132e-10, 1.8626e-09, ..., -2.7940e-09, + 1.3039e-08, 1.0245e-08], + ..., + [ 9.3132e-10, -1.8626e-09, 4.1910e-08, ..., 9.3132e-08, + -7.2643e-08, -1.3225e-07], + [-2.5146e-08, 0.0000e+00, 2.7940e-09, ..., 1.6764e-08, + 1.8626e-09, 2.7940e-09], + [ 1.8626e-09, -5.5879e-09, -3.4459e-08, ..., -5.3085e-08, + 1.8626e-09, 4.6566e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0235, -0.0208, -0.0435, 0.0064, 0.0062, 0.0364, 0.0246, -0.0185, + -0.0090, -0.0124], device='cuda:0'), grad: tensor([ 2.7008e-08, -2.7474e-07, 7.3574e-08, 2.4214e-08, 9.0338e-08, + 7.6368e-08, 3.2596e-08, 1.0803e-07, -8.9407e-08, -5.8673e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 214.03, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4887 re_mapping 0.0041 re_causal 0.0128 /// teacc 99.07 lr 0.00010000 +Epoch 266, weight, value: tensor([[ 0.0240, -0.0963, -0.0705, ..., -0.2134, -0.0913, -0.1734], + [-0.0831, -0.0526, -0.0168, ..., 0.0408, 0.0270, -0.1235], + [ 0.0298, -0.1517, -0.1908, ..., 0.0164, -0.1552, 0.0786], + ..., + [-0.0250, -0.0280, -0.1810, ..., -0.1646, 0.0235, 0.1460], + [ 0.0706, -0.1350, -0.1156, ..., -0.1683, -0.1000, -0.1368], + [-0.2167, 0.0222, 0.1151, ..., 0.0873, -0.0937, -0.1695]], + device='cuda:0'), grad: tensor([[-2.1998e-06, 0.0000e+00, 1.1176e-08, ..., 1.0245e-08, + 0.0000e+00, 2.7940e-09], + [ 5.6811e-08, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 6.1467e-08, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, -2.7940e-08], + ..., + [ 1.5832e-08, 0.0000e+00, 1.1176e-08, ..., 1.3970e-08, + 0.0000e+00, 1.5832e-08], + [ 3.7253e-07, 0.0000e+00, 5.4948e-08, ..., 5.9418e-07, + 0.0000e+00, 2.7940e-09], + [ 7.3574e-08, 0.0000e+00, 1.0896e-07, ..., 1.3504e-07, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0233, -0.0217, -0.0423, 0.0063, 0.0071, 0.0364, 0.0250, -0.0192, + -0.0092, -0.0133], device='cuda:0'), grad: tensor([-5.8264e-06, 1.4622e-07, 1.3225e-07, 2.7288e-07, -3.2689e-07, + -3.8669e-06, 7.5363e-06, 9.4064e-08, 1.3439e-06, 4.7963e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 213.96, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4754 re_mapping 0.0041 re_causal 0.0125 /// teacc 99.07 lr 0.00010000 +Epoch 267, weight, value: tensor([[ 0.0240, -0.0963, -0.0707, ..., -0.2138, -0.0916, -0.1738], + [-0.0837, -0.0529, -0.0164, ..., 0.0410, 0.0271, -0.1235], + [ 0.0300, -0.1523, -0.1917, ..., 0.0164, -0.1556, 0.0789], + ..., + [-0.0234, -0.0276, -0.1813, ..., -0.1650, 0.0235, 0.1462], + [ 0.0690, -0.1354, -0.1160, ..., -0.1704, -0.1015, -0.1381], + [-0.2181, 0.0221, 0.1149, ..., 0.0871, -0.0937, -0.1698]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 0.0000e+00, 1.0245e-08, ..., 6.5193e-09, + 0.0000e+00, 2.0489e-08], + [ 8.3819e-09, 6.5193e-09, -8.3819e-09, ..., -1.3039e-08, + 0.0000e+00, 1.8347e-07], + [ 1.3039e-08, 3.7253e-09, 8.3819e-09, ..., 2.7940e-09, + 0.0000e+00, -3.0082e-07], + ..., + [ 1.8626e-09, -9.3132e-09, 2.0489e-08, ..., 2.6077e-08, + 0.0000e+00, -7.1712e-08], + [ 4.6566e-09, 9.3132e-10, 1.8626e-08, ..., 3.9116e-08, + 0.0000e+00, 2.1420e-08], + [ 4.6566e-09, -3.7253e-09, 1.2107e-08, ..., 1.7695e-08, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0232, -0.0217, -0.0420, 0.0064, 0.0072, 0.0369, 0.0249, -0.0194, + -0.0108, -0.0136], device='cuda:0'), grad: tensor([ 1.1455e-07, 2.8592e-07, -4.3679e-07, 2.0303e-07, 8.3819e-09, + -5.0291e-08, -3.0734e-07, 1.3970e-08, 1.0151e-07, 7.5437e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 214.37, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4666 re_mapping 0.0043 re_causal 0.0131 /// teacc 98.97 lr 0.00010000 +Epoch 268, weight, value: tensor([[ 0.0241, -0.0963, -0.0710, ..., -0.2141, -0.0916, -0.1741], + [-0.0844, -0.0530, -0.0161, ..., 0.0412, 0.0271, -0.1237], + [ 0.0298, -0.1524, -0.1926, ..., 0.0164, -0.1556, 0.0790], + ..., + [-0.0231, -0.0276, -0.1816, ..., -0.1653, 0.0236, 0.1463], + [ 0.0686, -0.1356, -0.1169, ..., -0.1706, -0.1017, -0.1391], + [-0.2187, 0.0221, 0.1145, ..., 0.0866, -0.0938, -0.1706]], + device='cuda:0'), grad: tensor([[ 5.0291e-08, 9.3132e-10, 1.2107e-08, ..., 4.6566e-09, + 0.0000e+00, 1.7881e-07], + [ 3.9116e-08, 2.1420e-08, -2.8871e-08, ..., -3.3528e-08, + 0.0000e+00, 4.8894e-07], + [ 1.2014e-07, 2.7940e-09, 6.5193e-09, ..., 4.6566e-09, + 0.0000e+00, 4.8149e-07], + ..., + [-7.9442e-07, -2.5146e-08, 1.2107e-08, ..., 1.3039e-08, + 0.0000e+00, -2.9001e-06], + [ 1.1176e-07, 1.8626e-09, 1.2107e-08, ..., 3.4459e-08, + 0.0000e+00, 2.5611e-07], + [ 3.2596e-08, 9.3132e-09, 8.3819e-09, ..., 4.6566e-09, + 0.0000e+00, 2.1607e-07]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0232, -0.0216, -0.0420, 0.0069, 0.0075, 0.0370, 0.0253, -0.0195, + -0.0115, -0.0142], device='cuda:0'), grad: tensor([ 3.7812e-07, 9.2015e-07, 9.5274e-07, 2.3749e-07, 1.0831e-06, + 1.8440e-07, 6.5658e-07, -5.4799e-06, 5.4948e-07, 4.8988e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.26, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4851 re_mapping 0.0040 re_causal 0.0127 /// teacc 98.97 lr 0.00010000 +Epoch 269, weight, value: tensor([[ 0.0239, -0.0963, -0.0717, ..., -0.2155, -0.0917, -0.1746], + [-0.0857, -0.0531, -0.0162, ..., 0.0412, 0.0271, -0.1239], + [ 0.0296, -0.1526, -0.1936, ..., 0.0164, -0.1556, 0.0788], + ..., + [-0.0234, -0.0276, -0.1817, ..., -0.1654, 0.0236, 0.1469], + [ 0.0681, -0.1358, -0.1180, ..., -0.1712, -0.1019, -0.1395], + [-0.2192, 0.0220, 0.1128, ..., 0.0854, -0.0938, -0.1709]], + device='cuda:0'), grad: tensor([[-2.9802e-08, 0.0000e+00, 5.5879e-09, ..., 1.3039e-08, + 0.0000e+00, 2.7008e-08], + [ 1.9558e-08, 1.8626e-09, 5.5879e-09, ..., 7.4506e-09, + 0.0000e+00, 1.0245e-07], + [ 6.5193e-09, 0.0000e+00, 4.6566e-09, ..., 6.5193e-09, + 0.0000e+00, -1.9092e-07], + ..., + [ 9.3132e-09, 1.8626e-09, 2.0489e-08, ..., 2.6077e-08, + 0.0000e+00, 2.6077e-08], + [-4.6566e-09, 9.3132e-10, 5.5879e-09, ..., 1.0524e-07, + 0.0000e+00, 5.2154e-08], + [ 8.4750e-08, -1.8626e-09, 2.6636e-07, ..., 9.1176e-07, + 0.0000e+00, 4.9360e-08]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0235, -0.0217, -0.0421, 0.0067, 0.0086, 0.0388, 0.0233, -0.0193, + -0.0120, -0.0153], device='cuda:0'), grad: tensor([-8.6613e-08, 3.2783e-07, -2.4401e-07, 4.7404e-07, -5.3644e-07, + -2.2538e-06, 2.0489e-07, 1.2573e-07, 1.1642e-07, 1.8710e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 214.16, cls_loss 0.0017 cls_loss_mapping 0.0034 cls_loss_causal 0.5146 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.09 lr 0.00010000 +Epoch 270, weight, value: tensor([[ 0.0240, -0.0965, -0.0722, ..., -0.2163, -0.0917, -0.1755], + [-0.0860, -0.0532, -0.0155, ..., 0.0414, 0.0271, -0.1269], + [ 0.0295, -0.1527, -0.1966, ..., 0.0163, -0.1558, 0.0789], + ..., + [-0.0237, -0.0275, -0.1822, ..., -0.1658, 0.0240, 0.1500], + [ 0.0678, -0.1360, -0.1198, ..., -0.1719, -0.1019, -0.1406], + [-0.2198, 0.0221, 0.1122, ..., 0.0848, -0.0938, -0.1711]], + device='cuda:0'), grad: tensor([[ 8.2981e-07, 2.2817e-07, 6.7987e-08, ..., 4.7032e-07, + 0.0000e+00, 9.3132e-10], + [ 5.0291e-08, 1.3039e-08, 1.5832e-08, ..., 3.7253e-08, + 0.0000e+00, 2.7940e-09], + [ 7.5437e-08, 1.8626e-08, 6.5193e-09, ..., 3.9116e-08, + 0.0000e+00, -1.8626e-08], + ..., + [ 8.3819e-09, 2.7940e-09, 1.6764e-08, ..., 2.0489e-08, + 0.0000e+00, 1.1176e-08], + [-2.7921e-06, -7.5251e-07, -1.7695e-07, ..., -1.5013e-06, + 0.0000e+00, 1.8626e-09], + [ 7.2364e-07, 1.9372e-07, -3.9116e-08, ..., 2.6729e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0236, -0.0235, -0.0421, 0.0067, 0.0094, 0.0390, 0.0229, -0.0168, + -0.0128, -0.0160], device='cuda:0'), grad: tensor([ 3.5875e-06, 2.4773e-07, 3.4366e-07, 8.6147e-07, 1.2573e-07, + 2.6058e-06, 1.3076e-06, 9.3132e-08, -1.2144e-05, 2.9653e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 214.18, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4671 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.03 lr 0.00010000 +Epoch 271, weight, value: tensor([[ 0.0240, -0.0972, -0.0726, ..., -0.2174, -0.0917, -0.1768], + [-0.0861, -0.0538, -0.0161, ..., 0.0413, 0.0271, -0.1270], + [ 0.0294, -0.1535, -0.1973, ..., 0.0163, -0.1558, 0.0784], + ..., + [-0.0237, -0.0274, -0.1828, ..., -0.1661, 0.0240, 0.1504], + [ 0.0680, -0.1362, -0.1188, ..., -0.1720, -0.1020, -0.1414], + [-0.2200, 0.0226, 0.1123, ..., 0.0849, -0.0938, -0.1707]], + device='cuda:0'), grad: tensor([[-2.2259e-07, 0.0000e+00, 4.0047e-08, ..., 3.6322e-08, + 9.3132e-10, 1.8626e-09], + [ 6.5193e-08, 9.3132e-10, -5.8673e-07, ..., -5.8487e-07, + 0.0000e+00, 1.7695e-08], + [ 4.9360e-08, 9.3132e-10, 2.7940e-09, ..., 2.7940e-09, + 0.0000e+00, -3.0734e-08], + ..., + [ 1.3970e-08, -9.3132e-10, 2.5332e-07, ..., 2.4866e-07, + 0.0000e+00, 9.3132e-10], + [-1.7695e-08, 0.0000e+00, 1.0990e-07, ..., 1.0710e-07, + 0.0000e+00, 6.5193e-09], + [ 1.4901e-08, -0.0000e+00, 6.3330e-08, ..., 5.4948e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0238, -0.0235, -0.0423, 0.0067, 0.0094, 0.0390, 0.0225, -0.0168, + -0.0118, -0.0159], device='cuda:0'), grad: tensor([-2.3134e-06, -8.7731e-07, 4.8988e-07, 2.6543e-07, 1.2107e-08, + 3.3528e-07, 6.0257e-07, 8.4285e-07, 3.2596e-07, 3.2689e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 214.27, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4921 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 272, weight, value: tensor([[ 0.0239, -0.0974, -0.0734, ..., -0.2190, -0.0918, -0.1777], + [-0.0863, -0.0541, -0.0155, ..., 0.0418, 0.0271, -0.1276], + [ 0.0292, -0.1539, -0.2002, ..., 0.0162, -0.1559, 0.0780], + ..., + [-0.0241, -0.0271, -0.1839, ..., -0.1674, 0.0241, 0.1512], + [ 0.0679, -0.1364, -0.1190, ..., -0.1722, -0.1021, -0.1420], + [-0.2204, 0.0226, 0.1122, ..., 0.0848, -0.0939, -0.1709]], + device='cuda:0'), grad: tensor([[-1.5832e-07, 0.0000e+00, 1.0710e-07, ..., 2.2352e-08, + 0.0000e+00, 3.7253e-08], + [ 3.1758e-07, 2.7940e-09, -2.0489e-07, ..., -2.5332e-07, + 0.0000e+00, -2.7660e-06], + [ 5.5879e-09, 0.0000e+00, 1.5832e-08, ..., 1.5832e-08, + 0.0000e+00, -1.1660e-06], + ..., + [ 1.2107e-08, 2.7940e-09, 1.3504e-07, ..., 1.5646e-07, + 0.0000e+00, 3.0305e-06], + [ 1.5832e-08, 0.0000e+00, 5.4017e-08, ..., 4.3772e-08, + 0.0000e+00, 1.3970e-08], + [ 9.3132e-09, -1.0245e-08, -3.5297e-07, ..., -3.3714e-07, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0243, -0.0234, -0.0425, 0.0068, 0.0095, 0.0396, 0.0218, -0.0169, + -0.0120, -0.0160], device='cuda:0'), grad: tensor([-7.3016e-07, -1.1697e-05, 4.1164e-07, 2.5220e-06, 7.6089e-07, + 1.2787e-06, -5.9232e-07, 8.4937e-06, 2.0862e-07, -6.1560e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 214.45, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4697 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.13 lr 0.00010000 +Epoch 273, weight, value: tensor([[ 0.0243, -0.0974, -0.0747, ..., -0.2186, -0.0919, -0.1781], + [-0.0866, -0.0544, -0.0154, ..., 0.0418, 0.0271, -0.1274], + [ 0.0291, -0.1542, -0.2007, ..., 0.0162, -0.1559, 0.0782], + ..., + [-0.0252, -0.0269, -0.1848, ..., -0.1681, 0.0239, 0.1511], + [ 0.0680, -0.1369, -0.1194, ..., -0.1723, -0.1023, -0.1428], + [-0.2218, 0.0229, 0.1123, ..., 0.0848, -0.0937, -0.1710]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 5.3551e-08, ..., 4.8894e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 6.5193e-09, -1.0766e-06, ..., -9.3505e-07, + 0.0000e+00, 4.3772e-08], + [ 1.0710e-08, 7.4506e-09, 1.2107e-08, ..., 4.1910e-08, + 0.0000e+00, 5.6345e-08], + ..., + [ 0.0000e+00, -1.8626e-08, 3.7719e-08, ..., -3.6787e-08, + 0.0000e+00, -1.4622e-07], + [-2.0023e-08, 9.3132e-10, 1.5367e-08, ..., 2.0023e-08, + 0.0000e+00, 3.7253e-09], + [ 4.1910e-09, -6.0536e-09, -2.4680e-08, ..., -2.0955e-08, + 0.0000e+00, 1.8161e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0242, -0.0233, -0.0424, 0.0068, 0.0095, 0.0394, 0.0220, -0.0173, + -0.0121, -0.0160], device='cuda:0'), grad: tensor([ 1.4342e-07, -2.7828e-06, 1.9604e-07, 4.8429e-08, 1.6857e-07, + 5.0291e-08, 2.4047e-06, -1.2573e-07, -8.1025e-08, -2.4680e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 214.17, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4629 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.02 lr 0.00010000 +Epoch 274, weight, value: tensor([[ 0.0234, -0.0974, -0.0772, ..., -0.2185, -0.0920, -0.1784], + [-0.0872, -0.0545, -0.0152, ..., 0.0422, 0.0271, -0.1275], + [ 0.0289, -0.1550, -0.2020, ..., 0.0162, -0.1558, 0.0781], + ..., + [-0.0251, -0.0267, -0.1852, ..., -0.1692, 0.0240, 0.1513], + [ 0.0682, -0.1370, -0.1196, ..., -0.1725, -0.1024, -0.1431], + [-0.2222, 0.0230, 0.1123, ..., 0.0848, -0.0937, -0.1712]], + device='cuda:0'), grad: tensor([[-1.2061e-07, 4.6566e-10, -1.6298e-08, ..., 1.2107e-08, + 0.0000e+00, 9.3132e-10], + [ 6.9849e-09, 1.0338e-07, 1.4901e-08, ..., 1.1781e-07, + 0.0000e+00, 1.8533e-07], + [ 4.0513e-08, 5.1223e-09, 6.1467e-08, ..., 4.9826e-08, + 0.0000e+00, 9.3132e-09], + ..., + [ 5.5879e-09, -1.6252e-07, -1.6764e-08, ..., -1.8906e-07, + 0.0000e+00, -2.8405e-07], + [ 2.8405e-08, 9.3132e-10, 4.6566e-09, ..., 5.0757e-08, + 0.0000e+00, 2.3283e-09], + [ 3.1199e-08, 5.6811e-08, 2.2817e-08, ..., 9.7789e-08, + 0.0000e+00, 1.0012e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0254, -0.0231, -0.0425, 0.0068, 0.0095, 0.0393, 0.0229, -0.0175, + -0.0118, -0.0160], device='cuda:0'), grad: tensor([-4.9407e-07, 7.6229e-07, 4.3819e-07, 9.4529e-08, -2.7940e-07, + -6.7102e-07, 5.8860e-07, -1.1213e-06, 9.5461e-08, 5.7369e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 213.83, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4897 re_mapping 0.0041 re_causal 0.0127 /// teacc 99.14 lr 0.00010000 +Epoch 275, weight, value: tensor([[ 0.0234, -0.0977, -0.0776, ..., -0.2191, -0.0921, -0.1787], + [-0.0875, -0.0548, -0.0154, ..., 0.0422, 0.0271, -0.1277], + [ 0.0289, -0.1552, -0.2025, ..., 0.0163, -0.1560, 0.0790], + ..., + [-0.0253, -0.0264, -0.1855, ..., -0.1694, 0.0239, 0.1514], + [ 0.0696, -0.1360, -0.1198, ..., -0.1723, -0.1025, -0.1437], + [-0.2237, 0.0229, 0.1123, ..., 0.0849, -0.0930, -0.1714]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 0.0000e+00, 5.5879e-09, ..., 4.6566e-10, + 0.0000e+00, 1.7229e-08], + [ 4.6566e-10, 4.6566e-10, -8.3819e-09, ..., -1.6764e-08, + 0.0000e+00, 6.9849e-09], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 4.6566e-10, + 0.0000e+00, -1.5600e-07], + ..., + [ 0.0000e+00, -4.6566e-10, 8.8476e-09, ..., 1.5832e-08, + 0.0000e+00, 4.6566e-09], + [-1.8626e-09, 0.0000e+00, 5.1223e-09, ..., 2.7940e-09, + 0.0000e+00, 9.7789e-09], + [ 1.8626e-09, 0.0000e+00, -3.2596e-09, ..., -1.8626e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0253, -0.0243, -0.0423, 0.0066, 0.0095, 0.0394, 0.0222, -0.0176, + -0.0086, -0.0160], device='cuda:0'), grad: tensor([ 3.3062e-08, -2.7474e-08, -2.2212e-07, 1.5888e-06, 1.6764e-08, + -1.3001e-06, -1.2433e-07, 4.7497e-08, -2.0023e-08, 1.1642e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.03, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4597 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.00 lr 0.00010000 +Epoch 276, weight, value: tensor([[ 0.0233, -0.0979, -0.0782, ..., -0.2196, -0.0923, -0.1790], + [-0.0876, -0.0551, -0.0151, ..., 0.0423, 0.0271, -0.1278], + [ 0.0287, -0.1559, -0.2031, ..., 0.0163, -0.1561, 0.0789], + ..., + [-0.0253, -0.0261, -0.1858, ..., -0.1696, 0.0240, 0.1516], + [ 0.0697, -0.1362, -0.1196, ..., -0.1717, -0.1021, -0.1445], + [-0.2241, 0.0231, 0.1124, ..., 0.0849, -0.0930, -0.1720]], + device='cuda:0'), grad: tensor([[-9.3132e-09, 0.0000e+00, 1.2107e-08, ..., 6.0536e-09, + 0.0000e+00, 4.6566e-09], + [ 6.9849e-09, 1.3970e-09, -7.9162e-09, ..., 2.2817e-08, + 0.0000e+00, 2.1001e-07], + [ 7.4506e-09, 4.6566e-10, 7.4506e-09, ..., -8.3819e-09, + 0.0000e+00, 1.0710e-08], + ..., + [ 5.5879e-09, 3.2596e-09, 4.5169e-08, ..., 9.3132e-09, + 0.0000e+00, -2.8592e-07], + [-1.8720e-07, 9.3132e-10, -4.0000e-07, ..., 9.3132e-09, + 0.0000e+00, -2.7008e-08], + [ 1.0710e-08, 4.6566e-10, -3.3947e-07, ..., -3.7346e-07, + 0.0000e+00, 4.7497e-08]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0256, -0.0242, -0.0424, 0.0066, 0.0094, 0.0392, 0.0227, -0.0176, + -0.0082, -0.0159], device='cuda:0'), grad: tensor([ 2.2352e-08, 4.4843e-07, 6.9384e-08, 6.4261e-08, 5.0059e-07, + 4.5169e-07, 1.3094e-06, -5.3272e-07, -1.8077e-06, -5.2573e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 214.32, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4783 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.05 lr 0.00010000 +Epoch 277, weight, value: tensor([[ 0.0234, -0.0987, -0.0787, ..., -0.2202, -0.0923, -0.1794], + [-0.0877, -0.0553, -0.0145, ..., 0.0425, 0.0271, -0.1278], + [ 0.0285, -0.1562, -0.2055, ..., 0.0161, -0.1561, 0.0790], + ..., + [-0.0255, -0.0273, -0.1862, ..., -0.1714, 0.0240, 0.1516], + [ 0.0695, -0.1364, -0.1203, ..., -0.1721, -0.1023, -0.1450], + [-0.2248, 0.0243, 0.1123, ..., 0.0850, -0.0930, -0.1720]], + device='cuda:0'), grad: tensor([[-2.8498e-07, 4.6566e-10, 2.7940e-09, ..., 3.7253e-09, + 0.0000e+00, 1.3970e-09], + [ 6.5193e-09, 7.4506e-09, 2.9802e-08, ..., 3.5856e-08, + 0.0000e+00, 3.0734e-08], + [ 4.6566e-10, 1.3970e-08, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 4.7963e-08], + ..., + [ 1.3970e-09, -2.8871e-08, 4.7963e-07, ..., 4.4098e-07, + 0.0000e+00, 1.1083e-07], + [ 6.7521e-08, 1.8626e-09, 7.4506e-09, ..., 3.1618e-07, + 0.0000e+00, 6.9849e-09], + [ 7.4506e-09, 5.5879e-09, 1.0338e-07, ..., 9.4529e-08, + 0.0000e+00, 6.6590e-08]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0256, -0.0231, -0.0433, 0.0065, 0.0094, 0.0392, 0.0225, -0.0182, + -0.0088, -0.0159], device='cuda:0'), grad: tensor([-8.6706e-07, 1.4482e-07, 7.8697e-08, 2.3516e-07, -1.8086e-06, + -7.2084e-07, 8.7125e-07, 1.1837e-06, 5.2992e-07, 3.6601e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 214.02, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4769 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.05 lr 0.00010000 +Epoch 278, weight, value: tensor([[ 0.0235, -0.0988, -0.0789, ..., -0.2204, -0.0925, -0.1798], + [-0.0879, -0.0557, -0.0145, ..., 0.0425, 0.0269, -0.1280], + [ 0.0284, -0.1566, -0.2058, ..., 0.0161, -0.1561, 0.0796], + ..., + [-0.0256, -0.0276, -0.1871, ..., -0.1721, 0.0241, 0.1516], + [ 0.0698, -0.1365, -0.1202, ..., -0.1721, -0.1023, -0.1451], + [-0.2255, 0.0245, 0.1128, ..., 0.0858, -0.0931, -0.1721]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.4435e-08, 5.5879e-09, ..., 1.3970e-09, + 1.3970e-09, 3.6787e-08], + [ 3.7253e-09, 1.0338e-07, -2.1886e-08, ..., -2.7008e-08, + 1.5367e-08, 2.3376e-07], + [ 3.7253e-09, 4.2375e-08, 5.1223e-09, ..., 4.1910e-09, + 1.3970e-09, 6.4261e-08], + ..., + [ 1.3970e-09, -2.6403e-07, 1.6764e-08, ..., 1.7695e-08, + -3.3528e-08, -5.7928e-07], + [-5.8953e-07, 4.8429e-08, -1.3132e-06, ..., -7.9721e-07, + 1.3970e-09, 1.1688e-07], + [ 9.3132e-10, 3.6787e-08, -3.7253e-09, ..., -6.5193e-09, + 4.1910e-09, 8.5216e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0256, -0.0231, -0.0432, 0.0065, 0.0089, 0.0390, 0.0221, -0.0185, + -0.0086, -0.0152], device='cuda:0'), grad: tensor([ 1.1036e-07, 5.4203e-07, 2.4587e-07, -1.9595e-06, 1.5786e-07, + 6.6794e-06, 1.6410e-06, -1.3877e-06, -6.2324e-06, 2.1933e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 214.15, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4932 re_mapping 0.0040 re_causal 0.0127 /// teacc 98.99 lr 0.00010000 +Epoch 279, weight, value: tensor([[ 0.0234, -0.0995, -0.0796, ..., -0.2216, -0.0925, -0.1813], + [-0.0880, -0.0563, -0.0144, ..., 0.0423, 0.0269, -0.1281], + [ 0.0283, -0.1572, -0.2061, ..., 0.0161, -0.1563, 0.0796], + ..., + [-0.0252, -0.0277, -0.1876, ..., -0.1730, 0.0242, 0.1519], + [ 0.0699, -0.1369, -0.1203, ..., -0.1723, -0.1026, -0.1462], + [-0.2257, 0.0253, 0.1128, ..., 0.0859, -0.0932, -0.1722]], + device='cuda:0'), grad: tensor([[-4.1910e-08, 1.3970e-09, 9.3132e-09, ..., 2.3283e-09, + 0.0000e+00, 1.4901e-08], + [ 5.1223e-09, 1.3039e-08, -1.5367e-08, ..., -8.8476e-09, + 0.0000e+00, 7.1246e-08], + [ 1.3970e-09, 8.3819e-09, 1.8626e-09, ..., 2.3283e-09, + 0.0000e+00, -8.7544e-08], + ..., + [ 1.3970e-09, -4.6100e-08, 1.9092e-08, ..., 1.1642e-08, + 0.0000e+00, -7.7300e-08], + [-0.0000e+00, 9.3132e-10, 8.8476e-09, ..., 6.0536e-09, + 0.0000e+00, 2.1886e-08], + [ 5.5879e-09, 7.4506e-09, -3.2596e-08, ..., -4.3772e-08, + 0.0000e+00, 3.8184e-08]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0260, -0.0231, -0.0432, 0.0064, 0.0089, 0.0391, 0.0221, -0.0186, + -0.0086, -0.0152], device='cuda:0'), grad: tensor([-1.7509e-07, 1.4203e-07, -1.4389e-07, -1.7835e-07, 5.9605e-08, + 1.3923e-07, 1.3737e-07, -9.0804e-08, 6.1467e-08, 5.0757e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 214.17, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.5023 re_mapping 0.0041 re_causal 0.0128 /// teacc 98.99 lr 0.00010000 +Epoch 280, weight, value: tensor([[ 0.0234, -0.0997, -0.0800, ..., -0.2222, -0.0931, -0.1820], + [-0.0881, -0.0586, -0.0162, ..., 0.0422, 0.0268, -0.1295], + [ 0.0283, -0.1572, -0.2069, ..., 0.0160, -0.1563, 0.0800], + ..., + [-0.0253, -0.0269, -0.1878, ..., -0.1736, 0.0244, 0.1529], + [ 0.0699, -0.1372, -0.1204, ..., -0.1724, -0.1026, -0.1470], + [-0.2259, 0.0262, 0.1128, ..., 0.0858, -0.0932, -0.1722]], + device='cuda:0'), grad: tensor([[ 5.0757e-08, 5.1223e-09, 1.3970e-07, ..., 2.7148e-07, + 6.9849e-09, 9.3132e-09], + [ 1.1316e-07, 7.3574e-08, -8.9407e-05, ..., -1.2445e-04, + 2.4214e-08, 1.1967e-07], + [ 1.1176e-08, 7.6834e-08, 1.3039e-07, ..., 1.9418e-07, + 1.3970e-09, 1.0896e-07], + ..., + [ 5.1223e-09, 2.4680e-08, 6.4254e-05, ..., 8.9645e-05, + 0.0000e+00, 4.2841e-08], + [-4.4610e-07, 1.6298e-08, 1.1781e-07, ..., 5.4250e-07, + 2.9802e-08, 8.3819e-09], + [ 5.5879e-09, -1.9092e-08, 7.1824e-06, ..., 1.0118e-05, + 4.6566e-10, 8.8476e-09]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0260, -0.0247, -0.0434, 0.0065, 0.0090, 0.0390, 0.0222, -0.0170, + -0.0087, -0.0152], device='cuda:0'), grad: tensor([ 9.0152e-07, -1.9717e-04, 7.5251e-07, 1.9632e-06, 3.8445e-05, + 6.1691e-05, -6.2466e-05, 1.4329e-04, -3.8892e-06, 1.6108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.26, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4671 re_mapping 0.0043 re_causal 0.0119 /// teacc 98.93 lr 0.00010000 +Epoch 281, weight, value: tensor([[ 0.0235, -0.1001, -0.0821, ..., -0.2214, -0.0934, -0.1829], + [-0.0882, -0.0603, -0.0155, ..., 0.0428, 0.0269, -0.1299], + [ 0.0282, -0.1560, -0.2082, ..., 0.0160, -0.1564, 0.0826], + ..., + [-0.0254, -0.0268, -0.1908, ..., -0.1761, 0.0244, 0.1525], + [ 0.0702, -0.1375, -0.1209, ..., -0.1726, -0.1029, -0.1480], + [-0.2262, 0.0271, 0.1128, ..., 0.0858, -0.0932, -0.1723]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 4.6566e-10, 4.9826e-08, ..., 6.9849e-09, + 1.3970e-09, 2.6543e-08], + [ 4.1910e-09, 1.0710e-08, -1.0664e-07, ..., -1.0943e-07, + -2.3283e-09, 8.4750e-08], + [ 2.3283e-09, 9.3132e-10, 1.3970e-08, ..., 2.7940e-09, + 9.3132e-10, -1.9139e-07], + ..., + [ 4.6566e-10, -7.9162e-09, 9.2201e-08, ..., 8.9873e-08, + -2.3283e-09, -2.7940e-09], + [ 3.7253e-09, 2.7940e-09, 4.6100e-08, ..., 1.1129e-07, + 1.3970e-09, 3.1665e-08], + [ 2.3283e-09, 1.8626e-09, 3.3062e-08, ..., 6.3330e-08, + 9.3132e-10, 8.3819e-09]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0258, -0.0249, -0.0429, 0.0058, 0.0091, 0.0392, 0.0222, -0.0176, + -0.0081, -0.0152], device='cuda:0'), grad: tensor([ 1.9372e-07, -1.7136e-07, -2.5239e-07, 2.5099e-07, 2.1420e-07, + -4.0606e-07, -4.7684e-07, 2.6682e-07, 2.1560e-07, 1.5972e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 214.39, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.4728 re_mapping 0.0046 re_causal 0.0125 /// teacc 99.00 lr 0.00010000 +Epoch 282, weight, value: tensor([[ 0.0236, -0.1010, -0.0826, ..., -0.2225, -0.0934, -0.1844], + [-0.0883, -0.0589, -0.0145, ..., 0.0434, 0.0301, -0.1301], + [ 0.0281, -0.1578, -0.2105, ..., 0.0159, -0.1563, 0.0816], + ..., + [-0.0256, -0.0278, -0.1935, ..., -0.1784, 0.0215, 0.1532], + [ 0.0696, -0.1385, -0.1223, ..., -0.1735, -0.1033, -0.1484], + [-0.2267, 0.0275, 0.1129, ..., 0.0860, -0.0936, -0.1728]], + device='cuda:0'), grad: tensor([[-1.0803e-07, 1.7695e-08, 9.6392e-08, ..., 2.1420e-08, + 0.0000e+00, 9.3132e-10], + [ 1.0245e-08, 6.5193e-09, 1.3504e-08, ..., 3.7253e-09, + 0.0000e+00, 1.2107e-08], + [ 2.7474e-08, 2.3283e-09, 1.0245e-08, ..., 4.6566e-09, + 0.0000e+00, -3.6322e-08], + ..., + [ 4.6566e-09, 1.3970e-08, 5.6811e-08, ..., 3.6322e-08, + 0.0000e+00, 1.3039e-08], + [ 1.3970e-08, 1.4435e-08, 5.3085e-08, ..., 3.2596e-08, + 0.0000e+00, 9.7789e-09], + [ 1.4435e-08, -3.3947e-07, -1.7760e-06, ..., -8.5030e-07, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0258, -0.0243, -0.0433, 0.0064, 0.0090, 0.0389, 0.0223, -0.0180, + -0.0084, -0.0151], device='cuda:0'), grad: tensor([-4.9500e-07, 1.3039e-07, 1.6484e-07, 1.4156e-07, 2.8946e-06, + -3.9581e-08, 1.7183e-07, 1.7742e-07, 7.3109e-08, -3.2373e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 214.76, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4518 re_mapping 0.0046 re_causal 0.0126 /// teacc 99.11 lr 0.00010000 +Epoch 283, weight, value: tensor([[ 0.0237, -0.1015, -0.0829, ..., -0.2235, -0.0934, -0.1853], + [-0.0886, -0.0593, -0.0158, ..., 0.0430, 0.0304, -0.1301], + [ 0.0280, -0.1591, -0.2104, ..., 0.0159, -0.1566, 0.0823], + ..., + [-0.0259, -0.0281, -0.1939, ..., -0.1787, 0.0212, 0.1530], + [ 0.0701, -0.1408, -0.1212, ..., -0.1735, -0.1041, -0.1497], + [-0.2295, 0.0308, 0.1134, ..., 0.0866, -0.0937, -0.1732]], + device='cuda:0'), grad: tensor([[-5.1223e-09, 2.3283e-09, 1.3970e-09, ..., 2.3283e-09, + 0.0000e+00, 5.1688e-08], + [ 9.3132e-10, 5.2154e-07, 3.9907e-07, ..., 1.5413e-06, + 0.0000e+00, 1.2722e-06], + [ 4.6566e-10, -2.9802e-08, 5.1223e-09, ..., 2.0955e-08, + 0.0000e+00, -1.0915e-06], + ..., + [ 4.6566e-10, -6.6869e-07, -4.8941e-07, ..., -1.9483e-06, + 0.0000e+00, -1.0151e-06], + [ 1.3970e-08, 1.1642e-08, 4.1444e-08, ..., 1.1269e-07, + 0.0000e+00, 1.8394e-07], + [ 4.1910e-09, 1.4808e-07, -4.6194e-07, ..., -7.0781e-08, + 0.0000e+00, 2.6403e-07]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0256, -0.0244, -0.0430, 0.0067, 0.0086, 0.0389, 0.0222, -0.0183, + -0.0081, -0.0147], device='cuda:0'), grad: tensor([ 1.7928e-07, 7.0222e-06, -4.0233e-06, 5.0478e-07, 9.4343e-07, + -1.4389e-07, 6.2166e-07, -6.5640e-06, 8.7451e-07, 5.8953e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.00, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4791 re_mapping 0.0042 re_causal 0.0120 /// teacc 99.06 lr 0.00010000 +Epoch 284, weight, value: tensor([[ 0.0236, -0.1025, -0.0852, ..., -0.2241, -0.0936, -0.1869], + [-0.0887, -0.0597, -0.0160, ..., 0.0430, 0.0305, -0.1312], + [ 0.0279, -0.1599, -0.2112, ..., 0.0159, -0.1572, 0.0832], + ..., + [-0.0260, -0.0276, -0.1941, ..., -0.1787, 0.0214, 0.1543], + [ 0.0703, -0.1407, -0.1233, ..., -0.1743, -0.1051, -0.1532], + [-0.2302, 0.0309, 0.1136, ..., 0.0868, -0.0943, -0.1739]], + device='cuda:0'), grad: tensor([[-9.1270e-08, 0.0000e+00, 2.7008e-08, ..., 9.3132e-10, + 0.0000e+00, 4.4703e-08], + [ 1.4435e-08, 2.3283e-09, -5.8208e-08, ..., -8.4750e-08, + 0.0000e+00, 5.2620e-08], + [ 5.5879e-09, 4.6566e-10, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, -3.6834e-07], + ..., + [ 2.4680e-08, -0.0000e+00, 5.4948e-08, ..., 7.2177e-08, + 0.0000e+00, 1.4296e-07], + [ 1.1642e-08, 4.6566e-10, 6.5193e-09, ..., 4.1910e-09, + 0.0000e+00, 4.9826e-08], + [ 1.2107e-08, 2.7940e-09, -8.8476e-09, ..., -1.2573e-08, + 0.0000e+00, 1.3504e-08]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0270, -0.0249, -0.0426, 0.0067, 0.0086, 0.0387, 0.0226, -0.0176, + -0.0089, -0.0145], device='cuda:0'), grad: tensor([-5.7509e-07, -7.6368e-08, -7.1619e-07, 2.2072e-07, 3.5856e-08, + 2.0210e-07, -4.0978e-08, 6.7055e-07, 1.7975e-07, 1.1222e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.25, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4797 re_mapping 0.0041 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 285, weight, value: tensor([[ 0.0236, -0.1032, -0.0856, ..., -0.2248, -0.0936, -0.1876], + [-0.0889, -0.0599, -0.0158, ..., 0.0431, 0.0306, -0.1313], + [ 0.0277, -0.1611, -0.2116, ..., 0.0159, -0.1570, 0.0798], + ..., + [-0.0261, -0.0272, -0.1944, ..., -0.1790, 0.0214, 0.1564], + [ 0.0713, -0.1385, -0.1230, ..., -0.1738, -0.1053, -0.1540], + [-0.2316, 0.0307, 0.1136, ..., 0.0868, -0.0944, -0.1741]], + device='cuda:0'), grad: tensor([[-2.6543e-08, 2.3283e-09, 5.1223e-09, ..., 8.8476e-09, + 0.0000e+00, 7.9162e-09], + [ 2.7940e-09, 2.6077e-08, -4.4238e-08, ..., -4.1910e-09, + 0.0000e+00, 8.2888e-08], + [ 1.8626e-09, 2.9337e-08, 5.3085e-08, ..., 1.1828e-07, + 0.0000e+00, -2.2817e-08], + ..., + [ 9.3132e-10, 2.5285e-07, 5.2201e-07, ..., 1.1977e-06, + 0.0000e+00, -8.7544e-08], + [ 5.1223e-09, 1.3178e-07, 2.4261e-07, ..., 5.6997e-07, + 0.0000e+00, 2.4214e-08], + [ 7.9162e-09, -6.2911e-07, -1.0645e-06, ..., -2.6934e-06, + 0.0000e+00, 1.6764e-08]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0273, -0.0248, -0.0442, 0.0067, 0.0086, 0.0384, 0.0226, -0.0169, + -0.0086, -0.0145], device='cuda:0'), grad: tensor([-1.2154e-07, 5.1688e-08, 2.4401e-07, 4.0513e-08, 1.6280e-06, + 1.0571e-07, 4.5169e-08, 2.6692e-06, 1.4100e-06, -6.0536e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 214.36, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4698 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.06 lr 0.00010000 +Epoch 286, weight, value: tensor([[ 0.0238, -0.1035, -0.0877, ..., -0.2248, -0.0936, -0.1880], + [-0.0892, -0.0600, -0.0156, ..., 0.0432, 0.0306, -0.1318], + [ 0.0275, -0.1614, -0.2123, ..., 0.0159, -0.1568, 0.0798], + ..., + [-0.0264, -0.0271, -0.1947, ..., -0.1793, 0.0214, 0.1570], + [ 0.0713, -0.1388, -0.1235, ..., -0.1741, -0.1071, -0.1550], + [-0.2319, 0.0308, 0.1137, ..., 0.0868, -0.0944, -0.1745]], + device='cuda:0'), grad: tensor([[-7.4506e-09, 1.3970e-09, 2.7474e-08, ..., 9.3132e-09, + 0.0000e+00, 1.1642e-08], + [ 6.9849e-09, 6.8452e-08, -1.7043e-07, ..., 7.1432e-07, + 0.0000e+00, 1.6242e-06], + [ 3.2596e-09, 1.2992e-07, 9.7789e-09, ..., -1.0757e-06, + 0.0000e+00, -1.6131e-06], + ..., + [ 1.8626e-09, -1.0841e-06, 1.3597e-07, ..., 2.8033e-07, + 0.0000e+00, -1.6810e-06], + [-4.6566e-09, 3.2596e-09, 3.2131e-08, ..., 2.4214e-08, + 0.0000e+00, 2.0489e-08], + [ 3.7253e-09, 8.6986e-07, -1.2061e-07, ..., -9.0804e-08, + 0.0000e+00, 1.6149e-06]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0284, -0.0250, -0.0443, 0.0066, 0.0085, 0.0384, 0.0228, -0.0167, + -0.0090, -0.0144], device='cuda:0'), grad: tensor([ 5.9605e-08, 5.0701e-06, -6.0499e-06, -4.2841e-08, 3.5716e-07, + 9.2667e-08, -1.6810e-07, -3.7551e-06, -2.6543e-08, 4.4703e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 214.32, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4746 re_mapping 0.0040 re_causal 0.0122 /// teacc 99.01 lr 0.00010000 +Epoch 287, weight, value: tensor([[ 0.0240, -0.1041, -0.0880, ..., -0.2253, -0.0936, -0.1893], + [-0.0894, -0.0602, -0.0152, ..., 0.0434, 0.0306, -0.1323], + [ 0.0273, -0.1619, -0.2129, ..., 0.0159, -0.1567, 0.0799], + ..., + [-0.0266, -0.0268, -0.1957, ..., -0.1800, 0.0214, 0.1573], + [ 0.0718, -0.1395, -0.1237, ..., -0.1743, -0.1071, -0.1557], + [-0.2323, 0.0307, 0.1137, ..., 0.0869, -0.0944, -0.1751]], + device='cuda:0'), grad: tensor([[-5.2620e-08, -2.9337e-08, 8.6147e-08, ..., 6.0536e-08, + 2.7940e-09, 1.8626e-08], + [ 1.2573e-08, 1.3039e-08, -2.2026e-07, ..., -1.7975e-07, + 1.8626e-09, 3.6322e-08], + [ 1.1176e-08, 6.5193e-09, 4.8429e-08, ..., -2.2817e-08, + -1.9092e-08, -1.2945e-07], + ..., + [ 2.7940e-09, -2.0489e-08, 3.2084e-07, ..., 2.4820e-07, + 2.7940e-09, -2.7008e-08], + [-9.3132e-09, 8.3819e-09, 1.0384e-07, ..., 7.1712e-08, + 1.3970e-09, 2.6077e-08], + [ 2.7940e-09, -4.1910e-09, 2.4512e-06, ..., 1.6075e-06, + 3.2596e-09, 2.4214e-08]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0284, -0.0249, -0.0448, 0.0073, 0.0085, 0.0381, 0.0225, -0.0165, + -0.0089, -0.0144], device='cuda:0'), grad: tensor([-3.9581e-08, -4.3865e-07, -5.2620e-08, 1.6624e-07, -5.5768e-06, + 3.3434e-07, 7.1945e-07, 9.4436e-07, -7.7439e-07, 4.7274e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 214.24, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4731 re_mapping 0.0040 re_causal 0.0120 /// teacc 98.97 lr 0.00010000 +Epoch 288, weight, value: tensor([[ 0.0242, -0.1042, -0.0884, ..., -0.2260, -0.0936, -0.1899], + [-0.0897, -0.0605, -0.0150, ..., 0.0437, 0.0306, -0.1332], + [ 0.0269, -0.1624, -0.2135, ..., 0.0159, -0.1567, 0.0800], + ..., + [-0.0272, -0.0267, -0.1962, ..., -0.1805, 0.0214, 0.1580], + [ 0.0720, -0.1399, -0.1254, ..., -0.1753, -0.1072, -0.1563], + [-0.2326, 0.0307, 0.1139, ..., 0.0871, -0.0944, -0.1754]], + device='cuda:0'), grad: tensor([[-3.2596e-09, 4.6566e-09, 5.1223e-09, ..., 4.6566e-10, + 0.0000e+00, 3.2596e-09], + [ 2.3283e-09, 3.7253e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08], + [ 3.1665e-08, 5.5879e-09, 2.3283e-09, ..., 4.6566e-10, + 0.0000e+00, -2.1886e-08], + ..., + [ 1.3970e-09, -1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -8.8476e-09], + [-4.8429e-08, 6.0536e-09, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 2.0955e-08], + [ 7.9162e-09, 2.7940e-09, -4.1910e-09, ..., 1.4435e-08, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0284, -0.0252, -0.0447, 0.0072, 0.0084, 0.0383, 0.0217, -0.0163, + -0.0092, -0.0142], device='cuda:0'), grad: tensor([-1.3504e-08, 6.0536e-08, 3.3388e-07, -8.3819e-09, 1.1642e-08, + 5.9605e-08, -3.3528e-08, -1.3970e-08, -4.5635e-07, 6.7987e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 214.50, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4728 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.01 lr 0.00010000 +Epoch 289, weight, value: tensor([[ 0.0241, -0.1044, -0.0889, ..., -0.2264, -0.0936, -0.1903], + [-0.0898, -0.0606, -0.0149, ..., 0.0436, 0.0306, -0.1334], + [ 0.0267, -0.1636, -0.2142, ..., 0.0160, -0.1573, 0.0799], + ..., + [-0.0274, -0.0264, -0.1964, ..., -0.1806, 0.0214, 0.1582], + [ 0.0721, -0.1403, -0.1255, ..., -0.1754, -0.1072, -0.1567], + [-0.2328, 0.0307, 0.1139, ..., 0.0871, -0.0944, -0.1757]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.2596e-09, 8.8476e-09, ..., 1.0710e-08, + 0.0000e+00, 5.5879e-09], + [ 9.3132e-10, 2.8871e-08, -2.5379e-07, ..., -1.8626e-09, + 9.3132e-10, 7.4506e-08], + [ 4.6566e-10, 8.8476e-09, 2.2352e-08, ..., 3.7253e-09, + 0.0000e+00, 7.9162e-09], + ..., + [ 1.3970e-09, -1.6298e-08, 1.3551e-07, ..., 1.4901e-08, + -1.3970e-09, -6.7987e-08], + [-9.3132e-10, 7.9162e-09, 8.4750e-08, ..., 1.4901e-08, + 0.0000e+00, 1.3039e-08], + [ 7.4506e-09, 1.3504e-08, 4.4238e-08, ..., 3.8184e-08, + 4.6566e-10, 3.0268e-08]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0287, -0.0253, -0.0446, 0.0073, 0.0084, 0.0383, 0.0216, -0.0163, + -0.0091, -0.0142], device='cuda:0'), grad: tensor([ 8.7544e-08, -2.2836e-06, 2.1420e-07, -2.0862e-07, -1.4855e-07, + -3.1851e-07, 3.1292e-07, 1.1055e-06, 9.9558e-07, 2.4587e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 214.41, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4453 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.01 lr 0.00010000 +Epoch 290, weight, value: tensor([[ 0.0244, -0.1049, -0.0900, ..., -0.2269, -0.0936, -0.1908], + [-0.0905, -0.0612, -0.0151, ..., 0.0437, 0.0306, -0.1335], + [ 0.0265, -0.1639, -0.2147, ..., 0.0160, -0.1574, 0.0800], + ..., + [-0.0278, -0.0263, -0.1966, ..., -0.1807, 0.0214, 0.1582], + [ 0.0721, -0.1409, -0.1256, ..., -0.1758, -0.1073, -0.1571], + [-0.2331, 0.0306, 0.1143, ..., 0.0875, -0.0944, -0.1760]], + device='cuda:0'), grad: tensor([[-4.1444e-08, 1.5367e-08, 1.1176e-08, ..., 2.2352e-08, + 4.6566e-10, 6.0070e-08], + [ 6.0536e-09, 5.2620e-08, -4.1910e-09, ..., 6.7055e-08, + 1.8626e-09, 1.8487e-07], + [ 1.8626e-09, 2.0489e-08, 5.5879e-09, ..., 3.2131e-08, + 0.0000e+00, -2.9337e-08], + ..., + [-1.8626e-09, 2.7474e-08, 1.1316e-07, ..., 3.0035e-07, + -5.1223e-09, 9.9186e-08], + [ 1.1176e-08, 8.2422e-08, 1.0245e-07, ..., 2.8918e-07, + 0.0000e+00, 2.7148e-07], + [ 1.3504e-08, 5.4017e-08, -2.8312e-07, ..., -6.1840e-07, + 2.3283e-09, 1.8440e-07]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0286, -0.0255, -0.0446, 0.0077, 0.0081, 0.0386, 0.0210, -0.0165, + -0.0085, -0.0138], device='cuda:0'), grad: tensor([-4.2841e-08, 5.6345e-07, -2.5611e-08, -8.7768e-06, 1.4231e-06, + 6.5789e-06, -9.9838e-07, 1.0822e-06, 1.4501e-06, -1.2442e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 214.10, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4838 re_mapping 0.0039 re_causal 0.0122 /// teacc 98.77 lr 0.00010000 +Epoch 291, weight, value: tensor([[ 0.0257, -0.1046, -0.0896, ..., -0.2275, -0.0937, -0.1912], + [-0.0917, -0.0613, -0.0149, ..., 0.0438, 0.0306, -0.1336], + [ 0.0262, -0.1636, -0.2156, ..., 0.0160, -0.1570, 0.0801], + ..., + [-0.0282, -0.0259, -0.1969, ..., -0.1809, 0.0215, 0.1583], + [ 0.0722, -0.1412, -0.1283, ..., -0.1779, -0.1080, -0.1578], + [-0.2343, 0.0304, 0.1156, ..., 0.0892, -0.0946, -0.1764]], + device='cuda:0'), grad: tensor([[-1.8626e-08, 1.3970e-09, 5.5879e-09, ..., 6.9849e-09, + 0.0000e+00, 5.5879e-09], + [ 1.8626e-09, 3.4925e-08, -5.0105e-07, ..., -2.2631e-07, + 0.0000e+00, 7.2643e-08], + [ 1.8626e-09, 1.1176e-08, 1.7229e-08, ..., -1.8720e-07, + 0.0000e+00, -1.1781e-07], + ..., + [ 2.3283e-09, -4.4238e-08, 7.6368e-08, ..., 4.2841e-08, + 0.0000e+00, -9.8720e-08], + [ 1.8626e-08, 2.2352e-08, 2.3842e-07, ..., 1.5507e-07, + 0.0000e+00, 3.7253e-08], + [ 6.5193e-09, 7.4506e-09, 3.2596e-09, ..., 1.8766e-07, + 0.0000e+00, 1.4435e-07]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0259, -0.0255, -0.0445, 0.0075, 0.0064, 0.0385, 0.0196, -0.0165, + -0.0095, -0.0121], device='cuda:0'), grad: tensor([-3.4459e-08, -1.4063e-06, -1.8179e-06, -1.0058e-07, 2.4727e-07, + -7.2177e-08, 3.0827e-07, -1.4435e-08, 9.8627e-07, 1.9073e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 214.13, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4590 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.03 lr 0.00010000 +Epoch 292, weight, value: tensor([[ 0.0254, -0.1049, -0.0896, ..., -0.2278, -0.0938, -0.1926], + [-0.0923, -0.0615, -0.0133, ..., 0.0438, 0.0320, -0.1337], + [ 0.0257, -0.1637, -0.2160, ..., 0.0159, -0.1563, 0.0803], + ..., + [-0.0287, -0.0262, -0.1988, ..., -0.1810, 0.0201, 0.1584], + [ 0.0726, -0.1415, -0.1309, ..., -0.1798, -0.1094, -0.1586], + [-0.2348, 0.0295, 0.1157, ..., 0.0891, -0.0942, -0.1772]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 3.2596e-09, ..., 4.7497e-08, + 0.0000e+00, 1.5832e-08], + [ 0.0000e+00, 9.3132e-10, -3.1199e-08, ..., -1.4007e-06, + 0.0000e+00, 6.0536e-09], + [ 4.6566e-10, -1.4901e-08, 2.7940e-09, ..., 6.8918e-07, + 0.0000e+00, -3.3202e-07], + ..., + [ 0.0000e+00, 1.3970e-09, 4.3772e-08, ..., 6.2399e-07, + 0.0000e+00, 7.4506e-09], + [-1.3970e-09, 2.7940e-09, 1.7229e-08, ..., 1.6298e-08, + 0.0000e+00, 4.6566e-08], + [ 4.6566e-10, 9.3132e-10, -3.6322e-08, ..., -4.2375e-08, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0262, -0.0243, -0.0443, 0.0047, 0.0065, 0.0413, 0.0192, -0.0178, + -0.0106, -0.0121], device='cuda:0'), grad: tensor([ 5.6904e-07, -1.4760e-05, 5.4464e-06, 1.2200e-06, 2.2817e-08, + 5.6252e-07, 1.5367e-07, 6.3293e-06, 4.4703e-07, -5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 214.16, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4610 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.08 lr 0.00010000 +Epoch 293, weight, value: tensor([[ 0.0246, -0.1051, -0.0898, ..., -0.2281, -0.0938, -0.1932], + [-0.0925, -0.0617, -0.0132, ..., 0.0439, 0.0320, -0.1338], + [ 0.0255, -0.1639, -0.2166, ..., 0.0159, -0.1555, 0.0806], + ..., + [-0.0289, -0.0250, -0.1992, ..., -0.1812, 0.0201, 0.1584], + [ 0.0725, -0.1418, -0.1312, ..., -0.1801, -0.1107, -0.1607], + [-0.2351, 0.0279, 0.1157, ..., 0.0891, -0.0942, -0.1781]], + device='cuda:0'), grad: tensor([[ 1.9395e-07, 0.0000e+00, 1.5204e-07, ..., 8.8941e-08, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 8.3819e-09, -1.0384e-07, ..., -1.2503e-07, + 0.0000e+00, 1.2573e-08], + [ 6.2864e-09, 4.6566e-10, 6.7521e-09, ..., 8.6147e-09, + 0.0000e+00, 6.9849e-10], + ..., + [ 9.3132e-10, -1.6065e-08, 3.7486e-08, ..., 4.0513e-08, + 0.0000e+00, -2.8638e-08], + [ 5.7276e-08, 2.3283e-10, 3.8184e-08, ..., 6.2166e-08, + 0.0000e+00, 2.3283e-10], + [ 1.7462e-08, 6.9849e-09, 3.5157e-08, ..., 3.6787e-08, + 0.0000e+00, 1.6298e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0269, -0.0242, -0.0441, 0.0045, 0.0065, 0.0416, 0.0197, -0.0176, + -0.0111, -0.0122], device='cuda:0'), grad: tensor([ 1.1157e-06, -2.4284e-07, 5.7975e-08, 1.2037e-07, 1.3644e-07, + 3.4086e-06, -5.0217e-06, 2.9569e-08, -1.0035e-07, 4.9919e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 214.24, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4860 re_mapping 0.0038 re_causal 0.0116 /// teacc 99.07 lr 0.00010000 +Epoch 294, weight, value: tensor([[ 0.0247, -0.1052, -0.0899, ..., -0.2290, -0.0939, -0.1936], + [-0.0927, -0.0617, -0.0130, ..., 0.0438, 0.0319, -0.1340], + [ 0.0254, -0.1640, -0.2160, ..., 0.0159, -0.1553, 0.0808], + ..., + [-0.0292, -0.0245, -0.1993, ..., -0.1812, 0.0202, 0.1585], + [ 0.0736, -0.1416, -0.1313, ..., -0.1798, -0.1108, -0.1612], + [-0.2371, 0.0275, 0.1157, ..., 0.0891, -0.0942, -0.1786]], + device='cuda:0'), grad: tensor([[-3.1199e-08, 0.0000e+00, 1.7928e-08, ..., 1.3970e-09, + 0.0000e+00, 3.7253e-09], + [ 4.1910e-09, 6.9849e-10, -3.9581e-08, ..., -2.7940e-08, + 0.0000e+00, 2.0955e-09], + [ 9.3132e-10, 2.3283e-10, 2.3283e-09, ..., 1.6298e-09, + 0.0000e+00, -3.3993e-08], + ..., + [ 1.1642e-09, 6.9849e-10, 1.2573e-08, ..., 1.0477e-08, + 0.0000e+00, 3.0966e-08], + [ 5.1223e-09, 2.3283e-10, 2.5611e-08, ..., 1.7695e-08, + 0.0000e+00, 1.6298e-09], + [ 1.3504e-08, -6.9849e-10, -3.0501e-08, ..., -2.6310e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0268, -0.0243, -0.0438, 0.0039, 0.0065, 0.0419, 0.0199, -0.0175, + -0.0103, -0.0123], device='cuda:0'), grad: tensor([-1.3644e-07, -7.2177e-08, -5.5414e-08, 2.3283e-09, 3.0734e-08, + 7.9395e-08, -3.1898e-08, 9.8487e-08, 7.3574e-08, 2.0489e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 214.36, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4713 re_mapping 0.0039 re_causal 0.0118 /// teacc 98.98 lr 0.00010000 +Epoch 295, weight, value: tensor([[ 0.0248, -0.1053, -0.0899, ..., -0.2293, -0.0940, -0.1938], + [-0.0933, -0.0619, -0.0116, ..., 0.0450, 0.0319, -0.1339], + [ 0.0253, -0.1646, -0.2171, ..., 0.0158, -0.1558, 0.0808], + ..., + [-0.0293, -0.0243, -0.2003, ..., -0.1822, 0.0202, 0.1585], + [ 0.0739, -0.1419, -0.1314, ..., -0.1800, -0.1108, -0.1614], + [-0.2374, 0.0275, 0.1156, ..., 0.0891, -0.0942, -0.1788]], + device='cuda:0'), grad: tensor([[ 2.0023e-08, 0.0000e+00, 3.1432e-08, ..., 6.0536e-09, + 2.4680e-08, 9.3132e-09], + [ 6.9849e-10, 3.4925e-09, -3.2433e-07, ..., -3.2084e-07, + 2.3283e-10, 1.2340e-08], + [ 3.2596e-09, 1.3970e-09, 3.4925e-09, ..., 2.7940e-09, + 4.6566e-10, 5.1223e-09], + ..., + [ 2.3283e-10, -1.1409e-08, 7.6368e-08, ..., 7.8231e-08, + 0.0000e+00, -4.1211e-08], + [-1.0710e-08, 1.1642e-09, 9.1037e-08, ..., 1.5087e-07, + 2.3283e-10, 4.4238e-09], + [ 1.1642e-09, 2.3283e-09, 2.0256e-08, ..., 3.4692e-08, + 0.0000e+00, 1.2573e-08]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0263, -0.0239, -0.0439, 0.0041, 0.0065, 0.0416, 0.0199, -0.0179, + -0.0100, -0.0124], device='cuda:0'), grad: tensor([ 1.3155e-07, -7.0268e-07, 4.1211e-08, 3.1432e-08, 2.4447e-07, + 4.5868e-08, -1.5204e-07, 6.7521e-08, 2.0210e-07, 1.1781e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 214.69, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4809 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.06 lr 0.00010000 +Epoch 296, weight, value: tensor([[ 0.0246, -0.1053, -0.0901, ..., -0.2323, -0.0940, -0.1942], + [-0.0936, -0.0619, -0.0113, ..., 0.0453, 0.0320, -0.1339], + [ 0.0249, -0.1648, -0.2177, ..., 0.0158, -0.1558, 0.0810], + ..., + [-0.0296, -0.0245, -0.2008, ..., -0.1829, 0.0201, 0.1584], + [ 0.0744, -0.1421, -0.1312, ..., -0.1798, -0.1109, -0.1617], + [-0.2385, 0.0277, 0.1157, ..., 0.0891, -0.0941, -0.1790]], + device='cuda:0'), grad: tensor([[-5.7742e-08, 9.3132e-10, -6.6822e-08, ..., -1.7695e-08, + 0.0000e+00, 3.2596e-09], + [ 3.5856e-08, 5.3551e-09, 4.7497e-08, ..., 1.1176e-08, + 0.0000e+00, 2.1653e-08], + [ 2.5611e-09, 9.3132e-10, 4.8894e-09, ..., 1.1642e-09, + 0.0000e+00, 2.3283e-09], + ..., + [ 2.0955e-09, 4.4238e-09, 6.9849e-09, ..., 4.4238e-09, + 0.0000e+00, 1.9558e-08], + [ 8.4750e-08, 5.3551e-09, 1.3039e-08, ..., 2.3074e-07, + 0.0000e+00, 2.4680e-08], + [ 1.1176e-08, 3.0268e-09, 1.3737e-08, ..., 5.5879e-09, + 0.0000e+00, 1.0710e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0264, -0.0236, -0.0440, 0.0038, 0.0065, 0.0417, 0.0197, -0.0181, + -0.0094, -0.0124], device='cuda:0'), grad: tensor([-1.2787e-06, 7.0734e-07, 4.4471e-08, -3.1339e-07, 2.4680e-08, + -3.2294e-07, 1.7253e-07, 9.7090e-08, 6.1281e-07, 2.8219e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 214.27, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4705 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.01 lr 0.00010000 +Epoch 297, weight, value: tensor([[ 0.0246, -0.1054, -0.0903, ..., -0.2324, -0.0941, -0.1946], + [-0.0938, -0.0621, -0.0113, ..., 0.0454, 0.0320, -0.1341], + [ 0.0251, -0.1652, -0.2190, ..., 0.0157, -0.1535, 0.0811], + ..., + [-0.0298, -0.0244, -0.2010, ..., -0.1830, 0.0201, 0.1585], + [ 0.0745, -0.1425, -0.1310, ..., -0.1800, -0.1123, -0.1623], + [-0.2390, 0.0276, 0.1157, ..., 0.0891, -0.0942, -0.1793]], + device='cuda:0'), grad: tensor([[-8.3819e-09, 9.3132e-10, 7.2177e-09, ..., 3.0268e-09, + 0.0000e+00, 6.4028e-08], + [ 2.5611e-09, 3.4925e-09, -9.5321e-07, ..., -2.5169e-07, + 0.0000e+00, 4.1211e-07], + [ 9.3132e-10, 3.5623e-08, 1.4203e-08, ..., 4.6566e-09, + 0.0000e+00, -1.7704e-06], + ..., + [ 1.8626e-09, 2.3283e-09, 5.4250e-07, ..., 1.4831e-07, + 0.0000e+00, 1.6205e-07], + [ 1.5134e-08, 6.5193e-09, 6.1467e-08, ..., 5.4715e-08, + 0.0000e+00, 1.3504e-08], + [ 2.0955e-08, 8.3819e-09, 2.3958e-07, ..., 5.8208e-08, + 0.0000e+00, 7.2177e-09]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0262, -0.0236, -0.0442, 0.0037, 0.0065, 0.0418, 0.0194, -0.0180, + -0.0091, -0.0125], device='cuda:0'), grad: tensor([ 1.3178e-07, -2.0210e-06, -5.2489e-06, 4.4405e-06, 1.6810e-07, + -1.0133e-06, 3.1665e-08, 2.2203e-06, 3.0105e-07, 1.0002e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 213.89, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4815 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.09 lr 0.00010000 +Epoch 298, weight, value: tensor([[ 0.0248, -0.1063, -0.0904, ..., -0.2329, -0.0942, -0.1957], + [-0.0944, -0.0626, -0.0120, ..., 0.0452, 0.0320, -0.1342], + [ 0.0244, -0.1662, -0.2204, ..., 0.0156, -0.1534, 0.0813], + ..., + [-0.0300, -0.0240, -0.2011, ..., -0.1830, 0.0201, 0.1588], + [ 0.0745, -0.1431, -0.1313, ..., -0.1804, -0.1127, -0.1639], + [-0.2394, 0.0278, 0.1158, ..., 0.0892, -0.0942, -0.1798]], + device='cuda:0'), grad: tensor([[-7.4506e-09, 0.0000e+00, 6.5193e-09, ..., 4.6566e-09, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 3.2596e-09, -1.5367e-08, ..., -1.1176e-08, + 0.0000e+00, 4.7963e-08], + [ 4.6566e-10, 1.8626e-09, 1.1176e-08, ..., 5.5879e-09, + 0.0000e+00, -6.2864e-08], + ..., + [ 1.8626e-09, -4.1910e-09, 7.3574e-08, ..., 3.4459e-08, + 0.0000e+00, -3.3528e-08], + [ 5.4017e-08, 9.3132e-10, 1.7695e-08, ..., 1.5553e-07, + 0.0000e+00, 3.7719e-08], + [ 7.9162e-09, 1.3970e-09, 4.1816e-07, ..., 1.7416e-07, + 0.0000e+00, 5.1223e-09]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0261, -0.0237, -0.0441, 0.0034, 0.0065, 0.0420, 0.0194, -0.0178, + -0.0097, -0.0124], device='cuda:0'), grad: tensor([-3.9581e-08, -8.8476e-08, 7.4971e-08, 1.4342e-07, -9.5926e-07, + -4.1239e-06, 3.5688e-06, 4.5635e-08, 5.2433e-07, 8.4983e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 213.79, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4543 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.01 lr 0.00010000 +Epoch 299, weight, value: tensor([[ 0.0232, -0.1066, -0.0923, ..., -0.2335, -0.0943, -0.1966], + [-0.0945, -0.0628, -0.0126, ..., 0.0447, 0.0322, -0.1343], + [ 0.0243, -0.1668, -0.2204, ..., 0.0156, -0.1540, 0.0813], + ..., + [-0.0302, -0.0239, -0.2027, ..., -0.1844, 0.0200, 0.1589], + [ 0.0746, -0.1434, -0.1317, ..., -0.1809, -0.1129, -0.1647], + [-0.2399, 0.0280, 0.1159, ..., 0.0894, -0.0942, -0.1802]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 4.6566e-10, 6.5193e-09, ..., 4.1910e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 4.1910e-08, 1.5832e-08, ..., 8.8476e-09, + 0.0000e+00, 1.2014e-07], + [ 0.0000e+00, 3.2596e-09, 1.3970e-09, ..., 2.3283e-09, + 0.0000e+00, 1.2107e-08], + ..., + [ 0.0000e+00, -6.1002e-08, 2.0955e-08, ..., 1.8161e-08, + 0.0000e+00, -9.7603e-07], + [ 1.8626e-09, 3.2596e-09, 6.9849e-09, ..., 4.9360e-08, + 0.0000e+00, 9.3132e-09], + [ 9.3132e-10, 4.9360e-08, 3.3248e-07, ..., 2.0722e-07, + 0.0000e+00, 2.5099e-07]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0283, -0.0239, -0.0441, 0.0026, 0.0065, 0.0428, 0.0205, -0.0181, + -0.0099, -0.0122], device='cuda:0'), grad: tensor([-4.1910e-09, 3.2410e-07, 2.9802e-08, 1.3923e-06, -1.4901e-08, + -1.3504e-06, 7.7765e-08, -1.8487e-06, 1.7090e-07, 1.2331e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 213.96, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4882 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 diff --git a/Meta-causal/code-withStyleAttack/66528.error b/Meta-causal/code-withStyleAttack/66528.error new file mode 100644 index 0000000000000000000000000000000000000000..94d491ce2985c8bf1a4205b3f4d5587cb60cb55d --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66528.error @@ -0,0 +1,65 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 664, in + experiment() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 269, in experiment + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_FA, y, epsilon_list) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 111, in adversarial_attack_Incre + ori_loss = cls_criterion(x_ori_output, y_ori) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/modules/loss.py", line 1185, in forward + return F.cross_entropy(input, target, weight=self.weight, + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/functional.py", line 3086, in cross_entropy + return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index, label_smoothing) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +ValueError: Expected input batch_size (448) to match target batch_size (32). +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 145, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 29, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 45, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/serialization.py", line 997, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/serialization.py", line 444, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/serialization.py", line 425, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_FA/best_cls_net.pkl' +srun: error: gcp-us-0: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-withStyleAttack/66528.log b/Meta-causal/code-withStyleAttack/66528.log new file mode 100644 index 0000000000000000000000000000000000000000..1d96267a317e292c2fbd28d3dbca4a1ea5c414f4 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66528.log @@ -0,0 +1,25 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_FA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0147, 0.0303, -0.0251, ..., -0.0302, -0.0033, -0.0088], + [ 0.0007, -0.0181, 0.0010, ..., -0.0094, 0.0037, 0.0299], + [ 0.0096, -0.0300, 0.0145, ..., -0.0281, -0.0227, 0.0236], + ..., + [ 0.0306, -0.0302, -0.0116, ..., 0.0051, 0.0029, -0.0268], + [-0.0200, -0.0005, 0.0098, ..., 0.0062, 0.0308, 0.0146], + [ 0.0129, -0.0243, -0.0199, ..., -0.0191, 0.0098, -0.0306]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0234, -0.0300, -0.0249, -0.0259, -0.0306, 0.0035, 0.0253, 0.0311, + -0.0174, 0.0284], device='cuda:0'), grad: None +100 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_FA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_FA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best diff --git a/Meta-causal/code-withStyleAttack/66530.error b/Meta-causal/code-withStyleAttack/66530.error new file mode 100644 index 0000000000000000000000000000000000000000..4f72ff7ab3202956a4b544976dc55294cc3aa260 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66530.error @@ -0,0 +1,65 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 668, in + experiment() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 271, in experiment + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_CA, y, epsilon_list) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 111, in adversarial_attack_Incre + ori_loss = cls_criterion(x_ori_output, y_ori) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/modules/loss.py", line 1185, in forward + return F.cross_entropy(input, target, weight=self.weight, + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/nn/functional.py", line 3086, in cross_entropy + return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index, label_smoothing) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +ValueError: Expected input batch_size (4928) to match target batch_size (32). +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 145, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 29, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 45, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/serialization.py", line 997, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/serialization.py", line 444, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3/lib/python3.11/site-packages/torch/serialization.py", line 425, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_CA/best_cls_net.pkl' +srun: error: gcp-us-0: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-withStyleAttack/66530.log b/Meta-causal/code-withStyleAttack/66530.log new file mode 100644 index 0000000000000000000000000000000000000000..03e3e5ac7e65ae1978a5d2a0bbd0fc649d8547ec --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66530.log @@ -0,0 +1,25 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_CA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0277, 0.0167, -0.0013, ..., 0.0189, -0.0297, -0.0272], + [-0.0239, 0.0024, -0.0033, ..., -0.0273, 0.0171, -0.0036], + [-0.0175, -0.0069, -0.0023, ..., 0.0224, 0.0170, 0.0235], + ..., + [ 0.0276, 0.0035, 0.0014, ..., -0.0038, -0.0009, -0.0128], + [ 0.0178, -0.0050, -0.0292, ..., -0.0179, 0.0209, 0.0095], + [ 0.0199, 0.0231, -0.0225, ..., 0.0291, -0.0220, 0.0027]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0288, -0.0253, 0.0215, -0.0196, 0.0060, -0.0289, -0.0188, -0.0047, + -0.0187, -0.0223], device='cuda:0'), grad: None +100 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_CA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_CA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best diff --git a/Meta-causal/code-withStyleAttack/66534.error b/Meta-causal/code-withStyleAttack/66534.error new file mode 100644 index 0000000000000000000000000000000000000000..47c523a6371154433042b34829a97e2f15d79dc6 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66534.error @@ -0,0 +1,4 @@ +run_my_joint_test.sh: line 33: randn}_str3_WithStyleAttackExp1_onlyblock1: command not found +slurmstepd: error: *** STEP 66534.0 ON gcp-us-0 CANCELLED AT 2024-07-21T15:38:51 DUE TO TIME LIMIT *** +slurmstepd: error: *** JOB 66534 ON gcp-us-0 CANCELLED AT 2024-07-21T15:38:51 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-withStyleAttack/66534.log b/Meta-causal/code-withStyleAttack/66534.log new file mode 100644 index 0000000000000000000000000000000000000000..3016a063f556f2647ab6f5b6de089bd1c9c8ab2d --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66534.log @@ -0,0 +1,22413 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_onlyblock1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0173, 0.0306, -0.0042, ..., -0.0188, -0.0066, 0.0181], + [ 0.0281, 0.0010, 0.0214, ..., 0.0243, 0.0040, 0.0179], + [-0.0077, -0.0215, 0.0062, ..., -0.0144, -0.0264, -0.0055], + ..., + [ 0.0224, 0.0166, -0.0166, ..., 0.0290, 0.0155, 0.0165], + [-0.0193, 0.0286, 0.0257, ..., -0.0082, 0.0264, -0.0109], + [-0.0253, 0.0006, -0.0054, ..., -0.0236, 0.0260, -0.0084]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0255, -0.0243, 0.0124, 0.0280, -0.0300, 0.0068, -0.0089, 0.0085, + 0.0148, 0.0024], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 231.27, cls_loss 1.4751 cls_loss_mapping 1.9400 cls_loss_causal 2.2389 re_mapping 0.1137 re_causal 0.1194 /// teacc 84.13 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0133, 0.0299, -0.0042, ..., -0.0238, -0.0077, 0.0233], + [ 0.0266, 0.0002, 0.0214, ..., 0.0301, -0.0032, 0.0105], + [-0.0028, -0.0208, 0.0062, ..., -0.0190, -0.0348, -0.0113], + ..., + [ 0.0157, 0.0158, -0.0166, ..., 0.0328, 0.0181, 0.0169], + [-0.0201, 0.0293, 0.0257, ..., -0.0133, 0.0269, -0.0149], + [-0.0255, -0.0002, -0.0054, ..., -0.0239, 0.0291, -0.0084]], + device='cuda:0'), grad: tensor([[ 2.4967e-03, 0.0000e+00, 0.0000e+00, ..., 4.3988e-04, + 3.1681e-03, -1.4107e-02], + [ 1.6384e-03, 0.0000e+00, 0.0000e+00, ..., 7.8440e-05, + 2.8885e-02, 3.9787e-03], + [-5.4455e-04, 0.0000e+00, 0.0000e+00, ..., 6.7253e-03, + 5.4855e-03, 1.0399e-02], + ..., + [ 1.3294e-03, 0.0000e+00, 0.0000e+00, ..., -8.6288e-03, + -8.6288e-03, -3.0060e-03], + [ 3.6144e-03, 0.0000e+00, 0.0000e+00, ..., -1.3275e-03, + 5.7125e-04, -1.9360e-03], + [ 2.4452e-03, 0.0000e+00, 0.0000e+00, ..., -8.1940e-03, + -7.5867e-02, 4.2076e-03]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0287, -0.0235, 0.0136, 0.0277, -0.0296, 0.0074, -0.0087, 0.0083, + 0.0148, 0.0025], device='cuda:0'), grad: tensor([-4.5128e-03, 4.0192e-02, -9.7275e-05, 3.7476e-02, 5.2094e-02, + -1.0628e-02, -3.1311e-02, -3.4161e-03, -6.4468e-03, -7.3303e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 231.25, cls_loss 0.4496 cls_loss_mapping 0.7952 cls_loss_causal 1.9208 re_mapping 0.2048 re_causal 0.2571 /// teacc 91.96 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0113, 0.0281, -0.0042, ..., -0.0263, -0.0091, 0.0256], + [ 0.0259, -0.0008, 0.0214, ..., 0.0364, -0.0061, 0.0075], + [-0.0015, -0.0225, 0.0062, ..., -0.0249, -0.0381, -0.0132], + ..., + [ 0.0135, 0.0148, -0.0166, ..., 0.0323, 0.0188, 0.0165], + [-0.0231, 0.0272, 0.0257, ..., -0.0165, 0.0278, -0.0184], + [-0.0304, -0.0013, -0.0054, ..., -0.0254, 0.0302, -0.0065]], + device='cuda:0'), grad: tensor([[ 7.9060e-04, 6.3912e-08, 0.0000e+00, ..., 5.9128e-04, + 5.7411e-03, 2.6684e-03], + [-7.2403e-03, 1.0827e-07, 0.0000e+00, ..., -3.5248e-02, + -3.4165e-04, 3.4618e-03], + [ 1.7529e-03, 4.5379e-07, 0.0000e+00, ..., 4.0321e-03, + -2.6321e-03, -2.0218e-03], + ..., + [ 1.1425e-03, 8.0094e-08, 0.0000e+00, ..., 1.6190e-02, + 4.8859e-02, 3.9795e-02], + [ 6.6376e-03, 6.3004e-07, 0.0000e+00, ..., 1.1307e-02, + 2.8534e-02, 2.4780e-02], + [ 3.6907e-03, 2.6263e-07, 0.0000e+00, ..., -1.3596e-02, + -4.4983e-02, -1.8082e-02]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0295, -0.0240, 0.0135, 0.0274, -0.0294, 0.0087, -0.0091, 0.0080, + 0.0149, 0.0028], device='cuda:0'), grad: tensor([ 0.0059, -0.0243, -0.0141, -0.0109, 0.0059, -0.0525, 0.0269, 0.0481, + 0.0435, -0.0286], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 230.12, cls_loss 0.2639 cls_loss_mapping 0.4504 cls_loss_causal 1.6539 re_mapping 0.1510 re_causal 0.2428 /// teacc 94.47 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0107, 0.0269, -0.0137, ..., -0.0287, -0.0102, 0.0271], + [ 0.0260, -0.0124, 0.0275, ..., 0.0406, -0.0077, 0.0060], + [-0.0010, -0.0233, 0.0040, ..., -0.0279, -0.0403, -0.0142], + ..., + [ 0.0119, 0.0040, -0.0148, ..., 0.0314, 0.0189, 0.0160], + [-0.0251, 0.0193, 0.0213, ..., -0.0192, 0.0280, -0.0212], + [-0.0337, -0.0143, -0.0120, ..., -0.0265, 0.0303, -0.0052]], + device='cuda:0'), grad: tensor([[ 9.4986e-04, 1.4651e-04, 0.0000e+00, ..., 5.0879e-04, + 3.8013e-03, 4.7684e-03], + [-2.6093e-03, 4.9412e-05, 0.0000e+00, ..., -1.5808e-02, + 1.6994e-03, 1.7805e-03], + [ 3.1796e-03, 7.7724e-05, 0.0000e+00, ..., 4.6939e-05, + 2.7771e-03, 2.1725e-03], + ..., + [ 2.8687e-03, 1.3447e-04, 0.0000e+00, ..., 3.2482e-03, + -1.3056e-03, -7.1945e-03], + [-4.0627e-03, 9.9421e-05, 0.0000e+00, ..., 3.5896e-03, + -3.4698e-02, -1.2833e-02], + [ 1.5427e-02, 1.5140e-04, 0.0000e+00, ..., 3.0346e-03, + 4.0314e-02, 2.0691e-02]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0295, -0.0240, 0.0138, 0.0274, -0.0291, 0.0089, -0.0095, 0.0082, + 0.0149, 0.0025], device='cuda:0'), grad: tensor([ 0.0054, -0.0058, -0.0110, -0.0102, -0.0265, 0.0318, -0.0035, 0.0080, + -0.0385, 0.0503], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 230.50, cls_loss 0.1886 cls_loss_mapping 0.3218 cls_loss_causal 1.4555 re_mapping 0.1124 re_causal 0.2161 /// teacc 95.11 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 0.0096, 0.0280, -0.0157, ..., -0.0304, -0.0110, 0.0282], + [ 0.0265, -0.0194, 0.0267, ..., 0.0443, -0.0090, 0.0043], + [-0.0012, -0.0267, 0.0054, ..., -0.0287, -0.0421, -0.0156], + ..., + [ 0.0107, -0.0005, -0.0145, ..., 0.0298, 0.0191, 0.0148], + [-0.0256, 0.0141, 0.0215, ..., -0.0209, 0.0276, -0.0230], + [-0.0364, -0.0251, -0.0142, ..., -0.0267, 0.0309, -0.0032]], + device='cuda:0'), grad: tensor([[ 0.0006, 0.0000, 0.0003, ..., 0.0025, 0.0061, 0.0123], + [ 0.0015, 0.0000, 0.0003, ..., 0.0003, 0.0019, 0.0022], + [ 0.0061, 0.0000, 0.0039, ..., 0.0010, 0.0061, -0.0020], + ..., + [ 0.0011, 0.0000, 0.0027, ..., -0.0009, 0.0025, 0.0093], + [-0.0044, 0.0000, -0.0107, ..., 0.0017, -0.0045, -0.0048], + [ 0.0067, 0.0000, 0.0032, ..., 0.0021, -0.0006, -0.0030]], + device='cuda:0') +Epoch 5, bias, value: tensor([-0.0295, -0.0238, 0.0141, 0.0273, -0.0293, 0.0087, -0.0101, 0.0078, + 0.0153, 0.0028], device='cuda:0'), grad: tensor([ 0.0132, 0.0048, 0.0251, 0.0001, -0.0099, -0.0006, -0.0026, 0.0138, + -0.0546, 0.0106], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 230.46, cls_loss 0.1512 cls_loss_mapping 0.2504 cls_loss_causal 1.3337 re_mapping 0.0898 re_causal 0.1937 /// teacc 96.41 lr 0.00010000 +Epoch 6, weight, value: tensor([[ 0.0094, 0.0278, -0.0200, ..., -0.0307, -0.0122, 0.0290], + [ 0.0257, -0.0178, 0.0243, ..., 0.0463, -0.0100, 0.0022], + [-0.0013, -0.0270, 0.0064, ..., -0.0297, -0.0440, -0.0179], + ..., + [ 0.0095, -0.0040, -0.0122, ..., 0.0290, 0.0196, 0.0143], + [-0.0264, 0.0106, 0.0217, ..., -0.0220, 0.0274, -0.0251], + [-0.0382, -0.0256, -0.0164, ..., -0.0290, 0.0314, -0.0012]], + device='cuda:0'), grad: tensor([[ 8.4996e-05, 2.0981e-05, 1.0990e-05, ..., 3.5644e-04, + 7.4673e-04, -1.8234e-03], + [-6.3705e-04, 5.0440e-06, 7.8082e-05, ..., -7.6180e-03, + 1.4734e-03, 3.2234e-04], + [-1.8921e-03, 3.7998e-05, -4.5252e-04, ..., -1.3870e-02, + -1.2337e-02, -4.7374e-04], + ..., + [ 4.7040e-04, 2.0981e-05, -5.6982e-05, ..., 7.6180e-03, + 3.4447e-03, 4.6182e-04], + [ 3.9554e-04, 4.4763e-05, 2.2495e-04, ..., 6.6910e-03, + 1.4734e-03, 1.8108e-04], + [ 4.5085e-04, 2.5734e-05, 5.2929e-05, ..., 9.3651e-04, + 8.4019e-04, 5.4455e-04]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0290, -0.0241, 0.0139, 0.0276, -0.0295, 0.0086, -0.0102, 0.0082, + 0.0153, 0.0028], device='cuda:0'), grad: tensor([ 0.0004, -0.0020, -0.0290, 0.0102, 0.0014, -0.0007, 0.0033, 0.0100, + 0.0041, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 230.74, cls_loss 0.1284 cls_loss_mapping 0.2038 cls_loss_causal 1.2811 re_mapping 0.0751 re_causal 0.1766 /// teacc 97.03 lr 0.00010000 +Epoch 7, weight, value: tensor([[ 0.0090, 0.0275, -0.0261, ..., -0.0321, -0.0134, 0.0295], + [ 0.0253, -0.0169, 0.0234, ..., 0.0491, -0.0115, 0.0009], + [-0.0012, -0.0286, 0.0080, ..., -0.0310, -0.0455, -0.0200], + ..., + [ 0.0085, -0.0077, -0.0103, ..., 0.0280, 0.0203, 0.0137], + [-0.0263, 0.0046, 0.0200, ..., -0.0240, 0.0267, -0.0269], + [-0.0401, -0.0252, -0.0186, ..., -0.0302, 0.0317, 0.0007]], + device='cuda:0'), grad: tensor([[ 1.0996e-03, 2.5809e-05, -2.8014e-04, ..., 9.1839e-04, + 6.6090e-04, -2.5463e-03], + [ 3.5706e-03, 4.2588e-05, 9.0122e-05, ..., 8.6975e-04, + 7.0906e-04, 3.3836e-03], + [ 3.6545e-03, 7.8022e-05, -9.8896e-04, ..., 4.3144e-03, + 3.1328e-04, 3.0785e-03], + ..., + [ 1.7281e-03, 5.5641e-05, -1.6224e-04, ..., 1.9817e-03, + 4.4847e-04, 7.7534e-04], + [ 3.1147e-03, 1.0502e-04, 2.8872e-04, ..., 5.3978e-03, + 1.3161e-03, 3.4294e-03], + [ 1.1765e-02, 2.5773e-04, 4.0269e-04, ..., 1.0605e-02, + 2.5043e-03, 5.0831e-04]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0291, -0.0238, 0.0134, 0.0276, -0.0295, 0.0083, -0.0105, 0.0081, + 0.0156, 0.0031], device='cuda:0'), grad: tensor([-0.0019, 0.0109, 0.0075, -0.0144, -0.0455, 0.0057, 0.0050, 0.0049, + 0.0115, 0.0161], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 231.06, cls_loss 0.1087 cls_loss_mapping 0.1717 cls_loss_causal 1.1810 re_mapping 0.0637 re_causal 0.1592 /// teacc 97.37 lr 0.00010000 +Epoch 8, weight, value: tensor([[ 0.0082, 0.0254, -0.0309, ..., -0.0326, -0.0139, 0.0307], + [ 0.0247, -0.0162, 0.0240, ..., 0.0509, -0.0125, -0.0010], + [-0.0014, -0.0290, 0.0139, ..., -0.0320, -0.0471, -0.0211], + ..., + [ 0.0086, -0.0152, -0.0109, ..., 0.0273, 0.0204, 0.0127], + [-0.0272, -0.0032, 0.0197, ..., -0.0259, 0.0265, -0.0291], + [-0.0418, -0.0235, -0.0204, ..., -0.0312, 0.0319, 0.0027]], + device='cuda:0'), grad: tensor([[ 2.0659e-04, 1.3776e-05, 1.1206e-04, ..., 2.7084e-04, + 8.8501e-04, -1.4124e-03], + [ 3.8385e-04, 1.0200e-05, 7.7934e-03, ..., -4.5319e-03, + 1.2426e-03, 1.2283e-03], + [ 4.0388e-04, 9.1121e-06, -9.3079e-03, ..., -5.7030e-03, + 5.2071e-03, 1.1215e-03], + ..., + [ 4.0817e-04, 8.3223e-06, 2.6751e-04, ..., -2.7351e-03, + -1.0582e-02, -6.1750e-04], + [-2.2564e-03, 8.2627e-06, -1.6813e-03, ..., 6.9199e-03, + -1.2314e-02, -7.4959e-03], + [ 1.0405e-03, 4.9233e-05, 9.2220e-04, ..., 9.6464e-04, + 3.8605e-03, 2.2106e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0288, -0.0239, 0.0134, 0.0277, -0.0295, 0.0082, -0.0108, 0.0080, + 0.0157, 0.0034], device='cuda:0'), grad: tensor([ 0.0003, 0.0056, -0.0089, 0.0345, 0.0042, -0.0150, 0.0040, -0.0097, + -0.0233, 0.0083], device='cuda:0') +100 +0.0001 +changing lr +epoch 7, time 214.68, cls_loss 0.0823 cls_loss_mapping 0.1428 cls_loss_causal 1.1545 re_mapping 0.0577 re_causal 0.1501 /// teacc 97.34 lr 0.00010000 +Epoch 9, weight, value: tensor([[ 0.0073, 0.0247, -0.0334, ..., -0.0337, -0.0150, 0.0312], + [ 0.0238, -0.0150, 0.0206, ..., 0.0527, -0.0134, -0.0025], + [-0.0017, -0.0287, 0.0195, ..., -0.0332, -0.0484, -0.0220], + ..., + [ 0.0081, -0.0187, -0.0115, ..., 0.0265, 0.0213, 0.0123], + [-0.0268, -0.0070, 0.0190, ..., -0.0272, 0.0266, -0.0306], + [-0.0435, -0.0242, -0.0235, ..., -0.0326, 0.0318, 0.0035]], + device='cuda:0'), grad: tensor([[ 4.4137e-05, 2.3633e-05, 6.8247e-05, ..., 1.5807e-04, + 6.1035e-04, 2.3067e-04], + [ 8.1286e-06, 4.2140e-05, 4.3726e-04, ..., 6.9618e-04, + 2.5063e-03, 4.3488e-04], + [ 3.5435e-05, 4.6909e-05, -1.0433e-03, ..., 1.8644e-04, + 8.8501e-04, 3.9744e-04], + ..., + [ 8.0884e-05, 5.2959e-05, 3.1018e-04, ..., -8.0156e-04, + -2.8038e-03, 4.3030e-03], + [ 5.5730e-05, 1.3793e-04, -3.6550e-04, ..., -1.2932e-03, + -3.2592e-04, 1.4553e-03], + [ 2.0528e-04, 4.6158e-04, 1.4091e-04, ..., 3.6788e-04, + -5.8327e-03, -9.0714e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0289, -0.0239, 0.0134, 0.0278, -0.0294, 0.0080, -0.0110, 0.0082, + 0.0161, 0.0030], device='cuda:0'), grad: tensor([ 0.0009, 0.0063, 0.0006, 0.0009, 0.0051, 0.0042, -0.0014, -0.0037, + -0.0067, -0.0062], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 231.07, cls_loss 0.0974 cls_loss_mapping 0.1531 cls_loss_causal 1.1205 re_mapping 0.0502 re_causal 0.1329 /// teacc 97.53 lr 0.00010000 +Epoch 10, weight, value: tensor([[ 0.0072, 0.0233, -0.0351, ..., -0.0344, -0.0160, 0.0317], + [ 0.0225, -0.0134, 0.0201, ..., 0.0539, -0.0142, -0.0035], + [-0.0015, -0.0283, 0.0218, ..., -0.0336, -0.0495, -0.0220], + ..., + [ 0.0079, -0.0234, -0.0112, ..., 0.0261, 0.0215, 0.0112], + [-0.0272, -0.0138, 0.0204, ..., -0.0285, 0.0262, -0.0319], + [-0.0449, -0.0235, -0.0250, ..., -0.0335, 0.0321, 0.0049]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0003, 0.0008, ..., 0.0004, 0.0013, 0.0034], + [ 0.0043, 0.0003, 0.0006, ..., 0.0092, 0.0029, 0.0006], + [-0.0104, 0.0009, -0.0106, ..., 0.0009, 0.0014, 0.0018], + ..., + [ 0.0029, 0.0015, 0.0014, ..., 0.0009, -0.0028, 0.0022], + [-0.0035, 0.0021, -0.0004, ..., -0.0122, -0.0012, -0.0011], + [ 0.0013, 0.0014, 0.0006, ..., 0.0016, 0.0046, 0.0031]], + device='cuda:0') +Epoch 10, bias, value: tensor([-0.0288, -0.0241, 0.0137, 0.0277, -0.0297, 0.0080, -0.0111, 0.0081, + 0.0163, 0.0032], device='cuda:0'), grad: tensor([ 0.0056, 0.0144, -0.0101, -0.0159, 0.0124, -0.0007, 0.0043, 0.0037, + -0.0229, 0.0091], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 230.97, cls_loss 0.0790 cls_loss_mapping 0.1295 cls_loss_causal 1.1225 re_mapping 0.0458 re_causal 0.1269 /// teacc 97.95 lr 0.00010000 +Epoch 11, weight, value: tensor([[ 0.0065, 0.0222, -0.0369, ..., -0.0343, -0.0169, 0.0323], + [ 0.0220, -0.0125, 0.0181, ..., 0.0553, -0.0153, -0.0044], + [-0.0014, -0.0284, 0.0257, ..., -0.0338, -0.0507, -0.0228], + ..., + [ 0.0071, -0.0255, -0.0116, ..., 0.0246, 0.0224, 0.0109], + [-0.0276, -0.0192, 0.0203, ..., -0.0299, 0.0260, -0.0330], + [-0.0466, -0.0224, -0.0260, ..., -0.0344, 0.0321, 0.0061]], + device='cuda:0'), grad: tensor([[-3.1185e-04, 4.9257e-04, -2.9831e-03, ..., -2.0015e-04, + 1.8823e-04, -1.4885e-02], + [ 5.7340e-05, 1.9228e-04, 5.2500e-04, ..., -5.2071e-04, + 4.3750e-04, 5.1022e-04], + [ 4.1783e-05, 8.4877e-04, 6.0081e-03, ..., 9.0504e-04, + 1.7653e-03, 4.0207e-03], + ..., + [ 2.0897e-04, 2.7394e-04, -3.2539e-03, ..., -7.1955e-04, + -3.2024e-03, 2.6226e-03], + [ 1.0622e-04, 2.7046e-03, -4.4632e-03, ..., 6.5744e-05, + 4.3774e-04, 3.3512e-03], + [ 5.2547e-04, 2.4529e-03, 9.9850e-04, ..., 3.6693e-04, + 2.9993e-04, 4.5090e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0289, -0.0242, 0.0135, 0.0281, -0.0295, 0.0080, -0.0113, 0.0081, + 0.0164, 0.0031], device='cuda:0'), grad: tensor([-0.0171, 0.0017, 0.0174, -0.0117, 0.0024, 0.0072, 0.0050, -0.0046, + -0.0073, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 231.03, cls_loss 0.0718 cls_loss_mapping 0.1160 cls_loss_causal 1.1162 re_mapping 0.0435 re_causal 0.1199 /// teacc 97.97 lr 0.00010000 +Epoch 12, weight, value: tensor([[ 0.0058, 0.0214, -0.0391, ..., -0.0350, -0.0177, 0.0331], + [ 0.0206, -0.0118, 0.0171, ..., 0.0563, -0.0163, -0.0057], + [-0.0013, -0.0283, 0.0293, ..., -0.0341, -0.0515, -0.0238], + ..., + [ 0.0064, -0.0279, -0.0120, ..., 0.0241, 0.0229, 0.0107], + [-0.0279, -0.0214, 0.0190, ..., -0.0311, 0.0258, -0.0338], + [-0.0479, -0.0231, -0.0282, ..., -0.0351, 0.0320, 0.0072]], + device='cuda:0'), grad: tensor([[ 2.3913e-04, 1.4622e-06, 1.0180e-04, ..., 2.3198e-04, + 1.6432e-03, 4.9400e-03], + [ 4.5419e-04, 5.9277e-05, 1.0014e-03, ..., -2.4378e-04, + 6.2799e-04, 5.5742e-04], + [-7.3814e-04, -7.1943e-05, -2.2659e-03, ..., -6.0177e-04, + 1.8883e-03, 1.8740e-03], + ..., + [ 2.6894e-04, 1.0140e-05, 6.6757e-04, ..., 7.9060e-04, + 8.1482e-03, 2.3041e-02], + [ 1.7941e-04, 2.1402e-06, -7.3791e-05, ..., 1.1390e-04, + -2.1229e-03, 3.7932e-04], + [ 6.6662e-04, 1.8030e-06, 6.2585e-05, ..., 2.7227e-04, + -9.3918e-03, -2.7954e-02]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0290, -0.0247, 0.0137, 0.0280, -0.0293, 0.0078, -0.0115, 0.0083, + 0.0169, 0.0029], device='cuda:0'), grad: tensor([ 0.0055, 0.0035, 0.0018, -0.0109, 0.0015, 0.0068, 0.0005, 0.0245, + -0.0081, -0.0253], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 228.28, cls_loss 0.0624 cls_loss_mapping 0.1048 cls_loss_causal 1.0583 re_mapping 0.0395 re_causal 0.1123 /// teacc 98.06 lr 0.00010000 +Epoch 13, weight, value: tensor([[ 0.0055, 0.0208, -0.0399, ..., -0.0355, -0.0183, 0.0335], + [ 0.0199, -0.0115, 0.0155, ..., 0.0573, -0.0169, -0.0060], + [-0.0017, -0.0271, 0.0319, ..., -0.0353, -0.0525, -0.0242], + ..., + [ 0.0060, -0.0299, -0.0118, ..., 0.0239, 0.0236, 0.0101], + [-0.0282, -0.0227, 0.0185, ..., -0.0321, 0.0255, -0.0352], + [-0.0488, -0.0242, -0.0291, ..., -0.0363, 0.0319, 0.0085]], + device='cuda:0'), grad: tensor([[ 1.4067e-04, 5.7276e-07, -3.7163e-05, ..., -4.7040e-04, + 3.6240e-04, -2.5997e-03], + [ 1.8835e-04, 8.9258e-06, 2.3699e-04, ..., 6.3086e-04, + 1.4210e-03, 6.8903e-04], + [ 7.1943e-05, 2.3134e-06, -1.1176e-04, ..., 4.1533e-04, + 4.5896e-04, 9.1124e-04], + ..., + [ 4.7970e-04, 2.0981e-05, -3.2640e-04, ..., -1.2999e-03, + -2.5940e-03, -1.6699e-03], + [ 1.2094e-04, 2.8554e-06, 5.8591e-05, ..., 3.5858e-04, + 2.3210e-04, 6.2180e-04], + [-5.5389e-03, 1.3225e-05, 2.5317e-05, ..., -3.8834e-03, + -1.3266e-03, -5.9586e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0292, -0.0249, 0.0137, 0.0282, -0.0295, 0.0077, -0.0114, 0.0086, + 0.0168, 0.0031], device='cuda:0'), grad: tensor([-0.0042, 0.0036, 0.0021, -0.0016, 0.0144, 0.0008, 0.0009, -0.0084, + 0.0014, -0.0090], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 214.18, cls_loss 0.0559 cls_loss_mapping 0.0966 cls_loss_causal 1.0011 re_mapping 0.0375 re_causal 0.1037 /// teacc 97.91 lr 0.00010000 +Epoch 14, weight, value: tensor([[ 0.0050, 0.0205, -0.0415, ..., -0.0358, -0.0191, 0.0343], + [ 0.0189, -0.0113, 0.0146, ..., 0.0584, -0.0177, -0.0067], + [-0.0020, -0.0271, 0.0353, ..., -0.0363, -0.0537, -0.0249], + ..., + [ 0.0051, -0.0297, -0.0129, ..., 0.0229, 0.0247, 0.0096], + [-0.0284, -0.0235, 0.0166, ..., -0.0326, 0.0257, -0.0355], + [-0.0498, -0.0245, -0.0286, ..., -0.0369, 0.0316, 0.0086]], + device='cuda:0'), grad: tensor([[ 1.9512e-03, 3.2131e-06, 1.6785e-04, ..., 1.1234e-03, + 7.3576e-04, 2.2926e-03], + [ 1.6391e-04, 1.8077e-06, 1.5104e-04, ..., -2.1496e-03, + 2.0099e-04, 2.3603e-04], + [ 4.7326e-04, 9.2983e-06, -7.1716e-04, ..., 4.7779e-04, + 2.4378e-04, 4.7779e-04], + ..., + [ 2.1756e-04, 3.9898e-06, 2.2173e-04, ..., 4.0388e-04, + 3.7289e-04, 1.1110e-03], + [ 6.1560e-04, 8.0541e-06, -9.9421e-05, ..., 1.2112e-03, + 4.2033e-04, 6.2370e-04], + [ 1.7519e-03, 1.0781e-05, -5.7650e-04, ..., 9.7466e-04, + -1.6632e-03, -3.2368e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0288, -0.0250, 0.0137, 0.0283, -0.0294, 0.0077, -0.0119, 0.0087, + 0.0168, 0.0030], device='cuda:0'), grad: tensor([ 0.0060, -0.0013, 0.0011, -0.0048, -0.0099, 0.0042, 0.0033, 0.0021, + 0.0013, -0.0020], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 214.32, cls_loss 0.0546 cls_loss_mapping 0.0968 cls_loss_causal 0.9870 re_mapping 0.0357 re_causal 0.0996 /// teacc 97.84 lr 0.00010000 +Epoch 15, weight, value: tensor([[ 0.0051, 0.0197, -0.0417, ..., -0.0350, -0.0197, 0.0347], + [ 0.0186, -0.0121, 0.0125, ..., 0.0593, -0.0179, -0.0077], + [-0.0024, -0.0270, 0.0389, ..., -0.0364, -0.0544, -0.0256], + ..., + [ 0.0047, -0.0320, -0.0136, ..., 0.0224, 0.0252, 0.0091], + [-0.0288, -0.0260, 0.0153, ..., -0.0339, 0.0256, -0.0359], + [-0.0506, -0.0243, -0.0294, ..., -0.0379, 0.0313, 0.0095]], + device='cuda:0'), grad: tensor([[ 1.3602e-04, 4.5031e-05, 3.8952e-05, ..., 1.0127e-04, + 8.5354e-05, -2.6751e-04], + [ 2.6107e-04, 8.2850e-05, 8.0228e-05, ..., 1.0633e-04, + 1.1951e-04, 8.7261e-05], + [ 5.9938e-04, -3.7122e-04, -2.0084e-03, ..., 5.2786e-04, + 2.1780e-04, 1.5819e-04], + ..., + [ 1.9145e-04, 1.4442e-02, 9.4473e-05, ..., 1.9264e-04, + 2.5421e-02, 9.4604e-03], + [ 8.1360e-05, 7.8261e-05, -3.4690e-05, ..., 1.6093e-04, + 7.4089e-05, 1.3435e-04], + [ 1.0151e-04, 1.4734e-04, 2.9936e-05, ..., 1.3888e-04, + 2.0838e-04, 1.9133e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0287, -0.0252, 0.0139, 0.0286, -0.0296, 0.0074, -0.0119, 0.0087, + 0.0170, 0.0028], device='cuda:0'), grad: tensor([-5.5969e-05, 6.4278e-04, -5.5075e-04, -2.2446e-02, -6.3801e-04, + 5.7554e-04, -7.2002e-04, 2.2964e-02, -4.8351e-04, 7.1621e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 214.33, cls_loss 0.0468 cls_loss_mapping 0.0861 cls_loss_causal 0.9745 re_mapping 0.0334 re_causal 0.0961 /// teacc 98.05 lr 0.00010000 +Epoch 16, weight, value: tensor([[ 0.0045, 0.0192, -0.0432, ..., -0.0363, -0.0205, 0.0350], + [ 0.0178, -0.0112, 0.0113, ..., 0.0600, -0.0188, -0.0084], + [-0.0025, -0.0273, 0.0420, ..., -0.0357, -0.0551, -0.0261], + ..., + [ 0.0041, -0.0331, -0.0145, ..., 0.0216, 0.0258, 0.0086], + [-0.0291, -0.0269, 0.0147, ..., -0.0347, 0.0258, -0.0367], + [-0.0512, -0.0248, -0.0307, ..., -0.0384, 0.0313, 0.0103]], + device='cuda:0'), grad: tensor([[ 2.6798e-04, 0.0000e+00, 1.2064e-04, ..., 7.6413e-05, + 2.3723e-04, -3.7212e-03], + [ 6.1154e-05, 0.0000e+00, 6.7770e-05, ..., -9.0301e-06, + 2.0134e-04, 3.4022e-04], + [ 1.4901e-04, 0.0000e+00, -7.6008e-04, ..., 5.8085e-05, + 3.2973e-04, 1.3447e-03], + ..., + [ 4.6372e-05, 0.0000e+00, 7.8619e-05, ..., 2.8402e-05, + -1.1320e-03, -1.8311e-04], + [-2.6488e-04, 0.0000e+00, -5.3704e-05, ..., 4.4137e-05, + -2.1577e-04, -1.8537e-04], + [ 1.3518e-04, 0.0000e+00, 8.1301e-05, ..., 5.2691e-05, + 7.5817e-04, 6.7806e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0292, -0.0254, 0.0145, 0.0285, -0.0296, 0.0073, -0.0120, 0.0086, + 0.0174, 0.0028], device='cuda:0'), grad: tensor([-0.0048, 0.0009, 0.0026, 0.0020, 0.0007, -0.0001, 0.0021, -0.0007, + -0.0048, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 230.54, cls_loss 0.0490 cls_loss_mapping 0.0858 cls_loss_causal 0.9942 re_mapping 0.0306 re_causal 0.0928 /// teacc 98.42 lr 0.00010000 +Epoch 17, weight, value: tensor([[ 0.0040, 0.0188, -0.0446, ..., -0.0367, -0.0211, 0.0354], + [ 0.0168, -0.0109, 0.0113, ..., 0.0608, -0.0196, -0.0091], + [-0.0030, -0.0269, 0.0448, ..., -0.0366, -0.0555, -0.0265], + ..., + [ 0.0035, -0.0334, -0.0159, ..., 0.0214, 0.0262, 0.0075], + [-0.0293, -0.0278, 0.0150, ..., -0.0356, 0.0254, -0.0380], + [-0.0525, -0.0254, -0.0322, ..., -0.0399, 0.0318, 0.0116]], + device='cuda:0'), grad: tensor([[ 1.6057e-04, 3.1497e-06, 2.2542e-04, ..., 1.0049e-04, + 2.0242e-04, 2.5773e-04], + [ 3.0541e-04, 1.8418e-05, 2.9802e-05, ..., -1.9045e-03, + 1.7858e-04, 9.9480e-05], + [ 2.4402e-04, -3.8922e-05, -1.7452e-03, ..., 9.5415e-04, + 1.6773e-04, 6.7413e-05], + ..., + [ 6.3753e-04, 4.4033e-06, 2.7013e-04, ..., 4.6277e-04, + 2.1398e-04, 6.1131e-04], + [ 7.8297e-04, 3.6061e-05, 2.0945e-04, ..., 4.7708e-04, + 1.0347e-03, 6.9141e-04], + [-9.7885e-03, 6.1467e-06, 1.5152e-04, ..., -1.2054e-03, + -4.0627e-03, -1.2016e-02]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0292, -0.0256, 0.0147, 0.0286, -0.0293, 0.0073, -0.0120, 0.0083, + 0.0173, 0.0027], device='cuda:0'), grad: tensor([ 0.0008, -0.0012, -0.0010, 0.0034, 0.0087, -0.0010, 0.0024, 0.0017, + 0.0019, -0.0156], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 214.80, cls_loss 0.0466 cls_loss_mapping 0.0831 cls_loss_causal 0.9424 re_mapping 0.0305 re_causal 0.0863 /// teacc 98.30 lr 0.00010000 +Epoch 18, weight, value: tensor([[ 0.0037, 0.0189, -0.0435, ..., -0.0370, -0.0216, 0.0357], + [ 0.0165, -0.0106, 0.0103, ..., 0.0616, -0.0203, -0.0096], + [-0.0033, -0.0265, 0.0474, ..., -0.0371, -0.0563, -0.0271], + ..., + [ 0.0035, -0.0338, -0.0156, ..., 0.0210, 0.0266, 0.0069], + [-0.0298, -0.0292, 0.0147, ..., -0.0365, 0.0257, -0.0387], + [-0.0535, -0.0258, -0.0333, ..., -0.0407, 0.0315, 0.0127]], + device='cuda:0'), grad: tensor([[ 9.9957e-05, 0.0000e+00, 5.3436e-05, ..., 1.6069e-04, + 2.5725e-04, 1.3247e-05], + [ 2.2674e-04, 0.0000e+00, -1.9598e-04, ..., -6.7291e-03, + 8.9169e-05, 2.0564e-04], + [ 2.4390e-04, 0.0000e+00, -4.0512e-03, ..., 3.1929e-03, + 5.7316e-04, 3.8385e-04], + ..., + [ 4.6873e-04, 0.0000e+00, 2.4259e-05, ..., 4.0197e-04, + 3.0565e-04, 1.1387e-03], + [ 1.7416e-04, 0.0000e+00, 3.7136e-03, ..., 1.8606e-03, + -7.9811e-05, 5.3835e-04], + [ 1.1492e-03, 0.0000e+00, 2.3708e-05, ..., 8.1062e-04, + 2.7537e-04, -1.0710e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0290, -0.0257, 0.0147, 0.0287, -0.0293, 0.0071, -0.0119, 0.0083, + 0.0173, 0.0027], device='cuda:0'), grad: tensor([ 0.0006, -0.0082, -0.0008, 0.0003, -0.0039, 0.0004, 0.0028, 0.0017, + 0.0069, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 230.83, cls_loss 0.0416 cls_loss_mapping 0.0765 cls_loss_causal 0.9491 re_mapping 0.0285 re_causal 0.0845 /// teacc 98.57 lr 0.00010000 +Epoch 19, weight, value: tensor([[ 0.0034, 0.0191, -0.0448, ..., -0.0371, -0.0222, 0.0360], + [ 0.0153, -0.0104, 0.0090, ..., 0.0624, -0.0208, -0.0102], + [-0.0034, -0.0264, 0.0500, ..., -0.0375, -0.0573, -0.0279], + ..., + [ 0.0036, -0.0339, -0.0150, ..., 0.0206, 0.0274, 0.0063], + [-0.0304, -0.0296, 0.0139, ..., -0.0373, 0.0256, -0.0396], + [-0.0546, -0.0259, -0.0350, ..., -0.0417, 0.0314, 0.0137]], + device='cuda:0'), grad: tensor([[ 2.6250e-04, 1.2070e-05, 1.6737e-04, ..., 9.4414e-05, + 1.0389e-04, -3.4308e-04], + [ 6.6638e-05, 1.1757e-05, 1.1915e-04, ..., -1.3912e-04, + 2.2101e-04, 6.7294e-05], + [ 1.5283e-04, 1.8090e-05, -1.0204e-03, ..., 1.1706e-04, + 8.3864e-05, 9.0897e-05], + ..., + [ 5.9664e-05, 5.5507e-06, 8.6606e-05, ..., 4.8816e-05, + 1.0586e-04, 8.7261e-05], + [ 9.3460e-05, 1.7107e-05, -4.7162e-06, ..., 8.5354e-05, + 7.4482e-04, 3.2926e-04], + [ 3.8832e-05, 3.3110e-05, 3.8445e-05, ..., 2.8834e-05, + 1.0061e-04, 1.2815e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0292, -0.0258, 0.0147, 0.0286, -0.0291, 0.0072, -0.0119, 0.0085, + 0.0172, 0.0027], device='cuda:0'), grad: tensor([ 3.7718e-04, 6.8760e-04, -1.4963e-03, -2.4357e-03, -1.2207e-03, + -5.4866e-05, 9.6369e-04, 8.4400e-04, 1.8358e-03, 4.9543e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 214.25, cls_loss 0.0428 cls_loss_mapping 0.0776 cls_loss_causal 0.9661 re_mapping 0.0271 re_causal 0.0810 /// teacc 98.26 lr 0.00010000 +Epoch 20, weight, value: tensor([[ 0.0032, 0.0192, -0.0442, ..., -0.0376, -0.0228, 0.0368], + [ 0.0140, -0.0103, 0.0078, ..., 0.0630, -0.0215, -0.0111], + [-0.0034, -0.0263, 0.0516, ..., -0.0377, -0.0583, -0.0285], + ..., + [ 0.0030, -0.0339, -0.0153, ..., 0.0199, 0.0279, 0.0060], + [-0.0307, -0.0298, 0.0129, ..., -0.0376, 0.0258, -0.0404], + [-0.0555, -0.0260, -0.0361, ..., -0.0418, 0.0315, 0.0145]], + device='cuda:0'), grad: tensor([[ 3.2735e-04, 5.1921e-08, 3.9250e-05, ..., 9.7036e-04, + 7.1383e-04, 1.6327e-03], + [ 2.4533e-04, 1.1071e-07, 6.6459e-05, ..., 1.3375e-04, + 2.6393e-04, 1.4830e-04], + [ 4.9305e-04, 1.6496e-07, 1.4102e-04, ..., 2.7895e-04, + 4.6039e-04, 6.1989e-04], + ..., + [ 1.0967e-03, 1.4342e-07, -4.2462e-04, ..., 5.6076e-04, + -1.2217e-03, 2.5809e-05], + [ 9.8050e-05, 2.7171e-07, 6.9082e-05, ..., 8.2374e-05, + 1.8132e-04, 5.0068e-04], + [ 3.2592e-04, 1.0757e-06, 4.0650e-05, ..., 2.1398e-04, + 7.0000e-04, 8.1968e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0290, -0.0262, 0.0146, 0.0287, -0.0290, 0.0071, -0.0124, 0.0083, + 0.0176, 0.0030], device='cuda:0'), grad: tensor([ 2.5883e-03, 8.6260e-04, 2.1019e-03, -4.8294e-03, -2.7599e-03, + 1.4362e-03, -1.0462e-03, -6.5136e-04, 6.4559e-06, 2.2926e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 231.31, cls_loss 0.0361 cls_loss_mapping 0.0672 cls_loss_causal 0.9177 re_mapping 0.0267 re_causal 0.0819 /// teacc 98.70 lr 0.00010000 +Epoch 21, weight, value: tensor([[ 0.0028, 0.0192, -0.0433, ..., -0.0384, -0.0234, 0.0370], + [ 0.0135, -0.0097, 0.0074, ..., 0.0637, -0.0223, -0.0110], + [-0.0038, -0.0267, 0.0537, ..., -0.0386, -0.0589, -0.0288], + ..., + [ 0.0027, -0.0341, -0.0145, ..., 0.0202, 0.0288, 0.0056], + [-0.0311, -0.0302, 0.0116, ..., -0.0383, 0.0257, -0.0414], + [-0.0564, -0.0261, -0.0370, ..., -0.0430, 0.0312, 0.0148]], + device='cuda:0'), grad: tensor([[ 3.1851e-06, 1.3590e-05, -1.1407e-05, ..., -2.5511e-04, + 7.5758e-05, -8.8549e-04], + [ 5.4948e-06, 6.8843e-05, 1.1587e-04, ..., -1.7250e-04, + 1.7118e-04, 6.3610e-04], + [ 1.5628e-06, -8.2016e-04, -1.4057e-03, ..., 1.0067e-04, + -4.8685e-04, -6.4735e-03], + ..., + [ 1.2949e-05, 1.4389e-04, 2.6202e-04, ..., 1.1271e-04, + 4.7183e-04, 1.6041e-03], + [ 3.2604e-05, 2.5764e-05, 4.0412e-05, ..., 3.1590e-05, + 2.0993e-04, 5.6791e-04], + [-6.5613e-04, 4.8578e-05, 9.1374e-05, ..., -3.0637e-05, + -2.6155e-04, -1.5955e-03]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0292, -0.0262, 0.0145, 0.0289, -0.0289, 0.0068, -0.0122, 0.0088, + 0.0174, 0.0027], device='cuda:0'), grad: tensor([-0.0005, 0.0022, -0.0245, 0.0069, 0.0042, 0.0068, 0.0006, 0.0054, + 0.0012, -0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 214.82, cls_loss 0.0323 cls_loss_mapping 0.0641 cls_loss_causal 0.9011 re_mapping 0.0253 re_causal 0.0781 /// teacc 98.62 lr 0.00010000 +Epoch 22, weight, value: tensor([[ 0.0023, 0.0192, -0.0433, ..., -0.0393, -0.0241, 0.0375], + [ 0.0132, -0.0089, 0.0063, ..., 0.0646, -0.0228, -0.0117], + [-0.0041, -0.0270, 0.0555, ..., -0.0389, -0.0598, -0.0293], + ..., + [ 0.0022, -0.0343, -0.0144, ..., 0.0194, 0.0294, 0.0053], + [-0.0313, -0.0301, 0.0115, ..., -0.0397, 0.0262, -0.0420], + [-0.0569, -0.0262, -0.0391, ..., -0.0434, 0.0310, 0.0154]], + device='cuda:0'), grad: tensor([[ 6.1989e-05, 1.3828e-05, 4.1509e-04, ..., 3.6806e-05, + 3.7575e-04, 2.1172e-04], + [ 8.8573e-05, 3.4403e-06, 1.6201e-04, ..., -3.9792e-04, + 6.2513e-04, 5.2881e-04], + [-4.1008e-04, -4.0412e-05, -4.0550e-03, ..., 1.0848e-04, + -1.1902e-03, 2.7037e-04], + ..., + [ 1.2994e-04, 6.5006e-07, 6.1417e-04, ..., 8.3148e-05, + 6.5374e-04, 7.2575e-04], + [ 1.4532e-04, 1.2532e-05, 9.2316e-04, ..., 5.7727e-05, + -4.1084e-03, -4.1771e-03], + [ 1.3399e-04, 3.2643e-07, 3.6865e-05, ..., 2.1076e-04, + -9.5320e-04, -1.9474e-03]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0291, -0.0259, 0.0145, 0.0286, -0.0290, 0.0068, -0.0123, 0.0088, + 0.0177, 0.0026], device='cuda:0'), grad: tensor([ 0.0012, 0.0015, -0.0035, 0.0009, 0.0009, 0.0082, 0.0016, 0.0017, + -0.0113, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 214.62, cls_loss 0.0252 cls_loss_mapping 0.0508 cls_loss_causal 0.8864 re_mapping 0.0240 re_causal 0.0751 /// teacc 98.64 lr 0.00010000 +Epoch 23, weight, value: tensor([[ 0.0019, 0.0196, -0.0438, ..., -0.0393, -0.0249, 0.0379], + [ 0.0126, -0.0091, 0.0047, ..., 0.0652, -0.0233, -0.0120], + [-0.0043, -0.0270, 0.0582, ..., -0.0391, -0.0605, -0.0296], + ..., + [ 0.0019, -0.0345, -0.0148, ..., 0.0188, 0.0299, 0.0048], + [-0.0316, -0.0304, 0.0108, ..., -0.0405, 0.0259, -0.0428], + [-0.0575, -0.0263, -0.0412, ..., -0.0439, 0.0309, 0.0160]], + device='cuda:0'), grad: tensor([[ 1.1081e-04, 4.3074e-08, 1.4305e-04, ..., -6.6161e-05, + -5.9366e-04, -2.2659e-03], + [ 1.0145e-04, 1.0844e-07, 8.4221e-05, ..., 9.5785e-05, + 1.0097e-04, 2.5964e-04], + [ 5.0426e-05, 1.2142e-07, -3.1261e-03, ..., 1.3494e-04, + 9.6083e-05, -1.1692e-03], + ..., + [ 4.0054e-05, 1.5006e-07, 3.0875e-05, ..., 2.2590e-04, + -2.7180e-04, 2.5082e-04], + [ 8.9645e-05, 3.0850e-07, 5.4789e-04, ..., 2.0826e-04, + 1.5473e-04, 4.9114e-04], + [ 1.0328e-03, 2.7381e-07, 3.7670e-05, ..., 8.4305e-04, + 8.3447e-04, -5.8985e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0293, -0.0260, 0.0147, 0.0289, -0.0290, 0.0072, -0.0125, 0.0086, + 0.0175, 0.0026], device='cuda:0'), grad: tensor([-0.0014, 0.0005, -0.0035, 0.0033, -0.0028, 0.0009, -0.0012, 0.0001, + 0.0012, 0.0031], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 214.69, cls_loss 0.0232 cls_loss_mapping 0.0481 cls_loss_causal 0.8837 re_mapping 0.0233 re_causal 0.0717 /// teacc 98.64 lr 0.00010000 +Epoch 24, weight, value: tensor([[ 0.0014, 0.0201, -0.0443, ..., -0.0394, -0.0255, 0.0382], + [ 0.0118, -0.0090, 0.0034, ..., 0.0660, -0.0237, -0.0128], + [-0.0048, -0.0271, 0.0606, ..., -0.0397, -0.0611, -0.0301], + ..., + [ 0.0015, -0.0348, -0.0153, ..., 0.0184, 0.0302, 0.0045], + [-0.0318, -0.0310, 0.0104, ..., -0.0412, 0.0258, -0.0431], + [-0.0582, -0.0264, -0.0422, ..., -0.0444, 0.0308, 0.0165]], + device='cuda:0'), grad: tensor([[ 1.7270e-05, 4.0000e-07, 1.3423e-04, ..., -3.1620e-05, + 2.4796e-04, -3.2187e-04], + [ 4.7892e-05, 1.5562e-06, 2.3174e-04, ..., 1.7792e-05, + 1.8358e-04, 9.0659e-05], + [ 2.6256e-05, -1.7500e-06, -2.3127e-04, ..., -2.5749e-05, + 1.7333e-04, 1.2445e-04], + ..., + [ 4.9323e-05, 1.1381e-06, 1.3504e-03, ..., 4.6700e-05, + 8.2779e-04, 5.8460e-04], + [ 2.3603e-05, 2.1886e-06, -3.0289e-03, ..., 4.6343e-05, + -3.2539e-03, -1.5841e-03], + [ 1.6856e-04, 4.7102e-07, 9.6560e-04, ..., 1.4269e-04, + 1.3037e-03, 4.6396e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0295, -0.0260, 0.0149, 0.0291, -0.0292, 0.0074, -0.0126, 0.0084, + 0.0176, 0.0025], device='cuda:0'), grad: tensor([ 0.0005, 0.0009, 0.0001, 0.0007, -0.0005, 0.0011, 0.0003, 0.0055, + -0.0135, 0.0048], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 214.82, cls_loss 0.0275 cls_loss_mapping 0.0576 cls_loss_causal 0.8671 re_mapping 0.0223 re_causal 0.0675 /// teacc 98.62 lr 0.00010000 +Epoch 25, weight, value: tensor([[ 0.0011, 0.0200, -0.0444, ..., -0.0387, -0.0260, 0.0386], + [ 0.0114, -0.0088, 0.0028, ..., 0.0668, -0.0245, -0.0133], + [-0.0049, -0.0275, 0.0624, ..., -0.0405, -0.0618, -0.0304], + ..., + [ 0.0010, -0.0349, -0.0152, ..., 0.0177, 0.0309, 0.0043], + [-0.0320, -0.0312, 0.0100, ..., -0.0419, 0.0256, -0.0440], + [-0.0591, -0.0265, -0.0442, ..., -0.0450, 0.0307, 0.0173]], + device='cuda:0'), grad: tensor([[ 1.6645e-05, 2.8580e-08, 1.7691e-04, ..., 5.0604e-05, + 1.0574e-04, 9.4995e-08], + [ 2.3171e-05, 4.6450e-08, 9.7215e-05, ..., 2.3117e-03, + 2.0657e-03, 2.3937e-03], + [ 4.6015e-05, 7.0257e-08, -1.0147e-03, ..., 7.5996e-05, + 8.3566e-05, 4.8667e-05], + ..., + [ 2.4259e-05, 8.6671e-08, 4.5466e-04, ..., 6.5470e-04, + -3.6359e-05, 6.3372e-04], + [ 1.2487e-05, 1.2713e-07, 4.9174e-05, ..., 3.9005e-04, + 3.5715e-04, 4.0722e-04], + [ 6.4410e-06, 1.8219e-07, 4.4376e-05, ..., -3.7727e-03, + -3.0651e-03, -3.9330e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0294, -0.0260, 0.0149, 0.0292, -0.0291, 0.0073, -0.0127, 0.0085, + 0.0178, 0.0022], device='cuda:0'), grad: tensor([ 0.0004, 0.0079, -0.0007, 0.0007, 0.0006, 0.0006, -0.0003, 0.0019, + 0.0013, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 214.53, cls_loss 0.0252 cls_loss_mapping 0.0548 cls_loss_causal 0.8652 re_mapping 0.0218 re_causal 0.0684 /// teacc 98.56 lr 0.00010000 +Epoch 26, weight, value: tensor([[ 0.0007, 0.0200, -0.0445, ..., -0.0391, -0.0264, 0.0389], + [ 0.0105, -0.0088, 0.0020, ..., 0.0675, -0.0252, -0.0139], + [-0.0048, -0.0275, 0.0638, ..., -0.0412, -0.0625, -0.0309], + ..., + [ 0.0002, -0.0351, -0.0143, ..., 0.0178, 0.0315, 0.0040], + [-0.0319, -0.0317, 0.0102, ..., -0.0422, 0.0257, -0.0446], + [-0.0598, -0.0265, -0.0469, ..., -0.0458, 0.0305, 0.0182]], + device='cuda:0'), grad: tensor([[ 9.4622e-06, 1.3243e-06, 1.0405e-03, ..., -7.5102e-05, + 8.0168e-05, -6.8128e-05], + [ 6.4075e-05, -6.1810e-05, -5.5075e-04, ..., -2.7227e-04, + 1.3709e-04, 4.1068e-05], + [ 7.8917e-05, 3.4839e-05, -4.0894e-03, ..., 4.6921e-04, + 1.8024e-04, -1.3227e-03], + ..., + [ 4.8019e-06, 8.6129e-06, 1.6010e-04, ..., 5.0396e-05, + -3.7003e-04, 2.3806e-04], + [ 4.9084e-05, 2.6412e-06, 2.1286e-03, ..., 1.6069e-04, + 3.8624e-04, 9.8228e-04], + [ 6.9976e-05, 1.2498e-06, 1.8620e-04, ..., 1.0037e-04, + -1.1522e-04, -4.5395e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0296, -0.0264, 0.0151, 0.0292, -0.0290, 0.0068, -0.0125, 0.0091, + 0.0180, 0.0020], device='cuda:0'), grad: tensor([ 0.0021, -0.0079, -0.0042, 0.0030, 0.0004, 0.0006, -0.0005, 0.0009, + 0.0054, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 214.72, cls_loss 0.0244 cls_loss_mapping 0.0504 cls_loss_causal 0.8138 re_mapping 0.0216 re_causal 0.0655 /// teacc 98.60 lr 0.00010000 +Epoch 27, weight, value: tensor([[ 0.0007, 0.0200, -0.0438, ..., -0.0388, -0.0268, 0.0391], + [ 0.0101, -0.0085, 0.0004, ..., 0.0680, -0.0259, -0.0142], + [-0.0048, -0.0277, 0.0663, ..., -0.0421, -0.0631, -0.0311], + ..., + [ 0.0002, -0.0349, -0.0149, ..., 0.0182, 0.0321, 0.0037], + [-0.0321, -0.0322, 0.0094, ..., -0.0430, 0.0254, -0.0452], + [-0.0609, -0.0267, -0.0469, ..., -0.0467, 0.0305, 0.0190]], + device='cuda:0'), grad: tensor([[ 2.1785e-05, 0.0000e+00, -7.0214e-05, ..., -5.6326e-06, + 8.1301e-05, -1.9932e-04], + [ 1.1331e-04, 0.0000e+00, 2.1863e-04, ..., 9.8169e-05, + 1.0407e-04, 3.8177e-05], + [ 2.4423e-05, 0.0000e+00, -7.8583e-04, ..., 2.8372e-05, + 2.4283e-04, 7.9334e-05], + ..., + [ 6.1512e-05, 0.0000e+00, 3.1447e-04, ..., 5.3376e-05, + -7.7105e-04, 1.4722e-04], + [ 2.8715e-05, 0.0000e+00, 7.7307e-05, ..., 3.9935e-05, + 7.0930e-05, -3.0327e-04], + [-4.0233e-05, 0.0000e+00, 3.1531e-05, ..., -7.2241e-05, + 6.5804e-04, -6.8665e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0297, -0.0265, 0.0152, 0.0289, -0.0292, 0.0070, -0.0125, 0.0092, + 0.0180, 0.0022], device='cuda:0'), grad: tensor([-3.3319e-05, 6.3133e-04, 3.1042e-04, -6.7062e-03, 1.0900e-03, + 7.5626e-04, 9.0933e-04, -3.0637e-04, 2.6569e-03, 6.9523e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 230.31, cls_loss 0.0193 cls_loss_mapping 0.0413 cls_loss_causal 0.8155 re_mapping 0.0201 re_causal 0.0634 /// teacc 98.79 lr 0.00010000 +Epoch 28, weight, value: tensor([[ 0.0003, 0.0199, -0.0435, ..., -0.0393, -0.0272, 0.0396], + [ 0.0092, -0.0080, 0.0003, ..., 0.0687, -0.0265, -0.0145], + [-0.0047, -0.0277, 0.0680, ..., -0.0428, -0.0637, -0.0314], + ..., + [-0.0001, -0.0353, -0.0155, ..., 0.0177, 0.0326, 0.0032], + [-0.0324, -0.0324, 0.0091, ..., -0.0438, 0.0255, -0.0458], + [-0.0616, -0.0268, -0.0476, ..., -0.0474, 0.0303, 0.0198]], + device='cuda:0'), grad: tensor([[ 5.8919e-05, 0.0000e+00, -4.7594e-05, ..., 7.1406e-05, + 7.5340e-05, -1.7905e-04], + [ 1.5903e-04, 0.0000e+00, 2.5347e-05, ..., -4.8488e-05, + 9.4175e-05, 1.1720e-05], + [ 1.2076e-04, 0.0000e+00, -5.6190e-03, ..., 9.5785e-05, + -1.5516e-03, 4.6462e-05], + ..., + [ 4.7892e-05, 0.0000e+00, 5.3787e-03, ..., 4.5240e-05, + 1.4391e-03, 1.0741e-04], + [-1.1482e-03, 0.0000e+00, 6.4373e-05, ..., -4.8518e-04, + -2.4581e-04, 2.3639e-04], + [ 1.4365e-04, 0.0000e+00, 6.3181e-05, ..., 6.1929e-05, + 1.8269e-05, -1.9443e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0295, -0.0265, 0.0154, 0.0288, -0.0290, 0.0068, -0.0125, 0.0089, + 0.0182, 0.0021], device='cuda:0'), grad: tensor([ 2.0280e-05, 2.1732e-04, -4.3449e-03, 6.4278e-04, 2.9325e-04, + -9.7990e-05, 7.3910e-04, 4.5242e-03, -2.0294e-03, 3.2872e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 214.55, cls_loss 0.0158 cls_loss_mapping 0.0356 cls_loss_causal 0.8069 re_mapping 0.0199 re_causal 0.0643 /// teacc 98.76 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0003, 0.0198, -0.0434, ..., -0.0396, -0.0274, 0.0400], + [ 0.0088, -0.0079, -0.0009, ..., 0.0691, -0.0271, -0.0149], + [-0.0051, -0.0276, 0.0694, ..., -0.0434, -0.0645, -0.0315], + ..., + [-0.0006, -0.0361, -0.0152, ..., 0.0177, 0.0330, 0.0029], + [-0.0322, -0.0326, 0.0085, ..., -0.0442, 0.0253, -0.0465], + [-0.0627, -0.0270, -0.0482, ..., -0.0477, 0.0303, 0.0203]], + device='cuda:0'), grad: tensor([[ 1.7405e-05, 0.0000e+00, -2.3857e-05, ..., 1.7941e-05, + 9.9719e-05, 1.6078e-05], + [ 2.6539e-05, 0.0000e+00, 6.1035e-05, ..., -1.4162e-04, + 3.4183e-05, 5.4628e-05], + [ 1.8269e-05, 0.0000e+00, -1.1377e-03, ..., 3.3379e-05, + -4.7088e-05, 4.3958e-05], + ..., + [ 2.1294e-05, 0.0000e+00, 5.1641e-04, ..., 3.5137e-05, + 1.8616e-03, 2.5520e-03], + [ 3.7074e-05, 0.0000e+00, 1.5354e-04, ..., 2.5585e-05, + 9.7394e-05, 1.4758e-04], + [-2.8372e-04, 0.0000e+00, -4.0233e-05, ..., -3.2783e-05, + -2.3193e-03, -3.5610e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0295, -0.0268, 0.0151, 0.0293, -0.0290, 0.0066, -0.0127, 0.0092, + 0.0182, 0.0020], device='cuda:0'), grad: tensor([ 1.0771e-04, 9.6858e-06, -1.0481e-03, 5.4741e-04, 8.3494e-04, + 1.0329e-04, -6.0499e-05, 3.4904e-03, 4.3774e-04, -4.4289e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 231.24, cls_loss 0.0141 cls_loss_mapping 0.0330 cls_loss_causal 0.8207 re_mapping 0.0193 re_causal 0.0625 /// teacc 98.83 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0006, 0.0199, -0.0429, ..., -0.0393, -0.0278, 0.0402], + [ 0.0083, -0.0079, -0.0013, ..., 0.0698, -0.0275, -0.0147], + [-0.0050, -0.0275, 0.0714, ..., -0.0439, -0.0650, -0.0316], + ..., + [-0.0011, -0.0362, -0.0152, ..., 0.0175, 0.0334, 0.0022], + [-0.0323, -0.0326, 0.0083, ..., -0.0443, 0.0251, -0.0471], + [-0.0636, -0.0270, -0.0505, ..., -0.0483, 0.0303, 0.0209]], + device='cuda:0'), grad: tensor([[ 1.3821e-05, 0.0000e+00, 8.7261e-05, ..., 1.2830e-05, + 4.4197e-05, -2.2836e-06], + [ 8.3148e-05, 0.0000e+00, 3.9744e-04, ..., -6.4671e-05, + 6.0230e-05, 3.6001e-05], + [ 3.8892e-05, 0.0000e+00, -5.2834e-03, ..., 1.4722e-05, + 3.2455e-05, 2.6733e-05], + ..., + [ 4.3958e-05, 0.0000e+00, 1.1988e-03, ..., 2.1666e-05, + -1.9622e-04, -4.9314e-07], + [ 1.0215e-05, 0.0000e+00, 2.2793e-03, ..., 2.1830e-05, + -5.5730e-05, 3.1918e-05], + [ 4.5896e-05, 0.0000e+00, 4.7493e-04, ..., 3.2336e-05, + 2.1076e-03, 3.0975e-03]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0295, -0.0267, 0.0152, 0.0293, -0.0291, 0.0069, -0.0128, 0.0091, + 0.0183, 0.0018], device='cuda:0'), grad: tensor([ 0.0003, 0.0009, -0.0109, 0.0016, 0.0002, -0.0035, 0.0003, 0.0025, + 0.0036, 0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 214.59, cls_loss 0.0159 cls_loss_mapping 0.0374 cls_loss_causal 0.7784 re_mapping 0.0186 re_causal 0.0582 /// teacc 98.66 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0009, 0.0199, -0.0413, ..., -0.0400, -0.0282, 0.0405], + [ 0.0069, -0.0079, -0.0015, ..., 0.0704, -0.0279, -0.0151], + [-0.0049, -0.0275, 0.0722, ..., -0.0443, -0.0655, -0.0322], + ..., + [-0.0018, -0.0363, -0.0152, ..., 0.0170, 0.0340, 0.0019], + [-0.0326, -0.0327, 0.0074, ..., -0.0450, 0.0252, -0.0477], + [-0.0643, -0.0270, -0.0507, ..., -0.0488, 0.0302, 0.0217]], + device='cuda:0'), grad: tensor([[ 1.8072e-04, 5.2678e-08, -6.2943e-05, ..., 1.7846e-04, + 3.5584e-05, -6.2943e-05], + [ 3.2306e-04, 1.4540e-07, 1.0061e-04, ..., 4.3839e-05, + 1.0967e-04, 6.4909e-05], + [ 2.4986e-04, 1.4994e-07, -3.9744e-04, ..., 2.1970e-04, + 6.0976e-05, 1.1897e-04], + ..., + [ 1.9324e-04, 2.8033e-07, 1.8942e-04, ..., 3.0494e-04, + -3.4404e-04, 3.3349e-05], + [ 1.3506e-04, 2.2002e-07, 2.9922e-05, ..., 1.5664e-04, + 4.3035e-05, 1.2189e-04], + [ 1.0824e-03, 8.7486e-08, 1.0496e-04, ..., 9.6703e-04, + 2.5392e-04, 6.4731e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0293, -0.0267, 0.0151, 0.0292, -0.0291, 0.0066, -0.0127, 0.0091, + 0.0184, 0.0020], device='cuda:0'), grad: tensor([ 0.0004, 0.0007, 0.0004, 0.0003, -0.0057, -0.0010, 0.0012, 0.0003, + 0.0005, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 215.19, cls_loss 0.0149 cls_loss_mapping 0.0343 cls_loss_causal 0.7769 re_mapping 0.0188 re_causal 0.0583 /// teacc 98.72 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0012, 0.0204, -0.0398, ..., -0.0399, -0.0288, 0.0406], + [ 0.0064, -0.0080, -0.0025, ..., 0.0706, -0.0285, -0.0154], + [-0.0052, -0.0275, 0.0734, ..., -0.0454, -0.0660, -0.0325], + ..., + [-0.0023, -0.0363, -0.0155, ..., 0.0174, 0.0345, 0.0017], + [-0.0326, -0.0331, 0.0076, ..., -0.0451, 0.0250, -0.0485], + [-0.0650, -0.0271, -0.0517, ..., -0.0493, 0.0302, 0.0224]], + device='cuda:0'), grad: tensor([[ 2.1305e-03, 0.0000e+00, 7.1168e-05, ..., 2.8095e-03, + 2.9874e-04, 1.7996e-03], + [ 3.2353e-04, 0.0000e+00, 1.2410e-04, ..., -3.1567e-04, + 5.7727e-05, 6.5088e-05], + [ 1.7190e-04, 0.0000e+00, -2.3198e-04, ..., 3.5715e-04, + 3.4595e-04, 1.5640e-04], + ..., + [-6.3324e-04, 0.0000e+00, -2.1935e-03, ..., -2.0275e-03, + -3.7308e-03, -1.4961e-04], + [ 2.9755e-04, 0.0000e+00, 1.3685e-04, ..., 4.2748e-04, + 1.7166e-04, 3.3116e-04], + [ 8.0872e-04, 0.0000e+00, -3.0547e-05, ..., 1.0996e-03, + 8.5735e-04, 6.9714e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0293, -0.0273, 0.0150, 0.0291, -0.0291, 0.0069, -0.0125, 0.0092, + 0.0186, 0.0020], device='cuda:0'), grad: tensor([ 0.0056, 0.0002, 0.0005, -0.0005, 0.0098, 0.0003, -0.0115, -0.0083, + 0.0012, 0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 214.89, cls_loss 0.0165 cls_loss_mapping 0.0359 cls_loss_causal 0.7826 re_mapping 0.0178 re_causal 0.0561 /// teacc 98.58 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0018, 0.0203, -0.0400, ..., -0.0407, -0.0293, 0.0411], + [ 0.0054, -0.0080, -0.0032, ..., 0.0712, -0.0289, -0.0157], + [-0.0049, -0.0273, 0.0753, ..., -0.0451, -0.0667, -0.0326], + ..., + [-0.0027, -0.0366, -0.0156, ..., 0.0171, 0.0348, 0.0012], + [-0.0329, -0.0334, 0.0067, ..., -0.0456, 0.0249, -0.0496], + [-0.0658, -0.0271, -0.0523, ..., -0.0499, 0.0300, 0.0228]], + device='cuda:0'), grad: tensor([[ 5.7310e-05, 4.9211e-06, 4.7755e-04, ..., 4.3005e-05, + 3.9387e-04, 8.0633e-04], + [ 1.2314e-04, 2.6543e-06, 1.3602e-04, ..., 6.6519e-05, + 1.7011e-04, 9.9778e-05], + [ 1.5402e-04, -2.2936e-04, -2.4354e-04, ..., 1.5271e-04, + 2.7466e-04, 3.9077e-04], + ..., + [ 1.6937e-03, 8.3372e-06, 5.1975e-04, ..., 9.8896e-04, + 3.0017e-04, 4.2176e-04], + [ 6.1929e-05, 1.9419e-04, 7.7438e-04, ..., -1.5819e-04, + -1.0830e-04, 2.7871e-04], + [ 4.0030e-04, 2.0477e-07, 1.9884e-04, ..., 2.4939e-04, + 3.4523e-04, -6.5565e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0296, -0.0275, 0.0155, 0.0295, -0.0291, 0.0070, -0.0125, 0.0091, + 0.0181, 0.0019], device='cuda:0'), grad: tensor([ 0.0015, 0.0008, 0.0006, 0.0029, -0.0044, -0.0056, 0.0005, 0.0034, + -0.0006, 0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 214.85, cls_loss 0.0144 cls_loss_mapping 0.0344 cls_loss_causal 0.7362 re_mapping 0.0173 re_causal 0.0536 /// teacc 98.83 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0022, 0.0201, -0.0400, ..., -0.0408, -0.0295, 0.0412], + [ 0.0048, -0.0081, -0.0044, ..., 0.0718, -0.0291, -0.0168], + [-0.0053, -0.0270, 0.0766, ..., -0.0455, -0.0672, -0.0329], + ..., + [-0.0032, -0.0366, -0.0161, ..., 0.0166, 0.0355, 0.0009], + [-0.0329, -0.0337, 0.0070, ..., -0.0460, 0.0248, -0.0503], + [-0.0665, -0.0272, -0.0522, ..., -0.0503, 0.0298, 0.0237]], + device='cuda:0'), grad: tensor([[ 1.4737e-05, 0.0000e+00, -7.9632e-05, ..., 4.6402e-05, + 2.4676e-05, -1.0186e-04], + [ 3.5644e-05, 0.0000e+00, 6.7949e-06, ..., 1.6451e-04, + 1.8966e-04, 5.8889e-04], + [ 1.4707e-05, 0.0000e+00, 2.3559e-05, ..., 2.8297e-05, + 7.3195e-05, 6.5506e-05], + ..., + [ 2.9489e-05, 0.0000e+00, -2.4512e-05, ..., 9.9182e-05, + -9.3579e-05, 2.0349e-04], + [ 1.8820e-05, 0.0000e+00, 1.3690e-06, ..., 1.4889e-04, + 9.9301e-05, 2.7132e-04], + [ 1.0306e-04, 0.0000e+00, 2.8670e-05, ..., -5.2166e-04, + -2.7061e-04, -1.3466e-03]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0300, -0.0277, 0.0153, 0.0297, -0.0291, 0.0073, -0.0128, 0.0091, + 0.0184, 0.0021], device='cuda:0'), grad: tensor([-1.5974e-05, 1.1120e-03, 2.5940e-04, -7.8773e-04, 3.1185e-04, + 8.2874e-04, -1.1230e-04, 2.4152e-04, 5.3120e-04, -2.3670e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 214.69, cls_loss 0.0119 cls_loss_mapping 0.0304 cls_loss_causal 0.7257 re_mapping 0.0165 re_causal 0.0524 /// teacc 98.79 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0028, 0.0200, -0.0395, ..., -0.0412, -0.0298, 0.0415], + [ 0.0044, -0.0082, -0.0052, ..., 0.0722, -0.0293, -0.0172], + [-0.0055, -0.0270, 0.0780, ..., -0.0460, -0.0675, -0.0333], + ..., + [-0.0038, -0.0365, -0.0164, ..., 0.0168, 0.0359, 0.0005], + [-0.0334, -0.0337, 0.0064, ..., -0.0462, 0.0243, -0.0512], + [-0.0671, -0.0273, -0.0527, ..., -0.0508, 0.0297, 0.0243]], + device='cuda:0'), grad: tensor([[ 7.1600e-06, 1.1525e-08, -1.1406e-03, ..., -1.2264e-05, + -8.0764e-05, -4.8304e-04], + [ 1.0237e-05, 4.1793e-08, 2.0623e-05, ..., -9.2313e-06, + 3.4451e-05, 4.2289e-05], + [ 6.7689e-06, 8.9232e-08, 9.1410e-04, ..., 1.6317e-05, + 1.3769e-04, 4.2582e-04], + ..., + [ 1.6719e-05, 3.1549e-08, -5.1260e-06, ..., 2.5690e-05, + -1.5855e-04, 9.9421e-05], + [ 2.0042e-05, 6.5775e-08, -6.1356e-06, ..., 5.2243e-05, + -1.0383e-04, 1.8966e-04], + [ 2.5868e-05, 1.3923e-07, 9.9957e-05, ..., -4.1783e-05, + 5.5820e-05, -3.7432e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0301, -0.0278, 0.0153, 0.0294, -0.0288, 0.0072, -0.0123, 0.0092, + 0.0183, 0.0019], device='cuda:0'), grad: tensor([-1.1959e-03, 9.3460e-05, 1.0500e-03, -9.7942e-04, 4.2748e-04, + 8.5402e-04, -2.8563e-04, 4.3184e-05, 2.5615e-05, -3.2008e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 214.74, cls_loss 0.0168 cls_loss_mapping 0.0401 cls_loss_causal 0.7513 re_mapping 0.0155 re_causal 0.0509 /// teacc 98.68 lr 0.00010000 +Epoch 36, weight, value: tensor([[-3.1189e-03, 2.0274e-02, -3.8550e-02, ..., -4.1353e-02, + -2.9999e-02, 4.1864e-02], + [ 4.1261e-03, -8.0893e-03, -6.1065e-03, ..., 7.2704e-02, + -2.9789e-02, -1.7147e-02], + [-5.7507e-03, -2.6828e-02, 7.9174e-02, ..., -4.6470e-02, + -6.7990e-02, -3.4014e-02], + ..., + [-4.0540e-03, -3.7017e-02, -1.6908e-02, ..., 1.6857e-02, + 3.6527e-02, -1.8020e-05], + [-3.3640e-02, -3.3073e-02, 5.8164e-03, ..., -4.6488e-02, + 2.4486e-02, -5.1553e-02], + [-6.8085e-02, -2.7266e-02, -5.3921e-02, ..., -5.1335e-02, + 2.9475e-02, 2.5001e-02]], device='cuda:0'), grad: tensor([[ 1.0870e-05, 2.3007e-05, 1.8482e-03, ..., 1.4007e-04, + 3.9637e-05, 5.0783e-04], + [ 1.5117e-05, 1.7490e-03, 5.6458e-03, ..., 1.9445e-03, + 7.1108e-05, 3.1680e-05], + [ 1.1235e-05, -2.1782e-03, -6.5880e-03, ..., -2.8515e-03, + 2.4259e-05, 1.4913e-04], + ..., + [ 2.0027e-05, 1.7583e-04, 5.5885e-04, ..., 2.9707e-04, + -2.2459e-04, 3.7700e-05], + [ 8.1182e-05, 2.1487e-05, -4.6806e-03, ..., 1.9360e-04, + 7.8440e-05, -1.0920e-03], + [-5.0843e-05, 1.2100e-05, 2.5511e-04, ..., 5.9783e-05, + 4.9472e-05, -4.8184e-04]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0301, -0.0277, 0.0151, 0.0292, -0.0288, 0.0073, -0.0122, 0.0092, + 0.0185, 0.0017], device='cuda:0'), grad: tensor([ 0.0054, 0.0108, -0.0125, 0.0036, 0.0004, 0.0010, 0.0032, 0.0010, + -0.0121, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 230.88, cls_loss 0.0119 cls_loss_mapping 0.0323 cls_loss_causal 0.7519 re_mapping 0.0153 re_causal 0.0503 /// teacc 98.86 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0035, 0.0205, -0.0380, ..., -0.0415, -0.0302, 0.0422], + [ 0.0037, -0.0085, -0.0077, ..., 0.0732, -0.0302, -0.0175], + [-0.0061, -0.0265, 0.0805, ..., -0.0467, -0.0685, -0.0343], + ..., + [-0.0046, -0.0370, -0.0164, ..., 0.0166, 0.0369, -0.0006], + [-0.0342, -0.0335, 0.0055, ..., -0.0468, 0.0246, -0.0521], + [-0.0691, -0.0273, -0.0554, ..., -0.0523, 0.0296, 0.0259]], + device='cuda:0'), grad: tensor([[ 2.2650e-05, 0.0000e+00, 9.6858e-06, ..., 5.9307e-05, + 1.0274e-05, -2.3052e-05], + [ 2.9176e-05, 0.0000e+00, -1.1265e-04, ..., -3.2520e-03, + 6.7018e-06, 1.7956e-05], + [ 5.9247e-05, 0.0000e+00, -1.7726e-04, ..., 2.1267e-03, + 5.9642e-06, -5.2243e-05], + ..., + [ 1.1063e-04, 0.0000e+00, 4.7594e-05, ..., 5.0545e-04, + 3.0845e-05, 2.4557e-04], + [ 3.7044e-05, 0.0000e+00, 6.1274e-05, ..., 2.4939e-04, + 5.7459e-05, 4.2737e-05], + [ 1.4532e-04, 0.0000e+00, 2.7984e-05, ..., 7.2479e-05, + 4.5568e-05, -8.3017e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0302, -0.0280, 0.0151, 0.0294, -0.0283, 0.0070, -0.0124, 0.0094, + 0.0188, 0.0016], device='cuda:0'), grad: tensor([ 5.9873e-05, -3.5458e-03, 2.1439e-03, 2.8157e-04, 5.5933e-04, + 7.0095e-05, -1.2112e-04, 1.0300e-03, 4.2462e-04, -9.0408e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 214.27, cls_loss 0.0120 cls_loss_mapping 0.0308 cls_loss_causal 0.7554 re_mapping 0.0157 re_causal 0.0514 /// teacc 98.79 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0039, 0.0205, -0.0375, ..., -0.0416, -0.0305, 0.0423], + [ 0.0037, -0.0085, -0.0094, ..., 0.0740, -0.0306, -0.0180], + [-0.0064, -0.0265, 0.0817, ..., -0.0469, -0.0691, -0.0346], + ..., + [-0.0052, -0.0369, -0.0157, ..., 0.0165, 0.0375, -0.0008], + [-0.0346, -0.0336, 0.0047, ..., -0.0476, 0.0244, -0.0527], + [-0.0697, -0.0275, -0.0551, ..., -0.0531, 0.0296, 0.0267]], + device='cuda:0'), grad: tensor([[ 9.7454e-06, 0.0000e+00, 2.3674e-06, ..., -3.2258e-04, + 4.0196e-06, -6.7759e-04], + [ 6.3360e-05, 0.0000e+00, -2.4517e-07, ..., -5.8055e-05, + 8.1584e-06, 8.3745e-06], + [ 2.0576e-04, 0.0000e+00, -7.4208e-05, ..., 1.5318e-04, + 8.0466e-06, 1.5229e-05], + ..., + [ 7.3969e-05, 0.0000e+00, 1.3061e-05, ..., 9.0361e-05, + -2.4274e-05, 4.1842e-05], + [ 2.0504e-05, 0.0000e+00, 1.8865e-05, ..., 4.2737e-05, + 5.0366e-06, -3.6478e-05], + [ 5.4330e-05, 0.0000e+00, 7.2941e-06, ..., 6.4433e-05, + -1.9953e-05, 2.8864e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0304, -0.0281, 0.0150, 0.0290, -0.0283, 0.0074, -0.0126, 0.0096, + 0.0187, 0.0018], device='cuda:0'), grad: tensor([-6.7139e-04, -5.0336e-05, 2.0242e-04, -1.0681e-04, -4.1199e-04, + 1.7679e-04, 5.5695e-04, 1.5664e-04, -4.3821e-04, 5.8508e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 230.22, cls_loss 0.0145 cls_loss_mapping 0.0326 cls_loss_causal 0.7395 re_mapping 0.0161 re_causal 0.0485 /// teacc 98.95 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0042, 0.0203, -0.0363, ..., -0.0420, -0.0309, 0.0430], + [ 0.0024, -0.0085, -0.0102, ..., 0.0745, -0.0307, -0.0184], + [-0.0062, -0.0265, 0.0828, ..., -0.0476, -0.0700, -0.0348], + ..., + [-0.0055, -0.0375, -0.0149, ..., 0.0168, 0.0379, -0.0016], + [-0.0348, -0.0340, 0.0036, ..., -0.0483, 0.0244, -0.0530], + [-0.0712, -0.0277, -0.0563, ..., -0.0536, 0.0294, 0.0271]], + device='cuda:0'), grad: tensor([[ 2.3592e-04, 4.3539e-08, -9.5665e-06, ..., 2.1327e-04, + 2.5317e-05, 4.8816e-05], + [ 4.5490e-04, 1.6554e-07, 6.3598e-05, ..., 3.2163e-04, + 1.6212e-05, -3.2723e-05], + [ 2.8634e-04, 4.0280e-08, 3.2842e-05, ..., 3.1114e-04, + 1.3731e-05, 2.6897e-05], + ..., + [ 1.6105e-04, 2.1758e-07, -5.2750e-05, ..., 6.8963e-05, + -9.7811e-05, 2.5010e-04], + [ 1.8024e-04, 4.8545e-08, 2.3901e-05, ..., 1.8990e-04, + 1.7986e-05, 9.9778e-05], + [ 4.8351e-04, 2.5909e-06, 1.8343e-05, ..., 1.2517e-04, + 1.4293e-04, -7.1096e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0301, -0.0283, 0.0151, 0.0292, -0.0282, 0.0073, -0.0122, 0.0097, + 0.0187, 0.0013], device='cuda:0'), grad: tensor([ 0.0004, 0.0007, 0.0006, 0.0002, 0.0006, 0.0002, -0.0034, 0.0004, + 0.0005, -0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 214.64, cls_loss 0.0111 cls_loss_mapping 0.0279 cls_loss_causal 0.7313 re_mapping 0.0151 re_causal 0.0481 /// teacc 98.69 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0044, 0.0202, -0.0360, ..., -0.0420, -0.0310, 0.0430], + [ 0.0015, -0.0087, -0.0105, ..., 0.0752, -0.0312, -0.0187], + [-0.0057, -0.0260, 0.0843, ..., -0.0483, -0.0703, -0.0352], + ..., + [-0.0061, -0.0378, -0.0152, ..., 0.0165, 0.0386, -0.0021], + [-0.0349, -0.0343, 0.0029, ..., -0.0487, 0.0244, -0.0538], + [-0.0717, -0.0278, -0.0561, ..., -0.0537, 0.0292, 0.0283]], + device='cuda:0'), grad: tensor([[ 3.6091e-05, 1.3283e-07, -5.2035e-05, ..., 1.4514e-05, + 2.1026e-05, -6.3658e-05], + [ 5.9932e-05, 3.5297e-07, 1.7211e-05, ..., 8.6650e-06, + 6.3956e-05, 1.8597e-05], + [ 4.0555e-04, 3.3155e-07, -4.4793e-05, ..., 1.0115e-04, + 9.4175e-05, 2.9996e-05], + ..., + [ 1.4639e-04, 1.1306e-06, -7.7069e-05, ..., 4.4584e-05, + -4.7040e-04, 7.6234e-05], + [ 6.5207e-05, 4.0000e-07, 2.0698e-05, ..., 2.6822e-05, + 8.7261e-05, 7.0870e-05], + [ 5.8115e-05, 6.6450e-07, 2.2039e-05, ..., 2.0579e-05, + 5.0449e-04, 1.9789e-04]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0302, -0.0284, 0.0153, 0.0291, -0.0285, 0.0069, -0.0123, 0.0096, + 0.0184, 0.0021], device='cuda:0'), grad: tensor([-4.7445e-05, 2.0039e-04, 6.5327e-04, 7.0810e-04, -1.2007e-03, + -8.1825e-04, 1.5187e-04, -4.7898e-04, 1.4961e-04, 6.8140e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 214.35, cls_loss 0.0118 cls_loss_mapping 0.0314 cls_loss_causal 0.7194 re_mapping 0.0149 re_causal 0.0470 /// teacc 98.88 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0053, 0.0203, -0.0354, ..., -0.0425, -0.0313, 0.0432], + [ 0.0013, -0.0087, -0.0105, ..., 0.0767, -0.0316, -0.0191], + [-0.0060, -0.0259, 0.0853, ..., -0.0494, -0.0706, -0.0356], + ..., + [-0.0071, -0.0380, -0.0150, ..., 0.0157, 0.0394, -0.0023], + [-0.0352, -0.0349, 0.0023, ..., -0.0494, 0.0242, -0.0539], + [-0.0724, -0.0287, -0.0570, ..., -0.0539, 0.0290, 0.0286]], + device='cuda:0'), grad: tensor([[ 4.5806e-05, 3.1125e-06, -1.2115e-05, ..., 2.7522e-05, + 1.4819e-05, 3.8266e-05], + [ 3.5256e-05, 9.3505e-06, 4.7654e-05, ..., -8.9407e-05, + 2.3946e-05, 2.5764e-05], + [ 1.4770e-04, 2.1204e-05, 5.9575e-05, ..., 3.6895e-05, + 4.8131e-05, 2.3022e-05], + ..., + [ 1.2481e-04, 3.9816e-05, 9.0659e-05, ..., 1.0383e-04, + 6.0588e-05, 1.4745e-05], + [-2.1374e-04, 3.0160e-05, 3.9935e-05, ..., 6.7472e-05, + 2.5010e-04, -4.3654e-04], + [ 2.2340e-04, 3.5673e-05, 4.7863e-05, ..., 1.1694e-04, + 1.3791e-05, 2.2709e-04]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0305, -0.0278, 0.0149, 0.0293, -0.0283, 0.0068, -0.0123, 0.0096, + 0.0185, 0.0018], device='cuda:0'), grad: tensor([ 2.6584e-04, 7.1943e-05, 3.6883e-04, 1.2684e-04, -6.7091e-04, + 3.9368e-03, -3.6793e-03, 5.2929e-04, -1.9474e-03, 9.9850e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 214.43, cls_loss 0.0106 cls_loss_mapping 0.0263 cls_loss_causal 0.7191 re_mapping 0.0142 re_causal 0.0464 /// teacc 98.92 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0058, 0.0201, -0.0348, ..., -0.0427, -0.0314, 0.0435], + [ 0.0009, -0.0088, -0.0119, ..., 0.0770, -0.0320, -0.0190], + [-0.0063, -0.0255, 0.0869, ..., -0.0496, -0.0710, -0.0357], + ..., + [-0.0077, -0.0383, -0.0147, ..., 0.0161, 0.0398, -0.0026], + [-0.0357, -0.0362, 0.0022, ..., -0.0501, 0.0238, -0.0542], + [-0.0728, -0.0286, -0.0588, ..., -0.0545, 0.0289, 0.0288]], + device='cuda:0'), grad: tensor([[ 5.5172e-06, 1.0524e-06, -3.1739e-05, ..., -2.7031e-05, + 7.0892e-06, -5.7340e-05], + [ 1.2368e-05, 1.2722e-06, 1.1474e-05, ..., -3.3192e-06, + 1.2890e-05, 2.1487e-05], + [ 3.0492e-06, 1.2733e-05, 1.4845e-06, ..., 7.2606e-06, + 1.9252e-05, 2.7716e-05], + ..., + [ 2.8443e-06, 5.7332e-06, 1.5192e-05, ..., 1.0900e-05, + 7.5847e-06, 2.4796e-05], + [ 3.5651e-06, 4.9993e-06, 1.5318e-05, ..., 8.5160e-06, + 9.2909e-06, 1.6242e-05], + [ 1.2219e-04, 4.1793e-07, 1.1973e-05, ..., 7.8142e-05, + 2.6673e-05, 4.9829e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0305, -0.0279, 0.0150, 0.0293, -0.0283, 0.0070, -0.0121, 0.0101, + 0.0184, 0.0014], device='cuda:0'), grad: tensor([-6.8545e-05, 4.0650e-05, 9.4831e-05, -1.8060e-04, -1.6820e-04, + -1.0055e-04, 4.1217e-05, 6.8426e-05, 6.1989e-05, 2.1029e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 214.39, cls_loss 0.0084 cls_loss_mapping 0.0233 cls_loss_causal 0.7036 re_mapping 0.0140 re_causal 0.0463 /// teacc 98.72 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0059, 0.0201, -0.0339, ..., -0.0428, -0.0316, 0.0438], + [ 0.0005, -0.0088, -0.0129, ..., 0.0773, -0.0326, -0.0191], + [-0.0063, -0.0250, 0.0883, ..., -0.0492, -0.0712, -0.0359], + ..., + [-0.0084, -0.0390, -0.0148, ..., 0.0159, 0.0400, -0.0029], + [-0.0359, -0.0368, 0.0022, ..., -0.0506, 0.0239, -0.0547], + [-0.0734, -0.0286, -0.0601, ..., -0.0550, 0.0287, 0.0292]], + device='cuda:0'), grad: tensor([[ 2.8498e-06, 0.0000e+00, -5.1707e-06, ..., 2.7388e-05, + 4.3333e-05, 6.7115e-05], + [ 4.1127e-06, 0.0000e+00, -1.3793e-04, ..., -2.9850e-03, + 1.8388e-05, 1.3702e-05], + [ 2.4661e-06, 0.0000e+00, 1.3089e-04, ..., 1.7395e-03, + 2.6554e-05, 4.0233e-05], + ..., + [ 7.4692e-06, 0.0000e+00, -9.0480e-05, ..., 3.7885e-04, + -1.7226e-04, 6.1691e-06], + [ 4.8950e-06, 0.0000e+00, 1.3769e-05, ..., 6.0022e-05, + 5.9694e-05, 3.6335e-04], + [-2.1827e-04, 0.0000e+00, 2.0400e-05, ..., 3.4362e-05, + 5.9903e-05, -7.4673e-04]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0304, -0.0280, 0.0154, 0.0292, -0.0284, 0.0071, -0.0122, 0.0097, + 0.0186, 0.0014], device='cuda:0'), grad: tensor([ 0.0002, -0.0081, 0.0050, 0.0042, 0.0013, -0.0030, 0.0002, 0.0009, + 0.0006, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.27, cls_loss 0.0117 cls_loss_mapping 0.0308 cls_loss_causal 0.7577 re_mapping 0.0139 re_causal 0.0460 /// teacc 98.78 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0062, 0.0200, -0.0327, ..., -0.0436, -0.0318, 0.0438], + [ 0.0002, -0.0089, -0.0145, ..., 0.0778, -0.0327, -0.0191], + [-0.0065, -0.0251, 0.0908, ..., -0.0502, -0.0716, -0.0352], + ..., + [-0.0088, -0.0391, -0.0145, ..., 0.0164, 0.0404, -0.0035], + [-0.0359, -0.0373, 0.0014, ..., -0.0508, 0.0236, -0.0552], + [-0.0743, -0.0287, -0.0616, ..., -0.0558, 0.0286, 0.0296]], + device='cuda:0'), grad: tensor([[ 5.8055e-05, 0.0000e+00, -1.4031e-04, ..., 2.1905e-05, + 4.7654e-05, -4.1556e-04], + [ 1.8156e-04, 0.0000e+00, 3.2455e-05, ..., 5.8126e-04, + 2.5797e-04, 1.1750e-05], + [ 1.9634e-04, 0.0000e+00, -6.2101e-06, ..., 2.4962e-04, + 1.4091e-04, 2.1294e-05], + ..., + [-1.1730e-04, 0.0000e+00, -6.4135e-04, ..., -1.7822e-04, + -1.7204e-03, 1.5423e-05], + [ 1.1623e-04, 0.0000e+00, 2.0847e-05, ..., 1.4853e-04, + -3.5822e-05, 5.0515e-05], + [ 1.0294e-04, 0.0000e+00, 4.4495e-05, ..., 1.0484e-04, + 1.2648e-04, -1.4707e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0307, -0.0283, 0.0158, 0.0294, -0.0283, 0.0073, -0.0120, 0.0100, + 0.0183, 0.0008], device='cuda:0'), grad: tensor([-0.0004, 0.0011, 0.0007, 0.0003, -0.0056, 0.0004, 0.0062, -0.0029, + -0.0004, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 214.60, cls_loss 0.0095 cls_loss_mapping 0.0265 cls_loss_causal 0.7041 re_mapping 0.0136 re_causal 0.0429 /// teacc 98.85 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0064, 0.0200, -0.0314, ..., -0.0436, -0.0321, 0.0442], + [-0.0005, -0.0089, -0.0152, ..., 0.0783, -0.0331, -0.0193], + [-0.0065, -0.0251, 0.0915, ..., -0.0507, -0.0721, -0.0356], + ..., + [-0.0092, -0.0407, -0.0149, ..., 0.0167, 0.0405, -0.0043], + [-0.0360, -0.0375, 0.0015, ..., -0.0512, 0.0235, -0.0555], + [-0.0750, -0.0286, -0.0613, ..., -0.0562, 0.0286, 0.0308]], + device='cuda:0'), grad: tensor([[ 1.4424e-05, 0.0000e+00, -2.5153e-05, ..., 1.2644e-05, + 3.7670e-05, 2.6315e-05], + [ 1.6809e-05, 0.0000e+00, 6.4746e-06, ..., -4.8846e-05, + 9.5487e-05, 1.0175e-04], + [ 8.4713e-06, 0.0000e+00, -3.3021e-05, ..., 2.1771e-05, + 1.0842e-04, 1.0347e-04], + ..., + [ 1.5222e-05, 0.0000e+00, 6.4783e-06, ..., -1.2545e-06, + 1.4618e-05, 2.0742e-04], + [ 1.0937e-05, 0.0000e+00, 1.2614e-05, ..., 1.1563e-05, + 6.5446e-05, 1.0121e-04], + [-5.0306e-04, 0.0000e+00, 1.0423e-05, ..., -1.8048e-04, + 8.9407e-05, -1.5888e-03]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0305, -0.0286, 0.0155, 0.0295, -0.0283, 0.0073, -0.0123, 0.0095, + 0.0186, 0.0015], device='cuda:0'), grad: tensor([ 8.1599e-05, 2.7990e-04, 3.4928e-04, -1.1492e-03, 2.7161e-03, + -2.8872e-04, 3.0011e-05, 4.6134e-04, 1.7488e-04, -2.6550e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 44---------------------------------------------------- +epoch 44, time 230.66, cls_loss 0.0082 cls_loss_mapping 0.0222 cls_loss_causal 0.7015 re_mapping 0.0133 re_causal 0.0437 /// teacc 98.97 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0071, 0.0200, -0.0312, ..., -0.0442, -0.0323, 0.0443], + [-0.0013, -0.0090, -0.0158, ..., 0.0792, -0.0336, -0.0190], + [-0.0063, -0.0252, 0.0933, ..., -0.0514, -0.0724, -0.0358], + ..., + [-0.0094, -0.0408, -0.0149, ..., 0.0164, 0.0411, -0.0046], + [-0.0360, -0.0376, 0.0001, ..., -0.0506, 0.0236, -0.0559], + [-0.0754, -0.0286, -0.0620, ..., -0.0568, 0.0284, 0.0316]], + device='cuda:0'), grad: tensor([[-2.0247e-06, 0.0000e+00, -1.4529e-05, ..., -2.7325e-06, + 1.4260e-05, -6.5565e-05], + [ 1.3024e-05, 0.0000e+00, 3.1799e-05, ..., -5.7399e-05, + 3.4332e-05, 1.2472e-05], + [ 5.9344e-06, 0.0000e+00, 2.8014e-05, ..., 1.9640e-05, + 3.3110e-05, 2.2516e-05], + ..., + [ 7.7784e-06, 0.0000e+00, -1.8334e-04, ..., 1.8580e-06, + -1.8239e-04, -6.5982e-05], + [ 6.6943e-06, 0.0000e+00, 9.5516e-06, ..., 1.4350e-05, + 5.1856e-06, 1.9327e-05], + [ 1.3739e-05, 0.0000e+00, 3.1769e-05, ..., 1.0841e-05, + 2.9847e-05, -1.8373e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0308, -0.0286, 0.0158, 0.0294, -0.0287, 0.0070, -0.0125, 0.0096, + 0.0193, 0.0015], device='cuda:0'), grad: tensor([-3.2067e-05, 1.9535e-05, 1.0538e-04, 1.3721e-04, -3.9525e-06, + 4.4465e-05, 5.0843e-05, -4.2748e-04, 4.4048e-05, 6.1750e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 214.58, cls_loss 0.0072 cls_loss_mapping 0.0225 cls_loss_causal 0.6976 re_mapping 0.0126 re_causal 0.0417 /// teacc 98.85 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0073, 0.0200, -0.0307, ..., -0.0441, -0.0326, 0.0449], + [-0.0015, -0.0091, -0.0164, ..., 0.0801, -0.0341, -0.0199], + [-0.0066, -0.0251, 0.0946, ..., -0.0522, -0.0725, -0.0363], + ..., + [-0.0098, -0.0405, -0.0144, ..., 0.0164, 0.0417, -0.0044], + [-0.0363, -0.0379, -0.0014, ..., -0.0511, 0.0234, -0.0563], + [-0.0758, -0.0291, -0.0628, ..., -0.0571, 0.0282, 0.0323]], + device='cuda:0'), grad: tensor([[ 2.2367e-05, 2.1071e-08, -1.4029e-05, ..., 3.0339e-05, + 1.6809e-05, 1.9938e-05], + [ 5.4568e-05, 1.5285e-07, 3.1553e-06, ..., -9.4891e-05, + 1.5646e-05, -8.3968e-06], + [ 7.5459e-05, 9.4180e-08, 1.1444e-05, ..., 3.5673e-05, + 2.4766e-05, 2.2113e-05], + ..., + [ 2.2793e-04, 3.9651e-07, 3.0100e-06, ..., 8.2374e-05, + 4.0799e-05, 1.0349e-05], + [ 4.7237e-05, 3.9581e-08, -2.4334e-05, ..., 5.4359e-05, + 7.4692e-06, 3.6985e-05], + [ 3.6001e-04, 3.4715e-07, 1.0639e-05, ..., 1.3232e-04, + 3.8922e-05, -4.4048e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0307, -0.0286, 0.0159, 0.0292, -0.0289, 0.0071, -0.0127, 0.0102, + 0.0188, 0.0016], device='cuda:0'), grad: tensor([ 8.0109e-05, -9.6440e-05, 1.6332e-04, 2.3055e-04, -1.1473e-03, + -9.6798e-05, -1.3328e-04, 3.1233e-04, 1.0788e-04, 5.8031e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 214.72, cls_loss 0.0075 cls_loss_mapping 0.0210 cls_loss_causal 0.7203 re_mapping 0.0127 re_causal 0.0426 /// teacc 98.88 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0074, 0.0201, -0.0304, ..., -0.0440, -0.0328, 0.0449], + [-0.0021, -0.0091, -0.0168, ..., 0.0806, -0.0345, -0.0202], + [-0.0070, -0.0253, 0.0957, ..., -0.0526, -0.0725, -0.0369], + ..., + [-0.0104, -0.0411, -0.0148, ..., 0.0162, 0.0422, -0.0050], + [-0.0366, -0.0381, -0.0015, ..., -0.0515, 0.0233, -0.0568], + [-0.0765, -0.0290, -0.0629, ..., -0.0574, 0.0280, 0.0333]], + device='cuda:0'), grad: tensor([[ 1.7852e-05, 3.7160e-07, 5.5122e-04, ..., 1.4193e-05, + 1.4524e-03, 3.9482e-03], + [ 1.4573e-05, 1.5879e-07, 3.6843e-06, ..., -4.6670e-05, + 2.4587e-05, 3.6955e-05], + [ 1.0572e-05, 6.4168e-07, 2.4870e-05, ..., 1.9848e-05, + 8.4937e-05, 1.4699e-04], + ..., + [ 4.3325e-06, 9.0757e-07, -3.3706e-05, ..., 8.1807e-06, + -3.1620e-05, 6.6876e-05], + [ 8.8438e-06, 1.2349e-06, 1.4521e-05, ..., 2.7984e-05, + 6.0022e-05, 9.4831e-05], + [ 7.0259e-06, 4.8196e-07, 4.6104e-05, ..., 3.5167e-06, + 1.4925e-04, 3.0208e-04]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0307, -0.0286, 0.0156, 0.0291, -0.0288, 0.0071, -0.0126, 0.0100, + 0.0188, 0.0019], device='cuda:0'), grad: tensor([ 5.7068e-03, 2.9564e-05, 2.9635e-04, -5.5194e-05, 7.1287e-05, + -7.2098e-03, 4.3035e-04, 5.6267e-05, 1.6224e-04, 5.1737e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 214.66, cls_loss 0.0075 cls_loss_mapping 0.0207 cls_loss_causal 0.6672 re_mapping 0.0131 re_causal 0.0413 /// teacc 98.85 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0080, 0.0204, -0.0304, ..., -0.0443, -0.0332, 0.0450], + [-0.0022, -0.0092, -0.0177, ..., 0.0815, -0.0348, -0.0204], + [-0.0073, -0.0255, 0.0971, ..., -0.0527, -0.0729, -0.0373], + ..., + [-0.0109, -0.0433, -0.0144, ..., 0.0154, 0.0425, -0.0054], + [-0.0367, -0.0386, -0.0028, ..., -0.0521, 0.0230, -0.0573], + [-0.0773, -0.0294, -0.0637, ..., -0.0581, 0.0277, 0.0336]], + device='cuda:0'), grad: tensor([[ 2.4363e-05, 1.6543e-07, -1.7323e-06, ..., 2.0280e-05, + 1.0782e-04, 1.6141e-04], + [ 4.0323e-05, 4.5099e-07, -5.9128e-04, ..., -1.0786e-03, + 1.9938e-05, -1.3039e-05], + [ 8.5056e-05, 2.3499e-05, 7.7486e-04, ..., 8.6880e-04, + 1.0854e-04, 1.6141e-04], + ..., + [ 2.8896e-04, -3.9458e-05, -4.0221e-04, ..., 1.5962e-04, + -4.4197e-05, 1.6233e-06], + [-1.0468e-05, 2.6799e-07, -6.1274e-05, ..., 5.3495e-05, + 2.2089e-04, 4.4131e-04], + [ 8.8425e-03, 7.1106e-07, 4.5687e-05, ..., 2.4164e-04, + 6.0260e-05, 3.8433e-03]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0311, -0.0283, 0.0159, 0.0299, -0.0288, 0.0070, -0.0125, 0.0097, + 0.0186, 0.0015], device='cuda:0'), grad: tensor([ 0.0008, -0.0019, 0.0023, 0.0008, -0.0133, -0.0037, 0.0008, 0.0003, + 0.0015, 0.0125], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 214.72, cls_loss 0.0073 cls_loss_mapping 0.0199 cls_loss_causal 0.6514 re_mapping 0.0125 re_causal 0.0395 /// teacc 98.83 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0084, 0.0202, -0.0296, ..., -0.0445, -0.0335, 0.0454], + [-0.0023, -0.0092, -0.0182, ..., 0.0821, -0.0353, -0.0207], + [-0.0074, -0.0254, 0.0982, ..., -0.0531, -0.0735, -0.0376], + ..., + [-0.0114, -0.0433, -0.0145, ..., 0.0153, 0.0430, -0.0056], + [-0.0371, -0.0387, -0.0035, ..., -0.0526, 0.0228, -0.0579], + [-0.0787, -0.0290, -0.0649, ..., -0.0583, 0.0277, 0.0341]], + device='cuda:0'), grad: tensor([[ 6.6936e-05, 0.0000e+00, -3.9861e-06, ..., 9.6142e-05, + 1.2510e-05, -4.3333e-05], + [ 8.0764e-05, 0.0000e+00, 7.4208e-06, ..., -5.7268e-04, + 7.9572e-06, 1.1727e-05], + [ 4.2677e-04, 0.0000e+00, 5.3227e-05, ..., 8.7261e-04, + 1.0639e-05, 3.0696e-05], + ..., + [ 8.2031e-06, 0.0000e+00, 3.1412e-05, ..., 7.4804e-05, + -6.0916e-05, 2.1290e-06], + [ 6.6161e-05, 0.0000e+00, 1.1319e-04, ..., 1.1677e-04, + 1.2055e-05, 2.0728e-05], + [ 1.7688e-05, 0.0000e+00, 2.1771e-05, ..., 3.0875e-05, + 1.8463e-05, 2.2158e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0308, -0.0282, 0.0160, 0.0298, -0.0285, 0.0073, -0.0127, 0.0098, + 0.0181, 0.0014], device='cuda:0'), grad: tensor([ 1.6057e-04, -5.9700e-04, 1.2722e-03, 4.8071e-05, 3.5691e-04, + 1.2434e-04, -1.9131e-03, 1.0258e-04, 3.1757e-04, 1.2839e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 214.59, cls_loss 0.0063 cls_loss_mapping 0.0189 cls_loss_causal 0.6648 re_mapping 0.0130 re_causal 0.0395 /// teacc 98.84 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0091, 0.0202, -0.0286, ..., -0.0457, -0.0337, 0.0454], + [-0.0027, -0.0093, -0.0184, ..., 0.0829, -0.0357, -0.0210], + [-0.0077, -0.0254, 0.0991, ..., -0.0538, -0.0736, -0.0380], + ..., + [-0.0118, -0.0433, -0.0146, ..., 0.0150, 0.0434, -0.0058], + [-0.0377, -0.0388, -0.0043, ..., -0.0529, 0.0227, -0.0583], + [-0.0794, -0.0289, -0.0656, ..., -0.0588, 0.0275, 0.0345]], + device='cuda:0'), grad: tensor([[ 1.4648e-05, 2.5658e-07, -5.6773e-05, ..., -2.1152e-03, + 1.1668e-05, -3.3493e-03], + [ 2.1830e-05, 1.2608e-07, 2.1219e-05, ..., 8.5980e-06, + 8.0466e-06, 2.5138e-05], + [ 1.6525e-05, 1.0012e-07, -3.2663e-05, ..., 1.7211e-05, + 4.4219e-06, 3.6031e-05], + ..., + [ 4.0412e-05, 5.2806e-07, -7.6229e-07, ..., 1.7539e-05, + -1.5512e-05, 4.2647e-05], + [ 1.2308e-05, 2.9407e-07, 5.3570e-06, ..., 1.8790e-05, + 4.8190e-05, 1.0210e-04], + [ 3.2753e-05, -2.3600e-06, 8.1137e-06, ..., 1.1826e-04, + 4.7743e-05, 2.0802e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0312, -0.0281, 0.0158, 0.0299, -0.0288, 0.0075, -0.0124, 0.0096, + 0.0184, 0.0012], device='cuda:0'), grad: tensor([-3.9978e-03, 9.4116e-05, 6.8069e-05, 2.5749e-04, -1.3554e-04, + -3.4094e-04, 3.8013e-03, 1.3971e-04, 1.2100e-04, -6.0573e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 214.70, cls_loss 0.0055 cls_loss_mapping 0.0192 cls_loss_causal 0.6834 re_mapping 0.0122 re_causal 0.0398 /// teacc 98.83 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0095, 0.0202, -0.0288, ..., -0.0459, -0.0340, 0.0454], + [-0.0026, -0.0093, -0.0193, ..., 0.0839, -0.0360, -0.0213], + [-0.0080, -0.0254, 0.1003, ..., -0.0542, -0.0740, -0.0383], + ..., + [-0.0121, -0.0433, -0.0144, ..., 0.0146, 0.0439, -0.0062], + [-0.0382, -0.0390, -0.0044, ..., -0.0536, 0.0226, -0.0587], + [-0.0798, -0.0289, -0.0663, ..., -0.0590, 0.0275, 0.0353]], + device='cuda:0'), grad: tensor([[-3.8803e-05, 3.6089e-09, -9.1672e-05, ..., 1.0908e-05, + 4.0643e-06, -8.3864e-05], + [ 8.2329e-06, 1.2806e-09, 1.7524e-05, ..., -5.1945e-05, + 6.9439e-05, 7.5027e-06], + [ 1.2137e-05, 9.3132e-10, 8.2105e-06, ..., 1.5192e-05, + 2.7984e-05, 2.0340e-05], + ..., + [-5.4628e-05, 7.2177e-09, -1.3757e-04, ..., -4.7028e-05, + -4.2033e-04, -7.6815e-06], + [ 1.6987e-05, 5.0059e-09, 2.7165e-05, ..., 2.9057e-05, + 2.0757e-05, 7.0930e-05], + [ 8.5235e-06, -6.6822e-08, 1.7703e-05, ..., 9.2983e-06, + 3.5554e-05, -1.1110e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0315, -0.0277, 0.0157, 0.0299, -0.0290, 0.0075, -0.0124, 0.0096, + 0.0185, 0.0014], device='cuda:0'), grad: tensor([-1.0473e-04, 6.1989e-05, 9.4116e-05, -6.0111e-05, 3.0828e-04, + 2.5272e-04, 6.0126e-06, -6.9427e-04, 1.6713e-04, -3.1710e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 214.63, cls_loss 0.0061 cls_loss_mapping 0.0173 cls_loss_causal 0.6699 re_mapping 0.0118 re_causal 0.0386 /// teacc 98.92 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0098, 0.0201, -0.0282, ..., -0.0459, -0.0342, 0.0456], + [-0.0028, -0.0093, -0.0205, ..., 0.0841, -0.0365, -0.0217], + [-0.0082, -0.0255, 0.1014, ..., -0.0542, -0.0743, -0.0387], + ..., + [-0.0118, -0.0418, -0.0140, ..., 0.0150, 0.0445, -0.0065], + [-0.0384, -0.0392, -0.0048, ..., -0.0541, 0.0223, -0.0589], + [-0.0805, -0.0288, -0.0667, ..., -0.0595, 0.0272, 0.0359]], + device='cuda:0'), grad: tensor([[ 1.5132e-05, 0.0000e+00, -1.2979e-05, ..., 5.4576e-06, + 8.3670e-06, -2.1711e-05], + [ 1.4745e-05, 0.0000e+00, 3.7439e-06, ..., -3.8981e-05, + 3.0160e-04, 8.1509e-06], + [ 7.2494e-06, 0.0000e+00, -1.4998e-05, ..., 8.1658e-06, + 1.4402e-05, 1.0461e-05], + ..., + [ 3.5673e-05, 0.0000e+00, -1.3903e-05, ..., 1.9655e-05, + -4.4894e-04, 8.4937e-05], + [ 1.2651e-05, 0.0000e+00, 1.5005e-05, ..., 1.2912e-05, + 6.4969e-05, 5.0843e-05], + [ 1.5821e-03, 0.0000e+00, 1.1578e-05, ..., 3.9972e-06, + -1.9267e-05, 4.6086e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0314, -0.0280, 0.0158, 0.0300, -0.0292, 0.0078, -0.0127, 0.0101, + 0.0185, 0.0011], device='cuda:0'), grad: tensor([ 1.9670e-05, 6.5708e-04, 5.4091e-05, 1.1110e-04, -2.8038e-03, + 7.2479e-05, -2.8819e-05, -8.2541e-04, 2.2149e-04, 2.5234e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 214.62, cls_loss 0.0049 cls_loss_mapping 0.0141 cls_loss_causal 0.6569 re_mapping 0.0116 re_causal 0.0381 /// teacc 98.82 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0102, 0.0201, -0.0275, ..., -0.0461, -0.0344, 0.0459], + [-0.0032, -0.0094, -0.0207, ..., 0.0848, -0.0369, -0.0218], + [-0.0083, -0.0255, 0.1020, ..., -0.0547, -0.0746, -0.0390], + ..., + [-0.0125, -0.0418, -0.0140, ..., 0.0145, 0.0448, -0.0068], + [-0.0387, -0.0393, -0.0053, ..., -0.0547, 0.0222, -0.0595], + [-0.0812, -0.0288, -0.0672, ..., -0.0598, 0.0270, 0.0362]], + device='cuda:0'), grad: tensor([[ 9.4809e-07, 0.0000e+00, -8.5309e-06, ..., 1.9614e-06, + 7.3127e-06, 1.0043e-05], + [ 2.7474e-06, 0.0000e+00, 1.4648e-05, ..., -3.6269e-05, + 1.3277e-05, 5.3607e-06], + [ 1.2629e-06, 0.0000e+00, -7.3671e-05, ..., 6.4448e-06, + 2.2367e-05, 2.2143e-05], + ..., + [ 2.3432e-06, 0.0000e+00, 3.0136e-04, ..., 9.7379e-06, + 4.7874e-04, 2.3878e-04], + [ 1.5926e-06, 0.0000e+00, 1.5795e-05, ..., 9.9465e-06, + 1.2197e-05, 2.1532e-05], + [ 2.7514e-04, 0.0000e+00, -4.5210e-05, ..., 3.7968e-05, + 3.3360e-06, -1.1820e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0312, -0.0276, 0.0154, 0.0300, -0.0288, 0.0078, -0.0129, 0.0099, + 0.0184, 0.0009], device='cuda:0'), grad: tensor([ 1.9565e-05, 1.6182e-07, 4.3869e-05, -1.1988e-03, -3.3331e-04, + 2.2709e-05, 2.1785e-05, 1.2245e-03, 6.3837e-05, 1.3697e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 214.37, cls_loss 0.0053 cls_loss_mapping 0.0160 cls_loss_causal 0.6427 re_mapping 0.0121 re_causal 0.0377 /// teacc 98.75 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0104, 0.0200, -0.0272, ..., -0.0463, -0.0345, 0.0461], + [-0.0038, -0.0094, -0.0216, ..., 0.0851, -0.0372, -0.0241], + [-0.0085, -0.0256, 0.1034, ..., -0.0549, -0.0750, -0.0392], + ..., + [-0.0133, -0.0420, -0.0140, ..., 0.0141, 0.0454, -0.0068], + [-0.0390, -0.0394, -0.0064, ..., -0.0556, 0.0220, -0.0598], + [-0.0818, -0.0288, -0.0678, ..., -0.0588, 0.0266, 0.0370]], + device='cuda:0'), grad: tensor([[ 1.4976e-06, 0.0000e+00, -2.0266e-05, ..., 3.2298e-06, + 1.1213e-06, -1.8835e-05], + [ 2.2203e-06, 0.0000e+00, 6.7577e-06, ..., -2.5168e-05, + 3.4180e-06, 3.1404e-06], + [ 1.7956e-06, 0.0000e+00, 7.2658e-05, ..., 1.0423e-05, + 4.7684e-06, 7.4543e-06], + ..., + [ 5.7230e-07, 0.0000e+00, 3.7346e-06, ..., 7.7784e-06, + -3.9935e-06, 1.3024e-05], + [ 9.4390e-07, 0.0000e+00, -1.3041e-04, ..., 4.7497e-06, + -6.2168e-05, 5.1223e-06], + [ 8.2701e-06, 0.0000e+00, 3.3885e-05, ..., 5.0589e-06, + 1.4668e-06, -3.5197e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0312, -0.0282, 0.0158, 0.0298, -0.0284, 0.0080, -0.0128, 0.0100, + 0.0181, 0.0010], device='cuda:0'), grad: tensor([-2.5839e-05, 4.2878e-06, 2.0993e-04, 3.4952e-04, 1.5572e-05, + 1.7619e-04, -2.0638e-05, 3.1233e-05, -8.0299e-04, 6.2406e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 214.65, cls_loss 0.0088 cls_loss_mapping 0.0218 cls_loss_causal 0.6801 re_mapping 0.0115 re_causal 0.0362 /// teacc 98.82 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0110, 0.0200, -0.0266, ..., -0.0467, -0.0349, 0.0461], + [-0.0043, -0.0094, -0.0224, ..., 0.0855, -0.0377, -0.0244], + [-0.0086, -0.0256, 0.1048, ..., -0.0551, -0.0756, -0.0395], + ..., + [-0.0134, -0.0428, -0.0145, ..., 0.0133, 0.0458, -0.0075], + [-0.0394, -0.0396, -0.0072, ..., -0.0559, 0.0217, -0.0606], + [-0.0826, -0.0286, -0.0673, ..., -0.0592, 0.0268, 0.0375]], + device='cuda:0'), grad: tensor([[ 4.4815e-06, 0.0000e+00, -7.6294e-06, ..., 1.2361e-05, + 2.6286e-05, -1.9550e-05], + [ 1.0878e-05, 0.0000e+00, -4.2272e-04, ..., -1.1101e-03, + 7.2098e-04, 2.9653e-05], + [ 2.8964e-06, 0.0000e+00, 1.0338e-03, ..., 1.9569e-03, + 1.4839e-03, 2.0608e-05], + ..., + [ 2.4468e-05, 0.0000e+00, -1.0281e-03, ..., -1.3189e-03, + -3.6850e-03, -1.2897e-05], + [ 2.1696e-05, 0.0000e+00, 7.1347e-05, ..., 1.6487e-04, + 4.7135e-04, 3.5644e-05], + [-5.0038e-05, 0.0000e+00, 1.0663e-04, ..., 2.9370e-05, + 4.1866e-04, -6.9571e-04]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0315, -0.0285, 0.0159, 0.0302, -0.0278, 0.0078, -0.0126, 0.0095, + 0.0179, 0.0011], device='cuda:0'), grad: tensor([ 6.2346e-05, 1.5664e-04, 6.3400e-03, 1.0605e-03, 1.1158e-03, + 3.3331e-04, 1.0043e-04, -1.0155e-02, 1.2093e-03, -2.2638e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 214.59, cls_loss 0.0066 cls_loss_mapping 0.0176 cls_loss_causal 0.6638 re_mapping 0.0117 re_causal 0.0374 /// teacc 98.80 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0114, 0.0199, -0.0259, ..., -0.0468, -0.0351, 0.0466], + [-0.0057, -0.0095, -0.0226, ..., 0.0866, -0.0382, -0.0241], + [-0.0080, -0.0257, 0.1052, ..., -0.0554, -0.0765, -0.0400], + ..., + [-0.0141, -0.0427, -0.0136, ..., 0.0128, 0.0469, -0.0077], + [-0.0399, -0.0397, -0.0074, ..., -0.0565, 0.0217, -0.0611], + [-0.0832, -0.0286, -0.0688, ..., -0.0596, 0.0265, 0.0381]], + device='cuda:0'), grad: tensor([[ 4.1835e-06, 1.2049e-07, 4.4368e-06, ..., 8.0392e-06, + 2.7232e-06, 9.6671e-07], + [ 1.8194e-05, 2.6054e-07, 1.0008e-04, ..., 3.3677e-05, + 3.6657e-05, 5.6364e-06], + [ 1.0461e-04, 4.7730e-07, -2.5225e-04, ..., -5.3272e-06, + 7.7710e-06, 3.8855e-06], + ..., + [ 1.1310e-05, 9.4250e-07, -1.1154e-05, ..., -1.2264e-05, + -5.0038e-05, 1.9833e-05], + [ 4.1015e-06, 1.2910e-07, 2.6271e-05, ..., 1.2666e-05, + 3.2745e-06, 2.7139e-06], + [ 1.7494e-05, 1.0440e-06, 7.3686e-06, ..., 1.0826e-05, + 1.8060e-05, 1.7166e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0315, -0.0280, 0.0154, 0.0298, -0.0276, 0.0079, -0.0130, 0.0100, + 0.0178, 0.0011], device='cuda:0'), grad: tensor([ 5.4866e-05, 2.5678e-04, -2.0111e-04, -1.0377e-04, -2.3341e-04, + 1.9535e-05, 9.9242e-05, 2.2650e-06, -3.8654e-05, 1.4412e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 214.82, cls_loss 0.0082 cls_loss_mapping 0.0246 cls_loss_causal 0.6794 re_mapping 0.0113 re_causal 0.0365 /// teacc 98.90 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0119, 0.0202, -0.0251, ..., -0.0450, -0.0353, 0.0462], + [-0.0064, -0.0096, -0.0239, ..., 0.0866, -0.0388, -0.0247], + [-0.0084, -0.0257, 0.1061, ..., -0.0560, -0.0768, -0.0403], + ..., + [-0.0147, -0.0429, -0.0133, ..., 0.0124, 0.0474, -0.0080], + [-0.0404, -0.0404, -0.0081, ..., -0.0577, 0.0212, -0.0621], + [-0.0839, -0.0280, -0.0683, ..., -0.0602, 0.0263, 0.0393]], + device='cuda:0'), grad: tensor([[ 3.6621e-03, 0.0000e+00, 2.7835e-05, ..., 3.7766e-03, + 6.5640e-06, 4.4022e-03], + [ 1.1957e-04, 0.0000e+00, 1.6376e-05, ..., -1.4772e-03, + -2.0278e-04, 7.6652e-05], + [ 6.4850e-05, 0.0000e+00, -2.4986e-04, ..., 5.8085e-05, + 5.1185e-06, 6.3598e-05], + ..., + [ 7.7128e-05, 0.0000e+00, -1.1818e-06, ..., 1.5717e-03, + 1.8334e-04, 1.1683e-04], + [ 1.4436e-04, 0.0000e+00, 6.6817e-05, ..., 1.6487e-04, + 2.1055e-05, 1.9038e-04], + [ 9.1028e-04, 0.0000e+00, 3.5584e-05, ..., 4.1747e-04, + -1.6481e-05, 2.6011e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0315, -0.0284, 0.0153, 0.0299, -0.0273, 0.0078, -0.0132, 0.0100, + 0.0179, 0.0014], device='cuda:0'), grad: tensor([ 0.0094, -0.0016, -0.0002, 0.0001, -0.0016, 0.0016, -0.0118, 0.0020, + 0.0005, 0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 214.51, cls_loss 0.0059 cls_loss_mapping 0.0180 cls_loss_causal 0.6809 re_mapping 0.0110 re_causal 0.0370 /// teacc 98.93 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0131, 0.0209, -0.0242, ..., -0.0460, -0.0354, 0.0464], + [-0.0069, -0.0096, -0.0242, ..., 0.0866, -0.0393, -0.0253], + [-0.0086, -0.0256, 0.1069, ..., -0.0565, -0.0771, -0.0407], + ..., + [-0.0153, -0.0429, -0.0130, ..., 0.0125, 0.0479, -0.0084], + [-0.0404, -0.0405, -0.0086, ..., -0.0578, 0.0210, -0.0627], + [-0.0849, -0.0280, -0.0693, ..., -0.0607, 0.0263, 0.0400]], + device='cuda:0'), grad: tensor([[-2.4378e-05, 0.0000e+00, -1.0151e-06, ..., -2.6390e-05, + 1.9986e-06, -1.3936e-04], + [ 7.4357e-06, 0.0000e+00, 9.5740e-06, ..., -2.0981e-04, + 8.2701e-06, 2.1443e-05], + [ 7.3649e-06, 0.0000e+00, -5.1641e-04, ..., 5.3287e-05, + 3.8017e-06, 1.6049e-05], + ..., + [ 6.3740e-06, 0.0000e+00, -9.9689e-06, ..., 2.6882e-05, + -2.2888e-05, 4.8354e-06], + [-1.3657e-05, 0.0000e+00, 1.7023e-04, ..., -1.8328e-05, + 5.2080e-06, -4.3213e-05], + [ 2.1338e-04, 0.0000e+00, 8.9109e-06, ..., 2.9519e-05, + 1.1668e-05, 1.8165e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0317, -0.0290, 0.0150, 0.0293, -0.0270, 0.0076, -0.0122, 0.0102, + 0.0185, 0.0013], device='cuda:0'), grad: tensor([-1.4329e-04, 3.8415e-05, -8.4591e-04, 9.4652e-04, -6.2920e-06, + -1.1992e-04, 1.1897e-04, 2.4706e-05, -3.4070e-04, 3.2735e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 58---------------------------------------------------- +epoch 58, time 230.85, cls_loss 0.0047 cls_loss_mapping 0.0146 cls_loss_causal 0.6211 re_mapping 0.0109 re_causal 0.0351 /// teacc 99.03 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0135, 0.0209, -0.0241, ..., -0.0463, -0.0356, 0.0465], + [-0.0080, -0.0096, -0.0248, ..., 0.0874, -0.0396, -0.0259], + [-0.0080, -0.0256, 0.1079, ..., -0.0568, -0.0773, -0.0410], + ..., + [-0.0156, -0.0429, -0.0128, ..., 0.0122, 0.0482, -0.0079], + [-0.0407, -0.0406, -0.0087, ..., -0.0586, 0.0209, -0.0634], + [-0.0854, -0.0280, -0.0707, ..., -0.0611, 0.0260, 0.0404]], + device='cuda:0'), grad: tensor([[ 6.7532e-05, 2.7881e-08, -2.6003e-06, ..., 7.9215e-05, + 6.9253e-06, 1.0848e-05], + [ 5.3108e-05, 2.4389e-08, 2.5220e-06, ..., 4.9859e-05, + 3.0585e-06, 2.7530e-06], + [ 4.1455e-05, 4.7963e-08, -3.0100e-06, ..., 5.8949e-05, + 4.9062e-06, 9.3356e-06], + ..., + [ 1.6555e-05, 2.6496e-07, -4.8466e-06, ..., 1.0014e-05, + -1.0736e-05, 5.5172e-06], + [ 3.0756e-05, 7.7940e-08, -2.4498e-05, ..., 3.7283e-05, + 1.0759e-05, 1.1541e-05], + [ 4.7237e-05, -3.8929e-07, 1.8165e-05, ..., 4.1872e-05, + 3.7476e-06, 1.3813e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0319, -0.0290, 0.0153, 0.0293, -0.0268, 0.0076, -0.0122, 0.0104, + 0.0183, 0.0009], device='cuda:0'), grad: tensor([ 2.3842e-04, 1.5199e-04, 1.7333e-04, -3.0160e-04, 8.4352e-04, + 2.6894e-04, -1.6394e-03, 2.4572e-05, 7.3135e-05, 1.6844e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 214.86, cls_loss 0.0066 cls_loss_mapping 0.0189 cls_loss_causal 0.6634 re_mapping 0.0108 re_causal 0.0360 /// teacc 98.99 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0139, 0.0204, -0.0237, ..., -0.0465, -0.0358, 0.0469], + [-0.0083, -0.0098, -0.0254, ..., 0.0881, -0.0400, -0.0261], + [-0.0080, -0.0251, 0.1090, ..., -0.0572, -0.0777, -0.0417], + ..., + [-0.0160, -0.0436, -0.0129, ..., 0.0118, 0.0487, -0.0084], + [-0.0409, -0.0422, -0.0095, ..., -0.0593, 0.0205, -0.0639], + [-0.0859, -0.0284, -0.0713, ..., -0.0614, 0.0259, 0.0408]], + device='cuda:0'), grad: tensor([[ 1.3486e-05, 0.0000e+00, -1.4221e-06, ..., 3.6657e-05, + 1.8686e-05, 5.2124e-05], + [ 5.5246e-06, 0.0000e+00, 7.6890e-06, ..., -5.0068e-05, + 8.5384e-06, 5.9120e-06], + [ 3.4925e-06, 0.0000e+00, 4.0114e-05, ..., 1.5482e-05, + 2.3678e-05, 3.2298e-06], + ..., + [ 3.6210e-06, 0.0000e+00, -5.0694e-05, ..., 2.1875e-05, + -4.5568e-05, 1.6138e-05], + [ 6.9998e-06, 0.0000e+00, -2.6777e-05, ..., 1.0371e-05, + -1.5423e-05, 7.5996e-05], + [ 5.7556e-06, 0.0000e+00, 5.9046e-06, ..., 1.0870e-05, + 1.3202e-05, -8.1062e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0317, -0.0289, 0.0152, 0.0302, -0.0272, 0.0071, -0.0122, 0.0102, + 0.0183, 0.0009], device='cuda:0'), grad: tensor([ 1.1271e-04, -3.9428e-05, 1.2541e-04, 1.0014e-04, 6.0230e-05, + 1.4305e-03, -1.6127e-03, -1.3404e-05, -7.3195e-05, -9.0659e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 214.61, cls_loss 0.0050 cls_loss_mapping 0.0144 cls_loss_causal 0.6244 re_mapping 0.0106 re_causal 0.0338 /// teacc 98.76 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0143, 0.0202, -0.0227, ..., -0.0469, -0.0359, 0.0470], + [-0.0085, -0.0099, -0.0258, ..., 0.0890, -0.0404, -0.0263], + [-0.0084, -0.0251, 0.1103, ..., -0.0578, -0.0778, -0.0422], + ..., + [-0.0167, -0.0443, -0.0135, ..., 0.0112, 0.0491, -0.0093], + [-0.0411, -0.0429, -0.0104, ..., -0.0596, 0.0205, -0.0643], + [-0.0866, -0.0287, -0.0714, ..., -0.0620, 0.0257, 0.0417]], + device='cuda:0'), grad: tensor([[ 5.6475e-05, 4.0163e-08, 7.0751e-05, ..., 4.1462e-06, + 1.9111e-06, 2.8586e-04], + [ 3.2745e-06, 7.8406e-08, 9.1866e-06, ..., -2.2903e-05, + 5.1372e-06, 3.0369e-05], + [ 7.8008e-06, 1.2270e-07, 5.8450e-06, ..., 8.5309e-06, + 2.0079e-06, 3.8624e-05], + ..., + [ 6.1207e-06, -7.0175e-07, 2.7046e-06, ..., 3.5018e-07, + -1.4126e-05, 4.1336e-05], + [ 1.0028e-05, 3.5448e-08, 1.3776e-05, ..., 1.4685e-05, + 6.4112e-06, 5.6684e-05], + [-1.4687e-04, -2.7078e-07, -2.0504e-04, ..., 1.1371e-06, + 1.9930e-07, -7.6818e-04]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0318, -0.0287, 0.0154, 0.0306, -0.0269, 0.0070, -0.0127, 0.0096, + 0.0184, 0.0010], device='cuda:0'), grad: tensor([ 6.0797e-04, 5.6088e-05, 9.4593e-05, -8.8263e-04, 6.2656e-04, + 8.4925e-04, 2.6301e-05, 6.0022e-05, 1.0520e-04, -1.5430e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 214.24, cls_loss 0.0049 cls_loss_mapping 0.0152 cls_loss_causal 0.6451 re_mapping 0.0104 re_causal 0.0338 /// teacc 98.85 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0146, 0.0205, -0.0225, ..., -0.0471, -0.0361, 0.0469], + [-0.0089, -0.0099, -0.0266, ..., 0.0894, -0.0413, -0.0271], + [-0.0086, -0.0251, 0.1115, ..., -0.0579, -0.0785, -0.0424], + ..., + [-0.0173, -0.0442, -0.0135, ..., 0.0108, 0.0498, -0.0092], + [-0.0416, -0.0430, -0.0111, ..., -0.0601, 0.0205, -0.0647], + [-0.0873, -0.0287, -0.0722, ..., -0.0624, 0.0254, 0.0420]], + device='cuda:0'), grad: tensor([[ 8.2701e-06, 0.0000e+00, -3.6638e-06, ..., 5.8636e-06, + 3.2820e-06, -1.3001e-05], + [ 5.9903e-06, 0.0000e+00, 2.6505e-06, ..., -3.2037e-05, + 9.8646e-06, 2.2519e-06], + [ 1.1846e-05, 0.0000e+00, 1.0207e-06, ..., 2.1920e-05, + 4.5188e-06, 2.0470e-06], + ..., + [ 1.1742e-05, 0.0000e+00, -1.2621e-05, ..., 1.3649e-05, + -3.0935e-05, 1.1347e-05], + [ 1.5825e-05, 0.0000e+00, 8.3633e-07, ..., 1.1697e-05, + 1.5721e-05, 1.9714e-05], + [ 1.0461e-05, 0.0000e+00, 4.1015e-06, ..., 6.8322e-06, + 1.3135e-05, -4.8608e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0322, -0.0291, 0.0157, 0.0305, -0.0267, 0.0073, -0.0124, 0.0098, + 0.0185, 0.0006], device='cuda:0'), grad: tensor([ 6.0871e-06, -8.7619e-06, 4.2826e-05, 1.5691e-05, -2.8789e-05, + -9.7275e-05, 5.5134e-05, -2.1577e-05, 7.1466e-05, -3.4750e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.53, cls_loss 0.0037 cls_loss_mapping 0.0127 cls_loss_causal 0.6240 re_mapping 0.0104 re_causal 0.0334 /// teacc 98.98 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0146, 0.0205, -0.0222, ..., -0.0472, -0.0364, 0.0471], + [-0.0091, -0.0099, -0.0269, ..., 0.0902, -0.0417, -0.0275], + [-0.0088, -0.0250, 0.1121, ..., -0.0582, -0.0789, -0.0427], + ..., + [-0.0177, -0.0442, -0.0131, ..., 0.0107, 0.0504, -0.0094], + [-0.0419, -0.0430, -0.0113, ..., -0.0605, 0.0204, -0.0651], + [-0.0876, -0.0286, -0.0723, ..., -0.0629, 0.0252, 0.0427]], + device='cuda:0'), grad: tensor([[ 3.1255e-06, 0.0000e+00, 4.5705e-07, ..., 3.8370e-06, + 1.8291e-06, 4.6752e-06], + [ 4.5747e-06, 0.0000e+00, 5.8115e-06, ..., -2.8368e-06, + 1.5702e-06, 3.0119e-06], + [ 2.8536e-06, 0.0000e+00, -4.5538e-05, ..., 1.4221e-06, + 1.8179e-06, 3.0492e-06], + ..., + [ 3.8855e-06, 0.0000e+00, -9.7696e-07, ..., 1.4594e-06, + -2.0154e-06, 3.8028e-05], + [ 1.9103e-05, 0.0000e+00, 7.1675e-06, ..., 2.5630e-06, + -8.0541e-06, 5.6624e-05], + [-1.2219e-05, 0.0000e+00, 3.8669e-06, ..., 3.5837e-06, + 3.7421e-06, -1.6189e-04]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0322, -0.0289, 0.0154, 0.0305, -0.0270, 0.0072, -0.0127, 0.0101, + 0.0186, 0.0007], device='cuda:0'), grad: tensor([ 1.8656e-05, 1.5825e-05, -4.9323e-05, 7.2122e-05, 3.6180e-05, + 9.4175e-06, 8.3745e-06, 6.8903e-05, 1.2934e-04, -3.0947e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 214.32, cls_loss 0.0043 cls_loss_mapping 0.0131 cls_loss_causal 0.6234 re_mapping 0.0099 re_causal 0.0322 /// teacc 98.85 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0150, 0.0205, -0.0216, ..., -0.0472, -0.0366, 0.0473], + [-0.0094, -0.0099, -0.0278, ..., 0.0909, -0.0415, -0.0281], + [-0.0091, -0.0250, 0.1124, ..., -0.0583, -0.0794, -0.0431], + ..., + [-0.0183, -0.0442, -0.0119, ..., 0.0100, 0.0507, -0.0097], + [-0.0424, -0.0430, -0.0120, ..., -0.0610, 0.0203, -0.0655], + [-0.0884, -0.0286, -0.0730, ..., -0.0634, 0.0250, 0.0432]], + device='cuda:0'), grad: tensor([[ 2.9057e-06, 2.4331e-08, -1.6004e-05, ..., 2.1234e-05, + 1.1496e-05, -1.5840e-05], + [ 1.3612e-05, 5.4133e-09, -1.5569e-04, ..., -1.6193e-03, + -7.6723e-04, -5.1498e-04], + [ 8.6427e-05, 8.9640e-09, -4.6253e-05, ..., 6.3717e-05, + 4.9978e-05, 1.7539e-05], + ..., + [ 7.9811e-05, 2.6776e-08, 1.2076e-04, ..., 1.2980e-03, + 5.9175e-04, 4.2582e-04], + [ 4.6492e-06, 5.5297e-08, 1.4260e-05, ..., 1.3284e-05, + 2.9191e-05, 4.5002e-05], + [ 1.3031e-05, -3.9348e-07, 1.6898e-05, ..., 6.6936e-05, + 3.6657e-05, -1.0848e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0321, -0.0287, 0.0152, 0.0303, -0.0267, 0.0072, -0.0126, 0.0102, + 0.0186, 0.0005], device='cuda:0'), grad: tensor([ 3.4660e-05, -4.9629e-03, 1.5831e-04, 7.7724e-05, 4.6682e-04, + -3.6985e-05, 6.4492e-05, 4.0054e-03, 1.3149e-04, 5.8115e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 214.40, cls_loss 0.0047 cls_loss_mapping 0.0131 cls_loss_causal 0.6064 re_mapping 0.0101 re_causal 0.0312 /// teacc 98.99 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0153, 0.0204, -0.0213, ..., -0.0473, -0.0369, 0.0476], + [-0.0106, -0.0100, -0.0283, ..., 0.0918, -0.0417, -0.0288], + [-0.0084, -0.0249, 0.1130, ..., -0.0594, -0.0800, -0.0433], + ..., + [-0.0187, -0.0443, -0.0111, ..., 0.0103, 0.0515, -0.0102], + [-0.0428, -0.0432, -0.0126, ..., -0.0617, 0.0200, -0.0662], + [-0.0891, -0.0287, -0.0737, ..., -0.0634, 0.0248, 0.0442]], + device='cuda:0'), grad: tensor([[ 6.7195e-07, 0.0000e+00, 1.4894e-05, ..., 9.1791e-06, + 9.3784e-07, 1.0341e-05], + [ 3.1292e-06, 0.0000e+00, 7.9423e-06, ..., -8.0466e-05, + 1.6475e-06, -1.1481e-05], + [ 8.9221e-07, 0.0000e+00, -1.1903e-04, ..., 3.2876e-06, + 1.2163e-06, -8.1584e-06], + ..., + [ 3.6545e-06, 0.0000e+00, 1.9908e-05, ..., 1.3210e-05, + -3.5726e-06, 8.8960e-06], + [ 1.4007e-06, 0.0000e+00, -1.2547e-05, ..., 9.3728e-06, + 2.5854e-06, -4.3288e-06], + [ 2.1785e-05, 0.0000e+00, 2.2739e-05, ..., 3.7313e-05, + 5.5991e-06, -2.7701e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0321, -0.0287, 0.0147, 0.0303, -0.0269, 0.0072, -0.0127, 0.0109, + 0.0182, 0.0008], device='cuda:0'), grad: tensor([ 6.9082e-05, -7.7248e-05, -1.6582e-04, 4.4942e-05, 1.9342e-05, + 5.8502e-05, 4.3847e-06, 5.9158e-05, -1.8322e-04, 1.7071e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 214.57, cls_loss 0.0048 cls_loss_mapping 0.0134 cls_loss_causal 0.6791 re_mapping 0.0093 re_causal 0.0315 /// teacc 99.00 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0154, 0.0204, -0.0207, ..., -0.0472, -0.0371, 0.0478], + [-0.0109, -0.0100, -0.0285, ..., 0.0923, -0.0424, -0.0291], + [-0.0087, -0.0250, 0.1145, ..., -0.0601, -0.0802, -0.0435], + ..., + [-0.0193, -0.0437, -0.0114, ..., 0.0103, 0.0522, -0.0104], + [-0.0433, -0.0432, -0.0132, ..., -0.0620, 0.0198, -0.0667], + [-0.0905, -0.0287, -0.0743, ..., -0.0640, 0.0244, 0.0445]], + device='cuda:0'), grad: tensor([[ 2.0899e-06, 0.0000e+00, -3.1680e-05, ..., -7.0408e-06, + 5.7071e-06, -1.8334e-04], + [ 9.0757e-07, 0.0000e+00, 3.9846e-05, ..., -2.1088e-04, + 2.2635e-05, 1.3143e-05], + [ 1.3104e-06, 0.0000e+00, -1.0264e-04, ..., 7.3671e-05, + 9.0674e-06, 2.5049e-05], + ..., + [ 3.2107e-07, 0.0000e+00, -7.3075e-05, ..., 5.9932e-05, + -8.5771e-05, -2.7254e-05], + [ 1.4380e-06, 0.0000e+00, 2.5287e-05, ..., 1.4715e-05, + 4.6380e-06, 3.0026e-05], + [ 2.1774e-06, 0.0000e+00, 3.6836e-05, ..., 7.3425e-06, + 3.8207e-05, 3.6478e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0318, -0.0292, 0.0150, 0.0307, -0.0262, 0.0065, -0.0125, 0.0112, + 0.0182, 0.0001], device='cuda:0'), grad: tensor([-3.3212e-04, -1.9431e-04, 1.8954e-05, -1.5044e-04, 6.8188e-05, + 2.3508e-04, 2.5558e-04, -1.8775e-04, 1.0860e-04, 1.7822e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.39, cls_loss 0.0053 cls_loss_mapping 0.0157 cls_loss_causal 0.6377 re_mapping 0.0098 re_causal 0.0313 /// teacc 98.89 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0161, 0.0204, -0.0198, ..., -0.0474, -0.0374, 0.0481], + [-0.0114, -0.0100, -0.0292, ..., 0.0940, -0.0427, -0.0292], + [-0.0090, -0.0249, 0.1149, ..., -0.0605, -0.0806, -0.0440], + ..., + [-0.0198, -0.0437, -0.0106, ..., 0.0084, 0.0528, -0.0109], + [-0.0438, -0.0433, -0.0132, ..., -0.0626, 0.0195, -0.0675], + [-0.0904, -0.0287, -0.0754, ..., -0.0644, 0.0243, 0.0453]], + device='cuda:0'), grad: tensor([[ 2.7195e-07, 0.0000e+00, -5.2422e-05, ..., 1.1645e-05, + 1.1483e-06, -6.8545e-05], + [ 4.5309e-07, 0.0000e+00, 9.9912e-06, ..., -1.7464e-05, + 3.1311e-06, 3.6359e-06], + [ 7.4564e-08, 0.0000e+00, -4.6343e-05, ..., 1.7527e-06, + 2.0545e-06, 6.8992e-06], + ..., + [ 5.0664e-07, 0.0000e+00, 6.5751e-07, ..., 9.5926e-07, + -9.0301e-06, 1.0669e-05], + [ 3.7951e-07, 0.0000e+00, 2.1830e-05, ..., 1.0803e-05, + 1.3057e-06, 2.4036e-05], + [-1.5339e-06, 0.0000e+00, 1.1604e-06, ..., 1.3439e-06, + 2.7777e-07, -2.6390e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0318, -0.0280, 0.0145, 0.0303, -0.0270, 0.0067, -0.0121, 0.0107, + 0.0185, 0.0003], device='cuda:0'), grad: tensor([-1.0133e-04, 1.0759e-05, -3.1590e-05, -1.5318e-05, 2.6852e-05, + 3.7491e-05, 9.6932e-06, 1.0364e-05, 8.6010e-05, -3.2693e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.56, cls_loss 0.0044 cls_loss_mapping 0.0114 cls_loss_causal 0.6227 re_mapping 0.0096 re_causal 0.0311 /// teacc 98.98 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0165, 0.0204, -0.0190, ..., -0.0474, -0.0376, 0.0485], + [-0.0119, -0.0100, -0.0295, ..., 0.0945, -0.0433, -0.0294], + [-0.0092, -0.0249, 0.1159, ..., -0.0611, -0.0812, -0.0442], + ..., + [-0.0200, -0.0437, -0.0100, ..., 0.0081, 0.0534, -0.0112], + [-0.0438, -0.0433, -0.0141, ..., -0.0630, 0.0192, -0.0681], + [-0.0910, -0.0287, -0.0760, ..., -0.0651, 0.0241, 0.0457]], + device='cuda:0'), grad: tensor([[ 5.3018e-05, 0.0000e+00, 1.2219e-05, ..., 7.4983e-05, + 1.0826e-05, -1.3402e-06], + [-1.6868e-04, 0.0000e+00, -4.3325e-06, ..., -4.8971e-04, + 2.6271e-05, 1.7313e-06], + [ 1.3685e-04, 0.0000e+00, 2.7969e-05, ..., 3.4595e-04, + 5.6505e-05, 2.3227e-06], + ..., + [ 1.8060e-05, 0.0000e+00, -1.8609e-04, ..., 4.5419e-05, + -2.0313e-04, 6.9961e-06], + [ 7.1712e-06, 0.0000e+00, 2.3767e-05, ..., 1.4156e-05, + 1.8716e-05, 6.6496e-06], + [ 6.9678e-05, 0.0000e+00, 5.2124e-05, ..., 1.3083e-05, + 5.1469e-05, -1.4715e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-3.1470e-02, -2.8339e-02, 1.4176e-02, 3.0584e-02, -2.6885e-02, + 6.6535e-03, -1.2476e-02, 1.1432e-02, 1.8322e-02, -4.2948e-05], + device='cuda:0'), grad: tensor([ 1.5688e-04, -5.0354e-04, 4.8923e-04, 6.9082e-05, 2.5064e-05, + 5.0575e-05, -2.3139e-04, -4.0603e-04, 6.8963e-05, 2.8110e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 214.67, cls_loss 0.0041 cls_loss_mapping 0.0124 cls_loss_causal 0.6462 re_mapping 0.0094 re_causal 0.0303 /// teacc 98.94 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0172, 0.0203, -0.0181, ..., -0.0481, -0.0378, 0.0487], + [-0.0124, -0.0100, -0.0304, ..., 0.0951, -0.0437, -0.0296], + [-0.0096, -0.0248, 0.1169, ..., -0.0615, -0.0815, -0.0445], + ..., + [-0.0203, -0.0437, -0.0107, ..., 0.0077, 0.0537, -0.0118], + [-0.0441, -0.0433, -0.0145, ..., -0.0637, 0.0191, -0.0685], + [-0.0924, -0.0288, -0.0753, ..., -0.0653, 0.0241, 0.0463]], + device='cuda:0'), grad: tensor([[ 1.1288e-05, 0.0000e+00, -2.1216e-06, ..., 4.3362e-06, + 2.6096e-06, -1.0550e-05], + [ 2.5690e-05, 0.0000e+00, 2.3544e-04, ..., 3.1978e-05, + 4.8965e-05, 1.9316e-06], + [ 2.3395e-05, 0.0000e+00, -6.0606e-04, ..., -5.2422e-05, + 8.3596e-06, 1.0081e-05], + ..., + [ 1.3089e-04, 0.0000e+00, 1.3900e-04, ..., 2.0489e-05, + -1.1533e-04, 2.1830e-05], + [ 4.8816e-05, 0.0000e+00, 7.3016e-05, ..., 2.5302e-05, + 2.6852e-05, 8.0690e-06], + [ 5.1737e-04, 0.0000e+00, 1.3344e-05, ..., 2.9325e-05, + 1.7896e-05, -7.4506e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0315, -0.0284, 0.0142, 0.0305, -0.0265, 0.0068, -0.0123, 0.0107, + 0.0181, 0.0006], device='cuda:0'), grad: tensor([ 1.4484e-05, 4.6372e-04, -6.8855e-04, 2.0742e-04, -1.1730e-03, + -5.0455e-05, 6.0260e-05, 1.3292e-04, 2.5463e-04, 7.8106e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 214.61, cls_loss 0.0041 cls_loss_mapping 0.0103 cls_loss_causal 0.6180 re_mapping 0.0091 re_causal 0.0296 /// teacc 98.97 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0178, 0.0203, -0.0178, ..., -0.0484, -0.0381, 0.0487], + [-0.0127, -0.0100, -0.0312, ..., 0.0956, -0.0442, -0.0301], + [-0.0098, -0.0248, 0.1180, ..., -0.0618, -0.0818, -0.0449], + ..., + [-0.0206, -0.0437, -0.0106, ..., 0.0075, 0.0542, -0.0122], + [-0.0443, -0.0433, -0.0148, ..., -0.0643, 0.0187, -0.0691], + [-0.0921, -0.0288, -0.0757, ..., -0.0658, 0.0240, 0.0470]], + device='cuda:0'), grad: tensor([[ 8.4862e-06, 0.0000e+00, 1.5851e-06, ..., 1.1258e-05, + 3.0212e-06, 5.8226e-06], + [ 1.4849e-05, 0.0000e+00, 9.2462e-06, ..., -8.2282e-07, + 2.9709e-06, 1.7872e-06], + [ 8.2627e-06, 0.0000e+00, -2.0474e-05, ..., 1.0632e-05, + 2.7120e-06, 6.1505e-06], + ..., + [ 8.9929e-06, 0.0000e+00, 6.6310e-06, ..., 7.0781e-06, + -2.1048e-06, 1.9953e-05], + [ 1.9506e-05, 0.0000e+00, -8.6986e-07, ..., 3.1352e-05, + 9.9614e-06, 2.0966e-05], + [ 1.1250e-05, 0.0000e+00, -1.4096e-05, ..., 9.4548e-06, + -5.3234e-06, -8.3804e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0320, -0.0285, 0.0141, 0.0306, -0.0273, 0.0071, -0.0126, 0.0107, + 0.0186, 0.0009], device='cuda:0'), grad: tensor([ 5.6416e-05, 3.5793e-05, 3.9101e-05, -1.4096e-05, 1.0431e-04, + 8.9884e-05, -2.3711e-04, 6.2466e-05, -7.3433e-05, -6.3300e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 70---------------------------------------------------- +epoch 70, time 232.14, cls_loss 0.0035 cls_loss_mapping 0.0113 cls_loss_causal 0.5866 re_mapping 0.0098 re_causal 0.0303 /// teacc 99.08 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0184, 0.0203, -0.0167, ..., -0.0486, -0.0383, 0.0490], + [-0.0130, -0.0100, -0.0327, ..., 0.0959, -0.0449, -0.0303], + [-0.0106, -0.0248, 0.1187, ..., -0.0621, -0.0825, -0.0452], + ..., + [-0.0213, -0.0437, -0.0099, ..., 0.0073, 0.0550, -0.0124], + [-0.0437, -0.0434, -0.0153, ..., -0.0649, 0.0184, -0.0696], + [-0.0929, -0.0288, -0.0766, ..., -0.0660, 0.0237, 0.0474]], + device='cuda:0'), grad: tensor([[ 1.3746e-05, 0.0000e+00, -6.1840e-06, ..., 3.6154e-06, + 2.8200e-06, -1.6326e-06], + [ 1.5236e-05, 0.0000e+00, 1.2219e-05, ..., -5.5954e-06, + 2.3723e-05, 8.3372e-06], + [ 2.3767e-05, 0.0000e+00, -2.9817e-05, ..., 9.8944e-06, + 6.2101e-06, 9.8348e-06], + ..., + [ 1.0982e-05, 0.0000e+00, -7.3075e-05, ..., 5.5581e-06, + -2.6917e-04, -4.0829e-05], + [ 9.2685e-06, 0.0000e+00, 7.8380e-06, ..., 3.2689e-06, + 7.6815e-06, 2.2739e-05], + [ 2.3723e-05, 0.0000e+00, 5.3167e-05, ..., 3.5856e-06, + 2.0826e-04, 1.0455e-04]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0319, -0.0289, 0.0138, 0.0309, -0.0270, 0.0071, -0.0128, 0.0111, + 0.0185, 0.0006], device='cuda:0'), grad: tensor([ 2.6956e-05, 1.1027e-04, 8.2970e-05, -3.2616e-04, 9.9614e-06, + 9.7528e-06, 3.5763e-05, -7.3147e-04, 8.3804e-05, 6.9809e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 214.36, cls_loss 0.0034 cls_loss_mapping 0.0097 cls_loss_causal 0.6076 re_mapping 0.0091 re_causal 0.0294 /// teacc 98.93 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0188, 0.0203, -0.0167, ..., -0.0489, -0.0386, 0.0492], + [-0.0134, -0.0100, -0.0335, ..., 0.0969, -0.0452, -0.0300], + [-0.0108, -0.0247, 0.1202, ..., -0.0628, -0.0829, -0.0455], + ..., + [-0.0228, -0.0438, -0.0100, ..., 0.0070, 0.0554, -0.0128], + [-0.0442, -0.0434, -0.0160, ..., -0.0653, 0.0182, -0.0700], + [-0.0936, -0.0288, -0.0773, ..., -0.0663, 0.0235, 0.0479]], + device='cuda:0'), grad: tensor([[ 1.5080e-05, 0.0000e+00, -1.3220e-04, ..., 1.1325e-05, + 2.6673e-06, -1.1724e-04], + [ 3.0339e-05, 0.0000e+00, 4.9323e-06, ..., 1.2465e-05, + 5.0403e-06, 4.8354e-06], + [ 1.4074e-05, 0.0000e+00, 8.0168e-06, ..., 7.4767e-06, + 3.1628e-06, 8.1241e-05], + ..., + [ 1.5616e-05, 0.0000e+00, 9.1493e-06, ..., 8.2776e-06, + 2.6338e-06, 1.1586e-05], + [ 1.5169e-05, 0.0000e+00, 6.4731e-05, ..., 9.3579e-06, + 4.2804e-06, 1.9819e-06], + [ 1.4678e-05, 0.0000e+00, 1.0177e-05, ..., 6.7241e-06, + 2.6673e-06, -6.7502e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0320, -0.0284, 0.0140, 0.0307, -0.0264, 0.0072, -0.0129, 0.0107, + 0.0184, 0.0005], device='cuda:0'), grad: tensor([-1.6367e-04, 5.9366e-05, 9.8884e-05, -6.5386e-05, -3.1567e-04, + 8.9943e-05, 2.0266e-04, 5.4508e-05, 8.6725e-06, 3.0696e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.63, cls_loss 0.0035 cls_loss_mapping 0.0098 cls_loss_causal 0.6221 re_mapping 0.0090 re_causal 0.0301 /// teacc 98.81 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0196, 0.0203, -0.0165, ..., -0.0493, -0.0389, 0.0486], + [-0.0141, -0.0100, -0.0342, ..., 0.0970, -0.0455, -0.0302], + [-0.0105, -0.0247, 0.1217, ..., -0.0627, -0.0832, -0.0456], + ..., + [-0.0238, -0.0438, -0.0103, ..., 0.0069, 0.0555, -0.0133], + [-0.0444, -0.0434, -0.0168, ..., -0.0656, 0.0181, -0.0704], + [-0.0939, -0.0288, -0.0778, ..., -0.0664, 0.0233, 0.0489]], + device='cuda:0'), grad: tensor([[ 4.7381e-08, 0.0000e+00, 1.9651e-06, ..., 2.2501e-06, + 8.7032e-07, -5.4948e-06], + [ 2.4680e-07, 0.0000e+00, 4.7833e-06, ..., -6.0797e-06, + 1.8999e-06, 3.3099e-06], + [ 2.2561e-07, 0.0000e+00, -1.6296e-04, ..., 3.1106e-06, + 7.9069e-07, 3.6340e-06], + ..., + [ 1.2666e-07, 0.0000e+00, 6.5088e-05, ..., 1.4910e-06, + -8.6352e-06, -6.1514e-07], + [ 9.4355e-08, 0.0000e+00, 1.4700e-05, ..., 4.3064e-06, + 6.4038e-06, 1.4178e-05], + [-1.1800e-06, 0.0000e+00, 7.9125e-06, ..., 1.1474e-06, + 7.4208e-06, -1.8790e-05]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0329, -0.0288, 0.0145, 0.0309, -0.0263, 0.0075, -0.0128, 0.0104, + 0.0181, 0.0008], device='cuda:0'), grad: tensor([ 4.8093e-06, 5.6587e-06, -1.6332e-04, 9.2685e-05, 2.7955e-05, + -6.2704e-05, -3.3490e-06, 5.4777e-05, 4.4197e-05, -7.4925e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 214.72, cls_loss 0.0034 cls_loss_mapping 0.0100 cls_loss_causal 0.6064 re_mapping 0.0090 re_causal 0.0294 /// teacc 99.00 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0198, 0.0203, -0.0161, ..., -0.0498, -0.0392, 0.0487], + [-0.0146, -0.0100, -0.0349, ..., 0.0972, -0.0458, -0.0305], + [-0.0109, -0.0247, 0.1237, ..., -0.0635, -0.0830, -0.0458], + ..., + [-0.0241, -0.0438, -0.0112, ..., 0.0073, 0.0558, -0.0135], + [-0.0445, -0.0434, -0.0174, ..., -0.0660, 0.0180, -0.0709], + [-0.0944, -0.0288, -0.0785, ..., -0.0667, 0.0230, 0.0492]], + device='cuda:0'), grad: tensor([[ 4.1351e-07, 0.0000e+00, -5.8353e-05, ..., 7.1432e-07, + -5.3458e-06, -2.2486e-05], + [ 2.4177e-06, 0.0000e+00, 4.1611e-06, ..., -1.5581e-06, + 9.4296e-07, 2.0843e-06], + [ 1.0412e-06, 0.0000e+00, 6.9886e-06, ..., 7.6415e-07, + 2.6654e-06, 4.5337e-06], + ..., + [ 8.5533e-06, 0.0000e+00, 1.4417e-06, ..., 3.6019e-07, + -1.5146e-07, 6.6273e-06], + [ 1.0170e-06, 0.0000e+00, 2.2441e-05, ..., -5.5656e-06, + 4.9844e-06, 3.5460e-07], + [ 2.9616e-07, 0.0000e+00, 1.2070e-05, ..., 6.0536e-07, + 1.1232e-06, -4.1515e-05]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0332, -0.0293, 0.0152, 0.0307, -0.0264, 0.0075, -0.0118, 0.0103, + 0.0179, 0.0007], device='cuda:0'), grad: tensor([-8.3447e-05, 1.2755e-05, 2.2456e-05, 4.0889e-05, 5.1081e-05, + -1.0319e-05, 2.2158e-05, 2.3872e-05, -1.4648e-05, -6.4909e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 214.59, cls_loss 0.0037 cls_loss_mapping 0.0104 cls_loss_causal 0.6117 re_mapping 0.0091 re_causal 0.0293 /// teacc 98.95 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0201, 0.0203, -0.0149, ..., -0.0501, -0.0393, 0.0497], + [-0.0148, -0.0100, -0.0358, ..., 0.0977, -0.0459, -0.0307], + [-0.0111, -0.0247, 0.1238, ..., -0.0637, -0.0848, -0.0461], + ..., + [-0.0244, -0.0438, -0.0098, ..., 0.0071, 0.0565, -0.0141], + [-0.0447, -0.0434, -0.0183, ..., -0.0665, 0.0178, -0.0708], + [-0.0949, -0.0288, -0.0795, ..., -0.0669, 0.0230, 0.0495]], + device='cuda:0'), grad: tensor([[ 3.5409e-06, 0.0000e+00, -5.9567e-06, ..., 6.0610e-06, + 1.6401e-06, -3.9861e-06], + [-3.6180e-05, 0.0000e+00, -3.2573e-07, ..., -1.4031e-04, + -1.7971e-05, -3.5346e-05], + [ 1.8524e-06, 0.0000e+00, -1.7822e-05, ..., 1.2733e-05, + 1.7677e-06, 1.4929e-06], + ..., + [ 6.1207e-06, 0.0000e+00, 3.8780e-06, ..., 1.5289e-05, + -1.1057e-05, 6.1877e-06], + [ 1.4625e-05, 0.0000e+00, 1.1578e-05, ..., 4.7415e-05, + 1.6555e-05, 3.3557e-05], + [ 8.4043e-06, 0.0000e+00, -1.4566e-06, ..., 2.1368e-05, + 7.7337e-06, -1.3314e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0325, -0.0291, 0.0143, 0.0308, -0.0264, 0.0072, -0.0120, 0.0108, + 0.0181, 0.0006], device='cuda:0'), grad: tensor([ 5.5842e-06, -3.3617e-04, 3.2550e-07, 9.3997e-05, 7.4208e-05, + -7.7903e-05, 3.0752e-06, 3.2693e-05, 1.6868e-04, 3.5226e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 214.32, cls_loss 0.0035 cls_loss_mapping 0.0108 cls_loss_causal 0.6186 re_mapping 0.0089 re_causal 0.0294 /// teacc 98.99 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0205, 0.0203, -0.0146, ..., -0.0504, -0.0397, 0.0500], + [-0.0138, -0.0100, -0.0366, ..., 0.0988, -0.0461, -0.0309], + [-0.0113, -0.0247, 0.1247, ..., -0.0638, -0.0853, -0.0465], + ..., + [-0.0251, -0.0438, -0.0095, ..., 0.0068, 0.0573, -0.0144], + [-0.0451, -0.0434, -0.0190, ..., -0.0677, 0.0179, -0.0711], + [-0.0954, -0.0288, -0.0799, ..., -0.0672, 0.0226, 0.0499]], + device='cuda:0'), grad: tensor([[ 8.2143e-07, 0.0000e+00, 1.4305e-06, ..., 3.9749e-06, + 1.7723e-06, -1.6745e-06], + [ 1.0598e-06, 0.0000e+00, 1.1665e-04, ..., 2.2128e-05, + 3.5316e-06, 7.3761e-06], + [ 4.4797e-07, 0.0000e+00, -1.6177e-04, ..., -4.3780e-05, + 3.7979e-06, 1.0990e-05], + ..., + [ 2.8615e-07, 0.0000e+00, 6.6124e-06, ..., 2.7064e-06, + 1.8269e-05, 3.9756e-05], + [ 4.9639e-07, 0.0000e+00, 1.3942e-06, ..., 6.1542e-06, + 4.5598e-06, 6.4373e-06], + [ 4.0885e-07, 0.0000e+00, 5.1111e-06, ..., 2.5593e-06, + -1.8418e-05, -4.3064e-05]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0327, -0.0285, 0.0143, 0.0306, -0.0269, 0.0071, -0.0120, 0.0112, + 0.0181, 0.0005], device='cuda:0'), grad: tensor([ 1.7405e-05, 1.4496e-04, -1.2791e-04, -2.2042e-04, 2.9981e-05, + 8.9765e-05, 6.5565e-06, 9.4593e-05, 1.0774e-05, -4.6045e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 214.52, cls_loss 0.0033 cls_loss_mapping 0.0108 cls_loss_causal 0.5884 re_mapping 0.0089 re_causal 0.0284 /// teacc 99.01 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0211, 0.0203, -0.0137, ..., -0.0504, -0.0400, 0.0500], + [-0.0148, -0.0100, -0.0370, ..., 0.0995, -0.0469, -0.0304], + [-0.0106, -0.0247, 0.1254, ..., -0.0640, -0.0857, -0.0468], + ..., + [-0.0264, -0.0438, -0.0094, ..., 0.0062, 0.0577, -0.0149], + [-0.0452, -0.0434, -0.0198, ..., -0.0680, 0.0180, -0.0716], + [-0.0961, -0.0288, -0.0802, ..., -0.0678, 0.0226, 0.0507]], + device='cuda:0'), grad: tensor([[ 7.7439e-07, 0.0000e+00, -1.7852e-05, ..., 4.5933e-06, + 8.0653e-07, -1.8048e-04], + [ 7.4552e-07, 0.0000e+00, 2.6803e-06, ..., -4.8848e-07, + 9.8646e-06, 1.0230e-05], + [ 8.8150e-07, 0.0000e+00, 1.5954e-06, ..., 2.3786e-06, + 6.1952e-06, 5.2080e-06], + ..., + [ 2.7055e-07, 0.0000e+00, 2.6897e-06, ..., 8.3633e-07, + -3.1531e-05, 1.8224e-05], + [ 8.2850e-06, 0.0000e+00, 2.7493e-06, ..., 2.0310e-05, + 4.1306e-05, 3.3855e-05], + [ 4.5495e-07, 0.0000e+00, 2.9691e-06, ..., 2.8331e-06, + 1.8805e-05, -3.8594e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0326, -0.0282, 0.0144, 0.0303, -0.0269, 0.0068, -0.0116, 0.0108, + 0.0180, 0.0007], device='cuda:0'), grad: tensor([-1.7202e-04, 3.8505e-05, 3.1173e-05, 6.6876e-05, 6.4552e-05, + -6.7661e-07, -9.4891e-05, 9.9372e-07, 1.0282e-04, -3.7044e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 77---------------------------------------------------- +epoch 77, time 230.46, cls_loss 0.0037 cls_loss_mapping 0.0104 cls_loss_causal 0.5952 re_mapping 0.0091 re_causal 0.0288 /// teacc 99.13 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0217, 0.0203, -0.0123, ..., -0.0504, -0.0401, 0.0504], + [-0.0153, -0.0100, -0.0375, ..., 0.0998, -0.0476, -0.0299], + [-0.0108, -0.0247, 0.1258, ..., -0.0647, -0.0859, -0.0477], + ..., + [-0.0274, -0.0438, -0.0101, ..., 0.0062, 0.0578, -0.0167], + [-0.0453, -0.0434, -0.0209, ..., -0.0674, 0.0177, -0.0723], + [-0.0968, -0.0288, -0.0789, ..., -0.0682, 0.0229, 0.0514]], + device='cuda:0'), grad: tensor([[ 1.2740e-06, 0.0000e+00, -2.3574e-05, ..., 8.7870e-07, + 1.3011e-06, -1.2092e-05], + [ 6.8173e-06, 0.0000e+00, 1.3582e-05, ..., -7.5474e-06, + 1.7341e-06, 4.0233e-06], + [ 1.7602e-06, 0.0000e+00, -1.6570e-04, ..., 9.9093e-07, + 2.0079e-06, 9.5218e-06], + ..., + [ 1.8016e-05, 0.0000e+00, -6.6161e-05, ..., 3.0212e-06, + -1.4865e-04, -1.6856e-04], + [ 3.9078e-06, 0.0000e+00, 1.8626e-05, ..., 4.0904e-06, + 2.0966e-07, 4.4674e-05], + [-3.3557e-05, 0.0000e+00, 1.2153e-04, ..., 1.0096e-05, + 1.5426e-04, -1.4853e-04]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0323, -0.0283, 0.0141, 0.0311, -0.0264, 0.0064, -0.0118, 0.0099, + 0.0181, 0.0012], device='cuda:0'), grad: tensor([-1.5080e-05, 2.4796e-05, -1.0204e-04, -6.5994e-04, 5.3453e-04, + 4.1437e-04, 2.0221e-05, -2.3615e-04, 1.2672e-04, -1.0824e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 214.38, cls_loss 0.0025 cls_loss_mapping 0.0084 cls_loss_causal 0.6127 re_mapping 0.0088 re_causal 0.0279 /// teacc 99.03 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0221, 0.0203, -0.0116, ..., -0.0504, -0.0402, 0.0505], + [-0.0154, -0.0100, -0.0378, ..., 0.1005, -0.0480, -0.0302], + [-0.0113, -0.0247, 0.1264, ..., -0.0651, -0.0863, -0.0481], + ..., + [-0.0280, -0.0438, -0.0099, ..., 0.0060, 0.0585, -0.0168], + [-0.0455, -0.0434, -0.0217, ..., -0.0679, 0.0176, -0.0726], + [-0.0975, -0.0288, -0.0793, ..., -0.0684, 0.0226, 0.0519]], + device='cuda:0'), grad: tensor([[ 8.9593e-07, 0.0000e+00, 4.9531e-05, ..., 3.3691e-07, + 9.9242e-06, 1.1705e-05], + [ 2.2557e-06, 0.0000e+00, 8.4937e-06, ..., 1.3388e-08, + 9.2089e-06, 7.3351e-06], + [-2.8417e-05, 0.0000e+00, -1.5128e-04, ..., -5.5917e-06, + 2.1055e-05, -1.3188e-06], + ..., + [ 8.4937e-06, 0.0000e+00, -6.3002e-05, ..., 4.1933e-07, + -3.9607e-05, 6.9961e-06], + [-6.3740e-06, 0.0000e+00, 5.0291e-06, ..., 5.1260e-06, + 6.6280e-05, 6.2406e-05], + [ 1.8794e-06, 0.0000e+00, 1.5117e-05, ..., 2.5867e-07, + 4.8935e-05, 8.9481e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0323, -0.0281, 0.0141, 0.0308, -0.0263, 0.0062, -0.0119, 0.0103, + 0.0180, 0.0011], device='cuda:0'), grad: tensor([ 7.9751e-05, 3.7223e-05, -1.2171e-04, 3.3641e-04, 1.2612e-04, + -7.6771e-04, 1.0741e-04, -5.7399e-05, 1.6928e-04, 9.0599e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.32, cls_loss 0.0038 cls_loss_mapping 0.0106 cls_loss_causal 0.6113 re_mapping 0.0092 re_causal 0.0280 /// teacc 98.93 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0227, 0.0203, -0.0112, ..., -0.0508, -0.0405, 0.0509], + [-0.0160, -0.0100, -0.0384, ..., 0.1008, -0.0502, -0.0305], + [-0.0113, -0.0247, 0.1274, ..., -0.0654, -0.0869, -0.0483], + ..., + [-0.0285, -0.0438, -0.0096, ..., 0.0062, 0.0599, -0.0166], + [-0.0462, -0.0434, -0.0223, ..., -0.0685, 0.0174, -0.0731], + [-0.0984, -0.0288, -0.0802, ..., -0.0688, 0.0223, 0.0519]], + device='cuda:0'), grad: tensor([[ 3.5670e-07, 1.0472e-07, -8.0884e-05, ..., 1.1355e-05, + 9.1828e-07, -7.4863e-05], + [ 3.9116e-06, 1.3169e-06, 1.3083e-05, ..., 2.7865e-06, + 5.3048e-06, 7.6815e-06], + [ 9.3644e-07, 3.0780e-07, -4.1890e-04, ..., 6.1579e-06, + 2.4643e-06, 2.1592e-05], + ..., + [ 9.5088e-07, 2.8452e-07, 4.0126e-04, ..., 2.7078e-07, + -7.8231e-06, 4.6730e-05], + [ 1.7020e-07, 4.4645e-08, 1.7300e-05, ..., 2.4244e-05, + 7.4320e-07, 4.5627e-05], + [ 1.1129e-06, 3.2759e-07, 5.7295e-06, ..., 1.2638e-06, + -4.1537e-06, -3.0965e-05]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0322, -0.0287, 0.0143, 0.0305, -0.0262, 0.0066, -0.0117, 0.0111, + 0.0176, 0.0005], device='cuda:0'), grad: tensor([-3.4928e-05, 3.4422e-05, -3.3617e-04, 1.5363e-05, 8.5309e-06, + 5.0604e-05, -3.1137e-04, 4.2963e-04, 1.8215e-04, -3.7730e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 214.24, cls_loss 0.0034 cls_loss_mapping 0.0112 cls_loss_causal 0.6377 re_mapping 0.0080 re_causal 0.0276 /// teacc 99.08 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0235, 0.0203, -0.0107, ..., -0.0511, -0.0407, 0.0510], + [-0.0167, -0.0100, -0.0402, ..., 0.1009, -0.0505, -0.0308], + [-0.0113, -0.0246, 0.1288, ..., -0.0649, -0.0873, -0.0488], + ..., + [-0.0287, -0.0438, -0.0095, ..., 0.0058, 0.0605, -0.0168], + [-0.0468, -0.0434, -0.0226, ..., -0.0690, 0.0174, -0.0735], + [-0.1000, -0.0288, -0.0807, ..., -0.0706, 0.0221, 0.0526]], + device='cuda:0'), grad: tensor([[ 7.7416e-08, 0.0000e+00, -8.5905e-06, ..., 2.5192e-07, + 3.0897e-07, -7.4394e-06], + [ 3.7113e-07, 0.0000e+00, 1.6196e-06, ..., -5.1111e-06, + 8.9221e-07, 3.6415e-07], + [ 6.7346e-08, 0.0000e+00, -1.0483e-05, ..., 5.7416e-07, + 1.0235e-06, 1.3560e-06], + ..., + [ 2.4564e-07, 0.0000e+00, 2.1420e-06, ..., 1.8310e-06, + -2.8238e-06, 1.6345e-06], + [ 1.4261e-07, 0.0000e+00, 5.6624e-06, ..., 1.1092e-06, + 1.1707e-06, 4.3325e-06], + [ 8.8848e-07, 0.0000e+00, 3.6173e-06, ..., 5.2340e-07, + -5.7928e-07, -6.6310e-06]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0325, -0.0291, 0.0150, 0.0303, -0.0251, 0.0064, -0.0115, 0.0111, + 0.0176, -0.0001], device='cuda:0'), grad: tensor([-1.4409e-05, -3.5502e-06, 7.7719e-07, 1.7688e-05, 3.4831e-07, + -1.6034e-05, 4.5113e-06, 5.6289e-06, 1.2375e-05, -7.3500e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 214.25, cls_loss 0.0033 cls_loss_mapping 0.0101 cls_loss_causal 0.5708 re_mapping 0.0088 re_causal 0.0279 /// teacc 98.91 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0237, 0.0202, -0.0099, ..., -0.0513, -0.0409, 0.0512], + [-0.0175, -0.0101, -0.0406, ..., 0.1015, -0.0514, -0.0334], + [-0.0116, -0.0246, 0.1296, ..., -0.0654, -0.0877, -0.0490], + ..., + [-0.0292, -0.0438, -0.0087, ..., 0.0056, 0.0614, -0.0168], + [-0.0471, -0.0434, -0.0232, ..., -0.0695, 0.0174, -0.0741], + [-0.1003, -0.0288, -0.0820, ..., -0.0712, 0.0215, 0.0533]], + device='cuda:0'), grad: tensor([[ 3.2149e-06, 0.0000e+00, 6.0834e-06, ..., 7.5772e-06, + 1.0394e-06, 6.0834e-06], + [ 3.6415e-06, 0.0000e+00, -3.3259e-05, ..., -1.0353e-04, + 2.4550e-06, 1.1520e-06], + [ 1.5656e-06, 0.0000e+00, -6.6236e-06, ..., 4.2975e-05, + 1.2089e-06, -1.1589e-07], + ..., + [ 4.2282e-06, 0.0000e+00, 1.2927e-05, ..., 4.2140e-05, + -7.2643e-06, 9.2834e-06], + [ 4.7311e-06, 0.0000e+00, 1.0021e-05, ..., 1.1206e-05, + 2.2445e-06, 1.7360e-05], + [ 1.0565e-05, 0.0000e+00, -3.3956e-06, ..., 4.0047e-06, + -4.9062e-06, -5.1290e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0324, -0.0295, 0.0149, 0.0304, -0.0250, 0.0070, -0.0123, 0.0115, + 0.0175, -0.0002], device='cuda:0'), grad: tensor([ 2.7299e-05, -1.7416e-04, 6.0707e-05, 2.3603e-05, -3.5129e-06, + 9.0599e-06, -1.5602e-05, 8.5711e-05, 5.4389e-05, -6.7353e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 214.40, cls_loss 0.0028 cls_loss_mapping 0.0086 cls_loss_causal 0.6071 re_mapping 0.0084 re_causal 0.0279 /// teacc 99.00 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0239, 0.0202, -0.0089, ..., -0.0502, -0.0413, 0.0515], + [-0.0178, -0.0101, -0.0410, ..., 0.1017, -0.0517, -0.0345], + [-0.0120, -0.0246, 0.1301, ..., -0.0658, -0.0880, -0.0494], + ..., + [-0.0299, -0.0438, -0.0087, ..., 0.0053, 0.0618, -0.0173], + [-0.0474, -0.0435, -0.0234, ..., -0.0694, 0.0172, -0.0744], + [-0.1014, -0.0288, -0.0824, ..., -0.0712, 0.0210, 0.0542]], + device='cuda:0'), grad: tensor([[ 4.6426e-07, 5.8208e-11, -3.2056e-06, ..., 1.2219e-06, + 1.1250e-05, 2.4959e-06], + [ 3.1572e-07, 7.5670e-10, 1.9401e-05, ..., -3.5968e-06, + 2.4498e-05, 5.2825e-06], + [ 2.1444e-07, 5.8208e-11, -3.9369e-05, ..., 1.3104e-06, + 7.6815e-06, 4.3735e-06], + ..., + [ 1.2980e-08, 1.7462e-10, -4.6790e-06, ..., 1.4296e-06, + 4.0460e-04, 3.9935e-04], + [ 2.5402e-07, 5.8208e-11, 8.0466e-06, ..., 2.3134e-06, + 8.0347e-05, 6.3300e-05], + [ 4.5227e-08, 7.5670e-10, 3.0231e-06, ..., 4.2259e-07, + -7.6675e-04, -6.4278e-04]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0318, -0.0299, 0.0147, 0.0305, -0.0244, 0.0072, -0.0130, 0.0112, + 0.0176, -0.0002], device='cuda:0'), grad: tensor([ 2.5809e-05, 8.4519e-05, -3.6240e-05, 6.0976e-05, 2.1636e-04, + 3.3593e-04, 4.6939e-05, 1.2150e-03, 2.1279e-04, -2.1610e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 214.47, cls_loss 0.0028 cls_loss_mapping 0.0094 cls_loss_causal 0.5857 re_mapping 0.0085 re_causal 0.0278 /// teacc 98.98 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0241, 0.0202, -0.0088, ..., -0.0504, -0.0415, 0.0516], + [-0.0186, -0.0101, -0.0413, ..., 0.1021, -0.0519, -0.0347], + [-0.0123, -0.0246, 0.1303, ..., -0.0662, -0.0885, -0.0496], + ..., + [-0.0295, -0.0438, -0.0085, ..., 0.0049, 0.0624, -0.0178], + [-0.0476, -0.0435, -0.0232, ..., -0.0701, 0.0171, -0.0748], + [-0.1021, -0.0289, -0.0829, ..., -0.0714, 0.0206, 0.0551]], + device='cuda:0'), grad: tensor([[ 1.4380e-06, 0.0000e+00, -1.7174e-06, ..., 2.7698e-06, + 1.9372e-06, -1.9427e-06], + [ 1.8720e-06, 0.0000e+00, 2.0787e-06, ..., 1.7211e-06, + 1.1502e-06, 9.0292e-07], + [ 9.0199e-07, 0.0000e+00, -1.5855e-05, ..., 2.0340e-06, + 2.0955e-06, 2.4810e-06], + ..., + [ 1.0030e-06, 0.0000e+00, 5.1446e-06, ..., 1.2228e-06, + 1.8096e-06, 3.0976e-06], + [ 1.3364e-06, 0.0000e+00, 1.9316e-06, ..., 4.9509e-06, + 8.7991e-06, 1.5378e-05], + [ 2.3078e-06, 0.0000e+00, 5.6904e-07, ..., 1.3690e-06, + -3.7868e-06, -1.8865e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([-3.2018e-02, -2.9788e-02, 1.4144e-02, 3.0806e-02, -2.4935e-02, + 6.9101e-03, -1.2462e-02, 1.1350e-02, 1.7934e-02, 1.4836e-05], + device='cuda:0'), grad: tensor([ 7.1302e-06, 1.1362e-05, 7.1414e-06, -6.1691e-05, 5.8003e-06, + -6.2166e-07, -3.1054e-05, 2.4065e-05, 6.9559e-05, -3.1710e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 214.36, cls_loss 0.0031 cls_loss_mapping 0.0093 cls_loss_causal 0.5945 re_mapping 0.0083 re_causal 0.0271 /// teacc 99.04 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0246, 0.0202, -0.0082, ..., -0.0505, -0.0423, 0.0518], + [-0.0190, -0.0101, -0.0417, ..., 0.1024, -0.0522, -0.0349], + [-0.0124, -0.0246, 0.1312, ..., -0.0664, -0.0887, -0.0499], + ..., + [-0.0291, -0.0438, -0.0087, ..., 0.0046, 0.0630, -0.0182], + [-0.0476, -0.0435, -0.0236, ..., -0.0704, 0.0164, -0.0757], + [-0.1028, -0.0289, -0.0831, ..., -0.0716, 0.0206, 0.0559]], + device='cuda:0'), grad: tensor([[ 5.3202e-08, 2.4447e-09, 1.0449e-06, ..., 4.8541e-06, + 6.2212e-07, 7.4459e-07], + [ 3.7486e-08, 8.9640e-09, 8.2627e-06, ..., -1.3554e-04, + 2.5015e-06, -1.5005e-05], + [-1.1770e-07, -6.0187e-08, -3.1173e-05, ..., 9.6858e-06, + 2.7753e-06, 1.7313e-06], + ..., + [ 2.3202e-07, 1.3970e-09, 3.4254e-06, ..., 4.3988e-05, + -9.7081e-06, 1.3508e-05], + [ 1.5553e-07, 6.9849e-10, 5.5283e-06, ..., -7.3807e-08, + -5.6773e-06, 7.7635e-06], + [-2.6617e-06, 0.0000e+00, 1.6233e-06, ..., 1.1884e-05, + 3.9600e-06, -3.4153e-05]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0324, -0.0300, 0.0143, 0.0302, -0.0252, 0.0069, -0.0116, 0.0113, + 0.0183, 0.0003], device='cuda:0'), grad: tensor([ 1.8731e-05, -2.4414e-04, 2.1204e-05, 2.4647e-05, 1.0490e-04, + 6.7890e-05, 3.7670e-05, 1.0747e-04, -1.2374e-04, -1.4439e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 214.18, cls_loss 0.0025 cls_loss_mapping 0.0083 cls_loss_causal 0.5765 re_mapping 0.0085 re_causal 0.0277 /// teacc 99.02 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0249, 0.0202, -0.0078, ..., -0.0508, -0.0428, 0.0519], + [-0.0197, -0.0101, -0.0415, ..., 0.1029, -0.0524, -0.0353], + [-0.0128, -0.0246, 0.1318, ..., -0.0672, -0.0890, -0.0503], + ..., + [-0.0295, -0.0438, -0.0087, ..., 0.0042, 0.0635, -0.0184], + [-0.0475, -0.0435, -0.0245, ..., -0.0706, 0.0161, -0.0760], + [-0.1036, -0.0289, -0.0835, ..., -0.0720, 0.0204, 0.0562]], + device='cuda:0'), grad: tensor([[ 2.0373e-07, 1.7462e-08, 3.3025e-06, ..., 4.0792e-06, + 2.1029e-06, 3.1013e-06], + [ 3.6485e-07, 1.7462e-08, 5.1521e-06, ..., -1.6056e-06, + 3.1944e-06, 1.7239e-06], + [ 2.6803e-06, -3.5390e-07, -1.4372e-05, ..., 5.1074e-06, + 1.7226e-05, 3.2019e-06], + ..., + [ 8.9174e-08, 4.6799e-08, -5.7667e-05, ..., 1.0114e-06, + -6.9737e-05, 1.5441e-06], + [ 2.4796e-07, 4.9127e-08, 7.2196e-06, ..., 3.7849e-05, + 5.6252e-06, 2.5764e-05], + [ 1.9418e-07, 9.7789e-09, 2.8759e-05, ..., 1.5367e-06, + 3.3110e-05, -1.3448e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([-3.2597e-02, -3.0042e-02, 1.3809e-02, 3.0340e-02, -2.4986e-02, + 7.1356e-03, -1.1128e-02, 1.1158e-02, 1.8457e-02, -2.6169e-05], + device='cuda:0'), grad: tensor([ 2.5585e-05, 1.8731e-05, 3.3259e-05, -1.5751e-05, 1.6183e-05, + 2.4819e-04, -4.4727e-04, -1.2517e-04, 1.9073e-04, 5.5015e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 214.34, cls_loss 0.0026 cls_loss_mapping 0.0081 cls_loss_causal 0.5831 re_mapping 0.0083 re_causal 0.0265 /// teacc 99.12 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0252, 0.0204, -0.0073, ..., -0.0511, -0.0429, 0.0521], + [-0.0200, -0.0101, -0.0421, ..., 0.1032, -0.0527, -0.0356], + [-0.0131, -0.0245, 0.1326, ..., -0.0674, -0.0893, -0.0506], + ..., + [-0.0298, -0.0438, -0.0085, ..., 0.0040, 0.0639, -0.0190], + [-0.0486, -0.0435, -0.0248, ..., -0.0711, 0.0151, -0.0781], + [-0.1038, -0.0289, -0.0839, ..., -0.0723, 0.0203, 0.0570]], + device='cuda:0'), grad: tensor([[ 5.2825e-06, 1.1642e-10, -5.4330e-05, ..., 4.3064e-06, + -3.1352e-05, -2.3872e-05], + [ 4.6879e-05, 1.5134e-09, 1.3717e-05, ..., 1.9938e-05, + 1.0200e-05, 6.8881e-06], + [ 2.1812e-06, 1.1642e-10, 1.7732e-05, ..., 2.0452e-06, + 2.2680e-05, 4.9882e-06], + ..., + [ 7.0855e-06, 4.6566e-10, -8.4102e-05, ..., 5.4501e-06, + -7.1824e-05, 2.2147e-06], + [ 1.3299e-05, 1.1642e-10, 2.1920e-05, ..., 9.1121e-06, + 1.3910e-05, 5.9232e-06], + [ 7.8976e-05, 2.7940e-09, 3.3796e-05, ..., 4.4018e-05, + 2.6971e-05, 1.0885e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0327, -0.0304, 0.0138, 0.0306, -0.0251, 0.0075, -0.0106, 0.0111, + 0.0171, 0.0006], device='cuda:0'), grad: tensor([-1.2505e-04, 9.4831e-05, 5.7161e-05, 1.7717e-05, -3.1781e-04, + 5.1200e-05, 1.0943e-04, -1.4985e-04, 6.5565e-05, 1.9729e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 214.39, cls_loss 0.0041 cls_loss_mapping 0.0122 cls_loss_causal 0.5907 re_mapping 0.0085 re_causal 0.0262 /// teacc 98.95 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0257, 0.0204, -0.0070, ..., -0.0515, -0.0432, 0.0523], + [-0.0209, -0.0101, -0.0427, ..., 0.1028, -0.0532, -0.0370], + [-0.0121, -0.0245, 0.1340, ..., -0.0672, -0.0897, -0.0507], + ..., + [-0.0301, -0.0439, -0.0082, ..., 0.0036, 0.0645, -0.0202], + [-0.0482, -0.0435, -0.0261, ..., -0.0715, 0.0147, -0.0787], + [-0.1043, -0.0290, -0.0843, ..., -0.0724, 0.0201, 0.0581]], + device='cuda:0'), grad: tensor([[ 5.6298e-07, 0.0000e+00, -2.7008e-08, ..., 6.1560e-07, + 2.6310e-07, 1.2573e-08], + [ 6.3051e-07, 0.0000e+00, 7.9907e-07, ..., -2.8744e-05, + 1.2177e-07, 1.7951e-07], + [ 3.3970e-07, 0.0000e+00, -4.7833e-06, ..., 7.5111e-07, + 1.7113e-07, 1.8766e-07], + ..., + [ 1.2144e-06, 0.0000e+00, 1.2154e-06, ..., 2.2903e-05, + -3.4925e-07, 2.2631e-06], + [ 9.8627e-07, 0.0000e+00, 3.3458e-07, ..., 6.9803e-07, + 2.6636e-07, 1.0561e-06], + [ 9.0837e-05, 0.0000e+00, -6.5891e-08, ..., 2.9616e-06, + 2.3562e-07, -1.4044e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0330, -0.0314, 0.0147, 0.0309, -0.0253, 0.0077, -0.0104, 0.0105, + 0.0167, 0.0014], device='cuda:0'), grad: tensor([ 1.9595e-06, -3.4600e-05, -1.0496e-06, 5.3421e-06, -1.4496e-04, + 1.3858e-05, -7.0632e-06, 3.6120e-05, -3.8221e-06, 1.3411e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 214.41, cls_loss 0.0029 cls_loss_mapping 0.0096 cls_loss_causal 0.6272 re_mapping 0.0080 re_causal 0.0264 /// teacc 98.91 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0264, 0.0203, -0.0062, ..., -0.0519, -0.0433, 0.0526], + [-0.0212, -0.0101, -0.0446, ..., 0.1024, -0.0534, -0.0384], + [-0.0120, -0.0243, 0.1349, ..., -0.0669, -0.0907, -0.0510], + ..., + [-0.0306, -0.0439, -0.0073, ..., 0.0026, 0.0655, -0.0201], + [-0.0484, -0.0436, -0.0264, ..., -0.0723, 0.0149, -0.0790], + [-0.1048, -0.0290, -0.0851, ..., -0.0729, 0.0196, 0.0584]], + device='cuda:0'), grad: tensor([[ 8.5216e-08, 0.0000e+00, -1.9800e-06, ..., 1.1483e-06, + 1.4063e-07, -1.6559e-06], + [ 1.0850e-07, 0.0000e+00, 5.0217e-06, ..., -1.9550e-05, + 6.3330e-07, 3.1199e-07], + [ 1.5413e-07, 0.0000e+00, -5.2541e-05, ..., 2.1700e-06, + -8.5356e-07, 5.7463e-07], + ..., + [ 4.3400e-07, 0.0000e+00, 2.5660e-05, ..., 2.0191e-06, + -1.2703e-06, 2.5965e-06], + [ 6.0583e-07, 0.0000e+00, -5.4063e-07, ..., 6.8396e-06, + 1.1129e-06, 1.6717e-06], + [ 1.1595e-07, 0.0000e+00, 1.2638e-06, ..., 6.8778e-07, + 9.9372e-07, -5.9940e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0331, -0.0321, 0.0149, 0.0307, -0.0256, 0.0084, -0.0106, 0.0111, + 0.0170, 0.0011], device='cuda:0'), grad: tensor([-4.0466e-07, -2.8640e-05, -3.7014e-05, 1.0526e-04, 9.7454e-06, + -8.6904e-05, 1.4663e-05, 3.0637e-05, -2.0657e-06, -5.4650e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 213.98, cls_loss 0.0027 cls_loss_mapping 0.0081 cls_loss_causal 0.6064 re_mapping 0.0084 re_causal 0.0267 /// teacc 99.13 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0267, 0.0207, -0.0064, ..., -0.0521, -0.0435, 0.0523], + [-0.0215, -0.0101, -0.0449, ..., 0.1030, -0.0536, -0.0385], + [-0.0121, -0.0243, 0.1357, ..., -0.0673, -0.0912, -0.0513], + ..., + [-0.0315, -0.0439, -0.0072, ..., 0.0022, 0.0659, -0.0204], + [-0.0485, -0.0436, -0.0270, ..., -0.0727, 0.0148, -0.0793], + [-0.1066, -0.0290, -0.0854, ..., -0.0731, 0.0195, 0.0594]], + device='cuda:0'), grad: tensor([[ 1.4618e-05, 0.0000e+00, -7.1190e-06, ..., 9.7305e-06, + 4.6343e-06, -1.5423e-05], + [ 6.6943e-06, 0.0000e+00, 2.0102e-05, ..., -1.2290e-04, + 1.8731e-05, -2.5302e-05], + [ 4.2021e-05, 0.0000e+00, 1.5235e-04, ..., 2.1324e-05, + 1.4043e-04, 6.1691e-06], + ..., + [-8.0407e-05, 0.0000e+00, -4.3678e-04, ..., 5.0306e-05, + -4.0889e-04, 1.9670e-05], + [ 1.3113e-05, 0.0000e+00, 4.9055e-05, ..., 7.3500e-06, + 4.5508e-05, 2.7537e-05], + [ 5.2415e-06, 0.0000e+00, 4.3899e-05, ..., 7.8753e-06, + 3.4899e-05, -1.7655e-04]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0332, -0.0317, 0.0148, 0.0308, -0.0246, 0.0081, -0.0108, 0.0109, + 0.0170, 0.0005], device='cuda:0'), grad: tensor([ 1.5691e-05, -3.3140e-04, 4.5180e-04, -2.6543e-06, 1.2722e-03, + 4.7654e-05, -4.6778e-04, -8.6117e-04, 1.9872e-04, -3.2234e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 214.25, cls_loss 0.0044 cls_loss_mapping 0.0121 cls_loss_causal 0.5896 re_mapping 0.0091 re_causal 0.0268 /// teacc 98.96 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0273, 0.0223, -0.0052, ..., -0.0525, -0.0437, 0.0528], + [-0.0217, -0.0101, -0.0453, ..., 0.1035, -0.0539, -0.0386], + [-0.0122, -0.0238, 0.1375, ..., -0.0676, -0.0908, -0.0520], + ..., + [-0.0320, -0.0442, -0.0084, ..., 0.0019, 0.0658, -0.0208], + [-0.0492, -0.0436, -0.0266, ..., -0.0733, 0.0153, -0.0787], + [-0.1071, -0.0294, -0.0863, ..., -0.0732, 0.0192, 0.0598]], + device='cuda:0'), grad: tensor([[ 8.7777e-07, 0.0000e+00, 2.4959e-06, ..., 1.0692e-05, + 8.8848e-07, 4.9546e-06], + [ 2.7847e-06, 0.0000e+00, 1.8915e-06, ..., -5.8487e-06, + 5.5833e-07, 4.9686e-07], + [ 1.0235e-06, 0.0000e+00, -2.8744e-05, ..., 4.4378e-07, + 6.0722e-07, 5.6019e-07], + ..., + [ 1.3210e-05, 0.0000e+00, 4.6566e-06, ..., 1.4473e-06, + 1.1977e-06, 9.4855e-07], + [ 1.7704e-06, 0.0000e+00, 5.1819e-06, ..., 9.9186e-07, + 8.1258e-07, 2.0545e-06], + [ 1.9118e-05, 0.0000e+00, 9.2806e-07, ..., 8.5635e-07, + 2.6636e-06, -3.6322e-06]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0330, -0.0317, 0.0149, 0.0331, -0.0251, 0.0079, -0.0112, 0.0101, + 0.0163, 0.0004], device='cuda:0'), grad: tensor([ 2.2411e-05, 1.3188e-06, -2.6524e-05, 3.1888e-05, -4.4554e-05, + -2.1473e-05, -7.0073e-06, 2.3708e-05, 2.5667e-06, 1.7673e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 214.28, cls_loss 0.0029 cls_loss_mapping 0.0083 cls_loss_causal 0.5912 re_mapping 0.0089 re_causal 0.0267 /// teacc 98.97 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0277, 0.0223, -0.0050, ..., -0.0530, -0.0441, 0.0527], + [-0.0223, -0.0101, -0.0458, ..., 0.1042, -0.0544, -0.0387], + [-0.0122, -0.0237, 0.1390, ..., -0.0680, -0.0909, -0.0524], + ..., + [-0.0326, -0.0442, -0.0087, ..., 0.0013, 0.0664, -0.0209], + [-0.0494, -0.0437, -0.0275, ..., -0.0739, 0.0157, -0.0788], + [-0.1073, -0.0294, -0.0866, ..., -0.0727, 0.0188, 0.0605]], + device='cuda:0'), grad: tensor([[ 1.9614e-06, 0.0000e+00, -8.4266e-06, ..., 1.5870e-06, + -4.0606e-07, -4.2804e-06], + [ 2.1402e-06, 0.0000e+00, 5.9325e-07, ..., -9.1314e-05, + 1.7025e-06, 1.4771e-06], + [ 7.0222e-07, 0.0000e+00, -4.5169e-08, ..., 2.1517e-05, + 1.9046e-06, 1.4016e-06], + ..., + [ 6.9067e-06, 0.0000e+00, -9.8795e-06, ..., 2.0742e-05, + -4.7058e-05, 3.2336e-06], + [ 1.9538e-04, 0.0000e+00, 2.8610e-06, ..., 1.3612e-05, + 6.5044e-06, 3.6740e-04], + [-2.0456e-04, 0.0000e+00, 6.4559e-06, ..., 6.2399e-06, + 3.1173e-05, -4.2725e-04]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0335, -0.0314, 0.0153, 0.0328, -0.0255, 0.0078, -0.0117, 0.0101, + 0.0166, 0.0009], device='cuda:0'), grad: tensor([-5.9456e-06, -1.7393e-04, 4.8995e-05, 2.5606e-04, 1.0610e-05, + -7.1764e-05, 2.0742e-05, -4.4256e-05, 1.0557e-03, -1.0967e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 214.38, cls_loss 0.0032 cls_loss_mapping 0.0102 cls_loss_causal 0.5873 re_mapping 0.0079 re_causal 0.0255 /// teacc 98.94 lr 0.00010000 +Epoch 94, weight, value: tensor([[-2.8471e-02, 2.2298e-02, -4.5561e-03, ..., -5.3449e-02, + -4.4358e-02, 5.2974e-02], + [-2.2698e-02, -1.0099e-02, -4.6294e-02, ..., 1.0538e-01, + -5.3089e-02, -3.8805e-02], + [-1.2412e-02, -2.3606e-02, 1.4001e-01, ..., -6.8627e-02, + -9.1246e-02, -5.2659e-02], + ..., + [-3.3376e-02, -4.4267e-02, -8.4083e-03, ..., -3.0199e-05, + 6.6641e-02, -2.1043e-02], + [-4.9329e-02, -4.3721e-02, -2.8206e-02, ..., -7.4380e-02, + 1.5710e-02, -7.8905e-02], + [-1.0929e-01, -2.9342e-02, -8.7437e-02, ..., -7.3637e-02, + 1.8309e-02, 6.0592e-02]], device='cuda:0'), grad: tensor([[ 4.6846e-07, 0.0000e+00, -5.1688e-08, ..., 1.2768e-06, + 2.0079e-06, 6.2771e-07], + [ 1.4799e-06, 0.0000e+00, 1.2256e-05, ..., 5.0031e-06, + 3.8385e-05, 9.2434e-07], + [ 5.6624e-07, 0.0000e+00, 5.2759e-07, ..., 9.9931e-07, + 8.1211e-06, 4.7870e-07], + ..., + [ 6.1654e-07, 0.0000e+00, -3.4511e-05, ..., -1.2763e-05, + -1.3876e-04, -5.3877e-07], + [ 2.5909e-06, 0.0000e+00, 8.8476e-07, ..., 1.2806e-06, + 6.7875e-06, 7.0892e-06], + [-1.6302e-05, 0.0000e+00, 1.3590e-05, ..., 4.7907e-06, + 4.3303e-05, -4.6343e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0334, -0.0302, 0.0153, 0.0323, -0.0244, 0.0077, -0.0123, 0.0096, + 0.0170, -0.0005], device='cuda:0'), grad: tensor([ 6.0312e-06, 8.8096e-05, 1.8165e-05, 7.9811e-05, 7.7963e-05, + 9.4203e-07, -2.7660e-06, -2.9254e-04, 2.2113e-05, 2.0508e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 214.40, cls_loss 0.0039 cls_loss_mapping 0.0109 cls_loss_causal 0.5587 re_mapping 0.0077 re_causal 0.0245 /// teacc 99.03 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0285, 0.0223, -0.0050, ..., -0.0512, -0.0446, 0.0524], + [-0.0229, -0.0101, -0.0485, ..., 0.1046, -0.0536, -0.0389], + [-0.0126, -0.0232, 0.1415, ..., -0.0692, -0.0922, -0.0530], + ..., + [-0.0341, -0.0444, -0.0098, ..., -0.0003, 0.0665, -0.0211], + [-0.0498, -0.0438, -0.0281, ..., -0.0728, 0.0160, -0.0800], + [-0.1096, -0.0294, -0.0873, ..., -0.0740, 0.0178, 0.0616]], + device='cuda:0'), grad: tensor([[ 6.9709e-07, 0.0000e+00, 7.5158e-07, ..., 1.0580e-06, + 1.7881e-07, -1.7136e-07], + [ 5.7407e-06, 0.0000e+00, 4.0680e-06, ..., -7.0743e-06, + 9.8534e-07, 2.0061e-06], + [ 3.5064e-07, 0.0000e+00, -4.2886e-05, ..., -5.2992e-07, + 3.9628e-07, 7.0920e-07], + ..., + [ 3.2652e-06, 0.0000e+00, 1.8850e-06, ..., 2.7567e-06, + -2.5518e-06, 1.3635e-06], + [ 6.1654e-07, 0.0000e+00, 1.6451e-05, ..., 4.1462e-06, + 6.8545e-07, 2.9895e-06], + [ 7.6294e-05, 0.0000e+00, 6.5984e-07, ..., 3.1054e-05, + 6.0489e-07, 1.7941e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([-3.3316e-02, -3.1589e-02, 1.4672e-02, 3.1611e-02, -2.4567e-02, + 8.4197e-03, -1.3645e-02, 8.8930e-03, 1.9047e-02, -8.8583e-05], + device='cuda:0'), grad: tensor([ 4.0047e-06, 3.8669e-06, -6.3956e-05, 1.3478e-05, -1.7130e-04, + 1.2413e-05, 8.6194e-07, 1.0163e-05, 3.6269e-05, 1.5426e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 214.09, cls_loss 0.0030 cls_loss_mapping 0.0096 cls_loss_causal 0.5918 re_mapping 0.0079 re_causal 0.0251 /// teacc 98.93 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0291, 0.0227, -0.0049, ..., -0.0513, -0.0450, 0.0513], + [-0.0226, -0.0102, -0.0487, ..., 0.1055, -0.0538, -0.0389], + [-0.0130, -0.0226, 0.1418, ..., -0.0700, -0.0930, -0.0532], + ..., + [-0.0345, -0.0445, -0.0092, ..., -0.0005, 0.0673, -0.0213], + [-0.0503, -0.0441, -0.0281, ..., -0.0738, 0.0154, -0.0804], + [-0.1100, -0.0295, -0.0879, ..., -0.0745, 0.0175, 0.0633]], + device='cuda:0'), grad: tensor([[ 2.4233e-06, 0.0000e+00, -5.6960e-06, ..., 3.8631e-06, + 4.7171e-07, -3.6880e-07], + [ 2.7165e-05, 0.0000e+00, 1.4892e-06, ..., -2.5019e-05, + 8.5821e-07, -8.2925e-06], + [ 3.6489e-06, 0.0000e+00, -1.9707e-06, ..., 2.0936e-06, + 1.0217e-06, 2.1588e-06], + ..., + [ 4.0025e-05, 0.0000e+00, 3.9190e-06, ..., 1.2711e-05, + -7.2308e-06, 3.1441e-06], + [ 1.2383e-05, 0.0000e+00, -2.9817e-05, ..., 5.4277e-06, + -5.0701e-06, -1.4350e-05], + [ 3.9369e-05, 0.0000e+00, 1.9982e-05, ..., 6.1803e-06, + 6.9067e-06, 3.2354e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0346, -0.0310, 0.0141, 0.0306, -0.0250, 0.0083, -0.0133, 0.0094, + 0.0198, 0.0004], device='cuda:0'), grad: tensor([ 1.3694e-05, -3.8683e-05, 1.2577e-05, 1.2882e-05, -1.4889e-04, + 3.9607e-05, 2.2709e-05, 8.9586e-05, -1.3006e-04, 1.2660e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 214.58, cls_loss 0.0025 cls_loss_mapping 0.0081 cls_loss_causal 0.5397 re_mapping 0.0082 re_causal 0.0256 /// teacc 99.01 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0295, 0.0226, -0.0043, ..., -0.0515, -0.0454, 0.0516], + [-0.0236, -0.0102, -0.0492, ..., 0.1058, -0.0540, -0.0387], + [-0.0119, -0.0223, 0.1429, ..., -0.0696, -0.0931, -0.0534], + ..., + [-0.0353, -0.0452, -0.0096, ..., -0.0007, 0.0673, -0.0223], + [-0.0507, -0.0443, -0.0284, ..., -0.0739, 0.0149, -0.0815], + [-0.1107, -0.0296, -0.0883, ..., -0.0751, 0.0175, 0.0637]], + device='cuda:0'), grad: tensor([[ 3.8324e-07, 4.1910e-09, 8.5384e-06, ..., 6.2166e-07, + 6.7521e-07, 6.7474e-07], + [ 1.1381e-06, 4.6566e-10, 7.8231e-07, ..., -7.3239e-06, + 4.6846e-07, 3.9069e-07], + [ 5.2014e-07, 4.6566e-10, -1.9595e-05, ..., 8.2003e-07, + 1.8384e-06, 1.5572e-06], + ..., + [ 2.0824e-06, 3.2596e-09, 2.8647e-06, ..., 9.3179e-07, + 1.2293e-06, 3.1050e-06], + [ 7.2457e-07, 1.8626e-09, -6.0111e-05, ..., 4.6678e-06, + -7.6771e-05, -7.8559e-05], + [ 5.2005e-06, 1.5367e-08, 5.5581e-05, ..., 1.2256e-06, + 6.7711e-05, 6.7413e-05]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0346, -0.0312, 0.0148, 0.0316, -0.0248, 0.0070, -0.0130, 0.0088, + 0.0197, 0.0003], device='cuda:0'), grad: tensor([ 1.3635e-05, -9.5069e-06, -9.4920e-06, 1.1861e-05, -1.2033e-05, + 3.2902e-05, -3.5074e-06, 1.8165e-05, -4.6444e-04, 4.2295e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 214.90, cls_loss 0.0022 cls_loss_mapping 0.0077 cls_loss_causal 0.5657 re_mapping 0.0077 re_causal 0.0247 /// teacc 99.03 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0298, 0.0225, -0.0035, ..., -0.0517, -0.0457, 0.0520], + [-0.0244, -0.0102, -0.0499, ..., 0.1062, -0.0545, -0.0386], + [-0.0111, -0.0223, 0.1435, ..., -0.0696, -0.0938, -0.0541], + ..., + [-0.0361, -0.0454, -0.0091, ..., -0.0008, 0.0685, -0.0221], + [-0.0510, -0.0446, -0.0286, ..., -0.0743, 0.0145, -0.0819], + [-0.1113, -0.0296, -0.0888, ..., -0.0754, 0.0167, 0.0640]], + device='cuda:0'), grad: tensor([[ 2.5379e-07, 0.0000e+00, -3.2276e-05, ..., 2.3982e-07, + 1.3122e-06, -3.6538e-05], + [ 9.1828e-07, 0.0000e+00, 1.1874e-06, ..., -3.8631e-06, + 1.3644e-06, 1.5022e-06], + [ 5.7416e-07, 0.0000e+00, -3.9041e-06, ..., 5.4715e-07, + 1.3169e-06, 3.6787e-06], + ..., + [-7.0920e-07, 0.0000e+00, 4.4256e-06, ..., 5.1083e-07, + -2.1234e-06, 1.0192e-05], + [ 5.5227e-07, 0.0000e+00, 8.9407e-08, ..., 1.6280e-06, + 2.0973e-06, 6.3442e-06], + [ 1.5823e-06, 0.0000e+00, 2.1994e-05, ..., 5.2853e-07, + -7.6368e-06, -9.5218e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0345, -0.0311, 0.0149, 0.0315, -0.0250, 0.0069, -0.0126, 0.0093, + 0.0195, 0.0003], device='cuda:0'), grad: tensor([-4.3988e-05, -1.1064e-06, 5.0515e-06, 2.8640e-05, 5.4628e-05, + -1.6487e-04, 1.3781e-04, 2.0087e-05, 2.4922e-06, -3.8803e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 214.80, cls_loss 0.0025 cls_loss_mapping 0.0077 cls_loss_causal 0.5657 re_mapping 0.0079 re_causal 0.0240 /// teacc 98.97 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0304, 0.0225, -0.0033, ..., -0.0516, -0.0460, 0.0524], + [-0.0251, -0.0102, -0.0502, ..., 0.1067, -0.0549, -0.0386], + [-0.0112, -0.0222, 0.1439, ..., -0.0700, -0.0944, -0.0544], + ..., + [-0.0370, -0.0454, -0.0086, ..., -0.0010, 0.0695, -0.0227], + [-0.0510, -0.0447, -0.0289, ..., -0.0752, 0.0138, -0.0830], + [-0.1119, -0.0295, -0.0890, ..., -0.0756, 0.0165, 0.0646]], + device='cuda:0'), grad: tensor([[ 4.8243e-07, 2.4820e-07, 1.8582e-05, ..., 8.1882e-06, + 2.6487e-06, 9.6112e-06], + [ 6.0424e-06, 8.8476e-09, 1.9986e-06, ..., -3.2615e-06, + 6.2492e-07, 9.4809e-07], + [ 1.8906e-07, -3.5763e-07, -9.2089e-05, ..., 1.6484e-06, + 6.2119e-07, 8.2189e-07], + ..., + [ 4.4797e-07, 9.7323e-08, 9.1419e-06, ..., 3.8818e-06, + -9.2238e-06, -2.0443e-07], + [ 2.7008e-06, 7.4506e-08, 4.4465e-05, ..., 9.8944e-06, + 2.6599e-06, 1.3441e-05], + [ 9.9558e-07, -9.9838e-07, 3.8221e-06, ..., 9.6206e-07, + 4.0233e-06, -8.4564e-06]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0343, -0.0307, 0.0146, 0.0312, -0.0250, 0.0071, -0.0130, 0.0095, + 0.0191, 0.0004], device='cuda:0'), grad: tensor([ 5.8770e-05, 4.6045e-06, -1.1247e-04, 2.4900e-05, -2.4065e-05, + 3.4660e-05, -9.2089e-05, 1.1258e-05, 9.1672e-05, 2.6785e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 214.26, cls_loss 0.0028 cls_loss_mapping 0.0075 cls_loss_causal 0.5596 re_mapping 0.0077 re_causal 0.0241 /// teacc 98.99 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0307, 0.0225, -0.0051, ..., -0.0517, -0.0465, 0.0508], + [-0.0258, -0.0102, -0.0507, ..., 0.1070, -0.0552, -0.0387], + [-0.0111, -0.0220, 0.1436, ..., -0.0703, -0.0965, -0.0550], + ..., + [-0.0373, -0.0455, -0.0088, ..., -0.0008, 0.0702, -0.0232], + [-0.0510, -0.0449, -0.0293, ..., -0.0756, 0.0136, -0.0834], + [-0.1123, -0.0291, -0.0874, ..., -0.0760, 0.0163, 0.0668]], + device='cuda:0'), grad: tensor([[ 6.6170e-07, 0.0000e+00, -3.9767e-07, ..., 7.4226e-07, + 3.9488e-07, 5.4389e-07], + [ 2.7921e-06, 0.0000e+00, 6.1886e-07, ..., 1.1856e-06, + 1.4501e-06, 6.2399e-07], + [ 1.5479e-06, 0.0000e+00, -1.6065e-06, ..., 9.0571e-07, + 1.0468e-06, 9.5740e-07], + ..., + [ 1.3411e-06, 0.0000e+00, -5.3048e-06, ..., 5.8487e-07, + -1.7434e-05, 1.5413e-06], + [ 8.3493e-07, 0.0000e+00, 2.0042e-06, ..., 6.7567e-07, + 3.8296e-06, 6.2659e-06], + [ 4.3400e-06, 0.0000e+00, 1.2964e-06, ..., 1.1427e-06, + 2.8349e-06, -2.6710e-06]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0361, -0.0308, 0.0140, 0.0322, -0.0251, 0.0066, -0.0129, 0.0098, + 0.0190, 0.0017], device='cuda:0'), grad: tensor([ 6.7092e-06, 9.4771e-06, 8.0243e-06, -1.0455e-04, -1.9804e-05, + 5.8085e-05, 6.5230e-06, -1.5274e-05, 3.6269e-05, 1.4454e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 214.68, cls_loss 0.0022 cls_loss_mapping 0.0073 cls_loss_causal 0.5828 re_mapping 0.0074 re_causal 0.0243 /// teacc 98.98 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0309, 0.0210, -0.0039, ..., -0.0518, -0.0474, 0.0511], + [-0.0262, -0.0103, -0.0510, ..., 0.1075, -0.0554, -0.0390], + [-0.0112, -0.0221, 0.1428, ..., -0.0706, -0.0977, -0.0558], + ..., + [-0.0380, -0.0447, -0.0080, ..., -0.0013, 0.0712, -0.0232], + [-0.0512, -0.0451, -0.0295, ..., -0.0759, 0.0133, -0.0839], + [-0.1129, -0.0294, -0.0880, ..., -0.0762, 0.0157, 0.0667]], + device='cuda:0'), grad: tensor([[ 7.6229e-07, 0.0000e+00, 4.9956e-06, ..., 7.2597e-07, + 2.4345e-06, 5.0068e-05], + [ 1.6894e-06, 0.0000e+00, 1.2152e-05, ..., -8.4639e-06, + 1.0699e-05, 1.1012e-05], + [ 7.5391e-07, 0.0000e+00, 5.0217e-05, ..., 6.9244e-07, + 4.4316e-05, 6.6906e-06], + ..., + [ 1.9986e-06, 0.0000e+00, 7.1228e-05, ..., 8.3772e-07, + -5.3048e-05, 7.1144e-04], + [ 1.7453e-06, 0.0000e+00, 8.4713e-06, ..., 6.1840e-06, + 7.3649e-06, 9.1046e-06], + [ 1.4916e-05, 0.0000e+00, -1.6570e-04, ..., 2.5183e-06, + -1.6376e-05, -8.8263e-04]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0360, -0.0306, 0.0130, 0.0327, -0.0249, 0.0066, -0.0131, 0.0103, + 0.0189, 0.0012], device='cuda:0'), grad: tensor([ 8.7082e-05, 2.8670e-05, 1.0622e-04, 6.9797e-05, 5.9813e-05, + 1.9103e-05, 1.4037e-05, 1.0824e-03, 4.5270e-05, -1.5135e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.48, cls_loss 0.0024 cls_loss_mapping 0.0079 cls_loss_causal 0.5651 re_mapping 0.0072 re_causal 0.0226 /// teacc 99.06 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0319, 0.0211, -0.0036, ..., -0.0520, -0.0478, 0.0513], + [-0.0267, -0.0103, -0.0515, ..., 0.1082, -0.0557, -0.0391], + [-0.0107, -0.0222, 0.1431, ..., -0.0709, -0.0979, -0.0563], + ..., + [-0.0386, -0.0447, -0.0078, ..., -0.0017, 0.0717, -0.0241], + [-0.0514, -0.0452, -0.0297, ..., -0.0768, 0.0132, -0.0844], + [-0.1146, -0.0293, -0.0881, ..., -0.0764, 0.0153, 0.0671]], + device='cuda:0'), grad: tensor([[ 3.9767e-07, 4.6566e-10, -3.2280e-06, ..., 6.7288e-07, + 1.2126e-06, -2.5015e-06], + [ 1.2582e-06, 9.3132e-10, 3.3667e-07, ..., 1.2107e-07, + 1.1371e-06, 6.5891e-07], + [ 4.1444e-07, 2.6077e-08, 1.8254e-06, ..., 6.0908e-07, + 2.3805e-06, 1.4324e-06], + ..., + [ 4.6659e-07, 4.6566e-10, -4.3921e-06, ..., 3.3993e-07, + -4.2319e-06, -1.5832e-06], + [ 8.0513e-07, 1.3970e-09, 8.8010e-07, ..., 1.4026e-06, + 8.8811e-06, 4.2357e-06], + [ 4.9695e-06, 4.6566e-10, 4.5113e-06, ..., 2.5462e-06, + 4.3735e-06, 1.7416e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0362, -0.0303, 0.0129, 0.0328, -0.0240, 0.0068, -0.0134, 0.0102, + 0.0189, 0.0006], device='cuda:0'), grad: tensor([ 1.4976e-06, 4.8839e-06, 1.6004e-05, 4.2282e-06, -7.8753e-06, + -8.3745e-05, 3.7760e-05, -5.0887e-06, 1.4573e-05, 1.7673e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 214.56, cls_loss 0.0023 cls_loss_mapping 0.0075 cls_loss_causal 0.6185 re_mapping 0.0076 re_causal 0.0239 /// teacc 99.02 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0323, 0.0210, -0.0030, ..., -0.0521, -0.0481, 0.0515], + [-0.0261, -0.0103, -0.0519, ..., 0.1089, -0.0560, -0.0392], + [-0.0109, -0.0220, 0.1434, ..., -0.0713, -0.0981, -0.0567], + ..., + [-0.0392, -0.0447, -0.0072, ..., -0.0019, 0.0722, -0.0246], + [-0.0522, -0.0453, -0.0300, ..., -0.0777, 0.0130, -0.0850], + [-0.1168, -0.0292, -0.0884, ..., -0.0769, 0.0148, 0.0671]], + device='cuda:0'), grad: tensor([[ 3.5204e-07, -1.5333e-05, -2.4319e-04, ..., 1.0598e-06, + 9.1735e-07, -2.5177e-04], + [ 1.5544e-06, 3.6787e-08, 1.4260e-05, ..., 7.0408e-07, + 2.4308e-06, 2.0880e-06], + [ 2.7893e-07, 7.7020e-07, -3.3617e-05, ..., 1.8440e-07, + 1.0477e-06, 6.5267e-05], + ..., + [ 1.0617e-06, 3.6787e-08, 5.4359e-05, ..., -9.8534e-07, + -1.1288e-05, 2.0247e-06], + [ 1.6401e-06, 1.2545e-06, 6.7651e-05, ..., 8.1724e-07, + 2.3358e-06, 5.6118e-05], + [ 6.0610e-06, 9.1642e-06, 4.7326e-05, ..., 1.8999e-06, + 5.6289e-06, 6.0201e-05]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0361, -0.0300, 0.0127, 0.0327, -0.0233, 0.0071, -0.0129, 0.0108, + 0.0184, -0.0004], device='cuda:0'), grad: tensor([-4.6301e-04, 4.5985e-05, -8.0585e-05, 9.6560e-05, 3.1795e-06, + -1.1556e-05, 8.6367e-05, 1.0371e-04, 8.7261e-05, 1.3244e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 214.60, cls_loss 0.0021 cls_loss_mapping 0.0064 cls_loss_causal 0.5883 re_mapping 0.0073 re_causal 0.0238 /// teacc 99.03 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0327, 0.0212, -0.0025, ..., -0.0523, -0.0483, 0.0517], + [-0.0264, -0.0103, -0.0524, ..., 0.1091, -0.0562, -0.0394], + [-0.0110, -0.0218, 0.1440, ..., -0.0716, -0.0981, -0.0571], + ..., + [-0.0400, -0.0447, -0.0074, ..., -0.0020, 0.0727, -0.0247], + [-0.0524, -0.0454, -0.0298, ..., -0.0777, 0.0129, -0.0847], + [-0.1171, -0.0293, -0.0893, ..., -0.0773, 0.0146, 0.0671]], + device='cuda:0'), grad: tensor([[ 1.5087e-07, 0.0000e+00, 9.3551e-07, ..., 1.1632e-06, + 1.3448e-06, -8.6473e-07], + [ 5.3085e-08, 0.0000e+00, 1.4929e-06, ..., -1.3269e-05, + 1.8002e-06, 3.2457e-07], + [ 7.9628e-08, 0.0000e+00, -3.1423e-06, ..., 2.1905e-06, + 3.9227e-06, 4.1537e-07], + ..., + [ 1.8626e-09, 0.0000e+00, -2.2531e-05, ..., 3.6489e-06, + -3.1233e-05, 7.1190e-06], + [ 2.1653e-07, 0.0000e+00, 1.8086e-06, ..., 2.9355e-06, + 4.6417e-06, 1.7118e-06], + [ 1.1176e-08, 0.0000e+00, 1.6049e-05, ..., 6.6077e-07, + 1.7464e-05, -1.2137e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0361, -0.0303, 0.0129, 0.0325, -0.0235, 0.0070, -0.0124, 0.0109, + 0.0191, -0.0010], device='cuda:0'), grad: tensor([ 7.3053e-06, -1.3642e-05, 1.1750e-05, 4.7415e-05, 9.4026e-06, + -2.3946e-05, -3.3453e-06, -5.0515e-05, -1.6317e-05, 3.1739e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 214.51, cls_loss 0.0018 cls_loss_mapping 0.0066 cls_loss_causal 0.5784 re_mapping 0.0078 re_causal 0.0238 /// teacc 99.11 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0331, 0.0212, -0.0018, ..., -0.0525, -0.0485, 0.0519], + [-0.0265, -0.0104, -0.0529, ..., 0.1095, -0.0565, -0.0395], + [-0.0111, -0.0218, 0.1444, ..., -0.0718, -0.0982, -0.0574], + ..., + [-0.0405, -0.0447, -0.0073, ..., -0.0022, 0.0732, -0.0252], + [-0.0533, -0.0456, -0.0300, ..., -0.0781, 0.0125, -0.0859], + [-0.1171, -0.0293, -0.0898, ..., -0.0776, 0.0144, 0.0679]], + device='cuda:0'), grad: tensor([[ 1.4305e-06, 0.0000e+00, -4.4629e-06, ..., -8.0513e-07, + -3.9767e-07, -1.0647e-05], + [ 9.9931e-07, 0.0000e+00, 1.2815e-05, ..., 2.5809e-05, + 4.4137e-05, 7.1600e-06], + [ 1.8440e-07, 0.0000e+00, -2.2277e-05, ..., 9.2536e-06, + 1.4283e-05, 1.6149e-06], + ..., + [ 1.8952e-07, 0.0000e+00, -1.1913e-05, ..., -4.4197e-05, + -8.1658e-05, 1.7986e-05], + [-8.1360e-06, 0.0000e+00, 3.7272e-06, ..., 8.5728e-07, + -3.5223e-06, -2.5779e-06], + [ 6.4299e-06, 0.0000e+00, 3.7588e-06, ..., 2.3562e-06, + 6.9737e-06, -4.9770e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0360, -0.0303, 0.0129, 0.0324, -0.0237, 0.0072, -0.0122, 0.0109, + 0.0186, -0.0005], device='cuda:0'), grad: tensor([-4.2468e-06, 3.2783e-04, 8.4758e-05, 9.3639e-05, 9.3818e-05, + 3.6150e-05, -1.8673e-06, -5.4264e-04, -2.3454e-05, -6.3956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 214.61, cls_loss 0.0017 cls_loss_mapping 0.0058 cls_loss_causal 0.5377 re_mapping 0.0079 re_causal 0.0233 /// teacc 99.07 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0340, 0.0212, -0.0010, ..., -0.0530, -0.0491, 0.0522], + [-0.0257, -0.0104, -0.0533, ..., 0.1101, -0.0569, -0.0396], + [-0.0113, -0.0218, 0.1447, ..., -0.0721, -0.0984, -0.0577], + ..., + [-0.0410, -0.0446, -0.0071, ..., -0.0024, 0.0741, -0.0254], + [-0.0541, -0.0457, -0.0301, ..., -0.0785, 0.0123, -0.0864], + [-0.1177, -0.0292, -0.0906, ..., -0.0781, 0.0141, 0.0680]], + device='cuda:0'), grad: tensor([[ 1.5507e-07, 3.4459e-08, 1.0021e-05, ..., 9.4064e-08, + 1.6354e-06, 2.2296e-06], + [ 1.4715e-07, 1.9558e-08, 3.2671e-06, ..., -1.0449e-06, + 1.6419e-06, 1.0449e-06], + [ 5.0291e-08, -1.1176e-07, -6.7663e-04, ..., 7.5437e-08, + 2.4855e-05, -8.1956e-05], + ..., + [ 7.8231e-08, 6.0536e-09, 1.0893e-05, ..., 1.9511e-07, + 1.8515e-06, 3.1255e-06], + [ 6.8452e-08, 1.5367e-08, -7.4320e-06, ..., 4.2608e-07, + -4.7535e-05, -2.0519e-05], + [ 2.5192e-07, 4.6566e-10, 6.2656e-04, ..., 1.3737e-07, + 4.3996e-06, 9.0420e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0360, -0.0302, 0.0130, 0.0323, -0.0237, 0.0071, -0.0116, 0.0112, + 0.0185, -0.0008], device='cuda:0'), grad: tensor([ 2.4170e-05, 1.5453e-05, -4.0889e-04, 2.3410e-05, 1.5825e-05, + 9.5427e-05, 1.7107e-05, 4.8757e-05, -4.6945e-04, 6.3848e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 214.30, cls_loss 0.0025 cls_loss_mapping 0.0065 cls_loss_causal 0.5995 re_mapping 0.0074 re_causal 0.0236 /// teacc 98.91 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0344, 0.0212, -0.0008, ..., -0.0532, -0.0493, 0.0524], + [-0.0265, -0.0104, -0.0554, ..., 0.1103, -0.0579, -0.0396], + [-0.0099, -0.0218, 0.1457, ..., -0.0721, -0.0985, -0.0577], + ..., + [-0.0411, -0.0447, -0.0066, ..., -0.0017, 0.0752, -0.0260], + [-0.0546, -0.0457, -0.0302, ..., -0.0791, 0.0118, -0.0869], + [-0.1175, -0.0291, -0.0912, ..., -0.0785, 0.0137, 0.0686]], + device='cuda:0'), grad: tensor([[ 1.1139e-06, 0.0000e+00, 3.3110e-05, ..., 1.5544e-06, + 4.1753e-05, 1.6034e-05], + [ 2.6841e-06, 0.0000e+00, 6.3255e-06, ..., -4.6976e-06, + 7.0520e-06, 3.1404e-06], + [ 9.9838e-06, 0.0000e+00, 4.2617e-05, ..., 3.2187e-06, + 1.0937e-05, 3.3587e-05], + ..., + [ 3.5726e-06, 0.0000e+00, 8.8573e-05, ..., 1.3262e-06, + -1.3041e-04, 1.1134e-04], + [-1.1072e-05, 0.0000e+00, 1.1191e-05, ..., 1.4845e-06, + 5.4426e-06, 7.7188e-06], + [ 3.5800e-06, 0.0000e+00, -2.5105e-04, ..., 6.5332e-07, + 2.2084e-05, -2.1529e-04]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0360, -0.0310, 0.0136, 0.0319, -0.0247, 0.0079, -0.0121, 0.0122, + 0.0185, -0.0006], device='cuda:0'), grad: tensor([ 1.0633e-04, 1.4029e-05, 2.2495e-04, 7.5340e-05, 1.2517e-05, + 6.3479e-05, 2.5302e-05, 3.8952e-05, -1.6856e-04, -3.9291e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 214.41, cls_loss 0.0034 cls_loss_mapping 0.0083 cls_loss_causal 0.5666 re_mapping 0.0070 re_causal 0.0231 /// teacc 99.04 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0353, 0.0226, -0.0007, ..., -0.0536, -0.0500, 0.0525], + [-0.0263, -0.0104, -0.0556, ..., 0.1110, -0.0582, -0.0397], + [-0.0102, -0.0217, 0.1464, ..., -0.0727, -0.0988, -0.0583], + ..., + [-0.0419, -0.0472, -0.0088, ..., -0.0020, 0.0742, -0.0294], + [-0.0549, -0.0462, -0.0306, ..., -0.0795, 0.0112, -0.0876], + [-0.1175, -0.0304, -0.0900, ..., -0.0789, 0.0137, 0.0700]], + device='cuda:0'), grad: tensor([[ 1.8161e-08, 0.0000e+00, 2.7612e-05, ..., 1.3504e-08, + 2.6403e-07, 1.1921e-07], + [ 3.1199e-08, 0.0000e+00, 6.3926e-06, ..., -1.0896e-07, + 3.5837e-06, 1.6624e-07], + [ 9.4064e-08, 0.0000e+00, -1.2302e-04, ..., 1.2992e-07, + 1.8114e-06, 1.3039e-07], + ..., + [ 1.0710e-08, 0.0000e+00, -5.8413e-06, ..., 1.0245e-08, + -1.7583e-05, 4.1677e-07], + [ 5.4482e-08, 0.0000e+00, 1.2249e-05, ..., 1.2666e-07, + 1.2733e-05, 1.3284e-05], + [ 1.0524e-07, 0.0000e+00, 3.1982e-06, ..., 1.7695e-08, + 3.5502e-06, -4.9658e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0361, -0.0307, 0.0136, 0.0336, -0.0253, 0.0077, -0.0116, 0.0098, + 0.0182, 0.0004], device='cuda:0'), grad: tensor([ 4.6939e-05, 2.8864e-05, -1.8680e-04, 1.3196e-04, 9.5069e-06, + -2.0057e-05, 1.3158e-05, -8.3327e-05, 4.8995e-05, 1.0632e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 214.36, cls_loss 0.0021 cls_loss_mapping 0.0063 cls_loss_causal 0.5247 re_mapping 0.0070 re_causal 0.0218 /// teacc 99.04 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0359, 0.0232, -0.0002, ..., -0.0537, -0.0502, 0.0526], + [-0.0266, -0.0105, -0.0560, ..., 0.1112, -0.0586, -0.0398], + [-0.0109, -0.0217, 0.1468, ..., -0.0734, -0.0990, -0.0587], + ..., + [-0.0409, -0.0478, -0.0082, ..., -0.0020, 0.0744, -0.0298], + [-0.0552, -0.0464, -0.0308, ..., -0.0798, 0.0110, -0.0881], + [-0.1181, -0.0309, -0.0905, ..., -0.0793, 0.0131, 0.0704]], + device='cuda:0'), grad: tensor([[ 4.2431e-06, 0.0000e+00, 2.2557e-06, ..., 1.0477e-07, + -4.0606e-06, -1.1750e-05], + [ 3.6228e-07, 0.0000e+00, 8.2776e-06, ..., -1.1297e-06, + 1.5302e-06, 1.6317e-06], + [ 9.0618e-07, 0.0000e+00, -5.5581e-05, ..., 2.3609e-07, + 8.6008e-07, 1.4091e-06], + ..., + [ 2.6450e-07, 0.0000e+00, -6.8963e-05, ..., 7.6834e-08, + -2.0936e-05, -1.6779e-05], + [-1.2539e-05, 0.0000e+00, 1.9088e-05, ..., 2.5611e-07, + 1.0226e-06, -3.8333e-06], + [-1.1079e-05, 0.0000e+00, 6.3241e-05, ..., 5.7742e-08, + 1.7613e-05, -1.6600e-05]], device='cuda:0') +Epoch 109, bias, value: tensor([-3.6206e-02, -3.0941e-02, 1.3253e-02, 3.3824e-02, -2.6164e-02, + 7.6395e-03, -1.1116e-02, 1.0606e-02, 1.8092e-02, 3.6727e-05], + device='cuda:0'), grad: tensor([ 7.9535e-07, 1.7911e-05, -7.5161e-05, 4.3571e-05, 6.3777e-05, + 1.1355e-05, 1.9655e-05, -1.7166e-04, -2.6673e-05, 1.1623e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 214.51, cls_loss 0.0019 cls_loss_mapping 0.0062 cls_loss_causal 0.5641 re_mapping 0.0073 re_causal 0.0227 /// teacc 99.04 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0368, 0.0230, 0.0002, ..., -0.0537, -0.0505, 0.0527], + [-0.0270, -0.0105, -0.0557, ..., 0.1119, -0.0587, -0.0401], + [-0.0113, -0.0217, 0.1473, ..., -0.0740, -0.0992, -0.0590], + ..., + [-0.0422, -0.0478, -0.0083, ..., -0.0025, 0.0745, -0.0300], + [-0.0547, -0.0466, -0.0309, ..., -0.0801, 0.0111, -0.0884], + [-0.1189, -0.0311, -0.0907, ..., -0.0796, 0.0127, 0.0707]], + device='cuda:0'), grad: tensor([[ 3.5902e-07, 6.9849e-10, -9.3831e-07, ..., 3.4226e-07, + 4.6380e-07, 2.6776e-07], + [ 2.7032e-07, 1.1642e-09, 1.3784e-06, ..., -8.1398e-07, + 6.3982e-07, 4.6752e-07], + [ 3.1944e-07, 1.1642e-09, -1.4827e-05, ..., 3.8301e-07, + 9.3086e-07, 1.1884e-06], + ..., + [ 2.5379e-08, 3.2596e-09, 5.3644e-06, ..., 1.4761e-07, + -3.9674e-06, 6.3796e-08], + [ 5.0059e-08, 1.6298e-09, 1.3923e-06, ..., 6.7288e-07, + 5.2042e-06, 4.4070e-06], + [-5.9465e-07, 3.2596e-09, 1.3663e-06, ..., 7.5204e-08, + 1.1306e-06, -9.1940e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0363, -0.0304, 0.0132, 0.0338, -0.0257, 0.0077, -0.0112, 0.0103, + 0.0182, -0.0003], device='cuda:0'), grad: tensor([ 2.0768e-06, 2.3358e-06, -1.0245e-05, 1.0408e-05, 4.9397e-06, + -1.1712e-05, 1.5702e-06, 8.6892e-07, 1.0766e-05, -1.1079e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 214.37, cls_loss 0.0021 cls_loss_mapping 0.0065 cls_loss_causal 0.5559 re_mapping 0.0074 re_causal 0.0226 /// teacc 99.03 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0379, 0.0230, 0.0008, ..., -0.0539, -0.0510, 0.0531], + [-0.0256, -0.0105, -0.0561, ..., 0.1128, -0.0591, -0.0402], + [-0.0117, -0.0217, 0.1479, ..., -0.0743, -0.0994, -0.0594], + ..., + [-0.0425, -0.0478, -0.0079, ..., -0.0026, 0.0751, -0.0302], + [-0.0551, -0.0473, -0.0312, ..., -0.0808, 0.0109, -0.0889], + [-0.1212, -0.0314, -0.0915, ..., -0.0798, 0.0119, 0.0707]], + device='cuda:0'), grad: tensor([[ 2.1886e-07, 0.0000e+00, 5.8748e-06, ..., 8.1444e-07, + 2.1141e-06, -5.1921e-08], + [ 6.5705e-07, 0.0000e+00, 4.2796e-05, ..., -4.0904e-06, + 2.8387e-06, 8.9733e-07], + [ 5.3411e-07, 0.0000e+00, -1.1307e-04, ..., 1.2545e-06, + 2.0619e-06, 6.1700e-07], + ..., + [ 4.8755e-07, 0.0000e+00, 2.4170e-05, ..., 1.4296e-06, + -1.1176e-05, 1.0086e-06], + [ 9.0338e-07, 0.0000e+00, 2.8551e-05, ..., 3.0473e-06, + 1.1064e-05, 4.8019e-06], + [-1.7630e-06, 0.0000e+00, 1.0049e-06, ..., 4.7777e-07, + 2.7157e-06, -9.3505e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0362, -0.0300, 0.0132, 0.0336, -0.0248, 0.0080, -0.0118, 0.0107, + 0.0180, -0.0014], device='cuda:0'), grad: tensor([ 1.9789e-05, 8.7976e-05, -2.3341e-04, 2.8729e-05, 1.8016e-05, + -7.2539e-05, 2.6554e-05, 4.1544e-05, 8.7261e-05, -3.8855e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 110---------------------------------------------------- +epoch 110, time 230.57, cls_loss 0.0027 cls_loss_mapping 0.0064 cls_loss_causal 0.5615 re_mapping 0.0071 re_causal 0.0219 /// teacc 99.14 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0384, 0.0240, 0.0022, ..., -0.0539, -0.0497, 0.0530], + [-0.0271, -0.0105, -0.0570, ..., 0.1128, -0.0595, -0.0404], + [-0.0118, -0.0216, 0.1486, ..., -0.0745, -0.0996, -0.0602], + ..., + [-0.0436, -0.0478, -0.0079, ..., -0.0028, 0.0756, -0.0307], + [-0.0556, -0.0475, -0.0313, ..., -0.0813, 0.0104, -0.0894], + [-0.1218, -0.0317, -0.0919, ..., -0.0800, 0.0118, 0.0725]], + device='cuda:0'), grad: tensor([[ 1.3737e-07, 0.0000e+00, 1.1059e-07, ..., 1.6182e-07, + 5.3924e-07, 2.1188e-08], + [ 3.3248e-07, 0.0000e+00, 1.2226e-05, ..., -1.1232e-06, + 1.8299e-05, 2.1188e-07], + [ 1.9255e-07, 0.0000e+00, -1.1828e-06, ..., 1.9721e-07, + 8.7777e-07, 1.9255e-07], + ..., + [ 1.0692e-06, 0.0000e+00, -1.7941e-05, ..., 4.1886e-07, + -3.1322e-05, -1.1977e-06], + [ 1.1479e-07, 0.0000e+00, 7.4785e-07, ..., 3.0873e-07, + 1.6838e-06, 1.6773e-06], + [-4.7917e-07, 0.0000e+00, 2.5779e-06, ..., 8.7079e-08, + 5.4277e-06, -3.2373e-06]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0359, -0.0304, 0.0134, 0.0333, -0.0244, 0.0072, -0.0114, 0.0107, + 0.0179, -0.0007], device='cuda:0'), grad: tensor([ 2.3283e-06, 5.3197e-05, 1.2200e-06, 7.1637e-06, 5.2974e-06, + -4.6901e-06, 1.2489e-06, -7.8917e-05, 8.2254e-06, 4.8056e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 214.79, cls_loss 0.0019 cls_loss_mapping 0.0058 cls_loss_causal 0.5630 re_mapping 0.0069 re_causal 0.0214 /// teacc 98.95 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0391, 0.0240, 0.0036, ..., -0.0535, -0.0496, 0.0536], + [-0.0281, -0.0105, -0.0575, ..., 0.1133, -0.0601, -0.0405], + [-0.0117, -0.0215, 0.1493, ..., -0.0747, -0.0997, -0.0611], + ..., + [-0.0443, -0.0478, -0.0078, ..., -0.0031, 0.0760, -0.0309], + [-0.0555, -0.0476, -0.0319, ..., -0.0820, 0.0102, -0.0898], + [-0.1220, -0.0317, -0.0925, ..., -0.0807, 0.0115, 0.0727]], + device='cuda:0'), grad: tensor([[ 1.0774e-05, 0.0000e+00, 3.6694e-06, ..., 4.8708e-07, + 3.5856e-07, -2.6543e-08], + [ 1.0701e-06, 0.0000e+00, 2.6211e-05, ..., -8.0001e-07, + 1.3020e-06, 3.3225e-07], + [ 2.9523e-06, 0.0000e+00, -3.1590e-04, ..., -4.6305e-06, + 1.7667e-06, 1.2200e-07], + ..., + [ 7.6322e-07, 0.0000e+00, -3.6974e-06, ..., 2.0233e-07, + -6.8285e-06, 1.1153e-07], + [-1.9580e-05, 0.0000e+00, 2.6393e-04, ..., 4.8243e-06, + 9.5926e-07, 9.3319e-07], + [ 1.3918e-05, 0.0000e+00, 6.4075e-06, ..., 1.3364e-07, + 2.0806e-06, -5.5842e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0351, -0.0306, 0.0137, 0.0331, -0.0246, 0.0068, -0.0115, 0.0109, + 0.0177, -0.0009], device='cuda:0'), grad: tensor([ 5.6446e-05, 4.4674e-05, -4.9734e-04, 2.4930e-05, 3.4943e-06, + 2.6799e-07, 1.0759e-05, -8.5831e-06, 3.4237e-04, 2.3752e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 214.53, cls_loss 0.0018 cls_loss_mapping 0.0053 cls_loss_causal 0.5786 re_mapping 0.0068 re_causal 0.0226 /// teacc 99.09 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0397, 0.0240, 0.0038, ..., -0.0537, -0.0500, 0.0537], + [-0.0286, -0.0106, -0.0580, ..., 0.1138, -0.0605, -0.0405], + [-0.0115, -0.0215, 0.1500, ..., -0.0751, -0.0998, -0.0615], + ..., + [-0.0444, -0.0478, -0.0076, ..., -0.0035, 0.0763, -0.0313], + [-0.0557, -0.0476, -0.0324, ..., -0.0818, 0.0100, -0.0902], + [-0.1219, -0.0317, -0.0929, ..., -0.0808, 0.0112, 0.0734]], + device='cuda:0'), grad: tensor([[ 8.6846e-07, -8.6846e-08, -1.6168e-05, ..., 7.7486e-07, + 5.2946e-07, -3.6448e-05], + [ 2.0340e-06, 2.3283e-10, 5.7742e-06, ..., -1.3495e-06, + 4.7544e-07, 4.4354e-07], + [ 1.4797e-05, 1.6298e-09, -8.6606e-05, ..., 2.2184e-06, + 1.8533e-07, 1.1083e-06], + ..., + [ 9.1176e-07, 4.6566e-10, 3.7193e-05, ..., 2.2841e-07, + -2.0489e-05, 1.0170e-06], + [-2.5276e-06, 4.8894e-09, 1.9670e-05, ..., 1.6969e-06, + 1.1306e-06, 4.0941e-06], + [-7.5810e-06, 5.7509e-08, 1.3351e-05, ..., 1.1595e-07, + 2.1011e-06, -5.0068e-05]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0352, -0.0305, 0.0139, 0.0330, -0.0252, 0.0066, -0.0120, 0.0110, + 0.0180, -0.0006], device='cuda:0'), grad: tensor([-5.2482e-05, 1.1146e-05, -1.0812e-04, 5.9992e-05, 6.7949e-05, + -2.3134e-06, 2.2538e-06, 3.8952e-05, 1.4700e-05, -3.2157e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 214.59, cls_loss 0.0020 cls_loss_mapping 0.0058 cls_loss_causal 0.5537 re_mapping 0.0068 re_causal 0.0216 /// teacc 99.07 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0394, 0.0239, 0.0046, ..., -0.0536, -0.0506, 0.0538], + [-0.0291, -0.0106, -0.0601, ..., 0.1141, -0.0629, -0.0406], + [-0.0115, -0.0215, 0.1504, ..., -0.0754, -0.1001, -0.0619], + ..., + [-0.0448, -0.0478, -0.0068, ..., -0.0033, 0.0774, -0.0314], + [-0.0560, -0.0476, -0.0327, ..., -0.0826, 0.0090, -0.0909], + [-0.1224, -0.0316, -0.0932, ..., -0.0814, 0.0110, 0.0739]], + device='cuda:0'), grad: tensor([[ 3.1292e-07, 0.0000e+00, -1.0252e-05, ..., -2.0657e-06, + 2.5402e-07, -1.2591e-05], + [ 1.1716e-06, 0.0000e+00, 4.9779e-07, ..., -1.9111e-06, + 3.2503e-07, 7.8091e-07], + [ 2.6450e-07, 0.0000e+00, 9.7416e-07, ..., 2.9569e-07, + 2.2538e-07, 2.1979e-06], + ..., + [ 3.9954e-07, 0.0000e+00, 5.7276e-08, ..., 1.4910e-06, + 4.1723e-07, 5.9381e-06], + [ 2.4145e-07, 0.0000e+00, 9.5647e-07, ..., 3.0012e-07, + 1.7565e-06, 3.3174e-06], + [ 8.0559e-07, 0.0000e+00, 2.7604e-06, ..., 6.7288e-07, + -1.1362e-06, -8.9258e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0350, -0.0321, 0.0139, 0.0328, -0.0250, 0.0064, -0.0121, 0.0124, + 0.0177, -0.0005], device='cuda:0'), grad: tensor([-2.2531e-05, -9.7230e-07, 3.7719e-06, 1.1481e-05, -1.1083e-06, + -1.1832e-05, 1.2249e-05, 1.3143e-05, 7.4394e-06, -1.1660e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 214.25, cls_loss 0.0018 cls_loss_mapping 0.0058 cls_loss_causal 0.5629 re_mapping 0.0066 re_causal 0.0214 /// teacc 99.12 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0406, 0.0239, 0.0047, ..., -0.0543, -0.0508, 0.0538], + [-0.0298, -0.0106, -0.0607, ..., 0.1148, -0.0620, -0.0407], + [-0.0116, -0.0215, 0.1512, ..., -0.0755, -0.1004, -0.0623], + ..., + [-0.0452, -0.0478, -0.0064, ..., -0.0046, 0.0775, -0.0317], + [-0.0561, -0.0476, -0.0330, ..., -0.0830, 0.0088, -0.0914], + [-0.1226, -0.0316, -0.0936, ..., -0.0817, 0.0106, 0.0741]], + device='cuda:0'), grad: tensor([[ 2.2305e-07, 0.0000e+00, -1.4789e-06, ..., 1.0207e-06, + 2.2370e-06, 5.3272e-07], + [ 2.0023e-08, 0.0000e+00, 2.7865e-06, ..., -6.0387e-06, + 2.1551e-06, 7.2084e-07], + [ 3.3993e-08, 0.0000e+00, 1.5825e-05, ..., 2.6030e-07, + 1.8775e-05, 1.8477e-06], + ..., + [ 2.4680e-08, 0.0000e+00, -3.6627e-05, ..., 2.3898e-06, + -2.2069e-05, 1.2880e-06], + [ 9.3132e-08, 0.0000e+00, 6.6198e-06, ..., 7.5158e-07, + 4.5709e-06, 7.0110e-06], + [ 8.6613e-08, 0.0000e+00, 3.5577e-06, ..., 5.2620e-07, + 1.0431e-05, 6.2250e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0354, -0.0313, 0.0143, 0.0331, -0.0250, 0.0059, -0.0118, 0.0118, + 0.0176, -0.0006], device='cuda:0'), grad: tensor([ 5.0776e-06, -3.0436e-06, 1.5333e-05, 8.4877e-05, 6.4373e-06, + -1.2422e-04, 2.9672e-06, -3.0413e-05, -1.8314e-05, 6.1274e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 214.21, cls_loss 0.0021 cls_loss_mapping 0.0059 cls_loss_causal 0.5699 re_mapping 0.0064 re_causal 0.0216 /// teacc 99.02 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0410, 0.0239, 0.0053, ..., -0.0544, -0.0513, 0.0540], + [-0.0302, -0.0106, -0.0583, ..., 0.1163, -0.0625, -0.0409], + [-0.0117, -0.0214, 0.1513, ..., -0.0780, -0.1005, -0.0626], + ..., + [-0.0467, -0.0478, -0.0067, ..., -0.0050, 0.0778, -0.0319], + [-0.0564, -0.0477, -0.0335, ..., -0.0832, 0.0084, -0.0921], + [-0.1229, -0.0316, -0.0944, ..., -0.0819, 0.0099, 0.0745]], + device='cuda:0'), grad: tensor([[ 2.0256e-07, 0.0000e+00, -1.7090e-06, ..., 5.5414e-07, + 4.2422e-07, -1.4035e-06], + [ 7.4040e-08, 0.0000e+00, 4.5411e-06, ..., -2.6301e-06, + 3.3844e-06, 3.1479e-07], + [ 5.5414e-08, 0.0000e+00, -1.1615e-05, ..., 2.8405e-07, + 5.6773e-06, 5.5972e-07], + ..., + [ 6.0536e-09, 0.0000e+00, -2.0176e-05, ..., 1.2591e-06, + -2.0772e-05, 1.7416e-06], + [ 7.0734e-07, 0.0000e+00, 1.4029e-05, ..., 1.2675e-06, + 7.2569e-06, 2.3723e-05], + [-7.4552e-07, 0.0000e+00, 3.2615e-06, ..., 6.1467e-08, + 2.2594e-06, -2.6077e-05]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0353, -0.0303, 0.0136, 0.0330, -0.0247, 0.0062, -0.0120, 0.0116, + 0.0174, -0.0010], device='cuda:0'), grad: tensor([ 9.6858e-07, 3.4198e-06, -5.0999e-06, 2.0862e-06, 7.1004e-06, + 4.5784e-06, -1.4231e-06, -3.8654e-05, 6.4194e-05, -3.7283e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 214.31, cls_loss 0.0016 cls_loss_mapping 0.0060 cls_loss_causal 0.5764 re_mapping 0.0066 re_causal 0.0220 /// teacc 99.00 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0415, 0.0238, 0.0054, ..., -0.0546, -0.0515, 0.0541], + [-0.0308, -0.0106, -0.0593, ..., 0.1166, -0.0626, -0.0409], + [-0.0112, -0.0213, 0.1523, ..., -0.0778, -0.1007, -0.0629], + ..., + [-0.0474, -0.0478, -0.0067, ..., -0.0052, 0.0780, -0.0322], + [-0.0568, -0.0477, -0.0338, ..., -0.0836, 0.0082, -0.0921], + [-0.1238, -0.0315, -0.0946, ..., -0.0822, 0.0096, 0.0745]], + device='cuda:0'), grad: tensor([[ 1.0105e-07, 0.0000e+00, 2.4587e-07, ..., 1.1185e-06, + 3.4040e-07, 2.5127e-06], + [ 4.6147e-07, 0.0000e+00, 5.8394e-07, ..., -1.1427e-06, + 1.8524e-06, 3.6089e-07], + [ 1.3271e-07, 0.0000e+00, -2.1746e-07, ..., 9.5647e-07, + 5.1456e-07, 4.4284e-07], + ..., + [ 4.4797e-07, 0.0000e+00, -4.7609e-06, ..., 5.8208e-07, + -1.2554e-05, 3.8091e-06], + [ 3.2503e-07, 0.0000e+00, 8.9314e-07, ..., 9.8422e-06, + 1.6559e-06, 8.6334e-07], + [ 2.1048e-06, 0.0000e+00, 4.5309e-07, ..., 1.4668e-07, + 3.1404e-06, -2.0206e-05]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0354, -0.0306, 0.0146, 0.0330, -0.0242, 0.0065, -0.0122, 0.0114, + 0.0175, -0.0016], device='cuda:0'), grad: tensor([ 8.7023e-06, 5.8562e-06, 3.5390e-06, 1.6809e-05, 1.8373e-05, + 1.3173e-05, -3.6001e-05, -1.4238e-05, 2.1160e-05, -3.7432e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 214.23, cls_loss 0.0024 cls_loss_mapping 0.0081 cls_loss_causal 0.5666 re_mapping 0.0070 re_causal 0.0209 /// teacc 99.08 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0418, 0.0242, 0.0057, ..., -0.0544, -0.0519, 0.0542], + [-0.0308, -0.0106, -0.0598, ..., 0.1172, -0.0630, -0.0411], + [-0.0113, -0.0213, 0.1536, ..., -0.0780, -0.1008, -0.0638], + ..., + [-0.0481, -0.0478, -0.0071, ..., -0.0053, 0.0784, -0.0324], + [-0.0570, -0.0478, -0.0341, ..., -0.0839, 0.0079, -0.0931], + [-0.1242, -0.0315, -0.0949, ..., -0.0825, 0.0095, 0.0754]], + device='cuda:0'), grad: tensor([[ 4.5076e-07, 0.0000e+00, -1.7751e-06, ..., 1.0617e-06, + 9.6206e-07, -9.1456e-07], + [ 6.7567e-07, 0.0000e+00, 5.8711e-06, ..., -3.0641e-06, + 5.4166e-06, 6.7316e-06], + [ 6.4075e-07, 0.0000e+00, 4.7274e-06, ..., 1.7332e-06, + 5.0776e-06, 2.3916e-06], + ..., + [ 1.2331e-06, 0.0000e+00, -2.0012e-05, ..., 9.0292e-07, + -2.4110e-05, 1.0170e-05], + [ 4.1956e-07, 0.0000e+00, 3.1870e-06, ..., 1.0300e-06, + 1.6857e-06, 2.4121e-06], + [ 1.2517e-06, 0.0000e+00, 7.5670e-07, ..., 1.3830e-07, + 5.5917e-06, -2.4974e-05]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0355, -0.0306, 0.0152, 0.0327, -0.0249, 0.0062, -0.0131, 0.0115, + 0.0179, -0.0009], device='cuda:0'), grad: tensor([ 5.2378e-06, 3.6329e-05, 2.6435e-05, 2.7418e-05, 3.5334e-06, + -5.2713e-06, -3.1888e-06, -5.4091e-05, 1.6108e-05, -5.2482e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 214.40, cls_loss 0.0015 cls_loss_mapping 0.0048 cls_loss_causal 0.5370 re_mapping 0.0068 re_causal 0.0212 /// teacc 99.10 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0422, 0.0242, 0.0060, ..., -0.0545, -0.0522, 0.0543], + [-0.0311, -0.0106, -0.0600, ..., 0.1175, -0.0631, -0.0411], + [-0.0116, -0.0213, 0.1546, ..., -0.0783, -0.1014, -0.0624], + ..., + [-0.0488, -0.0478, -0.0064, ..., -0.0057, 0.0789, -0.0325], + [-0.0573, -0.0478, -0.0342, ..., -0.0844, 0.0076, -0.0935], + [-0.1245, -0.0315, -0.0971, ..., -0.0828, 0.0092, 0.0754]], + device='cuda:0'), grad: tensor([[ 4.7823e-07, 0.0000e+00, 2.7958e-06, ..., 2.3562e-07, + 3.5241e-06, 8.6147e-08], + [ 3.2429e-06, 0.0000e+00, 2.3060e-06, ..., 1.2275e-06, + 2.6748e-06, 3.2177e-07], + [ 1.2750e-06, 0.0000e+00, 1.4836e-06, ..., 7.5763e-07, + 2.2538e-06, 5.2527e-07], + ..., + [ 1.4231e-06, 0.0000e+00, -1.5318e-05, ..., 5.6112e-07, + -1.8016e-05, -6.8080e-07], + [ 7.2177e-07, 0.0000e+00, 8.4564e-07, ..., 3.4226e-07, + 9.7696e-07, 2.4354e-07], + [ 2.0191e-06, 0.0000e+00, 1.9856e-06, ..., 4.7358e-07, + 2.2165e-06, 3.1805e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0355, -0.0305, 0.0158, 0.0324, -0.0247, 0.0064, -0.0129, 0.0117, + 0.0179, -0.0016], device='cuda:0'), grad: tensor([ 6.5938e-06, 9.4920e-06, 8.7321e-06, 2.7269e-06, -2.5854e-05, + 1.3612e-05, 1.6615e-05, -2.5630e-05, -1.3955e-05, 7.6741e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 214.51, cls_loss 0.0018 cls_loss_mapping 0.0055 cls_loss_causal 0.5706 re_mapping 0.0067 re_causal 0.0219 /// teacc 99.03 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0424, 0.0242, 0.0071, ..., -0.0546, -0.0514, 0.0544], + [-0.0325, -0.0106, -0.0601, ..., 0.1178, -0.0636, -0.0414], + [-0.0114, -0.0212, 0.1550, ..., -0.0789, -0.1017, -0.0630], + ..., + [-0.0495, -0.0478, -0.0063, ..., -0.0057, 0.0794, -0.0329], + [-0.0580, -0.0478, -0.0345, ..., -0.0848, 0.0074, -0.0939], + [-0.1255, -0.0315, -0.0972, ..., -0.0835, 0.0085, 0.0761]], + device='cuda:0'), grad: tensor([[ 2.7474e-08, 0.0000e+00, 1.4110e-07, ..., 1.1595e-07, + 3.3528e-07, 9.0152e-07], + [ 7.0315e-08, 0.0000e+00, 6.9290e-07, ..., -3.3248e-06, + 1.6335e-06, 2.4438e-06], + [ 3.0268e-08, 0.0000e+00, -1.3135e-05, ..., 5.0571e-07, + 4.5775e-07, 9.6858e-07], + ..., + [ 3.9116e-08, 0.0000e+00, 2.7250e-06, ..., 1.0738e-06, + -4.6164e-05, -1.0997e-05], + [ 5.8673e-08, 0.0000e+00, 6.8210e-06, ..., 5.5321e-07, + 5.7789e-07, 2.1141e-06], + [ 2.5099e-07, 0.0000e+00, -4.1397e-07, ..., 1.7602e-07, + 3.9749e-06, -2.1115e-05]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0353, -0.0307, 0.0157, 0.0326, -0.0243, 0.0057, -0.0128, 0.0118, + 0.0177, -0.0016], device='cuda:0'), grad: tensor([ 2.8256e-06, 2.7139e-06, -1.2398e-05, 9.5546e-05, 3.9972e-06, + 3.9525e-06, -4.1053e-06, -7.1943e-05, 1.2808e-05, -3.3259e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 214.69, cls_loss 0.0025 cls_loss_mapping 0.0059 cls_loss_causal 0.5991 re_mapping 0.0067 re_causal 0.0217 /// teacc 99.01 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0430, 0.0247, 0.0072, ..., -0.0547, -0.0521, 0.0543], + [-0.0336, -0.0106, -0.0603, ..., 0.1181, -0.0638, -0.0417], + [-0.0118, -0.0212, 0.1554, ..., -0.0791, -0.1020, -0.0634], + ..., + [-0.0500, -0.0478, -0.0050, ..., -0.0059, 0.0820, -0.0301], + [-0.0585, -0.0480, -0.0347, ..., -0.0852, 0.0075, -0.0943], + [-0.1258, -0.0315, -0.0992, ..., -0.0839, 0.0056, 0.0740]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -2.6096e-06, ..., 2.4168e-07, + 1.0384e-07, -2.8238e-06], + [ 1.0245e-08, 0.0000e+00, 1.2433e-07, ..., -3.9348e-07, + 2.0023e-07, 2.9849e-07], + [ 6.0536e-09, 0.0000e+00, -1.6866e-06, ..., 8.1025e-08, + 2.4680e-07, 1.0384e-06], + ..., + [ 1.5367e-08, 0.0000e+00, -2.2585e-07, ..., 1.0664e-07, + -4.5775e-07, 5.6950e-07], + [ 8.3819e-09, 0.0000e+00, 2.1048e-06, ..., 8.8336e-07, + 9.4203e-07, 4.3213e-06], + [ 1.6764e-08, 0.0000e+00, 4.8103e-07, ..., 2.9802e-08, + 1.7462e-07, -1.4696e-06]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0355, -0.0307, 0.0155, 0.0326, -0.0246, 0.0057, -0.0131, 0.0138, + 0.0178, -0.0034], device='cuda:0'), grad: tensor([-2.5574e-06, 7.1526e-07, 6.2119e-07, -5.3085e-08, 1.8114e-06, + 7.8455e-06, -6.7577e-06, 7.7672e-07, -5.1595e-07, -1.8813e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 214.62, cls_loss 0.0019 cls_loss_mapping 0.0051 cls_loss_causal 0.5472 re_mapping 0.0065 re_causal 0.0208 /// teacc 98.83 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0433, 0.0272, 0.0080, ..., -0.0548, -0.0519, 0.0546], + [-0.0342, -0.0106, -0.0610, ..., 0.1204, -0.0614, -0.0421], + [-0.0119, -0.0213, 0.1560, ..., -0.0790, -0.1027, -0.0638], + ..., + [-0.0509, -0.0478, -0.0049, ..., -0.0087, 0.0806, -0.0304], + [-0.0588, -0.0485, -0.0350, ..., -0.0859, 0.0073, -0.0947], + [-0.1268, -0.0320, -0.0991, ..., -0.0835, 0.0058, 0.0743]], + device='cuda:0'), grad: tensor([[ 3.0035e-07, 0.0000e+00, -5.4017e-07, ..., 3.0501e-07, + 1.8626e-07, 3.9004e-06], + [ 3.0035e-07, 0.0000e+00, 4.6426e-07, ..., -1.0431e-06, + 1.0207e-06, 1.4119e-06], + [ 1.7555e-07, 0.0000e+00, 1.4203e-07, ..., 2.3749e-07, + 3.6927e-07, 1.2275e-06], + ..., + [ 7.9675e-07, 0.0000e+00, -1.3215e-06, ..., 3.6787e-08, + -1.6550e-06, 3.8482e-06], + [ 7.2923e-07, 0.0000e+00, 1.1735e-07, ..., 4.1863e-07, + 7.8697e-07, 1.1012e-05], + [-4.7207e-05, 0.0000e+00, 5.9092e-07, ..., 4.2841e-08, + -1.7449e-05, -9.0599e-05]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0352, -0.0284, 0.0158, 0.0319, -0.0242, 0.0060, -0.0131, 0.0119, + 0.0177, -0.0030], device='cuda:0'), grad: tensor([ 1.6555e-05, 7.1898e-06, 1.1444e-05, -1.4579e-04, 1.7130e-04, + 3.3379e-05, -1.2778e-06, 9.7826e-06, 4.9710e-05, -1.5235e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 214.55, cls_loss 0.0020 cls_loss_mapping 0.0053 cls_loss_causal 0.5457 re_mapping 0.0066 re_causal 0.0207 /// teacc 99.05 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0440, 0.0271, 0.0083, ..., -0.0545, -0.0520, 0.0542], + [-0.0345, -0.0106, -0.0618, ..., 0.1204, -0.0617, -0.0423], + [-0.0122, -0.0212, 0.1588, ..., -0.0792, -0.1005, -0.0642], + ..., + [-0.0515, -0.0478, -0.0071, ..., -0.0086, 0.0803, -0.0305], + [-0.0591, -0.0486, -0.0353, ..., -0.0861, 0.0071, -0.0949], + [-0.1274, -0.0320, -0.0992, ..., -0.0839, 0.0058, 0.0747]], + device='cuda:0'), grad: tensor([[ 1.0096e-06, 0.0000e+00, -3.5148e-06, ..., 8.4052e-07, + 6.2073e-07, -7.5530e-07], + [ 3.8929e-07, 0.0000e+00, 1.1986e-06, ..., -7.2606e-06, + 2.4270e-06, 7.3155e-07], + [ 4.3586e-07, 0.0000e+00, 1.4538e-06, ..., 6.4261e-07, + 3.7756e-06, 1.5683e-06], + ..., + [ 4.9546e-07, 0.0000e+00, -1.5840e-05, ..., 1.7239e-06, + -3.4690e-05, 1.0598e-06], + [ 1.9688e-06, 0.0000e+00, 2.4661e-06, ..., 2.0936e-06, + 2.1309e-06, 5.0627e-06], + [-3.3766e-05, 0.0000e+00, 1.2331e-06, ..., 2.2119e-07, + 2.1905e-06, -7.1406e-05]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0356, -0.0288, 0.0174, 0.0317, -0.0265, 0.0063, -0.0108, 0.0115, + 0.0178, -0.0028], device='cuda:0'), grad: tensor([ 2.4214e-06, -9.2387e-06, 1.1727e-05, 4.8399e-05, 8.6069e-05, + 1.3568e-05, 1.9014e-05, -5.2243e-05, -2.4773e-06, -1.1718e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 215.00, cls_loss 0.0017 cls_loss_mapping 0.0060 cls_loss_causal 0.5353 re_mapping 0.0061 re_causal 0.0196 /// teacc 99.12 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0438, 0.0275, 0.0100, ..., -0.0545, -0.0524, 0.0546], + [-0.0346, -0.0107, -0.0623, ..., 0.1206, -0.0620, -0.0424], + [-0.0124, -0.0212, 0.1591, ..., -0.0795, -0.1006, -0.0654], + ..., + [-0.0521, -0.0479, -0.0070, ..., -0.0084, 0.0803, -0.0310], + [-0.0593, -0.0487, -0.0356, ..., -0.0866, 0.0068, -0.0954], + [-0.1285, -0.0322, -0.0994, ..., -0.0840, 0.0062, 0.0752]], + device='cuda:0'), grad: tensor([[ 2.7381e-07, -1.2573e-08, 1.9670e-06, ..., 2.3143e-07, + 2.6878e-06, -1.2154e-07], + [ 1.5637e-06, 4.6566e-10, 1.4089e-05, ..., -5.9837e-07, + 1.8522e-05, 2.8918e-07], + [ 3.3155e-07, 2.7940e-09, -2.5406e-06, ..., 4.7497e-08, + 4.1947e-06, 5.5414e-07], + ..., + [ 5.4715e-07, 0.0000e+00, -3.5554e-05, ..., 2.9523e-07, + -5.2333e-05, 2.1048e-07], + [ 2.1886e-07, 1.8626e-09, 1.6084e-06, ..., 1.4016e-07, + 1.0449e-06, 7.9395e-07], + [ 4.1239e-06, 1.8626e-09, 4.5449e-06, ..., 3.2224e-07, + 5.9605e-06, -3.8277e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0351, -0.0290, 0.0175, 0.0317, -0.0262, 0.0060, -0.0110, 0.0114, + 0.0176, -0.0025], device='cuda:0'), grad: tensor([ 9.0823e-06, 5.3376e-05, 8.5607e-06, -1.2279e-05, 1.4856e-05, + 1.9506e-05, 3.4198e-06, -1.3125e-04, 1.1086e-05, 2.3499e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 214.21, cls_loss 0.0022 cls_loss_mapping 0.0060 cls_loss_causal 0.5636 re_mapping 0.0063 re_causal 0.0198 /// teacc 99.04 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0454, 0.0275, 0.0114, ..., -0.0549, -0.0531, 0.0551], + [-0.0348, -0.0107, -0.0624, ..., 0.1210, -0.0619, -0.0429], + [-0.0129, -0.0211, 0.1594, ..., -0.0797, -0.1010, -0.0660], + ..., + [-0.0534, -0.0479, -0.0069, ..., -0.0088, 0.0804, -0.0312], + [-0.0595, -0.0488, -0.0358, ..., -0.0869, 0.0072, -0.0964], + [-0.1305, -0.0322, -0.0999, ..., -0.0843, 0.0064, 0.0756]], + device='cuda:0'), grad: tensor([[ 3.5763e-07, 0.0000e+00, -1.3281e-06, ..., 1.1642e-06, + -4.7497e-08, -8.4098e-07], + [ 2.3879e-06, 0.0000e+00, 2.9057e-06, ..., -1.1455e-06, + 2.6217e-07, 1.1109e-05], + [ 2.4168e-07, 0.0000e+00, -3.2224e-06, ..., 4.8010e-07, + 1.1874e-07, 4.5262e-07], + ..., + [ 2.7847e-07, 0.0000e+00, 1.6950e-07, ..., 1.4687e-06, + -3.6089e-07, 8.4797e-07], + [ 3.1712e-07, 0.0000e+00, 2.4913e-07, ..., 2.7958e-06, + 2.3227e-06, 3.5018e-06], + [-2.9095e-06, 0.0000e+00, 4.8941e-07, ..., -9.1176e-07, + 3.7765e-07, -1.4380e-05]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0348, -0.0286, 0.0173, 0.0314, -0.0256, 0.0054, -0.0110, 0.0109, + 0.0181, -0.0024], device='cuda:0'), grad: tensor([ 6.9337e-07, 2.4676e-05, -1.1735e-06, -1.7732e-06, 7.0222e-06, + -4.8913e-06, -1.2606e-05, 4.9695e-06, 1.5512e-05, -3.2425e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 214.34, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5715 re_mapping 0.0066 re_causal 0.0214 /// teacc 99.04 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0463, 0.0276, 0.0120, ..., -0.0552, -0.0535, 0.0553], + [-0.0353, -0.0107, -0.0629, ..., 0.1211, -0.0621, -0.0434], + [-0.0133, -0.0211, 0.1596, ..., -0.0798, -0.1016, -0.0666], + ..., + [-0.0531, -0.0479, -0.0065, ..., -0.0088, 0.0809, -0.0312], + [-0.0587, -0.0489, -0.0364, ..., -0.0870, 0.0072, -0.0968], + [-0.1323, -0.0323, -0.1016, ..., -0.0846, 0.0062, 0.0757]], + device='cuda:0'), grad: tensor([[ 4.7963e-08, 0.0000e+00, -4.0889e-05, ..., 1.7136e-07, + 1.5926e-07, -5.3078e-05], + [ 6.4261e-08, 0.0000e+00, 1.3039e-06, ..., -1.8785e-06, + 7.5204e-07, 1.4864e-06], + [ 7.6368e-08, 0.0000e+00, 2.1197e-06, ..., 1.1325e-06, + 4.7544e-07, 2.7921e-06], + ..., + [ 7.5903e-08, 0.0000e+00, 9.5274e-07, ..., 3.1665e-08, + -1.5600e-06, 4.7460e-06], + [ 4.5635e-08, 0.0000e+00, 2.9653e-06, ..., 7.7486e-07, + 3.2550e-07, 4.2878e-06], + [ 8.1956e-08, 0.0000e+00, 1.6540e-05, ..., 3.3993e-08, + 3.2131e-07, 1.7494e-05]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0349, -0.0290, 0.0174, 0.0314, -0.0250, 0.0047, -0.0107, 0.0112, + 0.0181, -0.0029], device='cuda:0'), grad: tensor([-1.5140e-04, 1.6904e-06, 1.0863e-05, 1.9118e-05, 7.9125e-06, + 1.0259e-05, 2.7075e-05, 7.7859e-06, 1.1966e-05, 5.4777e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 214.53, cls_loss 0.0016 cls_loss_mapping 0.0049 cls_loss_causal 0.5665 re_mapping 0.0067 re_causal 0.0208 /// teacc 99.00 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0468, 0.0276, 0.0123, ..., -0.0555, -0.0537, 0.0555], + [-0.0361, -0.0107, -0.0631, ..., 0.1213, -0.0622, -0.0437], + [-0.0136, -0.0211, 0.1600, ..., -0.0802, -0.1019, -0.0670], + ..., + [-0.0544, -0.0479, -0.0065, ..., -0.0090, 0.0811, -0.0314], + [-0.0592, -0.0489, -0.0364, ..., -0.0874, 0.0074, -0.0971], + [-0.1330, -0.0323, -0.1017, ..., -0.0845, 0.0061, 0.0759]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, 0.0000e+00, 1.6764e-08, ..., 1.3923e-07, + 3.6322e-08, 1.5041e-07], + [ 8.8476e-08, 0.0000e+00, 8.8010e-08, ..., -2.0135e-06, + 4.8429e-08, 2.8219e-07], + [-7.8231e-08, 0.0000e+00, -1.9167e-06, ..., 1.7416e-07, + 2.0023e-08, 1.7695e-07], + ..., + [ 2.0862e-07, 0.0000e+00, 2.1514e-07, ..., 5.3644e-07, + -2.9337e-08, 2.7418e-06], + [ 5.7276e-08, 0.0000e+00, 4.1584e-07, ..., 5.5740e-07, + 1.5972e-07, 2.0172e-06], + [-7.1153e-07, 0.0000e+00, -2.2305e-07, ..., 2.2026e-07, + 4.6566e-08, -1.4663e-05]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0349, -0.0291, 0.0174, 0.0312, -0.0247, 0.0048, -0.0105, 0.0111, + 0.0185, -0.0030], device='cuda:0'), grad: tensor([ 9.6485e-07, -3.6117e-06, -1.6838e-06, 1.1437e-05, 1.8403e-05, + 4.3865e-07, 1.5590e-06, 6.9849e-06, -6.6496e-06, -2.7791e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 214.55, cls_loss 0.0017 cls_loss_mapping 0.0050 cls_loss_causal 0.5642 re_mapping 0.0066 re_causal 0.0209 /// teacc 99.05 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0472, 0.0276, 0.0126, ..., -0.0557, -0.0538, 0.0556], + [-0.0370, -0.0107, -0.0634, ..., 0.1215, -0.0623, -0.0446], + [-0.0138, -0.0208, 0.1605, ..., -0.0804, -0.1021, -0.0672], + ..., + [-0.0540, -0.0479, -0.0063, ..., -0.0091, 0.0814, -0.0315], + [-0.0597, -0.0490, -0.0367, ..., -0.0876, 0.0072, -0.0979], + [-0.1327, -0.0323, -0.1019, ..., -0.0845, 0.0060, 0.0764]], + device='cuda:0'), grad: tensor([[ 3.0734e-07, 6.0536e-09, -1.5460e-06, ..., 1.1036e-06, + 2.1467e-07, -4.7730e-07], + [ 2.1001e-07, 2.7008e-08, 4.0419e-07, ..., -4.0513e-07, + 5.4948e-07, 4.1164e-07], + [ 2.2678e-07, -5.4389e-07, -3.0529e-06, ..., 5.0198e-07, + 1.3690e-07, 3.7719e-07], + ..., + [ 2.0349e-07, 2.6310e-07, 1.5255e-06, ..., 3.3807e-07, + 6.4494e-07, 2.2333e-06], + [ 2.9197e-07, 4.9826e-08, 1.7229e-07, ..., 4.4936e-07, + 5.4017e-07, 3.7998e-06], + [ 5.1921e-07, 2.3283e-09, 4.5775e-07, ..., 1.0664e-07, + 1.5190e-06, -1.0930e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0349, -0.0293, 0.0176, 0.0308, -0.0254, 0.0054, -0.0107, 0.0112, + 0.0182, -0.0024], device='cuda:0'), grad: tensor([ 9.9465e-07, 1.5358e-06, -7.7719e-07, 2.1473e-05, -3.2596e-07, + -3.3110e-05, 1.2182e-05, 7.5884e-06, 7.7337e-06, -1.7330e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 214.53, cls_loss 0.0017 cls_loss_mapping 0.0050 cls_loss_causal 0.5520 re_mapping 0.0066 re_causal 0.0197 /// teacc 99.03 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0478, 0.0276, 0.0133, ..., -0.0560, -0.0541, 0.0558], + [-0.0375, -0.0107, -0.0634, ..., 0.1224, -0.0624, -0.0448], + [-0.0144, -0.0202, 0.1606, ..., -0.0818, -0.1024, -0.0678], + ..., + [-0.0544, -0.0480, -0.0062, ..., -0.0096, 0.0815, -0.0316], + [-0.0602, -0.0492, -0.0369, ..., -0.0878, 0.0069, -0.0988], + [-0.1330, -0.0324, -0.1020, ..., -0.0851, 0.0060, 0.0768]], + device='cuda:0'), grad: tensor([[ 6.9849e-07, 1.3970e-09, 5.5181e-07, ..., 2.8731e-07, + 5.7789e-07, 2.0955e-08], + [ 1.3709e-05, 3.2596e-09, 1.2973e-06, ..., 4.5747e-06, + 1.4424e-05, 1.3970e-08], + [ 8.9873e-07, 2.1886e-08, -3.4571e-05, ..., 2.5053e-07, + -4.6603e-06, 2.2352e-08], + ..., + [ 7.6033e-06, 2.2259e-07, 9.4920e-06, ..., 1.1194e-06, + -3.0220e-05, 2.2678e-07], + [ 5.2340e-06, 1.0710e-08, 8.3372e-06, ..., 1.2675e-06, + 1.1874e-06, 4.8429e-08], + [ 7.9349e-06, 4.7963e-08, 1.2301e-05, ..., 1.4864e-06, + 9.7975e-06, -1.2852e-07]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0349, -0.0291, 0.0168, 0.0313, -0.0255, 0.0051, -0.0104, 0.0110, + 0.0180, -0.0023], device='cuda:0'), grad: tensor([ 4.2468e-06, 6.7413e-05, -5.8144e-05, 3.8654e-05, -6.4075e-05, + 6.5565e-06, 4.8168e-06, -4.3035e-05, -1.5274e-05, 5.8681e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 214.32, cls_loss 0.0025 cls_loss_mapping 0.0073 cls_loss_causal 0.5400 re_mapping 0.0065 re_causal 0.0194 /// teacc 99.05 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0492, 0.0276, 0.0161, ..., -0.0559, -0.0546, 0.0574], + [-0.0382, -0.0108, -0.0626, ..., 0.1237, -0.0625, -0.0453], + [-0.0139, -0.0177, 0.1593, ..., -0.0836, -0.1049, -0.0683], + ..., + [-0.0553, -0.0496, -0.0050, ..., -0.0099, 0.0819, -0.0317], + [-0.0610, -0.0494, -0.0371, ..., -0.0885, 0.0066, -0.0996], + [-0.1343, -0.0325, -0.1037, ..., -0.0856, 0.0059, 0.0765]], + device='cuda:0'), grad: tensor([[ 9.1735e-08, 0.0000e+00, 3.6031e-05, ..., 1.1316e-07, + 4.1872e-05, 1.1295e-04], + [ 8.8941e-07, 0.0000e+00, 1.5926e-06, ..., -1.2154e-06, + 2.0694e-06, 1.9046e-06], + [ 7.6275e-07, 0.0000e+00, 6.7195e-07, ..., 3.5111e-07, + 1.2117e-06, 3.7961e-06], + ..., + [ 4.6473e-07, 0.0000e+00, -1.7071e-06, ..., 6.4122e-07, + -2.9560e-06, 4.4741e-06], + [ 1.2759e-07, 0.0000e+00, 4.0419e-07, ..., 2.0675e-07, + 3.3202e-07, 5.1036e-06], + [ 8.8336e-07, 0.0000e+00, 2.2054e-06, ..., 2.2305e-07, + 2.6282e-06, -1.6717e-06]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0336, -0.0284, 0.0142, 0.0319, -0.0246, 0.0054, -0.0101, 0.0110, + 0.0178, -0.0031], device='cuda:0'), grad: tensor([ 2.1493e-04, 6.9216e-06, 1.9044e-05, -7.0333e-05, 1.2778e-06, + -2.3639e-04, 5.9046e-06, 1.7047e-05, 3.1471e-05, 1.0051e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 214.17, cls_loss 0.0016 cls_loss_mapping 0.0051 cls_loss_causal 0.5607 re_mapping 0.0063 re_causal 0.0204 /// teacc 99.13 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0500, 0.0276, 0.0159, ..., -0.0561, -0.0552, 0.0565], + [-0.0387, -0.0108, -0.0630, ..., 0.1237, -0.0628, -0.0457], + [-0.0146, -0.0176, 0.1596, ..., -0.0838, -0.1050, -0.0686], + ..., + [-0.0555, -0.0496, -0.0048, ..., -0.0099, 0.0822, -0.0317], + [-0.0614, -0.0495, -0.0372, ..., -0.0888, 0.0064, -0.1002], + [-0.1344, -0.0324, -0.1035, ..., -0.0862, 0.0058, 0.0774]], + device='cuda:0'), grad: tensor([[-2.1439e-06, 0.0000e+00, -1.2982e-04, ..., -5.4911e-06, + 1.2489e-06, -1.4675e-04], + [ 1.3178e-07, 0.0000e+00, 4.0740e-05, ..., 2.4997e-06, + 4.2915e-06, 3.2578e-06], + [ 1.0896e-07, -1.8626e-09, -3.3021e-05, ..., -8.9221e-07, + 5.7779e-06, 4.0084e-06], + ..., + [ 9.5461e-08, 0.0000e+00, -1.6496e-05, ..., 2.7567e-07, + -3.0488e-05, 1.8226e-06], + [ 1.1465e-06, 0.0000e+00, 7.6592e-06, ..., 8.8243e-07, + -4.3325e-06, 1.8194e-05], + [-6.1188e-07, 0.0000e+00, 1.0806e-04, ..., 1.0803e-06, + 9.5069e-06, 9.4116e-05]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0344, -0.0286, 0.0140, 0.0320, -0.0249, 0.0050, -0.0098, 0.0113, + 0.0176, -0.0026], device='cuda:0'), grad: tensor([-2.5511e-04, 6.1214e-05, -2.4259e-05, 3.0473e-05, 1.9550e-05, + 1.0177e-05, 2.9534e-05, -5.1171e-05, -4.6402e-05, 2.2578e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 214.52, cls_loss 0.0018 cls_loss_mapping 0.0046 cls_loss_causal 0.5686 re_mapping 0.0062 re_causal 0.0199 /// teacc 99.03 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0509, 0.0276, 0.0165, ..., -0.0572, -0.0567, 0.0566], + [-0.0392, -0.0109, -0.0637, ..., 0.1241, -0.0630, -0.0457], + [-0.0148, -0.0174, 0.1598, ..., -0.0841, -0.1056, -0.0689], + ..., + [-0.0559, -0.0496, -0.0042, ..., -0.0100, 0.0827, -0.0318], + [-0.0618, -0.0495, -0.0379, ..., -0.0900, 0.0066, -0.1002], + [-0.1349, -0.0325, -0.1040, ..., -0.0868, 0.0056, 0.0773]], + device='cuda:0'), grad: tensor([[-1.5600e-06, 0.0000e+00, -6.4187e-06, ..., -1.8980e-06, + 7.7765e-08, -1.2435e-05], + [ 5.7742e-08, 0.0000e+00, 5.5041e-07, ..., -6.8638e-07, + 2.1374e-07, 4.0140e-07], + [ 3.3528e-08, 0.0000e+00, -1.7323e-06, ..., 1.8161e-07, + 3.0128e-07, 3.1525e-07], + ..., + [ 9.3132e-09, 0.0000e+00, -1.0133e-06, ..., 4.4145e-07, + -1.1642e-06, 2.1420e-08], + [ 2.8405e-08, 0.0000e+00, 6.8778e-07, ..., 3.4133e-07, + 1.4529e-07, 2.9150e-07], + [ 1.6065e-07, 0.0000e+00, 1.2415e-06, ..., 2.3283e-07, + 2.5192e-07, 1.8654e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0346, -0.0288, 0.0136, 0.0331, -0.0248, 0.0023, -0.0090, 0.0118, + 0.0179, -0.0030], device='cuda:0'), grad: tensor([-2.3797e-05, 4.6426e-07, -9.4902e-07, 2.8107e-06, 1.1325e-06, + 1.6894e-06, 1.4633e-05, -1.2703e-06, 1.1846e-06, 4.0792e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 214.28, cls_loss 0.0015 cls_loss_mapping 0.0052 cls_loss_causal 0.5114 re_mapping 0.0069 re_causal 0.0194 /// teacc 98.96 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0514, 0.0275, 0.0175, ..., -0.0573, -0.0571, 0.0569], + [-0.0398, -0.0109, -0.0646, ..., 0.1247, -0.0631, -0.0459], + [-0.0153, -0.0174, 0.1607, ..., -0.0842, -0.1060, -0.0694], + ..., + [-0.0563, -0.0496, -0.0042, ..., -0.0100, 0.0828, -0.0319], + [-0.0624, -0.0496, -0.0388, ..., -0.0916, 0.0067, -0.1007], + [-0.1352, -0.0325, -0.1045, ..., -0.0875, 0.0055, 0.0775]], + device='cuda:0'), grad: tensor([[ 1.8114e-07, 1.3039e-08, 1.6298e-06, ..., 1.6065e-07, + 1.0151e-06, 4.3847e-06], + [ 7.6834e-07, 3.5856e-08, 5.3179e-07, ..., -2.8824e-07, + 3.2550e-07, 7.1665e-07], + [ 6.2631e-07, 6.4727e-08, -2.8312e-06, ..., 2.9290e-07, + 3.2643e-07, 1.0068e-06], + ..., + [ 4.3958e-07, 3.6787e-08, -8.4639e-06, ..., 2.4913e-07, + -4.8690e-06, -5.9605e-06], + [ 3.1479e-07, 1.7229e-08, 5.8189e-06, ..., 3.2037e-07, + 2.3320e-06, 7.9051e-06], + [ 4.3027e-07, 5.7276e-08, -1.0461e-05, ..., 2.7474e-08, + -6.2063e-06, -4.2617e-05]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0340, -0.0290, 0.0142, 0.0332, -0.0246, 0.0021, -0.0090, 0.0117, + 0.0175, -0.0032], device='cuda:0'), grad: tensor([ 7.0892e-06, 2.0377e-06, 1.2051e-06, 4.0472e-05, -2.7195e-06, + 1.1139e-05, 6.8499e-07, -1.8880e-05, 9.4250e-06, -5.0426e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 133---------------------------------------------------- +epoch 133, time 230.75, cls_loss 0.0015 cls_loss_mapping 0.0049 cls_loss_causal 0.5189 re_mapping 0.0061 re_causal 0.0191 /// teacc 99.18 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0517, 0.0275, 0.0178, ..., -0.0578, -0.0584, 0.0569], + [-0.0401, -0.0109, -0.0650, ..., 0.1255, -0.0633, -0.0461], + [-0.0170, -0.0175, 0.1611, ..., -0.0847, -0.1062, -0.0699], + ..., + [-0.0569, -0.0497, -0.0040, ..., -0.0102, 0.0833, -0.0319], + [-0.0627, -0.0498, -0.0390, ..., -0.0922, 0.0053, -0.1019], + [-0.1356, -0.0327, -0.1048, ..., -0.0878, 0.0053, 0.0777]], + device='cuda:0'), grad: tensor([[ 5.0142e-06, 0.0000e+00, -5.3532e-06, ..., 8.9929e-06, + 6.8266e-07, -1.1511e-05], + [ 2.3693e-06, 0.0000e+00, 6.8406e-07, ..., 3.6284e-06, + 8.6101e-07, 1.3085e-07], + [ 3.6228e-07, 0.0000e+00, -2.4643e-06, ..., 1.3160e-06, + 3.1944e-07, 1.8161e-07], + ..., + [ 1.4948e-07, 0.0000e+00, -1.3858e-06, ..., 2.8638e-07, + -2.2743e-06, 6.5705e-07], + [ 1.7826e-06, 0.0000e+00, 1.0328e-06, ..., 3.3379e-06, + 4.3027e-07, 7.4599e-07], + [ 7.9023e-07, 0.0000e+00, 5.9567e-06, ..., 1.5851e-06, + 2.0117e-07, 8.0764e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0342, -0.0289, 0.0140, 0.0329, -0.0244, 0.0021, -0.0088, 0.0119, + 0.0172, -0.0032], device='cuda:0'), grad: tensor([ 5.0291e-06, 1.0513e-05, 2.0638e-06, 3.4012e-06, 2.7612e-05, + 6.0648e-06, -7.3552e-05, -3.3267e-06, 7.0706e-06, 1.5080e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 214.30, cls_loss 0.0014 cls_loss_mapping 0.0045 cls_loss_causal 0.5120 re_mapping 0.0063 re_causal 0.0193 /// teacc 99.13 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0520, 0.0275, 0.0182, ..., -0.0579, -0.0590, 0.0571], + [-0.0406, -0.0109, -0.0653, ..., 0.1261, -0.0635, -0.0485], + [-0.0171, -0.0174, 0.1614, ..., -0.0851, -0.1064, -0.0705], + ..., + [-0.0574, -0.0497, -0.0038, ..., -0.0105, 0.0836, -0.0319], + [-0.0630, -0.0498, -0.0394, ..., -0.0923, 0.0048, -0.1031], + [-0.1358, -0.0327, -0.1051, ..., -0.0880, 0.0052, 0.0783]], + device='cuda:0'), grad: tensor([[ 5.5879e-07, 0.0000e+00, -2.2743e-06, ..., 5.4063e-07, + 1.3108e-07, -2.8722e-06], + [ 1.1297e-06, 0.0000e+00, 1.1278e-06, ..., 4.0443e-07, + 6.9244e-07, 6.2911e-07], + [ 1.5628e-06, 0.0000e+00, -9.5647e-07, ..., 1.1381e-06, + 4.1886e-07, 5.0198e-07], + ..., + [ 1.3094e-06, 0.0000e+00, -2.4810e-06, ..., 4.1444e-08, + -1.6522e-06, 2.8405e-08], + [ 8.3772e-07, 0.0000e+00, 2.9453e-07, ..., 5.3551e-07, + 7.1712e-07, 1.2740e-06], + [ 9.3412e-07, 0.0000e+00, 2.8592e-06, ..., 5.6578e-08, + 1.8813e-06, 3.7421e-06]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0341, -0.0292, 0.0138, 0.0328, -0.0245, 0.0020, -0.0089, 0.0120, + 0.0172, -0.0027], device='cuda:0'), grad: tensor([-4.6082e-06, 8.6948e-06, 1.4611e-05, -1.1235e-04, 1.9353e-06, + 6.5625e-05, -5.9828e-06, 5.5134e-06, 9.4026e-06, 1.7121e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 214.42, cls_loss 0.0014 cls_loss_mapping 0.0050 cls_loss_causal 0.5429 re_mapping 0.0061 re_causal 0.0195 /// teacc 99.08 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0524, 0.0275, 0.0192, ..., -0.0574, -0.0592, 0.0575], + [-0.0411, -0.0109, -0.0656, ..., 0.1265, -0.0636, -0.0486], + [-0.0172, -0.0172, 0.1619, ..., -0.0853, -0.1064, -0.0709], + ..., + [-0.0580, -0.0496, -0.0039, ..., -0.0107, 0.0838, -0.0321], + [-0.0633, -0.0499, -0.0401, ..., -0.0927, 0.0047, -0.1040], + [-0.1363, -0.0327, -0.1053, ..., -0.0884, 0.0052, 0.0783]], + device='cuda:0'), grad: tensor([[ 7.0548e-08, 0.0000e+00, 1.4445e-06, ..., 8.6380e-08, + 1.2042e-06, -3.8976e-07], + [ 4.8196e-07, 0.0000e+00, 1.2666e-05, ..., -3.0957e-06, + 1.0543e-05, 1.3318e-07], + [ 7.7998e-08, 0.0000e+00, -1.6261e-06, ..., 1.5441e-06, + 9.2015e-06, 1.2829e-07], + ..., + [ 7.1153e-07, 0.0000e+00, -4.9710e-05, ..., 3.7649e-07, + -4.5925e-05, 1.4883e-06], + [ 4.7637e-07, 0.0000e+00, 8.7917e-06, ..., 4.6194e-07, + 2.0377e-06, 9.8627e-07], + [ 6.8406e-07, 0.0000e+00, 3.4329e-06, ..., 7.2410e-08, + 3.1516e-06, -3.7421e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0335, -0.0293, 0.0139, 0.0327, -0.0243, 0.0026, -0.0095, 0.0119, + 0.0169, -0.0028], device='cuda:0'), grad: tensor([ 3.3919e-06, 2.7418e-05, 1.1861e-05, 2.4095e-05, 2.0400e-05, + 1.3776e-05, 3.2987e-06, -1.2279e-04, 1.5646e-05, 2.9262e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 214.47, cls_loss 0.0016 cls_loss_mapping 0.0049 cls_loss_causal 0.5331 re_mapping 0.0060 re_causal 0.0184 /// teacc 99.13 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0531, 0.0275, 0.0204, ..., -0.0577, -0.0583, 0.0578], + [-0.0415, -0.0110, -0.0659, ..., 0.1271, -0.0638, -0.0487], + [-0.0175, -0.0172, 0.1621, ..., -0.0856, -0.1067, -0.0718], + ..., + [-0.0572, -0.0497, -0.0036, ..., -0.0110, 0.0842, -0.0322], + [-0.0637, -0.0501, -0.0404, ..., -0.0932, 0.0044, -0.1044], + [-0.1366, -0.0329, -0.1056, ..., -0.0887, 0.0051, 0.0786]], + device='cuda:0'), grad: tensor([[ 1.6158e-07, 0.0000e+00, -2.9569e-07, ..., 3.6298e-07, + 2.5565e-07, 4.7637e-07], + [ 6.4960e-08, 0.0000e+00, 1.3309e-06, ..., -9.0292e-07, + 3.2573e-07, 3.2783e-07], + [ 7.2410e-08, 0.0000e+00, -3.1088e-06, ..., 1.4273e-07, + 4.1537e-07, 3.9348e-07], + ..., + [ 4.7497e-08, 0.0000e+00, -4.7944e-06, ..., 1.0920e-07, + -1.2759e-07, 1.1496e-05], + [ 7.6648e-07, 0.0000e+00, 2.3772e-07, ..., 3.7090e-07, + 1.1595e-07, 1.4175e-06], + [ 1.3714e-07, 0.0000e+00, 3.7458e-06, ..., 6.9151e-08, + -2.2445e-06, -1.5959e-05]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0330, -0.0292, 0.0137, 0.0339, -0.0246, 0.0003, -0.0095, 0.0122, + 0.0168, -0.0027], device='cuda:0'), grad: tensor([ 2.6040e-06, 1.1995e-06, -1.0226e-06, 5.8189e-06, 1.3094e-06, + 1.5758e-06, -1.2740e-06, 1.3821e-05, 1.0412e-06, -2.5108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 214.44, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5480 re_mapping 0.0055 re_causal 0.0187 /// teacc 99.16 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0536, 0.0274, 0.0201, ..., -0.0579, -0.0587, 0.0578], + [-0.0417, -0.0110, -0.0668, ..., 0.1275, -0.0642, -0.0490], + [-0.0176, -0.0172, 0.1628, ..., -0.0859, -0.1069, -0.0721], + ..., + [-0.0574, -0.0499, -0.0033, ..., -0.0111, 0.0846, -0.0322], + [-0.0640, -0.0505, -0.0410, ..., -0.0935, 0.0040, -0.1047], + [-0.1368, -0.0333, -0.1057, ..., -0.0891, 0.0050, 0.0788]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, 0.0000e+00, -1.3728e-06, ..., 1.0431e-07, + 5.4715e-08, -2.7250e-06], + [ 6.2864e-09, 0.0000e+00, 1.7747e-05, ..., -2.0452e-06, + 4.1607e-07, 1.1525e-07], + [ 2.7940e-09, 0.0000e+00, -9.3818e-05, ..., 2.2934e-07, + 1.4566e-06, 8.6753e-07], + ..., + [ 4.4238e-09, 0.0000e+00, 5.8830e-05, ..., 1.0328e-06, + -2.7381e-06, 3.4226e-07], + [ 2.7940e-09, 0.0000e+00, 9.5591e-06, ..., 2.3446e-07, + 4.7963e-07, 5.6345e-08], + [ 3.9581e-09, 0.0000e+00, 3.3267e-06, ..., 1.2293e-07, + 1.2550e-07, 3.2922e-07]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0333, -0.0296, 0.0140, 0.0337, -0.0246, 0.0008, -0.0097, 0.0125, + 0.0165, -0.0027], device='cuda:0'), grad: tensor([-2.2184e-06, 1.7285e-05, -1.0651e-04, 5.9493e-06, 3.6918e-06, + -1.4231e-06, 1.9446e-06, 6.9201e-05, 7.4655e-06, 4.7199e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 214.25, cls_loss 0.0015 cls_loss_mapping 0.0050 cls_loss_causal 0.5341 re_mapping 0.0056 re_causal 0.0187 /// teacc 99.11 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0541, 0.0274, 0.0203, ..., -0.0579, -0.0592, 0.0579], + [-0.0424, -0.0111, -0.0675, ..., 0.1278, -0.0644, -0.0491], + [-0.0180, -0.0172, 0.1635, ..., -0.0860, -0.1072, -0.0734], + ..., + [-0.0565, -0.0498, -0.0030, ..., -0.0111, 0.0849, -0.0322], + [-0.0647, -0.0507, -0.0417, ..., -0.0939, 0.0046, -0.1060], + [-0.1379, -0.0334, -0.1060, ..., -0.0893, 0.0049, 0.0790]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 0.0000e+00, -5.2853e-08, ..., 3.2154e-07, + 9.0105e-08, -1.9791e-08], + [ 1.7928e-08, 0.0000e+00, 6.2352e-07, ..., -4.9360e-06, + 4.9872e-07, 6.9384e-08], + [ 1.8859e-08, 0.0000e+00, 7.4320e-07, ..., 7.1712e-07, + 5.5414e-07, 9.5461e-08], + ..., + [ 4.0047e-08, 0.0000e+00, -2.8219e-06, ..., 1.2750e-06, + -2.4512e-06, 1.9162e-07], + [ 1.7928e-08, 0.0000e+00, 1.2503e-07, ..., 1.8962e-06, + 9.6951e-07, 9.7416e-07], + [ 2.8405e-07, 0.0000e+00, 6.7567e-07, ..., 2.0303e-07, + 7.7905e-07, -3.6042e-07]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0333, -0.0298, 0.0143, 0.0334, -0.0245, 0.0014, -0.0107, 0.0129, + 0.0163, -0.0027], device='cuda:0'), grad: tensor([ 1.4948e-06, -1.0625e-05, 3.7849e-06, 5.0738e-06, 9.2713e-07, + -5.6252e-06, -4.6231e-06, -3.0976e-06, 1.0058e-05, 2.6245e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 214.22, cls_loss 0.0014 cls_loss_mapping 0.0063 cls_loss_causal 0.5208 re_mapping 0.0060 re_causal 0.0182 /// teacc 99.12 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0562, 0.0274, 0.0207, ..., -0.0582, -0.0594, 0.0580], + [-0.0427, -0.0111, -0.0674, ..., 0.1285, -0.0644, -0.0492], + [-0.0183, -0.0172, 0.1637, ..., -0.0863, -0.1073, -0.0738], + ..., + [-0.0573, -0.0498, -0.0030, ..., -0.0115, 0.0851, -0.0323], + [-0.0650, -0.0508, -0.0420, ..., -0.0945, 0.0045, -0.1066], + [-0.1388, -0.0334, -0.1062, ..., -0.0897, 0.0048, 0.0792]], + device='cuda:0'), grad: tensor([[ 5.5600e-07, 0.0000e+00, -8.8150e-07, ..., 2.9150e-07, + 1.1083e-07, -2.7707e-07], + [ 1.0626e-06, 0.0000e+00, 1.7113e-07, ..., -1.9628e-07, + 1.2130e-07, 1.3085e-06], + [ 4.7777e-07, 0.0000e+00, 4.4331e-07, ..., 6.8219e-08, + 1.1222e-07, 1.4892e-06], + ..., + [ 1.5544e-06, 0.0000e+00, -5.4436e-07, ..., 4.7032e-08, + -1.2396e-06, 2.7362e-06], + [ 4.3772e-07, 0.0000e+00, 1.0710e-07, ..., 2.3213e-07, + 4.2375e-08, 1.7378e-06], + [ 7.1153e-06, 0.0000e+00, 4.2445e-07, ..., 9.8255e-08, + 1.5763e-07, -4.2841e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0336, -0.0296, 0.0141, 0.0330, -0.0242, 0.0008, -0.0091, 0.0128, + 0.0163, -0.0028], device='cuda:0'), grad: tensor([ 1.4901e-06, 6.9141e-06, 1.1355e-05, -2.9564e-05, -1.3217e-05, + 5.1931e-06, -1.8142e-06, 5.7817e-06, 7.4580e-06, 6.3516e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 214.41, cls_loss 0.0017 cls_loss_mapping 0.0042 cls_loss_causal 0.5145 re_mapping 0.0058 re_causal 0.0176 /// teacc 99.15 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0569, 0.0274, 0.0226, ..., -0.0584, -0.0598, 0.0586], + [-0.0434, -0.0111, -0.0690, ..., 0.1298, -0.0645, -0.0494], + [-0.0185, -0.0174, 0.1645, ..., -0.0880, -0.1075, -0.0748], + ..., + [-0.0584, -0.0499, -0.0030, ..., -0.0118, 0.0853, -0.0325], + [-0.0660, -0.0511, -0.0422, ..., -0.0951, 0.0044, -0.1072], + [-0.1420, -0.0334, -0.1066, ..., -0.0900, 0.0047, 0.0791]], + device='cuda:0'), grad: tensor([[ 2.2142e-07, 0.0000e+00, 1.4119e-06, ..., 6.0955e-07, + 2.1118e-07, 2.6729e-06], + [ 5.1828e-07, 0.0000e+00, 8.7395e-06, ..., 7.5623e-07, + 8.0839e-07, 2.1253e-06], + [ 2.0559e-07, 0.0000e+00, -1.9193e-05, ..., -4.2133e-06, + 1.2396e-06, 1.2424e-06], + ..., + [ 3.2294e-07, 0.0000e+00, -7.4506e-06, ..., 4.5728e-07, + -8.4937e-06, 9.6411e-06], + [ 4.4098e-07, 0.0000e+00, 3.3770e-06, ..., 7.3481e-07, + 5.9744e-07, 1.8224e-05], + [ 1.9884e-07, 0.0000e+00, -4.4107e-06, ..., 1.0571e-07, + 6.6217e-07, -6.2168e-05]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0328, -0.0294, 0.0139, 0.0328, -0.0217, 0.0012, -0.0101, 0.0127, + 0.0161, -0.0042], device='cuda:0'), grad: tensor([ 1.0163e-05, 1.7211e-05, -2.4304e-05, 4.7088e-05, 2.7597e-05, + 4.9137e-06, -7.9628e-08, -5.1260e-06, 3.4541e-05, -1.1206e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 214.40, cls_loss 0.0016 cls_loss_mapping 0.0052 cls_loss_causal 0.5105 re_mapping 0.0061 re_causal 0.0191 /// teacc 99.01 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0584, 0.0273, 0.0223, ..., -0.0588, -0.0602, 0.0570], + [-0.0437, -0.0112, -0.0695, ..., 0.1302, -0.0646, -0.0491], + [-0.0187, -0.0174, 0.1645, ..., -0.0882, -0.1078, -0.0751], + ..., + [-0.0591, -0.0499, -0.0021, ..., -0.0121, 0.0856, -0.0326], + [-0.0664, -0.0512, -0.0431, ..., -0.0957, 0.0042, -0.1075], + [-0.1422, -0.0335, -0.1061, ..., -0.0902, 0.0046, 0.0803]], + device='cuda:0'), grad: tensor([[ 5.1595e-07, 0.0000e+00, -9.3222e-05, ..., 9.0003e-06, + 6.9337e-07, -1.1015e-04], + [ 8.5076e-07, 6.9849e-10, 7.2364e-07, ..., 1.2785e-05, + 1.0459e-06, 8.5682e-07], + [ 2.9267e-07, 2.3283e-10, 1.5302e-06, ..., 5.1744e-06, + 4.2282e-07, 1.7630e-06], + ..., + [ 2.4983e-07, 9.3132e-10, 1.3621e-07, ..., 2.0918e-06, + 1.5832e-08, 8.6566e-07], + [ 2.8387e-05, 2.3283e-10, 7.4387e-05, ..., 6.1083e-04, + 3.8713e-05, 8.8632e-05], + [ 6.6310e-07, 6.7521e-09, 8.9034e-06, ..., 5.6550e-06, + 5.5972e-07, 1.0543e-05]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0346, -0.0293, 0.0137, 0.0326, -0.0218, 0.0011, -0.0098, 0.0129, + 0.0163, -0.0035], device='cuda:0'), grad: tensor([-2.4354e-04, 5.6952e-05, 2.4885e-05, 4.7892e-05, 3.8385e-05, + 7.8321e-05, -2.7027e-03, 9.1568e-06, 2.6455e-03, 4.8757e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 214.36, cls_loss 0.0015 cls_loss_mapping 0.0057 cls_loss_causal 0.5023 re_mapping 0.0061 re_causal 0.0190 /// teacc 99.14 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0587, 0.0272, 0.0230, ..., -0.0588, -0.0604, 0.0574], + [-0.0443, -0.0112, -0.0699, ..., 0.1304, -0.0647, -0.0496], + [-0.0186, -0.0172, 0.1643, ..., -0.0887, -0.1078, -0.0757], + ..., + [-0.0577, -0.0499, -0.0009, ..., -0.0124, 0.0866, -0.0327], + [-0.0657, -0.0519, -0.0437, ..., -0.0966, 0.0041, -0.1076], + [-0.1426, -0.0323, -0.1064, ..., -0.0905, 0.0046, 0.0805]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, -1.2317e-07, ..., 5.3551e-08, + 2.1677e-07, 1.0780e-07], + [ 2.3283e-10, 0.0000e+00, 2.2841e-07, ..., -1.0608e-06, + 3.2363e-07, 1.9511e-07], + [ 6.9849e-10, 0.0000e+00, -2.1663e-06, ..., 2.2375e-07, + 3.7136e-07, 1.8068e-07], + ..., + [ 0.0000e+00, 0.0000e+00, -1.1902e-06, ..., 9.0338e-08, + -1.4314e-06, 5.5274e-07], + [ 4.6566e-10, 0.0000e+00, 6.3842e-07, ..., 2.9011e-07, + 1.1306e-06, 1.4147e-06], + [ 0.0000e+00, 0.0000e+00, 3.6275e-07, ..., 3.2829e-08, + 4.9872e-07, -7.9256e-07]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0342, -0.0301, 0.0132, 0.0325, -0.0239, 0.0020, -0.0104, 0.0143, + 0.0173, -0.0036], device='cuda:0'), grad: tensor([ 9.8161e-07, -7.6741e-07, -8.7777e-07, 4.0501e-05, 1.0226e-06, + -7.2002e-05, 2.9549e-05, -1.4640e-06, 9.9652e-07, 2.0657e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 214.25, cls_loss 0.0012 cls_loss_mapping 0.0034 cls_loss_causal 0.5289 re_mapping 0.0059 re_causal 0.0189 /// teacc 99.15 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0593, 0.0278, 0.0230, ..., -0.0593, -0.0611, 0.0574], + [-0.0447, -0.0112, -0.0701, ..., 0.1308, -0.0648, -0.0494], + [-0.0189, -0.0172, 0.1643, ..., -0.0890, -0.1081, -0.0762], + ..., + [-0.0577, -0.0500, -0.0005, ..., -0.0124, 0.0869, -0.0328], + [-0.0660, -0.0525, -0.0438, ..., -0.0970, 0.0049, -0.1084], + [-0.1427, -0.0316, -0.1067, ..., -0.0907, 0.0045, 0.0807]], + device='cuda:0'), grad: tensor([[ 8.1491e-08, 0.0000e+00, -2.9523e-07, ..., 8.8010e-08, + 3.7253e-08, -4.1490e-07], + [ 2.0815e-07, 0.0000e+00, -3.2596e-08, ..., -2.0117e-06, + 2.0815e-07, 3.0734e-08], + [ 7.8231e-08, 0.0000e+00, 9.6858e-08, ..., 9.2341e-07, + 2.0070e-07, 3.1665e-08], + ..., + [ 4.5169e-07, 0.0000e+00, -1.2508e-06, ..., 4.8662e-07, + -1.5711e-06, 2.1094e-07], + [ 1.5693e-07, 0.0000e+00, 4.4052e-07, ..., 1.8068e-07, + 1.4110e-07, 1.6624e-07], + [ 3.5204e-07, 0.0000e+00, 1.4827e-06, ..., 7.3574e-08, + 1.0207e-06, -8.1956e-08]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0346, -0.0299, 0.0127, 0.0322, -0.0239, 0.0021, -0.0105, 0.0147, + 0.0175, -0.0036], device='cuda:0'), grad: tensor([-1.9651e-07, -3.6750e-06, 3.0752e-06, -6.8806e-06, -1.7378e-06, + 5.5460e-07, 5.7975e-07, 1.6503e-06, 3.2727e-06, 3.3155e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 214.45, cls_loss 0.0014 cls_loss_mapping 0.0060 cls_loss_causal 0.5314 re_mapping 0.0059 re_causal 0.0186 /// teacc 98.98 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0596, 0.0277, 0.0233, ..., -0.0594, -0.0616, 0.0576], + [-0.0456, -0.0113, -0.0702, ..., 0.1314, -0.0648, -0.0495], + [-0.0188, -0.0171, 0.1646, ..., -0.0892, -0.1084, -0.0769], + ..., + [-0.0580, -0.0500, -0.0004, ..., -0.0133, 0.0873, -0.0332], + [-0.0664, -0.0526, -0.0439, ..., -0.0973, 0.0048, -0.1094], + [-0.1432, -0.0316, -0.1071, ..., -0.0910, 0.0045, 0.0814]], + device='cuda:0'), grad: tensor([[-5.8999e-07, 0.0000e+00, -1.5438e-04, ..., 6.6124e-08, + 2.6915e-07, -3.9607e-05], + [ 6.0536e-08, 0.0000e+00, 1.5423e-06, ..., -6.7614e-07, + 3.7672e-07, 3.5390e-07], + [ 6.8825e-07, 0.0000e+00, 1.2934e-04, ..., 7.3574e-08, + 2.1188e-07, 3.7193e-05], + ..., + [ 1.0571e-07, 0.0000e+00, 6.3144e-06, ..., 1.3085e-07, + -1.1362e-07, 1.0440e-06], + [ 3.8650e-08, 0.0000e+00, 6.3479e-06, ..., 2.3190e-07, + 8.5449e-07, 1.5860e-06], + [ 2.6124e-07, 0.0000e+00, 7.5391e-07, ..., 6.9849e-08, + 1.1632e-06, -7.3714e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0345, -0.0297, 0.0126, 0.0316, -0.0239, 0.0021, -0.0105, 0.0144, + 0.0175, -0.0030], device='cuda:0'), grad: tensor([-1.7953e-04, 2.7455e-06, 1.5211e-04, -8.6203e-06, 2.1514e-06, + -1.6436e-05, 1.0096e-05, 1.2308e-05, 2.2545e-05, 2.8256e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 214.31, cls_loss 0.0015 cls_loss_mapping 0.0042 cls_loss_causal 0.5495 re_mapping 0.0059 re_causal 0.0191 /// teacc 99.03 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0599, 0.0277, 0.0238, ..., -0.0595, -0.0619, 0.0577], + [-0.0463, -0.0114, -0.0709, ..., 0.1316, -0.0650, -0.0498], + [-0.0192, -0.0174, 0.1651, ..., -0.0893, -0.1087, -0.0775], + ..., + [-0.0583, -0.0500, -0.0006, ..., -0.0133, 0.0872, -0.0335], + [-0.0670, -0.0530, -0.0444, ..., -0.0976, 0.0046, -0.1101], + [-0.1444, -0.0316, -0.1066, ..., -0.0912, 0.0048, 0.0817]], + device='cuda:0'), grad: tensor([[ 1.2526e-07, 0.0000e+00, 1.7276e-07, ..., 2.7986e-07, + 2.5146e-07, -2.1160e-06], + [ 1.3616e-06, 0.0000e+00, 5.1083e-07, ..., -2.3711e-06, + 2.4959e-06, 1.7462e-07], + [ 1.2480e-07, 0.0000e+00, -2.2035e-06, ..., 5.6578e-07, + 2.7940e-07, 1.1874e-07], + ..., + [-1.2420e-05, 0.0000e+00, -1.1772e-06, ..., 2.4121e-07, + -2.3589e-05, 1.8626e-07], + [ 2.9756e-07, 0.0000e+00, 6.3702e-07, ..., 9.2760e-07, + 5.5926e-07, 1.7928e-07], + [ 7.4646e-07, 0.0000e+00, 9.5088e-07, ..., 4.0838e-07, + 3.4347e-06, -1.7509e-06]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0345, -0.0301, 0.0124, 0.0319, -0.0233, 0.0021, -0.0106, 0.0130, + 0.0173, -0.0014], device='cuda:0'), grad: tensor([-3.3854e-07, 4.3400e-06, 3.4440e-06, 9.7379e-06, 6.0558e-05, + -2.7083e-06, 2.0377e-06, -7.8619e-05, -6.9328e-06, 8.3670e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 214.00, cls_loss 0.0018 cls_loss_mapping 0.0049 cls_loss_causal 0.5408 re_mapping 0.0058 re_causal 0.0183 /// teacc 99.12 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0603, 0.0277, 0.0237, ..., -0.0595, -0.0626, 0.0574], + [-0.0497, -0.0114, -0.0724, ..., 0.1312, -0.0653, -0.0507], + [-0.0162, -0.0174, 0.1665, ..., -0.0877, -0.1089, -0.0780], + ..., + [-0.0585, -0.0500, -0.0003, ..., -0.0135, 0.0879, -0.0335], + [-0.0691, -0.0531, -0.0454, ..., -0.0985, 0.0044, -0.1104], + [-0.1446, -0.0316, -0.1069, ..., -0.0915, 0.0044, 0.0826]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 2.0023e-08, -7.2643e-07, ..., 2.3413e-06, + 2.4214e-08, -1.0282e-06], + [ 8.0094e-08, 1.1642e-08, 2.5798e-07, ..., -3.6299e-05, + 5.0804e-07, -2.1188e-07], + [ 4.8289e-07, -2.5751e-07, -4.7311e-06, ..., 1.4482e-06, + 3.3062e-08, 2.3749e-08], + ..., + [ 2.3004e-07, 5.6811e-08, -7.6788e-07, ..., 3.0771e-06, + -1.4473e-06, -1.8766e-07], + [ 1.1548e-07, 8.8476e-08, 3.1609e-06, ..., 4.7572e-06, + 2.6170e-07, 3.2131e-07], + [ 1.0869e-06, 2.0489e-08, 1.9539e-06, ..., 2.2668e-06, + 8.2888e-07, 1.1055e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0349, -0.0324, 0.0152, 0.0318, -0.0253, 0.0018, -0.0106, 0.0134, + 0.0167, -0.0003], device='cuda:0'), grad: tensor([ 3.5707e-06, -7.1526e-05, -2.3171e-06, 4.2375e-07, 8.0988e-06, + 5.7183e-06, 2.7403e-05, 2.7008e-06, 1.5065e-05, 1.0870e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 214.43, cls_loss 0.0026 cls_loss_mapping 0.0072 cls_loss_causal 0.5606 re_mapping 0.0059 re_causal 0.0185 /// teacc 98.99 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0608, 0.0277, 0.0237, ..., -0.0601, -0.0640, 0.0574], + [-0.0499, -0.0114, -0.0730, ..., 0.1314, -0.0659, -0.0499], + [-0.0163, -0.0174, 0.1667, ..., -0.0895, -0.1093, -0.0793], + ..., + [-0.0596, -0.0500, 0.0003, ..., -0.0118, 0.0884, -0.0336], + [-0.0698, -0.0532, -0.0459, ..., -0.0994, 0.0043, -0.1116], + [-0.1472, -0.0316, -0.1071, ..., -0.0929, 0.0042, 0.0824]], + device='cuda:0'), grad: tensor([[ 1.0524e-07, 0.0000e+00, -2.3562e-07, ..., 1.0943e-07, + 1.0291e-07, 4.2375e-08], + [ 1.7975e-07, 0.0000e+00, 4.0606e-07, ..., -4.2329e-07, + 3.1898e-07, 3.2596e-07], + [ 4.3539e-07, 0.0000e+00, 6.2361e-06, ..., 1.7369e-07, + 3.8818e-06, 8.8103e-07], + ..., + [ 9.2713e-07, 0.0000e+00, -1.0006e-05, ..., 3.4319e-07, + -6.1728e-06, 1.7863e-06], + [ 4.1630e-07, 0.0000e+00, 1.6950e-07, ..., 3.1628e-06, + 3.0100e-06, 2.9150e-06], + [ 2.7139e-06, 0.0000e+00, 5.3039e-07, ..., 2.6356e-07, + 3.1805e-07, 3.1535e-06]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0353, -0.0328, 0.0146, 0.0328, -0.0227, 0.0008, -0.0099, 0.0141, + 0.0162, -0.0026], device='cuda:0'), grad: tensor([ 7.9535e-07, 1.7658e-06, 1.6481e-05, -2.9668e-05, 1.2461e-06, + -1.7926e-05, -5.1670e-06, -1.0461e-05, 2.8327e-05, 1.4536e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 214.09, cls_loss 0.0015 cls_loss_mapping 0.0057 cls_loss_causal 0.5447 re_mapping 0.0061 re_causal 0.0190 /// teacc 99.05 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0613, 0.0276, 0.0240, ..., -0.0609, -0.0643, 0.0576], + [-0.0499, -0.0114, -0.0731, ..., 0.1333, -0.0660, -0.0500], + [-0.0164, -0.0188, 0.1666, ..., -0.0901, -0.1103, -0.0798], + ..., + [-0.0599, -0.0494, 0.0013, ..., -0.0133, 0.0892, -0.0337], + [-0.0709, -0.0532, -0.0460, ..., -0.1008, 0.0039, -0.1124], + [-0.1475, -0.0312, -0.1075, ..., -0.0935, 0.0040, 0.0826]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0449e-06, ..., 1.5832e-08, + 3.3993e-08, -1.6242e-06], + [ 1.3970e-09, 0.0000e+00, 4.7730e-07, ..., -3.8650e-07, + 2.8033e-07, 1.9837e-07], + [ 2.7940e-09, 0.0000e+00, -5.2154e-06, ..., 1.1688e-07, + 1.3504e-08, 1.8766e-07], + ..., + [ 4.6566e-10, 0.0000e+00, 3.6824e-06, ..., 1.6158e-07, + -1.0515e-06, 3.3434e-07], + [ 3.2596e-09, 0.0000e+00, -5.9092e-07, ..., 4.2841e-08, + -9.7789e-09, 6.5751e-07], + [ 4.6566e-10, 0.0000e+00, 1.4538e-06, ..., 1.3504e-08, + 4.1490e-07, 3.5949e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0354, -0.0321, 0.0143, 0.0323, -0.0229, 0.0012, -0.0096, 0.0142, + 0.0157, -0.0026], device='cuda:0'), grad: tensor([-1.6605e-06, 1.2256e-06, -4.9584e-06, -2.1443e-05, 5.9977e-07, + 1.9655e-05, 4.9965e-07, 4.5896e-06, -1.9781e-06, 3.4589e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 214.07, cls_loss 0.0017 cls_loss_mapping 0.0040 cls_loss_causal 0.5188 re_mapping 0.0059 re_causal 0.0178 /// teacc 99.07 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0622, 0.0277, 0.0244, ..., -0.0607, -0.0646, 0.0573], + [-0.0500, -0.0115, -0.0758, ..., 0.1326, -0.0661, -0.0503], + [-0.0164, -0.0188, 0.1688, ..., -0.0885, -0.1103, -0.0808], + ..., + [-0.0601, -0.0494, 0.0011, ..., -0.0138, 0.0895, -0.0338], + [-0.0715, -0.0533, -0.0465, ..., -0.1022, 0.0031, -0.1140], + [-0.1476, -0.0313, -0.1076, ..., -0.0941, 0.0038, 0.0830]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, 8.1491e-08, ..., 2.0908e-07, + 4.6566e-09, 6.2864e-08], + [ 1.3970e-09, 0.0000e+00, -8.6147e-07, ..., -9.9540e-05, + 2.7940e-08, 7.3109e-08], + [ 1.8626e-09, 0.0000e+00, -3.2615e-06, ..., 9.0063e-05, + 4.2375e-08, 1.0012e-07], + ..., + [ 4.1910e-09, 0.0000e+00, 1.7257e-06, ..., 1.4370e-06, + -1.7742e-07, 5.6578e-07], + [ 5.5879e-09, 0.0000e+00, 1.5460e-06, ..., 2.5537e-06, + 2.0955e-08, 2.8685e-07], + [ 4.6566e-10, 0.0000e+00, 1.2200e-07, ..., 1.1828e-06, + 3.9116e-08, -1.7863e-06]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0356, -0.0332, 0.0156, 0.0322, -0.0230, 0.0016, -0.0101, 0.0141, + 0.0151, -0.0025], device='cuda:0'), grad: tensor([ 7.7067e-07, -2.1219e-04, 1.8823e-04, 2.1420e-08, 5.0664e-06, + 1.3085e-06, 2.9355e-06, 6.0983e-06, 8.0466e-06, 3.4459e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 214.11, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5362 re_mapping 0.0059 re_causal 0.0178 /// teacc 99.14 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0630, 0.0277, 0.0260, ..., -0.0611, -0.0649, 0.0584], + [-0.0503, -0.0115, -0.0757, ..., 0.1334, -0.0662, -0.0508], + [-0.0164, -0.0188, 0.1688, ..., -0.0892, -0.1109, -0.0826], + ..., + [-0.0602, -0.0494, 0.0014, ..., -0.0141, 0.0899, -0.0341], + [-0.0720, -0.0534, -0.0467, ..., -0.1029, 0.0028, -0.1150], + [-0.1477, -0.0314, -0.1079, ..., -0.0943, 0.0038, 0.0835]], + device='cuda:0'), grad: tensor([[ 1.5367e-08, 0.0000e+00, -3.2596e-08, ..., 9.1270e-08, + 3.4459e-08, 5.1642e-07], + [ 2.5611e-08, 0.0000e+00, 4.6007e-07, ..., -1.0747e-06, + 2.3982e-07, 1.7835e-07], + [ 2.3283e-08, 0.0000e+00, 9.0059e-07, ..., 6.3796e-07, + 5.2014e-07, 2.5239e-07], + ..., + [ 5.8208e-08, 0.0000e+00, -1.9446e-06, ..., 2.4494e-07, + -1.0394e-06, 2.3516e-07], + [ 2.1420e-08, 0.0000e+00, -7.9162e-09, ..., 7.5437e-08, + 1.0943e-07, 2.1029e-06], + [ 9.6392e-08, 0.0000e+00, 2.2398e-07, ..., 6.7987e-08, + 1.6717e-07, -8.6054e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0349, -0.0330, 0.0152, 0.0321, -0.0231, 0.0014, -0.0096, 0.0141, + 0.0150, -0.0023], device='cuda:0'), grad: tensor([ 1.7826e-06, -5.3691e-07, 7.3239e-06, 1.4372e-05, 3.9488e-06, + -1.1586e-06, 3.1758e-07, -2.3115e-06, -3.7234e-06, -2.0102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 214.15, cls_loss 0.0016 cls_loss_mapping 0.0062 cls_loss_causal 0.5251 re_mapping 0.0057 re_causal 0.0176 /// teacc 98.99 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0629, 0.0277, 0.0273, ..., -0.0603, -0.0654, 0.0589], + [-0.0505, -0.0115, -0.0766, ..., 0.1341, -0.0662, -0.0509], + [-0.0165, -0.0187, 0.1688, ..., -0.0896, -0.1123, -0.0834], + ..., + [-0.0605, -0.0494, 0.0024, ..., -0.0146, 0.0906, -0.0341], + [-0.0722, -0.0534, -0.0470, ..., -0.1037, 0.0020, -0.1166], + [-0.1480, -0.0314, -0.1085, ..., -0.0948, 0.0035, 0.0836]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, 1.4901e-08, -2.2855e-06, ..., 1.5832e-07, + 1.0198e-07, -2.4773e-06], + [ 2.2817e-08, 1.3970e-08, 1.3914e-06, ..., -4.3958e-07, + 5.8301e-07, 2.5332e-07], + [ 3.0268e-08, 5.5879e-09, -4.0859e-05, ..., 2.4447e-07, + 7.6042e-07, 5.6205e-07], + ..., + [ 1.8626e-08, 8.5169e-07, -4.4741e-06, ..., 1.2806e-07, + -3.5875e-06, 9.2536e-06], + [ 2.0489e-08, 6.0536e-09, 3.7819e-05, ..., 2.3935e-07, + -4.3353e-07, 2.2771e-07], + [ 1.2806e-07, -1.0896e-06, 2.4326e-06, ..., 2.8871e-08, + 8.9686e-07, -1.0930e-05]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0343, -0.0329, 0.0142, 0.0320, -0.0231, 0.0016, -0.0096, 0.0151, + 0.0144, -0.0023], device='cuda:0'), grad: tensor([-1.5777e-06, 2.9057e-06, -6.1452e-05, 9.5293e-06, 2.7977e-06, + 6.2883e-06, 7.2159e-06, 1.0565e-05, 4.2349e-05, -1.8582e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 214.27, cls_loss 0.0015 cls_loss_mapping 0.0045 cls_loss_causal 0.5166 re_mapping 0.0058 re_causal 0.0176 /// teacc 99.13 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0632, 0.0277, 0.0278, ..., -0.0604, -0.0660, 0.0590], + [-0.0506, -0.0117, -0.0766, ..., 0.1347, -0.0664, -0.0512], + [-0.0167, -0.0189, 0.1682, ..., -0.0901, -0.1131, -0.0846], + ..., + [-0.0607, -0.0492, 0.0028, ..., -0.0148, 0.0911, -0.0343], + [-0.0725, -0.0541, -0.0453, ..., -0.1043, 0.0017, -0.1159], + [-0.1481, -0.0312, -0.1088, ..., -0.0951, 0.0033, 0.0841]], + device='cuda:0'), grad: tensor([[ 1.3597e-07, -9.3132e-10, 6.9942e-07, ..., 3.4459e-08, + 1.7975e-07, 5.8208e-08], + [ 5.8254e-07, 0.0000e+00, 1.7788e-06, ..., -7.0920e-07, + 6.0583e-07, 2.3516e-07], + [ 1.6857e-07, 0.0000e+00, -7.7337e-06, ..., 4.0513e-07, + 1.7975e-07, 7.9162e-08], + ..., + [ 1.1800e-06, -0.0000e+00, 8.1025e-07, ..., 1.0710e-07, + 1.9968e-06, 2.6524e-06], + [ 1.5600e-07, 0.0000e+00, 3.4720e-06, ..., 4.0513e-08, + -2.8461e-06, 6.1048e-07], + [ 1.7295e-06, 9.3132e-10, 1.0431e-07, ..., 1.9558e-08, + 5.6298e-07, -4.9062e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0343, -0.0328, 0.0132, 0.0316, -0.0233, 0.0017, -0.0093, 0.0154, + 0.0161, -0.0022], device='cuda:0'), grad: tensor([ 1.8310e-06, 4.0941e-06, -1.0230e-05, 3.6173e-06, -7.7486e-07, + -5.9530e-06, 2.8405e-06, 1.3538e-05, -1.1176e-08, -9.0152e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 214.19, cls_loss 0.0015 cls_loss_mapping 0.0036 cls_loss_causal 0.5228 re_mapping 0.0055 re_causal 0.0172 /// teacc 99.10 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0635, 0.0298, 0.0282, ..., -0.0608, -0.0669, 0.0593], + [-0.0507, -0.0118, -0.0765, ..., 0.1357, -0.0665, -0.0515], + [-0.0167, -0.0192, 0.1686, ..., -0.0907, -0.1130, -0.0854], + ..., + [-0.0611, -0.0495, 0.0027, ..., -0.0153, 0.0912, -0.0348], + [-0.0730, -0.0553, -0.0458, ..., -0.1044, 0.0018, -0.1160], + [-0.1483, -0.0320, -0.1093, ..., -0.0957, 0.0026, 0.0838]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, -1.5879e-07, -4.9500e-07, ..., 1.9558e-08, + 2.5798e-07, 2.9942e-07], + [ 9.3132e-09, 2.0023e-08, 1.2927e-06, ..., -1.2806e-07, + 1.2107e-06, 5.1484e-06], + [ 3.2596e-09, 2.4680e-08, -1.1530e-06, ..., 3.1665e-08, + 5.4669e-07, 9.3319e-07], + ..., + [ 5.3551e-08, 2.3283e-09, -4.5672e-06, ..., 7.2177e-08, + -4.9360e-06, 7.1935e-06], + [ 3.2596e-09, 9.7789e-09, 6.6916e-07, ..., 2.4680e-08, + 2.9104e-07, 3.7216e-06], + [ 6.0536e-09, 4.4238e-08, 7.7626e-07, ..., 1.0245e-08, + 1.0645e-06, -4.5985e-05]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0343, -0.0323, 0.0131, 0.0315, -0.0234, 0.0024, -0.0100, 0.0150, + 0.0166, -0.0029], device='cuda:0'), grad: tensor([ 1.0040e-06, 1.3866e-05, 2.9635e-06, 1.3284e-05, 6.7428e-06, + 2.9907e-05, 1.0822e-06, 8.5980e-06, 9.7081e-06, -8.7082e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 214.53, cls_loss 0.0014 cls_loss_mapping 0.0038 cls_loss_causal 0.5122 re_mapping 0.0056 re_causal 0.0171 /// teacc 99.09 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0644, 0.0297, 0.0281, ..., -0.0612, -0.0680, 0.0592], + [-0.0508, -0.0124, -0.0768, ..., 0.1364, -0.0668, -0.0517], + [-0.0170, -0.0192, 0.1690, ..., -0.0907, -0.1133, -0.0862], + ..., + [-0.0616, -0.0495, 0.0029, ..., -0.0162, 0.0917, -0.0349], + [-0.0736, -0.0557, -0.0463, ..., -0.1046, 0.0016, -0.1164], + [-0.1489, -0.0321, -0.1094, ..., -0.0963, 0.0024, 0.0841]], + device='cuda:0'), grad: tensor([[ 2.0750e-06, -4.1910e-09, 1.3681e-06, ..., 4.3362e-06, + 7.8231e-08, 3.4273e-06], + [ 6.7055e-08, 0.0000e+00, 1.2387e-07, ..., -8.8476e-09, + 2.9569e-07, 1.3178e-07], + [ 5.2620e-08, 4.6566e-10, -4.8429e-07, ..., 8.1025e-08, + 2.4214e-08, 7.8697e-08], + ..., + [ 4.7032e-08, 0.0000e+00, -1.4715e-07, ..., 1.5832e-08, + -5.8673e-07, 6.8173e-07], + [ 3.4645e-07, 0.0000e+00, -1.2871e-06, ..., 6.2445e-07, + 3.8091e-07, 1.3644e-06], + [ 7.6834e-08, 1.3970e-09, 1.3877e-07, ..., 4.6100e-08, + 1.4016e-07, -3.1926e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0349, -0.0322, 0.0132, 0.0315, -0.0232, 0.0024, -0.0102, 0.0150, + 0.0166, -0.0030], device='cuda:0'), grad: tensor([ 2.4021e-05, 1.1530e-06, 9.5926e-08, -5.8487e-06, 4.1351e-06, + 1.1232e-06, -2.0966e-05, 6.0908e-07, -5.4482e-08, -4.2655e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 214.54, cls_loss 0.0012 cls_loss_mapping 0.0033 cls_loss_causal 0.5003 re_mapping 0.0059 re_causal 0.0176 /// teacc 99.00 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0644, 0.0308, 0.0280, ..., -0.0616, -0.0686, 0.0591], + [-0.0509, -0.0130, -0.0770, ..., 0.1367, -0.0672, -0.0522], + [-0.0171, -0.0191, 0.1696, ..., -0.0908, -0.1133, -0.0865], + ..., + [-0.0620, -0.0495, 0.0027, ..., -0.0165, 0.0931, -0.0350], + [-0.0746, -0.0561, -0.0470, ..., -0.1049, 0.0010, -0.1172], + [-0.1490, -0.0322, -0.1095, ..., -0.0965, 0.0023, 0.0847]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 0.0000e+00, 3.1199e-08, ..., -4.6566e-10, + 1.2387e-07, -8.8476e-08], + [ 2.2445e-07, 0.0000e+00, 9.8348e-07, ..., -4.4703e-08, + 5.6531e-07, 1.4901e-08], + [ 1.3597e-07, 0.0000e+00, 7.1293e-07, ..., 2.7940e-08, + 7.4413e-07, 2.0489e-08], + ..., + [ 4.9826e-07, 0.0000e+00, -1.4547e-06, ..., 4.9360e-08, + -2.5518e-06, 2.0443e-07], + [ 5.5414e-08, 0.0000e+00, 1.0980e-06, ..., 1.0245e-08, + 8.1444e-07, 4.4238e-08], + [ 2.0023e-08, 0.0000e+00, 4.3586e-07, ..., 5.5879e-09, + 3.4506e-07, -2.3516e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0351, -0.0322, 0.0133, 0.0311, -0.0233, 0.0024, -0.0102, 0.0156, + 0.0159, -0.0026], device='cuda:0'), grad: tensor([ 9.5228e-07, 2.9150e-06, 8.0168e-06, -1.5318e-05, -1.5274e-06, + -6.8955e-06, 2.5667e-06, -1.6624e-07, 8.4937e-06, 8.9128e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 214.57, cls_loss 0.0011 cls_loss_mapping 0.0028 cls_loss_causal 0.5258 re_mapping 0.0054 re_causal 0.0176 /// teacc 99.02 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0648, 0.0309, 0.0285, ..., -0.0617, -0.0690, 0.0594], + [-0.0511, -0.0131, -0.0772, ..., 0.1371, -0.0674, -0.0522], + [-0.0171, -0.0191, 0.1697, ..., -0.0909, -0.1134, -0.0870], + ..., + [-0.0622, -0.0494, 0.0027, ..., -0.0170, 0.0933, -0.0352], + [-0.0749, -0.0561, -0.0469, ..., -0.1050, 0.0008, -0.1175], + [-0.1489, -0.0322, -0.1096, ..., -0.0967, 0.0023, 0.0855]], + device='cuda:0'), grad: tensor([[ 2.1663e-06, 2.3283e-09, -1.2852e-07, ..., 9.2015e-07, + 9.3598e-08, -2.9616e-07], + [ 6.3896e-04, 2.7940e-09, 1.5739e-07, ..., 2.3985e-04, + 7.0315e-08, 3.8184e-08], + [ 1.8589e-06, -1.6298e-08, -5.8068e-07, ..., 6.9803e-07, + 2.1188e-07, 8.5682e-08], + ..., + [ 1.4164e-05, 4.1910e-09, -3.5856e-08, ..., 5.1036e-06, + -2.7055e-07, 2.3749e-08], + [ 2.0593e-05, 9.3132e-10, 5.7276e-08, ..., 7.6368e-06, + 1.6112e-07, 2.0536e-07], + [ 6.0946e-06, 0.0000e+00, 1.1222e-07, ..., 1.7006e-06, + 1.5600e-07, 1.9278e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0349, -0.0322, 0.0132, 0.0312, -0.0238, 0.0024, -0.0102, 0.0154, + 0.0161, -0.0020], device='cuda:0'), grad: tensor([ 5.9195e-06, 1.5030e-03, 5.2415e-06, -8.9854e-06, -1.6298e-03, + 5.9791e-06, 1.9357e-05, 3.2991e-05, 4.9919e-05, 1.5110e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 214.38, cls_loss 0.0017 cls_loss_mapping 0.0061 cls_loss_causal 0.5358 re_mapping 0.0056 re_causal 0.0174 /// teacc 99.14 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0655, 0.0309, 0.0286, ..., -0.0623, -0.0694, 0.0594], + [-0.0524, -0.0132, -0.0777, ..., 0.1373, -0.0683, -0.0524], + [-0.0172, -0.0191, 0.1700, ..., -0.0911, -0.1136, -0.0877], + ..., + [-0.0629, -0.0496, 0.0030, ..., -0.0169, 0.0938, -0.0358], + [-0.0756, -0.0562, -0.0475, ..., -0.1058, 0.0006, -0.1179], + [-0.1483, -0.0322, -0.1092, ..., -0.0977, 0.0019, 0.0868]], + device='cuda:0'), grad: tensor([[ 2.0768e-07, 0.0000e+00, 8.5235e-06, ..., 7.2084e-07, + 5.1223e-09, 1.6391e-07], + [ 2.2352e-07, 0.0000e+00, 2.5053e-07, ..., 3.6508e-07, + 1.2573e-08, 5.9139e-08], + [ 1.7555e-07, 0.0000e+00, -2.8357e-05, ..., 5.3551e-07, + 2.7940e-09, -6.8918e-07], + ..., + [ 7.3574e-08, 0.0000e+00, 1.4246e-05, ..., 1.2759e-07, + 3.2596e-09, 3.7998e-07], + [ 2.5844e-07, 0.0000e+00, 2.1812e-06, ..., 8.5961e-07, + 1.1642e-08, 9.9186e-08], + [ 5.1921e-07, 0.0000e+00, 1.2452e-06, ..., 1.5460e-07, + 4.6566e-09, -6.5193e-08]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0353, -0.0325, 0.0131, 0.0314, -0.0242, 0.0024, -0.0105, 0.0150, + 0.0157, -0.0009], device='cuda:0'), grad: tensor([ 1.8060e-05, 7.1302e-06, -3.1352e-05, 4.6566e-06, 2.3209e-06, + 1.4216e-05, 2.4885e-06, 1.9908e-05, -4.2170e-05, 4.7721e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 214.36, cls_loss 0.0019 cls_loss_mapping 0.0046 cls_loss_causal 0.5236 re_mapping 0.0054 re_causal 0.0174 /// teacc 99.05 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0679, 0.0309, 0.0288, ..., -0.0628, -0.0702, 0.0594], + [-0.0525, -0.0132, -0.0788, ..., 0.1385, -0.0669, -0.0525], + [-0.0173, -0.0190, 0.1704, ..., -0.0912, -0.1138, -0.0884], + ..., + [-0.0635, -0.0496, 0.0037, ..., -0.0181, 0.0920, -0.0360], + [-0.0762, -0.0564, -0.0477, ..., -0.1063, 0.0002, -0.1183], + [-0.1504, -0.0323, -0.1096, ..., -0.0986, 0.0017, 0.0867]], + device='cuda:0'), grad: tensor([[ 4.5123e-07, -3.1758e-07, -6.9737e-06, ..., 1.1455e-07, + 7.1945e-07, -3.2634e-06], + [ 2.9011e-07, 5.4948e-08, 5.8450e-06, ..., -1.7229e-08, + 2.9951e-06, 6.1952e-06], + [ 3.0883e-06, 5.8673e-08, 1.1817e-05, ..., 1.4575e-07, + 5.3719e-06, 1.1139e-06], + ..., + [-1.6108e-05, 3.3993e-08, -7.1585e-05, ..., 1.7695e-08, + -3.5524e-05, 4.3726e-07], + [ 6.3144e-07, 2.7940e-08, 3.0175e-06, ..., 1.4482e-07, + 6.5519e-07, 7.5139e-06], + [-2.8033e-07, 6.1467e-08, 8.5011e-06, ..., 6.0536e-09, + 4.3474e-06, -2.0206e-05]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0356, -0.0306, 0.0131, 0.0314, -0.0228, 0.0023, -0.0098, 0.0129, + 0.0159, -0.0022], device='cuda:0'), grad: tensor([-1.3679e-05, 2.7880e-05, 2.7031e-05, 7.5102e-06, 9.8825e-05, + 6.1281e-06, 2.6589e-07, -1.5247e-04, 2.1711e-05, -2.3469e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 214.72, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5165 re_mapping 0.0057 re_causal 0.0172 /// teacc 99.03 lr 0.00010000 +Epoch 161, weight, value: tensor([[-6.8279e-02, 3.0579e-02, 2.9146e-02, ..., -6.3350e-02, + -7.1427e-02, 5.9344e-02], + [-5.2589e-02, -1.3298e-02, -7.9189e-02, ..., 1.3854e-01, + -6.6791e-02, -5.2733e-02], + [-1.7388e-02, -1.8905e-02, 1.7098e-01, ..., -9.1212e-02, + -1.1421e-01, -8.8876e-02], + ..., + [-6.3684e-02, -4.9622e-02, 3.8739e-03, ..., -1.8021e-02, + 9.2139e-02, -3.5906e-02], + [-7.6401e-02, -5.7245e-02, -4.8480e-02, ..., -1.0694e-01, + -6.7365e-05, -1.1950e-01], + [-1.5077e-01, -3.2491e-02, -1.1020e-01, ..., -9.9275e-02, + 1.1000e-03, 8.7266e-02]], device='cuda:0'), grad: tensor([[ 2.0955e-08, 0.0000e+00, 2.1886e-08, ..., 2.4680e-08, + 5.4482e-08, 8.7544e-08], + [ 1.8161e-07, 0.0000e+00, 2.8079e-07, ..., -8.7079e-08, + 3.4273e-07, 2.4168e-07], + [ 7.2643e-08, 0.0000e+00, -4.1910e-09, ..., 5.5879e-08, + 3.0082e-07, 7.6368e-08], + ..., + [ 3.8324e-07, 0.0000e+00, -7.4413e-07, ..., 8.3353e-08, + -7.3109e-07, 2.4913e-07], + [ 5.1688e-08, 0.0000e+00, 1.0245e-07, ..., 4.2375e-08, + 2.1607e-07, 2.5937e-07], + [ 8.7544e-07, 0.0000e+00, 3.3993e-08, ..., 6.0070e-08, + 3.2177e-07, -6.0583e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0360, -0.0307, 0.0133, 0.0315, -0.0227, 0.0026, -0.0107, 0.0131, + 0.0153, -0.0024], device='cuda:0'), grad: tensor([ 6.5705e-07, 2.4606e-06, 1.9763e-06, -1.3143e-05, 1.4622e-07, + -4.5486e-06, 2.3823e-06, 1.3690e-07, -3.0994e-06, 1.3031e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 214.46, cls_loss 0.0012 cls_loss_mapping 0.0044 cls_loss_causal 0.5080 re_mapping 0.0055 re_causal 0.0175 /// teacc 99.14 lr 0.00010000 +Epoch 162, weight, value: tensor([[-6.8843e-02, 3.0587e-02, 2.9510e-02, ..., -6.3816e-02, + -7.1728e-02, 5.9449e-02], + [-5.2665e-02, -1.3302e-02, -7.9312e-02, ..., 1.3877e-01, + -6.6852e-02, -5.3085e-02], + [-1.7459e-02, -1.8901e-02, 1.7113e-01, ..., -9.1352e-02, + -1.1448e-01, -8.9509e-02], + ..., + [-6.4011e-02, -4.9622e-02, 4.0680e-03, ..., -1.8030e-02, + 9.2546e-02, -3.5394e-02], + [-7.6856e-02, -5.7283e-02, -4.8657e-02, ..., -1.0736e-01, + -3.7131e-04, -1.2013e-01], + [-1.5100e-01, -3.2489e-02, -1.1037e-01, ..., -9.9880e-02, + -1.0259e-04, 8.7237e-02]], device='cuda:0'), grad: tensor([[ 1.7229e-08, 0.0000e+00, -3.1851e-07, ..., 6.9849e-09, + 5.3085e-08, -2.6170e-07], + [ 2.6636e-07, 0.0000e+00, 1.0151e-07, ..., -4.6566e-10, + 6.9803e-07, 6.0536e-08], + [ 5.3085e-08, 0.0000e+00, -1.0058e-06, ..., 2.4214e-08, + 8.9873e-08, 9.6858e-08], + ..., + [-2.8173e-07, 0.0000e+00, 3.5344e-07, ..., -2.7381e-07, + -2.4829e-06, 1.3644e-07], + [ 9.6392e-08, 0.0000e+00, 1.0198e-07, ..., 2.4214e-08, + 2.7008e-07, 2.6263e-07], + [ 3.9209e-07, 0.0000e+00, 1.7416e-07, ..., 3.4459e-08, + 3.9302e-07, -2.0023e-08]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0360, -0.0308, 0.0132, 0.0314, -0.0228, 0.0028, -0.0106, 0.0138, + 0.0151, -0.0031], device='cuda:0'), grad: tensor([-2.0955e-07, 2.5760e-06, -7.5530e-07, 1.5832e-06, 1.3486e-06, + -1.1474e-06, 8.3214e-07, -6.9961e-06, 1.1548e-06, 1.6149e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 214.44, cls_loss 0.0013 cls_loss_mapping 0.0038 cls_loss_causal 0.5454 re_mapping 0.0054 re_causal 0.0169 /// teacc 99.09 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0692, 0.0307, 0.0300, ..., -0.0643, -0.0722, 0.0596], + [-0.0528, -0.0133, -0.0787, ..., 0.1398, -0.0669, -0.0535], + [-0.0175, -0.0189, 0.1710, ..., -0.0924, -0.1146, -0.0900], + ..., + [-0.0643, -0.0497, 0.0039, ..., -0.0183, 0.0927, -0.0355], + [-0.0767, -0.0575, -0.0490, ..., -0.1076, -0.0008, -0.1205], + [-0.1512, -0.0327, -0.1104, ..., -0.1005, -0.0003, 0.0875]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, 0.0000e+00, -3.7719e-08, ..., 5.7882e-07, + 1.5274e-07, 1.6913e-06], + [ 1.4156e-07, 0.0000e+00, 2.0452e-06, ..., -1.4147e-06, + 1.0831e-06, 4.9593e-07], + [ 6.0070e-08, 0.0000e+00, 5.7463e-07, ..., 8.6753e-07, + 2.6962e-07, 4.6566e-07], + ..., + [ 1.1735e-07, 0.0000e+00, -5.5172e-06, ..., 7.4413e-07, + -2.8014e-06, 4.5588e-07], + [ 9.3598e-08, 0.0000e+00, 4.1910e-08, ..., 9.2713e-07, + 1.0291e-07, 7.5810e-07], + [ 9.9838e-07, 0.0000e+00, 1.4147e-06, ..., 2.3888e-07, + 7.7207e-07, -1.1697e-05]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0360, -0.0306, 0.0126, 0.0316, -0.0228, 0.0027, -0.0107, 0.0137, + 0.0153, -0.0031], device='cuda:0'), grad: tensor([ 4.8578e-06, 5.2042e-06, 4.8503e-06, 7.2382e-06, 1.8235e-06, + 1.2219e-05, -1.4603e-05, -1.0476e-05, 3.7737e-06, -1.4931e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 214.74, cls_loss 0.0012 cls_loss_mapping 0.0038 cls_loss_causal 0.5392 re_mapping 0.0056 re_causal 0.0176 /// teacc 99.06 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0692, 0.0313, 0.0320, ..., -0.0647, -0.0725, 0.0604], + [-0.0529, -0.0135, -0.0789, ..., 0.1401, -0.0670, -0.0538], + [-0.0176, -0.0186, 0.1713, ..., -0.0926, -0.1148, -0.0916], + ..., + [-0.0652, -0.0499, 0.0039, ..., -0.0184, 0.0928, -0.0357], + [-0.0772, -0.0583, -0.0497, ..., -0.1083, -0.0007, -0.1222], + [-0.1515, -0.0332, -0.1112, ..., -0.1008, -0.0006, 0.0878]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -3.5334e-06, ..., 5.9977e-07, + 4.2887e-07, -1.6037e-06], + [ 1.8626e-09, 0.0000e+00, 1.2992e-06, ..., -7.7439e-07, + 1.4547e-06, 8.6147e-08], + [ 4.6566e-10, 0.0000e+00, 6.4913e-07, ..., 2.0955e-07, + 3.5111e-07, 1.9511e-07], + ..., + [ 4.1910e-09, 0.0000e+00, -4.0978e-06, ..., 3.7253e-07, + -4.3735e-06, 2.1281e-07], + [ 0.0000e+00, 0.0000e+00, 1.8207e-06, ..., 8.6892e-07, + 3.6461e-07, 1.2126e-06], + [ 4.1910e-09, 0.0000e+00, 1.8906e-06, ..., 1.2387e-07, + 5.2107e-07, 6.4773e-07]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0349, -0.0306, 0.0126, 0.0315, -0.0226, 0.0029, -0.0108, 0.0137, + 0.0148, -0.0033], device='cuda:0'), grad: tensor([-3.7160e-06, 2.7753e-06, 2.0117e-06, 7.2382e-06, 5.8021e-07, + 6.7465e-06, -1.3962e-05, -1.0207e-05, 4.5300e-06, 4.0159e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 214.49, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5046 re_mapping 0.0054 re_causal 0.0166 /// teacc 99.10 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0697, 0.0315, 0.0320, ..., -0.0650, -0.0731, 0.0605], + [-0.0532, -0.0135, -0.0790, ..., 0.1403, -0.0671, -0.0541], + [-0.0174, -0.0186, 0.1719, ..., -0.0926, -0.1148, -0.0919], + ..., + [-0.0659, -0.0499, 0.0038, ..., -0.0185, 0.0929, -0.0359], + [-0.0778, -0.0584, -0.0502, ..., -0.1086, -0.0005, -0.1226], + [-0.1516, -0.0333, -0.1115, ..., -0.1013, -0.0007, 0.0881]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 8.9407e-08, -6.5193e-07, ..., 2.2352e-08, + 3.3341e-07, -2.7195e-07], + [ 2.4587e-07, 5.0291e-08, 3.3528e-07, ..., -1.2834e-06, + 1.6112e-06, 4.9546e-07], + [ 1.4342e-07, 1.7881e-07, -1.9185e-07, ..., 1.9558e-07, + 3.7067e-07, 5.5134e-07], + ..., + [ 5.5507e-07, 2.2352e-07, -8.9779e-07, ..., 9.6112e-07, + -5.4948e-06, 8.2515e-07], + [ 8.1956e-08, 7.8231e-08, 2.6822e-07, ..., 1.6764e-07, + 1.7677e-06, 1.1958e-06], + [ 2.8387e-06, 7.0781e-08, 5.4576e-07, ..., 8.7544e-08, + 1.4156e-06, 6.0722e-07]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0351, -0.0306, 0.0128, 0.0314, -0.0226, 0.0031, -0.0112, 0.0135, + 0.0146, -0.0032], device='cuda:0'), grad: tensor([ 2.1085e-06, 6.0499e-06, 5.6960e-06, 1.8924e-05, -3.3118e-06, + -7.5102e-05, 4.0948e-05, -3.0212e-06, -4.1574e-06, 1.1809e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.53, cls_loss 0.0014 cls_loss_mapping 0.0043 cls_loss_causal 0.5109 re_mapping 0.0055 re_causal 0.0165 /// teacc 98.97 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0700, 0.0317, 0.0332, ..., -0.0646, -0.0730, 0.0616], + [-0.0534, -0.0138, -0.0794, ..., 0.1408, -0.0671, -0.0549], + [-0.0176, -0.0186, 0.1728, ..., -0.0927, -0.1152, -0.0925], + ..., + [-0.0666, -0.0499, 0.0039, ..., -0.0189, 0.0931, -0.0360], + [-0.0783, -0.0594, -0.0506, ..., -0.1093, -0.0008, -0.1247], + [-0.1529, -0.0342, -0.1121, ..., -0.1025, -0.0012, 0.0883]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.7253e-09, -1.8626e-08, ..., -7.2643e-08, + 2.5705e-07, 3.7067e-07], + [ 3.5390e-08, 3.7253e-09, 4.2841e-08, ..., -5.4203e-07, + 3.9861e-07, 9.7975e-07], + [ 1.8626e-09, 0.0000e+00, -7.6368e-08, ..., 1.0058e-07, + 1.7136e-07, 2.6263e-07], + ..., + [ 3.6135e-07, 9.6858e-08, -2.2352e-08, ..., 3.0175e-07, + 5.0291e-07, 6.1244e-06], + [ 9.3132e-09, 1.8626e-09, 3.7253e-09, ..., 1.0431e-07, + 9.9838e-07, 1.5795e-06], + [-6.8545e-07, -1.9558e-07, 2.2352e-08, ..., 6.5193e-08, + 2.7232e-06, -5.9046e-06]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0339, -0.0306, 0.0131, 0.0315, -0.0219, 0.0032, -0.0121, 0.0134, + 0.0137, -0.0037], device='cuda:0'), grad: tensor([ 1.0245e-06, 4.5560e-06, 1.2144e-06, 8.1509e-06, 1.7434e-05, + -2.8297e-05, 2.8275e-06, 2.4557e-05, -1.3430e-06, -3.0130e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 214.33, cls_loss 0.0011 cls_loss_mapping 0.0026 cls_loss_causal 0.5184 re_mapping 0.0056 re_causal 0.0175 /// teacc 98.99 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0702, 0.0314, 0.0332, ..., -0.0652, -0.0734, 0.0616], + [-0.0534, -0.0139, -0.0793, ..., 0.1419, -0.0671, -0.0550], + [-0.0176, -0.0186, 0.1728, ..., -0.0937, -0.1158, -0.0924], + ..., + [-0.0671, -0.0500, 0.0041, ..., -0.0193, 0.0932, -0.0362], + [-0.0785, -0.0599, -0.0506, ..., -0.1100, -0.0012, -0.1250], + [-0.1535, -0.0320, -0.1124, ..., -0.1030, -0.0015, 0.0884]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 0.0000e+00, -4.3452e-05, ..., -3.7253e-08, + 5.5879e-09, -2.3007e-05], + [ 1.6764e-08, 0.0000e+00, 2.4401e-07, ..., -2.0489e-08, + 4.4703e-08, 8.1956e-08], + [ 7.2643e-08, 0.0000e+00, 1.2871e-06, ..., 1.8626e-08, + -2.6077e-08, 1.1045e-06], + ..., + [ 5.0291e-08, 0.0000e+00, 1.4715e-06, ..., 1.1176e-08, + -1.6205e-07, 3.9116e-07], + [ 1.1176e-08, 0.0000e+00, 5.4613e-06, ..., 1.4901e-08, + 3.5390e-08, 2.3209e-06], + [ 3.5390e-08, 0.0000e+00, 2.3991e-05, ..., 1.1176e-08, + 3.7253e-08, 9.3430e-06]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0344, -0.0303, 0.0128, 0.0316, -0.0215, 0.0032, -0.0119, 0.0133, + 0.0139, -0.0042], device='cuda:0'), grad: tensor([-8.3268e-05, 6.7987e-07, 3.6713e-06, 1.8813e-06, 1.6205e-07, + 1.7285e-05, 2.0526e-06, 2.9895e-06, 9.8348e-06, 4.4584e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 214.29, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.5018 re_mapping 0.0055 re_causal 0.0167 /// teacc 98.66 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0702, 0.0312, 0.0335, ..., -0.0652, -0.0737, 0.0619], + [-0.0535, -0.0139, -0.0794, ..., 0.1425, -0.0672, -0.0552], + [-0.0176, -0.0186, 0.1734, ..., -0.0938, -0.1157, -0.0928], + ..., + [-0.0672, -0.0500, 0.0040, ..., -0.0198, 0.0934, -0.0362], + [-0.0782, -0.0607, -0.0509, ..., -0.1107, -0.0016, -0.1257], + [-0.1537, -0.0320, -0.1127, ..., -0.1037, -0.0018, 0.0891]], + device='cuda:0'), grad: tensor([[ 5.9977e-07, 0.0000e+00, 8.5682e-08, ..., 1.0878e-06, + 9.3132e-09, 3.6694e-07], + [ 4.7684e-07, 0.0000e+00, 2.5705e-07, ..., -4.3139e-06, + 2.0489e-07, 2.8871e-07], + [ 3.2783e-07, 0.0000e+00, -1.1772e-06, ..., 4.7311e-07, + 2.4214e-08, 8.7544e-08], + ..., + [ 3.0845e-06, 0.0000e+00, 1.3597e-07, ..., 3.7812e-06, + -2.9802e-07, 4.7572e-06], + [ 7.3016e-07, 0.0000e+00, 1.4715e-07, ..., 4.8615e-07, + 2.0489e-08, 5.8301e-07], + [-3.2872e-05, 0.0000e+00, 3.7253e-08, ..., 1.5646e-07, + 2.2352e-08, -5.0694e-05]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0343, -0.0302, 0.0130, 0.0316, -0.0219, 0.0031, -0.0118, 0.0133, + 0.0135, -0.0040], device='cuda:0'), grad: tensor([ 3.6620e-06, -7.2308e-06, -4.6566e-07, 2.3097e-06, 1.5235e-04, + 3.6303e-06, -4.3958e-05, 2.0415e-05, 2.8312e-06, -1.3340e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 214.54, cls_loss 0.0018 cls_loss_mapping 0.0047 cls_loss_causal 0.5254 re_mapping 0.0053 re_causal 0.0166 /// teacc 99.12 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0710, 0.0323, 0.0320, ..., -0.0655, -0.0739, 0.0604], + [-0.0537, -0.0142, -0.0795, ..., 0.1429, -0.0672, -0.0574], + [-0.0178, -0.0185, 0.1740, ..., -0.0943, -0.1158, -0.0934], + ..., + [-0.0675, -0.0489, 0.0042, ..., -0.0200, 0.0936, -0.0367], + [-0.0789, -0.0620, -0.0514, ..., -0.1110, -0.0013, -0.1262], + [-0.1518, -0.0332, -0.1116, ..., -0.1046, -0.0018, 0.0924]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -3.9302e-07, ..., 3.7253e-08, + 4.6566e-08, -5.8599e-06], + [ 1.6764e-08, 0.0000e+00, 6.6869e-07, ..., -1.5032e-06, + 8.8848e-07, 4.4703e-08], + [ 2.7940e-08, 0.0000e+00, -3.5595e-06, ..., 8.7544e-08, + 2.3097e-07, 2.4214e-08], + ..., + [ 5.5879e-09, 1.8626e-09, 2.3469e-07, ..., 6.3889e-07, + -6.9030e-06, 3.7253e-08], + [ 3.7253e-09, 0.0000e+00, 3.5949e-07, ..., 3.2037e-07, + 1.7509e-07, 4.6566e-08], + [ 2.4587e-07, -1.4901e-08, 3.8370e-07, ..., 1.5646e-07, + 4.0755e-06, 1.0803e-07]], device='cuda:0') +Epoch 169, bias, value: tensor([-3.6021e-02, -3.0319e-02, 1.2927e-02, 3.1843e-02, -2.5279e-02, + 2.8745e-03, -1.1649e-02, 1.3134e-02, 1.3757e-02, 9.2690e-05], + device='cuda:0'), grad: tensor([-7.4208e-06, 2.1793e-07, -4.8839e-06, 3.6918e-06, 2.3358e-06, + 4.4703e-06, 5.1484e-06, -1.5914e-05, -4.3027e-07, 1.2755e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 214.12, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5571 re_mapping 0.0056 re_causal 0.0170 /// teacc 99.06 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0718, 0.0323, 0.0316, ..., -0.0636, -0.0747, 0.0600], + [-0.0539, -0.0144, -0.0800, ..., 0.1434, -0.0676, -0.0575], + [-0.0179, -0.0184, 0.1743, ..., -0.0945, -0.1164, -0.0943], + ..., + [-0.0677, -0.0489, 0.0052, ..., -0.0197, 0.0942, -0.0368], + [-0.0793, -0.0623, -0.0519, ..., -0.1135, -0.0021, -0.1278], + [-0.1516, -0.0332, -0.1114, ..., -0.1103, -0.0021, 0.0934]], + device='cuda:0'), grad: tensor([[ 4.2841e-08, 0.0000e+00, 1.2740e-06, ..., 4.4703e-08, + 5.5879e-09, -2.3283e-07], + [ 2.5891e-07, 0.0000e+00, 1.0878e-06, ..., -4.1164e-07, + 1.3039e-08, 1.3039e-08], + [ 4.2841e-08, 0.0000e+00, -2.1964e-05, ..., 6.3330e-08, + 1.8626e-09, -1.8626e-08], + ..., + [ 2.0303e-07, 0.0000e+00, 2.1681e-06, ..., 1.8440e-07, + 1.3039e-08, 2.4214e-08], + [ 8.3819e-08, 0.0000e+00, 1.3024e-05, ..., 1.3225e-07, + -6.7055e-08, 2.4214e-08], + [ 2.2482e-06, -1.8626e-09, 4.8243e-07, ..., 1.3411e-07, + 6.3330e-08, 1.8626e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0362, -0.0304, 0.0128, 0.0317, -0.0255, 0.0028, -0.0112, 0.0136, + 0.0127, 0.0003], device='cuda:0'), grad: tensor([ 3.1367e-06, 2.6543e-06, -5.3376e-05, 3.8706e-06, -3.6955e-06, + 4.1537e-07, 4.2170e-06, 6.0685e-06, 3.0994e-05, 5.7518e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.31, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5443 re_mapping 0.0057 re_causal 0.0171 /// teacc 99.06 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0725, 0.0322, 0.0318, ..., -0.0637, -0.0757, 0.0601], + [-0.0541, -0.0147, -0.0802, ..., 0.1438, -0.0677, -0.0577], + [-0.0180, -0.0184, 0.1761, ..., -0.0947, -0.1161, -0.0928], + ..., + [-0.0688, -0.0490, 0.0044, ..., -0.0202, 0.0945, -0.0371], + [-0.0796, -0.0625, -0.0521, ..., -0.1138, -0.0013, -0.1280], + [-0.1520, -0.0320, -0.1127, ..., -0.1107, -0.0025, 0.0935]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, 0.0000e+00, 8.1584e-07, ..., 1.3076e-06, + 2.0489e-07, 4.2841e-08], + [ 9.3132e-09, 0.0000e+00, -5.5939e-05, ..., -9.1970e-05, + 2.0843e-06, 1.6764e-08], + [ 1.8626e-09, 0.0000e+00, 5.3346e-06, ..., 9.3952e-06, + 1.7509e-07, 7.4506e-09], + ..., + [ 3.7253e-09, 3.7253e-09, 3.8922e-05, ..., 6.3419e-05, + -9.0897e-07, 2.6077e-07], + [ 1.8626e-09, 0.0000e+00, 5.1968e-07, ..., 1.3858e-06, + 7.6666e-06, 3.5390e-08], + [ 9.4995e-08, -7.4506e-09, 4.4815e-06, ..., 7.2010e-06, + 4.9360e-07, -4.3586e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0362, -0.0304, 0.0134, 0.0314, -0.0254, 0.0028, -0.0106, 0.0134, + 0.0135, 0.0001], device='cuda:0'), grad: tensor([ 4.5970e-06, -2.3186e-04, 2.5094e-05, 1.4141e-05, 8.5607e-06, + -9.2089e-05, 6.4611e-05, 1.6630e-04, 1.9878e-05, 2.1130e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 214.52, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.5266 re_mapping 0.0056 re_causal 0.0169 /// teacc 99.17 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0733, 0.0326, 0.0320, ..., -0.0641, -0.0766, 0.0602], + [-0.0544, -0.0150, -0.0802, ..., 0.1444, -0.0678, -0.0579], + [-0.0181, -0.0184, 0.1763, ..., -0.0951, -0.1164, -0.0931], + ..., + [-0.0701, -0.0490, 0.0046, ..., -0.0206, 0.0948, -0.0372], + [-0.0800, -0.0629, -0.0523, ..., -0.1139, -0.0019, -0.1283], + [-0.1523, -0.0331, -0.1131, ..., -0.1117, -0.0033, 0.0934]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, 5.5879e-08, ..., 2.0489e-08, + 8.0094e-08, -1.4901e-08], + [ 5.2154e-08, 5.5879e-09, 4.7311e-07, ..., -8.4564e-07, + 3.7253e-07, 2.0489e-08], + [ 3.1665e-08, 1.8626e-09, 1.3039e-08, ..., 1.1362e-07, + 5.9791e-07, 1.8626e-08], + ..., + [ 6.3330e-08, -3.9116e-08, -2.5239e-06, ..., 2.9802e-07, + -2.7008e-06, 2.6077e-08], + [ 2.2352e-08, 0.0000e+00, 1.7881e-07, ..., 1.8254e-07, + 7.2457e-07, 8.7544e-08], + [ 1.4901e-07, 2.6077e-08, 1.3467e-06, ..., 1.1176e-07, + 1.4827e-06, 2.0489e-08]], device='cuda:0') +Epoch 172, bias, value: tensor([-3.6108e-02, -3.0491e-02, 1.3221e-02, 3.1096e-02, -2.5314e-02, + 3.2963e-03, -1.0604e-02, 1.3409e-02, 1.3877e-02, -5.4188e-05], + device='cuda:0'), grad: tensor([ 2.1644e-06, 2.6766e-06, 3.7812e-06, 4.2677e-05, 1.6019e-07, + -1.1623e-04, 1.1563e-05, -2.2706e-06, 4.9263e-05, 6.1207e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 214.33, cls_loss 0.0011 cls_loss_mapping 0.0035 cls_loss_causal 0.5147 re_mapping 0.0057 re_causal 0.0169 /// teacc 99.11 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0739, 0.0330, 0.0320, ..., -0.0655, -0.0780, 0.0603], + [-0.0552, -0.0152, -0.0802, ..., 0.1454, -0.0679, -0.0581], + [-0.0182, -0.0183, 0.1766, ..., -0.0958, -0.1169, -0.0934], + ..., + [-0.0713, -0.0488, 0.0053, ..., -0.0211, 0.0952, -0.0373], + [-0.0803, -0.0632, -0.0526, ..., -0.1145, -0.0037, -0.1303], + [-0.1527, -0.0335, -0.1135, ..., -0.1132, -0.0037, 0.0934]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, -3.7253e-09, -1.8626e-08, ..., 6.7800e-07, + 2.2911e-07, 4.8429e-08], + [ 2.0117e-07, 0.0000e+00, 5.4389e-07, ..., -4.9829e-05, + -3.6191e-06, 8.5682e-08], + [ 1.0245e-07, 0.0000e+00, -2.0489e-08, ..., 2.2314e-06, + 2.4028e-07, 2.6077e-08], + ..., + [ 2.7008e-07, 0.0000e+00, -7.7300e-07, ..., 3.2067e-05, + 3.7700e-06, 2.0470e-06], + [ 1.0617e-07, 0.0000e+00, 1.6764e-08, ..., 1.3962e-05, + 4.1351e-07, 5.5879e-08], + [ 2.1793e-06, 3.7253e-09, 1.1921e-07, ..., 7.6368e-07, + -1.3299e-06, -5.2415e-06]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0362, -0.0303, 0.0130, 0.0311, -0.0251, 0.0037, -0.0109, 0.0135, + 0.0127, -0.0002], device='cuda:0'), grad: tensor([ 2.6133e-06, -1.0633e-04, 6.8620e-06, 3.0417e-06, -5.2713e-07, + 1.3717e-05, -1.4074e-05, 7.1645e-05, 2.5138e-05, -2.2147e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 214.72, cls_loss 0.0013 cls_loss_mapping 0.0040 cls_loss_causal 0.5477 re_mapping 0.0054 re_causal 0.0173 /// teacc 99.07 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0743, 0.0370, 0.0337, ..., -0.0678, -0.0785, 0.0597], + [-0.0561, -0.0155, -0.0811, ..., 0.1458, -0.0680, -0.0589], + [-0.0177, -0.0182, 0.1743, ..., -0.0959, -0.1202, -0.0935], + ..., + [-0.0723, -0.0488, 0.0085, ..., -0.0214, 0.0973, -0.0375], + [-0.0795, -0.0638, -0.0524, ..., -0.1151, -0.0038, -0.1310], + [-0.1530, -0.0366, -0.1142, ..., -0.1139, -0.0040, 0.0933]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.5146e-07, -2.4661e-06, ..., 1.6764e-08, + 2.0489e-08, -1.8403e-06], + [ 1.8626e-09, 2.6077e-08, 2.9802e-07, ..., -2.3283e-07, + 1.3225e-07, 6.2212e-07], + [ 5.5879e-09, 4.4703e-08, 7.0781e-08, ..., 7.0781e-08, + 8.7544e-08, 4.2282e-07], + ..., + [ 1.8626e-09, 2.0489e-08, 2.5705e-07, ..., 9.4995e-08, + -2.6077e-08, 9.1270e-06], + [ 3.7253e-09, 1.4901e-08, 2.7195e-07, ..., 5.0291e-08, + 2.3469e-07, 4.3772e-07], + [ 1.8626e-09, 5.5879e-08, 6.3144e-07, ..., 1.6764e-08, + 1.6578e-07, -9.5591e-06]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0382, -0.0305, 0.0107, 0.0331, -0.0250, 0.0038, -0.0109, 0.0153, + 0.0130, -0.0004], device='cuda:0'), grad: tensor([-7.0259e-06, 1.8030e-06, 1.3486e-06, 4.0010e-06, 9.7603e-07, + -4.2841e-06, 1.5423e-06, 2.7150e-05, 1.7714e-06, -2.7269e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 214.24, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5137 re_mapping 0.0057 re_causal 0.0166 /// teacc 99.04 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0756, 0.0402, 0.0340, ..., -0.0680, -0.0792, 0.0600], + [-0.0563, -0.0156, -0.0814, ..., 0.1461, -0.0688, -0.0589], + [-0.0177, -0.0182, 0.1745, ..., -0.0963, -0.1204, -0.0941], + ..., + [-0.0733, -0.0489, 0.0085, ..., -0.0211, 0.0981, -0.0378], + [-0.0797, -0.0642, -0.0529, ..., -0.1155, -0.0042, -0.1313], + [-0.1536, -0.0396, -0.1146, ..., -0.1148, -0.0043, 0.0933]], + device='cuda:0'), grad: tensor([[ 5.5879e-07, 0.0000e+00, 5.4017e-08, ..., 4.7684e-07, + 1.9185e-07, 3.2410e-07], + [ 1.8440e-07, 0.0000e+00, 1.8254e-07, ..., -1.0282e-06, + 3.5390e-07, 3.6135e-07], + [ 2.4401e-07, 0.0000e+00, 8.2515e-07, ..., 5.6811e-07, + 6.3702e-07, 2.0489e-07], + ..., + [ 1.1735e-07, 0.0000e+00, -2.0489e-06, ..., 4.0792e-07, + -7.3947e-07, 1.5855e-05], + [ 2.1234e-07, 0.0000e+00, 1.3039e-08, ..., 1.7695e-07, + 1.7453e-06, 3.2969e-07], + [ 2.5332e-06, 0.0000e+00, 7.6555e-07, ..., 1.5087e-07, + 6.1654e-07, -1.8299e-05]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0380, -0.0308, 0.0109, 0.0329, -0.0249, 0.0023, -0.0081, 0.0155, + 0.0128, -0.0007], device='cuda:0'), grad: tensor([ 3.6638e-06, 7.6368e-07, 5.5432e-06, 2.6319e-06, 1.6436e-05, + -1.5378e-05, -1.6764e-05, 5.0038e-05, 1.4175e-06, -4.8310e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 213.92, cls_loss 0.0014 cls_loss_mapping 0.0037 cls_loss_causal 0.5295 re_mapping 0.0052 re_causal 0.0159 /// teacc 99.08 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0761, 0.0418, 0.0342, ..., -0.0681, -0.0796, 0.0603], + [-0.0564, -0.0157, -0.0815, ..., 0.1474, -0.0688, -0.0592], + [-0.0178, -0.0184, 0.1746, ..., -0.0965, -0.1204, -0.0946], + ..., + [-0.0738, -0.0488, 0.0085, ..., -0.0223, 0.0983, -0.0381], + [-0.0797, -0.0649, -0.0524, ..., -0.1156, -0.0046, -0.1310], + [-0.1540, -0.0400, -0.1150, ..., -0.1156, -0.0048, 0.0934]], + device='cuda:0'), grad: tensor([[ 1.6391e-07, 0.0000e+00, -8.9407e-08, ..., 8.1956e-08, + 1.6578e-07, 2.2165e-07], + [ 8.3819e-08, 0.0000e+00, 1.9744e-07, ..., -7.4506e-09, + 5.0478e-07, 2.1420e-07], + [ 4.2841e-08, 0.0000e+00, 2.2352e-08, ..., 7.8231e-08, + 3.7812e-07, 1.0058e-07], + ..., + [ 4.9919e-07, 0.0000e+00, -8.3447e-07, ..., 2.6077e-08, + -1.6559e-06, 1.0598e-06], + [ 1.5274e-07, 0.0000e+00, 8.9407e-08, ..., 1.6205e-07, + 5.7928e-07, 7.6927e-07], + [-2.8729e-05, 0.0000e+00, 2.7381e-07, ..., 1.6764e-08, + 4.3586e-07, -5.2392e-05]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0378, -0.0304, 0.0109, 0.0326, -0.0248, 0.0024, -0.0090, 0.0153, + 0.0148, -0.0009], device='cuda:0'), grad: tensor([ 1.8422e-06, 2.4363e-06, 2.0377e-06, -4.3772e-07, 1.0806e-04, + -6.8545e-06, -1.5590e-06, -3.3528e-08, 5.2899e-06, -1.1098e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 214.30, cls_loss 0.0012 cls_loss_mapping 0.0041 cls_loss_causal 0.5207 re_mapping 0.0054 re_causal 0.0165 /// teacc 99.08 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0769, 0.0419, 0.0339, ..., -0.0682, -0.0800, 0.0604], + [-0.0565, -0.0157, -0.0816, ..., 0.1477, -0.0689, -0.0594], + [-0.0178, -0.0183, 0.1750, ..., -0.0967, -0.1205, -0.0948], + ..., + [-0.0737, -0.0488, 0.0084, ..., -0.0225, 0.0985, -0.0382], + [-0.0795, -0.0654, -0.0538, ..., -0.1157, -0.0028, -0.1312], + [-0.1541, -0.0401, -0.1151, ..., -0.1158, -0.0052, 0.0937]], + device='cuda:0'), grad: tensor([[ 1.5274e-07, 0.0000e+00, -1.7509e-07, ..., 1.6950e-07, + 6.4448e-07, -8.1956e-08], + [ 8.7544e-08, 0.0000e+00, 2.6636e-07, ..., -2.1327e-06, + 5.1595e-07, -3.1665e-08], + [ 4.5262e-07, 0.0000e+00, 7.4506e-07, ..., 3.2410e-07, + 2.3823e-06, 2.3469e-07], + ..., + [ 3.1460e-06, 0.0000e+00, -2.2892e-06, ..., 3.6694e-07, + 1.1496e-05, 1.5497e-06], + [ 3.8370e-07, 0.0000e+00, 2.7753e-07, ..., 4.0606e-07, + 1.7453e-06, 2.9244e-07], + [ 1.7881e-07, 0.0000e+00, 7.8417e-07, ..., 3.4086e-07, + 1.0189e-06, 7.0781e-08]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0379, -0.0304, 0.0113, 0.0326, -0.0249, 0.0024, -0.0107, 0.0152, + 0.0165, -0.0008], device='cuda:0'), grad: tensor([ 2.0862e-06, -3.8072e-06, 8.1658e-06, 2.0579e-05, 1.1865e-06, + -8.0109e-05, 4.2468e-06, 3.6001e-05, 7.3612e-06, 4.2170e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 214.45, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.4854 re_mapping 0.0054 re_causal 0.0159 /// teacc 99.05 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0773, 0.0419, 0.0340, ..., -0.0682, -0.0807, 0.0604], + [-0.0566, -0.0158, -0.0818, ..., 0.1484, -0.0690, -0.0594], + [-0.0180, -0.0183, 0.1752, ..., -0.0973, -0.1205, -0.0951], + ..., + [-0.0746, -0.0489, 0.0083, ..., -0.0227, 0.0986, -0.0383], + [-0.0798, -0.0655, -0.0542, ..., -0.1184, -0.0020, -0.1328], + [-0.1544, -0.0401, -0.1153, ..., -0.1173, -0.0055, 0.0938]], + device='cuda:0'), grad: tensor([[ 1.6950e-06, -1.1176e-08, -7.4506e-08, ..., 7.7672e-07, + 1.1176e-08, -3.5390e-08], + [ 1.5676e-05, 0.0000e+00, 4.3027e-07, ..., 7.1488e-06, + 1.4342e-07, 9.8720e-08], + [ 4.8243e-06, 0.0000e+00, -2.5295e-06, ..., 2.2668e-06, + 2.7940e-08, 3.5390e-08], + ..., + [ 4.8056e-06, 0.0000e+00, 1.9707e-06, ..., 8.1211e-07, + -2.7381e-07, 1.3784e-07], + [ 1.6578e-06, 0.0000e+00, 1.1362e-07, ..., 7.9535e-07, + 1.5460e-07, 7.8231e-08], + [ 4.2349e-05, 7.4506e-09, 1.1176e-07, ..., 1.9372e-05, + 1.1176e-07, -7.4506e-08]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0379, -0.0302, 0.0112, 0.0325, -0.0248, 0.0025, -0.0113, 0.0151, + 0.0171, -0.0009], device='cuda:0'), grad: tensor([ 4.9248e-06, 4.7475e-05, 1.0937e-05, 5.4128e-06, -2.1386e-04, + -2.2501e-06, 5.0552e-06, 1.1578e-05, 4.7907e-06, 1.2589e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 214.44, cls_loss 0.0014 cls_loss_mapping 0.0041 cls_loss_causal 0.5411 re_mapping 0.0050 re_causal 0.0159 /// teacc 99.08 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0783, 0.0422, 0.0342, ..., -0.0686, -0.0817, 0.0605], + [-0.0569, -0.0158, -0.0820, ..., 0.1487, -0.0690, -0.0595], + [-0.0181, -0.0184, 0.1754, ..., -0.0973, -0.1205, -0.0956], + ..., + [-0.0768, -0.0489, 0.0083, ..., -0.0230, 0.0986, -0.0385], + [-0.0804, -0.0659, -0.0546, ..., -0.1189, -0.0019, -0.1338], + [-0.1548, -0.0402, -0.1156, ..., -0.1194, -0.0060, 0.0941]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -9.3132e-09, -1.8999e-07, ..., 1.5628e-06, + 6.3702e-07, 1.4342e-06], + [ 3.7253e-08, 0.0000e+00, 1.0431e-07, ..., -3.0044e-06, + 1.7136e-07, -8.7544e-08], + [ 2.0489e-08, 0.0000e+00, -1.2666e-07, ..., 8.3260e-07, + 1.1548e-07, 2.7567e-07], + ..., + [ 1.4901e-08, 0.0000e+00, -2.5332e-07, ..., 3.2596e-07, + -6.7614e-07, 2.3656e-07], + [ 3.7253e-09, 0.0000e+00, 1.4901e-07, ..., 2.0657e-06, + 1.0207e-06, 2.0694e-06], + [ 9.4995e-08, 3.7253e-09, 2.2165e-07, ..., 9.8348e-07, + 1.7136e-07, -4.6566e-08]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0379, -0.0304, 0.0113, 0.0338, -0.0247, 0.0012, -0.0111, 0.0148, + 0.0177, -0.0011], device='cuda:0'), grad: tensor([ 6.3181e-06, -5.2191e-06, 2.3171e-06, 2.5574e-06, 1.2722e-06, + 5.4017e-06, -2.0787e-05, -2.5518e-07, 5.7593e-06, 2.6226e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 214.36, cls_loss 0.0012 cls_loss_mapping 0.0037 cls_loss_causal 0.5420 re_mapping 0.0052 re_causal 0.0165 /// teacc 99.13 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0784, 0.0432, 0.0346, ..., -0.0686, -0.0833, 0.0606], + [-0.0571, -0.0163, -0.0838, ..., 0.1472, -0.0709, -0.0596], + [-0.0181, -0.0185, 0.1755, ..., -0.0977, -0.1206, -0.0961], + ..., + [-0.0776, -0.0489, 0.0086, ..., -0.0207, 0.1000, -0.0390], + [-0.0807, -0.0664, -0.0550, ..., -0.1206, -0.0014, -0.1346], + [-0.1552, -0.0408, -0.1159, ..., -0.1216, -0.0062, 0.0944]], + device='cuda:0'), grad: tensor([[ 1.1362e-07, 0.0000e+00, 1.0412e-06, ..., 5.9605e-08, + 2.1234e-07, -4.0978e-08], + [ 6.4448e-07, 1.1176e-08, 5.4017e-08, ..., -2.0191e-06, + 9.9093e-07, 3.7253e-09], + [ 5.0291e-07, 1.8626e-09, -4.0270e-06, ..., 3.5763e-07, + 1.0934e-06, 7.4506e-09], + ..., + [ 4.6566e-07, -2.4214e-08, 3.0734e-07, ..., 1.4119e-06, + 3.8370e-07, 2.2352e-08], + [-4.8913e-06, 0.0000e+00, 2.4959e-07, ..., 1.1362e-07, + -2.0802e-05, 5.5879e-08], + [ 2.2147e-06, 3.7253e-09, 6.3330e-08, ..., 8.3819e-08, + 3.5949e-07, 2.6077e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0378, -0.0323, 0.0114, 0.0336, -0.0246, 0.0013, -0.0109, 0.0161, + 0.0177, -0.0011], device='cuda:0'), grad: tensor([ 4.5970e-06, 9.2313e-06, 9.4622e-06, -3.0696e-06, -1.5646e-07, + 2.3794e-04, 9.9689e-06, 1.1064e-05, -2.8634e-04, 6.6534e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 214.18, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5093 re_mapping 0.0053 re_causal 0.0165 /// teacc 99.14 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0792, 0.0434, 0.0347, ..., -0.0688, -0.0847, 0.0606], + [-0.0574, -0.0165, -0.0839, ..., 0.1501, -0.0699, -0.0597], + [-0.0182, -0.0184, 0.1756, ..., -0.0985, -0.1206, -0.0962], + ..., + [-0.0782, -0.0489, 0.0085, ..., -0.0234, 0.0992, -0.0393], + [-0.0834, -0.0670, -0.0559, ..., -0.1215, -0.0019, -0.1354], + [-0.1556, -0.0408, -0.1160, ..., -0.1226, -0.0065, 0.0944]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 0.0000e+00, -9.7603e-07, ..., 1.8626e-08, + -5.3644e-07, -2.5909e-06], + [ 8.0094e-08, 0.0000e+00, 3.9116e-08, ..., -3.1665e-08, + 9.6858e-08, 5.4017e-08], + [ 4.4703e-08, 0.0000e+00, -2.8498e-07, ..., 1.4901e-08, + 2.0489e-08, 2.2352e-08], + ..., + [ 1.3039e-07, 0.0000e+00, -2.0489e-08, ..., 1.1176e-08, + -1.2796e-06, -4.4703e-07], + [ 2.0489e-08, 0.0000e+00, 6.8918e-08, ..., 4.2841e-08, + 1.0431e-07, 1.1176e-07], + [ 1.7695e-07, 0.0000e+00, 3.3155e-07, ..., 2.7940e-08, + 1.1269e-06, 6.0350e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0378, -0.0302, 0.0112, 0.0334, -0.0244, 0.0018, -0.0111, 0.0144, + 0.0174, -0.0013], device='cuda:0'), grad: tensor([-4.2915e-06, 5.0105e-07, -1.5460e-07, -2.9616e-07, -4.1910e-07, + 1.4976e-06, 1.6689e-06, -2.3562e-06, 5.6252e-07, 3.2596e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 214.42, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.5382 re_mapping 0.0052 re_causal 0.0160 /// teacc 99.18 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0805, 0.0447, 0.0349, ..., -0.0689, -0.0851, 0.0608], + [-0.0578, -0.0171, -0.0840, ..., 0.1508, -0.0698, -0.0598], + [-0.0183, -0.0155, 0.1758, ..., -0.0988, -0.1206, -0.0964], + ..., + [-0.0790, -0.0499, 0.0085, ..., -0.0240, 0.0992, -0.0395], + [-0.0844, -0.0683, -0.0564, ..., -0.1221, -0.0026, -0.1372], + [-0.1546, -0.0424, -0.1163, ..., -0.1233, -0.0067, 0.0953]], + device='cuda:0'), grad: tensor([[ 5.5879e-08, 0.0000e+00, 3.4459e-07, ..., 2.0489e-07, + 8.5682e-08, -9.8161e-07], + [-1.6205e-07, 0.0000e+00, 1.5665e-06, ..., -5.9865e-06, + 3.1292e-06, 3.1292e-07], + [ 9.1270e-08, 0.0000e+00, -4.0717e-06, ..., 2.4773e-07, + 2.7940e-07, 1.1735e-07], + ..., + [ 5.4389e-07, 0.0000e+00, -2.0191e-06, ..., 1.7993e-06, + -5.6997e-06, 1.1921e-07], + [ 5.0291e-08, 0.0000e+00, 1.6019e-07, ..., 6.9849e-07, + 1.6075e-06, 1.3448e-06], + [ 5.4576e-07, 0.0000e+00, 8.8103e-07, ..., 3.2037e-07, + 1.9874e-06, -1.5926e-06]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0377, -0.0299, 0.0113, 0.0334, -0.0251, 0.0022, -0.0112, 0.0141, + 0.0173, -0.0008], device='cuda:0'), grad: tensor([-6.9290e-07, -5.3942e-06, -4.1053e-06, 9.1046e-06, 7.4282e-06, + -1.3903e-05, 5.9605e-06, -9.7081e-06, 1.1824e-05, -6.3144e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 214.62, cls_loss 0.0011 cls_loss_mapping 0.0036 cls_loss_causal 0.5025 re_mapping 0.0052 re_causal 0.0154 /// teacc 98.99 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0809, 0.0448, 0.0349, ..., -0.0689, -0.0859, 0.0608], + [-0.0580, -0.0174, -0.0842, ..., 0.1510, -0.0700, -0.0603], + [-0.0184, -0.0154, 0.1761, ..., -0.0989, -0.1206, -0.0979], + ..., + [-0.0793, -0.0500, 0.0084, ..., -0.0241, 0.0995, -0.0396], + [-0.0845, -0.0688, -0.0570, ..., -0.1223, -0.0027, -0.1380], + [-0.1545, -0.0426, -0.1165, ..., -0.1238, -0.0076, 0.0957]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -7.8231e-08, -9.4436e-07, ..., 1.0747e-06, + 7.4506e-09, -6.9849e-07], + [ 6.5193e-08, 7.2643e-08, 3.9116e-07, ..., 6.4969e-06, + 7.4506e-08, 3.1292e-07], + [ 2.4214e-08, -2.4699e-06, -1.2085e-05, ..., 4.5076e-07, + 1.4901e-08, 1.5460e-07], + ..., + [ 1.2293e-07, 2.1160e-06, 1.0461e-05, ..., 1.0617e-07, + -1.4715e-07, 1.1157e-06], + [ 7.8231e-08, 6.5193e-08, 3.4831e-07, ..., -9.8906e-07, + -1.6950e-07, 3.8184e-07], + [-1.4529e-07, 5.0291e-08, 4.8243e-07, ..., 8.5682e-08, + 9.1270e-08, -1.5087e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0377, -0.0297, 0.0114, 0.0332, -0.0253, 0.0025, -0.0112, 0.0140, + 0.0173, -0.0007], device='cuda:0'), grad: tensor([ 3.7812e-07, 2.0429e-05, -1.5497e-05, -2.6971e-05, 1.3690e-06, + 6.7875e-06, -8.7693e-06, 2.9370e-05, -1.1742e-05, 4.5598e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 214.45, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.5067 re_mapping 0.0050 re_causal 0.0154 /// teacc 99.09 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0816, 0.0451, 0.0358, ..., -0.0689, -0.0872, 0.0613], + [-0.0589, -0.0176, -0.0844, ..., 0.1516, -0.0700, -0.0608], + [-0.0187, -0.0153, 0.1763, ..., -0.0991, -0.1207, -0.0996], + ..., + [-0.0804, -0.0501, 0.0082, ..., -0.0248, 0.0995, -0.0399], + [-0.0854, -0.0691, -0.0587, ..., -0.1226, -0.0034, -0.1397], + [-0.1546, -0.0428, -0.1168, ..., -0.1242, -0.0078, 0.0956]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, -0.0000e+00, -9.3132e-08, ..., 1.4156e-07, + 1.4901e-08, -6.5193e-08], + [ 6.7428e-07, 0.0000e+00, 1.9930e-07, ..., -3.7253e-09, + 2.0489e-07, 9.3132e-09], + [ 2.9989e-07, 0.0000e+00, -3.2410e-07, ..., 9.1270e-08, + 5.9605e-08, 1.1176e-08], + ..., + [ 1.9595e-06, 0.0000e+00, 4.7125e-07, ..., 3.7253e-08, + 2.9989e-07, 5.4017e-08], + [ 3.9674e-07, 0.0000e+00, 1.1362e-07, ..., 1.8440e-07, + 8.5682e-08, 2.9802e-08], + [ 4.2543e-06, 0.0000e+00, 8.1770e-07, ..., 1.8626e-08, + 9.4250e-07, -1.0990e-07]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0372, -0.0295, 0.0114, 0.0322, -0.0251, 0.0040, -0.0115, 0.0135, + 0.0169, -0.0008], device='cuda:0'), grad: tensor([ 3.1851e-07, 1.3690e-06, 2.0675e-07, 1.8626e-07, -1.2040e-05, + 3.7625e-07, -1.2126e-06, 3.3639e-06, 6.9663e-07, 6.6906e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 214.64, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5257 re_mapping 0.0049 re_causal 0.0156 /// teacc 99.15 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0842, 0.0450, 0.0360, ..., -0.0693, -0.0880, 0.0614], + [-0.0593, -0.0177, -0.0848, ..., 0.1519, -0.0701, -0.0610], + [-0.0189, -0.0153, 0.1766, ..., -0.0992, -0.1207, -0.0996], + ..., + [-0.0817, -0.0501, 0.0081, ..., -0.0250, 0.0996, -0.0401], + [-0.0860, -0.0693, -0.0592, ..., -0.1230, -0.0034, -0.1401], + [-0.1548, -0.0428, -0.1175, ..., -0.1246, -0.0082, 0.0958]], + device='cuda:0'), grad: tensor([[ 9.4064e-08, 1.8626e-09, -2.3209e-06, ..., 2.9802e-08, + 1.3039e-08, -6.8396e-06], + [ 1.2908e-06, 0.0000e+00, 1.6838e-06, ..., 1.8068e-07, + 3.9116e-08, 3.3528e-08], + [ 1.9092e-07, 0.0000e+00, -3.3993e-06, ..., 6.7987e-08, + 1.2107e-07, 2.0768e-07], + ..., + [ 3.9488e-07, 3.7253e-09, 9.3132e-08, ..., 9.7789e-08, + -5.7556e-07, 5.6252e-07], + [-3.7625e-07, 0.0000e+00, 4.1258e-07, ..., 3.3528e-08, + 4.0047e-08, 1.7788e-07], + [ 4.7404e-07, -1.4901e-08, 1.2601e-06, ..., 9.6858e-08, + 3.0734e-08, 2.5872e-06]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0373, -0.0295, 0.0116, 0.0321, -0.0250, 0.0040, -0.0113, 0.0133, + 0.0170, -0.0009], device='cuda:0'), grad: tensor([-1.0550e-05, 4.0494e-06, -3.5595e-06, 2.3879e-06, -1.8980e-06, + 3.7663e-06, 1.8245e-06, 1.1427e-06, -2.2426e-06, 5.0589e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 214.49, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.5263 re_mapping 0.0048 re_causal 0.0154 /// teacc 99.12 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0868, 0.0450, 0.0363, ..., -0.0697, -0.0892, 0.0615], + [-0.0600, -0.0180, -0.0853, ..., 0.1527, -0.0698, -0.0614], + [-0.0190, -0.0154, 0.1769, ..., -0.0994, -0.1207, -0.0999], + ..., + [-0.0833, -0.0500, 0.0081, ..., -0.0256, 0.0995, -0.0403], + [-0.0872, -0.0695, -0.0603, ..., -0.1239, -0.0038, -0.1406], + [-0.1555, -0.0428, -0.1178, ..., -0.1255, -0.0086, 0.0959]], + device='cuda:0'), grad: tensor([[ 4.3772e-08, 3.7253e-09, 1.4901e-08, ..., 3.4645e-07, + 8.3819e-09, 2.4494e-07], + [ 4.8764e-06, 2.7940e-09, 5.7369e-07, ..., 1.1288e-06, + 3.5949e-07, 1.4529e-07], + [ 2.8778e-07, 9.3132e-10, -5.5693e-07, ..., 2.3842e-07, + 6.4261e-08, 5.1223e-08], + ..., + [ 3.2131e-07, 1.4901e-08, -3.6415e-07, ..., 5.6624e-07, + -7.2084e-07, 3.3900e-07], + [ 4.1910e-08, 9.3132e-10, 9.0338e-08, ..., 4.3958e-07, + 1.4901e-08, 1.6391e-07], + [-1.4808e-07, -9.8720e-08, 4.0978e-08, ..., 1.2852e-07, + 6.6124e-08, -1.9893e-06]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0373, -0.0290, 0.0118, 0.0322, -0.0247, 0.0038, -0.0113, 0.0128, + 0.0166, -0.0010], device='cuda:0'), grad: tensor([ 1.5227e-06, 7.9796e-06, 2.5406e-06, 1.0571e-06, -5.5246e-06, + 4.0010e-06, -7.0632e-06, 3.5018e-07, -2.8927e-06, -1.9707e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 185---------------------------------------------------- +epoch 185, time 231.16, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4823 re_mapping 0.0047 re_causal 0.0147 /// teacc 99.20 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0876, 0.0450, 0.0367, ..., -0.0700, -0.0904, 0.0616], + [-0.0605, -0.0198, -0.0855, ..., 0.1529, -0.0699, -0.0617], + [-0.0192, -0.0152, 0.1769, ..., -0.0996, -0.1208, -0.1004], + ..., + [-0.0838, -0.0499, 0.0082, ..., -0.0257, 0.1000, -0.0404], + [-0.0856, -0.0697, -0.0595, ..., -0.1246, -0.0039, -0.1409], + [-0.1560, -0.0428, -0.1182, ..., -0.1259, -0.0094, 0.0960]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -3.7253e-09, -4.0699e-07, ..., 3.4459e-08, + 3.7253e-09, -4.1537e-07], + [ 4.9360e-08, 0.0000e+00, 2.9802e-08, ..., -1.4901e-08, + 9.9652e-08, 1.3970e-08], + [ 1.5832e-08, 9.3132e-10, 3.2596e-08, ..., 8.1025e-08, + 2.9802e-08, 2.4214e-08], + ..., + [ 3.6322e-08, 9.3132e-10, -2.2352e-08, ..., 4.5635e-08, + -1.3597e-07, 5.0012e-07], + [ 2.7940e-08, 0.0000e+00, -1.8626e-08, ..., 1.3132e-07, + -9.3132e-10, 1.0245e-08], + [ 1.8533e-07, 1.8626e-09, 3.1106e-07, ..., 1.0245e-08, + -6.5193e-08, -2.9989e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0372, -0.0290, 0.0115, 0.0319, -0.0246, 0.0037, -0.0112, 0.0131, + 0.0171, -0.0013], device='cuda:0'), grad: tensor([-9.0711e-07, 3.7625e-07, 4.5262e-07, -3.3900e-07, -6.1467e-08, + 4.0047e-07, -8.2515e-07, 4.6100e-07, 1.5832e-07, 2.7195e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 214.51, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.4959 re_mapping 0.0049 re_causal 0.0151 /// teacc 99.02 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0883, 0.0451, 0.0369, ..., -0.0704, -0.0908, 0.0617], + [-0.0610, -0.0199, -0.0858, ..., 0.1534, -0.0699, -0.0618], + [-0.0193, -0.0152, 0.1771, ..., -0.1001, -0.1208, -0.1007], + ..., + [-0.0843, -0.0499, 0.0082, ..., -0.0259, 0.1000, -0.0406], + [-0.0856, -0.0700, -0.0599, ..., -0.1257, -0.0040, -0.1413], + [-0.1561, -0.0427, -0.1184, ..., -0.1266, -0.0095, 0.0962]], + device='cuda:0'), grad: tensor([[ 1.6205e-07, -6.0536e-08, -5.8673e-07, ..., 3.6322e-08, + 1.8626e-09, -2.9802e-07], + [ 1.5460e-07, 0.0000e+00, -4.6566e-09, ..., -2.5667e-06, + 5.5879e-08, 6.5193e-09], + [ 2.5146e-07, 2.9802e-08, 1.5181e-07, ..., 1.6904e-06, + 2.7008e-08, 1.4342e-07], + ..., + [ 1.3225e-07, 0.0000e+00, -4.7497e-08, ..., 3.9581e-07, + -2.3004e-07, 5.3085e-08], + [ 5.0291e-08, 1.8626e-09, 4.7497e-08, ..., 2.3842e-07, + 2.7008e-08, 3.0734e-08], + [ 5.2620e-07, 1.9558e-08, 2.1420e-07, ..., 4.6566e-08, + 2.5146e-08, -1.0245e-08]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0372, -0.0288, 0.0114, 0.0317, -0.0245, 0.0038, -0.0110, 0.0130, + 0.0169, -0.0013], device='cuda:0'), grad: tensor([-7.4878e-07, -6.0275e-06, 5.1484e-06, 1.6391e-07, -1.2703e-06, + 8.5030e-07, -2.3097e-07, 9.6299e-07, -3.2596e-08, 1.1567e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 214.45, cls_loss 0.0015 cls_loss_mapping 0.0038 cls_loss_causal 0.5594 re_mapping 0.0048 re_causal 0.0155 /// teacc 99.13 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0892, 0.0453, 0.0372, ..., -0.0704, -0.0915, 0.0618], + [-0.0623, -0.0207, -0.0840, ..., 0.1567, -0.0700, -0.0621], + [-0.0185, -0.0153, 0.1768, ..., -0.1044, -0.1209, -0.1011], + ..., + [-0.0856, -0.0497, 0.0081, ..., -0.0260, 0.1002, -0.0408], + [-0.0868, -0.0711, -0.0610, ..., -0.1298, -0.0042, -0.1420], + [-0.1565, -0.0432, -0.1186, ..., -0.1277, -0.0098, 0.0964]], + device='cuda:0'), grad: tensor([[ 2.3097e-07, 0.0000e+00, 7.2643e-08, ..., 4.1444e-07, + 4.0047e-08, 3.7253e-08], + [ 8.3633e-07, 0.0000e+00, 8.5682e-08, ..., 3.8892e-06, + 1.3970e-07, 1.5739e-07], + [ 3.5390e-08, 0.0000e+00, -1.4268e-06, ..., 1.5181e-07, + 2.0023e-07, 4.7497e-08], + ..., + [ 9.1270e-08, 0.0000e+00, 8.9593e-07, ..., 4.0978e-08, + 1.6298e-07, 1.2964e-06], + [ 3.1665e-08, 0.0000e+00, 1.1921e-07, ..., 9.4995e-08, + -9.6858e-07, 1.6112e-07], + [ 1.6391e-07, 0.0000e+00, 2.7008e-08, ..., 7.6368e-08, + 1.4808e-07, -2.1476e-06]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0371, -0.0267, 0.0093, 0.0310, -0.0245, 0.0024, -0.0085, 0.0129, + 0.0158, -0.0012], device='cuda:0'), grad: tensor([ 1.7649e-06, 1.0177e-05, 2.3283e-06, -5.9679e-06, 2.7008e-08, + 2.9802e-06, -8.7842e-06, 6.2473e-06, -6.2436e-06, -2.5555e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 214.45, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4857 re_mapping 0.0050 re_causal 0.0149 /// teacc 99.03 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0911, 0.0435, 0.0374, ..., -0.0713, -0.0934, 0.0618], + [-0.0628, -0.0217, -0.0842, ..., 0.1567, -0.0701, -0.0624], + [-0.0187, -0.0153, 0.1770, ..., -0.1044, -0.1209, -0.1014], + ..., + [-0.0860, -0.0497, 0.0081, ..., -0.0260, 0.1005, -0.0411], + [-0.0879, -0.0745, -0.0615, ..., -0.1304, -0.0044, -0.1424], + [-0.1566, -0.0433, -0.1188, ..., -0.1282, -0.0101, 0.0966]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, -9.3132e-09, -4.2003e-07, ..., 6.2399e-08, + 1.2107e-08, -2.9150e-07], + [ 9.3132e-08, 1.8626e-09, 1.9167e-06, ..., 2.5146e-08, + 8.4750e-08, 8.3819e-09], + [ 5.4948e-08, 4.6566e-09, -1.0267e-05, ..., 4.2841e-08, + 4.9360e-08, 6.7987e-08], + ..., + [ 2.4959e-07, 2.5146e-08, 3.8594e-06, ..., 8.3819e-09, + -4.4331e-07, 4.6566e-09], + [ 5.4017e-08, 1.8626e-09, 1.3644e-06, ..., 1.6112e-07, + 2.7008e-08, 4.9360e-08], + [ 1.7788e-07, 3.7253e-09, 4.9081e-07, ..., 1.3970e-08, + 1.6857e-07, 7.5437e-08]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0373, -0.0269, 0.0093, 0.0319, -0.0246, 0.0019, -0.0085, 0.0130, + 0.0155, -0.0012], device='cuda:0'), grad: tensor([-5.8953e-07, 2.6636e-06, -1.2949e-05, 3.3509e-06, -9.9931e-07, + 5.5693e-07, -5.0012e-07, 4.6156e-06, 2.4214e-06, 1.4212e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 214.59, cls_loss 0.0011 cls_loss_mapping 0.0027 cls_loss_causal 0.5451 re_mapping 0.0047 re_causal 0.0147 /// teacc 99.17 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0924, 0.0439, 0.0378, ..., -0.0720, -0.0936, 0.0619], + [-0.0631, -0.0220, -0.0845, ..., 0.1569, -0.0702, -0.0627], + [-0.0187, -0.0154, 0.1771, ..., -0.1045, -0.1211, -0.1018], + ..., + [-0.0867, -0.0497, 0.0082, ..., -0.0261, 0.1007, -0.0414], + [-0.0855, -0.0747, -0.0622, ..., -0.1311, -0.0045, -0.1430], + [-0.1569, -0.0434, -0.1191, ..., -0.1291, -0.0103, 0.0968]], + device='cuda:0'), grad: tensor([[ 6.7987e-08, -2.7940e-09, -5.8673e-08, ..., 9.3132e-09, + 8.1956e-08, 2.9337e-07], + [ 3.3062e-07, 0.0000e+00, 3.4459e-07, ..., -2.2911e-07, + 2.4680e-07, 1.0058e-07], + [ 1.1176e-07, 0.0000e+00, 3.1665e-07, ..., 3.6322e-08, + 1.0524e-06, 1.2293e-07], + ..., + [ 1.4715e-07, 0.0000e+00, -1.3625e-06, ..., 2.6077e-08, + 4.5542e-07, 3.6787e-07], + [ 1.0710e-07, 0.0000e+00, 2.5425e-07, ..., 9.3132e-09, + 6.7987e-08, 2.3004e-07], + [ 5.2992e-07, 1.8626e-09, 2.4680e-07, ..., 8.3819e-09, + 3.7067e-07, -6.2510e-06]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0372, -0.0269, 0.0093, 0.0305, -0.0246, 0.0027, -0.0085, 0.0131, + 0.0162, -0.0012], device='cuda:0'), grad: tensor([ 1.3383e-06, 1.8524e-06, 4.4033e-06, -2.3305e-05, -6.5099e-07, + 1.1131e-05, 7.8790e-07, 1.4350e-05, 6.1281e-07, -1.0543e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 214.46, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5417 re_mapping 0.0046 re_causal 0.0151 /// teacc 99.08 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0910, 0.0443, 0.0391, ..., -0.0719, -0.0946, 0.0620], + [-0.0644, -0.0226, -0.0850, ..., 0.1570, -0.0703, -0.0632], + [-0.0199, -0.0153, 0.1772, ..., -0.1044, -0.1213, -0.1025], + ..., + [-0.0878, -0.0497, 0.0083, ..., -0.0263, 0.1009, -0.0419], + [-0.0857, -0.0749, -0.0632, ..., -0.1316, -0.0047, -0.1437], + [-0.1572, -0.0435, -0.1196, ..., -0.1295, -0.0105, 0.0976]], + device='cuda:0'), grad: tensor([[ 8.0094e-08, 9.3132e-10, 7.4506e-08, ..., 1.6764e-08, + 3.8184e-08, 8.3819e-09], + [ 2.5705e-07, 3.7253e-09, 8.6520e-07, ..., -5.0850e-07, + 1.3150e-06, 9.3132e-09], + [ 1.9092e-07, 2.7940e-09, -6.6962e-07, ..., 2.7008e-08, + 6.6776e-07, 8.3819e-09], + ..., + [ 3.5390e-07, 5.5879e-09, -1.9316e-06, ..., 3.8184e-07, + -3.6862e-06, -2.4214e-08], + [ 1.0245e-07, 9.3132e-10, 1.5367e-07, ..., 3.7253e-08, + 1.0524e-07, 2.6077e-08], + [ 8.3540e-07, 2.7940e-08, 3.9581e-07, ..., 1.3970e-08, + 6.0163e-07, 4.3772e-08]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0370, -0.0271, 0.0092, 0.0304, -0.0244, 0.0027, -0.0084, 0.0131, + 0.0161, -0.0010], device='cuda:0'), grad: tensor([ 3.1851e-07, 2.6152e-06, 3.4366e-07, 3.4422e-06, -3.1013e-06, + -1.0878e-06, 5.4389e-07, -6.2473e-06, 5.4203e-07, 2.6263e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 214.24, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5325 re_mapping 0.0048 re_causal 0.0152 /// teacc 99.15 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0913, 0.0448, 0.0394, ..., -0.0719, -0.0953, 0.0621], + [-0.0648, -0.0241, -0.0853, ..., 0.1570, -0.0710, -0.0639], + [-0.0201, -0.0151, 0.1774, ..., -0.1044, -0.1214, -0.1031], + ..., + [-0.0883, -0.0498, 0.0084, ..., -0.0262, 0.1017, -0.0420], + [-0.0858, -0.0752, -0.0636, ..., -0.1319, -0.0049, -0.1442], + [-0.1576, -0.0436, -0.1200, ..., -0.1298, -0.0106, 0.0980]], + device='cuda:0'), grad: tensor([[ 2.1681e-06, -1.3039e-08, -1.2359e-06, ..., 1.2210e-06, + 1.7695e-08, -7.2643e-08], + [ 3.0641e-07, 0.0000e+00, 4.2841e-07, ..., 5.0385e-07, + 8.1025e-08, 2.7940e-09], + [ 1.1204e-06, 9.3132e-10, 4.5728e-07, ..., 1.2731e-06, + 1.1828e-07, 7.4506e-09], + ..., + [ 3.9116e-08, 0.0000e+00, -4.6566e-08, ..., 4.1910e-08, + -3.4086e-07, 1.2107e-08], + [ 5.2620e-07, 0.0000e+00, -1.0030e-06, ..., 6.6683e-07, + -1.9465e-07, 2.0489e-08], + [ 1.5646e-07, 4.6566e-09, 5.9512e-07, ..., 1.0617e-07, + 2.1420e-07, 2.2352e-08]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0369, -0.0276, 0.0093, 0.0304, -0.0243, 0.0029, -0.0087, 0.0136, + 0.0161, -0.0011], device='cuda:0'), grad: tensor([ 6.4597e-06, 2.9933e-06, 7.5363e-06, 3.9712e-06, 1.0066e-05, + 4.9584e-06, -3.5286e-05, -2.0955e-07, -2.5854e-06, 2.0973e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 214.45, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.5225 re_mapping 0.0050 re_causal 0.0152 /// teacc 99.15 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0925, 0.0450, 0.0394, ..., -0.0725, -0.0962, 0.0622], + [-0.0652, -0.0247, -0.0854, ..., 0.1571, -0.0711, -0.0646], + [-0.0200, -0.0148, 0.1777, ..., -0.1045, -0.1215, -0.1035], + ..., + [-0.0890, -0.0500, 0.0084, ..., -0.0262, 0.1019, -0.0423], + [-0.0858, -0.0756, -0.0639, ..., -0.1322, -0.0050, -0.1445], + [-0.1580, -0.0437, -0.1203, ..., -0.1300, -0.0113, 0.0982]], + device='cuda:0'), grad: tensor([[-8.4285e-07, -4.6007e-07, -2.9832e-05, ..., -7.5437e-08, + 6.4261e-08, -1.2383e-05], + [ 1.7975e-07, 8.3819e-09, 6.8452e-07, ..., 1.0151e-07, + 1.0431e-07, 6.0443e-07], + [ 1.0431e-07, 3.4738e-07, 1.9372e-05, ..., 1.3784e-07, + 4.6566e-08, 6.8694e-06], + ..., + [ 3.8184e-08, 1.3039e-08, 1.9278e-07, ..., 8.2888e-08, + -6.7055e-08, 1.3784e-07], + [ 1.0245e-08, 7.4506e-09, 2.5611e-07, ..., 4.0606e-07, + 1.6764e-07, 1.0151e-07], + [ 2.5891e-07, 3.9116e-08, 6.2920e-06, ..., 1.3504e-07, + 5.7742e-08, 3.2522e-06]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0370, -0.0276, 0.0094, 0.0306, -0.0242, 0.0027, -0.0087, 0.0137, + 0.0160, -0.0012], device='cuda:0'), grad: tensor([-5.0068e-05, 2.8834e-06, 2.9534e-05, 1.8794e-06, 2.3022e-06, + 4.1127e-06, -4.9472e-06, 7.7765e-07, 1.7630e-06, 1.1772e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.64, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.5137 re_mapping 0.0050 re_causal 0.0155 /// teacc 99.07 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0927, 0.0450, 0.0403, ..., -0.0725, -0.0973, 0.0627], + [-0.0658, -0.0252, -0.0856, ..., 0.1577, -0.0712, -0.0645], + [-0.0205, -0.0146, 0.1779, ..., -0.1046, -0.1216, -0.1041], + ..., + [-0.0906, -0.0498, 0.0086, ..., -0.0270, 0.1024, -0.0423], + [-0.0859, -0.0761, -0.0644, ..., -0.1324, -0.0051, -0.1427], + [-0.1593, -0.0437, -0.1217, ..., -0.1312, -0.0122, 0.0976]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, -1.8813e-07, ..., 2.9802e-08, + 1.3970e-08, -2.4773e-07], + [ 8.2888e-08, 0.0000e+00, 1.0245e-07, ..., -9.3132e-09, + 1.0338e-07, 5.4948e-08], + [ 3.6322e-08, 0.0000e+00, -1.8813e-07, ..., 2.5146e-08, + 4.3772e-08, 2.9802e-08], + ..., + [ 8.4750e-08, 9.3132e-10, -3.0175e-07, ..., 1.2107e-08, + -4.7591e-07, 3.8370e-07], + [ 1.6764e-08, 0.0000e+00, 4.2841e-08, ..., 3.6322e-08, + -1.7229e-07, 6.6124e-08], + [ 5.6811e-08, -1.8626e-09, 2.6636e-07, ..., 4.6566e-09, + 2.4680e-07, -6.9849e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0366, -0.0271, 0.0094, 0.0303, -0.0236, 0.0027, -0.0087, 0.0132, + 0.0172, -0.0025], device='cuda:0'), grad: tensor([-3.4831e-07, 1.0123e-06, 1.1213e-06, 6.6400e-05, 7.7300e-08, + 1.4156e-06, -1.2200e-07, 5.6624e-07, -6.9499e-05, -5.7742e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 214.30, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.4945 re_mapping 0.0048 re_causal 0.0146 /// teacc 99.02 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0931, 0.0459, 0.0408, ..., -0.0725, -0.1000, 0.0629], + [-0.0663, -0.0257, -0.0859, ..., 0.1579, -0.0713, -0.0652], + [-0.0206, -0.0144, 0.1781, ..., -0.1045, -0.1218, -0.1062], + ..., + [-0.0911, -0.0498, 0.0087, ..., -0.0273, 0.1027, -0.0426], + [-0.0862, -0.0763, -0.0666, ..., -0.1328, -0.0058, -0.1443], + [-0.1600, -0.0447, -0.1230, ..., -0.1319, -0.0131, 0.0977]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -2.4121e-07, ..., 1.4901e-08, + 9.3132e-09, 2.0489e-07], + [ 1.2107e-08, 0.0000e+00, 6.9849e-08, ..., -1.5460e-07, + 8.8476e-08, 1.6671e-07], + [ 2.7940e-09, 0.0000e+00, 1.9465e-07, ..., 3.4459e-08, + 7.9162e-08, 4.2655e-07], + ..., + [ 4.6566e-09, 0.0000e+00, 7.2923e-07, ..., 3.3528e-08, + 6.6496e-07, 4.9826e-07], + [ 1.8626e-09, 0.0000e+00, 2.2724e-07, ..., 4.2841e-08, + 2.3283e-07, 8.5216e-07], + [ 2.5146e-08, 0.0000e+00, 6.9849e-08, ..., 5.5879e-09, + 6.5193e-08, -1.3784e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0364, -0.0271, 0.0095, 0.0297, -0.0234, 0.0034, -0.0087, 0.0132, + 0.0169, -0.0030], device='cuda:0'), grad: tensor([ 1.3849e-06, 1.0571e-06, 2.9840e-06, -3.2723e-05, 2.5369e-06, + 1.8161e-06, 3.8221e-06, 1.2174e-05, 6.6534e-06, 2.8219e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 214.37, cls_loss 0.0022 cls_loss_mapping 0.0058 cls_loss_causal 0.5270 re_mapping 0.0050 re_causal 0.0151 /// teacc 99.05 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0938, 0.0460, 0.0409, ..., -0.0731, -0.1021, 0.0629], + [-0.0669, -0.0262, -0.0863, ..., 0.1581, -0.0715, -0.0655], + [-0.0212, -0.0143, 0.1785, ..., -0.1047, -0.1219, -0.1069], + ..., + [-0.0918, -0.0495, 0.0068, ..., -0.0274, 0.1000, -0.0435], + [-0.0867, -0.0764, -0.0679, ..., -0.1331, -0.0032, -0.1453], + [-0.1604, -0.0447, -0.1233, ..., -0.1326, -0.0137, 0.0981]], + device='cuda:0'), grad: tensor([[-3.0547e-07, 0.0000e+00, 2.1420e-08, ..., -5.7742e-08, + 1.6792e-06, -1.5954e-06], + [ 8.7544e-08, 0.0000e+00, 1.3104e-06, ..., -3.1106e-07, + 5.1372e-06, 3.5390e-08], + [ 4.0047e-08, 0.0000e+00, -1.8468e-06, ..., 1.0151e-07, + 3.8706e-06, -1.5460e-07], + ..., + [ 4.9360e-08, 0.0000e+00, -2.9698e-05, ..., 2.7008e-08, + -1.2124e-04, 1.9558e-08], + [ 3.1665e-08, 0.0000e+00, 1.7369e-06, ..., 8.7544e-08, + 6.8769e-06, 1.8626e-08], + [ 7.8883e-07, 0.0000e+00, 1.7434e-05, ..., 3.2596e-08, + 6.0678e-05, 3.7346e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0366, -0.0273, 0.0095, 0.0325, -0.0229, 0.0053, -0.0090, 0.0103, + 0.0152, -0.0031], device='cuda:0'), grad: tensor([ 1.8040e-06, 1.5065e-05, 9.1121e-06, 1.0324e-04, 2.8938e-05, + 3.3081e-06, 4.4610e-07, -3.6407e-04, 1.5348e-05, 1.8680e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 214.22, cls_loss 0.0009 cls_loss_mapping 0.0030 cls_loss_causal 0.5045 re_mapping 0.0047 re_causal 0.0153 /// teacc 99.05 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0947, 0.0462, 0.0411, ..., -0.0733, -0.1031, 0.0629], + [-0.0672, -0.0267, -0.0867, ..., 0.1582, -0.0716, -0.0660], + [-0.0213, -0.0142, 0.1789, ..., -0.1047, -0.1219, -0.1074], + ..., + [-0.0919, -0.0496, 0.0067, ..., -0.0274, 0.1001, -0.0443], + [-0.0867, -0.0766, -0.0681, ..., -0.1339, -0.0032, -0.1462], + [-0.1610, -0.0472, -0.1235, ..., -0.1330, -0.0143, 0.0987]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 0.0000e+00, -5.8673e-08, ..., 5.7742e-08, + 9.1270e-08, -3.6042e-07], + [ 6.7055e-08, 0.0000e+00, 1.4026e-06, ..., 6.7987e-08, + 9.0897e-07, 6.7987e-08], + [ 1.3784e-07, 0.0000e+00, 1.8980e-06, ..., 1.7323e-07, + 3.6135e-07, 4.4703e-08], + ..., + [ 1.5832e-08, 0.0000e+00, 3.6322e-07, ..., 5.5879e-09, + -2.4103e-06, 2.5164e-06], + [ 5.1223e-08, 0.0000e+00, 7.8510e-07, ..., 8.1025e-08, + 1.0617e-07, 6.9849e-08], + [ 2.1141e-07, 0.0000e+00, 2.4308e-07, ..., 2.7940e-09, + 5.8766e-07, -3.3919e-06]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0366, -0.0274, 0.0097, 0.0324, -0.0228, 0.0055, -0.0088, 0.0104, + 0.0152, -0.0032], device='cuda:0'), grad: tensor([ 2.2538e-07, 6.6645e-06, 8.9705e-06, -1.9267e-05, 3.1814e-06, + -6.8266e-07, 3.7067e-07, 4.6939e-06, 2.9355e-06, -7.0892e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 214.32, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4984 re_mapping 0.0046 re_causal 0.0151 /// teacc 99.00 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0951, 0.0467, 0.0414, ..., -0.0731, -0.1034, 0.0631], + [-0.0673, -0.0273, -0.0871, ..., 0.1583, -0.0717, -0.0664], + [-0.0214, -0.0142, 0.1792, ..., -0.1047, -0.1220, -0.1081], + ..., + [-0.0917, -0.0496, 0.0067, ..., -0.0274, 0.1002, -0.0448], + [-0.0867, -0.0767, -0.0681, ..., -0.1341, -0.0032, -0.1465], + [-0.1614, -0.0475, -0.1236, ..., -0.1339, -0.0145, 0.0994]], + device='cuda:0'), grad: tensor([[ 7.5437e-08, 1.4529e-07, 8.1025e-07, ..., 1.4249e-07, + 8.1770e-07, 1.6671e-07], + [ 3.1386e-07, 1.9558e-08, 1.3597e-07, ..., -1.0664e-06, + 1.1832e-05, 1.5460e-07], + [ 2.7940e-08, -4.1537e-07, -2.5388e-06, ..., 9.7789e-08, + 7.8976e-07, 5.9605e-08], + ..., + [-4.0419e-07, 5.4948e-08, 3.4459e-07, ..., 5.4948e-08, + -1.8626e-05, 1.3625e-06], + [ 3.2596e-08, 1.4901e-08, 9.9652e-08, ..., 1.6298e-07, + 4.0606e-07, 3.4086e-07], + [ 4.8429e-08, 1.6764e-08, 1.2759e-07, ..., 3.2596e-08, + 7.2550e-07, -4.0121e-06]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0364, -0.0276, 0.0098, 0.0323, -0.0230, 0.0055, -0.0088, 0.0104, + 0.0152, -0.0030], device='cuda:0'), grad: tensor([ 4.3325e-06, 3.3885e-05, -1.0738e-06, 6.2361e-06, 6.6198e-06, + 6.2361e-06, -1.0431e-07, -5.2005e-05, 2.3544e-06, -6.4149e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 214.36, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5103 re_mapping 0.0048 re_causal 0.0148 /// teacc 99.18 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0956, 0.0466, 0.0399, ..., -0.0730, -0.1041, 0.0607], + [-0.0691, -0.0270, -0.0880, ..., 0.1581, -0.0719, -0.0670], + [-0.0205, -0.0140, 0.1797, ..., -0.1042, -0.1220, -0.1086], + ..., + [-0.0921, -0.0496, 0.0067, ..., -0.0275, 0.1003, -0.0453], + [-0.0868, -0.0775, -0.0681, ..., -0.1339, -0.0032, -0.1469], + [-0.1620, -0.0483, -0.1220, ..., -0.1345, -0.0149, 0.1021]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, -8.1025e-08, -1.7900e-06, ..., 2.7940e-09, + 2.7008e-08, -1.6037e-06], + [ 1.3225e-07, 5.5879e-09, 1.5730e-06, ..., -1.0431e-07, + 8.1286e-06, 4.9360e-08], + [ 1.3411e-07, 7.4506e-09, 1.8626e-07, ..., 9.3132e-09, + 4.5076e-07, 1.1176e-07], + ..., + [ 2.8312e-07, 1.8626e-09, -2.6766e-06, ..., 4.4703e-08, + -1.4514e-05, 9.4995e-08], + [ 4.8429e-08, 1.8626e-09, 2.9802e-07, ..., 2.7008e-08, + 1.3681e-06, 9.0338e-08], + [ 6.0815e-07, 4.8429e-08, 1.1642e-06, ..., 1.2107e-08, + 9.7603e-07, 6.9756e-07]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0382, -0.0280, 0.0103, 0.0323, -0.0229, 0.0054, -0.0088, 0.0104, + 0.0152, -0.0019], device='cuda:0'), grad: tensor([-3.6359e-06, 3.1680e-05, 2.4382e-06, 1.4342e-05, -1.3309e-06, + 4.3958e-07, 1.1455e-06, -5.5492e-05, 4.1761e-06, 6.2436e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 214.30, cls_loss 0.0008 cls_loss_mapping 0.0028 cls_loss_causal 0.5012 re_mapping 0.0048 re_causal 0.0150 /// teacc 99.11 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0926, 0.0467, 0.0418, ..., -0.0731, -0.1048, 0.0606], + [-0.0699, -0.0282, -0.0884, ..., 0.1582, -0.0721, -0.0672], + [-0.0207, -0.0134, 0.1799, ..., -0.1041, -0.1221, -0.1091], + ..., + [-0.0926, -0.0496, 0.0068, ..., -0.0276, 0.1005, -0.0457], + [-0.0871, -0.0779, -0.0682, ..., -0.1341, -0.0033, -0.1476], + [-0.1625, -0.0485, -0.1221, ..., -0.1362, -0.0158, 0.1025]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.0396e-07, ..., 1.2107e-08, + 5.2154e-08, 1.1176e-08], + [ 9.3132e-10, 0.0000e+00, 2.6636e-07, ..., -3.6694e-07, + 9.1270e-07, 2.8871e-08], + [ 1.8626e-09, 0.0000e+00, -2.3562e-06, ..., 2.2538e-07, + 8.7358e-07, 7.4506e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 4.5635e-07, ..., 5.4017e-08, + -6.0461e-06, 1.2629e-06], + [ 1.8626e-09, 0.0000e+00, 3.5577e-07, ..., 4.0978e-08, + 4.1910e-08, 1.7695e-08], + [ 3.7253e-09, 0.0000e+00, 1.8720e-07, ..., 6.5193e-09, + 3.5688e-06, -2.4177e-06]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0373, -0.0281, 0.0103, 0.0323, -0.0231, 0.0053, -0.0091, 0.0105, + 0.0152, -0.0020], device='cuda:0'), grad: tensor([ 5.9977e-07, 2.1048e-06, -1.4715e-06, 8.9593e-07, 1.4156e-06, + 1.4836e-06, 1.4808e-07, -1.1213e-05, 9.5926e-07, 5.0403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 214.41, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5247 re_mapping 0.0046 re_causal 0.0146 /// teacc 99.20 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0927, 0.0474, 0.0420, ..., -0.0729, -0.1061, 0.0607], + [-0.0702, -0.0300, -0.0886, ..., 0.1584, -0.0723, -0.0674], + [-0.0209, -0.0128, 0.1802, ..., -0.1043, -0.1221, -0.1096], + ..., + [-0.0934, -0.0499, 0.0067, ..., -0.0277, 0.1006, -0.0464], + [-0.0874, -0.0784, -0.0683, ..., -0.1345, -0.0033, -0.1488], + [-0.1647, -0.0497, -0.1231, ..., -0.1371, -0.0165, 0.1028]], + device='cuda:0'), grad: tensor([[ 2.0210e-07, 1.6950e-07, -4.5635e-08, ..., 1.0245e-08, + 7.4506e-09, 1.7043e-07], + [ 2.5146e-08, 1.4901e-08, 5.4017e-08, ..., -9.6485e-07, + 9.6858e-08, 3.8184e-08], + [ 1.3690e-07, 1.0245e-07, 6.0536e-08, ..., 3.0734e-08, + 9.1270e-08, 1.2666e-07], + ..., + [ 2.8871e-08, 1.5832e-08, 1.0356e-06, ..., 8.0839e-07, + 1.1502e-06, 9.2201e-08], + [ 3.0734e-08, 2.3283e-08, 1.0338e-07, ..., 1.5832e-08, + 1.2293e-07, 1.0151e-07], + [ 2.1886e-07, 1.9558e-08, 5.6811e-08, ..., 1.2107e-08, + 5.3085e-08, -4.3027e-07]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0373, -0.0282, 0.0104, 0.0324, -0.0223, 0.0051, -0.0089, 0.0105, + 0.0152, -0.0026], device='cuda:0'), grad: tensor([ 1.2927e-06, -1.1800e-06, 1.2368e-06, -1.3590e-05, 2.4308e-07, + 5.0366e-06, 7.9162e-08, 6.6310e-06, 7.0594e-07, -4.7497e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.43, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4953 re_mapping 0.0045 re_causal 0.0140 /// teacc 99.13 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0927, 0.0475, 0.0423, ..., -0.0724, -0.1072, 0.0608], + [-0.0708, -0.0304, -0.0888, ..., 0.1586, -0.0728, -0.0683], + [-0.0211, -0.0125, 0.1804, ..., -0.1045, -0.1222, -0.1103], + ..., + [-0.0942, -0.0500, 0.0068, ..., -0.0279, 0.1008, -0.0464], + [-0.0877, -0.0786, -0.0683, ..., -0.1347, -0.0033, -0.1494], + [-0.1651, -0.0497, -0.1234, ..., -0.1378, -0.0173, 0.1032]], + device='cuda:0'), grad: tensor([[ 3.8184e-08, -2.3004e-07, -1.4575e-06, ..., 1.8813e-07, + 9.3132e-10, -1.6382e-06], + [ 1.0896e-07, 1.3970e-08, 6.0443e-07, ..., -7.4599e-07, + 2.1420e-08, 1.8720e-07], + [-1.6578e-07, 2.3283e-08, -1.2405e-06, ..., 2.5891e-07, + 4.3772e-08, 4.7497e-07], + ..., + [ 1.8813e-07, 4.6566e-09, 1.1083e-07, ..., 3.0454e-07, + -3.4459e-08, 3.4459e-07], + [ 1.6484e-07, 2.2352e-08, -4.3493e-07, ..., 4.3772e-08, + -7.1712e-08, 1.1772e-06], + [ 1.1651e-06, 9.0338e-08, 4.7591e-07, ..., 6.5193e-08, + 5.5879e-09, -1.1846e-05]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0372, -0.0285, 0.0104, 0.0324, -0.0224, 0.0050, -0.0088, 0.0107, + 0.0152, -0.0025], device='cuda:0'), grad: tensor([-2.6822e-06, 1.2750e-06, 1.2359e-06, 7.2718e-06, 2.9840e-06, + 2.5332e-05, -3.2410e-07, 2.4550e-06, -3.2969e-06, -3.4243e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 214.65, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.4769 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.09 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0930, 0.0477, 0.0425, ..., -0.0750, -0.1087, 0.0609], + [-0.0726, -0.0305, -0.0883, ..., 0.1605, -0.0732, -0.0648], + [-0.0199, -0.0125, 0.1804, ..., -0.1063, -0.1225, -0.1140], + ..., + [-0.0956, -0.0500, 0.0069, ..., -0.0283, 0.1010, -0.0472], + [-0.0876, -0.0787, -0.0684, ..., -0.1319, -0.0033, -0.1512], + [-0.1656, -0.0498, -0.1244, ..., -0.1430, -0.0175, 0.1035]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, -2.7940e-09, 3.1013e-07, ..., 1.8626e-09, + 1.3970e-08, -1.6764e-08], + [ 2.0862e-07, 0.0000e+00, 3.7588e-06, ..., -2.5146e-08, + 9.3319e-07, 8.6334e-07], + [-4.4797e-07, 9.3132e-10, -1.4110e-06, ..., 1.8626e-09, + 3.1572e-07, -8.7544e-08], + ..., + [ 8.4750e-08, 0.0000e+00, -7.2904e-06, ..., 1.3039e-08, + -1.7639e-06, 8.8662e-07], + [ 1.2759e-07, 0.0000e+00, 2.2613e-06, ..., 3.0734e-08, + 1.3597e-07, 3.7625e-07], + [ 3.2224e-07, 9.3132e-10, 9.4809e-07, ..., 9.3132e-10, + 1.4994e-07, -2.5872e-06]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0374, -0.0276, 0.0095, 0.0323, -0.0225, 0.0047, -0.0111, 0.0108, + 0.0160, -0.0026], device='cuda:0'), grad: tensor([ 8.4471e-07, 2.6986e-05, 1.0014e-05, 2.3395e-06, 1.6605e-06, + -2.0862e-07, 1.2778e-06, -4.9740e-05, 9.4175e-06, -2.6803e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.27, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.5004 re_mapping 0.0047 re_causal 0.0143 /// teacc 99.14 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0931, 0.0483, 0.0430, ..., -0.0745, -0.1094, 0.0612], + [-0.0728, -0.0314, -0.0884, ..., 0.1612, -0.0734, -0.0644], + [-0.0200, -0.0124, 0.1808, ..., -0.1068, -0.1226, -0.1145], + ..., + [-0.0963, -0.0502, 0.0070, ..., -0.0286, 0.1012, -0.0475], + [-0.0878, -0.0792, -0.0686, ..., -0.1319, -0.0036, -0.1538], + [-0.1660, -0.0499, -0.1250, ..., -0.1443, -0.0154, 0.1053]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, -4.6566e-10, 4.7730e-07, ..., 3.5390e-08, + 2.4680e-08, -6.0536e-09], + [ 1.1781e-07, 0.0000e+00, 2.6356e-07, ..., 1.3644e-07, + 2.0256e-07, 9.3132e-10], + [ 7.1712e-08, 0.0000e+00, -3.5129e-06, ..., 1.4761e-07, + 1.2061e-07, 4.6566e-10], + ..., + [ 3.4645e-07, 0.0000e+00, -1.2387e-07, ..., 1.0198e-07, + -7.7672e-07, 3.2596e-09], + [ 8.3819e-09, 0.0000e+00, 4.2841e-07, ..., 4.8429e-07, + 3.7253e-09, 9.3132e-10], + [ 2.7474e-08, 0.0000e+00, 1.3085e-07, ..., 1.1176e-08, + 1.2387e-07, -4.1910e-09]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0370, -0.0273, 0.0094, 0.0323, -0.0226, 0.0037, -0.0112, 0.0109, + 0.0156, -0.0011], device='cuda:0'), grad: tensor([ 8.2469e-07, 1.5376e-06, -3.5539e-06, 6.3702e-07, -1.5227e-07, + 4.6007e-07, -1.2591e-06, -2.8266e-07, 1.3672e-06, 3.9814e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 214.39, cls_loss 0.0008 cls_loss_mapping 0.0029 cls_loss_causal 0.4953 re_mapping 0.0050 re_causal 0.0153 /// teacc 99.11 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0933, 0.0495, 0.0428, ..., -0.0727, -0.1103, 0.0612], + [-0.0725, -0.0348, -0.0884, ..., 0.1617, -0.0735, -0.0645], + [-0.0207, -0.0122, 0.1819, ..., -0.1074, -0.1228, -0.1147], + ..., + [-0.0970, -0.0503, 0.0065, ..., -0.0287, 0.1012, -0.0478], + [-0.0882, -0.0796, -0.0688, ..., -0.1319, -0.0036, -0.1539], + [-0.1660, -0.0499, -0.1251, ..., -0.1447, -0.0154, 0.1055]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, -0.0000e+00, -2.2352e-08, ..., 1.8626e-09, + 7.9162e-09, -1.2573e-08], + [ 3.4459e-08, 0.0000e+00, 1.6717e-07, ..., -1.8626e-08, + 1.1362e-07, 2.0489e-08], + [ 1.8626e-08, 0.0000e+00, -7.7346e-07, ..., 5.5879e-09, + 8.4285e-08, 6.8452e-08], + ..., + [ 3.5390e-08, 0.0000e+00, -1.2759e-07, ..., 1.0710e-08, + -2.6822e-07, 2.2817e-08], + [ 1.7229e-08, 0.0000e+00, 5.7602e-07, ..., 3.7253e-09, + 4.1444e-08, 5.9139e-08], + [ 5.0291e-07, 0.0000e+00, 5.4017e-08, ..., 9.3132e-10, + 8.2422e-08, 1.8161e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0368, -0.0272, 0.0099, 0.0324, -0.0228, 0.0037, -0.0112, 0.0106, + 0.0156, -0.0009], device='cuda:0'), grad: tensor([ 9.1270e-08, 7.9768e-07, -7.7533e-07, 6.5286e-07, -7.3900e-07, + -1.1213e-06, 2.9430e-07, -2.5332e-07, 5.1688e-08, 9.9838e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 214.44, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.4940 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.17 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0935, 0.0498, 0.0427, ..., -0.0734, -0.1114, 0.0595], + [-0.0740, -0.0350, -0.0887, ..., 0.1623, -0.0735, -0.0645], + [-0.0222, -0.0119, 0.1821, ..., -0.1080, -0.1230, -0.1149], + ..., + [-0.0963, -0.0503, 0.0065, ..., -0.0294, 0.1014, -0.0479], + [-0.0889, -0.0797, -0.0690, ..., -0.1320, -0.0036, -0.1541], + [-0.1695, -0.0501, -0.1257, ..., -0.1452, -0.0171, 0.1068]], + device='cuda:0'), grad: tensor([[ 1.5367e-08, -4.6566e-10, -1.2387e-06, ..., 6.3796e-08, + 1.1642e-08, -1.5255e-06], + [ 8.8010e-08, 0.0000e+00, 1.6904e-06, ..., 9.6764e-07, + 2.8480e-06, 5.2154e-08], + [ 4.8429e-08, 0.0000e+00, 3.6042e-07, ..., 2.3097e-07, + 6.8964e-07, 3.3062e-08], + ..., + [ 2.6543e-07, 0.0000e+00, -2.6971e-06, ..., -1.4212e-06, + -4.6194e-06, 6.3004e-07], + [ 3.7719e-08, 0.0000e+00, 7.4040e-08, ..., 1.5367e-07, + 1.1176e-07, 7.1246e-08], + [ 4.6566e-07, 0.0000e+00, 1.2442e-06, ..., 9.6858e-08, + 2.6077e-07, 5.7369e-07]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0386, -0.0270, 0.0096, 0.0323, -0.0203, 0.0036, -0.0106, 0.0106, + 0.0155, -0.0018], device='cuda:0'), grad: tensor([-3.5446e-06, 7.8455e-06, 1.9614e-06, 6.8685e-07, 1.6950e-07, + 3.4925e-08, -8.9500e-07, -7.7114e-06, 7.4925e-07, 6.7241e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 214.35, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4958 re_mapping 0.0044 re_causal 0.0138 /// teacc 99.16 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0936, 0.0529, 0.0437, ..., -0.0738, -0.1117, 0.0599], + [-0.0744, -0.0354, -0.0892, ..., 0.1625, -0.0739, -0.0647], + [-0.0224, -0.0115, 0.1830, ..., -0.1082, -0.1227, -0.1151], + ..., + [-0.0965, -0.0504, 0.0064, ..., -0.0297, 0.1015, -0.0483], + [-0.0874, -0.0812, -0.0693, ..., -0.1321, -0.0036, -0.1542], + [-0.1699, -0.0516, -0.1267, ..., -0.1457, -0.0172, 0.1068]], + device='cuda:0'), grad: tensor([[ 9.7416e-07, 0.0000e+00, 5.9512e-07, ..., 9.7789e-09, + 1.8440e-07, 8.8476e-09], + [ 3.8045e-07, 0.0000e+00, 1.3225e-07, ..., -9.1270e-08, + 8.1770e-07, 1.3970e-09], + [ 3.1888e-06, 0.0000e+00, 8.3260e-07, ..., 3.8184e-08, + 6.8406e-07, 6.5193e-09], + ..., + [ 7.4552e-07, 0.0000e+00, -8.1444e-07, ..., 2.6543e-08, + -1.2934e-05, 4.6566e-09], + [ 1.9232e-07, 0.0000e+00, -7.4506e-07, ..., 3.2596e-08, + -5.4948e-08, 4.1910e-09], + [ 4.2953e-06, 0.0000e+00, 3.6135e-07, ..., 4.1910e-09, + 1.7146e-06, -1.1642e-08]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0381, -0.0272, 0.0100, 0.0323, -0.0207, 0.0035, -0.0103, 0.0106, + 0.0155, -0.0020], device='cuda:0'), grad: tensor([ 4.9658e-06, 2.9262e-06, 9.5516e-06, -3.6173e-06, 3.2354e-06, + 2.1290e-06, 2.0191e-06, -3.1441e-05, -2.2873e-06, 1.2487e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 214.83, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5285 re_mapping 0.0045 re_causal 0.0138 /// teacc 99.14 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.0937, 0.0533, 0.0442, ..., -0.0738, -0.1131, 0.0601], + [-0.0749, -0.0350, -0.0894, ..., 0.1629, -0.0741, -0.0647], + [-0.0227, -0.0113, 0.1832, ..., -0.1083, -0.1230, -0.1152], + ..., + [-0.0984, -0.0508, 0.0066, ..., -0.0301, 0.1017, -0.0487], + [-0.0877, -0.0818, -0.0694, ..., -0.1322, -0.0037, -0.1543], + [-0.1701, -0.0519, -0.1273, ..., -0.1462, -0.0176, 0.1069]], + device='cuda:0'), grad: tensor([[ 5.1688e-08, 8.3819e-09, -5.4948e-08, ..., 1.4668e-07, + 0.0000e+00, -5.3551e-08], + [ 6.8452e-08, 1.0710e-08, 1.3458e-07, ..., -5.9530e-06, + 0.0000e+00, 1.9046e-07], + [ 5.0757e-08, -8.2888e-08, -1.4752e-06, ..., 4.4033e-06, + -4.6566e-10, 1.3411e-07], + ..., + [ 3.2131e-08, 3.7253e-09, 5.6997e-07, ..., 3.2084e-07, + 4.6566e-10, 3.5390e-08], + [ 9.7323e-08, 1.3039e-08, 1.3178e-07, ..., 4.2468e-07, + 4.6566e-10, 1.5646e-07], + [ 4.6100e-08, 4.6566e-10, 9.1735e-08, ..., 5.3085e-08, + 4.6566e-10, -1.6680e-06]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0379, -0.0273, 0.0098, 0.0323, -0.0206, 0.0037, -0.0103, 0.0107, + 0.0155, -0.0022], device='cuda:0'), grad: tensor([ 6.0583e-07, -1.2740e-05, 1.0014e-05, -2.1532e-06, 2.5593e-06, + 4.9621e-06, 4.1258e-07, 1.7695e-06, 2.7195e-07, -5.7071e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 208---------------------------------------------------- +epoch 208, time 231.03, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.4980 re_mapping 0.0046 re_causal 0.0142 /// teacc 99.21 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.0937, 0.0532, 0.0445, ..., -0.0738, -0.1137, 0.0603], + [-0.0750, -0.0356, -0.0899, ..., 0.1633, -0.0742, -0.0650], + [-0.0223, -0.0108, 0.1847, ..., -0.1084, -0.1230, -0.1153], + ..., + [-0.0995, -0.0508, 0.0066, ..., -0.0306, 0.1018, -0.0492], + [-0.0879, -0.0822, -0.0704, ..., -0.1322, -0.0037, -0.1544], + [-0.1704, -0.0521, -0.1276, ..., -0.1464, -0.0180, 0.1070]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -8.9779e-07, ..., 3.5856e-08, + 7.9162e-09, -8.2562e-07], + [ 1.0245e-08, 0.0000e+00, 3.2876e-07, ..., -3.0128e-07, + 1.7881e-07, 5.0524e-07], + [ 9.7789e-09, 0.0000e+00, -1.6019e-06, ..., 6.9384e-08, + 1.3970e-09, 5.6345e-08], + ..., + [ 9.7789e-09, 0.0000e+00, 8.0699e-07, ..., 7.8231e-08, + 6.5193e-09, 4.6240e-07], + [ 4.6566e-09, 0.0000e+00, 3.2922e-07, ..., 1.1222e-07, + 8.5216e-08, 2.6496e-07], + [-5.3551e-08, 0.0000e+00, 1.8813e-07, ..., 3.0268e-08, + 2.5611e-08, -2.1886e-07]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0376, -0.0274, 0.0111, 0.0322, -0.0206, 0.0036, -0.0102, 0.0107, + 0.0152, -0.0023], device='cuda:0'), grad: tensor([-1.8757e-06, 8.2143e-07, -1.9819e-06, -8.6706e-07, 1.2489e-06, + -1.8198e-06, 2.2221e-06, 1.9167e-06, 4.0280e-07, -8.3353e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 214.77, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.5246 re_mapping 0.0043 re_causal 0.0144 /// teacc 99.18 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0936, 0.0554, 0.0459, ..., -0.0731, -0.1145, 0.0608], + [-0.0752, -0.0358, -0.0901, ..., 0.1634, -0.0744, -0.0651], + [-0.0225, -0.0114, 0.1847, ..., -0.1086, -0.1233, -0.1154], + ..., + [-0.0995, -0.0508, 0.0066, ..., -0.0308, 0.1018, -0.0495], + [-0.0879, -0.0825, -0.0705, ..., -0.1323, -0.0037, -0.1545], + [-0.1705, -0.0530, -0.1282, ..., -0.1467, -0.0182, 0.1071]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 0.0000e+00, -5.5172e-06, ..., 1.8626e-08, + 1.0245e-08, -7.4171e-06], + [ 2.5611e-08, 0.0000e+00, 2.3842e-07, ..., 1.3039e-08, + 7.8650e-07, 6.9384e-08], + [ 1.9372e-07, 0.0000e+00, 2.1514e-07, ..., 3.4925e-08, + 5.4203e-07, 4.1910e-08], + ..., + [ 1.0710e-08, 0.0000e+00, -7.5065e-07, ..., 4.1910e-09, + -2.3469e-06, 3.1432e-07], + [ 6.6124e-07, 0.0000e+00, 1.0896e-07, ..., 4.6566e-08, + 1.8300e-07, 9.4529e-08], + [ 7.4506e-09, 0.0000e+00, 2.5798e-06, ..., 2.3283e-09, + 4.8382e-07, 2.4680e-06]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0369, -0.0274, 0.0109, 0.0324, -0.0207, 0.0036, -0.0103, 0.0106, + 0.0152, -0.0024], device='cuda:0'), grad: tensor([-1.3307e-05, 3.8296e-06, 6.4932e-06, -2.1577e-05, 1.9651e-06, + 1.1690e-05, -1.4948e-07, -8.5384e-06, 1.3597e-05, 5.9642e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 214.26, cls_loss 0.0010 cls_loss_mapping 0.0028 cls_loss_causal 0.5213 re_mapping 0.0046 re_causal 0.0148 /// teacc 99.09 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0937, 0.0583, 0.0469, ..., -0.0740, -0.1165, 0.0609], + [-0.0764, -0.0362, -0.0911, ..., 0.1633, -0.0746, -0.0652], + [-0.0218, -0.0113, 0.1848, ..., -0.1083, -0.1239, -0.1157], + ..., + [-0.0997, -0.0509, 0.0074, ..., -0.0309, 0.1022, -0.0491], + [-0.0883, -0.0845, -0.0706, ..., -0.1323, -0.0037, -0.1549], + [-0.1697, -0.0555, -0.1302, ..., -0.1477, -0.0185, 0.1081]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.7695e-07, -4.2468e-07, ..., 9.7789e-09, + 8.8476e-09, -3.0780e-07], + [ 8.8941e-08, 6.0536e-09, 2.4308e-06, ..., -4.9872e-07, + 4.1118e-07, 1.6298e-08], + [ 1.3970e-09, 1.3039e-08, -7.8306e-06, ..., 9.5461e-08, + -1.1269e-06, 2.3749e-08], + ..., + [ 3.4925e-08, 2.2352e-08, 3.3397e-06, ..., 2.7940e-07, + 3.9255e-07, 6.9849e-08], + [ 1.3039e-08, 2.9337e-08, 2.4121e-07, ..., 5.2154e-08, + 2.7008e-08, 5.9605e-08], + [ 1.2899e-07, 6.2399e-08, 2.7148e-07, ..., 1.3970e-08, + 4.7963e-08, 8.5682e-08]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0366, -0.0278, 0.0108, 0.0325, -0.0216, 0.0034, -0.0102, 0.0110, + 0.0151, -0.0023], device='cuda:0'), grad: tensor([-1.2023e-06, 1.7527e-06, -8.2701e-06, 1.5646e-06, -1.2154e-07, + 5.3272e-07, 2.6682e-07, 4.3288e-06, 3.6508e-07, 7.7207e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 214.32, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.5126 re_mapping 0.0048 re_causal 0.0147 /// teacc 99.20 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0938, 0.0583, 0.0478, ..., -0.0762, -0.1174, 0.0609], + [-0.0768, -0.0377, -0.0915, ..., 0.1638, -0.0749, -0.0676], + [-0.0220, -0.0101, 0.1858, ..., -0.1085, -0.1242, -0.1157], + ..., + [-0.1000, -0.0510, 0.0071, ..., -0.0314, 0.1023, -0.0507], + [-0.0884, -0.0847, -0.0706, ..., -0.1324, -0.0038, -0.1556], + [-0.1699, -0.0555, -0.1308, ..., -0.1481, -0.0184, 0.1091]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 2.3283e-09, 1.8813e-07, ..., 6.9849e-09, + 3.0082e-07, -6.8918e-08], + [ 1.2061e-07, 1.2573e-08, 1.3243e-06, ..., -1.1548e-06, + 1.5646e-06, 4.7963e-08], + [ 2.5611e-08, -5.4482e-08, -2.2668e-06, ..., 9.3179e-07, + 5.4901e-07, 1.2107e-08], + ..., + [-2.6869e-07, 6.9849e-09, -1.8012e-06, ..., 0.0000e+00, + -4.1015e-06, 6.6496e-07], + [ 1.0245e-08, 1.1642e-08, 6.4727e-07, ..., 1.4435e-08, + 1.8300e-07, 2.8871e-08], + [ 2.4214e-08, 1.8626e-09, 3.3900e-07, ..., 3.7253e-09, + 2.5844e-07, -1.0198e-06]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0365, -0.0283, 0.0113, 0.0323, -0.0207, 0.0036, -0.0108, 0.0108, + 0.0151, -0.0017], device='cuda:0'), grad: tensor([ 1.3132e-06, 4.8466e-06, -1.1623e-06, 1.9949e-06, 1.7183e-06, + 1.2452e-06, 4.3493e-07, -1.2614e-05, 2.4214e-06, -2.1886e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 214.45, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.4916 re_mapping 0.0046 re_causal 0.0139 /// teacc 99.15 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0938, 0.0583, 0.0482, ..., -0.0764, -0.1180, 0.0610], + [-0.0769, -0.0378, -0.0919, ..., 0.1642, -0.0753, -0.0678], + [-0.0221, -0.0099, 0.1864, ..., -0.1086, -0.1243, -0.1158], + ..., + [-0.1005, -0.0511, 0.0071, ..., -0.0316, 0.1025, -0.0519], + [-0.0887, -0.0848, -0.0708, ..., -0.1325, -0.0038, -0.1557], + [-0.1706, -0.0555, -0.1315, ..., -0.1486, -0.0186, 0.1094]], + device='cuda:0'), grad: tensor([[ 3.9581e-08, 0.0000e+00, -8.7619e-06, ..., 7.4506e-09, + 3.6787e-08, -7.4320e-06], + [ 2.2445e-07, 0.0000e+00, 1.4948e-07, ..., -8.4983e-07, + 4.3772e-07, 1.6810e-07], + [ 9.8720e-08, 0.0000e+00, 2.5649e-06, ..., 6.7055e-08, + 2.7940e-08, 2.5034e-06], + ..., + [ 2.6543e-07, 0.0000e+00, 1.9465e-07, ..., 4.4331e-07, + -4.0559e-07, 8.5216e-08], + [ 1.0245e-07, 0.0000e+00, 1.2778e-06, ..., 3.9116e-08, + 1.2480e-07, 1.1539e-06], + [ 7.6294e-06, 0.0000e+00, 1.5832e-06, ..., 1.2107e-08, + 4.4936e-07, 1.1288e-06]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0364, -0.0284, 0.0116, 0.0323, -0.0205, 0.0035, -0.0109, 0.0108, + 0.0151, -0.0018], device='cuda:0'), grad: tensor([-2.0877e-05, 6.7661e-07, 7.0110e-06, 8.1733e-06, -1.2778e-05, + -5.6103e-06, 5.6997e-06, 3.4971e-07, 3.2019e-06, 1.4104e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 215.01, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4989 re_mapping 0.0047 re_causal 0.0145 /// teacc 99.16 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0940, 0.0583, 0.0489, ..., -0.0770, -0.1193, 0.0612], + [-0.0773, -0.0388, -0.0924, ..., 0.1643, -0.0756, -0.0678], + [-0.0227, -0.0104, 0.1866, ..., -0.1087, -0.1246, -0.1161], + ..., + [-0.1017, -0.0513, 0.0072, ..., -0.0317, 0.1027, -0.0527], + [-0.0892, -0.0854, -0.0710, ..., -0.1326, -0.0038, -0.1560], + [-0.1711, -0.0559, -0.1320, ..., -0.1490, -0.0187, 0.1095]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 3.2596e-09, ..., 2.2817e-08, + 1.4435e-08, 3.7253e-09], + [ 3.7719e-08, 0.0000e+00, 1.7509e-07, ..., -4.5262e-07, + 2.5658e-07, -5.6345e-08], + [ 3.7253e-08, 0.0000e+00, 1.7369e-07, ..., 3.1199e-08, + 2.5146e-07, 1.0245e-08], + ..., + [-1.0198e-07, 0.0000e+00, -1.3653e-06, ..., 6.8918e-08, + -2.0210e-06, 2.1886e-08], + [ 5.6811e-08, 0.0000e+00, 4.5961e-07, ..., 1.0757e-07, + 6.9756e-07, 4.5635e-08], + [ 2.0070e-07, 0.0000e+00, 9.0804e-08, ..., 1.9558e-08, + 1.2806e-07, -4.0978e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0363, -0.0287, 0.0113, 0.0323, -0.0202, 0.0044, -0.0116, 0.0110, + 0.0150, -0.0019], device='cuda:0'), grad: tensor([ 1.1828e-07, -2.0256e-07, 6.6450e-07, -6.5193e-07, 5.1735e-07, + 9.3272e-07, 3.3062e-07, -3.7961e-06, 1.3774e-06, 7.0408e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.59, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.5255 re_mapping 0.0046 re_causal 0.0146 /// teacc 99.17 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.0939, 0.0584, 0.0497, ..., -0.0766, -0.1195, 0.0613], + [-0.0780, -0.0399, -0.0936, ..., 0.1643, -0.0761, -0.0679], + [-0.0229, -0.0097, 0.1872, ..., -0.1085, -0.1248, -0.1162], + ..., + [-0.1023, -0.0519, 0.0073, ..., -0.0318, 0.1030, -0.0534], + [-0.0893, -0.0857, -0.0711, ..., -0.1326, -0.0038, -0.1563], + [-0.1712, -0.0545, -0.1325, ..., -0.1498, -0.0192, 0.1100]], + device='cuda:0'), grad: tensor([[ 3.9116e-08, 0.0000e+00, -7.1246e-08, ..., 5.0757e-08, + 5.8673e-08, 5.3551e-08], + [ 1.0943e-07, 0.0000e+00, -1.1185e-06, ..., -6.6422e-06, + 8.3819e-09, 1.2107e-08], + [ 9.8255e-08, 0.0000e+00, 6.5006e-07, ..., 3.6713e-06, + 5.1223e-09, 1.6764e-08], + ..., + [ 4.6846e-07, 0.0000e+00, 6.8778e-07, ..., 1.8291e-06, + 3.6694e-07, 1.0161e-06], + [ 8.3819e-09, 0.0000e+00, 5.1688e-08, ..., 7.2643e-08, + 7.0781e-08, 8.6147e-08], + [ 5.6028e-06, 0.0000e+00, -5.5041e-07, ..., 2.4680e-08, + -5.5926e-07, -1.0654e-06]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0360, -0.0292, 0.0116, 0.0322, -0.0204, 0.0042, -0.0115, 0.0112, + 0.0150, -0.0017], device='cuda:0'), grad: tensor([ 1.8906e-07, -9.8124e-06, 5.6662e-06, 1.2433e-06, -8.4639e-06, + 1.1660e-06, -1.9651e-07, 5.0217e-06, 2.5379e-07, 4.9174e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 214.37, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4931 re_mapping 0.0045 re_causal 0.0136 /// teacc 99.18 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0945, 0.0584, 0.0498, ..., -0.0782, -0.1208, 0.0613], + [-0.0782, -0.0399, -0.0939, ..., 0.1646, -0.0765, -0.0680], + [-0.0233, -0.0092, 0.1864, ..., -0.1086, -0.1270, -0.1163], + ..., + [-0.1029, -0.0520, 0.0084, ..., -0.0320, 0.1037, -0.0538], + [-0.0886, -0.0858, -0.0713, ..., -0.1327, -0.0038, -0.1564], + [-0.1719, -0.0545, -0.1327, ..., -0.1502, -0.0199, 0.1101]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, -2.7940e-09, -2.4494e-07, ..., 3.4925e-08, + 9.3132e-10, -2.4075e-07], + [ 9.4995e-08, 4.6566e-10, 6.6496e-07, ..., -4.2841e-08, + 2.3283e-09, 7.4506e-09], + [-1.2945e-07, 2.3283e-09, -2.2594e-06, ..., -5.2620e-08, + 5.5879e-09, 1.4435e-08], + ..., + [ 1.7229e-08, 4.6566e-10, 6.5239e-07, ..., 2.1886e-08, + 2.6077e-08, 2.1886e-08], + [ 1.6764e-08, 9.3132e-10, 6.9197e-07, ..., 4.9826e-08, + 2.3283e-09, 4.2375e-08], + [ 8.2888e-08, 2.3283e-09, 1.4715e-07, ..., 4.1910e-09, + 1.3970e-09, 7.9628e-08]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0363, -0.0293, 0.0108, 0.0321, -0.0201, 0.0044, -0.0115, 0.0117, + 0.0151, -0.0019], device='cuda:0'), grad: tensor([-2.4261e-07, 1.7518e-06, -5.8450e-06, -1.3523e-06, 1.2480e-07, + 5.2899e-07, 1.6298e-08, 2.6878e-06, 2.0303e-06, 3.0128e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 216---------------------------------------------------- +epoch 216, time 230.96, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5178 re_mapping 0.0043 re_causal 0.0136 /// teacc 99.24 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0946, 0.0588, 0.0506, ..., -0.0784, -0.1212, 0.0616], + [-0.0785, -0.0411, -0.0941, ..., 0.1647, -0.0767, -0.0680], + [-0.0235, -0.0058, 0.1871, ..., -0.1087, -0.1270, -0.1164], + ..., + [-0.1033, -0.0538, 0.0083, ..., -0.0321, 0.1038, -0.0543], + [-0.0890, -0.0891, -0.0716, ..., -0.1327, -0.0039, -0.1566], + [-0.1723, -0.0546, -0.1331, ..., -0.1505, -0.0201, 0.1102]], + device='cuda:0'), grad: tensor([[ 9.6392e-08, 8.0559e-08, -2.4550e-06, ..., 1.5367e-08, + 1.3504e-08, -3.0175e-06], + [ 6.2399e-08, 5.1223e-08, 4.9686e-07, ..., -1.5553e-07, + 6.9849e-08, 1.0664e-07], + [-2.2817e-06, -1.9204e-06, -7.4729e-06, ..., 2.7940e-08, + 5.1223e-08, 3.1665e-07], + ..., + [ 2.4214e-07, 2.0303e-07, 9.9652e-07, ..., 1.3225e-07, + -1.4435e-07, 2.1374e-07], + [ 1.3690e-07, 1.1455e-07, 9.6764e-07, ..., 1.3504e-08, + 6.7055e-08, 2.2724e-07], + [ 8.3819e-09, 5.1223e-09, 9.9186e-07, ..., 1.8626e-09, + 4.0513e-08, 6.0489e-07]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0360, -0.0294, 0.0112, 0.0320, -0.0200, 0.0045, -0.0115, 0.0117, + 0.0150, -0.0020], device='cuda:0'), grad: tensor([-8.0764e-06, 1.0151e-06, -1.0565e-05, 5.4948e-07, 5.3421e-06, + 2.8256e-06, 3.1311e-06, 1.8002e-06, 1.4566e-06, 2.5071e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 214.45, cls_loss 0.0008 cls_loss_mapping 0.0030 cls_loss_causal 0.4959 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.14 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0948, 0.0588, 0.0511, ..., -0.0787, -0.1226, 0.0618], + [-0.0793, -0.0437, -0.0952, ..., 0.1645, -0.0773, -0.0681], + [-0.0237, -0.0042, 0.1874, ..., -0.1087, -0.1274, -0.1166], + ..., + [-0.1035, -0.0543, 0.0085, ..., -0.0321, 0.1043, -0.0545], + [-0.0899, -0.0897, -0.0717, ..., -0.1328, -0.0039, -0.1568], + [-0.1725, -0.0544, -0.1334, ..., -0.1508, -0.0204, 0.1104]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, 4.6566e-10, 1.6764e-08, ..., 2.0908e-07, + 4.2375e-08, -9.3132e-10], + [ 3.6787e-08, 4.6566e-10, 1.8487e-07, ..., -7.4971e-08, + 1.0990e-07, 1.1642e-08], + [ 4.0513e-08, 1.8626e-09, 1.9930e-06, ..., 1.7090e-07, + 1.2135e-06, 2.6077e-08], + ..., + [ 2.3283e-08, 1.6764e-08, -3.5986e-06, ..., 8.8941e-08, + -2.1867e-06, 8.5682e-08], + [ 6.7055e-08, 1.3970e-09, 9.6112e-07, ..., 3.0408e-07, + 5.6298e-07, 1.8813e-07], + [ 1.8952e-07, 4.6566e-10, -4.2841e-08, ..., 2.6077e-08, + 9.4529e-08, -3.0221e-07]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0359, -0.0300, 0.0112, 0.0321, -0.0200, 0.0040, -0.0112, 0.0120, + 0.0150, -0.0019], device='cuda:0'), grad: tensor([ 7.3668e-07, 3.4552e-07, 4.0531e-06, 9.6019e-07, 5.6997e-07, + -1.3653e-06, -3.1795e-06, -4.9025e-06, 2.8387e-06, -5.9605e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 214.54, cls_loss 0.0008 cls_loss_mapping 0.0030 cls_loss_causal 0.5143 re_mapping 0.0042 re_causal 0.0136 /// teacc 99.16 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.0949, 0.0586, 0.0516, ..., -0.0791, -0.1231, 0.0618], + [-0.0799, -0.0438, -0.0954, ..., 0.1653, -0.0774, -0.0682], + [-0.0242, -0.0042, 0.1875, ..., -0.1088, -0.1275, -0.1168], + ..., + [-0.1043, -0.0543, 0.0085, ..., -0.0336, 0.1044, -0.0549], + [-0.0906, -0.0899, -0.0717, ..., -0.1330, -0.0039, -0.1571], + [-0.1729, -0.0545, -0.1337, ..., -0.1511, -0.0207, 0.1106]], + device='cuda:0'), grad: tensor([[ 1.2293e-07, 0.0000e+00, -2.7809e-06, ..., -2.0564e-06, + 3.2969e-07, -1.2934e-05], + [ 5.9744e-07, 0.0000e+00, 3.7858e-07, ..., 4.5169e-08, + 1.6633e-06, 6.2957e-07], + [ 2.4494e-07, 0.0000e+00, 8.2422e-07, ..., 2.8871e-08, + 1.2387e-06, 2.6356e-07], + ..., + [-2.2370e-06, 0.0000e+00, -1.6922e-06, ..., 8.1025e-08, + -6.9067e-06, -3.8138e-07], + [ 6.1933e-08, 0.0000e+00, 1.2945e-07, ..., 9.0338e-08, + 1.4342e-07, 5.2247e-07], + [ 3.6974e-07, 0.0000e+00, 2.0750e-06, ..., 1.3970e-06, + 9.3877e-07, 8.8066e-06]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0358, -0.0296, 0.0111, 0.0321, -0.0199, 0.0040, -0.0111, 0.0118, + 0.0149, -0.0019], device='cuda:0'), grad: tensor([-6.0648e-05, 6.1952e-06, 3.6471e-06, -6.4448e-06, 1.6745e-06, + 1.7181e-05, 7.1004e-06, -1.5289e-05, 2.7474e-06, 4.3809e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 214.40, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.5042 re_mapping 0.0044 re_causal 0.0143 /// teacc 99.17 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.0951, 0.0586, 0.0520, ..., -0.0806, -0.1240, 0.0618], + [-0.0803, -0.0439, -0.0958, ..., 0.1654, -0.0776, -0.0682], + [-0.0248, -0.0042, 0.1874, ..., -0.1095, -0.1277, -0.1169], + ..., + [-0.1051, -0.0544, 0.0089, ..., -0.0325, 0.1045, -0.0552], + [-0.0912, -0.0899, -0.0718, ..., -0.1331, -0.0039, -0.1573], + [-0.1739, -0.0545, -0.1344, ..., -0.1516, -0.0213, 0.1106]], + device='cuda:0'), grad: tensor([[ 5.5367e-07, 0.0000e+00, 4.7777e-07, ..., 1.1176e-07, + 4.6100e-08, 8.7405e-07], + [ 1.5832e-08, 0.0000e+00, 1.1316e-07, ..., -1.3970e-09, + 6.4727e-08, 1.4435e-08], + [ 1.1642e-08, -4.6566e-10, -2.6263e-07, ..., 1.0245e-08, + 9.9186e-08, 3.9116e-08], + ..., + [ 5.0757e-08, 0.0000e+00, -1.7695e-07, ..., 2.3283e-09, + 8.8476e-09, 1.3318e-07], + [ 4.1444e-08, 4.6566e-10, 1.6531e-07, ..., 1.0245e-08, + -2.3236e-07, 4.1444e-08], + [ 8.0094e-08, 0.0000e+00, 3.9581e-08, ..., 1.8626e-09, + 4.0047e-08, -4.0047e-07]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0359, -0.0300, 0.0098, 0.0322, -0.0195, 0.0041, -0.0108, 0.0125, + 0.0148, -0.0023], device='cuda:0'), grad: tensor([ 4.3809e-06, 3.0408e-07, -2.0023e-08, 3.7067e-07, -7.7439e-07, + -5.5460e-07, -3.0156e-06, 1.5423e-06, -1.5050e-06, -7.2503e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.55, cls_loss 0.0007 cls_loss_mapping 0.0026 cls_loss_causal 0.5285 re_mapping 0.0043 re_causal 0.0142 /// teacc 99.19 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0952, 0.0586, 0.0523, ..., -0.0811, -0.1252, 0.0619], + [-0.0809, -0.0439, -0.0961, ..., 0.1655, -0.0778, -0.0683], + [-0.0250, -0.0042, 0.1874, ..., -0.1095, -0.1279, -0.1171], + ..., + [-0.1061, -0.0544, 0.0091, ..., -0.0325, 0.1047, -0.0558], + [-0.0913, -0.0899, -0.0718, ..., -0.1332, -0.0040, -0.1572], + [-0.1744, -0.0545, -0.1353, ..., -0.1523, -0.0222, 0.1107]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 0.0000e+00, -5.0664e-07, ..., 4.0978e-08, + 8.4285e-08, -6.2818e-07], + [ 4.3306e-08, 0.0000e+00, 3.0082e-07, ..., -2.0023e-08, + 8.9407e-07, 1.1316e-07], + [ 4.1444e-08, -4.6566e-10, 1.2899e-07, ..., 4.4238e-08, + 2.3236e-07, 8.8476e-08], + ..., + [ 4.9360e-08, 0.0000e+00, -1.3262e-06, ..., 4.7963e-08, + 3.5437e-07, 1.0051e-05], + [-4.3306e-08, 0.0000e+00, 4.2841e-07, ..., 8.8941e-08, + 1.5311e-06, 5.2620e-07], + [ 9.1875e-07, 0.0000e+00, 6.1514e-07, ..., 1.2107e-08, + -4.1164e-06, -1.0602e-05]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0359, -0.0301, 0.0096, 0.0324, -0.0192, 0.0037, -0.0108, 0.0127, + 0.0148, -0.0027], device='cuda:0'), grad: tensor([-4.5681e-07, 4.1425e-06, 1.4957e-06, -3.7923e-06, 5.9232e-07, + 4.4927e-06, -2.1756e-06, 9.7975e-06, 7.3574e-06, -2.1458e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.26, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4828 re_mapping 0.0044 re_causal 0.0133 /// teacc 99.18 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0955, 0.0586, 0.0522, ..., -0.0816, -0.1270, 0.0619], + [-0.0813, -0.0442, -0.0990, ..., 0.1645, -0.0780, -0.0683], + [-0.0256, -0.0041, 0.1893, ..., -0.1081, -0.1280, -0.1172], + ..., + [-0.1073, -0.0545, 0.0090, ..., -0.0330, 0.1049, -0.0571], + [-0.0915, -0.0899, -0.0718, ..., -0.1334, -0.0039, -0.1573], + [-0.1755, -0.0545, -0.1358, ..., -0.1527, -0.0224, 0.1110]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 0.0000e+00, -1.3178e-07, ..., 6.7521e-08, + 1.3970e-08, -6.5658e-08], + [-3.1479e-07, 0.0000e+00, -5.7742e-08, ..., -6.6869e-06, + 1.5041e-07, 2.1886e-08], + [ 3.0734e-08, 0.0000e+00, -2.7474e-08, ..., 1.4622e-07, + 9.5461e-08, 3.3528e-08], + ..., + [ 1.7742e-07, 0.0000e+00, -6.0629e-07, ..., 1.2387e-06, + -1.4389e-06, 6.9663e-07], + [ 2.5146e-08, 0.0000e+00, 2.9523e-07, ..., 4.7348e-06, + 5.3691e-07, 6.8452e-08], + [ 1.3085e-07, 0.0000e+00, 1.3970e-07, ..., 2.9337e-08, + 1.8999e-07, -8.8708e-07]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0361, -0.0314, 0.0112, 0.0324, -0.0189, 0.0035, -0.0107, 0.0125, + 0.0149, -0.0029], device='cuda:0'), grad: tensor([ 5.4017e-08, -1.5661e-05, 6.0257e-07, 1.1735e-07, 1.4603e-06, + 9.1363e-07, -6.5658e-08, 2.0172e-06, 1.1504e-05, -9.4017e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 214.31, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4830 re_mapping 0.0043 re_causal 0.0136 /// teacc 99.14 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0962, 0.0586, 0.0527, ..., -0.0827, -0.1276, 0.0620], + [-0.0814, -0.0445, -0.0990, ..., 0.1649, -0.0782, -0.0684], + [-0.0258, -0.0041, 0.1896, ..., -0.1082, -0.1281, -0.1174], + ..., + [-0.1078, -0.0546, 0.0089, ..., -0.0333, 0.1049, -0.0585], + [-0.0920, -0.0900, -0.0719, ..., -0.1337, -0.0039, -0.1576], + [-0.1755, -0.0546, -0.1364, ..., -0.1532, -0.0222, 0.1113]], + device='cuda:0'), grad: tensor([[-1.7369e-06, 0.0000e+00, -1.8589e-06, ..., -2.4214e-08, + 3.2596e-09, -2.2035e-06], + [ 1.0571e-07, 0.0000e+00, 1.4855e-07, ..., -1.3877e-07, + 3.1199e-08, 1.1967e-07], + [ 1.8161e-07, 0.0000e+00, -8.5216e-08, ..., 2.9337e-08, + 9.7789e-09, 1.9697e-07], + ..., + [ 5.3085e-08, 0.0000e+00, 1.0710e-07, ..., 3.2596e-08, + -9.1270e-08, 1.7742e-07], + [ 5.3085e-07, 0.0000e+00, 5.3784e-07, ..., 1.0803e-07, + 2.4680e-08, 5.2201e-07], + [ 3.5623e-07, 0.0000e+00, 3.8836e-07, ..., 1.9558e-08, + 1.3504e-08, 1.1735e-07]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0361, -0.0312, 0.0113, 0.0324, -0.0191, 0.0037, -0.0106, 0.0123, + 0.0148, -0.0027], device='cuda:0'), grad: tensor([-6.6049e-06, 2.9197e-07, 5.2946e-07, -2.3730e-06, 7.3994e-07, + 2.1160e-06, 8.9919e-07, 9.4110e-07, 2.8573e-06, 5.7835e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 214.26, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.4966 re_mapping 0.0045 re_causal 0.0135 /// teacc 99.12 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.0963, 0.0586, 0.0537, ..., -0.0829, -0.1287, 0.0622], + [-0.0814, -0.0446, -0.0992, ..., 0.1656, -0.0783, -0.0685], + [-0.0261, -0.0040, 0.1898, ..., -0.1083, -0.1282, -0.1177], + ..., + [-0.1088, -0.0546, 0.0089, ..., -0.0339, 0.1049, -0.0600], + [-0.0920, -0.0901, -0.0720, ..., -0.1338, -0.0040, -0.1578], + [-0.1757, -0.0546, -0.1370, ..., -0.1536, -0.0224, 0.1118]], + device='cuda:0'), grad: tensor([[ 6.3777e-06, 0.0000e+00, 7.3109e-08, ..., 1.7835e-07, + 9.6392e-07, 1.4901e-07], + [ 2.5798e-07, 0.0000e+00, 2.6217e-07, ..., -5.0431e-07, + 4.7125e-07, 1.4831e-07], + [ 1.4924e-07, 0.0000e+00, -7.6741e-07, ..., 2.0606e-07, + 3.8743e-07, 3.0501e-08], + ..., + [ 3.0361e-07, 2.0955e-09, -3.0012e-07, ..., 1.8394e-07, + -8.4797e-07, 7.8510e-07], + [ 1.2256e-06, 0.0000e+00, 3.4412e-07, ..., 1.7090e-06, + 2.5076e-07, 1.9907e-07], + [ 4.3702e-07, 0.0000e+00, 6.6822e-08, ..., 4.1211e-08, + -1.8561e-06, -8.9854e-06]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0358, -0.0311, 0.0112, 0.0325, -0.0193, 0.0038, -0.0107, 0.0121, + 0.0148, -0.0025], device='cuda:0'), grad: tensor([ 9.4101e-06, 2.2259e-07, -4.6100e-08, -1.6000e-06, -2.3752e-05, + 1.4640e-05, 8.2180e-06, 5.0897e-07, 5.6177e-06, -1.3247e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.29, cls_loss 0.0009 cls_loss_mapping 0.0028 cls_loss_causal 0.4815 re_mapping 0.0046 re_causal 0.0134 /// teacc 99.14 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.0967, 0.0584, 0.0538, ..., -0.0839, -0.1324, 0.0622], + [-0.0817, -0.0454, -0.0994, ..., 0.1660, -0.0786, -0.0685], + [-0.0259, -0.0035, 0.1904, ..., -0.1085, -0.1284, -0.1179], + ..., + [-0.1092, -0.0564, 0.0092, ..., -0.0341, 0.1053, -0.0610], + [-0.0923, -0.0912, -0.0723, ..., -0.1339, -0.0041, -0.1585], + [-0.1766, -0.0546, -0.1376, ..., -0.1550, -0.0233, 0.1120]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 1.3970e-09, -7.4226e-07, ..., 3.4226e-08, + 5.3551e-09, -6.4587e-07], + [ 2.0955e-09, 4.6566e-10, 7.4971e-08, ..., -4.6752e-07, + 5.4948e-08, 2.3516e-08], + [ 9.3132e-10, 2.3283e-10, 1.2270e-07, ..., 1.1246e-07, + 6.8452e-08, 6.4960e-08], + ..., + [ 1.1642e-09, 4.6566e-10, -1.9511e-07, ..., 3.5157e-08, + -2.0675e-07, 6.0536e-09], + [ 1.1642e-09, 9.3132e-10, -2.3236e-07, ..., 2.2422e-07, + 2.7940e-09, 5.7509e-08], + [ 6.2864e-09, -1.3737e-08, 5.0059e-07, ..., 6.7521e-09, + 4.4238e-08, 1.2293e-07]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0360, -0.0312, 0.0115, 0.0324, -0.0188, 0.0041, -0.0107, 0.0123, + 0.0146, -0.0028], device='cuda:0'), grad: tensor([-1.3728e-06, -6.6496e-07, 8.2981e-07, 1.6205e-06, 3.0361e-07, + 2.1118e-07, 1.8403e-06, -8.6613e-08, -3.7774e-06, 1.0896e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 214.31, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5020 re_mapping 0.0046 re_causal 0.0137 /// teacc 99.17 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0969, 0.0588, 0.0546, ..., -0.0848, -0.1326, 0.0626], + [-0.0829, -0.0459, -0.0996, ..., 0.1670, -0.0788, -0.0686], + [-0.0267, -0.0035, 0.1905, ..., -0.1089, -0.1286, -0.1180], + ..., + [-0.1097, -0.0566, 0.0093, ..., -0.0348, 0.1055, -0.0614], + [-0.0925, -0.0914, -0.0722, ..., -0.1332, -0.0041, -0.1587], + [-0.1785, -0.0547, -0.1382, ..., -0.1558, -0.0236, 0.1123]], + device='cuda:0'), grad: tensor([[ 1.6484e-07, 0.0000e+00, 4.9872e-07, ..., 1.5786e-07, + 4.1910e-09, 2.2119e-07], + [ 4.1141e-07, 0.0000e+00, 2.4680e-08, ..., -9.1502e-07, + 1.6531e-08, 2.5611e-08], + [ 1.2969e-07, 0.0000e+00, -3.6648e-07, ..., 7.1153e-07, + 1.1874e-08, 1.9325e-08], + ..., + [ 4.1211e-08, 0.0000e+00, 1.2689e-07, ..., 1.6787e-07, + -8.3819e-09, 6.9151e-08], + [ 5.2266e-06, 0.0000e+00, -4.3921e-06, ..., 3.9302e-06, + 1.1409e-07, -1.9521e-06], + [ 8.0559e-08, 0.0000e+00, 3.6452e-06, ..., 6.6357e-08, + 3.7253e-08, 1.0086e-06]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0359, -0.0308, 0.0112, 0.0325, -0.0181, 0.0036, -0.0125, 0.0121, + 0.0155, -0.0035], device='cuda:0'), grad: tensor([ 2.2259e-06, -1.6876e-06, 1.8887e-06, 1.2573e-06, 3.8520e-06, + 6.2697e-06, -2.4721e-05, 9.9093e-07, -2.1737e-06, 1.2055e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 214.58, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4835 re_mapping 0.0045 re_causal 0.0136 /// teacc 99.09 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0972, 0.0585, 0.0549, ..., -0.0860, -0.1331, 0.0625], + [-0.0836, -0.0460, -0.0998, ..., 0.1677, -0.0790, -0.0687], + [-0.0264, -0.0034, 0.1909, ..., -0.1091, -0.1288, -0.1181], + ..., + [-0.1101, -0.0567, 0.0094, ..., -0.0354, 0.1057, -0.0617], + [-0.0930, -0.0915, -0.0723, ..., -0.1333, -0.0042, -0.1588], + [-0.1793, -0.0547, -0.1390, ..., -0.1563, -0.0241, 0.1123]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 0.0000e+00, -2.8173e-08, ..., 6.7521e-09, + 3.2596e-09, -3.5390e-08], + [ 7.3342e-08, 0.0000e+00, 4.3074e-08, ..., -8.2422e-08, + 4.0978e-08, 2.0955e-09], + [ 3.9581e-08, 0.0000e+00, -7.2317e-07, ..., 2.6077e-08, + 1.1642e-08, 2.0955e-09], + ..., + [ 3.4925e-09, 0.0000e+00, 5.5647e-07, ..., 5.1223e-08, + -1.3737e-08, 4.6566e-09], + [ 1.9791e-08, 0.0000e+00, 1.7695e-08, ..., 2.9569e-08, + 2.6776e-08, 1.1176e-08], + [ 3.7719e-08, 0.0000e+00, 3.1199e-08, ..., 6.0536e-09, + 6.9849e-09, 8.3819e-09]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0361, -0.0306, 0.0112, 0.0325, -0.0179, 0.0036, -0.0123, 0.0122, + 0.0155, -0.0040], device='cuda:0'), grad: tensor([ 2.8871e-08, 5.0291e-07, -1.4035e-06, -1.2200e-06, 7.6368e-08, + 4.6683e-07, 3.7719e-07, 1.4026e-06, -4.2818e-07, 1.9465e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 214.34, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4800 re_mapping 0.0043 re_causal 0.0137 /// teacc 99.11 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0974, 0.0585, 0.0553, ..., -0.0867, -0.1337, 0.0626], + [-0.0864, -0.0463, -0.1024, ..., 0.1667, -0.0792, -0.0688], + [-0.0274, -0.0033, 0.1927, ..., -0.1079, -0.1290, -0.1183], + ..., + [-0.1119, -0.0567, 0.0093, ..., -0.0369, 0.1057, -0.0631], + [-0.0935, -0.0912, -0.0723, ..., -0.1333, -0.0042, -0.1591], + [-0.1818, -0.0548, -0.1392, ..., -0.1572, -0.0242, 0.1120]], + device='cuda:0'), grad: tensor([[ 2.7474e-08, 0.0000e+00, 2.2352e-08, ..., 2.1420e-08, + 1.0710e-08, 1.1176e-08], + [ 2.3767e-06, 0.0000e+00, 1.0896e-07, ..., 1.1837e-06, + 2.4401e-07, 1.1176e-08], + [ 4.1910e-08, 0.0000e+00, 2.4680e-08, ..., 3.2131e-08, + 2.5006e-07, 1.0245e-08], + ..., + [ 6.4448e-07, 0.0000e+00, -8.6939e-07, ..., 3.2503e-07, + -2.0582e-06, 8.2888e-08], + [ 1.4734e-06, 0.0000e+00, -5.5879e-09, ..., 7.9488e-07, + -3.2270e-07, -3.3528e-08], + [ 8.1025e-07, 0.0000e+00, 8.7544e-08, ..., 4.0513e-07, + 2.0396e-07, -1.2247e-07]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0361, -0.0326, 0.0127, 0.0326, -0.0152, 0.0036, -0.0124, 0.0117, + 0.0155, -0.0056], device='cuda:0'), grad: tensor([ 2.1793e-07, 7.8902e-06, 8.4238e-07, 3.9861e-06, -1.6257e-05, + 1.3039e-06, 1.0543e-06, -2.3656e-06, -8.3912e-07, 4.1649e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 214.56, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4951 re_mapping 0.0046 re_causal 0.0142 /// teacc 99.13 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0983, 0.0587, 0.0543, ..., -0.0879, -0.1340, 0.0627], + [-0.0838, -0.0469, -0.1026, ..., 0.1680, -0.0795, -0.0679], + [-0.0257, -0.0014, 0.1935, ..., -0.1079, -0.1290, -0.1184], + ..., + [-0.1125, -0.0569, 0.0092, ..., -0.0384, 0.1059, -0.0655], + [-0.0945, -0.0913, -0.0724, ..., -0.1334, -0.0042, -0.1593], + [-0.1819, -0.0548, -0.1393, ..., -0.1579, -0.0246, 0.1125]], + device='cuda:0'), grad: tensor([[ 1.5413e-06, 0.0000e+00, -4.6846e-07, ..., 1.0356e-06, + 9.3132e-10, -1.3504e-07], + [ 2.1327e-06, 0.0000e+00, 1.3318e-07, ..., 1.2554e-06, + 1.5460e-07, 2.3935e-07], + [ 1.7639e-06, 0.0000e+00, 1.8720e-07, ..., 2.3190e-07, + 8.4750e-08, 1.2852e-07], + ..., + [ 6.8825e-07, 0.0000e+00, -2.2212e-07, ..., -3.7253e-08, + -2.8964e-07, 1.7695e-08], + [ 2.1001e-07, 0.0000e+00, 5.1223e-08, ..., 5.0897e-07, + 5.1223e-09, 1.2014e-07], + [ 7.3155e-07, 0.0000e+00, 1.2713e-07, ..., 2.4214e-07, + 6.5193e-09, 1.1409e-07]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0366, -0.0301, 0.0131, 0.0324, -0.0175, 0.0036, -0.0121, 0.0111, + 0.0155, -0.0054], device='cuda:0'), grad: tensor([ 2.5444e-06, 4.7386e-06, 3.4012e-06, 2.7642e-06, 4.2021e-06, + 4.9882e-06, -2.3842e-05, 5.0990e-07, -8.0839e-07, 1.5488e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 214.37, cls_loss 0.0009 cls_loss_mapping 0.0028 cls_loss_causal 0.4958 re_mapping 0.0045 re_causal 0.0136 /// teacc 98.99 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0984, 0.0591, 0.0556, ..., -0.0882, -0.1345, 0.0632], + [-0.0838, -0.0475, -0.1037, ..., 0.1685, -0.0799, -0.0671], + [-0.0259, -0.0013, 0.1944, ..., -0.1076, -0.1293, -0.1187], + ..., + [-0.1133, -0.0569, 0.0093, ..., -0.0399, 0.1062, -0.0672], + [-0.0954, -0.0914, -0.0725, ..., -0.1335, -0.0042, -0.1592], + [-0.1820, -0.0549, -0.1401, ..., -0.1588, -0.0249, 0.1127]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.5053e-07, -1.4799e-06, ..., -1.4435e-08, + 9.3132e-09, -7.6275e-07], + [ 3.7253e-09, 7.9162e-09, 1.6112e-07, ..., -1.2526e-07, + 8.8010e-08, 4.1910e-08], + [ 3.7253e-09, 9.6392e-08, -1.9744e-07, ..., 3.4925e-08, + 4.4238e-08, 2.7474e-07], + ..., + [ 1.3970e-09, 7.9162e-09, -7.4506e-09, ..., 1.4435e-08, + -1.8859e-07, 8.7544e-08], + [ 2.3283e-09, 4.6566e-08, 5.4622e-07, ..., 4.9826e-08, + 4.2142e-07, 4.0280e-07], + [ 1.8626e-08, 9.3132e-09, 1.1362e-07, ..., 3.7253e-09, + 2.4214e-08, -2.5518e-07]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0362, -0.0302, 0.0137, 0.0328, -0.0175, 0.0016, -0.0114, 0.0106, + 0.0155, -0.0053], device='cuda:0'), grad: tensor([-2.1998e-06, 1.8999e-07, -3.6089e-07, 1.3374e-06, 2.5425e-07, + -2.7884e-06, 1.5497e-06, 1.8626e-08, 2.3060e-06, -3.2643e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 214.38, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.5241 re_mapping 0.0046 re_causal 0.0143 /// teacc 99.15 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.1017, 0.0561, 0.0543, ..., -0.0894, -0.1355, 0.0608], + [-0.0840, -0.0490, -0.1051, ..., 0.1671, -0.0805, -0.0673], + [-0.0250, -0.0015, 0.1956, ..., -0.1064, -0.1296, -0.1195], + ..., + [-0.1135, -0.0570, 0.0095, ..., -0.0389, 0.1065, -0.0681], + [-0.0975, -0.0946, -0.0730, ..., -0.1342, -0.0045, -0.1611], + [-0.1822, -0.0552, -0.1407, ..., -0.1600, -0.0252, 0.1130]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, -5.5879e-09, ..., 1.0571e-07, + 3.2596e-09, 8.5682e-08], + [ 9.3132e-10, 0.0000e+00, 1.4435e-08, ..., 1.8626e-08, + 2.8405e-08, 2.9802e-08], + [ 5.5879e-09, 0.0000e+00, -3.6415e-07, ..., 9.3132e-09, + -4.3306e-08, 7.9162e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.1805e-07, ..., 5.1223e-09, + 4.6566e-10, 3.2596e-09], + [-1.8859e-07, 0.0000e+00, 1.0245e-08, ..., 7.5903e-08, + 3.4459e-08, 4.5169e-08], + [ 3.2596e-09, 0.0000e+00, 4.6566e-09, ..., 1.9092e-08, + 1.1642e-08, 1.7695e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0390, -0.0315, 0.0147, 0.0353, -0.0176, 0.0043, -0.0131, 0.0110, + 0.0149, -0.0052], device='cuda:0'), grad: tensor([ 2.1793e-07, 1.7742e-07, -2.6543e-07, 1.8962e-06, 6.7987e-08, + 3.0734e-08, -2.0601e-06, 3.5297e-07, -5.4436e-07, 1.2433e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 214.37, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.5042 re_mapping 0.0041 re_causal 0.0132 /// teacc 99.09 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.1017, 0.0561, 0.0551, ..., -0.0894, -0.1357, 0.0608], + [-0.0841, -0.0493, -0.1056, ..., 0.1673, -0.0814, -0.0674], + [-0.0253, -0.0015, 0.1959, ..., -0.1065, -0.1299, -0.1198], + ..., + [-0.1143, -0.0571, 0.0097, ..., -0.0393, 0.1069, -0.0685], + [-0.0978, -0.0946, -0.0731, ..., -0.1343, -0.0044, -0.1623], + [-0.1831, -0.0554, -0.1416, ..., -0.1608, -0.0258, 0.1135]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -4.6566e-10, -1.1194e-06, ..., 1.9558e-08, + 7.1246e-08, -3.8138e-07], + [ 1.3039e-08, 0.0000e+00, 5.4529e-07, ..., -2.5425e-07, + 6.8359e-07, 4.3306e-08], + [ 1.3970e-08, 0.0000e+00, 7.9069e-07, ..., 1.6578e-07, + 9.6392e-07, 1.2014e-07], + ..., + [ 2.3283e-09, 0.0000e+00, -2.4792e-06, ..., 6.7521e-08, + -3.6079e-06, 4.7497e-08], + [ 1.7229e-08, 0.0000e+00, 2.4773e-07, ..., 6.4727e-08, + 5.9931e-07, 2.0675e-07], + [ 7.4506e-09, 0.0000e+00, 5.8906e-07, ..., 7.9162e-09, + 7.0175e-07, 7.7765e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0389, -0.0318, 0.0147, 0.0353, -0.0174, 0.0038, -0.0128, 0.0111, + 0.0149, -0.0054], device='cuda:0'), grad: tensor([-1.9819e-06, 2.2165e-06, 4.1313e-06, 8.7544e-06, 4.0093e-07, + -5.4538e-06, 3.8967e-06, -1.0677e-05, -4.3251e-06, 3.0510e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 214.20, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4760 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.02 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.1017, 0.0561, 0.0555, ..., -0.0894, -0.1360, 0.0609], + [-0.0841, -0.0495, -0.1058, ..., 0.1674, -0.0828, -0.0675], + [-0.0258, -0.0016, 0.1959, ..., -0.1067, -0.1302, -0.1201], + ..., + [-0.1136, -0.0574, 0.0096, ..., -0.0394, 0.1075, -0.0694], + [-0.0984, -0.0946, -0.0735, ..., -0.1344, -0.0045, -0.1627], + [-0.1834, -0.0555, -0.1397, ..., -0.1616, -0.0260, 0.1140]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, -8.3819e-08, ..., 1.2573e-08, + 1.3970e-09, -1.6764e-08], + [ 3.7253e-09, 0.0000e+00, 2.4214e-08, ..., -7.9162e-09, + 2.0023e-08, 2.8871e-08], + [ 7.4506e-09, 0.0000e+00, 1.3970e-09, ..., 1.3970e-08, + 2.9337e-08, 2.4214e-08], + ..., + [ 6.0536e-09, 0.0000e+00, -4.8894e-08, ..., 1.2107e-08, + -7.5437e-08, 3.3341e-07], + [ 5.8673e-08, 0.0000e+00, 1.7695e-08, ..., 5.2154e-08, + 1.6298e-08, 3.5996e-07], + [-1.7835e-07, 0.0000e+00, 3.1199e-08, ..., 1.3970e-09, + 1.1176e-08, -1.8310e-06]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0389, -0.0324, 0.0145, 0.0353, -0.0175, 0.0039, -0.0127, 0.0115, + 0.0146, -0.0040], device='cuda:0'), grad: tensor([-1.5367e-08, 1.4994e-07, 1.5181e-07, 1.6438e-07, 2.4848e-06, + 6.6217e-07, 1.1781e-07, 6.5239e-07, 8.6240e-07, -5.2415e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.45, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4836 re_mapping 0.0045 re_causal 0.0138 /// teacc 99.12 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.1017, 0.0561, 0.0566, ..., -0.0894, -0.1366, 0.0611], + [-0.0842, -0.0499, -0.1060, ..., 0.1676, -0.0830, -0.0676], + [-0.0256, -0.0017, 0.1962, ..., -0.1067, -0.1306, -0.1209], + ..., + [-0.1138, -0.0574, 0.0098, ..., -0.0398, 0.1080, -0.0700], + [-0.0988, -0.0947, -0.0737, ..., -0.1345, -0.0045, -0.1631], + [-0.1831, -0.0562, -0.1402, ..., -0.1622, -0.0266, 0.1146]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 0.0000e+00, -3.4589e-06, ..., -1.1828e-07, + 2.3283e-09, -1.8300e-06], + [-2.1560e-07, 0.0000e+00, 4.8103e-07, ..., -9.4064e-08, + 1.6764e-08, 3.4273e-07], + [ 1.1502e-07, 0.0000e+00, 1.8580e-07, ..., 8.7544e-08, + 1.1176e-08, 1.1921e-07], + ..., + [ 5.5414e-08, 0.0000e+00, 9.7789e-08, ..., 2.1141e-07, + -5.1223e-08, 5.0291e-08], + [ 6.9849e-09, 0.0000e+00, 2.4354e-07, ..., 7.3295e-07, + 2.4075e-07, 2.1141e-07], + [ 3.2131e-08, 0.0000e+00, 6.4587e-07, ..., 4.3772e-08, + 1.0245e-08, 3.0082e-07]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0388, -0.0324, 0.0145, 0.0353, -0.0177, 0.0037, -0.0128, 0.0115, + 0.0145, -0.0037], device='cuda:0'), grad: tensor([-9.3430e-06, 8.5589e-07, 9.3924e-07, 3.8147e-06, 1.6950e-06, + 4.9956e-06, -7.6443e-06, 7.7114e-07, 2.1029e-06, 1.8468e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.78, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4670 re_mapping 0.0042 re_causal 0.0130 /// teacc 99.17 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.1017, 0.0561, 0.0568, ..., -0.0894, -0.1370, 0.0612], + [-0.0842, -0.0503, -0.1062, ..., 0.1676, -0.0833, -0.0676], + [-0.0259, -0.0015, 0.1966, ..., -0.1068, -0.1307, -0.1211], + ..., + [-0.1140, -0.0574, 0.0096, ..., -0.0397, 0.1082, -0.0703], + [-0.0991, -0.0947, -0.0738, ..., -0.1346, -0.0045, -0.1632], + [-0.1834, -0.0564, -0.1405, ..., -0.1630, -0.0271, 0.1147]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 9.3132e-10, -1.2107e-07, ..., 9.3132e-10, + 4.4703e-08, -3.6694e-07], + [ 1.7229e-08, 2.3283e-09, 6.1467e-07, ..., -3.2596e-09, + 5.1968e-07, 1.1642e-08], + [ 8.8476e-09, 9.3132e-10, 3.7579e-07, ..., 6.0536e-09, + 3.8277e-07, 8.3819e-09], + ..., + [ 8.3819e-09, 4.6566e-09, -1.7621e-06, ..., 1.3970e-09, + -1.5479e-06, 3.7253e-09], + [ 1.0710e-08, 2.7940e-09, 6.5193e-08, ..., 1.3039e-08, + 2.7474e-08, 2.6543e-08], + [ 1.1800e-06, 5.1223e-09, 3.1805e-07, ..., 1.3970e-09, + 2.7753e-07, 1.0710e-08]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0388, -0.0325, 0.0145, 0.0353, -0.0177, 0.0037, -0.0127, 0.0115, + 0.0145, -0.0038], device='cuda:0'), grad: tensor([-6.6776e-07, 2.1588e-06, 1.4286e-06, 5.0152e-07, -2.5965e-06, + 3.7486e-07, 6.1560e-07, -5.7369e-06, 2.0396e-07, 3.7365e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 214.59, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5152 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.07 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.1017, 0.0562, 0.0587, ..., -0.0895, -0.1371, 0.0615], + [-0.0843, -0.0510, -0.1064, ..., 0.1676, -0.0837, -0.0678], + [-0.0266, -0.0019, 0.1970, ..., -0.1069, -0.1316, -0.1234], + ..., + [-0.1147, -0.0565, 0.0095, ..., -0.0399, 0.1085, -0.0708], + [-0.0996, -0.0948, -0.0742, ..., -0.1347, -0.0046, -0.1634], + [-0.1841, -0.0590, -0.1423, ..., -0.1637, -0.0276, 0.1144]], + device='cuda:0'), grad: tensor([[ 5.9232e-07, 2.1886e-08, 6.6590e-08, ..., 8.1398e-07, + 1.3178e-07, 8.8476e-09], + [ 3.7719e-08, 2.3283e-09, 1.1828e-07, ..., -8.8941e-08, + 1.1642e-07, 5.1223e-09], + [ 1.7975e-07, 1.8626e-08, 1.9502e-06, ..., 3.1898e-07, + 1.6000e-06, 8.8476e-09], + ..., + [ 2.3283e-09, 4.8662e-07, -9.2806e-07, ..., 4.4238e-08, + 4.7125e-07, 1.3039e-08], + [ 2.3749e-08, 5.8208e-08, 2.2724e-07, ..., 3.7719e-08, + 3.3434e-07, 1.4435e-08], + [ 1.2573e-08, 6.0536e-09, 6.8452e-08, ..., 1.2107e-08, + 9.7323e-08, 2.0489e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0386, -0.0327, 0.0145, 0.0352, -0.0179, 0.0042, -0.0120, 0.0116, + 0.0144, -0.0043], device='cuda:0'), grad: tensor([ 3.6284e-06, 2.2957e-07, 4.7050e-06, -7.1563e-06, 2.2305e-07, + -4.7544e-07, -6.8806e-06, 4.1351e-06, 1.2172e-06, 3.4878e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 214.23, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.5221 re_mapping 0.0048 re_causal 0.0148 /// teacc 99.16 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.1018, 0.0562, 0.0590, ..., -0.0895, -0.1401, 0.0615], + [-0.0843, -0.0510, -0.1066, ..., 0.1671, -0.0840, -0.0680], + [-0.0268, -0.0019, 0.1957, ..., -0.1071, -0.1325, -0.1242], + ..., + [-0.1149, -0.0567, 0.0101, ..., -0.0399, 0.1090, -0.0711], + [-0.0995, -0.0948, -0.0721, ..., -0.1349, -0.0046, -0.1635], + [-0.1836, -0.0592, -0.1435, ..., -0.1649, -0.0283, 0.1153]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.6566e-10, -1.1856e-06, ..., -7.1386e-07, + 4.1910e-09, -4.6045e-06], + [ 9.3132e-10, 0.0000e+00, 1.1958e-06, ..., 6.4308e-07, + 1.4482e-07, 3.2131e-06], + [ 0.0000e+00, 0.0000e+00, -3.2131e-08, ..., 6.8918e-08, + 2.7008e-08, 1.7229e-08], + ..., + [ 4.6566e-10, 0.0000e+00, -5.7835e-07, ..., -3.8696e-07, + -3.5902e-07, 9.9558e-07], + [ 4.6566e-10, 0.0000e+00, 1.0896e-07, ..., 9.3598e-08, + 3.1199e-08, 1.3551e-07], + [ 3.7253e-09, 0.0000e+00, 4.3167e-07, ..., 2.9011e-07, + 1.4715e-07, 1.4948e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0386, -0.0332, 0.0124, 0.0351, -0.0186, 0.0047, -0.0110, 0.0119, + 0.0157, -0.0042], device='cuda:0'), grad: tensor([-8.4266e-06, 1.0990e-05, 9.6299e-07, -3.8184e-08, 3.6508e-07, + -8.4750e-08, -5.0291e-08, -1.0930e-05, 1.1008e-06, 6.1169e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 214.57, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.5058 re_mapping 0.0042 re_causal 0.0136 /// teacc 99.11 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.1019, 0.0562, 0.0592, ..., -0.0896, -0.1409, 0.0616], + [-0.0843, -0.0512, -0.1069, ..., 0.1673, -0.0848, -0.0683], + [-0.0273, -0.0016, 0.1958, ..., -0.1075, -0.1330, -0.1243], + ..., + [-0.1150, -0.0569, 0.0108, ..., -0.0393, 0.1095, -0.0716], + [-0.0998, -0.0948, -0.0723, ..., -0.1357, -0.0047, -0.1655], + [-0.1837, -0.0592, -0.1441, ..., -0.1657, -0.0288, 0.1157]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.3271e-07, ..., 5.1223e-09, + 2.7474e-08, -7.2643e-08], + [ 1.2107e-08, 0.0000e+00, 5.6345e-08, ..., -3.1199e-07, + 5.4017e-08, 2.1420e-08], + [ 5.1223e-09, 0.0000e+00, -6.6590e-08, ..., 2.3283e-08, + 2.4214e-08, 2.2817e-08], + ..., + [ 9.3132e-09, 0.0000e+00, -8.0047e-07, ..., 7.6368e-08, + -7.9302e-07, 3.8184e-08], + [ 4.1910e-09, 0.0000e+00, 6.5193e-08, ..., 9.8255e-08, + 9.4064e-08, 1.3364e-07], + [ 8.0559e-08, 0.0000e+00, 7.3109e-07, ..., 3.6322e-08, + 8.7405e-07, 3.0454e-07]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0386, -0.0334, 0.0121, 0.0351, -0.0188, 0.0058, -0.0114, 0.0125, + 0.0155, -0.0042], device='cuda:0'), grad: tensor([-1.7090e-07, -4.8988e-07, 3.4459e-08, 7.1153e-07, 7.1712e-08, + -3.0249e-06, 9.5926e-07, -1.6838e-06, 7.2876e-07, 2.8685e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.52, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4823 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.09 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.1020, 0.0562, 0.0592, ..., -0.0896, -0.1426, 0.0615], + [-0.0844, -0.0513, -0.1069, ..., 0.1676, -0.0851, -0.0683], + [-0.0282, -0.0025, 0.1958, ..., -0.1077, -0.1334, -0.1245], + ..., + [-0.1152, -0.0576, 0.0111, ..., -0.0398, 0.1098, -0.0720], + [-0.0990, -0.0948, -0.0723, ..., -0.1358, -0.0047, -0.1657], + [-0.1834, -0.0593, -0.1443, ..., -0.1667, -0.0292, 0.1163]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8217e-06, ..., 2.7940e-09, + -4.6566e-08, -1.1399e-06], + [ 2.7940e-09, 0.0000e+00, 2.7008e-08, ..., -6.9849e-08, + 9.3132e-09, 1.2107e-08], + [ 1.8626e-09, 0.0000e+00, 4.2189e-07, ..., 1.5832e-08, + -2.1420e-08, 3.5204e-07], + ..., + [ 2.7940e-09, 0.0000e+00, 2.3656e-07, ..., 2.5146e-08, + 2.4214e-08, 8.1956e-08], + [ 2.7940e-09, 0.0000e+00, 1.1735e-07, ..., 2.1420e-08, + 9.3132e-09, 6.7055e-08], + [ 1.9558e-08, 0.0000e+00, 6.5099e-07, ..., 6.5193e-09, + 1.9558e-08, 3.7905e-07]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0387, -0.0333, 0.0118, 0.0351, -0.0192, 0.0059, -0.0113, 0.0126, + 0.0157, -0.0038], device='cuda:0'), grad: tensor([-3.1237e-06, -6.9849e-08, 8.9873e-07, -1.2433e-06, 4.9360e-08, + 1.4119e-06, 2.3283e-08, 5.0664e-07, 4.2468e-07, 1.1064e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 214.81, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.5118 re_mapping 0.0043 re_causal 0.0134 /// teacc 99.01 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.1021, 0.0562, 0.0594, ..., -0.0896, -0.1429, 0.0616], + [-0.0844, -0.0513, -0.1070, ..., 0.1678, -0.0853, -0.0684], + [-0.0293, -0.0025, 0.1960, ..., -0.1078, -0.1336, -0.1246], + ..., + [-0.1157, -0.0576, 0.0110, ..., -0.0399, 0.1100, -0.0726], + [-0.0975, -0.0948, -0.0724, ..., -0.1360, -0.0047, -0.1659], + [-0.1843, -0.0593, -0.1445, ..., -0.1673, -0.0295, 0.1166]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -6.0257e-07, ..., 2.5146e-08, + 9.3132e-10, -2.1886e-07], + [ 4.6566e-09, 0.0000e+00, 5.1223e-08, ..., -2.9802e-08, + 2.6077e-08, 1.8626e-09], + [ 2.7940e-09, 0.0000e+00, 2.3935e-07, ..., 2.8871e-08, + 2.6077e-08, 5.6811e-08], + ..., + [ 5.5879e-09, 0.0000e+00, -1.4994e-07, ..., 1.4901e-08, + -8.0094e-08, 1.8626e-09], + [ 2.7940e-09, 0.0000e+00, 2.0582e-07, ..., 1.0245e-08, + 9.3132e-09, 6.9849e-08], + [ 8.1025e-08, 0.0000e+00, 9.4995e-08, ..., 9.3132e-10, + 1.9558e-08, 3.1665e-08]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0388, -0.0333, 0.0117, 0.0351, -0.0189, 0.0055, -0.0111, 0.0125, + 0.0159, -0.0041], device='cuda:0'), grad: tensor([-9.7603e-07, 9.4064e-08, 4.7591e-07, 3.7067e-07, -8.4750e-08, + 6.6124e-08, 9.8441e-07, -3.3528e-07, -9.0431e-07, 3.0827e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 214.81, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4934 re_mapping 0.0043 re_causal 0.0132 /// teacc 99.05 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.1021, 0.0562, 0.0606, ..., -0.0896, -0.1435, 0.0620], + [-0.0844, -0.0513, -0.1071, ..., 0.1681, -0.0857, -0.0685], + [-0.0295, -0.0025, 0.1961, ..., -0.1080, -0.1338, -0.1251], + ..., + [-0.1182, -0.0577, 0.0110, ..., -0.0402, 0.1102, -0.0736], + [-0.0978, -0.0948, -0.0725, ..., -0.1361, -0.0048, -0.1660], + [-0.1841, -0.0593, -0.1458, ..., -0.1679, -0.0297, 0.1168]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -9.0059e-07, ..., 0.0000e+00, + 6.7987e-08, -7.5530e-07], + [ 1.5832e-08, 0.0000e+00, 5.1223e-08, ..., -4.8429e-08, + 7.9162e-08, 6.0536e-08], + [ 5.5879e-09, 0.0000e+00, -7.9162e-07, ..., 8.3819e-09, + 8.0094e-08, 8.2888e-08], + ..., + [ 1.1176e-08, 0.0000e+00, 6.2399e-08, ..., 2.6077e-08, + -3.7160e-07, 9.2201e-08], + [ 8.3819e-09, 0.0000e+00, 4.5262e-07, ..., 5.5879e-09, + 1.7136e-07, 3.4552e-07], + [ 6.0443e-07, 0.0000e+00, 3.6787e-07, ..., 3.7253e-09, + 8.0839e-07, 1.5879e-06]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0385, -0.0332, 0.0115, 0.0351, -0.0192, 0.0056, -0.0114, 0.0125, + 0.0158, -0.0040], device='cuda:0'), grad: tensor([-2.4866e-07, 1.0282e-06, 6.5938e-07, -3.2298e-06, -7.3016e-07, + -7.4357e-06, 3.5111e-06, -6.8173e-07, 1.0496e-06, 6.0983e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 214.70, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5117 re_mapping 0.0043 re_causal 0.0135 /// teacc 98.85 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.1021, 0.0562, 0.0612, ..., -0.0896, -0.1443, 0.0621], + [-0.0845, -0.0514, -0.1072, ..., 0.1685, -0.0861, -0.0686], + [-0.0296, -0.0025, 0.1958, ..., -0.1081, -0.1356, -0.1253], + ..., + [-0.1194, -0.0578, 0.0120, ..., -0.0405, 0.1110, -0.0738], + [-0.0982, -0.0948, -0.0727, ..., -0.1363, -0.0048, -0.1661], + [-0.1867, -0.0595, -0.1463, ..., -0.1688, -0.0304, 0.1169]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -3.6415e-07, ..., -6.5193e-09, + 9.3132e-09, -1.7323e-07], + [ 1.2107e-08, 0.0000e+00, 9.1363e-07, ..., -8.6706e-07, + 1.2480e-06, 1.5832e-08], + [ 1.4901e-08, 0.0000e+00, 2.7381e-07, ..., 5.0291e-08, + 3.2969e-07, 1.3970e-08], + ..., + [ 8.3819e-09, 0.0000e+00, -1.6410e-06, ..., 7.0315e-07, + -2.2948e-06, 2.2352e-08], + [-4.6566e-09, 0.0000e+00, 1.6298e-07, ..., 4.9360e-08, + 2.0955e-07, 1.0245e-07], + [ 2.4028e-07, 0.0000e+00, 1.1642e-07, ..., 1.8626e-08, + 4.5635e-08, 5.9605e-08]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0385, -0.0331, 0.0110, 0.0351, -0.0170, 0.0053, -0.0114, 0.0131, + 0.0159, -0.0065], device='cuda:0'), grad: tensor([-2.2631e-07, 2.1271e-06, 1.9334e-06, -3.7868e-06, 1.0664e-06, + -2.5053e-07, 3.3248e-07, -3.9451e-06, 1.8440e-06, 8.9128e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 214.85, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.5150 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.18 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.1021, 0.0562, 0.0615, ..., -0.0896, -0.1452, 0.0621], + [-0.0844, -0.0514, -0.1075, ..., 0.1689, -0.0864, -0.0688], + [-0.0298, -0.0024, 0.1962, ..., -0.1082, -0.1358, -0.1254], + ..., + [-0.1198, -0.0579, 0.0121, ..., -0.0411, 0.1113, -0.0740], + [-0.0994, -0.0948, -0.0729, ..., -0.1365, -0.0049, -0.1680], + [-0.1875, -0.0595, -0.1464, ..., -0.1697, -0.0307, 0.1184]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, -2.9057e-07, ..., 5.7742e-08, + 2.7940e-09, -1.5274e-07], + [ 1.0617e-07, 0.0000e+00, 9.3132e-09, ..., -1.6335e-06, + 9.3132e-09, 1.3039e-08], + [ 4.0047e-08, 0.0000e+00, 8.3819e-09, ..., 1.4529e-07, + 9.3132e-09, 1.3039e-08], + ..., + [ 2.0582e-07, 0.0000e+00, -1.0245e-08, ..., 7.6368e-07, + -3.1665e-08, 1.8626e-08], + [ 3.0734e-08, 0.0000e+00, 1.0245e-08, ..., 1.8533e-07, + 3.5390e-08, 6.5193e-08], + [ 1.7220e-06, 0.0000e+00, 2.9802e-08, ..., 5.9605e-08, + 1.0245e-08, -4.1816e-07]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0385, -0.0330, 0.0110, 0.0351, -0.0164, 0.0057, -0.0117, 0.0131, + 0.0155, -0.0070], device='cuda:0'), grad: tensor([-3.4831e-07, -4.1202e-06, 5.1409e-07, 1.0664e-06, -2.3432e-06, + -3.7160e-07, 3.2224e-07, 2.4289e-06, 6.2864e-07, 2.2110e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 214.36, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.5120 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.15 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.1021, 0.0562, 0.0623, ..., -0.0896, -0.1462, 0.0622], + [-0.0845, -0.0514, -0.1077, ..., 0.1693, -0.0868, -0.0690], + [-0.0299, -0.0024, 0.1966, ..., -0.1082, -0.1357, -0.1257], + ..., + [-0.1203, -0.0579, 0.0119, ..., -0.0418, 0.1116, -0.0745], + [-0.0997, -0.0948, -0.0730, ..., -0.1367, -0.0048, -0.1682], + [-0.1875, -0.0595, -0.1467, ..., -0.1705, -0.0312, 0.1188]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, -2.0768e-07, ..., 9.3132e-10, + 5.5879e-09, -1.7136e-07], + [ 9.4995e-08, 0.0000e+00, 4.4703e-08, ..., -1.8626e-08, + 1.9651e-07, 5.9605e-08], + [ 5.5879e-08, 0.0000e+00, -1.1269e-07, ..., 5.5879e-09, + 1.3970e-08, 2.3283e-08], + ..., + [ 4.3772e-08, 0.0000e+00, -1.7509e-07, ..., 5.5879e-09, + -3.5483e-07, 1.3039e-08], + [ 5.9605e-08, 0.0000e+00, 4.0978e-08, ..., 1.8626e-08, + 9.7789e-08, 9.4064e-08], + [ 3.8221e-06, 0.0000e+00, 2.9244e-07, ..., 2.7940e-09, + 3.3062e-07, 9.2387e-07]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0384, -0.0329, 0.0111, 0.0350, -0.0164, 0.0051, -0.0112, 0.0131, + 0.0156, -0.0070], device='cuda:0'), grad: tensor([-3.6601e-07, 9.5367e-07, 4.2841e-08, 5.2564e-06, -9.6187e-06, + -1.3202e-05, 5.3160e-06, -7.4692e-07, 5.8487e-07, 1.1802e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 214.12, cls_loss 0.0010 cls_loss_mapping 0.0047 cls_loss_causal 0.4836 re_mapping 0.0043 re_causal 0.0131 /// teacc 99.07 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.1022, 0.0562, 0.0627, ..., -0.0896, -0.1503, 0.0623], + [-0.0849, -0.0514, -0.1083, ..., 0.1693, -0.0873, -0.0692], + [-0.0275, -0.0023, 0.1974, ..., -0.1080, -0.1358, -0.1259], + ..., + [-0.1207, -0.0579, 0.0140, ..., -0.0420, 0.1147, -0.0775], + [-0.1003, -0.0949, -0.0733, ..., -0.1369, -0.0047, -0.1692], + [-0.1875, -0.0596, -0.1498, ..., -0.1716, -0.0342, 0.1209]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -9.3132e-10, -2.8871e-08, ..., 1.5832e-08, + 2.7940e-09, -5.4017e-07], + [ 2.0489e-08, 0.0000e+00, 9.3132e-10, ..., -2.6077e-07, + 3.7253e-09, 2.7008e-08], + [ 7.0781e-08, 0.0000e+00, 3.7253e-09, ..., 1.8626e-08, + 9.3132e-09, 2.0489e-08], + ..., + [ 3.7253e-08, 0.0000e+00, -5.5879e-09, ..., 3.8184e-08, + -1.0245e-08, 1.3411e-07], + [-3.3528e-08, 0.0000e+00, 3.7253e-09, ..., 8.4750e-08, + -1.1921e-07, 4.1910e-08], + [ 9.2201e-08, 0.0000e+00, 6.5193e-09, ..., 1.3970e-08, + 7.4506e-09, -1.4715e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0384, -0.0333, 0.0117, 0.0348, -0.0167, 0.0047, -0.0110, 0.0155, + 0.0154, -0.0078], device='cuda:0'), grad: tensor([-2.0768e-06, -4.4517e-07, 2.5705e-07, 1.4994e-06, 1.1362e-07, + 3.7998e-07, 9.9093e-07, 5.9791e-07, -1.3616e-06, 4.0047e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 214.12, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.5051 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.06 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.1022, 0.0562, 0.0632, ..., -0.0897, -0.1508, 0.0624], + [-0.0850, -0.0515, -0.1088, ..., 0.1692, -0.0900, -0.0693], + [-0.0276, -0.0020, 0.1977, ..., -0.1080, -0.1365, -0.1265], + ..., + [-0.1215, -0.0580, 0.0146, ..., -0.0412, 0.1156, -0.0785], + [-0.1006, -0.0949, -0.0735, ..., -0.1371, -0.0047, -0.1695], + [-0.1876, -0.0596, -0.1499, ..., -0.1720, -0.0344, 0.1216]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 3.7253e-09, 7.4506e-08, ..., 6.1467e-08, + 3.7253e-09, 0.0000e+00], + [ 1.1362e-07, 9.3132e-09, 1.7881e-07, ..., -1.7378e-06, + 3.1665e-08, 0.0000e+00], + [ 4.6566e-08, -1.1176e-07, -1.9297e-06, ..., 1.9651e-07, + 5.2154e-08, 0.0000e+00], + ..., + [ 9.2201e-08, 4.8429e-08, 7.1432e-07, ..., 2.1700e-07, + -1.1921e-07, 0.0000e+00], + [ 2.7008e-08, 2.3283e-08, 4.1723e-07, ..., 8.1863e-07, + -1.8440e-07, 0.0000e+00], + [ 2.9262e-06, 9.3132e-10, 2.7008e-08, ..., 3.6322e-08, + 5.1223e-08, 0.0000e+00]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0384, -0.0342, 0.0117, 0.0348, -0.0166, 0.0048, -0.0112, 0.0162, + 0.0153, -0.0078], device='cuda:0'), grad: tensor([ 3.1386e-07, -3.3900e-06, -5.2527e-07, 2.7698e-06, -7.3761e-06, + 2.6077e-08, 1.3104e-06, 1.6307e-06, -2.0899e-06, 7.3239e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.43, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4996 re_mapping 0.0044 re_causal 0.0132 /// teacc 99.11 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.1022, 0.0562, 0.0634, ..., -0.0897, -0.1519, 0.0625], + [-0.0849, -0.0518, -0.1090, ..., 0.1696, -0.0904, -0.0694], + [-0.0277, -0.0022, 0.1981, ..., -0.1080, -0.1369, -0.1269], + ..., + [-0.1216, -0.0568, 0.0145, ..., -0.0415, 0.1158, -0.0788], + [-0.1008, -0.0949, -0.0735, ..., -0.1372, -0.0045, -0.1697], + [-0.1876, -0.0596, -0.1499, ..., -0.1727, -0.0345, 0.1222]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, 0.0000e+00, -9.6709e-06, ..., 1.9558e-08, + 1.2107e-08, -5.3234e-06], + [ 2.6077e-08, 0.0000e+00, 4.3120e-07, ..., -5.8673e-08, + 1.9278e-07, 1.2107e-08], + [ 4.1910e-08, 0.0000e+00, 3.2019e-06, ..., 4.2841e-08, + 1.7127e-06, 2.8312e-07], + ..., + [ 2.0955e-07, 0.0000e+00, -3.8557e-06, ..., 5.4948e-08, + -2.3469e-06, 1.2107e-08], + [-2.0489e-08, 0.0000e+00, 7.7114e-07, ..., 3.6322e-08, + 2.9523e-07, 3.2969e-07], + [ 3.4086e-07, 0.0000e+00, 2.9616e-07, ..., 2.9802e-08, + 5.2154e-08, 1.4063e-07]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0384, -0.0341, 0.0118, 0.0348, -0.0167, 0.0046, -0.0113, 0.0162, + 0.0155, -0.0077], device='cuda:0'), grad: tensor([-2.2605e-05, 1.2415e-06, 7.6257e-06, 2.0966e-05, -7.0594e-07, + -2.3749e-07, -1.1921e-07, -9.0674e-06, 1.5954e-06, 1.2731e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 214.35, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5066 re_mapping 0.0042 re_causal 0.0133 /// teacc 98.99 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.1022, 0.0562, 0.0637, ..., -0.0897, -0.1531, 0.0626], + [-0.0849, -0.0518, -0.1092, ..., 0.1701, -0.0905, -0.0697], + [-0.0278, -0.0023, 0.1987, ..., -0.1082, -0.1371, -0.1271], + ..., + [-0.1220, -0.0573, 0.0143, ..., -0.0421, 0.1158, -0.0793], + [-0.1013, -0.0949, -0.0737, ..., -0.1376, -0.0044, -0.1699], + [-0.1876, -0.0596, -0.1500, ..., -0.1734, -0.0345, 0.1230]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 1.0245e-08, ..., 4.2096e-07, + 9.3132e-09, 3.5390e-08], + [ 4.6566e-09, 0.0000e+00, 8.3819e-09, ..., -4.2059e-06, + 2.6077e-08, 1.8626e-08], + [ 4.6566e-09, 0.0000e+00, -2.0303e-07, ..., 3.7439e-07, + 1.9558e-08, 1.4901e-08], + ..., + [ 3.3528e-08, 0.0000e+00, 3.7253e-09, ..., 1.8384e-06, + -8.9407e-08, 6.4261e-08], + [ 3.7253e-09, 0.0000e+00, 1.4622e-07, ..., 2.5518e-07, + 6.1467e-08, 5.3085e-08], + [ 1.3970e-08, 0.0000e+00, 4.6566e-09, ..., 2.5891e-07, + 5.0291e-08, -1.0338e-07]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0384, -0.0339, 0.0119, 0.0348, -0.0168, 0.0044, -0.0114, 0.0159, + 0.0155, -0.0076], device='cuda:0'), grad: tensor([ 1.0682e-06, -8.9481e-06, 6.8452e-07, 1.5693e-06, 3.2783e-07, + -6.9849e-07, 3.4831e-07, 4.4927e-06, 7.8417e-07, 3.5111e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 214.39, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.5101 re_mapping 0.0043 re_causal 0.0130 /// teacc 99.08 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.1023, 0.0562, 0.0640, ..., -0.0897, -0.1541, 0.0626], + [-0.0850, -0.0519, -0.1093, ..., 0.1706, -0.0906, -0.0700], + [-0.0279, -0.0021, 0.1991, ..., -0.1082, -0.1380, -0.1272], + ..., + [-0.1224, -0.0575, 0.0143, ..., -0.0428, 0.1160, -0.0803], + [-0.1017, -0.0949, -0.0740, ..., -0.1379, -0.0045, -0.1707], + [-0.1877, -0.0596, -0.1501, ..., -0.1742, -0.0345, 0.1239]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 0.0000e+00, 2.5146e-08, ..., 9.3132e-09, + 1.5832e-08, -1.3039e-08], + [ 4.8429e-08, 0.0000e+00, 2.6170e-07, ..., -5.5879e-09, + 3.8184e-08, 2.7940e-09], + [ 4.7497e-08, 0.0000e+00, -2.7530e-06, ..., 2.0489e-08, + 2.2352e-08, 1.8626e-09], + ..., + [ 5.2154e-08, 0.0000e+00, 1.7323e-07, ..., 1.3970e-08, + -1.6764e-07, 7.4506e-09], + [ 8.6613e-08, 0.0000e+00, 2.1253e-06, ..., 5.2154e-08, + 1.5087e-07, 7.4506e-09], + [ 4.1258e-07, 0.0000e+00, 2.1420e-08, ..., 2.7940e-09, + 9.4064e-08, 3.2596e-08]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0384, -0.0337, 0.0120, 0.0349, -0.0169, 0.0046, -0.0118, 0.0157, + 0.0152, -0.0075], device='cuda:0'), grad: tensor([ 1.0990e-07, 6.0443e-07, -4.7907e-06, 4.1630e-07, -9.2667e-07, + -1.1921e-07, -3.9861e-07, 3.2037e-07, 3.9414e-06, 8.2795e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 214.36, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4950 re_mapping 0.0041 re_causal 0.0126 /// teacc 99.10 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.1023, 0.0562, 0.0646, ..., -0.0897, -0.1551, 0.0629], + [-0.0853, -0.0520, -0.1100, ..., 0.1710, -0.0907, -0.0700], + [-0.0257, -0.0021, 0.2001, ..., -0.1082, -0.1381, -0.1275], + ..., + [-0.1234, -0.0576, 0.0143, ..., -0.0434, 0.1162, -0.0811], + [-0.1028, -0.0949, -0.0744, ..., -0.1380, -0.0046, -0.1709], + [-0.1877, -0.0596, -0.1502, ..., -0.1749, -0.0345, 0.1241]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 0.0000e+00, -3.6638e-06, ..., 9.3132e-09, + 1.3039e-08, -2.0117e-06], + [ 8.3819e-09, 0.0000e+00, 3.2689e-07, ..., -1.5274e-07, + 6.0536e-08, 3.0734e-08], + [ 4.6566e-09, 0.0000e+00, 8.9128e-07, ..., 3.4459e-08, + 1.8626e-08, 8.8103e-07], + ..., + [ 4.6566e-09, 0.0000e+00, 8.7544e-08, ..., 2.7008e-08, + -1.1921e-07, 2.2352e-08], + [ 9.3132e-09, 0.0000e+00, 4.5728e-07, ..., 5.2154e-08, + 2.6729e-07, 2.0023e-07], + [ 5.6811e-08, 0.0000e+00, 4.6287e-07, ..., 5.5879e-09, + 3.1106e-07, 6.7707e-07]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0383, -0.0340, 0.0127, 0.0349, -0.0168, 0.0045, -0.0120, 0.0156, + 0.0151, -0.0075], device='cuda:0'), grad: tensor([-6.5789e-06, 4.2934e-07, 1.9185e-06, 1.4380e-06, 1.4901e-08, + -2.6263e-06, 1.2936e-06, 3.2596e-08, 2.0154e-06, 2.0452e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 214.51, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4754 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.15 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.1023, 0.0562, 0.0646, ..., -0.0898, -0.1558, 0.0625], + [-0.0854, -0.0520, -0.1102, ..., 0.1713, -0.0912, -0.0701], + [-0.0255, -0.0020, 0.1999, ..., -0.1084, -0.1397, -0.1277], + ..., + [-0.1238, -0.0576, 0.0153, ..., -0.0436, 0.1168, -0.0816], + [-0.1027, -0.0949, -0.0746, ..., -0.1381, -0.0046, -0.1710], + [-0.1877, -0.0596, -0.1502, ..., -0.1753, -0.0347, 0.1257]], + device='cuda:0'), grad: tensor([[ 6.0257e-07, 0.0000e+00, -1.3784e-07, ..., 6.5193e-09, + 2.7008e-08, -5.9605e-08], + [ 1.5274e-07, 9.3132e-10, 1.6764e-08, ..., -3.0175e-07, + 3.8184e-08, 1.7695e-08], + [ 8.9128e-07, 0.0000e+00, -3.5390e-08, ..., 3.7253e-08, + 3.0734e-08, 3.9116e-08], + ..., + [ 8.0094e-08, 9.3132e-10, 0.0000e+00, ..., 9.1270e-08, + -3.6322e-08, 1.6019e-07], + [-2.7210e-05, 1.4901e-08, 2.0489e-08, ..., 8.8476e-08, + 1.4529e-07, -7.8045e-07], + [ 1.1008e-06, 0.0000e+00, 3.4459e-08, ..., 1.1176e-08, + 6.0536e-08, -4.9639e-07]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0384, -0.0341, 0.0122, 0.0349, -0.0170, 0.0046, -0.0119, 0.0163, + 0.0151, -0.0074], device='cuda:0'), grad: tensor([ 2.0154e-06, -5.4203e-07, 3.4459e-06, 1.0771e-04, 1.9744e-06, + -2.2650e-05, 5.2992e-07, 1.4016e-06, -9.5725e-05, 1.7975e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 214.14, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.5012 re_mapping 0.0040 re_causal 0.0127 /// teacc 99.10 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.1024, 0.0562, 0.0655, ..., -0.0897, -0.1573, 0.0626], + [-0.0855, -0.0520, -0.1105, ..., 0.1716, -0.0914, -0.0704], + [-0.0257, -0.0019, 0.2001, ..., -0.1086, -0.1400, -0.1280], + ..., + [-0.1245, -0.0576, 0.0154, ..., -0.0435, 0.1173, -0.0827], + [-0.1016, -0.0949, -0.0747, ..., -0.1384, -0.0047, -0.1713], + [-0.1877, -0.0597, -0.1503, ..., -0.1760, -0.0347, 0.1280]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, -9.3132e-10, -1.1176e-08, ..., 8.3819e-08, + 9.3132e-10, -6.5193e-09], + [ 1.8626e-09, 0.0000e+00, 7.1712e-08, ..., 2.5146e-08, + 3.0734e-08, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -1.3411e-07, ..., 2.1420e-08, + 5.5879e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 2.2352e-08, + -7.4506e-08, 1.6764e-08], + [ 1.6764e-08, 0.0000e+00, 6.5193e-09, ..., 1.2107e-07, + 3.7253e-09, 6.5193e-09], + [ 9.3132e-10, 0.0000e+00, 1.5832e-08, ..., 8.3819e-09, + 2.1420e-08, -4.8429e-08]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0383, -0.0343, 0.0119, 0.0348, -0.0177, 0.0046, -0.0119, 0.0169, + 0.0151, -0.0068], device='cuda:0'), grad: tensor([ 5.1036e-07, 4.3772e-07, -8.1956e-08, -2.4475e-06, 1.2945e-07, + 8.3819e-07, -2.7753e-07, 7.8231e-08, 7.6927e-07, 3.8184e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 214.35, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4701 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.14 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.1024, 0.0562, 0.0663, ..., -0.0897, -0.1578, 0.0629], + [-0.0855, -0.0521, -0.1106, ..., 0.1718, -0.0916, -0.0705], + [-0.0259, -0.0018, 0.2002, ..., -0.1087, -0.1403, -0.1282], + ..., + [-0.1244, -0.0576, 0.0155, ..., -0.0437, 0.1170, -0.0856], + [-0.1022, -0.0949, -0.0749, ..., -0.1385, -0.0048, -0.1716], + [-0.1877, -0.0597, -0.1505, ..., -0.1766, -0.0341, 0.1299]], + device='cuda:0'), grad: tensor([[ 5.0291e-08, 0.0000e+00, 2.4214e-08, ..., 8.9407e-08, + 8.3819e-08, 1.0245e-07], + [ 8.9407e-08, 0.0000e+00, 4.8243e-07, ..., -5.8115e-07, + 4.1723e-07, 4.0978e-08], + [ 1.0803e-07, 0.0000e+00, 1.4566e-06, ..., 1.1362e-07, + 1.0785e-06, 9.1270e-08], + ..., + [ 5.7742e-08, 0.0000e+00, -4.1611e-06, ..., 1.6764e-07, + -4.3772e-06, 9.3132e-09], + [ 1.2293e-07, 0.0000e+00, 5.6438e-07, ..., 3.3528e-07, + 4.3213e-07, 2.0862e-07], + [ 5.2899e-07, 0.0000e+00, 1.0189e-06, ..., 5.2154e-08, + 1.5795e-06, 2.0303e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0382, -0.0342, 0.0118, 0.0349, -0.0178, 0.0043, -0.0118, 0.0159, + 0.0152, -0.0064], device='cuda:0'), grad: tensor([ 1.0896e-06, -1.6019e-07, 4.2990e-06, 1.0114e-06, -8.4937e-07, + 1.8775e-05, -2.1979e-05, -1.1154e-05, 3.3677e-06, 5.5581e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 214.27, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4933 re_mapping 0.0042 re_causal 0.0130 /// teacc 99.16 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.1025, 0.0562, 0.0670, ..., -0.0897, -0.1582, 0.0630], + [-0.0856, -0.0551, -0.1108, ..., 0.1720, -0.0919, -0.0707], + [-0.0263, -0.0011, 0.2009, ..., -0.1090, -0.1407, -0.1284], + ..., + [-0.1259, -0.0580, 0.0153, ..., -0.0435, 0.1175, -0.0863], + [-0.1029, -0.0951, -0.0747, ..., -0.1389, -0.0048, -0.1717], + [-0.1878, -0.0598, -0.1509, ..., -0.1775, -0.0343, 0.1307]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.4457e-06, -1.5721e-05, ..., 5.5879e-09, + -1.1586e-06, -6.9067e-06], + [ 9.3132e-09, 3.7253e-09, 1.5460e-07, ..., -1.6764e-07, + 7.4506e-09, 9.3132e-09], + [ 5.5879e-09, 9.8720e-08, 3.5428e-06, ..., 3.3528e-08, + 9.6858e-08, 2.7567e-07], + ..., + [ 1.3039e-08, 2.6077e-08, 1.6764e-08, ..., 2.9802e-08, + -1.0990e-07, 8.0094e-08], + [ 1.1176e-08, 1.6578e-07, 1.3281e-06, ..., 1.1362e-07, + 8.0094e-08, 5.0105e-07], + [ 3.1665e-08, 1.6876e-06, 7.6517e-06, ..., 9.3132e-09, + 8.3074e-07, 4.7162e-06]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0382, -0.0344, 0.0117, 0.0349, -0.0179, 0.0042, -0.0114, 0.0160, + 0.0152, -0.0063], device='cuda:0'), grad: tensor([-3.2544e-05, -2.4959e-07, 8.0764e-06, 3.7197e-06, -3.1665e-08, + 3.5949e-07, 2.0675e-06, 1.2852e-07, 3.2689e-06, 1.5154e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 214.48, cls_loss 0.0010 cls_loss_mapping 0.0037 cls_loss_causal 0.4969 re_mapping 0.0041 re_causal 0.0127 /// teacc 99.13 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.1026, 0.0562, 0.0672, ..., -0.0898, -0.1590, 0.0631], + [-0.0857, -0.0552, -0.1108, ..., 0.1712, -0.0921, -0.0756], + [-0.0268, -0.0010, 0.2013, ..., -0.1092, -0.1411, -0.1289], + ..., + [-0.1299, -0.0582, 0.0154, ..., -0.0425, 0.1176, -0.0867], + [-0.1035, -0.0951, -0.0751, ..., -0.1380, -0.0048, -0.1674], + [-0.1878, -0.0602, -0.1510, ..., -0.1812, -0.0344, 0.1310]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -6.3889e-07, ..., 1.8626e-09, + 2.7940e-08, -2.1048e-07], + [ 2.7940e-08, 0.0000e+00, 5.2154e-08, ..., -7.8231e-08, + 1.7136e-07, 5.0291e-08], + [ 1.6764e-08, 0.0000e+00, 3.6508e-07, ..., 2.6077e-08, + 2.2352e-08, 1.8440e-07], + ..., + [ 6.1430e-06, 0.0000e+00, -1.1176e-07, ..., -3.9116e-08, + 1.0240e-04, 1.0937e-04], + [ 2.7940e-08, 0.0000e+00, 8.5682e-08, ..., 4.0978e-08, + 9.6858e-08, 1.0245e-07], + [ 3.5390e-08, 0.0000e+00, 5.0291e-08, ..., 2.2352e-08, + -1.0091e-04, -1.0985e-04]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0383, -0.0358, 0.0115, 0.0336, -0.0177, 0.0059, -0.0118, 0.0161, + 0.0168, -0.0063], device='cuda:0'), grad: tensor([-6.9663e-07, 4.4703e-07, 7.4320e-07, 7.0408e-07, -1.1563e-05, + 1.7136e-07, 2.5891e-07, 5.4693e-04, 8.8103e-07, -5.3740e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 214.69, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4737 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.11 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.1026, 0.0562, 0.0671, ..., -0.0898, -0.1610, 0.0630], + [-0.0857, -0.0552, -0.1111, ..., 0.1715, -0.0923, -0.0757], + [-0.0270, -0.0009, 0.2019, ..., -0.1093, -0.1414, -0.1290], + ..., + [-0.1320, -0.0582, 0.0154, ..., -0.0429, 0.1175, -0.0878], + [-0.1042, -0.0951, -0.0754, ..., -0.1382, -0.0050, -0.1675], + [-0.1879, -0.0602, -0.1511, ..., -0.1825, -0.0340, 0.1320]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 0.0000e+00, 1.1921e-07, ..., 6.8918e-08, + 6.5193e-08, 0.0000e+00], + [ 4.8429e-08, 0.0000e+00, 3.9116e-08, ..., -9.7975e-07, + 9.3132e-09, 9.3132e-09], + [ 3.9116e-08, 0.0000e+00, -2.0303e-07, ..., 1.5460e-07, + 3.8370e-07, 7.4506e-09], + ..., + [ 1.3784e-07, 0.0000e+00, -7.3761e-07, ..., 1.4342e-07, + -5.5507e-07, 1.3039e-08], + [ 5.9605e-08, 0.0000e+00, 3.7625e-07, ..., 4.6194e-07, + -2.0489e-08, 3.9302e-07], + [ 3.7253e-07, 0.0000e+00, 1.0803e-07, ..., 2.0489e-08, + 7.6368e-08, -7.6927e-07]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0383, -0.0357, 0.0117, 0.0327, -0.0177, 0.0074, -0.0116, 0.0155, + 0.0166, -0.0060], device='cuda:0'), grad: tensor([ 7.4506e-07, -2.1718e-06, 1.0598e-06, -4.9919e-07, -7.4878e-07, + 8.4750e-07, 8.3074e-07, -9.5181e-07, 1.4883e-06, -6.0536e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 214.28, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.5049 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.16 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.1027, 0.0563, 0.0674, ..., -0.0899, -0.1625, 0.0630], + [-0.0859, -0.0557, -0.1111, ..., 0.1723, -0.0925, -0.0757], + [-0.0276, -0.0010, 0.2022, ..., -0.1096, -0.1418, -0.1293], + ..., + [-0.1339, -0.0583, 0.0154, ..., -0.0440, 0.1177, -0.0880], + [-0.1051, -0.0951, -0.0756, ..., -0.1385, -0.0052, -0.1676], + [-0.1881, -0.0602, -0.1513, ..., -0.1853, -0.0343, 0.1323]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, 6.1467e-08, ..., 1.8626e-09, + 4.2841e-08, 8.0094e-08], + [ 5.5879e-09, 0.0000e+00, 4.0978e-08, ..., -2.4028e-07, + 3.7253e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -3.7998e-07, ..., 9.3132e-09, + 1.8626e-09, 1.8626e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 1.5460e-07, ..., 1.6578e-07, + 3.7253e-09, 3.7253e-09], + [ 5.5879e-09, 0.0000e+00, 2.6077e-08, ..., 3.3528e-08, + 1.3970e-07, 1.5274e-07], + [-1.8626e-08, 0.0000e+00, 1.4901e-08, ..., 1.8626e-09, + 3.7253e-08, -1.3039e-08]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0384, -0.0353, 0.0113, 0.0332, -0.0175, 0.0065, -0.0104, 0.0151, + 0.0164, -0.0062], device='cuda:0'), grad: tensor([ 3.0734e-07, -4.1164e-07, -3.3714e-07, -2.9989e-07, 1.3970e-07, + -8.1025e-07, 3.5018e-07, 4.9174e-07, 5.6997e-07, 1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 214.46, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5018 re_mapping 0.0042 re_causal 0.0122 /// teacc 99.19 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.1028, 0.0563, 0.0672, ..., -0.0899, -0.1643, 0.0625], + [-0.0859, -0.0557, -0.1107, ..., 0.1740, -0.0926, -0.0757], + [-0.0279, -0.0010, 0.2025, ..., -0.1114, -0.1422, -0.1299], + ..., + [-0.1345, -0.0583, 0.0126, ..., -0.0453, 0.1166, -0.0882], + [-0.1058, -0.0951, -0.0731, ..., -0.1387, -0.0031, -0.1677], + [-0.1881, -0.0602, -0.1512, ..., -0.1878, -0.0344, 0.1334]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, -7.4506e-09, ..., 5.5879e-09, + 3.7253e-09, -7.4506e-09], + [ 2.0489e-08, 0.0000e+00, 6.7055e-08, ..., -2.1234e-07, + 8.9407e-08, 5.5879e-09], + [ 1.1176e-08, 0.0000e+00, -3.9302e-07, ..., 6.7055e-08, + -7.2643e-08, 1.8626e-09], + ..., + [ 2.5146e-07, 0.0000e+00, 9.4995e-08, ..., 9.3132e-08, + 3.9116e-08, 1.1735e-07], + [ 2.7940e-08, 0.0000e+00, 2.4214e-08, ..., 1.2666e-07, + 8.7544e-08, 9.3132e-09], + [ 4.8652e-06, 0.0000e+00, 2.2352e-08, ..., 9.3132e-09, + 2.0526e-06, -2.9244e-07]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0387, -0.0343, 0.0101, 0.0333, -0.0176, 0.0063, -0.0103, 0.0132, + 0.0176, -0.0060], device='cuda:0'), grad: tensor([ 3.7253e-08, -2.3097e-07, -2.0675e-07, 1.8440e-07, -1.6600e-05, + -8.9966e-07, 3.8370e-07, 1.3225e-06, 7.3761e-07, 1.5303e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 214.62, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.5092 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.15 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.1028, 0.0563, 0.0673, ..., -0.0900, -0.1673, 0.0623], + [-0.0859, -0.0557, -0.1114, ..., 0.1750, -0.0936, -0.0756], + [-0.0280, -0.0009, 0.2034, ..., -0.1118, -0.1426, -0.1302], + ..., + [-0.1350, -0.0583, 0.0131, ..., -0.0465, 0.1188, -0.0893], + [-0.1061, -0.0951, -0.0732, ..., -0.1389, -0.0033, -0.1680], + [-0.1882, -0.0602, -0.1525, ..., -0.1913, -0.0368, 0.1342]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.2224e-07, + 1.1176e-08, 4.1723e-07], + [ 9.3132e-09, 0.0000e+00, 1.8626e-09, ..., -1.3411e-07, + 3.7253e-09, 4.8429e-08], + [ 3.7253e-09, 0.0000e+00, -1.3597e-07, ..., 3.1665e-08, + 3.7253e-09, 1.1176e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 8.3819e-08, + 5.5879e-09, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 9.4995e-08, ..., 9.1270e-07, + 1.2480e-07, 1.1884e-06], + [ 1.4901e-08, 0.0000e+00, 1.8626e-09, ..., 5.4017e-08, + 1.8626e-08, 5.2154e-08]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0388, -0.0340, 0.0102, 0.0333, -0.0177, 0.0063, -0.0096, 0.0151, + 0.0173, -0.0073], device='cuda:0'), grad: tensor([ 1.5348e-06, -2.1420e-07, -9.8720e-08, 1.3150e-06, 5.9605e-08, + 6.3181e-06, -1.4156e-05, 2.4401e-07, 4.7050e-06, 2.6636e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 214.22, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4916 re_mapping 0.0042 re_causal 0.0135 /// teacc 99.07 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.1028, 0.0563, 0.0679, ..., -0.0901, -0.1687, 0.0624], + [-0.0860, -0.0558, -0.1115, ..., 0.1754, -0.0938, -0.0756], + [-0.0288, -0.0003, 0.2037, ..., -0.1123, -0.1426, -0.1308], + ..., + [-0.1354, -0.0585, 0.0130, ..., -0.0469, 0.1189, -0.0895], + [-0.1077, -0.0952, -0.0733, ..., -0.1391, -0.0034, -0.1689], + [-0.1879, -0.0603, -0.1526, ..., -0.1914, -0.0370, 0.1371]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.5682e-08, ..., 8.5682e-08, + 0.0000e+00, -6.1467e-08], + [ 1.8626e-09, 0.0000e+00, 9.3132e-09, ..., 4.9546e-07, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.0226e-06, + 1.1176e-08, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., 3.5390e-07, + -1.3411e-07, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 2.8685e-07, + 1.1176e-08, 1.1176e-08], + [ 7.4506e-09, 0.0000e+00, 3.9116e-08, ..., 1.4901e-08, + 5.2154e-08, 1.4901e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0388, -0.0339, 0.0100, 0.0334, -0.0187, 0.0063, -0.0092, 0.0150, + 0.0169, -0.0063], device='cuda:0'), grad: tensor([ 4.4890e-07, 3.8445e-06, 7.7486e-06, -1.9878e-05, 3.9116e-07, + 2.7865e-06, 2.0489e-08, 2.4587e-06, 1.8645e-06, 3.0175e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 214.66, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4879 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.17 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.1028, 0.0563, 0.0691, ..., -0.0901, -0.1692, 0.0627], + [-0.0861, -0.0558, -0.1121, ..., 0.1753, -0.0950, -0.0757], + [-0.0294, -0.0002, 0.2049, ..., -0.1124, -0.1428, -0.1314], + ..., + [-0.1362, -0.0585, 0.0128, ..., -0.0462, 0.1193, -0.0896], + [-0.1079, -0.0952, -0.0734, ..., -0.1393, -0.0035, -0.1689], + [-0.1879, -0.0603, -0.1531, ..., -0.1917, -0.0371, 0.1373]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 9.1270e-08, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 3.4273e-07, ..., -2.9802e-08, + 2.5146e-07, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, -3.6694e-07, ..., 1.1176e-08, + 5.7369e-07, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -2.1048e-07, ..., 3.7253e-09, + 4.6194e-07, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 7.0781e-08, ..., 9.3132e-09, + 7.4506e-07, 8.7544e-08], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 0.0000e+00, + 1.3411e-07, -2.9616e-07]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0387, -0.0343, 0.0105, 0.0334, -0.0188, 0.0063, -0.0091, 0.0151, + 0.0170, -0.0063], device='cuda:0'), grad: tensor([ 7.7300e-07, 2.1979e-06, 2.8815e-06, -2.5123e-05, 3.2037e-07, + 7.4394e-06, 2.5705e-07, 5.9009e-06, 4.9844e-06, 3.2037e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.28, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4872 re_mapping 0.0041 re_causal 0.0128 /// teacc 99.11 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.1029, 0.0563, 0.0705, ..., -0.0901, -0.1696, 0.0634], + [-0.0862, -0.0591, -0.1129, ..., 0.1752, -0.0955, -0.0758], + [-0.0294, 0.0027, 0.2056, ..., -0.1123, -0.1432, -0.1320], + ..., + [-0.1368, -0.0585, 0.0128, ..., -0.0459, 0.1196, -0.0901], + [-0.1080, -0.0952, -0.0734, ..., -0.1395, -0.0037, -0.1692], + [-0.1879, -0.0604, -0.1534, ..., -0.1920, -0.0372, 0.1375]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.5682e-08, ..., 0.0000e+00, + 5.5879e-09, -7.8231e-08], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., -9.1270e-08, + 3.5390e-08, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 2.7940e-08, + 2.4214e-08, 5.5879e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 2.4214e-08, + -1.3784e-07, -2.9802e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 2.6077e-08, + -5.2340e-07, 5.5879e-09], + [ 1.1176e-08, 0.0000e+00, 4.2841e-08, ..., 5.5879e-09, + 1.7509e-07, 3.5390e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0384, -0.0346, 0.0108, 0.0334, -0.0188, 0.0065, -0.0097, 0.0151, + 0.0169, -0.0064], device='cuda:0'), grad: tensor([-1.2666e-07, -1.8626e-09, 1.9372e-07, 8.8103e-07, 1.1176e-08, + 1.3858e-06, 8.0094e-08, -1.2293e-07, -3.0734e-06, 7.5623e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 214.18, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.5050 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.14 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.1029, 0.0563, 0.0710, ..., -0.0901, -0.1708, 0.0637], + [-0.0864, -0.0591, -0.1131, ..., 0.1754, -0.0958, -0.0758], + [-0.0299, 0.0028, 0.2065, ..., -0.1124, -0.1432, -0.1320], + ..., + [-0.1374, -0.0588, 0.0127, ..., -0.0460, 0.1198, -0.0903], + [-0.1077, -0.0952, -0.0735, ..., -0.1395, -0.0038, -0.1694], + [-0.1881, -0.0604, -0.1539, ..., -0.1923, -0.0373, 0.1376]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, 0.0000e+00, 1.6764e-08, ..., 1.6950e-07, + 3.7253e-09, 8.0094e-08], + [ 5.5879e-09, 0.0000e+00, 2.2352e-08, ..., -3.5763e-07, + 1.4901e-08, 3.7253e-09], + [ 7.4506e-09, -0.0000e+00, 2.0489e-08, ..., 2.8312e-07, + 1.1548e-07, 3.7253e-09], + ..., + [ 7.4506e-09, 0.0000e+00, -1.2107e-07, ..., 2.0489e-08, + -1.7509e-07, 7.4506e-09], + [-2.7940e-08, 0.0000e+00, 2.4214e-08, ..., 1.7136e-07, + 3.5390e-08, -3.3528e-08], + [ 1.0990e-07, 0.0000e+00, 5.5879e-09, ..., 5.5879e-09, + 5.5879e-09, -1.6764e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0383, -0.0347, 0.0110, 0.0334, -0.0187, 0.0066, -0.0097, 0.0152, + 0.0168, -0.0067], device='cuda:0'), grad: tensor([ 5.1595e-07, -3.2596e-07, 5.3085e-07, -1.6969e-06, -8.1956e-08, + 1.9278e-06, -6.7241e-07, -2.8498e-07, -4.0792e-07, 4.7870e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 214.40, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4787 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.09 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.1029, 0.0563, 0.0711, ..., -0.0901, -0.1749, 0.0638], + [-0.0864, -0.0592, -0.1134, ..., 0.1756, -0.0962, -0.0759], + [-0.0301, 0.0028, 0.2070, ..., -0.1125, -0.1435, -0.1321], + ..., + [-0.1375, -0.0589, 0.0127, ..., -0.0461, 0.1202, -0.0904], + [-0.1078, -0.0953, -0.0736, ..., -0.1396, -0.0040, -0.1695], + [-0.1882, -0.0606, -0.1541, ..., -0.1925, -0.0374, 0.1377]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, -7.2643e-08, ..., 3.7253e-09, + 9.3132e-09, -7.0781e-08], + [ 9.3132e-09, 0.0000e+00, 5.5879e-09, ..., -3.7253e-07, + 1.4901e-08, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, -2.0489e-07, ..., 1.6391e-07, + 1.3039e-08, 1.8626e-09], + ..., + [ 5.4017e-08, 0.0000e+00, 1.6764e-07, ..., 4.4703e-08, + 2.9802e-08, 2.0489e-08], + [ 9.3132e-09, 0.0000e+00, 1.1176e-08, ..., 7.6368e-08, + 1.6764e-08, 2.2352e-08], + [ 1.1176e-08, 0.0000e+00, 4.6566e-08, ..., 5.5879e-09, + 1.8626e-08, -7.4506e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0383, -0.0348, 0.0110, 0.0336, -0.0186, 0.0064, -0.0096, 0.0153, + 0.0166, -0.0068], device='cuda:0'), grad: tensor([-6.3330e-08, -7.5437e-07, 1.6950e-07, -6.9439e-06, 1.0431e-07, + 6.7353e-06, 5.9605e-08, 6.5193e-07, -5.5879e-09, 5.2154e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 214.31, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4606 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.12 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.1030, 0.0563, 0.0722, ..., -0.0902, -0.1760, 0.0643], + [-0.0866, -0.0592, -0.1136, ..., 0.1758, -0.0967, -0.0760], + [-0.0305, 0.0028, 0.2080, ..., -0.1127, -0.1438, -0.1325], + ..., + [-0.1381, -0.0589, 0.0127, ..., -0.0463, 0.1205, -0.0905], + [-0.1059, -0.0953, -0.0737, ..., -0.1397, -0.0041, -0.1696], + [-0.1883, -0.0613, -0.1548, ..., -0.1927, -0.0375, 0.1376]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.9605e-08, ..., 1.3039e-08, + 1.4156e-07, 2.4959e-07], + [ 7.4506e-09, 0.0000e+00, 1.6764e-08, ..., -4.8839e-06, + 9.3132e-09, 2.0489e-08], + [ 1.8626e-09, 0.0000e+00, -7.4506e-08, ..., 3.6918e-06, + 2.2352e-08, 2.6077e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 1.0040e-06, + -2.9802e-08, 7.0781e-08], + [-3.7253e-09, 0.0000e+00, 1.8626e-09, ..., 5.4017e-08, + 1.8813e-07, 4.0233e-07], + [ 2.0489e-08, 0.0000e+00, 2.7940e-08, ..., 1.4901e-08, + 1.4901e-08, -2.1607e-07]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0381, -0.0348, 0.0114, 0.0335, -0.0186, 0.0064, -0.0100, 0.0153, + 0.0169, -0.0070], device='cuda:0'), grad: tensor([ 1.1921e-06, -7.7486e-06, 6.7689e-06, 9.3505e-07, 2.1607e-07, + -4.6864e-06, 1.7174e-06, 1.9800e-06, -1.4156e-07, -2.6077e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 214.42, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4857 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.07 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.1031, 0.0563, 0.0727, ..., -0.0902, -0.1771, 0.0644], + [-0.0867, -0.0593, -0.1140, ..., 0.1760, -0.0974, -0.0761], + [-0.0309, 0.0030, 0.2086, ..., -0.1130, -0.1452, -0.1327], + ..., + [-0.1392, -0.0589, 0.0132, ..., -0.0461, 0.1218, -0.0906], + [-0.1068, -0.0953, -0.0739, ..., -0.1400, -0.0045, -0.1697], + [-0.1884, -0.0614, -0.1552, ..., -0.1931, -0.0378, 0.1377]], + device='cuda:0'), grad: tensor([[ 1.4137e-06, 0.0000e+00, -3.3155e-07, ..., 3.7625e-07, + 1.8626e-09, -2.7008e-07], + [ 1.3039e-08, 0.0000e+00, 3.3528e-08, ..., -2.9802e-08, + 7.4506e-09, 3.5390e-08], + [ 8.5682e-08, 0.0000e+00, 1.0431e-07, ..., 2.6077e-08, + 1.1176e-08, 7.0781e-08], + ..., + [ 1.8626e-08, 0.0000e+00, -4.6566e-08, ..., 3.7253e-09, + -1.0058e-07, 4.8429e-08], + [ 1.1362e-06, 0.0000e+00, 3.7812e-07, ..., 2.9244e-07, + 9.3132e-09, 4.6566e-08], + [ 3.1665e-08, 0.0000e+00, 9.3132e-08, ..., 1.8626e-09, + 5.5879e-09, -4.1723e-07]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0382, -0.0351, 0.0112, 0.0323, -0.0186, 0.0079, -0.0094, 0.0163, + 0.0165, -0.0071], device='cuda:0'), grad: tensor([ 2.7008e-06, 1.3039e-07, 6.1467e-07, 3.1106e-07, 1.4249e-06, + 2.3283e-07, -8.0913e-06, -7.0781e-08, 3.4217e-06, -7.0035e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 214.43, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4789 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.1032, 0.0563, 0.0735, ..., -0.0903, -0.1773, 0.0647], + [-0.0868, -0.0593, -0.1142, ..., 0.1765, -0.0975, -0.0761], + [-0.0305, 0.0032, 0.2089, ..., -0.1131, -0.1453, -0.1332], + ..., + [-0.1397, -0.0589, 0.0132, ..., -0.0464, 0.1220, -0.0907], + [-0.1072, -0.0954, -0.0740, ..., -0.1402, -0.0046, -0.1699], + [-0.1884, -0.0614, -0.1553, ..., -0.1935, -0.0378, 0.1382]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 1.0058e-07, ..., 7.4506e-09, + 2.7940e-08, 9.6858e-08], + [-5.5879e-09, 0.0000e+00, 1.3039e-07, ..., -8.5682e-08, + 4.0978e-08, 1.3039e-08], + [ 9.3132e-09, 0.0000e+00, -1.8217e-06, ..., 2.4214e-08, + 1.7695e-07, 1.6764e-08], + ..., + [ 5.5879e-09, 0.0000e+00, -1.6764e-08, ..., 2.6077e-08, + -1.8980e-06, 2.1793e-07], + [ 9.3132e-09, 0.0000e+00, 4.6566e-08, ..., 2.9802e-08, + 8.3819e-08, 5.9605e-08], + [ 1.8626e-09, 0.0000e+00, 2.0862e-07, ..., 1.8626e-09, + 2.6263e-07, -5.0291e-07]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0381, -0.0350, 0.0112, 0.0320, -0.0187, 0.0082, -0.0095, 0.0163, + 0.0164, -0.0069], device='cuda:0'), grad: tensor([ 7.2084e-07, 2.2538e-07, -3.4235e-06, 7.0781e-07, 2.5686e-06, + 1.7807e-06, -1.6764e-08, -2.8815e-06, 5.4762e-07, -2.4587e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.49, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4727 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.16 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.1034, 0.0563, 0.0740, ..., -0.0904, -0.1778, 0.0648], + [-0.0868, -0.0593, -0.1143, ..., 0.1770, -0.0977, -0.0762], + [-0.0308, 0.0033, 0.2091, ..., -0.1136, -0.1457, -0.1337], + ..., + [-0.1402, -0.0589, 0.0133, ..., -0.0468, 0.1222, -0.0907], + [-0.1076, -0.0954, -0.0740, ..., -0.1404, -0.0050, -0.1703], + [-0.1884, -0.0614, -0.1555, ..., -0.1927, -0.0380, 0.1393]], + device='cuda:0'), grad: tensor([[ 1.1921e-07, 0.0000e+00, -5.5879e-09, ..., 8.3819e-08, + 3.1665e-08, 1.3039e-08], + [ 2.4214e-07, 0.0000e+00, 0.0000e+00, ..., -2.0489e-07, + 3.9116e-08, -2.7940e-08], + [ 7.6368e-08, 0.0000e+00, 0.0000e+00, ..., 3.3528e-08, + 1.5087e-07, 3.7253e-09], + ..., + [ 1.9372e-07, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, + -1.9576e-06, 1.4901e-08], + [ 2.9057e-07, 0.0000e+00, 0.0000e+00, ..., 2.5146e-07, + 8.7544e-08, 4.2841e-08], + [ 1.4663e-05, 0.0000e+00, 3.7253e-09, ..., 9.3132e-09, + 8.5682e-08, -1.8999e-07]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0381, -0.0349, 0.0106, 0.0322, -0.0188, 0.0078, -0.0090, 0.0165, + 0.0159, -0.0066], device='cuda:0'), grad: tensor([ 7.4133e-07, 6.4261e-07, 9.7044e-07, 4.5449e-06, -5.3376e-05, + 1.1045e-06, -1.9837e-06, -5.4426e-06, 1.4529e-06, 5.1439e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 214.44, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4846 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.18 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.1035, 0.0563, 0.0729, ..., -0.0904, -0.1774, 0.0626], + [-0.0869, -0.0593, -0.1145, ..., 0.1782, -0.0978, -0.0762], + [-0.0312, 0.0033, 0.2105, ..., -0.1139, -0.1459, -0.1340], + ..., + [-0.1411, -0.0590, 0.0130, ..., -0.0496, 0.1223, -0.0909], + [-0.1083, -0.0955, -0.0742, ..., -0.1407, -0.0050, -0.1704], + [-0.1884, -0.0614, -0.1546, ..., -0.1929, -0.0380, 0.1420]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 0.0000e+00, 1.8626e-09, ..., 4.4703e-08, + 5.5879e-09, -9.3132e-09], + [ 1.8626e-08, 0.0000e+00, 1.3411e-07, ..., -2.2352e-08, + 3.3528e-08, 0.0000e+00], + [ 9.3132e-09, 0.0000e+00, -3.0734e-07, ..., -2.6077e-08, + 8.5682e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.9244e-07, ..., 2.4214e-08, + -6.2771e-07, 0.0000e+00], + [ 1.1921e-07, 0.0000e+00, 1.4156e-07, ..., 1.4901e-07, + 6.3330e-08, 1.8626e-09], + [ 1.4901e-08, 0.0000e+00, 2.3842e-07, ..., 1.3039e-08, + 3.7439e-07, 0.0000e+00]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0394, -0.0345, 0.0112, 0.0322, -0.0190, 0.0078, -0.0089, 0.0159, + 0.0159, -0.0054], device='cuda:0'), grad: tensor([ 2.8312e-07, 3.7998e-07, -1.8254e-07, -4.8801e-07, 1.0058e-07, + 7.9535e-07, -1.1418e-06, -1.5534e-06, 5.7183e-07, 1.2349e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 214.51, cls_loss 0.0022 cls_loss_mapping 0.0042 cls_loss_causal 0.5151 re_mapping 0.0041 re_causal 0.0127 /// teacc 99.03 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.1035, 0.0563, 0.0701, ..., -0.0902, -0.1777, 0.0598], + [-0.0870, -0.0593, -0.1118, ..., 0.1821, -0.0980, -0.0765], + [-0.0316, 0.0033, 0.2113, ..., -0.1150, -0.1461, -0.1351], + ..., + [-0.1420, -0.0590, 0.0128, ..., -0.0509, 0.1224, -0.0912], + [-0.1087, -0.0955, -0.0769, ..., -0.1446, -0.0050, -0.1709], + [-0.1885, -0.0615, -0.1518, ..., -0.1946, -0.0381, 0.1449]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.7917e-07, ..., 7.4506e-09, + 0.0000e+00, -1.3970e-06], + [ 1.6764e-08, 0.0000e+00, 5.1968e-07, ..., -2.6636e-07, + 3.7253e-09, 3.9116e-08], + [-1.1176e-08, 1.8626e-09, -8.3819e-07, ..., 9.6858e-08, + 5.5879e-09, 2.9802e-08], + ..., + [ 1.8626e-09, 0.0000e+00, 2.6077e-08, ..., 6.3330e-08, + -1.8626e-08, 5.4017e-08], + [ 5.5879e-09, 0.0000e+00, 1.6019e-07, ..., 5.9605e-08, + 1.8626e-09, 1.2666e-07], + [ 0.0000e+00, 0.0000e+00, 4.6380e-07, ..., 1.8626e-09, + 3.7253e-09, 7.2643e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0422, -0.0311, 0.0107, 0.0326, -0.0191, 0.0070, -0.0082, 0.0155, + 0.0125, -0.0026], device='cuda:0'), grad: tensor([-3.4422e-06, 3.6508e-07, -9.2760e-07, 1.9539e-06, 4.8801e-07, + 3.3155e-07, 1.6019e-07, 2.7940e-07, 5.9977e-07, 1.5832e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 214.41, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.5245 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.11 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.1036, 0.0563, 0.0701, ..., -0.0902, -0.1782, 0.0598], + [-0.0873, -0.0593, -0.1119, ..., 0.1822, -0.0984, -0.0766], + [-0.0326, 0.0034, 0.2091, ..., -0.1151, -0.1502, -0.1357], + ..., + [-0.1433, -0.0590, 0.0146, ..., -0.0518, 0.1236, -0.0915], + [-0.1091, -0.0955, -0.0770, ..., -0.1446, -0.0051, -0.1710], + [-0.1886, -0.0621, -0.1518, ..., -0.1950, -0.0383, 0.1450]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -6.5193e-08, ..., 7.4506e-09, + 7.4506e-09, -1.0803e-07], + [ 9.3132e-09, 0.0000e+00, 1.8626e-08, ..., -9.3132e-08, + 7.4506e-09, 7.4506e-09], + [ 1.8626e-09, 0.0000e+00, -1.4529e-07, ..., 1.6764e-08, + 1.8626e-09, 1.8626e-09], + ..., + [ 7.4506e-09, 0.0000e+00, -3.1665e-08, ..., 2.9802e-08, + -3.8929e-07, 5.7742e-08], + [ 7.4506e-09, 0.0000e+00, 3.3528e-08, ..., 3.3528e-08, + 5.5879e-09, 1.6764e-08], + [ 4.6566e-08, 0.0000e+00, 1.0990e-07, ..., 5.5879e-09, + 1.8626e-07, -8.5682e-08]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0422, -0.0311, 0.0083, 0.0329, -0.0184, 0.0070, -0.0101, 0.0164, + 0.0125, -0.0026], device='cuda:0'), grad: tensor([-9.6858e-08, 4.8429e-08, 3.3714e-07, 4.5076e-07, 1.2480e-07, + 1.1716e-06, 1.8813e-07, -4.1910e-07, -2.1104e-06, 2.8126e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 214.39, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4642 re_mapping 0.0041 re_causal 0.0120 /// teacc 99.15 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.1037, 0.0563, 0.0701, ..., -0.0902, -0.1789, 0.0598], + [-0.0878, -0.0593, -0.1120, ..., 0.1822, -0.0987, -0.0767], + [-0.0334, 0.0034, 0.2097, ..., -0.1157, -0.1505, -0.1360], + ..., + [-0.1473, -0.0590, 0.0146, ..., -0.0530, 0.1237, -0.0918], + [-0.1096, -0.0955, -0.0771, ..., -0.1446, -0.0051, -0.1712], + [-0.1886, -0.0621, -0.1518, ..., -0.1953, -0.0383, 0.1450]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4959e-07, ..., -5.0291e-08, + 1.4715e-07, 4.0978e-08], + [ 3.7253e-09, 0.0000e+00, 4.6566e-08, ..., 1.4342e-07, + 1.8626e-08, 5.2154e-08], + [ 0.0000e+00, 0.0000e+00, 2.2724e-07, ..., 9.8720e-08, + 1.1921e-07, 3.0175e-07], + ..., + [ 1.8626e-09, 0.0000e+00, 3.7253e-08, ..., 2.6077e-08, + 7.2643e-08, 7.2643e-08], + [ 0.0000e+00, 0.0000e+00, 6.1467e-08, ..., 1.0990e-07, + 1.2480e-07, 8.1956e-08], + [ 7.4506e-09, 0.0000e+00, 2.0675e-07, ..., 4.2841e-08, + 6.1467e-08, 2.5891e-07]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0422, -0.0313, 0.0075, 0.0340, -0.0184, 0.0068, -0.0107, 0.0158, + 0.0125, -0.0026], device='cuda:0'), grad: tensor([ 2.3469e-06, 6.9104e-07, 2.6282e-06, -3.7283e-05, 9.8720e-08, + 2.8104e-05, -8.6054e-07, 1.2275e-06, 2.0079e-06, 1.0431e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 214.30, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4886 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.1038, 0.0563, 0.0701, ..., -0.0901, -0.1797, 0.0598], + [-0.0880, -0.0593, -0.1120, ..., 0.1823, -0.0990, -0.0767], + [-0.0336, 0.0034, 0.2098, ..., -0.1159, -0.1509, -0.1364], + ..., + [-0.1492, -0.0590, 0.0148, ..., -0.0532, 0.1241, -0.0919], + [-0.1104, -0.0955, -0.0771, ..., -0.1447, -0.0052, -0.1710], + [-0.1887, -0.0621, -0.1518, ..., -0.1957, -0.0385, 0.1450]], + device='cuda:0'), grad: tensor([[ 8.0746e-07, 0.0000e+00, -1.5777e-06, ..., -6.5193e-09, + -4.3027e-07, -1.0356e-06], + [ 7.6648e-07, 0.0000e+00, 3.6322e-08, ..., 2.7940e-09, + 8.3819e-09, 1.3039e-08], + [ 2.1160e-06, 0.0000e+00, 2.9150e-07, ..., 5.5879e-09, + 8.8476e-08, 1.9465e-07], + ..., + [ 3.9116e-08, 0.0000e+00, 6.8918e-08, ..., 2.1420e-08, + 1.3970e-08, 4.8429e-08], + [ 1.5832e-08, 0.0000e+00, 1.4156e-07, ..., 7.2643e-08, + 6.0536e-08, 1.0524e-07], + [ 5.3272e-07, 0.0000e+00, 7.6182e-07, ..., 1.3970e-08, + 2.1420e-07, 4.7870e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0422, -0.0312, 0.0073, 0.0342, -0.0183, 0.0065, -0.0106, 0.0158, + 0.0125, -0.0026], device='cuda:0'), grad: tensor([-2.5537e-06, 1.7239e-06, 3.5465e-06, -5.5209e-06, -5.6699e-06, + 1.1371e-06, 6.8545e-07, 6.1654e-07, 3.4459e-06, 2.5854e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 214.40, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.5245 re_mapping 0.0038 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.1040, 0.0563, 0.0701, ..., -0.0905, -0.1800, 0.0598], + [-0.0885, -0.0594, -0.1122, ..., 0.1824, -0.0994, -0.0767], + [-0.0336, 0.0041, 0.2107, ..., -0.1169, -0.1510, -0.1367], + ..., + [-0.1511, -0.0597, 0.0148, ..., -0.0534, 0.1244, -0.0920], + [-0.1109, -0.0956, -0.0773, ..., -0.1447, -0.0053, -0.1711], + [-0.1888, -0.0621, -0.1518, ..., -0.1962, -0.0387, 0.1450]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 1.1176e-08, ..., 9.3132e-10, + 2.5146e-08, 4.0978e-08], + [ 9.0338e-08, 0.0000e+00, 4.9081e-07, ..., 2.6450e-07, + 6.2399e-07, 3.7253e-09], + [ 6.5193e-09, 0.0000e+00, -2.5891e-07, ..., 1.3039e-08, + 2.7008e-08, 2.3283e-08], + ..., + [ 1.9558e-08, 0.0000e+00, -5.8487e-07, ..., -3.0641e-07, + -7.8417e-07, 1.8626e-08], + [ 2.3283e-08, 0.0000e+00, 3.4459e-08, ..., 2.3283e-08, + 1.8626e-08, 5.5879e-09], + [ 3.3714e-07, 0.0000e+00, 1.9278e-07, ..., 1.4901e-08, + 1.1744e-06, 2.5835e-06]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0422, -0.0312, 0.0070, 0.0341, -0.0182, 0.0064, -0.0100, 0.0158, + 0.0126, -0.0026], device='cuda:0'), grad: tensor([ 1.5553e-07, 2.1011e-06, -3.0361e-07, -7.1712e-08, -1.1148e-06, + -6.3404e-06, -1.5460e-07, -1.8682e-06, 3.4459e-07, 7.2233e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.36, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5090 re_mapping 0.0041 re_causal 0.0127 /// teacc 98.94 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.1044, 0.0563, 0.0701, ..., -0.0907, -0.1811, 0.0598], + [-0.0887, -0.0594, -0.1125, ..., 0.1825, -0.1008, -0.0767], + [-0.0340, 0.0041, 0.2111, ..., -0.1179, -0.1511, -0.1376], + ..., + [-0.1515, -0.0597, 0.0153, ..., -0.0527, 0.1249, -0.0921], + [-0.1114, -0.0956, -0.0774, ..., -0.1447, -0.0054, -0.1711], + [-0.1889, -0.0622, -0.1518, ..., -0.1969, -0.0390, 0.1449]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, -8.1956e-08, ..., -0.0000e+00, + 0.0000e+00, -1.0338e-07], + [ 1.7695e-08, 0.0000e+00, -8.2180e-06, ..., -1.3016e-05, + 1.3039e-08, 2.7940e-09], + [ 4.6566e-09, 0.0000e+00, 7.7039e-06, ..., 1.2137e-05, + 1.8626e-09, 8.3819e-09], + ..., + [ 1.3039e-08, 0.0000e+00, 4.2003e-07, ..., 6.8825e-07, + -2.8871e-08, 6.9849e-08], + [ 2.1420e-08, 0.0000e+00, 4.0978e-08, ..., 2.5146e-08, + -1.3039e-08, 7.6368e-08], + [ 1.3877e-07, 0.0000e+00, 1.3225e-07, ..., 9.3132e-08, + 1.2107e-08, 4.6566e-08]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0422, -0.0313, 0.0065, 0.0342, -0.0182, 0.0063, -0.0097, 0.0168, + 0.0126, -0.0027], device='cuda:0'), grad: tensor([ 1.3039e-08, -2.6211e-05, 2.4557e-05, -4.5914e-07, -2.1607e-07, + 1.0990e-07, 8.9407e-08, 1.5935e-06, -2.0023e-07, 6.8452e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 214.58, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4829 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.1049, 0.0564, 0.0702, ..., -0.0907, -0.1814, 0.0599], + [-0.0893, -0.0594, -0.1125, ..., 0.1829, -0.1019, -0.0768], + [-0.0353, 0.0041, 0.2114, ..., -0.1189, -0.1513, -0.1384], + ..., + [-0.1527, -0.0598, 0.0155, ..., -0.0553, 0.1254, -0.0923], + [-0.1089, -0.0957, -0.0775, ..., -0.1443, -0.0055, -0.1713], + [-0.1891, -0.0623, -0.1519, ..., -0.1979, -0.0391, 0.1450]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.2107e-08, -2.0396e-07, ..., 0.0000e+00, + 1.3039e-08, -2.8498e-07], + [ 1.1176e-08, 0.0000e+00, 5.1223e-08, ..., -3.5390e-08, + 3.2596e-08, 3.7253e-09], + [ 2.7940e-09, 0.0000e+00, -2.3004e-07, ..., 3.7253e-09, + 3.0734e-08, 4.6566e-09], + ..., + [ 9.3132e-09, 0.0000e+00, 6.5193e-09, ..., 1.4901e-08, + -3.2503e-07, 9.3132e-09], + [ 3.7253e-09, 9.3132e-10, 2.7008e-08, ..., 1.0245e-08, + 1.0245e-08, 5.8673e-08], + [ 3.4831e-07, 9.3132e-09, 2.7288e-07, ..., 1.8626e-09, + 2.5146e-08, 9.6858e-08]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0422, -0.0311, 0.0061, 0.0336, -0.0181, 0.0075, -0.0137, 0.0164, + 0.0134, -0.0027], device='cuda:0'), grad: tensor([-5.4296e-07, 1.7416e-07, -1.2387e-07, 2.1327e-06, -3.7160e-07, + -4.3027e-07, 3.1665e-08, -2.1830e-06, 2.5425e-07, 1.0580e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 214.28, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4936 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.15 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.1049, 0.0564, 0.0702, ..., -0.0900, -0.1832, 0.0599], + [-0.0895, -0.0594, -0.1126, ..., 0.1831, -0.1029, -0.0769], + [-0.0354, 0.0041, 0.2118, ..., -0.1195, -0.1532, -0.1421], + ..., + [-0.1536, -0.0598, 0.0157, ..., -0.0559, 0.1265, -0.0925], + [-0.1092, -0.0957, -0.0780, ..., -0.1444, -0.0056, -0.1718], + [-0.1894, -0.0624, -0.1519, ..., -0.1976, -0.0401, 0.1450]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9092e-07, ..., 4.4703e-08, + 4.6566e-08, -4.1537e-07], + [ 6.5193e-09, 0.0000e+00, 4.3679e-07, ..., -2.2352e-08, + 9.1735e-07, 5.3085e-08], + [ 1.8626e-09, 0.0000e+00, 2.0582e-07, ..., 1.8626e-08, + 5.1595e-07, 6.5193e-09], + ..., + [ 5.5879e-09, 0.0000e+00, -4.1910e-06, ..., 9.3132e-09, + -9.3356e-06, -5.6159e-07], + [ 4.6566e-09, 0.0000e+00, 1.9465e-07, ..., 1.5274e-07, + 2.9802e-07, 3.2596e-08], + [ 1.7695e-08, 0.0000e+00, 2.9691e-06, ..., 8.3819e-09, + 6.8545e-06, 4.7963e-07]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0421, -0.0310, 0.0056, 0.0334, -0.0183, 0.0081, -0.0137, 0.0172, + 0.0133, -0.0027], device='cuda:0'), grad: tensor([-6.0163e-07, 2.5518e-06, 1.8515e-06, 7.6741e-07, 1.0133e-06, + 5.2992e-07, -4.3772e-08, -2.6122e-05, 1.2647e-06, 1.8761e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 214.44, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.4883 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.18 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.1050, 0.0564, 0.0702, ..., -0.0899, -0.1839, 0.0599], + [-0.0899, -0.0594, -0.1128, ..., 0.1831, -0.1038, -0.0770], + [-0.0380, 0.0041, 0.2132, ..., -0.1198, -0.1537, -0.1422], + ..., + [-0.1570, -0.0600, 0.0165, ..., -0.0560, 0.1242, -0.0959], + [-0.1096, -0.0957, -0.0783, ..., -0.1444, -0.0059, -0.1722], + [-0.1897, -0.0625, -0.1519, ..., -0.1967, -0.0380, 0.1464]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -1.5274e-07, -1.3411e-07, ..., -1.8626e-09, + 2.7008e-08, -2.4773e-07], + [ 3.9116e-08, 2.7940e-09, 1.6298e-07, ..., -8.3819e-09, + 2.5146e-08, 5.5879e-09], + [ 7.4506e-09, 3.7253e-09, -1.3812e-06, ..., 1.2107e-08, + 1.0058e-07, 1.3039e-08], + ..., + [ 6.5193e-09, 3.7253e-09, 1.9837e-07, ..., 3.7253e-09, + 3.9581e-07, 4.0047e-08], + [ 5.2154e-08, 2.5146e-08, 3.7905e-07, ..., 5.6811e-08, + 1.0617e-07, 5.4948e-08], + [ 1.2293e-07, 5.7742e-08, 1.5926e-07, ..., 1.5832e-08, + 1.2480e-07, 1.0617e-07]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0421, -0.0311, 0.0057, 0.0329, -0.0177, 0.0077, -0.0138, 0.0150, + 0.0133, -0.0003], device='cuda:0'), grad: tensor([-3.3714e-07, 4.6566e-07, -1.7835e-06, -4.8801e-07, -1.0151e-07, + -1.5525e-06, 5.4948e-08, 1.6475e-06, 1.2005e-06, 8.9128e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 214.49, cls_loss 0.0006 cls_loss_mapping 0.0022 cls_loss_causal 0.4930 re_mapping 0.0041 re_causal 0.0125 /// teacc 99.10 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.1050, 0.0564, 0.0702, ..., -0.0909, -0.1860, 0.0599], + [-0.0902, -0.0595, -0.1129, ..., 0.1831, -0.1042, -0.0797], + [-0.0392, 0.0044, 0.2136, ..., -0.1201, -0.1539, -0.1424], + ..., + [-0.1567, -0.0600, 0.0166, ..., -0.0562, 0.1243, -0.0959], + [-0.1098, -0.0957, -0.0784, ..., -0.1444, -0.0060, -0.1724], + [-0.1897, -0.0628, -0.1520, ..., -0.1951, -0.0380, 0.1464]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.9558e-08, + 1.8626e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 8.8476e-08, ..., -3.9581e-07, + 8.3819e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, -1.5041e-06, ..., 1.0617e-07, + 6.5193e-09, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.7497e-08, ..., 5.0291e-08, + -3.7253e-09, 2.1886e-07], + [ 9.3132e-10, 0.0000e+00, 9.6485e-07, ..., 9.2201e-08, + 1.5832e-08, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 2.8871e-08, ..., 1.1176e-08, + 5.1223e-08, -1.4715e-07]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0421, -0.0313, 0.0055, 0.0328, -0.0178, 0.0077, -0.0136, 0.0150, + 0.0133, -0.0003], device='cuda:0'), grad: tensor([ 1.4529e-07, -6.8732e-07, -2.4848e-06, 7.0129e-07, 1.2759e-07, + -4.9919e-07, 1.4808e-07, 7.7393e-07, 2.1160e-06, -3.4086e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.34, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4854 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.07 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.1051, 0.0564, 0.0703, ..., -0.0910, -0.1863, 0.0599], + [-0.0905, -0.0595, -0.1130, ..., 0.1831, -0.1051, -0.0798], + [-0.0396, 0.0044, 0.2146, ..., -0.1202, -0.1541, -0.1426], + ..., + [-0.1570, -0.0601, 0.0169, ..., -0.0561, 0.1243, -0.0959], + [-0.1098, -0.0958, -0.0785, ..., -0.1445, -0.0063, -0.1727], + [-0.1901, -0.0628, -0.1520, ..., -0.1950, -0.0380, 0.1465]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.6124e-08, ..., 4.6566e-09, + 1.3039e-08, -3.2596e-08], + [ 9.3132e-10, 0.0000e+00, 2.7008e-08, ..., -6.9849e-08, + 3.7253e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 4.0047e-08, ..., 1.8626e-08, + 5.1223e-08, 4.6566e-09], + ..., + [ 9.3132e-10, 0.0000e+00, -2.0768e-07, ..., 2.6077e-08, + -2.7753e-07, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.0245e-08, + -8.3819e-09, 4.6566e-09], + [ 6.5193e-09, 0.0000e+00, 4.5635e-08, ..., 9.3132e-10, + 5.0291e-08, -4.4703e-08]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0421, -0.0314, 0.0058, 0.0322, -0.0174, 0.0075, -0.0133, 0.0151, + 0.0133, -0.0004], device='cuda:0'), grad: tensor([-7.6368e-08, 4.6566e-08, 2.7940e-07, 3.2689e-07, 3.6322e-08, + 1.8720e-07, 7.8231e-08, -7.0687e-07, -2.9989e-07, 1.2480e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 214.21, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4856 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.04 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.1051, 0.0564, 0.0703, ..., -0.0908, -0.1856, 0.0599], + [-0.0908, -0.0595, -0.1153, ..., 0.1805, -0.1085, -0.0799], + [-0.0412, 0.0042, 0.2148, ..., -0.1205, -0.1544, -0.1427], + ..., + [-0.1571, -0.0610, 0.0195, ..., -0.0531, 0.1252, -0.0959], + [-0.1098, -0.0959, -0.0786, ..., -0.1445, -0.0064, -0.1730], + [-0.1903, -0.0632, -0.1520, ..., -0.1953, -0.0380, 0.1465]], + device='cuda:0'), grad: tensor([[ 1.1548e-07, 0.0000e+00, -2.1793e-07, ..., 1.8720e-07, + 3.7253e-09, -1.5460e-07], + [ 1.2200e-07, 0.0000e+00, 6.8918e-08, ..., -3.0156e-06, + 6.5193e-09, 9.3132e-10], + [ 1.8263e-06, 0.0000e+00, 6.6031e-07, ..., 2.2743e-06, + 1.5832e-08, 1.7695e-08], + ..., + [ 5.5879e-09, 0.0000e+00, -3.1292e-07, ..., 4.9267e-07, + -4.2282e-07, 0.0000e+00], + [ 1.1828e-07, 0.0000e+00, 1.5832e-08, ..., 1.4799e-06, + -5.2154e-08, 1.4901e-08], + [ 2.6077e-08, 0.0000e+00, 5.7742e-07, ..., 4.9360e-07, + 3.9116e-07, 6.7055e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0421, -0.0342, 0.0050, 0.0321, -0.0172, 0.0074, -0.0136, 0.0162, + 0.0135, -0.0004], device='cuda:0'), grad: tensor([ 1.0524e-07, -6.8732e-06, 6.6012e-06, -6.4261e-08, 7.4506e-06, + 5.1595e-07, -1.3947e-05, 4.7684e-07, 2.7530e-06, 2.9318e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 214.29, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.5131 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.15 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.1051, 0.0564, 0.0703, ..., -0.0907, -0.1860, 0.0600], + [-0.0911, -0.0595, -0.1155, ..., 0.1806, -0.1086, -0.0799], + [-0.0415, 0.0040, 0.2146, ..., -0.1213, -0.1547, -0.1434], + ..., + [-0.1576, -0.0614, 0.0197, ..., -0.0532, 0.1252, -0.0959], + [-0.1102, -0.0960, -0.0787, ..., -0.1445, -0.0064, -0.1736], + [-0.1904, -0.0636, -0.1520, ..., -0.1964, -0.0380, 0.1465]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -3.8464e-07, ..., 1.0245e-08, + 9.3132e-10, -6.6310e-07], + [ 2.4214e-08, 0.0000e+00, -6.7055e-08, ..., -4.4480e-06, + 9.3132e-09, 8.3819e-09], + [ 6.8918e-08, 0.0000e+00, 8.1025e-08, ..., 3.4589e-06, + 5.7742e-08, 8.3819e-09], + ..., + [ 1.3970e-08, 0.0000e+00, -6.0536e-08, ..., 7.5530e-07, + -9.8720e-08, 7.1712e-08], + [ 2.6077e-08, 0.0000e+00, 5.6811e-08, ..., 7.1712e-08, + 1.3970e-08, 3.0734e-08], + [ 6.4261e-08, 0.0000e+00, 2.6263e-07, ..., 2.2352e-08, + 1.2107e-08, 2.1141e-07]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0421, -0.0342, 0.0042, 0.0333, -0.0170, 0.0066, -0.0136, 0.0161, + 0.0135, -0.0004], device='cuda:0'), grad: tensor([-1.4314e-06, -9.6336e-06, 7.8008e-06, 3.7998e-07, -2.9802e-08, + 9.4995e-08, 2.9244e-07, 1.6773e-06, 1.0896e-07, 7.6275e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.45, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4744 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.21 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.1052, 0.0564, 0.0704, ..., -0.0907, -0.1862, 0.0600], + [-0.0914, -0.0595, -0.1155, ..., 0.1806, -0.1086, -0.0800], + [-0.0420, 0.0041, 0.2158, ..., -0.1220, -0.1550, -0.1436], + ..., + [-0.1579, -0.0616, 0.0195, ..., -0.0532, 0.1252, -0.0959], + [-0.1106, -0.0961, -0.0785, ..., -0.1446, -0.0048, -0.1738], + [-0.1905, -0.0643, -0.1521, ..., -0.1967, -0.0381, 0.1464]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, -6.8918e-08, -2.9430e-07, ..., 0.0000e+00, + 0.0000e+00, -1.8347e-07], + [ 4.6566e-09, 1.8626e-09, 1.1176e-08, ..., -5.3272e-07, + 0.0000e+00, 2.7008e-08], + [ 1.0245e-08, 1.8626e-09, 2.6077e-08, ..., 1.6950e-07, + 0.0000e+00, 2.7008e-08], + ..., + [ 1.8626e-09, 1.8626e-09, 5.5879e-09, ..., 2.5425e-07, + 0.0000e+00, 2.7008e-07], + [-2.3283e-08, 2.2352e-08, 5.3085e-08, ..., 2.2352e-08, + 1.8626e-09, 1.3784e-07], + [ 1.2107e-08, 3.9116e-08, 9.9652e-08, ..., 2.7940e-09, + 0.0000e+00, -3.9767e-07]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0421, -0.0342, 0.0043, 0.0322, -0.0171, 0.0068, -0.0134, 0.0161, + 0.0143, -0.0004], device='cuda:0'), grad: tensor([-5.1130e-07, -4.8149e-07, 5.7183e-07, -5.1968e-07, 4.9453e-07, + 5.5507e-07, 1.1362e-07, 1.3569e-06, -1.9465e-07, -1.3830e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.43, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4795 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.1053, 0.0565, 0.0704, ..., -0.0908, -0.1863, 0.0600], + [-0.0916, -0.0595, -0.1155, ..., 0.1807, -0.1086, -0.0801], + [-0.0422, 0.0041, 0.2160, ..., -0.1225, -0.1552, -0.1436], + ..., + [-0.1585, -0.0617, 0.0195, ..., -0.0533, 0.1252, -0.0959], + [-0.1105, -0.0962, -0.0787, ..., -0.1446, -0.0049, -0.1740], + [-0.1905, -0.0645, -0.1521, ..., -0.1968, -0.0381, 0.1464]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-09], + [ 3.7253e-09, 0.0000e+00, -4.6566e-09, ..., -5.0291e-08, + 1.8626e-09, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -3.4459e-08, ..., 2.4214e-08, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, -1.0245e-08, ..., 1.2107e-08, + -2.2352e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 7.4506e-09, + -0.0000e+00, 5.5879e-09], + [ 1.2107e-08, 0.0000e+00, 1.5832e-08, ..., 3.7253e-09, + 1.7695e-08, -1.0245e-08]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0421, -0.0342, 0.0039, 0.0323, -0.0170, 0.0067, -0.0133, 0.0161, + 0.0143, -0.0004], device='cuda:0'), grad: tensor([-3.7253e-09, -8.5682e-08, -2.0489e-08, -6.6124e-08, -1.7695e-08, + 3.6042e-07, -3.2596e-07, -8.3819e-09, 1.3225e-07, 4.2841e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 214.31, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.5057 re_mapping 0.0036 re_causal 0.0116 /// teacc 99.22 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.1054, 0.0565, 0.0704, ..., -0.0908, -0.1869, 0.0600], + [-0.0918, -0.0595, -0.1157, ..., 0.1808, -0.1086, -0.0801], + [-0.0423, 0.0041, 0.2160, ..., -0.1228, -0.1555, -0.1437], + ..., + [-0.1590, -0.0617, 0.0199, ..., -0.0534, 0.1252, -0.0959], + [-0.1108, -0.0963, -0.0788, ..., -0.1446, -0.0050, -0.1743], + [-0.1907, -0.0654, -0.1521, ..., -0.1977, -0.0381, 0.1464]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -4.4703e-08, ..., 1.8626e-08, + 4.2841e-08, -3.5390e-08], + [ 5.2154e-08, 0.0000e+00, 4.1910e-08, ..., 2.5053e-07, + 1.1548e-07, 4.7497e-08], + [ 1.1176e-08, 0.0000e+00, 2.7940e-09, ..., 5.8673e-08, + 5.5879e-09, 2.7940e-09], + ..., + [ 9.3132e-10, 0.0000e+00, -1.4901e-07, ..., 3.7253e-09, + -4.0326e-07, -1.6950e-07], + [ 1.6764e-08, 0.0000e+00, 1.2107e-08, ..., 1.2573e-07, + 6.4261e-08, 1.1176e-08], + [ 3.7253e-08, 0.0000e+00, 1.0990e-07, ..., 3.7253e-09, + 1.9744e-07, 1.1642e-07]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0420, -0.0341, 0.0036, 0.0324, -0.0167, 0.0062, -0.0125, 0.0161, + 0.0142, -0.0005], device='cuda:0'), grad: tensor([ 3.4459e-08, 8.0932e-07, 1.6578e-07, 2.4214e-08, 4.7497e-08, + 4.5821e-07, -1.5115e-06, -1.1921e-06, 3.8836e-07, 7.6834e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 214.15, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4711 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.11 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.1055, 0.0565, 0.0704, ..., -0.0900, -0.1860, 0.0601], + [-0.0919, -0.0595, -0.1158, ..., 0.1809, -0.1086, -0.0800], + [-0.0423, 0.0041, 0.2164, ..., -0.1230, -0.1557, -0.1438], + ..., + [-0.1591, -0.0618, 0.0199, ..., -0.0534, 0.1252, -0.0960], + [-0.1109, -0.0963, -0.0789, ..., -0.1447, -0.0054, -0.1753], + [-0.1908, -0.0655, -0.1521, ..., -0.1989, -0.0381, 0.1465]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 5.5879e-09, ..., 1.4901e-08, + 1.6764e-08, -3.7253e-09], + [ 1.6764e-08, 0.0000e+00, 4.8429e-08, ..., 7.0781e-08, + 5.5879e-08, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 7.4506e-08, ..., 2.2352e-08, + 9.1270e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, -2.5705e-07, ..., 2.4214e-08, + -3.1479e-07, 3.7253e-09], + [ 1.8626e-09, 0.0000e+00, 6.7055e-08, ..., 4.4703e-08, + 9.3132e-08, 3.7253e-09], + [ 4.4703e-08, 0.0000e+00, 3.7253e-08, ..., 3.7253e-09, + 4.4703e-08, -5.5879e-09]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0420, -0.0341, 0.0034, 0.0327, -0.0168, 0.0062, -0.0130, 0.0160, + 0.0140, -0.0004], device='cuda:0'), grad: tensor([ 1.3970e-07, 7.6555e-07, 3.8557e-07, -1.6596e-06, -1.0245e-07, + 1.6205e-07, -1.6764e-08, -4.5449e-07, 5.6811e-07, 1.9744e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 214.38, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4478 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.12 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.1055, 0.0566, 0.0705, ..., -0.0907, -0.1856, 0.0601], + [-0.0921, -0.0595, -0.1159, ..., 0.1809, -0.1086, -0.0800], + [-0.0424, 0.0041, 0.2165, ..., -0.1232, -0.1559, -0.1442], + ..., + [-0.1591, -0.0618, 0.0205, ..., -0.0535, 0.1253, -0.0961], + [-0.1112, -0.0965, -0.0793, ..., -0.1447, -0.0064, -0.1760], + [-0.1908, -0.0676, -0.1522, ..., -0.1992, -0.0381, 0.1465]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.4703e-08, ..., 3.7253e-09, + 3.7253e-09, -5.4017e-08], + [ 1.8626e-09, 0.0000e+00, 1.3784e-07, ..., -1.5832e-07, + 3.7253e-08, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, -5.9605e-07, ..., 2.0489e-08, + -4.2841e-08, 3.7253e-09], + ..., + [ 3.7253e-09, 0.0000e+00, 8.7544e-08, ..., 2.9802e-08, + -9.3132e-08, 4.0978e-08], + [ 1.8626e-09, 0.0000e+00, 2.1607e-07, ..., 5.5879e-08, + 1.6019e-07, 6.5193e-08], + [-6.4448e-07, 0.0000e+00, 7.8231e-08, ..., 9.3132e-09, + 1.0431e-07, -2.0843e-06]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0420, -0.0341, 0.0027, 0.0318, -0.0169, 0.0073, -0.0127, 0.0161, + 0.0140, -0.0004], device='cuda:0'), grad: tensor([-6.1467e-08, 2.7940e-08, -9.0711e-07, 6.1840e-07, 5.6140e-06, + -1.2685e-06, 4.5076e-07, 1.8254e-07, 7.2643e-07, -5.3719e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 214.08, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4736 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.11 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.1056, 0.0566, 0.0705, ..., -0.0909, -0.1858, 0.0601], + [-0.0924, -0.0595, -0.1160, ..., 0.1814, -0.1086, -0.0801], + [-0.0426, 0.0042, 0.2173, ..., -0.1235, -0.1561, -0.1443], + ..., + [-0.1591, -0.0619, 0.0208, ..., -0.0535, 0.1253, -0.0961], + [-0.1115, -0.0966, -0.0795, ..., -0.1456, -0.0068, -0.1763], + [-0.1910, -0.0681, -0.1522, ..., -0.1996, -0.0381, 0.1465]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -7.2271e-07, ..., 5.5879e-09, + -8.9407e-08, -4.6380e-07], + [ 3.7253e-09, 0.0000e+00, 3.3528e-08, ..., -1.8068e-07, + 1.4901e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -6.8918e-08, ..., 1.6764e-08, + 1.1176e-08, 1.8626e-08], + ..., + [ 1.8626e-09, 0.0000e+00, 1.1176e-08, ..., 5.7742e-08, + -7.2643e-08, 2.9802e-08], + [ 1.8626e-09, 0.0000e+00, 1.5087e-07, ..., 3.5390e-08, + 2.7940e-08, 1.0058e-07], + [ 4.4703e-08, 0.0000e+00, 3.2224e-07, ..., 5.5879e-09, + 4.6566e-08, 1.6578e-07]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0420, -0.0338, 0.0029, 0.0320, -0.0167, 0.0073, -0.0126, 0.0161, + 0.0131, -0.0005], device='cuda:0'), grad: tensor([-1.8440e-06, -3.2224e-07, 7.4506e-09, 3.8184e-07, 1.6205e-07, + 1.7323e-07, -2.2352e-08, 9.6858e-08, 4.9919e-07, 8.4750e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 214.12, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4895 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.09 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.1056, 0.0567, 0.0706, ..., -0.0909, -0.1862, 0.0601], + [-0.0926, -0.0596, -0.1160, ..., 0.1815, -0.1087, -0.0801], + [-0.0426, 0.0043, 0.2179, ..., -0.1238, -0.1561, -0.1445], + ..., + [-0.1591, -0.0620, 0.0208, ..., -0.0535, 0.1253, -0.0961], + [-0.1116, -0.0966, -0.0796, ..., -0.1457, -0.0069, -0.1766], + [-0.1911, -0.0694, -0.1522, ..., -0.2008, -0.0381, 0.1465]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -4.6566e-08, ..., 1.8626e-09, + 3.7253e-09, -6.1467e-08], + [ 9.1270e-08, 0.0000e+00, 3.1665e-08, ..., 3.7253e-09, + 4.2841e-08, 5.5879e-09], + [ 3.7253e-09, 0.0000e+00, -4.2841e-08, ..., 1.8626e-09, + 9.3132e-09, 1.8626e-09], + ..., + [ 4.8429e-08, 0.0000e+00, -9.3132e-08, ..., 1.4901e-08, + -1.9372e-07, -1.1176e-08], + [ 2.7940e-08, 0.0000e+00, -9.1270e-08, ..., 5.5879e-09, + -6.1467e-08, -7.4506e-09], + [ 3.5577e-07, 0.0000e+00, 1.7323e-07, ..., 5.5879e-08, + 1.8440e-07, 6.1467e-08]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0419, -0.0337, 0.0026, 0.0316, -0.0166, 0.0075, -0.0125, 0.0161, + 0.0133, -0.0005], device='cuda:0'), grad: tensor([-5.7742e-08, 3.0734e-07, 1.3597e-07, -1.6838e-06, -8.6613e-07, + 1.6112e-06, 4.2841e-08, 1.0990e-07, -1.2219e-06, 1.6317e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 214.17, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4826 re_mapping 0.0034 re_causal 0.0116 /// teacc 99.05 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.1058, 0.0571, 0.0706, ..., -0.0910, -0.1860, 0.0602], + [-0.0931, -0.0596, -0.1161, ..., 0.1815, -0.1087, -0.0803], + [-0.0434, 0.0042, 0.2196, ..., -0.1242, -0.1561, -0.1448], + ..., + [-0.1591, -0.0621, 0.0208, ..., -0.0536, 0.1253, -0.0961], + [-0.1118, -0.0967, -0.0799, ..., -0.1457, -0.0072, -0.1771], + [-0.1917, -0.0721, -0.1522, ..., -0.2012, -0.0381, 0.1465]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 1.8626e-09, + 0.0000e+00, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -2.9802e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -7.4506e-08, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 2.4214e-08, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 5.5879e-08, ..., -8.1956e-08, + 2.0489e-08, 1.6764e-08], + [ 5.5879e-09, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 1.8626e-09, 3.7253e-09]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0419, -0.0337, 0.0032, 0.0316, -0.0161, 0.0074, -0.0129, 0.0160, + 0.0133, -0.0005], device='cuda:0'), grad: tensor([ 1.4901e-08, 1.5460e-07, -2.6077e-08, -1.4901e-08, 0.0000e+00, + -6.8918e-08, 1.2033e-06, 1.1548e-07, -1.4547e-06, 4.6566e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 214.36, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.5091 re_mapping 0.0037 re_causal 0.0119 /// teacc 99.11 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.1059, 0.0575, 0.0708, ..., -0.0913, -0.1868, 0.0603], + [-0.0934, -0.0596, -0.1164, ..., 0.1819, -0.1087, -0.0806], + [-0.0437, 0.0043, 0.2205, ..., -0.1244, -0.1563, -0.1455], + ..., + [-0.1591, -0.0623, 0.0210, ..., -0.0540, 0.1254, -0.0961], + [-0.1124, -0.0972, -0.0801, ..., -0.1457, -0.0077, -0.1777], + [-0.1919, -0.0746, -0.1523, ..., -0.2020, -0.0381, 0.1464]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-08, ..., 5.5879e-09, + 0.0000e+00, -4.8429e-08], + [ 1.6764e-08, 0.0000e+00, 1.4901e-08, ..., -4.8429e-08, + 0.0000e+00, 1.1176e-08], + [ 1.8626e-09, 0.0000e+00, -3.9116e-08, ..., 1.4901e-08, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 9.3132e-09, ..., 2.6077e-08, + -1.8626e-09, 3.3528e-08], + [ 1.8626e-09, 0.0000e+00, 7.4506e-09, ..., 3.5390e-08, + -0.0000e+00, 1.3039e-08], + [ 2.2724e-07, 0.0000e+00, 2.0489e-08, ..., 9.3132e-09, + 1.8626e-09, -8.5682e-08]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0418, -0.0335, 0.0030, 0.0317, -0.0161, 0.0085, -0.0144, 0.0160, + 0.0132, -0.0006], device='cuda:0'), grad: tensor([-1.0803e-07, -4.8429e-08, -2.4214e-08, 1.0058e-07, -1.8626e-07, + 2.7567e-07, -2.6822e-07, 1.3970e-07, -6.8918e-08, 1.8254e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 214.66, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4920 re_mapping 0.0037 re_causal 0.0120 /// teacc 99.11 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.1060, 0.0575, 0.0713, ..., -0.0882, -0.1871, 0.0609], + [-0.0943, -0.0596, -0.1166, ..., 0.1820, -0.1087, -0.0809], + [-0.0422, 0.0045, 0.2213, ..., -0.1240, -0.1564, -0.1459], + ..., + [-0.1592, -0.0623, 0.0211, ..., -0.0541, 0.1254, -0.0961], + [-0.1131, -0.0972, -0.0803, ..., -0.1458, -0.0078, -0.1783], + [-0.1922, -0.0746, -0.1525, ..., -0.2041, -0.0381, 0.1463]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6263e-07, ..., -0.0000e+00, + 1.3039e-08, -2.2724e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -6.1467e-08, + 3.7253e-09, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 9.3132e-09, + 3.7253e-09, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -6.7055e-08, ..., 4.8429e-08, + -4.7497e-07, 8.9407e-08], + [-1.8626e-09, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 1.8626e-09, 1.3225e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-08, ..., 0.0000e+00, + 8.9407e-08, -5.0850e-07]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0412, -0.0336, 0.0035, 0.0317, -0.0160, 0.0084, -0.0171, 0.0160, + 0.0132, -0.0006], device='cuda:0'), grad: tensor([-4.5449e-07, -5.9605e-08, 5.0291e-08, 3.1665e-07, 1.1325e-06, + 6.9290e-07, 1.1176e-08, -5.1036e-07, 4.6194e-07, -1.6391e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 214.30, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4472 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.13 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.1062, 0.0576, 0.0713, ..., -0.0881, -0.1870, 0.0609], + [-0.0948, -0.0596, -0.1167, ..., 0.1820, -0.1087, -0.0809], + [-0.0423, 0.0046, 0.2242, ..., -0.1241, -0.1564, -0.1460], + ..., + [-0.1594, -0.0625, 0.0210, ..., -0.0541, 0.1254, -0.0961], + [-0.1141, -0.0972, -0.0817, ..., -0.1458, -0.0079, -0.1786], + [-0.1923, -0.0747, -0.1525, ..., -0.2043, -0.0381, 0.1463]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 3.7253e-09, + 0.0000e+00, -2.2352e-08], + [ 3.7253e-09, 0.0000e+00, 9.3132e-09, ..., -3.1665e-08, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.9558e-07, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 1.3039e-08, ..., 2.6077e-08, + -1.3039e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.1735e-07, ..., 2.9802e-08, + 3.7253e-09, 7.4506e-09], + [ 2.4214e-08, 0.0000e+00, 1.1176e-08, ..., 3.7253e-09, + 5.5879e-09, -1.3039e-08]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0412, -0.0335, 0.0054, 0.0315, -0.0159, 0.0081, -0.0168, 0.0160, + 0.0129, -0.0007], device='cuda:0'), grad: tensor([-3.3528e-08, -3.3528e-08, -3.3714e-07, 9.8720e-08, 7.4506e-09, + -9.3132e-09, -9.8720e-08, 8.5682e-08, 2.7753e-07, 1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 214.20, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.5004 re_mapping 0.0036 re_causal 0.0114 /// teacc 98.97 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.1063, 0.0576, 0.0713, ..., -0.0881, -0.1874, 0.0609], + [-0.0950, -0.0597, -0.1182, ..., 0.1820, -0.1087, -0.0811], + [-0.0425, 0.0046, 0.2267, ..., -0.1232, -0.1566, -0.1461], + ..., + [-0.1594, -0.0629, 0.0210, ..., -0.0541, 0.1254, -0.0962], + [-0.1144, -0.0973, -0.0818, ..., -0.1460, -0.0081, -0.1789], + [-0.1927, -0.0747, -0.1525, ..., -0.2046, -0.0381, 0.1465]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 3.7253e-08, ..., 7.4506e-09, + 1.8626e-09, -3.7253e-09], + [ 7.4506e-09, 0.0000e+00, 1.1455e-06, ..., 3.5018e-06, + 1.6950e-07, 5.1036e-06], + [ 3.7253e-09, 0.0000e+00, -3.5707e-06, ..., -5.2527e-07, + 8.1956e-08, 3.7253e-09], + ..., + [ 3.7253e-09, 0.0000e+00, 5.1036e-07, ..., 7.4506e-08, + -5.4948e-07, 8.1956e-08], + [-6.2212e-07, 0.0000e+00, 1.0692e-06, ..., 6.1467e-08, + 3.7253e-08, -4.9546e-07], + [ 3.7253e-08, 0.0000e+00, 9.6858e-08, ..., -6.0536e-06, + 1.2107e-07, -9.8348e-06]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0412, -0.0337, 0.0074, 0.0316, -0.0164, 0.0079, -0.0167, 0.0160, + 0.0128, -0.0005], device='cuda:0'), grad: tensor([ 1.0617e-07, 1.9118e-05, -5.5283e-06, 1.9185e-07, 1.5691e-05, + 3.3900e-07, 3.4217e-06, 8.1956e-08, -1.5758e-06, -3.1918e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 214.45, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4490 re_mapping 0.0037 re_causal 0.0111 /// teacc 98.94 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.1063, 0.0576, 0.0714, ..., -0.0881, -0.1876, 0.0610], + [-0.0953, -0.0597, -0.1189, ..., 0.1820, -0.1088, -0.0813], + [-0.0427, 0.0046, 0.2280, ..., -0.1232, -0.1569, -0.1466], + ..., + [-0.1595, -0.0630, 0.0187, ..., -0.0541, 0.1253, -0.0962], + [-0.1145, -0.0973, -0.0797, ..., -0.1464, -0.0054, -0.1792], + [-0.1915, -0.0747, -0.1526, ..., -0.2039, -0.0382, 0.1468]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.6368e-08, ..., 1.8626e-09, + 1.8626e-09, -7.6368e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -2.4214e-08, + 3.7253e-09, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 5.5879e-09, + 1.8626e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 9.3132e-09, + -1.1176e-08, 1.1176e-08], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 7.4506e-09, + 9.3132e-09, 1.6764e-08], + [ 3.7253e-09, 0.0000e+00, 5.0291e-08, ..., -3.7253e-09, + 1.4901e-08, 2.0489e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0411, -0.0339, 0.0083, 0.0315, -0.0183, 0.0077, -0.0161, 0.0158, + 0.0132, -0.0003], device='cuda:0'), grad: tensor([-1.2852e-07, 2.7940e-08, 2.4773e-07, -4.9174e-07, 2.0489e-08, + -1.6019e-07, 1.4529e-07, 6.8918e-08, 2.5705e-07, 1.4901e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 214.39, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.5165 re_mapping 0.0034 re_causal 0.0114 /// teacc 99.12 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.1064, 0.0576, 0.0713, ..., -0.0881, -0.1881, 0.0607], + [-0.0958, -0.0597, -0.1190, ..., 0.1821, -0.1088, -0.0814], + [-0.0440, 0.0046, 0.2285, ..., -0.1238, -0.1572, -0.1469], + ..., + [-0.1598, -0.0630, 0.0187, ..., -0.0542, 0.1254, -0.0962], + [-0.1148, -0.0973, -0.0798, ..., -0.1465, -0.0055, -0.1792], + [-0.1923, -0.0747, -0.1525, ..., -0.2041, -0.0382, 0.1471]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.1467e-08, -4.8056e-07, ..., 3.7253e-09, + 0.0000e+00, -1.0133e-06], + [ 1.8626e-09, 1.8626e-09, 9.3132e-09, ..., -6.8173e-07, + 1.8626e-09, 1.4901e-08], + [ 0.0000e+00, 1.8626e-09, -2.7381e-07, ..., 2.9989e-07, + 1.8626e-09, 3.5390e-08], + ..., + [ 0.0000e+00, 5.5879e-09, 7.0781e-08, ..., 3.1292e-07, + -5.5879e-09, 9.4995e-08], + [ 0.0000e+00, 3.7253e-09, 1.2666e-07, ..., 2.9802e-08, + 0.0000e+00, 5.9605e-08], + [ 9.3132e-09, 2.0489e-08, 2.8685e-07, ..., 1.6764e-08, + 1.8626e-09, 4.3213e-07]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0414, -0.0339, 0.0081, 0.0318, -0.0178, 0.0076, -0.0159, 0.0158, + 0.0130, -0.0002], device='cuda:0'), grad: tensor([-2.8759e-06, -1.6596e-06, 4.0606e-07, 1.0151e-06, 8.7544e-08, + 1.9185e-07, 3.7253e-08, 1.1101e-06, 4.1537e-07, 1.2554e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 214.40, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4731 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.1066, 0.0576, 0.0713, ..., -0.0881, -0.1882, 0.0607], + [-0.0959, -0.0597, -0.1190, ..., 0.1822, -0.1088, -0.0815], + [-0.0442, 0.0046, 0.2289, ..., -0.1240, -0.1577, -0.1472], + ..., + [-0.1599, -0.0630, 0.0188, ..., -0.0542, 0.1254, -0.0962], + [-0.1151, -0.0973, -0.0798, ..., -0.1466, -0.0056, -0.1796], + [-0.1924, -0.0748, -0.1525, ..., -0.2044, -0.0382, 0.1471]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.4703e-08, ..., 0.0000e+00, + 3.7253e-09, -4.8429e-08], + [ 1.8626e-09, 0.0000e+00, 2.4214e-08, ..., -2.2352e-08, + 2.6077e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.8231e-08, ..., 3.7253e-09, + 9.4995e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -2.0862e-07, ..., 1.1176e-08, + -2.0489e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 3.7253e-09, + 1.1176e-08, 3.7253e-09], + [ 9.3132e-09, 0.0000e+00, 9.6858e-08, ..., 0.0000e+00, + 5.7742e-08, 3.1665e-08]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0414, -0.0337, 0.0081, 0.0304, -0.0178, 0.0092, -0.0158, 0.0158, + 0.0129, -0.0002], device='cuda:0'), grad: tensor([-8.9407e-08, 3.3528e-08, 1.2293e-07, 1.7881e-07, 1.3039e-08, + 4.4703e-08, 9.3132e-09, -3.8184e-07, -1.6205e-07, 2.2538e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 214.21, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.4725 re_mapping 0.0035 re_causal 0.0110 /// teacc 99.02 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.1068, 0.0576, 0.0713, ..., -0.0887, -0.1887, 0.0606], + [-0.0962, -0.0597, -0.1192, ..., 0.1826, -0.1088, -0.0817], + [-0.0444, 0.0046, 0.2287, ..., -0.1251, -0.1599, -0.1474], + ..., + [-0.1600, -0.0630, 0.0193, ..., -0.0546, 0.1255, -0.0962], + [-0.1155, -0.0973, -0.0799, ..., -0.1466, -0.0057, -0.1800], + [-0.1925, -0.0748, -0.1526, ..., -0.2072, -0.0382, 0.1471]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 0.0000e+00, -5.2899e-07, ..., 9.3132e-09, + 5.5879e-09, -4.3772e-07], + [ 5.9605e-08, 0.0000e+00, 1.6764e-08, ..., -1.3039e-08, + 2.9802e-08, 5.5879e-09], + [ 6.1467e-08, 0.0000e+00, 7.6368e-08, ..., 2.2352e-08, + 2.4214e-08, 7.8231e-08], + ..., + [ 6.3330e-08, 0.0000e+00, -3.7253e-08, ..., 1.3039e-08, + -3.3528e-08, 9.3132e-09], + [ 3.9116e-08, 0.0000e+00, 4.8429e-08, ..., 2.9802e-08, + 6.1467e-08, 4.0978e-08], + [ 1.0245e-07, 0.0000e+00, 1.8068e-07, ..., 1.8626e-09, + 4.4703e-08, 2.0489e-08]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0416, -0.0335, 0.0074, 0.0316, -0.0178, 0.0082, -0.0142, 0.0157, + 0.0129, -0.0003], device='cuda:0'), grad: tensor([-1.2945e-06, 1.3597e-07, 3.9116e-07, 1.4529e-07, -4.6939e-07, + 3.3900e-07, -2.7381e-07, 6.1467e-08, 4.8801e-07, 4.5449e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 214.12, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4944 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.12 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.1071, 0.0576, 0.0718, ..., -0.0886, -0.1846, 0.0612], + [-0.0965, -0.0597, -0.1195, ..., 0.1826, -0.1089, -0.0819], + [-0.0447, 0.0046, 0.2285, ..., -0.1254, -0.1611, -0.1476], + ..., + [-0.1601, -0.0630, 0.0200, ..., -0.0547, 0.1256, -0.0962], + [-0.1157, -0.0973, -0.0797, ..., -0.1466, -0.0054, -0.1790], + [-0.1926, -0.0748, -0.1531, ..., -0.2073, -0.0386, 0.1466]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 1.3039e-08, ..., 2.2352e-08, + 3.7253e-09, 2.4214e-08], + [ 2.2352e-08, 0.0000e+00, 2.6077e-08, ..., -1.5907e-06, + 1.8626e-09, 3.7253e-09], + [ 1.4901e-08, 0.0000e+00, -3.1292e-07, ..., 4.6007e-07, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 1.6764e-08, ..., 3.6508e-07, + 1.8626e-09, 1.0431e-07], + [ 1.8626e-09, 0.0000e+00, 4.6566e-08, ..., 1.9744e-07, + 3.7253e-09, 1.8626e-08], + [ 3.1665e-08, 0.0000e+00, 7.4506e-09, ..., 5.7742e-08, + -1.1176e-08, -1.7323e-07]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0411, -0.0335, 0.0069, 0.0316, -0.0178, 0.0079, -0.0143, 0.0158, + 0.0131, -0.0007], device='cuda:0'), grad: tensor([ 2.1607e-07, -3.7905e-06, 4.7684e-07, -8.7544e-08, -2.3730e-06, + 4.4145e-07, 3.4273e-06, 1.1548e-06, 6.1467e-07, -8.3819e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 214.18, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4343 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.10 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.1073, 0.0576, 0.0718, ..., -0.0887, -0.1846, 0.0612], + [-0.0967, -0.0597, -0.1195, ..., 0.1837, -0.1090, -0.0819], + [-0.0449, 0.0046, 0.2289, ..., -0.1268, -0.1619, -0.1475], + ..., + [-0.1603, -0.0630, 0.0212, ..., -0.0547, 0.1257, -0.0962], + [-0.1162, -0.0973, -0.0808, ..., -0.1468, -0.0066, -0.1792], + [-0.1927, -0.0748, -0.1531, ..., -0.2084, -0.0386, 0.1466]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 1.8626e-09, + 1.8626e-09, -2.1048e-07], + [ 7.4506e-09, 0.0000e+00, 5.5879e-09, ..., 3.1665e-08, + 1.6764e-08, 1.3039e-08], + [ 1.8626e-09, 0.0000e+00, -5.5879e-09, ..., 1.3039e-08, + 5.4017e-08, 4.2841e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + -1.3039e-08, 9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., -2.5798e-06, + 7.4506e-09, 1.8626e-08], + [ 4.2841e-08, 0.0000e+00, 9.3132e-09, ..., 1.8626e-09, + 1.3039e-08, 4.4703e-08]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0411, -0.0328, 0.0066, 0.0316, -0.0177, 0.0081, -0.0156, 0.0159, + 0.0126, -0.0007], device='cuda:0'), grad: tensor([-2.3469e-07, 3.2969e-07, 4.3213e-07, 3.2410e-07, -7.2643e-08, + -5.5134e-07, 1.2837e-05, 7.4506e-09, -1.3247e-05, 1.7881e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 214.57, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4560 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.00 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.1076, 0.0576, 0.0719, ..., -0.0888, -0.1846, 0.0612], + [-0.0969, -0.0597, -0.1205, ..., 0.1838, -0.1090, -0.0823], + [-0.0451, 0.0046, 0.2313, ..., -0.1275, -0.1622, -0.1479], + ..., + [-0.1604, -0.0630, 0.0204, ..., -0.0548, 0.1258, -0.0962], + [-0.1165, -0.0973, -0.0808, ..., -0.1469, -0.0065, -0.1795], + [-0.1930, -0.0748, -0.1532, ..., -0.2086, -0.0387, 0.1466]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.0489e-06, ..., 8.1956e-08, + 0.0000e+00, 7.8976e-07], + [ 0.0000e+00, 0.0000e+00, 1.1921e-07, ..., -1.4342e-07, + 2.7008e-07, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, -4.2692e-06, ..., 1.3411e-07, + 3.7253e-09, -1.6745e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-08, ..., 9.3132e-09, + -4.7497e-07, 6.5193e-08], + [ 1.8626e-09, 0.0000e+00, 1.0394e-06, ..., 1.3970e-07, + 9.3132e-09, 4.5635e-07], + [ 0.0000e+00, 0.0000e+00, 1.8999e-07, ..., 3.7253e-09, + 1.5460e-07, 5.5879e-09]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0411, -0.0330, 0.0080, 0.0313, -0.0183, 0.0080, -0.0149, 0.0158, + 0.0127, -0.0007], device='cuda:0'), grad: tensor([ 4.0755e-06, 4.6566e-08, -7.0632e-06, 1.0394e-06, 1.9372e-07, + 4.3958e-07, -6.5751e-07, -6.3889e-07, 2.2966e-06, 2.5518e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 214.44, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4651 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.07 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.1078, 0.0576, 0.0719, ..., -0.0889, -0.1846, 0.0613], + [-0.0971, -0.0597, -0.1207, ..., 0.1839, -0.1090, -0.0824], + [-0.0453, 0.0046, 0.2318, ..., -0.1277, -0.1627, -0.1482], + ..., + [-0.1605, -0.0630, 0.0204, ..., -0.0548, 0.1258, -0.0962], + [-0.1168, -0.0973, -0.0809, ..., -0.1472, -0.0069, -0.1803], + [-0.1933, -0.0748, -0.1532, ..., -0.2088, -0.0387, 0.1466]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 1.3039e-08, + 1.8626e-09, -9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 1.3039e-08, ..., -1.4901e-08, + 3.3528e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 5.5879e-09, 1.8626e-09], + ..., + [ 1.8626e-09, 0.0000e+00, -7.0781e-08, ..., 1.1176e-08, + -1.5274e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.8626e-08, + 4.6566e-08, 3.1665e-08], + [ 3.5390e-08, 0.0000e+00, 4.2841e-08, ..., 1.8626e-09, + 9.6858e-08, 0.0000e+00]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0410, -0.0330, 0.0081, 0.0308, -0.0182, 0.0071, -0.0136, 0.0158, + 0.0123, -0.0007], device='cuda:0'), grad: tensor([ 6.4634e-07, 3.3341e-07, 1.1362e-07, 2.9430e-06, 1.8626e-08, + -6.5565e-07, 6.5379e-07, -1.9744e-07, -4.2953e-06, 4.1164e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 214.33, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4687 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.07 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.1107, 0.0576, 0.0719, ..., -0.0889, -0.1846, 0.0613], + [-0.0985, -0.0597, -0.1210, ..., 0.1840, -0.1091, -0.0829], + [-0.0461, 0.0046, 0.2315, ..., -0.1286, -0.1644, -0.1486], + ..., + [-0.1605, -0.0630, 0.0213, ..., -0.0549, 0.1259, -0.0963], + [-0.1172, -0.0973, -0.0810, ..., -0.1472, -0.0070, -0.1808], + [-0.1940, -0.0748, -0.1532, ..., -0.2081, -0.0387, 0.1467]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., 3.7253e-09, + 1.8626e-09, 5.0291e-08], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 0.0000e+00, + 4.0978e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + 7.4506e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., 1.8626e-09, + -5.0291e-08, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.3039e-08, + 2.6077e-08, 4.4703e-08], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 4.4703e-08, -8.9407e-08]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0411, -0.0331, 0.0075, 0.0306, -0.0180, 0.0070, -0.0135, 0.0159, + 0.0125, -0.0007], device='cuda:0'), grad: tensor([ 3.7123e-06, 1.6205e-07, 7.2643e-08, -5.9679e-06, 2.1979e-07, + -1.4156e-07, 9.6858e-08, 3.7253e-09, 1.9018e-06, -6.5193e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 214.53, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4840 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.1131, 0.0576, 0.0720, ..., -0.0890, -0.1847, 0.0613], + [-0.1014, -0.0597, -0.1213, ..., 0.1841, -0.1092, -0.0831], + [-0.0477, 0.0046, 0.2324, ..., -0.1291, -0.1649, -0.1490], + ..., + [-0.1607, -0.0630, 0.0210, ..., -0.0549, 0.1260, -0.0963], + [-0.1178, -0.0973, -0.0811, ..., -0.1473, -0.0070, -0.1812], + [-0.1941, -0.0748, -0.1532, ..., -0.2082, -0.0387, 0.1467]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -1.2871e-06, ..., 1.4901e-08, + 1.3039e-08, -1.4696e-06], + [ 1.8626e-09, 0.0000e+00, 2.0489e-08, ..., -5.4017e-08, + 1.1176e-08, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 7.4506e-09, + 1.3039e-08, 2.0489e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., 2.0489e-08, + -4.4703e-08, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 5.0291e-08, ..., 1.1176e-08, + 3.7253e-09, 5.4017e-08], + [ 5.5879e-09, 0.0000e+00, 4.0233e-07, ..., 3.7253e-09, + 1.1176e-08, 4.5821e-07]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0411, -0.0332, 0.0077, 0.0303, -0.0176, 0.0071, -0.0134, 0.0158, + 0.0125, -0.0007], device='cuda:0'), grad: tensor([-2.7306e-06, -2.2352e-08, 1.0990e-07, 3.0361e-07, 7.4506e-09, + 1.8999e-07, 1.3597e-06, -2.9802e-08, -1.0803e-07, 9.1642e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 214.41, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.5106 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.07 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.1133, 0.0576, 0.0720, ..., -0.0890, -0.1847, 0.0614], + [-0.1018, -0.0597, -0.1214, ..., 0.1841, -0.1092, -0.0833], + [-0.0477, 0.0046, 0.2327, ..., -0.1293, -0.1651, -0.1493], + ..., + [-0.1608, -0.0630, 0.0209, ..., -0.0550, 0.1260, -0.0963], + [-0.1183, -0.0973, -0.0813, ..., -0.1473, -0.0070, -0.1820], + [-0.1944, -0.0748, -0.1533, ..., -0.2090, -0.0387, 0.1468]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, -1.1735e-07, ..., 7.4506e-09, + 0.0000e+00, -8.3819e-08], + [ 1.2852e-07, 0.0000e+00, 3.5390e-08, ..., -2.4214e-07, + 2.9802e-08, 3.7253e-08], + [ 1.8626e-08, 0.0000e+00, -3.5390e-08, ..., 2.7940e-08, + 2.6077e-08, 5.5879e-09], + ..., + [ 5.0291e-08, 0.0000e+00, -1.0245e-07, ..., 9.1270e-08, + -1.4901e-07, 4.8429e-08], + [ 1.4901e-08, 0.0000e+00, 1.8626e-08, ..., 9.3132e-08, + 5.5879e-09, 9.8720e-08], + [ 1.2964e-06, 0.0000e+00, 1.4715e-07, ..., 3.5390e-08, + 7.2643e-08, -2.4587e-07]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0411, -0.0332, 0.0078, 0.0303, -0.0175, 0.0069, -0.0135, 0.0158, + 0.0125, -0.0007], device='cuda:0'), grad: tensor([-1.5646e-07, -2.0303e-07, 1.0990e-07, 1.3039e-07, -2.5071e-06, + 2.3283e-07, 1.6205e-07, 9.6858e-08, 3.4273e-07, 1.7788e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 215.90, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4903 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.03 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.1134, 0.0576, 0.0721, ..., -0.0891, -0.1847, 0.0614], + [-0.1021, -0.0597, -0.1214, ..., 0.1843, -0.1093, -0.0835], + [-0.0481, 0.0046, 0.2330, ..., -0.1309, -0.1655, -0.1496], + ..., + [-0.1610, -0.0630, 0.0210, ..., -0.0550, 0.1264, -0.0964], + [-0.1191, -0.0973, -0.0816, ..., -0.1474, -0.0072, -0.1831], + [-0.1947, -0.0748, -0.1533, ..., -0.2091, -0.0388, 0.1467]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 2.0862e-07, ..., 2.3283e-08, + 8.1584e-06, 1.6123e-05], + [ 0.0000e+00, 0.0000e+00, 3.9861e-07, ..., 1.8626e-09, + 2.3283e-08, 5.1223e-08], + [ 0.0000e+00, 0.0000e+00, -2.3078e-06, ..., -7.1712e-08, + 1.0245e-08, 2.1420e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.4063e-06, ..., 6.0536e-08, + 6.0536e-08, 1.3318e-07], + [ 0.0000e+00, 0.0000e+00, 3.1944e-07, ..., 3.1665e-08, + 3.7253e-08, 8.1025e-08], + [ 9.3132e-10, 1.8626e-09, 1.0803e-07, ..., 1.8626e-09, + 1.0598e-06, 2.1327e-06]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0410, -0.0331, 0.0076, 0.0302, -0.0186, 0.0071, -0.0135, 0.0161, + 0.0125, -0.0007], device='cuda:0'), grad: tensor([ 2.4155e-05, 1.3048e-06, -6.9104e-06, 5.0291e-07, 4.1723e-07, + -3.0160e-05, 1.9055e-06, 4.4964e-06, 1.0366e-06, 3.2913e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 216.37, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4619 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.10 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.1134, 0.0576, 0.0722, ..., -0.0891, -0.1850, 0.0615], + [-0.1021, -0.0598, -0.1216, ..., 0.1844, -0.1093, -0.0841], + [-0.0482, 0.0057, 0.2365, ..., -0.1311, -0.1629, -0.1498], + ..., + [-0.1610, -0.0642, 0.0188, ..., -0.0551, 0.1263, -0.0964], + [-0.1194, -0.0976, -0.0824, ..., -0.1474, -0.0073, -0.1838], + [-0.1948, -0.0748, -0.1534, ..., -0.2093, -0.0388, 0.1467]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.4214e-08, ..., 6.5193e-09, + 2.5146e-08, 1.1502e-06], + [ 4.6566e-09, 0.0000e+00, -2.2165e-07, ..., -9.5181e-07, + 9.3132e-09, 1.6764e-07], + [ 1.8626e-09, 0.0000e+00, 4.4703e-08, ..., 1.9651e-07, + 8.3819e-09, 4.0978e-08], + ..., + [ 5.5879e-09, 0.0000e+00, 1.4715e-07, ..., 6.5472e-07, + 9.3132e-10, 1.4259e-06], + [ 1.1176e-08, 0.0000e+00, 1.3970e-08, ..., 4.0978e-08, + 5.4948e-08, 1.1735e-07], + [ 8.0094e-08, 0.0000e+00, 1.7695e-08, ..., 9.3132e-09, + 1.1828e-07, -5.0999e-06]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0410, -0.0331, 0.0109, 0.0301, -0.0186, 0.0076, -0.0137, 0.0159, + 0.0121, -0.0008], device='cuda:0'), grad: tensor([ 7.9870e-06, -1.5125e-06, 6.2864e-07, -4.8690e-06, 1.1563e-05, + -1.8692e-06, 5.1036e-07, 7.3239e-06, 6.8638e-07, -2.0504e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 218.67, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4621 re_mapping 0.0036 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.1135, 0.0576, 0.0722, ..., -0.0892, -0.1850, 0.0615], + [-0.1023, -0.0598, -0.1244, ..., 0.1825, -0.1112, -0.0842], + [-0.0482, 0.0058, 0.2365, ..., -0.1314, -0.1631, -0.1501], + ..., + [-0.1612, -0.0644, 0.0214, ..., -0.0529, 0.1270, -0.0964], + [-0.1198, -0.0977, -0.0825, ..., -0.1475, -0.0075, -0.1841], + [-0.1950, -0.0748, -0.1535, ..., -0.2098, -0.0388, 0.1467]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + 9.3132e-10, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 6.3330e-08, ..., -2.7008e-08, + 9.3132e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 5.4389e-07, ..., 1.0245e-08, + 4.0513e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -7.4971e-07, ..., 8.3819e-09, + -4.8429e-07, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 6.5193e-09, + 1.9558e-08, 2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 6.3330e-08, ..., 1.8626e-09, + 3.9116e-08, -6.5193e-08]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0410, -0.0352, 0.0107, 0.0299, -0.0185, 0.0079, -0.0138, 0.0170, + 0.0120, -0.0008], device='cuda:0'), grad: tensor([ 1.7695e-08, 9.3132e-08, 8.8569e-07, -3.5390e-08, 1.1176e-07, + 5.1223e-08, 5.5879e-09, -1.0850e-06, 5.0291e-08, -8.6613e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 215.04, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4690 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.14 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.1135, 0.0576, 0.0723, ..., -0.0892, -0.1850, 0.0616], + [-0.1024, -0.0598, -0.1245, ..., 0.1828, -0.1113, -0.0844], + [-0.0481, 0.0058, 0.2361, ..., -0.1334, -0.1643, -0.1502], + ..., + [-0.1613, -0.0644, 0.0222, ..., -0.0531, 0.1271, -0.0964], + [-0.1199, -0.0977, -0.0826, ..., -0.1475, -0.0077, -0.1843], + [-0.1951, -0.0748, -0.1535, ..., -0.2100, -0.0388, 0.1466]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -1.3970e-08, ..., 8.3819e-09, + 1.8626e-09, -1.3039e-08], + [ 2.7940e-08, 0.0000e+00, 3.8184e-08, ..., -1.8626e-09, + 1.8626e-08, 3.7253e-09], + [ 2.7940e-09, 0.0000e+00, -3.1013e-07, ..., 3.1665e-08, + 1.8626e-09, 1.8626e-09], + ..., + [ 4.6566e-09, 0.0000e+00, 1.4715e-07, ..., 3.7253e-09, + -6.2399e-08, -2.7940e-09], + [ 4.6566e-09, 0.0000e+00, 3.2596e-08, ..., 1.8626e-08, + 5.5879e-09, 1.8626e-09], + [ 1.2033e-06, 0.0000e+00, 6.1467e-08, ..., 0.0000e+00, + 7.1712e-08, -5.5879e-09]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0409, -0.0350, 0.0098, 0.0298, -0.0185, 0.0076, -0.0138, 0.0170, + 0.0121, -0.0009], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.6205e-07, -5.4203e-07, -5.9605e-08, -2.4252e-06, + 7.9162e-08, -1.4901e-07, 3.3993e-07, 1.4249e-07, 2.4512e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 215.27, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4654 re_mapping 0.0035 re_causal 0.0110 /// teacc 99.22 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.1136, 0.0576, 0.0725, ..., -0.0892, -0.1851, 0.0617], + [-0.1028, -0.0598, -0.1246, ..., 0.1829, -0.1113, -0.0848], + [-0.0483, 0.0058, 0.2363, ..., -0.1336, -0.1643, -0.1506], + ..., + [-0.1614, -0.0644, 0.0221, ..., -0.0532, 0.1272, -0.0964], + [-0.1202, -0.0977, -0.0827, ..., -0.1476, -0.0078, -0.1849], + [-0.1967, -0.0748, -0.1536, ..., -0.2104, -0.0388, 0.1466]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 3.7253e-09, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -2.7940e-09, + 7.4506e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -2.7008e-08, ..., 9.3132e-10, + 7.4506e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -8.0094e-08, ..., 1.8626e-09, + -8.7544e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 2.7940e-09, + 1.1176e-08, 3.7253e-09], + [ 1.8626e-09, 0.0000e+00, 6.8918e-08, ..., 0.0000e+00, + 5.1223e-08, -9.3132e-09]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0408, -0.0349, 0.0098, 0.0298, -0.0178, 0.0074, -0.0138, 0.0170, + 0.0120, -0.0010], device='cuda:0'), grad: tensor([-9.3132e-10, 3.1665e-08, -2.0489e-08, 2.1420e-08, 3.0734e-08, + 2.9802e-08, -1.6764e-08, -2.0023e-07, 7.4506e-09, 1.0990e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 214.45, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4786 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.11 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.1137, 0.0576, 0.0725, ..., -0.0893, -0.1852, 0.0618], + [-0.1030, -0.0598, -0.1247, ..., 0.1829, -0.1114, -0.0850], + [-0.0486, 0.0059, 0.2368, ..., -0.1339, -0.1644, -0.1506], + ..., + [-0.1616, -0.0645, 0.0222, ..., -0.0532, 0.1273, -0.0965], + [-0.1207, -0.0977, -0.0825, ..., -0.1477, -0.0080, -0.1853], + [-0.1971, -0.0748, -0.1538, ..., -0.2109, -0.0389, 0.1466]], + device='cuda:0'), grad: tensor([[ 4.7497e-08, 0.0000e+00, -2.8871e-08, ..., 2.6729e-07, + 9.3132e-10, -1.5274e-07], + [ 7.3947e-07, 0.0000e+00, 2.8592e-07, ..., -4.1910e-08, + 7.4506e-09, 2.6450e-07], + [ 1.1176e-08, 0.0000e+00, -2.8554e-06, ..., 2.7008e-08, + 1.8626e-09, 1.3039e-08], + ..., + [ 3.9116e-08, 0.0000e+00, 4.3586e-07, ..., 9.4995e-08, + -2.9802e-08, 2.0489e-08], + [ 3.1665e-08, 0.0000e+00, 1.3774e-06, ..., 1.5832e-07, + 6.5193e-09, 7.0781e-08], + [ 2.0582e-06, 0.0000e+00, 1.3225e-07, ..., 5.5321e-07, + 1.4901e-08, 3.6042e-07]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0408, -0.0350, 0.0097, 0.0295, -0.0174, 0.0075, -0.0139, 0.0170, + 0.0122, -0.0011], device='cuda:0'), grad: tensor([ 7.6648e-07, 1.4883e-06, -5.3383e-06, 1.1548e-07, -4.9770e-06, + 4.4517e-07, -1.5432e-06, 1.3597e-06, 3.3956e-06, 4.2468e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 214.69, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4689 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.05 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.1138, 0.0576, 0.0726, ..., -0.0893, -0.1852, 0.0618], + [-0.1036, -0.0598, -0.1247, ..., 0.1830, -0.1114, -0.0853], + [-0.0489, 0.0059, 0.2373, ..., -0.1342, -0.1644, -0.1511], + ..., + [-0.1619, -0.0645, 0.0223, ..., -0.0533, 0.1274, -0.0965], + [-0.1213, -0.0977, -0.0828, ..., -0.1478, -0.0081, -0.1860], + [-0.1973, -0.0748, -0.1538, ..., -0.2114, -0.0389, 0.1466]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.0827e-07, ..., 9.3132e-10, + 1.0245e-08, 6.6962e-07], + [ 0.0000e+00, 0.0000e+00, -6.6496e-07, ..., -1.1642e-06, + 1.3039e-08, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 2.2165e-07, ..., 9.9652e-08, + 3.7253e-09, 3.4086e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5786e-07, ..., 1.0412e-06, + -1.3597e-07, 1.1176e-07], + [ 0.0000e+00, 0.0000e+00, 3.6322e-08, ..., 5.5879e-09, + 2.7008e-08, 1.0058e-07], + [ 0.0000e+00, 0.0000e+00, -6.7521e-07, ..., 1.8626e-09, + 1.2480e-07, -1.6009e-06]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0408, -0.0349, 0.0099, 0.0296, -0.0173, 0.0074, -0.0139, 0.0170, + 0.0121, -0.0012], device='cuda:0'), grad: tensor([ 1.4789e-06, -2.4680e-06, 9.8161e-07, 2.1458e-06, 6.9849e-08, + -1.6354e-06, 1.0896e-07, 2.3190e-06, 2.3842e-07, -3.2578e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 214.51, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4901 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.18 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.1142, 0.0576, 0.0727, ..., -0.0894, -0.1852, 0.0619], + [-0.1041, -0.0598, -0.1248, ..., 0.1831, -0.1114, -0.0854], + [-0.0496, 0.0061, 0.2373, ..., -0.1343, -0.1646, -0.1518], + ..., + [-0.1622, -0.0645, 0.0228, ..., -0.0533, 0.1275, -0.0965], + [-0.1216, -0.0977, -0.0830, ..., -0.1478, -0.0084, -0.1865], + [-0.1975, -0.0748, -0.1540, ..., -0.2116, -0.0390, 0.1466]], + device='cuda:0'), grad: tensor([[ 8.6613e-08, 0.0000e+00, -2.8126e-07, ..., -1.8626e-09, + 5.5879e-09, -2.3004e-07], + [ 1.0896e-07, 0.0000e+00, 1.0962e-06, ..., 1.0896e-07, + 8.6986e-07, 2.8871e-08], + [ 4.6566e-09, 0.0000e+00, 5.5879e-08, ..., 3.5390e-08, + 2.0489e-08, 2.7940e-08], + ..., + [ 1.4901e-08, 0.0000e+00, -1.1595e-06, ..., -1.8813e-07, + -9.6112e-07, 4.6566e-09], + [ 1.0245e-08, 0.0000e+00, 6.0536e-08, ..., 2.9802e-08, + 5.5879e-09, 3.4459e-08], + [ 5.2247e-07, 0.0000e+00, 1.0058e-07, ..., 2.0489e-08, + 1.5832e-08, -1.8626e-09]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0407, -0.0349, 0.0090, 0.0301, -0.0174, 0.0072, -0.0137, 0.0171, + 0.0120, -0.0012], device='cuda:0'), grad: tensor([-5.0943e-07, 3.1050e-06, 2.1886e-07, 2.5146e-08, -1.1595e-06, + 3.1758e-07, 2.2445e-07, -3.3472e-06, 2.4401e-07, 8.6706e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 214.55, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4877 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.23 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.1148, 0.0576, 0.0727, ..., -0.0894, -0.1853, 0.0619], + [-0.1042, -0.0599, -0.1248, ..., 0.1834, -0.1115, -0.0855], + [-0.0497, 0.0062, 0.2375, ..., -0.1355, -0.1647, -0.1520], + ..., + [-0.1623, -0.0646, 0.0229, ..., -0.0536, 0.1276, -0.0965], + [-0.1220, -0.0977, -0.0831, ..., -0.1480, -0.0086, -0.1866], + [-0.1978, -0.0748, -0.1540, ..., -0.2123, -0.0391, 0.1466]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -6.3330e-08, ..., -1.3039e-08, + 2.6077e-08, -8.6613e-08], + [ 3.2596e-08, 0.0000e+00, 4.6566e-08, ..., -1.1176e-08, + 9.3132e-08, 1.3039e-08], + [ 9.3132e-10, 0.0000e+00, 1.3970e-08, ..., 3.7253e-09, + 2.2352e-08, 3.7253e-09], + ..., + [ 7.4506e-09, 0.0000e+00, -9.8906e-07, ..., 7.4506e-09, + -1.9632e-06, 1.5832e-08], + [ 2.7940e-09, 0.0000e+00, 2.9802e-08, ..., 2.7940e-09, + 5.2154e-08, 4.1910e-08], + [ 2.6077e-08, 0.0000e+00, 6.1840e-07, ..., 1.8626e-09, + 1.1278e-06, -3.2410e-07]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0407, -0.0347, 0.0084, 0.0306, -0.0173, 0.0067, -0.0133, 0.0170, + 0.0121, -0.0013], device='cuda:0'), grad: tensor([-1.7229e-07, 3.1292e-07, 8.3819e-08, 4.6007e-07, 9.7789e-08, + 1.3085e-06, 1.2480e-07, -4.6901e-06, 2.1514e-07, 2.2482e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 214.31, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4859 re_mapping 0.0035 re_causal 0.0110 /// teacc 99.03 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.1150, 0.0576, 0.0728, ..., -0.0892, -0.1853, 0.0620], + [-0.1047, -0.0599, -0.1250, ..., 0.1835, -0.1116, -0.0884], + [-0.0498, 0.0063, 0.2382, ..., -0.1357, -0.1649, -0.1523], + ..., + [-0.1626, -0.0647, 0.0236, ..., -0.0537, 0.1280, -0.0966], + [-0.1221, -0.0977, -0.0835, ..., -0.1480, -0.0089, -0.1872], + [-0.1979, -0.0748, -0.1543, ..., -0.2094, -0.0393, 0.1467]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 7.4506e-09, ..., 5.8673e-08, + 1.3039e-08, -2.7940e-09], + [ 8.3819e-09, 0.0000e+00, 1.1921e-07, ..., -9.1270e-08, + 8.4750e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.5274e-07, ..., 1.8626e-08, + 1.0710e-07, 9.3132e-10], + ..., + [ 1.8626e-09, 0.0000e+00, -6.0629e-07, ..., 3.6322e-08, + -4.2375e-07, 3.1665e-08], + [ 3.7253e-09, 0.0000e+00, 1.2293e-07, ..., 9.3132e-08, + 8.6613e-08, 6.5193e-09], + [ 1.4901e-08, 0.0000e+00, 1.2387e-07, ..., 3.7253e-09, + 8.0094e-08, -1.0431e-07]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0406, -0.0349, 0.0088, 0.0305, -0.0177, 0.0057, -0.0134, 0.0173, + 0.0119, -0.0014], device='cuda:0'), grad: tensor([ 1.8906e-07, 1.4901e-07, 4.3400e-07, 3.0734e-07, 1.9465e-07, + 1.0803e-07, -4.5262e-07, -1.3402e-06, -1.7136e-07, 5.8673e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 214.61, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4479 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.09 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.1151, 0.0576, 0.0728, ..., -0.0893, -0.1853, 0.0620], + [-0.1034, -0.0599, -0.1250, ..., 0.1838, -0.1117, -0.0887], + [-0.0515, 0.0064, 0.2383, ..., -0.1368, -0.1651, -0.1527], + ..., + [-0.1635, -0.0647, 0.0236, ..., -0.0538, 0.1281, -0.0966], + [-0.1227, -0.0978, -0.0836, ..., -0.1481, -0.0090, -0.1877], + [-0.1984, -0.0749, -0.1544, ..., -0.2099, -0.0393, 0.1468]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.9558e-08], + [ 2.1420e-08, 0.0000e+00, 2.7008e-08, ..., -2.6077e-08, + 6.5193e-09, 6.5193e-09], + [ 1.0245e-08, 0.0000e+00, -2.8126e-07, ..., 2.7940e-09, + 1.8626e-08, 2.5146e-08], + ..., + [ 1.4901e-08, 0.0000e+00, -3.7253e-09, ..., 1.4901e-08, + -5.4948e-08, 1.8626e-09], + [ 1.1176e-08, 0.0000e+00, 1.0803e-07, ..., 3.7253e-09, + 6.5193e-09, 3.7253e-09], + [ 6.5193e-08, 0.0000e+00, 5.9605e-08, ..., 0.0000e+00, + 1.5832e-08, 4.0047e-08]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0406, -0.0347, 0.0084, 0.0304, -0.0182, 0.0056, -0.0135, 0.0173, + 0.0121, -0.0013], device='cuda:0'), grad: tensor([ 1.3784e-06, 1.3849e-06, 4.5672e-06, -3.4198e-06, 1.8626e-08, + -4.0233e-07, 2.0787e-06, 7.1712e-07, -1.3120e-05, 6.7651e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 214.52, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4643 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.11 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.1152, 0.0576, 0.0729, ..., -0.0893, -0.1853, 0.0621], + [-0.1036, -0.0599, -0.1252, ..., 0.1839, -0.1117, -0.0889], + [-0.0516, 0.0064, 0.2389, ..., -0.1382, -0.1652, -0.1534], + ..., + [-0.1642, -0.0648, 0.0237, ..., -0.0538, 0.1281, -0.0968], + [-0.1231, -0.0979, -0.0840, ..., -0.1481, -0.0093, -0.1890], + [-0.1987, -0.0751, -0.1544, ..., -0.2101, -0.0392, 0.1469]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.4180e-07, ..., -4.6566e-09, + 9.3132e-10, -4.0885e-07], + [ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., -2.7940e-08, + 7.4506e-09, 4.6566e-09], + [ 9.3132e-10, 0.0000e+00, 1.2107e-08, ..., 9.3132e-09, + 1.8626e-09, 1.4901e-08], + ..., + [ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., 9.3132e-09, + 2.0489e-08, 1.3504e-07], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 2.1420e-08, + 9.3132e-09, 9.3132e-09], + [ 6.5193e-09, 0.0000e+00, 5.6811e-08, ..., 1.8626e-09, + -1.5832e-08, -9.2201e-08]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0406, -0.0348, 0.0086, 0.0272, -0.0181, 0.0080, -0.0132, 0.0172, + 0.0119, -0.0013], device='cuda:0'), grad: tensor([-9.9372e-07, -5.5879e-09, 7.0781e-08, -5.8860e-07, -5.5879e-09, + 3.3434e-07, 5.3644e-07, 6.6496e-07, 1.1921e-07, -1.2200e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 214.52, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4786 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.03 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.1153, 0.0577, 0.0730, ..., -0.0894, -0.1854, 0.0621], + [-0.1038, -0.0600, -0.1252, ..., 0.1842, -0.1118, -0.0890], + [-0.0518, 0.0065, 0.2390, ..., -0.1400, -0.1656, -0.1540], + ..., + [-0.1645, -0.0649, 0.0238, ..., -0.0540, 0.1282, -0.0968], + [-0.1237, -0.0979, -0.0840, ..., -0.1482, -0.0095, -0.1891], + [-0.1990, -0.0753, -0.1545, ..., -0.2106, -0.0393, 0.1469]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, -7.4543e-06, ..., 1.7695e-08, + -2.2165e-06, -3.8594e-06], + [ 9.3132e-10, 0.0000e+00, 8.5682e-08, ..., 8.3819e-09, + 2.6077e-08, 4.4703e-08], + [ 0.0000e+00, -0.0000e+00, 4.7218e-07, ..., 1.8626e-09, + 1.4156e-07, 2.4587e-07], + ..., + [ 9.3132e-10, 0.0000e+00, 1.9837e-06, ..., 9.3132e-10, + 5.8953e-07, 1.0272e-06], + [ 9.3132e-10, 0.0000e+00, 2.0303e-07, ..., 9.5926e-08, + 6.2399e-08, 1.1362e-07], + [ 1.1176e-08, 0.0000e+00, 2.1793e-06, ..., 9.3132e-10, + 6.4820e-07, 1.1297e-06]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0406, -0.0346, 0.0079, 0.0272, -0.0180, 0.0082, -0.0131, 0.0172, + 0.0123, -0.0013], device='cuda:0'), grad: tensor([-2.5988e-05, 3.8091e-07, 1.7285e-06, 3.3025e-06, 2.7660e-07, + 5.0887e-06, 2.0023e-07, 6.9998e-06, 2.9337e-07, 7.7337e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 214.35, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4884 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.03 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.1157, 0.0577, 0.0731, ..., -0.0895, -0.1853, 0.0622], + [-0.1042, -0.0601, -0.1258, ..., 0.1840, -0.1118, -0.0906], + [-0.0519, 0.0071, 0.2404, ..., -0.1411, -0.1657, -0.1543], + ..., + [-0.1649, -0.0650, 0.0232, ..., -0.0541, 0.1282, -0.0969], + [-0.1242, -0.0981, -0.0841, ..., -0.1484, -0.0095, -0.1894], + [-0.2000, -0.0754, -0.1544, ..., -0.2076, -0.0393, 0.1469]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, -7.4506e-09, ..., 3.7253e-09, + 1.8626e-09, 9.3132e-10], + [ 2.4214e-08, 0.0000e+00, 5.5879e-09, ..., -4.5635e-08, + 6.5193e-09, 3.7253e-09], + [ 6.5193e-09, 0.0000e+00, 1.6764e-08, ..., 1.1176e-08, + 1.9558e-08, 9.3132e-10], + ..., + [ 2.9802e-08, 0.0000e+00, -6.0536e-08, ..., 1.5832e-08, + -5.4948e-08, 1.2107e-08], + [-1.8626e-08, 0.0000e+00, 3.7253e-09, ..., 1.3970e-08, + 7.4506e-09, 5.5879e-09], + [ 2.6077e-08, 0.0000e+00, 1.5832e-08, ..., 3.7253e-09, + 1.6764e-08, -3.7253e-08]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0405, -0.0351, 0.0086, 0.0272, -0.0177, 0.0081, -0.0128, 0.0170, + 0.0120, -0.0012], device='cuda:0'), grad: tensor([ 2.8871e-08, 3.7253e-09, 9.9652e-08, 3.8184e-08, -9.2201e-08, + -1.3858e-06, 1.3858e-06, 2.7008e-08, -9.1270e-08, -2.7940e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 214.58, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4764 re_mapping 0.0033 re_causal 0.0105 /// teacc 99.04 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.1158, 0.0577, 0.0732, ..., -0.0896, -0.1854, 0.0623], + [-0.1049, -0.0601, -0.1260, ..., 0.1844, -0.1119, -0.0904], + [-0.0522, 0.0074, 0.2410, ..., -0.1415, -0.1659, -0.1548], + ..., + [-0.1652, -0.0651, 0.0235, ..., -0.0545, 0.1284, -0.0969], + [-0.1250, -0.0982, -0.0845, ..., -0.1485, -0.0097, -0.1905], + [-0.2009, -0.0754, -0.1545, ..., -0.2079, -0.0394, 0.1470]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 0.0000e+00, -2.8592e-07, ..., 7.4506e-09, + 0.0000e+00, -4.2934e-07], + [ 1.6484e-07, 0.0000e+00, 0.0000e+00, ..., -1.1325e-06, + 0.0000e+00, 9.3132e-10], + [ 5.7742e-08, 0.0000e+00, -3.7253e-09, ..., 1.3411e-07, + 0.0000e+00, 9.3132e-10], + ..., + [ 8.6613e-08, 0.0000e+00, 9.3132e-10, ..., 8.3540e-07, + 0.0000e+00, 2.5146e-08], + [ 1.8347e-07, 0.0000e+00, 1.8626e-09, ..., 6.2399e-08, + 9.3132e-10, 4.6566e-09], + [ 4.8708e-07, 0.0000e+00, 2.7567e-07, ..., 5.8673e-08, + 0.0000e+00, 3.8743e-07]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0404, -0.0349, 0.0090, 0.0272, -0.0177, 0.0082, -0.0133, 0.0171, + 0.0119, -0.0013], device='cuda:0'), grad: tensor([-1.0058e-06, -2.5406e-06, 4.5821e-07, 3.9581e-07, -1.8710e-06, + 4.0978e-08, 3.2969e-07, 2.3171e-06, -4.4703e-08, 1.9036e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 214.35, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4844 re_mapping 0.0034 re_causal 0.0107 /// teacc 99.07 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.1159, 0.0577, 0.0733, ..., -0.0897, -0.1854, 0.0623], + [-0.1051, -0.0601, -0.1261, ..., 0.1853, -0.1119, -0.0894], + [-0.0524, 0.0104, 0.2434, ..., -0.1425, -0.1646, -0.1557], + ..., + [-0.1652, -0.0655, 0.0240, ..., -0.0549, 0.1286, -0.0969], + [-0.1256, -0.0982, -0.0848, ..., -0.1487, -0.0100, -0.1912], + [-0.2011, -0.0755, -0.1547, ..., -0.2093, -0.0394, 0.1470]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -9.7789e-08, ..., 2.7940e-09, + 9.3132e-10, -7.4506e-08], + [ 2.7940e-09, 0.0000e+00, 1.5832e-08, ..., -2.9150e-07, + 6.1467e-08, 3.8184e-08], + [ 9.3132e-10, 0.0000e+00, -2.6077e-08, ..., 5.5879e-09, + 1.3039e-08, 4.6566e-09], + ..., + [ 8.3819e-09, 0.0000e+00, -9.3132e-09, ..., 2.1514e-07, + -9.4995e-08, 4.9639e-07], + [ 2.7940e-09, 0.0000e+00, 1.0245e-08, ..., 9.3132e-09, + 3.7253e-09, 6.2399e-08], + [ 1.1176e-08, 0.0000e+00, 9.8720e-08, ..., 5.8673e-08, + 1.5832e-08, -1.2368e-06]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0404, -0.0342, 0.0106, 0.0265, -0.0180, 0.0082, -0.0133, 0.0170, + 0.0118, -0.0014], device='cuda:0'), grad: tensor([-1.7229e-07, -2.7381e-07, 4.1910e-08, 1.0710e-07, 2.7381e-06, + 5.4948e-08, 2.1420e-08, 2.2911e-06, 1.2759e-07, -4.9397e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 214.58, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4674 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.08 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.1160, 0.0578, 0.0735, ..., -0.0898, -0.1854, 0.0624], + [-0.1055, -0.0602, -0.1263, ..., 0.1853, -0.1122, -0.0893], + [-0.0526, 0.0104, 0.2433, ..., -0.1426, -0.1650, -0.1565], + ..., + [-0.1660, -0.0657, 0.0243, ..., -0.0549, 0.1287, -0.0970], + [-0.1263, -0.0982, -0.0849, ..., -0.1487, -0.0102, -0.1918], + [-0.2017, -0.0757, -0.1548, ..., -0.2095, -0.0395, 0.1469]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.2236e-07, ..., 7.4506e-09, + -9.0338e-08, -3.1572e-07], + [ 0.0000e+00, 0.0000e+00, -5.9605e-08, ..., -9.6764e-07, + -3.2596e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 3.7625e-07, ..., 1.9930e-07, + 2.8871e-08, 8.1956e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 2.3562e-07, ..., 6.8918e-07, + 3.8184e-08, 6.4261e-08], + [ 0.0000e+00, 0.0000e+00, -4.2655e-07, ..., 3.1665e-08, + 5.5879e-09, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 2.8871e-07, ..., 1.1176e-08, + 2.5146e-08, 2.9802e-08]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0403, -0.0344, 0.0103, 0.0264, -0.0178, 0.0083, -0.0133, 0.0171, + 0.0121, -0.0015], device='cuda:0'), grad: tensor([-1.0487e-06, -1.4622e-06, 1.8403e-06, 1.7043e-07, 8.0094e-07, + 3.3528e-07, 6.6031e-07, 2.2240e-06, -4.0829e-06, 5.4669e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 214.42, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4775 re_mapping 0.0034 re_causal 0.0107 /// teacc 99.08 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.1161, 0.0578, 0.0752, ..., -0.0898, -0.1837, 0.0642], + [-0.1056, -0.0603, -0.1265, ..., 0.1856, -0.1123, -0.0924], + [-0.0527, 0.0105, 0.2436, ..., -0.1438, -0.1653, -0.1572], + ..., + [-0.1660, -0.0657, 0.0246, ..., -0.0550, 0.1289, -0.0970], + [-0.1268, -0.0984, -0.0856, ..., -0.1494, -0.0109, -0.1924], + [-0.2020, -0.0757, -0.1549, ..., -0.2084, -0.0396, 0.1475]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -9.3132e-10, 5.5879e-09, ..., 1.8626e-09, + 5.5879e-09, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 6.5193e-08, ..., -1.0245e-08, + 1.7695e-08, 1.8626e-09], + [-9.3132e-10, -9.3132e-10, -1.4249e-07, ..., 9.3132e-10, + 1.3970e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, -2.8219e-07, ..., 1.1176e-08, + -5.8021e-07, 7.4506e-09], + [ 1.8626e-09, 0.0000e+00, 5.8673e-08, ..., 5.5879e-09, + 4.7497e-08, 7.4506e-09], + [ 7.8231e-08, 9.3132e-10, 2.6729e-07, ..., 9.3132e-10, + 5.0385e-07, -3.4459e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0387, -0.0347, 0.0104, 0.0265, -0.0179, 0.0059, -0.0133, 0.0172, + 0.0113, -0.0014], device='cuda:0'), grad: tensor([ 1.1362e-07, 2.3283e-07, -1.7881e-07, -1.5637e-06, -6.7055e-08, + 1.3197e-06, -9.4064e-08, -9.7975e-07, 1.2387e-07, 1.1027e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 214.41, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4751 re_mapping 0.0034 re_causal 0.0104 /// teacc 99.17 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.1166, 0.0579, 0.0753, ..., -0.0896, -0.1838, 0.0642], + [-0.1063, -0.0605, -0.1268, ..., 0.1857, -0.1124, -0.0926], + [-0.0530, 0.0105, 0.2437, ..., -0.1439, -0.1660, -0.1583], + ..., + [-0.1687, -0.0661, 0.0252, ..., -0.0551, 0.1291, -0.0970], + [-0.1285, -0.0986, -0.0861, ..., -0.1496, -0.0110, -0.1933], + [-0.2024, -0.0762, -0.1550, ..., -0.2085, -0.0396, 0.1475]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 9.3132e-10, + 4.6566e-09, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 3.4459e-08, ..., -4.2841e-08, + 1.5832e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.0023e-07, ..., 1.8626e-09, + 8.3819e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.0245e-07, ..., 3.0734e-08, + -2.3004e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0896e-07, ..., 2.7940e-09, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0710e-07, ..., 2.7940e-09, + 1.6950e-07, 1.8626e-09]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0386, -0.0348, 0.0102, 0.0267, -0.0176, 0.0058, -0.0131, 0.0172, + 0.0112, -0.0014], device='cuda:0'), grad: tensor([ 4.7497e-08, 3.1665e-08, -2.0023e-07, -3.0454e-07, 2.5146e-08, + 8.4750e-08, 3.1665e-08, -3.2689e-07, 2.1607e-07, 3.9581e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 214.40, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4903 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.13 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.1167, 0.0580, 0.0754, ..., -0.0897, -0.1838, 0.0643], + [-0.1065, -0.0605, -0.1270, ..., 0.1858, -0.1127, -0.0927], + [-0.0531, 0.0106, 0.2438, ..., -0.1441, -0.1662, -0.1588], + ..., + [-0.1688, -0.0666, 0.0254, ..., -0.0552, 0.1293, -0.0971], + [-0.1288, -0.0987, -0.0863, ..., -0.1496, -0.0111, -0.1944], + [-0.2027, -0.0764, -0.1551, ..., -0.2086, -0.0397, 0.1475]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 9.3132e-10, + 0.0000e+00, -4.6566e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -1.1176e-08, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -2.1793e-07, ..., 2.7940e-09, + 0.0000e+00, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 4.6566e-09, + 9.3132e-09, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, 1.4249e-07, ..., 1.8626e-09, + 0.0000e+00, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, -1.0803e-07]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0386, -0.0349, 0.0102, 0.0267, -0.0176, 0.0058, -0.0131, 0.0172, + 0.0110, -0.0015], device='cuda:0'), grad: tensor([-7.6368e-08, -1.8626e-09, -3.7160e-07, 3.7253e-08, 8.3819e-08, + 9.9652e-08, 6.5193e-09, 2.0117e-07, 3.1479e-07, -2.8685e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 214.52, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4933 re_mapping 0.0037 re_causal 0.0115 /// teacc 98.98 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.1168, 0.0580, 0.0755, ..., -0.0897, -0.1838, 0.0643], + [-0.1070, -0.0606, -0.1270, ..., 0.1859, -0.1128, -0.0927], + [-0.0531, 0.0108, 0.2437, ..., -0.1443, -0.1664, -0.1612], + ..., + [-0.1690, -0.0684, 0.0255, ..., -0.0553, 0.1293, -0.0972], + [-0.1291, -0.0987, -0.0861, ..., -0.1496, -0.0108, -0.1966], + [-0.2024, -0.0764, -0.1552, ..., -0.2086, -0.0397, 0.1479]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 9.3132e-10, + 1.0245e-08, 1.4901e-08], + [-5.8673e-08, 0.0000e+00, 1.3318e-07, ..., -1.6857e-07, + 3.4459e-07, 1.9558e-08], + [ 9.3132e-10, 0.0000e+00, -2.0489e-08, ..., 3.7253e-09, + 2.3283e-08, 1.2107e-08], + ..., + [ 1.3970e-08, 0.0000e+00, -2.0675e-07, ..., 4.5635e-08, + -5.2992e-07, 2.9430e-07], + [ 1.3039e-08, 0.0000e+00, 7.4506e-09, ..., 3.9116e-08, + 5.5879e-08, 3.0268e-06], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 9.3132e-10, + 3.2596e-08, -3.4459e-06]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0385, -0.0349, 0.0100, 0.0269, -0.0199, 0.0057, -0.0131, 0.0171, + 0.0114, -0.0010], device='cuda:0'), grad: tensor([ 1.0524e-07, 3.3062e-07, 5.5879e-08, 1.1083e-07, 5.0757e-07, + -9.8720e-08, 8.0094e-08, 2.8592e-07, 1.2375e-05, -1.3754e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 214.50, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.4749 re_mapping 0.0038 re_causal 0.0111 /// teacc 99.05 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.1176, 0.0583, 0.0758, ..., -0.0895, -0.1839, 0.0648], + [-0.1112, -0.0607, -0.1271, ..., 0.1878, -0.1128, -0.0928], + [-0.0555, 0.0111, 0.2414, ..., -0.1452, -0.1693, -0.1639], + ..., + [-0.1705, -0.0695, 0.0279, ..., -0.0567, 0.1301, -0.0973], + [-0.1311, -0.0988, -0.0864, ..., -0.1509, -0.0112, -0.1991], + [-0.2067, -0.0784, -0.1555, ..., -0.2119, -0.0398, 0.1482]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.3551e-07, ..., -5.2154e-08, + 0.0000e+00, -4.6473e-07], + [ 0.0000e+00, 0.0000e+00, 4.2841e-08, ..., -3.7253e-09, + 1.8626e-09, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 8.3819e-09, + 0.0000e+00, 2.2352e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 7.4506e-09, + -6.5193e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., -8.3819e-09, + 9.3132e-10, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 3.3900e-07, ..., 3.2596e-08, + 2.7940e-09, 3.0734e-07]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0381, -0.0340, 0.0071, 0.0269, -0.0165, 0.0057, -0.0134, 0.0172, + 0.0096, -0.0036], device='cuda:0'), grad: tensor([-1.3057e-06, 1.7416e-07, 2.1234e-07, -5.7649e-07, 1.4901e-08, + 1.6671e-07, 1.9837e-07, 2.5239e-07, -2.7940e-08, 8.7544e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 214.31, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4736 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.15 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.1193, 0.0583, 0.0760, ..., -0.0908, -0.1839, 0.0649], + [-0.1121, -0.0608, -0.1273, ..., 0.1878, -0.1129, -0.0933], + [-0.0567, 0.0119, 0.2409, ..., -0.1453, -0.1701, -0.1644], + ..., + [-0.1709, -0.0722, 0.0286, ..., -0.0567, 0.1304, -0.0974], + [-0.1323, -0.0991, -0.0870, ..., -0.1510, -0.0116, -0.2000], + [-0.2068, -0.0785, -0.1557, ..., -0.2120, -0.0398, 0.1483]], + device='cuda:0'), grad: tensor([[ 3.7160e-07, 0.0000e+00, -9.3132e-10, ..., 2.1420e-07, + 1.8626e-09, 1.8626e-09], + [ 3.1665e-08, 0.0000e+00, 2.5146e-08, ..., -4.9360e-08, + 8.3819e-09, 3.7253e-09], + [ 5.4017e-08, 0.0000e+00, -6.9197e-07, ..., 5.4948e-08, + 2.6077e-08, 9.3132e-10], + ..., + [ 1.3970e-08, 0.0000e+00, -3.7253e-08, ..., 2.8871e-08, + -1.0058e-07, 1.2107e-08], + [ 1.0487e-06, 0.0000e+00, 6.6496e-07, ..., 5.9977e-07, + 5.5879e-08, 8.7544e-08], + [ 1.6764e-08, 0.0000e+00, 1.5832e-08, ..., 2.2352e-08, + 3.7253e-09, -2.4308e-07]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0382, -0.0341, 0.0066, 0.0279, -0.0165, 0.0056, -0.0142, 0.0174, + 0.0092, -0.0036], device='cuda:0'), grad: tensor([ 1.3644e-06, -7.3574e-08, -1.2238e-06, 8.8476e-08, 1.4365e-05, + 1.4091e-06, -2.1160e-05, 2.3190e-07, 5.3644e-06, -3.8464e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 214.17, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4872 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.10 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.1206, 0.0583, 0.0762, ..., -0.0909, -0.1840, 0.0650], + [-0.1131, -0.0608, -0.1275, ..., 0.1879, -0.1130, -0.0934], + [-0.0572, 0.0120, 0.2410, ..., -0.1454, -0.1702, -0.1657], + ..., + [-0.1692, -0.0723, 0.0287, ..., -0.0568, 0.1309, -0.0974], + [-0.1340, -0.0991, -0.0874, ..., -0.1511, -0.0119, -0.2041], + [-0.2069, -0.0785, -0.1559, ..., -0.2121, -0.0398, 0.1484]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -3.7253e-09, ..., 1.8626e-09, + 3.2596e-08, 1.1176e-08], + [ 1.6857e-07, 5.5879e-09, 2.8871e-08, ..., 2.7008e-08, + 8.3819e-09, 4.6566e-09], + [-1.4901e-08, -7.4506e-09, -5.3085e-08, ..., 1.8626e-09, + 2.7940e-09, 9.3132e-10], + ..., + [ 3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 1.3039e-08, + 2.9523e-07, 2.0768e-07], + [ 7.4506e-09, 0.0000e+00, 1.3970e-08, ..., 3.7253e-09, + 1.7695e-08, 2.0489e-08], + [ 2.5425e-07, 0.0000e+00, 5.5879e-09, ..., 6.9849e-08, + -4.0606e-07, -2.9523e-07]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0381, -0.0340, 0.0067, 0.0278, -0.0165, 0.0056, -0.0142, 0.0175, + 0.0082, -0.0037], device='cuda:0'), grad: tensor([ 3.6508e-07, 3.5483e-07, -5.4017e-08, -3.2037e-07, -7.0781e-08, + 1.2945e-07, 5.2154e-08, 3.3956e-06, 2.2911e-07, -4.0755e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 214.11, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4571 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.15 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.1209, 0.0583, 0.0763, ..., -0.0909, -0.1841, 0.0652], + [-0.1134, -0.0609, -0.1286, ..., 0.1882, -0.1131, -0.0934], + [-0.0573, 0.0120, 0.2415, ..., -0.1430, -0.1703, -0.1674], + ..., + [-0.1692, -0.0723, 0.0288, ..., -0.0574, 0.1310, -0.0975], + [-0.1320, -0.0991, -0.0878, ..., -0.1512, -0.0121, -0.2049], + [-0.2069, -0.0785, -0.1560, ..., -0.2121, -0.0399, 0.1484]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-08, ..., 1.8626e-09, + 9.3132e-10, -1.1921e-07], + [ 0.0000e+00, 9.3132e-10, 1.4901e-08, ..., 1.8626e-09, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, -2.9802e-08, -1.4529e-07, ..., 1.8626e-09, + 0.0000e+00, 8.3819e-09], + ..., + [ 9.3132e-10, 2.3283e-08, 9.3132e-08, ..., 1.8626e-09, + 6.5193e-09, 2.2352e-08], + [ 9.3132e-10, 1.8626e-09, 4.1910e-08, ..., -2.6077e-08, + 1.8626e-09, 1.8626e-08], + [-2.3283e-08, 0.0000e+00, 4.8429e-08, ..., 1.8626e-09, + -9.3132e-10, -7.4506e-08]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0380, -0.0342, 0.0076, 0.0271, -0.0165, 0.0061, -0.0143, 0.0174, + 0.0085, -0.0037], device='cuda:0'), grad: tensor([-3.5577e-07, 6.7987e-08, -1.4901e-07, -1.5926e-07, 2.8033e-07, + 1.9930e-07, 7.9162e-08, 2.3656e-07, -5.7742e-08, -1.4808e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 214.05, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4558 re_mapping 0.0034 re_causal 0.0104 /// teacc 99.14 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.1213, 0.0583, 0.0764, ..., -0.0910, -0.1841, 0.0652], + [-0.1141, -0.0610, -0.1288, ..., 0.1887, -0.1132, -0.0935], + [-0.0593, 0.0121, 0.2415, ..., -0.1432, -0.1703, -0.1689], + ..., + [-0.1696, -0.0726, 0.0288, ..., -0.0579, 0.1311, -0.0976], + [-0.1319, -0.0991, -0.0883, ..., -0.1512, -0.0122, -0.2053], + [-0.2071, -0.0786, -0.1560, ..., -0.2124, -0.0399, 0.1484]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., 1.8626e-09, + 9.3132e-10, -1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -1.7092e-05, + 3.7253e-09, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 2.5611e-07, ..., 7.4506e-09, + 2.7940e-08, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2014e-07, ..., 1.6749e-05, + -1.9558e-08, 4.9919e-07], + [ 0.0000e+00, 0.0000e+00, -2.4773e-07, ..., 5.9605e-08, + -2.6077e-08, 7.3574e-08], + [ 0.0000e+00, 0.0000e+00, -1.4529e-07, ..., 1.6950e-07, + 1.2107e-08, -6.6962e-07]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0379, -0.0341, 0.0075, 0.0266, -0.0163, 0.0064, -0.0144, 0.0173, + 0.0087, -0.0039], device='cuda:0'), grad: tensor([-4.6566e-09, -3.2872e-05, 1.3113e-06, 2.7008e-08, 2.7660e-07, + 1.1548e-07, 4.1910e-08, 3.4302e-05, -9.5181e-07, -2.3283e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 214.06, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4618 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.17 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.1222, 0.0583, 0.0764, ..., -0.0911, -0.1842, 0.0652], + [-0.1146, -0.0611, -0.1289, ..., 0.1892, -0.1133, -0.0935], + [-0.0607, 0.0117, 0.2416, ..., -0.1439, -0.1703, -0.1693], + ..., + [-0.1698, -0.0728, 0.0289, ..., -0.0584, 0.1312, -0.0977], + [-0.1324, -0.0992, -0.0884, ..., -0.1508, -0.0123, -0.2059], + [-0.2071, -0.0789, -0.1561, ..., -0.2125, -0.0399, 0.1486]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, -1.0245e-08, ..., 1.8626e-09, + 2.0489e-08, 9.3132e-10], + [ 3.7253e-09, 0.0000e+00, 5.5879e-09, ..., -2.1420e-08, + 8.3819e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., 9.3132e-10, + 2.7940e-09, 2.7940e-09], + ..., + [ 3.7253e-09, 0.0000e+00, -3.5390e-08, ..., 1.3970e-08, + -1.8254e-07, 1.2107e-08], + [ 1.8626e-09, 0.0000e+00, 4.6566e-09, ..., 1.8626e-09, + 1.0245e-08, 3.7253e-09], + [ 2.7940e-09, 9.3132e-10, 5.2154e-08, ..., 9.3132e-10, + 7.9162e-08, 2.1420e-08]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0380, -0.0339, 0.0073, 0.0265, -0.0162, 0.0064, -0.0148, 0.0171, + 0.0098, -0.0039], device='cuda:0'), grad: tensor([-1.9558e-08, 2.8871e-08, -8.3819e-09, 7.3574e-08, 1.8068e-07, + -1.9651e-07, 5.5879e-08, -3.0454e-07, -2.3283e-08, 2.1514e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 214.33, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4757 re_mapping 0.0035 re_causal 0.0109 /// teacc 99.12 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.1224, 0.0583, 0.0765, ..., -0.0914, -0.1844, 0.0653], + [-0.1152, -0.0614, -0.1290, ..., 0.1893, -0.1133, -0.0936], + [-0.0608, 0.0117, 0.2421, ..., -0.1445, -0.1700, -0.1667], + ..., + [-0.1707, -0.0728, 0.0289, ..., -0.0584, 0.1313, -0.0978], + [-0.1324, -0.0992, -0.0887, ..., -0.1506, -0.0126, -0.2067], + [-0.2071, -0.0790, -0.1565, ..., -0.2126, -0.0401, 0.1485]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -5.5879e-09, ..., 4.6566e-09, + 9.3132e-10, -0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 1.5832e-08, ..., -2.0489e-08, + 1.7695e-08, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 2.7940e-09, + 2.5146e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, -1.1269e-07, ..., 5.5879e-09, + -1.4622e-07, 3.1665e-08], + [ 9.3132e-10, 0.0000e+00, 5.4948e-08, ..., 4.6566e-09, + 5.1223e-08, 6.5193e-09], + [ 4.6566e-09, 0.0000e+00, 1.9558e-08, ..., 9.3132e-10, + 1.6764e-08, -5.9605e-08]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0380, -0.0340, 0.0077, 0.0263, -0.0161, 0.0064, -0.0149, 0.0171, + 0.0105, -0.0040], device='cuda:0'), grad: tensor([ 1.7695e-08, 1.6764e-08, 1.5832e-08, -2.8871e-08, 2.2352e-08, + 5.6811e-08, 7.4506e-09, -1.7975e-07, 1.7975e-07, -1.0896e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 214.52, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4760 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.13 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.1226, 0.0583, 0.0767, ..., -0.0916, -0.1844, 0.0654], + [-0.1155, -0.0615, -0.1300, ..., 0.1892, -0.1139, -0.0936], + [-0.0615, 0.0123, 0.2422, ..., -0.1448, -0.1701, -0.1668], + ..., + [-0.1710, -0.0748, 0.0297, ..., -0.0583, 0.1319, -0.0975], + [-0.1326, -0.0995, -0.0891, ..., -0.1507, -0.0129, -0.2074], + [-0.2072, -0.0791, -0.1571, ..., -0.2126, -0.0406, 0.1484]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.6566e-09, ..., 1.4901e-08, + 9.3132e-10, -1.8626e-09], + [ 2.7940e-09, 2.7940e-09, 8.3819e-09, ..., 1.9558e-07, + 6.5193e-09, 2.7940e-09], + [ 9.3132e-10, -1.8626e-09, -7.4506e-09, ..., 1.8626e-08, + 2.7940e-09, 9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 1.3039e-08, + -2.7940e-09, 4.6566e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -2.5705e-07, + 6.4261e-08, 2.7008e-08], + [ 1.2107e-08, 0.0000e+00, 5.5879e-09, ..., 1.2107e-08, + 7.4506e-09, -1.2107e-08]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0379, -0.0343, 0.0077, 0.0265, -0.0161, 0.0062, -0.0147, 0.0176, + 0.0104, -0.0043], device='cuda:0'), grad: tensor([ 2.1886e-07, 2.9691e-06, 2.2259e-07, 2.5295e-06, 1.2107e-08, + -2.5425e-06, 6.3889e-07, 1.0803e-07, -4.3139e-06, 1.5553e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 214.51, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4620 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.09 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.1227, 0.0583, 0.0768, ..., -0.0917, -0.1845, 0.0656], + [-0.1156, -0.0616, -0.1302, ..., 0.1893, -0.1139, -0.0937], + [-0.0617, 0.0123, 0.2425, ..., -0.1451, -0.1701, -0.1668], + ..., + [-0.1712, -0.0748, 0.0298, ..., -0.0584, 0.1320, -0.0976], + [-0.1327, -0.0995, -0.0893, ..., -0.1507, -0.0130, -0.2080], + [-0.2072, -0.0791, -0.1573, ..., -0.2127, -0.0406, 0.1484]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 7.6368e-08, -1.3597e-07, ..., 5.5879e-09, + 0.0000e+00, -3.7253e-08], + [ 1.8626e-09, 5.5879e-09, 2.5146e-08, ..., -2.0489e-08, + 5.5879e-09, 1.2107e-08], + [ 9.3132e-10, 2.7940e-09, -9.7789e-08, ..., 9.3132e-09, + 2.7940e-09, 7.4506e-09], + ..., + [ 0.0000e+00, 2.6077e-08, 1.3970e-08, ..., 8.3819e-09, + -1.0245e-08, 8.0094e-08], + [ 1.8626e-09, 1.6764e-07, 2.4214e-08, ..., -5.5879e-09, + 1.8626e-09, 2.0582e-07], + [ 3.7253e-09, 1.0524e-07, 1.2480e-07, ..., 1.8626e-09, + 4.6566e-09, 6.9849e-08]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0377, -0.0343, 0.0078, 0.0264, -0.0161, 0.0062, -0.0147, 0.0176, + 0.0105, -0.0043], device='cuda:0'), grad: tensor([ 1.4529e-07, 7.3574e-08, -1.7323e-07, -2.2445e-06, 1.8161e-07, + 1.5553e-07, 3.6322e-08, 3.8650e-07, 1.0058e-06, 4.1723e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 214.22, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4815 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.11 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.1232, 0.0583, 0.0769, ..., -0.0919, -0.1846, 0.0659], + [-0.1163, -0.0617, -0.1301, ..., 0.1896, -0.1140, -0.0938], + [-0.0618, 0.0123, 0.2426, ..., -0.1467, -0.1702, -0.1670], + ..., + [-0.1716, -0.0748, 0.0298, ..., -0.0585, 0.1321, -0.0977], + [-0.1333, -0.1000, -0.0896, ..., -0.1508, -0.0134, -0.2090], + [-0.2072, -0.0793, -0.1574, ..., -0.2130, -0.0407, 0.1484]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, -1.8626e-08, ..., 9.3132e-10, + 9.3132e-10, -8.1956e-08], + [ 1.3039e-08, 0.0000e+00, 3.2596e-08, ..., -1.0245e-08, + 1.4901e-08, 5.5879e-09], + [ 4.6566e-09, 0.0000e+00, -3.6601e-07, ..., 1.8626e-09, + 8.7544e-08, 1.8626e-09], + ..., + [ 8.4750e-08, 0.0000e+00, 1.6764e-07, ..., 7.4506e-09, + -1.7229e-07, 3.9116e-08], + [ 7.4506e-09, 0.0000e+00, 1.4901e-07, ..., -9.3132e-10, + 2.8871e-08, 8.3819e-09], + [ 1.1269e-07, 0.0000e+00, 1.6764e-08, ..., 9.3132e-10, + 1.2107e-08, -2.8871e-08]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0375, -0.0342, 0.0077, 0.0247, -0.0161, 0.0074, -0.0143, 0.0175, + 0.0103, -0.0044], device='cuda:0'), grad: tensor([-1.9278e-07, 8.8476e-08, -8.0653e-07, 1.3132e-07, -5.4482e-07, + 2.2352e-08, 2.1514e-07, 6.4075e-07, 2.7567e-07, 1.7695e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 214.18, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4681 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.05 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.1242, 0.0583, 0.0769, ..., -0.0920, -0.1848, 0.0659], + [-0.1168, -0.0617, -0.1302, ..., 0.1897, -0.1141, -0.0938], + [-0.0627, 0.0123, 0.2426, ..., -0.1475, -0.1703, -0.1673], + ..., + [-0.1719, -0.0748, 0.0299, ..., -0.0586, 0.1322, -0.0981], + [-0.1322, -0.1000, -0.0896, ..., -0.1509, -0.0144, -0.2102], + [-0.2073, -0.0793, -0.1574, ..., -0.2130, -0.0407, 0.1486]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.0338e-08, ..., 0.0000e+00, + 9.3132e-10, -7.5437e-08], + [ 2.7940e-08, 0.0000e+00, 2.7940e-09, ..., -1.6764e-08, + 3.7253e-09, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -1.4901e-08, ..., 4.6566e-09, + -2.5146e-08, 2.7940e-09], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, + 2.7940e-09, 6.5193e-09], + [ 1.3970e-08, 0.0000e+00, 5.4948e-08, ..., 9.3132e-10, + 2.0489e-08, 2.9802e-08]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0376, -0.0342, 0.0074, 0.0241, -0.0161, 0.0081, -0.0140, 0.0174, + 0.0101, -0.0043], device='cuda:0'), grad: tensor([-1.9651e-07, 2.5146e-08, 1.1176e-08, 4.0047e-08, -3.2596e-08, + -1.8440e-07, 1.9372e-07, -2.2352e-08, 1.3039e-08, 1.4529e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 214.76, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4782 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.11 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.1244, 0.0583, 0.0770, ..., -0.0920, -0.1849, 0.0659], + [-0.1176, -0.0617, -0.1305, ..., 0.1900, -0.1142, -0.0938], + [-0.0628, 0.0124, 0.2431, ..., -0.1478, -0.1702, -0.1675], + ..., + [-0.1725, -0.0749, 0.0297, ..., -0.0589, 0.1324, -0.0982], + [-0.1329, -0.1004, -0.0898, ..., -0.1510, -0.0143, -0.2109], + [-0.2073, -0.0793, -0.1576, ..., -0.2132, -0.0408, 0.1486]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -8.3819e-09, ..., 1.0245e-08, + 9.3132e-10, -2.7940e-09], + [ 1.8626e-09, 0.0000e+00, 9.3132e-09, ..., -2.0489e-07, + 2.7940e-09, 2.7940e-09], + [ 2.7940e-09, 0.0000e+00, -4.5635e-08, ..., 8.3819e-09, + 9.3132e-10, 9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, 1.4901e-08, ..., 1.7788e-07, + -7.4506e-09, 5.5879e-08], + [ 5.5879e-09, 0.0000e+00, 1.7695e-08, ..., 2.2352e-08, + 2.7940e-09, 3.7253e-09], + [ 8.3819e-09, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 2.7940e-09, -9.4995e-08]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0376, -0.0342, 0.0076, 0.0243, -0.0161, 0.0080, -0.0137, 0.0173, + 0.0101, -0.0044], device='cuda:0'), grad: tensor([ 2.4214e-08, -3.0641e-07, -5.2154e-08, -2.5146e-08, 3.9116e-07, + 3.7253e-08, -3.9954e-07, 4.9639e-07, 1.1269e-07, -2.7940e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 214.53, cls_loss 0.0006 cls_loss_mapping 0.0026 cls_loss_causal 0.4518 re_mapping 0.0034 re_causal 0.0104 /// teacc 99.13 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.1244, 0.0583, 0.0772, ..., -0.0921, -0.1850, 0.0661], + [-0.1182, -0.0617, -0.1309, ..., 0.1901, -0.1144, -0.0938], + [-0.0628, 0.0124, 0.2464, ..., -0.1484, -0.1672, -0.1681], + ..., + [-0.1727, -0.0749, 0.0266, ..., -0.0589, 0.1297, -0.0982], + [-0.1331, -0.1004, -0.0885, ..., -0.1511, -0.0152, -0.2120], + [-0.2075, -0.0793, -0.1579, ..., -0.2133, -0.0410, 0.1487]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -8.4750e-08, ..., 0.0000e+00, + 0.0000e+00, -7.2643e-08], + [ 1.1176e-08, 0.0000e+00, 6.5193e-09, ..., -3.7253e-09, + 7.4506e-09, 6.5193e-09], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 6.5193e-09], + ..., + [ 1.5832e-08, 0.0000e+00, -9.3132e-10, ..., 2.7940e-09, + -2.3283e-08, 1.3411e-07], + [ 4.6566e-09, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 1.8626e-09, 4.6566e-09], + [ 5.4017e-08, 0.0000e+00, 5.4017e-08, ..., 9.3132e-10, + 9.3132e-09, -1.3225e-07]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0374, -0.0343, 0.0107, 0.0242, -0.0159, 0.0080, -0.0141, 0.0146, + 0.0104, -0.0046], device='cuda:0'), grad: tensor([-2.8592e-07, 4.9360e-08, 3.8184e-08, 7.6368e-08, -6.4261e-08, + 1.4901e-08, 4.6566e-08, 2.6450e-07, -3.0734e-08, -1.0338e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 214.40, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4807 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.15 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.1246, 0.0583, 0.0774, ..., -0.0921, -0.1852, 0.0662], + [-0.1185, -0.0618, -0.1313, ..., 0.1904, -0.1145, -0.0939], + [-0.0628, 0.0124, 0.2464, ..., -0.1483, -0.1672, -0.1686], + ..., + [-0.1729, -0.0749, 0.0267, ..., -0.0592, 0.1298, -0.0983], + [-0.1335, -0.1004, -0.0884, ..., -0.1512, -0.0152, -0.2130], + [-0.2074, -0.0794, -0.1581, ..., -0.2135, -0.0411, 0.1487]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8871e-08, ..., 0.0000e+00, + -0.0000e+00, -3.6322e-08], + [ 1.8626e-09, 0.0000e+00, 2.9802e-08, ..., 9.3132e-10, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.5460e-07, ..., 3.7253e-09, + 2.3283e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 1.9558e-08, ..., -5.5879e-09, + -3.3528e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 5.5879e-08, ..., 1.8626e-09, + 3.7253e-09, 1.8626e-09], + [ 1.1176e-08, 0.0000e+00, 2.7008e-08, ..., 0.0000e+00, + 1.8626e-08, 3.6322e-08]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0373, -0.0342, 0.0107, 0.0242, -0.0160, 0.0080, -0.0141, 0.0146, + 0.0105, -0.0046], device='cuda:0'), grad: tensor([-6.5193e-08, 7.5437e-08, -3.3714e-07, -4.6566e-09, -1.7695e-08, + 9.3132e-10, 3.3528e-08, 3.8184e-08, 1.3877e-07, 1.3877e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 214.45, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4379 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.09 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.1247, 0.0583, 0.0775, ..., -0.0922, -0.1851, 0.0663], + [-0.1188, -0.0618, -0.1315, ..., 0.1907, -0.1146, -0.0939], + [-0.0627, 0.0125, 0.2463, ..., -0.1486, -0.1674, -0.1690], + ..., + [-0.1736, -0.0749, 0.0268, ..., -0.0595, 0.1301, -0.0984], + [-0.1336, -0.1004, -0.0886, ..., -0.1515, -0.0153, -0.2137], + [-0.2074, -0.0794, -0.1582, ..., -0.2136, -0.0411, 0.1489]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.4901e-08, ..., 6.5193e-09, + 0.0000e+00, -9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., -9.2201e-08, + 9.3132e-10, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, -3.7253e-09, ..., 2.3283e-08, + -9.3132e-10, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.1176e-08, + 8.3819e-09, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 1.7695e-08, ..., 9.3132e-10, + 3.7253e-09, -1.8626e-09]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0372, -0.0341, 0.0105, 0.0242, -0.0161, 0.0080, -0.0139, 0.0147, + 0.0105, -0.0045], device='cuda:0'), grad: tensor([ 1.3718e-06, 7.0781e-08, 1.1176e-07, 1.5814e-06, 1.5832e-08, + 1.1642e-07, 2.6636e-07, 1.5926e-07, -3.8221e-06, 1.3225e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 214.20, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.5045 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.15 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.1248, 0.0583, 0.0777, ..., -0.0922, -0.1851, 0.0665], + [-0.1191, -0.0618, -0.1316, ..., 0.1910, -0.1147, -0.0940], + [-0.0629, 0.0124, 0.2463, ..., -0.1493, -0.1674, -0.1694], + ..., + [-0.1737, -0.0748, 0.0269, ..., -0.0597, 0.1301, -0.0986], + [-0.1340, -0.1005, -0.0890, ..., -0.1515, -0.0166, -0.2144], + [-0.2075, -0.0795, -0.1584, ..., -0.2137, -0.0412, 0.1489]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8871e-08, ..., -2.7940e-09, + 2.7940e-09, 6.8918e-08], + [ 1.8626e-09, 0.0000e+00, 2.1420e-08, ..., -1.7695e-08, + 5.2154e-08, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, -2.6450e-07, ..., 9.3132e-10, + 5.5879e-09, 1.8626e-09], + ..., + [ 2.7940e-09, 0.0000e+00, -5.4017e-08, ..., 4.6566e-09, + -1.5274e-07, 3.9116e-08], + [ 9.3132e-10, 0.0000e+00, 2.5518e-07, ..., 8.3819e-09, + 4.6566e-09, 2.4866e-07], + [ 1.3039e-08, 0.0000e+00, 4.0047e-08, ..., 2.7940e-09, + 5.6811e-08, -4.2468e-07]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0370, -0.0340, 0.0105, 0.0243, -0.0161, 0.0082, -0.0150, 0.0147, + 0.0101, -0.0046], device='cuda:0'), grad: tensor([ 1.1828e-07, 1.0431e-07, -5.4482e-07, 5.2154e-08, 7.9162e-08, + 5.3085e-08, 1.5832e-08, -2.0489e-07, 1.0310e-06, -6.9849e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 214.00, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4379 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.16 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.1251, 0.0583, 0.0777, ..., -0.0925, -0.1851, 0.0665], + [-0.1193, -0.0618, -0.1318, ..., 0.1911, -0.1148, -0.0941], + [-0.0633, 0.0124, 0.2463, ..., -0.1497, -0.1675, -0.1697], + ..., + [-0.1739, -0.0748, 0.0269, ..., -0.0598, 0.1302, -0.0988], + [-0.1324, -0.1005, -0.0888, ..., -0.1515, -0.0151, -0.2154], + [-0.2076, -0.0795, -0.1586, ..., -0.2139, -0.0413, 0.1491]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.4703e-08, + 1.0245e-08, 2.6822e-07], + [ 3.7253e-09, 0.0000e+00, 4.5635e-08, ..., 5.2992e-07, + 5.7742e-08, 4.4256e-06], + [ 0.0000e+00, 0.0000e+00, 6.1467e-08, ..., 4.5635e-08, + 6.6124e-08, 4.0978e-08], + ..., + [ 1.8626e-09, 0.0000e+00, -2.9802e-07, ..., 8.6706e-07, + -3.6508e-07, 7.2718e-06], + [ 0.0000e+00, 0.0000e+00, 7.5437e-08, ..., 3.3062e-07, + 1.2852e-07, 2.5630e-06], + [ 1.5832e-08, 0.0000e+00, 7.4506e-08, ..., -1.9502e-06, + 8.8476e-08, -1.6391e-05]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0371, -0.0341, 0.0105, 0.0246, -0.0161, 0.0080, -0.0148, 0.0147, + 0.0113, -0.0046], device='cuda:0'), grad: tensor([ 1.1036e-06, 1.7419e-05, 3.9767e-07, 7.3761e-07, 6.1840e-06, + 2.6077e-08, -2.5146e-07, 2.7686e-05, 1.0408e-05, -6.3837e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 213.93, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5162 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.15 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.1253, 0.0583, 0.0779, ..., -0.0926, -0.1852, 0.0666], + [-0.1201, -0.0620, -0.1322, ..., 0.1914, -0.1152, -0.0944], + [-0.0637, 0.0134, 0.2463, ..., -0.1510, -0.1676, -0.1702], + ..., + [-0.1742, -0.0754, 0.0299, ..., -0.0598, 0.1335, -0.0993], + [-0.1325, -0.1006, -0.0922, ..., -0.1518, -0.0187, -0.2156], + [-0.2078, -0.0795, -0.1588, ..., -0.2139, -0.0415, 0.1494]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.1176e-08, ..., 9.3132e-10, + 8.3819e-09, -1.3970e-08], + [ 5.1223e-08, 0.0000e+00, 1.6764e-08, ..., -7.1712e-08, + 5.9418e-07, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -1.3784e-07, ..., 1.0245e-08, + 6.0536e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 1.0245e-08, ..., 4.0978e-08, + -3.0808e-06, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 4.9360e-08, ..., 1.2107e-08, + 4.8429e-08, 0.0000e+00], + [ 1.5832e-08, 0.0000e+00, 1.3970e-08, ..., 1.8626e-09, + 1.2489e-06, 1.0245e-08]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0370, -0.0344, 0.0104, 0.0245, -0.0158, 0.0080, -0.0148, 0.0179, + 0.0077, -0.0048], device='cuda:0'), grad: tensor([ 2.0489e-08, 2.5891e-06, 3.3528e-08, 2.0117e-07, 4.6194e-06, + 1.6484e-07, 1.4901e-08, -1.3605e-05, 3.3062e-07, 5.6401e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 213.90, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4611 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.1255, 0.0584, 0.0781, ..., -0.0929, -0.1853, 0.0667], + [-0.1207, -0.0620, -0.1325, ..., 0.1920, -0.1154, -0.0945], + [-0.0638, 0.0134, 0.2461, ..., -0.1517, -0.1678, -0.1707], + ..., + [-0.1745, -0.0754, 0.0300, ..., -0.0604, 0.1335, -0.0994], + [-0.1332, -0.1006, -0.0922, ..., -0.1522, -0.0187, -0.2164], + [-0.2078, -0.0796, -0.1590, ..., -0.2140, -0.0413, 0.1495]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.5832e-08, ..., 1.8626e-09, + 9.3132e-10, -6.5193e-09], + [ 6.5193e-09, 0.0000e+00, 1.8626e-08, ..., 9.3132e-10, + 1.4901e-08, 4.6566e-09], + [ 9.3132e-10, 0.0000e+00, 1.2945e-07, ..., 9.3132e-10, + 8.1025e-08, 9.3132e-10], + ..., + [ 7.4506e-09, 0.0000e+00, -1.6205e-07, ..., 0.0000e+00, + -1.0617e-07, 4.3772e-08], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 8.3819e-09, 2.6077e-08], + [ 2.7008e-08, 0.0000e+00, 1.9558e-08, ..., 0.0000e+00, + 1.1176e-08, -2.0117e-07]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0369, -0.0342, 0.0101, 0.0245, -0.0158, 0.0080, -0.0142, 0.0179, + 0.0077, -0.0046], device='cuda:0'), grad: tensor([-7.4506e-09, 7.7300e-08, 3.3993e-07, -1.4529e-07, 3.2876e-07, + 1.8626e-09, 2.3283e-08, -2.6263e-07, 2.0303e-07, -5.5134e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 213.93, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4755 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.22 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.1256, 0.0584, 0.0782, ..., -0.0931, -0.1854, 0.0668], + [-0.1212, -0.0627, -0.1333, ..., 0.1921, -0.1155, -0.0946], + [-0.0639, 0.0135, 0.2462, ..., -0.1520, -0.1678, -0.1708], + ..., + [-0.1748, -0.0754, 0.0299, ..., -0.0606, 0.1334, -0.0996], + [-0.1337, -0.1006, -0.0922, ..., -0.1522, -0.0187, -0.2172], + [-0.2078, -0.0796, -0.1591, ..., -0.2141, -0.0413, 0.1499]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -4.6566e-10, 2.4214e-08, ..., 4.1910e-09, + 5.3551e-08, -3.2131e-08], + [ 3.2596e-09, 0.0000e+00, 1.2061e-07, ..., -2.4214e-08, + 1.3318e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.4075e-07, ..., 5.1223e-09, + 2.7800e-07, 4.6566e-10], + ..., + [ 1.8626e-09, 0.0000e+00, -5.2005e-06, ..., 1.2573e-08, + -5.5730e-06, 1.5367e-08], + [ 3.5390e-08, 0.0000e+00, 2.1420e-08, ..., 2.4680e-08, + 2.4680e-08, 2.3283e-08], + [ 1.7695e-08, 4.6566e-10, 4.7386e-06, ..., 9.3132e-10, + 5.0142e-06, -2.1420e-08]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0368, -0.0345, 0.0102, 0.0253, -0.0166, 0.0079, -0.0140, 0.0178, + 0.0078, -0.0039], device='cuda:0'), grad: tensor([ 9.3132e-08, 4.4145e-07, 7.3202e-07, 7.9162e-09, 5.4948e-08, + 1.7928e-07, -1.4575e-07, -2.5094e-05, 2.2771e-07, 2.3574e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 214.21, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4560 re_mapping 0.0032 re_causal 0.0102 /// teacc 99.13 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.1257, 0.0584, 0.0782, ..., -0.0929, -0.1855, 0.0669], + [-0.1220, -0.0657, -0.1363, ..., 0.1916, -0.1169, -0.0948], + [-0.0640, 0.0164, 0.2463, ..., -0.1495, -0.1678, -0.1711], + ..., + [-0.1750, -0.0754, 0.0299, ..., -0.0606, 0.1334, -0.0998], + [-0.1340, -0.1006, -0.0922, ..., -0.1523, -0.0187, -0.2176], + [-0.2078, -0.0796, -0.1596, ..., -0.2142, -0.0415, 0.1500]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 0.0000e+00, -1.3970e-09, ..., 9.3132e-10, + 5.1223e-09, -4.6566e-09], + [ 1.2573e-08, 0.0000e+00, 6.9849e-09, ..., -1.3970e-08, + 1.8626e-08, 0.0000e+00], + [ 3.2596e-09, 0.0000e+00, -1.9558e-08, ..., 6.9849e-09, + 6.0536e-09, 0.0000e+00], + ..., + [ 1.4203e-07, 0.0000e+00, -2.6543e-08, ..., 1.1176e-08, + 6.2864e-08, 4.6566e-10], + [ 1.2573e-08, 0.0000e+00, 2.3283e-09, ..., -2.7940e-09, + 1.2107e-08, 0.0000e+00], + [ 2.9337e-08, 0.0000e+00, 2.4680e-08, ..., 9.3132e-10, + 4.6100e-08, 5.5879e-09]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0368, -0.0357, 0.0103, 0.0254, -0.0166, 0.0078, -0.0142, 0.0178, + 0.0078, -0.0041], device='cuda:0'), grad: tensor([ 1.8161e-08, 7.1712e-08, 9.3132e-10, 6.0070e-08, -7.0501e-07, + -2.5146e-08, 1.7975e-07, 3.5018e-07, -1.0710e-07, 1.6252e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 214.29, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4651 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.07 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.1281, 0.0584, 0.0786, ..., -0.0937, -0.1853, 0.0672], + [-0.1228, -0.0657, -0.1366, ..., 0.1931, -0.1162, -0.0948], + [-0.0623, 0.0164, 0.2463, ..., -0.1499, -0.1678, -0.1715], + ..., + [-0.1752, -0.0755, 0.0299, ..., -0.0621, 0.1334, -0.0999], + [-0.1346, -0.1006, -0.0922, ..., -0.1523, -0.0187, -0.2180], + [-0.2079, -0.0796, -0.1598, ..., -0.2144, -0.0417, 0.1499]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8254e-07, ..., 1.1036e-07, + 5.1688e-08, -2.2864e-07], + [ 5.5879e-09, 0.0000e+00, 4.1910e-09, ..., -6.0070e-08, + 9.3132e-09, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 3.2596e-08, + 1.6298e-08, 7.9162e-09], + ..., + [ 4.6566e-10, 0.0000e+00, -1.4901e-08, ..., 5.5879e-08, + -1.4901e-08, 3.6787e-08], + [-1.8161e-08, 0.0000e+00, 8.8476e-09, ..., 2.5611e-08, + 1.9139e-07, 9.3598e-08], + [ 1.3970e-09, 0.0000e+00, 1.3644e-07, ..., 2.9802e-08, + 1.0571e-07, 1.7602e-07]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0366, -0.0334, 0.0103, 0.0258, -0.0165, 0.0078, -0.0150, 0.0178, + 0.0078, -0.0043], device='cuda:0'), grad: tensor([-2.0582e-07, -1.3039e-08, 1.9325e-07, 2.1104e-06, 1.3690e-07, + -2.9132e-06, -1.0571e-06, 2.6869e-07, 5.4110e-07, 9.4203e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 214.21, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4718 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.09 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.1283, 0.0584, 0.0787, ..., -0.0939, -0.1853, 0.0673], + [-0.1230, -0.0657, -0.1365, ..., 0.1939, -0.1162, -0.0949], + [-0.0625, 0.0164, 0.2463, ..., -0.1504, -0.1678, -0.1719], + ..., + [-0.1754, -0.0756, 0.0299, ..., -0.0628, 0.1334, -0.1000], + [-0.1349, -0.1006, -0.0922, ..., -0.1531, -0.0187, -0.2187], + [-0.2084, -0.0796, -0.1599, ..., -0.2146, -0.0417, 0.1501]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -9.3132e-10, ..., 4.6566e-10, + 4.6566e-10, 1.8161e-08], + [ 5.8208e-08, 0.0000e+00, 2.7940e-09, ..., -6.9849e-09, + 9.7789e-09, 5.1223e-09], + [ 9.3132e-10, 0.0000e+00, 1.0710e-08, ..., 3.2596e-09, + 1.0245e-08, 1.3970e-09], + ..., + [ 1.0710e-08, 0.0000e+00, -2.1886e-08, ..., 3.2596e-09, + -2.3283e-08, 3.5856e-08], + [ 1.8161e-08, 0.0000e+00, 3.2596e-09, ..., 4.6566e-09, + 6.9849e-09, 9.3132e-09], + [ 4.6892e-07, 0.0000e+00, 5.1223e-09, ..., 1.3039e-08, + 5.2620e-08, -1.5413e-07]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0366, -0.0327, 0.0103, 0.0257, -0.0158, 0.0077, -0.0140, 0.0178, + 0.0078, -0.0048], device='cuda:0'), grad: tensor([ 4.9826e-08, 1.5274e-07, 4.0978e-08, 1.7229e-08, -1.5274e-06, + 5.8208e-08, 2.1048e-07, 5.6345e-08, 9.5926e-08, 8.4937e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 214.35, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4702 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.11 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.1288, 0.0578, 0.0788, ..., -0.0938, -0.1854, 0.0674], + [-0.1246, -0.0657, -0.1390, ..., 0.1919, -0.1190, -0.0950], + [-0.0628, 0.0163, 0.2464, ..., -0.1509, -0.1679, -0.1727], + ..., + [-0.1759, -0.0757, 0.0300, ..., -0.0606, 0.1335, -0.1002], + [-0.1356, -0.1008, -0.0922, ..., -0.1535, -0.0187, -0.2194], + [-0.2086, -0.0798, -0.1600, ..., -0.2148, -0.0418, 0.1501]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 9.3132e-10, + -1.1176e-08, -3.7253e-08], + [ 4.6566e-10, 0.0000e+00, 2.7940e-09, ..., -1.9092e-08, + 4.6566e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -1.2107e-08, ..., 2.7940e-09, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 1.1642e-08, + -9.7789e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., 1.3970e-09, + 0.0000e+00, 1.3970e-09], + [ 4.1910e-09, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 3.7253e-09, -1.0245e-08]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0364, -0.0350, 0.0103, 0.0249, -0.0156, 0.0072, -0.0141, 0.0179, + 0.0078, -0.0051], device='cuda:0'), grad: tensor([-1.4575e-07, 7.6834e-08, 4.7497e-07, 1.3458e-07, 2.2817e-08, + 1.9604e-07, 3.8184e-08, 9.2201e-08, -9.3132e-07, 4.4238e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 214.21, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4638 re_mapping 0.0032 re_causal 0.0105 /// teacc 99.12 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.1289, 0.0577, 0.0789, ..., -0.0938, -0.1854, 0.0675], + [-0.1248, -0.0657, -0.1390, ..., 0.1927, -0.1190, -0.0951], + [-0.0630, 0.0162, 0.2464, ..., -0.1533, -0.1679, -0.1731], + ..., + [-0.1761, -0.0757, 0.0301, ..., -0.0610, 0.1335, -0.1003], + [-0.1357, -0.1009, -0.0922, ..., -0.1536, -0.0187, -0.2200], + [-0.2087, -0.0798, -0.1601, ..., -0.2150, -0.0418, 0.1501]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., 6.9849e-09, + 6.7521e-08, 4.1910e-09], + [ 4.6566e-10, 0.0000e+00, 6.0536e-09, ..., -5.1223e-09, + 1.8161e-08, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-09, + 1.6764e-08, 1.8626e-09], + ..., + [ 4.6566e-10, 0.0000e+00, -1.6764e-08, ..., 2.3283e-09, + 6.3796e-08, 3.4925e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-09, + 3.4459e-08, 3.3993e-08], + [ 1.8626e-09, 0.0000e+00, 5.1223e-09, ..., 1.3970e-09, + 1.0664e-07, -1.2852e-07]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0363, -0.0345, 0.0102, 0.0249, -0.0156, 0.0069, -0.0138, 0.0179, + 0.0078, -0.0051], device='cuda:0'), grad: tensor([ 3.4738e-07, 1.7649e-07, 1.0757e-07, 1.0626e-06, 3.1898e-07, + -2.3674e-06, 7.9116e-07, 5.9977e-07, -1.1027e-06, 6.2864e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 214.51, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4676 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.08 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.1290, 0.0577, 0.0790, ..., -0.0940, -0.1855, 0.0676], + [-0.1255, -0.0658, -0.1392, ..., 0.1929, -0.1191, -0.0960], + [-0.0631, 0.0163, 0.2464, ..., -0.1535, -0.1679, -0.1733], + ..., + [-0.1763, -0.0758, 0.0300, ..., -0.0611, 0.1336, -0.1004], + [-0.1358, -0.1009, -0.0922, ..., -0.1537, -0.0187, -0.2205], + [-0.2091, -0.0798, -0.1602, ..., -0.2151, -0.0419, 0.1504]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -1.2806e-07, + 5.5879e-09, -1.5274e-07], + [ 3.3993e-08, 0.0000e+00, 2.7474e-08, ..., 6.2864e-08, + 4.8429e-08, 7.6834e-08], + [ 3.2596e-09, 0.0000e+00, 2.6543e-08, ..., 7.9162e-09, + 4.6100e-08, 6.0536e-09], + ..., + [ 1.8626e-08, 0.0000e+00, -1.4156e-07, ..., 1.5832e-08, + -2.4820e-07, 8.8476e-09], + [ 3.2596e-09, 0.0000e+00, 1.4901e-08, ..., 1.0710e-08, + 2.5611e-08, 1.8626e-09], + [ 3.8836e-07, 0.0000e+00, 5.9139e-08, ..., 3.5856e-08, + 1.0058e-07, 1.5832e-08]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0363, -0.0346, 0.0102, 0.0251, -0.0152, 0.0069, -0.0138, 0.0179, + 0.0078, -0.0054], device='cuda:0'), grad: tensor([-9.4064e-07, 7.0268e-07, 1.6531e-07, 1.4296e-07, -1.1455e-06, + 7.7765e-08, -3.9116e-08, -5.1735e-07, 9.9186e-08, 1.4678e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 214.31, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4519 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.14 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.1292, 0.0577, 0.0791, ..., -0.0941, -0.1856, 0.0677], + [-0.1287, -0.0658, -0.1394, ..., 0.1926, -0.1192, -0.0962], + [-0.0633, 0.0163, 0.2464, ..., -0.1537, -0.1679, -0.1736], + ..., + [-0.1769, -0.0758, 0.0301, ..., -0.0614, 0.1336, -0.1005], + [-0.1359, -0.1009, -0.0922, ..., -0.1536, -0.0187, -0.2211], + [-0.2092, -0.0798, -0.1604, ..., -0.2159, -0.0420, 0.1506]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-08, ..., 4.1910e-09, + 4.6566e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -6.9849e-09, + 4.1910e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 1.3970e-09, + 3.2596e-09, 4.6566e-10], + ..., + [ 1.8626e-09, 0.0000e+00, -1.5367e-08, ..., 2.7940e-09, + -2.3749e-08, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 9.3132e-09, + 2.7940e-09, 9.3132e-10], + [-2.6543e-08, 0.0000e+00, 3.4459e-08, ..., 4.6566e-10, + 1.5367e-08, -4.1164e-07]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0363, -0.0349, 0.0102, 0.0252, -0.0149, 0.0067, -0.0144, 0.0179, + 0.0078, -0.0054], device='cuda:0'), grad: tensor([-2.0629e-07, 3.2131e-08, 3.4925e-08, 4.2841e-08, 3.8408e-06, + -1.3970e-09, -3.2131e-08, 2.1514e-07, 6.4727e-08, -3.9861e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 214.08, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4571 re_mapping 0.0032 re_causal 0.0101 /// teacc 99.17 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.1293, 0.0577, 0.0792, ..., -0.0967, -0.1858, 0.0664], + [-0.1288, -0.0658, -0.1406, ..., 0.1926, -0.1198, -0.0964], + [-0.0635, 0.0163, 0.2465, ..., -0.1541, -0.1679, -0.1739], + ..., + [-0.1772, -0.0758, 0.0301, ..., -0.0612, 0.1336, -0.1006], + [-0.1362, -0.1009, -0.0922, ..., -0.1532, -0.0186, -0.2205], + [-0.2092, -0.0798, -0.1606, ..., -0.2162, -0.0421, 0.1507]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 3.2596e-09, 0.0000e+00, 6.9849e-09, ..., -7.4506e-09, + 1.2107e-08, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 3.4925e-08, ..., 1.8626e-09, + 5.4482e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -6.1933e-08, ..., 1.3970e-09, + -1.0617e-07, 5.5879e-09], + [ 4.6566e-10, 0.0000e+00, 6.9849e-09, ..., 1.3970e-09, + 1.0710e-08, 9.3132e-09], + [ 1.5367e-08, 0.0000e+00, 9.7789e-09, ..., 1.3970e-09, + 1.9092e-08, -1.8626e-08]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0384, -0.0354, 0.0102, 0.0250, -0.0149, 0.0061, -0.0117, 0.0179, + 0.0079, -0.0055], device='cuda:0'), grad: tensor([ 2.0489e-08, 2.1420e-08, 1.1502e-07, -4.9127e-07, -2.8405e-08, + 2.2817e-08, 2.7940e-08, -1.6252e-07, 4.8755e-07, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 214.30, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4544 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.20 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.1294, 0.0577, 0.0793, ..., -0.0972, -0.1859, 0.0663], + [-0.1288, -0.0658, -0.1409, ..., 0.1927, -0.1199, -0.0965], + [-0.0636, 0.0163, 0.2465, ..., -0.1544, -0.1679, -0.1745], + ..., + [-0.1774, -0.0758, 0.0301, ..., -0.0613, 0.1336, -0.1009], + [-0.1366, -0.1009, -0.0922, ..., -0.1534, -0.0186, -0.2221], + [-0.2096, -0.0798, -0.1607, ..., -0.2164, -0.0422, 0.1510]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -4.6566e-10, ..., 9.3132e-10, + 2.3283e-09, 4.6566e-10], + [ 2.3283e-09, 0.0000e+00, 1.6764e-08, ..., 1.5507e-07, + 5.8953e-07, 4.6566e-10], + [-2.7940e-09, 0.0000e+00, -1.5832e-08, ..., 2.3283e-09, + 1.9558e-08, 4.6566e-10], + ..., + [ 1.8626e-09, 0.0000e+00, -3.4925e-08, ..., -1.9511e-07, + -7.5810e-07, 7.9162e-09], + [ 1.3970e-09, 0.0000e+00, 1.0710e-08, ..., 1.2107e-08, + 5.8673e-08, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 6.5193e-09, ..., 9.7789e-09, + 4.0513e-08, -1.6764e-08]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0388, -0.0355, 0.0102, 0.0248, -0.0146, 0.0061, -0.0113, 0.0179, + 0.0079, -0.0056], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.8170e-06, -3.3528e-08, 3.0175e-07, 3.9116e-08, + -1.1548e-07, -5.1223e-09, -2.2836e-06, 1.7835e-07, 9.4064e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 214.14, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4925 re_mapping 0.0033 re_causal 0.0102 /// teacc 99.16 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.1295, 0.0577, 0.0796, ..., -0.0971, -0.1859, 0.0667], + [-0.1288, -0.0658, -0.1412, ..., 0.1928, -0.1200, -0.0966], + [-0.0637, 0.0163, 0.2466, ..., -0.1548, -0.1679, -0.1747], + ..., + [-0.1776, -0.0758, 0.0301, ..., -0.0614, 0.1336, -0.1010], + [-0.1358, -0.1009, -0.0922, ..., -0.1537, -0.0186, -0.2231], + [-0.2096, -0.0798, -0.1611, ..., -0.2168, -0.0424, 0.1508]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 1.3970e-09, 0.0000e+00, 2.5425e-07, ..., 9.2667e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.9337e-07, ..., -1.0710e-07, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-09]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0386, -0.0356, 0.0103, 0.0252, -0.0147, 0.0056, -0.0111, 0.0179, + 0.0079, -0.0059], device='cuda:0'), grad: tensor([ 6.9849e-09, 6.0257e-07, -6.8406e-07, 7.9162e-09, 7.4506e-09, + -2.5146e-08, 7.1246e-08, 2.2817e-08, 3.2596e-09, -4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 214.10, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4765 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.09 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.1296, 0.0577, 0.0798, ..., -0.0971, -0.1861, 0.0665], + [-0.1289, -0.0658, -0.1415, ..., 0.1929, -0.1201, -0.0975], + [-0.0635, 0.0163, 0.2466, ..., -0.1553, -0.1680, -0.1767], + ..., + [-0.1778, -0.0759, 0.0301, ..., -0.0614, 0.1336, -0.1013], + [-0.1361, -0.1010, -0.0922, ..., -0.1541, -0.0186, -0.2205], + [-0.2097, -0.0798, -0.1613, ..., -0.2170, -0.0425, 0.1511]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -3.0734e-08, ..., -6.0536e-09, + 4.6566e-10, -6.6124e-08], + [-5.1223e-09, 0.0000e+00, 6.9849e-09, ..., -2.0023e-08, + 9.3132e-10, 1.8626e-08], + [ 1.8626e-09, 0.0000e+00, -4.6566e-10, ..., 8.8476e-09, + 1.3970e-09, 6.0536e-09], + ..., + [ 1.3970e-09, 0.0000e+00, -4.1910e-09, ..., 8.3819e-09, + -1.0710e-08, 1.4901e-08], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 4.1910e-09, + 9.3132e-10, 1.4435e-08], + [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 9.3132e-10, + 7.4506e-09, -3.2596e-09]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0394, -0.0359, 0.0101, 0.0247, -0.0147, 0.0088, -0.0141, 0.0179, + 0.0080, -0.0059], device='cuda:0'), grad: tensor([-1.4622e-07, 1.3970e-09, 2.6077e-08, 2.1886e-08, 3.3062e-08, + 1.8626e-09, 1.3970e-09, 5.3085e-08, 1.6298e-08, -1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 214.68, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4657 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.14 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.1303, 0.0578, 0.0798, ..., -0.0971, -0.1862, 0.0665], + [-0.1290, -0.0658, -0.1416, ..., 0.1944, -0.1201, -0.0973], + [-0.0631, 0.0163, 0.2466, ..., -0.1554, -0.1680, -0.1768], + ..., + [-0.1782, -0.0759, 0.0301, ..., -0.0628, 0.1336, -0.1018], + [-0.1369, -0.1010, -0.0922, ..., -0.1542, -0.0186, -0.2206], + [-0.2098, -0.0799, -0.1615, ..., -0.2176, -0.0427, 0.1515]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.3970e-09, ..., 9.3132e-10, + 4.6566e-09, 1.1176e-08], + [ 7.4506e-09, 0.0000e+00, 9.3132e-10, ..., 3.2596e-09, + 1.8626e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-09, 2.3283e-09], + ..., + [ 2.3283e-09, 0.0000e+00, -4.6566e-10, ..., 0.0000e+00, + 1.8626e-09, 1.4435e-08], + [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., -1.8626e-09, + 9.3132e-10, 4.1910e-09], + [ 8.3819e-09, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 1.3970e-08, -1.6298e-08]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0394, -0.0353, 0.0102, 0.0246, -0.0149, 0.0088, -0.0140, 0.0179, + 0.0080, -0.0057], device='cuda:0'), grad: tensor([ 3.9116e-08, 4.4703e-08, 1.0245e-08, 7.9628e-08, -8.3819e-09, + -1.9511e-07, 2.6543e-08, 4.1444e-08, -2.2817e-08, -7.9162e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 214.39, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4869 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.1305, 0.0578, 0.0800, ..., -0.0971, -0.1864, 0.0667], + [-0.1293, -0.0658, -0.1417, ..., 0.1946, -0.1202, -0.0974], + [-0.0632, 0.0164, 0.2467, ..., -0.1555, -0.1680, -0.1771], + ..., + [-0.1791, -0.0759, 0.0301, ..., -0.0631, 0.1336, -0.1027], + [-0.1382, -0.1010, -0.0922, ..., -0.1545, -0.0186, -0.2208], + [-0.2101, -0.0799, -0.1616, ..., -0.2177, -0.0427, 0.1520]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, 0.0000e+00, -9.3132e-09, ..., -4.6566e-10, + 1.0245e-08, -5.5879e-09], + [ 1.5693e-07, 0.0000e+00, 3.4925e-08, ..., -1.3504e-08, + 1.2480e-07, 2.3283e-09], + [ 1.4901e-08, 0.0000e+00, 2.6077e-08, ..., 1.8626e-09, + 4.6566e-08, 2.3283e-09], + ..., + [ 2.1420e-07, 0.0000e+00, -1.4529e-07, ..., 5.5879e-09, + -9.1456e-07, 3.4925e-08], + [ 4.0047e-08, 0.0000e+00, 1.1176e-08, ..., 3.7253e-09, + 2.4214e-08, 1.3970e-09], + [ 1.4137e-06, 0.0000e+00, 1.7136e-07, ..., 9.3132e-10, + 6.4913e-07, -6.8918e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0393, -0.0354, 0.0102, 0.0247, -0.0145, 0.0088, -0.0142, 0.0179, + 0.0080, -0.0058], device='cuda:0'), grad: tensor([ 6.3796e-08, 8.6194e-07, 2.3190e-07, 3.6322e-08, -3.8184e-06, + 1.0431e-07, 9.5461e-08, -3.4198e-06, 2.0768e-07, 5.6587e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 214.09, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4513 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.14 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.1305, 0.0578, 0.0802, ..., -0.0972, -0.1865, 0.0667], + [-0.1294, -0.0658, -0.1419, ..., 0.1946, -0.1203, -0.0985], + [-0.0633, 0.0164, 0.2470, ..., -0.1556, -0.1681, -0.1776], + ..., + [-0.1802, -0.0759, 0.0300, ..., -0.0631, 0.1336, -0.1030], + [-0.1387, -0.1010, -0.0922, ..., -0.1548, -0.0186, -0.2211], + [-0.2107, -0.0799, -0.1619, ..., -0.2178, -0.0428, 0.1526]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.7742e-08, ..., -5.5879e-09, + 4.6566e-10, -7.4971e-08], + [ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., -2.3283e-09, + 9.3132e-09, 2.7940e-09], + [-9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, + 3.2596e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -3.1199e-08, ..., 9.3132e-10, + -5.9605e-08, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, + 2.7940e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 5.7742e-08, ..., 5.1223e-09, + 3.3993e-08, 5.2620e-08]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0393, -0.0356, 0.0105, 0.0245, -0.0139, 0.0088, -0.0141, 0.0179, + 0.0080, -0.0061], device='cuda:0'), grad: tensor([-2.0396e-07, 3.3528e-08, 1.0710e-08, -1.6950e-07, 1.1642e-08, + 1.2992e-07, 2.3283e-08, -1.2154e-07, 5.8673e-08, 2.4121e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 214.30, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4797 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.14 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.1309, 0.0578, 0.0805, ..., -0.0971, -0.1865, 0.0670], + [-0.1294, -0.0658, -0.1420, ..., 0.1963, -0.1203, -0.0989], + [-0.0633, 0.0164, 0.2471, ..., -0.1559, -0.1681, -0.1797], + ..., + [-0.1805, -0.0759, 0.0300, ..., -0.0648, 0.1336, -0.1033], + [-0.1391, -0.1010, -0.0923, ..., -0.1551, -0.0186, -0.2214], + [-0.2110, -0.0799, -0.1621, ..., -0.2185, -0.0431, 0.1529]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -4.6566e-10, + 4.6566e-10, 9.3132e-10], + [-0.0000e+00, 0.0000e+00, 3.6322e-08, ..., -4.1444e-08, + 2.3749e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, -1.1455e-07, ..., 2.3283e-09, + 8.3819e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., 1.6298e-08, + -6.7055e-08, 4.7963e-08], + [ 0.0000e+00, 0.0000e+00, 2.1886e-08, ..., 6.9849e-09, + 8.8476e-09, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 3.7719e-08, ..., 2.3283e-09, + 1.7229e-08, -1.2247e-07]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0391, -0.0346, 0.0106, 0.0241, -0.0138, 0.0089, -0.0141, 0.0178, + 0.0080, -0.0063], device='cuda:0'), grad: tensor([ 4.7032e-08, 4.2841e-08, -1.3085e-07, -1.5460e-07, 3.7299e-07, + 8.2888e-08, 1.7229e-08, 2.0908e-07, 2.0163e-07, -6.8778e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 214.47, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4838 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.18 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.1324, 0.0574, 0.0809, ..., -0.0971, -0.1869, 0.0674], + [-0.1291, -0.0659, -0.1421, ..., 0.1968, -0.1209, -0.0994], + [-0.0634, 0.0163, 0.2472, ..., -0.1563, -0.1681, -0.1803], + ..., + [-0.1807, -0.0763, 0.0300, ..., -0.0649, 0.1337, -0.1035], + [-0.1401, -0.1011, -0.0923, ..., -0.1561, -0.0187, -0.2217], + [-0.2111, -0.0800, -0.1625, ..., -0.2190, -0.0435, 0.1533]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -2.3283e-09], + [ 9.3132e-10, 0.0000e+00, 1.4901e-08, ..., -1.5832e-08, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, -1.0710e-07, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 1.5367e-08, ..., 1.1642e-08, + 0.0000e+00, 1.3970e-09], + [-4.6566e-10, 0.0000e+00, 4.3306e-08, ..., 4.6566e-10, + -0.0000e+00, 1.3970e-09], + [ 3.1665e-08, 0.0000e+00, 3.7253e-09, ..., 4.6566e-10, + 0.0000e+00, -8.8476e-09]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0389, -0.0347, 0.0106, 0.0243, -0.0143, 0.0088, -0.0141, 0.0178, + 0.0080, -0.0062], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.0245e-08, -1.9837e-07, -6.9849e-08, -1.0710e-08, + 5.9605e-08, -4.6566e-10, 8.5682e-08, 9.0804e-08, 4.4238e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 214.28, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4841 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.15 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.1330, 0.0573, 0.0813, ..., -0.0971, -0.1871, 0.0676], + [-0.1282, -0.0659, -0.1423, ..., 0.1978, -0.1209, -0.0996], + [-0.0637, 0.0163, 0.2472, ..., -0.1566, -0.1681, -0.1805], + ..., + [-0.1814, -0.0763, 0.0300, ..., -0.0651, 0.1337, -0.1042], + [-0.1412, -0.1011, -0.0923, ..., -0.1568, -0.0187, -0.2217], + [-0.2093, -0.0800, -0.1628, ..., -0.2193, -0.0435, 0.1543]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 0.0000e+00, -1.3970e-09, ..., 1.3970e-08, + 1.8626e-09, -1.8626e-09], + [ 4.1910e-09, 0.0000e+00, 1.2573e-08, ..., -6.0536e-09, + 1.6764e-08, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 2.0489e-08, ..., 4.1910e-09, + 8.3819e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -5.4482e-08, ..., 2.7940e-09, + -8.1956e-08, 9.3132e-10], + [ 3.0734e-08, 0.0000e+00, -2.0955e-08, ..., 2.6543e-08, + 5.5879e-09, 0.0000e+00], + [ 5.1223e-09, 0.0000e+00, 3.8184e-08, ..., 9.3132e-10, + 9.0338e-08, -1.8626e-09]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0389, -0.0337, 0.0106, 0.0260, -0.0177, 0.0084, -0.0141, 0.0178, + 0.0080, -0.0034], device='cuda:0'), grad: tensor([ 7.5437e-08, 1.1874e-07, 8.3447e-07, 1.6904e-07, 1.6065e-07, + -7.9628e-08, -3.0920e-07, -6.2864e-08, -1.1474e-06, 2.5099e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 214.25, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4790 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.12 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.1334, 0.0573, 0.0816, ..., -0.0972, -0.1875, 0.0677], + [-0.1262, -0.0659, -0.1428, ..., 0.1990, -0.1211, -0.0997], + [-0.0639, 0.0163, 0.2472, ..., -0.1564, -0.1682, -0.1812], + ..., + [-0.1818, -0.0763, 0.0300, ..., -0.0652, 0.1337, -0.1043], + [-0.1399, -0.1011, -0.0923, ..., -0.1572, -0.0187, -0.2218], + [-0.2094, -0.0800, -0.1631, ..., -0.2196, -0.0437, 0.1543]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -4.4703e-08, ..., -2.7008e-08, + 4.1910e-09, -7.5437e-08], + [ 1.9092e-08, 0.0000e+00, 2.2724e-07, ..., 9.3132e-10, + 1.4575e-07, 1.8626e-09], + [ 1.2107e-08, 0.0000e+00, 3.6135e-07, ..., 4.6566e-10, + 2.7101e-07, 4.6566e-10], + ..., + [-3.2131e-08, 0.0000e+00, -1.1120e-06, ..., 4.6566e-10, + -8.1770e-07, 1.3970e-09], + [ 5.5879e-09, 0.0000e+00, 1.0012e-07, ..., 4.6566e-10, + 8.0559e-08, 1.3970e-09], + [ 4.1444e-08, 0.0000e+00, 9.8720e-08, ..., 1.9092e-08, + 4.7032e-08, 5.1223e-08]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0388, -0.0322, 0.0106, 0.0262, -0.0187, 0.0084, -0.0141, 0.0178, + 0.0080, -0.0035], device='cuda:0'), grad: tensor([-2.3749e-07, 6.0536e-07, 1.5069e-06, 2.5425e-06, 5.2620e-07, + 5.1223e-08, 9.5461e-08, -2.6412e-06, -2.9244e-06, 4.6659e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 214.17, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4667 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.09 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.1340, 0.0573, 0.0832, ..., -0.0970, -0.1876, 0.0697], + [-0.1262, -0.0659, -0.1433, ..., 0.2009, -0.1211, -0.1003], + [-0.0644, 0.0163, 0.2471, ..., -0.1569, -0.1683, -0.1823], + ..., + [-0.1821, -0.0763, 0.0301, ..., -0.0673, 0.1337, -0.1047], + [-0.1414, -0.1011, -0.0923, ..., -0.1579, -0.0187, -0.2221], + [-0.2095, -0.0800, -0.1647, ..., -0.2213, -0.0437, 0.1527]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -9.3132e-10, ..., 1.9092e-08, + 2.3283e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 6.9849e-09, ..., 5.5879e-09, + 6.5193e-09, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 1.3504e-08, ..., 4.6566e-09, + 1.2107e-08, 4.6566e-10], + ..., + [ 1.3970e-09, 0.0000e+00, -4.2841e-08, ..., 9.3132e-10, + -4.0047e-08, 1.6764e-08], + [ 4.6566e-10, 0.0000e+00, 3.7253e-09, ..., 1.5832e-08, + 2.7940e-09, 3.2596e-09], + [ 8.2422e-08, 0.0000e+00, 1.3504e-08, ..., 4.6566e-10, + 1.4435e-08, -4.0513e-08]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0378, -0.0315, 0.0105, 0.0261, -0.0188, 0.0083, -0.0140, 0.0177, + 0.0080, -0.0039], device='cuda:0'), grad: tensor([ 1.0151e-07, 5.4482e-08, 5.7276e-08, -1.4901e-08, -8.3819e-08, + 4.5169e-08, -2.6077e-07, -2.5611e-08, 9.5461e-08, 4.3772e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 214.31, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4943 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.09 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.1363, 0.0573, 0.0838, ..., -0.0971, -0.1877, 0.0703], + [-0.1262, -0.0659, -0.1436, ..., 0.2010, -0.1213, -0.1003], + [-0.0646, 0.0163, 0.2472, ..., -0.1573, -0.1683, -0.1826], + ..., + [-0.1825, -0.0763, 0.0301, ..., -0.0674, 0.1337, -0.1048], + [-0.1426, -0.1011, -0.0923, ..., -0.1582, -0.0187, -0.2223], + [-0.2095, -0.0800, -0.1655, ..., -0.2215, -0.0438, 0.1523]], + device='cuda:0'), grad: tensor([[ 6.8545e-07, 0.0000e+00, 2.0023e-08, ..., 7.5847e-06, + 3.4925e-09, 2.1816e-07], + [ 1.6531e-08, 0.0000e+00, 2.2585e-08, ..., 4.5635e-08, + 5.5879e-09, 1.9558e-08], + [-1.3970e-08, -0.0000e+00, -1.5134e-08, ..., 3.0966e-08, + 5.5879e-09, 1.7462e-08], + ..., + [ 2.3283e-10, 0.0000e+00, 3.4552e-07, ..., 3.7253e-09, + 6.0489e-07, 1.8161e-08], + [ 8.3819e-09, 0.0000e+00, 2.8638e-08, ..., 2.0606e-07, + 3.9581e-09, 5.4715e-08], + [ 5.5879e-09, 0.0000e+00, -4.7497e-08, ..., 8.0094e-08, + 3.0268e-09, -4.7148e-07]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0376, -0.0316, 0.0104, 0.0261, -0.0189, 0.0083, -0.0138, 0.0177, + 0.0080, -0.0040], device='cuda:0'), grad: tensor([ 3.0965e-05, 2.9919e-07, 1.6019e-07, -4.6790e-06, 1.6475e-06, + 8.0373e-07, -3.3855e-05, 4.4182e-06, 1.0375e-06, -7.9768e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 214.69, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4550 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.1385, 0.0573, 0.0841, ..., -0.0973, -0.1878, 0.0706], + [-0.1263, -0.0659, -0.1437, ..., 0.2010, -0.1214, -0.1007], + [-0.0651, 0.0163, 0.2471, ..., -0.1575, -0.1685, -0.1830], + ..., + [-0.1829, -0.0763, 0.0301, ..., -0.0675, 0.1337, -0.1049], + [-0.1406, -0.1011, -0.0923, ..., -0.1582, -0.0187, -0.2222], + [-0.2096, -0.0800, -0.1657, ..., -0.2216, -0.0439, 0.1523]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.3551e-09, ..., 0.0000e+00, + 0.0000e+00, -3.9581e-09], + [ 2.3283e-10, 0.0000e+00, 9.5461e-09, ..., 2.3283e-10, + 9.3132e-09, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 2.3283e-10, + 2.3982e-08, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., 2.3283e-09, + -4.1910e-08, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 6.9849e-10, + 2.3283e-09, 6.9849e-10], + [ 2.3283e-09, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 5.1223e-09, -2.7940e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0376, -0.0317, 0.0104, 0.0257, -0.0188, 0.0083, -0.0167, 0.0177, + 0.0095, -0.0041], device='cuda:0'), grad: tensor([-1.7462e-08, 2.3516e-08, 4.7497e-08, 9.3132e-09, 4.6566e-10, + 3.7253e-09, 2.3283e-09, -7.9861e-08, 1.0012e-08, 1.2806e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 214.34, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4664 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.13 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.1401, 0.0573, 0.0843, ..., -0.0973, -0.1879, 0.0707], + [-0.1265, -0.0659, -0.1442, ..., 0.2011, -0.1217, -0.1009], + [-0.0672, 0.0163, 0.2471, ..., -0.1583, -0.1685, -0.1834], + ..., + [-0.1833, -0.0763, 0.0302, ..., -0.0677, 0.1337, -0.1051], + [-0.1386, -0.1011, -0.0923, ..., -0.1571, -0.0187, -0.2220], + [-0.2097, -0.0800, -0.1664, ..., -0.2220, -0.0443, 0.1522]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -1.1176e-08, + 1.8626e-09, -1.7229e-08], + [ 9.3132e-10, 0.0000e+00, 1.7183e-07, ..., 7.3109e-08, + 2.3190e-07, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 1.0850e-07, ..., 0.0000e+00, + 8.7544e-08, 0.0000e+00], + ..., + [-1.1642e-08, 0.0000e+00, -4.6100e-07, ..., -6.5658e-08, + -5.5321e-07, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 1.7695e-08, ..., -1.1642e-08, + 2.1886e-08, 1.3970e-09], + [ 7.4506e-09, 0.0000e+00, 7.8697e-08, ..., 7.4506e-09, + 8.2888e-08, -8.8476e-09]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0376, -0.0319, 0.0102, 0.0260, -0.0188, 0.0082, -0.0167, 0.0178, + 0.0096, -0.0043], device='cuda:0'), grad: tensor([-4.4238e-08, 6.6962e-07, 2.2212e-07, 2.6124e-07, 6.0536e-08, + 2.7474e-08, 3.3528e-08, -1.4035e-06, -4.6566e-08, 2.1700e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 214.44, cls_loss 0.0025 cls_loss_mapping 0.0035 cls_loss_causal 0.4780 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.14 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.1417, 0.0573, 0.0848, ..., -0.0973, -0.1881, 0.0715], + [-0.1264, -0.0659, -0.1460, ..., 0.2034, -0.1200, -0.1011], + [-0.0698, 0.0163, 0.2469, ..., -0.1608, -0.1687, -0.1839], + ..., + [-0.1872, -0.0763, 0.0307, ..., -0.0698, 0.1316, -0.1053], + [-0.1392, -0.1011, -0.0923, ..., -0.1580, -0.0187, -0.2223], + [-0.2098, -0.0800, -0.1671, ..., -0.2226, -0.0452, 0.1520]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, -1.3970e-09, ..., 2.1886e-08, + 2.7940e-09, 2.3283e-09], + [ 4.0513e-08, 0.0000e+00, 1.1176e-08, ..., 2.3283e-09, + 2.6543e-08, 1.3970e-09], + [ 1.0245e-08, 0.0000e+00, 3.2596e-09, ..., 9.7789e-09, + 6.9849e-09, 9.3132e-10], + ..., + [ 1.0245e-08, 0.0000e+00, -4.0047e-08, ..., 4.6566e-10, + -9.5926e-08, 1.0710e-08], + [ 1.5832e-08, 0.0000e+00, 1.3970e-09, ..., 1.6764e-08, + 4.6566e-09, 2.3283e-09], + [ 1.8673e-07, 0.0000e+00, 2.0023e-08, ..., 4.6566e-10, + -4.0978e-08, -1.9139e-07]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0373, -0.0299, 0.0098, 0.0263, -0.0189, 0.0081, -0.0167, 0.0158, + 0.0095, -0.0047], device='cuda:0'), grad: tensor([ 9.3132e-08, 1.6484e-07, 7.6368e-08, 1.0617e-07, 2.4680e-08, + -6.6590e-08, -1.4435e-07, -1.6857e-07, 9.6392e-08, -1.6950e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 214.44, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.4802 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.03 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.1421, 0.0573, 0.0849, ..., -0.0973, -0.1885, 0.0716], + [-0.1265, -0.0659, -0.1480, ..., 0.2018, -0.1199, -0.1013], + [-0.0734, 0.0164, 0.2466, ..., -0.1645, -0.1691, -0.1842], + ..., + [-0.1885, -0.0765, 0.0316, ..., -0.0683, 0.1317, -0.1056], + [-0.1408, -0.1011, -0.0923, ..., -0.1594, -0.0187, -0.2235], + [-0.2099, -0.0800, -0.1673, ..., -0.2233, -0.0455, 0.1527]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -3.0268e-08, ..., 4.1910e-09, + 0.0000e+00, -2.8405e-08], + [ 2.7940e-09, 0.0000e+00, 4.6566e-10, ..., -7.7300e-08, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., 3.3062e-08, + 0.0000e+00, 4.6566e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 3.2131e-08, + 4.6566e-10, 5.5879e-09], + [ 4.6566e-10, 0.0000e+00, 1.4435e-08, ..., 3.7253e-09, + -9.3132e-10, 5.1223e-09], + [ 1.1642e-08, 0.0000e+00, 1.5367e-08, ..., 1.8626e-09, + 0.0000e+00, -1.1642e-08]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0373, -0.0301, 0.0088, 0.0286, -0.0192, 0.0058, -0.0167, 0.0160, + 0.0095, -0.0047], device='cuda:0'), grad: tensor([-5.3085e-08, -1.5041e-07, 3.3062e-08, 2.6077e-08, 1.9558e-08, + 6.9849e-09, 1.1642e-08, 9.9186e-08, 3.7719e-08, -2.0023e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 214.12, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4395 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.15 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.1425, 0.0573, 0.0850, ..., -0.0974, -0.1887, 0.0716], + [-0.1267, -0.0659, -0.1480, ..., 0.2018, -0.1200, -0.1013], + [-0.0735, 0.0164, 0.2462, ..., -0.1648, -0.1697, -0.1845], + ..., + [-0.1890, -0.0765, 0.0318, ..., -0.0683, 0.1317, -0.1057], + [-0.1410, -0.1012, -0.0923, ..., -0.1609, -0.0187, -0.2236], + [-0.2100, -0.0801, -0.1675, ..., -0.2237, -0.0457, 0.1529]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -1.1176e-08, ..., 1.8626e-09, + 0.0000e+00, -1.3970e-09], + [ 1.5367e-08, 0.0000e+00, 3.2596e-09, ..., -7.9162e-09, + 1.8626e-09, 1.0245e-08], + [ 3.2596e-09, 0.0000e+00, -3.7253e-09, ..., 2.7940e-09, + 4.6566e-10, 9.3132e-10], + ..., + [ 7.9162e-09, 0.0000e+00, -3.7253e-09, ..., 5.1223e-09, + -5.1223e-09, 2.3283e-08], + [ 9.7789e-09, 0.0000e+00, 3.2596e-09, ..., 1.2573e-08, + -1.3039e-08, 7.9162e-09], + [ 3.5204e-07, 0.0000e+00, 4.8429e-08, ..., 0.0000e+00, + 1.1176e-08, -2.3609e-07]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0374, -0.0301, 0.0083, 0.0286, -0.0191, 0.0058, -0.0167, 0.0160, + 0.0095, -0.0048], device='cuda:0'), grad: tensor([ 2.3283e-08, 9.1270e-08, 4.5355e-07, 2.0443e-07, -2.5844e-07, + 1.1176e-07, 2.3609e-07, 1.8254e-07, -1.0766e-06, 4.0978e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 214.55, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4662 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.08 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.1427, 0.0573, 0.0850, ..., -0.0975, -0.1888, 0.0716], + [-0.1268, -0.0660, -0.1481, ..., 0.2018, -0.1200, -0.1024], + [-0.0736, 0.0165, 0.2464, ..., -0.1649, -0.1697, -0.1876], + ..., + [-0.1899, -0.0765, 0.0317, ..., -0.0683, 0.1317, -0.1061], + [-0.1417, -0.1012, -0.0924, ..., -0.1616, -0.0187, -0.2240], + [-0.2101, -0.0801, -0.1677, ..., -0.2233, -0.0459, 0.1534]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 9.3132e-10], + [ 3.7253e-09, 0.0000e+00, 5.5879e-09, ..., -9.3132e-10, + 7.4506e-09, 9.3132e-10], + [ 2.7940e-09, 0.0000e+00, -4.6566e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + -1.1176e-08, 1.8626e-09], + [ 2.7940e-09, 0.0000e+00, 2.7940e-09, ..., 4.6566e-09, + 4.6566e-09, 2.7940e-09], + [-4.1164e-07, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, -3.8184e-07]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0375, -0.0301, 0.0084, 0.0285, -0.0191, 0.0057, -0.0167, 0.0160, + 0.0095, -0.0045], device='cuda:0'), grad: tensor([ 1.9558e-08, 2.4214e-08, 0.0000e+00, 1.1083e-07, 1.4054e-06, + -1.0990e-07, -1.3039e-08, -1.3039e-08, 9.3132e-10, -1.4286e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 372---------------------------------------------------- +epoch 372, time 231.60, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4416 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.26 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.1441, 0.0564, 0.0853, ..., -0.0976, -0.1888, 0.0717], + [-0.1269, -0.0660, -0.1481, ..., 0.2019, -0.1200, -0.1025], + [-0.0737, 0.0165, 0.2464, ..., -0.1668, -0.1697, -0.1881], + ..., + [-0.1909, -0.0766, 0.0317, ..., -0.0683, 0.1317, -0.1062], + [-0.1411, -0.1012, -0.0924, ..., -0.1618, -0.0187, -0.2240], + [-0.2103, -0.0801, -0.1680, ..., -0.2233, -0.0460, 0.1534]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, -2.7940e-09], + [-1.8626e-09, 1.8626e-09, 9.3132e-09, ..., -2.6077e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, -9.3132e-10, -2.2352e-08, ..., 1.3970e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 6.5193e-09, ..., 1.2107e-08, + 0.0000e+00, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 6.5193e-09, ..., 1.0245e-08, + 0.0000e+00, 0.0000e+00], + [ 1.1176e-08, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 9.3132e-10, -4.6566e-09]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0376, -0.0301, 0.0083, 0.0286, -0.0190, 0.0057, -0.0167, 0.0160, + 0.0095, -0.0047], device='cuda:0'), grad: tensor([-3.7253e-09, -5.6811e-08, 2.7940e-09, 2.6077e-08, 1.5553e-07, + 4.6566e-09, -1.8720e-07, 5.4017e-08, 3.9116e-08, -1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 214.32, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4753 re_mapping 0.0032 re_causal 0.0101 /// teacc 99.17 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.1434, 0.0564, 0.0858, ..., -0.0975, -0.1884, 0.0721], + [-0.1269, -0.0660, -0.1481, ..., 0.2019, -0.1200, -0.1026], + [-0.0742, 0.0166, 0.2465, ..., -0.1668, -0.1698, -0.1883], + ..., + [-0.1914, -0.0766, 0.0318, ..., -0.0683, 0.1318, -0.1066], + [-0.1412, -0.1012, -0.0924, ..., -0.1621, -0.0187, -0.2244], + [-0.2106, -0.0801, -0.1685, ..., -0.2235, -0.0463, 0.1536]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 0.0000e+00, -1.8626e-09, ..., 3.2596e-08, + 9.3132e-10, -2.7940e-09], + [ 7.3574e-08, 0.0000e+00, 1.3039e-08, ..., -9.3132e-10, + 7.4506e-09, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, -4.1910e-08, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 6.5193e-09, 0.0000e+00, -6.5193e-09, ..., 4.6566e-09, + -1.4901e-08, 1.8626e-09], + [ 2.4214e-08, 0.0000e+00, 1.5832e-08, ..., 2.0489e-08, + 1.8626e-09, 9.3132e-10], + [ 6.6124e-08, 0.0000e+00, 7.4506e-09, ..., 2.7940e-09, + 4.6566e-09, -9.3132e-10]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0373, -0.0301, 0.0083, 0.0286, -0.0190, 0.0058, -0.0167, 0.0160, + 0.0095, -0.0048], device='cuda:0'), grad: tensor([ 1.3225e-07, 1.2200e-07, -1.0617e-07, 3.4459e-08, 7.8231e-08, + 8.6613e-08, -5.9139e-07, 1.8626e-09, 1.4435e-07, 9.4064e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 214.18, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4706 re_mapping 0.0032 re_causal 0.0101 /// teacc 99.15 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.1438, 0.0561, 0.0862, ..., -0.0978, -0.1890, 0.0723], + [-0.1270, -0.0660, -0.1482, ..., 0.2019, -0.1200, -0.1027], + [-0.0748, 0.0166, 0.2464, ..., -0.1679, -0.1701, -0.1905], + ..., + [-0.1923, -0.0766, 0.0318, ..., -0.0683, 0.1318, -0.1068], + [-0.1417, -0.1013, -0.0924, ..., -0.1646, -0.0187, -0.2249], + [-0.2107, -0.0801, -0.1688, ..., -0.2237, -0.0464, 0.1537]], + device='cuda:0'), grad: tensor([[-1.8626e-09, 0.0000e+00, -0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, -2.7940e-09], + [ 1.8626e-09, 0.0000e+00, 1.4901e-08, ..., -5.8673e-08, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.5367e-07, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 1.0617e-07, ..., 5.4948e-08, + 0.0000e+00, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 2.4214e-08, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, -9.3132e-10]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0374, -0.0301, 0.0080, 0.0286, -0.0190, 0.0058, -0.0165, 0.0160, + 0.0094, -0.0048], device='cuda:0'), grad: tensor([-1.3970e-08, -1.8068e-07, -1.9744e-07, -9.2201e-08, 3.7253e-09, + 4.0047e-08, 0.0000e+00, 3.8184e-07, 4.6566e-08, 2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 214.35, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4398 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.22 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.1440, 0.0561, 0.0864, ..., -0.0978, -0.1891, 0.0725], + [-0.1271, -0.0661, -0.1483, ..., 0.2019, -0.1200, -0.1030], + [-0.0747, 0.0167, 0.2466, ..., -0.1682, -0.1701, -0.1905], + ..., + [-0.1949, -0.0767, 0.0319, ..., -0.0683, 0.1318, -0.1070], + [-0.1424, -0.1013, -0.0925, ..., -0.1650, -0.0187, -0.2251], + [-0.2111, -0.0801, -0.1693, ..., -0.2238, -0.0467, 0.1538]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -1.3970e-07, ..., 1.8626e-09, + 3.7253e-09, -2.3097e-07], + [ 4.1910e-08, 0.0000e+00, 1.9651e-07, ..., 7.6368e-08, + 4.3027e-07, 2.4308e-07], + [ 1.8626e-09, 0.0000e+00, -1.9092e-07, ..., 9.3132e-10, + 3.7439e-07, 2.7940e-09], + ..., + [ 3.7253e-09, 0.0000e+00, -6.9290e-07, ..., 3.7253e-09, + -2.0824e-06, 5.5879e-09], + [ 8.3819e-09, 0.0000e+00, 2.1420e-08, ..., 3.7253e-09, + 5.0291e-08, 2.7940e-09], + [ 1.1455e-07, 0.0000e+00, 1.8440e-07, ..., 7.4506e-09, + 4.4703e-08, 2.3283e-07]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0373, -0.0301, 0.0081, 0.0286, -0.0188, 0.0058, -0.0165, 0.0160, + 0.0094, -0.0050], device='cuda:0'), grad: tensor([-3.7998e-07, 5.6401e-06, 5.9139e-07, -7.3388e-06, -1.2387e-07, + 6.1132e-06, 1.3039e-08, -5.9195e-06, 2.2445e-07, 1.1474e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 214.41, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4417 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.07 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.1449, 0.0560, 0.0864, ..., -0.0978, -0.1894, 0.0727], + [-0.1276, -0.0661, -0.1486, ..., 0.2019, -0.1201, -0.1053], + [-0.0752, 0.0170, 0.2470, ..., -0.1684, -0.1701, -0.1907], + ..., + [-0.1988, -0.0769, 0.0319, ..., -0.0683, 0.1318, -0.1076], + [-0.1435, -0.1014, -0.0925, ..., -0.1653, -0.0187, -0.2225], + [-0.2115, -0.0801, -0.1698, ..., -0.2241, -0.0475, 0.1558]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 1.8626e-09, + 1.0245e-08, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.2666e-07, ..., -2.3283e-08, + 7.2643e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.1130e-07, ..., 2.7940e-09, + 2.0117e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -9.5554e-07, ..., 1.5832e-08, + -4.1164e-07, 5.5879e-09], + [-9.3132e-10, 0.0000e+00, 1.0245e-08, ..., 1.8626e-09, + 8.3819e-09, 9.3132e-10], + [ 5.5879e-09, 0.0000e+00, 5.7742e-08, ..., 9.3132e-10, + 3.0734e-08, 6.5193e-09]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0373, -0.0301, 0.0086, 0.0285, -0.0195, 0.0051, -0.0165, 0.0160, + 0.0096, -0.0040], device='cuda:0'), grad: tensor([ 6.4261e-08, 3.2503e-07, 1.3467e-06, 4.4797e-07, 2.9802e-08, + 1.6391e-07, -9.9652e-08, -2.4401e-06, -3.3528e-08, 2.0582e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 214.09, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4560 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.17 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.1473, 0.0560, 0.0860, ..., -0.0986, -0.1897, 0.0726], + [-0.1278, -0.0660, -0.1487, ..., 0.2019, -0.1201, -0.1062], + [-0.0751, 0.0169, 0.2470, ..., -0.1687, -0.1703, -0.1911], + ..., + [-0.1999, -0.0772, 0.0318, ..., -0.0683, 0.1316, -0.1079], + [-0.1436, -0.1014, -0.0925, ..., -0.1656, -0.0187, -0.2230], + [-0.2128, -0.0801, -0.1704, ..., -0.2269, -0.0480, 0.1558]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 1.0245e-08, ..., 1.3970e-08, + 0.0000e+00, -2.7940e-09], + [ 2.7940e-09, 0.0000e+00, 2.9802e-08, ..., -7.4506e-09, + 1.3039e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -1.9837e-07, ..., 4.6566e-09, + 1.4901e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 2.8051e-06, ..., 1.8626e-09, + 2.2184e-06, 9.3132e-10], + [-9.3132e-10, 0.0000e+00, 1.0245e-07, ..., 9.3132e-10, + 7.4506e-09, 0.0000e+00], + [ 3.3528e-08, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 8.3819e-09, 2.7940e-09]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0379, -0.0302, 0.0084, 0.0307, -0.0188, 0.0053, -0.0163, 0.0159, + 0.0096, -0.0050], device='cuda:0'), grad: tensor([ 1.4529e-07, 1.7136e-07, -1.6009e-06, -1.1563e-05, 1.6019e-07, + 4.1910e-08, 1.0431e-07, 1.1638e-05, 7.5717e-07, 1.2852e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 214.31, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4545 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.18 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.1472, 0.0560, 0.0864, ..., -0.0987, -0.1900, 0.0729], + [-0.1279, -0.0659, -0.1487, ..., 0.2019, -0.1201, -0.1066], + [-0.0752, 0.0169, 0.2470, ..., -0.1693, -0.1705, -0.1920], + ..., + [-0.2003, -0.0774, 0.0318, ..., -0.0683, 0.1316, -0.1085], + [-0.1427, -0.1015, -0.0925, ..., -0.1655, -0.0187, -0.2222], + [-0.2127, -0.0801, -0.1708, ..., -0.2272, -0.0484, 0.1574]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -5.8673e-08, ..., 4.8429e-08, + 9.3132e-10, -1.5832e-08], + [ 9.3132e-10, 0.0000e+00, 6.5193e-09, ..., -1.5832e-08, + 5.5879e-09, 4.6566e-09], + [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 4.6566e-09, + 9.3132e-10, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7008e-08, ..., 1.7695e-08, + -1.0245e-08, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.9558e-08, + 3.7253e-09, 7.4506e-09], + [ 6.5193e-09, 0.0000e+00, 1.1176e-08, ..., 3.7253e-09, + 4.6566e-09, -5.2154e-08]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0377, -0.0302, 0.0082, 0.0306, -0.0193, 0.0053, -0.0163, 0.0159, + 0.0096, -0.0046], device='cuda:0'), grad: tensor([ 8.9407e-08, 6.5193e-09, 6.8918e-08, 1.1269e-07, 3.2596e-08, + 2.6077e-08, -2.6729e-07, 1.6764e-07, -1.0431e-07, -1.0990e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 214.48, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4538 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.18 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.1474, 0.0560, 0.0881, ..., -0.0987, -0.1881, 0.0743], + [-0.1279, -0.0659, -0.1487, ..., 0.2019, -0.1201, -0.1067], + [-0.0753, 0.0170, 0.2470, ..., -0.1694, -0.1705, -0.1926], + ..., + [-0.2011, -0.0774, 0.0319, ..., -0.0683, 0.1316, -0.1086], + [-0.1420, -0.1015, -0.0925, ..., -0.1657, -0.0187, -0.2219], + [-0.2128, -0.0802, -0.1722, ..., -0.2274, -0.0492, 0.1568]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -6.0350e-07, ..., 5.5879e-09, + 1.3039e-08, -6.1002e-07], + [ 5.4017e-08, 0.0000e+00, 7.9162e-08, ..., 8.0094e-08, + 1.1921e-07, 1.6578e-07], + [ 9.3132e-10, 0.0000e+00, 1.4622e-07, ..., 6.5193e-09, + 1.7695e-08, 1.3784e-07], + ..., + [ 9.3132e-10, 0.0000e+00, -2.3749e-07, ..., 4.6566e-09, + -6.4354e-07, 4.6566e-08], + [-7.8231e-08, 0.0000e+00, 4.3772e-08, ..., -9.4064e-08, + 6.8918e-08, -1.5087e-07], + [ 9.3132e-09, 0.0000e+00, 1.8347e-07, ..., 1.5832e-08, + 2.8778e-07, 1.1455e-07]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0370, -0.0302, 0.0082, 0.0324, -0.0193, 0.0031, -0.0163, 0.0159, + 0.0096, -0.0049], device='cuda:0'), grad: tensor([-2.3656e-06, 1.2964e-06, 6.3702e-07, 7.4226e-07, 3.4552e-07, + -7.2643e-08, 7.7952e-07, -1.7500e-06, -8.4098e-07, 1.2312e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 214.25, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4529 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.11 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.1451, 0.0560, 0.0882, ..., -0.0986, -0.1876, 0.0736], + [-0.1279, -0.0659, -0.1488, ..., 0.2019, -0.1201, -0.1068], + [-0.0764, 0.0170, 0.2471, ..., -0.1702, -0.1705, -0.1932], + ..., + [-0.2020, -0.0775, 0.0319, ..., -0.0683, 0.1317, -0.1088], + [-0.1408, -0.1015, -0.0926, ..., -0.1651, -0.0187, -0.2220], + [-0.2129, -0.0802, -0.1712, ..., -0.2275, -0.0494, 0.1584]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.5926e-07, ..., 9.3132e-09, + 0.0000e+00, -2.1793e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., -1.2573e-07, + 2.7940e-09, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.0245e-08, + 1.8626e-09, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 9.6858e-08, + -1.0245e-08, 9.3132e-09], + [-0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 1.9558e-08, + 9.3132e-10, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 7.5437e-08, ..., 4.6566e-09, + 5.5879e-09, 8.9407e-08]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0378, -0.0302, 0.0082, 0.0325, -0.0193, 0.0030, -0.0164, 0.0159, + 0.0096, -0.0036], device='cuda:0'), grad: tensor([-4.5262e-07, -1.4342e-07, 2.1700e-06, 2.6543e-07, 6.1467e-08, + 2.1607e-07, 4.5635e-08, 5.0943e-07, -3.1739e-06, 5.1130e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 214.53, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4569 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.16 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.1454, 0.0560, 0.0872, ..., -0.0987, -0.1878, 0.0724], + [-0.1280, -0.0659, -0.1488, ..., 0.2019, -0.1201, -0.1069], + [-0.0765, 0.0170, 0.2471, ..., -0.1703, -0.1706, -0.1951], + ..., + [-0.2031, -0.0775, 0.0319, ..., -0.0683, 0.1317, -0.1093], + [-0.1409, -0.1015, -0.0926, ..., -0.1655, -0.0187, -0.2221], + [-0.2129, -0.0802, -0.1706, ..., -0.2276, -0.0497, 0.1596]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -8.3819e-09, ..., 1.8626e-09, + 9.3132e-10, -1.5832e-08], + [-0.0000e+00, 0.0000e+00, 1.7416e-07, ..., -1.0245e-08, + 1.2480e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 0.0000e+00, + 2.1420e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.2445e-07, ..., 8.3819e-09, + -1.6391e-07, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 1.6764e-08, ..., 2.7940e-09, + 7.4506e-09, 1.8626e-09], + [-9.3132e-10, 0.0000e+00, 1.7695e-08, ..., 0.0000e+00, + 7.4506e-09, -7.4506e-09]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0390, -0.0302, 0.0081, 0.0324, -0.0197, 0.0033, -0.0164, 0.0159, + 0.0097, -0.0026], device='cuda:0'), grad: tensor([-2.9802e-08, 3.7719e-07, 4.1910e-08, 2.1420e-08, 3.5390e-08, + 4.3772e-08, -4.6566e-08, -4.8522e-07, 5.2154e-08, -9.3132e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 214.30, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4541 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.15 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.1455, 0.0560, 0.0874, ..., -0.0988, -0.1880, 0.0724], + [-0.1280, -0.0660, -0.1489, ..., 0.2019, -0.1201, -0.1071], + [-0.0767, 0.0170, 0.2471, ..., -0.1709, -0.1707, -0.1953], + ..., + [-0.2033, -0.0772, 0.0320, ..., -0.0683, 0.1317, -0.1096], + [-0.1411, -0.1015, -0.0926, ..., -0.1661, -0.0187, -0.2222], + [-0.2128, -0.0802, -0.1707, ..., -0.2277, -0.0500, 0.1597]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.8429e-08, ..., 0.0000e+00, + 1.4901e-08, -1.0710e-07], + [ 7.4506e-09, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 7.0781e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., 0.0000e+00, + 4.0047e-08, 1.4901e-08], + ..., + [ 9.3132e-09, 0.0000e+00, -1.3970e-08, ..., 9.3132e-10, + -1.0720e-06, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 8.3819e-09, 4.6566e-09], + [ 2.5146e-08, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 9.8720e-08, 3.6322e-08]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0390, -0.0302, 0.0079, 0.0321, -0.0199, 0.0036, -0.0164, 0.0159, + 0.0096, -0.0025], device='cuda:0'), grad: tensor([-3.5111e-07, 1.6671e-07, 7.7300e-08, 5.5879e-08, 2.3283e-08, + 1.5330e-06, 2.1420e-07, -2.1439e-06, 6.0536e-08, 3.6787e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 214.30, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4275 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.18 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.1456, 0.0560, 0.0875, ..., -0.0988, -0.1882, 0.0725], + [-0.1282, -0.0660, -0.1489, ..., 0.2019, -0.1201, -0.1072], + [-0.0769, 0.0170, 0.2471, ..., -0.1712, -0.1707, -0.1955], + ..., + [-0.2046, -0.0772, 0.0320, ..., -0.0683, 0.1317, -0.1098], + [-0.1414, -0.1016, -0.0926, ..., -0.1662, -0.0187, -0.2222], + [-0.2129, -0.0802, -0.1708, ..., -0.2277, -0.0501, 0.1597]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -1.1176e-08, ..., 9.3132e-10, + 9.3132e-10, -2.6748e-05], + [ 4.7497e-08, 0.0000e+00, 2.7940e-09, ..., -1.8626e-09, + 2.7940e-09, 7.6368e-08], + [ 9.3132e-10, 0.0000e+00, -9.3132e-09, ..., 2.7940e-09, + 9.3132e-10, 6.5193e-09], + ..., + [ 8.3819e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + -5.5879e-09, 7.4506e-08], + [ 3.7253e-09, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 9.3132e-10, 1.0338e-07], + [ 7.4320e-07, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 2.3041e-06]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0389, -0.0302, 0.0079, 0.0321, -0.0199, 0.0035, -0.0164, 0.0159, + 0.0097, -0.0025], device='cuda:0'), grad: tensor([-5.3436e-05, 2.4680e-07, -5.5879e-09, 1.8114e-06, -1.6475e-06, + 4.3631e-05, 2.8778e-06, 1.8254e-07, 2.1607e-07, 6.1467e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 214.34, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4512 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.07 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.1462, 0.0560, 0.0877, ..., -0.0988, -0.1883, 0.0726], + [-0.1298, -0.0660, -0.1491, ..., 0.2019, -0.1201, -0.1078], + [-0.0771, 0.0170, 0.2474, ..., -0.1710, -0.1708, -0.1962], + ..., + [-0.2064, -0.0772, 0.0320, ..., -0.0683, 0.1317, -0.1101], + [-0.1421, -0.1016, -0.0926, ..., -0.1666, -0.0187, -0.2225], + [-0.2130, -0.0802, -0.1710, ..., -0.2285, -0.0503, 0.1601]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.0827e-07, ..., 4.1910e-08, + 0.0000e+00, -2.6636e-07], + [ 1.2107e-08, 0.0000e+00, 2.0489e-08, ..., 4.0792e-07, + 9.3132e-09, 6.5193e-09], + [ 9.3132e-10, 0.0000e+00, -1.0803e-07, ..., 4.6566e-09, + 6.5193e-09, 9.3132e-09], + ..., + [ 4.6566e-09, 0.0000e+00, 9.8720e-08, ..., 0.0000e+00, + -2.3283e-08, 2.7940e-09], + [-2.3283e-08, 0.0000e+00, 1.5832e-08, ..., 1.0245e-07, + 9.3132e-10, 1.1176e-08], + [ 2.7940e-09, 0.0000e+00, 2.5611e-07, ..., 0.0000e+00, + 2.7940e-09, 2.1514e-07]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0388, -0.0302, 0.0079, 0.0321, -0.0193, 0.0034, -0.0164, 0.0159, + 0.0097, -0.0023], device='cuda:0'), grad: tensor([-6.8080e-07, 1.1520e-06, -1.0710e-07, 1.4249e-07, 5.5879e-09, + 2.2445e-07, -1.4687e-06, 2.1420e-07, -1.6578e-07, 6.7055e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 214.35, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4880 re_mapping 0.0031 re_causal 0.0096 /// teacc 99.09 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.1467, 0.0560, 0.0897, ..., -0.0988, -0.1885, 0.0729], + [-0.1301, -0.0660, -0.1491, ..., 0.2019, -0.1201, -0.1086], + [-0.0772, 0.0170, 0.2475, ..., -0.1716, -0.1709, -0.1979], + ..., + [-0.2070, -0.0772, 0.0320, ..., -0.0683, 0.1318, -0.1105], + [-0.1424, -0.1016, -0.0926, ..., -0.1668, -0.0187, -0.2228], + [-0.2131, -0.0802, -0.1724, ..., -0.2287, -0.0507, 0.1600]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 1.1176e-08, ..., -5.1595e-07, + 1.7695e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -5.3085e-08, ..., 2.7940e-09, + 3.7253e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -5.5879e-09, ..., 4.4424e-07, + -3.6322e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 2.7940e-09, + 2.7940e-09, 9.3132e-10], + [ 9.0338e-08, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 9.3132e-09, 3.7253e-09]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0383, -0.0302, 0.0079, 0.0315, -0.0193, 0.0040, -0.0164, 0.0159, + 0.0097, -0.0025], device='cuda:0'), grad: tensor([ 7.4506e-09, -1.2852e-06, -9.4064e-08, 2.7940e-09, 4.0978e-08, + -1.0803e-07, 2.3283e-08, 1.1660e-06, 8.2888e-08, 1.7043e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 214.20, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4757 re_mapping 0.0031 re_causal 0.0096 /// teacc 99.15 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.1468, 0.0560, 0.0898, ..., -0.0988, -0.1887, 0.0729], + [-0.1302, -0.0660, -0.1491, ..., 0.2019, -0.1202, -0.1087], + [-0.0775, 0.0170, 0.2475, ..., -0.1722, -0.1709, -0.1980], + ..., + [-0.2075, -0.0772, 0.0320, ..., -0.0683, 0.1318, -0.1109], + [-0.1426, -0.1016, -0.0927, ..., -0.1673, -0.0187, -0.2229], + [-0.2132, -0.0802, -0.1726, ..., -0.2289, -0.0511, 0.1601]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 1.2107e-08], + [ 1.8626e-09, 0.0000e+00, 5.5879e-09, ..., 1.6764e-08, + 5.7742e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, -6.7055e-08, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 6.5193e-09, ..., -4.2841e-08, + -1.3448e-06, 1.7695e-08], + [-2.7008e-08, 0.0000e+00, 4.4703e-08, ..., 9.3132e-09, + 3.6228e-07, 3.7253e-09], + [ 2.1420e-08, 0.0000e+00, 1.8626e-09, ..., 1.0245e-08, + 2.8685e-07, -1.0896e-07]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0382, -0.0302, 0.0079, 0.0306, -0.0194, 0.0050, -0.0164, 0.0159, + 0.0097, -0.0025], device='cuda:0'), grad: tensor([ 4.7497e-08, 2.1663e-06, -1.2293e-07, -2.7940e-09, 4.8708e-07, + 8.0094e-08, 4.8429e-08, -4.9397e-06, 1.3523e-06, 8.8289e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 214.16, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4695 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.1469, 0.0560, 0.0904, ..., -0.0989, -0.1888, 0.0731], + [-0.1302, -0.0660, -0.1491, ..., 0.2019, -0.1202, -0.1087], + [-0.0776, 0.0170, 0.2477, ..., -0.1728, -0.1710, -0.1984], + ..., + [-0.2085, -0.0772, 0.0320, ..., -0.0683, 0.1318, -0.1110], + [-0.1432, -0.1016, -0.0927, ..., -0.1679, -0.0187, -0.2229], + [-0.2133, -0.0802, -0.1732, ..., -0.2290, -0.0516, 0.1600]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0381, -0.0302, 0.0079, 0.0300, -0.0193, 0.0056, -0.0164, 0.0159, + 0.0097, -0.0027], device='cuda:0'), grad: tensor([ 2.7940e-09, 8.3819e-09, -2.8871e-08, -1.1176e-08, 2.7940e-09, + 5.5879e-09, 9.3132e-10, 3.7253e-09, 2.4214e-08, 1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 214.46, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4517 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.15 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.1470, 0.0560, 0.0906, ..., -0.0990, -0.1890, 0.0731], + [-0.1303, -0.0660, -0.1492, ..., 0.2019, -0.1202, -0.1088], + [-0.0777, 0.0170, 0.2477, ..., -0.1730, -0.1711, -0.1990], + ..., + [-0.2085, -0.0772, 0.0322, ..., -0.0684, 0.1320, -0.1087], + [-0.1435, -0.1016, -0.0927, ..., -0.1682, -0.0187, -0.2230], + [-0.2135, -0.0802, -0.1748, ..., -0.2291, -0.0558, 0.1599]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -9.2201e-08, ..., 4.6566e-09, + 0.0000e+00, -7.3574e-08], + [ 9.3132e-10, 0.0000e+00, 1.7695e-08, ..., -4.3772e-08, + 4.9360e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 4.6566e-09, + 5.5879e-09, 1.2107e-08], + ..., + [ 9.3132e-10, 0.0000e+00, -3.7253e-08, ..., 3.2596e-08, + -1.1455e-07, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 1.9558e-08, + 5.5879e-09, 2.7940e-09], + [ 6.5193e-09, 0.0000e+00, 4.0047e-08, ..., 1.8626e-09, + 2.5146e-08, 2.8871e-08]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0380, -0.0302, 0.0079, 0.0300, -0.0194, 0.0056, -0.0164, 0.0160, + 0.0097, -0.0037], device='cuda:0'), grad: tensor([-1.6391e-07, 4.0978e-08, 2.7474e-07, -4.1723e-07, 5.5879e-09, + 1.7416e-07, -6.0536e-08, -4.7497e-08, 2.8871e-08, 1.6857e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 214.46, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4578 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.17 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.1471, 0.0560, 0.0910, ..., -0.0990, -0.1891, 0.0733], + [-0.1303, -0.0660, -0.1493, ..., 0.2020, -0.1202, -0.1089], + [-0.0778, 0.0170, 0.2479, ..., -0.1731, -0.1711, -0.1996], + ..., + [-0.2086, -0.0772, 0.0322, ..., -0.0684, 0.1320, -0.1089], + [-0.1435, -0.1016, -0.0927, ..., -0.1688, -0.0187, -0.2230], + [-0.2135, -0.0802, -0.1751, ..., -0.2291, -0.0562, 0.1598]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 1.6764e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 3.9861e-07, ..., 0.0000e+00, + 2.9895e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.3886e-06, ..., 0.0000e+00, + 1.0403e-06, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -2.5779e-06, ..., 9.3132e-10, + -1.9316e-06, 6.5193e-09], + [ 1.8626e-09, 0.0000e+00, 3.3807e-07, ..., 0.0000e+00, + 2.5332e-07, 7.4506e-09], + [-8.6613e-08, 0.0000e+00, 2.1607e-07, ..., 0.0000e+00, + 1.5926e-07, -4.3306e-07]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0379, -0.0302, 0.0081, 0.0300, -0.0194, 0.0055, -0.0165, 0.0160, + 0.0097, -0.0038], device='cuda:0'), grad: tensor([ 9.6858e-08, 1.4864e-06, 5.1670e-06, 5.2527e-07, 1.5358e-06, + 4.3772e-08, 1.0245e-08, -9.5293e-06, 1.2713e-06, -5.8953e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 214.35, cls_loss 0.0004 cls_loss_mapping 0.0008 cls_loss_causal 0.4642 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.19 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.1472, 0.0560, 0.0914, ..., -0.0991, -0.1893, 0.0733], + [-0.1303, -0.0660, -0.1494, ..., 0.2020, -0.1202, -0.1089], + [-0.0789, 0.0170, 0.2479, ..., -0.1742, -0.1713, -0.2006], + ..., + [-0.2088, -0.0772, 0.0323, ..., -0.0684, 0.1321, -0.1091], + [-0.1438, -0.1016, -0.0927, ..., -0.1691, -0.0187, -0.2231], + [-0.2137, -0.0802, -0.1753, ..., -0.2292, -0.0565, 0.1599]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 0.0000e+00, -3.7253e-09, ..., 1.0245e-08, + 9.3132e-10, -1.8626e-09], + [-8.7544e-08, 0.0000e+00, 0.0000e+00, ..., -3.2689e-07, + 0.0000e+00, 0.0000e+00], + [ 7.5437e-08, 0.0000e+00, 0.0000e+00, ..., 2.1886e-07, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-08, 0.0000e+00, 0.0000e+00, ..., 7.7300e-08, + 0.0000e+00, 5.5879e-09], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 6.5193e-09, + 9.3132e-10, 9.3132e-10], + [ 5.0291e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, -1.8626e-09]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0378, -0.0302, 0.0078, 0.0300, -0.0194, 0.0055, -0.0164, 0.0160, + 0.0097, -0.0039], device='cuda:0'), grad: tensor([ 9.3132e-08, -6.3051e-07, 4.5449e-07, 3.4459e-08, -1.3411e-07, + -2.2352e-08, -5.2154e-08, 1.6950e-07, 2.2352e-08, 7.1712e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 214.36, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4652 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.15 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.1474, 0.0561, 0.0918, ..., -0.0990, -0.1894, 0.0734], + [-0.1303, -0.0660, -0.1505, ..., 0.2019, -0.1202, -0.1104], + [-0.0793, 0.0170, 0.2491, ..., -0.1714, -0.1713, -0.2010], + ..., + [-0.2090, -0.0773, 0.0323, ..., -0.0684, 0.1321, -0.1092], + [-0.1438, -0.1016, -0.0928, ..., -0.1697, -0.0187, -0.2232], + [-0.2137, -0.0803, -0.1755, ..., -0.2293, -0.0567, 0.1599]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -2.6077e-08, ..., 1.8626e-09, + 0.0000e+00, -2.6543e-08], + [-1.8626e-09, 0.0000e+00, 5.1223e-09, ..., -4.0047e-08, + 2.3283e-09, 4.6566e-10], + [ 1.6764e-08, 0.0000e+00, 2.0023e-08, ..., 1.6298e-08, + 9.7789e-09, 1.8626e-09], + ..., + [ 8.8476e-09, 0.0000e+00, -3.2131e-08, ..., 2.6543e-08, + -1.0245e-08, 1.3970e-09], + [-4.6566e-10, 0.0000e+00, 2.3283e-09, ..., -4.6566e-10, + 9.3132e-10, 4.6566e-10], + [ 1.4110e-07, 0.0000e+00, 1.8161e-08, ..., 9.3132e-10, + 5.5879e-09, 1.4901e-08]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0379, -0.0302, 0.0096, 0.0300, -0.0195, 0.0054, -0.0164, 0.0160, + 0.0097, -0.0039], device='cuda:0'), grad: tensor([-1.2247e-07, -4.7032e-08, 8.6613e-08, 2.7474e-08, -1.6484e-07, + 4.8894e-08, 1.4901e-08, 4.9360e-08, -1.4901e-07, 2.6636e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 214.34, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4761 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.18 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.1475, 0.0561, 0.0920, ..., -0.0991, -0.1894, 0.0736], + [-0.1303, -0.0660, -0.1505, ..., 0.2020, -0.1202, -0.1105], + [-0.0794, 0.0170, 0.2492, ..., -0.1715, -0.1714, -0.2012], + ..., + [-0.2100, -0.0768, 0.0323, ..., -0.0684, 0.1321, -0.1094], + [-0.1433, -0.1016, -0.0928, ..., -0.1703, -0.0187, -0.2232], + [-0.2138, -0.0804, -0.1757, ..., -0.2295, -0.0569, 0.1599]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., 1.8626e-09, + 0.0000e+00, -4.3772e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -1.6764e-08, + 7.4506e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 9.3132e-10, + 2.4214e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -7.2643e-08, ..., 1.3039e-08, + -8.4750e-08, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.2107e-08, + 7.4506e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 0.0000e+00, + 1.2107e-08, -1.8626e-08]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0377, -0.0302, 0.0096, 0.0300, -0.0196, 0.0054, -0.0163, 0.0160, + 0.0097, -0.0040], device='cuda:0'), grad: tensor([-9.3132e-08, -1.8626e-08, 5.3085e-08, 7.9162e-08, 5.6811e-08, + 5.4948e-08, -9.8720e-08, -9.8720e-08, 1.0990e-07, -3.1665e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 214.17, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4643 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.14 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.1478, 0.0561, 0.0919, ..., -0.0994, -0.1897, 0.0736], + [-0.1304, -0.0661, -0.1507, ..., 0.2020, -0.1202, -0.1105], + [-0.0794, 0.0177, 0.2498, ..., -0.1716, -0.1715, -0.2013], + ..., + [-0.2105, -0.0771, 0.0323, ..., -0.0684, 0.1321, -0.1098], + [-0.1434, -0.1017, -0.0928, ..., -0.1707, -0.0187, -0.2233], + [-0.2139, -0.0805, -0.1760, ..., -0.2296, -0.0569, 0.1600]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 2.7940e-09, + 9.3132e-10, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 9.3132e-10, + -2.7940e-09, 1.7695e-08], + [-1.8626e-09, 0.0000e+00, 5.5879e-09, ..., 9.3132e-10, + 3.1665e-08, -1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, -0.0000e+00]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0378, -0.0302, 0.0102, 0.0299, -0.0196, 0.0054, -0.0163, 0.0160, + 0.0097, -0.0040], device='cuda:0'), grad: tensor([-1.8626e-09, 1.4901e-08, -2.6077e-08, 6.2399e-08, 1.3970e-08, + -1.8999e-07, 3.5390e-08, 7.4506e-08, -1.3597e-07, 1.6019e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 214.24, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4971 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.07 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.1479, 0.0561, 0.0921, ..., -0.0995, -0.1903, 0.0736], + [-0.1304, -0.0661, -0.1508, ..., 0.2020, -0.1203, -0.1107], + [-0.0795, 0.0177, 0.2497, ..., -0.1717, -0.1719, -0.2022], + ..., + [-0.2109, -0.0771, 0.0324, ..., -0.0684, 0.1322, -0.1102], + [-0.1430, -0.1017, -0.0928, ..., -0.1713, -0.0187, -0.2233], + [-0.2140, -0.0805, -0.1763, ..., -0.2297, -0.0572, 0.1601]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., -1.1176e-08, + 8.3819e-09, -7.4506e-08], + [ 5.7742e-08, 0.0000e+00, 6.0163e-07, ..., -1.5832e-08, + 8.0932e-07, 4.6566e-09], + [ 2.7940e-09, 0.0000e+00, 2.4214e-08, ..., 3.7253e-09, + 4.0047e-08, 9.3132e-10], + ..., + [-8.1025e-08, 0.0000e+00, -9.0711e-07, ..., 5.5879e-09, + -1.2275e-06, 2.2352e-08], + [ 4.6566e-09, 0.0000e+00, 5.5879e-08, ..., 4.6566e-09, + 8.0094e-08, 1.9558e-08], + [ 3.7532e-07, 0.0000e+00, 1.0803e-07, ..., 0.0000e+00, + 1.3970e-07, -4.3493e-07]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0377, -0.0302, 0.0097, 0.0299, -0.0196, 0.0054, -0.0164, 0.0160, + 0.0097, -0.0041], device='cuda:0'), grad: tensor([ 4.7497e-08, 3.0678e-06, 2.1327e-07, -1.5318e-04, 1.0040e-06, + 1.5104e-04, 2.0452e-06, -4.4331e-06, 2.2259e-07, -1.9372e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 214.15, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4873 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.15 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.1484, 0.0561, 0.0922, ..., -0.1001, -0.1906, 0.0737], + [-0.1305, -0.0661, -0.1511, ..., 0.2020, -0.1203, -0.1109], + [-0.0798, 0.0177, 0.2497, ..., -0.1719, -0.1720, -0.2026], + ..., + [-0.2114, -0.0771, 0.0326, ..., -0.0684, 0.1323, -0.1106], + [-0.1441, -0.1017, -0.0928, ..., -0.1731, -0.0187, -0.2236], + [-0.2143, -0.0805, -0.1765, ..., -0.2299, -0.0574, 0.1602]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 8.3819e-09, ..., 9.3132e-10, + 1.0245e-08, -1.3039e-08], + [-9.3132e-10, 0.0000e+00, 5.5879e-08, ..., -9.3132e-09, + 2.2352e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -1.5087e-07, ..., 0.0000e+00, + 6.5193e-09, 9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -1.3970e-08, ..., 5.5879e-09, + -3.1665e-08, 2.7940e-09], + [-7.4506e-09, 0.0000e+00, 1.3039e-08, ..., 2.7940e-09, + 2.5146e-08, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 9.3132e-10, + 3.5390e-08, 1.2107e-08]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0379, -0.0302, 0.0096, 0.0303, -0.0198, 0.0050, -0.0162, 0.0160, + 0.0097, -0.0041], device='cuda:0'), grad: tensor([ 6.7987e-08, 1.4715e-07, -3.2783e-07, 1.0550e-05, 1.5832e-08, + -1.0587e-05, 7.7300e-08, -2.7940e-08, 1.8626e-09, 1.3225e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 214.15, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4369 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.13 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.1490, 0.0561, 0.0925, ..., -0.1004, -0.1908, 0.0738], + [-0.1306, -0.0661, -0.1512, ..., 0.2020, -0.1203, -0.1109], + [-0.0800, 0.0177, 0.2499, ..., -0.1720, -0.1721, -0.2030], + ..., + [-0.2119, -0.0771, 0.0326, ..., -0.0684, 0.1323, -0.1109], + [-0.1441, -0.1017, -0.0929, ..., -0.1736, -0.0187, -0.2238], + [-0.2145, -0.0805, -0.1768, ..., -0.2301, -0.0576, 0.1602]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 6.4261e-08, ..., -3.7253e-09, + 8.5682e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-08, ..., 0.0000e+00, + 8.7544e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.8440e-07, ..., 3.7253e-09, + -5.0105e-07, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 7.6368e-08, ..., 0.0000e+00, + 2.1886e-07, 1.8626e-09], + [ 2.7940e-09, 0.0000e+00, 4.7497e-08, ..., 0.0000e+00, + 8.7544e-08, 1.8626e-09]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0379, -0.0302, 0.0098, 0.0304, -0.0198, 0.0049, -0.0161, 0.0160, + 0.0096, -0.0042], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.6915e-07, 1.0896e-07, 2.7940e-08, 3.5390e-08, + 1.8626e-09, 2.7940e-09, -1.9148e-06, 1.1111e-06, 3.4180e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 214.21, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4417 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.15 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.1498, 0.0561, 0.0925, ..., -0.1008, -0.1916, 0.0737], + [-0.1306, -0.0661, -0.1513, ..., 0.2020, -0.1203, -0.1110], + [-0.0802, 0.0177, 0.2500, ..., -0.1722, -0.1722, -0.2034], + ..., + [-0.2130, -0.0772, 0.0327, ..., -0.0685, 0.1323, -0.1116], + [-0.1445, -0.1017, -0.0929, ..., -0.1741, -0.0187, -0.2240], + [-0.2148, -0.0805, -0.1771, ..., -0.2302, -0.0581, 0.1604]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 2.7940e-09, ..., 0.0000e+00, + 5.5879e-09, -5.8673e-08], + [ 9.3132e-10, 9.3132e-10, 2.3283e-08, ..., -1.3039e-08, + 5.9605e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 0.0000e+00, + 1.2107e-08, 9.3132e-10], + ..., + [ 9.3132e-10, 1.8626e-09, -1.5553e-07, ..., 8.3819e-09, + -4.3865e-07, 2.0489e-08], + [ 0.0000e+00, 9.3132e-10, -4.8429e-08, ..., 1.8626e-09, + 2.1420e-08, 9.3132e-09], + [ 9.3132e-09, 1.8626e-09, 1.3225e-07, ..., 0.0000e+00, + 3.1572e-07, 2.5146e-08]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0381, -0.0302, 0.0097, 0.0304, -0.0198, 0.0049, -0.0161, 0.0160, + 0.0096, -0.0041], device='cuda:0'), grad: tensor([ 2.0489e-08, 2.1141e-07, 1.5367e-07, 6.5565e-07, -9.3132e-10, + -4.8149e-07, 6.2399e-08, -1.4966e-06, -3.9767e-07, 1.2796e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 214.42, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4619 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.18 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.1500, 0.0561, 0.0926, ..., -0.1010, -0.1919, 0.0737], + [-0.1308, -0.0662, -0.1513, ..., 0.2020, -0.1203, -0.1111], + [-0.0804, 0.0155, 0.2499, ..., -0.1742, -0.1725, -0.2036], + ..., + [-0.2156, -0.0775, 0.0328, ..., -0.0685, 0.1324, -0.1117], + [-0.1445, -0.1020, -0.0929, ..., -0.1746, -0.0187, -0.2242], + [-0.2152, -0.0806, -0.1779, ..., -0.2303, -0.0588, 0.1605]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, -3.7253e-09], + [-0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -2.8405e-07, + 1.8626e-09, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 2.0955e-07, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 5.0291e-08, + -9.3132e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 1.9558e-08, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 7.4506e-09, 2.7940e-09]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0383, -0.0302, 0.0083, 0.0306, -0.0195, 0.0062, -0.0171, 0.0160, + 0.0097, -0.0044], device='cuda:0'), grad: tensor([-1.8626e-09, -7.5530e-07, 5.6811e-07, 3.1665e-08, -1.3970e-08, + 1.8626e-08, -3.7253e-09, 1.1548e-07, -2.7940e-09, 5.4017e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 214.20, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4853 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.10 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.1502, 0.0563, 0.0928, ..., -0.1011, -0.1920, 0.0738], + [-0.1312, -0.0662, -0.1514, ..., 0.2020, -0.1204, -0.1112], + [-0.0805, 0.0155, 0.2503, ..., -0.1746, -0.1726, -0.2035], + ..., + [-0.2168, -0.0775, 0.0328, ..., -0.0685, 0.1324, -0.1119], + [-0.1459, -0.1020, -0.0929, ..., -0.1747, -0.0187, -0.2244], + [-0.2168, -0.0808, -0.1782, ..., -0.2307, -0.0590, 0.1605]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.9895e-07, ..., 3.7253e-09, + 9.3132e-10, -3.7998e-07], + [ 1.8626e-08, 0.0000e+00, 4.0047e-08, ..., -3.1665e-08, + 4.9360e-08, 1.1176e-08], + [ 2.7940e-09, 0.0000e+00, 1.1176e-08, ..., 1.8626e-09, + 7.4506e-09, 9.3132e-09], + ..., + [-8.8476e-08, 0.0000e+00, -1.3597e-07, ..., 1.3039e-08, + -2.2165e-07, 7.4506e-09], + [ 1.3970e-08, 0.0000e+00, 6.2399e-08, ..., 7.4506e-09, + 3.2596e-08, 5.6811e-08], + [ 1.4901e-08, 0.0000e+00, 2.2817e-07, ..., 9.3132e-10, + 6.5193e-08, 1.9092e-07]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0382, -0.0302, 0.0086, 0.0305, -0.0183, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0052], device='cuda:0'), grad: tensor([-1.1995e-06, 1.2573e-07, 6.1467e-08, 1.1455e-07, 2.9337e-07, + 5.8673e-08, 6.8918e-08, -7.1246e-07, 3.1665e-07, 8.7917e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 214.15, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4562 re_mapping 0.0028 re_causal 0.0088 /// teacc 99.13 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.1503, 0.0563, 0.0929, ..., -0.1011, -0.1922, 0.0738], + [-0.1312, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1112], + [-0.0805, 0.0155, 0.2503, ..., -0.1746, -0.1726, -0.2036], + ..., + [-0.2169, -0.0776, 0.0328, ..., -0.0685, 0.1324, -0.1119], + [-0.1460, -0.1021, -0.0930, ..., -0.1748, -0.0187, -0.2244], + [-0.2167, -0.0809, -0.1783, ..., -0.2307, -0.0591, 0.1606]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8871e-08, ..., 2.7008e-08, + 4.5635e-08, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -1.0245e-08, + 4.7497e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 1.8626e-08, 9.3132e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -5.9605e-08, ..., 1.3970e-08, + -1.0189e-06, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.9558e-08, + 1.3970e-08, 1.3039e-08], + [ 9.3132e-10, 0.0000e+00, 2.8871e-08, ..., 4.0047e-08, + 1.0617e-07, 1.0896e-07]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0382, -0.0302, 0.0086, 0.0305, -0.0185, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 1.4249e-07, 1.1735e-07, 8.6613e-08, 3.8184e-06, 6.8918e-08, + -4.9472e-06, 2.0340e-06, -1.8887e-06, 1.0710e-07, 4.5914e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 214.28, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4665 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.13 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.1503, 0.0563, 0.0929, ..., -0.1011, -0.1922, 0.0739], + [-0.1312, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1112], + [-0.0806, 0.0155, 0.2503, ..., -0.1746, -0.1726, -0.2036], + ..., + [-0.2169, -0.0776, 0.0328, ..., -0.0685, 0.1324, -0.1120], + [-0.1460, -0.1021, -0.0930, ..., -0.1749, -0.0187, -0.2245], + [-0.2167, -0.0809, -0.1783, ..., -0.2307, -0.0591, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, -3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, -3.7253e-09], + [-0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -2.9802e-08, + 1.0245e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -3.7253e-09, ..., 4.6566e-09, + 2.7940e-09, 0.0000e+00], + ..., + [-4.6566e-09, 0.0000e+00, -1.1176e-08, ..., 1.5832e-08, + -2.8871e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 5.5879e-09, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 5.5879e-09, 3.7253e-09]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0382, -0.0302, 0.0086, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([-4.6566e-09, -4.9360e-08, 1.1176e-08, 3.5390e-08, 1.2107e-08, + -2.5146e-08, 6.5193e-09, -1.3039e-08, -1.8626e-09, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 214.28, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4391 re_mapping 0.0026 re_causal 0.0087 /// teacc 99.12 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.1503, 0.0563, 0.0930, ..., -0.1011, -0.1922, 0.0739], + [-0.1313, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1112], + [-0.0806, 0.0155, 0.2504, ..., -0.1746, -0.1726, -0.2036], + ..., + [-0.2169, -0.0776, 0.0328, ..., -0.0685, 0.1324, -0.1121], + [-0.1461, -0.1021, -0.0930, ..., -0.1750, -0.0187, -0.2245], + [-0.2167, -0.0809, -0.1784, ..., -0.2307, -0.0591, 0.1607]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, -9.7789e-07], + [ 1.8626e-09, 0.0000e+00, 5.2154e-08, ..., 3.1665e-08, + 1.0617e-07, 9.1270e-08], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 1.2107e-08], + ..., + [ 3.7253e-09, 0.0000e+00, -6.1467e-08, ..., -3.5390e-08, + -1.2200e-07, 6.9849e-08], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 2.7940e-09, 8.1025e-08], + [ 9.3132e-09, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 4.6566e-09, 2.3469e-07]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0381, -0.0302, 0.0086, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([-2.5555e-06, 6.1188e-07, 4.3772e-08, 2.4587e-07, 4.7777e-07, + 2.0023e-07, 5.2061e-07, -1.3784e-07, 2.6543e-07, 3.1851e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 214.21, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4024 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.15 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.1503, 0.0563, 0.0930, ..., -0.1011, -0.1922, 0.0739], + [-0.1313, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1726, -0.2036], + ..., + [-0.2170, -0.0776, 0.0328, ..., -0.0685, 0.1324, -0.1121], + [-0.1461, -0.1021, -0.0930, ..., -0.1750, -0.0187, -0.2245], + [-0.2167, -0.0809, -0.1784, ..., -0.2307, -0.0591, 0.1607]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, -1.8626e-09], + [ 9.3132e-09, 0.0000e+00, 6.5193e-09, ..., -9.3132e-09, + 1.1176e-08, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, -2.9802e-08, ..., 6.5193e-09, + -5.5879e-08, -3.7253e-09], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 3.7253e-09, + 4.6566e-09, 1.8626e-09], + [ 2.9802e-08, 0.0000e+00, 2.3283e-08, ..., 0.0000e+00, + 3.4459e-08, 4.6566e-09]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0381, -0.0302, 0.0086, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([-9.3132e-10, 2.2352e-08, 4.6566e-09, -3.3528e-08, -8.6613e-08, + 3.7253e-09, 1.7695e-08, -8.0094e-08, 2.7940e-08, 1.4249e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 214.29, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4724 re_mapping 0.0026 re_causal 0.0090 /// teacc 99.10 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.1503, 0.0564, 0.0931, ..., -0.1011, -0.1922, 0.0739], + [-0.1313, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1726, -0.2037], + ..., + [-0.2170, -0.0776, 0.0328, ..., -0.0685, 0.1324, -0.1121], + [-0.1461, -0.1021, -0.0930, ..., -0.1751, -0.0187, -0.2246], + [-0.2167, -0.0809, -0.1784, ..., -0.2307, -0.0591, 0.1607]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.7940e-08, ..., 1.8626e-09, + 1.2107e-08, -1.2107e-08], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., -7.4506e-08, + 1.8626e-09, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -5.5879e-09, ..., 1.8626e-09, + 3.7253e-09, 7.4506e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 6.6124e-08, + -2.7940e-09, 8.3819e-09], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, + 3.7253e-09, 7.4506e-09], + [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., 9.3132e-10, + 1.8626e-09, 8.3819e-09]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0381, -0.0302, 0.0086, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([-4.0047e-08, -2.0210e-07, 1.8626e-08, 4.2841e-08, 1.7695e-08, + -2.8592e-07, 1.6764e-07, 2.2538e-07, -1.8626e-09, 4.8429e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 214.30, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4506 re_mapping 0.0026 re_causal 0.0088 /// teacc 99.13 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.1504, 0.0564, 0.0931, ..., -0.1012, -0.1923, 0.0740], + [-0.1313, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1727, -0.2037], + ..., + [-0.2170, -0.0776, 0.0328, ..., -0.0685, 0.1324, -0.1122], + [-0.1461, -0.1021, -0.0930, ..., -0.1751, -0.0187, -0.2246], + [-0.2167, -0.0810, -0.1785, ..., -0.2307, -0.0591, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3213e-07, ..., 9.3132e-10, + 0.0000e+00, -4.1537e-07], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., -1.0803e-07, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 9.6858e-08, + -2.7940e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 2.7940e-09, + 0.0000e+00, 1.0245e-08], + [ 4.6566e-09, 0.0000e+00, 1.4715e-07, ..., 9.3132e-10, + 9.3132e-10, 1.3690e-07]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0381, -0.0302, 0.0087, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([-1.3318e-06, -2.9150e-07, 1.6764e-08, 7.4040e-07, -1.8626e-09, + 6.8918e-08, 2.5146e-08, 3.2131e-07, -3.2596e-08, 4.8336e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 214.54, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4528 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.13 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.1504, 0.0564, 0.0932, ..., -0.1012, -0.1923, 0.0740], + [-0.1313, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1727, -0.2037], + ..., + [-0.2170, -0.0777, 0.0328, ..., -0.0685, 0.1324, -0.1122], + [-0.1461, -0.1021, -0.0930, ..., -0.1752, -0.0187, -0.2246], + [-0.2167, -0.0810, -0.1785, ..., -0.2307, -0.0592, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.9185e-07, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 1.2573e-07, ..., 0.0000e+00, + -1.7695e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.5635e-08, ..., 9.3132e-10, + 2.7940e-09, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-09, -1.1176e-08]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0381, -0.0302, 0.0087, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 1.3970e-08, 2.6077e-08, -3.2689e-07, -6.7055e-08, 1.8626e-09, + 5.3085e-08, 1.8626e-09, 2.2445e-07, 8.3819e-08, -1.0245e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 214.17, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4557 re_mapping 0.0026 re_causal 0.0087 /// teacc 99.12 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.1504, 0.0564, 0.0932, ..., -0.1012, -0.1923, 0.0740], + [-0.1313, -0.0662, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1727, -0.2037], + ..., + [-0.2170, -0.0777, 0.0328, ..., -0.0685, 0.1324, -0.1123], + [-0.1461, -0.1021, -0.0930, ..., -0.1753, -0.0187, -0.2246], + [-0.2167, -0.0810, -0.1786, ..., -0.2308, -0.0592, 0.1607]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.6566e-09, ..., 2.7940e-09, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 5.0291e-08, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 5.5879e-09, + -1.5832e-08, 2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., -6.3330e-08, + 3.7253e-09, 9.3132e-10], + [ 2.7940e-09, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 5.5879e-09, -3.7253e-09]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0381, -0.0302, 0.0087, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.2806e-06, 1.6764e-07, 2.1420e-08, 1.7695e-08, + 9.3132e-10, -2.6077e-08, -7.4506e-09, -1.4808e-06, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 214.44, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4398 re_mapping 0.0025 re_causal 0.0087 /// teacc 99.13 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.1504, 0.0564, 0.0932, ..., -0.1012, -0.1924, 0.0740], + [-0.1313, -0.0661, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1727, -0.2037], + ..., + [-0.2171, -0.0778, 0.0328, ..., -0.0685, 0.1324, -0.1123], + [-0.1462, -0.1021, -0.0930, ..., -0.1753, -0.0187, -0.2247], + [-0.2167, -0.0810, -0.1786, ..., -0.2308, -0.0592, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.2969e-07, ..., -5.5879e-09, + 0.0000e+00, -4.1351e-07], + [ 1.8626e-09, 0.0000e+00, 1.1176e-08, ..., -6.5193e-09, + 8.3819e-09, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 9.3132e-10, 1.2107e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 5.5879e-09, + -1.3970e-08, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 5.4017e-08, ..., 3.3528e-08, + -0.0000e+00, 9.4995e-08], + [ 1.2107e-08, 0.0000e+00, 1.9558e-07, ..., 5.5879e-09, + 3.7253e-09, 1.8720e-07]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0381, -0.0302, 0.0087, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([-1.2517e-06, 6.6124e-08, 4.2841e-08, 2.4028e-07, 6.5193e-09, + 8.8476e-08, -6.7055e-08, 3.7253e-08, 2.0582e-07, 6.3982e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 214.42, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4439 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.14 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.1504, 0.0564, 0.0933, ..., -0.1012, -0.1924, 0.0740], + [-0.1313, -0.0661, -0.1515, ..., 0.2020, -0.1204, -0.1113], + [-0.0806, 0.0155, 0.2504, ..., -0.1747, -0.1727, -0.2038], + ..., + [-0.2171, -0.0778, 0.0328, ..., -0.0685, 0.1325, -0.1124], + [-0.1462, -0.1021, -0.0930, ..., -0.1754, -0.0187, -0.2247], + [-0.2167, -0.0810, -0.1786, ..., -0.2308, -0.0592, 0.1607]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -4.6566e-09, ..., 1.8626e-09, + 3.7253e-09, -7.4506e-09], + [ 5.5879e-09, 0.0000e+00, 5.3085e-08, ..., -2.7940e-09, + 7.9162e-08, 9.3132e-10], + [ 2.7940e-09, 0.0000e+00, 1.2759e-07, ..., 0.0000e+00, + 1.5832e-07, 0.0000e+00], + ..., + [ 2.3283e-08, 0.0000e+00, -3.3807e-07, ..., 1.8626e-09, + -4.7870e-07, 1.8626e-09], + [ 5.5879e-09, 0.0000e+00, 7.4506e-08, ..., 3.7253e-09, + 9.1270e-08, 1.0245e-08], + [ 9.3132e-09, 0.0000e+00, 5.6811e-08, ..., 0.0000e+00, + 1.0896e-07, -1.1176e-08]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([-1.8626e-09, 2.6356e-07, 4.9453e-07, 1.1921e-07, -4.9360e-08, + -8.7544e-08, 8.3819e-08, -1.5162e-06, 3.3155e-07, 3.5670e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 214.46, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4052 re_mapping 0.0024 re_causal 0.0083 /// teacc 99.15 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.1504, 0.0564, 0.0933, ..., -0.1012, -0.1924, 0.0740], + [-0.1313, -0.0661, -0.1515, ..., 0.2020, -0.1204, -0.1114], + [-0.0806, 0.0155, 0.2505, ..., -0.1748, -0.1727, -0.2038], + ..., + [-0.2171, -0.0778, 0.0328, ..., -0.0685, 0.1325, -0.1124], + [-0.1462, -0.1021, -0.0930, ..., -0.1755, -0.0187, -0.2247], + [-0.2167, -0.0810, -0.1787, ..., -0.2308, -0.0592, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 1.5832e-08, + 0.0000e+00, -9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., -9.3132e-10, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -9.3132e-09, ..., 9.3132e-10, + -1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 1.8626e-09, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 3.7253e-09, 9.3132e-10]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0305, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0050], device='cuda:0'), grad: tensor([ 6.3330e-08, 2.1420e-08, 6.5193e-09, -1.8626e-09, -5.5879e-09, + 1.1176e-08, -1.0245e-07, -3.1665e-08, 2.2352e-08, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 214.25, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4574 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.1505, 0.0564, 0.0933, ..., -0.1012, -0.1925, 0.0740], + [-0.1313, -0.0661, -0.1515, ..., 0.2020, -0.1204, -0.1114], + [-0.0806, 0.0155, 0.2505, ..., -0.1748, -0.1727, -0.2038], + ..., + [-0.2172, -0.0778, 0.0328, ..., -0.0685, 0.1325, -0.1124], + [-0.1462, -0.1021, -0.0930, ..., -0.1756, -0.0187, -0.2248], + [-0.2168, -0.0810, -0.1787, ..., -0.2308, -0.0593, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.2352e-08, + 0.0000e+00, 7.4506e-09], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -1.1921e-07, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 9.9652e-08, ..., 9.3132e-10, + -8.3819e-09, 7.4506e-09], + [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 6.5193e-08, + 5.5879e-09, 3.3528e-08], + [ 5.5879e-09, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 3.7253e-09, -1.0245e-08]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0304, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 7.9162e-08, 1.6764e-08, -2.1141e-07, 1.3970e-08, 0.0000e+00, + -1.7695e-08, -3.3621e-07, 1.8813e-07, 2.6543e-07, -8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 214.42, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4779 re_mapping 0.0024 re_causal 0.0088 /// teacc 99.12 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.1505, 0.0564, 0.0933, ..., -0.1013, -0.1925, 0.0740], + [-0.1314, -0.0661, -0.1515, ..., 0.2020, -0.1204, -0.1114], + [-0.0807, 0.0155, 0.2505, ..., -0.1748, -0.1727, -0.2038], + ..., + [-0.2172, -0.0778, 0.0328, ..., -0.0685, 0.1325, -0.1125], + [-0.1462, -0.1021, -0.0930, ..., -0.1756, -0.0187, -0.2248], + [-0.2168, -0.0810, -0.1787, ..., -0.2309, -0.0593, 0.1607]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, -5.1223e-09], + [ 2.3283e-09, 0.0000e+00, 1.8626e-09, ..., -3.2131e-08, + 1.3970e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -2.8638e-07, ..., 6.0536e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 3.2596e-09, 0.0000e+00, 7.0781e-08, ..., 1.8161e-08, + 1.1642e-08, 4.6566e-10], + [ 1.8626e-09, 0.0000e+00, 7.9162e-09, ..., 7.9162e-09, + 1.3970e-09, 4.6566e-10], + [ 2.0582e-07, 0.0000e+00, 5.1223e-09, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0304, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 3.4925e-08, -4.7032e-08, -5.4063e-07, 3.4133e-07, -3.9162e-07, + 1.2107e-08, 3.6322e-08, 2.9337e-07, -1.4156e-07, 4.1118e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 214.51, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4180 re_mapping 0.0024 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.1505, 0.0564, 0.0934, ..., -0.1013, -0.1925, 0.0741], + [-0.1314, -0.0661, -0.1515, ..., 0.2020, -0.1204, -0.1114], + [-0.0807, 0.0155, 0.2505, ..., -0.1748, -0.1728, -0.2038], + ..., + [-0.2172, -0.0778, 0.0328, ..., -0.0685, 0.1325, -0.1125], + [-0.1462, -0.1021, -0.0930, ..., -0.1757, -0.0187, -0.2248], + [-0.2168, -0.0810, -0.1788, ..., -0.2309, -0.0593, 0.1607]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -0.0000e+00, -9.3132e-10, ..., 3.2596e-09, + 0.0000e+00, 2.3283e-09], + [ 1.4110e-07, 0.0000e+00, 8.3819e-09, ..., 3.1199e-08, + 7.9162e-09, 1.8626e-09], + [ 4.6566e-09, 0.0000e+00, 5.3551e-08, ..., 1.8626e-09, + 4.3772e-08, 0.0000e+00], + ..., + [ 1.6298e-08, 0.0000e+00, -7.3109e-08, ..., 6.5193e-09, + -6.1002e-08, 1.8626e-09], + [ 2.3283e-09, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, + 2.3283e-09, 4.6566e-10], + [ 3.9581e-08, 0.0000e+00, 3.2596e-09, ..., 8.3819e-09, + 1.8626e-09, -6.8452e-08]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 4.2375e-08, 2.4633e-07, 1.7509e-07, 2.5611e-07, -3.7206e-07, + 2.7474e-08, 1.7602e-07, -1.0990e-07, -3.9907e-07, -4.7497e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 214.66, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4421 re_mapping 0.0023 re_causal 0.0084 /// teacc 99.18 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.1506, 0.0564, 0.0934, ..., -0.1013, -0.1925, 0.0741], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1115], + [-0.0807, 0.0155, 0.2505, ..., -0.1748, -0.1728, -0.2039], + ..., + [-0.2172, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1126], + [-0.1463, -0.1021, -0.0930, ..., -0.1758, -0.0187, -0.2249], + [-0.2168, -0.0810, -0.1788, ..., -0.2309, -0.0594, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 4.1910e-09, ..., -1.4435e-08, + 9.3132e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 4.6566e-10, + 7.4506e-09, 4.6566e-10], + ..., + [ 1.3970e-09, 0.0000e+00, -1.6298e-08, ..., 1.3504e-08, + -3.4925e-08, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, + 2.7940e-09, 2.3283e-09], + [ 6.0536e-09, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 1.0710e-08, -1.4901e-08]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.2573e-08, 2.0023e-08, 3.8184e-08, 6.5193e-09, + -6.2399e-08, 2.5611e-08, -2.6543e-08, 2.0023e-08, -1.3970e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 214.33, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4264 re_mapping 0.0024 re_causal 0.0083 /// teacc 99.18 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.1506, 0.0565, 0.0934, ..., -0.1013, -0.1926, 0.0741], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1115], + [-0.0807, 0.0155, 0.2505, ..., -0.1749, -0.1728, -0.2039], + ..., + [-0.2173, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1126], + [-0.1463, -0.1021, -0.0930, ..., -0.1758, -0.0187, -0.2249], + [-0.2169, -0.0810, -0.1789, ..., -0.2309, -0.0594, 0.1607]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 4.6566e-10, ..., 2.7940e-09, + 4.6566e-10, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 5.1223e-09, ..., -6.0536e-09, + 5.5879e-09, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.1910e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.3749e-08, ..., 5.5879e-09, + -2.9802e-08, 4.1910e-09], + [ 2.3283e-09, 0.0000e+00, 4.1910e-09, ..., 7.6368e-08, + 2.3283e-09, 2.4680e-08], + [ 6.0536e-09, 0.0000e+00, 1.1642e-08, ..., 9.3132e-10, + 1.3970e-08, -2.5611e-08]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 1.3504e-08, -4.6566e-10, 5.1223e-09, 4.6566e-09, -3.6787e-08, + 1.2107e-08, -2.2352e-07, -3.9581e-08, 2.8964e-07, -1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 214.15, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4251 re_mapping 0.0023 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.1506, 0.0565, 0.0935, ..., -0.1014, -0.1926, 0.0741], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1115], + [-0.0807, 0.0155, 0.2505, ..., -0.1749, -0.1728, -0.2039], + ..., + [-0.2173, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1127], + [-0.1463, -0.1021, -0.0930, ..., -0.1759, -0.0187, -0.2249], + [-0.2169, -0.0811, -0.1790, ..., -0.2309, -0.0594, 0.1607]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 2.3283e-09, + 9.3132e-10, -7.4506e-08], + [ 1.3970e-09, 0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 4.1910e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -6.2864e-08, ..., 4.6566e-10, + 2.3283e-09, 1.8626e-09], + ..., + [ 2.7940e-09, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + -1.0710e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.8429e-08, + 2.7940e-09, 1.0245e-08], + [ 8.3819e-09, 0.0000e+00, 2.1886e-08, ..., 4.6566e-10, + 3.7253e-09, 4.0978e-08]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0380, -0.0302, 0.0087, 0.0304, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([-1.1455e-07, 2.8871e-08, -9.3598e-08, -1.7695e-08, -1.3970e-08, + -2.1420e-08, -8.3819e-08, 1.5832e-08, 1.8626e-07, 1.2293e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 214.20, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4581 re_mapping 0.0023 re_causal 0.0085 /// teacc 99.19 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.1506, 0.0565, 0.0935, ..., -0.1014, -0.1926, 0.0741], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1116], + [-0.0807, 0.0155, 0.2505, ..., -0.1749, -0.1728, -0.2039], + ..., + [-0.2173, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1128], + [-0.1463, -0.1021, -0.0930, ..., -0.1760, -0.0187, -0.2250], + [-0.2169, -0.0811, -0.1790, ..., -0.2309, -0.0594, 0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., 0.0000e+00, + 9.3132e-10, -1.6298e-08], + [ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., -1.8626e-09, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -3.4738e-07, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0291e-07, ..., 1.3970e-09, + -1.8626e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 2.1234e-07, ..., 4.6566e-10, + 4.4284e-07, 4.8429e-08], + [ 4.6566e-10, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 1.3970e-09, -4.6566e-10]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0379, -0.0302, 0.0086, 0.0304, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([-3.3993e-08, 2.0023e-08, -6.7102e-07, 8.4285e-08, 1.1176e-08, + -1.1716e-06, 1.1921e-07, 2.2398e-07, 1.4128e-06, 6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 214.23, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4037 re_mapping 0.0023 re_causal 0.0081 /// teacc 99.20 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.1506, 0.0565, 0.0935, ..., -0.1014, -0.1926, 0.0741], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1116], + [-0.0807, 0.0155, 0.2505, ..., -0.1749, -0.1729, -0.2039], + ..., + [-0.2174, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1129], + [-0.1464, -0.1021, -0.0930, ..., -0.1761, -0.0187, -0.2250], + [-0.2169, -0.0811, -0.1791, ..., -0.2310, -0.0595, 0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., 8.8476e-09, + 0.0000e+00, -4.6566e-09], + [ 4.6566e-10, 0.0000e+00, 1.2573e-08, ..., -4.6566e-10, + 1.1176e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 6.9849e-09, 4.6566e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -2.6543e-08, ..., 3.2596e-09, + -2.6543e-08, 2.5146e-08], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 1.0245e-08, + 2.7940e-09, 2.7940e-09], + [ 3.2596e-09, 0.0000e+00, 6.0536e-09, ..., 9.3132e-10, + 2.7940e-09, -4.2375e-08]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0379, -0.0302, 0.0086, 0.0304, -0.0186, 0.0063, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 2.6077e-08, 4.4238e-08, 2.0023e-08, 2.4214e-08, 6.7055e-08, + 5.0757e-08, -1.3411e-07, 2.6543e-08, 1.7695e-08, -1.2852e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 214.40, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4039 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.17 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.1507, 0.0565, 0.0936, ..., -0.1014, -0.1927, 0.0742], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1116], + [-0.0808, 0.0155, 0.2505, ..., -0.1749, -0.1729, -0.2040], + ..., + [-0.2174, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1129], + [-0.1464, -0.1021, -0.0930, ..., -0.1762, -0.0187, -0.2250], + [-0.2169, -0.0811, -0.1791, ..., -0.2310, -0.0595, 0.1608]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 1.6298e-08, + 0.0000e+00, -9.7789e-09], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., -2.7474e-08, + 3.7253e-09, 2.7940e-09], + [-0.0000e+00, 0.0000e+00, -2.1420e-08, ..., 6.0536e-09, + 2.7940e-09, 4.6566e-10], + ..., + [ 1.3970e-09, 0.0000e+00, 1.1176e-08, ..., 1.7695e-08, + -1.4435e-08, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 1.2107e-08, + 1.3970e-09, 1.0245e-08], + [ 7.4506e-09, 0.0000e+00, -4.1910e-09, ..., 1.3970e-09, + 4.1910e-09, -4.8429e-08]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0379, -0.0302, 0.0086, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 7.4040e-08, -5.1688e-08, 4.6566e-09, 1.4435e-08, 5.8208e-08, + 1.6298e-08, -1.3411e-07, 5.4948e-08, 8.1956e-08, -1.0943e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 214.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4245 re_mapping 0.0023 re_causal 0.0083 /// teacc 99.19 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.1507, 0.0565, 0.0936, ..., -0.1014, -0.1927, 0.0742], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1116], + [-0.0808, 0.0155, 0.2506, ..., -0.1749, -0.1729, -0.2040], + ..., + [-0.2175, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1130], + [-0.1464, -0.1021, -0.0930, ..., -0.1763, -0.0188, -0.2251], + [-0.2170, -0.0811, -0.1792, ..., -0.2310, -0.0596, 0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, -2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, -9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 1.3970e-09, ..., -3.7253e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 4.1910e-09, ..., 3.2596e-09, + 0.0000e+00, 5.5879e-09], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 2.7940e-09, + 0.0000e+00, 1.3970e-09], + [ 2.3283e-09, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -1.7229e-08]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0379, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0051], device='cuda:0'), grad: tensor([ 2.2817e-08, 5.1223e-09, -9.3132e-10, -5.5879e-07, 2.1420e-08, + 4.9826e-07, -1.7229e-08, 5.4017e-08, 2.1886e-08, -3.6322e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 214.20, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4209 re_mapping 0.0023 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.1507, 0.0565, 0.0936, ..., -0.1014, -0.1927, 0.0742], + [-0.1314, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1117], + [-0.0808, 0.0155, 0.2506, ..., -0.1750, -0.1729, -0.2040], + ..., + [-0.2175, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1130], + [-0.1464, -0.1021, -0.0930, ..., -0.1764, -0.0188, -0.2251], + [-0.2170, -0.0811, -0.1792, ..., -0.2310, -0.0596, 0.1608]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, -1.6298e-08, ..., 6.0536e-09, + 0.0000e+00, -7.9162e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., -7.4506e-09, + 8.3819e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 9.3132e-10, + 4.6566e-09, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, -5.5879e-09, ..., 7.4506e-09, + 4.6566e-10, 3.7253e-09], + [ 3.2596e-09, 0.0000e+00, 3.7253e-09, ..., 4.6566e-09, + 6.9849e-09, 2.3749e-08], + [-6.5193e-09, 0.0000e+00, 1.3970e-08, ..., 0.0000e+00, + 1.0245e-08, -3.7253e-08]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0379, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0052], device='cuda:0'), grad: tensor([-2.7940e-09, 1.5367e-08, 2.1886e-08, -1.3923e-07, 2.7474e-08, + -1.3970e-09, -9.7789e-09, 1.6065e-07, -2.9337e-08, -4.1444e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 214.25, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4482 re_mapping 0.0023 re_causal 0.0085 /// teacc 99.18 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.1507, 0.0565, 0.0937, ..., -0.1015, -0.1927, 0.0742], + [-0.1315, -0.0661, -0.1516, ..., 0.2021, -0.1204, -0.1117], + [-0.0808, 0.0155, 0.2506, ..., -0.1750, -0.1730, -0.2040], + ..., + [-0.2176, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1131], + [-0.1465, -0.1021, -0.0930, ..., -0.1765, -0.0188, -0.2251], + [-0.2170, -0.0811, -0.1792, ..., -0.2311, -0.0596, 0.1608]], + device='cuda:0'), grad: tensor([[ 3.8184e-08, 0.0000e+00, -1.8626e-09, ..., 2.4214e-08, + 0.0000e+00, 9.3132e-10], + [ 4.3772e-08, 0.0000e+00, 3.3528e-08, ..., 1.1176e-08, + 3.6322e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 3.7253e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -5.0291e-08, ..., 1.2107e-08, + -5.4017e-08, 3.7253e-09], + [ 2.4214e-08, 0.0000e+00, 1.8626e-09, ..., 1.6764e-08, + 1.8626e-09, 9.3132e-10], + [ 8.3819e-09, 0.0000e+00, 1.2107e-08, ..., 2.7940e-09, + 1.1176e-08, -2.8871e-08]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0379, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0052], device='cuda:0'), grad: tensor([ 1.1735e-07, 2.0862e-07, 2.8871e-08, 5.3085e-08, 6.0536e-07, + 2.4401e-07, -1.1446e-06, -1.3225e-07, 1.1176e-08, -6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 214.50, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4251 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.20 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.1508, 0.0566, 0.0937, ..., -0.1015, -0.1927, 0.0742], + [-0.1315, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1117], + [-0.0808, 0.0155, 0.2506, ..., -0.1750, -0.1730, -0.2040], + ..., + [-0.2176, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1132], + [-0.1465, -0.1021, -0.0930, ..., -0.1766, -0.0188, -0.2251], + [-0.2170, -0.0811, -0.1793, ..., -0.2311, -0.0597, 0.1608]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -5.5879e-09, ..., 9.3132e-09, + 9.3132e-10, 9.3132e-10], + [ 2.3283e-08, 0.0000e+00, 1.6764e-08, ..., -1.3039e-08, + 2.7940e-08, 3.7253e-09], + [ 4.6566e-09, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 9.3132e-09, 9.3132e-10], + ..., + [ 1.3039e-08, 0.0000e+00, -3.9116e-08, ..., 8.3819e-09, + -6.3330e-08, 1.6764e-08], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 9.3132e-09, + 5.5879e-09, 3.7253e-09], + [ 3.0734e-08, 0.0000e+00, 1.3039e-08, ..., 9.3132e-10, + 1.5832e-08, -5.1223e-08]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0379, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0052], device='cuda:0'), grad: tensor([ 3.1665e-08, 9.1270e-08, 4.0978e-08, 2.6077e-08, -7.3574e-08, + -3.7253e-08, -1.2107e-08, -9.3132e-08, 5.3085e-08, -1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 214.47, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4216 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.19 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.1508, 0.0566, 0.0938, ..., -0.1015, -0.1927, 0.0742], + [-0.1315, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1117], + [-0.0808, 0.0155, 0.2506, ..., -0.1751, -0.1730, -0.2041], + ..., + [-0.2177, -0.0778, 0.0329, ..., -0.0685, 0.1325, -0.1132], + [-0.1465, -0.1021, -0.0930, ..., -0.1767, -0.0188, -0.2252], + [-0.2170, -0.0812, -0.1794, ..., -0.2311, -0.0597, 0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 0.0000e+00, -9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + 9.3132e-10, 1.8626e-09], + [-1.8626e-09, 0.0000e+00, -1.6764e-08, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, 2.7940e-08], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., -4.6566e-09, + 0.0000e+00, 9.3132e-10], + [ 1.5460e-07, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0379, -0.0302, 0.0087, 0.0304, -0.0186, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0052], device='cuda:0'), grad: tensor([ 3.0734e-08, 1.2014e-07, -2.7940e-08, -6.2399e-08, -2.2911e-07, + 7.5437e-08, -1.3970e-08, 1.0803e-07, -1.3970e-07, 1.5274e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 214.49, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4086 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.19 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.1508, 0.0566, 0.0939, ..., -0.1015, -0.1928, 0.0743], + [-0.1315, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1118], + [-0.0809, 0.0155, 0.2506, ..., -0.1751, -0.1730, -0.2041], + ..., + [-0.2177, -0.0779, 0.0329, ..., -0.0685, 0.1325, -0.1133], + [-0.1465, -0.1021, -0.0931, ..., -0.1768, -0.0188, -0.2252], + [-0.2171, -0.0812, -0.1795, ..., -0.2312, -0.0597, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.9802e-08, ..., 0.0000e+00, + -3.7253e-09, -5.0291e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -5.5879e-09, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 2.7940e-09, + -4.6566e-09, 2.1420e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -2.7940e-09, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, -3.3528e-08]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0378, -0.0302, 0.0087, 0.0304, -0.0185, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0052], device='cuda:0'), grad: tensor([-8.3819e-08, 9.3132e-10, 1.5832e-08, 7.3574e-08, 3.3528e-08, + 1.0896e-07, 3.8184e-08, 7.9162e-08, -1.7695e-07, -8.3819e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 214.30, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4395 re_mapping 0.0023 re_causal 0.0083 /// teacc 99.19 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.1509, 0.0566, 0.0941, ..., -0.1016, -0.1928, 0.0744], + [-0.1315, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1119], + [-0.0809, 0.0155, 0.2506, ..., -0.1752, -0.1730, -0.2042], + ..., + [-0.2178, -0.0779, 0.0329, ..., -0.0685, 0.1325, -0.1134], + [-0.1465, -0.1021, -0.0931, ..., -0.1769, -0.0188, -0.2253], + [-0.2171, -0.0812, -0.1797, ..., -0.2312, -0.0598, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., 9.3132e-10, + 9.3132e-10, -4.6566e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-07, ..., -2.4214e-08, + 1.0245e-07, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -2.3283e-07, ..., 2.2352e-08, + -2.1048e-07, 3.3621e-07], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 9.3132e-10, + 5.5879e-09, 7.4506e-09], + [ 9.3132e-10, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 9.3132e-09, -3.5577e-07]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0377, -0.0302, 0.0086, 0.0304, -0.0185, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0053], device='cuda:0'), grad: tensor([-1.2107e-07, 5.6904e-07, 3.3528e-08, 3.9116e-08, 2.7195e-07, + 5.5879e-09, 1.3039e-08, -5.6811e-08, 2.4214e-08, -7.7486e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 214.31, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4230 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.20 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.1509, 0.0566, 0.0942, ..., -0.1016, -0.1928, 0.0744], + [-0.1315, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1119], + [-0.0809, 0.0155, 0.2507, ..., -0.1752, -0.1731, -0.2042], + ..., + [-0.2179, -0.0779, 0.0330, ..., -0.0685, 0.1325, -0.1136], + [-0.1466, -0.1021, -0.0931, ..., -0.1771, -0.0188, -0.2254], + [-0.2172, -0.0812, -0.1798, ..., -0.2312, -0.0598, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 1.3970e-08, + 0.0000e+00, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -2.6077e-08, + 4.6566e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 5.5879e-09, + 2.7940e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., 1.6764e-08, + -1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.7695e-08, + 2.7940e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 2.7940e-09, 7.4506e-09]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0377, -0.0302, 0.0087, 0.0304, -0.0185, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0053], device='cuda:0'), grad: tensor([ 4.9360e-08, -5.1223e-08, 1.9558e-08, 8.3819e-09, 7.4506e-09, + 6.3330e-08, -2.1234e-07, 1.1176e-08, 7.8231e-08, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 214.35, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4236 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.19 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.1510, 0.0566, 0.0943, ..., -0.1016, -0.1928, 0.0744], + [-0.1316, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1119], + [-0.0810, 0.0155, 0.2507, ..., -0.1752, -0.1731, -0.2042], + ..., + [-0.2180, -0.0779, 0.0330, ..., -0.0685, 0.1325, -0.1137], + [-0.1467, -0.1021, -0.0931, ..., -0.1772, -0.0188, -0.2254], + [-0.2172, -0.0812, -0.1799, ..., -0.2312, -0.0599, 0.1607]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 1.7695e-08, ..., -3.7253e-09, + 3.2596e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -3.0734e-08, ..., 2.7940e-09, + -5.3085e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 1.8626e-09, 2.7940e-09], + [ 2.7940e-09, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 1.1176e-08, -9.3132e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0377, -0.0302, 0.0087, 0.0304, -0.0185, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0053], device='cuda:0'), grad: tensor([ 8.3819e-09, 6.5193e-08, 1.2107e-08, 1.4901e-08, 9.3132e-09, + 1.3039e-08, -2.4214e-08, -9.5926e-08, -6.5193e-09, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 214.44, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4420 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.19 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.1510, 0.0566, 0.0944, ..., -0.1017, -0.1929, 0.0745], + [-0.1316, -0.0661, -0.1517, ..., 0.2021, -0.1204, -0.1120], + [-0.0810, 0.0155, 0.2507, ..., -0.1753, -0.1731, -0.2043], + ..., + [-0.2180, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1138], + [-0.1468, -0.1021, -0.0931, ..., -0.1774, -0.0188, -0.2255], + [-0.2173, -0.0812, -0.1800, ..., -0.2313, -0.0599, 0.1607]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -9.3132e-10, ..., 7.4506e-09, + 1.8626e-09, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., -8.3819e-09, + 3.7253e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, + 4.6566e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, -1.3970e-08, ..., 7.4506e-09, + -1.9558e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.0245e-08, ..., 4.6566e-09, + 1.0245e-08, 6.5193e-09], + [ 7.4506e-09, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 3.7253e-09]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0376, -0.0302, 0.0086, 0.0303, -0.0185, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0053], device='cuda:0'), grad: tensor([ 6.6124e-08, 4.6566e-09, 5.5879e-09, -4.6566e-08, -1.5832e-08, + -1.3877e-07, 5.4017e-08, -1.3039e-08, 6.9849e-08, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 214.45, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4368 re_mapping 0.0021 re_causal 0.0081 /// teacc 99.23 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.1511, 0.0566, 0.0945, ..., -0.1017, -0.1929, 0.0745], + [-0.1316, -0.0661, -0.1518, ..., 0.2021, -0.1204, -0.1120], + [-0.0810, 0.0155, 0.2507, ..., -0.1753, -0.1731, -0.2043], + ..., + [-0.2181, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1140], + [-0.1468, -0.1021, -0.0931, ..., -0.1775, -0.0188, -0.2256], + [-0.2173, -0.0812, -0.1800, ..., -0.2313, -0.0599, 0.1607]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.1176e-08, ..., 9.3132e-10, + 0.0000e+00, -1.3039e-08], + [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., -1.5832e-08, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + -0.0000e+00, 2.7940e-09], + [ 7.1712e-08, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0376, -0.0302, 0.0086, 0.0303, -0.0185, 0.0064, -0.0172, 0.0160, + 0.0097, -0.0054], device='cuda:0'), grad: tensor([-2.0489e-08, -1.3970e-08, 7.4506e-09, -2.5146e-08, -1.8533e-07, + 1.6764e-08, 1.6764e-08, 5.3085e-08, 1.4901e-08, 1.5460e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 214.30, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4083 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.21 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.1511, 0.0566, 0.0945, ..., -0.1018, -0.1929, 0.0745], + [-0.1316, -0.0661, -0.1518, ..., 0.2021, -0.1204, -0.1120], + [-0.0810, 0.0155, 0.2507, ..., -0.1753, -0.1732, -0.2044], + ..., + [-0.2182, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1141], + [-0.1468, -0.1021, -0.0931, ..., -0.1777, -0.0188, -0.2256], + [-0.2174, -0.0812, -0.1801, ..., -0.2314, -0.0600, 0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.6566e-09, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 2.7940e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.7940e-09, + 9.3132e-10, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 0.0000e+00, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -3.9116e-08]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0376, -0.0302, 0.0086, 0.0303, -0.0185, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0054], device='cuda:0'), grad: tensor([ 3.6322e-08, 1.3039e-08, -1.8626e-09, 1.0245e-08, 1.7695e-08, + 6.5193e-09, -2.7008e-08, 8.4750e-08, -4.0047e-08, -9.4064e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 214.28, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4193 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.19 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.1512, 0.0566, 0.0946, ..., -0.1018, -0.1930, 0.0745], + [-0.1317, -0.0661, -0.1518, ..., 0.2021, -0.1204, -0.1121], + [-0.0810, 0.0155, 0.2508, ..., -0.1754, -0.1732, -0.2044], + ..., + [-0.2183, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1143], + [-0.1469, -0.1021, -0.0931, ..., -0.1778, -0.0188, -0.2256], + [-0.2174, -0.0812, -0.1802, ..., -0.2314, -0.0600, 0.1608]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 0.0000e+00, 2.9802e-08, ..., 9.3132e-09, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-09, 0.0000e+00, 3.7253e-09, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -7.5437e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 2.0489e-08, ..., 1.8626e-09, + 0.0000e+00, 4.6566e-09], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 5.5879e-09, + -0.0000e+00, 9.3132e-10], + [ 4.2282e-07, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -8.3819e-09]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0376, -0.0302, 0.0086, 0.0303, -0.0185, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0054], device='cuda:0'), grad: tensor([ 1.0058e-07, 1.7695e-08, -1.1642e-07, 1.3039e-08, -5.9418e-07, + 6.5193e-08, -7.1712e-08, 6.7055e-08, -8.3819e-09, 5.3924e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 214.36, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4527 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.1512, 0.0566, 0.0946, ..., -0.1018, -0.1930, 0.0746], + [-0.1317, -0.0661, -0.1518, ..., 0.2021, -0.1204, -0.1121], + [-0.0811, 0.0155, 0.2508, ..., -0.1754, -0.1732, -0.2044], + ..., + [-0.2183, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1144], + [-0.1469, -0.1021, -0.0931, ..., -0.1779, -0.0188, -0.2257], + [-0.2175, -0.0812, -0.1802, ..., -0.2314, -0.0601, 0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.9802e-08, ..., 8.3819e-09, + 0.0000e+00, -9.3132e-08], + [ 7.4506e-09, 0.0000e+00, 3.7253e-09, ..., -1.0245e-07, + 9.3132e-10, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, -8.5682e-08, ..., 1.8626e-09, + 9.3132e-10, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 2.5146e-08, + -0.0000e+00, 6.1467e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 4.0047e-08, + 0.0000e+00, 1.1176e-08], + [ 1.3039e-08, 0.0000e+00, 2.9802e-08, ..., 1.8626e-09, + -9.3132e-10, 2.7940e-09]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0375, -0.0302, 0.0086, 0.0303, -0.0184, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0054], device='cuda:0'), grad: tensor([-1.4715e-07, -1.6484e-07, -1.5367e-07, 9.6858e-08, -1.6764e-08, + 1.0990e-07, -6.7987e-08, 2.1700e-07, 9.2201e-08, 4.7497e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 214.46, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4365 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.17 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.1512, 0.0567, 0.0947, ..., -0.1019, -0.1930, 0.0746], + [-0.1317, -0.0661, -0.1519, ..., 0.2021, -0.1204, -0.1122], + [-0.0811, 0.0155, 0.2508, ..., -0.1754, -0.1733, -0.2044], + ..., + [-0.2184, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1146], + [-0.1469, -0.1021, -0.0931, ..., -0.1781, -0.0188, -0.2257], + [-0.2176, -0.0812, -0.1803, ..., -0.2314, -0.0601, 0.1608]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., -9.3132e-10, + 1.1176e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -9.3132e-09, ..., 1.8626e-09, + -3.0734e-08, 1.0245e-08], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 1.8626e-09, + 2.7940e-09, 7.4506e-09], + [ 1.8626e-09, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 9.3132e-09, -4.2841e-08]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0375, -0.0302, 0.0086, 0.0303, -0.0184, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0054], device='cuda:0'), grad: tensor([ 4.5635e-08, 5.2154e-08, -2.7940e-08, 2.8871e-08, 4.0047e-08, + 6.7987e-08, -1.0245e-08, -1.8626e-09, -1.4715e-07, -5.8673e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 214.28, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4038 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.19 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.1513, 0.0567, 0.0948, ..., -0.1019, -0.1930, 0.0746], + [-0.1317, -0.0661, -0.1519, ..., 0.2021, -0.1204, -0.1122], + [-0.0811, 0.0155, 0.2509, ..., -0.1755, -0.1733, -0.2044], + ..., + [-0.2186, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1147], + [-0.1470, -0.1021, -0.0931, ..., -0.1782, -0.0188, -0.2258], + [-0.2176, -0.0812, -0.1804, ..., -0.2314, -0.0602, 0.1608]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 7.4506e-09, + 0.0000e+00, -8.3819e-09], + [ 1.6764e-08, 0.0000e+00, 1.7695e-08, ..., 6.5193e-09, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -4.4703e-08, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 2.1420e-08, 0.0000e+00, 5.5879e-09, ..., 5.5879e-09, + -4.6566e-09, 3.7253e-09], + [ 2.7940e-09, 0.0000e+00, 1.0245e-08, ..., 2.7940e-09, + 4.6566e-09, 1.8626e-09], + [ 1.6764e-08, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 9.3132e-10, -3.7253e-09]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0375, -0.0302, 0.0086, 0.0303, -0.0184, 0.0063, -0.0171, 0.0160, + 0.0097, -0.0055], device='cuda:0'), grad: tensor([ 1.2107e-08, 7.7300e-08, -7.4506e-08, 3.0734e-08, -7.9162e-08, + 1.6019e-07, -2.3562e-07, 5.7742e-08, 4.6566e-08, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 214.25, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4324 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.19 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.1513, 0.0567, 0.0948, ..., -0.1020, -0.1930, 0.0747], + [-0.1318, -0.0661, -0.1519, ..., 0.2021, -0.1205, -0.1122], + [-0.0811, 0.0155, 0.2509, ..., -0.1755, -0.1733, -0.2045], + ..., + [-0.2186, -0.0779, 0.0330, ..., -0.0685, 0.1326, -0.1149], + [-0.1470, -0.1021, -0.0931, ..., -0.1784, -0.0188, -0.2258], + [-0.2177, -0.0812, -0.1804, ..., -0.2315, -0.0603, 0.1608]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -1.8626e-09, ..., 4.6566e-09, + 9.3132e-10, -9.3132e-10], + [ 2.8871e-08, 0.0000e+00, 3.7253e-09, ..., -5.5879e-09, + 1.2107e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 8.3726e-07, 0.0000e+00, -1.8626e-09, ..., 4.6566e-09, + 1.5274e-07, 1.5832e-08], + [ 7.4506e-09, 0.0000e+00, 9.3132e-10, ..., 7.4506e-09, + 3.7253e-09, 1.8626e-09], + [ 6.5845e-07, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 1.3132e-07, 5.5879e-09]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0375, -0.0302, 0.0086, 0.0303, -0.0184, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0055], device='cuda:0'), grad: tensor([ 2.0489e-08, 5.7742e-08, -6.5193e-09, 2.9802e-08, -2.8685e-06, + -5.5879e-09, -3.5390e-08, 1.5786e-06, 2.7940e-09, 1.2247e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 214.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4058 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.18 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.1514, 0.0567, 0.0949, ..., -0.1020, -0.1931, 0.0747], + [-0.1318, -0.0661, -0.1519, ..., 0.2021, -0.1205, -0.1123], + [-0.0811, 0.0155, 0.2509, ..., -0.1755, -0.1734, -0.2045], + ..., + [-0.2188, -0.0779, 0.0331, ..., -0.0685, 0.1326, -0.1150], + [-0.1471, -0.1021, -0.0931, ..., -0.1785, -0.0188, -0.2259], + [-0.2178, -0.0812, -0.1805, ..., -0.2315, -0.0603, 0.1608]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -8.6613e-08, ..., 1.1176e-08, + 0.0000e+00, -1.5646e-07], + [ 1.0245e-08, 0.0000e+00, 7.4506e-09, ..., -2.7940e-09, + 9.3132e-10, 2.7940e-09], + [ 2.0489e-08, 0.0000e+00, 1.3039e-08, ..., 6.0536e-08, + 9.3132e-10, 2.7940e-09], + ..., + [ 3.7253e-09, 0.0000e+00, -0.0000e+00, ..., 6.5193e-09, + -1.8626e-09, 1.2107e-08], + [ 5.5879e-09, 0.0000e+00, 1.7695e-08, ..., 1.8626e-08, + 0.0000e+00, 2.0489e-08], + [ 3.9116e-08, 0.0000e+00, 5.8673e-08, ..., 9.3132e-10, + -9.3132e-10, 8.5682e-08]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0375, -0.0302, 0.0086, 0.0302, -0.0183, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0055], device='cuda:0'), grad: tensor([-4.4238e-07, 1.8626e-08, 1.6484e-07, 4.4703e-08, -1.0245e-08, + 5.4948e-08, -3.0920e-07, 4.8429e-08, 9.4995e-08, 3.4273e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 214.16, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4427 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.21 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.1514, 0.0567, 0.0949, ..., -0.1020, -0.1931, 0.0747], + [-0.1318, -0.0661, -0.1520, ..., 0.2021, -0.1205, -0.1123], + [-0.0812, 0.0155, 0.2510, ..., -0.1755, -0.1734, -0.2045], + ..., + [-0.2188, -0.0779, 0.0331, ..., -0.0685, 0.1326, -0.1151], + [-0.1471, -0.1021, -0.0931, ..., -0.1786, -0.0188, -0.2259], + [-0.2178, -0.0812, -0.1806, ..., -0.2316, -0.0604, 0.1609]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, -2.1420e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.9802e-08, + 4.6566e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.7940e-08, + -9.3132e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 9.3132e-10, + 2.7940e-09, -2.2352e-08]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0375, -0.0302, 0.0087, 0.0302, -0.0183, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0055], device='cuda:0'), grad: tensor([-5.0291e-08, -4.2841e-08, 4.6566e-09, 1.8626e-08, 1.4901e-08, + 6.1467e-08, 6.5193e-09, 3.3528e-08, 1.0245e-08, -3.6322e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 214.26, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4465 re_mapping 0.0021 re_causal 0.0083 /// teacc 99.20 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.1515, 0.0567, 0.0950, ..., -0.1021, -0.1932, 0.0747], + [-0.1318, -0.0661, -0.1520, ..., 0.2021, -0.1205, -0.1124], + [-0.0812, 0.0155, 0.2510, ..., -0.1756, -0.1734, -0.2045], + ..., + [-0.2189, -0.0779, 0.0331, ..., -0.0685, 0.1326, -0.1153], + [-0.1472, -0.1021, -0.0931, ..., -0.1788, -0.0188, -0.2260], + [-0.2179, -0.0812, -0.1807, ..., -0.2316, -0.0605, 0.1609]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 2.7940e-09, + 0.0000e+00, -1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 5.5879e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 4.6566e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, -6.5193e-09, ..., 9.3132e-10, + -1.3039e-08, 8.3819e-09], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + 9.3132e-10, 3.7253e-09], + [ 2.9802e-08, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 1.8626e-09, -2.9802e-08]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0375, -0.0302, 0.0087, 0.0302, -0.0183, 0.0064, -0.0171, 0.0160, + 0.0097, -0.0055], device='cuda:0'), grad: tensor([ 3.7253e-09, 3.2596e-08, 1.2107e-08, 5.5879e-09, 2.1420e-08, + 2.7847e-07, -2.9895e-07, 1.4901e-08, 1.8626e-09, -6.1467e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 214.36, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4025 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.1515, 0.0567, 0.0951, ..., -0.1021, -0.1932, 0.0748], + [-0.1319, -0.0661, -0.1520, ..., 0.2021, -0.1205, -0.1124], + [-0.0812, 0.0155, 0.2510, ..., -0.1756, -0.1735, -0.2045], + ..., + [-0.2190, -0.0779, 0.0331, ..., -0.0685, 0.1327, -0.1154], + [-0.1472, -0.1021, -0.0931, ..., -0.1789, -0.0188, -0.2261], + [-0.2179, -0.0812, -0.1808, ..., -0.2316, -0.0606, 0.1609]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.4901e-08, ..., 1.8626e-09, + 0.0000e+00, -4.5635e-08], + [-1.8626e-08, 0.0000e+00, 9.3132e-09, ..., -1.1269e-07, + 2.0489e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 2.7940e-09, 9.3132e-10], + ..., + [ 4.1910e-08, 0.0000e+00, -1.6764e-08, ..., 2.4214e-08, + -3.2596e-08, 8.3819e-09], + [ 1.4901e-08, 0.0000e+00, 8.3819e-09, ..., 7.0781e-08, + 8.3819e-09, 9.3132e-09], + [ 6.7055e-08, 0.0000e+00, 1.0245e-08, ..., 3.7253e-09, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0374, -0.0302, 0.0087, 0.0302, -0.0183, 0.0064, -0.0171, 0.0160, + 0.0096, -0.0055], device='cuda:0'), grad: tensor([-8.0094e-08, -2.5518e-07, 1.1176e-08, 7.4506e-09, -1.2014e-07, + 1.0245e-08, 5.5879e-09, 5.4948e-08, 2.3935e-07, 1.1362e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 214.21, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4300 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.20 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.1516, 0.0567, 0.0952, ..., -0.1022, -0.1932, 0.0748], + [-0.1319, -0.0661, -0.1520, ..., 0.2021, -0.1205, -0.1124], + [-0.0812, 0.0155, 0.2511, ..., -0.1756, -0.1735, -0.2046], + ..., + [-0.2190, -0.0779, 0.0331, ..., -0.0685, 0.1327, -0.1155], + [-0.1473, -0.1021, -0.0932, ..., -0.1791, -0.0188, -0.2262], + [-0.2179, -0.0813, -0.1810, ..., -0.2317, -0.0607, 0.1609]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -2.7940e-09, ..., 2.7940e-09, + 1.8626e-09, -3.7253e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., -7.4506e-09, + 1.4901e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 4.6566e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -4.5635e-08, ..., 8.3819e-09, + -9.2201e-08, 5.5879e-09], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 3.7253e-09, + 1.1176e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 1.3970e-08, 5.5879e-09]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0374, -0.0302, 0.0087, 0.0302, -0.0184, 0.0064, -0.0171, 0.0160, + 0.0096, -0.0056], device='cuda:0'), grad: tensor([-4.4703e-08, 3.5390e-08, 1.4901e-08, 5.2620e-07, 9.8720e-08, + -6.0163e-07, 4.5635e-08, -1.6112e-07, 5.2154e-08, 4.7497e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 213.85, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4211 re_mapping 0.0021 re_causal 0.0081 /// teacc 99.17 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.1517, 0.0567, 0.0953, ..., -0.1023, -0.1933, 0.0748], + [-0.1319, -0.0661, -0.1521, ..., 0.2021, -0.1205, -0.1124], + [-0.0812, 0.0155, 0.2511, ..., -0.1757, -0.1736, -0.2046], + ..., + [-0.2191, -0.0779, 0.0332, ..., -0.0685, 0.1327, -0.1156], + [-0.1473, -0.1021, -0.0932, ..., -0.1793, -0.0188, -0.2263], + [-0.2180, -0.0813, -0.1811, ..., -0.2317, -0.0608, 0.1609]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 1.3970e-09, + 2.3283e-09, -4.6566e-10], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 4.6566e-10], + [-9.3132e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00], + ..., + [ 5.5879e-09, 0.0000e+00, -5.5879e-09, ..., 9.3132e-10, + -1.1642e-08, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -2.1420e-08, ..., 3.7253e-09, + 1.9558e-08, 2.3283e-08], + [ 1.2107e-08, 0.0000e+00, 1.5367e-08, ..., 0.0000e+00, + 1.1642e-08, 4.6566e-09]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0374, -0.0302, 0.0087, 0.0302, -0.0184, 0.0064, -0.0170, 0.0160, + 0.0096, -0.0056], device='cuda:0'), grad: tensor([ 1.6764e-08, 1.6298e-08, 8.8476e-09, 5.5414e-08, -3.4459e-08, + -1.8161e-07, 1.6764e-08, 1.5367e-08, -2.7940e-09, 9.5926e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 214.15, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4438 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.1517, 0.0567, 0.0954, ..., -0.1023, -0.1933, 0.0749], + [-0.1320, -0.0661, -0.1521, ..., 0.2021, -0.1205, -0.1124], + [-0.0813, 0.0155, 0.2511, ..., -0.1757, -0.1737, -0.2046], + ..., + [-0.2193, -0.0779, 0.0332, ..., -0.0685, 0.1327, -0.1158], + [-0.1474, -0.1021, -0.0932, ..., -0.1795, -0.0188, -0.2263], + [-0.2181, -0.0813, -0.1813, ..., -0.2318, -0.0609, 0.1609]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.2596e-09, ..., 9.3132e-10, + 9.3132e-10, -9.7789e-09], + [ 0.0000e+00, 0.0000e+00, 2.0023e-08, ..., -1.6764e-08, + 3.2596e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -1.6112e-07, ..., -1.7229e-08, + 9.3132e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 1.5832e-08, ..., 9.3132e-09, + -6.5193e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 9.2667e-08, ..., 1.8161e-08, + 2.7940e-09, 1.3970e-09], + [ 1.3970e-09, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 9.3132e-10, -9.3132e-10]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0374, -0.0302, 0.0086, 0.0301, -0.0183, 0.0065, -0.0170, 0.0160, + 0.0096, -0.0057], device='cuda:0'), grad: tensor([-1.2573e-08, 4.6566e-09, -3.3667e-07, 2.3982e-07, 4.0978e-08, + -2.2352e-07, 2.3749e-08, 5.2154e-08, 2.2072e-07, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 214.32, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4214 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.20 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.1518, 0.0567, 0.0955, ..., -0.1024, -0.1934, 0.0749], + [-0.1320, -0.0661, -0.1522, ..., 0.2021, -0.1205, -0.1125], + [-0.0813, 0.0155, 0.2511, ..., -0.1757, -0.1738, -0.2047], + ..., + [-0.2193, -0.0779, 0.0332, ..., -0.0685, 0.1327, -0.1159], + [-0.1474, -0.1021, -0.0932, ..., -0.1797, -0.0188, -0.2264], + [-0.2181, -0.0813, -0.1814, ..., -0.2318, -0.0610, 0.1609]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2107e-08, ..., 0.0000e+00, + 4.6566e-10, -1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -1.3970e-09, + 2.1420e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 1.2573e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -4.7497e-08, ..., 9.3132e-10, + -7.4506e-08, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 6.5193e-09, 1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 0.0000e+00, + 2.0489e-08, -2.2352e-08]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0373, -0.0302, 0.0086, 0.0301, -0.0183, 0.0065, -0.0170, 0.0160, + 0.0096, -0.0057], device='cuda:0'), grad: tensor([-4.7032e-08, 6.1002e-08, 2.5611e-08, -1.3039e-08, 4.0047e-08, + 4.6566e-08, 2.0955e-08, -1.7229e-07, 5.3085e-08, -1.2573e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 214.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4284 re_mapping 0.0020 re_causal 0.0080 /// teacc 99.19 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.1519, 0.0567, 0.0956, ..., -0.1024, -0.1936, 0.0750], + [-0.1320, -0.0661, -0.1522, ..., 0.2022, -0.1205, -0.1125], + [-0.0813, 0.0155, 0.2511, ..., -0.1758, -0.1738, -0.2047], + ..., + [-0.2194, -0.0779, 0.0333, ..., -0.0686, 0.1328, -0.1160], + [-0.1475, -0.1021, -0.0932, ..., -0.1799, -0.0188, -0.2265], + [-0.2182, -0.0813, -0.1816, ..., -0.2319, -0.0612, 0.1609]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.7789e-09, ..., 0.0000e+00, + 4.6566e-10, -1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., -1.0664e-07, + 3.3062e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -6.1002e-08, ..., 3.2596e-09, + 6.0536e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 9.2201e-08, + -7.3109e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 5.5879e-09, + 5.1223e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 2.3283e-09, + 2.2817e-08, 4.6566e-09]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0373, -0.0302, 0.0085, 0.0301, -0.0183, 0.0065, -0.0170, 0.0160, + 0.0096, -0.0057], device='cuda:0'), grad: tensor([-3.3993e-08, -2.4680e-07, -8.0094e-08, -9.3132e-09, 1.0710e-08, + 9.3132e-09, 1.6298e-08, 1.9930e-07, 6.4727e-08, 7.5437e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 214.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4156 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.1519, 0.0567, 0.0957, ..., -0.1025, -0.1937, 0.0750], + [-0.1321, -0.0661, -0.1523, ..., 0.2022, -0.1206, -0.1126], + [-0.0814, 0.0155, 0.2511, ..., -0.1759, -0.1739, -0.2048], + ..., + [-0.2194, -0.0779, 0.0333, ..., -0.0686, 0.1328, -0.1162], + [-0.1475, -0.1021, -0.0932, ..., -0.1801, -0.0188, -0.2267], + [-0.2182, -0.0813, -0.1818, ..., -0.2319, -0.0613, 0.1610]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 5.1223e-09, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 1.0245e-08, ..., -7.4506e-09, + 2.5611e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 3.7253e-09, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, -1.4435e-08, ..., 6.0536e-09, + -4.1910e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -3.7253e-09, ..., 9.7789e-09, + 2.7940e-09, 1.3970e-09], + [ 2.7940e-08, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 7.4506e-09, 9.3132e-10]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0372, -0.0302, 0.0085, 0.0301, -0.0183, 0.0065, -0.0170, 0.0160, + 0.0096, -0.0058], device='cuda:0'), grad: tensor([ 2.4214e-08, 7.9162e-08, 2.5146e-08, 2.7474e-08, -3.1199e-08, + 2.5146e-08, -4.8429e-08, -1.1362e-07, -3.3993e-08, 7.0315e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 214.43, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3940 re_mapping 0.0021 re_causal 0.0078 /// teacc 99.18 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.1519, 0.0567, 0.0959, ..., -0.1025, -0.1938, 0.0751], + [-0.1321, -0.0661, -0.1523, ..., 0.2022, -0.1206, -0.1126], + [-0.0814, 0.0155, 0.2511, ..., -0.1759, -0.1740, -0.2049], + ..., + [-0.2195, -0.0779, 0.0334, ..., -0.0686, 0.1328, -0.1164], + [-0.1475, -0.1021, -0.0932, ..., -0.1802, -0.0188, -0.2267], + [-0.2183, -0.0813, -0.1819, ..., -0.2320, -0.0615, 0.1610]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.1223e-09, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 8.3819e-09, 0.0000e+00, 1.2573e-08, ..., -6.0536e-09, + 1.3504e-08, 1.3970e-09], + [ 2.3283e-09, 0.0000e+00, 7.4506e-09, ..., 4.6566e-10, + 5.5879e-09, 0.0000e+00], + ..., + [-3.0734e-08, 0.0000e+00, -6.1467e-08, ..., 4.1910e-09, + -7.0315e-08, 1.8626e-08], + [ 4.1910e-09, 0.0000e+00, 1.0245e-08, ..., 9.3132e-10, + 9.7789e-09, 5.1223e-09], + [ 6.0536e-09, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 4.6566e-10, -3.8184e-08]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0372, -0.0302, 0.0085, 0.0300, -0.0183, 0.0065, -0.0170, 0.0160, + 0.0096, -0.0058], device='cuda:0'), grad: tensor([-1.0710e-08, 3.6787e-08, 2.7474e-08, 9.3132e-10, 1.0291e-07, + 1.0245e-08, 3.7253e-09, -1.1642e-07, 1.7229e-08, -6.9849e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 214.12, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4496 re_mapping 0.0021 re_causal 0.0085 /// teacc 99.18 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.1520, 0.0567, 0.0960, ..., -0.1025, -0.1939, 0.0751], + [-0.1321, -0.0661, -0.1523, ..., 0.2022, -0.1206, -0.1127], + [-0.0814, 0.0155, 0.2511, ..., -0.1760, -0.1740, -0.2049], + ..., + [-0.2195, -0.0779, 0.0334, ..., -0.0686, 0.1328, -0.1166], + [-0.1475, -0.1021, -0.0932, ..., -0.1803, -0.0188, -0.2268], + [-0.2183, -0.0813, -0.1821, ..., -0.2320, -0.0616, 0.1610]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7229e-08, ..., 4.6566e-10, + 4.6566e-10, -2.4214e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., -3.1665e-08, + 2.7940e-09, -6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., 1.3970e-08, + -9.3132e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-09, + 1.8626e-09, 2.3283e-09], + [ 4.6566e-10, 0.0000e+00, 1.8161e-08, ..., 9.3132e-10, + 2.7940e-09, 2.3283e-08]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0371, -0.0302, 0.0084, 0.0300, -0.0184, 0.0065, -0.0170, 0.0160, + 0.0096, -0.0058], device='cuda:0'), grad: tensor([-4.8894e-08, -1.0105e-07, 4.2375e-08, -3.0734e-08, 4.6566e-09, + -1.1548e-07, 6.1002e-08, 8.5216e-08, 4.5635e-08, 7.0315e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 214.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3814 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.19 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.1520, 0.0568, 0.0962, ..., -0.1026, -0.1939, 0.0752], + [-0.1321, -0.0661, -0.1524, ..., 0.2022, -0.1206, -0.1127], + [-0.0814, 0.0155, 0.2512, ..., -0.1760, -0.1741, -0.2049], + ..., + [-0.2196, -0.0779, 0.0334, ..., -0.0686, 0.1329, -0.1167], + [-0.1475, -0.1021, -0.0932, ..., -0.1805, -0.0188, -0.2269], + [-0.2184, -0.0813, -0.1823, ..., -0.2321, -0.0618, 0.1610]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, -1.3970e-09, ..., 5.4017e-08, + 9.3132e-10, 6.5193e-09], + [ 2.7940e-09, 0.0000e+00, 3.7253e-09, ..., -6.0070e-07, + 7.9162e-09, -6.3330e-08], + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 3.2596e-09, + 2.3283e-09, 4.6566e-10], + ..., + [ 2.3283e-09, 0.0000e+00, -1.2573e-08, ..., 5.5786e-07, + -2.7008e-08, 7.5903e-08], + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 4.5635e-08, + 2.7940e-09, 8.8476e-09], + [ 1.8626e-08, 0.0000e+00, 6.9849e-09, ..., 3.6787e-08, + 1.2573e-08, -7.1712e-08]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0371, -0.0302, 0.0085, 0.0299, -0.0184, 0.0066, -0.0170, 0.0160, + 0.0096, -0.0059], device='cuda:0'), grad: tensor([ 1.4948e-07, -1.0990e-06, 1.6764e-08, 8.1258e-07, 1.4668e-07, + -8.2236e-07, -3.1246e-07, 1.0356e-06, 1.3458e-07, -6.4727e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 214.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4062 re_mapping 0.0021 re_causal 0.0078 /// teacc 99.17 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.1521, 0.0568, 0.0963, ..., -0.1026, -0.1940, 0.0752], + [-0.1321, -0.0661, -0.1524, ..., 0.2022, -0.1206, -0.1128], + [-0.0815, 0.0155, 0.2513, ..., -0.1761, -0.1741, -0.2050], + ..., + [-0.2196, -0.0779, 0.0334, ..., -0.0686, 0.1329, -0.1169], + [-0.1475, -0.1021, -0.0933, ..., -0.1806, -0.0188, -0.2270], + [-0.2184, -0.0813, -0.1824, ..., -0.2321, -0.0619, 0.1610]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 2.7940e-09, + 9.3132e-10, -1.1176e-08], + [-4.6566e-10, 0.0000e+00, 5.1223e-09, ..., -6.5193e-09, + 6.9849e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -5.1223e-09, ..., 9.3132e-10, + 6.5193e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -1.1642e-08, ..., 5.1223e-09, + -2.9802e-08, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 6.0536e-09, + 2.3283e-09, 9.3132e-10], + [ 2.7940e-09, 0.0000e+00, 1.3504e-08, ..., 4.6566e-10, + 1.0710e-08, 2.3283e-09]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0370, -0.0302, 0.0084, 0.0299, -0.0184, 0.0066, -0.0170, 0.0160, + 0.0096, -0.0059], device='cuda:0'), grad: tensor([-1.3039e-08, 4.1910e-09, 5.5879e-09, 3.6787e-08, 6.0536e-09, + -1.8626e-09, -5.7276e-08, -3.9116e-08, 3.0268e-08, 4.0513e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 214.31, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4136 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.21 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.1521, 0.0568, 0.0963, ..., -0.1027, -0.1941, 0.0752], + [-0.1322, -0.0661, -0.1525, ..., 0.2022, -0.1206, -0.1128], + [-0.0815, 0.0155, 0.2513, ..., -0.1761, -0.1742, -0.2050], + ..., + [-0.2197, -0.0779, 0.0335, ..., -0.0686, 0.1329, -0.1172], + [-0.1475, -0.1021, -0.0933, ..., -0.1806, -0.0188, -0.2270], + [-0.2184, -0.0813, -0.1825, ..., -0.2321, -0.0620, 0.1610]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -9.3132e-10, ..., 4.6566e-10, + 4.6566e-10, -1.3970e-09], + [ 1.9558e-08, 0.0000e+00, 6.5193e-09, ..., 4.6566e-10, + 1.0710e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -2.3283e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + ..., + [ 4.7963e-07, 0.0000e+00, -1.7695e-08, ..., 0.0000e+00, + 2.8871e-08, 1.3970e-09], + [ 5.1223e-09, 0.0000e+00, 5.1223e-09, ..., 6.0536e-09, + 4.1910e-09, 0.0000e+00], + [ 3.7672e-07, 0.0000e+00, 6.0536e-09, ..., 0.0000e+00, + 5.0291e-08, -0.0000e+00]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0371, -0.0302, 0.0084, 0.0299, -0.0184, 0.0067, -0.0170, 0.0160, + 0.0096, -0.0059], device='cuda:0'), grad: tensor([ 2.7940e-09, 6.1467e-08, 3.7253e-09, 1.7229e-08, -1.6391e-06, + 6.0536e-09, -1.2573e-08, 8.1398e-07, 3.9116e-08, 7.1153e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 214.24, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4523 re_mapping 0.0020 re_causal 0.0082 /// teacc 99.17 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.1522, 0.0568, 0.0964, ..., -0.1027, -0.1941, 0.0753], + [-0.1322, -0.0661, -0.1525, ..., 0.2022, -0.1206, -0.1129], + [-0.0815, 0.0155, 0.2514, ..., -0.1761, -0.1743, -0.2050], + ..., + [-0.2199, -0.0779, 0.0335, ..., -0.0686, 0.1329, -0.1174], + [-0.1475, -0.1021, -0.0933, ..., -0.1807, -0.0188, -0.2271], + [-0.2185, -0.0813, -0.1826, ..., -0.2322, -0.0621, 0.1611]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5367e-08, ..., 9.3132e-10, + 0.0000e+00, -2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -5.5879e-09, + 2.7940e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -5.1223e-09, ..., 4.6566e-09, + -6.0536e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 5.1223e-09, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 1.5367e-08, ..., 0.0000e+00, + 3.2596e-09, 2.1420e-08]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0370, -0.0302, 0.0085, 0.0299, -0.0183, 0.0067, -0.0170, 0.0160, + 0.0096, -0.0059], device='cuda:0'), grad: tensor([-5.7742e-08, -2.3283e-09, 7.4506e-09, 1.7695e-08, 6.9849e-09, + -2.5611e-08, 3.7253e-09, 4.6566e-09, -1.1176e-08, 6.5193e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 214.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4166 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.19 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.1522, 0.0568, 0.0965, ..., -0.1027, -0.1941, 0.0753], + [-0.1323, -0.0661, -0.1525, ..., 0.2022, -0.1206, -0.1129], + [-0.0815, 0.0155, 0.2515, ..., -0.1762, -0.1743, -0.2051], + ..., + [-0.2200, -0.0779, 0.0335, ..., -0.0686, 0.1329, -0.1175], + [-0.1475, -0.1021, -0.0933, ..., -0.1809, -0.0188, -0.2272], + [-0.2186, -0.0813, -0.1829, ..., -0.2322, -0.0622, 0.1611]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3283e-09, ..., 4.6566e-10, + 4.6566e-10, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -6.6590e-08, + 3.1199e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 6.5193e-09, + 3.7253e-09, 4.6566e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -1.0245e-08, ..., 4.7497e-08, + -6.1933e-08, -1.1176e-08], + [ 4.6566e-10, 0.0000e+00, 3.7253e-09, ..., 7.9162e-09, + 2.7940e-09, 5.5879e-09], + [-2.7940e-09, 0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 1.9092e-08, -4.0978e-08]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0370, -0.0302, 0.0086, 0.0299, -0.0183, 0.0067, -0.0170, 0.0160, + 0.0096, -0.0060], device='cuda:0'), grad: tensor([-9.3132e-10, -9.0804e-08, 1.4435e-08, 1.6298e-08, 7.0781e-08, + 2.7940e-08, 4.6566e-09, -3.3528e-08, 4.7497e-08, -4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 214.33, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4118 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.15 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.1523, 0.0568, 0.0966, ..., -0.1028, -0.1942, 0.0754], + [-0.1323, -0.0661, -0.1526, ..., 0.2022, -0.1206, -0.1129], + [-0.0815, 0.0155, 0.2516, ..., -0.1762, -0.1744, -0.2051], + ..., + [-0.2201, -0.0779, 0.0335, ..., -0.0686, 0.1329, -0.1177], + [-0.1475, -0.1021, -0.0933, ..., -0.1811, -0.0189, -0.2272], + [-0.2187, -0.0813, -0.1830, ..., -0.2323, -0.0623, 0.1611]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.1642e-08, ..., -1.2945e-07, + 1.6298e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 7.6368e-08, + 7.9162e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.3749e-08, ..., 4.4238e-08, + -4.5169e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + 7.9162e-09, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 1.3970e-08, -4.6566e-09]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0370, -0.0302, 0.0086, 0.0298, -0.0183, 0.0067, -0.0170, 0.0160, + 0.0096, -0.0061], device='cuda:0'), grad: tensor([ 1.0710e-08, -2.3749e-07, 1.7090e-07, 1.2573e-08, 1.5367e-08, + -5.2154e-08, 3.1665e-08, -9.3132e-10, 3.6322e-08, 3.4925e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 214.40, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4023 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.18 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.1523, 0.0568, 0.0967, ..., -0.1029, -0.1943, 0.0754], + [-0.1323, -0.0661, -0.1526, ..., 0.2022, -0.1206, -0.1130], + [-0.0815, 0.0155, 0.2517, ..., -0.1762, -0.1744, -0.2051], + ..., + [-0.2201, -0.0779, 0.0335, ..., -0.0686, 0.1330, -0.1179], + [-0.1475, -0.1021, -0.0933, ..., -0.1813, -0.0189, -0.2273], + [-0.2188, -0.0813, -0.1831, ..., -0.2323, -0.0624, 0.1611]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.1910e-09, ..., 4.6566e-10, + 0.0000e+00, -7.4506e-09], + [ 4.1910e-09, 0.0000e+00, 1.3039e-08, ..., -2.7940e-09, + 3.2596e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -3.4925e-08, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 4.1910e-09, 0.0000e+00, 2.7940e-09, ..., 1.3970e-09, + -8.8476e-09, 1.3970e-09], + [ 1.3970e-09, 0.0000e+00, 1.3970e-08, ..., 9.3132e-10, + 1.3970e-09, 0.0000e+00], + [ 7.4971e-08, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 3.2596e-09, 3.2596e-09]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0370, -0.0302, 0.0087, 0.0298, -0.0183, 0.0068, -0.0171, 0.0160, + 0.0096, -0.0061], device='cuda:0'), grad: tensor([-1.0710e-08, 3.3528e-08, -6.2399e-08, -3.6135e-07, -1.8440e-07, + 3.2317e-07, 2.7474e-08, 2.0489e-08, 4.3306e-08, 1.7695e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 214.22, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4221 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.15 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.1523, 0.0568, 0.0968, ..., -0.1030, -0.1944, 0.0754], + [-0.1324, -0.0661, -0.1527, ..., 0.2022, -0.1206, -0.1130], + [-0.0815, 0.0155, 0.2517, ..., -0.1763, -0.1745, -0.2051], + ..., + [-0.2202, -0.0779, 0.0335, ..., -0.0686, 0.1330, -0.1181], + [-0.1475, -0.1021, -0.0933, ..., -0.1815, -0.0189, -0.2274], + [-0.2188, -0.0813, -0.1833, ..., -0.2323, -0.0626, 0.1611]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-09, 0.0000e+00, 8.8476e-09, ..., -2.3283e-09, + 1.5367e-08, 5.5879e-09], + [ 9.3132e-10, 0.0000e+00, 2.3283e-09, ..., 4.6566e-10, + 3.2596e-09, 0.0000e+00], + ..., + [ 4.1910e-09, 0.0000e+00, -2.3283e-08, ..., 2.3283e-09, + -3.4459e-08, 8.8476e-09], + [ 2.7940e-09, 0.0000e+00, 3.2596e-09, ..., 9.3132e-09, + 4.6566e-09, 6.5193e-09], + [ 9.2201e-08, 0.0000e+00, 6.9849e-09, ..., 0.0000e+00, + 1.0245e-08, -2.9337e-08]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0370, -0.0302, 0.0087, 0.0298, -0.0183, 0.0068, -0.0171, 0.0160, + 0.0096, -0.0061], device='cuda:0'), grad: tensor([ 2.0489e-08, 6.8452e-08, 1.3970e-08, -3.3993e-08, -2.1420e-07, + 6.0536e-08, -6.9849e-08, -4.0978e-08, 9.7323e-08, 1.0757e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 214.12, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4394 re_mapping 0.0020 re_causal 0.0082 /// teacc 99.16 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.1524, 0.0568, 0.0969, ..., -0.1030, -0.1944, 0.0755], + [-0.1324, -0.0661, -0.1527, ..., 0.2022, -0.1207, -0.1131], + [-0.0816, 0.0155, 0.2518, ..., -0.1763, -0.1746, -0.2052], + ..., + [-0.2202, -0.0779, 0.0335, ..., -0.0686, 0.1330, -0.1183], + [-0.1475, -0.1021, -0.0933, ..., -0.1817, -0.0189, -0.2275], + [-0.2190, -0.0813, -0.1834, ..., -0.2324, -0.0627, 0.1612]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.7695e-08, 0.0000e+00, 9.7789e-09, ..., 2.3283e-09, + 8.3819e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -1.5367e-08, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + ..., + [ 3.0873e-07, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 1.0710e-08, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 2.7940e-09, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 3.7812e-07, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 4.3772e-08, 0.0000e+00]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0369, -0.0302, 0.0087, 0.0298, -0.0183, 0.0068, -0.0171, 0.0160, + 0.0096, -0.0062], device='cuda:0'), grad: tensor([ 6.0536e-09, 1.0291e-07, -2.1886e-08, -3.2596e-09, -1.3085e-06, + 1.5832e-08, 8.8476e-09, 5.4156e-07, -3.0268e-08, 7.0455e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 214.28, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4442 re_mapping 0.0020 re_causal 0.0081 /// teacc 99.18 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.1524, 0.0568, 0.0970, ..., -0.1031, -0.1945, 0.0756], + [-0.1324, -0.0661, -0.1528, ..., 0.2022, -0.1207, -0.1131], + [-0.0816, 0.0155, 0.2518, ..., -0.1763, -0.1746, -0.2052], + ..., + [-0.2204, -0.0779, 0.0336, ..., -0.0686, 0.1330, -0.1184], + [-0.1475, -0.1021, -0.0934, ..., -0.1819, -0.0189, -0.2276], + [-0.2191, -0.0813, -0.1835, ..., -0.2324, -0.0629, 0.1612]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -8.3819e-09, ..., 5.5879e-09, + 0.0000e+00, -1.1176e-08], + [ 9.3132e-10, 0.0000e+00, 1.3970e-09, ..., -2.3283e-09, + 2.7940e-09, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -2.1420e-08, ..., 1.8626e-09, + -3.6787e-08, 3.2596e-09], + [ 4.6566e-10, 0.0000e+00, 9.7789e-09, ..., 4.1910e-09, + 1.4435e-08, 2.7940e-09], + [ 2.3283e-09, 0.0000e+00, 8.8476e-09, ..., 0.0000e+00, + 6.5193e-09, -2.3283e-09]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0369, -0.0302, 0.0087, 0.0298, -0.0182, 0.0068, -0.0171, 0.0160, + 0.0096, -0.0062], device='cuda:0'), grad: tensor([-1.3970e-08, 2.3283e-09, 1.2573e-08, -7.5437e-08, 6.9849e-09, + 4.9360e-08, -3.2596e-08, -2.6077e-08, 5.5879e-08, 2.9337e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 214.45, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4037 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.19 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.1524, 0.0568, 0.0972, ..., -0.1031, -0.1946, 0.0756], + [-0.1325, -0.0661, -0.1528, ..., 0.2022, -0.1207, -0.1132], + [-0.0816, 0.0155, 0.2519, ..., -0.1764, -0.1747, -0.2052], + ..., + [-0.2205, -0.0780, 0.0336, ..., -0.0686, 0.1331, -0.1186], + [-0.1475, -0.1021, -0.0934, ..., -0.1822, -0.0189, -0.2277], + [-0.2192, -0.0813, -0.1838, ..., -0.2325, -0.0631, 0.1612]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -1.3039e-08, ..., 9.3132e-10, + 0.0000e+00, -1.0245e-08], + [ 3.7253e-09, 0.0000e+00, -2.3283e-09, ..., -3.3993e-08, + 0.0000e+00, 2.3283e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-09, 0.0000e+00, 2.7940e-09, ..., 2.8871e-08, + 0.0000e+00, 2.1886e-08], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, 8.3819e-09], + [ 1.2573e-08, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, -2.0070e-07]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0368, -0.0302, 0.0087, 0.0298, -0.0182, 0.0068, -0.0171, 0.0160, + 0.0096, -0.0064], device='cuda:0'), grad: tensor([-9.7789e-09, -6.1002e-08, 1.0245e-08, -2.1560e-07, 3.4133e-07, + 2.3143e-07, -4.1910e-09, 1.4063e-07, 2.6543e-08, -4.5216e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 214.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4448 re_mapping 0.0021 re_causal 0.0082 /// teacc 99.18 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.1525, 0.0568, 0.0974, ..., -0.1032, -0.1947, 0.0757], + [-0.1325, -0.0661, -0.1529, ..., 0.2023, -0.1207, -0.1132], + [-0.0816, 0.0155, 0.2520, ..., -0.1764, -0.1748, -0.2052], + ..., + [-0.2205, -0.0780, 0.0337, ..., -0.0686, 0.1331, -0.1188], + [-0.1475, -0.1021, -0.0934, ..., -0.1824, -0.0189, -0.2278], + [-0.2192, -0.0813, -0.1841, ..., -0.2325, -0.0633, 0.1612]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4342e-07, ..., 1.3970e-09, + 9.3132e-10, -2.2817e-07], + [-4.6566e-10, 0.0000e+00, 2.7940e-09, ..., -2.7940e-09, + 9.3132e-10, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.2596e-09, + 4.6566e-10, 4.1910e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 8.8476e-09, ..., 1.4435e-08, + 4.6566e-10, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 6.0536e-09, ..., -1.3039e-08, + 9.3132e-10, 1.1176e-08], + [ 5.5879e-09, 0.0000e+00, 4.7032e-08, ..., 0.0000e+00, + 4.6566e-10, 6.5658e-08]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0367, -0.0302, 0.0088, 0.0298, -0.0182, 0.0068, -0.0171, 0.0160, + 0.0096, -0.0064], device='cuda:0'), grad: tensor([-7.8278e-07, 2.8592e-07, 8.7544e-08, 1.6252e-07, 9.3132e-10, + 1.8300e-07, 8.5216e-08, 1.0058e-07, -3.5809e-07, 2.4633e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 214.24, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4075 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.17 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.1525, 0.0568, 0.0974, ..., -0.1033, -0.1947, 0.0758], + [-0.1325, -0.0661, -0.1530, ..., 0.2023, -0.1207, -0.1133], + [-0.0816, 0.0155, 0.2521, ..., -0.1765, -0.1749, -0.2053], + ..., + [-0.2205, -0.0780, 0.0337, ..., -0.0686, 0.1331, -0.1190], + [-0.1476, -0.1021, -0.0935, ..., -0.1826, -0.0189, -0.2279], + [-0.2193, -0.0813, -0.1842, ..., -0.2325, -0.0635, 0.1612]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 4.6566e-10, -4.1910e-09], + [ 4.6566e-10, 0.0000e+00, 5.5879e-09, ..., -1.8626e-09, + 6.0536e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 4.6566e-10, + 3.7253e-09, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, -6.0536e-09, ..., 2.7940e-09, + -1.3504e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 6.9849e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 2.6543e-08, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.3283e-09, 2.7940e-09]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0367, -0.0302, 0.0088, 0.0298, -0.0182, 0.0068, -0.0170, 0.0160, + 0.0096, -0.0064], device='cuda:0'), grad: tensor([ 2.3283e-09, 2.4680e-08, -1.4435e-08, -1.9185e-07, -3.4925e-08, + 2.5146e-08, 1.3039e-08, 6.8452e-08, 6.4727e-08, 5.4948e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 214.22, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4494 re_mapping 0.0020 re_causal 0.0082 /// teacc 99.18 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.1525, 0.0568, 0.0975, ..., -0.1034, -0.1948, 0.0759], + [-0.1326, -0.0661, -0.1531, ..., 0.2023, -0.1207, -0.1134], + [-0.0817, 0.0155, 0.2522, ..., -0.1766, -0.1750, -0.2053], + ..., + [-0.2206, -0.0780, 0.0338, ..., -0.0686, 0.1332, -0.1193], + [-0.1476, -0.1021, -0.0935, ..., -0.1829, -0.0189, -0.2280], + [-0.2194, -0.0813, -0.1844, ..., -0.2326, -0.0636, 0.1613]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -4.6566e-10, ..., 9.3132e-10, + 6.9849e-10, -9.3132e-10], + [ 7.6834e-09, 0.0000e+00, 2.3516e-08, ..., -3.2363e-08, + 7.4506e-09, 9.3132e-10], + [ 6.9849e-10, 0.0000e+00, -1.4203e-08, ..., 3.4925e-09, + 1.0710e-08, 0.0000e+00], + ..., + [ 6.7521e-09, 0.0000e+00, -1.5134e-08, ..., 2.5146e-08, + -2.7707e-08, 3.7253e-09], + [ 6.9849e-09, 0.0000e+00, 5.1223e-09, ..., 2.7940e-09, + 6.9849e-09, 1.3970e-09], + [ 6.9849e-10, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, + 2.5611e-09, -6.7521e-09]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0367, -0.0302, 0.0088, 0.0298, -0.0183, 0.0068, -0.0170, 0.0160, + 0.0095, -0.0064], device='cuda:0'), grad: tensor([ 6.5193e-09, -1.9092e-08, 0.0000e+00, 2.2817e-08, -4.1211e-08, + -3.9348e-08, 2.4214e-08, 2.2585e-08, 4.5402e-08, -5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 214.70, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4140 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.22 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.1526, 0.0568, 0.0976, ..., -0.1035, -0.1949, 0.0759], + [-0.1326, -0.0661, -0.1531, ..., 0.2023, -0.1208, -0.1134], + [-0.0817, 0.0155, 0.2522, ..., -0.1766, -0.1751, -0.2053], + ..., + [-0.2206, -0.0780, 0.0338, ..., -0.0686, 0.1332, -0.1196], + [-0.1477, -0.1021, -0.0935, ..., -0.1831, -0.0189, -0.2283], + [-0.2195, -0.0813, -0.1845, ..., -0.2326, -0.0638, 0.1613]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, -1.1642e-09, -1.1176e-08, ..., 2.3283e-10, + 1.6298e-09, -1.1874e-08], + [ 1.1642e-09, 0.0000e+00, 6.9849e-09, ..., -4.6566e-09, + 4.8894e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -2.5611e-09, ..., 4.6566e-10, + 7.4506e-09, 6.9849e-10], + ..., + [-6.9849e-10, 0.0000e+00, -2.4680e-08, ..., 4.6566e-09, + -2.7474e-08, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 1.3970e-09, + 1.1642e-09, 6.9849e-10], + [ 3.9581e-09, 6.9849e-10, 1.6065e-08, ..., 0.0000e+00, + 5.5879e-09, 9.3132e-09]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0367, -0.0302, 0.0088, 0.0298, -0.0183, 0.0069, -0.0170, 0.0160, + 0.0095, -0.0064], device='cuda:0'), grad: tensor([-5.4948e-08, 1.9791e-08, 1.3970e-08, 2.4633e-07, 4.4238e-09, + -2.4377e-07, 1.3271e-08, -5.6578e-08, -3.7253e-09, 8.7311e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 214.35, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4222 re_mapping 0.0020 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.1526, 0.0568, 0.0977, ..., -0.1036, -0.1949, 0.0759], + [-0.1327, -0.0661, -0.1532, ..., 0.2023, -0.1208, -0.1135], + [-0.0818, 0.0155, 0.2523, ..., -0.1767, -0.1752, -0.2053], + ..., + [-0.2208, -0.0780, 0.0339, ..., -0.0687, 0.1333, -0.1197], + [-0.1478, -0.1021, -0.0936, ..., -0.1835, -0.0189, -0.2284], + [-0.2197, -0.0813, -0.1848, ..., -0.2327, -0.0640, 0.1614]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., -1.8626e-08, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.4668e-08, + 0.0000e+00, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 1.3970e-09, + 4.6566e-10, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 2.3283e-10, -2.3050e-08]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0367, -0.0302, 0.0088, 0.0298, -0.0182, 0.0069, -0.0170, 0.0160, + 0.0095, -0.0065], device='cuda:0'), grad: tensor([ 7.2177e-09, -3.6787e-08, 5.1223e-09, -3.2596e-09, 3.6554e-08, + -3.9348e-08, 7.4506e-09, 8.8708e-08, 1.6531e-08, -6.1700e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 214.16, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4284 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.17 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.1527, 0.0568, 0.0979, ..., -0.1037, -0.1950, 0.0760], + [-0.1328, -0.0661, -0.1533, ..., 0.2023, -0.1208, -0.1136], + [-0.0818, 0.0155, 0.2524, ..., -0.1768, -0.1754, -0.2054], + ..., + [-0.2209, -0.0780, 0.0339, ..., -0.0687, 0.1333, -0.1199], + [-0.1478, -0.1021, -0.0936, ..., -0.1838, -0.0189, -0.2286], + [-0.2198, -0.0813, -0.1849, ..., -0.2328, -0.0641, 0.1614]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.8894e-09, ..., 2.3283e-10, + 2.3283e-10, -8.6147e-09], + [ 0.0000e+00, 0.0000e+00, 3.4925e-09, ..., -1.4203e-08, + 5.8208e-09, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 6.9849e-10, + 4.6566e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.0012e-08, ..., 1.0710e-08, + -1.9092e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.0268e-09, ..., 3.9581e-09, + 3.4925e-09, 9.3132e-10], + [ 2.3283e-10, 0.0000e+00, 2.3283e-09, ..., 6.9849e-10, + 5.3551e-09, -2.3283e-10]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0367, -0.0302, 0.0088, 0.0296, -0.0182, 0.0071, -0.0170, 0.0160, + 0.0095, -0.0066], device='cuda:0'), grad: tensor([-2.0256e-08, -1.6997e-08, 1.1176e-08, 1.0710e-08, 5.8208e-09, + 1.2806e-08, 2.5379e-08, 5.5879e-09, -3.8650e-08, 1.5600e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 214.23, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4434 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.21 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.1528, 0.0569, 0.0980, ..., -0.1040, -0.1951, 0.0761], + [-0.1328, -0.0661, -0.1534, ..., 0.2023, -0.1208, -0.1137], + [-0.0818, 0.0155, 0.2525, ..., -0.1768, -0.1755, -0.2054], + ..., + [-0.2210, -0.0780, 0.0340, ..., -0.0687, 0.1333, -0.1202], + [-0.1479, -0.1021, -0.0936, ..., -0.1841, -0.0190, -0.2287], + [-0.2200, -0.0813, -0.1851, ..., -0.2328, -0.0643, 0.1614]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -3.9581e-09, ..., 2.3283e-10, + 1.1642e-09, -8.6147e-09], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., -7.2177e-09, + 7.4506e-09, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 4.6566e-10, + 2.0955e-09, 2.3283e-10], + ..., + [ 2.7940e-09, 0.0000e+00, 4.6566e-10, ..., 6.2864e-09, + -8.3819e-09, 1.1176e-08], + [ 2.3283e-10, 0.0000e+00, 1.3970e-09, ..., 2.5611e-09, + -1.2806e-08, 9.3132e-10], + [ 3.4925e-09, 0.0000e+00, 3.9581e-09, ..., 2.3283e-10, + 4.1910e-09, -8.8476e-09]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0367, -0.0302, 0.0089, 0.0296, -0.0182, 0.0071, -0.0169, 0.0160, + 0.0094, -0.0067], device='cuda:0'), grad: tensor([ 2.3283e-10, 3.3295e-08, 8.8476e-09, 1.0268e-07, 2.1886e-08, + 3.1665e-08, 1.2573e-08, 1.0291e-07, -2.6822e-07, -2.1653e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 214.09, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4311 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.20 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.1528, 0.0569, 0.0981, ..., -0.1042, -0.1951, 0.0761], + [-0.1329, -0.0661, -0.1535, ..., 0.2023, -0.1209, -0.1138], + [-0.0818, 0.0155, 0.2526, ..., -0.1768, -0.1756, -0.2054], + ..., + [-0.2211, -0.0780, 0.0340, ..., -0.0687, 0.1334, -0.1204], + [-0.1479, -0.1021, -0.0936, ..., -0.1844, -0.0190, -0.2289], + [-0.2201, -0.0814, -0.1853, ..., -0.2329, -0.0645, 0.1615]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 1.3970e-09, -1.1642e-09], + [ 4.6566e-10, 0.0000e+00, 2.3050e-08, ..., -3.2131e-08, + 3.6787e-08, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 2.7940e-09, + 1.6298e-09, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, -1.3504e-08, ..., 2.2119e-08, + -3.4459e-08, 1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 2.5611e-09, ..., 1.1642e-09, + 4.6566e-09, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 3.0268e-09, + 4.8894e-09, -3.2596e-09]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0367, -0.0302, 0.0089, 0.0295, -0.0182, 0.0071, -0.0168, 0.0160, + 0.0094, -0.0067], device='cuda:0'), grad: tensor([ 3.7253e-09, 3.3062e-08, -7.9162e-09, -2.7707e-08, 1.4668e-08, + -4.6566e-08, 1.2107e-08, -2.3283e-10, 2.1420e-08, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 214.35, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4306 re_mapping 0.0020 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.1529, 0.0569, 0.0983, ..., -0.1043, -0.1952, 0.0762], + [-0.1329, -0.0661, -0.1536, ..., 0.2024, -0.1209, -0.1139], + [-0.0819, 0.0155, 0.2527, ..., -0.1769, -0.1757, -0.2055], + ..., + [-0.2212, -0.0780, 0.0341, ..., -0.0687, 0.1334, -0.1207], + [-0.1479, -0.1021, -0.0937, ..., -0.1847, -0.0190, -0.2290], + [-0.2202, -0.0814, -0.1856, ..., -0.2330, -0.0646, 0.1615]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 2.3283e-10, 7.2177e-09], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., -5.5879e-09, + 4.6566e-10, 1.6298e-09], + [ 0.0000e+00, 0.0000e+00, -2.2817e-08, ..., 6.9849e-10, + 2.3283e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 5.3551e-09, + 2.3283e-10, 1.8161e-08], + [ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., 3.2596e-09, + 1.3970e-09, 7.6834e-09], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, + 2.3283e-10, -5.2387e-08]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0366, -0.0302, 0.0089, 0.0294, -0.0182, 0.0071, -0.0168, 0.0160, + 0.0094, -0.0068], device='cuda:0'), grad: tensor([ 2.6077e-08, 1.6065e-08, -3.7253e-09, 5.0291e-08, 4.5635e-08, + 4.7963e-08, -3.3528e-08, 1.1316e-07, -1.1409e-07, -1.2922e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 214.44, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4433 re_mapping 0.0020 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.1529, 0.0569, 0.0985, ..., -0.1045, -0.1953, 0.0763], + [-0.1330, -0.0661, -0.1537, ..., 0.2024, -0.1209, -0.1139], + [-0.0819, 0.0155, 0.2528, ..., -0.1770, -0.1758, -0.2056], + ..., + [-0.2213, -0.0780, 0.0342, ..., -0.0687, 0.1335, -0.1209], + [-0.1480, -0.1021, -0.0937, ..., -0.1850, -0.0190, -0.2291], + [-0.2203, -0.0814, -0.1858, ..., -0.2331, -0.0648, 0.1615]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -9.3132e-10, ..., 1.6298e-09, + 2.3283e-10, 2.3283e-10], + [ 1.1642e-09, 0.0000e+00, 1.6531e-08, ..., -2.7940e-09, + 1.4901e-08, 2.5611e-09], + [ 2.3283e-10, 0.0000e+00, 1.3504e-08, ..., 2.3283e-10, + 1.1874e-08, 2.3283e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -5.6112e-08, ..., 1.6298e-09, + -5.1688e-08, 1.6065e-08], + [ 1.1642e-09, 0.0000e+00, 1.0012e-08, ..., 3.7253e-09, + 6.7521e-09, 5.1223e-09], + [ 1.9558e-08, 0.0000e+00, 1.1409e-08, ..., 0.0000e+00, + 1.1176e-08, -1.8859e-08]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0366, -0.0302, 0.0089, 0.0294, -0.0182, 0.0071, -0.0167, 0.0160, + 0.0094, -0.0069], device='cuda:0'), grad: tensor([ 1.0245e-08, 4.6799e-08, 3.7020e-08, 1.7462e-08, -2.0256e-08, + 3.7253e-09, -3.7719e-08, -8.6613e-08, 4.0978e-08, -2.5611e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 214.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4368 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.17 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.1531, 0.0569, 0.0986, ..., -0.1046, -0.1953, 0.0764], + [-0.1331, -0.0661, -0.1538, ..., 0.2024, -0.1209, -0.1140], + [-0.0819, 0.0155, 0.2528, ..., -0.1770, -0.1759, -0.2056], + ..., + [-0.2214, -0.0780, 0.0342, ..., -0.0687, 0.1335, -0.1211], + [-0.1481, -0.1021, -0.0937, ..., -0.1852, -0.0190, -0.2292], + [-0.2204, -0.0814, -0.1860, ..., -0.2331, -0.0650, 0.1616]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -1.8626e-09, ..., 3.4925e-09, + 0.0000e+00, -1.3039e-08], + [ 1.1642e-09, 0.0000e+00, 1.0943e-08, ..., -1.1642e-09, + 4.8894e-09, 9.3132e-10], + [ 2.3283e-10, 0.0000e+00, -9.0804e-09, ..., 1.8626e-09, + 4.6566e-10, 2.3283e-10], + ..., + [ 1.1642e-09, 0.0000e+00, -2.0955e-09, ..., 3.0268e-09, + -5.8208e-09, 2.8871e-08], + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 3.0268e-09, + 4.6566e-10, 6.9849e-10], + [ 7.6834e-09, 0.0000e+00, 2.3283e-09, ..., 4.6566e-10, + 2.3283e-09, -3.4692e-08]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0365, -0.0302, 0.0089, 0.0294, -0.0182, 0.0072, -0.0167, 0.0160, + 0.0094, -0.0069], device='cuda:0'), grad: tensor([-9.7789e-09, 3.1199e-08, -5.3551e-09, 5.1223e-09, 5.3551e-09, + 1.7462e-08, -4.0745e-08, 5.4948e-08, 1.2573e-08, -4.4471e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 214.38, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4017 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.17 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.1532, 0.0569, 0.0987, ..., -0.1047, -0.1954, 0.0764], + [-0.1332, -0.0661, -0.1539, ..., 0.2024, -0.1209, -0.1141], + [-0.0819, 0.0155, 0.2529, ..., -0.1770, -0.1760, -0.2056], + ..., + [-0.2216, -0.0780, 0.0343, ..., -0.0687, 0.1335, -0.1214], + [-0.1481, -0.1021, -0.0938, ..., -0.1853, -0.0190, -0.2293], + [-0.2206, -0.0814, -0.1861, ..., -0.2332, -0.0652, 0.1617]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., 1.3970e-09, + -9.3132e-10, -3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -8.3819e-09, + 4.6566e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 2.3283e-09, + 2.3283e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.4238e-09, ..., 1.6298e-09, + 2.3283e-10, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 3.0268e-09, + 2.3283e-10, 3.0268e-09], + [ 0.0000e+00, 0.0000e+00, 1.2340e-08, ..., 0.0000e+00, + 4.6566e-10, -2.5611e-09]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0365, -0.0302, 0.0090, 0.0292, -0.0182, 0.0072, -0.0166, 0.0160, + 0.0094, -0.0070], device='cuda:0'), grad: tensor([-7.7533e-08, -1.1409e-08, -8.6147e-09, 8.1491e-09, 2.2585e-08, + 1.2806e-08, 2.3982e-08, 3.7486e-08, 1.3271e-08, -1.0245e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 214.19, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4359 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.19 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.1533, 0.0569, 0.0989, ..., -0.1049, -0.1954, 0.0765], + [-0.1334, -0.0661, -0.1540, ..., 0.2024, -0.1210, -0.1142], + [-0.0820, 0.0155, 0.2531, ..., -0.1771, -0.1761, -0.2057], + ..., + [-0.2218, -0.0780, 0.0343, ..., -0.0687, 0.1336, -0.1216], + [-0.1482, -0.1021, -0.0938, ..., -0.1856, -0.0190, -0.2294], + [-0.2208, -0.0814, -0.1863, ..., -0.2333, -0.0654, 0.1617]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 9.3132e-10, + 0.0000e+00, -6.9849e-10], + [ 7.4506e-09, 0.0000e+00, 1.8626e-09, ..., -8.6147e-09, + 3.4925e-09, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 7.6834e-09, ..., 6.9849e-10, + 6.9849e-10, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 1.2573e-08, ..., 6.2864e-09, + -5.1223e-09, 9.0804e-09], + [ 1.8626e-09, 0.0000e+00, -2.5611e-08, ..., 3.2596e-09, + 4.6566e-10, 9.3132e-10], + [ 9.0804e-09, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 2.0955e-09, -1.8394e-08]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0365, -0.0302, 0.0090, 0.0292, -0.0181, 0.0072, -0.0165, 0.0160, + 0.0094, -0.0071], device='cuda:0'), grad: tensor([ 1.6298e-09, -2.3283e-10, 5.7509e-08, 1.5367e-08, -5.3551e-09, + 6.5193e-09, 1.3970e-09, 1.0501e-07, -1.4110e-07, -2.3050e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 214.17, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3927 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.1534, 0.0569, 0.0990, ..., -0.1050, -0.1955, 0.0765], + [-0.1334, -0.0661, -0.1542, ..., 0.2024, -0.1210, -0.1142], + [-0.0820, 0.0155, 0.2532, ..., -0.1772, -0.1763, -0.2057], + ..., + [-0.2219, -0.0780, 0.0344, ..., -0.0687, 0.1337, -0.1219], + [-0.1482, -0.1021, -0.0938, ..., -0.1858, -0.0190, -0.2296], + [-0.2209, -0.0814, -0.1865, ..., -0.2333, -0.0656, 0.1618]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 4.6566e-10, + 3.9581e-09, 3.9581e-09], + [ 0.0000e+00, 0.0000e+00, 1.6298e-08, ..., -1.1642e-09, + 3.4925e-08, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, 1.1874e-08, ..., 2.3283e-10, + 2.4447e-08, 2.3283e-10], + ..., + [ 2.3283e-10, 0.0000e+00, -9.6625e-08, ..., 6.9849e-10, + -1.8859e-07, 9.3132e-09], + [ 2.3283e-10, 0.0000e+00, 6.2864e-09, ..., 2.3283e-09, + 1.4435e-08, 1.8626e-09], + [ 2.5146e-08, 0.0000e+00, 5.0524e-08, ..., 0.0000e+00, + 9.9884e-08, -2.0023e-08]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0365, -0.0302, 0.0090, 0.0292, -0.0181, 0.0072, -0.0165, 0.0160, + 0.0094, -0.0071], device='cuda:0'), grad: tensor([ 2.3749e-08, 1.0314e-07, 7.9628e-08, 7.5437e-08, -8.8476e-09, + -8.3447e-07, 8.0001e-07, -5.4389e-07, 4.9593e-08, 2.6892e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 214.75, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4387 re_mapping 0.0020 re_causal 0.0081 /// teacc 99.19 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.1534, 0.0569, 0.0991, ..., -0.1053, -0.1956, 0.0766], + [-0.1335, -0.0661, -0.1543, ..., 0.2024, -0.1211, -0.1143], + [-0.0820, 0.0155, 0.2533, ..., -0.1772, -0.1765, -0.2057], + ..., + [-0.2220, -0.0780, 0.0345, ..., -0.0688, 0.1337, -0.1221], + [-0.1483, -0.1021, -0.0939, ..., -0.1860, -0.0190, -0.2297], + [-0.2212, -0.0814, -0.1867, ..., -0.2334, -0.0659, 0.1618]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.8626e-08, ..., 4.6566e-10, + 2.3283e-10, -2.8871e-08], + [ 5.5879e-09, 0.0000e+00, 1.1409e-08, ..., -3.0268e-09, + 2.2119e-08, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 5.1223e-09, 2.3283e-10], + ..., + [ 7.4506e-09, 0.0000e+00, -2.0256e-08, ..., 2.3283e-10, + -4.1444e-08, 4.6566e-09], + [ 6.9849e-10, 0.0000e+00, 5.8208e-09, ..., 2.7940e-09, + 6.2864e-09, 4.4238e-09], + [ 1.0105e-07, 0.0000e+00, 8.6147e-09, ..., 9.3132e-10, + 5.5879e-09, 2.0955e-09]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0365, -0.0303, 0.0091, 0.0292, -0.0180, 0.0072, -0.0165, 0.0161, + 0.0093, -0.0073], device='cuda:0'), grad: tensor([-5.8208e-08, 5.7509e-08, 1.7229e-08, 3.9348e-08, -1.9604e-07, + 2.9569e-08, 3.3295e-08, -6.4727e-08, -4.9360e-08, 2.0582e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 214.52, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4228 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.18 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.1535, 0.0569, 0.0993, ..., -0.1055, -0.1956, 0.0767], + [-0.1336, -0.0662, -0.1545, ..., 0.2024, -0.1211, -0.1144], + [-0.0821, 0.0155, 0.2534, ..., -0.1773, -0.1766, -0.2058], + ..., + [-0.2221, -0.0780, 0.0346, ..., -0.0688, 0.1338, -0.1224], + [-0.1483, -0.1021, -0.0940, ..., -0.1863, -0.0191, -0.2299], + [-0.2214, -0.0814, -0.1870, ..., -0.2334, -0.0662, 0.1619]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, -3.0268e-09, ..., 0.0000e+00, + 4.6566e-10, -1.3970e-09], + [ 1.6298e-09, 0.0000e+00, 6.7521e-09, ..., -7.6834e-09, + 9.5461e-09, 2.0955e-09], + [ 2.3283e-10, 0.0000e+00, -9.3132e-09, ..., 4.6566e-10, + 6.0536e-09, 4.6566e-10], + ..., + [ 4.6566e-09, 0.0000e+00, -1.3970e-08, ..., 6.0536e-09, + -2.4680e-08, 3.2363e-08], + [ 2.3283e-10, 0.0000e+00, 8.3819e-09, ..., 1.1642e-09, + 2.3283e-09, 2.7940e-09], + [ 7.4506e-09, 0.0000e+00, 4.8894e-09, ..., 2.3283e-10, + 6.5193e-09, -5.1688e-08]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0365, -0.0303, 0.0091, 0.0290, -0.0179, 0.0074, -0.0164, 0.0161, + 0.0093, -0.0074], device='cuda:0'), grad: tensor([-1.1642e-09, 1.5134e-08, -1.1874e-08, 2.3050e-08, 1.3970e-08, + -1.2340e-08, 1.4203e-08, 5.5414e-08, 2.9104e-08, -1.1711e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 214.25, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4266 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.18 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.1535, 0.0569, 0.0994, ..., -0.1056, -0.1957, 0.0768], + [-0.1337, -0.0662, -0.1546, ..., 0.2024, -0.1211, -0.1144], + [-0.0821, 0.0155, 0.2536, ..., -0.1773, -0.1767, -0.2058], + ..., + [-0.2222, -0.0780, 0.0347, ..., -0.0688, 0.1339, -0.1227], + [-0.1484, -0.1021, -0.0940, ..., -0.1866, -0.0191, -0.2301], + [-0.2216, -0.0814, -0.1871, ..., -0.2335, -0.0664, 0.1620]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -1.4668e-08, ..., -4.6566e-10, + 4.6566e-10, -2.3749e-08], + [ 1.1642e-09, 0.0000e+00, 2.0955e-09, ..., -5.2620e-08, + 2.7940e-09, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 4.6566e-10, + 1.6298e-09, 2.3283e-10], + ..., + [ 6.9849e-10, 0.0000e+00, -8.1491e-09, ..., 4.7265e-08, + -9.7789e-09, 2.1420e-08], + [ 2.3283e-10, 0.0000e+00, 2.0955e-09, ..., 3.9581e-09, + 3.0268e-09, 2.7940e-09], + [ 2.8638e-08, 0.0000e+00, 1.1642e-08, ..., 9.3132e-10, + -2.3283e-09, -1.9791e-08]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0365, -0.0303, 0.0092, 0.0290, -0.0179, 0.0074, -0.0164, 0.0161, + 0.0093, -0.0075], device='cuda:0'), grad: tensor([-4.8429e-08, -1.3318e-07, 7.2177e-09, 1.3504e-08, -6.5193e-09, + 1.7928e-08, 1.3504e-08, 2.0280e-07, 2.3749e-08, -8.1258e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 214.38, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4171 re_mapping 0.0019 re_causal 0.0077 /// teacc 99.18 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.1536, 0.0569, 0.0996, ..., -0.1058, -0.1958, 0.0769], + [-0.1339, -0.0662, -0.1547, ..., 0.2025, -0.1212, -0.1145], + [-0.0822, 0.0155, 0.2537, ..., -0.1774, -0.1769, -0.2058], + ..., + [-0.2223, -0.0780, 0.0348, ..., -0.0688, 0.1339, -0.1230], + [-0.1485, -0.1021, -0.0940, ..., -0.1868, -0.0191, -0.2302], + [-0.2220, -0.0814, -0.1874, ..., -0.2336, -0.0667, 0.1620]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, -3.7253e-09], + [ 6.2864e-09, 0.0000e+00, -9.3132e-10, ..., -2.3912e-07, + -2.4680e-08, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 1.0012e-08, 0.0000e+00, 0.0000e+00, ..., 2.3679e-07, + 2.2585e-08, 2.3283e-10], + [ 4.6566e-10, 0.0000e+00, 2.3283e-10, ..., 1.8626e-09, + 6.9849e-10, 0.0000e+00], + [ 1.3970e-08, 0.0000e+00, 1.1642e-09, ..., 1.1642e-09, + 6.9849e-10, 2.3283e-09]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0365, -0.0303, 0.0092, 0.0289, -0.0177, 0.0074, -0.0163, 0.0161, + 0.0092, -0.0076], device='cuda:0'), grad: tensor([-3.0268e-09, -5.3365e-07, 8.6147e-09, 3.2131e-08, -4.3772e-08, + 1.4226e-07, -9.1735e-08, 5.6811e-07, -1.0151e-07, 3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 214.33, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4075 re_mapping 0.0019 re_causal 0.0077 /// teacc 99.16 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.1536, 0.0569, 0.0999, ..., -0.1058, -0.1958, 0.0770], + [-0.1340, -0.0662, -0.1548, ..., 0.2025, -0.1212, -0.1146], + [-0.0822, 0.0155, 0.2539, ..., -0.1775, -0.1770, -0.2059], + ..., + [-0.2225, -0.0780, 0.0348, ..., -0.0688, 0.1340, -0.1232], + [-0.1485, -0.1021, -0.0941, ..., -0.1870, -0.0191, -0.2303], + [-0.2222, -0.0814, -0.1877, ..., -0.2337, -0.0668, 0.1621]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7229e-08, ..., 1.6298e-09, + 6.9849e-10, -2.4447e-08], + [ 0.0000e+00, 0.0000e+00, 1.6065e-08, ..., -1.8859e-08, + 3.0966e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 2.3283e-09, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -2.2817e-08, ..., 1.5600e-08, + -5.5879e-08, 6.7521e-09], + [ 0.0000e+00, 0.0000e+00, 1.6298e-09, ..., 3.2596e-09, + 2.7940e-09, 8.1491e-09], + [ 4.6566e-10, 0.0000e+00, 1.4668e-08, ..., 4.6566e-10, + 1.4203e-08, -2.3283e-09]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0364, -0.0303, 0.0093, 0.0290, -0.0176, 0.0074, -0.0163, 0.0161, + 0.0092, -0.0078], device='cuda:0'), grad: tensor([-4.7497e-08, 5.8440e-08, 8.3819e-09, 2.3283e-08, 2.9802e-08, + -8.7544e-08, 6.3097e-08, -9.3598e-08, 5.0291e-08, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 214.43, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4120 re_mapping 0.0020 re_causal 0.0076 /// teacc 99.17 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.1537, 0.0569, 0.1001, ..., -0.1059, -0.1959, 0.0772], + [-0.1340, -0.0662, -0.1550, ..., 0.2025, -0.1212, -0.1147], + [-0.0822, 0.0155, 0.2541, ..., -0.1776, -0.1772, -0.2059], + ..., + [-0.2226, -0.0780, 0.0348, ..., -0.0688, 0.1340, -0.1235], + [-0.1485, -0.1021, -0.0942, ..., -0.1872, -0.0191, -0.2305], + [-0.2224, -0.0814, -0.1881, ..., -0.2338, -0.0671, 0.1621]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -2.5611e-09, 0.0000e+00, ..., 4.6566e-10, + 3.9581e-09, -4.1910e-09], + [ 8.3819e-09, 1.3970e-09, 0.0000e+00, ..., -2.4424e-07, + 6.9849e-10, 6.0536e-09], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 1.8626e-09, 0.0000e+00, 2.3283e-10, ..., 2.2491e-07, + 2.3283e-09, 3.9581e-09], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.5611e-09, + 1.1642e-09, 1.1642e-09], + [ 6.1002e-08, 6.9849e-10, 0.0000e+00, ..., 6.2864e-09, + 1.6298e-09, -1.6764e-08]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0362, -0.0303, 0.0095, 0.0289, -0.0176, 0.0074, -0.0164, 0.0161, + 0.0092, -0.0080], device='cuda:0'), grad: tensor([ 3.4925e-09, -9.6392e-07, 2.1653e-08, 5.7044e-08, -9.5228e-08, + -1.1339e-07, 6.8452e-08, 9.6858e-07, -4.2608e-08, 9.1968e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 214.21, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4119 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.16 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.1538, 0.0569, 0.1006, ..., -0.1060, -0.1959, 0.0775], + [-0.1341, -0.0662, -0.1551, ..., 0.2025, -0.1213, -0.1148], + [-0.0823, 0.0155, 0.2542, ..., -0.1776, -0.1773, -0.2060], + ..., + [-0.2228, -0.0780, 0.0349, ..., -0.0688, 0.1341, -0.1238], + [-0.1486, -0.1021, -0.0942, ..., -0.1875, -0.0191, -0.2306], + [-0.2226, -0.0814, -0.1884, ..., -0.2339, -0.0673, 0.1620]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 9.7789e-09, ..., -1.9558e-08, + 1.0245e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.5367e-08, ..., 4.6566e-10, + 1.4435e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -3.2131e-08, ..., 3.7253e-09, + -3.1199e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 3.7253e-09, + 4.1910e-09, 0.0000e+00], + [ 2.3283e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 4.1910e-09, -5.5879e-09]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0359, -0.0303, 0.0095, 0.0289, -0.0175, 0.0075, -0.0164, 0.0161, + 0.0092, -0.0082], device='cuda:0'), grad: tensor([ 5.5879e-09, 6.5658e-08, 5.3551e-08, 6.9849e-09, 5.5879e-09, + 3.7253e-09, 3.1199e-08, -8.3819e-08, -6.8452e-08, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 214.33, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4424 re_mapping 0.0019 re_causal 0.0081 /// teacc 99.20 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.1538, 0.0569, 0.1008, ..., -0.1061, -0.1961, 0.0776], + [-0.1342, -0.0662, -0.1553, ..., 0.2025, -0.1213, -0.1150], + [-0.0823, 0.0155, 0.2543, ..., -0.1777, -0.1774, -0.2060], + ..., + [-0.2230, -0.0780, 0.0349, ..., -0.0688, 0.1342, -0.1240], + [-0.1486, -0.1021, -0.0942, ..., -0.1876, -0.0191, -0.2309], + [-0.2229, -0.0814, -0.1887, ..., -0.2339, -0.0676, 0.1621]], + device='cuda:0'), grad: tensor([[-1.8626e-09, -4.6566e-09, -1.3039e-08, ..., 4.6566e-10, + 9.3132e-10, -1.8161e-08], + [ 1.8626e-09, 0.0000e+00, 1.4435e-08, ..., 0.0000e+00, + 2.6543e-08, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 5.1223e-09, 0.0000e+00], + ..., + [-6.9849e-09, 0.0000e+00, -1.3411e-07, ..., 9.3132e-10, + -2.2445e-07, 4.6566e-10], + [ 1.3970e-09, 9.3132e-10, 9.3132e-09, ..., 1.8626e-09, + 1.2573e-08, 2.7940e-09], + [-8.8476e-09, 1.3970e-09, 1.0477e-07, ..., 0.0000e+00, + 1.5972e-07, -1.6764e-08]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0359, -0.0303, 0.0095, 0.0288, -0.0174, 0.0076, -0.0164, 0.0161, + 0.0092, -0.0083], device='cuda:0'), grad: tensor([-5.0757e-08, 8.3819e-08, 1.8161e-08, 1.9558e-08, 1.2061e-07, + 1.6764e-08, 1.8626e-09, -5.4296e-07, 4.7497e-08, 2.9011e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 214.52, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4079 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.19 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.1538, 0.0569, 0.1011, ..., -0.1063, -0.1962, 0.0778], + [-0.1343, -0.0662, -0.1555, ..., 0.2026, -0.1214, -0.1150], + [-0.0823, 0.0155, 0.2545, ..., -0.1778, -0.1776, -0.2060], + ..., + [-0.2231, -0.0780, 0.0351, ..., -0.0688, 0.1343, -0.1243], + [-0.1487, -0.1021, -0.0943, ..., -0.1878, -0.0192, -0.2310], + [-0.2231, -0.0814, -0.1890, ..., -0.2340, -0.0678, 0.1622]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -2.3283e-09, ..., 4.6566e-10, + 9.3132e-10, -2.3283e-09], + [ 1.3970e-09, 0.0000e+00, 2.1886e-08, ..., -6.0536e-09, + 3.5856e-08, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 9.3132e-09, 0.0000e+00], + ..., + [ 4.6566e-09, 0.0000e+00, -5.1223e-08, ..., 4.1910e-09, + -8.3353e-08, 5.1223e-09], + [ 9.3132e-10, 0.0000e+00, 3.2596e-09, ..., 1.8626e-09, + 5.5879e-09, 3.7253e-09], + [ 6.5193e-09, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 2.2817e-08, -4.1910e-08]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0358, -0.0303, 0.0095, 0.0287, -0.0174, 0.0078, -0.0165, 0.0161, + 0.0092, -0.0084], device='cuda:0'), grad: tensor([ 9.3132e-10, 1.3364e-07, 3.8184e-08, 6.3330e-08, -3.7253e-09, + 1.5832e-08, 4.6566e-09, -2.8498e-07, 2.2352e-08, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 214.61, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4204 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.18 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.1539, 0.0569, 0.1013, ..., -0.1064, -0.1963, 0.0780], + [-0.1344, -0.0662, -0.1557, ..., 0.2026, -0.1214, -0.1151], + [-0.0824, 0.0155, 0.2545, ..., -0.1778, -0.1778, -0.2061], + ..., + [-0.2233, -0.0780, 0.0352, ..., -0.0689, 0.1343, -0.1245], + [-0.1487, -0.1021, -0.0943, ..., -0.1880, -0.0192, -0.2312], + [-0.2234, -0.0814, -0.1893, ..., -0.2340, -0.0680, 0.1622]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, 9.3132e-10, ..., 6.9849e-09, + 1.8626e-09, 1.8626e-09], + [ 6.2864e-08, 0.0000e+00, 5.3551e-08, ..., -2.9337e-08, + 1.0617e-07, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 1.3970e-08, ..., 2.7940e-09, + 2.2817e-08, 0.0000e+00], + ..., + [ 5.5879e-08, 0.0000e+00, -1.4342e-07, ..., 1.5367e-08, + -2.5425e-07, 0.0000e+00], + [ 3.6322e-08, 0.0000e+00, 6.1933e-08, ..., 2.0955e-08, + 1.0105e-07, 1.3970e-09], + [ 1.1502e-07, 0.0000e+00, 7.4506e-09, ..., 4.6566e-10, + 1.2573e-08, 0.0000e+00]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0357, -0.0303, 0.0094, 0.0287, -0.0173, 0.0079, -0.0165, 0.0161, + 0.0091, -0.0086], device='cuda:0'), grad: tensor([ 3.4459e-08, 3.4412e-07, 9.5461e-08, 5.5879e-08, -4.4517e-07, + -4.0513e-08, -5.7742e-08, -7.2457e-07, 4.9081e-07, 2.5565e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 214.59, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4155 re_mapping 0.0019 re_causal 0.0077 /// teacc 99.18 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.1539, 0.0569, 0.1015, ..., -0.1065, -0.1964, 0.0781], + [-0.1344, -0.0662, -0.1559, ..., 0.2026, -0.1215, -0.1152], + [-0.0824, 0.0155, 0.2546, ..., -0.1780, -0.1780, -0.2061], + ..., + [-0.2233, -0.0780, 0.0354, ..., -0.0689, 0.1344, -0.1247], + [-0.1488, -0.1021, -0.0944, ..., -0.1883, -0.0192, -0.2314], + [-0.2236, -0.0814, -0.1896, ..., -0.2341, -0.0683, 0.1623]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 5.5879e-09, ..., -7.5437e-08, + 1.0245e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.2596e-09, ..., 3.1199e-08, + 4.6566e-10, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, -4.6566e-09, ..., 2.7008e-08, + -2.2352e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., 6.0536e-09, + -3.2596e-09, 4.6566e-10], + [ 2.3283e-09, 0.0000e+00, 5.1223e-09, ..., 1.3970e-09, + 6.9849e-09, -2.3283e-09]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0356, -0.0303, 0.0094, 0.0287, -0.0173, 0.0079, -0.0165, 0.0161, + 0.0091, -0.0087], device='cuda:0'), grad: tensor([ 8.8476e-09, -2.0443e-07, 7.0781e-08, 2.1886e-08, 2.0489e-08, + 3.2596e-09, -1.3970e-09, 7.3574e-08, 6.0536e-09, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 214.96, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4219 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.20 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.1541, 0.0569, 0.1017, ..., -0.1067, -0.1965, 0.0782], + [-0.1346, -0.0662, -0.1561, ..., 0.2027, -0.1215, -0.1153], + [-0.0825, 0.0155, 0.2547, ..., -0.1781, -0.1781, -0.2061], + ..., + [-0.2236, -0.0780, 0.0355, ..., -0.0689, 0.1345, -0.1250], + [-0.1489, -0.1021, -0.0944, ..., -0.1885, -0.0192, -0.2315], + [-0.2239, -0.0814, -0.1898, ..., -0.2342, -0.0685, 0.1624]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -2.7940e-09, ..., 4.6566e-09, + 2.3283e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -1.2107e-08, + 3.2596e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3970e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 7.4506e-09, + -4.1910e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.8405e-08, + 2.7940e-09, 5.1223e-09], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 4.6566e-10, + 1.3970e-09, -4.6566e-09]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0356, -0.0303, 0.0093, 0.0287, -0.0171, 0.0078, -0.0164, 0.0161, + 0.0091, -0.0089], device='cuda:0'), grad: tensor([ 1.6764e-08, -1.7229e-08, 7.4506e-09, 1.6298e-08, 2.7940e-09, + 2.9569e-07, -4.4843e-07, 2.1420e-08, 1.1083e-07, -4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 214.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4218 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.18 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.1542, 0.0569, 0.1018, ..., -0.1068, -0.1965, 0.0783], + [-0.1348, -0.0662, -0.1562, ..., 0.2027, -0.1216, -0.1154], + [-0.0825, 0.0155, 0.2548, ..., -0.1782, -0.1782, -0.2062], + ..., + [-0.2237, -0.0780, 0.0356, ..., -0.0689, 0.1346, -0.1252], + [-0.1490, -0.1021, -0.0945, ..., -0.1888, -0.0192, -0.2317], + [-0.2241, -0.0814, -0.1900, ..., -0.2343, -0.0688, 0.1625]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.8429e-08, ..., -2.3283e-09, + 4.6566e-10, -1.5367e-07], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -5.5879e-09, + 4.6566e-10, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -1.0291e-07, ..., 4.6566e-10, + -3.2596e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 8.8010e-08, ..., 5.5879e-09, + 3.7253e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.2596e-09, + 9.3132e-10, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 3.3993e-08, ..., 1.8626e-09, + -4.6566e-10, 9.3132e-08]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0356, -0.0303, 0.0093, 0.0286, -0.0171, 0.0079, -0.0164, 0.0161, + 0.0090, -0.0090], device='cuda:0'), grad: tensor([-3.9209e-07, 9.3132e-09, -2.1840e-07, 2.7940e-08, 1.1176e-08, + 1.2107e-08, 5.4017e-08, 2.1094e-07, 4.3772e-08, 2.4773e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 214.45, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4111 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.1543, 0.0569, 0.1019, ..., -0.1070, -0.1966, 0.0784], + [-0.1349, -0.0662, -0.1564, ..., 0.2027, -0.1216, -0.1154], + [-0.0826, 0.0155, 0.2549, ..., -0.1782, -0.1784, -0.2062], + ..., + [-0.2238, -0.0780, 0.0357, ..., -0.0690, 0.1347, -0.1255], + [-0.1490, -0.1021, -0.0946, ..., -0.1890, -0.0193, -0.2318], + [-0.2243, -0.0814, -0.1903, ..., -0.2344, -0.0691, 0.1626]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0245e-08, ..., 4.6566e-10, + 4.6566e-10, -9.3132e-09], + [ 4.6566e-10, 0.0000e+00, 1.1176e-08, ..., -2.3749e-08, + 9.7789e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 4.6566e-09, + 9.7789e-09, 9.3132e-10], + ..., + [ 4.1910e-09, 0.0000e+00, -3.3062e-08, ..., 1.4901e-08, + -3.0734e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., 4.1910e-09, + 5.5879e-09, 1.3970e-09], + [ 9.3132e-10, 0.0000e+00, 8.8476e-09, ..., 0.0000e+00, + 4.6566e-09, 5.5879e-09]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0356, -0.0303, 0.0093, 0.0286, -0.0171, 0.0079, -0.0163, 0.0161, + 0.0090, -0.0091], device='cuda:0'), grad: tensor([-2.7008e-08, -3.1665e-08, 3.9116e-08, 5.6811e-08, -9.3132e-10, + -5.6345e-08, 1.1642e-08, -3.6322e-08, 2.7940e-08, 2.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 214.55, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4249 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.1543, 0.0569, 0.1021, ..., -0.1072, -0.1967, 0.0785], + [-0.1350, -0.0662, -0.1565, ..., 0.2027, -0.1217, -0.1156], + [-0.0826, 0.0155, 0.2549, ..., -0.1783, -0.1786, -0.2063], + ..., + [-0.2239, -0.0780, 0.0359, ..., -0.0690, 0.1348, -0.1258], + [-0.1490, -0.1021, -0.0946, ..., -0.1894, -0.0193, -0.2320], + [-0.2245, -0.0814, -0.1905, ..., -0.2344, -0.0693, 0.1628]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 0.0000e+00, -4.1910e-09, ..., 4.6566e-09, + 0.0000e+00, -9.3132e-10], + [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., -2.0862e-07, + 2.3283e-09, 4.6566e-10], + [ 3.7253e-09, 0.0000e+00, 4.6566e-10, ..., 5.1223e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, -1.3970e-09, ..., 1.6764e-07, + -2.7940e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.2573e-08, + 9.3132e-10, 1.3970e-09], + [ 4.6566e-09, 0.0000e+00, 3.7253e-09, ..., 8.3819e-09, + 1.3970e-09, -3.2596e-09]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0355, -0.0303, 0.0092, 0.0286, -0.0171, 0.0079, -0.0162, 0.0161, + 0.0090, -0.0092], device='cuda:0'), grad: tensor([ 3.7719e-08, -8.6473e-07, 2.0023e-08, -7.9162e-09, 2.9802e-08, + 3.4459e-08, -3.8184e-08, 7.1526e-07, 5.0757e-08, 3.0734e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 214.42, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4155 re_mapping 0.0020 re_causal 0.0076 /// teacc 99.17 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.1544, 0.0569, 0.1023, ..., -0.1073, -0.1968, 0.0786], + [-0.1351, -0.0662, -0.1567, ..., 0.2028, -0.1218, -0.1157], + [-0.0826, 0.0155, 0.2550, ..., -0.1784, -0.1787, -0.2064], + ..., + [-0.2240, -0.0780, 0.0360, ..., -0.0690, 0.1349, -0.1260], + [-0.1491, -0.1021, -0.0947, ..., -0.1896, -0.0193, -0.2322], + [-0.2248, -0.0814, -0.1906, ..., -0.2345, -0.0695, 0.1629]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -1.0710e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 6.0536e-09], + [ 4.6566e-10, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0355, -0.0303, 0.0091, 0.0285, -0.0171, 0.0080, -0.0162, 0.0161, + 0.0090, -0.0092], device='cuda:0'), grad: tensor([ 3.7253e-09, 6.5193e-09, -1.7695e-08, -1.8626e-08, -1.8626e-09, + 6.5193e-09, 3.2596e-09, 3.6787e-08, 9.3132e-09, -2.3749e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 214.52, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4319 re_mapping 0.0019 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.1545, 0.0569, 0.1023, ..., -0.1075, -0.1968, 0.0786], + [-0.1352, -0.0662, -0.1569, ..., 0.2028, -0.1218, -0.1158], + [-0.0827, 0.0155, 0.2553, ..., -0.1786, -0.1789, -0.2064], + ..., + [-0.2242, -0.0780, 0.0361, ..., -0.0691, 0.1350, -0.1264], + [-0.1491, -0.1021, -0.0948, ..., -0.1898, -0.0193, -0.2324], + [-0.2250, -0.0814, -0.1908, ..., -0.2345, -0.0696, 0.1632]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.7229e-08, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 0.0000e+00, 1.8626e-09, ..., -2.7008e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.6450e-07, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 1.9092e-08, + -4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3190e-07, ..., 2.3283e-09, + 1.3970e-09, 4.6566e-10], + [ 1.3039e-08, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 1.3970e-09, 4.6566e-10]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0356, -0.0303, 0.0093, 0.0284, -0.0171, 0.0080, -0.0161, 0.0161, + 0.0089, -0.0092], device='cuda:0'), grad: tensor([ 4.0513e-08, -9.8720e-08, -4.3120e-07, 2.1420e-08, 4.1910e-09, + 1.3504e-08, 3.9581e-08, 1.1455e-07, 2.3376e-07, 7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 214.38, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4171 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.23 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.1545, 0.0569, 0.1023, ..., -0.1077, -0.1970, 0.0787], + [-0.1353, -0.0662, -0.1572, ..., 0.2029, -0.1219, -0.1159], + [-0.0827, 0.0155, 0.2554, ..., -0.1788, -0.1790, -0.2064], + ..., + [-0.2242, -0.0780, 0.0363, ..., -0.0691, 0.1352, -0.1266], + [-0.1492, -0.1021, -0.0949, ..., -0.1902, -0.0193, -0.2325], + [-0.2251, -0.0814, -0.1912, ..., -0.2346, -0.0701, 0.1633]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -2.0489e-08, + 1.3504e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 8.8476e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -4.5635e-08, ..., 2.0023e-08, + -6.6590e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 4.6566e-10, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3504e-08, ..., 0.0000e+00, + 1.8626e-08, -3.2596e-09]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0357, -0.0303, 0.0093, 0.0284, -0.0171, 0.0080, -0.0160, 0.0161, + 0.0089, -0.0094], device='cuda:0'), grad: tensor([ 9.3132e-10, -6.0536e-09, 2.3283e-08, 1.0245e-08, 2.0023e-08, + 2.7940e-09, 1.8626e-09, -1.2992e-07, 3.8184e-08, 4.0047e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 214.53, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4412 re_mapping 0.0019 re_causal 0.0079 /// teacc 99.20 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.1546, 0.0569, 0.1024, ..., -0.1079, -0.1970, 0.0787], + [-0.1353, -0.0662, -0.1575, ..., 0.2029, -0.1220, -0.1160], + [-0.0827, 0.0155, 0.2555, ..., -0.1789, -0.1792, -0.2065], + ..., + [-0.2243, -0.0780, 0.0365, ..., -0.0691, 0.1353, -0.1268], + [-0.1493, -0.1021, -0.0949, ..., -0.1906, -0.0194, -0.2327], + [-0.2253, -0.0814, -0.1914, ..., -0.2347, -0.0703, 0.1635]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., 0.0000e+00, + 4.6566e-10, -4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.3970e-09, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, -9.3132e-10, ..., 1.3970e-09, + -3.7253e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 1.3970e-09, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, -0.0000e+00]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0358, -0.0303, 0.0092, 0.0284, -0.0172, 0.0080, -0.0158, 0.0162, + 0.0088, -0.0094], device='cuda:0'), grad: tensor([-8.8476e-09, 1.3970e-09, 1.8626e-09, 6.0536e-09, 3.7253e-09, + -9.7789e-09, 3.2596e-09, 2.7940e-09, 2.3283e-09, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 214.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4086 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.1547, 0.0569, 0.1026, ..., -0.1081, -0.1970, 0.0788], + [-0.1353, -0.0662, -0.1578, ..., 0.2030, -0.1221, -0.1160], + [-0.0828, 0.0155, 0.2556, ..., -0.1792, -0.1794, -0.2065], + ..., + [-0.2244, -0.0780, 0.0367, ..., -0.0692, 0.1355, -0.1270], + [-0.1493, -0.1021, -0.0950, ..., -0.1909, -0.0194, -0.2329], + [-0.2255, -0.0814, -0.1916, ..., -0.2347, -0.0705, 0.1637]], + device='cuda:0'), grad: tensor([[-0.0000e+00, -4.6566e-10, -1.0245e-08, ..., 1.8626e-09, + 0.0000e+00, -9.7789e-09], + [-3.2596e-09, 0.0000e+00, 4.6566e-10, ..., -6.0536e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 1.8626e-09, 0.0000e+00, 4.6566e-10, ..., 4.1910e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 3.7253e-09, + 4.6566e-10, 3.2596e-09], + [ 4.6566e-10, 4.6566e-10, 5.1223e-09, ..., 4.6566e-10, + 4.6566e-10, 5.5879e-09]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0359, -0.0303, 0.0091, 0.0284, -0.0173, 0.0079, -0.0157, 0.0162, + 0.0087, -0.0094], device='cuda:0'), grad: tensor([-2.0023e-08, -1.8161e-08, 3.2596e-09, 4.0978e-08, 8.8476e-09, + 5.6811e-08, -8.0094e-08, 2.3749e-08, -3.4459e-08, 2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 215.17, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4234 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.1547, 0.0569, 0.1028, ..., -0.1082, -0.1971, 0.0789], + [-0.1355, -0.0662, -0.1580, ..., 0.2031, -0.1222, -0.1161], + [-0.0828, 0.0155, 0.2558, ..., -0.1793, -0.1796, -0.2065], + ..., + [-0.2245, -0.0780, 0.0369, ..., -0.0693, 0.1357, -0.1273], + [-0.1493, -0.1021, -0.0952, ..., -0.1912, -0.0195, -0.2330], + [-0.2258, -0.0814, -0.1919, ..., -0.2348, -0.0708, 0.1639]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 1.7229e-08, ..., -4.0513e-08, + 2.5146e-08, -2.7940e-09], + [-0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 1.1176e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, -7.9162e-08, ..., 2.5146e-08, + -1.2340e-07, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 1.6298e-08, ..., 6.0536e-09, + 2.3283e-08, 4.6566e-10], + [ 1.1642e-08, 0.0000e+00, 2.9802e-08, ..., 9.3132e-10, + 4.9826e-08, 4.6566e-10]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0359, -0.0303, 0.0091, 0.0284, -0.0174, 0.0079, -0.0156, 0.0162, + 0.0087, -0.0095], device='cuda:0'), grad: tensor([ 4.9360e-08, -3.1665e-08, 4.7497e-08, -3.4552e-07, 2.3749e-08, + 1.7695e-07, 1.9092e-08, -2.1700e-07, 1.0245e-07, 1.8440e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 214.53, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4060 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.1548, 0.0569, 0.1030, ..., -0.1084, -0.1972, 0.0790], + [-0.1356, -0.0662, -0.1582, ..., 0.2032, -0.1223, -0.1161], + [-0.0829, 0.0155, 0.2559, ..., -0.1794, -0.1798, -0.2065], + ..., + [-0.2246, -0.0780, 0.0371, ..., -0.0693, 0.1358, -0.1275], + [-0.1493, -0.1021, -0.0952, ..., -0.1916, -0.0195, -0.2332], + [-0.2261, -0.0814, -0.1922, ..., -0.2349, -0.0711, 0.1640]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., 9.7789e-09, + 9.3132e-10, -1.8626e-09], + [ 1.0710e-08, 0.0000e+00, 1.4435e-08, ..., 6.0536e-09, + 1.8161e-08, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 2.0489e-08, ..., 7.4506e-09, + 2.3283e-08, 0.0000e+00], + ..., + [ 6.0536e-09, 0.0000e+00, -7.5437e-08, ..., 1.8626e-09, + -9.9652e-08, 1.3970e-09], + [ 9.3132e-10, 0.0000e+00, 9.7789e-09, ..., 9.3132e-09, + 1.0710e-08, 4.6566e-10], + [ 4.6100e-08, 0.0000e+00, 2.1420e-08, ..., 4.6566e-10, + 3.3993e-08, 1.3970e-09]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0359, -0.0303, 0.0091, 0.0284, -0.0173, 0.0078, -0.0155, 0.0162, + 0.0086, -0.0098], device='cuda:0'), grad: tensor([ 3.9581e-08, 9.9652e-08, 1.0524e-07, 3.6787e-08, -8.1956e-08, + 1.3970e-09, -1.3784e-07, -2.6450e-07, 5.0757e-08, 1.6298e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 214.58, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4089 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.15 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.1548, 0.0569, 0.1032, ..., -0.1086, -0.1973, 0.0792], + [-0.1357, -0.0662, -0.1584, ..., 0.2033, -0.1224, -0.1162], + [-0.0829, 0.0155, 0.2560, ..., -0.1795, -0.1800, -0.2066], + ..., + [-0.2248, -0.0780, 0.0373, ..., -0.0694, 0.1359, -0.1277], + [-0.1493, -0.1021, -0.0953, ..., -0.1918, -0.0195, -0.2334], + [-0.2263, -0.0814, -0.1925, ..., -0.2350, -0.0713, 0.1641]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 4.6566e-10, + 4.6566e-10, -1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., -1.3970e-09, + 4.1910e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, -2.3283e-09, ..., 2.3283e-09, + -4.6566e-09, 1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3970e-09, + 2.3283e-09, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 2.7940e-09, -1.1176e-08]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0358, -0.0303, 0.0090, 0.0286, -0.0173, 0.0077, -0.0155, 0.0162, + 0.0086, -0.0099], device='cuda:0'), grad: tensor([ 2.3283e-09, 1.1642e-08, 5.1223e-09, -1.6298e-08, 6.9849e-09, + -4.0978e-08, 1.2573e-08, 3.6322e-08, 2.2352e-08, -2.0955e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 214.57, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4205 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.1549, 0.0569, 0.1033, ..., -0.1087, -0.1973, 0.0793], + [-0.1357, -0.0662, -0.1587, ..., 0.2033, -0.1225, -0.1163], + [-0.0829, 0.0155, 0.2562, ..., -0.1796, -0.1801, -0.2066], + ..., + [-0.2249, -0.0781, 0.0374, ..., -0.0694, 0.1360, -0.1279], + [-0.1493, -0.1021, -0.0954, ..., -0.1920, -0.0195, -0.2335], + [-0.2265, -0.0814, -0.1927, ..., -0.2350, -0.0714, 0.1643]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 4.6566e-10, + 5.5879e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.5390e-08, ..., -1.3970e-09, + 6.7521e-08, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 1.7229e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.0477e-07, ..., 1.3970e-09, + -2.0768e-07, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., 3.2596e-09, + 1.3970e-08, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 4.5169e-08, ..., 0.0000e+00, + 8.7079e-08, 2.7940e-09]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0359, -0.0303, 0.0091, 0.0286, -0.0173, 0.0078, -0.0155, 0.0162, + 0.0086, -0.0099], device='cuda:0'), grad: tensor([ 2.2352e-08, 2.2585e-07, 5.4482e-08, 4.0978e-08, 1.7695e-08, + 2.2817e-08, -1.8161e-08, -6.8126e-07, 5.4482e-08, 2.8964e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 215.01, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4121 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.18 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.1549, 0.0569, 0.1035, ..., -0.1088, -0.1974, 0.0794], + [-0.1358, -0.0662, -0.1590, ..., 0.2034, -0.1226, -0.1164], + [-0.0830, 0.0155, 0.2561, ..., -0.1797, -0.1806, -0.2066], + ..., + [-0.2250, -0.0781, 0.0378, ..., -0.0694, 0.1363, -0.1282], + [-0.1494, -0.1022, -0.0955, ..., -0.1922, -0.0196, -0.2337], + [-0.2266, -0.0814, -0.1930, ..., -0.2350, -0.0717, 0.1644]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.1665e-08, ..., -1.3970e-08, + 4.6566e-10, -9.3132e-08], + [ 4.6566e-10, 0.0000e+00, 2.7940e-08, ..., 9.7789e-09, + 1.1642e-08, 5.8208e-08], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 6.0536e-09, 4.6566e-10], + ..., + [ 9.3132e-10, 0.0000e+00, -2.1886e-08, ..., 9.3132e-10, + -3.4925e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.3970e-09, + 4.6566e-09, 1.8626e-09], + [ 2.3283e-09, 0.0000e+00, 9.3132e-09, ..., 1.8626e-09, + 6.0536e-09, 1.1176e-08]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0358, -0.0303, 0.0088, 0.0286, -0.0174, 0.0077, -0.0155, 0.0162, + 0.0085, -0.0100], device='cuda:0'), grad: tensor([-3.0361e-07, 2.4354e-07, 9.3132e-10, 1.3039e-08, 1.5367e-08, + 9.7789e-09, 5.4482e-08, -9.9186e-08, 2.7474e-08, 5.5414e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 214.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4155 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_onlyblock1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0283, 0.0057, -0.0019, ..., -0.0274, -0.0022, 0.0037], + [ 0.0120, 0.0231, -0.0203, ..., 0.0132, -0.0086, 0.0019], + [ 0.0071, 0.0276, -0.0189, ..., 0.0208, -0.0272, 0.0256], + ..., + [-0.0037, 0.0258, 0.0140, ..., -0.0099, 0.0187, 0.0058], + [-0.0209, 0.0135, 0.0214, ..., 0.0033, -0.0044, -0.0090], + [ 0.0023, -0.0041, -0.0280, ..., 0.0172, 0.0045, -0.0192]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0060, -0.0184, -0.0143, 0.0292, -0.0303, 0.0218, -0.0259, -0.0152, + 0.0269, -0.0212], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 231.99, cls_loss 1.2937 cls_loss_mapping 1.8333 cls_loss_causal 2.2064 re_mapping 0.1579 re_causal 0.1671 /// teacc 86.98 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0195, -0.0011, -0.0099, ..., -0.0329, 0.0003, 0.0100], + [ 0.0168, 0.0274, -0.0152, ..., 0.0081, -0.0159, -0.0059], + [ 0.0113, 0.0264, -0.0222, ..., 0.0159, -0.0319, 0.0208], + ..., + [-0.0086, 0.0230, 0.0178, ..., -0.0082, 0.0152, 0.0093], + [-0.0183, 0.0173, 0.0235, ..., -0.0021, -0.0087, -0.0125], + [-0.0050, -0.0019, -0.0277, ..., 0.0171, 0.0072, -0.0164]], + device='cuda:0'), grad: tensor([[ 1.1139e-02, 4.3640e-03, 3.0651e-03, ..., 6.8724e-05, + 4.3259e-03, -1.6613e-03], + [ 5.9998e-02, 2.7451e-02, 2.7939e-02, ..., 2.6512e-04, + 3.8929e-03, 2.0065e-03], + [-3.1982e-02, 2.0313e-03, 4.4632e-03, ..., 3.1328e-04, + 2.0065e-02, 5.4703e-03], + ..., + [-1.0094e-02, 6.6643e-03, -8.4229e-03, ..., 4.8637e-04, + 1.2207e-02, 1.7365e-02], + [-1.4168e-02, -4.0833e-02, -3.7201e-02, ..., 1.4133e-03, + -1.1932e-02, -1.1238e-02], + [ 3.8204e-03, 1.4664e-02, 1.7899e-02, ..., 3.3798e-03, + -7.8659e-03, -2.0279e-02]], device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0043, -0.0173, -0.0152, 0.0291, -0.0306, 0.0230, -0.0267, -0.0152, + 0.0266, -0.0203], device='cuda:0'), grad: tensor([ 0.0111, 0.0471, 0.0069, -0.0267, -0.0258, 0.0181, 0.0187, 0.0158, + -0.0539, -0.0115], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 231.99, cls_loss 0.4344 cls_loss_mapping 0.7874 cls_loss_causal 1.8913 re_mapping 0.2053 re_causal 0.2645 /// teacc 91.77 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0158, -0.0029, -0.0120, ..., -0.0389, 0.0002, 0.0142], + [ 0.0179, 0.0290, -0.0131, ..., 0.0030, -0.0194, -0.0084], + [ 0.0129, 0.0268, -0.0217, ..., 0.0154, -0.0347, 0.0196], + ..., + [-0.0099, 0.0199, 0.0186, ..., -0.0096, 0.0136, 0.0105], + [-0.0168, 0.0210, 0.0265, ..., -0.0091, -0.0121, -0.0159], + [-0.0092, -0.0010, -0.0307, ..., 0.0140, 0.0047, -0.0143]], + device='cuda:0'), grad: tensor([[ 0.0018, 0.0004, 0.0010, ..., 0.0005, 0.0015, -0.0047], + [ 0.0047, 0.0016, 0.0022, ..., 0.0007, 0.0019, 0.0005], + [-0.0011, -0.0251, -0.0250, ..., -0.0050, -0.0090, -0.0049], + ..., + [ 0.0037, -0.0069, -0.0083, ..., 0.0024, 0.0034, -0.0180], + [ 0.0114, 0.0285, 0.0298, ..., 0.0047, 0.0155, 0.0140], + [-0.0162, 0.0347, 0.0147, ..., 0.0229, 0.0208, 0.0150]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0042, -0.0173, -0.0154, 0.0290, -0.0305, 0.0239, -0.0272, -0.0162, + 0.0264, -0.0195], device='cuda:0'), grad: tensor([ 0.0022, 0.0080, -0.0358, -0.0238, -0.0003, -0.0328, 0.0172, -0.0135, + 0.0502, 0.0287], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 230.89, cls_loss 0.3076 cls_loss_mapping 0.4906 cls_loss_causal 1.6779 re_mapping 0.1548 re_causal 0.2422 /// teacc 94.04 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0142, -0.0037, -0.0140, ..., -0.0465, 0.0010, 0.0166], + [ 0.0191, 0.0310, -0.0108, ..., 0.0085, -0.0221, -0.0105], + [ 0.0142, 0.0265, -0.0212, ..., 0.0152, -0.0376, 0.0193], + ..., + [-0.0103, 0.0186, 0.0200, ..., -0.0103, 0.0129, 0.0114], + [-0.0162, 0.0236, 0.0287, ..., -0.0110, -0.0127, -0.0176], + [-0.0115, -0.0013, -0.0345, ..., 0.0108, 0.0021, -0.0130]], + device='cuda:0'), grad: tensor([[ 7.3624e-03, 2.4147e-03, 2.1534e-03, ..., 1.0662e-03, + 2.3308e-03, 5.8860e-06], + [ 1.1559e-02, 3.3340e-03, 4.0359e-03, ..., 1.1997e-03, + 2.5444e-03, 2.9039e-04], + [-9.6863e-02, -1.4381e-02, -1.4992e-02, ..., -8.7280e-03, + -4.8141e-03, 5.6076e-04], + ..., + [ 8.5754e-03, 3.0384e-03, 5.6953e-03, ..., -8.4763e-03, + -7.2556e-03, -4.9591e-04], + [ 2.7679e-02, -7.7248e-03, 1.0376e-02, ..., -3.3245e-03, + -9.3613e-03, 1.6985e-03], + [ 5.0049e-03, -1.3931e-02, 4.0627e-03, ..., -1.6800e-02, + -4.6310e-03, -8.2703e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0047, -0.0169, -0.0156, 0.0288, -0.0310, 0.0241, -0.0278, -0.0163, + 0.0267, -0.0193], device='cuda:0'), grad: tensor([ 0.0080, 0.0088, -0.0566, -0.0040, 0.0675, 0.0193, -0.0138, -0.0052, + -0.0022, -0.0216], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 230.82, cls_loss 0.1945 cls_loss_mapping 0.3135 cls_loss_causal 1.5133 re_mapping 0.1276 re_causal 0.2317 /// teacc 95.09 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 0.0127, -0.0040, -0.0156, ..., -0.0509, 0.0015, 0.0189], + [ 0.0196, 0.0322, -0.0096, ..., 0.0149, -0.0240, -0.0130], + [ 0.0151, 0.0263, -0.0206, ..., 0.0143, -0.0399, 0.0183], + ..., + [-0.0100, 0.0169, 0.0209, ..., -0.0120, 0.0115, 0.0113], + [-0.0156, 0.0257, 0.0306, ..., -0.0139, -0.0139, -0.0199], + [-0.0135, -0.0015, -0.0374, ..., 0.0084, 0.0004, -0.0112]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0007, 0.0011, ..., 0.0002, -0.0038, -0.0168], + [-0.0022, -0.0039, -0.0023, ..., -0.0018, 0.0001, 0.0009], + [-0.0004, -0.0060, -0.0090, ..., -0.0016, 0.0004, -0.0004], + ..., + [-0.0010, 0.0049, -0.0049, ..., 0.0023, 0.0004, 0.0019], + [ 0.0055, 0.0012, 0.0018, ..., 0.0014, 0.0016, 0.0039], + [ 0.0088, 0.0002, 0.0041, ..., 0.0014, 0.0010, -0.0010]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0050, -0.0169, -0.0156, 0.0289, -0.0311, 0.0242, -0.0282, -0.0165, + 0.0266, -0.0191], device='cuda:0'), grad: tensor([-0.0098, -0.0013, -0.0026, -0.0139, 0.0045, 0.0018, 0.0060, 0.0021, + 0.0070, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 229.97, cls_loss 0.1757 cls_loss_mapping 0.2546 cls_loss_causal 1.3651 re_mapping 0.1016 re_causal 0.1965 /// teacc 96.59 lr 0.00010000 +Epoch 6, weight, value: tensor([[ 0.0113, -0.0053, -0.0164, ..., -0.0576, 0.0009, 0.0211], + [ 0.0204, 0.0336, -0.0081, ..., 0.0198, -0.0259, -0.0149], + [ 0.0159, 0.0265, -0.0204, ..., 0.0145, -0.0417, 0.0168], + ..., + [-0.0102, 0.0151, 0.0212, ..., -0.0132, 0.0108, 0.0115], + [-0.0149, 0.0278, 0.0332, ..., -0.0150, -0.0148, -0.0218], + [-0.0157, -0.0016, -0.0411, ..., 0.0058, -0.0011, -0.0093]], + device='cuda:0'), grad: tensor([[ 7.1669e-04, -9.8228e-04, 4.6039e-04, ..., 4.1962e-04, + -5.6190e-03, -6.1760e-03], + [-1.1383e-02, -2.2659e-02, -7.9880e-03, ..., 3.7581e-05, + 3.4981e-03, 5.7793e-04], + [ 7.3967e-03, 4.3411e-03, 4.0741e-03, ..., 1.1892e-03, + 1.7633e-03, 8.8358e-04], + ..., + [ 2.4395e-03, 3.4904e-03, 1.9026e-03, ..., 2.0676e-03, + 3.5534e-03, 2.1248e-03], + [ 3.4119e-02, 3.0869e-02, 1.8158e-02, ..., 6.8169e-03, + 1.0918e-02, 2.0428e-03], + [ 7.1640e-03, 8.7204e-03, 3.9864e-03, ..., 4.5509e-03, + 1.0605e-02, 1.1234e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0052, -0.0166, -0.0160, 0.0288, -0.0313, 0.0239, -0.0280, -0.0165, + 0.0270, -0.0190], device='cuda:0'), grad: tensor([-0.0066, -0.0114, 0.0071, -0.0211, -0.0257, -0.0105, 0.0057, 0.0066, + 0.0380, 0.0178], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 228.69, cls_loss 0.1227 cls_loss_mapping 0.1963 cls_loss_causal 1.3091 re_mapping 0.0877 re_causal 0.1888 /// teacc 96.97 lr 0.00010000 +Epoch 7, weight, value: tensor([[ 0.0099, -0.0063, -0.0176, ..., -0.0597, 0.0017, 0.0229], + [ 0.0216, 0.0348, -0.0071, ..., 0.0255, -0.0275, -0.0157], + [ 0.0165, 0.0272, -0.0205, ..., 0.0139, -0.0431, 0.0164], + ..., + [-0.0108, 0.0134, 0.0216, ..., -0.0152, 0.0098, 0.0113], + [-0.0143, 0.0294, 0.0356, ..., -0.0172, -0.0158, -0.0239], + [-0.0183, -0.0022, -0.0437, ..., 0.0035, -0.0026, -0.0079]], + device='cuda:0'), grad: tensor([[ 5.2547e-04, 1.7238e-04, 5.6076e-04, ..., 8.1301e-04, + 1.9608e-02, 7.3700e-03], + [ 2.6764e-02, 8.1558e-03, 2.4536e-02, ..., 4.0245e-03, + 4.0698e-04, 7.1812e-04], + [ 9.5215e-03, 2.5463e-03, 7.8659e-03, ..., -4.7386e-05, + 3.1710e-04, 1.7176e-03], + ..., + [-2.2869e-03, -1.3161e-03, -1.0071e-02, ..., -4.3793e-03, + 3.5048e-04, -9.9945e-03], + [-7.6981e-03, -9.5978e-03, -1.2505e-02, ..., -3.7742e-04, + 9.8324e-04, 2.5902e-03], + [ 3.4218e-03, -7.9956e-03, 4.2114e-03, ..., -8.4915e-03, + -3.1738e-03, 3.2715e-02]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0058, -0.0161, -0.0161, 0.0288, -0.0313, 0.0239, -0.0283, -0.0167, + 0.0271, -0.0195], device='cuda:0'), grad: tensor([ 0.0134, 0.0200, 0.0075, -0.0239, -0.0157, 0.0022, -0.0143, -0.0118, + -0.0039, 0.0265], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 231.00, cls_loss 0.1235 cls_loss_mapping 0.1809 cls_loss_causal 1.2598 re_mapping 0.0725 re_causal 0.1606 /// teacc 97.22 lr 0.00010000 +Epoch 8, weight, value: tensor([[ 0.0087, -0.0077, -0.0190, ..., -0.0631, 0.0012, 0.0243], + [ 0.0221, 0.0360, -0.0062, ..., 0.0289, -0.0290, -0.0173], + [ 0.0169, 0.0272, -0.0208, ..., 0.0133, -0.0445, 0.0154], + ..., + [-0.0114, 0.0119, 0.0216, ..., -0.0166, 0.0089, 0.0115], + [-0.0134, 0.0313, 0.0377, ..., -0.0196, -0.0167, -0.0255], + [-0.0199, -0.0026, -0.0460, ..., 0.0018, -0.0041, -0.0062]], + device='cuda:0'), grad: tensor([[ 1.5819e-04, -1.9140e-03, 9.8288e-05, ..., 8.1718e-05, + -1.2047e-02, -1.0170e-02], + [-9.6560e-04, -1.1635e-03, -1.5917e-03, ..., -1.5030e-03, + 1.3500e-05, 5.3453e-04], + [ 1.1978e-03, 6.4230e-04, 1.0929e-03, ..., 1.6069e-04, + 8.1682e-04, 9.5844e-04], + ..., + [-8.7690e-04, 5.4312e-04, -1.5545e-03, ..., 1.8406e-04, + 3.1519e-04, 9.2030e-04], + [ 1.1740e-03, 2.9507e-03, 4.9740e-05, ..., 6.7854e-04, + 1.1415e-03, 3.4370e-03], + [-2.5082e-04, -6.4659e-03, -6.6757e-04, ..., 3.1638e-04, + 6.9475e-04, -7.2060e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0056, -0.0161, -0.0163, 0.0290, -0.0312, 0.0235, -0.0285, -0.0167, + 0.0275, -0.0192], device='cuda:0'), grad: tensor([-0.0138, -0.0006, 0.0022, -0.0002, 0.0040, 0.0033, 0.0089, 0.0007, + 0.0064, -0.0108], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 228.21, cls_loss 0.1154 cls_loss_mapping 0.1719 cls_loss_causal 1.1922 re_mapping 0.0663 re_causal 0.1498 /// teacc 97.30 lr 0.00010000 +Epoch 9, weight, value: tensor([[ 0.0076, -0.0087, -0.0198, ..., -0.0631, 0.0014, 0.0252], + [ 0.0220, 0.0362, -0.0058, ..., 0.0305, -0.0305, -0.0192], + [ 0.0177, 0.0274, -0.0216, ..., 0.0136, -0.0462, 0.0157], + ..., + [-0.0115, 0.0108, 0.0224, ..., -0.0181, 0.0087, 0.0117], + [-0.0127, 0.0326, 0.0393, ..., -0.0205, -0.0182, -0.0272], + [-0.0218, -0.0026, -0.0476, ..., 0.0004, -0.0057, -0.0047]], + device='cuda:0'), grad: tensor([[ 4.3631e-04, 2.1422e-04, 1.2326e-04, ..., 1.8859e-04, + 1.6940e-04, -4.0221e-04], + [ 3.6449e-03, -6.8617e-04, 1.6804e-03, ..., -2.6274e-04, + 1.4114e-04, 1.3554e-04], + [-8.5220e-03, -1.2598e-03, -5.0735e-03, ..., -1.1053e-03, + 1.4949e-04, 1.7297e-04], + ..., + [ 2.5349e-03, 1.1787e-03, 2.8634e-04, ..., 8.5497e-04, + 6.2585e-05, 2.7347e-04], + [-2.0123e-03, -4.9934e-03, -7.0915e-03, ..., 5.4073e-04, + 5.9700e-04, -2.7676e-03], + [ 5.3139e-03, 2.2449e-03, 5.5046e-03, ..., -1.0061e-03, + 7.7295e-04, -2.2507e-04]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0058, -0.0165, -0.0160, 0.0292, -0.0313, 0.0231, -0.0286, -0.0165, + 0.0276, -0.0192], device='cuda:0'), grad: tensor([ 0.0003, 0.0023, -0.0044, -0.0017, 0.0037, 0.0019, -0.0011, 0.0018, + -0.0073, 0.0044], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 228.98, cls_loss 0.0898 cls_loss_mapping 0.1338 cls_loss_causal 1.1258 re_mapping 0.0587 re_causal 0.1372 /// teacc 97.71 lr 0.00010000 +Epoch 10, weight, value: tensor([[ 0.0065, -0.0102, -0.0210, ..., -0.0642, 0.0015, 0.0265], + [ 0.0222, 0.0369, -0.0056, ..., 0.0331, -0.0313, -0.0203], + [ 0.0182, 0.0274, -0.0222, ..., 0.0134, -0.0471, 0.0149], + ..., + [-0.0116, 0.0101, 0.0229, ..., -0.0191, 0.0082, 0.0123], + [-0.0121, 0.0343, 0.0417, ..., -0.0216, -0.0192, -0.0281], + [-0.0234, -0.0028, -0.0499, ..., -0.0009, -0.0066, -0.0036]], + device='cuda:0'), grad: tensor([[ 6.4898e-04, 5.1928e-04, 4.2367e-04, ..., 2.3460e-04, + 4.7135e-04, -8.6188e-05], + [ 3.5114e-03, 4.6577e-03, 4.2229e-03, ..., -1.1616e-03, + 1.7347e-03, 7.9441e-04], + [ 1.3895e-03, 9.8324e-04, 1.4963e-03, ..., 4.5466e-04, + 3.9053e-04, 1.7548e-04], + ..., + [-2.8658e-04, 1.5049e-03, -6.1369e-04, ..., 5.9652e-04, + 8.0490e-04, 9.3746e-04], + [-7.4539e-03, -1.0101e-02, -8.0795e-03, ..., 8.0729e-04, + 2.3365e-04, 1.0242e-03], + [ 9.8038e-04, -2.5940e-03, 7.8964e-04, ..., -1.0643e-03, + -9.5367e-04, -3.1643e-03]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0060, -0.0165, -0.0161, 0.0294, -0.0318, 0.0229, -0.0288, -0.0162, + 0.0281, -0.0195], device='cuda:0'), grad: tensor([ 0.0013, 0.0081, 0.0019, -0.0064, 0.0051, 0.0051, 0.0007, 0.0024, + -0.0098, -0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 214.92, cls_loss 0.0777 cls_loss_mapping 0.1180 cls_loss_causal 1.0594 re_mapping 0.0545 re_causal 0.1298 /// teacc 97.57 lr 0.00010000 +Epoch 11, weight, value: tensor([[ 0.0058, -0.0112, -0.0219, ..., -0.0652, 0.0016, 0.0273], + [ 0.0225, 0.0369, -0.0052, ..., 0.0347, -0.0328, -0.0216], + [ 0.0187, 0.0275, -0.0221, ..., 0.0125, -0.0483, 0.0140], + ..., + [-0.0120, 0.0087, 0.0229, ..., -0.0201, 0.0083, 0.0124], + [-0.0119, 0.0354, 0.0432, ..., -0.0228, -0.0202, -0.0290], + [-0.0243, -0.0027, -0.0512, ..., -0.0020, -0.0078, -0.0020]], + device='cuda:0'), grad: tensor([[ 2.6727e-04, 1.6725e-04, 1.4603e-04, ..., 2.0564e-04, + 1.9717e-04, -1.4753e-03], + [ 7.7629e-04, -4.0591e-05, 7.5176e-06, ..., -4.4107e-04, + 3.3355e-04, 2.8539e-04], + [-1.0757e-03, -1.2201e-04, 2.1994e-05, ..., 3.1996e-04, + 2.1851e-04, 5.7697e-04], + ..., + [ 1.0729e-03, -1.0002e-04, -1.6432e-03, ..., -1.0138e-03, + 9.4235e-05, -8.3542e-04], + [ 1.7977e-03, -5.2959e-05, -4.4441e-04, ..., 5.9128e-04, + 1.1911e-03, 1.2751e-03], + [ 1.2703e-03, 3.2973e-04, 1.2531e-03, ..., 5.2500e-04, + 3.1829e-04, -5.1785e-04]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0060, -0.0168, -0.0160, 0.0292, -0.0316, 0.0232, -0.0292, -0.0163, + 0.0280, -0.0191], device='cuda:0'), grad: tensor([-6.4278e-04, 1.0014e-03, -6.0737e-05, -1.2226e-03, 1.1005e-03, + -3.6278e-03, -1.4198e-04, -4.6611e-04, 3.1166e-03, 9.4461e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 226.90, cls_loss 0.0687 cls_loss_mapping 0.1066 cls_loss_causal 1.0942 re_mapping 0.0480 re_causal 0.1214 /// teacc 97.76 lr 0.00010000 +Epoch 12, weight, value: tensor([[ 0.0047, -0.0118, -0.0222, ..., -0.0654, 0.0013, 0.0285], + [ 0.0225, 0.0372, -0.0048, ..., 0.0361, -0.0343, -0.0224], + [ 0.0194, 0.0278, -0.0227, ..., 0.0122, -0.0493, 0.0129], + ..., + [-0.0123, 0.0077, 0.0234, ..., -0.0213, 0.0078, 0.0124], + [-0.0114, 0.0362, 0.0445, ..., -0.0241, -0.0215, -0.0305], + [-0.0257, -0.0028, -0.0523, ..., -0.0031, -0.0087, -0.0012]], + device='cuda:0'), grad: tensor([[ 3.2926e-04, 5.4026e-04, 8.6451e-04, ..., 1.0991e-04, + -2.7919e-04, -2.1577e-04], + [-1.2421e-02, 1.3878e-02, 1.3016e-02, ..., -8.5144e-03, + 5.3263e-04, 1.2140e-03], + [ 2.1240e-02, 3.9978e-03, 5.5580e-03, ..., 6.3820e-03, + 1.7965e-04, -1.1784e-04], + ..., + [ 4.0779e-03, 6.8398e-03, 4.7264e-03, ..., 6.7091e-04, + -1.2312e-03, 2.6588e-03], + [-2.0782e-02, -3.1036e-02, -3.0350e-02, ..., -1.1019e-05, + 3.1567e-04, 1.4315e-03], + [ 2.3067e-04, -1.7395e-03, 1.0681e-03, ..., -3.7479e-04, + 2.7728e-04, -1.0612e-02]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0061, -0.0169, -0.0160, 0.0295, -0.0317, 0.0234, -0.0291, -0.0164, + 0.0279, -0.0194], device='cuda:0'), grad: tensor([ 0.0004, 0.0002, 0.0153, 0.0083, 0.0033, 0.0011, 0.0017, 0.0161, + -0.0257, -0.0208], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 231.33, cls_loss 0.0671 cls_loss_mapping 0.1050 cls_loss_causal 1.0557 re_mapping 0.0455 re_causal 0.1133 /// teacc 97.85 lr 0.00010000 +Epoch 13, weight, value: tensor([[ 0.0036, -0.0132, -0.0229, ..., -0.0657, 0.0017, 0.0297], + [ 0.0226, 0.0373, -0.0044, ..., 0.0368, -0.0356, -0.0230], + [ 0.0197, 0.0280, -0.0231, ..., 0.0118, -0.0501, 0.0123], + ..., + [-0.0124, 0.0065, 0.0238, ..., -0.0209, 0.0084, 0.0126], + [-0.0108, 0.0374, 0.0463, ..., -0.0251, -0.0228, -0.0320], + [-0.0269, -0.0026, -0.0539, ..., -0.0043, -0.0098, 0.0002]], + device='cuda:0'), grad: tensor([[ 3.8147e-04, 8.7738e-04, 9.3937e-04, ..., 4.2605e-04, + 9.2506e-05, 3.4690e-05], + [-1.4696e-03, -8.5306e-04, -3.2592e-04, ..., -8.4877e-04, + 1.3880e-05, 7.1239e-04], + [ 1.2350e-03, 1.4381e-03, 1.4677e-03, ..., 6.7329e-04, + 1.3065e-04, 7.8773e-04], + ..., + [ 4.0531e-04, 1.9779e-03, 1.1644e-03, ..., 1.0471e-03, + 2.3699e-04, 1.4505e-03], + [-4.0855e-03, -7.9269e-03, -9.6893e-03, ..., -2.6054e-03, + 1.9484e-03, -5.1003e-03], + [ 1.3075e-03, 2.9812e-03, 2.2297e-03, ..., 1.3323e-03, + 8.3494e-04, 2.0313e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0060, -0.0173, -0.0161, 0.0293, -0.0317, 0.0232, -0.0291, -0.0161, + 0.0282, -0.0193], device='cuda:0'), grad: tensor([ 0.0012, -0.0005, 0.0024, 0.0032, -0.0014, -0.0036, 0.0024, 0.0025, + -0.0115, 0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 215.23, cls_loss 0.0634 cls_loss_mapping 0.0970 cls_loss_causal 1.0147 re_mapping 0.0415 re_causal 0.1077 /// teacc 97.65 lr 0.00010000 +Epoch 14, weight, value: tensor([[ 0.0029, -0.0142, -0.0232, ..., -0.0666, 0.0016, 0.0305], + [ 0.0227, 0.0372, -0.0036, ..., 0.0376, -0.0365, -0.0242], + [ 0.0203, 0.0284, -0.0236, ..., 0.0123, -0.0508, 0.0116], + ..., + [-0.0122, 0.0062, 0.0249, ..., -0.0218, 0.0080, 0.0128], + [-0.0103, 0.0385, 0.0474, ..., -0.0259, -0.0230, -0.0329], + [-0.0280, -0.0026, -0.0550, ..., -0.0054, -0.0110, 0.0013]], + device='cuda:0'), grad: tensor([[ 1.3447e-03, 7.3767e-04, 5.6267e-04, ..., 2.0659e-04, + 7.5674e-04, 7.1347e-05], + [-5.8508e-04, -1.1187e-03, -4.1389e-04, ..., -1.5078e-03, + 1.1104e-04, 8.0407e-05], + [ 6.5460e-03, 3.5801e-03, 3.0880e-03, ..., 1.3819e-03, + 8.0538e-04, 1.6582e-04], + ..., + [ 1.3008e-03, 1.9445e-03, 2.3305e-04, ..., 3.3798e-03, + 4.3716e-03, 6.6662e-04], + [-4.9820e-03, -3.5076e-03, -2.2850e-03, ..., 5.4693e-04, + 4.7607e-03, 7.7677e-04], + [ 1.7738e-03, 1.0406e-02, 5.7840e-04, ..., 8.9188e-03, + 1.5099e-02, 5.3253e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0061, -0.0175, -0.0161, 0.0292, -0.0320, 0.0231, -0.0290, -0.0158, + 0.0286, -0.0194], device='cuda:0'), grad: tensor([ 0.0024, -0.0004, 0.0093, -0.0086, -0.0267, -0.0250, 0.0245, 0.0065, + -0.0014, 0.0194], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 231.46, cls_loss 0.0508 cls_loss_mapping 0.0822 cls_loss_causal 0.9909 re_mapping 0.0390 re_causal 0.1015 /// teacc 98.03 lr 0.00010000 +Epoch 15, weight, value: tensor([[ 0.0022, -0.0154, -0.0236, ..., -0.0678, 0.0018, 0.0315], + [ 0.0228, 0.0379, -0.0032, ..., 0.0382, -0.0372, -0.0253], + [ 0.0206, 0.0284, -0.0246, ..., 0.0126, -0.0518, 0.0118], + ..., + [-0.0123, 0.0055, 0.0258, ..., -0.0222, 0.0067, 0.0127], + [-0.0099, 0.0390, 0.0486, ..., -0.0265, -0.0240, -0.0338], + [-0.0291, -0.0025, -0.0563, ..., -0.0061, -0.0113, 0.0020]], + device='cuda:0'), grad: tensor([[ 1.1826e-03, 1.0786e-03, 4.8190e-05, ..., 1.9515e-04, + 2.0385e-05, -4.6825e-04], + [-1.6415e-04, -1.2236e-03, -1.4362e-03, ..., -1.7853e-03, + -6.5756e-04, -2.8804e-05], + [-5.0621e-03, -3.8280e-03, -1.8573e-04, ..., -1.0939e-03, + 1.6332e-04, 2.6250e-04], + ..., + [ 7.9346e-04, 1.0300e-03, 4.6015e-04, ..., 6.6423e-04, + 1.4973e-04, 9.1934e-04], + [ 4.2009e-04, 3.7003e-04, -5.3453e-04, ..., 7.2002e-04, + 6.1512e-04, 2.6655e-04], + [ 3.3402e-04, -6.1703e-04, 2.8419e-04, ..., -6.6280e-05, + 9.9277e-04, -3.5992e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0063, -0.0176, -0.0159, 0.0294, -0.0321, 0.0232, -0.0290, -0.0159, + 0.0286, -0.0195], device='cuda:0'), grad: tensor([ 0.0012, -0.0010, -0.0054, 0.0013, 0.0019, 0.0011, 0.0006, 0.0020, + 0.0011, -0.0029], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 231.06, cls_loss 0.0531 cls_loss_mapping 0.0789 cls_loss_causal 0.9585 re_mapping 0.0372 re_causal 0.0959 /// teacc 98.13 lr 0.00010000 +Epoch 16, weight, value: tensor([[ 0.0017, -0.0163, -0.0240, ..., -0.0697, 0.0019, 0.0326], + [ 0.0230, 0.0386, -0.0029, ..., 0.0393, -0.0382, -0.0246], + [ 0.0212, 0.0283, -0.0248, ..., 0.0122, -0.0527, 0.0108], + ..., + [-0.0127, 0.0046, 0.0261, ..., -0.0222, 0.0068, 0.0128], + [-0.0096, 0.0397, 0.0494, ..., -0.0271, -0.0245, -0.0352], + [-0.0298, -0.0027, -0.0570, ..., -0.0073, -0.0125, 0.0030]], + device='cuda:0'), grad: tensor([[ 5.3835e-04, 8.8453e-04, 4.8566e-04, ..., 3.9029e-04, + 1.2894e-03, 9.9897e-05], + [-1.5678e-03, -5.1117e-03, -2.0676e-03, ..., -3.2654e-03, + -1.6332e-04, 1.0079e-04], + [ 8.1024e-03, 5.0468e-03, 6.8932e-03, ..., -9.8705e-05, + 5.3674e-05, 2.1589e-04], + ..., + [ 5.5161e-03, 2.0905e-03, 3.2120e-03, ..., 2.8563e-04, + 3.1561e-05, -8.8692e-04], + [-2.9507e-03, -7.5989e-03, -6.4392e-03, ..., 1.5297e-03, + -2.0552e-04, 7.0669e-06], + [ 1.9245e-03, 7.5960e-04, 3.9911e-04, ..., 1.3947e-04, + 7.0095e-05, 1.7681e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0066, -0.0174, -0.0159, 0.0293, -0.0322, 0.0231, -0.0290, -0.0161, + 0.0286, -0.0196], device='cuda:0'), grad: tensor([ 0.0015, -0.0034, 0.0091, -0.0182, 0.0008, 0.0014, 0.0060, 0.0046, + -0.0062, 0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 215.12, cls_loss 0.0525 cls_loss_mapping 0.0857 cls_loss_causal 0.9667 re_mapping 0.0353 re_causal 0.0921 /// teacc 98.09 lr 0.00010000 +Epoch 17, weight, value: tensor([[ 0.0011, -0.0177, -0.0244, ..., -0.0704, 0.0016, 0.0336], + [ 0.0233, 0.0389, -0.0022, ..., 0.0402, -0.0393, -0.0249], + [ 0.0216, 0.0290, -0.0251, ..., 0.0121, -0.0535, 0.0097], + ..., + [-0.0132, 0.0027, 0.0260, ..., -0.0235, 0.0068, 0.0127], + [-0.0090, 0.0409, 0.0505, ..., -0.0283, -0.0258, -0.0362], + [-0.0307, -0.0023, -0.0578, ..., -0.0076, -0.0131, 0.0038]], + device='cuda:0'), grad: tensor([[ 5.5701e-05, 2.3961e-04, 1.0751e-05, ..., 2.5123e-05, + 4.2707e-05, 3.4475e-04], + [-1.7881e-04, -3.8475e-05, -1.9240e-04, ..., -1.2481e-04, + 1.4341e-04, 7.2718e-05], + [ 3.7169e-04, 3.2997e-04, 2.1982e-04, ..., 1.5676e-04, + 7.1645e-05, 2.8896e-04], + ..., + [ 7.4565e-05, 1.8740e-04, -1.8525e-04, ..., 1.0735e-04, + 8.8394e-05, 2.2340e-04], + [ 2.2602e-04, 1.5574e-03, 5.0701e-06, ..., 2.3878e-04, + 5.2881e-04, 2.3689e-03], + [-1.9515e-04, -3.1433e-03, 3.8445e-05, ..., 6.7091e-04, + 1.0738e-03, -4.5891e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0067, -0.0174, -0.0161, 0.0294, -0.0318, 0.0229, -0.0292, -0.0165, + 0.0289, -0.0195], device='cuda:0'), grad: tensor([ 9.1124e-04, 7.7069e-05, 1.1501e-03, 7.0620e-04, 2.1439e-03, + -7.5912e-04, 9.4831e-05, 3.5644e-04, 5.6686e-03, -1.0353e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 224.71, cls_loss 0.0459 cls_loss_mapping 0.0742 cls_loss_causal 0.9690 re_mapping 0.0325 re_causal 0.0910 /// teacc 98.17 lr 0.00010000 +Epoch 18, weight, value: tensor([[ 0.0002, -0.0183, -0.0247, ..., -0.0705, 0.0015, 0.0343], + [ 0.0235, 0.0391, -0.0017, ..., 0.0409, -0.0397, -0.0252], + [ 0.0220, 0.0290, -0.0254, ..., 0.0114, -0.0544, 0.0100], + ..., + [-0.0136, 0.0018, 0.0262, ..., -0.0237, 0.0064, 0.0130], + [-0.0084, 0.0417, 0.0516, ..., -0.0288, -0.0266, -0.0371], + [-0.0314, -0.0023, -0.0583, ..., -0.0089, -0.0136, 0.0046]], + device='cuda:0'), grad: tensor([[-8.8692e-04, -2.2984e-03, 4.0889e-05, ..., 3.0190e-05, + 1.5903e-04, -7.0801e-03], + [-3.3020e-02, -1.4282e-02, -1.3641e-02, ..., -3.4213e-05, + 9.1493e-05, -3.7937e-03], + [ 3.1952e-02, 1.4763e-02, 1.3283e-02, ..., 3.4779e-05, + 1.7500e-04, 7.2212e-03], + ..., + [ 9.7847e-04, 5.8365e-04, 1.4889e-04, ..., 3.2008e-05, + 9.0122e-05, 6.5804e-04], + [ 2.8248e-03, -7.0715e-04, -3.1924e-04, ..., 7.9095e-05, + 6.9857e-04, 1.4400e-03], + [ 1.0185e-03, 5.2929e-04, 1.3721e-04, ..., 9.1732e-05, + 5.4646e-04, 5.4693e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0068, -0.0175, -0.0160, 0.0291, -0.0318, 0.0231, -0.0296, -0.0163, + 0.0289, -0.0194], device='cuda:0'), grad: tensor([-0.0062, -0.0246, 0.0258, -0.0078, 0.0003, 0.0048, 0.0011, 0.0017, + 0.0029, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 231.18, cls_loss 0.0493 cls_loss_mapping 0.0760 cls_loss_causal 0.9548 re_mapping 0.0332 re_causal 0.0872 /// teacc 98.38 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0003, -0.0189, -0.0247, ..., -0.0708, 0.0011, 0.0351], + [ 0.0234, 0.0392, -0.0016, ..., 0.0414, -0.0404, -0.0255], + [ 0.0223, 0.0287, -0.0259, ..., 0.0111, -0.0552, 0.0091], + ..., + [-0.0136, 0.0011, 0.0270, ..., -0.0241, 0.0062, 0.0127], + [-0.0079, 0.0427, 0.0528, ..., -0.0296, -0.0273, -0.0384], + [-0.0324, -0.0022, -0.0594, ..., -0.0098, -0.0141, 0.0053]], + device='cuda:0'), grad: tensor([[ 9.3758e-05, -8.0013e-04, -3.6788e-04, ..., 6.9499e-05, + -1.0424e-03, -5.0812e-03], + [ 9.8169e-05, 6.8784e-05, 1.0657e-04, ..., -9.1910e-05, + 1.2219e-04, 1.4889e-04], + [ 7.2956e-04, 5.8603e-04, 9.2983e-04, ..., 1.5342e-04, + 1.8096e-04, 5.5218e-04], + ..., + [-1.1854e-03, -1.2369e-03, -1.9293e-03, ..., 1.0896e-04, + 1.2434e-04, -6.9571e-04], + [ 4.9353e-04, 1.5421e-03, 6.5374e-04, ..., 5.1451e-04, + 2.4548e-03, 4.8141e-03], + [ 4.0102e-04, 2.7599e-03, 4.0936e-04, ..., 2.8343e-03, + 1.6794e-03, 1.4997e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0067, -0.0177, -0.0161, 0.0291, -0.0317, 0.0232, -0.0292, -0.0163, + 0.0289, -0.0195], device='cuda:0'), grad: tensor([-0.0047, 0.0004, 0.0018, -0.0004, -0.0017, 0.0001, -0.0018, -0.0026, + 0.0062, 0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 215.06, cls_loss 0.0409 cls_loss_mapping 0.0697 cls_loss_causal 0.8914 re_mapping 0.0304 re_causal 0.0880 /// teacc 98.36 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0007, -0.0197, -0.0250, ..., -0.0724, 0.0009, 0.0354], + [ 0.0234, 0.0395, -0.0014, ..., 0.0425, -0.0402, -0.0263], + [ 0.0225, 0.0294, -0.0263, ..., 0.0107, -0.0559, 0.0088], + ..., + [-0.0139, 0.0007, 0.0275, ..., -0.0250, 0.0058, 0.0130], + [-0.0077, 0.0430, 0.0534, ..., -0.0303, -0.0276, -0.0383], + [-0.0331, -0.0024, -0.0601, ..., -0.0102, -0.0150, 0.0063]], + device='cuda:0'), grad: tensor([[ 8.7023e-04, 4.3273e-05, 8.1778e-05, ..., 3.0184e-04, + -1.2033e-06, 3.4285e-04], + [ 3.3021e-04, -2.0373e-04, -2.6870e-04, ..., -1.8418e-04, + 3.2306e-05, 4.4727e-04], + [-1.0864e-02, 2.9182e-04, -9.2936e-04, ..., -1.8158e-03, + 1.5438e-05, -4.9973e-03], + ..., + [ 1.2007e-03, 2.5773e-04, -1.0568e-04, ..., 4.6945e-04, + 1.8969e-05, 6.5041e-04], + [ 2.9297e-03, 4.5967e-04, 3.1918e-05, ..., 2.5725e-04, + 4.6819e-05, 3.8719e-04], + [ 4.8423e-04, 8.1599e-05, 1.6391e-04, ..., 3.3736e-04, + 1.3840e-04, 3.9250e-05]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0063, -0.0177, -0.0164, 0.0294, -0.0317, 0.0233, -0.0295, -0.0163, + 0.0291, -0.0194], device='cuda:0'), grad: tensor([ 0.0020, 0.0010, -0.0245, 0.0107, 0.0004, 0.0024, 0.0007, 0.0026, + 0.0039, 0.0008], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 230.92, cls_loss 0.0366 cls_loss_mapping 0.0600 cls_loss_causal 0.8907 re_mapping 0.0295 re_causal 0.0805 /// teacc 98.56 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0012, -0.0203, -0.0253, ..., -0.0728, 0.0009, 0.0360], + [ 0.0234, 0.0395, -0.0014, ..., 0.0427, -0.0412, -0.0274], + [ 0.0228, 0.0295, -0.0266, ..., 0.0112, -0.0561, 0.0087], + ..., + [-0.0143, -0.0004, 0.0277, ..., -0.0249, 0.0062, 0.0130], + [-0.0074, 0.0438, 0.0543, ..., -0.0310, -0.0284, -0.0388], + [-0.0339, -0.0019, -0.0606, ..., -0.0111, -0.0158, 0.0074]], + device='cuda:0'), grad: tensor([[ 5.4646e-04, 1.3399e-04, 5.8204e-05, ..., 8.6069e-04, + 1.0862e-03, -3.8713e-05], + [-3.3875e-03, -2.5845e-03, -5.9557e-04, ..., -1.5802e-03, + -1.5726e-03, 5.9456e-05], + [ 1.1665e-02, 6.4433e-05, 7.8797e-05, ..., 1.6603e-03, + 2.1572e-03, 7.4100e-04], + ..., + [ 4.9639e-04, 1.0461e-04, -8.8215e-05, ..., 1.7965e-04, + 2.1565e-04, -1.1429e-05], + [ 3.2673e-03, 2.4986e-04, -1.8513e-04, ..., 4.9067e-04, + 5.7268e-04, 2.0254e-04], + [ 3.0470e-04, 2.3413e-04, 8.2254e-05, ..., 3.9768e-04, + 4.3058e-04, -1.9956e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0063, -0.0182, -0.0163, 0.0294, -0.0322, 0.0235, -0.0297, -0.0160, + 0.0293, -0.0192], device='cuda:0'), grad: tensor([ 0.0018, -0.0043, 0.0197, 0.0069, 0.0056, -0.0284, -0.0077, 0.0007, + 0.0049, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 214.83, cls_loss 0.0401 cls_loss_mapping 0.0592 cls_loss_causal 0.8446 re_mapping 0.0291 re_causal 0.0799 /// teacc 98.37 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0016, -0.0212, -0.0256, ..., -0.0731, 0.0007, 0.0362], + [ 0.0240, 0.0400, -0.0005, ..., 0.0436, -0.0421, -0.0276], + [ 0.0227, 0.0291, -0.0275, ..., 0.0103, -0.0570, 0.0079], + ..., + [-0.0143, -0.0006, 0.0283, ..., -0.0248, 0.0059, 0.0132], + [-0.0074, 0.0444, 0.0546, ..., -0.0315, -0.0287, -0.0397], + [-0.0349, -0.0017, -0.0616, ..., -0.0122, -0.0162, 0.0090]], + device='cuda:0'), grad: tensor([[-2.0005e-06, -2.9564e-03, 1.8865e-05, ..., -8.0824e-04, + -1.0977e-03, -5.6648e-03], + [-6.8732e-06, -1.4740e-02, -1.6418e-02, ..., -1.9592e-02, + 1.9741e-04, -1.3535e-02], + [ 2.1529e-04, 4.9067e-04, 7.7903e-05, ..., 1.3316e-04, + 5.7364e-04, 8.5926e-04], + ..., + [ 1.7536e-04, 1.3176e-02, 1.4229e-02, ..., 1.7181e-02, + 2.1052e-04, 1.2131e-02], + [ 5.8222e-04, 6.3610e-04, 1.0544e-04, ..., 4.9639e-04, + 8.5163e-04, 7.8249e-04], + [ 1.2010e-04, 1.9951e-03, 1.5574e-03, ..., 2.1553e-03, + 4.8709e-04, 2.8820e-03]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0060, -0.0182, -0.0165, 0.0292, -0.0325, 0.0236, -0.0295, -0.0153, + 0.0293, -0.0192], device='cuda:0'), grad: tensor([-0.0053, -0.0250, 0.0015, -0.0002, 0.0004, -0.0008, 0.0004, 0.0226, + 0.0023, 0.0040], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 214.91, cls_loss 0.0364 cls_loss_mapping 0.0639 cls_loss_causal 0.8884 re_mapping 0.0279 re_causal 0.0794 /// teacc 98.46 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0021, -0.0223, -0.0260, ..., -0.0741, 0.0005, 0.0369], + [ 0.0241, 0.0403, -0.0002, ..., 0.0444, -0.0430, -0.0277], + [ 0.0233, 0.0293, -0.0276, ..., 0.0100, -0.0577, 0.0078], + ..., + [-0.0150, -0.0017, 0.0281, ..., -0.0254, 0.0054, 0.0131], + [-0.0072, 0.0449, 0.0556, ..., -0.0325, -0.0289, -0.0408], + [-0.0356, -0.0015, -0.0626, ..., -0.0138, -0.0170, 0.0097]], + device='cuda:0'), grad: tensor([[ 7.0691e-05, -1.1325e-05, 1.6049e-05, ..., 2.6375e-05, + -1.0481e-03, -2.1782e-03], + [-3.4308e-04, -5.4026e-04, -2.6584e-04, ..., -8.2970e-04, + 1.1675e-05, 2.4229e-05], + [ 9.4414e-05, 1.0532e-04, 1.5438e-04, ..., 2.0075e-04, + 3.4720e-05, 8.7619e-05], + ..., + [ 1.5104e-04, 1.1742e-04, -4.5866e-05, ..., 1.2362e-04, + 2.3440e-05, 5.9083e-06], + [ 1.7393e-04, 2.8163e-05, 3.2425e-05, ..., 4.1038e-05, + 1.6737e-04, 2.9874e-04], + [ 6.6662e-04, -2.3341e-04, 6.0368e-04, ..., 2.0415e-05, + 8.6188e-05, 7.2241e-05]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0058, -0.0182, -0.0162, 0.0295, -0.0322, 0.0233, -0.0288, -0.0161, + 0.0292, -0.0194], device='cuda:0'), grad: tensor([-0.0021, -0.0007, 0.0003, -0.0016, 0.0007, -0.0004, 0.0019, 0.0002, + 0.0006, 0.0011], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 230.67, cls_loss 0.0305 cls_loss_mapping 0.0520 cls_loss_causal 0.8605 re_mapping 0.0272 re_causal 0.0787 /// teacc 98.66 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0027, -0.0223, -0.0267, ..., -0.0744, 0.0006, 0.0374], + [ 0.0245, 0.0405, 0.0006, ..., 0.0451, -0.0434, -0.0287], + [ 0.0235, 0.0298, -0.0284, ..., 0.0104, -0.0582, 0.0074], + ..., + [-0.0151, -0.0024, 0.0285, ..., -0.0255, 0.0051, 0.0134], + [-0.0070, 0.0451, 0.0563, ..., -0.0334, -0.0293, -0.0416], + [-0.0364, -0.0015, -0.0642, ..., -0.0148, -0.0175, 0.0102]], + device='cuda:0'), grad: tensor([[ 4.3660e-05, 6.6876e-05, 3.6389e-05, ..., 5.9187e-05, + 1.7121e-05, -2.0385e-04], + [-1.8001e-05, -1.2112e-04, -1.4782e-04, ..., -2.5845e-04, + 4.1515e-05, 1.5095e-05], + [-4.6444e-04, -1.1975e-04, -7.8082e-05, ..., 1.3244e-04, + 1.0961e-04, 5.8144e-05], + ..., + [ 1.6797e-04, 2.5654e-04, -5.9545e-05, ..., 2.0015e-04, + 9.5248e-05, -3.2514e-05], + [-6.0749e-04, -3.1781e-04, -3.7289e-04, ..., 8.1301e-05, + 7.0286e-04, 3.9673e-04], + [ 1.2290e-04, 3.8929e-03, 9.5904e-05, ..., 3.3684e-03, + 1.9350e-03, 2.9564e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0059, -0.0181, -0.0160, 0.0295, -0.0322, 0.0239, -0.0296, -0.0159, + 0.0290, -0.0196], device='cuda:0'), grad: tensor([-4.5985e-05, -1.3602e-04, -2.2340e-04, 8.0156e-04, -3.8166e-03, + -1.3971e-03, 4.4751e-04, 1.2708e-04, 2.2686e-04, 4.0169e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 230.52, cls_loss 0.0282 cls_loss_mapping 0.0484 cls_loss_causal 0.8106 re_mapping 0.0256 re_causal 0.0735 /// teacc 98.73 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0031, -0.0230, -0.0272, ..., -0.0752, 0.0006, 0.0382], + [ 0.0249, 0.0407, 0.0016, ..., 0.0456, -0.0445, -0.0294], + [ 0.0234, 0.0295, -0.0296, ..., 0.0102, -0.0587, 0.0073], + ..., + [-0.0153, -0.0029, 0.0290, ..., -0.0264, 0.0048, 0.0133], + [-0.0066, 0.0457, 0.0569, ..., -0.0337, -0.0301, -0.0419], + [-0.0372, -0.0015, -0.0655, ..., -0.0158, -0.0182, 0.0108]], + device='cuda:0'), grad: tensor([[-1.3389e-05, -1.3292e-04, 4.2409e-05, ..., 3.0845e-05, + 7.4625e-05, -1.0700e-03], + [-6.0797e-04, -5.0974e-04, -5.0163e-04, ..., -6.5994e-04, + -2.1085e-05, 6.6698e-05], + [ 2.6673e-05, 1.9169e-04, 1.4102e-04, ..., 5.4359e-05, + 3.2604e-05, 9.3222e-04], + ..., + [-3.0383e-05, 9.5010e-05, -3.6478e-04, ..., 1.1754e-04, + 5.1618e-05, -5.8556e-04], + [-2.0294e-03, -4.4136e-03, -2.3499e-03, ..., 1.4389e-04, + -1.4114e-03, 9.5367e-05], + [ 1.5378e-04, 1.4949e-04, 3.1757e-04, ..., 4.9055e-05, + 7.1883e-05, 2.6226e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0061, -0.0180, -0.0166, 0.0296, -0.0321, 0.0236, -0.0291, -0.0157, + 0.0293, -0.0200], device='cuda:0'), grad: tensor([-0.0009, -0.0009, 0.0010, 0.0025, 0.0003, 0.0018, 0.0014, -0.0007, + -0.0051, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 214.82, cls_loss 0.0240 cls_loss_mapping 0.0469 cls_loss_causal 0.8217 re_mapping 0.0250 re_causal 0.0730 /// teacc 98.59 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0033, -0.0235, -0.0273, ..., -0.0755, 0.0005, 0.0383], + [ 0.0248, 0.0410, 0.0019, ..., 0.0461, -0.0453, -0.0293], + [ 0.0239, 0.0295, -0.0296, ..., 0.0104, -0.0589, 0.0064], + ..., + [-0.0155, -0.0034, 0.0299, ..., -0.0275, 0.0045, 0.0132], + [-0.0065, 0.0463, 0.0574, ..., -0.0339, -0.0306, -0.0426], + [-0.0382, -0.0014, -0.0665, ..., -0.0169, -0.0187, 0.0119]], + device='cuda:0'), grad: tensor([[ 8.8692e-05, 2.4819e-04, 1.1170e-04, ..., 3.7074e-05, + 3.5614e-05, -6.2063e-06], + [-1.1903e-04, -1.9062e-04, -2.5582e-04, ..., -4.1771e-04, + 3.0641e-06, 5.7556e-06], + [ 2.9063e-04, 9.4533e-05, 2.2972e-04, ..., 6.5327e-05, + 1.2636e-05, 6.6012e-06], + ..., + [-1.1778e-04, 9.1851e-05, -2.8849e-04, ..., 1.1230e-04, + 3.1143e-05, 5.0627e-06], + [-7.9215e-05, -6.2132e-04, -3.2711e-04, ..., 3.6716e-05, + 1.5140e-04, 1.7440e-04], + [ 2.7204e-04, 1.9503e-04, 1.2255e-04, ..., 1.1432e-04, + 1.4114e-04, -2.3514e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0061, -0.0182, -0.0164, 0.0296, -0.0321, 0.0235, -0.0293, -0.0155, + 0.0293, -0.0198], device='cuda:0'), grad: tensor([ 2.8181e-04, -2.7990e-04, 4.7016e-04, -6.6805e-04, -4.9710e-05, + -1.5187e-04, 1.7297e-04, -2.0444e-04, -1.1271e-04, 5.4169e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 215.18, cls_loss 0.0294 cls_loss_mapping 0.0497 cls_loss_causal 0.8256 re_mapping 0.0228 re_causal 0.0675 /// teacc 98.58 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0038, -0.0240, -0.0275, ..., -0.0755, 0.0003, 0.0389], + [ 0.0247, 0.0410, 0.0020, ..., 0.0466, -0.0455, -0.0301], + [ 0.0243, 0.0296, -0.0302, ..., 0.0108, -0.0594, 0.0064], + ..., + [-0.0154, -0.0038, 0.0305, ..., -0.0280, 0.0042, 0.0135], + [-0.0059, 0.0469, 0.0585, ..., -0.0342, -0.0313, -0.0434], + [-0.0390, -0.0013, -0.0667, ..., -0.0177, -0.0194, 0.0122]], + device='cuda:0'), grad: tensor([[ 7.3314e-05, 7.9870e-05, 6.1035e-05, ..., 1.4201e-05, + -8.0943e-05, -2.1446e-04], + [-1.7679e-04, -2.0385e-04, -2.1327e-04, ..., -4.1056e-04, + -2.8946e-06, 2.4065e-05], + [ 2.0349e-04, 1.9765e-04, 2.0874e-04, ..., 5.4777e-05, + 1.5162e-05, 8.0585e-05], + ..., + [ 4.0889e-04, 4.4632e-04, 2.6417e-04, ..., 1.9968e-04, + 1.2733e-05, 3.3927e-04], + [-3.0975e-03, -4.1580e-03, -3.4409e-03, ..., -2.7761e-05, + -1.3420e-06, -7.6103e-04], + [ 1.8139e-03, 2.4090e-03, 2.1343e-03, ..., 4.5717e-05, + 1.9908e-05, -7.0035e-06]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0062, -0.0187, -0.0162, 0.0293, -0.0320, 0.0235, -0.0298, -0.0151, + 0.0297, -0.0200], device='cuda:0'), grad: tensor([-4.5627e-05, -1.7953e-04, 4.6706e-04, -4.2009e-04, 4.0603e-04, + 2.0409e-03, 8.2016e-05, 1.0691e-03, -6.9847e-03, 3.5706e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 214.68, cls_loss 0.0243 cls_loss_mapping 0.0449 cls_loss_causal 0.8386 re_mapping 0.0224 re_causal 0.0679 /// teacc 98.68 lr 0.00010000 +Epoch 28, weight, value: tensor([[-4.3355e-03, -2.4791e-02, -2.7566e-02, ..., -7.6454e-02, + -3.7999e-05, 3.9317e-02], + [ 2.5029e-02, 4.1094e-02, 2.8390e-03, ..., 4.6810e-02, + -4.6100e-02, -3.0793e-02], + [ 2.4820e-02, 3.0031e-02, -3.0453e-02, ..., 1.0935e-02, + -6.0216e-02, 6.0827e-03], + ..., + [-1.5795e-02, -4.6244e-03, 3.0911e-02, ..., -2.8623e-02, + 3.8748e-03, 1.3631e-02], + [-6.0238e-03, 4.7136e-02, 5.8948e-02, ..., -3.4919e-02, + -3.1793e-02, -4.3988e-02], + [-3.9808e-02, -1.0863e-03, -6.7694e-02, ..., -1.8067e-02, + -2.0181e-02, 1.2714e-02]], device='cuda:0'), grad: tensor([[ 3.5226e-05, 1.7509e-05, 2.2456e-05, ..., 9.0227e-06, + -8.3089e-05, -5.6028e-04], + [-1.6499e-04, -3.1042e-04, -1.6117e-04, ..., -2.7108e-04, + -2.2590e-05, 3.3855e-05], + [ 9.5749e-04, 1.5736e-04, 8.7070e-04, ..., 5.7995e-05, + 5.1409e-05, 1.7929e-04], + ..., + [-1.5478e-03, -7.2658e-05, -1.4200e-03, ..., 1.1158e-04, + 7.7546e-05, -4.8168e-06], + [-1.8165e-05, -8.6188e-05, 4.3303e-05, ..., 5.9277e-05, + 1.1945e-04, 1.4627e-04], + [-2.9516e-04, -4.3130e-04, 1.2982e-04, ..., 2.0111e-04, + 2.4962e-04, -3.8767e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0061, -0.0187, -0.0159, 0.0297, -0.0322, 0.0237, -0.0296, -0.0151, + 0.0292, -0.0202], device='cuda:0'), grad: tensor([-0.0006, -0.0003, 0.0016, 0.0028, -0.0001, -0.0014, 0.0004, -0.0020, + 0.0005, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 214.94, cls_loss 0.0235 cls_loss_mapping 0.0425 cls_loss_causal 0.7796 re_mapping 0.0227 re_causal 0.0684 /// teacc 98.57 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0048, -0.0256, -0.0280, ..., -0.0769, 0.0006, 0.0400], + [ 0.0254, 0.0412, 0.0035, ..., 0.0475, -0.0468, -0.0314], + [ 0.0247, 0.0302, -0.0312, ..., 0.0105, -0.0606, 0.0058], + ..., + [-0.0160, -0.0054, 0.0310, ..., -0.0292, 0.0038, 0.0134], + [-0.0058, 0.0480, 0.0598, ..., -0.0356, -0.0322, -0.0443], + [-0.0406, -0.0014, -0.0687, ..., -0.0189, -0.0204, 0.0134]], + device='cuda:0'), grad: tensor([[ 3.7313e-05, 6.6698e-05, 2.0877e-05, ..., 4.9531e-05, + -7.5102e-05, -2.1398e-04], + [-3.6806e-05, -3.4034e-05, -5.0962e-05, ..., -7.8082e-05, + 3.6061e-05, 1.6227e-05], + [-2.1820e-03, -5.9605e-05, -1.2083e-03, ..., 3.7432e-05, + 3.1650e-05, 1.8641e-05], + ..., + [ 1.9388e-03, 1.0294e-04, 9.0265e-04, ..., 4.6343e-05, + 2.4900e-05, -2.6631e-04], + [-4.3869e-04, -6.7091e-04, -3.8242e-04, ..., -4.7636e-04, + -2.3043e-04, 5.2780e-05], + [ 6.4611e-05, -5.9791e-06, 1.4567e-04, ..., 2.9755e-04, + 5.1022e-04, 3.5071e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0062, -0.0186, -0.0162, 0.0299, -0.0322, 0.0235, -0.0295, -0.0154, + 0.0294, -0.0201], device='cuda:0'), grad: tensor([-2.3514e-05, 6.0014e-06, -4.6082e-03, 2.6751e-04, 1.5783e-04, + -2.1017e-04, 2.1338e-04, 3.7594e-03, -3.9124e-04, 8.2970e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 214.69, cls_loss 0.0241 cls_loss_mapping 0.0435 cls_loss_causal 0.7899 re_mapping 0.0215 re_causal 0.0635 /// teacc 98.52 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0056, -0.0258, -0.0285, ..., -0.0778, 0.0004, 0.0411], + [ 0.0253, 0.0412, 0.0039, ..., 0.0475, -0.0473, -0.0318], + [ 0.0251, 0.0306, -0.0317, ..., 0.0109, -0.0610, 0.0055], + ..., + [-0.0163, -0.0062, 0.0313, ..., -0.0299, 0.0035, 0.0135], + [-0.0055, 0.0483, 0.0604, ..., -0.0361, -0.0329, -0.0453], + [-0.0415, -0.0015, -0.0698, ..., -0.0196, -0.0210, 0.0139]], + device='cuda:0'), grad: tensor([[ 5.3346e-05, 4.3988e-05, 5.5879e-05, ..., 3.2544e-05, + 2.5809e-05, -4.8399e-05], + [-5.6148e-05, -1.5903e-04, -6.0737e-05, ..., -4.5848e-04, + 2.5973e-05, 1.7717e-05], + [ 1.4699e-04, 1.0145e-04, 1.9836e-04, ..., 5.5224e-05, + 1.2413e-05, 2.8491e-05], + ..., + [-3.2616e-03, -1.3571e-03, -4.5624e-03, ..., 1.5414e-04, + 7.3314e-05, 3.1859e-05], + [ 1.8656e-04, 3.1535e-06, 3.1209e-04, ..., 8.7261e-05, + 9.1791e-05, 2.2650e-04], + [ 1.5678e-03, 4.4894e-04, 1.6584e-03, ..., 4.6706e-04, + 2.6250e-04, -5.5075e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0064, -0.0189, -0.0159, 0.0302, -0.0321, 0.0239, -0.0295, -0.0160, + 0.0290, -0.0203], device='cuda:0'), grad: tensor([ 9.4533e-05, 7.7486e-05, 3.5715e-04, 2.7523e-03, -5.4121e-04, + 4.0507e-04, 3.9268e-04, -7.1640e-03, 8.5831e-04, 2.7733e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 215.02, cls_loss 0.0240 cls_loss_mapping 0.0412 cls_loss_causal 0.7836 re_mapping 0.0216 re_causal 0.0641 /// teacc 98.54 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0062, -0.0265, -0.0289, ..., -0.0778, 0.0006, 0.0414], + [ 0.0255, 0.0417, 0.0043, ..., 0.0480, -0.0480, -0.0321], + [ 0.0254, 0.0305, -0.0323, ..., 0.0107, -0.0616, 0.0056], + ..., + [-0.0155, -0.0068, 0.0326, ..., -0.0302, 0.0032, 0.0140], + [-0.0053, 0.0487, 0.0608, ..., -0.0361, -0.0331, -0.0461], + [-0.0424, -0.0014, -0.0709, ..., -0.0205, -0.0216, 0.0144]], + device='cuda:0'), grad: tensor([[ 5.5164e-05, 5.2154e-05, 3.6210e-05, ..., -6.9797e-05, + 2.8467e-04, -5.7369e-05], + [ 1.2141e-04, 7.6473e-05, 7.3195e-05, ..., 6.7413e-05, + 1.0455e-04, 4.2021e-05], + [ 5.0974e-04, 2.6727e-04, 3.5667e-04, ..., 2.4056e-04, + 7.6890e-05, 5.4359e-05], + ..., + [ 1.9944e-04, 5.2512e-05, 8.9228e-05, ..., 8.4460e-05, + 1.9282e-05, 6.4135e-05], + [-1.1244e-03, -6.2084e-04, -8.8215e-04, ..., -1.3328e-04, + 2.7204e-04, 1.6022e-04], + [ 1.0125e-05, -1.6413e-03, 1.0800e-04, ..., 6.8188e-05, + -4.6849e-05, -4.4708e-03]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0065, -0.0186, -0.0156, 0.0299, -0.0322, 0.0241, -0.0298, -0.0156, + 0.0289, -0.0205], device='cuda:0'), grad: tensor([ 4.8190e-05, 3.2139e-04, 8.9025e-04, 5.3558e-03, 4.7612e-04, + 5.9032e-04, -1.1225e-03, 5.3883e-04, -8.6069e-04, -6.2370e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 214.88, cls_loss 0.0188 cls_loss_mapping 0.0343 cls_loss_causal 0.7565 re_mapping 0.0203 re_causal 0.0621 /// teacc 98.66 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0066, -0.0270, -0.0291, ..., -0.0785, 0.0002, 0.0416], + [ 0.0257, 0.0420, 0.0047, ..., 0.0483, -0.0482, -0.0324], + [ 0.0253, 0.0305, -0.0328, ..., 0.0106, -0.0621, 0.0049], + ..., + [-0.0156, -0.0077, 0.0330, ..., -0.0305, 0.0032, 0.0140], + [-0.0050, 0.0489, 0.0613, ..., -0.0366, -0.0337, -0.0470], + [-0.0431, -0.0012, -0.0718, ..., -0.0212, -0.0222, 0.0153]], + device='cuda:0'), grad: tensor([[ 8.1122e-05, 3.4499e-04, 9.3043e-05, ..., 6.4945e-04, + 1.5318e-04, -1.0061e-03], + [-2.1756e-04, -1.6880e-04, -2.0671e-04, ..., -2.0933e-04, + 1.5581e-04, 3.5137e-05], + [-1.0719e-03, 1.0145e-04, -5.4264e-04, ..., 1.0753e-04, + 1.3578e-04, 5.9754e-05], + ..., + [ 1.0500e-03, 1.5604e-04, 2.7204e-04, ..., 1.2219e-04, + 1.1784e-04, -2.6274e-04], + [-6.1560e-04, -5.8174e-04, -1.1950e-03, ..., 2.2531e-04, + 8.7842e-06, 5.5432e-05], + [ 1.6320e-04, 3.0684e-04, 3.6263e-04, ..., 3.3736e-04, + 6.5041e-04, 6.2037e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0063, -0.0185, -0.0162, 0.0301, -0.0323, 0.0246, -0.0297, -0.0154, + 0.0287, -0.0205], device='cuda:0'), grad: tensor([-0.0005, -0.0002, -0.0011, 0.0003, 0.0007, 0.0012, -0.0020, 0.0011, + -0.0012, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 214.69, cls_loss 0.0244 cls_loss_mapping 0.0459 cls_loss_causal 0.7624 re_mapping 0.0191 re_causal 0.0585 /// teacc 98.65 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0073, -0.0282, -0.0297, ..., -0.0793, -0.0001, 0.0420], + [ 0.0254, 0.0418, 0.0045, ..., 0.0483, -0.0486, -0.0327], + [ 0.0255, 0.0308, -0.0334, ..., 0.0104, -0.0626, 0.0048], + ..., + [-0.0155, -0.0082, 0.0337, ..., -0.0314, 0.0026, 0.0140], + [-0.0046, 0.0496, 0.0624, ..., -0.0368, -0.0345, -0.0481], + [-0.0436, -0.0007, -0.0725, ..., -0.0217, -0.0226, 0.0162]], + device='cuda:0'), grad: tensor([[ 3.7700e-05, 3.1590e-04, 1.4178e-05, ..., 1.5898e-03, + 2.7618e-03, 1.9395e-04], + [-3.6430e-04, -3.0494e-04, -3.6287e-04, ..., -1.8144e-04, + 2.0647e-04, 5.6662e-06], + [-4.6827e-06, 1.5640e-04, 6.0946e-05, ..., 2.3377e-04, + 1.7142e-04, 3.7730e-05], + ..., + [ 9.9123e-05, 1.6296e-04, 8.4877e-05, ..., 1.3256e-04, + 5.0128e-05, 9.6917e-05], + [-2.8268e-05, 6.9976e-05, -4.7535e-05, ..., 1.7524e-04, + 1.0592e-04, 3.0085e-05], + [ 3.7044e-05, -3.7718e-04, 2.9877e-05, ..., -2.3395e-05, + -9.3699e-05, -5.4836e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0060, -0.0189, -0.0163, 0.0301, -0.0323, 0.0245, -0.0294, -0.0153, + 0.0286, -0.0200], device='cuda:0'), grad: tensor([ 0.0025, -0.0002, 0.0003, 0.0004, 0.0021, 0.0005, -0.0052, 0.0003, + 0.0002, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 215.04, cls_loss 0.0163 cls_loss_mapping 0.0285 cls_loss_causal 0.7549 re_mapping 0.0193 re_causal 0.0575 /// teacc 98.34 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0078, -0.0288, -0.0300, ..., -0.0799, -0.0005, 0.0423], + [ 0.0257, 0.0423, 0.0053, ..., 0.0490, -0.0493, -0.0327], + [ 0.0258, 0.0309, -0.0343, ..., 0.0108, -0.0632, 0.0045], + ..., + [-0.0159, -0.0091, 0.0341, ..., -0.0323, 0.0023, 0.0142], + [-0.0042, 0.0503, 0.0634, ..., -0.0377, -0.0353, -0.0486], + [-0.0445, -0.0005, -0.0734, ..., -0.0224, -0.0231, 0.0166]], + device='cuda:0'), grad: tensor([[ 4.2766e-06, -2.0657e-03, 1.3061e-05, ..., -8.7118e-04, + -2.4033e-03, -2.7466e-03], + [ 5.1290e-05, 1.8609e-04, 2.1368e-05, ..., 1.0151e-04, + 1.1832e-04, 6.2108e-05], + [-2.4170e-05, 3.9786e-05, 2.1636e-05, ..., 3.9965e-05, + 5.6058e-05, 2.4295e-04], + ..., + [ 3.8773e-05, 5.9187e-05, 4.0606e-06, ..., 3.7372e-05, + 4.8757e-05, 9.8944e-05], + [ 5.6803e-05, 4.8161e-04, 1.0035e-07, ..., 1.3328e-04, + 3.7789e-04, 2.4247e-04], + [ 3.4142e-04, 3.2496e-04, 5.5462e-05, ..., 2.0373e-04, + 3.4404e-04, 2.4891e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0058, -0.0187, -0.0164, 0.0302, -0.0324, 0.0242, -0.0292, -0.0151, + 0.0288, -0.0203], device='cuda:0'), grad: tensor([-4.9095e-03, 2.4652e-04, 3.2854e-04, -4.9591e-04, -4.3750e-05, + 1.1742e-04, 2.8095e-03, 2.1756e-04, 7.5817e-04, 9.7036e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 231.18, cls_loss 0.0168 cls_loss_mapping 0.0328 cls_loss_causal 0.7739 re_mapping 0.0186 re_causal 0.0582 /// teacc 98.74 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0083, -0.0293, -0.0305, ..., -0.0803, -0.0009, 0.0429], + [ 0.0261, 0.0426, 0.0060, ..., 0.0496, -0.0500, -0.0332], + [ 0.0260, 0.0309, -0.0346, ..., 0.0107, -0.0636, 0.0042], + ..., + [-0.0163, -0.0099, 0.0343, ..., -0.0330, 0.0022, 0.0142], + [-0.0040, 0.0509, 0.0639, ..., -0.0384, -0.0356, -0.0484], + [-0.0451, -0.0003, -0.0742, ..., -0.0225, -0.0236, 0.0165]], + device='cuda:0'), grad: tensor([[ 5.2303e-05, 7.4804e-05, 1.5765e-05, ..., 6.2466e-05, + 7.8321e-05, -2.1052e-04], + [ 1.4246e-04, 5.9319e-04, 3.0726e-05, ..., 7.4387e-04, + 3.8028e-04, 8.3804e-05], + [ 7.0047e-04, 1.3614e-04, 8.2791e-05, ..., 7.4744e-05, + 5.9694e-05, 5.0575e-05], + ..., + [ 6.2799e-04, 6.2883e-05, -5.7220e-05, ..., 5.4479e-05, + 1.3851e-05, -2.8446e-05], + [-6.5899e-04, -6.6566e-04, -4.1080e-04, ..., -1.3614e-04, + 1.0514e-04, 8.5294e-05], + [ 1.6451e-04, 1.0462e-03, 7.0632e-05, ..., 1.3847e-03, + 6.8092e-04, 1.8227e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0059, -0.0184, -0.0163, 0.0303, -0.0322, 0.0239, -0.0292, -0.0155, + 0.0292, -0.0207], device='cuda:0'), grad: tensor([-4.6313e-05, 1.1234e-03, 1.0090e-03, -1.4944e-03, -2.6169e-03, + 1.4269e-04, -5.4866e-05, 6.2704e-04, -6.8235e-04, 1.9894e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 214.52, cls_loss 0.0177 cls_loss_mapping 0.0322 cls_loss_causal 0.7633 re_mapping 0.0190 re_causal 0.0573 /// teacc 98.47 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0090, -0.0298, -0.0306, ..., -0.0799, -0.0004, 0.0437], + [ 0.0264, 0.0428, 0.0068, ..., 0.0505, -0.0505, -0.0338], + [ 0.0267, 0.0309, -0.0349, ..., 0.0105, -0.0641, 0.0037], + ..., + [-0.0161, -0.0104, 0.0346, ..., -0.0335, 0.0020, 0.0144], + [-0.0037, 0.0518, 0.0645, ..., -0.0383, -0.0360, -0.0488], + [-0.0459, -0.0003, -0.0749, ..., -0.0232, -0.0241, 0.0169]], + device='cuda:0'), grad: tensor([[ 8.1122e-05, 7.1526e-05, 3.9250e-05, ..., 1.3411e-04, + 1.0860e-04, -1.5116e-04], + [-1.5011e-03, -7.3099e-04, -8.3113e-04, ..., -1.5106e-03, + 6.4731e-05, 2.5734e-05], + [ 7.7009e-04, 5.4502e-04, 9.3031e-04, ..., 5.9414e-04, + 9.8050e-05, 1.2469e-04], + ..., + [-1.5664e-04, 1.0705e-04, -7.0429e-04, ..., 2.7108e-04, + 4.6343e-05, -5.3257e-05], + [-3.3975e-04, -2.6083e-04, -1.7774e-04, ..., 5.6839e-04, + 9.0778e-05, 6.7532e-05], + [ 3.0804e-04, 2.2805e-04, 1.8287e-04, ..., 2.8086e-04, + 2.1601e-04, 7.0155e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0063, -0.0182, -0.0160, 0.0299, -0.0324, 0.0242, -0.0299, -0.0155, + 0.0295, -0.0209], device='cuda:0'), grad: tensor([ 9.7156e-05, -1.5726e-03, 1.6537e-03, 1.5717e-03, 2.3127e-04, + -7.3576e-04, -8.2541e-04, -6.2180e-04, -7.1001e-04, 9.1124e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 214.87, cls_loss 0.0200 cls_loss_mapping 0.0363 cls_loss_causal 0.7911 re_mapping 0.0184 re_causal 0.0553 /// teacc 98.74 lr 0.00010000 +Epoch 37, weight, value: tensor([[-9.5998e-03, -3.1228e-02, -3.1322e-02, ..., -8.0318e-02, + -2.4818e-05, 4.3616e-02], + [ 2.6448e-02, 4.2757e-02, 6.7754e-03, ..., 5.0934e-02, + -5.1067e-02, -3.3524e-02], + [ 2.6664e-02, 3.0788e-02, -3.5675e-02, ..., 1.0049e-02, + -6.4840e-02, 2.8858e-03], + ..., + [-1.6137e-02, -1.0560e-02, 3.5503e-02, ..., -3.4136e-02, + 2.0247e-03, 1.4273e-02], + [-3.4047e-03, 5.2185e-02, 6.5332e-02, ..., -3.8659e-02, + -3.6651e-02, -5.0084e-02], + [-4.6303e-02, 1.9000e-05, -7.5602e-02, ..., -2.3930e-02, + -2.4843e-02, 1.8312e-02]], device='cuda:0'), grad: tensor([[ 7.7188e-05, 3.4243e-05, 7.0035e-05, ..., 2.4468e-05, + 1.8314e-05, -3.4779e-05], + [-1.5408e-05, -2.1100e-04, -5.7817e-05, ..., -3.0375e-04, + 1.0923e-05, 7.1041e-06], + [ 1.3649e-04, 2.3198e-04, 1.0719e-03, ..., 4.1723e-05, + 2.0042e-05, 3.8803e-05], + ..., + [ 1.0178e-02, -4.7660e-04, 1.0986e-02, ..., 6.8188e-05, + 2.1264e-05, -1.0473e-04], + [ 6.9094e-04, 1.8764e-04, 9.2506e-04, ..., 2.6464e-04, + 3.0947e-04, 3.1650e-05], + [ 1.3697e-04, 1.2584e-05, 1.3149e-04, ..., 3.8296e-05, + 2.7239e-05, -7.2122e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0060, -0.0185, -0.0164, 0.0301, -0.0324, 0.0246, -0.0300, -0.0153, + 0.0293, -0.0206], device='cuda:0'), grad: tensor([ 1.4198e-04, 1.2837e-05, 7.8678e-04, -2.7466e-02, 3.0828e-04, + -8.2970e-04, 1.6975e-04, 2.4887e-02, 1.8702e-03, 1.3423e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 215.16, cls_loss 0.0184 cls_loss_mapping 0.0312 cls_loss_causal 0.7204 re_mapping 0.0181 re_causal 0.0518 /// teacc 98.74 lr 0.00010000 +Epoch 38, weight, value: tensor([[-1.0111e-02, -3.2068e-02, -3.2052e-02, ..., -8.1040e-02, + -1.8171e-04, 4.4021e-02], + [ 2.6635e-02, 4.2824e-02, 7.2321e-03, ..., 5.1268e-02, + -5.2195e-02, -3.3956e-02], + [ 2.7244e-02, 3.1234e-02, -3.6081e-02, ..., 9.9591e-03, + -6.5384e-02, 2.3150e-03], + ..., + [-1.6545e-02, -1.1047e-02, 3.5818e-02, ..., -3.5005e-02, + 1.6204e-03, 1.4801e-02], + [-3.5565e-03, 5.2415e-02, 6.5866e-02, ..., -3.9234e-02, + -3.6967e-02, -5.0758e-02], + [-4.6745e-02, -3.0989e-05, -7.6173e-02, ..., -2.4561e-02, + -2.5270e-02, 1.8507e-02]], device='cuda:0'), grad: tensor([[ 6.4075e-05, 7.6890e-05, 8.5592e-05, ..., 8.6606e-05, + -5.9837e-07, -2.5243e-05], + [-1.9226e-03, -1.9588e-03, -2.4624e-03, ..., -3.2120e-03, + -2.6375e-05, 2.4348e-05], + [ 7.2122e-05, 1.4305e-04, 2.3997e-04, ..., 3.3283e-04, + 7.8976e-06, 4.0233e-05], + ..., + [ 6.0588e-05, 1.3864e-04, -1.1021e-04, ..., 9.9778e-05, + 1.7241e-05, -2.0540e-04], + [ 8.6260e-04, 8.9931e-04, 1.0433e-03, ..., 1.8082e-03, + 1.4775e-05, 6.1154e-05], + [ 1.0371e-04, -1.0004e-03, 2.2495e-04, ..., -7.4863e-04, + -7.0751e-05, -1.4696e-03]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0058, -0.0185, -0.0161, 0.0300, -0.0323, 0.0243, -0.0296, -0.0150, + 0.0291, -0.0208], device='cuda:0'), grad: tensor([ 0.0001, -0.0034, 0.0003, 0.0004, 0.0026, 0.0005, 0.0005, -0.0002, + 0.0013, -0.0021], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 215.28, cls_loss 0.0183 cls_loss_mapping 0.0345 cls_loss_causal 0.7599 re_mapping 0.0179 re_causal 0.0532 /// teacc 98.70 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0107, -0.0330, -0.0322, ..., -0.0815, -0.0002, 0.0443], + [ 0.0269, 0.0427, 0.0078, ..., 0.0514, -0.0530, -0.0357], + [ 0.0277, 0.0313, -0.0367, ..., 0.0095, -0.0658, 0.0018], + ..., + [-0.0171, -0.0117, 0.0363, ..., -0.0351, 0.0013, 0.0151], + [-0.0034, 0.0527, 0.0666, ..., -0.0400, -0.0374, -0.0519], + [-0.0471, 0.0006, -0.0771, ..., -0.0245, -0.0258, 0.0195]], + device='cuda:0'), grad: tensor([[ 4.1872e-05, 3.2306e-05, -1.6227e-05, ..., 1.8150e-05, + 2.5064e-05, -1.5509e-04], + [ 1.0595e-03, 1.7762e-04, -7.3195e-05, ..., 1.4162e-04, + 1.9178e-05, 5.3740e-04], + [-1.4191e-03, 5.2124e-05, 9.5665e-05, ..., -1.0532e-04, + 5.3167e-05, -7.4148e-04], + ..., + [ 1.0180e-04, 1.0383e-04, -5.3024e-04, ..., 7.0155e-05, + 3.1948e-05, 1.4162e-04], + [ 1.8811e-04, -1.6129e-04, 4.2692e-06, ..., 1.3936e-04, + 5.6982e-04, 2.4486e-04], + [ 1.3471e-04, 2.1648e-04, 9.2268e-05, ..., 2.5749e-04, + 1.2946e-04, -1.5736e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0057, -0.0189, -0.0158, 0.0299, -0.0319, 0.0245, -0.0298, -0.0150, + 0.0288, -0.0206], device='cuda:0'), grad: tensor([-6.0767e-05, 1.0004e-03, -7.8964e-04, 1.1263e-03, -8.6546e-04, + -1.5783e-03, 2.8586e-04, -3.7813e-04, 8.6784e-04, 3.9196e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 232.66, cls_loss 0.0199 cls_loss_mapping 0.0351 cls_loss_causal 0.7496 re_mapping 0.0177 re_causal 0.0536 /// teacc 98.76 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0110, -0.0338, -0.0320, ..., -0.0824, -0.0002, 0.0449], + [ 0.0269, 0.0424, 0.0081, ..., 0.0509, -0.0535, -0.0374], + [ 0.0280, 0.0317, -0.0373, ..., 0.0093, -0.0662, 0.0017], + ..., + [-0.0169, -0.0117, 0.0366, ..., -0.0358, 0.0008, 0.0152], + [-0.0034, 0.0527, 0.0676, ..., -0.0401, -0.0380, -0.0519], + [-0.0477, 0.0013, -0.0783, ..., -0.0239, -0.0261, 0.0204]], + device='cuda:0'), grad: tensor([[ 7.8297e-04, 1.9932e-03, 3.6865e-05, ..., 7.0810e-04, + 9.4473e-05, -8.1211e-06], + [-2.4605e-04, -8.9765e-05, -2.4772e-04, ..., -2.5630e-04, + 1.2137e-05, -3.1083e-07], + [ 1.2302e-04, 1.7965e-04, 1.6797e-04, ..., 1.2922e-04, + 2.6047e-05, 8.9034e-06], + ..., + [ 1.2541e-04, 9.7096e-05, 3.9250e-05, ..., 7.9572e-05, + 1.6615e-05, 9.4697e-06], + [-1.0338e-03, -2.5215e-03, -1.1855e-04, ..., -8.4734e-04, + 1.9789e-04, 7.4387e-05], + [ 4.2170e-05, 6.6817e-05, 2.5138e-05, ..., 5.8204e-05, + 7.6771e-05, 5.0843e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0057, -0.0194, -0.0160, 0.0304, -0.0317, 0.0241, -0.0301, -0.0148, + 0.0289, -0.0204], device='cuda:0'), grad: tensor([ 0.0036, -0.0003, 0.0002, 0.0002, 0.0001, -0.0003, 0.0002, 0.0003, + -0.0043, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 231.60, cls_loss 0.0163 cls_loss_mapping 0.0324 cls_loss_causal 0.7365 re_mapping 0.0172 re_causal 0.0531 /// teacc 98.79 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0109, -0.0339, -0.0323, ..., -0.0822, -0.0006, 0.0456], + [ 0.0269, 0.0424, 0.0085, ..., 0.0511, -0.0544, -0.0370], + [ 0.0283, 0.0317, -0.0378, ..., 0.0092, -0.0666, 0.0015], + ..., + [-0.0168, -0.0122, 0.0372, ..., -0.0360, 0.0006, 0.0151], + [-0.0030, 0.0533, 0.0683, ..., -0.0403, -0.0385, -0.0530], + [-0.0489, 0.0010, -0.0797, ..., -0.0246, -0.0267, 0.0206]], + device='cuda:0'), grad: tensor([[ 2.0294e-03, 1.0920e-03, 1.2971e-05, ..., 5.7995e-05, + 2.0182e-04, 9.9850e-04], + [-8.2397e-04, -5.1498e-03, -5.5695e-03, ..., -5.7335e-03, + -2.3499e-05, -2.6131e-03], + [-2.7122e-03, -2.5868e-04, 4.6134e-05, ..., 6.8367e-05, + 3.9935e-05, -3.7241e-04], + ..., + [ 2.5940e-04, 1.3399e-03, 7.3862e-04, ..., 1.5011e-03, + 4.7624e-05, -8.0729e-04], + [-6.8009e-05, -7.8964e-04, 2.3472e-04, ..., 2.6155e-04, + -9.9778e-05, -6.9094e-04], + [ 5.1546e-04, 3.1452e-03, 3.6430e-03, ..., 3.7804e-03, + 1.0147e-03, 3.6812e-03]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0063, -0.0193, -0.0162, 0.0302, -0.0318, 0.0244, -0.0297, -0.0148, + 0.0289, -0.0210], device='cuda:0'), grad: tensor([ 0.0055, -0.0098, -0.0040, 0.0006, -0.0006, 0.0002, 0.0002, 0.0006, + -0.0023, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 215.13, cls_loss 0.0169 cls_loss_mapping 0.0307 cls_loss_causal 0.7116 re_mapping 0.0174 re_causal 0.0520 /// teacc 98.77 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0116, -0.0346, -0.0328, ..., -0.0826, -0.0010, 0.0460], + [ 0.0269, 0.0426, 0.0088, ..., 0.0516, -0.0550, -0.0372], + [ 0.0286, 0.0316, -0.0382, ..., 0.0087, -0.0672, 0.0012], + ..., + [-0.0171, -0.0126, 0.0380, ..., -0.0368, 0.0002, 0.0164], + [-0.0027, 0.0537, 0.0689, ..., -0.0410, -0.0390, -0.0544], + [-0.0488, 0.0015, -0.0809, ..., -0.0254, -0.0268, 0.0206]], + device='cuda:0'), grad: tensor([[ 2.3887e-05, 2.4334e-05, 2.1830e-05, ..., -2.1219e-05, + -8.0526e-05, -1.2207e-04], + [-3.5787e-04, -4.5466e-04, -7.6866e-04, ..., -5.9843e-04, + -4.0280e-08, 3.2224e-06], + [-1.6022e-03, -8.1491e-04, -1.1158e-03, ..., 9.9897e-05, + 2.0221e-05, 1.7226e-05], + ..., + [ 1.6994e-03, 1.0672e-03, 1.5249e-03, ..., 3.5405e-04, + 5.2527e-06, 1.2077e-05], + [ 1.6689e-05, 3.6098e-06, 2.6494e-05, ..., 5.1975e-05, + 4.1366e-05, -4.5747e-05], + [ 5.2363e-05, 3.4004e-05, 5.2631e-05, ..., 2.0519e-05, + 1.8299e-05, 1.6406e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0060, -0.0193, -0.0161, 0.0298, -0.0317, 0.0243, -0.0297, -0.0141, + 0.0289, -0.0211], device='cuda:0'), grad: tensor([-1.3828e-04, -1.0519e-03, -2.2163e-03, 1.2410e-04, 1.2422e-04, + 3.8099e-04, -6.1810e-05, 2.6569e-03, 3.8356e-05, 1.4389e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 41---------------------------------------------------- +epoch 41, time 231.37, cls_loss 0.0153 cls_loss_mapping 0.0282 cls_loss_causal 0.6977 re_mapping 0.0163 re_causal 0.0505 /// teacc 98.86 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0120, -0.0356, -0.0334, ..., -0.0834, -0.0008, 0.0462], + [ 0.0271, 0.0433, 0.0099, ..., 0.0518, -0.0557, -0.0374], + [ 0.0291, 0.0321, -0.0382, ..., 0.0095, -0.0678, 0.0011], + ..., + [-0.0178, -0.0134, 0.0380, ..., -0.0377, -0.0001, 0.0162], + [-0.0024, 0.0538, 0.0691, ..., -0.0418, -0.0393, -0.0545], + [-0.0494, 0.0018, -0.0817, ..., -0.0260, -0.0274, 0.0210]], + device='cuda:0'), grad: tensor([[ 2.6345e-05, 6.5446e-05, 3.3170e-05, ..., 1.1492e-04, + 7.8976e-05, -6.8426e-05], + [-2.7442e-04, -1.8942e-04, -3.4308e-04, ..., -6.1941e-04, + 1.1124e-05, 1.1124e-05], + [ 2.5630e-04, 4.5037e-04, 2.7823e-04, ..., 4.7469e-04, + 2.6870e-04, 3.4690e-05], + ..., + [ 7.4267e-05, 9.3162e-05, -4.7255e-04, ..., -2.1136e-04, + -1.8191e-04, -5.4717e-05], + [-6.5851e-04, -1.2369e-03, -5.4646e-04, ..., 7.7128e-05, + -2.8729e-04, 4.6253e-05], + [ 5.7906e-05, 9.7156e-05, 1.8322e-04, ..., 1.3447e-04, + 1.0335e-04, 5.1439e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0060, -0.0191, -0.0157, 0.0304, -0.0322, 0.0239, -0.0295, -0.0145, + 0.0289, -0.0213], device='cuda:0'), grad: tensor([ 1.1224e-04, -5.7077e-04, 1.0548e-03, 4.7112e-04, 6.8617e-04, + 3.5584e-05, 5.6314e-04, -8.5831e-04, -1.9627e-03, 4.6897e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.76, cls_loss 0.0154 cls_loss_mapping 0.0299 cls_loss_causal 0.6804 re_mapping 0.0163 re_causal 0.0474 /// teacc 98.85 lr 0.00010000 +Epoch 44, weight, value: tensor([[-1.2396e-02, -3.6518e-02, -3.3394e-02, ..., -8.3373e-02, + -1.3638e-03, 4.5714e-02], + [ 2.6665e-02, 4.2753e-02, 9.6181e-03, ..., 5.2228e-02, + -5.6279e-02, -3.7944e-02], + [ 2.9197e-02, 3.2610e-02, -3.8158e-02, ..., 9.7132e-03, + -6.8514e-02, 4.8080e-04], + ..., + [-1.7907e-02, -1.3572e-02, 3.8605e-02, ..., -3.8024e-02, + -4.6799e-05, 1.6866e-02], + [-1.8463e-03, 5.4187e-02, 6.9907e-02, ..., -4.2407e-02, + -3.9806e-02, -5.5476e-02], + [-5.0160e-02, 2.3720e-03, -8.2402e-02, ..., -2.6779e-02, + -2.7807e-02, 2.2074e-02]], device='cuda:0'), grad: tensor([[ 1.5482e-05, 8.3506e-05, 1.9386e-05, ..., 1.5631e-05, + 1.4913e-04, -2.8920e-04], + [-5.3215e-04, -4.2295e-04, -1.0023e-03, ..., -8.9502e-04, + 1.9446e-05, 6.8173e-06], + [-1.4877e-04, 5.9247e-05, 1.4186e-04, ..., 2.0802e-04, + 1.3307e-05, 1.6820e-04], + ..., + [ 5.1785e-04, 3.6645e-04, 5.7220e-04, ..., 5.9319e-04, + 5.5164e-05, 2.1458e-05], + [ 2.9892e-05, -5.5507e-07, 1.4473e-06, ..., 6.8367e-05, + 6.2644e-05, 5.5850e-05], + [ 6.7115e-05, 1.6248e-04, 5.0753e-05, ..., 3.2401e-04, + 2.5058e-04, 6.1989e-06]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0054, -0.0196, -0.0158, 0.0305, -0.0325, 0.0241, -0.0295, -0.0141, + 0.0290, -0.0209], device='cuda:0'), grad: tensor([-1.5390e-04, -1.3065e-03, 2.5702e-04, 1.3804e-04, -5.1594e-04, + -1.9252e-04, 2.5500e-06, 1.0881e-03, 1.8573e-04, 4.9686e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 214.86, cls_loss 0.0187 cls_loss_mapping 0.0324 cls_loss_causal 0.7650 re_mapping 0.0166 re_causal 0.0485 /// teacc 98.86 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0136, -0.0380, -0.0341, ..., -0.0840, -0.0016, 0.0461], + [ 0.0266, 0.0426, 0.0097, ..., 0.0527, -0.0571, -0.0382], + [ 0.0298, 0.0327, -0.0388, ..., 0.0100, -0.0690, 0.0009], + ..., + [-0.0182, -0.0137, 0.0391, ..., -0.0388, -0.0005, 0.0169], + [-0.0016, 0.0548, 0.0710, ..., -0.0430, -0.0409, -0.0564], + [-0.0510, 0.0018, -0.0834, ..., -0.0280, -0.0286, 0.0222]], + device='cuda:0'), grad: tensor([[ 9.9018e-06, 7.2047e-06, 1.4529e-05, ..., 2.2016e-06, + -1.8291e-06, -4.6158e-04], + [ 3.6461e-07, 3.9756e-05, 4.8429e-05, ..., -3.6001e-05, + 1.7598e-05, 9.0718e-05], + [ 1.6010e-04, 1.5831e-04, 5.6553e-04, ..., 1.5348e-05, + 1.0893e-05, 3.9190e-05], + ..., + [-3.3903e-04, -1.7893e-04, -1.2026e-03, ..., 3.0443e-05, + 1.8388e-05, 4.6313e-05], + [ 1.0151e-04, 2.2864e-04, 3.9053e-04, ..., 3.9190e-05, + 4.4525e-05, 2.4843e-04], + [-1.5593e-04, -6.8092e-04, 1.1295e-04, ..., 2.3580e-04, + 2.2399e-04, -1.3599e-03]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0048, -0.0198, -0.0153, 0.0306, -0.0322, 0.0245, -0.0294, -0.0140, + 0.0287, -0.0215], device='cuda:0'), grad: tensor([-0.0006, 0.0002, 0.0006, 0.0018, -0.0007, 0.0003, 0.0001, -0.0011, + 0.0009, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 215.12, cls_loss 0.0159 cls_loss_mapping 0.0289 cls_loss_causal 0.7095 re_mapping 0.0163 re_causal 0.0465 /// teacc 98.80 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0139, -0.0386, -0.0347, ..., -0.0845, -0.0016, 0.0466], + [ 0.0270, 0.0425, 0.0101, ..., 0.0529, -0.0575, -0.0390], + [ 0.0297, 0.0321, -0.0403, ..., 0.0102, -0.0695, 0.0007], + ..., + [-0.0181, -0.0137, 0.0402, ..., -0.0390, -0.0009, 0.0166], + [-0.0014, 0.0553, 0.0714, ..., -0.0437, -0.0413, -0.0571], + [-0.0515, 0.0019, -0.0836, ..., -0.0287, -0.0292, 0.0229]], + device='cuda:0'), grad: tensor([[ 3.8773e-05, 3.1531e-05, 3.2187e-06, ..., 2.5317e-05, + -5.9032e-04, -1.1950e-03], + [ 8.3596e-06, 9.1717e-06, -2.7940e-05, ..., -3.9190e-06, + 2.7031e-05, 2.9027e-05], + [ 9.2387e-05, 5.3674e-05, 3.0503e-05, ..., 1.0067e-04, + 1.8072e-04, 2.9349e-04], + ..., + [ 5.4628e-05, 9.1076e-05, 2.7776e-05, ..., 9.8944e-05, + 9.0420e-05, 1.3709e-04], + [-5.1677e-05, -2.6032e-05, -1.8573e-04, ..., 5.2541e-05, + 1.4901e-04, 3.9315e-04], + [ 6.9082e-05, 6.8784e-05, 1.3225e-05, ..., 1.1885e-04, + 2.9063e-04, -3.2043e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0048, -0.0200, -0.0158, 0.0303, -0.0319, 0.0251, -0.0297, -0.0136, + 0.0285, -0.0213], device='cuda:0'), grad: tensor([-2.0370e-03, 1.1426e-04, 9.5510e-04, -9.6703e-04, -2.6631e-04, + 6.5041e-04, 4.7898e-04, 4.8089e-04, 6.5231e-04, -6.5982e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 215.21, cls_loss 0.0128 cls_loss_mapping 0.0251 cls_loss_causal 0.6841 re_mapping 0.0161 re_causal 0.0475 /// teacc 98.71 lr 0.00010000 +Epoch 47, weight, value: tensor([[-1.4242e-02, -3.9043e-02, -3.4831e-02, ..., -8.4665e-02, + -1.3678e-03, 4.7049e-02], + [ 2.7770e-02, 4.3270e-02, 1.1276e-02, ..., 5.4427e-02, + -5.7738e-02, -3.9174e-02], + [ 3.0257e-02, 3.2167e-02, -4.0397e-02, ..., 9.3126e-03, + -7.0274e-02, -8.2971e-06], + ..., + [-1.8472e-02, -1.4219e-02, 4.0401e-02, ..., -3.9918e-02, + -8.9771e-04, 1.6664e-02], + [-2.1364e-03, 5.5181e-02, 7.1218e-02, ..., -4.4232e-02, + -4.2041e-02, -5.8093e-02], + [-5.2141e-02, 1.7271e-03, -8.4349e-02, ..., -2.9315e-02, + -2.9778e-02, 2.3907e-02]], device='cuda:0'), grad: tensor([[ 1.7703e-05, 1.6138e-05, 9.5144e-06, ..., 4.7758e-06, + 1.9357e-05, -6.3896e-05], + [ 2.7823e-04, 8.1003e-05, 2.3365e-05, ..., 8.9645e-05, + 1.2882e-05, 6.0678e-05], + [-4.0364e-04, -1.1694e-04, 2.8178e-05, ..., -1.0228e-04, + 1.2137e-05, -4.4048e-05], + ..., + [ 8.5413e-05, 7.5161e-05, -5.0664e-05, ..., 5.0992e-05, + 3.5167e-05, 2.3708e-05], + [-3.2902e-04, -2.4939e-04, -4.7684e-04, ..., 5.2340e-06, + -6.6400e-05, 7.7605e-05], + [ 2.6926e-05, -1.2124e-04, 7.4878e-06, ..., 3.3736e-05, + 3.1084e-05, -1.6356e-04]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0049, -0.0192, -0.0159, 0.0304, -0.0322, 0.0249, -0.0291, -0.0138, + 0.0277, -0.0212], device='cuda:0'), grad: tensor([-2.8446e-05, 3.0065e-04, -3.2759e-04, 2.4390e-04, 1.8731e-05, + 1.9357e-05, 4.3893e-04, 1.0550e-04, -5.4646e-04, -2.2423e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 215.21, cls_loss 0.0150 cls_loss_mapping 0.0249 cls_loss_causal 0.6937 re_mapping 0.0156 re_causal 0.0448 /// teacc 98.85 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0149, -0.0403, -0.0355, ..., -0.0855, -0.0012, 0.0479], + [ 0.0280, 0.0432, 0.0119, ..., 0.0547, -0.0580, -0.0401], + [ 0.0312, 0.0327, -0.0403, ..., 0.0093, -0.0708, -0.0006], + ..., + [-0.0191, -0.0149, 0.0404, ..., -0.0409, -0.0013, 0.0168], + [-0.0023, 0.0557, 0.0717, ..., -0.0444, -0.0423, -0.0584], + [-0.0529, 0.0019, -0.0854, ..., -0.0298, -0.0306, 0.0245]], + device='cuda:0'), grad: tensor([[ 1.6615e-05, 5.4359e-05, 9.7305e-06, ..., 6.7651e-05, + 1.9765e-04, 8.2135e-05], + [-6.8009e-05, 1.6764e-05, -6.9082e-05, ..., 2.8208e-05, + 1.4997e-04, 4.6521e-05], + [-9.5248e-05, 7.4916e-06, 9.7990e-05, ..., 5.0724e-05, + 1.6987e-04, 4.0144e-05], + ..., + [ 5.1320e-05, 9.1314e-05, 3.6567e-05, ..., 1.0699e-04, + 1.0538e-04, 5.7280e-05], + [ 2.0683e-05, 5.9366e-05, -2.3589e-05, ..., 1.2350e-04, + 3.2306e-04, 2.5272e-04], + [ 2.8536e-05, 4.2945e-05, 5.8740e-05, ..., 1.0622e-04, + 2.2876e-04, 2.9147e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0049, -0.0191, -0.0152, 0.0305, -0.0319, 0.0247, -0.0293, -0.0144, + 0.0279, -0.0213], device='cuda:0'), grad: tensor([ 0.0003, 0.0001, -0.0002, 0.0006, 0.0004, -0.0018, -0.0006, 0.0003, + 0.0007, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 47---------------------------------------------------- +epoch 47, time 230.24, cls_loss 0.0102 cls_loss_mapping 0.0188 cls_loss_causal 0.6851 re_mapping 0.0150 re_causal 0.0473 /// teacc 98.91 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0151, -0.0409, -0.0357, ..., -0.0858, -0.0009, 0.0486], + [ 0.0282, 0.0432, 0.0120, ..., 0.0550, -0.0585, -0.0407], + [ 0.0310, 0.0325, -0.0409, ..., 0.0089, -0.0714, -0.0013], + ..., + [-0.0191, -0.0153, 0.0408, ..., -0.0416, -0.0022, 0.0167], + [-0.0019, 0.0564, 0.0725, ..., -0.0445, -0.0429, -0.0587], + [-0.0539, 0.0021, -0.0865, ..., -0.0302, -0.0298, 0.0253]], + device='cuda:0'), grad: tensor([[ 8.1360e-06, 2.2963e-05, 7.3463e-06, ..., 2.7671e-05, + 6.6698e-05, 9.4622e-06], + [-1.5765e-05, -2.0340e-05, -3.6031e-05, ..., -2.9206e-05, + 2.3097e-05, 1.4305e-05], + [ 3.7760e-05, 1.6451e-05, 1.7852e-05, ..., 2.9176e-05, + 2.3276e-05, 6.9663e-06], + ..., + [ 3.3855e-05, 2.9683e-05, -2.0186e-07, ..., 2.0310e-05, + 1.8492e-05, 1.5628e-06], + [-1.1790e-04, 5.6505e-05, -2.1899e-04, ..., 3.9428e-05, + 1.5032e-04, 1.8513e-04], + [ 4.5657e-05, -1.6308e-04, 3.7640e-05, ..., 3.2306e-05, + 6.2466e-05, -1.2052e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0054, -0.0192, -0.0159, 0.0306, -0.0320, 0.0251, -0.0299, -0.0146, + 0.0281, -0.0210], device='cuda:0'), grad: tensor([ 9.1136e-05, 1.5259e-05, 9.7215e-05, 5.7966e-05, 8.4162e-05, + -2.8777e-04, -1.3089e-04, 6.2287e-05, 1.1367e-04, -1.0341e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 48---------------------------------------------------- +epoch 48, time 231.47, cls_loss 0.0120 cls_loss_mapping 0.0224 cls_loss_causal 0.6668 re_mapping 0.0151 re_causal 0.0438 /// teacc 98.96 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0156, -0.0417, -0.0361, ..., -0.0861, -0.0016, 0.0485], + [ 0.0281, 0.0429, 0.0121, ..., 0.0546, -0.0590, -0.0420], + [ 0.0312, 0.0324, -0.0411, ..., 0.0087, -0.0719, -0.0018], + ..., + [-0.0196, -0.0158, 0.0411, ..., -0.0416, -0.0021, 0.0171], + [-0.0013, 0.0572, 0.0733, ..., -0.0442, -0.0434, -0.0589], + [-0.0540, 0.0026, -0.0876, ..., -0.0305, -0.0305, 0.0256]], + device='cuda:0'), grad: tensor([[ 4.2804e-06, 7.8529e-06, 3.0659e-06, ..., 4.8690e-06, + 6.4299e-06, -4.1723e-05], + [-5.1171e-05, -7.0274e-05, -1.1063e-04, ..., -1.0401e-04, + -6.9067e-06, 3.0790e-06], + [ 1.2450e-05, 8.3447e-06, 6.0499e-06, ..., 1.4409e-05, + 1.2651e-05, 4.8615e-06], + ..., + [ 1.9476e-05, 2.9534e-05, 2.1651e-05, ..., 3.9160e-05, + 4.1157e-05, 1.4395e-05], + [ 2.1458e-05, 3.7402e-05, 2.1175e-05, ..., 2.3931e-05, + 6.5267e-05, 3.9965e-05], + [ 3.9607e-05, -2.6894e-04, 1.3337e-05, ..., 1.7956e-05, + -1.9479e-04, -3.1114e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0046, -0.0199, -0.0159, 0.0302, -0.0322, 0.0257, -0.0295, -0.0142, + 0.0286, -0.0211], device='cuda:0'), grad: tensor([-1.3560e-05, -1.4305e-04, 4.1217e-05, -3.4034e-05, 7.5769e-04, + -1.8072e-04, 2.8580e-05, 1.0383e-04, 1.5950e-04, -7.1955e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 214.73, cls_loss 0.0110 cls_loss_mapping 0.0231 cls_loss_causal 0.6769 re_mapping 0.0144 re_causal 0.0438 /// teacc 98.85 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0161, -0.0422, -0.0361, ..., -0.0863, -0.0013, 0.0494], + [ 0.0285, 0.0435, 0.0130, ..., 0.0557, -0.0594, -0.0423], + [ 0.0316, 0.0328, -0.0414, ..., 0.0087, -0.0724, -0.0021], + ..., + [-0.0202, -0.0169, 0.0406, ..., -0.0435, -0.0024, 0.0171], + [-0.0012, 0.0575, 0.0737, ..., -0.0448, -0.0439, -0.0593], + [-0.0549, 0.0026, -0.0888, ..., -0.0311, -0.0310, 0.0258]], + device='cuda:0'), grad: tensor([[ 4.1217e-05, 8.1718e-05, 1.5900e-05, ..., 1.5080e-04, + 1.6713e-04, -3.3736e-04], + [ 1.2028e-04, 7.5102e-05, 1.0151e-04, ..., 1.5192e-05, + 2.4438e-05, 1.3582e-05], + [-3.1614e-04, -2.3913e-04, -7.9930e-05, ..., 3.4422e-05, + 3.1322e-05, 7.0930e-05], + ..., + [ 1.7023e-04, 1.0014e-04, 3.4496e-06, ..., 1.7375e-05, + 5.3376e-05, 6.7055e-05], + [-5.1051e-05, -5.5879e-05, -2.5296e-04, ..., -8.8453e-05, + 1.0139e-04, 1.1641e-04], + [ 5.1737e-05, 9.3281e-06, 2.8327e-05, ..., 1.2405e-05, + 2.5010e-04, 1.9014e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0052, -0.0193, -0.0158, 0.0308, -0.0321, 0.0255, -0.0297, -0.0149, + 0.0286, -0.0214], device='cuda:0'), grad: tensor([ 4.8205e-06, 2.3007e-04, -8.8120e-04, -5.4598e-05, 2.8658e-04, + -8.1539e-04, -9.4235e-05, 4.3917e-04, 4.4107e-04, 4.4489e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 215.03, cls_loss 0.0099 cls_loss_mapping 0.0207 cls_loss_causal 0.6505 re_mapping 0.0143 re_causal 0.0424 /// teacc 98.86 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0165, -0.0428, -0.0363, ..., -0.0870, -0.0013, 0.0503], + [ 0.0286, 0.0435, 0.0129, ..., 0.0559, -0.0599, -0.0426], + [ 0.0317, 0.0326, -0.0420, ..., 0.0088, -0.0730, -0.0022], + ..., + [-0.0197, -0.0169, 0.0419, ..., -0.0431, -0.0021, 0.0169], + [-0.0009, 0.0580, 0.0742, ..., -0.0450, -0.0446, -0.0604], + [-0.0561, 0.0026, -0.0899, ..., -0.0317, -0.0313, 0.0265]], + device='cuda:0'), grad: tensor([[ 2.0301e-04, 1.4746e-04, 2.9832e-05, ..., 2.1875e-05, + 8.2374e-05, 6.5744e-05], + [-1.5162e-06, -1.2286e-05, -5.8293e-05, ..., -7.1406e-05, + 4.0263e-05, 2.0757e-05], + [-2.5082e-03, -9.1076e-04, -1.5986e-04, ..., 9.8825e-05, + 6.5744e-05, 1.9506e-05], + ..., + [ 8.5211e-04, 1.1975e-04, -1.3375e-04, ..., 1.7273e-04, + 8.2433e-05, -4.7255e-04], + [ 4.7255e-04, 2.1529e-04, -9.3579e-05, ..., 3.3230e-05, + 6.5625e-05, 4.9710e-05], + [ 1.1152e-04, 8.9467e-05, 2.6011e-04, ..., 5.9158e-05, + 7.4685e-05, 3.4070e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0055, -0.0195, -0.0159, 0.0308, -0.0328, 0.0251, -0.0296, -0.0140, + 0.0284, -0.0213], device='cuda:0'), grad: tensor([ 4.6229e-04, 5.5104e-05, -3.4904e-03, 1.3523e-03, -3.1495e-04, + -9.9301e-05, 1.6456e-06, 5.6458e-04, 8.2397e-04, 6.4611e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 215.07, cls_loss 0.0090 cls_loss_mapping 0.0197 cls_loss_causal 0.6784 re_mapping 0.0137 re_causal 0.0434 /// teacc 98.95 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0170, -0.0432, -0.0368, ..., -0.0875, -0.0009, 0.0508], + [ 0.0289, 0.0437, 0.0133, ..., 0.0563, -0.0602, -0.0425], + [ 0.0319, 0.0327, -0.0426, ..., 0.0090, -0.0734, -0.0026], + ..., + [-0.0201, -0.0171, 0.0422, ..., -0.0440, -0.0025, 0.0169], + [-0.0008, 0.0582, 0.0747, ..., -0.0456, -0.0451, -0.0612], + [-0.0568, 0.0030, -0.0911, ..., -0.0319, -0.0312, 0.0272]], + device='cuda:0'), grad: tensor([[ 1.5989e-05, 1.5962e-04, 8.0585e-05, ..., 9.1642e-06, + 5.6811e-06, 1.6665e-04], + [-2.5058e-04, -1.3435e-04, -1.9765e-04, ..., -2.0564e-04, + 8.6576e-06, 2.6822e-05], + [-5.9158e-06, 1.5652e-04, 1.5283e-04, ..., 4.0174e-04, + 1.9121e-04, -4.0472e-05], + ..., + [ 1.3721e-04, 1.3757e-04, -2.0772e-05, ..., 4.4286e-05, + -2.9159e-04, -4.6253e-04], + [ 3.5971e-05, -1.0977e-03, -7.4053e-04, ..., 2.3022e-05, + 1.3791e-05, -1.1263e-03], + [ 9.5904e-05, 5.0688e-04, 4.8923e-04, ..., -3.6098e-06, + 5.5999e-05, 6.2227e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0056, -0.0193, -0.0158, 0.0310, -0.0328, 0.0251, -0.0302, -0.0140, + 0.0281, -0.0211], device='cuda:0'), grad: tensor([ 4.1008e-04, -2.6584e-04, 2.3115e-04, -2.0218e-04, 9.5272e-04, + 5.2023e-04, 3.5375e-05, -7.9489e-04, -2.4509e-03, 1.5640e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 215.01, cls_loss 0.0088 cls_loss_mapping 0.0171 cls_loss_causal 0.6465 re_mapping 0.0139 re_causal 0.0410 /// teacc 98.94 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0175, -0.0439, -0.0372, ..., -0.0879, -0.0012, 0.0510], + [ 0.0301, 0.0448, 0.0141, ..., 0.0572, -0.0606, -0.0428], + [ 0.0313, 0.0316, -0.0437, ..., 0.0080, -0.0740, -0.0028], + ..., + [-0.0202, -0.0177, 0.0427, ..., -0.0448, -0.0026, 0.0170], + [-0.0006, 0.0586, 0.0752, ..., -0.0463, -0.0453, -0.0614], + [-0.0575, 0.0029, -0.0921, ..., -0.0327, -0.0315, 0.0277]], + device='cuda:0'), grad: tensor([[ 1.0421e-06, 4.6156e-06, 4.9099e-06, ..., 2.2538e-06, + -1.6326e-06, -8.9779e-06], + [-5.5842e-06, 1.9550e-05, -4.2212e-07, ..., -1.4380e-05, + 6.9365e-06, 3.9428e-05], + [-2.6124e-07, 4.5709e-06, 1.3866e-05, ..., 4.6939e-06, + 3.5204e-06, 9.9987e-06], + ..., + [ 1.1772e-06, 2.4378e-05, -1.0920e-04, ..., 1.0200e-05, + 3.8892e-06, -9.9003e-05], + [-2.3794e-04, -3.9315e-04, -5.7554e-04, ..., -3.3188e-04, + -2.0492e-04, 7.3969e-05], + [ 4.9211e-06, -9.4414e-05, 5.5432e-05, ..., 1.0625e-05, + 8.4341e-06, -9.0301e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0053, -0.0182, -0.0169, 0.0310, -0.0328, 0.0249, -0.0297, -0.0139, + 0.0281, -0.0213], device='cuda:0'), grad: tensor([ 5.9558e-07, 3.0756e-05, 2.2307e-05, 3.0965e-05, 5.0604e-05, + 6.8843e-05, 3.5977e-04, -1.4400e-04, -3.5071e-04, -6.8486e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 215.09, cls_loss 0.0095 cls_loss_mapping 0.0203 cls_loss_causal 0.6717 re_mapping 0.0141 re_causal 0.0407 /// teacc 98.88 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0177, -0.0441, -0.0373, ..., -0.0880, -0.0017, 0.0512], + [ 0.0298, 0.0445, 0.0141, ..., 0.0568, -0.0611, -0.0430], + [ 0.0318, 0.0323, -0.0439, ..., 0.0090, -0.0744, -0.0029], + ..., + [-0.0202, -0.0181, 0.0432, ..., -0.0453, -0.0028, 0.0166], + [-0.0005, 0.0588, 0.0755, ..., -0.0468, -0.0459, -0.0625], + [-0.0581, 0.0032, -0.0925, ..., -0.0332, -0.0318, 0.0285]], + device='cuda:0'), grad: tensor([[ 5.5730e-06, 1.3947e-05, 4.8839e-06, ..., 7.9349e-06, + 3.7979e-06, -1.9133e-05], + [-1.3128e-05, 2.6263e-06, -1.9297e-05, ..., -8.2105e-06, + 2.0117e-05, 5.7071e-06], + [ 2.0146e-05, 3.3915e-05, 5.8591e-05, ..., 2.2978e-05, + 1.6093e-05, 6.1333e-05], + ..., + [-7.2792e-06, 3.1650e-05, -9.9361e-05, ..., 3.4273e-05, + 2.3782e-05, -1.2147e-04], + [-9.1851e-05, -1.3900e-04, -5.8532e-05, ..., 1.2055e-05, + 2.7865e-05, 8.0094e-06], + [ 5.0783e-05, 7.3671e-05, 5.5194e-05, ..., 1.8924e-05, + 2.3231e-05, 4.3094e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0053, -0.0189, -0.0162, 0.0310, -0.0327, 0.0250, -0.0297, -0.0139, + 0.0276, -0.0211], device='cuda:0'), grad: tensor([ 2.6766e-06, 4.1425e-06, 1.7011e-04, -1.7852e-05, -9.8825e-05, + -3.1114e-05, 8.5533e-05, -1.9777e-04, -9.8884e-05, 1.8144e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 215.40, cls_loss 0.0094 cls_loss_mapping 0.0190 cls_loss_causal 0.6525 re_mapping 0.0140 re_causal 0.0430 /// teacc 98.83 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0180, -0.0446, -0.0375, ..., -0.0881, -0.0017, 0.0514], + [ 0.0299, 0.0442, 0.0146, ..., 0.0565, -0.0619, -0.0441], + [ 0.0317, 0.0320, -0.0448, ..., 0.0087, -0.0748, -0.0033], + ..., + [-0.0202, -0.0186, 0.0436, ..., -0.0458, -0.0032, 0.0163], + [-0.0006, 0.0589, 0.0758, ..., -0.0471, -0.0464, -0.0633], + [-0.0584, 0.0033, -0.0931, ..., -0.0331, -0.0321, 0.0290]], + device='cuda:0'), grad: tensor([[ 1.7658e-05, 4.4294e-06, 2.4825e-05, ..., 6.5565e-06, + -4.6223e-05, -1.5330e-04], + [ 1.3113e-03, -1.1258e-05, 3.9558e-03, ..., -1.7032e-05, + 1.2398e-05, 8.3828e-04], + [-9.7811e-05, 8.1733e-06, -2.4345e-06, ..., 1.0654e-05, + 1.1079e-05, -2.8402e-05], + ..., + [-1.4963e-03, 2.5705e-05, -4.8370e-03, ..., 3.4958e-05, + 3.2544e-05, -1.0433e-03], + [ 7.0035e-05, 1.6332e-05, 1.6093e-04, ..., 1.1444e-05, + 3.6985e-05, 1.0997e-04], + [ 4.6283e-05, 1.9148e-05, 2.8896e-04, ..., 1.0115e-04, + 1.4746e-04, 2.7597e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0052, -0.0192, -0.0169, 0.0311, -0.0318, 0.0258, -0.0296, -0.0138, + 0.0272, -0.0212], device='cuda:0'), grad: tensor([-1.1939e-04, 4.7112e-03, -1.2684e-04, 5.5647e-04, -2.2113e-04, + 2.7800e-07, 3.0413e-05, -5.8022e-03, 3.3665e-04, 6.3467e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 215.01, cls_loss 0.0095 cls_loss_mapping 0.0207 cls_loss_causal 0.6416 re_mapping 0.0134 re_causal 0.0387 /// teacc 98.95 lr 0.00010000 +Epoch 57, weight, value: tensor([[-1.8379e-02, -4.5490e-02, -3.7972e-02, ..., -8.8597e-02, + -1.7580e-03, 5.1718e-02], + [ 3.0103e-02, 4.4690e-02, 1.4979e-02, ..., 5.7536e-02, + -6.2475e-02, -4.4123e-02], + [ 3.1884e-02, 3.2076e-02, -4.5331e-02, ..., 8.2651e-03, + -7.5247e-02, -3.5751e-03], + ..., + [-2.0220e-02, -1.9835e-02, 4.3834e-02, ..., -4.6941e-02, + -3.3882e-03, 1.6225e-02], + [-8.5864e-05, 5.9917e-02, 7.7632e-02, ..., -4.7578e-02, + -4.6818e-02, -6.3885e-02], + [-5.9354e-02, 3.4117e-03, -9.5307e-02, ..., -3.4152e-02, + -3.2611e-02, 2.9759e-02]], device='cuda:0'), grad: tensor([[ 2.0713e-05, 2.6330e-05, 1.9252e-05, ..., 1.9595e-05, + 2.1666e-05, -4.4107e-05], + [-4.0817e-04, -2.5463e-04, -4.1270e-04, ..., -4.0603e-04, + 1.7226e-05, 7.7546e-05], + [ 1.4529e-05, 2.2009e-05, 1.0943e-04, ..., 2.0337e-04, + 5.7787e-05, 3.5256e-05], + ..., + [ 1.9741e-04, 2.0361e-04, 1.6689e-04, ..., 1.7309e-04, + 3.1382e-05, 3.5596e-04], + [ 9.3699e-05, 5.5981e-04, 1.8227e-04, ..., 4.6939e-05, + 1.1009e-04, 5.9414e-04], + [ 2.3365e-05, -6.5231e-04, -1.5998e-04, ..., 2.4974e-05, + 4.6760e-05, -1.1072e-03]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0051, -0.0190, -0.0172, 0.0310, -0.0320, 0.0253, -0.0295, -0.0138, + 0.0280, -0.0214], device='cuda:0'), grad: tensor([ 3.1918e-05, -4.5896e-04, 7.9453e-05, 1.1885e-04, 1.7273e-04, + -6.0606e-04, 3.1090e-04, 9.8324e-04, 1.1129e-03, -1.7471e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 56---------------------------------------------------- +epoch 56, time 231.81, cls_loss 0.0073 cls_loss_mapping 0.0154 cls_loss_causal 0.6423 re_mapping 0.0132 re_causal 0.0395 /// teacc 98.99 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0186, -0.0459, -0.0382, ..., -0.0889, -0.0017, 0.0521], + [ 0.0303, 0.0448, 0.0153, ..., 0.0580, -0.0629, -0.0446], + [ 0.0323, 0.0321, -0.0453, ..., 0.0079, -0.0757, -0.0037], + ..., + [-0.0208, -0.0205, 0.0439, ..., -0.0474, -0.0036, 0.0161], + [ 0.0001, 0.0601, 0.0780, ..., -0.0478, -0.0477, -0.0649], + [-0.0596, 0.0040, -0.0959, ..., -0.0346, -0.0329, 0.0304]], + device='cuda:0'), grad: tensor([[-7.1168e-05, -1.5473e-04, -3.7372e-05, ..., 6.7167e-06, + 1.6227e-05, -3.7050e-04], + [ 1.1422e-05, -1.6347e-05, -2.4512e-05, ..., -3.3140e-05, + 9.9614e-06, 3.2902e-05], + [ 1.0616e-04, 1.1081e-04, 5.3376e-05, ..., 1.3053e-05, + 1.6734e-05, 1.2207e-04], + ..., + [ 4.2588e-05, 1.1481e-05, -6.4731e-05, ..., 9.5740e-06, + 6.8545e-06, 3.3706e-05], + [-5.3227e-05, -1.4916e-05, -4.6164e-05, ..., 1.6555e-05, + 5.3525e-05, 1.2207e-04], + [ 4.7833e-05, -2.4438e-04, 6.1750e-05, ..., 7.7784e-06, + -4.3839e-05, -2.5606e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0054, -0.0191, -0.0168, 0.0310, -0.0320, 0.0256, -0.0297, -0.0143, + 0.0277, -0.0211], device='cuda:0'), grad: tensor([-5.0020e-04, 1.2147e-04, 4.1151e-04, -8.3399e-04, 3.9387e-04, + 3.7050e-04, -1.7002e-05, 1.5700e-04, 2.8896e-04, -3.9196e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 215.21, cls_loss 0.0107 cls_loss_mapping 0.0213 cls_loss_causal 0.6895 re_mapping 0.0129 re_causal 0.0401 /// teacc 98.89 lr 0.00010000 +Epoch 59, weight, value: tensor([[-1.9487e-02, -4.7681e-02, -3.8944e-02, ..., -9.0730e-02, + -2.3453e-03, 5.2544e-02], + [ 3.0224e-02, 4.5037e-02, 1.5254e-02, ..., 5.8157e-02, + -6.3600e-02, -4.4661e-02], + [ 3.3061e-02, 3.2428e-02, -4.5605e-02, ..., 8.2104e-03, + -7.6339e-02, -4.1266e-03], + ..., + [-2.0608e-02, -1.9524e-02, 4.5335e-02, ..., -4.7856e-02, + -3.8880e-03, 1.6375e-02], + [ 9.5812e-05, 5.9923e-02, 7.8367e-02, ..., -4.8292e-02, + -4.8293e-02, -6.5755e-02], + [-6.0527e-02, 3.9531e-03, -9.7806e-02, ..., -3.5054e-02, + -3.3623e-02, 3.0858e-02]], device='cuda:0'), grad: tensor([[ 9.3207e-06, 1.8016e-05, 1.0006e-05, ..., 1.0647e-05, + 3.6180e-05, -2.4006e-05], + [ 5.3704e-05, 1.0931e-04, 1.9777e-04, ..., 1.9044e-05, + 3.5882e-05, 2.7612e-05], + [ 5.2482e-05, 3.1471e-05, 5.8144e-05, ..., 3.9369e-05, + 5.8770e-05, 3.6985e-05], + ..., + [-1.7866e-05, -6.4254e-05, -2.8872e-04, ..., 8.0615e-06, + 4.1932e-05, -2.5600e-05], + [-1.0204e-04, -6.8188e-05, -1.9038e-04, ..., 1.1349e-04, + 4.8184e-04, 3.0756e-04], + [ 7.0035e-05, 8.4460e-05, 9.3341e-05, ..., 8.5294e-05, + 2.7919e-04, 1.9169e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0050, -0.0191, -0.0159, 0.0308, -0.0321, 0.0255, -0.0295, -0.0139, + 0.0273, -0.0214], device='cuda:0'), grad: tensor([ 4.1634e-05, 3.3689e-04, 1.9610e-04, 3.8910e-04, 3.5572e-04, + -3.0289e-03, 9.4652e-04, -3.0661e-04, 4.6134e-04, 6.0463e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 214.59, cls_loss 0.0089 cls_loss_mapping 0.0195 cls_loss_causal 0.6592 re_mapping 0.0126 re_causal 0.0395 /// teacc 98.89 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0204, -0.0484, -0.0397, ..., -0.0920, -0.0028, 0.0531], + [ 0.0310, 0.0456, 0.0166, ..., 0.0584, -0.0643, -0.0449], + [ 0.0331, 0.0324, -0.0463, ..., 0.0080, -0.0772, -0.0047], + ..., + [-0.0213, -0.0204, 0.0451, ..., -0.0482, -0.0043, 0.0162], + [ 0.0001, 0.0602, 0.0789, ..., -0.0491, -0.0494, -0.0667], + [-0.0617, 0.0041, -0.0994, ..., -0.0353, -0.0345, 0.0313]], + device='cuda:0'), grad: tensor([[ 1.8273e-06, 1.3247e-05, 5.6811e-06, ..., 1.1273e-05, + 2.4781e-05, 4.0121e-06], + [-5.6736e-06, -2.1178e-06, 3.0976e-06, ..., -5.8040e-06, + 7.5847e-06, 9.2015e-06], + [-4.3027e-06, 2.6867e-05, 8.9630e-06, ..., 4.4517e-07, + 9.6336e-06, 3.8743e-05], + ..., + [ 3.8017e-06, 5.6960e-06, -6.7234e-05, ..., 6.1430e-06, + 6.9365e-06, -7.0930e-05], + [ 1.1340e-05, 1.4400e-04, 4.0621e-05, ..., 9.7096e-05, + 3.8123e-04, 2.9778e-04], + [-5.3160e-06, -2.0385e-04, 2.3022e-05, ..., -2.6941e-05, + -4.7415e-05, -1.9753e-04]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0048, -0.0183, -0.0161, 0.0310, -0.0323, 0.0267, -0.0300, -0.0145, + 0.0268, -0.0216], device='cuda:0'), grad: tensor([ 2.6330e-05, 1.1817e-05, 4.7922e-05, -2.7463e-05, 2.4509e-04, + -9.6416e-04, 5.7793e-04, -7.4148e-05, 4.4417e-04, -2.8682e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 215.11, cls_loss 0.0069 cls_loss_mapping 0.0150 cls_loss_causal 0.6185 re_mapping 0.0123 re_causal 0.0392 /// teacc 98.91 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0206, -0.0488, -0.0400, ..., -0.0922, -0.0027, 0.0537], + [ 0.0309, 0.0453, 0.0167, ..., 0.0581, -0.0648, -0.0455], + [ 0.0333, 0.0322, -0.0469, ..., 0.0081, -0.0778, -0.0048], + ..., + [-0.0214, -0.0203, 0.0451, ..., -0.0488, -0.0044, 0.0157], + [ 0.0003, 0.0604, 0.0791, ..., -0.0497, -0.0500, -0.0678], + [-0.0621, 0.0044, -0.0999, ..., -0.0355, -0.0348, 0.0320]], + device='cuda:0'), grad: tensor([[ 3.0603e-06, 9.4697e-06, 3.2578e-06, ..., 4.9695e-06, + 5.4613e-06, -8.0019e-06], + [-3.0577e-05, -4.8161e-05, -6.9439e-05, ..., -5.2691e-05, + 3.6117e-06, 3.0585e-06], + [ 3.6322e-06, -2.5090e-06, 1.2964e-05, ..., 5.3421e-06, + 3.9227e-06, 3.7048e-06], + ..., + [ 4.9919e-06, 1.9148e-05, -3.4779e-05, ..., 5.4054e-06, + 2.2929e-06, 3.7253e-06], + [-6.4969e-06, 2.8443e-04, 5.3823e-05, ..., 1.9953e-05, + 3.9697e-05, 1.6904e-04], + [ 1.7926e-05, -3.4118e-04, -4.7356e-05, ..., 4.1053e-06, + 4.8131e-06, -1.6785e-04]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0050, -0.0188, -0.0161, 0.0318, -0.0324, 0.0268, -0.0297, -0.0150, + 0.0265, -0.0215], device='cuda:0'), grad: tensor([ 2.4661e-06, -5.8591e-05, 2.1383e-05, -1.0066e-05, 5.8234e-05, + -1.4782e-04, 1.4961e-04, -3.8743e-06, 5.6791e-04, -5.7840e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 214.90, cls_loss 0.0097 cls_loss_mapping 0.0215 cls_loss_causal 0.6051 re_mapping 0.0127 re_causal 0.0363 /// teacc 98.72 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0214, -0.0497, -0.0405, ..., -0.0941, -0.0035, 0.0532], + [ 0.0316, 0.0457, 0.0176, ..., 0.0589, -0.0650, -0.0460], + [ 0.0334, 0.0321, -0.0482, ..., 0.0078, -0.0783, -0.0058], + ..., + [-0.0218, -0.0202, 0.0455, ..., -0.0490, -0.0048, 0.0160], + [ 0.0011, 0.0611, 0.0801, ..., -0.0508, -0.0503, -0.0681], + [-0.0631, 0.0041, -0.1014, ..., -0.0358, -0.0352, 0.0329]], + device='cuda:0'), grad: tensor([[ 5.7705e-06, 8.0764e-06, 4.8801e-06, ..., 2.5332e-06, + -3.3388e-07, -2.2560e-05], + [-4.9882e-06, -9.2899e-07, -2.2382e-05, ..., -3.6627e-05, + 5.6848e-06, 5.4762e-07], + [ 6.3300e-05, 5.9158e-05, 1.0937e-04, ..., 1.0081e-05, + 1.8790e-05, 1.9446e-06], + ..., + [ 9.3505e-06, 2.0063e-04, 3.4451e-05, ..., 1.7083e-04, + 1.3340e-04, 1.2498e-06], + [-9.9659e-05, -6.4194e-05, -1.0860e-04, ..., 2.6166e-05, + 6.7055e-05, 2.9728e-06], + [ 1.3359e-05, 6.5744e-05, 3.8236e-05, ..., 5.0753e-05, + 4.4912e-05, 1.2152e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0040, -0.0187, -0.0164, 0.0314, -0.0325, 0.0273, -0.0299, -0.0145, + 0.0272, -0.0217], device='cuda:0'), grad: tensor([-3.2131e-06, -1.0498e-05, 1.4985e-04, 1.8537e-04, -5.3406e-04, + -4.2224e-04, 2.1672e-04, 3.2091e-04, -4.4823e-05, 1.4150e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 214.95, cls_loss 0.0127 cls_loss_mapping 0.0220 cls_loss_causal 0.6529 re_mapping 0.0139 re_causal 0.0366 /// teacc 98.75 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0224, -0.0508, -0.0415, ..., -0.0945, -0.0040, 0.0526], + [ 0.0310, 0.0454, 0.0175, ..., 0.0590, -0.0653, -0.0467], + [ 0.0338, 0.0322, -0.0488, ..., 0.0077, -0.0793, -0.0065], + ..., + [-0.0221, -0.0203, 0.0465, ..., -0.0496, -0.0050, 0.0167], + [ 0.0020, 0.0620, 0.0810, ..., -0.0512, -0.0513, -0.0692], + [-0.0645, 0.0040, -0.1032, ..., -0.0364, -0.0352, 0.0341]], + device='cuda:0'), grad: tensor([[-3.4213e-05, -8.6248e-05, -2.4751e-05, ..., 4.7088e-06, + 1.1148e-06, -2.0540e-04], + [-3.0845e-05, -1.8761e-05, -2.7359e-05, ..., -4.8369e-05, + 7.3612e-06, 2.7090e-05], + [ 6.8486e-05, 6.7890e-05, 8.1420e-05, ..., 9.7379e-06, + 1.3329e-05, 1.2124e-04], + ..., + [-2.5177e-04, 1.5378e-05, -5.7602e-04, ..., 1.3843e-05, + 5.2676e-06, -9.8407e-05], + [ 1.6257e-05, 2.4587e-06, -3.5781e-06, ..., 1.1258e-05, + 1.3947e-04, 1.7035e-04], + [ 4.3988e-05, 1.9342e-05, 6.7651e-05, ..., 5.4948e-06, + 1.8626e-05, 5.1320e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0029, -0.0194, -0.0167, 0.0328, -0.0325, 0.0276, -0.0303, -0.0142, + 0.0273, -0.0219], device='cuda:0'), grad: tensor([-2.4509e-04, 1.0081e-05, 2.8372e-04, 7.6914e-04, 6.6340e-05, + -9.1493e-05, -2.0242e-04, -1.1425e-03, 3.5810e-04, 1.9324e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.98, cls_loss 0.0088 cls_loss_mapping 0.0179 cls_loss_causal 0.6209 re_mapping 0.0132 re_causal 0.0372 /// teacc 98.97 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0227, -0.0526, -0.0405, ..., -0.0941, -0.0042, 0.0519], + [ 0.0315, 0.0455, 0.0179, ..., 0.0594, -0.0664, -0.0470], + [ 0.0337, 0.0318, -0.0498, ..., 0.0069, -0.0799, -0.0067], + ..., + [-0.0222, -0.0207, 0.0472, ..., -0.0507, -0.0056, 0.0179], + [ 0.0022, 0.0624, 0.0815, ..., -0.0515, -0.0516, -0.0699], + [-0.0646, 0.0043, -0.1049, ..., -0.0370, -0.0356, 0.0350]], + device='cuda:0'), grad: tensor([[ 1.4067e-05, 4.8205e-06, 7.6964e-06, ..., 3.8370e-06, + -2.7850e-05, -1.8609e-04], + [-1.5378e-04, -8.3923e-05, -2.3675e-04, ..., -2.0361e-04, + 3.1851e-06, 8.6576e-06], + [ 4.9531e-05, 6.4038e-06, -1.4037e-05, ..., 3.3677e-05, + 4.6417e-06, 4.7117e-05], + ..., + [ 3.5191e-04, 1.2177e-04, 1.0920e-04, ..., 9.7513e-05, + 1.6317e-05, 9.0003e-05], + [ 1.8084e-04, 6.6519e-05, 6.5088e-05, ..., 4.0859e-05, + 9.2238e-06, 4.1008e-05], + [-1.6928e-05, -3.5453e-04, 1.2204e-05, ..., -5.3674e-05, + -8.1062e-05, -3.3498e-04]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0025, -0.0193, -0.0170, 0.0323, -0.0326, 0.0274, -0.0298, -0.0135, + 0.0274, -0.0219], device='cuda:0'), grad: tensor([-1.9133e-04, -2.4891e-04, 2.1958e-04, -5.4502e-04, 5.2309e-04, + 1.2100e-04, 5.5641e-05, 6.5613e-04, 3.1805e-04, -9.0933e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 215.14, cls_loss 0.0062 cls_loss_mapping 0.0130 cls_loss_causal 0.6330 re_mapping 0.0121 re_causal 0.0380 /// teacc 98.95 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0226, -0.0527, -0.0407, ..., -0.0930, -0.0033, 0.0529], + [ 0.0321, 0.0459, 0.0186, ..., 0.0603, -0.0668, -0.0474], + [ 0.0338, 0.0318, -0.0501, ..., 0.0065, -0.0804, -0.0074], + ..., + [-0.0226, -0.0212, 0.0474, ..., -0.0512, -0.0058, 0.0177], + [ 0.0023, 0.0627, 0.0817, ..., -0.0519, -0.0517, -0.0705], + [-0.0649, 0.0047, -0.1052, ..., -0.0375, -0.0361, 0.0355]], + device='cuda:0'), grad: tensor([[ 9.8124e-06, 1.4611e-05, 7.7710e-06, ..., 1.7777e-05, + 1.8388e-05, -3.2689e-07], + [-2.1562e-05, 5.3458e-07, -2.1517e-05, ..., 3.9972e-06, + 3.5018e-05, 1.2554e-05], + [ 1.3016e-05, 4.3571e-05, 5.8681e-05, ..., 2.6360e-05, + 1.5333e-05, 9.2462e-06], + ..., + [-8.6725e-05, 5.3495e-05, -1.9503e-04, ..., 6.9439e-05, + 4.9263e-05, -6.7316e-06], + [-7.2420e-05, -9.4235e-05, -9.2745e-05, ..., 4.2439e-05, + 2.2158e-05, 5.9791e-06], + [ 6.1214e-05, 6.3848e-04, 6.9618e-05, ..., 1.0576e-03, + 7.8630e-04, 2.9063e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0039, -0.0189, -0.0173, 0.0319, -0.0329, 0.0268, -0.0298, -0.0137, + 0.0273, -0.0215], device='cuda:0'), grad: tensor([ 4.9680e-05, 6.3598e-05, 1.4746e-04, 1.6820e-04, -1.8950e-03, + 1.0175e-04, -1.1873e-04, -3.0851e-04, -8.3566e-05, 1.8740e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 64---------------------------------------------------- +epoch 64, time 231.92, cls_loss 0.0089 cls_loss_mapping 0.0174 cls_loss_causal 0.6318 re_mapping 0.0115 re_causal 0.0345 /// teacc 99.01 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0228, -0.0529, -0.0407, ..., -0.0928, -0.0035, 0.0531], + [ 0.0323, 0.0457, 0.0195, ..., 0.0601, -0.0673, -0.0483], + [ 0.0358, 0.0331, -0.0498, ..., 0.0068, -0.0812, -0.0085], + ..., + [-0.0231, -0.0218, 0.0473, ..., -0.0527, -0.0064, 0.0174], + [ 0.0007, 0.0625, 0.0816, ..., -0.0521, -0.0521, -0.0709], + [-0.0650, 0.0041, -0.1057, ..., -0.0382, -0.0365, 0.0368]], + device='cuda:0'), grad: tensor([[ 4.0740e-05, 2.6301e-05, 4.9546e-06, ..., 2.4170e-05, + 2.5108e-05, -8.7202e-05], + [-2.2892e-06, 2.6356e-06, -1.6525e-05, ..., 6.8545e-05, + 3.5346e-05, -1.0207e-06], + [ 4.7827e-04, 1.0633e-03, 9.5892e-04, ..., 2.0206e-05, + 1.0118e-05, -4.6730e-05], + ..., + [ 1.5223e-04, 9.7871e-05, 4.9621e-05, ..., 1.7738e-04, + 6.2108e-05, 6.1616e-06], + [-7.3242e-04, -1.1358e-03, -1.0481e-03, ..., 1.9282e-05, + 2.5988e-05, 6.6698e-05], + [ 3.3557e-05, -6.5081e-06, 2.4810e-05, ..., 8.7380e-05, + 3.6448e-05, 1.3500e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0041, -0.0189, -0.0165, 0.0316, -0.0321, 0.0269, -0.0299, -0.0142, + 0.0265, -0.0213], device='cuda:0'), grad: tensor([ 2.9311e-05, 1.5390e-04, 1.5202e-03, 4.8690e-06, -6.2609e-04, + 1.1754e-04, -6.2466e-05, 6.2656e-04, -1.9817e-03, 2.1744e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 214.96, cls_loss 0.0062 cls_loss_mapping 0.0109 cls_loss_causal 0.6301 re_mapping 0.0119 re_causal 0.0361 /// teacc 98.98 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0231, -0.0531, -0.0411, ..., -0.0932, -0.0031, 0.0538], + [ 0.0324, 0.0458, 0.0196, ..., 0.0605, -0.0676, -0.0485], + [ 0.0361, 0.0330, -0.0508, ..., 0.0066, -0.0818, -0.0086], + ..., + [-0.0229, -0.0218, 0.0480, ..., -0.0532, -0.0067, 0.0173], + [ 0.0008, 0.0627, 0.0821, ..., -0.0526, -0.0527, -0.0718], + [-0.0655, 0.0039, -0.1067, ..., -0.0393, -0.0376, 0.0370]], + device='cuda:0'), grad: tensor([[ 6.6049e-06, 4.2953e-06, 4.1239e-06, ..., 1.8515e-06, + 2.1160e-06, 6.0303e-07], + [ 1.9968e-06, 3.9488e-06, -3.2019e-06, ..., -2.4904e-06, + 9.1642e-06, 2.0135e-06], + [ 5.6237e-05, 3.3379e-05, 2.6971e-05, ..., 3.1423e-06, + 2.2016e-06, 4.4890e-06], + ..., + [ 6.9797e-05, 3.5167e-05, -1.9684e-05, ..., 3.5822e-05, + 2.3365e-05, -1.1936e-05], + [-6.5625e-05, -7.5758e-05, -2.9996e-05, ..., 5.3942e-06, + 1.1221e-05, 1.3495e-06], + [ 1.7121e-05, 9.0957e-05, 3.0905e-05, ..., 1.5688e-04, + 1.6701e-04, 1.9073e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0049, -0.0190, -0.0167, 0.0312, -0.0321, 0.0270, -0.0298, -0.0138, + 0.0262, -0.0217], device='cuda:0'), grad: tensor([ 1.5065e-05, 2.6539e-05, 8.1420e-05, -1.9002e-04, -4.9353e-04, + 3.4034e-05, 4.3929e-05, 1.1617e-04, -3.6716e-05, 4.0364e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.95, cls_loss 0.0092 cls_loss_mapping 0.0173 cls_loss_causal 0.6648 re_mapping 0.0120 re_causal 0.0357 /// teacc 98.93 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0239, -0.0538, -0.0418, ..., -0.0935, -0.0035, 0.0536], + [ 0.0322, 0.0457, 0.0198, ..., 0.0598, -0.0693, -0.0485], + [ 0.0361, 0.0324, -0.0522, ..., 0.0066, -0.0827, -0.0091], + ..., + [-0.0232, -0.0224, 0.0484, ..., -0.0540, -0.0072, 0.0174], + [ 0.0018, 0.0640, 0.0834, ..., -0.0525, -0.0532, -0.0726], + [-0.0666, 0.0046, -0.1077, ..., -0.0395, -0.0371, 0.0382]], + device='cuda:0'), grad: tensor([[ 1.1669e-06, 3.9823e-06, 4.1537e-06, ..., 2.8927e-06, + -1.8477e-06, -1.8701e-05], + [-8.0690e-06, -3.5595e-06, -1.0937e-05, ..., -9.0823e-06, + 4.9993e-06, 5.6326e-06], + [-9.1270e-06, 3.4031e-06, 2.1178e-06, ..., 9.2164e-06, + 6.9402e-06, 3.0603e-06], + ..., + [ 8.6948e-06, 7.5139e-06, 8.8289e-06, ..., 7.3761e-06, + 4.5970e-06, 2.0429e-05], + [ 4.0270e-06, 5.9068e-05, 2.8744e-05, ..., 5.3458e-06, + 1.4472e-04, 2.7990e-04], + [ 2.6207e-06, 5.4277e-06, -1.3411e-07, ..., 1.6645e-05, + 4.2856e-05, 2.9892e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0045, -0.0196, -0.0172, 0.0314, -0.0322, 0.0265, -0.0293, -0.0136, + 0.0270, -0.0213], device='cuda:0'), grad: tensor([-1.6242e-06, 1.2834e-06, 4.4107e-06, 1.4615e-04, -3.9011e-05, + -9.0313e-04, 4.3511e-05, 8.8394e-05, 6.0034e-04, 6.0230e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.94, cls_loss 0.0062 cls_loss_mapping 0.0141 cls_loss_causal 0.6216 re_mapping 0.0122 re_causal 0.0367 /// teacc 98.91 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0245, -0.0543, -0.0424, ..., -0.0939, -0.0034, 0.0539], + [ 0.0327, 0.0463, 0.0208, ..., 0.0603, -0.0694, -0.0480], + [ 0.0363, 0.0325, -0.0525, ..., 0.0065, -0.0830, -0.0094], + ..., + [-0.0236, -0.0228, 0.0485, ..., -0.0544, -0.0074, 0.0175], + [ 0.0018, 0.0642, 0.0834, ..., -0.0532, -0.0535, -0.0737], + [-0.0670, 0.0040, -0.1087, ..., -0.0407, -0.0376, 0.0386]], + device='cuda:0'), grad: tensor([[-6.3479e-05, -2.1851e-04, 1.3202e-05, ..., 2.9728e-06, + -2.8387e-05, -8.8978e-04], + [-1.5855e-04, -7.5638e-05, -1.7071e-04, ..., -3.3528e-05, + 3.2131e-06, 7.7337e-06], + [ 6.5506e-05, 2.0123e-04, 9.8646e-06, ..., 3.0994e-06, + 3.3528e-06, 7.3051e-04], + ..., + [ 9.7528e-06, 6.7316e-06, -6.8963e-05, ..., -6.9365e-06, + -2.0824e-06, -5.1618e-05], + [ 2.7969e-05, 1.4208e-05, 2.4393e-05, ..., 5.8003e-06, + 1.9044e-05, 3.5018e-05], + [ 1.2472e-05, 6.4336e-06, 4.4346e-05, ..., 3.8967e-06, + 1.8626e-05, 1.0943e-04]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0046, -0.0190, -0.0171, 0.0307, -0.0313, 0.0269, -0.0296, -0.0137, + 0.0267, -0.0218], device='cuda:0'), grad: tensor([-1.0576e-03, -3.3927e-04, 9.1267e-04, 2.3127e-04, 7.6175e-05, + -1.2898e-04, 1.3638e-04, -9.0420e-05, 1.0294e-04, 1.5724e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 215.15, cls_loss 0.0087 cls_loss_mapping 0.0177 cls_loss_causal 0.6360 re_mapping 0.0121 re_causal 0.0365 /// teacc 98.83 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0250, -0.0549, -0.0430, ..., -0.0943, -0.0038, 0.0544], + [ 0.0330, 0.0471, 0.0220, ..., 0.0618, -0.0704, -0.0466], + [ 0.0362, 0.0325, -0.0532, ..., 0.0063, -0.0838, -0.0111], + ..., + [-0.0241, -0.0243, 0.0479, ..., -0.0574, -0.0077, 0.0165], + [ 0.0020, 0.0645, 0.0837, ..., -0.0534, -0.0547, -0.0749], + [-0.0670, 0.0046, -0.1085, ..., -0.0409, -0.0373, 0.0394]], + device='cuda:0'), grad: tensor([[ 1.0437e-04, 8.7857e-05, 1.1498e-04, ..., 1.5453e-05, + 1.6121e-06, -2.5749e-05], + [-7.6485e-04, -7.6962e-04, -9.3937e-04, ..., -1.4353e-04, + 7.3053e-06, -5.0163e-04], + [ 6.1274e-05, 6.1333e-05, 5.9873e-05, ..., 2.0668e-05, + 4.1649e-06, 4.4078e-05], + ..., + [ 3.1322e-05, 5.1916e-05, 3.0786e-05, ..., 2.9802e-05, + 8.5682e-06, 1.5482e-05], + [ 1.6046e-04, 2.6989e-04, 2.1613e-04, ..., 1.1235e-04, + 5.8472e-05, 1.4436e-04], + [ 3.8475e-05, 9.2268e-05, 5.9217e-05, ..., 6.2883e-05, + 3.2753e-05, 2.5705e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0045, -0.0179, -0.0176, 0.0307, -0.0315, 0.0271, -0.0293, -0.0150, + 0.0262, -0.0209], device='cuda:0'), grad: tensor([ 2.1851e-04, -2.5120e-03, 2.0111e-04, 2.5406e-05, -1.9109e-04, + 2.7919e-04, 9.1648e-04, 1.1659e-04, 7.4244e-04, 2.0361e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 215.21, cls_loss 0.0086 cls_loss_mapping 0.0159 cls_loss_causal 0.6207 re_mapping 0.0117 re_causal 0.0357 /// teacc 98.91 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0257, -0.0556, -0.0434, ..., -0.0940, -0.0035, 0.0547], + [ 0.0334, 0.0471, 0.0222, ..., 0.0620, -0.0715, -0.0466], + [ 0.0359, 0.0324, -0.0537, ..., 0.0062, -0.0842, -0.0117], + ..., + [-0.0244, -0.0243, 0.0484, ..., -0.0572, -0.0078, 0.0172], + [ 0.0022, 0.0649, 0.0842, ..., -0.0538, -0.0554, -0.0757], + [-0.0678, 0.0047, -0.1093, ..., -0.0413, -0.0380, 0.0393]], + device='cuda:0'), grad: tensor([[ 1.8962e-06, 2.2247e-05, 4.3549e-06, ..., 8.6427e-06, + -3.7421e-06, 2.0228e-06], + [ 6.3837e-05, 3.8671e-04, 2.4915e-04, ..., 3.0732e-04, + 2.6488e-04, 9.6299e-07], + [-2.1920e-05, -4.9137e-06, 6.0014e-06, ..., 7.9796e-06, + 4.1798e-06, 4.7050e-06], + ..., + [ 8.1956e-06, 6.8508e-06, 3.7774e-06, ..., 4.6082e-06, + 9.8161e-07, 3.6638e-06], + [-1.2350e-04, -6.0320e-04, -4.7398e-04, ..., -5.0068e-04, + -4.0889e-04, 1.6153e-04], + [ 7.1079e-06, -5.6177e-05, 1.9625e-05, ..., 2.2694e-05, + 1.7956e-05, -2.0993e-04]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0043, -0.0176, -0.0183, 0.0318, -0.0319, 0.0276, -0.0299, -0.0145, + 0.0260, -0.0215], device='cuda:0'), grad: tensor([-8.8196e-07, 6.8760e-04, -8.6203e-06, 3.8177e-05, 7.7248e-05, + 9.1434e-05, 2.4772e-04, 1.8030e-05, -9.7561e-04, -1.7500e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 70---------------------------------------------------- +epoch 70, time 231.16, cls_loss 0.0074 cls_loss_mapping 0.0155 cls_loss_causal 0.6133 re_mapping 0.0116 re_causal 0.0345 /// teacc 99.04 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0260, -0.0561, -0.0437, ..., -0.0948, -0.0039, 0.0550], + [ 0.0332, 0.0467, 0.0217, ..., 0.0618, -0.0728, -0.0469], + [ 0.0362, 0.0328, -0.0543, ..., 0.0064, -0.0849, -0.0122], + ..., + [-0.0241, -0.0245, 0.0495, ..., -0.0578, -0.0081, 0.0164], + [ 0.0022, 0.0652, 0.0844, ..., -0.0544, -0.0562, -0.0766], + [-0.0682, 0.0053, -0.1100, ..., -0.0416, -0.0381, 0.0405]], + device='cuda:0'), grad: tensor([[ 1.5395e-06, -2.3423e-07, 2.9746e-06, ..., -3.1083e-07, + -3.5226e-05, -5.7042e-05], + [-1.6004e-05, -9.4920e-06, -9.8050e-06, ..., -1.0557e-05, + 2.9635e-06, 5.3234e-06], + [ 6.2352e-07, 6.9439e-06, 6.0014e-06, ..., 1.0110e-05, + 3.1777e-06, 4.2878e-06], + ..., + [ 7.7412e-06, 2.0385e-05, -6.4433e-05, ..., 2.7001e-05, + 5.7556e-06, -2.4885e-05], + [ 5.1875e-07, -7.0827e-07, -4.9034e-07, ..., 4.8243e-06, + 2.5705e-06, 4.9546e-06], + [ 6.0759e-06, 2.5213e-05, 4.9025e-05, ..., 5.2452e-05, + 2.4915e-05, 3.1382e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0041, -0.0185, -0.0181, 0.0317, -0.0323, 0.0279, -0.0300, -0.0144, + 0.0256, -0.0207], device='cuda:0'), grad: tensor([-6.7115e-05, -3.6806e-06, 1.8433e-05, 8.9360e-07, -1.1277e-04, + 1.1869e-05, 4.2886e-05, -5.1409e-05, 1.3009e-05, 1.4782e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 214.79, cls_loss 0.0059 cls_loss_mapping 0.0111 cls_loss_causal 0.5999 re_mapping 0.0112 re_causal 0.0344 /// teacc 98.97 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0264, -0.0562, -0.0443, ..., -0.0952, -0.0037, 0.0554], + [ 0.0334, 0.0467, 0.0223, ..., 0.0620, -0.0735, -0.0463], + [ 0.0367, 0.0332, -0.0549, ..., 0.0066, -0.0853, -0.0123], + ..., + [-0.0244, -0.0248, 0.0498, ..., -0.0586, -0.0083, 0.0164], + [ 0.0023, 0.0655, 0.0850, ..., -0.0549, -0.0566, -0.0770], + [-0.0688, 0.0051, -0.1112, ..., -0.0422, -0.0386, 0.0406]], + device='cuda:0'), grad: tensor([[ 1.8543e-06, 1.0580e-05, 4.4107e-06, ..., 8.1435e-06, + 1.2033e-05, -8.0885e-07], + [ 4.1366e-05, 1.2183e-04, 1.4555e-04, ..., 1.0109e-04, + 1.1128e-04, 1.1502e-06], + [-4.7833e-06, 7.3714e-07, 4.0174e-05, ..., 1.5013e-05, + 5.4948e-06, -1.5739e-07], + ..., + [-1.6674e-05, 2.3991e-06, -6.8784e-05, ..., -1.7375e-05, + 1.0543e-06, -2.7269e-06], + [-8.0407e-05, -3.8052e-04, -4.3011e-04, ..., -3.1495e-04, + -3.5048e-04, 2.8573e-06], + [ 7.7114e-06, 1.7583e-05, 2.7567e-05, ..., 1.9878e-05, + 2.0802e-05, -6.0238e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0043, -0.0183, -0.0176, 0.0318, -0.0322, 0.0273, -0.0293, -0.0147, + 0.0257, -0.0210], device='cuda:0'), grad: tensor([ 1.9595e-05, 2.7394e-04, 5.6654e-05, 5.2378e-06, 1.2505e-04, + 8.7142e-05, 2.4581e-04, -1.2058e-04, -7.3719e-04, 4.4405e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.87, cls_loss 0.0058 cls_loss_mapping 0.0128 cls_loss_causal 0.6079 re_mapping 0.0108 re_causal 0.0321 /// teacc 99.02 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0269, -0.0566, -0.0441, ..., -0.0948, -0.0038, 0.0557], + [ 0.0335, 0.0469, 0.0225, ..., 0.0626, -0.0743, -0.0466], + [ 0.0366, 0.0330, -0.0561, ..., 0.0061, -0.0857, -0.0130], + ..., + [-0.0244, -0.0252, 0.0505, ..., -0.0591, -0.0090, 0.0166], + [ 0.0026, 0.0659, 0.0857, ..., -0.0552, -0.0573, -0.0778], + [-0.0695, 0.0048, -0.1121, ..., -0.0432, -0.0392, 0.0407]], + device='cuda:0'), grad: tensor([[-1.0589e-06, -1.7703e-05, 3.8110e-06, ..., 1.0438e-05, + -7.2606e-06, -7.9989e-05], + [-8.4797e-07, 2.6122e-05, -9.0674e-06, ..., -6.0163e-07, + 5.1856e-06, 9.7007e-06], + [ 5.3793e-05, 1.8656e-04, 6.2525e-05, ..., 1.2529e-04, + 1.6883e-05, 2.4199e-05], + ..., + [ 7.2345e-06, 7.4506e-06, 5.4725e-06, ..., 7.1339e-06, + 1.2666e-06, -6.4492e-05], + [-1.6451e-04, 5.2929e-05, -7.7367e-05, ..., 4.4972e-05, + 1.0109e-04, 1.0923e-05], + [ 2.9072e-05, 3.2425e-05, 1.4558e-05, ..., 7.7710e-06, + 1.1079e-05, 8.0764e-05]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0044, -0.0184, -0.0183, 0.0321, -0.0321, 0.0275, -0.0291, -0.0143, + 0.0256, -0.0215], device='cuda:0'), grad: tensor([-8.1778e-05, 2.9013e-05, 2.7299e-04, 1.1146e-04, 9.1553e-05, + 9.8467e-05, -6.0129e-04, -4.8697e-05, -4.5300e-06, 1.3268e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 214.76, cls_loss 0.0055 cls_loss_mapping 0.0131 cls_loss_causal 0.6123 re_mapping 0.0110 re_causal 0.0331 /// teacc 98.87 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0273, -0.0580, -0.0445, ..., -0.0950, -0.0036, 0.0552], + [ 0.0333, 0.0461, 0.0222, ..., 0.0617, -0.0766, -0.0471], + [ 0.0370, 0.0334, -0.0562, ..., 0.0062, -0.0863, -0.0133], + ..., + [-0.0248, -0.0255, 0.0508, ..., -0.0592, -0.0086, 0.0166], + [ 0.0027, 0.0662, 0.0864, ..., -0.0554, -0.0580, -0.0788], + [-0.0705, 0.0051, -0.1130, ..., -0.0436, -0.0399, 0.0417]], + device='cuda:0'), grad: tensor([[ 2.7828e-06, 4.5933e-06, 4.7609e-06, ..., 4.9397e-06, + 1.7462e-08, -5.1409e-06], + [-1.8373e-05, -1.1587e-04, -1.3912e-04, ..., -1.5748e-04, + -8.5235e-05, 2.7567e-07], + [-3.1024e-05, -6.0946e-06, 1.0297e-05, ..., -4.7460e-06, + 2.7996e-06, 6.1328e-07], + ..., + [ 4.9509e-06, 1.0796e-05, -3.3140e-05, ..., 1.0706e-05, + 5.7518e-06, 2.8661e-07], + [-2.3451e-06, 2.7381e-06, 3.6461e-07, ..., 1.2405e-05, + 5.2825e-06, 4.2133e-06], + [ 6.1020e-06, 7.3500e-06, 1.4678e-05, ..., 1.8254e-05, + 1.3143e-05, -7.1302e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0037, -0.0194, -0.0179, 0.0325, -0.0322, 0.0277, -0.0284, -0.0142, + 0.0254, -0.0214], device='cuda:0'), grad: tensor([ 5.0478e-06, -2.1553e-04, -3.8296e-05, 5.1349e-05, 6.3002e-05, + 8.6129e-06, 1.1951e-04, -4.0472e-05, 1.6168e-05, 3.0607e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 214.49, cls_loss 0.0072 cls_loss_mapping 0.0156 cls_loss_causal 0.6171 re_mapping 0.0112 re_causal 0.0330 /// teacc 98.85 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0278, -0.0583, -0.0448, ..., -0.0954, -0.0035, 0.0560], + [ 0.0346, 0.0470, 0.0244, ..., 0.0619, -0.0771, -0.0469], + [ 0.0374, 0.0332, -0.0577, ..., 0.0061, -0.0883, -0.0137], + ..., + [-0.0261, -0.0265, 0.0506, ..., -0.0601, -0.0091, 0.0166], + [ 0.0029, 0.0666, 0.0870, ..., -0.0557, -0.0584, -0.0798], + [-0.0713, 0.0049, -0.1145, ..., -0.0442, -0.0407, 0.0418]], + device='cuda:0'), grad: tensor([[ 2.2408e-06, 3.0603e-06, 2.1290e-06, ..., 1.7611e-06, + 9.6578e-07, -1.4521e-05], + [-6.6757e-05, -7.2718e-05, -1.0532e-04, ..., -9.4771e-05, + -2.2873e-05, 1.0384e-06], + [ 1.6659e-05, 6.8806e-06, 6.1169e-06, ..., 3.6564e-06, + 2.3730e-06, 2.6487e-06], + ..., + [ 6.1020e-06, 3.7495e-06, 3.5157e-07, ..., 2.3507e-06, + 3.0138e-06, 2.8573e-06], + [ 2.6792e-05, 1.5132e-05, 9.4250e-06, ..., 1.0207e-05, + 2.2992e-05, 5.4762e-06], + [ 6.3814e-06, -2.3935e-06, 3.2280e-06, ..., 1.1548e-06, + 7.8678e-06, 2.8070e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0043, -0.0179, -0.0181, 0.0320, -0.0321, 0.0278, -0.0281, -0.0149, + 0.0254, -0.0220], device='cuda:0'), grad: tensor([-5.8562e-06, -1.3757e-04, 2.5481e-05, -4.9025e-05, 1.6719e-05, + -3.1322e-05, 9.1612e-05, 1.1444e-05, 6.5744e-05, 1.2569e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 214.72, cls_loss 0.0066 cls_loss_mapping 0.0141 cls_loss_causal 0.5959 re_mapping 0.0110 re_causal 0.0320 /// teacc 98.86 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0285, -0.0586, -0.0448, ..., -0.0954, -0.0034, 0.0565], + [ 0.0354, 0.0470, 0.0254, ..., 0.0620, -0.0776, -0.0477], + [ 0.0366, 0.0324, -0.0592, ..., 0.0058, -0.0886, -0.0139], + ..., + [-0.0266, -0.0267, 0.0509, ..., -0.0607, -0.0097, 0.0169], + [ 0.0042, 0.0685, 0.0880, ..., -0.0561, -0.0569, -0.0793], + [-0.0720, 0.0060, -0.1155, ..., -0.0443, -0.0412, 0.0421]], + device='cuda:0'), grad: tensor([[ 2.3618e-06, 2.1551e-06, 4.8161e-05, ..., 2.2594e-06, + 4.6045e-06, 1.6540e-05], + [-7.6413e-05, -6.6102e-05, -7.8440e-05, ..., -8.2135e-05, + -5.3197e-06, 4.1686e-06], + [-1.2659e-05, -8.7321e-06, 2.8268e-05, ..., 6.3144e-06, + 9.0599e-06, 1.9774e-05], + ..., + [ 8.9630e-06, 7.6294e-06, -1.0598e-04, ..., 7.2457e-06, + -1.4752e-05, -6.9559e-05], + [ 3.5793e-05, 3.1173e-05, 3.1918e-05, ..., 3.2067e-05, + 2.4542e-05, 1.6600e-05], + [ 6.0871e-06, 1.0073e-05, 1.6496e-05, ..., 1.5140e-05, + 2.3559e-05, 2.4840e-05]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0047, -0.0174, -0.0188, 0.0325, -0.0326, 0.0265, -0.0291, -0.0146, + 0.0271, -0.0218], device='cuda:0'), grad: tensor([ 7.3075e-05, -1.1867e-04, 2.8297e-05, 2.0809e-03, -3.6061e-06, + -2.1095e-03, 3.7491e-05, -1.7846e-04, 1.1241e-04, 7.9870e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 215.06, cls_loss 0.0054 cls_loss_mapping 0.0117 cls_loss_causal 0.6320 re_mapping 0.0107 re_causal 0.0338 /// teacc 98.81 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0290, -0.0593, -0.0455, ..., -0.0958, -0.0035, 0.0566], + [ 0.0358, 0.0469, 0.0260, ..., 0.0623, -0.0780, -0.0481], + [ 0.0373, 0.0330, -0.0593, ..., 0.0056, -0.0890, -0.0143], + ..., + [-0.0271, -0.0270, 0.0511, ..., -0.0611, -0.0099, 0.0171], + [ 0.0040, 0.0684, 0.0882, ..., -0.0565, -0.0573, -0.0800], + [-0.0724, 0.0072, -0.1157, ..., -0.0441, -0.0422, 0.0426]], + device='cuda:0'), grad: tensor([[ 2.7549e-06, 2.3246e-05, 5.9605e-06, ..., 8.1584e-07, + 3.8520e-06, 2.1562e-05], + [-1.5736e-05, -2.7511e-06, -1.5795e-05, ..., -7.0147e-06, + 4.8950e-06, -1.0123e-06], + [ 3.2961e-05, 2.3991e-05, 2.0757e-05, ..., 8.0094e-06, + 1.7416e-06, 8.9481e-06], + ..., + [ 4.8019e-06, 3.9749e-06, 3.6899e-06, ..., 1.8161e-06, + 1.1278e-06, 2.5183e-06], + [-2.6003e-05, -8.5890e-05, -1.6820e-04, ..., -1.3700e-06, + -1.1837e-04, 2.9579e-05], + [ 1.6168e-05, -2.9996e-05, 1.4327e-05, ..., 2.1663e-06, + 4.9807e-06, -6.8247e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0045, -0.0172, -0.0180, 0.0319, -0.0332, 0.0269, -0.0293, -0.0148, + 0.0266, -0.0213], device='cuda:0'), grad: tensor([ 3.6985e-05, -1.5929e-05, 5.4300e-05, -4.8965e-05, -1.4203e-06, + 8.8274e-05, 1.3101e-04, 1.1161e-05, -2.0492e-04, -5.0575e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 214.78, cls_loss 0.0061 cls_loss_mapping 0.0134 cls_loss_causal 0.6168 re_mapping 0.0103 re_causal 0.0312 /// teacc 99.00 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0294, -0.0597, -0.0461, ..., -0.0965, -0.0039, 0.0570], + [ 0.0355, 0.0471, 0.0259, ..., 0.0628, -0.0788, -0.0479], + [ 0.0374, 0.0332, -0.0605, ..., 0.0057, -0.0896, -0.0144], + ..., + [-0.0264, -0.0275, 0.0525, ..., -0.0620, -0.0097, 0.0175], + [ 0.0042, 0.0686, 0.0887, ..., -0.0570, -0.0579, -0.0807], + [-0.0730, 0.0069, -0.1171, ..., -0.0450, -0.0433, 0.0424]], + device='cuda:0'), grad: tensor([[ 1.6615e-05, 6.1765e-06, 2.1365e-06, ..., 6.3796e-07, + -2.9162e-05, -2.1029e-04], + [ 3.7223e-05, 1.5102e-05, 1.7202e-04, ..., -1.0848e-05, + 2.3097e-06, 3.5942e-05], + [-7.4990e-06, -1.2472e-05, 3.2872e-05, ..., 1.7080e-06, + 1.1791e-06, 1.5974e-05], + ..., + [-3.2276e-05, -2.2039e-05, -3.2020e-04, ..., 5.0999e-06, + 1.6103e-06, -4.7207e-05], + [ 1.4961e-05, 5.4613e-06, 2.3887e-05, ..., 1.6103e-06, + 1.6112e-06, 1.9014e-05], + [ 2.2769e-05, -1.3746e-05, 4.5300e-05, ..., 5.7258e-06, + 2.7463e-05, 1.3936e-04]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0044, -0.0175, -0.0185, 0.0319, -0.0329, 0.0270, -0.0288, -0.0136, + 0.0264, -0.0221], device='cuda:0'), grad: tensor([-1.0026e-04, 3.1042e-04, -2.1353e-05, -2.9826e-04, 3.5733e-05, + 2.2247e-05, 5.6982e-05, -3.2616e-04, 1.1379e-04, 2.0695e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 214.83, cls_loss 0.0061 cls_loss_mapping 0.0133 cls_loss_causal 0.6032 re_mapping 0.0108 re_causal 0.0309 /// teacc 98.94 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0298, -0.0606, -0.0467, ..., -0.0967, -0.0035, 0.0573], + [ 0.0360, 0.0471, 0.0266, ..., 0.0629, -0.0793, -0.0488], + [ 0.0377, 0.0333, -0.0610, ..., 0.0054, -0.0900, -0.0147], + ..., + [-0.0272, -0.0278, 0.0523, ..., -0.0619, -0.0097, 0.0174], + [ 0.0045, 0.0690, 0.0897, ..., -0.0579, -0.0587, -0.0824], + [-0.0735, 0.0068, -0.1178, ..., -0.0460, -0.0441, 0.0432]], + device='cuda:0'), grad: tensor([[ 2.7921e-06, 1.6764e-06, 2.4494e-06, ..., 5.6848e-06, + -1.0565e-05, -2.1994e-04], + [-6.5279e-04, -1.6438e-06, -6.7091e-04, ..., -5.0735e-04, + 1.6624e-06, 8.8513e-06], + [ 2.2674e-04, 2.0098e-06, 2.5463e-04, ..., 1.8716e-04, + 3.2075e-06, 2.9281e-05], + ..., + [ 3.5310e-04, 1.9372e-05, 3.3736e-04, ..., 2.6822e-04, + 1.3057e-06, -1.7206e-07], + [ 3.5111e-06, 2.6468e-06, 1.4994e-06, ..., 4.7721e-06, + 5.0753e-05, 9.5725e-05], + [ 8.3894e-06, -3.1769e-05, 1.0706e-05, ..., 5.9679e-06, + 2.3738e-05, 1.3638e-04]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0046, -0.0171, -0.0183, 0.0320, -0.0329, 0.0268, -0.0288, -0.0142, + 0.0262, -0.0219], device='cuda:0'), grad: tensor([-2.0146e-04, -1.0939e-03, 4.2558e-04, 2.7493e-05, 4.4078e-05, + -5.8264e-05, -1.7611e-06, 5.9557e-04, 1.2875e-04, 1.3316e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.98, cls_loss 0.0052 cls_loss_mapping 0.0097 cls_loss_causal 0.5764 re_mapping 0.0101 re_causal 0.0301 /// teacc 98.90 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0304, -0.0606, -0.0473, ..., -0.0974, -0.0030, 0.0585], + [ 0.0361, 0.0473, 0.0268, ..., 0.0633, -0.0799, -0.0491], + [ 0.0378, 0.0332, -0.0616, ..., 0.0049, -0.0907, -0.0148], + ..., + [-0.0276, -0.0278, 0.0527, ..., -0.0625, -0.0102, 0.0178], + [ 0.0047, 0.0696, 0.0904, ..., -0.0583, -0.0591, -0.0833], + [-0.0744, 0.0067, -0.1189, ..., -0.0466, -0.0446, 0.0429]], + device='cuda:0'), grad: tensor([[ 8.3521e-06, 9.0003e-06, -1.9670e-05, ..., 4.4405e-06, + 2.8145e-06, -8.7976e-05], + [ 3.2149e-06, 1.6699e-06, 1.8487e-06, ..., -8.1025e-07, + 2.0657e-06, 1.2733e-05], + [ 1.6674e-05, 3.3230e-05, 3.3945e-05, ..., 1.3709e-06, + 5.1893e-06, 4.6343e-06], + ..., + [ 7.2718e-06, 9.5293e-06, 3.5223e-06, ..., 5.9195e-06, + 6.3777e-06, 1.7777e-05], + [ 9.7081e-06, -1.9479e-04, -1.5640e-04, ..., 5.4128e-06, + -1.6451e-05, 2.9549e-05], + [ 1.1578e-05, -1.6003e-03, 1.9655e-05, ..., -9.3842e-04, + -1.0347e-03, -1.6088e-03]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0054, -0.0171, -0.0185, 0.0323, -0.0326, 0.0269, -0.0291, -0.0142, + 0.0261, -0.0225], device='cuda:0'), grad: tensor([-5.7012e-05, 2.2113e-05, 3.1561e-05, -8.6367e-05, 3.6774e-03, + 1.9222e-05, 1.9360e-04, 4.9919e-05, -1.8251e-04, -3.6716e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 214.87, cls_loss 0.0051 cls_loss_mapping 0.0105 cls_loss_causal 0.5879 re_mapping 0.0098 re_causal 0.0293 /// teacc 98.93 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0310, -0.0608, -0.0480, ..., -0.0979, -0.0030, 0.0589], + [ 0.0361, 0.0473, 0.0267, ..., 0.0635, -0.0804, -0.0494], + [ 0.0382, 0.0334, -0.0617, ..., 0.0047, -0.0911, -0.0151], + ..., + [-0.0277, -0.0277, 0.0537, ..., -0.0629, -0.0108, 0.0185], + [ 0.0046, 0.0697, 0.0906, ..., -0.0587, -0.0597, -0.0845], + [-0.0750, 0.0070, -0.1207, ..., -0.0466, -0.0448, 0.0432]], + device='cuda:0'), grad: tensor([[ 6.0834e-06, 2.8275e-06, 2.5202e-06, ..., 1.0459e-06, + 2.8331e-06, -8.2999e-06], + [-1.9714e-05, -4.8459e-05, -7.8738e-05, ..., -3.7670e-05, + 1.3895e-06, -3.1352e-05], + [ 8.3506e-05, 1.5497e-05, 8.5011e-06, ..., 4.3064e-06, + 3.2168e-06, 5.9120e-06], + ..., + [ 4.3809e-05, 3.5137e-05, 4.5091e-05, ..., 1.9386e-05, + 1.4836e-06, 1.8209e-05], + [ 4.3124e-05, 9.4235e-05, 3.8296e-05, ..., 5.9791e-06, + 6.5714e-06, 7.2896e-05], + [ 8.6650e-06, -1.0502e-04, -1.2644e-05, ..., 1.0610e-05, + 6.1952e-06, -8.0705e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0054, -0.0174, -0.0183, 0.0324, -0.0329, 0.0273, -0.0292, -0.0135, + 0.0255, -0.0229], device='cuda:0'), grad: tensor([ 1.7881e-05, -7.9751e-05, 2.1327e-04, -4.4918e-04, 4.5449e-05, + 7.4267e-05, 6.9626e-06, 9.4473e-05, 2.2519e-04, -1.4830e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 215.08, cls_loss 0.0044 cls_loss_mapping 0.0097 cls_loss_causal 0.5968 re_mapping 0.0099 re_causal 0.0311 /// teacc 98.88 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0313, -0.0604, -0.0469, ..., -0.0973, -0.0033, 0.0593], + [ 0.0363, 0.0473, 0.0270, ..., 0.0630, -0.0816, -0.0494], + [ 0.0383, 0.0333, -0.0619, ..., 0.0046, -0.0912, -0.0154], + ..., + [-0.0281, -0.0282, 0.0538, ..., -0.0634, -0.0113, 0.0183], + [ 0.0045, 0.0698, 0.0906, ..., -0.0593, -0.0600, -0.0853], + [-0.0754, 0.0068, -0.1215, ..., -0.0473, -0.0454, 0.0433]], + device='cuda:0'), grad: tensor([[ 6.6962e-07, 2.1036e-07, 3.9511e-07, ..., 9.4948e-07, + 1.7770e-06, -6.2212e-06], + [-2.5127e-06, -2.0470e-06, -5.8226e-06, ..., -3.8054e-06, + 2.0880e-06, 1.7351e-06], + [ 8.3772e-07, -5.2061e-07, 1.0673e-06, ..., 9.9465e-07, + 1.6764e-06, 2.3153e-06], + ..., + [ 1.6131e-06, 1.5171e-06, 9.1642e-07, ..., 1.4808e-06, + 2.6487e-06, 2.1569e-06], + [ 1.4845e-06, 1.1707e-06, 1.8545e-07, ..., 1.6950e-06, + 1.9357e-05, 1.7703e-05], + [ 1.8533e-06, 4.8988e-06, 9.4762e-07, ..., 1.0632e-05, + 1.2800e-05, 3.0529e-06]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0059, -0.0175, -0.0183, 0.0324, -0.0330, 0.0276, -0.0281, -0.0138, + 0.0250, -0.0233], device='cuda:0'), grad: tensor([-3.0734e-06, -6.5193e-07, 8.3521e-06, -6.4671e-06, -1.2010e-05, + -6.9320e-05, 1.0751e-05, 8.0615e-06, 3.8296e-05, 2.6017e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 214.92, cls_loss 0.0045 cls_loss_mapping 0.0081 cls_loss_causal 0.6120 re_mapping 0.0099 re_causal 0.0305 /// teacc 98.91 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0317, -0.0607, -0.0473, ..., -0.0977, -0.0036, 0.0595], + [ 0.0363, 0.0472, 0.0271, ..., 0.0629, -0.0820, -0.0496], + [ 0.0387, 0.0335, -0.0619, ..., 0.0047, -0.0917, -0.0162], + ..., + [-0.0285, -0.0288, 0.0541, ..., -0.0635, -0.0118, 0.0188], + [ 0.0047, 0.0702, 0.0909, ..., -0.0600, -0.0608, -0.0865], + [-0.0759, 0.0070, -0.1223, ..., -0.0478, -0.0457, 0.0439]], + device='cuda:0'), grad: tensor([[ 1.6034e-05, 1.1705e-05, 3.8818e-06, ..., 1.1427e-06, + -1.4557e-06, -3.9637e-05], + [ 6.5081e-06, 5.4277e-06, 5.1372e-06, ..., 1.3225e-06, + 2.3637e-06, 4.2096e-06], + [-3.6329e-05, -3.4630e-05, -1.6227e-05, ..., -2.0228e-06, + 1.2424e-06, -2.0619e-06], + ..., + [ 2.0906e-05, 8.0094e-06, -5.7779e-06, ..., 4.1239e-06, + 1.8664e-06, -1.2591e-05], + [-2.8923e-05, -3.2425e-05, -5.9962e-05, ..., 4.1164e-06, + 1.9018e-06, 9.3803e-06], + [ 1.2785e-05, 1.6212e-05, 1.4052e-05, ..., 2.1294e-05, + 1.1526e-05, 1.5900e-05]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0059, -0.0178, -0.0182, 0.0323, -0.0321, 0.0275, -0.0286, -0.0136, + 0.0247, -0.0233], device='cuda:0'), grad: tensor([-5.7399e-05, 2.5928e-05, -8.6784e-05, 9.2834e-06, -2.5779e-06, + 1.8746e-05, 2.2322e-05, 2.8387e-05, -4.0084e-05, 8.1837e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 214.48, cls_loss 0.0056 cls_loss_mapping 0.0103 cls_loss_causal 0.5978 re_mapping 0.0104 re_causal 0.0307 /// teacc 98.76 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0324, -0.0615, -0.0477, ..., -0.0984, -0.0038, 0.0597], + [ 0.0362, 0.0478, 0.0283, ..., 0.0643, -0.0823, -0.0492], + [ 0.0392, 0.0338, -0.0618, ..., 0.0048, -0.0923, -0.0162], + ..., + [-0.0291, -0.0305, 0.0532, ..., -0.0657, -0.0117, 0.0183], + [ 0.0048, 0.0708, 0.0918, ..., -0.0602, -0.0612, -0.0871], + [-0.0764, 0.0072, -0.1230, ..., -0.0480, -0.0459, 0.0446]], + device='cuda:0'), grad: tensor([[ 3.0864e-06, 1.6978e-06, 7.0371e-06, ..., 7.2837e-05, + 1.2326e-04, 1.3122e-06], + [-4.2878e-06, -1.1392e-05, -1.2048e-05, ..., -2.3544e-05, + 2.2985e-06, 2.7511e-06], + [-6.5148e-05, 1.2582e-06, -9.3654e-06, ..., 5.5395e-06, + 2.4009e-06, 6.1579e-06], + ..., + [ 2.1353e-05, 6.3032e-06, -2.6584e-04, ..., -3.4332e-05, + 2.9597e-06, -8.8513e-05], + [ 8.6473e-07, -3.8832e-05, -2.2754e-05, ..., 4.9211e-06, + -4.1462e-06, 5.7705e-06], + [ 1.0818e-05, 3.1948e-05, 1.0622e-04, ..., 2.6554e-05, + 4.3541e-05, 6.2048e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0056, -0.0172, -0.0177, 0.0331, -0.0329, 0.0272, -0.0283, -0.0147, + 0.0247, -0.0232], device='cuda:0'), grad: tensor([ 1.4329e-04, -6.3553e-06, -1.1277e-04, 2.1315e-04, 5.8413e-05, + -4.2558e-05, -3.7462e-05, -4.5371e-04, -4.4376e-05, 2.8229e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 214.99, cls_loss 0.0054 cls_loss_mapping 0.0099 cls_loss_causal 0.5800 re_mapping 0.0101 re_causal 0.0298 /// teacc 98.89 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0329, -0.0616, -0.0483, ..., -0.0988, -0.0038, 0.0600], + [ 0.0363, 0.0477, 0.0284, ..., 0.0642, -0.0831, -0.0496], + [ 0.0393, 0.0344, -0.0630, ..., 0.0066, -0.0910, -0.0167], + ..., + [-0.0293, -0.0308, 0.0537, ..., -0.0658, -0.0123, 0.0184], + [ 0.0055, 0.0719, 0.0931, ..., -0.0602, -0.0615, -0.0875], + [-0.0769, 0.0075, -0.1240, ..., -0.0480, -0.0450, 0.0458]], + device='cuda:0'), grad: tensor([[ 1.0021e-06, 8.0049e-05, 1.6009e-06, ..., 1.5013e-05, + 1.7357e-04, 3.3808e-04], + [ 2.3711e-06, 5.7667e-06, 6.6869e-06, ..., 6.9477e-07, + 2.1867e-06, 3.0473e-06], + [-1.7369e-06, 1.3657e-05, 6.7577e-06, ..., 2.5388e-06, + 2.6792e-05, 5.2094e-05], + ..., + [ 2.1774e-06, 4.9770e-06, -8.2403e-06, ..., 8.0280e-07, + 3.5334e-06, 2.3991e-06], + [-2.5570e-05, -5.9813e-05, -6.3598e-05, ..., 7.1852e-07, + 1.0706e-05, 1.9699e-05], + [ 1.2778e-05, -9.2328e-05, 3.4958e-05, ..., -1.9312e-05, + -2.6584e-04, -5.3263e-04]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0056, -0.0174, -0.0168, 0.0331, -0.0345, 0.0270, -0.0286, -0.0145, + 0.0255, -0.0226], device='cuda:0'), grad: tensor([ 5.7793e-04, 1.6928e-05, 8.9705e-05, 3.0011e-05, 1.4663e-04, + 2.7865e-05, 1.7509e-05, 6.4857e-06, -7.8321e-05, -8.3494e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 214.83, cls_loss 0.0051 cls_loss_mapping 0.0101 cls_loss_causal 0.5906 re_mapping 0.0101 re_causal 0.0308 /// teacc 98.86 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0335, -0.0616, -0.0480, ..., -0.0983, -0.0038, 0.0601], + [ 0.0366, 0.0477, 0.0284, ..., 0.0641, -0.0848, -0.0499], + [ 0.0393, 0.0341, -0.0639, ..., 0.0063, -0.0918, -0.0170], + ..., + [-0.0293, -0.0313, 0.0546, ..., -0.0664, -0.0133, 0.0187], + [ 0.0056, 0.0719, 0.0932, ..., -0.0624, -0.0624, -0.0883], + [-0.0774, 0.0079, -0.1248, ..., -0.0481, -0.0447, 0.0466]], + device='cuda:0'), grad: tensor([[ 1.2256e-06, 3.9250e-05, 3.8370e-06, ..., 3.0436e-06, + 1.6809e-04, 7.1228e-05], + [-2.0433e-06, 4.0904e-06, 2.2829e-05, ..., 9.4026e-06, + 1.4290e-05, 2.2128e-05], + [-1.2135e-06, 1.3802e-06, 2.3052e-05, ..., 3.6061e-06, + 3.8892e-06, 2.0295e-05], + ..., + [ 9.9279e-07, 3.3099e-06, -8.9931e-04, ..., 6.7130e-06, + 6.1132e-06, -7.4291e-04], + [-9.9838e-06, 3.1833e-06, 1.1556e-05, ..., 1.6406e-05, + 2.2590e-05, 2.5690e-05], + [ 2.5127e-06, 1.9744e-05, 7.1907e-04, ..., 3.7432e-05, + 3.4094e-05, 5.8126e-04]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0059, -0.0177, -0.0172, 0.0327, -0.0346, 0.0274, -0.0285, -0.0141, + 0.0251, -0.0223], device='cuda:0'), grad: tensor([ 2.2864e-04, 8.2731e-05, 6.4313e-05, 5.1528e-05, 3.8773e-05, + 9.3818e-05, -2.0421e-04, -2.4395e-03, 7.9393e-05, 2.0027e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 215.23, cls_loss 0.0053 cls_loss_mapping 0.0101 cls_loss_causal 0.6068 re_mapping 0.0103 re_causal 0.0304 /// teacc 98.83 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0342, -0.0622, -0.0488, ..., -0.0990, -0.0039, 0.0605], + [ 0.0356, 0.0473, 0.0272, ..., 0.0640, -0.0851, -0.0506], + [ 0.0403, 0.0354, -0.0626, ..., 0.0070, -0.0923, -0.0168], + ..., + [-0.0283, -0.0313, 0.0565, ..., -0.0669, -0.0135, 0.0191], + [ 0.0052, 0.0715, 0.0930, ..., -0.0635, -0.0631, -0.0893], + [-0.0780, 0.0075, -0.1266, ..., -0.0499, -0.0470, 0.0465]], + device='cuda:0'), grad: tensor([[ 3.1758e-06, 5.4166e-06, 9.9689e-06, ..., 2.6952e-06, + -4.1686e-06, -1.9997e-05], + [ 5.9642e-06, 4.5709e-06, 3.3885e-05, ..., 1.4510e-06, + 3.3528e-06, 2.6263e-06], + [-3.0566e-06, 7.7300e-07, 1.3947e-04, ..., 1.8282e-06, + 3.1460e-06, 5.3123e-06], + ..., + [-8.0645e-05, 1.4296e-06, -4.6706e-04, ..., 4.2841e-07, + 7.6229e-07, 7.6890e-06], + [ 1.1407e-05, 5.3436e-05, 1.9938e-05, ..., 2.7373e-05, + 3.3975e-05, 5.8152e-06], + [ 9.7305e-06, -1.5115e-06, 4.7177e-05, ..., 1.1902e-06, + 1.5404e-06, -1.5914e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0057, -0.0192, -0.0159, 0.0329, -0.0329, 0.0272, -0.0288, -0.0128, + 0.0240, -0.0237], device='cuda:0'), grad: tensor([ 6.0387e-06, 9.9719e-05, 3.5310e-04, 4.8518e-04, 8.2135e-05, + 6.9559e-05, -9.8586e-05, -1.2579e-03, 1.4639e-04, 1.1313e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 215.11, cls_loss 0.0041 cls_loss_mapping 0.0100 cls_loss_causal 0.6003 re_mapping 0.0094 re_causal 0.0289 /// teacc 98.88 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0346, -0.0630, -0.0490, ..., -0.0995, -0.0044, 0.0611], + [ 0.0355, 0.0471, 0.0273, ..., 0.0639, -0.0856, -0.0513], + [ 0.0406, 0.0354, -0.0627, ..., 0.0071, -0.0929, -0.0174], + ..., + [-0.0286, -0.0319, 0.0565, ..., -0.0671, -0.0138, 0.0186], + [ 0.0056, 0.0723, 0.0938, ..., -0.0639, -0.0632, -0.0898], + [-0.0785, 0.0075, -0.1271, ..., -0.0504, -0.0479, 0.0469]], + device='cuda:0'), grad: tensor([[ 2.4196e-06, 4.0680e-06, 5.8413e-06, ..., 1.0775e-06, + 9.6262e-06, 9.1195e-06], + [-3.0518e-05, -2.3708e-05, -3.9250e-05, ..., -3.0637e-05, + 8.1351e-07, 2.0955e-06], + [ 8.4192e-06, 6.5304e-06, 1.2264e-05, ..., 5.9269e-06, + 1.2480e-06, 3.2429e-06], + ..., + [ 7.1526e-06, 1.4827e-05, 3.8594e-06, ..., 6.5342e-06, + 2.1216e-06, 1.7928e-07], + [-1.7631e-04, -5.1117e-04, -5.9891e-04, ..., 4.8093e-06, + 8.7842e-06, -2.7776e-04], + [ 1.7178e-04, 4.8161e-04, 5.7745e-04, ..., 4.2543e-06, + 1.3979e-06, 2.5201e-04]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0058, -0.0194, -0.0159, 0.0320, -0.0329, 0.0280, -0.0281, -0.0131, + 0.0242, -0.0239], device='cuda:0'), grad: tensor([ 2.3663e-05, -5.7220e-05, 2.2978e-05, 5.3287e-05, 1.6659e-05, + -8.9824e-05, 7.5996e-05, 1.7643e-05, -1.2541e-03, 1.1902e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 214.95, cls_loss 0.0044 cls_loss_mapping 0.0105 cls_loss_causal 0.5636 re_mapping 0.0097 re_causal 0.0278 /// teacc 98.97 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0353, -0.0635, -0.0497, ..., -0.0996, -0.0044, 0.0613], + [ 0.0358, 0.0471, 0.0275, ..., 0.0640, -0.0861, -0.0514], + [ 0.0404, 0.0351, -0.0637, ..., 0.0068, -0.0936, -0.0179], + ..., + [-0.0286, -0.0319, 0.0574, ..., -0.0673, -0.0145, 0.0187], + [ 0.0057, 0.0723, 0.0942, ..., -0.0645, -0.0637, -0.0907], + [-0.0786, 0.0076, -0.1283, ..., -0.0513, -0.0483, 0.0476]], + device='cuda:0'), grad: tensor([[ 2.3246e-06, 3.7681e-06, 2.8238e-06, ..., 4.0531e-06, + 8.2850e-06, 1.8617e-06], + [-6.2510e-06, -3.3341e-06, -8.3521e-06, ..., -9.8869e-06, + 4.6976e-06, 1.0189e-06], + [-2.4483e-05, 1.4836e-06, 4.5337e-06, ..., 3.2391e-06, + 2.7958e-06, 6.7521e-09], + ..., + [ 3.0205e-05, 9.8944e-06, 2.9132e-06, ..., 6.0350e-06, + 4.3437e-06, -3.8221e-06], + [-4.8548e-05, -5.2392e-05, -6.8486e-05, ..., 7.3910e-06, + -1.5631e-05, 5.4128e-06], + [ 1.3836e-05, 2.1353e-05, 1.9714e-05, ..., 1.7822e-05, + 1.7554e-05, -1.0356e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0059, -0.0193, -0.0164, 0.0320, -0.0327, 0.0276, -0.0277, -0.0127, + 0.0239, -0.0239], device='cuda:0'), grad: tensor([ 1.4536e-05, -6.9290e-07, -8.1897e-05, 5.0128e-05, 7.8529e-06, + 7.5586e-06, -1.9781e-06, 9.4056e-05, -1.6081e-04, 7.1108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 215.00, cls_loss 0.0050 cls_loss_mapping 0.0083 cls_loss_causal 0.5832 re_mapping 0.0097 re_causal 0.0287 /// teacc 98.84 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0358, -0.0640, -0.0501, ..., -0.0999, -0.0064, 0.0599], + [ 0.0371, 0.0479, 0.0286, ..., 0.0645, -0.0865, -0.0507], + [ 0.0399, 0.0345, -0.0649, ..., 0.0066, -0.0942, -0.0183], + ..., + [-0.0294, -0.0330, 0.0576, ..., -0.0674, -0.0129, 0.0181], + [ 0.0058, 0.0726, 0.0949, ..., -0.0650, -0.0644, -0.0912], + [-0.0792, 0.0067, -0.1292, ..., -0.0523, -0.0493, 0.0481]], + device='cuda:0'), grad: tensor([[ 3.1721e-06, 8.0645e-05, 7.4040e-07, ..., 1.1260e-06, + 3.0428e-05, 1.6296e-04], + [ 2.1495e-06, 3.4962e-06, 2.6543e-06, ..., -8.8941e-07, + 2.3991e-06, 4.1164e-06], + [ 7.4729e-06, 4.5955e-05, 6.5938e-06, ..., 2.0396e-07, + 2.2370e-06, 8.6486e-05], + ..., + [ 9.2387e-06, 1.0751e-05, -1.0341e-05, ..., 6.9477e-07, + 2.3302e-06, 7.8008e-06], + [ 2.6655e-04, 9.8324e-04, 7.0810e-05, ..., 1.6382e-06, + 3.7813e-04, 5.5885e-04], + [-3.1281e-04, -1.3075e-03, -8.5413e-05, ..., 1.3309e-06, + -4.1294e-04, -9.7179e-04]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0042, -0.0183, -0.0170, 0.0318, -0.0326, 0.0289, -0.0274, -0.0128, + 0.0237, -0.0245], device='cuda:0'), grad: tensor([ 1.9920e-04, 1.7613e-05, 1.4114e-04, -7.9572e-05, 1.1599e-04, + 6.0415e-04, -4.3988e-04, 3.0443e-05, 1.6136e-03, -2.2011e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 215.02, cls_loss 0.0048 cls_loss_mapping 0.0089 cls_loss_causal 0.5938 re_mapping 0.0095 re_causal 0.0292 /// teacc 98.84 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0363, -0.0646, -0.0501, ..., -0.1002, -0.0065, 0.0603], + [ 0.0379, 0.0486, 0.0295, ..., 0.0645, -0.0870, -0.0519], + [ 0.0398, 0.0342, -0.0661, ..., 0.0065, -0.0950, -0.0190], + ..., + [-0.0292, -0.0331, 0.0582, ..., -0.0672, -0.0136, 0.0171], + [ 0.0052, 0.0713, 0.0942, ..., -0.0653, -0.0658, -0.0937], + [-0.0796, 0.0073, -0.1292, ..., -0.0533, -0.0490, 0.0503]], + device='cuda:0'), grad: tensor([[ 2.5798e-06, 3.1497e-06, 2.1607e-06, ..., 8.4331e-07, + 5.7817e-06, 1.5488e-06], + [ 9.5926e-07, 9.1037e-07, 3.8624e-05, ..., 5.0804e-07, + 2.4177e-06, 2.5824e-05], + [-8.7842e-06, -4.5262e-06, 5.3383e-06, ..., 5.6019e-07, + 2.3246e-06, 1.6624e-06], + ..., + [ 3.9488e-07, 4.0187e-07, -6.3419e-05, ..., 2.6426e-07, + 8.9221e-07, -3.9726e-05], + [ 7.3500e-06, 9.0301e-06, 2.2035e-06, ..., 7.7561e-06, + 1.5303e-05, 6.6869e-06], + [ 2.7195e-06, 1.1013e-07, 1.0811e-05, ..., 1.5981e-06, + 7.5512e-06, 8.9854e-06]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0043, -0.0179, -0.0175, 0.0313, -0.0325, 0.0292, -0.0273, -0.0127, + 0.0216, -0.0232], device='cuda:0'), grad: tensor([ 1.3776e-05, 9.2208e-05, 2.5611e-06, 1.1384e-05, 3.3658e-06, + 1.3900e-04, -2.0003e-04, -1.3649e-04, 4.0650e-05, 3.3289e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 215.11, cls_loss 0.0050 cls_loss_mapping 0.0115 cls_loss_causal 0.5966 re_mapping 0.0094 re_causal 0.0283 /// teacc 98.79 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0373, -0.0649, -0.0508, ..., -0.1015, -0.0066, 0.0606], + [ 0.0379, 0.0494, 0.0296, ..., 0.0647, -0.0866, -0.0516], + [ 0.0406, 0.0343, -0.0664, ..., 0.0065, -0.0962, -0.0198], + ..., + [-0.0293, -0.0338, 0.0589, ..., -0.0677, -0.0143, 0.0179], + [ 0.0049, 0.0715, 0.0944, ..., -0.0656, -0.0659, -0.0949], + [-0.0809, 0.0071, -0.1300, ..., -0.0541, -0.0494, 0.0503]], + device='cuda:0'), grad: tensor([[ 1.2191e-06, 5.7463e-07, 8.9500e-07, ..., 3.0361e-07, + -5.0217e-06, -3.8922e-05], + [ 6.2538e-07, 9.3319e-07, 2.0061e-06, ..., -6.9663e-07, + 7.3668e-07, 1.1157e-06], + [-7.0810e-05, -2.7101e-07, -3.5822e-05, ..., 6.7987e-07, + 9.5367e-07, 2.7176e-06], + ..., + [ 1.1381e-06, 2.9039e-06, -2.4334e-05, ..., 4.2887e-07, + 4.0582e-07, 1.8897e-06], + [ 3.5971e-05, -5.2124e-05, -2.6658e-05, ..., -1.4924e-05, + -1.1642e-06, 3.9451e-06], + [ 5.5227e-07, -3.7700e-06, 4.4852e-06, ..., 2.1909e-07, + 1.8915e-06, 2.6282e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0042, -0.0177, -0.0172, 0.0319, -0.0324, 0.0284, -0.0271, -0.0123, + 0.0210, -0.0237], device='cuda:0'), grad: tensor([-4.2677e-05, 5.9344e-06, -7.9751e-05, 7.2002e-05, 1.7643e-05, + -1.9276e-04, 2.4748e-04, -3.3468e-05, -6.8322e-06, 1.1988e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 215.06, cls_loss 0.0049 cls_loss_mapping 0.0103 cls_loss_causal 0.5732 re_mapping 0.0097 re_causal 0.0280 /// teacc 98.96 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0383, -0.0653, -0.0515, ..., -0.1016, -0.0064, 0.0609], + [ 0.0383, 0.0499, 0.0302, ..., 0.0653, -0.0869, -0.0522], + [ 0.0403, 0.0344, -0.0664, ..., 0.0063, -0.0966, -0.0196], + ..., + [-0.0306, -0.0357, 0.0586, ..., -0.0687, -0.0147, 0.0169], + [ 0.0055, 0.0721, 0.0952, ..., -0.0659, -0.0660, -0.0933], + [-0.0832, 0.0075, -0.1306, ..., -0.0543, -0.0498, 0.0508]], + device='cuda:0'), grad: tensor([[ 1.5311e-06, 1.5069e-06, 1.5810e-05, ..., 1.3188e-06, + 4.9323e-06, 8.1286e-06], + [ 6.6357e-07, 6.3842e-07, 1.3910e-05, ..., -4.8801e-06, + 2.3786e-06, 1.1504e-05], + [-2.8964e-06, 1.7695e-06, 1.3314e-05, ..., 1.0682e-06, + 3.3956e-06, 7.6666e-06], + ..., + [-5.1335e-06, 1.6866e-06, -1.9193e-04, ..., -1.5758e-06, + -4.8339e-05, -1.4997e-04], + [-1.2435e-05, -1.1660e-05, -1.1787e-05, ..., 9.5461e-07, + 2.0396e-07, 5.5805e-06], + [ 1.0155e-05, 2.8536e-06, 6.2108e-05, ..., 1.6345e-06, + 1.3292e-05, 4.0621e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0042, -0.0174, -0.0175, 0.0329, -0.0324, 0.0288, -0.0277, -0.0136, + 0.0225, -0.0240], device='cuda:0'), grad: tensor([ 3.4869e-05, 2.9311e-05, 1.8910e-05, 1.7691e-04, 2.0698e-05, + 5.0068e-05, -4.4256e-06, -4.5490e-04, -1.2018e-05, 1.4102e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 214.97, cls_loss 0.0041 cls_loss_mapping 0.0105 cls_loss_causal 0.5656 re_mapping 0.0095 re_causal 0.0271 /// teacc 98.79 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0389, -0.0659, -0.0525, ..., -0.1018, -0.0055, 0.0616], + [ 0.0389, 0.0498, 0.0308, ..., 0.0657, -0.0873, -0.0520], + [ 0.0398, 0.0341, -0.0675, ..., 0.0063, -0.0972, -0.0204], + ..., + [-0.0315, -0.0361, 0.0586, ..., -0.0692, -0.0149, 0.0173], + [ 0.0065, 0.0729, 0.0968, ..., -0.0660, -0.0665, -0.0929], + [-0.0840, 0.0074, -0.1319, ..., -0.0548, -0.0504, 0.0506]], + device='cuda:0'), grad: tensor([[ 1.8002e-06, 2.4568e-06, 2.6748e-06, ..., 1.0198e-06, + 6.3051e-07, 1.3821e-06], + [ 3.1090e-04, 1.1688e-04, 5.8174e-04, ..., 1.2234e-05, + 1.8366e-06, 3.0436e-06], + [ 2.1935e-05, 3.5226e-05, 4.8786e-05, ..., 1.4059e-05, + 1.1409e-06, 9.8571e-06], + ..., + [-3.7122e-04, -7.6234e-05, -7.3910e-04, ..., 9.3356e-06, + 2.5574e-06, -1.1377e-05], + [-5.4874e-06, -6.2883e-05, 3.3319e-05, ..., -3.9935e-05, + 9.0003e-06, 4.1753e-05], + [ 1.3798e-05, -1.1426e-04, -7.3537e-06, ..., 2.6226e-06, + -2.9624e-05, -1.4842e-04]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0046, -0.0171, -0.0183, 0.0330, -0.0323, 0.0290, -0.0279, -0.0136, + 0.0233, -0.0247], device='cuda:0'), grad: tensor([ 7.6890e-06, 6.1178e-04, 7.0274e-05, 2.1204e-05, 1.9562e-04, + 3.3140e-05, 8.2096e-07, -7.2336e-04, 5.4508e-05, -2.7156e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 215.14, cls_loss 0.0039 cls_loss_mapping 0.0079 cls_loss_causal 0.6011 re_mapping 0.0095 re_causal 0.0289 /// teacc 98.78 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0393, -0.0662, -0.0529, ..., -0.1026, -0.0055, 0.0619], + [ 0.0388, 0.0495, 0.0306, ..., 0.0657, -0.0878, -0.0520], + [ 0.0396, 0.0337, -0.0690, ..., 0.0058, -0.0986, -0.0207], + ..., + [-0.0313, -0.0363, 0.0594, ..., -0.0691, -0.0143, 0.0174], + [ 0.0078, 0.0740, 0.0987, ..., -0.0652, -0.0665, -0.0933], + [-0.0844, 0.0074, -0.1325, ..., -0.0553, -0.0510, 0.0510]], + device='cuda:0'), grad: tensor([[ 2.3171e-06, 2.6822e-06, 3.2801e-06, ..., 4.6529e-06, + 3.6154e-06, -7.6830e-05], + [-1.1325e-04, -8.6427e-05, -1.0693e-04, ..., -1.1456e-04, + 5.1223e-07, 1.5631e-05], + [ 1.4615e-04, 4.0948e-05, 8.2254e-05, ..., 4.2319e-05, + 1.7919e-06, 6.4552e-05], + ..., + [-2.9728e-05, -6.7353e-05, -1.2898e-04, ..., 5.1968e-06, + 3.2922e-07, -5.2154e-05], + [ 7.7784e-05, 4.7237e-05, 6.9499e-05, ..., 4.8935e-05, + 2.5928e-06, 5.9716e-06], + [ 2.7105e-05, 3.5703e-05, 7.2360e-05, ..., 1.2908e-06, + 1.9046e-07, 2.4751e-05]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0047, -0.0176, -0.0188, 0.0323, -0.0324, 0.0290, -0.0278, -0.0128, + 0.0244, -0.0247], device='cuda:0'), grad: tensor([-1.0043e-04, -1.1122e-04, 3.4857e-04, -2.3329e-04, 2.9564e-05, + 2.6315e-05, 2.1756e-06, -2.4581e-04, 1.3840e-04, 1.4555e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 214.71, cls_loss 0.0037 cls_loss_mapping 0.0089 cls_loss_causal 0.5976 re_mapping 0.0089 re_causal 0.0275 /// teacc 98.86 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0399, -0.0667, -0.0533, ..., -0.1030, -0.0056, 0.0622], + [ 0.0389, 0.0495, 0.0310, ..., 0.0664, -0.0868, -0.0503], + [ 0.0401, 0.0337, -0.0692, ..., 0.0056, -0.0991, -0.0211], + ..., + [-0.0315, -0.0364, 0.0597, ..., -0.0697, -0.0149, 0.0171], + [ 0.0078, 0.0742, 0.0988, ..., -0.0652, -0.0671, -0.0938], + [-0.0849, 0.0076, -0.1336, ..., -0.0557, -0.0515, 0.0511]], + device='cuda:0'), grad: tensor([[ 6.4867e-07, 1.0310e-06, 1.0720e-06, ..., 2.7139e-06, + 4.8317e-06, -6.8033e-07], + [ 3.6601e-06, 4.1835e-06, 6.4485e-06, ..., 1.7490e-06, + 1.1921e-06, 4.9407e-07], + [-1.4910e-06, 1.7798e-06, 2.7586e-06, ..., 8.6008e-07, + 1.0449e-06, 6.2445e-07], + ..., + [ 1.8068e-06, 2.1830e-06, -1.0477e-07, ..., 6.5099e-07, + 4.3213e-07, 1.7844e-06], + [-2.6822e-05, -2.1875e-05, -3.0816e-05, ..., -1.2452e-06, + -2.6263e-06, -4.5635e-07], + [ 2.8647e-06, -1.7146e-06, 3.2261e-06, ..., 1.4538e-06, + 1.9427e-06, -8.2031e-06]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0046, -0.0168, -0.0185, 0.0326, -0.0325, 0.0288, -0.0279, -0.0132, + 0.0241, -0.0251], device='cuda:0'), grad: tensor([ 6.1467e-06, 9.6262e-06, -2.4028e-07, 3.0503e-05, 1.7602e-06, + 1.2003e-05, -8.2925e-06, 4.5970e-06, -5.3495e-05, -2.5965e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 214.98, cls_loss 0.0039 cls_loss_mapping 0.0083 cls_loss_causal 0.5718 re_mapping 0.0088 re_causal 0.0266 /// teacc 98.94 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0402, -0.0672, -0.0535, ..., -0.1049, -0.0064, 0.0629], + [ 0.0387, 0.0490, 0.0310, ..., 0.0663, -0.0873, -0.0513], + [ 0.0402, 0.0336, -0.0697, ..., 0.0056, -0.0996, -0.0222], + ..., + [-0.0317, -0.0370, 0.0599, ..., -0.0700, -0.0151, 0.0169], + [ 0.0085, 0.0748, 0.0996, ..., -0.0653, -0.0672, -0.0941], + [-0.0852, 0.0088, -0.1342, ..., -0.0559, -0.0518, 0.0522]], + device='cuda:0'), grad: tensor([[ 1.3262e-06, 2.1867e-06, 4.1863e-07, ..., 1.7956e-06, + -6.8367e-05, -2.7585e-04], + [-5.4985e-06, -2.9132e-06, -1.7630e-06, ..., -3.7178e-06, + 7.6219e-06, 5.6662e-06], + [-1.2301e-05, -6.1095e-06, 1.1614e-06, ..., 3.0808e-06, + 1.5765e-05, 4.6462e-05], + ..., + [ 2.3618e-06, 3.3639e-06, -8.6948e-06, ..., 6.3740e-06, + 4.5635e-06, 2.4047e-06], + [ 7.0073e-06, 4.3921e-06, -6.7567e-07, ..., 1.5460e-06, + 3.1918e-05, 7.1645e-05], + [ 2.4289e-06, 3.3844e-06, 5.6475e-06, ..., 4.4033e-06, + 2.5973e-05, 7.6115e-05]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0046, -0.0174, -0.0187, 0.0327, -0.0329, 0.0280, -0.0271, -0.0132, + 0.0244, -0.0243], device='cuda:0'), grad: tensor([-3.1638e-04, 8.9630e-06, 3.8087e-05, -1.1645e-05, -8.3447e-06, + -1.3018e-04, 1.9765e-04, 4.4927e-06, 1.1128e-04, 1.0610e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 215.13, cls_loss 0.0039 cls_loss_mapping 0.0091 cls_loss_causal 0.5324 re_mapping 0.0091 re_causal 0.0253 /// teacc 98.88 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0406, -0.0673, -0.0530, ..., -0.1052, -0.0060, 0.0634], + [ 0.0386, 0.0488, 0.0309, ..., 0.0667, -0.0876, -0.0517], + [ 0.0405, 0.0337, -0.0699, ..., 0.0054, -0.1002, -0.0226], + ..., + [-0.0314, -0.0369, 0.0605, ..., -0.0701, -0.0153, 0.0171], + [ 0.0089, 0.0765, 0.1001, ..., -0.0656, -0.0668, -0.0946], + [-0.0858, 0.0085, -0.1351, ..., -0.0568, -0.0522, 0.0526]], + device='cuda:0'), grad: tensor([[ 2.4974e-05, 1.4976e-05, 1.2256e-05, ..., 8.4564e-07, + -1.8319e-06, -6.6608e-06], + [ 1.4178e-05, 9.3803e-06, 1.3992e-05, ..., 1.7611e-06, + 1.8282e-06, 7.5996e-06], + [-5.6744e-05, 2.5909e-06, 8.7172e-06, ..., 1.6876e-06, + 2.8014e-06, -8.8587e-06], + ..., + [ 2.7284e-05, 6.6608e-06, -2.2858e-05, ..., 4.5076e-06, + 2.2948e-06, -1.3933e-05], + [-1.1855e-04, -8.8394e-05, -8.2374e-05, ..., -9.6206e-07, + 3.8184e-07, 1.8310e-06], + [ 2.1622e-05, 2.2441e-05, 2.1368e-05, ..., 1.3225e-05, + 7.2829e-06, 1.0096e-05]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0053, -0.0179, -0.0185, 0.0322, -0.0325, 0.0282, -0.0284, -0.0127, + 0.0252, -0.0246], device='cuda:0'), grad: tensor([ 6.5863e-05, 6.3181e-05, -6.4552e-05, 2.1851e-04, -2.1532e-05, + -1.7107e-05, 2.6077e-05, 5.0617e-07, -3.8314e-04, 1.1182e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 214.67, cls_loss 0.0039 cls_loss_mapping 0.0085 cls_loss_causal 0.5938 re_mapping 0.0086 re_causal 0.0269 /// teacc 98.96 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0415, -0.0676, -0.0539, ..., -0.1059, -0.0061, 0.0633], + [ 0.0390, 0.0490, 0.0312, ..., 0.0672, -0.0881, -0.0517], + [ 0.0409, 0.0338, -0.0701, ..., 0.0050, -0.1006, -0.0238], + ..., + [-0.0319, -0.0373, 0.0606, ..., -0.0707, -0.0154, 0.0171], + [ 0.0088, 0.0766, 0.1006, ..., -0.0657, -0.0675, -0.0951], + [-0.0865, 0.0086, -0.1361, ..., -0.0571, -0.0527, 0.0531]], + device='cuda:0'), grad: tensor([[ 5.0031e-06, 3.4198e-06, 3.1153e-07, ..., 2.9430e-07, + -9.5554e-07, 2.3422e-03], + [ 1.8245e-06, 4.8894e-07, -2.9802e-06, ..., -5.0366e-06, + 2.0489e-07, 2.9709e-06], + [-6.5088e-05, -2.2233e-05, -3.1088e-06, ..., 9.4436e-07, + 3.2503e-07, 3.0726e-05], + ..., + [ 2.4810e-05, 4.3437e-06, 6.1840e-07, ..., 2.3916e-06, + 2.2678e-07, 8.1956e-06], + [ 2.2694e-05, 1.2323e-05, 1.3765e-06, ..., 7.7253e-07, + 5.2201e-07, 1.9163e-05], + [ 1.2470e-06, -1.1824e-05, 1.1632e-06, ..., 3.8603e-07, + 7.2503e-07, -2.4872e-03]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0051, -0.0177, -0.0184, 0.0320, -0.0327, 0.0289, -0.0281, -0.0130, + 0.0249, -0.0247], device='cuda:0'), grad: tensor([ 2.4681e-03, 7.9870e-06, -1.2118e-04, 1.1593e-05, 6.7174e-05, + 1.7628e-05, 2.5243e-05, 6.6280e-05, 6.9380e-05, -2.6131e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 214.82, cls_loss 0.0044 cls_loss_mapping 0.0091 cls_loss_causal 0.5661 re_mapping 0.0090 re_causal 0.0263 /// teacc 98.88 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0425, -0.0682, -0.0553, ..., -0.1066, -0.0062, 0.0629], + [ 0.0391, 0.0489, 0.0311, ..., 0.0669, -0.0888, -0.0518], + [ 0.0413, 0.0342, -0.0701, ..., 0.0053, -0.1010, -0.0248], + ..., + [-0.0322, -0.0382, 0.0611, ..., -0.0707, -0.0155, 0.0166], + [ 0.0090, 0.0773, 0.1015, ..., -0.0664, -0.0678, -0.0955], + [-0.0867, 0.0083, -0.1375, ..., -0.0581, -0.0534, 0.0542]], + device='cuda:0'), grad: tensor([[ 1.6009e-06, 3.3770e-06, 1.5721e-06, ..., 2.6301e-06, + 2.6133e-06, -1.7527e-06], + [-1.0379e-05, -9.4920e-06, -1.4611e-05, ..., -8.0615e-06, + 1.9278e-06, 1.5441e-06], + [ 7.7635e-06, 1.1198e-05, 7.2531e-06, ..., 7.9498e-06, + 4.3288e-06, 3.8855e-06], + ..., + [ 7.2718e-06, 2.9840e-06, 3.2987e-06, ..., 2.7679e-06, + 4.2543e-06, 8.1435e-06], + [ 1.9178e-05, 5.2117e-06, 1.0818e-05, ..., 5.2825e-06, + 1.1317e-05, 1.7822e-05], + [ 1.7747e-05, 2.4308e-06, 8.6129e-06, ..., 2.1253e-06, + 9.1791e-06, 2.1681e-05]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0042, -0.0181, -0.0181, 0.0315, -0.0319, 0.0294, -0.0288, -0.0132, + 0.0253, -0.0244], device='cuda:0'), grad: tensor([ 1.0043e-05, -1.0714e-05, 3.8356e-05, -2.9516e-04, 5.9791e-06, + 3.9399e-05, -3.3528e-05, 3.4899e-05, 1.0884e-04, 1.0192e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.89, cls_loss 0.0041 cls_loss_mapping 0.0098 cls_loss_causal 0.5628 re_mapping 0.0088 re_causal 0.0258 /// teacc 99.02 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0431, -0.0704, -0.0559, ..., -0.1072, -0.0063, 0.0605], + [ 0.0397, 0.0490, 0.0314, ..., 0.0674, -0.0895, -0.0523], + [ 0.0412, 0.0341, -0.0711, ..., 0.0047, -0.1016, -0.0256], + ..., + [-0.0323, -0.0385, 0.0618, ..., -0.0710, -0.0156, 0.0168], + [ 0.0087, 0.0771, 0.1013, ..., -0.0674, -0.0687, -0.0964], + [-0.0870, 0.0086, -0.1384, ..., -0.0585, -0.0537, 0.0569]], + device='cuda:0'), grad: tensor([[ 5.2154e-07, 1.6272e-05, 4.7917e-07, ..., 3.0510e-06, + 1.9625e-05, -1.1164e-04], + [-2.0146e-05, -1.2547e-05, -1.9506e-05, ..., -2.0102e-05, + 1.5674e-06, 1.9670e-06], + [ 1.0572e-05, 1.1466e-05, 1.4760e-05, ..., 1.3396e-05, + 3.0678e-06, 4.7117e-05], + ..., + [ 4.9584e-06, 1.1101e-05, -1.0759e-05, ..., 1.8865e-05, + 9.5665e-06, -3.7160e-06], + [-9.9745e-07, 8.0885e-07, -1.9334e-06, ..., 2.0042e-06, + 6.8545e-06, 9.1940e-06], + [ 1.0300e-06, -7.8417e-07, 2.1495e-06, ..., 1.5013e-06, + 3.2634e-06, 2.7969e-05]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0018, -0.0179, -0.0188, 0.0317, -0.0320, 0.0295, -0.0280, -0.0127, + 0.0241, -0.0230], device='cuda:0'), grad: tensor([-1.4460e-04, -3.1650e-05, 1.0139e-04, 1.4976e-05, -8.0094e-06, + 7.5221e-05, -8.6844e-05, 5.2415e-06, 2.2292e-05, 5.2035e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 214.98, cls_loss 0.0033 cls_loss_mapping 0.0070 cls_loss_causal 0.5860 re_mapping 0.0085 re_causal 0.0259 /// teacc 98.88 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0437, -0.0707, -0.0562, ..., -0.1072, -0.0048, 0.0612], + [ 0.0399, 0.0490, 0.0315, ..., 0.0676, -0.0901, -0.0523], + [ 0.0414, 0.0341, -0.0714, ..., 0.0044, -0.1021, -0.0258], + ..., + [-0.0326, -0.0390, 0.0619, ..., -0.0715, -0.0160, 0.0167], + [ 0.0086, 0.0772, 0.1017, ..., -0.0681, -0.0692, -0.0969], + [-0.0874, 0.0079, -0.1390, ..., -0.0599, -0.0551, 0.0568]], + device='cuda:0'), grad: tensor([[ 6.8545e-07, 6.9663e-07, 7.0315e-08, ..., -7.5158e-07, + -3.1739e-06, -5.7369e-06], + [-1.1083e-07, 2.3190e-07, -3.9954e-07, ..., -2.9569e-07, + 2.4447e-07, 8.2701e-07], + [-6.1430e-06, -4.9435e-06, 2.8405e-06, ..., 5.9046e-07, + 5.7463e-07, 7.5344e-07], + ..., + [-5.1446e-06, 2.1141e-06, -1.0855e-05, ..., 1.0803e-06, + 6.2026e-07, 3.5278e-06], + [ 9.4399e-06, 8.1584e-06, 5.6112e-07, ..., 8.1817e-07, + 1.3784e-06, 5.9009e-06], + [ 7.1013e-07, -9.9763e-06, 4.8010e-07, ..., 9.8441e-07, + 1.0012e-06, -1.7911e-05]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0028, -0.0179, -0.0186, 0.0319, -0.0307, 0.0293, -0.0285, -0.0130, + 0.0237, -0.0238], device='cuda:0'), grad: tensor([-6.7167e-06, 2.4997e-06, -3.2075e-06, 8.2403e-06, 1.6078e-05, + 5.4538e-06, 2.1905e-06, -9.3728e-06, 3.1441e-05, -4.6641e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 215.23, cls_loss 0.0038 cls_loss_mapping 0.0073 cls_loss_causal 0.5548 re_mapping 0.0090 re_causal 0.0262 /// teacc 98.94 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0441, -0.0710, -0.0567, ..., -0.1075, -0.0047, 0.0614], + [ 0.0405, 0.0490, 0.0317, ..., 0.0680, -0.0911, -0.0530], + [ 0.0412, 0.0339, -0.0719, ..., 0.0039, -0.1026, -0.0264], + ..., + [-0.0332, -0.0392, 0.0618, ..., -0.0720, -0.0163, 0.0168], + [ 0.0088, 0.0775, 0.1022, ..., -0.0684, -0.0698, -0.0971], + [-0.0880, 0.0081, -0.1398, ..., -0.0602, -0.0551, 0.0571]], + device='cuda:0'), grad: tensor([[ 2.0817e-05, 6.1607e-07, 1.3299e-06, ..., 1.4780e-06, + -1.9297e-06, -8.6948e-06], + [ 6.0275e-06, 1.0729e-06, 1.4782e-05, ..., 3.5558e-06, + 4.0010e-06, 7.4180e-07], + [ 9.7454e-06, 1.2899e-06, 7.6294e-05, ..., 3.4831e-06, + 1.7937e-06, -6.5118e-06], + ..., + [-7.1943e-05, 1.5562e-06, -1.6701e-04, ..., 4.2766e-06, + 1.8394e-06, -1.0645e-06], + [ 9.6187e-06, 4.5681e-07, 1.6242e-05, ..., 4.6343e-06, + 5.9009e-06, 2.2147e-06], + [ 1.5110e-05, 3.9667e-05, 3.0190e-05, ..., 1.1355e-04, + 5.4181e-05, 5.0291e-06]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0032, -0.0180, -0.0192, 0.0326, -0.0310, 0.0302, -0.0287, -0.0140, + 0.0236, -0.0236], device='cuda:0'), grad: tensor([ 1.0604e-04, 5.4628e-05, 1.5870e-05, 4.9740e-05, -2.4676e-04, + 1.2860e-05, 1.4350e-05, -3.5858e-04, 5.7966e-05, 2.9349e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 215.32, cls_loss 0.0038 cls_loss_mapping 0.0088 cls_loss_causal 0.5732 re_mapping 0.0085 re_causal 0.0252 /// teacc 98.85 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0450, -0.0713, -0.0564, ..., -0.1078, -0.0048, 0.0616], + [ 0.0405, 0.0492, 0.0318, ..., 0.0686, -0.0915, -0.0532], + [ 0.0413, 0.0337, -0.0725, ..., 0.0035, -0.1033, -0.0267], + ..., + [-0.0327, -0.0395, 0.0626, ..., -0.0721, -0.0164, 0.0170], + [ 0.0089, 0.0776, 0.1026, ..., -0.0688, -0.0704, -0.0976], + [-0.0887, 0.0069, -0.1410, ..., -0.0620, -0.0564, 0.0573]], + device='cuda:0'), grad: tensor([[ 3.5986e-06, 5.6103e-06, 8.6203e-06, ..., 2.4997e-06, + -2.5928e-06, -1.0155e-05], + [-6.8665e-05, -5.7220e-05, -1.7476e-04, ..., -9.7454e-05, + -9.3207e-06, 6.3442e-06], + [-1.4156e-07, 3.0082e-06, 6.9849e-06, ..., 3.3360e-06, + 9.3039e-07, 9.1922e-07], + ..., + [ 2.7139e-06, 1.7956e-06, -1.8090e-05, ..., 2.4494e-06, + -6.6962e-07, -5.1335e-06], + [ 9.4771e-06, 2.4159e-06, 2.8029e-05, ..., 1.8865e-05, + 1.6525e-05, 9.3132e-06], + [ 1.6289e-06, 1.2685e-06, 8.4043e-06, ..., 1.3085e-06, + 6.2644e-05, 6.7949e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0035, -0.0182, -0.0192, 0.0328, -0.0293, 0.0301, -0.0290, -0.0135, + 0.0232, -0.0248], device='cuda:0'), grad: tensor([ 2.6803e-06, -2.4652e-04, 8.5533e-06, 1.0544e-04, 3.3438e-05, + -3.3021e-04, 2.5320e-04, -2.8059e-05, 6.1631e-05, 1.3983e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 215.08, cls_loss 0.0038 cls_loss_mapping 0.0084 cls_loss_causal 0.5827 re_mapping 0.0081 re_causal 0.0252 /// teacc 98.93 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0459, -0.0721, -0.0571, ..., -0.1082, -0.0062, 0.0613], + [ 0.0401, 0.0486, 0.0317, ..., 0.0692, -0.0917, -0.0534], + [ 0.0416, 0.0340, -0.0727, ..., 0.0034, -0.1037, -0.0267], + ..., + [-0.0331, -0.0397, 0.0628, ..., -0.0724, -0.0166, 0.0169], + [ 0.0102, 0.0779, 0.1040, ..., -0.0693, -0.0721, -0.0983], + [-0.0890, 0.0070, -0.1413, ..., -0.0625, -0.0571, 0.0576]], + device='cuda:0'), grad: tensor([[ 3.5437e-07, 5.9698e-07, 2.3609e-07, ..., 4.3958e-07, + -5.4482e-07, -7.6257e-06], + [-6.2361e-06, -1.7677e-06, -7.7188e-06, ..., -2.2110e-06, + 7.1013e-07, -3.0976e-06], + [-8.6799e-07, 1.1642e-07, 1.2955e-06, ..., 8.8802e-07, + 8.7731e-07, 7.7905e-07], + ..., + [ 1.9707e-06, 8.4797e-07, -2.7148e-07, ..., 1.2964e-06, + 8.8103e-07, 5.2294e-07], + [ 5.1633e-06, 3.0734e-06, 2.2482e-06, ..., 2.3432e-06, + 3.1404e-06, 1.9614e-06], + [ 2.3153e-06, 1.3057e-06, 3.6061e-06, ..., 1.5376e-06, + 1.2619e-06, 3.7737e-06]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0027, -0.0191, -0.0186, 0.0329, -0.0291, 0.0303, -0.0285, -0.0136, + 0.0235, -0.0248], device='cuda:0'), grad: tensor([-7.7263e-06, -9.2462e-06, 1.9632e-06, -4.7088e-06, 9.3598e-07, + -2.5239e-06, -7.1265e-06, 2.5779e-06, 1.6421e-05, 9.3579e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 215.22, cls_loss 0.0034 cls_loss_mapping 0.0066 cls_loss_causal 0.5553 re_mapping 0.0084 re_causal 0.0244 /// teacc 98.92 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0463, -0.0726, -0.0575, ..., -0.1085, -0.0056, 0.0618], + [ 0.0402, 0.0486, 0.0319, ..., 0.0695, -0.0922, -0.0535], + [ 0.0420, 0.0343, -0.0731, ..., 0.0033, -0.1044, -0.0273], + ..., + [-0.0331, -0.0400, 0.0636, ..., -0.0725, -0.0166, 0.0170], + [ 0.0098, 0.0778, 0.1039, ..., -0.0698, -0.0729, -0.0989], + [-0.0897, 0.0070, -0.1426, ..., -0.0630, -0.0576, 0.0578]], + device='cuda:0'), grad: tensor([[ 2.0918e-06, -2.9698e-05, 4.1500e-06, ..., 1.4277e-06, + -9.5069e-05, -1.5354e-04], + [ 6.1691e-06, -4.4480e-06, 2.0683e-05, ..., -5.6550e-06, + 1.1940e-06, 2.4168e-07], + [ 2.9981e-05, 2.3618e-06, 6.3062e-05, ..., 1.0952e-06, + 2.4550e-06, 3.8296e-06], + ..., + [-1.6510e-04, 2.1141e-06, -3.4356e-04, ..., 1.9390e-06, + 6.3423e-07, 2.2314e-06], + [ 4.9889e-05, 1.9148e-05, 1.0151e-04, ..., 1.2023e-06, + 1.6272e-05, 3.6150e-05], + [ 7.3351e-06, -2.6613e-05, 1.2584e-05, ..., 1.1725e-06, + 1.5527e-05, -9.1791e-06]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0034, -0.0192, -0.0186, 0.0333, -0.0296, 0.0298, -0.0280, -0.0129, + 0.0227, -0.0250], device='cuda:0'), grad: tensor([-3.5930e-04, 3.0816e-05, 9.7990e-05, 2.5630e-04, 7.1108e-05, + 6.2704e-05, 8.5175e-05, -4.8089e-04, 2.3127e-04, 4.4703e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 215.04, cls_loss 0.0031 cls_loss_mapping 0.0074 cls_loss_causal 0.5595 re_mapping 0.0082 re_causal 0.0253 /// teacc 98.90 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0468, -0.0731, -0.0580, ..., -0.1088, -0.0056, 0.0618], + [ 0.0410, 0.0491, 0.0328, ..., 0.0703, -0.0924, -0.0535], + [ 0.0421, 0.0344, -0.0736, ..., 0.0031, -0.1047, -0.0273], + ..., + [-0.0333, -0.0405, 0.0638, ..., -0.0731, -0.0169, 0.0168], + [ 0.0097, 0.0774, 0.1036, ..., -0.0708, -0.0735, -0.0993], + [-0.0899, 0.0075, -0.1439, ..., -0.0632, -0.0577, 0.0585]], + device='cuda:0'), grad: tensor([[ 6.1141e-07, 1.3486e-06, 4.9779e-07, ..., 7.4552e-07, + 2.3879e-06, -2.0722e-07], + [-4.3027e-06, -1.9781e-06, -5.4091e-06, ..., -3.6377e-06, + 8.0513e-07, 1.0412e-06], + [-3.8370e-06, -1.0813e-06, 4.9807e-06, ..., 1.2023e-06, + 4.0932e-07, 4.0652e-07], + ..., + [-2.4606e-06, 6.7009e-07, -1.4238e-05, ..., 6.5379e-07, + 2.4214e-07, -8.9332e-06], + [-7.2233e-06, -2.1338e-05, -1.1623e-05, ..., 9.8646e-06, + 4.7922e-05, 1.2480e-06], + [ 2.1160e-05, 2.8059e-05, 2.1964e-05, ..., 3.9581e-07, + 1.4231e-06, 1.4216e-05]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0033, -0.0184, -0.0183, 0.0337, -0.0301, 0.0297, -0.0278, -0.0135, + 0.0218, -0.0246], device='cuda:0'), grad: tensor([ 4.9882e-06, -1.2247e-06, 3.0473e-06, -7.6443e-06, 9.2238e-06, + -1.2350e-04, 3.5614e-05, -2.4438e-05, 3.0160e-05, 7.3910e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 215.16, cls_loss 0.0042 cls_loss_mapping 0.0085 cls_loss_causal 0.5949 re_mapping 0.0085 re_causal 0.0260 /// teacc 98.89 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0477, -0.0748, -0.0585, ..., -0.1083, -0.0051, 0.0620], + [ 0.0422, 0.0493, 0.0334, ..., 0.0706, -0.0929, -0.0537], + [ 0.0412, 0.0340, -0.0749, ..., 0.0029, -0.1051, -0.0277], + ..., + [-0.0336, -0.0408, 0.0643, ..., -0.0734, -0.0171, 0.0169], + [ 0.0116, 0.0803, 0.1050, ..., -0.0710, -0.0728, -0.1001], + [-0.0908, 0.0074, -0.1451, ..., -0.0634, -0.0581, 0.0590]], + device='cuda:0'), grad: tensor([[ 5.0396e-05, 7.7933e-06, 1.7323e-07, ..., 1.9912e-06, + 3.6936e-06, 3.8324e-07], + [-1.5153e-06, 1.5534e-06, -4.2580e-06, ..., -7.8790e-07, + 1.8226e-06, 4.8475e-07], + [ 7.8827e-06, 7.7933e-06, 1.2666e-06, ..., 3.3528e-06, + 4.6790e-06, 9.4995e-07], + ..., + [ 2.1011e-06, 8.4005e-07, -1.1781e-06, ..., 1.2964e-06, + 1.6103e-06, 2.1234e-06], + [ 1.8012e-06, 1.1206e-05, -1.9744e-07, ..., 5.5097e-06, + 9.2760e-06, 3.6601e-06], + [ 9.5926e-07, 5.9884e-07, 1.1716e-06, ..., 6.4215e-07, + 3.2019e-06, 3.9786e-06]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0032, -0.0180, -0.0193, 0.0337, -0.0301, 0.0277, -0.0287, -0.0134, + 0.0245, -0.0247], device='cuda:0'), grad: tensor([ 1.7095e-04, 1.2182e-06, 3.8475e-05, -1.8716e-04, 1.4797e-05, + -2.8282e-05, -5.7966e-05, 4.5002e-06, 3.2932e-05, 1.0572e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 214.96, cls_loss 0.0037 cls_loss_mapping 0.0095 cls_loss_causal 0.5665 re_mapping 0.0081 re_causal 0.0240 /// teacc 98.87 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0486, -0.0754, -0.0591, ..., -0.1086, -0.0049, 0.0623], + [ 0.0424, 0.0494, 0.0336, ..., 0.0709, -0.0932, -0.0542], + [ 0.0416, 0.0350, -0.0761, ..., 0.0028, -0.1055, -0.0280], + ..., + [-0.0335, -0.0414, 0.0650, ..., -0.0736, -0.0172, 0.0172], + [ 0.0110, 0.0801, 0.1053, ..., -0.0715, -0.0730, -0.1008], + [-0.0915, 0.0074, -0.1460, ..., -0.0637, -0.0583, 0.0593]], + device='cuda:0'), grad: tensor([[ 1.4575e-06, 3.6240e-05, 2.2650e-06, ..., 1.2800e-05, + 5.3823e-05, 7.6830e-05], + [ 5.0999e-06, 8.4415e-06, 9.8497e-06, ..., -1.3113e-06, + 1.7472e-06, 2.6692e-06], + [ 4.0326e-07, 6.4448e-06, 3.2596e-06, ..., 1.8161e-06, + 7.8231e-06, 1.1891e-05], + ..., + [ 2.8014e-06, 4.7535e-06, 2.0172e-06, ..., 8.9314e-07, + 2.8610e-06, 3.7346e-06], + [-2.1845e-05, -6.0648e-06, -3.8922e-05, ..., 8.7321e-06, + 3.9637e-05, 5.7995e-05], + [ 1.0841e-06, -1.5247e-04, 2.2762e-06, ..., -5.6207e-05, + -2.5368e-04, -3.7527e-04]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0035, -0.0183, -0.0191, 0.0337, -0.0302, 0.0273, -0.0286, -0.0126, + 0.0238, -0.0246], device='cuda:0'), grad: tensor([ 1.7273e-04, 1.7464e-05, 2.7314e-05, 4.0233e-05, 1.6117e-04, + 2.3782e-04, 5.6207e-05, 1.2033e-05, 7.2479e-05, -7.9823e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 214.96, cls_loss 0.0036 cls_loss_mapping 0.0091 cls_loss_causal 0.5630 re_mapping 0.0082 re_causal 0.0247 /// teacc 98.90 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0496, -0.0747, -0.0600, ..., -0.1104, -0.0037, 0.0632], + [ 0.0427, 0.0494, 0.0336, ..., 0.0712, -0.0940, -0.0548], + [ 0.0423, 0.0359, -0.0768, ..., 0.0023, -0.1062, -0.0289], + ..., + [-0.0337, -0.0416, 0.0656, ..., -0.0734, -0.0172, 0.0175], + [ 0.0106, 0.0804, 0.1062, ..., -0.0709, -0.0733, -0.1015], + [-0.0920, 0.0078, -0.1472, ..., -0.0637, -0.0588, 0.0596]], + device='cuda:0'), grad: tensor([[ 4.6333e-07, 5.7835e-07, 4.8755e-07, ..., 2.4354e-07, + 8.6147e-07, -1.3886e-06], + [ 1.0349e-05, 1.2301e-05, 1.3568e-05, ..., 2.2491e-07, + 5.3970e-07, 4.7404e-07], + [-7.8529e-06, -4.0270e-06, 1.1371e-06, ..., 8.0653e-07, + 7.0920e-07, 4.3446e-07], + ..., + [ 2.1271e-06, 2.2426e-06, -6.7912e-06, ..., 8.2003e-07, + 8.6008e-07, -7.4878e-07], + [-8.6352e-06, -2.6062e-05, -3.7909e-05, ..., 9.3877e-07, + 6.2883e-06, 3.1721e-06], + [ 6.3377e-07, 7.7812e-07, 3.0361e-06, ..., 9.7789e-07, + 2.8163e-06, 2.6450e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0046, -0.0185, -0.0185, 0.0331, -0.0309, 0.0277, -0.0294, -0.0121, + 0.0238, -0.0246], device='cuda:0'), grad: tensor([-1.0990e-06, 2.6286e-05, -7.5586e-06, 3.8028e-05, -5.3495e-06, + -1.4499e-05, 1.2726e-05, -2.0325e-05, -4.2886e-05, 1.4693e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 215.30, cls_loss 0.0057 cls_loss_mapping 0.0110 cls_loss_causal 0.5656 re_mapping 0.0087 re_causal 0.0240 /// teacc 99.04 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0510, -0.0774, -0.0608, ..., -0.1133, -0.0049, 0.0634], + [ 0.0433, 0.0507, 0.0339, ..., 0.0728, -0.0951, -0.0542], + [ 0.0423, 0.0372, -0.0784, ..., 0.0003, -0.1071, -0.0299], + ..., + [-0.0327, -0.0418, 0.0671, ..., -0.0729, -0.0176, 0.0174], + [ 0.0098, 0.0791, 0.1060, ..., -0.0714, -0.0746, -0.1042], + [-0.0926, 0.0085, -0.1483, ..., -0.0642, -0.0585, 0.0607]], + device='cuda:0'), grad: tensor([[ 1.4342e-06, 5.9977e-07, 2.4512e-06, ..., 8.4564e-07, + 4.7497e-07, 1.8537e-04], + [-2.5053e-07, -9.2387e-06, 5.6773e-05, ..., -2.0832e-05, + 1.2517e-06, 5.6863e-05], + [ 1.2636e-04, 2.0526e-06, 2.2376e-04, ..., 4.9472e-06, + 9.2015e-07, 3.2365e-05], + ..., + [-1.7810e-04, 3.2242e-06, -4.2963e-04, ..., -2.1070e-05, + -1.9997e-05, -1.3018e-04], + [ 1.3955e-05, 1.8859e-06, 1.9982e-05, ..., 3.0641e-06, + 1.7611e-06, 1.0625e-05], + [ 1.2219e-05, -4.4741e-06, 5.0038e-05, ..., 7.9349e-06, + 6.2846e-06, -2.5153e-04]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0044, -0.0177, -0.0188, 0.0327, -0.0314, 0.0284, -0.0299, -0.0109, + 0.0222, -0.0241], device='cuda:0'), grad: tensor([ 2.3878e-04, 2.8968e-04, 4.7612e-04, 7.7963e-05, 1.7107e-04, + 8.8736e-06, 4.2468e-06, -1.1635e-03, 5.7459e-05, -1.6105e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 214.89, cls_loss 0.0039 cls_loss_mapping 0.0069 cls_loss_causal 0.5394 re_mapping 0.0085 re_causal 0.0245 /// teacc 98.88 lr 0.00010000 +Epoch 113, weight, value: tensor([[-5.1787e-02, -7.7825e-02, -6.1426e-02, ..., -1.1396e-01, + -5.2823e-03, 6.3526e-02], + [ 4.3733e-02, 5.1056e-02, 3.5040e-02, ..., 7.3322e-02, + -9.5601e-02, -5.4597e-02], + [ 4.2485e-02, 3.7121e-02, -7.9012e-02, ..., 1.1022e-04, + -1.0772e-01, -3.0432e-02], + ..., + [-3.3911e-02, -4.2557e-02, 6.6904e-02, ..., -7.2592e-02, + -1.7735e-02, 1.7373e-02], + [ 9.4391e-03, 7.8871e-02, 1.0631e-01, ..., -7.2099e-02, + -7.5467e-02, -1.0526e-01], + [-9.2680e-02, 9.2090e-03, -1.5009e-01, ..., -6.4699e-02, + -5.8391e-02, 6.1145e-02]], device='cuda:0'), grad: tensor([[ 5.3551e-07, 4.0280e-07, -2.8079e-07, ..., -8.0559e-08, + 6.9618e-05, -4.1276e-06], + [-5.8711e-06, -3.4589e-06, -5.1521e-06, ..., -5.8524e-06, + 4.4890e-06, 9.3207e-06], + [-5.2676e-06, 3.5157e-07, 8.5169e-07, ..., 1.0543e-06, + 7.6294e-06, 6.9551e-06], + ..., + [ 1.2675e-06, 1.6131e-06, -6.5938e-07, ..., 1.1437e-06, + 2.7474e-06, 9.3579e-06], + [ 2.1942e-06, 1.1306e-06, 8.1398e-07, ..., 8.6287e-07, + 1.6123e-05, 2.3812e-05], + [ 3.4692e-07, -9.7677e-06, -1.2210e-06, ..., 9.9465e-07, + 1.0155e-05, -3.1739e-05]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0042, -0.0174, -0.0189, 0.0331, -0.0314, 0.0295, -0.0310, -0.0110, + 0.0214, -0.0240], device='cuda:0'), grad: tensor([ 1.4387e-05, 9.4920e-06, 6.6273e-06, 4.1795e-04, 5.7787e-05, + 2.1191e-03, -2.6455e-03, 2.2858e-05, 5.7936e-05, -6.2704e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 214.60, cls_loss 0.0036 cls_loss_mapping 0.0071 cls_loss_causal 0.5872 re_mapping 0.0080 re_causal 0.0246 /// teacc 99.01 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0524, -0.0782, -0.0616, ..., -0.1147, -0.0045, 0.0640], + [ 0.0434, 0.0506, 0.0346, ..., 0.0730, -0.0962, -0.0555], + [ 0.0429, 0.0375, -0.0791, ..., 0.0003, -0.1084, -0.0306], + ..., + [-0.0340, -0.0431, 0.0674, ..., -0.0737, -0.0187, 0.0174], + [ 0.0097, 0.0793, 0.1071, ..., -0.0724, -0.0757, -0.1053], + [-0.0930, 0.0082, -0.1508, ..., -0.0654, -0.0590, 0.0614]], + device='cuda:0'), grad: tensor([[ 4.1239e-06, 6.2352e-07, 6.2585e-07, ..., 7.0594e-07, + 6.3702e-07, -7.1200e-07], + [ 1.4469e-05, 2.3423e-07, -2.5914e-07, ..., -1.5497e-06, + 5.0571e-07, 2.6845e-07], + [ 2.0161e-05, 3.2540e-06, 5.5768e-06, ..., 2.5108e-06, + 1.6708e-06, 4.8336e-07], + ..., + [ 3.3140e-05, 8.5356e-07, -8.5354e-05, ..., -1.0121e-04, + -8.8453e-05, -6.2108e-05], + [ 2.4676e-04, 8.1360e-06, 2.3812e-05, ..., 7.2690e-07, + 1.5171e-06, 1.3700e-06], + [ 7.6368e-06, 1.0608e-06, 3.7979e-06, ..., 2.8666e-06, + 3.5167e-06, 4.2208e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0046, -0.0181, -0.0184, 0.0326, -0.0294, 0.0291, -0.0307, -0.0111, + 0.0219, -0.0251], device='cuda:0'), grad: tensor([ 8.9258e-06, 3.0443e-05, -2.4199e-05, -6.1989e-04, 3.7527e-04, + 2.1890e-05, -8.6427e-06, -2.6226e-04, 4.4823e-04, 2.9340e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 215.12, cls_loss 0.0033 cls_loss_mapping 0.0066 cls_loss_causal 0.5626 re_mapping 0.0076 re_causal 0.0234 /// teacc 98.93 lr 0.00010000 +Epoch 115, weight, value: tensor([[-5.3061e-02, -7.8882e-02, -6.2168e-02, ..., -1.1525e-01, + -4.6398e-03, 6.4190e-02], + [ 4.3721e-02, 5.0802e-02, 3.5325e-02, ..., 7.4097e-02, + -9.6614e-02, -5.5407e-02], + [ 4.2918e-02, 3.7386e-02, -7.9829e-02, ..., 2.4241e-05, + -1.0887e-01, -3.1259e-02], + ..., + [-3.4388e-02, -4.3693e-02, 6.7664e-02, ..., -7.4840e-02, + -1.8598e-02, 1.7805e-02], + [ 1.0271e-02, 8.0415e-02, 1.0868e-01, ..., -7.1779e-02, + -7.5909e-02, -1.0610e-01], + [-9.3884e-02, 8.4570e-03, -1.5292e-01, ..., -6.5949e-02, + -5.9305e-02, 6.1568e-02]], device='cuda:0'), grad: tensor([[ 2.9430e-07, 1.7717e-05, 2.7916e-07, ..., 1.7462e-06, + 1.1754e-04, 7.9334e-05], + [ 3.2806e-07, 8.1444e-07, -1.4505e-07, ..., 2.7893e-07, + 9.7603e-07, 9.6392e-07], + [-1.8231e-07, 1.2936e-06, 5.7695e-07, ..., 8.1491e-07, + 1.6484e-06, 7.7998e-07], + ..., + [ 4.3004e-07, 9.4203e-07, 3.0920e-07, ..., 1.2349e-06, + 7.5623e-07, 1.7136e-07], + [-1.3784e-06, -1.7118e-06, -1.8608e-06, ..., 5.2340e-07, + 4.1723e-06, 2.9076e-06], + [ 1.2815e-06, 3.5074e-06, 3.2876e-07, ..., 3.5930e-06, + 5.5693e-06, 2.8051e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0045, -0.0176, -0.0186, 0.0330, -0.0299, 0.0286, -0.0311, -0.0111, + 0.0226, -0.0255], device='cuda:0'), grad: tensor([ 1.4329e-04, 4.3996e-06, 3.0715e-06, -4.8876e-06, -2.4110e-05, + 5.2974e-06, -1.4961e-04, 3.5986e-06, 2.0433e-06, 1.7121e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 215.33, cls_loss 0.0038 cls_loss_mapping 0.0077 cls_loss_causal 0.5406 re_mapping 0.0079 re_causal 0.0230 /// teacc 98.81 lr 0.00010000 +Epoch 116, weight, value: tensor([[-5.2841e-02, -7.9164e-02, -6.2563e-02, ..., -1.1506e-01, + -3.5961e-03, 6.3209e-02], + [ 4.3692e-02, 5.0865e-02, 3.5259e-02, ..., 7.4509e-02, + -9.7065e-02, -5.5508e-02], + [ 4.2961e-02, 3.7401e-02, -8.0385e-02, ..., -4.0967e-05, + -1.0948e-01, -3.2223e-02], + ..., + [-3.4191e-02, -4.4112e-02, 6.8397e-02, ..., -7.5396e-02, + -1.8862e-02, 1.7854e-02], + [ 9.7838e-03, 8.0252e-02, 1.0902e-01, ..., -7.2082e-02, + -7.6545e-02, -1.0778e-01], + [-9.4056e-02, 8.9261e-03, -1.5377e-01, ..., -6.6156e-02, + -5.8473e-02, 6.3807e-02]], device='cuda:0'), grad: tensor([[ 3.4296e-07, 2.0619e-06, 2.7777e-07, ..., 4.2655e-07, + 1.5199e-06, -8.6054e-06], + [-2.1588e-06, 1.0598e-06, -3.8780e-06, ..., -4.8010e-07, + 2.4699e-06, 5.5879e-07], + [-1.2428e-05, -1.3094e-06, -7.9069e-07, ..., 1.4575e-06, + 2.0172e-06, 1.9260e-06], + ..., + [-1.6272e-05, 2.1830e-06, -6.5029e-05, ..., 2.8275e-06, + 1.4156e-06, -1.9819e-05], + [-1.9237e-05, 4.6313e-05, -7.7933e-06, ..., 6.0499e-06, + 2.8387e-05, 2.7586e-06], + [ 1.3039e-06, 1.9534e-07, 2.3264e-06, ..., 2.4885e-06, + 7.1265e-06, 1.2867e-05]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0037, -0.0177, -0.0187, 0.0332, -0.0301, 0.0284, -0.0319, -0.0108, + 0.0218, -0.0238], device='cuda:0'), grad: tensor([-5.7556e-06, 4.5784e-06, -1.0550e-05, 2.0719e-04, -6.6310e-06, + -1.4072e-06, -1.2052e-04, -1.5140e-04, 6.1333e-05, 2.3291e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 215.01, cls_loss 0.0035 cls_loss_mapping 0.0076 cls_loss_causal 0.5774 re_mapping 0.0079 re_causal 0.0246 /// teacc 98.95 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0516, -0.0792, -0.0630, ..., -0.1151, -0.0025, 0.0641], + [ 0.0441, 0.0506, 0.0366, ..., 0.0745, -0.0976, -0.0544], + [ 0.0430, 0.0371, -0.0807, ..., -0.0003, -0.1104, -0.0337], + ..., + [-0.0361, -0.0450, 0.0672, ..., -0.0757, -0.0191, 0.0165], + [ 0.0105, 0.0802, 0.1100, ..., -0.0725, -0.0774, -0.1090], + [-0.0945, 0.0098, -0.1535, ..., -0.0666, -0.0592, 0.0640]], + device='cuda:0'), grad: tensor([[ 1.7183e-07, 5.6112e-08, 1.0966e-07, ..., 3.9227e-06, + 6.0759e-06, -2.5574e-06], + [ 4.4936e-07, 2.4773e-07, 6.2631e-08, ..., -1.5460e-07, + 4.6566e-07, 8.2050e-07], + [-2.2333e-06, -3.8766e-07, 2.5169e-07, ..., 2.9448e-06, + 4.3288e-06, 4.9733e-07], + ..., + [ 2.3399e-07, 1.6345e-07, -4.2990e-06, ..., 1.7462e-07, + 6.1234e-08, -2.6673e-06], + [ 3.1432e-08, 2.7474e-08, -6.5053e-07, ..., 1.9791e-07, + 3.0641e-07, 8.2795e-07], + [ 1.5832e-07, -5.3737e-07, 3.3192e-06, ..., 1.2498e-06, + 6.6822e-07, 1.8743e-07]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0049, -0.0172, -0.0190, 0.0329, -0.0301, 0.0280, -0.0313, -0.0117, + 0.0212, -0.0235], device='cuda:0'), grad: tensor([ 9.8795e-06, 3.0566e-06, 6.2138e-06, 2.6338e-06, 2.5973e-05, + 7.1116e-06, -5.5224e-05, -5.7667e-06, 1.9986e-06, 4.2394e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 214.82, cls_loss 0.0030 cls_loss_mapping 0.0053 cls_loss_causal 0.5355 re_mapping 0.0076 re_causal 0.0230 /// teacc 98.93 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0519, -0.0790, -0.0632, ..., -0.1151, -0.0022, 0.0644], + [ 0.0436, 0.0505, 0.0358, ..., 0.0747, -0.0983, -0.0538], + [ 0.0432, 0.0372, -0.0810, ..., -0.0003, -0.1109, -0.0340], + ..., + [-0.0352, -0.0455, 0.0680, ..., -0.0761, -0.0193, 0.0163], + [ 0.0106, 0.0801, 0.1102, ..., -0.0738, -0.0781, -0.1094], + [-0.0950, 0.0095, -0.1548, ..., -0.0672, -0.0596, 0.0640]], + device='cuda:0'), grad: tensor([[ 1.9185e-07, 9.8161e-07, 1.4738e-07, ..., 5.8394e-07, + 1.6140e-06, -2.4009e-06], + [ 1.1995e-06, 2.6077e-08, 3.3132e-07, ..., -1.8789e-07, + 1.8571e-06, 1.8235e-06], + [-3.8999e-07, 6.4960e-07, 2.1956e-07, ..., 4.4145e-07, + 1.9595e-06, 1.5423e-06], + ..., + [ 1.1753e-06, 3.9209e-07, 4.0978e-07, ..., 4.8941e-07, + 1.3569e-06, 1.3830e-06], + [ 3.1409e-07, 7.1106e-07, -4.0513e-08, ..., 4.3283e-07, + 2.9895e-06, 2.8852e-06], + [ 9.3924e-07, 3.2806e-07, 6.8871e-07, ..., 9.4995e-07, + 2.3767e-05, 2.3738e-05]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0054, -0.0177, -0.0189, 0.0325, -0.0296, 0.0280, -0.0313, -0.0111, + 0.0211, -0.0239], device='cuda:0'), grad: tensor([ 1.4175e-06, 1.0692e-05, 5.6550e-06, 6.4430e-03, 5.8189e-06, + -6.5651e-03, -8.7321e-06, 7.6108e-06, 1.0908e-05, 8.7798e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 215.19, cls_loss 0.0032 cls_loss_mapping 0.0073 cls_loss_causal 0.5806 re_mapping 0.0072 re_causal 0.0222 /// teacc 98.92 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0526, -0.0799, -0.0643, ..., -0.1157, -0.0021, 0.0646], + [ 0.0434, 0.0507, 0.0348, ..., 0.0751, -0.0997, -0.0540], + [ 0.0433, 0.0372, -0.0815, ..., -0.0006, -0.1113, -0.0346], + ..., + [-0.0345, -0.0459, 0.0689, ..., -0.0763, -0.0191, 0.0164], + [ 0.0105, 0.0800, 0.1106, ..., -0.0741, -0.0785, -0.1100], + [-0.0956, 0.0095, -0.1555, ..., -0.0676, -0.0600, 0.0642]], + device='cuda:0'), grad: tensor([[ 5.8487e-06, 4.1388e-06, 6.4485e-06, ..., 3.1181e-06, + -8.1258e-08, -1.5190e-06], + [-2.2554e-04, -2.5129e-04, -3.9649e-04, ..., -1.9586e-04, + 1.6997e-07, -1.2957e-05], + [ 2.3544e-05, 1.5318e-05, 1.6540e-05, ..., 1.8403e-05, + 1.8626e-07, 9.4920e-06], + ..., + [ 1.6347e-05, 1.7062e-05, 2.3142e-05, ..., 1.2599e-05, + 2.1956e-07, 3.9954e-07], + [ 4.9978e-05, 6.2287e-05, 1.0020e-04, ..., 4.5955e-05, + 2.4354e-07, 1.9725e-06], + [ 6.5938e-06, 2.5984e-06, 1.2308e-05, ..., 6.5379e-06, + 3.1516e-06, -7.4059e-06]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0054, -0.0183, -0.0190, 0.0318, -0.0298, 0.0287, -0.0309, -0.0105, + 0.0211, -0.0241], device='cuda:0'), grad: tensor([ 2.2575e-05, -8.8501e-04, 3.3557e-05, 1.4886e-05, 1.7560e-04, + 6.2525e-05, 2.6941e-04, 6.5386e-05, 2.2256e-04, 1.8060e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 215.35, cls_loss 0.0035 cls_loss_mapping 0.0065 cls_loss_causal 0.5305 re_mapping 0.0073 re_causal 0.0223 /// teacc 98.93 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0525, -0.0806, -0.0645, ..., -0.1156, -0.0018, 0.0649], + [ 0.0434, 0.0505, 0.0349, ..., 0.0754, -0.1003, -0.0543], + [ 0.0432, 0.0367, -0.0822, ..., -0.0008, -0.1120, -0.0354], + ..., + [-0.0347, -0.0461, 0.0690, ..., -0.0765, -0.0197, 0.0160], + [ 0.0111, 0.0831, 0.1114, ..., -0.0745, -0.0763, -0.1080], + [-0.0962, 0.0100, -0.1562, ..., -0.0680, -0.0604, 0.0645]], + device='cuda:0'), grad: tensor([[ 1.2806e-07, -2.1905e-05, -3.0011e-05, ..., 5.7044e-08, + -7.4552e-07, -5.7429e-05], + [ 4.7684e-07, 4.1686e-06, 3.8520e-06, ..., 1.8557e-07, + 5.5134e-07, 6.5006e-06], + [-1.0030e-06, 6.8918e-06, 7.3127e-06, ..., 2.1025e-07, + 5.8720e-07, 9.4697e-06], + ..., + [ 2.4051e-07, 1.6587e-06, 8.1304e-07, ..., 1.9628e-07, + 4.3330e-07, 2.0061e-06], + [-2.2724e-06, -1.5005e-05, -5.7928e-06, ..., -6.8173e-07, + 2.8685e-07, 2.8878e-05], + [ 6.5193e-07, -4.0591e-05, 3.5353e-06, ..., -2.3283e-06, + -1.1511e-05, -7.0870e-05]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0058, -0.0185, -0.0193, 0.0322, -0.0302, 0.0273, -0.0322, -0.0106, + 0.0231, -0.0239], device='cuda:0'), grad: tensor([-1.1826e-04, 1.6943e-05, 2.7686e-05, 3.8929e-06, 1.5092e-04, + 8.8522e-07, 5.2214e-05, 6.1095e-06, -6.0946e-06, -1.3471e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 215.06, cls_loss 0.0035 cls_loss_mapping 0.0077 cls_loss_causal 0.5666 re_mapping 0.0077 re_causal 0.0226 /// teacc 98.82 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0537, -0.0813, -0.0647, ..., -0.1156, -0.0017, 0.0651], + [ 0.0438, 0.0509, 0.0355, ..., 0.0760, -0.1008, -0.0544], + [ 0.0432, 0.0365, -0.0830, ..., -0.0013, -0.1131, -0.0360], + ..., + [-0.0349, -0.0465, 0.0691, ..., -0.0765, -0.0199, 0.0161], + [ 0.0109, 0.0836, 0.1115, ..., -0.0750, -0.0762, -0.1082], + [-0.0972, 0.0100, -0.1570, ..., -0.0685, -0.0611, 0.0645]], + device='cuda:0'), grad: tensor([[ 8.0109e-05, 1.4126e-05, 2.5705e-07, ..., 7.1265e-06, + 1.7405e-05, 3.2991e-05], + [ 4.7730e-07, 7.3900e-07, 1.4110e-07, ..., -2.7963e-07, + 1.0720e-06, 9.5088e-07], + [ 3.7570e-06, 1.2247e-06, 1.6550e-06, ..., 4.3982e-07, + 1.3206e-06, 1.9260e-06], + ..., + [ 1.6810e-07, 1.7397e-06, -3.2745e-06, ..., 1.8626e-07, + 1.4603e-06, 1.6605e-06], + [ 8.5756e-06, 2.9746e-06, -4.2375e-07, ..., 1.1269e-06, + 5.9046e-06, 6.3553e-06], + [ 3.2987e-06, -5.3197e-05, 8.0327e-07, ..., 3.9861e-07, + -3.2634e-05, -6.2168e-05]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0061, -0.0181, -0.0194, 0.0318, -0.0311, 0.0282, -0.0324, -0.0106, + 0.0231, -0.0241], device='cuda:0'), grad: tensor([ 3.0375e-04, 4.7237e-06, 1.5706e-05, -3.5906e-04, 1.5426e-04, + -5.3823e-05, 3.4958e-05, 2.9858e-06, 4.3273e-05, -1.4663e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 214.96, cls_loss 0.0029 cls_loss_mapping 0.0071 cls_loss_causal 0.5658 re_mapping 0.0078 re_causal 0.0236 /// teacc 98.94 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0550, -0.0828, -0.0654, ..., -0.1167, -0.0022, 0.0652], + [ 0.0440, 0.0507, 0.0357, ..., 0.0767, -0.1019, -0.0546], + [ 0.0437, 0.0370, -0.0833, ..., -0.0018, -0.1137, -0.0356], + ..., + [-0.0354, -0.0466, 0.0691, ..., -0.0770, -0.0200, 0.0155], + [ 0.0113, 0.0838, 0.1118, ..., -0.0756, -0.0766, -0.1086], + [-0.0998, 0.0099, -0.1579, ..., -0.0691, -0.0614, 0.0649]], + device='cuda:0'), grad: tensor([[ 8.9547e-07, 5.3924e-07, 3.5833e-07, ..., 3.7788e-07, + 3.1921e-07, 1.0943e-07], + [-5.5045e-05, -5.0604e-05, -7.6890e-05, ..., -6.0558e-05, + 2.7427e-07, -6.2704e-05], + [ 9.8944e-06, 5.2899e-07, 5.0152e-07, ..., 3.6322e-07, + 1.2456e-07, 2.6403e-07], + ..., + [ 3.4899e-05, 2.2128e-05, 3.2455e-05, ..., 2.5764e-05, + 1.3411e-07, 2.8968e-05], + [ 1.1019e-05, -9.8124e-06, -3.7011e-06, ..., 1.7388e-06, + -1.1034e-05, 1.6121e-06], + [ 3.1382e-05, 2.5421e-05, 3.8236e-05, ..., 3.0443e-05, + 5.6112e-07, 2.7120e-05]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0058, -0.0181, -0.0189, 0.0319, -0.0315, 0.0283, -0.0321, -0.0110, + 0.0231, -0.0241], device='cuda:0'), grad: tensor([ 2.6319e-06, -1.5688e-04, 2.3559e-05, -8.5473e-05, 8.0690e-06, + 2.1815e-05, 9.5218e-06, 9.7871e-05, -4.8093e-06, 8.3625e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 215.16, cls_loss 0.0035 cls_loss_mapping 0.0068 cls_loss_causal 0.5677 re_mapping 0.0071 re_causal 0.0229 /// teacc 98.92 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0555, -0.0840, -0.0658, ..., -0.1175, -0.0025, 0.0654], + [ 0.0441, 0.0507, 0.0359, ..., 0.0773, -0.1022, -0.0553], + [ 0.0437, 0.0376, -0.0844, ..., -0.0018, -0.1140, -0.0373], + ..., + [-0.0348, -0.0473, 0.0695, ..., -0.0775, -0.0204, 0.0164], + [ 0.0111, 0.0854, 0.1117, ..., -0.0780, -0.0752, -0.1091], + [-0.1005, 0.0097, -0.1595, ..., -0.0698, -0.0622, 0.0650]], + device='cuda:0'), grad: tensor([[ 2.1979e-05, 1.2573e-06, 1.7583e-05, ..., 1.9395e-07, + 6.4112e-06, 1.3120e-05], + [ 7.6108e-06, 1.4901e-06, 9.7007e-06, ..., -5.1269e-07, + 4.0345e-06, 7.5996e-06], + [-2.4700e-04, 1.3538e-05, -7.7009e-05, ..., 2.5090e-06, + 2.3078e-06, -4.8578e-05], + ..., + [ 8.8096e-05, 1.7919e-06, -2.2262e-05, ..., 4.6869e-07, + -2.2680e-05, -2.0057e-05], + [ 2.3544e-05, -3.3170e-05, -2.6673e-06, ..., 4.2981e-07, + 3.1330e-06, 1.0498e-05], + [ 8.5890e-05, -5.2117e-06, 3.5644e-05, ..., 1.2387e-06, + -4.3958e-06, 4.6170e-07]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0058, -0.0182, -0.0192, 0.0320, -0.0320, 0.0283, -0.0333, -0.0103, + 0.0241, -0.0244], device='cuda:0'), grad: tensor([ 8.4996e-05, 4.3631e-05, -5.1975e-04, 5.6803e-05, 1.5236e-05, + 6.0499e-05, 1.1213e-05, 3.0994e-05, 3.3259e-05, 1.8275e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 215.28, cls_loss 0.0028 cls_loss_mapping 0.0068 cls_loss_causal 0.5474 re_mapping 0.0073 re_causal 0.0224 /// teacc 99.01 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0562, -0.0853, -0.0665, ..., -0.1183, -0.0029, 0.0657], + [ 0.0443, 0.0510, 0.0366, ..., 0.0779, -0.1004, -0.0530], + [ 0.0438, 0.0375, -0.0853, ..., -0.0021, -0.1148, -0.0372], + ..., + [-0.0349, -0.0476, 0.0695, ..., -0.0779, -0.0221, 0.0156], + [ 0.0110, 0.0853, 0.1118, ..., -0.0784, -0.0754, -0.1097], + [-0.1018, 0.0093, -0.1623, ..., -0.0706, -0.0630, 0.0649]], + device='cuda:0'), grad: tensor([[ 1.3295e-07, 1.5954e-06, 6.2399e-08, ..., 1.2387e-07, + 1.6354e-06, -6.0443e-07], + [-1.3853e-07, 1.3765e-06, -2.6543e-08, ..., 4.8056e-07, + 1.4575e-06, 4.9360e-07], + [-1.2685e-06, 1.8161e-08, 1.3877e-07, ..., 3.6275e-07, + 7.2969e-07, 1.1064e-06], + ..., + [ 9.7416e-07, 1.7155e-06, -3.7206e-07, ..., 5.7137e-07, + 5.0943e-07, 9.0897e-06], + [ 6.2445e-07, 1.0297e-05, 2.2352e-08, ..., 3.4319e-07, + 1.0453e-05, 4.0680e-06], + [ 4.1444e-08, -2.5984e-07, -5.8068e-07, ..., 1.6633e-06, + 2.7362e-06, -2.6196e-05]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0057, -0.0171, -0.0193, 0.0324, -0.0319, 0.0284, -0.0331, -0.0107, + 0.0239, -0.0253], device='cuda:0'), grad: tensor([ 2.8182e-06, 3.6992e-06, 1.9241e-06, 2.0236e-05, 4.5709e-06, + 7.2382e-06, -3.7283e-05, 2.2575e-05, 2.6584e-05, -5.2422e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 123---------------------------------------------------- +epoch 123, time 231.23, cls_loss 0.0024 cls_loss_mapping 0.0056 cls_loss_causal 0.5703 re_mapping 0.0077 re_causal 0.0238 /// teacc 99.05 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0566, -0.0858, -0.0667, ..., -0.1183, -0.0028, 0.0658], + [ 0.0446, 0.0509, 0.0367, ..., 0.0782, -0.1010, -0.0532], + [ 0.0438, 0.0374, -0.0857, ..., -0.0025, -0.1156, -0.0375], + ..., + [-0.0350, -0.0480, 0.0698, ..., -0.0779, -0.0221, 0.0156], + [ 0.0103, 0.0850, 0.1118, ..., -0.0789, -0.0756, -0.1104], + [-0.1022, 0.0098, -0.1629, ..., -0.0711, -0.0630, 0.0652]], + device='cuda:0'), grad: tensor([[ 8.4750e-08, 1.3381e-05, -2.5973e-05, ..., 1.0664e-07, + 9.2462e-06, -6.4433e-05], + [ 6.6496e-07, 1.3495e-06, 1.3001e-06, ..., 3.8650e-08, + 6.3423e-07, 2.7101e-06], + [-1.1148e-06, 2.0061e-06, 1.4585e-06, ..., 1.7160e-07, + 1.6568e-06, 3.6228e-06], + ..., + [-5.3830e-07, 3.1409e-07, 1.3851e-05, ..., 2.1118e-07, + 1.0757e-07, 3.7193e-05], + [-5.1456e-07, 3.1799e-05, 1.3299e-06, ..., 6.3377e-07, + 2.0728e-05, 7.7039e-06], + [ 4.5565e-07, 3.0585e-06, 2.9895e-06, ..., 7.9442e-07, + 1.9222e-06, 7.8306e-06]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0058, -0.0171, -0.0195, 0.0329, -0.0321, 0.0294, -0.0337, -0.0105, + 0.0233, -0.0251], device='cuda:0'), grad: tensor([-7.9632e-05, 9.2685e-06, 9.0301e-06, 1.2264e-05, 3.5651e-06, + 1.0878e-05, -1.3173e-04, 5.7518e-05, 8.7798e-05, 2.0802e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 215.13, cls_loss 0.0027 cls_loss_mapping 0.0061 cls_loss_causal 0.5583 re_mapping 0.0077 re_causal 0.0231 /// teacc 98.86 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0574, -0.0873, -0.0670, ..., -0.1193, -0.0029, 0.0660], + [ 0.0443, 0.0489, 0.0364, ..., 0.0761, -0.1030, -0.0533], + [ 0.0441, 0.0383, -0.0857, ..., -0.0011, -0.1162, -0.0379], + ..., + [-0.0351, -0.0485, 0.0699, ..., -0.0781, -0.0224, 0.0154], + [ 0.0104, 0.0848, 0.1120, ..., -0.0795, -0.0759, -0.1111], + [-0.1021, 0.0111, -0.1633, ..., -0.0705, -0.0624, 0.0657]], + device='cuda:0'), grad: tensor([[ 4.8988e-07, 5.2946e-07, 3.2037e-07, ..., 3.5902e-07, + -1.7717e-05, -2.4438e-05], + [ 1.2126e-06, 1.1707e-06, 4.9807e-06, ..., 5.1875e-07, + 6.5472e-07, 4.1425e-06], + [-2.6282e-06, 4.2352e-07, 3.7223e-05, ..., 1.1921e-06, + 6.8313e-07, 2.7061e-05], + ..., + [-8.8066e-06, 9.9000e-07, -8.0824e-05, ..., 7.6927e-07, + 5.2061e-07, -5.7131e-05], + [ 1.0179e-06, -1.1278e-06, -1.3858e-06, ..., 1.4873e-06, + 9.4343e-07, 2.5257e-06], + [ 3.6452e-06, 3.4142e-04, 2.7031e-05, ..., 2.9707e-04, + 1.9884e-04, 1.2255e-04]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0058, -0.0182, -0.0188, 0.0329, -0.0330, 0.0296, -0.0334, -0.0105, + 0.0231, -0.0243], device='cuda:0'), grad: tensor([-2.6748e-05, 1.8537e-05, 8.4758e-05, 5.8413e-06, -7.0906e-04, + 1.5177e-05, 2.1875e-05, -2.1183e-04, 1.0423e-05, 7.8964e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 215.00, cls_loss 0.0024 cls_loss_mapping 0.0052 cls_loss_causal 0.5332 re_mapping 0.0077 re_causal 0.0225 /// teacc 99.05 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0577, -0.0880, -0.0671, ..., -0.1203, -0.0030, 0.0662], + [ 0.0442, 0.0489, 0.0361, ..., 0.0763, -0.1034, -0.0535], + [ 0.0441, 0.0381, -0.0867, ..., -0.0015, -0.1166, -0.0385], + ..., + [-0.0348, -0.0488, 0.0705, ..., -0.0785, -0.0223, 0.0153], + [ 0.0106, 0.0849, 0.1121, ..., -0.0802, -0.0761, -0.1115], + [-0.1027, 0.0112, -0.1639, ..., -0.0707, -0.0626, 0.0662]], + device='cuda:0'), grad: tensor([[ 2.8429e-07, 6.9477e-07, 4.1607e-07, ..., 1.9418e-07, + -1.4249e-07, -2.6859e-06], + [-1.9241e-06, 4.0233e-06, 2.8848e-07, ..., 4.5300e-06, + 2.1961e-06, 1.7299e-07], + [ 1.0114e-06, 2.4866e-06, 1.7406e-06, ..., 1.3169e-06, + 6.1467e-07, 5.9931e-07], + ..., + [ 1.2005e-06, 1.8803e-06, 2.2016e-06, ..., 4.1015e-06, + 1.3039e-06, 2.1653e-07], + [-5.0887e-06, -1.1817e-05, -6.1691e-06, ..., 1.8217e-06, + 1.9856e-06, 1.1576e-06], + [ 4.1611e-06, 1.9297e-05, 6.4336e-06, ..., 2.1145e-05, + 1.4730e-05, 2.3507e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0059, -0.0186, -0.0189, 0.0331, -0.0334, 0.0296, -0.0334, -0.0100, + 0.0230, -0.0242], device='cuda:0'), grad: tensor([-2.9523e-06, 6.9998e-06, 6.7465e-06, 2.5213e-05, -6.4433e-05, + -3.3915e-05, 3.3081e-06, 7.8753e-06, -1.4655e-05, 6.5744e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 126---------------------------------------------------- +epoch 126, time 231.36, cls_loss 0.0031 cls_loss_mapping 0.0062 cls_loss_causal 0.5484 re_mapping 0.0075 re_causal 0.0213 /// teacc 99.08 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0583, -0.0907, -0.0675, ..., -0.1207, -0.0044, 0.0659], + [ 0.0450, 0.0507, 0.0377, ..., 0.0771, -0.1036, -0.0534], + [ 0.0443, 0.0375, -0.0874, ..., -0.0018, -0.1173, -0.0390], + ..., + [-0.0349, -0.0497, 0.0706, ..., -0.0790, -0.0225, 0.0151], + [ 0.0100, 0.0845, 0.1113, ..., -0.0814, -0.0763, -0.1118], + [-0.1048, 0.0114, -0.1652, ..., -0.0710, -0.0629, 0.0663]], + device='cuda:0'), grad: tensor([[ 9.0078e-06, 4.7218e-07, 2.5751e-07, ..., 2.0792e-07, + -2.3879e-06, -2.2233e-04], + [ 1.4462e-05, -1.9488e-07, -7.2876e-07, ..., 2.5798e-06, + 7.2643e-07, 2.7157e-06], + [ 8.8811e-05, -1.1558e-06, 2.5108e-06, ..., -4.3437e-06, + 7.9628e-07, 1.8954e-05], + ..., + [ 4.1798e-06, 5.5926e-07, -9.2462e-06, ..., 5.5647e-07, + 1.2740e-06, 1.9580e-05], + [ 9.3877e-07, 2.4531e-06, -8.9407e-08, ..., 5.1130e-07, + 2.9635e-06, 2.4304e-05], + [ 3.2783e-06, -3.4813e-06, 3.3304e-06, ..., 2.2049e-07, + 2.3767e-06, 1.1581e-04]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0047, -0.0176, -0.0188, 0.0329, -0.0339, 0.0298, -0.0326, -0.0101, + 0.0224, -0.0244], device='cuda:0'), grad: tensor([-3.4952e-04, 3.4422e-05, 2.6035e-04, -3.0875e-04, 1.5616e-05, + 3.3081e-05, 3.9160e-05, 1.6138e-05, 4.4703e-05, 2.1505e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 214.84, cls_loss 0.0022 cls_loss_mapping 0.0052 cls_loss_causal 0.5277 re_mapping 0.0074 re_causal 0.0220 /// teacc 98.89 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0589, -0.0918, -0.0679, ..., -0.1215, -0.0049, 0.0661], + [ 0.0450, 0.0509, 0.0379, ..., 0.0776, -0.1039, -0.0535], + [ 0.0447, 0.0375, -0.0880, ..., -0.0019, -0.1178, -0.0395], + ..., + [-0.0349, -0.0502, 0.0710, ..., -0.0789, -0.0217, 0.0150], + [ 0.0099, 0.0844, 0.1113, ..., -0.0824, -0.0766, -0.1124], + [-0.1054, 0.0107, -0.1662, ..., -0.0722, -0.0639, 0.0663]], + device='cuda:0'), grad: tensor([[ 1.1805e-07, 2.8864e-05, 3.9162e-07, ..., 1.7905e-07, + 1.5661e-05, 7.4767e-06], + [ 1.1045e-06, 2.5835e-06, 4.2617e-06, ..., -3.0268e-08, + 1.1623e-06, 4.3139e-06], + [ 1.9725e-06, 1.1623e-06, 6.8508e-06, ..., 1.4529e-07, + 7.5717e-07, 2.1700e-06], + ..., + [-5.7891e-06, 1.0975e-05, -2.1785e-05, ..., 9.5228e-08, + 7.7486e-07, 2.8312e-05], + [ 1.7718e-07, 3.2157e-05, 4.7567e-07, ..., 3.6065e-07, + 1.9461e-05, 1.9342e-05], + [ 1.6741e-07, -2.8059e-05, 6.1048e-07, ..., 9.1270e-08, + -2.7614e-07, -7.6830e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0047, -0.0175, -0.0186, 0.0328, -0.0336, 0.0298, -0.0322, -0.0099, + 0.0221, -0.0250], device='cuda:0'), grad: tensor([ 5.5015e-05, 1.6063e-05, 1.4737e-05, 4.3333e-05, 2.7403e-05, + 2.8357e-05, -1.4293e-04, 2.3782e-05, 7.7248e-05, -1.4317e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 215.04, cls_loss 0.0027 cls_loss_mapping 0.0063 cls_loss_causal 0.5542 re_mapping 0.0071 re_causal 0.0215 /// teacc 98.99 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0594, -0.0922, -0.0683, ..., -0.1217, -0.0050, 0.0664], + [ 0.0452, 0.0510, 0.0382, ..., 0.0778, -0.1038, -0.0531], + [ 0.0449, 0.0374, -0.0881, ..., -0.0021, -0.1188, -0.0403], + ..., + [-0.0351, -0.0508, 0.0711, ..., -0.0793, -0.0221, 0.0153], + [ 0.0101, 0.0842, 0.1114, ..., -0.0831, -0.0770, -0.1133], + [-0.1061, 0.0119, -0.1671, ..., -0.0739, -0.0645, 0.0666]], + device='cuda:0'), grad: tensor([[ 5.0440e-06, 8.6427e-06, 3.7365e-06, ..., 5.0198e-07, + 3.7178e-06, 5.1875e-07], + [ 4.8093e-06, 7.7337e-06, 1.2793e-05, ..., 2.2389e-06, + 8.0140e-07, 2.3213e-07], + [ 4.8429e-06, 4.7982e-06, 2.9225e-06, ..., 5.1130e-07, + 6.3656e-07, 8.2189e-08], + ..., + [ 9.3365e-07, 1.4389e-06, 6.8871e-07, ..., 4.1560e-07, + 1.7206e-07, 1.3532e-06], + [-2.8104e-05, -2.6658e-05, -3.0637e-05, ..., -3.8780e-06, + 2.7828e-06, 1.2517e-06], + [ 1.5749e-06, 1.1344e-06, 1.2694e-06, ..., 1.2238e-06, + 1.4836e-06, -2.8852e-06]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0049, -0.0173, -0.0185, 0.0326, -0.0333, 0.0296, -0.0320, -0.0099, + 0.0219, -0.0251], device='cuda:0'), grad: tensor([ 2.6822e-05, 1.8030e-05, 9.9763e-06, 1.8939e-05, -5.3411e-07, + 6.4187e-06, -1.3046e-05, 4.2245e-06, -7.4804e-05, 3.9823e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 214.72, cls_loss 0.0030 cls_loss_mapping 0.0062 cls_loss_causal 0.5275 re_mapping 0.0075 re_causal 0.0217 /// teacc 98.92 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0610, -0.0929, -0.0698, ..., -0.1224, -0.0052, 0.0673], + [ 0.0437, 0.0511, 0.0365, ..., 0.0788, -0.1034, -0.0532], + [ 0.0448, 0.0370, -0.0887, ..., -0.0025, -0.1201, -0.0415], + ..., + [-0.0329, -0.0513, 0.0734, ..., -0.0800, -0.0219, 0.0172], + [ 0.0107, 0.0844, 0.1119, ..., -0.0833, -0.0776, -0.1143], + [-0.1062, 0.0115, -0.1702, ..., -0.0747, -0.0650, 0.0656]], + device='cuda:0'), grad: tensor([[ 1.6415e-07, 7.2364e-07, 7.7765e-07, ..., 2.0955e-08, + 6.3563e-08, -8.8429e-07], + [ 5.9931e-07, -6.8732e-07, 2.1271e-06, ..., -1.5162e-06, + 1.8859e-08, 5.4482e-07], + [ 4.9593e-07, -2.0652e-07, 3.9525e-06, ..., 2.4727e-07, + 1.1595e-07, 6.6729e-07], + ..., + [-4.5486e-06, 9.1130e-07, -1.0759e-05, ..., 3.0198e-07, + 2.2119e-08, -9.8720e-07], + [ 8.4192e-07, -8.0466e-07, -1.4007e-06, ..., 7.0548e-07, + 2.7590e-07, 4.2049e-07], + [ 3.7835e-07, -4.9593e-07, 1.3774e-06, ..., 8.5682e-08, + 7.2177e-08, -1.1297e-06]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0055, -0.0185, -0.0189, 0.0327, -0.0332, 0.0288, -0.0317, -0.0075, + 0.0219, -0.0266], device='cuda:0'), grad: tensor([ 4.3050e-07, 3.9041e-06, 3.8743e-06, 5.1782e-06, 1.4929e-06, + 1.2480e-06, -1.1092e-06, -1.4678e-05, -1.9767e-07, -2.0792e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 215.09, cls_loss 0.0027 cls_loss_mapping 0.0060 cls_loss_causal 0.5368 re_mapping 0.0071 re_causal 0.0206 /// teacc 99.00 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0617, -0.0947, -0.0701, ..., -0.1248, -0.0071, 0.0667], + [ 0.0439, 0.0513, 0.0368, ..., 0.0795, -0.1037, -0.0531], + [ 0.0451, 0.0370, -0.0887, ..., -0.0028, -0.1217, -0.0430], + ..., + [-0.0333, -0.0525, 0.0732, ..., -0.0812, -0.0224, 0.0168], + [ 0.0109, 0.0844, 0.1121, ..., -0.0839, -0.0777, -0.1148], + [-0.1066, 0.0116, -0.1707, ..., -0.0752, -0.0656, 0.0663]], + device='cuda:0'), grad: tensor([[ 4.7311e-07, 5.3179e-07, 3.0943e-07, ..., 2.8894e-07, + 4.1653e-07, -2.2724e-06], + [ 3.3472e-06, 1.0300e-06, 1.8636e-06, ..., 2.2678e-07, + 5.0664e-07, 4.1118e-07], + [-3.8683e-05, 1.1250e-06, 1.9055e-06, ..., 6.2585e-07, + 5.4110e-07, 3.5134e-07], + ..., + [ 4.1217e-05, 7.2177e-07, -1.1154e-05, ..., 4.0978e-07, + 1.2340e-07, -1.5087e-06], + [ 1.3970e-06, 2.9281e-06, 1.3998e-06, ..., 1.2089e-06, + 2.0806e-06, 1.6438e-06], + [ 5.0105e-07, -1.0617e-06, 1.7518e-06, ..., 3.9511e-07, + 3.9442e-07, -1.3364e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0036, -0.0183, -0.0191, 0.0326, -0.0333, 0.0290, -0.0305, -0.0078, + 0.0218, -0.0263], device='cuda:0'), grad: tensor([-1.3122e-06, 1.0043e-05, -1.4746e-04, -6.5379e-06, 2.8517e-06, + 6.0424e-06, -9.3430e-06, 1.3304e-04, 1.1101e-05, 1.4938e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 215.20, cls_loss 0.0025 cls_loss_mapping 0.0050 cls_loss_causal 0.5509 re_mapping 0.0066 re_causal 0.0206 /// teacc 98.93 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0624, -0.0951, -0.0703, ..., -0.1250, -0.0071, 0.0673], + [ 0.0444, 0.0512, 0.0370, ..., 0.0803, -0.1040, -0.0538], + [ 0.0447, 0.0370, -0.0896, ..., -0.0034, -0.1224, -0.0436], + ..., + [-0.0336, -0.0541, 0.0731, ..., -0.0821, -0.0231, 0.0160], + [ 0.0111, 0.0844, 0.1123, ..., -0.0844, -0.0781, -0.1162], + [-0.1065, 0.0106, -0.1710, ..., -0.0767, -0.0668, 0.0666]], + device='cuda:0'), grad: tensor([[ 2.0699e-07, 9.0199e-07, 7.5204e-08, ..., 3.4906e-06, + 6.8769e-06, -1.7835e-06], + [-6.4597e-06, -3.3434e-06, -8.8662e-06, ..., 5.2564e-06, + 1.6049e-05, 1.5576e-07], + [-8.5775e-07, 1.3178e-06, 9.8255e-07, ..., 6.6757e-06, + 1.0610e-05, 2.4261e-07], + ..., + [ 5.0925e-06, 3.6769e-06, 6.3665e-06, ..., 1.4031e-04, + 2.0444e-04, 7.8045e-07], + [ 1.4808e-07, 5.5879e-06, 3.9651e-07, ..., 8.6650e-06, + 1.7390e-05, 9.7323e-07], + [ 4.5495e-07, 1.1316e-06, 4.6496e-07, ..., 5.9634e-05, + 9.0003e-05, 1.1204e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0040, -0.0182, -0.0194, 0.0327, -0.0320, 0.0289, -0.0303, -0.0082, + 0.0215, -0.0268], device='cuda:0'), grad: tensor([ 1.4871e-05, 2.0906e-05, 2.3067e-05, 1.7196e-05, -8.3256e-04, + 5.0116e-04, -4.1294e-04, 4.3631e-04, 4.0799e-05, 1.8930e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 215.03, cls_loss 0.0027 cls_loss_mapping 0.0057 cls_loss_causal 0.5555 re_mapping 0.0075 re_causal 0.0210 /// teacc 99.06 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0630, -0.0953, -0.0709, ..., -0.1252, -0.0070, 0.0674], + [ 0.0445, 0.0512, 0.0371, ..., 0.0806, -0.1049, -0.0541], + [ 0.0452, 0.0367, -0.0901, ..., -0.0037, -0.1231, -0.0440], + ..., + [-0.0339, -0.0553, 0.0732, ..., -0.0832, -0.0243, 0.0159], + [ 0.0115, 0.0846, 0.1126, ..., -0.0846, -0.0783, -0.1170], + [-0.1069, 0.0101, -0.1716, ..., -0.0783, -0.0688, 0.0671]], + device='cuda:0'), grad: tensor([[ 5.6718e-07, 2.1942e-06, 4.4564e-07, ..., 3.9255e-07, + 1.6941e-06, -1.6484e-06], + [-1.7419e-05, -7.6070e-06, -6.8992e-06, ..., 1.0036e-05, + 8.9481e-06, 4.2655e-07], + [ 6.1840e-06, 7.1935e-06, 5.9195e-06, ..., 3.9451e-06, + 1.9670e-06, 2.4843e-07], + ..., + [ 7.1572e-07, 2.6934e-06, 4.9593e-07, ..., 4.4629e-06, + 2.0862e-06, -1.5441e-06], + [ 1.2495e-05, 1.6272e-05, 7.2569e-06, ..., 4.5784e-06, + 1.6958e-05, 1.5777e-06], + [ 4.4913e-07, 1.5022e-06, 1.9446e-06, ..., 1.8319e-06, + 1.8748e-06, 1.2871e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0041, -0.0183, -0.0192, 0.0323, -0.0306, 0.0290, -0.0303, -0.0084, + 0.0215, -0.0274], device='cuda:0'), grad: tensor([ 4.9993e-06, -7.4878e-06, 1.9550e-05, 9.7901e-06, -3.6031e-05, + -9.1493e-05, -1.6779e-05, 5.4315e-06, 1.0329e-04, 8.6799e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 214.86, cls_loss 0.0032 cls_loss_mapping 0.0073 cls_loss_causal 0.5532 re_mapping 0.0070 re_causal 0.0207 /// teacc 98.91 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0638, -0.0956, -0.0719, ..., -0.1253, -0.0068, 0.0678], + [ 0.0446, 0.0512, 0.0375, ..., 0.0811, -0.1056, -0.0544], + [ 0.0454, 0.0366, -0.0912, ..., -0.0044, -0.1240, -0.0447], + ..., + [-0.0344, -0.0577, 0.0731, ..., -0.0833, -0.0243, 0.0145], + [ 0.0119, 0.0850, 0.1131, ..., -0.0850, -0.0784, -0.1174], + [-0.1058, 0.0107, -0.1720, ..., -0.0785, -0.0691, 0.0684]], + device='cuda:0'), grad: tensor([[ 3.8487e-07, 1.5032e-06, 6.1933e-07, ..., 2.7311e-07, + 1.8431e-06, 4.9509e-06], + [ 1.9372e-06, 2.4494e-06, 4.0457e-06, ..., 3.9604e-07, + 3.0547e-07, 2.4177e-06], + [-3.1069e-06, 3.3919e-06, -2.2678e-07, ..., 6.6869e-07, + 3.6228e-07, 2.3767e-06], + ..., + [ 2.6986e-05, 3.0771e-06, 2.0653e-05, ..., 6.8685e-07, + 5.2527e-07, -1.5181e-07], + [-4.4465e-05, -3.3551e-07, -5.7220e-05, ..., -2.2515e-07, + 3.3993e-07, 3.2894e-06], + [ 5.0776e-06, -3.6621e-03, 1.2115e-05, ..., -5.9462e-04, + 6.8307e-05, -1.4019e-03]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0044, -0.0187, -0.0190, 0.0315, -0.0312, 0.0296, -0.0304, -0.0086, + 0.0217, -0.0266], device='cuda:0'), grad: tensor([ 1.0766e-05, 1.3277e-05, 3.5539e-06, 3.0145e-05, 7.5951e-03, + -2.8825e-04, 5.0813e-06, 4.6223e-05, -8.4221e-05, -7.3280e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 214.99, cls_loss 0.0024 cls_loss_mapping 0.0061 cls_loss_causal 0.5915 re_mapping 0.0072 re_causal 0.0219 /// teacc 98.99 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0645, -0.0966, -0.0726, ..., -0.1256, -0.0072, 0.0678], + [ 0.0447, 0.0513, 0.0376, ..., 0.0817, -0.1058, -0.0546], + [ 0.0455, 0.0363, -0.0920, ..., -0.0047, -0.1251, -0.0458], + ..., + [-0.0346, -0.0585, 0.0734, ..., -0.0842, -0.0241, 0.0151], + [ 0.0126, 0.0852, 0.1137, ..., -0.0853, -0.0786, -0.1179], + [-0.1065, 0.0118, -0.1738, ..., -0.0788, -0.0704, 0.0685]], + device='cuda:0'), grad: tensor([[ 3.0594e-07, 7.9069e-07, 9.0972e-06, ..., 5.6028e-06, + 9.7416e-07, 2.6878e-06], + [-3.6880e-06, -1.4469e-05, -3.6955e-04, ..., -2.2626e-04, + -3.2097e-05, -1.2624e-04], + [ 1.6214e-06, 2.2016e-06, 9.5293e-06, ..., 5.7369e-06, + 9.5554e-07, 2.7306e-06], + ..., + [ 2.0172e-06, 1.1563e-05, 2.8896e-04, ..., 1.7643e-04, + 2.5958e-05, 1.0061e-04], + [-3.8408e-06, -3.7830e-06, 3.3546e-06, ..., 5.1409e-06, + 1.6820e-06, 4.0159e-06], + [ 8.5728e-07, -4.5262e-07, 1.5557e-05, ..., 1.0356e-05, + 2.3209e-06, 1.9148e-06]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0042, -0.0187, -0.0193, 0.0314, -0.0319, 0.0297, -0.0303, -0.0082, + 0.0218, -0.0264], device='cuda:0'), grad: tensor([ 2.0444e-05, -8.2493e-04, 2.1651e-05, 1.1977e-06, 7.5161e-05, + 7.6741e-06, 6.6869e-06, 6.5041e-04, 8.1882e-06, 3.3140e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 214.86, cls_loss 0.0023 cls_loss_mapping 0.0044 cls_loss_causal 0.5580 re_mapping 0.0073 re_causal 0.0219 /// teacc 98.93 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0649, -0.0977, -0.0738, ..., -0.1258, -0.0072, 0.0678], + [ 0.0448, 0.0516, 0.0381, ..., 0.0830, -0.1062, -0.0548], + [ 0.0456, 0.0362, -0.0922, ..., -0.0048, -0.1257, -0.0464], + ..., + [-0.0347, -0.0604, 0.0731, ..., -0.0865, -0.0243, 0.0136], + [ 0.0127, 0.0850, 0.1138, ..., -0.0862, -0.0791, -0.1187], + [-0.1070, 0.0114, -0.1735, ..., -0.0797, -0.0713, 0.0703]], + device='cuda:0'), grad: tensor([[ 1.0175e-07, 2.0023e-08, 1.7695e-07, ..., 2.4447e-08, + -1.0291e-07, -2.0012e-05], + [-2.8126e-07, 1.6298e-08, 5.3346e-06, ..., 3.0398e-06, + 7.4040e-07, 2.8759e-06], + [ 5.9837e-07, 7.7998e-08, 5.8450e-06, ..., 1.1548e-07, + 7.1246e-08, 1.3366e-05], + ..., + [-1.9725e-06, -3.2154e-07, -4.3452e-05, ..., -5.6773e-06, + -1.2722e-06, -4.4852e-05], + [ 5.8720e-07, 9.6019e-07, 9.7603e-07, ..., 5.7416e-07, + 1.8589e-06, 3.5968e-06], + [ 1.6661e-06, 1.3923e-07, 2.2277e-05, ..., 1.5888e-06, + 7.4180e-07, 3.3081e-05]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0040, -0.0185, -0.0193, 0.0314, -0.0312, 0.0297, -0.0300, -0.0096, + 0.0216, -0.0257], device='cuda:0'), grad: tensor([-3.7491e-05, 1.3568e-05, 3.8207e-05, 9.9652e-07, 3.0786e-05, + -6.0815e-07, 3.4459e-06, -1.6880e-04, 1.1295e-05, 1.0860e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 215.13, cls_loss 0.0030 cls_loss_mapping 0.0065 cls_loss_causal 0.5213 re_mapping 0.0071 re_causal 0.0206 /// teacc 98.93 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0659, -0.0993, -0.0745, ..., -0.1269, -0.0075, 0.0681], + [ 0.0448, 0.0514, 0.0380, ..., 0.0832, -0.1068, -0.0550], + [ 0.0451, 0.0357, -0.0930, ..., -0.0050, -0.1264, -0.0469], + ..., + [-0.0347, -0.0624, 0.0722, ..., -0.0867, -0.0255, 0.0123], + [ 0.0133, 0.0853, 0.1142, ..., -0.0868, -0.0793, -0.1196], + [-0.1072, 0.0121, -0.1712, ..., -0.0799, -0.0709, 0.0716]], + device='cuda:0'), grad: tensor([[ 6.4727e-07, 1.9856e-06, 1.3807e-07, ..., 8.9640e-08, + 6.8685e-07, -7.6413e-05], + [-1.1595e-07, 3.9511e-07, -1.0822e-06, ..., -3.9022e-07, + 2.1793e-06, 1.3541e-06], + [-9.3132e-06, -3.9972e-06, 5.3132e-07, ..., 2.8475e-07, + 4.0256e-07, 9.5293e-06], + ..., + [ 6.0024e-07, 5.1502e-07, -6.1793e-07, ..., 2.3935e-07, + 4.6054e-07, 4.7404e-07], + [ 5.1558e-06, -1.2673e-05, -1.3756e-06, ..., 6.0443e-07, + 5.1856e-06, 6.8918e-06], + [ 9.7137e-07, 1.5408e-05, 1.6084e-06, ..., 4.9220e-07, + 1.8673e-06, 3.4004e-05]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0039, -0.0187, -0.0201, 0.0336, -0.0317, 0.0292, -0.0296, -0.0118, + 0.0216, -0.0238], device='cuda:0'), grad: tensor([-1.1647e-04, 6.0238e-06, -2.9933e-06, 7.5474e-06, 6.1560e-07, + -4.0829e-05, 6.0678e-05, 1.7108e-06, -9.6112e-06, 9.3281e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 215.28, cls_loss 0.0028 cls_loss_mapping 0.0063 cls_loss_causal 0.5700 re_mapping 0.0073 re_causal 0.0213 /// teacc 98.92 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0664, -0.1003, -0.0748, ..., -0.1276, -0.0077, 0.0683], + [ 0.0456, 0.0518, 0.0386, ..., 0.0834, -0.1073, -0.0558], + [ 0.0453, 0.0356, -0.0938, ..., -0.0054, -0.1278, -0.0474], + ..., + [-0.0352, -0.0634, 0.0720, ..., -0.0866, -0.0255, 0.0122], + [ 0.0125, 0.0845, 0.1141, ..., -0.0884, -0.0800, -0.1208], + [-0.1073, 0.0119, -0.1716, ..., -0.0817, -0.0726, 0.0721]], + device='cuda:0'), grad: tensor([[ 5.7137e-07, 8.4704e-07, 3.5367e-07, ..., 7.9954e-07, + 6.6236e-06, 6.8061e-06], + [-2.5164e-06, -1.8580e-06, -2.0396e-06, ..., 3.2317e-07, + 4.4294e-06, 1.2983e-06], + [ 5.2713e-07, 1.2470e-06, 9.2527e-07, ..., 1.3448e-06, + 2.0694e-06, 5.4808e-07], + ..., + [ 1.3383e-06, 2.3581e-06, -4.3988e-05, ..., -1.4627e-04, + -2.5678e-04, -2.1420e-06], + [-3.2429e-06, -1.2992e-06, -1.5153e-06, ..., 1.8459e-06, + 3.8445e-06, 2.4140e-06], + [ 1.3467e-06, -6.6683e-06, 6.6794e-06, ..., 1.0550e-05, + 1.4313e-05, -8.5533e-06]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0038, -0.0184, -0.0204, 0.0333, -0.0308, 0.0290, -0.0287, -0.0119, + 0.0205, -0.0239], device='cuda:0'), grad: tensor([ 1.4029e-05, 4.7944e-06, 6.6720e-06, 3.3617e-05, 4.8876e-04, + -3.8445e-05, -4.5635e-06, -5.4026e-04, 4.1053e-06, 3.1322e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 215.15, cls_loss 0.0031 cls_loss_mapping 0.0056 cls_loss_causal 0.5699 re_mapping 0.0075 re_causal 0.0211 /// teacc 98.98 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0681, -0.1010, -0.0757, ..., -0.1275, -0.0066, 0.0692], + [ 0.0452, 0.0515, 0.0378, ..., 0.0837, -0.1078, -0.0568], + [ 0.0461, 0.0374, -0.0945, ..., -0.0054, -0.1286, -0.0471], + ..., + [-0.0354, -0.0639, 0.0728, ..., -0.0862, -0.0245, 0.0130], + [ 0.0123, 0.0845, 0.1151, ..., -0.0885, -0.0803, -0.1214], + [-0.1092, 0.0116, -0.1733, ..., -0.0828, -0.0735, 0.0717]], + device='cuda:0'), grad: tensor([[ 4.1793e-07, 7.8278e-07, 9.0571e-08, ..., 3.1781e-07, + 5.2713e-07, -5.6392e-07], + [-1.4585e-06, -3.2205e-06, -4.5374e-06, ..., -2.6710e-06, + 5.5693e-07, 1.8557e-07], + [-1.3255e-05, -7.7719e-07, 5.8440e-07, ..., 5.2620e-07, + 3.3015e-07, 1.4366e-07], + ..., + [ 2.1141e-06, 1.1539e-06, -2.5635e-07, ..., 3.4529e-07, + 1.3108e-07, 4.4424e-07], + [ 5.9530e-06, 4.7237e-06, 2.8107e-06, ..., 2.0918e-06, + 1.5544e-06, 2.6487e-06], + [-1.7928e-07, -2.2538e-06, 4.7311e-07, ..., 5.6764e-07, + 3.8673e-07, -8.7693e-06]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0043, -0.0192, -0.0199, 0.0341, -0.0307, 0.0290, -0.0291, -0.0111, + 0.0205, -0.0249], device='cuda:0'), grad: tensor([ 1.2182e-06, -9.7882e-07, -2.3380e-05, 7.3500e-06, 4.3884e-06, + 1.3337e-05, -8.6427e-06, 4.8280e-06, 1.7449e-05, -1.5587e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 214.80, cls_loss 0.0023 cls_loss_mapping 0.0045 cls_loss_causal 0.5434 re_mapping 0.0076 re_causal 0.0215 /// teacc 99.01 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0692, -0.1013, -0.0759, ..., -0.1274, -0.0064, 0.0694], + [ 0.0453, 0.0514, 0.0378, ..., 0.0841, -0.1084, -0.0569], + [ 0.0470, 0.0386, -0.0935, ..., -0.0051, -0.1306, -0.0479], + ..., + [-0.0357, -0.0648, 0.0730, ..., -0.0867, -0.0248, 0.0137], + [ 0.0120, 0.0846, 0.1151, ..., -0.0897, -0.0804, -0.1218], + [-0.1098, 0.0115, -0.1742, ..., -0.0834, -0.0738, 0.0714]], + device='cuda:0'), grad: tensor([[ 1.6764e-06, 7.2829e-07, 2.0023e-06, ..., 8.7451e-07, + -1.3062e-07, 7.6741e-07], + [-8.4400e-05, -3.2514e-05, -1.0312e-04, ..., -4.6760e-05, + 1.0850e-07, -8.4698e-05], + [ 8.6799e-06, 3.8892e-06, 9.7230e-06, ..., 4.0382e-06, + 9.1502e-08, 6.6385e-06], + ..., + [ 6.1274e-05, 2.3693e-05, 7.3254e-05, ..., 3.3617e-05, + 7.7533e-08, 6.2108e-05], + [-1.4631e-06, -9.4855e-07, -8.6613e-08, ..., 1.0626e-06, + 6.0350e-07, 2.0452e-06], + [ 1.0125e-05, 3.5539e-06, 1.2979e-05, ..., 5.9195e-06, + 7.1432e-07, 9.1195e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0044, -0.0193, -0.0190, 0.0339, -0.0306, 0.0288, -0.0290, -0.0107, + 0.0204, -0.0256], device='cuda:0'), grad: tensor([ 3.4198e-06, -2.1315e-04, 2.0862e-05, 4.8727e-06, 2.6375e-06, + -4.0117e-07, 2.1104e-06, 1.5438e-04, 1.0920e-07, 2.5198e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 215.29, cls_loss 0.0026 cls_loss_mapping 0.0057 cls_loss_causal 0.5399 re_mapping 0.0067 re_causal 0.0206 /// teacc 98.92 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0699, -0.1018, -0.0766, ..., -0.1277, -0.0065, 0.0696], + [ 0.0462, 0.0515, 0.0394, ..., 0.0871, -0.1087, -0.0542], + [ 0.0470, 0.0385, -0.0939, ..., -0.0054, -0.1325, -0.0484], + ..., + [-0.0367, -0.0656, 0.0720, ..., -0.0892, -0.0218, 0.0123], + [ 0.0118, 0.0848, 0.1155, ..., -0.0900, -0.0806, -0.1216], + [-0.1103, 0.0121, -0.1748, ..., -0.0834, -0.0743, 0.0718]], + device='cuda:0'), grad: tensor([[ 3.1665e-07, 5.8562e-06, 1.8273e-06, ..., 1.6391e-07, + 9.6336e-06, 2.8647e-06], + [-9.7044e-07, 6.0583e-07, -8.6194e-07, ..., -1.3877e-06, + 9.9838e-07, 5.2014e-07], + [-1.7677e-06, 1.5702e-06, 9.0897e-07, ..., 7.4366e-07, + 1.5125e-06, 5.5181e-07], + ..., + [ 1.0636e-06, 5.0804e-07, -1.9395e-07, ..., 4.2724e-07, + 2.1420e-07, 7.0920e-07], + [-8.0690e-06, -1.1820e-04, -6.6698e-05, ..., 3.6811e-07, + -4.3929e-05, -3.0603e-06], + [ 3.3202e-07, 2.3842e-06, 8.4564e-07, ..., 3.1348e-06, + 2.0452e-06, -4.2617e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0044, -0.0171, -0.0193, 0.0348, -0.0315, 0.0279, -0.0290, -0.0121, + 0.0203, -0.0254], device='cuda:0'), grad: tensor([ 2.3380e-05, 2.1812e-06, -5.9344e-06, 4.2319e-06, -2.8405e-06, + 8.8871e-05, 1.5473e-04, 5.5283e-06, -2.7490e-04, 4.4294e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 215.11, cls_loss 0.0024 cls_loss_mapping 0.0050 cls_loss_causal 0.5474 re_mapping 0.0068 re_causal 0.0205 /// teacc 99.07 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0706, -0.1019, -0.0770, ..., -0.1283, -0.0066, 0.0703], + [ 0.0466, 0.0519, 0.0398, ..., 0.0880, -0.1090, -0.0539], + [ 0.0468, 0.0380, -0.0946, ..., -0.0061, -0.1347, -0.0520], + ..., + [-0.0368, -0.0665, 0.0721, ..., -0.0896, -0.0223, 0.0121], + [ 0.0120, 0.0847, 0.1158, ..., -0.0908, -0.0811, -0.1229], + [-0.1102, 0.0115, -0.1752, ..., -0.0862, -0.0760, 0.0720]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 1.9535e-05, 1.4831e-07, ..., 2.0433e-06, + 1.7166e-05, 3.1531e-05], + [-2.0908e-07, 6.1747e-07, 5.0990e-08, ..., 1.5507e-07, + 8.8476e-07, 2.6706e-07], + [ 7.1479e-08, 2.1569e-06, 9.1502e-08, ..., 7.2876e-07, + 9.3970e-07, 4.8764e-06], + ..., + [ 7.3342e-08, 2.9709e-07, 8.1258e-08, ..., 6.3796e-08, + 2.9081e-07, 6.7335e-07], + [-4.9174e-07, 6.5006e-06, 9.5461e-08, ..., 1.6643e-06, + 5.4017e-06, 1.2554e-06], + [ 3.2899e-07, -9.2834e-06, 5.1409e-07, ..., 1.4179e-07, + 5.6485e-07, -3.2991e-05]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0049, -0.0167, -0.0200, 0.0345, -0.0307, 0.0277, -0.0285, -0.0123, + 0.0199, -0.0258], device='cuda:0'), grad: tensor([ 6.7294e-05, 1.9064e-06, 7.3314e-06, 8.1584e-06, 6.4299e-06, + -8.5682e-06, -6.4611e-05, 1.2703e-06, 1.4335e-05, -3.3557e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 215.12, cls_loss 0.0017 cls_loss_mapping 0.0045 cls_loss_causal 0.5469 re_mapping 0.0065 re_causal 0.0208 /// teacc 99.01 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0717, -0.1024, -0.0775, ..., -0.1282, -0.0063, 0.0708], + [ 0.0469, 0.0521, 0.0399, ..., 0.0882, -0.1097, -0.0540], + [ 0.0464, 0.0374, -0.0954, ..., -0.0068, -0.1360, -0.0527], + ..., + [-0.0368, -0.0668, 0.0723, ..., -0.0897, -0.0223, 0.0122], + [ 0.0123, 0.0849, 0.1161, ..., -0.0911, -0.0812, -0.1234], + [-0.1104, 0.0116, -0.1758, ..., -0.0864, -0.0764, 0.0720]], + device='cuda:0'), grad: tensor([[ 1.3830e-07, 1.9511e-07, 1.1525e-07, ..., 6.1234e-08, + 1.9348e-07, 7.7300e-08], + [ 2.5146e-07, -4.3493e-07, -7.4971e-08, ..., -8.8988e-07, + 8.1258e-08, 9.7509e-07], + [-2.4326e-06, -5.4715e-08, 4.3679e-07, ..., 8.1956e-08, + 6.4261e-08, 2.1094e-07], + ..., + [ 1.6186e-06, 2.8801e-07, -2.7735e-06, ..., 3.6252e-07, + 4.0745e-08, -2.3767e-06], + [ 1.2144e-06, 2.3395e-06, 4.7241e-07, ..., 6.0862e-07, + 3.0752e-06, 1.4501e-06], + [ 3.1269e-07, 3.6624e-07, 6.6962e-07, ..., 6.6636e-07, + 4.6729e-07, 1.5413e-07]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0052, -0.0166, -0.0207, 0.0343, -0.0308, 0.0277, -0.0286, -0.0121, + 0.0200, -0.0259], device='cuda:0'), grad: tensor([ 8.1770e-07, 2.6152e-06, -2.7120e-06, -1.5665e-06, 1.9185e-07, + 6.7987e-07, -7.2159e-06, -4.4443e-06, 9.1270e-06, 2.4773e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 214.90, cls_loss 0.0023 cls_loss_mapping 0.0060 cls_loss_causal 0.5673 re_mapping 0.0063 re_causal 0.0211 /// teacc 98.95 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0725, -0.1029, -0.0778, ..., -0.1285, -0.0064, 0.0709], + [ 0.0472, 0.0522, 0.0402, ..., 0.0883, -0.1103, -0.0538], + [ 0.0467, 0.0373, -0.0955, ..., -0.0069, -0.1368, -0.0524], + ..., + [-0.0373, -0.0676, 0.0720, ..., -0.0897, -0.0224, 0.0121], + [ 0.0128, 0.0850, 0.1169, ..., -0.0916, -0.0815, -0.1239], + [-0.1121, 0.0086, -0.1767, ..., -0.0866, -0.0770, 0.0712]], + device='cuda:0'), grad: tensor([[ 1.0408e-07, 3.8953e-07, 1.5949e-07, ..., 2.7288e-07, + 2.6473e-07, -1.3011e-06], + [ 1.5437e-07, 4.4227e-05, 2.0087e-05, ..., 5.4061e-05, + 5.9038e-05, 5.8161e-07], + [ 4.4634e-07, 4.7521e-07, 8.4704e-07, ..., 1.1851e-07, + 1.4179e-07, 3.3132e-07], + ..., + [ 1.9977e-07, 2.4587e-07, -1.2666e-06, ..., 2.4913e-08, + 7.1712e-08, 1.0421e-06], + [-1.4305e-06, 1.2189e-05, 4.3362e-06, ..., 1.5691e-05, + 1.7568e-05, 1.0189e-06], + [ 1.5181e-07, -2.2613e-06, 2.3399e-07, ..., 9.4762e-08, + -8.6846e-08, -9.9391e-06]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0051, -0.0165, -0.0203, 0.0337, -0.0283, 0.0284, -0.0282, -0.0124, + 0.0202, -0.0281], device='cuda:0'), grad: tensor([-1.5199e-06, 1.1158e-04, 1.6280e-06, -1.3970e-08, 9.8646e-06, + 1.4305e-05, -1.5724e-04, 1.2247e-07, 3.1799e-05, -1.0423e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 215.27, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5317 re_mapping 0.0063 re_causal 0.0202 /// teacc 99.03 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0731, -0.1033, -0.0783, ..., -0.1286, -0.0065, 0.0702], + [ 0.0472, 0.0519, 0.0401, ..., 0.0883, -0.1114, -0.0539], + [ 0.0471, 0.0373, -0.0957, ..., -0.0070, -0.1373, -0.0529], + ..., + [-0.0375, -0.0682, 0.0719, ..., -0.0898, -0.0225, 0.0118], + [ 0.0131, 0.0860, 0.1182, ..., -0.0916, -0.0807, -0.1246], + [-0.1131, 0.0086, -0.1769, ..., -0.0868, -0.0772, 0.0722]], + device='cuda:0'), grad: tensor([[ 6.8452e-08, 1.8533e-07, 9.3132e-08, ..., 6.9151e-08, + 2.9337e-08, -1.4715e-06], + [ 3.0734e-08, -1.1642e-08, 4.7730e-07, ..., -1.0664e-07, + 4.4703e-08, 2.5216e-07], + [-1.3048e-06, 3.4156e-07, 2.3772e-07, ..., 2.1746e-07, + 4.0699e-07, 2.8173e-08], + ..., + [ 1.8557e-07, 7.4971e-08, -2.1514e-06, ..., 3.9814e-08, + 2.0023e-08, -6.7288e-07], + [ 4.7311e-07, 6.6869e-07, 1.5600e-08, ..., 2.1700e-07, + 6.2492e-07, 8.4983e-07], + [ 2.8685e-07, -6.8359e-07, 7.5018e-07, ..., 1.8394e-08, + 2.3283e-09, 4.6566e-08]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0044, -0.0167, -0.0201, 0.0337, -0.0282, 0.0288, -0.0293, -0.0127, + 0.0212, -0.0276], device='cuda:0'), grad: tensor([-1.3253e-06, 1.4510e-06, 1.0012e-08, 7.6182e-07, 1.1949e-06, + 5.5414e-07, -3.0175e-06, -4.1872e-06, 3.0380e-06, 1.5199e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 214.86, cls_loss 0.0028 cls_loss_mapping 0.0058 cls_loss_causal 0.5218 re_mapping 0.0065 re_causal 0.0191 /// teacc 98.98 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0754, -0.1039, -0.0794, ..., -0.1285, -0.0064, 0.0717], + [ 0.0475, 0.0516, 0.0402, ..., 0.0877, -0.1122, -0.0536], + [ 0.0466, 0.0372, -0.0966, ..., -0.0071, -0.1387, -0.0566], + ..., + [-0.0381, -0.0693, 0.0718, ..., -0.0899, -0.0226, 0.0117], + [ 0.0135, 0.0864, 0.1188, ..., -0.0922, -0.0808, -0.1254], + [-0.1136, 0.0087, -0.1775, ..., -0.0871, -0.0771, 0.0721]], + device='cuda:0'), grad: tensor([[ 1.3970e-07, 4.4331e-07, 2.4680e-07, ..., 1.0012e-08, + 8.3353e-08, 2.1537e-07], + [ 3.3574e-07, 5.6950e-07, 6.7614e-06, ..., 6.3330e-08, + 6.7055e-08, 7.8231e-06], + [-4.8382e-07, 3.3365e-07, 3.3085e-07, ..., 1.9441e-07, + 1.2713e-07, 3.5716e-07], + ..., + [ 2.4703e-07, 1.4133e-07, -1.9714e-05, ..., 6.3563e-08, + 3.1898e-08, -2.0370e-05], + [ 4.8801e-07, 2.2035e-06, 5.3318e-07, ..., 8.4285e-08, + 1.5344e-07, 3.0287e-06], + [-2.0973e-06, -1.3389e-05, 1.0267e-05, ..., 4.6752e-06, + 1.8673e-06, -7.3537e-06]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0055, -0.0169, -0.0213, 0.0337, -0.0284, 0.0290, -0.0291, -0.0129, + 0.0215, -0.0278], device='cuda:0'), grad: tensor([ 1.3039e-06, 1.8671e-05, 6.4820e-07, 1.6823e-05, -1.7062e-06, + 8.8140e-06, -1.1940e-06, -4.7624e-05, 7.5214e-06, -3.2522e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 215.10, cls_loss 0.0017 cls_loss_mapping 0.0039 cls_loss_causal 0.5269 re_mapping 0.0066 re_causal 0.0204 /// teacc 99.00 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0769, -0.1043, -0.0808, ..., -0.1285, -0.0063, 0.0725], + [ 0.0466, 0.0494, 0.0390, ..., 0.0858, -0.1133, -0.0536], + [ 0.0471, 0.0376, -0.0970, ..., -0.0070, -0.1391, -0.0573], + ..., + [-0.0382, -0.0696, 0.0722, ..., -0.0898, -0.0226, 0.0117], + [ 0.0138, 0.0864, 0.1191, ..., -0.0928, -0.0811, -0.1262], + [-0.1150, 0.0087, -0.1780, ..., -0.0875, -0.0775, 0.0718]], + device='cuda:0'), grad: tensor([[ 3.0007e-06, 1.6401e-06, 3.2037e-07, ..., 4.4797e-07, + 1.9670e-06, 1.5413e-06], + [ 4.6417e-06, 9.1549e-07, -3.6228e-07, ..., -1.6950e-07, + 2.7753e-07, 8.1025e-08], + [-7.3910e-05, -1.4365e-05, 3.9209e-07, ..., 2.0862e-07, + 1.2480e-07, -1.8086e-06], + ..., + [ 7.1898e-06, 8.2981e-07, -6.3237e-07, ..., 1.1176e-07, + 3.5390e-08, 1.2945e-07], + [ 2.1011e-05, 3.7495e-06, -3.1441e-06, ..., 4.3586e-07, + 1.2508e-06, 1.4678e-06], + [ 1.3700e-06, 1.6578e-07, 5.4203e-07, ..., 3.7998e-07, + 3.3062e-07, -5.6345e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0060, -0.0184, -0.0210, 0.0335, -0.0286, 0.0293, -0.0275, -0.0127, + 0.0214, -0.0282], device='cuda:0'), grad: tensor([ 1.2934e-05, 1.1079e-05, -1.5223e-04, 6.3479e-05, 2.6803e-06, + 5.1446e-06, -6.9030e-06, 1.4916e-05, 4.5806e-05, 3.0976e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 214.77, cls_loss 0.0029 cls_loss_mapping 0.0051 cls_loss_causal 0.5006 re_mapping 0.0067 re_causal 0.0195 /// teacc 99.01 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0796, -0.1048, -0.0841, ..., -0.1288, -0.0066, 0.0725], + [ 0.0465, 0.0497, 0.0387, ..., 0.0863, -0.1137, -0.0536], + [ 0.0471, 0.0374, -0.0977, ..., -0.0075, -0.1399, -0.0575], + ..., + [-0.0378, -0.0705, 0.0729, ..., -0.0901, -0.0230, 0.0115], + [ 0.0161, 0.0864, 0.1213, ..., -0.0939, -0.0813, -0.1273], + [-0.1153, 0.0085, -0.1788, ..., -0.0885, -0.0782, 0.0721]], + device='cuda:0'), grad: tensor([[ 2.1793e-07, 1.1083e-07, 1.3504e-07, ..., 9.4995e-08, + -2.4959e-07, -1.6419e-06], + [-5.1819e-06, -3.6806e-06, -6.7391e-06, ..., -4.2729e-06, + 3.1665e-08, 1.3225e-07], + [-5.4464e-06, 1.0431e-07, 1.3784e-07, ..., 1.1828e-07, + 1.3970e-08, 7.8231e-08], + ..., + [ 4.2766e-06, 4.7032e-07, -3.1590e-06, ..., 5.6997e-07, + 1.2107e-08, -1.1120e-06], + [ 2.7809e-06, 2.8834e-06, 3.3174e-06, ..., 2.0117e-06, + 1.9446e-06, 1.3132e-06], + [ 1.1548e-06, 3.5204e-07, 2.1234e-06, ..., 4.3493e-07, + 9.6858e-08, 7.9256e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0056, -0.0184, -0.0210, 0.0328, -0.0283, 0.0284, -0.0276, -0.0126, + 0.0228, -0.0283], device='cuda:0'), grad: tensor([-2.9057e-06, -8.6948e-06, -6.7577e-06, 3.3807e-07, 4.4983e-07, + 1.1828e-06, -2.5332e-07, -1.0626e-06, 1.2398e-05, 5.2415e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 215.10, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5751 re_mapping 0.0068 re_causal 0.0204 /// teacc 99.03 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0808, -0.1053, -0.0851, ..., -0.1292, -0.0068, 0.0725], + [ 0.0469, 0.0498, 0.0389, ..., 0.0866, -0.1145, -0.0537], + [ 0.0474, 0.0372, -0.0981, ..., -0.0081, -0.1407, -0.0580], + ..., + [-0.0383, -0.0710, 0.0733, ..., -0.0903, -0.0231, 0.0117], + [ 0.0159, 0.0864, 0.1214, ..., -0.0949, -0.0816, -0.1283], + [-0.1156, 0.0087, -0.1796, ..., -0.0889, -0.0785, 0.0724]], + device='cuda:0'), grad: tensor([[ 1.7788e-07, 3.4738e-07, -1.2107e-08, ..., 2.7660e-07, + -3.9339e-06, -5.8822e-06], + [ 2.0023e-07, 3.5297e-07, 9.3505e-07, ..., 2.4959e-07, + 7.7207e-07, 1.1073e-06], + [-8.1863e-07, 5.3365e-07, 2.0303e-07, ..., 4.6846e-07, + 9.2294e-07, 5.1036e-07], + ..., + [ 6.0070e-07, 2.5518e-07, -4.1239e-06, ..., 2.4028e-07, + 5.7556e-07, -9.6485e-07], + [ 2.6170e-07, 1.6866e-06, -1.1828e-07, ..., 1.4761e-06, + 2.3507e-06, 5.1595e-07], + [ 6.9477e-07, 1.4022e-05, 1.0962e-06, ..., 1.1906e-05, + 1.1265e-05, -2.9746e-06]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0054, -0.0183, -0.0212, 0.0323, -0.0283, 0.0289, -0.0275, -0.0123, + 0.0225, -0.0283], device='cuda:0'), grad: tensor([-1.3627e-05, 4.5896e-06, 1.9558e-06, -1.5117e-05, -5.3287e-05, + 1.7166e-05, 1.9804e-05, -3.2671e-06, 7.4208e-06, 3.4273e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 214.70, cls_loss 0.0028 cls_loss_mapping 0.0055 cls_loss_causal 0.5183 re_mapping 0.0072 re_causal 0.0197 /// teacc 98.90 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0821, -0.1056, -0.0861, ..., -0.1293, -0.0064, 0.0738], + [ 0.0474, 0.0498, 0.0398, ..., 0.0871, -0.1153, -0.0519], + [ 0.0480, 0.0368, -0.0988, ..., -0.0087, -0.1436, -0.0592], + ..., + [-0.0388, -0.0717, 0.0732, ..., -0.0908, -0.0232, 0.0110], + [ 0.0160, 0.0864, 0.1214, ..., -0.0958, -0.0821, -0.1291], + [-0.1166, 0.0088, -0.1812, ..., -0.0897, -0.0795, 0.0718]], + device='cuda:0'), grad: tensor([[ 3.6508e-07, 7.6555e-07, 5.0571e-07, ..., 3.4180e-07, + -9.3132e-09, -9.9838e-07], + [ 2.8778e-07, 2.9877e-06, 1.8161e-06, ..., 5.1484e-06, + 4.4033e-06, 1.5058e-05], + [ 2.9579e-06, 7.2736e-07, 2.2762e-06, ..., 5.5134e-07, + 8.3540e-07, 1.9949e-06], + ..., + [ 6.8266e-07, 2.2557e-06, -3.0976e-06, ..., 9.6951e-07, + 2.5984e-06, 3.2592e-04], + [-4.1723e-05, -4.9442e-05, -4.7475e-05, ..., 4.1351e-07, + -5.0366e-05, 1.8505e-06], + [ 1.7202e-06, -7.3761e-06, 3.0734e-06, ..., 4.3660e-06, + -1.8515e-06, -4.1533e-04]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0064, -0.0172, -0.0208, 0.0321, -0.0280, 0.0290, -0.0275, -0.0131, + 0.0222, -0.0293], device='cuda:0'), grad: tensor([ 1.6065e-06, 3.1024e-05, 9.0301e-06, 3.6150e-05, 6.2883e-05, + 2.5213e-05, 1.4234e-04, 3.9196e-04, -1.9133e-04, -5.1022e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 214.93, cls_loss 0.0025 cls_loss_mapping 0.0058 cls_loss_causal 0.5231 re_mapping 0.0069 re_causal 0.0196 /// teacc 98.99 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0839, -0.1063, -0.0888, ..., -0.1295, -0.0065, 0.0740], + [ 0.0475, 0.0505, 0.0399, ..., 0.0878, -0.1147, -0.0522], + [ 0.0478, 0.0361, -0.1000, ..., -0.0093, -0.1457, -0.0598], + ..., + [-0.0385, -0.0721, 0.0740, ..., -0.0907, -0.0226, 0.0110], + [ 0.0169, 0.0866, 0.1219, ..., -0.0970, -0.0826, -0.1298], + [-0.1177, 0.0088, -0.1820, ..., -0.0899, -0.0801, 0.0722]], + device='cuda:0'), grad: tensor([[ 1.6298e-07, 1.2843e-06, 3.9395e-07, ..., 5.5600e-07, + -4.3306e-07, -4.6529e-06], + [-7.8604e-06, -9.8441e-07, -1.1176e-05, ..., -5.9009e-06, + 2.4028e-07, -1.6624e-06], + [ 1.3970e-07, 2.3283e-07, 1.9744e-06, ..., 9.3225e-07, + 2.3190e-07, 7.0594e-07], + ..., + [ 3.6731e-06, 3.9302e-07, -8.7768e-06, ..., 2.3413e-06, + 1.1455e-07, -1.1861e-05], + [ 8.9873e-07, 1.6121e-06, 1.4631e-06, ..., 9.2108e-07, + 2.5406e-06, 1.4063e-06], + [ 4.5728e-07, 1.4994e-07, 1.1273e-05, ..., 4.2655e-07, + 1.9893e-06, 1.4149e-05]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0061, -0.0171, -0.0214, 0.0322, -0.0281, 0.0293, -0.0283, -0.0127, + 0.0225, -0.0293], device='cuda:0'), grad: tensor([-3.2634e-06, -1.3307e-05, 1.1623e-06, -1.2219e-06, -2.3283e-07, + 9.0599e-06, -1.1653e-05, -1.4856e-05, 8.2776e-06, 2.5973e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 215.14, cls_loss 0.0022 cls_loss_mapping 0.0046 cls_loss_causal 0.5462 re_mapping 0.0063 re_causal 0.0194 /// teacc 99.00 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0843, -0.1069, -0.0874, ..., -0.1286, -0.0060, 0.0746], + [ 0.0476, 0.0506, 0.0401, ..., 0.0881, -0.1154, -0.0524], + [ 0.0481, 0.0361, -0.1008, ..., -0.0095, -0.1472, -0.0603], + ..., + [-0.0385, -0.0726, 0.0744, ..., -0.0907, -0.0217, 0.0110], + [ 0.0167, 0.0865, 0.1218, ..., -0.0983, -0.0829, -0.1312], + [-0.1181, 0.0090, -0.1825, ..., -0.0902, -0.0809, 0.0725]], + device='cuda:0'), grad: tensor([[ 2.7753e-07, 1.3784e-07, 1.2526e-06, ..., 1.4715e-07, + 9.1828e-07, -9.8720e-08], + [-7.2084e-07, -6.4261e-07, 5.9418e-06, ..., 2.7940e-08, + 6.6794e-06, 4.4629e-06], + [ 1.5385e-06, 4.7870e-07, 1.6484e-06, ..., 5.1875e-07, + 3.6601e-07, 2.7288e-07], + ..., + [ 1.5106e-06, 1.3895e-06, -1.2502e-05, ..., 2.5705e-07, + -1.1481e-05, -8.2478e-06], + [ 5.9009e-06, 1.4156e-07, 4.5188e-06, ..., 3.3621e-07, + 4.7404e-07, 4.4517e-07], + [ 9.7137e-07, 1.8440e-07, 1.2554e-06, ..., 7.6927e-07, + 1.0962e-06, -8.5402e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0068, -0.0172, -0.0213, 0.0320, -0.0285, 0.0297, -0.0284, -0.0125, + 0.0221, -0.0293], device='cuda:0'), grad: tensor([ 3.8184e-06, 2.5824e-05, 5.0515e-06, -4.6730e-05, 5.1819e-06, + 2.2382e-05, 3.5670e-07, -4.1008e-05, 2.0176e-05, 4.9546e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 215.30, cls_loss 0.0018 cls_loss_mapping 0.0041 cls_loss_causal 0.5383 re_mapping 0.0063 re_causal 0.0197 /// teacc 99.03 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0851, -0.1072, -0.0864, ..., -0.1285, -0.0055, 0.0753], + [ 0.0478, 0.0506, 0.0402, ..., 0.0881, -0.1159, -0.0525], + [ 0.0481, 0.0357, -0.1013, ..., -0.0096, -0.1480, -0.0614], + ..., + [-0.0389, -0.0732, 0.0742, ..., -0.0908, -0.0214, 0.0109], + [ 0.0168, 0.0866, 0.1219, ..., -0.0987, -0.0833, -0.1323], + [-0.1186, 0.0093, -0.1828, ..., -0.0904, -0.0806, 0.0729]], + device='cuda:0'), grad: tensor([[ 6.6310e-07, 2.2911e-07, 2.4214e-08, ..., 7.0781e-08, + 3.8184e-08, -2.5053e-07], + [ 7.4133e-07, 5.6252e-07, -3.9674e-07, ..., -1.4529e-07, + 1.9372e-07, 8.9407e-08], + [-1.0751e-05, -5.0142e-06, 1.0524e-07, ..., 4.9453e-07, + 4.7591e-07, 3.6322e-08], + ..., + [ 6.0815e-07, 1.9465e-07, -7.7300e-08, ..., 7.7300e-08, + 4.8429e-08, -7.5437e-08], + [ 7.9721e-06, 5.4948e-06, 1.6484e-07, ..., 1.2852e-06, + 4.8876e-06, 1.0561e-06], + [ 5.2899e-07, 1.4808e-07, 9.2201e-08, ..., 1.4529e-07, + 2.4494e-07, -2.0396e-07]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0073, -0.0172, -0.0216, 0.0323, -0.0288, 0.0295, -0.0283, -0.0127, + 0.0219, -0.0291], device='cuda:0'), grad: tensor([ 1.3048e-06, 2.1625e-06, -2.1324e-05, -1.8943e-06, 1.9856e-06, + -5.3532e-06, -4.4182e-06, 1.1846e-06, 2.5064e-05, 1.3188e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 215.27, cls_loss 0.0028 cls_loss_mapping 0.0051 cls_loss_causal 0.5502 re_mapping 0.0061 re_causal 0.0183 /// teacc 98.95 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0852, -0.1075, -0.0847, ..., -0.1280, -0.0053, 0.0758], + [ 0.0492, 0.0514, 0.0406, ..., 0.0887, -0.1166, -0.0521], + [ 0.0466, 0.0335, -0.1032, ..., -0.0112, -0.1499, -0.0639], + ..., + [-0.0391, -0.0749, 0.0744, ..., -0.0915, -0.0214, 0.0111], + [ 0.0167, 0.0866, 0.1219, ..., -0.1002, -0.0840, -0.1336], + [-0.1188, 0.0089, -0.1839, ..., -0.0908, -0.0819, 0.0728]], + device='cuda:0'), grad: tensor([[ 1.5460e-07, 2.7381e-06, 1.7118e-06, ..., 2.8126e-07, + -9.6858e-08, -4.4703e-06], + [ 5.3272e-07, 1.2154e-06, 1.1176e-07, ..., 1.1288e-06, + 9.3505e-07, 1.2303e-06], + [-4.8205e-06, 6.0722e-07, 6.3330e-08, ..., 5.2527e-07, + 4.6659e-07, 3.7439e-07], + ..., + [ 1.4631e-06, 2.4773e-06, 6.3330e-08, ..., 1.1148e-06, + 9.0618e-07, 8.7917e-06], + [-2.2352e-06, -1.3761e-05, -2.0236e-05, ..., -6.9514e-06, + -5.2117e-06, 8.4098e-07], + [ 1.1828e-07, -2.6841e-06, 2.8554e-06, ..., 1.8030e-06, + 2.4084e-06, -5.9128e-05]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0079, -0.0164, -0.0240, 0.0324, -0.0282, 0.0301, -0.0285, -0.0127, + 0.0214, -0.0297], device='cuda:0'), grad: tensor([ 5.0217e-06, 7.8529e-06, -1.5676e-05, 2.8964e-06, 8.4281e-05, + 1.7494e-05, 4.9323e-06, 2.8133e-05, -2.9162e-05, -1.0592e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 215.08, cls_loss 0.0025 cls_loss_mapping 0.0051 cls_loss_causal 0.5260 re_mapping 0.0063 re_causal 0.0197 /// teacc 98.91 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0860, -0.1086, -0.0852, ..., -0.1283, -0.0060, 0.0751], + [ 0.0502, 0.0522, 0.0410, ..., 0.0891, -0.1166, -0.0518], + [ 0.0457, 0.0317, -0.1055, ..., -0.0120, -0.1506, -0.0654], + ..., + [-0.0393, -0.0751, 0.0748, ..., -0.0916, -0.0214, 0.0112], + [ 0.0169, 0.0865, 0.1222, ..., -0.1006, -0.0846, -0.1346], + [-0.1198, 0.0086, -0.1845, ..., -0.0918, -0.0830, 0.0732]], + device='cuda:0'), grad: tensor([[ 2.8126e-07, 1.5460e-07, 1.6578e-07, ..., 4.0978e-08, + 1.3411e-07, 1.0524e-07], + [ 3.1292e-06, -9.5926e-08, -1.9092e-07, ..., -3.1758e-07, + 1.2293e-07, 3.4273e-07], + [-5.9083e-06, -3.1106e-07, 2.1514e-07, ..., 4.5635e-08, + 4.0978e-08, 6.1467e-08], + ..., + [ 1.5870e-06, 3.5297e-07, 2.0210e-07, ..., 1.9651e-07, + 2.5705e-07, 6.2678e-07], + [-3.8557e-06, -3.7737e-06, -3.8780e-06, ..., 1.5646e-07, + 2.9802e-07, 5.7183e-07], + [ 1.7649e-06, 8.5682e-08, 1.0738e-06, ..., 1.1269e-07, + 3.8370e-07, -2.1253e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0071, -0.0159, -0.0256, 0.0324, -0.0281, 0.0301, -0.0282, -0.0123, + 0.0211, -0.0300], device='cuda:0'), grad: tensor([ 7.7672e-07, 8.7991e-06, -1.4462e-05, 2.3171e-06, 4.3958e-06, + -2.1327e-07, 1.5758e-06, 4.1127e-06, -7.3463e-06, -3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 215.24, cls_loss 0.0020 cls_loss_mapping 0.0037 cls_loss_causal 0.5165 re_mapping 0.0060 re_causal 0.0182 /// teacc 98.99 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0862, -0.1091, -0.0845, ..., -0.1272, -0.0065, 0.0747], + [ 0.0499, 0.0521, 0.0408, ..., 0.0891, -0.1170, -0.0519], + [ 0.0463, 0.0317, -0.1054, ..., -0.0122, -0.1519, -0.0655], + ..., + [-0.0395, -0.0757, 0.0753, ..., -0.0917, -0.0214, 0.0115], + [ 0.0171, 0.0865, 0.1223, ..., -0.1015, -0.0850, -0.1351], + [-0.1204, 0.0104, -0.1853, ..., -0.0904, -0.0813, 0.0739]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 3.6322e-07, -1.8114e-06, ..., 1.7788e-07, + -3.8669e-06, -7.6368e-06], + [ 1.0133e-06, -3.0734e-08, 4.4294e-06, ..., -3.3062e-07, + 7.9162e-08, 3.6601e-07], + [-1.7742e-06, -1.3877e-07, 8.1025e-08, ..., 5.9605e-08, + 1.8254e-07, 1.3877e-07], + ..., + [-7.5437e-08, 1.3318e-07, -5.4054e-06, ..., 1.4529e-07, + 2.1327e-07, 2.0321e-06], + [ 1.1548e-07, 2.0582e-07, 1.3420e-06, ..., 1.4249e-07, + 4.1276e-06, 5.6587e-06], + [ 1.6112e-07, 7.8231e-07, 5.1968e-07, ..., 4.5542e-07, + 1.9874e-06, -1.4650e-06]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0067, -0.0162, -0.0250, 0.0324, -0.0301, 0.0301, -0.0280, -0.0119, + 0.0209, -0.0285], device='cuda:0'), grad: tensor([-1.2770e-05, 6.2175e-06, -3.6769e-06, 1.7416e-06, -5.6326e-06, + -1.7546e-06, 4.2189e-07, -9.4064e-07, 1.3158e-05, 3.2689e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 215.17, cls_loss 0.0020 cls_loss_mapping 0.0038 cls_loss_causal 0.5072 re_mapping 0.0062 re_causal 0.0182 /// teacc 98.97 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0870, -0.1094, -0.0846, ..., -0.1274, -0.0066, 0.0749], + [ 0.0505, 0.0523, 0.0418, ..., 0.0895, -0.1171, -0.0517], + [ 0.0467, 0.0318, -0.1054, ..., -0.0122, -0.1529, -0.0654], + ..., + [-0.0403, -0.0770, 0.0747, ..., -0.0921, -0.0220, 0.0111], + [ 0.0174, 0.0866, 0.1226, ..., -0.1025, -0.0853, -0.1355], + [-0.1213, 0.0105, -0.1861, ..., -0.0907, -0.0822, 0.0740]], + device='cuda:0'), grad: tensor([[ 3.8184e-08, 4.7497e-08, 5.0291e-08, ..., 3.6322e-08, + -4.8988e-07, -1.4706e-06], + [-5.6252e-07, -4.3400e-07, -6.9197e-07, ..., -3.9209e-07, + 2.4587e-07, -3.7253e-08], + [ 8.3819e-08, 9.3132e-08, 1.5739e-07, ..., 1.1176e-07, + 7.7300e-08, 9.8720e-08], + ..., + [ 1.2014e-07, 2.5332e-07, -1.2014e-07, ..., 8.0187e-07, + 1.0235e-06, -1.0617e-07], + [-3.8333e-06, -5.0999e-06, -6.6981e-06, ..., -2.2296e-06, + -2.2333e-06, 8.7358e-07], + [ 1.1716e-06, 1.0710e-06, 1.1334e-06, ..., 3.6322e-07, + 6.3609e-07, 3.1758e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0067, -0.0158, -0.0244, 0.0323, -0.0300, 0.0301, -0.0281, -0.0123, + 0.0209, -0.0287], device='cuda:0'), grad: tensor([-2.0154e-06, -8.6520e-07, 4.8336e-07, 2.4792e-06, -2.9020e-06, + 1.9297e-06, 5.3905e-06, 1.8887e-06, -1.2696e-05, 6.3069e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 215.08, cls_loss 0.0019 cls_loss_mapping 0.0047 cls_loss_causal 0.5500 re_mapping 0.0060 re_causal 0.0189 /// teacc 98.87 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0876, -0.1094, -0.0849, ..., -0.1275, -0.0065, 0.0754], + [ 0.0504, 0.0522, 0.0416, ..., 0.0895, -0.1177, -0.0517], + [ 0.0468, 0.0315, -0.1059, ..., -0.0125, -0.1546, -0.0656], + ..., + [-0.0404, -0.0776, 0.0752, ..., -0.0922, -0.0215, 0.0114], + [ 0.0168, 0.0861, 0.1227, ..., -0.1031, -0.0856, -0.1358], + [-0.1219, 0.0102, -0.1875, ..., -0.0916, -0.0841, 0.0737]], + device='cuda:0'), grad: tensor([[ 4.8522e-07, 3.8087e-05, 7.8604e-07, ..., 5.6811e-08, + 7.6056e-05, 3.9488e-05], + [ 7.0408e-07, 3.5670e-06, 1.4333e-06, ..., 2.6915e-07, + 5.8487e-06, 3.5949e-06], + [ 4.8317e-06, 3.2745e-06, 9.7901e-06, ..., 1.1362e-07, + 3.7067e-06, 2.0012e-05], + ..., + [ 5.3644e-07, 1.2349e-06, -5.8673e-08, ..., 8.9407e-07, + 2.3972e-06, 2.1886e-06], + [-2.9467e-06, 4.9412e-05, -1.9558e-06, ..., 6.4168e-07, + 1.0186e-04, 5.3942e-05], + [ 2.3469e-07, 6.5304e-06, 4.1071e-07, ..., 5.9139e-07, + 1.2964e-05, 6.5565e-06]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0070, -0.0160, -0.0245, 0.0325, -0.0295, 0.0301, -0.0277, -0.0119, + 0.0203, -0.0295], device='cuda:0'), grad: tensor([ 1.5378e-04, 1.5408e-05, 6.5446e-05, 1.2994e-04, 2.2687e-06, + 2.6798e-04, -8.7929e-04, 9.9093e-06, 2.0778e-04, 2.7075e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 214.82, cls_loss 0.0016 cls_loss_mapping 0.0037 cls_loss_causal 0.5001 re_mapping 0.0059 re_causal 0.0184 /// teacc 98.97 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0880, -0.1093, -0.0847, ..., -0.1279, -0.0065, 0.0760], + [ 0.0504, 0.0522, 0.0417, ..., 0.0895, -0.1181, -0.0518], + [ 0.0476, 0.0313, -0.1048, ..., -0.0126, -0.1554, -0.0658], + ..., + [-0.0413, -0.0784, 0.0748, ..., -0.0924, -0.0211, 0.0114], + [ 0.0170, 0.0846, 0.1216, ..., -0.1059, -0.0870, -0.1364], + [-0.1220, 0.0105, -0.1878, ..., -0.0917, -0.0841, 0.0740]], + device='cuda:0'), grad: tensor([[ 1.9092e-07, 7.5437e-07, 1.1455e-07, ..., 3.5390e-08, + -1.4808e-07, 1.1306e-06], + [-2.8033e-07, -4.6566e-09, -3.5297e-07, ..., -2.5984e-07, + 1.1269e-07, 3.8836e-07], + [-1.5832e-07, 1.0431e-07, 2.4773e-07, ..., 7.2643e-08, + 4.4703e-08, 1.6298e-07], + ..., + [ 5.4855e-07, 1.8226e-06, 1.2107e-08, ..., 2.0117e-07, + 6.6124e-08, 4.2021e-06], + [-1.5823e-06, -3.3434e-07, -1.9670e-06, ..., 2.0117e-07, + 2.1514e-07, 5.4669e-07], + [-2.4959e-07, -3.7439e-06, 2.1048e-07, ..., 2.4959e-07, + 2.5518e-07, -9.7975e-06]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0074, -0.0160, -0.0234, 0.0324, -0.0299, 0.0301, -0.0263, -0.0125, + 0.0191, -0.0294], device='cuda:0'), grad: tensor([ 1.7406e-06, 2.1793e-07, 1.1548e-07, 1.4845e-06, 2.4959e-06, + 1.2685e-06, 1.0813e-06, 6.4969e-06, -2.0750e-06, -1.2837e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 214.94, cls_loss 0.0020 cls_loss_mapping 0.0039 cls_loss_causal 0.5524 re_mapping 0.0063 re_causal 0.0196 /// teacc 99.04 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0885, -0.1129, -0.0850, ..., -0.1282, -0.0092, 0.0737], + [ 0.0511, 0.0524, 0.0426, ..., 0.0896, -0.1185, -0.0515], + [ 0.0479, 0.0312, -0.1051, ..., -0.0127, -0.1569, -0.0661], + ..., + [-0.0426, -0.0804, 0.0742, ..., -0.0924, -0.0203, 0.0113], + [ 0.0172, 0.0841, 0.1220, ..., -0.1060, -0.0877, -0.1388], + [-0.1220, 0.0111, -0.1887, ..., -0.0919, -0.0841, 0.0743]], + device='cuda:0'), grad: tensor([[ 1.4529e-07, 2.0489e-07, 1.2387e-07, ..., 2.6077e-08, + 1.0431e-07, -1.0524e-07], + [-1.9986e-06, -1.1651e-06, -2.0470e-06, ..., -1.0701e-06, + 2.7567e-07, 3.6135e-07], + [ 6.0070e-07, 9.0897e-07, 7.9907e-07, ..., 1.2480e-07, + 6.7055e-08, 1.1735e-07], + ..., + [ 1.2526e-06, 7.5158e-07, -1.6829e-06, ..., 2.5798e-07, + -6.2678e-07, -2.2966e-06], + [-9.8627e-07, -3.3509e-06, -1.9539e-06, ..., 1.9837e-07, + -1.9465e-07, 5.1968e-07], + [ 1.6792e-06, 6.2957e-07, 2.7139e-06, ..., 2.3004e-07, + 5.5972e-07, 5.8394e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0050, -0.0157, -0.0233, 0.0326, -0.0302, 0.0300, -0.0247, -0.0129, + 0.0188, -0.0293], device='cuda:0'), grad: tensor([ 3.4552e-07, -2.6524e-06, 1.8850e-06, -4.9919e-06, 3.2969e-06, + 4.8801e-07, 1.5451e-06, -3.6750e-06, -4.1761e-06, 7.9572e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 214.97, cls_loss 0.0023 cls_loss_mapping 0.0046 cls_loss_causal 0.5244 re_mapping 0.0058 re_causal 0.0174 /// teacc 98.96 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0894, -0.1126, -0.0853, ..., -0.1284, -0.0077, 0.0750], + [ 0.0508, 0.0524, 0.0424, ..., 0.0897, -0.1187, -0.0515], + [ 0.0485, 0.0315, -0.1050, ..., -0.0124, -0.1579, -0.0665], + ..., + [-0.0425, -0.0812, 0.0745, ..., -0.0925, -0.0204, 0.0109], + [ 0.0175, 0.0846, 0.1225, ..., -0.1061, -0.0881, -0.1399], + [-0.1225, 0.0108, -0.1894, ..., -0.0925, -0.0856, 0.0744]], + device='cuda:0'), grad: tensor([[ 5.1595e-07, 1.6792e-06, 2.4680e-07, ..., 1.5832e-08, + 3.5297e-06, 1.3597e-06], + [ 2.6450e-07, 1.0245e-07, 1.2666e-07, ..., -1.0990e-07, + 3.5111e-07, 4.5449e-07], + [-2.3562e-07, 1.8626e-07, 1.6205e-07, ..., 1.3039e-08, + 4.3213e-07, 3.6787e-07], + ..., + [ 5.8301e-07, 6.7055e-08, -5.5041e-07, ..., 1.2107e-08, + 1.8720e-07, 4.1910e-08], + [ 4.5355e-07, -1.5683e-06, -1.6866e-06, ..., 2.0489e-08, + 2.7400e-06, 1.2359e-06], + [ 8.0094e-07, 7.9907e-07, 8.4750e-07, ..., 3.0734e-08, + 2.7101e-07, 4.0326e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0060, -0.0159, -0.0229, 0.0326, -0.0301, 0.0283, -0.0229, -0.0129, + 0.0190, -0.0297], device='cuda:0'), grad: tensor([ 7.3612e-06, 2.5146e-06, 1.4286e-06, -7.8738e-05, 4.7218e-07, + 6.7294e-05, -8.5533e-06, 1.7071e-06, 1.5208e-06, 4.9472e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 215.09, cls_loss 0.0026 cls_loss_mapping 0.0043 cls_loss_causal 0.5123 re_mapping 0.0059 re_causal 0.0172 /// teacc 98.94 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0910, -0.1129, -0.0857, ..., -0.1285, -0.0077, 0.0751], + [ 0.0508, 0.0527, 0.0421, ..., 0.0899, -0.1193, -0.0515], + [ 0.0489, 0.0313, -0.1065, ..., -0.0127, -0.1587, -0.0664], + ..., + [-0.0425, -0.0820, 0.0755, ..., -0.0926, -0.0207, 0.0117], + [ 0.0180, 0.0847, 0.1233, ..., -0.1062, -0.0885, -0.1407], + [-0.1244, 0.0127, -0.1912, ..., -0.0912, -0.0840, 0.0752]], + device='cuda:0'), grad: tensor([[ 2.4103e-06, 2.0154e-06, 5.0142e-06, ..., 6.7428e-07, + -8.3260e-07, 1.2284e-06], + [-1.5247e-04, -1.0467e-04, -1.5545e-04, ..., -3.6865e-05, + 2.0023e-07, -3.0011e-05], + [ 3.4571e-06, 1.6112e-06, 3.2485e-06, ..., 9.2387e-07, + 1.6205e-07, 2.3246e-06], + ..., + [ 1.0669e-05, 4.2915e-06, -6.7800e-07, ..., 1.7975e-06, + 1.4249e-07, -2.5511e-05], + [ 1.2136e-04, 9.0241e-05, 1.2577e-04, ..., 3.1352e-05, + 4.0513e-07, 2.8789e-05], + [ 3.5539e-06, -8.5216e-07, 8.5756e-06, ..., 1.5944e-06, + -6.4448e-07, 5.5805e-06]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0059, -0.0162, -0.0228, 0.0326, -0.0319, 0.0284, -0.0231, -0.0122, + 0.0191, -0.0283], device='cuda:0'), grad: tensor([ 1.8269e-05, -3.3951e-04, 1.1906e-05, 1.6198e-05, 1.2338e-05, + 2.3186e-05, 4.1425e-06, -7.1824e-05, 2.8968e-04, 3.5346e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 214.92, cls_loss 0.0021 cls_loss_mapping 0.0045 cls_loss_causal 0.5357 re_mapping 0.0059 re_causal 0.0182 /// teacc 98.89 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0891, -0.1134, -0.0863, ..., -0.1311, -0.0087, 0.0755], + [ 0.0510, 0.0526, 0.0419, ..., 0.0895, -0.1211, -0.0519], + [ 0.0496, 0.0313, -0.1069, ..., -0.0129, -0.1597, -0.0664], + ..., + [-0.0426, -0.0834, 0.0757, ..., -0.0930, -0.0215, 0.0116], + [ 0.0180, 0.0847, 0.1234, ..., -0.1065, -0.0890, -0.1416], + [-0.1276, 0.0133, -0.1918, ..., -0.0914, -0.0838, 0.0758]], + device='cuda:0'), grad: tensor([[ 1.1599e-04, 2.0730e-04, 8.2888e-08, ..., 1.5926e-07, + 5.5134e-07, 9.2363e-04], + [ 1.5363e-05, 1.0267e-05, 1.6108e-05, ..., -7.4506e-08, + 1.3784e-07, 1.4417e-06], + [ 2.3976e-05, 3.3170e-05, 9.9838e-06, ..., 7.9162e-08, + 6.0536e-08, 1.2141e-04], + ..., + [ 1.1763e-06, 1.3513e-06, 6.5193e-08, ..., 3.2596e-08, + 1.3039e-08, 4.0978e-06], + [-2.7016e-05, -1.8597e-05, -3.1263e-05, ..., 7.2829e-07, + 2.0638e-06, 2.2221e-06], + [-1.3363e-04, -2.3890e-04, 3.2317e-07, ..., 1.3132e-07, + 2.0582e-07, -1.0662e-03]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0057, -0.0167, -0.0225, 0.0326, -0.0321, 0.0285, -0.0230, -0.0123, + 0.0189, -0.0279], device='cuda:0'), grad: tensor([ 1.1873e-03, 2.8849e-05, 1.7095e-04, 4.5858e-06, 1.3195e-05, + 6.4299e-06, -2.8759e-06, 6.1058e-06, -4.6164e-05, -1.3695e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 214.72, cls_loss 0.0021 cls_loss_mapping 0.0042 cls_loss_causal 0.5118 re_mapping 0.0064 re_causal 0.0186 /// teacc 98.86 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0895, -0.1139, -0.0860, ..., -0.1303, -0.0089, 0.0750], + [ 0.0504, 0.0522, 0.0410, ..., 0.0892, -0.1224, -0.0527], + [ 0.0497, 0.0311, -0.1077, ..., -0.0131, -0.1609, -0.0669], + ..., + [-0.0420, -0.0844, 0.0770, ..., -0.0930, -0.0215, 0.0117], + [ 0.0183, 0.0845, 0.1239, ..., -0.1067, -0.0901, -0.1433], + [-0.1272, 0.0134, -0.1925, ..., -0.0924, -0.0841, 0.0767]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, 2.3283e-08, 3.7253e-09, ..., 1.2107e-08, + 1.6764e-08, -1.2200e-07], + [-9.3132e-09, 7.4506e-09, -1.2107e-08, ..., -6.1467e-08, + 5.8673e-08, 1.2759e-07], + [-6.6124e-07, -1.3039e-08, 1.8626e-08, ..., 4.6566e-09, + 9.3132e-09, 3.2596e-08], + ..., + [ 8.2888e-08, 1.6019e-07, -2.0396e-07, ..., 2.1420e-08, + 2.4214e-08, 4.5728e-07], + [ 2.5425e-07, 2.3283e-08, 3.4459e-08, ..., 2.6077e-08, + 2.4959e-07, 3.9395e-07], + [ 3.4459e-08, -1.0151e-07, 5.2154e-08, ..., 3.0175e-07, + 2.0023e-07, -9.1176e-07]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0054, -0.0180, -0.0227, 0.0325, -0.0319, 0.0290, -0.0232, -0.0111, + 0.0185, -0.0278], device='cuda:0'), grad: tensor([-4.0047e-08, 3.2317e-07, -1.3048e-06, 6.1933e-07, -7.3574e-08, + -2.0452e-06, 1.0217e-06, 6.2026e-07, 1.5171e-06, -6.2957e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.97, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.4995 re_mapping 0.0062 re_causal 0.0182 /// teacc 99.04 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0901, -0.1141, -0.0864, ..., -0.1303, -0.0091, 0.0741], + [ 0.0505, 0.0523, 0.0410, ..., 0.0892, -0.1229, -0.0533], + [ 0.0485, 0.0304, -0.1080, ..., -0.0138, -0.1619, -0.0696], + ..., + [-0.0424, -0.0852, 0.0771, ..., -0.0931, -0.0216, 0.0109], + [ 0.0186, 0.0847, 0.1243, ..., -0.1069, -0.0905, -0.1440], + [-0.1263, 0.0131, -0.1930, ..., -0.0933, -0.0844, 0.0784]], + device='cuda:0'), grad: tensor([[ 1.2480e-07, 4.8056e-07, 4.8429e-08, ..., 8.4750e-08, + 8.3447e-07, 4.3027e-07], + [ 3.5390e-08, 1.1083e-07, 1.8906e-07, ..., 4.3772e-08, + 4.9733e-07, 8.0373e-07], + [-1.3718e-06, 3.8184e-08, 9.1270e-08, ..., 1.8068e-07, + 4.2096e-07, 2.3842e-07], + ..., + [ 1.3504e-07, 8.0094e-08, -1.4398e-06, ..., 1.4342e-07, + 4.7591e-07, -3.2224e-07], + [-1.2573e-06, 5.9325e-07, -2.4904e-06, ..., -1.1083e-07, + 3.7085e-06, 3.9712e-06], + [ 1.3132e-07, 2.8219e-07, 1.0086e-06, ..., 2.4121e-07, + 9.5181e-07, 2.4345e-06]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0046, -0.0182, -0.0246, 0.0324, -0.0314, 0.0295, -0.0231, -0.0116, + 0.0186, -0.0276], device='cuda:0'), grad: tensor([ 3.4049e-06, 3.3360e-06, -1.5376e-06, 9.2834e-06, -6.3330e-08, + -3.4183e-05, -2.0079e-06, -1.9092e-07, 1.3396e-05, 8.5160e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 214.69, cls_loss 0.0020 cls_loss_mapping 0.0044 cls_loss_causal 0.5251 re_mapping 0.0058 re_causal 0.0176 /// teacc 98.95 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0910, -0.1143, -0.0868, ..., -0.1304, -0.0090, 0.0744], + [ 0.0511, 0.0529, 0.0414, ..., 0.0896, -0.1236, -0.0532], + [ 0.0500, 0.0310, -0.1082, ..., -0.0141, -0.1630, -0.0695], + ..., + [-0.0434, -0.0861, 0.0773, ..., -0.0933, -0.0219, 0.0107], + [ 0.0177, 0.0840, 0.1242, ..., -0.1075, -0.0909, -0.1455], + [-0.1268, 0.0131, -0.1936, ..., -0.0938, -0.0845, 0.0788]], + device='cuda:0'), grad: tensor([[ 9.7789e-08, 1.4342e-07, 2.3283e-08, ..., -4.3772e-07, + -3.5167e-06, -6.5938e-06], + [-2.1979e-07, -2.6915e-07, -4.7963e-07, ..., -6.5751e-07, + 1.1362e-07, 5.4203e-07], + [-1.7881e-07, 1.2945e-07, 1.3690e-07, ..., 8.2888e-08, + 6.7055e-08, 8.5682e-08], + ..., + [ 4.8243e-07, 2.0023e-07, -5.2154e-08, ..., 1.8161e-07, + 5.1223e-08, 5.1316e-07], + [-6.7987e-08, 2.1048e-07, -2.8964e-07, ..., 2.2911e-07, + 6.7428e-07, 1.4901e-06], + [-1.4985e-06, -2.1569e-06, 1.2759e-07, ..., 2.0117e-07, + 2.2259e-07, -8.7097e-06]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0046, -0.0179, -0.0234, 0.0327, -0.0313, 0.0293, -0.0231, -0.0121, + 0.0179, -0.0276], device='cuda:0'), grad: tensor([-9.3356e-06, 2.3283e-07, 1.3411e-07, 1.6401e-06, 1.9055e-06, + 4.8578e-06, 9.0301e-06, 1.5311e-06, 2.5313e-06, -1.2510e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 214.67, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.5137 re_mapping 0.0060 re_causal 0.0185 /// teacc 98.92 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0914, -0.1143, -0.0869, ..., -0.1305, -0.0088, 0.0745], + [ 0.0511, 0.0529, 0.0414, ..., 0.0897, -0.1239, -0.0533], + [ 0.0502, 0.0310, -0.1085, ..., -0.0145, -0.1639, -0.0696], + ..., + [-0.0437, -0.0866, 0.0775, ..., -0.0935, -0.0219, 0.0105], + [ 0.0181, 0.0842, 0.1245, ..., -0.1076, -0.0912, -0.1460], + [-0.1274, 0.0131, -0.1944, ..., -0.0940, -0.0846, 0.0793]], + device='cuda:0'), grad: tensor([[ 3.7113e-07, 4.6333e-07, 5.8813e-07, ..., 2.4540e-07, + 2.5332e-07, -2.6217e-07], + [-1.4752e-05, -1.0446e-05, -1.9222e-05, ..., -1.4484e-05, + -4.2096e-07, 4.9360e-08], + [ 1.8086e-06, 1.0487e-06, 1.9521e-06, ..., 8.5216e-07, + 5.0757e-07, 2.4680e-08], + ..., + [ 1.0855e-05, 6.7577e-06, 1.3076e-05, ..., 8.9109e-06, + 4.9081e-07, 2.1700e-07], + [-9.8720e-06, -5.8860e-06, -1.7419e-05, ..., -4.2059e-06, + -1.7518e-06, 1.1967e-07], + [ 1.0403e-06, 1.8077e-06, 1.3765e-06, ..., 2.1625e-06, + 1.5143e-06, -1.0617e-06]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0047, -0.0180, -0.0233, 0.0327, -0.0313, 0.0292, -0.0232, -0.0123, + 0.0180, -0.0275], device='cuda:0'), grad: tensor([ 1.3672e-06, -3.7044e-05, 6.1020e-06, -2.1346e-06, -1.2144e-06, + 3.3975e-05, 2.9355e-06, 2.6897e-05, -3.4481e-05, 3.6117e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 215.12, cls_loss 0.0018 cls_loss_mapping 0.0040 cls_loss_causal 0.5291 re_mapping 0.0059 re_causal 0.0189 /// teacc 98.96 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0921, -0.1145, -0.0872, ..., -0.1307, -0.0089, 0.0747], + [ 0.0546, 0.0539, 0.0426, ..., 0.0906, -0.1243, -0.0522], + [ 0.0476, 0.0300, -0.1100, ..., -0.0145, -0.1636, -0.0696], + ..., + [-0.0450, -0.0884, 0.0768, ..., -0.0947, -0.0221, 0.0095], + [ 0.0185, 0.0844, 0.1250, ..., -0.1076, -0.0915, -0.1467], + [-0.1272, 0.0133, -0.1951, ..., -0.0938, -0.0847, 0.0799]], + device='cuda:0'), grad: tensor([[ 7.3994e-07, 1.4296e-07, 2.4820e-07, ..., 1.9558e-07, + 7.3574e-08, 1.8068e-07], + [-6.0014e-06, -3.1963e-06, -6.9775e-06, ..., -7.4282e-06, + 1.7695e-08, 5.3551e-07], + [-8.5384e-06, 3.2596e-08, 1.5786e-06, ..., 8.3121e-07, + 1.0710e-08, -8.9081e-07], + ..., + [ 2.0210e-07, 2.3423e-07, -8.5086e-06, ..., 3.5716e-07, + 9.3132e-10, -2.9206e-06], + [ 3.1758e-06, 1.1073e-06, 1.3914e-06, ..., 2.7549e-06, + 5.2620e-08, 8.4424e-07], + [ 1.7500e-06, -4.0513e-07, 1.7053e-06, ..., 1.7090e-07, + 6.9849e-09, 8.8243e-07]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0047, -0.0158, -0.0252, 0.0326, -0.0315, 0.0290, -0.0231, -0.0133, + 0.0182, -0.0272], device='cuda:0'), grad: tensor([ 2.1346e-06, -7.7486e-06, -2.2158e-05, 9.1791e-06, 7.0408e-06, + 8.6427e-06, 6.1281e-06, -1.3985e-05, 5.1968e-06, 5.5581e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 214.69, cls_loss 0.0020 cls_loss_mapping 0.0046 cls_loss_causal 0.5386 re_mapping 0.0059 re_causal 0.0176 /// teacc 98.90 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0922, -0.1146, -0.0872, ..., -0.1305, -0.0087, 0.0750], + [ 0.0544, 0.0544, 0.0429, ..., 0.0917, -0.1245, -0.0511], + [ 0.0478, 0.0301, -0.1107, ..., -0.0148, -0.1646, -0.0700], + ..., + [-0.0446, -0.0904, 0.0773, ..., -0.0963, -0.0227, 0.0088], + [ 0.0187, 0.0847, 0.1253, ..., -0.1077, -0.0919, -0.1475], + [-0.1280, 0.0132, -0.1965, ..., -0.0943, -0.0851, 0.0800]], + device='cuda:0'), grad: tensor([[ 1.4389e-07, 8.9267e-07, 2.6124e-07, ..., 4.0745e-07, + 9.7789e-07, -3.6648e-07], + [ 8.0559e-07, 3.8976e-07, 6.7009e-07, ..., 1.5274e-07, + 2.9569e-07, 2.2259e-07], + [ 7.6648e-07, 5.3132e-07, 1.0971e-06, ..., 5.6345e-08, + 1.4622e-07, 7.2643e-08], + ..., + [ 2.1216e-06, 9.4296e-07, -7.5018e-07, ..., 2.8359e-07, + 1.9465e-07, -3.5902e-07], + [-6.6422e-06, -2.3656e-06, -6.8322e-06, ..., 3.3714e-07, + 7.6368e-07, 4.2142e-07], + [ 5.2899e-07, 4.9686e-07, 1.5479e-06, ..., 7.7765e-07, + 6.7614e-07, -1.2992e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0051, -0.0153, -0.0254, 0.0320, -0.0312, 0.0296, -0.0234, -0.0135, + 0.0182, -0.0276], device='cuda:0'), grad: tensor([ 1.9837e-06, 3.6992e-06, 2.2091e-06, 3.1441e-06, -3.0883e-06, + 4.2021e-06, -4.1872e-06, -4.6007e-07, -1.4126e-05, 6.5453e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.83, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5314 re_mapping 0.0057 re_causal 0.0178 /// teacc 99.05 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0920, -0.1147, -0.0873, ..., -0.1309, -0.0090, 0.0754], + [ 0.0541, 0.0543, 0.0423, ..., 0.0918, -0.1251, -0.0512], + [ 0.0481, 0.0302, -0.1110, ..., -0.0152, -0.1658, -0.0704], + ..., + [-0.0442, -0.0907, 0.0780, ..., -0.0964, -0.0230, 0.0088], + [ 0.0187, 0.0857, 0.1256, ..., -0.1079, -0.0917, -0.1482], + [-0.1285, 0.0132, -0.1973, ..., -0.0944, -0.0852, 0.0800]], + device='cuda:0'), grad: tensor([[ 8.4378e-07, 6.2399e-08, 3.7719e-08, ..., -6.3796e-07, + -2.6654e-06, -3.9563e-06], + [ 8.9779e-07, -6.6077e-07, 5.2107e-07, ..., -2.7400e-06, + 1.9977e-07, -8.9826e-07], + [-6.4149e-06, -1.8049e-06, -1.0077e-06, ..., 1.1874e-07, + 6.7521e-08, 3.5297e-07], + ..., + [ 1.3290e-06, 2.7753e-07, -1.8226e-06, ..., 6.0396e-07, + 2.4354e-07, 1.4296e-07], + [ 8.7246e-06, 1.8077e-06, 1.4668e-06, ..., 1.5693e-07, + 2.1653e-07, 8.7405e-07], + [ 1.3784e-06, 3.8557e-07, 5.8021e-07, ..., 1.4585e-06, + 6.3283e-07, -3.0100e-06]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0054, -0.0158, -0.0253, 0.0320, -0.0312, 0.0297, -0.0238, -0.0132, + 0.0188, -0.0277], device='cuda:0'), grad: tensor([-4.7013e-06, -1.2722e-06, -6.7092e-06, -1.8358e-05, 2.4354e-07, + 1.3700e-06, 9.1046e-06, 2.2277e-06, 1.6153e-05, 1.8459e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 215.09, cls_loss 0.0016 cls_loss_mapping 0.0038 cls_loss_causal 0.5300 re_mapping 0.0059 re_causal 0.0182 /// teacc 99.06 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0924, -0.1148, -0.0875, ..., -0.1308, -0.0090, 0.0756], + [ 0.0545, 0.0547, 0.0427, ..., 0.0923, -0.1257, -0.0510], + [ 0.0484, 0.0300, -0.1114, ..., -0.0158, -0.1664, -0.0703], + ..., + [-0.0450, -0.0919, 0.0779, ..., -0.0969, -0.0232, 0.0085], + [ 0.0187, 0.0857, 0.1257, ..., -0.1080, -0.0920, -0.1490], + [-0.1303, 0.0130, -0.1979, ..., -0.0956, -0.0858, 0.0800]], + device='cuda:0'), grad: tensor([[ 3.0780e-07, 1.1856e-06, 3.3574e-07, ..., 5.8440e-07, + 5.1036e-07, 7.0781e-08], + [ 1.3812e-06, 3.4142e-06, 3.1665e-06, ..., 2.2836e-06, + 1.6820e-06, 3.1758e-07], + [ 1.1576e-06, 3.3267e-06, 1.8068e-06, ..., 1.6335e-06, + 1.3411e-06, 1.7090e-07], + ..., + [ 5.9512e-07, 7.0548e-07, 7.9162e-08, ..., 3.2503e-07, + 1.2154e-07, -3.9442e-07], + [-9.9093e-06, -1.4119e-05, -7.6741e-06, ..., 1.4743e-06, + 3.1311e-06, -8.6520e-07], + [ 1.1986e-06, 2.2054e-06, 9.3598e-07, ..., 4.9174e-07, + 2.0731e-06, 4.6268e-06]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0055, -0.0156, -0.0250, 0.0324, -0.0309, 0.0295, -0.0238, -0.0135, + 0.0186, -0.0282], device='cuda:0'), grad: tensor([ 2.6897e-06, 9.6112e-06, 7.6964e-06, 2.9057e-06, 4.4182e-06, + -3.0398e-06, -9.5293e-06, 5.2387e-07, -2.5302e-05, 9.9540e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 215.24, cls_loss 0.0018 cls_loss_mapping 0.0051 cls_loss_causal 0.5476 re_mapping 0.0056 re_causal 0.0172 /// teacc 99.07 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0929, -0.1149, -0.0873, ..., -0.1306, -0.0086, 0.0766], + [ 0.0552, 0.0551, 0.0438, ..., 0.0922, -0.1267, -0.0511], + [ 0.0486, 0.0303, -0.1123, ..., -0.0160, -0.1660, -0.0705], + ..., + [-0.0466, -0.0938, 0.0769, ..., -0.0968, -0.0232, 0.0085], + [ 0.0190, 0.0857, 0.1261, ..., -0.1081, -0.0929, -0.1505], + [-0.1311, 0.0129, -0.1990, ..., -0.0961, -0.0864, 0.0794]], + device='cuda:0'), grad: tensor([[ 3.5763e-07, -1.0803e-07, 2.0349e-07, ..., 2.7148e-07, + -9.5554e-07, -1.3344e-05], + [-2.0210e-06, -2.0657e-06, -2.4531e-06, ..., -3.1069e-06, + -1.5646e-07, -1.5786e-07], + [-2.1979e-06, -2.1141e-07, -1.7602e-07, ..., -1.8254e-07, + 2.4587e-07, 1.5944e-06], + ..., + [ 1.1129e-06, 5.2107e-07, -7.8417e-07, ..., 7.4087e-07, + 8.0932e-07, -1.0384e-07], + [ 7.4925e-07, 8.8476e-07, 2.2026e-07, ..., 6.7940e-07, + 3.9935e-06, 4.5635e-06], + [ 9.5740e-07, 8.7591e-07, 1.5674e-06, ..., 9.3877e-07, + 1.1995e-06, 7.3835e-06]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0062, -0.0153, -0.0250, 0.0326, -0.0309, 0.0297, -0.0237, -0.0141, + 0.0184, -0.0287], device='cuda:0'), grad: tensor([-2.3752e-05, -3.0305e-06, -1.8328e-06, 1.5043e-05, 7.7486e-07, + -2.6494e-05, 7.5251e-06, 1.8198e-06, 1.3307e-05, 1.6570e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 214.99, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5287 re_mapping 0.0060 re_causal 0.0169 /// teacc 99.02 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0936, -0.1150, -0.0880, ..., -0.1306, -0.0081, 0.0775], + [ 0.0551, 0.0554, 0.0436, ..., 0.0929, -0.1277, -0.0512], + [ 0.0495, 0.0297, -0.1119, ..., -0.0178, -0.1673, -0.0695], + ..., + [-0.0467, -0.0944, 0.0775, ..., -0.0973, -0.0237, 0.0087], + [ 0.0193, 0.0858, 0.1262, ..., -0.1083, -0.0934, -0.1515], + [-0.1336, 0.0128, -0.2007, ..., -0.0972, -0.0870, 0.0786]], + device='cuda:0'), grad: tensor([[ 3.3341e-07, 4.5169e-08, 3.2363e-07, ..., -8.1630e-07, + -1.8990e-06, -4.2506e-06], + [ 1.5451e-06, 1.6764e-08, 1.5192e-05, ..., 1.2191e-06, + 7.2643e-08, 4.8429e-08], + [-7.1526e-06, 1.0617e-07, 2.1923e-06, ..., 3.6974e-07, + 7.4506e-08, 3.4459e-08], + ..., + [-1.2051e-06, -5.9465e-07, -3.8832e-05, ..., -4.3511e-06, + 1.8161e-08, 8.2888e-08], + [ 2.2929e-06, 2.9011e-07, 7.4320e-06, ..., 1.0468e-06, + 2.1188e-07, 3.2457e-07], + [ 6.1886e-07, 4.7497e-08, 5.7220e-06, ..., 9.7975e-07, + 2.3702e-07, -5.3132e-07]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0069, -0.0157, -0.0236, 0.0325, -0.0306, 0.0297, -0.0237, -0.0140, + 0.0181, -0.0295], device='cuda:0'), grad: tensor([-5.4277e-06, 2.7657e-05, -8.9332e-06, 4.6119e-06, 9.8869e-06, + 6.6124e-06, 1.9893e-06, -6.2943e-05, 1.6674e-05, 9.8497e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 215.09, cls_loss 0.0020 cls_loss_mapping 0.0032 cls_loss_causal 0.5392 re_mapping 0.0065 re_causal 0.0186 /// teacc 98.99 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0949, -0.1152, -0.0890, ..., -0.1307, -0.0081, 0.0775], + [ 0.0554, 0.0557, 0.0439, ..., 0.0936, -0.1282, -0.0510], + [ 0.0496, 0.0289, -0.1142, ..., -0.0204, -0.1677, -0.0697], + ..., + [-0.0469, -0.0950, 0.0783, ..., -0.0976, -0.0239, 0.0102], + [ 0.0199, 0.0865, 0.1275, ..., -0.1070, -0.0937, -0.1525], + [-0.1345, 0.0128, -0.2033, ..., -0.0978, -0.0872, 0.0777]], + device='cuda:0'), grad: tensor([[ 1.5320e-07, 6.5658e-08, 1.0990e-07, ..., -8.1956e-08, + -1.5683e-06, -2.9933e-06], + [-3.1851e-07, 1.5991e-06, -3.1013e-07, ..., 1.5227e-06, + 9.3551e-07, 3.0734e-07], + [ 1.0524e-06, 5.7276e-08, 7.5530e-07, ..., 8.8010e-08, + 5.9139e-08, 1.4342e-07], + ..., + [ 3.9581e-07, 5.6112e-07, 2.7055e-07, ..., 5.9791e-07, + 1.8673e-07, 4.3539e-07], + [-7.3388e-06, 7.4366e-07, -4.5858e-06, ..., 8.5682e-07, + 6.4122e-07, 8.4843e-07], + [ 9.7789e-08, 3.2466e-06, 6.3470e-07, ..., 4.2133e-06, + 1.9595e-06, -1.6894e-06]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0067, -0.0156, -0.0241, 0.0327, -0.0304, 0.0287, -0.0229, -0.0130, + 0.0188, -0.0302], device='cuda:0'), grad: tensor([-5.1744e-06, 4.4331e-06, 2.2091e-06, 9.4995e-06, -1.7807e-05, + 3.1926e-06, 3.3174e-06, 2.4475e-06, -9.0376e-06, 6.9328e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 215.07, cls_loss 0.0018 cls_loss_mapping 0.0035 cls_loss_causal 0.5435 re_mapping 0.0059 re_causal 0.0183 /// teacc 99.07 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0958, -0.1153, -0.0896, ..., -0.1310, -0.0083, 0.0772], + [ 0.0554, 0.0558, 0.0436, ..., 0.0936, -0.1290, -0.0511], + [ 0.0494, 0.0288, -0.1149, ..., -0.0206, -0.1687, -0.0700], + ..., + [-0.0468, -0.0955, 0.0789, ..., -0.0977, -0.0239, 0.0103], + [ 0.0197, 0.0865, 0.1277, ..., -0.1072, -0.0941, -0.1541], + [-0.1349, 0.0126, -0.2040, ..., -0.0986, -0.0876, 0.0780]], + device='cuda:0'), grad: tensor([[ 1.5460e-07, 4.2245e-06, 1.1362e-07, ..., 1.8887e-06, + 2.6971e-06, 3.7719e-08], + [ 2.0349e-07, -9.7789e-09, 1.5665e-06, ..., -2.3236e-07, + 1.1269e-07, 2.9383e-07], + [-4.5123e-07, 1.5153e-06, 3.9302e-07, ..., 7.2410e-07, + 9.3272e-07, 1.2433e-07], + ..., + [-2.7986e-07, 3.5390e-08, -6.7353e-06, ..., 6.6590e-08, + 1.8626e-08, -8.0606e-07], + [ 8.3074e-07, 1.9427e-06, 2.3767e-06, ..., 9.4157e-07, + 1.2321e-06, 1.6997e-07], + [ 5.3924e-07, 1.5320e-07, 8.5589e-07, ..., 2.0908e-07, + 1.4529e-07, 1.6252e-07]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0062, -0.0160, -0.0246, 0.0329, -0.0300, 0.0291, -0.0229, -0.0126, + 0.0180, -0.0304], device='cuda:0'), grad: tensor([ 1.1742e-05, 3.3267e-06, 3.6377e-06, -3.1032e-06, 1.0483e-05, + 3.4478e-06, -3.3140e-05, -9.2089e-06, 9.8273e-06, 2.9579e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 215.15, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.5226 re_mapping 0.0059 re_causal 0.0178 /// teacc 99.01 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0965, -0.1156, -0.0900, ..., -0.1312, -0.0085, 0.0770], + [ 0.0553, 0.0559, 0.0437, ..., 0.0937, -0.1298, -0.0512], + [ 0.0492, 0.0289, -0.1166, ..., -0.0209, -0.1696, -0.0701], + ..., + [-0.0463, -0.0957, 0.0797, ..., -0.0977, -0.0236, 0.0103], + [ 0.0193, 0.0864, 0.1276, ..., -0.1074, -0.0944, -0.1552], + [-0.1350, 0.0126, -0.2043, ..., -0.0991, -0.0879, 0.0784]], + device='cuda:0'), grad: tensor([[-3.1851e-07, -6.4261e-08, 2.2817e-08, ..., -7.3109e-08, + -7.7393e-07, -2.4475e-06], + [-4.1956e-07, -2.8312e-07, -6.1886e-07, ..., -5.0105e-07, + -2.4680e-08, 3.4925e-08], + [-5.9279e-07, 5.6345e-08, 6.1467e-08, ..., 8.4750e-08, + 2.2817e-08, 4.9360e-08], + ..., + [ 2.7381e-07, 6.3330e-08, 9.2201e-08, ..., 1.0291e-07, + 2.5611e-08, 2.0955e-08], + [ 4.0838e-07, 5.5414e-08, 4.8894e-08, ..., 8.5216e-08, + 6.8219e-07, 8.5821e-07], + [ 1.5879e-07, 1.7695e-08, 4.2841e-08, ..., 1.2666e-07, + 5.1316e-07, 1.7118e-06]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0058, -0.0162, -0.0255, 0.0331, -0.0299, 0.0292, -0.0229, -0.0118, + 0.0177, -0.0304], device='cuda:0'), grad: tensor([-3.4124e-06, -1.0766e-06, -8.3027e-07, 5.0571e-07, 1.6298e-07, + -1.3048e-06, 7.3714e-07, 5.1130e-07, 2.2836e-06, 2.4214e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 214.89, cls_loss 0.0019 cls_loss_mapping 0.0032 cls_loss_causal 0.5360 re_mapping 0.0057 re_causal 0.0173 /// teacc 99.04 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0971, -0.1157, -0.0899, ..., -0.1313, -0.0084, 0.0773], + [ 0.0543, 0.0556, 0.0430, ..., 0.0938, -0.1302, -0.0515], + [ 0.0500, 0.0292, -0.1164, ..., -0.0211, -0.1701, -0.0702], + ..., + [-0.0465, -0.0960, 0.0804, ..., -0.0978, -0.0240, 0.0100], + [ 0.0195, 0.0863, 0.1279, ..., -0.1074, -0.0948, -0.1560], + [-0.1351, 0.0127, -0.2047, ..., -0.0991, -0.0879, 0.0790]], + device='cuda:0'), grad: tensor([[ 3.1199e-08, 8.1491e-08, 1.7229e-08, ..., 1.1316e-07, + 4.0652e-07, 3.7719e-08], + [-5.2666e-07, -1.7975e-07, -6.7661e-07, ..., -2.3935e-07, + 3.6322e-07, 9.6392e-08], + [-5.4017e-08, 6.6590e-08, 3.5856e-08, ..., 1.1735e-07, + 2.9011e-07, 4.3306e-08], + ..., + [ 1.4389e-07, 1.9511e-07, 5.4017e-08, ..., 3.4971e-07, + 8.1956e-07, 6.6962e-07], + [ 1.7136e-07, 2.2771e-07, 1.3178e-07, ..., 3.2084e-07, + 1.3327e-06, 9.2341e-07], + [ 1.6298e-07, -2.1188e-07, 1.9977e-07, ..., 1.0356e-06, + 1.6615e-06, -1.0878e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0060, -0.0174, -0.0244, 0.0337, -0.0300, 0.0294, -0.0234, -0.0120, + 0.0174, -0.0302], device='cuda:0'), grad: tensor([ 1.2256e-06, 4.7032e-08, 9.7603e-07, 3.2783e-06, -1.1727e-05, + -4.4964e-06, -9.7603e-07, 4.2543e-06, 4.6305e-06, 2.7977e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 214.96, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.5322 re_mapping 0.0059 re_causal 0.0178 /// teacc 99.06 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0969, -0.1158, -0.0888, ..., -0.1312, -0.0084, 0.0776], + [ 0.0548, 0.0557, 0.0431, ..., 0.0942, -0.1315, -0.0515], + [ 0.0506, 0.0291, -0.1155, ..., -0.0222, -0.1709, -0.0689], + ..., + [-0.0474, -0.0961, 0.0805, ..., -0.0980, -0.0240, 0.0098], + [ 0.0194, 0.0861, 0.1279, ..., -0.1077, -0.0951, -0.1568], + [-0.1355, 0.0129, -0.2058, ..., -0.0992, -0.0881, 0.0790]], + device='cuda:0'), grad: tensor([[ 4.3306e-08, 1.2014e-07, 6.1002e-08, ..., 1.5507e-07, + 1.5227e-07, 3.3062e-08], + [-8.0839e-07, -2.5332e-07, -1.0049e-06, ..., -5.4576e-07, + 2.9383e-07, 9.9652e-08], + [ 6.1467e-08, 8.4285e-08, 2.3888e-07, ..., 1.7416e-07, + 1.0990e-07, 9.5461e-08], + ..., + [ 1.8440e-07, 2.0349e-07, -4.1584e-07, ..., 3.0315e-07, + 6.0070e-08, 1.1828e-07], + [ 9.0804e-08, 1.4305e-06, 3.4645e-07, ..., 2.6077e-07, + 1.6717e-07, 3.4031e-06], + [ 1.1083e-07, -1.7453e-06, 1.4063e-07, ..., 2.5751e-07, + 9.5926e-08, -4.6305e-06]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0064, -0.0172, -0.0237, 0.0336, -0.0302, 0.0297, -0.0236, -0.0123, + 0.0170, -0.0303], device='cuda:0'), grad: tensor([ 5.4669e-07, -1.5367e-06, 7.8836e-07, 5.5181e-07, -1.0105e-07, + 4.2561e-07, 2.6729e-07, 3.5344e-07, 9.1493e-06, -1.0431e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 214.95, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.5465 re_mapping 0.0059 re_causal 0.0178 /// teacc 98.96 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0977, -0.1160, -0.0899, ..., -0.1316, -0.0083, 0.0776], + [ 0.0550, 0.0555, 0.0432, ..., 0.0944, -0.1326, -0.0516], + [ 0.0502, 0.0289, -0.1171, ..., -0.0226, -0.1716, -0.0691], + ..., + [-0.0470, -0.0965, 0.0813, ..., -0.0982, -0.0247, 0.0101], + [ 0.0200, 0.0868, 0.1288, ..., -0.1077, -0.0951, -0.1573], + [-0.1359, 0.0129, -0.2068, ..., -0.0998, -0.0886, 0.0789]], + device='cuda:0'), grad: tensor([[ 7.5437e-08, 1.9791e-07, 9.6392e-08, ..., 2.1933e-07, + 3.4785e-07, -3.6322e-08], + [ 3.1060e-07, 2.3823e-06, 1.6587e-06, ..., 3.8054e-06, + 1.3979e-06, 1.6578e-07], + [ 1.7229e-07, 3.1758e-07, 2.0396e-07, ..., 3.3155e-07, + 2.1467e-07, 6.4261e-08], + ..., + [ 1.4389e-07, 8.7218e-07, -6.0536e-08, ..., 1.0170e-06, + 6.1188e-07, 1.6764e-07], + [-1.5814e-06, -2.7940e-09, -2.0135e-06, ..., 1.3225e-06, + 2.9597e-06, 2.1197e-06], + [ 3.0175e-07, 4.4852e-06, 9.1270e-07, ..., 5.5544e-06, + 2.6133e-06, 3.2457e-07]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0062, -0.0172, -0.0246, 0.0332, -0.0299, 0.0298, -0.0237, -0.0115, + 0.0176, -0.0307], device='cuda:0'), grad: tensor([ 1.0906e-06, 8.9779e-06, 1.7257e-06, 4.3213e-06, -3.5644e-05, + -1.8165e-05, 1.0170e-05, 3.7104e-06, 4.8131e-06, 1.8984e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 179---------------------------------------------------- +epoch 179, time 231.17, cls_loss 0.0018 cls_loss_mapping 0.0034 cls_loss_causal 0.5439 re_mapping 0.0052 re_causal 0.0166 /// teacc 99.12 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0980, -0.1162, -0.0903, ..., -0.1318, -0.0081, 0.0781], + [ 0.0528, 0.0557, 0.0405, ..., 0.0949, -0.1327, -0.0517], + [ 0.0503, 0.0289, -0.1174, ..., -0.0228, -0.1723, -0.0692], + ..., + [-0.0444, -0.0968, 0.0843, ..., -0.0984, -0.0250, 0.0096], + [ 0.0198, 0.0868, 0.1284, ..., -0.1082, -0.0956, -0.1584], + [-0.1359, 0.0128, -0.2069, ..., -0.1001, -0.0889, 0.0795]], + device='cuda:0'), grad: tensor([[ 6.0955e-07, 3.6787e-08, 3.9116e-07, ..., -6.5826e-06, + -1.2778e-05, -4.5508e-05], + [-4.2200e-05, -6.1607e-07, -4.4793e-05, ..., -4.4078e-05, + 1.6289e-06, -1.2986e-05], + [-4.7721e-06, 1.0710e-07, 2.2948e-06, ..., 4.1723e-06, + 1.1502e-07, 3.0417e-06], + ..., + [ 1.7345e-05, 3.8836e-07, 1.6332e-05, ..., 1.6391e-05, + 1.9418e-07, 7.2829e-06], + [ 7.0259e-06, 7.7300e-08, 2.8070e-06, ..., 2.2557e-06, + 7.6322e-07, 3.1646e-06], + [ 3.0734e-06, 1.0980e-06, 3.0771e-06, ..., 4.1723e-06, + 1.8990e-06, 1.2284e-06]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0065, -0.0195, -0.0247, 0.0331, -0.0297, 0.0300, -0.0239, -0.0091, + 0.0169, -0.0306], device='cuda:0'), grad: tensor([-1.0949e-04, -1.1688e-04, -2.0303e-06, 2.4915e-05, -6.4299e-06, + 4.5270e-05, 7.3075e-05, 5.2392e-05, 2.1741e-05, 1.7345e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 214.78, cls_loss 0.0020 cls_loss_mapping 0.0035 cls_loss_causal 0.5121 re_mapping 0.0057 re_causal 0.0168 /// teacc 99.07 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0998, -0.1175, -0.0910, ..., -0.1341, -0.0096, 0.0777], + [ 0.0533, 0.0570, 0.0409, ..., 0.0962, -0.1334, -0.0514], + [ 0.0507, 0.0281, -0.1177, ..., -0.0238, -0.1756, -0.0695], + ..., + [-0.0449, -0.0990, 0.0841, ..., -0.0999, -0.0255, 0.0092], + [ 0.0198, 0.0865, 0.1285, ..., -0.1087, -0.0961, -0.1602], + [-0.1366, 0.0131, -0.2072, ..., -0.1008, -0.0892, 0.0804]], + device='cuda:0'), grad: tensor([[ 1.6531e-07, 2.4354e-07, 1.8161e-08, ..., 3.4459e-08, + 1.3737e-07, 8.7637e-07], + [ 2.7474e-07, 2.9290e-07, -3.5902e-07, ..., -2.6124e-07, + 1.8440e-07, 1.8515e-06], + [-1.4529e-06, 1.1595e-07, -3.2363e-07, ..., 2.9337e-08, + 7.1712e-08, 5.4436e-07], + ..., + [ 1.2461e-06, 3.4180e-07, 4.1490e-07, ..., 1.3225e-07, + 1.0058e-07, 1.0170e-06], + [ 1.3672e-06, 6.8471e-06, -2.9337e-08, ..., 2.4103e-06, + 3.6936e-06, 9.3058e-06], + [-8.8289e-06, -1.2927e-05, 4.2375e-08, ..., 7.7765e-08, + -2.2072e-06, -5.4687e-05]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0055, -0.0190, -0.0249, 0.0333, -0.0296, 0.0299, -0.0235, -0.0096, + 0.0163, -0.0303], device='cuda:0'), grad: tensor([ 2.1085e-06, 3.9525e-06, -3.8147e-06, 3.0160e-05, 4.8019e-06, + 5.3942e-05, -9.0897e-06, 5.8152e-06, 3.0339e-05, -1.1808e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 214.83, cls_loss 0.0021 cls_loss_mapping 0.0052 cls_loss_causal 0.5175 re_mapping 0.0059 re_causal 0.0167 /// teacc 99.04 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1005, -0.1177, -0.0915, ..., -0.1341, -0.0092, 0.0761], + [ 0.0539, 0.0582, 0.0416, ..., 0.0969, -0.1336, -0.0517], + [ 0.0511, 0.0280, -0.1183, ..., -0.0244, -0.1760, -0.0699], + ..., + [-0.0452, -0.1006, 0.0838, ..., -0.1002, -0.0258, 0.0101], + [ 0.0200, 0.0869, 0.1287, ..., -0.1089, -0.0963, -0.1613], + [-0.1381, 0.0131, -0.2088, ..., -0.1011, -0.0894, 0.0828]], + device='cuda:0'), grad: tensor([[ 9.7882e-07, 1.0142e-06, 1.1045e-06, ..., 1.7416e-07, + 3.5856e-08, -8.0839e-07], + [-2.4378e-05, -1.5065e-05, -2.6196e-05, ..., -5.7817e-06, + 1.2061e-07, 8.4285e-08], + [ 9.7156e-06, 6.5193e-06, 1.0341e-05, ..., 2.1737e-06, + 4.2841e-08, 5.1688e-08], + ..., + [ 7.7039e-06, 4.6641e-06, 7.7188e-06, ..., 1.6997e-06, + 9.8720e-08, 3.9116e-08], + [-4.4778e-06, -5.8770e-05, -3.5465e-05, ..., 1.0924e-06, + 1.7462e-07, 1.7229e-07], + [ 3.2764e-06, 2.4792e-06, 1.9204e-06, ..., 9.5228e-07, + 1.1809e-06, 3.1665e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0040, -0.0184, -0.0247, 0.0333, -0.0297, 0.0295, -0.0239, -0.0096, + 0.0165, -0.0293], device='cuda:0'), grad: tensor([ 2.5295e-06, -4.8220e-05, 2.0713e-05, 3.3733e-06, -5.0068e-06, + 4.2439e-05, 1.2553e-04, 1.6019e-05, -1.6999e-04, 1.2688e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 214.92, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5225 re_mapping 0.0056 re_causal 0.0165 /// teacc 99.08 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1012, -0.1177, -0.0919, ..., -0.1340, -0.0087, 0.0768], + [ 0.0550, 0.0583, 0.0418, ..., 0.0969, -0.1344, -0.0520], + [ 0.0496, 0.0268, -0.1191, ..., -0.0252, -0.1763, -0.0702], + ..., + [-0.0455, -0.1019, 0.0838, ..., -0.1003, -0.0262, 0.0090], + [ 0.0209, 0.0884, 0.1295, ..., -0.1090, -0.0967, -0.1597], + [-0.1384, 0.0130, -0.2100, ..., -0.1014, -0.0897, 0.0831]], + device='cuda:0'), grad: tensor([[ 3.9972e-06, 2.5090e-06, 1.4622e-06, ..., 1.2526e-07, + 2.8592e-07, 1.4761e-06], + [ 7.6881e-07, 4.0187e-07, 4.0419e-07, ..., 1.0710e-08, + 2.9337e-08, 2.1141e-07], + [ 1.2621e-05, 7.3314e-06, 4.7460e-06, ..., 4.1910e-08, + 7.5437e-08, 9.0338e-08], + ..., + [ 8.6473e-07, 4.8475e-07, -1.2154e-07, ..., 3.7253e-09, + 6.5193e-09, -3.3947e-07], + [-3.8058e-05, -2.1324e-05, -1.3977e-05, ..., 9.5461e-08, + 1.8300e-07, 1.5553e-07], + [ 6.2101e-06, 3.2745e-06, 2.4680e-06, ..., 8.3819e-09, + -5.2620e-08, -2.0191e-06]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0048, -0.0179, -0.0262, 0.0334, -0.0302, 0.0295, -0.0242, -0.0100, + 0.0182, -0.0294], device='cuda:0'), grad: tensor([ 1.1817e-05, 2.2985e-06, 3.1233e-05, 1.7732e-05, 1.5534e-06, + 3.9935e-06, 9.1195e-06, 1.1791e-06, -9.1910e-05, 1.3024e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 214.83, cls_loss 0.0015 cls_loss_mapping 0.0036 cls_loss_causal 0.5219 re_mapping 0.0053 re_causal 0.0165 /// teacc 99.11 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1018, -0.1178, -0.0923, ..., -0.1342, -0.0087, 0.0770], + [ 0.0553, 0.0585, 0.0419, ..., 0.0973, -0.1350, -0.0519], + [ 0.0497, 0.0266, -0.1194, ..., -0.0256, -0.1764, -0.0702], + ..., + [-0.0458, -0.1025, 0.0838, ..., -0.1007, -0.0265, 0.0087], + [ 0.0214, 0.0883, 0.1301, ..., -0.1095, -0.0978, -0.1607], + [-0.1388, 0.0130, -0.2106, ..., -0.1018, -0.0902, 0.0832]], + device='cuda:0'), grad: tensor([[ 2.3749e-07, 6.1141e-07, 7.6368e-08, ..., 1.9139e-07, + 1.5832e-08, -1.1865e-06], + [-4.0699e-07, 1.4929e-06, 1.3243e-06, ..., 2.8685e-06, + 2.0284e-06, 6.0536e-08], + [-2.6748e-06, 1.0571e-07, 2.0489e-07, ..., 2.8592e-07, + 3.8929e-07, 1.4156e-07], + ..., + [ 3.6042e-07, 5.6718e-07, 5.5041e-07, ..., 8.8429e-07, + 5.2759e-07, 1.5460e-07], + [ 1.4324e-06, 1.3942e-06, 1.1204e-06, ..., 2.2724e-06, + 5.7966e-06, 3.4776e-06], + [ 2.6403e-07, 6.6450e-07, 5.1502e-07, ..., 7.6694e-07, + 9.3691e-07, 2.4354e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0049, -0.0178, -0.0261, 0.0333, -0.0301, 0.0297, -0.0241, -0.0101, + 0.0181, -0.0296], device='cuda:0'), grad: tensor([-3.5949e-07, 5.6773e-06, -3.0026e-06, 1.0103e-05, -1.1735e-05, + -3.2306e-05, 9.3952e-06, 2.4457e-06, 1.6242e-05, 3.4831e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 214.89, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.4814 re_mapping 0.0055 re_causal 0.0158 /// teacc 99.11 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1029, -0.1207, -0.0929, ..., -0.1369, -0.0114, 0.0748], + [ 0.0556, 0.0587, 0.0420, ..., 0.0980, -0.1353, -0.0517], + [ 0.0501, 0.0267, -0.1195, ..., -0.0261, -0.1770, -0.0700], + ..., + [-0.0460, -0.1029, 0.0838, ..., -0.1011, -0.0267, 0.0085], + [ 0.0207, 0.0880, 0.1301, ..., -0.1100, -0.0986, -0.1620], + [-0.1395, 0.0131, -0.2112, ..., -0.1022, -0.0905, 0.0835]], + device='cuda:0'), grad: tensor([[ 7.0296e-06, 9.7677e-06, 4.1816e-07, ..., -2.7008e-08, + -2.6030e-07, -7.8836e-07], + [-5.2527e-07, 4.1910e-09, -6.1700e-07, ..., -1.1548e-06, + 6.2399e-08, 2.0070e-07], + [ 4.6901e-06, 4.5784e-06, 4.4145e-06, ..., 1.3830e-07, + 2.5611e-08, 6.3330e-08], + ..., + [ 8.5169e-07, 4.9965e-07, 2.8359e-07, ..., 6.1654e-07, + 9.3132e-09, -3.8138e-07], + [-2.5779e-05, -8.6576e-06, -5.1558e-05, ..., 1.3644e-07, + 1.8766e-07, 3.2317e-07], + [ 3.2224e-07, 3.1572e-07, 8.9267e-07, ..., 5.7276e-08, + 3.3993e-08, -2.7474e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0022, -0.0176, -0.0258, 0.0335, -0.0303, 0.0298, -0.0218, -0.0103, + 0.0174, -0.0295], device='cuda:0'), grad: tensor([ 6.0380e-05, 2.0787e-06, 4.1246e-05, -1.5378e-04, 1.8878e-06, + 2.0432e-04, 1.0747e-04, 1.7975e-06, -2.6751e-04, 2.5257e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 214.86, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5030 re_mapping 0.0052 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1031, -0.1212, -0.0918, ..., -0.1372, -0.0113, 0.0751], + [ 0.0561, 0.0593, 0.0424, ..., 0.0985, -0.1361, -0.0515], + [ 0.0503, 0.0263, -0.1195, ..., -0.0265, -0.1773, -0.0701], + ..., + [-0.0465, -0.1037, 0.0837, ..., -0.1017, -0.0271, 0.0083], + [ 0.0201, 0.0881, 0.1302, ..., -0.1104, -0.0993, -0.1638], + [-0.1388, 0.0133, -0.2121, ..., -0.1026, -0.0908, 0.0839]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 2.7195e-07, 8.8476e-09, ..., 2.0489e-07, + 2.4494e-07, -6.1002e-08], + [-3.8482e-06, -2.1746e-07, -2.1756e-06, ..., -8.5542e-07, + 5.9605e-08, 6.7987e-08], + [ 8.1072e-07, 1.6531e-07, 5.1595e-07, ..., 2.5285e-07, + 7.2643e-08, 1.7695e-08], + ..., + [ 2.3805e-06, 2.1234e-07, 1.2182e-06, ..., 5.4855e-07, + 1.6810e-07, 9.9763e-06], + [ 3.3062e-08, 1.0999e-06, 2.3283e-09, ..., 9.7509e-07, + 1.1418e-06, 5.0291e-08], + [ 3.3062e-08, -4.1444e-08, 8.2888e-08, ..., 1.0710e-08, + -1.5786e-07, -1.0401e-05]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0024, -0.0173, -0.0258, 0.0337, -0.0301, 0.0296, -0.0221, -0.0105, + 0.0170, -0.0294], device='cuda:0'), grad: tensor([ 5.0105e-07, -5.2154e-06, 1.3588e-06, 8.4518e-07, 5.0385e-07, + 1.4696e-06, -4.8652e-06, 1.5467e-05, 2.5313e-06, -1.2599e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 214.92, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4855 re_mapping 0.0053 re_causal 0.0162 /// teacc 99.11 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1041, -0.1213, -0.0915, ..., -0.1372, -0.0113, 0.0752], + [ 0.0562, 0.0593, 0.0424, ..., 0.0986, -0.1367, -0.0516], + [ 0.0505, 0.0261, -0.1196, ..., -0.0267, -0.1775, -0.0703], + ..., + [-0.0469, -0.1041, 0.0834, ..., -0.1020, -0.0278, 0.0080], + [ 0.0207, 0.0882, 0.1318, ..., -0.1105, -0.0997, -0.1642], + [-0.1382, 0.0134, -0.2125, ..., -0.1030, -0.0911, 0.0842]], + device='cuda:0'), grad: tensor([[ 1.5530e-07, 2.0396e-07, 1.8138e-07, ..., 8.2189e-08, + -1.1027e-06, -3.7085e-06], + [ 4.4424e-07, 1.5367e-08, 4.8429e-07, ..., -3.0966e-08, + 1.1828e-07, 5.0524e-08], + [-4.4219e-06, 1.1199e-07, -2.5034e-06, ..., 1.0524e-07, + 8.6846e-08, 3.5856e-08], + ..., + [ 2.9057e-06, 3.7020e-08, -4.7917e-07, ..., 6.0303e-08, + 2.6776e-08, 2.3982e-08], + [ 2.3562e-07, 9.0199e-07, 5.4436e-07, ..., 6.4261e-07, + 9.1363e-07, 2.4354e-07], + [-2.4005e-07, 2.2026e-07, 2.0070e-07, ..., 6.7102e-07, + 6.3702e-07, -5.8906e-07]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0024, -0.0173, -0.0257, 0.0335, -0.0300, 0.0301, -0.0223, -0.0110, + 0.0175, -0.0293], device='cuda:0'), grad: tensor([-6.1058e-06, 1.9064e-06, -8.4266e-06, 7.1377e-06, -2.9653e-06, + 1.2834e-06, 1.3988e-06, 5.8068e-07, 3.3583e-06, 1.7807e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 215.13, cls_loss 0.0015 cls_loss_mapping 0.0039 cls_loss_causal 0.5109 re_mapping 0.0055 re_causal 0.0170 /// teacc 99.00 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1050, -0.1214, -0.0921, ..., -0.1372, -0.0113, 0.0753], + [ 0.0565, 0.0590, 0.0425, ..., 0.0993, -0.1372, -0.0515], + [ 0.0511, 0.0258, -0.1201, ..., -0.0270, -0.1779, -0.0704], + ..., + [-0.0470, -0.1045, 0.0835, ..., -0.1022, -0.0278, 0.0079], + [ 0.0206, 0.0888, 0.1326, ..., -0.1110, -0.1004, -0.1653], + [-0.1387, 0.0133, -0.2133, ..., -0.1034, -0.0913, 0.0843]], + device='cuda:0'), grad: tensor([[ 3.8184e-08, 2.1420e-08, 3.6787e-08, ..., 3.8184e-08, + 1.8161e-08, -8.8476e-09], + [-1.5199e-06, -7.9302e-07, -1.7304e-06, ..., -1.4743e-06, + 4.5635e-08, -1.0710e-08], + [ 6.6962e-07, 2.4354e-07, 9.2108e-07, ..., 4.1258e-07, + 2.7940e-08, 1.1176e-08], + ..., + [-2.9281e-06, 1.7090e-07, -8.8960e-06, ..., 3.0547e-07, + 8.3353e-08, 8.0559e-08], + [ 1.0170e-06, 2.9895e-07, 1.0626e-06, ..., 5.1409e-07, + 2.0163e-07, 4.1956e-07], + [ 2.0256e-07, 1.6810e-07, 3.3714e-07, ..., 4.8196e-07, + 3.6089e-07, -3.3993e-08]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0025, -0.0172, -0.0258, 0.0332, -0.0298, 0.0306, -0.0224, -0.0108, + 0.0172, -0.0296], device='cuda:0'), grad: tensor([ 1.1316e-07, -3.0901e-06, 1.9297e-06, 1.1377e-05, -1.8058e-06, + -5.3924e-07, 1.0133e-06, -1.4022e-05, 3.2205e-06, 1.8179e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 214.90, cls_loss 0.0016 cls_loss_mapping 0.0032 cls_loss_causal 0.5258 re_mapping 0.0053 re_causal 0.0162 /// teacc 99.09 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1068, -0.1217, -0.0922, ..., -0.1386, -0.0115, 0.0754], + [ 0.0574, 0.0598, 0.0428, ..., 0.1001, -0.1377, -0.0520], + [ 0.0511, 0.0257, -0.1207, ..., -0.0273, -0.1783, -0.0705], + ..., + [-0.0473, -0.1051, 0.0836, ..., -0.1027, -0.0281, 0.0076], + [ 0.0201, 0.0887, 0.1328, ..., -0.1117, -0.1008, -0.1661], + [-0.1393, 0.0130, -0.2152, ..., -0.1049, -0.0919, 0.0845]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 4.6566e-09, 5.1223e-09, ..., -6.0536e-09, + -8.8476e-09, -8.4750e-08], + [-8.2608e-07, -2.3888e-07, -9.7789e-07, ..., -1.2498e-06, + 1.3970e-08, -1.0710e-08], + [-3.8603e-07, 1.3970e-09, 9.2667e-08, ..., 9.8255e-08, + 6.9849e-09, 1.1642e-08], + ..., + [ 3.2550e-07, 9.2667e-08, 4.2841e-08, ..., 3.4459e-07, + 1.1176e-08, 9.3132e-09], + [ 1.8161e-08, 7.9162e-09, 1.7695e-08, ..., 9.1270e-08, + 1.1083e-07, 1.3923e-07], + [ 5.0291e-08, 1.1642e-07, 7.5903e-08, ..., 2.0489e-07, + 6.7987e-08, 1.4901e-08]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0023, -0.0169, -0.0259, 0.0334, -0.0292, 0.0305, -0.0224, -0.0109, + 0.0169, -0.0299], device='cuda:0'), grad: tensor([-8.6613e-08, -1.7369e-06, -3.5437e-07, 4.9733e-07, 6.4448e-07, + -3.8184e-07, 1.3784e-07, 4.1723e-07, 3.6415e-07, 4.9593e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 215.22, cls_loss 0.0018 cls_loss_mapping 0.0042 cls_loss_causal 0.5323 re_mapping 0.0057 re_causal 0.0171 /// teacc 98.86 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1075, -0.1218, -0.0924, ..., -0.1386, -0.0114, 0.0755], + [ 0.0578, 0.0597, 0.0430, ..., 0.1003, -0.1391, -0.0522], + [ 0.0510, 0.0255, -0.1209, ..., -0.0277, -0.1785, -0.0709], + ..., + [-0.0476, -0.1053, 0.0836, ..., -0.1030, -0.0280, 0.0074], + [ 0.0206, 0.0890, 0.1333, ..., -0.1121, -0.1019, -0.1662], + [-0.1408, 0.0125, -0.2164, ..., -0.1073, -0.0928, 0.0841]], + device='cuda:0'), grad: tensor([[ 3.5577e-06, 1.1967e-07, 1.1967e-07, ..., 6.6590e-08, + 2.2352e-08, 2.0135e-06], + [ 1.2154e-06, 1.3039e-08, -5.0059e-07, ..., -6.8406e-07, + 3.7253e-08, 8.4285e-07], + [ 6.9094e-04, 1.6555e-05, 5.6997e-06, ..., 8.0327e-07, + 1.9558e-08, 4.1294e-04], + ..., + [ 7.5847e-06, 2.5518e-07, 2.6124e-07, ..., 2.4121e-07, + 8.0559e-08, 4.5747e-06], + [ 2.5854e-05, -2.0191e-05, -7.5661e-06, ..., -7.2550e-07, + 1.7760e-06, 2.5049e-05], + [ 1.0729e-06, 2.2072e-07, 1.2759e-07, ..., 2.6962e-07, + 1.5786e-07, -2.1886e-07]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0023, -0.0168, -0.0262, 0.0338, -0.0284, 0.0303, -0.0223, -0.0110, + 0.0172, -0.0308], device='cuda:0'), grad: tensor([ 1.4827e-05, 6.1393e-06, 2.8305e-03, 3.7694e-04, 2.6301e-06, + -3.4370e-03, 2.2218e-05, 3.2008e-05, 1.4913e-04, 3.6526e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 214.79, cls_loss 0.0016 cls_loss_mapping 0.0037 cls_loss_causal 0.5242 re_mapping 0.0056 re_causal 0.0168 /// teacc 98.90 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1081, -0.1218, -0.0927, ..., -0.1387, -0.0114, 0.0755], + [ 0.0573, 0.0570, 0.0429, ..., 0.0981, -0.1398, -0.0556], + [ 0.0503, 0.0255, -0.1212, ..., -0.0280, -0.1791, -0.0727], + ..., + [-0.0478, -0.1057, 0.0837, ..., -0.1033, -0.0287, 0.0080], + [ 0.0206, 0.0888, 0.1336, ..., -0.1128, -0.1048, -0.1676], + [-0.1388, 0.0141, -0.2167, ..., -0.1053, -0.0933, 0.0862]], + device='cuda:0'), grad: tensor([[ 7.7300e-08, -1.1027e-05, 6.2864e-08, ..., 1.1176e-08, + -1.9565e-05, -5.8621e-05], + [-6.0070e-07, -2.1886e-07, -6.1933e-07, ..., -5.4017e-07, + 4.2375e-08, -2.2352e-08], + [ 1.4855e-07, 3.3295e-07, 1.8021e-07, ..., 6.1002e-08, + 4.4936e-07, 1.4259e-06], + ..., + [ 2.8312e-07, 2.8871e-07, -3.1898e-07, ..., 1.2852e-07, + 1.7276e-07, 7.3947e-07], + [-2.0918e-06, 5.5088e-07, -1.4976e-06, ..., 2.3702e-07, + 2.7940e-06, 7.7635e-06], + [ 5.0897e-07, 9.1270e-07, 6.8266e-07, ..., 2.9802e-08, + 2.5872e-06, 6.0759e-06]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0023, -0.0187, -0.0277, 0.0331, -0.0283, 0.0325, -0.0224, -0.0109, + 0.0159, -0.0291], device='cuda:0'), grad: tensor([-8.7976e-05, -9.6671e-07, 2.4959e-06, 8.4490e-06, 3.1590e-06, + -3.8855e-06, 5.9992e-05, 6.1840e-07, 8.0466e-06, 1.0036e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 215.01, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.5111 re_mapping 0.0054 re_causal 0.0163 /// teacc 99.06 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.1085, -0.1219, -0.0928, ..., -0.1389, -0.0116, 0.0756], + [ 0.0575, 0.0571, 0.0430, ..., 0.0982, -0.1402, -0.0557], + [ 0.0510, 0.0269, -0.1207, ..., -0.0275, -0.1795, -0.0729], + ..., + [-0.0480, -0.1060, 0.0839, ..., -0.1034, -0.0283, 0.0083], + [ 0.0201, 0.0881, 0.1334, ..., -0.1134, -0.1055, -0.1684], + [-0.1387, 0.0143, -0.2178, ..., -0.1054, -0.0930, 0.0866]], + device='cuda:0'), grad: tensor([[ 9.1782e-07, 4.8801e-07, 1.3970e-08, ..., 4.1910e-09, + 2.0862e-07, 1.9185e-07], + [-1.1083e-07, -3.2596e-08, -1.7742e-07, ..., -1.2899e-07, + 2.9337e-08, -8.8476e-09], + [-2.0657e-06, -6.8638e-07, 6.1002e-08, ..., 1.9092e-08, + 4.5635e-08, -6.4773e-07], + ..., + [ 2.4363e-06, 1.1316e-07, 6.4149e-06, ..., 7.2177e-08, + 8.3819e-09, 2.6878e-06], + [ 4.4191e-07, 4.8196e-07, -3.9348e-07, ..., 1.9092e-08, + 6.5146e-07, 4.3586e-07], + [ 2.2212e-07, 1.6624e-07, 1.9418e-07, ..., 4.3772e-08, + 4.9360e-08, 2.4214e-08]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0023, -0.0187, -0.0271, 0.0329, -0.0285, 0.0326, -0.0224, -0.0106, + 0.0151, -0.0289], device='cuda:0'), grad: tensor([ 3.0808e-06, -1.9372e-07, -5.6773e-06, -1.5184e-05, 2.4028e-07, + 1.4575e-06, -3.3602e-06, 1.5378e-05, 3.4086e-06, 8.1724e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 214.76, cls_loss 0.0016 cls_loss_mapping 0.0036 cls_loss_causal 0.5408 re_mapping 0.0056 re_causal 0.0168 /// teacc 98.84 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1095, -0.1217, -0.0925, ..., -0.1390, -0.0112, 0.0759], + [ 0.0578, 0.0572, 0.0433, ..., 0.0986, -0.1402, -0.0556], + [ 0.0504, 0.0267, -0.1200, ..., -0.0277, -0.1798, -0.0729], + ..., + [-0.0485, -0.1055, 0.0839, ..., -0.1037, -0.0289, 0.0085], + [ 0.0199, 0.0886, 0.1329, ..., -0.1136, -0.1059, -0.1690], + [-0.1396, 0.0141, -0.2202, ..., -0.1057, -0.0925, 0.0868]], + device='cuda:0'), grad: tensor([[-9.1363e-07, 5.4017e-07, -2.5611e-06, ..., 7.9162e-09, + -1.3737e-07, -9.5442e-06], + [-8.5086e-06, -4.5039e-06, -5.0031e-06, ..., -2.1476e-06, + 3.3295e-07, 4.7833e-06], + [ 1.9372e-07, 1.1148e-06, 1.9204e-06, ..., 1.3970e-07, + 2.6077e-08, 4.9500e-07], + ..., + [ 8.5607e-06, 4.3511e-06, 6.0983e-06, ..., 1.7481e-06, + 8.7544e-08, 2.6394e-06], + [-2.4755e-06, -1.8803e-06, -2.7660e-06, ..., 5.6345e-08, + 7.3435e-07, 1.0105e-06], + [ 1.1455e-06, -4.8801e-06, 9.9652e-07, ..., 8.9407e-08, + -1.0468e-06, -1.0453e-05]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0026, -0.0185, -0.0271, 0.0340, -0.0291, 0.0325, -0.0228, -0.0110, + 0.0150, -0.0289], device='cuda:0'), grad: tensor([-1.2413e-05, -1.0870e-05, 3.1926e-06, 1.1697e-06, 2.2963e-05, + 8.2329e-06, -8.1211e-06, 2.0295e-05, -6.8545e-06, -1.7643e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.77, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5221 re_mapping 0.0056 re_causal 0.0165 /// teacc 99.04 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1101, -0.1218, -0.0927, ..., -0.1390, -0.0112, 0.0760], + [ 0.0579, 0.0571, 0.0433, ..., 0.0986, -0.1413, -0.0557], + [ 0.0505, 0.0265, -0.1202, ..., -0.0283, -0.1803, -0.0731], + ..., + [-0.0487, -0.1056, 0.0841, ..., -0.1040, -0.0293, 0.0087], + [ 0.0198, 0.0883, 0.1328, ..., -0.1137, -0.1066, -0.1702], + [-0.1394, 0.0143, -0.2213, ..., -0.1060, -0.0926, 0.0871]], + device='cuda:0'), grad: tensor([[ 6.6590e-08, 1.0245e-07, 8.4285e-08, ..., 2.3749e-08, + 2.8163e-06, 2.4021e-05], + [-1.4575e-07, -8.9873e-08, -4.7451e-07, ..., -4.3493e-07, + 7.6368e-08, 2.3423e-07], + [ 8.3819e-08, 1.3877e-07, 2.1746e-07, ..., 8.1956e-08, + 9.6858e-08, 7.2410e-07], + ..., + [ 1.4482e-07, 2.7986e-07, 8.4750e-08, ..., 1.3644e-07, + 6.4261e-08, 7.9023e-07], + [-2.4168e-07, 5.9744e-07, -4.5076e-07, ..., 7.9162e-08, + 5.8534e-07, 1.9185e-06], + [ 1.1036e-07, -1.4408e-06, 9.3598e-08, ..., 1.8161e-08, + -3.1758e-06, -3.0324e-05]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0026, -0.0186, -0.0271, 0.0342, -0.0290, 0.0324, -0.0228, -0.0109, + 0.0145, -0.0288], device='cuda:0'), grad: tensor([ 3.7193e-05, -8.3819e-08, 1.4501e-06, -7.6182e-07, 4.8429e-06, + -1.5516e-06, 7.4459e-07, 1.5870e-06, 3.8892e-06, -4.7326e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 215.26, cls_loss 0.0017 cls_loss_mapping 0.0036 cls_loss_causal 0.5500 re_mapping 0.0052 re_causal 0.0165 /// teacc 99.09 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1109, -0.1220, -0.0930, ..., -0.1391, -0.0111, 0.0763], + [ 0.0577, 0.0571, 0.0433, ..., 0.0986, -0.1422, -0.0559], + [ 0.0517, 0.0266, -0.1207, ..., -0.0283, -0.1810, -0.0731], + ..., + [-0.0488, -0.1060, 0.0845, ..., -0.1045, -0.0309, 0.0091], + [ 0.0199, 0.0887, 0.1330, ..., -0.1140, -0.1072, -0.1711], + [-0.1415, 0.0140, -0.2255, ..., -0.1068, -0.0923, 0.0870]], + device='cuda:0'), grad: tensor([[ 1.7202e-06, 1.2526e-06, 8.1491e-08, ..., 9.0618e-07, + -3.7532e-07, -3.0315e-07], + [ 1.6963e-04, 1.2493e-04, -1.2163e-06, ..., 9.0122e-05, + 3.2097e-05, 1.2541e-04], + [-1.9264e-04, -1.4162e-04, 7.1106e-07, ..., -1.0234e-04, + -3.6329e-05, -1.4198e-04], + ..., + [ 8.0001e-07, 5.2294e-07, 4.6566e-10, ..., 4.5169e-07, + 1.1083e-07, 1.6252e-07], + [ 9.8441e-07, 4.0187e-07, -1.3066e-06, ..., 6.1886e-07, + 3.0315e-07, 1.2927e-06], + [ 1.7444e-06, 1.2880e-06, 2.8824e-07, ..., 9.4762e-07, + 3.7486e-07, 1.4240e-06]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0027, -0.0189, -0.0265, 0.0343, -0.0281, 0.0323, -0.0230, -0.0105, + 0.0144, -0.0297], device='cuda:0'), grad: tensor([ 6.1393e-06, 8.2302e-04, -9.3269e-04, 8.7395e-06, 3.6448e-05, + 2.1473e-05, 1.9789e-05, 2.7139e-06, 5.5991e-06, 8.8215e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 215.06, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.4920 re_mapping 0.0051 re_causal 0.0158 /// teacc 99.08 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1115, -0.1222, -0.0937, ..., -0.1392, -0.0112, 0.0762], + [ 0.0578, 0.0573, 0.0443, ..., 0.0986, -0.1417, -0.0550], + [ 0.0523, 0.0271, -0.1212, ..., -0.0278, -0.1807, -0.0724], + ..., + [-0.0492, -0.1076, 0.0839, ..., -0.1048, -0.0330, 0.0079], + [ 0.0203, 0.0886, 0.1334, ..., -0.1141, -0.1080, -0.1723], + [-0.1421, 0.0143, -0.2266, ..., -0.1071, -0.0921, 0.0876]], + device='cuda:0'), grad: tensor([[ 1.4761e-07, 8.3353e-08, 9.2201e-08, ..., 4.6100e-08, + -2.6543e-08, -4.2049e-07], + [-2.2352e-08, 2.9802e-07, 7.2177e-08, ..., 1.4482e-07, + 2.0210e-07, 6.0070e-08], + [-7.0315e-08, 7.4506e-08, 5.4948e-08, ..., 7.4506e-08, + 4.2375e-08, 2.0955e-08], + ..., + [ 6.2818e-07, 5.6298e-07, 1.2703e-06, ..., 2.0768e-07, + 8.8476e-08, 2.0489e-08], + [-3.4384e-06, -1.0328e-06, -8.1956e-06, ..., 1.1222e-06, + 7.6508e-07, 3.8277e-07], + [ 6.4261e-08, 1.8813e-06, 9.5926e-08, ..., 3.9935e-06, + 2.7958e-06, -1.9651e-06]], device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0027, -0.0179, -0.0259, 0.0343, -0.0288, 0.0336, -0.0238, -0.0117, + 0.0144, -0.0294], device='cuda:0'), grad: tensor([ 4.3772e-08, 1.0710e-06, 1.0338e-07, 4.8429e-08, -1.1668e-05, + 1.0073e-05, 1.3206e-06, 2.5686e-06, -9.3728e-06, 5.8375e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 214.86, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.5239 re_mapping 0.0054 re_causal 0.0168 /// teacc 99.02 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1123, -0.1225, -0.0941, ..., -0.1393, -0.0112, 0.0762], + [ 0.0580, 0.0576, 0.0445, ..., 0.0987, -0.1423, -0.0548], + [ 0.0520, 0.0268, -0.1218, ..., -0.0281, -0.1811, -0.0726], + ..., + [-0.0494, -0.1081, 0.0839, ..., -0.1051, -0.0333, 0.0074], + [ 0.0209, 0.0888, 0.1339, ..., -0.1145, -0.1084, -0.1733], + [-0.1422, 0.0143, -0.2270, ..., -0.1075, -0.0924, 0.0881]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 2.2817e-08, 1.2573e-08, ..., 1.8626e-08, + -8.1491e-08, -2.6636e-07], + [ 1.2852e-07, 3.1013e-07, 2.2817e-07, ..., 4.7404e-07, + 4.5262e-07, 3.6787e-08], + [ 6.4448e-07, 5.3551e-08, 3.9581e-08, ..., 3.5390e-08, + 3.1199e-08, 1.3970e-08], + ..., + [ 3.5483e-07, 1.3551e-07, -4.1677e-07, ..., 1.6205e-07, + 1.3970e-07, 4.0978e-08], + [ 2.0536e-07, 2.9197e-07, 2.8126e-07, ..., 5.7742e-07, + 5.4203e-07, 1.0477e-07], + [ 4.1910e-08, 3.2764e-06, 1.4426e-06, ..., 5.5619e-06, + 5.0627e-06, -1.2107e-07]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0026, -0.0177, -0.0261, 0.0343, -0.0288, 0.0335, -0.0238, -0.0120, + 0.0145, -0.0293], device='cuda:0'), grad: tensor([-2.6310e-07, 1.6531e-06, 1.6615e-06, -5.7667e-06, -1.6153e-05, + 2.7735e-06, 6.8592e-07, 6.2492e-07, 2.2240e-06, 1.2554e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 214.72, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.4986 re_mapping 0.0056 re_causal 0.0161 /// teacc 99.06 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1127, -0.1225, -0.0945, ..., -0.1393, -0.0108, 0.0767], + [ 0.0583, 0.0576, 0.0447, ..., 0.0988, -0.1429, -0.0549], + [ 0.0519, 0.0266, -0.1225, ..., -0.0287, -0.1826, -0.0731], + ..., + [-0.0495, -0.1084, 0.0840, ..., -0.1052, -0.0332, 0.0073], + [ 0.0210, 0.0890, 0.1342, ..., -0.1148, -0.1091, -0.1745], + [-0.1423, 0.0140, -0.2277, ..., -0.1081, -0.0934, 0.0881]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, 3.9116e-08, 3.8650e-08, ..., 2.7008e-08, + 2.6543e-08, -3.7719e-08], + [ 1.6699e-06, 1.1921e-07, 6.0201e-06, ..., 2.2678e-07, + 1.7229e-07, 2.1905e-06], + [ 2.6450e-07, 1.2992e-07, 5.4250e-07, ..., 1.2619e-07, + 8.4750e-08, 7.7300e-08], + ..., + [-3.8035e-06, 8.8476e-09, -1.4648e-05, ..., 5.5879e-09, + 2.6543e-08, 1.6717e-06], + [ 1.1120e-06, -1.0151e-07, 4.2357e-06, ..., 4.6566e-08, + 1.6857e-07, 3.2224e-07], + [ 6.5798e-07, 1.0710e-07, 1.7984e-06, ..., 1.3737e-07, + 1.0617e-07, -7.0259e-06]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0029, -0.0176, -0.0265, 0.0347, -0.0281, 0.0346, -0.0249, -0.0120, + 0.0142, -0.0297], device='cuda:0'), grad: tensor([ 2.9569e-07, 1.4260e-05, 1.6149e-06, 2.0042e-06, 4.5635e-06, + -2.2091e-06, 5.6904e-07, -1.9833e-05, 7.6294e-06, -8.9332e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 215.26, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5050 re_mapping 0.0054 re_causal 0.0162 /// teacc 99.11 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1105, -0.1227, -0.0923, ..., -0.1395, -0.0105, 0.0771], + [ 0.0582, 0.0574, 0.0445, ..., 0.0985, -0.1439, -0.0551], + [ 0.0518, 0.0259, -0.1232, ..., -0.0303, -0.1839, -0.0733], + ..., + [-0.0494, -0.1088, 0.0844, ..., -0.1052, -0.0331, 0.0077], + [ 0.0212, 0.0889, 0.1343, ..., -0.1150, -0.1098, -0.1753], + [-0.1426, 0.0137, -0.2286, ..., -0.1088, -0.0946, 0.0882]], + device='cuda:0'), grad: tensor([[ 2.2538e-07, 4.7963e-08, 7.3109e-08, ..., 2.3283e-08, + -9.8720e-08, -3.0780e-07], + [-1.4126e-05, -3.5852e-05, -7.3612e-05, ..., -1.4193e-05, + 2.1420e-08, -1.9461e-05], + [-7.6462e-07, 9.9186e-08, 1.5553e-07, ..., 5.6811e-08, + 3.4925e-08, 2.3749e-08], + ..., + [ 9.9093e-06, 2.5079e-05, 5.1439e-05, ..., 9.9316e-06, + 2.5611e-08, 1.3575e-05], + [ 7.0315e-08, 6.8266e-07, 1.4743e-06, ..., 4.2981e-07, + 3.4925e-08, 6.0303e-07], + [ 3.0845e-06, 7.7933e-06, 1.5974e-05, ..., 3.1516e-06, + 1.0990e-07, 4.2394e-06]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0034, -0.0181, -0.0271, 0.0347, -0.0272, 0.0341, -0.0246, -0.0115, + 0.0140, -0.0304], device='cuda:0'), grad: tensor([ 3.0501e-07, -1.1569e-04, -2.3972e-06, 2.2668e-06, 4.9211e-06, + 1.3206e-06, 5.8161e-07, 8.1062e-05, 2.5164e-06, 2.5392e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 215.09, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5235 re_mapping 0.0055 re_causal 0.0165 /// teacc 99.10 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1111, -0.1229, -0.0927, ..., -0.1396, -0.0104, 0.0773], + [ 0.0578, 0.0582, 0.0439, ..., 0.0988, -0.1445, -0.0552], + [ 0.0532, 0.0257, -0.1235, ..., -0.0306, -0.1842, -0.0723], + ..., + [-0.0484, -0.1097, 0.0855, ..., -0.1056, -0.0332, 0.0079], + [ 0.0200, 0.0880, 0.1334, ..., -0.1158, -0.1105, -0.1763], + [-0.1449, 0.0127, -0.2293, ..., -0.1107, -0.0957, 0.0876]], + device='cuda:0'), grad: tensor([[ 3.7253e-08, 5.4017e-08, 9.3598e-08, ..., 4.8894e-08, + 2.5146e-08, 1.6857e-07], + [ 5.5414e-08, 1.2061e-07, 1.3346e-06, ..., 1.4156e-07, + 5.2620e-08, 2.5295e-06], + [ 2.8964e-07, 6.4727e-08, 1.3132e-07, ..., 4.7963e-08, + 1.8626e-08, 1.3970e-07], + ..., + [ 3.1432e-07, 1.2433e-07, -6.3032e-06, ..., 7.2177e-08, + 3.1199e-08, -1.2219e-05], + [-2.5984e-07, -3.7253e-08, -1.8440e-07, ..., 1.1222e-07, + 5.2154e-08, 2.5984e-07], + [ 4.7032e-08, 2.4959e-06, 2.2482e-06, ..., 3.3807e-06, + 1.1697e-06, 3.8221e-06]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0035, -0.0185, -0.0261, 0.0349, -0.0255, 0.0338, -0.0246, -0.0106, + 0.0126, -0.0320], device='cuda:0'), grad: tensor([ 5.2666e-07, 5.4464e-06, 9.8441e-07, 2.8051e-06, -9.7603e-06, + 5.5656e-06, 1.3653e-06, -2.3067e-05, 1.7649e-07, 1.5959e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 215.10, cls_loss 0.0015 cls_loss_mapping 0.0027 cls_loss_causal 0.4936 re_mapping 0.0053 re_causal 0.0161 /// teacc 99.05 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1119, -0.1230, -0.0929, ..., -0.1397, -0.0104, 0.0773], + [ 0.0568, 0.0572, 0.0434, ..., 0.0988, -0.1457, -0.0566], + [ 0.0541, 0.0269, -0.1241, ..., -0.0307, -0.1835, -0.0721], + ..., + [-0.0478, -0.1102, 0.0864, ..., -0.1058, -0.0337, 0.0087], + [ 0.0203, 0.0882, 0.1336, ..., -0.1160, -0.1109, -0.1771], + [-0.1450, 0.0128, -0.2305, ..., -0.1111, -0.0961, 0.0881]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 2.6077e-08, 2.0023e-08, ..., 2.8405e-08, + -5.0664e-07, -1.5358e-06], + [-6.7847e-07, -4.6100e-08, -4.3446e-07, ..., -1.3877e-07, + 6.1002e-08, -1.8114e-07], + [-2.0489e-08, 4.8429e-08, 9.3598e-08, ..., 8.2888e-08, + 1.9046e-07, 5.1502e-07], + ..., + [ 4.9733e-07, 2.9011e-07, 3.9814e-07, ..., 3.0268e-07, + 1.5693e-07, 1.3364e-07], + [-6.5193e-09, 1.3318e-07, 8.7544e-08, ..., 1.7928e-07, + 1.6438e-07, 9.7323e-08], + [ 5.8673e-08, 7.0920e-07, 4.2981e-07, ..., 8.7544e-07, + 7.9395e-07, 7.7998e-07]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0034, -0.0196, -0.0255, 0.0357, -0.0252, 0.0328, -0.0243, -0.0096, + 0.0124, -0.0320], device='cuda:0'), grad: tensor([-3.7290e-06, -7.5717e-07, 1.2051e-06, 7.2410e-07, -3.7439e-06, + -5.2527e-07, 4.4657e-07, 1.4538e-06, 7.0082e-07, 4.2170e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.90, cls_loss 0.0012 cls_loss_mapping 0.0027 cls_loss_causal 0.4848 re_mapping 0.0055 re_causal 0.0159 /// teacc 99.04 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1122, -0.1235, -0.0932, ..., -0.1402, -0.0106, 0.0774], + [ 0.0570, 0.0573, 0.0436, ..., 0.0989, -0.1467, -0.0567], + [ 0.0544, 0.0267, -0.1240, ..., -0.0309, -0.1840, -0.0722], + ..., + [-0.0482, -0.1108, 0.0863, ..., -0.1063, -0.0341, 0.0086], + [ 0.0200, 0.0882, 0.1336, ..., -0.1168, -0.1119, -0.1776], + [-0.1440, 0.0128, -0.2307, ..., -0.1116, -0.0964, 0.0884]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 5.4017e-08, 8.3819e-09, ..., 1.7602e-07, + 1.0105e-07, -9.1717e-06], + [-7.3574e-08, 1.9977e-07, -1.4435e-08, ..., 2.8405e-08, + 6.5193e-08, 2.1048e-07], + [ 1.9325e-07, 1.0477e-07, 1.2014e-07, ..., 7.3109e-08, + 3.2131e-08, 1.1353e-06], + ..., + [ 5.0291e-08, 3.4226e-07, 5.4482e-08, ..., 1.9092e-08, + 1.4901e-08, 6.2492e-07], + [-6.2259e-07, 5.0943e-07, -2.4354e-07, ..., 3.0454e-07, + 7.8790e-07, 1.6559e-06], + [ 1.2573e-08, -3.8743e-06, -9.5833e-07, ..., 1.1036e-07, + 1.1083e-07, -4.4927e-06]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0034, -0.0196, -0.0254, 0.0355, -0.0248, 0.0324, -0.0240, -0.0098, + 0.0120, -0.0318], device='cuda:0'), grad: tensor([-1.6749e-05, 5.1642e-07, 2.6599e-06, 2.4401e-06, 1.7405e-05, + -3.9302e-06, 1.3076e-06, 1.2945e-06, 3.4906e-06, -8.4788e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 214.94, cls_loss 0.0014 cls_loss_mapping 0.0036 cls_loss_causal 0.4930 re_mapping 0.0055 re_causal 0.0155 /// teacc 99.02 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1123, -0.1236, -0.0934, ..., -0.1403, -0.0106, 0.0777], + [ 0.0570, 0.0574, 0.0436, ..., 0.0995, -0.1470, -0.0568], + [ 0.0544, 0.0264, -0.1246, ..., -0.0315, -0.1842, -0.0724], + ..., + [-0.0480, -0.1112, 0.0865, ..., -0.1068, -0.0344, 0.0090], + [ 0.0206, 0.0890, 0.1344, ..., -0.1170, -0.1127, -0.1792], + [-0.1436, 0.0128, -0.2312, ..., -0.1119, -0.0964, 0.0886]], + device='cuda:0'), grad: tensor([[ 1.1781e-07, 1.2713e-07, 7.2643e-08, ..., -4.1910e-09, + -5.6345e-08, -3.3015e-07], + [-1.6531e-07, 3.9581e-08, -3.0873e-07, ..., -3.8929e-07, + 1.0245e-08, 7.4506e-08], + [-5.1223e-09, 3.3434e-07, 2.1514e-07, ..., 3.3062e-08, + 5.5879e-09, 1.1642e-08], + ..., + [ 4.0326e-07, 1.2945e-07, -4.2375e-08, ..., 2.6543e-07, + 9.7789e-09, -4.7823e-07], + [-1.7574e-06, -1.8859e-06, -1.2126e-06, ..., 3.1665e-08, + 2.5611e-08, 2.6543e-08], + [ 1.9185e-07, 3.3574e-07, 4.3120e-07, ..., 5.4017e-07, + 1.2061e-07, 3.8045e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0035, -0.0196, -0.0256, 0.0348, -0.0250, 0.0327, -0.0242, -0.0095, + 0.0122, -0.0318], device='cuda:0'), grad: tensor([-3.5856e-08, -4.3772e-08, 3.8091e-07, 3.5651e-06, -8.9267e-07, + -1.2163e-06, 1.7043e-06, 8.0094e-08, -5.9195e-06, 2.3618e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.94, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5382 re_mapping 0.0053 re_causal 0.0159 /// teacc 99.04 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1134, -0.1238, -0.0941, ..., -0.1404, -0.0107, 0.0771], + [ 0.0570, 0.0574, 0.0435, ..., 0.0995, -0.1484, -0.0572], + [ 0.0536, 0.0263, -0.1251, ..., -0.0317, -0.1844, -0.0728], + ..., + [-0.0479, -0.1118, 0.0867, ..., -0.1070, -0.0348, 0.0089], + [ 0.0216, 0.0904, 0.1356, ..., -0.1171, -0.1130, -0.1793], + [-0.1435, 0.0131, -0.2316, ..., -0.1121, -0.0967, 0.0901]], + device='cuda:0'), grad: tensor([[ 9.2667e-08, 1.5413e-07, 3.0734e-08, ..., 2.5611e-07, + 2.3516e-07, 2.2072e-07], + [-7.2643e-08, 8.9407e-08, -1.4715e-07, ..., 2.0117e-07, + 1.4575e-07, 6.6590e-08], + [-3.6322e-08, 8.8010e-08, 3.1199e-08, ..., 2.8312e-07, + 1.1548e-07, 2.6543e-08], + ..., + [ 9.3132e-08, 8.0559e-08, -1.7183e-07, ..., 7.9162e-09, + 5.9605e-08, 6.2399e-08], + [-1.8952e-07, 3.8184e-08, -2.9569e-07, ..., 1.5320e-07, + 1.1548e-07, 7.2643e-08], + [-7.4971e-08, 2.7847e-06, 5.4017e-08, ..., 3.8128e-06, + 9.1176e-07, 5.3737e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0031, -0.0200, -0.0261, 0.0348, -0.0260, 0.0332, -0.0242, -0.0092, + 0.0130, -0.0310], device='cuda:0'), grad: tensor([ 8.1910e-07, 4.3772e-07, 4.4657e-07, 4.3884e-06, -9.8944e-06, + -4.1090e-06, -8.8243e-07, -1.5227e-07, -3.0082e-07, 9.2387e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 214.80, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.5135 re_mapping 0.0050 re_causal 0.0155 /// teacc 99.09 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1139, -0.1241, -0.0942, ..., -0.1404, -0.0107, 0.0772], + [ 0.0569, 0.0575, 0.0437, ..., 0.0998, -0.1493, -0.0572], + [ 0.0535, 0.0263, -0.1260, ..., -0.0318, -0.1847, -0.0728], + ..., + [-0.0481, -0.1128, 0.0868, ..., -0.1079, -0.0351, 0.0088], + [ 0.0227, 0.0905, 0.1364, ..., -0.1173, -0.1133, -0.1800], + [-0.1439, 0.0130, -0.2319, ..., -0.1123, -0.0970, 0.0904]], + device='cuda:0'), grad: tensor([[ 1.3551e-07, 1.8533e-07, 7.8231e-08, ..., 2.3283e-08, + 1.2480e-07, -6.8545e-07], + [ 3.7998e-06, 8.4285e-08, 1.0282e-05, ..., 8.5775e-07, + 8.9407e-08, 8.5682e-08], + [ 2.3469e-07, 2.2352e-08, 2.0005e-06, ..., 1.7602e-07, + 2.7008e-08, 4.2375e-08], + ..., + [-6.4932e-06, 2.0023e-08, -1.8030e-05, ..., -1.4780e-06, + 1.7229e-08, -1.4901e-08], + [ 1.8878e-06, 3.5623e-07, 4.3511e-06, ..., 3.9535e-07, + 5.6578e-07, 2.9523e-07], + [ 1.1409e-07, 1.9837e-07, 1.8533e-07, ..., 5.0291e-08, + 3.5157e-07, 2.2491e-07]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0031, -0.0201, -0.0261, 0.0349, -0.0263, 0.0332, -0.0240, -0.0093, + 0.0135, -0.0311], device='cuda:0'), grad: tensor([-1.6438e-07, 2.0444e-05, 2.9746e-06, 6.1747e-07, 1.4221e-06, + 2.2314e-06, -4.2804e-06, -3.5018e-05, 1.0535e-05, 1.2498e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 214.61, cls_loss 0.0014 cls_loss_mapping 0.0036 cls_loss_causal 0.5177 re_mapping 0.0050 re_causal 0.0153 /// teacc 99.02 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1120, -0.1245, -0.0920, ..., -0.1413, -0.0111, 0.0773], + [ 0.0574, 0.0578, 0.0440, ..., 0.1004, -0.1499, -0.0571], + [ 0.0539, 0.0261, -0.1263, ..., -0.0321, -0.1850, -0.0729], + ..., + [-0.0487, -0.1133, 0.0866, ..., -0.1084, -0.0354, 0.0087], + [ 0.0225, 0.0913, 0.1369, ..., -0.1178, -0.1139, -0.1808], + [-0.1442, 0.0134, -0.2323, ..., -0.1124, -0.0973, 0.0909]], + device='cuda:0'), grad: tensor([[ 2.9448e-06, 3.3136e-06, 3.6322e-08, ..., 1.8161e-08, + -5.9605e-08, 1.7750e-04], + [-6.7102e-07, -1.9418e-07, -9.5647e-07, ..., -5.4901e-07, + 6.0536e-09, 1.4249e-07], + [-1.2882e-05, -1.4178e-05, 8.5682e-08, ..., 4.9360e-08, + 3.7253e-09, 1.8347e-06], + ..., + [ 6.4354e-07, 1.2619e-07, 6.2166e-07, ..., 3.4273e-07, + 3.7253e-09, 1.0934e-06], + [ 8.5384e-06, 9.4473e-06, -4.6566e-10, ..., 3.5856e-08, + 4.6566e-09, 1.5507e-07], + [ 1.0822e-06, 1.1036e-06, 7.2643e-08, ..., 9.3598e-08, + 2.8405e-08, -1.8370e-04]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0034, -0.0199, -0.0260, 0.0346, -0.0273, 0.0336, -0.0242, -0.0096, + 0.0135, -0.0306], device='cuda:0'), grad: tensor([ 2.4652e-04, -1.3728e-06, -6.9857e-05, -7.9768e-07, 3.1609e-06, + 1.5823e-06, 8.7218e-07, 3.2187e-06, 4.8250e-05, -2.3198e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 215.30, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5227 re_mapping 0.0051 re_causal 0.0153 /// teacc 98.99 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1128, -0.1248, -0.0924, ..., -0.1415, -0.0112, 0.0771], + [ 0.0574, 0.0581, 0.0440, ..., 0.1010, -0.1504, -0.0570], + [ 0.0543, 0.0263, -0.1270, ..., -0.0324, -0.1855, -0.0730], + ..., + [-0.0486, -0.1140, 0.0870, ..., -0.1088, -0.0346, 0.0090], + [ 0.0226, 0.0916, 0.1372, ..., -0.1179, -0.1141, -0.1816], + [-0.1446, 0.0136, -0.2333, ..., -0.1127, -0.0976, 0.0913]], + device='cuda:0'), grad: tensor([[ 2.8405e-07, 3.6322e-08, 1.2107e-07, ..., 3.8184e-08, + -1.5832e-08, -5.0757e-08], + [-9.3132e-09, -2.4075e-07, -6.3377e-07, ..., -4.8103e-07, + 1.9092e-08, 2.1746e-07], + [-3.5912e-06, -6.5705e-07, 1.2144e-06, ..., -1.0394e-06, + 3.8184e-08, 2.3516e-07], + ..., + [ 1.6019e-07, 1.6717e-07, -1.2470e-06, ..., 1.9930e-07, + -4.6566e-09, -6.6916e-07], + [ 1.7723e-06, 9.7044e-07, -8.1118e-07, ..., 1.1222e-06, + 3.1199e-08, 7.1013e-07], + [ 1.1735e-07, -4.0280e-07, 1.5553e-07, ..., 1.0757e-07, + 7.7765e-08, -7.7114e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0032, -0.0198, -0.0257, 0.0347, -0.0275, 0.0333, -0.0242, -0.0091, + 0.0134, -0.0306], device='cuda:0'), grad: tensor([ 5.5879e-07, 3.3993e-07, -1.0811e-05, 1.9204e-06, 8.6380e-07, + 9.1642e-07, 4.2003e-07, -3.3993e-06, 1.0177e-05, -9.7975e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 215.12, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.5247 re_mapping 0.0052 re_causal 0.0159 /// teacc 98.93 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1116, -0.1248, -0.0908, ..., -0.1409, -0.0110, 0.0774], + [ 0.0571, 0.0582, 0.0438, ..., 0.1008, -0.1519, -0.0572], + [ 0.0545, 0.0262, -0.1275, ..., -0.0325, -0.1859, -0.0732], + ..., + [-0.0481, -0.1143, 0.0876, ..., -0.1088, -0.0348, 0.0092], + [ 0.0228, 0.0914, 0.1374, ..., -0.1183, -0.1150, -0.1823], + [-0.1453, 0.0139, -0.2340, ..., -0.1130, -0.0967, 0.0917]], + device='cuda:0'), grad: tensor([[ 1.2619e-07, 8.5449e-08, 8.7311e-08, ..., 8.5449e-08, + -3.2294e-07, -5.5926e-07], + [-3.4762e-07, 2.1583e-07, -3.3737e-07, ..., 4.2538e-07, + 4.2142e-07, 1.0151e-07], + [ 1.8487e-07, 2.1840e-07, 1.4435e-07, ..., 2.6729e-07, + 1.9814e-07, 1.9418e-07], + ..., + [ 2.4587e-07, 2.7288e-07, 1.8184e-07, ..., 3.5716e-07, + 2.0792e-07, 1.6764e-08], + [-5.2387e-07, 4.7963e-08, -7.8790e-07, ..., 3.8673e-07, + 1.0775e-06, 3.9348e-07], + [-2.3451e-06, -3.3830e-07, 7.7533e-08, ..., 4.9174e-07, + 3.7462e-07, -4.8503e-06]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0037, -0.0203, -0.0258, 0.0349, -0.0277, 0.0331, -0.0245, -0.0084, + 0.0129, -0.0304], device='cuda:0'), grad: tensor([-4.4052e-07, 1.3206e-06, 1.6456e-06, 9.7603e-06, -7.1526e-06, + -5.3681e-06, 9.0003e-06, 1.4696e-06, 1.7909e-06, -1.2048e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 215.01, cls_loss 0.0016 cls_loss_mapping 0.0023 cls_loss_causal 0.5124 re_mapping 0.0054 re_causal 0.0158 /// teacc 98.97 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1126, -0.1250, -0.0913, ..., -0.1410, -0.0110, 0.0771], + [ 0.0594, 0.0602, 0.0461, ..., 0.1019, -0.1525, -0.0562], + [ 0.0540, 0.0254, -0.1288, ..., -0.0326, -0.1863, -0.0738], + ..., + [-0.0486, -0.1156, 0.0862, ..., -0.1107, -0.0350, 0.0071], + [ 0.0210, 0.0895, 0.1367, ..., -0.1184, -0.1156, -0.1839], + [-0.1452, 0.0139, -0.2346, ..., -0.1135, -0.0971, 0.0927]], + device='cuda:0'), grad: tensor([[ 1.5739e-07, 5.7742e-08, 7.1712e-08, ..., 5.4482e-08, + -6.2631e-07, -2.6822e-06], + [-2.4289e-06, -6.7009e-07, 6.5528e-06, ..., -1.9185e-06, + 6.0070e-08, 1.2303e-06], + [-1.7993e-06, 4.1584e-07, 1.3132e-06, ..., 9.4203e-07, + 7.4971e-08, 1.0440e-06], + ..., + [ 5.5553e-07, 3.6880e-07, -4.5091e-05, ..., 5.6764e-07, + 1.2992e-07, -1.3530e-05], + [ 2.9756e-07, 9.7789e-08, 6.5379e-07, ..., 1.4948e-07, + 1.1222e-07, 1.2228e-06], + [ 1.7928e-07, 9.4483e-07, 3.0756e-05, ..., 7.0455e-07, + 6.1886e-07, 9.5069e-06]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0034, -0.0175, -0.0264, 0.0346, -0.0274, 0.0333, -0.0245, -0.0105, + 0.0110, -0.0301], device='cuda:0'), grad: tensor([-4.6864e-06, 1.7181e-05, 1.3579e-06, 1.2316e-05, -4.6566e-06, + 6.1058e-06, 1.6047e-06, -1.2517e-04, 4.9211e-06, 9.0837e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 215.01, cls_loss 0.0013 cls_loss_mapping 0.0038 cls_loss_causal 0.5043 re_mapping 0.0055 re_causal 0.0157 /// teacc 99.02 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1130, -0.1252, -0.0914, ..., -0.1411, -0.0112, 0.0764], + [ 0.0603, 0.0608, 0.0462, ..., 0.1025, -0.1534, -0.0563], + [ 0.0535, 0.0248, -0.1297, ..., -0.0335, -0.1869, -0.0741], + ..., + [-0.0484, -0.1158, 0.0874, ..., -0.1110, -0.0357, 0.0079], + [ 0.0209, 0.0892, 0.1348, ..., -0.1188, -0.1163, -0.1872], + [-0.1454, 0.0138, -0.2357, ..., -0.1138, -0.0972, 0.0937]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, 1.9558e-08, 3.2596e-08, ..., 1.2107e-08, + 0.0000e+00, 6.3330e-08], + [-7.3202e-07, -2.4028e-07, -3.0641e-07, ..., -1.9185e-07, + 4.8429e-08, 5.1316e-07], + [ 4.3679e-07, 1.5367e-07, 2.6077e-07, ..., 1.1455e-07, + 1.5739e-07, 1.2442e-06], + ..., + [ 5.7090e-07, 8.6613e-08, -7.7784e-06, ..., 4.4703e-08, + 1.2731e-06, -8.6799e-06], + [-4.4145e-07, -4.1164e-07, -5.1595e-07, ..., 2.0489e-08, + 7.9162e-08, 5.2340e-07], + [ 1.7323e-07, -8.8476e-08, 8.0243e-06, ..., 5.5879e-08, + -2.3097e-07, 1.8358e-05]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0029, -0.0172, -0.0269, 0.0346, -0.0275, 0.0330, -0.0243, -0.0091, + 0.0093, -0.0298], device='cuda:0'), grad: tensor([ 1.1884e-06, 4.0419e-07, 5.1744e-06, 1.0200e-05, 1.3281e-06, + -6.1631e-05, 3.5651e-06, 4.2431e-06, 8.0559e-07, 3.4690e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 215.25, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4927 re_mapping 0.0055 re_causal 0.0153 /// teacc 99.09 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1132, -0.1253, -0.0915, ..., -0.1411, -0.0110, 0.0768], + [ 0.0605, 0.0609, 0.0462, ..., 0.1028, -0.1539, -0.0565], + [ 0.0538, 0.0243, -0.1300, ..., -0.0351, -0.1888, -0.0744], + ..., + [-0.0486, -0.1160, 0.0876, ..., -0.1110, -0.0362, 0.0081], + [ 0.0210, 0.0900, 0.1349, ..., -0.1190, -0.1162, -0.1876], + [-0.1458, 0.0136, -0.2363, ..., -0.1141, -0.0975, 0.0937]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 6.0536e-08, 1.3039e-08, ..., 7.3574e-08, + 8.4750e-08, -2.7940e-09], + [-1.4901e-07, -2.4214e-08, -1.8626e-09, ..., -2.6450e-07, + 1.1548e-07, 1.7602e-07], + [-8.4750e-08, 3.6322e-08, 5.0291e-08, ..., 5.8673e-08, + 4.2841e-08, 4.2841e-08], + ..., + [ 2.7008e-08, 6.0536e-08, -5.4017e-07, ..., 6.3330e-08, + -0.0000e+00, -3.5949e-07], + [ 2.6077e-08, 4.5635e-08, 1.1176e-08, ..., 5.4017e-08, + 9.2201e-08, 3.5390e-08], + [ 3.3528e-08, -3.1851e-07, 2.6729e-07, ..., 8.6613e-08, + 2.7008e-08, -3.3155e-07]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0032, -0.0172, -0.0271, 0.0353, -0.0275, 0.0322, -0.0242, -0.0090, + 0.0098, -0.0302], device='cuda:0'), grad: tensor([ 2.1700e-07, 1.6950e-07, 5.0291e-08, 3.1758e-07, 1.5358e-06, + 3.0827e-07, -1.5879e-06, -1.0943e-06, 2.6356e-07, -1.9372e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 214.91, cls_loss 0.0014 cls_loss_mapping 0.0034 cls_loss_causal 0.5270 re_mapping 0.0049 re_causal 0.0155 /// teacc 98.97 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1135, -0.1255, -0.0916, ..., -0.1411, -0.0110, 0.0771], + [ 0.0605, 0.0613, 0.0463, ..., 0.1036, -0.1538, -0.0566], + [ 0.0547, 0.0247, -0.1301, ..., -0.0356, -0.1892, -0.0746], + ..., + [-0.0491, -0.1169, 0.0877, ..., -0.1113, -0.0350, 0.0085], + [ 0.0209, 0.0905, 0.1355, ..., -0.1193, -0.1165, -0.1880], + [-0.1461, 0.0157, -0.2376, ..., -0.1146, -0.0957, 0.0953]], + device='cuda:0'), grad: tensor([[ 1.0803e-07, 1.1642e-07, 1.3970e-08, ..., 6.9849e-08, + 1.8626e-07, -1.8906e-07], + [ 4.9360e-08, -6.6124e-08, -1.4715e-07, ..., -1.4156e-07, + 1.0431e-07, 6.4261e-08], + [ 1.0002e-06, 2.9802e-08, 6.2864e-07, ..., 1.6764e-08, + 2.5146e-08, 9.7416e-07], + ..., + [ 2.6636e-07, 6.7055e-08, -8.5030e-07, ..., 8.3819e-08, + 2.7008e-08, -1.5143e-06], + [ 1.8626e-09, 4.1723e-07, -1.9185e-07, ..., 4.9360e-08, + 4.1462e-06, 6.1933e-07], + [ 1.8440e-07, 2.8871e-08, 2.3376e-07, ..., 3.9116e-08, + 5.4017e-08, 3.6974e-07]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0033, -0.0171, -0.0265, 0.0356, -0.0300, 0.0320, -0.0245, -0.0090, + 0.0101, -0.0282], device='cuda:0'), grad: tensor([ 2.8498e-07, 3.8184e-07, 4.1015e-06, -1.2562e-05, 7.1712e-08, + -1.4985e-06, 2.6710e-06, -2.5816e-06, 7.8827e-06, 1.2135e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 214.93, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5342 re_mapping 0.0053 re_causal 0.0152 /// teacc 99.02 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1140, -0.1258, -0.0918, ..., -0.1412, -0.0109, 0.0786], + [ 0.0607, 0.0611, 0.0465, ..., 0.1036, -0.1551, -0.0568], + [ 0.0551, 0.0248, -0.1308, ..., -0.0372, -0.1921, -0.0749], + ..., + [-0.0494, -0.1177, 0.0878, ..., -0.1116, -0.0353, 0.0088], + [ 0.0219, 0.0922, 0.1372, ..., -0.1198, -0.1159, -0.1885], + [-0.1468, 0.0149, -0.2404, ..., -0.1167, -0.0969, 0.0943]], + device='cuda:0'), grad: tensor([[ 2.0396e-07, 6.9849e-08, 5.4017e-08, ..., 4.1910e-08, + 1.5739e-07, 6.6124e-08], + [ 1.4799e-06, 3.5483e-07, 9.4995e-07, ..., -1.6950e-07, + 2.8871e-08, 3.6415e-07], + [-4.9211e-06, -1.3262e-06, -2.7232e-06, ..., 5.1223e-08, + 4.0978e-08, -7.4413e-07], + ..., + [ 1.9763e-06, 5.8673e-07, -2.3749e-07, ..., 3.8184e-08, + 2.0489e-08, 1.9465e-07], + [ 7.3388e-07, 2.7567e-07, 4.1910e-07, ..., 7.8231e-08, + 1.1921e-07, 1.3318e-07], + [ 2.8405e-07, 8.1956e-08, 2.4959e-07, ..., 6.3330e-08, + 4.7497e-08, 7.3574e-08]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0043, -0.0172, -0.0268, 0.0359, -0.0290, 0.0314, -0.0252, -0.0090, + 0.0119, -0.0297], device='cuda:0'), grad: tensor([ 7.7486e-07, 5.3979e-06, -1.4707e-05, 3.6135e-07, 1.5616e-05, + -2.8871e-07, -6.5099e-07, -1.0356e-05, 2.6170e-06, 1.1986e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 215.26, cls_loss 0.0018 cls_loss_mapping 0.0041 cls_loss_causal 0.5341 re_mapping 0.0049 re_causal 0.0148 /// teacc 99.04 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1152, -0.1265, -0.0934, ..., -0.1413, -0.0120, 0.0780], + [ 0.0612, 0.0613, 0.0466, ..., 0.1040, -0.1565, -0.0570], + [ 0.0547, 0.0239, -0.1318, ..., -0.0382, -0.1926, -0.0752], + ..., + [-0.0498, -0.1180, 0.0879, ..., -0.1118, -0.0359, 0.0088], + [ 0.0231, 0.0926, 0.1381, ..., -0.1204, -0.1175, -0.1897], + [-0.1472, 0.0148, -0.2408, ..., -0.1173, -0.0975, 0.0945]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, 2.6077e-08, 1.5832e-08, ..., 1.4901e-08, + -1.0896e-07, -9.6578e-07], + [-9.3132e-10, -3.6322e-08, -1.5553e-07, ..., -1.1642e-07, + 4.2841e-08, 4.1910e-08], + [-9.7789e-08, 4.1910e-08, 3.5390e-08, ..., 3.7253e-08, + 2.3283e-08, 2.6077e-08], + ..., + [ 1.0310e-06, 9.8720e-08, 1.7695e-07, ..., 9.3132e-08, + 3.2596e-08, 8.8476e-08], + [ 4.6287e-07, 7.9162e-08, 9.8720e-08, ..., 4.6566e-08, + 9.4064e-08, 8.4750e-08], + [ 2.6077e-07, 3.3915e-05, 8.4750e-08, ..., 2.5526e-05, + 2.1711e-05, 7.8380e-06]], device='cuda:0') +Epoch 215, bias, value: tensor([ 0.0035, -0.0171, -0.0275, 0.0353, -0.0290, 0.0318, -0.0246, -0.0087, + 0.0127, -0.0299], device='cuda:0'), grad: tensor([-1.0421e-06, 5.5041e-07, -4.3679e-07, -9.9838e-06, -1.2946e-04, + 3.9302e-06, 1.0086e-06, 3.9376e-06, 1.9372e-06, 1.2946e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.93, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4790 re_mapping 0.0051 re_causal 0.0150 /// teacc 99.03 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1155, -0.1267, -0.0933, ..., -0.1414, -0.0119, 0.0785], + [ 0.0601, 0.0608, 0.0456, ..., 0.1038, -0.1578, -0.0571], + [ 0.0550, 0.0235, -0.1323, ..., -0.0385, -0.1929, -0.0756], + ..., + [-0.0485, -0.1171, 0.0889, ..., -0.1118, -0.0369, 0.0088], + [ 0.0235, 0.0927, 0.1383, ..., -0.1211, -0.1181, -0.1907], + [-0.1472, 0.0146, -0.2410, ..., -0.1179, -0.0981, 0.0946]], + device='cuda:0'), grad: tensor([[ 2.7008e-08, 9.3132e-09, 1.3970e-08, ..., 7.4506e-09, + 7.4506e-09, -9.3132e-10], + [-6.1467e-08, -6.0536e-08, -1.0896e-07, ..., -6.5193e-08, + 6.5193e-09, -1.3970e-08], + [-7.4506e-09, 8.3819e-09, 2.7940e-08, ..., 2.2352e-08, + 9.3132e-09, -1.8626e-08], + ..., + [ 5.3085e-08, 2.8871e-08, 1.8626e-08, ..., 3.5390e-08, + 9.3132e-10, 1.7695e-08], + [-2.9895e-07, -6.0536e-08, -3.5577e-07, ..., 1.1176e-08, + 3.7253e-09, 1.1176e-08], + [ 5.4017e-08, 2.9802e-08, 4.0978e-08, ..., 6.2399e-08, + 2.1420e-08, -1.4901e-08]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0037, -0.0181, -0.0274, 0.0347, -0.0286, 0.0317, -0.0247, -0.0076, + 0.0127, -0.0301], device='cuda:0'), grad: tensor([ 6.6124e-08, -1.0617e-07, 4.2841e-08, -9.3132e-10, -3.4831e-07, + 4.3400e-07, 2.4121e-07, 5.9605e-08, -6.0070e-07, 2.2165e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 215.17, cls_loss 0.0014 cls_loss_mapping 0.0035 cls_loss_causal 0.5162 re_mapping 0.0055 re_causal 0.0160 /// teacc 98.96 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1163, -0.1272, -0.0946, ..., -0.1420, -0.0120, 0.0786], + [ 0.0604, 0.0606, 0.0453, ..., 0.1028, -0.1595, -0.0572], + [ 0.0552, 0.0232, -0.1327, ..., -0.0398, -0.1940, -0.0759], + ..., + [-0.0486, -0.1173, 0.0895, ..., -0.1102, -0.0369, 0.0088], + [ 0.0238, 0.0932, 0.1392, ..., -0.1211, -0.1183, -0.1912], + [-0.1476, 0.0154, -0.2420, ..., -0.1175, -0.0977, 0.0956]], + device='cuda:0'), grad: tensor([[ 1.0524e-07, 2.6263e-07, 1.3039e-08, ..., 7.4506e-09, + 2.9095e-06, -3.0454e-06], + [ 2.1700e-07, 3.3528e-08, 1.2266e-06, ..., 3.8184e-08, + 6.6124e-08, 7.3016e-07], + [-3.7868e-06, -4.4610e-07, 6.3330e-08, ..., 2.8871e-08, + 2.0489e-08, 9.3132e-10], + ..., + [ 1.0962e-06, 1.0431e-07, -2.2650e-06, ..., 7.5437e-08, + 2.0489e-08, -1.0710e-07], + [ 1.5600e-06, 2.4587e-07, 9.3132e-09, ..., 2.7008e-08, + 2.2445e-07, 4.7591e-07], + [ 5.9605e-08, 1.8626e-07, 6.2585e-07, ..., 6.2678e-07, + 3.0268e-07, 3.6787e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0036, -0.0185, -0.0275, 0.0352, -0.0294, 0.0308, -0.0242, -0.0071, + 0.0135, -0.0297], device='cuda:0'), grad: tensor([ 4.9248e-06, 3.5614e-06, -7.4394e-06, 1.5292e-06, -1.6456e-06, + 3.0696e-06, -1.0401e-05, -1.1791e-06, 4.5635e-06, 3.0287e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 214.82, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4455 re_mapping 0.0050 re_causal 0.0147 /// teacc 98.97 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1169, -0.1286, -0.0951, ..., -0.1422, -0.0137, 0.0762], + [ 0.0600, 0.0609, 0.0457, ..., 0.1030, -0.1599, -0.0570], + [ 0.0564, 0.0232, -0.1327, ..., -0.0400, -0.1943, -0.0761], + ..., + [-0.0491, -0.1176, 0.0894, ..., -0.1105, -0.0371, 0.0088], + [ 0.0239, 0.0931, 0.1396, ..., -0.1213, -0.1188, -0.1921], + [-0.1475, 0.0152, -0.2422, ..., -0.1196, -0.0986, 0.0969]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 3.7160e-07, 1.4901e-08, ..., 1.5460e-07, + 2.6543e-07, -3.1292e-07], + [-4.4610e-07, -2.8964e-07, -3.4552e-07, ..., -5.2247e-07, + 5.0291e-08, 1.8626e-08], + [ 2.0117e-07, 4.8988e-07, 2.1514e-07, ..., 4.0978e-07, + 2.4214e-07, 5.8673e-08], + ..., + [ 1.4249e-07, 9.5926e-08, 9.3132e-08, ..., 1.1083e-07, + 3.7253e-09, 1.8626e-07], + [-2.5053e-07, 7.5903e-07, -4.7870e-07, ..., 3.6601e-07, + 6.4820e-07, 4.9360e-08], + [ 2.1420e-08, 1.7695e-08, 1.9558e-08, ..., 5.7742e-08, + 4.0978e-08, -1.9930e-07]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0016, -0.0185, -0.0267, 0.0354, -0.0289, 0.0304, -0.0230, -0.0074, + 0.0134, -0.0296], device='cuda:0'), grad: tensor([ 3.4086e-07, -1.0962e-06, 1.3718e-06, 3.1665e-07, 3.7737e-06, + 1.0654e-06, -7.3984e-06, 6.1281e-07, 1.1222e-06, -1.0710e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 214.94, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5051 re_mapping 0.0050 re_causal 0.0148 /// teacc 98.95 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1171, -0.1288, -0.0951, ..., -0.1423, -0.0135, 0.0767], + [ 0.0587, 0.0595, 0.0458, ..., 0.1022, -0.1611, -0.0570], + [ 0.0578, 0.0235, -0.1337, ..., -0.0377, -0.1946, -0.0765], + ..., + [-0.0494, -0.1179, 0.0893, ..., -0.1109, -0.0376, 0.0084], + [ 0.0250, 0.0944, 0.1406, ..., -0.1215, -0.1196, -0.1928], + [-0.1475, 0.0158, -0.2422, ..., -0.1199, -0.0986, 0.0975]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 5.7742e-08, 1.5832e-08, ..., 3.6322e-08, + 6.6124e-08, 3.3528e-08], + [ 4.0978e-08, 7.5903e-07, 1.1176e-08, ..., 5.0385e-07, + 4.5076e-07, 1.6484e-07], + [ 9.8720e-08, 3.1386e-07, 7.8231e-08, ..., 2.0862e-07, + 1.8161e-07, 9.3132e-09], + ..., + [ 4.5635e-08, 1.6671e-07, 2.3283e-08, ..., 9.3132e-08, + 1.7975e-07, 2.8033e-07], + [-3.2224e-07, -1.3784e-07, -3.3248e-07, ..., 5.4948e-08, + 1.3318e-07, 2.8126e-07], + [ 1.3225e-07, -2.3656e-07, 9.9652e-08, ..., 1.8161e-07, + 2.1048e-07, -7.4971e-07]], device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0019, -0.0192, -0.0260, 0.0354, -0.0294, 0.0304, -0.0231, -0.0077, + 0.0142, -0.0292], device='cuda:0'), grad: tensor([ 2.2352e-07, 1.9912e-06, 8.2236e-07, -1.1269e-07, 3.3714e-07, + -8.7451e-07, -2.2762e-06, 1.3923e-06, 3.0361e-07, -1.7975e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 215.15, cls_loss 0.0014 cls_loss_mapping 0.0034 cls_loss_causal 0.4952 re_mapping 0.0051 re_causal 0.0150 /// teacc 99.01 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1173, -0.1292, -0.0952, ..., -0.1426, -0.0137, 0.0768], + [ 0.0585, 0.0589, 0.0459, ..., 0.1019, -0.1628, -0.0574], + [ 0.0585, 0.0237, -0.1342, ..., -0.0371, -0.1951, -0.0770], + ..., + [-0.0503, -0.1190, 0.0888, ..., -0.1112, -0.0388, 0.0065], + [ 0.0254, 0.0946, 0.1413, ..., -0.1220, -0.1217, -0.1947], + [-0.1473, 0.0159, -0.2418, ..., -0.1211, -0.0993, 0.0986]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, 3.8184e-08, 1.3039e-07, ..., 6.5193e-09, + 3.7253e-08, 5.4017e-08], + [-4.8801e-07, -2.4773e-07, -2.6822e-07, ..., -2.1979e-07, + 2.7008e-08, 6.5193e-08], + [ 1.2852e-07, 9.2201e-08, 6.1467e-07, ..., 5.2154e-08, + 1.3039e-08, 1.6298e-07], + ..., + [ 2.7679e-06, 3.6377e-06, 2.9683e-04, ..., 6.7055e-08, + 2.9802e-08, 9.7513e-05], + [-3.9153e-06, -4.6752e-06, -3.4738e-04, ..., 2.2352e-08, + -8.6613e-08, -1.1349e-04], + [ 4.5355e-07, 6.3330e-07, 4.3690e-05, ..., 1.2573e-07, + 2.2631e-07, 1.4350e-05]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0019, -0.0196, -0.0254, 0.0361, -0.0291, 0.0305, -0.0231, -0.0090, + 0.0140, -0.0287], device='cuda:0'), grad: tensor([ 3.1386e-07, -6.6124e-07, 9.9372e-07, -1.2629e-06, 3.6322e-07, + 5.6028e-06, 2.6673e-06, 3.8075e-04, -4.4584e-04, 5.6624e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 215.35, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4735 re_mapping 0.0051 re_causal 0.0149 /// teacc 99.07 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1176, -0.1295, -0.0953, ..., -0.1429, -0.0140, 0.0766], + [ 0.0597, 0.0590, 0.0468, ..., 0.1019, -0.1638, -0.0574], + [ 0.0580, 0.0235, -0.1361, ..., -0.0369, -0.1959, -0.0789], + ..., + [-0.0519, -0.1197, 0.0878, ..., -0.1115, -0.0393, 0.0055], + [ 0.0250, 0.0945, 0.1422, ..., -0.1226, -0.1224, -0.1946], + [-0.1477, 0.0157, -0.2423, ..., -0.1217, -0.0997, 0.0993]], + device='cuda:0'), grad: tensor([[ 4.8243e-07, 2.0489e-08, 4.3772e-08, ..., 9.3132e-09, + 1.5367e-06, 1.9064e-06], + [-1.2107e-08, -5.8673e-08, -2.3562e-07, ..., -1.4249e-07, + 2.3283e-08, 7.2643e-08], + [-7.5214e-06, 1.3039e-08, 8.4750e-08, ..., 9.3132e-09, + 3.8184e-08, 5.0291e-08], + ..., + [ 7.0315e-07, 1.2014e-07, -7.2923e-07, ..., 9.0338e-08, + 2.9802e-08, 1.0664e-06], + [ 1.7798e-06, 4.0419e-07, 5.4296e-07, ..., 2.3283e-08, + 6.5193e-08, 1.2144e-06], + [ 2.5984e-07, -2.0713e-06, -2.6096e-06, ..., 3.4459e-08, + 2.2072e-07, -8.1584e-06]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0016, -0.0189, -0.0264, 0.0361, -0.0289, 0.0315, -0.0230, -0.0102, + 0.0139, -0.0286], device='cuda:0'), grad: tensor([ 5.2750e-06, 4.7125e-07, -3.1590e-05, 1.6734e-05, 1.3590e-05, + -2.8051e-06, 8.5495e-07, 2.6673e-06, 9.7156e-06, -1.4916e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.63, cls_loss 0.0017 cls_loss_mapping 0.0023 cls_loss_causal 0.4931 re_mapping 0.0049 re_causal 0.0145 /// teacc 99.00 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1181, -0.1299, -0.0954, ..., -0.1430, -0.0135, 0.0771], + [ 0.0627, 0.0587, 0.0471, ..., 0.1019, -0.1645, -0.0570], + [ 0.0535, 0.0234, -0.1390, ..., -0.0368, -0.1968, -0.0824], + ..., + [-0.0497, -0.1199, 0.0885, ..., -0.1113, -0.0382, 0.0059], + [ 0.0246, 0.0934, 0.1420, ..., -0.1229, -0.1231, -0.1983], + [-0.1477, 0.0164, -0.2423, ..., -0.1222, -0.1004, 0.1004]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 3.2410e-07, 1.8626e-09, ..., 1.8626e-09, + 1.0617e-06, 2.8498e-07], + [-3.0734e-08, -1.5832e-08, -4.3772e-08, ..., -2.5146e-08, + 9.0338e-08, 1.3970e-07], + [-1.5832e-08, 6.5193e-09, 9.3132e-09, ..., 5.5879e-09, + 1.6764e-08, 1.6764e-08], + ..., + [ 8.6613e-08, 3.1665e-08, -7.1712e-08, ..., 3.0734e-08, + 3.7253e-08, -5.5879e-08], + [ 2.8871e-08, 2.7940e-08, -1.8626e-09, ..., 6.5193e-09, + 1.6298e-07, 1.1083e-07], + [ 6.7055e-08, 2.7288e-07, 8.0094e-08, ..., 2.2445e-07, + 2.8219e-07, 1.4715e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0019, -0.0164, -0.0318, 0.0358, -0.0289, 0.0314, -0.0232, -0.0076, + 0.0125, -0.0281], device='cuda:0'), grad: tensor([ 1.6866e-06, 2.3935e-07, 7.4506e-09, 2.2799e-05, -8.1863e-07, + -2.3574e-05, -1.9819e-06, -1.6764e-08, 3.6974e-07, 1.2731e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.82, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4803 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.07 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1186, -0.1304, -0.0957, ..., -0.1432, -0.0136, 0.0772], + [ 0.0628, 0.0586, 0.0472, ..., 0.1018, -0.1659, -0.0573], + [ 0.0535, 0.0232, -0.1395, ..., -0.0371, -0.1977, -0.0826], + ..., + [-0.0498, -0.1203, 0.0885, ..., -0.1117, -0.0388, 0.0059], + [ 0.0246, 0.0935, 0.1424, ..., -0.1231, -0.1232, -0.1986], + [-0.1478, 0.0168, -0.2423, ..., -0.1227, -0.1006, 0.1009]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 3.9116e-08, 1.0245e-08, ..., 8.3819e-09, + -2.0582e-07, -7.7859e-07], + [ 1.1176e-08, 2.7381e-07, 1.1828e-07, ..., 9.2201e-08, + 2.3097e-07, 6.7055e-08], + [ 8.3819e-09, 1.7695e-08, 1.0245e-08, ..., 4.6566e-09, + 2.0489e-08, 2.1420e-08], + ..., + [ 8.3819e-09, 1.2107e-08, -1.9558e-08, ..., 6.5193e-09, + 2.5146e-08, 3.6322e-08], + [-1.8626e-08, -2.1514e-07, -4.2189e-07, ..., 1.2014e-07, + 3.6974e-07, 1.4994e-07], + [ 1.2107e-08, -2.7940e-09, 4.0978e-08, ..., 8.8476e-08, + 7.8045e-07, 1.0766e-06]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0017, -0.0164, -0.0319, 0.0358, -0.0292, 0.0316, -0.0232, -0.0075, + 0.0126, -0.0277], device='cuda:0'), grad: tensor([-8.9500e-07, 7.8510e-07, 1.0338e-07, 5.1316e-07, 2.9244e-07, + -5.2713e-07, -2.0675e-06, 8.0094e-08, -1.8161e-07, 1.8831e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 215.17, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4843 re_mapping 0.0047 re_causal 0.0138 /// teacc 99.12 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1190, -0.1309, -0.0960, ..., -0.1434, -0.0137, 0.0773], + [ 0.0627, 0.0579, 0.0468, ..., 0.1007, -0.1673, -0.0574], + [ 0.0536, 0.0229, -0.1397, ..., -0.0370, -0.1982, -0.0828], + ..., + [-0.0497, -0.1205, 0.0890, ..., -0.1121, -0.0394, 0.0060], + [ 0.0246, 0.0935, 0.1425, ..., -0.1235, -0.1237, -0.1989], + [-0.1481, 0.0174, -0.2428, ..., -0.1233, -0.1009, 0.1015]], + device='cuda:0'), grad: tensor([[ 3.1851e-07, 5.4017e-08, 6.0536e-08, ..., 1.2107e-08, + -5.5879e-09, -1.4063e-07], + [-5.5768e-06, -1.1995e-06, -9.1568e-06, ..., -1.1176e-08, + 4.0885e-07, -2.8275e-06], + [-1.2033e-05, 4.1910e-08, 8.4750e-08, ..., 1.3039e-08, + 1.5832e-08, -4.8894e-07], + ..., + [ 2.1644e-06, 6.0257e-07, 3.4068e-06, ..., 4.0047e-08, + 8.3819e-09, 1.1157e-06], + [ 1.4566e-06, 6.0350e-07, 1.4855e-06, ..., 6.7055e-08, + 2.8126e-07, 9.3505e-07], + [ 1.9781e-06, -2.0489e-08, 3.0231e-06, ..., 7.3574e-08, + 7.0781e-08, -6.5193e-08]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0016, -0.0167, -0.0319, 0.0365, -0.0296, 0.0311, -0.0230, -0.0072, + 0.0125, -0.0272], device='cuda:0'), grad: tensor([ 5.6531e-07, -1.2614e-05, -2.3335e-05, 2.2247e-05, 1.6056e-06, + 2.9616e-06, -4.1761e-06, 5.8152e-06, 4.2692e-06, 2.6003e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 223---------------------------------------------------- +epoch 223, time 231.54, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5027 re_mapping 0.0048 re_causal 0.0145 /// teacc 99.13 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1195, -0.1311, -0.0963, ..., -0.1435, -0.0136, 0.0775], + [ 0.0627, 0.0569, 0.0469, ..., 0.1006, -0.1678, -0.0575], + [ 0.0536, 0.0240, -0.1401, ..., -0.0363, -0.1984, -0.0824], + ..., + [-0.0499, -0.1211, 0.0887, ..., -0.1124, -0.0395, 0.0055], + [ 0.0257, 0.0949, 0.1435, ..., -0.1238, -0.1239, -0.1988], + [-0.1486, 0.0174, -0.2434, ..., -0.1238, -0.1013, 0.1017]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 6.5193e-09, 3.7253e-09, ..., 8.3819e-09, + -1.0906e-06, -3.5185e-06], + [ 1.8626e-08, 2.3842e-07, 2.2165e-07, ..., 4.3027e-07, + 1.0990e-07, 7.9162e-08], + [-1.6857e-07, 5.6811e-07, 3.7253e-08, ..., 1.2433e-06, + 2.9597e-06, 9.2201e-08], + ..., + [ 6.6124e-08, 1.8068e-07, 9.8720e-08, ..., 3.2596e-07, + 7.0781e-08, 7.9162e-08], + [ 1.3132e-07, 2.2352e-08, 1.3970e-08, ..., 7.2643e-08, + 3.0734e-08, 6.9849e-08], + [ 4.1910e-08, 1.2945e-07, 1.4063e-07, ..., 2.3562e-07, + 1.2852e-07, -8.0187e-07]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0017, -0.0168, -0.0318, 0.0371, -0.0295, 0.0305, -0.0236, -0.0076, + 0.0136, -0.0272], device='cuda:0'), grad: tensor([-5.5656e-06, 1.1986e-06, 4.7237e-06, -5.8860e-07, -6.7130e-06, + -8.6240e-07, 6.6720e-06, 8.9128e-07, 5.8673e-07, -3.4738e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.87, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4935 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.11 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1201, -0.1316, -0.0967, ..., -0.1436, -0.0126, 0.0798], + [ 0.0627, 0.0566, 0.0472, ..., 0.1011, -0.1683, -0.0576], + [ 0.0536, 0.0244, -0.1405, ..., -0.0362, -0.1990, -0.0822], + ..., + [-0.0502, -0.1216, 0.0886, ..., -0.1127, -0.0392, 0.0054], + [ 0.0252, 0.0955, 0.1441, ..., -0.1240, -0.1243, -0.1991], + [-0.1489, 0.0175, -0.2440, ..., -0.1240, -0.1032, 0.1000]], + device='cuda:0'), grad: tensor([[ 5.2154e-08, 9.1270e-08, 1.0524e-07, ..., 4.7497e-08, + 8.3819e-09, 7.4506e-08], + [-1.5780e-05, -2.6554e-05, -3.2067e-05, ..., -1.3962e-05, + 4.6566e-09, -2.2694e-05], + [ 5.8673e-08, 9.8720e-08, 1.2014e-07, ..., 5.3085e-08, + 9.3132e-10, 8.4750e-08], + ..., + [ 3.1833e-06, 5.3570e-06, 6.3889e-06, ..., 2.8182e-06, + 6.5193e-09, 4.5411e-06], + [ 7.0594e-07, 1.2526e-06, 1.4463e-06, ..., 6.5658e-07, + 1.7602e-07, 1.1120e-06], + [ 1.0952e-05, 1.8448e-05, 2.2322e-05, ..., 9.7081e-06, + 1.1176e-08, 1.5795e-05]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0032, -0.0168, -0.0318, 0.0376, -0.0295, 0.0305, -0.0239, -0.0077, + 0.0136, -0.0283], device='cuda:0'), grad: tensor([ 2.6543e-07, -7.6532e-05, 2.8964e-07, 3.1106e-07, 3.3639e-06, + -3.5651e-06, 3.4906e-06, 1.5289e-05, 3.8091e-06, 5.3287e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 215.06, cls_loss 0.0015 cls_loss_mapping 0.0032 cls_loss_causal 0.5292 re_mapping 0.0045 re_causal 0.0146 /// teacc 98.97 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1203, -0.1320, -0.0969, ..., -0.1437, -0.0131, 0.0794], + [ 0.0631, 0.0585, 0.0477, ..., 0.1026, -0.1664, -0.0566], + [ 0.0535, 0.0220, -0.1417, ..., -0.0383, -0.2020, -0.0837], + ..., + [-0.0505, -0.1224, 0.0885, ..., -0.1130, -0.0396, 0.0042], + [ 0.0250, 0.0957, 0.1445, ..., -0.1247, -0.1248, -0.1999], + [-0.1493, 0.0180, -0.2445, ..., -0.1244, -0.1028, 0.1013]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, 1.2293e-07, 5.3085e-08, ..., 7.4506e-09, + -2.3358e-06, -3.9712e-06], + [-2.4494e-07, -5.5879e-09, -2.6543e-07, ..., -2.1793e-07, + 1.2573e-07, 2.6263e-07], + [ 6.7987e-08, 4.0978e-08, 4.9360e-08, ..., 1.4901e-08, + 4.4703e-08, 7.6368e-08], + ..., + [ 1.0431e-07, 3.8184e-08, -3.5670e-07, ..., 4.9360e-08, + 1.2480e-07, -1.7602e-07], + [-6.1560e-07, -2.2743e-06, -1.7025e-06, ..., 5.6811e-08, + 5.1316e-07, 2.8219e-07], + [ 8.1677e-07, 2.1346e-06, 1.8496e-06, ..., 1.5832e-08, + 1.2107e-07, 8.5495e-07]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0029, -0.0162, -0.0321, 0.0376, -0.0301, 0.0298, -0.0227, -0.0083, + 0.0133, -0.0277], device='cuda:0'), grad: tensor([-8.2329e-06, -1.5832e-07, 5.3644e-07, -1.6168e-05, 3.9302e-07, + 1.9833e-05, -6.5193e-08, -5.4017e-08, -2.8200e-06, 6.7651e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 215.12, cls_loss 0.0019 cls_loss_mapping 0.0041 cls_loss_causal 0.5330 re_mapping 0.0050 re_causal 0.0148 /// teacc 99.04 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.1205, -0.1323, -0.0970, ..., -0.1439, -0.0130, 0.0797], + [ 0.0628, 0.0575, 0.0474, ..., 0.1007, -0.1696, -0.0556], + [ 0.0537, 0.0220, -0.1413, ..., -0.0386, -0.2023, -0.0834], + ..., + [-0.0503, -0.1231, 0.0890, ..., -0.1134, -0.0401, 0.0023], + [ 0.0251, 0.0960, 0.1451, ..., -0.1254, -0.1275, -0.2012], + [-0.1497, 0.0172, -0.2451, ..., -0.1257, -0.1040, 0.1011]], + device='cuda:0'), grad: tensor([[ 4.9267e-07, 6.8918e-08, 5.7742e-08, ..., 6.6124e-08, + 3.4459e-08, 1.7509e-07], + [ 2.7940e-08, 1.8030e-06, 1.7341e-06, ..., 2.2370e-06, + 7.0967e-07, 1.2899e-06], + [-1.6009e-06, 1.2945e-07, 4.3679e-07, ..., 1.4901e-07, + 5.0291e-08, 6.6683e-07], + ..., + [ 2.8685e-07, 1.6950e-07, -4.1239e-06, ..., 2.0210e-07, + -3.1665e-08, -8.4192e-06], + [ 3.2410e-07, 4.9639e-07, 2.1048e-07, ..., 7.3481e-07, + 2.6822e-07, 2.1420e-07], + [ 5.6811e-08, 8.3540e-07, 2.8666e-06, ..., 1.2880e-06, + 4.5728e-07, 4.8168e-06]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0031, -0.0174, -0.0310, 0.0362, -0.0289, 0.0262, -0.0182, -0.0095, + 0.0127, -0.0288], device='cuda:0'), grad: tensor([ 2.6915e-06, 9.3356e-06, -4.5374e-06, 1.0747e-06, -1.3441e-05, + 2.0042e-06, 3.6750e-06, -1.9357e-05, 3.6564e-06, 1.4856e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 215.08, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4976 re_mapping 0.0050 re_causal 0.0140 /// teacc 99.02 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.1211, -0.1326, -0.0972, ..., -0.1439, -0.0124, 0.0796], + [ 0.0628, 0.0577, 0.0476, ..., 0.1011, -0.1699, -0.0558], + [ 0.0541, 0.0220, -0.1408, ..., -0.0387, -0.2027, -0.0830], + ..., + [-0.0507, -0.1248, 0.0887, ..., -0.1147, -0.0411, 0.0018], + [ 0.0253, 0.0967, 0.1457, ..., -0.1257, -0.1278, -0.2021], + [-0.1504, 0.0172, -0.2453, ..., -0.1267, -0.1045, 0.1025]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 9.0245e-07, 8.6613e-08, ..., 1.8440e-07, + 7.3649e-06, 4.4927e-06], + [-7.1712e-08, 3.5856e-07, -1.3784e-07, ..., 2.5798e-07, + 6.9756e-07, 2.6170e-07], + [-1.2312e-06, 1.1865e-06, 1.2480e-07, ..., 3.7253e-07, + 1.0198e-06, 4.3772e-08], + ..., + [ 1.0803e-07, 8.2701e-07, 4.0978e-08, ..., 4.5914e-07, + 1.3532e-06, 5.8021e-07], + [ 8.6613e-08, -7.4029e-05, -6.6496e-06, ..., 3.1944e-07, + -4.3541e-05, 6.0126e-06], + [ 5.9605e-08, 8.5980e-06, 1.7975e-07, ..., 1.0863e-05, + 1.0170e-05, -2.8685e-07]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0030, -0.0174, -0.0301, 0.0358, -0.0289, 0.0265, -0.0186, -0.0105, + 0.0128, -0.0284], device='cuda:0'), grad: tensor([ 1.8701e-05, 3.0976e-06, 2.3767e-06, 2.7224e-05, -7.2539e-05, + 9.9540e-06, 1.5903e-04, 6.1281e-06, -2.2149e-04, 6.7651e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 214.88, cls_loss 0.0015 cls_loss_mapping 0.0034 cls_loss_causal 0.5013 re_mapping 0.0051 re_causal 0.0142 /// teacc 99.07 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.1218, -0.1331, -0.0978, ..., -0.1442, -0.0124, 0.0797], + [ 0.0627, 0.0577, 0.0476, ..., 0.1015, -0.1700, -0.0564], + [ 0.0539, 0.0219, -0.1419, ..., -0.0392, -0.2031, -0.0835], + ..., + [-0.0497, -0.1253, 0.0895, ..., -0.1151, -0.0417, 0.0029], + [ 0.0254, 0.0974, 0.1464, ..., -0.1262, -0.1276, -0.2034], + [-0.1506, 0.0173, -0.2477, ..., -0.1272, -0.1047, 0.1026]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 2.4214e-08, 3.7253e-09, ..., 4.6566e-09, + -1.2107e-08, -1.4529e-07], + [-7.7300e-08, -3.3528e-08, -9.8720e-08, ..., -3.6322e-08, + 1.7695e-08, 5.6811e-08], + [-2.3581e-06, 1.7695e-08, -8.7637e-07, ..., 6.5193e-09, + 1.3039e-08, 4.1910e-08], + ..., + [ 9.7789e-08, 4.8429e-08, -2.6077e-08, ..., 1.6764e-08, + 1.5832e-08, 1.8226e-06], + [ 2.2575e-06, 1.5274e-07, 8.7358e-07, ..., 2.9802e-08, + 1.4808e-07, 2.5705e-07], + [ 1.3970e-08, -3.5670e-07, 4.2841e-08, ..., 4.6566e-09, + 8.0094e-08, -2.4810e-06]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0029, -0.0177, -0.0306, 0.0352, -0.0290, 0.0268, -0.0188, -0.0091, + 0.0135, -0.0285], device='cuda:0'), grad: tensor([-2.1327e-07, -7.4506e-09, -3.8035e-06, 3.9861e-07, 6.7893e-07, + -4.3120e-07, -6.6124e-08, 3.2037e-06, 4.4294e-06, -4.1723e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 215.06, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4926 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.02 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.1219, -0.1340, -0.0979, ..., -0.1445, -0.0126, 0.0798], + [ 0.0623, 0.0571, 0.0471, ..., 0.1015, -0.1702, -0.0568], + [ 0.0545, 0.0228, -0.1421, ..., -0.0392, -0.2034, -0.0830], + ..., + [-0.0497, -0.1260, 0.0904, ..., -0.1154, -0.0419, 0.0034], + [ 0.0256, 0.0973, 0.1469, ..., -0.1269, -0.1284, -0.2040], + [-0.1508, 0.0175, -0.2498, ..., -0.1274, -0.1047, 0.1028]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 3.2596e-08, 1.8626e-09, ..., 1.3970e-08, + 2.8405e-08, -2.5006e-07], + [-3.2596e-08, 1.2061e-07, -7.1246e-08, ..., -2.4214e-08, + 2.0862e-07, 2.7008e-08], + [ 9.3132e-10, 6.5193e-09, 9.7789e-09, ..., 3.3993e-08, + 9.0804e-08, 3.5390e-08], + ..., + [ 8.1956e-08, 2.7474e-08, 5.7742e-08, ..., 4.8894e-08, + 2.2259e-07, 2.4214e-08], + [-1.9092e-08, 1.7043e-07, -1.4435e-07, ..., 1.7695e-08, + 5.9465e-07, 4.3027e-07], + [ 3.7253e-08, 2.7474e-08, 6.8452e-08, ..., 3.5390e-08, + 1.0496e-06, 6.2678e-07]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0028, -0.0184, -0.0301, 0.0359, -0.0291, 0.0266, -0.0188, -0.0088, + 0.0135, -0.0287], device='cuda:0'), grad: tensor([-3.6834e-07, 5.3644e-07, 4.0093e-07, 7.5903e-08, -3.6806e-06, + -1.2880e-06, -2.2762e-06, 1.0189e-06, 1.6689e-06, 3.9153e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 215.08, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4869 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.11 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.1224, -0.1344, -0.0982, ..., -0.1447, -0.0138, 0.0777], + [ 0.0627, 0.0576, 0.0498, ..., 0.1016, -0.1704, -0.0546], + [ 0.0552, 0.0236, -0.1424, ..., -0.0394, -0.2036, -0.0825], + ..., + [-0.0518, -0.1290, 0.0877, ..., -0.1168, -0.0447, 0.0009], + [ 0.0252, 0.0967, 0.1470, ..., -0.1271, -0.1321, -0.2058], + [-0.1532, 0.0178, -0.2504, ..., -0.1272, -0.1035, 0.1050]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 1.8626e-09, 1.8626e-09, ..., 1.3970e-09, + -1.2945e-06, -4.6268e-06], + [-6.6124e-08, -4.2841e-08, -8.5216e-08, ..., -5.2154e-08, + 4.2375e-08, 1.4482e-07], + [-9.3132e-10, 2.0489e-08, 5.2620e-08, ..., 2.4214e-08, + 1.4808e-07, 5.1595e-07], + ..., + [ 6.7987e-08, 2.2817e-08, -1.8626e-09, ..., 3.9581e-08, + 1.8999e-07, 6.2818e-07], + [ 1.4529e-07, 3.5856e-08, 4.1910e-08, ..., 6.2399e-08, + 1.2666e-07, 3.7951e-07], + [ 5.6345e-08, 2.4680e-08, 2.0023e-08, ..., 9.0338e-08, + 1.1455e-07, 3.1013e-07]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0005, -0.0170, -0.0296, 0.0368, -0.0297, 0.0273, -0.0191, -0.0113, + 0.0112, -0.0269], device='cuda:0'), grad: tensor([-9.6336e-06, 2.0675e-07, 1.1213e-06, -9.7323e-08, -4.0652e-07, + 1.6103e-06, 3.4980e-06, 1.4864e-06, 1.2470e-06, 9.6858e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 214.55, cls_loss 0.0014 cls_loss_mapping 0.0021 cls_loss_causal 0.5008 re_mapping 0.0048 re_causal 0.0142 /// teacc 99.12 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.1231, -0.1349, -0.0984, ..., -0.1450, -0.0138, 0.0778], + [ 0.0627, 0.0563, 0.0497, ..., 0.1007, -0.1712, -0.0548], + [ 0.0551, 0.0234, -0.1429, ..., -0.0397, -0.2043, -0.0825], + ..., + [-0.0518, -0.1291, 0.0879, ..., -0.1169, -0.0451, 0.0011], + [ 0.0253, 0.0963, 0.1476, ..., -0.1276, -0.1345, -0.2068], + [-0.1537, 0.0179, -0.2510, ..., -0.1273, -0.1036, 0.1052]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 9.1735e-08, 7.9162e-09, ..., 9.3132e-09, + 3.2131e-08, 4.4052e-07], + [-9.2201e-08, 7.8697e-08, -1.2480e-07, ..., -1.8161e-08, + 1.0384e-07, 3.7253e-09], + [ 1.6764e-08, 6.1002e-08, 6.0536e-08, ..., 1.3504e-08, + 2.0023e-08, 8.4285e-08], + ..., + [ 5.7742e-08, 3.3062e-08, -2.9663e-07, ..., 1.7229e-08, + 2.3283e-09, 2.2817e-08], + [-3.4552e-06, -2.8461e-06, -4.7199e-06, ..., 6.9849e-08, + 2.2957e-07, -2.9709e-07], + [ 3.3546e-06, 2.9858e-06, 4.6864e-06, ..., 2.2817e-08, + 4.0978e-08, -2.5192e-07]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0005, -0.0175, -0.0298, 0.0369, -0.0301, 0.0272, -0.0186, -0.0111, + 0.0106, -0.0268], device='cuda:0'), grad: tensor([ 6.3190e-07, 2.3330e-07, 2.4727e-07, 1.0245e-06, 5.8720e-07, + -9.2061e-07, -1.4314e-06, -6.1234e-07, -9.4920e-06, 9.7305e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 214.60, cls_loss 0.0010 cls_loss_mapping 0.0029 cls_loss_causal 0.4767 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.07 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.1233, -0.1367, -0.0985, ..., -0.1451, -0.0147, 0.0773], + [ 0.0629, 0.0566, 0.0499, ..., 0.1008, -0.1713, -0.0547], + [ 0.0549, 0.0228, -0.1440, ..., -0.0400, -0.2044, -0.0826], + ..., + [-0.0521, -0.1294, 0.0875, ..., -0.1171, -0.0454, 0.0009], + [ 0.0253, 0.0962, 0.1493, ..., -0.1277, -0.1353, -0.2073], + [-0.1541, 0.0177, -0.2513, ..., -0.1279, -0.1039, 0.1055]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -3.0361e-07, 2.0489e-08, ..., 1.8626e-09, + -2.5257e-06, -5.1931e-06], + [ 2.1886e-07, 1.1735e-07, 2.1402e-06, ..., -1.9558e-08, + 1.6764e-08, 6.7148e-07], + [ 2.9150e-07, 2.1979e-07, 2.8852e-06, ..., 7.4506e-09, + 1.2107e-08, 8.5123e-07], + ..., + [-6.9812e-06, -3.9116e-06, -6.1154e-05, ..., 7.4506e-09, + 4.0978e-08, -1.8045e-05], + [ 3.3583e-06, 1.7118e-06, 3.0696e-05, ..., -3.9116e-08, + 1.2014e-07, 9.4324e-06], + [ 2.7698e-06, 1.5432e-06, 2.3887e-05, ..., 6.5193e-09, + 2.9895e-07, 7.4208e-06]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0002, -0.0173, -0.0299, 0.0370, -0.0299, 0.0274, -0.0186, -0.0116, + 0.0110, -0.0268], device='cuda:0'), grad: tensor([-8.4117e-06, 2.9374e-06, 3.8557e-06, 9.1083e-07, 1.8766e-06, + -1.6661e-06, 8.6054e-06, -8.2493e-05, 4.1425e-05, 3.3021e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.95, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4994 re_mapping 0.0050 re_causal 0.0145 /// teacc 99.05 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.1237, -0.1371, -0.0989, ..., -0.1455, -0.0148, 0.0772], + [ 0.0637, 0.0565, 0.0498, ..., 0.1009, -0.1714, -0.0550], + [ 0.0543, 0.0229, -0.1457, ..., -0.0403, -0.2047, -0.0833], + ..., + [-0.0522, -0.1296, 0.0877, ..., -0.1174, -0.0457, 0.0012], + [ 0.0247, 0.0952, 0.1491, ..., -0.1284, -0.1365, -0.2083], + [-0.1542, 0.0178, -0.2513, ..., -0.1285, -0.1040, 0.1058]], + device='cuda:0'), grad: tensor([[-4.5635e-08, 3.3528e-08, 1.4901e-08, ..., 2.6077e-08, + -3.5949e-07, -1.5208e-06], + [-4.6566e-08, 1.7695e-08, 1.7509e-06, ..., -7.4506e-09, + 1.8254e-07, 1.0338e-06], + [ 2.2352e-08, 8.6613e-08, 1.6484e-07, ..., 6.7987e-08, + 8.2888e-08, 1.6391e-07], + ..., + [ 1.4156e-07, 1.5832e-08, -9.7379e-06, ..., 1.8626e-08, + 9.3132e-09, -4.8913e-06], + [ 6.4820e-07, 4.1910e-08, 1.2442e-06, ..., 4.8429e-08, + 9.4343e-07, 1.5786e-06], + [ 3.2596e-08, 2.2259e-07, 7.0967e-06, ..., 2.3097e-07, + 9.2201e-08, 3.6061e-06]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0004, -0.0168, -0.0306, 0.0372, -0.0295, 0.0275, -0.0186, -0.0115, + 0.0099, -0.0267], device='cuda:0'), grad: tensor([-2.7716e-06, 3.7923e-06, 7.4226e-07, -4.1053e-06, -3.5577e-07, + -5.1968e-06, 3.7141e-06, -1.6481e-05, 7.5176e-06, 1.3165e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.95, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4940 re_mapping 0.0045 re_causal 0.0141 /// teacc 99.04 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.1229, -0.1394, -0.0983, ..., -0.1480, -0.0160, 0.0773], + [ 0.0639, 0.0565, 0.0499, ..., 0.1008, -0.1715, -0.0551], + [ 0.0543, 0.0227, -0.1471, ..., -0.0409, -0.2053, -0.0835], + ..., + [-0.0527, -0.1299, 0.0878, ..., -0.1177, -0.0461, 0.0012], + [ 0.0251, 0.0954, 0.1497, ..., -0.1277, -0.1368, -0.2089], + [-0.1543, 0.0177, -0.2520, ..., -0.1292, -0.1045, 0.1060]], + device='cuda:0'), grad: tensor([[ 6.7055e-08, 1.1176e-08, 1.3039e-08, ..., 2.7940e-09, + -9.3132e-10, 1.2107e-08], + [-2.7008e-08, -9.9652e-08, -2.2631e-07, ..., -9.9652e-08, + 2.7940e-09, 1.0245e-08], + [-9.7509e-07, -9.3132e-09, 9.4064e-08, ..., 2.4214e-08, + 4.6566e-09, 1.6764e-08], + ..., + [ 2.8498e-07, 3.2596e-08, -1.2107e-08, ..., 2.3283e-08, + 9.3132e-10, 1.8626e-09], + [ 9.6858e-08, -2.8871e-08, -8.4750e-08, ..., 3.2596e-08, + 5.5879e-09, 2.2352e-08], + [ 3.9116e-08, 4.6566e-09, 4.6566e-08, ..., 1.6764e-08, + 7.4506e-09, -1.9930e-07]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0009, -0.0168, -0.0305, 0.0374, -0.0293, 0.0274, -0.0185, -0.0118, + 0.0101, -0.0269], device='cuda:0'), grad: tensor([ 2.0768e-07, 1.5646e-07, -2.8722e-06, 9.5367e-07, 1.9651e-07, + 2.1234e-07, 2.1141e-07, 7.0315e-07, 3.9954e-07, -1.5367e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 214.91, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4842 re_mapping 0.0046 re_causal 0.0139 /// teacc 98.99 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.1234, -0.1397, -0.0985, ..., -0.1481, -0.0160, 0.0775], + [ 0.0614, 0.0568, 0.0472, ..., 0.1009, -0.1717, -0.0552], + [ 0.0542, 0.0224, -0.1482, ..., -0.0411, -0.2058, -0.0838], + ..., + [-0.0496, -0.1301, 0.0907, ..., -0.1178, -0.0463, 0.0013], + [ 0.0241, 0.0952, 0.1496, ..., -0.1284, -0.1375, -0.2092], + [-0.1546, 0.0176, -0.2527, ..., -0.1293, -0.1047, 0.1062]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 5.5879e-09, 5.5879e-09, ..., 5.5879e-09, + 3.4366e-07, 4.4890e-07], + [-5.1409e-07, -2.5611e-07, -7.2829e-07, ..., -4.7777e-07, + 9.3132e-09, 1.1176e-08], + [ 5.8673e-08, 4.0978e-08, 8.1956e-08, ..., 5.5879e-08, + 2.7940e-08, 2.6077e-08], + ..., + [ 3.0920e-07, 1.4342e-07, 3.7905e-07, ..., 2.6729e-07, + 5.5879e-09, 2.3283e-08], + [ 8.5682e-08, 5.2154e-08, 7.1712e-08, ..., 6.1467e-08, + 2.9802e-08, 2.8871e-08], + [ 2.7940e-08, 5.6811e-08, 3.5390e-08, ..., 1.4715e-07, + 2.1327e-07, 1.4622e-07]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0008, -0.0194, -0.0307, 0.0377, -0.0295, 0.0274, -0.0184, -0.0087, + 0.0094, -0.0269], device='cuda:0'), grad: tensor([ 7.8604e-07, -1.1623e-06, 2.1048e-07, -1.4715e-07, -1.9092e-07, + -1.4063e-06, 1.5832e-07, 7.8510e-07, 3.2503e-07, 6.4354e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 214.94, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.4981 re_mapping 0.0048 re_causal 0.0147 /// teacc 98.96 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.1237, -0.1398, -0.0987, ..., -0.1481, -0.0158, 0.0779], + [ 0.0614, 0.0568, 0.0472, ..., 0.1009, -0.1721, -0.0552], + [ 0.0557, 0.0225, -0.1444, ..., -0.0414, -0.2061, -0.0835], + ..., + [-0.0504, -0.1310, 0.0905, ..., -0.1183, -0.0467, 0.0008], + [ 0.0245, 0.0953, 0.1503, ..., -0.1291, -0.1376, -0.2099], + [-0.1549, 0.0173, -0.2531, ..., -0.1301, -0.1053, 0.1063]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 3.7253e-09, 2.7940e-09, ..., 1.8626e-09, + -9.3132e-10, -3.7253e-08], + [-1.0710e-07, -1.1642e-07, -1.6205e-07, ..., -9.6858e-08, + 3.7253e-09, 2.7940e-09], + [ 1.8626e-09, 2.7008e-08, 2.8871e-08, ..., 1.4901e-08, + 6.5193e-09, -0.0000e+00], + ..., + [ 3.9116e-08, 2.7940e-08, 4.0978e-08, ..., 2.6077e-08, + 1.8626e-09, -9.3132e-10], + [-2.0489e-08, 6.1467e-08, -3.1665e-08, ..., 2.8871e-08, + 5.0291e-08, 3.6322e-08], + [ 1.1176e-08, 1.0245e-08, 1.5832e-08, ..., 3.0734e-08, + 2.1420e-08, 5.5879e-09]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0005, -0.0194, -0.0276, 0.0376, -0.0288, 0.0274, -0.0184, -0.0103, + 0.0094, -0.0272], device='cuda:0'), grad: tensor([-3.1665e-08, -2.5611e-07, 2.6077e-08, 2.3190e-07, -3.1665e-08, + -5.3737e-07, 2.8871e-07, 9.0338e-08, 1.3318e-07, 7.9162e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 215.00, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4743 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.04 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.1219, -0.1394, -0.0989, ..., -0.1482, -0.0158, 0.0779], + [ 0.0614, 0.0566, 0.0472, ..., 0.1013, -0.1721, -0.0550], + [ 0.0554, 0.0227, -0.1452, ..., -0.0423, -0.2071, -0.0847], + ..., + [-0.0504, -0.1313, 0.0905, ..., -0.1194, -0.0485, 0.0008], + [ 0.0242, 0.0953, 0.1505, ..., -0.1292, -0.1379, -0.2103], + [-0.1554, 0.0174, -0.2534, ..., -0.1306, -0.1055, 0.1066]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 1.5832e-08, 1.2107e-08, ..., 2.1420e-08, + 2.7008e-08, 1.9558e-08], + [ 2.2352e-08, 5.4017e-08, 9.1176e-07, ..., 9.4995e-08, + 1.3597e-07, 3.5670e-07], + [ 1.6671e-07, 4.8429e-08, 1.7788e-07, ..., 2.9802e-08, + 3.9116e-08, 4.3772e-08], + ..., + [ 1.7136e-07, 8.5682e-08, -1.9670e-06, ..., 8.6613e-08, + 8.7544e-08, -4.1816e-07], + [-1.1083e-07, -8.3819e-09, 5.6811e-08, ..., 2.4773e-07, + 3.5577e-07, 4.4145e-07], + [ 5.8673e-08, -2.6298e-04, 2.8033e-07, ..., -3.0899e-04, + -4.9782e-04, -4.8637e-04]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0001, -0.0195, -0.0279, 0.0377, -0.0289, 0.0273, -0.0183, -0.0103, + 0.0092, -0.0270], device='cuda:0'), grad: tensor([ 1.7602e-07, 2.3246e-06, 7.0408e-07, -7.2271e-07, 1.9970e-03, + 1.3523e-06, 1.0338e-06, -2.7716e-06, 1.8934e-06, -2.0027e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.53, cls_loss 0.0015 cls_loss_mapping 0.0036 cls_loss_causal 0.4815 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.03 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.1224, -0.1397, -0.0995, ..., -0.1484, -0.0139, 0.0796], + [ 0.0615, 0.0566, 0.0473, ..., 0.1020, -0.1723, -0.0562], + [ 0.0557, 0.0226, -0.1455, ..., -0.0427, -0.2081, -0.0853], + ..., + [-0.0505, -0.1320, 0.0905, ..., -0.1201, -0.0489, 0.0011], + [ 0.0240, 0.0957, 0.1504, ..., -0.1306, -0.1390, -0.2117], + [-0.1549, 0.0186, -0.2536, ..., -0.1292, -0.1038, 0.1080]], + device='cuda:0'), grad: tensor([[ 3.1013e-07, 7.2643e-08, 1.7229e-07, ..., 4.0047e-08, + 4.5635e-08, -1.8720e-07], + [ 1.7695e-08, -1.0617e-07, -5.7463e-07, ..., -2.5798e-07, + 1.6941e-06, 1.4957e-06], + [-1.5739e-06, 3.3528e-08, 1.1455e-07, ..., 3.5390e-08, + 3.8184e-08, 4.4703e-08], + ..., + [ 4.4703e-07, 1.2852e-07, 6.2734e-06, ..., 2.7288e-07, + 5.8673e-08, 1.9167e-06], + [-5.7090e-07, -2.2911e-07, -3.3062e-07, ..., 2.7940e-08, + 2.2165e-07, 2.9709e-07], + [ 1.2852e-07, 4.5635e-08, 9.8720e-08, ..., 2.1700e-07, + 1.8440e-07, 5.1223e-08]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0013, -0.0196, -0.0277, 0.0340, -0.0307, 0.0291, -0.0188, -0.0103, + 0.0087, -0.0254], device='cuda:0'), grad: tensor([ 7.6648e-07, 5.3979e-06, -2.9299e-06, -1.3605e-05, -5.5321e-07, + -9.4101e-06, 1.1586e-06, 1.8209e-05, -3.1665e-08, 1.0030e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 214.96, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4990 re_mapping 0.0046 re_causal 0.0143 /// teacc 98.98 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.1227, -0.1397, -0.0997, ..., -0.1483, -0.0133, 0.0801], + [ 0.0616, 0.0567, 0.0474, ..., 0.1021, -0.1731, -0.0563], + [ 0.0555, 0.0225, -0.1458, ..., -0.0435, -0.2087, -0.0857], + ..., + [-0.0506, -0.1329, 0.0905, ..., -0.1222, -0.0506, 0.0012], + [ 0.0237, 0.0954, 0.1504, ..., -0.1324, -0.1403, -0.2122], + [-0.1552, 0.0183, -0.2539, ..., -0.1303, -0.1044, 0.1080]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.7253e-09, 9.3132e-10, ..., -3.7253e-09, + -1.1083e-06, -2.1961e-06], + [-1.2107e-08, -7.4506e-09, -9.3132e-09, ..., -9.3132e-10, + 1.9558e-08, 2.0489e-08], + [-2.7940e-09, 4.6566e-09, 1.3039e-08, ..., 8.3819e-09, + 9.3132e-09, 1.4901e-08], + ..., + [ 9.3132e-09, 1.3970e-08, -2.7940e-08, ..., 1.0245e-08, + 2.7940e-09, 7.4506e-09], + [ 2.7940e-09, 1.3039e-08, 2.7940e-09, ..., 6.5193e-09, + 1.3039e-08, 2.7940e-08], + [ 9.3132e-10, -3.1665e-08, 9.3132e-09, ..., 4.9360e-08, + 6.3330e-08, -2.4214e-08]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0016, -0.0196, -0.0279, 0.0336, -0.0301, 0.0294, -0.0188, -0.0103, + 0.0082, -0.0258], device='cuda:0'), grad: tensor([-3.5502e-06, 3.7253e-08, 4.7497e-08, 1.8626e-09, -9.0338e-08, + 2.0396e-07, 3.2671e-06, 2.7940e-09, 7.8231e-08, 2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 215.02, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4997 re_mapping 0.0044 re_causal 0.0141 /// teacc 99.05 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.1235, -0.1399, -0.1001, ..., -0.1484, -0.0132, 0.0802], + [ 0.0622, 0.0580, 0.0475, ..., 0.1022, -0.1732, -0.0561], + [ 0.0545, 0.0205, -0.1475, ..., -0.0436, -0.2090, -0.0868], + ..., + [-0.0506, -0.1333, 0.0906, ..., -0.1224, -0.0503, 0.0010], + [ 0.0239, 0.0970, 0.1507, ..., -0.1328, -0.1392, -0.2123], + [-0.1557, 0.0179, -0.2543, ..., -0.1309, -0.1048, 0.1085]], + device='cuda:0'), grad: tensor([[ 2.8871e-08, 8.3819e-09, 1.8626e-08, ..., 3.7253e-09, + -2.0210e-07, -6.6590e-07], + [ 1.3504e-07, -1.3039e-08, 1.6224e-06, ..., -3.2596e-08, + 1.7695e-08, 1.9278e-06], + [ 1.2107e-07, 8.3819e-09, 2.4680e-07, ..., 1.8626e-09, + 1.3039e-08, 2.7660e-07], + ..., + [-3.2596e-08, 8.3819e-09, -3.7085e-06, ..., 9.3132e-09, + 1.3039e-08, -4.1723e-06], + [ 1.2293e-07, 4.6566e-09, 6.9849e-08, ..., 3.7253e-09, + 3.0734e-08, 1.4994e-07], + [ 1.4622e-07, 3.7253e-09, 1.5488e-06, ..., 3.7253e-09, + 5.8673e-08, 1.9595e-06]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0016, -0.0192, -0.0288, 0.0338, -0.0300, 0.0294, -0.0190, -0.0103, + 0.0088, -0.0258], device='cuda:0'), grad: tensor([-1.4957e-06, 5.1931e-06, 9.9372e-07, -4.4107e-06, 3.9488e-07, + -6.1467e-06, 9.8422e-06, -1.0453e-05, 7.1339e-07, 5.3383e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 214.84, cls_loss 0.0015 cls_loss_mapping 0.0039 cls_loss_causal 0.4975 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.03 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.1244, -0.1411, -0.1010, ..., -0.1496, -0.0135, 0.0805], + [ 0.0624, 0.0590, 0.0476, ..., 0.1031, -0.1733, -0.0561], + [ 0.0548, 0.0210, -0.1475, ..., -0.0438, -0.2093, -0.0848], + ..., + [-0.0508, -0.1361, 0.0906, ..., -0.1252, -0.0514, 0.0014], + [ 0.0231, 0.0968, 0.1506, ..., -0.1334, -0.1397, -0.2139], + [-0.1575, 0.0169, -0.2572, ..., -0.1335, -0.1055, 0.1075]], + device='cuda:0'), grad: tensor([[ 4.4890e-07, 3.8836e-07, 3.4459e-08, ..., 7.7300e-08, + 2.5146e-08, -4.9360e-08], + [ 4.3493e-07, 2.2594e-06, 6.7800e-07, ..., 1.3569e-06, + 4.4610e-07, 1.8347e-07], + [-4.4331e-07, 2.1514e-07, 1.8533e-07, ..., 8.1025e-08, + 2.6077e-08, 1.0524e-07], + ..., + [-2.5146e-07, 7.6368e-08, -2.0713e-06, ..., 4.6566e-09, + 2.7940e-09, -1.2033e-06], + [ 7.8604e-07, 5.1316e-07, 7.7300e-08, ..., 3.2596e-08, + 1.5832e-08, 2.2724e-07], + [ 2.7474e-07, -2.1514e-07, 9.3970e-07, ..., 6.5193e-09, + -4.0978e-08, -3.8370e-07]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0015, -0.0190, -0.0284, 0.0339, -0.0293, 0.0295, -0.0191, -0.0104, + 0.0080, -0.0270], device='cuda:0'), grad: tensor([ 1.3579e-06, 5.5432e-06, -4.5076e-07, -4.4554e-06, 1.4361e-06, + 1.5069e-06, -4.1462e-06, -4.8243e-06, 2.8946e-06, 1.1250e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 215.01, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.5128 re_mapping 0.0042 re_causal 0.0134 /// teacc 98.92 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.1248, -0.1413, -0.1012, ..., -0.1492, -0.0129, 0.0810], + [ 0.0625, 0.0590, 0.0476, ..., 0.1034, -0.1734, -0.0563], + [ 0.0549, 0.0207, -0.1478, ..., -0.0442, -0.2094, -0.0848], + ..., + [-0.0508, -0.1366, 0.0906, ..., -0.1256, -0.0517, 0.0015], + [ 0.0228, 0.0963, 0.1507, ..., -0.1337, -0.1427, -0.2169], + [-0.1573, 0.0170, -0.2576, ..., -0.1339, -0.1058, 0.1076]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 1.6764e-08, 9.3132e-09, ..., 6.5193e-09, + 1.8626e-09, -1.7695e-08], + [-2.2538e-07, -1.8347e-07, -4.0419e-07, ..., -1.5832e-07, + 4.6566e-09, 8.3819e-09], + [-5.3085e-08, 5.1223e-08, 5.0291e-08, ..., 2.2352e-08, + 3.7253e-09, 2.2352e-08], + ..., + [ 1.3597e-07, 7.9162e-08, 1.9278e-07, ..., 7.6368e-08, + 9.3132e-10, -1.8626e-09], + [ 4.9360e-08, 6.5193e-08, 6.0536e-08, ..., 3.0734e-08, + 1.7695e-08, 3.6322e-08], + [ 1.5832e-08, -5.5134e-07, 2.2352e-08, ..., 2.3283e-08, + 1.5832e-08, -8.0373e-07]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0019, -0.0190, -0.0285, 0.0339, -0.0293, 0.0299, -0.0192, -0.0103, + 0.0060, -0.0271], device='cuda:0'), grad: tensor([ 3.7253e-08, -5.9605e-07, -9.0338e-08, 2.2352e-07, 1.3225e-07, + 2.0433e-06, 2.0489e-08, 4.1258e-07, 2.8219e-07, -2.4457e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 214.65, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4906 re_mapping 0.0045 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.1250, -0.1415, -0.1015, ..., -0.1492, -0.0129, 0.0812], + [ 0.0632, 0.0605, 0.0475, ..., 0.1032, -0.1738, -0.0565], + [ 0.0538, 0.0182, -0.1480, ..., -0.0443, -0.2098, -0.0859], + ..., + [-0.0507, -0.1349, 0.0910, ..., -0.1257, -0.0525, 0.0032], + [ 0.0223, 0.0964, 0.1510, ..., -0.1333, -0.1431, -0.2174], + [-0.1580, 0.0180, -0.2602, ..., -0.1342, -0.1059, 0.1075]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, -4.9360e-08, 8.3819e-09, ..., 0.0000e+00, + -4.1258e-07, -1.1083e-06], + [-3.9488e-07, -1.7416e-07, -7.2643e-08, ..., -1.1176e-08, + 9.3132e-09, 7.5437e-08], + [ 6.6031e-07, 2.2911e-07, 2.7195e-07, ..., 9.3132e-10, + 2.7940e-09, 7.2643e-08], + ..., + [ 9.4995e-08, 1.6764e-08, -1.0151e-07, ..., 1.1176e-08, + 2.3283e-08, -1.2480e-07], + [-1.3746e-06, -4.7125e-07, -7.5996e-07, ..., 1.8626e-09, + 3.7253e-08, 2.2352e-08], + [ 7.8231e-08, 3.0734e-08, 5.4017e-08, ..., 1.3411e-07, + 4.0419e-07, 1.4901e-07]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0019, -0.0187, -0.0294, 0.0339, -0.0298, 0.0299, -0.0192, -0.0099, + 0.0057, -0.0272], device='cuda:0'), grad: tensor([-1.3420e-06, -4.0140e-07, 1.1949e-06, -1.2703e-05, -7.4506e-07, + 1.3553e-05, 1.2452e-06, 2.5146e-08, -2.0117e-06, 1.1707e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 214.99, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5169 re_mapping 0.0044 re_causal 0.0137 /// teacc 99.02 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.1257, -0.1414, -0.1024, ..., -0.1492, -0.0151, 0.0781], + [ 0.0633, 0.0609, 0.0475, ..., 0.1037, -0.1738, -0.0564], + [ 0.0540, 0.0180, -0.1481, ..., -0.0447, -0.2101, -0.0862], + ..., + [-0.0508, -0.1356, 0.0910, ..., -0.1264, -0.0527, 0.0033], + [ 0.0228, 0.0989, 0.1528, ..., -0.1335, -0.1426, -0.2175], + [-0.1589, 0.0177, -0.2610, ..., -0.1345, -0.1039, 0.1107]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.3283e-08, 1.8626e-09, ..., 3.7253e-08, + 2.5146e-08, -4.3772e-08], + [-3.8184e-08, -1.2107e-08, -5.7742e-08, ..., -1.3039e-08, + 1.6764e-08, 1.8626e-08], + [-1.8626e-09, 1.7323e-07, 3.7253e-09, ..., 6.7428e-07, + 3.0454e-07, 4.6566e-09], + ..., + [ 1.4901e-08, 1.1735e-07, 5.1223e-08, ..., 2.6077e-08, + 1.4901e-08, 1.7323e-07], + [ 8.3819e-09, 8.7544e-08, 2.1420e-08, ..., 2.6077e-08, + 7.6368e-08, 5.4948e-08], + [ 6.5193e-09, -2.0396e-07, -1.1083e-07, ..., 1.8720e-07, + 1.1362e-07, -5.0478e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0012, -0.0185, -0.0294, 0.0341, -0.0299, 0.0298, -0.0194, -0.0099, + 0.0068, -0.0242], device='cuda:0'), grad: tensor([ 5.2154e-08, -1.1176e-08, 1.3066e-06, -5.4948e-08, -1.4352e-06, + -9.0338e-08, -1.1362e-07, 4.2561e-07, 3.2224e-07, -4.0606e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 215.06, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5192 re_mapping 0.0043 re_causal 0.0133 /// teacc 99.02 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.1258, -0.1415, -0.1011, ..., -0.1494, -0.0150, 0.0781], + [ 0.0644, 0.0615, 0.0490, ..., 0.1046, -0.1742, -0.0544], + [ 0.0539, 0.0179, -0.1489, ..., -0.0451, -0.2106, -0.0864], + ..., + [-0.0519, -0.1374, 0.0897, ..., -0.1278, -0.0523, 0.0016], + [ 0.0227, 0.0986, 0.1531, ..., -0.1338, -0.1434, -0.2190], + [-0.1585, 0.0180, -0.2614, ..., -0.1350, -0.1042, 0.1107]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 5.5879e-09, 1.7695e-08, ..., 2.7940e-09, + 1.8626e-09, 9.3132e-10], + [-2.1979e-07, 4.8336e-07, 9.3430e-06, ..., 1.7136e-07, + 9.3132e-10, 2.4587e-06], + [ 2.7940e-09, 1.1176e-08, 5.2154e-08, ..., 6.5193e-09, + 9.3132e-10, 1.3970e-08], + ..., + [ 1.0431e-07, -6.3144e-07, -1.0684e-05, ..., -2.5798e-07, + 9.3132e-10, -2.7884e-06], + [ 1.6764e-08, 9.3132e-09, 5.3085e-08, ..., 6.5193e-09, + 9.3132e-09, 1.8626e-08], + [ 5.2154e-08, 8.9407e-08, 9.6858e-07, ..., 5.7742e-08, + 7.4506e-09, 2.3097e-07]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0011, -0.0170, -0.0296, 0.0347, -0.0298, 0.0294, -0.0194, -0.0113, + 0.0063, -0.0242], device='cuda:0'), grad: tensor([ 4.5635e-08, 2.1964e-05, 1.0524e-07, 1.5646e-07, 2.9523e-07, + 3.5390e-08, 5.7742e-08, -2.5064e-05, 1.5553e-07, 2.2836e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.87, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4651 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.04 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.1276, -0.1428, -0.1026, ..., -0.1497, -0.0150, 0.0781], + [ 0.0646, 0.0616, 0.0491, ..., 0.1048, -0.1745, -0.0546], + [ 0.0538, 0.0177, -0.1493, ..., -0.0455, -0.2108, -0.0865], + ..., + [-0.0520, -0.1381, 0.0897, ..., -0.1279, -0.0526, 0.0015], + [ 0.0232, 0.0987, 0.1538, ..., -0.1341, -0.1438, -0.2197], + [-0.1583, 0.0178, -0.2613, ..., -0.1357, -0.1039, 0.1109]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, 3.8184e-08, 4.5635e-08, ..., 2.5146e-08, + -8.6613e-08, -3.2503e-07], + [-1.2573e-07, -6.6124e-08, 4.6752e-07, ..., -4.0978e-08, + 6.1467e-08, 3.9022e-07], + [-1.2489e-06, 8.4750e-08, 1.3970e-07, ..., 1.9558e-08, + 1.1176e-08, 2.7940e-08], + ..., + [ 1.2200e-06, 7.5437e-08, -8.3540e-07, ..., 5.2154e-08, + 3.3528e-08, -5.0757e-07], + [-5.5321e-07, -2.7567e-06, -5.8152e-06, ..., 1.9558e-08, + 2.3283e-08, 2.8871e-08], + [ 3.2596e-08, 6.0908e-07, 3.1572e-07, ..., 6.4727e-07, + 4.3958e-07, 1.8440e-07]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0012, -0.0169, -0.0297, 0.0347, -0.0302, 0.0294, -0.0194, -0.0114, + 0.0066, -0.0241], device='cuda:0'), grad: tensor([-2.9989e-07, 1.7220e-06, -6.5565e-06, 1.0014e-05, -2.7046e-06, + 1.5991e-06, 1.8226e-06, 3.9525e-06, -1.2904e-05, 3.2894e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 214.87, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.5128 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.1278, -0.1430, -0.1029, ..., -0.1499, -0.0147, 0.0782], + [ 0.0647, 0.0615, 0.0491, ..., 0.1045, -0.1750, -0.0547], + [ 0.0538, 0.0177, -0.1506, ..., -0.0461, -0.2112, -0.0867], + ..., + [-0.0520, -0.1383, 0.0898, ..., -0.1284, -0.0544, 0.0015], + [ 0.0235, 0.0994, 0.1542, ..., -0.1343, -0.1437, -0.2200], + [-0.1588, 0.0176, -0.2617, ..., -0.1369, -0.1043, 0.1109]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 2.6077e-08, 9.3132e-10, ..., 2.4214e-08, + 4.0978e-08, 6.5193e-09], + [-1.1176e-08, 2.8498e-07, -3.5390e-08, ..., 2.5425e-07, + 4.8243e-07, 1.3039e-08], + [-2.0768e-07, -3.0734e-08, 8.3819e-09, ..., 1.0245e-08, + 1.1176e-08, -2.8871e-08], + ..., + [ 4.7497e-08, 6.5193e-09, 9.3132e-10, ..., 7.4506e-09, + 0.0000e+00, 2.5146e-08], + [ 3.3155e-07, 4.2841e-08, 5.4948e-08, ..., 1.3039e-08, + 1.7695e-08, 1.6764e-08], + [ 1.0245e-08, -1.7695e-08, 6.5193e-09, ..., 4.6566e-09, + 2.7940e-09, -1.6578e-07]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0011, -0.0170, -0.0302, 0.0345, -0.0297, 0.0295, -0.0196, -0.0111, + 0.0069, -0.0242], device='cuda:0'), grad: tensor([ 1.3039e-07, 1.1064e-06, -4.1164e-07, -9.0711e-07, 2.1700e-07, + 3.8650e-07, -1.3569e-06, 1.2759e-07, 9.0618e-07, -1.9372e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 215.18, cls_loss 0.0015 cls_loss_mapping 0.0027 cls_loss_causal 0.5035 re_mapping 0.0045 re_causal 0.0134 /// teacc 98.96 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.1292, -0.1445, -0.1040, ..., -0.1499, -0.0144, 0.0784], + [ 0.0647, 0.0613, 0.0491, ..., 0.1045, -0.1751, -0.0556], + [ 0.0537, 0.0178, -0.1507, ..., -0.0464, -0.2118, -0.0866], + ..., + [-0.0521, -0.1385, 0.0898, ..., -0.1287, -0.0557, 0.0014], + [ 0.0231, 0.1008, 0.1546, ..., -0.1348, -0.1435, -0.2204], + [-0.1594, 0.0189, -0.2623, ..., -0.1371, -0.1044, 0.1113]], + device='cuda:0'), grad: tensor([[ 6.0629e-07, 1.9558e-08, 1.8626e-09, ..., 1.1176e-08, + -1.1176e-08, -1.5087e-07], + [ 7.3276e-06, 3.7253e-09, -6.0536e-08, ..., -3.5390e-08, + 3.9116e-08, 4.6566e-09], + [ 4.0859e-05, -5.5879e-09, 1.5832e-08, ..., 1.6764e-08, + 7.4506e-09, 9.3132e-09], + ..., + [ 3.4701e-06, 1.3039e-08, 1.1176e-08, ..., 2.1420e-08, + 8.3819e-09, 1.1176e-08], + [ 4.2003e-07, -8.1956e-08, -6.6124e-08, ..., 2.2352e-08, + 2.6077e-08, 7.4506e-09], + [ 2.5164e-06, 1.1828e-07, 5.0291e-08, ..., 4.3027e-07, + 4.3958e-07, 9.0338e-08]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0011, -0.0172, -0.0304, 0.0356, -0.0307, 0.0286, -0.0191, -0.0111, + 0.0075, -0.0239], device='cuda:0'), grad: tensor([ 9.2853e-07, 1.3188e-05, 7.3195e-05, -1.1092e-04, -8.6706e-07, + 1.1936e-05, -3.7160e-07, 6.2585e-06, 5.9325e-07, 6.1356e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 215.39, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.5153 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.08 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.1296, -0.1463, -0.1044, ..., -0.1501, -0.0152, 0.0783], + [ 0.0647, 0.0622, 0.0493, ..., 0.1053, -0.1748, -0.0552], + [ 0.0537, 0.0179, -0.1512, ..., -0.0468, -0.2119, -0.0867], + ..., + [-0.0522, -0.1387, 0.0899, ..., -0.1280, -0.0545, 0.0016], + [ 0.0233, 0.1010, 0.1554, ..., -0.1355, -0.1440, -0.2207], + [-0.1604, 0.0186, -0.2633, ..., -0.1375, -0.1049, 0.1113]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.8626e-08, 1.0245e-08, ..., 0.0000e+00, + 8.3819e-09, 2.7940e-09], + [ 1.4156e-07, 7.3574e-08, 0.0000e+00, ..., -1.6764e-08, + 1.1176e-08, 1.7695e-08], + [-2.9989e-07, -1.0803e-07, 2.6077e-08, ..., 1.8626e-09, + 2.7940e-09, 4.6566e-09], + ..., + [ 4.3772e-08, 3.0734e-08, -1.8626e-09, ..., 9.3132e-09, + 2.7940e-09, 1.1176e-08], + [-1.2666e-07, 1.1083e-07, -3.1479e-07, ..., 8.3819e-09, + 1.7416e-07, 1.3690e-07], + [ 1.1455e-07, -1.5739e-07, 1.7881e-07, ..., 1.0245e-07, + 8.1956e-08, -7.2923e-07]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0013, -0.0170, -0.0304, 0.0360, -0.0307, 0.0285, -0.0190, -0.0111, + 0.0076, -0.0240], device='cuda:0'), grad: tensor([ 7.0781e-08, 3.7905e-07, -5.5134e-07, 3.2596e-07, 8.9221e-07, + 1.3132e-07, -5.4855e-07, 1.1362e-07, -1.1642e-07, -7.0222e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 215.35, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.4737 re_mapping 0.0044 re_causal 0.0133 /// teacc 98.96 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.1298, -0.1476, -0.1045, ..., -0.1502, -0.0157, 0.0783], + [ 0.0648, 0.0623, 0.0492, ..., 0.1056, -0.1749, -0.0555], + [ 0.0542, 0.0192, -0.1520, ..., -0.0468, -0.2121, -0.0867], + ..., + [-0.0522, -0.1389, 0.0901, ..., -0.1281, -0.0539, 0.0024], + [ 0.0220, 0.1001, 0.1556, ..., -0.1358, -0.1443, -0.2211], + [-0.1616, 0.0182, -0.2648, ..., -0.1381, -0.1051, 0.1112]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -2.4214e-08, -3.4459e-08], + [ 2.7940e-09, 9.3132e-10, 2.2538e-07, ..., 0.0000e+00, + 1.8626e-09, 1.2107e-08], + [-1.1176e-08, 2.7940e-09, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 2.9802e-08], + ..., + [ 4.6566e-09, 9.3132e-09, -2.4773e-07, ..., 1.8626e-09, + -0.0000e+00, 2.6263e-07], + [ 9.3132e-10, 9.3132e-10, 1.3970e-08, ..., 0.0000e+00, + 7.4506e-09, 1.2107e-08], + [ 0.0000e+00, -3.5390e-08, 4.6566e-09, ..., 7.4506e-09, + 1.4901e-08, -4.0233e-07]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0014, -0.0171, -0.0299, 0.0361, -0.0300, 0.0284, -0.0189, -0.0106, + 0.0065, -0.0244], device='cuda:0'), grad: tensor([-5.4017e-08, 2.7753e-07, 2.6077e-08, 3.1665e-08, 1.2387e-07, + -5.9605e-08, 6.8918e-08, 9.5926e-08, 4.1910e-08, -5.4762e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 215.02, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4742 re_mapping 0.0048 re_causal 0.0138 /// teacc 99.07 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.1261, -0.1474, -0.1021, ..., -0.1503, -0.0154, 0.0787], + [ 0.0647, 0.0623, 0.0493, ..., 0.1059, -0.1751, -0.0558], + [ 0.0545, 0.0193, -0.1523, ..., -0.0470, -0.2131, -0.0875], + ..., + [-0.0522, -0.1391, 0.0901, ..., -0.1283, -0.0543, 0.0027], + [ 0.0207, 0.1006, 0.1556, ..., -0.1365, -0.1446, -0.2243], + [-0.1627, 0.0181, -0.2656, ..., -0.1384, -0.1054, 0.1113]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.3819e-09, 9.3132e-10, ..., 3.7253e-09, + 1.0245e-08, -3.8184e-08], + [-1.0245e-08, 2.7940e-08, -9.3132e-09, ..., 2.7940e-08, + 1.5646e-07, 1.1176e-08], + [ 9.3132e-10, 6.5193e-09, 5.5879e-09, ..., 6.5193e-09, + 1.0245e-08, 4.6566e-09], + ..., + [ 5.5879e-09, 3.0734e-08, 5.5879e-09, ..., 1.1176e-08, + 2.7008e-08, 4.9360e-08], + [-1.6764e-08, 3.8836e-07, -3.6322e-08, ..., 4.9453e-07, + 2.2538e-06, 4.6566e-08], + [ 1.1176e-08, -2.5984e-07, -2.5146e-08, ..., 3.9488e-07, + 2.6077e-08, -7.8883e-07]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0007, -0.0172, -0.0298, 0.0360, -0.0300, 0.0286, -0.0194, -0.0104, + 0.0048, -0.0245], device='cuda:0'), grad: tensor([-1.1176e-08, 3.4645e-07, 5.2154e-08, 4.5747e-06, -5.2806e-07, + -1.1533e-05, 2.8331e-06, 1.5087e-07, 4.8056e-06, -7.1898e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 215.16, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4999 re_mapping 0.0043 re_causal 0.0141 /// teacc 98.95 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.1262, -0.1476, -0.1021, ..., -0.1504, -0.0154, 0.0787], + [ 0.0648, 0.0623, 0.0493, ..., 0.1061, -0.1752, -0.0558], + [ 0.0545, 0.0193, -0.1524, ..., -0.0474, -0.2137, -0.0875], + ..., + [-0.0523, -0.1393, 0.0901, ..., -0.1286, -0.0543, 0.0027], + [ 0.0207, 0.1006, 0.1560, ..., -0.1372, -0.1451, -0.2245], + [-0.1629, 0.0178, -0.2662, ..., -0.1393, -0.1058, 0.1113]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 1.6158e-07, 7.7765e-08, ..., 1.0524e-07, + 9.4995e-08, -1.2107e-08], + [-4.6100e-08, 2.4680e-08, -3.2596e-08, ..., -2.9337e-08, + 4.1910e-08, 1.0245e-08], + [ 2.0489e-08, 8.8941e-08, 5.1223e-08, ..., 6.4261e-08, + 5.0291e-08, 3.7253e-09], + ..., + [ 3.0268e-08, 2.8871e-08, 1.9092e-08, ..., 2.4680e-08, + 6.5193e-09, 5.6811e-08], + [ 1.7229e-08, 2.3469e-06, 1.1008e-06, ..., 1.5600e-06, + 1.4342e-06, 3.3528e-08], + [ 2.7008e-08, -2.7008e-08, 2.2352e-08, ..., 1.3504e-08, + 8.8476e-09, -1.3737e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0007, -0.0171, -0.0297, 0.0361, -0.0295, 0.0285, -0.0195, -0.0106, + 0.0048, -0.0246], device='cuda:0'), grad: tensor([ 4.9407e-07, 9.4064e-08, 3.2457e-07, -1.9092e-08, 5.4017e-07, + 3.9535e-07, -9.3207e-06, 2.0163e-07, 7.4394e-06, -1.1828e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 214.86, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.4712 re_mapping 0.0045 re_causal 0.0135 /// teacc 98.98 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.1262, -0.1478, -0.1022, ..., -0.1505, -0.0154, 0.0788], + [ 0.0648, 0.0623, 0.0492, ..., 0.1064, -0.1753, -0.0562], + [ 0.0579, 0.0193, -0.1481, ..., -0.0475, -0.2139, -0.0872], + ..., + [-0.0550, -0.1395, 0.0878, ..., -0.1289, -0.0543, 0.0027], + [ 0.0209, 0.1032, 0.1584, ..., -0.1376, -0.1451, -0.2250], + [-0.1644, 0.0166, -0.2690, ..., -0.1404, -0.1066, 0.1113]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 8.3819e-09, 7.4506e-09, ..., 9.3132e-10, + 1.6764e-08, 3.2596e-09], + [-2.5611e-08, -2.2352e-08, -4.6566e-08, ..., -3.0734e-08, + 2.7940e-09, 1.2107e-08], + [ 4.6566e-09, 5.1223e-09, 1.2573e-08, ..., 4.6566e-09, + 4.6566e-10, 2.3283e-09], + ..., + [ 1.3039e-08, 1.6764e-08, 2.7940e-09, ..., 1.3039e-08, + 9.3132e-10, 2.7940e-08], + [-1.8626e-08, -1.0245e-08, -1.4901e-08, ..., 5.1223e-09, + 1.3970e-08, 4.9360e-08], + [ 9.3132e-09, -3.3528e-08, 9.3132e-09, ..., 1.7229e-08, + 1.0710e-08, -4.3912e-07]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0006, -0.0174, -0.0257, 0.0358, -0.0291, 0.0287, -0.0197, -0.0141, + 0.0067, -0.0252], device='cuda:0'), grad: tensor([ 5.3085e-08, -6.7521e-08, 3.7719e-08, 1.0664e-07, 7.1153e-07, + 1.0245e-07, -5.9139e-08, 7.7300e-08, -7.6834e-08, -8.7311e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 214.53, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.5074 re_mapping 0.0043 re_causal 0.0130 /// teacc 98.97 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.1263, -0.1478, -0.1023, ..., -0.1506, -0.0154, 0.0788], + [ 0.0647, 0.0624, 0.0492, ..., 0.1066, -0.1756, -0.0565], + [ 0.0581, 0.0193, -0.1480, ..., -0.0477, -0.2143, -0.0874], + ..., + [-0.0551, -0.1399, 0.0878, ..., -0.1288, -0.0544, 0.0027], + [ 0.0208, 0.1032, 0.1590, ..., -0.1379, -0.1457, -0.2256], + [-0.1647, 0.0167, -0.2693, ..., -0.1406, -0.1067, 0.1114]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 1.2107e-08, 6.5193e-09, ..., 2.7940e-09, + 4.4983e-07, -8.9593e-07], + [-3.2596e-08, -6.0536e-09, 1.5367e-08, ..., -4.1910e-08, + 3.1199e-07, 5.5600e-07], + [-1.3448e-06, 1.1642e-08, -3.2131e-08, ..., 1.3504e-08, + 6.9384e-08, 3.1292e-07], + ..., + [ 3.7719e-08, 1.7229e-08, -3.0128e-07, ..., 2.7008e-08, + 1.4203e-07, 2.4540e-07], + [ 1.1967e-06, -2.0210e-07, -1.2573e-08, ..., -5.0291e-08, + 7.2969e-07, 8.8243e-07], + [ 9.7789e-09, 7.9628e-08, 4.9360e-08, ..., 2.0582e-07, + 1.5208e-06, 2.1271e-06]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0006, -0.0175, -0.0256, 0.0356, -0.0292, 0.0292, -0.0200, -0.0141, + 0.0066, -0.0252], device='cuda:0'), grad: tensor([-5.5274e-07, 1.7453e-06, -1.5683e-06, 7.0715e-04, -3.7532e-07, + -7.2384e-04, 4.8988e-06, 1.7835e-07, 4.9472e-06, 7.3910e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 215.10, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4510 re_mapping 0.0043 re_causal 0.0131 /// teacc 98.96 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.1260, -0.1479, -0.1019, ..., -0.1504, -0.0152, 0.0789], + [ 0.0648, 0.0624, 0.0492, ..., 0.1064, -0.1759, -0.0567], + [ 0.0581, 0.0194, -0.1482, ..., -0.0474, -0.2146, -0.0878], + ..., + [-0.0550, -0.1401, 0.0880, ..., -0.1290, -0.0545, 0.0028], + [ 0.0207, 0.1029, 0.1592, ..., -0.1385, -0.1464, -0.2261], + [-0.1648, 0.0166, -0.2695, ..., -0.1409, -0.1069, 0.1114]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.9162e-09, 4.6566e-09, ..., 9.3132e-10, + 9.7789e-09, 9.3132e-09], + [-6.9849e-09, 2.5611e-08, 4.1910e-08, ..., -2.0955e-08, + 1.6298e-08, 1.4435e-08], + [-2.8871e-08, 2.3283e-09, 8.4285e-08, ..., 4.1910e-09, + 3.2596e-09, 6.5193e-09], + ..., + [ 6.9849e-09, 6.5193e-09, -2.0862e-07, ..., 6.9849e-09, + 1.8626e-09, -3.7719e-08], + [-2.2817e-08, -5.0757e-08, -3.4925e-08, ..., 5.1223e-09, + 4.7963e-08, 1.3039e-08], + [ 5.1223e-09, -5.5879e-09, 4.8894e-08, ..., 2.7940e-09, + 9.3132e-09, -2.8405e-08]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0005, -0.0176, -0.0256, 0.0346, -0.0291, 0.0297, -0.0201, -0.0141, + 0.0062, -0.0252], device='cuda:0'), grad: tensor([ 4.9360e-08, 1.4296e-07, 6.3796e-08, 1.0310e-06, 1.0245e-07, + -4.8848e-07, -5.2573e-07, -3.9255e-07, -2.6077e-08, 6.5658e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 215.30, cls_loss 0.0012 cls_loss_mapping 0.0029 cls_loss_causal 0.5078 re_mapping 0.0042 re_causal 0.0130 /// teacc 98.96 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.1262, -0.1481, -0.1021, ..., -0.1510, -0.0152, 0.0790], + [ 0.0646, 0.0621, 0.0491, ..., 0.1059, -0.1763, -0.0572], + [ 0.0582, 0.0199, -0.1482, ..., -0.0464, -0.2150, -0.0877], + ..., + [-0.0547, -0.1384, 0.0886, ..., -0.1293, -0.0542, 0.0030], + [ 0.0190, 0.1014, 0.1561, ..., -0.1392, -0.1479, -0.2272], + [-0.1653, 0.0164, -0.2700, ..., -0.1415, -0.1076, 0.1112]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + 1.7695e-08, 1.1642e-08], + [-2.0489e-08, -9.7789e-09, 5.7276e-08, ..., -2.2352e-08, + 7.9162e-09, 5.1223e-09], + [-2.5611e-08, 3.7253e-09, 1.4435e-08, ..., 6.0536e-09, + 2.3283e-09, 9.3132e-10], + ..., + [ 9.7789e-09, 4.1910e-09, -8.8941e-08, ..., 7.9162e-09, + 1.4435e-08, 1.1176e-08], + [-5.1223e-08, -2.6077e-08, -6.1467e-08, ..., 2.3283e-09, + 6.2864e-08, 5.0291e-08], + [ 4.1910e-09, -1.8626e-09, 1.1176e-08, ..., 1.1642e-08, + 1.2573e-08, -7.4506e-09]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0004, -0.0179, -0.0256, 0.0350, -0.0289, 0.0298, -0.0198, -0.0137, + 0.0034, -0.0255], device='cuda:0'), grad: tensor([ 5.1223e-08, 1.9884e-07, -1.6298e-08, 2.9290e-07, -2.7940e-09, + -5.3458e-07, 1.8394e-07, -2.1327e-07, -2.2817e-08, 5.2620e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 215.00, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4720 re_mapping 0.0044 re_causal 0.0125 /// teacc 99.03 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.1265, -0.1495, -0.1023, ..., -0.1531, -0.0176, 0.0787], + [ 0.0648, 0.0618, 0.0490, ..., 0.1068, -0.1767, -0.0581], + [ 0.0577, 0.0200, -0.1486, ..., -0.0473, -0.2155, -0.0882], + ..., + [-0.0544, -0.1381, 0.0890, ..., -0.1308, -0.0544, 0.0037], + [ 0.0199, 0.1033, 0.1570, ..., -0.1367, -0.1487, -0.2275], + [-0.1658, 0.0164, -0.2706, ..., -0.1419, -0.1077, 0.1113]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.4435e-08, 2.3749e-08, ..., 8.8476e-09, + 6.0536e-09, 1.2107e-08], + [-6.7521e-08, -2.5611e-08, 7.6834e-08, ..., 3.5390e-07, + 6.0210e-07, 3.1758e-07], + [-5.9605e-08, 2.0023e-08, 3.7719e-08, ..., 2.5611e-08, + 3.7253e-09, 6.5193e-09], + ..., + [ 1.1874e-07, 4.9826e-08, 2.9802e-08, ..., 7.4971e-08, + 1.0245e-08, 1.3039e-08], + [-4.1910e-09, -2.4401e-07, -4.8941e-07, ..., 9.8720e-08, + 1.0850e-07, -1.5274e-07], + [ 1.8161e-08, 1.2657e-06, 3.8557e-07, ..., 1.5749e-06, + 4.5495e-07, 1.4249e-07]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0009, -0.0181, -0.0259, 0.0326, -0.0290, 0.0316, -0.0199, -0.0134, + 0.0041, -0.0255], device='cuda:0'), grad: tensor([ 1.0151e-07, 1.2917e-06, -8.1956e-08, -1.8906e-07, -6.2138e-06, + -2.9802e-06, 1.7760e-06, 3.5204e-07, -8.2701e-07, 6.7949e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 215.11, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.4673 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.09 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.1269, -0.1496, -0.1026, ..., -0.1533, -0.0175, 0.0788], + [ 0.0645, 0.0625, 0.0493, ..., 0.1063, -0.1776, -0.0586], + [ 0.0588, 0.0204, -0.1486, ..., -0.0451, -0.2169, -0.0885], + ..., + [-0.0547, -0.1391, 0.0889, ..., -0.1313, -0.0490, 0.0048], + [ 0.0187, 0.1031, 0.1568, ..., -0.1374, -0.1491, -0.2279], + [-0.1662, 0.0160, -0.2730, ..., -0.1424, -0.1086, 0.1112]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 5.4110e-07, 1.8626e-09, ..., 9.2667e-08, + 6.6729e-07, -2.7474e-08], + [-4.4703e-08, 3.2596e-09, -3.7253e-08, ..., -3.9581e-08, + 3.3528e-08, 1.3830e-07], + [ 1.3970e-09, 1.7695e-08, 2.7474e-08, ..., -2.7940e-08, + 2.5611e-08, 1.4901e-08], + ..., + [ 5.0291e-08, 3.3993e-08, -1.3271e-07, ..., 3.3993e-08, + 8.8476e-09, -4.4703e-08], + [-6.8452e-08, 5.7463e-07, -8.1956e-08, ..., 1.0571e-07, + 7.6462e-07, 2.6077e-08], + [ 1.6764e-08, -1.1642e-08, 9.2201e-08, ..., 1.8626e-08, + 6.8918e-08, -5.1642e-07]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0009, -0.0183, -0.0257, 0.0329, -0.0290, 0.0314, -0.0199, -0.0133, + 0.0035, -0.0259], device='cuda:0'), grad: tensor([ 2.2110e-06, 3.2689e-07, 1.5134e-07, -1.9595e-06, 9.4343e-07, + 6.2361e-06, -9.4771e-06, -1.8999e-07, 2.2985e-06, -5.6718e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 215.07, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4846 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.10 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.1270, -0.1497, -0.1027, ..., -0.1537, -0.0177, 0.0787], + [ 0.0648, 0.0632, 0.0495, ..., 0.1074, -0.1778, -0.0588], + [ 0.0586, 0.0201, -0.1487, ..., -0.0457, -0.2186, -0.0892], + ..., + [-0.0549, -0.1403, 0.0888, ..., -0.1340, -0.0495, 0.0044], + [ 0.0186, 0.1030, 0.1571, ..., -0.1377, -0.1500, -0.2287], + [-0.1663, 0.0158, -0.2735, ..., -0.1437, -0.1090, 0.1114]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + -1.9614e-06, -3.2578e-06], + [-2.2352e-08, -1.9092e-08, -4.0978e-08, ..., -2.0489e-08, + 6.5658e-08, 1.0850e-07], + [-1.2014e-07, 2.3283e-09, 2.7940e-09, ..., 2.7940e-09, + 5.7742e-08, 9.4529e-08], + ..., + [ 1.0896e-07, 1.0710e-08, 2.0955e-08, ..., 1.2573e-08, + 3.6322e-08, 6.2399e-08], + [ 5.1223e-09, 1.2573e-08, 1.8626e-09, ..., 8.8476e-09, + 9.8255e-08, 1.4948e-07], + [ 4.1910e-09, -1.3970e-09, 4.6566e-09, ..., 1.4435e-08, + 6.6310e-07, 1.0878e-06]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0009, -0.0181, -0.0258, 0.0329, -0.0288, 0.0319, -0.0201, -0.0134, + 0.0032, -0.0259], device='cuda:0'), grad: tensor([-1.0714e-05, 2.9989e-07, 6.4261e-08, 5.2061e-07, 7.5903e-08, + 1.4780e-06, 3.6955e-06, 4.1071e-07, 5.2759e-07, 3.6433e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 215.07, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.4797 re_mapping 0.0042 re_causal 0.0126 /// teacc 98.95 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.1273, -0.1504, -0.1029, ..., -0.1546, -0.0181, 0.0787], + [ 0.0650, 0.0634, 0.0499, ..., 0.1078, -0.1787, -0.0592], + [ 0.0586, 0.0197, -0.1488, ..., -0.0466, -0.2205, -0.0894], + ..., + [-0.0551, -0.1407, 0.0887, ..., -0.1346, -0.0497, 0.0042], + [ 0.0198, 0.1061, 0.1579, ..., -0.1377, -0.1474, -0.2258], + [-0.1666, 0.0158, -0.2737, ..., -0.1442, -0.1091, 0.1116]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 7.9162e-09, 2.8871e-08, ..., 9.4064e-08, + -1.0431e-07, -3.7206e-07], + [-4.1910e-09, 1.3504e-08, 9.6858e-07, ..., 3.3323e-06, + 2.8163e-06, 6.3470e-07], + [-1.3970e-09, 2.4214e-08, 9.9652e-08, ..., 1.6158e-07, + 2.1746e-07, 4.4657e-07], + ..., + [ 1.8626e-09, 2.6543e-08, -4.1444e-08, ..., 2.9476e-07, + 9.8869e-06, 4.9889e-05], + [-1.2107e-08, -3.5129e-06, -5.3979e-06, ..., 1.1651e-06, + 9.7603e-07, 1.5460e-07], + [ 3.7253e-09, -6.4308e-07, 5.0338e-07, ..., 1.2303e-06, + -9.2238e-06, -5.4806e-05]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0010, -0.0179, -0.0258, 0.0323, -0.0288, 0.0304, -0.0204, -0.0135, + 0.0063, -0.0258], device='cuda:0'), grad: tensor([-3.0082e-07, 7.4431e-06, 1.4640e-06, 3.1013e-06, -9.5889e-06, + 1.5855e-05, 1.1846e-06, 1.2314e-04, -1.1228e-05, -1.3149e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.77, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4742 re_mapping 0.0041 re_causal 0.0126 /// teacc 99.06 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.1274, -0.1536, -0.1031, ..., -0.1563, -0.0210, 0.0784], + [ 0.0651, 0.0633, 0.0499, ..., 0.1081, -0.1790, -0.0596], + [ 0.0587, 0.0196, -0.1488, ..., -0.0472, -0.2226, -0.0898], + ..., + [-0.0552, -0.1410, 0.0887, ..., -0.1354, -0.0507, 0.0037], + [ 0.0201, 0.1061, 0.1582, ..., -0.1378, -0.1478, -0.2260], + [-0.1668, 0.0157, -0.2742, ..., -0.1450, -0.1093, 0.1117]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 9.3132e-09, 6.0536e-09, ..., 4.6566e-09, + -4.2049e-07, -7.7114e-07], + [-7.4506e-09, -6.2864e-08, -1.7043e-07, ..., -4.4238e-08, + 2.7940e-08, -3.6787e-08], + [-4.2748e-07, 1.7229e-08, 2.3749e-08, ..., 5.5879e-09, + 3.3528e-08, 5.4482e-08], + ..., + [ 3.7299e-07, 3.3062e-08, -1.4901e-08, ..., 2.3283e-08, + 1.8626e-08, 3.3993e-08], + [-1.5367e-08, -1.4901e-08, -1.5367e-08, ..., 1.4901e-08, + 7.0315e-08, 9.4995e-08], + [ 2.1886e-08, 8.7544e-08, 1.3178e-07, ..., 1.1874e-07, + 6.9849e-08, 7.9628e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0021, -0.0179, -0.0259, 0.0312, -0.0296, 0.0315, -0.0192, -0.0135, + 0.0063, -0.0257], device='cuda:0'), grad: tensor([-1.3532e-06, -2.0629e-07, -1.4305e-06, 7.3109e-07, -2.2631e-07, + -1.7956e-06, 2.0470e-06, 1.3728e-06, 2.3935e-07, 6.2445e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 215.23, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4849 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.03 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.1275, -0.1537, -0.1032, ..., -0.1563, -0.0210, 0.0785], + [ 0.0653, 0.0640, 0.0500, ..., 0.1093, -0.1794, -0.0599], + [ 0.0586, 0.0195, -0.1489, ..., -0.0480, -0.2231, -0.0901], + ..., + [-0.0553, -0.1419, 0.0887, ..., -0.1369, -0.0510, 0.0039], + [ 0.0198, 0.1060, 0.1581, ..., -0.1384, -0.1481, -0.2261], + [-0.1670, 0.0156, -0.2748, ..., -0.1456, -0.1096, 0.1117]], + device='cuda:0'), grad: tensor([[ 2.6543e-08, 5.7276e-08, 1.7229e-08, ..., 1.1642e-08, + 2.7008e-08, -6.8918e-08], + [-1.9651e-06, -1.0040e-06, -1.9576e-06, ..., -6.6310e-07, + 3.5390e-08, 1.4901e-08], + [ 1.7239e-06, 6.8126e-07, 1.3653e-06, ..., 3.6042e-07, + 1.3039e-08, 1.3504e-08], + ..., + [ 5.4063e-07, 2.5611e-07, -6.0536e-09, ..., 1.9465e-07, + 9.3132e-10, -5.0291e-08], + [ 9.1270e-08, 1.1921e-07, 5.2620e-08, ..., 6.5193e-08, + 9.7789e-08, 1.9558e-08], + [ 4.1444e-08, 2.7008e-08, 1.7043e-07, ..., 1.3039e-08, + 1.0245e-08, 3.9116e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0020, -0.0178, -0.0259, 0.0301, -0.0296, 0.0322, -0.0192, -0.0136, + 0.0061, -0.0258], device='cuda:0'), grad: tensor([ 8.8941e-08, -3.7886e-06, 4.4890e-06, -2.5388e-06, 1.2573e-07, + 1.2061e-06, -8.4471e-07, 2.5146e-07, 5.7509e-07, 4.3679e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 215.24, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.4891 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.08 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.1274, -0.1536, -0.1034, ..., -0.1563, -0.0209, 0.0788], + [ 0.0660, 0.0658, 0.0505, ..., 0.1126, -0.1796, -0.0599], + [ 0.0587, 0.0194, -0.1489, ..., -0.0489, -0.2237, -0.0903], + ..., + [-0.0554, -0.1430, 0.0887, ..., -0.1384, -0.0511, 0.0038], + [ 0.0196, 0.1060, 0.1580, ..., -0.1397, -0.1483, -0.2262], + [-0.1677, 0.0134, -0.2755, ..., -0.1507, -0.1110, 0.1116]], + device='cuda:0'), grad: tensor([[ 1.8720e-07, 3.4785e-07, 1.4808e-07, ..., 2.4447e-07, + 6.0955e-07, 5.4017e-08], + [-9.2667e-08, 1.1921e-07, -6.7754e-07, ..., -1.7323e-07, + 3.3993e-08, 1.0245e-08], + [-9.6411e-06, -1.0267e-05, 9.1735e-08, ..., 2.5574e-06, + 9.1866e-06, 1.1642e-08], + ..., + [ 2.0536e-07, 2.2911e-07, 1.0710e-07, ..., 7.9162e-08, + 1.3970e-08, 3.7253e-09], + [ 7.5102e-06, 1.0483e-05, -6.4727e-08, ..., 1.3178e-06, + 9.5926e-08, 5.1223e-09], + [ 2.2771e-07, 4.9872e-07, 1.8766e-07, ..., 4.5868e-07, + 9.5740e-07, 1.1036e-07]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0017, -0.0172, -0.0259, 0.0301, -0.0261, 0.0320, -0.0194, -0.0136, + 0.0060, -0.0277], device='cuda:0'), grad: tensor([ 1.7146e-06, -2.3190e-07, -7.6517e-06, 3.0566e-06, -2.0742e-05, + 6.9514e-06, -5.7332e-06, 5.7789e-07, 1.9357e-05, 2.7530e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 215.07, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4814 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.08 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.1278, -0.1542, -0.1036, ..., -0.1577, -0.0217, 0.0788], + [ 0.0660, 0.0655, 0.0505, ..., 0.1126, -0.1798, -0.0602], + [ 0.0589, 0.0193, -0.1490, ..., -0.0493, -0.2256, -0.0922], + ..., + [-0.0555, -0.1435, 0.0887, ..., -0.1389, -0.0512, 0.0036], + [ 0.0205, 0.1066, 0.1591, ..., -0.1391, -0.1484, -0.2266], + [-0.1676, 0.0154, -0.2750, ..., -0.1512, -0.1110, 0.1123]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 2.7008e-08, 1.0710e-08, ..., 1.4435e-08, + 2.7474e-08, -1.3178e-07], + [-1.0990e-07, 1.9092e-08, 2.1867e-06, ..., 5.7137e-07, + 3.3062e-08, 1.6764e-08], + [ 3.7253e-08, 2.5146e-08, 8.7544e-08, ..., 3.0734e-08, + 2.0489e-08, 1.3504e-08], + ..., + [ 7.3109e-08, -7.0315e-08, -3.7253e-06, ..., -1.0589e-06, + 2.7940e-09, -1.6298e-08], + [ 7.8697e-08, 1.7276e-07, 1.0021e-06, ..., 3.9907e-07, + 2.4727e-07, 3.4925e-08], + [ 8.8476e-09, 2.3749e-08, 1.8068e-07, ..., 3.2131e-08, + 1.8626e-08, 6.9849e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0022, -0.0174, -0.0259, 0.0303, -0.0265, 0.0313, -0.0188, -0.0136, + 0.0062, -0.0269], device='cuda:0'), grad: tensor([-1.4342e-07, 3.6620e-06, 2.4075e-07, -1.7975e-07, 5.4529e-07, + -1.6056e-06, 7.8604e-07, -6.0834e-06, 2.3320e-06, 4.2329e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 215.48, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4837 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.04 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.1279, -0.1544, -0.1036, ..., -0.1578, -0.0220, 0.0787], + [ 0.0656, 0.0656, 0.0500, ..., 0.1126, -0.1801, -0.0606], + [ 0.0587, 0.0191, -0.1490, ..., -0.0503, -0.2271, -0.0924], + ..., + [-0.0550, -0.1437, 0.0893, ..., -0.1382, -0.0513, 0.0040], + [ 0.0207, 0.1062, 0.1591, ..., -0.1404, -0.1495, -0.2268], + [-0.1680, 0.0154, -0.2776, ..., -0.1538, -0.1132, 0.1123]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.4680e-08, 3.7253e-08, ..., 3.5856e-08, + -9.1735e-07, -3.8482e-06], + [-2.3423e-07, -3.5390e-08, -8.3353e-08, ..., -1.0012e-07, + 2.8405e-08, 9.2201e-08], + [ 4.9826e-08, 2.0489e-08, 9.5461e-08, ..., 4.9826e-08, + 2.9337e-08, 4.8429e-08], + ..., + [ 1.2293e-07, 2.4214e-08, -2.7986e-07, ..., 6.2399e-08, + 1.1642e-08, 3.8650e-08], + [ 1.3970e-08, 2.3609e-07, 9.1270e-08, ..., 1.3271e-07, + 1.5926e-07, 9.7323e-08], + [ 1.0245e-08, 4.1910e-09, 7.8231e-08, ..., 2.2352e-08, + 1.2107e-07, 3.7905e-07]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0024, -0.0183, -0.0260, 0.0304, -0.0253, 0.0312, -0.0181, -0.0132, + 0.0058, -0.0278], device='cuda:0'), grad: tensor([-5.6624e-06, 4.1910e-08, 3.2317e-07, 2.1234e-07, -2.0722e-07, + 3.9227e-06, 3.7253e-08, -4.5262e-07, 1.0412e-06, 7.4040e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 215.44, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4880 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.02 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.1278, -0.1544, -0.1038, ..., -0.1578, -0.0220, 0.0789], + [ 0.0657, 0.0658, 0.0500, ..., 0.1131, -0.1803, -0.0612], + [ 0.0588, 0.0189, -0.1491, ..., -0.0509, -0.2286, -0.0933], + ..., + [-0.0551, -0.1440, 0.0894, ..., -0.1389, -0.0514, 0.0038], + [ 0.0209, 0.1061, 0.1593, ..., -0.1406, -0.1502, -0.2270], + [-0.1684, 0.0149, -0.2780, ..., -0.1541, -0.1136, 0.1124]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.6764e-08, 1.8626e-09, ..., 4.6566e-10, + -2.9337e-08, -1.8720e-06], + [-7.3574e-08, -3.1199e-08, -6.0536e-09, ..., -6.1933e-08, + 4.6566e-09, 1.8300e-07], + [ 9.3132e-09, 5.5879e-09, 1.6112e-07, ..., 5.5879e-09, + 3.7253e-09, 1.0338e-06], + ..., + [ 3.0734e-08, 2.1420e-08, -3.3388e-07, ..., 3.4459e-08, + 2.7940e-09, -9.5926e-08], + [ 1.0710e-08, 4.6566e-09, 3.6787e-08, ..., 1.0245e-08, + 2.0955e-08, 8.5216e-08], + [ 4.6566e-09, 6.5193e-09, 8.9873e-08, ..., 1.8626e-09, + 1.7229e-08, 4.2794e-07]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0023, -0.0184, -0.0260, 0.0303, -0.0250, 0.0311, -0.0180, -0.0131, + 0.0057, -0.0280], device='cuda:0'), grad: tensor([-3.9414e-06, 2.8312e-07, 2.4848e-06, 1.0096e-06, 7.3574e-08, + -7.8976e-07, 2.2678e-07, -5.5693e-07, 2.2259e-07, 9.8720e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.86, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.5120 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.06 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.1275, -0.1544, -0.1038, ..., -0.1578, -0.0219, 0.0790], + [ 0.0662, 0.0659, 0.0508, ..., 0.1144, -0.1798, -0.0604], + [ 0.0589, 0.0191, -0.1491, ..., -0.0511, -0.2293, -0.0932], + ..., + [-0.0555, -0.1443, 0.0889, ..., -0.1411, -0.0522, 0.0034], + [ 0.0206, 0.1058, 0.1593, ..., -0.1409, -0.1509, -0.2273], + [-0.1689, 0.0149, -0.2787, ..., -0.1543, -0.1140, 0.1123]], + device='cuda:0'), grad: tensor([[-1.2666e-07, 3.6322e-08, 3.2596e-09, ..., 6.0536e-09, + -1.9418e-07, -6.8964e-07], + [-6.9849e-09, 1.2107e-08, 1.3225e-07, ..., 2.4214e-08, + 8.5216e-08, 6.7521e-08], + [-6.3796e-08, 2.0023e-08, 1.8161e-08, ..., -1.8626e-08, + 2.1886e-08, 5.4482e-08], + ..., + [ 1.7695e-08, 1.1642e-08, -3.6228e-07, ..., -8.1025e-08, + -1.6112e-07, -8.1491e-08], + [-2.4866e-06, -6.6906e-06, 3.7253e-09, ..., 7.4506e-09, + -1.8571e-06, -1.6931e-06], + [ 9.6858e-08, -1.2573e-08, 7.5437e-08, ..., 3.7253e-08, + 1.8347e-07, -5.4017e-08]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0021, -0.0175, -0.0260, 0.0303, -0.0248, 0.0311, -0.0179, -0.0133, + 0.0054, -0.0283], device='cuda:0'), grad: tensor([-1.4678e-06, 7.9488e-07, 9.3132e-08, 3.5297e-07, 1.4585e-06, + 2.2218e-05, 9.5740e-07, -1.5479e-06, -2.3067e-05, 2.4773e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 215.49, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4848 re_mapping 0.0039 re_causal 0.0125 /// teacc 98.98 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.1275, -0.1544, -0.1041, ..., -0.1578, -0.0219, 0.0791], + [ 0.0667, 0.0665, 0.0513, ..., 0.1158, -0.1801, -0.0599], + [ 0.0588, 0.0191, -0.1493, ..., -0.0518, -0.2299, -0.0934], + ..., + [-0.0558, -0.1452, 0.0887, ..., -0.1425, -0.0521, 0.0032], + [ 0.0212, 0.1060, 0.1597, ..., -0.1420, -0.1512, -0.2274], + [-0.1704, 0.0141, -0.2804, ..., -0.1552, -0.1146, 0.1122]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 8.8476e-09, ..., 6.9849e-09, + 1.8626e-09, -1.2573e-08], + [-2.6589e-07, -1.1828e-07, -2.3702e-07, ..., -1.5926e-07, + 3.2596e-08, 2.3749e-08], + [ 5.9605e-08, 2.5146e-08, 1.6624e-07, ..., 8.3819e-08, + 2.8871e-08, 2.8871e-08], + ..., + [ 8.8941e-08, 4.1444e-08, -1.7695e-07, ..., 2.3143e-07, + 1.3551e-07, -7.9628e-08], + [ 3.9116e-08, 1.8161e-08, 5.8673e-08, ..., 2.8405e-08, + 1.3970e-09, 8.8476e-09], + [ 1.8626e-08, 4.1910e-09, 7.4506e-08, ..., 6.6590e-08, + 4.2841e-08, 1.1176e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0020, -0.0171, -0.0261, 0.0309, -0.0243, 0.0309, -0.0179, -0.0135, + 0.0055, -0.0287], device='cuda:0'), grad: tensor([-8.3819e-09, -3.4552e-07, 4.7311e-07, 3.4459e-08, -6.5146e-07, + 7.1712e-08, 7.9162e-08, -7.6834e-08, 1.4063e-07, 2.9569e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 215.14, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4851 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.00 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.1267, -0.1546, -0.1033, ..., -0.1576, -0.0217, 0.0793], + [ 0.0670, 0.0675, 0.0514, ..., 0.1166, -0.1806, -0.0600], + [ 0.0579, 0.0187, -0.1495, ..., -0.0532, -0.2330, -0.0941], + ..., + [-0.0553, -0.1441, 0.0896, ..., -0.1431, -0.0523, 0.0033], + [ 0.0192, 0.1056, 0.1575, ..., -0.1414, -0.1509, -0.2276], + [-0.1708, 0.0142, -0.2807, ..., -0.1554, -0.1149, 0.1123]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 1.0710e-08, 9.3132e-09, ..., 1.0245e-08, + 9.3132e-09, -3.1199e-08], + [-1.3458e-07, -7.5437e-08, -1.5646e-07, ..., -1.2480e-07, + 1.0245e-08, 6.9849e-09], + [-2.7614e-07, 1.6764e-08, 2.7940e-08, ..., 3.6322e-08, + 1.1642e-08, 5.1223e-09], + ..., + [ 2.9616e-07, 3.0734e-08, 2.2352e-08, ..., 5.3551e-08, + 3.2596e-09, -1.6298e-08], + [ 1.6298e-08, -5.5879e-09, -2.5146e-08, ..., 1.1642e-08, + 2.8405e-08, 2.3283e-09], + [ 2.0955e-08, 2.5611e-08, 2.9337e-08, ..., 9.8720e-08, + 4.0978e-08, 1.8626e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0017, -0.0169, -0.0265, 0.0315, -0.0244, 0.0310, -0.0183, -0.0129, + 0.0042, -0.0287], device='cuda:0'), grad: tensor([ 8.5682e-08, -2.2538e-07, -6.0024e-07, -3.9348e-07, -1.5087e-07, + 8.8755e-07, -7.3714e-07, 7.1991e-07, 1.4948e-07, 2.7660e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 214.88, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.5156 re_mapping 0.0040 re_causal 0.0127 /// teacc 98.98 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.1269, -0.1546, -0.1035, ..., -0.1576, -0.0217, 0.0794], + [ 0.0669, 0.0676, 0.0514, ..., 0.1170, -0.1807, -0.0602], + [ 0.0587, 0.0195, -0.1494, ..., -0.0521, -0.2333, -0.0943], + ..., + [-0.0554, -0.1444, 0.0897, ..., -0.1438, -0.0525, 0.0036], + [ 0.0177, 0.1057, 0.1573, ..., -0.1425, -0.1509, -0.2277], + [-0.1710, 0.0144, -0.2811, ..., -0.1554, -0.1149, 0.1124]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 4.6566e-09, 2.3283e-09, ..., 2.3283e-09, + -1.1828e-07, -4.3167e-07], + [ 3.7253e-08, 2.0955e-08, 2.7940e-08, ..., 1.3504e-08, + 4.0513e-08, 6.3796e-08], + [-2.7008e-08, 9.3132e-10, 4.1910e-09, ..., -3.2596e-09, + 7.4506e-09, 1.6764e-08], + ..., + [ 7.4506e-09, 9.3132e-10, -2.3283e-09, ..., 9.3132e-10, + 8.8476e-09, 1.9092e-08], + [-1.3504e-08, 1.8626e-09, -2.2817e-08, ..., 6.9849e-09, + 3.3528e-08, 2.7008e-08], + [ 1.0245e-08, 3.7253e-09, 9.7789e-09, ..., 1.8626e-09, + 5.8208e-08, 1.6065e-07]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0016, -0.0170, -0.0263, 0.0309, -0.0246, 0.0310, -0.0184, -0.0128, + 0.0042, -0.0287], device='cuda:0'), grad: tensor([-6.9849e-07, 2.6356e-07, -9.3132e-10, -4.1910e-08, 1.9092e-08, + -9.9838e-07, 1.0505e-06, 5.4017e-08, 4.7963e-08, 3.1991e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 215.13, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.5041 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.00 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.1289, -0.1552, -0.1037, ..., -0.1576, -0.0220, 0.0791], + [ 0.0668, 0.0673, 0.0513, ..., 0.1172, -0.1811, -0.0610], + [ 0.0589, 0.0197, -0.1494, ..., -0.0520, -0.2334, -0.0949], + ..., + [-0.0554, -0.1445, 0.0899, ..., -0.1434, -0.0526, 0.0039], + [ 0.0180, 0.1062, 0.1576, ..., -0.1425, -0.1512, -0.2279], + [-0.1690, 0.0153, -0.2819, ..., -0.1554, -0.1150, 0.1132]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 9.3132e-10, 1.3970e-09, ..., 1.3970e-09, + -6.3330e-06, -1.6242e-05], + [-8.8476e-08, -6.6590e-08, -1.4342e-07, ..., -1.1548e-07, + 9.3598e-08, 2.4913e-07], + [-7.9162e-09, 6.0536e-09, 1.0710e-08, ..., 1.1176e-08, + 3.3062e-08, 8.1025e-08], + ..., + [ 3.8650e-08, 2.9337e-08, 4.9826e-08, ..., 4.9360e-08, + 1.4901e-08, 3.1665e-08], + [-9.7789e-09, -1.2107e-08, -1.3970e-08, ..., 9.3132e-09, + 3.6787e-08, 1.0012e-07], + [ 5.1223e-09, -8.3819e-09, 9.3132e-09, ..., 9.7789e-09, + 1.9744e-07, 4.7637e-07]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0021, -0.0173, -0.0263, 0.0305, -0.0247, 0.0306, -0.0181, -0.0128, + 0.0045, -0.0281], device='cuda:0'), grad: tensor([-3.3200e-05, 2.6356e-07, 1.6810e-07, 1.2480e-07, 1.1642e-07, + 4.1723e-07, 3.0756e-05, 1.6857e-07, 1.6764e-07, 9.9652e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 215.09, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4959 re_mapping 0.0042 re_causal 0.0129 /// teacc 98.97 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.1295, -0.1552, -0.1040, ..., -0.1576, -0.0219, 0.0794], + [ 0.0667, 0.0680, 0.0514, ..., 0.1179, -0.1837, -0.0630], + [ 0.0587, 0.0189, -0.1495, ..., -0.0532, -0.2338, -0.0956], + ..., + [-0.0555, -0.1459, 0.0900, ..., -0.1465, -0.0535, 0.0043], + [ 0.0185, 0.1066, 0.1578, ..., -0.1429, -0.1514, -0.2280], + [-0.1691, 0.0131, -0.2842, ..., -0.1577, -0.1154, 0.1124]], + device='cuda:0'), grad: tensor([[ 3.0268e-07, 1.4110e-07, 1.5879e-07, ..., 6.1234e-08, + 1.8626e-09, -8.5449e-08], + [-2.5854e-06, -1.3160e-06, -1.5246e-06, ..., -5.2992e-07, + 2.5611e-09, 1.6997e-08], + [-8.9919e-07, 9.7323e-08, 1.2526e-07, ..., 5.1688e-08, + 1.8626e-09, 1.6997e-08], + ..., + [ 6.7148e-07, 3.1665e-08, -1.9558e-08, ..., 3.1432e-08, + 1.6298e-09, -6.3097e-08], + [ 2.5542e-07, 7.9628e-08, 8.8243e-08, ..., 3.7951e-08, + 2.0955e-09, 1.9325e-08], + [ 4.5868e-08, 2.8801e-07, 3.7719e-08, ..., 1.2498e-06, + 2.0955e-07, 5.5227e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0019, -0.0177, -0.0264, 0.0295, -0.0222, 0.0314, -0.0183, -0.0127, + 0.0047, -0.0298], device='cuda:0'), grad: tensor([ 5.9186e-07, -6.0275e-06, -2.1663e-06, 4.0140e-07, -2.7306e-06, + 2.7264e-07, 4.6380e-06, 1.4966e-06, 6.6496e-07, 2.8666e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 215.15, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4867 re_mapping 0.0042 re_causal 0.0128 /// teacc 98.96 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.1298, -0.1553, -0.1042, ..., -0.1577, -0.0220, 0.0795], + [ 0.0673, 0.0684, 0.0515, ..., 0.1195, -0.1838, -0.0639], + [ 0.0585, 0.0181, -0.1497, ..., -0.0564, -0.2341, -0.0977], + ..., + [-0.0557, -0.1463, 0.0901, ..., -0.1474, -0.0539, 0.0056], + [ 0.0185, 0.1066, 0.1580, ..., -0.1433, -0.1518, -0.2283], + [-0.1694, 0.0121, -0.2847, ..., -0.1587, -0.1158, 0.1120]], + device='cuda:0'), grad: tensor([[ 5.3504e-07, 2.9448e-06, 3.6694e-07, ..., 2.6310e-08, + 2.5351e-06, 3.4366e-06], + [ 1.1339e-07, 1.6927e-07, 1.3574e-07, ..., -6.5193e-09, + 7.3342e-08, 1.0524e-07], + [-4.8196e-08, 1.3970e-08, 3.2829e-08, ..., 1.8626e-09, + 8.8476e-09, 2.5844e-08], + ..., + [ 1.6764e-08, 7.2177e-09, -1.4179e-07, ..., 5.1223e-09, + 3.2596e-09, -8.9407e-08], + [-1.1325e-06, 2.7474e-06, -8.9966e-07, ..., 1.0477e-08, + 2.9523e-06, 3.8743e-06], + [ 2.9593e-07, 4.9174e-07, 2.6519e-07, ..., 2.0955e-09, + 3.6461e-07, 5.2620e-07]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0019, -0.0176, -0.0265, 0.0280, -0.0212, 0.0322, -0.0182, -0.0126, + 0.0045, -0.0305], device='cuda:0'), grad: tensor([ 1.1817e-05, 7.3528e-07, -3.4925e-09, 3.0175e-07, 2.0117e-07, + 5.1819e-06, -2.7299e-05, -3.3877e-07, 6.7316e-06, 2.6803e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.83, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4844 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.01 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.1304, -0.1552, -0.1046, ..., -0.1579, -0.0220, 0.0798], + [ 0.0673, 0.0680, 0.0515, ..., 0.1196, -0.1842, -0.0646], + [ 0.0590, 0.0188, -0.1497, ..., -0.0555, -0.2337, -0.0963], + ..., + [-0.0560, -0.1465, 0.0901, ..., -0.1483, -0.0540, 0.0052], + [ 0.0187, 0.1063, 0.1580, ..., -0.1439, -0.1527, -0.2287], + [-0.1695, 0.0123, -0.2850, ..., -0.1587, -0.1159, 0.1120]], + device='cuda:0'), grad: tensor([[-1.3458e-07, -1.9791e-08, 3.2363e-08, ..., -3.2829e-08, + -2.1746e-07, -1.8626e-06], + [-4.9826e-08, -2.7707e-08, -7.5204e-08, ..., -7.5670e-08, + 1.8394e-08, 4.0513e-08], + [-7.4040e-08, 9.0804e-09, 1.0151e-07, ..., 1.5134e-08, + 1.2806e-08, 6.8685e-08], + ..., + [ 8.5216e-08, 1.1409e-08, 1.1339e-07, ..., 1.3737e-08, + 9.3132e-09, 2.5379e-08], + [-6.5425e-08, 1.9232e-07, -2.8359e-07, ..., 1.1642e-08, + 3.9325e-07, 5.3039e-07], + [ 1.0780e-07, 2.1118e-07, 2.5379e-08, ..., 3.0035e-08, + 4.5612e-07, 1.4547e-06]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0017, -0.0177, -0.0264, 0.0280, -0.0214, 0.0323, -0.0181, -0.0127, + 0.0042, -0.0305], device='cuda:0'), grad: tensor([-2.9430e-06, -4.9127e-08, 1.1944e-07, 1.7765e-07, 8.9407e-08, + -3.8445e-06, 2.6673e-06, 3.4645e-07, 6.6496e-07, 2.7735e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 215.37, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.5021 re_mapping 0.0041 re_causal 0.0126 /// teacc 98.96 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.1308, -0.1553, -0.1049, ..., -0.1568, -0.0206, 0.0803], + [ 0.0675, 0.0685, 0.0516, ..., 0.1210, -0.1846, -0.0650], + [ 0.0589, 0.0186, -0.1498, ..., -0.0561, -0.2339, -0.0968], + ..., + [-0.0560, -0.1468, 0.0902, ..., -0.1494, -0.0539, 0.0019], + [ 0.0186, 0.1063, 0.1581, ..., -0.1445, -0.1530, -0.2289], + [-0.1697, 0.0123, -0.2852, ..., -0.1588, -0.1160, 0.1129]], + device='cuda:0'), grad: tensor([[ 2.3795e-07, 4.0513e-08, 3.8184e-08, ..., 2.3749e-08, + -4.6566e-10, -2.8871e-08], + [ 1.3066e-06, 3.5856e-07, 3.8277e-07, ..., 2.3656e-07, + 9.3132e-10, 2.3283e-09], + [-4.3102e-06, 2.4075e-07, 4.7637e-07, ..., 3.1106e-07, + 4.6566e-10, 5.1223e-09], + ..., + [ 1.9092e-07, 2.9337e-08, 1.3970e-08, ..., 2.4680e-08, + 4.6566e-10, -2.3283e-09], + [ 5.5842e-06, 3.6275e-07, 2.1188e-07, ..., 1.7555e-07, + 5.5879e-09, 8.8476e-09], + [ 2.1085e-06, 2.5611e-08, 3.1199e-08, ..., 1.0245e-08, + 1.8626e-09, 3.1758e-07]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0003, -0.0176, -0.0264, 0.0283, -0.0216, 0.0323, -0.0198, -0.0130, + 0.0041, -0.0298], device='cuda:0'), grad: tensor([ 6.4541e-07, 3.3733e-06, -1.2957e-05, -2.6047e-05, 3.4273e-07, + 4.8093e-06, 2.3190e-07, 5.0897e-07, 1.6019e-05, 1.3039e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 214.87, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4744 re_mapping 0.0041 re_causal 0.0125 /// teacc 99.10 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.1302, -0.1553, -0.1038, ..., -0.1568, -0.0204, 0.0807], + [ 0.0676, 0.0686, 0.0516, ..., 0.1212, -0.1853, -0.0653], + [ 0.0590, 0.0184, -0.1499, ..., -0.0564, -0.2344, -0.0971], + ..., + [-0.0561, -0.1470, 0.0902, ..., -0.1494, -0.0540, 0.0015], + [ 0.0181, 0.1057, 0.1579, ..., -0.1458, -0.1542, -0.2293], + [-0.1700, 0.0124, -0.2854, ..., -0.1590, -0.1162, 0.1128]], + device='cuda:0'), grad: tensor([[ 2.6543e-08, 2.9337e-08, 1.3039e-08, ..., 3.7253e-09, + 2.7474e-08, -2.7008e-08], + [ 4.0047e-08, 1.3458e-07, 1.0896e-07, ..., 2.7474e-08, + 1.3411e-07, 4.3306e-08], + [-4.8894e-07, -1.4901e-08, 1.1967e-07, ..., 6.0536e-09, + 3.7253e-09, 4.6566e-09], + ..., + [ 9.4064e-08, 1.3970e-08, -3.3574e-07, ..., 6.6124e-08, + 2.6543e-08, -3.2596e-09], + [ 1.8580e-07, 2.8638e-07, 1.1548e-07, ..., 7.5903e-08, + 2.3702e-07, 7.5437e-08], + [ 5.5879e-09, 1.9558e-08, 2.5146e-08, ..., 2.5611e-07, + 8.6147e-08, 1.5367e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0003, -0.0179, -0.0265, 0.0291, -0.0217, 0.0323, -0.0198, -0.0131, + 0.0034, -0.0299], device='cuda:0'), grad: tensor([ 8.1491e-08, 5.6392e-07, -7.6089e-07, 1.2014e-07, -1.3700e-06, + 2.6263e-06, -3.4980e-06, -6.8918e-08, 1.2834e-06, 1.0096e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 214.92, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.5003 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.02 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.1308, -0.1554, -0.1036, ..., -0.1569, -0.0204, 0.0808], + [ 0.0686, 0.0686, 0.0520, ..., 0.1240, -0.1856, -0.0651], + [ 0.0583, 0.0189, -0.1501, ..., -0.0599, -0.2347, -0.0974], + ..., + [-0.0562, -0.1475, 0.0902, ..., -0.1498, -0.0541, 0.0015], + [ 0.0182, 0.1058, 0.1581, ..., -0.1460, -0.1546, -0.2297], + [-0.1706, 0.0095, -0.2858, ..., -0.1594, -0.1164, 0.1126]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 2.4773e-07, 3.2596e-09, ..., 2.8796e-06, + 2.0355e-05, 1.4223e-05], + [-1.2573e-08, -2.7940e-09, -1.9092e-08, ..., -5.1223e-09, + 1.3551e-07, 1.2945e-07], + [-3.6322e-08, 6.0536e-09, 1.1176e-08, ..., 1.0245e-08, + 6.8452e-08, 1.2107e-07], + ..., + [ 1.1642e-08, 5.1223e-09, -1.5832e-08, ..., 6.0536e-09, + 3.3993e-08, 1.6671e-07], + [-1.8626e-08, -9.2015e-07, -3.5297e-07, ..., -5.9139e-08, + 3.8836e-07, 3.2783e-07], + [ 4.6566e-09, 1.4435e-08, 2.1886e-08, ..., 2.4214e-08, + 1.8813e-07, 2.5984e-07]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0005, -0.0173, -0.0266, 0.0293, -0.0192, 0.0323, -0.0200, -0.0131, + 0.0032, -0.0322], device='cuda:0'), grad: tensor([ 6.6280e-05, 4.7451e-07, 2.2724e-07, 1.2061e-07, 9.0664e-07, + 8.8140e-06, -7.7128e-05, 2.8359e-07, -7.9907e-07, 7.3202e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 214.99, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.4940 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.1312, -0.1552, -0.1024, ..., -0.1566, -0.0204, 0.0815], + [ 0.0688, 0.0703, 0.0521, ..., 0.1268, -0.1858, -0.0656], + [ 0.0581, 0.0186, -0.1503, ..., -0.0610, -0.2351, -0.0986], + ..., + [-0.0559, -0.1480, 0.0906, ..., -0.1490, -0.0541, 0.0018], + [ 0.0183, 0.1058, 0.1582, ..., -0.1463, -0.1550, -0.2304], + [-0.1710, 0.0094, -0.2868, ..., -0.1597, -0.1166, 0.1127]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 6.2864e-08, 2.5611e-08, ..., 1.0245e-08, + 8.7544e-08, 9.3132e-09], + [ 5.6671e-07, 2.1793e-07, 3.7104e-05, ..., 7.4506e-09, + 9.3132e-09, 6.0536e-09], + [-4.6566e-10, 8.3819e-09, 1.1930e-06, ..., 3.2596e-09, + 3.7253e-09, 4.6566e-10], + ..., + [-5.8673e-07, -2.1793e-07, -3.9011e-05, ..., 2.0489e-08, + 1.0245e-08, 1.1176e-08], + [ 5.1223e-09, 1.1642e-08, 2.0256e-07, ..., 7.4506e-09, + 1.4901e-08, 1.2107e-08], + [ 4.1910e-09, 5.1223e-09, 2.5705e-07, ..., 3.9581e-08, + 2.3749e-08, -1.9558e-08]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0014, -0.0174, -0.0268, 0.0291, -0.0195, 0.0318, -0.0203, -0.0127, + 0.0030, -0.0322], device='cuda:0'), grad: tensor([ 2.9849e-07, 5.4181e-05, 1.7127e-06, 1.4156e-07, 4.6100e-08, + 2.2352e-07, -5.3598e-07, -5.6833e-05, 3.6694e-07, 4.4284e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.72, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4998 re_mapping 0.0040 re_causal 0.0129 /// teacc 98.99 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.1316, -0.1551, -0.1025, ..., -0.1567, -0.0204, 0.0818], + [ 0.0683, 0.0701, 0.0520, ..., 0.1275, -0.1866, -0.0679], + [ 0.0591, 0.0180, -0.1504, ..., -0.0613, -0.2355, -0.0974], + ..., + [-0.0563, -0.1487, 0.0907, ..., -0.1502, -0.0546, 0.0017], + [ 0.0192, 0.1065, 0.1586, ..., -0.1469, -0.1556, -0.2301], + [-0.1737, 0.0093, -0.2873, ..., -0.1599, -0.1168, 0.1125]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.3283e-08, 9.3132e-10, ..., 4.6566e-10, + 2.8871e-08, -1.1781e-07], + [-1.9558e-08, -4.6566e-09, -1.4435e-08, ..., -2.0955e-08, + 6.0536e-09, 6.9849e-09], + [-2.9337e-08, 3.2596e-09, 5.3085e-08, ..., 8.8476e-09, + 1.3970e-09, 1.4435e-08], + ..., + [ 2.9802e-08, 3.2596e-09, -7.5437e-08, ..., 4.1910e-09, + 1.8626e-09, 8.8476e-09], + [-3.2596e-08, -2.2212e-07, -3.4878e-07, ..., -1.8394e-07, + -6.3330e-08, 2.4680e-08], + [ 2.3283e-09, 9.3132e-10, 2.3749e-08, ..., 2.7940e-09, + 4.1910e-09, 6.9849e-08]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0016, -0.0184, -0.0265, 0.0287, -0.0195, 0.0307, -0.0186, -0.0127, + 0.0034, -0.0325], device='cuda:0'), grad: tensor([-1.1828e-07, -2.3283e-09, 4.0047e-08, 5.3085e-08, 4.7963e-08, + 1.4855e-07, 1.4063e-07, -9.4529e-08, -3.6415e-07, 1.6252e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 214.67, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4965 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.1317, -0.1553, -0.1025, ..., -0.1569, -0.0198, 0.0824], + [ 0.0688, 0.0696, 0.0523, ..., 0.1280, -0.1873, -0.0682], + [ 0.0595, 0.0171, -0.1507, ..., -0.0618, -0.2362, -0.0985], + ..., + [-0.0564, -0.1490, 0.0909, ..., -0.1507, -0.0550, 0.0018], + [ 0.0207, 0.1076, 0.1594, ..., -0.1474, -0.1561, -0.2294], + [-0.1739, 0.0092, -0.2882, ..., -0.1603, -0.1175, 0.1124]], + device='cuda:0'), grad: tensor([[-2.3283e-09, 7.9162e-09, 4.1910e-09, ..., 6.9849e-09, + 5.1223e-09, -2.1420e-08], + [-3.9581e-08, -2.7008e-08, -3.6322e-08, ..., -2.1420e-08, + 1.3970e-09, 3.2596e-09], + [ 4.1910e-09, 5.1223e-09, 1.7695e-08, ..., 1.0710e-08, + 1.8626e-09, 9.3132e-09], + ..., + [ 8.3819e-09, 9.7789e-09, -4.0513e-08, ..., 2.6077e-08, + 4.6566e-10, 2.7940e-09], + [-7.4506e-08, -8.5682e-08, -1.5646e-07, ..., 1.5367e-08, + -7.0781e-08, 1.1176e-08], + [ 6.9849e-09, -6.9849e-09, 1.2573e-08, ..., 6.0024e-07, + 3.2596e-09, 9.9186e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0021, -0.0182, -0.0264, 0.0274, -0.0195, 0.0302, -0.0188, -0.0127, + 0.0042, -0.0327], device='cuda:0'), grad: tensor([ 1.3504e-08, -3.7253e-08, 7.6834e-08, 4.6473e-07, -3.6173e-06, + 1.2154e-07, 8.2422e-08, 8.8941e-08, -5.2806e-07, 3.3639e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 215.14, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5086 re_mapping 0.0038 re_causal 0.0123 /// teacc 98.98 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.1309, -0.1553, -0.1027, ..., -0.1566, -0.0195, 0.0825], + [ 0.0689, 0.0708, 0.0524, ..., 0.1286, -0.1880, -0.0680], + [ 0.0596, 0.0163, -0.1511, ..., -0.0622, -0.2371, -0.0991], + ..., + [-0.0564, -0.1509, 0.0912, ..., -0.1517, -0.0553, 0.0018], + [ 0.0207, 0.1078, 0.1595, ..., -0.1479, -0.1563, -0.2297], + [-0.1738, 0.0094, -0.2889, ..., -0.1606, -0.1176, 0.1127]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, 8.3819e-09, 2.1886e-08, ..., 6.5193e-09, + 3.2596e-09, -2.1327e-07], + [-2.4214e-08, -1.3784e-07, -2.3609e-07, ..., -2.0768e-07, + 3.7253e-09, 8.3819e-09], + [ 2.5518e-07, 1.8440e-07, 9.1270e-08, ..., 1.3504e-08, + 4.6566e-09, -9.3132e-10], + ..., + [ 1.0151e-07, 7.4506e-08, 6.3330e-08, ..., 9.8720e-08, + 3.1199e-08, 2.7474e-08], + [-3.2596e-09, -1.2713e-07, -1.3039e-07, ..., 2.8871e-08, + 8.3819e-09, 2.6077e-08], + [ 4.4238e-08, 1.6764e-08, 3.5856e-08, ..., 3.3993e-08, + 1.4435e-08, 6.5193e-08]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0024, -0.0181, -0.0265, 0.0283, -0.0196, 0.0290, -0.0189, -0.0125, + 0.0041, -0.0325], device='cuda:0'), grad: tensor([-1.7555e-07, -1.7090e-07, 9.0804e-07, -2.4922e-06, -4.0187e-07, + 7.4599e-07, 4.2189e-07, 6.8964e-07, 9.4529e-08, 3.9116e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.61, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4728 re_mapping 0.0040 re_causal 0.0124 /// teacc 99.08 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.1316, -0.1559, -0.1029, ..., -0.1566, -0.0201, 0.0824], + [ 0.0707, 0.0737, 0.0554, ..., 0.1288, -0.1883, -0.0654], + [ 0.0598, 0.0166, -0.1512, ..., -0.0626, -0.2377, -0.0994], + ..., + [-0.0581, -0.1541, 0.0884, ..., -0.1514, -0.0552, -0.0008], + [ 0.0205, 0.1079, 0.1595, ..., -0.1481, -0.1567, -0.2300], + [-0.1736, 0.0098, -0.2894, ..., -0.1605, -0.1178, 0.1132]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 7.9162e-09, 9.3132e-10, ..., 1.8626e-09, + -6.9849e-09, -8.8010e-08], + [-1.1642e-08, -9.7789e-09, -2.7940e-08, ..., -2.1886e-08, + 4.1910e-09, 1.7229e-08], + [-2.3283e-09, 3.7253e-09, 4.6566e-09, ..., 3.2596e-09, + 2.3283e-09, 3.2596e-08], + ..., + [ 1.2107e-08, 9.3132e-09, 1.3504e-08, ..., 1.2107e-08, + 9.3132e-10, 7.7765e-08], + [ 2.3283e-09, 9.8720e-08, -4.6566e-10, ..., 1.5367e-08, + 7.9628e-08, 1.7229e-08], + [ 4.6566e-09, 8.3819e-09, 4.6566e-09, ..., 2.1420e-08, + 9.3132e-09, -3.3434e-07]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0021, -0.0152, -0.0264, 0.0300, -0.0201, 0.0280, -0.0189, -0.0149, + 0.0038, -0.0322], device='cuda:0'), grad: tensor([-1.8859e-07, 0.0000e+00, 1.0338e-07, -1.0245e-08, 2.5798e-07, + 1.3364e-07, -2.8685e-07, 1.5367e-07, 2.5285e-07, -4.0000e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.91, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4537 re_mapping 0.0041 re_causal 0.0124 /// teacc 98.93 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.1323, -0.1561, -0.1036, ..., -0.1567, -0.0201, 0.0810], + [ 0.0684, 0.0738, 0.0551, ..., 0.1294, -0.1894, -0.0655], + [ 0.0591, 0.0157, -0.1516, ..., -0.0635, -0.2372, -0.0988], + ..., + [-0.0556, -0.1542, 0.0888, ..., -0.1516, -0.0552, -0.0008], + [ 0.0208, 0.1078, 0.1597, ..., -0.1490, -0.1574, -0.2303], + [-0.1737, 0.0099, -0.2901, ..., -0.1609, -0.1181, 0.1148]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 7.9162e-09, 2.8405e-08, ..., -6.1700e-07, + -1.9465e-06, -5.7705e-06], + [-1.1727e-05, -1.7015e-06, -1.0006e-05, ..., -5.4128e-06, + 2.6077e-08, 6.4261e-08], + [ 4.4294e-06, 6.7381e-07, 3.7756e-06, ..., 2.2352e-06, + 2.6682e-07, 1.1642e-08], + ..., + [ 6.3255e-06, 9.2015e-07, 5.3756e-06, ..., 2.9225e-06, + 5.5879e-09, -1.3504e-08], + [ 2.3283e-08, 6.7055e-08, 1.2107e-08, ..., 1.8626e-08, + 2.1420e-08, 1.4249e-07], + [ 2.5146e-08, -7.4971e-08, 2.7008e-08, ..., 9.4529e-08, + 1.7323e-07, 2.2631e-07]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0011, -0.0160, -0.0267, 0.0299, -0.0203, 0.0280, -0.0187, -0.0141, + 0.0038, -0.0315], device='cuda:0'), grad: tensor([-9.7901e-06, -2.8759e-05, 1.1414e-05, 2.0191e-06, -4.5821e-07, + 9.6392e-08, 9.1046e-06, 1.5527e-05, 4.2235e-07, 4.1211e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 215.28, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4786 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.94 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.1326, -0.1561, -0.1038, ..., -0.1567, -0.0199, 0.0814], + [ 0.0684, 0.0738, 0.0551, ..., 0.1297, -0.1899, -0.0658], + [ 0.0589, 0.0151, -0.1520, ..., -0.0638, -0.2375, -0.0988], + ..., + [-0.0554, -0.1543, 0.0889, ..., -0.1524, -0.0555, -0.0007], + [ 0.0214, 0.1081, 0.1601, ..., -0.1492, -0.1581, -0.2305], + [-0.1742, 0.0099, -0.2909, ..., -0.1610, -0.1185, 0.1147]], + device='cuda:0'), grad: tensor([[-1.4435e-08, -2.2631e-07, 1.8626e-09, ..., -1.1176e-08, + -5.7090e-07, -1.7732e-06], + [-7.9162e-09, 2.3283e-09, -4.6566e-09, ..., -9.7789e-09, + 2.2352e-08, 1.2573e-08], + [-1.2061e-07, 6.5193e-09, -2.0862e-07, ..., 9.3132e-10, + 1.0710e-08, 2.0023e-08], + ..., + [ 1.0710e-08, 5.1223e-09, 3.7253e-09, ..., 4.1910e-09, + 1.4901e-08, 3.7253e-09], + [ 1.1036e-07, 4.7823e-07, 2.2957e-07, ..., 2.7940e-09, + 4.6939e-07, 8.9873e-08], + [ 1.0710e-08, 1.4529e-07, 3.7253e-09, ..., 1.1176e-08, + 3.8650e-07, 1.0934e-06]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0015, -0.0160, -0.0268, 0.0297, -0.0204, 0.0281, -0.0188, -0.0140, + 0.0039, -0.0316], device='cuda:0'), grad: tensor([-1.9632e-06, 5.3551e-08, -9.2760e-07, 2.1467e-07, 8.5682e-08, + -2.8592e-07, -5.7882e-07, 8.2422e-08, 2.0675e-06, 1.2591e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 215.05, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4881 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.05 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.1333, -0.1561, -0.1037, ..., -0.1569, -0.0197, 0.0816], + [ 0.0686, 0.0740, 0.0552, ..., 0.1316, -0.1903, -0.0660], + [ 0.0590, 0.0147, -0.1522, ..., -0.0648, -0.2384, -0.0988], + ..., + [-0.0555, -0.1545, 0.0889, ..., -0.1538, -0.0561, -0.0010], + [ 0.0213, 0.1082, 0.1600, ..., -0.1515, -0.1584, -0.2307], + [-0.1744, 0.0105, -0.2911, ..., -0.1611, -0.1189, 0.1153]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-09, 8.8476e-09, ..., 1.3970e-09, + -9.7789e-09, -6.5658e-08], + [-1.7695e-08, -2.1886e-08, -3.1199e-08, ..., -2.5611e-08, + -0.0000e+00, 4.1910e-09], + [ 7.4506e-09, 5.5879e-09, 1.4901e-08, ..., 1.8626e-09, + 1.8626e-09, 6.5193e-09], + ..., + [ 8.3819e-09, 8.8476e-09, -2.8405e-08, ..., 9.7789e-09, + -1.6298e-08, -4.2375e-08], + [-1.4901e-08, -1.5832e-08, -2.0489e-08, ..., 3.2596e-09, + 9.3132e-10, -0.0000e+00], + [ 8.3819e-09, 1.0710e-08, 1.5832e-08, ..., 5.5879e-09, + 1.1176e-08, 4.5169e-08]], device='cuda:0') +Epoch 287, bias, value: tensor([ 0.0016, -0.0159, -0.0268, 0.0305, -0.0211, 0.0274, -0.0188, -0.0140, + 0.0037, -0.0311], device='cuda:0'), grad: tensor([-6.7987e-08, -5.4482e-08, 3.8184e-08, -2.1420e-08, 1.5832e-08, + 9.8720e-08, 4.8894e-08, -9.5461e-08, -4.5169e-08, 1.0012e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 286---------------------------------------------------- +epoch 286, time 231.55, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4769 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.15 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.1336, -0.1563, -0.1038, ..., -0.1571, -0.0199, 0.0815], + [ 0.0686, 0.0742, 0.0552, ..., 0.1322, -0.1903, -0.0662], + [ 0.0596, 0.0146, -0.1522, ..., -0.0645, -0.2391, -0.0986], + ..., + [-0.0557, -0.1546, 0.0890, ..., -0.1549, -0.0573, -0.0004], + [ 0.0214, 0.1083, 0.1603, ..., -0.1522, -0.1588, -0.2308], + [-0.1749, 0.0106, -0.2928, ..., -0.1614, -0.1191, 0.1156]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.0006e-05, 1.3970e-09, ..., 3.2596e-09, + 3.9577e-05, 2.9653e-05], + [ 4.6566e-09, 1.3970e-08, -4.6566e-09, ..., 5.8673e-08, + 5.8208e-08, 4.7963e-08], + [-7.2177e-08, -4.2841e-08, 4.5635e-08, ..., -3.4133e-07, + 1.7695e-08, 1.6298e-08], + ..., + [ 1.3970e-08, 6.5193e-09, -1.0524e-07, ..., 3.7253e-09, + 2.7940e-09, -2.9337e-08], + [ 2.8871e-08, 2.5146e-08, 2.5611e-08, ..., 5.1223e-09, + 1.0571e-07, 8.3819e-08], + [ 4.1910e-09, 9.7789e-09, 2.4214e-08, ..., 6.1467e-08, + 8.0094e-08, 4.5169e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0015, -0.0160, -0.0265, 0.0307, -0.0213, 0.0273, -0.0187, -0.0140, + 0.0038, -0.0312], device='cuda:0'), grad: tensor([ 1.1677e-04, 4.0280e-07, -9.0245e-07, -1.0896e-07, 4.3260e-07, + 7.2271e-06, -1.2445e-04, -1.7323e-07, 4.2282e-07, 2.7567e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 214.66, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.4992 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.1339, -0.1565, -0.1042, ..., -0.1572, -0.0193, 0.0827], + [ 0.0687, 0.0742, 0.0553, ..., 0.1328, -0.1906, -0.0663], + [ 0.0596, 0.0125, -0.1523, ..., -0.0664, -0.2422, -0.0992], + ..., + [-0.0559, -0.1564, 0.0877, ..., -0.1553, -0.0574, -0.0004], + [ 0.0217, 0.1099, 0.1635, ..., -0.1525, -0.1616, -0.2313], + [-0.1751, 0.0106, -0.2939, ..., -0.1615, -0.1199, 0.1154]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.3039e-08, 2.1420e-08, ..., 2.1420e-08, + -8.4117e-06, -2.2069e-05], + [-3.2037e-07, -2.3097e-07, -4.9546e-07, ..., -4.7404e-07, + 5.7742e-08, 1.3318e-07], + [ 8.2888e-08, 5.0291e-08, 1.0617e-07, ..., 9.8720e-08, + 2.0489e-08, 7.4506e-08], + ..., + [ 1.3597e-07, 9.4995e-08, 1.8533e-07, ..., 1.9185e-07, + 8.3819e-08, 2.2165e-07], + [ 4.4703e-08, 1.4901e-08, 3.7253e-09, ..., 4.6566e-08, + 6.0536e-08, 8.4750e-08], + [ 2.2352e-08, 1.7695e-08, 3.2596e-08, ..., 6.4261e-08, + 4.3958e-07, 1.0710e-06]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0026, -0.0160, -0.0266, 0.0307, -0.0215, 0.0259, -0.0170, -0.0152, + 0.0063, -0.0314], device='cuda:0'), grad: tensor([-3.0845e-05, -1.1027e-06, 4.3213e-07, -9.4995e-08, 1.2293e-07, + 2.7567e-05, 1.0915e-06, 8.3819e-07, 2.9895e-07, 1.6717e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 214.74, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5031 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.02 lr 0.00010000 +Epoch 290, weight, value: tensor([[-1.3443e-01, -1.5672e-01, -1.0448e-01, ..., -1.5753e-01, + -1.9234e-02, 8.3040e-02], + [ 6.9241e-02, 7.3948e-02, 5.5257e-02, ..., 1.3274e-01, + -1.9202e-01, -6.6994e-02], + [ 5.9453e-02, 1.3777e-02, -1.5242e-01, ..., -6.5674e-02, + -2.4154e-01, -9.8993e-02], + ..., + [-5.6366e-02, -1.5647e-01, 8.7792e-02, ..., -1.5595e-01, + -5.7150e-02, 1.4807e-04], + [ 2.1640e-02, 1.0983e-01, 1.6347e-01, ..., -1.5362e-01, + -1.6274e-01, -2.3175e-01], + [-1.7536e-01, 1.0526e-02, -2.9553e-01, ..., -1.6216e-01, + -1.2109e-01, 1.1529e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.5832e-08, 4.6566e-09, ..., 8.3819e-09, + -1.1707e-06, -1.1131e-05], + [ 8.3819e-08, 6.2399e-08, 7.1246e-07, ..., 4.2841e-08, + 4.1910e-08, 6.7055e-08], + [-3.2876e-07, 3.9116e-08, 2.4866e-07, ..., 6.5193e-09, + 1.0245e-08, 6.4261e-08], + ..., + [-6.9849e-08, 2.3283e-08, -1.4789e-06, ..., 1.5832e-08, + 8.3819e-09, 6.7055e-08], + [ 1.3318e-07, -1.8161e-07, -1.3132e-07, ..., 1.4901e-08, + 3.2596e-08, 1.9185e-07], + [ 1.0245e-08, 1.9558e-08, 1.0524e-07, ..., 1.2200e-07, + 1.1269e-07, 3.2317e-07]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0027, -0.0161, -0.0267, 0.0307, -0.0214, 0.0264, -0.0173, -0.0151, + 0.0062, -0.0317], device='cuda:0'), grad: tensor([-2.3097e-05, 2.2314e-06, 1.9092e-07, 1.9461e-05, -6.4075e-07, + -1.7524e-05, 2.1234e-05, -3.8929e-06, 7.3668e-07, 1.3113e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 214.99, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4878 re_mapping 0.0040 re_causal 0.0122 /// teacc 99.07 lr 0.00010000 +Epoch 291, weight, value: tensor([[-1.3515e-01, -1.5674e-01, -1.0474e-01, ..., -1.5767e-01, + -1.8932e-02, 8.3604e-02], + [ 6.8428e-02, 7.1924e-02, 5.4388e-02, ..., 1.3270e-01, + -1.9291e-01, -6.7345e-02], + [ 5.9822e-02, 1.4390e-02, -1.5252e-01, ..., -6.5321e-02, + -2.4136e-01, -9.8587e-02], + ..., + [-5.5715e-02, -1.5471e-01, 8.8623e-02, ..., -1.5641e-01, + -5.7351e-02, 1.6924e-04], + [ 2.1442e-02, 1.0967e-01, 1.6345e-01, ..., -1.5411e-01, + -1.6336e-01, -2.3223e-01], + [-1.7552e-01, 1.0634e-02, -2.9585e-01, ..., -1.6250e-01, + -1.2187e-01, 1.1524e-01]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.9558e-08, 1.8626e-09, ..., 2.7940e-09, + 8.3819e-09, 3.2596e-08], + [ 7.4971e-07, 1.3486e-06, 4.6194e-07, ..., 1.2293e-07, + 1.8626e-09, 2.5611e-07], + [-4.0978e-08, 6.3330e-08, 2.2352e-08, ..., 9.3132e-09, + 3.7253e-09, 1.5832e-08], + ..., + [ 6.6124e-08, 7.4506e-09, 1.0245e-08, ..., 1.0245e-08, + 9.3132e-10, 9.3132e-10], + [-7.8976e-07, -1.5069e-06, -5.3924e-07, ..., -1.5646e-07, + 3.7253e-09, -2.5239e-07], + [ 2.5146e-08, -2.7940e-08, 3.7253e-09, ..., 1.7695e-08, + 9.3132e-10, -1.1921e-07]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0033, -0.0171, -0.0266, 0.0308, -0.0216, 0.0265, -0.0175, -0.0143, + 0.0060, -0.0317], device='cuda:0'), grad: tensor([ 7.7300e-08, 2.5891e-06, -9.4995e-08, -5.9605e-07, 1.5832e-08, + 4.2748e-07, 6.9849e-08, 2.1607e-07, -2.6729e-06, -1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 214.79, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4710 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.04 lr 0.00010000 +Epoch 292, weight, value: tensor([[-1.3559e-01, -1.5682e-01, -1.0495e-01, ..., -1.5799e-01, + -1.9032e-02, 8.3450e-02], + [ 6.8657e-02, 7.2140e-02, 5.4298e-02, ..., 1.3281e-01, + -1.9338e-01, -6.6961e-02], + [ 5.9897e-02, 1.4317e-02, -1.5270e-01, ..., -6.5466e-02, + -2.4169e-01, -9.8813e-02], + ..., + [-5.5905e-02, -1.5489e-01, 8.8811e-02, ..., -1.5641e-01, + -5.6047e-02, -7.9418e-05], + [ 2.1276e-02, 1.0959e-01, 1.6343e-01, ..., -1.5443e-01, + -1.6410e-01, -2.3310e-01], + [-1.7561e-01, 1.0689e-02, -2.9705e-01, ..., -1.6274e-01, + -1.2237e-01, 1.1554e-01]], device='cuda:0'), grad: tensor([[-1.8254e-07, 2.4214e-08, 3.5390e-08, ..., 2.7940e-09, + 2.0489e-08, -1.5609e-06], + [ 5.6103e-06, 2.9262e-06, 5.9418e-07, ..., 3.5111e-07, + 1.5590e-06, 1.1109e-05], + [-5.9698e-07, -2.7008e-08, 8.7544e-08, ..., 5.5879e-09, + 9.3132e-09, 9.2760e-07], + ..., + [ 2.5798e-07, 6.8918e-08, -2.7008e-07, ..., 1.5181e-07, + 3.7253e-09, 6.8918e-08], + [ 7.3109e-07, 4.3772e-07, 2.1048e-07, ..., 2.1700e-07, + 2.2259e-07, 1.4687e-06], + [ 4.4797e-07, 3.1106e-07, 2.2724e-07, ..., 2.6450e-07, + 1.5646e-07, 1.0831e-06]], device='cuda:0') +Epoch 292, bias, value: tensor([ 0.0031, -0.0172, -0.0266, 0.0313, -0.0217, 0.0262, -0.0176, -0.0141, + 0.0058, -0.0317], device='cuda:0'), grad: tensor([-2.5388e-06, 2.5019e-05, 5.5693e-07, 5.9158e-06, -1.6158e-06, + -3.3349e-05, -2.3004e-07, -2.9709e-07, 3.7197e-06, 2.8573e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 214.81, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.4640 re_mapping 0.0042 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 293, weight, value: tensor([[-1.3632e-01, -1.5712e-01, -1.0567e-01, ..., -1.5847e-01, + -1.9248e-02, 8.3247e-02], + [ 6.8751e-02, 7.2107e-02, 5.4228e-02, ..., 1.3315e-01, + -1.9393e-01, -6.7348e-02], + [ 6.0101e-02, 1.3927e-02, -1.5285e-01, ..., -6.5706e-02, + -2.4231e-01, -1.0004e-01], + ..., + [-5.5976e-02, -1.5490e-01, 8.8984e-02, ..., -1.5608e-01, + -5.6295e-02, 1.2639e-04], + [ 2.1577e-02, 1.0883e-01, 1.6343e-01, ..., -1.5673e-01, + -1.6627e-01, -2.3445e-01], + [-1.7575e-01, 1.0570e-02, -2.9813e-01, ..., -1.6323e-01, + -1.2357e-01, 1.1591e-01]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-08, 1.3039e-08, ..., 1.3970e-08, + -1.4110e-06, -4.4592e-06], + [-6.9011e-07, -1.1055e-06, -1.0235e-06, ..., -1.6773e-06, + 2.7940e-09, 2.0489e-08], + [ 1.2573e-07, 2.8498e-07, 3.6228e-07, ..., 4.0326e-07, + 3.7253e-09, 1.1176e-08], + ..., + [ 2.3283e-08, 1.0245e-08, -2.5611e-07, ..., 1.3039e-08, + 0.0000e+00, 1.3970e-08], + [ 1.3970e-08, -3.6322e-08, 3.2596e-08, ..., 2.0489e-08, + 3.7253e-09, 9.3132e-09], + [ 1.8626e-09, 1.8626e-08, 3.8184e-08, ..., 2.7940e-09, + 2.7008e-08, -2.7940e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0029, -0.0173, -0.0267, 0.0315, -0.0216, 0.0237, -0.0148, -0.0140, + 0.0054, -0.0318], device='cuda:0'), grad: tensor([-9.0003e-06, -2.4065e-06, 9.6019e-07, 7.3574e-08, 2.3376e-07, + 1.5739e-07, 1.0632e-05, -5.3365e-07, -2.9616e-07, 2.1141e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 214.80, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4917 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.01 lr 0.00010000 +Epoch 294, weight, value: tensor([[-1.3660e-01, -1.5720e-01, -1.0594e-01, ..., -1.5865e-01, + -1.9034e-02, 8.0817e-02], + [ 6.8832e-02, 7.2308e-02, 5.4293e-02, ..., 1.3454e-01, + -1.9389e-01, -6.7257e-02], + [ 6.0198e-02, 1.3472e-02, -1.5313e-01, ..., -6.6301e-02, + -2.4295e-01, -1.0051e-01], + ..., + [-5.6034e-02, -1.5495e-01, 8.9089e-02, ..., -1.5631e-01, + -5.6435e-02, 1.5250e-04], + [ 2.1636e-02, 1.0864e-01, 1.6340e-01, ..., -1.5763e-01, + -1.6701e-01, -2.3481e-01], + [-1.7586e-01, 1.0716e-02, -2.9873e-01, ..., -1.6336e-01, + -1.2378e-01, 1.1865e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 9.3132e-09, 3.3528e-08, ..., 1.3970e-08, + 9.3132e-09, -1.9558e-08], + [-3.1665e-08, -2.5705e-07, 3.3230e-06, ..., -1.6950e-07, + 5.9232e-07, 1.1483e-06], + [-3.7253e-08, 9.3132e-09, 8.8476e-08, ..., 2.1420e-08, + 1.5832e-08, 1.9558e-08], + ..., + [ 2.7008e-08, 5.7742e-08, -2.2128e-05, ..., -1.7723e-06, + -3.9004e-06, -6.6198e-06], + [-5.9605e-08, -3.3528e-08, -4.3772e-08, ..., 2.8871e-08, + 1.9558e-08, 2.4214e-08], + [ 1.8626e-08, 1.3970e-08, 1.7300e-05, ..., 1.4678e-06, + 3.0417e-06, 5.1223e-06]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0009, -0.0173, -0.0268, 0.0317, -0.0219, 0.0237, -0.0148, -0.0139, + 0.0051, -0.0300], device='cuda:0'), grad: tensor([ 4.4703e-08, 6.4336e-06, 6.3330e-08, 6.4448e-07, 1.2349e-06, + 1.3597e-07, 6.8545e-07, -4.1753e-05, -1.7975e-07, 3.2663e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 214.69, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4798 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.12 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.1365, -0.1576, -0.1056, ..., -0.1582, -0.0177, 0.0815], + [ 0.0685, 0.0725, 0.0539, ..., 0.1348, -0.1946, -0.0660], + [ 0.0600, 0.0133, -0.1538, ..., -0.0666, -0.2434, -0.1005], + ..., + [-0.0556, -0.1553, 0.0896, ..., -0.1566, -0.0557, -0.0011], + [ 0.0220, 0.1089, 0.1634, ..., -0.1574, -0.1672, -0.2352], + [-0.1760, 0.0119, -0.2992, ..., -0.1635, -0.1224, 0.1188]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 5.5879e-09, 5.5879e-09, ..., 3.7253e-09, + 6.5193e-09, -4.6566e-09], + [-4.9360e-08, -5.4017e-08, -1.1269e-07, ..., -6.5193e-08, + 1.8626e-09, 9.3132e-10], + [-3.0734e-08, 2.3283e-08, 3.9116e-08, ..., 1.6764e-08, + 3.7253e-09, 0.0000e+00], + ..., + [ 4.6566e-08, 4.2841e-08, 7.4506e-08, ..., 3.4459e-08, + 1.8626e-09, 5.5879e-09], + [-1.3132e-07, -1.9930e-07, -2.8592e-07, ..., 1.0245e-08, + 3.7253e-09, 2.7940e-09], + [ 1.1176e-08, 1.4901e-08, 2.3283e-08, ..., 6.8918e-08, + 4.6566e-08, -6.5193e-09]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0018, -0.0179, -0.0271, 0.0314, -0.0225, 0.0237, -0.0149, -0.0133, + 0.0051, -0.0296], device='cuda:0'), grad: tensor([ 2.6077e-08, -1.6671e-07, -1.9558e-08, -1.4808e-07, -2.9150e-07, + 8.0839e-07, 3.6322e-08, 1.8906e-07, -7.0315e-07, 2.8312e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 214.28, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5003 re_mapping 0.0039 re_causal 0.0122 /// teacc 98.99 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.1367, -0.1576, -0.1052, ..., -0.1580, -0.0172, 0.0818], + [ 0.0667, 0.0707, 0.0520, ..., 0.1350, -0.1961, -0.0644], + [ 0.0601, 0.0130, -0.1540, ..., -0.0668, -0.2438, -0.1007], + ..., + [-0.0538, -0.1537, 0.0914, ..., -0.1575, -0.0573, -0.0028], + [ 0.0219, 0.1099, 0.1634, ..., -0.1580, -0.1669, -0.2347], + [-0.1762, 0.0120, -0.2998, ..., -0.1637, -0.1231, 0.1188]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.3970e-08, 9.3132e-10, ..., 9.3132e-10, + -3.2596e-08, -2.9616e-07], + [-3.7253e-09, -9.3132e-10, 1.3970e-08, ..., -5.5879e-09, + 2.7940e-09, 1.7695e-08], + [-1.2107e-08, 2.7940e-09, 3.7253e-09, ..., 9.3132e-10, + 1.8626e-09, 3.4459e-08], + ..., + [ 1.4901e-08, 1.0245e-08, -3.2596e-08, ..., 3.7253e-09, + 6.5193e-09, 3.0734e-08], + [-9.3132e-10, -5.5879e-09, -9.3132e-09, ..., -9.3132e-10, + 4.6566e-09, 3.0734e-08], + [ 2.7940e-09, -3.8743e-07, 8.3819e-09, ..., 2.7940e-09, + -4.8056e-07, -2.2724e-06]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0023, -0.0197, -0.0271, 0.0314, -0.0226, 0.0237, -0.0149, -0.0116, + 0.0055, -0.0297], device='cuda:0'), grad: tensor([-3.7067e-07, 6.5193e-08, 3.4459e-08, -6.6124e-08, 4.2021e-06, + 8.7544e-08, 2.2072e-07, 3.2596e-08, 4.2841e-08, -4.2543e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 214.85, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.4897 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.11 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.1368, -0.1577, -0.1049, ..., -0.1581, -0.0171, 0.0820], + [ 0.0669, 0.0708, 0.0520, ..., 0.1354, -0.1972, -0.0646], + [ 0.0603, 0.0128, -0.1542, ..., -0.0670, -0.2442, -0.1006], + ..., + [-0.0540, -0.1538, 0.0916, ..., -0.1578, -0.0559, -0.0024], + [ 0.0221, 0.1110, 0.1635, ..., -0.1583, -0.1667, -0.2343], + [-0.1766, 0.0151, -0.3021, ..., -0.1606, -0.1202, 0.1218]], + device='cuda:0'), grad: tensor([[ 2.7008e-08, 1.5832e-08, 9.3132e-09, ..., 2.7940e-09, + 7.4506e-09, 7.4506e-09], + [-2.8405e-07, -2.3190e-07, -2.7008e-07, ..., -1.4249e-07, + 1.0245e-08, 1.2107e-08], + [ 9.2201e-08, -4.6566e-09, 6.6124e-08, ..., 3.1665e-08, + 7.4506e-09, 8.3819e-09], + ..., + [ 1.1455e-07, 3.3528e-08, 1.3970e-08, ..., 5.5879e-09, + 1.6764e-08, 2.7008e-08], + [-2.7567e-07, -1.1576e-06, -9.7882e-07, ..., 7.7300e-08, + 3.0734e-08, -3.1013e-07], + [ 4.0419e-07, 1.0906e-06, 9.1176e-07, ..., 4.6566e-09, + 1.7416e-07, 4.7684e-07]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0024, -0.0198, -0.0271, 0.0313, -0.0258, 0.0237, -0.0150, -0.0115, + 0.0060, -0.0266], device='cuda:0'), grad: tensor([ 9.4995e-08, -4.4610e-07, 3.3062e-07, -1.4221e-06, 1.8068e-07, + -3.1479e-07, 8.0559e-07, 3.9674e-07, -3.2373e-06, 3.6117e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 214.79, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4846 re_mapping 0.0038 re_causal 0.0120 /// teacc 99.03 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.1372, -0.1578, -0.1055, ..., -0.1584, -0.0171, 0.0820], + [ 0.0669, 0.0703, 0.0520, ..., 0.1335, -0.2006, -0.0645], + [ 0.0600, 0.0127, -0.1549, ..., -0.0673, -0.2445, -0.1008], + ..., + [-0.0539, -0.1538, 0.0917, ..., -0.1580, -0.0561, -0.0024], + [ 0.0223, 0.1110, 0.1635, ..., -0.1594, -0.1671, -0.2347], + [-0.1767, 0.0152, -0.3026, ..., -0.1606, -0.1203, 0.1218]], + device='cuda:0'), grad: tensor([[ 4.0047e-08, -5.4948e-08, 3.6322e-08, ..., 5.5879e-09, + 1.8626e-09, -6.5099e-07], + [-4.1693e-05, -8.0019e-06, 4.5598e-06, ..., -4.8056e-07, + 0.0000e+00, 7.4431e-06], + [ 8.4937e-06, 1.2377e-06, 6.0201e-06, ..., 3.9302e-07, + 4.6566e-09, 7.9814e-07], + ..., + [ 3.2723e-05, 6.6943e-06, -1.1705e-05, ..., 8.0094e-08, + 0.0000e+00, -9.1791e-06], + [-8.3819e-08, -1.1735e-07, -2.3283e-07, ..., 2.7940e-09, + 9.3132e-10, 1.4901e-08], + [ 4.3772e-08, 6.3330e-08, 8.2236e-07, ..., 2.7940e-09, + 9.3132e-10, 1.4957e-06]], device='cuda:0') diff --git a/Meta-causal/code-withStyleAttack/66539.error b/Meta-causal/code-withStyleAttack/66539.error new file mode 100644 index 0000000000000000000000000000000000000000..00da0f285352fc1c57390b613b1dc4eccf2320e7 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66539.error @@ -0,0 +1,303 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3 +error libmamba Could not lock non-existing path '/scratch/yuqian_fu/micromamba/pkgs' +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-uvqtxnlanqi3 + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 h4bc722e_7 conda-forge + + ca-certificates 2024.7.4 hbcca054_0 conda-forge + + certifi 2024.7.4 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + click 8.1.7 unix_pyh707e725_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.1.7 0 nvidia + + libcurand 10.3.6.82 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 14.1.0 h77fa898_0 conda-forge + + libgfortran-ng 14.1.0 h69a702a_0 conda-forge + + libgfortran5 14.1.0 hc5f4f2c_0 /work/conda/cache/conda-forge + + libhwloc 2.11.1 default_hecaa2ac_1000 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 14.1.0 hc0a3c3a_0 /work/conda/cache/conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4bc722e_2 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 /work/conda/cache/conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.3.1 py3.11_cuda12.1_cudnn8.9.2_0 /work/conda/cache/pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 71.0.1 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.13.0 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h434a139_3 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 2.3.1 py311 /work/conda/cache/pytorch + + torchvision 0.18.1 py311_cu121 /work/conda/cache/pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.23.0 py311h5cd10c7_0 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 119 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.7.4-hbcca054_0 +Linking libgcc-ng-14.1.0-h77fa898_0 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4bc722e_2 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-14.1.0-hc5f4f2c_0 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-14.1.0-hc0a3c3a_0 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-h4bc722e_7 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-14.1.0-h69a702a_0 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.11.1-default_hecaa2ac_1000 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h434a139_3 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.82-0 +Linking libcufile-1.10.1.7-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-71.0.1-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.7.4-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking click-8.1.7-unix_pyh707e725_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.23.0-py311h5cd10c7_0 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.13.0-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.3.1-py3.11_cuda12.1_cudnn8.9.2_0 +Linking torchtriton-2.3.1-py311 +Linking torchvision-0.18.1-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-uvqtxnlanqi3 + +Or to execute a single command in this environment, use: + + mamba run -n auto-uvqtxnlanqi3 mycommand + +Installing pip packages +WARNING: The candidate selected for download or install is a yanked version: 'opencv-python' candidate (version 4.5.5.62 at https://files.pythonhosted.org/packages/9d/98/36bfcbff30da27dd6922ed73ca7802c37d87f77daf4c569da3dcb87b4296/opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from https://pypi.org/simple/opencv-python/) (requires-python:>=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +run_my_joint_test.sh: line 33: randn}_str3_WithStyleAttackExp1_onlyblock2: command not found +slurmstepd: error: *** STEP 66539.0 ON gcp-us-1 CANCELLED AT 2024-07-21T15:44:21 DUE TO TIME LIMIT *** +slurmstepd: error: *** JOB 66539 ON gcp-us-1 CANCELLED AT 2024-07-21T15:44:21 DUE TO TIME LIMIT *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. diff --git a/Meta-causal/code-withStyleAttack/66539.log b/Meta-causal/code-withStyleAttack/66539.log new file mode 100644 index 0000000000000000000000000000000000000000..48f06c01c3afbcaf54806c29008ae087a125da50 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66539.log @@ -0,0 +1,19244 @@ +Collecting h5py>=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 3.9 MB/s eta 0:00:00 + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.0) +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 3.9 MB/s eta 0:00:00 +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 11.3 MB/s eta 0:00:00 +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (71.0.1) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.3.1) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.18.1) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.24.0-py3-none-any.whl.metadata (13 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 18.9 MB/s eta 0:00:00 +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.13.0) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.7.4) +Requirement already satisfied: mpmath<1.4,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 80.7 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 195.2 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 99.0 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 185.5 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 32.9 MB/s eta 0:00:00 +Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 142.5 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 41.6 MB/s eta 0:00:00 +Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.7 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.7/5.7 MB 186.4 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 31.0 MB/s eta 0:00:00 +Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl (294 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 76.8 MB/s eta 0:00:00 +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 181.2 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 62.6 MB/s eta 0:00:00 +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.24.0-py3-none-any.whl (419 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 419.0/419.0 kB 96.6 MB/s eta 0:00:00 +Downloading packaging-24.1-py3-none-any.whl (53 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 10.3 MB/s eta 0:00:00 +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 143.5 MB/s eta 0:00:00 +Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 177.6/177.6 kB 50.9 MB/s eta 0:00:00 +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 26.3 MB/s eta 0:00:00 +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94508 sha256=310dbc935b23fb0dacb072b058c1f197374aa7241d1279503e2bd3bb0bab5ea9 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.65.1 h5py-3.11.0 huggingface_hub-0.24.0 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.3 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_onlyblock2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-2.5139e-02, -1.7924e-02, 2.9307e-02, ..., -6.4647e-03, + -1.8291e-02, -2.4441e-02], + [-2.2718e-02, -2.9409e-02, 2.0061e-02, ..., -7.4217e-03, + 3.3684e-05, 2.0003e-02], + [ 2.5372e-02, 1.9253e-02, -4.3399e-03, ..., -2.4109e-02, + -2.7648e-02, -2.6141e-02], + ..., + [-1.5319e-02, -5.5781e-03, -2.4777e-02, ..., 1.2113e-02, + 9.7509e-03, 5.7319e-03], + [-2.1644e-02, -3.1244e-02, 2.4834e-03, ..., 4.4214e-03, + 4.7801e-03, 2.4475e-02], + [ 4.0269e-03, -9.8763e-03, -5.6816e-03, ..., 2.5718e-02, + 1.1947e-02, -2.4943e-02]], device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0088, 0.0188, 0.0148, -0.0022, 0.0142, -0.0066, -0.0144, -0.0159, + 0.0256, -0.0067], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 278.09, cls_loss 1.3028 cls_loss_mapping 1.8310 cls_loss_causal 2.2140 re_mapping 0.1459 re_causal 0.1544 /// teacc 87.55 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0304, -0.0127, 0.0363, ..., -0.0007, -0.0263, -0.0246], + [-0.0255, -0.0372, 0.0133, ..., -0.0162, -0.0066, 0.0178], + [ 0.0306, 0.0129, -0.0094, ..., -0.0265, -0.0362, -0.0276], + ..., + [-0.0109, -0.0030, -0.0264, ..., 0.0158, 0.0134, 0.0089], + [-0.0236, -0.0367, -0.0051, ..., -0.0020, 0.0042, 0.0201], + [ 0.0010, -0.0127, -0.0044, ..., 0.0254, 0.0161, -0.0236]], + device='cuda:0'), grad: tensor([[ 0.0012, 0.0124, -0.0088, ..., -0.0057, 0.0126, 0.0000], + [ 0.0092, 0.0006, 0.0007, ..., 0.0030, 0.0177, 0.0000], + [-0.0139, -0.0013, 0.0054, ..., 0.0070, 0.0114, 0.0000], + ..., + [-0.0021, 0.0012, 0.0028, ..., 0.0145, 0.0536, 0.0000], + [ 0.0046, 0.0085, -0.0041, ..., -0.0115, -0.0413, 0.0000], + [ 0.0007, 0.0008, -0.0082, ..., -0.0449, -0.0875, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0077, 0.0202, 0.0141, -0.0018, 0.0147, -0.0067, -0.0156, -0.0145, + 0.0248, -0.0068], device='cuda:0'), grad: tensor([ 0.0065, 0.0371, -0.0193, 0.0364, 0.0056, -0.0031, -0.0156, 0.0386, + -0.0386, -0.0476], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 277.76, cls_loss 0.4191 cls_loss_mapping 0.7467 cls_loss_causal 1.8580 re_mapping 0.2066 re_causal 0.2696 /// teacc 91.72 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0329, -0.0113, 0.0411, ..., 0.0012, -0.0298, -0.0248], + [-0.0284, -0.0379, 0.0104, ..., -0.0198, -0.0092, 0.0177], + [ 0.0338, 0.0105, -0.0126, ..., -0.0277, -0.0388, -0.0277], + ..., + [-0.0079, -0.0035, -0.0299, ..., 0.0157, 0.0133, 0.0091], + [-0.0279, -0.0409, -0.0081, ..., -0.0057, 0.0051, 0.0199], + [ 0.0020, -0.0143, -0.0029, ..., 0.0280, 0.0175, -0.0235]], + device='cuda:0'), grad: tensor([[ 0.0032, 0.0082, 0.0032, ..., 0.0125, 0.0078, 0.0000], + [ 0.0008, 0.0021, 0.0004, ..., 0.0027, 0.0031, 0.0000], + [-0.0242, 0.0007, 0.0025, ..., 0.0046, 0.0020, 0.0000], + ..., + [ 0.0063, -0.0016, 0.0081, ..., -0.0039, -0.0333, 0.0000], + [ 0.0065, -0.0195, -0.0086, ..., -0.0059, -0.0210, 0.0000], + [-0.0145, -0.0126, -0.0169, ..., -0.0911, -0.0687, 0.0000]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0077, 0.0202, 0.0136, -0.0020, 0.0149, -0.0061, -0.0164, -0.0147, + 0.0249, -0.0064], device='cuda:0'), grad: tensor([ 0.0165, -0.0090, -0.0070, 0.0492, 0.0646, -0.0028, -0.0120, -0.0011, + -0.0295, -0.0690], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 278.39, cls_loss 0.2513 cls_loss_mapping 0.4353 cls_loss_causal 1.7004 re_mapping 0.1454 re_causal 0.2390 /// teacc 94.84 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0348, -0.0109, 0.0455, ..., 0.0026, -0.0317, -0.0246], + [-0.0302, -0.0395, 0.0067, ..., -0.0231, -0.0112, 0.0177], + [ 0.0370, 0.0089, -0.0152, ..., -0.0285, -0.0405, -0.0277], + ..., + [-0.0074, -0.0020, -0.0323, ..., 0.0153, 0.0135, 0.0091], + [-0.0308, -0.0436, -0.0095, ..., -0.0078, 0.0055, 0.0199], + [ 0.0016, -0.0162, -0.0015, ..., 0.0302, 0.0183, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.3237e-03, -5.3749e-03, -1.2703e-02, ..., -7.3166e-03, + 2.7561e-03, -2.6379e-03], + [ 4.9171e-03, 7.1478e-04, 3.9625e-04, ..., 2.2221e-03, + 4.5013e-03, 5.6595e-05], + [-3.6255e-02, 3.2949e-04, -3.7861e-04, ..., -4.9829e-04, + -3.9101e-03, 2.6420e-05], + ..., + [ 9.7122e-03, 2.9716e-03, 8.1558e-03, ..., 6.2218e-03, + 1.6069e-03, 1.4963e-03], + [ 3.3798e-03, 1.9779e-03, -4.4525e-05, ..., 3.2654e-03, + 2.7008e-03, 6.1929e-05], + [-1.4162e-04, -1.3313e-02, 2.3365e-03, ..., 3.2940e-03, + 5.4016e-03, 3.2234e-04]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0080, 0.0203, 0.0137, -0.0024, 0.0150, -0.0061, -0.0166, -0.0148, + 0.0247, -0.0060], device='cuda:0'), grad: tensor([-0.0065, 0.0036, -0.0241, 0.0354, -0.0230, 0.0006, 0.0029, 0.0122, + 0.0023, -0.0033], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 278.35, cls_loss 0.1958 cls_loss_mapping 0.3174 cls_loss_causal 1.4986 re_mapping 0.1158 re_causal 0.2061 /// teacc 95.47 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0366, -0.0104, 0.0495, ..., 0.0037, -0.0338, -0.0216], + [-0.0309, -0.0406, 0.0027, ..., -0.0262, -0.0129, 0.0146], + [ 0.0399, 0.0072, -0.0172, ..., -0.0291, -0.0429, -0.0341], + ..., + [-0.0074, -0.0010, -0.0350, ..., 0.0149, 0.0139, 0.0096], + [-0.0323, -0.0443, -0.0099, ..., -0.0092, 0.0056, 0.0160], + [ 0.0006, -0.0168, -0.0014, ..., 0.0315, 0.0185, -0.0248]], + device='cuda:0'), grad: tensor([[ 5.5962e-03, 6.8486e-05, 1.6289e-03, ..., 5.3787e-04, + 1.3046e-03, 4.5466e-04], + [ 1.1168e-03, 5.2834e-04, 3.5739e-04, ..., 9.1553e-04, + 1.8396e-03, 5.5552e-05], + [-3.1204e-02, -6.1646e-03, -7.9727e-03, ..., -3.8795e-03, + -3.2654e-03, 2.3782e-04], + ..., + [ 2.9316e-03, 6.1989e-04, 1.2541e-03, ..., 7.2136e-03, + 1.1116e-02, 1.6138e-05], + [ 5.4932e-03, 2.3537e-03, 2.2945e-03, ..., 5.1384e-03, + 6.1378e-03, 3.2496e-04], + [ 6.5470e-04, -3.0403e-03, -1.6451e-04, ..., -3.3173e-02, + -5.7648e-02, 2.9624e-05]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0082, 0.0206, 0.0134, -0.0022, 0.0152, -0.0065, -0.0171, -0.0146, + 0.0250, -0.0064], device='cuda:0'), grad: tensor([ 6.0921e-03, 6.5416e-06, -2.9526e-02, 1.8600e-02, 2.3712e-02, + 9.3536e-03, -5.4054e-03, 9.9182e-03, 1.0880e-02, -4.3610e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 277.86, cls_loss 0.1580 cls_loss_mapping 0.2540 cls_loss_causal 1.4358 re_mapping 0.0939 re_causal 0.1804 /// teacc 96.88 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0371, -0.0106, 0.0526, ..., 0.0051, -0.0359, -0.0241], + [-0.0316, -0.0421, 0.0021, ..., -0.0296, -0.0148, 0.0132], + [ 0.0419, 0.0069, -0.0193, ..., -0.0297, -0.0445, -0.0408], + ..., + [-0.0070, -0.0005, -0.0375, ..., 0.0142, 0.0141, 0.0103], + [-0.0341, -0.0450, -0.0113, ..., -0.0109, 0.0054, 0.0093], + [-0.0003, -0.0177, -0.0017, ..., 0.0328, 0.0191, -0.0273]], + device='cuda:0'), grad: tensor([[ 4.4107e-04, 2.5392e-04, 6.0539e-03, ..., 4.4365e-03, + 3.5172e-03, 4.9829e-04], + [ 1.0284e-02, 6.4075e-05, -1.1568e-03, ..., 1.1606e-03, + 1.6823e-03, 1.9205e-04], + [-1.8890e-02, 7.5340e-04, 4.4060e-03, ..., 4.4518e-03, + 5.4245e-03, 7.7057e-04], + ..., + [ 1.7633e-03, -4.0680e-05, 1.3533e-03, ..., 4.3726e-04, + -2.3251e-03, 3.4243e-05], + [ 3.8528e-03, -4.7684e-04, -1.2901e-02, ..., -1.0254e-02, + -1.4946e-02, -1.4949e-04], + [ 2.5535e-04, 1.4877e-04, 2.3727e-03, ..., 4.8332e-03, + 5.4779e-03, 1.2046e-04]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0083, 0.0203, 0.0133, -0.0024, 0.0153, -0.0067, -0.0168, -0.0141, + 0.0249, -0.0064], device='cuda:0'), grad: tensor([ 0.0097, 0.0146, -0.0173, 0.0037, -0.0011, 0.0049, 0.0012, 0.0003, + -0.0247, 0.0086], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 278.31, cls_loss 0.1141 cls_loss_mapping 0.1882 cls_loss_causal 1.3510 re_mapping 0.0799 re_causal 0.1651 /// teacc 97.10 lr 0.00010000 +Epoch 7, weight, value: tensor([[-3.8017e-02, -1.0378e-02, 5.4926e-02, ..., 6.2616e-03, + -3.7454e-02, -2.8002e-02], + [-3.1948e-02, -4.2724e-02, 2.8633e-05, ..., -3.2009e-02, + -1.6452e-02, 1.2362e-02], + [ 4.4339e-02, 5.8609e-03, -2.1181e-02, ..., -3.0784e-02, + -4.6275e-02, -4.8904e-02], + ..., + [-7.3374e-03, -3.2995e-04, -3.8411e-02, ..., 1.3766e-02, + 1.4295e-02, 1.2877e-02], + [-3.6592e-02, -4.5429e-02, -1.0921e-02, ..., -1.2124e-02, + 5.6063e-03, 5.9953e-03], + [-6.4927e-04, -1.8044e-02, -2.7281e-03, ..., 3.3696e-02, + 1.9460e-02, -2.9106e-02]], device='cuda:0'), grad: tensor([[ 1.0413e-04, 3.6061e-05, -2.1210e-03, ..., -6.5136e-04, + 1.0719e-03, 1.0177e-05], + [ 3.7932e-04, 7.9751e-05, -3.3784e-04, ..., 8.4877e-05, + -9.9301e-05, -9.6142e-05], + [ 3.6430e-04, 1.9574e-04, 1.1282e-03, ..., 6.9332e-04, + 1.6088e-03, 9.3162e-05], + ..., + [ 6.6137e-04, 9.1434e-05, 5.6171e-04, ..., 1.2026e-03, + 4.3559e-04, 2.3350e-05], + [ 1.5039e-03, 1.6809e-04, 8.0884e-05, ..., 6.8474e-04, + 7.5912e-04, 1.5819e-04], + [-5.2309e-04, 4.4680e-04, -8.1062e-04, ..., -2.5253e-03, + -1.1396e-03, 2.9951e-05]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0084, 0.0201, 0.0129, -0.0021, 0.0154, -0.0069, -0.0170, -0.0142, + 0.0253, -0.0063], device='cuda:0'), grad: tensor([ 0.0002, -0.0031, 0.0030, 0.0030, 0.0009, -0.0077, 0.0006, 0.0011, + 0.0027, -0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 6, time 261.34, cls_loss 0.1024 cls_loss_mapping 0.1556 cls_loss_causal 1.2240 re_mapping 0.0693 re_causal 0.1462 /// teacc 97.09 lr 0.00010000 +Epoch 8, weight, value: tensor([[-3.9099e-02, -9.8635e-03, 5.6857e-02, ..., 7.2413e-03, + -3.8952e-02, -2.9268e-02], + [-3.2591e-02, -4.2773e-02, 2.7749e-04, ..., -3.3093e-02, + -1.7628e-02, 1.0653e-02], + [ 4.6330e-02, 4.7390e-03, -2.2939e-02, ..., -3.1849e-02, + -4.7772e-02, -5.3329e-02], + ..., + [-8.2794e-03, -7.3237e-05, -4.0368e-02, ..., 1.3732e-02, + 1.5140e-02, 1.3445e-02], + [-3.8000e-02, -4.5898e-02, -1.0978e-02, ..., -1.3401e-02, + 5.6859e-03, 2.0614e-03], + [-1.0782e-03, -1.8746e-02, -2.6144e-03, ..., 3.4746e-02, + 1.9653e-02, -3.0435e-02]], device='cuda:0'), grad: tensor([[ 2.1877e-03, 8.6248e-05, -1.1658e-02, ..., -1.8287e-04, + 5.7077e-04, -1.3914e-03], + [-3.0136e-02, 1.0991e-04, 7.0572e-04, ..., 5.9986e-04, + 3.9196e-04, 5.3078e-05], + [ 1.3916e-02, -4.7994e-04, 1.6079e-03, ..., 3.2091e-04, + 7.0524e-04, 3.0041e-04], + ..., + [ 1.6842e-03, 9.1076e-05, 5.8031e-04, ..., 7.3385e-04, + 7.0274e-05, 3.9458e-05], + [ 2.7218e-03, 2.7037e-04, 2.6302e-03, ..., 1.0633e-03, + 1.2598e-03, 3.5977e-04], + [ 5.4407e-04, 2.8968e-04, 1.4687e-03, ..., 1.1871e-02, + 1.3969e-02, 9.3043e-05]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0085, 0.0201, 0.0130, -0.0020, 0.0154, -0.0069, -0.0171, -0.0142, + 0.0253, -0.0065], device='cuda:0'), grad: tensor([-0.0052, -0.0317, 0.0171, 0.0055, -0.0139, 0.0022, 0.0048, 0.0018, + 0.0060, 0.0133], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 278.54, cls_loss 0.1057 cls_loss_mapping 0.1583 cls_loss_causal 1.1747 re_mapping 0.0598 re_causal 0.1285 /// teacc 97.21 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0406, -0.0095, 0.0590, ..., 0.0081, -0.0404, -0.0299], + [-0.0323, -0.0436, 0.0006, ..., -0.0354, -0.0193, 0.0095], + [ 0.0482, 0.0039, -0.0247, ..., -0.0325, -0.0494, -0.0596], + ..., + [-0.0087, 0.0006, -0.0414, ..., 0.0133, 0.0157, 0.0181], + [-0.0390, -0.0468, -0.0114, ..., -0.0143, 0.0059, -0.0013], + [-0.0018, -0.0192, -0.0028, ..., 0.0356, 0.0198, -0.0341]], + device='cuda:0'), grad: tensor([[ 5.8794e-04, 7.5912e-04, -1.3614e-04, ..., -2.4900e-05, + 1.1492e-03, 7.3862e-04], + [ 1.2827e-03, 1.2815e-04, 5.6744e-04, ..., 1.8561e-04, + 5.0449e-04, 2.8953e-05], + [-5.6496e-03, 2.6608e-04, -2.0275e-03, ..., 1.8549e-04, + 7.8392e-04, 9.2387e-05], + ..., + [ 1.0729e-03, 2.4045e-04, 2.0206e-04, ..., 1.8377e-03, + 3.2005e-03, 3.9227e-06], + [ 4.5166e-03, 2.2964e-03, 7.8440e-04, ..., 1.4420e-03, + 7.7400e-03, 1.1414e-04], + [ 3.5763e-04, 4.3702e-04, 4.1962e-04, ..., 1.7462e-03, + 2.3899e-03, 9.5591e-06]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0085, 0.0200, 0.0130, -0.0020, 0.0157, -0.0071, -0.0174, -0.0139, + 0.0256, -0.0067], device='cuda:0'), grad: tensor([ 0.0024, 0.0008, -0.0057, -0.0112, -0.0064, -0.0063, -0.0002, 0.0040, + 0.0181, 0.0044], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 277.99, cls_loss 0.0795 cls_loss_mapping 0.1290 cls_loss_causal 1.1320 re_mapping 0.0545 re_causal 0.1243 /// teacc 97.73 lr 0.00010000 +Epoch 10, weight, value: tensor([[-4.1535e-02, -9.4166e-03, 6.0846e-02, ..., 9.1568e-03, + -4.1750e-02, -3.0630e-02], + [-3.3310e-02, -4.4122e-02, 4.3387e-07, ..., -3.6372e-02, + -2.0060e-02, 1.0328e-02], + [ 4.9945e-02, 3.2577e-03, -2.6320e-02, ..., -3.3269e-02, + -5.0451e-02, -6.3274e-02], + ..., + [-9.1274e-03, 8.3013e-04, -4.2307e-02, ..., 1.2600e-02, + 1.5928e-02, 1.9006e-02], + [-4.0328e-02, -4.7789e-02, -1.1606e-02, ..., -1.5415e-02, + 5.8589e-03, -3.0546e-03], + [-2.0778e-03, -1.9355e-02, -3.0932e-03, ..., 3.6396e-02, + 1.9966e-02, -3.5436e-02]], device='cuda:0'), grad: tensor([[ 1.4677e-03, -1.9681e-04, -1.6832e-04, ..., 5.5122e-04, + 5.0974e-04, 1.7202e-04], + [ 2.7485e-03, 1.1212e-04, 4.4012e-04, ..., 7.2813e-04, + 2.2495e-04, 3.8552e-04], + [ 3.2482e-03, 2.2757e-04, 1.0004e-03, ..., 4.3416e-04, + 1.1978e-03, 1.2827e-04], + ..., + [ 3.0422e-03, 3.4189e-04, 2.8586e-04, ..., 1.9875e-03, + 1.5259e-03, 8.3745e-06], + [ 1.1578e-03, 5.9366e-04, -6.7997e-04, ..., 1.7595e-04, + 2.3308e-03, 4.9973e-04], + [ 3.0828e-04, 4.3440e-04, 2.8706e-04, ..., -4.2000e-03, + -3.5553e-03, 2.5362e-05]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0086, 0.0199, 0.0132, -0.0021, 0.0161, -0.0068, -0.0179, -0.0140, + 0.0254, -0.0068], device='cuda:0'), grad: tensor([ 0.0018, 0.0026, 0.0091, -0.0127, -0.0003, -0.0030, -0.0043, 0.0060, + 0.0044, -0.0034], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 278.37, cls_loss 0.0743 cls_loss_mapping 0.1175 cls_loss_causal 1.1044 re_mapping 0.0474 re_causal 0.1133 /// teacc 97.74 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0427, -0.0095, 0.0623, ..., 0.0102, -0.0433, -0.0320], + [-0.0329, -0.0444, 0.0002, ..., -0.0385, -0.0215, 0.0090], + [ 0.0513, 0.0026, -0.0286, ..., -0.0345, -0.0516, -0.0673], + ..., + [-0.0098, 0.0016, -0.0438, ..., 0.0120, 0.0166, 0.0189], + [-0.0412, -0.0486, -0.0122, ..., -0.0165, 0.0059, -0.0073], + [-0.0024, -0.0201, -0.0031, ..., 0.0375, 0.0201, -0.0384]], + device='cuda:0'), grad: tensor([[ 9.6917e-05, 3.4928e-05, -3.9721e-04, ..., -1.0210e-04, + 2.5249e-04, 1.1760e-04], + [ 2.7204e-04, 1.9777e-04, -1.1319e-04, ..., 8.3268e-05, + 3.0732e-04, -9.3639e-05], + [-7.8011e-04, 4.4274e-04, 1.1444e-04, ..., 3.2282e-04, + 6.3086e-04, 5.2482e-05], + ..., + [ 5.2595e-04, 8.8215e-04, 4.8727e-05, ..., 7.7629e-04, + 5.5742e-04, 1.8343e-05], + [ 8.2493e-04, 1.6375e-03, 1.3483e-04, ..., 1.0996e-03, + 1.4086e-03, 1.3053e-04], + [ 7.4863e-04, 1.7424e-03, -1.9923e-05, ..., 2.5368e-04, + 8.8930e-04, 1.8060e-05]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0086, 0.0199, 0.0129, -0.0022, 0.0162, -0.0071, -0.0175, -0.0136, + 0.0255, -0.0070], device='cuda:0'), grad: tensor([ 0.0002, -0.0015, 0.0008, -0.0086, 0.0011, -0.0040, 0.0048, 0.0016, + 0.0034, 0.0024], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 278.08, cls_loss 0.0635 cls_loss_mapping 0.0979 cls_loss_causal 1.0332 re_mapping 0.0441 re_causal 0.1056 /// teacc 98.03 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0436, -0.0095, 0.0641, ..., 0.0113, -0.0443, -0.0322], + [-0.0339, -0.0451, 0.0005, ..., -0.0406, -0.0227, 0.0110], + [ 0.0527, 0.0015, -0.0296, ..., -0.0350, -0.0527, -0.0709], + ..., + [-0.0098, 0.0019, -0.0448, ..., 0.0112, 0.0166, 0.0204], + [-0.0422, -0.0491, -0.0125, ..., -0.0172, 0.0063, -0.0073], + [-0.0030, -0.0206, -0.0040, ..., 0.0381, 0.0202, -0.0410]], + device='cuda:0'), grad: tensor([[ 4.1556e-04, -3.6597e-05, -1.6117e-03, ..., -8.3685e-04, + 4.4894e-04, 7.0512e-05], + [ 5.7840e-04, 6.2466e-05, 1.1551e-04, ..., 3.1137e-04, + 6.4898e-04, 1.5736e-05], + [-3.5229e-03, 1.0532e-04, 1.5974e-04, ..., 6.2275e-04, + -5.0592e-04, 6.9886e-06], + ..., + [ 8.4019e-04, 4.0889e-05, 2.3222e-04, ..., 1.8370e-04, + -1.1158e-03, 1.8388e-05], + [ 1.0710e-03, 1.6105e-04, 2.1911e-04, ..., 2.5868e-04, + 2.3985e-04, 8.0109e-05], + [ 2.2626e-04, 1.4663e-04, 1.1533e-04, ..., 1.3771e-03, + 2.9659e-03, 3.3140e-05]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0088, 0.0198, 0.0132, -0.0021, 0.0161, -0.0070, -0.0178, -0.0139, + 0.0257, -0.0072], device='cuda:0'), grad: tensor([-3.6538e-05, 9.8801e-04, -4.3449e-03, 1.6727e-03, -2.3022e-03, + -3.0384e-03, 1.3638e-03, 1.5247e-04, 1.5469e-03, 4.0016e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 261.47, cls_loss 0.0624 cls_loss_mapping 0.0958 cls_loss_causal 1.0565 re_mapping 0.0389 re_causal 0.0994 /// teacc 97.79 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0440, -0.0096, 0.0652, ..., 0.0121, -0.0456, -0.0344], + [-0.0341, -0.0454, -0.0007, ..., -0.0415, -0.0237, 0.0070], + [ 0.0539, 0.0012, -0.0313, ..., -0.0358, -0.0536, -0.0745], + ..., + [-0.0098, 0.0021, -0.0458, ..., 0.0105, 0.0171, 0.0210], + [-0.0433, -0.0496, -0.0128, ..., -0.0182, 0.0064, -0.0087], + [-0.0037, -0.0210, -0.0038, ..., 0.0390, 0.0205, -0.0414]], + device='cuda:0'), grad: tensor([[ 4.0030e-04, 8.2195e-05, -1.4377e-04, ..., 3.1376e-04, + 6.1464e-04, 1.6844e-04], + [ 9.2030e-05, 3.9756e-05, -2.5349e-03, ..., 1.7917e-04, + 5.6458e-04, -1.2302e-03], + [-1.4591e-03, -8.8692e-05, 2.9683e-04, ..., -2.8778e-06, + 3.1781e-04, 1.0830e-04], + ..., + [ 2.0719e-04, 8.3566e-05, -3.7789e-04, ..., -5.7373e-03, + -9.7427e-03, 3.3259e-05], + [ 2.9826e-04, 1.6093e-04, 3.8552e-04, ..., 6.3276e-04, + 3.8552e-04, 3.4690e-04], + [ 2.2733e-04, 1.9467e-04, 4.4298e-04, ..., 2.5444e-03, + 6.6986e-03, 4.7654e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0088, 0.0193, 0.0135, -0.0023, 0.0161, -0.0072, -0.0175, -0.0137, + 0.0259, -0.0072], device='cuda:0'), grad: tensor([ 1.3981e-03, -4.5013e-03, -5.0366e-05, 4.0245e-04, 2.1725e-03, + -6.3753e-04, 3.2921e-03, -1.2642e-02, 1.6241e-03, 8.9493e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 261.50, cls_loss 0.0551 cls_loss_mapping 0.0888 cls_loss_causal 0.9801 re_mapping 0.0372 re_causal 0.0957 /// teacc 98.02 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0443, -0.0095, 0.0664, ..., 0.0132, -0.0467, -0.0357], + [-0.0349, -0.0454, -0.0004, ..., -0.0420, -0.0250, 0.0076], + [ 0.0552, 0.0005, -0.0323, ..., -0.0368, -0.0546, -0.0768], + ..., + [-0.0097, 0.0026, -0.0470, ..., 0.0097, 0.0179, 0.0199], + [-0.0442, -0.0501, -0.0128, ..., -0.0195, 0.0066, -0.0099], + [-0.0043, -0.0215, -0.0041, ..., 0.0399, 0.0207, -0.0432]], + device='cuda:0'), grad: tensor([[ 3.6001e-04, -3.1042e-04, -1.2617e-03, ..., -5.1737e-04, + 1.0687e-04, 8.9705e-05], + [ 1.8895e-04, 1.1486e-04, -7.5400e-05, ..., -3.4595e-04, + 3.3140e-04, -3.0065e-04], + [-3.4389e-03, 5.5552e-05, 3.3021e-04, ..., 2.0432e-04, + 2.5415e-04, 1.7834e-04], + ..., + [ 6.9571e-04, -3.4779e-05, 9.2268e-05, ..., 8.6069e-05, + -9.0027e-04, 6.7770e-05], + [ 7.0763e-04, 1.7381e-04, 1.1265e-04, ..., 1.9109e-04, + 3.6502e-04, 1.5438e-04], + [ 2.4986e-04, 1.5938e-04, 1.8418e-04, ..., 3.6073e-04, + 7.3242e-04, 2.7269e-05]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0089, 0.0196, 0.0132, -0.0024, 0.0160, -0.0072, -0.0178, -0.0134, + 0.0259, -0.0072], device='cuda:0'), grad: tensor([-0.0003, -0.0010, -0.0015, -0.0215, 0.0004, 0.0201, 0.0006, -0.0001, + 0.0016, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 277.69, cls_loss 0.0544 cls_loss_mapping 0.0889 cls_loss_causal 0.9948 re_mapping 0.0349 re_causal 0.0915 /// teacc 98.39 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0453, -0.0097, 0.0675, ..., 0.0136, -0.0478, -0.0372], + [-0.0352, -0.0457, 0.0002, ..., -0.0425, -0.0259, 0.0062], + [ 0.0565, -0.0003, -0.0341, ..., -0.0375, -0.0554, -0.0794], + ..., + [-0.0104, 0.0026, -0.0483, ..., 0.0091, 0.0184, 0.0198], + [-0.0454, -0.0504, -0.0125, ..., -0.0198, 0.0068, -0.0111], + [-0.0049, -0.0217, -0.0042, ..., 0.0408, 0.0209, -0.0434]], + device='cuda:0'), grad: tensor([[-4.3201e-04, -6.2609e-04, -4.6959e-03, ..., -1.4410e-03, + 8.6427e-05, 1.4946e-05], + [ 1.6689e-05, 5.8338e-06, -2.2709e-04, ..., 1.4818e-04, + 3.1734e-04, 3.0939e-06], + [ 3.0065e-04, 4.7743e-05, 1.2465e-03, ..., 1.0471e-03, + 4.9400e-04, 2.3559e-05], + ..., + [ 1.0170e-05, 1.5236e-05, 9.4354e-05, ..., 7.2289e-04, + 2.7251e-04, 4.2189e-07], + [ 1.3280e-04, 1.4603e-04, 5.8794e-04, ..., 1.1663e-03, + 1.7929e-03, 1.3083e-05], + [ 1.1361e-04, 8.7440e-05, 4.3488e-04, ..., 4.9171e-03, + 5.4398e-03, 2.3935e-06]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0089, 0.0194, 0.0131, -0.0023, 0.0156, -0.0069, -0.0179, -0.0133, + 0.0260, -0.0071], device='cuda:0'), grad: tensor([-0.0031, -0.0012, 0.0020, 0.0002, -0.0103, 0.0002, 0.0014, 0.0008, + 0.0031, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 261.51, cls_loss 0.0468 cls_loss_mapping 0.0780 cls_loss_causal 0.9790 re_mapping 0.0337 re_causal 0.0910 /// teacc 98.19 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0461, -0.0094, 0.0684, ..., 0.0140, -0.0490, -0.0383], + [-0.0356, -0.0462, 0.0001, ..., -0.0432, -0.0264, 0.0071], + [ 0.0580, -0.0005, -0.0350, ..., -0.0381, -0.0565, -0.0815], + ..., + [-0.0113, 0.0020, -0.0492, ..., 0.0085, 0.0186, 0.0188], + [-0.0460, -0.0505, -0.0123, ..., -0.0202, 0.0070, -0.0118], + [-0.0054, -0.0224, -0.0050, ..., 0.0412, 0.0208, -0.0452]], + device='cuda:0'), grad: tensor([[ 3.7551e-05, -1.6391e-04, -1.1444e-03, ..., 3.4362e-05, + 2.5535e-04, 3.3528e-05], + [ 8.3637e-04, 3.2759e-04, 9.6083e-05, ..., 3.2210e-04, + 9.4128e-04, 1.5283e-04], + [-1.0300e-03, 6.8998e-04, 2.7156e-04, ..., 5.3406e-04, + 1.3151e-03, 4.6015e-05], + ..., + [ 7.7069e-05, -2.5558e-03, 4.6402e-05, ..., -3.4714e-03, + -7.3357e-03, 2.9430e-06], + [ 6.3658e-05, 5.4169e-04, 2.6655e-04, ..., 5.7888e-04, + 1.2693e-03, 2.2066e-04], + [ 1.8597e-05, 1.2863e-04, 5.6267e-05, ..., -6.3181e-04, + -4.5633e-04, 1.0006e-05]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0087, 0.0196, 0.0131, -0.0020, 0.0160, -0.0073, -0.0178, -0.0135, + 0.0264, -0.0076], device='cuda:0'), grad: tensor([-2.8801e-04, 2.3460e-03, 2.4471e-03, 7.8773e-04, 7.5912e-03, + 1.1139e-03, -5.8899e-03, -1.0933e-02, 2.9011e-03, -6.8843e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 260.58, cls_loss 0.0425 cls_loss_mapping 0.0693 cls_loss_causal 0.9267 re_mapping 0.0306 re_causal 0.0850 /// teacc 98.37 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0468, -0.0093, 0.0696, ..., 0.0147, -0.0498, -0.0391], + [-0.0361, -0.0466, -0.0004, ..., -0.0444, -0.0272, 0.0057], + [ 0.0595, -0.0010, -0.0364, ..., -0.0387, -0.0573, -0.0835], + ..., + [-0.0117, 0.0024, -0.0498, ..., 0.0076, 0.0191, 0.0185], + [-0.0464, -0.0507, -0.0124, ..., -0.0213, 0.0070, -0.0129], + [-0.0054, -0.0225, -0.0054, ..., 0.0421, 0.0210, -0.0461]], + device='cuda:0'), grad: tensor([[ 1.8132e-04, 8.6874e-06, -1.4043e-04, ..., -1.3340e-04, + 1.1462e-04, 3.3307e-04], + [ 2.3556e-03, 1.5102e-05, 2.6703e-03, ..., 2.7955e-05, + 2.2447e-04, 5.5313e-03], + [ 6.7663e-04, 1.0818e-04, 5.2977e-04, ..., 1.0782e-04, + 1.4677e-03, 1.0271e-03], + ..., + [-6.6090e-04, 1.8001e-05, 2.7850e-05, ..., -5.7846e-05, + -1.7281e-03, -1.4454e-05], + [ 1.0405e-03, 2.3440e-05, 1.0023e-03, ..., 7.9155e-05, + 3.7372e-05, 2.2335e-03], + [ 1.5259e-04, 4.8488e-05, 9.5308e-05, ..., 1.6725e-04, + 4.5037e-04, 4.7117e-05]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0088, 0.0195, 0.0131, -0.0025, 0.0162, -0.0071, -0.0177, -0.0133, + 0.0262, -0.0075], device='cuda:0'), grad: tensor([ 0.0005, 0.0099, 0.0044, 0.0003, -0.0012, 0.0012, -0.0174, -0.0025, + 0.0040, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 275.63, cls_loss 0.0409 cls_loss_mapping 0.0722 cls_loss_causal 0.9112 re_mapping 0.0299 re_causal 0.0832 /// teacc 98.40 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0474, -0.0092, 0.0705, ..., 0.0148, -0.0507, -0.0400], + [-0.0360, -0.0468, -0.0008, ..., -0.0456, -0.0283, 0.0043], + [ 0.0601, -0.0015, -0.0374, ..., -0.0394, -0.0580, -0.0864], + ..., + [-0.0120, 0.0022, -0.0505, ..., 0.0075, 0.0196, 0.0171], + [-0.0471, -0.0508, -0.0122, ..., -0.0226, 0.0067, -0.0150], + [-0.0057, -0.0230, -0.0051, ..., 0.0425, 0.0210, -0.0478]], + device='cuda:0'), grad: tensor([[ 1.0639e-04, 9.6485e-06, 1.2255e-04, ..., 8.1539e-04, + 8.5878e-04, 8.5652e-05], + [ 2.6345e-04, 2.2128e-05, 1.2118e-04, ..., 2.3615e-04, + 2.5225e-04, 1.2040e-05], + [-8.9264e-04, 1.1361e-04, 2.4986e-04, ..., 5.3167e-05, + 3.1161e-04, 4.8466e-06], + ..., + [ 2.2197e-04, 5.8264e-05, 4.6879e-05, ..., -2.3975e-03, + -2.0580e-03, -1.8919e-04], + [ 1.9109e-04, 5.5462e-05, -2.0885e-03, ..., -3.9268e-04, + -1.2131e-03, 6.1095e-05], + [ 6.4790e-05, 8.7142e-05, 2.5940e-04, ..., 1.2980e-03, + 1.0681e-03, 1.0526e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0087, 0.0195, 0.0127, -0.0021, 0.0162, -0.0071, -0.0178, -0.0131, + 0.0259, -0.0074], device='cuda:0'), grad: tensor([ 1.9703e-03, 7.6008e-04, 2.6393e-04, -6.9320e-05, 1.6761e-04, + 2.0561e-03, 2.0754e-04, -4.3411e-03, -4.0703e-03, 3.0537e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 254.21, cls_loss 0.0326 cls_loss_mapping 0.0553 cls_loss_causal 0.9035 re_mapping 0.0290 re_causal 0.0834 /// teacc 98.35 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0481, -0.0092, 0.0715, ..., 0.0155, -0.0514, -0.0405], + [-0.0368, -0.0471, -0.0015, ..., -0.0460, -0.0291, 0.0032], + [ 0.0612, -0.0023, -0.0380, ..., -0.0399, -0.0588, -0.0876], + ..., + [-0.0123, 0.0028, -0.0511, ..., 0.0073, 0.0202, 0.0177], + [-0.0479, -0.0511, -0.0119, ..., -0.0236, 0.0067, -0.0153], + [-0.0063, -0.0237, -0.0058, ..., 0.0428, 0.0211, -0.0485]], + device='cuda:0'), grad: tensor([[ 7.7128e-05, 1.0349e-05, -1.1765e-02, ..., 1.4019e-04, + -3.6201e-03, -6.1073e-03], + [-4.2458e-03, -2.5606e-04, -1.5097e-03, ..., -1.7703e-04, + 4.5609e-04, -5.0735e-04], + [ 2.4261e-03, 7.5817e-05, 3.3188e-04, ..., -1.1215e-03, + -6.1035e-04, 4.6206e-04], + ..., + [ 4.4084e-04, 2.8789e-05, 4.9448e-04, ..., 2.1057e-03, + 1.8988e-03, 5.1618e-05], + [ 6.9332e-04, 1.3161e-04, 5.9557e-04, ..., 1.1768e-03, + 1.4591e-03, 1.8322e-04], + [ 4.9323e-05, 1.9774e-05, 2.2423e-04, ..., -3.1757e-03, + -3.5381e-03, 5.5224e-05]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0087, 0.0191, 0.0128, -0.0024, 0.0164, -0.0067, -0.0176, -0.0128, + 0.0259, -0.0076], device='cuda:0'), grad: tensor([-0.0099, -0.0070, -0.0102, 0.0018, 0.0011, 0.0104, 0.0029, 0.0066, + 0.0073, -0.0029], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 268.36, cls_loss 0.0291 cls_loss_mapping 0.0543 cls_loss_causal 0.8886 re_mapping 0.0271 re_causal 0.0799 /// teacc 98.44 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0486, -0.0093, 0.0723, ..., 0.0157, -0.0521, -0.0409], + [-0.0372, -0.0474, -0.0015, ..., -0.0466, -0.0297, 0.0031], + [ 0.0622, -0.0027, -0.0391, ..., -0.0402, -0.0594, -0.0889], + ..., + [-0.0126, 0.0028, -0.0518, ..., 0.0067, 0.0204, 0.0175], + [-0.0482, -0.0513, -0.0119, ..., -0.0243, 0.0067, -0.0154], + [-0.0067, -0.0239, -0.0060, ..., 0.0434, 0.0214, -0.0489]], + device='cuda:0'), grad: tensor([[ 2.2888e-05, 4.6402e-05, -1.9193e-04, ..., -2.4110e-05, + 1.3828e-04, 1.7300e-05], + [ 1.3840e-04, 2.2447e-04, 1.8045e-05, ..., 1.4758e-04, + 3.5596e-04, 1.2904e-05], + [-9.3102e-05, 2.3949e-04, 6.0618e-05, ..., 2.5344e-04, + 4.5633e-04, 1.8805e-05], + ..., + [ 8.1122e-05, -8.7440e-05, 2.4021e-05, ..., 8.2169e-03, + 6.8359e-03, 2.4512e-06], + [-4.9353e-05, 4.9019e-04, 6.0461e-06, ..., 4.5371e-04, + 1.0157e-03, -3.0905e-05], + [ 7.7188e-05, -5.5647e-04, 4.3422e-05, ..., -9.1248e-03, + -8.7433e-03, 7.5251e-06]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0088, 0.0189, 0.0127, -0.0024, 0.0162, -0.0068, -0.0175, -0.0128, + 0.0262, -0.0076], device='cuda:0'), grad: tensor([ 1.9145e-04, 7.4339e-04, 9.8705e-04, -2.6512e-04, 8.2672e-05, + -6.5088e-04, 6.3181e-04, 4.5624e-03, 2.0599e-03, -8.3389e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 251.02, cls_loss 0.0297 cls_loss_mapping 0.0518 cls_loss_causal 0.8685 re_mapping 0.0260 re_causal 0.0770 /// teacc 98.44 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0491, -0.0094, 0.0729, ..., 0.0157, -0.0530, -0.0420], + [-0.0379, -0.0475, -0.0016, ..., -0.0465, -0.0302, 0.0020], + [ 0.0634, -0.0033, -0.0401, ..., -0.0409, -0.0598, -0.0904], + ..., + [-0.0129, 0.0027, -0.0527, ..., 0.0059, 0.0206, 0.0177], + [-0.0487, -0.0518, -0.0120, ..., -0.0246, 0.0069, -0.0158], + [-0.0074, -0.0241, -0.0052, ..., 0.0442, 0.0213, -0.0496]], + device='cuda:0'), grad: tensor([[-1.9699e-05, -1.9610e-04, -6.4611e-04, ..., 6.0380e-05, + 2.1553e-04, 1.2130e-04], + [ 3.4720e-05, 2.9191e-05, -1.0081e-05, ..., 8.7798e-05, + -1.0364e-05, 2.0400e-05], + [ 8.8573e-05, 1.3876e-04, 3.8934e-04, ..., 4.6039e-04, + 4.7326e-04, 2.1651e-05], + ..., + [-2.2840e-04, -1.0371e-04, 1.3542e-04, ..., 2.1243e-04, + -3.4118e-04, 1.1936e-05], + [-7.8619e-05, -4.9114e-04, 5.8270e-04, ..., 8.7500e-04, + 3.6979e-04, 1.5926e-04], + [ 2.1219e-05, 6.4790e-05, 1.1349e-03, ..., 1.6603e-03, + 7.7486e-04, 6.1095e-05]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0087, 0.0191, 0.0127, -0.0027, 0.0162, -0.0063, -0.0176, -0.0129, + 0.0261, -0.0078], device='cuda:0'), grad: tensor([-1.7479e-05, -2.0962e-03, 1.3027e-03, 2.0733e-03, -4.7531e-03, + -1.2169e-03, 1.1215e-03, -3.4022e-04, 1.2465e-03, 2.6817e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 267.63, cls_loss 0.0279 cls_loss_mapping 0.0494 cls_loss_causal 0.8384 re_mapping 0.0250 re_causal 0.0735 /// teacc 98.66 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0504, -0.0094, 0.0737, ..., 0.0162, -0.0538, -0.0427], + [-0.0383, -0.0478, -0.0019, ..., -0.0470, -0.0308, 0.0013], + [ 0.0645, -0.0038, -0.0406, ..., -0.0410, -0.0601, -0.0925], + ..., + [-0.0134, 0.0030, -0.0529, ..., 0.0054, 0.0207, 0.0179], + [-0.0490, -0.0520, -0.0122, ..., -0.0252, 0.0068, -0.0158], + [-0.0076, -0.0244, -0.0054, ..., 0.0447, 0.0214, -0.0501]], + device='cuda:0'), grad: tensor([[ 6.3181e-05, -1.5765e-05, -1.1873e-03, ..., -4.1080e-04, + 1.2338e-04, 4.6343e-05], + [ 1.5104e-04, 3.0071e-05, 3.5316e-05, ..., 1.1951e-04, + 1.2791e-04, 1.5885e-05], + [-9.6464e-04, 3.9697e-05, 1.6725e-04, ..., 1.5330e-04, + 1.0496e-04, 1.5020e-05], + ..., + [ 2.4056e-04, 2.8715e-05, 4.4852e-05, ..., 7.6437e-04, + 5.5981e-04, 6.4187e-06], + [ 3.4261e-04, 1.8132e-04, 9.6321e-05, ..., 4.8828e-04, + 4.9973e-04, 8.8513e-05], + [ 1.9491e-05, 2.8789e-05, 1.2362e-04, ..., -4.9438e-03, + -3.5496e-03, 1.4216e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0085, 0.0192, 0.0131, -0.0025, 0.0162, -0.0067, -0.0174, -0.0128, + 0.0260, -0.0078], device='cuda:0'), grad: tensor([-5.8794e-04, 4.0531e-04, -5.6219e-04, 5.6177e-05, 3.8719e-03, + 1.3483e-04, 4.5609e-04, 1.3571e-03, 1.2903e-03, -6.4201e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 251.05, cls_loss 0.0314 cls_loss_mapping 0.0568 cls_loss_causal 0.8744 re_mapping 0.0236 re_causal 0.0714 /// teacc 98.64 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0510, -0.0091, 0.0747, ..., 0.0167, -0.0544, -0.0440], + [-0.0389, -0.0482, -0.0022, ..., -0.0469, -0.0312, -0.0017], + [ 0.0659, -0.0042, -0.0417, ..., -0.0413, -0.0607, -0.0938], + ..., + [-0.0137, 0.0025, -0.0533, ..., 0.0048, 0.0209, 0.0180], + [-0.0500, -0.0523, -0.0122, ..., -0.0261, 0.0070, -0.0158], + [-0.0081, -0.0250, -0.0060, ..., 0.0455, 0.0216, -0.0511]], + device='cuda:0'), grad: tensor([[ 6.2644e-05, 7.7343e-04, 1.2732e-03, ..., 7.0858e-04, + 2.8038e-04, 7.4983e-05], + [ 1.8522e-05, 5.2750e-05, 7.8201e-05, ..., 8.8334e-05, + 9.2328e-05, 2.1830e-05], + [-1.0071e-03, -1.3009e-05, 2.1362e-04, ..., -2.1005e-04, + 1.4436e-04, 2.3440e-05], + ..., + [ 8.6010e-05, 2.8896e-04, 4.7016e-04, ..., 3.7456e-04, + 1.2696e-04, 3.7085e-06], + [ 1.1486e-04, 1.4954e-03, 2.5177e-03, ..., 1.7939e-03, + 9.2316e-04, 1.5581e-04], + [ 1.0395e-04, 2.4092e-04, 3.8290e-04, ..., 8.0204e-04, + 5.6839e-04, 1.8895e-05]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0085, 0.0187, 0.0135, -0.0027, 0.0161, -0.0068, -0.0172, -0.0127, + 0.0261, -0.0079], device='cuda:0'), grad: tensor([ 0.0031, 0.0002, -0.0002, 0.0036, -0.0015, -0.0172, 0.0026, 0.0013, + 0.0065, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 251.55, cls_loss 0.0260 cls_loss_mapping 0.0504 cls_loss_causal 0.8111 re_mapping 0.0238 re_causal 0.0691 /// teacc 98.43 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0519, -0.0094, 0.0751, ..., 0.0169, -0.0550, -0.0447], + [-0.0393, -0.0486, -0.0016, ..., -0.0475, -0.0319, -0.0009], + [ 0.0672, -0.0044, -0.0426, ..., -0.0413, -0.0610, -0.0971], + ..., + [-0.0142, 0.0023, -0.0539, ..., 0.0045, 0.0212, 0.0175], + [-0.0503, -0.0523, -0.0125, ..., -0.0269, 0.0072, -0.0162], + [-0.0084, -0.0254, -0.0062, ..., 0.0461, 0.0216, -0.0516]], + device='cuda:0'), grad: tensor([[-2.4602e-05, -3.3528e-05, -1.5247e-04, ..., 3.1143e-05, + 1.4913e-04, 6.7532e-05], + [-4.7088e-06, 2.8491e-05, -7.4565e-05, ..., 1.5393e-05, + 1.0884e-04, -7.0274e-05], + [ 5.8746e-04, 3.4833e-04, 7.4327e-05, ..., 3.6216e-04, + 4.7421e-04, 4.6700e-05], + ..., + [ 5.9932e-05, 1.6367e-04, 1.5631e-05, ..., 5.4312e-04, + 4.8304e-04, -3.7819e-05], + [ 2.3949e-04, 4.4608e-04, 1.9681e-04, ..., 3.6788e-04, + 1.0042e-03, 3.8600e-04], + [ 7.9095e-05, 1.5068e-04, 7.0989e-05, ..., 3.8218e-04, + 4.6206e-04, 6.2823e-05]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0082, 0.0186, 0.0137, -0.0029, 0.0161, -0.0065, -0.0172, -0.0125, + 0.0265, -0.0082], device='cuda:0'), grad: tensor([ 0.0004, -0.0003, 0.0028, 0.0023, 0.0020, -0.0147, -0.0008, 0.0028, + 0.0037, 0.0018], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 251.25, cls_loss 0.0266 cls_loss_mapping 0.0545 cls_loss_causal 0.8280 re_mapping 0.0223 re_causal 0.0669 /// teacc 98.63 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0528, -0.0094, 0.0760, ..., 0.0177, -0.0555, -0.0452], + [-0.0395, -0.0487, -0.0014, ..., -0.0479, -0.0324, -0.0013], + [ 0.0678, -0.0054, -0.0434, ..., -0.0419, -0.0620, -0.0968], + ..., + [-0.0139, 0.0028, -0.0546, ..., 0.0041, 0.0217, 0.0179], + [-0.0511, -0.0527, -0.0122, ..., -0.0279, 0.0072, -0.0174], + [-0.0090, -0.0259, -0.0070, ..., 0.0465, 0.0214, -0.0537]], + device='cuda:0'), grad: tensor([[ 2.0102e-05, 3.2485e-05, 1.1069e-04, ..., 1.1677e-04, + 1.0437e-04, 1.2708e-04], + [-9.0837e-05, 9.6038e-06, 2.5317e-05, ..., -1.1273e-05, + 1.9562e-04, -7.8857e-05], + [-6.2609e-04, 1.8954e-05, 3.3945e-05, ..., 7.6890e-05, + 1.3208e-04, 2.5377e-05], + ..., + [ 1.7130e-04, -6.7711e-05, 1.5467e-05, ..., 1.0324e-04, + -7.7915e-04, 6.0976e-05], + [ 1.4913e-04, 1.2711e-05, 4.5478e-05, ..., 1.0383e-04, + 1.9586e-04, 9.5785e-05], + [ 1.7717e-05, 1.6034e-05, 2.0996e-05, ..., 3.5793e-05, + 1.3840e-04, 2.3097e-05]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0085, 0.0190, 0.0132, -0.0027, 0.0162, -0.0066, -0.0174, -0.0122, + 0.0264, -0.0087], device='cuda:0'), grad: tensor([ 0.0003, -0.0019, 0.0005, 0.0010, -0.0006, 0.0004, -0.0004, -0.0010, + 0.0011, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 268.03, cls_loss 0.0216 cls_loss_mapping 0.0408 cls_loss_causal 0.7950 re_mapping 0.0212 re_causal 0.0650 /// teacc 98.69 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0534, -0.0094, 0.0763, ..., 0.0180, -0.0562, -0.0464], + [-0.0402, -0.0492, -0.0014, ..., -0.0482, -0.0329, -0.0020], + [ 0.0689, -0.0055, -0.0442, ..., -0.0423, -0.0627, -0.0964], + ..., + [-0.0140, 0.0031, -0.0551, ..., 0.0034, 0.0219, 0.0180], + [-0.0516, -0.0530, -0.0118, ..., -0.0288, 0.0071, -0.0173], + [-0.0093, -0.0262, -0.0074, ..., 0.0472, 0.0217, -0.0542]], + device='cuda:0'), grad: tensor([[ 2.8953e-05, 3.8892e-06, -8.5890e-05, ..., 4.0196e-06, + 1.7321e-04, 6.9261e-05], + [ 4.8846e-05, 3.4153e-05, 1.5001e-03, ..., -3.8910e-03, + -5.1618e-05, 3.2663e-05], + [-6.4802e-04, 3.1888e-05, 1.3626e-04, ..., 1.9455e-04, + 2.6035e-04, 7.0632e-05], + ..., + [ 1.7536e-04, -3.0017e-04, 6.1333e-05, ..., 4.2992e-03, + 2.3901e-04, 1.0863e-05], + [ 1.9443e-04, 7.4096e-06, -2.1324e-03, ..., -1.6212e-03, + -6.7663e-04, 4.9257e-04], + [ 1.4320e-05, 5.2214e-05, 5.0592e-04, ..., 2.9802e-05, + 8.5652e-05, 1.2034e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0081, 0.0190, 0.0134, -0.0030, 0.0166, -0.0066, -0.0176, -0.0121, + 0.0262, -0.0085], device='cuda:0'), grad: tensor([ 0.0003, -0.0062, 0.0001, 0.0013, 0.0012, -0.0144, 0.0136, 0.0070, + -0.0039, 0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 251.29, cls_loss 0.0267 cls_loss_mapping 0.0460 cls_loss_causal 0.7893 re_mapping 0.0205 re_causal 0.0623 /// teacc 98.51 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0539, -0.0093, 0.0766, ..., 0.0181, -0.0565, -0.0479], + [-0.0409, -0.0495, -0.0017, ..., -0.0483, -0.0335, -0.0027], + [ 0.0699, -0.0058, -0.0450, ..., -0.0429, -0.0633, -0.0978], + ..., + [-0.0143, 0.0032, -0.0557, ..., 0.0028, 0.0219, 0.0170], + [-0.0522, -0.0532, -0.0115, ..., -0.0295, 0.0072, -0.0181], + [-0.0094, -0.0265, -0.0071, ..., 0.0478, 0.0218, -0.0561]], + device='cuda:0'), grad: tensor([[-1.1368e-02, 6.5231e-04, -1.2047e-02, ..., -2.1103e-02, + 4.6164e-05, 4.8399e-05], + [-4.3392e-05, 1.0587e-05, -1.5676e-04, ..., 4.5419e-05, + 2.7704e-04, -4.2170e-05], + [ 8.1482e-03, -2.9579e-05, 1.0208e-02, ..., 1.6815e-02, + 1.1849e-04, 2.0885e-04], + ..., + [ 3.8600e-04, 2.8849e-04, 5.5647e-04, ..., 5.4407e-04, + -1.3704e-03, -4.7851e-04], + [ 2.3317e-04, 4.2558e-05, 3.2902e-04, ..., 3.3951e-04, + 6.0844e-04, 3.0828e-04], + [ 1.7226e-04, 1.0169e-04, 4.2987e-04, ..., 4.4417e-04, + 3.6407e-04, 1.1063e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0081, 0.0187, 0.0136, -0.0028, 0.0167, -0.0066, -0.0177, -0.0124, + 0.0264, -0.0084], device='cuda:0'), grad: tensor([-0.0235, -0.0005, 0.0220, 0.0033, 0.0055, -0.0108, 0.0005, -0.0002, + 0.0020, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 251.13, cls_loss 0.0219 cls_loss_mapping 0.0425 cls_loss_causal 0.8226 re_mapping 0.0194 re_causal 0.0621 /// teacc 98.67 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0537, -0.0093, 0.0773, ..., 0.0189, -0.0571, -0.0485], + [-0.0413, -0.0497, -0.0020, ..., -0.0482, -0.0342, -0.0040], + [ 0.0706, -0.0060, -0.0463, ..., -0.0440, -0.0638, -0.0995], + ..., + [-0.0146, 0.0035, -0.0562, ..., 0.0026, 0.0225, 0.0164], + [-0.0525, -0.0535, -0.0111, ..., -0.0300, 0.0071, -0.0187], + [-0.0099, -0.0267, -0.0076, ..., 0.0485, 0.0220, -0.0578]], + device='cuda:0'), grad: tensor([[ 1.2153e-04, -3.0935e-05, 1.9730e-02, ..., 1.7975e-02, + 6.3777e-05, 5.5194e-05], + [ 1.9765e-04, 2.2724e-06, 4.3929e-05, ..., 6.6340e-05, + 2.6822e-04, 1.2982e-04], + [-2.9430e-03, 5.6960e-06, 1.9455e-04, ..., 1.2219e-04, + -1.0425e-04, -3.4571e-05], + ..., + [ 1.0328e-03, 5.8226e-06, 9.4891e-05, ..., 1.0526e-04, + 6.2466e-05, 7.5638e-05], + [ 8.3780e-04, 7.6517e-06, 2.1434e-04, ..., 9.1910e-05, + -2.0266e-04, -1.3351e-04], + [ 1.3185e-04, 6.3665e-06, -2.0584e-02, ..., -1.7471e-02, + 5.1641e-04, 4.1097e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0082, 0.0185, 0.0132, -0.0029, 0.0163, -0.0068, -0.0170, -0.0120, + 0.0263, -0.0082], device='cuda:0'), grad: tensor([ 0.0303, 0.0009, -0.0041, 0.0003, -0.0009, 0.0010, -0.0001, 0.0012, + 0.0009, -0.0296], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 251.08, cls_loss 0.0219 cls_loss_mapping 0.0464 cls_loss_causal 0.8260 re_mapping 0.0200 re_causal 0.0615 /// teacc 98.58 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0543, -0.0091, 0.0786, ..., 0.0197, -0.0579, -0.0493], + [-0.0418, -0.0501, -0.0022, ..., -0.0484, -0.0348, -0.0046], + [ 0.0718, -0.0063, -0.0474, ..., -0.0446, -0.0643, -0.1003], + ..., + [-0.0151, 0.0037, -0.0570, ..., 0.0022, 0.0227, 0.0158], + [-0.0530, -0.0539, -0.0111, ..., -0.0308, 0.0071, -0.0175], + [-0.0105, -0.0271, -0.0078, ..., 0.0490, 0.0220, -0.0585]], + device='cuda:0'), grad: tensor([[ 3.7611e-05, 3.6508e-06, -6.2037e-04, ..., -2.4700e-04, + 4.3392e-05, 1.3173e-05], + [ 1.6525e-05, 2.7306e-06, 2.0421e-04, ..., 1.2898e-04, + 3.3349e-05, -2.3544e-05], + [-8.1182e-05, 3.3695e-06, 1.2147e-04, ..., 7.2122e-05, + 7.7367e-05, 2.2724e-05], + ..., + [ 3.9190e-05, 1.3208e-04, 2.9042e-05, ..., 4.8876e-04, + 4.5729e-04, 4.0948e-05], + [ 1.3877e-06, 1.2748e-05, 9.0659e-05, ..., 3.6240e-04, + 2.7585e-04, 3.3200e-05], + [-3.1114e-05, -1.9205e-04, 4.1157e-05, ..., -1.2398e-03, + -1.2732e-03, -1.0669e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0086, 0.0185, 0.0134, -0.0026, 0.0163, -0.0065, -0.0175, -0.0121, + 0.0263, -0.0086], device='cuda:0'), grad: tensor([-2.9230e-04, -8.6948e-06, 2.2686e-04, 1.3804e-04, 5.0306e-04, + 6.7174e-05, 6.2108e-05, 8.0919e-04, 3.5644e-04, -1.8616e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 268.25, cls_loss 0.0222 cls_loss_mapping 0.0426 cls_loss_causal 0.8053 re_mapping 0.0194 re_causal 0.0594 /// teacc 98.82 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0549, -0.0093, 0.0790, ..., 0.0201, -0.0588, -0.0506], + [-0.0422, -0.0505, -0.0021, ..., -0.0491, -0.0358, -0.0052], + [ 0.0732, -0.0075, -0.0484, ..., -0.0451, -0.0649, -0.1013], + ..., + [-0.0154, 0.0046, -0.0578, ..., 0.0015, 0.0229, 0.0158], + [-0.0536, -0.0540, -0.0111, ..., -0.0316, 0.0070, -0.0172], + [-0.0112, -0.0275, -0.0082, ..., 0.0493, 0.0221, -0.0586]], + device='cuda:0'), grad: tensor([[ 7.0810e-05, 6.9886e-06, -2.8372e-04, ..., -2.1446e-04, + 2.3976e-05, 5.8413e-05], + [ 2.0969e-04, 2.4676e-05, 3.2097e-05, ..., 8.4400e-05, + 5.9515e-05, 2.6441e-04], + [-1.4138e-04, 2.7552e-05, 1.1426e-04, ..., 1.3733e-04, + 8.0526e-05, 2.0707e-04], + ..., + [ 1.3304e-04, 2.5295e-06, 2.1338e-05, ..., 5.8800e-05, + -4.6670e-05, 2.0042e-05], + [ 2.3437e-04, 3.5197e-05, 2.9802e-05, ..., 1.5986e-04, + 1.8156e-04, 8.4519e-05], + [ 2.2650e-04, 2.7850e-05, 5.0277e-05, ..., 5.6744e-04, + 3.5381e-04, 1.4514e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0082, 0.0183, 0.0135, -0.0027, 0.0166, -0.0063, -0.0176, -0.0120, + 0.0263, -0.0087], device='cuda:0'), grad: tensor([-0.0002, 0.0005, 0.0007, -0.0016, -0.0004, 0.0002, -0.0018, 0.0003, + 0.0009, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 251.21, cls_loss 0.0144 cls_loss_mapping 0.0322 cls_loss_causal 0.8132 re_mapping 0.0186 re_causal 0.0598 /// teacc 98.71 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0557, -0.0094, 0.0798, ..., 0.0206, -0.0594, -0.0513], + [-0.0425, -0.0509, -0.0022, ..., -0.0498, -0.0364, -0.0057], + [ 0.0740, -0.0077, -0.0492, ..., -0.0450, -0.0655, -0.1026], + ..., + [-0.0158, 0.0046, -0.0583, ..., 0.0012, 0.0234, 0.0157], + [-0.0542, -0.0542, -0.0114, ..., -0.0323, 0.0069, -0.0170], + [-0.0117, -0.0277, -0.0086, ..., 0.0497, 0.0222, -0.0589]], + device='cuda:0'), grad: tensor([[-3.3712e-04, -1.1005e-05, -1.3599e-03, ..., -3.1638e-04, + -2.7627e-05, 1.4029e-05], + [ 5.0247e-05, 1.5363e-05, 1.2517e-05, ..., 5.9783e-05, + 5.9694e-05, 1.5218e-06], + [ 1.0329e-04, 4.2528e-05, 3.8147e-04, ..., 1.7357e-04, + 9.8050e-05, 1.2539e-05], + ..., + [ 1.4508e-04, 4.2655e-06, 2.5749e-05, ..., 1.0192e-04, + 1.6659e-05, 1.1206e-05], + [ 2.2209e-04, 2.9385e-05, 4.2987e-04, ..., 2.3830e-04, + 1.1617e-04, 1.0896e-07], + [ 8.3983e-05, 3.5167e-05, 1.7107e-04, ..., -5.7173e-04, + -3.8624e-04, -7.3493e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0083, 0.0180, 0.0136, -0.0026, 0.0165, -0.0061, -0.0176, -0.0119, + 0.0261, -0.0088], device='cuda:0'), grad: tensor([-1.7090e-03, 2.2396e-05, 8.2254e-04, -5.3835e-04, -7.7724e-05, + 5.2118e-04, 2.4939e-04, 3.6216e-04, 8.8215e-04, -5.3501e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 251.20, cls_loss 0.0200 cls_loss_mapping 0.0382 cls_loss_causal 0.7281 re_mapping 0.0190 re_causal 0.0545 /// teacc 98.61 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0562, -0.0094, 0.0800, ..., 0.0206, -0.0602, -0.0524], + [-0.0432, -0.0512, -0.0021, ..., -0.0502, -0.0370, -0.0057], + [ 0.0748, -0.0080, -0.0500, ..., -0.0454, -0.0659, -0.1030], + ..., + [-0.0160, 0.0047, -0.0588, ..., 0.0008, 0.0236, 0.0156], + [-0.0546, -0.0544, -0.0115, ..., -0.0332, 0.0070, -0.0180], + [-0.0121, -0.0279, -0.0079, ..., 0.0502, 0.0222, -0.0599]], + device='cuda:0'), grad: tensor([[ 1.7738e-04, -1.5557e-05, -3.2139e-04, ..., -2.6369e-04, + 3.2043e-04, 1.1295e-04], + [ 2.2426e-05, 7.8455e-06, -3.3323e-06, ..., -1.1899e-05, + 2.1055e-05, 8.4750e-07], + [ 6.1810e-05, 4.0293e-05, 3.6430e-04, ..., 2.1231e-04, + 2.6870e-04, 1.2267e-04], + ..., + [ 5.9277e-05, 3.2336e-05, 7.6890e-05, ..., 1.5700e-04, + 5.0552e-06, -1.8477e-05], + [ 1.0014e-04, 2.0042e-05, 7.1883e-05, ..., 1.5473e-04, + 2.5725e-04, 4.2796e-05], + [ 2.4259e-05, -1.5441e-06, 1.0061e-04, ..., -5.0455e-05, + -7.3016e-05, 3.3528e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0077, 0.0176, 0.0139, -0.0022, 0.0166, -0.0062, -0.0177, -0.0118, + 0.0262, -0.0086], device='cuda:0'), grad: tensor([ 3.5524e-04, -7.1335e-04, 8.7452e-04, -9.2745e-04, -1.0723e-04, + -5.0038e-05, -4.5037e-04, 1.8454e-04, 7.1859e-04, 1.1450e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 251.53, cls_loss 0.0179 cls_loss_mapping 0.0347 cls_loss_causal 0.7735 re_mapping 0.0183 re_causal 0.0572 /// teacc 98.78 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0572, -0.0093, 0.0805, ..., 0.0210, -0.0609, -0.0534], + [-0.0435, -0.0514, -0.0019, ..., -0.0505, -0.0375, -0.0065], + [ 0.0754, -0.0092, -0.0508, ..., -0.0462, -0.0667, -0.1032], + ..., + [-0.0158, 0.0049, -0.0596, ..., 0.0002, 0.0238, 0.0158], + [-0.0556, -0.0544, -0.0116, ..., -0.0336, 0.0069, -0.0183], + [-0.0120, -0.0283, -0.0077, ..., 0.0507, 0.0225, -0.0591]], + device='cuda:0'), grad: tensor([[-4.7773e-05, 7.9870e-06, 2.0294e-03, ..., 8.3971e-04, + 1.3399e-03, 2.7046e-03], + [ 8.0690e-06, 1.9088e-05, 1.3888e-04, ..., 7.0572e-05, + 1.1772e-04, 1.3268e-04], + [ 3.5107e-05, 2.3022e-05, 1.5795e-04, ..., 1.6403e-04, + 7.8261e-05, 5.2005e-05], + ..., + [ 1.7017e-05, -8.7082e-05, 6.2823e-05, ..., 1.6582e-04, + -3.0088e-04, -2.5809e-05], + [ 9.2089e-06, 2.0146e-05, 2.1589e-04, ..., 1.1414e-04, + 2.3150e-04, 1.6320e-04], + [ 4.6849e-05, 7.0632e-05, 2.6512e-04, ..., 1.7548e-04, + 1.9252e-04, 1.0437e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0077, 0.0178, 0.0133, -0.0025, 0.0168, -0.0065, -0.0172, -0.0115, + 0.0260, -0.0082], device='cuda:0'), grad: tensor([ 0.0028, 0.0001, 0.0003, 0.0001, 0.0005, -0.0002, -0.0048, -0.0002, + 0.0006, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 267.79, cls_loss 0.0178 cls_loss_mapping 0.0357 cls_loss_causal 0.7641 re_mapping 0.0177 re_causal 0.0552 /// teacc 98.85 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0579, -0.0093, 0.0811, ..., 0.0215, -0.0614, -0.0549], + [-0.0439, -0.0518, -0.0027, ..., -0.0504, -0.0375, -0.0069], + [ 0.0763, -0.0094, -0.0515, ..., -0.0465, -0.0672, -0.1044], + ..., + [-0.0166, 0.0045, -0.0603, ..., -0.0007, 0.0237, 0.0154], + [-0.0562, -0.0550, -0.0111, ..., -0.0343, 0.0066, -0.0186], + [-0.0123, -0.0281, -0.0078, ..., 0.0513, 0.0227, -0.0592]], + device='cuda:0'), grad: tensor([[-9.2760e-06, 7.7812e-07, -1.2932e-03, ..., -7.9679e-04, + 1.5065e-05, 2.6450e-05], + [ 5.1446e-06, 3.0790e-06, -4.1574e-05, ..., 1.9893e-05, + 6.4552e-05, 4.8816e-05], + [ 2.6405e-05, 1.3262e-05, 1.4973e-04, ..., 9.7156e-05, + 2.4706e-05, 1.7166e-05], + ..., + [ 1.0513e-05, -1.1362e-06, 4.6402e-05, ..., 4.2170e-05, + -1.2648e-04, -4.2021e-05], + [ 8.6129e-06, 4.8205e-06, 1.4460e-04, ..., 1.3745e-04, + 1.0538e-04, 1.3173e-04], + [ 1.3821e-05, 9.9912e-06, 5.7793e-04, ..., 2.6631e-04, + -1.3256e-04, -5.8025e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0073, 0.0177, 0.0133, -0.0024, 0.0168, -0.0058, -0.0171, -0.0120, + 0.0257, -0.0079], device='cuda:0'), grad: tensor([-1.1797e-03, -1.1629e-04, 2.6679e-04, 1.3804e-04, -5.6118e-05, + 5.6601e-04, -3.9959e-04, -6.6757e-05, 4.7708e-04, 3.7003e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 251.15, cls_loss 0.0166 cls_loss_mapping 0.0340 cls_loss_causal 0.7346 re_mapping 0.0169 re_causal 0.0537 /// teacc 98.82 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0587, -0.0094, 0.0818, ..., 0.0225, -0.0619, -0.0561], + [-0.0444, -0.0521, -0.0028, ..., -0.0509, -0.0381, -0.0072], + [ 0.0773, -0.0095, -0.0520, ..., -0.0470, -0.0676, -0.1055], + ..., + [-0.0172, 0.0047, -0.0607, ..., -0.0011, 0.0241, 0.0156], + [-0.0568, -0.0552, -0.0108, ..., -0.0346, 0.0066, -0.0190], + [-0.0121, -0.0285, -0.0088, ..., 0.0516, 0.0230, -0.0585]], + device='cuda:0'), grad: tensor([[ 4.1388e-06, 5.0887e-06, -1.5831e-04, ..., -6.0558e-05, + 2.6762e-05, 2.6390e-05], + [ 5.3942e-06, 1.3255e-05, -8.8274e-05, ..., -3.3468e-05, + 2.5347e-05, 5.6736e-06], + [-4.6194e-05, 1.1675e-05, 9.6619e-05, ..., 4.8637e-05, + 4.0501e-05, 3.1531e-05], + ..., + [ 1.3471e-05, 1.6224e-06, 1.6481e-05, ..., 3.0667e-05, + -3.0294e-05, 1.2768e-06], + [ 1.6361e-05, 1.5914e-05, 7.3195e-05, ..., 7.8261e-05, + 6.3837e-05, 3.4183e-05], + [ 2.5779e-06, 5.3525e-05, 4.1664e-05, ..., 6.6876e-05, + 4.0591e-05, 3.5852e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0075, 0.0175, 0.0136, -0.0029, 0.0170, -0.0060, -0.0171, -0.0119, + 0.0258, -0.0080], device='cuda:0'), grad: tensor([-3.1769e-05, -3.2496e-04, 2.2042e-04, -1.2743e-04, 2.2113e-04, + -2.8038e-04, -2.1398e-04, 2.7761e-05, 1.5342e-04, 3.5453e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 251.21, cls_loss 0.0146 cls_loss_mapping 0.0335 cls_loss_causal 0.7366 re_mapping 0.0162 re_causal 0.0512 /// teacc 98.58 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0592, -0.0094, 0.0828, ..., 0.0230, -0.0621, -0.0554], + [-0.0442, -0.0524, -0.0027, ..., -0.0509, -0.0381, -0.0075], + [ 0.0782, -0.0096, -0.0525, ..., -0.0475, -0.0683, -0.1065], + ..., + [-0.0176, 0.0044, -0.0611, ..., -0.0018, 0.0242, 0.0156], + [-0.0577, -0.0554, -0.0112, ..., -0.0353, 0.0066, -0.0195], + [-0.0121, -0.0285, -0.0092, ..., 0.0523, 0.0231, -0.0590]], + device='cuda:0'), grad: tensor([[ 1.9819e-05, 1.3066e-06, -1.0663e-04, ..., -1.0431e-06, + 3.9160e-05, 1.3523e-05], + [ 3.0369e-05, 3.8669e-06, 1.2210e-06, ..., 1.0729e-04, + 3.4046e-04, 6.4149e-06], + [ 1.6463e-04, 6.2063e-06, 2.6554e-05, ..., 2.1410e-04, + 1.0508e-04, 3.7774e-06], + ..., + [ 2.5466e-05, -4.8190e-05, 4.1761e-06, ..., -6.5207e-05, + -5.1022e-04, 4.8615e-07], + [ 3.4124e-05, 4.1872e-06, 2.5198e-05, ..., 2.0123e-04, + 2.2328e-04, 1.7807e-05], + [ 2.3857e-05, 1.8582e-05, 2.2367e-05, ..., 1.5745e-03, + 7.2384e-04, 4.0010e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0079, 0.0183, 0.0135, -0.0029, 0.0168, -0.0062, -0.0170, -0.0120, + 0.0253, -0.0080], device='cuda:0'), grad: tensor([ 6.9618e-05, 2.4776e-03, 5.6744e-04, 2.4261e-03, -1.9484e-03, + -6.6299e-03, 1.5855e-04, -7.1573e-04, 1.3323e-03, 2.2602e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 251.46, cls_loss 0.0120 cls_loss_mapping 0.0265 cls_loss_causal 0.7327 re_mapping 0.0158 re_causal 0.0511 /// teacc 98.77 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0598, -0.0094, 0.0834, ..., 0.0235, -0.0627, -0.0561], + [-0.0446, -0.0528, -0.0027, ..., -0.0513, -0.0389, -0.0079], + [ 0.0789, -0.0103, -0.0530, ..., -0.0476, -0.0687, -0.1070], + ..., + [-0.0180, 0.0050, -0.0615, ..., -0.0024, 0.0247, 0.0157], + [-0.0581, -0.0555, -0.0114, ..., -0.0360, 0.0065, -0.0195], + [-0.0124, -0.0289, -0.0095, ..., 0.0526, 0.0233, -0.0588]], + device='cuda:0'), grad: tensor([[ 5.6356e-05, -6.1178e-04, -6.0616e-03, ..., -3.5686e-03, + -1.4372e-03, 2.9281e-05], + [ 1.6093e-04, 9.3043e-05, 1.3620e-05, ..., 1.5602e-05, + 2.2757e-04, 9.3505e-06], + [ 1.8740e-03, 3.9268e-04, 3.3617e-04, ..., 2.0289e-04, + 1.2980e-03, 7.9215e-05], + ..., + [-2.1839e-03, -2.4643e-03, 3.7014e-05, ..., 6.8128e-05, + -5.1613e-03, 8.8364e-06], + [ 1.9140e-03, 2.1973e-03, 3.0785e-03, ..., 1.8053e-03, + 4.8637e-03, 5.8919e-05], + [ 5.6952e-05, 5.9694e-05, 4.9925e-04, ..., 4.1771e-04, + 1.4782e-04, 1.2644e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0080, 0.0179, 0.0135, -0.0027, 0.0167, -0.0064, -0.0169, -0.0113, + 0.0252, -0.0081], device='cuda:0'), grad: tensor([-7.0190e-03, 4.7088e-04, 4.1542e-03, -3.7575e-03, -7.0989e-05, + 1.7929e-03, 2.0905e-03, -1.1101e-02, 1.2581e-02, 8.3923e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 251.32, cls_loss 0.0149 cls_loss_mapping 0.0345 cls_loss_causal 0.7418 re_mapping 0.0151 re_causal 0.0499 /// teacc 98.83 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0604, -0.0095, 0.0840, ..., 0.0238, -0.0634, -0.0567], + [-0.0453, -0.0533, -0.0030, ..., -0.0517, -0.0397, -0.0092], + [ 0.0803, -0.0107, -0.0537, ..., -0.0480, -0.0690, -0.1069], + ..., + [-0.0184, 0.0051, -0.0620, ..., -0.0028, 0.0251, 0.0154], + [-0.0589, -0.0557, -0.0115, ..., -0.0363, 0.0067, -0.0188], + [-0.0130, -0.0293, -0.0101, ..., 0.0528, 0.0234, -0.0596]], + device='cuda:0'), grad: tensor([[ 9.1642e-06, -3.8892e-06, -2.4176e-04, ..., -8.9467e-05, + 4.1038e-05, 4.8041e-05], + [ 1.4342e-05, 2.0582e-06, -6.0126e-06, ..., 3.2876e-06, + 4.0740e-05, 2.9162e-05], + [-2.1183e-04, 2.7847e-06, 6.0558e-05, ..., 1.4737e-05, + 1.4469e-05, 3.6117e-06], + ..., + [ 1.5244e-05, -2.5108e-06, 3.2812e-05, ..., 1.9118e-05, + -3.4630e-05, 1.0766e-05], + [ 1.2088e-04, 1.8328e-05, 2.9898e-04, ..., 5.6416e-05, + 5.8985e-04, 7.1669e-04], + [ 1.6708e-06, 6.3367e-06, 6.1750e-05, ..., -9.3043e-05, + -8.3983e-05, 1.6183e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0078, 0.0176, 0.0140, -0.0027, 0.0167, -0.0065, -0.0166, -0.0112, + 0.0254, -0.0086], device='cuda:0'), grad: tensor([-1.1152e-04, -1.2636e-04, -1.3292e-04, 9.9754e-04, 1.8358e-04, + -2.0313e-03, -3.7813e-04, 4.3660e-05, 1.5888e-03, -3.2276e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 269.98, cls_loss 0.0137 cls_loss_mapping 0.0287 cls_loss_causal 0.7411 re_mapping 0.0152 re_causal 0.0483 /// teacc 98.87 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0610, -0.0096, 0.0846, ..., 0.0242, -0.0642, -0.0577], + [-0.0457, -0.0537, -0.0027, ..., -0.0522, -0.0405, -0.0098], + [ 0.0810, -0.0113, -0.0547, ..., -0.0485, -0.0697, -0.1078], + ..., + [-0.0186, 0.0046, -0.0628, ..., -0.0034, 0.0251, 0.0155], + [-0.0594, -0.0559, -0.0112, ..., -0.0366, 0.0067, -0.0188], + [-0.0130, -0.0295, -0.0105, ..., 0.0531, 0.0235, -0.0598]], + device='cuda:0'), grad: tensor([[ 3.2336e-05, 6.4448e-06, -2.2256e-04, ..., -8.0824e-05, + 1.9148e-05, 9.4771e-06], + [ 7.4327e-05, 8.3968e-06, 6.0834e-06, ..., 9.6783e-06, + 4.2617e-05, 1.3746e-06], + [-4.6849e-04, -3.2812e-05, 1.8165e-05, ..., 1.8835e-05, + 3.4988e-05, 7.6508e-07], + ..., + [ 1.2624e-04, 1.8150e-05, 4.9584e-06, ..., 7.9721e-06, + -1.9717e-04, 1.8356e-06], + [ 5.8800e-05, 2.2995e-04, 2.7761e-05, ..., 1.0675e-04, + 3.1877e-04, 6.9916e-05], + [ 1.8865e-05, 8.3089e-05, 1.2636e-04, ..., 1.3220e-04, + 1.4579e-04, 2.5585e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0078, 0.0173, 0.0140, -0.0027, 0.0172, -0.0068, -0.0163, -0.0112, + 0.0252, -0.0084], device='cuda:0'), grad: tensor([-8.5354e-05, 1.9205e-04, -3.5548e-04, 3.3307e-04, -3.1531e-05, + -1.1768e-03, 1.2791e-04, -4.6539e-04, 9.3126e-04, 5.3072e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 251.41, cls_loss 0.0169 cls_loss_mapping 0.0346 cls_loss_causal 0.7538 re_mapping 0.0148 re_causal 0.0478 /// teacc 98.86 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0616, -0.0097, 0.0850, ..., 0.0240, -0.0648, -0.0589], + [-0.0465, -0.0540, -0.0019, ..., -0.0525, -0.0412, -0.0104], + [ 0.0817, -0.0120, -0.0556, ..., -0.0493, -0.0702, -0.1077], + ..., + [-0.0184, 0.0046, -0.0636, ..., -0.0037, 0.0254, 0.0159], + [-0.0599, -0.0564, -0.0114, ..., -0.0372, 0.0065, -0.0188], + [-0.0132, -0.0299, -0.0105, ..., 0.0538, 0.0236, -0.0609]], + device='cuda:0'), grad: tensor([[ 8.3521e-06, 1.3098e-05, 1.0222e-04, ..., -2.4274e-05, + 5.7578e-05, 1.6320e-04], + [ 5.4277e-06, 3.9563e-06, 3.7532e-06, ..., 1.3523e-05, + 1.9014e-05, 1.1586e-05], + [-3.3832e-04, -1.0192e-04, 4.3631e-05, ..., 9.7871e-05, + -1.6376e-05, 2.3946e-05], + ..., + [ 2.6441e-04, 6.3837e-05, 6.8024e-06, ..., 1.4460e-04, + 1.0210e-04, 1.4290e-05], + [-1.1332e-05, 3.3863e-06, -2.5570e-05, ..., 1.4901e-05, + -2.5153e-05, -7.6964e-06], + [ 4.6492e-06, 1.1832e-05, 2.3320e-05, ..., 4.5747e-05, + 3.7313e-05, 2.2054e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0073, 0.0165, 0.0142, -0.0024, 0.0171, -0.0068, -0.0161, -0.0110, + 0.0249, -0.0080], device='cuda:0'), grad: tensor([ 1.8764e-04, -1.1759e-03, 4.2152e-04, 2.0576e-04, -1.7846e-04, + -8.7619e-05, -7.9453e-05, 6.1560e-04, -1.4484e-04, 2.3580e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 251.28, cls_loss 0.0109 cls_loss_mapping 0.0264 cls_loss_causal 0.6991 re_mapping 0.0145 re_causal 0.0465 /// teacc 98.81 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0620, -0.0098, 0.0857, ..., 0.0248, -0.0655, -0.0600], + [-0.0465, -0.0542, -0.0017, ..., -0.0531, -0.0415, -0.0107], + [ 0.0827, -0.0123, -0.0564, ..., -0.0500, -0.0706, -0.1085], + ..., + [-0.0189, 0.0047, -0.0640, ..., -0.0040, 0.0257, 0.0151], + [-0.0603, -0.0566, -0.0111, ..., -0.0378, 0.0067, -0.0178], + [-0.0134, -0.0302, -0.0110, ..., 0.0537, 0.0236, -0.0619]], + device='cuda:0'), grad: tensor([[ 1.9133e-05, 2.8126e-07, -5.8115e-07, ..., -2.6952e-06, + 3.5942e-05, 3.1114e-05], + [ 9.7334e-05, 4.0643e-06, -5.3465e-05, ..., 4.6529e-06, + 2.7359e-05, -6.9141e-05], + [-5.2333e-05, 5.7146e-06, 2.1413e-05, ..., 5.8562e-06, + 1.8492e-05, 1.1161e-05], + ..., + [ 8.1956e-05, 4.0047e-06, 5.3234e-06, ..., 2.2128e-05, + -1.8919e-04, 1.0826e-05], + [ 4.4137e-05, 2.4177e-06, 2.7418e-05, ..., 5.3465e-05, + 6.3598e-05, 6.3360e-05], + [ 3.9697e-05, 8.0690e-06, 1.0103e-05, ..., -2.8539e-04, + 1.3120e-05, 2.4021e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0075, 0.0172, 0.0141, -0.0025, 0.0171, -0.0069, -0.0160, -0.0116, + 0.0250, -0.0082], device='cuda:0'), grad: tensor([ 1.7655e-04, -1.9848e-04, 3.1376e-04, 2.3136e-03, 3.3569e-04, + -3.2940e-03, -4.5598e-05, -3.9363e-04, 4.9734e-04, 2.9159e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 269.33, cls_loss 0.0108 cls_loss_mapping 0.0255 cls_loss_causal 0.6745 re_mapping 0.0147 re_causal 0.0457 /// teacc 98.92 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0624, -0.0099, 0.0865, ..., 0.0251, -0.0660, -0.0602], + [-0.0466, -0.0545, -0.0023, ..., -0.0536, -0.0417, -0.0105], + [ 0.0835, -0.0132, -0.0570, ..., -0.0507, -0.0710, -0.1095], + ..., + [-0.0196, 0.0049, -0.0643, ..., -0.0043, 0.0259, 0.0148], + [-0.0607, -0.0565, -0.0113, ..., -0.0385, 0.0066, -0.0177], + [-0.0139, -0.0305, -0.0110, ..., 0.0545, 0.0238, -0.0629]], + device='cuda:0'), grad: tensor([[ 9.8348e-06, 6.1542e-06, 2.9922e-05, ..., 8.3372e-06, + 2.0742e-05, 1.6123e-05], + [ 1.5378e-04, 1.7062e-05, 1.0356e-05, ..., -2.5362e-05, + 5.0753e-05, -6.4909e-05], + [-3.1853e-04, 1.8507e-05, 1.0140e-05, ..., 1.0647e-05, + 3.4332e-05, 2.2247e-05], + ..., + [ 5.9545e-05, 3.1620e-05, 1.2713e-06, ..., 8.2374e-05, + 1.4871e-05, 2.8998e-05], + [ 6.5804e-05, 5.6416e-05, 3.2067e-05, ..., 1.2743e-04, + 2.0647e-04, 3.6508e-05], + [ 1.5058e-05, 1.2529e-04, 5.0999e-06, ..., -2.0182e-04, + -1.5736e-04, 1.8075e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0075, 0.0174, 0.0142, -0.0026, 0.0168, -0.0067, -0.0159, -0.0116, + 0.0248, -0.0081], device='cuda:0'), grad: tensor([ 1.0157e-04, -3.1185e-04, -2.6655e-04, -2.8400e-03, 5.5122e-04, + 2.2697e-03, -4.8566e-04, 2.3186e-04, 7.4100e-04, 9.6709e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 251.87, cls_loss 0.0130 cls_loss_mapping 0.0275 cls_loss_causal 0.7024 re_mapping 0.0142 re_causal 0.0460 /// teacc 98.74 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0628, -0.0098, 0.0871, ..., 0.0256, -0.0665, -0.0607], + [-0.0469, -0.0549, -0.0023, ..., -0.0553, -0.0423, -0.0110], + [ 0.0843, -0.0136, -0.0576, ..., -0.0511, -0.0716, -0.1096], + ..., + [-0.0198, 0.0049, -0.0646, ..., -0.0048, 0.0263, 0.0143], + [-0.0613, -0.0567, -0.0114, ..., -0.0391, 0.0068, -0.0170], + [-0.0145, -0.0310, -0.0117, ..., 0.0549, 0.0237, -0.0653]], + device='cuda:0'), grad: tensor([[ 8.2478e-06, 6.4597e-06, -4.0233e-05, ..., -1.0341e-05, + 3.6508e-05, 3.4600e-05], + [ 3.0249e-05, 2.0981e-05, 2.3678e-05, ..., 1.5676e-05, + 4.7863e-05, 3.3170e-05], + [ 2.3440e-05, 3.5226e-05, 4.0412e-05, ..., 2.3946e-05, + 5.2512e-05, 4.3511e-05], + ..., + [ 3.3557e-05, -8.9943e-05, 4.3213e-06, ..., -4.6939e-05, + -4.2963e-04, -3.3557e-05], + [ 4.3422e-05, 3.7253e-05, -4.6015e-05, ..., 2.5973e-05, + 8.3089e-05, 2.6524e-05], + [ 1.5870e-05, 5.6267e-05, 1.3106e-05, ..., 6.8247e-06, + 1.7118e-04, 2.6181e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0077, 0.0169, 0.0146, -0.0028, 0.0172, -0.0063, -0.0157, -0.0118, + 0.0248, -0.0084], device='cuda:0'), grad: tensor([ 1.1945e-04, 2.4533e-04, 3.1209e-04, -2.7180e-04, 8.6427e-05, + -2.5415e-04, 2.2876e-04, -1.1177e-03, 1.9133e-04, 4.6110e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 251.33, cls_loss 0.0105 cls_loss_mapping 0.0265 cls_loss_causal 0.6975 re_mapping 0.0137 re_causal 0.0430 /// teacc 98.68 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0632, -0.0099, 0.0876, ..., 0.0259, -0.0670, -0.0613], + [-0.0470, -0.0554, -0.0017, ..., -0.0554, -0.0428, -0.0114], + [ 0.0850, -0.0142, -0.0583, ..., -0.0515, -0.0721, -0.1106], + ..., + [-0.0201, 0.0053, -0.0648, ..., -0.0055, 0.0268, 0.0145], + [-0.0617, -0.0569, -0.0116, ..., -0.0399, 0.0068, -0.0173], + [-0.0146, -0.0313, -0.0120, ..., 0.0558, 0.0239, -0.0659]], + device='cuda:0'), grad: tensor([[ 3.6627e-05, 4.3400e-06, -1.3813e-05, ..., -8.1956e-06, + 1.7911e-05, 1.8194e-05], + [ 4.9543e-04, 1.1958e-05, 3.1497e-06, ..., 1.9008e-06, + 2.1443e-05, 2.6420e-05], + [-1.8253e-03, -1.1194e-04, 7.9796e-06, ..., 6.7018e-06, + 1.9237e-05, -1.3113e-04], + ..., + [ 5.5265e-04, 9.5606e-05, 4.0121e-06, ..., 2.5883e-05, + 1.2450e-05, 1.1459e-05], + [ 4.3488e-04, 3.1888e-05, -5.5265e-04, ..., 5.9828e-06, + -8.3685e-04, -8.5449e-04], + [ 1.8492e-05, 8.5160e-06, 1.5169e-05, ..., -2.0325e-05, + 6.8128e-05, 4.6074e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0076, 0.0172, 0.0144, -0.0024, 0.0169, -0.0069, -0.0156, -0.0112, + 0.0245, -0.0085], device='cuda:0'), grad: tensor([ 7.5281e-05, 5.9795e-04, -2.7485e-03, -1.3769e-04, 8.1348e-04, + 1.8239e-04, 1.4658e-03, 7.5245e-04, -1.2274e-03, 2.2650e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 251.54, cls_loss 0.0091 cls_loss_mapping 0.0234 cls_loss_causal 0.6902 re_mapping 0.0132 re_causal 0.0418 /// teacc 98.83 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0636, -0.0099, 0.0883, ..., 0.0266, -0.0675, -0.0619], + [-0.0475, -0.0560, -0.0017, ..., -0.0553, -0.0434, -0.0114], + [ 0.0856, -0.0152, -0.0590, ..., -0.0521, -0.0728, -0.1117], + ..., + [-0.0202, 0.0058, -0.0653, ..., -0.0061, 0.0273, 0.0143], + [-0.0621, -0.0567, -0.0114, ..., -0.0403, 0.0066, -0.0175], + [-0.0148, -0.0317, -0.0123, ..., 0.0558, 0.0238, -0.0666]], + device='cuda:0'), grad: tensor([[ 1.5512e-05, 6.7474e-07, -3.4302e-05, ..., -1.1206e-05, + 1.1116e-05, 5.4166e-06], + [ 1.4730e-05, 3.1032e-06, -3.2574e-05, ..., 7.9215e-05, + 1.0673e-06, -2.7925e-05], + [-1.3602e-04, 1.3426e-05, 1.4298e-05, ..., 2.5988e-05, + 3.0205e-05, 1.9684e-05], + ..., + [ 3.9250e-05, 7.5176e-06, 7.9721e-06, ..., 7.1466e-05, + -2.0608e-05, 1.1191e-05], + [ 4.9770e-05, 7.1526e-06, 7.8827e-06, ..., 2.3961e-05, + 1.7837e-05, 1.9614e-06], + [ 5.3421e-06, 6.1616e-06, 1.0103e-05, ..., -3.6740e-04, + -1.0180e-04, 6.1244e-06]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0078, 0.0171, 0.0140, -0.0026, 0.0171, -0.0065, -0.0158, -0.0106, + 0.0245, -0.0090], device='cuda:0'), grad: tensor([ 2.6360e-05, -2.6441e-04, 2.8133e-05, -3.7681e-06, 1.6284e-04, + 9.0122e-05, 1.2589e-04, 1.6522e-04, 1.1390e-04, -4.4394e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 251.32, cls_loss 0.0099 cls_loss_mapping 0.0215 cls_loss_causal 0.6998 re_mapping 0.0132 re_causal 0.0423 /// teacc 98.83 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0642, -0.0102, 0.0885, ..., 0.0267, -0.0680, -0.0626], + [-0.0477, -0.0564, -0.0016, ..., -0.0561, -0.0442, -0.0125], + [ 0.0868, -0.0157, -0.0597, ..., -0.0525, -0.0729, -0.1109], + ..., + [-0.0213, 0.0055, -0.0656, ..., -0.0065, 0.0275, 0.0136], + [-0.0628, -0.0571, -0.0111, ..., -0.0408, 0.0064, -0.0180], + [-0.0147, -0.0315, -0.0125, ..., 0.0562, 0.0239, -0.0665]], + device='cuda:0'), grad: tensor([[ 1.3731e-05, 5.3123e-06, -3.1054e-05, ..., -2.5593e-06, + 3.5137e-05, 1.9178e-05], + [ 4.9400e-04, 6.2108e-05, 7.0632e-06, ..., 3.1769e-05, + 1.7369e-04, 1.1548e-05], + [-1.5569e-04, 1.1176e-04, 2.0444e-05, ..., 1.0118e-05, + 2.0576e-04, 2.4036e-05], + ..., + [-5.4169e-04, -1.8871e-04, 1.6801e-06, ..., -4.4882e-05, + -6.2037e-04, 5.7667e-06], + [ 1.3423e-04, 8.1122e-05, 2.0161e-05, ..., 1.3232e-05, + 8.1480e-05, 3.8117e-05], + [ 3.3200e-05, 2.1651e-05, 1.1072e-05, ..., 2.3901e-04, + 1.5724e-04, 1.7613e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0073, 0.0168, 0.0146, -0.0024, 0.0174, -0.0061, -0.0161, -0.0109, + 0.0242, -0.0089], device='cuda:0'), grad: tensor([ 8.6188e-05, 7.6723e-04, 2.3425e-04, 3.7646e-04, -1.9717e-04, + -3.9434e-04, -2.1148e-04, -1.5202e-03, 3.5119e-04, 5.0783e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 251.63, cls_loss 0.0095 cls_loss_mapping 0.0226 cls_loss_causal 0.7181 re_mapping 0.0129 re_causal 0.0437 /// teacc 98.75 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0654, -0.0102, 0.0888, ..., 0.0264, -0.0683, -0.0628], + [-0.0479, -0.0569, -0.0018, ..., -0.0562, -0.0447, -0.0128], + [ 0.0875, -0.0161, -0.0602, ..., -0.0530, -0.0734, -0.1119], + ..., + [-0.0215, 0.0057, -0.0659, ..., -0.0071, 0.0278, 0.0135], + [-0.0632, -0.0574, -0.0112, ..., -0.0412, 0.0066, -0.0180], + [-0.0138, -0.0316, -0.0119, ..., 0.0571, 0.0240, -0.0670]], + device='cuda:0'), grad: tensor([[ 8.9765e-05, 1.1055e-06, -8.6725e-05, ..., -2.3603e-05, + 3.5077e-05, 3.2455e-05], + [ 2.8074e-05, 3.5204e-06, 1.4611e-05, ..., 4.6968e-05, + 4.2945e-05, 1.3322e-05], + [-9.0790e-04, 1.2435e-05, 2.2098e-05, ..., -3.9291e-04, + -8.2254e-05, 1.1392e-05], + ..., + [ 3.0923e-04, 2.1551e-06, 1.2316e-05, ..., 2.2054e-04, + 8.3447e-05, 7.2382e-06], + [-4.7870e-06, -6.6450e-07, 1.9699e-05, ..., -3.8505e-05, + -1.3471e-04, -3.0696e-05], + [ 2.6274e-04, -4.3362e-06, 3.4541e-05, ..., -1.7118e-04, + -1.1730e-04, 8.8215e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0066, 0.0172, 0.0145, -0.0027, 0.0174, -0.0063, -0.0162, -0.0111, + 0.0244, -0.0082], device='cuda:0'), grad: tensor([ 1.1069e-04, 1.6248e-04, -1.3008e-03, 2.2590e-04, 3.6287e-04, + 2.4390e-04, -1.9357e-05, 6.5708e-04, -7.4434e-04, 3.0112e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 46---------------------------------------------------- +epoch 46, time 268.10, cls_loss 0.0107 cls_loss_mapping 0.0216 cls_loss_causal 0.6859 re_mapping 0.0127 re_causal 0.0399 /// teacc 99.00 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0661, -0.0103, 0.0898, ..., 0.0272, -0.0689, -0.0632], + [-0.0485, -0.0572, -0.0024, ..., -0.0572, -0.0453, -0.0132], + [ 0.0883, -0.0163, -0.0608, ..., -0.0536, -0.0740, -0.1126], + ..., + [-0.0217, 0.0060, -0.0664, ..., -0.0075, 0.0281, 0.0133], + [-0.0637, -0.0576, -0.0108, ..., -0.0414, 0.0068, -0.0182], + [-0.0137, -0.0322, -0.0126, ..., 0.0574, 0.0241, -0.0679]], + device='cuda:0'), grad: tensor([[ 6.8128e-05, 1.6615e-05, -4.9055e-05, ..., -3.2723e-05, + 5.6773e-05, 1.5289e-05], + [ 9.5218e-06, 2.4922e-06, 6.4857e-06, ..., 3.9428e-05, + -1.3733e-03, 2.2836e-06], + [ 6.1095e-06, 9.5740e-06, 5.0068e-05, ..., 4.5836e-05, + 4.4346e-05, 1.6857e-06], + ..., + [ 1.0557e-05, -6.5327e-05, 7.0035e-06, ..., 4.8950e-06, + 6.0320e-04, 1.2470e-06], + [ 8.2493e-05, 2.2903e-05, 8.0585e-05, ..., 1.6439e-04, + 1.6332e-04, 9.3877e-06], + [ 2.3633e-05, 5.5343e-05, 7.4804e-05, ..., 8.8024e-04, + 1.2026e-03, 7.6517e-06]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0070, 0.0168, 0.0144, -0.0027, 0.0178, -0.0063, -0.0165, -0.0110, + 0.0246, -0.0084], device='cuda:0'), grad: tensor([ 1.5163e-04, -1.6891e-02, 2.1470e-04, 3.6573e-04, -1.7185e-03, + -3.7789e-05, 8.5890e-05, 1.3008e-02, 7.9966e-04, 4.0169e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 251.62, cls_loss 0.0081 cls_loss_mapping 0.0210 cls_loss_causal 0.6867 re_mapping 0.0129 re_causal 0.0407 /// teacc 98.99 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0666, -0.0105, 0.0902, ..., 0.0274, -0.0694, -0.0641], + [-0.0490, -0.0578, -0.0025, ..., -0.0578, -0.0456, -0.0135], + [ 0.0890, -0.0169, -0.0613, ..., -0.0540, -0.0744, -0.1133], + ..., + [-0.0220, 0.0063, -0.0667, ..., -0.0080, 0.0284, 0.0132], + [-0.0641, -0.0575, -0.0107, ..., -0.0414, 0.0068, -0.0183], + [-0.0142, -0.0327, -0.0131, ..., 0.0573, 0.0241, -0.0683]], + device='cuda:0'), grad: tensor([[ 2.2501e-06, 1.8273e-06, -8.1062e-05, ..., 4.6976e-06, + 1.7166e-05, 4.2357e-06], + [ 1.0662e-05, 8.3372e-06, 1.8924e-05, ..., 2.9594e-05, + 2.7075e-05, 1.4016e-06], + [-2.5719e-05, 1.3486e-05, 3.0130e-05, ..., 3.4213e-05, + 3.1650e-05, 1.6363e-06], + ..., + [ 1.0453e-05, 3.1013e-06, 1.8537e-05, ..., 1.8549e-04, + 1.2207e-04, 3.6322e-06], + [ 3.0726e-05, 1.5780e-05, -7.5281e-05, ..., -2.1998e-06, + 3.0071e-05, 1.1042e-05], + [ 4.4480e-06, -1.5542e-05, 3.8922e-05, ..., 1.7333e-04, + -1.8072e-04, -3.1907e-06]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0067, 0.0169, 0.0145, -0.0031, 0.0182, -0.0059, -0.0163, -0.0113, + 0.0249, -0.0089], device='cuda:0'), grad: tensor([ 4.9770e-05, 8.2791e-05, 2.7466e-04, 4.6939e-05, -4.5609e-04, + 1.2457e-04, 3.2872e-05, 5.2834e-04, -7.8678e-04, 1.0151e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 48---------------------------------------------------- +epoch 48, time 267.60, cls_loss 0.0091 cls_loss_mapping 0.0235 cls_loss_causal 0.7136 re_mapping 0.0121 re_causal 0.0395 /// teacc 99.03 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0670, -0.0106, 0.0905, ..., 0.0270, -0.0699, -0.0651], + [-0.0493, -0.0583, -0.0027, ..., -0.0584, -0.0461, -0.0139], + [ 0.0898, -0.0169, -0.0620, ..., -0.0544, -0.0749, -0.1142], + ..., + [-0.0223, 0.0064, -0.0671, ..., -0.0088, 0.0284, 0.0130], + [-0.0648, -0.0581, -0.0109, ..., -0.0425, 0.0061, -0.0184], + [-0.0140, -0.0326, -0.0131, ..., 0.0584, 0.0248, -0.0683]], + device='cuda:0'), grad: tensor([[ 2.8580e-05, 3.5856e-07, -1.0389e-04, ..., -2.8685e-05, + 5.6624e-05, 1.2785e-05], + [ 2.2638e-04, 7.7412e-06, 4.8578e-06, ..., 5.2229e-06, + 3.7146e-04, 1.8086e-06], + [ 8.0585e-05, 1.5989e-05, 2.4021e-05, ..., 1.7926e-05, + 2.1422e-04, 1.9670e-05], + ..., + [ 1.5545e-03, 9.0897e-06, 6.0946e-06, ..., 9.8869e-06, + 2.5749e-03, 5.8999e-07], + [ 1.7822e-04, 4.1902e-05, 5.1767e-05, ..., 2.3797e-05, + 2.6751e-04, 3.8922e-05], + [-2.1286e-03, 1.9237e-05, 2.4229e-05, ..., 1.7822e-05, + -3.5934e-03, 3.1646e-06]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0063, 0.0168, 0.0144, -0.0033, 0.0183, -0.0058, -0.0162, -0.0115, + 0.0240, -0.0076], device='cuda:0'), grad: tensor([ 0.0001, 0.0016, 0.0010, 0.0002, 0.0005, 0.0001, -0.0001, 0.0120, + 0.0012, -0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 248.13, cls_loss 0.0079 cls_loss_mapping 0.0200 cls_loss_causal 0.6615 re_mapping 0.0126 re_causal 0.0383 /// teacc 98.74 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0676, -0.0107, 0.0910, ..., 0.0272, -0.0705, -0.0658], + [-0.0499, -0.0586, -0.0028, ..., -0.0585, -0.0465, -0.0143], + [ 0.0906, -0.0171, -0.0626, ..., -0.0547, -0.0752, -0.1151], + ..., + [-0.0224, 0.0064, -0.0675, ..., -0.0091, 0.0289, 0.0128], + [-0.0653, -0.0585, -0.0105, ..., -0.0431, 0.0059, -0.0187], + [-0.0141, -0.0327, -0.0136, ..., 0.0586, 0.0247, -0.0688]], + device='cuda:0'), grad: tensor([[ 9.9361e-05, 5.0664e-05, 1.1951e-04, ..., 1.1528e-04, + 5.0247e-05, 8.1539e-05], + [ 5.2899e-05, 4.8459e-05, 4.0084e-05, ..., 5.4777e-05, + 1.0043e-04, 2.6274e-04], + [ 1.0943e-04, 1.8728e-04, 4.8190e-05, ..., 2.3639e-04, + 1.1247e-04, 4.1664e-05], + ..., + [ 2.0790e-04, 6.5088e-04, 7.0892e-06, ..., 1.9109e-04, + 3.6073e-04, 8.5235e-05], + [ 1.2591e-05, 1.3912e-04, -2.6512e-04, ..., 8.6784e-05, + -2.4199e-04, -3.2425e-04], + [-8.4817e-05, 4.9770e-05, 2.7567e-05, ..., -4.9162e-04, + -1.4845e-06, 4.4554e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0062, 0.0170, 0.0144, -0.0033, 0.0178, -0.0055, -0.0159, -0.0109, + 0.0237, -0.0079], device='cuda:0'), grad: tensor([ 0.0007, 0.0006, 0.0009, -0.0027, -0.0008, 0.0049, -0.0037, 0.0017, + -0.0012, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 251.58, cls_loss 0.0115 cls_loss_mapping 0.0264 cls_loss_causal 0.7045 re_mapping 0.0127 re_causal 0.0390 /// teacc 98.98 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0688, -0.0108, 0.0920, ..., 0.0278, -0.0709, -0.0663], + [-0.0508, -0.0590, -0.0028, ..., -0.0583, -0.0470, -0.0146], + [ 0.0915, -0.0172, -0.0631, ..., -0.0552, -0.0756, -0.1152], + ..., + [-0.0221, 0.0066, -0.0675, ..., -0.0094, 0.0293, 0.0124], + [-0.0658, -0.0586, -0.0107, ..., -0.0436, 0.0059, -0.0191], + [-0.0139, -0.0333, -0.0142, ..., 0.0594, 0.0248, -0.0695]], + device='cuda:0'), grad: tensor([[ 1.2666e-06, -8.8513e-06, -2.4962e-04, ..., -5.6028e-05, + 2.4050e-05, -6.0201e-05], + [ 1.1504e-05, 6.5193e-06, -1.2526e-07, ..., 2.1875e-05, + 1.5289e-05, 2.7884e-06], + [-3.2693e-05, 1.0490e-05, 9.1642e-06, ..., 9.2760e-06, + 1.0915e-05, 3.7458e-06], + ..., + [ 1.8254e-05, 2.9251e-05, 4.9472e-06, ..., 3.8624e-05, + 3.2485e-05, 5.9567e-06], + [ 1.5959e-05, 3.0786e-05, 1.8582e-05, ..., 1.5795e-04, + 1.0085e-04, 1.4372e-05], + [ 7.8678e-06, -4.5508e-05, 3.1382e-05, ..., -1.4620e-03, + -7.4434e-04, 2.1428e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0062, 0.0167, 0.0143, -0.0037, 0.0178, -0.0051, -0.0158, -0.0108, + 0.0236, -0.0077], device='cuda:0'), grad: tensor([-2.8443e-04, -8.2016e-05, 6.1244e-06, 2.8658e-04, 6.2466e-04, + 8.9884e-04, 4.9695e-06, 1.5199e-04, 2.9612e-04, -1.9026e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 251.59, cls_loss 0.0063 cls_loss_mapping 0.0157 cls_loss_causal 0.6804 re_mapping 0.0122 re_causal 0.0390 /// teacc 98.76 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0692, -0.0109, 0.0927, ..., 0.0277, -0.0714, -0.0666], + [-0.0508, -0.0593, -0.0026, ..., -0.0584, -0.0473, -0.0147], + [ 0.0917, -0.0179, -0.0637, ..., -0.0556, -0.0760, -0.1155], + ..., + [-0.0224, 0.0063, -0.0679, ..., -0.0099, 0.0294, 0.0122], + [-0.0661, -0.0587, -0.0107, ..., -0.0441, 0.0061, -0.0189], + [-0.0142, -0.0337, -0.0147, ..., 0.0598, 0.0247, -0.0700]], + device='cuda:0'), grad: tensor([[ 1.0222e-05, 2.6543e-06, -1.1486e-04, ..., -7.6711e-05, + 8.9332e-06, 4.6566e-06], + [ 2.5868e-05, 3.5018e-06, 6.6608e-06, ..., 8.5682e-06, + 1.3277e-05, 5.4054e-06], + [-1.2374e-04, 8.1211e-06, 2.4393e-05, ..., 1.1526e-05, + 2.5466e-05, 2.0429e-05], + ..., + [ 1.7866e-05, -2.3931e-05, 5.8115e-06, ..., 2.6882e-05, + -4.0799e-05, 8.4005e-07], + [ 7.7412e-06, 4.2357e-06, 1.5348e-05, ..., 2.7925e-05, + 4.4495e-05, 2.6062e-05], + [ 3.2261e-06, 4.2878e-06, 4.4882e-05, ..., -5.8711e-06, + -3.1829e-05, 2.6040e-06]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0060, 0.0174, 0.0138, -0.0030, 0.0183, -0.0057, -0.0160, -0.0111, + 0.0239, -0.0080], device='cuda:0'), grad: tensor([-8.8394e-05, 5.4270e-05, -5.3227e-05, 1.5414e-04, -1.3456e-05, + -4.4972e-05, -7.2479e-05, -5.2691e-05, 1.1146e-04, 5.5358e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 251.75, cls_loss 0.0083 cls_loss_mapping 0.0192 cls_loss_causal 0.6576 re_mapping 0.0113 re_causal 0.0359 /// teacc 98.90 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0697, -0.0109, 0.0934, ..., 0.0279, -0.0719, -0.0672], + [-0.0506, -0.0598, -0.0030, ..., -0.0587, -0.0481, -0.0170], + [ 0.0924, -0.0183, -0.0642, ..., -0.0559, -0.0763, -0.1163], + ..., + [-0.0228, 0.0061, -0.0683, ..., -0.0102, 0.0295, 0.0121], + [-0.0671, -0.0592, -0.0107, ..., -0.0450, 0.0062, -0.0182], + [-0.0146, -0.0338, -0.0151, ..., 0.0600, 0.0248, -0.0706]], + device='cuda:0'), grad: tensor([[ 3.8117e-05, 4.8578e-06, -2.0072e-05, ..., 2.0955e-06, + 2.8074e-05, 6.3106e-06], + [ 2.1413e-05, 8.1435e-06, -5.2378e-06, ..., 1.0259e-05, + 2.0370e-05, 7.4785e-07], + [-3.8892e-05, 8.0466e-06, 2.9206e-06, ..., 2.6211e-05, + 1.5467e-05, 1.5600e-06], + ..., + [ 3.5346e-05, 1.9476e-05, 1.3262e-06, ..., 1.5318e-05, + 8.8988e-07, 7.2131e-07], + [ 3.1322e-05, 2.3991e-06, 4.5933e-06, ..., 4.2439e-05, + 3.7014e-05, 1.2778e-06], + [-2.2590e-04, 7.1824e-06, 8.2925e-06, ..., -3.5262e-04, + -1.1092e-04, 3.1721e-06]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0060, 0.0172, 0.0138, -0.0031, 0.0182, -0.0054, -0.0159, -0.0113, + 0.0243, -0.0081], device='cuda:0'), grad: tensor([ 1.1927e-04, 6.2990e-04, 2.8700e-05, 5.6362e-04, 5.7936e-05, + -1.3266e-03, 6.6996e-05, 7.8261e-05, 3.8600e-04, -6.0225e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 251.69, cls_loss 0.0082 cls_loss_mapping 0.0217 cls_loss_causal 0.6682 re_mapping 0.0122 re_causal 0.0373 /// teacc 98.99 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0708, -0.0110, 0.0945, ..., 0.0278, -0.0726, -0.0676], + [-0.0509, -0.0600, -0.0045, ..., -0.0593, -0.0486, -0.0174], + [ 0.0936, -0.0184, -0.0649, ..., -0.0566, -0.0768, -0.1168], + ..., + [-0.0233, 0.0056, -0.0689, ..., -0.0112, 0.0295, 0.0118], + [-0.0681, -0.0596, -0.0108, ..., -0.0460, 0.0062, -0.0183], + [-0.0144, -0.0341, -0.0154, ..., 0.0613, 0.0249, -0.0714]], + device='cuda:0'), grad: tensor([[ 1.7628e-05, 1.3094e-06, -6.1274e-04, ..., -3.0065e-04, + -1.4573e-05, -2.3320e-05], + [ 1.1265e-04, 6.4187e-06, 1.5229e-05, ..., 8.3864e-05, + 4.2289e-05, 2.9523e-06], + [-2.4402e-04, 1.1869e-05, 6.3777e-05, ..., 4.8220e-05, + 1.3061e-05, 7.3016e-06], + ..., + [ 5.4598e-05, 7.7635e-06, 1.8239e-05, ..., 5.5885e-04, + 3.0279e-04, 1.6298e-06], + [ 4.3243e-05, 8.1807e-06, 8.2612e-05, ..., 7.4744e-05, + 8.0764e-06, -4.0932e-07], + [ 2.0951e-05, 9.2238e-06, 8.8394e-05, ..., -6.7568e-04, + -4.3201e-04, 7.8529e-06]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0061, 0.0168, 0.0143, -0.0032, 0.0179, -0.0052, -0.0158, -0.0117, + 0.0239, -0.0076], device='cuda:0'), grad: tensor([-8.6689e-04, 1.8728e-04, -1.2189e-04, 3.2544e-05, 8.7321e-05, + 1.2052e-04, 3.5524e-04, 1.0147e-03, 2.0599e-04, -1.0157e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 54---------------------------------------------------- +epoch 54, time 268.65, cls_loss 0.0087 cls_loss_mapping 0.0201 cls_loss_causal 0.6687 re_mapping 0.0112 re_causal 0.0357 /// teacc 99.06 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0713, -0.0113, 0.0956, ..., 0.0283, -0.0730, -0.0681], + [-0.0517, -0.0605, -0.0046, ..., -0.0590, -0.0491, -0.0176], + [ 0.0944, -0.0201, -0.0656, ..., -0.0570, -0.0773, -0.1186], + ..., + [-0.0237, 0.0051, -0.0694, ..., -0.0122, 0.0296, 0.0117], + [-0.0687, -0.0601, -0.0102, ..., -0.0466, 0.0062, -0.0179], + [-0.0151, -0.0341, -0.0160, ..., 0.0611, 0.0254, -0.0717]], + device='cuda:0'), grad: tensor([[ 9.0972e-06, 8.2841e-07, -1.7345e-04, ..., -5.5313e-05, + 7.4767e-06, -2.9989e-07], + [ 6.3717e-05, 1.4538e-06, -2.9534e-05, ..., 2.3544e-05, + 1.8731e-05, -5.4464e-06], + [-1.5080e-04, 7.8008e-06, 2.2188e-05, ..., -1.6531e-07, + 1.4387e-05, 3.1423e-06], + ..., + [ 3.4153e-05, 2.5406e-06, 4.0419e-06, ..., 1.3404e-05, + -5.6714e-05, 5.5227e-07], + [ 1.9297e-05, 2.8647e-06, 3.4064e-05, ..., 1.6898e-05, + 1.4067e-05, 9.6112e-06], + [ 5.7966e-06, 2.9858e-06, 4.9531e-05, ..., 3.1829e-05, + 2.7597e-05, 2.7306e-06]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0064, 0.0167, 0.0142, -0.0028, 0.0184, -0.0049, -0.0164, -0.0118, + 0.0238, -0.0079], device='cuda:0'), grad: tensor([-1.8322e-04, 3.4809e-05, -8.6546e-05, 5.6982e-05, -1.2118e-04, + 1.9825e-04, -3.8117e-05, -1.0133e-04, 1.0967e-04, 1.3030e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 251.08, cls_loss 0.0088 cls_loss_mapping 0.0218 cls_loss_causal 0.6582 re_mapping 0.0113 re_causal 0.0363 /// teacc 98.89 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0718, -0.0114, 0.0964, ..., 0.0289, -0.0739, -0.0696], + [-0.0524, -0.0606, -0.0046, ..., -0.0584, -0.0489, -0.0178], + [ 0.0952, -0.0203, -0.0668, ..., -0.0578, -0.0778, -0.1194], + ..., + [-0.0241, 0.0050, -0.0699, ..., -0.0129, 0.0296, 0.0114], + [-0.0690, -0.0602, -0.0097, ..., -0.0479, 0.0063, -0.0175], + [-0.0154, -0.0342, -0.0170, ..., 0.0612, 0.0256, -0.0720]], + device='cuda:0'), grad: tensor([[ 2.4717e-06, 1.9148e-06, -4.3005e-05, ..., -3.9041e-05, + 1.0975e-05, 1.0520e-05], + [ 4.7907e-06, 2.0321e-06, -3.1263e-05, ..., -2.2173e-05, + 9.9689e-06, -1.8533e-06], + [ 6.1631e-05, 4.1366e-05, 1.2346e-05, ..., 1.4640e-05, + 1.8477e-05, 2.9933e-06], + ..., + [ 1.7062e-05, 5.4389e-06, 7.5810e-06, ..., 1.2092e-05, + -4.8488e-05, 5.5647e-07], + [ 2.8923e-05, 1.4007e-05, 2.1532e-05, ..., 2.1890e-05, + 1.5751e-05, 1.2547e-05], + [ 1.2055e-05, 7.8082e-06, 2.6703e-05, ..., 1.9267e-05, + 1.5795e-05, 2.5947e-06]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0065, 0.0172, 0.0142, -0.0029, 0.0184, -0.0053, -0.0162, -0.0121, + 0.0239, -0.0080], device='cuda:0'), grad: tensor([-5.0180e-06, -1.2338e-04, 1.7774e-04, -8.6725e-05, 8.0615e-06, + -5.1588e-05, -1.4782e-05, -1.4865e-04, 1.2398e-04, 1.2016e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 251.67, cls_loss 0.0065 cls_loss_mapping 0.0163 cls_loss_causal 0.6549 re_mapping 0.0109 re_causal 0.0354 /// teacc 98.83 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0724, -0.0114, 0.0967, ..., 0.0289, -0.0746, -0.0704], + [-0.0533, -0.0609, -0.0045, ..., -0.0588, -0.0497, -0.0196], + [ 0.0963, -0.0206, -0.0674, ..., -0.0583, -0.0784, -0.1174], + ..., + [-0.0246, 0.0049, -0.0704, ..., -0.0132, 0.0299, 0.0103], + [-0.0694, -0.0604, -0.0102, ..., -0.0484, 0.0063, -0.0181], + [-0.0156, -0.0345, -0.0173, ..., 0.0617, 0.0257, -0.0730]], + device='cuda:0'), grad: tensor([[ 4.0948e-05, 1.1273e-05, -1.1826e-04, ..., -6.0648e-05, + 1.6034e-05, 6.3516e-06], + [ 7.0274e-05, 2.9013e-05, 3.6955e-06, ..., 1.6242e-05, + 3.4928e-05, 7.2680e-06], + [-7.1859e-04, 7.4923e-05, 1.9893e-05, ..., 1.4350e-05, + 8.9765e-05, 9.6858e-06], + ..., + [ 5.4538e-05, -2.1303e-04, 5.9977e-06, ..., 6.5826e-06, + -3.7456e-04, 8.4490e-06], + [ 4.1127e-05, 1.0438e-05, 1.4752e-06, ..., 7.9051e-06, + 4.4964e-06, -2.3432e-06], + [ 3.2043e-04, 1.2629e-05, 4.5270e-05, ..., 3.0965e-05, + 2.1651e-05, 1.1452e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0061, 0.0165, 0.0147, -0.0029, 0.0181, -0.0053, -0.0154, -0.0118, + 0.0237, -0.0079], device='cuda:0'), grad: tensor([-2.9653e-06, 2.3222e-04, -7.7820e-04, 4.4733e-05, 1.4603e-04, + 3.2258e-04, 1.2493e-04, -7.6437e-04, 6.1333e-05, 6.1464e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 251.38, cls_loss 0.0065 cls_loss_mapping 0.0171 cls_loss_causal 0.6743 re_mapping 0.0111 re_causal 0.0355 /// teacc 98.98 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0733, -0.0115, 0.0970, ..., 0.0292, -0.0754, -0.0716], + [-0.0538, -0.0612, -0.0045, ..., -0.0590, -0.0502, -0.0201], + [ 0.0974, -0.0207, -0.0672, ..., -0.0586, -0.0788, -0.1182], + ..., + [-0.0250, 0.0049, -0.0713, ..., -0.0133, 0.0306, 0.0105], + [-0.0702, -0.0605, -0.0096, ..., -0.0492, 0.0065, -0.0174], + [-0.0161, -0.0347, -0.0179, ..., 0.0617, 0.0254, -0.0747]], + device='cuda:0'), grad: tensor([[ 7.2690e-07, 1.3104e-06, -1.3304e-04, ..., -6.8545e-05, + 1.9372e-05, -4.2468e-05], + [ 1.0738e-06, 1.9558e-06, -6.6273e-06, ..., 4.8816e-05, + 2.9072e-05, -9.5461e-09], + [-3.6117e-06, 2.5649e-06, 1.2942e-05, ..., 2.3484e-05, + 1.1794e-05, 7.6219e-06], + ..., + [ 1.5292e-06, 1.0884e-04, 1.4812e-05, ..., 6.1655e-04, + 6.2990e-04, 5.7258e-06], + [ 2.0657e-06, 5.0701e-06, 2.4468e-05, ..., 6.0588e-05, + 4.3631e-05, 1.4566e-05], + [ 1.3607e-06, -1.1659e-04, 2.5421e-05, ..., -4.9973e-04, + -7.0000e-04, 9.7081e-06]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0059, 0.0162, 0.0151, -0.0023, 0.0182, -0.0060, -0.0152, -0.0115, + 0.0239, -0.0083], device='cuda:0'), grad: tensor([-1.1808e-04, 7.7128e-05, 5.5075e-05, 7.5221e-05, -5.1212e-04, + 1.2589e-04, 3.1620e-05, 1.3876e-03, 1.7428e-04, -1.2951e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 251.40, cls_loss 0.0060 cls_loss_mapping 0.0167 cls_loss_causal 0.6390 re_mapping 0.0112 re_causal 0.0347 /// teacc 98.89 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0737, -0.0114, 0.0976, ..., 0.0291, -0.0760, -0.0720], + [-0.0541, -0.0616, -0.0043, ..., -0.0593, -0.0507, -0.0206], + [ 0.0983, -0.0211, -0.0678, ..., -0.0592, -0.0793, -0.1186], + ..., + [-0.0255, 0.0052, -0.0717, ..., -0.0138, 0.0310, 0.0093], + [-0.0707, -0.0608, -0.0099, ..., -0.0498, 0.0063, -0.0180], + [-0.0162, -0.0350, -0.0185, ..., 0.0623, 0.0256, -0.0757]], + device='cuda:0'), grad: tensor([[ 1.0490e-05, 4.8615e-07, -4.9496e-04, ..., 1.0127e-04, + -3.0100e-05, -4.7874e-04], + [ 3.8415e-05, 1.4640e-06, 2.8491e-05, ..., 1.8924e-05, + 1.9312e-05, 2.3410e-05], + [-1.6642e-04, 2.6580e-06, 5.4598e-05, ..., 2.2531e-05, + 3.4481e-05, 3.3408e-05], + ..., + [ 1.7270e-05, 3.8967e-06, 1.8731e-05, ..., 3.5763e-04, + 3.2592e-04, 1.4767e-05], + [ 5.3883e-05, 2.8871e-06, -1.3649e-05, ..., 1.3478e-05, + -3.3677e-05, 5.7936e-05], + [ 4.8988e-06, 3.2764e-06, 1.0657e-04, ..., -2.8491e-05, + 2.0072e-05, 7.3016e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0057, 0.0162, 0.0152, -0.0024, 0.0180, -0.0057, -0.0152, -0.0115, + 0.0237, -0.0082], device='cuda:0'), grad: tensor([-9.6416e-04, 1.0800e-04, -4.5657e-05, 6.1750e-05, -7.3767e-04, + 5.1498e-04, -4.4763e-05, 9.7609e-04, -1.2040e-05, 1.4555e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 251.63, cls_loss 0.0058 cls_loss_mapping 0.0152 cls_loss_causal 0.6424 re_mapping 0.0104 re_causal 0.0341 /// teacc 99.03 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0741, -0.0115, 0.0985, ..., 0.0297, -0.0767, -0.0721], + [-0.0546, -0.0618, -0.0047, ..., -0.0599, -0.0512, -0.0212], + [ 0.0986, -0.0234, -0.0687, ..., -0.0598, -0.0802, -0.1184], + ..., + [-0.0256, 0.0052, -0.0722, ..., -0.0141, 0.0315, 0.0089], + [-0.0712, -0.0609, -0.0101, ..., -0.0507, 0.0062, -0.0183], + [-0.0164, -0.0349, -0.0187, ..., 0.0629, 0.0258, -0.0763]], + device='cuda:0'), grad: tensor([[ 3.0883e-06, 1.7837e-05, 5.5999e-05, ..., 1.4506e-05, + 3.1084e-05, 2.7210e-05], + [ 4.4443e-06, 2.4792e-06, -5.7481e-06, ..., -9.3132e-07, + 5.0291e-06, 3.1833e-06], + [-2.6256e-05, 9.1344e-06, 3.2753e-05, ..., 1.2368e-05, + 1.1854e-05, 8.4490e-06], + ..., + [ 4.3288e-06, 1.0142e-06, 7.7784e-06, ..., 2.9001e-06, + -1.3625e-06, 6.9384e-07], + [ 5.2936e-06, 6.9514e-06, 2.2447e-04, ..., -1.8775e-05, + 1.2600e-04, 2.3723e-04], + [ 8.1025e-07, 1.6363e-06, 7.1704e-05, ..., 3.0264e-05, + 6.7770e-05, 6.1952e-06]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0061, 0.0161, 0.0147, -0.0020, 0.0176, -0.0059, -0.0150, -0.0111, + 0.0232, -0.0078], device='cuda:0'), grad: tensor([ 1.5640e-04, -5.5790e-05, 8.1718e-05, 2.2382e-05, 1.3077e-04, + 2.6774e-04, -1.0319e-03, 3.0667e-05, 4.3690e-05, 3.5405e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 251.39, cls_loss 0.0055 cls_loss_mapping 0.0131 cls_loss_causal 0.6390 re_mapping 0.0103 re_causal 0.0338 /// teacc 98.92 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0746, -0.0115, 0.0986, ..., 0.0294, -0.0773, -0.0729], + [-0.0550, -0.0620, -0.0047, ..., -0.0603, -0.0516, -0.0219], + [ 0.0996, -0.0235, -0.0690, ..., -0.0600, -0.0804, -0.1182], + ..., + [-0.0262, 0.0051, -0.0726, ..., -0.0143, 0.0317, 0.0085], + [-0.0720, -0.0612, -0.0102, ..., -0.0510, 0.0061, -0.0184], + [-0.0169, -0.0350, -0.0188, ..., 0.0635, 0.0260, -0.0771]], + device='cuda:0'), grad: tensor([[-8.8066e-06, -6.7241e-06, -1.7655e-04, ..., -1.0693e-04, + 2.2396e-05, 2.5466e-05], + [ 1.7332e-06, 2.5220e-06, 9.3505e-06, ..., 3.2056e-06, + 2.8268e-05, 2.8238e-05], + [ 5.5246e-06, 5.1484e-06, 4.5151e-05, ..., 3.0011e-05, + 2.2486e-05, 2.2978e-05], + ..., + [ 4.8429e-06, -1.3709e-06, 8.2329e-06, ..., 4.9025e-06, + -2.8148e-05, 4.0233e-06], + [ 3.1311e-06, 4.1202e-06, 3.5465e-05, ..., 5.1141e-05, + 5.2094e-05, 4.1693e-05], + [ 9.8720e-06, 9.0450e-06, 8.0228e-05, ..., -1.1415e-03, + -5.2547e-04, 2.4006e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0056, 0.0160, 0.0152, -0.0016, 0.0175, -0.0062, -0.0148, -0.0111, + 0.0229, -0.0078], device='cuda:0'), grad: tensor([-1.4198e-04, -8.9929e-06, 1.0914e-04, 1.0557e-03, 1.3771e-03, + -1.0147e-03, -2.1696e-04, -1.6227e-05, 1.4615e-04, -1.2884e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 251.12, cls_loss 0.0059 cls_loss_mapping 0.0171 cls_loss_causal 0.6245 re_mapping 0.0104 re_causal 0.0321 /// teacc 98.99 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0747, -0.0115, 0.0992, ..., 0.0294, -0.0778, -0.0737], + [-0.0551, -0.0623, -0.0047, ..., -0.0609, -0.0522, -0.0226], + [ 0.1002, -0.0237, -0.0700, ..., -0.0606, -0.0808, -0.1190], + ..., + [-0.0264, 0.0051, -0.0732, ..., -0.0147, 0.0321, 0.0081], + [-0.0724, -0.0613, -0.0100, ..., -0.0516, 0.0063, -0.0179], + [-0.0172, -0.0352, -0.0193, ..., 0.0637, 0.0259, -0.0785]], + device='cuda:0'), grad: tensor([[ 3.0443e-05, 9.0292e-07, -1.8273e-06, ..., -1.2569e-05, + 1.0572e-05, 1.7598e-05], + [ 6.0424e-06, 2.8126e-07, 2.1216e-06, ..., -2.2678e-07, + 3.3733e-06, 2.7344e-06], + [-8.5735e-04, 4.4564e-07, 1.6496e-05, ..., -2.2605e-05, + -1.8090e-05, 4.3698e-06], + ..., + [ 8.5473e-05, 1.6857e-07, 2.5928e-06, ..., 3.5837e-06, + -3.2242e-06, 5.4017e-07], + [ 5.0098e-05, 3.4086e-06, 2.0042e-06, ..., 1.5147e-05, + -1.7649e-06, 1.2413e-05], + [ 3.4928e-04, 5.9791e-07, 1.3337e-05, ..., -1.6734e-05, + 1.1124e-05, 5.3458e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0055, 0.0157, 0.0150, -0.0016, 0.0179, -0.0063, -0.0149, -0.0110, + 0.0233, -0.0079], device='cuda:0'), grad: tensor([ 6.7592e-05, -1.3836e-05, -9.7561e-04, 3.6478e-04, 1.0175e-04, + 7.2837e-05, -1.1611e-04, 1.0945e-05, 1.9342e-05, 4.6730e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 251.56, cls_loss 0.0058 cls_loss_mapping 0.0144 cls_loss_causal 0.6661 re_mapping 0.0101 re_causal 0.0330 /// teacc 98.99 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0753, -0.0116, 0.1000, ..., 0.0292, -0.0782, -0.0750], + [-0.0552, -0.0625, -0.0056, ..., -0.0616, -0.0525, -0.0231], + [ 0.1013, -0.0239, -0.0709, ..., -0.0608, -0.0808, -0.1200], + ..., + [-0.0275, 0.0051, -0.0737, ..., -0.0157, 0.0321, 0.0079], + [-0.0732, -0.0613, -0.0098, ..., -0.0523, 0.0062, -0.0177], + [-0.0174, -0.0354, -0.0197, ..., 0.0644, 0.0260, -0.0793]], + device='cuda:0'), grad: tensor([[ 3.4878e-07, 5.2154e-08, -2.4810e-05, ..., -1.6659e-05, + 4.8801e-06, 4.1351e-06], + [ 9.9465e-07, 2.3982e-07, 1.3150e-06, ..., 9.0245e-07, + 6.9737e-06, 1.1260e-06], + [ 4.6566e-07, 2.7474e-07, 3.2485e-06, ..., 4.5002e-06, + 2.7299e-05, 1.4696e-06], + ..., + [-2.5351e-06, 2.5611e-07, 1.8505e-06, ..., -5.0627e-06, + -7.5042e-05, 3.9721e-07], + [-5.8971e-06, 2.9337e-07, 3.7514e-06, ..., 3.7365e-06, + -2.9638e-05, 3.4962e-06], + [ 4.9919e-06, 3.7253e-07, 1.1146e-05, ..., 2.6766e-06, + 4.4137e-05, 2.4121e-06]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0050, 0.0157, 0.0154, -0.0013, 0.0179, -0.0068, -0.0148, -0.0115, + 0.0235, -0.0076], device='cuda:0'), grad: tensor([-1.3471e-05, 6.9812e-06, 9.5904e-05, 4.7147e-05, 1.5974e-05, + 2.6338e-06, -7.9116e-07, -1.6165e-04, -1.5259e-04, 1.5950e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 251.38, cls_loss 0.0058 cls_loss_mapping 0.0172 cls_loss_causal 0.6736 re_mapping 0.0102 re_causal 0.0326 /// teacc 98.89 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0763, -0.0117, 0.1003, ..., 0.0293, -0.0787, -0.0758], + [-0.0559, -0.0627, -0.0058, ..., -0.0615, -0.0525, -0.0236], + [ 0.1022, -0.0241, -0.0713, ..., -0.0612, -0.0813, -0.1208], + ..., + [-0.0278, 0.0048, -0.0740, ..., -0.0162, 0.0323, 0.0077], + [-0.0737, -0.0615, -0.0100, ..., -0.0531, 0.0060, -0.0177], + [-0.0174, -0.0355, -0.0202, ..., 0.0644, 0.0258, -0.0798]], + device='cuda:0'), grad: tensor([[ 2.3282e-04, 1.5413e-07, -5.6314e-04, ..., -9.0075e-04, + 2.4974e-05, 2.4587e-05], + [ 5.1372e-06, 8.8848e-07, 8.4043e-06, ..., 6.8136e-06, + 9.6560e-06, 5.8301e-06], + [-4.5919e-04, 1.2098e-06, -2.3377e-04, ..., -1.1635e-04, + 7.3351e-06, 5.1707e-06], + ..., + [ 5.1782e-06, 1.2340e-07, 8.0392e-06, ..., 1.1832e-05, + -1.3687e-05, 6.3609e-07], + [ 1.3351e-05, 1.1800e-06, 1.6451e-05, ..., 1.9804e-05, + -6.0769e-07, -7.3425e-06], + [ 8.8960e-06, 4.5337e-06, 6.9714e-04, ..., 8.7309e-04, + -6.7987e-06, 5.1893e-06]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0048, 0.0156, 0.0156, -0.0009, 0.0183, -0.0064, -0.0146, -0.0118, + 0.0233, -0.0080], device='cuda:0'), grad: tensor([-1.0233e-03, 1.0200e-05, -5.9509e-04, 2.3174e-04, 4.3333e-05, + 6.2943e-05, -7.3969e-05, 2.0921e-05, -5.9903e-06, 1.3285e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 251.62, cls_loss 0.0048 cls_loss_mapping 0.0139 cls_loss_causal 0.5864 re_mapping 0.0101 re_causal 0.0305 /// teacc 98.75 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0775, -0.0117, 0.1003, ..., 0.0298, -0.0792, -0.0778], + [-0.0562, -0.0627, -0.0058, ..., -0.0616, -0.0531, -0.0238], + [ 0.1028, -0.0244, -0.0710, ..., -0.0616, -0.0818, -0.1217], + ..., + [-0.0280, 0.0048, -0.0746, ..., -0.0164, 0.0329, 0.0080], + [-0.0741, -0.0618, -0.0102, ..., -0.0536, 0.0057, -0.0183], + [-0.0178, -0.0358, -0.0209, ..., 0.0646, 0.0258, -0.0809]], + device='cuda:0'), grad: tensor([[-3.2922e-07, 2.1942e-06, -1.7971e-05, ..., -2.2113e-05, + 7.1190e-06, 5.0217e-06], + [ 5.1362e-07, -7.3671e-05, 1.1679e-06, ..., -1.2480e-07, + -5.0336e-05, -1.1593e-04], + [-1.1690e-05, 1.7449e-05, 3.2127e-05, ..., 1.0088e-05, + 2.5526e-05, 3.1531e-05], + ..., + [ 1.8505e-06, 1.4305e-05, 3.2410e-06, ..., 8.7917e-06, + -8.7097e-06, 2.2188e-05], + [ 1.1977e-06, 9.1046e-06, -5.3495e-05, ..., 3.8780e-06, + -4.1388e-06, 1.9103e-05], + [ 6.7055e-07, 1.2126e-06, 2.0817e-05, ..., -1.2122e-05, + -3.4869e-06, 2.1998e-06]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0042, 0.0155, 0.0156, -0.0010, 0.0181, -0.0060, -0.0141, -0.0110, + 0.0230, -0.0083], device='cuda:0'), grad: tensor([-9.1642e-06, -5.7030e-04, 1.8191e-04, 1.8215e-04, 2.6152e-05, + 7.7128e-05, 9.9391e-06, 7.8917e-05, -1.7527e-06, 2.5317e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 251.58, cls_loss 0.0057 cls_loss_mapping 0.0139 cls_loss_causal 0.6449 re_mapping 0.0104 re_causal 0.0311 /// teacc 98.85 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0781, -0.0118, 0.1013, ..., 0.0305, -0.0792, -0.0778], + [-0.0569, -0.0628, -0.0059, ..., -0.0618, -0.0533, -0.0242], + [ 0.1039, -0.0247, -0.0716, ..., -0.0618, -0.0823, -0.1228], + ..., + [-0.0285, 0.0050, -0.0751, ..., -0.0166, 0.0332, 0.0077], + [-0.0746, -0.0623, -0.0097, ..., -0.0544, 0.0056, -0.0183], + [-0.0180, -0.0359, -0.0216, ..., 0.0647, 0.0259, -0.0816]], + device='cuda:0'), grad: tensor([[ 2.8517e-06, 1.5963e-06, -1.9431e-05, ..., -1.0371e-05, + 2.6226e-06, 2.2147e-06], + [ 4.9734e-04, 9.1083e-07, 3.0883e-06, ..., -6.3069e-06, + 1.5758e-06, -2.8200e-06], + [-5.7650e-04, 7.5856e-07, 3.7365e-06, ..., 4.7088e-06, + 1.1437e-06, -5.1688e-08], + ..., + [ 4.4346e-05, 3.4422e-06, 1.2685e-06, ..., 8.8811e-06, + 3.9451e-06, 2.3376e-06], + [ 9.8050e-06, 1.9409e-06, 1.3979e-06, ..., 9.0674e-06, + 1.5441e-06, -2.4326e-06], + [ 1.3914e-06, 3.5316e-06, 3.9376e-06, ..., -1.0394e-05, + -7.2569e-06, 2.0321e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0047, 0.0153, 0.0162, -0.0018, 0.0181, -0.0061, -0.0142, -0.0113, + 0.0234, -0.0083], device='cuda:0'), grad: tensor([ 2.2322e-05, 1.6451e-03, -7.4816e-04, 3.8815e-04, 1.5751e-05, + 1.4573e-05, 2.9057e-05, -1.4458e-03, 4.8697e-05, 3.2455e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 251.57, cls_loss 0.0065 cls_loss_mapping 0.0158 cls_loss_causal 0.6482 re_mapping 0.0093 re_causal 0.0289 /// teacc 99.01 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0788, -0.0119, 0.1025, ..., 0.0310, -0.0798, -0.0781], + [-0.0573, -0.0630, -0.0059, ..., -0.0638, -0.0540, -0.0246], + [ 0.1040, -0.0250, -0.0729, ..., -0.0631, -0.0827, -0.1228], + ..., + [-0.0273, 0.0048, -0.0755, ..., -0.0168, 0.0335, 0.0073], + [-0.0753, -0.0624, -0.0101, ..., -0.0553, 0.0056, -0.0184], + [-0.0183, -0.0364, -0.0223, ..., 0.0653, 0.0260, -0.0822]], + device='cuda:0'), grad: tensor([[ 2.0042e-06, 1.6615e-06, -1.2740e-05, ..., -6.2808e-06, + 6.2361e-06, 3.9600e-06], + [ 3.1069e-06, 1.0887e-06, -4.4107e-05, ..., 2.0694e-06, + 2.5965e-06, -1.1399e-05], + [-2.3544e-05, 3.1237e-06, 5.0925e-06, ..., 2.8938e-05, + 1.0118e-05, 1.9111e-06], + ..., + [ 5.6103e-06, 3.2395e-05, 3.6806e-06, ..., 1.4842e-05, + 4.0114e-05, 1.9781e-06], + [ 6.5677e-06, 1.6568e-06, 1.2442e-05, ..., 6.8173e-06, + -1.7524e-05, -3.2615e-06], + [ 1.4668e-06, 3.5819e-06, 8.9481e-06, ..., -5.9530e-06, + -8.1658e-06, 2.6003e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0055, 0.0144, 0.0151, -0.0017, 0.0179, -0.0062, -0.0147, -0.0099, + 0.0233, -0.0080], device='cuda:0'), grad: tensor([ 1.5929e-05, -1.1420e-04, 6.0111e-05, -1.7154e-04, -2.4462e-04, + 5.8353e-05, 2.3234e-04, 1.8013e-04, -3.0145e-05, 1.3635e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 251.62, cls_loss 0.0043 cls_loss_mapping 0.0133 cls_loss_causal 0.6305 re_mapping 0.0098 re_causal 0.0313 /// teacc 98.93 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0796, -0.0120, 0.1031, ..., 0.0315, -0.0804, -0.0786], + [-0.0578, -0.0632, -0.0067, ..., -0.0634, -0.0541, -0.0250], + [ 0.1047, -0.0251, -0.0732, ..., -0.0636, -0.0829, -0.1237], + ..., + [-0.0276, 0.0046, -0.0758, ..., -0.0175, 0.0338, 0.0069], + [-0.0755, -0.0624, -0.0103, ..., -0.0560, 0.0055, -0.0185], + [-0.0185, -0.0366, -0.0227, ..., 0.0652, 0.0258, -0.0829]], + device='cuda:0'), grad: tensor([[-7.6415e-07, 1.1986e-06, 1.3037e-03, ..., -2.2078e-04, + 1.8463e-05, 1.2369e-03], + [ 9.3281e-06, 3.7625e-06, 1.2875e-05, ..., 4.8466e-06, + 5.4948e-06, 7.4692e-06], + [-8.6278e-06, 6.9514e-06, 7.5996e-05, ..., 3.0309e-05, + 7.7114e-06, 2.3231e-05], + ..., + [ 1.2524e-05, -6.8434e-06, 1.2152e-05, ..., 2.7299e-05, + -4.2588e-05, 5.4017e-06], + [ 1.3128e-05, 2.9616e-06, 1.8883e-04, ..., 7.0035e-05, + 8.1882e-06, 3.9458e-05], + [ 3.7700e-06, 4.2729e-06, 1.3614e-04, ..., -2.7686e-05, + -1.0550e-05, 9.6038e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0056, 0.0145, 0.0153, -0.0021, 0.0183, -0.0057, -0.0146, -0.0099, + 0.0232, -0.0087], device='cuda:0'), grad: tensor([ 1.6851e-03, 2.0564e-05, 1.4067e-04, 7.3016e-05, 3.4511e-05, + 3.6478e-04, -2.6703e-03, -6.8545e-05, 3.1567e-04, 1.0341e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 251.48, cls_loss 0.0044 cls_loss_mapping 0.0112 cls_loss_causal 0.6032 re_mapping 0.0096 re_causal 0.0305 /// teacc 98.98 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0801, -0.0121, 0.1038, ..., 0.0321, -0.0809, -0.0794], + [-0.0578, -0.0632, -0.0069, ..., -0.0637, -0.0543, -0.0250], + [ 0.1055, -0.0252, -0.0741, ..., -0.0641, -0.0832, -0.1244], + ..., + [-0.0279, 0.0043, -0.0761, ..., -0.0178, 0.0340, 0.0065], + [-0.0769, -0.0623, -0.0092, ..., -0.0572, 0.0059, -0.0170], + [-0.0188, -0.0368, -0.0230, ..., 0.0656, 0.0260, -0.0837]], + device='cuda:0'), grad: tensor([[ 2.0154e-06, 8.6613e-08, -5.0306e-05, ..., -2.8849e-05, + 1.0744e-05, 1.6272e-05], + [ 2.6137e-05, 1.1688e-07, 2.3022e-06, ..., 1.7762e-05, + 6.0558e-05, 9.1642e-06], + [-1.6585e-05, 3.2829e-07, 3.8259e-06, ..., 1.7419e-05, + 1.8179e-05, 1.6605e-06], + ..., + [-4.5925e-05, 2.7195e-07, 3.0380e-06, ..., 2.0832e-05, + -3.4571e-05, 3.4161e-06], + [ 1.1005e-05, 5.3551e-08, 2.6256e-05, ..., 2.3976e-05, + -7.3731e-05, -3.6985e-05], + [ 1.8608e-06, 3.1665e-07, 1.1235e-05, ..., 7.8440e-04, + 4.2129e-04, 1.1235e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0059, 0.0151, 0.0153, -0.0016, 0.0181, -0.0060, -0.0155, -0.0099, + 0.0231, -0.0088], device='cuda:0'), grad: tensor([-5.8502e-05, 1.9646e-04, 4.1127e-05, 9.4652e-05, -1.1053e-03, + 3.2693e-05, 2.1383e-05, -1.1939e-04, -2.5368e-04, 1.1511e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 251.68, cls_loss 0.0046 cls_loss_mapping 0.0135 cls_loss_causal 0.6688 re_mapping 0.0090 re_causal 0.0303 /// teacc 98.93 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0806, -0.0121, 0.1045, ..., 0.0324, -0.0811, -0.0796], + [-0.0586, -0.0634, -0.0067, ..., -0.0637, -0.0550, -0.0256], + [ 0.1066, -0.0253, -0.0743, ..., -0.0646, -0.0835, -0.1244], + ..., + [-0.0277, 0.0042, -0.0764, ..., -0.0185, 0.0341, 0.0061], + [-0.0783, -0.0625, -0.0095, ..., -0.0583, 0.0058, -0.0171], + [-0.0191, -0.0369, -0.0233, ..., 0.0657, 0.0259, -0.0846]], + device='cuda:0'), grad: tensor([[ 5.7258e-06, 2.8983e-06, -4.1723e-06, ..., -2.7064e-06, + 1.0252e-05, 3.4049e-06], + [ 2.6062e-05, 1.1511e-06, 3.7067e-07, ..., 1.4640e-06, + 5.2899e-06, 2.2445e-06], + [-1.0031e-04, 1.7080e-06, 1.5572e-06, ..., 2.4904e-06, + 4.2506e-06, 1.7397e-06], + ..., + [ 2.1651e-05, 1.8654e-06, 6.7009e-07, ..., 7.8678e-06, + 5.8822e-06, 2.4289e-06], + [ 9.4324e-06, -1.4174e-04, -3.1352e-05, ..., 2.0832e-05, + -3.7622e-04, -4.6551e-05], + [ 2.7474e-06, 2.4792e-06, 2.3991e-06, ..., -6.3777e-05, + -4.5866e-05, 3.5372e-06]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0061, 0.0148, 0.0158, -0.0018, 0.0185, -0.0060, -0.0154, -0.0097, + 0.0227, -0.0090], device='cuda:0'), grad: tensor([ 3.3945e-05, -7.3385e-04, -8.2403e-06, 2.2459e-04, 9.0301e-05, + 7.5674e-04, 1.1528e-04, 1.7035e-04, -5.9843e-04, -5.1320e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 251.68, cls_loss 0.0043 cls_loss_mapping 0.0119 cls_loss_causal 0.6433 re_mapping 0.0093 re_causal 0.0295 /// teacc 98.97 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0811, -0.0123, 0.1052, ..., 0.0325, -0.0816, -0.0798], + [-0.0591, -0.0639, -0.0065, ..., -0.0637, -0.0553, -0.0256], + [ 0.1072, -0.0258, -0.0751, ..., -0.0654, -0.0840, -0.1251], + ..., + [-0.0275, 0.0041, -0.0770, ..., -0.0189, 0.0340, 0.0058], + [-0.0795, -0.0627, -0.0098, ..., -0.0593, 0.0057, -0.0172], + [-0.0194, -0.0371, -0.0237, ..., 0.0665, 0.0262, -0.0853]], + device='cuda:0'), grad: tensor([[ 3.7625e-06, 2.9989e-07, 3.8236e-05, ..., 2.7148e-07, + 5.1796e-05, 8.4639e-05], + [ 5.7928e-06, 3.6322e-07, 4.2990e-06, ..., 1.7118e-06, + 1.4961e-04, 8.8289e-06], + [-2.1040e-05, 2.3656e-07, 5.5060e-06, ..., 1.6876e-06, + 1.5616e-05, 2.4870e-05], + ..., + [-2.6114e-06, 2.5285e-07, 2.0824e-06, ..., 2.7135e-05, + -1.5771e-04, 2.0731e-06], + [ 1.4929e-06, 5.4296e-07, 1.2182e-05, ..., 4.0710e-05, + 5.7220e-05, 1.0937e-05], + [ 2.9337e-06, 2.1374e-07, 2.7809e-06, ..., -1.1343e-04, + -8.8930e-05, 4.9025e-06]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0060, 0.0150, 0.0157, -0.0011, 0.0180, -0.0062, -0.0151, -0.0098, + 0.0221, -0.0086], device='cuda:0'), grad: tensor([ 1.5283e-04, 2.4128e-04, 7.5638e-05, 4.7415e-05, 2.6584e-04, + 9.5844e-05, -4.6945e-04, -3.2163e-04, 7.0691e-05, -1.5783e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 251.59, cls_loss 0.0040 cls_loss_mapping 0.0118 cls_loss_causal 0.6051 re_mapping 0.0097 re_causal 0.0292 /// teacc 99.05 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0817, -0.0123, 0.1055, ..., 0.0326, -0.0824, -0.0808], + [-0.0599, -0.0641, -0.0064, ..., -0.0639, -0.0559, -0.0263], + [ 0.1087, -0.0260, -0.0756, ..., -0.0657, -0.0842, -0.1247], + ..., + [-0.0281, 0.0040, -0.0772, ..., -0.0199, 0.0341, 0.0055], + [-0.0800, -0.0629, -0.0100, ..., -0.0601, 0.0055, -0.0173], + [-0.0197, -0.0373, -0.0242, ..., 0.0670, 0.0267, -0.0865]], + device='cuda:0'), grad: tensor([[ 1.9297e-06, 9.7789e-07, -1.2470e-06, ..., 7.6694e-07, + 5.7258e-06, 1.5730e-06], + [ 2.3264e-06, 8.8522e-07, -1.2770e-05, ..., -1.0535e-05, + 2.3842e-06, 1.5032e-06], + [ 7.6294e-06, 7.3574e-06, 6.5863e-06, ..., 2.7381e-06, + 1.9476e-05, 5.9754e-06], + ..., + [ 8.4192e-06, -1.8049e-06, 1.3523e-06, ..., 7.6056e-05, + 3.0324e-05, 1.1744e-06], + [ 3.4217e-06, 1.0021e-06, 1.5628e-06, ..., 5.6289e-06, + -5.7310e-05, -3.7193e-05], + [ 2.7139e-06, 2.4252e-06, 2.1104e-06, ..., -9.6500e-05, + -5.5224e-05, 1.3383e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0057, 0.0149, 0.0164, -0.0017, 0.0177, -0.0060, -0.0148, -0.0101, + 0.0219, -0.0082], device='cuda:0'), grad: tensor([ 2.4244e-05, -6.4671e-05, 9.1493e-05, -8.7202e-05, 3.7998e-05, + 2.4259e-04, 2.0072e-05, 1.3137e-04, -2.3413e-04, -1.6165e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 251.50, cls_loss 0.0039 cls_loss_mapping 0.0123 cls_loss_causal 0.6096 re_mapping 0.0094 re_causal 0.0299 /// teacc 98.95 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0820, -0.0123, 0.1061, ..., 0.0331, -0.0828, -0.0809], + [-0.0600, -0.0642, -0.0063, ..., -0.0638, -0.0562, -0.0275], + [ 0.1092, -0.0262, -0.0762, ..., -0.0660, -0.0846, -0.1243], + ..., + [-0.0283, 0.0038, -0.0775, ..., -0.0201, 0.0344, 0.0053], + [-0.0806, -0.0633, -0.0101, ..., -0.0607, 0.0054, -0.0171], + [-0.0203, -0.0374, -0.0246, ..., 0.0670, 0.0267, -0.0872]], + device='cuda:0'), grad: tensor([[ 5.5693e-06, 3.1758e-07, -3.1404e-06, ..., 1.4615e-04, + 3.1322e-05, 6.9812e-06], + [ 3.8743e-06, 6.1560e-07, -4.5784e-06, ..., 2.9225e-06, + 2.9534e-05, 4.5419e-05], + [-9.1255e-05, 1.6177e-06, 1.0587e-05, ..., 1.1265e-05, + 1.1660e-05, 1.3903e-05], + ..., + [ 4.4018e-05, -1.4305e-06, 1.4249e-06, ..., 1.5378e-05, + -1.5646e-05, 2.1644e-06], + [ 6.2510e-06, -2.5099e-07, 9.8124e-06, ..., 8.9481e-06, + -3.2067e-05, -6.0529e-05], + [ 6.9439e-06, 9.4296e-07, 4.8615e-06, ..., -2.2268e-04, + -2.9534e-05, 3.2149e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0060, 0.0149, 0.0165, -0.0013, 0.0179, -0.0061, -0.0150, -0.0101, + 0.0218, -0.0086], device='cuda:0'), grad: tensor([ 3.0756e-04, 5.3835e-04, -5.8442e-05, 4.7714e-05, 6.4373e-05, + 1.6260e-04, -3.1024e-05, 7.3373e-05, -7.4196e-04, -3.6144e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 251.60, cls_loss 0.0042 cls_loss_mapping 0.0107 cls_loss_causal 0.5900 re_mapping 0.0092 re_causal 0.0281 /// teacc 98.89 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0825, -0.0124, 0.1066, ..., 0.0334, -0.0836, -0.0816], + [-0.0599, -0.0645, -0.0061, ..., -0.0640, -0.0566, -0.0283], + [ 0.1099, -0.0264, -0.0768, ..., -0.0668, -0.0853, -0.1247], + ..., + [-0.0287, 0.0037, -0.0779, ..., -0.0201, 0.0348, 0.0052], + [-0.0815, -0.0635, -0.0102, ..., -0.0612, 0.0052, -0.0170], + [-0.0206, -0.0377, -0.0252, ..., 0.0672, 0.0269, -0.0879]], + device='cuda:0'), grad: tensor([[ 5.6764e-07, 5.4762e-07, -2.1458e-05, ..., -2.2873e-05, + 2.0210e-06, 1.1073e-06], + [ 2.4866e-06, 1.9502e-06, 1.0040e-06, ..., 1.4305e-06, + 1.7118e-06, -4.6231e-06], + [ 2.8927e-06, 4.0196e-06, 3.0212e-06, ..., 3.7123e-06, + 3.1069e-06, 1.7891e-06], + ..., + [ 4.8131e-06, 3.8259e-06, 1.3970e-06, ..., 7.5139e-06, + 7.5214e-06, 7.1488e-06], + [ 3.5055e-06, 2.1718e-06, 1.6270e-06, ..., 3.1609e-06, + 2.7139e-06, 1.4203e-06], + [ 6.2101e-06, 5.0738e-06, 9.5144e-06, ..., 2.8666e-06, + -4.7218e-07, 2.7716e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0058, 0.0159, 0.0162, -0.0015, 0.0177, -0.0060, -0.0147, -0.0103, + 0.0216, -0.0087], device='cuda:0'), grad: tensor([-3.0547e-05, -8.6367e-05, 2.0415e-05, -5.3763e-05, -9.2462e-06, + 2.2098e-05, -1.3318e-06, 9.7871e-05, 1.4901e-05, 2.5988e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 251.42, cls_loss 0.0055 cls_loss_mapping 0.0136 cls_loss_causal 0.6052 re_mapping 0.0091 re_causal 0.0282 /// teacc 98.90 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0832, -0.0124, 0.1074, ..., 0.0337, -0.0845, -0.0823], + [-0.0605, -0.0643, -0.0057, ..., -0.0642, -0.0574, -0.0287], + [ 0.1105, -0.0267, -0.0774, ..., -0.0673, -0.0861, -0.1251], + ..., + [-0.0286, 0.0036, -0.0784, ..., -0.0195, 0.0358, 0.0047], + [-0.0824, -0.0637, -0.0103, ..., -0.0619, 0.0054, -0.0166], + [-0.0210, -0.0378, -0.0259, ..., 0.0673, 0.0266, -0.0897]], + device='cuda:0'), grad: tensor([[ 7.8082e-06, 4.6333e-07, -3.7923e-06, ..., 1.0771e-04, + 3.3945e-05, 2.0526e-06], + [ 3.7067e-06, 1.1818e-06, -2.1327e-06, ..., 1.7602e-06, + 3.7905e-06, -2.8824e-07], + [-3.7134e-05, 2.7083e-06, 1.1427e-06, ..., 3.4347e-06, + 2.4531e-06, 3.8929e-06], + ..., + [ 1.9625e-05, 2.4475e-06, 4.5588e-07, ..., 9.1866e-06, + 7.1675e-06, 2.7753e-06], + [ 1.0766e-05, 2.1737e-06, 2.1402e-06, ..., 7.7337e-06, + 3.7774e-06, -2.6096e-06], + [ 7.6257e-06, 2.9542e-06, 1.5898e-06, ..., -1.3900e-04, + -3.7819e-05, 4.2170e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0061, 0.0151, 0.0159, -0.0016, 0.0175, -0.0059, -0.0146, -0.0088, + 0.0219, -0.0095], device='cuda:0'), grad: tensor([ 2.1088e-04, -3.0056e-05, -2.9296e-05, 1.1331e-04, 2.4125e-05, + -2.1005e-04, 3.0324e-05, 7.9155e-05, 2.9519e-05, -2.1756e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 251.12, cls_loss 0.0055 cls_loss_mapping 0.0139 cls_loss_causal 0.6210 re_mapping 0.0094 re_causal 0.0279 /// teacc 98.99 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0835, -0.0125, 0.1074, ..., 0.0337, -0.0853, -0.0831], + [-0.0616, -0.0645, -0.0060, ..., -0.0647, -0.0581, -0.0293], + [ 0.1113, -0.0269, -0.0783, ..., -0.0679, -0.0868, -0.1254], + ..., + [-0.0287, 0.0033, -0.0788, ..., -0.0199, 0.0362, 0.0041], + [-0.0827, -0.0640, -0.0096, ..., -0.0613, 0.0049, -0.0169], + [-0.0212, -0.0380, -0.0262, ..., 0.0675, 0.0266, -0.0902]], + device='cuda:0'), grad: tensor([[ 2.7306e-06, 5.6950e-07, -8.6129e-06, ..., -4.1611e-06, + 4.4815e-06, 1.9614e-06], + [ 1.3001e-06, 4.4517e-07, -6.8918e-08, ..., 4.4797e-07, + 1.1884e-05, 6.2864e-07], + [-2.5686e-06, 8.0001e-07, 1.0114e-06, ..., 3.7672e-07, + 7.9945e-06, 2.6431e-06], + ..., + [ 4.2915e-06, 1.2377e-06, 2.4121e-07, ..., 8.4117e-06, + -6.5207e-05, 1.9046e-06], + [ 2.5615e-05, 5.5544e-06, 1.6270e-06, ..., 2.6617e-06, + 6.5804e-05, 2.2888e-05], + [ 7.2867e-06, 2.1756e-06, 1.5255e-06, ..., -1.7732e-05, + 1.3947e-05, 6.2883e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0055, 0.0151, 0.0160, -0.0028, 0.0178, -0.0050, -0.0142, -0.0091, + 0.0221, -0.0095], device='cuda:0'), grad: tensor([ 2.0023e-06, 1.4871e-05, 1.9923e-05, 6.4659e-03, 1.7151e-05, + -6.5765e-03, 3.7309e-06, -1.5354e-04, 1.6153e-04, 4.3839e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 250.82, cls_loss 0.0042 cls_loss_mapping 0.0121 cls_loss_causal 0.6225 re_mapping 0.0087 re_causal 0.0275 /// teacc 99.00 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0841, -0.0125, 0.1080, ..., 0.0341, -0.0859, -0.0833], + [-0.0616, -0.0647, -0.0062, ..., -0.0648, -0.0587, -0.0300], + [ 0.1121, -0.0271, -0.0786, ..., -0.0682, -0.0878, -0.1266], + ..., + [-0.0293, 0.0031, -0.0791, ..., -0.0204, 0.0365, 0.0050], + [-0.0834, -0.0643, -0.0098, ..., -0.0622, 0.0049, -0.0167], + [-0.0215, -0.0382, -0.0267, ..., 0.0683, 0.0271, -0.0909]], + device='cuda:0'), grad: tensor([[ 2.3827e-05, 3.5437e-07, -1.1124e-05, ..., -6.1989e-06, + 7.4953e-06, -1.9446e-06], + [ 1.0528e-05, 2.1309e-06, 1.3104e-06, ..., 2.6058e-06, + 4.4331e-06, -8.2981e-07], + [-5.1880e-04, -1.4566e-05, 1.2137e-05, ..., 6.9551e-06, + 1.2249e-05, 2.4494e-06], + ..., + [ 3.7169e-04, 4.0606e-06, 7.2410e-07, ..., 4.8518e-05, + 1.6754e-06, 1.0310e-06], + [ 8.1360e-05, 1.4968e-05, 8.3521e-06, ..., 5.3458e-06, + -5.8934e-06, 3.1609e-06], + [ 3.3509e-06, 3.1926e-06, 2.4941e-06, ..., 1.1109e-05, + 1.5661e-05, 1.2862e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0058, 0.0151, 0.0160, -0.0035, 0.0172, -0.0045, -0.0139, -0.0094, + 0.0222, -0.0092], device='cuda:0'), grad: tensor([ 3.6836e-05, -7.7114e-06, -6.7711e-04, -1.0088e-05, -1.1081e-04, + 5.1051e-05, -3.9995e-05, 5.9748e-04, 9.4712e-05, 6.6042e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 251.11, cls_loss 0.0043 cls_loss_mapping 0.0115 cls_loss_causal 0.6136 re_mapping 0.0084 re_causal 0.0269 /// teacc 98.93 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0852, -0.0127, 0.1081, ..., 0.0342, -0.0865, -0.0840], + [-0.0620, -0.0650, -0.0060, ..., -0.0641, -0.0589, -0.0319], + [ 0.1132, -0.0272, -0.0789, ..., -0.0687, -0.0883, -0.1270], + ..., + [-0.0300, 0.0029, -0.0794, ..., -0.0214, 0.0363, 0.0048], + [-0.0840, -0.0648, -0.0104, ..., -0.0637, 0.0043, -0.0171], + [-0.0214, -0.0383, -0.0270, ..., 0.0687, 0.0271, -0.0919]], + device='cuda:0'), grad: tensor([[-7.3649e-06, 6.5193e-08, -1.6022e-04, ..., -6.0529e-05, + 2.3078e-06, -2.0638e-05], + [ 4.6939e-06, 3.6694e-07, 3.7886e-06, ..., -1.2279e-05, + 6.4850e-05, -3.7968e-05], + [ 4.1015e-06, 7.5717e-07, 8.0541e-06, ..., 7.8678e-06, + 1.9699e-05, 5.4911e-06], + ..., + [-1.9237e-05, 4.0978e-07, 2.5071e-06, ..., -2.8927e-06, + -1.1933e-04, 3.1237e-06], + [ 1.3355e-06, 1.3597e-07, 1.1586e-05, ..., 1.4752e-05, + -3.6173e-06, 7.0632e-06], + [ 7.2680e-06, 6.8545e-07, 2.0668e-05, ..., 1.2062e-05, + 1.8030e-05, 8.6203e-06]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0055, 0.0154, 0.0164, -0.0033, 0.0174, -0.0046, -0.0127, -0.0103, + 0.0213, -0.0090], device='cuda:0'), grad: tensor([-1.3852e-04, -2.8834e-05, 1.3077e-04, 1.1426e-04, 4.1872e-05, + 8.0705e-05, 8.8990e-05, -5.2929e-04, 9.4771e-05, 1.4532e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 251.11, cls_loss 0.0055 cls_loss_mapping 0.0140 cls_loss_causal 0.6175 re_mapping 0.0086 re_causal 0.0274 /// teacc 98.94 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0859, -0.0128, 0.1087, ..., 0.0344, -0.0869, -0.0843], + [-0.0611, -0.0651, -0.0056, ..., -0.0652, -0.0600, -0.0314], + [ 0.1135, -0.0274, -0.0793, ..., -0.0682, -0.0888, -0.1280], + ..., + [-0.0308, 0.0027, -0.0797, ..., -0.0217, 0.0370, 0.0051], + [-0.0844, -0.0652, -0.0106, ..., -0.0642, 0.0042, -0.0170], + [-0.0219, -0.0385, -0.0274, ..., 0.0699, 0.0272, -0.0930]], + device='cuda:0'), grad: tensor([[ 3.8967e-06, 6.5677e-06, -2.4855e-05, ..., -1.0118e-05, + 5.3644e-06, 9.0152e-06], + [ 3.5353e-06, 5.2340e-06, -3.5793e-05, ..., -2.6718e-05, + 4.9658e-06, -2.7686e-05], + [ 2.4103e-06, 2.6878e-06, 7.9572e-06, ..., 4.5709e-06, + 9.7305e-06, 5.7034e-06], + ..., + [-1.3737e-08, 2.1700e-06, 4.1462e-06, ..., 6.9216e-06, + -5.8189e-06, 1.8151e-06], + [ 6.5789e-06, 1.0870e-05, 1.8880e-05, ..., 8.5011e-06, + 5.8599e-06, 9.7305e-06], + [ 1.1578e-05, 1.8582e-05, 2.4199e-05, ..., 6.5446e-05, + 4.0859e-05, 1.2174e-05]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0055, 0.0158, 0.0160, -0.0023, 0.0166, -0.0052, -0.0130, -0.0099, + 0.0212, -0.0086], device='cuda:0'), grad: tensor([ 4.0010e-06, -1.1671e-04, 5.1856e-05, -5.2404e-04, -7.6294e-05, + 3.3212e-04, 6.9618e-05, -2.1420e-06, 7.6711e-05, 1.8513e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 79---------------------------------------------------- +epoch 79, time 268.73, cls_loss 0.0046 cls_loss_mapping 0.0108 cls_loss_causal 0.6410 re_mapping 0.0084 re_causal 0.0272 /// teacc 99.08 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0859, -0.0130, 0.1092, ..., 0.0338, -0.0876, -0.0848], + [-0.0621, -0.0654, -0.0056, ..., -0.0655, -0.0608, -0.0315], + [ 0.1142, -0.0276, -0.0798, ..., -0.0691, -0.0893, -0.1285], + ..., + [-0.0301, 0.0024, -0.0801, ..., -0.0220, 0.0375, 0.0046], + [-0.0850, -0.0653, -0.0103, ..., -0.0649, 0.0043, -0.0167], + [-0.0234, -0.0390, -0.0274, ..., 0.0701, 0.0268, -0.0945]], + device='cuda:0'), grad: tensor([[ 3.5875e-06, 6.5798e-07, -3.1024e-05, ..., -1.2882e-05, + 8.9705e-06, 5.0198e-07], + [-1.0651e-04, 1.1092e-06, 4.3842e-07, ..., 3.1382e-05, + -1.8513e-04, 2.5192e-07], + [ 3.3677e-05, 7.0147e-06, 1.3625e-06, ..., 1.3545e-05, + 8.6546e-05, 4.0862e-07], + ..., + [ 1.4551e-05, 1.3970e-06, 5.2992e-07, ..., 9.1195e-05, + 7.6771e-05, 1.9418e-07], + [ 8.3089e-05, 2.2307e-05, 3.6061e-06, ..., 9.6142e-05, + 1.7011e-04, 1.1669e-06], + [ 4.6045e-06, 2.4121e-06, 1.7360e-05, ..., -2.9731e-04, + -2.2590e-04, 6.0536e-07]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0049, 0.0152, 0.0159, -0.0022, 0.0174, -0.0052, -0.0134, -0.0093, + 0.0215, -0.0089], device='cuda:0'), grad: tensor([ 3.2149e-06, -1.3905e-03, 5.5218e-04, -5.2243e-05, 1.1981e-04, + 5.9068e-05, 1.7059e-04, 3.2783e-04, 9.3603e-04, -7.2622e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 251.49, cls_loss 0.0039 cls_loss_mapping 0.0101 cls_loss_causal 0.6096 re_mapping 0.0089 re_causal 0.0266 /// teacc 99.06 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0862, -0.0131, 0.1097, ..., 0.0341, -0.0883, -0.0851], + [-0.0630, -0.0657, -0.0051, ..., -0.0650, -0.0613, -0.0319], + [ 0.1153, -0.0280, -0.0804, ..., -0.0693, -0.0898, -0.1288], + ..., + [-0.0299, 0.0019, -0.0804, ..., -0.0222, 0.0379, 0.0044], + [-0.0853, -0.0655, -0.0105, ..., -0.0654, 0.0042, -0.0166], + [-0.0250, -0.0394, -0.0280, ..., 0.0703, 0.0269, -0.0953]], + device='cuda:0'), grad: tensor([[ 3.8669e-06, 1.5404e-06, 4.8801e-07, ..., 1.2303e-06, + 2.9728e-06, 1.9027e-06], + [ 1.2323e-05, 4.6454e-06, 1.4063e-07, ..., -3.6787e-06, + 4.0904e-06, 3.0752e-06], + [ 2.7016e-05, 1.2167e-05, 7.7020e-07, ..., 1.4743e-06, + 2.1122e-06, 1.3104e-06], + ..., + [ 1.4059e-05, 5.1931e-06, 6.1654e-07, ..., 2.6952e-06, + 1.6270e-06, 1.1139e-06], + [ 5.0403e-06, 2.2706e-06, 5.5581e-06, ..., 1.9819e-06, + 1.4231e-05, 1.4223e-05], + [ 8.3148e-06, 3.3304e-06, 1.9297e-06, ..., 3.1758e-06, + 3.2820e-06, 2.0117e-06]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0049, 0.0150, 0.0160, -0.0019, 0.0170, -0.0055, -0.0131, -0.0087, + 0.0216, -0.0093], device='cuda:0'), grad: tensor([ 1.4551e-05, -9.2089e-06, 6.8069e-05, -1.9276e-04, -7.9453e-05, + 1.0294e-04, -4.1090e-06, 4.8220e-05, 1.9923e-05, 3.1710e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 251.73, cls_loss 0.0041 cls_loss_mapping 0.0121 cls_loss_causal 0.5936 re_mapping 0.0082 re_causal 0.0254 /// teacc 99.05 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0868, -0.0132, 0.1101, ..., 0.0344, -0.0892, -0.0860], + [-0.0633, -0.0660, -0.0052, ..., -0.0652, -0.0616, -0.0321], + [ 0.1159, -0.0282, -0.0810, ..., -0.0696, -0.0903, -0.1297], + ..., + [-0.0303, 0.0015, -0.0808, ..., -0.0224, 0.0381, 0.0042], + [-0.0856, -0.0658, -0.0108, ..., -0.0666, 0.0035, -0.0168], + [-0.0251, -0.0395, -0.0281, ..., 0.0705, 0.0270, -0.0959]], + device='cuda:0'), grad: tensor([[ 7.4878e-06, 3.3062e-08, -5.2862e-06, ..., -2.3283e-06, + 2.6971e-06, 5.7183e-07], + [ 4.2021e-05, 8.4750e-08, 4.3167e-07, ..., 4.1984e-06, + 2.7269e-05, 3.1590e-06], + [-8.7357e-04, 1.0757e-07, -4.7646e-06, ..., 1.2740e-06, + 9.4026e-06, -6.7532e-05], + ..., + [ 3.5971e-05, 1.9558e-07, 1.9409e-06, ..., 7.2047e-06, + -1.1218e-04, 7.8231e-07], + [ 7.5674e-04, 3.6601e-07, 2.5779e-06, ..., 1.4186e-05, + 2.0891e-05, 6.3002e-05], + [ 4.7348e-06, 1.9744e-07, 1.8617e-06, ..., -1.0687e-04, + 4.1351e-06, 1.5534e-06]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0049, 0.0152, 0.0159, -0.0010, 0.0167, -0.0053, -0.0134, -0.0087, + 0.0209, -0.0093], device='cuda:0'), grad: tensor([ 1.7315e-05, 1.7273e-04, -1.6432e-03, 1.1557e-04, 7.9989e-05, + 2.9221e-05, 1.6317e-05, -3.2687e-04, 1.5306e-03, 6.9179e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 251.31, cls_loss 0.0047 cls_loss_mapping 0.0118 cls_loss_causal 0.6014 re_mapping 0.0083 re_causal 0.0259 /// teacc 99.00 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0875, -0.0134, 0.1107, ..., 0.0347, -0.0902, -0.0866], + [-0.0637, -0.0662, -0.0057, ..., -0.0653, -0.0620, -0.0321], + [ 0.1168, -0.0283, -0.0820, ..., -0.0708, -0.0907, -0.1307], + ..., + [-0.0308, 0.0015, -0.0810, ..., -0.0231, 0.0382, 0.0039], + [-0.0862, -0.0662, -0.0107, ..., -0.0673, 0.0037, -0.0165], + [-0.0252, -0.0396, -0.0283, ..., 0.0697, 0.0268, -0.0966]], + device='cuda:0'), grad: tensor([[ 5.5544e-06, -4.0699e-07, 1.4102e-04, ..., -2.2054e-05, + 1.1569e-04, 1.1331e-04], + [-1.4842e-05, 1.5683e-06, 8.4788e-06, ..., 1.1079e-05, + 1.4424e-05, 9.6634e-06], + [-1.3018e-04, 1.2899e-06, 2.0728e-05, ..., -7.8082e-05, + 1.8492e-05, 1.5065e-05], + ..., + [ 7.9989e-05, 1.7332e-06, 5.5209e-06, ..., 7.9274e-05, + 8.9332e-06, 3.0492e-06], + [-4.6864e-06, 4.0010e-06, 2.3052e-05, ..., 1.1764e-05, + -2.5153e-05, -1.8418e-05], + [ 2.9400e-05, 2.6245e-06, 1.2226e-05, ..., 4.8250e-05, + 1.1221e-05, 9.2685e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0047, 0.0150, 0.0160, -0.0013, 0.0186, -0.0051, -0.0139, -0.0087, + 0.0213, -0.0106], device='cuda:0'), grad: tensor([ 3.5691e-04, -1.0121e-04, -8.8632e-05, 5.1928e-04, 2.0182e-04, + -4.1032e-04, -7.6389e-04, 2.2554e-04, -7.6771e-05, 1.3757e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 251.59, cls_loss 0.0043 cls_loss_mapping 0.0096 cls_loss_causal 0.5985 re_mapping 0.0082 re_causal 0.0255 /// teacc 99.06 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0883, -0.0136, 0.1109, ..., 0.0350, -0.0913, -0.0878], + [-0.0644, -0.0666, -0.0055, ..., -0.0646, -0.0624, -0.0327], + [ 0.1172, -0.0285, -0.0828, ..., -0.0712, -0.0912, -0.1313], + ..., + [-0.0304, 0.0011, -0.0815, ..., -0.0237, 0.0384, 0.0039], + [-0.0866, -0.0663, -0.0106, ..., -0.0677, 0.0037, -0.0164], + [-0.0254, -0.0399, -0.0286, ..., 0.0700, 0.0268, -0.0971]], + device='cuda:0'), grad: tensor([[ 1.1418e-06, -1.0030e-06, -4.1515e-05, ..., -2.3887e-05, + 1.4352e-06, 2.0601e-06], + [ 1.1109e-05, 2.5332e-07, 1.3169e-06, ..., -6.4522e-06, + 1.4324e-06, 1.7723e-06], + [-5.0336e-05, 3.3807e-07, 8.5756e-06, ..., 5.4017e-06, + 1.4687e-06, 1.7649e-06], + ..., + [ 1.3821e-05, 4.2934e-07, 3.9339e-06, ..., 1.0878e-05, + 1.5553e-07, 1.8720e-07], + [ 1.4193e-05, 4.4983e-07, 8.3447e-06, ..., 6.2659e-06, + 4.0159e-06, 7.4282e-06], + [ 1.1194e-06, 5.6624e-07, 6.5118e-06, ..., -8.9779e-06, + -5.1782e-06, 7.4180e-07]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0043, 0.0141, 0.0157, -0.0014, 0.0184, -0.0052, -0.0133, -0.0075, + 0.0215, -0.0107], device='cuda:0'), grad: tensor([-4.1664e-05, -5.8025e-05, -1.2958e-04, 1.6332e-05, 8.6725e-06, + 7.2382e-06, 8.6725e-06, 9.3997e-05, 7.5996e-05, 1.8299e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 251.42, cls_loss 0.0031 cls_loss_mapping 0.0107 cls_loss_causal 0.6105 re_mapping 0.0083 re_causal 0.0258 /// teacc 98.98 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0886, -0.0137, 0.1111, ..., 0.0352, -0.0919, -0.0890], + [-0.0652, -0.0667, -0.0058, ..., -0.0649, -0.0630, -0.0328], + [ 0.1183, -0.0289, -0.0827, ..., -0.0715, -0.0916, -0.1321], + ..., + [-0.0311, 0.0006, -0.0821, ..., -0.0240, 0.0387, 0.0036], + [-0.0869, -0.0666, -0.0109, ..., -0.0682, 0.0038, -0.0164], + [-0.0256, -0.0400, -0.0289, ..., 0.0703, 0.0268, -0.0981]], + device='cuda:0'), grad: tensor([[ 2.9840e-06, 1.0151e-06, -2.2337e-05, ..., -7.1041e-06, + 1.5087e-07, -1.9539e-06], + [ 1.7779e-06, 6.4913e-07, 1.6689e-06, ..., 7.8231e-06, + 2.5213e-05, 8.3297e-06], + [ 6.3255e-06, 2.5388e-06, 3.3882e-06, ..., 5.8440e-07, + 1.4668e-06, 2.7893e-07], + ..., + [ 3.4198e-06, 1.2061e-06, 1.1884e-06, ..., -2.5257e-05, + -8.4817e-05, -2.0444e-05], + [ 4.4052e-07, 1.7509e-07, -1.1817e-05, ..., 9.3924e-07, + -4.1686e-06, -7.3574e-06], + [ 5.9977e-07, 2.7055e-07, 8.0839e-06, ..., 1.6108e-05, + 5.1886e-05, 1.4476e-05]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0042, 0.0147, 0.0164, -0.0010, 0.0183, -0.0054, -0.0130, -0.0087, + 0.0213, -0.0107], device='cuda:0'), grad: tensor([-9.8422e-06, 6.3539e-05, 2.7373e-05, -1.6630e-05, 2.4773e-06, + -7.0110e-06, 3.1590e-05, -1.7965e-04, -5.2810e-05, 1.4067e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 251.53, cls_loss 0.0043 cls_loss_mapping 0.0127 cls_loss_causal 0.6067 re_mapping 0.0082 re_causal 0.0260 /// teacc 99.00 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0889, -0.0137, 0.1128, ..., 0.0344, -0.0922, -0.0891], + [-0.0654, -0.0669, -0.0079, ..., -0.0656, -0.0638, -0.0340], + [ 0.1188, -0.0291, -0.0839, ..., -0.0722, -0.0919, -0.1326], + ..., + [-0.0315, 0.0005, -0.0830, ..., -0.0242, 0.0392, 0.0038], + [-0.0871, -0.0668, -0.0109, ..., -0.0691, 0.0038, -0.0162], + [-0.0259, -0.0402, -0.0289, ..., 0.0715, 0.0267, -0.0992]], + device='cuda:0'), grad: tensor([[ 1.8766e-07, 3.6787e-08, -4.5188e-06, ..., -2.5816e-06, + 5.7034e-06, 1.3746e-06], + [ 1.4026e-06, 9.5461e-08, -2.3656e-07, ..., 2.6021e-06, + 1.4305e-05, 2.4494e-07], + [-1.1642e-08, 6.7754e-07, 9.8124e-06, ..., 2.7623e-06, + 1.3456e-05, 1.2657e-06], + ..., + [ 2.4755e-06, 2.0023e-07, 7.9116e-07, ..., -4.5240e-05, + -1.9741e-04, 1.5832e-07], + [-1.2312e-06, 2.0536e-07, -1.2897e-05, ..., 2.8256e-06, + -3.3975e-05, -1.9576e-06], + [ 7.8650e-07, 3.7812e-07, 2.3637e-06, ..., 1.6600e-05, + 9.8407e-05, 6.1188e-07]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0042, 0.0141, 0.0162, -0.0010, 0.0180, -0.0054, -0.0134, -0.0086, + 0.0220, -0.0100], device='cuda:0'), grad: tensor([ 8.4639e-06, 2.1487e-05, 8.2374e-05, -7.3127e-06, 1.8787e-04, + 6.7234e-05, 1.1079e-05, -4.9162e-04, -1.4377e-04, 2.6393e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 251.50, cls_loss 0.0052 cls_loss_mapping 0.0129 cls_loss_causal 0.5978 re_mapping 0.0081 re_causal 0.0245 /// teacc 98.83 lr 0.00010000 +Epoch 88, weight, value: tensor([[-9.0728e-02, -1.3838e-02, 1.1302e-01, ..., 3.4492e-02, + -9.2871e-02, -8.9492e-02], + [-6.4390e-02, -6.7070e-02, -7.1980e-03, ..., -6.6131e-02, + -6.4682e-02, -3.3944e-02], + [ 1.1879e-01, -2.9280e-02, -8.4015e-02, ..., -7.2789e-02, + -9.2746e-02, -1.3326e-01], + ..., + [-3.1956e-02, 8.8850e-05, -8.3897e-02, ..., -2.3905e-02, + 3.9172e-02, 3.4793e-03], + [-8.7722e-02, -6.6979e-02, -1.0803e-02, ..., -6.9290e-02, + 3.7055e-03, -1.6317e-02], + [-2.6299e-02, -4.0429e-02, -2.9215e-02, ..., 7.2029e-02, + 2.8190e-02, -9.9485e-02]], device='cuda:0'), grad: tensor([[ 7.9162e-07, 8.0653e-07, 1.1122e-04, ..., 9.9838e-06, + 8.4341e-05, 8.7738e-05], + [ 2.2780e-06, 2.2575e-06, 7.4804e-06, ..., 3.4496e-06, + 1.2428e-05, 7.9945e-06], + [ 5.2415e-06, 5.0217e-06, 4.9807e-06, ..., 1.8086e-06, + 1.0796e-05, 3.4962e-06], + ..., + [ 9.0823e-06, 7.9423e-06, 2.4345e-06, ..., 1.0625e-05, + -3.1173e-05, 1.8682e-06], + [ 3.1684e-06, 3.3192e-06, 8.1509e-06, ..., 1.1235e-05, + 8.6054e-06, 1.4259e-06], + [ 1.4873e-06, 1.5153e-06, 8.5905e-06, ..., -7.4863e-05, + -4.3750e-05, 2.2016e-06]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0037, 0.0153, 0.0146, -0.0011, 0.0173, -0.0057, -0.0133, -0.0086, + 0.0222, -0.0092], device='cuda:0'), grad: tensor([ 1.9717e-04, 2.7746e-05, 5.8681e-05, -5.2124e-05, 1.8680e-04, + 8.8394e-05, -3.4356e-04, -8.3089e-05, 2.8908e-05, -1.0902e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 251.68, cls_loss 0.0027 cls_loss_mapping 0.0090 cls_loss_causal 0.5971 re_mapping 0.0083 re_causal 0.0255 /// teacc 98.95 lr 0.00010000 +Epoch 89, weight, value: tensor([[-9.1141e-02, -1.3863e-02, 1.1343e-01, ..., 3.4667e-02, + -9.3379e-02, -8.9823e-02], + [-6.4809e-02, -6.7402e-02, -7.2452e-03, ..., -6.6407e-02, + -6.5195e-02, -3.4276e-02], + [ 1.1990e-01, -2.9514e-02, -8.4112e-02, ..., -7.2821e-02, + -9.3190e-02, -1.3319e-01], + ..., + [-3.2554e-02, -1.0247e-04, -8.4402e-02, ..., -2.4411e-02, + 3.9205e-02, 3.0725e-03], + [-8.8292e-02, -6.7388e-02, -1.0793e-02, ..., -6.9799e-02, + 3.6850e-03, -1.6147e-02], + [-2.6505e-02, -4.0651e-02, -2.9509e-02, ..., 7.2278e-02, + 2.8510e-02, -9.9802e-02]], device='cuda:0'), grad: tensor([[ 4.0196e-06, 7.7020e-07, 8.2096e-07, ..., 1.5944e-06, + 2.8126e-06, 1.2498e-06], + [ 1.2154e-06, 3.2363e-07, 2.1188e-07, ..., 7.4282e-06, + 9.1270e-06, 2.5835e-06], + [-7.8917e-05, -6.1244e-06, 3.8138e-07, ..., 3.5167e-06, + 3.9749e-06, 7.9675e-07], + ..., + [ 6.6571e-06, 6.3190e-07, 3.3528e-08, ..., 5.3011e-06, + -4.5449e-06, 1.6550e-06], + [ 2.1100e-05, 3.8110e-06, 4.0643e-06, ..., 6.7316e-06, + 1.2688e-05, 6.3628e-06], + [ 4.1304e-07, 8.8010e-07, 1.6019e-07, ..., 6.1095e-05, + 6.4671e-05, 2.1726e-05]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0038, 0.0151, 0.0153, -0.0002, 0.0172, -0.0066, -0.0134, -0.0090, + 0.0225, -0.0090], device='cuda:0'), grad: tensor([ 1.6078e-05, 2.0847e-05, -9.9838e-05, 9.9957e-05, -4.9400e-04, + 6.8367e-05, 1.1748e-04, -1.3635e-05, 7.0751e-05, 2.1327e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 251.68, cls_loss 0.0032 cls_loss_mapping 0.0096 cls_loss_causal 0.6074 re_mapping 0.0077 re_causal 0.0249 /// teacc 98.98 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0919, -0.0140, 0.1140, ..., 0.0348, -0.0948, -0.0903], + [-0.0650, -0.0676, -0.0078, ..., -0.0666, -0.0658, -0.0345], + [ 0.1207, -0.0297, -0.0848, ..., -0.0729, -0.0939, -0.1329], + ..., + [-0.0328, -0.0002, -0.0850, ..., -0.0248, 0.0397, 0.0028], + [-0.0889, -0.0679, -0.0109, ..., -0.0705, 0.0037, -0.0163], + [-0.0253, -0.0408, -0.0284, ..., 0.0731, 0.0293, -0.1000]], + device='cuda:0'), grad: tensor([[ 1.7174e-06, 4.2189e-07, -2.7448e-05, ..., 1.0371e-05, + 6.4000e-06, 3.2391e-06], + [ 8.6986e-07, 1.1558e-06, 1.2973e-06, ..., 4.1500e-06, + 3.3062e-06, 2.5053e-06], + [ 6.0797e-06, 7.2643e-07, 1.2569e-05, ..., 6.5982e-05, + 1.4611e-05, 1.1278e-06], + ..., + [ 2.2184e-06, 6.7577e-06, 3.3900e-06, ..., 1.7077e-05, + 1.8194e-05, 4.6864e-06], + [ 1.3057e-06, 8.6203e-06, 5.1707e-06, ..., 1.2219e-05, + 1.6063e-05, 2.4751e-05], + [ 3.7216e-06, 4.7907e-06, 1.4044e-05, ..., 3.1918e-05, + 1.1772e-05, 1.0327e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0036, 0.0149, 0.0154, -0.0002, 0.0170, -0.0081, -0.0136, -0.0088, + 0.0226, -0.0079], device='cuda:0'), grad: tensor([ 1.2785e-05, 1.7151e-05, 1.4961e-04, -1.8549e-04, -3.7336e-04, + -1.1957e-04, 1.4031e-04, 1.0121e-04, 1.2469e-04, 1.3280e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 251.30, cls_loss 0.0032 cls_loss_mapping 0.0088 cls_loss_causal 0.5751 re_mapping 0.0077 re_causal 0.0243 /// teacc 98.95 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0927, -0.0140, 0.1147, ..., 0.0353, -0.0966, -0.0915], + [-0.0653, -0.0679, -0.0072, ..., -0.0662, -0.0665, -0.0347], + [ 0.1219, -0.0300, -0.0859, ..., -0.0732, -0.0943, -0.1335], + ..., + [-0.0336, -0.0006, -0.0857, ..., -0.0253, 0.0394, 0.0025], + [-0.0892, -0.0684, -0.0112, ..., -0.0713, 0.0033, -0.0165], + [-0.0254, -0.0412, -0.0291, ..., 0.0730, 0.0295, -0.1008]], + device='cuda:0'), grad: tensor([[ 9.0338e-08, 3.9116e-08, -5.1409e-06, ..., -5.0925e-06, + 1.5516e-06, 1.6792e-06], + [ 2.3935e-07, 2.3190e-07, 3.4999e-06, ..., 1.0310e-06, + 2.5220e-06, 2.4401e-06], + [-7.2271e-07, 4.9733e-07, 5.2191e-06, ..., 1.2424e-06, + 3.1199e-06, 3.2093e-06], + ..., + [ 3.1702e-06, 3.9004e-06, 1.0906e-06, ..., 1.4091e-06, + -3.5893e-06, 2.6133e-06], + [ 1.0217e-06, 5.6624e-07, 4.7311e-06, ..., 3.4124e-06, + 3.1516e-06, 2.5723e-06], + [ 4.5355e-07, 8.5402e-07, 1.6838e-06, ..., -4.1723e-07, + 3.5632e-06, 4.8801e-07]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0039, 0.0149, 0.0156, -0.0008, 0.0175, -0.0075, -0.0136, -0.0094, + 0.0222, -0.0077], device='cuda:0'), grad: tensor([-3.8594e-06, -6.2212e-06, 1.6525e-05, -1.6794e-05, -1.7677e-06, + 8.1211e-06, -3.7432e-05, 6.6981e-06, 1.9699e-05, 1.4953e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 251.54, cls_loss 0.0045 cls_loss_mapping 0.0116 cls_loss_causal 0.5618 re_mapping 0.0080 re_causal 0.0238 /// teacc 98.99 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0931, -0.0141, 0.1153, ..., 0.0355, -0.0971, -0.0918], + [-0.0658, -0.0681, -0.0073, ..., -0.0677, -0.0680, -0.0350], + [ 0.1230, -0.0301, -0.0866, ..., -0.0735, -0.0948, -0.1341], + ..., + [-0.0341, -0.0008, -0.0863, ..., -0.0249, 0.0399, 0.0025], + [-0.0894, -0.0687, -0.0117, ..., -0.0719, 0.0030, -0.0167], + [-0.0262, -0.0415, -0.0295, ..., 0.0723, 0.0289, -0.1013]], + device='cuda:0'), grad: tensor([[ 1.7509e-06, 9.3132e-07, 1.8969e-05, ..., -1.9856e-06, + 2.4438e-05, 3.3408e-05], + [ 1.6382e-06, 1.2498e-06, -2.9914e-06, ..., 4.0047e-06, + 2.3283e-06, 1.6056e-06], + [-1.5438e-05, 2.4401e-06, 3.4850e-06, ..., 3.0771e-06, + 3.2354e-06, 3.0473e-06], + ..., + [ 3.3136e-06, 3.0193e-06, 7.8138e-07, ..., 7.6070e-06, + 2.7250e-06, 9.9279e-07], + [ 1.0066e-05, 2.1420e-06, 1.8045e-05, ..., 2.8089e-06, + 1.2688e-05, 1.9878e-05], + [ 1.7295e-06, 2.9430e-06, 5.5768e-06, ..., -2.5719e-05, + -7.5512e-06, 4.2357e-06]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0040, 0.0139, 0.0162, -0.0013, 0.0189, -0.0073, -0.0132, -0.0089, + 0.0223, -0.0089], device='cuda:0'), grad: tensor([ 6.6698e-05, -6.6042e-04, 4.3130e-04, 1.0031e-04, 1.0587e-05, + 4.2319e-04, -5.0449e-04, 6.4731e-05, 6.0976e-05, 6.7391e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 251.52, cls_loss 0.0025 cls_loss_mapping 0.0074 cls_loss_causal 0.5723 re_mapping 0.0078 re_causal 0.0237 /// teacc 99.00 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0936, -0.0143, 0.1158, ..., 0.0357, -0.0974, -0.0920], + [-0.0660, -0.0684, -0.0073, ..., -0.0677, -0.0682, -0.0351], + [ 0.1234, -0.0304, -0.0873, ..., -0.0740, -0.0952, -0.1345], + ..., + [-0.0341, -0.0011, -0.0871, ..., -0.0253, 0.0401, 0.0025], + [-0.0896, -0.0689, -0.0117, ..., -0.0726, 0.0029, -0.0170], + [-0.0264, -0.0416, -0.0298, ..., 0.0725, 0.0289, -0.1018]], + device='cuda:0'), grad: tensor([[ 2.3749e-07, 1.7695e-08, -6.3241e-05, ..., -6.4790e-05, + 6.5938e-07, 6.1654e-07], + [ 4.2748e-07, 1.4249e-07, 1.9297e-06, ..., 1.7937e-06, + 2.1495e-06, 3.5577e-07], + [-4.4145e-06, 2.1886e-07, 4.1649e-06, ..., 3.9861e-06, + 8.1211e-07, 7.6648e-07], + ..., + [ 8.8289e-07, 1.7881e-07, 3.6001e-05, ..., 4.9859e-05, + 2.5984e-06, 2.0489e-08], + [ 2.7232e-06, 1.7416e-07, 6.7130e-06, ..., 1.3307e-05, + 1.8105e-05, 1.5637e-06], + [ 1.7416e-07, 1.2014e-07, 7.5586e-06, ..., -8.6054e-06, + -2.5257e-05, 1.2759e-07]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0041, 0.0142, 0.0159, -0.0018, 0.0191, -0.0069, -0.0136, -0.0088, + 0.0223, -0.0090], device='cuda:0'), grad: tensor([-1.4186e-04, -4.9509e-06, 7.2345e-06, 1.9092e-06, 2.8759e-06, + 1.1005e-05, -1.4286e-06, 1.1510e-04, 6.7472e-05, -5.7667e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 251.33, cls_loss 0.0030 cls_loss_mapping 0.0076 cls_loss_causal 0.5763 re_mapping 0.0074 re_causal 0.0234 /// teacc 98.98 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0938, -0.0144, 0.1161, ..., 0.0358, -0.0980, -0.0925], + [-0.0661, -0.0687, -0.0074, ..., -0.0678, -0.0686, -0.0354], + [ 0.1239, -0.0308, -0.0878, ..., -0.0744, -0.0955, -0.1348], + ..., + [-0.0346, -0.0014, -0.0875, ..., -0.0255, 0.0403, 0.0023], + [-0.0902, -0.0696, -0.0121, ..., -0.0738, 0.0029, -0.0170], + [-0.0265, -0.0417, -0.0300, ..., 0.0730, 0.0293, -0.1035]], + device='cuda:0'), grad: tensor([[ 8.3819e-07, 1.6764e-08, 7.5437e-07, ..., -2.4028e-07, + 2.1160e-06, 2.3618e-06], + [ 6.2492e-07, 1.3970e-08, -1.1269e-06, ..., 4.5355e-07, + 2.3320e-06, 6.3144e-07], + [-3.4068e-06, 3.9116e-08, 2.9933e-06, ..., 1.3486e-06, + 2.9728e-06, 2.5388e-06], + ..., + [ 1.2908e-06, 2.3283e-08, 3.7067e-07, ..., 5.4836e-06, + -3.4645e-06, 2.0303e-07], + [ 6.6273e-06, 1.7695e-08, 1.5823e-06, ..., 5.0962e-06, + 7.9051e-06, 7.2196e-06], + [ 2.3562e-06, 1.7695e-08, 8.2701e-07, ..., -5.2184e-05, + -1.6034e-05, 4.8578e-06]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0040, 0.0141, 0.0158, -0.0014, 0.0188, -0.0074, -0.0127, -0.0089, + 0.0221, -0.0089], device='cuda:0'), grad: tensor([ 8.3596e-06, -1.0610e-05, 1.3530e-05, 1.3089e-04, 7.0751e-05, + -1.9813e-04, 2.8238e-05, -1.7047e-05, 4.0352e-05, -6.6400e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 251.53, cls_loss 0.0032 cls_loss_mapping 0.0092 cls_loss_causal 0.6008 re_mapping 0.0071 re_causal 0.0232 /// teacc 99.04 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0941, -0.0145, 0.1165, ..., 0.0361, -0.0984, -0.0929], + [-0.0663, -0.0689, -0.0073, ..., -0.0681, -0.0689, -0.0355], + [ 0.1253, -0.0308, -0.0886, ..., -0.0752, -0.0957, -0.1355], + ..., + [-0.0353, -0.0016, -0.0878, ..., -0.0266, 0.0400, 0.0025], + [-0.0922, -0.0699, -0.0123, ..., -0.0746, 0.0026, -0.0171], + [-0.0268, -0.0419, -0.0303, ..., 0.0736, 0.0299, -0.1040]], + device='cuda:0'), grad: tensor([[ 1.2383e-05, 3.0547e-07, -5.6982e-05, ..., -2.1994e-05, + 5.5693e-07, 7.1526e-07], + [ 7.3127e-06, 1.1642e-06, 9.9465e-07, ..., 7.3016e-07, + 5.4948e-07, 2.4494e-07], + [-5.8353e-05, 4.0755e-06, 3.5111e-06, ..., -1.0943e-06, + 1.4408e-06, 5.8208e-07], + ..., + [ 1.4544e-05, 1.0123e-06, 8.8848e-07, ..., 1.0682e-06, + -7.9274e-06, 8.1025e-08], + [ 1.1332e-05, 3.3788e-06, 3.3360e-06, ..., 2.2613e-06, + 1.3141e-06, 6.2585e-07], + [ 4.9733e-06, 1.3383e-06, 1.6063e-05, ..., 2.0470e-06, + 2.1886e-06, 3.0734e-07]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0043, 0.0140, 0.0164, -0.0016, 0.0186, -0.0070, -0.0125, -0.0096, + 0.0208, -0.0081], device='cuda:0'), grad: tensor([-6.0230e-05, 3.7372e-05, -1.6287e-05, 3.0249e-05, 2.4617e-05, + 3.1114e-05, 1.7032e-05, -1.4436e-04, 3.2365e-05, 4.7922e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 251.30, cls_loss 0.0026 cls_loss_mapping 0.0087 cls_loss_causal 0.5810 re_mapping 0.0075 re_causal 0.0234 /// teacc 98.99 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0946, -0.0146, 0.1170, ..., 0.0364, -0.0987, -0.0931], + [-0.0664, -0.0691, -0.0074, ..., -0.0682, -0.0692, -0.0357], + [ 0.1265, -0.0306, -0.0893, ..., -0.0756, -0.0959, -0.1362], + ..., + [-0.0361, -0.0018, -0.0882, ..., -0.0268, 0.0404, 0.0023], + [-0.0927, -0.0702, -0.0125, ..., -0.0754, 0.0024, -0.0172], + [-0.0270, -0.0421, -0.0305, ..., 0.0739, 0.0297, -0.1042]], + device='cuda:0'), grad: tensor([[ 6.2957e-06, 3.0175e-07, -7.2002e-05, ..., -8.7440e-05, + 1.9465e-06, 9.5926e-07], + [ 4.7274e-06, 2.4959e-07, 2.8200e-06, ..., 1.1094e-05, + 4.7050e-06, 1.5739e-07], + [-1.3208e-04, 2.7940e-07, 6.0424e-06, ..., 8.6278e-06, + 1.0636e-06, 5.3458e-07], + ..., + [ 1.0198e-04, 5.4296e-07, 3.0287e-06, ..., 2.8521e-05, + 7.5847e-06, -1.9278e-07], + [ 8.6874e-06, 6.1281e-07, 4.4480e-06, ..., 1.4782e-05, + 5.8748e-06, -2.3674e-06], + [ 2.8275e-06, 4.6100e-07, 1.6034e-05, ..., -8.6427e-05, + -4.7415e-05, 1.6494e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0043, 0.0142, 0.0166, -0.0020, 0.0184, -0.0070, -0.0124, -0.0097, + 0.0206, -0.0081], device='cuda:0'), grad: tensor([-2.2149e-04, 1.8105e-05, -2.0480e-04, 1.2720e-04, 1.8358e-05, + 8.4341e-05, 5.0753e-05, 2.3675e-04, 2.1234e-05, -1.2958e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 251.01, cls_loss 0.0029 cls_loss_mapping 0.0087 cls_loss_causal 0.5776 re_mapping 0.0072 re_causal 0.0234 /// teacc 99.02 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0952, -0.0153, 0.1175, ..., 0.0367, -0.0994, -0.0936], + [-0.0664, -0.0694, -0.0065, ..., -0.0682, -0.0694, -0.0359], + [ 0.1273, -0.0308, -0.0908, ..., -0.0758, -0.0959, -0.1366], + ..., + [-0.0366, -0.0019, -0.0886, ..., -0.0274, 0.0400, 0.0024], + [-0.0930, -0.0694, -0.0123, ..., -0.0756, 0.0023, -0.0173], + [-0.0282, -0.0426, -0.0309, ..., 0.0741, 0.0301, -0.1049]], + device='cuda:0'), grad: tensor([[ 2.7493e-06, 5.5879e-09, -3.3885e-05, ..., -7.7412e-06, + 1.8999e-06, -1.2405e-06], + [ 1.1604e-06, 1.4901e-08, -5.6438e-06, ..., -2.3972e-06, + 2.8014e-06, 1.7332e-06], + [-1.3903e-05, 3.3528e-08, 2.3525e-06, ..., 6.5099e-07, + 5.7295e-06, 1.4668e-06], + ..., + [-3.4459e-08, 3.8184e-08, 1.5832e-06, ..., 1.0412e-06, + -1.6406e-05, 9.1270e-08], + [ 3.3975e-06, 4.5635e-08, 4.6417e-06, ..., 9.1642e-07, + 4.3362e-06, 3.1590e-06], + [ 3.4273e-07, 5.0291e-08, 4.7423e-06, ..., -2.1495e-06, + 6.3516e-06, 2.1681e-06]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0040, 0.0150, 0.0162, -0.0016, 0.0182, -0.0072, -0.0122, -0.0099, + 0.0211, -0.0082], device='cuda:0'), grad: tensor([-2.1324e-05, -3.4869e-05, 3.9488e-06, 3.1561e-05, 1.5333e-05, + -1.1697e-05, -1.5507e-06, -3.5524e-05, 2.0370e-05, 3.3706e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 251.52, cls_loss 0.0035 cls_loss_mapping 0.0082 cls_loss_causal 0.5595 re_mapping 0.0074 re_causal 0.0231 /// teacc 98.98 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0956, -0.0153, 0.1178, ..., 0.0352, -0.1000, -0.0940], + [-0.0667, -0.0696, -0.0067, ..., -0.0687, -0.0702, -0.0362], + [ 0.1282, -0.0304, -0.0914, ..., -0.0764, -0.0963, -0.1370], + ..., + [-0.0366, -0.0021, -0.0893, ..., -0.0272, 0.0407, 0.0027], + [-0.0936, -0.0698, -0.0125, ..., -0.0766, 0.0019, -0.0175], + [-0.0289, -0.0428, -0.0307, ..., 0.0767, 0.0311, -0.1060]], + device='cuda:0'), grad: tensor([[ 2.3574e-05, 6.9849e-08, 6.6981e-06, ..., -3.2783e-06, + 1.0487e-06, 1.0980e-06], + [ 5.3197e-06, 2.3562e-07, -2.4214e-05, ..., -3.2224e-06, + 1.2377e-06, -1.1168e-05], + [-1.0413e-04, 1.0291e-06, -4.2200e-05, ..., 3.2149e-06, + 1.5739e-06, 4.5784e-06], + ..., + [ 3.4720e-06, 4.7032e-07, 2.1458e-06, ..., 2.9989e-06, + 2.2817e-06, 9.7975e-07], + [ 4.9233e-05, 1.1884e-06, 3.3587e-05, ..., 6.4746e-06, + -2.7716e-06, 1.0766e-06], + [ 1.9949e-06, 5.6997e-07, 6.5491e-06, ..., -1.0598e-04, + -8.5294e-05, 2.2780e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0026, 0.0148, 0.0164, -0.0029, 0.0167, -0.0063, -0.0122, -0.0094, + 0.0207, -0.0065], device='cuda:0'), grad: tensor([ 3.1978e-05, -1.6654e-04, -8.6188e-05, 4.6305e-06, 2.1744e-04, + 4.9621e-05, 2.4945e-05, 2.2665e-05, 1.0264e-04, -2.0123e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 251.21, cls_loss 0.0026 cls_loss_mapping 0.0077 cls_loss_causal 0.5573 re_mapping 0.0073 re_causal 0.0234 /// teacc 99.02 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0957, -0.0154, 0.1185, ..., 0.0354, -0.1004, -0.0942], + [-0.0668, -0.0698, -0.0063, ..., -0.0684, -0.0706, -0.0362], + [ 0.1290, -0.0306, -0.0918, ..., -0.0765, -0.0968, -0.1374], + ..., + [-0.0369, -0.0022, -0.0900, ..., -0.0275, 0.0414, 0.0030], + [-0.0949, -0.0700, -0.0128, ..., -0.0780, 0.0016, -0.0176], + [-0.0293, -0.0430, -0.0311, ..., 0.0763, 0.0304, -0.1071]], + device='cuda:0'), grad: tensor([[ 1.1977e-06, 1.5274e-07, -2.6729e-06, ..., 2.7567e-07, + 2.5444e-06, 1.0403e-06], + [ 4.2357e-06, 1.3970e-08, 8.8066e-06, ..., 1.3169e-06, + 8.1658e-06, 6.2883e-06], + [-6.2436e-06, 5.6811e-08, 3.8017e-06, ..., 1.5358e-06, + 5.4911e-06, 2.8219e-06], + ..., + [ 4.2003e-07, 1.0245e-07, 1.4063e-07, ..., 4.5039e-06, + 1.2293e-07, -1.7881e-07], + [ 7.0129e-07, 1.6764e-08, 2.9895e-07, ..., -7.4506e-09, + -1.1146e-05, 4.0047e-07], + [ 4.9546e-07, 6.2399e-08, 2.3320e-06, ..., 2.3067e-05, + 1.7747e-05, 1.9465e-07]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0029, 0.0150, 0.0167, -0.0023, 0.0177, -0.0067, -0.0123, -0.0092, + 0.0197, -0.0074], device='cuda:0'), grad: tensor([ 1.3158e-05, 3.9160e-05, 1.7479e-05, 3.8117e-05, -7.5102e-05, + -1.7919e-06, -3.3706e-05, -1.8729e-06, -6.9857e-05, 7.4387e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 251.17, cls_loss 0.0026 cls_loss_mapping 0.0090 cls_loss_causal 0.5805 re_mapping 0.0075 re_causal 0.0232 /// teacc 98.98 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0968, -0.0155, 0.1197, ..., 0.0356, -0.1013, -0.0945], + [-0.0671, -0.0699, -0.0074, ..., -0.0685, -0.0712, -0.0370], + [ 0.1295, -0.0328, -0.0927, ..., -0.0776, -0.0973, -0.1377], + ..., + [-0.0375, -0.0025, -0.0905, ..., -0.0278, 0.0416, 0.0029], + [-0.0952, -0.0702, -0.0127, ..., -0.0781, 0.0015, -0.0175], + [-0.0294, -0.0432, -0.0318, ..., 0.0765, 0.0303, -0.1076]], + device='cuda:0'), grad: tensor([[ 1.0366e-06, 9.2201e-08, -1.0766e-05, ..., -5.3644e-06, + 2.8741e-06, 7.0781e-07], + [ 2.5984e-06, 3.1292e-07, -3.8296e-06, ..., 1.5553e-06, + 2.7772e-06, 4.7497e-07], + [ 7.6741e-06, 2.1122e-06, 2.6487e-06, ..., 5.5172e-06, + 1.8790e-05, 3.2317e-07], + ..., + [ 4.4703e-06, 1.6587e-06, 1.1651e-06, ..., -2.4021e-05, + -1.0967e-04, 1.5553e-07], + [ 1.7481e-06, 2.2259e-07, 1.2005e-06, ..., 3.3733e-06, + 5.0329e-06, 7.7952e-07], + [ 8.6501e-06, 9.7230e-07, 4.2170e-06, ..., 1.3053e-05, + 1.6466e-05, 5.9232e-07]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0030, 0.0150, 0.0163, -0.0022, 0.0178, -0.0067, -0.0121, -0.0093, + 0.0202, -0.0076], device='cuda:0'), grad: tensor([-2.4028e-06, -2.6298e-04, 7.7426e-05, -2.9176e-05, 1.1927e-04, + -2.4065e-05, 2.3931e-05, -3.8952e-05, 2.0966e-05, 1.1593e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 251.16, cls_loss 0.0026 cls_loss_mapping 0.0076 cls_loss_causal 0.5785 re_mapping 0.0072 re_causal 0.0229 /// teacc 99.04 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0970, -0.0155, 0.1215, ..., 0.0363, -0.1018, -0.0952], + [-0.0672, -0.0701, -0.0075, ..., -0.0682, -0.0724, -0.0371], + [ 0.1299, -0.0330, -0.0937, ..., -0.0784, -0.0979, -0.1380], + ..., + [-0.0375, -0.0027, -0.0911, ..., -0.0285, 0.0423, 0.0026], + [-0.0954, -0.0703, -0.0127, ..., -0.0780, 0.0016, -0.0174], + [-0.0297, -0.0434, -0.0328, ..., 0.0768, 0.0303, -0.1082]], + device='cuda:0'), grad: tensor([[ 1.4659e-06, 4.5169e-07, -4.1649e-06, ..., -2.2650e-06, + 1.3206e-06, 1.4259e-06], + [ 4.8019e-06, 1.6047e-06, 9.3598e-07, ..., -4.4145e-07, + 2.9374e-06, 1.2694e-06], + [ 7.7426e-05, 2.5362e-05, 1.4631e-06, ..., 2.5015e-06, + 1.8254e-05, 1.9316e-06], + ..., + [-2.1696e-05, 5.6289e-06, 3.8370e-07, ..., 2.0023e-06, + -2.5839e-05, 2.3376e-07], + [ 5.6252e-06, 1.9595e-06, -7.4469e-06, ..., -1.4119e-06, + -1.1846e-05, -1.4089e-05], + [ 8.2925e-06, 2.0545e-06, 5.8897e-06, ..., 6.5193e-06, + 7.8827e-06, 4.8950e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0037, 0.0153, 0.0156, -0.0024, 0.0173, -0.0069, -0.0125, -0.0091, + 0.0210, -0.0076], device='cuda:0'), grad: tensor([ 3.3826e-06, -1.5318e-05, 1.5402e-04, -1.1295e-04, 5.0366e-06, + 2.3305e-05, 1.3031e-05, -6.3837e-05, -6.4850e-05, 5.8204e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 100---------------------------------------------------- +epoch 100, time 268.46, cls_loss 0.0034 cls_loss_mapping 0.0093 cls_loss_causal 0.5889 re_mapping 0.0071 re_causal 0.0225 /// teacc 99.09 lr 0.00010000 +Epoch 102, weight, value: tensor([[-9.7476e-02, -1.5566e-02, 1.2246e-01, ..., 3.6668e-02, + -1.0234e-01, -9.5498e-02], + [-6.7772e-02, -7.0457e-02, -7.0759e-03, ..., -6.8958e-02, + -7.3398e-02, -3.7248e-02], + [ 1.3051e-01, -3.3285e-02, -9.4475e-02, ..., -7.8999e-02, + -9.8376e-02, -1.3857e-01], + ..., + [-3.7536e-02, -3.2772e-03, -9.1633e-02, ..., -2.8687e-02, + 4.2397e-02, 1.6970e-03], + [-9.5801e-02, -7.0848e-02, -1.4436e-02, ..., -7.8969e-02, + 7.6544e-05, -1.9684e-02], + [-3.0357e-02, -4.3802e-02, -3.3353e-02, ..., 7.6568e-02, + 2.9771e-02, -1.0947e-01]], device='cuda:0'), grad: tensor([[ 1.2340e-06, 1.0710e-06, -1.0312e-05, ..., -4.5374e-06, + 6.4727e-07, 6.9849e-07], + [ 1.7527e-06, 1.4789e-06, 8.1956e-07, ..., 2.1793e-06, + 1.3253e-06, 3.0175e-07], + [ 1.2472e-05, 1.1176e-05, 1.6773e-06, ..., 1.2480e-06, + 1.2768e-06, 1.3802e-06], + ..., + [ 6.5714e-06, 5.7369e-06, 4.3586e-07, ..., 1.3169e-06, + -7.4413e-07, 6.3889e-07], + [ 4.1425e-06, 3.6154e-06, 1.9092e-06, ..., 3.6806e-06, + 3.9823e-06, 5.0105e-06], + [ 6.3367e-06, 5.6438e-06, 3.8184e-06, ..., 4.4778e-06, + 3.0119e-06, 3.0641e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0041, 0.0149, 0.0154, -0.0013, 0.0185, -0.0077, -0.0116, -0.0088, + 0.0199, -0.0086], device='cuda:0'), grad: tensor([-7.2494e-06, 3.6247e-06, 4.8876e-05, -1.1683e-04, -2.9519e-05, + -7.8008e-06, 9.5367e-06, 2.1443e-05, 3.5405e-05, 4.2528e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 251.57, cls_loss 0.0025 cls_loss_mapping 0.0069 cls_loss_causal 0.5556 re_mapping 0.0068 re_causal 0.0215 /// teacc 98.94 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0978, -0.0156, 0.1233, ..., 0.0368, -0.1028, -0.0959], + [-0.0681, -0.0708, -0.0070, ..., -0.0694, -0.0739, -0.0374], + [ 0.1312, -0.0335, -0.0952, ..., -0.0791, -0.0988, -0.1391], + ..., + [-0.0378, -0.0036, -0.0923, ..., -0.0289, 0.0428, 0.0017], + [-0.0962, -0.0712, -0.0150, ..., -0.0799, -0.0005, -0.0200], + [-0.0307, -0.0443, -0.0336, ..., 0.0772, 0.0302, -0.1094]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -8.1837e-05, ..., -3.6716e-05, + 1.1176e-08, 1.8906e-07], + [ 1.7695e-08, 0.0000e+00, 7.4506e-08, ..., -2.1197e-06, + 3.1274e-06, 4.2841e-07], + [-9.4995e-08, 0.0000e+00, 2.0638e-05, ..., 9.3281e-06, + 8.2981e-07, 1.5739e-07], + ..., + [ 2.5146e-08, 0.0000e+00, 7.8231e-07, ..., 6.7316e-06, + -1.0483e-05, 9.4995e-08], + [ 2.0489e-08, 0.0000e+00, 1.8507e-05, ..., 8.8662e-06, + 1.6019e-07, -2.2072e-06], + [ 1.8626e-09, 0.0000e+00, 1.5229e-05, ..., -1.1642e-06, + -3.4291e-06, 6.7241e-07]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0043, 0.0149, 0.0154, -0.0012, 0.0181, -0.0079, -0.0115, -0.0087, + 0.0194, -0.0082], device='cuda:0'), grad: tensor([-1.2624e-04, -1.3947e-05, 3.7938e-05, 2.6777e-05, 6.5453e-06, + 6.7241e-06, 3.5226e-05, -1.7181e-05, 2.4304e-05, 1.9908e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 251.68, cls_loss 0.0024 cls_loss_mapping 0.0074 cls_loss_causal 0.5858 re_mapping 0.0074 re_causal 0.0234 /// teacc 98.94 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0983, -0.0157, 0.1239, ..., 0.0370, -0.1034, -0.0964], + [-0.0683, -0.0711, -0.0072, ..., -0.0697, -0.0744, -0.0379], + [ 0.1318, -0.0336, -0.0959, ..., -0.0795, -0.0996, -0.1406], + ..., + [-0.0383, -0.0041, -0.0928, ..., -0.0290, 0.0432, 0.0018], + [-0.0970, -0.0719, -0.0149, ..., -0.0804, -0.0007, -0.0202], + [-0.0309, -0.0450, -0.0339, ..., 0.0773, 0.0304, -0.1098]], + device='cuda:0'), grad: tensor([[ 6.2399e-08, 2.3022e-06, -4.2096e-07, ..., -1.2852e-07, + 1.4063e-06, 7.3388e-07], + [ 1.5832e-07, 1.3597e-06, 3.1665e-08, ..., 2.3376e-07, + 1.1222e-06, 4.4238e-07], + [-1.1688e-06, 3.4366e-07, 2.6077e-08, ..., 1.4063e-07, + 3.2410e-07, 1.2480e-07], + ..., + [ 4.9174e-07, 1.3486e-06, 1.6764e-08, ..., 4.4145e-07, + -3.0082e-07, 4.2655e-07], + [-4.5355e-07, 5.0277e-05, 1.5739e-07, ..., 3.2187e-06, + 3.0965e-05, 1.8850e-05], + [ 9.2201e-08, 7.5400e-06, 2.3097e-07, ..., -7.9423e-06, + 2.1271e-06, 2.8014e-06]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0045, 0.0149, 0.0151, -0.0024, 0.0181, -0.0057, -0.0125, -0.0087, + 0.0192, -0.0083], device='cuda:0'), grad: tensor([ 5.7369e-06, 4.5523e-06, 1.1725e-06, 6.3419e-04, 4.8578e-06, + -7.9107e-04, 3.2689e-06, 1.5665e-06, 1.3006e-04, 5.6773e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 251.52, cls_loss 0.0027 cls_loss_mapping 0.0075 cls_loss_causal 0.5576 re_mapping 0.0074 re_causal 0.0224 /// teacc 98.90 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0991, -0.0159, 0.1242, ..., 0.0371, -0.1042, -0.0972], + [-0.0685, -0.0715, -0.0074, ..., -0.0696, -0.0746, -0.0377], + [ 0.1324, -0.0338, -0.0969, ..., -0.0798, -0.1005, -0.1417], + ..., + [-0.0386, -0.0043, -0.0931, ..., -0.0291, 0.0436, 0.0017], + [-0.0975, -0.0722, -0.0146, ..., -0.0809, -0.0006, -0.0198], + [-0.0310, -0.0454, -0.0342, ..., 0.0774, 0.0302, -0.1104]], + device='cuda:0'), grad: tensor([[ 8.8476e-08, -6.9477e-07, -1.3709e-05, ..., -6.1207e-06, + 1.9874e-06, 1.6335e-06], + [ 1.7136e-07, 1.3132e-07, 2.4904e-06, ..., 8.2329e-07, + 4.6864e-06, 5.7779e-06], + [-4.7218e-07, 2.9616e-07, 1.4622e-06, ..., 8.7917e-07, + 4.4797e-07, 4.5821e-07], + ..., + [ 4.1164e-07, 3.8370e-07, 8.7172e-07, ..., 1.6717e-06, + 5.6718e-07, 2.1048e-07], + [ 2.2911e-07, 3.5390e-07, 4.8280e-06, ..., 6.7316e-06, + 6.1952e-06, 4.0419e-06], + [ 1.1921e-07, 4.4424e-07, 5.0776e-06, ..., -1.0565e-05, + -9.9614e-06, -9.6858e-07]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0043, 0.0141, 0.0164, -0.0027, 0.0184, -0.0061, -0.0123, -0.0087, + 0.0198, -0.0087], device='cuda:0'), grad: tensor([-1.4886e-05, 8.8811e-06, 6.6794e-06, 1.7881e-05, 1.6913e-05, + 1.2927e-05, -5.3823e-05, 5.7593e-06, 2.4125e-05, -2.4557e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 251.78, cls_loss 0.0025 cls_loss_mapping 0.0094 cls_loss_causal 0.5725 re_mapping 0.0071 re_causal 0.0211 /// teacc 98.94 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0995, -0.0160, 0.1249, ..., 0.0373, -0.1048, -0.0980], + [-0.0686, -0.0718, -0.0075, ..., -0.0699, -0.0750, -0.0379], + [ 0.1334, -0.0349, -0.0972, ..., -0.0801, -0.1005, -0.1420], + ..., + [-0.0395, -0.0045, -0.0939, ..., -0.0294, 0.0439, 0.0016], + [-0.0977, -0.0740, -0.0147, ..., -0.0820, -0.0010, -0.0202], + [-0.0312, -0.0457, -0.0346, ..., 0.0775, 0.0302, -0.1108]], + device='cuda:0'), grad: tensor([[ 2.9150e-06, -2.7567e-07, -4.6156e-06, ..., -1.3774e-06, + 1.4948e-06, 7.2364e-07], + [ 1.5451e-06, 7.8790e-07, 2.7381e-07, ..., 3.8017e-06, + 4.2282e-06, 5.3924e-07], + [-1.1288e-05, 5.1670e-06, 6.5193e-07, ..., 8.1304e-07, + 1.1541e-05, 6.8732e-07], + ..., + [-3.4552e-06, -2.2516e-05, 1.0151e-07, ..., 2.2259e-06, + -5.9545e-05, 3.2783e-07], + [ 1.2238e-06, 6.1374e-07, 2.1514e-07, ..., 9.8497e-06, + 4.8541e-06, 1.0580e-06], + [ 6.3423e-07, 1.2917e-06, 2.2836e-06, ..., 1.9763e-06, + 1.5154e-05, 4.7777e-07]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0045, 0.0141, 0.0163, -0.0029, 0.0187, -0.0055, -0.0125, -0.0085, + 0.0190, -0.0089], device='cuda:0'), grad: tensor([ 6.4671e-06, 2.0027e-05, 8.3447e-06, 1.5008e-04, -7.2896e-05, + -7.2658e-05, 1.3918e-05, -1.4806e-04, 4.1395e-05, 5.3227e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 251.45, cls_loss 0.0024 cls_loss_mapping 0.0075 cls_loss_causal 0.5546 re_mapping 0.0072 re_causal 0.0213 /// teacc 98.99 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0999, -0.0160, 0.1266, ..., 0.0379, -0.1055, -0.0984], + [-0.0689, -0.0722, -0.0081, ..., -0.0700, -0.0757, -0.0384], + [ 0.1344, -0.0349, -0.0982, ..., -0.0805, -0.1010, -0.1419], + ..., + [-0.0400, -0.0046, -0.0945, ..., -0.0301, 0.0439, 0.0018], + [-0.0981, -0.0736, -0.0145, ..., -0.0824, -0.0012, -0.0202], + [-0.0314, -0.0468, -0.0357, ..., 0.0777, 0.0306, -0.1120]], + device='cuda:0'), grad: tensor([[ 3.5502e-06, 9.5833e-07, -5.3644e-06, ..., -2.9542e-06, + 3.5036e-06, 1.5181e-06], + [ 4.5002e-06, 1.7909e-06, 4.3679e-07, ..., 8.5011e-06, + 3.7104e-05, 5.9232e-06], + [-2.3973e-04, -4.6939e-06, 1.4044e-06, ..., -3.7193e-05, + 1.0327e-05, 1.1940e-06], + ..., + [ 1.5711e-06, -2.6245e-06, 2.1420e-07, ..., -5.0589e-06, + -8.0585e-05, -7.5847e-06], + [ 1.1504e-05, -1.0446e-05, 9.0431e-07, ..., 2.8443e-06, + 3.2317e-06, -8.3745e-06], + [ 2.2316e-04, 1.0222e-05, 2.2780e-06, ..., 6.0380e-05, + 2.0653e-05, 3.3807e-06]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0054, 0.0144, 0.0165, -0.0032, 0.0185, -0.0055, -0.0131, -0.0090, + 0.0191, -0.0090], device='cuda:0'), grad: tensor([ 1.5706e-05, 1.7595e-04, -4.6372e-04, 1.8448e-05, -1.7792e-05, + -2.2054e-05, 2.1771e-05, -3.0494e-04, -1.5274e-05, 5.9080e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 251.77, cls_loss 0.0028 cls_loss_mapping 0.0092 cls_loss_causal 0.5830 re_mapping 0.0072 re_causal 0.0222 /// teacc 99.02 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.1003, -0.0160, 0.1248, ..., 0.0379, -0.1064, -0.1015], + [-0.0683, -0.0725, -0.0084, ..., -0.0703, -0.0765, -0.0388], + [ 0.1347, -0.0349, -0.0988, ..., -0.0808, -0.1014, -0.1423], + ..., + [-0.0401, -0.0049, -0.0949, ..., -0.0303, 0.0441, 0.0023], + [-0.0989, -0.0740, -0.0122, ..., -0.0827, -0.0007, -0.0186], + [-0.0319, -0.0471, -0.0363, ..., 0.0779, 0.0308, -0.1134]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 3.4459e-08, -6.1154e-05, ..., -4.0203e-05, + 3.2783e-07, -6.7893e-07], + [ 7.1712e-07, 1.1548e-07, 8.0746e-07, ..., -6.0797e-05, + 3.4645e-07, 9.9652e-08], + [-2.4326e-06, 1.4529e-07, 9.2760e-06, ..., 6.8583e-06, + 4.2748e-07, 2.2445e-07], + ..., + [ 9.4157e-07, 2.3749e-07, 2.9262e-06, ..., 2.0489e-05, + -1.0310e-06, -1.6764e-07], + [ 2.9616e-07, 1.7323e-07, 8.9109e-06, ..., 7.8306e-06, + 1.0785e-06, 6.5286e-07], + [ 2.0489e-07, 4.5635e-08, 1.1809e-05, ..., 1.7256e-05, + -6.5453e-06, 5.3924e-07]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0035, 0.0152, 0.0159, -0.0029, 0.0184, -0.0059, -0.0122, -0.0093, + 0.0204, -0.0089], device='cuda:0'), grad: tensor([-1.1945e-04, -2.2042e-04, 1.9208e-05, 9.3356e-06, 7.5817e-05, + 1.8269e-05, 3.1143e-05, 7.5340e-05, 2.7165e-05, 8.3268e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 251.76, cls_loss 0.0026 cls_loss_mapping 0.0076 cls_loss_causal 0.5618 re_mapping 0.0069 re_causal 0.0214 /// teacc 99.01 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.1007, -0.0159, 0.1258, ..., 0.0377, -0.1069, -0.1017], + [-0.0695, -0.0728, -0.0085, ..., -0.0708, -0.0771, -0.0390], + [ 0.1353, -0.0350, -0.0994, ..., -0.0812, -0.1019, -0.1427], + ..., + [-0.0391, -0.0051, -0.0956, ..., -0.0308, 0.0442, 0.0023], + [-0.0998, -0.0744, -0.0124, ..., -0.0834, -0.0007, -0.0185], + [-0.0321, -0.0474, -0.0365, ..., 0.0782, 0.0306, -0.1142]], + device='cuda:0'), grad: tensor([[ 2.8610e-06, 1.0394e-06, -3.6713e-06, ..., -2.4028e-07, + 1.3513e-06, 1.3784e-07], + [ 2.2687e-06, 7.0687e-07, -1.7043e-07, ..., 3.0603e-06, + 1.2713e-06, 6.3330e-08], + [-7.0870e-05, 1.6037e-06, 2.4121e-07, ..., 9.2108e-07, + 9.8068e-07, 6.6124e-08], + ..., + [ 6.1691e-05, 6.4038e-06, 3.3528e-07, ..., 1.6401e-06, + 3.5278e-06, 2.3283e-08], + [ 6.1318e-06, 2.6412e-06, 4.1164e-07, ..., 4.8056e-06, + 1.2796e-06, -7.4320e-07], + [ 6.7167e-06, 5.8189e-06, 2.0918e-06, ..., -1.2779e-04, + -3.5703e-05, 4.2189e-07]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0036, 0.0138, 0.0159, -0.0025, 0.0188, -0.0064, -0.0123, -0.0082, + 0.0206, -0.0090], device='cuda:0'), grad: tensor([ 6.3479e-06, 4.9323e-06, -1.2338e-04, -6.8665e-03, 1.4329e-04, + 6.8130e-03, 6.3851e-06, 1.4269e-04, 1.7881e-05, -1.4460e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 108---------------------------------------------------- +epoch 108, time 268.22, cls_loss 0.0019 cls_loss_mapping 0.0063 cls_loss_causal 0.5519 re_mapping 0.0067 re_causal 0.0209 /// teacc 99.12 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.1011, -0.0160, 0.1267, ..., 0.0383, -0.1076, -0.1019], + [-0.0696, -0.0732, -0.0086, ..., -0.0710, -0.0773, -0.0391], + [ 0.1356, -0.0361, -0.0998, ..., -0.0815, -0.1027, -0.1435], + ..., + [-0.0395, -0.0055, -0.0962, ..., -0.0310, 0.0445, 0.0025], + [-0.1003, -0.0747, -0.0125, ..., -0.0841, -0.0010, -0.0184], + [-0.0324, -0.0480, -0.0381, ..., 0.0779, 0.0305, -0.1151]], + device='cuda:0'), grad: tensor([[ 3.6648e-07, 3.1665e-08, -7.2122e-05, ..., -1.5289e-05, + 8.0839e-07, 3.3900e-07], + [ 1.2061e-06, 3.8184e-08, 1.1986e-06, ..., 2.2557e-06, + 1.2023e-06, 1.3784e-07], + [-3.7663e-06, 8.8476e-08, 4.2617e-06, ..., 1.5767e-06, + 3.0510e-06, 3.9078e-06], + ..., + [ 1.2387e-06, 5.2061e-07, 6.6962e-07, ..., 4.0270e-06, + -2.6450e-06, 1.9930e-07], + [ 3.7858e-07, 7.2177e-08, 4.2692e-06, ..., 6.4149e-06, + 3.3155e-06, 6.0583e-07], + [ 1.2200e-07, 1.5507e-07, 9.6560e-06, ..., -1.6168e-05, + -1.0401e-05, 1.0477e-07]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0041, 0.0139, 0.0158, -0.0013, 0.0191, -0.0068, -0.0123, -0.0083, + 0.0205, -0.0096], device='cuda:0'), grad: tensor([-9.4056e-05, -6.0126e-06, 1.1399e-05, 1.0885e-05, 1.5661e-05, + 5.0575e-05, 1.1355e-05, -1.3243e-06, 2.0549e-05, -1.9073e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 251.23, cls_loss 0.0023 cls_loss_mapping 0.0066 cls_loss_causal 0.5749 re_mapping 0.0065 re_causal 0.0209 /// teacc 99.11 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.1018, -0.0158, 0.1284, ..., 0.0387, -0.1073, -0.1020], + [-0.0700, -0.0733, -0.0082, ..., -0.0711, -0.0778, -0.0388], + [ 0.1365, -0.0361, -0.1017, ..., -0.0824, -0.1032, -0.1442], + ..., + [-0.0395, -0.0056, -0.0970, ..., -0.0314, 0.0454, 0.0027], + [-0.1012, -0.0751, -0.0125, ..., -0.0848, -0.0013, -0.0185], + [-0.0325, -0.0481, -0.0384, ..., 0.0782, 0.0305, -0.1146]], + device='cuda:0'), grad: tensor([[ 2.6990e-06, 4.4703e-08, -2.7288e-07, ..., 9.3132e-09, + 1.3057e-06, 1.0254e-06], + [ 1.1688e-06, 6.9849e-08, 7.6368e-08, ..., 1.0366e-06, + 2.9821e-06, 6.3051e-07], + [-9.0674e-06, 4.0326e-07, 3.4645e-07, ..., 8.1491e-07, + 2.6003e-06, 1.5302e-06], + ..., + [ 5.6904e-07, 8.7544e-08, 5.3085e-08, ..., -4.0174e-05, + -1.4234e-04, -5.2899e-05], + [ 3.4962e-06, 1.9651e-07, 4.7199e-06, ..., 1.0906e-06, + 3.7365e-06, 3.8929e-06], + [ 2.7753e-07, 5.5879e-08, 3.2783e-07, ..., 3.6061e-05, + 1.3340e-04, 5.1081e-05]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0047, 0.0146, 0.0161, -0.0016, 0.0188, -0.0072, -0.0126, -0.0085, + 0.0199, -0.0096], device='cuda:0'), grad: tensor([ 9.3579e-06, 3.2596e-06, -6.9775e-06, 1.2323e-05, 2.7698e-06, + 1.3426e-05, -1.3486e-05, -4.8065e-04, 1.5339e-06, 4.5800e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 251.25, cls_loss 0.0023 cls_loss_mapping 0.0069 cls_loss_causal 0.5397 re_mapping 0.0064 re_causal 0.0198 /// teacc 99.05 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.1025, -0.0167, 0.1285, ..., 0.0388, -0.1080, -0.1023], + [-0.0701, -0.0736, -0.0083, ..., -0.0714, -0.0787, -0.0393], + [ 0.1373, -0.0363, -0.1022, ..., -0.0831, -0.1036, -0.1448], + ..., + [-0.0399, -0.0058, -0.0979, ..., -0.0317, 0.0460, 0.0045], + [-0.1017, -0.0752, -0.0125, ..., -0.0854, -0.0022, -0.0195], + [-0.0324, -0.0483, -0.0388, ..., 0.0788, 0.0307, -0.1151]], + device='cuda:0'), grad: tensor([[ 1.3877e-07, 9.7789e-08, -8.8755e-07, ..., 7.3574e-08, + 9.3691e-07, 3.2317e-07], + [ 1.6764e-07, 4.2468e-07, -7.9442e-07, ..., 3.1497e-06, + 4.4517e-06, 6.6869e-07], + [ 6.8918e-07, 8.2236e-07, 9.0804e-07, ..., 9.3412e-07, + 1.3085e-06, 4.9546e-07], + ..., + [ 4.1071e-07, 5.4017e-07, 2.1607e-07, ..., -9.6951e-07, + -6.6161e-06, 4.7963e-07], + [ 3.0920e-07, 3.4459e-07, 7.4506e-07, ..., 5.7556e-06, + 6.1747e-07, 1.0915e-06], + [ 5.1688e-07, 5.8208e-07, 3.7625e-07, ..., -9.0301e-05, + -6.3956e-05, 1.9744e-07]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0044, 0.0147, 0.0160, -0.0026, 0.0184, -0.0060, -0.0127, -0.0081, + 0.0192, -0.0093], device='cuda:0'), grad: tensor([ 2.6412e-06, -3.0734e-06, 2.0415e-05, 6.8638e-07, 1.7166e-04, + 1.8388e-05, -7.3016e-06, -3.4034e-05, -4.2133e-06, -1.6510e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 251.21, cls_loss 0.0018 cls_loss_mapping 0.0063 cls_loss_causal 0.5464 re_mapping 0.0069 re_causal 0.0214 /// teacc 98.99 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.1030, -0.0167, 0.1289, ..., 0.0391, -0.1090, -0.1027], + [-0.0706, -0.0738, -0.0083, ..., -0.0719, -0.0794, -0.0398], + [ 0.1382, -0.0365, -0.1027, ..., -0.0834, -0.1041, -0.1453], + ..., + [-0.0405, -0.0061, -0.0983, ..., -0.0319, 0.0468, 0.0062], + [-0.1020, -0.0754, -0.0127, ..., -0.0859, -0.0024, -0.0195], + [-0.0327, -0.0486, -0.0392, ..., 0.0786, 0.0301, -0.1166]], + device='cuda:0'), grad: tensor([[ 4.6007e-07, 1.1269e-07, 5.5786e-07, ..., 4.8149e-07, + 4.5635e-07, 5.1688e-07], + [ 8.7731e-07, 3.9767e-07, 2.6356e-07, ..., -1.2234e-05, + 1.9297e-06, 9.8255e-07], + [-3.0920e-07, 1.2964e-06, 9.0897e-06, ..., 4.3679e-07, + 1.2014e-06, 5.3868e-06], + ..., + [ 1.9111e-06, 8.1398e-07, 9.4995e-08, ..., 1.0490e-05, + -7.8008e-06, -7.3481e-07], + [ 1.0133e-06, 5.7928e-07, 1.7341e-06, ..., 2.7008e-06, + 1.2470e-06, -1.8720e-07], + [ 1.0189e-06, 1.1008e-06, 1.9465e-07, ..., -2.5034e-05, + -8.1807e-06, -2.9933e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0045, 0.0146, 0.0162, -0.0027, 0.0186, -0.0059, -0.0125, -0.0079, + 0.0190, -0.0098], device='cuda:0'), grad: tensor([ 4.5747e-06, -3.2514e-05, 4.3064e-05, -5.1260e-06, 5.5462e-05, + 1.2115e-05, -3.3587e-05, 3.8669e-06, -8.2999e-06, -3.9577e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 251.60, cls_loss 0.0023 cls_loss_mapping 0.0083 cls_loss_causal 0.5365 re_mapping 0.0068 re_causal 0.0206 /// teacc 98.90 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.1036, -0.0169, 0.1295, ..., 0.0395, -0.1095, -0.1029], + [-0.0711, -0.0743, -0.0083, ..., -0.0740, -0.0798, -0.0402], + [ 0.1394, -0.0367, -0.1033, ..., -0.0837, -0.1026, -0.1452], + ..., + [-0.0407, -0.0065, -0.0989, ..., -0.0324, 0.0463, 0.0057], + [-0.1024, -0.0758, -0.0128, ..., -0.0866, -0.0022, -0.0193], + [-0.0335, -0.0491, -0.0396, ..., 0.0791, 0.0302, -0.1168]], + device='cuda:0'), grad: tensor([[ 1.6559e-06, 4.7218e-07, -1.5469e-06, ..., -1.8906e-07, + 1.1669e-06, 6.2957e-07], + [ 2.2054e-05, 2.2221e-06, 7.5437e-08, ..., 1.3206e-06, + 2.2165e-06, 1.1930e-06], + [ 1.6958e-05, 3.8929e-06, 1.6391e-07, ..., 1.5534e-06, + 2.5649e-06, 4.6566e-07], + ..., + [-4.7922e-05, 1.6123e-05, 1.3970e-08, ..., 3.7812e-06, + 7.1637e-06, 1.5106e-06], + [ 5.4389e-06, 5.3607e-06, -1.3784e-07, ..., 6.1058e-06, + 5.6699e-06, 5.5321e-07], + [ 8.9407e-07, 8.2329e-07, 1.1148e-06, ..., -7.6517e-06, + -6.4075e-06, -5.1446e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0046, 0.0142, 0.0172, -0.0031, 0.0184, -0.0055, -0.0128, -0.0086, + 0.0192, -0.0095], device='cuda:0'), grad: tensor([ 2.2560e-05, 2.7943e-04, 2.4915e-04, 1.2147e-04, -5.3197e-06, + 3.0100e-05, 1.0006e-05, -7.4291e-04, 4.7207e-05, -1.1332e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 251.30, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5498 re_mapping 0.0065 re_causal 0.0209 /// teacc 98.97 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.1046, -0.0170, 0.1300, ..., 0.0392, -0.1100, -0.1030], + [-0.0715, -0.0752, -0.0086, ..., -0.0744, -0.0803, -0.0405], + [ 0.1402, -0.0369, -0.1037, ..., -0.0837, -0.1029, -0.1458], + ..., + [-0.0413, -0.0068, -0.0994, ..., -0.0327, 0.0470, 0.0067], + [-0.1030, -0.0763, -0.0129, ..., -0.0870, -0.0026, -0.0193], + [-0.0337, -0.0493, -0.0399, ..., 0.0795, 0.0301, -0.1173]], + device='cuda:0'), grad: tensor([[ 6.6590e-07, 1.9651e-07, -1.3364e-06, ..., -8.5216e-07, + 1.4333e-06, 1.4119e-06], + [ 1.3122e-06, 8.4843e-07, 5.1782e-07, ..., 6.6403e-07, + 8.3297e-06, 1.0587e-05], + [ 5.0925e-06, 3.1181e-06, 4.0606e-07, ..., 9.2573e-07, + 6.5379e-06, 8.7097e-06], + ..., + [ 2.4326e-06, 1.9036e-06, 2.3283e-07, ..., 6.6794e-06, + 3.1646e-06, 7.9256e-07], + [ 5.4482e-07, 5.6066e-07, 1.1459e-05, ..., 1.1744e-06, + 6.0678e-05, 6.7651e-05], + [ 6.9942e-07, -1.3011e-06, 3.6880e-07, ..., -1.0431e-05, + -7.5847e-06, -6.1467e-08]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0044, 0.0141, 0.0175, -0.0032, 0.0184, -0.0053, -0.0131, -0.0085, + 0.0189, -0.0095], device='cuda:0'), grad: tensor([ 4.2245e-06, 5.1737e-05, 5.8323e-05, -1.7345e-05, 6.2168e-05, + -2.4331e-04, -1.3661e-04, 1.9237e-05, 2.1255e-04, -1.1489e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 250.87, cls_loss 0.0031 cls_loss_mapping 0.0085 cls_loss_causal 0.5928 re_mapping 0.0061 re_causal 0.0200 /// teacc 98.92 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.1050, -0.0173, 0.1309, ..., 0.0400, -0.1109, -0.1033], + [-0.0726, -0.0759, -0.0087, ..., -0.0745, -0.0809, -0.0407], + [ 0.1398, -0.0385, -0.1044, ..., -0.0841, -0.1042, -0.1462], + ..., + [-0.0397, -0.0052, -0.1002, ..., -0.0328, 0.0477, 0.0068], + [-0.1037, -0.0767, -0.0131, ..., -0.0876, -0.0028, -0.0195], + [-0.0342, -0.0498, -0.0422, ..., 0.0790, 0.0300, -0.1178]], + device='cuda:0'), grad: tensor([[ 7.1339e-07, 3.1665e-08, -2.2352e-07, ..., 1.2107e-07, + 5.4482e-07, 5.4855e-07], + [ 1.2238e-06, 6.9849e-08, 4.0047e-08, ..., 6.2771e-07, + 7.2923e-07, 4.1630e-07], + [-6.9328e-06, 6.6124e-08, 1.2852e-07, ..., 8.3167e-07, + 7.8324e-07, 4.2096e-07], + ..., + [ 7.1619e-07, 9.2201e-08, 2.9802e-08, ..., 7.6462e-07, + -1.2843e-06, 8.7544e-08], + [ 1.7555e-06, 5.1223e-08, 5.7742e-08, ..., 6.4448e-07, + 2.8107e-06, 3.2876e-06], + [ 2.0862e-07, 6.9849e-08, 1.2107e-07, ..., 3.3434e-06, + 1.8505e-06, 9.2573e-07]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0051, 0.0153, 0.0161, -0.0025, 0.0189, -0.0052, -0.0129, -0.0082, + 0.0187, -0.0106], device='cuda:0'), grad: tensor([ 3.1032e-06, 2.0601e-06, -5.8301e-06, 7.7114e-06, -2.1875e-05, + -2.2560e-05, 1.5423e-05, -1.3513e-06, 1.2144e-05, 1.1064e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 251.00, cls_loss 0.0040 cls_loss_mapping 0.0106 cls_loss_causal 0.5624 re_mapping 0.0067 re_causal 0.0198 /// teacc 99.01 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.1059, -0.0174, 0.1311, ..., 0.0397, -0.1117, -0.1037], + [-0.0730, -0.0762, -0.0088, ..., -0.0750, -0.0817, -0.0428], + [ 0.1407, -0.0385, -0.1048, ..., -0.0855, -0.1047, -0.1461], + ..., + [-0.0397, -0.0054, -0.1013, ..., -0.0340, 0.0480, 0.0068], + [-0.1050, -0.0770, -0.0130, ..., -0.0886, -0.0030, -0.0185], + [-0.0343, -0.0498, -0.0426, ..., 0.0815, 0.0308, -0.1181]], + device='cuda:0'), grad: tensor([[ 5.4762e-07, -2.5705e-07, -2.9597e-06, ..., 1.4693e-05, + 1.5512e-05, 5.8115e-07], + [ 9.6764e-07, 5.3737e-07, 9.7137e-07, ..., 4.5858e-06, + 2.8983e-06, 1.4901e-07], + [ 4.6380e-06, 3.0752e-06, 5.0217e-06, ..., 1.1586e-05, + 9.1791e-06, 2.2259e-07], + ..., + [ 1.2340e-06, 7.5251e-07, 4.0978e-06, ..., 1.6615e-05, + 9.9465e-06, 1.0803e-07], + [ 4.6100e-07, 3.3434e-07, 4.0680e-05, ..., 1.1390e-04, + 9.5129e-05, 1.5832e-08], + [ 5.5041e-07, 3.7067e-07, -1.0335e-04, ..., -2.7847e-04, + -2.5177e-04, 2.4494e-07]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0047, 0.0144, 0.0163, -0.0030, 0.0167, -0.0051, -0.0132, -0.0078, + 0.0197, -0.0085], device='cuda:0'), grad: tensor([ 8.5473e-05, 1.7300e-05, 7.5579e-05, 9.2611e-06, 7.3373e-05, + 5.6028e-04, 4.8071e-05, 7.9989e-05, 6.0320e-04, -1.5526e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 251.16, cls_loss 0.0022 cls_loss_mapping 0.0068 cls_loss_causal 0.5564 re_mapping 0.0063 re_causal 0.0201 /// teacc 98.99 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.1068, -0.0175, 0.1313, ..., 0.0398, -0.1124, -0.1038], + [-0.0732, -0.0765, -0.0093, ..., -0.0752, -0.0824, -0.0430], + [ 0.1416, -0.0386, -0.1054, ..., -0.0862, -0.1051, -0.1464], + ..., + [-0.0398, -0.0055, -0.1020, ..., -0.0344, 0.0484, 0.0070], + [-0.1053, -0.0772, -0.0131, ..., -0.0892, -0.0034, -0.0186], + [-0.0347, -0.0499, -0.0428, ..., 0.0812, 0.0304, -0.1186]], + device='cuda:0'), grad: tensor([[ 7.7300e-08, 1.5926e-07, 1.5390e-04, ..., -3.1944e-07, + 2.9624e-05, 7.5996e-05], + [ 1.8161e-07, 2.9616e-07, -5.1595e-07, ..., 1.4044e-06, + 3.2000e-06, 1.2089e-06], + [ 8.3167e-07, 2.6356e-07, 3.0696e-06, ..., 2.3842e-07, + 2.9951e-06, 2.1961e-06], + ..., + [-1.5795e-05, 1.3597e-07, 5.4296e-07, ..., -4.8466e-06, + -4.6432e-05, -3.6974e-06], + [ 5.6811e-08, 1.5814e-06, 1.7658e-06, ..., 2.9057e-07, + 2.2314e-06, 3.9786e-06], + [ 5.4948e-07, 3.4273e-07, 1.6419e-06, ..., 3.8017e-06, + 6.9179e-06, 1.7611e-06]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0044, 0.0142, 0.0163, -0.0018, 0.0171, -0.0062, -0.0133, -0.0076, + 0.0197, -0.0089], device='cuda:0'), grad: tensor([ 3.4785e-04, 1.4938e-06, 1.6168e-05, 7.9274e-05, 1.0997e-05, + 1.1332e-05, -3.7193e-04, -1.4794e-04, 1.9863e-05, 3.3379e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 251.11, cls_loss 0.0017 cls_loss_mapping 0.0059 cls_loss_causal 0.5639 re_mapping 0.0065 re_causal 0.0206 /// teacc 99.09 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.1073, -0.0176, 0.1315, ..., 0.0400, -0.1130, -0.1043], + [-0.0735, -0.0767, -0.0094, ..., -0.0756, -0.0828, -0.0431], + [ 0.1429, -0.0387, -0.1062, ..., -0.0865, -0.1055, -0.1463], + ..., + [-0.0401, -0.0056, -0.1023, ..., -0.0348, 0.0488, 0.0069], + [-0.1055, -0.0774, -0.0132, ..., -0.0900, -0.0037, -0.0187], + [-0.0371, -0.0500, -0.0430, ..., 0.0811, 0.0304, -0.1193]], + device='cuda:0'), grad: tensor([[ 7.7393e-07, 8.3819e-09, -5.5954e-06, ..., -1.3150e-06, + 8.9593e-07, 3.2969e-07], + [ 4.0159e-06, 3.6322e-08, 2.3991e-06, ..., 7.7765e-07, + 4.2766e-06, 3.4831e-07], + [-2.0638e-05, 7.7300e-08, 7.9256e-07, ..., 4.5076e-07, + 1.3761e-05, 4.4517e-07], + ..., + [-2.2128e-05, 4.1910e-08, 1.3318e-07, ..., 2.2016e-06, + -2.2948e-05, -1.2759e-07], + [ 1.7462e-06, 1.4901e-08, 8.2701e-07, ..., 2.8461e-06, + -2.4028e-06, -2.3171e-06], + [ 5.5507e-07, 3.1665e-08, 3.8743e-07, ..., -3.9078e-06, + -1.8422e-06, 3.4459e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0043, 0.0142, 0.0167, -0.0021, 0.0174, -0.0061, -0.0130, -0.0077, + 0.0198, -0.0092], device='cuda:0'), grad: tensor([-5.3346e-06, 2.3633e-05, 2.4825e-05, 5.3883e-05, -5.7928e-07, + 1.4603e-05, 5.3234e-06, -9.9719e-05, -1.0028e-05, -6.6347e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 251.15, cls_loss 0.0018 cls_loss_mapping 0.0064 cls_loss_causal 0.5562 re_mapping 0.0064 re_causal 0.0202 /// teacc 99.00 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.1078, -0.0179, 0.1319, ..., 0.0402, -0.1134, -0.1044], + [-0.0737, -0.0770, -0.0100, ..., -0.0772, -0.0834, -0.0438], + [ 0.1433, -0.0389, -0.1067, ..., -0.0873, -0.1057, -0.1468], + ..., + [-0.0403, -0.0059, -0.1029, ..., -0.0357, 0.0488, 0.0069], + [-0.1059, -0.0775, -0.0132, ..., -0.0908, -0.0037, -0.0186], + [-0.0371, -0.0502, -0.0432, ..., 0.0811, 0.0304, -0.1197]], + device='cuda:0'), grad: tensor([[ 1.1176e-07, 2.0489e-08, -2.1651e-05, ..., -1.3866e-05, + 2.8871e-07, 8.4750e-08], + [ 4.0699e-07, 6.5193e-08, 1.1828e-07, ..., -2.0023e-07, + 3.6508e-07, 6.7055e-08], + [-1.0533e-06, 1.0803e-07, 1.0403e-06, ..., 9.1270e-07, + 8.6054e-07, 2.0768e-07], + ..., + [ 2.7660e-07, 7.2643e-08, 2.5425e-07, ..., 6.0536e-07, + -1.1865e-06, 4.1910e-08], + [ 4.0513e-07, -8.0094e-08, 5.1446e-06, ..., 3.8147e-06, + -5.2862e-06, -1.2098e-06], + [ 3.0361e-07, 1.7602e-07, 1.2085e-05, ..., 1.1235e-05, + 2.1327e-06, 1.5367e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0044, 0.0143, 0.0166, -0.0024, 0.0179, -0.0060, -0.0134, -0.0081, + 0.0200, -0.0093], device='cuda:0'), grad: tensor([-2.8312e-05, -1.0028e-05, 6.5304e-06, 5.1707e-06, -6.9775e-06, + 1.8612e-05, 5.8934e-06, -5.9344e-06, -1.5870e-05, 3.0905e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 251.45, cls_loss 0.0018 cls_loss_mapping 0.0048 cls_loss_causal 0.5526 re_mapping 0.0061 re_causal 0.0193 /// teacc 99.00 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.1084, -0.0180, 0.1323, ..., 0.0405, -0.1140, -0.1045], + [-0.0744, -0.0772, -0.0101, ..., -0.0760, -0.0840, -0.0440], + [ 0.1441, -0.0392, -0.1071, ..., -0.0876, -0.1060, -0.1468], + ..., + [-0.0405, -0.0060, -0.1032, ..., -0.0368, 0.0494, 0.0071], + [-0.1063, -0.0777, -0.0137, ..., -0.0924, -0.0043, -0.0189], + [-0.0374, -0.0505, -0.0435, ..., 0.0810, 0.0300, -0.1203]], + device='cuda:0'), grad: tensor([[ 3.5148e-06, 1.4901e-08, 2.2985e-06, ..., 2.1327e-07, + 3.1106e-07, 3.1386e-07], + [ 2.9523e-07, 6.0536e-08, 1.9092e-07, ..., -3.1292e-07, + 2.8498e-07, 1.7416e-07], + [-8.4750e-08, 1.0710e-07, 8.2795e-07, ..., 3.8650e-07, + 2.3469e-07, 3.2783e-07], + ..., + [ 5.1409e-07, 4.8429e-08, 2.9337e-07, ..., 3.8184e-07, + -8.1677e-07, 2.0489e-08], + [ 1.2303e-06, 6.3330e-08, 1.2033e-06, ..., 3.5763e-07, + 4.7870e-07, 4.7777e-07], + [ 3.8058e-05, 1.3597e-07, 2.7806e-05, ..., 2.5723e-06, + 7.5437e-08, 9.7789e-08]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0045, 0.0150, 0.0168, -0.0026, 0.0182, -0.0059, -0.0131, -0.0086, + 0.0195, -0.0097], device='cuda:0'), grad: tensor([ 1.9655e-05, -1.1094e-05, 6.9812e-06, 1.2726e-05, 3.3826e-06, + -2.4390e-04, -2.4438e-06, 2.6543e-06, 7.1079e-06, 2.0456e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 250.94, cls_loss 0.0024 cls_loss_mapping 0.0064 cls_loss_causal 0.5702 re_mapping 0.0057 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.1088, -0.0180, 0.1327, ..., 0.0405, -0.1144, -0.1046], + [-0.0747, -0.0774, -0.0100, ..., -0.0761, -0.0844, -0.0442], + [ 0.1444, -0.0393, -0.1077, ..., -0.0882, -0.1063, -0.1473], + ..., + [-0.0405, -0.0061, -0.1039, ..., -0.0370, 0.0497, 0.0072], + [-0.1064, -0.0778, -0.0139, ..., -0.0934, -0.0046, -0.0191], + [-0.0377, -0.0507, -0.0437, ..., 0.0809, 0.0301, -0.1207]], + device='cuda:0'), grad: tensor([[ 5.2340e-07, 0.0000e+00, -4.9826e-07, ..., 1.8440e-07, + 4.0513e-07, 4.9360e-08], + [ 8.7917e-07, 1.8626e-09, 2.1420e-08, ..., 6.6217e-07, + 8.3260e-07, 1.4901e-08], + [-5.4203e-06, 2.7940e-09, 6.8918e-08, ..., 4.6100e-07, + 4.8336e-07, 4.0047e-08], + ..., + [ 9.9838e-07, 1.8626e-09, 3.6322e-08, ..., 8.3633e-07, + -5.0105e-07, 8.3819e-09], + [ 1.5264e-06, 9.3132e-10, 1.0524e-07, ..., 3.7681e-06, + 3.0845e-06, 7.9162e-08], + [ 3.8184e-08, 9.3132e-10, 1.7136e-07, ..., 4.3869e-05, + 3.6001e-05, 1.9558e-08]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0045, 0.0147, 0.0162, -0.0027, 0.0187, -0.0058, -0.0132, -0.0080, + 0.0191, -0.0102], device='cuda:0'), grad: tensor([ 1.3877e-06, 2.2762e-06, -6.5863e-06, 6.0275e-06, -1.0997e-04, + -1.5907e-06, 2.5675e-05, -1.2508e-06, 1.3806e-05, 7.0333e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 251.26, cls_loss 0.0017 cls_loss_mapping 0.0041 cls_loss_causal 0.5591 re_mapping 0.0061 re_causal 0.0195 /// teacc 98.90 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.1096, -0.0181, 0.1324, ..., 0.0405, -0.1151, -0.1054], + [-0.0749, -0.0775, -0.0110, ..., -0.0762, -0.0848, -0.0446], + [ 0.1451, -0.0393, -0.1081, ..., -0.0887, -0.1066, -0.1477], + ..., + [-0.0409, -0.0062, -0.1048, ..., -0.0374, 0.0501, 0.0073], + [-0.1073, -0.0779, -0.0143, ..., -0.0939, -0.0051, -0.0193], + [-0.0379, -0.0507, -0.0440, ..., 0.0809, 0.0300, -0.1210]], + device='cuda:0'), grad: tensor([[ 1.2526e-06, 3.1665e-08, -4.6305e-06, ..., -3.0342e-06, + 6.2771e-07, 1.7043e-07], + [ 1.2852e-06, 1.0058e-07, -1.3597e-07, ..., -3.6974e-07, + 2.2631e-06, 1.3039e-08], + [-1.1288e-05, 2.0489e-07, 6.1281e-07, ..., 4.1816e-07, + 8.3074e-06, 3.5390e-08], + ..., + [ 4.9882e-06, 9.4995e-08, 2.7101e-07, ..., 7.5717e-07, + -3.4552e-06, 7.4506e-09], + [-4.2841e-06, 7.7020e-07, 3.3528e-07, ..., 4.2655e-07, + -1.7568e-05, 7.4506e-09], + [ 8.5868e-07, 1.2480e-07, 2.9653e-06, ..., 1.5181e-07, + 1.4296e-06, 2.6077e-08]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0042, 0.0149, 0.0164, -0.0030, 0.0188, -0.0054, -0.0123, -0.0083, + 0.0193, -0.0104], device='cuda:0'), grad: tensor([ 4.7758e-06, 1.7419e-05, 4.8518e-05, 4.0174e-05, 2.2247e-05, + 5.1022e-05, 1.7017e-05, 4.1127e-06, -2.6751e-04, 6.2168e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 251.47, cls_loss 0.0019 cls_loss_mapping 0.0052 cls_loss_causal 0.5382 re_mapping 0.0056 re_causal 0.0188 /// teacc 99.00 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.1113, -0.0179, 0.1341, ..., 0.0411, -0.1137, -0.1048], + [-0.0752, -0.0778, -0.0116, ..., -0.0764, -0.0859, -0.0448], + [ 0.1457, -0.0395, -0.1092, ..., -0.0897, -0.1072, -0.1480], + ..., + [-0.0411, -0.0064, -0.1070, ..., -0.0382, 0.0503, 0.0073], + [-0.1075, -0.0782, -0.0146, ..., -0.0950, -0.0054, -0.0195], + [-0.0374, -0.0510, -0.0445, ..., 0.0809, 0.0299, -0.1212]], + device='cuda:0'), grad: tensor([[ 2.2035e-06, 1.7695e-08, -8.8383e-07, ..., -2.8498e-07, + 1.5004e-06, 8.9779e-07], + [ 1.9334e-06, 4.2841e-08, 1.7788e-07, ..., 1.5674e-06, + 8.1025e-07, 1.9372e-07], + [-4.0084e-05, 7.6368e-08, 5.7183e-07, ..., 7.7486e-07, + 5.2527e-07, 1.8068e-07], + ..., + [ 2.3797e-05, 1.5087e-07, 2.9616e-07, ..., 2.8070e-06, + 1.1064e-06, 5.3085e-08], + [ 8.5309e-06, 1.2573e-07, 3.9451e-06, ..., 1.0878e-06, + 2.5723e-06, 1.8850e-06], + [ 2.8778e-07, 5.0291e-08, 8.0839e-07, ..., 2.7031e-05, + 1.2375e-05, 8.4750e-08]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0054, 0.0149, 0.0161, -0.0033, 0.0189, -0.0058, -0.0121, -0.0084, + 0.0196, -0.0105], device='cuda:0'), grad: tensor([ 8.9630e-06, -2.7977e-06, -1.1486e-04, 6.7540e-06, -6.3419e-05, + 2.8625e-05, -2.9713e-05, 8.0705e-05, 3.5167e-05, 5.0604e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 251.66, cls_loss 0.0016 cls_loss_mapping 0.0058 cls_loss_causal 0.5395 re_mapping 0.0063 re_causal 0.0199 /// teacc 99.02 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.1128, -0.0179, 0.1347, ..., 0.0413, -0.1141, -0.1048], + [-0.0755, -0.0780, -0.0118, ..., -0.0766, -0.0864, -0.0449], + [ 0.1468, -0.0396, -0.1098, ..., -0.0901, -0.1074, -0.1481], + ..., + [-0.0407, -0.0065, -0.1078, ..., -0.0391, 0.0503, 0.0072], + [-0.1088, -0.0784, -0.0146, ..., -0.0958, -0.0058, -0.0196], + [-0.0405, -0.0512, -0.0448, ..., 0.0809, 0.0300, -0.1212]], + device='cuda:0'), grad: tensor([[ 6.1467e-08, 7.4506e-09, 1.8440e-07, ..., 7.4506e-08, + 3.9116e-07, 3.4552e-07], + [ 2.1700e-07, 3.0734e-08, 6.1467e-08, ..., 8.8476e-08, + 9.4995e-07, 7.9162e-08], + [-6.0815e-07, 9.0338e-08, 7.8231e-08, ..., 9.2201e-08, + 2.9989e-07, 8.0094e-08], + ..., + [ 4.5355e-07, -1.3970e-08, 6.5193e-09, ..., 4.9733e-07, + -3.8929e-06, 2.7940e-08], + [ 6.9849e-08, 3.5390e-08, 2.8592e-07, ..., 5.9605e-07, + 6.2678e-07, 4.0885e-07], + [ 4.7497e-08, 4.0047e-08, 5.8673e-08, ..., 1.6481e-05, + 9.0674e-06, 1.0431e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0056, 0.0149, 0.0163, -0.0037, 0.0192, -0.0059, -0.0120, -0.0083, + 0.0191, -0.0108], device='cuda:0'), grad: tensor([ 1.6456e-06, 1.3346e-06, 7.9721e-07, 2.2035e-06, -3.4273e-05, + -2.6673e-06, -1.7257e-06, -1.5169e-05, 3.6247e-06, 4.4167e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 251.28, cls_loss 0.0018 cls_loss_mapping 0.0060 cls_loss_causal 0.5423 re_mapping 0.0061 re_causal 0.0193 /// teacc 98.95 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.1138, -0.0181, 0.1353, ..., 0.0415, -0.1145, -0.1049], + [-0.0759, -0.0781, -0.0118, ..., -0.0768, -0.0870, -0.0449], + [ 0.1475, -0.0397, -0.1102, ..., -0.0906, -0.1076, -0.1483], + ..., + [-0.0410, -0.0072, -0.1084, ..., -0.0387, 0.0516, 0.0073], + [-0.1091, -0.0785, -0.0147, ..., -0.0967, -0.0060, -0.0197], + [-0.0405, -0.0515, -0.0454, ..., 0.0808, 0.0297, -0.1214]], + device='cuda:0'), grad: tensor([[ 1.9465e-07, -9.3132e-09, -1.4482e-06, ..., 1.4985e-06, + 6.8173e-07, 2.9989e-07], + [ 3.3155e-07, 9.1270e-08, 1.5739e-07, ..., 1.2312e-06, + 7.9069e-07, 5.4948e-08], + [-6.0536e-07, 3.4366e-07, 9.6858e-08, ..., 5.4482e-07, + 2.7139e-06, 9.3132e-08], + ..., + [-5.0757e-07, -4.3306e-07, 2.3842e-07, ..., 8.8811e-06, + -2.5108e-06, 2.5146e-08], + [ 2.3283e-07, 6.2399e-08, -3.0383e-05, ..., 2.7716e-06, + -1.3486e-05, -1.7524e-05], + [ 6.5193e-08, 7.1712e-08, 1.1576e-06, ..., -4.3452e-05, + -6.2287e-06, 2.6263e-07]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0057, 0.0148, 0.0165, -0.0036, 0.0193, -0.0061, -0.0120, -0.0080, + 0.0195, -0.0114], device='cuda:0'), grad: tensor([ 5.1074e-06, 7.3854e-07, 7.9423e-06, 1.2934e-05, 5.7489e-05, + 2.1502e-05, 5.8144e-05, 1.0341e-05, -8.1360e-05, -9.2745e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 251.17, cls_loss 0.0020 cls_loss_mapping 0.0061 cls_loss_causal 0.5413 re_mapping 0.0061 re_causal 0.0189 /// teacc 98.90 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.1156, -0.0186, 0.1355, ..., 0.0416, -0.1154, -0.1052], + [-0.0761, -0.0787, -0.0122, ..., -0.0772, -0.0876, -0.0452], + [ 0.1481, -0.0398, -0.1100, ..., -0.0911, -0.1079, -0.1486], + ..., + [-0.0412, -0.0077, -0.1091, ..., -0.0391, 0.0517, 0.0069], + [-0.1096, -0.0791, -0.0153, ..., -0.0975, -0.0065, -0.0198], + [-0.0406, -0.0519, -0.0457, ..., 0.0817, 0.0306, -0.1217]], + device='cuda:0'), grad: tensor([[ 4.5355e-07, 3.7253e-09, -2.2929e-06, ..., -9.3412e-07, + 2.4382e-06, 2.5760e-06], + [ 5.0291e-08, 3.7253e-09, 3.2969e-07, ..., 8.3260e-07, + 8.9221e-07, 5.3179e-07], + [ 1.1520e-06, 5.5879e-09, 2.5313e-06, ..., 8.2701e-07, + 9.8124e-06, 1.0349e-05], + ..., + [ 1.2852e-07, 7.4506e-09, 5.3551e-07, ..., 3.4589e-06, + 1.7993e-06, 3.9674e-07], + [ 1.7975e-07, 6.5193e-09, 9.0152e-07, ..., 1.0421e-06, + -4.1127e-06, -1.0198e-06], + [ 2.2817e-07, 3.7253e-09, 1.0822e-06, ..., 4.6529e-06, + 3.6471e-06, 1.8757e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0055, 0.0147, 0.0166, -0.0035, 0.0186, -0.0056, -0.0120, -0.0080, + 0.0190, -0.0107], device='cuda:0'), grad: tensor([ 6.3777e-06, -3.3863e-06, 3.3945e-05, 3.8475e-05, 7.3910e-06, + -9.3803e-06, -6.5148e-05, 1.4067e-05, -6.8188e-05, 4.5836e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 250.98, cls_loss 0.0020 cls_loss_mapping 0.0053 cls_loss_causal 0.5343 re_mapping 0.0062 re_causal 0.0194 /// teacc 98.94 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.1166, -0.0187, 0.1358, ..., 0.0418, -0.1164, -0.1055], + [-0.0763, -0.0789, -0.0117, ..., -0.0776, -0.0903, -0.0453], + [ 0.1496, -0.0400, -0.1107, ..., -0.0919, -0.1083, -0.1489], + ..., + [-0.0420, -0.0080, -0.1098, ..., -0.0396, 0.0530, 0.0069], + [-0.1112, -0.0794, -0.0157, ..., -0.0983, -0.0068, -0.0199], + [-0.0405, -0.0521, -0.0460, ..., 0.0814, 0.0304, -0.1219]], + device='cuda:0'), grad: tensor([[ 6.7428e-07, 7.2177e-07, -2.4475e-06, ..., -1.1632e-06, + 2.4773e-07, 4.2841e-08], + [ 2.2072e-07, 1.5926e-07, 9.2201e-08, ..., 6.7428e-07, + 5.2433e-07, 8.3819e-09], + [ 2.1048e-07, 5.9512e-07, 3.6322e-07, ..., 5.0198e-07, + 5.8580e-07, 1.4901e-08], + ..., + [ 3.1572e-07, 2.5425e-07, 1.0151e-07, ..., 2.6338e-06, + -1.7229e-07, 3.7253e-09], + [ 1.2107e-06, 1.3085e-06, 2.9989e-07, ..., 7.5530e-07, + -8.4005e-07, 5.2154e-08], + [ 2.7791e-06, 3.0361e-06, 6.7428e-07, ..., -7.1675e-06, + -1.6466e-06, 1.6764e-08]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0054, 0.0143, 0.0177, -0.0041, 0.0192, -0.0059, -0.0114, -0.0083, + 0.0195, -0.0112], device='cuda:0'), grad: tensor([ 3.4384e-06, 3.6210e-06, 8.3968e-06, -5.0455e-05, 4.7833e-06, + 9.3505e-06, 1.3150e-06, 1.9837e-06, 2.8349e-06, 1.4715e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 251.40, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5312 re_mapping 0.0056 re_causal 0.0179 /// teacc 99.04 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.1199, -0.0196, 0.1365, ..., 0.0421, -0.1177, -0.1064], + [-0.0772, -0.0796, -0.0123, ..., -0.0767, -0.0907, -0.0454], + [ 0.1502, -0.0400, -0.1118, ..., -0.0924, -0.1087, -0.1492], + ..., + [-0.0416, -0.0081, -0.1099, ..., -0.0412, 0.0534, 0.0068], + [-0.1127, -0.0799, -0.0161, ..., -0.1003, -0.0071, -0.0201], + [-0.0409, -0.0533, -0.0463, ..., 0.0815, 0.0304, -0.1223]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, 1.4156e-07, -3.3528e-06, ..., -5.8115e-07, + 4.4610e-07, 9.8720e-08], + [ 1.2387e-07, 3.3900e-07, -5.2154e-08, ..., 1.6885e-06, + 7.2550e-07, 3.6322e-08], + [ 5.4017e-07, 2.1625e-06, 9.1176e-07, ..., 1.1079e-05, + 4.4629e-06, 1.5926e-07], + ..., + [ 6.8359e-07, 2.5071e-06, 2.3376e-07, ..., 1.4238e-05, + 5.6289e-06, 4.0047e-08], + [ 1.3784e-07, 2.2631e-07, 6.1002e-07, ..., 1.4920e-06, + 6.2026e-07, -1.2293e-07], + [ 4.2375e-07, 1.7211e-06, 8.3726e-07, ..., 6.7241e-06, + 2.6282e-06, 7.3574e-08]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0048, 0.0141, 0.0176, -0.0033, 0.0192, -0.0056, -0.0115, -0.0080, + 0.0188, -0.0112], device='cuda:0'), grad: tensor([-1.8170e-06, 3.2037e-06, 2.4199e-05, -4.0270e-06, -7.5817e-05, + 6.7353e-06, 3.0156e-06, 2.8804e-05, 2.1346e-06, 1.3545e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 251.61, cls_loss 0.0018 cls_loss_mapping 0.0057 cls_loss_causal 0.5298 re_mapping 0.0057 re_causal 0.0179 /// teacc 99.04 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.1206, -0.0199, 0.1372, ..., 0.0405, -0.1184, -0.1064], + [-0.0778, -0.0802, -0.0127, ..., -0.0772, -0.0911, -0.0456], + [ 0.1509, -0.0405, -0.1126, ..., -0.0938, -0.1093, -0.1495], + ..., + [-0.0418, -0.0051, -0.1110, ..., -0.0410, 0.0544, 0.0069], + [-0.1132, -0.0809, -0.0165, ..., -0.1008, -0.0075, -0.0203], + [-0.0410, -0.0538, -0.0465, ..., 0.0823, 0.0303, -0.1227]], + device='cuda:0'), grad: tensor([[ 1.6391e-07, 4.8429e-08, -3.9674e-06, ..., -2.2706e-06, + 7.8976e-07, 9.1642e-07], + [ 1.4463e-06, 1.0524e-07, -1.4529e-07, ..., 9.9652e-08, + 9.0618e-07, 7.9442e-07], + [-5.7444e-06, 1.6950e-07, 6.2771e-07, ..., 4.2282e-07, + 5.8487e-07, 4.2189e-07], + ..., + [ 5.1595e-07, 9.4995e-08, 3.2783e-07, ..., 1.5190e-06, + -9.6299e-07, 9.5926e-08], + [ 4.1313e-06, 3.7905e-07, 7.9535e-07, ..., 1.8440e-07, + 1.3830e-06, 1.9409e-06], + [ 2.8405e-07, 9.4995e-08, 2.2035e-06, ..., 1.6754e-06, + 2.5313e-06, 1.6559e-06]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0035, 0.0141, 0.0177, -0.0036, 0.0192, -0.0055, -0.0116, -0.0078, + 0.0187, -0.0107], device='cuda:0'), grad: tensor([ 4.8131e-06, -5.2862e-06, -5.4538e-06, -1.2666e-07, -2.1178e-06, + 3.6918e-06, -1.9863e-05, 6.3032e-06, -1.0580e-05, 2.8566e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 251.70, cls_loss 0.0016 cls_loss_mapping 0.0061 cls_loss_causal 0.5686 re_mapping 0.0059 re_causal 0.0190 /// teacc 99.01 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.1212, -0.0201, 0.1377, ..., 0.0408, -0.1194, -0.1067], + [-0.0785, -0.0811, -0.0131, ..., -0.0775, -0.0915, -0.0458], + [ 0.1523, -0.0405, -0.1137, ..., -0.0944, -0.1095, -0.1497], + ..., + [-0.0418, -0.0052, -0.1117, ..., -0.0414, 0.0547, 0.0068], + [-0.1136, -0.0814, -0.0165, ..., -0.1010, -0.0076, -0.0204], + [-0.0426, -0.0545, -0.0469, ..., 0.0823, 0.0302, -0.1231]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, 2.0210e-07, 1.2405e-06, ..., 2.6729e-07, + 9.9186e-07, 1.9763e-06], + [ 3.4785e-07, 1.7229e-08, 1.0012e-07, ..., 6.2445e-07, + 4.0606e-07, 1.2573e-07], + [-3.3751e-06, 1.3039e-08, 7.0781e-08, ..., 1.0803e-07, + 2.0443e-07, 9.5461e-08], + ..., + [ 1.8068e-07, 8.8476e-09, 9.7789e-09, ..., 3.8892e-06, + 2.2724e-06, 1.7229e-08], + [ 2.3860e-06, 6.2399e-08, 3.6601e-07, ..., 1.2834e-06, + 4.7451e-07, 5.8115e-07], + [ 2.1420e-08, 1.3504e-08, 4.7497e-08, ..., -9.5740e-06, + -4.1723e-06, 1.2061e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0037, 0.0142, 0.0179, -0.0036, 0.0192, -0.0055, -0.0121, -0.0078, + 0.0190, -0.0110], device='cuda:0'), grad: tensor([ 3.9637e-06, -1.3066e-06, -2.5313e-06, 1.2040e-05, 4.9882e-06, + -1.0714e-05, -4.1015e-06, 1.0714e-05, 1.0841e-05, -2.3961e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 251.44, cls_loss 0.0019 cls_loss_mapping 0.0065 cls_loss_causal 0.5534 re_mapping 0.0059 re_causal 0.0182 /// teacc 99.02 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.1215, -0.0199, 0.1387, ..., 0.0411, -0.1198, -0.1069], + [-0.0788, -0.0818, -0.0130, ..., -0.0778, -0.0921, -0.0459], + [ 0.1529, -0.0408, -0.1147, ..., -0.0956, -0.1101, -0.1501], + ..., + [-0.0420, -0.0053, -0.1126, ..., -0.0417, 0.0553, 0.0067], + [-0.1145, -0.0813, -0.0168, ..., -0.1017, -0.0079, -0.0208], + [-0.0427, -0.0555, -0.0484, ..., 0.0818, 0.0292, -0.1235]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 9.3132e-09, -5.4389e-07, ..., -6.4261e-08, + 5.8394e-07, 1.0245e-07], + [ 7.8883e-07, 5.0757e-08, 9.3132e-09, ..., 1.3234e-06, + 5.3160e-06, 4.0513e-08], + [-9.2434e-07, 4.7497e-08, 8.9873e-08, ..., 4.6520e-07, + 1.4743e-06, 8.6147e-08], + ..., + [ 2.0163e-07, 5.9139e-08, 2.9337e-08, ..., -1.2028e-04, + -5.0640e-04, 5.2154e-08], + [ 2.8871e-08, -5.6345e-07, 8.0094e-08, ..., -1.0226e-06, + 3.0361e-07, -1.2731e-06], + [ 1.8161e-08, 3.4692e-07, 1.8068e-07, ..., 1.1629e-04, + 4.9448e-04, 8.9360e-07]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0042, 0.0144, 0.0179, -0.0029, 0.0200, -0.0069, -0.0119, -0.0077, + 0.0185, -0.0118], device='cuda:0'), grad: tensor([ 1.7136e-06, 1.5840e-05, 3.7495e-06, 4.1015e-06, 1.1750e-05, + 3.3863e-06, 1.5702e-06, -1.3952e-03, -2.0325e-05, 1.3742e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 251.64, cls_loss 0.0016 cls_loss_mapping 0.0049 cls_loss_causal 0.5526 re_mapping 0.0061 re_causal 0.0192 /// teacc 98.92 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.1218, -0.0200, 0.1373, ..., 0.0413, -0.1203, -0.1099], + [-0.0791, -0.0823, -0.0133, ..., -0.0787, -0.0929, -0.0460], + [ 0.1536, -0.0411, -0.1154, ..., -0.0962, -0.1106, -0.1509], + ..., + [-0.0425, -0.0053, -0.1133, ..., -0.0407, 0.0571, 0.0070], + [-0.1148, -0.0815, -0.0171, ..., -0.1028, -0.0084, -0.0210], + [-0.0427, -0.0563, -0.0487, ..., 0.0819, 0.0290, -0.1238]], + device='cuda:0'), grad: tensor([[ 2.0210e-07, 1.3504e-08, -5.6326e-06, ..., -5.2266e-06, + 2.4633e-07, 5.4948e-08], + [ 2.3236e-07, 5.7276e-08, 1.9604e-07, ..., 3.5297e-07, + 8.7544e-08, 1.0245e-08], + [ 2.7614e-07, 7.8697e-08, 5.7137e-07, ..., 1.1222e-06, + 1.3411e-07, 7.4040e-08], + ..., + [ 4.4610e-07, 1.1036e-07, 1.5087e-07, ..., 2.8927e-06, + 1.5385e-06, 7.4506e-09], + [ 2.8498e-07, 1.9092e-08, 3.2270e-07, ..., 2.5779e-06, + 1.0505e-06, 2.3283e-09], + [-2.7250e-06, 7.9628e-08, 1.5274e-06, ..., -1.4849e-05, + -4.4405e-06, 1.6764e-08]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0030, 0.0143, 0.0180, -0.0035, 0.0197, -0.0063, -0.0100, -0.0073, + 0.0179, -0.0118], device='cuda:0'), grad: tensor([-1.0453e-05, -3.9011e-05, 5.8711e-06, 1.3106e-05, 5.9195e-06, + 4.5598e-06, 5.9605e-06, 3.9130e-05, 9.9912e-06, -3.5197e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 251.62, cls_loss 0.0016 cls_loss_mapping 0.0054 cls_loss_causal 0.5640 re_mapping 0.0059 re_causal 0.0187 /// teacc 98.95 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.1233, -0.0201, 0.1377, ..., 0.0415, -0.1207, -0.1102], + [-0.0794, -0.0828, -0.0133, ..., -0.0792, -0.0938, -0.0461], + [ 0.1541, -0.0412, -0.1165, ..., -0.0975, -0.1112, -0.1511], + ..., + [-0.0427, -0.0066, -0.1141, ..., -0.0412, 0.0568, 0.0069], + [-0.1156, -0.0817, -0.0171, ..., -0.1051, -0.0094, -0.0210], + [-0.0424, -0.0566, -0.0494, ..., 0.0822, 0.0295, -0.1239]], + device='cuda:0'), grad: tensor([[ 6.1467e-08, 1.1642e-08, -9.5461e-07, ..., -7.6927e-07, + 4.0280e-07, 3.7998e-07], + [ 1.5041e-07, 6.6124e-08, 1.7043e-07, ..., -4.9220e-07, + 1.0058e-06, 3.0082e-07], + [-3.2876e-07, 6.7055e-08, 4.4098e-07, ..., 1.9046e-07, + 5.1828e-07, 3.1246e-07], + ..., + [ 1.1735e-07, 1.5832e-08, 6.6124e-08, ..., 5.1875e-07, + -1.9729e-05, 1.3784e-07], + [ 1.9558e-07, 4.7497e-08, -8.0690e-06, ..., 3.7998e-07, + 3.7067e-06, -5.0701e-06], + [ 8.0094e-08, 5.6811e-08, 1.3430e-06, ..., 9.1735e-08, + 8.9705e-06, 1.3970e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0030, 0.0141, 0.0178, -0.0026, 0.0197, -0.0063, -0.0103, -0.0076, + 0.0171, -0.0112], device='cuda:0'), grad: tensor([ 5.5786e-07, -1.5786e-07, 2.7008e-06, 4.0643e-06, 6.8434e-06, + 1.9535e-05, 9.7603e-06, -6.4850e-05, -1.6078e-05, 3.7551e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 251.57, cls_loss 0.0016 cls_loss_mapping 0.0040 cls_loss_causal 0.5475 re_mapping 0.0056 re_causal 0.0180 /// teacc 99.04 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.1241, -0.0201, 0.1380, ..., 0.0418, -0.1211, -0.1104], + [-0.0823, -0.0832, -0.0130, ..., -0.0800, -0.0943, -0.0459], + [ 0.1550, -0.0414, -0.1175, ..., -0.0981, -0.1115, -0.1524], + ..., + [-0.0416, -0.0067, -0.1152, ..., -0.0418, 0.0569, 0.0066], + [-0.1159, -0.0820, -0.0167, ..., -0.1060, -0.0096, -0.0204], + [-0.0426, -0.0568, -0.0500, ..., 0.0823, 0.0297, -0.1246]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, 7.5903e-08, 1.1781e-06, ..., -2.0582e-07, + 6.3423e-07, 8.5542e-07], + [ 2.9150e-07, 1.1781e-07, 1.1828e-07, ..., 2.2817e-07, + 2.1979e-07, 1.2387e-07], + [-4.0093e-07, 8.7544e-08, 5.5088e-07, ..., 1.8999e-07, + 3.6694e-07, 5.2387e-07], + ..., + [ 3.3434e-07, 1.7555e-07, 3.0734e-08, ..., 6.1654e-07, + 3.7532e-07, 1.8906e-07], + [ 4.9220e-07, 9.2201e-08, 3.3341e-07, ..., 9.0711e-07, + 7.1060e-07, 1.0291e-07], + [ 1.1129e-07, 1.1362e-07, 4.1537e-07, ..., -2.6505e-06, + -1.3476e-06, 4.5262e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0030, 0.0132, 0.0180, -0.0027, 0.0198, -0.0066, -0.0106, -0.0070, + 0.0179, -0.0112], device='cuda:0'), grad: tensor([ 4.9211e-06, -2.1607e-06, 4.7050e-06, 1.1593e-05, -5.5879e-09, + -2.2963e-05, -3.0398e-06, 5.2936e-06, 3.3192e-06, -1.6503e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 251.84, cls_loss 0.0016 cls_loss_mapping 0.0056 cls_loss_causal 0.5659 re_mapping 0.0057 re_causal 0.0189 /// teacc 99.06 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.1246, -0.0201, 0.1386, ..., 0.0421, -0.1215, -0.1104], + [-0.0825, -0.0835, -0.0134, ..., -0.0793, -0.0949, -0.0460], + [ 0.1555, -0.0417, -0.1192, ..., -0.0990, -0.1120, -0.1529], + ..., + [-0.0417, -0.0067, -0.1158, ..., -0.0427, 0.0572, 0.0064], + [-0.1161, -0.0822, -0.0166, ..., -0.1065, -0.0097, -0.0204], + [-0.0427, -0.0570, -0.0504, ..., 0.0825, 0.0299, -0.1248]], + device='cuda:0'), grad: tensor([[ 4.2841e-08, 7.8231e-08, -1.0294e-04, ..., -2.2888e-05, + 1.1967e-06, 4.1397e-07], + [ 7.9628e-08, 7.5903e-08, 5.5600e-07, ..., 8.5216e-07, + 1.0356e-06, 1.9465e-07], + [ 3.8836e-07, 1.5367e-07, 1.6734e-05, ..., 8.7023e-06, + 2.4419e-06, 4.8522e-07], + ..., + [ 7.9162e-08, -2.9989e-06, 7.3062e-07, ..., -3.2354e-06, + -3.1203e-05, 1.6764e-08], + [ 4.9826e-08, 2.0023e-08, 3.0566e-06, ..., 9.2667e-07, + 6.2026e-07, 1.0878e-06], + [ 9.1735e-08, 2.9337e-07, 1.3173e-05, ..., 4.5821e-06, + 4.1351e-06, 8.5216e-08]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0033, 0.0138, 0.0177, -0.0009, 0.0194, -0.0081, -0.0111, -0.0073, + 0.0181, -0.0112], device='cuda:0'), grad: tensor([-1.8477e-04, 1.8915e-06, 4.7475e-05, 9.9838e-06, 2.1178e-06, + 5.7578e-05, 1.1420e-04, -9.0778e-05, 1.0617e-06, 4.1127e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 250.76, cls_loss 0.0019 cls_loss_mapping 0.0057 cls_loss_causal 0.5243 re_mapping 0.0054 re_causal 0.0171 /// teacc 99.00 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.1256, -0.0204, 0.1391, ..., 0.0423, -0.1219, -0.1105], + [-0.0828, -0.0840, -0.0133, ..., -0.0798, -0.0960, -0.0462], + [ 0.1566, -0.0419, -0.1199, ..., -0.0998, -0.1126, -0.1536], + ..., + [-0.0420, -0.0072, -0.1167, ..., -0.0438, 0.0577, 0.0064], + [-0.1169, -0.0828, -0.0167, ..., -0.1071, -0.0099, -0.0203], + [-0.0430, -0.0581, -0.0508, ..., 0.0827, 0.0299, -0.1252]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 4.1910e-09, -2.5988e-05, ..., -5.9940e-06, + 1.1036e-07, 7.0315e-08], + [ 2.1886e-08, 2.1420e-08, 2.3376e-07, ..., 9.3132e-07, + 4.0745e-07, 3.0268e-08], + [ 2.0489e-08, 4.3772e-08, 1.4948e-06, ..., 7.2364e-07, + 4.0978e-07, 7.8231e-08], + ..., + [ 3.3528e-08, -9.3598e-08, 2.7008e-07, ..., 3.7439e-07, + -2.8946e-06, 1.6298e-08], + [ 3.5390e-08, 2.9802e-08, 2.1029e-06, ..., 9.5740e-07, + 5.8068e-07, 7.0315e-08], + [ 2.1420e-08, 1.8161e-08, 1.7686e-06, ..., 7.4282e-06, + 1.3988e-06, 1.3690e-07]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0034, 0.0138, 0.0179, -0.0030, 0.0196, -0.0061, -0.0111, -0.0074, + 0.0180, -0.0112], device='cuda:0'), grad: tensor([-3.9935e-05, 1.7453e-06, 3.9265e-06, 1.3471e-05, -1.5259e-05, + 1.0068e-06, 2.4110e-05, -8.2254e-06, 5.2117e-06, 1.3895e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 246.95, cls_loss 0.0016 cls_loss_mapping 0.0045 cls_loss_causal 0.5449 re_mapping 0.0058 re_causal 0.0183 /// teacc 99.02 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.1262, -0.0205, 0.1396, ..., 0.0424, -0.1224, -0.1105], + [-0.0825, -0.0845, -0.0137, ..., -0.0800, -0.0970, -0.0476], + [ 0.1570, -0.0423, -0.1205, ..., -0.1007, -0.1133, -0.1538], + ..., + [-0.0424, -0.0074, -0.1175, ..., -0.0439, 0.0586, 0.0062], + [-0.1186, -0.0832, -0.0178, ..., -0.1104, -0.0110, -0.0208], + [-0.0423, -0.0585, -0.0505, ..., 0.0829, 0.0298, -0.1256]], + device='cuda:0'), grad: tensor([[ 1.0598e-06, 1.4901e-08, -4.5374e-06, ..., -7.3090e-06, + 4.2375e-08, 2.6543e-08], + [ 6.3740e-06, 2.9337e-08, 1.2154e-07, ..., 2.0117e-07, + 4.8429e-08, 1.1176e-08], + [-1.3515e-05, 3.0734e-08, 8.7544e-08, ..., 1.4761e-07, + 7.3109e-08, 4.3772e-08], + ..., + [ 2.9113e-06, 3.9116e-08, 7.2177e-08, ..., 4.9407e-07, + 2.2491e-07, 3.7253e-08], + [ 1.3504e-06, 7.7393e-07, 1.2107e-07, ..., 2.9337e-07, + 9.0431e-07, 8.1584e-07], + [ 1.1781e-07, 9.7789e-08, 3.9786e-06, ..., 4.8839e-06, + -5.8021e-07, 9.7323e-08]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0035, 0.0142, 0.0176, -0.0029, 0.0197, -0.0053, -0.0125, -0.0074, + 0.0162, -0.0112], device='cuda:0'), grad: tensor([-9.4548e-06, 9.0376e-06, -2.1189e-05, 2.4550e-06, 2.2147e-06, + -5.0031e-06, 6.3470e-07, 6.6534e-06, 6.9402e-06, 7.7263e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 251.56, cls_loss 0.0009 cls_loss_mapping 0.0036 cls_loss_causal 0.5372 re_mapping 0.0059 re_causal 0.0193 /// teacc 98.98 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.1266, -0.0205, 0.1398, ..., 0.0426, -0.1228, -0.1106], + [-0.0828, -0.0849, -0.0139, ..., -0.0801, -0.0976, -0.0481], + [ 0.1579, -0.0422, -0.1209, ..., -0.1012, -0.1133, -0.1542], + ..., + [-0.0428, -0.0075, -0.1179, ..., -0.0441, 0.0586, 0.0060], + [-0.1194, -0.0833, -0.0179, ..., -0.1107, -0.0102, -0.0208], + [-0.0424, -0.0587, -0.0507, ..., 0.0828, 0.0295, -0.1260]], + device='cuda:0'), grad: tensor([[ 6.3255e-06, 1.8626e-09, -6.3330e-08, ..., 5.7742e-08, + 8.4750e-08, 1.5832e-08], + [ 6.8024e-06, 1.0245e-08, -7.1712e-08, ..., -2.0489e-07, + 2.1141e-07, -9.0338e-08], + [-7.1466e-05, 1.6764e-08, 1.4901e-08, ..., 8.0094e-08, + 1.3132e-07, 2.0489e-08], + ..., + [ 9.1046e-06, 1.7695e-08, 1.2107e-08, ..., 8.8662e-07, + 5.1223e-08, 1.6764e-08], + [ 2.4401e-07, 1.1176e-08, 3.1665e-08, ..., 6.7893e-07, + 5.8673e-07, 8.2888e-08], + [ 1.9819e-06, 9.3132e-09, 2.8871e-08, ..., -4.7088e-06, + -3.3043e-06, 3.9116e-08]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0036, 0.0141, 0.0179, -0.0029, 0.0198, -0.0052, -0.0126, -0.0075, + 0.0168, -0.0116], device='cuda:0'), grad: tensor([ 7.5549e-06, 4.9025e-06, -7.9870e-05, 4.4823e-05, 9.1717e-06, + 7.1414e-06, 9.6858e-07, 1.1832e-05, 2.9188e-06, -9.3654e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 251.90, cls_loss 0.0016 cls_loss_mapping 0.0054 cls_loss_causal 0.5641 re_mapping 0.0057 re_causal 0.0178 /// teacc 99.11 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1279, -0.0206, 0.1403, ..., 0.0428, -0.1231, -0.1107], + [-0.0830, -0.0852, -0.0144, ..., -0.0796, -0.0986, -0.0481], + [ 0.1587, -0.0425, -0.1218, ..., -0.1022, -0.1141, -0.1552], + ..., + [-0.0433, -0.0077, -0.1186, ..., -0.0447, 0.0593, 0.0056], + [-0.1199, -0.0829, -0.0178, ..., -0.1112, -0.0106, -0.0207], + [-0.0425, -0.0590, -0.0510, ..., 0.0827, 0.0295, -0.1274]], + device='cuda:0'), grad: tensor([[ 3.8370e-07, 2.3283e-08, -3.0827e-07, ..., 4.1351e-07, + 4.6007e-07, 2.1420e-08], + [ 6.9384e-07, 6.5193e-08, -1.0245e-08, ..., 1.0123e-06, + 9.9093e-07, 8.3819e-09], + [-1.5693e-06, 1.7323e-07, 3.7253e-08, ..., 4.7311e-07, + 4.2375e-07, 7.4506e-09], + ..., + [ 6.4075e-07, 1.1642e-07, 3.7253e-08, ..., 7.1883e-05, + 7.1228e-05, 9.3132e-09], + [ 5.1502e-07, 5.6811e-08, 4.6566e-08, ..., 7.8380e-06, + 7.8976e-06, 2.1979e-07], + [ 4.0233e-07, 1.2014e-07, 8.0094e-08, ..., -8.7321e-05, + -8.4817e-05, 6.9849e-08]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0037, 0.0142, 0.0178, -0.0031, 0.0200, -0.0052, -0.0124, -0.0075, + 0.0174, -0.0120], device='cuda:0'), grad: tensor([ 2.6934e-06, 1.2508e-06, 7.7579e-07, -4.1798e-06, 1.4447e-05, + 6.2361e-06, 1.0375e-06, 2.9683e-04, 4.2498e-05, -3.6168e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 251.67, cls_loss 0.0017 cls_loss_mapping 0.0056 cls_loss_causal 0.5770 re_mapping 0.0053 re_causal 0.0176 /// teacc 98.99 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.1289, -0.0206, 0.1413, ..., 0.0433, -0.1237, -0.1108], + [-0.0833, -0.0856, -0.0147, ..., -0.0792, -0.0987, -0.0488], + [ 0.1593, -0.0427, -0.1237, ..., -0.1035, -0.1147, -0.1565], + ..., + [-0.0436, -0.0079, -0.1196, ..., -0.0458, 0.0593, 0.0053], + [-0.1205, -0.0836, -0.0186, ..., -0.1120, -0.0117, -0.0216], + [-0.0425, -0.0595, -0.0515, ..., 0.0830, 0.0300, -0.1282]], + device='cuda:0'), grad: tensor([[ 2.0228e-06, 2.7008e-08, 1.6931e-06, ..., 6.5565e-07, + 1.1129e-06, 1.2731e-06], + [ 9.2238e-06, 9.3132e-08, -5.0478e-07, ..., 1.4445e-06, + 7.8883e-07, 6.4168e-07], + [-2.5600e-05, 2.8219e-07, 2.0117e-06, ..., 5.3458e-07, + 8.7824e-07, 1.0934e-06], + ..., + [ 1.1986e-06, 3.7067e-07, 1.1642e-07, ..., 1.0971e-06, + 4.1910e-08, 1.5087e-07], + [ 1.1278e-06, 1.0245e-07, 1.7481e-06, ..., 1.3448e-06, + -1.0896e-06, -8.6054e-07], + [ 2.7753e-07, 3.8184e-08, 2.3004e-07, ..., -4.1723e-06, + 9.8161e-07, 1.3532e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0041, 0.0148, 0.0176, -0.0027, 0.0196, -0.0051, -0.0128, -0.0081, + 0.0169, -0.0117], device='cuda:0'), grad: tensor([ 1.1630e-05, 6.1169e-06, -2.3007e-05, -2.1178e-06, 1.1824e-05, + 1.0304e-05, -1.8746e-05, 6.1914e-06, -4.9248e-06, 2.6803e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 251.41, cls_loss 0.0022 cls_loss_mapping 0.0060 cls_loss_causal 0.5114 re_mapping 0.0058 re_causal 0.0173 /// teacc 99.02 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.1297, -0.0207, 0.1425, ..., 0.0423, -0.1242, -0.1120], + [-0.0834, -0.0862, -0.0143, ..., -0.0797, -0.0991, -0.0483], + [ 0.1601, -0.0428, -0.1259, ..., -0.1049, -0.1151, -0.1586], + ..., + [-0.0438, -0.0076, -0.1217, ..., -0.0465, 0.0596, 0.0051], + [-0.1212, -0.0838, -0.0197, ..., -0.1126, -0.0118, -0.0221], + [-0.0427, -0.0601, -0.0537, ..., 0.0835, 0.0301, -0.1305]], + device='cuda:0'), grad: tensor([[ 6.2305e-07, 1.5460e-07, 1.5832e-08, ..., 5.2154e-07, + 1.5832e-07, 3.2596e-08], + [ 1.3700e-06, 1.4622e-07, -2.2072e-07, ..., -1.9502e-06, + 1.9744e-07, -1.3597e-07], + [-6.1207e-06, 3.4925e-07, 8.5682e-08, ..., 4.0699e-07, + 3.0547e-07, 9.5926e-08], + ..., + [ 1.0971e-06, 2.7101e-07, 9.5926e-08, ..., 8.0280e-07, + -2.0023e-07, 4.4703e-08], + [ 1.4845e-06, 1.4808e-07, 3.5390e-08, ..., 3.0175e-07, + -1.1930e-06, -8.6240e-07], + [ 1.3076e-06, 3.5297e-07, 1.3132e-07, ..., 2.2594e-06, + 7.4226e-07, 6.6124e-08]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0034, 0.0154, 0.0174, -0.0032, 0.0194, -0.0049, -0.0115, -0.0085, + 0.0164, -0.0116], device='cuda:0'), grad: tensor([ 5.8636e-06, -4.1068e-05, -6.7279e-06, -7.0333e-06, -8.8140e-06, + 7.9051e-06, 1.3575e-05, 1.3486e-05, 1.1906e-05, 1.0818e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 251.53, cls_loss 0.0018 cls_loss_mapping 0.0068 cls_loss_causal 0.5631 re_mapping 0.0057 re_causal 0.0179 /// teacc 99.09 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1310, -0.0208, 0.1430, ..., 0.0424, -0.1246, -0.1121], + [-0.0839, -0.0865, -0.0143, ..., -0.0799, -0.0996, -0.0485], + [ 0.1612, -0.0430, -0.1268, ..., -0.1056, -0.1155, -0.1593], + ..., + [-0.0442, -0.0077, -0.1225, ..., -0.0472, 0.0597, 0.0050], + [-0.1231, -0.0840, -0.0178, ..., -0.1147, -0.0121, -0.0214], + [-0.0421, -0.0603, -0.0527, ..., 0.0833, 0.0299, -0.1327]], + device='cuda:0'), grad: tensor([[ 1.1735e-07, 1.6764e-08, 1.0701e-06, ..., 1.6019e-07, + 3.7067e-07, 3.0641e-07], + [ 2.7008e-07, 2.2352e-08, -5.5879e-09, ..., 2.1514e-07, + 2.3004e-07, 3.9116e-08], + [-1.3467e-06, -2.9802e-08, 5.8673e-08, ..., 1.1828e-07, + 1.6578e-07, 3.9116e-08], + ..., + [ 1.8813e-07, 1.5832e-08, 1.2107e-08, ..., 3.3993e-07, + -5.4576e-07, -3.7253e-09], + [ 1.0990e-07, -4.0047e-08, 3.1013e-07, ..., 1.3113e-06, + 4.4983e-07, -2.5146e-08], + [ 6.7055e-08, 1.4901e-08, 5.0291e-08, ..., -6.4038e-06, + -1.9185e-06, 3.7253e-08]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0035, 0.0160, 0.0176, -0.0030, 0.0202, -0.0050, -0.0123, -0.0092, + 0.0168, -0.0120], device='cuda:0'), grad: tensor([ 4.2357e-06, -3.4533e-06, 2.9281e-06, 6.2548e-06, 8.6278e-06, + 2.7809e-06, -1.7807e-06, -1.1392e-05, 3.5912e-06, -1.1846e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 251.26, cls_loss 0.0013 cls_loss_mapping 0.0043 cls_loss_causal 0.5588 re_mapping 0.0056 re_causal 0.0181 /// teacc 98.99 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1316, -0.0208, 0.1434, ..., 0.0427, -0.1250, -0.1121], + [-0.0842, -0.0867, -0.0141, ..., -0.0799, -0.0999, -0.0486], + [ 0.1622, -0.0432, -0.1275, ..., -0.1063, -0.1162, -0.1597], + ..., + [-0.0443, -0.0077, -0.1232, ..., -0.0489, 0.0595, 0.0049], + [-0.1253, -0.0845, -0.0178, ..., -0.1152, -0.0121, -0.0214], + [-0.0421, -0.0604, -0.0533, ..., 0.0834, 0.0301, -0.1328]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 0.0000e+00, -3.9861e-07, ..., -1.7788e-07, + 1.4156e-07, 1.3784e-07], + [ 2.8405e-07, 3.7253e-09, 5.9605e-08, ..., 4.7777e-07, + 2.0210e-07, 9.4064e-08], + [ 1.0990e-06, 3.7253e-09, 8.4750e-07, ..., 1.4948e-06, + 7.3016e-07, 7.7859e-07], + ..., + [ 2.0210e-07, 1.3039e-08, 2.3283e-08, ..., 2.1234e-07, + -2.0396e-07, 2.9802e-08], + [ 1.9558e-08, 1.8626e-09, 1.1083e-07, ..., 8.4750e-08, + 3.3304e-06, 6.3516e-06], + [ 5.4948e-08, 3.7253e-09, 2.1607e-07, ..., -1.5274e-07, + 1.5646e-07, 1.9278e-07]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0036, 0.0161, 0.0179, -0.0029, 0.0204, -0.0051, -0.0123, -0.0095, + 0.0165, -0.0119], device='cuda:0'), grad: tensor([ 1.3132e-07, 1.5507e-06, 7.6257e-06, -7.6462e-07, -3.9265e-06, + -2.3156e-05, -3.1684e-06, 3.6601e-07, 2.0042e-05, 1.2973e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 251.76, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5304 re_mapping 0.0058 re_causal 0.0177 /// teacc 99.04 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1321, -0.0208, 0.1441, ..., 0.0430, -0.1255, -0.1121], + [-0.0853, -0.0869, -0.0141, ..., -0.0808, -0.1004, -0.0488], + [ 0.1646, -0.0433, -0.1286, ..., -0.1070, -0.1170, -0.1610], + ..., + [-0.0456, -0.0078, -0.1239, ..., -0.0488, 0.0599, 0.0050], + [-0.1259, -0.0848, -0.0179, ..., -0.1157, -0.0123, -0.0214], + [-0.0423, -0.0606, -0.0540, ..., 0.0835, 0.0300, -0.1332]], + device='cuda:0'), grad: tensor([[ 2.0117e-07, 4.5635e-08, -1.0598e-06, ..., 5.8226e-06, + 9.4343e-07, 3.9302e-07], + [ 1.3886e-06, 2.7567e-07, 8.3819e-08, ..., 2.0470e-06, + 1.4165e-06, 4.4610e-07], + [ 9.3877e-07, 4.8801e-07, 1.8906e-07, ..., 1.2321e-06, + 1.7025e-06, 6.7707e-07], + ..., + [ 1.9372e-06, 4.8336e-07, 5.2154e-08, ..., 7.1600e-06, + 1.8450e-06, 4.6752e-07], + [ 2.5891e-07, 1.0896e-07, 6.2399e-08, ..., 3.1162e-06, + 3.2987e-06, 2.2911e-06], + [ 1.0226e-06, 2.7567e-07, 7.7207e-07, ..., -3.8505e-05, + -3.4124e-05, 3.8296e-06]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0039, 0.0158, 0.0190, -0.0029, 0.0205, -0.0056, -0.0121, -0.0098, + 0.0167, -0.0121], device='cuda:0'), grad: tensor([ 1.5080e-05, 1.1854e-05, 1.3702e-05, -2.1666e-05, 3.4899e-05, + -2.5451e-05, 1.0818e-05, 1.8910e-05, 1.4521e-05, -7.2479e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 251.63, cls_loss 0.0018 cls_loss_mapping 0.0049 cls_loss_causal 0.5219 re_mapping 0.0055 re_causal 0.0168 /// teacc 99.07 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1333, -0.0209, 0.1452, ..., 0.0434, -0.1266, -0.1123], + [-0.0861, -0.0874, -0.0148, ..., -0.0816, -0.1018, -0.0494], + [ 0.1655, -0.0435, -0.1311, ..., -0.1093, -0.1192, -0.1637], + ..., + [-0.0457, -0.0081, -0.1247, ..., -0.0495, 0.0608, 0.0073], + [-0.1267, -0.0851, -0.0180, ..., -0.1163, -0.0124, -0.0212], + [-0.0425, -0.0609, -0.0542, ..., 0.0837, 0.0304, -0.1348]], + device='cuda:0'), grad: tensor([[ 4.7497e-08, 1.5832e-08, 7.0781e-08, ..., 1.1735e-07, + 1.4808e-07, 9.0338e-08], + [ 2.7381e-07, 9.7789e-08, 2.7008e-08, ..., 4.3306e-07, + 7.7672e-07, 6.2399e-08], + [ 4.3586e-07, 1.4249e-07, 4.4703e-08, ..., 1.5274e-07, + 4.9546e-07, 1.4715e-07], + ..., + [ 1.8813e-07, 6.1467e-08, 0.0000e+00, ..., 2.9895e-07, + -1.0999e-06, 4.3772e-08], + [ 2.7940e-07, 4.4703e-08, 4.3772e-08, ..., 3.5297e-07, + 3.4831e-07, 3.5390e-07], + [ 3.3341e-07, 9.2201e-08, 4.6566e-09, ..., 7.9349e-07, + 6.1002e-07, 1.3504e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0042, 0.0153, 0.0184, -0.0028, 0.0205, -0.0057, -0.0123, -0.0096, + 0.0170, -0.0117], device='cuda:0'), grad: tensor([ 1.0654e-06, -6.8657e-06, 5.0291e-06, -7.5586e-06, -6.9626e-06, + 2.8163e-06, 1.2759e-06, 9.4064e-08, 5.2899e-06, 5.8115e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 251.97, cls_loss 0.0014 cls_loss_mapping 0.0052 cls_loss_causal 0.5501 re_mapping 0.0054 re_causal 0.0177 /// teacc 99.04 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.1346, -0.0210, 0.1459, ..., 0.0436, -0.1281, -0.1126], + [-0.0868, -0.0877, -0.0151, ..., -0.0818, -0.1026, -0.0500], + [ 0.1667, -0.0438, -0.1319, ..., -0.1098, -0.1203, -0.1646], + ..., + [-0.0462, -0.0083, -0.1257, ..., -0.0500, 0.0614, 0.0074], + [-0.1271, -0.0855, -0.0180, ..., -0.1168, -0.0126, -0.0211], + [-0.0427, -0.0614, -0.0544, ..., 0.0839, 0.0304, -0.1353]], + device='cuda:0'), grad: tensor([[ 1.0394e-06, 3.7253e-09, -4.8056e-07, ..., -2.3004e-07, + 6.7055e-08, 7.3109e-07], + [ 2.7753e-07, 2.7940e-09, 5.5879e-08, ..., 4.6566e-08, + 2.0768e-07, 1.6764e-08], + [-1.5246e-06, 4.6566e-09, 7.9162e-08, ..., 6.1467e-08, + 1.7323e-07, 2.3283e-08], + ..., + [ 6.8080e-07, 1.8626e-09, 3.6322e-08, ..., 8.1025e-08, + 5.2154e-08, 2.5146e-08], + [ 1.3225e-07, 6.5193e-09, 7.4506e-08, ..., 1.6764e-07, + -8.5309e-07, 3.7253e-08], + [ 3.3434e-07, 4.6566e-09, 1.4249e-07, ..., -4.4703e-07, + -3.6322e-08, 2.3283e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0043, 0.0152, 0.0186, -0.0029, 0.0201, -0.0054, -0.0121, -0.0097, + 0.0170, -0.0116], device='cuda:0'), grad: tensor([ 4.3884e-06, -4.3839e-05, 3.4094e-05, 6.8396e-06, 1.4575e-06, + -1.0937e-05, 1.1679e-06, 4.2915e-06, -1.0431e-07, 2.6673e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 252.05, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.5173 re_mapping 0.0052 re_causal 0.0167 /// teacc 99.02 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.1348, -0.0211, 0.1462, ..., 0.0438, -0.1290, -0.1128], + [-0.0870, -0.0880, -0.0159, ..., -0.0823, -0.1031, -0.0504], + [ 0.1670, -0.0439, -0.1325, ..., -0.1106, -0.1208, -0.1651], + ..., + [-0.0463, -0.0083, -0.1263, ..., -0.0502, 0.0619, 0.0074], + [-0.1273, -0.0857, -0.0181, ..., -0.1176, -0.0127, -0.0212], + [-0.0429, -0.0616, -0.0549, ..., 0.0837, 0.0301, -0.1356]], + device='cuda:0'), grad: tensor([[ 3.5577e-07, 6.7987e-08, -5.2992e-07, ..., 8.3819e-08, + 3.2689e-07, 1.9558e-07], + [ 3.7160e-07, 7.0781e-08, 4.6566e-08, ..., 2.3022e-06, + 1.3774e-06, -1.5926e-07], + [-2.0079e-06, 1.0058e-07, 4.6566e-08, ..., 8.3819e-07, + 1.1567e-06, 2.3562e-07], + ..., + [ 2.2445e-07, 6.7987e-08, 2.9802e-08, ..., -1.1595e-06, + -2.3693e-06, 6.8918e-08], + [ 1.4510e-06, 3.3062e-07, 6.8918e-08, ..., 5.1036e-07, + 4.2003e-07, 8.5495e-07], + [ 1.3877e-07, 6.5193e-08, 2.0303e-07, ..., 3.5197e-05, + 1.2003e-05, 2.9709e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0043, 0.0151, 0.0183, -0.0030, 0.0204, -0.0054, -0.0120, -0.0095, + 0.0170, -0.0118], device='cuda:0'), grad: tensor([ 2.3097e-06, 4.2059e-06, 2.1346e-06, -3.9674e-06, -6.5267e-05, + -8.2180e-06, 2.2277e-06, -6.7167e-06, 1.1995e-05, 6.1333e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 251.69, cls_loss 0.0013 cls_loss_mapping 0.0050 cls_loss_causal 0.5143 re_mapping 0.0053 re_causal 0.0165 /// teacc 98.99 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1353, -0.0212, 0.1468, ..., 0.0441, -0.1299, -0.1129], + [-0.0882, -0.0883, -0.0170, ..., -0.0832, -0.1039, -0.0508], + [ 0.1682, -0.0445, -0.1342, ..., -0.1116, -0.1216, -0.1659], + ..., + [-0.0467, -0.0078, -0.1273, ..., -0.0503, 0.0625, 0.0070], + [-0.1277, -0.0858, -0.0182, ..., -0.1173, -0.0123, -0.0212], + [-0.0430, -0.0617, -0.0552, ..., 0.0837, 0.0297, -0.1364]], + device='cuda:0'), grad: tensor([[ 8.8755e-07, 8.3819e-09, -1.0869e-06, ..., -4.7032e-07, + 4.8429e-08, 3.2596e-08], + [ 2.1234e-07, 2.4214e-08, 6.7987e-08, ..., 2.9616e-07, + 1.3504e-07, 3.9116e-08], + [-5.7146e-06, 3.7253e-08, 1.6391e-07, ..., 2.0955e-07, + 8.2888e-08, 5.2154e-08], + ..., + [ 1.1036e-06, 3.2596e-08, 2.1607e-07, ..., 5.0850e-07, + 1.6671e-07, 4.9360e-08], + [ 3.1572e-07, 7.6368e-08, 1.7416e-07, ..., 2.6263e-07, + 1.3225e-07, 1.2480e-07], + [ 1.2107e-07, 4.1910e-08, 5.7742e-08, ..., 1.7151e-05, + 7.4059e-06, 9.4995e-08]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0045, 0.0150, 0.0183, -0.0033, 0.0204, -0.0051, -0.0119, -0.0094, + 0.0170, -0.0123], device='cuda:0'), grad: tensor([-6.9570e-07, -7.5437e-07, -5.4240e-06, -1.5587e-05, -3.8415e-05, + 1.8120e-05, 8.8196e-07, 3.0808e-06, 1.6177e-06, 3.7223e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 251.82, cls_loss 0.0015 cls_loss_mapping 0.0047 cls_loss_causal 0.5280 re_mapping 0.0053 re_causal 0.0171 /// teacc 99.04 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1363, -0.0212, 0.1473, ..., 0.0425, -0.1306, -0.1129], + [-0.0906, -0.0885, -0.0174, ..., -0.0834, -0.1067, -0.0510], + [ 0.1723, -0.0446, -0.1351, ..., -0.1114, -0.1228, -0.1674], + ..., + [-0.0492, -0.0078, -0.1317, ..., -0.0513, 0.0647, 0.0084], + [-0.1287, -0.0859, -0.0182, ..., -0.1177, -0.0124, -0.0212], + [-0.0431, -0.0618, -0.0556, ..., 0.0848, 0.0298, -0.1366]], + device='cuda:0'), grad: tensor([[ 6.5938e-07, 6.5193e-09, -1.4789e-06, ..., -1.0701e-06, + 1.0990e-07, 8.7544e-08], + [ 2.8498e-07, 2.4214e-08, 1.8161e-07, ..., 5.5879e-09, + 2.0955e-07, 2.5146e-08], + [-1.0446e-05, 8.3819e-08, 1.4054e-06, ..., 9.8720e-08, + 3.1479e-07, 2.9802e-08], + ..., + [ 1.7323e-07, 2.2352e-08, 9.8720e-08, ..., 1.1455e-07, + -8.2329e-07, 6.5193e-09], + [ 4.3437e-06, 2.0489e-08, -4.2133e-06, ..., 1.2759e-07, + 1.2293e-07, 2.5053e-07], + [ 6.6124e-08, 1.8626e-08, 3.0752e-06, ..., 3.6601e-07, + 1.5367e-07, 9.4995e-08]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0034, 0.0140, 0.0205, -0.0029, 0.0201, -0.0056, -0.0118, -0.0094, + 0.0170, -0.0116], device='cuda:0'), grad: tensor([ 1.3467e-06, 2.8312e-07, -5.0925e-06, 9.1195e-06, 1.3951e-06, + -3.4273e-06, 3.0622e-06, -5.9884e-07, -2.4170e-05, 1.8045e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 250.73, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5518 re_mapping 0.0053 re_causal 0.0168 /// teacc 99.05 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1374, -0.0213, 0.1481, ..., 0.0429, -0.1313, -0.1130], + [-0.0910, -0.0889, -0.0175, ..., -0.0834, -0.1071, -0.0511], + [ 0.1728, -0.0451, -0.1341, ..., -0.1127, -0.1245, -0.1688], + ..., + [-0.0490, -0.0088, -0.1327, ..., -0.0516, 0.0653, 0.0104], + [-0.1297, -0.0862, -0.0183, ..., -0.1182, -0.0127, -0.0213], + [-0.0434, -0.0620, -0.0567, ..., 0.0847, 0.0295, -0.1369]], + device='cuda:0'), grad: tensor([[ 5.7742e-08, 3.2596e-08, -1.3504e-06, ..., -2.2911e-07, + 2.3656e-07, 5.4017e-08], + [ 1.8626e-07, 9.4995e-08, -5.5879e-09, ..., 6.3702e-07, + 7.9814e-07, 4.3772e-08], + [-1.6335e-06, 1.8720e-07, 2.0768e-07, ..., 5.6624e-07, + 6.8173e-07, 7.7300e-08], + ..., + [ 1.0561e-06, 1.1362e-07, 8.2888e-08, ..., 6.5491e-06, + 2.9542e-06, 5.1223e-08], + [ 7.3295e-07, 1.1735e-07, 2.0768e-07, ..., 1.4743e-06, + 1.0356e-06, 8.9407e-08], + [ 7.8231e-08, 1.4249e-07, 2.2724e-07, ..., 2.6450e-06, + 3.0752e-06, 1.0151e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0037, 0.0139, 0.0199, -0.0024, 0.0203, -0.0057, -0.0120, -0.0089, + 0.0168, -0.0119], device='cuda:0'), grad: tensor([-2.7288e-07, 3.9749e-06, -1.8254e-06, -2.8647e-06, -5.5760e-05, + 1.4883e-06, 6.1318e-06, 2.3782e-05, 7.5065e-06, 1.7881e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 249.58, cls_loss 0.0017 cls_loss_mapping 0.0052 cls_loss_causal 0.5293 re_mapping 0.0054 re_causal 0.0166 /// teacc 99.00 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1380, -0.0213, 0.1493, ..., 0.0414, -0.1320, -0.1133], + [-0.0913, -0.0893, -0.0195, ..., -0.0842, -0.1080, -0.0515], + [ 0.1737, -0.0453, -0.1352, ..., -0.1137, -0.1251, -0.1694], + ..., + [-0.0496, -0.0088, -0.1336, ..., -0.0515, 0.0665, 0.0114], + [-0.1306, -0.0864, -0.0184, ..., -0.1188, -0.0127, -0.0213], + [-0.0434, -0.0622, -0.0572, ..., 0.0864, 0.0296, -0.1387]], + device='cuda:0'), grad: tensor([[ 3.5483e-07, 1.8626e-09, -5.5581e-06, ..., -5.6028e-06, + 4.0196e-06, 2.7716e-06], + [ 2.3283e-06, 2.7008e-08, 8.7637e-07, ..., 6.1188e-07, + 1.1418e-06, 9.4436e-07], + [ 3.3658e-06, 4.1910e-08, 1.4361e-06, ..., 7.7579e-07, + 1.7658e-06, 9.7323e-07], + ..., + [ 3.3043e-06, 5.0291e-08, 5.4296e-07, ..., 5.4054e-06, + 2.1141e-06, 5.5041e-07], + [ 2.4624e-06, 5.3085e-08, 3.0138e-06, ..., 7.8976e-07, + 3.5614e-06, 4.2804e-06], + [ 1.9036e-06, 1.0245e-08, 5.3272e-06, ..., -1.1839e-05, + -2.7325e-06, 1.0962e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0027, 0.0137, 0.0199, -0.0022, 0.0194, -0.0061, -0.0117, -0.0085, + 0.0167, -0.0108], device='cuda:0'), grad: tensor([-3.7104e-06, 1.3433e-05, 3.1888e-05, -8.8215e-05, 2.1502e-05, + 2.2066e-04, -2.4056e-04, 3.1024e-05, 2.7001e-05, -1.3053e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 248.50, cls_loss 0.0012 cls_loss_mapping 0.0048 cls_loss_causal 0.5426 re_mapping 0.0050 re_causal 0.0165 /// teacc 99.00 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1389, -0.0214, 0.1515, ..., 0.0418, -0.1327, -0.1134], + [-0.0920, -0.0895, -0.0189, ..., -0.0854, -0.1085, -0.0518], + [ 0.1749, -0.0454, -0.1359, ..., -0.1149, -0.1255, -0.1694], + ..., + [-0.0498, -0.0090, -0.1346, ..., -0.0518, 0.0668, 0.0114], + [-0.1314, -0.0865, -0.0187, ..., -0.1202, -0.0127, -0.0213], + [-0.0440, -0.0624, -0.0608, ..., 0.0866, 0.0296, -0.1393]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 1.2107e-08, 1.8189e-06, ..., -2.2817e-07, + 9.3412e-07, 1.9027e-06], + [ 4.8429e-08, 1.4901e-08, 4.2561e-07, ..., 1.1642e-07, + 3.9302e-07, 2.2817e-07], + [-1.2200e-07, 1.7695e-08, 4.4610e-07, ..., 1.6578e-07, + 4.0140e-07, 2.0768e-07], + ..., + [ 6.4261e-08, 1.5832e-08, 6.3330e-08, ..., 8.1956e-08, + 1.0157e-04, 1.9558e-08], + [ 1.9558e-08, 6.5193e-09, 4.0326e-07, ..., 9.6858e-08, + -1.0389e-04, 1.8254e-07], + [ 5.1223e-08, 4.3772e-08, 5.8673e-07, ..., 4.2189e-07, + 1.4808e-07, 7.9162e-08]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0035, 0.0136, 0.0203, -0.0022, 0.0198, -0.0062, -0.0121, -0.0085, + 0.0165, -0.0110], device='cuda:0'), grad: tensor([ 7.6219e-06, 2.4810e-06, 2.1327e-06, 4.5896e-06, 8.1807e-06, + 3.0808e-06, -2.2426e-05, 3.5906e-04, -3.6645e-04, 1.7621e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 250.22, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5152 re_mapping 0.0053 re_causal 0.0163 /// teacc 99.09 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.1395, -0.0215, 0.1519, ..., 0.0420, -0.1344, -0.1138], + [-0.0921, -0.0899, -0.0196, ..., -0.0839, -0.1072, -0.0520], + [ 0.1754, -0.0456, -0.1367, ..., -0.1162, -0.1264, -0.1697], + ..., + [-0.0500, -0.0091, -0.1355, ..., -0.0534, 0.0677, 0.0113], + [-0.1319, -0.0868, -0.0188, ..., -0.1207, -0.0139, -0.0214], + [-0.0442, -0.0633, -0.0610, ..., 0.0867, 0.0294, -0.1398]], + device='cuda:0'), grad: tensor([[ 7.1712e-08, 3.6322e-08, -2.0768e-06, ..., -1.7080e-06, + 3.7998e-07, 3.0734e-07], + [ 2.2538e-07, 1.7323e-07, 1.0803e-07, ..., 6.8918e-08, + 7.3295e-07, 1.2293e-07], + [-1.8068e-06, 9.1270e-08, 2.0396e-07, ..., 1.2852e-07, + 4.1910e-07, 8.9407e-08], + ..., + [ 1.1465e-06, 2.0955e-07, 1.0524e-07, ..., 1.5181e-07, + -2.2668e-06, 1.1269e-07], + [ 1.7695e-07, 1.1455e-07, 5.4948e-08, ..., 1.2387e-07, + -9.4716e-07, -5.7649e-07], + [ 1.0803e-07, 1.8254e-07, 9.7696e-07, ..., 5.9977e-07, + 9.4902e-07, 1.3039e-07]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0033, 0.0152, 0.0198, -0.0023, 0.0198, -0.0065, -0.0118, -0.0089, + 0.0157, -0.0111], device='cuda:0'), grad: tensor([-1.1064e-06, 3.7495e-06, -9.2015e-07, -6.9179e-06, 9.7230e-07, + 8.0168e-06, 2.6263e-06, -3.6675e-06, -8.8736e-06, 6.0908e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 153---------------------------------------------------- +epoch 153, time 268.49, cls_loss 0.0012 cls_loss_mapping 0.0036 cls_loss_causal 0.5272 re_mapping 0.0049 re_causal 0.0159 /// teacc 99.14 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1400, -0.0222, 0.1516, ..., 0.0422, -0.1356, -0.1146], + [-0.0924, -0.0906, -0.0196, ..., -0.0841, -0.1077, -0.0524], + [ 0.1759, -0.0458, -0.1373, ..., -0.1168, -0.1271, -0.1702], + ..., + [-0.0502, -0.0094, -0.1359, ..., -0.0538, 0.0681, 0.0114], + [-0.1324, -0.0871, -0.0188, ..., -0.1211, -0.0140, -0.0214], + [-0.0444, -0.0636, -0.0618, ..., 0.0870, 0.0298, -0.1403]], + device='cuda:0'), grad: tensor([[ 2.0117e-07, 9.3132e-10, 3.6974e-07, ..., 2.9802e-08, + 7.5437e-08, 1.1828e-07], + [ 7.4040e-07, 2.5146e-08, 2.0992e-06, ..., 2.6450e-07, + 6.1374e-07, 5.6997e-07], + [-2.0456e-04, 1.6764e-08, 3.7160e-07, ..., 1.0617e-07, + 2.8964e-07, 1.0803e-07], + ..., + [ 1.1241e-06, -6.5193e-08, 5.5879e-09, ..., 8.1584e-07, + -5.1409e-07, 5.5879e-09], + [ 3.2224e-07, 1.0245e-08, 1.0710e-07, ..., 8.3819e-08, + 2.1327e-07, 9.9652e-08], + [ 1.9479e-04, 1.6764e-08, 2.1420e-08, ..., -1.0990e-07, + 4.1351e-07, 3.3528e-08]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0030, 0.0153, 0.0196, -0.0024, 0.0194, -0.0067, -0.0114, -0.0090, + 0.0160, -0.0108], device='cuda:0'), grad: tensor([ 1.3076e-06, 5.5432e-06, -3.6979e-04, 5.1856e-06, -4.7125e-06, + 6.9290e-06, -5.3868e-06, 3.0883e-06, 1.7853e-06, 3.5620e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 251.83, cls_loss 0.0014 cls_loss_mapping 0.0043 cls_loss_causal 0.5012 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.06 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1406, -0.0222, 0.1527, ..., 0.0415, -0.1360, -0.1147], + [-0.0930, -0.0914, -0.0198, ..., -0.0845, -0.1087, -0.0531], + [ 0.1772, -0.0461, -0.1380, ..., -0.1176, -0.1278, -0.1697], + ..., + [-0.0504, -0.0094, -0.1366, ..., -0.0549, 0.0685, 0.0116], + [-0.1327, -0.0874, -0.0189, ..., -0.1216, -0.0141, -0.0216], + [-0.0460, -0.0642, -0.0629, ..., 0.0884, 0.0312, -0.1416]], + device='cuda:0'), grad: tensor([[ 5.4855e-07, 1.7881e-07, -1.8273e-06, ..., -1.0794e-06, + 1.4622e-07, 7.2643e-08], + [ 6.2346e-05, 3.0547e-07, 4.2841e-08, ..., 3.7253e-09, + 3.8091e-07, 6.4261e-08], + [-6.5625e-05, 1.7695e-07, 1.1269e-07, ..., 5.8673e-08, + 1.6391e-07, 6.7055e-08], + ..., + [ 7.5996e-07, 5.4482e-07, 2.0489e-08, ..., 2.0489e-08, + -1.0729e-06, -1.7043e-07], + [ 8.1398e-07, 2.8312e-07, 2.0768e-07, ..., 1.0058e-07, + 1.9278e-07, 9.8720e-08], + [ 5.6531e-07, 1.5348e-06, 1.0701e-06, ..., 6.4448e-07, + 1.4985e-06, 2.9057e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0028, 0.0148, 0.0199, -0.0025, 0.0184, -0.0065, -0.0115, -0.0090, + 0.0161, -0.0095], device='cuda:0'), grad: tensor([-4.4610e-07, 1.9944e-04, -2.0671e-04, -9.6112e-06, 1.1437e-06, + 5.0105e-07, 8.1677e-07, 5.9605e-08, 3.7607e-06, 1.0811e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 251.77, cls_loss 0.0019 cls_loss_mapping 0.0051 cls_loss_causal 0.5568 re_mapping 0.0052 re_causal 0.0171 /// teacc 99.03 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1415, -0.0223, 0.1543, ..., 0.0415, -0.1367, -0.1148], + [-0.0940, -0.0924, -0.0239, ..., -0.0859, -0.1105, -0.0557], + [ 0.1788, -0.0479, -0.1394, ..., -0.1186, -0.1297, -0.1704], + ..., + [-0.0514, -0.0078, -0.1372, ..., -0.0556, 0.0693, 0.0125], + [-0.1334, -0.0883, -0.0189, ..., -0.1225, -0.0145, -0.0216], + [-0.0462, -0.0634, -0.0635, ..., 0.0894, 0.0320, -0.1413]], + device='cuda:0'), grad: tensor([[ 2.4587e-07, 2.9895e-07, -1.7509e-07, ..., 1.6121e-06, + 1.1567e-06, 4.7963e-07], + [ 6.1747e-07, 6.0722e-07, 7.4506e-08, ..., 5.6960e-06, + 2.2389e-06, 5.9512e-07], + [ 4.4238e-07, 3.1590e-06, 6.7055e-08, ..., 5.5805e-06, + 1.3456e-05, 2.5164e-06], + ..., + [ 8.7544e-07, -1.5318e-05, 2.5146e-08, ..., 2.8118e-05, + -7.4744e-05, -1.2450e-05], + [ 1.3877e-07, 1.5739e-07, 1.5460e-07, ..., 2.6077e-06, + 1.0189e-06, 3.7625e-07], + [ 5.3830e-07, 7.7672e-07, 4.8708e-07, ..., 1.1787e-03, + 3.4475e-04, 9.5189e-05]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0032, 0.0136, 0.0202, -0.0025, 0.0177, -0.0066, -0.0111, -0.0088, + 0.0160, -0.0088], device='cuda:0'), grad: tensor([ 6.8583e-06, 1.4544e-05, 4.4048e-05, -7.9572e-06, -2.0237e-03, + 1.6928e-05, 4.6715e-06, -1.1408e-04, 6.5267e-06, 2.0504e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 251.55, cls_loss 0.0014 cls_loss_mapping 0.0048 cls_loss_causal 0.5464 re_mapping 0.0049 re_causal 0.0161 /// teacc 98.98 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.1425, -0.0223, 0.1562, ..., 0.0418, -0.1373, -0.1149], + [-0.0946, -0.0937, -0.0249, ..., -0.0866, -0.1110, -0.0559], + [ 0.1798, -0.0481, -0.1416, ..., -0.1202, -0.1302, -0.1698], + ..., + [-0.0520, -0.0086, -0.1381, ..., -0.0560, 0.0697, 0.0112], + [-0.1340, -0.0873, -0.0193, ..., -0.1235, -0.0154, -0.0222], + [-0.0465, -0.0656, -0.0643, ..., 0.0889, 0.0314, -0.1424]], + device='cuda:0'), grad: tensor([[ 2.5984e-07, 3.9209e-07, -5.5041e-07, ..., -3.6694e-07, + 6.5565e-07, 4.9174e-07], + [ 4.4610e-07, 1.6298e-07, 1.3225e-07, ..., 9.2201e-08, + 2.8312e-07, 2.7753e-07], + [-1.7798e-06, 2.2352e-07, 1.8440e-07, ..., 3.4459e-08, + 2.2538e-07, 2.8219e-07], + ..., + [ 3.2783e-07, 8.1025e-08, 4.9360e-08, ..., 1.0245e-07, + 3.6322e-08, 4.1910e-08], + [ 3.4086e-07, 1.3448e-06, 6.0536e-07, ..., 9.2201e-08, + 2.3413e-06, 1.9372e-06], + [ 1.7509e-07, 2.0117e-07, 4.5542e-07, ..., -3.0920e-07, + 4.5169e-07, 3.5111e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0039, 0.0130, 0.0198, -0.0022, 0.0185, -0.0080, -0.0101, -0.0087, + 0.0160, -0.0094], device='cuda:0'), grad: tensor([ 2.1551e-06, 1.9744e-06, -1.2610e-06, 4.6007e-06, 2.0005e-06, + -2.5332e-05, 5.8487e-07, 1.2433e-06, 1.0364e-05, 3.6377e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 251.73, cls_loss 0.0028 cls_loss_mapping 0.0061 cls_loss_causal 0.5370 re_mapping 0.0051 re_causal 0.0165 /// teacc 99.09 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1433, -0.0225, 0.1566, ..., 0.0420, -0.1379, -0.1151], + [-0.0946, -0.0946, -0.0250, ..., -0.0876, -0.1119, -0.0561], + [ 0.1801, -0.0482, -0.1424, ..., -0.1212, -0.1308, -0.1703], + ..., + [-0.0521, -0.0086, -0.1390, ..., -0.0563, 0.0704, 0.0113], + [-0.1346, -0.0875, -0.0194, ..., -0.1238, -0.0156, -0.0224], + [-0.0467, -0.0664, -0.0648, ..., 0.0881, 0.0305, -0.1430]], + device='cuda:0'), grad: tensor([[ 4.9826e-08, -1.6950e-07, -9.5088e-07, ..., -2.9104e-07, + 8.8941e-08, 1.4761e-07], + [ 8.7079e-08, 2.7474e-08, 3.9116e-08, ..., 9.4995e-08, + 5.5879e-08, -2.1476e-06], + [-4.5681e-07, 4.2841e-08, 1.7323e-07, ..., 6.9849e-08, + 6.5193e-08, 9.2201e-08], + ..., + [ 2.1234e-07, 3.3528e-08, 1.1316e-07, ..., 2.9337e-07, + 7.0781e-08, 9.2853e-07], + [ 6.0070e-08, 1.2573e-08, 2.0443e-07, ..., 4.8894e-08, + 2.5611e-08, 8.6147e-08], + [ 7.5717e-07, 6.9803e-07, 1.5507e-07, ..., -5.5879e-07, + -2.8685e-07, 1.4249e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0039, 0.0119, 0.0191, -0.0021, 0.0196, -0.0085, -0.0097, -0.0074, + 0.0160, -0.0104], device='cuda:0'), grad: tensor([-6.9663e-07, -2.8655e-05, 8.8196e-07, -2.3302e-06, 4.9407e-07, + 1.4514e-05, -1.8729e-06, 1.3649e-05, 1.0505e-06, 2.9169e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 251.51, cls_loss 0.0012 cls_loss_mapping 0.0036 cls_loss_causal 0.5242 re_mapping 0.0052 re_causal 0.0166 /// teacc 98.95 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.1442, -0.0230, 0.1567, ..., 0.0420, -0.1384, -0.1153], + [-0.0949, -0.0953, -0.0248, ..., -0.0879, -0.1124, -0.0560], + [ 0.1818, -0.0484, -0.1429, ..., -0.1216, -0.1311, -0.1713], + ..., + [-0.0525, -0.0087, -0.1403, ..., -0.0568, 0.0707, 0.0112], + [-0.1372, -0.0855, -0.0196, ..., -0.1242, -0.0150, -0.0221], + [-0.0482, -0.0695, -0.0652, ..., 0.0880, 0.0297, -0.1458]], + device='cuda:0'), grad: tensor([[ 2.6403e-07, 4.6566e-10, -3.5577e-07, ..., 6.4354e-07, + 2.8266e-07, 9.2201e-08], + [ 1.9558e-08, 1.2107e-08, 8.8476e-09, ..., 1.1846e-06, + 6.2073e-07, 3.2131e-08], + [ 3.6741e-07, 1.8626e-09, 8.5682e-08, ..., 1.9539e-06, + 6.5379e-07, 4.2841e-08], + ..., + [ 1.1688e-07, 2.3982e-07, 2.0489e-08, ..., 2.6897e-06, + 2.0042e-06, 1.0896e-07], + [ 4.1910e-08, 9.3132e-10, 5.9139e-08, ..., 5.5740e-07, + 3.1013e-07, 1.5460e-07], + [-2.1011e-06, 4.1910e-09, 2.4820e-07, ..., -1.2200e-06, + 1.0384e-06, 8.9407e-08]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0038, 0.0117, 0.0195, -0.0025, 0.0198, -0.0078, -0.0096, -0.0071, + 0.0163, -0.0114], device='cuda:0'), grad: tensor([ 2.8461e-06, 4.1872e-06, 9.0525e-06, -9.9018e-06, -2.8268e-05, + -4.5588e-07, 6.6981e-06, 1.9476e-05, -6.5984e-07, -2.9560e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 251.70, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.5220 re_mapping 0.0053 re_causal 0.0180 /// teacc 99.04 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1446, -0.0232, 0.1571, ..., 0.0421, -0.1388, -0.1154], + [-0.0953, -0.0955, -0.0247, ..., -0.0895, -0.1127, -0.0561], + [ 0.1822, -0.0485, -0.1434, ..., -0.1226, -0.1315, -0.1724], + ..., + [-0.0525, -0.0089, -0.1413, ..., -0.0571, 0.0715, 0.0115], + [-0.1374, -0.0856, -0.0196, ..., -0.1244, -0.0151, -0.0221], + [-0.0489, -0.0696, -0.0659, ..., 0.0878, 0.0294, -0.1462]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 2.7940e-09, -7.2177e-08, ..., 6.9384e-08, + 1.6391e-07, 2.6543e-08], + [ 3.5856e-08, 9.3132e-09, 2.3283e-09, ..., 3.1199e-08, + 3.4086e-07, 4.4703e-08], + [ 9.3132e-08, 1.8161e-08, 1.2573e-08, ..., 1.9558e-08, + 3.8650e-08, 2.0489e-08], + ..., + [ 4.1444e-08, 1.2107e-08, 4.1910e-09, ..., -2.3097e-07, + -7.2364e-07, -8.7544e-08], + [ 1.0710e-08, 2.7940e-09, 1.2107e-08, ..., 8.8476e-09, + -4.5635e-08, -1.3644e-07], + [ 1.6624e-07, 6.1467e-08, 4.7032e-08, ..., 1.1129e-07, + 1.6205e-07, 1.1548e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0039, 0.0116, 0.0192, -0.0028, 0.0210, -0.0074, -0.0097, -0.0071, + 0.0163, -0.0120], device='cuda:0'), grad: tensor([ 1.3337e-06, 8.1956e-07, 6.6450e-07, -9.7696e-07, -2.4680e-08, + 7.9768e-07, 1.3364e-07, -3.6284e-06, -2.7530e-06, 3.6340e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 251.77, cls_loss 0.0012 cls_loss_mapping 0.0033 cls_loss_causal 0.5139 re_mapping 0.0049 re_causal 0.0161 /// teacc 98.96 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1451, -0.0234, 0.1577, ..., 0.0423, -0.1399, -0.1156], + [-0.0955, -0.0958, -0.0248, ..., -0.0898, -0.1132, -0.0560], + [ 0.1829, -0.0486, -0.1442, ..., -0.1237, -0.1319, -0.1729], + ..., + [-0.0531, -0.0090, -0.1419, ..., -0.0573, 0.0721, 0.0110], + [-0.1377, -0.0858, -0.0196, ..., -0.1248, -0.0150, -0.0221], + [-0.0492, -0.0698, -0.0664, ..., 0.0877, 0.0292, -0.1464]], + device='cuda:0'), grad: tensor([[ 2.4447e-07, 7.8697e-08, -1.5832e-08, ..., 5.3551e-08, + 1.7881e-07, 1.4016e-07], + [ 6.3749e-07, 7.7300e-08, 2.2817e-08, ..., 3.7253e-07, + 2.7381e-07, 3.8184e-08], + [-1.3262e-05, 4.2235e-07, 2.7940e-08, ..., 2.3050e-07, + 3.4133e-07, 2.7940e-08], + ..., + [ 1.3299e-05, 1.7136e-07, 4.0513e-08, ..., 5.1968e-07, + -3.1013e-07, 2.7008e-08], + [ 1.7928e-07, 2.6450e-07, 2.0722e-07, ..., 4.5449e-07, + -8.9873e-08, -1.0803e-07], + [ 1.2806e-07, 9.4995e-08, 1.6950e-07, ..., 3.0752e-06, + 9.8627e-07, 1.6997e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0039, 0.0114, 0.0192, -0.0025, 0.0214, -0.0075, -0.0100, -0.0070, + 0.0164, -0.0123], device='cuda:0'), grad: tensor([ 1.5171e-06, 1.9465e-06, -1.9729e-05, 1.5870e-06, -1.2122e-05, + -7.1190e-06, 2.1905e-06, 2.4691e-05, -1.7462e-06, 8.7172e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 161---------------------------------------------------- +epoch 161, time 268.22, cls_loss 0.0013 cls_loss_mapping 0.0040 cls_loss_causal 0.5487 re_mapping 0.0050 re_causal 0.0160 /// teacc 99.19 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1457, -0.0235, 0.1583, ..., 0.0426, -0.1406, -0.1158], + [-0.0958, -0.0962, -0.0252, ..., -0.0902, -0.1135, -0.0561], + [ 0.1843, -0.0488, -0.1451, ..., -0.1246, -0.1325, -0.1740], + ..., + [-0.0536, -0.0091, -0.1438, ..., -0.0580, 0.0723, 0.0119], + [-0.1381, -0.0860, -0.0197, ..., -0.1253, -0.0151, -0.0221], + [-0.0514, -0.0699, -0.0675, ..., 0.0872, 0.0291, -0.1465]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 7.4506e-09, -1.3504e-06, ..., -5.8534e-07, + 1.2387e-07, 1.1502e-07], + [ 3.2596e-08, 1.3970e-08, 4.6100e-08, ..., -2.4680e-08, + 2.9337e-07, -1.0710e-08], + [ 1.0245e-08, 2.9802e-08, 1.1967e-07, ..., 6.3796e-08, + 1.1316e-07, 6.5193e-08], + ..., + [ 7.9628e-08, 2.7008e-08, 4.9826e-08, ..., 1.6158e-07, + -1.1362e-06, 2.7008e-08], + [ 7.8231e-08, 7.2177e-08, 2.8173e-07, ..., 2.1420e-07, + 3.0827e-07, 3.7951e-07], + [ 3.0734e-08, 2.8405e-08, 9.7882e-07, ..., -1.3970e-09, + 4.7684e-07, 3.4459e-08]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0041, 0.0115, 0.0192, -0.0036, 0.0219, -0.0064, -0.0100, -0.0070, + 0.0165, -0.0131], device='cuda:0'), grad: tensor([-1.4044e-06, -4.7032e-07, 9.2620e-07, -2.2743e-06, 6.0489e-07, + 2.6207e-06, -3.6042e-06, -3.1181e-06, 3.1590e-06, 3.5521e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 251.81, cls_loss 0.0010 cls_loss_mapping 0.0040 cls_loss_causal 0.5267 re_mapping 0.0051 re_causal 0.0164 /// teacc 99.15 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1466, -0.0237, 0.1583, ..., 0.0426, -0.1411, -0.1162], + [-0.0962, -0.0966, -0.0247, ..., -0.0898, -0.1137, -0.0558], + [ 0.1848, -0.0487, -0.1456, ..., -0.1250, -0.1328, -0.1744], + ..., + [-0.0539, -0.0094, -0.1450, ..., -0.0583, 0.0726, 0.0120], + [-0.1384, -0.0861, -0.0197, ..., -0.1261, -0.0152, -0.0221], + [-0.0515, -0.0700, -0.0680, ..., 0.0873, 0.0292, -0.1468]], + device='cuda:0'), grad: tensor([[ 9.3132e-08, 1.3504e-08, -1.0291e-07, ..., -8.3353e-08, + 1.5926e-07, 7.8697e-08], + [ 2.0443e-07, 6.2399e-08, -4.2375e-08, ..., 3.4459e-07, + 1.3039e-07, 1.7695e-08], + [-7.0082e-07, 8.3819e-08, 1.8673e-07, ..., 3.4785e-07, + 1.4482e-07, 2.6543e-08], + ..., + [ 2.7474e-07, 6.6124e-08, 1.4482e-07, ..., 3.5623e-07, + -3.4226e-07, 1.5832e-08], + [ 4.9593e-07, 2.9802e-08, 1.5739e-07, ..., 7.9442e-07, + 2.0722e-07, 1.3737e-07], + [-3.6452e-06, 2.3749e-08, 6.3330e-08, ..., -1.4976e-05, + -8.8941e-08, 2.1746e-07]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0039, 0.0117, 0.0192, -0.0034, 0.0218, -0.0069, -0.0096, -0.0072, + 0.0164, -0.0132], device='cuda:0'), grad: tensor([ 7.6136e-07, 1.7118e-06, 3.0696e-06, 8.0233e-07, 1.6838e-05, + 1.9884e-04, 9.6112e-07, 1.6000e-06, 7.1563e-06, -2.3139e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 251.36, cls_loss 0.0011 cls_loss_mapping 0.0039 cls_loss_causal 0.5292 re_mapping 0.0052 re_causal 0.0166 /// teacc 99.06 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1473, -0.0240, 0.1588, ..., 0.0428, -0.1420, -0.1163], + [-0.0963, -0.0970, -0.0247, ..., -0.0899, -0.1141, -0.0558], + [ 0.1853, -0.0488, -0.1462, ..., -0.1256, -0.1332, -0.1747], + ..., + [-0.0542, -0.0096, -0.1463, ..., -0.0588, 0.0729, 0.0120], + [-0.1386, -0.0863, -0.0210, ..., -0.1265, -0.0160, -0.0224], + [-0.0517, -0.0700, -0.0692, ..., 0.0871, 0.0291, -0.1468]], + device='cuda:0'), grad: tensor([[ 1.5879e-07, 5.4948e-08, 3.5949e-06, ..., -1.3039e-08, + 9.9372e-07, 6.2538e-07], + [ 9.8255e-08, 2.9802e-08, 3.7439e-07, ..., 4.0513e-08, + 2.2305e-07, 2.1746e-07], + [ 1.8347e-07, 1.6764e-07, 6.9756e-07, ..., 3.9581e-08, + 2.1374e-07, 1.8813e-07], + ..., + [ 6.7521e-07, 2.4447e-07, 9.3132e-08, ..., 7.4040e-08, + -4.9779e-07, -2.0955e-08], + [ 1.5274e-07, 1.9092e-08, 5.9716e-06, ..., 8.4750e-08, + 1.6624e-06, 4.4703e-07], + [ 3.8650e-08, 1.3039e-08, 1.1595e-07, ..., -2.5192e-07, + -6.9849e-09, 3.4459e-08]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0040, 0.0113, 0.0189, -0.0034, 0.0221, -0.0071, -0.0084, -0.0067, + 0.0153, -0.0135], device='cuda:0'), grad: tensor([ 7.4431e-06, 3.2812e-05, 3.2969e-06, 7.3481e-07, 3.2950e-06, + -6.2168e-05, -1.9833e-05, 3.2242e-06, 2.4870e-05, 6.1877e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 251.27, cls_loss 0.0015 cls_loss_mapping 0.0047 cls_loss_causal 0.5421 re_mapping 0.0049 re_causal 0.0156 /// teacc 99.15 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1495, -0.0241, 0.1606, ..., 0.0410, -0.1430, -0.1165], + [-0.0963, -0.0976, -0.0249, ..., -0.0896, -0.1145, -0.0560], + [ 0.1865, -0.0476, -0.1472, ..., -0.1263, -0.1337, -0.1749], + ..., + [-0.0545, -0.0100, -0.1480, ..., -0.0595, 0.0736, 0.0120], + [-0.1390, -0.0866, -0.0212, ..., -0.1272, -0.0163, -0.0226], + [-0.0522, -0.0702, -0.0722, ..., 0.0886, 0.0287, -0.1468]], + device='cuda:0'), grad: tensor([[ 4.2841e-08, 2.7940e-09, -9.4250e-07, ..., -5.2806e-07, + 4.2375e-08, 1.2573e-08], + [ 1.1129e-07, 2.7940e-09, -7.4040e-08, ..., 1.0850e-07, + 4.2841e-08, 4.1910e-09], + [-7.7719e-07, 1.3970e-09, 2.9802e-08, ..., 8.0559e-08, + 4.2375e-08, 5.1223e-09], + ..., + [ 1.2573e-07, 5.5879e-09, 7.3109e-08, ..., 2.0536e-07, + -4.0978e-08, 9.3132e-10], + [ 8.6613e-08, 2.7940e-09, 6.2864e-08, ..., 6.6590e-08, + -7.2177e-08, -4.6566e-10], + [ 2.6077e-08, 9.3132e-10, 6.7614e-07, ..., -1.5460e-07, + -1.2107e-08, 3.2596e-09]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0024, 0.0114, 0.0195, -0.0024, 0.0223, -0.0082, -0.0085, -0.0067, + 0.0151, -0.0127], device='cuda:0'), grad: tensor([-1.0580e-06, -1.0263e-06, -5.5879e-07, 2.7381e-07, -5.9372e-07, + 1.1995e-06, 6.5332e-07, 1.3625e-06, -1.8626e-06, 1.5954e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 251.69, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.5375 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.19 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1491, -0.0236, 0.1619, ..., 0.0412, -0.1435, -0.1166], + [-0.0977, -0.0980, -0.0256, ..., -0.0898, -0.1148, -0.0561], + [ 0.1876, -0.0477, -0.1483, ..., -0.1276, -0.1346, -0.1757], + ..., + [-0.0546, -0.0102, -0.1492, ..., -0.0597, 0.0744, 0.0125], + [-0.1394, -0.0867, -0.0213, ..., -0.1277, -0.0164, -0.0226], + [-0.0528, -0.0705, -0.0729, ..., 0.0885, 0.0285, -0.1469]], + device='cuda:0'), grad: tensor([[ 2.5658e-07, 1.0710e-08, -5.4482e-08, ..., -4.8429e-08, + 2.6543e-08, 2.2352e-08], + [ 8.2888e-08, 1.1642e-08, -8.1025e-08, ..., -3.6322e-08, + 7.3574e-08, 2.7940e-09], + [-9.2480e-07, 3.5390e-08, 1.8161e-08, ..., 9.3132e-09, + 5.7742e-08, 4.1910e-09], + ..., + [ 6.0443e-07, 2.1420e-08, 3.3993e-08, ..., 2.7474e-08, + -3.7486e-07, 0.0000e+00], + [ 7.4971e-08, 1.8161e-08, 2.3283e-08, ..., 9.3132e-09, + 2.4680e-08, 9.7789e-09], + [ 1.3690e-07, 8.1491e-08, 5.5879e-08, ..., 4.2561e-07, + 2.5285e-07, 1.8626e-09]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0027, 0.0116, 0.0189, -0.0026, 0.0224, -0.0082, -0.0088, -0.0067, + 0.0152, -0.0131], device='cuda:0'), grad: tensor([ 7.0268e-07, -2.3041e-06, -1.1921e-06, -1.2778e-06, -7.4040e-08, + 1.7369e-07, 3.9535e-07, 7.7439e-07, 6.1700e-07, 2.1830e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 251.88, cls_loss 0.0009 cls_loss_mapping 0.0032 cls_loss_causal 0.5277 re_mapping 0.0048 re_causal 0.0158 /// teacc 99.06 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1493, -0.0237, 0.1623, ..., 0.0412, -0.1441, -0.1166], + [-0.0981, -0.0985, -0.0256, ..., -0.0899, -0.1151, -0.0562], + [ 0.1881, -0.0481, -0.1488, ..., -0.1280, -0.1349, -0.1757], + ..., + [-0.0547, -0.0104, -0.1499, ..., -0.0608, 0.0744, 0.0125], + [-0.1403, -0.0868, -0.0213, ..., -0.1282, -0.0164, -0.0226], + [-0.0531, -0.0706, -0.0731, ..., 0.0884, 0.0286, -0.1469]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 9.7789e-09, 2.7940e-09, ..., 7.7300e-08, + 6.1002e-08, 1.3970e-08], + [ 5.7742e-08, 5.1223e-09, 1.3970e-09, ..., 3.7951e-07, + 4.6846e-07, 1.0245e-08], + [-1.0990e-07, 9.3132e-09, 1.8626e-09, ..., 7.7765e-08, + 1.2992e-07, 2.1886e-08], + ..., + [ 4.3772e-08, 3.3528e-08, 0.0000e+00, ..., 3.2969e-07, + -2.0768e-06, 1.5832e-08], + [ 6.4261e-08, 8.8476e-08, 9.3132e-10, ..., 9.2108e-07, + 2.7521e-07, -1.1269e-07], + [-9.4529e-08, 1.6764e-08, 0.0000e+00, ..., -2.9728e-06, + 9.1223e-07, 1.0245e-08]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0027, 0.0118, 0.0185, -0.0028, 0.0227, -0.0078, -0.0091, -0.0068, + 0.0151, -0.0132], device='cuda:0'), grad: tensor([ 4.9965e-07, 2.4512e-06, 9.0152e-07, 2.8349e-06, 7.9535e-07, + -2.9840e-06, 1.0682e-06, -4.3996e-06, 1.2927e-06, -2.4531e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 251.70, cls_loss 0.0011 cls_loss_mapping 0.0049 cls_loss_causal 0.5312 re_mapping 0.0051 re_causal 0.0160 /// teacc 99.07 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1505, -0.0241, 0.1627, ..., 0.0413, -0.1449, -0.1167], + [-0.0986, -0.0991, -0.0256, ..., -0.0910, -0.1161, -0.0563], + [ 0.1887, -0.0483, -0.1492, ..., -0.1285, -0.1355, -0.1761], + ..., + [-0.0551, -0.0109, -0.1516, ..., -0.0620, 0.0752, 0.0128], + [-0.1407, -0.0873, -0.0220, ..., -0.1291, -0.0174, -0.0231], + [-0.0532, -0.0708, -0.0735, ..., 0.0886, 0.0287, -0.1469]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 0.0000e+00, 3.2987e-06, ..., 7.9945e-06, + 9.2341e-07, 1.5832e-08], + [ 1.5367e-08, 0.0000e+00, -5.4482e-08, ..., 3.7067e-07, + 8.8941e-08, 7.9162e-09], + [-3.2596e-07, 1.3970e-09, 3.8836e-07, ..., 2.0815e-07, + 2.3190e-07, 2.7940e-09], + ..., + [ 2.4354e-07, 9.3132e-10, 1.2107e-08, ..., 3.8231e-07, + -3.3975e-06, 1.8626e-09], + [ 1.8161e-08, 0.0000e+00, 4.7535e-06, ..., 5.4482e-07, + 1.2927e-06, 1.8626e-09], + [ 4.6566e-09, 0.0000e+00, 9.8255e-08, ..., 9.0227e-06, + 1.3737e-07, 1.8626e-09]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0027, 0.0118, 0.0184, -0.0022, 0.0228, -0.0079, -0.0087, -0.0068, + 0.0146, -0.0132], device='cuda:0'), grad: tensor([ 2.2605e-05, -2.5344e-04, 3.1665e-06, 6.5923e-05, -3.9369e-05, + 2.5276e-06, -1.6034e-05, 1.8311e-04, 1.1489e-05, 1.9699e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 251.83, cls_loss 0.0011 cls_loss_mapping 0.0046 cls_loss_causal 0.5317 re_mapping 0.0053 re_causal 0.0170 /// teacc 99.05 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1508, -0.0246, 0.1659, ..., 0.0417, -0.1454, -0.1166], + [-0.1012, -0.0999, -0.0289, ..., -0.0925, -0.1169, -0.0577], + [ 0.1907, -0.0494, -0.1501, ..., -0.1292, -0.1357, -0.1764], + ..., + [-0.0554, -0.0112, -0.1523, ..., -0.0634, 0.0752, 0.0128], + [-0.1408, -0.0884, -0.0223, ..., -0.1303, -0.0177, -0.0232], + [-0.0534, -0.0711, -0.0747, ..., 0.0888, 0.0289, -0.1470]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 5.5879e-09, -5.7742e-08, ..., 1.3048e-06, + 9.3132e-08, 8.3819e-09], + [ 9.3132e-09, 1.0245e-08, 2.4214e-08, ..., 6.7987e-07, + 1.5181e-07, 6.5193e-09], + [ 4.8429e-08, 5.2154e-08, 2.7008e-08, ..., 6.0722e-07, + 1.2852e-07, 7.4506e-09], + ..., + [ 3.7253e-09, 5.5879e-09, 4.2841e-08, ..., 1.4324e-06, + 3.1665e-07, 1.3039e-08], + [ 1.4249e-07, 1.5646e-07, 2.7008e-08, ..., 9.8348e-07, + 2.1141e-07, 5.5879e-08], + [ 1.8626e-09, 1.8626e-09, 1.1642e-07, ..., 5.8115e-07, + 4.3120e-07, 2.0489e-08]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0038, 0.0115, 0.0191, -0.0013, 0.0225, -0.0080, -0.0088, -0.0071, + 0.0143, -0.0130], device='cuda:0'), grad: tensor([ 2.8946e-06, 2.0154e-06, 3.0305e-06, -8.0373e-07, -2.6599e-05, + -7.1339e-07, 1.1951e-05, 1.4752e-06, 3.7514e-06, 3.0193e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 251.58, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.4971 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.12 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1513, -0.0251, 0.1662, ..., 0.0418, -0.1470, -0.1174], + [-0.1014, -0.1004, -0.0289, ..., -0.0927, -0.1174, -0.0579], + [ 0.1911, -0.0497, -0.1515, ..., -0.1301, -0.1360, -0.1768], + ..., + [-0.0555, -0.0113, -0.1542, ..., -0.0639, 0.0756, 0.0127], + [-0.1408, -0.0886, -0.0224, ..., -0.1308, -0.0179, -0.0233], + [-0.0537, -0.0713, -0.0751, ..., 0.0889, 0.0289, -0.1471]], + device='cuda:0'), grad: tensor([[ 2.4494e-07, 8.1956e-08, -2.4308e-07, ..., -1.5832e-08, + 3.2596e-08, 3.5390e-08], + [ 1.7192e-06, 3.6322e-08, 3.0734e-08, ..., 6.7893e-07, + 9.9652e-08, 2.7008e-08], + [-4.0643e-06, 6.1467e-08, 4.0978e-08, ..., 8.3819e-08, + 6.0536e-08, 2.0489e-08], + ..., + [ 6.5286e-07, 7.4506e-08, 3.5390e-08, ..., 7.2271e-07, + -2.6077e-08, 5.6811e-08], + [ 8.2608e-07, 8.1956e-08, 7.8231e-08, ..., 3.3155e-07, + 4.6566e-09, 4.4703e-08], + [ 1.7509e-07, 1.4622e-07, 1.2480e-07, ..., -3.6042e-07, + -2.0955e-07, 1.6764e-08]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0038, 0.0115, 0.0185, -0.0013, 0.0222, -0.0082, -0.0084, -0.0071, + 0.0145, -0.0130], device='cuda:0'), grad: tensor([ 5.4296e-07, 5.5134e-06, -6.0983e-06, -1.1176e-07, -5.4576e-06, + -3.5204e-07, 1.2554e-06, 2.6636e-06, 1.4277e-06, 6.2492e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 251.56, cls_loss 0.0012 cls_loss_mapping 0.0041 cls_loss_causal 0.5399 re_mapping 0.0051 re_causal 0.0157 /// teacc 98.97 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1519, -0.0260, 0.1664, ..., 0.0418, -0.1485, -0.1179], + [-0.1017, -0.1012, -0.0289, ..., -0.0920, -0.1179, -0.0579], + [ 0.1937, -0.0498, -0.1523, ..., -0.1292, -0.1346, -0.1772], + ..., + [-0.0559, -0.0111, -0.1557, ..., -0.0648, 0.0760, 0.0129], + [-0.1417, -0.0889, -0.0225, ..., -0.1320, -0.0181, -0.0234], + [-0.0536, -0.0715, -0.0755, ..., 0.0892, 0.0292, -0.1472]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 0.0000e+00, 3.8184e-08, ..., 3.2596e-08, + 1.6205e-07, 9.6858e-08], + [ 4.6566e-09, 2.7940e-09, 1.8626e-09, ..., 1.4808e-07, + 2.5984e-07, 3.0734e-08], + [ 1.7695e-08, 1.8626e-09, 1.0245e-08, ..., 7.7300e-08, + 1.1083e-07, 2.0489e-08], + ..., + [ 3.1665e-08, 2.7940e-09, 5.5879e-09, ..., 3.5577e-07, + -4.3958e-07, 1.3970e-08], + [ 2.1420e-08, 1.2107e-08, 2.7940e-08, ..., 1.7881e-07, + 3.6787e-07, 1.7323e-07], + [ 1.3784e-07, 9.3132e-10, 5.5879e-09, ..., 1.9744e-07, + 1.4156e-07, 7.4506e-09]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0038, 0.0116, 0.0197, -0.0017, 0.0215, -0.0082, -0.0088, -0.0071, + 0.0144, -0.0127], device='cuda:0'), grad: tensor([ 1.3709e-06, -4.4018e-05, 5.9828e-06, 2.5760e-06, 2.0787e-06, + -4.9360e-06, 2.1793e-06, -4.6194e-07, 3.3557e-05, 1.6587e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 251.58, cls_loss 0.0012 cls_loss_mapping 0.0044 cls_loss_causal 0.5582 re_mapping 0.0047 re_causal 0.0149 /// teacc 99.03 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1508, -0.0264, 0.1672, ..., 0.0420, -0.1494, -0.1181], + [-0.1021, -0.1020, -0.0289, ..., -0.0922, -0.1183, -0.0576], + [ 0.1945, -0.0499, -0.1550, ..., -0.1308, -0.1343, -0.1777], + ..., + [-0.0555, -0.0109, -0.1563, ..., -0.0651, 0.0765, 0.0129], + [-0.1427, -0.0894, -0.0226, ..., -0.1327, -0.0183, -0.0236], + [-0.0540, -0.0719, -0.0762, ..., 0.0891, 0.0294, -0.1473]], + device='cuda:0'), grad: tensor([[ 6.0536e-08, 3.8184e-08, 1.4473e-06, ..., -7.0781e-08, + 6.5286e-07, 9.7416e-07], + [ 3.8184e-08, 1.5832e-08, 6.2957e-07, ..., 2.3097e-06, + 3.7439e-06, 4.6939e-07], + [-4.5635e-08, 1.2107e-08, 1.1362e-07, ..., 9.4995e-08, + 1.3039e-07, 7.3574e-08], + ..., + [ 7.7300e-08, 1.3970e-08, 4.4703e-08, ..., -3.4198e-06, + -5.3495e-06, 3.1665e-08], + [ 3.6322e-08, 2.6915e-07, 4.1611e-06, ..., 1.2200e-07, + 2.4345e-06, 3.7309e-06], + [-5.4017e-08, 9.6858e-08, 2.7567e-07, ..., 3.6322e-07, + 1.4594e-06, 1.7043e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0040, 0.0117, 0.0198, -0.0020, 0.0216, -0.0082, -0.0089, -0.0070, + 0.0142, -0.0130], device='cuda:0'), grad: tensor([ 4.0568e-06, 2.3931e-05, 1.3141e-06, 4.8056e-06, 2.4512e-06, + -9.1344e-06, -1.5855e-05, -3.8207e-05, 1.5154e-05, 1.1533e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 251.50, cls_loss 0.0011 cls_loss_mapping 0.0030 cls_loss_causal 0.5185 re_mapping 0.0046 re_causal 0.0148 /// teacc 99.12 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1512, -0.0269, 0.1677, ..., 0.0421, -0.1503, -0.1183], + [-0.1026, -0.1034, -0.0292, ..., -0.0927, -0.1186, -0.0580], + [ 0.1952, -0.0501, -0.1564, ..., -0.1316, -0.1346, -0.1771], + ..., + [-0.0559, -0.0118, -0.1569, ..., -0.0661, 0.0765, 0.0130], + [-0.1431, -0.0898, -0.0226, ..., -0.1335, -0.0185, -0.0237], + [-0.0542, -0.0719, -0.0765, ..., 0.0889, 0.0295, -0.1473]], + device='cuda:0'), grad: tensor([[ 1.9558e-08, 5.5879e-09, -2.8908e-05, ..., -1.4998e-05, + 7.8231e-08, 6.5193e-09], + [ 4.8429e-08, 1.0245e-08, 8.1584e-07, ..., 5.1130e-07, + 9.3970e-07, -4.3772e-08], + [ 5.9232e-07, 1.4901e-07, 3.1032e-06, ..., 1.6456e-06, + 5.7649e-07, 5.5879e-09], + ..., + [ 1.1176e-08, 2.6077e-08, 1.9893e-06, ..., 1.0980e-06, + -4.1164e-06, 8.3819e-09], + [ 2.3283e-08, 5.5879e-09, 1.2815e-06, ..., 6.8825e-07, + 2.8871e-08, 2.2352e-08], + [ 1.0245e-08, 2.7940e-09, 1.3120e-05, ..., 6.7465e-06, + 4.4797e-07, 5.5879e-09]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0041, 0.0117, 0.0200, -0.0018, 0.0221, -0.0077, -0.0093, -0.0071, + 0.0141, -0.0133], device='cuda:0'), grad: tensor([-6.2287e-05, 4.7013e-06, 1.5736e-05, 1.4320e-05, 1.6717e-06, + 5.2936e-06, 1.0833e-05, -2.0608e-05, -3.5092e-06, 3.3915e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 251.06, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.5372 re_mapping 0.0047 re_causal 0.0153 /// teacc 99.05 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1518, -0.0274, 0.1685, ..., 0.0422, -0.1518, -0.1185], + [-0.1028, -0.1044, -0.0298, ..., -0.0936, -0.1190, -0.0579], + [ 0.1954, -0.0503, -0.1572, ..., -0.1323, -0.1349, -0.1777], + ..., + [-0.0561, -0.0124, -0.1583, ..., -0.0671, 0.0765, 0.0131], + [-0.1434, -0.0902, -0.0228, ..., -0.1342, -0.0187, -0.0238], + [-0.0526, -0.0719, -0.0770, ..., 0.0891, 0.0293, -0.1474]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, -1.4063e-07, -1.5525e-06, ..., -3.4366e-07, + 8.1025e-08, 8.6613e-08], + [ 3.1665e-08, 1.2107e-08, -2.6077e-08, ..., 1.3039e-08, + 1.0245e-07, 5.9605e-08], + [-3.0734e-08, 4.0978e-08, 1.6205e-07, ..., 5.1223e-08, + 5.9605e-08, 6.7055e-08], + ..., + [ 7.1712e-08, 3.6322e-08, 9.3132e-08, ..., -1.2107e-08, + -2.3469e-07, 1.8626e-08], + [ 3.8184e-08, 7.0781e-08, 5.6438e-07, ..., 1.3318e-07, + 8.3819e-09, -1.8720e-07], + [ 4.4703e-08, 3.4459e-08, 2.9709e-07, ..., 3.7253e-08, + 1.5367e-07, 6.4261e-08]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0043, 0.0118, 0.0197, -0.0021, 0.0223, -0.0075, -0.0095, -0.0073, + 0.0140, -0.0130], device='cuda:0'), grad: tensor([-3.0361e-06, -1.3784e-07, 7.1805e-07, -3.7253e-08, 4.1537e-07, + -1.9670e-06, 3.3733e-06, -3.5949e-07, -4.7404e-07, 1.4668e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 251.63, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.5300 re_mapping 0.0050 re_causal 0.0156 /// teacc 99.08 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1524, -0.0291, 0.1687, ..., 0.0422, -0.1525, -0.1186], + [-0.1021, -0.1056, -0.0298, ..., -0.0935, -0.1194, -0.0579], + [ 0.1951, -0.0514, -0.1578, ..., -0.1328, -0.1358, -0.1794], + ..., + [-0.0561, -0.0123, -0.1599, ..., -0.0679, 0.0770, 0.0139], + [-0.1437, -0.0904, -0.0228, ..., -0.1351, -0.0189, -0.0238], + [-0.0527, -0.0725, -0.0777, ..., 0.0893, 0.0296, -0.1476]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.3318e-07, -9.4995e-08, ..., -8.1025e-08, + 1.9465e-07, 1.3039e-08], + [ 5.1223e-08, 2.3283e-08, -1.7229e-07, ..., 2.9802e-08, + 8.6613e-08, 1.1176e-08], + [ 2.2352e-08, 3.6322e-08, 1.0896e-07, ..., 3.9116e-08, + 2.6077e-08, 1.2107e-08], + ..., + [ 4.9360e-08, 3.6322e-08, 7.0781e-08, ..., 1.9837e-07, + -8.2888e-08, 6.5193e-09], + [ 4.7497e-08, 4.7777e-07, 3.5949e-07, ..., 1.6205e-07, + 6.2771e-07, 3.4459e-08], + [ 2.4214e-08, 7.0781e-08, 1.6298e-07, ..., -9.1642e-07, + -1.5646e-07, 4.6566e-09]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0042, 0.0121, 0.0178, -0.0022, 0.0220, -0.0073, -0.0093, -0.0074, + 0.0146, -0.0130], device='cuda:0'), grad: tensor([ 8.7265e-07, -2.0582e-06, 1.0561e-06, -5.9195e-06, 9.1828e-07, + 1.5749e-06, 1.1716e-06, 1.1465e-06, 3.2783e-06, -2.0824e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 251.68, cls_loss 0.0010 cls_loss_mapping 0.0032 cls_loss_causal 0.5041 re_mapping 0.0047 re_causal 0.0146 /// teacc 99.06 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1534, -0.0292, 0.1691, ..., 0.0423, -0.1534, -0.1188], + [-0.1035, -0.1065, -0.0298, ..., -0.0937, -0.1197, -0.0581], + [ 0.1956, -0.0518, -0.1590, ..., -0.1337, -0.1361, -0.1783], + ..., + [-0.0551, -0.0124, -0.1608, ..., -0.0682, 0.0773, 0.0127], + [-0.1443, -0.0906, -0.0227, ..., -0.1360, -0.0190, -0.0238], + [-0.0531, -0.0728, -0.0782, ..., 0.0894, 0.0297, -0.1478]], + device='cuda:0'), grad: tensor([[ 9.6858e-08, 4.2841e-08, -3.7532e-07, ..., -2.1514e-07, + 1.4156e-07, 6.0536e-08], + [ 4.1723e-07, 1.7695e-08, 6.7987e-08, ..., 2.5146e-08, + 1.1362e-07, 2.2352e-08], + [-3.0547e-07, 3.5390e-08, 8.9407e-08, ..., 1.9372e-07, + 2.0955e-07, 1.4901e-08], + ..., + [-1.7509e-06, 2.9802e-08, 3.1665e-08, ..., 3.6322e-08, + -1.0449e-06, 9.3132e-09], + [ 1.6019e-07, 6.5193e-08, 3.1758e-07, ..., 1.2200e-07, + 2.0955e-07, 1.3318e-07], + [ 1.0310e-06, 3.4459e-08, 2.1514e-07, ..., -6.1002e-07, + 6.4634e-07, 2.7940e-08]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0041, 0.0119, 0.0176, -0.0024, 0.0217, -0.0070, -0.0091, -0.0071, + 0.0149, -0.0130], device='cuda:0'), grad: tensor([ 1.2666e-07, -2.9225e-06, 1.7220e-06, 9.3039e-07, 1.0645e-06, + 1.5274e-07, -1.0720e-06, -5.8860e-06, 1.6261e-06, 4.2319e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 251.80, cls_loss 0.0014 cls_loss_mapping 0.0032 cls_loss_causal 0.5021 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.10 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1550, -0.0293, 0.1699, ..., 0.0408, -0.1543, -0.1190], + [-0.1025, -0.1074, -0.0298, ..., -0.0943, -0.1201, -0.0581], + [ 0.1942, -0.0544, -0.1600, ..., -0.1360, -0.1366, -0.1787], + ..., + [-0.0564, -0.0129, -0.1622, ..., -0.0704, 0.0777, 0.0126], + [-0.1451, -0.0908, -0.0228, ..., -0.1373, -0.0194, -0.0241], + [-0.0511, -0.0730, -0.0798, ..., 0.0913, 0.0299, -0.1478]], + device='cuda:0'), grad: tensor([[ 8.7544e-08, 0.0000e+00, -4.2096e-07, ..., -2.4121e-07, + 4.0047e-08, 3.3528e-08], + [ 1.0151e-07, 9.3132e-10, 4.9360e-08, ..., 7.6368e-08, + 2.0303e-07, 4.8429e-08], + [-6.8825e-07, 1.8626e-09, 9.2201e-08, ..., 2.7008e-08, + 6.7987e-08, -2.0489e-08], + ..., + [ 3.8650e-07, 1.8626e-09, 1.4901e-08, ..., 1.0990e-07, + -4.7870e-07, 2.7940e-08], + [ 4.0047e-08, 0.0000e+00, -3.2224e-07, ..., -6.6124e-08, + -6.7055e-08, -4.1910e-08], + [ 4.2841e-08, 9.3132e-10, 4.3400e-07, ..., -1.9558e-07, + 4.2841e-08, 4.0978e-08]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0025, 0.0124, 0.0147, -0.0015, 0.0212, -0.0069, -0.0089, -0.0076, + 0.0144, -0.0106], device='cuda:0'), grad: tensor([-3.4459e-08, 1.5469e-06, 1.8068e-07, 1.6298e-06, 7.3854e-07, + 4.6846e-07, 1.6019e-07, -9.3412e-07, -7.8008e-06, 4.0345e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 251.13, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.4937 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.03 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1559, -0.0300, 0.1704, ..., 0.0408, -0.1556, -0.1193], + [-0.1029, -0.1084, -0.0299, ..., -0.0943, -0.1206, -0.0585], + [ 0.1947, -0.0546, -0.1609, ..., -0.1365, -0.1368, -0.1787], + ..., + [-0.0561, -0.0131, -0.1635, ..., -0.0705, 0.0784, 0.0129], + [-0.1460, -0.0910, -0.0229, ..., -0.1379, -0.0196, -0.0243], + [-0.0518, -0.0736, -0.0807, ..., 0.0914, 0.0298, -0.1479]], + device='cuda:0'), grad: tensor([[ 5.1130e-07, 9.3132e-10, -3.6322e-08, ..., 6.5658e-07, + 9.8161e-07, 3.6322e-08], + [-2.3860e-06, 6.5193e-09, 6.1467e-08, ..., 8.6799e-07, + 4.5858e-06, 1.3970e-08], + [ 3.5353e-06, 3.7253e-09, 3.5390e-08, ..., 3.5204e-07, + 6.2492e-07, 6.5193e-09], + ..., + [-4.8578e-06, 4.6566e-09, 1.7695e-08, ..., -4.1500e-06, + -1.0155e-05, 1.8626e-09], + [ 7.7300e-08, 1.8626e-09, 4.3958e-07, ..., 2.1234e-07, + 1.9558e-07, 3.4459e-08], + [ 7.8417e-07, 6.5193e-09, 1.0151e-07, ..., 2.1279e-05, + 4.9658e-06, 3.7253e-08]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0024, 0.0124, 0.0144, -0.0020, 0.0216, -0.0063, -0.0092, -0.0074, + 0.0142, -0.0107], device='cuda:0'), grad: tensor([ 5.8226e-06, -4.6372e-05, 5.3138e-05, 1.1802e-05, -2.6196e-05, + 3.5856e-06, -4.3120e-07, -4.8131e-05, 6.0629e-07, 4.6104e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 251.65, cls_loss 0.0010 cls_loss_mapping 0.0033 cls_loss_causal 0.5307 re_mapping 0.0046 re_causal 0.0149 /// teacc 98.91 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.1558, -0.0300, 0.1716, ..., 0.0408, -0.1566, -0.1194], + [-0.1026, -0.1097, -0.0299, ..., -0.0922, -0.1212, -0.0586], + [ 0.1950, -0.0545, -0.1618, ..., -0.1376, -0.1370, -0.1788], + ..., + [-0.0563, -0.0134, -0.1648, ..., -0.0729, 0.0787, 0.0129], + [-0.1462, -0.0908, -0.0230, ..., -0.1393, -0.0197, -0.0244], + [-0.0527, -0.0749, -0.0826, ..., 0.0914, 0.0285, -0.1496]], + device='cuda:0'), grad: tensor([[ 7.3109e-07, 1.1176e-08, -3.4459e-08, ..., 1.2573e-07, + 2.5705e-07, 3.5390e-08], + [ 1.3253e-06, 1.2200e-07, 8.3819e-09, ..., 1.1176e-07, + 5.7463e-07, 2.3469e-07], + [-4.6752e-07, 2.7101e-07, 1.3970e-08, ..., 4.0978e-08, + 8.0094e-08, 1.6764e-08], + ..., + [ 1.0226e-06, 3.2596e-08, 1.8626e-09, ..., 9.1270e-08, + -1.2415e-06, -5.4017e-07], + [ 4.1071e-07, 9.3132e-09, 1.3039e-08, ..., 1.5926e-07, + 7.3574e-08, -1.1642e-07], + [ 1.1744e-06, 1.7695e-08, 5.1223e-08, ..., -3.1106e-07, + 5.1875e-07, 3.0734e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0026, 0.0131, 0.0138, -0.0023, 0.0215, -0.0050, -0.0093, -0.0079, + 0.0146, -0.0118], device='cuda:0'), grad: tensor([ 2.3246e-06, 4.7013e-06, 1.0524e-07, -3.4831e-07, 1.2442e-06, + -8.6650e-06, 9.7137e-07, -2.7437e-06, -1.5143e-06, 3.8780e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 251.14, cls_loss 0.0011 cls_loss_mapping 0.0039 cls_loss_causal 0.5655 re_mapping 0.0046 re_causal 0.0148 /// teacc 99.15 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1568, -0.0301, 0.1720, ..., 0.0409, -0.1582, -0.1196], + [-0.1038, -0.1116, -0.0300, ..., -0.0925, -0.1222, -0.0588], + [ 0.1958, -0.0546, -0.1622, ..., -0.1381, -0.1372, -0.1794], + ..., + [-0.0560, -0.0136, -0.1653, ..., -0.0730, 0.0797, 0.0131], + [-0.1465, -0.0911, -0.0230, ..., -0.1405, -0.0198, -0.0244], + [-0.0530, -0.0751, -0.0834, ..., 0.0917, 0.0285, -0.1498]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 4.6566e-09, -3.1851e-07, ..., -1.1828e-07, + 3.2596e-08, 1.4901e-08], + [ 4.5635e-08, 9.3132e-09, -2.7940e-09, ..., 6.2399e-08, + 6.2399e-08, 5.5879e-09], + [-2.1886e-07, 3.7253e-09, 2.7008e-08, ..., 1.4901e-08, + 1.7695e-08, 5.5879e-09], + ..., + [ 1.2200e-07, 1.5832e-08, 2.1420e-08, ..., 9.4064e-08, + -6.1467e-08, 1.8626e-09], + [ 2.9802e-08, 2.7008e-08, 1.0338e-07, ..., 2.5239e-07, + 1.7695e-07, 3.7253e-08], + [ 1.4901e-08, 4.3772e-08, 1.0245e-07, ..., -4.7497e-07, + 7.6368e-08, 8.3819e-09]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0027, 0.0128, 0.0140, -0.0022, 0.0212, -0.0054, -0.0093, -0.0076, + 0.0151, -0.0116], device='cuda:0'), grad: tensor([-2.9150e-07, -4.0978e-07, -9.9652e-08, 1.6093e-06, 6.7987e-07, + -3.6955e-06, 4.6566e-08, 6.2212e-07, 1.8114e-06, -2.9616e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 251.42, cls_loss 0.0013 cls_loss_mapping 0.0043 cls_loss_causal 0.5228 re_mapping 0.0048 re_causal 0.0151 /// teacc 99.02 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.1572, -0.0301, 0.1729, ..., 0.0409, -0.1599, -0.1198], + [-0.1046, -0.1130, -0.0301, ..., -0.0932, -0.1231, -0.0589], + [ 0.1963, -0.0548, -0.1627, ..., -0.1385, -0.1375, -0.1794], + ..., + [-0.0560, -0.0137, -0.1663, ..., -0.0745, 0.0791, 0.0131], + [-0.1471, -0.0915, -0.0232, ..., -0.1415, -0.0200, -0.0244], + [-0.0532, -0.0754, -0.0839, ..., 0.0911, 0.0277, -0.1499]], + device='cuda:0'), grad: tensor([[ 2.3376e-07, 4.6566e-09, -6.7987e-07, ..., -2.0582e-07, + 1.0245e-07, 6.5193e-09], + [ 4.8280e-06, 1.0245e-08, -2.2445e-07, ..., -1.1828e-07, + 6.9384e-07, 2.7940e-09], + [-1.5795e-05, 1.1176e-08, 3.5390e-08, ..., 5.3085e-08, + 4.4145e-07, 3.7253e-09], + ..., + [ 7.5623e-06, 1.4901e-08, 1.7509e-07, ..., 2.3656e-07, + -1.5292e-06, 0.0000e+00], + [ 2.0713e-06, 0.0000e+00, 1.0338e-07, ..., 3.7067e-07, + 2.6450e-07, 3.7253e-09], + [ 1.2945e-07, 1.7695e-08, 2.9616e-07, ..., 6.8825e-07, + 2.7474e-07, 9.3132e-10]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0028, 0.0119, 0.0142, -0.0022, 0.0220, -0.0058, -0.0090, -0.0070, + 0.0150, -0.0122], device='cuda:0'), grad: tensor([ 3.9209e-07, 9.6858e-06, -2.6733e-05, 2.3358e-06, -1.8151e-06, + 2.1234e-07, 1.3784e-06, 3.5670e-06, 6.9663e-06, 4.0308e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 251.77, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5478 re_mapping 0.0044 re_causal 0.0144 /// teacc 98.92 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1581, -0.0315, 0.1762, ..., 0.0410, -0.1574, -0.1208], + [-0.1049, -0.1150, -0.0301, ..., -0.0935, -0.1239, -0.0590], + [ 0.1969, -0.0550, -0.1638, ..., -0.1391, -0.1378, -0.1795], + ..., + [-0.0562, -0.0143, -0.1676, ..., -0.0750, 0.0798, 0.0131], + [-0.1478, -0.0920, -0.0233, ..., -0.1425, -0.0202, -0.0244], + [-0.0543, -0.0780, -0.0845, ..., 0.0913, 0.0279, -0.1499]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, -3.0734e-08, -1.0682e-06, ..., -3.5111e-07, + 9.5926e-08, 3.7253e-09], + [ 1.4901e-07, 1.4901e-08, 3.0734e-08, ..., 2.1420e-07, + 5.6252e-07, 9.3132e-10], + [-4.7870e-07, 9.3132e-09, 8.2888e-08, ..., 1.7881e-07, + 8.8476e-08, 9.3132e-10], + ..., + [ 1.3411e-07, 1.6764e-08, 9.3132e-08, ..., 3.6880e-07, + -5.4389e-07, 0.0000e+00], + [ 4.8429e-08, 1.3970e-08, 8.2888e-08, ..., 4.0233e-07, + 2.0862e-07, 1.8626e-09], + [ 3.1665e-08, 2.1420e-08, 2.5798e-07, ..., -2.1867e-06, + -1.1278e-06, 9.3132e-10]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0044, 0.0118, 0.0141, -0.0015, 0.0218, -0.0056, -0.0111, -0.0069, + 0.0150, -0.0123], device='cuda:0'), grad: tensor([-9.7975e-07, 2.6859e-06, 5.5227e-07, 8.2795e-07, 1.5339e-06, + 1.5507e-06, 7.8231e-07, -7.9256e-07, -1.6000e-06, -4.5784e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 251.90, cls_loss 0.0011 cls_loss_mapping 0.0040 cls_loss_causal 0.4919 re_mapping 0.0045 re_causal 0.0146 /// teacc 98.92 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1594, -0.0317, 0.1763, ..., 0.0411, -0.1577, -0.1211], + [-0.1052, -0.1159, -0.0303, ..., -0.0936, -0.1247, -0.0591], + [ 0.1973, -0.0552, -0.1647, ..., -0.1394, -0.1382, -0.1794], + ..., + [-0.0559, -0.0142, -0.1687, ..., -0.0750, 0.0804, 0.0131], + [-0.1482, -0.0923, -0.0233, ..., -0.1435, -0.0204, -0.0244], + [-0.0547, -0.0782, -0.0848, ..., 0.0909, 0.0273, -0.1500]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, -1.5004e-06, ..., -3.6415e-07, + 3.1665e-08, 2.7940e-09], + [ 4.3772e-08, 1.1176e-08, 3.5390e-08, ..., 3.6135e-07, + 7.3388e-07, 3.0734e-08], + [ 2.4214e-08, 1.3970e-08, 1.6764e-08, ..., 1.4901e-08, + 1.7695e-08, 1.2107e-08], + ..., + [ 1.1921e-07, 2.2352e-08, 5.2154e-08, ..., -7.4506e-09, + -7.4506e-07, 9.3132e-10], + [ 4.1910e-08, 1.3970e-08, 8.0094e-08, ..., 1.9651e-07, + 2.0117e-07, 1.7043e-07], + [ 1.3039e-08, 4.6566e-09, 6.0722e-07, ..., -1.2424e-06, + -5.0198e-07, 1.1176e-08]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0043, 0.0118, 0.0139, -0.0016, 0.0224, -0.0060, -0.0104, -0.0068, + 0.0151, -0.0130], device='cuda:0'), grad: tensor([-2.0713e-06, 2.0750e-06, 2.6636e-07, 1.5302e-06, 1.4128e-06, + -3.0883e-06, 7.2643e-07, -1.5274e-06, 2.2184e-06, -1.5413e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 251.71, cls_loss 0.0008 cls_loss_mapping 0.0028 cls_loss_causal 0.4967 re_mapping 0.0047 re_causal 0.0149 /// teacc 99.01 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1597, -0.0315, 0.1764, ..., 0.0411, -0.1578, -0.1212], + [-0.1054, -0.1164, -0.0302, ..., -0.0937, -0.1253, -0.0598], + [ 0.1977, -0.0554, -0.1650, ..., -0.1398, -0.1386, -0.1796], + ..., + [-0.0562, -0.0144, -0.1702, ..., -0.0753, 0.0808, 0.0131], + [-0.1491, -0.0925, -0.0234, ..., -0.1457, -0.0209, -0.0246], + [-0.0545, -0.0784, -0.0856, ..., 0.0910, 0.0273, -0.1501]], + device='cuda:0'), grad: tensor([[ 5.5879e-08, 9.3132e-10, -3.2783e-07, ..., -2.8126e-07, + 1.3039e-07, 1.6112e-07], + [ 1.0058e-07, 9.3132e-10, 5.0291e-08, ..., -9.3132e-10, + 3.4571e-06, 1.2200e-07], + [-1.0207e-06, 9.3132e-10, 6.8918e-08, ..., 3.7253e-08, + 3.6322e-08, 1.3039e-08], + ..., + [ 2.6915e-07, 9.3132e-10, 6.1467e-08, ..., 1.3039e-07, + -3.8072e-06, 1.6764e-08], + [ 1.6298e-07, 1.8626e-09, 1.2480e-07, ..., 4.4703e-08, + 1.4622e-07, 1.6391e-07], + [ 2.3283e-08, 9.3132e-10, 2.9523e-07, ..., 1.7602e-07, + 1.0617e-07, 4.9360e-08]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0044, 0.0119, 0.0139, -0.0017, 0.0224, -0.0056, -0.0107, -0.0068, + 0.0147, -0.0130], device='cuda:0'), grad: tensor([-2.4587e-07, 2.1935e-05, -1.3700e-06, 1.7313e-06, -2.5332e-07, + -1.0319e-06, -2.6450e-07, -2.3291e-05, 1.4212e-06, 1.3355e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 252.01, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5379 re_mapping 0.0044 re_causal 0.0147 /// teacc 99.09 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1602, -0.0316, 0.1765, ..., 0.0412, -0.1580, -0.1215], + [-0.1060, -0.1175, -0.0300, ..., -0.0933, -0.1265, -0.0601], + [ 0.1983, -0.0555, -0.1657, ..., -0.1400, -0.1392, -0.1800], + ..., + [-0.0563, -0.0144, -0.1720, ..., -0.0760, 0.0818, 0.0135], + [-0.1497, -0.0925, -0.0235, ..., -0.1468, -0.0210, -0.0245], + [-0.0547, -0.0789, -0.0859, ..., 0.0910, 0.0273, -0.1502]], + device='cuda:0'), grad: tensor([[ 1.6671e-07, 5.5879e-09, 2.3376e-06, ..., -5.5134e-07, + 8.4843e-07, 2.1420e-06], + [ 1.1502e-06, 6.5193e-09, 2.4959e-07, ..., -7.0781e-08, + 3.2876e-07, 1.6950e-07], + [-3.1181e-06, 5.5879e-09, 8.4937e-07, ..., 1.7229e-07, + 2.8126e-07, 4.0140e-07], + ..., + [ 1.2871e-06, 1.7695e-08, 1.4622e-07, ..., 2.4680e-07, + -9.2667e-07, 6.5193e-09], + [-3.2131e-07, 9.3132e-09, 4.7777e-07, ..., 3.5949e-07, + 1.5553e-07, 1.3318e-07], + [ 1.5926e-07, 6.5193e-09, 4.2375e-07, ..., -4.6659e-07, + -5.3085e-08, 3.4459e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0043, 0.0118, 0.0138, -0.0017, 0.0224, -0.0056, -0.0108, -0.0066, + 0.0151, -0.0132], device='cuda:0'), grad: tensor([ 4.2617e-06, 1.0990e-07, -3.5502e-06, 1.0151e-06, 2.5909e-06, + 4.9472e-06, -1.1563e-05, 1.6112e-06, -1.7509e-06, 2.2799e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 251.81, cls_loss 0.0009 cls_loss_mapping 0.0036 cls_loss_causal 0.5405 re_mapping 0.0047 re_causal 0.0151 /// teacc 98.97 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1607, -0.0316, 0.1768, ..., 0.0413, -0.1581, -0.1217], + [-0.1067, -0.1185, -0.0302, ..., -0.0934, -0.1271, -0.0602], + [ 0.1990, -0.0558, -0.1665, ..., -0.1404, -0.1398, -0.1802], + ..., + [-0.0565, -0.0140, -0.1737, ..., -0.0761, 0.0828, 0.0139], + [-0.1500, -0.0928, -0.0232, ..., -0.1475, -0.0211, -0.0245], + [-0.0548, -0.0791, -0.0873, ..., 0.0912, 0.0280, -0.1503]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, 1.4901e-08, -2.5146e-08, ..., -9.3132e-09, + 2.1420e-08, 2.7940e-09], + [ 1.2573e-07, 4.2841e-08, -3.7253e-09, ..., 6.5193e-09, + 7.7300e-08, 9.3132e-10], + [-3.3528e-06, 5.3085e-08, 7.4506e-09, ..., 2.7940e-09, + 3.8184e-08, 1.8626e-09], + ..., + [ 1.9185e-07, 4.0047e-08, 2.7940e-09, ..., 8.3819e-09, + -3.3621e-07, 9.3132e-10], + [ 3.2131e-07, 2.2352e-08, 6.5193e-09, ..., 8.3819e-09, + 2.1420e-08, 1.8626e-09], + [ 2.6077e-08, 2.3283e-08, 2.1420e-08, ..., -1.7788e-07, + 8.1956e-08, 9.3132e-10]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0045, 0.0117, 0.0140, -0.0021, 0.0219, -0.0055, -0.0109, -0.0065, + 0.0158, -0.0132], device='cuda:0'), grad: tensor([ 1.9185e-07, 2.3190e-07, -3.7830e-06, -2.5406e-06, 3.2689e-06, + 1.9372e-06, 8.1025e-08, -4.2934e-07, 7.4320e-07, 2.9057e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 251.77, cls_loss 0.0015 cls_loss_mapping 0.0038 cls_loss_causal 0.5297 re_mapping 0.0045 re_causal 0.0137 /// teacc 99.03 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1615, -0.0336, 0.1765, ..., 0.0415, -0.1587, -0.1230], + [-0.1071, -0.1211, -0.0300, ..., -0.0935, -0.1285, -0.0603], + [ 0.1997, -0.0559, -0.1683, ..., -0.1414, -0.1402, -0.1805], + ..., + [-0.0569, -0.0155, -0.1764, ..., -0.0770, 0.0825, 0.0140], + [-0.1511, -0.0942, -0.0236, ..., -0.1509, -0.0220, -0.0246], + [-0.0549, -0.0791, -0.0890, ..., 0.0916, 0.0288, -0.1504]], + device='cuda:0'), grad: tensor([[ 7.1712e-08, 2.6077e-08, -3.7532e-07, ..., 3.4180e-07, + 1.3877e-07, 2.7940e-09], + [ 1.7975e-07, 7.3574e-08, -1.1176e-08, ..., 4.1630e-07, + 5.0291e-07, 2.1420e-08], + [ 1.6671e-07, 2.1141e-07, 1.7695e-08, ..., 4.4610e-07, + 1.6391e-07, 2.7940e-09], + ..., + [ 3.0920e-07, 1.2759e-07, 3.1665e-08, ..., 9.9652e-07, + -7.7393e-07, -3.2596e-08], + [ 1.1083e-07, 3.2596e-08, 5.5879e-08, ..., 9.2015e-07, + 4.8243e-07, 9.3132e-10], + [ 5.8673e-08, 2.7008e-08, 1.3318e-07, ..., -1.3439e-06, + -9.3691e-07, 6.5193e-09]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0043, 0.0117, 0.0141, -0.0026, 0.0217, -0.0043, -0.0103, -0.0067, + 0.0152, -0.0130], device='cuda:0'), grad: tensor([ 1.2424e-06, 9.8627e-07, 2.4289e-06, -1.7760e-06, -7.0632e-06, + 1.3961e-06, 1.3188e-06, -5.9605e-08, 3.6731e-06, -2.1793e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 251.51, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.5084 re_mapping 0.0045 re_causal 0.0143 /// teacc 99.12 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1622, -0.0337, 0.1767, ..., 0.0395, -0.1588, -0.1233], + [-0.1073, -0.1224, -0.0300, ..., -0.0936, -0.1299, -0.0605], + [ 0.2003, -0.0561, -0.1697, ..., -0.1421, -0.1409, -0.1806], + ..., + [-0.0574, -0.0167, -0.1780, ..., -0.0774, 0.0828, 0.0142], + [-0.1522, -0.0950, -0.0236, ..., -0.1514, -0.0222, -0.0247], + [-0.0550, -0.0792, -0.0897, ..., 0.0933, 0.0287, -0.1504]], + device='cuda:0'), grad: tensor([[ 2.3656e-07, 1.4901e-08, -5.6811e-08, ..., -1.6950e-07, + 6.0536e-08, 9.4995e-08], + [-1.2787e-06, 8.0094e-08, 1.8626e-07, ..., 3.3528e-08, + 1.2945e-07, 4.2841e-08], + [-1.9521e-06, -1.0245e-08, 4.6566e-08, ..., 7.4506e-09, + 1.5926e-07, 1.4901e-08], + ..., + [ 1.8552e-06, 2.2724e-07, 2.1420e-08, ..., 5.1223e-08, + -4.6939e-07, 9.3132e-10], + [ 5.7928e-07, 6.2399e-08, 2.1979e-07, ..., 3.7253e-08, + 5.8673e-08, 6.5193e-08], + [ 7.3574e-08, 9.4995e-08, 1.9837e-07, ..., -1.6298e-07, + 3.3528e-08, 4.6566e-09]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0024, 0.0117, 0.0142, -0.0033, 0.0216, -0.0031, -0.0101, -0.0067, + 0.0149, -0.0117], device='cuda:0'), grad: tensor([ 7.3668e-07, -2.0072e-05, 1.7695e-06, -4.2655e-06, 2.6915e-06, + 3.1181e-06, -2.2352e-06, 1.5438e-05, 1.9539e-06, 8.6892e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 251.42, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.5049 re_mapping 0.0046 re_causal 0.0147 /// teacc 98.95 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1626, -0.0337, 0.1769, ..., 0.0395, -0.1591, -0.1236], + [-0.1075, -0.1234, -0.0302, ..., -0.0937, -0.1305, -0.0606], + [ 0.2025, -0.0564, -0.1712, ..., -0.1429, -0.1414, -0.1809], + ..., + [-0.0582, -0.0168, -0.1802, ..., -0.0770, 0.0847, 0.0142], + [-0.1557, -0.0954, -0.0241, ..., -0.1517, -0.0223, -0.0248], + [-0.0554, -0.0793, -0.0904, ..., 0.0932, 0.0278, -0.1505]], + device='cuda:0'), grad: tensor([[ 1.5367e-08, 6.5193e-09, 1.5832e-07, ..., 1.4435e-08, + 7.7765e-08, 7.2643e-08], + [ 4.7963e-08, 1.5832e-08, -6.9849e-09, ..., 2.6077e-08, + 3.1665e-08, -2.5388e-06], + [-7.4506e-08, 1.2573e-08, 2.7940e-08, ..., 2.1420e-08, + 2.2352e-08, 5.8673e-08], + ..., + [ 8.4750e-08, 4.3772e-08, 1.0710e-08, ..., 4.6659e-07, + 2.8219e-07, 4.1258e-07], + [ 3.0734e-08, 5.6345e-08, 1.7276e-07, ..., 4.1910e-08, + 2.1514e-07, 1.0161e-06], + [ 7.4506e-09, 4.1910e-09, 2.6543e-08, ..., 9.8255e-08, + 5.8673e-08, 4.0513e-08]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0024, 0.0119, 0.0161, -0.0032, 0.0215, -0.0033, -0.0096, -0.0067, + 0.0130, -0.0121], device='cuda:0'), grad: tensor([ 1.3644e-06, -1.3769e-04, 2.8387e-06, -4.6566e-08, 2.6658e-05, + 1.5870e-05, 1.1757e-05, 2.3857e-05, 5.3197e-05, 2.2110e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 251.64, cls_loss 0.0008 cls_loss_mapping 0.0029 cls_loss_causal 0.4971 re_mapping 0.0048 re_causal 0.0148 /// teacc 98.99 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1628, -0.0338, 0.1767, ..., 0.0393, -0.1594, -0.1247], + [-0.1079, -0.1241, -0.0303, ..., -0.0940, -0.1311, -0.0602], + [ 0.2029, -0.0565, -0.1717, ..., -0.1433, -0.1419, -0.1810], + ..., + [-0.0582, -0.0169, -0.1813, ..., -0.0775, 0.0850, 0.0141], + [-0.1558, -0.0960, -0.0243, ..., -0.1520, -0.0226, -0.0248], + [-0.0555, -0.0794, -0.0908, ..., 0.0937, 0.0281, -0.1505]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, -4.8429e-07, ..., -1.3318e-07, + 1.2573e-07, 2.5425e-07], + [ 2.6077e-08, 3.7253e-09, 3.5018e-07, ..., 2.1700e-07, + 5.3365e-07, 6.5845e-07], + [-7.5437e-08, 9.3132e-10, 4.0326e-07, ..., 1.9558e-08, + 3.5204e-07, 9.1735e-07], + ..., + [ 3.7253e-08, 5.5879e-09, 4.0978e-08, ..., 7.9907e-07, + -2.2724e-07, -4.6566e-09], + [ 3.7253e-09, 3.7253e-09, 3.1367e-06, ..., 1.3970e-08, + 1.9819e-06, 7.0632e-06], + [ 9.3132e-10, 4.6566e-09, 2.0582e-07, ..., -1.0030e-06, + -5.1968e-07, 2.7008e-08]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0019, 0.0120, 0.0161, -0.0032, 0.0213, -0.0032, -0.0093, -0.0068, + 0.0128, -0.0117], device='cuda:0'), grad: tensor([ 2.8871e-07, 1.6876e-06, 3.7439e-06, 2.4233e-06, -1.1921e-07, + 5.8524e-06, -3.7879e-05, 1.0943e-06, 2.5809e-05, -2.8983e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 251.57, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5013 re_mapping 0.0045 re_causal 0.0137 /// teacc 98.85 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1632, -0.0338, 0.1765, ..., 0.0393, -0.1597, -0.1261], + [-0.1086, -0.1252, -0.0303, ..., -0.0943, -0.1320, -0.0592], + [ 0.2029, -0.0594, -0.1734, ..., -0.1449, -0.1428, -0.1810], + ..., + [-0.0580, -0.0167, -0.1832, ..., -0.0779, 0.0856, 0.0135], + [-0.1559, -0.0962, -0.0245, ..., -0.1525, -0.0227, -0.0251], + [-0.0554, -0.0794, -0.0914, ..., 0.0911, 0.0250, -0.1506]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 9.3132e-10, -1.6131e-06, ..., -5.2433e-07, + 4.9360e-08, 6.7055e-08], + [ 6.3330e-08, 6.5193e-09, 1.0245e-07, ..., 1.1176e-07, + 6.3330e-08, 2.8871e-08], + [-1.7416e-07, 5.5879e-09, 2.4028e-07, ..., 8.8476e-08, + 2.8871e-08, 6.3330e-08], + ..., + [ 2.5146e-08, 7.4506e-09, 6.7987e-08, ..., 6.0257e-07, + 2.8964e-07, 3.7253e-09], + [ 3.0734e-08, 2.7940e-09, 1.3132e-07, ..., 1.1921e-07, + 5.3085e-08, 3.2596e-08], + [ 4.6566e-09, 9.3132e-10, 4.4703e-07, ..., -1.9576e-06, + -1.0831e-06, 1.8626e-09]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0018, 0.0121, 0.0156, -0.0030, 0.0244, -0.0035, -0.0091, -0.0067, + 0.0127, -0.0144], device='cuda:0'), grad: tensor([-2.5332e-06, -6.6962e-07, 6.6217e-07, 6.3423e-07, 4.8652e-06, + 8.7731e-07, -1.6391e-07, 2.6189e-06, -1.0859e-06, -5.2303e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 251.72, cls_loss 0.0010 cls_loss_mapping 0.0041 cls_loss_causal 0.5120 re_mapping 0.0048 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.1637, -0.0339, 0.1749, ..., 0.0393, -0.1612, -0.1305], + [-0.1088, -0.1260, -0.0304, ..., -0.0951, -0.1328, -0.0598], + [ 0.2031, -0.0595, -0.1752, ..., -0.1458, -0.1437, -0.1824], + ..., + [-0.0581, -0.0168, -0.1841, ..., -0.0791, 0.0856, 0.0134], + [-0.1561, -0.0960, -0.0247, ..., -0.1529, -0.0228, -0.0252], + [-0.0555, -0.0797, -0.0916, ..., 0.0916, 0.0251, -0.1507]], + device='cuda:0'), grad: tensor([[ 2.7195e-07, 2.6356e-07, -6.6236e-06, ..., -1.9316e-06, + 5.4948e-08, 7.2643e-08], + [ 1.3970e-07, 1.6019e-07, 2.5332e-07, ..., 2.8871e-08, + 6.9849e-08, 8.1025e-08], + [ 1.8671e-05, 4.1753e-05, 1.1707e-06, ..., 2.3190e-07, + 2.9802e-08, 2.1420e-08], + ..., + [ 3.2783e-07, 7.2084e-07, 1.2107e-07, ..., 6.5193e-08, + -5.4948e-08, 2.7940e-09], + [-1.2349e-06, -9.7323e-07, 2.1216e-06, ..., 1.1334e-06, + 7.3574e-08, 1.0058e-07], + [ 3.1013e-07, 2.7288e-07, 1.8850e-06, ..., 1.3690e-07, + -7.7300e-08, 3.7253e-09]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0007, 0.0121, 0.0152, -0.0041, 0.0239, -0.0027, -0.0074, -0.0069, + 0.0130, -0.0138], device='cuda:0'), grad: tensor([-1.0349e-05, 2.1625e-06, 8.4281e-05, -8.1658e-05, 8.5868e-07, + -2.3738e-05, 3.3915e-05, 1.9670e-06, -1.6168e-05, 8.6501e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 251.27, cls_loss 0.0010 cls_loss_mapping 0.0035 cls_loss_causal 0.5266 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.08 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1641, -0.0340, 0.1752, ..., 0.0392, -0.1614, -0.1306], + [-0.1092, -0.1273, -0.0309, ..., -0.0954, -0.1337, -0.0600], + [ 0.2032, -0.0603, -0.1773, ..., -0.1462, -0.1450, -0.1837], + ..., + [-0.0575, -0.0173, -0.1858, ..., -0.0794, 0.0860, 0.0135], + [-0.1562, -0.0961, -0.0250, ..., -0.1533, -0.0228, -0.0251], + [-0.0567, -0.0799, -0.0922, ..., 0.0917, 0.0253, -0.1508]], + device='cuda:0'), grad: tensor([[-3.5390e-08, 9.3132e-10, 5.6066e-06, ..., -4.8894e-07, + 1.6019e-07, 1.6158e-06], + [ 2.5146e-08, 4.6566e-09, 9.4995e-08, ..., 5.0291e-08, + 2.8871e-08, 3.9116e-08], + [ 5.5879e-09, 3.7253e-09, 1.6857e-07, ..., 1.1548e-07, + 1.3970e-08, 1.0245e-08], + ..., + [ 2.5146e-08, 5.5879e-09, 3.6322e-08, ..., 4.3772e-08, + -1.4901e-08, 1.8626e-09], + [ 3.5390e-08, 4.6566e-09, 3.1944e-07, ..., 5.2154e-08, + 8.4750e-08, 9.8720e-08], + [ 1.1269e-07, 1.3970e-08, 4.1444e-07, ..., 4.2934e-07, + 1.5832e-08, 1.6764e-08]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0007, 0.0121, 0.0146, -0.0038, 0.0237, -0.0030, -0.0074, -0.0068, + 0.0133, -0.0137], device='cuda:0'), grad: tensor([ 8.8587e-06, -3.5353e-06, 8.8476e-07, -2.5127e-06, -1.4128e-06, + 1.9185e-06, -9.0599e-06, 2.2314e-06, 4.2934e-07, 2.1532e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 251.80, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4930 re_mapping 0.0043 re_causal 0.0140 /// teacc 99.11 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1646, -0.0338, 0.1756, ..., 0.0393, -0.1616, -0.1307], + [-0.1103, -0.1285, -0.0308, ..., -0.0956, -0.1338, -0.0596], + [ 0.2037, -0.0604, -0.1780, ..., -0.1466, -0.1453, -0.1841], + ..., + [-0.0568, -0.0174, -0.1891, ..., -0.0802, 0.0857, 0.0135], + [-0.1562, -0.0963, -0.0250, ..., -0.1538, -0.0229, -0.0250], + [-0.0572, -0.0799, -0.0930, ..., 0.0918, 0.0254, -0.1509]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.8626e-09, -6.3702e-07, ..., -2.2445e-07, + 5.6811e-08, 6.5193e-09], + [ 5.2154e-08, 1.2107e-08, 2.6077e-08, ..., 1.3039e-07, + 3.1292e-07, 2.7940e-08], + [-4.6566e-09, 7.4506e-09, 2.1420e-08, ..., 8.9407e-08, + 1.8813e-07, 1.6764e-08], + ..., + [-6.1002e-07, 1.3039e-08, 2.1420e-08, ..., -1.5022e-06, + -3.6657e-06, -2.9709e-07], + [ 3.9116e-08, 1.4529e-07, 3.9116e-08, ..., 2.7176e-06, + 1.5963e-06, 6.5193e-09], + [ 4.3493e-07, -1.3877e-07, 2.2445e-07, ..., -2.1793e-06, + 5.9046e-07, 2.0023e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0009, 0.0123, 0.0145, -0.0046, 0.0237, -0.0024, -0.0077, -0.0070, + 0.0136, -0.0137], device='cuda:0'), grad: tensor([-6.6403e-07, 1.1064e-06, 6.2864e-07, -2.6915e-07, 2.7455e-06, + 1.0561e-06, 3.9954e-07, -1.0505e-05, 6.8471e-06, -1.3895e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 251.58, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4918 re_mapping 0.0044 re_causal 0.0139 /// teacc 99.14 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1657, -0.0338, 0.1761, ..., 0.0394, -0.1617, -0.1306], + [-0.1110, -0.1331, -0.0310, ..., -0.0955, -0.1344, -0.0598], + [ 0.2045, -0.0605, -0.1789, ..., -0.1472, -0.1455, -0.1842], + ..., + [-0.0572, -0.0170, -0.1906, ..., -0.0808, 0.0857, 0.0137], + [-0.1565, -0.0987, -0.0234, ..., -0.1521, -0.0232, -0.0249], + [-0.0577, -0.0806, -0.0962, ..., 0.0918, 0.0256, -0.1510]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, 1.6764e-08, 8.8476e-08, ..., 1.1642e-06, + 2.8498e-07, 9.7789e-08], + [ 2.9895e-07, 5.7742e-08, 3.7253e-09, ..., 5.8115e-07, + 1.0338e-07, 1.3877e-07], + [-1.4659e-06, -1.1362e-07, 1.5832e-08, ..., 7.5903e-07, + 1.5739e-07, 1.2573e-07], + ..., + [ 1.0198e-06, 1.1083e-07, 2.9802e-08, ..., 9.0897e-07, + 1.9558e-08, 2.7474e-07], + [ 8.0094e-08, 3.5390e-08, 3.0734e-08, ..., 3.2596e-06, + 8.4564e-07, 1.3411e-07], + [ 1.3970e-08, 1.4901e-08, 2.7940e-08, ..., -1.1005e-05, + -2.4512e-06, 4.2375e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0011, 0.0123, 0.0147, -0.0046, 0.0236, -0.0022, -0.0083, -0.0071, + 0.0153, -0.0139], device='cuda:0'), grad: tensor([ 2.8051e-06, 1.2983e-06, -1.4808e-07, -1.2573e-07, 5.1223e-07, + 9.6187e-06, 3.8128e-06, 4.4517e-06, -3.0436e-06, -1.9237e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 251.51, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.5240 re_mapping 0.0044 re_causal 0.0142 /// teacc 99.07 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1677, -0.0335, 0.1768, ..., 0.0398, -0.1618, -0.1307], + [-0.1111, -0.1352, -0.0313, ..., -0.0960, -0.1353, -0.0600], + [ 0.2051, -0.0604, -0.1810, ..., -0.1488, -0.1458, -0.1845], + ..., + [-0.0582, -0.0180, -0.1926, ..., -0.0813, 0.0858, 0.0139], + [-0.1565, -0.0995, -0.0233, ..., -0.1525, -0.0237, -0.0253], + [-0.0580, -0.0815, -0.0988, ..., 0.0918, 0.0258, -0.1510]], + device='cuda:0'), grad: tensor([[ 3.2503e-07, 0.0000e+00, 6.5565e-07, ..., 2.4680e-07, + 2.6263e-07, 1.6019e-07], + [-1.9856e-06, 0.0000e+00, -2.4047e-06, ..., 1.5274e-07, + 1.2014e-07, 1.9558e-08], + [ 8.4098e-07, 9.3132e-10, 1.0505e-06, ..., 9.2201e-08, + 3.9116e-08, 2.0489e-08], + ..., + [ 2.7101e-07, 9.3132e-10, 2.5798e-07, ..., 5.4482e-07, + 1.5460e-07, -1.3039e-08], + [ 5.2806e-07, 9.3132e-10, 7.2550e-07, ..., 2.8778e-07, + 1.3597e-07, 4.8429e-08], + [-8.7544e-08, 0.0000e+00, 1.1362e-07, ..., -2.7586e-06, + -1.0859e-06, 4.0978e-08]], device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0015, 0.0123, 0.0145, -0.0041, 0.0236, -0.0036, -0.0072, -0.0072, + 0.0156, -0.0141], device='cuda:0'), grad: tensor([ 8.1733e-06, -4.5955e-05, 1.9863e-05, 1.5618e-06, 3.2187e-06, + 9.5740e-07, -1.2238e-06, 6.2510e-06, 1.2979e-05, -5.8822e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 251.48, cls_loss 0.0008 cls_loss_mapping 0.0032 cls_loss_causal 0.5345 re_mapping 0.0045 re_causal 0.0148 /// teacc 99.15 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1688, -0.0340, 0.1767, ..., 0.0397, -0.1623, -0.1307], + [-0.1112, -0.1368, -0.0312, ..., -0.0962, -0.1362, -0.0602], + [ 0.2054, -0.0604, -0.1820, ..., -0.1494, -0.1461, -0.1847], + ..., + [-0.0585, -0.0194, -0.1950, ..., -0.0819, 0.0857, 0.0141], + [-0.1567, -0.1001, -0.0233, ..., -0.1527, -0.0237, -0.0253], + [-0.0591, -0.0819, -0.0992, ..., 0.0918, 0.0258, -0.1511]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, 0.0000e+00, -9.0338e-08, ..., 1.1176e-08, + 4.2841e-08, 8.3819e-09], + [ 9.5926e-08, 1.8626e-09, 2.7940e-09, ..., 1.5087e-07, + 9.5926e-08, 2.7940e-09], + [-9.8348e-07, 1.8626e-09, 9.3132e-09, ..., 1.0990e-07, + 5.7742e-08, 1.8626e-09], + ..., + [ 4.2375e-07, 1.8626e-09, 9.3132e-10, ..., 1.8990e-06, + 8.9686e-07, 0.0000e+00], + [ 1.9930e-07, 0.0000e+00, 1.7695e-08, ..., 3.5577e-07, + 1.8533e-07, 6.5193e-09], + [ 8.3819e-09, 0.0000e+00, 3.9116e-08, ..., 2.5481e-06, + 1.2824e-06, 1.8626e-09]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0012, 0.0124, 0.0144, -0.0038, 0.0238, -0.0034, -0.0073, -0.0073, + 0.0159, -0.0142], device='cuda:0'), grad: tensor([ 1.3504e-07, 1.2107e-07, -1.3011e-06, 1.7695e-07, -1.8358e-05, + 7.4133e-07, 9.5181e-07, 7.3016e-06, 1.6447e-06, 8.5756e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 251.62, cls_loss 0.0006 cls_loss_mapping 0.0029 cls_loss_causal 0.5086 re_mapping 0.0045 re_causal 0.0146 /// teacc 99.06 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1691, -0.0340, 0.1769, ..., 0.0397, -0.1624, -0.1308], + [-0.1113, -0.1382, -0.0312, ..., -0.0963, -0.1372, -0.0607], + [ 0.2056, -0.0606, -0.1826, ..., -0.1506, -0.1466, -0.1851], + ..., + [-0.0585, -0.0197, -0.1965, ..., -0.0823, 0.0861, 0.0157], + [-0.1568, -0.1023, -0.0234, ..., -0.1529, -0.0240, -0.0254], + [-0.0591, -0.0820, -0.0995, ..., 0.0918, 0.0258, -0.1512]], + device='cuda:0'), grad: tensor([[ 1.0617e-07, 9.3132e-10, 5.3085e-08, ..., -1.4063e-07, + 2.7940e-09, 9.3132e-10], + [ 2.5891e-07, 5.5879e-09, -3.0845e-06, ..., 4.8429e-08, + 4.5635e-08, 5.5879e-09], + [-3.8855e-06, 9.3132e-10, 8.9128e-07, ..., 6.5193e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 2.5108e-06, -5.5879e-09, 1.8198e-06, ..., 2.7008e-08, + -1.0896e-07, -1.6764e-08], + [ 1.6671e-07, 4.2841e-08, 2.7008e-08, ..., 2.4214e-08, + 3.0734e-08, 1.7695e-08], + [ 7.4506e-09, 1.8626e-09, 1.1828e-07, ..., 2.7940e-09, + -4.0047e-08, 1.8626e-09]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0013, 0.0126, 0.0141, -0.0039, 0.0238, -0.0033, -0.0074, -0.0074, + 0.0157, -0.0143], device='cuda:0'), grad: tensor([ 9.2946e-07, -1.3046e-05, -2.9467e-06, 1.2536e-06, 2.2911e-07, + -5.2154e-07, 5.4203e-07, 1.2249e-05, 1.1371e-06, 1.5274e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 251.80, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.4991 re_mapping 0.0044 re_causal 0.0142 /// teacc 99.13 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1694, -0.0341, 0.1771, ..., 0.0398, -0.1625, -0.1308], + [-0.1116, -0.1401, -0.0319, ..., -0.0962, -0.1382, -0.0617], + [ 0.2061, -0.0608, -0.1831, ..., -0.1510, -0.1469, -0.1852], + ..., + [-0.0587, -0.0196, -0.1975, ..., -0.0826, 0.0865, 0.0159], + [-0.1571, -0.1038, -0.0234, ..., -0.1532, -0.0243, -0.0254], + [-0.0592, -0.0824, -0.0997, ..., 0.0919, 0.0259, -0.1513]], + device='cuda:0'), grad: tensor([[ 2.3823e-06, 7.4506e-09, 4.3772e-08, ..., 9.3132e-09, + 3.8184e-08, 1.3970e-08], + [ 3.5111e-07, 4.5635e-08, 2.8871e-08, ..., 5.0291e-08, + 8.4657e-07, 1.3039e-08], + [-1.0416e-05, 1.1828e-07, 9.3132e-09, ..., 1.8626e-09, + 2.5146e-08, 2.7940e-09], + ..., + [ 6.0908e-07, 6.5193e-08, 0.0000e+00, ..., -5.1223e-08, + -1.1018e-06, 0.0000e+00], + [ 5.1372e-06, 1.9558e-08, 2.3283e-08, ..., 4.9360e-08, + 3.8184e-08, 8.3819e-09], + [ 1.9185e-07, 1.8626e-08, 1.8626e-09, ..., -1.8533e-07, + 1.1642e-07, 9.3132e-10]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0014, 0.0126, 0.0141, -0.0037, 0.0237, -0.0035, -0.0073, -0.0075, + 0.0157, -0.0142], device='cuda:0'), grad: tensor([ 6.5416e-06, 5.7071e-06, -2.7671e-05, 3.5539e-06, 3.7998e-07, + 1.5814e-06, -1.1921e-07, -4.4927e-06, 1.3724e-05, 7.7859e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 251.56, cls_loss 0.0010 cls_loss_mapping 0.0041 cls_loss_causal 0.4995 re_mapping 0.0044 re_causal 0.0136 /// teacc 98.96 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1709, -0.0344, 0.1774, ..., 0.0399, -0.1627, -0.1310], + [-0.1133, -0.1425, -0.0333, ..., -0.0966, -0.1395, -0.0635], + [ 0.2073, -0.0610, -0.1823, ..., -0.1532, -0.1473, -0.1856], + ..., + [-0.0593, -0.0203, -0.1985, ..., -0.0836, 0.0872, 0.0159], + [-0.1575, -0.1036, -0.0233, ..., -0.1533, -0.0240, -0.0254], + [-0.0591, -0.0830, -0.1003, ..., 0.0921, 0.0262, -0.1513]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, -3.6322e-08, ..., 1.5832e-08, + 2.7940e-08, 7.4506e-09], + [ 3.3528e-08, 7.4506e-09, -0.0000e+00, ..., 4.7497e-07, + 1.9744e-07, 2.7940e-09], + [-5.0291e-08, 4.6566e-09, 5.5879e-09, ..., 4.3772e-08, + 2.7008e-08, 2.7940e-09], + ..., + [ 3.4459e-08, 2.3283e-08, 4.6566e-09, ..., 1.5181e-07, + 3.8184e-08, 2.7940e-09], + [ 4.6566e-09, 5.5879e-09, 1.6764e-08, ..., 4.1910e-08, + 4.8429e-08, 2.1420e-08], + [ 4.6566e-09, 5.5879e-09, 3.1665e-08, ..., 4.5970e-06, + 1.9027e-06, 6.5193e-09]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0014, 0.0122, 0.0145, -0.0030, 0.0235, -0.0042, -0.0071, -0.0073, + 0.0163, -0.0141], device='cuda:0'), grad: tensor([ 8.7544e-08, 1.1278e-06, 9.4064e-08, 2.0638e-06, -1.1101e-05, + -2.4512e-06, 5.7369e-07, 4.5262e-07, -1.6671e-07, 9.3356e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 251.42, cls_loss 0.0009 cls_loss_mapping 0.0031 cls_loss_causal 0.5043 re_mapping 0.0040 re_causal 0.0128 /// teacc 99.06 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1716, -0.0345, 0.1778, ..., 0.0401, -0.1628, -0.1311], + [-0.1135, -0.1436, -0.0336, ..., -0.0971, -0.1422, -0.0644], + [ 0.2076, -0.0611, -0.1844, ..., -0.1562, -0.1483, -0.1863], + ..., + [-0.0596, -0.0207, -0.1997, ..., -0.0849, 0.0878, 0.0161], + [-0.1577, -0.1036, -0.0235, ..., -0.1538, -0.0241, -0.0255], + [-0.0594, -0.0832, -0.1011, ..., 0.0924, 0.0269, -0.1515]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 1.3039e-08, -2.7381e-07, ..., -1.3411e-07, + 4.3772e-08, 6.4261e-08], + [ 1.9558e-07, 3.3528e-08, 4.6566e-08, ..., 5.8673e-08, + 1.8626e-07, 4.3772e-08], + [ 1.7975e-07, 7.4506e-08, 4.2841e-08, ..., 1.3039e-08, + 8.8476e-08, 2.9802e-08], + ..., + [-2.4773e-07, 7.3574e-08, 1.0245e-08, ..., 1.0245e-07, + -1.4994e-07, -1.5832e-08], + [ 1.6019e-07, 1.6019e-07, 6.5193e-08, ..., 5.0291e-08, + 5.1223e-08, 3.3528e-08], + [ 1.9558e-08, 1.2107e-08, 1.5181e-07, ..., -3.1479e-07, + -3.2037e-07, 3.7253e-09]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0017, 0.0127, 0.0138, -0.0031, 0.0233, -0.0040, -0.0076, -0.0078, + 0.0162, -0.0138], device='cuda:0'), grad: tensor([ 7.4506e-08, 2.4512e-06, 2.5257e-06, 1.6745e-06, 1.0664e-06, + 1.9651e-06, 4.3679e-07, -2.3320e-06, -9.6858e-06, 1.8021e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 251.90, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4989 re_mapping 0.0043 re_causal 0.0139 /// teacc 99.04 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1747, -0.0346, 0.1780, ..., 0.0402, -0.1629, -0.1312], + [-0.1141, -0.1453, -0.0337, ..., -0.0975, -0.1440, -0.0646], + [ 0.2088, -0.0611, -0.1849, ..., -0.1584, -0.1497, -0.1853], + ..., + [-0.0606, -0.0209, -0.2003, ..., -0.0862, 0.0884, 0.0160], + [-0.1580, -0.1039, -0.0237, ..., -0.1543, -0.0247, -0.0255], + [-0.0596, -0.0834, -0.1015, ..., 0.0925, 0.0269, -0.1515]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 2.7940e-09, -2.0768e-07, ..., -2.0768e-07, + 4.5635e-08, 2.4214e-08], + [ 1.9837e-07, 1.7695e-08, 2.3283e-08, ..., 6.5193e-08, + 6.6124e-08, 1.8626e-09], + [-1.1176e-08, 1.3970e-08, 1.5832e-08, ..., 5.3085e-08, + 2.1420e-08, 1.8626e-09], + ..., + [ 2.5146e-07, 9.3132e-08, 5.0291e-08, ..., 7.1712e-08, + -3.0734e-08, 0.0000e+00], + [ 2.5798e-07, 2.1420e-08, 7.3574e-08, ..., 3.3528e-08, + 3.5018e-07, 1.3970e-08], + [ 7.1712e-08, 6.5193e-09, 9.4064e-08, ..., 7.8231e-08, + 8.3819e-08, 9.3132e-10]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0017, 0.0123, 0.0140, -0.0032, 0.0235, -0.0041, -0.0074, -0.0077, + 0.0161, -0.0138], device='cuda:0'), grad: tensor([-3.2783e-07, -1.4849e-05, 6.0536e-07, -6.8247e-06, 8.5589e-07, + 6.3702e-07, 8.7824e-07, 4.1649e-06, 1.2673e-05, 2.1346e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 251.59, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.4875 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.01 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1753, -0.0346, 0.1781, ..., 0.0388, -0.1633, -0.1313], + [-0.1144, -0.1459, -0.0337, ..., -0.0982, -0.1454, -0.0649], + [ 0.2114, -0.0612, -0.1856, ..., -0.1587, -0.1493, -0.1856], + ..., + [-0.0637, -0.0211, -0.2012, ..., -0.0865, 0.0885, 0.0162], + [-0.1582, -0.1041, -0.0239, ..., -0.1547, -0.0252, -0.0257], + [-0.0602, -0.0835, -0.1018, ..., 0.0931, 0.0269, -0.1516]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, -4.5076e-07, ..., -3.2689e-07, + 2.4214e-08, 1.2107e-08], + [ 2.4214e-08, 2.7940e-09, -2.2724e-07, ..., 1.6205e-07, + 5.9605e-08, -3.4459e-08], + [-2.5332e-07, 2.7940e-09, 5.2154e-08, ..., 1.0990e-07, + 3.6322e-08, 7.4506e-09], + ..., + [ 3.3528e-08, 3.7253e-09, 2.4214e-08, ..., 1.0617e-07, + -1.4901e-08, 2.7940e-09], + [ 2.0582e-07, 4.6566e-09, 1.1548e-07, ..., 1.1176e-07, + 1.2480e-07, 4.3772e-08], + [ 1.1176e-08, 3.7253e-09, 4.1071e-07, ..., 2.0489e-06, + 9.0990e-07, 2.7940e-09]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0005, 0.0123, 0.0155, -0.0025, 0.0236, -0.0046, -0.0070, -0.0084, + 0.0160, -0.0135], device='cuda:0'), grad: tensor([-5.4669e-07, -6.5845e-07, 1.2014e-07, -3.5241e-06, -5.2005e-06, + 3.4235e-06, 6.8918e-08, 3.0641e-07, 1.0803e-06, 4.9099e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 251.37, cls_loss 0.0011 cls_loss_mapping 0.0034 cls_loss_causal 0.5562 re_mapping 0.0041 re_causal 0.0136 /// teacc 98.95 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1769, -0.0347, 0.1790, ..., 0.0386, -0.1634, -0.1313], + [-0.1147, -0.1481, -0.0336, ..., -0.0985, -0.1473, -0.0651], + [ 0.2118, -0.0613, -0.1864, ..., -0.1591, -0.1501, -0.1859], + ..., + [-0.0638, -0.0216, -0.2030, ..., -0.0870, 0.0892, 0.0167], + [-0.1585, -0.1044, -0.0239, ..., -0.1547, -0.0254, -0.0259], + [-0.0597, -0.0844, -0.1044, ..., 0.0948, 0.0287, -0.1518]], + device='cuda:0'), grad: tensor([[ 3.9116e-08, 9.3132e-10, -1.1697e-06, ..., -3.7439e-07, + -1.3597e-07, 1.1176e-08], + [ 7.6368e-08, 9.3132e-10, 1.9558e-08, ..., 2.7940e-08, + 2.1420e-08, 2.7940e-09], + [-4.0047e-07, 0.0000e+00, 2.3283e-08, ..., 8.5682e-08, + 8.3819e-09, 1.5832e-08], + ..., + [ 1.4901e-07, 9.3132e-10, 9.3132e-09, ..., 5.2154e-08, + 1.1176e-08, -9.3132e-10], + [ 4.4703e-08, 6.5193e-09, 3.1665e-08, ..., 1.1083e-07, + -3.0734e-08, 8.3819e-09], + [ 1.3039e-08, 1.8626e-09, 1.4529e-07, ..., -8.0094e-08, + 1.8626e-08, 3.7253e-09]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0004, 0.0125, 0.0152, -0.0021, 0.0223, -0.0048, -0.0075, -0.0083, + 0.0160, -0.0121], device='cuda:0'), grad: tensor([-1.4370e-06, -1.0179e-06, -4.4052e-07, 4.1816e-07, -3.1386e-07, + -3.0454e-07, 1.4985e-06, 1.3430e-06, 1.7509e-07, 7.7300e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 251.30, cls_loss 0.0013 cls_loss_mapping 0.0034 cls_loss_causal 0.5219 re_mapping 0.0042 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1776, -0.0348, 0.1794, ..., 0.0387, -0.1637, -0.1314], + [-0.1155, -0.1505, -0.0338, ..., -0.0982, -0.1510, -0.0658], + [ 0.2121, -0.0616, -0.1873, ..., -0.1593, -0.1509, -0.1861], + ..., + [-0.0635, -0.0216, -0.2039, ..., -0.0873, 0.0925, 0.0182], + [-0.1586, -0.1046, -0.0241, ..., -0.1554, -0.0258, -0.0259], + [-0.0614, -0.0852, -0.1052, ..., 0.0939, 0.0270, -0.1519]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, 0.0000e+00, 2.8778e-07, ..., -9.3132e-10, + 3.6508e-07, 6.7987e-08], + [ 9.4064e-08, 9.3132e-10, 4.6566e-08, ..., 1.2107e-08, + 9.7789e-08, 1.4901e-08], + [-1.2545e-06, 0.0000e+00, 1.5181e-07, ..., 3.7253e-09, + 1.6671e-07, 2.8871e-08], + ..., + [ 9.3970e-07, 9.3132e-10, 2.0489e-08, ..., 1.4901e-08, + -2.7940e-09, 3.7253e-09], + [ 3.0734e-08, 0.0000e+00, 2.5332e-07, ..., 5.6811e-08, + 3.7346e-07, 7.7300e-08], + [ 4.6566e-09, 0.0000e+00, 1.7695e-07, ..., -3.7253e-07, + 1.1083e-07, 3.5390e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0005, 0.0137, 0.0146, -0.0020, 0.0233, -0.0018, -0.0104, -0.0083, + 0.0163, -0.0139], device='cuda:0'), grad: tensor([ 1.2182e-06, -8.4750e-08, -1.4249e-06, 2.5332e-07, 1.2452e-06, + 8.4788e-06, -1.1019e-05, 1.6587e-06, -2.2445e-07, -1.6019e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 251.46, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4995 re_mapping 0.0042 re_causal 0.0135 /// teacc 99.09 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1785, -0.0349, 0.1795, ..., 0.0387, -0.1640, -0.1315], + [-0.1162, -0.1516, -0.0339, ..., -0.0989, -0.1534, -0.0659], + [ 0.2105, -0.0648, -0.1877, ..., -0.1595, -0.1533, -0.1860], + ..., + [-0.0633, -0.0203, -0.2045, ..., -0.0882, 0.0935, 0.0183], + [-0.1589, -0.1049, -0.0243, ..., -0.1562, -0.0264, -0.0260], + [-0.0615, -0.0853, -0.1054, ..., 0.0940, 0.0271, -0.1519]], + device='cuda:0'), grad: tensor([[2.7940e-08, 1.0245e-08, 5.4948e-08, ..., 9.3132e-10, 4.2841e-08, + 5.4017e-08], + [5.8673e-08, 3.2596e-08, 1.0338e-07, ..., 7.4506e-09, 6.8918e-08, + 1.1642e-07], + [2.5146e-08, 6.1467e-08, 6.1467e-08, ..., 8.3819e-09, 5.1223e-08, + 8.3819e-08], + ..., + [3.0082e-07, 2.2352e-07, 3.7253e-09, ..., 2.7940e-09, 3.4459e-08, + 1.9185e-07], + [3.0734e-08, 1.6764e-08, 1.1548e-07, ..., 1.1083e-07, 9.1270e-08, + 1.0896e-07], + [9.3132e-09, 5.5879e-09, 1.7695e-08, ..., 2.0210e-07, 6.7987e-08, + 1.4901e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0005, 0.0135, 0.0125, -0.0003, 0.0232, -0.0016, -0.0106, -0.0081, + 0.0163, -0.0137], device='cuda:0'), grad: tensor([ 2.6822e-07, 4.8336e-07, 4.1630e-07, -2.5816e-06, -1.6671e-07, + 6.1654e-07, -1.7826e-06, 1.3085e-06, 6.8173e-07, 7.6182e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 251.34, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4754 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.11 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1787, -0.0349, 0.1802, ..., 0.0381, -0.1639, -0.1315], + [-0.1165, -0.1526, -0.0352, ..., -0.0997, -0.1539, -0.0660], + [ 0.2108, -0.0648, -0.1884, ..., -0.1597, -0.1537, -0.1858], + ..., + [-0.0633, -0.0205, -0.2058, ..., -0.0886, 0.0936, 0.0183], + [-0.1595, -0.1051, -0.0244, ..., -0.1565, -0.0268, -0.0261], + [-0.0617, -0.0854, -0.1060, ..., 0.0942, 0.0270, -0.1520]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -2.7940e-09, -1.2673e-05, ..., -6.0052e-06, + 2.0489e-08, 6.5193e-09], + [ 2.3749e-07, 3.7253e-09, 9.2201e-08, ..., 5.2154e-08, + 8.5682e-08, 2.7940e-09], + [-3.7998e-07, 2.7940e-09, 7.2271e-07, ..., 3.5297e-07, + 2.0489e-08, 2.7940e-09], + ..., + [ 5.9605e-08, 5.5879e-09, 2.3469e-07, ..., 1.2573e-07, + -9.3132e-08, 0.0000e+00], + [ 5.4017e-08, 6.5193e-09, 4.4443e-06, ..., 2.0750e-06, + 3.7625e-07, 1.1083e-07], + [ 1.3039e-08, 9.3132e-09, 5.3830e-06, ..., 2.5127e-06, + -3.2596e-08, 9.3132e-10]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0002, 0.0136, 0.0125, -0.0003, 0.0234, -0.0016, -0.0106, -0.0082, + 0.0161, -0.0136], device='cuda:0'), grad: tensor([-2.9102e-05, 1.2713e-06, 9.1083e-07, 9.6578e-07, 2.1141e-07, + 3.4831e-07, 1.9558e-06, 8.8476e-08, 1.0937e-05, 1.2390e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 251.34, cls_loss 0.0008 cls_loss_mapping 0.0031 cls_loss_causal 0.5189 re_mapping 0.0040 re_causal 0.0132 /// teacc 99.06 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1793, -0.0349, 0.1810, ..., 0.0383, -0.1640, -0.1316], + [-0.1182, -0.1536, -0.0362, ..., -0.1003, -0.1550, -0.0662], + [ 0.2117, -0.0648, -0.1894, ..., -0.1600, -0.1541, -0.1861], + ..., + [-0.0636, -0.0206, -0.2071, ..., -0.0891, 0.0939, 0.0195], + [-0.1599, -0.1053, -0.0246, ..., -0.1571, -0.0273, -0.0262], + [-0.0616, -0.0855, -0.1065, ..., 0.0942, 0.0270, -0.1520]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 3.7253e-09, -2.3376e-07, ..., -1.9651e-07, + 3.4831e-07, 2.0862e-07], + [ 5.8673e-08, 3.9116e-08, 1.8906e-07, ..., 7.8697e-07, + 9.3970e-07, 1.1269e-07], + [ 1.2107e-08, 2.1420e-08, 6.7055e-08, ..., 6.5193e-09, + 1.0151e-07, 3.6322e-08], + ..., + [ 1.1362e-07, 7.5437e-08, 5.5879e-09, ..., 2.8498e-07, + -8.1956e-07, 5.7742e-08], + [ 6.5193e-08, 4.4703e-08, 2.5947e-06, ..., 3.3528e-08, + 1.8133e-06, 1.2955e-06], + [ 1.6764e-08, 1.3039e-08, 2.2259e-07, ..., -1.5767e-06, + -4.1537e-07, 2.5146e-08]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0006, 0.0135, 0.0129, -0.0002, 0.0235, -0.0018, -0.0106, -0.0083, + 0.0158, -0.0136], device='cuda:0'), grad: tensor([ 2.9895e-07, 5.6550e-06, 6.0070e-07, -3.4496e-06, 1.2210e-06, + 2.0817e-05, -2.4840e-05, -3.1963e-06, 6.5081e-06, -3.6545e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 251.28, cls_loss 0.0010 cls_loss_mapping 0.0035 cls_loss_causal 0.4957 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1800, -0.0349, 0.1818, ..., 0.0385, -0.1639, -0.1315], + [-0.1201, -0.1547, -0.0366, ..., -0.1016, -0.1578, -0.0665], + [ 0.2129, -0.0649, -0.1903, ..., -0.1602, -0.1546, -0.1863], + ..., + [-0.0639, -0.0207, -0.2082, ..., -0.0898, 0.0951, 0.0202], + [-0.1603, -0.1054, -0.0248, ..., -0.1584, -0.0279, -0.0263], + [-0.0622, -0.0856, -0.1073, ..., 0.0950, 0.0282, -0.1521]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -9.6858e-08, ..., -5.3085e-08, + 2.7940e-08, 2.0489e-08], + [ 1.2107e-08, 1.8626e-09, 3.7253e-09, ..., 5.3085e-08, + 1.2489e-06, 1.9446e-06], + [-1.1548e-07, 9.3132e-10, 2.6077e-08, ..., 3.9116e-08, + 1.2107e-08, 5.5879e-09], + ..., + [ 1.3970e-08, 2.7940e-09, 9.3132e-10, ..., 7.0781e-08, + 6.4261e-08, 5.6811e-08], + [ 9.6858e-08, 9.3132e-10, 2.9802e-08, ..., 9.0338e-08, + 1.7351e-06, 2.6822e-06], + [ 1.8626e-09, 9.3132e-10, 5.4948e-08, ..., -5.2620e-07, + -2.8685e-07, 7.4506e-09]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0012, 0.0129, 0.0140, 0.0007, 0.0223, -0.0020, -0.0106, -0.0081, + 0.0154, -0.0126], device='cuda:0'), grad: tensor([-6.2399e-08, 1.0230e-05, -2.2352e-08, 2.6636e-07, 3.4180e-07, + -4.0084e-05, 1.5572e-05, 4.6846e-07, 1.4298e-05, -1.0300e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 249.11, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.5007 re_mapping 0.0040 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1804, -0.0349, 0.1821, ..., 0.0386, -0.1642, -0.1316], + [-0.1202, -0.1552, -0.0384, ..., -0.1021, -0.1606, -0.0690], + [ 0.2121, -0.0649, -0.1926, ..., -0.1605, -0.1587, -0.1884], + ..., + [-0.0640, -0.0208, -0.2090, ..., -0.0901, 0.0954, 0.0200], + [-0.1608, -0.1056, -0.0252, ..., -0.1591, -0.0298, -0.0266], + [-0.0623, -0.0857, -0.1079, ..., 0.0950, 0.0282, -0.1522]], + device='cuda:0'), grad: tensor([[ 8.3726e-07, 3.6322e-08, 2.1104e-06, ..., -3.7253e-08, + 1.5954e-06, 1.7695e-08], + [ 1.2945e-07, 1.5553e-07, 4.5635e-08, ..., 1.6764e-08, + 2.9244e-07, 4.1910e-08], + [-2.1644e-06, 3.6322e-08, -3.5111e-07, ..., 4.6566e-09, + 9.3132e-08, 1.3039e-08], + ..., + [ 2.0210e-07, 1.5087e-07, 3.5390e-08, ..., 3.0734e-08, + 3.1665e-08, 3.7253e-08], + [ 5.3179e-07, 9.4064e-08, 4.2003e-07, ..., 1.2107e-08, + 3.4180e-07, 5.0291e-08], + [ 4.6566e-08, 4.9360e-08, 5.6811e-08, ..., 1.2759e-07, + 1.7788e-07, 1.3970e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0013, 0.0124, 0.0124, 0.0006, 0.0223, -0.0016, -0.0106, -0.0081, + 0.0149, -0.0125], device='cuda:0'), grad: tensor([ 7.5810e-06, 2.1458e-06, -3.5651e-06, -1.0140e-05, 8.8476e-08, + 4.7162e-06, -7.1861e-06, 2.1644e-06, 2.1905e-06, 1.9744e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 246.70, cls_loss 0.0010 cls_loss_mapping 0.0028 cls_loss_causal 0.4842 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.19 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1810, -0.0350, 0.1823, ..., 0.0386, -0.1648, -0.1318], + [-0.1226, -0.1559, -0.0385, ..., -0.1020, -0.1616, -0.0691], + [ 0.2137, -0.0649, -0.1937, ..., -0.1611, -0.1598, -0.1892], + ..., + [-0.0652, -0.0210, -0.2102, ..., -0.0906, 0.0957, 0.0202], + [-0.1613, -0.1057, -0.0247, ..., -0.1597, -0.0277, -0.0268], + [-0.0623, -0.0858, -0.1082, ..., 0.0949, 0.0281, -0.1523]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 3.7253e-09, ..., 2.7940e-09, + 2.7940e-08, 3.7253e-09], + [ 2.7940e-09, 3.7253e-09, -1.8626e-09, ..., 5.5879e-09, + 6.3330e-08, 4.6566e-09], + [-0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + 1.4901e-08, 9.3132e-10], + ..., + [ 3.7253e-09, 5.5879e-09, 1.8626e-09, ..., 9.3132e-09, + -5.0943e-07, 6.5193e-09], + [ 2.7940e-09, 8.3819e-09, -6.5193e-09, ..., 3.1665e-08, + 4.5635e-08, 1.3970e-08], + [ 1.8626e-09, 2.7940e-09, 2.7940e-09, ..., -6.2399e-08, + 3.7998e-07, 8.3819e-09]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0012, 0.0121, 0.0140, 0.0006, 0.0224, -0.0018, -0.0106, -0.0086, + 0.0165, -0.0127], device='cuda:0'), grad: tensor([ 1.7136e-07, -2.9467e-06, 1.9483e-06, 3.7402e-06, 9.0338e-08, + -4.7535e-06, 4.1164e-07, -6.7055e-07, 4.0606e-07, 1.5888e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 246.81, cls_loss 0.0007 cls_loss_mapping 0.0027 cls_loss_causal 0.4968 re_mapping 0.0044 re_causal 0.0138 /// teacc 99.08 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1816, -0.0350, 0.1825, ..., 0.0387, -0.1652, -0.1320], + [-0.1231, -0.1565, -0.0385, ..., -0.1022, -0.1629, -0.0693], + [ 0.2143, -0.0649, -0.1946, ..., -0.1613, -0.1606, -0.1896], + ..., + [-0.0655, -0.0213, -0.2109, ..., -0.0906, 0.0962, 0.0206], + [-0.1620, -0.1063, -0.0250, ..., -0.1602, -0.0283, -0.0270], + [-0.0625, -0.0859, -0.1086, ..., 0.0949, 0.0281, -0.1524]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2538e-07, ..., -1.9185e-07, + 5.0291e-08, 5.5879e-09], + [ 1.8626e-09, 1.8626e-09, 6.5193e-09, ..., -2.5146e-08, + 2.1886e-07, 9.3132e-10], + [-2.7940e-09, 0.0000e+00, 1.4901e-08, ..., 9.3132e-09, + 2.9802e-08, 1.8626e-09], + ..., + [ 3.7253e-09, 9.3132e-10, 1.8626e-09, ..., 3.5390e-08, + -3.9861e-07, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 2.5146e-08, ..., 5.8673e-08, + 1.3225e-07, 3.7253e-09], + [ 9.3132e-09, 8.3819e-09, 2.2631e-07, ..., 1.6019e-07, + 8.1956e-08, 9.3132e-10]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0013, 0.0117, 0.0142, 0.0007, 0.0224, -0.0019, -0.0105, -0.0078, + 0.0161, -0.0130], device='cuda:0'), grad: tensor([-1.6671e-07, 3.7625e-07, 2.0582e-07, 5.8115e-07, 1.2387e-07, + -2.8685e-07, -8.0839e-07, -1.0254e-06, 1.8347e-07, 8.2608e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 246.11, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.5036 re_mapping 0.0041 re_causal 0.0132 /// teacc 99.06 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1825, -0.0351, 0.1829, ..., 0.0390, -0.1657, -0.1321], + [-0.1236, -0.1575, -0.0386, ..., -0.1025, -0.1640, -0.0694], + [ 0.2148, -0.0649, -0.1953, ..., -0.1616, -0.1619, -0.1896], + ..., + [-0.0659, -0.0219, -0.2121, ..., -0.0933, 0.0958, 0.0207], + [-0.1628, -0.1067, -0.0249, ..., -0.1605, -0.0283, -0.0272], + [-0.0630, -0.0860, -0.1096, ..., 0.0950, 0.0283, -0.1524]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8405e-07, ..., -2.4401e-07, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, 3.7253e-09, ..., 5.5879e-09, + 5.5879e-09, 0.0000e+00], + [-2.2352e-08, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 1.8626e-08, 2.7940e-09, 4.6566e-09, ..., 6.5193e-09, + -1.0245e-08, 9.3132e-10], + [ 1.8626e-09, 1.8626e-09, 5.5879e-09, ..., 4.6566e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.0862e-07, ..., 2.0396e-07, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0014, 0.0114, 0.0141, 0.0006, 0.0225, -0.0018, -0.0105, -0.0078, + 0.0162, -0.0130], device='cuda:0'), grad: tensor([-4.6194e-07, -9.6858e-07, 4.1164e-07, 5.4017e-08, 7.4506e-09, + 5.8673e-08, 2.9244e-07, 5.5134e-07, -5.4855e-07, 5.9325e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 246.69, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.5134 re_mapping 0.0040 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1828, -0.0352, 0.1831, ..., 0.0388, -0.1660, -0.1322], + [-0.1240, -0.1582, -0.0386, ..., -0.1024, -0.1651, -0.0695], + [ 0.2149, -0.0649, -0.1959, ..., -0.1619, -0.1629, -0.1902], + ..., + [-0.0655, -0.0224, -0.2127, ..., -0.0937, 0.0964, 0.0212], + [-0.1631, -0.1071, -0.0252, ..., -0.1611, -0.0286, -0.0273], + [-0.0633, -0.0862, -0.1099, ..., 0.0951, 0.0282, -0.1524]], + device='cuda:0'), grad: tensor([[ 2.5146e-08, 5.6811e-08, 2.7940e-08, ..., 1.2107e-08, + 9.4064e-08, 9.3132e-10], + [ 1.3039e-08, 8.3819e-09, 5.5879e-09, ..., 4.9360e-08, + 3.1665e-08, 9.3132e-10], + [-1.2852e-07, 2.2352e-08, 4.6566e-09, ..., 2.1420e-08, + 2.3283e-08, 0.0000e+00], + ..., + [ 1.8626e-08, 3.3807e-07, 6.5193e-09, ..., 1.6261e-06, + 9.1270e-07, -9.3132e-10], + [ 3.7253e-08, 5.3085e-08, 8.1956e-08, ..., 5.5879e-08, + 8.7544e-08, 9.3132e-10], + [ 8.3819e-09, 1.3039e-08, 1.5832e-08, ..., -1.9241e-06, + -8.9314e-07, 0.0000e+00]], device='cuda:0') +Epoch 215, bias, value: tensor([ 1.2109e-03, 1.2355e-02, 1.3674e-02, 7.1428e-05, 2.2415e-02, + -1.7808e-03, -1.0427e-02, -7.6227e-03, 1.4705e-02, -1.3002e-02], + device='cuda:0'), grad: tensor([ 7.2457e-07, -1.7975e-07, 4.0047e-08, 1.3448e-06, 1.2899e-06, + -7.4916e-06, -2.8033e-07, 7.7114e-06, 8.3912e-07, -4.0159e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 246.85, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.5196 re_mapping 0.0039 re_causal 0.0127 /// teacc 99.10 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1834, -0.0356, 0.1834, ..., 0.0390, -0.1666, -0.1323], + [-0.1243, -0.1593, -0.0387, ..., -0.1040, -0.1703, -0.0700], + [ 0.2158, -0.0649, -0.1966, ..., -0.1617, -0.1634, -0.1904], + ..., + [-0.0662, -0.0228, -0.2142, ..., -0.0948, 0.1004, 0.0213], + [-0.1633, -0.1054, -0.0252, ..., -0.1615, -0.0307, -0.0274], + [-0.0649, -0.0866, -0.1106, ..., 0.0952, 0.0274, -0.1525]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.7253e-09, -1.0151e-07, ..., -1.9558e-08, + 5.5134e-07, 6.6310e-07], + [ 1.9558e-08, 9.5926e-08, 3.7253e-09, ..., 2.1048e-07, + 2.3022e-05, 2.9147e-05], + [-2.2352e-08, 2.7940e-08, 1.5832e-08, ..., 1.2107e-08, + 8.2981e-07, 9.7323e-07], + ..., + [-4.0978e-08, -2.3935e-07, 1.8626e-09, ..., 6.9570e-07, + -4.6566e-09, 2.4587e-07], + [ 1.8626e-09, 5.5879e-09, 1.5832e-08, ..., 1.3225e-07, + 4.0159e-06, 5.0403e-06], + [ 9.3132e-10, 6.5193e-09, 6.7987e-08, ..., -1.1940e-06, + -5.8766e-07, 1.4156e-07]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0013, 0.0111, 0.0139, -0.0003, 0.0223, -0.0018, -0.0105, -0.0050, + 0.0136, -0.0137], device='cuda:0'), grad: tensor([ 2.7865e-06, 1.2350e-04, 4.3362e-06, 1.3150e-06, 9.3952e-06, + 7.8380e-05, -2.3949e-04, 1.7565e-06, 2.1636e-05, -3.5968e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 246.47, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4891 re_mapping 0.0039 re_causal 0.0131 /// teacc 98.99 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1838, -0.0357, 0.1833, ..., 0.0390, -0.1684, -0.1340], + [-0.1248, -0.1609, -0.0387, ..., -0.1047, -0.1717, -0.0715], + [ 0.2162, -0.0649, -0.1971, ..., -0.1619, -0.1637, -0.1910], + ..., + [-0.0661, -0.0231, -0.2153, ..., -0.0953, 0.1006, 0.0223], + [-0.1637, -0.1053, -0.0254, ..., -0.1621, -0.0309, -0.0276], + [-0.0651, -0.0868, -0.1110, ..., 0.0953, 0.0275, -0.1527]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 1.2107e-08, ..., 6.5193e-09, + 5.6811e-08, 4.3772e-08], + [ 5.5879e-09, 8.3819e-09, -1.2964e-06, ..., -1.2489e-06, + 3.3528e-08, 2.9802e-08], + [ 1.1176e-08, 1.9558e-08, 3.7253e-08, ..., 2.0489e-08, + 9.3132e-09, 8.3819e-09], + ..., + [ 1.1176e-08, 1.6764e-08, 7.1712e-08, ..., 1.7136e-07, + 2.3283e-08, 9.3132e-10], + [ 1.2107e-08, 2.5146e-08, 5.8580e-07, ..., 5.6345e-07, + 9.9652e-08, 6.1467e-08], + [ 3.7253e-09, 8.3819e-09, 6.1654e-07, ..., -2.9616e-07, + -2.0396e-07, 4.6566e-09]], device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0009, 0.0107, 0.0138, -0.0007, 0.0224, -0.0018, -0.0103, -0.0047, + 0.0136, -0.0135], device='cuda:0'), grad: tensor([ 4.8615e-07, -8.8811e-06, 3.2876e-07, -4.1351e-07, 3.2447e-06, + 8.9500e-07, -7.1526e-07, 1.0552e-06, 3.0883e-06, 9.1363e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 246.63, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.5218 re_mapping 0.0041 re_causal 0.0137 /// teacc 99.06 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1837, -0.0357, 0.1844, ..., 0.0395, -0.1687, -0.1342], + [-0.1258, -0.1617, -0.0387, ..., -0.1049, -0.1723, -0.0718], + [ 0.2163, -0.0650, -0.1981, ..., -0.1624, -0.1657, -0.1913], + ..., + [-0.0664, -0.0233, -0.2165, ..., -0.0957, 0.1009, 0.0227], + [-0.1646, -0.1055, -0.0257, ..., -0.1627, -0.0309, -0.0277], + [-0.0655, -0.0870, -0.1136, ..., 0.0951, 0.0275, -0.1529]], + device='cuda:0'), grad: tensor([[ 1.3318e-07, 1.1828e-07, -3.4366e-07, ..., -2.6450e-07, + 5.5879e-09, 2.7940e-09], + [ 1.0245e-08, 2.7940e-09, 1.2107e-08, ..., 6.5193e-09, + 5.1223e-08, 1.8626e-09], + [-3.7253e-09, 1.2107e-08, 4.0047e-08, ..., 1.6764e-08, + 7.4506e-09, 1.8626e-09], + ..., + [ 3.3528e-08, 9.3132e-09, 2.0489e-08, ..., 1.8626e-08, + -2.6729e-07, 9.3132e-10], + [-6.5193e-09, 3.4459e-08, 1.0431e-07, ..., 3.1665e-08, + 3.2596e-08, 4.6566e-09], + [ 3.4459e-08, 3.0734e-08, 3.7439e-07, ..., 1.5832e-07, + 4.6566e-09, 2.7940e-09]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0016, 0.0103, 0.0135, -0.0008, 0.0224, -0.0017, -0.0103, -0.0045, + 0.0136, -0.0138], device='cuda:0'), grad: tensor([ 1.9558e-07, -2.6673e-06, 1.2899e-06, -3.7905e-07, 1.6950e-07, + 2.8405e-07, 1.6671e-07, 1.8463e-05, -1.8626e-05, 1.0701e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 246.42, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4871 re_mapping 0.0041 re_causal 0.0130 /// teacc 99.10 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1842, -0.0358, 0.1847, ..., 0.0396, -0.1689, -0.1344], + [-0.1253, -0.1620, -0.0387, ..., -0.1066, -0.1740, -0.0719], + [ 0.2164, -0.0650, -0.2002, ..., -0.1626, -0.1662, -0.1939], + ..., + [-0.0668, -0.0234, -0.2178, ..., -0.0976, 0.1007, 0.0230], + [-0.1649, -0.1056, -0.0260, ..., -0.1632, -0.0311, -0.0280], + [-0.0656, -0.0871, -0.1139, ..., 0.0949, 0.0274, -0.1530]], + device='cuda:0'), grad: tensor([[ 3.3341e-07, 0.0000e+00, -1.1269e-07, ..., -6.7987e-08, + 3.9116e-08, 1.2107e-08], + [ 6.3982e-07, 1.8626e-09, 2.7940e-09, ..., -1.3132e-07, + 8.9034e-07, 3.7253e-09], + [-6.2473e-06, 3.7253e-09, 5.5879e-09, ..., 2.7940e-09, + 6.1467e-08, 2.7940e-09], + ..., + [ 2.3190e-06, 1.8626e-09, 5.5879e-09, ..., 3.5390e-08, + -2.0210e-06, -5.5879e-09], + [ 4.2468e-07, 0.0000e+00, 2.5146e-08, ..., 1.2107e-08, + 2.1420e-08, 1.0245e-08], + [ 4.9360e-08, 9.3132e-10, 6.3330e-08, ..., 1.3970e-08, + 8.4378e-07, 3.7253e-09]], device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0016, 0.0110, 0.0123, -0.0011, 0.0228, -0.0017, -0.0102, -0.0054, + 0.0136, -0.0136], device='cuda:0'), grad: tensor([ 5.6624e-07, 2.4624e-06, -1.0826e-05, 1.7341e-06, 2.6114e-06, + 6.7800e-07, 9.7044e-07, -2.3656e-06, 8.6240e-07, 3.2838e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 246.11, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5103 re_mapping 0.0039 re_causal 0.0128 /// teacc 99.17 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1847, -0.0358, 0.1850, ..., 0.0397, -0.1689, -0.1344], + [-0.1250, -0.1627, -0.0387, ..., -0.1064, -0.1754, -0.0722], + [ 0.2167, -0.0650, -0.2010, ..., -0.1629, -0.1667, -0.1937], + ..., + [-0.0671, -0.0238, -0.2202, ..., -0.0983, 0.1011, 0.0230], + [-0.1659, -0.1057, -0.0261, ..., -0.1639, -0.0312, -0.0283], + [-0.0661, -0.0872, -0.1141, ..., 0.0951, 0.0276, -0.1531]], + device='cuda:0'), grad: tensor([[ 3.0734e-08, 8.3819e-09, -1.2573e-07, ..., -6.7055e-08, + 1.3039e-08, 0.0000e+00], + [ 8.2888e-08, 2.3283e-08, 2.4214e-08, ..., 2.5146e-08, + 7.6368e-08, 9.3132e-10], + [ 4.6566e-08, 1.9558e-08, 1.1269e-07, ..., 2.8871e-08, + 7.9162e-08, 0.0000e+00], + ..., + [ 1.3039e-07, 3.6322e-08, 6.5193e-09, ..., -8.3819e-09, + -2.6636e-07, -1.8626e-09], + [ 1.2945e-07, 3.6322e-08, -9.4902e-07, ..., 1.6019e-07, + -6.1840e-07, 9.3132e-10], + [ 1.0710e-07, 2.3283e-08, 9.7789e-08, ..., -2.4121e-06, + -7.1991e-07, 9.3132e-10]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0018, 0.0114, 0.0110, -0.0022, 0.0226, -0.0016, -0.0102, -0.0049, + 0.0135, -0.0136], device='cuda:0'), grad: tensor([ 3.3528e-08, 8.4005e-07, 1.4417e-06, -5.9940e-06, 4.3213e-06, + 4.4890e-06, 7.0557e-06, -4.4983e-07, -8.9332e-06, -2.8461e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 246.11, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4976 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.08 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1857, -0.0359, 0.1854, ..., 0.0398, -0.1692, -0.1345], + [-0.1251, -0.1631, -0.0387, ..., -0.1065, -0.1758, -0.0723], + [ 0.2172, -0.0650, -0.2019, ..., -0.1633, -0.1669, -0.1943], + ..., + [-0.0675, -0.0240, -0.2217, ..., -0.0990, 0.1010, 0.0230], + [-0.1665, -0.1058, -0.0264, ..., -0.1644, -0.0314, -0.0285], + [-0.0665, -0.0874, -0.1146, ..., 0.0953, 0.0278, -0.1533]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 0.0000e+00, -1.8450e-06, ..., -7.9535e-07, + 2.7567e-07, 3.7253e-09], + [ 1.0990e-07, 0.0000e+00, 2.6077e-08, ..., 2.6077e-08, + 5.7183e-07, 9.3132e-10], + [-8.4098e-07, 9.3132e-10, 4.0047e-08, ..., 1.3970e-08, + 4.1910e-08, 1.8626e-09], + ..., + [ 1.0803e-07, 9.3132e-10, 1.5832e-08, ..., 4.2841e-08, + -5.1595e-07, 9.3132e-10], + [ 2.0210e-07, 7.4506e-09, 5.3085e-08, ..., 3.0361e-07, + -5.5879e-07, 1.1176e-08], + [ 1.8626e-08, 0.0000e+00, 7.5903e-07, ..., -1.2107e-07, + -1.5367e-07, 0.0000e+00]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0019, 0.0115, 0.0112, -0.0028, 0.0224, -0.0017, -0.0098, -0.0052, + 0.0132, -0.0134], device='cuda:0'), grad: tensor([-9.1270e-08, 9.0823e-06, -1.0375e-06, -1.4128e-06, 5.6718e-07, + 9.7603e-07, 2.1867e-06, -7.9721e-06, -2.8461e-06, 5.2806e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 246.32, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5060 re_mapping 0.0038 re_causal 0.0124 /// teacc 99.07 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1887, -0.0361, 0.1851, ..., 0.0398, -0.1704, -0.1350], + [-0.1257, -0.1634, -0.0389, ..., -0.1073, -0.1790, -0.0726], + [ 0.2179, -0.0650, -0.2041, ..., -0.1636, -0.1677, -0.1960], + ..., + [-0.0682, -0.0243, -0.2234, ..., -0.1005, 0.1012, 0.0235], + [-0.1668, -0.1059, -0.0264, ..., -0.1648, -0.0310, -0.0284], + [-0.0667, -0.0874, -0.1150, ..., 0.0954, 0.0280, -0.1535]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5917e-06, ..., -4.0047e-08, + 3.7700e-06, 3.8631e-06], + [ 0.0000e+00, 0.0000e+00, 7.8231e-08, ..., 5.5879e-08, + 8.0094e-08, 5.0291e-08], + [ 0.0000e+00, 0.0000e+00, 5.3085e-08, ..., 1.2107e-08, + 3.7253e-08, 3.0734e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 3.8184e-08, + 8.3819e-09, 1.1176e-08], + [ 9.3132e-10, 0.0000e+00, 4.1164e-07, ..., 1.4249e-07, + 3.4552e-07, 2.9616e-07], + [ 9.3132e-10, 0.0000e+00, 8.8476e-08, ..., -6.9942e-07, + -2.2445e-07, 4.8429e-08]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0008, 0.0108, 0.0114, -0.0020, 0.0224, -0.0020, -0.0096, -0.0049, + 0.0138, -0.0133], device='cuda:0'), grad: tensor([ 1.2219e-05, 3.2131e-07, 1.6857e-07, 7.0781e-08, 1.2880e-06, + 2.5965e-06, -1.6123e-05, 1.2480e-07, 1.0459e-06, -1.7136e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 246.43, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.5171 re_mapping 0.0041 re_causal 0.0131 /// teacc 99.11 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1893, -0.0361, 0.1853, ..., 0.0399, -0.1709, -0.1352], + [-0.1263, -0.1639, -0.0390, ..., -0.1081, -0.1806, -0.0729], + [ 0.2185, -0.0650, -0.2046, ..., -0.1638, -0.1681, -0.1962], + ..., + [-0.0686, -0.0247, -0.2253, ..., -0.1003, 0.1022, 0.0240], + [-0.1670, -0.1060, -0.0267, ..., -0.1654, -0.0308, -0.0286], + [-0.0669, -0.0877, -0.1153, ..., 0.0948, 0.0275, -0.1536]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 2.7940e-09, -5.5693e-07, ..., 7.4506e-09, + 1.8347e-07, 2.0489e-08], + [ 8.1956e-08, 1.0245e-08, 1.8626e-08, ..., 2.7381e-07, + 2.5332e-07, 1.2107e-08], + [-4.0419e-07, 2.7008e-08, 2.0489e-08, ..., 6.8918e-08, + 5.6811e-08, 4.6566e-09], + ..., + [ 1.1921e-07, 3.0734e-08, 1.2107e-08, ..., 1.4435e-07, + -4.2561e-07, 9.3132e-10], + [ 1.8068e-07, 2.7008e-08, 9.4995e-08, ..., 3.9563e-06, + 2.9430e-06, 7.4506e-09], + [ 1.1176e-08, 3.7253e-09, 1.0617e-07, ..., -7.2010e-06, + -5.4836e-06, 1.8626e-09]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0009, 0.0108, 0.0116, -0.0016, 0.0231, -0.0022, -0.0095, -0.0050, + 0.0142, -0.0140], device='cuda:0'), grad: tensor([-1.3318e-07, 1.2908e-06, -3.1199e-07, 2.0061e-06, 6.8173e-06, + 1.6289e-06, 5.5879e-07, -7.4320e-07, 1.2159e-05, -2.3291e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 246.31, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4871 re_mapping 0.0039 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1896, -0.0361, 0.1860, ..., 0.0402, -0.1712, -0.1355], + [-0.1268, -0.1642, -0.0391, ..., -0.1088, -0.1817, -0.0736], + [ 0.2200, -0.0650, -0.2048, ..., -0.1641, -0.1681, -0.1947], + ..., + [-0.0704, -0.0248, -0.2277, ..., -0.1008, 0.1023, 0.0223], + [-0.1676, -0.1060, -0.0270, ..., -0.1666, -0.0312, -0.0284], + [-0.0679, -0.0877, -0.1157, ..., 0.0949, 0.0279, -0.1538]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 0.0000e+00, -2.3283e-07, ..., -1.2573e-07, + 2.7940e-08, 1.0245e-08], + [ 2.3432e-06, 9.3132e-10, 1.1176e-08, ..., 1.5832e-08, + 1.0747e-06, 1.8626e-09], + [ 1.3905e-06, 0.0000e+00, 1.0245e-08, ..., 4.0978e-08, + 6.3237e-07, 3.7253e-09], + ..., + [-4.4629e-06, 9.3132e-10, 5.5879e-09, ..., 3.0547e-07, + -1.7332e-06, 1.8626e-09], + [ 2.8871e-08, 0.0000e+00, 2.8871e-08, ..., 6.5193e-08, + 4.9360e-08, 1.3039e-08], + [ 2.7008e-08, 9.3132e-10, 1.4808e-07, ..., -9.6951e-07, + -1.6941e-06, 2.7940e-09]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0013, 0.0110, 0.0123, -0.0014, 0.0230, -0.0022, -0.0095, -0.0057, + 0.0143, -0.0139], device='cuda:0'), grad: tensor([-7.3574e-08, 8.2925e-06, 5.7444e-06, 2.5220e-06, 1.9893e-06, + -1.2014e-07, -5.1316e-07, -1.6287e-05, 6.3051e-07, -2.2314e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 246.41, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.5024 re_mapping 0.0040 re_causal 0.0134 /// teacc 99.07 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1899, -0.0359, 0.1883, ..., 0.0409, -0.1715, -0.1356], + [-0.1260, -0.1645, -0.0391, ..., -0.1090, -0.1825, -0.0737], + [ 0.2202, -0.0650, -0.2057, ..., -0.1643, -0.1684, -0.1947], + ..., + [-0.0710, -0.0249, -0.2298, ..., -0.1013, 0.1025, 0.0223], + [-0.1679, -0.1062, -0.0274, ..., -0.1674, -0.0313, -0.0285], + [-0.0680, -0.0878, -0.1174, ..., 0.0943, 0.0272, -0.1539]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, -1.6857e-06, ..., -9.6858e-07, + 4.0047e-08, 1.4901e-08], + [ 4.2096e-07, 0.0000e+00, 1.3039e-08, ..., 1.0431e-07, + 6.8266e-07, 3.8743e-07], + [ 2.5425e-07, 0.0000e+00, 9.7789e-08, ..., 5.8673e-08, + 6.3516e-07, 3.6880e-07], + ..., + [-1.2955e-06, 9.3132e-10, 2.7940e-08, ..., 2.4214e-08, + -2.2817e-06, -1.2992e-06], + [ 5.5879e-08, 0.0000e+00, 7.3574e-08, ..., 4.0047e-08, + 1.1176e-08, 1.3970e-08], + [ 2.7940e-08, 9.3132e-10, 1.3895e-06, ..., 8.6613e-07, + 6.1467e-08, 2.9802e-08]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0027, 0.0116, 0.0115, -0.0007, 0.0237, -0.0025, -0.0097, -0.0058, + 0.0144, -0.0146], device='cuda:0'), grad: tensor([-3.2745e-06, 3.8072e-06, 2.7269e-06, 5.0291e-07, 3.0510e-06, + -7.0687e-07, 9.7230e-07, -9.2834e-06, -1.3318e-06, 3.5204e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 246.20, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4964 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.02 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1915, -0.0360, 0.1888, ..., 0.0410, -0.1720, -0.1358], + [-0.1262, -0.1653, -0.0391, ..., -0.1098, -0.1835, -0.0738], + [ 0.2207, -0.0650, -0.2066, ..., -0.1647, -0.1688, -0.1949], + ..., + [-0.0712, -0.0252, -0.2344, ..., -0.1018, 0.1027, 0.0230], + [-0.1683, -0.1063, -0.0277, ..., -0.1683, -0.0315, -0.0287], + [-0.0683, -0.0881, -0.1181, ..., 0.0944, 0.0274, -0.1541]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, -3.1665e-08, ..., -3.7253e-08, + 3.2596e-08, 2.1420e-08], + [ 4.6566e-09, 9.3132e-10, -4.0326e-07, ..., 1.5832e-08, + 5.1223e-08, 2.7940e-09], + [ 3.3528e-08, 1.8626e-09, 1.4901e-08, ..., 3.0454e-07, + 1.1176e-08, 5.7742e-08], + ..., + [ 8.3819e-09, 3.7253e-09, 4.7497e-08, ..., 1.6764e-08, + -9.9652e-08, 1.8626e-09], + [ 3.7253e-09, 9.3132e-10, 1.5646e-07, ..., 1.3039e-08, + 6.2399e-08, 4.0978e-08], + [ 2.7940e-09, 0.0000e+00, 3.4459e-08, ..., -1.8533e-07, + 2.0489e-08, 4.6566e-09]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0029, 0.0120, 0.0114, -0.0007, 0.0235, -0.0025, -0.0097, -0.0062, + 0.0144, -0.0145], device='cuda:0'), grad: tensor([ 6.4261e-08, -1.4324e-06, 5.9325e-07, 1.9465e-07, -4.7311e-07, + -2.4214e-08, 4.2375e-07, -1.3690e-07, 7.7579e-07, 9.3132e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 246.10, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5311 re_mapping 0.0038 re_causal 0.0126 /// teacc 99.02 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1918, -0.0361, 0.1890, ..., 0.0411, -0.1724, -0.1359], + [-0.1271, -0.1660, -0.0394, ..., -0.1100, -0.1842, -0.0740], + [ 0.2214, -0.0650, -0.2074, ..., -0.1652, -0.1701, -0.1951], + ..., + [-0.0714, -0.0254, -0.2353, ..., -0.1021, 0.1032, 0.0233], + [-0.1687, -0.1063, -0.0280, ..., -0.1697, -0.0317, -0.0290], + [-0.0683, -0.0882, -0.1183, ..., 0.0944, 0.0274, -0.1542]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.8626e-09, -1.1176e-08, ..., -1.1176e-08, + 9.6858e-08, 1.3970e-08], + [ 2.0489e-08, 9.3132e-10, 2.7940e-09, ..., 5.5879e-09, + 2.7381e-07, 6.5193e-09], + [-1.5926e-07, 9.3132e-10, 1.8626e-09, ..., 1.8626e-09, + 1.1828e-07, 8.3819e-09], + ..., + [ 1.4901e-08, 9.3132e-10, 9.3132e-10, ..., 8.3819e-09, + -1.8962e-06, -1.0058e-07], + [ 1.0245e-08, 4.9360e-08, 1.9558e-08, ..., 4.7497e-08, + 4.3493e-07, 1.4901e-08], + [ 2.7940e-09, 1.8626e-09, 1.3970e-08, ..., -8.6613e-08, + 9.4250e-07, 5.9605e-08]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0029, 0.0121, 0.0113, -0.0007, 0.0236, -0.0025, -0.0097, -0.0061, + 0.0145, -0.0146], device='cuda:0'), grad: tensor([ 3.9488e-07, 9.8068e-07, 2.2259e-07, 1.3141e-06, 2.7195e-07, + -3.7290e-06, 1.9185e-07, -6.8918e-06, 3.7812e-06, 3.4608e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 246.74, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4831 re_mapping 0.0037 re_causal 0.0116 /// teacc 99.16 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.1922, -0.0361, 0.1898, ..., 0.0415, -0.1731, -0.1361], + [-0.1272, -0.1668, -0.0401, ..., -0.1109, -0.1859, -0.0743], + [ 0.2218, -0.0651, -0.2095, ..., -0.1660, -0.1706, -0.1955], + ..., + [-0.0714, -0.0257, -0.2372, ..., -0.1026, 0.1036, 0.0236], + [-0.1691, -0.1072, -0.0260, ..., -0.1714, -0.0296, -0.0270], + [-0.0686, -0.0884, -0.1187, ..., 0.0943, 0.0273, -0.1544]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, -1.5553e-07, ..., -2.9802e-08, + 7.4506e-08, 2.6077e-08], + [ 2.5146e-08, 2.7940e-09, 3.5390e-08, ..., 3.1665e-08, + 2.5798e-07, 1.0245e-08], + [-1.3690e-07, 1.8626e-09, 2.1420e-08, ..., 4.6566e-09, + 1.6764e-08, 2.7940e-09], + ..., + [ 6.7055e-08, 1.1176e-08, 3.2596e-08, ..., 1.8626e-08, + -2.3916e-06, -6.5193e-09], + [ 2.0489e-08, 3.1665e-08, 1.7695e-07, ..., 9.6858e-08, + 1.5926e-07, 1.3039e-08], + [ 2.7940e-09, 9.3132e-09, 2.4214e-08, ..., -7.8976e-07, + 1.8580e-06, 1.0245e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0035, 0.0131, 0.0105, -0.0013, 0.0237, -0.0025, -0.0102, -0.0067, + 0.0165, -0.0147], device='cuda:0'), grad: tensor([-6.7987e-08, 4.3120e-07, -1.3411e-07, 3.9712e-06, 1.4435e-06, + -5.0068e-06, 5.1223e-08, -7.2829e-06, 1.1101e-06, 5.4352e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 246.46, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4733 re_mapping 0.0038 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.1950, -0.0365, 0.1899, ..., 0.0416, -0.1745, -0.1368], + [-0.1276, -0.1700, -0.0401, ..., -0.1116, -0.1882, -0.0746], + [ 0.2225, -0.0651, -0.2101, ..., -0.1664, -0.1710, -0.1957], + ..., + [-0.0716, -0.0261, -0.2385, ..., -0.1033, 0.1042, 0.0241], + [-0.1694, -0.1085, -0.0264, ..., -0.1729, -0.0301, -0.0275], + [-0.0693, -0.0887, -0.1190, ..., 0.0938, 0.0270, -0.1546]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.0094e-08, -1.1921e-07, ..., 2.2352e-08, + 1.4529e-07, 7.5437e-08], + [ 1.8626e-09, 2.1420e-07, 1.2107e-08, ..., 9.8720e-08, + 3.5856e-07, 6.7987e-08], + [ 1.8626e-09, 1.2759e-07, 7.4506e-09, ..., 3.6322e-07, + 5.7835e-07, 1.9558e-08], + ..., + [ 4.6566e-09, 1.8515e-06, 3.7253e-09, ..., 1.4063e-07, + -1.2759e-06, -4.3958e-07], + [ 1.8626e-09, 1.1735e-07, 1.4994e-07, ..., 8.0094e-08, + -4.8522e-07, -2.7008e-08], + [ 0.0000e+00, 6.8918e-08, 1.6484e-07, ..., 1.8803e-06, + 3.8370e-07, 8.3819e-09]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0032, 0.0121, 0.0107, -0.0016, 0.0245, -0.0024, -0.0101, -0.0059, + 0.0163, -0.0154], device='cuda:0'), grad: tensor([ 8.7731e-07, 3.3733e-06, 2.3499e-05, -1.3128e-05, -6.8136e-06, + 6.1579e-06, 8.7172e-07, 5.0217e-06, -2.8208e-05, 8.2850e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 246.71, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5037 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.09 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.1958, -0.0367, 0.1937, ..., 0.0431, -0.1759, -0.1386], + [-0.1277, -0.1716, -0.0406, ..., -0.1097, -0.1888, -0.0748], + [ 0.2245, -0.0650, -0.2106, ..., -0.1666, -0.1714, -0.1960], + ..., + [-0.0719, -0.0266, -0.2399, ..., -0.1059, 0.1043, 0.0246], + [-0.1701, -0.1090, -0.0268, ..., -0.1740, -0.0305, -0.0280], + [-0.0726, -0.0889, -0.1254, ..., 0.0921, 0.0270, -0.1550]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -1.2945e-07, ..., -6.0536e-08, + 1.2107e-08, 1.8626e-09], + [ 7.4506e-09, 6.5193e-09, -6.5193e-09, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00], + [ 5.9605e-08, 5.4948e-08, 2.3283e-08, ..., 5.5879e-09, + 8.3819e-09, 9.3132e-10], + ..., + [ 3.9116e-08, 2.8871e-08, 9.3132e-09, ..., 1.3970e-08, + -4.1910e-08, 0.0000e+00], + [ 4.6566e-09, 2.7940e-09, 6.5193e-08, ..., 2.7940e-08, + 3.1665e-08, 1.8626e-09], + [ 9.3132e-10, 1.8626e-09, 7.2643e-08, ..., -6.9849e-08, + -3.8184e-08, 0.0000e+00]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0049, 0.0129, 0.0115, -0.0018, 0.0244, -0.0025, -0.0099, -0.0065, + 0.0160, -0.0176], device='cuda:0'), grad: tensor([-1.0803e-07, -1.2200e-06, 2.6450e-07, -2.4680e-07, 2.2724e-07, + 1.2387e-07, -9.6858e-08, 7.7859e-07, 3.8464e-07, -9.5926e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 246.65, cls_loss 0.0007 cls_loss_mapping 0.0030 cls_loss_causal 0.5095 re_mapping 0.0037 re_causal 0.0124 /// teacc 99.10 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.1961, -0.0377, 0.1940, ..., 0.0434, -0.1768, -0.1388], + [-0.1279, -0.1735, -0.0410, ..., -0.1110, -0.1899, -0.0749], + [ 0.2250, -0.0650, -0.2111, ..., -0.1670, -0.1718, -0.1962], + ..., + [-0.0722, -0.0282, -0.2413, ..., -0.1064, 0.1046, 0.0246], + [-0.1705, -0.1093, -0.0272, ..., -0.1751, -0.0308, -0.0281], + [-0.0730, -0.0894, -0.1257, ..., 0.0914, 0.0268, -0.1551]], + device='cuda:0'), grad: tensor([[ 9.3132e-08, 6.5193e-09, -3.2596e-08, ..., 2.1420e-08, + 3.3993e-08, 2.7940e-09], + [ 1.5739e-07, 2.5146e-08, 2.7940e-09, ..., 4.5169e-08, + 3.6787e-08, 4.6566e-10], + [-1.2163e-06, -3.1665e-08, 2.3283e-09, ..., 6.5193e-09, + 2.0489e-08, 0.0000e+00], + ..., + [ 1.9465e-07, 2.0955e-08, 2.3283e-09, ..., 1.7835e-07, + 3.6322e-08, 9.3132e-10], + [ 6.2678e-07, 3.6787e-08, 7.4506e-09, ..., 5.4482e-08, + -3.2596e-09, 9.7789e-09], + [ 1.1176e-08, 3.7253e-09, 1.2107e-08, ..., -1.1986e-06, + -4.4098e-07, 4.6566e-10]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0051, 0.0127, 0.0116, -0.0017, 0.0262, -0.0026, -0.0097, -0.0066, + 0.0160, -0.0187], device='cuda:0'), grad: tensor([ 3.3854e-07, 1.9185e-07, -2.1700e-06, 4.5635e-08, 2.2650e-06, + -9.0804e-08, 4.0932e-07, 1.1735e-06, 5.9651e-07, -2.7753e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 246.84, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.5102 re_mapping 0.0038 re_causal 0.0129 /// teacc 99.01 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.1973, -0.0366, 0.1943, ..., 0.0434, -0.1772, -0.1390], + [-0.1289, -0.1757, -0.0411, ..., -0.1116, -0.1912, -0.0750], + [ 0.2262, -0.0651, -0.2117, ..., -0.1667, -0.1720, -0.1964], + ..., + [-0.0730, -0.0292, -0.2441, ..., -0.1073, 0.1047, 0.0249], + [-0.1717, -0.1097, -0.0275, ..., -0.1766, -0.0311, -0.0282], + [-0.0730, -0.0902, -0.1258, ..., 0.0913, 0.0269, -0.1551]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, -4.0326e-07, ..., -4.3772e-08, + 1.5832e-08, 1.0245e-08], + [ 9.2201e-08, 5.1223e-08, 1.9558e-08, ..., -8.3819e-09, + 1.2107e-08, 9.3132e-10], + [ 9.3132e-10, 2.7940e-09, 6.5193e-09, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + ..., + [ 1.7695e-08, 1.3039e-08, 7.4506e-09, ..., 2.7940e-09, + -2.6077e-08, 9.3132e-10], + [ 3.7253e-09, 6.5193e-09, 2.8871e-08, ..., 2.7940e-09, + 1.3039e-08, 3.7253e-09], + [ 7.4506e-09, 1.1176e-08, 2.9802e-08, ..., 6.5193e-09, + 3.7253e-08, 1.8626e-09]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0052, 0.0119, 0.0124, -0.0017, 0.0265, -0.0026, -0.0098, -0.0063, + 0.0158, -0.0187], device='cuda:0'), grad: tensor([-5.3830e-07, 2.6915e-07, 5.8673e-08, -6.5565e-07, 8.1025e-08, + -2.3004e-07, 3.7160e-07, 2.2259e-07, 1.3411e-07, 2.9244e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 246.39, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5026 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.17 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.1977, -0.0362, 0.1946, ..., 0.0432, -0.1775, -0.1391], + [-0.1296, -0.1789, -0.0414, ..., -0.1092, -0.1918, -0.0751], + [ 0.2268, -0.0651, -0.2128, ..., -0.1673, -0.1730, -0.1966], + ..., + [-0.0732, -0.0261, -0.2464, ..., -0.1050, 0.1079, 0.0248], + [-0.1722, -0.1096, -0.0275, ..., -0.1779, -0.0313, -0.0282], + [-0.0731, -0.0936, -0.1259, ..., 0.0911, 0.0243, -0.1553]], + device='cuda:0'), grad: tensor([[ 1.4901e-07, 7.5437e-08, -5.0291e-08, ..., -1.7695e-08, + 3.7253e-09, 1.8626e-09], + [ 3.5949e-07, 1.7602e-07, 7.4506e-09, ..., 6.5193e-09, + 1.1176e-08, 9.3132e-10], + [ 1.8440e-07, 1.5274e-07, 2.7940e-09, ..., 1.8626e-09, + 9.3132e-09, 0.0000e+00], + ..., + [ 5.5321e-07, 2.5425e-07, 2.7940e-09, ..., 1.3039e-08, + -1.4901e-08, 0.0000e+00], + [ 1.5087e-07, 1.0338e-07, 9.3132e-09, ..., 2.7008e-08, + 6.5193e-09, 2.7940e-09], + [ 7.8231e-08, 4.9360e-08, 2.3283e-08, ..., -1.4994e-07, + -2.2352e-08, 0.0000e+00]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0051, 0.0113, 0.0124, -0.0020, 0.0261, -0.0026, -0.0098, -0.0035, + 0.0157, -0.0200], device='cuda:0'), grad: tensor([ 6.8452e-07, 1.3895e-06, 1.3281e-06, -9.9316e-06, 2.6356e-07, + 2.2538e-06, 1.3225e-07, 2.7530e-06, 9.8068e-07, 1.3504e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 246.54, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.4874 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.00 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.1990, -0.0363, 0.1948, ..., 0.0427, -0.1778, -0.1394], + [-0.1298, -0.1819, -0.0405, ..., -0.1070, -0.1924, -0.0754], + [ 0.2285, -0.0654, -0.2133, ..., -0.1677, -0.1739, -0.1964], + ..., + [-0.0755, -0.0257, -0.2490, ..., -0.1052, 0.1080, 0.0244], + [-0.1731, -0.1115, -0.0279, ..., -0.1843, -0.0323, -0.0289], + [-0.0731, -0.0936, -0.1262, ..., 0.0931, 0.0263, -0.1555]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-08, -1.0990e-07, ..., -7.8231e-08, + 1.8626e-08, 7.8231e-08], + [ 4.6566e-09, 9.3132e-09, 1.4901e-08, ..., 2.1420e-08, + 1.0338e-07, 1.3039e-06], + [-2.7940e-09, 4.1910e-08, 8.3819e-09, ..., 1.0245e-08, + 7.5437e-08, 1.0412e-06], + ..., + [ 1.1176e-08, 3.3528e-08, 2.8871e-08, ..., 2.3004e-07, + 2.7940e-08, 3.2596e-08], + [ 0.0000e+00, 3.9116e-08, 1.6764e-08, ..., 3.1665e-08, + -3.3528e-08, 9.3132e-10], + [ 0.0000e+00, 1.5832e-08, 3.8184e-08, ..., -1.2107e-08, + -5.5879e-08, 4.5635e-08]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0043, 0.0126, 0.0132, -0.0028, 0.0235, -0.0025, -0.0096, -0.0036, + 0.0139, -0.0178], device='cuda:0'), grad: tensor([ 3.0454e-07, 6.6683e-06, 5.5879e-06, 4.7963e-07, 1.1306e-06, + 4.5300e-06, -1.9461e-05, 8.4005e-07, -5.1782e-07, 4.1910e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 246.77, cls_loss 0.0008 cls_loss_mapping 0.0028 cls_loss_causal 0.5076 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.06 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.2002, -0.0367, 0.1949, ..., 0.0427, -0.1780, -0.1397], + [-0.1309, -0.1861, -0.0406, ..., -0.1070, -0.1928, -0.0759], + [ 0.2322, -0.0646, -0.2140, ..., -0.1681, -0.1742, -0.1963], + ..., + [-0.0764, -0.0259, -0.2501, ..., -0.1058, 0.1078, 0.0243], + [-0.1790, -0.1158, -0.0281, ..., -0.1847, -0.0325, -0.0291], + [-0.0732, -0.0936, -0.1263, ..., 0.0931, 0.0264, -0.1558]], + device='cuda:0'), grad: tensor([[ 1.1269e-07, 4.6566e-09, -7.5437e-08, ..., -6.7055e-08, + 7.4506e-09, 6.3330e-08], + [ 1.1981e-05, 2.7940e-09, 6.0536e-08, ..., 2.0489e-08, + 3.7253e-09, 1.0245e-08], + [-1.5497e-05, 6.5193e-09, 2.6077e-08, ..., 1.1176e-08, + 1.5832e-08, 1.1176e-08], + ..., + [ 2.1514e-07, 7.4506e-09, 4.6566e-09, ..., 1.8626e-09, + 2.7940e-09, 5.5879e-09], + [ 8.8569e-07, -2.4959e-07, 1.0990e-07, ..., 1.8626e-08, + -8.1025e-07, -5.0478e-07], + [ 9.0338e-08, 1.9558e-08, 2.5146e-08, ..., 1.2107e-08, + 1.1176e-08, 1.2107e-08]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0043, 0.0128, 0.0163, -0.0039, 0.0238, -0.0025, -0.0096, -0.0042, + 0.0120, -0.0178], device='cuda:0'), grad: tensor([ 3.9022e-07, 3.9846e-05, -5.1260e-05, 1.3663e-06, 4.1351e-06, + 3.2503e-06, 1.0729e-06, 7.5903e-07, -1.1921e-07, 5.5041e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 246.64, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.5255 re_mapping 0.0036 re_causal 0.0124 /// teacc 99.11 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.2005, -0.0369, 0.1949, ..., 0.0427, -0.1785, -0.1402], + [-0.1317, -0.1911, -0.0409, ..., -0.1065, -0.1932, -0.0765], + [ 0.2320, -0.0653, -0.2149, ..., -0.1686, -0.1749, -0.1977], + ..., + [-0.0764, -0.0260, -0.2514, ..., -0.1069, 0.1077, 0.0270], + [-0.1799, -0.1161, -0.0283, ..., -0.1849, -0.0326, -0.0294], + [-0.0732, -0.0937, -0.1263, ..., 0.0929, 0.0264, -0.1559]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, 1.6764e-08, 1.6764e-08, ..., -4.8429e-08, + 9.3132e-09, 2.7940e-08], + [ 1.0524e-07, 7.1712e-08, 8.3819e-09, ..., -6.3330e-08, + 1.6764e-08, 1.8626e-09], + [ 2.6450e-07, 1.3132e-07, 2.1048e-07, ..., 1.5832e-08, + 1.4901e-08, 4.1910e-08], + ..., + [ 6.8918e-08, 2.3283e-08, 4.6566e-09, ..., 1.2200e-07, + -6.5193e-09, 0.0000e+00], + [ 6.8918e-08, 5.1223e-08, 5.6718e-07, ..., 2.5146e-08, + 4.2841e-08, 1.2759e-07], + [ 5.5879e-09, 2.7940e-09, 6.0536e-08, ..., -2.0489e-07, + -5.7742e-08, 3.7253e-09]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0042, 0.0131, 0.0156, -0.0026, 0.0240, -0.0026, -0.0095, -0.0047, + 0.0117, -0.0178], device='cuda:0'), grad: tensor([ 3.1199e-07, -1.2293e-07, 1.9129e-06, -2.6673e-06, 3.7067e-07, + 2.0470e-06, -4.2059e-06, 6.1188e-07, 1.9893e-06, -2.3935e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 246.40, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4789 re_mapping 0.0035 re_causal 0.0114 /// teacc 99.12 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.2015, -0.0371, 0.1951, ..., 0.0427, -0.1786, -0.1404], + [-0.1343, -0.1943, -0.0413, ..., -0.1066, -0.1944, -0.0766], + [ 0.2337, -0.0654, -0.2153, ..., -0.1692, -0.1751, -0.1981], + ..., + [-0.0773, -0.0260, -0.2536, ..., -0.1086, 0.1065, 0.0270], + [-0.1802, -0.1166, -0.0286, ..., -0.1852, -0.0327, -0.0294], + [-0.0733, -0.0937, -0.1264, ..., 0.0932, 0.0275, -0.1560]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.6077e-08, -6.4448e-07, ..., -3.2317e-07, + 1.3039e-08, 1.8626e-09], + [ 1.6764e-08, 9.3132e-09, 3.7253e-09, ..., 9.3132e-10, + 2.1420e-08, 0.0000e+00], + [-1.9558e-08, 3.1665e-08, 2.9895e-07, ..., 1.8533e-07, + 3.7253e-09, 0.0000e+00], + ..., + [ 1.2107e-08, 2.7008e-08, 8.3819e-09, ..., 9.3132e-08, + -1.4901e-08, 0.0000e+00], + [ 6.5193e-09, 1.9092e-07, 1.7509e-07, ..., 1.6578e-07, + 2.0489e-08, 9.3132e-10], + [ 5.7742e-07, 1.4268e-06, 7.2643e-08, ..., -3.4925e-07, + -1.2480e-07, 9.3132e-10]], device='cuda:0') +Epoch 237, bias, value: tensor([ 0.0042, 0.0125, 0.0159, -0.0026, 0.0240, -0.0026, -0.0094, -0.0053, + 0.0116, -0.0172], device='cuda:0'), grad: tensor([-1.1688e-06, 6.7987e-08, 6.5751e-07, -5.1111e-06, 1.7229e-07, + -1.4538e-06, 2.0023e-07, 1.9278e-07, 2.4214e-06, 3.9861e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 246.57, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4685 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.2019, -0.0378, 0.1951, ..., 0.0427, -0.1790, -0.1406], + [-0.1348, -0.1963, -0.0415, ..., -0.1041, -0.1950, -0.0769], + [ 0.2340, -0.0655, -0.2175, ..., -0.1695, -0.1756, -0.1989], + ..., + [-0.0775, -0.0261, -0.2547, ..., -0.1100, 0.1065, 0.0281], + [-0.1802, -0.1161, -0.0288, ..., -0.1854, -0.0327, -0.0292], + [-0.0734, -0.0937, -0.1264, ..., 0.0931, 0.0276, -0.1562]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 1.7695e-08, ..., -5.5879e-09, + 1.4901e-08, 1.9558e-08], + [ 1.8626e-09, 4.6566e-09, 6.5193e-09, ..., 1.6764e-08, + 2.4214e-08, 3.7253e-09], + [ 1.8626e-09, 3.7253e-09, 1.0245e-08, ..., 1.3039e-08, + 6.5193e-09, 4.6566e-09], + ..., + [ 1.8626e-09, 3.7253e-09, 9.3132e-10, ..., 5.5879e-09, + -6.8918e-08, 9.3132e-10], + [ 1.8626e-09, -1.8626e-09, 2.0489e-08, ..., 4.6566e-09, + 8.3819e-09, 4.6566e-09], + [ 0.0000e+00, 2.7940e-09, 7.4506e-09, ..., 1.4063e-07, + 9.3132e-09, 1.8626e-09]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0041, 0.0152, 0.0157, -0.0026, 0.0239, -0.0029, -0.0093, -0.0071, + 0.0122, -0.0175], device='cuda:0'), grad: tensor([ 3.2037e-07, 6.7335e-07, 4.9639e-07, 1.2331e-06, -1.0803e-07, + -2.1886e-07, 1.8626e-09, -7.7114e-07, -2.1011e-06, 4.7497e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 246.57, cls_loss 0.0009 cls_loss_mapping 0.0035 cls_loss_causal 0.5220 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.07 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.2028, -0.0391, 0.1952, ..., 0.0427, -0.1794, -0.1409], + [-0.1353, -0.2009, -0.0416, ..., -0.1034, -0.1958, -0.0772], + [ 0.2344, -0.0658, -0.2199, ..., -0.1702, -0.1762, -0.2017], + ..., + [-0.0779, -0.0259, -0.2558, ..., -0.1105, 0.1066, 0.0284], + [-0.1805, -0.1176, -0.0306, ..., -0.1859, -0.0329, -0.0307], + [-0.0735, -0.0933, -0.1265, ..., 0.0931, 0.0276, -0.1571]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3039e-08, 6.5193e-09, ..., 2.0489e-08, + 1.3970e-08, 9.3132e-09], + [ 9.3132e-10, 1.2107e-08, -2.4214e-08, ..., 6.5193e-08, + 1.6764e-08, 6.5193e-09], + [ 0.0000e+00, 1.0245e-08, 9.3132e-10, ..., 6.5193e-09, + 6.5193e-09, 5.5879e-09], + ..., + [ 2.7940e-09, 2.5146e-08, 7.4506e-09, ..., 1.3039e-07, + 3.9116e-08, 1.3970e-08], + [ 0.0000e+00, 8.6799e-07, 1.8626e-09, ..., 7.2643e-08, + 5.1688e-07, 5.1502e-07], + [ 0.0000e+00, 5.0291e-08, 2.7940e-09, ..., -2.5239e-07, + -1.2852e-07, 2.5146e-08]], device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0041, 0.0157, 0.0154, -0.0029, 0.0239, -0.0029, -0.0089, -0.0073, + 0.0115, -0.0177], device='cuda:0'), grad: tensor([ 3.0547e-07, 2.2072e-07, 1.7323e-07, 7.6517e-06, 3.3341e-07, + -1.8403e-05, 1.5870e-06, 8.5961e-07, 7.9200e-06, -6.3796e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0029 cls_loss_causal 0.4674 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.2030, -0.0393, 0.1953, ..., 0.0427, -0.1798, -0.1411], + [-0.1357, -0.2031, -0.0417, ..., -0.1034, -0.1966, -0.0774], + [ 0.2347, -0.0658, -0.2207, ..., -0.1708, -0.1765, -0.2019], + ..., + [-0.0780, -0.0260, -0.2564, ..., -0.1106, 0.1066, 0.0285], + [-0.1806, -0.1180, -0.0309, ..., -0.1862, -0.0328, -0.0309], + [-0.0736, -0.0933, -0.1266, ..., 0.0941, 0.0282, -0.1575]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 9.3132e-10, -1.5926e-07, ..., -1.4249e-07, + 1.7695e-08, 7.6368e-08], + [ 1.3504e-07, 1.8626e-09, 1.0245e-08, ..., 1.9558e-08, + 3.7253e-09, 2.6077e-08], + [-1.8522e-05, 0.0000e+00, 3.1665e-08, ..., 2.5146e-08, + 2.7940e-09, 9.3132e-09], + ..., + [ 9.6858e-08, 3.7253e-09, 1.0245e-08, ..., 1.4435e-07, + 1.8626e-09, 1.8626e-09], + [ 1.8239e-05, 1.4901e-08, 3.8184e-08, ..., 2.8871e-08, + -1.8626e-09, -1.5460e-07], + [ 2.7940e-09, 1.0151e-07, 1.7788e-07, ..., -4.4890e-07, + 5.0291e-08, 9.3132e-10]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0040, 0.0156, 0.0153, -0.0024, 0.0227, -0.0030, -0.0088, -0.0073, + 0.0116, -0.0168], device='cuda:0'), grad: tensor([-3.4459e-08, 4.9360e-07, -3.2693e-05, 7.0687e-07, 6.6590e-07, + -1.4100e-06, 1.7323e-07, 4.5449e-07, 3.1143e-05, 5.5041e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 248.58, cls_loss 0.0006 cls_loss_mapping 0.0039 cls_loss_causal 0.4939 re_mapping 0.0038 re_causal 0.0125 /// teacc 99.13 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.2037, -0.0394, 0.1952, ..., 0.0427, -0.1801, -0.1413], + [-0.1362, -0.2046, -0.0417, ..., -0.1035, -0.1972, -0.0778], + [ 0.2352, -0.0658, -0.2221, ..., -0.1715, -0.1767, -0.2021], + ..., + [-0.0782, -0.0260, -0.2578, ..., -0.1108, 0.1066, 0.0285], + [-0.1804, -0.1183, -0.0309, ..., -0.1859, -0.0329, -0.0313], + [-0.0739, -0.0934, -0.1267, ..., 0.0939, 0.0280, -0.1576]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, -1.8161e-07, ..., -2.7008e-08, + 1.8626e-09, 1.8626e-09], + [ 4.6566e-09, 5.5879e-09, 1.1176e-08, ..., 3.7253e-08, + 1.1176e-08, 0.0000e+00], + [ 2.7940e-09, 2.7940e-09, 1.8626e-09, ..., 1.3039e-08, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.4901e-08, 1.3039e-08, 2.7940e-09, ..., 1.0338e-07, + 1.4901e-08, 0.0000e+00], + [ 5.5879e-09, 5.5879e-09, 1.8626e-09, ..., 1.3504e-07, + 3.7253e-09, 0.0000e+00], + [ 1.5832e-08, 3.2596e-08, 1.5832e-08, ..., -8.3353e-07, + -1.3970e-07, 0.0000e+00]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0040, 0.0156, 0.0153, -0.0026, 0.0229, -0.0031, -0.0085, -0.0073, + 0.0117, -0.0170], device='cuda:0'), grad: tensor([-1.6484e-07, 1.0245e-07, 5.2154e-08, -9.2108e-07, -1.1884e-06, + 3.2112e-06, 3.7625e-07, 2.8592e-07, 2.9802e-07, -2.0489e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 248.44, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.5154 re_mapping 0.0037 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.2054, -0.0397, 0.1953, ..., 0.0427, -0.1804, -0.1415], + [-0.1366, -0.2064, -0.0418, ..., -0.1036, -0.1982, -0.0810], + [ 0.2364, -0.0658, -0.2225, ..., -0.1718, -0.1770, -0.2024], + ..., + [-0.0802, -0.0261, -0.2589, ..., -0.1109, 0.1067, 0.0285], + [-0.1806, -0.1189, -0.0310, ..., -0.1861, -0.0329, -0.0317], + [-0.0741, -0.0935, -0.1267, ..., 0.0939, 0.0280, -0.1577]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 9.3132e-09, -2.9802e-07, ..., -1.7043e-07, + 5.5879e-09, 6.5193e-09], + [ 3.4273e-07, 6.6124e-08, 2.4214e-08, ..., 5.5879e-09, + 1.6764e-08, 2.7940e-09], + [-4.0699e-07, -5.4948e-08, 1.5832e-08, ..., 6.5193e-09, + 1.2107e-08, 9.3132e-10], + ..., + [ 4.5635e-08, 1.7695e-08, 5.5879e-09, ..., 9.3132e-10, + -4.6566e-08, 1.8626e-09], + [ 1.5832e-08, -1.6764e-08, -4.3772e-08, ..., -1.3039e-08, + 3.7253e-09, 9.3132e-09], + [ 6.5193e-09, 1.6764e-08, 2.7567e-07, ..., 1.4808e-07, + 9.3132e-10, 6.5193e-09]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0039, 0.0148, 0.0161, -0.0020, 0.0231, -0.0035, -0.0073, -0.0076, + 0.0116, -0.0171], device='cuda:0'), grad: tensor([-2.9802e-07, -3.5651e-06, 8.1956e-07, -6.3609e-07, 1.3132e-07, + -1.8254e-07, 7.5344e-07, 2.2314e-06, -6.0536e-08, 8.2143e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 248.46, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4886 re_mapping 0.0036 re_causal 0.0114 /// teacc 99.17 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.2055, -0.0398, 0.1956, ..., 0.0428, -0.1807, -0.1418], + [-0.1370, -0.2076, -0.0418, ..., -0.1036, -0.1992, -0.0811], + [ 0.2369, -0.0659, -0.2233, ..., -0.1721, -0.1772, -0.2028], + ..., + [-0.0805, -0.0261, -0.2600, ..., -0.1110, 0.1068, 0.0285], + [-0.1807, -0.1192, -0.0313, ..., -0.1864, -0.0330, -0.0322], + [-0.0743, -0.0935, -0.1268, ..., 0.0939, 0.0281, -0.1580]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 0.0000e+00, -1.6298e-07, ..., -7.5437e-08, + 9.3132e-10, 1.8626e-09], + [ 1.3877e-07, 1.8626e-09, 2.5146e-08, ..., 4.6473e-07, + 2.7940e-09, 0.0000e+00], + [-2.5984e-07, 1.8626e-09, 7.4506e-09, ..., 4.3772e-08, + 9.3132e-10, 9.3132e-10], + ..., + [ 5.5879e-08, 9.3132e-09, 6.5193e-09, ..., 8.3819e-09, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-08, -0.0000e+00, 8.3819e-09, ..., 2.0489e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 5.4948e-08, ..., 1.1642e-07, + -1.3970e-08, 0.0000e+00]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0040, 0.0147, 0.0161, 0.0009, 0.0230, -0.0052, -0.0071, -0.0076, + 0.0115, -0.0171], device='cuda:0'), grad: tensor([-1.3132e-07, 1.4119e-06, 2.0675e-07, 1.8440e-07, -9.4157e-07, + 1.9278e-07, 2.2817e-07, -1.1632e-06, -7.6927e-07, 7.7952e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 248.54, cls_loss 0.0006 cls_loss_mapping 0.0024 cls_loss_causal 0.4717 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.03 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.2059, -0.0399, 0.1957, ..., 0.0428, -0.1810, -0.1424], + [-0.1377, -0.2087, -0.0418, ..., -0.1039, -0.2007, -0.0811], + [ 0.2375, -0.0659, -0.2246, ..., -0.1728, -0.1777, -0.2031], + ..., + [-0.0809, -0.0262, -0.2618, ..., -0.1113, 0.1068, 0.0284], + [-0.1808, -0.1192, -0.0314, ..., -0.1866, -0.0331, -0.0323], + [-0.0744, -0.0935, -0.1269, ..., 0.0939, 0.0282, -0.1581]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.5146e-08, 1.8533e-07, ..., 0.0000e+00, + 5.5879e-09, 3.2596e-08], + [ 1.8626e-09, 1.7695e-08, 1.2200e-07, ..., 7.4506e-09, + 8.3819e-09, 3.3975e-06], + [-9.3132e-10, 1.5832e-08, 1.1642e-07, ..., 1.8626e-09, + 6.5193e-09, 3.0380e-06], + ..., + [ 9.3132e-10, 9.3132e-10, 3.7253e-09, ..., 2.7940e-09, + -1.1176e-08, 1.5832e-08], + [ 3.7253e-09, 3.7253e-09, 2.3283e-08, ..., 9.3132e-10, + 1.8626e-09, 1.2759e-07], + [ 9.3132e-10, 1.8626e-09, 1.0245e-08, ..., 6.3330e-08, + 1.8626e-09, 6.5193e-09]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0040, 0.0146, 0.0162, 0.0007, 0.0231, -0.0052, -0.0070, -0.0077, + 0.0117, -0.0170], device='cuda:0'), grad: tensor([ 1.5963e-06, 1.8939e-05, 1.6943e-05, 4.0699e-07, 3.4347e-06, + 3.6582e-06, -4.6164e-05, 5.5879e-08, 8.5961e-07, 2.9150e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 248.46, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.5240 re_mapping 0.0036 re_causal 0.0123 /// teacc 99.11 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.2061, -0.0401, 0.1958, ..., 0.0425, -0.1812, -0.1424], + [-0.1383, -0.2094, -0.0416, ..., -0.1045, -0.2017, -0.0813], + [ 0.2380, -0.0659, -0.2254, ..., -0.1731, -0.1780, -0.2039], + ..., + [-0.0809, -0.0262, -0.2626, ..., -0.1110, 0.1069, 0.0285], + [-0.1809, -0.1193, -0.0315, ..., -0.1869, -0.0331, -0.0323], + [-0.0745, -0.0935, -0.1270, ..., 0.0941, 0.0281, -0.1580]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, -9.8720e-08, ..., -5.1223e-08, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 2.7940e-09, 8.3819e-09, ..., 4.6566e-09, + 1.3970e-08, 9.3132e-10], + [-7.4506e-09, 9.3132e-10, 7.4506e-09, ..., 5.5879e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 6.5193e-09, 3.0734e-08, 1.1176e-08, ..., 1.0245e-08, + -2.3283e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 6.5193e-09, ..., 4.6566e-09, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 2.1420e-08, 3.9116e-08, ..., 6.1467e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0037, 0.0137, 0.0160, 0.0017, 0.0233, -0.0059, -0.0068, -0.0069, + 0.0118, -0.0169], device='cuda:0'), grad: tensor([-1.6484e-07, 1.1176e-07, 1.4901e-08, -3.5018e-07, -6.7987e-08, + 8.4750e-08, 1.3970e-08, 6.9849e-08, 2.7008e-08, 2.5798e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 248.12, cls_loss 0.0005 cls_loss_mapping 0.0019 cls_loss_causal 0.4723 re_mapping 0.0039 re_causal 0.0125 /// teacc 99.10 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.2048, -0.0403, 0.1973, ..., 0.0434, -0.1818, -0.1444], + [-0.1386, -0.2098, -0.0416, ..., -0.1045, -0.2021, -0.0813], + [ 0.2385, -0.0659, -0.2269, ..., -0.1739, -0.1781, -0.2042], + ..., + [-0.0813, -0.0262, -0.2633, ..., -0.1110, 0.1070, 0.0285], + [-0.1811, -0.1193, -0.0309, ..., -0.1866, -0.0332, -0.0325], + [-0.0751, -0.0936, -0.1290, ..., 0.0934, 0.0281, -0.1587]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 5.5879e-09, -2.1420e-08, ..., 9.3132e-10, + 3.7253e-09, 9.3132e-10], + [ 1.8161e-07, 3.1665e-08, -9.3132e-10, ..., 1.6764e-08, + 2.2352e-08, 9.3132e-10], + [-3.0547e-07, 2.5146e-08, 2.7940e-09, ..., 8.3819e-09, + 4.6566e-09, 1.8626e-09], + ..., + [ 5.5879e-08, 8.4750e-08, 9.3132e-10, ..., -3.7253e-09, + -6.3330e-08, 9.3132e-10], + [ 2.2352e-08, 1.9558e-08, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 2.7940e-09, 3.7253e-09, 7.4506e-09, ..., 9.3132e-10, + 2.1420e-08, 1.8626e-09]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0046, 0.0137, 0.0162, 0.0016, 0.0233, -0.0059, -0.0064, -0.0070, + 0.0121, -0.0177], device='cuda:0'), grad: tensor([ 7.7300e-08, 7.1898e-07, -5.8953e-07, -1.1735e-06, -1.6764e-08, + 3.7905e-07, 2.5332e-07, -1.4901e-07, 1.7416e-07, 3.3900e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 248.64, cls_loss 0.0006 cls_loss_mapping 0.0025 cls_loss_causal 0.5059 re_mapping 0.0038 re_causal 0.0128 /// teacc 99.08 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.2046, -0.0407, 0.1982, ..., 0.0437, -0.1824, -0.1449], + [-0.1388, -0.2106, -0.0417, ..., -0.1047, -0.2035, -0.0814], + [ 0.2387, -0.0659, -0.2285, ..., -0.1747, -0.1785, -0.2045], + ..., + [-0.0815, -0.0263, -0.2646, ..., -0.1110, 0.1071, 0.0285], + [-0.1812, -0.1197, -0.0305, ..., -0.1868, -0.0333, -0.0328], + [-0.0753, -0.0936, -0.1300, ..., 0.0932, 0.0281, -0.1590]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 0.0000e+00, -1.2917e-06, ..., -5.0198e-07, + 1.8626e-09, -9.3132e-10], + [ 1.8626e-09, 9.3132e-10, 6.4261e-08, ..., 5.9605e-08, + 5.6811e-08, 1.2107e-08], + [-5.1223e-08, 0.0000e+00, 1.0990e-07, ..., 8.5682e-08, + 8.3819e-09, 1.3039e-08], + ..., + [ 7.4506e-09, 9.3132e-10, 1.1735e-07, ..., 9.1270e-08, + -7.7300e-08, 4.6566e-09], + [ 1.8626e-08, 9.3132e-10, 1.3411e-07, ..., 1.1828e-07, + 4.4703e-08, 6.0536e-08], + [ 0.0000e+00, 9.3132e-10, 3.0175e-07, ..., 1.0058e-07, + 1.8626e-08, 1.4901e-08]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0050, 0.0136, 0.0160, 0.0013, 0.0233, -0.0057, -0.0063, -0.0069, + 0.0124, -0.0181], device='cuda:0'), grad: tensor([-2.4028e-06, 1.5153e-06, 3.0547e-07, 3.3807e-07, -9.4343e-07, + -7.0687e-07, 1.8161e-06, -1.5236e-06, 5.1595e-07, 1.0785e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 248.41, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4806 re_mapping 0.0036 re_causal 0.0115 /// teacc 99.14 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.2046, -0.0409, 0.1985, ..., 0.0438, -0.1826, -0.1450], + [-0.1390, -0.2114, -0.0420, ..., -0.1049, -0.2042, -0.0816], + [ 0.2405, -0.0659, -0.2295, ..., -0.1756, -0.1789, -0.2047], + ..., + [-0.0818, -0.0263, -0.2677, ..., -0.1111, 0.1071, 0.0284], + [-0.1838, -0.1199, -0.0304, ..., -0.1873, -0.0337, -0.0334], + [-0.0754, -0.0936, -0.1301, ..., 0.0931, 0.0282, -0.1593]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, -8.3521e-06, ..., -5.0776e-06, + -9.3132e-10, 1.8626e-09], + [ 1.8626e-09, 9.3132e-10, 2.8964e-07, ..., 1.9558e-07, + 2.6077e-08, 9.3132e-10], + [-1.3970e-08, 9.3132e-10, 1.5646e-07, ..., 9.8720e-08, + 1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 9.3132e-10, 3.1106e-07, ..., 2.1234e-07, + -4.8429e-08, 0.0000e+00], + [ 8.3819e-09, 3.7253e-09, 1.9465e-07, ..., 1.0896e-07, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.3416e-06, ..., 2.2296e-06, + 8.3819e-09, 9.3132e-10]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0051, 0.0131, 0.0170, 0.0010, 0.0235, -0.0054, -0.0064, -0.0065, + 0.0114, -0.0182], device='cuda:0'), grad: tensor([-2.3499e-05, 1.0105e-06, 4.5449e-07, 1.3355e-06, 1.0710e-07, + 4.9695e-06, 4.3400e-06, 6.7614e-07, 5.8021e-07, 1.0036e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 248.47, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4942 re_mapping 0.0036 re_causal 0.0115 /// teacc 99.12 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.2047, -0.0411, 0.1987, ..., 0.0438, -0.1834, -0.1454], + [-0.1394, -0.2120, -0.0424, ..., -0.1051, -0.2049, -0.0816], + [ 0.2408, -0.0660, -0.2311, ..., -0.1762, -0.1794, -0.2051], + ..., + [-0.0817, -0.0263, -0.2717, ..., -0.1113, 0.1072, 0.0284], + [-0.1839, -0.1204, -0.0308, ..., -0.1882, -0.0340, -0.0338], + [-0.0755, -0.0934, -0.1302, ..., 0.0933, 0.0282, -0.1597]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 9.4995e-08, ..., -4.0978e-08, + 5.6811e-08, 9.0338e-08], + [ 1.8626e-09, 5.5879e-09, 1.6764e-08, ..., 1.7695e-08, + 7.8231e-08, 8.3819e-09], + [-2.7940e-09, 4.6566e-09, 1.0245e-08, ..., 6.5193e-09, + 1.8626e-08, 2.7940e-09], + ..., + [ 3.7253e-09, 1.4901e-08, 9.3132e-09, ..., 0.0000e+00, + -3.5856e-07, 0.0000e+00], + [ 2.7940e-09, 1.1176e-08, 4.5635e-08, ..., 2.1420e-08, + 2.9802e-08, 2.4214e-08], + [ 9.3132e-10, 2.7940e-09, 3.6322e-08, ..., -0.0000e+00, + 2.3749e-07, 9.3132e-10]], device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0050, 0.0132, 0.0168, 0.0005, 0.0238, -0.0055, -0.0060, -0.0067, + 0.0113, -0.0182], device='cuda:0'), grad: tensor([ 3.0734e-07, -1.2837e-05, 1.8906e-07, -2.6263e-07, 2.7008e-08, + 3.8370e-07, -6.6310e-07, 1.1638e-05, 2.6356e-07, 9.7137e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 247.42, cls_loss 0.0007 cls_loss_mapping 0.0027 cls_loss_causal 0.4820 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.13 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.2051, -0.0415, 0.1979, ..., 0.0438, -0.1857, -0.1486], + [-0.1402, -0.2131, -0.0436, ..., -0.1053, -0.2060, -0.0817], + [ 0.2412, -0.0661, -0.2313, ..., -0.1765, -0.1805, -0.2054], + ..., + [-0.0818, -0.0263, -0.2733, ..., -0.1114, 0.1074, 0.0283], + [-0.1839, -0.1202, -0.0312, ..., -0.1899, -0.0352, -0.0350], + [-0.0755, -0.0934, -0.1303, ..., 0.0934, 0.0284, -0.1601]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 9.3132e-10, -3.1386e-07, ..., -1.0245e-08, + 9.3132e-09, -3.4459e-08], + [ 1.2621e-05, 6.5193e-09, 4.6566e-09, ..., 3.9116e-08, + 1.7695e-07, 9.3132e-10], + [ 3.5390e-08, -1.3039e-08, 3.7253e-09, ..., 5.5879e-09, + 5.5879e-09, 9.3132e-10], + ..., + [-1.2726e-05, 8.3819e-09, 1.8626e-09, ..., 2.7940e-08, + -2.0768e-07, 0.0000e+00], + [ 3.7253e-09, 1.2107e-08, 8.3819e-09, ..., 1.7043e-07, + 4.0978e-08, 1.8626e-09], + [ 4.6566e-09, -3.0734e-08, 1.5832e-08, ..., -8.9314e-07, + -1.5087e-07, 1.8626e-09]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0045, 0.0130, 0.0169, 0.0005, 0.0237, -0.0057, -0.0049, -0.0067, + 0.0103, -0.0179], device='cuda:0'), grad: tensor([-2.7288e-07, 1.0926e-04, 1.5842e-06, 7.6462e-07, 1.5842e-06, + 2.0023e-07, 3.1479e-07, -1.1122e-04, 4.7777e-07, -2.5723e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 246.71, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4779 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.13 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.2052, -0.0417, 0.1980, ..., 0.0437, -0.1860, -0.1486], + [-0.1411, -0.2135, -0.0443, ..., -0.1056, -0.2080, -0.0818], + [ 0.2415, -0.0661, -0.2317, ..., -0.1769, -0.1811, -0.2055], + ..., + [-0.0816, -0.0263, -0.2742, ..., -0.1117, 0.1075, 0.0280], + [-0.1840, -0.1201, -0.0313, ..., -0.1903, -0.0351, -0.0351], + [-0.0752, -0.0935, -0.1303, ..., 0.0936, 0.0285, -0.1604]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.6764e-08, -6.5193e-09, ..., 1.3039e-08, + 5.5879e-09, 9.3132e-10], + [-2.5146e-08, 4.6566e-09, 0.0000e+00, ..., -4.4703e-08, + 4.9360e-08, 0.0000e+00], + [ 2.7008e-08, 2.7008e-08, 9.3132e-10, ..., 2.7940e-08, + 1.3970e-08, 9.3132e-10], + ..., + [ 2.1420e-08, 1.4901e-08, 0.0000e+00, ..., 1.4808e-07, + -1.3411e-07, 0.0000e+00], + [ 1.2107e-08, 1.5832e-08, 9.3132e-10, ..., 5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 1.6764e-08, 1.8626e-08, 2.7940e-09, ..., 2.4214e-08, + 5.0291e-08, 9.3132e-10]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0044, 0.0125, 0.0168, 0.0003, 0.0236, -0.0053, -0.0052, -0.0065, + 0.0104, -0.0177], device='cuda:0'), grad: tensor([ 2.8871e-07, -3.2596e-07, 6.2957e-07, 2.8498e-06, -4.7870e-07, + -4.3698e-06, 4.8708e-07, 6.7987e-08, 2.2817e-07, 6.3609e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 246.83, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4668 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.17 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.2055, -0.0420, 0.1980, ..., 0.0437, -0.1864, -0.1487], + [-0.1435, -0.2144, -0.0440, ..., -0.1054, -0.2086, -0.0818], + [ 0.2427, -0.0661, -0.2320, ..., -0.1774, -0.1815, -0.2057], + ..., + [-0.0819, -0.0263, -0.2752, ..., -0.1118, 0.1076, 0.0280], + [-0.1841, -0.1200, -0.0308, ..., -0.1910, -0.0350, -0.0352], + [-0.0752, -0.0935, -0.1303, ..., 0.0935, 0.0285, -0.1605]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -5.3551e-08, ..., -8.3819e-09, + -4.6566e-10, 3.7253e-09], + [ 7.4971e-08, 4.6566e-10, 2.3283e-09, ..., 1.1688e-07, + 4.1444e-08, 9.3132e-10], + [-1.9511e-07, 1.8626e-09, 1.8626e-09, ..., 6.4261e-08, + 4.4703e-08, 0.0000e+00], + ..., + [ 9.1270e-08, 4.6566e-10, 1.8626e-09, ..., 6.1933e-08, + -1.0012e-07, 0.0000e+00], + [ 8.8476e-09, -6.9849e-09, 2.0489e-08, ..., 3.6787e-08, + 1.0710e-08, 1.8626e-09], + [ 1.3970e-09, 4.6566e-10, 1.7229e-08, ..., 5.6028e-06, + 1.6252e-07, 0.0000e+00]], device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0044, 0.0126, 0.0174, 0.0001, 0.0238, -0.0054, -0.0053, -0.0065, + 0.0107, -0.0179], device='cuda:0'), grad: tensor([ 6.1933e-08, 8.6799e-07, 3.2829e-07, 1.9791e-07, -1.5363e-05, + 9.5926e-08, 2.8778e-07, -1.0263e-06, -1.6531e-07, 1.4678e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 245.88, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4886 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.15 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.2060, -0.0427, 0.1980, ..., 0.0438, -0.1881, -0.1497], + [-0.1443, -0.2166, -0.0440, ..., -0.1054, -0.2093, -0.0818], + [ 0.2434, -0.0667, -0.2327, ..., -0.1780, -0.1826, -0.2062], + ..., + [-0.0823, -0.0267, -0.2766, ..., -0.1120, 0.1077, 0.0281], + [-0.1843, -0.1207, -0.0311, ..., -0.1916, -0.0352, -0.0355], + [-0.0750, -0.0935, -0.1307, ..., 0.0935, 0.0285, -0.1607]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 2.7940e-09, -1.9465e-07, ..., -1.7788e-07, + 3.5390e-08, 4.0978e-08], + [ 7.5903e-07, 1.8626e-09, 2.1420e-08, ..., 7.1712e-08, + 1.7695e-08, 2.7940e-09], + [-8.6613e-07, 9.3132e-10, 9.9652e-08, ..., 7.7300e-08, + 5.5879e-09, 1.8626e-09], + ..., + [ 5.0291e-08, 1.5832e-08, 1.7695e-08, ..., 1.3039e-07, + -2.3283e-08, 0.0000e+00], + [ 6.5193e-09, 9.3132e-10, 4.6566e-09, ..., 5.9605e-08, + 6.5193e-09, 3.7253e-09], + [ 1.8626e-09, -3.6322e-08, 3.6322e-08, ..., -3.2503e-07, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 253, bias, value: tensor([ 0.0043, 0.0125, 0.0174, 0.0007, 0.0238, -0.0056, -0.0047, -0.0066, + 0.0104, -0.0179], device='cuda:0'), grad: tensor([-2.7567e-07, 2.3879e-06, -2.1495e-06, 2.6636e-07, 1.2852e-07, + 5.1968e-07, 2.5425e-07, 3.0547e-07, -1.1120e-06, -3.2689e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 246.14, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.5114 re_mapping 0.0035 re_causal 0.0119 /// teacc 99.11 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.2062, -0.0433, 0.1981, ..., 0.0437, -0.1884, -0.1497], + [-0.1446, -0.2176, -0.0442, ..., -0.1055, -0.2114, -0.0819], + [ 0.2437, -0.0667, -0.2336, ..., -0.1788, -0.1834, -0.2065], + ..., + [-0.0824, -0.0268, -0.2776, ..., -0.1122, 0.1080, 0.0281], + [-0.1844, -0.1217, -0.0313, ..., -0.1919, -0.0352, -0.0356], + [-0.0750, -0.0935, -0.1307, ..., 0.0937, 0.0285, -0.1608]], + device='cuda:0'), grad: tensor([[-1.9558e-08, 0.0000e+00, -9.9093e-07, ..., -4.0419e-07, + 9.3132e-09, 4.6566e-09], + [ 2.7940e-09, 9.3132e-10, 1.3970e-08, ..., 1.3132e-07, + 2.3283e-08, 1.8626e-09], + [-6.5193e-09, 1.8626e-09, 8.9407e-08, ..., 1.6112e-07, + 7.4506e-09, 1.8626e-09], + ..., + [ 1.6764e-08, 3.7253e-09, 1.5832e-08, ..., 4.4145e-07, + 1.0338e-07, 0.0000e+00], + [ 4.6566e-09, 9.3132e-10, 2.7008e-08, ..., 6.3330e-08, + 0.0000e+00, 9.3132e-10], + [ 3.7253e-09, -4.6566e-09, 1.3690e-07, ..., -1.3132e-07, + -1.5646e-07, 0.0000e+00]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0042, 0.0124, 0.0172, 0.0006, 0.0237, -0.0056, -0.0045, -0.0063, + 0.0102, -0.0179], device='cuda:0'), grad: tensor([-1.8338e-06, -4.8429e-08, 6.5379e-07, 1.9278e-07, -4.0531e-06, + 1.4715e-07, 3.3211e-06, 1.3215e-06, -1.0151e-07, 3.9767e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 253---------------------------------------------------- +epoch 253, time 262.94, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4621 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.20 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.2065, -0.0437, 0.1983, ..., 0.0437, -0.1886, -0.1497], + [-0.1452, -0.2182, -0.0443, ..., -0.1056, -0.2120, -0.0819], + [ 0.2443, -0.0668, -0.2344, ..., -0.1793, -0.1836, -0.2067], + ..., + [-0.0828, -0.0268, -0.2798, ..., -0.1124, 0.1080, 0.0281], + [-0.1846, -0.1219, -0.0315, ..., -0.1921, -0.0349, -0.0356], + [-0.0753, -0.0939, -0.1308, ..., 0.0937, 0.0286, -0.1608]], + device='cuda:0'), grad: tensor([[-2.6077e-08, 0.0000e+00, -8.0559e-07, ..., -4.1910e-07, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 1.6764e-08, + 2.6356e-07, 0.0000e+00], + [ 1.3970e-08, 0.0000e+00, 6.7055e-08, ..., 4.6566e-08, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.3039e-08, + -3.1386e-07, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 6.5193e-08, ..., 7.9162e-08, + 1.3970e-08, 9.3132e-10], + [ 6.5193e-09, -0.0000e+00, 5.0943e-07, ..., 1.3318e-07, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0043, 0.0124, 0.0176, 0.0012, 0.0237, -0.0059, -0.0047, -0.0066, + 0.0105, -0.0179], device='cuda:0'), grad: tensor([-1.5106e-06, -1.9837e-07, 6.9011e-07, 2.2538e-07, 2.1141e-07, + 5.5879e-08, 1.7136e-07, -5.5321e-07, 3.2037e-07, 6.1560e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 246.02, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4952 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.20 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.2063, -0.0438, 0.1987, ..., 0.0438, -0.1891, -0.1499], + [-0.1457, -0.2192, -0.0450, ..., -0.1056, -0.2142, -0.0819], + [ 0.2448, -0.0670, -0.2356, ..., -0.1800, -0.1854, -0.2068], + ..., + [-0.0831, -0.0267, -0.2809, ..., -0.1126, 0.1083, 0.0280], + [-0.1848, -0.1221, -0.0316, ..., -0.1925, -0.0351, -0.0357], + [-0.0753, -0.0939, -0.1310, ..., 0.0939, 0.0287, -0.1609]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8626e-09, -7.5437e-08, ..., -4.2841e-08, + 8.3819e-09, 4.6566e-09], + [ 7.4506e-09, 3.7253e-09, 5.5879e-09, ..., 8.3819e-09, + 1.8626e-09, 9.3132e-10], + [-6.5193e-09, 6.5193e-09, 5.5879e-09, ..., 8.3819e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.3039e-08, 9.3132e-09, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 1.3970e-08, 5.5879e-09, 1.1176e-08, ..., 9.3132e-09, + 2.7940e-09, 1.8626e-09], + [ 6.5193e-09, 5.5879e-09, 1.4901e-08, ..., 1.6671e-07, + 1.4901e-08, 0.0000e+00]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0045, 0.0124, 0.0175, 0.0010, 0.0235, -0.0058, -0.0048, -0.0066, + 0.0108, -0.0178], device='cuda:0'), grad: tensor([-9.0338e-08, 1.8626e-09, 2.3283e-08, -2.0489e-07, -4.3213e-07, + 4.0047e-08, 1.2852e-07, 8.1025e-08, 7.3574e-08, 3.8557e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 246.14, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4706 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.17 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.2070, -0.0443, 0.1987, ..., 0.0438, -0.1902, -0.1500], + [-0.1461, -0.2201, -0.0452, ..., -0.1056, -0.2163, -0.0819], + [ 0.2452, -0.0674, -0.2361, ..., -0.1804, -0.1861, -0.2070], + ..., + [-0.0835, -0.0267, -0.2814, ..., -0.1129, 0.1085, 0.0280], + [-0.1849, -0.1224, -0.0318, ..., -0.1928, -0.0354, -0.0359], + [-0.0756, -0.0939, -0.1310, ..., 0.0939, 0.0286, -0.1609]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -2.4214e-08, ..., -1.3970e-08, + 4.6566e-09, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 7.3574e-08, 0.0000e+00], + [-1.9372e-07, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 1.7043e-07, 0.0000e+00, 1.8626e-09, ..., 1.0245e-08, + -3.7346e-07, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 1.2107e-08, ..., 1.9558e-07, + 2.7940e-07, 0.0000e+00]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0043, 0.0127, 0.0143, 0.0011, 0.0235, -0.0057, -0.0048, -0.0055, + 0.0107, -0.0179], device='cuda:0'), grad: tensor([ 1.0468e-06, -1.2606e-05, -3.0454e-07, 3.9395e-07, -5.5321e-07, + 2.9989e-07, 3.1292e-07, 1.2498e-06, 8.4564e-06, 1.6829e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 246.01, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4805 re_mapping 0.0034 re_causal 0.0111 /// teacc 99.13 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.2074, -0.0445, 0.1983, ..., 0.0438, -0.1934, -0.1513], + [-0.1466, -0.2206, -0.0454, ..., -0.1057, -0.2175, -0.0820], + [ 0.2456, -0.0674, -0.2373, ..., -0.1812, -0.1867, -0.2071], + ..., + [-0.0835, -0.0267, -0.2830, ..., -0.1132, 0.1086, 0.0280], + [-0.1850, -0.1225, -0.0320, ..., -0.1930, -0.0358, -0.0362], + [-0.0743, -0.0939, -0.1310, ..., 0.0950, 0.0294, -0.1612]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 1.2107e-08, + 9.3132e-10, 0.0000e+00], + [-6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 1.8626e-09, ..., 1.9558e-08, + 7.4506e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.7695e-08, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0038, 0.0131, 0.0128, 0.0011, 0.0222, -0.0056, -0.0041, -0.0054, + 0.0105, -0.0167], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.8626e-09, -2.7940e-09, 3.1665e-08, -9.0338e-08, + -1.4715e-07, 1.1362e-07, 2.3283e-08, 8.7544e-08, -1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 246.21, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4792 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.17 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.2065, -0.0447, 0.1982, ..., 0.0440, -0.1945, -0.1523], + [-0.1471, -0.2212, -0.0482, ..., -0.1068, -0.2185, -0.0821], + [ 0.2461, -0.0675, -0.2379, ..., -0.1817, -0.1868, -0.2072], + ..., + [-0.0843, -0.0268, -0.2843, ..., -0.1134, 0.1082, 0.0280], + [-0.1850, -0.1227, -0.0321, ..., -0.1935, -0.0360, -0.0363], + [-0.0736, -0.0939, -0.1311, ..., 0.0956, 0.0299, -0.1613]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., 9.3132e-09, + 3.7253e-09, 9.3132e-10], + [ 8.3819e-09, 9.3132e-10, 1.8626e-09, ..., 2.7940e-08, + 6.5193e-09, 9.3132e-10], + [-3.1665e-08, 1.8626e-09, 9.3132e-10, ..., 2.7940e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 2.0489e-08, 1.8626e-09, 0.0000e+00, ..., 1.1176e-08, + -1.1176e-08, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, 2.7940e-09, ..., 7.4506e-09, + 1.8626e-09, 9.3132e-10], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., -9.6858e-08, + 3.7253e-09, 1.8626e-09]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0039, 0.0105, 0.0125, 0.0041, 0.0215, -0.0072, -0.0039, -0.0029, + 0.0106, -0.0160], device='cuda:0'), grad: tensor([ 5.1223e-08, 1.0291e-06, 3.2317e-07, 3.4273e-07, 2.3283e-08, + 4.1723e-07, 1.7975e-07, 4.5635e-08, -2.5444e-06, 1.2945e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 258---------------------------------------------------- +epoch 258, time 262.90, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.4868 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.22 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.2084, -0.0447, 0.1982, ..., 0.0440, -0.1949, -0.1526], + [-0.1472, -0.2222, -0.0481, ..., -0.1070, -0.2221, -0.0840], + [ 0.2471, -0.0675, -0.2385, ..., -0.1827, -0.1875, -0.2079], + ..., + [-0.0860, -0.0270, -0.2857, ..., -0.1139, 0.1082, 0.0278], + [-0.1852, -0.1227, -0.0324, ..., -0.1942, -0.0344, -0.0366], + [-0.0737, -0.0939, -0.1311, ..., 0.0955, 0.0299, -0.1615]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -6.5193e-09, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 4.6566e-09, 0.0000e+00], + [-5.1223e-08, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 7.4506e-09, 9.3132e-10, 0.0000e+00, ..., 9.5926e-08, + 3.8184e-08, 9.3132e-10], + [ 5.5879e-09, 1.8626e-09, 1.8626e-09, ..., 5.5879e-09, + 6.5193e-09, 3.7253e-09], + [ 3.7253e-09, 9.3132e-10, 2.7940e-09, ..., 2.7008e-08, + 4.6566e-09, 9.3132e-10]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0037, 0.0106, 0.0120, 0.0037, 0.0218, -0.0101, -0.0008, -0.0030, + 0.0122, -0.0161], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.0047e-08, -7.0781e-08, 6.7987e-08, -3.4738e-07, + -9.0990e-07, 7.6089e-07, 2.6450e-07, 7.8231e-08, 1.1735e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 246.11, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4982 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.11 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.2110, -0.0449, 0.1983, ..., 0.0430, -0.1957, -0.1533], + [-0.1477, -0.2234, -0.0488, ..., -0.1072, -0.2239, -0.0840], + [ 0.2478, -0.0676, -0.2388, ..., -0.1829, -0.1882, -0.2081], + ..., + [-0.0865, -0.0270, -0.2863, ..., -0.1142, 0.1084, 0.0282], + [-0.1855, -0.1231, -0.0327, ..., -0.1946, -0.0344, -0.0369], + [-0.0729, -0.0939, -0.1311, ..., 0.0962, 0.0301, -0.1630]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.7253e-09, ..., -1.8626e-09, + 2.7940e-09, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7008e-08, 9.3132e-10], + [-3.5390e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -5.3085e-08, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, -4.6566e-09], + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., -9.3132e-10, + 6.5193e-09, 1.8626e-09]], device='cuda:0') +Epoch 261, bias, value: tensor([ 0.0029, 0.0107, 0.0122, 0.0036, 0.0219, -0.0101, -0.0007, -0.0032, + 0.0121, -0.0156], device='cuda:0'), grad: tensor([ 1.9558e-08, -1.9714e-05, -4.6566e-08, 3.7514e-06, -1.4901e-08, + -1.2852e-07, -2.7008e-08, 1.5840e-05, -3.5390e-08, 3.4273e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 246.80, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4963 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.11 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.2112, -0.0450, 0.1978, ..., 0.0430, -0.1968, -0.1549], + [-0.1489, -0.2237, -0.0489, ..., -0.1073, -0.2247, -0.0842], + [ 0.2488, -0.0676, -0.2393, ..., -0.1832, -0.1885, -0.2082], + ..., + [-0.0867, -0.0270, -0.2868, ..., -0.1145, 0.1085, 0.0285], + [-0.1861, -0.1236, -0.0327, ..., -0.1949, -0.0351, -0.0382], + [-0.0729, -0.0939, -0.1312, ..., 0.0962, 0.0301, -0.1668]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -5.6811e-08, ..., -2.7940e-08, + 2.7940e-09, 1.8626e-09], + [ 7.4506e-09, 2.7940e-09, 5.5879e-09, ..., 2.7940e-09, + 5.5879e-09, 1.8626e-09], + [-2.8871e-08, 1.8626e-09, 8.3819e-09, ..., 2.7940e-09, + 2.7940e-09, 1.8626e-09], + ..., + [ 2.3283e-08, 1.3039e-08, 9.3132e-10, ..., 3.2596e-08, + 2.0489e-08, 9.3132e-10], + [ 9.3132e-09, 0.0000e+00, 2.0489e-08, ..., 4.6566e-09, + 8.3819e-09, 9.3132e-09], + [ 9.3132e-10, 1.8626e-09, 3.5390e-08, ..., 1.8161e-07, + 4.6566e-09, 9.3132e-10]], device='cuda:0') +Epoch 262, bias, value: tensor([ 0.0025, 0.0107, 0.0125, 0.0035, 0.0220, -0.0093, -0.0013, -0.0031, + 0.0117, -0.0157], device='cuda:0'), grad: tensor([-5.7742e-08, -4.0047e-07, -6.5193e-09, -5.5879e-08, -3.3621e-07, + -2.3562e-07, 1.1455e-07, 5.0850e-07, 1.0058e-07, 3.5018e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 246.08, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4695 re_mapping 0.0034 re_causal 0.0110 /// teacc 99.14 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.2111, -0.0451, 0.1980, ..., 0.0431, -0.1970, -0.1551], + [-0.1494, -0.2240, -0.0501, ..., -0.1075, -0.2253, -0.0848], + [ 0.2495, -0.0677, -0.2421, ..., -0.1846, -0.1888, -0.2096], + ..., + [-0.0869, -0.0270, -0.2885, ..., -0.1155, 0.1085, 0.0263], + [-0.1866, -0.1238, -0.0332, ..., -0.1955, -0.0358, -0.0400], + [-0.0732, -0.0939, -0.1313, ..., 0.0934, 0.0288, -0.1706]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -8.8476e-08, ..., -5.8673e-08, + 9.3132e-09, 6.5193e-09], + [ 3.7253e-09, 0.0000e+00, 1.7695e-08, ..., 2.7008e-08, + 7.4506e-09, 4.6566e-09], + [-3.0082e-07, -9.3132e-10, 1.3970e-08, ..., 4.6566e-09, + 5.5879e-09, 4.6566e-09], + ..., + [ 2.8685e-07, 9.3132e-10, 1.8626e-09, ..., 1.0245e-08, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 2.2352e-08, ..., 4.7497e-08, + 5.5879e-09, 4.6566e-09], + [ 1.8626e-09, 0.0000e+00, 2.8871e-08, ..., 3.3434e-07, + -1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0025, 0.0106, 0.0125, 0.0032, 0.0250, -0.0087, -0.0016, -0.0031, + 0.0109, -0.0187], device='cuda:0'), grad: tensor([-7.0781e-08, 1.0524e-07, -3.1479e-07, 3.8370e-06, 7.6741e-07, + -1.0796e-05, 8.1304e-07, 4.6752e-07, 9.5181e-07, 4.2506e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 245.97, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4872 re_mapping 0.0035 re_causal 0.0113 /// teacc 99.19 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.2098, -0.0452, 0.1983, ..., 0.0432, -0.1971, -0.1551], + [-0.1510, -0.2244, -0.0508, ..., -0.1080, -0.2260, -0.0850], + [ 0.2500, -0.0677, -0.2438, ..., -0.1860, -0.1890, -0.2099], + ..., + [-0.0871, -0.0271, -0.2899, ..., -0.1188, 0.1056, 0.0261], + [-0.1867, -0.1247, -0.0333, ..., -0.1961, -0.0360, -0.0402], + [-0.0732, -0.0940, -0.1314, ..., 0.0944, 0.0322, -0.1708]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.0245e-07, ..., -7.3574e-08, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 1.8626e-09, 4.6566e-09, ..., -4.6566e-09, + 9.3132e-10, 2.7940e-09], + [ 1.8626e-09, 9.3132e-10, 4.6566e-09, ..., 6.5193e-09, + 9.3132e-10, 1.8626e-09], + ..., + [ 4.6566e-09, 2.7940e-09, 6.5193e-09, ..., 1.1176e-08, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 2.5146e-08, ..., 1.3039e-08, + 4.6566e-09, -1.4901e-08], + [ 2.7940e-09, 3.7253e-09, 4.7497e-08, ..., 6.8918e-08, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 264, bias, value: tensor([ 0.0027, 0.0105, 0.0124, 0.0019, 0.0247, -0.0082, -0.0017, -0.0039, + 0.0107, -0.0174], device='cuda:0'), grad: tensor([-2.2631e-07, -9.3132e-09, 1.2293e-07, 9.3132e-10, -2.1048e-07, + 8.7544e-08, 2.2445e-07, 1.1269e-07, -3.8557e-07, 2.8312e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 246.17, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4788 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.04 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.2091, -0.0452, 0.1989, ..., 0.0434, -0.1979, -0.1561], + [-0.1512, -0.2249, -0.0516, ..., -0.1080, -0.2265, -0.0850], + [ 0.2503, -0.0677, -0.2451, ..., -0.1872, -0.1893, -0.2099], + ..., + [-0.0877, -0.0271, -0.2909, ..., -0.1202, 0.1052, 0.0262], + [-0.1868, -0.1249, -0.0334, ..., -0.1965, -0.0361, -0.0405], + [-0.0730, -0.0940, -0.1315, ..., 0.0946, 0.0326, -0.1708]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 0.0000e+00, -8.9128e-07, ..., -6.2957e-07, + 9.3132e-10, -1.3039e-08], + [ 3.9116e-08, 9.3132e-10, 2.4214e-08, ..., 5.5879e-08, + 3.7253e-08, 6.5193e-09], + [-3.7439e-07, 9.3132e-10, 2.4214e-08, ..., 3.8184e-08, + 9.3132e-10, -4.6566e-08], + ..., + [ 3.2037e-07, 1.8626e-09, 9.3132e-09, ..., -1.1176e-08, + -5.6811e-08, 4.0047e-08], + [ 4.6566e-09, 9.3132e-10, 4.0047e-08, ..., 3.1665e-08, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 9.3132e-10, 4.8056e-07, ..., 3.6694e-07, + 8.3819e-09, 1.8626e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0029, 0.0105, 0.0123, 0.0017, 0.0247, -0.0081, -0.0017, -0.0042, + 0.0105, -0.0171], device='cuda:0'), grad: tensor([-1.6913e-06, 4.6194e-07, -3.9395e-07, 1.4063e-07, -1.5367e-07, + 1.2293e-07, 4.1071e-07, -2.4680e-07, 1.0803e-07, 1.2526e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 246.22, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.5080 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.15 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.2089, -0.0453, 0.1997, ..., 0.0437, -0.1979, -0.1562], + [-0.1520, -0.2256, -0.0528, ..., -0.1088, -0.2280, -0.0850], + [ 0.2515, -0.0677, -0.2462, ..., -0.1886, -0.1899, -0.2096], + ..., + [-0.0891, -0.0272, -0.2932, ..., -0.1203, 0.1053, 0.0257], + [-0.1870, -0.1250, -0.0338, ..., -0.1988, -0.0362, -0.0406], + [-0.0729, -0.0940, -0.1316, ..., 0.0964, 0.0326, -0.1708]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 9.3132e-10, -2.7008e-08, ..., -3.4459e-08, + 2.6077e-08, 1.3970e-08], + [ 8.3819e-09, 9.3132e-10, 1.2107e-08, ..., 8.3819e-09, + 3.7253e-09, 9.3132e-10], + [-5.3085e-08, 1.8626e-09, 6.5193e-09, ..., 3.7253e-09, + 1.8626e-09, 1.8626e-09], + ..., + [ 3.3528e-08, 3.7253e-09, 9.3132e-10, ..., 6.5193e-09, + -2.7940e-09, 0.0000e+00], + [ 4.6566e-09, 9.3132e-10, 3.4459e-08, ..., 1.8626e-08, + 6.5193e-09, 3.7253e-09], + [ 1.0245e-08, 3.7253e-09, 3.9116e-08, ..., 7.2643e-08, + 4.6566e-09, 2.7940e-09]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0034, 0.0104, 0.0125, 0.0015, 0.0229, -0.0081, -0.0019, -0.0041, + 0.0100, -0.0153], device='cuda:0'), grad: tensor([-2.7008e-08, -1.4994e-07, -5.1223e-08, -2.9709e-07, -8.5682e-08, + 1.9614e-06, -1.9446e-06, 2.3749e-07, 1.3225e-07, 2.3469e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 245.67, cls_loss 0.0006 cls_loss_mapping 0.0023 cls_loss_causal 0.4885 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.2089, -0.0457, 0.1999, ..., 0.0437, -0.1981, -0.1564], + [-0.1527, -0.2262, -0.0530, ..., -0.1089, -0.2290, -0.0851], + [ 0.2521, -0.0678, -0.2471, ..., -0.1887, -0.1906, -0.2100], + ..., + [-0.0894, -0.0272, -0.2948, ..., -0.1203, 0.1054, 0.0260], + [-0.1872, -0.1252, -0.0340, ..., -0.1995, -0.0364, -0.0408], + [-0.0732, -0.0941, -0.1317, ..., 0.0964, 0.0325, -0.1708]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, -9.1270e-08, ..., -6.8918e-08, + 3.7253e-09, 9.3132e-10], + [-8.1025e-08, 8.3819e-09, 3.7253e-09, ..., 1.8626e-09, + 3.6322e-08, 0.0000e+00], + [ 1.1921e-07, 5.5879e-09, 3.7253e-09, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 2.1420e-08, 8.3819e-09, 3.7253e-09, ..., 3.7253e-09, + -5.9605e-08, 0.0000e+00], + [ 1.0245e-08, 2.7940e-09, 9.3132e-09, ..., 6.5193e-09, + 2.7940e-09, 9.3132e-10], + [ 1.2107e-08, 3.7253e-09, 4.0047e-08, ..., 2.1420e-08, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0035, 0.0105, 0.0124, 0.0015, 0.0230, -0.0081, -0.0019, -0.0041, + 0.0093, -0.0155], device='cuda:0'), grad: tensor([-1.5367e-07, -2.5090e-06, 2.7195e-06, -3.1386e-07, 2.4214e-08, + 1.9092e-07, -4.8429e-08, -1.2200e-07, 6.5193e-08, 1.6484e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 245.76, cls_loss 0.0006 cls_loss_mapping 0.0031 cls_loss_causal 0.4658 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.16 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.2095, -0.0458, 0.2004, ..., 0.0437, -0.1980, -0.1564], + [-0.1536, -0.2273, -0.0531, ..., -0.1090, -0.2307, -0.0854], + [ 0.2516, -0.0693, -0.2486, ..., -0.1900, -0.1914, -0.2105], + ..., + [-0.0891, -0.0274, -0.2956, ..., -0.1203, 0.1055, 0.0258], + [-0.1874, -0.1252, -0.0342, ..., -0.1999, -0.0365, -0.0419], + [-0.0732, -0.0941, -0.1320, ..., 0.0963, 0.0325, -0.1708]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.7940e-09, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-08, 4.6566e-09, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 9.3132e-10], + [-2.7008e-08, 8.3819e-09, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.9802e-08, 1.7695e-08, 0.0000e+00, ..., 3.7253e-09, + -9.3132e-10, 9.3132e-10], + [ 7.4506e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 1.0245e-08, 9.3132e-09, 0.0000e+00, ..., 5.5879e-09, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0036, 0.0105, 0.0112, 0.0020, 0.0230, -0.0079, -0.0020, -0.0041, + 0.0093, -0.0155], device='cuda:0'), grad: tensor([ 6.1467e-08, 6.4261e-08, -3.7253e-09, 2.6338e-06, 1.3039e-08, + -4.0345e-06, 2.0582e-07, 3.1292e-07, 4.5728e-07, 2.8405e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 246.50, cls_loss 0.0005 cls_loss_mapping 0.0030 cls_loss_causal 0.4801 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.09 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.2098, -0.0460, 0.2005, ..., 0.0435, -0.1981, -0.1564], + [-0.1541, -0.2283, -0.0531, ..., -0.1095, -0.2317, -0.0854], + [ 0.2523, -0.0694, -0.2490, ..., -0.1907, -0.1916, -0.2107], + ..., + [-0.0896, -0.0274, -0.2961, ..., -0.1204, 0.1056, 0.0258], + [-0.1878, -0.1254, -0.0341, ..., -0.2000, -0.0365, -0.0419], + [-0.0734, -0.0941, -0.1321, ..., 0.0964, 0.0325, -0.1708]], + device='cuda:0'), grad: tensor([[ 1.8161e-07, 0.0000e+00, 1.5832e-08, ..., 9.3132e-09, + 2.7940e-09, 3.7253e-09], + [ 7.4506e-09, 0.0000e+00, 1.2107e-08, ..., 1.8626e-09, + 1.3039e-08, 9.3132e-10], + [-4.9639e-07, -0.0000e+00, 1.2107e-08, ..., -3.4459e-08, + 3.7253e-09, 0.0000e+00], + ..., + [ 7.7300e-08, 1.8626e-09, 1.8626e-09, ..., 1.3039e-08, + -3.0734e-08, 0.0000e+00], + [ 8.3819e-09, 0.0000e+00, -8.6613e-08, ..., 1.1176e-08, + 6.5193e-09, 1.8626e-09], + [ 5.5879e-09, 0.0000e+00, 6.5193e-09, ..., -3.5390e-08, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 269, bias, value: tensor([ 0.0034, 0.0105, 0.0113, 0.0020, 0.0231, -0.0072, -0.0027, -0.0041, + 0.0093, -0.0155], device='cuda:0'), grad: tensor([ 7.0315e-07, 5.4948e-08, -6.9570e-07, 4.9826e-07, 5.1223e-08, + -1.5367e-07, 8.0466e-07, 2.5518e-07, -1.6205e-06, 1.0058e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 246.37, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.5013 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.18 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.2095, -0.0461, 0.2006, ..., 0.0436, -0.1984, -0.1565], + [-0.1544, -0.2289, -0.0535, ..., -0.1099, -0.2334, -0.0859], + [ 0.2533, -0.0694, -0.2501, ..., -0.1913, -0.1923, -0.2113], + ..., + [-0.0909, -0.0275, -0.2977, ..., -0.1205, 0.1057, 0.0256], + [-0.1879, -0.1254, -0.0340, ..., -0.2001, -0.0364, -0.0420], + [-0.0739, -0.0942, -0.1322, ..., 0.0964, 0.0324, -0.1708]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -2.7940e-08, ..., 1.8626e-09, + 2.7940e-09, 9.3132e-10], + [ 5.0012e-07, 0.0000e+00, 9.3132e-10, ..., 3.2969e-07, + 1.1176e-08, 0.0000e+00], + [ 2.4866e-07, 0.0000e+00, -0.0000e+00, ..., 4.6566e-09, + 9.3132e-10, 0.0000e+00], + ..., + [-7.7020e-07, 0.0000e+00, 9.3132e-10, ..., 2.7008e-08, + -1.9558e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.4901e-08, + 2.7940e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 5.5879e-09, ..., -3.8929e-07, + -2.4214e-08, 0.0000e+00]], device='cuda:0') +Epoch 270, bias, value: tensor([ 0.0035, 0.0105, 0.0113, 0.0018, 0.0230, -0.0071, -0.0027, -0.0041, + 0.0097, -0.0155], device='cuda:0'), grad: tensor([ 7.5437e-08, 8.7395e-06, 4.4517e-06, 1.4529e-07, -5.3272e-07, + 7.0781e-08, 8.6613e-08, -1.2293e-05, -1.1176e-08, -7.1153e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 246.71, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4739 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.10 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.2096, -0.0463, 0.2010, ..., 0.0435, -0.1987, -0.1566], + [-0.1552, -0.2300, -0.0542, ..., -0.1107, -0.2342, -0.0859], + [ 0.2543, -0.0694, -0.2512, ..., -0.1927, -0.1929, -0.2115], + ..., + [-0.0917, -0.0275, -0.3025, ..., -0.1206, 0.1057, 0.0252], + [-0.1881, -0.1256, -0.0342, ..., -0.2005, -0.0365, -0.0421], + [-0.0741, -0.0944, -0.1323, ..., 0.0965, 0.0325, -0.1708]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -2.4214e-08, + 4.6566e-09, -9.3132e-10], + [ 2.7940e-09, 9.3132e-10, 1.8626e-09, ..., 6.5193e-09, + 9.3132e-09, 0.0000e+00], + [-1.1176e-08, 9.3132e-10, 1.8626e-09, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + ..., + [ 1.2107e-08, 9.3132e-10, 1.8626e-09, ..., 3.6322e-08, + -6.5193e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 1.0524e-07, + 4.5635e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., -1.1083e-07, + -6.1467e-08, 9.3132e-10]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0035, 0.0105, 0.0114, 0.0017, 0.0231, -0.0071, -0.0027, -0.0042, + 0.0103, -0.0155], device='cuda:0'), grad: tensor([ 2.9895e-07, 3.0734e-08, 2.1420e-08, 6.7055e-08, -7.7300e-08, + 4.1910e-08, 4.7497e-08, 1.1176e-07, -2.6636e-07, -2.9523e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 246.30, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.5153 re_mapping 0.0036 re_causal 0.0116 /// teacc 99.09 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.2098, -0.0464, 0.2020, ..., 0.0435, -0.1991, -0.1562], + [-0.1557, -0.2313, -0.0543, ..., -0.1109, -0.2354, -0.0860], + [ 0.2546, -0.0696, -0.2529, ..., -0.1945, -0.1955, -0.2134], + ..., + [-0.0915, -0.0275, -0.3053, ..., -0.1210, 0.1056, 0.0274], + [-0.1882, -0.1261, -0.0343, ..., -0.2012, -0.0369, -0.0422], + [-0.0741, -0.0944, -0.1324, ..., 0.0965, 0.0327, -0.1708]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 1.5702e-06, ..., -7.4506e-08, + 7.7114e-07, 5.5879e-07], + [-6.8359e-07, 0.0000e+00, 3.2596e-08, ..., 3.0734e-08, + 2.4214e-08, 2.7940e-09], + [ 5.6066e-07, 0.0000e+00, 1.3970e-08, ..., 1.4901e-08, + 1.3039e-08, 2.7940e-09], + ..., + [ 8.1025e-08, 0.0000e+00, 8.3819e-09, ..., 1.4901e-07, + 2.1420e-08, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 3.2596e-08, ..., 2.6077e-08, + 2.3283e-08, 8.3819e-09], + [ 1.8626e-09, 0.0000e+00, 4.4703e-08, ..., 4.6566e-09, + 4.6566e-08, 2.7940e-09]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0039, 0.0104, 0.0110, 0.0015, 0.0232, -0.0070, -0.0030, -0.0042, + 0.0104, -0.0154], device='cuda:0'), grad: tensor([ 3.0026e-06, -7.1749e-06, 6.1691e-06, 1.8906e-07, -7.0129e-07, + -5.3365e-07, -2.3339e-06, 1.0747e-06, 2.3283e-07, 3.5390e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 245.98, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4690 re_mapping 0.0035 re_causal 0.0111 /// teacc 98.99 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.2102, -0.0469, 0.2023, ..., 0.0436, -0.1996, -0.1562], + [-0.1558, -0.2327, -0.0544, ..., -0.1110, -0.2367, -0.0861], + [ 0.2554, -0.0696, -0.2531, ..., -0.1956, -0.1963, -0.2138], + ..., + [-0.0918, -0.0275, -0.3080, ..., -0.1211, 0.1059, 0.0279], + [-0.1886, -0.1264, -0.0348, ..., -0.2014, -0.0371, -0.0425], + [-0.0742, -0.0944, -0.1326, ..., 0.0965, 0.0325, -0.1708]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, -1.3178e-07, ..., -8.6613e-08, + 1.6298e-08, 5.1223e-09], + [ 9.3132e-10, 9.3132e-10, 2.3283e-09, ..., 6.8918e-08, + 3.6322e-08, 1.6298e-08], + [ 0.0000e+00, 4.6566e-10, 4.1910e-09, ..., 2.7474e-08, + 1.3039e-08, 6.0536e-09], + ..., + [ 1.3970e-09, 1.8626e-09, 6.0536e-09, ..., -1.1977e-06, + -5.4482e-07, -2.9616e-07], + [ 4.6566e-10, 4.6566e-10, 6.5193e-09, ..., 7.4506e-09, + 5.1223e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 9.0804e-08, ..., 1.4715e-07, + 3.7253e-08, 2.2352e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([ 0.0040, 0.0104, 0.0109, 0.0014, 0.0232, -0.0069, -0.0031, -0.0042, + 0.0102, -0.0155], device='cuda:0'), grad: tensor([-4.5169e-08, 6.1980e-07, 2.9476e-07, 7.5549e-06, 1.1176e-07, + -1.6317e-06, 1.0133e-06, -8.7991e-06, 9.9186e-08, 7.8278e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 246.49, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.5005 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.2105, -0.0471, 0.2029, ..., 0.0437, -0.2003, -0.1565], + [-0.1566, -0.2333, -0.0555, ..., -0.1116, -0.2383, -0.0861], + [ 0.2576, -0.0697, -0.2539, ..., -0.1967, -0.1969, -0.2139], + ..., + [-0.0943, -0.0275, -0.3104, ..., -0.1209, 0.1069, 0.0291], + [-0.1886, -0.1265, -0.0344, ..., -0.2006, -0.0373, -0.0425], + [-0.0744, -0.0945, -0.1329, ..., 0.0956, 0.0306, -0.1709]], + device='cuda:0'), grad: tensor([[ 4.6100e-08, 0.0000e+00, -5.4948e-08, ..., -3.6322e-08, + 8.2422e-08, 3.7253e-08], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., -1.3970e-09, + 9.3132e-09, 4.6566e-10], + [ 7.4506e-09, 0.0000e+00, 2.3749e-08, ..., 2.2352e-08, + 1.9558e-08, 9.3132e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 1.6298e-08, + -1.8161e-08, 4.6566e-10], + [ 8.3819e-09, 0.0000e+00, 4.5635e-08, ..., 3.8184e-08, + 1.7229e-08, 7.9162e-09], + [ 9.3132e-10, 0.0000e+00, 8.7544e-08, ..., 5.1223e-08, + -1.3970e-09, 1.8626e-09]], device='cuda:0') +Epoch 274, bias, value: tensor([ 0.0043, 0.0104, 0.0118, 0.0005, 0.0240, -0.0069, -0.0032, -0.0041, + 0.0109, -0.0164], device='cuda:0'), grad: tensor([ 8.1956e-08, -3.8650e-08, 8.4285e-08, 4.9826e-08, 8.4285e-08, + 1.7090e-07, -7.0361e-07, 2.0955e-08, 1.2573e-07, 1.4435e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 246.91, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4930 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.17 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.2108, -0.0478, 0.2039, ..., 0.0441, -0.2010, -0.1566], + [-0.1573, -0.2349, -0.0557, ..., -0.1119, -0.2397, -0.0862], + [ 0.2598, -0.0697, -0.2562, ..., -0.1975, -0.1980, -0.2146], + ..., + [-0.0968, -0.0276, -0.3139, ..., -0.1209, 0.1070, 0.0290], + [-0.1891, -0.1272, -0.0351, ..., -0.2017, -0.0378, -0.0426], + [-0.0750, -0.0948, -0.1339, ..., 0.0955, 0.0306, -0.1709]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 1.3970e-09, -6.0024e-07, ..., -4.4471e-07, + 7.9162e-09, 3.7253e-09], + [ 4.1910e-09, 1.8626e-09, 1.3504e-08, ..., -5.1223e-09, + 9.3132e-10, 0.0000e+00], + [-8.3819e-09, 1.3970e-09, 3.8184e-08, ..., 2.5611e-08, + 1.3970e-09, 4.6566e-10], + ..., + [ 9.3132e-09, 3.2596e-09, 2.7940e-08, ..., 2.4680e-08, + -4.6566e-10, 0.0000e+00], + [ 4.1910e-09, 2.7940e-09, 2.1420e-08, ..., 3.1199e-08, + 8.3819e-09, 4.6566e-10], + [ 2.3283e-09, 1.8626e-09, 2.4494e-07, ..., 1.3271e-07, + -1.4435e-08, 4.6566e-10]], device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0049, 0.0105, 0.0128, 0.0004, 0.0240, -0.0069, -0.0031, -0.0042, + 0.0105, -0.0165], device='cuda:0'), grad: tensor([-1.2582e-06, -1.7416e-07, 7.4971e-08, -6.4261e-08, 9.1735e-08, + 3.4925e-08, 4.9220e-07, 1.7183e-07, 1.1036e-07, 5.2294e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 246.46, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4909 re_mapping 0.0035 re_causal 0.0113 /// teacc 99.21 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.2111, -0.0484, 0.2043, ..., 0.0437, -0.2014, -0.1567], + [-0.1572, -0.2373, -0.0558, ..., -0.1120, -0.2415, -0.0862], + [ 0.2601, -0.0699, -0.2574, ..., -0.1985, -0.1994, -0.2153], + ..., + [-0.0975, -0.0278, -0.3162, ..., -0.1209, 0.1072, 0.0289], + [-0.1893, -0.1294, -0.0358, ..., -0.2023, -0.0386, -0.0428], + [-0.0756, -0.0950, -0.1341, ..., 0.0956, 0.0304, -0.1709]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6124e-07, ..., -1.7183e-07, + 1.1176e-08, 1.8626e-09], + [ 4.6566e-10, 0.0000e+00, 1.2107e-08, ..., 1.9558e-08, + 2.3283e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 5.1223e-09, + 1.0245e-08, 4.6566e-10], + ..., + [ 4.6566e-10, 0.0000e+00, 4.6566e-09, ..., 7.4506e-09, + 1.6484e-07, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 1.5832e-08, ..., 3.4925e-08, + -2.6403e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.0198e-07, ..., 3.7532e-07, + 6.9384e-08, 4.6566e-10]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0047, 0.0105, 0.0110, -0.0010, 0.0241, -0.0049, -0.0032, -0.0042, + 0.0115, -0.0166], device='cuda:0'), grad: tensor([-4.4843e-07, -6.4913e-07, 8.5216e-08, 1.5879e-07, -6.5658e-07, + 7.6368e-08, 1.6624e-07, 1.9185e-06, -1.6671e-06, 1.0226e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 246.66, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4996 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.15 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.2114, -0.0492, 0.2045, ..., 0.0437, -0.2017, -0.1567], + [-0.1576, -0.2384, -0.0559, ..., -0.1122, -0.2429, -0.0863], + [ 0.2624, -0.0700, -0.2578, ..., -0.1988, -0.1998, -0.2156], + ..., + [-0.1005, -0.0279, -0.3170, ..., -0.1209, 0.1074, 0.0292], + [-0.1898, -0.1296, -0.0361, ..., -0.2042, -0.0391, -0.0432], + [-0.0755, -0.0950, -0.1342, ..., 0.0958, 0.0306, -0.1709]], + device='cuda:0'), grad: tensor([[ 2.3749e-08, 3.7253e-09, -5.0291e-08, ..., 1.9791e-07, + 1.2107e-08, 2.0489e-08], + [ 2.5611e-08, 3.7253e-09, 1.8626e-09, ..., 2.0489e-08, + 1.0710e-08, 5.1223e-09], + [-1.4435e-07, -9.3132e-10, 5.5879e-09, ..., 3.0268e-08, + 1.8626e-09, 1.2107e-08], + ..., + [ 5.7276e-08, 6.0536e-09, 9.3132e-10, ..., 4.6566e-09, + -1.6298e-08, 0.0000e+00], + [ 1.7695e-08, 4.6566e-10, 5.1223e-09, ..., 4.0513e-08, + 6.5193e-09, -2.7940e-09], + [ 6.9849e-09, 2.7940e-09, 2.5146e-08, ..., -2.5053e-07, + -2.6077e-08, 1.4901e-08]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0047, 0.0108, 0.0120, -0.0010, 0.0239, -0.0048, -0.0032, -0.0045, + 0.0109, -0.0165], device='cuda:0'), grad: tensor([ 4.2887e-07, 1.8300e-07, -1.9092e-07, -9.0804e-08, -5.8711e-06, + 7.8231e-08, 5.8524e-06, 3.9581e-08, 9.7789e-09, -4.2608e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 246.26, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4922 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.00 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.2120, -0.0493, 0.2047, ..., 0.0437, -0.2020, -0.1568], + [-0.1586, -0.2391, -0.0559, ..., -0.1125, -0.2452, -0.0865], + [ 0.2641, -0.0701, -0.2567, ..., -0.1997, -0.2003, -0.2139], + ..., + [-0.1012, -0.0279, -0.3178, ..., -0.1208, 0.1077, 0.0294], + [-0.1905, -0.1297, -0.0365, ..., -0.2051, -0.0397, -0.0423], + [-0.0765, -0.0951, -0.1343, ..., 0.0957, 0.0304, -0.1709]], + device='cuda:0'), grad: tensor([[ 3.9209e-07, 1.8626e-09, 1.5739e-07, ..., 3.5297e-07, + 1.5832e-08, 3.0361e-07], + [ 9.3132e-08, 9.3132e-10, 1.1176e-08, ..., 9.2201e-08, + 5.7742e-08, 1.3970e-08], + [ 6.1933e-07, 1.8626e-09, 3.1386e-07, ..., 7.2550e-07, + 4.9360e-08, 5.7463e-07], + ..., + [-2.7008e-08, 0.0000e+00, 1.8626e-09, ..., 1.6205e-07, + -9.2201e-08, 1.8626e-09], + [ 9.1270e-08, 9.3132e-10, 3.7253e-08, ..., 7.4506e-08, + 9.3132e-10, 6.1467e-08], + [ 3.4459e-08, 0.0000e+00, 2.1420e-08, ..., -3.8929e-07, + -2.7008e-08, 1.4901e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0046, 0.0110, 0.0128, -0.0011, 0.0240, -0.0047, -0.0034, -0.0047, + 0.0114, -0.0166], device='cuda:0'), grad: tensor([ 1.9781e-06, -7.9796e-06, 3.7029e-06, 7.7765e-07, 2.5928e-06, + 1.0401e-05, -1.2808e-05, 9.2909e-06, -9.6336e-06, 1.6065e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 246.28, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4906 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.15 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.2123, -0.0496, 0.2049, ..., 0.0434, -0.2026, -0.1570], + [-0.1625, -0.2397, -0.0561, ..., -0.1127, -0.2473, -0.0867], + [ 0.2676, -0.0702, -0.2579, ..., -0.1984, -0.2010, -0.2154], + ..., + [-0.1019, -0.0281, -0.3198, ..., -0.1209, 0.1079, 0.0297], + [-0.1908, -0.1299, -0.0373, ..., -0.2058, -0.0401, -0.0428], + [-0.0775, -0.0951, -0.1346, ..., 0.0958, 0.0305, -0.1709]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -2.1514e-07, ..., -1.2666e-07, + 9.3132e-10, -9.3132e-10], + [ 8.1956e-08, 0.0000e+00, 2.2352e-08, ..., 2.0489e-08, + 2.7940e-09, 0.0000e+00], + [-1.1269e-06, 0.0000e+00, 9.3132e-09, ..., 8.3819e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.7882e-07, 2.7940e-09, 1.0245e-08, ..., -9.3132e-10, + -1.0245e-08, 0.0000e+00], + [ 1.6764e-08, 0.0000e+00, 2.5146e-08, ..., 2.0489e-08, + -9.3132e-10, 0.0000e+00], + [ 9.3132e-09, 0.0000e+00, 8.5682e-08, ..., 2.8871e-08, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0045, 0.0102, 0.0150, -0.0012, 0.0240, -0.0048, -0.0032, -0.0041, + 0.0115, -0.0166], device='cuda:0'), grad: tensor([-3.8557e-07, 2.8871e-07, -2.2408e-06, 6.4261e-08, 5.6811e-08, + 3.7253e-08, 1.0338e-07, 1.7826e-06, 7.0781e-08, 2.0955e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 246.77, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4949 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.13 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.2130, -0.0497, 0.2049, ..., 0.0435, -0.2039, -0.1578], + [-0.1630, -0.2404, -0.0561, ..., -0.1147, -0.2488, -0.0868], + [ 0.2681, -0.0703, -0.2596, ..., -0.1989, -0.2021, -0.2171], + ..., + [-0.1022, -0.0282, -0.3211, ..., -0.1211, 0.1080, 0.0291], + [-0.1910, -0.1302, -0.0376, ..., -0.2066, -0.0416, -0.0449], + [-0.0792, -0.0953, -0.1347, ..., 0.0958, 0.0303, -0.1710]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, 9.3132e-10, -1.4789e-06, ..., 1.2293e-07, + -2.3283e-08, 8.3819e-09], + [ 2.7940e-09, 9.3132e-10, 1.3039e-08, ..., 1.3504e-07, + 4.6566e-09, 7.4506e-09], + [ 2.9802e-08, 2.7940e-09, 4.6566e-09, ..., 5.1223e-08, + 1.0245e-08, 2.7940e-09], + ..., + [-1.3039e-08, 9.3132e-10, 6.5193e-09, ..., 3.7625e-07, + 0.0000e+00, 1.6764e-08], + [ 9.3132e-09, 1.8626e-09, 2.6077e-08, ..., 4.0978e-08, + 5.5879e-09, 1.8626e-09], + [ 2.5146e-08, 1.8626e-09, 5.1223e-08, ..., 1.6853e-05, + 1.2107e-07, 9.3319e-07]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0043, 0.0099, 0.0149, -0.0011, 0.0241, -0.0048, -0.0030, -0.0040, + 0.0103, -0.0166], device='cuda:0'), grad: tensor([-2.4140e-06, -3.2596e-08, 4.1910e-07, -2.7940e-07, -4.2707e-05, + 2.7567e-07, 2.7530e-06, 8.7358e-07, 2.9150e-07, 4.0889e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4762 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.21 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.2134, -0.0499, 0.2057, ..., 0.0437, -0.2039, -0.1578], + [-0.1631, -0.2408, -0.0563, ..., -0.1151, -0.2504, -0.0870], + [ 0.2683, -0.0703, -0.2615, ..., -0.2003, -0.2032, -0.2177], + ..., + [-0.1023, -0.0282, -0.3226, ..., -0.1210, 0.1085, 0.0293], + [-0.1906, -0.1303, -0.0373, ..., -0.2067, -0.0423, -0.0457], + [-0.0796, -0.0953, -0.1353, ..., 0.0956, 0.0300, -0.1712]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 1.3039e-08, + 2.7940e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + 1.9558e-08, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.3283e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + -7.4506e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.8871e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -2.0675e-07, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0047, 0.0099, 0.0146, -0.0011, 0.0243, -0.0047, -0.0030, -0.0039, + 0.0105, -0.0169], device='cuda:0'), grad: tensor([ 2.6077e-08, 3.8184e-08, 1.5367e-07, 6.7987e-08, 2.8871e-07, + -9.3132e-09, 2.6077e-08, -3.5670e-07, 9.3132e-08, -3.2131e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 246.63, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4326 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.16 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.2137, -0.0507, 0.2059, ..., 0.0430, -0.2040, -0.1577], + [-0.1631, -0.2411, -0.0561, ..., -0.1152, -0.2516, -0.0872], + [ 0.2685, -0.0704, -0.2627, ..., -0.2013, -0.2041, -0.2181], + ..., + [-0.1023, -0.0282, -0.3242, ..., -0.1209, 0.1089, 0.0298], + [-0.1910, -0.1306, -0.0377, ..., -0.2073, -0.0426, -0.0459], + [-0.0798, -0.0954, -0.1354, ..., 0.0957, 0.0297, -0.1712]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -4.6566e-09, -5.4482e-07, ..., -2.4214e-07, + 2.7940e-09, -2.0489e-08], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 7.4506e-09, + 9.3132e-09, 9.3132e-10], + [-7.4506e-09, 9.3132e-10, 2.1420e-08, ..., 1.4901e-08, + 8.3819e-09, -2.7940e-09], + ..., + [ 1.8626e-09, 9.3132e-10, 1.4901e-08, ..., 1.1176e-08, + -3.9116e-08, 9.3132e-10], + [ 3.7253e-09, 2.7940e-09, 7.0781e-08, ..., 3.2596e-08, + 1.8626e-09, 3.7253e-09], + [ 9.3132e-10, 9.3132e-10, 2.1793e-07, ..., 2.0582e-07, + 3.2596e-08, 1.2107e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0042, 0.0099, 0.0140, -0.0012, 0.0243, -0.0047, -0.0031, -0.0037, + 0.0102, -0.0169], device='cuda:0'), grad: tensor([-1.1008e-06, 1.1269e-07, 4.8056e-07, 1.5274e-07, -2.9802e-07, + 2.7008e-07, 3.0920e-07, -1.0021e-06, 1.9465e-07, 8.8103e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 246.12, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4959 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.18 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.2149, -0.0508, 0.2060, ..., 0.0430, -0.2043, -0.1578], + [-0.1624, -0.2416, -0.0561, ..., -0.1154, -0.2520, -0.0873], + [ 0.2686, -0.0705, -0.2632, ..., -0.2009, -0.2052, -0.2186], + ..., + [-0.1027, -0.0283, -0.3250, ..., -0.1210, 0.1089, 0.0301], + [-0.1915, -0.1308, -0.0381, ..., -0.2076, -0.0426, -0.0462], + [-0.0800, -0.0954, -0.1355, ..., 0.0957, 0.0297, -0.1712]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0990e-07, ..., -1.6764e-08, + 1.8626e-09, 9.3132e-10], + [ 5.5879e-09, 9.3132e-10, 2.7940e-09, ..., 7.0781e-08, + 6.5193e-09, 0.0000e+00], + [-5.0291e-08, 2.7940e-09, 7.4506e-09, ..., 2.3283e-08, + 2.7940e-09, 0.0000e+00], + ..., + [ 5.9605e-08, 2.7940e-09, 4.6566e-09, ..., 3.1665e-08, + -6.5193e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.2107e-08, ..., -2.5146e-08, + 1.8626e-09, 1.8626e-09], + [ 1.8626e-09, 9.3132e-10, 2.4214e-08, ..., 5.0012e-07, + 1.4901e-08, 9.3132e-10]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0041, 0.0101, 0.0131, -0.0014, 0.0243, -0.0046, -0.0030, -0.0037, + 0.0102, -0.0169], device='cuda:0'), grad: tensor([-3.8184e-08, -6.2399e-08, -9.3132e-10, 6.0536e-08, -1.8794e-06, + -5.0291e-08, 2.0489e-07, 3.7160e-07, -3.0082e-07, 1.6931e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 246.46, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.5171 re_mapping 0.0033 re_causal 0.0112 /// teacc 99.11 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.2162, -0.0510, 0.2060, ..., 0.0431, -0.2049, -0.1584], + [-0.1625, -0.2427, -0.0562, ..., -0.1156, -0.2526, -0.0875], + [ 0.2696, -0.0706, -0.2637, ..., -0.2004, -0.2059, -0.2177], + ..., + [-0.1034, -0.0285, -0.3260, ..., -0.1210, 0.1090, 0.0285], + [-0.1917, -0.1315, -0.0385, ..., -0.2082, -0.0430, -0.0471], + [-0.0802, -0.0952, -0.1356, ..., 0.0957, 0.0297, -0.1713]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 2.7940e-09, -6.0722e-07, ..., -1.8533e-07, + 3.7253e-09, -8.1956e-08], + [ 8.9407e-08, 2.9802e-08, 2.7940e-09, ..., 1.4901e-08, + 1.8626e-08, 1.8626e-09], + [-5.5879e-09, 1.3970e-08, 1.8626e-09, ..., 3.7253e-09, + 9.3132e-10, -1.6764e-08], + ..., + [ 1.0896e-07, 3.6322e-08, 4.6566e-09, ..., 5.3085e-08, + -2.1420e-08, 2.7940e-09], + [ 2.9802e-08, 1.2107e-08, 9.3132e-09, ..., 1.6112e-07, + 2.6077e-08, 1.0245e-08], + [ 2.7008e-08, 6.5193e-09, 6.5193e-09, ..., -2.3283e-08, + -1.3039e-08, 1.8626e-09]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0040, 0.0101, 0.0130, -0.0015, 0.0244, -0.0046, -0.0030, -0.0038, + 0.0100, -0.0170], device='cuda:0'), grad: tensor([-9.9186e-07, 3.6974e-07, 1.3597e-07, -1.3346e-06, -5.4482e-07, + -1.8533e-07, 1.2293e-06, 5.6159e-07, 5.8487e-07, 1.6391e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 246.54, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.5079 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.12 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.2165, -0.0512, 0.2066, ..., 0.0432, -0.2052, -0.1585], + [-0.1626, -0.2434, -0.0563, ..., -0.1166, -0.2542, -0.0877], + [ 0.2710, -0.0704, -0.2645, ..., -0.2010, -0.2066, -0.2177], + ..., + [-0.1052, -0.0287, -0.3276, ..., -0.1216, 0.1090, 0.0286], + [-0.1919, -0.1315, -0.0389, ..., -0.2115, -0.0430, -0.0475], + [-0.0803, -0.0952, -0.1359, ..., 0.0960, 0.0298, -0.1713]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.7940e-09, -2.7008e-08, ..., -1.0245e-08, + 1.0245e-08, 4.6566e-09], + [ 4.6566e-09, 4.6566e-09, 5.5879e-09, ..., 4.3772e-08, + 1.9558e-08, 2.7940e-09], + [ 5.5879e-09, 9.3132e-09, 2.7940e-09, ..., 1.4901e-08, + 5.5879e-09, 9.3132e-10], + ..., + [ 1.7695e-08, 1.0245e-08, 1.8626e-09, ..., 3.0454e-07, + 1.4901e-07, 0.0000e+00], + [ 3.7253e-09, 3.7253e-09, 9.3132e-09, ..., 8.3819e-09, + 1.2107e-08, 5.5879e-09], + [ 1.0245e-08, 1.1176e-08, 1.1176e-08, ..., 1.8533e-07, + -2.0489e-08, 0.0000e+00]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0042, 0.0101, 0.0137, -0.0011, 0.0245, -0.0049, -0.0030, -0.0040, + 0.0085, -0.0167], device='cuda:0'), grad: tensor([ 1.4901e-08, 1.0431e-07, 1.2014e-07, -4.3027e-07, -2.4233e-06, + 4.2655e-07, 2.4773e-07, 1.1660e-06, -1.9744e-07, 9.7789e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 246.56, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4787 re_mapping 0.0035 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.2167, -0.0515, 0.2072, ..., 0.0414, -0.2055, -0.1586], + [-0.1627, -0.2445, -0.0565, ..., -0.1169, -0.2558, -0.0882], + [ 0.2713, -0.0705, -0.2654, ..., -0.2018, -0.2074, -0.2179], + ..., + [-0.1054, -0.0288, -0.3295, ..., -0.1216, 0.1092, 0.0292], + [-0.1921, -0.1318, -0.0397, ..., -0.2120, -0.0432, -0.0480], + [-0.0805, -0.0954, -0.1362, ..., 0.0964, 0.0297, -0.1713]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.9802e-08, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0028, 0.0102, 0.0135, -0.0011, 0.0245, -0.0050, -0.0030, -0.0040, + 0.0083, -0.0165], device='cuda:0'), grad: tensor([ 0.0000e+00, -8.1956e-08, 1.0245e-08, 4.6566e-09, 5.2154e-08, + -2.7940e-09, 6.5193e-09, 4.6566e-08, 1.1176e-08, -4.9360e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 246.64, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.5198 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.12 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.2167, -0.0518, 0.2075, ..., 0.0414, -0.2058, -0.1587], + [-0.1629, -0.2454, -0.0565, ..., -0.1170, -0.2567, -0.0883], + [ 0.2714, -0.0705, -0.2666, ..., -0.2026, -0.2086, -0.2187], + ..., + [-0.1054, -0.0294, -0.3340, ..., -0.1219, 0.1092, 0.0294], + [-0.1921, -0.1320, -0.0402, ..., -0.2123, -0.0434, -0.0481], + [-0.0806, -0.0955, -0.1364, ..., 0.0965, 0.0298, -0.1713]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4214e-08, ..., -1.2107e-08, + 2.7940e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.5832e-08, + 9.3132e-10, 3.7253e-09], + [ 9.3132e-10, 9.3132e-10, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 2.7940e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.5832e-08, + 5.5879e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -9.8720e-08, + -6.5193e-09, 2.7940e-09]], device='cuda:0') +Epoch 287, bias, value: tensor([ 0.0028, 0.0102, 0.0129, -0.0010, 0.0246, -0.0050, -0.0030, -0.0041, + 0.0086, -0.0164], device='cuda:0'), grad: tensor([-3.1665e-08, 1.9558e-08, 1.8626e-08, 2.0489e-08, 7.4506e-08, + -9.4995e-08, 1.1455e-07, 1.0617e-07, -4.8429e-08, -1.7323e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 246.47, cls_loss 0.0005 cls_loss_mapping 0.0024 cls_loss_causal 0.4721 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.03 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.2168, -0.0549, 0.2073, ..., 0.0411, -0.2061, -0.1588], + [-0.1634, -0.2461, -0.0567, ..., -0.1174, -0.2589, -0.0901], + [ 0.2720, -0.0706, -0.2673, ..., -0.2034, -0.2092, -0.2191], + ..., + [-0.1057, -0.0295, -0.3352, ..., -0.1217, 0.1099, 0.0321], + [-0.1928, -0.1323, -0.0407, ..., -0.2126, -0.0440, -0.0491], + [-0.0809, -0.0955, -0.1366, ..., 0.0964, 0.0293, -0.1716]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 9.3132e-10, -1.8533e-07, ..., -2.5425e-07, + 1.8626e-09, 1.9558e-08], + [ 5.5879e-09, 9.3132e-10, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [-2.8126e-07, 1.8626e-09, 9.3132e-10, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.6450e-07, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, + -2.7940e-09, 1.8626e-09], + [ 6.5193e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, 4.8429e-08], + [ 2.7940e-09, 0.0000e+00, 1.9651e-07, ..., 2.6356e-07, + 1.8626e-09, 1.8626e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0019, 0.0102, 0.0129, -0.0010, 0.0247, -0.0046, -0.0029, -0.0039, + 0.0080, -0.0167], device='cuda:0'), grad: tensor([-3.7532e-07, -2.7008e-08, -4.7125e-07, 2.9802e-08, -7.6368e-08, + -7.8883e-07, 1.9837e-07, 5.1130e-07, 3.6601e-07, 6.1747e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0023 cls_loss_causal 0.4984 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.11 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.2169, -0.0549, 0.2103, ..., 0.0439, -0.2062, -0.1585], + [-0.1641, -0.2465, -0.0592, ..., -0.1195, -0.2594, -0.0906], + [ 0.2733, -0.0707, -0.2680, ..., -0.2033, -0.2093, -0.2197], + ..., + [-0.1069, -0.0296, -0.3360, ..., -0.1219, 0.1100, 0.0320], + [-0.1929, -0.1324, -0.0407, ..., -0.2126, -0.0439, -0.0491], + [-0.0815, -0.0956, -0.1390, ..., 0.0952, 0.0293, -0.1716]], + device='cuda:0'), grad: tensor([[ 7.9162e-08, 1.8626e-09, -6.5193e-09, ..., 6.5193e-09, + 9.3132e-10, -1.8626e-09], + [ 3.0734e-08, 1.2107e-08, 4.6566e-09, ..., 5.5879e-09, + 1.8626e-09, 3.7253e-09], + [-4.6473e-07, 3.3528e-08, 0.0000e+00, ..., -7.8231e-08, + 9.3132e-10, 1.8626e-09], + ..., + [ 5.3085e-08, 2.3283e-08, 9.3132e-10, ..., 5.5879e-09, + -4.6566e-09, 9.3132e-09], + [ 3.4459e-08, 1.3970e-08, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, 2.7940e-09], + [ 1.3970e-08, 1.8626e-09, 1.8626e-09, ..., 1.7695e-08, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0049, 0.0098, 0.0135, -0.0013, 0.0247, -0.0043, -0.0032, -0.0039, + 0.0083, -0.0175], device='cuda:0'), grad: tensor([ 1.8254e-07, 1.2014e-07, -8.8569e-07, -1.9185e-07, -5.6811e-08, + 2.5798e-07, 1.9372e-07, 1.7229e-07, 9.9652e-08, 1.0431e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 246.98, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4949 re_mapping 0.0034 re_causal 0.0113 /// teacc 99.12 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.2172, -0.0551, 0.2105, ..., 0.0440, -0.2064, -0.1587], + [-0.1642, -0.2471, -0.0592, ..., -0.1196, -0.2600, -0.0907], + [ 0.2736, -0.0708, -0.2692, ..., -0.2032, -0.2097, -0.2199], + ..., + [-0.1071, -0.0297, -0.3368, ..., -0.1220, 0.1101, 0.0320], + [-0.1931, -0.1326, -0.0421, ..., -0.2127, -0.0439, -0.0490], + [-0.0819, -0.0956, -0.1392, ..., 0.0951, 0.0292, -0.1717]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 1.8626e-09, 9.3132e-09, ..., -0.0000e+00, + 2.9802e-08, 1.2107e-08], + [ 2.4214e-08, 1.8626e-09, 6.5193e-09, ..., 2.7008e-08, + 4.2003e-07, 5.5879e-09], + [-6.0350e-07, -5.7742e-08, 1.8626e-09, ..., 4.4703e-08, + 2.2352e-08, -3.2596e-08], + ..., + [ 3.7253e-08, 3.7253e-09, 9.3132e-10, ..., 1.8626e-09, + -6.7707e-07, 1.8626e-09], + [ 3.9488e-07, 4.0978e-08, 6.5193e-09, ..., 6.5193e-09, + 1.0245e-08, 2.7008e-08], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 4.8429e-08, + 1.4435e-07, 9.3132e-10]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0049, 0.0099, 0.0126, -0.0013, 0.0248, -0.0043, -0.0032, -0.0038, + 0.0089, -0.0176], device='cuda:0'), grad: tensor([ 2.2165e-07, 2.6785e-06, -1.9819e-06, 3.5949e-07, -1.3039e-07, + 2.3376e-07, -9.9652e-08, -3.9339e-06, 1.6056e-06, 1.0431e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 246.77, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4844 re_mapping 0.0032 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.2181, -0.0551, 0.2107, ..., 0.0440, -0.2068, -0.1590], + [-0.1641, -0.2476, -0.0601, ..., -0.1200, -0.2613, -0.0910], + [ 0.2740, -0.0708, -0.2705, ..., -0.2038, -0.2103, -0.2202], + ..., + [-0.1077, -0.0298, -0.3373, ..., -0.1221, 0.1102, 0.0320], + [-0.1933, -0.1328, -0.0423, ..., -0.2127, -0.0439, -0.0491], + [-0.0822, -0.0957, -0.1392, ..., 0.0950, 0.0292, -0.1717]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, -8.6725e-06, ..., -5.7369e-06, + 2.7940e-09, 2.7940e-09], + [ 1.3970e-08, 1.8626e-09, 3.2857e-06, ..., 2.1849e-06, + 5.5879e-08, 9.3132e-10], + [-1.6298e-07, 9.3132e-10, 4.3772e-08, ..., 8.2888e-08, + 6.5193e-09, 0.0000e+00], + ..., + [ 9.6858e-08, 5.5879e-09, 9.3132e-09, ..., 4.7497e-08, + -6.5193e-08, -1.8626e-09], + [ 5.5879e-09, 9.3132e-10, 7.4506e-09, ..., 8.3819e-09, + 9.3132e-10, 1.8626e-09], + [ 1.8626e-09, 9.3132e-10, 4.3772e-08, ..., 1.1548e-07, + 1.2107e-08, -9.3132e-10]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0049, 0.0099, 0.0123, -0.0014, 0.0249, -0.0043, -0.0032, -0.0038, + 0.0092, -0.0177], device='cuda:0'), grad: tensor([-2.1994e-05, 7.0147e-06, -1.7695e-08, 3.1665e-08, -1.6689e-06, + 1.2163e-06, 1.3471e-05, 1.4072e-06, 1.7788e-07, 3.8743e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 246.49, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4912 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.2179, -0.0554, 0.2113, ..., 0.0441, -0.2073, -0.1593], + [-0.1647, -0.2482, -0.0604, ..., -0.1203, -0.2621, -0.0907], + [ 0.2749, -0.0710, -0.2725, ..., -0.2048, -0.2110, -0.2210], + ..., + [-0.1084, -0.0300, -0.3386, ..., -0.1221, 0.1104, 0.0319], + [-0.1935, -0.1330, -0.0427, ..., -0.2130, -0.0438, -0.0493], + [-0.0829, -0.0957, -0.1396, ..., 0.0950, 0.0291, -0.1717]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 1.3411e-07, + 1.4901e-08, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.4808e-07, + 1.5553e-07, 0.0000e+00], + [-5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 9.7789e-08, + 1.2107e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., -1.9595e-06, + -3.5074e-06, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + 5.5879e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 1.1018e-06, + 3.2466e-06, 0.0000e+00]], device='cuda:0') +Epoch 292, bias, value: tensor([ 0.0050, 0.0099, 0.0125, -0.0014, 0.0250, -0.0042, -0.0033, -0.0039, + 0.0093, -0.0178], device='cuda:0'), grad: tensor([ 4.3027e-07, 1.0319e-06, 3.0175e-07, 3.2224e-07, 6.3051e-07, + 4.5262e-07, 8.4750e-08, -1.9625e-05, 9.5926e-08, 1.6257e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 246.99, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4834 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.10 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.2187, -0.0559, 0.2119, ..., 0.0442, -0.2077, -0.1594], + [-0.1649, -0.2492, -0.0607, ..., -0.1205, -0.2641, -0.0908], + [ 0.2752, -0.0711, -0.2740, ..., -0.2063, -0.2119, -0.2211], + ..., + [-0.1088, -0.0302, -0.3403, ..., -0.1239, 0.1089, 0.0319], + [-0.1942, -0.1333, -0.0431, ..., -0.2137, -0.0441, -0.0495], + [-0.0832, -0.0958, -0.1396, ..., 0.0953, 0.0306, -0.1717]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1921e-07, ..., -5.4948e-08, + 1.3970e-08, -4.6566e-09], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., -0.0000e+00, + 3.7253e-09, 9.3132e-10], + [-2.7940e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + ..., + [ 2.7940e-09, 9.3132e-10, 9.3132e-10, ..., 1.8626e-09, + -4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + -1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -1.8626e-08, + -3.7253e-09, 9.3132e-10]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0050, 0.0102, 0.0121, -0.0014, 0.0251, -0.0040, -0.0035, -0.0047, + 0.0087, -0.0174], device='cuda:0'), grad: tensor([-1.4994e-07, 4.6566e-09, 1.0245e-08, -4.0047e-08, 4.7497e-08, + 4.4331e-07, -2.6729e-07, 7.4506e-09, -4.5635e-08, -4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4715 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.12 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.2185, -0.0561, 0.2122, ..., 0.0442, -0.2079, -0.1595], + [-0.1650, -0.2502, -0.0609, ..., -0.1207, -0.2653, -0.0909], + [ 0.2758, -0.0711, -0.2749, ..., -0.2072, -0.2126, -0.2212], + ..., + [-0.1097, -0.0303, -0.3409, ..., -0.1227, 0.1100, 0.0320], + [-0.1946, -0.1337, -0.0435, ..., -0.2138, -0.0444, -0.0498], + [-0.0838, -0.0958, -0.1398, ..., 0.0949, 0.0296, -0.1717]], + device='cuda:0'), grad: tensor([[-3.5390e-08, 0.0000e+00, -1.1828e-07, ..., -7.0781e-08, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 1.2107e-08, + 7.4506e-09, 9.3132e-10], + [ 3.7253e-09, 0.0000e+00, 1.2107e-08, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 1.1176e-08, ..., 3.5390e-08, + -1.3039e-08, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 1.2107e-08, ..., 4.6566e-09, + 2.7940e-09, 2.7940e-09], + [ 1.4901e-08, 0.0000e+00, 5.1223e-08, ..., -5.0291e-08, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0050, 0.0103, 0.0120, -0.0014, 0.0252, -0.0039, -0.0034, -0.0042, + 0.0088, -0.0181], device='cuda:0'), grad: tensor([-3.2596e-07, -5.5879e-09, 9.4064e-08, 6.9477e-07, 1.4622e-07, + -7.0781e-07, 1.4435e-07, 1.5087e-07, -2.5705e-07, 6.1467e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 246.84, cls_loss 0.0005 cls_loss_mapping 0.0020 cls_loss_causal 0.5207 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.14 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.2187, -0.0561, 0.2102, ..., 0.0443, -0.2101, -0.1606], + [-0.1652, -0.2507, -0.0610, ..., -0.1209, -0.2678, -0.0910], + [ 0.2760, -0.0713, -0.2765, ..., -0.2079, -0.2136, -0.2212], + ..., + [-0.1096, -0.0304, -0.3428, ..., -0.1230, 0.1101, 0.0320], + [-0.1953, -0.1343, -0.0442, ..., -0.2143, -0.0445, -0.0501], + [-0.0839, -0.0955, -0.1399, ..., 0.0949, 0.0295, -0.1717]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 9.3132e-10, -1.0803e-07, ..., -2.7940e-08, + 0.0000e+00, 9.3132e-10], + [ 6.8918e-08, 1.8626e-09, 5.5879e-09, ..., -1.8626e-09, + 2.7940e-09, 1.8626e-09], + [-6.7893e-07, 1.6764e-08, 1.3970e-08, ..., 3.7253e-09, + 0.0000e+00, 3.7253e-09], + ..., + [ 4.3679e-07, 1.4901e-08, 6.5193e-09, ..., 2.7940e-09, + -6.5193e-09, 3.7253e-09], + [ 1.5832e-07, 2.7940e-09, 1.6764e-08, ..., 3.7253e-09, + 1.8626e-09, 1.8626e-09], + [ 2.7940e-08, 3.7253e-09, 3.4459e-08, ..., 1.9558e-08, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0037, 0.0102, 0.0114, -0.0015, 0.0253, -0.0038, -0.0023, -0.0041, + 0.0084, -0.0181], device='cuda:0'), grad: tensor([-6.9849e-08, -8.1584e-07, -1.8450e-06, -2.8592e-07, 2.4214e-08, + -4.7497e-08, 1.2852e-07, 1.3448e-06, 1.2908e-06, 2.7288e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 246.55, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4760 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.13 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.2190, -0.0562, 0.2100, ..., 0.0442, -0.2109, -0.1610], + [-0.1654, -0.2515, -0.0612, ..., -0.1239, -0.2690, -0.0912], + [ 0.2767, -0.0714, -0.2772, ..., -0.2087, -0.2142, -0.2213], + ..., + [-0.1100, -0.0306, -0.3438, ..., -0.1219, 0.1123, 0.0319], + [-0.1961, -0.1346, -0.0461, ..., -0.2157, -0.0451, -0.0511], + [-0.0842, -0.0956, -0.1396, ..., 0.0953, 0.0275, -0.1717]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, 0.0000e+00, ..., 3.6322e-08, + 4.6566e-09, 0.0000e+00], + [ 1.3411e-07, 0.0000e+00, 0.0000e+00, ..., 5.8673e-08, + 5.6811e-08, 0.0000e+00], + [ 6.5006e-07, 0.0000e+00, 0.0000e+00, ..., 3.9674e-07, + 3.9488e-07, 0.0000e+00], + ..., + [-1.0543e-06, 0.0000e+00, 0.0000e+00, ..., -5.3924e-07, + -5.7463e-07, 0.0000e+00], + [ 1.3039e-08, 0.0000e+00, 0.0000e+00, ..., 4.4703e-08, + 4.6566e-09, 0.0000e+00], + [ 6.6124e-08, 0.0000e+00, 9.3132e-10, ..., -1.4901e-07, + 2.8871e-08, 0.0000e+00]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0033, 0.0112, 0.0116, -0.0014, 0.0251, -0.0027, -0.0035, -0.0043, + 0.0074, -0.0186], device='cuda:0'), grad: tensor([ 1.3504e-07, 4.5076e-07, 4.2357e-06, 7.3854e-07, 2.9523e-07, + -6.8918e-08, 1.0524e-07, -6.0797e-06, 1.4249e-07, 3.5390e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 246.73, cls_loss 0.0006 cls_loss_mapping 0.0028 cls_loss_causal 0.4764 re_mapping 0.0033 re_causal 0.0102 /// teacc 99.20 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.2187, -0.0562, 0.2105, ..., 0.0439, -0.2111, -0.1612], + [-0.1656, -0.2520, -0.0612, ..., -0.1238, -0.2697, -0.0914], + [ 0.2770, -0.0715, -0.2783, ..., -0.2103, -0.2152, -0.2217], + ..., + [-0.1101, -0.0308, -0.3448, ..., -0.1222, 0.1123, 0.0317], + [-0.1963, -0.1348, -0.0465, ..., -0.2153, -0.0433, -0.0512], + [-0.0849, -0.0956, -0.1403, ..., 0.0956, 0.0277, -0.1718]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0896e-07, ..., -2.5146e-08, + 2.7940e-09, 7.4506e-09], + [ 1.0245e-08, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 4.6566e-09], + [-4.0978e-08, -1.8626e-09, 2.7940e-09, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + ..., + [ 8.3819e-09, 0.0000e+00, 4.6566e-09, ..., 2.9802e-08, + 3.7253e-09, 4.6566e-09], + [ 1.3970e-08, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.4064e-08, 1.5926e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -8.8476e-08, + 9.3132e-10, 1.1176e-08]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0031, 0.0116, 0.0113, -0.0015, 0.0251, -0.0026, -0.0039, -0.0046, + 0.0086, -0.0185], device='cuda:0'), grad: tensor([-1.7788e-07, -5.8673e-08, -7.9162e-08, 4.6566e-08, 1.7416e-07, + -2.2277e-06, 1.7285e-06, 1.9185e-07, 5.8580e-07, -1.7788e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4677 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.17 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.2182, -0.0562, 0.2118, ..., 0.0444, -0.2112, -0.1612], + [-0.1664, -0.2524, -0.0641, ..., -0.1252, -0.2705, -0.0915], + [ 0.2779, -0.0715, -0.2797, ..., -0.2123, -0.2158, -0.2223], + ..., + [-0.1105, -0.0309, -0.3465, ..., -0.1222, 0.1123, 0.0317], + [-0.1965, -0.1350, -0.0456, ..., -0.2154, -0.0427, -0.0503], + [-0.0854, -0.0957, -0.1406, ..., 0.0956, 0.0277, -0.1718]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 2.3283e-08, ..., 3.2596e-09, + 1.6764e-08, 9.3132e-09], + [ 2.0955e-08, 0.0000e+00, 1.7695e-08, ..., -1.8626e-09, + 2.3749e-08, 8.8476e-09], + [-3.9581e-08, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 4.1910e-09, 1.3970e-09], + ..., + [ 6.9849e-09, 4.6566e-10, 0.0000e+00, ..., 2.7474e-08, + -3.6322e-08, 0.0000e+00], + [ 5.1223e-09, 4.6566e-10, 2.0489e-08, ..., 2.7940e-09, + 1.6764e-08, 1.0245e-08], + [ 1.3970e-09, 0.0000e+00, 1.3970e-09, ..., -1.4435e-08, + 1.3970e-08, 0.0000e+00]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0041, 0.0113, 0.0118, -0.0015, 0.0251, -0.0026, -0.0045, -0.0046, + 0.0102, -0.0185], device='cuda:0'), grad: tensor([ 1.3551e-07, -1.7928e-07, -2.9337e-08, -3.4925e-08, 1.1642e-08, + 1.0338e-07, -3.6228e-07, 5.9139e-08, 1.5879e-07, 1.3504e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 246.73, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4799 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.12 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.2188, -0.0564, 0.2119, ..., 0.0445, -0.2115, -0.1614], + [-0.1668, -0.2528, -0.0641, ..., -0.1254, -0.2714, -0.0916], + [ 0.2787, -0.0716, -0.2812, ..., -0.2136, -0.2168, -0.2231], + ..., + [-0.1108, -0.0309, -0.3472, ..., -0.1223, 0.1125, 0.0317], + [-0.1969, -0.1352, -0.0465, ..., -0.2155, -0.0428, -0.0511], + [-0.0868, -0.0957, -0.1410, ..., 0.0956, 0.0277, -0.1718]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 2.7940e-09, 1.5367e-08, ..., 8.3819e-09, + 1.7229e-08, 4.6566e-09], + [ 9.3132e-09, 4.6566e-10, 7.9162e-09, ..., 1.3039e-08, + 5.2620e-08, 2.3283e-09], + [-4.8056e-07, 4.6566e-10, 4.1910e-09, ..., 4.1910e-09, + 1.7695e-08, 1.3970e-09], + ..., + [ 4.3027e-07, 9.3132e-10, 0.0000e+00, ..., 1.1642e-08, + -1.1176e-07, 0.0000e+00], + [ 3.3062e-08, 3.2596e-09, 8.3353e-08, ..., 1.0710e-08, + 1.2107e-08, 2.5611e-08], + [ 8.3819e-09, 6.5193e-09, 5.1223e-09, ..., -9.2201e-08, + 1.4901e-08, 1.3970e-09]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0041, 0.0112, 0.0116, -0.0016, 0.0252, -0.0028, -0.0043, -0.0045, + 0.0103, -0.0185], device='cuda:0'), grad: tensor([ 1.9372e-07, 3.6741e-07, -1.0096e-06, 4.0280e-07, 1.2759e-07, + -5.1036e-07, -1.4808e-07, 4.3353e-07, -1.4389e-07, 3.1106e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 246.94, cls_loss 0.0005 cls_loss_mapping 0.0021 cls_loss_causal 0.4658 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.2191, -0.0565, 0.2120, ..., 0.0445, -0.2119, -0.1616], + [-0.1672, -0.2538, -0.0645, ..., -0.1254, -0.2730, -0.0928], + [ 0.2798, -0.0717, -0.2819, ..., -0.2151, -0.2173, -0.2232], + ..., + [-0.1118, -0.0311, -0.3478, ..., -0.1225, 0.1125, 0.0316], + [-0.1973, -0.1358, -0.0465, ..., -0.2156, -0.0428, -0.0515], + [-0.0876, -0.0958, -0.1411, ..., 0.0957, 0.0278, -0.1718]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 7.9768e-07, ..., 4.7497e-08, + 1.5832e-07, 1.4715e-07], + [ 3.2596e-09, 0.0000e+00, 4.1910e-09, ..., 7.4971e-08, + 3.5390e-08, 9.3132e-10], + [-9.7789e-09, 0.0000e+00, 2.7940e-09, ..., 2.3283e-09, + 1.8626e-09, 4.6566e-10], + ..., + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., -5.6531e-07, + -4.8475e-07, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 4.6566e-09, ..., 5.1223e-09, + 6.5193e-09, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 3.2596e-09, ..., 6.1328e-07, + 4.4284e-07, 4.6566e-10]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0041, 0.0112, 0.0120, -0.0015, 0.0251, -0.0023, -0.0047, -0.0047, + 0.0107, -0.0185], device='cuda:0'), grad: tensor([ 1.4342e-06, 4.4797e-07, 1.1176e-08, 8.9407e-08, -3.1525e-07, + -1.3830e-07, -1.4417e-06, -3.1702e-06, 1.1595e-07, 2.9784e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 246.64, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4549 re_mapping 0.0034 re_causal 0.0103 /// teacc 99.13 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.2194, -0.0565, 0.2120, ..., 0.0442, -0.2123, -0.1619], + [-0.1684, -0.2543, -0.0647, ..., -0.1257, -0.2740, -0.0929], + [ 0.2814, -0.0717, -0.2826, ..., -0.2158, -0.2182, -0.2233], + ..., + [-0.1121, -0.0314, -0.3482, ..., -0.1226, 0.1126, 0.0316], + [-0.1980, -0.1359, -0.0468, ..., -0.2159, -0.0432, -0.0521], + [-0.0887, -0.0958, -0.1412, ..., 0.0959, 0.0278, -0.1718]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 0.0000e+00, 8.3819e-09, ..., 3.2596e-09, + 4.6566e-09, 1.6764e-08], + [ 6.7987e-08, 0.0000e+00, 2.7940e-09, ..., 1.3504e-08, + 1.2899e-07, 4.1910e-09], + [ 2.1886e-08, 0.0000e+00, 2.0023e-08, ..., 8.8476e-09, + 6.6124e-08, 3.8184e-08], + ..., + [-1.2433e-07, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + -2.5565e-07, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 6.0536e-09, ..., 4.1910e-09, + 1.4435e-08, 4.1910e-09], + [ 3.8184e-08, 0.0000e+00, 3.2596e-09, ..., 1.9372e-07, + 2.1886e-08, 1.8626e-09]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0037, 0.0110, 0.0118, -0.0016, 0.0253, -0.0024, -0.0047, -0.0045, + 0.0107, -0.0184], device='cuda:0'), grad: tensor([ 1.9977e-07, 1.0431e-06, 5.6392e-07, 1.5218e-06, -3.2037e-07, + -1.3150e-06, -7.3807e-07, -1.8440e-06, -9.3132e-09, 9.1456e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 246.40, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4682 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.13 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.2203, -0.0566, 0.2121, ..., 0.0442, -0.2127, -0.1621], + [-0.1686, -0.2548, -0.0647, ..., -0.1258, -0.2752, -0.0930], + [ 0.2820, -0.0718, -0.2835, ..., -0.2169, -0.2194, -0.2217], + ..., + [-0.1126, -0.0316, -0.3489, ..., -0.1228, 0.1126, 0.0314], + [-0.1983, -0.1361, -0.0475, ..., -0.2160, -0.0436, -0.0526], + [-0.0895, -0.0961, -0.1414, ..., 0.0959, 0.0278, -0.1718]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 4.6566e-09, + 1.2107e-08, 4.1910e-09], + [ 4.1910e-09, 0.0000e+00, 3.7253e-09, ..., 2.5611e-08, + 4.1910e-09, 2.3283e-09], + [-1.5041e-07, 0.0000e+00, 2.7940e-09, ..., 2.3749e-08, + 1.3970e-09, 1.8626e-09], + ..., + [ 1.3737e-07, 4.6566e-10, 0.0000e+00, ..., 5.2154e-08, + 8.3819e-09, 9.3132e-10], + [ 9.3132e-09, 0.0000e+00, 8.3819e-09, ..., 3.2131e-08, + 9.3132e-09, 2.3283e-09], + [ 9.3132e-10, 0.0000e+00, 6.9849e-09, ..., -3.3528e-08, + -3.7253e-09, 1.3970e-09]], device='cuda:0') +Epoch 302, bias, value: tensor([ 0.0036, 0.0114, 0.0105, -0.0015, 0.0254, -0.0031, -0.0037, -0.0047, + 0.0108, -0.0185], device='cuda:0'), grad: tensor([ 7.9628e-08, 8.5682e-08, -1.4761e-07, 2.9802e-08, -4.8382e-07, + 2.6356e-07, -2.1607e-07, 4.0233e-07, 2.9802e-08, -2.3749e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 246.42, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4819 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.07 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.2203, -0.0567, 0.2123, ..., 0.0443, -0.2131, -0.1622], + [-0.1689, -0.2558, -0.0648, ..., -0.1257, -0.2757, -0.0931], + [ 0.2824, -0.0720, -0.2839, ..., -0.2183, -0.2198, -0.2216], + ..., + [-0.1129, -0.0317, -0.3490, ..., -0.1229, 0.1127, 0.0315], + [-0.1988, -0.1365, -0.0479, ..., -0.2161, -0.0440, -0.0528], + [-0.0900, -0.0960, -0.1415, ..., 0.0958, 0.0278, -0.1718]], + device='cuda:0'), grad: tensor([[ 4.1444e-08, 1.3970e-09, -2.1607e-07, ..., -1.5041e-07, + 4.6566e-09, 1.3970e-09], + [ 2.4680e-08, 1.8626e-09, 3.7253e-09, ..., 2.7940e-09, + 2.8452e-07, 4.6566e-10], + [-2.7614e-07, -2.7940e-09, 1.3970e-09, ..., 1.3970e-09, + 9.3132e-10, 4.6566e-10], + ..., + [ 2.6543e-08, 4.6566e-09, 2.3283e-09, ..., 4.1910e-09, + -3.8883e-07, 4.6566e-10], + [ 1.9697e-07, 8.3819e-09, 3.2596e-09, ..., 1.3970e-09, + 2.3283e-09, 9.3132e-10], + [ 2.7940e-09, 4.6566e-10, 1.8114e-07, ..., 1.2061e-07, + 7.9628e-08, 0.0000e+00]], device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0036, 0.0129, 0.0104, -0.0013, 0.0255, -0.0033, -0.0035, -0.0061, + 0.0106, -0.0186], device='cuda:0'), grad: tensor([-1.2200e-07, 5.9158e-06, 8.4285e-08, 8.1491e-08, 1.0291e-07, + 5.8673e-08, 6.7055e-08, -1.0394e-05, 5.4250e-07, 3.6657e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 246.55, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4950 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.06 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.2208, -0.0567, 0.2123, ..., 0.0443, -0.2134, -0.1624], + [-0.1695, -0.2563, -0.0649, ..., -0.1258, -0.2780, -0.0935], + [ 0.2836, -0.0720, -0.2852, ..., -0.2194, -0.2209, -0.2217], + ..., + [-0.1131, -0.0317, -0.3493, ..., -0.1229, 0.1129, 0.0318], + [-0.2006, -0.1368, -0.0485, ..., -0.2163, -0.0444, -0.0532], + [-0.0903, -0.0961, -0.1415, ..., 0.0957, 0.0277, -0.1719]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + -5.5879e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0675e-07, + 7.8231e-08, 0.0000e+00]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0036, 0.0128, 0.0109, 0.0004, 0.0256, -0.0047, -0.0035, -0.0059, + 0.0101, -0.0188], device='cuda:0'), grad: tensor([ 2.7940e-09, 4.0978e-08, 5.5879e-09, 5.5879e-09, -4.6380e-07, + 1.8626e-09, 8.3819e-09, -7.4506e-09, 1.4901e-08, 3.9767e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 246.36, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4523 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.17 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.2201, -0.0567, 0.2125, ..., 0.0444, -0.2137, -0.1625], + [-0.1704, -0.2566, -0.0649, ..., -0.1258, -0.2789, -0.0936], + [ 0.2844, -0.0721, -0.2865, ..., -0.2201, -0.2216, -0.2217], + ..., + [-0.1133, -0.0317, -0.3495, ..., -0.1236, 0.1125, 0.0318], + [-0.2008, -0.1369, -0.0489, ..., -0.2164, -0.0446, -0.0537], + [-0.0909, -0.0961, -0.1418, ..., 0.0959, 0.0281, -0.1719]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 2.4214e-08, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4904e-06, + 5.8301e-07, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 7.4506e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-06, + 6.4727e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., 1.3039e-08, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -6.0089e-06, + -1.4007e-06, 9.3132e-10]], device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0036, 0.0105, 0.0112, 0.0006, 0.0256, -0.0046, -0.0037, -0.0039, + 0.0097, -0.0186], device='cuda:0'), grad: tensor([ 1.3225e-07, 1.0341e-05, 1.0338e-07, 7.7300e-08, 2.8554e-06, + 9.3132e-09, 5.2154e-08, 1.2375e-05, -3.9116e-08, -2.5883e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 304---------------------------------------------------- +epoch 304, time 262.99, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4683 re_mapping 0.0033 re_causal 0.0102 /// teacc 99.24 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.2217, -0.0567, 0.2125, ..., 0.0443, -0.2139, -0.1626], + [-0.1712, -0.2571, -0.0650, ..., -0.1261, -0.2808, -0.0938], + [ 0.2887, -0.0724, -0.2866, ..., -0.2204, -0.2230, -0.2217], + ..., + [-0.1176, -0.0319, -0.3498, ..., -0.1237, 0.1126, 0.0317], + [-0.2020, -0.1372, -0.0493, ..., -0.2167, -0.0449, -0.0539], + [-0.0913, -0.0962, -0.1419, ..., 0.0959, 0.0281, -0.1719]], + device='cuda:0'), grad: tensor([[ 9.8720e-08, 0.0000e+00, 4.0047e-08, ..., -9.3132e-10, + 1.7695e-08, 4.6566e-09], + [ 2.7940e-08, 0.0000e+00, 9.3132e-10, ..., -0.0000e+00, + 9.3132e-10, 9.3132e-10], + [-1.5814e-06, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 1.2387e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.5146e-08, 0.0000e+00, -8.3819e-09, ..., 0.0000e+00, + -1.0245e-08, 2.7940e-09], + [ 1.9558e-08, 0.0000e+00, 8.3819e-09, ..., 1.2107e-08, + 7.4506e-09, 9.3132e-10]], device='cuda:0') +Epoch 306, bias, value: tensor([ 0.0034, 0.0106, 0.0138, 0.0006, 0.0258, -0.0047, -0.0036, -0.0058, + 0.0095, -0.0187], device='cuda:0'), grad: tensor([ 3.7160e-07, 6.9849e-08, -3.4608e-06, 2.3153e-06, 1.9558e-08, + 2.5611e-07, -1.9465e-07, 5.4203e-07, -1.0524e-07, 1.6857e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 246.41, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4697 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.16 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.2226, -0.0567, 0.2129, ..., 0.0443, -0.2142, -0.1626], + [-0.1723, -0.2581, -0.0650, ..., -0.1263, -0.2842, -0.0939], + [ 0.2891, -0.0739, -0.2875, ..., -0.2208, -0.2250, -0.2210], + ..., + [-0.1178, -0.0320, -0.3513, ..., -0.1238, 0.1130, 0.0317], + [-0.2038, -0.1381, -0.0498, ..., -0.2171, -0.0454, -0.0544], + [-0.0918, -0.0960, -0.1421, ..., 0.0959, 0.0279, -0.1719]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, -1.4901e-08, ..., -4.6566e-09, + 1.0245e-08, 9.3132e-10], + [ 9.2201e-08, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 1.0710e-07, 0.0000e+00], + [-1.5274e-07, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + ..., + [ 2.5146e-08, 2.7940e-09, 0.0000e+00, ..., 3.7253e-09, + -2.0117e-07, 0.0000e+00], + [ 1.3970e-08, 9.3132e-10, 5.5879e-09, ..., 1.8626e-09, + 1.4901e-08, 4.6566e-09], + [ 5.5879e-09, 0.0000e+00, 5.5879e-09, ..., -7.4506e-09, + 4.1910e-08, 0.0000e+00]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0035, 0.0107, 0.0140, 0.0006, 0.0260, -0.0047, -0.0037, -0.0059, + 0.0089, -0.0188], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.0878e-06, -2.4028e-07, 2.6077e-08, 8.7544e-08, + -1.2480e-07, 1.0058e-07, -1.5385e-06, 1.5087e-07, 3.8557e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 246.43, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4284 re_mapping 0.0033 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.2229, -0.0568, 0.2132, ..., 0.0443, -0.2146, -0.1628], + [-0.1729, -0.2584, -0.0651, ..., -0.1268, -0.2895, -0.0941], + [ 0.2892, -0.0741, -0.2879, ..., -0.2211, -0.2271, -0.2191], + ..., + [-0.1178, -0.0322, -0.3521, ..., -0.1248, 0.1128, 0.0317], + [-0.2041, -0.1383, -0.0502, ..., -0.2173, -0.0461, -0.0549], + [-0.0926, -0.0961, -0.1425, ..., 0.0961, 0.0283, -0.1719]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.1176e-08, 3.7253e-09, ..., 3.7998e-07, + 4.6566e-09, 2.7940e-09], + [ 2.3283e-08, 3.2596e-08, 9.3132e-10, ..., 4.4703e-08, + 2.7008e-08, 1.8626e-09], + [ 7.0781e-08, 9.6764e-07, 0.0000e+00, ..., 2.7008e-08, + 2.2352e-08, 9.3132e-10], + ..., + [ 4.0047e-08, 1.0990e-07, 0.0000e+00, ..., 5.7742e-08, + -4.8429e-08, 9.3132e-10], + [ 2.7940e-09, 4.1910e-08, 1.8626e-09, ..., 2.2817e-07, + 1.4249e-07, 8.1025e-08], + [ 1.8626e-09, -3.4645e-07, 0.0000e+00, ..., -3.7253e-06, + -1.2573e-07, 3.7253e-09]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0037, 0.0106, 0.0140, 0.0003, 0.0261, -0.0045, -0.0036, -0.0060, + 0.0087, -0.0184], device='cuda:0'), grad: tensor([ 1.3718e-06, 3.8836e-07, 4.4070e-06, -2.1178e-06, 5.1558e-06, + -1.1576e-06, 6.5938e-07, 5.2247e-07, 1.3616e-06, -1.0580e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 246.12, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4796 re_mapping 0.0033 re_causal 0.0099 /// teacc 98.97 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.2233, -0.0568, 0.2135, ..., 0.0440, -0.2149, -0.1629], + [-0.1732, -0.2590, -0.0651, ..., -0.1270, -0.2943, -0.0941], + [ 0.2893, -0.0748, -0.2885, ..., -0.2215, -0.2300, -0.2189], + ..., + [-0.1179, -0.0323, -0.3528, ..., -0.1251, 0.1132, 0.0318], + [-0.2044, -0.1385, -0.0507, ..., -0.2182, -0.0466, -0.0553], + [-0.0927, -0.0959, -0.1428, ..., 0.0958, 0.0283, -0.1720]], + device='cuda:0'), grad: tensor([[-2.6077e-08, 0.0000e+00, 9.9652e-08, ..., 7.4785e-07, + 1.0245e-08, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 3.2596e-08, + 1.0245e-08, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 1.2107e-08, ..., 4.6566e-09, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.7940e-09, 0.0000e+00, 8.3819e-09, ..., 5.4948e-08, + 1.3039e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 2.3283e-08, + 4.6566e-09, 9.3132e-10], + [ 1.0245e-08, 0.0000e+00, -1.4063e-07, ..., -2.3916e-06, + -2.4401e-07, 0.0000e+00]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0034, 0.0105, 0.0140, 0.0001, 0.0269, -0.0045, -0.0036, -0.0060, + 0.0079, -0.0188], device='cuda:0'), grad: tensor([ 1.1083e-06, 8.1956e-08, 4.2841e-08, 3.9116e-08, 3.5688e-06, + 2.1048e-07, -1.5181e-07, 2.3097e-07, 8.1025e-08, -5.2229e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 246.61, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4903 re_mapping 0.0032 re_causal 0.0102 /// teacc 99.15 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.2226, -0.0568, 0.2150, ..., 0.0443, -0.2152, -0.1628], + [-0.1739, -0.2590, -0.0662, ..., -0.1272, -0.2954, -0.0943], + [ 0.2893, -0.0748, -0.2900, ..., -0.2223, -0.2316, -0.2184], + ..., + [-0.1179, -0.0323, -0.3545, ..., -0.1252, 0.1136, 0.0318], + [-0.2045, -0.1386, -0.0516, ..., -0.2185, -0.0472, -0.0565], + [-0.0933, -0.0960, -0.1439, ..., 0.0970, 0.0283, -0.1720]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., -1.7695e-08, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 9.3132e-09, + 2.2352e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, + -5.9605e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -9.3132e-09, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 310, bias, value: tensor([ 4.3602e-03, 1.0489e-02, 1.3952e-02, -5.2807e-05, 2.5495e-02, + -3.7253e-03, -4.8555e-03, -5.8842e-03, 7.9361e-03, -1.7865e-02], + device='cuda:0'), grad: tensor([-2.5146e-08, -4.4983e-07, 8.4750e-08, 8.0094e-08, 4.1910e-08, + -2.7940e-08, 3.8184e-08, -3.7253e-09, 3.8184e-08, 2.4680e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 246.43, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4725 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.18 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.2226, -0.0568, 0.2153, ..., 0.0443, -0.2157, -0.1628], + [-0.1744, -0.2591, -0.0663, ..., -0.1273, -0.2971, -0.0943], + [ 0.2893, -0.0748, -0.2907, ..., -0.2233, -0.2331, -0.2186], + ..., + [-0.1178, -0.0323, -0.3556, ..., -0.1252, 0.1143, 0.0317], + [-0.2050, -0.1387, -0.0527, ..., -0.2187, -0.0480, -0.0567], + [-0.0937, -0.0959, -0.1443, ..., 0.0970, 0.0279, -0.1720]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 0.0000e+00, -4.4703e-08, ..., -1.5832e-08, + 9.3132e-10, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 5.5879e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -2.0489e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -0.0000e+00, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.2107e-08, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0044, 0.0104, 0.0139, -0.0002, 0.0256, -0.0037, -0.0050, -0.0057, + 0.0077, -0.0180], device='cuda:0'), grad: tensor([-8.6613e-08, -2.6077e-08, 1.9558e-08, -1.2107e-08, 7.0781e-08, + 3.9116e-08, 2.4214e-08, -8.6613e-08, 9.3132e-10, 5.0291e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 246.43, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4976 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.14 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.2237, -0.0569, 0.2153, ..., 0.0444, -0.2171, -0.1647], + [-0.1747, -0.2596, -0.0664, ..., -0.1274, -0.2982, -0.0944], + [ 0.2894, -0.0748, -0.2931, ..., -0.2238, -0.2347, -0.2186], + ..., + [-0.1179, -0.0327, -0.3583, ..., -0.1253, 0.1144, 0.0317], + [-0.2067, -0.1394, -0.0537, ..., -0.2189, -0.0492, -0.0573], + [-0.0960, -0.0973, -0.1453, ..., 0.0980, 0.0292, -0.1720]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 3.0734e-08, 9.3132e-10], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -3.6322e-08, 0.0000e+00], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -6.5193e-09, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0043, 0.0104, 0.0139, -0.0002, 0.0246, -0.0034, -0.0047, -0.0057, + 0.0062, -0.0173], device='cuda:0'), grad: tensor([ 1.0245e-08, 2.0210e-07, 1.0058e-07, 2.1420e-08, 9.3132e-09, + -2.2631e-07, 2.0303e-07, -2.1979e-07, -9.4064e-08, -2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 246.64, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.5101 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.20 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.2242, -0.0570, 0.2158, ..., 0.0444, -0.2173, -0.1647], + [-0.1754, -0.2601, -0.0671, ..., -0.1277, -0.2990, -0.0946], + [ 0.2894, -0.0749, -0.2937, ..., -0.2242, -0.2367, -0.2187], + ..., + [-0.1178, -0.0327, -0.3588, ..., -0.1254, 0.1147, 0.0319], + [-0.2076, -0.1406, -0.0522, ..., -0.2193, -0.0498, -0.0574], + [-0.0970, -0.0982, -0.1455, ..., 0.0982, 0.0292, -0.1720]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.1130e-07, ..., -1.2107e-07, + 5.5879e-09, 3.7253e-09], + [ 3.7253e-09, 0.0000e+00, 1.7695e-08, ..., 2.2352e-08, + 1.4901e-08, 9.3132e-10], + [-2.7940e-08, 0.0000e+00, 2.7008e-08, ..., 9.3132e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 1.3970e-08, 0.0000e+00, 2.4214e-08, ..., 1.0151e-07, + 3.3528e-08, 2.7940e-09], + [ 1.8626e-09, 0.0000e+00, 2.1420e-08, ..., 1.2107e-08, + 7.4506e-09, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 1.8720e-07, ..., 4.2841e-08, + -3.0734e-08, -6.5193e-09]], device='cuda:0') +Epoch 313, bias, value: tensor([ 4.3289e-03, 1.0270e-02, 1.3852e-02, -6.8141e-05, 2.4522e-02, + -3.5653e-03, -4.8375e-03, -5.5576e-03, 6.8824e-03, -1.7254e-02], + device='cuda:0'), grad: tensor([-1.0822e-06, 2.4680e-07, 4.0978e-08, 1.2014e-07, -3.4552e-07, + 2.8498e-07, 1.5367e-07, 2.0489e-07, 1.2014e-07, 2.6543e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 246.08, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4798 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.12 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.2233, -0.0570, 0.2186, ..., 0.0447, -0.2176, -0.1650], + [-0.1758, -0.2605, -0.0672, ..., -0.1279, -0.2995, -0.0950], + [ 0.2892, -0.0749, -0.2973, ..., -0.2276, -0.2376, -0.2189], + ..., + [-0.1178, -0.0328, -0.3606, ..., -0.1277, 0.1141, 0.0319], + [-0.2088, -0.1408, -0.0505, ..., -0.2197, -0.0490, -0.0578], + [-0.0986, -0.0988, -0.1462, ..., 0.0979, 0.0288, -0.1721]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 1.8626e-09, -2.3656e-07, ..., -2.1420e-08, + 3.3528e-08, 1.8626e-08], + [ 1.9558e-08, 1.8626e-09, 2.6077e-08, ..., 7.4506e-09, + 3.7253e-09, 3.7253e-09], + [ 5.4948e-08, 3.7253e-09, 2.4214e-08, ..., 2.9802e-08, + 1.0245e-08, 3.7253e-09], + ..., + [ 5.8673e-08, 4.6566e-09, 2.6077e-08, ..., 1.3970e-08, + 3.7253e-09, 0.0000e+00], + [ 1.6764e-08, 9.3132e-10, -2.7940e-09, ..., 7.4506e-09, + -5.5879e-09, 6.5193e-09], + [-2.1420e-08, 3.7253e-09, 1.5087e-07, ..., 3.3528e-08, + -3.0734e-08, 1.8626e-09]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0066, 0.0085, 0.0135, 0.0002, 0.0258, -0.0038, -0.0065, -0.0041, + 0.0056, -0.0178], device='cuda:0'), grad: tensor([-3.9954e-07, 1.7881e-07, 3.4925e-07, -1.0934e-06, -3.1106e-07, + 1.7500e-06, -1.2508e-06, 3.3341e-07, -5.1502e-07, 9.3319e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 246.25, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4960 re_mapping 0.0032 re_causal 0.0102 /// teacc 99.18 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.2234, -0.0571, 0.2187, ..., 0.0447, -0.2184, -0.1652], + [-0.1760, -0.2606, -0.0673, ..., -0.1280, -0.3014, -0.0962], + [ 0.2893, -0.0750, -0.2976, ..., -0.2278, -0.2412, -0.2193], + ..., + [-0.1178, -0.0329, -0.3625, ..., -0.1278, 0.1155, 0.0324], + [-0.2094, -0.1409, -0.0497, ..., -0.2200, -0.0494, -0.0581], + [-0.0989, -0.0988, -0.1466, ..., 0.0975, 0.0276, -0.1722]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.0489e-08, ..., -7.4506e-09, + 9.3132e-10, 0.0000e+00], + [ 2.3283e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 5.5879e-09, 0.0000e+00], + [-5.2154e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + ..., + [ 8.3819e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + -1.1176e-08, 0.0000e+00], + [ 1.6764e-08, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 0.0000e+00, 8.3819e-09, ..., 3.6322e-08, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 315, bias, value: tensor([ 0.0065, 0.0103, 0.0134, 0.0002, 0.0264, -0.0038, -0.0064, -0.0051, + 0.0060, -0.0196], device='cuda:0'), grad: tensor([-3.7253e-09, 1.1735e-07, -1.6484e-07, 2.4401e-06, -6.6124e-08, + -3.1441e-06, 1.4715e-07, -1.8626e-09, 1.0338e-07, 5.6438e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 246.15, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4761 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.10 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.2236, -0.0571, 0.2191, ..., 0.0448, -0.2192, -0.1653], + [-0.1765, -0.2607, -0.0673, ..., -0.1282, -0.3036, -0.0973], + [ 0.2893, -0.0750, -0.2978, ..., -0.2279, -0.2430, -0.2193], + ..., + [-0.1179, -0.0329, -0.3637, ..., -0.1280, 0.1164, 0.0328], + [-0.2099, -0.1409, -0.0501, ..., -0.2202, -0.0497, -0.0585], + [-0.0994, -0.0989, -0.1469, ..., 0.0975, 0.0269, -0.1722]], + device='cuda:0'), grad: tensor([[ 4.6566e-08, 0.0000e+00, -1.1176e-08, ..., 2.2072e-07, + 1.3039e-07, 0.0000e+00], + [ 2.1420e-08, 0.0000e+00, 4.6566e-09, ..., 6.1467e-08, + 2.9802e-08, 0.0000e+00], + [-1.8626e-08, 0.0000e+00, 9.3132e-10, ..., 3.2596e-08, + 1.8626e-08, 0.0000e+00], + ..., + [ 4.0978e-08, 0.0000e+00, 9.3132e-10, ..., 1.9744e-07, + 8.5682e-08, 0.0000e+00], + [ 1.7695e-08, 0.0000e+00, -8.3819e-09, ..., 4.1910e-08, + -4.1910e-08, 0.0000e+00], + [-2.5239e-07, 0.0000e+00, 8.3819e-09, ..., -1.4165e-06, + -8.2888e-07, 0.0000e+00]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0066, 0.0103, 0.0134, 0.0009, 0.0265, -0.0042, -0.0066, -0.0050, + 0.0060, -0.0199], device='cuda:0'), grad: tensor([ 1.0729e-06, 2.9057e-07, 1.2666e-07, 6.1747e-07, 1.4426e-06, + 1.6140e-06, 5.8860e-07, 9.2480e-07, -2.1607e-07, -6.4895e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 245.92, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.5122 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.15 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.2238, -0.0571, 0.2204, ..., 0.0449, -0.2216, -0.1672], + [-0.1769, -0.2607, -0.0678, ..., -0.1286, -0.3052, -0.0975], + [ 0.2894, -0.0750, -0.2979, ..., -0.2279, -0.2438, -0.2194], + ..., + [-0.1179, -0.0329, -0.3655, ..., -0.1282, 0.1168, 0.0328], + [-0.2103, -0.1409, -0.0514, ..., -0.2205, -0.0499, -0.0588], + [-0.0996, -0.0989, -0.1472, ..., 0.0976, 0.0267, -0.1722]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 2.2352e-08, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 317, bias, value: tensor([ 0.0072, 0.0107, 0.0134, 0.0008, 0.0265, -0.0042, -0.0070, -0.0054, + 0.0067, -0.0199], device='cuda:0'), grad: tensor([ 1.9558e-08, -2.5891e-07, 2.6077e-08, 3.9116e-08, -8.6613e-08, + -9.7789e-08, 6.0536e-08, 1.6671e-07, 3.1665e-08, 1.0151e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 246.46, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4892 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.11 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.2238, -0.0571, 0.2204, ..., 0.0450, -0.2223, -0.1676], + [-0.1773, -0.2608, -0.0670, ..., -0.1287, -0.3062, -0.0976], + [ 0.2895, -0.0750, -0.2982, ..., -0.2281, -0.2456, -0.2194], + ..., + [-0.1180, -0.0329, -0.3667, ..., -0.1283, 0.1170, 0.0328], + [-0.2105, -0.1409, -0.0526, ..., -0.2211, -0.0501, -0.0589], + [-0.1003, -0.0989, -0.1474, ..., 0.0977, 0.0267, -0.1722]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 7.5437e-08, + 1.7695e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 2.4214e-08, + 4.6566e-08, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 1.2107e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + -1.8068e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 4.6566e-09, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -1.5395e-06, + 9.4064e-08, 0.0000e+00]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0072, 0.0108, 0.0135, 0.0006, 0.0265, -0.0041, -0.0070, -0.0054, + 0.0059, -0.0199], device='cuda:0'), grad: tensor([ 3.1479e-07, 2.2352e-08, 7.8231e-08, 2.5015e-06, 1.7099e-06, + 1.0896e-07, -1.3039e-08, -5.4762e-07, -1.3039e-08, -4.1649e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 246.08, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4694 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.2240, -0.0571, 0.2205, ..., 0.0450, -0.2227, -0.1676], + [-0.1777, -0.2612, -0.0673, ..., -0.1288, -0.3071, -0.0977], + [ 0.2897, -0.0750, -0.2982, ..., -0.2282, -0.2466, -0.2194], + ..., + [-0.1180, -0.0330, -0.3678, ..., -0.1286, 0.1170, 0.0329], + [-0.2110, -0.1410, -0.0514, ..., -0.2214, -0.0516, -0.0589], + [-0.1005, -0.0989, -0.1484, ..., 0.0977, 0.0267, -0.1722]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 5.5879e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -2.0489e-08, ..., 9.3132e-09, + 5.7742e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., -1.7695e-08, + -1.4435e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -5.1223e-08, ..., 1.8626e-09, + -3.5390e-08, 0.0000e+00], + [ 2.7940e-09, 9.3132e-10, 2.3283e-08, ..., -5.5879e-09, + 6.3330e-08, 0.0000e+00]], device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0071, 0.0108, 0.0135, -0.0033, 0.0265, -0.0003, -0.0071, -0.0054, + 0.0066, -0.0200], device='cuda:0'), grad: tensor([ 7.9162e-08, -3.9767e-07, 1.1642e-07, 1.1083e-07, 5.5879e-08, + 2.2538e-07, 1.0896e-07, -3.8370e-07, -8.5309e-07, 9.2480e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 246.61, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4738 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.2241, -0.0571, 0.2210, ..., 0.0453, -0.2237, -0.1678], + [-0.1778, -0.2613, -0.0694, ..., -0.1299, -0.3097, -0.0981], + [ 0.2898, -0.0750, -0.2985, ..., -0.2284, -0.2498, -0.2195], + ..., + [-0.1180, -0.0330, -0.3693, ..., -0.1286, 0.1174, 0.0333], + [-0.2123, -0.1410, -0.0511, ..., -0.2215, -0.0522, -0.0590], + [-0.1006, -0.0989, -0.1489, ..., 0.0977, 0.0265, -0.1723]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -2.7940e-09, ..., -1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 5.2154e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-09, 0.0000e+00], + [-4.5262e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.6857e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 1.2107e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + -6.5193e-09, 2.7940e-09], + [ 1.2666e-07, 0.0000e+00, 1.8626e-09, ..., 2.3283e-08, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0074, 0.0107, 0.0135, -0.0033, 0.0265, -0.0005, -0.0062, -0.0054, + 0.0072, -0.0203], device='cuda:0'), grad: tensor([ 1.0245e-08, 9.6858e-08, -9.3225e-07, 1.8533e-07, -2.9802e-08, + 9.0338e-08, 8.6613e-08, 4.0140e-07, -1.7043e-07, 2.6822e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 246.46, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4717 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.2241, -0.0571, 0.2210, ..., 0.0453, -0.2244, -0.1678], + [-0.1781, -0.2613, -0.0693, ..., -0.1299, -0.3106, -0.0981], + [ 0.2898, -0.0751, -0.2985, ..., -0.2285, -0.2507, -0.2195], + ..., + [-0.1181, -0.0330, -0.3707, ..., -0.1288, 0.1177, 0.0335], + [-0.2125, -0.1410, -0.0513, ..., -0.2216, -0.0520, -0.0591], + [-0.1009, -0.0989, -0.1490, ..., 0.0977, 0.0264, -0.1723]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -3.5390e-08, ..., -8.3819e-09, + 9.3132e-10, 0.0000e+00], + [ 7.8231e-08, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + 1.5832e-08, 0.0000e+00], + [-1.6484e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.3085e-08, 0.0000e+00, 1.8626e-09, ..., 1.5274e-07, + 1.1921e-07, 0.0000e+00], + [ 2.0489e-08, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + -0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.5832e-08, ..., -1.8533e-07, + -1.4808e-07, 0.0000e+00]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0072, 0.0109, 0.0135, -0.0034, 0.0266, -0.0005, -0.0061, -0.0054, + 0.0072, -0.0203], device='cuda:0'), grad: tensor([-6.6124e-08, 1.9185e-07, -3.1758e-07, -5.6811e-08, 9.5926e-08, + 3.7253e-08, 4.4703e-08, 9.3784e-07, 6.5193e-08, -9.1270e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 246.30, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4567 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.2243, -0.0571, 0.2212, ..., 0.0452, -0.2250, -0.1679], + [-0.1788, -0.2614, -0.0693, ..., -0.1302, -0.3120, -0.0982], + [ 0.2900, -0.0752, -0.2987, ..., -0.2286, -0.2520, -0.2195], + ..., + [-0.1181, -0.0331, -0.3716, ..., -0.1291, 0.1177, 0.0335], + [-0.2131, -0.1411, -0.0516, ..., -0.2220, -0.0522, -0.0592], + [-0.1012, -0.0989, -0.1492, ..., 0.0974, 0.0261, -0.1723]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 1.9558e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, + -4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2331e-06, + -4.6659e-07, 0.0000e+00]], device='cuda:0') +Epoch 322, bias, value: tensor([ 0.0070, 0.0113, 0.0135, -0.0034, 0.0269, -0.0005, -0.0060, -0.0059, + 0.0071, -0.0205], device='cuda:0'), grad: tensor([ 1.5832e-08, 7.9162e-08, 2.5146e-08, 4.5635e-08, 2.7418e-06, + -1.7602e-07, 6.8918e-08, -9.3132e-09, 2.7008e-08, -2.8238e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 246.05, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4564 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.2243, -0.0572, 0.2211, ..., 0.0440, -0.2260, -0.1679], + [-0.1790, -0.2615, -0.0693, ..., -0.1302, -0.3127, -0.0983], + [ 0.2900, -0.0752, -0.2988, ..., -0.2288, -0.2528, -0.2195], + ..., + [-0.1182, -0.0331, -0.3725, ..., -0.1292, 0.1178, 0.0336], + [-0.2132, -0.1412, -0.0525, ..., -0.2231, -0.0523, -0.0593], + [-0.1015, -0.0991, -0.1487, ..., 0.0977, 0.0257, -0.1723]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 0.0000e+00, 1.2061e-06, ..., 3.3528e-08, + 2.9057e-07, 0.0000e+00], + [ 8.8476e-08, 0.0000e+00, 2.7008e-08, ..., 1.8626e-09, + 1.5832e-08, 0.0000e+00], + [-3.2317e-06, 0.0000e+00, 1.0245e-08, ..., 9.3132e-10, + 6.5193e-09, 0.0000e+00], + ..., + [ 3.0920e-06, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + -2.6077e-08, 0.0000e+00], + [ 1.8626e-08, 0.0000e+00, 9.4995e-08, ..., 8.3819e-09, + 2.1420e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.7253e-08, ..., 1.3970e-08, + 2.0489e-08, 0.0000e+00]], device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0059, 0.0114, 0.0135, -0.0034, 0.0271, -0.0005, -0.0059, -0.0059, + 0.0069, -0.0203], device='cuda:0'), grad: tensor([ 2.2296e-06, 3.2783e-07, -9.9391e-06, 4.0047e-08, -6.2399e-08, + 3.9600e-06, -6.4559e-06, 9.4771e-06, 2.4214e-07, 1.4622e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 245.84, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5112 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.20 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.2243, -0.0572, 0.2177, ..., 0.0442, -0.2281, -0.1712], + [-0.1793, -0.2617, -0.0696, ..., -0.1304, -0.3152, -0.0984], + [ 0.2901, -0.0752, -0.2991, ..., -0.2291, -0.2576, -0.2197], + ..., + [-0.1182, -0.0332, -0.3759, ..., -0.1292, 0.1184, 0.0341], + [-0.2135, -0.1416, -0.0530, ..., -0.2237, -0.0531, -0.0596], + [-0.1020, -0.0989, -0.1492, ..., 0.0976, 0.0254, -0.1724]], + device='cuda:0'), grad: tensor([[-1.9092e-08, 4.6566e-10, -2.4913e-07, ..., -1.4808e-07, + 6.0536e-09, 4.6566e-10], + [ 1.0245e-08, 9.3132e-10, 1.4901e-08, ..., 8.8476e-09, + 3.8184e-08, 0.0000e+00], + [-6.1002e-08, 1.3970e-09, 1.7695e-08, ..., 1.0245e-08, + 5.4017e-08, 0.0000e+00], + ..., + [ 1.1642e-08, 2.3283e-09, 1.8626e-08, ..., 5.1223e-09, + -2.3935e-07, 0.0000e+00], + [ 6.5193e-09, 9.3132e-10, 1.7229e-08, ..., 3.3528e-08, + 4.9360e-08, 0.0000e+00], + [ 3.2131e-08, 1.3970e-09, 2.8405e-08, ..., -3.2596e-09, + 1.6298e-08, 0.0000e+00]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0025, 0.0109, 0.0135, -0.0035, 0.0273, -0.0003, -0.0026, -0.0055, + 0.0066, -0.0205], device='cuda:0'), grad: tensor([-7.3714e-07, 1.9884e-07, 1.5367e-07, 3.0687e-07, 3.4925e-08, + 2.8964e-07, 8.3819e-08, -7.9069e-07, 3.0082e-07, 1.7416e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 246.09, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4898 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.15 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.2244, -0.0572, 0.2178, ..., 0.0443, -0.2283, -0.1712], + [-0.1796, -0.2620, -0.0699, ..., -0.1306, -0.3171, -0.0985], + [ 0.2902, -0.0754, -0.2991, ..., -0.2292, -0.2583, -0.2197], + ..., + [-0.1183, -0.0334, -0.3787, ..., -0.1293, 0.1186, 0.0341], + [-0.2138, -0.1418, -0.0530, ..., -0.2241, -0.0532, -0.0596], + [-0.1022, -0.0990, -0.1497, ..., 0.0975, 0.0253, -0.1724]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 1.3970e-09, ..., -9.3132e-10, + 2.7940e-09, 1.8626e-09], + [ 6.5193e-09, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, + 1.3970e-09, 4.6566e-10], + [-2.1420e-08, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.5879e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-09, + 9.3132e-10, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 1.1176e-08, ..., 4.6566e-09, + 3.7253e-09, 1.8626e-09], + [ 4.6566e-10, 0.0000e+00, 8.3819e-09, ..., -6.5891e-07, + -2.2119e-07, 0.0000e+00]], device='cuda:0') +Epoch 325, bias, value: tensor([ 0.0026, 0.0109, 0.0135, -0.0035, 0.0274, -0.0003, -0.0026, -0.0055, + 0.0067, -0.0206], device='cuda:0'), grad: tensor([ 1.5367e-08, -1.0710e-08, -3.4459e-08, -1.1642e-08, 1.4054e-06, + 5.0757e-08, -6.4261e-08, 4.5635e-08, 3.1199e-08, -1.4119e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 246.32, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4760 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.11 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.2246, -0.0573, 0.2178, ..., 0.0444, -0.2290, -0.1712], + [-0.1799, -0.2627, -0.0699, ..., -0.1307, -0.3179, -0.0986], + [ 0.2902, -0.0758, -0.2999, ..., -0.2293, -0.2588, -0.2202], + ..., + [-0.1183, -0.0338, -0.3802, ..., -0.1293, 0.1187, 0.0341], + [-0.2141, -0.1427, -0.0532, ..., -0.2243, -0.0534, -0.0600], + [-0.1027, -0.0994, -0.1501, ..., 0.0973, 0.0251, -0.1724]], + device='cuda:0'), grad: tensor([[ 4.5635e-08, 3.5390e-08, 1.8626e-09, ..., 2.1094e-07, + 9.3132e-10, 4.6566e-10], + [ 2.7940e-08, 2.1420e-08, 9.3132e-10, ..., 2.7008e-08, + 1.8626e-09, 0.0000e+00], + [ 2.4680e-08, 2.5611e-08, 0.0000e+00, ..., 3.2596e-09, + 1.3970e-09, 0.0000e+00], + ..., + [ 4.3772e-08, 2.7474e-08, 0.0000e+00, ..., 4.2375e-07, + 5.1223e-09, 0.0000e+00], + [ 9.6858e-08, 7.5437e-08, -1.3970e-09, ..., 4.6566e-09, + 1.3970e-09, 4.6566e-10], + [ 6.2659e-06, 4.8950e-06, 1.3970e-09, ..., -8.6520e-07, + 1.2992e-07, 0.0000e+00]], device='cuda:0') +Epoch 326, bias, value: tensor([ 0.0026, 0.0109, 0.0134, -0.0034, 0.0280, -0.0004, -0.0028, -0.0055, + 0.0069, -0.0209], device='cuda:0'), grad: tensor([ 1.5572e-06, 3.2783e-07, 2.1141e-07, -5.7369e-05, 2.1644e-06, + 1.9610e-05, 1.4296e-07, 2.8163e-06, 3.8743e-07, 3.0175e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 246.38, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4840 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.2247, -0.0575, 0.2178, ..., 0.0444, -0.2293, -0.1712], + [-0.1829, -0.2640, -0.0700, ..., -0.1308, -0.3188, -0.0987], + [ 0.2909, -0.0760, -0.2997, ..., -0.2294, -0.2597, -0.2202], + ..., + [-0.1184, -0.0341, -0.3811, ..., -0.1294, 0.1189, 0.0341], + [-0.2176, -0.1439, -0.0535, ..., -0.2245, -0.0537, -0.0601], + [-0.1053, -0.1017, -0.1503, ..., 0.0972, 0.0249, -0.1724]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 0.0000e+00, -8.7079e-08, ..., -3.6787e-08, + 7.4506e-09, 0.0000e+00], + [ 1.8161e-08, 0.0000e+00, 6.0536e-09, ..., 1.0245e-08, + 1.4435e-08, 0.0000e+00], + [-1.0338e-07, -4.6566e-10, 1.5367e-08, ..., 2.3283e-09, + 1.4901e-08, 0.0000e+00], + ..., + [ 4.6566e-09, 0.0000e+00, 1.3970e-09, ..., 2.3283e-08, + -3.5856e-08, 0.0000e+00], + [ 1.6298e-08, 0.0000e+00, 7.9162e-09, ..., 4.4238e-08, + 1.8626e-09, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 7.8231e-08, ..., -1.6578e-07, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0026, 0.0107, 0.0140, -0.0031, 0.0281, -0.0005, -0.0028, -0.0056, + 0.0069, -0.0212], device='cuda:0'), grad: tensor([-8.1025e-08, 1.5739e-07, -1.3178e-07, 1.3132e-07, 5.3644e-07, + 4.2375e-08, -8.0094e-08, -3.6787e-08, 1.0524e-07, -6.5146e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 246.25, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4721 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.14 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.2250, -0.0576, 0.2180, ..., 0.0446, -0.2299, -0.1713], + [-0.1835, -0.2644, -0.0701, ..., -0.1311, -0.3204, -0.0988], + [ 0.2913, -0.0760, -0.3001, ..., -0.2298, -0.2607, -0.2203], + ..., + [-0.1184, -0.0342, -0.3849, ..., -0.1296, 0.1191, 0.0341], + [-0.2203, -0.1442, -0.0539, ..., -0.2253, -0.0542, -0.0602], + [-0.1056, -0.1017, -0.1509, ..., 0.0971, 0.0247, -0.1724]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 5.5879e-09, ..., 6.5193e-09, + 5.1223e-09, 9.3132e-10], + [ 9.3132e-10, 4.6566e-10, -3.9581e-08, ..., -1.3504e-08, + 9.0338e-08, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 3.2596e-09, 0.0000e+00], + ..., + [ 2.3283e-09, 9.3132e-10, 2.2352e-08, ..., 1.3039e-08, + -3.4040e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8626e-08, + 5.1223e-09, 4.6566e-10], + [ 9.3132e-10, 9.3132e-10, 6.5193e-09, ..., -4.1910e-08, + 1.7695e-08, 0.0000e+00]], device='cuda:0') +Epoch 328, bias, value: tensor([ 0.0028, 0.0108, 0.0142, -0.0032, 0.0283, -0.0004, -0.0029, -0.0057, + 0.0066, -0.0213], device='cuda:0'), grad: tensor([ 8.3819e-08, -3.8696e-07, 3.0268e-08, 7.3714e-07, 6.7055e-08, + 7.6834e-08, 1.8626e-09, -6.9756e-07, 6.1002e-08, 4.9360e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 246.47, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4553 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.10 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.2251, -0.0578, 0.2180, ..., 0.0447, -0.2305, -0.1713], + [-0.1836, -0.2651, -0.0703, ..., -0.1313, -0.3218, -0.1001], + [ 0.2913, -0.0761, -0.3003, ..., -0.2299, -0.2614, -0.2204], + ..., + [-0.1184, -0.0345, -0.3860, ..., -0.1302, 0.1190, 0.0340], + [-0.2205, -0.1448, -0.0521, ..., -0.2256, -0.0539, -0.0606], + [-0.1064, -0.1044, -0.1525, ..., 0.0970, 0.0248, -0.1724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + -3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -8.3819e-09, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0028, 0.0108, 0.0142, -0.0030, 0.0285, -0.0002, -0.0030, -0.0058, + 0.0075, -0.0220], device='cuda:0'), grad: tensor([ 2.3283e-09, -9.3132e-10, 3.2596e-09, -1.2573e-08, 8.8476e-09, + 9.3132e-09, 4.1910e-09, 6.9849e-09, 2.3283e-09, -6.5193e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 246.20, cls_loss 0.0006 cls_loss_mapping 0.0025 cls_loss_causal 0.4736 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.06 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.2252, -0.0579, 0.2181, ..., 0.0448, -0.2312, -0.1713], + [-0.1838, -0.2656, -0.0733, ..., -0.1315, -0.3231, -0.1003], + [ 0.2913, -0.0764, -0.3009, ..., -0.2301, -0.2629, -0.2205], + ..., + [-0.1184, -0.0348, -0.3876, ..., -0.1304, 0.1212, 0.0340], + [-0.2207, -0.1454, -0.0527, ..., -0.2259, -0.0546, -0.0609], + [-0.1065, -0.1044, -0.1530, ..., 0.0969, 0.0229, -0.1724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-08, ..., -8.3819e-09, + 3.7253e-09, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -2.8871e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-2.7940e-09, -9.3132e-10, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 1.7695e-08, ..., 1.8626e-09, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -5.5879e-09, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0028, 0.0103, 0.0140, -0.0029, 0.0286, -0.0002, -0.0021, -0.0053, + 0.0076, -0.0233], device='cuda:0'), grad: tensor([ 0.0000e+00, -2.4214e-07, 3.4459e-08, 2.0489e-08, 2.8871e-08, + 6.4261e-08, -1.0245e-08, 1.4529e-07, -5.8673e-08, 2.1420e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 246.53, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4883 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.15 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.2254, -0.0580, 0.2184, ..., 0.0451, -0.2315, -0.1713], + [-0.1840, -0.2661, -0.0735, ..., -0.1322, -0.3239, -0.1003], + [ 0.2914, -0.0764, -0.3012, ..., -0.2304, -0.2634, -0.2205], + ..., + [-0.1183, -0.0350, -0.3888, ..., -0.1308, 0.1222, 0.0340], + [-0.2210, -0.1454, -0.0531, ..., -0.2261, -0.0548, -0.0611], + [-0.1068, -0.1044, -0.1534, ..., 0.0971, 0.0221, -0.1724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., -1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 5.5879e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0030, 0.0089, 0.0139, -0.0029, 0.0284, -0.0002, -0.0023, -0.0039, + 0.0073, -0.0233], device='cuda:0'), grad: tensor([-5.7742e-08, -2.8592e-07, 7.4506e-09, -3.4273e-07, 9.3132e-09, + 7.8231e-08, 3.3528e-08, 2.5891e-07, 1.7881e-07, 1.2387e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 246.57, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4691 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.2255, -0.0580, 0.2184, ..., 0.0451, -0.2319, -0.1714], + [-0.1841, -0.2670, -0.0735, ..., -0.1322, -0.3253, -0.1003], + [ 0.2915, -0.0765, -0.3014, ..., -0.2305, -0.2654, -0.2206], + ..., + [-0.1184, -0.0353, -0.3894, ..., -0.1313, 0.1223, 0.0340], + [-0.2213, -0.1458, -0.0534, ..., -0.2279, -0.0557, -0.0613], + [-0.1069, -0.1046, -0.1538, ..., 0.0973, 0.0222, -0.1724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6729e-07, ..., -3.8184e-08, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8906e-07, 0.0000e+00], + [-6.5193e-09, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + -1.9278e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.8626e-09, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 332, bias, value: tensor([ 3.0687e-03, 9.0623e-03, 1.3859e-02, -3.0083e-03, 2.8432e-02, + -3.8579e-05, -2.2786e-03, -4.1072e-03, 7.0482e-03, -2.3238e-02], + device='cuda:0'), grad: tensor([-3.4925e-07, 1.2955e-06, -9.3132e-10, 5.3085e-08, -5.5879e-09, + -2.0862e-07, 4.3586e-07, -1.2955e-06, 9.3132e-10, 7.7300e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 246.09, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4880 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.06 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.2257, -0.0580, 0.2185, ..., 0.0424, -0.2331, -0.1714], + [-0.1842, -0.2685, -0.0735, ..., -0.1335, -0.3298, -0.1005], + [ 0.2916, -0.0768, -0.3017, ..., -0.2307, -0.2667, -0.2210], + ..., + [-0.1184, -0.0360, -0.3902, ..., -0.1317, 0.1229, 0.0342], + [-0.2218, -0.1473, -0.0538, ..., -0.2281, -0.0559, -0.0614], + [-0.1070, -0.1047, -0.1547, ..., 0.0995, 0.0222, -0.1724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + -9.3132e-10, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.7940e-09, ..., -0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 333, bias, value: tensor([ 1.7280e-03, 8.3839e-03, 1.3836e-02, -2.9308e-03, 2.7949e-02, + -9.5762e-05, -1.9174e-03, -3.5790e-03, 7.2022e-03, -2.1743e-02], + device='cuda:0'), grad: tensor([ 9.3132e-10, -3.9116e-08, 5.5879e-09, -7.9162e-08, 7.4506e-09, + 3.7253e-09, 2.3283e-08, 3.4459e-08, -8.3819e-09, 4.6566e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 246.19, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4518 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.16 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.2259, -0.0582, 0.2186, ..., 0.0424, -0.2353, -0.1715], + [-0.1843, -0.2695, -0.0744, ..., -0.1333, -0.3304, -0.1003], + [ 0.2921, -0.0768, -0.3019, ..., -0.2307, -0.2672, -0.2197], + ..., + [-0.1186, -0.0380, -0.3907, ..., -0.1321, 0.1228, 0.0328], + [-0.2219, -0.1449, -0.0553, ..., -0.2305, -0.0587, -0.0649], + [-0.1080, -0.1048, -0.1551, ..., 0.0987, 0.0225, -0.1724]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 0.0000e+00, -1.0245e-07, ..., -4.8429e-08, + 2.7940e-09, 1.8626e-09], + [ 2.7940e-09, 9.3132e-10, 3.7253e-09, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + [-2.7381e-07, -4.8429e-08, 1.0245e-08, ..., 6.5193e-09, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.0117e-07, 5.8673e-08, 9.3132e-10, ..., 9.3132e-10, + 3.7253e-09, 5.5879e-09], + [ 1.1176e-08, 1.8626e-09, 1.0245e-08, ..., 1.8626e-09, + 1.8626e-09, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 6.5193e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 334, bias, value: tensor([ 0.0018, 0.0063, 0.0141, -0.0027, 0.0291, -0.0003, -0.0015, -0.0017, + 0.0065, -0.0228], device='cuda:0'), grad: tensor([ 7.1712e-08, -3.9265e-06, -4.3772e-07, -1.7043e-07, 6.1467e-08, + 2.3954e-06, 7.1712e-07, 1.1846e-06, 4.1910e-08, 4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 246.41, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4719 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.12 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.2263, -0.0586, 0.2187, ..., 0.0424, -0.2359, -0.1715], + [-0.1847, -0.2715, -0.0744, ..., -0.1333, -0.3315, -0.1014], + [ 0.2928, -0.0758, -0.3016, ..., -0.2308, -0.2678, -0.2183], + ..., + [-0.1188, -0.0422, -0.3915, ..., -0.1329, 0.1225, 0.0315], + [-0.2222, -0.1447, -0.0559, ..., -0.2331, -0.0595, -0.0651], + [-0.1081, -0.1048, -0.1553, ..., 0.0989, 0.0232, -0.1724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.1514e-07, ..., -4.6566e-08, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 9.3132e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 2.7940e-09, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.8429e-08, ..., 3.7253e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0018, 0.0065, 0.0144, -0.0025, 0.0290, -0.0005, -0.0017, -0.0020, + 0.0066, -0.0222], device='cuda:0'), grad: tensor([-5.8673e-07, 1.1176e-08, 2.8871e-08, -1.8626e-08, -6.5193e-08, + 2.6170e-07, 1.2200e-07, 3.1665e-08, 2.7008e-08, 1.9837e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 246.47, cls_loss 0.0004 cls_loss_mapping 0.0018 cls_loss_causal 0.4679 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.11 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.2264, -0.0586, 0.2188, ..., 0.0424, -0.2360, -0.1715], + [-0.1849, -0.2721, -0.0745, ..., -0.1336, -0.3322, -0.1017], + [ 0.2929, -0.0759, -0.3020, ..., -0.2309, -0.2687, -0.2184], + ..., + [-0.1189, -0.0425, -0.3920, ..., -0.1331, 0.1226, 0.0315], + [-0.2224, -0.1449, -0.0568, ..., -0.2334, -0.0597, -0.0654], + [-0.1082, -0.1048, -0.1559, ..., 0.0988, 0.0232, -0.1725]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.1176e-08, 0.0000e+00], + [-1.9558e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + -4.6566e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.5146e-08, + -1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0019, 0.0066, 0.0144, -0.0024, 0.0293, -0.0006, -0.0017, -0.0021, + 0.0065, -0.0224], device='cuda:0'), grad: tensor([ 5.1223e-08, -3.1013e-06, -3.6322e-08, 7.3574e-08, -1.6857e-07, + -6.7987e-08, 2.2259e-07, 2.9542e-06, 6.7987e-08, -5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 246.57, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4808 re_mapping 0.0029 re_causal 0.0097 /// teacc 98.96 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.2265, -0.0587, 0.2189, ..., 0.0424, -0.2371, -0.1715], + [-0.1852, -0.2727, -0.0746, ..., -0.1338, -0.3313, -0.1017], + [ 0.2930, -0.0761, -0.3027, ..., -0.2311, -0.2697, -0.2185], + ..., + [-0.1189, -0.0427, -0.3932, ..., -0.1338, 0.1220, 0.0314], + [-0.2226, -0.1453, -0.0578, ..., -0.2336, -0.0599, -0.0653], + [-0.1084, -0.1049, -0.1564, ..., 0.0988, 0.0235, -0.1725]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, -1.4901e-08, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 2.7940e-09, 3.7253e-09, ..., 4.6566e-09, + -4.6566e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + -0.0000e+00, -1.8626e-09], + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., -1.3970e-08, + -2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 337, bias, value: tensor([ 0.0019, 0.0067, 0.0144, -0.0023, 0.0295, -0.0007, -0.0016, -0.0023, + 0.0066, -0.0224], device='cuda:0'), grad: tensor([ 5.5879e-09, -5.4017e-08, 1.0245e-08, -1.3327e-06, 3.7253e-08, + 1.3094e-06, 2.4214e-08, 5.3085e-08, -1.3039e-08, -2.8871e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 246.71, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4586 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.06 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.2266, -0.0589, 0.2196, ..., 0.0408, -0.2396, -0.1716], + [-0.1861, -0.2736, -0.0748, ..., -0.1347, -0.3339, -0.1025], + [ 0.2934, -0.0761, -0.3033, ..., -0.2317, -0.2738, -0.2199], + ..., + [-0.1192, -0.0429, -0.3944, ..., -0.1354, 0.1224, 0.0354], + [-0.2228, -0.1456, -0.0558, ..., -0.2378, -0.0600, -0.0662], + [-0.1086, -0.1048, -0.1589, ..., 0.1035, 0.0280, -0.1729]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, -3.3993e-07, ..., -1.0524e-07, + 8.3819e-09, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 6.5193e-09, 0.0000e+00], + [-2.2352e-08, -0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + 7.4506e-09, 2.7940e-09], + ..., + [ 1.1176e-08, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, + -1.0245e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 2.0489e-08, ..., 6.5193e-09, + 5.5879e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.8964e-07, ..., 8.9407e-08, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0010, 0.0067, 0.0144, -0.0038, 0.0250, -0.0030, -0.0004, -0.0024, + 0.0102, -0.0173], device='cuda:0'), grad: tensor([-4.8708e-07, 2.3283e-08, -5.5879e-09, 2.4214e-08, -1.3039e-08, + -2.6450e-07, 2.4214e-07, -4.6566e-09, 4.3772e-08, 4.4983e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 246.45, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4826 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.2267, -0.0592, 0.2193, ..., 0.0406, -0.2429, -0.1727], + [-0.1863, -0.2746, -0.0751, ..., -0.1352, -0.3349, -0.1038], + [ 0.2936, -0.0760, -0.3037, ..., -0.2320, -0.2745, -0.2201], + ..., + [-0.1193, -0.0431, -0.3952, ..., -0.1353, 0.1249, 0.0357], + [-0.2230, -0.1460, -0.0557, ..., -0.2380, -0.0598, -0.0667], + [-0.1095, -0.1050, -0.1595, ..., 0.1036, 0.0279, -0.1732]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -7.1339e-07, ..., -2.8498e-07, + 0.0000e+00, 9.3132e-10], + [ 1.8626e-09, 9.3132e-10, 1.7695e-08, ..., 9.3132e-09, + 1.2107e-08, 2.7940e-09], + [-4.6566e-09, 9.3132e-10, 7.4506e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 9.3132e-10, 2.7940e-08, ..., 9.3132e-09, + -1.2107e-08, -2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, + 1.8626e-09, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 4.3586e-07, ..., 1.7043e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 339, bias, value: tensor([ 0.0005, 0.0067, 0.0144, -0.0032, 0.0249, -0.0030, -0.0009, -0.0022, + 0.0107, -0.0172], device='cuda:0'), grad: tensor([-1.3588e-06, 6.4261e-08, 2.7940e-09, -3.4459e-08, 2.6077e-08, + -1.2107e-07, 5.2154e-07, 5.0291e-08, 2.5146e-08, 8.3353e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 246.62, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4748 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.18 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.2267, -0.0591, 0.2198, ..., 0.0412, -0.2429, -0.1727], + [-0.1864, -0.2769, -0.0751, ..., -0.1356, -0.3357, -0.1035], + [ 0.2937, -0.0763, -0.3044, ..., -0.2332, -0.2754, -0.2211], + ..., + [-0.1193, -0.0437, -0.3966, ..., -0.1353, 0.1255, 0.0373], + [-0.2237, -0.1463, -0.0561, ..., -0.2382, -0.0599, -0.0669], + [-0.1103, -0.1055, -0.1614, ..., 0.1036, 0.0278, -0.1740]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 2.7940e-09, + 1.8626e-09, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., -5.5879e-09, + 1.5460e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 2.7940e-09, 2.7940e-09], + ..., + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., 7.4506e-09, + -2.7847e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 5.5879e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.0489e-08, + 1.1455e-07, 2.7940e-09]], device='cuda:0') +Epoch 340, bias, value: tensor([ 1.1821e-03, 6.7544e-03, 1.4147e-02, 1.4303e-05, 2.4928e-02, + -5.7604e-03, -9.4102e-04, -2.1872e-03, 1.0707e-02, -1.7279e-02], + device='cuda:0'), grad: tensor([ 3.2596e-08, 3.1199e-07, 2.7008e-08, -9.3132e-10, -1.9465e-07, + 1.0803e-07, 1.6764e-08, -8.0466e-07, 2.7008e-08, 4.8708e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 246.49, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4536 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.13 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.2269, -0.0592, 0.2201, ..., 0.0415, -0.2430, -0.1727], + [-0.1890, -0.2777, -0.0751, ..., -0.1369, -0.3363, -0.1037], + [ 0.2944, -0.0765, -0.3048, ..., -0.2337, -0.2758, -0.2212], + ..., + [-0.1195, -0.0439, -0.3980, ..., -0.1355, 0.1257, 0.0374], + [-0.2241, -0.1464, -0.0564, ..., -0.2384, -0.0600, -0.0674], + [-0.1105, -0.1055, -0.1623, ..., 0.1036, 0.0278, -0.1741]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.7695e-08, -8.0094e-08, ..., -3.6322e-08, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 3.0734e-08, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.6524e-06, 9.0376e-06, 2.0489e-08, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.0117e-07, 6.5472e-07, 3.7253e-09, ..., 5.5879e-09, + -1.8626e-09, 0.0000e+00], + [ 3.7253e-08, 2.1420e-08, 2.0489e-08, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 2.9802e-08, ..., 2.0489e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 341, bias, value: tensor([ 1.5533e-03, 6.6092e-03, 1.4545e-02, -7.4696e-05, 2.4801e-02, + -5.6534e-03, 5.3902e-04, -2.1868e-03, 1.0739e-02, -1.7307e-02], + device='cuda:0'), grad: tensor([-1.5460e-07, 7.0781e-08, 2.3246e-05, -2.5496e-05, -2.7940e-08, + 3.6601e-07, 3.0734e-08, 1.7267e-06, 1.6298e-07, 9.8720e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 246.37, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4867 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.2270, -0.0595, 0.2202, ..., 0.0415, -0.2431, -0.1727], + [-0.1892, -0.2784, -0.0751, ..., -0.1342, -0.3370, -0.1038], + [ 0.2945, -0.0800, -0.3050, ..., -0.2346, -0.2765, -0.2211], + ..., + [-0.1196, -0.0446, -0.3988, ..., -0.1386, 0.1260, 0.0373], + [-0.2244, -0.1468, -0.0566, ..., -0.2387, -0.0602, -0.0692], + [-0.1106, -0.1055, -0.1624, ..., 0.1036, 0.0278, -0.1743]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 0.0000e+00, -2.1420e-08, ..., -1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 1.1176e-08, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [-9.8720e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 2.5146e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 4.4703e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 2.7940e-09], + [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., 2.0489e-08, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0016, 0.0082, 0.0143, -0.0003, 0.0248, -0.0054, 0.0007, -0.0036, + 0.0106, -0.0173], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.3970e-08, -2.8405e-07, 2.7940e-09, -1.8626e-08, + 1.8626e-09, 4.6566e-09, 7.4506e-08, 1.4435e-07, 7.0781e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 246.33, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4715 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.07 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.2272, -0.0596, 0.2202, ..., 0.0415, -0.2431, -0.1727], + [-0.1911, -0.2785, -0.0751, ..., -0.1342, -0.3388, -0.1046], + [ 0.2952, -0.0800, -0.3051, ..., -0.2349, -0.2770, -0.2206], + ..., + [-0.1198, -0.0447, -0.3990, ..., -0.1386, 0.1266, 0.0377], + [-0.2248, -0.1470, -0.0568, ..., -0.2390, -0.0603, -0.0696], + [-0.1109, -0.1054, -0.1625, ..., 0.1035, 0.0278, -0.1744]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., -2.7008e-08, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, -1.3039e-08, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-09, ..., 6.5193e-09, + -9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 5.5879e-09, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0016, 0.0082, 0.0147, -0.0004, 0.0248, -0.0054, 0.0008, -0.0036, + 0.0106, -0.0174], device='cuda:0'), grad: tensor([-1.0990e-07, -2.1793e-07, 1.0245e-08, -1.7695e-08, 2.3283e-08, + 4.0047e-08, 1.8626e-08, 1.4994e-07, 8.8476e-08, 2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 245.84, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4783 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.09 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.2274, -0.0594, 0.2203, ..., 0.0416, -0.2432, -0.1728], + [-0.1914, -0.2791, -0.0754, ..., -0.1342, -0.3398, -0.1058], + [ 0.2954, -0.0800, -0.3053, ..., -0.2349, -0.2772, -0.2206], + ..., + [-0.1198, -0.0449, -0.3997, ..., -0.1387, 0.1268, 0.0377], + [-0.2253, -0.1474, -0.0572, ..., -0.2392, -0.0604, -0.0712], + [-0.1110, -0.1055, -0.1626, ..., 0.1035, 0.0278, -0.1745]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5646e-07, ..., -1.1548e-07, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 1.3039e-08, 2.7940e-09], + [-1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, + 8.3819e-09, 1.8626e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., -2.7940e-09, + -4.4703e-08, -9.3132e-09], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.7544e-08, ..., 6.9849e-08, + 1.8626e-08, 3.7253e-09]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0016, 0.0081, 0.0148, -0.0003, 0.0248, -0.0054, 0.0009, -0.0036, + 0.0104, -0.0174], device='cuda:0'), grad: tensor([-3.5111e-07, 1.0617e-07, 1.7043e-07, 1.9558e-08, -9.3132e-09, + 3.8184e-08, 1.4529e-07, -1.8533e-07, -2.1886e-07, 2.7567e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 246.16, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4727 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.14 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.2275, -0.0595, 0.2202, ..., 0.0416, -0.2436, -0.1729], + [-0.1915, -0.2793, -0.0754, ..., -0.1342, -0.3405, -0.1060], + [ 0.2955, -0.0800, -0.3056, ..., -0.2352, -0.2780, -0.2208], + ..., + [-0.1199, -0.0450, -0.4000, ..., -0.1387, 0.1269, 0.0378], + [-0.2257, -0.1477, -0.0588, ..., -0.2398, -0.0609, -0.0724], + [-0.1111, -0.1054, -0.1627, ..., 0.1035, 0.0277, -0.1746]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -7.4506e-09, ..., -5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 2.6077e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-4.8429e-08, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.3970e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.1176e-08, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0015, 0.0083, 0.0148, -0.0003, 0.0249, -0.0054, 0.0011, -0.0038, + 0.0103, -0.0174], device='cuda:0'), grad: tensor([-7.4506e-09, -4.5449e-07, -7.8231e-08, -3.7253e-09, 1.1176e-08, + 4.3772e-08, -7.6368e-08, 4.7684e-07, 5.0291e-08, 4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 246.23, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4732 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.14 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.2278, -0.0596, 0.2209, ..., 0.0419, -0.2434, -0.1728], + [-0.1920, -0.2796, -0.0754, ..., -0.1351, -0.3422, -0.1063], + [ 0.2958, -0.0800, -0.3064, ..., -0.2361, -0.2782, -0.2208], + ..., + [-0.1200, -0.0451, -0.4008, ..., -0.1389, 0.1268, 0.0379], + [-0.2260, -0.1478, -0.0591, ..., -0.2403, -0.0611, -0.0727], + [-0.1115, -0.1054, -0.1635, ..., 0.1036, 0.0278, -0.1748]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.6566e-09, ..., 4.6566e-09, + 9.3132e-10, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-08, 0.0000e+00], + [-3.3807e-07, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 3.2969e-07, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, + -5.4948e-08, -9.3132e-10], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., -1.5832e-08, + 1.3039e-08, 1.8626e-09]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0022, 0.0081, 0.0149, -0.0004, 0.0249, -0.0051, 0.0003, -0.0039, + 0.0102, -0.0173], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.2480e-07, -7.4040e-07, 3.7253e-09, 2.0489e-08, + -9.3132e-10, 7.4506e-09, 5.2247e-07, 2.0489e-08, 3.9116e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 246.59, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4657 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.14 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.2283, -0.0603, 0.2211, ..., 0.0420, -0.2439, -0.1729], + [-0.1923, -0.2805, -0.0756, ..., -0.1352, -0.3440, -0.1070], + [ 0.2964, -0.0792, -0.3067, ..., -0.2364, -0.2766, -0.2208], + ..., + [-0.1202, -0.0455, -0.4017, ..., -0.1390, 0.1272, 0.0381], + [-0.2268, -0.1484, -0.0599, ..., -0.2412, -0.0620, -0.0731], + [-0.1117, -0.1055, -0.1639, ..., 0.1035, 0.0277, -0.1750]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, -9.6858e-08, ..., -3.0734e-08, + 3.7253e-09, 9.3132e-10], + [ 2.7940e-08, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, + 1.0245e-08, 9.3132e-10], + [-2.5798e-07, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, + -5.4948e-08, 0.0000e+00], + ..., + [ 1.7695e-07, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, + 3.3528e-08, -9.3132e-10], + [ 1.0245e-08, 0.0000e+00, 1.0245e-08, ..., 9.3132e-09, + 1.3039e-08, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 2.0489e-08, ..., 3.1013e-07, + 7.0781e-08, 9.3132e-10]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0022, 0.0080, 0.0153, -0.0005, 0.0250, -0.0050, 0.0004, -0.0038, + 0.0099, -0.0174], device='cuda:0'), grad: tensor([-1.9837e-07, 6.4261e-08, -7.3574e-07, 8.0094e-08, -6.5193e-07, + 7.9162e-08, 4.9360e-08, 5.3924e-07, 1.0803e-07, 6.6031e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 246.07, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4685 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.15 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.2285, -0.0604, 0.2214, ..., 0.0422, -0.2441, -0.1729], + [-0.1931, -0.2812, -0.0756, ..., -0.1352, -0.3456, -0.1078], + [ 0.2972, -0.0791, -0.3071, ..., -0.2370, -0.2755, -0.2206], + ..., + [-0.1207, -0.0458, -0.4037, ..., -0.1391, 0.1276, 0.0382], + [-0.2275, -0.1486, -0.0603, ..., -0.2416, -0.0623, -0.0732], + [-0.1119, -0.1057, -0.1643, ..., 0.1034, 0.0276, -0.1751]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-08, ..., -3.1665e-08, + 9.3132e-10, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 1.3970e-08, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, -0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 2.6077e-08, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -3.0734e-08, 5.5879e-09, ..., 2.7940e-09, + -3.2596e-08, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., -2.0489e-08, + -1.3970e-08, 0.0000e+00]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0026, 0.0079, 0.0157, -0.0006, 0.0250, -0.0050, 0.0003, -0.0038, + 0.0098, -0.0175], device='cuda:0'), grad: tensor([-1.3597e-07, 3.6322e-08, 8.3819e-09, 3.4180e-07, 4.6566e-09, + 9.5926e-08, 1.0058e-07, 5.4948e-08, -4.6473e-07, -3.7253e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 246.11, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4518 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.17 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.2298, -0.0608, 0.2214, ..., 0.0422, -0.2449, -0.1733], + [-0.1932, -0.2818, -0.0755, ..., -0.1353, -0.3468, -0.1081], + [ 0.2975, -0.0791, -0.3069, ..., -0.2375, -0.2762, -0.2207], + ..., + [-0.1207, -0.0460, -0.4051, ..., -0.1392, 0.1279, 0.0385], + [-0.2280, -0.1482, -0.0607, ..., -0.2422, -0.0626, -0.0733], + [-0.1120, -0.1057, -0.1646, ..., 0.1034, 0.0276, -0.1753]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 1.1176e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8720e-07, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.5146e-08, + 2.7940e-09, 9.3132e-10]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0024, 0.0079, 0.0158, -0.0006, 0.0251, -0.0050, 0.0002, -0.0038, + 0.0098, -0.0176], device='cuda:0'), grad: tensor([ 1.4901e-08, 6.2212e-07, 1.8626e-09, 2.6077e-08, -7.2177e-07, + -6.8918e-08, 2.9802e-08, 2.7940e-09, 9.3132e-09, 8.2888e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 245.89, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4748 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.09 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.2301, -0.0614, 0.2216, ..., 0.0422, -0.2450, -0.1733], + [-0.1934, -0.2823, -0.0755, ..., -0.1353, -0.3476, -0.1092], + [ 0.2976, -0.0791, -0.3071, ..., -0.2378, -0.2766, -0.2209], + ..., + [-0.1208, -0.0462, -0.4058, ..., -0.1392, 0.1282, 0.0385], + [-0.2284, -0.1485, -0.0610, ..., -0.2429, -0.0627, -0.0741], + [-0.1134, -0.1058, -0.1647, ..., 0.1034, 0.0276, -0.1755]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -2.3283e-09, ..., 9.3132e-10, + 1.3970e-09, 4.6566e-10], + [ 3.0268e-08, 0.0000e+00, -6.9849e-09, ..., 2.6543e-08, + 1.1642e-07, -0.0000e+00], + [ 2.6543e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 2.0023e-08, 0.0000e+00, 4.6566e-10, ..., -2.7474e-08, + -1.2573e-07, -4.6566e-10], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-09, 0.0000e+00, 1.8626e-09, ..., -3.7253e-09, + 6.0536e-09, 4.6566e-10]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0025, 0.0082, 0.0158, -0.0007, 0.0250, -0.0046, -0.0001, -0.0041, + 0.0097, -0.0175], device='cuda:0'), grad: tensor([ 2.7008e-08, 3.9907e-07, 1.3411e-07, -4.8056e-07, -1.3970e-09, + 5.7742e-08, 1.0803e-07, -3.1339e-07, 2.3283e-08, 3.6787e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 246.09, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4948 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.17 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.2303, -0.0619, 0.2220, ..., 0.0425, -0.2451, -0.1733], + [-0.1935, -0.2829, -0.0756, ..., -0.1354, -0.3491, -0.1101], + [ 0.2978, -0.0791, -0.3077, ..., -0.2386, -0.2781, -0.2211], + ..., + [-0.1209, -0.0466, -0.4075, ..., -0.1392, 0.1286, 0.0388], + [-0.2288, -0.1488, -0.0618, ..., -0.2436, -0.0636, -0.0748], + [-0.1137, -0.1058, -0.1650, ..., 0.1033, 0.0276, -0.1758]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0028, 0.0084, 0.0156, -0.0007, 0.0251, -0.0045, -0.0004, -0.0043, + 0.0095, -0.0177], device='cuda:0'), grad: tensor([ 8.3819e-09, -4.5449e-07, 3.7998e-07, 5.5879e-09, 3.5390e-08, + 2.7940e-09, 1.8626e-08, 2.0489e-08, -2.4214e-08, 6.5193e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 246.52, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4642 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.09 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.2306, -0.0621, 0.2224, ..., 0.0426, -0.2451, -0.1732], + [-0.1943, -0.2836, -0.0756, ..., -0.1355, -0.3504, -0.1103], + [ 0.2982, -0.0792, -0.3080, ..., -0.2379, -0.2791, -0.2212], + ..., + [-0.1210, -0.0471, -0.4086, ..., -0.1395, 0.1286, 0.0388], + [-0.2298, -0.1492, -0.0621, ..., -0.2440, -0.0641, -0.0751], + [-0.1156, -0.1059, -0.1653, ..., 0.1032, 0.0276, -0.1760]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.7940e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0030, 0.0076, 0.0158, -0.0009, 0.0252, -0.0042, -0.0007, -0.0036, + 0.0096, -0.0177], device='cuda:0'), grad: tensor([ 2.1420e-08, -6.4746e-06, 2.7008e-08, 2.7940e-09, -1.7695e-08, + 1.0245e-08, 5.7742e-08, 5.6438e-06, -2.2352e-08, 7.5903e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 246.40, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4495 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.20 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.2308, -0.0625, 0.2224, ..., 0.0426, -0.2454, -0.1733], + [-0.1946, -0.2846, -0.0756, ..., -0.1360, -0.3516, -0.1106], + [ 0.2987, -0.0792, -0.3082, ..., -0.2375, -0.2799, -0.2211], + ..., + [-0.1211, -0.0474, -0.4094, ..., -0.1396, 0.1286, 0.0388], + [-0.2303, -0.1495, -0.0625, ..., -0.2450, -0.0645, -0.0756], + [-0.1154, -0.1057, -0.1654, ..., 0.1032, 0.0276, -0.1761]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-07, ..., -6.6124e-08, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.6811e-08, ..., 2.6077e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 2.7940e-09, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0030, 0.0075, 0.0160, -0.0009, 0.0253, -0.0042, -0.0007, -0.0036, + 0.0094, -0.0177], device='cuda:0'), grad: tensor([-2.8964e-07, -9.3132e-10, 1.2107e-08, -5.5879e-09, 7.4506e-09, + 1.1176e-08, 1.3039e-07, -2.7940e-09, 1.1362e-07, 3.0734e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 246.32, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4719 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.16 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.2310, -0.0627, 0.2225, ..., 0.0427, -0.2456, -0.1733], + [-0.1947, -0.2850, -0.0756, ..., -0.1360, -0.3527, -0.1115], + [ 0.2991, -0.0793, -0.3084, ..., -0.2363, -0.2806, -0.2212], + ..., + [-0.1212, -0.0476, -0.4100, ..., -0.1397, 0.1289, 0.0388], + [-0.2295, -0.1501, -0.0627, ..., -0.2447, -0.0646, -0.0762], + [-0.1164, -0.1057, -0.1656, ..., 0.1032, 0.0275, -0.1765]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.7695e-08, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -9.3132e-10, ..., 9.3132e-10, + -9.3132e-10, 0.0000e+00], + [ 1.6764e-08, 1.1176e-08, 4.6566e-09, ..., -3.6228e-07, + -2.4214e-08, 0.0000e+00]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0030, 0.0058, 0.0159, -0.0022, 0.0253, -0.0060, 0.0023, -0.0019, + 0.0095, -0.0178], device='cuda:0'), grad: tensor([-3.7253e-09, -2.5146e-08, -2.6077e-08, 2.5146e-08, 6.9849e-07, + -1.2945e-07, 2.5146e-08, 2.8871e-08, 1.8626e-09, -5.9418e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 245.83, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4554 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.21 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.2311, -0.0628, 0.2226, ..., 0.0428, -0.2457, -0.1733], + [-0.1949, -0.2851, -0.0756, ..., -0.1362, -0.3537, -0.1120], + [ 0.2993, -0.0792, -0.3087, ..., -0.2365, -0.2816, -0.2214], + ..., + [-0.1212, -0.0477, -0.4115, ..., -0.1398, 0.1293, 0.0392], + [-0.2295, -0.1504, -0.0640, ..., -0.2450, -0.0660, -0.0771], + [-0.1172, -0.1058, -0.1658, ..., 0.1031, 0.0275, -0.1769]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 4.6566e-09, 0.0000e+00], + [-2.0489e-08, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -9.3132e-09, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.7695e-08, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0031, 0.0060, 0.0157, -0.0027, 0.0253, -0.0056, 0.0024, -0.0021, + 0.0088, -0.0178], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.4901e-08, -2.4214e-08, 1.2107e-08, -4.3772e-08, + -5.8673e-08, 3.0734e-08, -2.0489e-08, 3.7253e-08, 4.7497e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 246.04, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4621 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.13 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.2313, -0.0630, 0.2232, ..., 0.0410, -0.2459, -0.1734], + [-0.1951, -0.2854, -0.0771, ..., -0.1364, -0.3557, -0.1124], + [ 0.2994, -0.0792, -0.3102, ..., -0.2372, -0.2826, -0.2214], + ..., + [-0.1213, -0.0478, -0.4121, ..., -0.1398, 0.1307, 0.0393], + [-0.2300, -0.1507, -0.0643, ..., -0.2462, -0.0664, -0.0774], + [-0.1176, -0.1058, -0.1659, ..., 0.1033, 0.0274, -0.1770]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 4.2841e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.5832e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.3597e-07, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-08, + 4.5635e-08, 2.7940e-09]], device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0022, 0.0059, 0.0156, -0.0029, 0.0254, -0.0057, 0.0024, -0.0020, + 0.0086, -0.0177], device='cuda:0'), grad: tensor([ 2.0489e-08, 2.2817e-07, 1.0990e-07, 1.8626e-08, 1.3225e-07, + 1.5832e-08, -4.6566e-09, -6.4727e-07, -1.5926e-07, 2.9057e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 246.46, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4676 re_mapping 0.0034 re_causal 0.0103 /// teacc 99.14 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.2317, -0.0631, 0.2233, ..., 0.0410, -0.2463, -0.1734], + [-0.1954, -0.2859, -0.0772, ..., -0.1364, -0.3566, -0.1125], + [ 0.2997, -0.0795, -0.3108, ..., -0.2376, -0.2848, -0.2216], + ..., + [-0.1214, -0.0486, -0.4127, ..., -0.1398, 0.1310, 0.0395], + [-0.2317, -0.1513, -0.0648, ..., -0.2466, -0.0666, -0.0775], + [-0.1192, -0.1060, -0.1663, ..., 0.1032, 0.0274, -0.1771]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., -6.5193e-09, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.1420e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.4901e-08, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0021, 0.0059, 0.0156, -0.0053, 0.0254, -0.0040, 0.0027, -0.0020, + 0.0085, -0.0178], device='cuda:0'), grad: tensor([-2.5146e-08, -4.7497e-08, -4.6566e-08, -1.9558e-08, 9.3132e-10, + 1.3039e-08, -6.5193e-09, 7.9162e-08, 2.6077e-08, 3.5390e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 245.86, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4664 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.21 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.2322, -0.0633, 0.2234, ..., 0.0411, -0.2469, -0.1735], + [-0.1956, -0.2865, -0.0771, ..., -0.1364, -0.3580, -0.1129], + [ 0.3003, -0.0795, -0.3111, ..., -0.2377, -0.2876, -0.2217], + ..., + [-0.1216, -0.0488, -0.4131, ..., -0.1400, 0.1315, 0.0397], + [-0.2338, -0.1536, -0.0652, ..., -0.2472, -0.0669, -0.0776], + [-0.1194, -0.1060, -0.1664, ..., 0.1032, 0.0274, -0.1771]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, -2.7940e-09, ..., 1.1176e-08, + 5.5879e-09, 0.0000e+00], + [ 5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 8.3819e-09, + 1.8626e-09, 0.0000e+00], + [-3.5390e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.5146e-08, 9.3132e-10, 9.3132e-10, ..., 1.0245e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 9.3132e-10, ..., 3.2596e-08, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, -7.8231e-08, 2.7940e-09, ..., -6.9756e-07, + -1.4249e-07, 0.0000e+00]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0021, 0.0061, 0.0151, -0.0055, 0.0254, -0.0038, 0.0027, -0.0021, + 0.0083, -0.0178], device='cuda:0'), grad: tensor([ 3.4459e-08, 1.3039e-08, -7.9162e-08, 1.1176e-07, 1.0254e-06, + 2.9150e-07, -1.3970e-08, 9.3132e-08, 6.6124e-08, -1.5413e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 245.96, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4517 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.19 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.2327, -0.0635, 0.2235, ..., 0.0411, -0.2471, -0.1735], + [-0.1958, -0.2867, -0.0771, ..., -0.1364, -0.3590, -0.1137], + [ 0.3005, -0.0795, -0.3112, ..., -0.2378, -0.2878, -0.2217], + ..., + [-0.1217, -0.0490, -0.4139, ..., -0.1399, 0.1329, 0.0399], + [-0.2347, -0.1554, -0.0654, ..., -0.2473, -0.0656, -0.0778], + [-0.1194, -0.1059, -0.1665, ..., 0.1032, 0.0272, -0.1772]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -6.5193e-09, + 0.0000e+00, -9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., -4.9360e-08, + -1.0245e-08, 0.0000e+00]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0021, 0.0061, 0.0151, -0.0056, 0.0254, -0.0038, 0.0026, -0.0020, + 0.0091, -0.0179], device='cuda:0'), grad: tensor([-1.9558e-08, -6.7055e-08, 7.4506e-09, -3.2596e-08, 5.5879e-08, + 1.6764e-08, 1.3039e-08, 6.6124e-08, 9.7789e-08, -1.2759e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 246.02, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4759 re_mapping 0.0028 re_causal 0.0091 /// teacc 99.19 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.2333, -0.0638, 0.2236, ..., 0.0410, -0.2474, -0.1735], + [-0.1961, -0.2873, -0.0771, ..., -0.1366, -0.3611, -0.1138], + [ 0.3008, -0.0796, -0.3114, ..., -0.2401, -0.2883, -0.2217], + ..., + [-0.1218, -0.0493, -0.4146, ..., -0.1401, 0.1330, 0.0399], + [-0.2352, -0.1557, -0.0656, ..., -0.2479, -0.0657, -0.0779], + [-0.1197, -0.1059, -0.1668, ..., 0.1032, 0.0273, -0.1773]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.4703e-08, ..., -9.3132e-10, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.7253e-09, + 5.5879e-09, 0.0000e+00], + [-4.6566e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.5832e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0020, 0.0062, 0.0150, -0.0055, 0.0255, -0.0038, 0.0026, -0.0021, + 0.0090, -0.0179], device='cuda:0'), grad: tensor([ 1.1548e-07, -4.7497e-08, -2.7940e-09, 3.7253e-09, 1.7695e-08, + 5.2154e-08, -1.7136e-07, 2.7940e-09, -5.5879e-09, 3.9116e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 246.68, cls_loss 0.0004 cls_loss_mapping 0.0015 cls_loss_causal 0.4721 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.13 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.2337, -0.0643, 0.2237, ..., 0.0410, -0.2475, -0.1736], + [-0.1974, -0.2879, -0.0771, ..., -0.1366, -0.3621, -0.1143], + [ 0.3008, -0.0796, -0.3118, ..., -0.2408, -0.2883, -0.2218], + ..., + [-0.1214, -0.0498, -0.4148, ..., -0.1402, 0.1332, 0.0400], + [-0.2358, -0.1576, -0.0657, ..., -0.2481, -0.0658, -0.0780], + [-0.1200, -0.1090, -0.1669, ..., 0.1032, 0.0273, -0.1773]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 2.7940e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.9914e-06, + 2.7940e-09, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 8.3819e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.4925e-07, + 9.3132e-10, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 3.3528e-08, + 7.4506e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 2.3264e-06, + -1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0020, 0.0060, 0.0137, -0.0049, 0.0255, -0.0038, 0.0026, -0.0018, + 0.0084, -0.0180], device='cuda:0'), grad: tensor([ 6.5193e-09, -1.4573e-05, 7.0781e-08, 3.8650e-07, 1.1148e-06, + -4.0606e-07, 3.0734e-08, 2.4233e-06, 5.2154e-08, 1.0878e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 246.19, cls_loss 0.0003 cls_loss_mapping 0.0013 cls_loss_causal 0.4674 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.18 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.2338, -0.0644, 0.2239, ..., 0.0410, -0.2476, -0.1736], + [-0.1975, -0.2882, -0.0771, ..., -0.1365, -0.3636, -0.1149], + [ 0.3010, -0.0797, -0.3121, ..., -0.2410, -0.2884, -0.2218], + ..., + [-0.1215, -0.0501, -0.4153, ..., -0.1402, 0.1341, 0.0405], + [-0.2361, -0.1583, -0.0660, ..., -0.2486, -0.0661, -0.0783], + [-0.1201, -0.1090, -0.1675, ..., 0.1031, 0.0272, -0.1779]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, -5.5879e-09, ..., -6.5193e-09, + 5.5879e-09, 1.8626e-09], + [ 1.8626e-09, 1.8626e-09, 8.3819e-09, ..., 7.4506e-09, + 2.4214e-08, 1.8626e-09], + [ 3.7253e-09, 3.7253e-09, 4.6566e-09, ..., 2.7940e-09, + 3.7253e-09, 9.3132e-10], + ..., + [ 2.7940e-09, 3.7253e-09, 9.3132e-10, ..., 1.1176e-08, + -3.2596e-08, 0.0000e+00], + [ 2.7940e-09, 2.7940e-09, 6.5193e-09, ..., 6.5193e-09, + 1.0245e-08, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 1.2107e-08, ..., 3.9581e-07, + 6.0536e-08, 0.0000e+00]], device='cuda:0') +Epoch 362, bias, value: tensor([ 0.0021, 0.0060, 0.0136, -0.0048, 0.0255, -0.0038, 0.0026, -0.0017, + 0.0081, -0.0180], device='cuda:0'), grad: tensor([ 1.1176e-08, 6.5193e-09, 4.0978e-08, -1.0803e-07, -8.9966e-07, + 1.4063e-07, -1.9558e-07, 2.8871e-08, 6.7987e-08, 9.1270e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 245.75, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4743 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.13 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.2351, -0.0649, 0.2237, ..., 0.0410, -0.2488, -0.1742], + [-0.1982, -0.2887, -0.0767, ..., -0.1365, -0.3656, -0.1155], + [ 0.3016, -0.0797, -0.3123, ..., -0.2413, -0.2888, -0.2219], + ..., + [-0.1217, -0.0503, -0.4168, ..., -0.1403, 0.1352, 0.0412], + [-0.2368, -0.1586, -0.0663, ..., -0.2495, -0.0663, -0.0784], + [-0.1203, -0.1090, -0.1677, ..., 0.1031, 0.0271, -0.1784]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-09, + 9.3132e-10, 0.0000e+00], + [-1.6764e-08, -0.0000e+00, 4.6566e-09, ..., 2.1420e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + -5.5879e-09, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0017, 0.0051, 0.0136, -0.0048, 0.0256, -0.0037, 0.0024, -0.0007, + 0.0081, -0.0181], device='cuda:0'), grad: tensor([ 2.8871e-08, 8.3819e-09, 1.5832e-08, -2.5518e-07, -1.4808e-07, + 2.2538e-07, 6.9849e-08, 1.8626e-08, 1.2107e-08, 2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 246.08, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4764 re_mapping 0.0029 re_causal 0.0094 /// teacc 99.17 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.2353, -0.0651, 0.2241, ..., 0.0412, -0.2492, -0.1742], + [-0.1983, -0.2890, -0.0774, ..., -0.1366, -0.3679, -0.1158], + [ 0.3018, -0.0797, -0.3128, ..., -0.2418, -0.2898, -0.2220], + ..., + [-0.1219, -0.0510, -0.4177, ..., -0.1404, 0.1366, 0.0416], + [-0.2370, -0.1587, -0.0664, ..., -0.2494, -0.0664, -0.0786], + [-0.1205, -0.1090, -0.1680, ..., 0.1031, 0.0270, -0.1786]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.2107e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.4459e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1176e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 364, bias, value: tensor([ 0.0020, 0.0077, 0.0133, -0.0048, 0.0256, -0.0038, 0.0024, -0.0032, + 0.0083, -0.0182], device='cuda:0'), grad: tensor([-2.7940e-09, 1.7695e-08, -5.6811e-08, -4.6566e-09, -3.7253e-09, + 4.6566e-09, 0.0000e+00, 2.3283e-08, 1.4901e-08, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 246.11, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4607 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.15 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.2359, -0.0654, 0.2241, ..., 0.0412, -0.2498, -0.1744], + [-0.1990, -0.2893, -0.0773, ..., -0.1367, -0.3687, -0.1164], + [ 0.3027, -0.0803, -0.3130, ..., -0.2426, -0.2921, -0.2238], + ..., + [-0.1224, -0.0514, -0.4189, ..., -0.1410, 0.1360, 0.0436], + [-0.2374, -0.1591, -0.0670, ..., -0.2497, -0.0654, -0.0793], + [-0.1207, -0.1090, -0.1683, ..., 0.1031, 0.0272, -0.1787]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 7.4506e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -3.7253e-09, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.0245e-08, + -1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 365, bias, value: tensor([ 0.0019, 0.0078, 0.0134, -0.0048, 0.0257, -0.0038, 0.0025, -0.0034, + 0.0086, -0.0182], device='cuda:0'), grad: tensor([ 6.3330e-08, -8.3819e-09, 1.8626e-08, 5.5879e-09, 2.1420e-08, + 6.2399e-08, -1.0058e-07, 1.3039e-08, -1.6484e-07, 8.8476e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 245.89, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.5158 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.22 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.2364, -0.0663, 0.2243, ..., 0.0410, -0.2515, -0.1752], + [-0.1996, -0.2900, -0.0779, ..., -0.1367, -0.3704, -0.1202], + [ 0.3029, -0.0806, -0.3137, ..., -0.2441, -0.2940, -0.2252], + ..., + [-0.1225, -0.0517, -0.4196, ..., -0.1416, 0.1368, 0.0466], + [-0.2385, -0.1598, -0.0691, ..., -0.2523, -0.0671, -0.0802], + [-0.1198, -0.1090, -0.1692, ..., 0.1028, 0.0270, -0.1793]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.7987e-08, + 2.0489e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -4.7497e-08, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, 4.6566e-09, ..., -2.4587e-07, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0017, 0.0078, 0.0129, -0.0048, 0.0260, -0.0040, 0.0028, -0.0034, + 0.0084, -0.0185], device='cuda:0'), grad: tensor([ 1.5832e-08, 1.4715e-07, 8.2888e-08, 9.3132e-09, 4.4517e-07, + 2.7008e-08, -1.3970e-08, -9.4064e-08, 1.8626e-08, -6.3144e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 246.30, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4835 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.15 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.2368, -0.0667, 0.2244, ..., 0.0411, -0.2520, -0.1753], + [-0.2001, -0.2903, -0.0780, ..., -0.1368, -0.3713, -0.1210], + [ 0.3033, -0.0806, -0.3141, ..., -0.2444, -0.2943, -0.2251], + ..., + [-0.1226, -0.0520, -0.4200, ..., -0.1417, 0.1371, 0.0469], + [-0.2399, -0.1600, -0.0694, ..., -0.2528, -0.0671, -0.0805], + [-0.1203, -0.1090, -0.1695, ..., 0.1027, 0.0269, -0.1795]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + 7.4506e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.7742e-08, + 3.7253e-09, -9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.1910e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.6787e-07, + 1.8440e-07, 0.0000e+00]], device='cuda:0') +Epoch 367, bias, value: tensor([ 0.0017, 0.0084, 0.0124, -0.0047, 0.0261, -0.0040, 0.0029, -0.0040, + 0.0083, -0.0186], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.5635e-08, 9.3132e-10, 6.5193e-09, -8.7544e-07, + 1.8626e-09, 1.3039e-08, 1.1548e-07, 1.1735e-07, 5.7369e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 246.61, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4498 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.23 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.2376, -0.0668, 0.2244, ..., 0.0402, -0.2537, -0.1762], + [-0.2017, -0.2904, -0.0782, ..., -0.1369, -0.3719, -0.1212], + [ 0.3059, -0.0806, -0.3146, ..., -0.2453, -0.2934, -0.2250], + ..., + [-0.1249, -0.0521, -0.4203, ..., -0.1425, 0.1358, 0.0469], + [-0.2406, -0.1599, -0.0700, ..., -0.2526, -0.0673, -0.0809], + [-0.1206, -0.1091, -0.1697, ..., 0.1028, 0.0271, -0.1795]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 1.8626e-09, 9.3132e-10], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 3.7253e-09, + 2.7940e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., -8.3819e-09, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0009, 0.0084, 0.0133, -0.0046, 0.0262, -0.0043, 0.0028, -0.0042, + 0.0100, -0.0185], device='cuda:0'), grad: tensor([ 2.6077e-08, -4.0978e-08, 2.7008e-08, -3.3528e-08, 1.9558e-08, + -1.1083e-07, 3.0734e-08, -2.7940e-09, 6.5193e-09, 5.9605e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 246.58, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4869 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.17 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.2381, -0.0669, 0.2246, ..., 0.0400, -0.2540, -0.1763], + [-0.2023, -0.2907, -0.0782, ..., -0.1370, -0.3757, -0.1204], + [ 0.3075, -0.0803, -0.3149, ..., -0.2455, -0.2938, -0.2245], + ..., + [-0.1258, -0.0525, -0.4208, ..., -0.1425, 0.1378, 0.0466], + [-0.2441, -0.1602, -0.0702, ..., -0.2528, -0.0674, -0.0812], + [-0.1213, -0.1091, -0.1708, ..., 0.1029, 0.0270, -0.1797]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.7695e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 2.5146e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0007, 0.0081, 0.0142, -0.0045, 0.0261, -0.0044, 0.0030, -0.0040, + 0.0098, -0.0186], device='cuda:0'), grad: tensor([-4.1910e-08, -1.1548e-07, 4.6566e-08, 1.8626e-09, -4.4890e-07, + 1.4901e-08, 1.3970e-08, 1.1735e-07, -6.5193e-08, 4.9360e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 246.51, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.5057 re_mapping 0.0032 re_causal 0.0101 /// teacc 99.14 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.2387, -0.0670, 0.2249, ..., 0.0402, -0.2547, -0.1763], + [-0.2035, -0.2909, -0.0782, ..., -0.1370, -0.3766, -0.1221], + [ 0.3080, -0.0807, -0.3154, ..., -0.2457, -0.2969, -0.2275], + ..., + [-0.1258, -0.0529, -0.4214, ..., -0.1426, 0.1396, 0.0496], + [-0.2445, -0.1603, -0.0707, ..., -0.2532, -0.0676, -0.0806], + [-0.1215, -0.1091, -0.1717, ..., 0.1027, 0.0269, -0.1800]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-08, 0.0000e+00, -0.0000e+00, ..., 4.6566e-09, + 1.8626e-09, 0.0000e+00], + [-1.1548e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -0.0000e+00, 0.0000e+00], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, -0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -2.8871e-08, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0008, 0.0081, 0.0133, -0.0042, 0.0263, -0.0047, 0.0031, -0.0038, + 0.0098, -0.0188], device='cuda:0'), grad: tensor([ 1.3970e-08, 1.2387e-07, -2.4866e-07, 6.0536e-08, 4.5635e-08, + 1.6764e-08, 0.0000e+00, 5.8673e-08, 1.3039e-08, -7.6368e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 246.47, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4463 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.2397, -0.0682, 0.2252, ..., 0.0404, -0.2548, -0.1763], + [-0.2040, -0.2911, -0.0784, ..., -0.1372, -0.3770, -0.1232], + [ 0.3098, -0.0806, -0.3160, ..., -0.2459, -0.2971, -0.2274], + ..., + [-0.1262, -0.0531, -0.4228, ..., -0.1427, 0.1398, 0.0498], + [-0.2477, -0.1589, -0.0713, ..., -0.2535, -0.0677, -0.0809], + [-0.1218, -0.1091, -0.1723, ..., 0.1026, 0.0268, -0.1804]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-09, ..., 3.7253e-09, + 4.6566e-10, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.3993e-08, + 4.6566e-10, 0.0000e+00], + [-3.3062e-08, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.0268e-08, 0.0000e+00, 0.0000e+00, ..., 4.7637e-07, + -9.3132e-10, 0.0000e+00], + [ 4.6566e-10, -9.3132e-10, 0.0000e+00, ..., 1.5832e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -1.1111e-06, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0008, 0.0082, 0.0143, -0.0045, 0.0264, -0.0048, 0.0031, -0.0039, + 0.0100, -0.0188], device='cuda:0'), grad: tensor([ 9.3132e-09, 9.0338e-08, -5.2154e-08, 9.7789e-09, 1.7984e-06, + -8.3819e-09, 2.7940e-08, 1.3635e-06, 3.4925e-08, -3.2652e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 246.58, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4653 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.20 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.2399, -0.0683, 0.2254, ..., 0.0403, -0.2549, -0.1763], + [-0.2045, -0.2913, -0.0785, ..., -0.1373, -0.3788, -0.1239], + [ 0.3101, -0.0804, -0.3164, ..., -0.2462, -0.2974, -0.2275], + ..., + [-0.1263, -0.0532, -0.4233, ..., -0.1430, 0.1405, 0.0499], + [-0.2478, -0.1592, -0.0717, ..., -0.2542, -0.0678, -0.0810], + [-0.1220, -0.1091, -0.1725, ..., 0.1025, 0.0268, -0.1805]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -3.2596e-09, + 6.9849e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 9.3132e-10, + 1.3504e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 9.3132e-10, + 1.3970e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 5.1223e-09, + -2.9337e-08, 0.0000e+00], + [ 4.6566e-10, -0.0000e+00, 1.5367e-08, ..., 1.3970e-09, + 6.9849e-09, 4.6566e-10], + [ 0.0000e+00, -4.6566e-10, 6.9849e-09, ..., -1.2573e-08, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 372, bias, value: tensor([ 0.0007, 0.0081, 0.0134, -0.0045, 0.0265, -0.0048, 0.0030, -0.0039, + 0.0107, -0.0189], device='cuda:0'), grad: tensor([ 4.9826e-08, 5.7276e-08, 1.7229e-08, 2.3283e-08, 3.7579e-07, + 3.9116e-08, -5.4436e-07, -9.3598e-08, 6.8918e-08, 1.8626e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 246.36, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4545 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.2401, -0.0687, 0.2255, ..., 0.0403, -0.2550, -0.1763], + [-0.2046, -0.2915, -0.0785, ..., -0.1374, -0.3796, -0.1247], + [ 0.3105, -0.0796, -0.3168, ..., -0.2466, -0.2975, -0.2275], + ..., + [-0.1264, -0.0534, -0.4235, ..., -0.1431, 0.1409, 0.0500], + [-0.2483, -0.1604, -0.0721, ..., -0.2549, -0.0680, -0.0811], + [-0.1227, -0.1090, -0.1729, ..., 0.1025, 0.0268, -0.1806]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 9.3132e-10, -1.8626e-09, ..., 6.0070e-08, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 2.9337e-08, + 1.8626e-09, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 3.7253e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 2.2072e-07, + 9.7789e-09, 4.6566e-10], + [ 9.3132e-10, 1.0710e-08, 9.3132e-10, ..., 9.1270e-08, + 9.3132e-10, 1.3970e-09], + [ 4.6566e-10, -1.7229e-08, 4.1910e-09, ..., -6.2911e-07, + -1.7229e-08, 1.3970e-09]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0008, 0.0081, 0.0135, -0.0045, 0.0265, -0.0048, 0.0030, -0.0038, + 0.0105, -0.0190], device='cuda:0'), grad: tensor([ 2.2398e-07, 2.1933e-07, 3.0547e-07, -5.7705e-06, 6.4587e-07, + 5.6922e-06, 4.6566e-09, 3.5390e-07, 3.3993e-07, -2.0135e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 246.35, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4791 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.2404, -0.0691, 0.2267, ..., 0.0406, -0.2551, -0.1763], + [-0.2051, -0.2924, -0.0799, ..., -0.1377, -0.3835, -0.1229], + [ 0.3110, -0.0802, -0.3178, ..., -0.2471, -0.2985, -0.2280], + ..., + [-0.1267, -0.0531, -0.4245, ..., -0.1432, 0.1434, 0.0506], + [-0.2485, -0.1607, -0.0735, ..., -0.2561, -0.0682, -0.0817], + [-0.1228, -0.1089, -0.1737, ..., 0.1015, 0.0266, -0.1807]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -9.1363e-07, ..., -5.4063e-07, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 4.6566e-10, ..., 3.0268e-08, + 9.3132e-09, 9.3132e-10], + [-3.7253e-09, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 2.3283e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 4.6566e-10, 9.3132e-10, ..., 1.3039e-08, + -2.1886e-08, -1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 2.5611e-08, ..., 2.6543e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.7498e-07, ..., 1.9856e-06, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0018, 0.0073, 0.0133, -0.0044, 0.0274, -0.0047, 0.0026, -0.0031, + 0.0102, -0.0198], device='cuda:0'), grad: tensor([-1.6857e-06, 8.4750e-08, 1.9092e-08, 1.5832e-08, -2.5630e-06, + 2.4633e-07, 8.4285e-08, -4.6566e-08, -2.9430e-07, 4.1351e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 246.19, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4709 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.12 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.2406, -0.0692, 0.2271, ..., 0.0408, -0.2554, -0.1763], + [-0.2057, -0.2929, -0.0799, ..., -0.1378, -0.3837, -0.1231], + [ 0.3115, -0.0807, -0.3185, ..., -0.2475, -0.2991, -0.2284], + ..., + [-0.1269, -0.0524, -0.4258, ..., -0.1434, 0.1440, 0.0511], + [-0.2487, -0.1611, -0.0754, ..., -0.2572, -0.0685, -0.0819], + [-0.1239, -0.1086, -0.1745, ..., 0.1014, 0.0265, -0.1810]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.9828e-06, ..., -1.0841e-06, + 4.6566e-10, -1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 1.2573e-08, ..., 8.3819e-09, + 4.6566e-10, 0.0000e+00], + [-2.3283e-09, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, 3.7253e-09, ..., 2.1886e-08, + -4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 1.8626e-09, 2.4214e-08, ..., 2.3283e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.8673e-08, ..., -2.1560e-07, + -5.1223e-09, 0.0000e+00]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0021, 0.0072, 0.0133, -0.0044, 0.0275, -0.0046, 0.0023, -0.0029, + 0.0100, -0.0199], device='cuda:0'), grad: tensor([-1.3568e-05, 3.3993e-08, 1.2573e-08, 1.2200e-07, 4.1164e-07, + 1.7788e-06, 1.1414e-05, 6.6590e-08, 8.1491e-08, -3.7206e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 246.30, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4551 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.10 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.2405, -0.0693, 0.2301, ..., 0.0413, -0.2552, -0.1764], + [-0.2062, -0.2934, -0.0800, ..., -0.1380, -0.3839, -0.1233], + [ 0.3120, -0.0812, -0.3211, ..., -0.2485, -0.2994, -0.2284], + ..., + [-0.1271, -0.0523, -0.4269, ..., -0.1435, 0.1442, 0.0511], + [-0.2487, -0.1611, -0.0768, ..., -0.2576, -0.0688, -0.0822], + [-0.1261, -0.1100, -0.1749, ..., 0.1011, 0.0265, -0.1810]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 9.3132e-10, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, -0.0000e+00, ..., 3.2596e-09, + 7.4506e-09, 2.3283e-09], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 9.3132e-10], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.6764e-08, + -1.2573e-08, -4.6566e-09], + [ 2.3283e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -3.8650e-08, + -1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0050, 0.0066, 0.0132, -0.0044, 0.0278, -0.0046, 0.0011, -0.0023, + 0.0099, -0.0201], device='cuda:0'), grad: tensor([ 1.3039e-08, -7.3109e-08, 1.3504e-08, 3.8184e-08, 7.9162e-08, + -1.2061e-07, 1.0896e-07, 4.6100e-08, 2.9337e-08, -1.1409e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 246.39, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4495 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.12 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.2407, -0.0695, 0.2332, ..., 0.0386, -0.2555, -0.1764], + [-0.2064, -0.2937, -0.0803, ..., -0.1383, -0.3840, -0.1236], + [ 0.3121, -0.0813, -0.3219, ..., -0.2488, -0.2995, -0.2285], + ..., + [-0.1272, -0.0527, -0.4274, ..., -0.1437, 0.1444, 0.0512], + [-0.2488, -0.1612, -0.0770, ..., -0.2577, -0.0691, -0.0823], + [-0.1262, -0.1104, -0.1752, ..., 0.1019, 0.0265, -0.1811]], + device='cuda:0'), grad: tensor([[-0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 1.3970e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + -2.7940e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8859e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0059, 0.0065, 0.0131, -0.0041, 0.0278, -0.0048, -0.0011, -0.0023, + 0.0099, -0.0196], device='cuda:0'), grad: tensor([ 2.3283e-09, -3.7253e-08, 3.2596e-09, 4.6566e-09, -4.0419e-07, + 4.3772e-08, -3.4459e-08, 2.1420e-08, 3.7253e-09, 4.0093e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4695 re_mapping 0.0027 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.2410, -0.0696, 0.2332, ..., 0.0386, -0.2557, -0.1764], + [-0.2068, -0.2941, -0.0804, ..., -0.1392, -0.3841, -0.1237], + [ 0.3125, -0.0810, -0.3223, ..., -0.2491, -0.2996, -0.2285], + ..., + [-0.1274, -0.0531, -0.4280, ..., -0.1439, 0.1446, 0.0512], + [-0.2490, -0.1615, -0.0776, ..., -0.2585, -0.0696, -0.0825], + [-0.1276, -0.1107, -0.1755, ..., 0.1008, 0.0262, -0.1811]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3027e-07, ..., 0.0000e+00, + 4.1910e-09, -6.5658e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 5.1223e-09, ..., 1.3737e-07, + 4.6566e-10, 9.3132e-10]], device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0059, 0.0064, 0.0133, -0.0042, 0.0290, -0.0044, -0.0013, -0.0024, + 0.0097, -0.0206], device='cuda:0'), grad: tensor([-8.1351e-07, 3.1665e-08, 1.0710e-08, 4.1910e-09, -2.7334e-07, + 8.0047e-07, 1.8161e-08, -6.5193e-08, -5.7742e-08, 3.4552e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 246.48, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4634 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.13 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.2412, -0.0699, 0.2337, ..., 0.0384, -0.2555, -0.1757], + [-0.2073, -0.2946, -0.0805, ..., -0.1401, -0.3843, -0.1238], + [ 0.3129, -0.0810, -0.3232, ..., -0.2496, -0.2997, -0.2285], + ..., + [-0.1275, -0.0533, -0.4305, ..., -0.1441, 0.1448, 0.0512], + [-0.2491, -0.1617, -0.0784, ..., -0.2590, -0.0697, -0.0826], + [-0.1278, -0.1107, -0.1768, ..., 0.1008, 0.0262, -0.1812]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.3749e-08, ..., -7.9162e-09, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + 4.6566e-10, 0.0000e+00], + [-8.5682e-08, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 8.1025e-08, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 7.9162e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., -1.5367e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0060, 0.0057, 0.0133, -0.0032, 0.0291, -0.0054, -0.0013, -0.0018, + 0.0096, -0.0206], device='cuda:0'), grad: tensor([-4.1444e-08, -3.3993e-08, -2.1700e-07, 4.6566e-09, 2.9337e-08, + 2.3283e-09, 1.8626e-08, 2.6776e-07, 2.5611e-08, -3.9581e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 246.63, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4514 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.12 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.2414, -0.0700, 0.2337, ..., 0.0384, -0.2556, -0.1756], + [-0.2072, -0.2948, -0.0803, ..., -0.1405, -0.3846, -0.1239], + [ 0.3131, -0.0811, -0.3236, ..., -0.2502, -0.2998, -0.2285], + ..., + [-0.1277, -0.0536, -0.4322, ..., -0.1443, 0.1452, 0.0513], + [-0.2493, -0.1618, -0.0791, ..., -0.2611, -0.0702, -0.0827], + [-0.1278, -0.1106, -0.1770, ..., 0.1007, 0.0262, -0.1815]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -1.8626e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 3.7253e-09], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.9558e-08, -8.3819e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -5.5879e-09, + 3.7253e-09, 1.8626e-09]], device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0061, 0.0058, 0.0126, -0.0031, 0.0293, -0.0055, -0.0013, -0.0018, + 0.0093, -0.0208], device='cuda:0'), grad: tensor([ 8.3819e-09, -2.6822e-07, 1.3970e-08, 2.1420e-08, 2.6077e-08, + 2.1420e-08, -6.5193e-09, 1.0617e-07, 3.6322e-08, 4.1910e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 246.45, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4482 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.15 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.2420, -0.0703, 0.2338, ..., 0.0384, -0.2559, -0.1756], + [-0.2092, -0.2952, -0.0806, ..., -0.1426, -0.3858, -0.1273], + [ 0.3140, -0.0806, -0.3240, ..., -0.2504, -0.3000, -0.2286], + ..., + [-0.1280, -0.0540, -0.4326, ..., -0.1443, 0.1470, 0.0523], + [-0.2498, -0.1625, -0.0794, ..., -0.2619, -0.0707, -0.0839], + [-0.1280, -0.1108, -0.1774, ..., 0.1006, 0.0262, -0.1825]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -3.7253e-09, + 9.3132e-10, -9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., -9.3132e-10, + 5.5879e-09, 9.3132e-10], + [-2.7940e-09, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, + -3.0734e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -4.6566e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0061, 0.0054, 0.0129, -0.0032, 0.0296, -0.0054, -0.0011, -0.0016, + 0.0089, -0.0210], device='cuda:0'), grad: tensor([-1.0245e-08, -1.6857e-07, 2.0489e-08, 2.2352e-08, 8.1956e-08, + 2.2352e-08, 3.1665e-08, -4.1910e-08, 4.9360e-08, -9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 246.13, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4206 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.17 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.2425, -0.0707, 0.2338, ..., 0.0384, -0.2562, -0.1756], + [-0.2096, -0.2956, -0.0806, ..., -0.1427, -0.3859, -0.1273], + [ 0.3145, -0.0805, -0.3241, ..., -0.2505, -0.3000, -0.2285], + ..., + [-0.1285, -0.0547, -0.4329, ..., -0.1444, 0.1474, 0.0524], + [-0.2498, -0.1627, -0.0797, ..., -0.2620, -0.0708, -0.0840], + [-0.1281, -0.1108, -0.1776, ..., 0.1005, 0.0261, -0.1826]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-8.5682e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.4703e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 8.3819e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0060, 0.0055, 0.0128, -0.0028, 0.0297, -0.0056, -0.0011, -0.0017, + 0.0088, -0.0211], device='cuda:0'), grad: tensor([ 1.7695e-08, 6.5193e-09, -1.8720e-07, 8.6613e-08, 1.5832e-08, + -2.0117e-07, 5.9605e-08, 1.1548e-07, 7.5437e-08, 1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 245.86, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4788 re_mapping 0.0031 re_causal 0.0096 /// teacc 99.10 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.2433, -0.0709, 0.2338, ..., 0.0384, -0.2567, -0.1757], + [-0.2097, -0.2959, -0.0806, ..., -0.1428, -0.3861, -0.1277], + [ 0.3149, -0.0806, -0.3244, ..., -0.2508, -0.3001, -0.2285], + ..., + [-0.1287, -0.0552, -0.4335, ..., -0.1446, 0.1477, 0.0523], + [-0.2500, -0.1628, -0.0800, ..., -0.2627, -0.0706, -0.0843], + [-0.1282, -0.1108, -0.1778, ..., 0.1005, 0.0262, -0.1827]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -2.7940e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0060, 0.0054, 0.0119, -0.0029, 0.0297, -0.0054, -0.0011, -0.0015, + 0.0090, -0.0211], device='cuda:0'), grad: tensor([ 2.7940e-08, 3.5204e-07, 3.9116e-08, 6.5193e-09, 2.7008e-08, + 4.6566e-09, 9.3132e-10, -4.2096e-07, 3.7253e-09, -2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 245.85, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4317 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.20 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.2439, -0.0711, 0.2338, ..., 0.0384, -0.2572, -0.1757], + [-0.2137, -0.2967, -0.0811, ..., -0.1431, -0.3863, -0.1283], + [ 0.3174, -0.0805, -0.3251, ..., -0.2544, -0.3002, -0.2289], + ..., + [-0.1291, -0.0554, -0.4343, ..., -0.1447, 0.1480, 0.0529], + [-0.2501, -0.1631, -0.0809, ..., -0.2639, -0.0707, -0.0845], + [-0.1286, -0.1109, -0.1781, ..., 0.1004, 0.0262, -0.1828]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 4.0047e-08, ..., 1.3970e-08, + 1.1176e-08, 2.7940e-09], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., -1.9558e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0059, 0.0050, 0.0128, -0.0031, 0.0298, -0.0054, -0.0008, -0.0014, + 0.0090, -0.0212], device='cuda:0'), grad: tensor([ 2.4214e-08, 3.7253e-09, 1.8626e-08, 2.5444e-06, 6.5193e-09, + -3.2075e-06, 5.2992e-07, 1.0245e-08, 8.2888e-08, -1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 246.00, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4789 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.17 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.2442, -0.0712, 0.2338, ..., 0.0384, -0.2573, -0.1757], + [-0.2144, -0.2972, -0.0805, ..., -0.1431, -0.3864, -0.1283], + [ 0.3183, -0.0807, -0.3257, ..., -0.2546, -0.3002, -0.2289], + ..., + [-0.1296, -0.0560, -0.4358, ..., -0.1452, 0.1478, 0.0529], + [-0.2505, -0.1633, -0.0818, ..., -0.2649, -0.0707, -0.0846], + [-0.1287, -0.1109, -0.1785, ..., 0.1005, 0.0262, -0.1828]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0059, 0.0055, 0.0129, -0.0029, 0.0297, -0.0055, -0.0007, -0.0019, + 0.0091, -0.0211], device='cuda:0'), grad: tensor([ 9.3132e-10, 6.5193e-09, -5.5879e-09, 1.3970e-08, -1.5832e-08, + -1.1176e-08, 2.0862e-07, 9.3132e-09, -2.1420e-07, 1.3970e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 246.00, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4564 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.09 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.2445, -0.0713, 0.2338, ..., 0.0384, -0.2576, -0.1757], + [-0.2145, -0.2975, -0.0801, ..., -0.1432, -0.3866, -0.1292], + [ 0.3186, -0.0811, -0.3263, ..., -0.2547, -0.3003, -0.2290], + ..., + [-0.1297, -0.0563, -0.4366, ..., -0.1453, 0.1481, 0.0532], + [-0.2507, -0.1633, -0.0825, ..., -0.2649, -0.0707, -0.0848], + [-0.1304, -0.1109, -0.1787, ..., 0.1004, 0.0261, -0.1832]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-09, 0.0000e+00, -9.3132e-10, ..., -6.5193e-09, + 6.5193e-09, -0.0000e+00], + [-2.1420e-08, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, -0.0000e+00], + ..., + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + -1.6764e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 9.3132e-10, ..., -6.5193e-09, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0058, 0.0058, 0.0128, -0.0036, 0.0298, -0.0047, -0.0007, -0.0022, + 0.0094, -0.0213], device='cuda:0'), grad: tensor([ 3.1665e-08, -1.0058e-07, -1.7695e-08, 9.9652e-08, 1.4994e-07, + -1.3132e-07, 7.5437e-08, -3.6880e-07, 2.2352e-08, 2.3935e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 245.89, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4538 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.16 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.2455, -0.0714, 0.2339, ..., 0.0384, -0.2580, -0.1758], + [-0.2147, -0.2976, -0.0785, ..., -0.1430, -0.3868, -0.1280], + [ 0.3189, -0.0811, -0.3269, ..., -0.2552, -0.3004, -0.2290], + ..., + [-0.1298, -0.0565, -0.4375, ..., -0.1461, 0.1483, 0.0534], + [-0.2508, -0.1634, -0.0839, ..., -0.2654, -0.0709, -0.0849], + [-0.1306, -0.1109, -0.1792, ..., 0.1007, 0.0262, -0.1836]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.3970e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, + -9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.0978e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.7940e-09, + 0.0000e+00, 1.3970e-08]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0058, 0.0060, 0.0126, -0.0037, 0.0295, -0.0048, -0.0010, -0.0023, + 0.0095, -0.0211], device='cuda:0'), grad: tensor([-4.1910e-08, 6.3330e-08, 2.2352e-08, 3.5390e-08, 1.6764e-08, + -5.9418e-07, 1.2573e-07, -1.9092e-07, 3.4366e-07, 2.1979e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 246.17, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4660 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.08 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.2459, -0.0714, 0.2339, ..., 0.0385, -0.2584, -0.1760], + [-0.2151, -0.2977, -0.0786, ..., -0.1430, -0.3870, -0.1284], + [ 0.3196, -0.0811, -0.3275, ..., -0.2554, -0.3004, -0.2291], + ..., + [-0.1303, -0.0567, -0.4378, ..., -0.1463, 0.1486, 0.0537], + [-0.2510, -0.1634, -0.0843, ..., -0.2656, -0.0710, -0.0853], + [-0.1311, -0.1109, -0.1794, ..., 0.1007, 0.0261, -0.1838]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, -0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0057, 0.0052, 0.0124, -0.0036, 0.0295, -0.0048, -0.0010, -0.0014, + 0.0101, -0.0211], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.4901e-08, 1.8626e-09, 1.8626e-09, -2.0489e-08, + -3.7253e-09, 4.6566e-09, 1.3970e-08, 0.0000e+00, 1.3970e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 246.15, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4448 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.04 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.2471, -0.0715, 0.2352, ..., 0.0395, -0.2585, -0.1759], + [-0.2155, -0.2978, -0.0786, ..., -0.1433, -0.3870, -0.1286], + [ 0.3204, -0.0811, -0.3288, ..., -0.2557, -0.3004, -0.2291], + ..., + [-0.1308, -0.0574, -0.4382, ..., -0.1468, 0.1486, 0.0538], + [-0.2511, -0.1635, -0.0846, ..., -0.2667, -0.0711, -0.0853], + [-0.1313, -0.1110, -0.1818, ..., 0.1003, 0.0262, -0.1839]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7229e-07, ..., -1.1362e-07, + 1.8626e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 9.3132e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1921e-07, ..., 6.0536e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0068, 0.0051, 0.0128, -0.0036, 0.0294, -0.0048, -0.0011, -0.0014, + 0.0102, -0.0214], device='cuda:0'), grad: tensor([-4.0047e-07, 1.6764e-08, 8.3819e-09, 4.5635e-08, 1.5832e-08, + -3.9116e-08, 2.7940e-09, 2.7008e-08, 4.3772e-08, 2.8592e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 246.05, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4711 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.08 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.2477, -0.0716, 0.2365, ..., 0.0409, -0.2595, -0.1760], + [-0.2157, -0.2981, -0.0786, ..., -0.1435, -0.3871, -0.1286], + [ 0.3211, -0.0810, -0.3299, ..., -0.2561, -0.3005, -0.2291], + ..., + [-0.1314, -0.0583, -0.4388, ..., -0.1472, 0.1487, 0.0539], + [-0.2512, -0.1636, -0.0849, ..., -0.2679, -0.0712, -0.0859], + [-0.1325, -0.1109, -0.1835, ..., 0.0998, 0.0262, -0.1834]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [-5.7742e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.7695e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -0.0000e+00, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.4214e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0082, 0.0049, 0.0131, -0.0039, 0.0294, -0.0046, -0.0011, -0.0012, + 0.0101, -0.0220], device='cuda:0'), grad: tensor([ 1.2107e-08, -1.1176e-08, -1.1269e-07, 7.5437e-08, 6.2399e-08, + -3.9116e-08, 7.4506e-09, 4.3772e-08, 9.3132e-09, -4.3772e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 246.21, cls_loss 0.0004 cls_loss_mapping 0.0016 cls_loss_causal 0.4525 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.08 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.2500, -0.0717, 0.2364, ..., 0.0408, -0.2600, -0.1770], + [-0.2159, -0.2982, -0.0785, ..., -0.1436, -0.3872, -0.1288], + [ 0.3215, -0.0812, -0.3302, ..., -0.2564, -0.3005, -0.2291], + ..., + [-0.1317, -0.0584, -0.4393, ..., -0.1473, 0.1488, 0.0538], + [-0.2512, -0.1636, -0.0848, ..., -0.2680, -0.0709, -0.0860], + [-0.1329, -0.1109, -0.1835, ..., 0.0997, 0.0262, -0.1838]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., -2.7940e-09, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 6.5193e-09, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 9.3132e-10, 3.7253e-09, ..., 5.5879e-09, + -9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 4.0047e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.6077e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0079, 0.0049, 0.0133, -0.0040, 0.0296, -0.0048, -0.0008, -0.0013, + 0.0107, -0.0221], device='cuda:0'), grad: tensor([-3.3528e-08, -2.0415e-05, 1.0617e-07, 9.3132e-09, -3.3714e-07, + 2.4214e-08, 1.2107e-08, 2.0280e-05, 1.7881e-07, 1.4435e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 246.29, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4735 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.13 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.2519, -0.0717, 0.2338, ..., 0.0360, -0.2606, -0.1787], + [-0.2159, -0.2983, -0.0789, ..., -0.1438, -0.3877, -0.1298], + [ 0.3218, -0.0813, -0.3307, ..., -0.2569, -0.3007, -0.2295], + ..., + [-0.1319, -0.0586, -0.4400, ..., -0.1475, 0.1495, 0.0553], + [-0.2513, -0.1636, -0.0852, ..., -0.2687, -0.0710, -0.0867], + [-0.1332, -0.1110, -0.1808, ..., 0.1043, 0.0267, -0.1855]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.4459e-08, ..., 1.9558e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 9.3132e-10, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1420e-08, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., -5.8673e-08, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0034, 0.0050, 0.0132, -0.0042, 0.0286, -0.0045, -0.0006, -0.0014, + 0.0108, -0.0179], device='cuda:0'), grad: tensor([ 2.4214e-08, 2.7008e-08, 1.3039e-08, 2.1607e-07, -7.4506e-08, + -8.4005e-07, 5.6531e-07, 8.0094e-08, 3.3528e-08, -4.6566e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 246.29, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4634 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.11 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.2527, -0.0717, 0.2337, ..., 0.0357, -0.2607, -0.1786], + [-0.2160, -0.2984, -0.0789, ..., -0.1440, -0.3878, -0.1300], + [ 0.3220, -0.0814, -0.3317, ..., -0.2574, -0.3008, -0.2297], + ..., + [-0.1322, -0.0587, -0.4403, ..., -0.1478, 0.1498, 0.0558], + [-0.2515, -0.1637, -0.0861, ..., -0.2691, -0.0711, -0.0877], + [-0.1333, -0.1110, -0.1806, ..., 0.1046, 0.0267, -0.1869]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -5.5879e-08, ..., 5.3085e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 2.6077e-08, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + -3.2596e-08, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 3.7253e-09, 4.6566e-09], + [ 2.7940e-09, 0.0000e+00, 5.4017e-08, ..., -2.3004e-07, + 1.8626e-09, 2.7940e-09]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0031, 0.0050, 0.0131, -0.0041, 0.0286, -0.0046, -0.0006, -0.0014, + 0.0107, -0.0176], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.2573e-07, 1.0245e-08, 1.7136e-07, 4.7125e-07, + -3.4738e-07, 1.2480e-07, -1.4901e-07, 6.0536e-08, -5.3830e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 246.44, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4572 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.14 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.2538, -0.0719, 0.2337, ..., 0.0357, -0.2611, -0.1786], + [-0.2161, -0.2989, -0.0788, ..., -0.1443, -0.3881, -0.1301], + [ 0.3226, -0.0814, -0.3335, ..., -0.2584, -0.3009, -0.2298], + ..., + [-0.1325, -0.0591, -0.4406, ..., -0.1482, 0.1500, 0.0560], + [-0.2521, -0.1638, -0.0862, ..., -0.2697, -0.0712, -0.0879], + [-0.1335, -0.1110, -0.1806, ..., 0.1046, 0.0267, -0.1873]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.7695e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0031, 0.0050, 0.0132, -0.0038, 0.0287, -0.0047, -0.0005, -0.0015, + 0.0106, -0.0176], device='cuda:0'), grad: tensor([ 4.6566e-09, -5.7742e-08, 7.4506e-09, 2.9802e-08, -6.7987e-08, + -1.3039e-08, -1.5832e-08, 4.0978e-08, -1.0245e-08, 8.1025e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 246.38, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4630 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.2545, -0.0721, 0.2337, ..., 0.0357, -0.2616, -0.1784], + [-0.2163, -0.2993, -0.0790, ..., -0.1448, -0.3886, -0.1307], + [ 0.3235, -0.0833, -0.3353, ..., -0.2593, -0.3010, -0.2287], + ..., + [-0.1332, -0.0597, -0.4411, ..., -0.1490, 0.1508, 0.0564], + [-0.2525, -0.1638, -0.0872, ..., -0.2700, -0.0716, -0.0877], + [-0.1337, -0.1110, -0.1807, ..., 0.1046, 0.0266, -0.1891]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -6.6124e-08, ..., -2.2352e-08, + 9.3132e-10, -2.0489e-08], + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, + 2.7940e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -3.2596e-08, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0031, 0.0050, 0.0134, -0.0029, 0.0289, -0.0049, -0.0004, -0.0015, + 0.0107, -0.0176], device='cuda:0'), grad: tensor([-1.0151e-07, -2.1420e-08, 7.4506e-09, -6.2399e-08, 5.9605e-08, + 1.8813e-07, -8.0094e-08, 4.7497e-08, 3.0734e-08, -6.8918e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 246.03, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.5023 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.11 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.2550, -0.0726, 0.2337, ..., 0.0357, -0.2621, -0.1786], + [-0.2188, -0.3003, -0.0790, ..., -0.1439, -0.3891, -0.1308], + [ 0.3258, -0.0844, -0.3359, ..., -0.2594, -0.3012, -0.2289], + ..., + [-0.1335, -0.0608, -0.4417, ..., -0.1503, 0.1525, 0.0568], + [-0.2529, -0.1646, -0.0876, ..., -0.2707, -0.0718, -0.0880], + [-0.1342, -0.1111, -0.1807, ..., 0.1046, 0.0264, -0.1898]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.7940e-09, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0031, 0.0052, 0.0149, -0.0027, 0.0289, -0.0049, -0.0004, -0.0017, + 0.0109, -0.0176], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.8626e-09, -4.6566e-09, -1.2107e-08, -1.0245e-08, + 4.6566e-09, 9.3132e-10, 1.0245e-08, 1.8626e-09, 1.2107e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4716 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.10 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.2567, -0.0728, 0.2338, ..., 0.0357, -0.2626, -0.1786], + [-0.2189, -0.3006, -0.0791, ..., -0.1443, -0.3896, -0.1316], + [ 0.3261, -0.0851, -0.3364, ..., -0.2601, -0.3014, -0.2290], + ..., + [-0.1338, -0.0607, -0.4421, ..., -0.1509, 0.1533, 0.0581], + [-0.2531, -0.1650, -0.0883, ..., -0.2741, -0.0720, -0.0886], + [-0.1347, -0.1111, -0.1807, ..., 0.1046, 0.0263, -0.1908]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-08, ..., -1.3970e-08, + 0.0000e+00, -1.8626e-09], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1176e-08, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0031, 0.0056, 0.0148, -0.0028, 0.0282, -0.0047, -0.0002, -0.0021, + 0.0103, -0.0176], device='cuda:0'), grad: tensor([-1.5553e-07, -2.7940e-09, -1.7695e-08, 0.0000e+00, -4.6566e-09, + 1.2107e-08, 7.5437e-08, 3.2596e-08, 3.5390e-08, 2.9802e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 246.26, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4759 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.23 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.2567, -0.0730, 0.2338, ..., 0.0357, -0.2629, -0.1786], + [-0.2190, -0.3009, -0.0791, ..., -0.1445, -0.3897, -0.1317], + [ 0.3263, -0.0855, -0.3372, ..., -0.2606, -0.3015, -0.2290], + ..., + [-0.1343, -0.0607, -0.4422, ..., -0.1511, 0.1507, 0.0581], + [-0.2533, -0.1653, -0.0886, ..., -0.2751, -0.0692, -0.0888], + [-0.1358, -0.1111, -0.1807, ..., 0.1046, 0.0263, -0.1911]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -1.8626e-09, + 0.0000e+00, -0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0031, 0.0056, 0.0148, -0.0034, 0.0279, -0.0041, -0.0004, -0.0042, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-7.4506e-09, 0.0000e+00, -9.3132e-10, -4.4703e-08, -2.7940e-09, + 2.6077e-08, 0.0000e+00, 1.0245e-08, 4.6566e-09, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 246.32, cls_loss 0.0003 cls_loss_mapping 0.0006 cls_loss_causal 0.4783 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.10 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.2568, -0.0730, 0.2338, ..., 0.0357, -0.2634, -0.1787], + [-0.2190, -0.3011, -0.0791, ..., -0.1445, -0.3899, -0.1320], + [ 0.3268, -0.0855, -0.3378, ..., -0.2610, -0.3016, -0.2291], + ..., + [-0.1349, -0.0609, -0.4424, ..., -0.1512, 0.1505, 0.0585], + [-0.2535, -0.1654, -0.0902, ..., -0.2756, -0.0690, -0.0891], + [-0.1360, -0.1112, -0.1807, ..., 0.1046, 0.0263, -0.1915]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 3.4459e-08, ..., 0.0000e+00, + 1.9558e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.4901e-08, -9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + ..., + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 2.5146e-08, ..., 1.8626e-09, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0031, 0.0056, 0.0149, -0.0034, 0.0279, -0.0041, -0.0003, -0.0043, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 6.3330e-08, 6.5193e-09, -3.0734e-08, 1.8626e-09, -6.5193e-09, + 1.8626e-09, -1.0990e-07, 1.3970e-08, 5.4948e-08, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 246.87, cls_loss 0.0003 cls_loss_mapping 0.0007 cls_loss_causal 0.4623 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.09 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.2571, -0.0730, 0.2338, ..., 0.0357, -0.2635, -0.1788], + [-0.2191, -0.3012, -0.0791, ..., -0.1446, -0.3900, -0.1320], + [ 0.3275, -0.0855, -0.3379, ..., -0.2612, -0.3016, -0.2287], + ..., + [-0.1358, -0.0611, -0.4427, ..., -0.1531, 0.1505, 0.0581], + [-0.2536, -0.1654, -0.0905, ..., -0.2757, -0.0690, -0.0892], + [-0.1360, -0.1112, -0.1808, ..., 0.1046, 0.0264, -0.1916]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-08, ..., -1.0245e-08, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 1.3970e-09, ..., 9.3132e-10, + 7.9162e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + -4.2841e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 2.4214e-08, ..., 1.2480e-07, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0031, 0.0056, 0.0154, -0.0035, 0.0280, -0.0041, -0.0002, -0.0045, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([-6.2864e-08, 4.4703e-08, 8.8476e-09, 5.0757e-08, -2.1048e-07, + -2.1886e-08, 2.4680e-08, -1.2573e-07, -1.4435e-08, 3.1060e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 247.07, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4547 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.2572, -0.0731, 0.2339, ..., 0.0357, -0.2638, -0.1785], + [-0.2192, -0.3014, -0.0794, ..., -0.1449, -0.3903, -0.1322], + [ 0.3278, -0.0856, -0.3385, ..., -0.2616, -0.3019, -0.2290], + ..., + [-0.1361, -0.0625, -0.4433, ..., -0.1545, 0.1503, 0.0589], + [-0.2535, -0.1655, -0.0912, ..., -0.2762, -0.0690, -0.0895], + [-0.1370, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1929]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -1.3970e-09, ..., -9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-10], + [-3.9116e-08, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.2596e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 4.6566e-10], + [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -3.2596e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 401, bias, value: tensor([ 3.0899e-03, 5.6636e-03, 1.5365e-02, -4.5598e-03, 2.8849e-02, + -3.6655e-03, -6.2230e-05, -4.6025e-03, 1.3379e-02, -1.7585e-02], + device='cuda:0'), grad: tensor([ 2.7940e-09, -1.3504e-08, -6.3330e-08, -1.3970e-08, 2.2817e-08, + 2.3283e-08, 6.5658e-08, 1.1129e-07, -1.4808e-07, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 246.58, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4189 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.15 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.2572, -0.0731, 0.2339, ..., 0.0357, -0.2638, -0.1785], + [-0.2192, -0.3015, -0.0794, ..., -0.1449, -0.3904, -0.1322], + [ 0.3278, -0.0856, -0.3386, ..., -0.2616, -0.3019, -0.2290], + ..., + [-0.1361, -0.0626, -0.4433, ..., -0.1545, 0.1503, 0.0589], + [-0.2536, -0.1656, -0.0912, ..., -0.2763, -0.0690, -0.0895], + [-0.1370, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1929]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, -2.6077e-08, ..., -1.8161e-08, + 4.6566e-10, 4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 4.6566e-09, ..., 3.2596e-09, + 4.6566e-10, 4.6566e-10], + [-2.8871e-08, 0.0000e+00, 4.1910e-09, ..., 2.3283e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 8.8476e-09, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + 4.6566e-10, -1.3970e-09], + [ 9.3132e-10, -0.0000e+00, 3.7253e-09, ..., 2.3283e-09, + -3.2596e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 4.1910e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 402, bias, value: tensor([ 3.0888e-03, 5.6677e-03, 1.5326e-02, -4.4647e-03, 2.8847e-02, + -3.7790e-03, -4.4756e-05, -4.6102e-03, 1.3386e-02, -1.7584e-02], + device='cuda:0'), grad: tensor([-6.8452e-08, 2.3283e-09, -6.5658e-08, 2.5146e-08, 3.2596e-09, + 2.3749e-08, 1.3970e-08, 5.2154e-08, -1.3970e-08, 2.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 245.89, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4372 re_mapping 0.0024 re_causal 0.0080 /// teacc 99.14 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2639, -0.1784], + [-0.2192, -0.3015, -0.0795, ..., -0.1449, -0.3904, -0.1322], + [ 0.3279, -0.0857, -0.3385, ..., -0.2616, -0.3019, -0.2290], + ..., + [-0.1361, -0.0626, -0.4434, ..., -0.1545, 0.1503, 0.0590], + [-0.2536, -0.1656, -0.0913, ..., -0.2763, -0.0690, -0.0895], + [-0.1370, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1930]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.4261e-08, ..., -1.8626e-09, + 9.3132e-10, -2.6077e-08], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 2.3283e-09, + 2.2817e-08, 9.3132e-09], + [-1.8626e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.6566e-10, 9.3132e-10, 4.6566e-10, ..., 4.1910e-09, + -2.7940e-08, -1.1642e-08], + [ 1.8626e-09, 4.6566e-10, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 4.6566e-09, ..., 5.5879e-09, + 1.3970e-09, 2.3283e-09]], device='cuda:0') +Epoch 403, bias, value: tensor([ 3.0889e-03, 5.6687e-03, 1.5227e-02, -4.1460e-03, 2.8843e-02, + -4.0865e-03, -4.0300e-05, -4.6057e-03, 1.3388e-02, -1.7584e-02], + device='cuda:0'), grad: tensor([-1.4668e-07, 1.1455e-07, 3.7253e-09, -3.9581e-08, -4.6100e-08, + 1.1502e-07, 6.0536e-08, -1.2154e-07, 1.3504e-08, 6.4261e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 246.63, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4439 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.14 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2640, -0.1785], + [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3904, -0.1322], + [ 0.3279, -0.0857, -0.3385, ..., -0.2617, -0.3019, -0.2290], + ..., + [-0.1362, -0.0627, -0.4434, ..., -0.1545, 0.1503, 0.0590], + [-0.2536, -0.1656, -0.0913, ..., -0.2763, -0.0690, -0.0894], + [-0.1371, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1931]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., -2.3283e-09, + 1.1176e-08, 1.0245e-08], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [-6.8918e-08, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 2.7940e-09, 2.7940e-09], + ..., + [ 7.2643e-08, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-08, -1.6764e-08], + [ 3.2596e-09, 6.0536e-09, 9.3132e-10, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 9.3132e-10, ..., 2.3283e-09, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 404, bias, value: tensor([ 3.0889e-03, 5.6740e-03, 1.5234e-02, -4.1376e-03, 2.8839e-02, + -4.1107e-03, -1.0064e-05, -4.6119e-03, 1.3391e-02, -1.7584e-02], + device='cuda:0'), grad: tensor([ 5.2154e-08, 1.1967e-07, -6.5658e-08, -1.4994e-07, 1.8626e-09, + 3.4459e-08, -9.3132e-10, -1.8766e-07, 9.0804e-08, 1.0896e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 246.09, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4582 re_mapping 0.0024 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2640, -0.1785], + [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3904, -0.1322], + [ 0.3279, -0.0857, -0.3385, ..., -0.2617, -0.3019, -0.2290], + ..., + [-0.1362, -0.0628, -0.4434, ..., -0.1546, 0.1503, 0.0591], + [-0.2536, -0.1656, -0.0913, ..., -0.2763, -0.0690, -0.0894], + [-0.1372, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1932]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.7789e-09, ..., -5.5879e-09, + 0.0000e+00, -9.3132e-10], + [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, -0.0000e+00], + [-8.8476e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-10], + ..., + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 3.2596e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 405, bias, value: tensor([ 3.0889e-03, 5.6746e-03, 1.5228e-02, -4.0974e-03, 2.8843e-02, + -4.1383e-03, -2.2912e-05, -4.6139e-03, 1.3391e-02, -1.7584e-02], + device='cuda:0'), grad: tensor([-2.0955e-08, -1.0394e-06, 5.5879e-09, 6.0536e-08, 1.6298e-08, + -4.6566e-09, 1.3504e-08, 9.6392e-07, 8.8476e-09, 1.4435e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 246.23, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4607 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.16 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2640, -0.1785], + [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3905, -0.1323], + [ 0.3280, -0.0857, -0.3386, ..., -0.2619, -0.3019, -0.2288], + ..., + [-0.1364, -0.0628, -0.4435, ..., -0.1546, 0.1503, 0.0589], + [-0.2536, -0.1656, -0.0914, ..., -0.2764, -0.0690, -0.0895], + [-0.1374, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1932]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.3970e-09, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + 4.6566e-10, -4.6566e-10], + [-1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-09, 1.3970e-09, 0.0000e+00, ..., 1.1642e-08, + 9.3132e-10, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 4.2841e-08, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 406, bias, value: tensor([ 3.0890e-03, 5.6753e-03, 1.5285e-02, -4.0789e-03, 2.8846e-02, + -4.1272e-03, -2.5131e-05, -4.6234e-03, 1.3391e-02, -1.7584e-02], + device='cuda:0'), grad: tensor([ 3.0268e-08, -1.8626e-09, 1.1176e-08, -3.6787e-08, -2.9057e-07, + -3.9116e-08, 8.1956e-08, 7.9628e-08, 3.6787e-08, 1.4156e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 246.31, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4252 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.15 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1785], + [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3905, -0.1323], + [ 0.3281, -0.0857, -0.3386, ..., -0.2619, -0.3019, -0.2288], + ..., + [-0.1365, -0.0629, -0.4435, ..., -0.1546, 0.1503, 0.0589], + [-0.2537, -0.1656, -0.0914, ..., -0.2764, -0.0690, -0.0894], + [-0.1374, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1933]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., -4.6566e-09, + 0.0000e+00, -0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-09, -1.9558e-08], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + ..., + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.3970e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([ 3.0889e-03, 5.6765e-03, 1.5312e-02, -4.0540e-03, 2.8864e-02, + -4.1526e-03, -2.6312e-05, -4.6281e-03, 1.3393e-02, -1.7584e-02], + device='cuda:0'), grad: tensor([-2.4214e-08, -2.3004e-07, 3.7253e-08, 9.3132e-10, 1.1176e-08, + 2.9802e-08, 7.4506e-08, 8.4750e-08, -8.3819e-09, 3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 246.78, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4331 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.15 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.2574, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], + [-0.2192, -0.3016, -0.0796, ..., -0.1450, -0.3906, -0.1323], + [ 0.3282, -0.0857, -0.3387, ..., -0.2619, -0.3020, -0.2288], + ..., + [-0.1366, -0.0629, -0.4435, ..., -0.1546, 0.1503, 0.0589], + [-0.2537, -0.1656, -0.0915, ..., -0.2764, -0.0690, -0.0894], + [-0.1374, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1933]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., -4.6566e-09, + 0.0000e+00, -3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 3.7253e-09], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + ..., + [-2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + -1.5832e-08, -1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 408, bias, value: tensor([ 3.0890e-03, 5.6792e-03, 1.5310e-02, -4.0430e-03, 2.8865e-02, + -4.1699e-03, -1.0583e-05, -4.6305e-03, 1.3394e-02, -1.7585e-02], + device='cuda:0'), grad: tensor([-3.1665e-08, 4.0978e-08, 1.3039e-08, 3.7253e-09, 5.2154e-08, + -8.3819e-09, 2.4214e-08, -1.1176e-07, 1.8626e-09, 2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 246.23, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4672 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.15 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.2574, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], + [-0.2192, -0.3016, -0.0796, ..., -0.1450, -0.3906, -0.1323], + [ 0.3282, -0.0857, -0.3386, ..., -0.2619, -0.3020, -0.2288], + ..., + [-0.1367, -0.0629, -0.4435, ..., -0.1546, 0.1504, 0.0590], + [-0.2537, -0.1656, -0.0915, ..., -0.2765, -0.0690, -0.0894], + [-0.1376, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1934]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., -5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.0245e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 3.0734e-08, + 2.1607e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 4.6566e-09, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([ 3.0890e-03, 5.7177e-03, 1.5336e-02, -4.0331e-03, 2.8868e-02, + -4.1572e-03, -1.1627e-05, -4.6578e-03, 1.3392e-02, -1.7585e-02], + device='cuda:0'), grad: tensor([-1.4901e-08, 4.5635e-08, 9.3132e-10, -1.2508e-06, 0.0000e+00, + 1.7881e-07, -8.1025e-08, 1.0785e-06, 1.3970e-08, 3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 246.67, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4375 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.19 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.2574, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], + [-0.2193, -0.3017, -0.0796, ..., -0.1451, -0.3907, -0.1324], + [ 0.3284, -0.0857, -0.3386, ..., -0.2619, -0.3020, -0.2287], + ..., + [-0.1368, -0.0629, -0.4436, ..., -0.1547, 0.1504, 0.0589], + [-0.2538, -0.1656, -0.0915, ..., -0.2765, -0.0690, -0.0894], + [-0.1376, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1935]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, -0.0000e+00], + [ 9.3132e-10, 1.8626e-09, -1.8626e-09, ..., -9.3132e-10, + 9.3132e-10, -9.3132e-10], + [-1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-09, 5.5879e-09, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, 9.3132e-10], + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -6.5193e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 410, bias, value: tensor([ 3.0891e-03, 5.7314e-03, 1.5391e-02, -4.0099e-03, 2.8875e-02, + -4.1750e-03, -2.3204e-05, -4.6684e-03, 1.3391e-02, -1.7585e-02], + device='cuda:0'), grad: tensor([-9.3132e-10, -1.0245e-08, 1.9558e-08, -1.0990e-07, 1.3970e-08, + 2.8871e-08, 4.6566e-09, 2.6077e-08, 2.3283e-08, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 246.56, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4500 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.20 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.2575, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], + [-0.2193, -0.3017, -0.0797, ..., -0.1451, -0.3907, -0.1325], + [ 0.3284, -0.0857, -0.3386, ..., -0.2619, -0.3020, -0.2287], + ..., + [-0.1369, -0.0629, -0.4436, ..., -0.1547, 0.1504, 0.0590], + [-0.2538, -0.1656, -0.0916, ..., -0.2765, -0.0690, -0.0894], + [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1935]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.2165e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, -0.0000e+00], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.5053e-07, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 411, bias, value: tensor([ 3.0892e-03, 5.7306e-03, 1.5395e-02, -3.9917e-03, 2.8873e-02, + -4.1854e-03, -3.4837e-05, -4.6672e-03, 1.3390e-02, -1.7585e-02], + device='cuda:0'), grad: tensor([ 7.0222e-07, 3.7253e-09, -2.7940e-09, 1.6764e-08, 4.8429e-08, + -1.3970e-08, 1.8626e-09, 2.8871e-08, 4.6566e-09, -7.8045e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 246.40, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4305 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.19 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.2575, -0.0731, 0.2339, ..., 0.0357, -0.2642, -0.1784], + [-0.2193, -0.3017, -0.0798, ..., -0.1451, -0.3908, -0.1325], + [ 0.3285, -0.0857, -0.3387, ..., -0.2620, -0.3020, -0.2287], + ..., + [-0.1369, -0.0630, -0.4436, ..., -0.1547, 0.1504, 0.0590], + [-0.2538, -0.1656, -0.0916, ..., -0.2766, -0.0690, -0.0894], + [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1936]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.3085e-08, ..., -3.3528e-08, + 0.0000e+00, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 3.7253e-09, -0.0000e+00], + [-3.7253e-09, -0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + -6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 8.9407e-08, + 1.8626e-09, 2.7940e-09]], device='cuda:0') +Epoch 412, bias, value: tensor([ 3.0893e-03, 5.7400e-03, 1.5396e-02, -4.0086e-03, 2.8873e-02, + -4.1713e-03, -3.4275e-05, -4.6747e-03, 1.3390e-02, -1.7585e-02], + device='cuda:0'), grad: tensor([-1.4529e-07, -3.2596e-08, -1.8626e-09, 3.9116e-08, -2.0023e-07, + -3.0734e-08, 6.6124e-08, 3.5390e-08, 1.1176e-08, 2.6450e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 246.76, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4451 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.17 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.2575, -0.0731, 0.2339, ..., 0.0357, -0.2642, -0.1784], + [-0.2193, -0.3017, -0.0798, ..., -0.1451, -0.3909, -0.1326], + [ 0.3285, -0.0857, -0.3387, ..., -0.2620, -0.3020, -0.2287], + ..., + [-0.1370, -0.0630, -0.4436, ..., -0.1547, 0.1504, 0.0590], + [-0.2538, -0.1656, -0.0917, ..., -0.2766, -0.0690, -0.0894], + [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1936]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 0.0000e+00, + 8.3819e-09, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, -0.0000e+00], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 413, bias, value: tensor([ 3.0893e-03, 5.7327e-03, 1.5421e-02, -4.0923e-03, 2.8888e-02, + -4.0947e-03, -2.6335e-05, -4.6718e-03, 1.3388e-02, -1.7586e-02], + device='cuda:0'), grad: tensor([ 5.6811e-08, -5.3085e-08, -5.5879e-09, 1.9558e-08, -6.7055e-08, + -6.6124e-08, -2.6077e-08, 4.9360e-08, 2.0489e-08, 7.6368e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 246.59, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4111 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.2575, -0.0731, 0.2340, ..., 0.0357, -0.2642, -0.1784], + [-0.2193, -0.3017, -0.0798, ..., -0.1451, -0.3910, -0.1325], + [ 0.3286, -0.0857, -0.3387, ..., -0.2620, -0.3020, -0.2286], + ..., + [-0.1371, -0.0630, -0.4437, ..., -0.1547, 0.1504, 0.0589], + [-0.2539, -0.1656, -0.0917, ..., -0.2766, -0.0690, -0.0894], + [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1937]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0245e-08, ..., -7.4506e-09, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + 5.4948e-08, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -5.7742e-08, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, -9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 3.7253e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 414, bias, value: tensor([ 3.0893e-03, 5.7312e-03, 1.5461e-02, -4.0985e-03, 2.8898e-02, + -4.0841e-03, -2.6809e-05, -4.6730e-03, 1.3386e-02, -1.7586e-02], + device='cuda:0'), grad: tensor([-2.7008e-08, 1.2098e-06, 0.0000e+00, 1.2107e-08, 9.3132e-09, + 1.3970e-08, 0.0000e+00, -1.2405e-06, 2.1420e-08, 2.8871e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 246.50, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4214 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.23 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.2575, -0.0731, 0.2340, ..., 0.0357, -0.2642, -0.1784], + [-0.2193, -0.3017, -0.0799, ..., -0.1451, -0.3911, -0.1326], + [ 0.3287, -0.0857, -0.3387, ..., -0.2620, -0.3021, -0.2286], + ..., + [-0.1372, -0.0630, -0.4437, ..., -0.1547, 0.1504, 0.0589], + [-0.2539, -0.1657, -0.0918, ..., -0.2767, -0.0690, -0.0895], + [-0.1379, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1937]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -4.1910e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 415, bias, value: tensor([ 3.0893e-03, 5.7281e-03, 1.5516e-02, -4.0983e-03, 2.8898e-02, + -4.0690e-03, -9.1731e-06, -4.6711e-03, 1.3385e-02, -1.7586e-02], + device='cuda:0'), grad: tensor([ 1.1176e-08, 1.3970e-08, -5.5879e-09, 5.5879e-08, 8.1025e-08, + -6.7987e-08, -2.8871e-08, 2.6077e-08, 5.4017e-08, -1.2107e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 246.75, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4306 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.21 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.2575, -0.0731, 0.2340, ..., 0.0357, -0.2643, -0.1784], + [-0.2193, -0.3017, -0.0799, ..., -0.1452, -0.3911, -0.1326], + [ 0.3287, -0.0857, -0.3387, ..., -0.2621, -0.3021, -0.2286], + ..., + [-0.1372, -0.0631, -0.4437, ..., -0.1547, 0.1505, 0.0590], + [-0.2540, -0.1657, -0.0918, ..., -0.2767, -0.0690, -0.0895], + [-0.1379, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1938]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 3.7253e-09, + 0.0000e+00, -4.6566e-09], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [-1.1176e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -1.4901e-08, + -0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 416, bias, value: tensor([ 3.0894e-03, 5.7204e-03, 1.5551e-02, -4.1018e-03, 2.8903e-02, + -4.0634e-03, -3.9225e-06, -4.6679e-03, 1.3382e-02, -1.7586e-02], + device='cuda:0'), grad: tensor([-1.6764e-08, 1.0245e-08, -2.6543e-07, 1.1176e-08, 9.3132e-10, + -2.0489e-08, 3.1665e-08, 2.4866e-07, 2.6077e-08, -2.1420e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 246.63, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4529 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.16 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2643, -0.1784], + [-0.2193, -0.3017, -0.0799, ..., -0.1452, -0.3912, -0.1326], + [ 0.3288, -0.0857, -0.3388, ..., -0.2621, -0.3021, -0.2286], + ..., + [-0.1373, -0.0631, -0.4437, ..., -0.1547, 0.1505, 0.0590], + [-0.2540, -0.1657, -0.0919, ..., -0.2767, -0.0690, -0.0895], + [-0.1379, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1938]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -9.3132e-10, + 9.3132e-10, 1.8626e-09], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1176e-08, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., -9.3132e-10, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 417, bias, value: tensor([ 3.0894e-03, 5.7139e-03, 1.5585e-02, -4.1038e-03, 2.8905e-02, + -4.0572e-03, 7.7366e-06, -4.6640e-03, 1.3378e-02, -1.7586e-02], + device='cuda:0'), grad: tensor([ 4.6566e-09, -2.4214e-08, -1.2107e-08, -2.2352e-08, 1.1176e-08, + 2.0489e-08, -2.3283e-08, 6.2399e-08, -2.1420e-08, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 246.66, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4284 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2643, -0.1784], + [-0.2193, -0.3017, -0.0800, ..., -0.1452, -0.3913, -0.1327], + [ 0.3289, -0.0857, -0.3388, ..., -0.2621, -0.3021, -0.2286], + ..., + [-0.1374, -0.0631, -0.4438, ..., -0.1548, 0.1505, 0.0590], + [-0.2541, -0.1657, -0.0920, ..., -0.2768, -0.0690, -0.0895], + [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1938]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -3.7253e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.2573e-07, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 418, bias, value: tensor([ 3.0894e-03, 5.7157e-03, 1.5603e-02, -4.1047e-03, 2.8909e-02, + -4.0366e-03, 1.5412e-05, -4.6663e-03, 1.3377e-02, -1.7587e-02], + device='cuda:0'), grad: tensor([ 3.7253e-08, 2.0489e-08, 7.4506e-09, 3.3807e-07, -3.5111e-07, + -3.7346e-07, 1.1176e-08, -7.4506e-09, 5.5879e-09, 3.2783e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 246.86, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4226 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.17 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2644, -0.1784], + [-0.2194, -0.3017, -0.0800, ..., -0.1452, -0.3913, -0.1327], + [ 0.3289, -0.0857, -0.3389, ..., -0.2622, -0.3021, -0.2286], + ..., + [-0.1374, -0.0631, -0.4438, ..., -0.1548, 0.1505, 0.0590], + [-0.2541, -0.1657, -0.0920, ..., -0.2768, -0.0690, -0.0895], + [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1939]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 9.3132e-10, + 1.0245e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.4214e-08, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 419, bias, value: tensor([ 3.0894e-03, 5.7150e-03, 1.5618e-02, -4.1364e-03, 2.8925e-02, + -3.9985e-03, 7.3957e-06, -4.6687e-03, 1.3374e-02, -1.7587e-02], + device='cuda:0'), grad: tensor([ 6.8918e-08, 5.5879e-09, 0.0000e+00, 7.4506e-09, 3.2596e-08, + -7.4506e-09, -7.1712e-08, 2.5146e-08, 9.3132e-10, -6.9849e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 246.31, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4228 re_mapping 0.0022 re_causal 0.0077 /// teacc 99.19 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2644, -0.1784], + [-0.2194, -0.3017, -0.0801, ..., -0.1453, -0.3914, -0.1328], + [ 0.3290, -0.0857, -0.3389, ..., -0.2623, -0.3021, -0.2285], + ..., + [-0.1376, -0.0631, -0.4438, ..., -0.1548, 0.1505, 0.0590], + [-0.2541, -0.1657, -0.0921, ..., -0.2769, -0.0690, -0.0895], + [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1939]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 8.3819e-09, 0.0000e+00], + [-2.7940e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.2352e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -1.7695e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.4506e-09, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 420, bias, value: tensor([ 3.0895e-03, 5.7084e-03, 1.5689e-02, -4.1365e-03, 2.8935e-02, + -4.0021e-03, 1.2542e-05, -4.6701e-03, 1.3374e-02, -1.7587e-02], + device='cuda:0'), grad: tensor([ 3.7253e-09, 4.0978e-08, -5.4017e-08, 1.6764e-08, 1.7695e-08, + -1.8626e-09, 7.3574e-08, -2.4214e-08, -8.6613e-08, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 246.92, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4243 re_mapping 0.0021 re_causal 0.0078 /// teacc 99.21 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2645, -0.1784], + [-0.2194, -0.3017, -0.0801, ..., -0.1453, -0.3915, -0.1328], + [ 0.3291, -0.0857, -0.3389, ..., -0.2623, -0.3021, -0.2285], + ..., + [-0.1376, -0.0632, -0.4438, ..., -0.1548, 0.1505, 0.0590], + [-0.2542, -0.1657, -0.0922, ..., -0.2769, -0.0690, -0.0895], + [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1940]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -2.5146e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 421, bias, value: tensor([ 3.0895e-03, 5.7111e-03, 1.5691e-02, -4.1334e-03, 2.8948e-02, + -4.0063e-03, 5.6771e-05, -4.6741e-03, 1.3371e-02, -1.7587e-02], + device='cuda:0'), grad: tensor([-2.5146e-08, -1.8626e-09, -1.4901e-08, 9.3132e-10, 6.4261e-08, + 6.5193e-09, -9.3132e-10, 1.8626e-08, 2.7940e-09, -5.6811e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 246.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3887 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2645, -0.1784], + [-0.2194, -0.3017, -0.0801, ..., -0.1453, -0.3915, -0.1328], + [ 0.3291, -0.0857, -0.3390, ..., -0.2624, -0.3022, -0.2285], + ..., + [-0.1377, -0.0632, -0.4439, ..., -0.1549, 0.1506, 0.0590], + [-0.2542, -0.1657, -0.0923, ..., -0.2770, -0.0690, -0.0895], + [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1940]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, -0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -2.7940e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0031, 0.0057, 0.0157, -0.0041, 0.0289, -0.0040, 0.0001, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.7695e-08, -1.9558e-08, 5.5879e-09, 1.8626e-09, + -1.8626e-08, 2.0489e-08, 2.7940e-09, -1.2387e-07, 1.1921e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 246.18, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4038 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], + [-0.2194, -0.3018, -0.0801, ..., -0.1453, -0.3916, -0.1329], + [ 0.3292, -0.0857, -0.3390, ..., -0.2624, -0.3022, -0.2284], + ..., + [-0.1378, -0.0632, -0.4439, ..., -0.1549, 0.1506, 0.0589], + [-0.2542, -0.1657, -0.0924, ..., -0.2770, -0.0690, -0.0896], + [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.5437e-08, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.0094e-08, + -1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.2643e-08, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0031, 0.0057, 0.0157, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 2.0489e-08, -1.1073e-06, 4.9360e-08, 1.8626e-09, 3.6415e-07, + 1.3039e-08, 4.6566e-09, 7.5810e-07, 2.7940e-09, -1.0431e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 246.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4227 re_mapping 0.0021 re_causal 0.0076 /// teacc 99.22 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], + [-0.2194, -0.3018, -0.0801, ..., -0.1454, -0.3917, -0.1329], + [ 0.3294, -0.0857, -0.3391, ..., -0.2625, -0.3022, -0.2283], + ..., + [-0.1380, -0.0632, -0.4439, ..., -0.1549, 0.1506, 0.0588], + [-0.2542, -0.1657, -0.0925, ..., -0.2771, -0.0690, -0.0896], + [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.7253e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-8.3819e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + ..., + [ 5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + -9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -6.5193e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0031, 0.0057, 0.0158, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([-3.7253e-09, 3.7253e-09, -1.4901e-08, 2.7940e-09, -2.7940e-09, + -9.3132e-10, 5.5879e-09, 1.2107e-08, 1.5832e-08, -1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 246.62, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], + [-0.2194, -0.3018, -0.0802, ..., -0.1454, -0.3917, -0.1329], + [ 0.3295, -0.0857, -0.3391, ..., -0.2625, -0.3022, -0.2283], + ..., + [-0.1381, -0.0633, -0.4440, ..., -0.1550, 0.1506, 0.0588], + [-0.2542, -0.1657, -0.0926, ..., -0.2771, -0.0690, -0.0896], + [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [-4.6566e-09, -0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + -0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -2.1420e-08, + -2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0031, 0.0057, 0.0158, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 1.4901e-08, 5.5879e-09, 3.7253e-09, 3.7253e-09, 1.1176e-08, + 1.8626e-09, 1.1176e-08, 2.8871e-08, 7.4506e-09, -7.0781e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 246.11, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3980 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.17 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], + [-0.2195, -0.3018, -0.0802, ..., -0.1454, -0.3918, -0.1329], + [ 0.3295, -0.0857, -0.3392, ..., -0.2625, -0.3022, -0.2283], + ..., + [-0.1381, -0.0633, -0.4440, ..., -0.1550, 0.1506, 0.0588], + [-0.2543, -0.1658, -0.0927, ..., -0.2772, -0.0690, -0.0896], + [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -9.3132e-09, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -5.5879e-09, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0031, 0.0057, 0.0158, -0.0041, 0.0290, -0.0040, 0.0001, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.7940e-08, -1.2107e-08, -6.5193e-09, 6.5193e-09, + 2.2352e-08, -1.3970e-08, -9.4064e-08, 1.5832e-08, 5.4017e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 246.26, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4118 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.20 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.2578, -0.0731, 0.2340, ..., 0.0357, -0.2647, -0.1784], + [-0.2195, -0.3018, -0.0802, ..., -0.1454, -0.3918, -0.1329], + [ 0.3296, -0.0857, -0.3392, ..., -0.2625, -0.3023, -0.2282], + ..., + [-0.1382, -0.0633, -0.4440, ..., -0.1550, 0.1506, 0.0588], + [-0.2543, -0.1658, -0.0928, ..., -0.2773, -0.0690, -0.0896], + [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1942]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 0.0000e+00, -9.3132e-10, ..., -4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3970e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.1910e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0031, 0.0057, 0.0159, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 2.3749e-08, 1.3970e-09, -2.9802e-08, -6.6124e-08, 9.7789e-09, + 6.9384e-08, -1.3970e-08, 5.1223e-09, 1.3504e-08, -7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 246.71, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4215 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.2578, -0.0731, 0.2340, ..., 0.0357, -0.2647, -0.1784], + [-0.2195, -0.3018, -0.0803, ..., -0.1455, -0.3919, -0.1329], + [ 0.3297, -0.0857, -0.3393, ..., -0.2626, -0.3023, -0.2281], + ..., + [-0.1384, -0.0634, -0.4441, ..., -0.1551, 0.1506, 0.0587], + [-0.2544, -0.1658, -0.0930, ..., -0.2774, -0.0690, -0.0897], + [-0.1381, -0.1114, -0.1809, ..., 0.1046, 0.0260, -0.1942]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.1025e-08, ..., -5.1223e-08, + 1.3970e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.3283e-09, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., 2.7940e-09, + -9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 3.2596e-09, ..., 2.3283e-09, + 3.7253e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 3.1199e-08, ..., 7.9162e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, + 0.0134, -0.0176], device='cuda:0'), grad: tensor([-2.6915e-07, 1.2107e-08, 1.1176e-08, 2.3283e-09, 3.7719e-08, + 1.7602e-07, -7.4971e-08, 1.1176e-08, 2.8871e-08, 6.7521e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 245.81, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.3944 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.2578, -0.0731, 0.2341, ..., 0.0357, -0.2647, -0.1784], + [-0.2195, -0.3018, -0.0804, ..., -0.1455, -0.3920, -0.1330], + [ 0.3298, -0.0858, -0.3394, ..., -0.2626, -0.3023, -0.2281], + ..., + [-0.1384, -0.0634, -0.4441, ..., -0.1551, 0.1507, 0.0587], + [-0.2544, -0.1658, -0.0931, ..., -0.2775, -0.0691, -0.0897], + [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1942]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.4435e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 5.5879e-09, 8.7544e-08, 6.9849e-09, 2.5146e-08, 5.1223e-09, + -1.3504e-08, -6.0536e-09, -1.2945e-07, 1.1176e-08, 1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 245.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4130 re_mapping 0.0020 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.2578, -0.0731, 0.2341, ..., 0.0357, -0.2648, -0.1784], + [-0.2195, -0.3018, -0.0804, ..., -0.1455, -0.3921, -0.1330], + [ 0.3298, -0.0858, -0.3394, ..., -0.2627, -0.3023, -0.2281], + ..., + [-0.1384, -0.0634, -0.4442, ..., -0.1551, 0.1507, 0.0587], + [-0.2544, -0.1658, -0.0932, ..., -0.2775, -0.0691, -0.0897], + [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1943]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -9.3132e-10, + 1.8626e-09, 1.8626e-09], + [ 1.7229e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 3.7253e-09, 4.1910e-09], + [-2.7474e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 1.3970e-09], + ..., + [ 1.0710e-08, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + -2.2352e-08, -2.2817e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 6.9849e-09, + 4.6566e-10, 1.3970e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.2573e-08, 1.8440e-07, -4.8429e-08, 1.4435e-07, 3.0268e-08, + -5.7556e-07, 4.6426e-07, -2.6077e-07, 3.3528e-08, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 245.88, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4063 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.21 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.2578, -0.0731, 0.2341, ..., 0.0357, -0.2648, -0.1784], + [-0.2195, -0.3019, -0.0805, ..., -0.1455, -0.3922, -0.1330], + [ 0.3299, -0.0858, -0.3395, ..., -0.2627, -0.3023, -0.2281], + ..., + [-0.1385, -0.0635, -0.4442, ..., -0.1551, 0.1507, 0.0587], + [-0.2544, -0.1658, -0.0933, ..., -0.2776, -0.0691, -0.0898], + [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1943]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-09, ..., -9.3132e-10, + 0.0000e+00, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.9849e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 6.5193e-09, ..., -9.3132e-10, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-7.4506e-09, 7.9162e-09, 2.3283e-09, -1.0151e-07, -3.3993e-08, + 3.9116e-08, 3.1665e-08, 4.1910e-09, 4.0978e-08, 2.0955e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 246.03, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4198 re_mapping 0.0021 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2649, -0.1784], + [-0.2195, -0.3019, -0.0805, ..., -0.1456, -0.3923, -0.1330], + [ 0.3299, -0.0858, -0.3396, ..., -0.2628, -0.3024, -0.2281], + ..., + [-0.1385, -0.0635, -0.4442, ..., -0.1552, 0.1507, 0.0588], + [-0.2545, -0.1658, -0.0934, ..., -0.2776, -0.0691, -0.0898], + [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1943]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -0.0000e+00], + [-2.2817e-08, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 1.5367e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0031, 0.0058, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 5.5879e-09, -9.7789e-09, -4.3772e-08, 3.2596e-09, 1.2573e-08, + 1.1176e-08, -1.3970e-08, 4.2375e-08, 5.5879e-09, -1.3970e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 246.18, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4277 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.22 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2649, -0.1784], + [-0.2196, -0.3019, -0.0805, ..., -0.1456, -0.3925, -0.1330], + [ 0.3301, -0.0858, -0.3396, ..., -0.2628, -0.3024, -0.2280], + ..., + [-0.1388, -0.0635, -0.4442, ..., -0.1552, 0.1508, 0.0586], + [-0.2545, -0.1658, -0.0934, ..., -0.2777, -0.0691, -0.0898], + [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.6298e-08, ..., -1.1642e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 3.2596e-09, ..., 1.3970e-09, + 1.3970e-09, 0.0000e+00], + [-2.3283e-08, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-08, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + -2.7940e-09, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 7.9162e-09, ..., 1.3644e-07, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0031, 0.0058, 0.0161, -0.0041, 0.0290, -0.0041, 0.0003, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-4.5169e-08, -9.1270e-08, -5.2154e-08, 7.9162e-09, -2.9569e-07, + 5.1223e-09, 6.5193e-09, 1.1316e-07, 1.0710e-08, 3.5344e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 246.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4178 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2650, -0.1784], + [-0.2196, -0.3019, -0.0806, ..., -0.1456, -0.3926, -0.1330], + [ 0.3302, -0.0858, -0.3397, ..., -0.2629, -0.3024, -0.2280], + ..., + [-0.1388, -0.0635, -0.4443, ..., -0.1553, 0.1508, 0.0586], + [-0.2545, -0.1658, -0.0935, ..., -0.2778, -0.0691, -0.0899], + [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7695e-08, ..., 9.6858e-08, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, + -1.3970e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., -1.2945e-07, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0290, -0.0041, 0.0004, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.7649e-07, 1.2573e-08, 2.3283e-09, 3.7253e-09, 6.7521e-08, + -9.3132e-10, 7.4506e-09, 9.3132e-09, 2.4680e-08, -2.8266e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 246.03, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4126 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2650, -0.1784], + [-0.2196, -0.3019, -0.0806, ..., -0.1456, -0.3927, -0.1330], + [ 0.3303, -0.0858, -0.3397, ..., -0.2629, -0.3024, -0.2280], + ..., + [-0.1388, -0.0636, -0.4443, ..., -0.1553, 0.1508, 0.0586], + [-0.2546, -0.1658, -0.0936, ..., -0.2779, -0.0691, -0.0899], + [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 8.8476e-09, 9.3132e-10], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + -3.3062e-08, -1.3970e-09], + [ 4.6566e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., -1.3970e-09, + 2.0489e-08, 0.0000e+00]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0040, 0.0291, -0.0041, 0.0004, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.5367e-08, 3.2131e-08, -1.5367e-08, -7.0315e-08, 1.1642e-08, + 2.8405e-08, 9.3132e-10, -1.0803e-07, 2.8871e-08, 7.9628e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 246.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4203 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.2580, -0.0731, 0.2341, ..., 0.0357, -0.2652, -0.1785], + [-0.2196, -0.3019, -0.0806, ..., -0.1457, -0.3928, -0.1331], + [ 0.3303, -0.0858, -0.3398, ..., -0.2629, -0.3025, -0.2280], + ..., + [-0.1389, -0.0636, -0.4444, ..., -0.1554, 0.1509, 0.0586], + [-0.2546, -0.1658, -0.0938, ..., -0.2780, -0.0691, -0.0899], + [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0040, 0.0291, -0.0041, 0.0005, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-3.7253e-09, 2.3283e-09, 4.6566e-10, -1.8626e-09, -1.7229e-08, + 4.1910e-09, 7.9162e-09, 6.9849e-09, 3.2596e-09, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 246.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4230 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.2580, -0.0731, 0.2341, ..., 0.0357, -0.2652, -0.1785], + [-0.2196, -0.3019, -0.0806, ..., -0.1457, -0.3930, -0.1331], + [ 0.3304, -0.0858, -0.3398, ..., -0.2630, -0.3025, -0.2279], + ..., + [-0.1389, -0.0636, -0.4444, ..., -0.1555, 0.1509, 0.0586], + [-0.2546, -0.1659, -0.0938, ..., -0.2781, -0.0691, -0.0899], + [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1945]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.0245e-08, + 2.3283e-09, -9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 7.9162e-09], + ..., + [ 1.3970e-09, 0.0000e+00, 1.8626e-09, ..., 1.3970e-09, + -9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 5.1223e-09, ..., 4.6566e-10, + 2.3283e-09, -6.6124e-08], + [ 0.0000e+00, 0.0000e+00, 1.3504e-08, ..., -1.7695e-08, + 4.6566e-10, 9.3132e-10]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0040, 0.0291, -0.0041, 0.0005, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-4.3306e-08, -5.8394e-07, 1.4482e-07, -4.0047e-08, 6.6124e-08, + 9.8813e-07, 8.3819e-08, 5.5926e-07, -1.1902e-06, 2.0023e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 245.94, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4444 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.2580, -0.0731, 0.2341, ..., 0.0357, -0.2653, -0.1785], + [-0.2197, -0.3019, -0.0806, ..., -0.1457, -0.3931, -0.1331], + [ 0.3304, -0.0858, -0.3399, ..., -0.2630, -0.3025, -0.2279], + ..., + [-0.1389, -0.0636, -0.4444, ..., -0.1555, 0.1509, 0.0586], + [-0.2547, -0.1659, -0.0939, ..., -0.2783, -0.0691, -0.0899], + [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1945]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -8.8476e-09, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [-3.8650e-08, 0.0000e+00, 9.3132e-10, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-08, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -9.3132e-10, 0.0000e+00], + [ 6.9849e-09, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3039e-08, 0.0000e+00, 6.9849e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0291, -0.0041, 0.0005, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-1.8161e-08, 1.5832e-08, -8.2888e-08, 2.4214e-08, 1.1642e-08, + -1.9558e-08, 2.3283e-09, 3.6787e-08, 2.7474e-08, 1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 246.07, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3979 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.21 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.2581, -0.0731, 0.2341, ..., 0.0357, -0.2653, -0.1785], + [-0.2197, -0.3019, -0.0807, ..., -0.1458, -0.3932, -0.1332], + [ 0.3305, -0.0858, -0.3400, ..., -0.2631, -0.3025, -0.2280], + ..., + [-0.1390, -0.0637, -0.4444, ..., -0.1556, 0.1510, 0.0587], + [-0.2547, -0.1659, -0.0940, ..., -0.2784, -0.0691, -0.0900], + [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0258, -0.1945]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 2.4214e-08, ..., 9.3132e-09, + 0.0000e+00, 2.3283e-09], + [ 1.3970e-09, 1.3970e-09, -2.6543e-08, ..., -5.1223e-09, + 0.0000e+00, -2.3283e-09], + [-2.3749e-08, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.9558e-08, 4.6566e-09, 9.3132e-10, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-08, + 0.0000e+00, 0.0000e+00], + [-2.7940e-09, 4.6566e-10, 0.0000e+00, ..., -7.1246e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0291, -0.0041, 0.0005, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.0710e-07, -8.0094e-08, -6.1002e-08, -1.6391e-07, 1.2806e-07, + 8.7079e-08, 8.8476e-09, 1.3830e-07, 9.5461e-08, -2.5565e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 245.89, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4197 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2654, -0.1785], + [-0.2197, -0.3020, -0.0807, ..., -0.1458, -0.3933, -0.1332], + [ 0.3305, -0.0858, -0.3401, ..., -0.2631, -0.3026, -0.2280], + ..., + [-0.1390, -0.0637, -0.4445, ..., -0.1557, 0.1510, 0.0587], + [-0.2547, -0.1659, -0.0941, ..., -0.2785, -0.0691, -0.0900], + [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0258, -0.1945]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.6298e-08, + 4.6566e-10, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3749e-08, + 1.8626e-09, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.5146e-08, + -6.9849e-09, 0.0000e+00]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0291, -0.0041, 0.0005, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.8626e-08, -5.6345e-08, 6.5193e-09, 9.3132e-09, -6.1467e-08, + 1.3039e-08, 1.4435e-08, 1.3132e-07, 2.5146e-08, -7.6368e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 245.97, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4420 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2654, -0.1785], + [-0.2197, -0.3020, -0.0807, ..., -0.1459, -0.3934, -0.1332], + [ 0.3307, -0.0858, -0.3402, ..., -0.2631, -0.3026, -0.2279], + ..., + [-0.1391, -0.0637, -0.4445, ..., -0.1558, 0.1511, 0.0586], + [-0.2547, -0.1659, -0.0942, ..., -0.2786, -0.0691, -0.0900], + [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0258, -0.1946]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0048, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 2.7940e-09, -6.9849e-09, 3.7253e-09, 4.6566e-10, 1.8626e-09, + 7.9162e-09, -9.3132e-10, 6.0536e-09, -1.3039e-08, 1.2107e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 246.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4306 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.17 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2655, -0.1785], + [-0.2197, -0.3020, -0.0808, ..., -0.1459, -0.3935, -0.1332], + [ 0.3307, -0.0858, -0.3402, ..., -0.2632, -0.3027, -0.2279], + ..., + [-0.1392, -0.0637, -0.4445, ..., -0.1559, 0.1511, 0.0586], + [-0.2548, -0.1659, -0.0943, ..., -0.2787, -0.0691, -0.0900], + [-0.1384, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1946]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [-1.1642e-08, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + ..., + [ 1.1642e-08, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.3970e-09, -1.3039e-08, -2.5611e-08, 1.3970e-09, -3.7253e-09, + 2.1886e-08, -1.1642e-08, 3.0734e-08, -1.3504e-08, 1.5367e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 246.07, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4056 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2655, -0.1784], + [-0.2197, -0.3020, -0.0808, ..., -0.1460, -0.3936, -0.1332], + [ 0.3307, -0.0858, -0.3403, ..., -0.2632, -0.3027, -0.2279], + ..., + [-0.1392, -0.0638, -0.4446, ..., -0.1559, 0.1511, 0.0587], + [-0.2548, -0.1659, -0.0943, ..., -0.2788, -0.0691, -0.0900], + [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1946]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., -9.3132e-10, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-09, 4.6566e-10, 9.3132e-10, ..., 1.3970e-09, + 1.8626e-09, 0.0000e+00], + [-1.1642e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + -2.0955e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -1.7695e-08, + 1.9092e-08, 0.0000e+00]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.2107e-08, 1.2107e-08, -4.6566e-09, -8.3819e-09, 4.3772e-08, + 2.3283e-09, -2.5146e-08, -5.6811e-08, -6.0536e-09, 3.0734e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 246.29, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4119 re_mapping 0.0020 re_causal 0.0077 /// teacc 99.19 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.2582, -0.0731, 0.2343, ..., 0.0357, -0.2655, -0.1784], + [-0.2198, -0.3020, -0.0809, ..., -0.1460, -0.3936, -0.1333], + [ 0.3308, -0.0858, -0.3404, ..., -0.2633, -0.3027, -0.2279], + ..., + [-0.1392, -0.0638, -0.4446, ..., -0.1560, 0.1511, 0.0587], + [-0.2548, -0.1659, -0.0944, ..., -0.2788, -0.0691, -0.0900], + [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1947]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.7940e-09, + 5.1223e-09, 2.7940e-09], + [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 2.7940e-09, + 9.3132e-10, 4.6566e-10], + [-1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-08, + -9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 8.8476e-09, ..., 2.3283e-09, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 4.5169e-08, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 6.0536e-09, 1.4435e-08, -1.3970e-09, 2.0489e-08, -2.5285e-07, + -1.1176e-08, -4.4238e-08, 8.9873e-08, 2.4214e-08, 1.6438e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 246.37, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4123 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.18 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.2582, -0.0732, 0.2343, ..., 0.0357, -0.2656, -0.1784], + [-0.2198, -0.3020, -0.0809, ..., -0.1461, -0.3937, -0.1333], + [ 0.3309, -0.0858, -0.3405, ..., -0.2634, -0.3027, -0.2279], + ..., + [-0.1393, -0.0638, -0.4447, ..., -0.1561, 0.1511, 0.0587], + [-0.2548, -0.1659, -0.0944, ..., -0.2789, -0.0691, -0.0900], + [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1947]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + -9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0031, 0.0059, 0.0163, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 3.2596e-09, -2.7940e-09, -2.7940e-09, -1.7229e-08, 2.1886e-08, + 1.3504e-08, -7.4506e-09, 8.8476e-09, 5.1223e-09, -1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 245.96, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3998 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.2582, -0.0732, 0.2343, ..., 0.0357, -0.2656, -0.1784], + [-0.2199, -0.3020, -0.0810, ..., -0.1461, -0.3938, -0.1333], + [ 0.3310, -0.0858, -0.3405, ..., -0.2636, -0.3027, -0.2279], + ..., + [-0.1394, -0.0638, -0.4447, ..., -0.1562, 0.1512, 0.0587], + [-0.2549, -0.1659, -0.0945, ..., -0.2791, -0.0691, -0.0901], + [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1947]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-9.7789e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + ..., + [ 6.0536e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + 4.6566e-10, 4.6566e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 1.3970e-09, ..., -6.5193e-09, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0031, 0.0059, 0.0163, -0.0041, 0.0292, -0.0042, 0.0005, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-10, -2.9337e-08, -2.7474e-08, -6.9849e-09, 1.3504e-08, + -4.1910e-09, 1.1642e-08, 4.6566e-08, 8.3819e-09, -1.0710e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 246.24, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4244 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.2583, -0.0732, 0.2343, ..., 0.0357, -0.2657, -0.1784], + [-0.2199, -0.3020, -0.0810, ..., -0.1462, -0.3939, -0.1333], + [ 0.3311, -0.0858, -0.3406, ..., -0.2636, -0.3028, -0.2279], + ..., + [-0.1395, -0.0638, -0.4447, ..., -0.1563, 0.1512, 0.0587], + [-0.2549, -0.1659, -0.0946, ..., -0.2792, -0.0691, -0.0901], + [-0.1386, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1948]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -7.2177e-09, ..., -2.0955e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., -1.3970e-09, + 2.3283e-09, 2.3283e-10], + [-5.5879e-09, -9.3132e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.2596e-09, 2.3283e-10, 2.3283e-10, ..., 4.6566e-10, + -1.2340e-08, -1.6298e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.8208e-09, ..., -2.0955e-09, + 8.1491e-09, 1.1642e-09]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0031, 0.0059, 0.0163, -0.0041, 0.0292, -0.0042, 0.0006, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-6.9849e-10, -4.0280e-08, -6.9849e-09, 3.4692e-08, 2.5611e-08, + -2.9802e-08, 7.2177e-09, -6.5193e-09, 8.8476e-09, 2.7707e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 246.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4281 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.18 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.2583, -0.0732, 0.2343, ..., 0.0357, -0.2657, -0.1784], + [-0.2199, -0.3020, -0.0810, ..., -0.1462, -0.3940, -0.1333], + [ 0.3312, -0.0858, -0.3406, ..., -0.2636, -0.3028, -0.2279], + ..., + [-0.1395, -0.0639, -0.4448, ..., -0.1564, 0.1512, 0.0587], + [-0.2550, -0.1659, -0.0947, ..., -0.2793, -0.0691, -0.0901], + [-0.1386, -0.1114, -0.1810, ..., 0.1046, 0.0257, -0.1948]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0151e-07, ..., -4.1444e-08, + 4.6566e-10, -6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 2.1188e-08, ..., 1.1176e-08, + 1.8626e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 7.9162e-09, + 1.6298e-09, 6.9849e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 9.3132e-10, ..., 6.2864e-09, + -3.2596e-09, -2.0955e-09], + [ 0.0000e+00, 0.0000e+00, 4.8894e-09, ..., 6.0536e-09, + 6.9849e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.5157e-08, ..., -1.4435e-08, + -3.7253e-09, 2.3283e-10]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0041, 0.0292, -0.0042, 0.0005, -0.0049, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-2.2491e-07, 6.6590e-08, 5.4482e-08, 3.4925e-09, 6.3796e-08, + 8.6147e-09, 4.9127e-08, -3.7253e-09, 2.8871e-08, -2.9104e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 246.65, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4175 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.15 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.2583, -0.0732, 0.2344, ..., 0.0357, -0.2658, -0.1784], + [-0.2200, -0.3020, -0.0811, ..., -0.1462, -0.3940, -0.1333], + [ 0.3313, -0.0858, -0.3407, ..., -0.2637, -0.3028, -0.2279], + ..., + [-0.1396, -0.0639, -0.4448, ..., -0.1565, 0.1513, 0.0586], + [-0.2550, -0.1660, -0.0948, ..., -0.2793, -0.0691, -0.0901], + [-0.1386, -0.1114, -0.1810, ..., 0.1046, 0.0257, -0.1948]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + -0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., -0.0000e+00, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0041, 0.0292, -0.0042, 0.0006, -0.0050, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.6298e-09, 1.2619e-07, 2.5146e-08, -2.3283e-09, 1.3970e-09, + 7.2177e-09, 3.4925e-09, -1.3015e-07, -2.0256e-08, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 246.33, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3970 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.2583, -0.0732, 0.2344, ..., 0.0357, -0.2659, -0.1784], + [-0.2200, -0.3020, -0.0811, ..., -0.1462, -0.3941, -0.1333], + [ 0.3314, -0.0858, -0.3408, ..., -0.2638, -0.3028, -0.2279], + ..., + [-0.1397, -0.0639, -0.4449, ..., -0.1566, 0.1513, 0.0586], + [-0.2550, -0.1660, -0.0949, ..., -0.2794, -0.0691, -0.0901], + [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1949]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5611e-09, ..., 2.5611e-09, + 4.6566e-10, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 9.3132e-10, ..., 5.8208e-09, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.0955e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 3.9581e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -7.2177e-09, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0040, 0.0292, -0.0043, 0.0007, -0.0050, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 4.1910e-09, -7.2643e-08, 5.3551e-09, 1.8626e-08, -6.5193e-08, + 9.5461e-09, 1.1874e-08, 9.3365e-08, 1.2573e-08, -2.0955e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 245.83, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3925 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.2584, -0.0732, 0.2344, ..., 0.0357, -0.2659, -0.1784], + [-0.2200, -0.3020, -0.0811, ..., -0.1463, -0.3942, -0.1334], + [ 0.3315, -0.0858, -0.3409, ..., -0.2638, -0.3029, -0.2279], + ..., + [-0.1397, -0.0640, -0.4449, ..., -0.1567, 0.1513, 0.0586], + [-0.2551, -0.1660, -0.0950, ..., -0.2795, -0.0691, -0.0901], + [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1949]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 2.3283e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + -3.0268e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.5611e-09, ..., 2.3283e-10, + 1.1642e-09, 2.3283e-10]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0039, 0.0293, -0.0044, 0.0007, -0.0050, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-2.5611e-09, 9.0804e-09, 6.7521e-09, 1.6298e-09, 6.5193e-09, + 3.0268e-09, 3.2596e-09, -3.5856e-08, 1.8626e-09, 2.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 245.95, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4114 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.2584, -0.0732, 0.2344, ..., 0.0357, -0.2660, -0.1784], + [-0.2200, -0.3020, -0.0812, ..., -0.1463, -0.3942, -0.1334], + [ 0.3316, -0.0858, -0.3410, ..., -0.2638, -0.3029, -0.2279], + ..., + [-0.1398, -0.0640, -0.4450, ..., -0.1568, 0.1513, 0.0586], + [-0.2551, -0.1660, -0.0951, ..., -0.2796, -0.0691, -0.0901], + [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1949]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 0.0000e+00, -5.5879e-09, ..., -3.4925e-09, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 3.7253e-09, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + -4.6566e-09, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 3.2596e-09, ..., 2.7940e-09, + 6.9849e-10, 0.0000e+00]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0031, 0.0060, 0.0164, -0.0040, 0.0293, -0.0044, 0.0008, -0.0050, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-1.3737e-08, 2.0955e-08, 2.7940e-09, 0.0000e+00, -2.3283e-10, + -2.5611e-09, 2.7940e-09, -1.9791e-08, -1.8859e-08, 2.9569e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 245.81, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4426 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.22 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.2584, -0.0732, 0.2345, ..., 0.0357, -0.2661, -0.1785], + [-0.2201, -0.3021, -0.0812, ..., -0.1464, -0.3944, -0.1334], + [ 0.3316, -0.0858, -0.3410, ..., -0.2639, -0.3029, -0.2279], + ..., + [-0.1399, -0.0640, -0.4450, ..., -0.1569, 0.1514, 0.0586], + [-0.2552, -0.1660, -0.0952, ..., -0.2797, -0.0691, -0.0901], + [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1950]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [-1.3737e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.2340e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 1.8626e-09, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0031, 0.0060, 0.0164, -0.0040, 0.0294, -0.0043, 0.0007, -0.0050, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 2.3283e-10, -4.8894e-09, -2.7241e-08, -3.2363e-08, 4.6566e-10, + 3.5390e-08, 1.8626e-09, 3.3993e-08, 2.7940e-09, 9.5461e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 245.88, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4301 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.2584, -0.0732, 0.2345, ..., 0.0357, -0.2661, -0.1784], + [-0.2201, -0.3021, -0.0813, ..., -0.1464, -0.3945, -0.1334], + [ 0.3318, -0.0858, -0.3411, ..., -0.2641, -0.3029, -0.2278], + ..., + [-0.1400, -0.0641, -0.4451, ..., -0.1570, 0.1514, 0.0586], + [-0.2552, -0.1660, -0.0953, ..., -0.2798, -0.0691, -0.0901], + [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1950]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -6.5193e-09, ..., -3.9581e-09, + 0.0000e+00, -2.3283e-10], + [ 1.3970e-09, 4.6566e-10, 4.6566e-10, ..., 6.9849e-10, + 2.3283e-10, 2.3283e-10], + [-1.2107e-08, -2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.3551e-09, 9.3132e-10, 2.3283e-10, ..., 3.0268e-09, + 6.9849e-10, 0.0000e+00], + [ 8.3819e-09, 1.1642e-09, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, -6.9849e-10], + [ 2.3283e-10, 2.3283e-10, 6.5193e-09, ..., -2.2119e-08, + -6.9849e-10, 4.6566e-10]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0031, 0.0060, 0.0164, -0.0039, 0.0294, -0.0043, 0.0007, -0.0050, + 0.0133, -0.0176], device='cuda:0'), grad: tensor([-1.2107e-08, 7.4506e-09, -2.8871e-08, -6.3563e-08, 5.3085e-08, + 4.0745e-08, 6.9849e-10, 3.7486e-08, 1.8394e-08, -3.7951e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 245.87, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4589 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.2585, -0.0732, 0.2346, ..., 0.0357, -0.2663, -0.1785], + [-0.2201, -0.3021, -0.0813, ..., -0.1465, -0.3946, -0.1335], + [ 0.3319, -0.0858, -0.3412, ..., -0.2642, -0.3030, -0.2278], + ..., + [-0.1402, -0.0641, -0.4452, ..., -0.1572, 0.1515, 0.0586], + [-0.2553, -0.1660, -0.0955, ..., -0.2799, -0.0691, -0.0902], + [-0.1388, -0.1114, -0.1812, ..., 0.1046, 0.0256, -0.1950]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 2.3283e-10, -5.0478e-07, ..., -1.2945e-07, + 1.6298e-09, -4.0047e-08], + [ 1.3970e-09, 4.6566e-10, 2.5611e-09, ..., 9.3132e-10, + 2.3283e-10, 2.3283e-10], + [-1.1642e-09, 2.3283e-10, 1.4668e-08, ..., 3.7253e-09, + 4.6566e-10, 1.1642e-09], + ..., + [ 3.0268e-09, 1.1642e-09, 1.1642e-09, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 9.3132e-10, 2.3283e-10, 2.5146e-08, ..., 6.5193e-09, + 2.3283e-10, 1.8626e-09], + [ 2.3283e-10, 2.3283e-10, 1.2503e-07, ..., 3.4692e-08, + 2.3283e-10, 8.8476e-09]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0031, 0.0059, 0.0165, -0.0039, 0.0294, -0.0044, 0.0008, -0.0050, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([-1.0440e-06, 1.2340e-08, 3.3295e-08, -3.3295e-08, 9.4995e-08, + 2.9104e-08, 5.7789e-07, 1.8626e-08, 5.8673e-08, 2.6939e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 245.79, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4093 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.2585, -0.0732, 0.2347, ..., 0.0357, -0.2663, -0.1784], + [-0.2202, -0.3021, -0.0814, ..., -0.1465, -0.3948, -0.1335], + [ 0.3320, -0.0858, -0.3414, ..., -0.2642, -0.3030, -0.2278], + ..., + [-0.1402, -0.0642, -0.4453, ..., -0.1573, 0.1515, 0.0586], + [-0.2554, -0.1661, -0.0958, ..., -0.2801, -0.0691, -0.0902], + [-0.1389, -0.1114, -0.1812, ..., 0.1046, 0.0256, -0.1951]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 2.3283e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -2.3283e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.5611e-09, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.9581e-09, + 1.3970e-09, 2.3283e-10]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0039, 0.0294, -0.0043, 0.0006, -0.0050, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.2107e-08, 1.3039e-08, 2.4005e-07, 7.0082e-08, + -1.6550e-06, 1.3029e-06, -1.2340e-08, 2.4214e-08, 6.2864e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 246.50, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4372 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.2585, -0.0732, 0.2348, ..., 0.0357, -0.2664, -0.1784], + [-0.2202, -0.3021, -0.0815, ..., -0.1466, -0.3949, -0.1335], + [ 0.3321, -0.0858, -0.3415, ..., -0.2643, -0.3031, -0.2278], + ..., + [-0.1403, -0.0642, -0.4453, ..., -0.1574, 0.1516, 0.0585], + [-0.2554, -0.1661, -0.0959, ..., -0.2802, -0.0692, -0.0902], + [-0.1389, -0.1114, -0.1813, ..., 0.1046, 0.0256, -0.1951]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 9.3132e-10, -3.7253e-09, ..., 1.8626e-09, + 5.1223e-09, 4.6566e-10], + [ 2.3283e-09, 4.1910e-09, 4.1910e-09, ..., 2.7940e-09, + 3.7253e-09, 1.3970e-09], + [ 1.3970e-09, 2.7940e-09, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 1.3970e-09], + ..., + [ 1.3504e-08, 2.6077e-08, 0.0000e+00, ..., 4.6566e-09, + -1.1176e-08, 0.0000e+00], + [ 9.3132e-10, 1.3970e-09, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-10, 9.3132e-10], + [ 4.6566e-10, 9.3132e-10, 1.8626e-09, ..., 2.1840e-07, + 1.3970e-09, 4.6566e-10]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0039, 0.0294, -0.0043, 0.0006, -0.0050, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 1.3597e-07, 1.4110e-07, 4.7032e-08, -2.6310e-07, -4.6706e-07, + 1.0710e-07, -5.8208e-08, -1.4994e-07, 7.9162e-09, 4.9453e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 246.30, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4274 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.2586, -0.0732, 0.2349, ..., 0.0357, -0.2665, -0.1784], + [-0.2203, -0.3021, -0.0815, ..., -0.1466, -0.3950, -0.1336], + [ 0.3323, -0.0858, -0.3416, ..., -0.2644, -0.3031, -0.2278], + ..., + [-0.1404, -0.0643, -0.4454, ..., -0.1575, 0.1517, 0.0586], + [-0.2555, -0.1661, -0.0960, ..., -0.2803, -0.0692, -0.0902], + [-0.1389, -0.1114, -0.1813, ..., 0.1046, 0.0256, -0.1952]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 3.2596e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0031, 0.0059, 0.0165, -0.0039, 0.0295, -0.0043, 0.0006, -0.0050, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-10, -4.6566e-10, 1.8626e-09, -1.1642e-08, -2.3283e-09, + 1.4435e-08, -9.3132e-09, 2.3283e-09, 4.6566e-10, 1.5367e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 246.40, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4260 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.2586, -0.0732, 0.2349, ..., 0.0357, -0.2666, -0.1784], + [-0.2203, -0.3022, -0.0816, ..., -0.1468, -0.3951, -0.1336], + [ 0.3325, -0.0858, -0.3417, ..., -0.2644, -0.3031, -0.2278], + ..., + [-0.1405, -0.0643, -0.4454, ..., -0.1576, 0.1517, 0.0586], + [-0.2556, -0.1661, -0.0961, ..., -0.2804, -0.0692, -0.0903], + [-0.1390, -0.1114, -0.1813, ..., 0.1046, 0.0255, -0.1952]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., -4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -7.4506e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0040, 0.0296, -0.0042, 0.0005, -0.0051, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 5.1223e-09, -4.6566e-10, 2.3283e-09, 3.0734e-08, 1.8161e-08, + 3.7253e-08, -7.4506e-08, 1.0710e-08, -3.3993e-08, 1.9092e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 246.54, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4030 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.2587, -0.0732, 0.2350, ..., 0.0357, -0.2666, -0.1784], + [-0.2204, -0.3022, -0.0816, ..., -0.1467, -0.3952, -0.1336], + [ 0.3326, -0.0859, -0.3418, ..., -0.2645, -0.3032, -0.2278], + ..., + [-0.1406, -0.0644, -0.4455, ..., -0.1578, 0.1518, 0.0586], + [-0.2556, -0.1661, -0.0963, ..., -0.2805, -0.0692, -0.0903], + [-0.1390, -0.1114, -0.1814, ..., 0.1046, 0.0255, -0.1953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., 4.6566e-10, + 3.2596e-09, 0.0000e+00], + [ 1.8626e-09, 4.6566e-10, 1.3970e-09, ..., 4.6566e-10, + 1.3970e-09, 0.0000e+00], + [-2.7940e-09, -4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -1.3970e-09, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -6.0536e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -3.2596e-09, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0040, 0.0295, -0.0041, 0.0004, -0.0051, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 3.9116e-08, 1.9092e-08, 0.0000e+00, -1.8626e-09, 5.3551e-08, + 6.2864e-08, -1.4296e-07, 3.2596e-09, -3.2131e-08, 1.1642e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 246.63, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4065 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.2587, -0.0732, 0.2351, ..., 0.0357, -0.2666, -0.1784], + [-0.2204, -0.3022, -0.0818, ..., -0.1468, -0.3954, -0.1337], + [ 0.3327, -0.0858, -0.3419, ..., -0.2645, -0.3032, -0.2278], + ..., + [-0.1407, -0.0644, -0.4455, ..., -0.1579, 0.1519, 0.0586], + [-0.2557, -0.1661, -0.0964, ..., -0.2806, -0.0692, -0.0903], + [-0.1390, -0.1114, -0.1814, ..., 0.1046, 0.0255, -0.1953]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 1.1176e-08, + 4.6566e-10, 4.6566e-10], + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 3.2596e-09, ..., 1.1642e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0039, 0.0295, -0.0041, 0.0003, -0.0051, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([-4.6566e-09, 2.0023e-08, 3.7253e-09, -8.5216e-08, -7.5903e-08, + 4.0047e-08, 1.8626e-09, 4.6100e-08, 1.3970e-08, 4.5635e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 246.58, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4236 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.22 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.2587, -0.0732, 0.2352, ..., 0.0357, -0.2667, -0.1784], + [-0.2204, -0.3022, -0.0819, ..., -0.1469, -0.3955, -0.1337], + [ 0.3328, -0.0859, -0.3420, ..., -0.2646, -0.3033, -0.2278], + ..., + [-0.1408, -0.0644, -0.4456, ..., -0.1580, 0.1519, 0.0586], + [-0.2557, -0.1661, -0.0965, ..., -0.2807, -0.0692, -0.0903], + [-0.1390, -0.1114, -0.1815, ..., 0.1046, 0.0255, -0.1953]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.1642e-08, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0039, 0.0296, -0.0041, 0.0003, -0.0051, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 4.6566e-10, -9.7789e-09, 2.3283e-09, -9.3132e-10, -3.5390e-08, + 4.6566e-10, 2.7940e-09, 1.6764e-08, 6.0536e-09, 3.4459e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 246.45, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4600 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.2587, -0.0732, 0.2353, ..., 0.0357, -0.2667, -0.1783], + [-0.2205, -0.3022, -0.0821, ..., -0.1470, -0.3956, -0.1338], + [ 0.3329, -0.0859, -0.3421, ..., -0.2646, -0.3033, -0.2278], + ..., + [-0.1409, -0.0645, -0.4456, ..., -0.1581, 0.1520, 0.0586], + [-0.2558, -0.1661, -0.0965, ..., -0.2807, -0.0692, -0.0903], + [-0.1391, -0.1114, -0.1815, ..., 0.1046, 0.0254, -0.1954]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1874e-07, ..., -3.1199e-08, + 0.0000e+00, -2.1886e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 9.3132e-10, ..., 6.0536e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.5367e-08, ..., 2.7940e-08, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0031, 0.0059, 0.0167, -0.0039, 0.0296, -0.0041, 0.0003, -0.0051, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([-3.0687e-07, 1.2573e-08, 9.3132e-10, 1.7695e-08, -7.7765e-08, + 4.4703e-08, 2.0629e-07, 2.3749e-08, 4.1910e-09, 8.5216e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 246.57, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4144 re_mapping 0.0018 re_causal 0.0073 /// teacc 99.22 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.2587, -0.0732, 0.2353, ..., 0.0357, -0.2668, -0.1783], + [-0.2205, -0.3022, -0.0821, ..., -0.1472, -0.3957, -0.1338], + [ 0.3330, -0.0859, -0.3421, ..., -0.2647, -0.3033, -0.2278], + ..., + [-0.1410, -0.0645, -0.4457, ..., -0.1583, 0.1521, 0.0586], + [-0.2559, -0.1661, -0.0966, ..., -0.2809, -0.0692, -0.0904], + [-0.1391, -0.1114, -0.1816, ..., 0.1046, 0.0254, -0.1954]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.2596e-09, ..., -1.3970e-09, + 0.0000e+00, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -4.6566e-10, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 2.3283e-09, ..., -1.2573e-08, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0040, 0.0297, -0.0041, 0.0003, -0.0051, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([-5.1223e-09, 1.0245e-08, 9.3132e-09, -1.2200e-07, 3.4925e-08, + 1.0943e-07, 4.1910e-09, 1.9092e-08, -1.9092e-08, -2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 246.36, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4420 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.2588, -0.0732, 0.2354, ..., 0.0357, -0.2669, -0.1783], + [-0.2206, -0.3022, -0.0821, ..., -0.1472, -0.3959, -0.1338], + [ 0.3331, -0.0859, -0.3422, ..., -0.2647, -0.3033, -0.2278], + ..., + [-0.1410, -0.0645, -0.4457, ..., -0.1584, 0.1521, 0.0587], + [-0.2559, -0.1662, -0.0967, ..., -0.2810, -0.0692, -0.0904], + [-0.1391, -0.1114, -0.1816, ..., 0.1046, 0.0254, -0.1954]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -7.9162e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., -2.1886e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0039, 0.0297, -0.0041, 0.0003, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-1.2713e-07, -8.0559e-08, -9.3132e-10, 6.5193e-09, 6.1933e-08, + 2.2817e-08, 2.0023e-08, 2.7008e-08, 8.6147e-08, -6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 246.23, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4153 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.2588, -0.0732, 0.2355, ..., 0.0357, -0.2669, -0.1783], + [-0.2206, -0.3022, -0.0822, ..., -0.1472, -0.3960, -0.1338], + [ 0.3332, -0.0859, -0.3422, ..., -0.2647, -0.3034, -0.2279], + ..., + [-0.1411, -0.0645, -0.4457, ..., -0.1585, 0.1522, 0.0587], + [-0.2560, -0.1662, -0.0968, ..., -0.2810, -0.0692, -0.0904], + [-0.1392, -0.1114, -0.1816, ..., 0.1046, 0.0254, -0.1955]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.2817e-08, + 0.0000e+00, 0.0000e+00], + [ 3.5390e-08, 0.0000e+00, -4.6566e-10, ..., 5.1223e-09, + 4.6566e-10, 0.0000e+00], + [-3.8184e-08, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 4.6566e-10, 4.6566e-10, ..., 7.4506e-09, + -9.3132e-10, 0.0000e+00], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 9.3132e-10, ..., 3.7719e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0039, 0.0298, -0.0041, 0.0003, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 6.0536e-08, 1.0198e-07, -7.6368e-08, -6.5193e-09, -2.2957e-07, + 1.4435e-08, 3.4925e-08, 2.6543e-08, -6.0536e-09, 9.7789e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 246.27, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4099 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.2588, -0.0732, 0.2355, ..., 0.0357, -0.2670, -0.1783], + [-0.2206, -0.3022, -0.0822, ..., -0.1473, -0.3961, -0.1338], + [ 0.3333, -0.0859, -0.3423, ..., -0.2649, -0.3034, -0.2278], + ..., + [-0.1411, -0.0646, -0.4457, ..., -0.1586, 0.1523, 0.0587], + [-0.2560, -0.1662, -0.0968, ..., -0.2811, -0.0692, -0.0904], + [-0.1392, -0.1114, -0.1816, ..., 0.1046, 0.0253, -0.1955]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.3993e-08, ..., -1.7695e-08, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 5.5879e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.7789e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., -7.4506e-09, + 1.8626e-09, 4.6566e-10]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0298, -0.0041, 0.0002, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-7.2177e-08, 1.6298e-08, 3.7253e-09, 3.2596e-09, 3.7253e-09, + 9.7789e-09, 4.2375e-08, 8.8476e-09, 3.6787e-08, -3.5856e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 246.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4247 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.2588, -0.0732, 0.2355, ..., 0.0357, -0.2671, -0.1783], + [-0.2207, -0.3022, -0.0822, ..., -0.1474, -0.3962, -0.1338], + [ 0.3334, -0.0859, -0.3423, ..., -0.2649, -0.3034, -0.2279], + ..., + [-0.1412, -0.0646, -0.4458, ..., -0.1587, 0.1523, 0.0587], + [-0.2560, -0.1662, -0.0969, ..., -0.2812, -0.0692, -0.0905], + [-0.1392, -0.1114, -0.1817, ..., 0.1046, 0.0253, -0.1955]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-10, 0.0000e+00], + [-6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 2.9802e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 4.6566e-10, ..., -1.5367e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0299, -0.0041, 0.0002, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 1.4435e-08, 1.0710e-08, 3.4459e-08, -5.1223e-09, -3.2596e-08, + 6.5193e-09, 9.3132e-10, 4.7963e-08, 3.2596e-09, -7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 246.33, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3952 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.16 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.2589, -0.0732, 0.2356, ..., 0.0357, -0.2671, -0.1783], + [-0.2207, -0.3022, -0.0823, ..., -0.1475, -0.3963, -0.1338], + [ 0.3335, -0.0859, -0.3423, ..., -0.2650, -0.3035, -0.2279], + ..., + [-0.1412, -0.0646, -0.4458, ..., -0.1588, 0.1524, 0.0587], + [-0.2560, -0.1662, -0.0970, ..., -0.2813, -0.0692, -0.0905], + [-0.1392, -0.1114, -0.1817, ..., 0.1046, 0.0253, -0.1955]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0299, -0.0041, 0.0003, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 4.1910e-09, -5.1223e-09, 0.0000e+00, 1.1176e-08, 7.4506e-09, + -1.0710e-08, -1.8626e-09, 8.3819e-09, 4.6566e-10, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 246.26, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4212 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.2589, -0.0732, 0.2357, ..., 0.0357, -0.2672, -0.1784], + [-0.2207, -0.3022, -0.0823, ..., -0.1476, -0.3965, -0.1338], + [ 0.3336, -0.0859, -0.3424, ..., -0.2650, -0.3035, -0.2278], + ..., + [-0.1413, -0.0646, -0.4458, ..., -0.1591, 0.1525, 0.0587], + [-0.2561, -0.1662, -0.0970, ..., -0.2814, -0.0692, -0.0905], + [-0.1393, -0.1114, -0.1817, ..., 0.1046, 0.0253, -0.1955]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, -0.0000e+00], + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 4.1910e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0245e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0299, -0.0042, 0.0003, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 3.2596e-09, 3.2596e-09, -1.8626e-09, 2.7940e-09, 1.3039e-08, + 3.5856e-08, -2.9802e-08, 3.2596e-09, 7.9162e-09, -2.6543e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 245.93, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4597 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.17 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.2589, -0.0732, 0.2357, ..., 0.0357, -0.2673, -0.1784], + [-0.2208, -0.3023, -0.0823, ..., -0.1477, -0.3966, -0.1338], + [ 0.3337, -0.0859, -0.3424, ..., -0.2650, -0.3035, -0.2278], + ..., + [-0.1414, -0.0647, -0.4458, ..., -0.1592, 0.1526, 0.0587], + [-0.2561, -0.1662, -0.0971, ..., -0.2815, -0.0693, -0.0905], + [-0.1393, -0.1114, -0.1818, ..., 0.1046, 0.0252, -0.1956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 1.4901e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0031, 0.0058, 0.0165, -0.0040, 0.0299, -0.0043, 0.0003, -0.0051, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 7.4506e-09, 6.5193e-09, 4.6566e-10, 2.9337e-08, -8.8010e-08, + -3.5390e-08, 1.6764e-08, -7.4506e-09, 1.8161e-08, 5.8673e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 246.46, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4012 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.2590, -0.0732, 0.2358, ..., 0.0357, -0.2673, -0.1784], + [-0.2208, -0.3023, -0.0823, ..., -0.1477, -0.3967, -0.1339], + [ 0.3338, -0.0859, -0.3424, ..., -0.2651, -0.3036, -0.2278], + ..., + [-0.1415, -0.0647, -0.4459, ..., -0.1594, 0.1526, 0.0587], + [-0.2561, -0.1662, -0.0971, ..., -0.2816, -0.0693, -0.0905], + [-0.1393, -0.1114, -0.1818, ..., 0.1046, 0.0252, -0.1956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.8673e-08, ..., -3.5856e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., 2.3283e-09, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 5.5879e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 3.5390e-08, ..., 2.2352e-08, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0031, 0.0059, 0.0165, -0.0039, 0.0300, -0.0043, 0.0002, -0.0052, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-1.4808e-07, 6.5193e-09, 4.6566e-09, -9.7789e-09, -1.8626e-09, + 3.2596e-09, 2.4680e-08, 1.0245e-08, 2.4680e-08, 9.2201e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 246.39, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4211 re_mapping 0.0019 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.2590, -0.0732, 0.2358, ..., 0.0357, -0.2674, -0.1785], + [-0.2208, -0.3023, -0.0823, ..., -0.1478, -0.3968, -0.1339], + [ 0.3340, -0.0859, -0.3424, ..., -0.2651, -0.3036, -0.2278], + ..., + [-0.1416, -0.0647, -0.4459, ..., -0.1595, 0.1527, 0.0587], + [-0.2562, -0.1662, -0.0972, ..., -0.2817, -0.0693, -0.0905], + [-0.1393, -0.1113, -0.1818, ..., 0.1046, 0.0252, -0.1956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8161e-08, + -2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -2.8871e-08, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0031, 0.0058, 0.0167, -0.0039, 0.0300, -0.0043, 0.0003, -0.0052, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-9.3132e-10, 5.1223e-09, 0.0000e+00, 9.3132e-10, 2.8405e-08, + 8.3819e-09, -1.2107e-08, 1.3970e-08, 1.3970e-09, -3.3062e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 246.70, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4410 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.2590, -0.0732, 0.2359, ..., 0.0357, -0.2675, -0.1785], + [-0.2209, -0.3023, -0.0824, ..., -0.1478, -0.3969, -0.1339], + [ 0.3342, -0.0860, -0.3425, ..., -0.2651, -0.3036, -0.2278], + ..., + [-0.1417, -0.0648, -0.4459, ..., -0.1596, 0.1528, 0.0587], + [-0.2562, -0.1662, -0.0972, ..., -0.2817, -0.0693, -0.0906], + [-0.1393, -0.1113, -0.1819, ..., 0.1046, 0.0252, -0.1956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, -1.3970e-09, 2.7940e-09, ..., -1.0245e-08, + -2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0031, 0.0058, 0.0167, -0.0038, 0.0301, -0.0043, 0.0002, -0.0052, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-5.5879e-09, -3.5856e-08, 4.1910e-09, -1.3504e-08, -7.4506e-09, + -2.3935e-07, 2.4308e-07, 5.3085e-08, 1.3504e-08, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 246.20, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.2591, -0.0732, 0.2360, ..., 0.0357, -0.2675, -0.1785], + [-0.2209, -0.3023, -0.0824, ..., -0.1479, -0.3970, -0.1339], + [ 0.3343, -0.0859, -0.3425, ..., -0.2652, -0.3036, -0.2278], + ..., + [-0.1418, -0.0648, -0.4460, ..., -0.1597, 0.1529, 0.0587], + [-0.2563, -0.1662, -0.0973, ..., -0.2819, -0.0693, -0.0906], + [-0.1394, -0.1113, -0.1819, ..., 0.1046, 0.0251, -0.1956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.3062e-08, ..., -1.6298e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 2.3283e-09, + 1.3970e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + -1.2573e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, + 2.3283e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.4901e-08, + 1.0710e-08, 1.8626e-09]], device='cuda:0') +Epoch 475, bias, value: tensor([ 3.1182e-03, 5.8945e-03, 1.6783e-02, -3.8210e-03, 3.0269e-02, + -4.2500e-03, -1.5174e-06, -5.2641e-03, 1.3048e-02, -1.7632e-02], + device='cuda:0'), grad: tensor([-6.7987e-08, 1.3504e-08, 2.3283e-09, 2.3283e-09, -8.8476e-09, + -3.6322e-08, 2.5611e-08, -4.5169e-08, 3.1665e-08, 1.0245e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 246.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4262 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.2591, -0.0732, 0.2361, ..., 0.0357, -0.2676, -0.1786], + [-0.2210, -0.3023, -0.0824, ..., -0.1479, -0.3971, -0.1339], + [ 0.3345, -0.0860, -0.3426, ..., -0.2652, -0.3036, -0.2278], + ..., + [-0.1419, -0.0648, -0.4460, ..., -0.1600, 0.1529, 0.0587], + [-0.2564, -0.1663, -0.0974, ..., -0.2820, -0.0693, -0.0906], + [-0.1394, -0.1114, -0.1820, ..., 0.1046, 0.0251, -0.1957]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.1665e-08, ..., -2.0023e-08, + 0.0000e+00, -0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 4.6566e-10, 0.0000e+00], + [-6.5193e-09, 0.0000e+00, 4.1910e-09, ..., 2.7940e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 1.3970e-09, ..., 3.7253e-09, + -1.3970e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 4.5169e-08, + -1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 476, bias, value: tensor([ 3.1198e-03, 6.0766e-03, 1.6831e-02, -3.7820e-03, 3.0331e-02, + -4.2687e-03, -2.8614e-05, -5.4375e-03, 1.3033e-02, -1.7635e-02], + device='cuda:0'), grad: tensor([-8.9873e-08, 1.4901e-08, -4.6566e-10, 1.4901e-08, -1.0012e-07, + -4.6566e-10, 6.5193e-09, 1.6298e-08, 2.8405e-08, 1.2387e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 246.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.2592, -0.0732, 0.2361, ..., 0.0357, -0.2677, -0.1786], + [-0.2211, -0.3023, -0.0824, ..., -0.1480, -0.3972, -0.1339], + [ 0.3347, -0.0860, -0.3426, ..., -0.2652, -0.3036, -0.2278], + ..., + [-0.1420, -0.0648, -0.4461, ..., -0.1601, 0.1530, 0.0587], + [-0.2564, -0.1663, -0.0974, ..., -0.2820, -0.0693, -0.0906], + [-0.1394, -0.1113, -0.1821, ..., 0.1046, 0.0251, -0.1957]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., 3.9581e-09, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, -6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., -1.0477e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 477, bias, value: tensor([ 3.1199e-03, 6.0670e-03, 1.6961e-02, -3.6773e-03, 3.0383e-02, + -4.4487e-03, 6.0416e-05, -5.4583e-03, 1.3052e-02, -1.7636e-02], + device='cuda:0'), grad: tensor([ 2.3283e-09, 4.4238e-09, -9.7789e-09, 3.2596e-08, 1.9325e-08, + 7.9162e-09, 3.2596e-09, 3.0501e-08, -2.7707e-08, -3.8184e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 246.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4072 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.2592, -0.0732, 0.2362, ..., 0.0357, -0.2677, -0.1786], + [-0.2211, -0.3023, -0.0824, ..., -0.1480, -0.3972, -0.1339], + [ 0.3349, -0.0860, -0.3426, ..., -0.2652, -0.3037, -0.2278], + ..., + [-0.1421, -0.0649, -0.4461, ..., -0.1602, 0.1530, 0.0587], + [-0.2565, -0.1663, -0.0974, ..., -0.2821, -0.0693, -0.0906], + [-0.1394, -0.1113, -0.1821, ..., 0.1046, 0.0250, -0.1957]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.2864e-09, ..., 4.6566e-10, + 2.7940e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.7474e-08, + 2.3283e-10, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.3283e-10, + -2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., -4.7497e-08, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 478, bias, value: tensor([ 3.1208e-03, 6.2690e-03, 1.7072e-02, -3.6802e-03, 3.0329e-02, + -4.4400e-03, -2.1296e-05, -5.6580e-03, 1.3060e-02, -1.7636e-02], + device='cuda:0'), grad: tensor([ 1.6764e-08, 7.5670e-08, 2.3283e-10, 6.2864e-09, 5.2387e-08, + 8.8243e-08, -9.8720e-08, 9.5461e-09, -5.5879e-09, -1.3295e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 248.26, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3914 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.22 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.2593, -0.0732, 0.2363, ..., 0.0357, -0.2678, -0.1786], + [-0.2212, -0.3023, -0.0825, ..., -0.1481, -0.3973, -0.1339], + [ 0.3350, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], + ..., + [-0.1422, -0.0649, -0.4461, ..., -0.1603, 0.1531, 0.0587], + [-0.2565, -0.1663, -0.0974, ..., -0.2821, -0.0693, -0.0906], + [-0.1394, -0.1113, -0.1821, ..., 0.1046, 0.0250, -0.1957]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, -1.6298e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-08, + 2.3283e-10, 0.0000e+00], + [-7.9162e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.8894e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 1.3970e-09, ..., -1.8394e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 479, bias, value: tensor([ 3.1219e-03, 6.3710e-03, 1.7141e-02, -3.6522e-03, 3.0344e-02, + -4.5717e-03, -4.7893e-05, -5.7705e-03, 1.3113e-02, -1.7637e-02], + device='cuda:0'), grad: tensor([-1.1642e-09, 4.3539e-08, -2.3749e-08, 7.4506e-09, 2.4680e-08, + -1.1642e-08, 5.3551e-09, 1.9092e-08, 1.0710e-08, -5.4017e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 248.08, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4356 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.2593, -0.0732, 0.2364, ..., 0.0357, -0.2678, -0.1786], + [-0.2212, -0.3024, -0.0825, ..., -0.1482, -0.3974, -0.1339], + [ 0.3352, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], + ..., + [-0.1424, -0.0649, -0.4462, ..., -0.1604, 0.1532, 0.0587], + [-0.2566, -0.1663, -0.0974, ..., -0.2821, -0.0693, -0.0906], + [-0.1395, -0.1113, -0.1822, ..., 0.1046, 0.0250, -0.1957]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5379e-08, ..., -1.6531e-08, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 6.9849e-10, ..., 9.3132e-10, + 0.0000e+00, -6.9849e-10], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 4.6566e-10, 2.3283e-10, ..., 6.9849e-10, + 0.0000e+00, 9.3132e-10], + [ 2.3283e-10, 2.3283e-10, 6.9849e-10, ..., -9.5461e-09, + 2.3283e-10, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 2.4447e-08, ..., 2.9569e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0031, 0.0064, 0.0173, -0.0036, 0.0303, -0.0046, -0.0001, -0.0058, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([-5.7975e-08, -3.4925e-09, 1.2806e-08, -1.1642e-08, -1.0477e-08, + 4.4471e-08, 4.1910e-09, 2.7241e-08, -1.5995e-07, 1.6717e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 247.75, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4094 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.21 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.2593, -0.0732, 0.2365, ..., 0.0357, -0.2679, -0.1786], + [-0.2213, -0.3024, -0.0826, ..., -0.1482, -0.3974, -0.1339], + [ 0.3355, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], + ..., + [-0.1425, -0.0650, -0.4462, ..., -0.1605, 0.1532, 0.0587], + [-0.2566, -0.1663, -0.0975, ..., -0.2822, -0.0693, -0.0906], + [-0.1395, -0.1113, -0.1822, ..., 0.1046, 0.0250, -0.1957]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 6.2864e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 2.0955e-09, 0.0000e+00], + [-8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + -3.0268e-09, 0.0000e+00], + [ 6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., -2.5611e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0031, 0.0064, 0.0175, -0.0036, 0.0304, -0.0047, -0.0002, -0.0058, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 2.5611e-09, 3.4925e-08, -1.0477e-08, -3.0268e-08, 7.6834e-09, + 1.1642e-08, 1.1642e-09, -3.9581e-09, 6.0536e-09, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 248.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4267 re_mapping 0.0018 re_causal 0.0077 /// teacc 99.20 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.2594, -0.0732, 0.2366, ..., 0.0357, -0.2679, -0.1786], + [-0.2214, -0.3024, -0.0826, ..., -0.1482, -0.3975, -0.1339], + [ 0.3357, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], + ..., + [-0.1426, -0.0650, -0.4462, ..., -0.1607, 0.1533, 0.0587], + [-0.2567, -0.1663, -0.0975, ..., -0.2822, -0.0693, -0.0906], + [-0.1395, -0.1113, -0.1823, ..., 0.1046, 0.0249, -0.1957]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 0.0000e+00, -3.7253e-09, ..., -1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-7.6834e-09, 0.0000e+00, -1.6298e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 4.6566e-09, ..., 1.0710e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0031, 0.0065, 0.0176, -0.0035, 0.0303, -0.0048, -0.0003, -0.0059, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([-3.9581e-09, -3.7253e-09, -1.4435e-08, -3.2596e-09, -2.1188e-08, + 3.7253e-09, 7.4506e-09, 8.6147e-09, 2.7940e-09, 3.1432e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 248.09, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4102 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.2594, -0.0732, 0.2366, ..., 0.0357, -0.2680, -0.1786], + [-0.2215, -0.3024, -0.0826, ..., -0.1482, -0.3976, -0.1339], + [ 0.3359, -0.0860, -0.3428, ..., -0.2654, -0.3037, -0.2278], + ..., + [-0.1429, -0.0650, -0.4462, ..., -0.1608, 0.1534, 0.0587], + [-0.2567, -0.1663, -0.0975, ..., -0.2822, -0.0693, -0.0906], + [-0.1395, -0.1113, -0.1823, ..., 0.1046, 0.0249, -0.1957]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 2.3283e-10, ..., 1.3970e-09, + 2.3283e-10, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 5.3551e-09, + 1.1642e-09, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -1.1642e-09, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., -2.4680e-08, + -4.8894e-09, 0.0000e+00]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0031, 0.0065, 0.0178, -0.0035, 0.0306, -0.0049, -0.0002, -0.0060, + 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 3.0268e-09, -3.9581e-09, 1.3970e-09, -1.7462e-08, 5.2154e-08, + 1.7462e-08, 0.0000e+00, 3.1665e-08, -2.9802e-08, -4.3074e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 247.64, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4022 re_mapping 0.0018 re_causal 0.0073 /// teacc 99.19 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.2594, -0.0732, 0.2367, ..., 0.0357, -0.2680, -0.1786], + [-0.2215, -0.3024, -0.0827, ..., -0.1483, -0.3977, -0.1339], + [ 0.3360, -0.0860, -0.3428, ..., -0.2654, -0.3038, -0.2278], + ..., + [-0.1429, -0.0651, -0.4463, ..., -0.1609, 0.1535, 0.0587], + [-0.2568, -0.1663, -0.0976, ..., -0.2823, -0.0693, -0.0906], + [-0.1395, -0.1113, -0.1824, ..., 0.1046, 0.0249, -0.1958]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3050e-08, ..., -6.0536e-09, + 0.0000e+00, -9.3132e-10], + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 1.1642e-09, 0.0000e+00], + [-1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 1.1642e-09, 0.0000e+00, 9.3132e-10, ..., 2.3283e-10, + -8.6147e-09, 2.3283e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-10, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 1.7229e-08, ..., 4.8894e-09, + 6.5193e-09, 6.9849e-10]], device='cuda:0') +Epoch 484, bias, value: tensor([ 3.1262e-03, 6.4709e-03, 1.7825e-02, -3.4399e-03, 3.0611e-02, + -4.9969e-03, 3.0793e-06, -5.9801e-03, 1.3163e-02, -1.7649e-02], + device='cuda:0'), grad: tensor([-5.3551e-08, 3.7253e-09, -1.6298e-09, -2.0955e-09, 2.3283e-09, + 6.2864e-09, 7.9162e-09, -2.0256e-08, -1.6298e-09, 6.6124e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 247.72, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4330 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.21 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.2594, -0.0732, 0.2368, ..., 0.0357, -0.2681, -0.1786], + [-0.2216, -0.3024, -0.0827, ..., -0.1483, -0.3979, -0.1339], + [ 0.3361, -0.0860, -0.3428, ..., -0.2655, -0.3038, -0.2278], + ..., + [-0.1430, -0.0651, -0.4463, ..., -0.1610, 0.1536, 0.0587], + [-0.2568, -0.1663, -0.0977, ..., -0.2824, -0.0693, -0.0906], + [-0.1395, -0.1113, -0.1824, ..., 0.1046, 0.0248, -0.1958]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -2.3283e-09, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 485, bias, value: tensor([ 3.1284e-03, 6.5911e-03, 1.7799e-02, -3.3711e-03, 3.0659e-02, + -5.0840e-03, -3.9424e-05, -6.0899e-03, 1.3146e-02, -1.7653e-02], + device='cuda:0'), grad: tensor([ 6.9849e-10, 1.3271e-08, 4.6566e-10, 4.6566e-10, -6.2864e-09, + 4.6566e-10, 1.3970e-09, -2.0955e-09, 9.3132e-10, 1.0477e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 247.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4447 re_mapping 0.0018 re_causal 0.0077 /// teacc 99.17 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.2594, -0.0733, 0.2369, ..., 0.0357, -0.2681, -0.1786], + [-0.2216, -0.3024, -0.0827, ..., -0.1483, -0.3980, -0.1339], + [ 0.3363, -0.0860, -0.3429, ..., -0.2655, -0.3038, -0.2278], + ..., + [-0.1431, -0.0651, -0.4463, ..., -0.1611, 0.1537, 0.0587], + [-0.2569, -0.1664, -0.0977, ..., -0.2825, -0.0693, -0.0906], + [-0.1396, -0.1113, -0.1825, ..., 0.1046, 0.0248, -0.1958]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4913e-08, ..., -6.7521e-09, + 0.0000e+00, -4.8894e-09], + [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, + 6.9849e-10, 9.3132e-10], + [-1.6298e-09, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 1.1642e-09, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, + -2.3982e-08, -6.9849e-10], + [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 3.4925e-09, ..., 1.1642e-09, + 2.0955e-09, 6.9849e-10]], device='cuda:0') +Epoch 486, bias, value: tensor([ 3.1297e-03, 6.5977e-03, 1.7871e-02, -3.3474e-03, 3.0687e-02, + -5.0980e-03, -6.1985e-05, -6.1066e-03, 1.3108e-02, -1.7655e-02], + device='cuda:0'), grad: tensor([-4.7497e-08, 1.2573e-08, -9.3132e-10, -3.2131e-08, 1.0966e-07, + 5.5647e-08, 4.0745e-08, -1.3178e-07, -1.2340e-08, 1.8859e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 246.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3865 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.2595, -0.0733, 0.2369, ..., 0.0357, -0.2682, -0.1786], + [-0.2217, -0.3024, -0.0828, ..., -0.1484, -0.3981, -0.1339], + [ 0.3363, -0.0860, -0.3429, ..., -0.2655, -0.3038, -0.2278], + ..., + [-0.1432, -0.0652, -0.4463, ..., -0.1612, 0.1539, 0.0587], + [-0.2569, -0.1664, -0.0977, ..., -0.2826, -0.0693, -0.0906], + [-0.1396, -0.1113, -0.1825, ..., 0.1046, 0.0248, -0.1958]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 6.9849e-10, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.6298e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-10, + -2.0955e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 4.1910e-09, + 1.1642e-09, 0.0000e+00]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0031, 0.0066, 0.0178, -0.0032, 0.0307, -0.0051, -0.0001, -0.0061, + 0.0131, -0.0177], device='cuda:0'), grad: tensor([ 4.4238e-09, -3.1898e-08, 3.4925e-09, 5.8208e-09, -1.9092e-08, + -1.6298e-08, 2.5844e-08, 4.1677e-08, -1.2340e-08, 2.3982e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 246.15, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4571 re_mapping 0.0017 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.2595, -0.0733, 0.2370, ..., 0.0357, -0.2682, -0.1786], + [-0.2217, -0.3025, -0.0828, ..., -0.1485, -0.3983, -0.1339], + [ 0.3365, -0.0860, -0.3429, ..., -0.2655, -0.3039, -0.2278], + ..., + [-0.1432, -0.0652, -0.4464, ..., -0.1613, 0.1540, 0.0587], + [-0.2569, -0.1664, -0.0978, ..., -0.2826, -0.0693, -0.0906], + [-0.1396, -0.1113, -0.1826, ..., 0.1046, 0.0247, -0.1958]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-6.0536e-09, -6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.4238e-09, 9.3132e-10, 0.0000e+00, ..., 2.3283e-10, + -4.6566e-10, -2.3283e-10], + [ 3.2596e-09, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-10, 0.0000e+00], + [ 1.1642e-09, 4.6566e-10, 6.9849e-10, ..., 8.1491e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0031, 0.0066, 0.0177, -0.0032, 0.0306, -0.0051, -0.0001, -0.0062, + 0.0131, -0.0177], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.0268e-09, -7.9162e-09, -1.0012e-06, -1.7229e-08, + 9.8441e-07, 3.7253e-09, 1.3271e-08, -2.3283e-09, 2.6543e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 246.45, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4235 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.2595, -0.0733, 0.2371, ..., 0.0357, -0.2682, -0.1786], + [-0.2218, -0.3025, -0.0828, ..., -0.1486, -0.3984, -0.1339], + [ 0.3366, -0.0860, -0.3430, ..., -0.2656, -0.3039, -0.2278], + ..., + [-0.1433, -0.0652, -0.4464, ..., -0.1615, 0.1541, 0.0587], + [-0.2569, -0.1664, -0.0978, ..., -0.2827, -0.0693, -0.0907], + [-0.1396, -0.1113, -0.1827, ..., 0.1046, 0.0247, -0.1958]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.5623e-08, ..., -3.1898e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -3.0268e-09, + 4.1910e-09, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 3.0268e-09, + -6.9849e-09, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 4.6566e-10, + 9.3132e-10, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 3.5623e-08, ..., 2.9569e-08, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0031, 0.0066, 0.0176, -0.0032, 0.0304, -0.0052, -0.0002, -0.0062, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([-8.4052e-08, -7.4506e-09, 4.8894e-09, 5.8208e-09, 2.6543e-08, + 1.1874e-08, -4.2841e-08, -1.1642e-08, 2.7940e-09, 1.1642e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 246.44, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4426 re_mapping 0.0017 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.2595, -0.0733, 0.2373, ..., 0.0357, -0.2683, -0.1787], + [-0.2218, -0.3025, -0.0829, ..., -0.1487, -0.3984, -0.1339], + [ 0.3366, -0.0860, -0.3431, ..., -0.2656, -0.3039, -0.2278], + ..., + [-0.1434, -0.0652, -0.4464, ..., -0.1617, 0.1542, 0.0587], + [-0.2570, -0.1664, -0.0978, ..., -0.2828, -0.0694, -0.0907], + [-0.1396, -0.1113, -0.1828, ..., 0.1046, 0.0247, -0.1958]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 6.9849e-10, ..., 6.9849e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 5.8208e-09, + 1.3970e-09, 2.3283e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + -1.6298e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 3.0268e-09, ..., -2.8871e-08, + -7.2177e-09, 0.0000e+00]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0031, 0.0068, 0.0174, -0.0032, 0.0302, -0.0053, -0.0001, -0.0064, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 5.3551e-09, 6.2864e-09, 9.3132e-10, 4.4238e-09, 6.0303e-08, + -2.0256e-08, 4.6566e-09, 1.9558e-08, -4.4238e-09, -5.6811e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 246.52, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4015 re_mapping 0.0017 re_causal 0.0068 /// teacc 99.18 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.2595, -0.0733, 0.2374, ..., 0.0357, -0.2684, -0.1787], + [-0.2218, -0.3025, -0.0829, ..., -0.1488, -0.3986, -0.1339], + [ 0.3367, -0.0860, -0.3432, ..., -0.2657, -0.3040, -0.2278], + ..., + [-0.1435, -0.0653, -0.4465, ..., -0.1618, 0.1543, 0.0587], + [-0.2570, -0.1664, -0.0979, ..., -0.2829, -0.0694, -0.0907], + [-0.1397, -0.1113, -0.1829, ..., 0.1046, 0.0247, -0.1958]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 2.3283e-10, -2.3283e-10, ..., 2.3283e-10, + 6.9849e-10, 0.0000e+00], + [-4.6566e-09, -1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + [ 1.6298e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + 6.9849e-10, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., -1.1642e-09, + 6.9849e-10, 2.3283e-10]], device='cuda:0') +Epoch 491, bias, value: tensor([ 3.1347e-03, 6.8227e-03, 1.7222e-02, -3.2285e-03, 3.0162e-02, + -5.3294e-03, 1.7088e-05, -6.3709e-03, 1.3122e-02, -1.7644e-02], + device='cuda:0'), grad: tensor([ 1.1409e-08, -1.1642e-08, -8.8476e-09, 1.1874e-08, 6.0536e-09, + -6.8918e-08, 1.7462e-08, 1.9092e-08, 2.2585e-08, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 246.35, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4216 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.2595, -0.0733, 0.2376, ..., 0.0357, -0.2684, -0.1787], + [-0.2219, -0.3025, -0.0830, ..., -0.1489, -0.3987, -0.1339], + [ 0.3368, -0.0860, -0.3433, ..., -0.2657, -0.3040, -0.2278], + ..., + [-0.1435, -0.0653, -0.4465, ..., -0.1619, 0.1545, 0.0587], + [-0.2570, -0.1664, -0.0980, ..., -0.2830, -0.0694, -0.0907], + [-0.1397, -0.1113, -0.1830, ..., 0.1046, 0.0246, -0.1959]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.1223e-09, ..., -3.2596e-09, + 0.0000e+00, -2.3283e-10], + [ 2.7940e-09, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + 6.9849e-10, 2.7940e-09], + [-3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -9.3132e-10, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 3.4925e-09, ..., 7.2177e-09, + 2.3283e-10, 4.6566e-10]], device='cuda:0') +Epoch 492, bias, value: tensor([ 3.1399e-03, 6.8837e-03, 1.7148e-02, -3.1274e-03, 3.0145e-02, + -5.4663e-03, 7.6682e-05, -6.4609e-03, 1.3102e-02, -1.7647e-02], + device='cuda:0'), grad: tensor([-9.5461e-09, 1.0477e-07, -6.5193e-09, 5.3318e-08, -2.4214e-08, + -1.6182e-07, 7.6834e-09, 1.0943e-08, 7.6834e-09, 3.3760e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 246.76, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4164 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.23 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.2596, -0.0733, 0.2377, ..., 0.0357, -0.2685, -0.1787], + [-0.2219, -0.3025, -0.0830, ..., -0.1490, -0.3988, -0.1339], + [ 0.3369, -0.0860, -0.3433, ..., -0.2657, -0.3040, -0.2278], + ..., + [-0.1436, -0.0653, -0.4466, ..., -0.1620, 0.1546, 0.0587], + [-0.2570, -0.1664, -0.0981, ..., -0.2831, -0.0694, -0.0907], + [-0.1397, -0.1113, -0.1831, ..., 0.1046, 0.0246, -0.1959]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 6.2864e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-2.1188e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, 0.0000e+00], + ..., + [ 1.2573e-08, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 2.0955e-09, -1.8626e-09, 9.3132e-10, ..., -7.9162e-09, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0031, 0.0069, 0.0171, -0.0030, 0.0301, -0.0056, 0.0002, -0.0065, + 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.1176e-08, -4.9127e-08, 1.1874e-08, 8.6147e-09, + -5.0291e-08, 8.8476e-09, 4.2608e-08, 3.1432e-08, -6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 246.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4069 re_mapping 0.0017 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.2596, -0.0733, 0.2378, ..., 0.0357, -0.2686, -0.1788], + [-0.2220, -0.3025, -0.0831, ..., -0.1490, -0.3989, -0.1339], + [ 0.3371, -0.0860, -0.3434, ..., -0.2657, -0.3041, -0.2278], + ..., + [-0.1437, -0.0653, -0.4466, ..., -0.1621, 0.1548, 0.0588], + [-0.2571, -0.1664, -0.0982, ..., -0.2832, -0.0694, -0.0907], + [-0.1397, -0.1113, -0.1832, ..., 0.1046, 0.0245, -0.1959]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00], + [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + -6.9849e-10, 0.0000e+00], + [ 3.0268e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0031, 0.0069, 0.0171, -0.0030, 0.0300, -0.0057, 0.0003, -0.0065, + 0.0130, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-09, 9.3132e-10, -2.0023e-08, 1.4203e-08, 4.1910e-09, + -1.4668e-08, 8.8476e-09, 4.6566e-09, 6.2864e-09, 6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 246.71, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4303 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.2596, -0.0733, 0.2380, ..., 0.0357, -0.2687, -0.1788], + [-0.2220, -0.3025, -0.0831, ..., -0.1491, -0.3991, -0.1339], + [ 0.3373, -0.0860, -0.3434, ..., -0.2658, -0.3041, -0.2278], + ..., + [-0.1438, -0.0654, -0.4466, ..., -0.1623, 0.1550, 0.0588], + [-0.2572, -0.1665, -0.0982, ..., -0.2832, -0.0694, -0.0908], + [-0.1397, -0.1113, -0.1833, ..., 0.1046, 0.0245, -0.1959]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3772e-08, ..., -3.0035e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 1.3970e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.8626e-09, + -4.1910e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.1642e-09, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7951e-08, ..., -5.5879e-09, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0031, 0.0069, 0.0171, -0.0030, 0.0299, -0.0057, 0.0004, -0.0066, + 0.0130, -0.0176], device='cuda:0'), grad: tensor([-1.1059e-07, 1.6065e-08, 1.0943e-08, 3.7253e-09, 9.0105e-08, + -1.3970e-09, 9.3132e-09, -2.0489e-08, 6.2864e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 246.80, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4218 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.2596, -0.0733, 0.2381, ..., 0.0357, -0.2688, -0.1788], + [-0.2221, -0.3025, -0.0832, ..., -0.1492, -0.3992, -0.1340], + [ 0.3375, -0.0860, -0.3435, ..., -0.2658, -0.3041, -0.2278], + ..., + [-0.1439, -0.0654, -0.4467, ..., -0.1624, 0.1552, 0.0588], + [-0.2572, -0.1665, -0.0983, ..., -0.2833, -0.0695, -0.0908], + [-0.1397, -0.1113, -0.1834, ..., 0.1046, 0.0244, -0.1959]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 6.9849e-10, + 9.3132e-10, -2.3283e-09], + [-3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + -9.3132e-10, 2.3283e-10], + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 2.5611e-09, + 6.2864e-09, 1.6298e-09], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., -1.6298e-08, + -4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0031, 0.0070, 0.0172, -0.0030, 0.0298, -0.0057, 0.0005, -0.0066, + 0.0130, -0.0176], device='cuda:0'), grad: tensor([ 5.1223e-09, -6.5193e-08, -4.8894e-09, 1.6997e-08, 3.8184e-08, + -1.2550e-07, 4.6799e-08, 8.1491e-09, 1.2130e-07, -2.8405e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 246.16, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4127 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.19 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.2597, -0.0733, 0.2382, ..., 0.0357, -0.2689, -0.1788], + [-0.2222, -0.3025, -0.0832, ..., -0.1494, -0.3994, -0.1340], + [ 0.3376, -0.0860, -0.3435, ..., -0.2658, -0.3042, -0.2278], + ..., + [-0.1440, -0.0654, -0.4467, ..., -0.1626, 0.1554, 0.0588], + [-0.2572, -0.1665, -0.0984, ..., -0.2834, -0.0695, -0.0908], + [-0.1398, -0.1113, -0.1835, ..., 0.1046, 0.0244, -0.1960]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 8.1491e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 1.0477e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 1.1642e-10, 0.0000e+00], + ..., + [ 5.8208e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.2806e-09, 0.0000e+00], + [ 5.8208e-10, 0.0000e+00, 1.1642e-10, ..., 6.9849e-10, + 5.8208e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.5611e-09, + 8.1491e-10, 0.0000e+00]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0031, 0.0069, 0.0172, -0.0030, 0.0299, -0.0057, 0.0005, -0.0065, + 0.0129, -0.0177], device='cuda:0'), grad: tensor([ 4.6566e-09, 5.4715e-09, 8.1491e-10, 8.0327e-09, -1.2224e-08, + -1.9092e-08, -3.9581e-09, -3.6089e-09, 7.6834e-09, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 246.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4403 re_mapping 0.0017 re_causal 0.0069 /// teacc 99.19 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.2597, -0.0733, 0.2383, ..., 0.0357, -0.2689, -0.1788], + [-0.2223, -0.3025, -0.0833, ..., -0.1495, -0.3995, -0.1340], + [ 0.3379, -0.0861, -0.3435, ..., -0.2659, -0.3042, -0.2278], + ..., + [-0.1442, -0.0655, -0.4467, ..., -0.1627, 0.1555, 0.0588], + [-0.2573, -0.1665, -0.0984, ..., -0.2834, -0.0695, -0.0908], + [-0.1398, -0.1113, -0.1836, ..., 0.1046, 0.0243, -0.1960]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 1.3970e-09, 2.3283e-10], + [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-09, 9.3132e-10, 0.0000e+00, ..., 2.5611e-09, + -6.9849e-10, 0.0000e+00], + [ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8161e-08, + 6.9849e-10, 2.3283e-10]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0032, 0.0069, 0.0173, -0.0030, 0.0298, -0.0058, 0.0006, -0.0066, + 0.0130, -0.0177], device='cuda:0'), grad: tensor([ 4.1910e-09, -1.7928e-08, -1.7229e-08, -2.3283e-09, -4.0978e-08, + 4.8894e-09, -2.1653e-08, 3.4925e-08, 1.6298e-08, 5.1456e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 246.15, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4404 re_mapping 0.0016 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.2597, -0.0733, 0.2386, ..., 0.0357, -0.2691, -0.1788], + [-0.2224, -0.3026, -0.0833, ..., -0.1495, -0.3997, -0.1340], + [ 0.3382, -0.0861, -0.3436, ..., -0.2659, -0.3043, -0.2278], + ..., + [-0.1444, -0.0655, -0.4468, ..., -0.1628, 0.1557, 0.0588], + [-0.2574, -0.1665, -0.0985, ..., -0.2835, -0.0695, -0.0908], + [-0.1398, -0.1113, -0.1837, ..., 0.1046, 0.0243, -0.1960]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.0268e-09, ..., -2.3749e-08, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0032, 0.0070, 0.0174, -0.0031, 0.0297, -0.0055, 0.0005, -0.0068, + 0.0129, -0.0177], device='cuda:0'), grad: tensor([-1.6298e-09, -9.7789e-09, 4.6566e-10, 1.1176e-08, 7.3574e-08, + -1.8394e-08, 7.9162e-09, 2.0489e-08, 5.5879e-09, -7.3807e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 246.47, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4229 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.2598, -0.0733, 0.2387, ..., 0.0357, -0.2691, -0.1788], + [-0.2225, -0.3026, -0.0833, ..., -0.1497, -0.3999, -0.1340], + [ 0.3383, -0.0861, -0.3436, ..., -0.2659, -0.3043, -0.2278], + ..., + [-0.1445, -0.0655, -0.4468, ..., -0.1629, 0.1560, 0.0588], + [-0.2574, -0.1665, -0.0985, ..., -0.2836, -0.0696, -0.0908], + [-0.1398, -0.1113, -0.1838, ..., 0.1046, 0.0242, -0.1960]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.0734e-08, ..., -1.3039e-08, + 6.9849e-10, -0.0000e+00], + [ 2.3283e-10, 4.6566e-10, 1.3970e-09, ..., 9.3132e-10, + 2.3283e-10, 0.0000e+00], + [ 1.1642e-09, 3.4925e-09, 1.1642e-09, ..., 4.6566e-10, + 6.9849e-10, 2.3283e-10], + ..., + [ 4.6566e-10, 1.1642e-09, 2.3283e-09, ..., 2.5611e-09, + 6.9849e-10, 0.0000e+00], + [ 0.0000e+00, -1.1642e-09, 6.9849e-09, ..., 4.6566e-10, + -1.6298e-08, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 1.2340e-08, ..., 2.6310e-08, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0032, 0.0070, 0.0173, -0.0032, 0.0296, -0.0054, 0.0005, -0.0067, + 0.0130, -0.0177], device='cuda:0'), grad: tensor([-7.4040e-08, 6.7521e-09, 2.3283e-08, 3.3760e-08, -8.1258e-08, + 1.2806e-08, 1.3970e-08, 2.7474e-08, -7.1945e-08, 1.1921e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 246.31, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4185 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_onlyblock2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0169, -0.0253, 0.0165, ..., 0.0308, -0.0245, 0.0071], + [ 0.0253, -0.0156, -0.0133, ..., -0.0190, 0.0099, -0.0189], + [-0.0034, 0.0289, -0.0063, ..., -0.0258, -0.0044, -0.0031], + ..., + [-0.0209, 0.0037, 0.0139, ..., -0.0167, -0.0230, 0.0121], + [-0.0203, 0.0144, 0.0223, ..., -0.0020, 0.0059, -0.0016], + [-0.0243, -0.0195, 0.0228, ..., 0.0235, -0.0155, 0.0110]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0295, -0.0169, -0.0055, 0.0009, 0.0004, -0.0044, -0.0083, -0.0051, + 0.0301, 0.0197], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 276.71, cls_loss 1.2511 cls_loss_mapping 1.7983 cls_loss_causal 2.2017 re_mapping 0.1599 re_causal 0.1691 /// teacc 81.97 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0159, -0.0265, 0.0146, ..., 0.0314, -0.0300, 0.0096], + [ 0.0175, -0.0222, -0.0215, ..., -0.0196, 0.0111, -0.0269], + [ 0.0010, 0.0316, -0.0102, ..., -0.0265, -0.0117, -0.0075], + ..., + [-0.0158, 0.0036, 0.0140, ..., -0.0174, -0.0271, 0.0187], + [-0.0228, 0.0152, 0.0244, ..., -0.0026, 0.0058, -0.0085], + [-0.0246, -0.0182, 0.0245, ..., 0.0228, -0.0146, 0.0131]], + device='cuda:0'), grad: tensor([[ 2.3232e-03, 8.7690e-04, -2.7943e-03, ..., -7.7844e-05, + -7.9041e-03, -1.5823e-02], + [ 1.0414e-02, 4.3106e-03, 8.9417e-03, ..., 3.8580e-07, + 2.2644e-02, 1.2611e-02], + [ 6.0177e-04, 5.2605e-03, 1.0315e-02, ..., 6.9402e-06, + 2.1088e-02, 8.8196e-03], + ..., + [-4.5746e-02, -3.5515e-03, -3.8166e-03, ..., 1.3569e-06, + 1.0201e-02, -6.1707e-02], + [ 1.4015e-02, 1.4467e-03, 9.7351e-03, ..., 6.0350e-06, + 2.3148e-02, 2.8137e-02], + [ 3.6682e-02, 3.5610e-03, 4.6753e-02, ..., 2.8834e-06, + 4.0314e-02, 8.4595e-02]], device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0273, -0.0155, -0.0061, 0.0006, -0.0003, -0.0041, -0.0080, -0.0044, + 0.0297, 0.0196], device='cuda:0'), grad: tensor([-0.0194, 0.0277, 0.0198, -0.0734, -0.0612, 0.0463, -0.0369, -0.0263, + 0.0376, 0.0858], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 275.56, cls_loss 0.4374 cls_loss_mapping 0.7801 cls_loss_causal 1.9032 re_mapping 0.2112 re_causal 0.2743 /// teacc 91.82 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0167, -0.0268, 0.0123, ..., 0.0305, -0.0334, 0.0104], + [ 0.0147, -0.0253, -0.0245, ..., -0.0167, 0.0124, -0.0288], + [ 0.0039, 0.0329, -0.0128, ..., -0.0289, -0.0159, -0.0076], + ..., + [-0.0139, 0.0030, 0.0116, ..., -0.0196, -0.0330, 0.0209], + [-0.0220, 0.0185, 0.0271, ..., -0.0052, 0.0076, -0.0138], + [-0.0237, -0.0164, 0.0242, ..., 0.0211, -0.0172, 0.0153]], + device='cuda:0'), grad: tensor([[-2.0981e-03, 1.9484e-03, 2.2566e-04, ..., 3.1758e-07, + 1.1930e-03, -6.1750e-04], + [ 1.6602e-02, 9.0027e-03, 5.2147e-03, ..., -5.6237e-05, + 1.8692e-02, 1.8402e-02], + [-1.0246e-02, -7.2250e-03, 2.3499e-03, ..., 1.0781e-05, + 3.9558e-03, 1.5459e-03], + ..., + [-6.8893e-03, 1.4181e-03, 4.1733e-03, ..., 1.3402e-06, + 7.2403e-03, -2.3590e-02], + [ 4.2419e-03, 1.5480e-02, 2.5665e-02, ..., 3.6806e-05, + 2.1729e-02, -3.9635e-03], + [-2.6150e-03, -1.1200e-02, -8.7585e-03, ..., 3.9255e-07, + -6.9809e-03, -6.5193e-03]], device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0270, -0.0152, -0.0063, 0.0003, -0.0003, -0.0032, -0.0089, -0.0048, + 0.0301, 0.0200], device='cuda:0'), grad: tensor([ 0.0004, 0.0277, -0.0048, -0.0062, -0.0087, -0.0186, 0.0092, -0.0090, + 0.0159, -0.0059], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 275.31, cls_loss 0.2946 cls_loss_mapping 0.4847 cls_loss_causal 1.7281 re_mapping 0.1530 re_causal 0.2479 /// teacc 94.34 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0169, -0.0273, 0.0102, ..., 0.0296, -0.0365, 0.0103], + [ 0.0126, -0.0277, -0.0265, ..., -0.0140, 0.0128, -0.0307], + [ 0.0067, 0.0341, -0.0145, ..., -0.0334, -0.0197, -0.0077], + ..., + [-0.0129, 0.0024, 0.0107, ..., -0.0329, -0.0363, 0.0225], + [-0.0222, 0.0209, 0.0286, ..., -0.0074, 0.0086, -0.0162], + [-0.0245, -0.0163, 0.0237, ..., 0.0107, -0.0183, 0.0161]], + device='cuda:0'), grad: tensor([[-8.3313e-03, 1.1253e-03, 6.3658e-04, ..., -1.9073e-04, + -3.5820e-03, -8.5144e-03], + [ 2.1706e-03, 3.8319e-03, 4.9095e-03, ..., 1.6415e-04, + 9.4299e-03, 6.4516e-04], + [ 1.0557e-03, 3.4866e-03, 4.3869e-03, ..., 1.8330e-03, + 6.3858e-03, 1.3914e-03], + ..., + [ 1.3371e-03, 2.2907e-03, 1.6737e-03, ..., 6.1393e-05, + 2.7962e-03, 2.3329e-04], + [-1.7746e-02, -1.9470e-02, -6.0120e-03, ..., 2.0428e-03, + -1.1093e-02, -4.0932e-03], + [ 8.6136e-03, 1.0872e-02, 1.6754e-02, ..., 1.2791e-04, + 2.9449e-02, 6.5613e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0267, -0.0154, -0.0060, 0.0005, -0.0002, -0.0028, -0.0093, -0.0052, + 0.0303, 0.0199], device='cuda:0'), grad: tensor([-0.0111, 0.0097, 0.0061, 0.0003, -0.0330, 0.0131, -0.0012, 0.0043, + -0.0292, 0.0409], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 275.17, cls_loss 0.2248 cls_loss_mapping 0.3588 cls_loss_causal 1.5569 re_mapping 0.1233 re_causal 0.2187 /// teacc 94.49 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0167, -0.0278, 0.0086, ..., 0.0312, -0.0387, 0.0103], + [ 0.0115, -0.0291, -0.0284, ..., -0.0160, 0.0131, -0.0324], + [ 0.0083, 0.0344, -0.0164, ..., -0.0371, -0.0224, -0.0089], + ..., + [-0.0120, 0.0027, 0.0096, ..., -0.0394, -0.0391, 0.0235], + [-0.0226, 0.0225, 0.0301, ..., -0.0079, 0.0088, -0.0190], + [-0.0251, -0.0160, 0.0233, ..., 0.0091, -0.0188, 0.0174]], + device='cuda:0'), grad: tensor([[-1.3962e-03, 1.3437e-03, 9.4366e-04, ..., -5.6791e-04, + 7.7915e-04, -5.2567e-03], + [ 1.2985e-02, 9.7885e-03, 6.7787e-03, ..., 3.8099e-04, + 1.1681e-02, 4.8447e-03], + [ 1.0635e-02, -5.5199e-03, 1.2922e-03, ..., 2.0492e-04, + -3.4885e-03, 1.6418e-02], + ..., + [-4.4739e-02, -1.3489e-02, 2.5439e-04, ..., 2.2233e-05, + 6.1417e-04, -3.6621e-02], + [ 4.3564e-03, 5.6076e-04, 4.6349e-03, ..., 1.3723e-03, + 3.5515e-03, 3.1452e-03], + [ 5.5428e-03, 3.4943e-03, 1.2045e-03, ..., 1.3709e-04, + 1.0433e-03, 6.0501e-03]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0268, -0.0152, -0.0061, 0.0004, -0.0001, -0.0028, -0.0095, -0.0051, + 0.0300, 0.0202], device='cuda:0'), grad: tensor([-0.0038, 0.0334, -0.0076, 0.0182, 0.0019, -0.0165, 0.0015, -0.0421, + 0.0070, 0.0080], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 275.33, cls_loss 0.1651 cls_loss_mapping 0.2728 cls_loss_causal 1.4585 re_mapping 0.1028 re_causal 0.2002 /// teacc 96.07 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0164, -0.0282, 0.0070, ..., 0.0306, -0.0408, 0.0110], + [ 0.0100, -0.0311, -0.0302, ..., -0.0184, 0.0131, -0.0339], + [ 0.0103, 0.0358, -0.0182, ..., -0.0392, -0.0255, -0.0099], + ..., + [-0.0110, 0.0029, 0.0091, ..., -0.0416, -0.0409, 0.0245], + [-0.0223, 0.0240, 0.0310, ..., -0.0101, 0.0093, -0.0211], + [-0.0265, -0.0161, 0.0225, ..., 0.0083, -0.0197, 0.0178]], + device='cuda:0'), grad: tensor([[ 4.1351e-03, 9.9106e-03, 4.4174e-03, ..., 6.6042e-05, + 1.3571e-03, 1.6756e-03], + [-9.1267e-04, -5.9471e-03, -3.2749e-03, ..., 2.9579e-06, + -5.1689e-03, 3.9434e-04], + [ 2.2793e-03, 5.3329e-03, 3.4237e-03, ..., 2.2739e-05, + 3.2387e-03, 9.9754e-04], + ..., + [ 5.0316e-03, 5.3329e-03, 1.4944e-03, ..., 5.1269e-07, + 5.6152e-03, 1.1543e-02], + [-1.4236e-02, -3.2684e-02, -5.1666e-02, ..., 1.6034e-05, + -3.2288e-02, -8.8453e-04], + [ 8.9264e-03, 9.1705e-03, 2.9278e-03, ..., 5.0664e-06, + 5.1041e-03, 1.6739e-02]], device='cuda:0') +Epoch 6, bias, value: tensor([ 2.6959e-02, -1.5339e-02, -5.8197e-03, 1.9760e-04, -3.5663e-05, + -2.6938e-03, -9.9802e-03, -4.9399e-03, 3.0016e-02, 2.0198e-02], + device='cuda:0'), grad: tensor([ 0.0058, -0.0184, 0.0090, -0.0263, 0.0036, 0.0185, 0.0043, 0.0122, + -0.0270, 0.0182], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 275.76, cls_loss 0.1362 cls_loss_mapping 0.2158 cls_loss_causal 1.2994 re_mapping 0.0852 re_causal 0.1719 /// teacc 96.45 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0162, -0.0287, 0.0053, ..., 0.0312, -0.0428, 0.0110], + [ 0.0088, -0.0325, -0.0323, ..., -0.0203, 0.0123, -0.0347], + [ 0.0118, 0.0360, -0.0202, ..., -0.0399, -0.0283, -0.0110], + ..., + [-0.0097, 0.0040, 0.0083, ..., -0.0435, -0.0432, 0.0254], + [-0.0218, 0.0257, 0.0320, ..., -0.0124, 0.0099, -0.0227], + [-0.0276, -0.0165, 0.0222, ..., 0.0070, -0.0200, 0.0186]], + device='cuda:0'), grad: tensor([[-5.1842e-03, -2.6017e-05, 9.9182e-04, ..., 6.9261e-05, + 9.1028e-04, -1.3313e-03], + [ 1.1246e-02, 1.3443e-02, 2.9488e-03, ..., 1.2599e-05, + 1.8740e-03, 2.2755e-03], + [ 5.3329e-03, 1.4496e-02, 6.5842e-03, ..., 4.1097e-05, + 5.6343e-03, 1.4210e-03], + ..., + [ 5.6915e-03, 5.5923e-03, 5.3501e-04, ..., 2.8126e-06, + 1.5879e-03, 1.9398e-03], + [-2.5360e-02, -4.3243e-02, -1.6205e-02, ..., -9.5218e-06, + -1.3893e-02, -4.1237e-03], + [ 7.5493e-03, 1.8549e-03, 8.3303e-04, ..., 1.6302e-05, + 5.2834e-03, 6.4583e-03]], device='cuda:0') +Epoch 7, bias, value: tensor([ 2.7292e-02, -1.5538e-02, -6.1149e-03, 1.5834e-04, -5.4673e-05, + -2.4777e-03, -1.0435e-02, -4.7028e-03, 3.0156e-02, 2.0233e-02], + device='cuda:0'), grad: tensor([-0.0057, 0.0148, 0.0147, 0.0049, -0.0163, 0.0052, 0.0021, 0.0110, + -0.0480, 0.0173], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 275.37, cls_loss 0.1168 cls_loss_mapping 0.1849 cls_loss_causal 1.2598 re_mapping 0.0758 re_causal 0.1636 /// teacc 96.88 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0158, -0.0294, 0.0044, ..., 0.0310, -0.0444, 0.0110], + [ 0.0074, -0.0339, -0.0338, ..., -0.0220, 0.0116, -0.0350], + [ 0.0132, 0.0368, -0.0214, ..., -0.0410, -0.0301, -0.0114], + ..., + [-0.0093, 0.0043, 0.0080, ..., -0.0435, -0.0451, 0.0257], + [-0.0216, 0.0272, 0.0327, ..., -0.0151, 0.0102, -0.0243], + [-0.0284, -0.0169, 0.0219, ..., 0.0055, -0.0199, 0.0193]], + device='cuda:0'), grad: tensor([[ 1.6117e-04, 1.7271e-03, 3.1128e-03, ..., 1.2932e-03, + 2.5558e-03, 1.3103e-03], + [ 5.2977e-04, 3.5954e-04, 4.4060e-03, ..., 1.4436e-04, + 9.2163e-03, 2.8019e-03], + [-2.1954e-03, -2.5630e-04, 7.5340e-04, ..., 2.2316e-04, + 7.7629e-04, 3.6025e-04], + ..., + [-6.0368e-04, -7.8964e-04, 8.6546e-04, ..., 6.8605e-05, + 1.1816e-03, -7.6103e-03], + [ 5.0640e-04, -1.1462e-04, 1.9646e-03, ..., 3.8719e-04, + 3.6583e-03, 2.8534e-03], + [ 4.0674e-04, -8.3542e-04, 3.5229e-03, ..., 1.3733e-04, + 4.4632e-03, 3.6087e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([ 2.7674e-02, -1.5673e-02, -6.1870e-03, 6.2555e-04, 1.7484e-05, + -2.7191e-03, -1.0810e-02, -4.7630e-03, 3.0345e-02, 2.0064e-02], + device='cuda:0'), grad: tensor([ 3.7212e-03, 1.4145e-02, -8.8692e-05, -2.6760e-03, 4.8904e-03, + -2.1713e-02, -4.9896e-03, -4.4518e-03, 5.8861e-03, 5.2910e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 275.10, cls_loss 0.1027 cls_loss_mapping 0.1569 cls_loss_causal 1.2471 re_mapping 0.0680 re_causal 0.1488 /// teacc 97.03 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0156, -0.0299, 0.0031, ..., 0.0297, -0.0468, 0.0114], + [ 0.0063, -0.0354, -0.0353, ..., -0.0237, 0.0108, -0.0350], + [ 0.0142, 0.0367, -0.0226, ..., -0.0419, -0.0326, -0.0122], + ..., + [-0.0083, 0.0053, 0.0078, ..., -0.0443, -0.0463, 0.0264], + [-0.0212, 0.0285, 0.0337, ..., -0.0165, 0.0109, -0.0268], + [-0.0298, -0.0172, 0.0215, ..., 0.0049, -0.0202, 0.0196]], + device='cuda:0'), grad: tensor([[ 0.0045, 0.0041, 0.0011, ..., 0.0008, 0.0015, 0.0043], + [ 0.0033, 0.0057, 0.0042, ..., 0.0010, 0.0025, 0.0004], + [-0.0218, -0.0076, 0.0026, ..., 0.0016, 0.0008, 0.0013], + ..., + [-0.0017, -0.0050, 0.0027, ..., -0.0013, 0.0028, -0.0038], + [ 0.0105, -0.0063, -0.0068, ..., -0.0011, -0.0013, 0.0034], + [ 0.0126, 0.0043, 0.0100, ..., 0.0003, 0.0115, 0.0250]], + device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0275, -0.0158, -0.0063, 0.0006, 0.0002, -0.0028, -0.0108, -0.0046, + 0.0307, 0.0198], device='cuda:0'), grad: tensor([ 0.0071, 0.0078, -0.0177, -0.0286, 0.0060, 0.0042, 0.0003, -0.0055, + 0.0024, 0.0240], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 275.44, cls_loss 0.0948 cls_loss_mapping 0.1475 cls_loss_causal 1.1582 re_mapping 0.0588 re_causal 0.1363 /// teacc 97.49 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0159, -0.0307, 0.0018, ..., 0.0284, -0.0487, 0.0112], + [ 0.0049, -0.0366, -0.0364, ..., -0.0258, 0.0102, -0.0351], + [ 0.0153, 0.0370, -0.0241, ..., -0.0430, -0.0346, -0.0128], + ..., + [-0.0077, 0.0052, 0.0073, ..., -0.0423, -0.0474, 0.0269], + [-0.0211, 0.0294, 0.0339, ..., -0.0176, 0.0107, -0.0284], + [-0.0296, -0.0169, 0.0213, ..., 0.0035, -0.0204, 0.0202]], + device='cuda:0'), grad: tensor([[-9.4080e-04, 6.1035e-04, 1.6289e-03, ..., 2.5864e-03, + 1.5965e-03, 1.2102e-03], + [ 2.3592e-04, -1.4675e-04, -1.8759e-03, ..., 9.8467e-05, + -3.3894e-03, 8.9183e-06], + [-1.1005e-03, 1.7726e-04, 1.0118e-03, ..., 5.4932e-04, + 1.4162e-03, 1.1158e-04], + ..., + [ 4.2230e-05, 4.5359e-05, 1.6916e-04, ..., 6.6698e-05, + 3.3689e-04, -6.2513e-04], + [ 1.0502e-04, -3.4027e-03, -3.3131e-03, ..., 1.8907e-04, + -4.8180e-03, 2.4939e-04], + [ 5.8460e-04, 2.6584e-04, 4.5967e-04, ..., 1.7285e-04, + 1.3924e-04, -2.4247e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0272, -0.0162, -0.0062, 0.0007, 0.0005, -0.0029, -0.0110, -0.0046, + 0.0307, 0.0201], device='cuda:0'), grad: tensor([ 3.6240e-05, -1.0551e-02, 3.1033e-03, 2.2335e-03, 6.9542e-03, + 2.2984e-03, -2.6703e-03, 4.8518e-04, -3.0575e-03, 1.1654e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 259.07, cls_loss 0.0863 cls_loss_mapping 0.1368 cls_loss_causal 1.1418 re_mapping 0.0532 re_causal 0.1240 /// teacc 97.38 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0160, -0.0312, 0.0003, ..., 0.0278, -0.0507, 0.0104], + [ 0.0042, -0.0381, -0.0380, ..., -0.0272, 0.0094, -0.0347], + [ 0.0164, 0.0377, -0.0252, ..., -0.0448, -0.0364, -0.0136], + ..., + [-0.0072, 0.0056, 0.0071, ..., -0.0394, -0.0488, 0.0273], + [-0.0214, 0.0300, 0.0349, ..., -0.0188, 0.0113, -0.0301], + [-0.0303, -0.0171, 0.0210, ..., 0.0034, -0.0205, 0.0206]], + device='cuda:0'), grad: tensor([[-2.1534e-03, -5.0116e-04, 4.2892e-04, ..., 1.6057e-04, + 5.1308e-04, -2.2049e-03], + [ 8.5831e-04, 6.3992e-04, 1.8988e-03, ..., -2.0790e-04, + 2.4452e-03, 7.0620e-04], + [-1.3695e-03, -2.2292e-04, 2.9240e-03, ..., 4.0960e-04, + 3.6659e-03, 2.9430e-03], + ..., + [ 1.2608e-03, 7.4530e-04, 2.9731e-04, ..., 6.2227e-05, + 8.0490e-04, 1.6470e-03], + [ 3.3307e-04, -2.5196e-03, -1.0853e-03, ..., -1.3084e-03, + -2.2049e-03, 1.9054e-03], + [ 1.6022e-04, -1.5306e-04, 8.3590e-04, ..., 8.1420e-05, + 3.0041e-04, -3.6693e-04]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0272, -0.0163, -0.0059, 0.0012, 0.0007, -0.0031, -0.0112, -0.0047, + 0.0305, 0.0200], device='cuda:0'), grad: tensor([-0.0053, 0.0016, 0.0048, -0.0148, 0.0081, 0.0005, 0.0014, 0.0035, + -0.0003, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 276.13, cls_loss 0.0698 cls_loss_mapping 0.1133 cls_loss_causal 1.0684 re_mapping 0.0501 re_causal 0.1200 /// teacc 97.53 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0157, -0.0318, -0.0004, ..., 0.0270, -0.0518, 0.0098], + [ 0.0035, -0.0390, -0.0397, ..., -0.0275, 0.0083, -0.0343], + [ 0.0171, 0.0380, -0.0266, ..., -0.0450, -0.0378, -0.0143], + ..., + [-0.0067, 0.0056, 0.0067, ..., -0.0390, -0.0499, 0.0280], + [-0.0209, 0.0313, 0.0360, ..., -0.0191, 0.0123, -0.0307], + [-0.0311, -0.0173, 0.0208, ..., 0.0027, -0.0206, 0.0210]], + device='cuda:0'), grad: tensor([[-1.9658e-04, 5.9307e-05, 6.4611e-05, ..., -2.8804e-05, + 7.4029e-05, -7.6115e-05], + [ 4.1604e-04, 2.2936e-04, 7.2896e-05, ..., 1.5704e-07, + 1.4246e-04, 2.1124e-04], + [-1.3828e-04, -2.3472e-04, 2.4080e-04, ..., 3.2429e-06, + 2.5415e-04, 3.6740e-04], + ..., + [-4.5443e-04, -5.1737e-04, 1.2994e-04, ..., 4.3027e-07, + 2.0766e-04, -8.0395e-04], + [ 3.2616e-04, 2.5225e-04, 4.6062e-04, ..., 1.8766e-06, + 4.5991e-04, 4.1175e-04], + [ 2.9492e-04, 1.1909e-04, 2.9817e-05, ..., 1.6894e-06, + 7.3314e-06, 4.8965e-05]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0273, -0.0163, -0.0058, 0.0009, 0.0006, -0.0032, -0.0116, -0.0046, + 0.0311, 0.0200], device='cuda:0'), grad: tensor([-4.8041e-04, 3.2449e-04, 7.5960e-04, -1.2484e-03, -7.0906e-04, + -7.9393e-05, 5.6791e-04, -2.0337e-04, 8.4162e-04, 2.2495e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 275.13, cls_loss 0.0733 cls_loss_mapping 0.1179 cls_loss_causal 1.0573 re_mapping 0.0450 re_causal 0.1138 /// teacc 97.79 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0158, -0.0325, -0.0011, ..., 0.0252, -0.0530, 0.0094], + [ 0.0027, -0.0400, -0.0405, ..., -0.0281, 0.0080, -0.0346], + [ 0.0181, 0.0382, -0.0282, ..., -0.0455, -0.0396, -0.0149], + ..., + [-0.0066, 0.0057, 0.0059, ..., -0.0382, -0.0510, 0.0284], + [-0.0208, 0.0320, 0.0364, ..., -0.0200, 0.0124, -0.0319], + [-0.0319, -0.0176, 0.0207, ..., 0.0017, -0.0208, 0.0211]], + device='cuda:0'), grad: tensor([[-4.0293e-05, 1.3294e-03, 6.8521e-04, ..., 2.5063e-03, + 1.4372e-03, 1.1206e-03], + [ 3.0756e-04, 1.4949e-04, 1.5473e-04, ..., 2.2709e-05, + 2.1458e-05, -5.8031e-04], + [-2.1782e-03, 4.8614e-04, 2.3127e-04, ..., 8.4758e-05, + 2.5344e-04, 2.5105e-04], + ..., + [-1.8482e-03, -2.4815e-03, 2.4700e-04, ..., 1.1981e-05, + 2.2495e-04, -3.7346e-03], + [-1.3113e-04, -1.0214e-03, 1.0830e-04, ..., 1.1653e-04, + -5.3406e-04, 9.6178e-04], + [ 1.0929e-03, 1.1120e-03, 6.8951e-04, ..., 9.6202e-05, + 4.7135e-04, 1.5631e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0275, -0.0162, -0.0058, 0.0013, 0.0008, -0.0034, -0.0119, -0.0047, + 0.0310, 0.0198], device='cuda:0'), grad: tensor([ 2.3193e-03, -1.3065e-03, -5.6601e-04, 3.1261e-03, 1.6251e-03, + -2.4681e-03, -1.7252e-03, -2.4624e-03, -2.5451e-05, 1.4830e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 275.01, cls_loss 0.0677 cls_loss_mapping 0.1051 cls_loss_causal 1.0156 re_mapping 0.0419 re_causal 0.1046 /// teacc 97.96 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0159, -0.0332, -0.0023, ..., 0.0248, -0.0545, 0.0091], + [ 0.0016, -0.0416, -0.0421, ..., -0.0291, 0.0072, -0.0346], + [ 0.0189, 0.0386, -0.0294, ..., -0.0465, -0.0412, -0.0155], + ..., + [-0.0064, 0.0063, 0.0051, ..., -0.0377, -0.0522, 0.0287], + [-0.0206, 0.0332, 0.0372, ..., -0.0213, 0.0130, -0.0323], + [-0.0323, -0.0176, 0.0201, ..., 0.0008, -0.0214, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.3990e-03, 7.7133e-03, 4.5547e-03, ..., 3.2444e-03, + 4.1695e-03, 1.7941e-04], + [ 6.0129e-04, 6.5851e-04, 2.9945e-04, ..., 8.0943e-05, + 2.5225e-04, 2.2936e-04], + [ 1.3189e-03, 2.4948e-03, 6.9475e-04, ..., 3.8910e-04, + 6.5422e-04, 1.1539e-03], + ..., + [-3.8185e-03, -6.2895e-04, 3.1328e-04, ..., 2.8044e-05, + 3.4261e-04, -4.0932e-03], + [-1.9951e-03, 2.5024e-03, 1.0941e-02, ..., 5.8289e-03, + 1.0803e-02, 8.0681e-04], + [-9.2220e-04, -1.0592e-04, -1.7118e-03, ..., 1.9836e-04, + -3.2444e-03, -1.8850e-05]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0275, -0.0166, -0.0059, 0.0011, 0.0013, -0.0034, -0.0121, -0.0048, + 0.0314, 0.0198], device='cuda:0'), grad: tensor([ 0.0086, 0.0004, 0.0040, 0.0004, 0.0041, -0.0027, -0.0134, -0.0040, + 0.0061, -0.0036], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 259.21, cls_loss 0.0587 cls_loss_mapping 0.0894 cls_loss_causal 0.9640 re_mapping 0.0414 re_causal 0.1019 /// teacc 97.94 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0162, -0.0341, -0.0034, ..., 0.0234, -0.0562, 0.0086], + [ 0.0008, -0.0426, -0.0431, ..., -0.0305, 0.0063, -0.0338], + [ 0.0196, 0.0385, -0.0310, ..., -0.0478, -0.0428, -0.0163], + ..., + [-0.0059, 0.0071, 0.0048, ..., -0.0376, -0.0526, 0.0290], + [-0.0201, 0.0340, 0.0378, ..., -0.0226, 0.0137, -0.0334], + [-0.0327, -0.0175, 0.0196, ..., 0.0003, -0.0218, 0.0219]], + device='cuda:0'), grad: tensor([[-2.4796e-03, -1.9350e-03, -4.6396e-04, ..., 5.8442e-05, + 1.6081e-04, -2.4204e-03], + [ 4.5705e-04, 1.6272e-04, 1.8334e-04, ..., 2.9579e-05, + 1.8096e-04, 2.1964e-05], + [ 1.2755e-04, 6.9666e-04, 3.7432e-04, ..., 3.8892e-05, + 2.8539e-04, 3.1447e-04], + ..., + [ 3.2496e-04, 3.2276e-05, 1.6832e-04, ..., 1.0371e-05, + 2.5487e-04, -1.5993e-03], + [-1.0300e-03, -1.4858e-03, -7.5102e-04, ..., 7.2062e-05, + -7.7200e-04, 2.6107e-04], + [ 3.6240e-04, 4.4346e-04, 2.7990e-04, ..., 3.4750e-05, + 2.0468e-04, 1.4162e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0273, -0.0165, -0.0061, 0.0010, 0.0010, -0.0032, -0.0121, -0.0047, + 0.0318, 0.0198], device='cuda:0'), grad: tensor([-0.0072, 0.0001, 0.0013, 0.0072, -0.0003, -0.0020, 0.0002, -0.0007, + -0.0009, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 273.71, cls_loss 0.0544 cls_loss_mapping 0.0873 cls_loss_causal 0.9824 re_mapping 0.0371 re_causal 0.0995 /// teacc 98.09 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0161, -0.0346, -0.0045, ..., 0.0224, -0.0576, 0.0081], + [-0.0001, -0.0436, -0.0440, ..., -0.0319, 0.0057, -0.0328], + [ 0.0200, 0.0387, -0.0328, ..., -0.0487, -0.0445, -0.0171], + ..., + [-0.0051, 0.0074, 0.0043, ..., -0.0373, -0.0536, 0.0296], + [-0.0196, 0.0353, 0.0387, ..., -0.0229, 0.0144, -0.0343], + [-0.0337, -0.0178, 0.0194, ..., -0.0004, -0.0220, 0.0218]], + device='cuda:0'), grad: tensor([[ 2.3508e-04, 3.7694e-04, 5.9843e-04, ..., 1.8251e-04, + 5.4359e-04, 2.1160e-04], + [ 1.4219e-03, 4.7922e-04, 1.7297e-04, ..., 1.2279e-05, + 1.6761e-04, 1.3409e-03], + [-1.8711e-03, -6.0415e-04, 2.1315e-04, ..., 1.8567e-05, + 1.6785e-04, 1.6475e-04], + ..., + [-2.1744e-03, -8.3065e-04, 5.0735e-04, ..., 3.1330e-06, + 4.6039e-04, -3.9368e-03], + [-3.4885e-03, 1.0394e-05, -8.5020e-04, ..., 1.0908e-04, + 8.2016e-04, 7.2479e-04], + [ 8.4543e-04, -1.8082e-03, -3.4943e-03, ..., 1.7852e-05, + -6.1531e-03, 6.4754e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0275, -0.0163, -0.0066, 0.0013, 0.0011, -0.0034, -0.0123, -0.0043, + 0.0319, 0.0194], device='cuda:0'), grad: tensor([ 0.0008, 0.0022, -0.0034, 0.0082, 0.0017, -0.0023, 0.0025, -0.0035, + -0.0007, -0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 255.52, cls_loss 0.0535 cls_loss_mapping 0.0839 cls_loss_causal 0.9495 re_mapping 0.0369 re_causal 0.0959 /// teacc 98.03 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0161, -0.0355, -0.0055, ..., 0.0219, -0.0588, 0.0076], + [-0.0012, -0.0451, -0.0455, ..., -0.0341, 0.0048, -0.0325], + [ 0.0209, 0.0394, -0.0337, ..., -0.0489, -0.0461, -0.0179], + ..., + [-0.0050, 0.0077, 0.0038, ..., -0.0371, -0.0539, 0.0299], + [-0.0194, 0.0363, 0.0394, ..., -0.0231, 0.0149, -0.0351], + [-0.0334, -0.0175, 0.0193, ..., -0.0017, -0.0222, 0.0220]], + device='cuda:0'), grad: tensor([[ 8.1062e-04, 6.0886e-05, 7.7844e-05, ..., 4.9502e-05, + 6.5386e-05, 7.7784e-05], + [ 9.8646e-05, 1.2493e-04, 9.1553e-05, ..., 4.6730e-05, + 9.1672e-05, -7.5054e-04], + [-5.7869e-03, 1.0622e-04, 5.1498e-05, ..., 3.4809e-05, + 6.2287e-05, 4.8709e-04], + ..., + [-2.0826e-04, -6.9618e-04, 2.2948e-04, ..., -5.6401e-06, + 7.0989e-05, 4.9925e-04], + [ 2.9016e-04, 1.0705e-04, 2.0409e-04, ..., 6.8903e-05, + 1.7440e-04, 2.5558e-04], + [ 2.2678e-03, 7.3075e-05, 2.8000e-03, ..., 1.7270e-05, + 1.6174e-03, 5.2071e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0273, -0.0166, -0.0061, 0.0014, 0.0010, -0.0038, -0.0124, -0.0043, + 0.0318, 0.0197], device='cuda:0'), grad: tensor([ 0.0008, -0.0011, -0.0052, 0.0041, -0.0101, -0.0021, -0.0001, 0.0015, + 0.0007, 0.0117], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 266.73, cls_loss 0.0541 cls_loss_mapping 0.0815 cls_loss_causal 0.9760 re_mapping 0.0351 re_causal 0.0910 /// teacc 98.33 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0158, -0.0360, -0.0063, ..., 0.0210, -0.0598, 0.0075], + [-0.0024, -0.0461, -0.0466, ..., -0.0349, 0.0040, -0.0329], + [ 0.0219, 0.0398, -0.0343, ..., -0.0492, -0.0470, -0.0183], + ..., + [-0.0045, 0.0082, 0.0036, ..., -0.0362, -0.0550, 0.0306], + [-0.0193, 0.0369, 0.0397, ..., -0.0231, 0.0153, -0.0363], + [-0.0340, -0.0176, 0.0191, ..., -0.0032, -0.0221, 0.0221]], + device='cuda:0'), grad: tensor([[-1.6904e-04, 2.6059e-04, 1.0478e-04, ..., 1.0207e-05, + 1.5473e-04, -2.7204e-04], + [ 1.5950e-04, 2.2149e-04, 1.3673e-04, ..., 1.3456e-05, + -3.1638e-04, -5.9843e-04], + [-1.1053e-03, -4.2844e-04, 1.3351e-04, ..., 1.8314e-05, + 3.8862e-04, 5.2977e-04], + ..., + [ 1.2720e-04, 4.7088e-04, 1.1027e-04, ..., 2.7604e-06, + 2.4080e-04, 7.0286e-04], + [ 2.1291e-04, 1.5929e-05, 8.6010e-05, ..., 2.0254e-04, + 1.2672e-04, 5.3930e-04], + [-9.3341e-05, -2.0313e-03, 6.1846e-04, ..., 1.6630e-05, + -1.5008e-04, -2.4643e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0274, -0.0172, -0.0061, 0.0014, 0.0010, -0.0037, -0.0125, -0.0038, + 0.0319, 0.0197], device='cuda:0'), grad: tensor([-1.1120e-03, -1.1749e-03, -3.9399e-05, 4.1175e-04, 2.1667e-03, + -2.9516e-04, 1.1129e-03, 1.4925e-03, 9.9182e-04, -3.5572e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 248.92, cls_loss 0.0422 cls_loss_mapping 0.0719 cls_loss_causal 0.9324 re_mapping 0.0328 re_causal 0.0915 /// teacc 98.31 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0159, -0.0364, -0.0071, ..., 0.0199, -0.0609, 0.0074], + [-0.0032, -0.0469, -0.0471, ..., -0.0355, 0.0037, -0.0324], + [ 0.0227, 0.0400, -0.0353, ..., -0.0498, -0.0484, -0.0187], + ..., + [-0.0047, 0.0080, 0.0031, ..., -0.0358, -0.0558, 0.0310], + [-0.0188, 0.0376, 0.0405, ..., -0.0242, 0.0158, -0.0372], + [-0.0344, -0.0175, 0.0184, ..., -0.0034, -0.0229, 0.0222]], + device='cuda:0'), grad: tensor([[-1.3781e-03, 4.9293e-05, 1.1772e-04, ..., -8.6010e-05, + 9.2566e-05, -7.7009e-04], + [ 5.9456e-05, 1.5724e-04, 2.1446e-04, ..., 3.7163e-05, + 1.6594e-04, -1.8701e-05], + [ 2.9325e-04, 8.0764e-05, 5.5403e-05, ..., 3.3051e-05, + 4.4674e-05, 1.9526e-04], + ..., + [ 3.6764e-04, -1.5700e-04, 2.8923e-05, ..., 1.3009e-05, + 3.3170e-05, 3.9268e-04], + [ 6.3539e-05, -3.9291e-04, -3.4118e-04, ..., 1.3554e-04, + -1.8382e-04, 1.4913e-04], + [ 7.1228e-05, -6.2287e-05, 1.4293e-04, ..., 4.0114e-05, + 7.3671e-05, -1.8253e-03]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0275, -0.0172, -0.0063, 0.0017, 0.0011, -0.0038, -0.0127, -0.0041, + 0.0320, 0.0197], device='cuda:0'), grad: tensor([-2.4071e-03, 1.0598e-04, 6.0320e-04, 1.0967e-03, 1.0214e-03, + -9.9301e-05, 5.5981e-04, 1.1187e-03, -2.9898e-04, -1.6975e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 248.69, cls_loss 0.0388 cls_loss_mapping 0.0684 cls_loss_causal 0.9245 re_mapping 0.0321 re_causal 0.0886 /// teacc 97.92 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0160, -0.0373, -0.0083, ..., 0.0188, -0.0621, 0.0071], + [-0.0038, -0.0477, -0.0477, ..., -0.0358, 0.0035, -0.0314], + [ 0.0234, 0.0403, -0.0359, ..., -0.0505, -0.0495, -0.0199], + ..., + [-0.0044, 0.0083, 0.0027, ..., -0.0351, -0.0565, 0.0316], + [-0.0189, 0.0384, 0.0411, ..., -0.0240, 0.0163, -0.0382], + [-0.0349, -0.0175, 0.0182, ..., -0.0042, -0.0230, 0.0225]], + device='cuda:0'), grad: tensor([[-1.2422e-04, 2.4155e-05, 3.8862e-05, ..., 2.3797e-05, + 4.2737e-05, 7.8380e-06], + [ 8.8334e-05, -1.2636e-04, 1.9193e-04, ..., 1.5363e-05, + 3.2043e-04, -4.8685e-04], + [ 2.4915e-04, 2.8706e-04, 2.3627e-04, ..., 3.6955e-05, + 4.0126e-04, 5.2261e-04], + ..., + [-2.1803e-04, 1.2048e-05, 2.0432e-04, ..., 1.1012e-05, + 4.2081e-04, 4.2653e-04], + [ 1.6057e-04, 3.1185e-04, 2.1636e-04, ..., 1.0586e-04, + 2.8443e-04, 3.0160e-04], + [-1.1463e-03, -1.4400e-03, -7.0453e-05, ..., 2.1845e-05, + -7.5626e-04, -7.3242e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0271, -0.0169, -0.0065, 0.0018, 0.0014, -0.0037, -0.0127, -0.0039, + 0.0320, 0.0195], device='cuda:0'), grad: tensor([-0.0004, -0.0009, 0.0012, -0.0014, 0.0010, 0.0003, 0.0009, 0.0013, + 0.0007, -0.0028], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 265.54, cls_loss 0.0465 cls_loss_mapping 0.0705 cls_loss_causal 0.9298 re_mapping 0.0297 re_causal 0.0807 /// teacc 98.35 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0163, -0.0379, -0.0095, ..., 0.0180, -0.0633, 0.0068], + [-0.0052, -0.0483, -0.0485, ..., -0.0367, 0.0028, -0.0321], + [ 0.0241, 0.0399, -0.0372, ..., -0.0516, -0.0506, -0.0204], + ..., + [-0.0037, 0.0087, 0.0021, ..., -0.0348, -0.0574, 0.0326], + [-0.0186, 0.0392, 0.0417, ..., -0.0247, 0.0166, -0.0391], + [-0.0353, -0.0175, 0.0176, ..., -0.0050, -0.0234, 0.0225]], + device='cuda:0'), grad: tensor([[-9.7036e-04, -3.2640e-04, 6.1870e-05, ..., 3.1501e-05, + -1.2577e-04, 5.6565e-05], + [ 5.3465e-05, 1.5843e-04, 1.9825e-04, ..., 1.5843e-04, + 8.2016e-05, 8.2180e-06], + [ 1.0210e-04, 6.5029e-05, 1.0157e-04, ..., 4.9949e-05, + 6.1154e-05, 8.2195e-05], + ..., + [ 3.2544e-05, 4.5681e-04, 8.6784e-04, ..., 3.0518e-04, + 3.1543e-04, 1.0329e-04], + [ 9.1553e-05, -4.6587e-04, -4.8685e-04, ..., 7.5698e-05, + -1.9121e-04, 1.8632e-04], + [ 1.8704e-04, -1.8001e-04, 2.7370e-04, ..., 1.1414e-04, + 3.3522e-04, -8.9264e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0268, -0.0176, -0.0064, 0.0013, 0.0015, -0.0030, -0.0129, -0.0034, + 0.0319, 0.0196], device='cuda:0'), grad: tensor([-1.6756e-03, 5.9652e-04, 4.3035e-04, 8.8632e-05, -6.3782e-03, + 1.2884e-03, 3.1605e-03, 1.9875e-03, 7.6199e-04, -2.6464e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 265.11, cls_loss 0.0410 cls_loss_mapping 0.0690 cls_loss_causal 0.8732 re_mapping 0.0294 re_causal 0.0804 /// teacc 98.45 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0159, -0.0382, -0.0100, ..., 0.0174, -0.0642, 0.0067], + [-0.0061, -0.0493, -0.0493, ..., -0.0380, 0.0025, -0.0327], + [ 0.0245, 0.0401, -0.0379, ..., -0.0521, -0.0516, -0.0204], + ..., + [-0.0034, 0.0086, 0.0019, ..., -0.0343, -0.0583, 0.0327], + [-0.0182, 0.0399, 0.0419, ..., -0.0259, 0.0168, -0.0396], + [-0.0352, -0.0175, 0.0175, ..., -0.0057, -0.0237, 0.0234]], + device='cuda:0'), grad: tensor([[ 3.5262e-04, 8.0156e-04, 1.1501e-03, ..., 4.7684e-04, + 7.1669e-04, 4.0317e-04], + [ 4.8614e-04, 4.9162e-04, 1.4758e-04, ..., 6.2771e-06, + 1.7190e-04, 1.9085e-04], + [ 7.7209e-03, 6.3438e-03, 1.5914e-04, ..., 5.8934e-06, + 1.8752e-04, 5.5504e-03], + ..., + [-1.5793e-02, -1.2459e-02, 6.7353e-05, ..., 1.3132e-06, + -2.8777e-04, -1.1002e-02], + [ 6.0511e-04, 6.8724e-05, -4.5848e-04, ..., 1.1516e-04, + -4.6206e-04, 8.1968e-04], + [ 6.4087e-04, 3.4547e-04, 2.5797e-04, ..., 3.8058e-05, + 2.8110e-04, 2.1100e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0270, -0.0180, -0.0064, 0.0011, 0.0014, -0.0029, -0.0132, -0.0034, + 0.0319, 0.0201], device='cuda:0'), grad: tensor([ 0.0016, 0.0005, 0.0100, 0.0068, 0.0002, 0.0020, -0.0024, -0.0201, + 0.0010, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 248.71, cls_loss 0.0329 cls_loss_mapping 0.0601 cls_loss_causal 0.8950 re_mapping 0.0283 re_causal 0.0801 /// teacc 98.33 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0154, -0.0385, -0.0106, ..., 0.0164, -0.0652, 0.0066], + [-0.0070, -0.0502, -0.0500, ..., -0.0389, 0.0021, -0.0320], + [ 0.0254, 0.0405, -0.0382, ..., -0.0527, -0.0527, -0.0211], + ..., + [-0.0031, 0.0091, 0.0014, ..., -0.0345, -0.0588, 0.0330], + [-0.0182, 0.0404, 0.0422, ..., -0.0264, 0.0171, -0.0407], + [-0.0357, -0.0178, 0.0168, ..., -0.0051, -0.0241, 0.0238]], + device='cuda:0'), grad: tensor([[ 6.5923e-05, 3.4118e-04, 1.9467e-04, ..., 9.7334e-05, + 7.2300e-05, 7.4983e-05], + [ 2.6393e-04, 2.5797e-04, 1.5342e-04, ..., 4.2140e-05, + 5.9158e-05, -1.8568e-03], + [-1.5659e-03, -2.5787e-03, -8.6737e-04, ..., -7.2336e-04, + 4.8757e-05, 2.1279e-04], + ..., + [-4.4560e-04, -6.5851e-04, 1.0610e-04, ..., 3.5703e-05, + 5.7638e-05, -6.3753e-04], + [-5.2834e-04, -1.8835e-03, -1.9093e-03, ..., 1.7679e-04, + -1.7004e-03, 1.1510e-04], + [ 1.0519e-03, 2.4071e-03, 1.5182e-03, ..., 2.9802e-05, + 1.1587e-03, 7.0286e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0273, -0.0177, -0.0065, 0.0010, 0.0019, -0.0029, -0.0136, -0.0036, + 0.0317, 0.0201], device='cuda:0'), grad: tensor([ 0.0002, -0.0016, -0.0040, 0.0045, 0.0010, -0.0024, 0.0012, -0.0007, + -0.0017, 0.0036], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 265.08, cls_loss 0.0302 cls_loss_mapping 0.0552 cls_loss_causal 0.8820 re_mapping 0.0280 re_causal 0.0819 /// teacc 98.52 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0156, -0.0393, -0.0121, ..., 0.0154, -0.0668, 0.0063], + [-0.0079, -0.0510, -0.0506, ..., -0.0395, 0.0017, -0.0316], + [ 0.0258, 0.0405, -0.0390, ..., -0.0535, -0.0539, -0.0217], + ..., + [-0.0031, 0.0093, 0.0009, ..., -0.0326, -0.0594, 0.0335], + [-0.0181, 0.0411, 0.0426, ..., -0.0263, 0.0175, -0.0417], + [-0.0355, -0.0175, 0.0166, ..., -0.0067, -0.0242, 0.0242]], + device='cuda:0'), grad: tensor([[ 1.9288e-04, 3.1853e-04, 4.6849e-04, ..., 2.2495e-04, + 3.4189e-04, 1.9097e-04], + [ 1.0002e-04, 9.2566e-05, 6.3598e-05, ..., 1.1608e-05, + 6.3777e-05, -1.6987e-04], + [-3.0441e-03, -2.3632e-03, -6.2180e-04, ..., 1.3955e-05, + -3.0875e-04, 5.2392e-05], + ..., + [-1.4620e-03, -1.3533e-03, 4.2647e-05, ..., 2.0321e-06, + 4.6700e-05, -1.4238e-03], + [ 1.9369e-03, 1.5869e-03, 6.4135e-04, ..., 8.5175e-05, + 4.6968e-04, 1.0777e-04], + [ 1.5059e-03, 1.3914e-03, 2.5177e-04, ..., 1.3500e-05, + 3.3808e-04, 1.3018e-03]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0270, -0.0176, -0.0065, 0.0010, 0.0020, -0.0025, -0.0137, -0.0037, + 0.0315, 0.0203], device='cuda:0'), grad: tensor([ 6.6519e-04, -7.0751e-05, -3.8586e-03, 2.3317e-04, -4.5433e-03, + 6.1846e-04, -8.1968e-04, -1.2159e-03, 3.2139e-03, 5.7793e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 248.33, cls_loss 0.0266 cls_loss_mapping 0.0512 cls_loss_causal 0.8695 re_mapping 0.0272 re_causal 0.0773 /// teacc 98.31 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0154, -0.0398, -0.0129, ..., 0.0145, -0.0677, 0.0059], + [-0.0084, -0.0520, -0.0520, ..., -0.0406, 0.0009, -0.0303], + [ 0.0262, 0.0408, -0.0399, ..., -0.0541, -0.0552, -0.0226], + ..., + [-0.0026, 0.0099, 0.0003, ..., -0.0311, -0.0600, 0.0339], + [-0.0181, 0.0417, 0.0429, ..., -0.0270, 0.0178, -0.0423], + [-0.0359, -0.0175, 0.0166, ..., -0.0070, -0.0242, 0.0242]], + device='cuda:0'), grad: tensor([[-1.1568e-03, 1.0633e-04, -1.7405e-03, ..., -2.3818e-04, + -5.8746e-04, -1.0653e-03], + [ 1.2720e-04, 1.5497e-04, 1.9348e-04, ..., 3.5197e-05, + 2.1064e-04, -1.2266e-06], + [ 8.9884e-04, 1.0729e-03, 8.0633e-04, ..., 9.1672e-05, + 8.7547e-04, 8.9169e-05], + ..., + [ 2.8992e-04, 2.1660e-04, 2.5606e-04, ..., 9.4593e-05, + 1.8406e-04, 3.6263e-04], + [-1.6766e-03, -2.2373e-03, -1.6623e-03, ..., 4.2647e-05, + -1.6565e-03, -8.2910e-05], + [ 1.9705e-04, 1.9133e-04, 7.5531e-04, ..., 5.4926e-05, + 5.6982e-04, -3.4118e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0270, -0.0174, -0.0067, 0.0014, 0.0019, -0.0025, -0.0139, -0.0035, + 0.0314, 0.0200], device='cuda:0'), grad: tensor([-3.1815e-03, 3.1400e-04, 1.5593e-03, 1.1311e-03, 6.0111e-05, + 1.6904e-04, 1.0319e-03, 8.1491e-04, -2.4586e-03, 5.5885e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 248.55, cls_loss 0.0304 cls_loss_mapping 0.0529 cls_loss_causal 0.8876 re_mapping 0.0261 re_causal 0.0764 /// teacc 97.92 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0151, -0.0408, -0.0136, ..., 0.0138, -0.0688, 0.0056], + [-0.0094, -0.0527, -0.0527, ..., -0.0415, 0.0006, -0.0300], + [ 0.0267, 0.0410, -0.0407, ..., -0.0547, -0.0564, -0.0229], + ..., + [-0.0025, 0.0103, -0.0002, ..., -0.0309, -0.0605, 0.0341], + [-0.0174, 0.0425, 0.0436, ..., -0.0273, 0.0184, -0.0432], + [-0.0362, -0.0181, 0.0160, ..., -0.0077, -0.0247, 0.0243]], + device='cuda:0'), grad: tensor([[ 7.2837e-05, 1.9276e-04, 2.7373e-05, ..., 5.1379e-05, + 6.6459e-05, 8.2374e-05], + [ 2.9278e-04, 4.7493e-04, 1.5289e-05, ..., 6.7428e-06, + 1.0240e-04, 2.4915e-04], + [ 9.8133e-04, 1.3819e-03, 1.5542e-05, ..., 1.0543e-05, + 4.1127e-04, 1.1082e-03], + ..., + [-7.2575e-04, -1.4744e-03, 1.0200e-05, ..., 2.6100e-07, + 2.3472e-04, -9.4080e-04], + [ 4.9353e-04, 6.6566e-04, 9.0718e-05, ..., 5.1647e-05, + 4.2319e-04, 4.2772e-04], + [ 1.7190e-04, -8.9854e-06, 1.4269e-04, ..., 2.7977e-06, + 5.5456e-04, -1.9205e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0270, -0.0174, -0.0068, 0.0013, 0.0025, -0.0030, -0.0137, -0.0035, + 0.0318, 0.0196], device='cuda:0'), grad: tensor([ 1.9646e-04, 6.7043e-04, 2.3174e-03, -3.3455e-03, -1.5345e-03, + 3.9905e-05, 2.0593e-05, -1.4534e-03, 1.4467e-03, 1.6403e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 264.65, cls_loss 0.0275 cls_loss_mapping 0.0484 cls_loss_causal 0.8157 re_mapping 0.0243 re_causal 0.0696 /// teacc 98.56 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0147, -0.0416, -0.0145, ..., 0.0127, -0.0697, 0.0053], + [-0.0105, -0.0538, -0.0533, ..., -0.0422, 0.0002, -0.0298], + [ 0.0275, 0.0416, -0.0414, ..., -0.0551, -0.0572, -0.0236], + ..., + [-0.0024, 0.0104, -0.0006, ..., -0.0310, -0.0613, 0.0346], + [-0.0170, 0.0430, 0.0441, ..., -0.0270, 0.0189, -0.0434], + [-0.0366, -0.0178, 0.0156, ..., -0.0074, -0.0251, 0.0244]], + device='cuda:0'), grad: tensor([[ 2.6870e-04, 2.1863e-04, 2.1017e-04, ..., 6.9797e-05, + 1.2970e-04, 1.6117e-04], + [-3.7789e-04, -2.7180e-04, -2.0385e-04, ..., 4.2319e-05, + 8.8096e-05, -2.9888e-03], + [ 3.2020e-04, -2.2876e-04, 2.8992e-04, ..., -6.6817e-05, + 4.7612e-04, 1.0691e-03], + ..., + [ 2.7313e-03, 2.8419e-03, 2.5892e-04, ..., 1.5944e-05, + 3.6430e-04, 2.2526e-03], + [ 1.8959e-03, 2.5539e-03, 3.0518e-03, ..., 1.5783e-03, + 2.0542e-03, 1.4973e-03], + [ 1.5383e-03, 1.3189e-03, -9.0122e-04, ..., 1.4193e-05, + -3.7217e-04, 4.2319e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0269, -0.0175, -0.0065, 0.0010, 0.0022, -0.0025, -0.0137, -0.0033, + 0.0319, 0.0193], device='cuda:0'), grad: tensor([ 0.0006, -0.0063, 0.0019, -0.0033, -0.0092, 0.0023, -0.0027, 0.0066, + 0.0058, 0.0042], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 266.96, cls_loss 0.0283 cls_loss_mapping 0.0480 cls_loss_causal 0.8021 re_mapping 0.0240 re_causal 0.0671 /// teacc 98.64 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0148, -0.0420, -0.0153, ..., 0.0122, -0.0707, 0.0052], + [-0.0114, -0.0547, -0.0539, ..., -0.0427, -0.0003, -0.0297], + [ 0.0283, 0.0424, -0.0412, ..., -0.0560, -0.0576, -0.0238], + ..., + [-0.0019, 0.0105, -0.0013, ..., -0.0304, -0.0627, 0.0350], + [-0.0171, 0.0436, 0.0447, ..., -0.0264, 0.0194, -0.0441], + [-0.0370, -0.0178, 0.0154, ..., -0.0079, -0.0253, 0.0244]], + device='cuda:0'), grad: tensor([[-1.4496e-03, 4.0442e-05, 1.0258e-04, ..., 2.4110e-05, + 3.5912e-05, -5.0211e-04], + [ 9.3102e-05, 3.8326e-05, 3.6180e-05, ..., 3.6508e-06, + 4.6134e-05, 1.7464e-05], + [ 2.7275e-04, -1.7929e-04, 3.0443e-05, ..., -6.2771e-07, + 5.2184e-05, 3.0565e-04], + ..., + [ 5.5611e-05, 3.9697e-05, 5.5224e-05, ..., 7.2420e-06, + 6.0439e-05, -1.5378e-04], + [ 4.2105e-04, 8.3804e-05, 3.6240e-04, ..., 6.2287e-05, + 3.2258e-04, 2.5177e-04], + [-4.5240e-05, -1.0157e-04, 6.4731e-05, ..., 1.6332e-05, + 4.2379e-05, -7.9423e-06]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0267, -0.0177, -0.0063, 0.0009, 0.0024, -0.0023, -0.0141, -0.0030, + 0.0317, 0.0193], device='cuda:0'), grad: tensor([-1.6499e-03, 8.7619e-05, 3.5143e-04, 2.0771e-03, 4.3511e-04, + -2.7428e-03, 5.1260e-04, 1.5521e-04, 7.5006e-04, 2.3097e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 249.27, cls_loss 0.0229 cls_loss_mapping 0.0445 cls_loss_causal 0.7922 re_mapping 0.0240 re_causal 0.0684 /// teacc 98.44 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0150, -0.0426, -0.0161, ..., 0.0116, -0.0715, 0.0048], + [-0.0124, -0.0554, -0.0541, ..., -0.0433, -0.0003, -0.0296], + [ 0.0288, 0.0426, -0.0423, ..., -0.0566, -0.0592, -0.0245], + ..., + [-0.0017, 0.0108, -0.0019, ..., -0.0298, -0.0635, 0.0355], + [-0.0168, 0.0441, 0.0452, ..., -0.0270, 0.0197, -0.0449], + [-0.0371, -0.0179, 0.0154, ..., -0.0085, -0.0253, 0.0247]], + device='cuda:0'), grad: tensor([[ 1.4019e-03, 1.5793e-03, 8.0729e-04, ..., 1.2316e-05, + 6.7234e-04, 1.0834e-03], + [ 7.3624e-04, 5.3453e-04, 1.8799e-04, ..., 9.7454e-06, + 1.7715e-04, -3.5495e-05], + [-1.0681e-03, -6.6805e-04, 3.8552e-04, ..., 1.4432e-05, + 3.1400e-04, 3.3045e-04], + ..., + [ 3.8433e-04, 4.2272e-04, 2.9731e-04, ..., 1.2010e-05, + 2.8872e-04, 2.0826e-04], + [ 2.9297e-03, 2.1114e-03, 3.6774e-03, ..., 2.8539e-04, + 2.6093e-03, 2.2125e-03], + [-6.7940e-03, -7.9346e-03, -1.9550e-03, ..., 7.2479e-05, + -1.2817e-03, -9.5520e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0265, -0.0177, -0.0063, 0.0012, 0.0027, -0.0022, -0.0144, -0.0029, + 0.0317, 0.0191], device='cuda:0'), grad: tensor([ 0.0025, 0.0012, -0.0020, 0.0033, 0.0146, -0.0112, 0.0032, 0.0008, + 0.0064, -0.0189], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 266.38, cls_loss 0.0319 cls_loss_mapping 0.0554 cls_loss_causal 0.8528 re_mapping 0.0215 re_causal 0.0646 /// teacc 98.72 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0147, -0.0433, -0.0163, ..., 0.0112, -0.0725, 0.0047], + [-0.0132, -0.0565, -0.0553, ..., -0.0439, -0.0012, -0.0297], + [ 0.0293, 0.0429, -0.0431, ..., -0.0569, -0.0604, -0.0250], + ..., + [-0.0015, 0.0109, -0.0024, ..., -0.0297, -0.0640, 0.0355], + [-0.0168, 0.0445, 0.0453, ..., -0.0274, 0.0200, -0.0457], + [-0.0371, -0.0177, 0.0151, ..., -0.0088, -0.0257, 0.0251]], + device='cuda:0'), grad: tensor([[-1.7309e-03, -1.4868e-03, 3.8117e-05, ..., 5.2340e-06, + 5.0843e-05, -1.2608e-03], + [ 2.5272e-04, 1.0693e-04, -3.0351e-04, ..., -1.0028e-05, + -3.1924e-04, -5.7411e-04], + [ 2.2106e-03, 1.3380e-03, 7.4089e-05, ..., 1.9129e-06, + 6.6161e-05, 1.3008e-03], + ..., + [-2.4853e-03, -1.7023e-03, 1.7571e-04, ..., 1.7239e-06, + 1.3912e-04, -2.2411e-03], + [ 4.1652e-04, 2.5964e-04, 2.9016e-04, ..., 7.9051e-06, + 2.1601e-04, 8.4209e-04], + [ 7.5912e-04, 5.1165e-04, 1.7381e-04, ..., 7.4133e-07, + 6.4611e-05, 4.2558e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0262, -0.0181, -0.0066, 0.0013, 0.0025, -0.0019, -0.0146, -0.0027, + 0.0317, 0.0195], device='cuda:0'), grad: tensor([-0.0037, -0.0011, 0.0052, 0.0029, -0.0037, -0.0008, 0.0009, -0.0035, + 0.0020, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 250.53, cls_loss 0.0205 cls_loss_mapping 0.0383 cls_loss_causal 0.7987 re_mapping 0.0232 re_causal 0.0671 /// teacc 98.57 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0142, -0.0437, -0.0172, ..., 0.0101, -0.0734, 0.0042], + [-0.0140, -0.0576, -0.0563, ..., -0.0442, -0.0019, -0.0296], + [ 0.0299, 0.0428, -0.0443, ..., -0.0577, -0.0614, -0.0256], + ..., + [-0.0016, 0.0106, -0.0029, ..., -0.0297, -0.0651, 0.0356], + [-0.0160, 0.0459, 0.0462, ..., -0.0272, 0.0208, -0.0456], + [-0.0373, -0.0176, 0.0148, ..., -0.0091, -0.0259, 0.0253]], + device='cuda:0'), grad: tensor([[-3.7994e-03, -3.8643e-03, 1.5986e-04, ..., 9.9361e-05, + -6.7949e-04, 6.1095e-05], + [ 1.6994e-03, 2.1534e-03, 2.0921e-04, ..., 1.9982e-05, + 6.5136e-04, 7.4685e-05], + [ 6.5041e-04, 7.1239e-04, 1.2374e-04, ..., 5.1975e-05, + 2.5201e-04, 6.1154e-05], + ..., + [-6.6710e-04, -5.5075e-04, 9.0301e-05, ..., 3.2671e-06, + 1.2982e-04, -6.3133e-04], + [-8.6665e-05, -3.4595e-04, -7.9060e-04, ..., 6.2585e-05, + -8.4734e-04, 1.5366e-04], + [ 6.9046e-04, 5.4932e-04, 5.0306e-04, ..., 1.6898e-05, + 6.4993e-04, 2.5511e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0261, -0.0186, -0.0062, 0.0012, 0.0027, -0.0018, -0.0148, -0.0030, + 0.0323, 0.0194], device='cuda:0'), grad: tensor([-1.3702e-02, -4.4405e-05, 2.4719e-03, -5.6601e-04, 6.9923e-03, + 1.5545e-03, 1.6031e-03, -3.6335e-04, 6.5804e-05, 1.9855e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 251.03, cls_loss 0.0239 cls_loss_mapping 0.0422 cls_loss_causal 0.8037 re_mapping 0.0220 re_causal 0.0649 /// teacc 98.72 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0129, -0.0440, -0.0177, ..., 0.0088, -0.0743, 0.0045], + [-0.0147, -0.0583, -0.0575, ..., -0.0459, -0.0028, -0.0292], + [ 0.0302, 0.0431, -0.0445, ..., -0.0582, -0.0615, -0.0261], + ..., + [-0.0015, 0.0108, -0.0034, ..., -0.0290, -0.0658, 0.0358], + [-0.0158, 0.0465, 0.0465, ..., -0.0277, 0.0212, -0.0461], + [-0.0376, -0.0175, 0.0147, ..., -0.0097, -0.0260, 0.0253]], + device='cuda:0'), grad: tensor([[ 7.9334e-05, 1.2732e-04, 1.9228e-04, ..., 8.9258e-06, + 2.3615e-04, 2.9588e-04], + [-5.2738e-04, -6.5184e-04, -8.2636e-04, ..., 3.5614e-06, + -1.3819e-03, -2.2545e-03], + [ 1.5426e-04, 2.7609e-04, 4.4918e-04, ..., 6.3181e-06, + 6.6805e-04, 9.7179e-04], + ..., + [ 5.4568e-05, 4.9531e-05, 4.8786e-05, ..., 3.7197e-06, + 6.6280e-05, 1.0383e-04], + [ 1.6880e-04, 1.9109e-04, 3.6120e-04, ..., 1.4991e-05, + 4.9162e-04, 7.6056e-04], + [-1.0097e-04, -2.1362e-04, 5.7459e-05, ..., 2.2333e-06, + 5.5015e-05, -3.4356e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0269, -0.0186, -0.0063, 0.0013, 0.0025, -0.0020, -0.0143, -0.0029, + 0.0319, 0.0191], device='cuda:0'), grad: tensor([ 4.7994e-04, -4.2839e-03, 1.9703e-03, 6.7174e-05, 5.9509e-04, + -5.3310e-04, 6.2895e-04, 3.3593e-04, 1.5316e-03, -7.9203e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 251.47, cls_loss 0.0263 cls_loss_mapping 0.0437 cls_loss_causal 0.8222 re_mapping 0.0223 re_causal 0.0630 /// teacc 98.68 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0131, -0.0445, -0.0185, ..., 0.0083, -0.0753, 0.0041], + [-0.0158, -0.0585, -0.0577, ..., -0.0467, -0.0025, -0.0296], + [ 0.0309, 0.0432, -0.0455, ..., -0.0586, -0.0628, -0.0269], + ..., + [-0.0011, 0.0115, -0.0038, ..., -0.0288, -0.0668, 0.0362], + [-0.0155, 0.0471, 0.0468, ..., -0.0280, 0.0215, -0.0467], + [-0.0375, -0.0178, 0.0144, ..., -0.0101, -0.0265, 0.0262]], + device='cuda:0'), grad: tensor([[ 5.7966e-05, 1.4746e-04, 1.8775e-04, ..., 1.1194e-04, + 1.4865e-04, 5.0634e-05], + [ 3.4928e-04, 4.8161e-04, 5.2303e-05, ..., 1.7190e-04, + 8.5473e-05, 4.6700e-05], + [-6.1178e-04, 1.1939e-04, 1.7214e-04, ..., 1.0347e-03, + 5.6839e-04, -4.9680e-05], + ..., + [ 9.8884e-05, 1.5020e-04, 6.4671e-05, ..., 4.8369e-05, + 6.4790e-05, 3.9518e-05], + [ 7.7665e-05, -2.1935e-05, 1.4663e-04, ..., 1.0633e-04, + 3.0696e-05, 8.3089e-05], + [-7.5221e-05, -3.4988e-05, 6.0940e-04, ..., 3.2216e-05, + 5.8889e-04, 9.5367e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0267, -0.0186, -0.0063, 0.0010, 0.0025, -0.0019, -0.0142, -0.0026, + 0.0316, 0.0193], device='cuda:0'), grad: tensor([ 0.0005, 0.0013, 0.0008, -0.0310, -0.0047, 0.0296, 0.0013, 0.0005, + 0.0007, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 251.11, cls_loss 0.0226 cls_loss_mapping 0.0419 cls_loss_causal 0.7983 re_mapping 0.0215 re_causal 0.0620 /// teacc 98.54 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0128, -0.0455, -0.0195, ..., 0.0078, -0.0766, 0.0038], + [-0.0167, -0.0597, -0.0584, ..., -0.0472, -0.0032, -0.0297], + [ 0.0312, 0.0435, -0.0464, ..., -0.0590, -0.0634, -0.0269], + ..., + [-0.0006, 0.0120, -0.0041, ..., -0.0292, -0.0676, 0.0366], + [-0.0153, 0.0477, 0.0470, ..., -0.0285, 0.0216, -0.0477], + [-0.0375, -0.0175, 0.0141, ..., -0.0103, -0.0270, 0.0266]], + device='cuda:0'), grad: tensor([[-1.7989e-04, 1.4842e-04, 8.3387e-05, ..., 4.9360e-08, + 5.9754e-05, -7.1824e-05], + [ 9.0837e-05, -1.9226e-03, -8.7643e-04, ..., 1.0304e-05, + 7.2300e-05, -3.7899e-03], + [ 1.1578e-03, 3.1300e-03, 2.2526e-03, ..., 3.7253e-05, + 1.7557e-03, 4.4274e-04], + ..., + [ 5.5820e-05, 1.6403e-04, 8.2791e-05, ..., 1.8673e-06, + 3.7730e-05, 1.8179e-04], + [-1.3781e-03, -2.2449e-03, -2.0199e-03, ..., 9.4891e-05, + -2.0885e-03, 2.3060e-03], + [ 2.1935e-03, 3.1013e-03, 8.6427e-05, ..., 1.4596e-05, + 5.7250e-05, 5.2786e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0267, -0.0190, -0.0059, 0.0012, 0.0025, -0.0015, -0.0148, -0.0027, + 0.0314, 0.0195], device='cuda:0'), grad: tensor([-8.5235e-05, -9.4452e-03, 3.9291e-03, 1.4210e-04, -7.2784e-03, + 9.6941e-04, 8.6355e-04, 6.4516e-04, 2.9011e-03, 7.3662e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 251.02, cls_loss 0.0200 cls_loss_mapping 0.0398 cls_loss_causal 0.7622 re_mapping 0.0206 re_causal 0.0601 /// teacc 98.57 lr 0.00010000 +Epoch 35, weight, value: tensor([[-1.2679e-02, -4.6142e-02, -2.0089e-02, ..., 7.3313e-03, + -7.7372e-02, 3.5660e-03], + [-1.7384e-02, -6.0510e-02, -5.9200e-02, ..., -4.7669e-02, + -4.2675e-03, -2.9394e-02], + [ 3.1480e-02, 4.3955e-02, -4.6198e-02, ..., -5.9361e-02, + -6.2164e-02, -2.7430e-02], + ..., + [-1.9993e-05, 1.2685e-02, -4.4425e-03, ..., -2.9124e-02, + -6.8417e-02, 3.7467e-02], + [-1.5319e-02, 4.8000e-02, 4.7081e-02, ..., -2.9701e-02, + 2.1834e-02, -4.8492e-02], + [-3.7499e-02, -1.6989e-02, 1.4206e-02, ..., -9.1683e-03, + -2.7018e-02, 2.6614e-02]], device='cuda:0'), grad: tensor([[ 6.2823e-05, 2.7204e-04, 1.8573e-04, ..., 1.8120e-04, + 2.0158e-04, 8.3745e-05], + [ 4.1455e-05, 7.1764e-05, 3.7253e-05, ..., 2.6122e-05, + 3.6806e-05, -1.2141e-04], + [ 1.0478e-04, 3.6240e-04, 2.5749e-04, ..., 2.4378e-04, + 2.8491e-04, 1.3089e-04], + ..., + [-5.9032e-04, -7.3433e-04, 2.6986e-05, ..., 1.8150e-05, + 2.9907e-05, -4.7827e-04], + [-1.3769e-04, -1.0377e-04, -2.8396e-04, ..., 1.1867e-04, + -1.3876e-04, 8.0764e-05], + [ 3.4022e-04, 5.6887e-04, 1.9240e-04, ..., 1.5438e-04, + 1.9348e-04, 2.6679e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0264, -0.0192, -0.0055, 0.0012, 0.0021, -0.0016, -0.0149, -0.0018, + 0.0309, 0.0196], device='cuda:0'), grad: tensor([ 3.8791e-04, -1.3006e-04, 6.0892e-04, 1.3661e-04, 3.9506e-04, + 1.4029e-03, -2.8305e-03, -7.1478e-04, 1.5618e-06, 7.4005e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 251.40, cls_loss 0.0185 cls_loss_mapping 0.0336 cls_loss_causal 0.7600 re_mapping 0.0202 re_causal 0.0584 /// teacc 98.59 lr 0.00010000 +Epoch 36, weight, value: tensor([[-1.2800e-02, -4.7033e-02, -2.0935e-02, ..., 6.7311e-03, + -7.8751e-02, 3.1389e-03], + [-1.8015e-02, -6.1389e-02, -6.0033e-02, ..., -4.8466e-02, + -4.9184e-03, -2.9077e-02], + [ 3.1717e-02, 4.4053e-02, -4.6511e-02, ..., -5.9664e-02, + -6.2287e-02, -2.8211e-02], + ..., + [ 6.1960e-05, 1.3001e-02, -5.2057e-03, ..., -2.9075e-02, + -6.9221e-02, 3.7479e-02], + [-1.4722e-02, 4.9189e-02, 4.7954e-02, ..., -2.9769e-02, + 2.2954e-02, -4.8978e-02], + [-3.7085e-02, -1.6689e-02, 1.3799e-02, ..., -9.4828e-03, + -2.7587e-02, 2.6985e-02]], device='cuda:0'), grad: tensor([[ 2.6608e-04, 3.3617e-04, 3.5977e-04, ..., 2.9063e-04, + 3.2592e-04, 1.6820e-04], + [ 7.5579e-05, 8.8096e-05, 8.0764e-05, ..., 5.7369e-05, + 7.5877e-05, -2.7871e-04], + [ 4.7994e-04, 6.4230e-04, 5.6505e-04, ..., 5.5885e-04, + 5.9366e-04, 3.1090e-04], + ..., + [-7.2837e-05, -1.3340e-04, 3.3212e-04, ..., 4.5896e-05, + 6.5029e-05, 2.4393e-05], + [ 4.1342e-04, 4.2081e-04, 5.2071e-04, ..., 5.2214e-04, + 4.6301e-04, 3.5715e-04], + [ 6.6328e-04, 7.9775e-04, 8.7500e-04, ..., 6.0701e-04, + 7.3051e-04, 4.5109e-04]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0261, -0.0187, -0.0057, 0.0011, 0.0021, -0.0018, -0.0152, -0.0019, + 0.0314, 0.0197], device='cuda:0'), grad: tensor([ 0.0008, -0.0005, 0.0015, -0.0005, 0.0015, 0.0010, -0.0077, 0.0004, + 0.0014, 0.0020], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 253.75, cls_loss 0.0181 cls_loss_mapping 0.0343 cls_loss_causal 0.7322 re_mapping 0.0199 re_causal 0.0576 /// teacc 98.67 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0122, -0.0469, -0.0215, ..., 0.0064, -0.0794, 0.0031], + [-0.0188, -0.0624, -0.0602, ..., -0.0497, -0.0052, -0.0285], + [ 0.0325, 0.0445, -0.0469, ..., -0.0605, -0.0628, -0.0290], + ..., + [ 0.0007, 0.0136, -0.0054, ..., -0.0284, -0.0693, 0.0380], + [-0.0152, 0.0493, 0.0479, ..., -0.0302, 0.0229, -0.0499], + [-0.0374, -0.0168, 0.0131, ..., -0.0101, -0.0281, 0.0271]], + device='cuda:0'), grad: tensor([[ 9.4622e-06, 1.4439e-05, 3.3885e-05, ..., 2.2519e-06, + 3.1680e-05, 4.7423e-06], + [ 2.0659e-04, 1.1808e-04, 3.4392e-05, ..., 4.1537e-07, + 4.9800e-05, 3.1944e-06], + [-2.2233e-05, -1.5414e-04, 1.3196e-04, ..., 3.9954e-07, + 2.0897e-04, -2.1085e-05], + ..., + [ 2.4700e-04, 1.0395e-04, 1.1617e-04, ..., 3.7765e-07, + 1.5485e-04, 1.0115e-04], + [ 2.7442e-04, 5.7459e-05, 3.2711e-04, ..., 7.0743e-06, + 3.4308e-04, 4.3303e-05], + [ 1.4877e-04, -1.0198e-04, 2.4843e-04, ..., 3.4906e-06, + 2.4509e-04, -1.0592e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0263, -0.0183, -0.0056, 0.0009, 0.0020, -0.0014, -0.0153, -0.0017, + 0.0304, 0.0199], device='cuda:0'), grad: tensor([ 4.1544e-05, 2.1851e-04, 2.0683e-04, -2.0447e-03, 1.3149e-04, + -8.3303e-04, 6.1321e-04, 5.7411e-04, 7.4244e-04, 3.4833e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 256.61, cls_loss 0.0205 cls_loss_mapping 0.0349 cls_loss_causal 0.7607 re_mapping 0.0193 re_causal 0.0562 /// teacc 98.63 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0117, -0.0471, -0.0222, ..., 0.0063, -0.0800, 0.0031], + [-0.0197, -0.0636, -0.0607, ..., -0.0495, -0.0059, -0.0284], + [ 0.0327, 0.0442, -0.0482, ..., -0.0610, -0.0630, -0.0298], + ..., + [ 0.0010, 0.0140, -0.0058, ..., -0.0287, -0.0703, 0.0383], + [-0.0145, 0.0501, 0.0484, ..., -0.0310, 0.0234, -0.0505], + [-0.0380, -0.0174, 0.0126, ..., -0.0106, -0.0284, 0.0270]], + device='cuda:0'), grad: tensor([[-2.7776e-05, 5.8860e-05, 5.5015e-05, ..., 3.9637e-05, + 4.3839e-05, 8.3596e-06], + [ 2.5943e-05, 5.2214e-05, 5.8770e-05, ..., 2.5898e-05, + 5.2840e-05, 4.7803e-05], + [ 1.1581e-04, 1.3947e-04, 2.1696e-04, ..., 2.2516e-05, + 1.6391e-04, 2.6509e-05], + ..., + [-2.2388e-04, -9.3579e-05, 5.4061e-05, ..., 2.5108e-06, + 7.6890e-05, -4.8661e-04], + [-1.8537e-04, -1.7571e-04, -2.0087e-04, ..., 8.6904e-05, + -9.6738e-05, 7.8976e-05], + [ 2.3377e-04, -4.4256e-05, -5.6982e-05, ..., 7.5027e-06, + -1.0657e-04, 3.4881e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0264, -0.0186, -0.0056, 0.0014, 0.0021, -0.0014, -0.0160, -0.0016, + 0.0307, 0.0196], device='cuda:0'), grad: tensor([-4.0317e-04, 3.7599e-04, 5.1880e-04, -2.1923e-04, -2.3861e-03, + 4.3750e-04, -2.6792e-05, -7.9155e-04, -2.6926e-05, 2.5215e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 273.34, cls_loss 0.0145 cls_loss_mapping 0.0319 cls_loss_causal 0.7050 re_mapping 0.0182 re_causal 0.0542 /// teacc 98.73 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0113, -0.0477, -0.0228, ..., 0.0058, -0.0804, 0.0031], + [-0.0207, -0.0652, -0.0615, ..., -0.0502, -0.0067, -0.0284], + [ 0.0330, 0.0443, -0.0494, ..., -0.0615, -0.0635, -0.0297], + ..., + [ 0.0011, 0.0138, -0.0061, ..., -0.0288, -0.0712, 0.0386], + [-0.0140, 0.0512, 0.0489, ..., -0.0313, 0.0241, -0.0509], + [-0.0381, -0.0172, 0.0123, ..., -0.0110, -0.0286, 0.0272]], + device='cuda:0'), grad: tensor([[-8.1003e-05, 5.2750e-05, 6.4135e-05, ..., 2.6643e-05, + 4.6432e-05, -3.3118e-06], + [ 9.6440e-05, 7.8082e-05, 9.4473e-05, ..., 1.5110e-05, + 1.2422e-04, -2.8944e-04], + [-2.6932e-03, -3.7327e-03, -2.2850e-03, ..., 1.1243e-05, + 5.6297e-05, 3.1519e-04], + ..., + [ 5.9426e-05, 4.1336e-05, 3.6895e-05, ..., 2.5108e-06, + 3.2753e-05, 1.2118e-04], + [ 3.4885e-03, 3.7060e-03, 2.7046e-03, ..., 7.4983e-05, + 1.1702e-03, 1.3316e-04], + [ 4.2051e-05, -1.9118e-05, 2.0429e-05, ..., 6.6608e-06, + 1.2144e-05, -2.5868e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0257, -0.0192, -0.0055, 0.0012, 0.0022, -0.0012, -0.0157, -0.0016, + 0.0312, 0.0196], device='cuda:0'), grad: tensor([-1.5604e-04, -3.9148e-04, -2.4567e-03, -1.6499e-03, 1.9455e-04, + 3.0136e-04, -3.6097e-04, 3.0994e-04, 4.1313e-03, 7.3910e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 256.61, cls_loss 0.0142 cls_loss_mapping 0.0298 cls_loss_causal 0.6949 re_mapping 0.0183 re_causal 0.0540 /// teacc 98.67 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0107, -0.0478, -0.0233, ..., 0.0060, -0.0813, 0.0033], + [-0.0216, -0.0662, -0.0620, ..., -0.0504, -0.0069, -0.0282], + [ 0.0330, 0.0442, -0.0499, ..., -0.0620, -0.0642, -0.0310], + ..., + [ 0.0016, 0.0144, -0.0062, ..., -0.0289, -0.0715, 0.0391], + [-0.0135, 0.0518, 0.0489, ..., -0.0320, 0.0244, -0.0516], + [-0.0385, -0.0173, 0.0122, ..., -0.0106, -0.0286, 0.0273]], + device='cuda:0'), grad: tensor([[-1.3697e-04, 3.8207e-05, 8.3208e-05, ..., 4.3958e-05, + 7.2658e-05, 3.4243e-05], + [ 4.6402e-05, 2.9728e-05, 2.2829e-05, ..., 2.8789e-05, + -6.9916e-05, -5.2691e-04], + [-7.7665e-05, -6.2250e-06, 1.2094e-04, ..., 5.2303e-05, + 1.0484e-04, 5.8860e-05], + ..., + [ 2.8759e-06, 2.1234e-05, 1.9386e-05, ..., 5.0813e-06, + 3.0518e-05, 3.5226e-05], + [ 5.5879e-05, 3.1114e-04, 6.4039e-04, ..., 3.3450e-04, + 5.2977e-04, 2.0945e-04], + [-2.5854e-05, -8.9526e-05, 2.2963e-05, ..., 5.5805e-06, + -1.6227e-05, -1.1110e-04]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0260, -0.0192, -0.0060, 0.0013, 0.0025, -0.0012, -0.0159, -0.0013, + 0.0311, 0.0194], device='cuda:0'), grad: tensor([-2.2686e-04, -8.0061e-04, 4.2260e-05, 1.0628e-04, 2.9016e-04, + 9.3508e-04, -1.4267e-03, 2.0897e-04, 9.7322e-04, -9.9063e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 272.93, cls_loss 0.0165 cls_loss_mapping 0.0317 cls_loss_causal 0.7442 re_mapping 0.0176 re_causal 0.0529 /// teacc 98.76 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0107, -0.0485, -0.0241, ..., 0.0051, -0.0821, 0.0031], + [-0.0227, -0.0674, -0.0625, ..., -0.0517, -0.0072, -0.0270], + [ 0.0341, 0.0450, -0.0504, ..., -0.0622, -0.0646, -0.0316], + ..., + [ 0.0016, 0.0144, -0.0065, ..., -0.0289, -0.0718, 0.0391], + [-0.0138, 0.0518, 0.0491, ..., -0.0327, 0.0245, -0.0525], + [-0.0382, -0.0171, 0.0117, ..., -0.0111, -0.0289, 0.0278]], + device='cuda:0'), grad: tensor([[-2.6170e-06, 1.1522e-04, 1.7273e-04, ..., 3.4004e-05, + 1.3566e-04, 5.2631e-05], + [ 2.0599e-04, 2.9659e-04, 3.7283e-05, ..., 4.6864e-06, + 3.6240e-05, 1.9026e-04], + [-2.2388e-04, -4.4227e-04, 1.6242e-05, ..., -1.0580e-04, + 7.5579e-05, 1.0252e-04], + ..., + [-6.3229e-04, -9.3508e-04, 7.0333e-05, ..., 9.2685e-06, + -1.4758e-04, -9.5224e-04], + [ 4.4751e-04, 6.7091e-04, 4.3416e-04, ..., 9.0718e-05, + 4.6086e-04, 6.0320e-04], + [ 1.1599e-04, 8.4937e-05, 4.0197e-04, ..., 9.3430e-06, + 3.3188e-04, 1.3852e-04]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0259, -0.0188, -0.0057, 0.0011, 0.0026, -0.0009, -0.0154, -0.0021, + 0.0305, 0.0196], device='cuda:0'), grad: tensor([ 0.0002, 0.0003, -0.0005, 0.0009, 0.0002, -0.0018, 0.0004, -0.0017, + 0.0015, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 256.57, cls_loss 0.0183 cls_loss_mapping 0.0334 cls_loss_causal 0.7049 re_mapping 0.0178 re_causal 0.0517 /// teacc 98.66 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0100, -0.0491, -0.0239, ..., 0.0041, -0.0825, 0.0025], + [-0.0234, -0.0682, -0.0629, ..., -0.0523, -0.0075, -0.0267], + [ 0.0349, 0.0456, -0.0511, ..., -0.0620, -0.0652, -0.0307], + ..., + [ 0.0019, 0.0148, -0.0069, ..., -0.0290, -0.0723, 0.0394], + [-0.0136, 0.0525, 0.0494, ..., -0.0330, 0.0248, -0.0531], + [-0.0389, -0.0168, 0.0112, ..., -0.0095, -0.0293, 0.0281]], + device='cuda:0'), grad: tensor([[-9.2201e-07, 7.0858e-04, 1.1358e-03, ..., 4.7135e-04, + 9.0742e-04, 3.2711e-04], + [ 3.4213e-05, 5.4747e-05, 5.0545e-05, ..., 3.0294e-05, + 4.3780e-05, -4.4137e-05], + [-6.5279e-04, -2.3568e-04, 8.0347e-05, ..., 1.3769e-05, + 5.4002e-05, 4.3124e-05], + ..., + [ 6.3777e-05, 8.8289e-06, 1.9193e-05, ..., 8.0764e-06, + 1.1280e-05, -6.9588e-06], + [ 9.6679e-05, 4.3392e-04, 5.6171e-04, ..., 3.1805e-04, + 4.8566e-04, 2.3580e-04], + [ 6.8545e-05, 7.4208e-05, 1.4043e-04, ..., 4.2707e-05, + 7.1883e-05, 8.9556e-06]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0262, -0.0188, -0.0053, 0.0014, 0.0028, -0.0015, -0.0160, -0.0023, + 0.0306, 0.0198], device='cuda:0'), grad: tensor([ 1.2884e-03, -3.8099e-04, -7.9203e-04, 8.6164e-04, -1.0061e-04, + -9.1732e-05, -2.5196e-03, 2.1863e-04, 9.6893e-04, 5.4741e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 257.11, cls_loss 0.0179 cls_loss_mapping 0.0310 cls_loss_causal 0.7320 re_mapping 0.0169 re_causal 0.0499 /// teacc 98.73 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0101, -0.0499, -0.0248, ..., 0.0037, -0.0840, 0.0023], + [-0.0240, -0.0697, -0.0634, ..., -0.0529, -0.0079, -0.0266], + [ 0.0353, 0.0461, -0.0515, ..., -0.0621, -0.0655, -0.0309], + ..., + [ 0.0014, 0.0146, -0.0073, ..., -0.0289, -0.0728, 0.0397], + [-0.0129, 0.0536, 0.0499, ..., -0.0327, 0.0252, -0.0538], + [-0.0386, -0.0166, 0.0106, ..., -0.0101, -0.0298, 0.0287]], + device='cuda:0'), grad: tensor([[ 6.9320e-05, 1.3149e-04, 4.0245e-04, ..., 1.5891e-04, + 3.0327e-04, 3.6812e-04], + [ 9.9659e-05, 1.2457e-04, 1.3404e-05, ..., 4.8690e-06, + 1.4193e-05, -1.2293e-05], + [ 4.9442e-05, 9.4473e-05, 2.6554e-05, ..., -3.8892e-06, + 2.5630e-05, 1.9264e-04], + ..., + [-4.7517e-04, -5.9319e-04, 1.4558e-05, ..., 4.2729e-06, + 1.7673e-05, -4.1389e-04], + [-9.1502e-07, 2.7433e-05, -1.7118e-04, ..., 5.8621e-05, + -9.1791e-05, 5.2959e-05], + [-1.5509e-04, -3.1042e-04, 1.4997e-04, ..., 9.4716e-07, + 1.0884e-04, -9.1255e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0258, -0.0191, -0.0053, 0.0012, 0.0028, -0.0015, -0.0155, -0.0023, + 0.0308, 0.0200], device='cuda:0'), grad: tensor([ 8.7309e-04, 4.9531e-05, 1.9717e-04, 1.3733e-04, 1.2531e-03, + 7.5400e-05, -9.4032e-04, -8.5163e-04, 1.2052e-04, -9.1362e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 273.33, cls_loss 0.0146 cls_loss_mapping 0.0286 cls_loss_causal 0.6979 re_mapping 0.0176 re_causal 0.0504 /// teacc 98.80 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0098, -0.0507, -0.0253, ..., 0.0032, -0.0846, 0.0024], + [-0.0248, -0.0711, -0.0639, ..., -0.0538, -0.0084, -0.0267], + [ 0.0357, 0.0464, -0.0520, ..., -0.0619, -0.0658, -0.0314], + ..., + [ 0.0020, 0.0156, -0.0076, ..., -0.0290, -0.0733, 0.0402], + [-0.0125, 0.0542, 0.0505, ..., -0.0335, 0.0257, -0.0543], + [-0.0388, -0.0165, 0.0102, ..., -0.0105, -0.0301, 0.0287]], + device='cuda:0'), grad: tensor([[ 6.1207e-06, 1.1277e-04, 2.5019e-05, ..., 5.8115e-05, + 2.9460e-05, 2.7061e-05], + [ 6.5006e-06, 1.9848e-05, 1.5721e-05, ..., 3.5595e-06, + 1.5423e-05, -4.9978e-05], + [-3.9518e-05, -7.9861e-07, 1.9386e-05, ..., 3.9935e-06, + 2.9102e-05, 2.5615e-05], + ..., + [-2.5854e-05, -4.6134e-05, 1.1757e-05, ..., 3.5856e-07, + 1.3135e-05, -2.2396e-05], + [ 1.1124e-05, -7.3351e-06, 3.2812e-05, ..., 3.0041e-05, + 3.6925e-05, 1.6406e-05], + [ 2.1875e-05, 3.7879e-05, 3.0577e-05, ..., 1.1206e-05, + 2.6032e-05, 1.9744e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0259, -0.0193, -0.0051, 0.0011, 0.0027, -0.0018, -0.0156, -0.0016, + 0.0308, 0.0197], device='cuda:0'), grad: tensor([ 1.2028e-04, -9.7930e-05, 5.1945e-05, -3.4899e-05, 4.3511e-06, + -3.0786e-05, -1.9515e-04, -4.2804e-06, 7.6771e-05, 1.0955e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 256.20, cls_loss 0.0152 cls_loss_mapping 0.0291 cls_loss_causal 0.7285 re_mapping 0.0167 re_causal 0.0498 /// teacc 98.64 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0094, -0.0513, -0.0258, ..., 0.0025, -0.0851, 0.0022], + [-0.0254, -0.0716, -0.0645, ..., -0.0543, -0.0089, -0.0256], + [ 0.0359, 0.0464, -0.0526, ..., -0.0626, -0.0662, -0.0323], + ..., + [ 0.0027, 0.0164, -0.0078, ..., -0.0291, -0.0738, 0.0404], + [-0.0123, 0.0547, 0.0506, ..., -0.0339, 0.0259, -0.0548], + [-0.0396, -0.0170, 0.0096, ..., -0.0109, -0.0305, 0.0289]], + device='cuda:0'), grad: tensor([[ 7.0632e-05, 1.0157e-04, 7.3075e-05, ..., 1.4700e-05, + 5.3883e-05, 2.2873e-05], + [ 2.8372e-04, 3.0708e-04, 1.5509e-04, ..., 3.9972e-06, + 1.2553e-04, 9.0003e-05], + [-6.8951e-04, -2.7418e-04, 1.2147e-04, ..., 1.8533e-06, + 9.9719e-05, 1.7092e-05], + ..., + [ 3.5763e-04, 5.2834e-04, 8.1420e-05, ..., 1.6484e-07, + 6.0260e-05, 7.4530e-04], + [-3.1519e-04, -1.2045e-03, -8.5545e-04, ..., -6.0070e-07, + -6.8903e-04, -1.0735e-04], + [-2.0969e-04, -2.4652e-04, 7.6246e-04, ..., 1.7248e-06, + 5.6934e-04, -7.3624e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0256, -0.0187, -0.0055, 0.0011, 0.0024, -0.0015, -0.0156, -0.0014, + 0.0306, 0.0197], device='cuda:0'), grad: tensor([ 2.0015e-04, 6.0654e-04, -6.4421e-04, 1.8921e-02, -2.9549e-05, + -2.0172e-02, 2.5439e-04, 1.5249e-03, -1.1635e-03, 4.9400e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 256.47, cls_loss 0.0120 cls_loss_mapping 0.0246 cls_loss_causal 0.6910 re_mapping 0.0160 re_causal 0.0475 /// teacc 98.79 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0095, -0.0518, -0.0262, ..., 0.0017, -0.0857, 0.0017], + [-0.0265, -0.0722, -0.0649, ..., -0.0550, -0.0090, -0.0249], + [ 0.0361, 0.0463, -0.0533, ..., -0.0638, -0.0668, -0.0330], + ..., + [ 0.0024, 0.0165, -0.0082, ..., -0.0293, -0.0743, 0.0402], + [-0.0123, 0.0550, 0.0509, ..., -0.0339, 0.0261, -0.0555], + [-0.0397, -0.0169, 0.0092, ..., -0.0113, -0.0309, 0.0291]], + device='cuda:0'), grad: tensor([[-1.9986e-06, 3.4451e-05, 1.0818e-05, ..., 1.8980e-06, + 1.0133e-05, 2.2873e-05], + [ 3.0056e-05, -1.7440e-04, 4.5061e-05, ..., 9.4762e-07, + 4.8310e-05, -4.3821e-04], + [-3.9995e-05, -1.6894e-06, 2.2337e-05, ..., 4.8568e-07, + 2.5377e-05, 3.0547e-05], + ..., + [-1.5521e-04, -2.6584e-04, 2.1443e-05, ..., 9.5926e-08, + 1.7643e-05, 1.3612e-05], + [-5.9366e-05, -9.1970e-05, -8.6725e-05, ..., 6.9924e-06, + -1.0097e-04, 2.2545e-05], + [ 1.5426e-04, 9.2387e-05, -2.8968e-04, ..., 9.3551e-07, + -2.2638e-04, -1.1003e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0250, -0.0185, -0.0058, 0.0017, 0.0026, -0.0014, -0.0148, -0.0018, + 0.0302, 0.0196], device='cuda:0'), grad: tensor([-2.5146e-06, -9.7847e-04, 2.3454e-05, 3.3379e-04, 2.7132e-04, + 3.4118e-04, 9.8869e-06, 2.4140e-04, -3.2574e-05, -2.0766e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 256.69, cls_loss 0.0123 cls_loss_mapping 0.0253 cls_loss_causal 0.6936 re_mapping 0.0158 re_causal 0.0458 /// teacc 98.80 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0091, -0.0522, -0.0268, ..., 0.0012, -0.0867, 0.0019], + [-0.0273, -0.0732, -0.0661, ..., -0.0553, -0.0098, -0.0241], + [ 0.0367, 0.0470, -0.0526, ..., -0.0639, -0.0666, -0.0338], + ..., + [ 0.0025, 0.0163, -0.0086, ..., -0.0293, -0.0754, 0.0400], + [-0.0124, 0.0551, 0.0513, ..., -0.0343, 0.0267, -0.0559], + [-0.0398, -0.0164, 0.0090, ..., -0.0112, -0.0312, 0.0294]], + device='cuda:0'), grad: tensor([[ 7.3835e-06, -1.1559e-03, 5.9485e-05, ..., 2.4527e-05, + 6.0916e-05, 7.0743e-06], + [ 3.5852e-05, 7.3075e-05, 1.3304e-04, ..., 6.3926e-06, + 1.6928e-04, -2.6926e-05], + [ 1.7658e-05, 2.9111e-04, 7.8201e-05, ..., 1.3158e-05, + 9.2864e-05, 2.6435e-05], + ..., + [-1.3590e-04, -6.6400e-05, 5.2810e-05, ..., 1.2759e-06, + 5.8562e-05, -1.3173e-04], + [-1.6391e-04, -2.2674e-04, -5.1689e-04, ..., 2.3618e-05, + -5.3883e-04, 1.7956e-05], + [ 8.0824e-05, 2.5558e-04, 3.0017e-04, ..., 3.7868e-06, + 3.2163e-04, -7.0296e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0255, -0.0185, -0.0056, 0.0020, 0.0024, -0.0017, -0.0151, -0.0022, + 0.0302, 0.0197], device='cuda:0'), grad: tensor([-7.8659e-03, 3.8719e-04, 1.7910e-03, -6.4039e-04, 2.7561e-03, + 1.2074e-03, 9.9754e-04, 4.8995e-05, 2.9898e-04, 1.0157e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 46---------------------------------------------------- +epoch 46, time 272.31, cls_loss 0.0122 cls_loss_mapping 0.0228 cls_loss_causal 0.7073 re_mapping 0.0153 re_causal 0.0470 /// teacc 98.83 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0091, -0.0527, -0.0275, ..., 0.0009, -0.0875, 0.0015], + [-0.0283, -0.0735, -0.0665, ..., -0.0558, -0.0099, -0.0238], + [ 0.0369, 0.0466, -0.0539, ..., -0.0641, -0.0675, -0.0344], + ..., + [ 0.0027, 0.0167, -0.0089, ..., -0.0294, -0.0758, 0.0401], + [-0.0119, 0.0557, 0.0517, ..., -0.0348, 0.0270, -0.0566], + [-0.0397, -0.0165, 0.0086, ..., -0.0107, -0.0316, 0.0298]], + device='cuda:0'), grad: tensor([[ 6.6310e-07, 1.0431e-04, 4.8280e-05, ..., 2.0992e-06, + 2.7753e-06, 5.3018e-05], + [ 1.6361e-05, 3.4332e-05, 2.0549e-05, ..., 4.6706e-07, + 1.1303e-05, -5.0694e-05], + [-6.3237e-07, 1.0878e-04, 8.8871e-05, ..., 1.4659e-06, + 1.6659e-05, 9.3520e-05], + ..., + [-1.0872e-04, -1.1641e-04, 2.5198e-05, ..., 3.0152e-07, + 2.9206e-06, -1.2839e-04], + [ 6.0489e-07, 1.4651e-04, 1.7488e-04, ..., 4.4107e-06, + -5.6326e-05, 4.3333e-05], + [ 5.3138e-05, -3.6907e-04, 2.4006e-05, ..., 7.4785e-07, + 5.1260e-06, -3.3832e-04]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0255, -0.0182, -0.0061, 0.0016, 0.0032, -0.0012, -0.0152, -0.0023, + 0.0302, 0.0196], device='cuda:0'), grad: tensor([ 6.7592e-05, -1.3933e-05, 2.4581e-04, 3.3927e-04, 9.8288e-05, + -5.4407e-04, 1.1808e-04, -2.2745e-04, 3.1018e-04, -3.9482e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 256.65, cls_loss 0.0097 cls_loss_mapping 0.0216 cls_loss_causal 0.7042 re_mapping 0.0155 re_causal 0.0493 /// teacc 98.82 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0087, -0.0531, -0.0281, ..., 0.0003, -0.0882, 0.0012], + [-0.0290, -0.0746, -0.0670, ..., -0.0565, -0.0102, -0.0237], + [ 0.0374, 0.0468, -0.0544, ..., -0.0643, -0.0679, -0.0350], + ..., + [ 0.0029, 0.0174, -0.0093, ..., -0.0293, -0.0759, 0.0406], + [-0.0119, 0.0558, 0.0518, ..., -0.0346, 0.0271, -0.0573], + [-0.0400, -0.0160, 0.0083, ..., -0.0108, -0.0319, 0.0301]], + device='cuda:0'), grad: tensor([[-1.3798e-05, 3.2634e-06, 3.7253e-06, ..., 0.0000e+00, + 4.0568e-06, 2.5425e-06], + [ 1.8835e-05, 3.1680e-05, 2.4676e-05, ..., 0.0000e+00, + 3.1382e-05, 2.1040e-05], + [-9.5740e-07, 1.1036e-06, 7.0967e-06, ..., 0.0000e+00, + 9.3952e-06, 3.6899e-06], + ..., + [-2.6584e-05, -1.0416e-05, 7.3425e-06, ..., 0.0000e+00, + 1.0148e-05, -2.9504e-05], + [ 4.2133e-06, -1.7866e-05, 2.0564e-06, ..., 0.0000e+00, + -6.3796e-07, 9.8720e-06], + [-4.3735e-06, -4.8399e-05, 2.9266e-05, ..., 0.0000e+00, + 3.1829e-05, -3.8564e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0255, -0.0185, -0.0062, 0.0016, 0.0029, -0.0009, -0.0150, -0.0019, + 0.0299, 0.0196], device='cuda:0'), grad: tensor([-2.6718e-05, 7.0333e-05, 3.0085e-05, 4.2224e-04, -3.3408e-05, + -5.4407e-04, 4.3958e-05, 5.9754e-06, 4.1693e-05, -1.1235e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 256.99, cls_loss 0.0119 cls_loss_mapping 0.0245 cls_loss_causal 0.7021 re_mapping 0.0152 re_causal 0.0464 /// teacc 98.81 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0086, -0.0534, -0.0285, ..., -0.0005, -0.0887, 0.0009], + [-0.0303, -0.0756, -0.0676, ..., -0.0569, -0.0104, -0.0240], + [ 0.0379, 0.0469, -0.0551, ..., -0.0648, -0.0686, -0.0351], + ..., + [ 0.0029, 0.0172, -0.0096, ..., -0.0293, -0.0771, 0.0410], + [-0.0115, 0.0564, 0.0522, ..., -0.0346, 0.0275, -0.0580], + [-0.0400, -0.0158, 0.0081, ..., -0.0093, -0.0320, 0.0304]], + device='cuda:0'), grad: tensor([[ 1.8954e-05, 4.3571e-05, 5.6028e-05, ..., 3.0309e-05, + 2.8744e-05, 2.5511e-05], + [ 2.0850e-04, 1.3244e-04, 1.8671e-05, ..., 2.2482e-06, + 1.1221e-05, 4.6182e-04], + [-2.7847e-04, -1.5426e-04, -7.3731e-05, ..., 5.2340e-06, + -1.8999e-05, 1.0952e-05], + ..., + [-1.0328e-03, -6.6185e-04, 6.1952e-06, ..., 8.9547e-07, + -4.3437e-06, -2.7447e-03], + [ 1.9670e-04, 1.1468e-04, 8.1241e-05, ..., 9.3803e-06, + 2.7940e-05, 2.7514e-04], + [ 7.0238e-04, 3.8576e-04, 2.4900e-05, ..., 2.4643e-06, + 1.0148e-05, 1.6832e-03]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0249, -0.0190, -0.0060, 0.0017, 0.0036, -0.0010, -0.0152, -0.0018, + 0.0298, 0.0197], device='cuda:0'), grad: tensor([ 1.0157e-04, 8.9121e-04, -4.4584e-04, 2.4486e-04, 6.1214e-05, + 9.5665e-05, -1.4670e-05, -4.7073e-03, 7.1716e-04, 3.0556e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 49---------------------------------------------------- +epoch 49, time 273.89, cls_loss 0.0141 cls_loss_mapping 0.0260 cls_loss_causal 0.6992 re_mapping 0.0151 re_causal 0.0427 /// teacc 98.85 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0082, -0.0544, -0.0297, ..., -0.0014, -0.0907, 0.0010], + [-0.0314, -0.0765, -0.0682, ..., -0.0574, -0.0109, -0.0244], + [ 0.0385, 0.0468, -0.0558, ..., -0.0650, -0.0694, -0.0357], + ..., + [ 0.0032, 0.0178, -0.0099, ..., -0.0293, -0.0774, 0.0419], + [-0.0107, 0.0572, 0.0523, ..., -0.0357, 0.0280, -0.0588], + [-0.0405, -0.0162, 0.0077, ..., -0.0098, -0.0325, 0.0304]], + device='cuda:0'), grad: tensor([[ 8.0392e-06, 5.5671e-05, 4.0799e-05, ..., 7.5065e-06, + 2.8789e-05, 1.3210e-05], + [ 2.5535e-04, 1.0996e-03, 8.5974e-04, ..., 3.6985e-05, + 6.6376e-04, -1.7607e-04], + [ 1.3292e-04, 1.6165e-04, 1.6522e-04, ..., 1.3299e-05, + 1.4603e-04, 2.0945e-04], + ..., + [-4.4727e-04, -3.4690e-04, 7.5161e-05, ..., 6.4559e-06, + 3.3259e-05, -5.6219e-04], + [-3.4857e-04, -1.8396e-03, -1.5440e-03, ..., -1.0914e-04, + -1.1425e-03, 2.5749e-05], + [ 1.1563e-04, 2.7728e-04, 9.8884e-05, ..., 3.8296e-06, + 7.1943e-05, 2.4283e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0248, -0.0198, -0.0057, 0.0014, 0.0033, -0.0002, -0.0154, -0.0009, + 0.0296, 0.0195], device='cuda:0'), grad: tensor([ 7.3731e-05, 2.0199e-03, 7.0620e-04, 4.5347e-04, 1.9097e-04, + 1.8513e-04, 4.8923e-04, -9.7656e-04, -3.8376e-03, 6.9189e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 256.43, cls_loss 0.0106 cls_loss_mapping 0.0214 cls_loss_causal 0.6583 re_mapping 0.0148 re_causal 0.0445 /// teacc 98.76 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0083, -0.0547, -0.0303, ..., -0.0014, -0.0914, 0.0009], + [-0.0325, -0.0774, -0.0688, ..., -0.0576, -0.0115, -0.0239], + [ 0.0396, 0.0476, -0.0563, ..., -0.0652, -0.0699, -0.0357], + ..., + [ 0.0031, 0.0175, -0.0103, ..., -0.0295, -0.0777, 0.0416], + [-0.0107, 0.0576, 0.0527, ..., -0.0358, 0.0284, -0.0594], + [-0.0403, -0.0155, 0.0074, ..., -0.0099, -0.0328, 0.0309]], + device='cuda:0'), grad: tensor([[ 1.0028e-05, 2.6822e-05, 1.8314e-05, ..., 1.4879e-05, + 1.6928e-05, 1.5274e-05], + [ 1.1563e-05, 1.6734e-05, 5.5581e-06, ..., 1.7397e-06, + 7.2457e-06, -2.5332e-05], + [ 5.0020e-04, 5.7507e-04, 1.0826e-05, ..., 2.4009e-06, + 1.5423e-05, 2.5988e-04], + ..., + [-5.8603e-04, -6.4993e-04, 3.2574e-05, ..., 1.8277e-07, + 3.3528e-05, -2.7657e-04], + [-1.8311e-04, -3.5596e-04, -2.3615e-04, ..., 6.9439e-06, + -3.0899e-04, -5.0664e-05], + [ 1.2207e-04, 2.1958e-04, 2.0635e-04, ..., 1.1735e-06, + 2.4438e-04, -2.8491e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 2.4986e-02, -1.9963e-02, -5.0843e-03, 1.1135e-03, 2.9460e-03, + -5.9599e-05, -1.5603e-02, -8.7716e-04, 2.9459e-02, 1.9553e-02], + device='cuda:0'), grad: tensor([ 4.4823e-05, -5.0008e-05, 6.0368e-04, 4.2343e-04, 2.8685e-05, + -2.7871e-04, -2.4393e-05, -5.8556e-04, -2.8205e-04, 1.1867e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 256.54, cls_loss 0.0110 cls_loss_mapping 0.0227 cls_loss_causal 0.6552 re_mapping 0.0143 re_causal 0.0419 /// teacc 98.58 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0076, -0.0550, -0.0312, ..., -0.0019, -0.0925, 0.0008], + [-0.0331, -0.0779, -0.0695, ..., -0.0580, -0.0122, -0.0238], + [ 0.0402, 0.0480, -0.0569, ..., -0.0656, -0.0701, -0.0360], + ..., + [ 0.0027, 0.0173, -0.0107, ..., -0.0295, -0.0787, 0.0417], + [-0.0097, 0.0587, 0.0532, ..., -0.0367, 0.0291, -0.0597], + [-0.0405, -0.0156, 0.0069, ..., -0.0101, -0.0332, 0.0311]], + device='cuda:0'), grad: tensor([[-8.5682e-08, 3.6657e-05, 6.1929e-05, ..., 9.8571e-06, + 3.7283e-05, 1.6969e-06], + [ 9.6932e-06, 3.4630e-05, 5.5969e-05, ..., 1.2003e-05, + 3.0905e-05, -3.0184e-04], + [ 3.8564e-05, 1.0413e-04, 1.5998e-04, ..., 2.4408e-05, + 9.6262e-05, 4.4078e-05], + ..., + [-1.8239e-04, -2.2411e-04, 2.9728e-05, ..., 3.5241e-06, + 5.0068e-06, -1.2219e-04], + [ 3.3593e-04, -9.4986e-04, -2.3289e-03, ..., -9.5320e-04, + -9.9850e-04, 1.2770e-05], + [ 4.9204e-05, 8.2552e-05, 6.8784e-05, ..., 9.1270e-06, + 3.9369e-05, 4.9561e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([ 2.4774e-02, -1.9746e-02, -5.0063e-03, 1.0615e-03, 3.1965e-03, + -6.8045e-05, -1.5259e-02, -1.4891e-03, 2.9790e-02, 1.9421e-02], + device='cuda:0'), grad: tensor([ 4.3690e-05, -8.3351e-04, 2.7442e-04, 3.4595e-04, 4.8113e-04, + -1.6823e-03, 2.7122e-03, -1.7488e-04, -1.3523e-03, 1.9014e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 256.48, cls_loss 0.0099 cls_loss_mapping 0.0242 cls_loss_causal 0.6785 re_mapping 0.0151 re_causal 0.0448 /// teacc 98.82 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0065, -0.0554, -0.0315, ..., -0.0036, -0.0928, 0.0015], + [-0.0338, -0.0795, -0.0703, ..., -0.0578, -0.0129, -0.0237], + [ 0.0403, 0.0479, -0.0574, ..., -0.0660, -0.0708, -0.0364], + ..., + [ 0.0028, 0.0178, -0.0110, ..., -0.0298, -0.0795, 0.0419], + [-0.0095, 0.0594, 0.0538, ..., -0.0368, 0.0300, -0.0599], + [-0.0409, -0.0158, 0.0065, ..., -0.0102, -0.0337, 0.0310]], + device='cuda:0'), grad: tensor([[-6.2466e-05, 8.2612e-05, 3.8356e-05, ..., 9.6142e-05, + 3.5435e-05, -1.6363e-06], + [ 8.9705e-06, -2.7633e-04, 2.9117e-05, ..., 1.0389e-04, + 2.9534e-05, -2.2483e-04], + [-7.6532e-05, 7.1406e-05, 3.2693e-05, ..., 6.3300e-05, + 3.2514e-05, 2.2441e-05], + ..., + [ 3.3557e-05, 1.3602e-04, 1.1429e-05, ..., 1.6406e-05, + 1.1422e-05, 1.4913e-04], + [-2.2903e-05, 1.3232e-04, -1.6525e-05, ..., 8.2672e-05, + -1.8671e-05, 8.4519e-05], + [ 1.4566e-05, -1.1331e-04, 3.2634e-05, ..., 3.5286e-05, + 2.5839e-05, -2.1899e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0254, -0.0202, -0.0054, 0.0010, 0.0031, 0.0003, -0.0152, -0.0012, + 0.0303, 0.0189], device='cuda:0'), grad: tensor([ 0.0002, -0.0014, 0.0002, 0.0004, 0.0015, 0.0007, -0.0027, 0.0005, + 0.0008, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 256.63, cls_loss 0.0097 cls_loss_mapping 0.0218 cls_loss_causal 0.6871 re_mapping 0.0138 re_causal 0.0428 /// teacc 98.57 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0062, -0.0556, -0.0319, ..., -0.0041, -0.0933, 0.0016], + [-0.0343, -0.0795, -0.0705, ..., -0.0584, -0.0132, -0.0221], + [ 0.0406, 0.0479, -0.0579, ..., -0.0661, -0.0715, -0.0370], + ..., + [ 0.0033, 0.0181, -0.0113, ..., -0.0297, -0.0798, 0.0417], + [-0.0093, 0.0597, 0.0541, ..., -0.0374, 0.0303, -0.0607], + [-0.0411, -0.0157, 0.0061, ..., -0.0100, -0.0341, 0.0316]], + device='cuda:0'), grad: tensor([[-2.1920e-05, -4.6613e-07, -1.9129e-06, ..., -3.6191e-06, + 8.2254e-06, -6.3628e-06], + [ 3.3021e-05, 2.6911e-05, 1.0483e-05, ..., 1.4789e-06, + 1.3120e-05, -4.0196e-06], + [-1.4627e-04, -1.3947e-04, 7.1824e-06, ..., 1.9912e-06, + 9.2685e-06, -2.7511e-06], + ..., + [ 6.3255e-06, 1.3866e-05, 7.0892e-06, ..., 4.2561e-07, + 1.9697e-07, -1.9163e-05], + [ 8.0705e-05, 8.4162e-05, 3.7313e-05, ..., 7.6517e-06, + 2.9400e-05, 1.3083e-05], + [ 5.4352e-06, 2.8871e-06, 2.2084e-05, ..., 1.0598e-06, + 2.2009e-05, 1.4221e-06]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0257, -0.0192, -0.0056, 0.0016, 0.0026, -0.0003, -0.0152, -0.0015, + 0.0300, 0.0190], device='cuda:0'), grad: tensor([-8.0884e-05, 5.7131e-05, -1.8907e-04, 1.3804e-04, 1.9237e-05, + -2.2602e-04, 3.6478e-05, 8.7023e-06, 1.9240e-04, 4.3482e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 256.27, cls_loss 0.0101 cls_loss_mapping 0.0190 cls_loss_causal 0.6973 re_mapping 0.0147 re_causal 0.0428 /// teacc 98.75 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0057, -0.0559, -0.0323, ..., -0.0044, -0.0937, 0.0016], + [-0.0362, -0.0805, -0.0708, ..., -0.0582, -0.0134, -0.0224], + [ 0.0415, 0.0480, -0.0584, ..., -0.0663, -0.0722, -0.0368], + ..., + [ 0.0036, 0.0187, -0.0119, ..., -0.0301, -0.0804, 0.0423], + [-0.0092, 0.0601, 0.0542, ..., -0.0379, 0.0305, -0.0615], + [-0.0413, -0.0159, 0.0057, ..., -0.0103, -0.0345, 0.0319]], + device='cuda:0'), grad: tensor([[-4.1239e-06, 4.3660e-06, 6.6943e-06, ..., 4.4890e-06, + 4.7050e-06, 4.2059e-06], + [ 9.5665e-06, 1.1556e-05, 5.8338e-06, ..., 8.9081e-07, + 8.4862e-06, -2.4050e-05], + [ 1.0058e-05, 9.9987e-06, 9.0748e-06, ..., 1.1185e-06, + 1.0848e-05, 1.1101e-05], + ..., + [-5.5879e-05, -2.5779e-05, 3.1758e-06, ..., 8.0792e-08, + 6.6496e-06, -3.6031e-05], + [ 6.3935e-07, -9.0823e-06, -2.8089e-06, ..., 3.5092e-06, + 1.9581e-07, 1.0215e-05], + [ 1.6494e-06, -3.6992e-06, 5.6140e-06, ..., 4.5146e-07, + 6.4746e-06, 4.1560e-07]], device='cuda:0') +Epoch 56, bias, value: tensor([ 2.6025e-02, -1.9512e-02, -5.2482e-03, 1.1847e-03, 2.9592e-03, + 5.2053e-05, -1.5376e-02, -1.5131e-03, 2.9576e-02, 1.8909e-02], + device='cuda:0'), grad: tensor([ 6.3963e-06, -4.5359e-05, 5.3108e-05, -8.3670e-06, -1.3128e-05, + 3.2276e-05, -1.3664e-05, -6.8188e-05, 4.5329e-05, 1.1414e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 256.56, cls_loss 0.0080 cls_loss_mapping 0.0194 cls_loss_causal 0.6532 re_mapping 0.0137 re_causal 0.0410 /// teacc 98.75 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0053, -0.0566, -0.0332, ..., -0.0050, -0.0945, 0.0015], + [-0.0367, -0.0806, -0.0716, ..., -0.0593, -0.0137, -0.0221], + [ 0.0418, 0.0482, -0.0590, ..., -0.0669, -0.0730, -0.0372], + ..., + [ 0.0038, 0.0189, -0.0122, ..., -0.0301, -0.0808, 0.0427], + [-0.0090, 0.0605, 0.0544, ..., -0.0383, 0.0307, -0.0623], + [-0.0414, -0.0159, 0.0054, ..., -0.0106, -0.0348, 0.0319]], + device='cuda:0'), grad: tensor([[-3.6687e-05, 2.8327e-05, 6.9439e-05, ..., 8.3819e-06, + 8.4221e-05, 3.3051e-05], + [ 4.7177e-05, 6.3717e-05, 3.5584e-05, ..., 5.2564e-06, + 4.6641e-05, -3.5256e-05], + [-1.0079e-04, -1.1605e-04, 6.1154e-05, ..., 9.8869e-06, + 5.4270e-05, 3.4094e-05], + ..., + [ 2.5681e-07, 7.9200e-06, 2.3872e-05, ..., 1.0813e-06, + 4.1336e-05, 3.3170e-05], + [-7.2658e-05, -9.3341e-05, -3.3069e-04, ..., 1.4879e-05, + -3.2592e-04, 2.5108e-05], + [ 3.5733e-05, -9.9838e-06, 1.3411e-04, ..., 5.1595e-06, + 4.7922e-05, -3.4523e-04]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0261, -0.0193, -0.0053, 0.0011, 0.0034, 0.0001, -0.0154, -0.0016, + 0.0293, 0.0187], device='cuda:0'), grad: tensor([ 9.3222e-05, 9.5248e-05, -6.8426e-05, 4.0894e-03, 7.0953e-04, + -3.4027e-03, -4.0531e-04, 2.0909e-04, -1.4853e-04, -1.1683e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 56---------------------------------------------------- +epoch 56, time 272.32, cls_loss 0.0099 cls_loss_mapping 0.0215 cls_loss_causal 0.6489 re_mapping 0.0136 re_causal 0.0397 /// teacc 98.87 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0047, -0.0571, -0.0346, ..., -0.0052, -0.0960, 0.0012], + [-0.0378, -0.0816, -0.0732, ..., -0.0597, -0.0148, -0.0221], + [ 0.0421, 0.0484, -0.0596, ..., -0.0672, -0.0737, -0.0378], + ..., + [ 0.0042, 0.0192, -0.0126, ..., -0.0301, -0.0811, 0.0430], + [-0.0090, 0.0610, 0.0551, ..., -0.0386, 0.0315, -0.0632], + [-0.0414, -0.0157, 0.0048, ..., -0.0107, -0.0353, 0.0323]], + device='cuda:0'), grad: tensor([[ 2.0996e-05, 1.5855e-05, 2.0802e-05, ..., 5.8077e-06, + 1.9774e-05, 1.2510e-05], + [ 1.3843e-05, 8.2254e-06, 6.7763e-06, ..., 1.4007e-06, + 9.0972e-06, -6.4313e-05], + [ 4.0412e-05, 3.1769e-05, 1.5095e-05, ..., 1.1250e-06, + 2.6554e-05, 4.6581e-05], + ..., + [-5.6934e-04, -4.5687e-05, 1.9491e-05, ..., 3.7402e-06, + 2.3618e-05, -5.8413e-04], + [ 1.1575e-04, 4.2111e-05, 1.4508e-04, ..., 4.2439e-05, + 1.4496e-04, 4.2945e-05], + [-8.4519e-05, -2.3031e-04, 2.5883e-05, ..., 1.5534e-06, + -6.5491e-06, -5.8860e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0261, -0.0199, -0.0057, 0.0014, 0.0035, 0.0008, -0.0158, -0.0011, + 0.0292, 0.0186], device='cuda:0'), grad: tensor([ 5.3555e-05, -1.2040e-04, 1.2493e-04, 9.1374e-05, 9.8419e-04, + -1.3363e-04, 1.4663e-04, -9.8515e-04, 2.8419e-04, -4.4441e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 257.23, cls_loss 0.0075 cls_loss_mapping 0.0161 cls_loss_causal 0.6540 re_mapping 0.0135 re_causal 0.0404 /// teacc 98.85 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0043, -0.0574, -0.0350, ..., -0.0056, -0.0964, 0.0010], + [-0.0386, -0.0820, -0.0736, ..., -0.0602, -0.0151, -0.0220], + [ 0.0425, 0.0487, -0.0600, ..., -0.0674, -0.0742, -0.0382], + ..., + [ 0.0040, 0.0192, -0.0132, ..., -0.0303, -0.0815, 0.0433], + [-0.0089, 0.0611, 0.0552, ..., -0.0393, 0.0316, -0.0638], + [-0.0412, -0.0152, 0.0045, ..., -0.0109, -0.0355, 0.0326]], + device='cuda:0'), grad: tensor([[-4.0866e-06, 1.8269e-05, 1.7509e-05, ..., 7.7561e-06, + 8.9705e-06, -6.4410e-06], + [ 5.1886e-05, 6.9439e-05, 2.7969e-05, ..., 1.0476e-05, + 1.3188e-05, 5.6207e-05], + [ 5.7489e-05, 7.0870e-05, 3.5673e-05, ..., 1.1899e-05, + 1.6958e-05, 6.2287e-05], + ..., + [-3.2258e-04, -4.3583e-04, 4.8019e-06, ..., 1.1846e-06, + 2.5313e-06, -5.1641e-04], + [-5.5695e-04, -6.6614e-04, -9.9468e-04, ..., -3.0732e-04, + -4.1223e-04, 3.7432e-05], + [ 1.9050e-04, 2.0289e-04, 8.7738e-05, ..., 9.8124e-06, + 1.6958e-05, 2.3067e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0260, -0.0197, -0.0057, 0.0018, 0.0029, 0.0007, -0.0155, -0.0014, + 0.0289, 0.0189], device='cuda:0'), grad: tensor([-1.1377e-05, 1.4377e-04, 1.6296e-04, 1.6797e-04, -3.3259e-04, + 2.3234e-04, 9.4175e-04, -9.5749e-04, -1.0929e-03, 7.4482e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 256.63, cls_loss 0.0075 cls_loss_mapping 0.0174 cls_loss_causal 0.6943 re_mapping 0.0128 re_causal 0.0410 /// teacc 98.68 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0033, -0.0574, -0.0356, ..., -0.0059, -0.0973, 0.0013], + [-0.0390, -0.0822, -0.0740, ..., -0.0599, -0.0155, -0.0215], + [ 0.0427, 0.0484, -0.0607, ..., -0.0675, -0.0750, -0.0385], + ..., + [ 0.0044, 0.0196, -0.0135, ..., -0.0309, -0.0820, 0.0435], + [-0.0084, 0.0619, 0.0557, ..., -0.0397, 0.0323, -0.0643], + [-0.0413, -0.0151, 0.0041, ..., -0.0112, -0.0359, 0.0326]], + device='cuda:0'), grad: tensor([[-7.5758e-05, 1.5259e-05, 1.0878e-05, ..., 4.6156e-06, + 3.9414e-06, -1.5825e-05], + [ 1.4119e-05, 1.7986e-05, -2.9624e-05, ..., 1.5777e-06, + 8.2627e-06, -1.0288e-04], + [-1.2779e-04, -4.1306e-05, 1.6987e-05, ..., -2.5094e-05, + -3.5524e-05, 1.6376e-05], + ..., + [-1.2165e-04, -2.0647e-04, 3.8128e-06, ..., 1.2880e-06, + 4.6603e-06, -1.7536e-04], + [-3.0939e-06, -2.5183e-05, -7.7710e-06, ..., 9.4697e-06, + -3.1739e-05, 5.2720e-05], + [ 9.8467e-05, 1.5581e-04, 1.2472e-05, ..., 2.3041e-06, + 7.9051e-06, 1.4436e-04]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0263, -0.0194, -0.0060, 0.0016, 0.0022, 0.0005, -0.0152, -0.0016, + 0.0292, 0.0192], device='cuda:0'), grad: tensor([-9.4235e-05, -2.4128e-04, -2.2650e-04, 3.3760e-04, -1.3268e-04, + 9.8169e-05, 6.7532e-05, -2.4462e-04, 1.0645e-04, 3.2949e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 256.55, cls_loss 0.0119 cls_loss_mapping 0.0243 cls_loss_causal 0.6653 re_mapping 0.0133 re_causal 0.0384 /// teacc 98.83 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0025, -0.0581, -0.0366, ..., -0.0061, -0.0985, 0.0009], + [-0.0402, -0.0831, -0.0746, ..., -0.0614, -0.0161, -0.0213], + [ 0.0433, 0.0485, -0.0611, ..., -0.0679, -0.0753, -0.0390], + ..., + [ 0.0048, 0.0202, -0.0138, ..., -0.0311, -0.0824, 0.0438], + [-0.0084, 0.0626, 0.0563, ..., -0.0403, 0.0327, -0.0653], + [-0.0416, -0.0149, 0.0036, ..., -0.0117, -0.0364, 0.0323]], + device='cuda:0'), grad: tensor([[-4.7255e-04, -1.9872e-04, -2.6539e-05, ..., 1.3527e-07, + 5.0059e-07, -8.3625e-05], + [ 1.5914e-05, 4.2506e-06, 5.8413e-06, ..., 3.1199e-08, + 6.6347e-06, -3.2097e-05], + [-1.3459e-04, -3.9315e-04, 6.2026e-06, ..., 4.9360e-08, + -4.7177e-05, -8.6784e-05], + ..., + [ 8.6963e-05, 1.3745e-04, 2.9169e-06, ..., 3.8883e-08, + 2.1473e-05, 4.9531e-05], + [ 3.6359e-04, 2.5964e-04, 5.4955e-05, ..., 7.1339e-06, + 4.6432e-05, 9.1076e-05], + [ 1.2159e-05, -1.3433e-05, 9.2462e-06, ..., 2.0419e-07, + 8.3521e-06, -1.6347e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0256, -0.0197, -0.0064, 0.0016, 0.0039, 0.0004, -0.0149, -0.0014, + 0.0291, 0.0187], device='cuda:0'), grad: tensor([-7.3814e-04, -5.1796e-05, -3.6168e-04, 2.4116e-04, -1.7130e-04, + -2.6122e-05, 1.1581e-04, 2.5010e-04, 6.6566e-04, 7.7009e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 256.64, cls_loss 0.0105 cls_loss_mapping 0.0198 cls_loss_causal 0.6604 re_mapping 0.0135 re_causal 0.0376 /// teacc 98.67 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0020, -0.0589, -0.0371, ..., -0.0070, -0.0995, 0.0003], + [-0.0412, -0.0838, -0.0751, ..., -0.0620, -0.0167, -0.0210], + [ 0.0437, 0.0489, -0.0618, ..., -0.0684, -0.0758, -0.0395], + ..., + [ 0.0050, 0.0201, -0.0142, ..., -0.0312, -0.0836, 0.0437], + [-0.0082, 0.0630, 0.0564, ..., -0.0404, 0.0327, -0.0657], + [-0.0418, -0.0146, 0.0030, ..., -0.0120, -0.0370, 0.0328]], + device='cuda:0'), grad: tensor([[ 1.8597e-05, 2.8044e-05, 1.5989e-05, ..., 1.3925e-05, + 9.0748e-06, 2.0713e-05], + [ 3.1519e-04, 2.5225e-04, 5.1111e-06, ..., 2.6617e-06, + 3.2961e-05, 3.4285e-04], + [ 9.5558e-04, 1.7715e-04, 8.6501e-06, ..., 1.3057e-06, + 3.3379e-05, 5.5170e-04], + ..., + [-2.3117e-03, -1.0376e-03, 7.4059e-06, ..., 2.7730e-07, + -5.0694e-05, -1.8702e-03], + [ 2.5988e-04, 1.9896e-04, 2.8070e-06, ..., 6.0759e-06, + 3.0383e-05, 2.9850e-04], + [ 2.2262e-05, 7.5325e-06, 1.2174e-05, ..., 8.8429e-07, + 3.0473e-05, 3.4064e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0256, -0.0197, -0.0066, 0.0023, 0.0036, 0.0005, -0.0154, -0.0013, + 0.0289, 0.0189], device='cuda:0'), grad: tensor([ 7.1347e-05, 5.7030e-04, 1.3256e-03, -2.6989e-04, 6.0987e-04, + 7.8773e-04, -3.0375e-04, -3.3932e-03, 5.3120e-04, 6.8486e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 256.87, cls_loss 0.0075 cls_loss_mapping 0.0179 cls_loss_causal 0.6402 re_mapping 0.0133 re_causal 0.0395 /// teacc 98.83 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0015, -0.0593, -0.0376, ..., -0.0071, -0.1002, 0.0002], + [-0.0419, -0.0842, -0.0755, ..., -0.0623, -0.0170, -0.0208], + [ 0.0436, 0.0486, -0.0628, ..., -0.0686, -0.0765, -0.0403], + ..., + [ 0.0059, 0.0209, -0.0145, ..., -0.0312, -0.0840, 0.0442], + [-0.0077, 0.0637, 0.0572, ..., -0.0406, 0.0336, -0.0664], + [-0.0418, -0.0148, 0.0028, ..., -0.0119, -0.0373, 0.0332]], + device='cuda:0'), grad: tensor([[-9.2626e-05, -1.3493e-05, 1.6704e-05, ..., 1.5367e-08, + 1.1615e-05, -1.6997e-06], + [ 1.8314e-05, 2.9549e-05, 7.3910e-06, ..., 2.7940e-09, + 7.7859e-06, -3.8520e-06], + [-1.3441e-05, -6.7912e-06, 7.4394e-06, ..., 1.8626e-09, + 9.2983e-06, 1.1206e-05], + ..., + [-7.1228e-05, -1.7297e-04, 1.1593e-05, ..., 1.2340e-08, + 8.9630e-06, -9.3758e-05], + [-1.2845e-05, -4.3303e-05, -6.6519e-05, ..., 9.4064e-08, + -5.2661e-05, 5.9903e-06], + [ 1.1164e-04, 1.6224e-04, 1.0425e-04, ..., 7.7300e-08, + 7.6830e-05, 6.1631e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 2.5724e-02, -1.9717e-02, -7.1135e-03, 2.3144e-03, 4.1779e-03, + 7.1365e-05, -1.5483e-02, -6.1219e-04, 2.9038e-02, 1.8492e-02], + device='cuda:0'), grad: tensor([-3.6120e-04, 1.3590e-05, 3.4958e-05, 3.6329e-05, 2.3365e-05, + -1.4639e-04, 1.8370e-04, -1.6606e-04, 3.0324e-06, 3.7885e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 256.34, cls_loss 0.0081 cls_loss_mapping 0.0177 cls_loss_causal 0.6572 re_mapping 0.0129 re_causal 0.0375 /// teacc 98.83 lr 0.00010000 +Epoch 64, weight, value: tensor([[-1.2839e-03, -5.9813e-02, -3.8527e-02, ..., -7.4525e-03, + -1.0122e-01, -4.0430e-05], + [-4.2566e-02, -8.5597e-02, -7.6964e-02, ..., -6.4692e-02, + -1.8765e-02, -2.0637e-02], + [ 4.3938e-02, 4.8839e-02, -6.3455e-02, ..., -6.8890e-02, + -7.6510e-02, -4.0566e-02], + ..., + [ 6.4619e-03, 2.1703e-02, -1.4420e-02, ..., -3.1293e-02, + -8.3626e-02, 4.4538e-02], + [-7.8719e-03, 6.4043e-02, 5.7459e-02, ..., -4.0561e-02, + 3.3759e-02, -6.7257e-02], + [-4.2224e-02, -1.4898e-02, 2.3034e-03, ..., -1.2204e-02, + -3.7643e-02, 3.3122e-02]], device='cuda:0'), grad: tensor([[-6.7055e-06, 6.2548e-06, 9.1493e-06, ..., 1.3206e-06, + 8.6129e-06, 5.8077e-06], + [ 1.0341e-05, 1.2919e-05, 7.0855e-06, ..., -4.9360e-07, + 1.6138e-05, -4.0472e-05], + [ 1.9118e-05, 2.0444e-05, 1.1280e-05, ..., 7.3947e-07, + 4.4405e-05, 6.6876e-05], + ..., + [ 1.8284e-05, 1.0476e-05, 5.0813e-06, ..., 1.6298e-07, + 1.8612e-05, 2.9549e-05], + [ 3.1646e-06, -1.4147e-06, 4.5896e-05, ..., 7.4729e-06, + 3.3557e-05, 2.8133e-05], + [ 7.2047e-06, 1.6708e-06, 1.4082e-05, ..., 1.9511e-07, + 1.3314e-05, -1.1194e-06]], device='cuda:0') +Epoch 64, bias, value: tensor([ 2.5741e-02, -2.0571e-02, -6.7446e-03, 1.7965e-03, 4.4808e-03, + 7.8124e-04, -1.5096e-02, -3.4100e-05, 2.8628e-02, 1.7988e-02], + device='cuda:0'), grad: tensor([ 5.0999e-06, -3.8147e-05, 1.3852e-04, -2.4605e-04, 2.1905e-05, + -1.4853e-04, 8.8751e-05, 7.4267e-05, 7.9036e-05, 2.5004e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 255.97, cls_loss 0.0073 cls_loss_mapping 0.0165 cls_loss_causal 0.6767 re_mapping 0.0126 re_causal 0.0384 /// teacc 98.69 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0009, -0.0603, -0.0389, ..., -0.0079, -0.1018, -0.0001], + [-0.0431, -0.0852, -0.0774, ..., -0.0651, -0.0188, -0.0193], + [ 0.0445, 0.0494, -0.0641, ..., -0.0692, -0.0769, -0.0410], + ..., + [ 0.0058, 0.0214, -0.0163, ..., -0.0314, -0.0843, 0.0438], + [-0.0076, 0.0649, 0.0584, ..., -0.0404, 0.0344, -0.0680], + [-0.0422, -0.0144, 0.0020, ..., -0.0123, -0.0379, 0.0337]], + device='cuda:0'), grad: tensor([[-2.0906e-05, -4.0770e-04, -4.5323e-04, ..., 2.8908e-05, + 2.9922e-05, -6.9427e-04], + [ 5.1916e-05, 7.2598e-05, 1.4171e-05, ..., 2.0694e-06, + 7.3947e-06, -8.3745e-05], + [ 8.6650e-06, 1.5512e-05, 1.0177e-05, ..., 8.3540e-07, + 1.6272e-05, 4.1366e-05], + ..., + [-7.0274e-05, -9.0241e-05, 7.7188e-06, ..., 2.7288e-07, + 1.4760e-05, -1.2565e-04], + [ 2.7746e-05, 9.0003e-05, 1.3483e-04, ..., 4.6045e-05, + 9.2268e-05, 8.0764e-05], + [ 3.3118e-06, 3.3993e-06, 2.6301e-05, ..., 3.0231e-06, + 9.0972e-06, 2.4036e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0258, -0.0199, -0.0064, 0.0017, 0.0040, 0.0016, -0.0159, -0.0013, + 0.0288, 0.0185], device='cuda:0'), grad: tensor([-2.8629e-03, -2.4819e-04, 1.1051e-04, -6.5088e-05, 4.2701e-04, + 2.5635e-03, -6.2108e-05, -2.2542e-04, 2.5368e-04, 1.0800e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 256.28, cls_loss 0.0082 cls_loss_mapping 0.0175 cls_loss_causal 0.6467 re_mapping 0.0116 re_causal 0.0363 /// teacc 98.81 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0007, -0.0610, -0.0394, ..., -0.0085, -0.1027, -0.0002], + [-0.0435, -0.0846, -0.0780, ..., -0.0653, -0.0195, -0.0190], + [ 0.0446, 0.0491, -0.0648, ..., -0.0694, -0.0772, -0.0409], + ..., + [ 0.0057, 0.0212, -0.0170, ..., -0.0314, -0.0852, 0.0437], + [-0.0069, 0.0657, 0.0591, ..., -0.0409, 0.0351, -0.0693], + [-0.0418, -0.0139, 0.0020, ..., -0.0125, -0.0379, 0.0344]], + device='cuda:0'), grad: tensor([[-2.6189e-06, 5.5842e-06, 2.4159e-06, ..., 2.0806e-06, + 1.8552e-06, -8.0094e-06], + [ 1.2003e-05, 1.6242e-05, 9.6671e-07, ..., 1.6950e-07, + 1.6708e-06, -1.3029e-06], + [-6.3837e-05, -1.9446e-05, 7.3528e-07, ..., 9.6625e-08, + 2.2929e-06, 8.6650e-06], + ..., + [ 1.8282e-06, -2.2113e-05, 2.1467e-07, ..., 1.3271e-08, + 1.2062e-05, -3.2187e-05], + [ 6.0052e-06, 2.3115e-06, -7.0687e-07, ..., 9.7509e-07, + 1.1614e-06, 7.3835e-06], + [ 1.7628e-05, 3.7521e-05, 7.5856e-07, ..., 2.1746e-07, + 1.1818e-06, 3.8218e-04]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0259, -0.0201, -0.0060, 0.0016, 0.0039, 0.0014, -0.0165, -0.0016, + 0.0289, 0.0191], device='cuda:0'), grad: tensor([ 3.6389e-05, 1.6540e-05, -1.8668e-04, 5.8562e-05, -2.2316e-03, + 1.5795e-05, 1.5661e-05, 4.3154e-05, 3.8087e-05, 2.1915e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 65---------------------------------------------------- +epoch 65, time 273.92, cls_loss 0.0075 cls_loss_mapping 0.0167 cls_loss_causal 0.6222 re_mapping 0.0125 re_causal 0.0367 /// teacc 98.88 lr 0.00010000 +Epoch 67, weight, value: tensor([[-2.0603e-05, -6.1639e-02, -4.0704e-02, ..., -8.9565e-03, + -1.0458e-01, -2.6783e-04], + [-4.4231e-02, -8.5306e-02, -7.8906e-02, ..., -6.6170e-02, + -2.0162e-02, -1.9085e-02], + [ 4.4850e-02, 4.9048e-02, -6.5618e-02, ..., -6.9612e-02, + -7.8054e-02, -4.1381e-02], + ..., + [ 5.4832e-03, 2.1248e-02, -1.7342e-02, ..., -3.1538e-02, + -8.6029e-02, 4.3753e-02], + [-6.6382e-03, 6.6384e-02, 5.9615e-02, ..., -4.1311e-02, + 3.5609e-02, -6.9969e-02], + [-4.2128e-02, -1.3857e-02, 1.6435e-03, ..., -1.2707e-02, + -3.8092e-02, 3.4588e-02]], device='cuda:0'), grad: tensor([[ 3.7514e-06, 8.4639e-06, 1.0662e-05, ..., 2.0713e-06, + 1.2428e-05, 4.7274e-06], + [ 5.3912e-05, 5.1826e-05, 7.1079e-06, ..., 5.4203e-07, + 8.7842e-06, 5.4896e-05], + [-2.5645e-05, 4.3124e-05, 1.4111e-05, ..., 1.7486e-07, + -2.0936e-05, 2.5034e-05], + ..., + [-1.1146e-04, -1.0026e-04, 1.0125e-05, ..., 1.9092e-07, + 2.3052e-05, -1.4460e-04], + [ 1.1402e-04, 3.2282e-04, 1.0359e-04, ..., 3.8818e-06, + 1.3340e-04, 1.5092e-04], + [-1.3745e-04, -5.0020e-04, -1.6367e-04, ..., 5.4622e-07, + -1.2362e-04, -2.3615e-04]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0258, -0.0206, -0.0063, 0.0020, 0.0039, 0.0017, -0.0163, -0.0016, + 0.0291, 0.0191], device='cuda:0'), grad: tensor([ 3.0458e-05, 7.0810e-05, -1.8072e-04, -4.7088e-04, 1.7321e-04, + 5.4169e-04, -2.5779e-06, -1.3900e-04, 5.0306e-04, -5.2595e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 256.83, cls_loss 0.0078 cls_loss_mapping 0.0165 cls_loss_causal 0.6682 re_mapping 0.0121 re_causal 0.0377 /// teacc 98.65 lr 0.00010000 +Epoch 68, weight, value: tensor([[ 0.0007, -0.0622, -0.0405, ..., -0.0097, -0.1043, -0.0007], + [-0.0450, -0.0860, -0.0793, ..., -0.0666, -0.0204, -0.0186], + [ 0.0449, 0.0487, -0.0662, ..., -0.0697, -0.0787, -0.0422], + ..., + [ 0.0064, 0.0224, -0.0175, ..., -0.0316, -0.0862, 0.0442], + [-0.0064, 0.0670, 0.0602, ..., -0.0415, 0.0361, -0.0708], + [-0.0423, -0.0137, 0.0009, ..., -0.0130, -0.0387, 0.0351]], + device='cuda:0'), grad: tensor([[ 6.2436e-06, 9.9391e-06, 5.4985e-05, ..., 3.0044e-06, + 3.3170e-05, 3.3796e-05], + [ 1.5870e-05, -1.7166e-05, 2.4900e-05, ..., 6.3470e-07, + 1.4082e-05, -2.7999e-05], + [ 9.5785e-05, 1.3685e-04, 1.2703e-05, ..., 4.7497e-07, + 1.0140e-05, 2.4140e-05], + ..., + [-1.1319e-04, -1.0139e-04, 1.1958e-05, ..., 1.8231e-07, + 7.1637e-06, 1.6809e-04], + [ 2.2322e-05, 2.5719e-05, 8.5771e-05, ..., 2.9653e-06, + 4.6939e-05, 5.4091e-05], + [-1.1510e-04, -1.6177e-04, 2.9221e-05, ..., 1.0515e-06, + 1.7464e-05, -2.9731e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0267, -0.0206, -0.0065, 0.0020, 0.0034, 0.0011, -0.0169, -0.0007, + 0.0291, 0.0190], device='cuda:0'), grad: tensor([ 1.2094e-04, 5.5122e-04, 2.7823e-04, 5.2243e-05, -1.8225e-03, + -3.1877e-04, 2.3103e-04, 9.1267e-04, 2.0647e-04, -2.1327e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 255.57, cls_loss 0.0074 cls_loss_mapping 0.0166 cls_loss_causal 0.6581 re_mapping 0.0116 re_causal 0.0356 /// teacc 98.85 lr 0.00010000 +Epoch 69, weight, value: tensor([[ 8.7428e-04, -6.3210e-02, -4.1182e-02, ..., -1.0532e-02, + -1.0498e-01, -1.1126e-03], + [-4.5573e-02, -8.6667e-02, -7.9679e-02, ..., -6.7155e-02, + -2.0986e-02, -1.7908e-02], + [ 4.5019e-02, 4.8490e-02, -6.6956e-02, ..., -7.0138e-02, + -7.9265e-02, -4.2981e-02], + ..., + [ 6.9094e-03, 2.2966e-02, -1.7682e-02, ..., -3.1659e-02, + -8.6639e-02, 4.4504e-02], + [-6.0203e-03, 6.7481e-02, 6.0290e-02, ..., -4.2125e-02, + 3.6301e-02, -7.1359e-02], + [-4.2797e-02, -1.3476e-02, 2.9767e-05, ..., -1.3296e-02, + -3.9486e-02, 3.5437e-02]], device='cuda:0'), grad: tensor([[-2.0504e-05, 3.8669e-06, 2.4773e-06, ..., 5.5460e-07, + 2.3134e-06, 1.7704e-06], + [ 7.6517e-06, 4.7572e-06, 2.8946e-06, ..., 2.8964e-07, + 2.8685e-06, -4.8161e-05], + [-2.6986e-05, -2.1253e-06, 5.9754e-06, ..., 2.1909e-07, + 5.9344e-06, 5.1521e-06], + ..., + [-2.2084e-05, -4.3899e-05, 1.1936e-05, ..., 8.4983e-09, + 1.0431e-05, -5.8979e-05], + [-2.7046e-05, -4.3154e-05, -4.9353e-05, ..., 1.2079e-06, + -4.1157e-05, 1.0788e-05], + [ 5.0217e-05, 4.1306e-05, 1.9535e-05, ..., 1.1164e-07, + 1.7881e-05, 5.5730e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0263, -0.0202, -0.0072, 0.0020, 0.0043, 0.0013, -0.0179, -0.0003, + 0.0288, 0.0190], device='cuda:0'), grad: tensor([-5.9426e-05, -7.1466e-05, -3.1263e-05, 3.6865e-05, 1.7673e-05, + 2.6003e-05, 2.4095e-05, -6.1393e-05, -3.4779e-05, 1.5354e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 68---------------------------------------------------- +epoch 68, time 273.95, cls_loss 0.0060 cls_loss_mapping 0.0145 cls_loss_causal 0.6212 re_mapping 0.0112 re_causal 0.0345 /// teacc 98.94 lr 0.00010000 +Epoch 70, weight, value: tensor([[ 0.0015, -0.0638, -0.0416, ..., -0.0112, -0.1055, -0.0016], + [-0.0462, -0.0871, -0.0801, ..., -0.0674, -0.0214, -0.0176], + [ 0.0452, 0.0485, -0.0676, ..., -0.0705, -0.0800, -0.0434], + ..., + [ 0.0071, 0.0231, -0.0181, ..., -0.0317, -0.0870, 0.0447], + [-0.0055, 0.0683, 0.0607, ..., -0.0429, 0.0368, -0.0719], + [-0.0431, -0.0139, -0.0005, ..., -0.0134, -0.0397, 0.0354]], + device='cuda:0'), grad: tensor([[-1.0282e-06, 1.9986e-06, 1.3806e-05, ..., 2.5239e-07, + 1.3217e-05, 2.0247e-06], + [ 2.4959e-05, 2.0340e-05, 1.5572e-05, ..., 7.7649e-08, + 2.0251e-05, 1.4327e-05], + [ 7.6652e-05, 2.4602e-05, 1.6600e-05, ..., 6.3912e-08, + 1.2863e-04, 1.0020e-04], + ..., + [-1.5661e-05, -6.1244e-06, 9.0450e-06, ..., 1.9791e-09, + 1.5542e-05, -4.1485e-05], + [ 7.3947e-06, -2.0470e-06, 2.0280e-05, ..., 5.2433e-07, + 2.3544e-05, 9.7081e-06], + [ 2.0593e-05, 1.1557e-04, 4.8935e-05, ..., 3.5972e-08, + 4.3243e-05, 1.5497e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0260, -0.0201, -0.0076, 0.0015, 0.0039, 0.0014, -0.0168, -0.0002, + 0.0291, 0.0190], device='cuda:0'), grad: tensor([ 1.1653e-05, 7.4923e-05, 2.2936e-04, -2.4533e-04, -3.4332e-04, + -2.4843e-04, 8.2612e-05, 7.6964e-06, 5.7071e-05, 3.7432e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 256.70, cls_loss 0.0069 cls_loss_mapping 0.0145 cls_loss_causal 0.6239 re_mapping 0.0116 re_causal 0.0351 /// teacc 98.88 lr 0.00010000 +Epoch 71, weight, value: tensor([[ 0.0017, -0.0644, -0.0422, ..., -0.0119, -0.1060, -0.0017], + [-0.0473, -0.0880, -0.0806, ..., -0.0676, -0.0217, -0.0179], + [ 0.0456, 0.0488, -0.0681, ..., -0.0708, -0.0806, -0.0432], + ..., + [ 0.0072, 0.0229, -0.0183, ..., -0.0318, -0.0874, 0.0449], + [-0.0051, 0.0695, 0.0614, ..., -0.0429, 0.0377, -0.0723], + [-0.0432, -0.0140, -0.0009, ..., -0.0138, -0.0402, 0.0358]], + device='cuda:0'), grad: tensor([[ 1.4484e-05, 8.9407e-05, 9.5606e-05, ..., 1.7672e-07, + 4.8637e-05, 1.4476e-05], + [ 4.5508e-05, 1.1390e-04, 2.9877e-05, ..., 8.5449e-08, + 5.1051e-05, 8.2135e-05], + [ 2.1994e-05, 9.1076e-05, 7.3850e-05, ..., 9.1502e-08, + 6.4552e-05, 4.5240e-05], + ..., + [ 7.3090e-03, 1.6937e-02, 1.4432e-05, ..., 2.7940e-09, + 7.0686e-03, 1.3145e-02], + [-5.8603e-04, -3.4809e-03, -4.4937e-03, ..., 2.7800e-07, + -1.4191e-03, 2.2054e-04], + [ 1.1945e-04, 4.0030e-04, 2.9755e-04, ..., 3.4226e-08, + 2.0528e-04, 1.5044e-04]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0259, -0.0202, -0.0076, 0.0009, 0.0041, 0.0019, -0.0169, -0.0003, + 0.0298, 0.0188], device='cuda:0'), grad: tensor([ 1.3840e-04, 2.4796e-04, 2.1267e-04, -2.6596e-02, 2.3142e-05, + 5.4283e-03, 1.2010e-04, 2.5330e-02, -5.6419e-03, 7.3528e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 256.88, cls_loss 0.0064 cls_loss_mapping 0.0142 cls_loss_causal 0.6033 re_mapping 0.0112 re_causal 0.0343 /// teacc 98.80 lr 0.00010000 +Epoch 72, weight, value: tensor([[ 0.0021, -0.0647, -0.0426, ..., -0.0120, -0.1066, -0.0017], + [-0.0478, -0.0884, -0.0809, ..., -0.0679, -0.0220, -0.0174], + [ 0.0459, 0.0490, -0.0687, ..., -0.0711, -0.0814, -0.0435], + ..., + [ 0.0072, 0.0228, -0.0187, ..., -0.0319, -0.0884, 0.0443], + [-0.0050, 0.0700, 0.0618, ..., -0.0434, 0.0383, -0.0731], + [-0.0437, -0.0147, -0.0016, ..., -0.0143, -0.0410, 0.0358]], + device='cuda:0'), grad: tensor([[-1.7032e-05, 7.7263e-06, 7.0892e-06, ..., 5.2229e-06, + 5.9865e-06, 2.4401e-06], + [ 1.7108e-06, 8.0019e-06, 2.4755e-06, ..., 2.1271e-06, + 6.4336e-06, 4.3847e-06], + [ 1.3977e-05, -3.2457e-07, 2.6003e-06, ..., 1.3616e-06, + 3.4738e-06, 4.4443e-06], + ..., + [ 2.7090e-05, 1.5450e-04, 5.5227e-07, ..., 1.8929e-07, + 1.6296e-04, 2.8348e-04], + [ 3.8259e-06, 2.4527e-05, 1.9684e-05, ..., 1.2547e-05, + 1.6481e-05, 1.1839e-05], + [-3.6135e-06, -2.2091e-06, -1.3160e-06, ..., 1.0068e-06, + 9.8199e-06, 1.2092e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0262, -0.0200, -0.0073, 0.0015, 0.0044, 0.0019, -0.0172, -0.0008, + 0.0297, 0.0183], device='cuda:0'), grad: tensor([-2.9549e-05, 8.9332e-06, 4.4644e-05, -4.8447e-04, 3.3587e-05, + 2.8446e-05, -8.8096e-05, 4.3106e-04, 4.4107e-05, 1.1973e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 255.86, cls_loss 0.0074 cls_loss_mapping 0.0130 cls_loss_causal 0.6604 re_mapping 0.0108 re_causal 0.0352 /// teacc 98.85 lr 0.00010000 +Epoch 73, weight, value: tensor([[ 0.0026, -0.0649, -0.0431, ..., -0.0123, -0.1072, -0.0019], + [-0.0488, -0.0896, -0.0815, ..., -0.0688, -0.0242, -0.0180], + [ 0.0462, 0.0495, -0.0692, ..., -0.0714, -0.0803, -0.0425], + ..., + [ 0.0072, 0.0226, -0.0188, ..., -0.0317, -0.0904, 0.0441], + [-0.0050, 0.0702, 0.0617, ..., -0.0442, 0.0384, -0.0741], + [-0.0439, -0.0147, -0.0019, ..., -0.0145, -0.0414, 0.0360]], + device='cuda:0'), grad: tensor([[ 1.0550e-05, 1.4983e-05, 6.8322e-06, ..., 1.4342e-06, + 8.1956e-06, 9.9167e-06], + [ 3.3826e-05, 3.9607e-05, 5.6744e-05, ..., -8.9593e-07, + 7.1585e-05, 6.8247e-05], + [-7.0632e-05, -3.1948e-05, 2.3812e-05, ..., 1.2377e-06, + 3.0071e-05, 4.9889e-05], + ..., + [-1.8217e-06, -6.9261e-05, 4.4733e-05, ..., 7.1945e-08, + 5.6416e-05, -2.5734e-05], + [ 5.1826e-05, 8.3148e-05, 4.1157e-05, ..., 2.7753e-06, + 5.4926e-05, 7.0691e-05], + [-6.0588e-05, -1.1861e-04, 2.6926e-05, ..., 1.6647e-07, + 1.0200e-05, -3.5137e-05]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0265, -0.0208, -0.0062, 0.0015, 0.0045, 0.0027, -0.0173, -0.0014, + 0.0292, 0.0180], device='cuda:0'), grad: tensor([ 4.5210e-05, 2.2721e-04, -2.2158e-05, -8.1635e-04, 4.3750e-05, + 2.5940e-04, 3.6746e-05, 2.9325e-05, 2.2578e-04, -2.9221e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 256.37, cls_loss 0.0081 cls_loss_mapping 0.0160 cls_loss_causal 0.6303 re_mapping 0.0113 re_causal 0.0316 /// teacc 98.84 lr 0.00010000 +Epoch 74, weight, value: tensor([[ 0.0026, -0.0662, -0.0443, ..., -0.0127, -0.1088, -0.0027], + [-0.0494, -0.0903, -0.0823, ..., -0.0696, -0.0250, -0.0175], + [ 0.0468, 0.0500, -0.0696, ..., -0.0720, -0.0807, -0.0435], + ..., + [ 0.0073, 0.0231, -0.0192, ..., -0.0318, -0.0904, 0.0444], + [-0.0049, 0.0706, 0.0620, ..., -0.0443, 0.0389, -0.0748], + [-0.0439, -0.0144, -0.0025, ..., -0.0148, -0.0422, 0.0364]], + device='cuda:0'), grad: tensor([[ 1.8859e-08, 4.0799e-05, 2.1458e-06, ..., 1.6941e-06, + 3.2596e-06, 8.6948e-06], + [ 2.5351e-06, 9.7677e-06, 3.2783e-06, ..., 4.9500e-07, + 3.4049e-06, -4.7288e-07], + [-6.9253e-06, -2.4238e-07, 8.0466e-07, ..., 4.6776e-07, + 9.9465e-07, 3.3323e-06], + ..., + [-5.3160e-06, -2.4691e-05, 8.1677e-07, ..., 1.3039e-08, + 1.5786e-06, -2.3335e-05], + [ 6.6347e-06, 3.8207e-05, 5.6505e-05, ..., 1.7043e-06, + 4.3333e-05, 2.5302e-05], + [-5.5619e-06, -1.8537e-05, 1.3858e-05, ..., 1.2619e-07, + 2.6766e-06, -1.2152e-05]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0256, -0.0203, -0.0064, 0.0017, 0.0051, 0.0023, -0.0172, -0.0011, + 0.0289, 0.0179], device='cuda:0'), grad: tensor([ 6.4433e-05, 7.8082e-06, -3.1898e-07, 3.0518e-05, 1.3580e-03, + -9.1493e-05, -1.4238e-03, -4.1038e-05, 1.1528e-04, -1.7866e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 256.54, cls_loss 0.0065 cls_loss_mapping 0.0154 cls_loss_causal 0.5874 re_mapping 0.0112 re_causal 0.0323 /// teacc 98.89 lr 0.00010000 +Epoch 75, weight, value: tensor([[ 0.0028, -0.0667, -0.0446, ..., -0.0130, -0.1092, -0.0029], + [-0.0501, -0.0908, -0.0825, ..., -0.0702, -0.0253, -0.0179], + [ 0.0472, 0.0500, -0.0699, ..., -0.0727, -0.0810, -0.0438], + ..., + [ 0.0074, 0.0231, -0.0194, ..., -0.0318, -0.0911, 0.0447], + [-0.0048, 0.0704, 0.0620, ..., -0.0448, 0.0391, -0.0754], + [-0.0438, -0.0128, -0.0027, ..., -0.0151, -0.0420, 0.0372]], + device='cuda:0'), grad: tensor([[-8.6753e-07, 2.6479e-05, 9.8944e-05, ..., 1.6749e-05, + 7.8201e-05, 2.8193e-05], + [ 1.7034e-06, 5.9903e-06, 1.1668e-05, ..., 2.5090e-06, + 1.1027e-05, -1.7151e-05], + [ 1.2536e-06, 6.1616e-06, 1.2062e-05, ..., 1.8859e-06, + 1.1817e-05, 5.5432e-06], + ..., + [ 1.4156e-06, 1.3383e-06, 9.1195e-06, ..., 1.9395e-07, + 9.6560e-06, 2.5705e-06], + [ 7.3947e-06, -1.6373e-06, 1.3924e-04, ..., 1.3694e-05, + 1.0824e-04, 2.0355e-05], + [ 1.7926e-05, 1.5065e-05, 7.8619e-05, ..., 1.5935e-06, + 7.4565e-05, 4.1090e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0256, -0.0207, -0.0066, 0.0016, 0.0046, 0.0019, -0.0169, -0.0011, + 0.0282, 0.0195], device='cuda:0'), grad: tensor([ 1.1432e-04, -2.7210e-05, 2.9504e-05, -1.0484e-04, 3.6478e-05, + -1.0926e-04, -2.5773e-04, 2.4974e-05, 1.6916e-04, 1.2481e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 256.30, cls_loss 0.0083 cls_loss_mapping 0.0149 cls_loss_causal 0.6236 re_mapping 0.0110 re_causal 0.0311 /// teacc 98.92 lr 0.00010000 +Epoch 76, weight, value: tensor([[ 0.0028, -0.0675, -0.0455, ..., -0.0135, -0.1100, -0.0032], + [-0.0510, -0.0914, -0.0836, ..., -0.0712, -0.0259, -0.0177], + [ 0.0477, 0.0498, -0.0708, ..., -0.0732, -0.0816, -0.0443], + ..., + [ 0.0075, 0.0233, -0.0197, ..., -0.0318, -0.0914, 0.0449], + [-0.0044, 0.0716, 0.0629, ..., -0.0460, 0.0397, -0.0760], + [-0.0435, -0.0122, -0.0035, ..., -0.0154, -0.0429, 0.0377]], + device='cuda:0'), grad: tensor([[-9.0227e-06, 9.3579e-06, 2.3954e-06, ..., 1.5553e-06, + 9.4548e-06, -8.4490e-06], + [ 5.0701e-06, 5.4762e-06, 1.8002e-06, ..., 5.7183e-07, + 1.1966e-05, -3.4682e-06], + [-2.1422e-04, -1.5175e-04, 1.3448e-06, ..., -4.9965e-07, + -2.2354e-03, -5.9515e-05], + ..., + [ 2.7210e-05, -3.8967e-06, 6.5984e-07, ..., -3.3788e-06, + 1.2495e-05, -9.9540e-06], + [ 4.1097e-05, 4.4674e-05, 7.3053e-06, ..., 2.8592e-06, + 2.2843e-05, 2.4699e-06], + [ 2.8953e-05, 1.4096e-05, -3.7234e-06, ..., 3.2187e-06, + 2.1381e-03, 2.9370e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0242, -0.0205, -0.0066, 0.0014, 0.0045, 0.0018, -0.0167, -0.0010, + 0.0285, 0.0201], device='cuda:0'), grad: tensor([-5.1945e-05, 3.4511e-05, -1.2840e-02, 3.0589e-04, 3.7789e-05, + 1.1724e-04, 3.7923e-06, 5.9754e-05, 1.3220e-04, 1.2207e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 256.19, cls_loss 0.0061 cls_loss_mapping 0.0124 cls_loss_causal 0.6168 re_mapping 0.0116 re_causal 0.0338 /// teacc 98.74 lr 0.00010000 +Epoch 77, weight, value: tensor([[ 0.0033, -0.0679, -0.0471, ..., -0.0140, -0.1113, -0.0036], + [-0.0517, -0.0918, -0.0845, ..., -0.0718, -0.0263, -0.0174], + [ 0.0476, 0.0497, -0.0714, ..., -0.0731, -0.0817, -0.0448], + ..., + [ 0.0077, 0.0238, -0.0206, ..., -0.0318, -0.0916, 0.0452], + [-0.0038, 0.0723, 0.0634, ..., -0.0463, 0.0402, -0.0762], + [-0.0436, -0.0117, -0.0036, ..., -0.0157, -0.0435, 0.0375]], + device='cuda:0'), grad: tensor([[-6.2659e-06, 1.4305e-06, 1.5544e-06, ..., 7.1060e-07, + 1.0161e-06, 5.4808e-07], + [-1.3057e-06, 3.3434e-06, -1.6037e-06, ..., 4.6054e-07, + 2.1253e-06, -1.4819e-05], + [ 3.8818e-06, 5.9977e-06, 8.3447e-06, ..., 2.6505e-06, + 6.9551e-06, 3.1255e-06], + ..., + [-9.2804e-05, -1.5569e-04, 3.4310e-06, ..., 1.9791e-08, + 3.8370e-06, -1.1081e-04], + [ 5.3756e-06, 1.7673e-05, 2.9624e-05, ..., 1.0885e-05, + 2.1175e-05, 8.5384e-06], + [ 8.6606e-05, 1.3816e-04, 2.4103e-06, ..., 1.1758e-07, + 1.6652e-06, 1.0437e-04]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0242, -0.0207, -0.0066, 0.0009, 0.0043, 0.0026, -0.0162, -0.0011, + 0.0288, 0.0198], device='cuda:0'), grad: tensor([ 6.5938e-07, -3.6836e-05, 2.4170e-05, -2.3648e-05, -9.3699e-05, + 1.8567e-05, -2.8938e-05, -1.5020e-04, 5.3316e-05, 2.3627e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 256.68, cls_loss 0.0078 cls_loss_mapping 0.0145 cls_loss_causal 0.6116 re_mapping 0.0112 re_causal 0.0331 /// teacc 98.78 lr 0.00010000 +Epoch 78, weight, value: tensor([[ 0.0039, -0.0682, -0.0474, ..., -0.0142, -0.1115, -0.0036], + [-0.0528, -0.0926, -0.0865, ..., -0.0727, -0.0272, -0.0174], + [ 0.0474, 0.0489, -0.0728, ..., -0.0732, -0.0823, -0.0456], + ..., + [ 0.0081, 0.0244, -0.0210, ..., -0.0319, -0.0921, 0.0454], + [-0.0028, 0.0726, 0.0634, ..., -0.0471, 0.0402, -0.0769], + [-0.0438, -0.0119, -0.0048, ..., -0.0158, -0.0446, 0.0379]], + device='cuda:0'), grad: tensor([[ 3.1561e-05, 2.9370e-05, 2.5988e-05, ..., 1.4693e-05, + 2.0579e-05, 5.6103e-06], + [ 1.5748e-04, 1.4305e-04, 4.6563e-04, ..., 6.9067e-06, + 4.0507e-04, 6.4135e-05], + [-3.0351e-04, -8.1435e-06, 2.0266e-05, ..., 1.0505e-05, + 1.9982e-05, 6.6422e-06], + ..., + [ 1.0103e-04, 9.6485e-06, 2.8655e-05, ..., 4.1840e-07, + 2.8476e-05, -8.5831e-06], + [-1.4520e-04, -1.6940e-04, -6.2895e-04, ..., 1.8090e-05, + -4.8018e-04, -6.8188e-05], + [ 1.7092e-05, 1.3001e-05, 5.8621e-05, ..., 2.0675e-06, + 5.7399e-05, 1.1079e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([ 2.4505e-02, -2.1379e-02, -7.4757e-03, 1.2477e-03, 5.5767e-03, + 2.9672e-03, -1.6983e-02, -9.5242e-05, 2.8568e-02, 1.8986e-02], + device='cuda:0'), grad: tensor([ 1.3101e-04, 8.4686e-04, -5.9843e-04, -7.6115e-05, 1.0449e-04, + -1.3582e-05, -1.3614e-04, 2.5225e-04, -7.0381e-04, 1.9383e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 77---------------------------------------------------- +epoch 77, time 273.90, cls_loss 0.0068 cls_loss_mapping 0.0121 cls_loss_causal 0.5957 re_mapping 0.0109 re_causal 0.0316 /// teacc 98.95 lr 0.00010000 +Epoch 79, weight, value: tensor([[ 0.0040, -0.0688, -0.0480, ..., -0.0147, -0.1121, -0.0039], + [-0.0537, -0.0929, -0.0865, ..., -0.0730, -0.0275, -0.0170], + [ 0.0490, 0.0502, -0.0735, ..., -0.0735, -0.0827, -0.0457], + ..., + [ 0.0070, 0.0231, -0.0213, ..., -0.0320, -0.0926, 0.0452], + [-0.0026, 0.0733, 0.0648, ..., -0.0470, 0.0412, -0.0779], + [-0.0427, -0.0109, -0.0055, ..., -0.0161, -0.0451, 0.0391]], + device='cuda:0'), grad: tensor([[-3.3307e-04, -1.1379e-04, 1.5467e-05, ..., 6.0536e-07, + 7.7859e-06, -9.9897e-05], + [ 1.9884e-04, 6.9022e-05, 9.1642e-06, ..., 1.1711e-07, + 5.9381e-06, -4.9162e-04], + [ 3.1590e-05, 1.2085e-05, 5.1558e-06, ..., 1.2340e-07, + 3.1516e-06, 2.3913e-04], + ..., + [ 1.1824e-05, -1.3895e-06, 7.2233e-06, ..., 9.0338e-08, + 4.1462e-06, 1.9681e-04], + [ 6.5923e-05, 3.1814e-06, 3.0398e-04, ..., 3.2969e-06, + 1.4913e-04, 1.0476e-05], + [ 1.2957e-05, 2.9225e-06, 3.2932e-05, ..., 7.5949e-07, + 1.8641e-05, 4.2260e-05]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0245, -0.0211, -0.0066, 0.0014, 0.0050, 0.0024, -0.0175, -0.0003, + 0.0284, 0.0195], device='cuda:0'), grad: tensor([-0.0036, 0.0012, 0.0008, 0.0004, -0.0002, -0.0005, 0.0006, 0.0006, + 0.0005, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 255.43, cls_loss 0.0055 cls_loss_mapping 0.0140 cls_loss_causal 0.6005 re_mapping 0.0110 re_causal 0.0323 /// teacc 98.90 lr 0.00010000 +Epoch 80, weight, value: tensor([[ 0.0052, -0.0692, -0.0486, ..., -0.0152, -0.1127, -0.0041], + [-0.0542, -0.0932, -0.0867, ..., -0.0732, -0.0277, -0.0157], + [ 0.0490, 0.0501, -0.0746, ..., -0.0738, -0.0834, -0.0460], + ..., + [ 0.0068, 0.0231, -0.0215, ..., -0.0320, -0.0927, 0.0445], + [-0.0027, 0.0738, 0.0649, ..., -0.0479, 0.0414, -0.0785], + [-0.0430, -0.0111, -0.0060, ..., -0.0164, -0.0455, 0.0388]], + device='cuda:0'), grad: tensor([[ 1.4435e-07, 2.7344e-06, -5.7742e-06, ..., 8.7311e-07, + 1.2117e-06, -9.5665e-06], + [ 3.7730e-05, 4.0382e-05, 1.2200e-06, ..., 1.4692e-07, + 5.6624e-07, 1.1629e-04], + [ 1.7941e-05, 1.9625e-05, 5.7183e-07, ..., 1.4459e-07, + 4.8475e-07, 3.0130e-05], + ..., + [-1.5247e-04, -1.7619e-04, 4.5262e-07, ..., 3.6787e-08, + 2.2026e-07, -2.7776e-04], + [ 7.1861e-06, 9.6262e-06, 5.3607e-06, ..., 1.6559e-06, + 2.2240e-06, 5.5581e-06], + [ 1.6406e-05, 1.8388e-05, 3.8370e-06, ..., 4.7171e-07, + 1.9222e-06, 2.9400e-05]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0249, -0.0203, -0.0073, 0.0022, 0.0051, 0.0019, -0.0167, -0.0009, + 0.0282, 0.0190], device='cuda:0'), grad: tensor([-5.2780e-05, 2.2399e-04, 5.7489e-05, 1.9169e-04, 2.1055e-05, + 1.9632e-06, 2.9922e-05, -5.7650e-04, 3.0428e-05, 7.3016e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 256.76, cls_loss 0.0057 cls_loss_mapping 0.0145 cls_loss_causal 0.6349 re_mapping 0.0101 re_causal 0.0322 /// teacc 98.74 lr 0.00010000 +Epoch 81, weight, value: tensor([[ 0.0056, -0.0693, -0.0491, ..., -0.0155, -0.1134, -0.0039], + [-0.0554, -0.0940, -0.0872, ..., -0.0739, -0.0279, -0.0163], + [ 0.0496, 0.0502, -0.0756, ..., -0.0740, -0.0835, -0.0455], + ..., + [ 0.0073, 0.0237, -0.0216, ..., -0.0320, -0.0928, 0.0449], + [-0.0030, 0.0740, 0.0652, ..., -0.0481, 0.0416, -0.0790], + [-0.0432, -0.0113, -0.0065, ..., -0.0164, -0.0459, 0.0391]], + device='cuda:0'), grad: tensor([[-2.2367e-05, 3.9250e-05, 5.2124e-05, ..., 2.6613e-05, + 1.7852e-05, 6.2063e-06], + [ 6.1840e-06, 4.4592e-06, 1.7006e-06, ..., 3.7323e-07, + 1.7090e-06, -6.3144e-06], + [-2.5249e-04, -1.2970e-04, 9.9242e-06, ..., 6.6543e-07, + 1.1824e-05, 1.4640e-06], + ..., + [ 2.1172e-04, 1.2958e-04, 6.1933e-07, ..., 3.1199e-08, + 1.1558e-06, 5.6531e-07], + [ 1.0990e-05, -3.4243e-05, -2.8655e-05, ..., 1.1846e-06, + -3.8147e-05, 3.6042e-06], + [-4.3539e-07, -8.0094e-06, 1.5600e-06, ..., 3.7323e-07, + 1.5297e-07, -6.8769e-06]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0251, -0.0214, -0.0070, 0.0020, 0.0051, 0.0021, -0.0162, -0.0003, + 0.0279, 0.0190], device='cuda:0'), grad: tensor([ 5.7101e-05, 2.9220e-07, -4.1199e-04, 9.9480e-05, 1.1683e-05, + 2.5392e-05, -8.4519e-05, 3.2520e-04, -2.5362e-05, 3.0212e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 257.05, cls_loss 0.0047 cls_loss_mapping 0.0117 cls_loss_causal 0.6047 re_mapping 0.0107 re_causal 0.0331 /// teacc 98.91 lr 0.00010000 +Epoch 82, weight, value: tensor([[ 0.0058, -0.0700, -0.0509, ..., -0.0161, -0.1155, -0.0043], + [-0.0558, -0.0938, -0.0879, ..., -0.0741, -0.0284, -0.0161], + [ 0.0502, 0.0502, -0.0761, ..., -0.0741, -0.0835, -0.0460], + ..., + [ 0.0073, 0.0240, -0.0221, ..., -0.0320, -0.0929, 0.0451], + [-0.0031, 0.0742, 0.0653, ..., -0.0483, 0.0416, -0.0795], + [-0.0436, -0.0114, -0.0067, ..., -0.0167, -0.0457, 0.0391]], + device='cuda:0'), grad: tensor([[-7.2531e-06, -6.3656e-07, 2.3074e-07, ..., 8.2655e-08, + 6.2818e-07, 3.6322e-08], + [ 3.0152e-07, 2.8359e-07, 4.5006e-07, ..., 1.3504e-08, + 7.4226e-07, -4.6372e-05], + [ 1.1642e-06, 3.1688e-07, 1.1176e-06, ..., 6.7754e-08, + 1.6280e-06, 2.0385e-05], + ..., + [ 5.7416e-07, 3.8813e-07, 8.5123e-07, ..., 1.6298e-09, + 1.7975e-06, 4.2729e-06], + [-2.1346e-06, -5.0552e-06, -6.4485e-06, ..., 2.3725e-07, + -2.4326e-06, 1.3430e-06], + [ 1.5153e-06, -2.7940e-07, 1.7881e-06, ..., 1.5367e-08, + 3.3453e-06, 5.7602e-07]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0244, -0.0211, -0.0063, 0.0012, 0.0054, 0.0028, -0.0160, -0.0004, + 0.0273, 0.0188], device='cuda:0'), grad: tensor([-1.9729e-05, -1.2100e-04, 6.4075e-05, -3.9816e-05, -1.2949e-05, + 5.8651e-05, 2.7552e-05, 1.9163e-05, 6.4783e-06, 1.7777e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 256.66, cls_loss 0.0057 cls_loss_mapping 0.0124 cls_loss_causal 0.6072 re_mapping 0.0104 re_causal 0.0312 /// teacc 98.95 lr 0.00010000 +Epoch 83, weight, value: tensor([[ 0.0058, -0.0707, -0.0526, ..., -0.0166, -0.1172, -0.0049], + [-0.0562, -0.0953, -0.0879, ..., -0.0742, -0.0285, -0.0158], + [ 0.0504, 0.0510, -0.0771, ..., -0.0742, -0.0842, -0.0461], + ..., + [ 0.0077, 0.0243, -0.0226, ..., -0.0320, -0.0931, 0.0454], + [-0.0028, 0.0746, 0.0656, ..., -0.0487, 0.0419, -0.0802], + [-0.0436, -0.0111, -0.0072, ..., -0.0163, -0.0461, 0.0392]], + device='cuda:0'), grad: tensor([[-3.4302e-05, 2.2091e-06, 2.2631e-06, ..., 4.0862e-07, + 1.8161e-06, -2.2780e-06], + [ 4.3958e-06, 3.1162e-06, 2.2631e-06, ..., 1.0058e-07, + 1.8440e-06, -7.8082e-06], + [ 4.5389e-05, 3.8177e-05, 2.2985e-06, ..., 2.3399e-07, + 2.0973e-06, 7.4804e-05], + ..., + [ 2.1562e-05, 2.1249e-05, 1.1921e-06, ..., 6.7521e-09, + 2.5500e-06, 3.3051e-05], + [ 2.2575e-06, -3.7886e-06, -2.5332e-05, ..., 1.1567e-06, + -1.6585e-05, 1.1422e-05], + [-1.1310e-05, -3.5137e-05, 2.5686e-06, ..., 5.9139e-08, + -4.9919e-06, -1.6406e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0239, -0.0209, -0.0060, 0.0013, 0.0054, 0.0029, -0.0163, -0.0004, + 0.0272, 0.0188], device='cuda:0'), grad: tensor([-9.9242e-05, -1.6749e-05, 4.7326e-04, -6.6519e-04, 1.4797e-05, + 4.3571e-05, 3.8356e-05, 2.1040e-04, 2.7895e-05, -2.8282e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 257.17, cls_loss 0.0060 cls_loss_mapping 0.0123 cls_loss_causal 0.6140 re_mapping 0.0100 re_causal 0.0301 /// teacc 98.77 lr 0.00010000 +Epoch 84, weight, value: tensor([[ 0.0060, -0.0711, -0.0534, ..., -0.0169, -0.1180, -0.0051], + [-0.0567, -0.0966, -0.0901, ..., -0.0747, -0.0305, -0.0138], + [ 0.0504, 0.0506, -0.0778, ..., -0.0747, -0.0847, -0.0467], + ..., + [ 0.0080, 0.0247, -0.0230, ..., -0.0321, -0.0933, 0.0448], + [-0.0025, 0.0758, 0.0665, ..., -0.0498, 0.0432, -0.0813], + [-0.0435, -0.0111, -0.0075, ..., -0.0165, -0.0462, 0.0393]], + device='cuda:0'), grad: tensor([[-1.1422e-05, 4.6454e-06, 8.7842e-06, ..., 4.7572e-06, + 5.6848e-06, 6.3516e-07], + [ 1.1651e-06, -8.1509e-06, 3.5055e-06, ..., 1.9968e-06, + 3.6471e-06, -7.1406e-05], + [ 5.3160e-06, 1.8850e-05, 6.4299e-06, ..., 1.7434e-06, + 9.3728e-06, 6.8486e-05], + ..., + [ 5.6513e-06, -5.5246e-06, 6.7614e-07, ..., 6.3097e-08, + 1.8328e-05, 3.3498e-05], + [ 2.0228e-06, 7.9572e-06, 2.5764e-05, ..., 1.5721e-05, + 9.6411e-06, 7.5698e-06], + [ 4.5076e-06, 4.5672e-06, 3.6396e-06, ..., 8.7265e-07, + 4.6566e-06, 8.7395e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0240, -0.0208, -0.0063, 0.0013, 0.0050, 0.0026, -0.0158, -0.0010, + 0.0282, 0.0188], device='cuda:0'), grad: tensor([-2.5347e-05, -1.2565e-04, 1.3733e-04, -5.2810e-05, -2.9728e-06, + 9.7334e-05, -1.5128e-04, 4.7415e-05, 3.6031e-05, 3.9637e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 257.23, cls_loss 0.0053 cls_loss_mapping 0.0114 cls_loss_causal 0.6007 re_mapping 0.0107 re_causal 0.0318 /// teacc 98.86 lr 0.00010000 +Epoch 85, weight, value: tensor([[ 0.0066, -0.0714, -0.0538, ..., -0.0179, -0.1178, -0.0056], + [-0.0572, -0.0969, -0.0903, ..., -0.0749, -0.0306, -0.0137], + [ 0.0507, 0.0507, -0.0787, ..., -0.0754, -0.0854, -0.0471], + ..., + [ 0.0081, 0.0249, -0.0233, ..., -0.0321, -0.0935, 0.0450], + [-0.0026, 0.0758, 0.0664, ..., -0.0503, 0.0431, -0.0820], + [-0.0434, -0.0110, -0.0079, ..., -0.0169, -0.0465, 0.0396]], + device='cuda:0'), grad: tensor([[-4.0770e-05, 1.4463e-06, 2.3078e-06, ..., 1.8785e-06, + -4.5099e-07, 5.0478e-07], + [ 8.6650e-06, 8.9407e-06, 6.4224e-06, ..., 1.0328e-06, + 7.6443e-06, -1.6391e-05], + [-1.4044e-06, -2.2855e-06, 2.4989e-05, ..., 5.0664e-06, + 4.0263e-05, 1.7002e-05], + ..., + [-1.8895e-05, 5.7332e-06, 3.8520e-06, ..., 2.1607e-07, + 6.0014e-06, -2.2322e-05], + [-7.8902e-06, -1.4104e-05, -1.0066e-05, ..., 3.0082e-06, + -8.9034e-06, 4.7162e-06], + [-3.7197e-06, -3.7760e-05, -3.6806e-05, ..., 3.6927e-07, + -2.7165e-05, -1.1280e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0249, -0.0210, -0.0063, 0.0012, 0.0050, 0.0026, -0.0158, -0.0008, + 0.0277, 0.0188], device='cuda:0'), grad: tensor([-8.9109e-05, -1.0990e-05, 3.5644e-05, 2.4378e-05, 1.4067e-04, + 4.1038e-05, 5.3972e-05, -2.2814e-05, 5.3167e-05, -2.2602e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 84---------------------------------------------------- +epoch 84, time 275.12, cls_loss 0.0053 cls_loss_mapping 0.0102 cls_loss_causal 0.5924 re_mapping 0.0103 re_causal 0.0303 /// teacc 98.96 lr 0.00010000 +Epoch 86, weight, value: tensor([[ 0.0070, -0.0717, -0.0543, ..., -0.0179, -0.1184, -0.0057], + [-0.0575, -0.0971, -0.0905, ..., -0.0754, -0.0306, -0.0120], + [ 0.0508, 0.0503, -0.0795, ..., -0.0765, -0.0866, -0.0475], + ..., + [ 0.0083, 0.0246, -0.0236, ..., -0.0321, -0.0942, 0.0438], + [-0.0024, 0.0762, 0.0666, ..., -0.0507, 0.0434, -0.0825], + [-0.0430, -0.0102, -0.0084, ..., -0.0172, -0.0470, 0.0403]], + device='cuda:0'), grad: tensor([[-9.4017e-07, 5.5939e-05, 2.5019e-05, ..., 4.4614e-05, + 1.8910e-05, 1.5959e-05], + [ 4.1798e-06, 1.1057e-05, 3.2336e-06, ..., 2.3395e-06, + 2.2743e-06, -1.3340e-04], + [-4.6566e-08, 8.4341e-06, 4.6529e-06, ..., 4.6566e-06, + 3.3397e-06, 5.7034e-06], + ..., + [-2.2173e-05, -4.5985e-05, 1.3784e-06, ..., 2.2375e-07, + 9.5367e-07, 2.0079e-06], + [ 1.6941e-06, 2.3782e-05, 3.0965e-05, ..., 1.8567e-05, + 2.2218e-05, 9.6038e-06], + [ 1.1131e-05, 2.5049e-05, 6.3144e-06, ..., 3.1963e-06, + 4.1984e-06, 8.0645e-05]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0252, -0.0200, -0.0070, 0.0019, 0.0048, 0.0021, -0.0157, -0.0015, + 0.0276, 0.0189], device='cuda:0'), grad: tensor([ 9.3222e-05, -2.2554e-04, 2.1383e-05, 1.7357e-04, 1.0103e-05, + -1.7464e-04, -1.8120e-04, -1.0133e-06, 7.8857e-05, 2.0480e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 257.03, cls_loss 0.0047 cls_loss_mapping 0.0103 cls_loss_causal 0.6027 re_mapping 0.0100 re_causal 0.0296 /// teacc 98.87 lr 0.00010000 +Epoch 87, weight, value: tensor([[ 0.0074, -0.0719, -0.0550, ..., -0.0179, -0.1191, -0.0059], + [-0.0581, -0.0974, -0.0908, ..., -0.0758, -0.0308, -0.0119], + [ 0.0510, 0.0498, -0.0805, ..., -0.0771, -0.0874, -0.0481], + ..., + [ 0.0087, 0.0254, -0.0240, ..., -0.0321, -0.0946, 0.0446], + [-0.0015, 0.0768, 0.0669, ..., -0.0518, 0.0437, -0.0834], + [-0.0434, -0.0107, -0.0091, ..., -0.0176, -0.0476, 0.0400]], + device='cuda:0'), grad: tensor([[ 1.8209e-05, 9.8571e-06, 6.1810e-05, ..., 1.5460e-07, + 7.0393e-05, 8.2180e-06], + [ 1.1139e-05, 1.7524e-05, 1.6719e-05, ..., 3.1991e-07, + 1.5318e-05, -1.1083e-07], + [ 1.3776e-05, 7.6592e-06, 2.9370e-05, ..., 9.4436e-07, + 3.3885e-05, 5.8785e-06], + ..., + [ 1.6451e-05, 4.1783e-05, 8.6427e-06, ..., 3.5856e-08, + 7.6964e-06, 3.2932e-05], + [ 9.7990e-05, 1.2052e-04, 1.3061e-05, ..., 1.1027e-06, + 3.3110e-05, 9.6262e-05], + [-2.0170e-04, -3.5286e-04, 2.6822e-05, ..., 3.3597e-07, + 2.9147e-05, -2.3282e-04]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0253, -0.0200, -0.0075, 0.0015, 0.0051, 0.0025, -0.0160, -0.0009, + 0.0283, 0.0181], device='cuda:0'), grad: tensor([ 2.6369e-04, 5.3525e-05, 1.3244e-04, -4.3559e-04, 5.5313e-04, + 6.6817e-05, 5.7817e-05, 1.0705e-04, 3.7074e-04, -1.1692e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 257.45, cls_loss 0.0064 cls_loss_mapping 0.0138 cls_loss_causal 0.5793 re_mapping 0.0107 re_causal 0.0297 /// teacc 98.88 lr 0.00010000 +Epoch 88, weight, value: tensor([[ 0.0081, -0.0726, -0.0554, ..., -0.0182, -0.1196, -0.0060], + [-0.0586, -0.0976, -0.0912, ..., -0.0766, -0.0310, -0.0117], + [ 0.0514, 0.0498, -0.0810, ..., -0.0770, -0.0880, -0.0485], + ..., + [ 0.0102, 0.0280, -0.0242, ..., -0.0318, -0.0946, 0.0458], + [-0.0015, 0.0768, 0.0669, ..., -0.0525, 0.0438, -0.0842], + [-0.0447, -0.0126, -0.0095, ..., -0.0181, -0.0478, 0.0393]], + device='cuda:0'), grad: tensor([[ 1.6671e-06, 3.6508e-06, 1.3914e-06, ..., 4.3819e-07, + 2.1569e-06, 1.5553e-07], + [ 3.0398e-06, 5.0813e-06, 1.4743e-06, ..., 2.1863e-07, + 3.5856e-06, 8.1258e-08], + [-9.0837e-05, -8.3506e-05, 1.7285e-06, ..., 3.9418e-07, + -1.7896e-05, 1.4100e-06], + ..., + [ 2.8983e-05, 2.1815e-05, 5.3598e-07, ..., 2.2119e-08, + 1.3039e-05, -1.0118e-05], + [ 8.7991e-06, 7.7784e-06, -6.3032e-06, ..., 4.4890e-06, + -5.5693e-06, 5.7742e-07], + [ 7.8380e-06, 1.4693e-05, 8.1509e-06, ..., 1.1828e-07, + 1.1489e-05, 3.1553e-06]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0254, -0.0199, -0.0074, 0.0012, 0.0050, 0.0031, -0.0170, 0.0007, + 0.0277, 0.0170], device='cuda:0'), grad: tensor([ 3.9116e-06, 1.3597e-05, -1.4043e-04, 4.7028e-05, -4.4137e-05, + 2.3782e-05, -1.3774e-06, 4.9621e-05, 1.3024e-05, 3.5167e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 256.87, cls_loss 0.0054 cls_loss_mapping 0.0110 cls_loss_causal 0.6113 re_mapping 0.0096 re_causal 0.0296 /// teacc 98.94 lr 0.00010000 +Epoch 89, weight, value: tensor([[ 0.0095, -0.0721, -0.0558, ..., -0.0184, -0.1200, -0.0076], + [-0.0596, -0.0982, -0.0913, ..., -0.0769, -0.0312, -0.0114], + [ 0.0519, 0.0505, -0.0815, ..., -0.0772, -0.0885, -0.0490], + ..., + [ 0.0098, 0.0278, -0.0245, ..., -0.0319, -0.0951, 0.0456], + [-0.0013, 0.0774, 0.0674, ..., -0.0526, 0.0444, -0.0847], + [-0.0451, -0.0128, -0.0100, ..., -0.0184, -0.0483, 0.0399]], + device='cuda:0'), grad: tensor([[-7.8022e-05, 2.4624e-06, 1.3001e-06, ..., 1.0999e-06, + 1.1856e-06, 8.2050e-07], + [ 2.6021e-06, 2.0992e-06, 6.4960e-07, ..., -1.1222e-07, + 6.0257e-07, -4.1097e-05], + [ 8.0094e-06, -1.8803e-06, 1.0775e-06, ..., 2.7008e-07, + 9.6206e-07, 6.5267e-06], + ..., + [-3.0790e-06, -1.8030e-05, 2.0722e-07, ..., 9.5228e-08, + 1.9209e-07, -1.0334e-05], + [ 3.6955e-05, -3.8445e-06, -6.8396e-06, ..., 1.6456e-06, + -5.3421e-06, 5.1968e-06], + [ 9.4622e-06, 3.3472e-06, 4.6636e-07, ..., 2.1723e-07, + 4.2235e-07, 4.8243e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0246, -0.0201, -0.0069, 0.0014, 0.0055, 0.0032, -0.0174, 0.0002, + 0.0277, 0.0174], device='cuda:0'), grad: tensor([-1.3626e-04, -7.7486e-05, 1.7196e-05, 2.8387e-05, 5.2869e-05, + 2.2590e-05, 1.2659e-05, -2.6777e-05, 7.7367e-05, 2.9311e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 257.51, cls_loss 0.0048 cls_loss_mapping 0.0120 cls_loss_causal 0.6052 re_mapping 0.0094 re_causal 0.0286 /// teacc 98.92 lr 0.00010000 +Epoch 90, weight, value: tensor([[ 0.0098, -0.0727, -0.0564, ..., -0.0188, -0.1206, -0.0076], + [-0.0599, -0.0984, -0.0916, ..., -0.0773, -0.0317, -0.0111], + [ 0.0527, 0.0513, -0.0823, ..., -0.0776, -0.0891, -0.0482], + ..., + [ 0.0097, 0.0278, -0.0249, ..., -0.0320, -0.0947, 0.0453], + [-0.0011, 0.0777, 0.0674, ..., -0.0526, 0.0443, -0.0851], + [-0.0451, -0.0126, -0.0100, ..., -0.0158, -0.0482, 0.0400]], + device='cuda:0'), grad: tensor([[-3.3118e-06, 5.0850e-06, 9.6206e-07, ..., 5.9837e-07, + 1.1129e-06, 1.0923e-05], + [-5.0932e-05, -2.0787e-05, 1.0887e-06, ..., 8.2888e-08, + 1.5553e-06, -6.7532e-05], + [ 1.5995e-07, -5.0589e-06, 2.6133e-06, ..., 1.7486e-07, + 3.5428e-06, 2.7344e-05], + ..., + [ 8.9049e-05, 1.1885e-04, 8.3726e-07, ..., 1.0245e-08, + 1.2498e-06, 2.2125e-04], + [ 3.4064e-05, 1.7330e-05, 1.5413e-06, ..., 7.1106e-07, + 2.3916e-06, 4.3988e-05], + [-8.4341e-05, -1.2708e-04, 1.3821e-06, ..., 8.0094e-08, + 2.1067e-06, -8.7380e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([ 2.4833e-02, -2.0306e-02, -6.4494e-03, 2.0833e-03, 5.4204e-03, + 2.9967e-03, -1.7723e-02, -4.4118e-05, 2.7293e-02, 1.7605e-02], + device='cuda:0'), grad: tensor([ 1.2584e-05, -2.6798e-04, 1.2493e-04, -3.7029e-06, -9.7752e-04, + 7.1466e-05, 6.3479e-05, 7.0906e-04, 2.0468e-04, 6.2466e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 255.45, cls_loss 0.0050 cls_loss_mapping 0.0111 cls_loss_causal 0.6131 re_mapping 0.0096 re_causal 0.0292 /// teacc 98.93 lr 0.00010000 +Epoch 91, weight, value: tensor([[ 0.0110, -0.0724, -0.0570, ..., -0.0185, -0.1212, -0.0073], + [-0.0603, -0.0987, -0.0920, ..., -0.0782, -0.0320, -0.0106], + [ 0.0524, 0.0504, -0.0833, ..., -0.0780, -0.0903, -0.0486], + ..., + [ 0.0098, 0.0280, -0.0251, ..., -0.0320, -0.0949, 0.0453], + [-0.0004, 0.0785, 0.0676, ..., -0.0533, 0.0446, -0.0862], + [-0.0451, -0.0122, -0.0103, ..., -0.0159, -0.0482, 0.0401]], + device='cuda:0'), grad: tensor([[ 3.0901e-06, 8.2105e-06, 1.6719e-05, ..., 3.6098e-06, + 1.2934e-05, 1.3029e-06], + [ 1.2465e-05, 2.6911e-05, 1.9252e-05, ..., 3.3248e-07, + 1.8179e-05, 1.6451e-05], + [ 2.5287e-05, 4.7773e-05, 8.4937e-05, ..., 5.4296e-07, + 7.3493e-05, 1.1764e-05], + ..., + [-1.8850e-05, -6.2525e-05, 5.9456e-05, ..., 2.5262e-08, + 4.4465e-05, -7.3910e-05], + [ 1.8930e-04, 2.8062e-04, 7.6771e-04, ..., 6.6385e-06, + 6.6566e-04, 1.5292e-06], + [ 1.3709e-05, 1.9491e-05, 2.2516e-05, ..., 2.2701e-07, + 2.0847e-05, 1.1906e-05]], device='cuda:0') +Epoch 91, bias, value: tensor([ 2.5816e-02, -2.0139e-02, -7.2502e-03, 2.1645e-03, 5.9563e-03, + 2.8759e-03, -1.7688e-02, -6.0538e-05, 2.7556e-02, 1.7025e-02], + device='cuda:0'), grad: tensor([ 2.8461e-05, 6.8963e-05, 1.7428e-04, -1.7128e-03, -1.3532e-06, + 1.0622e-04, -1.6004e-05, -5.2303e-05, 1.3285e-03, 7.4327e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 257.27, cls_loss 0.0045 cls_loss_mapping 0.0112 cls_loss_causal 0.6029 re_mapping 0.0095 re_causal 0.0295 /// teacc 98.84 lr 0.00010000 +Epoch 92, weight, value: tensor([[ 0.0113, -0.0731, -0.0579, ..., -0.0192, -0.1221, -0.0074], + [-0.0607, -0.0989, -0.0925, ..., -0.0785, -0.0325, -0.0103], + [ 0.0529, 0.0507, -0.0830, ..., -0.0785, -0.0903, -0.0492], + ..., + [ 0.0099, 0.0281, -0.0255, ..., -0.0320, -0.0951, 0.0453], + [-0.0007, 0.0788, 0.0678, ..., -0.0537, 0.0449, -0.0866], + [-0.0449, -0.0119, -0.0104, ..., -0.0161, -0.0482, 0.0405]], + device='cuda:0'), grad: tensor([[-3.0875e-04, 4.2245e-06, 3.0613e-04, ..., -4.0799e-05, + 2.5582e-04, -5.1148e-06], + [ 1.1846e-05, 3.2280e-06, 1.8582e-05, ..., 5.3691e-07, + 1.4096e-05, -7.4983e-05], + [ 2.5678e-04, 1.9550e-05, 5.4657e-05, ..., 2.2091e-06, + 4.2409e-05, 5.7161e-05], + ..., + [-5.9545e-05, -1.0759e-04, 1.2529e-04, ..., 9.1270e-08, + 9.8705e-05, 2.6040e-06], + [ 3.0220e-05, -9.9465e-07, 3.0056e-05, ..., 3.8221e-06, + 2.1845e-05, 1.3426e-05], + [ 8.0466e-05, 7.8380e-05, 3.7044e-05, ..., 8.7591e-07, + 2.8268e-05, 4.6819e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0256, -0.0202, -0.0072, 0.0022, 0.0070, 0.0026, -0.0177, -0.0002, + 0.0273, 0.0170], device='cuda:0'), grad: tensor([-4.6492e-04, -4.9353e-05, 6.1893e-04, 4.4250e-04, 1.5572e-05, + -1.4229e-03, 4.3154e-04, 4.5061e-05, 1.2141e-04, 2.6250e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 257.00, cls_loss 0.0045 cls_loss_mapping 0.0102 cls_loss_causal 0.6137 re_mapping 0.0093 re_causal 0.0289 /// teacc 98.81 lr 0.00010000 +Epoch 93, weight, value: tensor([[ 0.0113, -0.0739, -0.0586, ..., -0.0199, -0.1228, -0.0077], + [-0.0616, -0.0992, -0.0928, ..., -0.0787, -0.0328, -0.0101], + [ 0.0539, 0.0508, -0.0837, ..., -0.0792, -0.0908, -0.0501], + ..., + [ 0.0101, 0.0281, -0.0264, ..., -0.0322, -0.0956, 0.0457], + [-0.0005, 0.0790, 0.0679, ..., -0.0544, 0.0451, -0.0874], + [-0.0446, -0.0115, -0.0106, ..., -0.0163, -0.0483, 0.0408]], + device='cuda:0'), grad: tensor([[ 1.4007e-06, 9.2387e-06, 1.0490e-05, ..., 1.3404e-05, + 1.3746e-05, 7.4692e-07], + [ 1.2793e-05, 1.6123e-05, 8.9332e-06, ..., 8.5980e-06, + 1.3441e-05, 3.2075e-06], + [-1.9252e-05, -7.9721e-06, 3.0130e-05, ..., 1.0110e-05, + 5.1618e-05, 1.9088e-05], + ..., + [-2.9191e-05, -5.7280e-05, 1.3657e-05, ..., 6.5891e-07, + 2.5198e-05, -1.0565e-05], + [ 2.1890e-05, 1.6257e-05, 8.2552e-06, ..., 4.4741e-06, + 1.6093e-05, 6.4969e-06], + [ 3.2365e-05, 4.0054e-05, 8.3596e-06, ..., 1.9539e-06, + 1.7479e-05, 1.0639e-05]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0250, -0.0202, -0.0069, 0.0019, 0.0067, 0.0027, -0.0185, 0.0003, + 0.0270, 0.0175], device='cuda:0'), grad: tensor([ 4.8071e-05, 7.9393e-05, -1.7196e-05, -1.4818e-04, 1.6057e-04, + 4.0114e-05, -3.6526e-04, -4.8995e-05, 8.1897e-05, 1.6904e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 92---------------------------------------------------- +epoch 92, time 273.97, cls_loss 0.0064 cls_loss_mapping 0.0119 cls_loss_causal 0.5649 re_mapping 0.0093 re_causal 0.0264 /// teacc 98.99 lr 0.00010000 +Epoch 94, weight, value: tensor([[ 0.0122, -0.0732, -0.0591, ..., -0.0195, -0.1232, -0.0071], + [-0.0621, -0.0994, -0.0930, ..., -0.0782, -0.0331, -0.0104], + [ 0.0545, 0.0511, -0.0841, ..., -0.0806, -0.0911, -0.0506], + ..., + [ 0.0103, 0.0284, -0.0270, ..., -0.0324, -0.0960, 0.0458], + [-0.0018, 0.0788, 0.0683, ..., -0.0550, 0.0455, -0.0885], + [-0.0441, -0.0113, -0.0115, ..., -0.0167, -0.0490, 0.0415]], + device='cuda:0'), grad: tensor([[-2.8148e-05, 3.6787e-06, 1.0431e-06, ..., 2.5216e-07, + 7.7346e-07, 4.1239e-06], + [ 4.2289e-05, 2.9564e-05, 1.3551e-06, ..., 6.7055e-08, + 1.0254e-06, 3.0726e-05], + [-1.0036e-05, -2.0042e-05, 1.9278e-06, ..., 5.4482e-08, + 1.0738e-06, 2.1264e-05], + ..., + [-2.6250e-04, -1.7989e-04, 1.6307e-06, ..., 1.8626e-09, + 1.3346e-06, -2.3746e-04], + [ 2.5734e-05, 1.7896e-05, -6.7838e-06, ..., 4.3260e-07, + -1.7392e-07, 2.2769e-05], + [ 5.7966e-05, 6.5006e-06, -2.5049e-05, ..., 6.4727e-08, + -3.0756e-05, 5.0485e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0252, -0.0203, -0.0072, 0.0020, 0.0060, 0.0027, -0.0190, 0.0004, + 0.0260, 0.0188], device='cuda:0'), grad: tensor([-7.7248e-05, 8.5592e-05, -4.9919e-05, 2.9063e-04, 2.1243e-04, + 4.2289e-05, 1.4670e-05, -5.8174e-04, 1.0294e-04, -4.0084e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 256.99, cls_loss 0.0038 cls_loss_mapping 0.0080 cls_loss_causal 0.5764 re_mapping 0.0097 re_causal 0.0288 /// teacc 98.97 lr 0.00010000 +Epoch 95, weight, value: tensor([[ 0.0144, -0.0725, -0.0596, ..., -0.0198, -0.1238, -0.0067], + [-0.0624, -0.0998, -0.0936, ..., -0.0792, -0.0337, -0.0103], + [ 0.0546, 0.0511, -0.0846, ..., -0.0791, -0.0914, -0.0508], + ..., + [ 0.0105, 0.0286, -0.0274, ..., -0.0324, -0.0965, 0.0460], + [-0.0014, 0.0800, 0.0689, ..., -0.0552, 0.0466, -0.0891], + [-0.0450, -0.0116, -0.0122, ..., -0.0161, -0.0495, 0.0414]], + device='cuda:0'), grad: tensor([[-2.0146e-05, 3.1870e-06, 2.9150e-06, ..., 1.9204e-06, + 2.8405e-06, -6.5677e-06], + [ 4.3996e-06, 4.4703e-06, 3.7830e-06, ..., 2.6822e-07, + 3.4589e-06, -3.7074e-05], + [ 1.5810e-05, 1.2875e-05, 1.0170e-05, ..., 5.0338e-07, + 8.0764e-06, 8.5309e-06], + ..., + [ 8.4788e-06, 6.8732e-06, 5.2899e-06, ..., 1.0012e-08, + 4.1574e-06, 2.0862e-05], + [-6.5148e-05, -7.3195e-05, -5.6654e-05, ..., 1.8021e-06, + -4.2975e-05, 3.9339e-06], + [ 8.1360e-06, 5.9139e-07, 2.7288e-06, ..., 9.1037e-08, + 3.0696e-06, 1.7788e-06]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0261, -0.0204, -0.0070, 0.0019, 0.0064, 0.0029, -0.0197, 0.0004, + 0.0265, 0.0179], device='cuda:0'), grad: tensor([-6.9976e-05, -1.3566e-04, 6.0797e-05, 2.7239e-05, 7.1824e-06, + 8.1062e-05, -1.0971e-06, 9.6738e-05, -9.9063e-05, 3.2872e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 257.19, cls_loss 0.0046 cls_loss_mapping 0.0116 cls_loss_causal 0.5844 re_mapping 0.0093 re_causal 0.0279 /// teacc 98.94 lr 0.00010000 +Epoch 96, weight, value: tensor([[ 0.0143, -0.0731, -0.0609, ..., -0.0203, -0.1257, -0.0071], + [-0.0625, -0.1001, -0.0941, ..., -0.0799, -0.0343, -0.0091], + [ 0.0548, 0.0511, -0.0852, ..., -0.0797, -0.0914, -0.0513], + ..., + [ 0.0107, 0.0290, -0.0277, ..., -0.0325, -0.0967, 0.0459], + [-0.0010, 0.0807, 0.0696, ..., -0.0558, 0.0472, -0.0900], + [-0.0452, -0.0117, -0.0126, ..., -0.0160, -0.0498, 0.0415]], + device='cuda:0'), grad: tensor([[-2.1413e-05, 1.5348e-06, 2.1663e-06, ..., 3.8557e-07, + 1.6065e-06, 3.1888e-06], + [ 2.6338e-06, 2.8759e-06, 3.2131e-06, ..., 1.1176e-07, + 2.4047e-06, 9.1612e-05], + [-9.3728e-06, -5.4762e-06, 2.3320e-06, ..., 1.3178e-07, + 1.8589e-06, -5.4538e-06], + ..., + [ 2.2706e-06, 4.3679e-07, 1.7751e-06, ..., 9.3132e-09, + 1.2433e-06, 2.5436e-05], + [-1.3955e-05, -4.2140e-05, -5.5611e-05, ..., 2.5202e-06, + -4.5955e-05, 1.0423e-05], + [ 1.3299e-05, 4.4480e-06, 2.9635e-06, ..., 8.1956e-08, + 2.2296e-06, 2.7037e-04]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0253, -0.0198, -0.0070, 0.0021, 0.0047, 0.0028, -0.0200, 0.0004, + 0.0267, 0.0191], device='cuda:0'), grad: tensor([-1.4794e-04, 2.2864e-04, -1.4052e-05, 2.6435e-05, -1.0004e-03, + 2.4036e-05, 1.2338e-04, 6.9439e-05, -1.6183e-05, 7.0620e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 256.68, cls_loss 0.0086 cls_loss_mapping 0.0156 cls_loss_causal 0.5567 re_mapping 0.0106 re_causal 0.0284 /// teacc 98.92 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 0.0148, -0.0736, -0.0620, ..., -0.0211, -0.1267, -0.0072], + [-0.0645, -0.1019, -0.0952, ..., -0.0807, -0.0350, -0.0100], + [ 0.0548, 0.0505, -0.0864, ..., -0.0803, -0.0920, -0.0524], + ..., + [ 0.0118, 0.0301, -0.0281, ..., -0.0326, -0.0971, 0.0472], + [-0.0012, 0.0807, 0.0698, ..., -0.0574, 0.0472, -0.0909], + [-0.0450, -0.0112, -0.0122, ..., -0.0163, -0.0487, 0.0417]], + device='cuda:0'), grad: tensor([[ 5.3132e-07, 5.0813e-06, 8.4639e-06, ..., 3.6228e-06, + 6.8992e-06, 1.9744e-07], + [ 5.3551e-07, 1.2601e-06, 1.6494e-06, ..., 3.0873e-07, + 9.2899e-07, -1.4931e-05], + [ 1.5991e-06, 4.2431e-06, 6.0201e-06, ..., 1.8859e-07, + 5.0738e-06, 5.0440e-06], + ..., + [-7.5717e-07, -4.3353e-07, 2.6375e-06, ..., 1.3504e-08, + 3.6675e-06, 2.9970e-06], + [-1.0647e-05, -1.8880e-05, -1.4782e-05, ..., 3.5241e-06, + -1.3173e-05, 1.1623e-06], + [ 2.3749e-06, 7.3165e-06, 7.2680e-06, ..., 6.0629e-07, + 7.9274e-06, 7.2457e-07]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0248, -0.0211, -0.0082, 0.0011, 0.0051, 0.0025, -0.0201, 0.0018, + 0.0263, 0.0203], device='cuda:0'), grad: tensor([ 1.0841e-05, -2.3291e-05, 1.7405e-05, 1.9923e-05, 7.0781e-06, + 2.7996e-06, -2.9474e-05, 1.2688e-05, -3.2961e-05, 1.5058e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 256.92, cls_loss 0.0035 cls_loss_mapping 0.0094 cls_loss_causal 0.5791 re_mapping 0.0101 re_causal 0.0300 /// teacc 98.93 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0154, -0.0739, -0.0635, ..., -0.0213, -0.1279, -0.0073], + [-0.0648, -0.1023, -0.0955, ..., -0.0812, -0.0353, -0.0094], + [ 0.0553, 0.0506, -0.0869, ..., -0.0804, -0.0924, -0.0527], + ..., + [ 0.0118, 0.0302, -0.0285, ..., -0.0326, -0.0974, 0.0470], + [-0.0009, 0.0810, 0.0693, ..., -0.0583, 0.0471, -0.0913], + [-0.0454, -0.0115, -0.0124, ..., -0.0163, -0.0491, 0.0415]], + device='cuda:0'), grad: tensor([[-4.6417e-06, 6.5565e-06, 3.1944e-06, ..., 3.1805e-07, + 3.6024e-06, 3.9563e-06], + [ 7.1339e-06, 9.2089e-06, 5.7071e-06, ..., 3.7346e-07, + 6.4522e-06, -1.7751e-06], + [-5.3108e-05, -3.3379e-05, 5.5432e-06, ..., 2.0908e-07, + 7.9572e-06, 1.3653e-06], + ..., + [ 3.0518e-05, 3.7611e-05, 1.7464e-05, ..., 1.3039e-08, + 1.8194e-05, -1.7151e-05], + [-3.3885e-05, -1.2672e-04, -1.0574e-04, ..., -5.1036e-07, + -1.0061e-04, 1.5264e-06], + [ 2.0564e-05, 1.7792e-05, 1.3813e-05, ..., 5.6811e-08, + 1.3947e-05, 6.9328e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0248, -0.0207, -0.0079, 0.0008, 0.0062, 0.0024, -0.0192, 0.0016, + 0.0259, 0.0192], device='cuda:0'), grad: tensor([-5.9605e-06, 1.9252e-05, -8.2552e-05, 2.2903e-05, 7.1060e-07, + 6.2585e-05, 1.1779e-05, 7.8261e-05, -1.6856e-04, 6.1393e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 256.86, cls_loss 0.0047 cls_loss_mapping 0.0110 cls_loss_causal 0.5596 re_mapping 0.0089 re_causal 0.0259 /// teacc 98.99 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0160, -0.0747, -0.0641, ..., -0.0215, -0.1287, -0.0074], + [-0.0642, -0.1016, -0.0960, ..., -0.0815, -0.0359, -0.0083], + [ 0.0558, 0.0510, -0.0869, ..., -0.0807, -0.0918, -0.0542], + ..., + [ 0.0116, 0.0299, -0.0289, ..., -0.0327, -0.0986, 0.0463], + [-0.0007, 0.0817, 0.0698, ..., -0.0587, 0.0479, -0.0917], + [-0.0452, -0.0111, -0.0128, ..., -0.0163, -0.0492, 0.0421]], + device='cuda:0'), grad: tensor([[ 4.3549e-06, 6.0983e-06, 1.8254e-06, ..., 1.0189e-06, + 1.9521e-06, 1.0036e-05], + [ 4.2468e-06, 5.3160e-06, 2.4941e-06, ..., 1.3318e-07, + 2.6468e-06, 5.1707e-06], + [-5.3272e-06, -2.7884e-06, 1.2303e-06, ..., 1.3364e-07, + 1.9930e-06, 1.3681e-06], + ..., + [ 1.3304e-04, 1.5509e-04, 1.2740e-06, ..., 1.5367e-08, + 2.0396e-06, 3.4547e-04], + [-5.5768e-06, -1.1988e-05, -1.2890e-05, ..., 3.3295e-07, + -9.4846e-06, 1.5385e-06], + [-1.4365e-04, -1.6940e-04, 4.1723e-06, ..., 1.6298e-07, + 3.7123e-06, -3.8719e-04]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0250, -0.0192, -0.0083, 0.0002, 0.0062, 0.0025, -0.0194, 0.0005, + 0.0262, 0.0193], device='cuda:0'), grad: tensor([ 3.1471e-05, 3.6985e-05, 1.7416e-04, 7.0073e-06, 3.8815e-04, + 7.9691e-05, 1.0356e-05, -2.6703e-04, -5.1968e-06, -4.5586e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 256.61, cls_loss 0.0044 cls_loss_mapping 0.0093 cls_loss_causal 0.5607 re_mapping 0.0093 re_causal 0.0274 /// teacc 98.96 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0161, -0.0753, -0.0652, ..., -0.0220, -0.1299, -0.0078], + [-0.0645, -0.1017, -0.0968, ..., -0.0823, -0.0375, -0.0067], + [ 0.0566, 0.0514, -0.0866, ..., -0.0801, -0.0905, -0.0535], + ..., + [ 0.0116, 0.0300, -0.0295, ..., -0.0327, -0.0989, 0.0453], + [-0.0008, 0.0828, 0.0707, ..., -0.0571, 0.0489, -0.0928], + [-0.0451, -0.0110, -0.0132, ..., -0.0165, -0.0494, 0.0420]], + device='cuda:0'), grad: tensor([[-5.7416e-07, 6.7567e-07, 3.0976e-06, ..., 2.9989e-07, + 2.8722e-06, 1.0589e-06], + [ 4.5784e-06, 1.0796e-05, 8.1584e-07, ..., 5.2620e-08, + 8.6147e-07, 2.3544e-06], + [-5.7928e-07, 3.4496e-06, 6.5286e-07, ..., 1.2992e-07, + 7.3807e-07, 2.7809e-06], + ..., + [-2.5019e-05, -6.4790e-05, 7.3994e-07, ..., 1.3970e-09, + 4.6892e-07, -2.7344e-05], + [ 1.2130e-05, 2.9102e-05, 3.4589e-06, ..., 1.1874e-07, + 3.1963e-06, 1.1735e-05], + [ 4.4256e-06, 9.7007e-06, 1.0923e-05, ..., 3.0268e-08, + 1.0289e-05, 8.0392e-06]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0249, -0.0186, -0.0071, 0.0006, 0.0059, 0.0027, -0.0200, -0.0004, + 0.0263, 0.0193], device='cuda:0'), grad: tensor([ 5.8301e-06, -6.7949e-05, 1.3128e-05, 4.0412e-05, 9.4026e-06, + -7.1406e-05, 3.5077e-05, -1.1075e-04, 1.0264e-04, 4.3541e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 256.68, cls_loss 0.0039 cls_loss_mapping 0.0095 cls_loss_causal 0.5947 re_mapping 0.0090 re_causal 0.0282 /// teacc 98.83 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0167, -0.0754, -0.0666, ..., -0.0219, -0.1311, -0.0081], + [-0.0660, -0.1025, -0.0972, ..., -0.0826, -0.0378, -0.0062], + [ 0.0576, 0.0517, -0.0874, ..., -0.0811, -0.0913, -0.0538], + ..., + [ 0.0114, 0.0299, -0.0301, ..., -0.0328, -0.0992, 0.0451], + [-0.0008, 0.0830, 0.0707, ..., -0.0570, 0.0491, -0.0939], + [-0.0452, -0.0109, -0.0138, ..., -0.0166, -0.0498, 0.0423]], + device='cuda:0'), grad: tensor([[-1.3178e-06, 7.0110e-06, 6.7465e-06, ..., 4.9211e-06, + 3.5670e-06, 1.1250e-06], + [ 6.8638e-07, 1.8328e-06, 6.1095e-07, ..., 3.7160e-07, + 9.2527e-07, -1.6401e-06], + [ 1.3607e-06, 3.1125e-06, 1.5460e-06, ..., 3.8603e-07, + 4.1313e-06, 1.8850e-06], + ..., + [-1.5926e-06, -5.3644e-06, 2.3609e-07, ..., 1.9092e-08, + -7.1526e-07, 6.2492e-07], + [ 1.0859e-06, 4.4368e-06, 3.4608e-06, ..., 2.2035e-06, + 2.9430e-06, 1.9111e-06], + [-2.8219e-06, -3.6396e-06, 1.3225e-06, ..., 4.1956e-07, + 1.4193e-06, -5.4091e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0249, -0.0184, -0.0066, 0.0004, 0.0061, 0.0029, -0.0197, -0.0007, + 0.0258, 0.0189], device='cuda:0'), grad: tensor([ 8.7395e-06, -1.5050e-06, 2.4989e-05, -3.3736e-05, 8.3819e-06, + 1.9580e-05, -3.6627e-05, 6.0461e-06, 1.1176e-05, -7.0594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 255.70, cls_loss 0.0036 cls_loss_mapping 0.0083 cls_loss_causal 0.6052 re_mapping 0.0092 re_causal 0.0270 /// teacc 98.86 lr 0.00010000 +Epoch 102, weight, value: tensor([[ 0.0169, -0.0760, -0.0672, ..., -0.0223, -0.1318, -0.0080], + [-0.0669, -0.1029, -0.0977, ..., -0.0831, -0.0382, -0.0061], + [ 0.0578, 0.0512, -0.0880, ..., -0.0816, -0.0919, -0.0546], + ..., + [ 0.0117, 0.0301, -0.0305, ..., -0.0327, -0.0995, 0.0456], + [-0.0006, 0.0832, 0.0710, ..., -0.0575, 0.0493, -0.0945], + [-0.0451, -0.0104, -0.0140, ..., -0.0168, -0.0500, 0.0425]], + device='cuda:0'), grad: tensor([[-9.8944e-06, -2.3916e-06, 3.2373e-06, ..., 1.7723e-06, + 1.9241e-06, -3.9898e-06], + [ 2.1793e-06, 3.6489e-06, 2.7269e-06, ..., 2.4680e-08, + 3.2596e-06, 6.7567e-07], + [ 4.8652e-06, 5.5619e-06, 4.1984e-06, ..., 2.9802e-08, + 5.2564e-06, 1.3905e-06], + ..., + [-1.0289e-05, -8.6278e-06, 9.8441e-07, ..., -3.5390e-08, + 1.3234e-06, -9.4250e-06], + [-3.5353e-06, -7.4357e-06, -1.9912e-06, ..., 2.7334e-07, + -1.8030e-06, 1.5087e-06], + [ 1.3359e-05, 1.1161e-05, 4.4778e-06, ..., 6.2864e-08, + 3.9972e-06, 1.4849e-05]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0250, -0.0186, -0.0070, 0.0002, 0.0063, 0.0026, -0.0193, -0.0004, + 0.0257, 0.0189], device='cuda:0'), grad: tensor([-3.1620e-05, 1.0096e-05, 1.9312e-05, -1.1399e-05, -4.1038e-05, + -2.5421e-05, 1.5661e-05, -1.8001e-05, 3.6415e-06, 7.8738e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 256.73, cls_loss 0.0056 cls_loss_mapping 0.0105 cls_loss_causal 0.5695 re_mapping 0.0091 re_causal 0.0260 /// teacc 98.94 lr 0.00010000 +Epoch 103, weight, value: tensor([[ 1.7405e-02, -7.6313e-02, -6.7860e-02, ..., -2.2697e-02, + -1.3247e-01, -7.9666e-03], + [-6.8657e-02, -1.0403e-01, -9.8322e-02, ..., -8.4685e-02, + -3.8752e-02, -6.2386e-03], + [ 5.8280e-02, 5.1710e-02, -8.8969e-02, ..., -8.2148e-02, + -9.2701e-02, -5.5193e-02], + ..., + [ 1.0156e-02, 2.8522e-02, -3.1229e-02, ..., -3.2869e-02, + -1.0060e-01, 4.5693e-02], + [ 9.1587e-06, 8.3869e-02, 7.1720e-02, ..., -5.8579e-02, + 5.0215e-02, -9.5644e-02], + [-4.2508e-02, -8.0927e-03, -1.4605e-02, ..., -1.7101e-02, + -5.0279e-02, 4.3528e-02]], device='cuda:0'), grad: tensor([[-6.1728e-06, 1.3663e-06, 1.5479e-06, ..., 5.5134e-07, + 1.5553e-06, 5.3551e-08], + [ 5.6624e-07, 4.8662e-07, 5.6252e-07, ..., 7.2177e-08, + 8.0513e-07, -1.3551e-06], + [ 1.9930e-06, 2.5686e-06, 2.5164e-06, ..., 2.2165e-07, + 3.3956e-06, 9.9652e-08], + ..., + [ 5.7975e-07, 3.9907e-07, 3.9954e-07, ..., 5.1223e-09, + 6.3051e-07, 3.8557e-07], + [-4.1202e-06, -6.9328e-06, -4.1015e-06, ..., 7.5251e-07, + -5.7220e-06, 2.1048e-07], + [ 2.4587e-06, 1.1958e-06, 2.6878e-06, ..., 2.4214e-08, + 3.2578e-06, -6.7009e-07]], device='cuda:0') +Epoch 103, bias, value: tensor([ 2.5178e-02, -1.9103e-02, -6.9718e-03, 4.0717e-05, 6.0884e-03, + 1.9722e-03, -1.8974e-02, -1.1598e-03, 2.5889e-02, 2.0023e-02], + device='cuda:0'), grad: tensor([-1.1913e-05, 1.1679e-06, 8.9183e-06, -2.5481e-06, -4.6760e-05, + 6.6217e-07, 3.8184e-06, 1.1869e-05, -5.4836e-06, 4.0174e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 256.70, cls_loss 0.0035 cls_loss_mapping 0.0079 cls_loss_causal 0.5910 re_mapping 0.0089 re_causal 0.0272 /// teacc 98.92 lr 0.00010000 +Epoch 104, weight, value: tensor([[ 0.0177, -0.0773, -0.0692, ..., -0.0253, -0.1336, -0.0085], + [-0.0695, -0.1046, -0.0989, ..., -0.0857, -0.0392, -0.0061], + [ 0.0585, 0.0518, -0.0895, ..., -0.0824, -0.0932, -0.0555], + ..., + [ 0.0105, 0.0287, -0.0316, ..., -0.0330, -0.1009, 0.0461], + [ 0.0002, 0.0841, 0.0716, ..., -0.0594, 0.0501, -0.0962], + [-0.0424, -0.0081, -0.0153, ..., -0.0172, -0.0509, 0.0434]], + device='cuda:0'), grad: tensor([[-1.5218e-06, 3.3621e-06, 2.4214e-06, ..., 9.0525e-07, + 2.1514e-06, 1.7826e-06], + [ 1.3579e-06, 2.1439e-06, 1.7136e-06, ..., 3.1106e-07, + 2.2799e-06, -2.7108e-04], + [ 4.1500e-06, 6.6943e-06, 4.6007e-06, ..., 3.7299e-07, + 5.4650e-06, 4.3064e-06], + ..., + [ 3.6508e-06, 5.4985e-06, 3.8743e-06, ..., 1.6764e-08, + 4.9099e-06, 2.3234e-04], + [-3.1114e-05, -4.6581e-05, -2.0593e-05, ..., 9.9558e-07, + -2.6599e-05, -9.2238e-06], + [ 1.8552e-05, 2.3693e-05, 1.4335e-05, ..., 1.8487e-07, + 1.8373e-05, 1.9178e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0248, -0.0192, -0.0071, -0.0003, 0.0061, 0.0028, -0.0185, -0.0009, + 0.0255, 0.0198], device='cuda:0'), grad: tensor([-1.7527e-06, -4.6420e-04, 1.6883e-05, -1.1176e-05, 3.6061e-05, + 9.5442e-06, -5.8673e-06, 4.1080e-04, -4.4316e-05, 5.3585e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 256.78, cls_loss 0.0045 cls_loss_mapping 0.0084 cls_loss_causal 0.5878 re_mapping 0.0090 re_causal 0.0270 /// teacc 98.92 lr 0.00010000 +Epoch 105, weight, value: tensor([[ 0.0190, -0.0777, -0.0698, ..., -0.0269, -0.1333, -0.0088], + [-0.0701, -0.1067, -0.0998, ..., -0.0884, -0.0402, -0.0049], + [ 0.0586, 0.0519, -0.0902, ..., -0.0821, -0.0938, -0.0562], + ..., + [ 0.0107, 0.0296, -0.0319, ..., -0.0331, -0.1014, 0.0458], + [ 0.0004, 0.0845, 0.0721, ..., -0.0599, 0.0507, -0.0967], + [-0.0428, -0.0082, -0.0159, ..., -0.0182, -0.0513, 0.0433]], + device='cuda:0'), grad: tensor([[-4.5866e-05, -3.3021e-05, -1.3784e-07, ..., 1.9092e-07, + -6.0759e-06, -2.3395e-05], + [ 3.4273e-07, 2.8545e-07, 9.2667e-08, ..., -2.1420e-06, + 3.1665e-07, -7.7710e-06], + [ 1.3802e-06, 9.6764e-07, 7.2177e-08, ..., 1.5507e-07, + 4.6333e-07, 1.6121e-06], + ..., + [-5.6438e-07, -5.9325e-07, 5.7742e-08, ..., 3.2736e-07, + -3.5018e-07, 3.1348e-06], + [ 2.3656e-06, 1.8319e-06, 1.8161e-07, ..., 1.4296e-07, + 6.8359e-07, 1.9725e-06], + [ 3.6687e-05, 2.6166e-05, 7.8697e-08, ..., 9.1735e-08, + 4.9695e-06, 2.3425e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0253, -0.0193, -0.0073, -0.0006, 0.0051, 0.0029, -0.0176, -0.0004, + 0.0256, 0.0198], device='cuda:0'), grad: tensor([-1.5068e-04, -2.1309e-05, 7.6592e-06, 8.5384e-06, -4.6879e-05, + 5.0999e-06, 1.1161e-05, 1.1846e-05, 1.0766e-05, 1.6379e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 104---------------------------------------------------- +epoch 104, time 273.50, cls_loss 0.0037 cls_loss_mapping 0.0098 cls_loss_causal 0.5752 re_mapping 0.0088 re_causal 0.0268 /// teacc 99.01 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 0.0195, -0.0782, -0.0697, ..., -0.0270, -0.1333, -0.0088], + [-0.0708, -0.1078, -0.1009, ..., -0.0895, -0.0416, -0.0046], + [ 0.0587, 0.0511, -0.0916, ..., -0.0839, -0.0947, -0.0568], + ..., + [ 0.0112, 0.0303, -0.0324, ..., -0.0329, -0.1013, 0.0459], + [ 0.0003, 0.0850, 0.0725, ..., -0.0602, 0.0515, -0.0972], + [-0.0432, -0.0088, -0.0164, ..., -0.0186, -0.0516, 0.0429]], + device='cuda:0'), grad: tensor([[ 9.6709e-06, 8.7470e-06, 2.9337e-06, ..., 8.7544e-06, + 1.8487e-07, 1.1653e-05], + [ 3.8333e-06, 2.3693e-06, 1.9018e-06, ..., 1.6261e-06, + 1.5404e-06, 4.8196e-07], + [-1.2898e-04, -6.1274e-05, 9.1130e-07, ..., 1.5413e-07, + 5.9418e-07, -1.2629e-05], + ..., + [ 1.2137e-05, 3.9637e-06, -6.6832e-06, ..., -3.3051e-05, + 6.4634e-07, -4.3601e-05], + [ 1.1988e-05, 6.4634e-06, 1.0349e-05, ..., 4.2096e-06, + 9.1642e-06, 2.6822e-06], + [ 2.2590e-05, 8.3297e-06, 5.7332e-06, ..., 1.0177e-05, + 2.2911e-06, 1.5661e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0255, -0.0197, -0.0080, -0.0008, 0.0057, 0.0027, -0.0171, 0.0002, + 0.0258, 0.0191], device='cuda:0'), grad: tensor([ 3.0667e-05, 8.1509e-06, -2.2018e-04, 1.0735e-04, 1.5721e-06, + -7.8976e-06, 1.5840e-05, -5.1826e-05, 4.4167e-05, 7.2122e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 105---------------------------------------------------- +epoch 105, time 273.16, cls_loss 0.0034 cls_loss_mapping 0.0088 cls_loss_causal 0.5901 re_mapping 0.0081 re_causal 0.0265 /// teacc 99.02 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0200, -0.0789, -0.0702, ..., -0.0277, -0.1334, -0.0090], + [-0.0713, -0.1083, -0.1013, ..., -0.0900, -0.0419, -0.0043], + [ 0.0596, 0.0518, -0.0926, ..., -0.0837, -0.0953, -0.0562], + ..., + [ 0.0110, 0.0303, -0.0329, ..., -0.0344, -0.1014, 0.0456], + [-0.0002, 0.0847, 0.0721, ..., -0.0609, 0.0509, -0.0980], + [-0.0432, -0.0088, -0.0159, ..., -0.0189, -0.0510, 0.0433]], + device='cuda:0'), grad: tensor([[-1.5497e-05, 4.7907e-06, 5.9307e-06, ..., 6.1989e-06, + 3.7365e-06, -1.9027e-06], + [ 2.2482e-06, 1.1735e-06, 2.0675e-06, ..., 9.7789e-07, + 1.6270e-06, -1.3085e-06], + [-2.1324e-05, -3.1712e-07, 1.2144e-06, ..., 7.9954e-07, + -4.7423e-06, 2.6636e-07], + ..., + [ 1.5616e-05, -1.7779e-06, 1.5451e-06, ..., 1.2433e-06, + 5.0440e-06, -8.9593e-07], + [ 4.5821e-06, 2.6956e-05, 3.5435e-05, ..., 3.2961e-05, + 2.8044e-05, 7.8883e-07], + [ 3.9972e-06, 1.0282e-06, 4.5523e-06, ..., 1.6820e-06, + 3.7123e-06, 1.2107e-07]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0256, -0.0195, -0.0077, 0.0003, 0.0060, 0.0023, -0.0171, -0.0001, + 0.0245, 0.0191], device='cuda:0'), grad: tensor([-2.4259e-05, 4.6790e-06, -4.7415e-05, 7.5102e-05, -1.8090e-05, + -6.4075e-05, -5.9336e-05, 5.1439e-05, 6.1572e-05, 2.0310e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 257.01, cls_loss 0.0028 cls_loss_mapping 0.0072 cls_loss_causal 0.5731 re_mapping 0.0083 re_causal 0.0263 /// teacc 99.02 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0205, -0.0792, -0.0706, ..., -0.0278, -0.1339, -0.0089], + [-0.0716, -0.1086, -0.1016, ..., -0.0906, -0.0423, -0.0041], + [ 0.0596, 0.0516, -0.0932, ..., -0.0839, -0.0962, -0.0567], + ..., + [ 0.0112, 0.0307, -0.0334, ..., -0.0342, -0.1017, 0.0457], + [-0.0002, 0.0850, 0.0722, ..., -0.0611, 0.0511, -0.0985], + [-0.0434, -0.0090, -0.0165, ..., -0.0190, -0.0515, 0.0432]], + device='cuda:0'), grad: tensor([[-1.1601e-05, 7.0296e-06, 7.9572e-06, ..., 3.9004e-06, + 5.8487e-06, 1.2992e-06], + [ 4.3884e-06, 5.9642e-06, 2.8219e-06, ..., 6.4680e-07, + 3.1609e-06, 4.1686e-06], + [ 5.2229e-06, 8.6352e-06, 5.3458e-06, ..., 2.4885e-06, + 4.2766e-06, 3.3490e-06], + ..., + [-6.6698e-05, -7.4804e-05, 6.0303e-07, ..., -1.0021e-06, + 6.8825e-07, -6.7830e-05], + [-3.5673e-05, -2.7359e-05, -7.1228e-06, ..., 2.4498e-05, + -1.4782e-05, 9.2527e-07], + [ 1.0365e-04, 1.3018e-04, 5.5969e-05, ..., 2.8256e-06, + 4.9055e-05, 5.2899e-05]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0258, -0.0192, -0.0080, 0.0005, 0.0065, 0.0025, -0.0173, -0.0003, + 0.0244, 0.0187], device='cuda:0'), grad: tensor([-3.3051e-05, 1.6749e-05, 1.9372e-05, -3.1348e-06, -9.6187e-06, + 6.5506e-05, -1.3018e-04, -1.4198e-04, -6.7472e-05, 2.8372e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 257.19, cls_loss 0.0053 cls_loss_mapping 0.0116 cls_loss_causal 0.5863 re_mapping 0.0083 re_causal 0.0246 /// teacc 98.93 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0211, -0.0779, -0.0709, ..., -0.0273, -0.1344, -0.0089], + [-0.0710, -0.1072, -0.1022, ..., -0.0918, -0.0426, -0.0032], + [ 0.0600, 0.0516, -0.0938, ..., -0.0843, -0.0969, -0.0570], + ..., + [ 0.0113, 0.0298, -0.0338, ..., -0.0336, -0.1022, 0.0444], + [-0.0009, 0.0843, 0.0724, ..., -0.0614, 0.0510, -0.1004], + [-0.0431, -0.0080, -0.0170, ..., -0.0195, -0.0518, 0.0451]], + device='cuda:0'), grad: tensor([[ 9.3458e-07, 2.0694e-06, 1.4678e-06, ..., 1.7118e-06, + 1.2554e-06, 1.6252e-07], + [ 5.7183e-06, 7.6741e-06, 1.4901e-06, ..., 1.6466e-06, + 1.4966e-06, 3.6024e-06], + [-1.8859e-06, 9.5461e-07, 2.3413e-06, ..., 1.6186e-06, + 2.0247e-06, 3.6694e-07], + ..., + [-7.4431e-06, -9.6112e-06, 9.7230e-07, ..., 1.6680e-06, + 5.9418e-07, -7.2531e-06], + [-8.4415e-06, -1.0036e-05, -1.2800e-05, ..., 8.8941e-07, + -1.7270e-05, 3.2224e-07], + [ 2.9821e-06, 9.1419e-06, 1.4566e-05, ..., 8.4415e-06, + 6.6124e-06, -1.2591e-06]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0265, -0.0181, -0.0078, 0.0003, 0.0065, 0.0025, -0.0178, -0.0019, + 0.0226, 0.0200], device='cuda:0'), grad: tensor([ 1.0751e-05, 2.7359e-05, 6.8396e-06, 1.6689e-05, -2.6894e-04, + 1.2182e-05, 1.2505e-04, -1.6227e-05, -8.9034e-06, 9.5189e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 256.96, cls_loss 0.0054 cls_loss_mapping 0.0108 cls_loss_causal 0.5704 re_mapping 0.0086 re_causal 0.0242 /// teacc 98.95 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 0.0232, -0.0784, -0.0715, ..., -0.0268, -0.1349, -0.0090], + [-0.0723, -0.1080, -0.1031, ..., -0.0923, -0.0435, -0.0032], + [ 0.0596, 0.0522, -0.0947, ..., -0.0846, -0.0973, -0.0568], + ..., + [ 0.0117, 0.0303, -0.0344, ..., -0.0333, -0.1026, 0.0444], + [-0.0003, 0.0851, 0.0731, ..., -0.0616, 0.0520, -0.1007], + [-0.0434, -0.0080, -0.0183, ..., -0.0212, -0.0533, 0.0453]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 8.5235e-06, 9.0748e-06, ..., 4.7870e-07, + 8.4415e-06, -1.0133e-06], + [ 6.1952e-06, 7.4320e-06, 7.5214e-06, ..., 4.7032e-08, + 8.4788e-06, -6.4913e-07], + [ 2.7046e-05, 3.4004e-05, 3.4034e-05, ..., 7.9162e-08, + 3.1978e-05, 2.0470e-06], + ..., + [-9.0823e-06, -8.6427e-06, 6.6869e-06, ..., 2.7940e-09, + 8.5011e-06, -1.4223e-05], + [-8.8811e-05, -1.1885e-04, -1.3411e-04, ..., 3.5530e-07, + -1.2827e-04, 1.2498e-06], + [ 1.3769e-05, 1.1198e-05, 1.0908e-05, ..., 4.8429e-08, + 1.2711e-05, 5.0217e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0248, -0.0191, -0.0057, -0.0003, 0.0069, 0.0032, -0.0197, -0.0018, + 0.0231, 0.0212], device='cuda:0'), grad: tensor([-2.0579e-05, 1.2346e-05, 8.1837e-05, 2.5406e-05, -2.1684e-04, + 7.5758e-05, 2.7150e-05, -2.4483e-05, -2.5010e-04, 2.8944e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 256.77, cls_loss 0.0035 cls_loss_mapping 0.0094 cls_loss_causal 0.5697 re_mapping 0.0083 re_causal 0.0254 /// teacc 98.97 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0240, -0.0788, -0.0723, ..., -0.0270, -0.1357, -0.0089], + [-0.0733, -0.1082, -0.1039, ..., -0.0931, -0.0438, -0.0036], + [ 0.0602, 0.0523, -0.0958, ..., -0.0849, -0.0977, -0.0573], + ..., + [ 0.0120, 0.0310, -0.0348, ..., -0.0332, -0.1031, 0.0454], + [-0.0002, 0.0850, 0.0734, ..., -0.0622, 0.0523, -0.1017], + [-0.0438, -0.0086, -0.0188, ..., -0.0215, -0.0538, 0.0444]], + device='cuda:0'), grad: tensor([[-1.2191e-06, 1.2983e-06, 7.2271e-07, ..., 1.9837e-07, + 6.7661e-07, -1.2945e-07], + [ 3.1060e-07, 8.1118e-07, 2.8312e-07, ..., 7.3574e-08, + 2.8312e-07, -2.5649e-06], + [ 7.9907e-07, 1.1325e-06, 1.1632e-06, ..., 6.3796e-08, + 1.1362e-06, 4.2841e-07], + ..., + [-1.5460e-07, -1.7649e-07, 1.1083e-07, ..., 1.8626e-09, + 1.1595e-07, 6.3982e-07], + [-3.7085e-06, -3.4682e-06, -4.5151e-06, ..., 4.3586e-07, + -4.3884e-06, 6.1514e-07], + [ 1.3197e-06, -1.5823e-06, 8.7963e-07, ..., 1.4435e-08, + 8.5449e-07, -8.5495e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0250, -0.0193, -0.0058, -0.0006, 0.0077, 0.0029, -0.0191, -0.0009, + 0.0226, 0.0201], device='cuda:0'), grad: tensor([-2.4959e-06, -4.9099e-06, 4.3772e-06, 2.9281e-06, -1.3947e-05, + 3.2969e-06, 1.0729e-05, 3.4086e-06, -8.5980e-06, 5.1968e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 255.11, cls_loss 0.0028 cls_loss_mapping 0.0067 cls_loss_causal 0.5605 re_mapping 0.0078 re_causal 0.0245 /// teacc 98.98 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0245, -0.0792, -0.0732, ..., -0.0265, -0.1366, -0.0092], + [-0.0736, -0.1086, -0.1044, ..., -0.0944, -0.0445, -0.0035], + [ 0.0604, 0.0523, -0.0961, ..., -0.0850, -0.0982, -0.0576], + ..., + [ 0.0123, 0.0314, -0.0351, ..., -0.0330, -0.1032, 0.0456], + [-0.0003, 0.0853, 0.0736, ..., -0.0625, 0.0523, -0.1021], + [-0.0441, -0.0087, -0.0192, ..., -0.0217, -0.0542, 0.0442]], + device='cuda:0'), grad: tensor([[ 1.5097e-06, 3.1255e-06, 4.4219e-06, ..., 1.4389e-06, + 4.4666e-06, 2.3935e-07], + [ 3.0864e-06, 4.9435e-06, 3.4086e-06, ..., 2.8405e-07, + 3.5334e-06, -1.0999e-06], + [-9.7789e-07, 1.3420e-06, 2.1383e-06, ..., 3.7719e-07, + 3.4850e-06, 1.1502e-06], + ..., + [-1.0990e-05, -2.2769e-05, 1.8254e-06, ..., 1.4435e-08, + 1.8422e-06, -9.6858e-06], + [-1.3039e-07, -1.0384e-06, 2.1592e-05, ..., 6.8545e-07, + 1.8314e-05, 1.0366e-06], + [ 5.3532e-06, 7.0296e-06, 3.0044e-06, ..., 1.0198e-07, + 2.8703e-06, 3.6675e-06]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0253, -0.0194, -0.0059, -0.0002, 0.0076, 0.0033, -0.0193, -0.0007, + 0.0222, 0.0199], device='cuda:0'), grad: tensor([ 1.0759e-05, 5.9903e-06, 2.3562e-06, 2.8923e-05, 3.6061e-06, + -2.6155e-04, 1.9228e-04, -2.8670e-05, 3.3051e-05, 1.3657e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 111---------------------------------------------------- +epoch 111, time 273.89, cls_loss 0.0029 cls_loss_mapping 0.0077 cls_loss_causal 0.5681 re_mapping 0.0080 re_causal 0.0244 /// teacc 99.08 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 0.0253, -0.0790, -0.0737, ..., -0.0269, -0.1370, -0.0092], + [-0.0739, -0.1084, -0.1053, ..., -0.0949, -0.0454, -0.0032], + [ 0.0603, 0.0520, -0.0972, ..., -0.0851, -0.0994, -0.0580], + ..., + [ 0.0125, 0.0315, -0.0355, ..., -0.0330, -0.1034, 0.0455], + [ 0.0002, 0.0861, 0.0744, ..., -0.0628, 0.0533, -0.1024], + [-0.0443, -0.0087, -0.0188, ..., -0.0215, -0.0540, 0.0441]], + device='cuda:0'), grad: tensor([[-5.1744e-06, 2.6729e-06, 1.9148e-06, ..., 1.4920e-06, + 1.7341e-06, 8.4471e-07], + [ 2.8666e-06, 4.3027e-06, 1.1884e-06, ..., 8.4564e-07, + 1.2415e-06, -7.2597e-07], + [-3.6247e-06, -1.5432e-06, 1.1381e-06, ..., 7.0082e-07, + 1.1399e-06, 1.1874e-06], + ..., + [ 1.9446e-06, 1.7062e-06, 1.9697e-07, ..., 4.7497e-08, + 2.3702e-07, 1.9837e-06], + [-3.0641e-06, -3.8780e-06, -3.5856e-06, ..., 3.8408e-06, + -2.4829e-06, 7.4413e-07], + [-3.7868e-06, -1.2405e-05, 6.5863e-06, ..., 1.8300e-07, + 5.5507e-06, -1.8224e-05]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0259, -0.0190, -0.0066, 0.0002, 0.0076, 0.0026, -0.0194, -0.0008, + 0.0227, 0.0196], device='cuda:0'), grad: tensor([-1.0170e-05, 9.3952e-06, -2.1402e-06, 9.8497e-06, 4.5836e-05, + 2.2203e-05, -2.0653e-05, 2.0877e-05, 7.6462e-07, -7.5996e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 256.48, cls_loss 0.0034 cls_loss_mapping 0.0073 cls_loss_causal 0.5579 re_mapping 0.0082 re_causal 0.0245 /// teacc 98.97 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0267, -0.0787, -0.0744, ..., -0.0265, -0.1378, -0.0090], + [-0.0745, -0.1092, -0.1061, ..., -0.0958, -0.0462, -0.0031], + [ 0.0608, 0.0518, -0.0983, ..., -0.0859, -0.0998, -0.0582], + ..., + [ 0.0126, 0.0317, -0.0361, ..., -0.0329, -0.1040, 0.0457], + [ 0.0003, 0.0865, 0.0746, ..., -0.0633, 0.0536, -0.1029], + [-0.0441, -0.0084, -0.0194, ..., -0.0216, -0.0546, 0.0442]], + device='cuda:0'), grad: tensor([[ 1.5413e-07, 3.0436e-06, 2.2370e-06, ..., 1.4259e-06, + 1.9390e-06, 1.0580e-06], + [ 4.6333e-07, 5.5246e-06, 1.2666e-06, ..., 8.4750e-07, + 1.3672e-06, 1.8617e-06], + [-1.9912e-06, 4.0652e-07, 1.4268e-06, ..., 1.0254e-06, + 1.8189e-06, 3.1432e-07], + ..., + [ 4.0047e-08, 1.7835e-06, 1.1865e-06, ..., 9.6392e-08, + 1.3309e-06, 1.7276e-06], + [ 1.1669e-06, 3.3583e-06, 4.4443e-06, ..., 1.2796e-06, + 4.3549e-06, 1.7593e-06], + [ 2.8498e-07, -5.2094e-05, 1.4184e-06, ..., -4.7088e-06, + 1.4585e-06, -2.9936e-05]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0267, -0.0193, -0.0066, 0.0005, 0.0072, 0.0026, -0.0192, -0.0008, + 0.0225, 0.0194], device='cuda:0'), grad: tensor([ 6.1132e-06, 1.2338e-05, 8.7991e-06, 7.0520e-06, 1.1700e-04, + -8.4490e-06, -2.2978e-05, 9.9540e-06, 1.1787e-05, -1.4162e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 256.84, cls_loss 0.0031 cls_loss_mapping 0.0083 cls_loss_causal 0.5639 re_mapping 0.0081 re_causal 0.0242 /// teacc 99.06 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 0.0270, -0.0790, -0.0753, ..., -0.0268, -0.1389, -0.0088], + [-0.0753, -0.1103, -0.1072, ..., -0.0960, -0.0475, -0.0031], + [ 0.0623, 0.0522, -0.0990, ..., -0.0863, -0.0998, -0.0587], + ..., + [ 0.0130, 0.0324, -0.0364, ..., -0.0319, -0.1043, 0.0459], + [ 0.0004, 0.0874, 0.0752, ..., -0.0639, 0.0545, -0.1031], + [-0.0443, -0.0087, -0.0199, ..., -0.0221, -0.0547, 0.0441]], + device='cuda:0'), grad: tensor([[-2.2613e-06, 5.0804e-07, 1.8412e-06, ..., 6.7661e-07, + 1.7146e-06, -2.4531e-06], + [ 5.4622e-07, 8.9081e-07, 8.0001e-07, ..., 1.1455e-07, + 6.8918e-07, -6.3982e-07], + [ 9.3132e-08, 3.7905e-07, 5.6531e-07, ..., 1.2852e-07, + 5.2107e-07, 5.2201e-07], + ..., + [-1.4501e-06, -3.4161e-06, 8.2003e-07, ..., 1.6810e-07, + 7.2224e-07, -1.7732e-06], + [ 6.5267e-06, 2.6226e-06, 2.6673e-05, ..., 1.0751e-05, + 2.8729e-05, 2.4680e-07], + [ 1.8496e-06, 1.6764e-06, 1.9241e-06, ..., 4.5681e-07, + 1.7472e-06, 2.3879e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0269, -0.0196, -0.0060, -0.0002, 0.0071, 0.0026, -0.0195, -0.0002, + 0.0230, 0.0189], device='cuda:0'), grad: tensor([-2.5406e-05, 2.4959e-07, 2.2911e-06, 4.4852e-05, -1.0002e-06, + -1.0097e-04, 2.4125e-05, -1.5656e-06, 3.7283e-05, 1.9982e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 257.04, cls_loss 0.0032 cls_loss_mapping 0.0070 cls_loss_causal 0.5699 re_mapping 0.0079 re_causal 0.0250 /// teacc 98.99 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0277, -0.0787, -0.0762, ..., -0.0265, -0.1399, -0.0090], + [-0.0760, -0.1111, -0.1090, ..., -0.0962, -0.0496, -0.0028], + [ 0.0625, 0.0516, -0.0999, ..., -0.0872, -0.1005, -0.0591], + ..., + [ 0.0135, 0.0330, -0.0367, ..., -0.0312, -0.1048, 0.0461], + [ 0.0002, 0.0876, 0.0751, ..., -0.0649, 0.0551, -0.1036], + [-0.0448, -0.0090, -0.0204, ..., -0.0224, -0.0552, 0.0440]], + device='cuda:0'), grad: tensor([[-3.9786e-05, 1.7583e-06, -6.7987e-06, ..., 1.0524e-06, + 1.5870e-06, -5.5693e-06], + [ 2.1253e-06, 1.4678e-06, 1.1828e-06, ..., 3.0268e-07, + 1.3309e-06, 4.2026e-07], + [-1.8077e-06, -1.6401e-06, 1.5441e-06, ..., 2.7684e-07, + 1.0626e-06, 3.7253e-07], + ..., + [ 2.2110e-06, 1.0282e-06, 3.1572e-07, ..., -2.2305e-07, + 1.0859e-06, 1.5623e-07], + [ 2.2929e-06, -1.7853e-06, -1.8533e-07, ..., 1.1390e-06, + -2.6673e-06, 5.7137e-07], + [ 2.5362e-05, 4.0256e-07, 6.5342e-06, ..., 1.5623e-07, + 1.1465e-06, 3.2410e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([ 2.7209e-02, -1.9717e-02, -6.2022e-03, -5.1442e-04, 7.5631e-03, + 1.9853e-03, -1.8267e-02, 7.6748e-05, 2.2704e-02, 1.8295e-02], + device='cuda:0'), grad: tensor([-1.1653e-04, 8.9854e-06, 6.8918e-07, -3.3192e-06, -6.4224e-06, + 1.6972e-05, -4.1886e-07, 1.0379e-05, 8.8289e-06, 8.0764e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 256.70, cls_loss 0.0032 cls_loss_mapping 0.0086 cls_loss_causal 0.5699 re_mapping 0.0082 re_causal 0.0250 /// teacc 98.97 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 0.0276, -0.0796, -0.0773, ..., -0.0270, -0.1417, -0.0088], + [-0.0766, -0.1112, -0.1098, ..., -0.0965, -0.0502, -0.0027], + [ 0.0624, 0.0512, -0.1013, ..., -0.0878, -0.1020, -0.0597], + ..., + [ 0.0130, 0.0329, -0.0388, ..., -0.0310, -0.1059, 0.0460], + [ 0.0011, 0.0882, 0.0752, ..., -0.0653, 0.0553, -0.1041], + [-0.0448, -0.0086, -0.0205, ..., -0.0227, -0.0551, 0.0443]], + device='cuda:0'), grad: tensor([[ 6.4122e-07, 6.0629e-07, 4.5425e-07, ..., 2.1420e-07, + 3.7882e-07, 1.1735e-07], + [ 4.3120e-07, 6.1700e-07, 2.8429e-07, ..., 1.8394e-08, + 3.0850e-07, -7.7998e-07], + [-1.3458e-06, -5.2946e-07, 4.8010e-07, ..., 2.0489e-08, + 6.0257e-07, 5.6205e-07], + ..., + [-7.6462e-07, -1.5926e-06, 5.6764e-07, ..., -1.1642e-09, + 7.2783e-07, -1.3607e-06], + [ 3.6950e-07, 4.4936e-08, 1.1791e-06, ..., 1.3015e-07, + 9.2946e-07, 1.8044e-07], + [ 8.6520e-07, 6.4960e-07, 6.5938e-07, ..., 1.3271e-08, + 7.9582e-07, 6.1002e-07]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0268, -0.0197, -0.0067, 0.0008, 0.0077, 0.0021, -0.0187, -0.0004, + 0.0229, 0.0187], device='cuda:0'), grad: tensor([ 2.1700e-06, -1.3579e-06, -1.1157e-06, -3.1050e-06, 2.6613e-07, + -1.5767e-06, -5.2620e-08, -4.6939e-07, 2.5351e-06, 2.6990e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 257.28, cls_loss 0.0029 cls_loss_mapping 0.0064 cls_loss_causal 0.5882 re_mapping 0.0080 re_causal 0.0249 /// teacc 98.96 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 0.0278, -0.0799, -0.0782, ..., -0.0273, -0.1424, -0.0089], + [-0.0771, -0.1117, -0.1100, ..., -0.0966, -0.0503, -0.0032], + [ 0.0626, 0.0509, -0.1018, ..., -0.0880, -0.1024, -0.0603], + ..., + [ 0.0134, 0.0335, -0.0393, ..., -0.0307, -0.1061, 0.0466], + [ 0.0019, 0.0890, 0.0763, ..., -0.0658, 0.0563, -0.1045], + [-0.0449, -0.0087, -0.0210, ..., -0.0227, -0.0556, 0.0451]], + device='cuda:0'), grad: tensor([[-1.1779e-05, 1.0384e-06, 2.1756e-05, ..., 3.4925e-07, + 1.6272e-05, 3.8650e-07], + [ 6.6683e-07, 5.3551e-07, 2.0027e-05, ..., 6.2864e-08, + 1.4983e-05, -2.4140e-06], + [-3.2559e-06, -3.0361e-06, 1.4558e-05, ..., 2.1537e-07, + 1.0908e-05, 4.1770e-07], + ..., + [ 1.6168e-06, 1.2014e-06, 2.4140e-05, ..., 8.4052e-08, + 1.8045e-05, 1.3392e-06], + [ 2.4475e-06, 2.4531e-06, 1.5628e-04, ..., 4.5006e-07, + 1.1635e-04, 9.6485e-07], + [ 7.0855e-06, -3.6769e-06, 2.6464e-05, ..., -2.0047e-07, + 1.9699e-05, -3.3099e-06]], device='cuda:0') +Epoch 118, bias, value: tensor([ 2.6891e-02, -2.0213e-02, -6.7734e-03, 8.9989e-05, 7.5903e-03, + 1.3307e-03, -1.8470e-02, 5.3537e-04, 2.3243e-02, 1.8801e-02], + device='cuda:0'), grad: tensor([-1.0617e-06, 2.5317e-05, 1.6421e-05, 1.2236e-03, 1.0014e-05, + -1.7233e-03, 1.1778e-04, 4.3124e-05, 2.3699e-04, 5.2124e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 256.93, cls_loss 0.0031 cls_loss_mapping 0.0067 cls_loss_causal 0.5500 re_mapping 0.0078 re_causal 0.0241 /// teacc 98.91 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0269, -0.0821, -0.0780, ..., -0.0279, -0.1430, -0.0090], + [-0.0776, -0.1122, -0.1103, ..., -0.0968, -0.0505, -0.0032], + [ 0.0628, 0.0509, -0.1022, ..., -0.0881, -0.1029, -0.0609], + ..., + [ 0.0135, 0.0339, -0.0398, ..., -0.0301, -0.1068, 0.0468], + [ 0.0021, 0.0889, 0.0758, ..., -0.0672, 0.0559, -0.1050], + [-0.0442, -0.0076, -0.0217, ..., -0.0231, -0.0563, 0.0450]], + device='cuda:0'), grad: tensor([[ 1.1558e-06, 4.6007e-06, 1.8608e-06, ..., 7.0687e-07, + 3.2894e-06, 7.4273e-08], + [ 1.1241e-06, 6.0648e-06, 3.0156e-06, ..., 3.0790e-06, + 4.8280e-06, -1.0142e-06], + [-2.3976e-05, -3.2485e-05, 1.7211e-06, ..., 5.7882e-07, + -4.8988e-06, 7.0175e-07], + ..., + [ 1.6950e-06, 2.6412e-06, 8.9314e-07, ..., 2.8405e-08, + 1.9576e-06, 2.3982e-08], + [-1.2048e-05, -6.4075e-05, -5.2959e-05, ..., 1.7826e-06, + -9.6679e-05, 2.1025e-07], + [ 8.9640e-07, 2.5821e-07, 5.8580e-07, ..., 1.1222e-07, + 8.4797e-07, -4.1421e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0264, -0.0199, -0.0073, 0.0003, 0.0074, 0.0014, -0.0182, 0.0006, + 0.0227, 0.0192], device='cuda:0'), grad: tensor([ 7.0669e-06, 1.0587e-05, -6.8307e-05, 5.1975e-05, 4.2608e-08, + 9.6738e-05, -1.1072e-05, 7.9870e-06, -9.8288e-05, 3.2261e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 257.14, cls_loss 0.0030 cls_loss_mapping 0.0060 cls_loss_causal 0.5426 re_mapping 0.0078 re_causal 0.0238 /// teacc 99.01 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0291, -0.0799, -0.0756, ..., -0.0230, -0.1395, -0.0076], + [-0.0779, -0.1121, -0.1106, ..., -0.0971, -0.0508, -0.0028], + [ 0.0630, 0.0511, -0.1027, ..., -0.0886, -0.1034, -0.0609], + ..., + [ 0.0134, 0.0339, -0.0401, ..., -0.0302, -0.1073, 0.0468], + [ 0.0020, 0.0885, 0.0753, ..., -0.0691, 0.0556, -0.1055], + [-0.0454, -0.0082, -0.0220, ..., -0.0259, -0.0561, 0.0449]], + device='cuda:0'), grad: tensor([[-2.4125e-05, 5.3465e-05, 7.1287e-05, ..., 1.1867e-04, + 1.5283e-04, 5.9465e-07], + [ 8.9407e-07, 1.3513e-06, 1.7080e-06, ..., 2.2128e-06, + 3.8855e-06, 8.9407e-08], + [ 1.3024e-05, -1.8403e-06, 1.5320e-06, ..., 1.8887e-06, + 4.1611e-06, 2.1863e-07], + ..., + [ 1.2089e-06, -9.3132e-09, 9.1409e-07, ..., 3.6764e-07, + 1.7369e-06, -8.7777e-08], + [ 1.1241e-06, 5.3532e-06, 9.2015e-06, ..., 8.8587e-06, + 1.0863e-05, 3.5693e-07], + [ 4.1984e-06, 1.8487e-06, 1.3858e-05, ..., 4.0606e-06, + 1.4447e-05, 5.9279e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0284, -0.0196, -0.0076, -0.0009, 0.0071, 0.0021, -0.0187, 0.0005, + 0.0220, 0.0183], device='cuda:0'), grad: tensor([ 2.8205e-04, 1.2092e-05, 3.6716e-05, -1.5236e-05, 8.7619e-06, + 4.6754e-04, -8.7547e-04, 8.1733e-06, 2.7046e-05, 4.8399e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 257.04, cls_loss 0.0027 cls_loss_mapping 0.0075 cls_loss_causal 0.5665 re_mapping 0.0077 re_causal 0.0247 /// teacc 98.83 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0295, -0.0799, -0.0762, ..., -0.0234, -0.1402, -0.0077], + [-0.0783, -0.1128, -0.1117, ..., -0.0975, -0.0520, -0.0021], + [ 0.0630, 0.0513, -0.1036, ..., -0.0888, -0.1043, -0.0612], + ..., + [ 0.0136, 0.0341, -0.0406, ..., -0.0297, -0.1081, 0.0462], + [ 0.0022, 0.0896, 0.0759, ..., -0.0694, 0.0565, -0.1059], + [-0.0457, -0.0086, -0.0227, ..., -0.0262, -0.0557, 0.0453]], + device='cuda:0'), grad: tensor([[-3.0752e-06, 2.3339e-06, 1.3532e-06, ..., 4.8429e-07, + 1.1418e-06, 3.2596e-09], + [ 9.4175e-06, 2.9191e-05, 1.3374e-05, ..., 5.9158e-06, + 8.4415e-06, 1.3746e-05], + [ 2.5053e-06, 7.1116e-06, 7.5176e-06, ..., 1.9260e-06, + 6.9737e-06, 1.2545e-06], + ..., + [ 7.3016e-06, 2.5481e-05, 1.2601e-06, ..., 3.7136e-07, + 1.0906e-06, 3.6091e-05], + [-1.3679e-05, -5.0157e-05, -2.2516e-05, ..., -1.5013e-05, + -7.7561e-06, -3.1367e-06], + [-7.6294e-06, -3.7432e-05, 1.4240e-06, ..., 1.5926e-07, + 1.5292e-06, -5.5075e-05]], device='cuda:0') +Epoch 121, bias, value: tensor([ 2.8732e-02, -1.9494e-02, -7.7390e-03, -1.5409e-03, 7.1881e-03, + 2.3704e-03, -1.8773e-02, 4.5337e-05, 2.2548e-02, 1.8267e-02], + device='cuda:0'), grad: tensor([-6.5491e-06, 4.8429e-05, 1.7241e-05, -3.4302e-05, 5.8599e-06, + 4.2170e-06, 2.9862e-05, 6.1214e-05, -4.6462e-05, -7.9632e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 257.02, cls_loss 0.0037 cls_loss_mapping 0.0070 cls_loss_causal 0.5394 re_mapping 0.0080 re_causal 0.0239 /// teacc 98.85 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0304, -0.0795, -0.0772, ..., -0.0222, -0.1397, -0.0077], + [-0.0805, -0.1151, -0.1123, ..., -0.0980, -0.0525, -0.0039], + [ 0.0626, 0.0508, -0.1045, ..., -0.0890, -0.1052, -0.0618], + ..., + [ 0.0149, 0.0359, -0.0411, ..., -0.0292, -0.1086, 0.0482], + [ 0.0028, 0.0906, 0.0760, ..., -0.0699, 0.0572, -0.1066], + [-0.0460, -0.0088, -0.0233, ..., -0.0263, -0.0558, 0.0452]], + device='cuda:0'), grad: tensor([[ 1.5059e-06, 1.3635e-06, 1.0198e-06, ..., 2.6124e-07, + 1.0030e-06, 6.4401e-07], + [ 1.6503e-06, 1.5441e-06, 1.0654e-06, ..., 1.4179e-07, + 1.1194e-06, -1.7034e-06], + [-1.2077e-05, -2.6245e-06, 1.3886e-06, ..., 2.2072e-07, + 1.7574e-06, 4.9314e-07], + ..., + [ 1.3802e-06, 2.1104e-06, 1.2042e-06, ..., 6.2864e-09, + 2.0936e-06, 2.7400e-06], + [ 6.8471e-06, 2.0098e-06, -9.2667e-08, ..., 7.3342e-08, + -6.0070e-08, 2.2780e-06], + [-2.8219e-06, -8.4043e-06, 1.1735e-06, ..., 1.5600e-08, + -9.1316e-07, -7.7188e-06]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0294, -0.0206, -0.0086, -0.0022, 0.0071, 0.0033, -0.0199, 0.0017, + 0.0227, 0.0178], device='cuda:0'), grad: tensor([ 6.0573e-06, 7.0594e-07, -1.7151e-05, -8.0988e-06, -3.9935e-05, + 7.8836e-07, 4.4733e-05, 9.8646e-06, 1.7598e-05, -1.4536e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 255.30, cls_loss 0.0031 cls_loss_mapping 0.0050 cls_loss_causal 0.5733 re_mapping 0.0072 re_causal 0.0240 /// teacc 98.91 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0304, -0.0798, -0.0781, ..., -0.0226, -0.1404, -0.0080], + [-0.0812, -0.1167, -0.1142, ..., -0.1008, -0.0547, -0.0041], + [ 0.0659, 0.0541, -0.1050, ..., -0.0892, -0.1056, -0.0608], + ..., + [ 0.0130, 0.0340, -0.0416, ..., -0.0290, -0.1089, 0.0482], + [ 0.0026, 0.0901, 0.0758, ..., -0.0711, 0.0568, -0.1079], + [-0.0462, -0.0087, -0.0236, ..., -0.0263, -0.0562, 0.0450]], + device='cuda:0'), grad: tensor([[ 3.2093e-06, 2.6107e-05, 2.6971e-05, ..., 2.0206e-05, + 2.0102e-05, 7.3165e-06], + [ 4.0093e-07, 1.6866e-06, 2.1327e-06, ..., 2.6217e-07, + 3.5223e-06, -1.3500e-05], + [-1.0185e-05, -1.0431e-05, 3.8184e-06, ..., 8.0233e-07, + 5.9456e-06, 1.0267e-05], + ..., + [-9.3970e-07, -5.5656e-06, 2.6021e-06, ..., 1.7323e-06, + 1.7295e-06, 4.2915e-06], + [ 4.5486e-06, 8.7097e-06, 6.6683e-06, ..., 2.4475e-06, + 5.4911e-06, 1.2089e-06], + [-8.7637e-07, -6.2361e-06, 5.4277e-06, ..., -3.6769e-06, + 6.7130e-06, -7.7635e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0292, -0.0214, -0.0061, -0.0026, 0.0088, 0.0034, -0.0185, 0.0007, + 0.0219, 0.0168], device='cuda:0'), grad: tensor([ 6.5327e-05, -2.1291e-04, 1.3638e-04, -9.4771e-05, 7.5251e-06, + 3.7718e-04, -3.6025e-04, 4.6968e-05, 3.2455e-05, 1.8384e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 256.79, cls_loss 0.0033 cls_loss_mapping 0.0054 cls_loss_causal 0.5703 re_mapping 0.0075 re_causal 0.0238 /// teacc 98.95 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0304, -0.0800, -0.0787, ..., -0.0228, -0.1407, -0.0082], + [-0.0818, -0.1172, -0.1147, ..., -0.1011, -0.0550, -0.0041], + [ 0.0668, 0.0547, -0.1058, ..., -0.0895, -0.1061, -0.0614], + ..., + [ 0.0130, 0.0340, -0.0421, ..., -0.0279, -0.1093, 0.0481], + [ 0.0015, 0.0896, 0.0761, ..., -0.0713, 0.0571, -0.1084], + [-0.0458, -0.0080, -0.0230, ..., -0.0254, -0.0562, 0.0463]], + device='cuda:0'), grad: tensor([[ 1.7919e-06, 6.1132e-06, 1.0096e-05, ..., 4.0978e-08, + 6.6757e-06, 1.0151e-07], + [ 3.7998e-07, 5.7649e-07, 4.2235e-07, ..., -3.4925e-08, + 3.3760e-07, -1.5339e-06], + [ 3.2540e-06, 3.5521e-06, 2.9188e-06, ..., 6.5193e-09, + 2.5760e-06, 4.8103e-07], + ..., + [-4.8988e-07, -9.6019e-07, 7.6275e-07, ..., 3.2596e-08, + 7.0455e-07, 3.1386e-07], + [-1.2681e-05, -1.5378e-05, -1.7032e-05, ..., 4.8708e-07, + -1.1958e-05, 9.1270e-08], + [ 2.9784e-06, 3.7672e-07, 1.8030e-06, ..., 3.7253e-08, + 1.0645e-06, -2.0955e-08]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0292, -0.0216, -0.0057, -0.0030, 0.0082, 0.0039, -0.0190, 0.0007, + 0.0208, 0.0177], device='cuda:0'), grad: tensor([ 2.9951e-06, -2.9076e-06, 8.4490e-06, 3.6396e-06, 2.7008e-08, + -3.9749e-06, 5.7556e-06, 4.8382e-07, -2.6643e-05, 1.2159e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 256.60, cls_loss 0.0030 cls_loss_mapping 0.0072 cls_loss_causal 0.5619 re_mapping 0.0083 re_causal 0.0242 /// teacc 99.00 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0304, -0.0803, -0.0796, ..., -0.0232, -0.1414, -0.0084], + [-0.0822, -0.1175, -0.1151, ..., -0.1011, -0.0553, -0.0040], + [ 0.0669, 0.0546, -0.1069, ..., -0.0899, -0.1073, -0.0619], + ..., + [ 0.0124, 0.0336, -0.0428, ..., -0.0278, -0.1100, 0.0475], + [ 0.0019, 0.0906, 0.0766, ..., -0.0713, 0.0575, -0.1090], + [-0.0448, -0.0064, -0.0228, ..., -0.0251, -0.0560, 0.0478]], + device='cuda:0'), grad: tensor([[ 4.5970e-06, 5.6662e-06, 1.2685e-06, ..., 3.0547e-07, + 1.5348e-06, 3.6880e-07], + [ 6.2957e-07, -2.3052e-05, 3.5716e-07, ..., -8.7917e-07, + 1.7509e-07, -3.2093e-06], + [-1.5022e-06, 5.4128e-06, 6.6608e-06, ..., 2.4820e-07, + 6.7800e-06, 9.3738e-07], + ..., + [-5.0478e-06, -4.4666e-06, 1.0030e-06, ..., 1.4668e-07, + 7.2643e-08, -1.4575e-07], + [-3.4105e-06, 6.4485e-06, -6.4522e-06, ..., 8.3027e-07, + -5.7369e-06, 7.5344e-07], + [ 9.9186e-07, 3.2280e-06, 8.3074e-07, ..., 1.3039e-07, + 1.0654e-06, 1.9651e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0291, -0.0213, -0.0063, -0.0026, 0.0076, 0.0043, -0.0193, 0.0002, + 0.0209, 0.0185], device='cuda:0'), grad: tensor([ 2.4214e-05, -2.5368e-04, 2.2128e-05, 5.9558e-07, 7.2755e-06, + 7.7784e-06, 1.0543e-05, 1.9759e-05, 1.2994e-04, 3.1173e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 256.54, cls_loss 0.0029 cls_loss_mapping 0.0075 cls_loss_causal 0.5420 re_mapping 0.0075 re_causal 0.0225 /// teacc 99.02 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0303, -0.0807, -0.0808, ..., -0.0240, -0.1425, -0.0087], + [-0.0826, -0.1175, -0.1169, ..., -0.1014, -0.0564, -0.0035], + [ 0.0669, 0.0544, -0.1088, ..., -0.0907, -0.1083, -0.0624], + ..., + [ 0.0127, 0.0338, -0.0432, ..., -0.0280, -0.1103, 0.0474], + [ 0.0030, 0.0927, 0.0782, ..., -0.0710, 0.0591, -0.1095], + [-0.0449, -0.0068, -0.0238, ..., -0.0253, -0.0570, 0.0479]], + device='cuda:0'), grad: tensor([[ 2.2631e-07, 6.0022e-05, 2.0176e-05, ..., 6.0678e-05, + 1.3836e-05, 1.5631e-05], + [ 8.0168e-06, 5.3197e-05, 2.3805e-06, ..., 8.0988e-06, + 9.9279e-07, 2.1122e-06], + [-5.5134e-06, -4.1097e-05, 1.7313e-06, ..., 8.3679e-07, + 1.1483e-06, 2.0536e-07], + ..., + [-2.0815e-07, 2.7120e-06, 2.8703e-06, ..., 6.0629e-07, + 1.6093e-06, 3.7719e-08], + [-2.4274e-05, -4.0203e-05, -3.4839e-05, ..., 8.1211e-06, + -1.6987e-05, 7.3807e-07], + [ 9.5665e-06, 3.0756e-05, 1.8328e-05, ..., 7.1377e-06, + 1.0915e-05, 1.3178e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0288, -0.0208, -0.0068, -0.0037, 0.0072, 0.0051, -0.0191, 0.0002, + 0.0225, 0.0183], device='cuda:0'), grad: tensor([ 1.7715e-04, 3.3641e-04, -2.9922e-04, 3.2373e-06, 2.6274e-04, + 1.2863e-04, -6.8998e-04, 1.8284e-05, -4.8786e-05, 1.1164e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 256.99, cls_loss 0.0023 cls_loss_mapping 0.0056 cls_loss_causal 0.5572 re_mapping 0.0069 re_causal 0.0229 /// teacc 98.95 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0306, -0.0809, -0.0816, ..., -0.0243, -0.1430, -0.0087], + [-0.0833, -0.1181, -0.1174, ..., -0.1025, -0.0568, -0.0035], + [ 0.0668, 0.0543, -0.1092, ..., -0.0909, -0.1088, -0.0633], + ..., + [ 0.0130, 0.0340, -0.0436, ..., -0.0281, -0.1107, 0.0474], + [ 0.0034, 0.0933, 0.0787, ..., -0.0717, 0.0596, -0.1098], + [-0.0450, -0.0069, -0.0245, ..., -0.0255, -0.0577, 0.0484]], + device='cuda:0'), grad: tensor([[-1.1520e-06, 6.8638e-07, 2.7288e-07, ..., 1.4668e-07, + 1.5041e-07, 2.2678e-07], + [ 3.8696e-07, 1.0887e-06, 6.1933e-08, ..., 1.8161e-08, + 4.5635e-08, -2.2333e-06], + [-5.5395e-06, -2.1815e-05, 5.3085e-08, ..., 1.7229e-08, + 4.6566e-08, 1.4193e-06], + ..., + [ 5.6736e-06, 2.1473e-05, 1.4575e-07, ..., 4.6566e-10, + 8.4750e-08, -1.4296e-06], + [ 1.2955e-06, 2.9318e-06, 2.6869e-07, ..., 3.9581e-08, + 1.4203e-07, 6.9244e-07], + [ 1.1427e-06, 1.2340e-06, 3.9535e-07, ..., 6.9849e-09, + 3.1805e-07, 1.4761e-07]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0286, -0.0208, -0.0071, -0.0038, 0.0075, 0.0048, -0.0188, 0.0002, + 0.0227, 0.0184], device='cuda:0'), grad: tensor([-1.2554e-06, 3.1926e-06, -1.9699e-05, 2.2426e-06, -1.6797e-04, + 4.3400e-06, 1.3018e-04, 3.0994e-05, 8.0094e-06, 9.9763e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 256.88, cls_loss 0.0024 cls_loss_mapping 0.0058 cls_loss_causal 0.5396 re_mapping 0.0070 re_causal 0.0220 /// teacc 99.02 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0307, -0.0812, -0.0825, ..., -0.0247, -0.1436, -0.0088], + [-0.0836, -0.1180, -0.1178, ..., -0.1027, -0.0568, -0.0033], + [ 0.0670, 0.0543, -0.1101, ..., -0.0910, -0.1095, -0.0635], + ..., + [ 0.0129, 0.0341, -0.0447, ..., -0.0281, -0.1113, 0.0474], + [ 0.0042, 0.0947, 0.0806, ..., -0.0716, 0.0610, -0.1115], + [-0.0450, -0.0067, -0.0248, ..., -0.0255, -0.0576, 0.0486]], + device='cuda:0'), grad: tensor([[ 2.8387e-06, 2.7716e-06, 3.6899e-06, ..., 5.3551e-08, + 7.1377e-06, 1.9446e-06], + [ 3.2801e-06, 1.0626e-06, 2.5723e-06, ..., -9.7789e-09, + 3.3919e-06, 1.7649e-06], + [ 4.3064e-06, 1.1614e-06, 4.7572e-06, ..., 9.3132e-09, + 6.4336e-06, 3.4086e-06], + ..., + [ 1.1168e-05, 1.4696e-06, 4.1164e-06, ..., 2.7940e-09, + 5.8785e-06, 1.8263e-06], + [ 9.4846e-06, 6.6534e-06, 9.7975e-06, ..., 3.0268e-08, + 2.0668e-05, 5.6252e-06], + [ 4.5337e-06, -1.4156e-05, -2.4110e-05, ..., 1.3970e-09, + -5.3495e-05, -1.1809e-05]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0287, -0.0206, -0.0071, -0.0035, 0.0074, 0.0038, -0.0188, 0.0001, + 0.0234, 0.0183], device='cuda:0'), grad: tensor([ 1.1697e-05, 2.0474e-05, 2.2188e-05, -5.2154e-06, -4.6182e-04, + 2.2441e-05, 1.2465e-05, 1.1665e-04, 5.6863e-05, 2.0468e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 256.87, cls_loss 0.0030 cls_loss_mapping 0.0057 cls_loss_causal 0.5374 re_mapping 0.0076 re_causal 0.0220 /// teacc 99.06 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0307, -0.0815, -0.0833, ..., -0.0250, -0.1442, -0.0091], + [-0.0842, -0.1183, -0.1182, ..., -0.1028, -0.0571, -0.0033], + [ 0.0676, 0.0545, -0.1109, ..., -0.0912, -0.1098, -0.0637], + ..., + [ 0.0133, 0.0345, -0.0454, ..., -0.0280, -0.1118, 0.0477], + [ 0.0043, 0.0950, 0.0807, ..., -0.0723, 0.0614, -0.1123], + [-0.0450, -0.0069, -0.0249, ..., -0.0254, -0.0578, 0.0486]], + device='cuda:0'), grad: tensor([[ 5.1036e-06, 1.7926e-05, 1.3039e-07, ..., 7.2643e-08, + 7.1200e-07, 7.6517e-06], + [ 3.6415e-07, 3.9581e-07, 7.2177e-08, ..., 5.5879e-09, + 7.3574e-08, -4.3586e-07], + [-2.3516e-07, -3.0752e-06, 4.2701e-07, ..., 2.0955e-08, + 3.4412e-07, 7.3574e-07], + ..., + [-1.2759e-07, 2.6915e-06, 1.0105e-07, ..., 9.3132e-10, + 5.3784e-07, -5.9325e-07], + [ 8.0019e-06, 1.0677e-05, 9.3598e-08, ..., -1.2107e-08, + 1.1707e-06, 4.8392e-06], + [-2.5183e-05, -3.5286e-05, 8.4424e-07, ..., 1.4435e-08, + -3.4552e-06, -1.5736e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0287, -0.0207, -0.0068, -0.0036, 0.0070, 0.0038, -0.0192, 0.0005, + 0.0229, 0.0185], device='cuda:0'), grad: tensor([ 1.3113e-06, -6.5239e-07, -3.2671e-06, 1.4126e-05, 6.8685e-07, + -4.9919e-06, 2.4080e-05, 1.7732e-06, 1.9535e-05, -5.2601e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 257.15, cls_loss 0.0036 cls_loss_mapping 0.0073 cls_loss_causal 0.5400 re_mapping 0.0071 re_causal 0.0216 /// teacc 98.95 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0309, -0.0817, -0.0855, ..., -0.0252, -0.1465, -0.0093], + [-0.0844, -0.1190, -0.1184, ..., -0.1029, -0.0584, -0.0027], + [ 0.0674, 0.0547, -0.1118, ..., -0.0914, -0.1086, -0.0663], + ..., + [ 0.0138, 0.0349, -0.0460, ..., -0.0275, -0.1123, 0.0477], + [ 0.0041, 0.0959, 0.0827, ..., -0.0726, 0.0624, -0.1130], + [-0.0449, -0.0067, -0.0253, ..., -0.0259, -0.0583, 0.0486]], + device='cuda:0'), grad: tensor([[-4.6045e-06, 1.0999e-06, -3.2689e-07, ..., 5.0385e-07, + -3.9442e-07, 6.2305e-07], + [ 5.6578e-07, 5.0552e-06, 1.1940e-06, ..., -9.3598e-08, + 1.0505e-06, -6.0489e-07], + [ 3.6135e-07, 2.9970e-06, 7.2923e-07, ..., 1.9139e-07, + 7.6042e-07, 7.3062e-07], + ..., + [-4.9621e-06, -1.8001e-05, 6.6170e-07, ..., 2.8405e-08, + 5.7556e-07, -1.1049e-05], + [ 1.4035e-06, 1.0677e-05, 2.3112e-05, ..., 1.0729e-06, + 1.9580e-05, 6.2166e-07], + [ 3.7011e-06, 5.5060e-06, 1.1260e-06, ..., 5.8673e-08, + 9.9000e-07, 5.7593e-06]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0285, -0.0192, -0.0083, -0.0033, 0.0067, 0.0042, -0.0202, 0.0006, + 0.0233, 0.0184], device='cuda:0'), grad: tensor([-1.9372e-05, 1.1824e-05, 1.1109e-05, -5.2124e-05, 1.6242e-06, + 2.9832e-05, 3.2894e-06, -5.7608e-05, 4.4584e-05, 2.6658e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 256.82, cls_loss 0.0039 cls_loss_mapping 0.0078 cls_loss_causal 0.5529 re_mapping 0.0071 re_causal 0.0211 /// teacc 99.00 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0311, -0.0821, -0.0862, ..., -0.0260, -0.1470, -0.0095], + [-0.0850, -0.1188, -0.1191, ..., -0.1034, -0.0588, -0.0017], + [ 0.0675, 0.0546, -0.1125, ..., -0.0918, -0.1092, -0.0675], + ..., + [ 0.0139, 0.0349, -0.0465, ..., -0.0276, -0.1128, 0.0471], + [ 0.0047, 0.0957, 0.0822, ..., -0.0743, 0.0622, -0.1135], + [-0.0443, -0.0053, -0.0263, ..., -0.0264, -0.0588, 0.0489]], + device='cuda:0'), grad: tensor([[-2.8266e-07, 2.7474e-06, 2.6729e-06, ..., 1.7397e-06, + 2.3674e-06, 8.7544e-07], + [ 1.4976e-06, 3.5260e-06, 5.0887e-06, ..., 2.5667e-06, + 4.2692e-06, -7.3165e-06], + [-1.8952e-06, 5.0627e-06, 5.5693e-06, ..., 3.9339e-06, + 5.0254e-06, 1.2340e-06], + ..., + [-5.8673e-08, 7.2736e-07, 2.5760e-06, ..., 1.3895e-06, + 2.1979e-06, 5.2378e-06], + [ 6.3796e-08, 6.0499e-05, 8.8573e-05, ..., 4.9144e-05, + 7.6413e-05, 2.3609e-07], + [-1.5162e-06, 1.1623e-06, 6.8061e-06, ..., 3.3565e-06, + 5.6028e-06, -3.3975e-06]], device='cuda:0') +Epoch 131, bias, value: tensor([ 2.8567e-02, -1.7859e-02, -9.0745e-03, -2.2606e-03, 5.1761e-03, + 3.1119e-03, -1.8991e-02, -8.3274e-05, 2.2161e-02, 1.9299e-02], + device='cuda:0'), grad: tensor([ 1.3098e-05, -1.4096e-05, 1.1221e-05, 5.0336e-05, 5.0180e-06, + -4.6825e-04, 1.8024e-04, 3.4422e-05, 1.9813e-04, -1.0580e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 257.20, cls_loss 0.0023 cls_loss_mapping 0.0069 cls_loss_causal 0.5395 re_mapping 0.0075 re_causal 0.0228 /// teacc 99.02 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0311, -0.0825, -0.0873, ..., -0.0264, -0.1477, -0.0098], + [-0.0854, -0.1189, -0.1194, ..., -0.1039, -0.0590, -0.0015], + [ 0.0677, 0.0546, -0.1128, ..., -0.0920, -0.1095, -0.0677], + ..., + [ 0.0139, 0.0349, -0.0468, ..., -0.0278, -0.1131, 0.0471], + [ 0.0058, 0.0964, 0.0828, ..., -0.0748, 0.0627, -0.1140], + [-0.0440, -0.0048, -0.0279, ..., -0.0270, -0.0598, 0.0492]], + device='cuda:0'), grad: tensor([[ 4.1025e-07, 8.4192e-07, 3.5036e-06, ..., 2.0117e-07, + 1.6894e-06, 1.2955e-06], + [ 3.7309e-06, -4.0606e-06, 5.0031e-06, ..., 6.7521e-08, + 3.9600e-06, -1.9446e-05], + [ 1.3569e-06, 3.0603e-06, 3.4329e-06, ..., 8.3819e-08, + 2.6934e-06, 3.6210e-06], + ..., + [-4.4405e-06, -2.9355e-06, 2.7250e-06, ..., 3.0734e-08, + 1.7928e-06, 1.3553e-05], + [ 5.4948e-08, -1.6717e-06, 2.3887e-05, ..., 5.5367e-07, + 1.4618e-05, 1.4799e-06], + [ 2.6897e-06, -5.6438e-07, 5.7779e-06, ..., 1.0151e-07, + 1.0610e-05, -3.1497e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0282, -0.0178, -0.0090, -0.0017, 0.0044, 0.0020, -0.0180, -0.0001, + 0.0226, 0.0198], device='cuda:0'), grad: tensor([ 1.7121e-05, -2.8148e-05, 2.8133e-05, -6.3300e-05, -7.6056e-05, + 9.8720e-06, 1.2398e-05, 4.8667e-05, 6.2287e-05, -1.1228e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 256.57, cls_loss 0.0034 cls_loss_mapping 0.0062 cls_loss_causal 0.5661 re_mapping 0.0069 re_causal 0.0204 /// teacc 99.00 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0315, -0.0828, -0.0877, ..., -0.0264, -0.1482, -0.0101], + [-0.0861, -0.1192, -0.1198, ..., -0.1042, -0.0593, -0.0009], + [ 0.0679, 0.0546, -0.1140, ..., -0.0926, -0.1102, -0.0682], + ..., + [ 0.0143, 0.0353, -0.0476, ..., -0.0263, -0.1132, 0.0468], + [ 0.0060, 0.0966, 0.0830, ..., -0.0751, 0.0629, -0.1148], + [-0.0442, -0.0049, -0.0290, ..., -0.0280, -0.0604, 0.0492]], + device='cuda:0'), grad: tensor([[ 4.9770e-05, 9.2341e-07, 5.9837e-07, ..., 8.1398e-07, + 6.6578e-05, 5.9575e-05], + [ 4.0010e-06, 1.8589e-06, 5.6950e-07, ..., 9.0338e-08, + 5.6736e-06, 4.3698e-06], + [ 2.5555e-05, -1.3178e-06, 4.3632e-07, ..., 5.9605e-08, + 3.6091e-05, 3.1233e-05], + ..., + [ 1.0364e-05, 5.6345e-07, 2.4913e-07, ..., 4.1910e-09, + 1.4193e-05, 1.1995e-05], + [ 1.7462e-06, -2.3041e-06, -1.0794e-06, ..., 3.2177e-07, + 1.0505e-06, 2.1476e-06], + [-7.7295e-04, 1.3085e-07, 1.2852e-07, ..., 4.3306e-08, + -1.0061e-03, -9.0694e-04]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0286, -0.0174, -0.0090, -0.0016, 0.0045, 0.0019, -0.0181, -0.0003, + 0.0221, 0.0195], device='cuda:0'), grad: tensor([ 6.7282e-04, 5.5850e-05, 3.6430e-04, 8.7051e-03, 1.5661e-05, + 5.8413e-04, 7.4245e-06, 1.4424e-04, 2.4527e-05, -1.0574e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 255.35, cls_loss 0.0024 cls_loss_mapping 0.0060 cls_loss_causal 0.5313 re_mapping 0.0072 re_causal 0.0219 /// teacc 99.02 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0316, -0.0830, -0.0880, ..., -0.0269, -0.1487, -0.0104], + [-0.0867, -0.1205, -0.1211, ..., -0.1047, -0.0600, -0.0012], + [ 0.0683, 0.0547, -0.1145, ..., -0.0928, -0.1107, -0.0683], + ..., + [ 0.0143, 0.0358, -0.0481, ..., -0.0264, -0.1139, 0.0471], + [ 0.0061, 0.0963, 0.0827, ..., -0.0756, 0.0624, -0.1151], + [-0.0445, -0.0051, -0.0295, ..., -0.0280, -0.0594, 0.0494]], + device='cuda:0'), grad: tensor([[ 3.1944e-06, 1.4775e-05, 1.4201e-05, ..., 3.1106e-06, + 9.4399e-06, 2.6431e-06], + [ 2.6390e-05, 5.1647e-05, 3.5256e-05, ..., 9.5041e-07, + 3.1263e-05, -1.3551e-06], + [ 5.2005e-05, 1.2863e-04, 1.1271e-04, ..., 2.6301e-05, + 7.8797e-05, 9.3058e-06], + ..., + [ 1.5378e-05, 2.4542e-05, 1.6317e-05, ..., 3.3099e-06, + 1.3076e-05, 8.3074e-06], + [-1.1051e-04, -2.2984e-04, -1.2672e-04, ..., -6.2846e-06, + -1.1253e-04, 9.0012e-07], + [ 5.2191e-06, 1.7300e-05, 1.8179e-05, ..., 3.7719e-06, + 1.3039e-05, -8.7395e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0288, -0.0181, -0.0089, -0.0022, 0.0055, 0.0020, -0.0184, 0.0003, + 0.0214, 0.0196], device='cuda:0'), grad: tensor([ 3.7968e-05, 1.0812e-04, 3.4761e-04, -1.8673e-03, 1.2003e-05, + 1.6394e-03, 7.0453e-05, 7.0453e-05, -3.8981e-04, -2.8506e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 256.79, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5320 re_mapping 0.0074 re_causal 0.0221 /// teacc 98.94 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0318, -0.0835, -0.0895, ..., -0.0287, -0.1497, -0.0105], + [-0.0871, -0.1209, -0.1215, ..., -0.1050, -0.0604, -0.0011], + [ 0.0685, 0.0545, -0.1156, ..., -0.0935, -0.1117, -0.0685], + ..., + [ 0.0146, 0.0363, -0.0487, ..., -0.0263, -0.1146, 0.0472], + [ 0.0052, 0.0955, 0.0828, ..., -0.0763, 0.0622, -0.1177], + [-0.0444, -0.0044, -0.0292, ..., -0.0287, -0.0584, 0.0497]], + device='cuda:0'), grad: tensor([[ 4.1053e-06, 5.7295e-06, 1.0896e-07, ..., 6.2864e-08, + 1.5227e-07, 3.4738e-07], + [ 4.6268e-06, 8.4117e-06, 8.1956e-08, ..., 7.4506e-09, + 2.6729e-07, -2.4401e-07], + [-1.8254e-06, 2.8402e-05, 2.6496e-07, ..., 1.7695e-08, + 6.7987e-07, 1.6019e-07], + ..., + [-2.6673e-05, -6.1393e-05, 1.4622e-07, ..., 4.6566e-10, + 4.7591e-07, 2.2762e-06], + [ 1.2070e-06, 1.9092e-06, -5.2387e-07, ..., 5.3085e-08, + -6.9011e-07, 9.0292e-07], + [-1.0831e-06, -9.2089e-06, 7.5437e-08, ..., 4.1910e-09, + 3.5577e-07, -6.1132e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0287, -0.0189, -0.0089, -0.0019, 0.0058, 0.0019, -0.0182, 0.0014, + 0.0206, 0.0195], device='cuda:0'), grad: tensor([ 1.7866e-05, 2.4021e-05, 5.8174e-05, 5.5671e-05, 9.9018e-06, + 3.8296e-06, 1.2167e-05, -1.7285e-04, 8.6725e-06, -1.7688e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 256.90, cls_loss 0.0030 cls_loss_mapping 0.0059 cls_loss_causal 0.5268 re_mapping 0.0077 re_causal 0.0220 /// teacc 99.02 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0319, -0.0840, -0.0903, ..., -0.0294, -0.1502, -0.0110], + [-0.0881, -0.1217, -0.1220, ..., -0.1055, -0.0608, -0.0016], + [ 0.0686, 0.0542, -0.1164, ..., -0.0937, -0.1124, -0.0692], + ..., + [ 0.0154, 0.0371, -0.0491, ..., -0.0253, -0.1154, 0.0486], + [ 0.0062, 0.0971, 0.0834, ..., -0.0758, 0.0633, -0.1176], + [-0.0448, -0.0057, -0.0296, ..., -0.0291, -0.0591, 0.0483]], + device='cuda:0'), grad: tensor([[-8.1509e-06, 2.7791e-06, 2.6375e-06, ..., 3.0715e-06, + 2.1271e-06, 8.4937e-07], + [ 1.3728e-06, 1.3765e-06, 1.1967e-07, ..., 4.8801e-07, + 1.3923e-07, -2.1569e-06], + [-2.2184e-06, -2.3581e-06, 1.3225e-07, ..., 1.7649e-07, + -1.4901e-07, 3.0780e-07], + ..., + [-1.8068e-06, -1.2340e-06, 1.3504e-08, ..., -1.2526e-07, + 2.6543e-08, -1.8012e-06], + [ 1.6559e-06, 2.9150e-06, 4.4191e-07, ..., 4.2003e-07, + 4.1351e-07, 7.8417e-07], + [ 6.3926e-06, 3.9488e-06, 1.7276e-07, ..., 4.3120e-07, + 1.4342e-07, 9.5740e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0285, -0.0193, -0.0093, -0.0021, 0.0066, 0.0016, -0.0181, 0.0029, + 0.0216, 0.0183], device='cuda:0'), grad: tensor([-4.8548e-05, -6.7614e-06, -8.1351e-07, 5.7407e-06, -5.7697e-05, + 3.7216e-06, 1.6704e-05, 1.0140e-05, 1.3337e-05, 6.4135e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 256.81, cls_loss 0.0030 cls_loss_mapping 0.0062 cls_loss_causal 0.5643 re_mapping 0.0071 re_causal 0.0218 /// teacc 98.97 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0318, -0.0845, -0.0916, ..., -0.0300, -0.1512, -0.0115], + [-0.0889, -0.1220, -0.1227, ..., -0.1062, -0.0615, -0.0014], + [ 0.0696, 0.0547, -0.1176, ..., -0.0940, -0.1136, -0.0693], + ..., + [ 0.0152, 0.0375, -0.0500, ..., -0.0251, -0.1178, 0.0487], + [ 0.0062, 0.0984, 0.0843, ..., -0.0755, 0.0640, -0.1179], + [-0.0443, -0.0063, -0.0295, ..., -0.0294, -0.0572, 0.0477]], + device='cuda:0'), grad: tensor([[ 4.3735e-06, 5.3532e-06, 3.7234e-06, ..., 2.6189e-06, + 3.6433e-06, 7.9535e-07], + [ 5.0068e-06, 4.4629e-06, 3.0342e-06, ..., 2.0657e-06, + 3.2373e-06, 2.1551e-06], + [-1.9848e-05, -6.0629e-07, 1.5227e-06, ..., -3.3621e-06, + 1.5618e-06, 4.0652e-07], + ..., + [ 3.5437e-07, 5.0478e-07, 5.3737e-07, ..., 1.7323e-07, + 5.7556e-07, -8.8476e-09], + [ 1.3905e-06, -2.1104e-06, -1.0848e-05, ..., 5.6773e-06, + -8.5011e-06, 2.3395e-06], + [-5.1968e-06, 1.0423e-05, 1.4566e-06, ..., 2.0815e-07, + 1.4324e-06, 4.1366e-05]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0281, -0.0194, -0.0089, -0.0019, 0.0064, 0.0013, -0.0182, 0.0026, + 0.0215, 0.0189], device='cuda:0'), grad: tensor([ 2.5555e-05, 6.2287e-05, -9.2447e-05, 4.1544e-05, -7.2336e-04, + 1.9774e-05, -1.0923e-05, 8.4043e-06, -5.5227e-07, 6.6996e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 257.41, cls_loss 0.0021 cls_loss_mapping 0.0049 cls_loss_causal 0.5401 re_mapping 0.0068 re_causal 0.0215 /// teacc 99.00 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0318, -0.0847, -0.0921, ..., -0.0303, -0.1517, -0.0117], + [-0.0896, -0.1223, -0.1230, ..., -0.1063, -0.0620, -0.0014], + [ 0.0699, 0.0548, -0.1185, ..., -0.0945, -0.1143, -0.0695], + ..., + [ 0.0155, 0.0376, -0.0505, ..., -0.0252, -0.1188, 0.0488], + [ 0.0062, 0.0986, 0.0842, ..., -0.0758, 0.0641, -0.1185], + [-0.0444, -0.0063, -0.0298, ..., -0.0295, -0.0571, 0.0475]], + device='cuda:0'), grad: tensor([[-7.3854e-07, 1.3756e-06, 1.7658e-06, ..., 1.1865e-06, + 1.6522e-06, 9.3598e-08], + [ 6.5425e-07, 7.0361e-07, 3.3993e-07, ..., 8.1025e-08, + 5.2201e-07, 6.3796e-08], + [ 4.0000e-07, 1.6531e-06, 1.8300e-06, ..., 1.9092e-07, + 2.5369e-06, 2.1746e-07], + ..., + [-7.4767e-06, -9.0376e-06, 6.6496e-07, ..., 5.1223e-09, + 1.0338e-06, -4.9807e-06], + [-1.2601e-06, -3.2559e-06, 3.1516e-06, ..., 6.7428e-07, + 2.0228e-06, 6.4261e-08], + [ 7.6815e-06, 8.4043e-06, 5.7407e-06, ..., 8.1025e-08, + 5.6513e-06, 4.1388e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0276, -0.0197, -0.0087, -0.0017, 0.0071, 0.0011, -0.0179, 0.0025, + 0.0212, 0.0189], device='cuda:0'), grad: tensor([ 7.5623e-07, 1.7006e-06, 4.1686e-06, 5.8897e-06, 8.9174e-07, + -2.2963e-05, -4.2394e-06, -1.2919e-05, 2.9933e-06, 2.3693e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 257.34, cls_loss 0.0022 cls_loss_mapping 0.0060 cls_loss_causal 0.5614 re_mapping 0.0069 re_causal 0.0218 /// teacc 98.90 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 0.0316, -0.0855, -0.0929, ..., -0.0307, -0.1531, -0.0128], + [-0.0903, -0.1224, -0.1233, ..., -0.1067, -0.0629, -0.0014], + [ 0.0702, 0.0549, -0.1193, ..., -0.0949, -0.1140, -0.0692], + ..., + [ 0.0159, 0.0379, -0.0509, ..., -0.0253, -0.1190, 0.0489], + [ 0.0064, 0.0989, 0.0843, ..., -0.0763, 0.0644, -0.1194], + [-0.0445, -0.0063, -0.0298, ..., -0.0293, -0.0567, 0.0474]], + device='cuda:0'), grad: tensor([[-3.1386e-07, 4.8950e-06, 4.1127e-06, ..., 2.3786e-06, + 5.7742e-06, 7.4506e-08], + [ 8.4162e-05, 8.0168e-05, 1.4659e-06, ..., 4.5914e-07, + 2.1793e-06, 5.2363e-05], + [ 1.3560e-06, 3.3248e-06, 2.1998e-06, ..., 4.6287e-07, + 3.4049e-06, 7.6182e-07], + ..., + [-9.9421e-05, -9.4354e-05, 1.6652e-06, ..., 3.1944e-07, + 2.2482e-06, -6.2227e-05], + [ 2.4810e-06, 6.5416e-06, 5.3868e-06, ..., 1.5842e-06, + 7.8455e-06, 1.0803e-06], + [ 1.0870e-05, 1.1332e-05, 4.5784e-06, ..., 9.4716e-07, + 6.3851e-06, 5.7966e-06]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0274, -0.0199, -0.0079, -0.0017, 0.0073, 0.0012, -0.0179, 0.0024, + 0.0211, 0.0186], device='cuda:0'), grad: tensor([ 1.0066e-05, 2.1124e-04, 1.1422e-05, -3.6627e-05, 8.1360e-06, + -1.5214e-05, -1.7639e-06, -2.4652e-04, 2.2039e-05, 3.7134e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 251.92, cls_loss 0.0022 cls_loss_mapping 0.0062 cls_loss_causal 0.5142 re_mapping 0.0070 re_causal 0.0204 /// teacc 98.98 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 0.0317, -0.0858, -0.0936, ..., -0.0312, -0.1537, -0.0129], + [-0.0910, -0.1233, -0.1240, ..., -0.1081, -0.0638, -0.0012], + [ 0.0711, 0.0551, -0.1200, ..., -0.0956, -0.1148, -0.0688], + ..., + [ 0.0159, 0.0383, -0.0514, ..., -0.0254, -0.1189, 0.0490], + [ 0.0064, 0.0991, 0.0843, ..., -0.0766, 0.0645, -0.1197], + [-0.0446, -0.0064, -0.0305, ..., -0.0297, -0.0571, 0.0473]], + device='cuda:0'), grad: tensor([[-6.4149e-06, 1.4165e-06, 9.3598e-07, ..., 6.3702e-07, + 8.5030e-07, 3.5763e-07], + [ 1.2163e-06, 2.3060e-06, 2.3842e-07, ..., 8.9407e-08, + 3.1013e-07, -6.5044e-06], + [ 6.5453e-06, 1.5028e-05, 4.2468e-07, ..., 1.3318e-07, + 5.2154e-07, 3.3807e-06], + ..., + [-7.9945e-06, -2.0698e-05, 1.1362e-07, ..., -1.3225e-07, + 1.3970e-07, -2.3283e-06], + [ 7.9814e-07, 5.0254e-06, 6.3740e-06, ..., 4.5747e-06, + 4.4815e-06, 8.5589e-07], + [ 1.5264e-06, 1.7229e-07, 3.7532e-07, ..., 9.2201e-08, + 4.3772e-07, 1.2815e-06]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0275, -0.0199, -0.0077, -0.0017, 0.0074, 0.0020, -0.0190, 0.0026, + 0.0207, 0.0181], device='cuda:0'), grad: tensor([-1.7047e-05, -1.9982e-05, 2.6599e-05, 6.9067e-06, 3.7476e-06, + 7.1302e-06, -6.5118e-06, -2.0206e-05, 9.5516e-06, 9.7975e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 251.57, cls_loss 0.0033 cls_loss_mapping 0.0067 cls_loss_causal 0.5489 re_mapping 0.0071 re_causal 0.0205 /// teacc 98.92 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 0.0318, -0.0865, -0.0949, ..., -0.0322, -0.1545, -0.0132], + [-0.0920, -0.1243, -0.1243, ..., -0.1084, -0.0644, -0.0014], + [ 0.0719, 0.0536, -0.1220, ..., -0.0961, -0.1197, -0.0669], + ..., + [ 0.0133, 0.0365, -0.0518, ..., -0.0239, -0.1223, 0.0489], + [ 0.0112, 0.1047, 0.0848, ..., -0.0768, 0.0697, -0.1181], + [-0.0447, -0.0063, -0.0303, ..., -0.0294, -0.0573, 0.0470]], + device='cuda:0'), grad: tensor([[ 6.6031e-07, 2.0228e-06, 2.2464e-06, ..., 2.7940e-08, + 3.0231e-06, 1.7947e-06], + [ 2.1569e-06, 5.2620e-07, 2.2817e-07, ..., 2.7940e-09, + -2.2985e-06, -1.0580e-06], + [-1.2107e-07, 5.3924e-07, 4.0047e-07, ..., 6.5193e-09, + 3.6508e-07, 9.2387e-07], + ..., + [-4.1164e-06, -8.4043e-06, 3.7439e-07, ..., 7.4506e-09, + 1.0654e-06, -6.4522e-06], + [ 6.3516e-07, 2.4345e-06, 6.9812e-06, ..., 4.0047e-08, + 7.0706e-06, 1.9763e-06], + [-2.0508e-06, -1.6764e-06, 1.0077e-06, ..., 1.0245e-08, + -2.5164e-06, -2.0768e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0275, -0.0205, -0.0078, -0.0022, 0.0072, 0.0020, -0.0192, 0.0016, + 0.0250, 0.0179], device='cuda:0'), grad: tensor([ 4.7460e-06, -6.9924e-06, 2.0750e-06, 8.4713e-06, 6.5081e-06, + -2.5049e-05, 1.4566e-05, -2.3186e-05, 1.9163e-05, -3.6135e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 257.00, cls_loss 0.0024 cls_loss_mapping 0.0059 cls_loss_causal 0.5502 re_mapping 0.0079 re_causal 0.0230 /// teacc 98.95 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 0.0326, -0.0865, -0.0952, ..., -0.0323, -0.1542, -0.0132], + [-0.0926, -0.1267, -0.1245, ..., -0.1087, -0.0645, -0.0018], + [ 0.0719, 0.0534, -0.1232, ..., -0.0965, -0.1204, -0.0674], + ..., + [ 0.0134, 0.0378, -0.0521, ..., -0.0236, -0.1225, 0.0496], + [ 0.0113, 0.1048, 0.0847, ..., -0.0771, 0.0698, -0.1185], + [-0.0452, -0.0065, -0.0306, ..., -0.0295, -0.0581, 0.0470]], + device='cuda:0'), grad: tensor([[ 3.5055e-06, 4.0904e-06, 1.5162e-06, ..., 1.0850e-06, + 1.6959e-06, 1.6578e-07], + [ 2.4494e-07, 6.2585e-07, 2.3749e-07, ..., 1.3411e-07, + 3.6880e-07, -5.2154e-08], + [ 6.6683e-07, -4.1164e-06, 2.0396e-07, ..., 5.6811e-08, + -1.7043e-07, 8.5682e-08], + ..., + [ 2.4531e-06, 3.2131e-06, 3.3341e-07, ..., 6.5193e-09, + 6.1095e-07, 1.1735e-07], + [ 1.1642e-07, 6.4168e-07, 2.5313e-06, ..., 4.0699e-07, + 2.5053e-06, 3.4459e-08], + [-1.2293e-05, -1.0923e-05, 7.8790e-07, ..., 2.5146e-08, + 1.4836e-06, -5.4948e-07]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0284, -0.0209, -0.0086, -0.0024, 0.0094, 0.0019, -0.0191, 0.0028, + 0.0247, 0.0158], device='cuda:0'), grad: tensor([ 1.8433e-05, 2.6859e-06, -8.1956e-06, 1.0431e-05, -2.5705e-06, + -7.8380e-06, -6.7614e-07, 1.4670e-05, 8.2329e-06, -3.5197e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 256.90, cls_loss 0.0031 cls_loss_mapping 0.0071 cls_loss_causal 0.5561 re_mapping 0.0074 re_causal 0.0213 /// teacc 98.99 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 0.0327, -0.0869, -0.0961, ..., -0.0327, -0.1552, -0.0136], + [-0.0928, -0.1264, -0.1247, ..., -0.1093, -0.0646, -0.0009], + [ 0.0729, 0.0534, -0.1259, ..., -0.0979, -0.1198, -0.0684], + ..., + [ 0.0133, 0.0380, -0.0528, ..., -0.0237, -0.1225, 0.0493], + [ 0.0116, 0.1053, 0.0860, ..., -0.0755, 0.0705, -0.1189], + [-0.0450, -0.0066, -0.0314, ..., -0.0296, -0.0586, 0.0471]], + device='cuda:0'), grad: tensor([[ 8.4937e-07, 1.4780e-06, 8.0373e-07, ..., 4.0699e-07, + 1.3392e-06, 5.4017e-07], + [ 4.5821e-07, 1.6363e-06, 5.7742e-07, ..., 1.2387e-07, + 1.3467e-06, 3.6322e-07], + [-3.1162e-06, -3.1237e-06, 2.6543e-07, ..., 1.3039e-07, + -1.7872e-06, 5.4482e-07], + ..., + [-6.4448e-07, -4.6492e-06, 1.7742e-06, ..., 7.4506e-09, + 2.6170e-06, -9.6019e-07], + [ 1.0999e-06, 5.2005e-06, 2.0713e-06, ..., 7.3854e-07, + 4.4145e-06, 1.2070e-06], + [-3.7253e-07, 2.0955e-07, 3.7514e-06, ..., 5.7742e-08, + 6.3963e-06, 1.4259e-06]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0282, -0.0190, -0.0093, -0.0029, 0.0089, 0.0014, -0.0191, 0.0025, + 0.0249, 0.0158], device='cuda:0'), grad: tensor([ 8.2925e-06, 5.7705e-06, -2.5496e-05, -2.2829e-05, 5.2005e-06, + 7.4394e-06, -5.3868e-06, -4.8894e-07, 1.6272e-05, 1.1191e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 257.07, cls_loss 0.0021 cls_loss_mapping 0.0052 cls_loss_causal 0.5403 re_mapping 0.0071 re_causal 0.0215 /// teacc 98.94 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0329, -0.0872, -0.0968, ..., -0.0330, -0.1559, -0.0133], + [-0.0931, -0.1266, -0.1258, ..., -0.1100, -0.0660, -0.0007], + [ 0.0731, 0.0536, -0.1271, ..., -0.0981, -0.1195, -0.0686], + ..., + [ 0.0133, 0.0382, -0.0538, ..., -0.0236, -0.1226, 0.0494], + [ 0.0117, 0.1053, 0.0858, ..., -0.0761, 0.0701, -0.1198], + [-0.0452, -0.0070, -0.0326, ..., -0.0297, -0.0592, 0.0471]], + device='cuda:0'), grad: tensor([[ 7.0781e-08, 4.5914e-07, 2.5239e-07, ..., 1.2852e-07, + 2.2352e-07, 1.7695e-07], + [ 1.2014e-07, 5.2620e-07, 2.8871e-07, ..., 1.9558e-08, + 3.0641e-07, -5.0012e-07], + [ 2.4214e-07, 1.1390e-06, 4.1537e-07, ..., 3.5390e-08, + 4.4331e-07, 5.3924e-07], + ..., + [-7.2550e-07, -3.6806e-06, 3.6415e-07, ..., 0.0000e+00, + 3.8929e-07, -8.0653e-07], + [ 2.6822e-07, 1.6708e-06, 3.0715e-06, ..., 2.7381e-07, + 3.1758e-06, 3.8370e-07], + [-9.4529e-07, -6.3702e-07, 1.3681e-06, ..., 6.5193e-09, + 1.4566e-06, -1.3355e-06]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0283, -0.0194, -0.0084, -0.0028, 0.0089, 0.0014, -0.0190, 0.0027, + 0.0245, 0.0154], device='cuda:0'), grad: tensor([ 7.2736e-07, -1.9558e-07, 2.4661e-06, -9.1866e-06, -1.2510e-05, + 2.4587e-06, 5.4948e-07, -9.2387e-07, 6.4000e-06, 1.0192e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 256.97, cls_loss 0.0024 cls_loss_mapping 0.0059 cls_loss_causal 0.5401 re_mapping 0.0071 re_causal 0.0209 /// teacc 98.96 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0331, -0.0879, -0.0990, ..., -0.0343, -0.1574, -0.0138], + [-0.0934, -0.1268, -0.1260, ..., -0.1104, -0.0660, -0.0004], + [ 0.0734, 0.0538, -0.1281, ..., -0.0983, -0.1196, -0.0689], + ..., + [ 0.0133, 0.0382, -0.0546, ..., -0.0238, -0.1227, 0.0493], + [ 0.0117, 0.1053, 0.0861, ..., -0.0765, 0.0703, -0.1206], + [-0.0451, -0.0067, -0.0334, ..., -0.0301, -0.0597, 0.0467]], + device='cuda:0'), grad: tensor([[ 2.7847e-06, 1.3430e-06, 8.1956e-08, ..., 5.0291e-08, + 7.3574e-08, 2.6058e-06], + [ 2.0303e-07, 2.5947e-06, 1.0896e-07, ..., 3.7253e-09, + 1.0803e-07, -1.0327e-05], + [-3.3155e-07, -2.6524e-06, 1.6391e-07, ..., 3.7253e-09, + 1.4901e-07, 4.6939e-06], + ..., + [ 1.9178e-05, 8.0243e-06, 2.0303e-07, ..., 0.0000e+00, + 1.8347e-07, 2.0221e-05], + [-4.0233e-07, -1.0729e-06, -2.0023e-06, ..., 1.9558e-08, + -1.6028e-06, 4.8801e-07], + [-2.3156e-05, -1.0513e-05, 1.7043e-07, ..., 9.3132e-10, + 1.4808e-07, -2.1279e-05]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0281, -0.0193, -0.0081, -0.0028, 0.0101, 0.0015, -0.0186, 0.0025, + 0.0242, 0.0144], device='cuda:0'), grad: tensor([ 8.6948e-06, -6.3896e-05, 2.3529e-05, 3.4794e-06, 8.7842e-06, + 6.5081e-06, 1.4948e-06, 8.0347e-05, 6.3982e-07, -6.9499e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 256.94, cls_loss 0.0023 cls_loss_mapping 0.0059 cls_loss_causal 0.5761 re_mapping 0.0072 re_causal 0.0222 /// teacc 98.81 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0334, -0.0882, -0.0995, ..., -0.0346, -0.1580, -0.0138], + [-0.0937, -0.1268, -0.1263, ..., -0.1107, -0.0662, -0.0003], + [ 0.0728, 0.0527, -0.1285, ..., -0.0986, -0.1198, -0.0693], + ..., + [ 0.0140, 0.0391, -0.0552, ..., -0.0238, -0.1227, 0.0494], + [ 0.0117, 0.1053, 0.0860, ..., -0.0770, 0.0702, -0.1210], + [-0.0453, -0.0066, -0.0345, ..., -0.0302, -0.0599, 0.0468]], + device='cuda:0'), grad: tensor([[-2.7567e-06, 8.9500e-07, 7.9628e-07, ..., 1.3318e-07, + 8.7731e-07, 1.8626e-08], + [ 4.7777e-07, 1.1232e-06, 1.0096e-06, ..., 5.7742e-08, + 1.1316e-06, 1.8626e-09], + [ 5.8673e-08, 1.4476e-05, 1.3523e-05, ..., 3.8184e-08, + 1.5102e-05, 1.6764e-08], + ..., + [ 1.0803e-07, 5.5507e-06, 5.1893e-06, ..., 1.8626e-09, + 5.7966e-06, 8.0094e-08], + [ 2.6915e-07, 1.3165e-05, 1.2331e-05, ..., 9.4995e-08, + 1.3709e-05, 1.2293e-07], + [ 1.0664e-06, 2.3376e-07, 1.9297e-06, ..., 2.8871e-08, + 2.1383e-06, -7.7486e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0277, -0.0191, -0.0095, -0.0028, 0.0101, 0.0017, -0.0185, 0.0035, + 0.0238, 0.0146], device='cuda:0'), grad: tensor([-4.3124e-05, 1.6361e-05, 1.8179e-05, -1.4150e-04, -4.5002e-06, + 5.8770e-05, 6.1952e-06, 2.5228e-05, 4.6164e-05, 1.8224e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 255.67, cls_loss 0.0029 cls_loss_mapping 0.0066 cls_loss_causal 0.5153 re_mapping 0.0069 re_causal 0.0196 /// teacc 99.02 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 3.3642e-02, -8.8174e-02, -1.0058e-01, ..., -3.4658e-02, + -1.5901e-01, -1.3827e-02], + [-9.4089e-02, -1.2732e-01, -1.2684e-01, ..., -1.1022e-01, + -6.6740e-02, -5.8788e-05], + [ 7.3148e-02, 5.2653e-02, -1.2904e-01, ..., -9.8792e-02, + -1.2040e-01, -6.9431e-02], + ..., + [ 1.3571e-02, 3.9036e-02, -5.6408e-02, ..., -2.3869e-02, + -1.2401e-01, 4.9015e-02], + [ 1.1777e-02, 1.0562e-01, 8.6505e-02, ..., -7.6968e-02, + 7.0496e-02, -1.2166e-01], + [-4.5356e-02, -6.4570e-03, -3.6120e-02, ..., -3.0314e-02, + -6.1014e-02, 4.7254e-02]], device='cuda:0'), grad: tensor([[-1.1735e-06, 1.2852e-06, 1.2834e-06, ..., 6.7335e-07, + 1.0589e-06, 6.7055e-08], + [ 3.8464e-07, 3.7309e-06, 4.0885e-07, ..., 1.4435e-07, + 3.8184e-07, -1.1638e-05], + [-1.8468e-06, -1.0608e-06, 2.5146e-07, ..., 4.4703e-08, + -6.0257e-07, 6.6049e-06], + ..., + [-9.3319e-07, -4.5896e-06, 1.0859e-06, ..., 6.5193e-09, + 7.1619e-07, 2.3656e-06], + [ 9.8906e-07, 3.5651e-06, 1.1273e-05, ..., 2.2352e-06, + 7.0482e-06, 9.6858e-08], + [ 1.1167e-06, 2.5146e-07, 1.9576e-06, ..., 1.1548e-07, + 1.0347e-06, 9.2201e-08]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0293, -0.0199, -0.0101, -0.0015, 0.0102, 0.0014, -0.0192, 0.0035, + 0.0240, 0.0144], device='cuda:0'), grad: tensor([-1.6503e-06, -5.6028e-05, 2.7627e-05, 1.2852e-05, -3.1292e-05, + -2.0713e-05, -3.2447e-06, 4.2647e-05, 2.0862e-05, 8.8662e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 256.98, cls_loss 0.0021 cls_loss_mapping 0.0045 cls_loss_causal 0.5277 re_mapping 0.0069 re_causal 0.0200 /// teacc 99.03 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0345, -0.0885, -0.1015, ..., -0.0355, -0.1596, -0.0142], + [-0.0947, -0.1283, -0.1272, ..., -0.1105, -0.0671, -0.0008], + [ 0.0739, 0.0531, -0.1294, ..., -0.0994, -0.1206, -0.0694], + ..., + [ 0.0132, 0.0391, -0.0574, ..., -0.0239, -0.1241, 0.0495], + [ 0.0115, 0.1055, 0.0859, ..., -0.0779, 0.0701, -0.1222], + [-0.0453, -0.0056, -0.0364, ..., -0.0304, -0.0615, 0.0483]], + device='cuda:0'), grad: tensor([[ 4.6566e-08, 5.5227e-07, 5.1782e-07, ..., 3.1572e-07, + 5.9791e-07, -2.1607e-07], + [ 3.5577e-07, 1.5274e-07, 1.8161e-07, ..., 8.3819e-09, + 2.2911e-07, -2.1011e-05], + [ 1.5553e-07, 6.5193e-09, 9.3319e-07, ..., 7.8231e-08, + 1.2210e-06, 9.2983e-06], + ..., + [ 1.0245e-08, -1.2387e-07, 1.3970e-07, ..., 0.0000e+00, + 1.7323e-07, 8.6650e-06], + [-1.2945e-06, -1.3700e-06, -1.7509e-07, ..., 4.3772e-08, + -6.9756e-07, 4.2375e-07], + [ 9.9652e-08, 7.2643e-08, 2.4401e-07, ..., 2.7940e-09, + 2.6729e-07, 2.6543e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0295, -0.0207, -0.0095, -0.0017, 0.0101, 0.0018, -0.0188, 0.0036, + 0.0234, 0.0147], device='cuda:0'), grad: tensor([-1.8440e-07, -1.0800e-04, 4.4465e-05, -3.1684e-06, 9.3356e-06, + 4.8280e-06, 4.2059e-06, 4.5836e-05, 7.7765e-07, 1.9092e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 257.16, cls_loss 0.0020 cls_loss_mapping 0.0055 cls_loss_causal 0.5517 re_mapping 0.0066 re_causal 0.0202 /// teacc 98.93 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0347, -0.0888, -0.1026, ..., -0.0363, -0.1602, -0.0146], + [-0.0952, -0.1286, -0.1278, ..., -0.1110, -0.0676, -0.0003], + [ 0.0741, 0.0532, -0.1303, ..., -0.1000, -0.1211, -0.0700], + ..., + [ 0.0133, 0.0391, -0.0585, ..., -0.0241, -0.1242, 0.0492], + [ 0.0115, 0.1048, 0.0844, ..., -0.0806, 0.0686, -0.1229], + [-0.0453, -0.0055, -0.0367, ..., -0.0306, -0.0616, 0.0486]], + device='cuda:0'), grad: tensor([[ 5.0571e-07, 2.2817e-07, 4.0978e-07, ..., 2.0955e-07, + 3.1199e-07, 4.9267e-07], + [ 3.6042e-07, 9.4064e-08, 1.3784e-07, ..., 1.9558e-08, + 1.2107e-07, -5.5768e-06], + [-3.7253e-08, 1.9558e-08, 1.1921e-07, ..., 3.2596e-08, + 1.6671e-07, 1.0943e-06], + ..., + [-4.3139e-06, -2.6543e-07, 3.2783e-07, ..., 4.6566e-09, + 1.3690e-07, -6.1188e-07], + [-1.7416e-07, -5.9605e-08, 1.6391e-07, ..., 2.1886e-07, + -2.5891e-07, 6.1467e-07], + [ 3.0119e-06, 2.3469e-07, 2.6766e-06, ..., 2.2352e-08, + 6.8825e-07, 2.7344e-06]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0295, -0.0204, -0.0097, -0.0019, 0.0100, 0.0022, -0.0173, 0.0032, + 0.0223, 0.0148], device='cuda:0'), grad: tensor([ 2.2855e-06, -1.7434e-05, 2.9206e-06, 7.0669e-06, 3.6731e-06, + -1.0647e-05, 4.8988e-07, -3.9078e-06, 1.9148e-06, 1.3612e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 257.00, cls_loss 0.0025 cls_loss_mapping 0.0055 cls_loss_causal 0.5375 re_mapping 0.0066 re_causal 0.0201 /// teacc 99.04 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0349, -0.0890, -0.1034, ..., -0.0370, -0.1609, -0.0147], + [-0.0958, -0.1292, -0.1298, ..., -0.1128, -0.0687, -0.0005], + [ 0.0740, 0.0526, -0.1314, ..., -0.1004, -0.1215, -0.0703], + ..., + [ 0.0134, 0.0397, -0.0595, ..., -0.0244, -0.1243, 0.0493], + [ 0.0115, 0.1044, 0.0820, ..., -0.0833, 0.0673, -0.1242], + [-0.0449, -0.0050, -0.0372, ..., -0.0308, -0.0608, 0.0498]], + device='cuda:0'), grad: tensor([[ 1.4054e-06, 6.5472e-07, 5.3458e-07, ..., 2.4214e-07, + 5.6718e-07, 1.1921e-06], + [ 1.4063e-06, 9.1735e-07, 6.8732e-07, ..., 6.5193e-09, + 8.3633e-07, 1.8440e-07], + [ 1.7844e-06, 1.8533e-06, 1.1148e-06, ..., 1.1176e-08, + 1.4100e-06, 1.3616e-06], + ..., + [-8.0168e-06, -2.7940e-06, 5.2992e-07, ..., 0.0000e+00, + 7.2829e-07, -6.8098e-06], + [-3.2429e-06, -7.9647e-06, -9.7677e-06, ..., 2.2352e-08, + -1.1154e-05, 2.2817e-07], + [ 9.1828e-07, 1.3802e-06, 1.1893e-06, ..., 4.6566e-09, + 1.4324e-06, 6.6962e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0294, -0.0203, -0.0106, -0.0017, 0.0095, 0.0038, -0.0182, 0.0035, + 0.0207, 0.0159], device='cuda:0'), grad: tensor([ 6.8247e-06, 5.4128e-06, 9.7081e-06, 1.3880e-05, -9.4473e-05, + 1.2025e-05, 5.3458e-06, -2.0817e-05, -1.8567e-05, 8.0585e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 257.05, cls_loss 0.0017 cls_loss_mapping 0.0046 cls_loss_causal 0.5280 re_mapping 0.0065 re_causal 0.0204 /// teacc 99.04 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0349, -0.0894, -0.1043, ..., -0.0376, -0.1619, -0.0152], + [-0.0963, -0.1293, -0.1305, ..., -0.1134, -0.0692, -0.0002], + [ 0.0741, 0.0526, -0.1324, ..., -0.1007, -0.1220, -0.0709], + ..., + [ 0.0135, 0.0397, -0.0605, ..., -0.0246, -0.1244, 0.0493], + [ 0.0116, 0.1044, 0.0817, ..., -0.0837, 0.0670, -0.1245], + [-0.0449, -0.0049, -0.0383, ..., -0.0308, -0.0606, 0.0501]], + device='cuda:0'), grad: tensor([[-3.3766e-05, 4.6659e-07, -5.9195e-06, ..., 1.5460e-07, + -4.0412e-05, 4.7497e-08], + [ 1.9409e-06, 3.9153e-06, 2.8852e-06, ..., 1.0543e-06, + 2.5816e-06, -2.2203e-06], + [ 4.3251e-06, 1.3802e-06, 1.8859e-06, ..., 4.3679e-07, + 6.1803e-06, 3.9488e-07], + ..., + [ 3.4366e-07, 3.6787e-07, 5.4110e-07, ..., 1.7416e-07, + 1.1493e-06, 3.6974e-07], + [ 9.3654e-06, 1.5646e-07, -5.1260e-06, ..., -2.4121e-06, + 1.0036e-05, 1.5106e-06], + [-1.6931e-06, -1.1839e-05, -2.9244e-07, ..., 5.4948e-08, + -4.8913e-06, -1.0794e-06]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0293, -0.0201, -0.0109, -0.0014, 0.0098, 0.0036, -0.0176, 0.0033, + 0.0204, 0.0158], device='cuda:0'), grad: tensor([-1.1760e-04, 1.0049e-06, 1.7658e-05, 5.9634e-05, 1.9502e-06, + 7.4618e-06, 4.6529e-06, 3.7067e-06, 3.8594e-05, -1.7092e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 256.29, cls_loss 0.0024 cls_loss_mapping 0.0054 cls_loss_causal 0.5286 re_mapping 0.0064 re_causal 0.0188 /// teacc 98.96 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0352, -0.0895, -0.1050, ..., -0.0379, -0.1619, -0.0148], + [-0.0971, -0.1295, -0.1312, ..., -0.1132, -0.0697, 0.0005], + [ 0.0741, 0.0526, -0.1338, ..., -0.1017, -0.1225, -0.0713], + ..., + [ 0.0136, 0.0397, -0.0616, ..., -0.0245, -0.1245, 0.0488], + [ 0.0118, 0.1043, 0.0815, ..., -0.0843, 0.0669, -0.1249], + [-0.0452, -0.0046, -0.0389, ..., -0.0307, -0.0605, 0.0505]], + device='cuda:0'), grad: tensor([[ 1.9614e-06, 7.9125e-06, 8.9332e-06, ..., 2.6934e-06, + 6.7577e-06, 5.1595e-07], + [ 7.3314e-06, 1.0356e-05, 4.0233e-07, ..., 1.7509e-07, + 2.6077e-07, 3.5819e-06], + [-1.3554e-04, -1.8990e-04, 3.0920e-06, ..., 1.9316e-06, + 1.3839e-06, -7.7784e-05], + ..., + [ 1.1742e-04, 1.6725e-04, 1.1269e-07, ..., 1.3970e-08, + 1.0338e-07, 6.7472e-05], + [ 5.5470e-06, 9.5218e-06, 1.8582e-05, ..., 2.6543e-07, + 1.9029e-05, 1.2806e-06], + [ 5.3756e-06, 7.4878e-06, 4.8522e-07, ..., 4.1910e-08, + 4.5728e-07, 2.7362e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0295, -0.0198, -0.0112, -0.0014, 0.0101, 0.0036, -0.0176, 0.0030, + 0.0203, 0.0158], device='cuda:0'), grad: tensor([ 2.6122e-05, 1.9804e-05, -3.4142e-04, -5.6803e-05, 9.1419e-06, + 1.8418e-05, -4.1127e-05, 3.1137e-04, 3.9518e-05, 1.4499e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 256.92, cls_loss 0.0018 cls_loss_mapping 0.0036 cls_loss_causal 0.5055 re_mapping 0.0064 re_causal 0.0193 /// teacc 98.93 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 0.0356, -0.0898, -0.1061, ..., -0.0379, -0.1627, -0.0155], + [-0.0976, -0.1296, -0.1315, ..., -0.1133, -0.0699, 0.0009], + [ 0.0748, 0.0530, -0.1346, ..., -0.1020, -0.1227, -0.0713], + ..., + [ 0.0134, 0.0398, -0.0624, ..., -0.0247, -0.1244, 0.0487], + [ 0.0118, 0.1043, 0.0816, ..., -0.0843, 0.0670, -0.1255], + [-0.0452, -0.0048, -0.0392, ..., -0.0308, -0.0608, 0.0506]], + device='cuda:0'), grad: tensor([[ 1.6019e-07, 8.2422e-07, 2.8498e-07, ..., 4.2841e-07, + 2.2165e-07, -3.8054e-06], + [ 1.9930e-07, 1.0943e-06, 4.5449e-07, ..., 5.5041e-07, + 3.7998e-07, 7.9907e-07], + [-3.1944e-07, 5.1409e-07, 1.1735e-07, ..., 8.8662e-07, + 1.4622e-07, 2.6077e-07], + ..., + [ 7.9721e-07, 1.8105e-06, 1.7043e-07, ..., 1.0617e-07, + 1.7136e-07, 4.7032e-07], + [ 3.1758e-07, 5.4799e-06, 2.1383e-05, ..., 1.5926e-07, + 1.1884e-05, 3.5856e-07], + [-1.7760e-06, -4.0308e-06, 8.8476e-07, ..., 4.7404e-07, + 5.3924e-07, -9.7789e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0294, -0.0196, -0.0108, -0.0012, 0.0116, 0.0034, -0.0176, 0.0027, + 0.0202, 0.0146], device='cuda:0'), grad: tensor([-3.0503e-05, 1.1660e-05, 4.5970e-06, 6.4746e-06, -4.7088e-05, + -2.9683e-05, 5.2452e-05, 5.5023e-06, 2.9013e-05, -2.3916e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 256.64, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5423 re_mapping 0.0066 re_causal 0.0197 /// teacc 98.97 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0359, -0.0902, -0.1067, ..., -0.0395, -0.1634, -0.0155], + [-0.0979, -0.1298, -0.1330, ..., -0.1136, -0.0715, 0.0009], + [ 0.0753, 0.0531, -0.1353, ..., -0.1021, -0.1231, -0.0717], + ..., + [ 0.0132, 0.0398, -0.0634, ..., -0.0249, -0.1246, 0.0487], + [ 0.0118, 0.1045, 0.0817, ..., -0.0843, 0.0673, -0.1260], + [-0.0453, -0.0027, -0.0404, ..., -0.0314, -0.0612, 0.0519]], + device='cuda:0'), grad: tensor([[ 1.5460e-07, 3.0082e-07, 2.0768e-07, ..., 6.5193e-08, + 1.6857e-07, 5.2154e-08], + [ 8.4378e-07, 1.7285e-06, 5.5321e-07, ..., -6.9849e-08, + 6.6310e-07, -3.1292e-07], + [-8.4192e-07, -4.6566e-09, 2.4494e-07, ..., 2.5146e-08, + 3.5763e-07, 6.7055e-08], + ..., + [ 4.0568e-06, 9.2536e-06, 2.0154e-06, ..., 2.2352e-08, + 2.6040e-06, 2.1718e-06], + [-7.0743e-06, -1.6347e-05, -4.2282e-06, ..., 8.6613e-08, + -5.9605e-06, 9.8720e-08], + [ 1.7602e-07, -6.0722e-07, 8.6427e-07, ..., 2.7008e-08, + 9.1922e-07, -2.4065e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0292, -0.0196, -0.0109, -0.0020, 0.0098, 0.0050, -0.0182, 0.0025, + 0.0202, 0.0160], device='cuda:0'), grad: tensor([ 5.7928e-07, 2.4643e-06, -9.6858e-07, 1.1295e-05, -1.7136e-07, + -2.8521e-05, 2.5347e-05, 1.8656e-05, -2.6822e-05, -1.9185e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 257.01, cls_loss 0.0031 cls_loss_mapping 0.0058 cls_loss_causal 0.5374 re_mapping 0.0065 re_causal 0.0189 /// teacc 99.02 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 3.6097e-02, -9.0778e-02, -1.0780e-01, ..., -4.0214e-02, + -1.6465e-01, -1.5980e-02], + [-9.8934e-02, -1.3101e-01, -1.3551e-01, ..., -1.1504e-01, + -7.4972e-02, 6.9601e-05], + [ 7.5909e-02, 5.3494e-02, -1.3598e-01, ..., -1.0218e-01, + -1.2304e-01, -7.2211e-02], + ..., + [ 1.3141e-02, 3.9604e-02, -6.4776e-02, ..., -2.5000e-02, + -1.2482e-01, 4.8568e-02], + [ 1.2561e-02, 1.0596e-01, 8.3668e-02, ..., -8.4157e-02, + 6.8812e-02, -1.2737e-01], + [-4.5115e-02, -2.3898e-03, -4.0734e-02, ..., -3.1698e-02, + -5.8896e-02, 5.4345e-02]], device='cuda:0'), grad: tensor([[-1.0338e-07, 4.4703e-08, 1.5739e-07, ..., 4.6566e-09, + 1.9278e-07, 9.3132e-10], + [ 1.3132e-07, 3.1199e-07, 1.0468e-06, ..., -1.6764e-08, + 1.3895e-06, -1.8161e-07], + [-1.8394e-06, -1.8794e-06, 5.6624e-07, ..., 2.7940e-09, + 7.4226e-07, 5.4948e-08], + ..., + [ 1.4324e-06, 1.0375e-06, 1.1344e-06, ..., 6.5193e-09, + 1.4799e-06, -3.0734e-08], + [ 6.3330e-08, 2.3283e-08, 4.6566e-06, ..., 9.3132e-10, + 5.4576e-06, 7.4506e-09], + [ 1.5926e-07, 2.4680e-07, 1.1753e-06, ..., 5.5879e-09, + 1.4743e-06, 1.1362e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0292, -0.0221, -0.0094, -0.0028, 0.0097, 0.0038, -0.0180, 0.0018, + 0.0221, 0.0177], device='cuda:0'), grad: tensor([ 1.6019e-07, 3.3975e-06, -3.2447e-06, -1.0502e-04, 1.0803e-07, + 8.0764e-05, 2.1700e-07, 6.6459e-06, 1.2673e-05, 4.1872e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 256.80, cls_loss 0.0024 cls_loss_mapping 0.0051 cls_loss_causal 0.5343 re_mapping 0.0067 re_causal 0.0198 /// teacc 98.88 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 0.0363, -0.0913, -0.1092, ..., -0.0413, -0.1655, -0.0162], + [-0.1008, -0.1322, -0.1371, ..., -0.1161, -0.0768, 0.0003], + [ 0.0761, 0.0533, -0.1371, ..., -0.1029, -0.1235, -0.0727], + ..., + [ 0.0137, 0.0401, -0.0659, ..., -0.0251, -0.1249, 0.0487], + [ 0.0129, 0.1065, 0.0841, ..., -0.0840, 0.0694, -0.1278], + [-0.0455, -0.0028, -0.0418, ..., -0.0320, -0.0592, 0.0540]], + device='cuda:0'), grad: tensor([[ 6.9197e-07, 2.5779e-06, 1.5851e-06, ..., 1.1353e-06, + 1.2834e-06, 1.2387e-07], + [ 1.7192e-06, 2.0321e-06, 1.9465e-07, ..., 1.2666e-07, + 1.5181e-07, -1.8626e-07], + [ 4.6223e-05, 7.8604e-06, 5.8021e-07, ..., 2.6077e-07, + 1.2154e-06, 2.6058e-06], + ..., + [-1.1347e-05, -3.3319e-05, 4.7218e-07, ..., 1.2107e-08, + -1.3504e-07, -1.3784e-05], + [ 1.6969e-06, 3.9004e-06, 2.1663e-06, ..., 1.6075e-06, + 1.7593e-06, 2.5425e-07], + [ 8.7097e-06, 4.7535e-06, 4.5169e-07, ..., 1.4249e-07, + 1.7136e-07, 6.0238e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0291, -0.0227, -0.0100, -0.0026, 0.0097, 0.0035, -0.0171, 0.0023, + 0.0232, 0.0172], device='cuda:0'), grad: tensor([ 7.1265e-06, 5.0627e-06, 1.9097e-04, 4.6045e-05, -2.7943e-04, + 1.6868e-05, -1.6823e-05, -3.3617e-05, 1.3895e-05, 4.9442e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 257.26, cls_loss 0.0018 cls_loss_mapping 0.0047 cls_loss_causal 0.5349 re_mapping 0.0067 re_causal 0.0208 /// teacc 98.98 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 0.0361, -0.0916, -0.1099, ..., -0.0417, -0.1659, -0.0163], + [-0.1011, -0.1323, -0.1375, ..., -0.1161, -0.0771, 0.0009], + [ 0.0762, 0.0533, -0.1377, ..., -0.1030, -0.1238, -0.0734], + ..., + [ 0.0137, 0.0402, -0.0669, ..., -0.0252, -0.1249, 0.0484], + [ 0.0131, 0.1067, 0.0840, ..., -0.0840, 0.0695, -0.1281], + [-0.0450, -0.0028, -0.0415, ..., -0.0320, -0.0595, 0.0542]], + device='cuda:0'), grad: tensor([[-4.9919e-06, -4.1723e-07, 1.3039e-07, ..., 1.3970e-08, + 1.2852e-07, -4.4610e-07], + [ 3.8557e-07, -1.5169e-05, 3.1944e-07, ..., 1.8626e-09, + 3.4366e-07, -7.2345e-06], + [-4.9081e-07, 4.7870e-06, 3.4086e-07, ..., 4.6566e-09, + 5.0012e-07, 2.4922e-06], + ..., + [ 5.2154e-07, 3.1739e-06, 1.1548e-07, ..., 0.0000e+00, + 1.5926e-07, 1.5628e-06], + [ 1.2945e-07, 6.5751e-07, -2.8778e-06, ..., 9.3132e-09, + -2.5742e-06, 1.1660e-06], + [ 1.2964e-06, 9.4902e-07, 3.1758e-07, ..., 1.8626e-09, + 3.7905e-07, 3.4552e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0289, -0.0222, -0.0103, -0.0026, 0.0095, 0.0039, -0.0173, 0.0021, + 0.0232, 0.0173], device='cuda:0'), grad: tensor([-1.5303e-05, -1.1456e-04, 4.0293e-05, 1.4221e-06, 1.8552e-06, + 1.1958e-05, 2.2829e-05, 2.4483e-05, 1.8507e-05, 8.6799e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 255.44, cls_loss 0.0017 cls_loss_mapping 0.0040 cls_loss_causal 0.5325 re_mapping 0.0060 re_causal 0.0194 /// teacc 98.98 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 0.0370, -0.0923, -0.1110, ..., -0.0422, -0.1668, -0.0160], + [-0.1021, -0.1326, -0.1379, ..., -0.1164, -0.0774, 0.0020], + [ 0.0766, 0.0536, -0.1381, ..., -0.1030, -0.1239, -0.0736], + ..., + [ 0.0150, 0.0412, -0.0675, ..., -0.0252, -0.1250, 0.0478], + [ 0.0132, 0.1067, 0.0839, ..., -0.0842, 0.0694, -0.1285], + [-0.0472, -0.0041, -0.0421, ..., -0.0321, -0.0599, 0.0536]], + device='cuda:0'), grad: tensor([[-3.9861e-07, 2.2165e-07, 5.6811e-08, ..., 3.9116e-08, + 6.4261e-08, 3.2596e-08], + [ 9.0338e-07, 2.5425e-06, 5.2154e-08, ..., -7.4506e-09, + 1.2759e-07, 9.6019e-07], + [ 5.1223e-08, 2.7195e-07, 6.3330e-08, ..., 7.4506e-09, + 1.0990e-07, 2.9430e-07], + ..., + [-2.2314e-06, -6.9849e-06, 5.9605e-08, ..., 2.7940e-09, + -3.7253e-09, -3.7886e-06], + [ 2.6915e-07, 3.0827e-07, 4.0047e-08, ..., 6.8918e-08, + 1.2014e-07, 3.5297e-07], + [ 7.0315e-07, 2.6915e-06, 1.7695e-08, ..., 2.7940e-09, + 1.3877e-07, 1.2843e-06]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0290, -0.0215, -0.0100, -0.0027, 0.0095, 0.0039, -0.0169, 0.0018, + 0.0231, 0.0167], device='cuda:0'), grad: tensor([ 1.2442e-06, 4.0345e-06, 1.6605e-06, 2.6189e-06, 3.3639e-06, + 6.1579e-06, 4.7497e-07, -1.1720e-05, 3.6694e-06, -1.1519e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 256.92, cls_loss 0.0020 cls_loss_mapping 0.0038 cls_loss_causal 0.5651 re_mapping 0.0062 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 0.0369, -0.0931, -0.1142, ..., -0.0423, -0.1698, -0.0173], + [-0.1029, -0.1330, -0.1387, ..., -0.1174, -0.0780, 0.0026], + [ 0.0769, 0.0539, -0.1387, ..., -0.1031, -0.1241, -0.0739], + ..., + [ 0.0150, 0.0413, -0.0683, ..., -0.0253, -0.1252, 0.0474], + [ 0.0134, 0.1067, 0.0838, ..., -0.0844, 0.0695, -0.1292], + [-0.0474, -0.0038, -0.0430, ..., -0.0322, -0.0602, 0.0539]], + device='cuda:0'), grad: tensor([[-1.1520e-06, 3.8696e-07, 3.9069e-07, ..., 4.2003e-07, + 4.9453e-07, 9.7789e-09], + [ 1.2899e-07, 3.2317e-07, 3.5344e-07, ..., 3.9535e-07, + 6.3051e-07, -8.2003e-07], + [-6.6310e-07, -2.4214e-07, 1.8813e-07, ..., 1.3877e-07, + 3.8324e-07, 5.7742e-08], + ..., + [ 3.3947e-07, 1.5832e-07, 1.0990e-07, ..., 6.9849e-09, + 3.0128e-07, 9.0338e-08], + [-5.4017e-08, -2.1514e-07, 8.6147e-08, ..., 3.9348e-07, + 1.0710e-08, 7.5437e-08], + [ 5.4203e-07, -1.6484e-07, 4.4145e-07, ..., 6.5193e-08, + 5.9698e-07, 1.3737e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0280, -0.0211, -0.0098, -0.0020, 0.0092, 0.0039, -0.0169, 0.0015, + 0.0229, 0.0169], device='cuda:0'), grad: tensor([-2.6692e-06, -1.3458e-07, -1.4137e-06, -2.5015e-06, -2.6338e-06, + 4.2934e-07, -1.2722e-06, 4.7460e-06, 1.3094e-06, 4.1313e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 256.99, cls_loss 0.0024 cls_loss_mapping 0.0042 cls_loss_causal 0.5503 re_mapping 0.0062 re_causal 0.0184 /// teacc 99.02 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0374, -0.0937, -0.1154, ..., -0.0433, -0.1706, -0.0170], + [-0.1038, -0.1333, -0.1395, ..., -0.1177, -0.0785, 0.0027], + [ 0.0773, 0.0542, -0.1394, ..., -0.1026, -0.1242, -0.0743], + ..., + [ 0.0154, 0.0417, -0.0694, ..., -0.0260, -0.1250, 0.0479], + [ 0.0131, 0.1065, 0.0836, ..., -0.0845, 0.0694, -0.1306], + [-0.0476, -0.0039, -0.0436, ..., -0.0325, -0.0604, 0.0534]], + device='cuda:0'), grad: tensor([[ 4.5747e-06, 9.8422e-06, 6.5342e-06, ..., 2.3982e-07, + 9.2164e-06, 6.4187e-06], + [ 1.3888e-05, 1.8701e-05, 1.9789e-05, ..., 4.7032e-08, + 1.7419e-05, -9.2566e-05], + [ 2.0768e-06, -1.6987e-05, 1.0416e-05, ..., 2.6543e-07, + 6.0014e-06, 5.2862e-06], + ..., + [ 1.4044e-05, 2.5868e-05, 3.4552e-06, ..., -5.6345e-08, + 5.1074e-06, 9.5516e-06], + [-5.5730e-05, -1.2088e-04, -7.0930e-05, ..., 5.8208e-08, + -9.7632e-05, -4.9034e-07], + [ 2.7612e-05, 6.8605e-05, 2.7582e-05, ..., 2.2352e-08, + 4.4644e-05, 1.4238e-05]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0276, -0.0211, -0.0092, -0.0022, 0.0096, 0.0042, -0.0169, 0.0016, + 0.0224, 0.0165], device='cuda:0'), grad: tensor([ 3.7342e-05, -1.3041e-04, 8.0645e-05, 2.2128e-05, -9.7930e-05, + 6.7890e-05, 6.3121e-05, 7.8678e-05, -2.2733e-04, 1.0586e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 257.34, cls_loss 0.0017 cls_loss_mapping 0.0044 cls_loss_causal 0.5315 re_mapping 0.0061 re_causal 0.0192 /// teacc 98.99 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0378, -0.0937, -0.1157, ..., -0.0434, -0.1709, -0.0169], + [-0.1041, -0.1336, -0.1400, ..., -0.1178, -0.0788, 0.0029], + [ 0.0774, 0.0542, -0.1406, ..., -0.1029, -0.1245, -0.0744], + ..., + [ 0.0154, 0.0416, -0.0708, ..., -0.0261, -0.1252, 0.0478], + [ 0.0135, 0.1069, 0.0841, ..., -0.0845, 0.0698, -0.1311], + [-0.0479, -0.0037, -0.0448, ..., -0.0326, -0.0610, 0.0535]], + device='cuda:0'), grad: tensor([[ 1.1146e-05, 1.1556e-05, 1.3739e-05, ..., 5.1223e-09, + 1.8179e-05, 1.4948e-07], + [ 7.8883e-07, 1.7788e-06, 1.1008e-06, ..., 1.3970e-09, + 1.4221e-06, -3.3192e-06], + [ 1.6792e-06, 1.7192e-06, 2.4810e-06, ..., 2.7940e-09, + 3.2391e-06, 4.7637e-07], + ..., + [ 3.9255e-07, -1.1306e-06, 6.3796e-07, ..., 0.0000e+00, + 7.1200e-07, -8.0373e-07], + [-2.6688e-05, -2.6658e-05, -2.9385e-05, ..., 2.8871e-08, + -4.0770e-05, 1.6205e-07], + [ 2.4848e-06, 2.1514e-06, 3.3118e-06, ..., 9.3132e-10, + 4.2059e-06, 6.3190e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0278, -0.0211, -0.0092, -0.0020, 0.0097, 0.0041, -0.0173, 0.0012, + 0.0226, 0.0165], device='cuda:0'), grad: tensor([ 5.7191e-05, -1.6764e-08, 1.2390e-05, 3.2932e-05, -9.4175e-06, + 7.2159e-06, 6.0275e-06, 8.1677e-07, -1.3137e-04, 2.4050e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 257.03, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5269 re_mapping 0.0060 re_causal 0.0190 /// teacc 99.00 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0382, -0.0940, -0.1162, ..., -0.0435, -0.1712, -0.0169], + [-0.1045, -0.1337, -0.1402, ..., -0.1179, -0.0789, 0.0034], + [ 0.0776, 0.0543, -0.1419, ..., -0.1032, -0.1249, -0.0748], + ..., + [ 0.0152, 0.0415, -0.0744, ..., -0.0263, -0.1256, 0.0475], + [ 0.0138, 0.1071, 0.0840, ..., -0.0846, 0.0699, -0.1316], + [-0.0480, -0.0038, -0.0453, ..., -0.0325, -0.0613, 0.0535]], + device='cuda:0'), grad: tensor([[-4.6864e-06, -3.2000e-06, 1.7229e-07, ..., 7.3574e-08, + 1.6065e-07, -9.3132e-09], + [ 2.1514e-07, 2.4633e-07, 1.3225e-07, ..., 1.5367e-08, + 1.5926e-07, -1.4761e-06], + [ 7.0408e-07, 6.0722e-07, 4.2329e-07, ..., 2.2817e-08, + 4.5542e-07, 2.0396e-07], + ..., + [-2.3236e-07, -6.4867e-07, 2.0163e-07, ..., -1.2107e-08, + 2.1188e-07, -4.3306e-08], + [ 4.2003e-07, 1.7649e-07, 1.1409e-06, ..., 1.3411e-07, + 8.9174e-07, 4.2794e-07], + [ 2.7400e-06, 2.3730e-06, 1.2806e-06, ..., 8.3819e-09, + 1.1260e-06, 5.0012e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0282, -0.0208, -0.0092, -0.0020, 0.0098, 0.0041, -0.0168, 0.0008, + 0.0226, 0.0163], device='cuda:0'), grad: tensor([-1.5810e-05, -1.9539e-06, 3.1628e-06, 8.5831e-06, 2.5891e-07, + -1.2860e-05, 1.3011e-06, 4.1537e-07, 3.9749e-06, 1.2897e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 257.02, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5066 re_mapping 0.0065 re_causal 0.0185 /// teacc 98.90 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0385, -0.0943, -0.1166, ..., -0.0437, -0.1716, -0.0171], + [-0.1046, -0.1338, -0.1405, ..., -0.1179, -0.0791, 0.0038], + [ 0.0760, 0.0521, -0.1478, ..., -0.1068, -0.1279, -0.0765], + ..., + [ 0.0152, 0.0416, -0.0758, ..., -0.0263, -0.1266, 0.0472], + [ 0.0151, 0.1079, 0.0842, ..., -0.0839, 0.0702, -0.1333], + [-0.0480, -0.0036, -0.0477, ..., -0.0326, -0.0621, 0.0531]], + device='cuda:0'), grad: tensor([[ 6.8080e-07, 1.4929e-06, 1.6568e-06, ..., 6.5193e-08, + 6.4215e-07, 7.6368e-08], + [ 3.3528e-07, 1.5069e-06, 7.8324e-07, ..., 8.3819e-09, + 5.0524e-07, 7.4971e-08], + [-2.3339e-06, -4.8727e-06, 7.1153e-07, ..., 3.7253e-09, + 2.8266e-07, 2.3888e-07], + ..., + [ 4.1444e-08, -2.2501e-06, 4.2282e-07, ..., 0.0000e+00, + 2.2165e-07, -8.7498e-07], + [ 8.8662e-07, 1.9595e-06, 6.3181e-05, ..., 3.0734e-08, + 2.0206e-05, 1.6764e-07], + [ 1.3918e-05, 2.4028e-07, -8.2552e-05, ..., 3.2596e-09, + -2.5660e-05, -1.7323e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0283, -0.0214, -0.0096, -0.0015, 0.0099, 0.0063, -0.0178, 0.0005, + 0.0229, 0.0159], device='cuda:0'), grad: tensor([ 2.7511e-06, 5.4240e-06, -6.6087e-06, 2.5049e-05, -1.5950e-04, + 2.2024e-05, 3.6694e-06, -3.0529e-06, 1.6773e-04, -5.7220e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 257.02, cls_loss 0.0023 cls_loss_mapping 0.0052 cls_loss_causal 0.5453 re_mapping 0.0068 re_causal 0.0195 /// teacc 98.94 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0384, -0.0950, -0.1177, ..., -0.0444, -0.1723, -0.0178], + [-0.1050, -0.1340, -0.1409, ..., -0.1181, -0.0792, 0.0040], + [ 0.0765, 0.0523, -0.1483, ..., -0.1072, -0.1283, -0.0769], + ..., + [ 0.0153, 0.0416, -0.0769, ..., -0.0260, -0.1268, 0.0460], + [ 0.0152, 0.1083, 0.0851, ..., -0.0840, 0.0708, -0.1348], + [-0.0480, -0.0036, -0.0484, ..., -0.0327, -0.0623, 0.0540]], + device='cuda:0'), grad: tensor([[-3.1255e-06, -1.8347e-07, 1.4063e-07, ..., 1.0710e-08, + 6.1560e-07, -2.9802e-08], + [ 6.7055e-08, 2.3749e-08, 4.7870e-07, ..., -4.3306e-08, + 2.5127e-06, -1.2591e-06], + [ 1.2694e-06, 8.9407e-08, 3.0827e-06, ..., 1.3504e-08, + 2.0057e-05, 2.9663e-07], + ..., + [ 4.9360e-08, 7.6368e-08, 1.8533e-07, ..., 1.5367e-08, + 3.6741e-07, 4.9826e-07], + [ 3.9209e-07, -3.4925e-08, 1.9427e-06, ..., 3.0268e-08, + 1.7090e-06, 1.8720e-07], + [ 1.0300e-06, -1.4482e-07, 2.4727e-07, ..., 2.3283e-09, + 3.0361e-07, -3.2596e-08]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0276, -0.0214, -0.0094, -0.0010, 0.0097, 0.0062, -0.0175, -0.0005, + 0.0232, 0.0164], device='cuda:0'), grad: tensor([-1.0930e-05, 4.8988e-06, 7.3969e-05, -8.6010e-05, 4.8149e-07, + 2.5071e-06, 9.8906e-07, 2.6934e-06, 6.5230e-06, 4.8392e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 257.32, cls_loss 0.0022 cls_loss_mapping 0.0042 cls_loss_causal 0.4982 re_mapping 0.0065 re_causal 0.0185 /// teacc 98.85 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0388, -0.0957, -0.1190, ..., -0.0459, -0.1733, -0.0187], + [-0.1065, -0.1343, -0.1413, ..., -0.1184, -0.0792, 0.0047], + [ 0.0767, 0.0522, -0.1490, ..., -0.1079, -0.1293, -0.0771], + ..., + [ 0.0158, 0.0422, -0.0774, ..., -0.0255, -0.1267, 0.0460], + [ 0.0149, 0.1081, 0.0848, ..., -0.0841, 0.0705, -0.1366], + [-0.0478, -0.0031, -0.0488, ..., -0.0329, -0.0620, 0.0542]], + device='cuda:0'), grad: tensor([[ 1.4268e-06, 4.7013e-06, 1.0259e-05, ..., 2.3935e-06, + 4.9174e-06, 2.0443e-07], + [ 1.5944e-06, 7.7635e-06, 6.9588e-06, ..., 3.0966e-07, + 8.4192e-06, 1.3225e-07], + [ 4.6100e-07, 3.4031e-06, 5.0068e-06, ..., 7.3109e-07, + 3.5875e-06, 1.2619e-07], + ..., + [ 1.0207e-06, 3.2857e-06, 2.0992e-06, ..., 2.0489e-08, + 2.7828e-06, 7.9535e-07], + [-4.0494e-06, -2.0146e-05, -1.1466e-05, ..., 1.0878e-06, + -1.9640e-05, -3.5670e-07], + [-5.6671e-07, -6.8033e-07, 1.4175e-06, ..., 9.5926e-08, + 1.1828e-06, -1.3700e-06]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0276, -0.0205, -0.0097, -0.0023, 0.0090, 0.0070, -0.0169, -0.0006, + 0.0225, 0.0169], device='cuda:0'), grad: tensor([ 1.7956e-05, 2.1100e-05, 9.3728e-06, 8.2105e-06, 1.3150e-05, + 3.1680e-05, -6.4850e-05, 8.6203e-06, -4.6343e-05, 1.1390e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 257.39, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5345 re_mapping 0.0063 re_causal 0.0186 /// teacc 99.05 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0388, -0.0966, -0.1203, ..., -0.0467, -0.1741, -0.0188], + [-0.1073, -0.1346, -0.1420, ..., -0.1189, -0.0795, 0.0051], + [ 0.0768, 0.0519, -0.1498, ..., -0.1085, -0.1305, -0.0783], + ..., + [ 0.0159, 0.0425, -0.0795, ..., -0.0256, -0.1269, 0.0459], + [ 0.0151, 0.1083, 0.0845, ..., -0.0845, 0.0704, -0.1371], + [-0.0477, -0.0034, -0.0497, ..., -0.0331, -0.0623, 0.0544]], + device='cuda:0'), grad: tensor([[ 1.4296e-07, 6.9626e-06, 1.0237e-05, ..., 9.3281e-06, + 1.1094e-05, -5.9605e-08], + [ 1.2638e-06, 3.5837e-06, 3.5204e-06, ..., 1.7760e-06, + 4.6231e-06, -2.4354e-07], + [ 7.3202e-07, 1.7555e-06, 1.4799e-06, ..., 6.0024e-07, + 2.0768e-06, 3.0734e-08], + ..., + [ 1.7090e-07, 2.3376e-07, 2.9290e-07, ..., 2.5611e-08, + 4.4936e-07, 3.3528e-08], + [-2.4941e-06, 2.1428e-05, 3.6031e-05, ..., 3.6687e-05, + 3.7193e-05, 8.5216e-08], + [ 7.1246e-08, 4.6333e-07, 7.0827e-07, ..., 3.6089e-07, + 7.5996e-07, 3.2596e-08]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0269, -0.0199, -0.0106, -0.0021, 0.0097, 0.0073, -0.0163, -0.0008, + 0.0222, 0.0167], device='cuda:0'), grad: tensor([ 1.5914e-05, 8.7246e-06, 4.6492e-06, 8.8988e-07, 2.4922e-06, + 8.8662e-06, -9.9719e-05, 1.1958e-06, 5.5313e-05, 1.7583e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 257.50, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5468 re_mapping 0.0055 re_causal 0.0181 /// teacc 99.00 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0384, -0.0971, -0.1215, ..., -0.0478, -0.1750, -0.0192], + [-0.1082, -0.1352, -0.1426, ..., -0.1196, -0.0799, 0.0052], + [ 0.0770, 0.0518, -0.1503, ..., -0.1089, -0.1310, -0.0787], + ..., + [ 0.0163, 0.0429, -0.0804, ..., -0.0261, -0.1269, 0.0461], + [ 0.0154, 0.1089, 0.0850, ..., -0.0837, 0.0709, -0.1377], + [-0.0472, -0.0035, -0.0501, ..., -0.0336, -0.0620, 0.0545]], + device='cuda:0'), grad: tensor([[-1.1660e-06, 7.4506e-06, 3.2131e-08, ..., 1.2107e-08, + 3.3062e-08, 5.4948e-08], + [ 2.1188e-07, 1.3420e-06, 2.5611e-08, ..., -2.3283e-09, + 7.3574e-08, 2.1560e-07], + [ 1.0896e-07, 2.3525e-06, 4.8429e-08, ..., 1.1642e-08, + 6.7521e-08, 1.5367e-06], + ..., + [-2.4727e-07, -2.9467e-06, 3.9116e-08, ..., 2.3283e-09, + -9.9186e-08, -2.6822e-06], + [ 2.1420e-07, 2.1840e-07, -1.6205e-07, ..., 2.4214e-08, + -2.6450e-07, 6.3796e-08], + [-3.0035e-07, -1.4350e-05, 2.2817e-07, ..., 2.3283e-09, + 2.8405e-07, 4.2794e-07]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0252, -0.0199, -0.0108, -0.0022, 0.0100, 0.0061, -0.0154, -0.0004, + 0.0223, 0.0175], device='cuda:0'), grad: tensor([ 2.8685e-05, 3.6545e-06, 7.6815e-06, 9.3356e-06, 2.0549e-05, + -2.8033e-07, 2.8033e-06, -3.8482e-06, 3.4329e-06, -7.2062e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 257.07, cls_loss 0.0020 cls_loss_mapping 0.0040 cls_loss_causal 0.5137 re_mapping 0.0059 re_causal 0.0181 /// teacc 98.88 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0387, -0.0972, -0.1219, ..., -0.0479, -0.1755, -0.0193], + [-0.1089, -0.1356, -0.1428, ..., -0.1217, -0.0801, 0.0053], + [ 0.0772, 0.0521, -0.1505, ..., -0.1080, -0.1314, -0.0791], + ..., + [ 0.0166, 0.0430, -0.0810, ..., -0.0263, -0.1270, 0.0462], + [ 0.0155, 0.1091, 0.0851, ..., -0.0841, 0.0710, -0.1382], + [-0.0473, -0.0035, -0.0503, ..., -0.0335, -0.0623, 0.0545]], + device='cuda:0'), grad: tensor([[ 1.1874e-07, 2.3469e-06, 1.7714e-06, ..., 1.9260e-06, + 1.4901e-06, 4.4936e-07], + [ 5.6811e-07, 1.3113e-06, 7.7020e-07, ..., 1.1083e-07, + 8.8103e-07, 2.4587e-07], + [ 2.5202e-06, 6.3032e-06, 2.2445e-06, ..., 8.4471e-07, + 1.8580e-06, 4.6659e-07], + ..., + [-3.3844e-06, -6.9477e-06, 5.4715e-07, ..., 4.5169e-08, + 5.4156e-07, -7.2969e-07], + [-6.6450e-07, -2.1905e-06, 3.0044e-06, ..., -4.8196e-07, + 2.9895e-06, 3.0035e-07], + [ 3.6461e-07, 9.0059e-07, 8.8476e-07, ..., 3.3807e-07, + 7.9209e-07, -2.5146e-08]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0254, -0.0195, -0.0109, -0.0021, 0.0105, 0.0060, -0.0158, -0.0002, + 0.0221, 0.0170], device='cuda:0'), grad: tensor([ 4.7609e-06, 4.2804e-06, 1.7226e-05, 1.7405e-05, 3.9749e-06, + -1.0304e-05, -2.7016e-05, -2.1294e-05, 6.3591e-06, 4.4927e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 255.59, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5080 re_mapping 0.0060 re_causal 0.0181 /// teacc 99.01 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0389, -0.0975, -0.1221, ..., -0.0481, -0.1757, -0.0194], + [-0.1096, -0.1357, -0.1432, ..., -0.1215, -0.0802, 0.0056], + [ 0.0772, 0.0509, -0.1509, ..., -0.1090, -0.1320, -0.0795], + ..., + [ 0.0168, 0.0439, -0.0819, ..., -0.0239, -0.1277, 0.0459], + [ 0.0156, 0.1090, 0.0846, ..., -0.0845, 0.0707, -0.1386], + [-0.0475, -0.0037, -0.0511, ..., -0.0337, -0.0627, 0.0544]], + device='cuda:0'), grad: tensor([[-1.0477e-06, 4.8149e-07, 9.8255e-08, ..., 1.4435e-08, + 6.5658e-08, 4.6473e-07], + [ 6.4867e-07, 2.5239e-06, 1.3132e-07, ..., 5.1223e-09, + 9.5926e-08, -1.7539e-05], + [-4.7358e-07, 1.6354e-06, 7.6368e-08, ..., 4.6566e-09, + 6.2864e-08, 1.2003e-05], + ..., + [-1.8030e-06, -1.3508e-05, 3.3760e-07, ..., 0.0000e+00, + 1.9697e-07, -1.6773e-06], + [ 6.4261e-07, 1.1688e-06, 5.9530e-06, ..., 1.0943e-07, + 3.3472e-06, 5.8068e-07], + [ 1.3830e-06, 5.3197e-06, 1.5926e-07, ..., 9.3132e-09, + 2.0582e-07, 3.7681e-06]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0255, -0.0191, -0.0122, -0.0016, 0.0114, 0.0064, -0.0158, 0.0002, + 0.0215, 0.0164], device='cuda:0'), grad: tensor([-1.4128e-06, -8.9943e-05, 5.4240e-05, 6.5714e-06, -1.0114e-06, + -7.3463e-06, 2.9691e-06, 8.7619e-06, 1.1042e-05, 1.6049e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 256.57, cls_loss 0.0015 cls_loss_mapping 0.0038 cls_loss_causal 0.5283 re_mapping 0.0060 re_causal 0.0184 /// teacc 98.94 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0390, -0.0978, -0.1231, ..., -0.0494, -0.1762, -0.0196], + [-0.1102, -0.1360, -0.1438, ..., -0.1229, -0.0805, 0.0058], + [ 0.0777, 0.0511, -0.1511, ..., -0.1096, -0.1322, -0.0798], + ..., + [ 0.0167, 0.0442, -0.0830, ..., -0.0230, -0.1278, 0.0459], + [ 0.0157, 0.1091, 0.0844, ..., -0.0846, 0.0706, -0.1390], + [-0.0475, -0.0038, -0.0518, ..., -0.0339, -0.0630, 0.0546]], + device='cuda:0'), grad: tensor([[-1.3039e-07, 1.7127e-06, 1.0170e-06, ..., 9.9931e-07, + 9.8255e-07, 2.2352e-08], + [ 1.5469e-06, 6.0052e-06, 2.6003e-06, ..., 6.7567e-07, + 1.7611e-06, 7.5530e-07], + [-2.4103e-06, 3.5446e-06, 7.9200e-06, ..., 2.2966e-06, + 5.1819e-06, 2.2259e-07], + ..., + [ 3.2727e-06, 1.3085e-06, 1.2033e-06, ..., 2.3888e-07, + 8.3447e-07, -2.1178e-06], + [-1.8030e-05, -5.1975e-05, -3.9190e-05, ..., -9.0301e-06, + -2.5153e-05, 1.7881e-07], + [ 1.1120e-06, 3.0175e-06, 2.2352e-06, ..., 4.6100e-07, + 1.5851e-06, 1.8161e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0254, -0.0190, -0.0122, -0.0017, 0.0112, 0.0067, -0.0155, 0.0002, + 0.0212, 0.0164], device='cuda:0'), grad: tensor([ 3.6098e-06, 1.2234e-05, -1.8030e-06, 2.5749e-05, 6.7987e-06, + 1.0774e-05, 1.1250e-05, -2.5053e-07, -7.4804e-05, 6.2212e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 256.87, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5350 re_mapping 0.0059 re_causal 0.0178 /// teacc 98.94 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0397, -0.0979, -0.1234, ..., -0.0499, -0.1766, -0.0198], + [-0.1110, -0.1369, -0.1444, ..., -0.1253, -0.0811, 0.0055], + [ 0.0781, 0.0515, -0.1514, ..., -0.1083, -0.1322, -0.0800], + ..., + [ 0.0167, 0.0445, -0.0837, ..., -0.0230, -0.1279, 0.0462], + [ 0.0158, 0.1095, 0.0848, ..., -0.0847, 0.0708, -0.1397], + [-0.0479, -0.0043, -0.0530, ..., -0.0340, -0.0637, 0.0554]], + device='cuda:0'), grad: tensor([[-2.1514e-07, 1.2154e-07, 1.6298e-07, ..., 7.7765e-08, + 1.3877e-07, 3.8184e-08], + [ 1.4063e-07, 2.9383e-07, 1.0896e-07, ..., 2.0023e-08, + 1.5926e-07, -1.2601e-06], + [-2.0675e-07, -1.1874e-07, 7.8697e-08, ..., 1.6298e-08, + 1.0012e-07, 1.1083e-07], + ..., + [-9.4995e-08, -3.8324e-07, 1.0617e-07, ..., 9.3132e-10, + 1.1548e-07, 3.7439e-07], + [-8.2422e-08, -3.3760e-07, -6.2864e-08, ..., 1.3458e-07, + -1.2852e-07, 1.5227e-07], + [ 2.4866e-07, 1.8952e-07, 1.0384e-06, ..., 1.8161e-08, + 8.7265e-07, 3.0873e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0257, -0.0201, -0.0113, -0.0019, 0.0112, 0.0064, -0.0148, 0.0005, + 0.0212, 0.0162], device='cuda:0'), grad: tensor([-4.5402e-07, -3.1274e-06, -3.3900e-07, 1.5572e-06, 7.5623e-07, + -3.5241e-06, -2.8685e-07, 1.8869e-06, 3.7905e-07, 3.1609e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 256.45, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5543 re_mapping 0.0059 re_causal 0.0174 /// teacc 98.94 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0401, -0.0984, -0.1235, ..., -0.0502, -0.1771, -0.0199], + [-0.1124, -0.1375, -0.1458, ..., -0.1267, -0.0820, 0.0057], + [ 0.0782, 0.0514, -0.1516, ..., -0.1084, -0.1325, -0.0807], + ..., + [ 0.0170, 0.0450, -0.0845, ..., -0.0230, -0.1278, 0.0463], + [ 0.0163, 0.1115, 0.0852, ..., -0.0850, 0.0722, -0.1392], + [-0.0487, -0.0070, -0.0569, ..., -0.0344, -0.0675, 0.0552]], + device='cuda:0'), grad: tensor([[-7.0035e-07, 3.7216e-06, 3.1721e-06, ..., 1.1399e-06, + 3.9414e-06, -1.0785e-06], + [ 1.1921e-07, 1.2010e-05, 1.2197e-05, ..., 1.9316e-06, + 1.2815e-05, -4.6007e-06], + [-2.4214e-08, 1.4678e-06, 1.7565e-06, ..., 6.8638e-07, + 1.8701e-06, 7.2736e-07], + ..., + [-1.4156e-07, 4.1537e-07, 8.0839e-07, ..., 2.1327e-07, + 1.3039e-06, 2.9057e-07], + [ 4.0513e-08, -4.2629e-04, -4.3273e-04, ..., -1.2779e-04, + -4.4346e-04, 2.1234e-07], + [ 3.7113e-07, 1.2051e-06, 1.8263e-06, ..., 4.0326e-07, + 1.9744e-06, 9.6206e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0261, -0.0200, -0.0115, -0.0002, 0.0119, 0.0047, -0.0147, 0.0003, + 0.0225, 0.0149], device='cuda:0'), grad: tensor([ 2.3283e-06, -2.2322e-05, 8.9258e-06, 3.4403e-06, 1.2442e-05, + 5.0962e-06, 9.2173e-04, 1.8403e-05, -9.6178e-04, 1.1981e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 256.79, cls_loss 0.0030 cls_loss_mapping 0.0055 cls_loss_causal 0.5372 re_mapping 0.0060 re_causal 0.0181 /// teacc 99.07 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0409, -0.0981, -0.1238, ..., -0.0501, -0.1773, -0.0193], + [-0.1131, -0.1378, -0.1461, ..., -0.1266, -0.0823, 0.0060], + [ 0.0784, 0.0514, -0.1520, ..., -0.1083, -0.1329, -0.0821], + ..., + [ 0.0172, 0.0450, -0.0856, ..., -0.0225, -0.1280, 0.0464], + [ 0.0165, 0.1122, 0.0863, ..., -0.0845, 0.0731, -0.1400], + [-0.0500, -0.0070, -0.0576, ..., -0.0362, -0.0677, 0.0556]], + device='cuda:0'), grad: tensor([[-2.6021e-06, 1.6578e-07, 6.2399e-08, ..., 4.6100e-08, + 9.2201e-08, 6.2399e-08], + [ 2.6356e-06, 8.7246e-06, 1.6037e-06, ..., 3.2596e-08, + 3.3733e-06, 1.6317e-06], + [ 1.7043e-06, 3.7588e-06, 1.9390e-06, ..., 3.1665e-08, + 3.9563e-06, 4.2049e-07], + ..., + [-1.8934e-06, -9.5442e-06, 1.5413e-07, ..., 0.0000e+00, + 1.3923e-07, -2.1309e-06], + [-2.6654e-06, -3.9190e-06, -3.8091e-06, ..., 7.3109e-08, + -7.7337e-06, 3.2131e-07], + [ 2.2445e-06, -1.5553e-07, 3.7253e-08, ..., 3.2596e-09, + -3.3202e-07, -7.9488e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0234, -0.0196, -0.0119, -0.0004, 0.0131, 0.0050, -0.0161, 0.0002, + 0.0231, 0.0173], device='cuda:0'), grad: tensor([-1.4469e-05, 1.8626e-05, 1.0550e-05, 1.0943e-06, 1.8850e-06, + 5.1223e-09, 2.4354e-07, -1.5363e-05, -1.3724e-05, 1.1124e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 256.68, cls_loss 0.0014 cls_loss_mapping 0.0036 cls_loss_causal 0.5386 re_mapping 0.0056 re_causal 0.0190 /// teacc 99.02 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0413, -0.0980, -0.1240, ..., -0.0505, -0.1773, -0.0189], + [-0.1137, -0.1400, -0.1466, ..., -0.1269, -0.0830, 0.0046], + [ 0.0786, 0.0517, -0.1520, ..., -0.1083, -0.1325, -0.0823], + ..., + [ 0.0174, 0.0463, -0.0861, ..., -0.0225, -0.1280, 0.0482], + [ 0.0167, 0.1126, 0.0870, ..., -0.0842, 0.0734, -0.1404], + [-0.0504, -0.0073, -0.0579, ..., -0.0363, -0.0679, 0.0553]], + device='cuda:0'), grad: tensor([[-1.3504e-07, 7.4953e-06, 1.9118e-05, ..., 1.3016e-05, + 2.1663e-06, -9.4064e-08], + [ 1.4622e-07, 1.9725e-06, 3.8696e-07, ..., 2.0256e-07, + 1.2387e-07, 1.6550e-06], + [-3.9209e-07, -2.7008e-07, 3.6089e-07, ..., 9.8720e-08, + 2.3516e-07, 8.8476e-09], + ..., + [ 1.3597e-07, -1.6121e-06, 9.6858e-08, ..., 2.2352e-08, + 6.2864e-08, -1.7332e-06], + [-7.2792e-06, -5.7667e-05, -7.0691e-05, ..., -1.9357e-05, + -4.3690e-05, 2.1420e-08], + [ 1.2573e-07, 2.4773e-07, 8.5402e-07, ..., 3.2224e-07, + 3.7486e-07, -4.6566e-10]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0236, -0.0212, -0.0114, -0.0005, 0.0129, 0.0049, -0.0165, 0.0019, + 0.0229, 0.0172], device='cuda:0'), grad: tensor([ 4.7654e-05, 4.9323e-06, -4.7218e-07, -4.1634e-05, 8.6986e-07, + 4.9949e-05, 1.3731e-05, -2.9281e-06, -7.5400e-05, 3.4235e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 256.48, cls_loss 0.0018 cls_loss_mapping 0.0043 cls_loss_causal 0.5135 re_mapping 0.0057 re_causal 0.0180 /// teacc 98.94 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0417, -0.0980, -0.1244, ..., -0.0507, -0.1775, -0.0191], + [-0.1140, -0.1402, -0.1463, ..., -0.1266, -0.0831, 0.0047], + [ 0.0789, 0.0518, -0.1523, ..., -0.1083, -0.1328, -0.0828], + ..., + [ 0.0176, 0.0466, -0.0878, ..., -0.0224, -0.1282, 0.0484], + [ 0.0166, 0.1126, 0.0870, ..., -0.0844, 0.0733, -0.1417], + [-0.0505, -0.0074, -0.0594, ..., -0.0368, -0.0684, 0.0552]], + device='cuda:0'), grad: tensor([[-7.7719e-07, 7.1246e-08, 7.6368e-08, ..., 4.6566e-09, + 1.2992e-07, 5.3085e-08], + [ 1.8906e-07, 2.8405e-07, 1.5972e-07, ..., 1.3970e-09, + 2.6915e-07, -2.4773e-07], + [ 1.2480e-07, 1.4203e-07, 1.8720e-07, ..., 3.7253e-09, + 3.1386e-07, 4.8894e-08], + ..., + [ 6.7521e-08, 1.2619e-07, 1.2852e-07, ..., 0.0000e+00, + 2.2026e-07, 2.7986e-07], + [-3.3714e-07, -5.5181e-07, -3.9162e-07, ..., 2.3749e-08, + -2.9104e-07, 2.7940e-07], + [-1.0710e-07, -1.2154e-07, 1.1073e-06, ..., 9.3132e-10, + 2.0470e-06, -8.8476e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0236, -0.0210, -0.0114, -0.0033, 0.0130, 0.0079, -0.0170, 0.0021, + 0.0224, 0.0170], device='cuda:0'), grad: tensor([-3.8743e-06, 9.6485e-07, 1.2936e-06, -1.8895e-05, -2.7478e-05, + 1.2361e-05, 1.9409e-06, 1.4696e-06, 1.8124e-06, 3.0428e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 256.65, cls_loss 0.0023 cls_loss_mapping 0.0043 cls_loss_causal 0.5455 re_mapping 0.0061 re_causal 0.0176 /// teacc 98.94 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0422, -0.0985, -0.1253, ..., -0.0518, -0.1781, -0.0195], + [-0.1147, -0.1405, -0.1468, ..., -0.1269, -0.0835, 0.0049], + [ 0.0789, 0.0518, -0.1528, ..., -0.1084, -0.1336, -0.0833], + ..., + [ 0.0178, 0.0465, -0.0895, ..., -0.0224, -0.1284, 0.0482], + [ 0.0167, 0.1128, 0.0869, ..., -0.0846, 0.0735, -0.1421], + [-0.0505, -0.0072, -0.0612, ..., -0.0355, -0.0691, 0.0555]], + device='cuda:0'), grad: tensor([[-2.3050e-07, 3.2922e-07, -1.2338e-04, ..., -5.0664e-05, + 2.5518e-07, 6.9989e-07], + [ 1.5693e-07, 2.2305e-07, 2.9244e-07, ..., 8.0559e-08, + 1.9604e-07, 1.0990e-07], + [-1.4110e-07, 7.4040e-08, 1.1222e-07, ..., 1.6764e-08, + 9.4995e-08, 3.0268e-08], + ..., + [-3.0734e-08, 1.8720e-07, 1.2713e-07, ..., 2.5611e-08, + 1.0245e-07, 5.9465e-07], + [-1.7090e-07, -5.6531e-07, 1.4249e-07, ..., 2.4308e-07, + -3.9488e-07, 1.7695e-08], + [ 2.6077e-07, -5.8301e-07, 1.7891e-06, ..., 6.1095e-07, + 3.8743e-07, -1.6652e-06]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0236, -0.0206, -0.0116, -0.0031, 0.0153, 0.0077, -0.0168, 0.0017, + 0.0222, 0.0164], device='cuda:0'), grad: tensor([-5.2452e-04, 2.8163e-06, -1.7667e-06, 2.0396e-06, 8.2096e-07, + -1.5885e-05, 5.3024e-04, 3.5185e-06, 7.9302e-07, 1.8720e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 256.99, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5168 re_mapping 0.0058 re_causal 0.0182 /// teacc 98.95 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0426, -0.0986, -0.1248, ..., -0.0514, -0.1778, -0.0199], + [-0.1152, -0.1408, -0.1473, ..., -0.1270, -0.0841, 0.0049], + [ 0.0789, 0.0519, -0.1531, ..., -0.1085, -0.1342, -0.0838], + ..., + [ 0.0178, 0.0466, -0.0908, ..., -0.0224, -0.1285, 0.0482], + [ 0.0167, 0.1130, 0.0871, ..., -0.0847, 0.0737, -0.1425], + [-0.0505, -0.0070, -0.0613, ..., -0.0349, -0.0691, 0.0563]], + device='cuda:0'), grad: tensor([[-1.0151e-07, 4.4703e-08, 1.8440e-07, ..., 7.4506e-09, + 2.3376e-07, 2.2817e-08], + [ 1.0384e-07, 1.8720e-07, 9.1270e-08, ..., 2.3283e-09, + 1.3271e-07, -4.4238e-08], + [ 7.6834e-08, 5.5972e-07, 7.5437e-08, ..., 1.3970e-09, + 2.9709e-07, 3.3062e-08], + ..., + [-6.2212e-07, -1.9167e-06, 1.1083e-07, ..., 0.0000e+00, + -5.1036e-07, -2.0489e-08], + [ 2.7427e-07, 1.2247e-06, 5.1409e-06, ..., 1.3039e-08, + 7.1935e-06, 4.2841e-08], + [ 6.9849e-08, -9.4064e-07, -6.1318e-06, ..., 4.6566e-10, + -8.8140e-06, -4.8988e-07]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0239, -0.0207, -0.0120, -0.0033, 0.0156, 0.0078, -0.0175, 0.0017, + 0.0221, 0.0164], device='cuda:0'), grad: tensor([ 8.1956e-08, 6.2073e-07, 9.8348e-07, 2.4047e-06, 2.6636e-06, + -1.1614e-06, 1.1101e-06, -3.3304e-06, 1.6227e-05, -1.9625e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 256.56, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5241 re_mapping 0.0055 re_causal 0.0173 /// teacc 98.94 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0427, -0.0989, -0.1250, ..., -0.0511, -0.1780, -0.0195], + [-0.1158, -0.1411, -0.1479, ..., -0.1277, -0.0846, 0.0063], + [ 0.0795, 0.0521, -0.1534, ..., -0.1084, -0.1347, -0.0839], + ..., + [ 0.0175, 0.0465, -0.0921, ..., -0.0224, -0.1288, 0.0467], + [ 0.0168, 0.1134, 0.0874, ..., -0.0846, 0.0740, -0.1429], + [-0.0505, -0.0065, -0.0615, ..., -0.0355, -0.0689, 0.0568]], + device='cuda:0'), grad: tensor([[ 7.3109e-08, 5.4948e-07, 7.4040e-08, ..., 6.7055e-08, + 4.5635e-08, 2.6496e-07], + [ 3.1618e-07, 1.8738e-06, 4.7032e-08, ..., 3.7253e-08, + 1.2806e-07, 5.4110e-07], + [ 1.6624e-07, 1.3411e-06, 1.8626e-08, ..., -1.5367e-08, + 5.3551e-08, 4.9453e-07], + ..., + [-2.6692e-06, -1.6913e-05, 1.0710e-08, ..., 3.2596e-09, + -1.2536e-06, -5.6177e-06], + [ 1.0654e-06, 1.1690e-05, 7.8827e-06, ..., 4.9993e-06, + 5.1335e-06, 1.8254e-06], + [ 7.7346e-07, 4.8913e-06, 8.7079e-08, ..., 4.7497e-08, + 3.8184e-07, 1.7723e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0241, -0.0201, -0.0115, -0.0033, 0.0153, 0.0077, -0.0178, 0.0004, + 0.0222, 0.0165], device='cuda:0'), grad: tensor([ 1.6298e-06, 3.1814e-06, 3.1628e-06, 9.6764e-07, 1.2964e-06, + 1.3812e-06, -1.0967e-05, -3.0875e-05, 2.0847e-05, 9.3430e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 256.44, cls_loss 0.0019 cls_loss_mapping 0.0035 cls_loss_causal 0.5120 re_mapping 0.0056 re_causal 0.0164 /// teacc 98.83 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0430, -0.0993, -0.1255, ..., -0.0515, -0.1786, -0.0199], + [-0.1172, -0.1420, -0.1486, ..., -0.1280, -0.0851, 0.0061], + [ 0.0800, 0.0525, -0.1538, ..., -0.1087, -0.1353, -0.0841], + ..., + [ 0.0181, 0.0478, -0.0929, ..., -0.0225, -0.1289, 0.0474], + [ 0.0164, 0.1129, 0.0871, ..., -0.0850, 0.0736, -0.1443], + [-0.0501, -0.0070, -0.0618, ..., -0.0343, -0.0686, 0.0564]], + device='cuda:0'), grad: tensor([[-8.2050e-07, 2.2119e-08, 1.6997e-08, ..., 9.5461e-09, + 1.5134e-08, 3.4459e-08], + [ 6.3563e-08, 4.3167e-07, 5.1223e-09, ..., 1.3970e-09, + 1.6298e-09, -7.3388e-07], + [ 2.1467e-07, 1.2713e-07, 1.7928e-08, ..., 9.7789e-09, + 2.4447e-08, 7.6368e-08], + ..., + [-3.1455e-07, -4.7348e-06, 3.9581e-09, ..., 0.0000e+00, + 1.3970e-09, -2.2892e-06], + [ 6.5425e-08, 7.0082e-08, 5.5647e-08, ..., 3.0734e-08, + 4.3306e-08, 1.6438e-07], + [ 5.3830e-07, 2.8890e-06, 8.8476e-09, ..., 6.9849e-10, + 9.0804e-09, 1.8980e-06]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0241, -0.0207, -0.0113, -0.0030, 0.0146, 0.0074, -0.0167, 0.0014, + 0.0212, 0.0166], device='cuda:0'), grad: tensor([-3.6471e-06, -2.2352e-06, 1.2340e-06, 2.4308e-06, 4.3376e-07, + 7.1665e-07, 4.3050e-07, -8.7246e-06, 1.0040e-06, 8.3372e-06], + device='cuda:0') +100 +0.0001 +changing lr diff --git a/Meta-causal/code-withStyleAttack/66550.error b/Meta-causal/code-withStyleAttack/66550.error new file mode 100644 index 0000000000000000000000000000000000000000..aa58699561182e03e316a8e1a4c5831cdb421167 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66550.error @@ -0,0 +1,7 @@ +Traceback (most recent call last): + File "/usr/bin/mkenv", line 544, in +srun: error: gcp-us-1: task 0: Exited with exit code 1 + main() + File "/usr/bin/mkenv", line 512, in main + raise RuntimeError('must run without an activated environment') +RuntimeError: must run without an activated environment diff --git a/Meta-causal/code-withStyleAttack/66550.log b/Meta-causal/code-withStyleAttack/66550.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/66551.error b/Meta-causal/code-withStyleAttack/66551.error new file mode 100644 index 0000000000000000000000000000000000000000..daba6d2f7ea07a3d4480fde2fa2c05bcbd2760bb --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66551.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 34: ta: command not found diff --git a/Meta-causal/code-withStyleAttack/66551.log b/Meta-causal/code-withStyleAttack/66551.log new file mode 100644 index 0000000000000000000000000000000000000000..f42220a26f6498b59696a47a76c3d10c397271e3 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66551.log @@ -0,0 +1,14133 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_adam', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0229, -0.0198, 0.0145, ..., -0.0150, -0.0229, 0.0150], + [ 0.0067, -0.0086, -0.0215, ..., 0.0307, 0.0172, 0.0031], + [-0.0304, -0.0185, -0.0098, ..., 0.0100, -0.0173, 0.0222], + ..., + [ 0.0125, -0.0113, -0.0201, ..., -0.0116, 0.0090, 0.0017], + [ 0.0270, 0.0120, -0.0142, ..., -0.0211, -0.0112, 0.0308], + [-0.0156, 0.0101, 0.0250, ..., -0.0249, 0.0107, -0.0086]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0058, -0.0305, 0.0100, -0.0208, 0.0150, 0.0014, 0.0218, -0.0126, + -0.0232, 0.0025], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 280.32, cls_loss 1.5221 cls_loss_mapping 1.9297 cls_loss_causal 2.2244 re_mapping 0.1031 re_causal 0.1042 /// teacc 82.82 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0187, -0.0223, 0.0113, ..., -0.0233, -0.0257, 0.0088], + [ 0.0056, -0.0064, -0.0187, ..., 0.0356, 0.0174, 0.0065], + [-0.0322, -0.0189, -0.0155, ..., 0.0165, -0.0119, 0.0174], + ..., + [ 0.0079, -0.0115, -0.0274, ..., -0.0111, 0.0017, 0.0076], + [ 0.0233, 0.0096, -0.0100, ..., -0.0189, -0.0133, 0.0258], + [-0.0200, 0.0071, 0.0202, ..., -0.0310, 0.0091, -0.0075]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.2368e-03, ..., 5.0354e-03, + -9.0866e-03, 4.6310e-07], + [ 0.0000e+00, 0.0000e+00, 9.5673e-03, ..., 1.8482e-03, + 1.3781e-03, -2.4796e-04], + [ 0.0000e+00, 0.0000e+00, -1.5747e-02, ..., -5.8258e-02, + -1.5022e-02, 2.7195e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5656e-02, ..., 1.6800e-02, + 2.0004e-02, 2.4736e-05], + [ 0.0000e+00, 0.0000e+00, -2.1423e-02, ..., 6.3515e-04, + 4.4823e-03, 4.9353e-05], + [ 0.0000e+00, 0.0000e+00, 2.0767e-02, ..., 1.0090e-03, + -1.9806e-02, 4.1455e-05]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0077, -0.0288, 0.0089, -0.0207, 0.0140, 0.0016, 0.0217, -0.0119, + -0.0237, 0.0025], device='cuda:0'), grad: tensor([-0.0174, 0.0233, -0.0216, 0.0623, 0.0347, -0.1466, 0.0386, 0.0470, + -0.0019, -0.0184], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 279.51, cls_loss 0.5193 cls_loss_mapping 0.8489 cls_loss_causal 1.9181 re_mapping 0.2062 re_causal 0.2506 /// teacc 90.62 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0131, -0.0223, 0.0066, ..., -0.0265, -0.0276, 0.0070], + [ 0.0052, -0.0064, -0.0181, ..., 0.0379, 0.0188, 0.0087], + [-0.0315, -0.0189, -0.0192, ..., 0.0190, -0.0104, 0.0200], + ..., + [ 0.0002, -0.0115, -0.0320, ..., -0.0099, -0.0008, 0.0109], + [ 0.0202, 0.0096, -0.0075, ..., -0.0198, -0.0153, 0.0207], + [-0.0269, 0.0071, 0.0190, ..., -0.0360, 0.0083, -0.0096]], + device='cuda:0'), grad: tensor([[-2.5997e-03, 0.0000e+00, 2.1439e-03, ..., 2.4738e-03, + 1.2684e-04, 5.6148e-05], + [ 5.8365e-04, 0.0000e+00, -1.1978e-02, ..., -3.0258e-02, + -1.8454e-03, -4.2992e-03], + [-2.1706e-03, 0.0000e+00, 6.4659e-04, ..., -4.2664e-02, + -1.0214e-03, -9.6273e-04], + ..., + [ 1.2836e-03, 0.0000e+00, 3.8624e-03, ..., 1.3199e-02, + 2.3232e-03, 4.7660e-04], + [-1.1740e-03, 0.0000e+00, -4.2915e-03, ..., 1.0429e-02, + 1.4200e-03, 1.0481e-03], + [ 4.7398e-04, 0.0000e+00, 7.9727e-03, ..., 1.1276e-02, + -5.7564e-03, -4.7922e-04]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0081, -0.0287, 0.0083, -0.0207, 0.0138, 0.0029, 0.0213, -0.0123, + -0.0239, 0.0030], device='cuda:0'), grad: tensor([-0.0053, -0.0143, -0.0182, 0.0160, -0.0222, 0.0097, 0.0101, 0.0087, + -0.0011, 0.0167], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 278.47, cls_loss 0.3500 cls_loss_mapping 0.4936 cls_loss_causal 1.6778 re_mapping 0.1573 re_causal 0.2369 /// teacc 93.59 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0103, -0.0223, 0.0036, ..., -0.0287, -0.0287, 0.0006], + [ 0.0057, -0.0064, -0.0182, ..., 0.0386, 0.0202, 0.0099], + [-0.0292, -0.0189, -0.0216, ..., 0.0207, -0.0094, 0.0201], + ..., + [-0.0062, -0.0115, -0.0360, ..., -0.0080, -0.0019, 0.0151], + [ 0.0193, 0.0096, -0.0069, ..., -0.0198, -0.0175, 0.0158], + [-0.0297, 0.0071, 0.0201, ..., -0.0389, 0.0078, -0.0117]], + device='cuda:0'), grad: tensor([[ 4.4179e-04, 0.0000e+00, 4.5180e-04, ..., 7.4625e-04, + 1.1253e-02, 3.5310e-04], + [-1.3762e-03, 0.0000e+00, 1.5144e-02, ..., 1.3113e-03, + 9.5978e-03, 6.7520e-03], + [ 5.4502e-04, 0.0000e+00, 2.3842e-03, ..., -6.1378e-03, + 2.9488e-03, -4.8709e-04], + ..., + [ 9.7215e-05, 0.0000e+00, 6.0959e-03, ..., -2.2542e-04, + 4.5700e-03, -6.2799e-04], + [ 1.1873e-03, 0.0000e+00, 6.7101e-03, ..., 1.1421e-02, + 5.5656e-03, -1.2693e-03], + [ 8.5890e-05, 0.0000e+00, -3.5004e-02, ..., -4.2381e-03, + 8.7357e-03, 1.7843e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0087, -0.0287, 0.0084, -0.0209, 0.0139, 0.0037, 0.0211, -0.0121, + -0.0242, 0.0031], device='cuda:0'), grad: tensor([ 0.0173, 0.0151, 0.0085, 0.0242, -0.0677, 0.0196, -0.0236, 0.0063, + 0.0038, -0.0034], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 278.78, cls_loss 0.2505 cls_loss_mapping 0.3509 cls_loss_causal 1.5292 re_mapping 0.1212 re_causal 0.2169 /// teacc 94.58 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0085, -0.0223, 0.0007, ..., -0.0307, -0.0284, -0.0034], + [ 0.0055, -0.0064, -0.0171, ..., 0.0400, 0.0222, 0.0099], + [-0.0293, -0.0189, -0.0246, ..., 0.0221, -0.0094, 0.0214], + ..., + [-0.0140, -0.0115, -0.0392, ..., -0.0075, -0.0040, 0.0179], + [ 0.0244, 0.0096, -0.0061, ..., -0.0199, -0.0191, 0.0140], + [-0.0356, 0.0071, 0.0215, ..., -0.0412, 0.0072, -0.0145]], + device='cuda:0'), grad: tensor([[ 6.2287e-05, 0.0000e+00, 4.5681e-04, ..., 1.0548e-03, + -1.6918e-03, 8.2031e-06], + [-9.3384e-03, 0.0000e+00, -1.2779e-02, ..., -7.9224e-02, + -1.2383e-02, 1.2174e-05], + [ 7.0686e-03, 0.0000e+00, 1.0117e-02, ..., 4.0863e-02, + 3.9062e-03, -4.0102e-04], + ..., + [ 1.6212e-04, 0.0000e+00, 1.8187e-03, ..., -1.3103e-03, + 1.1091e-03, 6.6042e-05], + [ 2.4652e-04, 0.0000e+00, 2.5749e-03, ..., 3.2120e-03, + 2.2488e-03, 4.2975e-05], + [ 1.7071e-04, 0.0000e+00, -8.1787e-03, ..., 1.1187e-03, + -5.7755e-03, 1.0431e-05]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0084, -0.0282, 0.0084, -0.0209, 0.0141, 0.0034, 0.0211, -0.0122, + -0.0246, 0.0030], device='cuda:0'), grad: tensor([-0.0055, -0.0304, 0.0171, 0.0146, 0.0064, 0.0022, 0.0062, 0.0020, + 0.0054, -0.0179], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 280.63, cls_loss 0.2013 cls_loss_mapping 0.2692 cls_loss_causal 1.3562 re_mapping 0.1003 re_causal 0.1970 /// teacc 95.39 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0058, -0.0223, -0.0011, ..., -0.0329, -0.0288, -0.0075], + [ 0.0085, -0.0064, -0.0168, ..., 0.0413, 0.0242, 0.0100], + [-0.0316, -0.0189, -0.0277, ..., 0.0234, -0.0081, 0.0215], + ..., + [-0.0200, -0.0115, -0.0408, ..., -0.0075, -0.0056, 0.0205], + [ 0.0268, 0.0096, -0.0055, ..., -0.0193, -0.0204, 0.0121], + [-0.0418, 0.0071, 0.0222, ..., -0.0437, 0.0067, -0.0172]], + device='cuda:0'), grad: tensor([[ 5.8651e-04, 0.0000e+00, 9.8896e-04, ..., 7.0429e-04, + 2.5864e-03, 4.5180e-05], + [-2.1286e-03, 0.0000e+00, -1.4162e-03, ..., -4.3983e-03, + -5.0850e-03, -5.0592e-04], + [ 8.8739e-04, 0.0000e+00, 2.5845e-03, ..., 8.9979e-04, + 5.0201e-03, 2.9182e-04], + ..., + [ 2.7180e-04, 0.0000e+00, 1.0967e-03, ..., -2.4915e-04, + -5.6076e-04, -3.5405e-04], + [ 2.2066e-04, 0.0000e+00, -5.7259e-03, ..., -6.0310e-03, + -9.4681e-03, 7.8058e-04], + [ 9.2649e-04, 0.0000e+00, -3.8929e-03, ..., 1.0090e-03, + -1.4992e-03, 2.9826e-04]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0082, -0.0279, 0.0085, -0.0209, 0.0142, 0.0030, 0.0207, -0.0125, + -0.0243, 0.0032], device='cuda:0'), grad: tensor([ 0.0021, -0.0036, 0.0055, -0.0034, -0.0021, 0.0094, 0.0069, -0.0034, + -0.0055, -0.0059], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 278.72, cls_loss 0.1822 cls_loss_mapping 0.2325 cls_loss_causal 1.3252 re_mapping 0.0856 re_causal 0.1840 /// teacc 96.41 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0048, -0.0223, -0.0029, ..., -0.0350, -0.0300, -0.0132], + [ 0.0102, -0.0064, -0.0165, ..., 0.0420, 0.0263, 0.0095], + [-0.0334, -0.0189, -0.0305, ..., 0.0243, -0.0072, 0.0208], + ..., + [-0.0229, -0.0115, -0.0416, ..., -0.0071, -0.0070, 0.0219], + [ 0.0299, 0.0096, -0.0046, ..., -0.0194, -0.0223, 0.0103], + [-0.0454, 0.0071, 0.0220, ..., -0.0465, 0.0057, -0.0200]], + device='cuda:0'), grad: tensor([[-0.0008, 0.0000, 0.0014, ..., 0.0017, -0.0031, 0.0002], + [ 0.0029, 0.0000, 0.0070, ..., 0.0149, -0.0004, -0.0002], + [ 0.0014, 0.0000, 0.0031, ..., -0.0083, 0.0003, -0.0008], + ..., + [ 0.0008, 0.0000, 0.0013, ..., 0.0039, 0.0040, 0.0002], + [-0.0004, 0.0000, -0.0041, ..., -0.0001, 0.0012, -0.0005], + [ 0.0008, 0.0000, 0.0015, ..., 0.0028, -0.0038, -0.0012]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0082, -0.0278, 0.0082, -0.0209, 0.0145, 0.0031, 0.0202, -0.0122, + -0.0238, 0.0027], device='cuda:0'), grad: tensor([-0.0098, 0.0083, -0.0008, -0.0146, 0.0015, 0.0237, -0.0142, 0.0126, + 0.0012, -0.0079], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 278.42, cls_loss 0.1492 cls_loss_mapping 0.1952 cls_loss_causal 1.2539 re_mapping 0.0763 re_causal 0.1698 /// teacc 96.50 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0034, -0.0223, -0.0051, ..., -0.0376, -0.0306, -0.0179], + [ 0.0120, -0.0064, -0.0159, ..., 0.0430, 0.0283, 0.0086], + [-0.0338, -0.0189, -0.0320, ..., 0.0253, -0.0059, 0.0204], + ..., + [-0.0255, -0.0115, -0.0432, ..., -0.0073, -0.0088, 0.0229], + [ 0.0313, 0.0096, -0.0039, ..., -0.0195, -0.0235, 0.0087], + [-0.0490, 0.0071, 0.0221, ..., -0.0479, 0.0052, -0.0221]], + device='cuda:0'), grad: tensor([[-6.9618e-04, 0.0000e+00, 4.6611e-04, ..., 3.6454e-04, + -8.1873e-04, 2.3916e-05], + [-1.0653e-03, 0.0000e+00, -8.9502e-04, ..., -1.1177e-03, + -3.5324e-03, -2.8357e-05], + [ 5.7268e-04, 0.0000e+00, 3.9649e-04, ..., -5.2299e-03, + -1.8473e-03, 1.1764e-05], + ..., + [ 1.6406e-05, 0.0000e+00, 1.4200e-03, ..., 3.4833e-04, + 6.3705e-04, -4.7743e-05], + [ 1.0405e-03, 0.0000e+00, 9.7809e-03, ..., 5.3101e-03, + 2.9888e-03, 4.3011e-04], + [ 1.7223e-03, 0.0000e+00, 1.4048e-03, ..., 2.6779e-03, + -2.5272e-03, -1.1355e-04]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0083, -0.0276, 0.0084, -0.0207, 0.0145, 0.0026, 0.0200, -0.0122, + -0.0235, 0.0026], device='cuda:0'), grad: tensor([-0.0018, -0.0020, -0.0042, -0.0106, 0.0061, 0.0014, 0.0023, -0.0003, + 0.0121, -0.0030], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 279.28, cls_loss 0.1329 cls_loss_mapping 0.1677 cls_loss_causal 1.2467 re_mapping 0.0697 re_causal 0.1641 /// teacc 96.76 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0019, -0.0223, -0.0066, ..., -0.0392, -0.0305, -0.0224], + [ 0.0127, -0.0064, -0.0156, ..., 0.0437, 0.0305, 0.0089], + [-0.0338, -0.0189, -0.0334, ..., 0.0263, -0.0056, 0.0197], + ..., + [-0.0276, -0.0115, -0.0439, ..., -0.0075, -0.0106, 0.0241], + [ 0.0321, 0.0096, -0.0035, ..., -0.0196, -0.0245, 0.0077], + [-0.0504, 0.0071, 0.0220, ..., -0.0490, 0.0053, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.9860e-04, 0.0000e+00, 8.2159e-04, ..., 4.7159e-04, + 6.7472e-05, 2.5824e-05], + [-1.2474e-03, 0.0000e+00, -1.9409e-02, ..., -5.4688e-02, + -2.6627e-02, -8.7023e-04], + [ 1.3523e-03, 0.0000e+00, 2.0142e-02, ..., 7.3608e-02, + 2.9358e-02, 3.0255e-04], + ..., + [ 2.2793e-04, 0.0000e+00, 1.4668e-03, ..., -2.4048e-02, + -3.1796e-03, 6.5565e-06], + [ 3.5501e-04, 0.0000e+00, 6.9475e-04, ..., 1.3647e-03, + 2.1133e-03, -1.2353e-05], + [-3.2687e-04, 0.0000e+00, -5.8670e-03, ..., -2.5215e-03, + -6.8779e-03, 1.9753e-04]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0079, -0.0274, 0.0085, -0.0209, 0.0146, 0.0023, 0.0198, -0.0121, + -0.0236, 0.0026], device='cuda:0'), grad: tensor([ 0.0006, -0.0255, 0.0376, 0.0006, 0.0075, 0.0014, -0.0009, -0.0115, + 0.0020, -0.0118], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 278.95, cls_loss 0.0998 cls_loss_mapping 0.1346 cls_loss_causal 1.1738 re_mapping 0.0622 re_causal 0.1539 /// teacc 97.20 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0009, -0.0223, -0.0080, ..., -0.0408, -0.0304, -0.0248], + [ 0.0123, -0.0064, -0.0155, ..., 0.0436, 0.0313, 0.0084], + [-0.0347, -0.0189, -0.0349, ..., 0.0273, -0.0048, 0.0195], + ..., + [-0.0305, -0.0115, -0.0448, ..., -0.0073, -0.0118, 0.0247], + [ 0.0348, 0.0096, -0.0033, ..., -0.0192, -0.0253, 0.0073], + [-0.0519, 0.0071, 0.0219, ..., -0.0504, 0.0056, -0.0244]], + device='cuda:0'), grad: tensor([[ 7.9155e-05, 0.0000e+00, 8.6403e-04, ..., 3.7122e-04, + 1.4389e-04, 6.3598e-05], + [ 2.0611e-04, 0.0000e+00, 1.6289e-03, ..., 3.4332e-03, + 3.0112e-04, 1.1339e-03], + [-4.8685e-04, 0.0000e+00, -2.1553e-03, ..., -1.5991e-02, + 1.0443e-04, -8.9188e-03], + ..., + [ 1.5247e-04, 0.0000e+00, 3.0479e-03, ..., 6.2904e-03, + 1.4896e-03, 2.5101e-03], + [ 8.8644e-04, 0.0000e+00, 4.4441e-03, ..., 1.6006e-02, + 1.7605e-03, 8.9111e-03], + [ 3.7456e-04, 0.0000e+00, 8.9979e-04, ..., 1.7252e-03, + -3.5167e-05, 3.4451e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0077, -0.0277, 0.0088, -0.0210, 0.0145, 0.0019, 0.0198, -0.0119, + -0.0232, 0.0025], device='cuda:0'), grad: tensor([-0.0019, 0.0027, -0.0081, -0.0084, 0.0010, -0.0204, 0.0108, 0.0074, + 0.0175, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 262.35, cls_loss 0.0925 cls_loss_mapping 0.1251 cls_loss_causal 1.1859 re_mapping 0.0567 re_causal 0.1430 /// teacc 96.75 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0003, -0.0223, -0.0092, ..., -0.0426, -0.0312, -0.0276], + [ 0.0129, -0.0064, -0.0153, ..., 0.0437, 0.0323, 0.0080], + [-0.0362, -0.0189, -0.0364, ..., 0.0277, -0.0046, 0.0186], + ..., + [-0.0329, -0.0115, -0.0454, ..., -0.0068, -0.0127, 0.0263], + [ 0.0364, 0.0096, -0.0025, ..., -0.0187, -0.0264, 0.0070], + [-0.0548, 0.0071, 0.0215, ..., -0.0524, 0.0055, -0.0256]], + device='cuda:0'), grad: tensor([[-1.4095e-03, 0.0000e+00, 1.8847e-04, ..., 2.6107e-04, + -7.4148e-04, 2.7716e-05], + [-4.8447e-04, 0.0000e+00, -1.3294e-03, ..., -6.9475e-04, + -2.6093e-03, 4.7684e-05], + [ 3.9846e-05, 0.0000e+00, 8.2374e-05, ..., -7.2670e-03, + -1.0414e-03, -7.2145e-04], + ..., + [ 3.7956e-04, 0.0000e+00, 5.7030e-04, ..., 3.3417e-03, + 1.4067e-03, -2.6727e-04], + [ 9.4748e-04, 0.0000e+00, 1.0208e-02, ..., 6.8512e-03, + 6.8016e-03, 1.8370e-04], + [ 5.3501e-04, 0.0000e+00, -1.0048e-02, ..., -4.3411e-03, + -6.8398e-03, 1.4925e-04]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0079, -0.0279, 0.0083, -0.0210, 0.0147, 0.0018, 0.0197, -0.0114, + -0.0227, 0.0021], device='cuda:0'), grad: tensor([-0.0051, -0.0022, -0.0031, 0.0017, 0.0056, 0.0016, 0.0007, 0.0033, + 0.0233, -0.0259], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 262.38, cls_loss 0.0920 cls_loss_mapping 0.1155 cls_loss_causal 1.0707 re_mapping 0.0534 re_causal 0.1240 /// teacc 97.04 lr 0.00010000 +Epoch 12, weight, value: tensor([[ 0.0005, -0.0223, -0.0104, ..., -0.0439, -0.0319, -0.0300], + [ 0.0130, -0.0064, -0.0156, ..., 0.0435, 0.0334, 0.0079], + [-0.0364, -0.0189, -0.0375, ..., 0.0288, -0.0039, 0.0184], + ..., + [-0.0349, -0.0115, -0.0465, ..., -0.0066, -0.0142, 0.0273], + [ 0.0375, 0.0096, -0.0020, ..., -0.0186, -0.0277, 0.0065], + [-0.0560, 0.0071, 0.0215, ..., -0.0536, 0.0056, -0.0266]], + device='cuda:0'), grad: tensor([[ 5.4464e-06, 0.0000e+00, 8.2159e-04, ..., 7.2432e-04, + 6.7568e-04, 9.8407e-05], + [-8.0645e-05, 0.0000e+00, -2.9278e-04, ..., -7.1764e-04, + -7.9393e-04, 3.4928e-04], + [ 1.6302e-05, 0.0000e+00, 2.3232e-03, ..., 7.1001e-04, + 2.0008e-03, 4.8423e-04], + ..., + [ 1.3165e-05, 0.0000e+00, 3.1376e-03, ..., 2.7447e-03, + 7.1383e-04, 9.5320e-04], + [ 1.8328e-05, 0.0000e+00, -3.6831e-03, ..., -2.0638e-03, + 7.7915e-04, 4.6539e-04], + [ 1.5402e-04, 0.0000e+00, 2.3041e-03, ..., 4.4899e-03, + 1.2684e-03, 1.8892e-03]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0078, -0.0280, 0.0086, -0.0208, 0.0145, 0.0016, 0.0198, -0.0114, + -0.0227, 0.0021], device='cuda:0'), grad: tensor([-0.0015, 0.0003, 0.0051, -0.0063, -0.0071, 0.0043, -0.0005, 0.0052, + -0.0061, 0.0068], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 278.61, cls_loss 0.0944 cls_loss_mapping 0.1198 cls_loss_causal 1.1229 re_mapping 0.0485 re_causal 0.1233 /// teacc 97.52 lr 0.00010000 +Epoch 13, weight, value: tensor([[ 0.0016, -0.0223, -0.0118, ..., -0.0459, -0.0326, -0.0320], + [ 0.0126, -0.0064, -0.0154, ..., 0.0441, 0.0350, 0.0068], + [-0.0375, -0.0189, -0.0393, ..., 0.0288, -0.0043, 0.0181], + ..., + [-0.0347, -0.0115, -0.0469, ..., -0.0063, -0.0150, 0.0286], + [ 0.0388, 0.0096, -0.0013, ..., -0.0180, -0.0285, 0.0056], + [-0.0565, 0.0071, 0.0211, ..., -0.0551, 0.0055, -0.0278]], + device='cuda:0'), grad: tensor([[ 3.1829e-04, 0.0000e+00, 6.0844e-04, ..., 1.4009e-03, + 1.6296e-04, 2.5168e-05], + [ 5.7906e-05, 0.0000e+00, -3.0255e-04, ..., -3.4022e-04, + -6.2799e-04, 9.5785e-05], + [ 1.5342e-04, 0.0000e+00, 5.0735e-04, ..., -1.8632e-04, + -5.5361e-04, 2.5010e-04], + ..., + [-1.2743e-04, 0.0000e+00, 2.8539e-04, ..., -5.7173e-04, + 3.7432e-04, -8.7738e-04], + [-3.8576e-04, 0.0000e+00, 2.7714e-03, ..., 3.3331e-04, + 1.9331e-03, 6.0201e-05], + [ 2.4652e-04, 0.0000e+00, 6.7902e-04, ..., 1.0383e-04, + -2.0370e-03, 1.2255e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0077, -0.0278, 0.0083, -0.0209, 0.0146, 0.0016, 0.0194, -0.0113, + -0.0224, 0.0020], device='cuda:0'), grad: tensor([ 0.0030, -0.0003, 0.0006, 0.0024, 0.0013, -0.0071, 0.0020, -0.0005, + -0.0008, -0.0007], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 278.72, cls_loss 0.0749 cls_loss_mapping 0.0972 cls_loss_causal 1.0655 re_mapping 0.0466 re_causal 0.1147 /// teacc 97.66 lr 0.00010000 +Epoch 14, weight, value: tensor([[ 0.0017, -0.0223, -0.0130, ..., -0.0473, -0.0332, -0.0341], + [ 0.0123, -0.0064, -0.0153, ..., 0.0444, 0.0360, 0.0067], + [-0.0381, -0.0189, -0.0408, ..., 0.0292, -0.0036, 0.0173], + ..., + [-0.0349, -0.0115, -0.0479, ..., -0.0064, -0.0161, 0.0294], + [ 0.0398, 0.0096, -0.0009, ..., -0.0178, -0.0291, 0.0050], + [-0.0576, 0.0071, 0.0210, ..., -0.0562, 0.0057, -0.0281]], + device='cuda:0'), grad: tensor([[-5.3883e-04, 0.0000e+00, 3.7980e-04, ..., 3.5071e-04, + 1.2740e-05, 1.8582e-05], + [ 1.4801e-03, 0.0000e+00, 1.8578e-03, ..., 2.1744e-03, + -9.0790e-04, 6.2823e-05], + [ 6.2227e-04, 0.0000e+00, 1.5860e-03, ..., 1.8559e-03, + 2.6464e-04, 9.5725e-05], + ..., + [-7.6175e-05, 0.0000e+00, 4.1986e-04, ..., -9.6607e-04, + 3.4904e-04, -6.8617e-04], + [-3.9825e-03, 0.0000e+00, -1.0170e-02, ..., -1.1017e-02, + -7.7069e-05, 2.8148e-05], + [ 4.4441e-04, 0.0000e+00, 4.9324e-03, ..., 4.3259e-03, + -1.1665e-04, 1.7023e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0076, -0.0279, 0.0084, -0.0205, 0.0142, 0.0019, 0.0192, -0.0114, + -0.0225, 0.0019], device='cuda:0'), grad: tensor([-0.0008, 0.0027, 0.0023, 0.0038, 0.0008, -0.0018, 0.0017, -0.0005, + -0.0165, 0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 262.33, cls_loss 0.0878 cls_loss_mapping 0.1078 cls_loss_causal 1.0380 re_mapping 0.0432 re_causal 0.1055 /// teacc 97.58 lr 0.00010000 +Epoch 15, weight, value: tensor([[ 0.0021, -0.0223, -0.0141, ..., -0.0488, -0.0337, -0.0358], + [ 0.0128, -0.0064, -0.0155, ..., 0.0445, 0.0369, 0.0066], + [-0.0392, -0.0189, -0.0420, ..., 0.0300, -0.0025, 0.0170], + ..., + [-0.0363, -0.0115, -0.0480, ..., -0.0063, -0.0173, 0.0304], + [ 0.0420, 0.0096, -0.0006, ..., -0.0175, -0.0301, 0.0042], + [-0.0595, 0.0071, 0.0209, ..., -0.0570, 0.0052, -0.0286]], + device='cuda:0'), grad: tensor([[ 3.3379e-04, 0.0000e+00, 5.1403e-04, ..., 1.1263e-03, + 1.1034e-03, 7.6694e-07], + [ 3.5357e-04, 0.0000e+00, 4.1509e-04, ..., 7.5293e-04, + 1.0228e-04, 1.3253e-06], + [ 1.1539e-03, 0.0000e+00, 1.4572e-03, ..., -8.4686e-04, + 1.0157e-03, -1.7062e-05], + ..., + [ 2.4259e-04, 0.0000e+00, 1.1501e-03, ..., 1.5516e-03, + 6.8617e-04, 4.2133e-06], + [-2.0580e-03, 0.0000e+00, -7.8249e-04, ..., -2.8801e-03, + 6.5422e-04, 1.0729e-06], + [ 4.6277e-04, 0.0000e+00, 9.4843e-04, ..., 1.1625e-03, + 9.8705e-04, 5.4948e-08]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0079, -0.0278, 0.0088, -0.0205, 0.0141, 0.0021, 0.0191, -0.0118, + -0.0225, 0.0020], device='cuda:0'), grad: tensor([ 0.0028, 0.0012, -0.0005, 0.0091, -0.0082, -0.0145, 0.0031, 0.0064, + 0.0008, -0.0001], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 279.63, cls_loss 0.0727 cls_loss_mapping 0.0954 cls_loss_causal 1.0262 re_mapping 0.0412 re_causal 0.1041 /// teacc 97.71 lr 0.00010000 +Epoch 16, weight, value: tensor([[ 2.9271e-03, -2.2335e-02, -1.5212e-02, ..., -5.0395e-02, + -3.3793e-02, -3.7773e-02], + [ 1.3338e-02, -6.4199e-03, -1.5480e-02, ..., 4.4424e-02, + 3.7605e-02, 5.9012e-03], + [-3.9176e-02, -1.8948e-02, -4.3374e-02, ..., 3.0824e-02, + -1.9028e-03, 1.6749e-02], + ..., + [-3.7710e-02, -1.1540e-02, -4.7206e-02, ..., -6.4257e-03, + -1.7850e-02, 3.0798e-02], + [ 4.2785e-02, 9.5671e-03, 2.2658e-05, ..., -1.7307e-02, + -3.1190e-02, 3.9399e-03], + [-6.0666e-02, 7.1317e-03, 2.1243e-02, ..., -5.7402e-02, + 5.3516e-03, -2.9453e-02]], device='cuda:0'), grad: tensor([[-1.5755e-03, 0.0000e+00, 2.9588e-04, ..., 1.2455e-03, + 6.3038e-04, 8.5175e-05], + [ 7.9918e-04, 0.0000e+00, 9.3126e-04, ..., 1.8797e-03, + 1.7083e-04, 2.4402e-04], + [ 1.3390e-03, 0.0000e+00, 2.1172e-03, ..., -1.4549e-02, + 6.3848e-04, 9.4700e-04], + ..., + [ 1.4095e-03, 0.0000e+00, 7.7200e-04, ..., 2.4548e-03, + 2.5387e-03, -1.6832e-04], + [-7.5436e-04, 0.0000e+00, -2.9163e-03, ..., -1.5612e-03, + -8.7309e-04, 1.3649e-04], + [-9.8801e-04, 0.0000e+00, 2.8687e-03, ..., 3.0670e-03, + -4.0207e-03, 1.7416e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0072, -0.0280, 0.0091, -0.0206, 0.0140, 0.0018, 0.0189, -0.0114, + -0.0225, 0.0017], device='cuda:0'), grad: tensor([ 0.0021, 0.0030, -0.0113, 0.0034, 0.0012, 0.0042, 0.0033, 0.0085, + -0.0066, -0.0078], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 279.05, cls_loss 0.0821 cls_loss_mapping 0.0971 cls_loss_causal 1.0063 re_mapping 0.0395 re_causal 0.0973 /// teacc 97.84 lr 0.00010000 +Epoch 17, weight, value: tensor([[ 0.0032, -0.0233, -0.0162, ..., -0.0521, -0.0342, -0.0400], + [ 0.0140, -0.0086, -0.0153, ..., 0.0444, 0.0381, 0.0053], + [-0.0397, -0.0121, -0.0448, ..., 0.0314, -0.0007, 0.0163], + ..., + [-0.0393, -0.0133, -0.0470, ..., -0.0061, -0.0187, 0.0314], + [ 0.0435, 0.0094, 0.0005, ..., -0.0168, -0.0320, 0.0038], + [-0.0627, 0.0065, 0.0208, ..., -0.0589, 0.0054, -0.0306]], + device='cuda:0'), grad: tensor([[ 1.0622e-04, 1.5929e-05, 1.6284e-04, ..., 2.5058e-04, + 3.2735e-04, 1.3387e-04], + [ 3.7861e-04, 2.4121e-06, 1.8702e-03, ..., 3.3283e-03, + 4.3845e-04, 1.3885e-03], + [-4.7874e-04, 1.0086e-06, -1.1322e-02, ..., -2.0889e-02, + -2.7523e-03, -9.0256e-03], + ..., + [ 4.2796e-04, 3.1125e-06, 7.3433e-03, ..., 1.3321e-02, + 1.6584e-03, 5.1079e-03], + [-2.1863e-04, 1.1399e-06, 7.9989e-05, ..., 1.4839e-03, + -1.3840e-04, 5.3883e-04], + [ 1.2350e-04, 3.2643e-07, 1.6427e-04, ..., 3.6788e-04, + 7.4804e-05, 1.7965e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0073, -0.0280, 0.0093, -0.0206, 0.0142, 0.0024, 0.0182, -0.0114, + -0.0223, 0.0011], device='cuda:0'), grad: tensor([ 4.0221e-04, 3.1872e-03, -1.6006e-02, 1.6384e-03, 1.2045e-03, + 4.5133e-04, -3.0065e-04, 8.5068e-03, -3.6389e-05, 9.5892e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 278.68, cls_loss 0.0692 cls_loss_mapping 0.0852 cls_loss_causal 0.9558 re_mapping 0.0389 re_causal 0.0921 /// teacc 98.14 lr 0.00010000 +Epoch 18, weight, value: tensor([[ 0.0034, -0.0281, -0.0170, ..., -0.0536, -0.0348, -0.0413], + [ 0.0140, -0.0052, -0.0149, ..., 0.0448, 0.0391, 0.0046], + [-0.0405, -0.0111, -0.0463, ..., 0.0316, -0.0004, 0.0161], + ..., + [-0.0408, -0.0174, -0.0472, ..., -0.0060, -0.0194, 0.0320], + [ 0.0445, 0.0047, 0.0012, ..., -0.0166, -0.0328, 0.0041], + [-0.0629, 0.0023, 0.0203, ..., -0.0601, 0.0056, -0.0315]], + device='cuda:0'), grad: tensor([[ 8.4415e-06, 2.2352e-08, 1.1241e-04, ..., 1.4734e-04, + 8.1360e-05, 8.0764e-06], + [-3.4750e-05, -1.6764e-06, 7.6532e-04, ..., 6.4421e-04, + -4.0412e-04, 9.5546e-05], + [ 3.6031e-05, 1.5926e-07, 3.7193e-04, ..., 4.8876e-04, + 3.2473e-04, 1.8850e-05], + ..., + [ 2.4915e-05, 4.4238e-08, -4.7760e-03, ..., -4.2572e-03, + 2.5272e-04, -6.4135e-04], + [ 2.6301e-05, 6.6031e-07, 2.6665e-03, ..., 2.2373e-03, + 3.5906e-04, 2.3258e-04], + [ 1.7807e-05, 3.4459e-08, 2.6560e-04, ..., 5.9795e-04, + -7.1526e-05, 8.8155e-05]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0075, -0.0280, 0.0088, -0.0202, 0.0146, 0.0022, 0.0181, -0.0114, + -0.0223, 0.0011], device='cuda:0'), grad: tensor([ 2.2817e-04, 8.4925e-04, 7.6199e-04, 3.8929e-03, 3.3975e-04, + -3.6201e-03, 9.8571e-06, -6.5651e-03, 3.4313e-03, 6.6566e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 261.99, cls_loss 0.0606 cls_loss_mapping 0.0783 cls_loss_causal 0.9925 re_mapping 0.0366 re_causal 0.0930 /// teacc 97.85 lr 0.00010000 +Epoch 19, weight, value: tensor([[ 0.0039, -0.0298, -0.0182, ..., -0.0551, -0.0355, -0.0428], + [ 0.0135, -0.0060, -0.0148, ..., 0.0447, 0.0393, 0.0043], + [-0.0403, -0.0082, -0.0468, ..., 0.0325, 0.0010, 0.0160], + ..., + [-0.0414, -0.0205, -0.0479, ..., -0.0059, -0.0206, 0.0334], + [ 0.0451, 0.0031, 0.0014, ..., -0.0169, -0.0339, 0.0032], + [-0.0639, 0.0037, 0.0204, ..., -0.0610, 0.0057, -0.0326]], + device='cuda:0'), grad: tensor([[ 6.4790e-05, 1.8207e-06, 6.0409e-05, ..., 2.4939e-04, + 5.0831e-04, 2.7999e-05], + [ 2.0294e-03, 2.2640e-03, 3.9458e-04, ..., 1.1299e-02, + 1.1597e-02, 1.4111e-05], + [-2.2202e-03, -2.3098e-03, 2.1970e-04, ..., -1.2459e-02, + -1.2932e-02, -1.2361e-05], + ..., + [-4.3678e-04, 2.6345e-05, -1.0614e-03, ..., -2.4581e-04, + -1.3900e-04, -1.4937e-04], + [ 3.7193e-05, 2.0918e-06, 1.0371e-04, ..., 1.7118e-04, + 2.4045e-04, 9.0227e-06], + [ 2.3639e-04, 2.4028e-07, 6.1178e-04, ..., 6.7377e-04, + 3.6430e-04, 8.0585e-05]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0074, -0.0280, 0.0096, -0.0204, 0.0143, 0.0028, 0.0179, -0.0112, + -0.0228, 0.0008], device='cuda:0'), grad: tensor([ 0.0012, 0.0116, -0.0111, 0.0005, 0.0013, -0.0052, 0.0026, -0.0038, + 0.0006, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 261.97, cls_loss 0.0531 cls_loss_mapping 0.0700 cls_loss_causal 0.9371 re_mapping 0.0353 re_causal 0.0912 /// teacc 97.92 lr 0.00010000 +Epoch 20, weight, value: tensor([[ 0.0046, -0.0391, -0.0189, ..., -0.0558, -0.0368, -0.0454], + [ 0.0135, -0.0071, -0.0155, ..., 0.0443, 0.0394, 0.0037], + [-0.0399, -0.0019, -0.0468, ..., 0.0332, 0.0020, 0.0156], + ..., + [-0.0415, -0.0305, -0.0483, ..., -0.0059, -0.0220, 0.0345], + [ 0.0452, -0.0007, 0.0018, ..., -0.0168, -0.0347, 0.0032], + [-0.0652, 0.0076, 0.0206, ..., -0.0618, 0.0067, -0.0331]], + device='cuda:0'), grad: tensor([[ 1.3411e-04, 3.3863e-06, 3.6597e-04, ..., 5.0163e-04, + 2.6798e-04, 1.0937e-04], + [ 1.9729e-04, -7.6666e-06, 2.8419e-04, ..., 2.8658e-04, + -6.9737e-05, 1.1289e-04], + [ 1.9722e-03, -4.5300e-05, 8.6403e-04, ..., 1.6129e-02, + 8.5373e-03, 3.0875e-04], + ..., + [-9.7036e-04, 1.5944e-05, 9.2030e-05, ..., -3.8767e-04, + 1.9848e-04, -1.6155e-03], + [-1.4839e-03, 5.3644e-06, -5.2605e-03, ..., -3.6888e-03, + -2.2984e-03, 8.5652e-05], + [ 5.6362e-04, 7.1600e-06, 1.6813e-03, ..., 1.0338e-03, + 7.2670e-03, 6.2132e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0077, -0.0286, 0.0101, -0.0202, 0.0140, 0.0022, 0.0184, -0.0113, + -0.0227, 0.0013], device='cuda:0'), grad: tensor([ 0.0011, 0.0005, 0.0157, -0.0131, -0.0097, 0.0027, 0.0008, -0.0020, + -0.0128, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 262.13, cls_loss 0.0633 cls_loss_mapping 0.0785 cls_loss_causal 0.9255 re_mapping 0.0324 re_causal 0.0829 /// teacc 98.14 lr 0.00010000 +Epoch 21, weight, value: tensor([[ 5.1018e-03, -4.2474e-02, -1.9843e-02, ..., -5.7534e-02, + -3.7260e-02, -4.7040e-02], + [ 1.3454e-02, -8.9705e-03, -1.5220e-02, ..., 4.4096e-02, + 3.9606e-02, 3.6152e-03], + [-4.1687e-02, 3.9066e-05, -4.7625e-02, ..., 3.3249e-02, + 2.6951e-03, 1.4517e-02], + ..., + [-4.2111e-02, -2.8523e-02, -4.8739e-02, ..., -5.7513e-03, + -2.2157e-02, 3.5070e-02], + [ 4.6602e-02, -1.8864e-03, 2.1780e-03, ..., -1.6189e-02, + -3.5679e-02, 3.4759e-03], + [-6.5758e-02, 5.3955e-03, 2.0645e-02, ..., -6.2398e-02, + 6.6634e-03, -3.3677e-02]], device='cuda:0'), grad: tensor([[-7.1287e-05, 7.7188e-06, 1.0215e-05, ..., 4.0352e-05, + -3.4750e-05, 5.3234e-06], + [ 8.9034e-06, 1.1832e-05, -1.3657e-05, ..., 3.9965e-05, + 2.0787e-05, 1.4573e-05], + [ 5.5850e-05, -1.1498e-04, 1.1164e-04, ..., -1.4198e-04, + -1.8001e-04, -1.4074e-05], + ..., + [ 6.8098e-06, 2.7835e-05, 1.4506e-05, ..., -1.9848e-04, + 1.5986e-04, -2.0206e-04], + [-4.5933e-06, 2.1145e-05, -1.2958e-04, ..., -6.1035e-05, + 1.1992e-04, 1.0863e-05], + [ 3.8296e-05, 8.9183e-06, 5.2500e-04, ..., 7.3004e-04, + 2.9430e-03, 8.1658e-05]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0077, -0.0288, 0.0096, -0.0203, 0.0145, 0.0025, 0.0179, -0.0111, + -0.0225, 0.0012], device='cuda:0'), grad: tensor([-1.5593e-04, 9.9361e-05, 2.9251e-05, 5.6839e-04, -7.6141e-03, + -4.2701e-04, -2.4843e-04, -1.4412e-04, 1.3328e-04, 7.7629e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 261.91, cls_loss 0.0411 cls_loss_mapping 0.0577 cls_loss_causal 0.8953 re_mapping 0.0320 re_causal 0.0870 /// teacc 97.95 lr 0.00010000 +Epoch 22, weight, value: tensor([[ 5.5057e-03, -4.8713e-02, -2.0457e-02, ..., -5.8301e-02, + -3.7313e-02, -4.7785e-02], + [ 1.3723e-02, -6.9754e-03, -1.4377e-02, ..., 4.4885e-02, + 4.0666e-02, 3.3136e-03], + [-4.1866e-02, 6.5748e-05, -4.9471e-02, ..., 3.3447e-02, + 2.2865e-03, 1.4338e-02], + ..., + [-4.3079e-02, -3.0418e-02, -4.8271e-02, ..., -5.4875e-03, + -2.2548e-02, 3.6107e-02], + [ 4.7456e-02, 2.1532e-04, 2.2991e-03, ..., -1.5973e-02, + -3.6440e-02, 2.9919e-03], + [-6.6437e-02, 5.4790e-03, 2.1240e-02, ..., -6.3581e-02, + 6.8969e-03, -3.4409e-02]], device='cuda:0'), grad: tensor([[-3.7074e-04, -1.8394e-04, 3.3569e-04, ..., 4.8280e-04, + 2.7716e-05, 3.2037e-06], + [ 6.5088e-05, -8.5950e-05, -3.3236e-04, ..., -2.5463e-04, + -6.9427e-04, 1.1601e-05], + [ 3.6144e-04, -1.4710e-04, 2.9802e-04, ..., 4.2319e-04, + 3.7336e-04, 1.4520e-04], + ..., + [ 1.5080e-04, 2.4170e-05, 1.6344e-04, ..., -6.8140e-04, + -4.5419e-04, -3.6979e-04], + [ 5.7840e-04, 1.4257e-04, 7.3862e-04, ..., 9.1553e-04, + 3.9601e-04, 3.0011e-05], + [ 8.3876e-04, 2.7561e-04, 1.4067e-03, ..., 1.2503e-03, + 1.8978e-03, 9.8109e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0075, -0.0285, 0.0093, -0.0210, 0.0143, 0.0027, 0.0182, -0.0107, + -0.0224, 0.0011], device='cuda:0'), grad: tensor([ 3.9577e-05, -6.1321e-04, 1.9302e-03, -8.4457e-03, -2.4300e-03, + 5.8250e-03, 6.4659e-04, -1.3876e-03, 1.8063e-03, 2.6245e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 262.43, cls_loss 0.0422 cls_loss_mapping 0.0582 cls_loss_causal 0.8725 re_mapping 0.0304 re_causal 0.0806 /// teacc 98.14 lr 0.00010000 +Epoch 23, weight, value: tensor([[ 0.0056, -0.0546, -0.0213, ..., -0.0602, -0.0383, -0.0509], + [ 0.0135, -0.0058, -0.0137, ..., 0.0453, 0.0419, 0.0030], + [-0.0428, 0.0011, -0.0510, ..., 0.0338, 0.0027, 0.0138], + ..., + [-0.0433, -0.0316, -0.0490, ..., -0.0057, -0.0237, 0.0371], + [ 0.0478, 0.0023, 0.0023, ..., -0.0162, -0.0374, 0.0024], + [-0.0668, 0.0062, 0.0216, ..., -0.0639, 0.0071, -0.0356]], + device='cuda:0'), grad: tensor([[ 2.8208e-05, -7.1973e-06, 1.0729e-04, ..., 7.6354e-05, + -1.6713e-04, 1.5542e-05], + [ 8.7380e-05, -2.8208e-05, -1.6456e-06, ..., -6.6340e-05, + -4.2510e-04, 9.8288e-05], + [ 8.0287e-05, 7.0184e-06, 2.4557e-04, ..., 2.9325e-04, + 1.1563e-04, 9.6977e-05], + ..., + [ 2.3842e-04, -1.8969e-05, 1.2436e-03, ..., 7.2050e-04, + 2.2709e-04, 1.4019e-04], + [-1.2657e-06, 2.2035e-06, 9.6560e-05, ..., -1.1545e-04, + 1.5140e-04, 1.9073e-05], + [ 3.2258e-04, 1.3649e-05, 6.4468e-04, ..., 9.7656e-04, + -1.5962e-04, 2.4176e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0078, -0.0282, 0.0095, -0.0205, 0.0148, 0.0022, 0.0180, -0.0111, + -0.0227, 0.0013], device='cuda:0'), grad: tensor([-2.8419e-04, -3.5375e-05, 5.3072e-04, -4.1809e-03, -8.9049e-05, + 9.5844e-04, 3.3402e-04, 1.4086e-03, 3.9458e-04, 9.6416e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 262.18, cls_loss 0.0496 cls_loss_mapping 0.0644 cls_loss_causal 0.8851 re_mapping 0.0295 re_causal 0.0766 /// teacc 98.10 lr 0.00010000 +Epoch 24, weight, value: tensor([[ 0.0060, -0.0574, -0.0220, ..., -0.0614, -0.0390, -0.0522], + [ 0.0129, -0.0070, -0.0138, ..., 0.0448, 0.0420, 0.0025], + [-0.0430, 0.0027, -0.0520, ..., 0.0343, 0.0036, 0.0135], + ..., + [-0.0440, -0.0316, -0.0493, ..., -0.0058, -0.0252, 0.0378], + [ 0.0490, 0.0015, 0.0030, ..., -0.0157, -0.0382, 0.0022], + [-0.0681, 0.0049, 0.0211, ..., -0.0650, 0.0081, -0.0365]], + device='cuda:0'), grad: tensor([[ 5.5462e-05, 4.6945e-04, 6.0940e-04, ..., 7.9584e-04, + 4.6682e-04, 2.5094e-05], + [ 1.4091e-04, 6.9678e-05, 6.8092e-04, ..., 1.0576e-03, + 2.3210e-04, 7.4089e-05], + [ 3.2735e-04, -1.2803e-04, 9.3603e-04, ..., -7.0453e-05, + 8.9049e-05, 3.3402e-04], + ..., + [-1.8225e-03, -3.1114e-04, 1.1665e-04, ..., -3.3054e-03, + 3.4165e-04, -1.5354e-03], + [ 8.0919e-04, 9.3231e-03, 1.1253e-02, ..., 1.2581e-02, + 4.5280e-03, 3.6144e-04], + [ 2.0623e-04, -7.6408e-03, -8.3313e-03, ..., -7.4196e-03, + 5.2490e-03, 2.2781e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0079, -0.0288, 0.0097, -0.0204, 0.0144, 0.0025, 0.0181, -0.0112, + -0.0226, 0.0015], device='cuda:0'), grad: tensor([ 0.0008, 0.0013, 0.0019, -0.0071, -0.0111, 0.0046, 0.0007, -0.0074, + 0.0321, -0.0157], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 279.15, cls_loss 0.0410 cls_loss_mapping 0.0548 cls_loss_causal 0.8936 re_mapping 0.0282 re_causal 0.0758 /// teacc 98.32 lr 0.00010000 +Epoch 25, weight, value: tensor([[ 0.0063, -0.0587, -0.0228, ..., -0.0625, -0.0391, -0.0550], + [ 0.0126, -0.0070, -0.0131, ..., 0.0450, 0.0429, 0.0022], + [-0.0438, 0.0048, -0.0533, ..., 0.0348, 0.0042, 0.0131], + ..., + [-0.0444, -0.0334, -0.0496, ..., -0.0062, -0.0264, 0.0388], + [ 0.0494, 0.0002, 0.0033, ..., -0.0155, -0.0398, 0.0020], + [-0.0689, 0.0077, 0.0208, ..., -0.0659, 0.0091, -0.0370]], + device='cuda:0'), grad: tensor([[ 7.8902e-06, 8.9183e-06, 6.7174e-05, ..., 7.5810e-06, + -7.7009e-05, 3.0905e-05], + [ 2.8878e-05, 2.6226e-03, 1.8680e-04, ..., 4.3106e-03, + 6.3248e-03, 8.9526e-05], + [-2.5654e-03, -3.1776e-03, -6.0158e-03, ..., -1.5808e-02, + -1.2970e-02, 1.2207e-04], + ..., + [-2.4110e-05, 5.2422e-05, -1.8799e-04, ..., -7.5626e-04, + 1.1152e-04, -1.1377e-03], + [ 2.3270e-03, 2.4211e-04, 4.7760e-03, ..., 1.0147e-02, + 5.4893e-03, 2.9624e-05], + [ 1.0449e-04, 5.9381e-06, 5.1594e-04, ..., 6.6471e-04, + -5.1260e-05, 5.6887e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0078, -0.0286, 0.0097, -0.0203, 0.0139, 0.0024, 0.0184, -0.0116, + -0.0229, 0.0021], device='cuda:0'), grad: tensor([-0.0047, 0.0050, -0.0126, 0.0013, 0.0005, 0.0008, 0.0031, -0.0027, + 0.0069, 0.0025], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 262.36, cls_loss 0.0401 cls_loss_mapping 0.0559 cls_loss_causal 0.8670 re_mapping 0.0277 re_causal 0.0744 /// teacc 98.13 lr 0.00010000 +Epoch 26, weight, value: tensor([[ 6.3072e-03, -5.9249e-02, -2.3751e-02, ..., -6.3550e-02, + -3.9737e-02, -5.6704e-02], + [ 1.1861e-02, -8.1189e-03, -1.2558e-02, ..., 4.4919e-02, + 4.2930e-02, 1.7998e-03], + [-4.3228e-02, 6.8351e-03, -5.4552e-02, ..., 3.5124e-02, + 5.1489e-03, 1.2409e-02], + ..., + [-4.4607e-02, -3.4165e-02, -4.9596e-02, ..., -5.7494e-03, + -2.6601e-02, 3.9994e-02], + [ 4.9796e-02, 4.4689e-05, 3.6707e-03, ..., -1.5601e-02, + -4.1024e-02, 1.8276e-03], + [-7.0225e-02, 7.0874e-03, 2.0145e-02, ..., -6.7072e-02, + 9.3765e-03, -3.8391e-02]], device='cuda:0'), grad: tensor([[ 3.9153e-06, 7.9036e-05, 5.5611e-05, ..., 1.0777e-04, + 2.2459e-04, 2.3246e-06], + [ 1.2565e-04, 1.0920e-03, 2.6226e-03, ..., 5.5466e-03, + 3.1834e-03, 2.4867e-04], + [ 1.8597e-04, 1.3466e-03, 8.6260e-04, ..., 1.7099e-03, + 4.8752e-03, 2.6539e-05], + ..., + [-6.9797e-05, -1.6317e-03, -4.9133e-03, ..., -9.8801e-03, + -6.5575e-03, -3.4547e-04], + [-6.6423e-04, 1.3161e-04, -6.6662e-04, ..., -1.2054e-03, + 4.3416e-04, 3.2578e-06], + [ 2.7156e-04, -2.3346e-03, 1.4753e-03, ..., 3.1528e-03, + -4.2458e-03, 2.9072e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0078, -0.0288, 0.0096, -0.0200, 0.0138, 0.0025, 0.0183, -0.0106, + -0.0231, 0.0015], device='cuda:0'), grad: tensor([ 0.0004, 0.0080, 0.0077, 0.0021, 0.0023, 0.0009, 0.0004, -0.0130, + -0.0004, -0.0083], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 261.99, cls_loss 0.0354 cls_loss_mapping 0.0457 cls_loss_causal 0.8578 re_mapping 0.0265 re_causal 0.0736 /// teacc 98.17 lr 0.00010000 +Epoch 27, weight, value: tensor([[ 0.0072, -0.0601, -0.0244, ..., -0.0639, -0.0394, -0.0577], + [ 0.0110, -0.0093, -0.0122, ..., 0.0448, 0.0434, 0.0014], + [-0.0435, 0.0074, -0.0557, ..., 0.0356, 0.0051, 0.0124], + ..., + [-0.0460, -0.0327, -0.0496, ..., -0.0055, -0.0269, 0.0404], + [ 0.0509, 0.0010, 0.0044, ..., -0.0151, -0.0414, 0.0018], + [-0.0713, 0.0069, 0.0196, ..., -0.0682, 0.0091, -0.0389]], + device='cuda:0'), grad: tensor([[ 1.7405e-05, 2.4308e-06, 5.7310e-05, ..., 7.9930e-05, + 3.8713e-05, 5.6326e-06], + [-7.8917e-04, -4.1890e-04, -2.3136e-03, ..., -2.4567e-03, + -2.5787e-03, 1.7852e-05], + [-7.4971e-07, 4.0054e-05, 1.1196e-03, ..., 5.7125e-04, + 4.1842e-04, 1.7524e-05], + ..., + [ 4.4346e-05, 4.0054e-05, 2.5249e-04, ..., -4.6223e-05, + 1.8716e-04, -5.6148e-05], + [ 6.3515e-04, 2.2507e-04, 4.6468e-04, ..., 1.4544e-03, + 1.0576e-03, 5.5507e-06], + [ 4.5627e-05, 3.1684e-06, 2.6837e-05, ..., 2.2328e-04, + 6.1572e-05, -9.6440e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0073, -0.0290, 0.0097, -0.0202, 0.0142, 0.0026, 0.0177, -0.0106, + -0.0228, 0.0011], device='cuda:0'), grad: tensor([-2.9534e-05, -3.7022e-03, 1.5240e-03, -1.8013e-04, 1.2007e-03, + 2.3675e-04, 1.5044e-04, -6.3002e-05, 9.8515e-04, -1.2600e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 278.81, cls_loss 0.0345 cls_loss_mapping 0.0466 cls_loss_causal 0.8280 re_mapping 0.0265 re_causal 0.0713 /// teacc 98.44 lr 0.00010000 +Epoch 28, weight, value: tensor([[ 0.0076, -0.0609, -0.0251, ..., -0.0647, -0.0402, -0.0584], + [ 0.0109, -0.0092, -0.0126, ..., 0.0445, 0.0441, 0.0009], + [-0.0448, 0.0078, -0.0560, ..., 0.0356, 0.0053, 0.0120], + ..., + [-0.0470, -0.0335, -0.0492, ..., -0.0051, -0.0270, 0.0410], + [ 0.0511, 0.0010, 0.0047, ..., -0.0151, -0.0421, 0.0018], + [-0.0720, 0.0072, 0.0197, ..., -0.0686, 0.0097, -0.0392]], + device='cuda:0'), grad: tensor([[ 2.8033e-06, 7.7039e-06, 1.2152e-05, ..., 2.9683e-05, + -1.5751e-05, 1.2340e-06], + [ 3.5129e-06, -2.5351e-06, -5.7161e-05, ..., -2.6003e-05, + -8.7678e-05, 1.0088e-05], + [ 2.1517e-05, -3.1859e-05, 3.0458e-05, ..., 9.7677e-06, + -1.0155e-05, 7.3433e-05], + ..., + [-3.6418e-05, -5.4166e-06, 5.8126e-04, ..., -5.5730e-05, + 7.8630e-04, -1.5759e-04], + [ 8.7637e-07, 8.6948e-06, -4.3184e-05, ..., 4.7088e-06, + 5.0306e-05, 8.6278e-06], + [ 4.3474e-06, 2.1011e-06, -5.8603e-04, ..., -1.1277e-04, + -8.6594e-04, 8.7321e-06]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0079, -0.0293, 0.0094, -0.0199, 0.0140, 0.0027, 0.0179, -0.0104, + -0.0230, 0.0015], device='cuda:0'), grad: tensor([-2.7823e-04, -2.9668e-05, 6.9141e-05, 5.9557e-04, 3.7718e-04, + -3.8505e-04, -4.6670e-05, 3.9043e-03, 9.2864e-05, -4.2992e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 262.28, cls_loss 0.0383 cls_loss_mapping 0.0468 cls_loss_causal 0.8278 re_mapping 0.0245 re_causal 0.0644 /// teacc 98.26 lr 0.00010000 +Epoch 29, weight, value: tensor([[ 0.0077, -0.0621, -0.0258, ..., -0.0655, -0.0403, -0.0597], + [ 0.0105, -0.0097, -0.0119, ..., 0.0449, 0.0447, 0.0007], + [-0.0445, 0.0095, -0.0573, ..., 0.0359, 0.0061, 0.0120], + ..., + [-0.0465, -0.0334, -0.0502, ..., -0.0054, -0.0281, 0.0412], + [ 0.0514, 0.0007, 0.0052, ..., -0.0147, -0.0430, 0.0022], + [-0.0735, 0.0068, 0.0192, ..., -0.0696, 0.0093, -0.0402]], + device='cuda:0'), grad: tensor([[ 5.0753e-05, 1.5438e-05, 7.3016e-05, ..., 6.7353e-05, + 3.7456e-04, 6.2697e-06], + [ 4.6074e-05, -1.2830e-05, -2.3210e-04, ..., -4.2737e-05, + -3.2449e-04, 1.2159e-05], + [ 1.1760e-04, -2.8706e-04, 3.4595e-04, ..., 1.4961e-04, + 2.7084e-04, 1.9062e-04], + ..., + [-2.6107e-05, 1.6499e-04, 8.6248e-05, ..., -2.3484e-04, + 1.4055e-04, -2.9159e-04], + [ 2.0170e-04, 5.0068e-05, 3.0470e-04, ..., 4.0078e-04, + 2.1315e-04, 5.5254e-05], + [ 8.4519e-05, -3.2969e-06, 2.5368e-04, ..., 1.0896e-04, + -8.6784e-04, 1.3426e-05]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0072, -0.0291, 0.0098, -0.0198, 0.0134, 0.0028, 0.0178, -0.0105, + -0.0230, 0.0010], device='cuda:0'), grad: tensor([ 8.9979e-04, -6.0976e-05, 9.1028e-04, 5.4312e-04, 4.7636e-04, + -1.2932e-03, -5.2357e-04, 3.8326e-05, 8.7404e-04, -1.8663e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 262.22, cls_loss 0.0367 cls_loss_mapping 0.0467 cls_loss_causal 0.8258 re_mapping 0.0253 re_causal 0.0682 /// teacc 98.23 lr 0.00010000 +Epoch 30, weight, value: tensor([[ 0.0082, -0.0637, -0.0264, ..., -0.0661, -0.0404, -0.0606], + [ 0.0104, -0.0104, -0.0115, ..., 0.0448, 0.0447, 0.0003], + [-0.0451, 0.0112, -0.0575, ..., 0.0364, 0.0069, 0.0113], + ..., + [-0.0466, -0.0327, -0.0508, ..., -0.0054, -0.0290, 0.0418], + [ 0.0519, 0.0002, 0.0058, ..., -0.0145, -0.0435, 0.0028], + [-0.0740, 0.0071, 0.0187, ..., -0.0706, 0.0094, -0.0407]], + device='cuda:0'), grad: tensor([[ 7.0827e-07, 1.2897e-05, 3.8266e-05, ..., 1.7449e-05, + -3.7879e-05, 1.9129e-06], + [-1.4435e-06, 4.4912e-05, 1.4591e-04, ..., -1.2383e-05, + -9.3520e-05, 2.6315e-05], + [ 5.8524e-06, 4.9210e-04, 3.1161e-04, ..., 6.3848e-04, + 4.3440e-04, 7.0989e-05], + ..., + [ 1.4044e-06, 1.1496e-05, -5.3585e-05, ..., -4.8327e-04, + 9.6142e-05, -2.8396e-04], + [ 2.6077e-06, -3.9554e-04, 1.4246e-04, ..., -3.8123e-04, + -1.4150e-04, 1.3754e-05], + [ 6.5751e-06, -1.7178e-04, -4.7398e-04, ..., 1.8203e-04, + -3.9530e-04, 4.7207e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0074, -0.0290, 0.0100, -0.0199, 0.0136, 0.0031, 0.0177, -0.0105, + -0.0230, 0.0007], device='cuda:0'), grad: tensor([-3.4857e-04, 3.2330e-04, 1.5545e-03, 3.2973e-04, 3.1137e-04, + -1.9073e-04, 7.8201e-05, -1.0452e-03, -4.4727e-04, -5.6458e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 262.24, cls_loss 0.0331 cls_loss_mapping 0.0445 cls_loss_causal 0.8056 re_mapping 0.0250 re_causal 0.0670 /// teacc 98.40 lr 0.00010000 +Epoch 31, weight, value: tensor([[ 0.0088, -0.0646, -0.0272, ..., -0.0668, -0.0416, -0.0622], + [ 0.0096, -0.0116, -0.0114, ..., 0.0445, 0.0450, -0.0004], + [-0.0447, 0.0114, -0.0584, ..., 0.0366, 0.0069, 0.0109], + ..., + [-0.0473, -0.0319, -0.0506, ..., -0.0050, -0.0294, 0.0428], + [ 0.0524, 0.0012, 0.0063, ..., -0.0143, -0.0436, 0.0023], + [-0.0744, 0.0071, 0.0184, ..., -0.0711, 0.0102, -0.0411]], + device='cuda:0'), grad: tensor([[ 1.2994e-05, 3.0443e-05, 2.2054e-04, ..., 1.0896e-04, + 1.1665e-04, 3.0220e-05], + [ 1.7703e-04, -3.3474e-03, -4.2610e-03, ..., -2.8458e-03, + -7.6065e-03, 3.1567e-04], + [ 2.9159e-04, 2.4274e-05, 8.0681e-04, ..., 3.7937e-03, + 8.8882e-04, 5.5981e-04], + ..., + [-7.6437e-04, 9.0003e-05, -2.7142e-03, ..., -2.1420e-03, + 4.6802e-04, -1.6193e-03], + [ 9.7081e-06, 2.4643e-03, 3.1338e-03, ..., 2.5692e-03, + 5.7411e-03, 5.5939e-05], + [ 3.1024e-05, 1.5469e-03, 3.7527e-04, ..., 1.8823e-04, + 9.6130e-03, 2.0719e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0079, -0.0294, 0.0096, -0.0199, 0.0134, 0.0030, 0.0183, -0.0099, + -0.0228, 0.0007], device='cuda:0'), grad: tensor([ 0.0010, -0.0120, 0.0035, -0.0008, -0.0104, 0.0089, 0.0026, -0.0158, + 0.0102, 0.0129], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 278.28, cls_loss 0.0268 cls_loss_mapping 0.0358 cls_loss_causal 0.7893 re_mapping 0.0241 re_causal 0.0665 /// teacc 98.54 lr 0.00010000 +Epoch 32, weight, value: tensor([[ 0.0089, -0.0642, -0.0278, ..., -0.0676, -0.0421, -0.0639], + [ 0.0100, -0.0096, -0.0109, ..., 0.0448, 0.0461, -0.0009], + [-0.0450, 0.0117, -0.0594, ..., 0.0367, 0.0071, 0.0107], + ..., + [-0.0479, -0.0321, -0.0504, ..., -0.0046, -0.0309, 0.0442], + [ 0.0526, 0.0011, 0.0063, ..., -0.0145, -0.0447, 0.0019], + [-0.0750, 0.0055, 0.0182, ..., -0.0719, 0.0105, -0.0416]], + device='cuda:0'), grad: tensor([[ 3.1710e-05, 6.2227e-05, 3.4362e-05, ..., 8.5473e-05, + 2.5535e-04, 9.8869e-06], + [ 1.2189e-04, 2.7132e-04, -1.4365e-04, ..., 1.3411e-05, + 6.4898e-04, 1.2301e-05], + [ 2.4581e-04, 1.5688e-04, 2.7442e-04, ..., 6.8903e-04, + 6.4182e-04, 2.9624e-05], + ..., + [ 5.7779e-06, 8.6799e-06, 1.5271e-04, ..., -3.4285e-04, + 5.5790e-04, -1.3733e-04], + [-2.0361e-04, 5.6803e-05, -3.9861e-06, ..., -6.6328e-04, + 2.5392e-04, 3.3379e-05], + [ 1.9848e-05, 3.7588e-06, -1.3697e-04, ..., 2.0957e-04, + -8.0061e-04, -1.7309e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0077, -0.0289, 0.0093, -0.0202, 0.0138, 0.0028, 0.0183, -0.0096, + -0.0230, 0.0005], device='cuda:0'), grad: tensor([ 3.4022e-04, 7.1859e-04, 1.4277e-03, 4.6468e-04, -1.0977e-03, + -2.9707e-04, -1.9054e-03, 2.0909e-04, -3.5644e-05, 1.7536e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 31---------------------------------------------------- +epoch 31, time 278.37, cls_loss 0.0342 cls_loss_mapping 0.0480 cls_loss_causal 0.8259 re_mapping 0.0226 re_causal 0.0620 /// teacc 98.56 lr 0.00010000 +Epoch 33, weight, value: tensor([[ 0.0093, -0.0653, -0.0286, ..., -0.0685, -0.0423, -0.0648], + [ 0.0094, -0.0098, -0.0108, ..., 0.0451, 0.0469, -0.0016], + [-0.0445, 0.0128, -0.0598, ..., 0.0368, 0.0075, 0.0105], + ..., + [-0.0483, -0.0320, -0.0504, ..., -0.0045, -0.0317, 0.0454], + [ 0.0531, 0.0002, 0.0070, ..., -0.0143, -0.0457, 0.0015], + [-0.0759, 0.0050, 0.0179, ..., -0.0723, 0.0105, -0.0427]], + device='cuda:0'), grad: tensor([[ 2.3991e-06, 2.2694e-05, 2.6196e-05, ..., 1.5259e-04, + 6.4611e-05, -2.2376e-04], + [-6.0886e-05, -8.6334e-07, -3.3927e-04, ..., -2.0015e-04, + -6.5708e-04, 1.7250e-04], + [ 9.0823e-06, -3.3808e-04, -1.1998e-04, ..., -2.0428e-03, + -8.6880e-04, 4.3154e-04], + ..., + [ 1.0148e-05, -1.7524e-04, -1.4186e-04, ..., -1.5879e-03, + 1.0423e-05, -1.4668e-03], + [-1.0923e-05, 2.7394e-04, 5.3978e-04, ..., 1.8740e-03, + 7.9632e-04, 8.8871e-05], + [ 9.2685e-06, 2.5868e-05, 6.9201e-05, ..., 1.8370e-04, + 2.3276e-05, 7.6294e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0072, -0.0291, 0.0094, -0.0198, 0.0136, 0.0021, 0.0184, -0.0092, + -0.0229, 0.0001], device='cuda:0'), grad: tensor([-0.0007, -0.0005, -0.0021, 0.0017, 0.0006, -0.0019, 0.0008, -0.0013, + 0.0031, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 262.58, cls_loss 0.0348 cls_loss_mapping 0.0443 cls_loss_causal 0.7970 re_mapping 0.0225 re_causal 0.0618 /// teacc 98.35 lr 0.00010000 +Epoch 34, weight, value: tensor([[ 0.0092, -0.0656, -0.0295, ..., -0.0692, -0.0425, -0.0667], + [ 0.0090, -0.0097, -0.0104, ..., 0.0457, 0.0479, -0.0020], + [-0.0446, 0.0140, -0.0600, ..., 0.0370, 0.0076, 0.0105], + ..., + [-0.0483, -0.0324, -0.0509, ..., -0.0050, -0.0332, 0.0457], + [ 0.0531, -0.0007, 0.0070, ..., -0.0143, -0.0466, 0.0012], + [-0.0763, 0.0058, 0.0177, ..., -0.0728, 0.0116, -0.0436]], + device='cuda:0'), grad: tensor([[ 1.4409e-05, 4.8578e-06, 1.2450e-05, ..., 4.8310e-05, + 2.1696e-05, 2.4661e-05], + [ 1.9744e-05, 1.9193e-05, -1.8156e-04, ..., 2.3258e-04, + -1.3638e-04, 1.3924e-04], + [ 5.7310e-05, -1.5363e-05, 5.3853e-05, ..., -1.5354e-03, + -1.2226e-03, -2.8825e-04], + ..., + [-2.1851e-04, -5.9873e-05, -3.5316e-05, ..., 2.1899e-04, + 9.8705e-04, -4.1461e-04], + [ 1.6764e-05, 6.4038e-06, 9.2238e-06, ..., 5.0932e-05, + 1.1241e-04, 4.0740e-05], + [ 3.8087e-05, 1.2472e-05, 1.2383e-05, ..., 1.3828e-04, + -4.6104e-05, 1.0234e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0072, -0.0288, 0.0095, -0.0195, 0.0136, 0.0024, 0.0181, -0.0094, + -0.0234, 0.0001], device='cuda:0'), grad: tensor([-6.0177e-04, 2.6250e-04, -1.0777e-03, 9.2506e-04, -8.2779e-03, + 7.2765e-04, 6.9737e-05, 7.5493e-03, 3.1543e-04, 1.0788e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 262.07, cls_loss 0.0319 cls_loss_mapping 0.0387 cls_loss_causal 0.7821 re_mapping 0.0226 re_causal 0.0613 /// teacc 98.25 lr 0.00010000 +Epoch 35, weight, value: tensor([[ 0.0099, -0.0654, -0.0302, ..., -0.0696, -0.0431, -0.0682], + [ 0.0087, -0.0097, -0.0097, ..., 0.0462, 0.0488, -0.0025], + [-0.0445, 0.0148, -0.0608, ..., 0.0376, 0.0083, 0.0105], + ..., + [-0.0492, -0.0325, -0.0512, ..., -0.0053, -0.0346, 0.0464], + [ 0.0536, -0.0010, 0.0075, ..., -0.0140, -0.0470, 0.0014], + [-0.0770, 0.0055, 0.0172, ..., -0.0739, 0.0119, -0.0449]], + device='cuda:0'), grad: tensor([[ 5.0634e-05, -1.6019e-05, 3.9697e-04, ..., 3.9792e-04, + 2.9221e-05, 3.7044e-05], + [ 1.8343e-05, 1.1399e-05, 1.0377e-04, ..., 2.0337e-04, + -2.9132e-05, 5.1349e-05], + [ 1.4675e-04, -2.6315e-05, 4.2629e-04, ..., 8.5926e-04, + 2.0519e-05, 2.8920e-04], + ..., + [-7.0892e-06, 1.5765e-05, -4.0144e-05, ..., -1.3714e-03, + 1.9717e-04, -7.1144e-04], + [-4.9686e-04, 5.1856e-06, -8.9264e-04, ..., -1.3390e-03, + 4.8488e-05, 2.8342e-05], + [ 4.3094e-05, 8.1724e-07, -6.8188e-05, ..., 3.9124e-04, + -1.0192e-04, 5.9426e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0064, -0.0285, 0.0098, -0.0199, 0.0135, 0.0022, 0.0178, -0.0096, + -0.0231, -0.0002], device='cuda:0'), grad: tensor([ 0.0004, 0.0003, 0.0012, 0.0241, 0.0001, -0.0286, 0.0041, -0.0007, + -0.0008, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 262.65, cls_loss 0.0290 cls_loss_mapping 0.0428 cls_loss_causal 0.7885 re_mapping 0.0223 re_causal 0.0620 /// teacc 98.52 lr 0.00010000 +Epoch 36, weight, value: tensor([[ 0.0098, -0.0660, -0.0314, ..., -0.0710, -0.0437, -0.0699], + [ 0.0086, -0.0104, -0.0098, ..., 0.0463, 0.0490, -0.0029], + [-0.0450, 0.0182, -0.0611, ..., 0.0386, 0.0092, 0.0098], + ..., + [-0.0489, -0.0332, -0.0522, ..., -0.0057, -0.0356, 0.0472], + [ 0.0544, -0.0027, 0.0086, ..., -0.0135, -0.0482, 0.0015], + [-0.0782, 0.0055, 0.0172, ..., -0.0744, 0.0130, -0.0451]], + device='cuda:0'), grad: tensor([[ 1.2212e-05, 2.7064e-06, 3.5554e-05, ..., 2.9817e-05, + 2.9638e-05, -7.8557e-07], + [-2.2113e-05, 5.4948e-06, -6.7770e-05, ..., -4.3541e-05, + -3.3712e-04, 6.1952e-06], + [ 6.2704e-05, -2.4050e-05, 1.8704e-04, ..., 1.9526e-04, + 2.2441e-05, 1.2450e-05], + ..., + [ 2.4974e-05, 1.3664e-05, -7.3671e-05, ..., -1.2960e-03, + 1.4031e-04, -3.1590e-05], + [-6.3717e-05, -6.8210e-06, -3.0947e-04, ..., -3.1757e-04, + 2.3293e-04, 1.1787e-05], + [ 4.7147e-05, 1.3532e-06, -2.9135e-04, ..., 6.8665e-04, + -1.4172e-03, 2.6356e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0067, -0.0288, 0.0105, -0.0203, 0.0131, 0.0023, 0.0175, -0.0095, + -0.0228, 0.0001], device='cuda:0'), grad: tensor([ 3.8683e-05, -2.7990e-04, 2.4390e-04, 1.0376e-03, 1.7805e-03, + -6.6519e-05, 1.9515e-04, -1.6413e-03, -2.2161e-04, -1.0853e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 262.24, cls_loss 0.0240 cls_loss_mapping 0.0323 cls_loss_causal 0.7610 re_mapping 0.0206 re_causal 0.0588 /// teacc 98.37 lr 0.00010000 +Epoch 37, weight, value: tensor([[ 0.0099, -0.0665, -0.0321, ..., -0.0721, -0.0440, -0.0702], + [ 0.0085, -0.0100, -0.0093, ..., 0.0467, 0.0499, -0.0035], + [-0.0454, 0.0186, -0.0623, ..., 0.0384, 0.0091, 0.0095], + ..., + [-0.0491, -0.0341, -0.0523, ..., -0.0054, -0.0364, 0.0484], + [ 0.0549, -0.0017, 0.0089, ..., -0.0133, -0.0492, 0.0011], + [-0.0792, 0.0053, 0.0167, ..., -0.0755, 0.0132, -0.0466]], + device='cuda:0'), grad: tensor([[ 1.4231e-05, 1.5303e-05, 3.1590e-05, ..., 4.5598e-05, + 2.9311e-05, 7.2308e-06], + [ 5.8293e-05, 5.8800e-05, 2.5439e-04, ..., 3.4070e-04, + 3.1471e-05, 3.5197e-05], + [ 7.8678e-05, -1.7178e-04, 1.0556e-04, ..., -9.6917e-05, + -1.4770e-04, 8.4341e-05], + ..., + [-2.5257e-05, -8.6352e-06, 6.6662e-04, ..., 2.9469e-04, + 5.0843e-05, -2.7680e-04], + [ 8.8394e-05, 3.6776e-05, 3.0994e-04, ..., 2.0039e-04, + 1.6713e-04, 9.8422e-06], + [ 3.9983e-04, 7.9796e-06, 1.0519e-03, ..., 9.3889e-04, + -7.8976e-05, 7.2777e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0066, -0.0283, 0.0101, -0.0199, 0.0132, 0.0020, 0.0180, -0.0096, + -0.0230, -0.0004], device='cuda:0'), grad: tensor([ 7.6115e-05, 3.9768e-04, -1.3657e-05, -2.3880e-03, 1.7309e-04, + 7.3075e-05, 1.1069e-04, 3.1304e-04, 5.6458e-04, 6.9332e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 262.20, cls_loss 0.0231 cls_loss_mapping 0.0312 cls_loss_causal 0.7677 re_mapping 0.0210 re_causal 0.0595 /// teacc 98.45 lr 0.00010000 +Epoch 38, weight, value: tensor([[ 0.0095, -0.0680, -0.0331, ..., -0.0728, -0.0446, -0.0709], + [ 0.0083, -0.0119, -0.0095, ..., 0.0464, 0.0499, -0.0045], + [-0.0457, 0.0208, -0.0623, ..., 0.0389, 0.0102, 0.0093], + ..., + [-0.0487, -0.0345, -0.0519, ..., -0.0049, -0.0374, 0.0498], + [ 0.0551, -0.0012, 0.0091, ..., -0.0134, -0.0499, 0.0007], + [-0.0792, 0.0056, 0.0166, ..., -0.0762, 0.0134, -0.0478]], + device='cuda:0'), grad: tensor([[ 3.6448e-05, -5.7109e-06, 6.2644e-05, ..., 1.8167e-04, + 1.1545e-04, 4.5747e-06], + [ 2.9683e-05, -1.9062e-04, -1.1778e-03, ..., -1.3657e-03, + -1.8549e-03, -9.0837e-05], + [-9.1028e-04, 4.2349e-05, -6.3133e-04, ..., -6.1378e-03, + -1.4839e-03, -8.8274e-05], + ..., + [ 9.1612e-05, 2.1040e-05, 1.5059e-03, ..., 1.7366e-03, + 1.8206e-03, 7.5996e-05], + [-8.4925e-04, -3.4839e-05, -4.9829e-04, ..., -1.1988e-03, + -7.5758e-05, 2.6181e-05], + [ 4.1223e-04, 5.2273e-05, -6.0749e-04, ..., 4.0621e-05, + -7.7581e-04, 1.8510e-07]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0070, -0.0287, 0.0104, -0.0202, 0.0131, 0.0023, 0.0180, -0.0090, + -0.0231, -0.0004], device='cuda:0'), grad: tensor([ 0.0003, -0.0018, -0.0051, 0.0065, 0.0007, 0.0007, -0.0002, 0.0035, + -0.0044, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 262.27, cls_loss 0.0274 cls_loss_mapping 0.0372 cls_loss_causal 0.7886 re_mapping 0.0199 re_causal 0.0571 /// teacc 98.49 lr 0.00010000 +Epoch 39, weight, value: tensor([[ 0.0094, -0.0696, -0.0339, ..., -0.0738, -0.0451, -0.0716], + [ 0.0091, -0.0125, -0.0087, ..., 0.0466, 0.0505, -0.0046], + [-0.0467, 0.0223, -0.0632, ..., 0.0390, 0.0104, 0.0085], + ..., + [-0.0491, -0.0350, -0.0521, ..., -0.0052, -0.0382, 0.0506], + [ 0.0554, -0.0015, 0.0100, ..., -0.0127, -0.0507, 0.0015], + [-0.0799, 0.0066, 0.0160, ..., -0.0767, 0.0144, -0.0494]], + device='cuda:0'), grad: tensor([[ 2.2557e-06, 5.9843e-05, 2.9683e-05, ..., 7.0035e-05, + 1.8787e-04, 3.2634e-06], + [ 1.3523e-05, 1.4174e-04, 1.0777e-03, ..., 1.8930e-04, + 2.3060e-03, 1.4909e-05], + [ 2.8864e-05, -9.0456e-04, 1.5628e-04, ..., -5.6744e-04, + -2.4872e-03, 2.5854e-05], + ..., + [-8.0347e-04, -1.0788e-05, -1.7281e-03, ..., -1.3304e-03, + -3.0403e-03, -1.4114e-04], + [ 5.3078e-05, -2.1911e-04, -9.1553e-04, ..., -5.5122e-04, + -1.6594e-04, 5.4479e-05], + [ 7.2527e-04, 2.1505e-04, 1.4029e-03, ..., 1.9321e-03, + 8.1301e-04, 2.0459e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0073, -0.0286, 0.0102, -0.0201, 0.0125, 0.0022, 0.0176, -0.0085, + -0.0226, -0.0003], device='cuda:0'), grad: tensor([ 0.0002, 0.0026, -0.0014, -0.0003, 0.0010, 0.0001, 0.0009, -0.0234, + -0.0015, 0.0218], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 262.34, cls_loss 0.0265 cls_loss_mapping 0.0396 cls_loss_causal 0.7396 re_mapping 0.0201 re_causal 0.0566 /// teacc 98.49 lr 0.00010000 +Epoch 40, weight, value: tensor([[ 0.0107, -0.0706, -0.0341, ..., -0.0740, -0.0451, -0.0723], + [ 0.0082, -0.0125, -0.0084, ..., 0.0465, 0.0503, -0.0049], + [-0.0465, 0.0235, -0.0643, ..., 0.0391, 0.0109, 0.0080], + ..., + [-0.0491, -0.0358, -0.0527, ..., -0.0053, -0.0391, 0.0512], + [ 0.0557, -0.0024, 0.0102, ..., -0.0127, -0.0516, 0.0012], + [-0.0799, 0.0064, 0.0166, ..., -0.0771, 0.0153, -0.0502]], + device='cuda:0'), grad: tensor([[ 3.7309e-06, 2.4736e-05, 4.0293e-05, ..., 8.5890e-05, + 8.2135e-05, 1.2256e-06], + [ 1.7017e-05, -1.1330e-03, 2.4834e-03, ..., -5.4970e-03, + -3.9444e-03, 1.7971e-05], + [ 7.0296e-06, 9.4271e-04, 5.0694e-05, ..., 5.7106e-03, + 4.6005e-03, 7.7188e-06], + ..., + [-1.3448e-05, 4.3124e-05, 1.3202e-05, ..., -5.6654e-05, + 2.0361e-04, -1.9789e-04], + [ 1.8194e-05, 5.9813e-05, 8.6498e-04, ..., 5.4169e-04, + 5.6887e-04, 7.8797e-05], + [ 1.1362e-05, 2.0340e-05, 3.5644e-04, ..., 3.5572e-04, + 3.7098e-04, 5.2452e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0071, -0.0289, 0.0101, -0.0200, 0.0125, 0.0019, 0.0176, -0.0088, + -0.0229, 0.0006], device='cuda:0'), grad: tensor([ 0.0001, -0.0032, 0.0062, 0.0032, -0.0003, -0.0085, 0.0003, -0.0002, + 0.0021, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 280.55, cls_loss 0.0220 cls_loss_mapping 0.0311 cls_loss_causal 0.7567 re_mapping 0.0193 re_causal 0.0572 /// teacc 98.60 lr 0.00010000 +Epoch 41, weight, value: tensor([[ 0.0106, -0.0714, -0.0349, ..., -0.0747, -0.0455, -0.0736], + [ 0.0092, -0.0129, -0.0090, ..., 0.0459, 0.0505, -0.0060], + [-0.0467, 0.0251, -0.0649, ..., 0.0395, 0.0114, 0.0078], + ..., + [-0.0507, -0.0368, -0.0520, ..., -0.0050, -0.0396, 0.0518], + [ 0.0559, -0.0025, 0.0107, ..., -0.0125, -0.0518, 0.0008], + [-0.0806, 0.0060, 0.0162, ..., -0.0778, 0.0154, -0.0511]], + device='cuda:0'), grad: tensor([[-3.0208e-04, 2.4885e-05, 1.2803e-04, ..., 1.1152e-04, + -3.7842e-03, 6.1765e-06], + [ 7.2896e-05, 9.0301e-06, 5.2929e-04, ..., 1.4830e-04, + 1.4162e-04, 2.2411e-04], + [ 7.7248e-05, -6.9916e-05, 9.9480e-05, ..., -1.0335e-04, + -4.4256e-05, 1.5751e-05], + ..., + [ 1.2927e-06, 1.7717e-05, -3.1700e-03, ..., -5.0497e-04, + 8.3208e-05, -1.4772e-03], + [-5.2452e-04, -5.7638e-05, -6.6948e-04, ..., -5.5647e-04, + 1.2302e-04, 1.4886e-05], + [ 2.0564e-05, 5.4874e-06, 2.2049e-03, ..., 3.6383e-04, + -1.4484e-04, 1.0357e-03]], device='cuda:0') +Epoch 41, bias, value: tensor([-7.2204e-03, -2.8990e-02, 1.0248e-02, -1.9564e-02, 1.3015e-02, + 2.1794e-03, 1.6977e-02, -8.7730e-03, -2.2834e-02, -2.4634e-05], + device='cuda:0'), grad: tensor([-8.8577e-03, 1.5793e-03, 9.6083e-05, 8.4448e-04, 1.1358e-03, + 2.7204e-04, 7.7362e-03, -7.5455e-03, -8.0299e-04, 5.5504e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 262.33, cls_loss 0.0192 cls_loss_mapping 0.0284 cls_loss_causal 0.7486 re_mapping 0.0181 re_causal 0.0551 /// teacc 98.58 lr 0.00010000 +Epoch 42, weight, value: tensor([[ 0.0109, -0.0719, -0.0355, ..., -0.0752, -0.0460, -0.0745], + [ 0.0096, -0.0134, -0.0086, ..., 0.0461, 0.0507, -0.0064], + [-0.0471, 0.0260, -0.0658, ..., 0.0394, 0.0113, 0.0071], + ..., + [-0.0507, -0.0371, -0.0518, ..., -0.0048, -0.0403, 0.0529], + [ 0.0559, -0.0036, 0.0108, ..., -0.0127, -0.0528, 0.0001], + [-0.0812, 0.0045, 0.0162, ..., -0.0787, 0.0161, -0.0520]], + device='cuda:0'), grad: tensor([[ 6.3442e-06, 2.3663e-05, 3.5554e-05, ..., 3.7640e-05, + -2.8357e-05, -5.3309e-06], + [-7.6145e-06, 4.4727e-04, -4.4131e-04, ..., 4.6301e-04, + -9.4116e-05, 1.3806e-05], + [ 3.5584e-05, -2.3556e-03, 6.8188e-05, ..., -1.5030e-03, + -3.0651e-03, 3.5856e-06], + ..., + [ 7.6229e-07, 7.2956e-05, -2.8658e-04, ..., 1.8120e-05, + 3.1185e-04, -4.8923e-04], + [ 2.8059e-05, 7.7844e-05, 2.6798e-04, ..., 1.8501e-04, + 5.5313e-04, 9.8720e-06], + [ 2.2277e-05, 5.2378e-06, 3.1757e-04, ..., 1.6916e-04, + -3.2991e-05, 3.2783e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([-7.5196e-03, -2.9081e-02, 9.6837e-03, -1.9521e-02, 1.2586e-02, + 2.6258e-03, 1.7738e-02, -8.3705e-03, -2.3273e-02, 8.4240e-05], + device='cuda:0'), grad: tensor([-3.2210e-04, -1.6034e-04, -2.6398e-03, 3.9071e-05, 5.3835e-04, + 1.2708e-04, 1.5869e-03, -4.8971e-04, 7.4387e-04, 5.7745e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 262.33, cls_loss 0.0250 cls_loss_mapping 0.0346 cls_loss_causal 0.7460 re_mapping 0.0184 re_causal 0.0515 /// teacc 98.38 lr 0.00010000 +Epoch 43, weight, value: tensor([[ 0.0109, -0.0737, -0.0363, ..., -0.0760, -0.0459, -0.0754], + [ 0.0100, -0.0132, -0.0083, ..., 0.0462, 0.0515, -0.0072], + [-0.0483, 0.0280, -0.0668, ..., 0.0400, 0.0116, 0.0072], + ..., + [-0.0508, -0.0386, -0.0517, ..., -0.0052, -0.0410, 0.0528], + [ 0.0561, -0.0037, 0.0105, ..., -0.0131, -0.0541, -0.0007], + [-0.0826, 0.0034, 0.0157, ..., -0.0794, 0.0161, -0.0528]], + device='cuda:0'), grad: tensor([[ 2.1994e-05, 8.9407e-06, 7.6652e-05, ..., 3.0935e-05, + 5.9456e-06, 4.9733e-07], + [ 7.2792e-06, 1.6198e-05, 1.8328e-05, ..., 3.8892e-05, + -6.5453e-06, 9.6485e-06], + [ 1.4879e-05, -6.5148e-05, 7.0810e-05, ..., 1.5423e-05, + -5.7638e-05, 6.0871e-06], + ..., + [ 4.4435e-05, 9.6560e-06, 2.8591e-03, ..., 4.0512e-03, + 3.9160e-05, 8.4496e-04], + [ 7.1049e-05, -3.7178e-06, -3.3112e-03, ..., -4.5319e-03, + 5.9187e-05, -9.2936e-04], + [ 8.4758e-05, 6.7912e-06, -1.9878e-05, ..., 8.5533e-05, + 8.7976e-05, 8.4490e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0076, -0.0290, 0.0101, -0.0188, 0.0130, 0.0026, 0.0180, -0.0086, + -0.0239, -0.0007], device='cuda:0'), grad: tensor([ 3.5226e-05, 9.3937e-05, 5.2750e-05, 5.7220e-04, -4.0340e-04, + -5.3596e-04, 2.3985e-04, 3.8757e-03, -4.1275e-03, 1.9801e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 262.40, cls_loss 0.0207 cls_loss_mapping 0.0305 cls_loss_causal 0.7319 re_mapping 0.0183 re_causal 0.0514 /// teacc 98.59 lr 0.00010000 +Epoch 44, weight, value: tensor([[ 0.0108, -0.0745, -0.0370, ..., -0.0763, -0.0464, -0.0758], + [ 0.0103, -0.0132, -0.0077, ..., 0.0466, 0.0524, -0.0074], + [-0.0483, 0.0288, -0.0676, ..., 0.0401, 0.0116, 0.0068], + ..., + [-0.0519, -0.0377, -0.0517, ..., -0.0047, -0.0414, 0.0536], + [ 0.0563, -0.0046, 0.0111, ..., -0.0129, -0.0547, -0.0008], + [-0.0839, 0.0027, 0.0155, ..., -0.0806, 0.0163, -0.0534]], + device='cuda:0'), grad: tensor([[ 9.9614e-06, 9.0105e-08, 9.7603e-06, ..., 7.8902e-06, + -4.5657e-05, 7.3388e-06], + [ 1.0401e-05, 1.6298e-09, 1.0081e-05, ..., 2.9653e-05, + -3.2433e-07, 2.5809e-05], + [ 2.2233e-05, 1.6298e-09, 6.5207e-05, ..., 1.3077e-04, + 2.7582e-05, 6.2346e-05], + ..., + [-4.3005e-05, 2.3283e-10, 2.2292e-05, ..., -1.8024e-04, + 1.8775e-04, -1.5295e-04], + [ 1.7345e-05, 1.6997e-08, 6.1572e-05, ..., 5.6893e-05, + 5.1737e-05, 1.9416e-05], + [ 8.6129e-06, 6.9849e-10, -2.9564e-04, ..., -7.9691e-05, + -6.2943e-04, -4.0203e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0073, -0.0289, 0.0097, -0.0189, 0.0131, 0.0025, 0.0182, -0.0084, + -0.0239, -0.0008], device='cuda:0'), grad: tensor([-1.6129e-04, 6.2704e-05, 1.9932e-04, 1.6177e-04, 7.3338e-04, + 6.5625e-05, 8.3372e-06, 2.8586e-04, 2.1064e-04, -1.5669e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 262.19, cls_loss 0.0196 cls_loss_mapping 0.0292 cls_loss_causal 0.7307 re_mapping 0.0180 re_causal 0.0509 /// teacc 98.56 lr 0.00010000 +Epoch 45, weight, value: tensor([[ 0.0108, -0.0752, -0.0374, ..., -0.0768, -0.0469, -0.0763], + [ 0.0103, -0.0147, -0.0073, ..., 0.0466, 0.0527, -0.0069], + [-0.0484, 0.0310, -0.0679, ..., 0.0407, 0.0122, 0.0064], + ..., + [-0.0516, -0.0380, -0.0526, ..., -0.0048, -0.0429, 0.0540], + [ 0.0560, -0.0049, 0.0112, ..., -0.0131, -0.0555, -0.0011], + [-0.0842, 0.0015, 0.0157, ..., -0.0810, 0.0163, -0.0539]], + device='cuda:0'), grad: tensor([[ 6.9290e-06, 4.6268e-06, 1.6898e-05, ..., 2.0012e-05, + 4.5031e-05, 8.9407e-08], + [ 3.7383e-06, 3.5733e-05, -1.9416e-05, ..., 8.4937e-05, + 4.6283e-05, -1.1288e-06], + [ 6.9201e-05, -2.1470e-04, 4.5180e-05, ..., -5.0259e-04, + -4.6229e-04, -4.1444e-07], + ..., + [ 5.8748e-06, 5.4687e-05, 2.6017e-05, ..., 9.6381e-05, + 1.0943e-04, -1.0822e-06], + [ 7.1943e-05, 9.6798e-05, 2.0492e-04, ..., 5.0449e-04, + 3.5739e-04, 2.7521e-07], + [ 2.6703e-05, 2.0228e-06, 3.7432e-05, ..., 5.4359e-05, + 1.5065e-05, 1.2657e-06]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0076, -0.0289, 0.0101, -0.0186, 0.0136, 0.0022, 0.0181, -0.0088, + -0.0243, -0.0005], device='cuda:0'), grad: tensor([ 9.2447e-05, 6.9082e-05, -4.7660e-04, -2.7442e-04, 6.7592e-05, + 6.9094e-04, -1.1997e-03, 1.5330e-04, 8.0442e-04, 7.2241e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 261.28, cls_loss 0.0216 cls_loss_mapping 0.0293 cls_loss_causal 0.7246 re_mapping 0.0179 re_causal 0.0507 /// teacc 98.52 lr 0.00010000 +Epoch 46, weight, value: tensor([[ 0.0108, -0.0758, -0.0372, ..., -0.0769, -0.0469, -0.0785], + [ 0.0099, -0.0152, -0.0070, ..., 0.0466, 0.0531, -0.0069], + [-0.0487, 0.0319, -0.0686, ..., 0.0405, 0.0121, 0.0058], + ..., + [-0.0522, -0.0375, -0.0530, ..., -0.0046, -0.0431, 0.0547], + [ 0.0563, -0.0053, 0.0117, ..., -0.0128, -0.0561, -0.0011], + [-0.0845, 0.0012, 0.0156, ..., -0.0816, 0.0168, -0.0544]], + device='cuda:0'), grad: tensor([[ 1.1045e-06, 2.7884e-06, 4.7758e-06, ..., 1.0207e-05, + -1.6764e-05, -1.6287e-05], + [ 2.9430e-06, 7.4327e-05, 1.3685e-04, ..., 3.1447e-04, + 9.7454e-05, 1.8036e-04], + [ 5.9344e-06, -4.0197e-04, -1.1042e-05, ..., -7.1001e-04, + -6.0654e-04, 2.1458e-05], + ..., + [ 5.0515e-06, 3.1114e-05, -2.0552e-04, ..., -2.3675e-04, + 5.0485e-05, -3.3522e-04], + [ 6.0722e-06, 2.5892e-04, 1.6257e-05, ..., 4.7827e-04, + 4.0579e-04, 3.0696e-05], + [ 4.1053e-06, 3.5930e-06, 2.4706e-05, ..., 3.2246e-05, + 2.7016e-05, 3.1322e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0076, -0.0289, 0.0096, -0.0185, 0.0135, 0.0025, 0.0181, -0.0085, + -0.0245, -0.0004], device='cuda:0'), grad: tensor([-3.9172e-04, 5.6648e-04, -8.8072e-04, 7.6532e-05, 1.9640e-05, + 1.2946e-04, 1.2779e-04, -5.7411e-04, 7.1859e-04, 2.0754e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 256.89, cls_loss 0.0170 cls_loss_mapping 0.0233 cls_loss_causal 0.7125 re_mapping 0.0177 re_causal 0.0495 /// teacc 98.60 lr 0.00010000 +Epoch 47, weight, value: tensor([[ 0.0107, -0.0769, -0.0376, ..., -0.0774, -0.0477, -0.0796], + [ 0.0099, -0.0162, -0.0064, ..., 0.0468, 0.0531, -0.0077], + [-0.0489, 0.0333, -0.0691, ..., 0.0407, 0.0127, 0.0054], + ..., + [-0.0521, -0.0381, -0.0534, ..., -0.0046, -0.0436, 0.0557], + [ 0.0565, -0.0055, 0.0119, ..., -0.0127, -0.0566, -0.0014], + [-0.0846, 0.0007, 0.0152, ..., -0.0823, 0.0174, -0.0550]], + device='cuda:0'), grad: tensor([[ 1.6391e-04, 2.9936e-05, 3.3307e-04, ..., 2.2149e-04, + 2.3711e-04, 1.0759e-05], + [-1.6940e-04, 2.3112e-05, -9.0265e-04, ..., -6.1464e-04, + -1.3409e-03, 6.5267e-05], + [ 3.4750e-05, -6.5684e-05, 1.2600e-04, ..., 4.5395e-04, + 4.6134e-04, 1.4668e-03], + ..., + [ 2.1502e-05, 9.3430e-06, 1.9228e-04, ..., -5.4741e-04, + -3.4571e-04, -2.3041e-03], + [-5.2414e-03, -6.5506e-05, -8.6060e-03, ..., -5.3596e-03, + 1.0157e-04, 2.3529e-05], + [ 8.5711e-05, 3.0175e-05, 2.1398e-04, ..., 1.8251e-04, + 1.2124e-04, 8.1658e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0079, -0.0290, 0.0096, -0.0185, 0.0137, 0.0023, 0.0184, -0.0085, + -0.0245, -0.0004], device='cuda:0'), grad: tensor([ 0.0009, -0.0022, 0.0029, 0.0013, 0.0011, 0.0232, 0.0009, -0.0036, + -0.0253, 0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 261.47, cls_loss 0.0195 cls_loss_mapping 0.0279 cls_loss_causal 0.7324 re_mapping 0.0171 re_causal 0.0502 /// teacc 98.50 lr 0.00010000 +Epoch 48, weight, value: tensor([[ 0.0106, -0.0781, -0.0389, ..., -0.0783, -0.0485, -0.0804], + [ 0.0096, -0.0177, -0.0065, ..., 0.0465, 0.0531, -0.0084], + [-0.0487, 0.0352, -0.0700, ..., 0.0410, 0.0135, 0.0054], + ..., + [-0.0523, -0.0399, -0.0535, ..., -0.0049, -0.0440, 0.0564], + [ 0.0572, -0.0049, 0.0127, ..., -0.0121, -0.0573, -0.0018], + [-0.0842, -0.0002, 0.0152, ..., -0.0830, 0.0174, -0.0564]], + device='cuda:0'), grad: tensor([[ 2.1482e-04, 1.9968e-05, 1.9856e-06, ..., 1.9324e-04, + 3.4273e-05, 7.5717e-07], + [ 1.2741e-03, 4.5300e-04, 4.2319e-06, ..., 1.3456e-03, + 1.9944e-04, 6.3963e-06], + [-7.7858e-03, -2.3537e-03, -1.7449e-05, ..., -7.1335e-03, + -4.7493e-04, 9.2611e-06], + ..., + [ 4.5700e-03, 1.2999e-03, 6.2771e-06, ..., 3.9101e-03, + 2.3782e-04, -4.0382e-05], + [ 3.9744e-04, 1.3685e-04, -1.3098e-05, ..., 3.8767e-04, + 6.3837e-05, 2.7493e-06], + [ 2.2089e-04, 7.5579e-05, -2.3786e-06, ..., 2.1291e-04, + -1.2648e-04, 1.6987e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0084, -0.0294, 0.0094, -0.0182, 0.0141, 0.0021, 0.0187, -0.0084, + -0.0240, -0.0007], device='cuda:0'), grad: tensor([-5.1439e-05, 2.0809e-03, -1.1047e-02, 1.4257e-03, 2.7680e-04, + 2.4772e-04, 6.9737e-05, 6.0272e-03, 6.7139e-04, 3.0017e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 262.21, cls_loss 0.0222 cls_loss_mapping 0.0301 cls_loss_causal 0.7183 re_mapping 0.0173 re_causal 0.0468 /// teacc 98.50 lr 0.00010000 +Epoch 49, weight, value: tensor([[ 0.0108, -0.0788, -0.0389, ..., -0.0788, -0.0484, -0.0815], + [ 0.0099, -0.0185, -0.0064, ..., 0.0461, 0.0530, -0.0092], + [-0.0474, 0.0367, -0.0703, ..., 0.0417, 0.0145, 0.0052], + ..., + [-0.0536, -0.0403, -0.0534, ..., -0.0047, -0.0451, 0.0574], + [ 0.0572, -0.0054, 0.0127, ..., -0.0123, -0.0588, -0.0021], + [-0.0844, -0.0008, 0.0152, ..., -0.0833, 0.0178, -0.0580]], + device='cuda:0'), grad: tensor([[ 1.7881e-07, 5.4836e-06, 1.1927e-04, ..., 6.9141e-05, + 2.0111e-04, 1.3754e-05], + [ 3.9674e-07, 4.2051e-05, -5.8441e-03, ..., -2.9831e-03, + -6.8016e-03, -1.7214e-03], + [ 4.8522e-07, 1.6749e-05, 1.9014e-04, ..., 1.4555e-04, + 2.2388e-04, 2.0370e-05], + ..., + [ 3.7672e-07, -9.4920e-06, 2.3899e-03, ..., 1.6575e-03, + 2.8362e-03, 1.0176e-03], + [ 6.8499e-07, -2.5082e-04, -5.6314e-04, ..., -1.7529e-03, + 1.3199e-03, 2.1219e-05], + [ 1.8971e-06, 3.6117e-06, 1.5993e-03, ..., 1.0014e-03, + 1.1826e-03, 3.6311e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0078, -0.0300, 0.0103, -0.0184, 0.0138, 0.0027, 0.0182, -0.0083, + -0.0248, -0.0006], device='cuda:0'), grad: tensor([ 0.0004, -0.0130, 0.0005, 0.0007, 0.0015, 0.0019, 0.0002, 0.0047, + -0.0002, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 262.24, cls_loss 0.0158 cls_loss_mapping 0.0204 cls_loss_causal 0.7318 re_mapping 0.0166 re_causal 0.0477 /// teacc 98.56 lr 0.00010000 +Epoch 50, weight, value: tensor([[ 0.0108, -0.0800, -0.0396, ..., -0.0796, -0.0481, -0.0821], + [ 0.0097, -0.0182, -0.0061, ..., 0.0464, 0.0538, -0.0096], + [-0.0481, 0.0369, -0.0716, ..., 0.0414, 0.0145, 0.0042], + ..., + [-0.0536, -0.0400, -0.0537, ..., -0.0045, -0.0457, 0.0589], + [ 0.0573, -0.0049, 0.0132, ..., -0.0118, -0.0594, -0.0019], + [-0.0847, -0.0014, 0.0148, ..., -0.0843, 0.0179, -0.0595]], + device='cuda:0'), grad: tensor([[ 7.0333e-04, 1.6546e-04, 7.5865e-04, ..., 1.7595e-04, + 1.8632e-04, 2.0061e-06], + [ 1.1630e-05, 1.1683e-05, -4.7445e-05, ..., 2.2426e-05, + -7.1943e-05, 1.8567e-05], + [ 3.0056e-05, -4.4441e-04, 6.3717e-05, ..., -5.3215e-04, + -5.8126e-04, 1.4380e-05], + ..., + [ 5.6595e-05, 6.4194e-05, 1.0365e-04, ..., 5.2720e-05, + 2.8276e-04, 3.3051e-05], + [-7.5197e-04, 1.1855e-04, -1.2426e-03, ..., -9.6369e-04, + 4.7117e-05, 1.9390e-06], + [ 3.8886e-04, 2.1741e-05, 4.6515e-04, ..., 1.2577e-04, + -2.9042e-05, 8.2180e-06]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0078, -0.0298, 0.0097, -0.0179, 0.0138, 0.0028, 0.0182, -0.0078, + -0.0248, -0.0013], device='cuda:0'), grad: tensor([ 3.5591e-03, 2.1160e-05, -1.7643e-03, -1.2230e-02, -2.0117e-05, + 1.0986e-02, 4.9067e-04, 7.8869e-04, -3.5458e-03, 1.7128e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 262.53, cls_loss 0.0181 cls_loss_mapping 0.0254 cls_loss_causal 0.6882 re_mapping 0.0166 re_causal 0.0454 /// teacc 98.53 lr 0.00010000 +Epoch 51, weight, value: tensor([[ 0.0106, -0.0825, -0.0405, ..., -0.0803, -0.0487, -0.0829], + [ 0.0100, -0.0191, -0.0052, ..., 0.0462, 0.0544, -0.0105], + [-0.0483, 0.0382, -0.0722, ..., 0.0417, 0.0150, 0.0039], + ..., + [-0.0537, -0.0402, -0.0540, ..., -0.0041, -0.0462, 0.0599], + [ 0.0571, -0.0051, 0.0133, ..., -0.0118, -0.0605, -0.0022], + [-0.0851, -0.0013, 0.0143, ..., -0.0848, 0.0182, -0.0601]], + device='cuda:0'), grad: tensor([[ 1.2830e-05, 2.0470e-06, 1.5453e-05, ..., 2.3067e-05, + 2.0534e-05, 3.6918e-06], + [ 9.0301e-06, 3.1516e-06, 7.5400e-05, ..., 1.0133e-04, + 8.1778e-05, 1.0991e-04], + [-2.0102e-05, -1.3702e-05, 3.4750e-05, ..., -7.4208e-05, + -2.7448e-05, -1.1161e-05], + ..., + [ 1.1303e-05, 5.0738e-06, -1.4663e-04, ..., -2.4390e-04, + -7.4983e-05, -3.2210e-04], + [ 5.8979e-05, -4.3213e-06, 5.9485e-05, ..., 2.5913e-05, + 3.3975e-05, 7.5400e-06], + [ 1.6272e-05, 1.1902e-06, 6.0081e-05, ..., 1.7416e-04, + -2.4962e-04, 1.7798e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0084, -0.0294, 0.0097, -0.0183, 0.0133, 0.0030, 0.0191, -0.0072, + -0.0252, -0.0015], device='cuda:0'), grad: tensor([ 7.7069e-05, 3.1400e-04, -6.1691e-05, 1.2236e-03, 3.1042e-04, + -2.0657e-03, 4.5061e-04, -5.1069e-04, 2.3937e-04, 2.4036e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 262.67, cls_loss 0.0166 cls_loss_mapping 0.0233 cls_loss_causal 0.7195 re_mapping 0.0158 re_causal 0.0451 /// teacc 98.57 lr 0.00010000 +Epoch 52, weight, value: tensor([[ 0.0108, -0.0839, -0.0405, ..., -0.0808, -0.0490, -0.0832], + [ 0.0098, -0.0198, -0.0049, ..., 0.0461, 0.0544, -0.0102], + [-0.0488, 0.0379, -0.0730, ..., 0.0417, 0.0154, 0.0033], + ..., + [-0.0537, -0.0408, -0.0543, ..., -0.0041, -0.0465, 0.0605], + [ 0.0579, -0.0040, 0.0142, ..., -0.0111, -0.0610, -0.0018], + [-0.0853, -0.0020, 0.0139, ..., -0.0860, 0.0183, -0.0611]], + device='cuda:0'), grad: tensor([[ 1.5041e-07, 6.3777e-06, 3.4440e-06, ..., 1.4111e-05, + 2.1905e-06, -7.5474e-06], + [ 1.0692e-06, 7.7412e-06, -4.2319e-05, ..., -2.3678e-05, + -5.4002e-05, 2.7269e-06], + [ 7.2131e-07, -9.7930e-05, -1.0803e-06, ..., -2.3687e-04, + -1.3363e-04, -1.4096e-05], + ..., + [ 8.7311e-07, 9.2313e-06, 1.4976e-05, ..., 1.6913e-05, + 3.9876e-05, -9.3728e-06], + [ 8.2422e-07, 4.4733e-05, 4.1991e-05, ..., 1.5521e-04, + 1.0604e-04, 1.2852e-05], + [ 9.4902e-07, 1.7826e-06, -3.1471e-05, ..., 9.7975e-06, + -5.3078e-05, 3.1870e-06]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0081, -0.0294, 0.0091, -0.0184, 0.0134, 0.0034, 0.0190, -0.0074, + -0.0243, -0.0020], device='cuda:0'), grad: tensor([-1.6379e-04, -3.4153e-05, -2.3627e-04, 1.3351e-04, 3.0845e-05, + -6.5088e-05, 4.9800e-05, 1.4138e-04, 2.7728e-04, -1.3280e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 262.65, cls_loss 0.0142 cls_loss_mapping 0.0218 cls_loss_causal 0.6826 re_mapping 0.0164 re_causal 0.0465 /// teacc 98.58 lr 0.00010000 +Epoch 53, weight, value: tensor([[ 0.0105, -0.0841, -0.0408, ..., -0.0811, -0.0486, -0.0840], + [ 0.0098, -0.0201, -0.0048, ..., 0.0460, 0.0544, -0.0106], + [-0.0490, 0.0384, -0.0736, ..., 0.0422, 0.0163, 0.0023], + ..., + [-0.0537, -0.0404, -0.0545, ..., -0.0040, -0.0475, 0.0617], + [ 0.0580, -0.0043, 0.0142, ..., -0.0114, -0.0620, -0.0022], + [-0.0853, -0.0023, 0.0139, ..., -0.0860, 0.0191, -0.0614]], + device='cuda:0'), grad: tensor([[ 1.1481e-05, 3.5614e-06, 9.9242e-06, ..., 3.8773e-05, + 5.1826e-05, 1.3530e-05], + [ 6.7540e-06, 6.6776e-07, -3.4332e-04, ..., -1.1702e-03, + -1.6327e-03, 8.9705e-06], + [ 4.1544e-05, -1.9044e-05, 3.0112e-04, ..., 1.0099e-03, + 1.2445e-03, 2.8685e-05], + ..., + [-1.0073e-04, 3.4757e-06, 3.7909e-05, ..., -8.0943e-05, + 1.0842e-04, -1.4734e-04], + [ 1.5467e-05, 1.5730e-06, 9.9421e-05, ..., 1.4305e-04, + 5.8562e-05, 2.5496e-05], + [ 1.3836e-05, 4.9779e-07, 3.4958e-05, ..., 5.4032e-05, + 4.4912e-05, 2.0519e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0077, -0.0296, 0.0093, -0.0180, 0.0137, 0.0032, 0.0182, -0.0074, + -0.0250, -0.0016], device='cuda:0'), grad: tensor([ 8.5771e-05, -1.7910e-03, 1.5020e-03, -3.1322e-05, 5.2713e-06, + 3.0790e-06, -2.7269e-05, -1.1945e-04, 2.3866e-04, 1.3220e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 52---------------------------------------------------- +epoch 52, time 279.35, cls_loss 0.0159 cls_loss_mapping 0.0222 cls_loss_causal 0.6894 re_mapping 0.0159 re_causal 0.0455 /// teacc 98.65 lr 0.00010000 +Epoch 54, weight, value: tensor([[ 0.0105, -0.0848, -0.0412, ..., -0.0818, -0.0489, -0.0846], + [ 0.0100, -0.0208, -0.0044, ..., 0.0463, 0.0552, -0.0112], + [-0.0489, 0.0393, -0.0745, ..., 0.0421, 0.0165, 0.0017], + ..., + [-0.0538, -0.0399, -0.0551, ..., -0.0040, -0.0478, 0.0622], + [ 0.0580, -0.0047, 0.0152, ..., -0.0106, -0.0632, -0.0011], + [-0.0858, -0.0026, 0.0136, ..., -0.0867, 0.0194, -0.0609]], + device='cuda:0'), grad: tensor([[ 1.0924e-06, 3.7299e-07, 5.2489e-06, ..., 8.1733e-06, + 7.3090e-06, 2.8852e-06], + [ 4.4741e-06, 8.5905e-06, -2.8238e-05, ..., -1.0328e-06, + -2.7195e-05, 7.7635e-06], + [-3.1237e-06, -5.4032e-05, 1.1575e-04, ..., 9.4295e-05, + -4.2677e-05, 1.1839e-05], + ..., + [ 2.8074e-05, 4.2140e-05, 8.9586e-05, ..., 1.3387e-04, + 2.8276e-04, -3.9369e-05], + [ 1.6481e-05, 1.0701e-06, 2.3293e-04, ..., 3.0875e-04, + 9.5725e-05, 4.6007e-06], + [ 1.3568e-05, 1.2014e-07, 2.9042e-05, ..., -7.6175e-05, + -1.6108e-03, 5.4985e-06]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0076, -0.0293, 0.0089, -0.0184, 0.0138, 0.0029, 0.0182, -0.0070, + -0.0247, -0.0017], device='cuda:0'), grad: tensor([-1.1966e-05, 2.8927e-06, 1.1712e-04, -5.0497e-04, 2.9488e-03, + 1.0681e-04, 2.0146e-05, 6.1560e-04, 4.0579e-04, -3.7003e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 262.87, cls_loss 0.0154 cls_loss_mapping 0.0214 cls_loss_causal 0.7139 re_mapping 0.0151 re_causal 0.0454 /// teacc 98.59 lr 0.00010000 +Epoch 55, weight, value: tensor([[ 0.0104, -0.0851, -0.0417, ..., -0.0825, -0.0503, -0.0860], + [ 0.0096, -0.0215, -0.0039, ..., 0.0466, 0.0558, -0.0111], + [-0.0487, 0.0404, -0.0749, ..., 0.0422, 0.0172, 0.0017], + ..., + [-0.0543, -0.0405, -0.0556, ..., -0.0041, -0.0492, 0.0625], + [ 0.0579, -0.0048, 0.0153, ..., -0.0107, -0.0647, -0.0004], + [-0.0863, -0.0029, 0.0133, ..., -0.0869, 0.0202, -0.0619]], + device='cuda:0'), grad: tensor([[ 1.6671e-07, 7.2867e-06, 9.5442e-06, ..., 3.4481e-05, + 2.8774e-05, 1.3098e-05], + [ 1.3988e-06, 1.0304e-05, 2.6211e-05, ..., 4.7654e-05, + 3.0637e-05, 2.9936e-05], + [ 3.3788e-06, -1.0357e-03, -1.0446e-05, ..., -4.3755e-03, + -5.8441e-03, 8.6486e-05], + ..., + [ 2.4289e-06, -3.5971e-05, -2.5678e-04, ..., -2.3413e-04, + -4.5657e-05, -4.8637e-04], + [ 1.0431e-05, 1.5378e-05, 1.9419e-04, ..., 2.8157e-04, + 1.3161e-04, 9.3162e-05], + [ 7.5670e-07, 9.8896e-04, 1.1551e-04, ..., 4.3182e-03, + 5.4970e-03, 1.1885e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0082, -0.0288, 0.0091, -0.0185, 0.0141, 0.0032, 0.0186, -0.0080, + -0.0244, -0.0016], device='cuda:0'), grad: tensor([ 5.7578e-05, 1.2743e-04, -5.2032e-03, 4.2534e-04, 1.0949e-04, + -7.3075e-05, -1.6212e-05, -1.1253e-03, 5.1355e-04, 5.1804e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 262.44, cls_loss 0.0153 cls_loss_mapping 0.0200 cls_loss_causal 0.6926 re_mapping 0.0151 re_causal 0.0447 /// teacc 98.57 lr 0.00010000 +Epoch 56, weight, value: tensor([[ 0.0102, -0.0857, -0.0417, ..., -0.0832, -0.0493, -0.0870], + [ 0.0097, -0.0219, -0.0038, ..., 0.0466, 0.0559, -0.0125], + [-0.0492, 0.0415, -0.0756, ..., 0.0425, 0.0175, 0.0015], + ..., + [-0.0542, -0.0413, -0.0556, ..., -0.0042, -0.0497, 0.0636], + [ 0.0581, -0.0055, 0.0153, ..., -0.0110, -0.0656, -0.0010], + [-0.0865, -0.0035, 0.0131, ..., -0.0879, 0.0199, -0.0632]], + device='cuda:0'), grad: tensor([[ 6.1840e-06, 1.6689e-06, 1.5163e-04, ..., 4.4465e-05, + 8.1837e-05, -1.2424e-06], + [-3.9488e-05, 2.2858e-05, -3.4833e-04, ..., -5.8085e-05, + -2.4748e-04, 1.3679e-05], + [ 4.3698e-06, -6.5506e-05, 3.6180e-05, ..., -6.7234e-05, + -7.0453e-05, 5.6438e-06], + ..., + [ 6.9290e-06, 3.3945e-05, 1.9044e-05, ..., 1.3418e-05, + 6.7174e-05, -1.3602e-04], + [ 7.8678e-06, 2.4866e-07, -3.7146e-04, ..., -2.5988e-04, + 6.9380e-05, 9.8646e-06], + [ 4.6492e-06, 6.0536e-07, 9.9778e-05, ..., 1.5068e-04, + -3.5548e-04, 5.2303e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0071, -0.0293, 0.0092, -0.0182, 0.0138, 0.0037, 0.0181, -0.0078, + -0.0249, -0.0020], device='cuda:0'), grad: tensor([ 1.7715e-04, -6.1464e-04, -3.8147e-05, 2.4354e-04, 1.2255e-03, + 1.8167e-04, 1.5891e-04, -1.3149e-04, -5.2214e-04, -6.7902e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 262.18, cls_loss 0.0170 cls_loss_mapping 0.0240 cls_loss_causal 0.7047 re_mapping 0.0150 re_causal 0.0422 /// teacc 98.51 lr 0.00010000 +Epoch 57, weight, value: tensor([[ 0.0100, -0.0868, -0.0422, ..., -0.0839, -0.0499, -0.0877], + [ 0.0100, -0.0229, -0.0038, ..., 0.0457, 0.0558, -0.0126], + [-0.0501, 0.0428, -0.0758, ..., 0.0430, 0.0183, 0.0010], + ..., + [-0.0544, -0.0417, -0.0554, ..., -0.0036, -0.0498, 0.0644], + [ 0.0590, -0.0051, 0.0157, ..., -0.0103, -0.0665, -0.0006], + [-0.0867, -0.0037, 0.0127, ..., -0.0888, 0.0204, -0.0641]], + device='cuda:0'), grad: tensor([[ 2.5220e-06, 7.8261e-05, 2.6282e-06, ..., 1.4007e-04, + 8.9109e-05, 5.2750e-06], + [-4.5776e-05, 9.1717e-06, -1.0747e-04, ..., -5.7667e-05, + -8.4400e-05, 4.7296e-05], + [-5.3123e-06, -3.3569e-04, 1.8671e-05, ..., -5.5933e-04, + -3.3069e-04, 2.4676e-05], + ..., + [ 6.6534e-06, 1.1259e-04, -1.6287e-05, ..., 6.0350e-05, + 7.7152e-04, -2.6083e-04], + [ 3.0994e-05, 6.5088e-05, 5.4389e-05, ..., 2.0456e-04, + 2.0874e-04, 1.9237e-05], + [ 3.7253e-06, 1.1966e-05, 7.7710e-06, ..., 3.7849e-05, + -7.6830e-05, 2.2545e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0075, -0.0296, 0.0096, -0.0187, 0.0138, 0.0039, 0.0183, -0.0075, + -0.0249, -0.0019], device='cuda:0'), grad: tensor([ 3.0756e-04, 8.1122e-05, -1.0672e-03, 2.1482e-04, -3.1872e-03, + 1.1480e-04, 1.3673e-04, 2.9964e-03, 4.8685e-04, -8.3864e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 262.51, cls_loss 0.0146 cls_loss_mapping 0.0217 cls_loss_causal 0.6835 re_mapping 0.0154 re_causal 0.0444 /// teacc 98.61 lr 0.00010000 +Epoch 58, weight, value: tensor([[ 0.0111, -0.0873, -0.0425, ..., -0.0846, -0.0500, -0.0886], + [ 0.0101, -0.0245, -0.0036, ..., 0.0456, 0.0559, -0.0137], + [-0.0494, 0.0446, -0.0758, ..., 0.0438, 0.0193, 0.0008], + ..., + [-0.0546, -0.0420, -0.0555, ..., -0.0035, -0.0504, 0.0652], + [ 0.0591, -0.0058, 0.0158, ..., -0.0106, -0.0678, -0.0011], + [-0.0869, -0.0039, 0.0123, ..., -0.0895, 0.0203, -0.0652]], + device='cuda:0'), grad: tensor([[-1.1384e-05, 1.1958e-05, 4.9174e-05, ..., 6.0469e-05, + -7.8380e-05, 4.7684e-07], + [-4.6790e-05, 2.6226e-05, -1.3676e-03, ..., -2.0580e-03, + -1.7090e-03, -8.0538e-04], + [ 9.7692e-05, 2.5082e-04, 3.0136e-04, ..., 5.3501e-04, + 5.3704e-05, 3.3248e-06], + ..., + [ 5.9992e-05, 6.1750e-04, 1.2770e-03, ..., 2.2488e-03, + 1.3323e-03, 6.9904e-04], + [-2.0385e-05, -1.0452e-03, -8.6486e-05, ..., -6.3944e-04, + 4.0799e-05, 6.0573e-06], + [ 1.9848e-05, 1.7866e-05, 1.9521e-05, ..., 1.2767e-04, + -1.2243e-04, 3.9011e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0070, -0.0297, 0.0104, -0.0187, 0.0144, 0.0038, 0.0183, -0.0075, + -0.0255, -0.0028], device='cuda:0'), grad: tensor([-4.6420e-04, -5.2643e-03, 1.1234e-03, -8.3494e-04, 7.1955e-04, + 9.7692e-05, 6.1560e-04, 5.7220e-03, -1.6613e-03, -4.7356e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 262.73, cls_loss 0.0167 cls_loss_mapping 0.0249 cls_loss_causal 0.7140 re_mapping 0.0150 re_causal 0.0427 /// teacc 98.61 lr 0.00010000 +Epoch 59, weight, value: tensor([[ 0.0106, -0.0879, -0.0439, ..., -0.0858, -0.0507, -0.0899], + [ 0.0099, -0.0234, -0.0031, ..., 0.0462, 0.0571, -0.0150], + [-0.0489, 0.0452, -0.0771, ..., 0.0436, 0.0190, 0.0007], + ..., + [-0.0552, -0.0433, -0.0552, ..., -0.0035, -0.0513, 0.0669], + [ 0.0593, -0.0059, 0.0162, ..., -0.0105, -0.0692, -0.0017], + [-0.0874, -0.0041, 0.0117, ..., -0.0900, 0.0206, -0.0662]], + device='cuda:0'), grad: tensor([[ 8.2050e-07, 8.7637e-07, 2.1681e-06, ..., 3.4980e-06, + -2.0824e-06, 1.8813e-06], + [ 5.0198e-07, 3.5334e-06, -2.8059e-05, ..., -4.8056e-06, + -3.0756e-05, 8.7097e-06], + [ 6.1747e-07, -6.1452e-05, -2.8551e-05, ..., -1.6797e-04, + -9.7394e-05, -3.3200e-05], + ..., + [ 6.3423e-07, 7.8157e-06, 2.1994e-05, ..., 1.4909e-05, + 3.8266e-04, 4.8727e-05], + [-3.0082e-06, 4.5478e-05, 7.8753e-06, ..., 1.3161e-04, + 1.1057e-04, 4.0114e-05], + [ 1.5637e-06, 1.6298e-07, 8.0317e-06, ..., 1.0744e-05, + -3.9887e-04, -7.9513e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0076, -0.0294, 0.0103, -0.0185, 0.0147, 0.0040, 0.0177, -0.0073, + -0.0255, -0.0030], device='cuda:0'), grad: tensor([-1.4819e-05, -2.1711e-05, -1.9157e-04, 7.9572e-05, 6.8128e-05, + -5.9187e-05, 3.3565e-06, 6.9714e-04, 1.9217e-04, -7.5436e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 262.25, cls_loss 0.0136 cls_loss_mapping 0.0200 cls_loss_causal 0.6408 re_mapping 0.0147 re_causal 0.0406 /// teacc 98.57 lr 0.00010000 +Epoch 60, weight, value: tensor([[ 0.0103, -0.0880, -0.0446, ..., -0.0864, -0.0518, -0.0911], + [ 0.0122, -0.0243, -0.0016, ..., 0.0468, 0.0579, -0.0139], + [-0.0493, 0.0463, -0.0781, ..., 0.0438, 0.0197, 0.0003], + ..., + [-0.0566, -0.0439, -0.0560, ..., -0.0039, -0.0533, 0.0671], + [ 0.0592, -0.0063, 0.0166, ..., -0.0107, -0.0701, -0.0025], + [-0.0878, -0.0043, 0.0110, ..., -0.0909, 0.0220, -0.0660]], + device='cuda:0'), grad: tensor([[ 4.7721e-06, 1.2629e-05, 3.2224e-06, ..., 2.6096e-06, + 2.1085e-05, 1.3243e-06], + [ 1.3607e-06, 9.5814e-06, 2.7250e-06, ..., 1.6183e-05, + 4.5687e-05, 6.6161e-06], + [ 1.7500e-06, -2.1130e-05, 9.3281e-06, ..., -2.0966e-05, + -2.2963e-05, 5.0738e-06], + ..., + [-7.4580e-06, 3.5539e-06, -1.9878e-05, ..., -4.2200e-05, + 7.0095e-05, -6.4492e-05], + [ 2.8592e-06, 4.7386e-06, -1.1683e-05, ..., -5.9605e-08, + 2.3413e-04, 2.0377e-06], + [ 4.8093e-06, 5.9698e-07, -1.0394e-05, ..., 1.6004e-05, + -2.1057e-03, 1.3955e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0084, -0.0284, 0.0102, -0.0178, 0.0143, 0.0038, 0.0180, -0.0081, + -0.0259, -0.0024], device='cuda:0'), grad: tensor([ 5.3972e-05, 1.2982e-04, 1.3456e-05, 1.1611e-04, 5.4741e-03, + 1.6129e-04, 4.6879e-05, 1.1122e-04, 7.5006e-04, -6.8550e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 262.20, cls_loss 0.0112 cls_loss_mapping 0.0182 cls_loss_causal 0.6588 re_mapping 0.0145 re_causal 0.0426 /// teacc 98.65 lr 0.00010000 +Epoch 61, weight, value: tensor([[ 0.0102, -0.0887, -0.0450, ..., -0.0868, -0.0519, -0.0918], + [ 0.0119, -0.0248, -0.0020, ..., 0.0464, 0.0577, -0.0146], + [-0.0490, 0.0474, -0.0776, ..., 0.0444, 0.0205, 0.0001], + ..., + [-0.0565, -0.0447, -0.0563, ..., -0.0041, -0.0541, 0.0678], + [ 0.0596, -0.0064, 0.0172, ..., -0.0105, -0.0703, -0.0029], + [-0.0888, -0.0046, 0.0106, ..., -0.0919, 0.0220, -0.0667]], + device='cuda:0'), grad: tensor([[-5.8152e-06, 4.3474e-06, 1.1986e-06, ..., 6.5789e-06, + 1.7500e-04, 3.7998e-06], + [ 1.2539e-05, 3.1710e-04, 1.7239e-06, ..., 2.3496e-04, + 3.7384e-04, 3.9846e-05], + [-9.7156e-06, -4.8304e-04, -1.0923e-05, ..., -3.8886e-04, + -5.9271e-04, -4.8459e-05], + ..., + [ 2.6748e-06, 1.3955e-05, -5.9605e-06, ..., -1.5453e-05, + 5.8502e-05, -3.5554e-05], + [-1.0826e-05, 9.4771e-06, -3.2425e-05, ..., -1.1407e-05, + 3.7313e-05, 3.1367e-06], + [ 9.3225e-07, 1.7257e-06, 8.6427e-06, ..., -2.9489e-05, + -2.5392e-04, -9.9480e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0083, -0.0290, 0.0108, -0.0176, 0.0144, 0.0040, 0.0176, -0.0081, + -0.0258, -0.0029], device='cuda:0'), grad: tensor([ 4.4417e-04, 3.9005e-04, -5.1022e-04, 1.3304e-04, 1.0071e-03, + 7.4506e-05, -9.6178e-04, 6.7890e-05, 1.3880e-05, -6.5899e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 261.89, cls_loss 0.0127 cls_loss_mapping 0.0201 cls_loss_causal 0.6618 re_mapping 0.0142 re_causal 0.0416 /// teacc 98.49 lr 0.00010000 +Epoch 62, weight, value: tensor([[ 1.0399e-02, -8.9173e-02, -4.5177e-02, ..., -8.7403e-02, + -5.2336e-02, -9.2642e-02], + [ 1.2629e-02, -2.4592e-02, -1.0421e-03, ..., 4.6919e-02, + 5.8763e-02, -1.3845e-02], + [-4.9594e-02, 4.7889e-02, -7.8553e-02, ..., 4.4304e-02, + 2.0586e-02, -1.3442e-05], + ..., + [-5.6891e-02, -4.5617e-02, -5.6746e-02, ..., -4.1491e-03, + -5.5773e-02, 6.8108e-02], + [ 5.9460e-02, -6.3847e-03, 1.7492e-02, ..., -1.0285e-02, + -7.0937e-02, -3.3056e-03], + [-8.9216e-02, -4.6511e-03, 1.0163e-02, ..., -9.2741e-02, + 2.3265e-02, -6.6444e-02]], device='cuda:0'), grad: tensor([[ 4.8913e-06, 7.8604e-06, 3.7700e-05, ..., 2.1636e-05, + 6.6161e-06, 1.9688e-06], + [ 3.6824e-06, 5.3495e-06, -7.7337e-06, ..., 3.7886e-06, + -7.5400e-06, 5.3085e-06], + [-1.8835e-05, -9.7632e-05, 2.8312e-05, ..., -1.6403e-04, + -1.4114e-04, -2.1487e-05], + ..., + [ 1.9133e-05, 8.0884e-05, 1.7539e-05, ..., 1.3697e-04, + 1.5664e-04, 1.0014e-05], + [ 6.7838e-06, -4.7326e-05, -5.8055e-05, ..., -1.4715e-05, + -1.3638e-04, 1.0291e-06], + [ 6.0946e-06, 3.5703e-05, 8.5056e-05, ..., 2.8193e-05, + 1.0610e-04, 1.6494e-06]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0083, -0.0283, 0.0106, -0.0170, 0.0133, 0.0034, 0.0180, -0.0085, + -0.0258, -0.0021], device='cuda:0'), grad: tensor([ 7.3314e-05, 1.4186e-05, -1.2386e-04, 1.5450e-04, -6.1929e-05, + -4.4632e-04, 8.1718e-05, 2.4056e-04, -3.0255e-04, 3.6931e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 61---------------------------------------------------- +epoch 61, time 279.38, cls_loss 0.0122 cls_loss_mapping 0.0188 cls_loss_causal 0.6811 re_mapping 0.0144 re_causal 0.0435 /// teacc 98.67 lr 0.00010000 +Epoch 63, weight, value: tensor([[ 0.0104, -0.0912, -0.0465, ..., -0.0878, -0.0523, -0.0936], + [ 0.0123, -0.0251, -0.0010, ..., 0.0467, 0.0586, -0.0144], + [-0.0494, 0.0486, -0.0790, ..., 0.0446, 0.0212, -0.0003], + ..., + [-0.0570, -0.0460, -0.0571, ..., -0.0044, -0.0565, 0.0686], + [ 0.0597, -0.0058, 0.0178, ..., -0.0101, -0.0715, -0.0036], + [-0.0895, -0.0050, 0.0100, ..., -0.0932, 0.0238, -0.0668]], + device='cuda:0'), grad: tensor([[ 2.2259e-07, 5.1688e-07, 1.5618e-06, ..., 3.2447e-06, + -2.4706e-05, 4.5486e-06], + [ 3.7998e-07, -2.2575e-06, -3.6340e-06, ..., 1.0412e-06, + -2.0396e-06, 3.4958e-05], + [ 7.7933e-06, 1.2787e-06, 3.6031e-05, ..., 5.9992e-05, + 1.5110e-05, 1.4395e-05], + ..., + [ 1.2526e-06, 1.1548e-06, 5.9903e-06, ..., -2.0728e-05, + -5.9813e-05, -1.3483e-04], + [ 4.2990e-06, -4.4703e-06, -1.0423e-05, ..., -3.7216e-06, + 2.1547e-05, 1.9968e-06], + [ 3.5204e-07, 3.6787e-07, 2.9597e-06, ..., 7.6890e-06, + 7.8201e-05, 4.8876e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0084, -0.0287, 0.0106, -0.0169, 0.0128, 0.0041, 0.0179, -0.0087, + -0.0259, -0.0019], device='cuda:0'), grad: tensor([-9.0003e-05, 5.3763e-05, 9.7692e-05, -3.5018e-05, -2.1785e-05, + 6.1572e-05, -8.1137e-06, -2.3770e-04, 3.1829e-05, 1.4770e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 262.46, cls_loss 0.0157 cls_loss_mapping 0.0189 cls_loss_causal 0.6852 re_mapping 0.0139 re_causal 0.0398 /// teacc 98.65 lr 0.00010000 +Epoch 64, weight, value: tensor([[ 0.0104, -0.0930, -0.0472, ..., -0.0885, -0.0524, -0.0944], + [ 0.0121, -0.0256, -0.0010, ..., 0.0464, 0.0584, -0.0152], + [-0.0487, 0.0494, -0.0796, ..., 0.0446, 0.0210, -0.0002], + ..., + [-0.0578, -0.0455, -0.0571, ..., -0.0037, -0.0557, 0.0695], + [ 0.0599, -0.0058, 0.0183, ..., -0.0099, -0.0725, -0.0038], + [-0.0898, -0.0034, 0.0099, ..., -0.0936, 0.0241, -0.0675]], + device='cuda:0'), grad: tensor([[ 8.9407e-07, 9.1076e-05, 2.7940e-06, ..., 5.3346e-06, + 4.7177e-05, 2.3488e-06], + [ 1.8151e-06, 1.2964e-06, 6.6042e-05, ..., 9.2804e-05, + -3.7402e-06, 5.1260e-05], + [ 6.7018e-06, -8.9854e-06, 2.6718e-05, ..., 1.9014e-05, + -1.4290e-05, 2.2337e-05], + ..., + [-1.5944e-05, -7.2457e-06, 1.2489e-02, ..., 1.6602e-02, + 1.8716e-05, 8.6746e-03], + [ 2.0135e-06, 8.7246e-06, -1.2840e-02, ..., -1.7090e-02, + 2.6211e-05, -8.9417e-03], + [ 2.9821e-06, 3.1684e-06, 1.0235e-06, ..., 7.3969e-05, + -5.5820e-05, 4.0263e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0081, -0.0295, 0.0104, -0.0175, 0.0132, 0.0041, 0.0184, -0.0079, + -0.0260, -0.0021], device='cuda:0'), grad: tensor([ 1.8907e-04, 1.0526e-04, 3.7253e-05, 3.3069e-04, 1.9646e-04, + -8.5413e-05, -2.3770e-04, 1.7593e-02, -1.8036e-02, -9.5069e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 262.45, cls_loss 0.0127 cls_loss_mapping 0.0163 cls_loss_causal 0.6758 re_mapping 0.0138 re_causal 0.0396 /// teacc 98.58 lr 0.00010000 +Epoch 65, weight, value: tensor([[ 0.0105, -0.0939, -0.0479, ..., -0.0891, -0.0526, -0.0954], + [ 0.0130, -0.0258, -0.0006, ..., 0.0464, 0.0586, -0.0160], + [-0.0489, 0.0504, -0.0803, ..., 0.0447, 0.0217, -0.0013], + ..., + [-0.0578, -0.0454, -0.0578, ..., -0.0041, -0.0565, 0.0704], + [ 0.0597, -0.0066, 0.0193, ..., -0.0092, -0.0739, -0.0027], + [-0.0900, -0.0036, 0.0098, ..., -0.0941, 0.0249, -0.0683]], + device='cuda:0'), grad: tensor([[ 1.6112e-07, 9.2909e-06, 4.9286e-06, ..., 1.2256e-05, + 1.3545e-05, 5.8208e-07], + [ 3.2689e-07, -1.2350e-04, -3.3998e-04, ..., -3.5143e-04, + -4.7803e-04, 4.4741e-06], + [ 3.3341e-07, -2.7537e-04, -4.4376e-05, ..., -3.2043e-04, + 3.1304e-04, 1.9759e-05], + ..., + [ 4.4797e-07, 2.7299e-05, 1.1861e-05, ..., 1.3351e-05, + 3.2336e-05, -2.4572e-05], + [ 6.9663e-07, 3.2187e-04, 3.4094e-04, ..., 5.4455e-04, + 6.3753e-04, 3.1646e-06], + [ 1.4612e-06, 4.5039e-06, 5.3905e-06, ..., 1.8671e-05, + 5.8770e-05, 1.2688e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0082, -0.0295, 0.0101, -0.0179, 0.0135, 0.0043, 0.0185, -0.0079, + -0.0257, -0.0020], device='cuda:0'), grad: tensor([-8.1301e-05, -6.7091e-04, 1.4007e-04, 1.3554e-04, -7.1383e-04, + -6.2585e-05, 1.4508e-04, -1.7375e-05, 9.9277e-04, 1.3435e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 64---------------------------------------------------- +epoch 64, time 278.49, cls_loss 0.0112 cls_loss_mapping 0.0147 cls_loss_causal 0.6958 re_mapping 0.0139 re_causal 0.0408 /// teacc 98.69 lr 0.00010000 +Epoch 66, weight, value: tensor([[ 1.0757e-02, -9.5198e-02, -4.8139e-02, ..., -8.9725e-02, + -5.3294e-02, -9.5830e-02], + [ 1.2627e-02, -2.6135e-02, -6.7352e-05, ..., 4.6324e-02, + 5.8831e-02, -1.6305e-02], + [-4.8842e-02, 5.1630e-02, -8.0917e-02, ..., 4.4990e-02, + 2.2209e-02, -1.8782e-03], + ..., + [-5.7479e-02, -4.5611e-02, -5.7908e-02, ..., -3.8529e-03, + -5.7171e-02, 7.1724e-02], + [ 5.9682e-02, -6.6708e-03, 1.9776e-02, ..., -9.1764e-03, + -7.4682e-02, -3.2358e-03], + [-8.9935e-02, -4.0307e-03, 9.5205e-03, ..., -9.4765e-02, + 2.5303e-02, -6.9225e-02]], device='cuda:0'), grad: tensor([[ 8.1025e-08, 5.6714e-05, 1.8971e-06, ..., 4.8727e-06, + 1.1778e-04, 8.4378e-07], + [ 7.0035e-07, 4.2766e-06, -1.5453e-05, ..., 6.0908e-07, + -1.7300e-05, 1.4573e-05], + [ 2.4773e-07, 8.0943e-05, 3.5763e-05, ..., 5.8830e-05, + 1.6057e-04, 2.2739e-05], + ..., + [ 3.1386e-07, -2.7549e-06, 9.4902e-07, ..., -8.7142e-05, + 8.2180e-06, -9.2328e-05], + [ 2.5146e-07, 4.2260e-05, -3.6359e-05, ..., -5.5194e-05, + 2.0528e-04, 5.7220e-06], + [ 2.8685e-07, -9.3520e-05, 8.5831e-06, ..., 3.6091e-05, + -1.4887e-03, 1.9222e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0088, -0.0294, 0.0103, -0.0181, 0.0140, 0.0039, 0.0186, -0.0078, + -0.0258, -0.0019], device='cuda:0'), grad: tensor([ 2.2221e-04, 2.4438e-05, 4.0460e-04, 7.1108e-05, 1.6813e-03, + 1.3056e-03, -2.6054e-03, -1.0604e-04, 3.6645e-04, -1.3628e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 262.47, cls_loss 0.0121 cls_loss_mapping 0.0172 cls_loss_causal 0.6641 re_mapping 0.0143 re_causal 0.0395 /// teacc 98.53 lr 0.00010000 +Epoch 67, weight, value: tensor([[ 0.0107, -0.0969, -0.0484, ..., -0.0902, -0.0537, -0.0965], + [ 0.0126, -0.0259, 0.0002, ..., 0.0463, 0.0591, -0.0169], + [-0.0484, 0.0546, -0.0798, ..., 0.0461, 0.0230, -0.0023], + ..., + [-0.0579, -0.0458, -0.0579, ..., -0.0038, -0.0574, 0.0723], + [ 0.0595, -0.0099, 0.0192, ..., -0.0098, -0.0763, -0.0034], + [-0.0901, -0.0053, 0.0091, ..., -0.0962, 0.0252, -0.0689]], + device='cuda:0'), grad: tensor([[ 1.2732e-04, 9.8169e-05, 6.8545e-05, ..., 9.2328e-05, + 5.7459e-05, 3.2306e-05], + [ 4.0680e-05, 6.1417e-04, 2.5196e-03, ..., 1.2722e-03, + 2.9354e-03, 9.9659e-05], + [-9.3699e-05, 8.1003e-05, 1.0276e-04, ..., -8.2207e-04, + -1.9205e-04, -4.5371e-04], + ..., + [ 2.8655e-05, 4.9137e-06, 4.9770e-05, ..., 1.8942e-04, + 1.2994e-04, 8.4758e-05], + [-2.2006e-04, -5.0515e-05, 5.2643e-04, ..., 4.3535e-04, + 7.2098e-04, 1.1325e-04], + [ 1.2010e-05, 7.8157e-06, -8.6948e-06, ..., 1.7792e-05, + -1.0711e-04, 6.0461e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0087, -0.0294, 0.0117, -0.0187, 0.0148, 0.0043, 0.0180, -0.0080, + -0.0270, -0.0021], device='cuda:0'), grad: tensor([ 0.0007, 0.0055, -0.0009, 0.0005, 0.0004, 0.0008, -0.0076, 0.0004, + 0.0006, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 262.55, cls_loss 0.0117 cls_loss_mapping 0.0152 cls_loss_causal 0.6478 re_mapping 0.0134 re_causal 0.0368 /// teacc 98.63 lr 0.00010000 +Epoch 68, weight, value: tensor([[ 0.0106, -0.0979, -0.0488, ..., -0.0909, -0.0540, -0.0967], + [ 0.0129, -0.0257, 0.0005, ..., 0.0466, 0.0600, -0.0167], + [-0.0490, 0.0546, -0.0807, ..., 0.0457, 0.0228, -0.0029], + ..., + [-0.0577, -0.0460, -0.0582, ..., -0.0040, -0.0591, 0.0730], + [ 0.0603, -0.0097, 0.0196, ..., -0.0092, -0.0771, -0.0037], + [-0.0907, -0.0056, 0.0086, ..., -0.0964, 0.0261, -0.0694]], + device='cuda:0'), grad: tensor([[ 1.0850e-06, 5.0105e-06, 1.0580e-05, ..., 1.2979e-05, + 1.9535e-05, 1.5181e-06], + [ 8.0764e-06, 2.8476e-05, 1.9088e-05, ..., 6.2406e-05, + 1.8731e-05, 1.8448e-05], + [ 5.7742e-06, 4.2655e-06, 4.1366e-05, ..., 4.2081e-05, + 2.6643e-05, 1.3135e-05], + ..., + [-6.1691e-05, -2.0134e-04, -2.0313e-04, ..., -4.1270e-04, + -2.0719e-04, -1.2201e-04], + [ 1.9912e-06, 3.6918e-06, -3.2723e-05, ..., -2.1935e-05, + 6.6102e-05, 3.8855e-06], + [ 3.9279e-05, 1.2577e-04, 1.8251e-04, ..., 2.8229e-04, + -1.3943e-03, 6.3598e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0085, -0.0291, 0.0110, -0.0184, 0.0145, 0.0044, 0.0179, -0.0083, + -0.0269, -0.0018], device='cuda:0'), grad: tensor([ 8.3327e-05, 2.9778e-04, 1.9491e-04, 2.2471e-04, 4.4785e-03, + -1.6558e-04, 5.5134e-05, -1.6289e-03, 2.6941e-04, -3.8052e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 67---------------------------------------------------- +epoch 67, time 279.17, cls_loss 0.0115 cls_loss_mapping 0.0146 cls_loss_causal 0.6184 re_mapping 0.0131 re_causal 0.0375 /// teacc 98.72 lr 0.00010000 +Epoch 69, weight, value: tensor([[ 0.0105, -0.0987, -0.0499, ..., -0.0917, -0.0549, -0.0978], + [ 0.0132, -0.0263, 0.0008, ..., 0.0462, 0.0605, -0.0172], + [-0.0488, 0.0548, -0.0813, ..., 0.0457, 0.0232, -0.0031], + ..., + [-0.0582, -0.0465, -0.0581, ..., -0.0033, -0.0595, 0.0744], + [ 0.0605, -0.0086, 0.0200, ..., -0.0088, -0.0776, -0.0037], + [-0.0913, -0.0062, 0.0078, ..., -0.0982, 0.0260, -0.0713]], + device='cuda:0'), grad: tensor([[ 9.2834e-06, 1.3255e-05, 4.3493e-07, ..., 3.8855e-06, + 4.1008e-05, 2.2165e-07], + [ 4.9710e-05, 7.0572e-05, 2.3115e-06, ..., 2.1741e-05, + 2.1899e-04, 2.7455e-06], + [ 5.0277e-05, 7.0512e-05, 4.1574e-06, ..., 2.1264e-05, + 2.2328e-04, 1.6782e-06], + ..., + [-1.4435e-07, 1.8626e-07, -1.5460e-07, ..., -7.6890e-06, + 1.3011e-06, -1.4283e-05], + [ 8.3297e-06, 1.1675e-05, -2.5406e-06, ..., 2.7120e-06, + 4.0859e-05, 1.3420e-06], + [ 1.1167e-06, 4.5914e-07, 2.3246e-06, ..., 5.1446e-06, + 2.0508e-06, 4.7274e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0091, -0.0291, 0.0106, -0.0184, 0.0141, 0.0046, 0.0189, -0.0075, + -0.0267, -0.0025], device='cuda:0'), grad: tensor([ 5.0694e-05, 3.0231e-04, 3.0732e-04, 9.9316e-06, 6.3062e-05, + 8.2105e-06, -7.9536e-04, -2.0072e-05, 5.8085e-05, 1.6108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 262.79, cls_loss 0.0095 cls_loss_mapping 0.0138 cls_loss_causal 0.6330 re_mapping 0.0128 re_causal 0.0384 /// teacc 98.72 lr 0.00010000 +Epoch 70, weight, value: tensor([[ 0.0106, -0.0991, -0.0501, ..., -0.0921, -0.0563, -0.0980], + [ 0.0131, -0.0265, 0.0007, ..., 0.0457, 0.0605, -0.0170], + [-0.0494, 0.0559, -0.0814, ..., 0.0465, 0.0233, -0.0028], + ..., + [-0.0584, -0.0468, -0.0580, ..., -0.0030, -0.0596, 0.0745], + [ 0.0613, -0.0094, 0.0203, ..., -0.0093, -0.0780, -0.0045], + [-0.0918, -0.0065, 0.0070, ..., -0.0991, 0.0266, -0.0716]], + device='cuda:0'), grad: tensor([[ 2.1458e-06, 5.7649e-07, 6.2920e-06, ..., 7.2084e-06, + 6.8061e-06, 1.3970e-07], + [ 2.1793e-07, -1.3839e-06, 2.1923e-04, ..., 1.3769e-04, + 5.0068e-04, 8.1304e-07], + [ 8.0019e-06, 3.1255e-06, 4.6730e-05, ..., 1.6004e-05, + -2.8014e-05, 1.0412e-06], + ..., + [ 2.5518e-07, 3.1479e-07, 1.8448e-05, ..., 1.0669e-05, + 4.6760e-05, -2.5835e-06], + [-1.6272e-05, -1.0682e-06, -6.7353e-05, ..., -8.8453e-05, + 3.1620e-05, 2.1886e-07], + [ 2.9597e-06, 1.3970e-07, 1.0604e-04, ..., 9.7871e-05, + 2.0182e-04, 2.0005e-06]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0097, -0.0292, 0.0109, -0.0177, 0.0145, 0.0040, 0.0189, -0.0077, + -0.0268, -0.0024], device='cuda:0'), grad: tensor([ 2.0742e-05, 7.8773e-04, 4.0293e-05, 1.0264e-04, -1.1959e-03, + 2.0817e-05, 4.0047e-08, 7.2718e-05, -1.1039e-04, 2.5940e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 69---------------------------------------------------- +epoch 69, time 279.29, cls_loss 0.0092 cls_loss_mapping 0.0142 cls_loss_causal 0.6208 re_mapping 0.0126 re_causal 0.0373 /// teacc 98.74 lr 0.00010000 +Epoch 71, weight, value: tensor([[ 0.0107, -0.0996, -0.0504, ..., -0.0926, -0.0570, -0.0981], + [ 0.0130, -0.0270, 0.0007, ..., 0.0454, 0.0609, -0.0174], + [-0.0497, 0.0566, -0.0814, ..., 0.0468, 0.0244, -0.0030], + ..., + [-0.0587, -0.0472, -0.0579, ..., -0.0030, -0.0603, 0.0754], + [ 0.0615, -0.0091, 0.0207, ..., -0.0090, -0.0785, -0.0046], + [-0.0919, -0.0068, 0.0068, ..., -0.0994, 0.0265, -0.0723]], + device='cuda:0'), grad: tensor([[ 1.2172e-06, 1.3746e-05, 2.4159e-06, ..., 3.0264e-05, + -4.7565e-05, 2.4047e-06], + [ 8.2981e-07, 1.2117e-06, -1.3411e-05, ..., -1.4948e-06, + -1.3120e-05, 3.2187e-06], + [ 3.8743e-06, -1.2970e-04, 9.6038e-06, ..., -2.4700e-04, + -3.6573e-04, 8.1584e-06], + ..., + [-1.9684e-05, 1.3866e-05, -1.0148e-05, ..., -1.1355e-05, + 4.9442e-05, -4.8995e-05], + [ 2.5649e-06, 8.5458e-06, -2.4289e-05, ..., -1.9312e-05, + 3.6597e-05, 2.7120e-06], + [-5.4054e-06, 3.9041e-05, -3.6001e-05, ..., 1.0544e-04, + 1.0449e-04, 2.3276e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0098, -0.0292, 0.0113, -0.0175, 0.0138, 0.0041, 0.0189, -0.0076, + -0.0264, -0.0026], device='cuda:0'), grad: tensor([-1.2362e-04, 5.9381e-06, -8.1539e-04, 2.6727e-04, 8.8990e-05, + 1.5008e-04, 1.2672e-04, -8.9526e-05, 1.9476e-05, 3.7003e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 262.26, cls_loss 0.0097 cls_loss_mapping 0.0152 cls_loss_causal 0.6165 re_mapping 0.0123 re_causal 0.0365 /// teacc 98.65 lr 0.00010000 +Epoch 72, weight, value: tensor([[ 0.0108, -0.1001, -0.0508, ..., -0.0934, -0.0572, -0.0985], + [ 0.0137, -0.0269, 0.0015, ..., 0.0458, 0.0620, -0.0170], + [-0.0503, 0.0566, -0.0816, ..., 0.0470, 0.0240, -0.0037], + ..., + [-0.0589, -0.0472, -0.0578, ..., -0.0026, -0.0603, 0.0764], + [ 0.0613, -0.0090, 0.0211, ..., -0.0087, -0.0792, -0.0049], + [-0.0919, -0.0070, 0.0061, ..., -0.1005, 0.0263, -0.0733]], + device='cuda:0'), grad: tensor([[ 9.1270e-06, 1.6578e-07, 1.2159e-05, ..., 1.9595e-05, + 5.2340e-06, 3.1479e-07], + [ 2.3082e-05, 2.2843e-05, 6.9141e-06, ..., 3.6478e-05, + 5.3197e-05, 1.3381e-05], + [ 1.0765e-04, -2.3484e-05, 8.6188e-05, ..., 1.8549e-04, + -1.0066e-05, 2.6971e-06], + ..., + [ 1.8358e-05, 1.5199e-06, 2.2173e-05, ..., 3.4779e-05, + 3.1441e-05, 3.4012e-06], + [ 2.8089e-05, 4.6752e-07, 6.8963e-05, ..., 6.9916e-05, + 4.0174e-05, 1.4296e-06], + [ 1.1075e-04, 1.8347e-07, 3.0398e-04, ..., 2.8706e-04, + 2.3127e-04, 1.2582e-06]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0096, -0.0286, 0.0112, -0.0188, 0.0138, 0.0042, 0.0189, -0.0069, + -0.0264, -0.0032], device='cuda:0'), grad: tensor([ 3.0667e-05, 1.0592e-04, 2.4772e-04, -1.2627e-03, -3.6025e-04, + 2.7752e-04, 1.8999e-05, 7.6234e-05, 1.4126e-04, 7.2527e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 262.52, cls_loss 0.0098 cls_loss_mapping 0.0141 cls_loss_causal 0.6240 re_mapping 0.0128 re_causal 0.0356 /// teacc 98.70 lr 0.00010000 +Epoch 73, weight, value: tensor([[ 0.0108, -0.1010, -0.0522, ..., -0.0946, -0.0570, -0.0990], + [ 0.0138, -0.0269, 0.0020, ..., 0.0462, 0.0628, -0.0179], + [-0.0520, 0.0563, -0.0828, ..., 0.0464, 0.0238, -0.0040], + ..., + [-0.0589, -0.0474, -0.0580, ..., -0.0027, -0.0609, 0.0765], + [ 0.0627, -0.0082, 0.0216, ..., -0.0081, -0.0802, -0.0055], + [-0.0924, -0.0069, 0.0055, ..., -0.1016, 0.0266, -0.0738]], + device='cuda:0'), grad: tensor([[ 1.4342e-05, 7.4133e-06, 4.5925e-05, ..., 4.3422e-05, + 1.0085e-04, 5.0180e-06], + [ 1.0235e-06, -8.6904e-05, -7.0274e-05, ..., -4.8065e-04, + -5.3310e-04, 2.8815e-06], + [ 6.0439e-05, 6.6102e-05, 1.8537e-04, ..., 4.2748e-04, + 3.4285e-04, 1.8552e-05], + ..., + [ 3.7253e-07, 2.1666e-05, 1.1221e-05, ..., 1.0121e-04, + 1.0884e-04, -4.2133e-06], + [-1.0204e-04, -3.1024e-05, -3.8481e-04, ..., -2.9778e-04, + -1.2666e-07, -3.3140e-05], + [ 4.1053e-06, 3.1162e-06, 7.9155e-05, ..., 7.1406e-05, + -4.6283e-05, 5.2154e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0093, -0.0285, 0.0105, -0.0184, 0.0140, 0.0043, 0.0185, -0.0070, + -0.0260, -0.0032], device='cuda:0'), grad: tensor([ 2.5344e-04, -6.4945e-04, 6.6996e-04, 3.0923e-04, 7.8380e-05, + -1.3542e-04, -1.0651e-04, 1.4830e-04, -6.8092e-04, 1.1259e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 262.54, cls_loss 0.0092 cls_loss_mapping 0.0138 cls_loss_causal 0.6373 re_mapping 0.0124 re_causal 0.0368 /// teacc 98.73 lr 0.00010000 +Epoch 74, weight, value: tensor([[ 0.0108, -0.1016, -0.0526, ..., -0.0954, -0.0569, -0.0998], + [ 0.0139, -0.0278, 0.0030, ..., 0.0460, 0.0633, -0.0181], + [-0.0522, 0.0570, -0.0835, ..., 0.0464, 0.0241, -0.0042], + ..., + [-0.0590, -0.0469, -0.0587, ..., -0.0022, -0.0614, 0.0773], + [ 0.0628, -0.0081, 0.0216, ..., -0.0081, -0.0815, -0.0059], + [-0.0926, -0.0076, 0.0046, ..., -0.1024, 0.0267, -0.0746]], + device='cuda:0'), grad: tensor([[ 3.3788e-06, 4.5672e-06, 1.4178e-05, ..., 2.0131e-05, + 2.2650e-05, 1.0803e-07], + [ 9.3412e-07, 1.2536e-06, -1.0729e-06, ..., 3.3546e-06, + 1.0759e-05, 6.2026e-07], + [ 3.5893e-06, -1.4808e-06, 2.3067e-05, ..., 1.0684e-05, + -4.1686e-06, -8.3819e-08], + ..., + [ 1.8114e-06, 3.7365e-06, 2.8357e-05, ..., 3.2812e-05, + 1.3733e-04, -4.2208e-06], + [ 1.8757e-06, -1.5154e-05, -1.1092e-04, ..., -1.6105e-04, + 1.8284e-05, 2.8312e-07], + [ 8.0466e-06, 1.1427e-06, 8.1003e-05, ..., 9.7394e-05, + 1.0085e-04, 1.8943e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0090, -0.0285, 0.0101, -0.0183, 0.0141, 0.0046, 0.0187, -0.0066, + -0.0266, -0.0035], device='cuda:0'), grad: tensor([ 8.9824e-05, 7.4089e-05, 7.0751e-05, -6.3610e-04, -8.4257e-04, + 1.0699e-04, 4.1175e-04, 5.6791e-04, -4.4131e-04, 5.9891e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 262.44, cls_loss 0.0101 cls_loss_mapping 0.0180 cls_loss_causal 0.6429 re_mapping 0.0122 re_causal 0.0352 /// teacc 98.70 lr 0.00010000 +Epoch 75, weight, value: tensor([[ 0.0119, -0.1025, -0.0519, ..., -0.0960, -0.0561, -0.1006], + [ 0.0137, -0.0285, 0.0035, ..., 0.0462, 0.0633, -0.0182], + [-0.0524, 0.0575, -0.0838, ..., 0.0464, 0.0247, -0.0047], + ..., + [-0.0586, -0.0464, -0.0592, ..., -0.0019, -0.0623, 0.0782], + [ 0.0631, -0.0082, 0.0226, ..., -0.0076, -0.0815, -0.0062], + [-0.0927, -0.0083, 0.0043, ..., -0.1030, 0.0271, -0.0753]], + device='cuda:0'), grad: tensor([[ 1.0245e-07, 6.8061e-06, 1.3905e-06, ..., 5.8711e-06, + 2.4036e-05, 4.6566e-06], + [ 2.5053e-07, 1.3843e-05, -9.4324e-06, ..., 2.7921e-06, + -1.3180e-05, 1.0304e-05], + [ 3.1572e-07, 3.3081e-05, 9.1642e-06, ..., 3.1680e-05, + 4.8608e-05, 2.5287e-05], + ..., + [ 3.8464e-07, -8.7202e-05, 5.8785e-06, ..., -6.8903e-05, + 7.3791e-05, -2.4274e-05], + [ 9.1456e-07, -3.4850e-06, -6.2063e-06, ..., -9.2387e-06, + 1.0557e-05, 4.1723e-06], + [ 2.9150e-07, 6.7279e-06, 5.0776e-06, ..., 1.9267e-05, + -2.2244e-04, -4.1038e-05]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0082, -0.0287, 0.0098, -0.0191, 0.0142, 0.0045, 0.0172, -0.0061, + -0.0253, -0.0037], device='cuda:0'), grad: tensor([ 5.0426e-05, 2.2814e-05, 1.5867e-04, 4.7326e-05, 9.7156e-05, + 3.9369e-05, -2.6412e-06, -4.5121e-05, 1.6183e-05, -3.8433e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 74---------------------------------------------------- +epoch 74, time 279.07, cls_loss 0.0071 cls_loss_mapping 0.0107 cls_loss_causal 0.6133 re_mapping 0.0120 re_causal 0.0369 /// teacc 98.82 lr 0.00010000 +Epoch 76, weight, value: tensor([[ 0.0119, -0.1031, -0.0528, ..., -0.0968, -0.0564, -0.1009], + [ 0.0140, -0.0292, 0.0039, ..., 0.0458, 0.0634, -0.0187], + [-0.0524, 0.0583, -0.0845, ..., 0.0465, 0.0248, -0.0052], + ..., + [-0.0588, -0.0467, -0.0593, ..., -0.0015, -0.0620, 0.0789], + [ 0.0628, -0.0085, 0.0226, ..., -0.0077, -0.0827, -0.0064], + [-0.0929, -0.0086, 0.0040, ..., -0.1035, 0.0273, -0.0757]], + device='cuda:0'), grad: tensor([[ 1.5832e-07, 4.1425e-06, 3.6582e-06, ..., 1.1332e-05, + 4.7684e-06, 8.4657e-07], + [-1.4625e-05, -8.0606e-07, -1.8275e-04, ..., -3.1567e-04, + -3.2496e-04, 9.1270e-06], + [ 1.0744e-05, -1.9014e-05, 1.1963e-04, ..., 2.0993e-04, + 1.7595e-04, 3.8922e-05], + ..., + [ 2.8983e-06, -2.6021e-06, 4.3392e-05, ..., 1.7673e-05, + 9.8646e-05, -6.9857e-05], + [ 6.8033e-07, 3.9488e-06, 1.7703e-05, ..., 2.3007e-05, + 1.5870e-05, 4.6343e-06], + [ 2.8545e-07, 1.3523e-06, 1.2763e-05, ..., -9.4548e-06, + -4.7505e-05, 6.4000e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0084, -0.0290, 0.0097, -0.0192, 0.0142, 0.0051, 0.0177, -0.0057, + -0.0259, -0.0039], device='cuda:0'), grad: tensor([-1.3657e-05, -4.1151e-04, 2.9802e-04, 1.9395e-04, 4.3631e-05, + -2.6393e-04, 2.8282e-05, 8.2433e-05, 9.0659e-05, -4.8131e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 262.72, cls_loss 0.0100 cls_loss_mapping 0.0139 cls_loss_causal 0.6361 re_mapping 0.0124 re_causal 0.0357 /// teacc 98.76 lr 0.00010000 +Epoch 77, weight, value: tensor([[ 0.0123, -0.1058, -0.0536, ..., -0.0981, -0.0569, -0.1014], + [ 0.0146, -0.0299, 0.0039, ..., 0.0456, 0.0636, -0.0195], + [-0.0523, 0.0588, -0.0851, ..., 0.0463, 0.0251, -0.0057], + ..., + [-0.0597, -0.0473, -0.0592, ..., -0.0012, -0.0635, 0.0800], + [ 0.0628, -0.0081, 0.0236, ..., -0.0073, -0.0831, -0.0066], + [-0.0935, -0.0068, 0.0036, ..., -0.1029, 0.0283, -0.0762]], + device='cuda:0'), grad: tensor([[-6.0759e-06, 5.7459e-05, 2.0508e-06, ..., 3.2689e-06, + 5.9664e-05, 2.2888e-05], + [ 3.3546e-06, 8.0392e-06, -1.4044e-05, ..., -1.1422e-05, + -3.2961e-05, 2.6032e-05], + [ 1.1837e-06, -1.9264e-04, 6.2473e-06, ..., 3.7309e-06, + -2.1994e-04, 9.6321e-05], + ..., + [ 1.1008e-06, 9.1255e-05, 1.6615e-05, ..., -1.2703e-05, + -8.1658e-05, -3.0003e-03], + [ 2.2892e-06, 2.2024e-05, -1.0252e-05, ..., -9.6858e-06, + 3.7402e-05, 1.1019e-05], + [ 9.1642e-07, 5.0366e-06, 7.3276e-06, ..., 1.3471e-05, + 1.1466e-05, 2.8059e-05]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0089, -0.0292, 0.0093, -0.0193, 0.0146, 0.0049, 0.0176, -0.0058, + -0.0254, -0.0033], device='cuda:0'), grad: tensor([ 1.7786e-04, 3.9279e-05, -2.7776e-04, 6.8784e-05, 6.2370e-03, + 4.9472e-05, 2.3171e-05, -6.5079e-03, 9.7871e-05, 9.3102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 262.28, cls_loss 0.0080 cls_loss_mapping 0.0124 cls_loss_causal 0.5894 re_mapping 0.0125 re_causal 0.0347 /// teacc 98.81 lr 0.00010000 +Epoch 78, weight, value: tensor([[ 0.0125, -0.1061, -0.0541, ..., -0.0987, -0.0572, -0.1017], + [ 0.0152, -0.0304, 0.0041, ..., 0.0457, 0.0639, -0.0198], + [-0.0527, 0.0599, -0.0849, ..., 0.0469, 0.0258, -0.0058], + ..., + [-0.0601, -0.0475, -0.0591, ..., -0.0014, -0.0647, 0.0809], + [ 0.0629, -0.0090, 0.0232, ..., -0.0082, -0.0846, -0.0068], + [-0.0938, -0.0062, 0.0030, ..., -0.1029, 0.0288, -0.0765]], + device='cuda:0'), grad: tensor([[-1.0073e-05, 2.3004e-06, -2.0750e-06, ..., 2.1085e-05, + 4.5300e-06, 6.2166e-07], + [ 1.2582e-06, 1.3340e-04, 2.3603e-04, ..., 1.2188e-03, + 5.6362e-04, 1.0721e-05], + [-4.3288e-06, 1.6439e-04, 3.3355e-04, ..., 1.7185e-03, + 7.4768e-04, 2.3484e-05], + ..., + [ 1.3383e-06, -3.6740e-04, -6.6328e-04, ..., -3.4561e-03, + -1.5421e-03, -8.5413e-05], + [ 5.9232e-06, 1.9863e-05, -6.7391e-06, ..., 2.7850e-05, + 3.9548e-05, 2.3674e-06], + [ 2.7642e-06, 2.9299e-06, 1.4968e-05, ..., 4.8608e-05, + 1.0177e-05, 3.2216e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0088, -0.0291, 0.0098, -0.0195, 0.0141, 0.0056, 0.0178, -0.0059, + -0.0265, -0.0030], device='cuda:0'), grad: tensor([-1.0580e-05, 2.1267e-03, 2.9736e-03, 6.2323e-04, 3.8087e-05, + 4.8369e-05, 2.3127e-05, -5.9700e-03, 4.9591e-05, 9.3818e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 262.48, cls_loss 0.0071 cls_loss_mapping 0.0116 cls_loss_causal 0.6216 re_mapping 0.0121 re_causal 0.0357 /// teacc 98.61 lr 0.00010000 +Epoch 79, weight, value: tensor([[ 0.0125, -0.1066, -0.0539, ..., -0.0993, -0.0567, -0.1023], + [ 0.0153, -0.0313, 0.0039, ..., 0.0454, 0.0638, -0.0201], + [-0.0527, 0.0613, -0.0851, ..., 0.0478, 0.0271, -0.0062], + ..., + [-0.0602, -0.0481, -0.0588, ..., -0.0012, -0.0651, 0.0817], + [ 0.0630, -0.0093, 0.0236, ..., -0.0080, -0.0857, -0.0066], + [-0.0939, -0.0065, 0.0028, ..., -0.1037, 0.0288, -0.0778]], + device='cuda:0'), grad: tensor([[ 1.4175e-06, 9.5367e-07, 7.3761e-06, ..., 7.9349e-06, + 1.3337e-05, 2.0675e-07], + [ 1.6978e-06, 4.5896e-06, -3.3319e-05, ..., -2.1398e-05, + -6.0707e-05, 6.2212e-07], + [ 1.2435e-05, -2.3693e-05, 4.3690e-05, ..., 1.0943e-06, + -1.3679e-05, 1.7984e-06], + ..., + [ 5.2750e-06, 1.7926e-05, 6.6347e-06, ..., 2.6673e-05, + 2.8118e-05, -1.8645e-06], + [ 3.1888e-05, 1.3001e-06, 8.2254e-05, ..., 7.7963e-05, + 4.6581e-05, 4.3912e-07], + [ 2.6114e-06, 1.0990e-06, 6.7428e-06, ..., 7.5847e-06, + 5.2713e-06, 1.9819e-06]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0075, -0.0298, 0.0107, -0.0198, 0.0133, 0.0052, 0.0180, -0.0053, + -0.0268, -0.0035], device='cuda:0'), grad: tensor([ 2.8133e-05, -7.5042e-05, 4.3750e-05, -2.0206e-04, -2.8163e-05, + -2.9597e-06, -1.5423e-05, 4.6462e-05, 1.8322e-04, 2.2054e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 262.52, cls_loss 0.0088 cls_loss_mapping 0.0117 cls_loss_causal 0.6554 re_mapping 0.0115 re_causal 0.0344 /// teacc 98.68 lr 0.00010000 +Epoch 80, weight, value: tensor([[ 0.0126, -0.1073, -0.0544, ..., -0.1000, -0.0569, -0.1037], + [ 0.0152, -0.0317, 0.0043, ..., 0.0458, 0.0637, -0.0206], + [-0.0529, 0.0622, -0.0857, ..., 0.0479, 0.0277, -0.0063], + ..., + [-0.0602, -0.0487, -0.0594, ..., -0.0020, -0.0661, 0.0817], + [ 0.0629, -0.0095, 0.0236, ..., -0.0081, -0.0871, -0.0067], + [-0.0942, -0.0065, 0.0032, ..., -0.1038, 0.0297, -0.0787]], + device='cuda:0'), grad: tensor([[ 8.2422e-08, 4.0047e-06, 6.5863e-06, ..., 1.5989e-05, + 4.4316e-05, 2.1514e-06], + [ 2.2165e-07, 2.6934e-06, -4.2558e-05, ..., -4.2319e-05, + -1.1659e-04, 5.0897e-07], + [-5.0431e-07, -2.1771e-05, 8.1956e-06, ..., -1.9029e-05, + -7.0557e-06, 1.9139e-07], + ..., + [ 5.4762e-07, 5.4017e-06, 8.8811e-06, ..., 1.5780e-05, + 3.4094e-05, 1.1828e-06], + [ 6.4261e-07, -1.7090e-07, -1.4015e-05, ..., -2.2918e-05, + 3.4511e-05, 3.7765e-07], + [ 2.5611e-07, 2.1271e-06, 3.1348e-06, ..., -4.6305e-06, + -7.3433e-05, 9.9000e-07]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0071, -0.0301, 0.0109, -0.0193, 0.0130, 0.0052, 0.0179, -0.0058, + -0.0273, -0.0029], device='cuda:0'), grad: tensor([ 1.2743e-04, -1.4830e-04, -1.0557e-05, 2.3842e-05, 1.0782e-04, + 1.2852e-05, -4.4554e-05, 1.2141e-04, 8.7321e-06, -1.9872e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 262.30, cls_loss 0.0083 cls_loss_mapping 0.0117 cls_loss_causal 0.6296 re_mapping 0.0118 re_causal 0.0347 /// teacc 98.65 lr 0.00010000 +Epoch 81, weight, value: tensor([[ 0.0138, -0.1080, -0.0549, ..., -0.1005, -0.0571, -0.1046], + [ 0.0152, -0.0300, 0.0050, ..., 0.0465, 0.0649, -0.0208], + [-0.0522, 0.0620, -0.0869, ..., 0.0477, 0.0274, -0.0073], + ..., + [-0.0605, -0.0490, -0.0596, ..., -0.0019, -0.0667, 0.0832], + [ 0.0629, -0.0093, 0.0239, ..., -0.0077, -0.0877, -0.0070], + [-0.0946, -0.0065, 0.0033, ..., -0.1045, 0.0296, -0.0790]], + device='cuda:0'), grad: tensor([[-1.2778e-05, 7.4180e-07, -2.9624e-05, ..., 3.2373e-06, + -1.4915e-03, 6.6590e-08], + [ 1.4044e-06, 1.2126e-06, 1.1422e-05, ..., 1.6138e-05, + 9.0003e-06, 7.0175e-07], + [ 1.9874e-06, -3.0309e-05, 1.5303e-05, ..., -8.1241e-05, + -8.7023e-05, 3.0734e-07], + ..., + [ 4.3167e-07, 1.4277e-06, 1.9863e-05, ..., 1.6898e-05, + 9.4846e-06, -2.9542e-06], + [ 8.9109e-06, 4.3884e-06, 2.4885e-05, ..., 2.2635e-05, + 7.9036e-05, 3.5949e-07], + [ 1.2144e-06, 2.8703e-06, 6.1616e-06, ..., 7.6056e-05, + -3.8967e-06, 3.5316e-06]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0066, -0.0295, 0.0105, -0.0194, 0.0128, 0.0044, 0.0183, -0.0056, + -0.0272, -0.0031], device='cuda:0'), grad: tensor([-2.9602e-03, 3.1263e-05, -3.5197e-05, 1.0693e-04, 9.2328e-05, + -5.4836e-04, 2.6894e-03, 4.3720e-05, 2.3437e-04, 3.4356e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 262.42, cls_loss 0.0070 cls_loss_mapping 0.0097 cls_loss_causal 0.5968 re_mapping 0.0113 re_causal 0.0345 /// teacc 98.80 lr 0.00010000 +Epoch 82, weight, value: tensor([[ 0.0140, -0.1082, -0.0550, ..., -0.1011, -0.0564, -0.1050], + [ 0.0152, -0.0301, 0.0052, ..., 0.0466, 0.0647, -0.0212], + [-0.0519, 0.0624, -0.0872, ..., 0.0480, 0.0278, -0.0072], + ..., + [-0.0607, -0.0492, -0.0600, ..., -0.0019, -0.0673, 0.0838], + [ 0.0629, -0.0096, 0.0240, ..., -0.0079, -0.0884, -0.0074], + [-0.0948, -0.0064, 0.0035, ..., -0.1048, 0.0302, -0.0794]], + device='cuda:0'), grad: tensor([[ 1.3830e-07, 2.0247e-06, 5.0291e-07, ..., 5.6922e-06, + 2.8647e-06, 1.9614e-06], + [ 9.6858e-08, 7.6294e-06, -9.4809e-07, ..., 1.9416e-05, + 2.8722e-06, 6.6943e-06], + [ 6.8452e-08, -1.3781e-04, 2.4457e-06, ..., -3.5167e-04, + -5.6624e-05, -9.0182e-05], + ..., + [ 3.0966e-07, 3.0071e-05, 1.6922e-06, ..., 5.9694e-05, + 1.5646e-05, -5.9493e-06], + [ 6.0163e-07, 7.6473e-05, 3.2224e-06, ..., 2.0492e-04, + 3.3319e-05, 5.4121e-05], + [ 6.1654e-07, 4.6194e-07, 3.3937e-06, ..., 1.3441e-05, + 1.9029e-05, 1.3165e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0057, -0.0298, 0.0107, -0.0200, 0.0125, 0.0050, 0.0181, -0.0059, + -0.0275, -0.0027], device='cuda:0'), grad: tensor([ 1.1876e-05, 2.6450e-05, -4.0627e-04, -1.5423e-05, -1.2435e-05, + -7.0632e-05, 5.0217e-05, 4.6581e-05, 2.8062e-04, 8.8990e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 262.41, cls_loss 0.0073 cls_loss_mapping 0.0104 cls_loss_causal 0.5684 re_mapping 0.0114 re_causal 0.0318 /// teacc 98.79 lr 0.00010000 +Epoch 83, weight, value: tensor([[ 0.0139, -0.1092, -0.0557, ..., -0.1021, -0.0566, -0.1057], + [ 0.0161, -0.0300, 0.0056, ..., 0.0468, 0.0655, -0.0210], + [-0.0520, 0.0629, -0.0876, ..., 0.0478, 0.0280, -0.0073], + ..., + [-0.0618, -0.0494, -0.0604, ..., -0.0016, -0.0679, 0.0841], + [ 0.0632, -0.0094, 0.0246, ..., -0.0077, -0.0891, -0.0075], + [-0.0952, -0.0064, 0.0032, ..., -0.1054, 0.0307, -0.0797]], + device='cuda:0'), grad: tensor([[-2.6554e-05, -2.9169e-06, 1.5218e-06, ..., 2.1886e-06, + 1.3560e-05, 1.0012e-07], + [ 6.2622e-06, -1.3568e-05, 2.7269e-06, ..., 1.8347e-07, + 2.2620e-05, 1.6525e-05], + [ 2.5872e-06, 9.4920e-06, 1.2919e-05, ..., 1.8597e-05, + 4.7088e-05, 7.1060e-07], + ..., + [ 7.2597e-07, 6.2492e-07, -5.5730e-06, ..., -3.3766e-05, + 1.5289e-05, -2.3350e-05], + [ 2.4121e-06, 7.0874e-07, 2.2352e-06, ..., 1.8375e-06, + 2.1085e-05, 4.9826e-07], + [ 2.1040e-05, 9.9745e-07, 8.2478e-06, ..., 1.4044e-05, + 4.5925e-05, 1.8338e-06]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0053, -0.0294, 0.0102, -0.0197, 0.0120, 0.0050, 0.0178, -0.0059, + -0.0273, -0.0026], device='cuda:0'), grad: tensor([-3.8266e-04, 1.7059e-04, 8.6010e-05, 9.5546e-05, -6.1083e-04, + 1.9252e-05, 3.4642e-04, -2.0012e-05, -7.0286e-04, 9.9850e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 82---------------------------------------------------- +epoch 82, time 279.62, cls_loss 0.0091 cls_loss_mapping 0.0129 cls_loss_causal 0.6147 re_mapping 0.0113 re_causal 0.0320 /// teacc 98.83 lr 0.00010000 +Epoch 84, weight, value: tensor([[ 0.0141, -0.1082, -0.0561, ..., -0.1028, -0.0569, -0.1061], + [ 0.0160, -0.0301, 0.0052, ..., 0.0461, 0.0651, -0.0223], + [-0.0520, 0.0634, -0.0880, ..., 0.0481, 0.0284, -0.0071], + ..., + [-0.0621, -0.0499, -0.0601, ..., -0.0019, -0.0673, 0.0843], + [ 0.0634, -0.0089, 0.0254, ..., -0.0064, -0.0899, -0.0061], + [-0.0955, -0.0072, 0.0029, ..., -0.1062, 0.0292, -0.0800]], + device='cuda:0'), grad: tensor([[ 4.2375e-08, 1.9027e-06, 4.0382e-06, ..., 8.6576e-06, + 3.2872e-05, 1.2852e-07], + [ 4.3306e-08, 1.2973e-06, -2.8610e-05, ..., -2.5705e-05, + -4.1068e-05, 1.2945e-07], + [ 1.8673e-07, -3.5375e-05, 1.1645e-05, ..., -4.1872e-05, + -3.7700e-05, 3.0780e-07], + ..., + [ 5.1688e-08, 2.3663e-05, 7.7561e-06, ..., 3.8892e-05, + 3.8207e-05, -5.0552e-06], + [-1.2010e-05, 2.9672e-06, -9.8228e-05, ..., -1.0014e-04, + 2.8595e-05, 4.1630e-07], + [ 5.8208e-08, 4.3120e-07, 1.0008e-04, ..., 1.4246e-04, + 1.4997e-04, 2.9113e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0045, -0.0304, 0.0102, -0.0193, 0.0131, 0.0051, 0.0176, -0.0058, + -0.0265, -0.0039], device='cuda:0'), grad: tensor([ 3.1447e-04, -4.8578e-05, -7.2956e-05, 2.3711e-04, -1.7002e-05, + -6.6566e-04, -3.2568e-04, 7.6056e-05, -8.7440e-05, 5.8889e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 262.22, cls_loss 0.0067 cls_loss_mapping 0.0091 cls_loss_causal 0.6275 re_mapping 0.0108 re_causal 0.0332 /// teacc 98.66 lr 0.00010000 +Epoch 85, weight, value: tensor([[ 0.0141, -0.1085, -0.0564, ..., -0.1033, -0.0574, -0.1064], + [ 0.0160, -0.0301, 0.0062, ..., 0.0471, 0.0659, -0.0224], + [-0.0517, 0.0623, -0.0887, ..., 0.0469, 0.0279, -0.0098], + ..., + [-0.0625, -0.0509, -0.0609, ..., -0.0027, -0.0684, 0.0850], + [ 0.0633, -0.0089, 0.0251, ..., -0.0069, -0.0920, -0.0062], + [-0.0958, -0.0074, 0.0026, ..., -0.1070, 0.0294, -0.0803]], + device='cuda:0'), grad: tensor([[-6.3516e-07, -5.3272e-07, 2.4997e-06, ..., 5.4352e-06, + 9.7081e-06, 5.3719e-06], + [ 2.2054e-06, 8.4471e-07, 6.0678e-05, ..., 1.2517e-04, + 3.2663e-05, 1.3888e-04], + [ 2.3730e-06, -1.6978e-06, 1.4059e-05, ..., 2.3335e-05, + 5.7854e-06, 2.1055e-05], + ..., + [ 4.5309e-07, 8.5356e-07, -4.3344e-04, ..., -8.8930e-04, + -2.2531e-04, -9.9659e-04], + [ 3.7104e-06, 2.6189e-06, 1.1966e-05, ..., 2.6777e-05, + 1.3143e-05, 3.1292e-05], + [ 2.2911e-06, 1.3076e-06, 2.4962e-04, ..., 5.1117e-04, + 1.1313e-04, 5.6934e-04]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0048, -0.0296, 0.0092, -0.0173, 0.0129, 0.0044, 0.0192, -0.0062, + -0.0278, -0.0039], device='cuda:0'), grad: tensor([-1.1855e-04, 3.9911e-04, 7.5519e-05, 5.9223e-04, 6.8843e-05, + 2.5138e-05, 2.3982e-07, -2.7485e-03, 1.0973e-04, 1.5984e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 262.17, cls_loss 0.0080 cls_loss_mapping 0.0114 cls_loss_causal 0.6499 re_mapping 0.0106 re_causal 0.0328 /// teacc 98.79 lr 0.00010000 +Epoch 86, weight, value: tensor([[ 0.0139, -0.1088, -0.0565, ..., -0.1037, -0.0572, -0.1071], + [ 0.0162, -0.0304, 0.0070, ..., 0.0482, 0.0663, -0.0206], + [-0.0519, 0.0628, -0.0893, ..., 0.0470, 0.0284, -0.0098], + ..., + [-0.0624, -0.0514, -0.0615, ..., -0.0036, -0.0694, 0.0850], + [ 0.0632, -0.0084, 0.0253, ..., -0.0065, -0.0925, -0.0064], + [-0.0965, -0.0080, 0.0017, ..., -0.1086, 0.0288, -0.0816]], + device='cuda:0'), grad: tensor([[ 3.4366e-07, 9.9912e-06, 1.1899e-05, ..., 2.3305e-05, + 1.8388e-05, 7.3016e-07], + [ 1.2899e-07, 4.6223e-05, 9.3877e-06, ..., 7.3910e-05, + 8.0824e-05, 3.3379e-06], + [ 4.4936e-07, -1.2088e-04, 2.6852e-05, ..., -1.5187e-04, + -1.9193e-04, 2.7213e-06], + ..., + [ 1.5646e-07, 4.2431e-06, 6.3442e-06, ..., -1.4290e-05, + 1.2688e-05, -1.9774e-05], + [ 2.6338e-06, 3.2157e-05, -2.3380e-05, ..., 2.5615e-05, + 4.7445e-05, 2.4326e-06], + [ 1.1493e-06, 7.9870e-06, 2.0790e-04, ..., 1.5152e-04, + 4.8965e-05, 6.1765e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0046, -0.0288, 0.0091, -0.0172, 0.0134, 0.0049, 0.0188, -0.0067, + -0.0277, -0.0049], device='cuda:0'), grad: tensor([ 5.1230e-05, 1.2624e-04, -2.4486e-04, -4.8494e-04, 1.4469e-05, + 7.5400e-05, -7.7859e-06, -2.2855e-06, 6.0141e-05, 4.1246e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 262.03, cls_loss 0.0086 cls_loss_mapping 0.0112 cls_loss_causal 0.6205 re_mapping 0.0108 re_causal 0.0326 /// teacc 98.74 lr 0.00010000 +Epoch 87, weight, value: tensor([[ 0.0136, -0.1099, -0.0573, ..., -0.1048, -0.0577, -0.1085], + [ 0.0169, -0.0307, 0.0051, ..., 0.0459, 0.0660, -0.0214], + [-0.0521, 0.0631, -0.0896, ..., 0.0471, 0.0291, -0.0104], + ..., + [-0.0625, -0.0507, -0.0597, ..., -0.0016, -0.0694, 0.0856], + [ 0.0628, -0.0086, 0.0260, ..., -0.0052, -0.0936, -0.0050], + [-0.0971, -0.0081, 0.0011, ..., -0.1092, 0.0292, -0.0822]], + device='cuda:0'), grad: tensor([[ 5.0897e-07, 3.3993e-07, 1.2759e-06, ..., 2.3022e-06, + 3.7514e-06, 1.3039e-08], + [-1.0990e-07, 3.5297e-07, -6.5714e-06, ..., -1.0133e-05, + -8.8736e-06, 3.2596e-08], + [ 4.7125e-07, -1.8366e-06, 3.2932e-06, ..., 3.1199e-06, + 2.5555e-06, 1.6298e-08], + ..., + [ 1.0645e-06, 4.3912e-07, 2.5090e-06, ..., 3.5800e-06, + 5.8226e-06, -4.2003e-07], + [ 5.2946e-07, -1.1409e-06, -6.4299e-06, ..., -1.1079e-05, + 1.9353e-06, 3.8184e-08], + [ 4.2422e-07, 9.7416e-07, 4.7907e-06, ..., 8.8066e-06, + 1.4961e-05, 1.6112e-07]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0048, -0.0304, 0.0092, -0.0175, 0.0133, 0.0049, 0.0193, -0.0054, + -0.0274, -0.0047], device='cuda:0'), grad: tensor([ 1.1571e-05, -1.5646e-05, 1.2144e-05, 2.7224e-05, -7.9930e-05, + -3.2336e-05, 1.8671e-05, 1.7181e-05, -1.1966e-05, 5.3018e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 262.79, cls_loss 0.0083 cls_loss_mapping 0.0106 cls_loss_causal 0.6494 re_mapping 0.0109 re_causal 0.0333 /// teacc 98.79 lr 0.00010000 +Epoch 88, weight, value: tensor([[ 0.0133, -0.1102, -0.0578, ..., -0.1063, -0.0592, -0.1114], + [ 0.0181, -0.0307, 0.0053, ..., 0.0460, 0.0664, -0.0221], + [-0.0527, 0.0634, -0.0903, ..., 0.0469, 0.0292, -0.0109], + ..., + [-0.0625, -0.0508, -0.0593, ..., -0.0009, -0.0694, 0.0872], + [ 0.0628, -0.0087, 0.0264, ..., -0.0054, -0.0946, -0.0053], + [-0.0976, -0.0082, 0.0004, ..., -0.1103, 0.0296, -0.0837]], + device='cuda:0'), grad: tensor([[-1.3271e-07, 1.4063e-07, 1.3607e-06, ..., 5.2080e-06, + -5.1785e-04, -1.8203e-04], + [ 1.6689e-06, 5.9092e-07, -1.3217e-05, ..., -2.4766e-05, + -3.3081e-05, 4.7721e-06], + [ 9.3598e-07, -1.8775e-06, 6.0424e-06, ..., 1.0028e-05, + 1.0943e-04, 3.3587e-05], + ..., + [ 1.6391e-07, 3.8743e-07, 9.6560e-06, ..., 1.5944e-05, + 3.2037e-05, -8.2795e-07], + [-6.2585e-06, -4.1258e-07, -1.8224e-05, ..., -3.5018e-06, + 4.0770e-05, 9.0897e-06], + [ 9.5461e-08, 6.9384e-08, 6.3218e-06, ..., 1.2383e-05, + 2.5168e-05, 8.3670e-06]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0058, -0.0304, 0.0088, -0.0169, 0.0133, 0.0048, 0.0189, -0.0046, + -0.0277, -0.0047], device='cuda:0'), grad: tensor([-1.0881e-03, -1.7077e-05, 2.3150e-04, 1.0437e-04, 2.0480e-04, + -2.1625e-04, 5.1403e-04, 7.4983e-05, 1.1945e-04, 7.1883e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 262.57, cls_loss 0.0082 cls_loss_mapping 0.0140 cls_loss_causal 0.6003 re_mapping 0.0112 re_causal 0.0323 /// teacc 98.70 lr 0.00010000 +Epoch 89, weight, value: tensor([[ 0.0136, -0.1100, -0.0582, ..., -0.1081, -0.0603, -0.1122], + [ 0.0185, -0.0300, 0.0056, ..., 0.0470, 0.0678, -0.0221], + [-0.0532, 0.0637, -0.0909, ..., 0.0467, 0.0290, -0.0110], + ..., + [-0.0627, -0.0516, -0.0597, ..., -0.0014, -0.0711, 0.0875], + [ 0.0630, -0.0088, 0.0268, ..., -0.0053, -0.0955, -0.0055], + [-0.0985, -0.0083, -0.0005, ..., -0.1110, 0.0310, -0.0835]], + device='cuda:0'), grad: tensor([[ 1.7323e-07, 9.5461e-08, 2.6776e-07, ..., 5.9931e-07, + 4.4750e-07, -1.2107e-07], + [ 8.3819e-09, 1.3066e-06, -7.7114e-07, ..., 4.8578e-06, + 2.6897e-06, 2.2929e-06], + [ 2.4820e-07, -2.4904e-06, 5.9092e-07, ..., -3.5018e-06, + -2.5127e-06, -1.2852e-07], + ..., + [ 3.3295e-07, 2.2864e-07, 8.7032e-07, ..., -1.4938e-05, + 2.0508e-06, -1.1377e-05], + [ 1.4734e-06, 1.6345e-07, 2.6058e-06, ..., 6.3889e-06, + 9.3132e-06, 2.3507e-06], + [ 1.6401e-06, 5.9139e-08, 2.1998e-06, ..., 3.1255e-06, + 2.8208e-05, 1.0524e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0063, -0.0297, 0.0086, -0.0172, 0.0125, 0.0054, 0.0183, -0.0054, + -0.0281, -0.0033], device='cuda:0'), grad: tensor([-4.1723e-06, 9.0450e-06, 2.7418e-06, 3.0659e-06, -1.1492e-04, + -5.5373e-05, 2.2173e-05, -1.1079e-05, 3.2753e-05, 1.1557e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 262.58, cls_loss 0.0069 cls_loss_mapping 0.0109 cls_loss_causal 0.5865 re_mapping 0.0110 re_causal 0.0312 /// teacc 98.77 lr 0.00010000 +Epoch 90, weight, value: tensor([[ 0.0137, -0.1104, -0.0586, ..., -0.1088, -0.0613, -0.1154], + [ 0.0184, -0.0301, 0.0059, ..., 0.0466, 0.0677, -0.0225], + [-0.0530, 0.0646, -0.0913, ..., 0.0474, 0.0304, -0.0112], + ..., + [-0.0627, -0.0520, -0.0598, ..., -0.0009, -0.0717, 0.0886], + [ 0.0631, -0.0089, 0.0273, ..., -0.0051, -0.0962, -0.0059], + [-0.0988, -0.0098, -0.0011, ..., -0.1125, 0.0314, -0.0826]], + device='cuda:0'), grad: tensor([[ 1.5236e-06, 1.1092e-06, 2.5705e-06, ..., 8.2776e-06, + 6.5088e-05, 1.2137e-05], + [ 1.9204e-06, 1.5711e-06, -7.3537e-06, ..., 3.1851e-06, + -5.9158e-06, 1.1772e-05], + [ 1.3085e-06, -2.4447e-07, 3.4813e-06, ..., 5.9530e-06, + 4.4107e-06, 6.6981e-06], + ..., + [-8.8140e-06, -6.2101e-06, 1.7975e-06, ..., -5.5969e-05, + 1.5637e-06, -1.3137e-04], + [ 2.7940e-07, 6.1607e-07, -8.2180e-06, ..., -1.2238e-06, + 6.0089e-06, 8.1956e-06], + [ 1.4696e-06, 1.0263e-06, 4.7274e-06, ..., 2.5392e-05, + -7.5281e-05, 6.9439e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0068, -0.0302, 0.0093, -0.0178, 0.0125, 0.0058, 0.0181, -0.0050, + -0.0281, -0.0030], device='cuda:0'), grad: tensor([ 1.5235e-04, 2.0996e-05, 3.0100e-05, 2.6718e-05, 9.3400e-05, + 3.9905e-05, -1.0192e-04, -3.6311e-04, 1.4238e-05, 8.7023e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 262.22, cls_loss 0.0053 cls_loss_mapping 0.0085 cls_loss_causal 0.5950 re_mapping 0.0106 re_causal 0.0315 /// teacc 98.71 lr 0.00010000 +Epoch 91, weight, value: tensor([[ 0.0139, -0.1102, -0.0595, ..., -0.1095, -0.0612, -0.1155], + [ 0.0192, -0.0301, 0.0065, ..., 0.0469, 0.0681, -0.0234], + [-0.0533, 0.0652, -0.0918, ..., 0.0476, 0.0303, -0.0115], + ..., + [-0.0629, -0.0519, -0.0600, ..., -0.0004, -0.0714, 0.0901], + [ 0.0630, -0.0096, 0.0280, ..., -0.0056, -0.0972, -0.0063], + [-0.0991, -0.0101, -0.0020, ..., -0.1132, 0.0310, -0.0835]], + device='cuda:0'), grad: tensor([[ 6.7055e-08, 3.3397e-06, 9.9372e-07, ..., 4.8801e-06, + 4.2617e-06, 1.9092e-08], + [ 1.1129e-07, 1.3663e-06, -2.9132e-05, ..., -2.5898e-05, + -4.7594e-05, 1.3737e-07], + [-3.6880e-07, -4.0084e-05, 3.1907e-06, ..., -4.4078e-05, + -3.7193e-05, 4.5169e-08], + ..., + [ 3.9674e-07, 5.4985e-06, 3.0156e-06, ..., 8.8438e-06, + 1.1273e-05, -1.8906e-07], + [ 2.5146e-08, 3.3230e-06, -4.3437e-06, ..., 8.1956e-07, + 2.0638e-05, 2.8405e-08], + [ 8.0559e-08, 2.1746e-07, 5.6857e-07, ..., 7.7253e-07, + 1.1045e-06, 3.2550e-07]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0064, -0.0301, 0.0092, -0.0185, 0.0126, 0.0061, 0.0186, -0.0042, + -0.0286, -0.0038], device='cuda:0'), grad: tensor([-2.0370e-05, -8.6546e-05, -9.4354e-05, 4.8906e-05, 2.0280e-05, + 5.6684e-05, 7.9647e-06, 2.3142e-05, 3.6120e-05, 8.1509e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 262.12, cls_loss 0.0069 cls_loss_mapping 0.0105 cls_loss_causal 0.6339 re_mapping 0.0104 re_causal 0.0320 /// teacc 98.83 lr 0.00010000 +Epoch 92, weight, value: tensor([[ 1.3766e-02, -1.1022e-01, -6.0017e-02, ..., -1.1029e-01, + -6.1421e-02, -1.1578e-01], + [ 2.0084e-02, -3.0368e-02, 6.6888e-03, ..., 4.6752e-02, + 6.8061e-02, -2.4182e-02], + [-5.3488e-02, 6.5497e-02, -9.2620e-02, ..., 4.7774e-02, + 3.1722e-02, -1.1159e-02], + ..., + [-6.3126e-02, -5.1962e-02, -5.9789e-02, ..., -1.1232e-04, + -7.1858e-02, 9.1025e-02], + [ 6.3028e-02, -9.6692e-03, 2.8991e-02, ..., -5.3622e-03, + -9.8100e-02, -6.8899e-03], + [-9.9297e-02, -1.0562e-02, -2.4371e-03, ..., -1.1417e-01, + 3.0349e-02, -8.4403e-02]], device='cuda:0'), grad: tensor([[ 1.3504e-08, 6.7567e-07, 8.2189e-07, ..., 1.7779e-06, + 2.2799e-06, 3.5111e-07], + [ 4.1444e-08, 5.0059e-07, -1.0401e-05, ..., -9.4399e-06, + -1.7196e-05, -5.7481e-06], + [ 5.3085e-08, -4.4238e-07, 2.7604e-06, ..., 3.4757e-06, + 5.8934e-06, 2.9914e-06], + ..., + [ 1.2340e-07, 8.6799e-07, 6.0759e-06, ..., 1.1604e-06, + 1.6913e-05, -2.1718e-06], + [ 1.2713e-07, 2.3618e-06, -9.7789e-08, ..., 3.8892e-06, + 8.7172e-06, 7.9582e-07], + [ 2.6543e-08, 1.1548e-06, -6.8285e-06, ..., 4.3921e-06, + -3.8952e-05, 1.6605e-06]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0061, -0.0303, 0.0093, -0.0183, 0.0130, 0.0059, 0.0182, -0.0040, + -0.0280, -0.0046], device='cuda:0'), grad: tensor([ 3.9265e-06, -2.1070e-05, 1.7047e-05, 2.6643e-05, 7.7724e-05, + -6.3360e-05, 1.4916e-05, 3.4332e-05, 2.6971e-05, -1.1688e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 262.70, cls_loss 0.0057 cls_loss_mapping 0.0074 cls_loss_causal 0.5860 re_mapping 0.0107 re_causal 0.0320 /// teacc 98.72 lr 0.00010000 +Epoch 93, weight, value: tensor([[ 0.0134, -0.1111, -0.0601, ..., -0.1113, -0.0613, -0.1160], + [ 0.0201, -0.0309, 0.0075, ..., 0.0470, 0.0683, -0.0240], + [-0.0551, 0.0651, -0.0935, ..., 0.0474, 0.0321, -0.0119], + ..., + [-0.0630, -0.0518, -0.0603, ..., -0.0004, -0.0721, 0.0912], + [ 0.0631, -0.0098, 0.0292, ..., -0.0055, -0.0990, -0.0075], + [-0.0997, -0.0107, -0.0032, ..., -0.1152, 0.0307, -0.0858]], + device='cuda:0'), grad: tensor([[ 2.4494e-07, 5.0664e-06, 7.0082e-07, ..., 1.0632e-05, + 7.5400e-06, 4.1211e-07], + [-3.4086e-07, 2.1756e-05, -1.4745e-05, ..., 2.1815e-05, + 1.0923e-05, -2.9691e-06], + [-3.0864e-06, -2.5272e-04, 1.3141e-06, ..., -4.2772e-04, + -2.5821e-04, -3.8594e-06], + ..., + [ 1.6056e-06, 8.2776e-06, 6.7130e-06, ..., 2.3693e-05, + 1.4566e-05, 1.9465e-06], + [ 2.3115e-06, 1.7536e-04, 3.2280e-06, ..., 2.8515e-04, + 1.7810e-04, 1.1306e-06], + [ 3.5390e-07, 1.8049e-06, 1.7332e-06, ..., 4.9211e-06, + 4.0121e-06, 5.7742e-07]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0060, -0.0301, 0.0085, -0.0170, 0.0125, 0.0054, 0.0188, -0.0041, + -0.0283, -0.0046], device='cuda:0'), grad: tensor([ 1.8582e-05, 2.4289e-05, -5.9175e-04, 9.6083e-05, 3.0342e-06, + 3.3200e-05, -3.4451e-05, 3.5465e-05, 4.0483e-04, 1.1049e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 262.36, cls_loss 0.0054 cls_loss_mapping 0.0082 cls_loss_causal 0.5731 re_mapping 0.0108 re_causal 0.0299 /// teacc 98.69 lr 0.00010000 +Epoch 94, weight, value: tensor([[ 0.0132, -0.1121, -0.0602, ..., -0.1124, -0.0613, -0.1162], + [ 0.0201, -0.0309, 0.0081, ..., 0.0475, 0.0690, -0.0242], + [-0.0549, 0.0659, -0.0944, ..., 0.0473, 0.0324, -0.0132], + ..., + [-0.0631, -0.0522, -0.0603, ..., -0.0002, -0.0726, 0.0924], + [ 0.0632, -0.0098, 0.0298, ..., -0.0051, -0.0995, -0.0074], + [-0.1000, -0.0124, -0.0037, ..., -0.1167, 0.0302, -0.0867]], + device='cuda:0'), grad: tensor([[ 9.8255e-08, 2.6554e-05, 2.8461e-05, ..., 4.1366e-05, + 3.1203e-05, 0.0000e+00], + [ 9.1270e-08, 4.1677e-07, 1.3809e-03, ..., 1.1950e-03, + 6.5613e-04, 0.0000e+00], + [-3.8091e-07, -6.2943e-05, 2.9624e-05, ..., -1.1779e-05, + -2.8580e-05, 0.0000e+00], + ..., + [ 5.2946e-07, 8.9034e-06, 1.1951e-05, ..., 1.9446e-05, + 2.5019e-05, 0.0000e+00], + [ 2.5239e-07, -9.2909e-06, -1.6518e-03, ..., -1.4362e-03, + -6.7329e-04, 0.0000e+00], + [ 2.8824e-07, 7.8529e-06, 2.8923e-05, ..., 3.1263e-05, + -6.7472e-05, 0.0000e+00]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0057, -0.0298, 0.0085, -0.0174, 0.0128, 0.0056, 0.0184, -0.0039, + -0.0280, -0.0053], device='cuda:0'), grad: tensor([ 6.3479e-05, 2.5673e-03, -5.8115e-05, 1.7297e-04, -4.9144e-05, + 1.5974e-04, -8.0541e-06, 7.6771e-05, -2.7752e-03, -1.5271e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 262.37, cls_loss 0.0072 cls_loss_mapping 0.0113 cls_loss_causal 0.6233 re_mapping 0.0103 re_causal 0.0313 /// teacc 98.79 lr 0.00010000 +Epoch 95, weight, value: tensor([[ 0.0131, -0.1121, -0.0609, ..., -0.1138, -0.0614, -0.1163], + [ 0.0210, -0.0310, 0.0075, ..., 0.0466, 0.0700, -0.0239], + [-0.0548, 0.0665, -0.0950, ..., 0.0477, 0.0328, -0.0133], + ..., + [-0.0640, -0.0526, -0.0594, ..., 0.0006, -0.0744, 0.0924], + [ 0.0635, -0.0099, 0.0320, ..., -0.0042, -0.0989, -0.0076], + [-0.1008, -0.0125, -0.0059, ..., -0.1170, 0.0299, -0.0863]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, 1.3970e-08, 5.6718e-07, ..., 1.0310e-06, + -3.6266e-06, 1.4203e-07], + [ 2.5332e-07, 2.7940e-08, -4.0457e-06, ..., -4.5188e-06, + -1.0960e-05, 1.8580e-07], + [ 1.1036e-07, -5.4436e-07, 1.6112e-06, ..., 1.3690e-06, + 3.2317e-06, 8.9873e-08], + ..., + [ 6.3051e-07, 9.4529e-08, 3.6936e-06, ..., 6.8136e-06, + 6.5416e-06, -6.2818e-07], + [ 4.4843e-07, 1.6298e-07, -4.2528e-05, ..., -9.2089e-05, + 3.3751e-06, 5.6345e-08], + [-8.5821e-07, 1.5367e-08, 2.6841e-06, ..., 6.1393e-06, + -4.7497e-06, 2.4354e-07]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0064, -0.0298, 0.0085, -0.0179, 0.0137, 0.0057, 0.0184, -0.0041, + -0.0267, -0.0059], device='cuda:0'), grad: tensor([-1.7270e-05, -8.4639e-06, 7.9647e-06, 3.3319e-05, 1.0626e-06, + 1.4853e-04, 2.0593e-05, 2.0981e-05, -2.1672e-04, 9.9391e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 262.11, cls_loss 0.0061 cls_loss_mapping 0.0105 cls_loss_causal 0.6150 re_mapping 0.0102 re_causal 0.0301 /// teacc 98.77 lr 0.00010000 +Epoch 96, weight, value: tensor([[ 0.0128, -0.1122, -0.0613, ..., -0.1157, -0.0623, -0.1165], + [ 0.0216, -0.0310, 0.0082, ..., 0.0466, 0.0702, -0.0244], + [-0.0551, 0.0670, -0.0955, ..., 0.0485, 0.0339, -0.0133], + ..., + [-0.0638, -0.0530, -0.0597, ..., 0.0006, -0.0751, 0.0930], + [ 0.0632, -0.0102, 0.0322, ..., -0.0045, -0.1002, -0.0078], + [-0.1013, -0.0126, -0.0067, ..., -0.1179, 0.0304, -0.0866]], + device='cuda:0'), grad: tensor([[-5.3585e-05, 3.0268e-08, 1.9029e-05, ..., 1.0833e-05, + -5.3421e-06, 3.8650e-08], + [ 8.8587e-06, 4.7497e-08, -1.1055e-06, ..., -1.3225e-06, + -4.3064e-06, 1.1176e-07], + [ 9.8497e-06, -7.8045e-07, 2.4363e-05, ..., 1.2554e-05, + 1.9688e-06, 3.7393e-07], + ..., + [ 3.9823e-06, 1.0990e-07, 8.2180e-06, ..., 4.3362e-06, + 3.4329e-06, -1.3020e-06], + [ 8.4937e-06, 4.6473e-07, 2.7716e-05, ..., 1.8448e-05, + 8.9332e-06, 7.9628e-08], + [ 2.4363e-05, 1.6298e-08, 8.6606e-05, ..., 5.0068e-05, + -2.1875e-05, 4.1537e-07]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0070, -0.0297, 0.0095, -0.0181, 0.0138, 0.0062, 0.0180, -0.0042, + -0.0274, -0.0057], device='cuda:0'), grad: tensor([-6.1393e-05, 1.7881e-05, 7.7546e-05, -4.5371e-04, 8.6188e-05, + 4.9770e-05, -2.9817e-05, 2.6450e-05, 9.9719e-05, 1.8740e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 95---------------------------------------------------- +epoch 95, time 278.80, cls_loss 0.0082 cls_loss_mapping 0.0111 cls_loss_causal 0.5944 re_mapping 0.0103 re_causal 0.0289 /// teacc 98.86 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 0.0128, -0.1125, -0.0623, ..., -0.1171, -0.0625, -0.1167], + [ 0.0220, -0.0311, 0.0079, ..., 0.0464, 0.0705, -0.0247], + [-0.0557, 0.0675, -0.0962, ..., 0.0487, 0.0343, -0.0135], + ..., + [-0.0639, -0.0532, -0.0597, ..., 0.0002, -0.0757, 0.0933], + [ 0.0641, -0.0105, 0.0332, ..., -0.0034, -0.1011, -0.0058], + [-0.1015, -0.0127, -0.0072, ..., -0.1183, 0.0305, -0.0866]], + device='cuda:0'), grad: tensor([[ 5.8711e-05, 1.1288e-06, 7.5102e-05, ..., 4.9584e-06, + 3.2540e-06, 6.7009e-07], + [ 1.4402e-05, 2.2560e-05, 4.1753e-05, ..., 6.9141e-05, + 7.3433e-05, 1.0312e-05], + [ 5.2124e-05, -4.4137e-05, 3.4690e-05, ..., -1.0401e-04, + -1.3053e-04, -1.5825e-05], + ..., + [ 4.1080e-04, 3.6843e-06, 2.8706e-04, ..., 8.6352e-06, + 1.5467e-05, -9.8348e-06], + [ 7.1704e-05, 1.4612e-06, -1.9386e-05, ..., -1.0842e-04, + -2.8163e-05, 1.6754e-06], + [ 9.5844e-05, 2.9290e-07, 5.9700e-04, ..., 4.1053e-06, + 5.5313e-05, 5.7258e-06]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0070, -0.0301, 0.0094, -0.0167, 0.0138, 0.0050, 0.0180, -0.0036, + -0.0270, -0.0064], device='cuda:0'), grad: tensor([ 0.0004, 0.0002, 0.0002, -0.0012, 0.0006, -0.0052, 0.0003, 0.0026, + 0.0004, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 96---------------------------------------------------- +epoch 96, time 278.41, cls_loss 0.0054 cls_loss_mapping 0.0084 cls_loss_causal 0.6091 re_mapping 0.0099 re_causal 0.0298 /// teacc 98.92 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0133, -0.1128, -0.0632, ..., -0.1181, -0.0627, -0.1168], + [ 0.0229, -0.0312, 0.0087, ..., 0.0467, 0.0714, -0.0247], + [-0.0554, 0.0680, -0.0967, ..., 0.0492, 0.0347, -0.0136], + ..., + [-0.0646, -0.0533, -0.0600, ..., 0.0003, -0.0766, 0.0936], + [ 0.0640, -0.0110, 0.0330, ..., -0.0044, -0.1036, -0.0058], + [-0.1011, -0.0127, -0.0068, ..., -0.1181, 0.0308, -0.0868]], + device='cuda:0'), grad: tensor([[ 6.6590e-08, 5.3551e-08, 3.4180e-07, ..., 6.3237e-07, + 5.3458e-06, 2.6869e-07], + [ 4.0978e-08, 2.7707e-07, 7.1712e-07, ..., 5.0813e-06, + 8.4564e-06, 3.6769e-06], + [-5.7742e-08, -2.5965e-06, 2.0728e-05, ..., 2.1547e-05, + 3.9116e-06, 1.7583e-06], + ..., + [ 8.6613e-08, 3.9814e-07, 3.4850e-06, ..., 2.3507e-06, + 1.3024e-05, -4.3772e-06], + [ 8.0047e-07, 5.7323e-07, -2.7031e-05, ..., -2.7537e-05, + -2.5816e-06, 3.3230e-06], + [ 2.9011e-07, 1.0245e-08, 1.5600e-06, ..., 4.0680e-06, + 1.0222e-05, 2.7400e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0068, -0.0296, 0.0098, -0.0174, 0.0137, 0.0057, 0.0180, -0.0037, + -0.0285, -0.0061], device='cuda:0'), grad: tensor([ 5.0455e-05, 3.2932e-05, 4.7594e-05, 4.8392e-06, -1.1551e-04, + -3.0667e-05, -5.9605e-05, 3.1561e-05, -2.7448e-05, 6.5744e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 262.30, cls_loss 0.0060 cls_loss_mapping 0.0083 cls_loss_causal 0.5627 re_mapping 0.0097 re_causal 0.0281 /// teacc 98.78 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0142, -0.1129, -0.0637, ..., -0.1188, -0.0645, -0.1168], + [ 0.0233, -0.0313, 0.0094, ..., 0.0473, 0.0721, -0.0246], + [-0.0555, 0.0682, -0.0974, ..., 0.0492, 0.0349, -0.0142], + ..., + [-0.0648, -0.0533, -0.0603, ..., 0.0004, -0.0773, 0.0942], + [ 0.0636, -0.0112, 0.0331, ..., -0.0047, -0.1045, -0.0058], + [-0.1019, -0.0129, -0.0076, ..., -0.1188, 0.0322, -0.0873]], + device='cuda:0'), grad: tensor([[ 6.4261e-08, 6.1002e-08, 2.3469e-07, ..., 4.0559e-07, + 6.2538e-07, 8.1025e-08], + [ 1.5274e-07, 1.5041e-07, -3.5502e-06, ..., -2.2203e-06, + -6.0573e-06, 4.2096e-07], + [-4.0699e-07, -6.6124e-07, 5.5274e-07, ..., -4.9360e-07, + -1.8161e-06, 4.0513e-07], + ..., + [-3.5856e-08, -9.5461e-08, 1.5711e-06, ..., -1.6624e-06, + 2.4084e-06, -2.5202e-06], + [ 7.0315e-08, 1.5367e-07, -1.9129e-06, ..., -1.1204e-06, + 3.0976e-06, 3.1199e-07], + [ 1.3318e-07, 3.4925e-08, 2.2864e-07, ..., 5.7742e-07, + 1.0254e-06, 4.6985e-07]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0079, -0.0292, 0.0096, -0.0176, 0.0133, 0.0056, 0.0185, -0.0038, + -0.0291, -0.0049], device='cuda:0'), grad: tensor([ 2.3982e-07, -6.1542e-06, -5.1968e-07, 4.4629e-06, -2.0694e-06, + 7.2345e-06, -5.2452e-06, -2.3767e-06, 3.1888e-06, 1.2368e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 262.07, cls_loss 0.0076 cls_loss_mapping 0.0099 cls_loss_causal 0.5878 re_mapping 0.0100 re_causal 0.0290 /// teacc 98.83 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0156, -0.1134, -0.0643, ..., -0.1197, -0.0651, -0.1169], + [ 0.0239, -0.0316, 0.0099, ..., 0.0475, 0.0727, -0.0247], + [-0.0560, 0.0688, -0.0983, ..., 0.0490, 0.0353, -0.0148], + ..., + [-0.0640, -0.0535, -0.0609, ..., -0.0011, -0.0780, 0.0932], + [ 0.0635, -0.0115, 0.0335, ..., -0.0045, -0.1056, -0.0056], + [-0.1036, -0.0133, -0.0071, ..., -0.1181, 0.0318, -0.0858]], + device='cuda:0'), grad: tensor([[ 3.6787e-07, 7.7765e-08, 2.5099e-07, ..., 5.2154e-07, + -1.2136e-04, 8.4285e-08], + [ 2.2165e-07, 1.1222e-07, -4.9314e-07, ..., 5.7090e-07, + -4.0606e-07, 6.2538e-07], + [ 3.0687e-07, -8.2003e-07, 9.5926e-07, ..., 5.2806e-07, + 3.0501e-07, 3.3155e-07], + ..., + [-3.2457e-07, 8.9873e-08, 9.2667e-07, ..., -7.6182e-06, + 1.6205e-06, -1.2897e-05], + [ 2.1569e-06, 1.7229e-08, 6.4187e-06, ..., 1.2264e-05, + 4.9770e-06, 6.2725e-07], + [ 5.0897e-07, 8.1025e-08, 4.9174e-06, ..., 1.3314e-05, + 1.1629e-04, 9.9167e-06]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0082, -0.0289, 0.0092, -0.0168, 0.0138, 0.0048, 0.0185, -0.0042, + -0.0297, -0.0044], device='cuda:0'), grad: tensor([-4.2701e-04, 5.5730e-06, 5.8338e-06, 5.4955e-05, 4.2021e-06, + -1.7154e-04, 2.6494e-05, -2.2143e-05, 6.4850e-05, 4.5872e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 262.23, cls_loss 0.0053 cls_loss_mapping 0.0078 cls_loss_causal 0.5646 re_mapping 0.0100 re_causal 0.0285 /// teacc 98.91 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0158, -0.1135, -0.0644, ..., -0.1203, -0.0650, -0.1170], + [ 0.0237, -0.0330, 0.0101, ..., 0.0477, 0.0727, -0.0251], + [-0.0554, 0.0703, -0.0986, ..., 0.0502, 0.0367, -0.0150], + ..., + [-0.0641, -0.0539, -0.0611, ..., -0.0012, -0.0784, 0.0937], + [ 0.0632, -0.0113, 0.0335, ..., -0.0044, -0.1051, -0.0057], + [-0.1042, -0.0138, -0.0077, ..., -0.1189, 0.0313, -0.0861]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 4.0047e-08, 1.0328e-06, ..., 7.9069e-07, + 1.6931e-06, 2.8871e-08], + [ 5.9605e-08, -3.4738e-07, -1.7866e-05, ..., -9.0972e-06, + -3.7849e-05, 2.8051e-06], + [ 1.9837e-07, -2.2259e-07, 4.0494e-06, ..., 3.2932e-06, + 4.0308e-06, 2.8871e-07], + ..., + [ 1.1642e-07, 3.6415e-07, 1.9446e-06, ..., -2.5257e-06, + 3.1330e-06, -4.1947e-06], + [ 7.3202e-07, 4.0047e-08, 1.8761e-05, ..., 1.4551e-05, + 1.7866e-05, 2.3749e-07], + [ 9.4995e-08, 1.3970e-08, 1.5441e-06, ..., 1.5004e-06, + -7.4971e-07, 1.6298e-07]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0076, -0.0291, 0.0104, -0.0173, 0.0142, 0.0053, 0.0178, -0.0044, + -0.0295, -0.0051], device='cuda:0'), grad: tensor([ 1.4221e-06, -4.3303e-05, 9.3132e-06, -1.6257e-05, 4.4368e-06, + -2.8275e-06, 1.1414e-05, -3.2652e-06, 3.8534e-05, 5.4389e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 262.57, cls_loss 0.0054 cls_loss_mapping 0.0087 cls_loss_causal 0.5706 re_mapping 0.0097 re_causal 0.0289 /// teacc 98.84 lr 0.00010000 +Epoch 102, weight, value: tensor([[ 0.0157, -0.1140, -0.0653, ..., -0.1212, -0.0650, -0.1171], + [ 0.0260, -0.0332, 0.0101, ..., 0.0473, 0.0745, -0.0245], + [-0.0555, 0.0705, -0.0997, ..., 0.0500, 0.0365, -0.0156], + ..., + [-0.0662, -0.0541, -0.0611, ..., -0.0008, -0.0809, 0.0939], + [ 0.0629, -0.0116, 0.0340, ..., -0.0045, -0.1054, -0.0056], + [-0.1053, -0.0139, -0.0088, ..., -0.1194, 0.0313, -0.0865]], + device='cuda:0'), grad: tensor([[-1.2387e-07, 3.4273e-07, 1.4696e-06, ..., 1.4026e-06, + 1.3616e-06, 2.0489e-08], + [-7.1414e-06, -4.9621e-05, -1.4818e-04, ..., -1.5938e-04, + -1.6308e-04, 2.2911e-07], + [ 5.0701e-06, 3.2246e-05, 9.4593e-05, ..., 1.0419e-04, + 1.0371e-04, 1.2461e-06], + ..., + [ 1.6522e-06, 1.0356e-05, 3.1888e-05, ..., 3.1292e-05, + 3.4958e-05, -2.5909e-06], + [ 1.8999e-07, 4.1723e-07, 1.6363e-06, ..., 1.5944e-06, + 3.3397e-06, 1.5832e-07], + [ 5.3272e-07, 1.2480e-07, 1.4342e-06, ..., 1.8477e-06, + -5.6159e-07, 6.1188e-07]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0077, -0.0281, 0.0098, -0.0168, 0.0143, 0.0055, 0.0181, -0.0051, + -0.0293, -0.0056], device='cuda:0'), grad: tensor([ 1.6596e-06, -2.9230e-04, 1.9205e-04, 2.2411e-05, 2.1420e-08, + 8.7619e-06, 4.3325e-06, 5.8204e-05, 6.8583e-06, -1.9297e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 262.01, cls_loss 0.0051 cls_loss_mapping 0.0071 cls_loss_causal 0.5731 re_mapping 0.0097 re_causal 0.0279 /// teacc 98.82 lr 0.00010000 +Epoch 103, weight, value: tensor([[ 0.0157, -0.1148, -0.0662, ..., -0.1221, -0.0650, -0.1173], + [ 0.0262, -0.0333, 0.0102, ..., 0.0473, 0.0744, -0.0249], + [-0.0552, 0.0705, -0.1006, ..., 0.0500, 0.0366, -0.0160], + ..., + [-0.0662, -0.0541, -0.0610, ..., -0.0004, -0.0815, 0.0948], + [ 0.0631, -0.0113, 0.0347, ..., -0.0047, -0.1062, -0.0067], + [-0.1053, -0.0138, -0.0089, ..., -0.1193, 0.0319, -0.0865]], + device='cuda:0'), grad: tensor([[ 2.4587e-07, 6.4727e-07, 4.3213e-07, ..., 1.2461e-06, + 1.5182e-03, 5.9698e-07], + [ 1.9036e-06, 1.3309e-06, 1.5080e-05, ..., 1.6401e-06, + 6.6280e-05, 7.2181e-05], + [ 3.0175e-07, -1.9878e-05, 5.7090e-07, ..., -2.6330e-05, + -1.6078e-05, 5.7928e-07], + ..., + [-2.6710e-06, 4.6864e-06, 7.0930e-06, ..., 4.8205e-06, + 4.5896e-05, 1.6868e-05], + [ 2.3376e-07, 1.0803e-05, 1.4352e-06, ..., 1.5408e-05, + 3.1918e-05, 1.8571e-06], + [ 1.4519e-06, 2.2072e-07, 9.9540e-06, ..., 3.5483e-06, + -1.5888e-03, 2.3618e-05]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0077, -0.0286, 0.0093, -0.0173, 0.0150, 0.0057, 0.0182, -0.0050, + -0.0294, -0.0052], device='cuda:0'), grad: tensor([ 2.2202e-03, 2.3687e-04, -3.5346e-05, -1.0282e-05, -2.7061e-04, + 5.5999e-05, -5.5730e-05, 1.1176e-04, 9.5904e-05, -2.3479e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 262.52, cls_loss 0.0058 cls_loss_mapping 0.0080 cls_loss_causal 0.6027 re_mapping 0.0095 re_causal 0.0274 /// teacc 98.78 lr 0.00010000 +Epoch 104, weight, value: tensor([[ 0.0156, -0.1153, -0.0674, ..., -0.1236, -0.0654, -0.1174], + [ 0.0263, -0.0346, 0.0102, ..., 0.0471, 0.0742, -0.0250], + [-0.0555, 0.0727, -0.1009, ..., 0.0509, 0.0381, -0.0152], + ..., + [-0.0658, -0.0563, -0.0611, ..., -0.0008, -0.0828, 0.0947], + [ 0.0630, -0.0113, 0.0349, ..., -0.0048, -0.1068, -0.0070], + [-0.1055, -0.0142, -0.0089, ..., -0.1196, 0.0313, -0.0866]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.0058e-07, 1.7015e-06, ..., 1.9502e-06, + 1.7388e-06, 2.7940e-09], + [ 4.6566e-09, -6.3777e-05, -1.9538e-04, ..., -3.7909e-04, + -4.5013e-04, 1.4901e-08], + [ 7.4506e-09, 5.6773e-05, 1.2755e-04, ..., 2.8229e-04, + 3.5739e-04, 8.3819e-09], + ..., + [-9.3132e-10, 5.2638e-06, 5.8860e-05, ..., 8.1956e-05, + 7.3075e-05, -7.2643e-08], + [ 2.7008e-08, 5.0291e-07, 7.7784e-06, ..., 9.2015e-06, + 1.2204e-05, 4.6566e-09], + [ 8.3819e-09, 1.2387e-07, 8.9779e-06, ..., 8.0243e-06, + -1.2957e-05, 2.5146e-08]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0081, -0.0288, 0.0104, -0.0174, 0.0166, 0.0062, 0.0186, -0.0055, + -0.0301, -0.0060], device='cuda:0'), grad: tensor([ 7.8231e-07, -5.5599e-04, 4.2558e-04, -8.4817e-05, 3.4243e-05, + 6.3062e-05, -1.3456e-05, 1.1736e-04, 4.2468e-05, -2.8491e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 262.80, cls_loss 0.0045 cls_loss_mapping 0.0078 cls_loss_causal 0.5787 re_mapping 0.0103 re_causal 0.0280 /// teacc 98.88 lr 0.00010000 +Epoch 105, weight, value: tensor([[ 0.0155, -0.1154, -0.0686, ..., -0.1252, -0.0657, -0.1175], + [ 0.0264, -0.0351, 0.0106, ..., 0.0471, 0.0744, -0.0251], + [-0.0559, 0.0736, -0.1015, ..., 0.0516, 0.0395, -0.0148], + ..., + [-0.0655, -0.0563, -0.0611, ..., -0.0008, -0.0834, 0.0954], + [ 0.0631, -0.0117, 0.0351, ..., -0.0052, -0.1077, -0.0073], + [-0.1060, -0.0146, -0.0092, ..., -0.1201, 0.0316, -0.0867]], + device='cuda:0'), grad: tensor([[ 2.8405e-07, 4.1351e-07, 1.5264e-06, ..., 1.7220e-06, + -5.2247e-07, 7.9162e-08], + [ 4.0699e-07, 3.0268e-07, 4.0047e-06, ..., 3.4105e-06, + 2.4550e-06, 1.8161e-07], + [-1.1036e-06, -8.1509e-06, 1.1407e-05, ..., -2.0210e-06, + -5.4240e-06, 9.9558e-07], + ..., + [ 1.2387e-06, 4.5784e-06, 1.9342e-05, ..., 1.9461e-05, + 1.8805e-05, 1.3374e-06], + [ 6.8638e-07, 1.2992e-06, 5.4799e-06, ..., 5.4613e-06, + 4.0904e-06, 3.2503e-07], + [ 9.8255e-07, 1.2387e-07, -3.5930e-06, ..., 7.0855e-06, + -4.5925e-05, 4.6846e-07]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0083, -0.0288, 0.0115, -0.0176, 0.0158, 0.0067, 0.0183, -0.0052, + -0.0307, -0.0061], device='cuda:0'), grad: tensor([-9.5144e-06, 9.0003e-06, -3.7905e-06, -5.1069e-04, 8.8334e-05, + 4.4894e-04, 2.6077e-06, 4.6164e-05, 1.3910e-05, -8.4400e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 262.77, cls_loss 0.0051 cls_loss_mapping 0.0077 cls_loss_causal 0.5943 re_mapping 0.0092 re_causal 0.0275 /// teacc 98.90 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 0.0156, -0.1157, -0.0696, ..., -0.1258, -0.0674, -0.1176], + [ 0.0265, -0.0353, 0.0107, ..., 0.0473, 0.0744, -0.0253], + [-0.0561, 0.0740, -0.1020, ..., 0.0517, 0.0397, -0.0151], + ..., + [-0.0654, -0.0564, -0.0612, ..., -0.0007, -0.0840, 0.0960], + [ 0.0631, -0.0119, 0.0357, ..., -0.0052, -0.1085, -0.0075], + [-0.1062, -0.0148, -0.0092, ..., -0.1200, 0.0331, -0.0870]], + device='cuda:0'), grad: tensor([[ 1.3318e-07, -5.4911e-06, 5.5972e-07, ..., 1.3448e-06, + -2.6580e-06, 2.8312e-07], + [ 2.8126e-07, 1.6084e-06, 3.9022e-07, ..., 4.1053e-06, + 4.8243e-06, 1.3700e-06], + [ 1.2433e-06, -1.1265e-05, -3.3844e-06, ..., -1.6898e-05, + -3.1739e-05, 9.7416e-07], + ..., + [ 2.1234e-07, 1.0766e-06, -2.3007e-05, ..., -9.6440e-05, + -5.8413e-05, -8.9705e-05], + [ 3.0175e-07, 6.8434e-06, 6.6347e-06, ..., 1.9565e-05, + 2.1651e-05, 9.1046e-06], + [ 9.4250e-07, 3.0994e-06, 1.8314e-05, ..., 7.0810e-05, + 4.8846e-05, 6.3896e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0096, -0.0290, 0.0113, -0.0179, 0.0158, 0.0066, 0.0187, -0.0051, + -0.0306, -0.0052], device='cuda:0'), grad: tensor([-6.5982e-05, 1.7166e-05, -3.2902e-05, 2.5421e-05, 5.3346e-05, + 3.7663e-06, 6.2957e-06, -5.0449e-04, 8.5652e-05, 4.1103e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 262.68, cls_loss 0.0054 cls_loss_mapping 0.0108 cls_loss_causal 0.6164 re_mapping 0.0099 re_causal 0.0289 /// teacc 98.90 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0155, -0.1162, -0.0711, ..., -0.1262, -0.0675, -0.1178], + [ 0.0266, -0.0354, 0.0123, ..., 0.0477, 0.0755, -0.0257], + [-0.0555, 0.0742, -0.1026, ..., 0.0516, 0.0395, -0.0153], + ..., + [-0.0652, -0.0565, -0.0611, ..., -0.0002, -0.0839, 0.0965], + [ 0.0623, -0.0116, 0.0361, ..., -0.0049, -0.1101, -0.0076], + [-0.1066, -0.0148, -0.0116, ..., -0.1220, 0.0328, -0.0873]], + device='cuda:0'), grad: tensor([[ 5.7742e-08, -6.6217e-07, 4.3623e-06, ..., 4.0345e-06, + 3.2634e-06, 3.1665e-07], + [ 8.2515e-07, 7.9162e-08, -8.7595e-04, ..., -7.4911e-04, + -8.5926e-04, 1.9558e-07], + [ 5.0329e-06, -9.5647e-07, 1.5509e-04, ..., 1.2326e-04, + 5.4240e-05, 2.5984e-07], + ..., + [ 2.1514e-07, 1.0803e-07, 5.2184e-05, ..., 4.0174e-05, + 4.4078e-05, -2.3972e-06], + [ 3.3155e-06, 7.5903e-07, 8.6451e-04, ..., 7.3004e-04, + 7.7868e-04, -3.5483e-07], + [ 1.1921e-07, 4.6566e-08, 1.1809e-05, ..., 1.0878e-05, + 8.5458e-06, 8.8941e-07]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0096, -0.0281, 0.0111, -0.0179, 0.0159, 0.0061, 0.0189, -0.0047, + -0.0309, -0.0058], device='cuda:0'), grad: tensor([-6.0678e-05, -1.7900e-03, 2.7537e-04, -3.5977e-04, 9.2834e-06, + -6.0908e-06, 2.5406e-06, 9.4712e-05, 1.8005e-03, 3.3706e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 262.70, cls_loss 0.0067 cls_loss_mapping 0.0102 cls_loss_causal 0.6286 re_mapping 0.0094 re_causal 0.0274 /// teacc 98.82 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0158, -0.1164, -0.0719, ..., -0.1271, -0.0678, -0.1180], + [ 0.0266, -0.0356, 0.0134, ..., 0.0482, 0.0762, -0.0255], + [-0.0559, 0.0744, -0.1036, ..., 0.0513, 0.0396, -0.0162], + ..., + [-0.0652, -0.0566, -0.0618, ..., -0.0003, -0.0849, 0.0967], + [ 0.0623, -0.0118, 0.0360, ..., -0.0051, -0.1114, -0.0077], + [-0.1069, -0.0149, -0.0113, ..., -0.1211, 0.0340, -0.0869]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 9.9652e-08, 1.3039e-07, ..., 2.4401e-07, + -2.1141e-06, 2.2352e-08], + [ 7.0781e-08, 1.9558e-08, -9.5833e-07, ..., -4.3772e-07, + -1.0263e-06, 1.5832e-07], + [ 3.8184e-08, -1.4752e-06, 6.0163e-07, ..., -4.1313e-06, + -4.4443e-06, 1.5181e-07], + ..., + [ 2.8871e-08, 2.0582e-07, 1.5143e-06, ..., 4.7032e-07, + 3.0957e-06, -4.1071e-07], + [-9.8906e-07, 6.4261e-08, -1.5758e-06, ..., -5.4017e-08, + 1.7826e-06, 3.6135e-07], + [ 1.4901e-08, 2.0489e-08, 6.3963e-06, ..., 1.1642e-06, + 1.8597e-05, 2.5164e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0092, -0.0278, 0.0103, -0.0183, 0.0146, 0.0057, 0.0199, -0.0053, + -0.0317, -0.0038], device='cuda:0'), grad: tensor([-6.8605e-05, 3.8333e-06, -4.4405e-06, 5.8450e-06, -6.2406e-05, + -2.8864e-05, 1.3947e-05, 8.8438e-06, 6.6832e-06, 1.2505e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 262.67, cls_loss 0.0049 cls_loss_mapping 0.0073 cls_loss_causal 0.5841 re_mapping 0.0093 re_causal 0.0265 /// teacc 98.90 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0159, -0.1167, -0.0724, ..., -0.1278, -0.0672, -0.1182], + [ 0.0265, -0.0363, 0.0132, ..., 0.0477, 0.0757, -0.0266], + [-0.0556, 0.0753, -0.1035, ..., 0.0522, 0.0411, -0.0161], + ..., + [-0.0656, -0.0570, -0.0616, ..., -0.0002, -0.0853, 0.0975], + [ 0.0624, -0.0118, 0.0367, ..., -0.0049, -0.1122, -0.0077], + [-0.1072, -0.0151, -0.0117, ..., -0.1213, 0.0333, -0.0869]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -4.4331e-07, 1.1092e-06, ..., 1.5702e-06, + 2.5332e-06, -3.4831e-07], + [ 3.8184e-08, 6.2659e-06, -6.0946e-06, ..., 2.6584e-05, + 2.5466e-05, 8.8941e-07], + [ 1.1455e-07, -9.0227e-06, 1.1511e-05, ..., -1.4730e-05, + -3.5673e-05, 2.1048e-06], + ..., + [-2.0582e-07, 1.7881e-06, -3.8091e-07, ..., -2.6003e-05, + 1.7896e-05, -2.5928e-05], + [ 1.3597e-07, 6.4168e-07, 8.9332e-06, ..., 8.3148e-06, + 1.1615e-05, 3.4459e-06], + [ 3.7253e-08, 7.1712e-08, 6.6981e-06, ..., 1.4462e-05, + 6.5118e-06, 1.3582e-05]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0076, -0.0284, 0.0112, -0.0185, 0.0150, 0.0058, 0.0194, -0.0052, + -0.0316, -0.0048], device='cuda:0'), grad: tensor([-8.4490e-06, 4.4197e-05, 1.5214e-05, 8.2105e-06, 6.8247e-05, + -4.1246e-04, 1.2708e-04, -4.8995e-05, 1.1569e-04, 9.1493e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 262.62, cls_loss 0.0044 cls_loss_mapping 0.0070 cls_loss_causal 0.5666 re_mapping 0.0092 re_causal 0.0273 /// teacc 98.76 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 1.6203e-02, -1.1709e-01, -7.1949e-02, ..., -1.2855e-01, + -6.7447e-02, -1.1835e-01], + [ 2.6430e-02, -3.6436e-02, 1.3321e-02, ..., 4.7838e-02, + 7.5649e-02, -2.6720e-02], + [-5.6039e-02, 7.5361e-02, -1.0440e-01, ..., 5.2119e-02, + 4.1277e-02, -1.6652e-02], + ..., + [-6.5382e-02, -5.6908e-02, -6.1856e-02, ..., -7.2966e-05, + -8.5671e-02, 9.8264e-02], + [ 6.2353e-02, -1.1939e-02, 3.6987e-02, ..., -4.7849e-03, + -1.1255e-01, -7.7472e-03], + [-1.0759e-01, -1.5421e-02, -1.2209e-02, ..., -1.2193e-01, + 3.3372e-02, -8.7601e-02]], device='cuda:0'), grad: tensor([[ 2.2817e-06, 3.0082e-07, 1.2159e-05, ..., 9.3579e-06, + 7.7933e-06, 9.3132e-09], + [ 1.5721e-06, 4.2282e-07, 6.0238e-06, ..., 8.1435e-06, + 8.1733e-06, 8.2888e-08], + [ 2.4531e-06, -2.6319e-06, 8.3223e-06, ..., 1.0416e-05, + 9.2909e-06, 3.6322e-08], + ..., + [ 9.4436e-07, 5.2527e-07, 2.8312e-06, ..., 4.5858e-06, + 6.5751e-06, -5.2527e-07], + [-3.2783e-07, 1.8068e-07, -3.3319e-05, ..., -1.9506e-05, + 3.5204e-06, 2.2352e-08], + [-2.0504e-05, 6.5286e-07, -3.6567e-05, ..., -8.7500e-05, + -1.1933e-04, 2.3190e-07]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0073, -0.0286, 0.0107, -0.0199, 0.0149, 0.0076, 0.0198, -0.0049, + -0.0317, -0.0052], device='cuda:0'), grad: tensor([ 5.3614e-05, 3.5703e-05, 4.8667e-05, 2.0266e-04, 2.4930e-05, + 1.0289e-05, 6.2466e-05, 2.0862e-05, -7.1287e-05, -3.8815e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 262.82, cls_loss 0.0042 cls_loss_mapping 0.0062 cls_loss_causal 0.5934 re_mapping 0.0091 re_causal 0.0271 /// teacc 98.85 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0161, -0.1179, -0.0728, ..., -0.1308, -0.0674, -0.1184], + [ 0.0269, -0.0371, 0.0126, ..., 0.0469, 0.0753, -0.0283], + [-0.0559, 0.0764, -0.1055, ..., 0.0525, 0.0420, -0.0167], + ..., + [-0.0660, -0.0574, -0.0610, ..., 0.0005, -0.0857, 0.0995], + [ 0.0624, -0.0121, 0.0374, ..., -0.0044, -0.1131, -0.0078], + [-0.1076, -0.0157, -0.0126, ..., -0.1220, 0.0331, -0.0881]], + device='cuda:0'), grad: tensor([[ 4.9360e-07, 2.9486e-06, 2.7940e-05, ..., 1.2517e-05, + 2.2709e-05, 4.3772e-08], + [ 1.0030e-06, 9.1735e-07, -1.2910e-04, ..., -4.7237e-05, + -9.8944e-05, 1.7602e-07], + [ 2.6450e-07, -3.2043e-04, -2.2364e-04, ..., -4.1199e-04, + -3.6478e-04, 5.4669e-07], + ..., + [ 5.0291e-07, 2.5518e-07, 1.3351e-05, ..., 6.9141e-06, + 1.1772e-05, -1.8468e-06], + [ 6.5193e-07, 3.0851e-04, 2.9325e-04, ..., 4.2462e-04, + 4.0865e-04, 9.9652e-08], + [ 8.2608e-07, 3.4180e-06, 4.5747e-06, ..., 5.8785e-06, + 5.1409e-06, 2.7381e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0072, -0.0296, 0.0111, -0.0198, 0.0151, 0.0075, 0.0202, -0.0044, + -0.0314, -0.0056], device='cuda:0'), grad: tensor([ 1.8721e-03, -3.3951e-04, -9.1743e-04, 1.7989e-04, 2.7016e-05, + -2.6360e-03, 1.9395e-04, 1.5152e-04, 1.1816e-03, 2.9016e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 262.55, cls_loss 0.0039 cls_loss_mapping 0.0060 cls_loss_causal 0.5753 re_mapping 0.0090 re_causal 0.0269 /// teacc 98.90 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0161, -0.1184, -0.0747, ..., -0.1317, -0.0675, -0.1185], + [ 0.0269, -0.0373, 0.0134, ..., 0.0474, 0.0759, -0.0285], + [-0.0553, 0.0767, -0.1060, ..., 0.0522, 0.0420, -0.0191], + ..., + [-0.0671, -0.0570, -0.0619, ..., 0.0006, -0.0860, 0.1011], + [ 0.0626, -0.0123, 0.0375, ..., -0.0045, -0.1140, -0.0078], + [-0.1077, -0.0161, -0.0127, ..., -0.1221, 0.0332, -0.0883]], + device='cuda:0'), grad: tensor([[-4.9919e-07, 4.4703e-08, 6.3796e-07, ..., 7.8976e-07, + -3.9116e-06, 4.8615e-07], + [ 2.6077e-08, 2.2352e-07, 5.0396e-05, ..., 5.4806e-05, + 1.5154e-05, 3.6150e-05], + [ 3.6322e-08, -5.5879e-08, 8.2478e-06, ..., 8.9929e-06, + 2.1756e-06, 6.8918e-06], + ..., + [ 8.9407e-08, -1.4920e-06, -2.2221e-04, ..., -2.4176e-04, + -6.6221e-05, -1.6284e-04], + [ 2.1886e-07, 2.9802e-07, 8.3297e-06, ..., 1.0274e-05, + 3.3304e-06, 7.7412e-06], + [ 9.1363e-07, 3.7067e-07, 9.3341e-05, ..., 9.9003e-05, + 2.8789e-05, 6.4731e-05]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0076, -0.0290, 0.0101, -0.0197, 0.0152, 0.0077, 0.0200, -0.0041, + -0.0316, -0.0056], device='cuda:0'), grad: tensor([-1.3523e-05, 1.4579e-04, 2.7403e-05, 2.1982e-04, 1.4022e-05, + -5.6088e-05, 2.7083e-06, -6.4373e-04, 2.9325e-05, 2.7418e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 262.24, cls_loss 0.0051 cls_loss_mapping 0.0068 cls_loss_causal 0.5767 re_mapping 0.0093 re_causal 0.0271 /// teacc 98.78 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 0.0164, -0.1189, -0.0747, ..., -0.1327, -0.0669, -0.1191], + [ 0.0264, -0.0388, 0.0147, ..., 0.0488, 0.0760, -0.0263], + [-0.0550, 0.0776, -0.1052, ..., 0.0523, 0.0433, -0.0193], + ..., + [-0.0674, -0.0569, -0.0641, ..., -0.0007, -0.0878, 0.0998], + [ 0.0627, -0.0127, 0.0380, ..., -0.0044, -0.1151, -0.0072], + [-0.1084, -0.0166, -0.0131, ..., -0.1226, 0.0335, -0.0888]], + device='cuda:0'), grad: tensor([[-2.0303e-07, 9.9652e-07, 4.3735e-06, ..., 2.6580e-06, + 5.2564e-06, 1.5087e-07], + [ 2.0489e-08, 1.3346e-06, -8.4102e-05, ..., -1.0163e-04, + -1.2791e-04, 5.3737e-07], + [ 5.4017e-08, -8.7172e-06, 3.9250e-05, ..., 3.1710e-05, + 4.1366e-05, 4.6846e-07], + ..., + [ 1.3970e-08, 2.1011e-06, 6.4299e-06, ..., 9.0525e-06, + 1.5065e-05, -2.5630e-06], + [ 2.9802e-08, 9.3598e-07, 9.7007e-06, ..., 1.0282e-05, + 1.4991e-05, 3.6787e-07], + [ 6.0536e-08, 3.7067e-07, 8.1137e-06, ..., 1.0297e-05, + 9.3132e-06, 5.1968e-07]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0067, -0.0283, 0.0104, -0.0196, 0.0150, 0.0078, 0.0194, -0.0052, + -0.0316, -0.0056], device='cuda:0'), grad: tensor([ 1.9707e-06, -1.6892e-04, 5.2303e-05, 3.8743e-05, 1.5028e-05, + -1.0878e-05, 9.1642e-06, 1.6659e-05, 3.3915e-05, 1.2383e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 262.63, cls_loss 0.0060 cls_loss_mapping 0.0089 cls_loss_causal 0.5642 re_mapping 0.0092 re_causal 0.0263 /// teacc 98.68 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0165, -0.1191, -0.0750, ..., -0.1336, -0.0671, -0.1194], + [ 0.0265, -0.0389, 0.0151, ..., 0.0488, 0.0759, -0.0261], + [-0.0546, 0.0782, -0.1060, ..., 0.0521, 0.0436, -0.0200], + ..., + [-0.0666, -0.0573, -0.0643, ..., -0.0008, -0.0885, 0.1004], + [ 0.0627, -0.0130, 0.0377, ..., -0.0048, -0.1171, -0.0073], + [-0.1096, -0.0168, -0.0137, ..., -0.1219, 0.0343, -0.0885]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, 8.6613e-08, 5.5507e-07, ..., 7.4040e-07, + 8.1882e-06, 4.1910e-08], + [ 8.1956e-08, 3.5390e-07, 8.4098e-07, ..., 2.0880e-06, + 1.4706e-06, 4.6846e-07], + [ 3.0547e-07, -3.4738e-06, 1.6578e-06, ..., -1.9539e-06, + 1.0267e-05, 2.2911e-07], + ..., + [ 2.6077e-07, 3.0082e-07, 2.0713e-06, ..., -1.6885e-06, + 2.8573e-06, -4.3362e-06], + [ 5.5600e-07, 1.8133e-06, 2.8014e-06, ..., 6.3218e-06, + 5.3346e-06, 1.6764e-07], + [ 1.6019e-07, 2.8871e-08, 1.5274e-06, ..., 3.1423e-06, + -1.3828e-05, 1.4901e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0066, -0.0286, 0.0101, -0.0195, 0.0134, 0.0082, 0.0201, -0.0048, + -0.0328, -0.0046], device='cuda:0'), grad: tensor([ 2.7139e-06, 8.0690e-06, 1.7956e-05, -7.5139e-06, -5.3421e-06, + -2.6867e-05, 1.1183e-05, -5.2620e-07, 2.2873e-05, -2.2620e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 262.32, cls_loss 0.0069 cls_loss_mapping 0.0081 cls_loss_causal 0.6039 re_mapping 0.0093 re_causal 0.0267 /// teacc 98.89 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 1.6279e-02, -1.2023e-01, -7.5680e-02, ..., -1.3473e-01, + -6.7798e-02, -1.1952e-01], + [ 2.6654e-02, -3.8665e-02, 1.4047e-02, ..., 4.7737e-02, + 7.5665e-02, -2.6870e-02], + [-5.4584e-02, 7.8155e-02, -1.0695e-01, ..., 5.2231e-02, + 4.3406e-02, -1.9961e-02], + ..., + [-6.6596e-02, -5.7833e-02, -6.3101e-02, ..., 1.1827e-04, + -8.8103e-02, 1.0358e-01], + [ 6.2843e-02, -1.2553e-02, 3.7690e-02, ..., -4.8990e-03, + -1.1811e-01, -7.4039e-03], + [-1.0964e-01, -1.6972e-02, -1.3936e-02, ..., -1.2239e-01, + 3.5119e-02, -8.9203e-02]], device='cuda:0'), grad: tensor([[ 2.0582e-07, 5.2620e-07, 6.0350e-07, ..., 1.0477e-06, + 2.2408e-06, 2.3283e-08], + [-9.2268e-05, -4.4048e-05, -8.0407e-05, ..., -2.1684e-04, + -2.4915e-04, 1.8347e-07], + [ 6.0588e-05, 2.1726e-05, 5.2392e-05, ..., 1.3506e-04, + 1.5140e-04, 3.9861e-07], + ..., + [ 2.5034e-05, 1.2569e-05, 2.3782e-05, ..., 5.8889e-05, + 6.8545e-05, -1.8319e-06], + [ 1.5646e-06, 4.2729e-06, -8.5402e-07, ..., 7.9423e-06, + 1.0446e-05, 3.8557e-07], + [ 2.1141e-07, 6.4541e-07, 1.9316e-06, ..., 2.2035e-06, + 7.2643e-07, 5.4389e-07]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0064, -0.0295, 0.0095, -0.0189, 0.0115, 0.0083, 0.0193, -0.0028, + -0.0328, -0.0043], device='cuda:0'), grad: tensor([ 1.9027e-06, -2.9969e-04, 1.8430e-04, 1.0198e-06, -1.3625e-06, + 1.0036e-05, 4.4554e-06, 8.2016e-05, 1.1921e-05, 5.3495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 262.76, cls_loss 0.0050 cls_loss_mapping 0.0067 cls_loss_causal 0.5725 re_mapping 0.0087 re_causal 0.0267 /// teacc 98.75 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0165, -0.1233, -0.0736, ..., -0.1355, -0.0673, -0.1198], + [ 0.0267, -0.0389, 0.0140, ..., 0.0474, 0.0754, -0.0271], + [-0.0547, 0.0783, -0.1073, ..., 0.0522, 0.0435, -0.0204], + ..., + [-0.0666, -0.0579, -0.0633, ..., -0.0002, -0.0885, 0.1018], + [ 0.0629, -0.0112, 0.0385, ..., -0.0045, -0.1182, -0.0078], + [-0.1098, -0.0173, -0.0144, ..., -0.1215, 0.0357, -0.0867]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 2.0117e-07, ..., 4.5262e-07, + 6.6102e-05, 3.0547e-07], + [ 2.6077e-08, 0.0000e+00, -2.8256e-06, ..., -8.2795e-07, + -3.4925e-06, 2.4457e-06], + [ 1.7695e-08, 9.3132e-10, 7.4785e-07, ..., 4.2021e-06, + 4.7356e-05, 2.5425e-06], + ..., + [ 1.2107e-08, 0.0000e+00, 1.0841e-06, ..., 2.0545e-06, + 2.5585e-05, 4.8578e-05], + [ 1.2107e-08, -9.3132e-10, -4.5542e-07, ..., 4.0885e-07, + 4.6778e-04, 6.5845e-07], + [ 1.0245e-08, 0.0000e+00, 7.1805e-07, ..., -9.9614e-06, + -8.1491e-04, -6.0171e-05]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0057, -0.0300, 0.0092, -0.0186, 0.0113, 0.0082, 0.0190, -0.0038, + -0.0324, -0.0031], device='cuda:0'), grad: tensor([ 1.7643e-04, 5.4538e-06, 1.3351e-04, 2.6032e-05, 1.0848e-04, + 4.0221e-04, 4.3035e-05, 2.1958e-04, 1.2589e-03, -2.3727e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 262.67, cls_loss 0.0041 cls_loss_mapping 0.0084 cls_loss_causal 0.5411 re_mapping 0.0093 re_causal 0.0263 /// teacc 98.86 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 1.5501e-02, -1.2369e-01, -7.4265e-02, ..., -1.3655e-01, + -6.7484e-02, -1.2011e-01], + [ 2.7600e-02, -3.8894e-02, 1.4475e-02, ..., 4.7533e-02, + 7.6454e-02, -2.6830e-02], + [-5.4890e-02, 7.9197e-02, -1.0808e-01, ..., 5.2533e-02, + 4.3578e-02, -2.0176e-02], + ..., + [-6.7356e-02, -5.8163e-02, -6.3525e-02, ..., -6.7993e-05, + -8.9330e-02, 1.0205e-01], + [ 6.4459e-02, -1.2048e-02, 3.9543e-02, ..., -4.5599e-03, + -1.1873e-01, -8.5964e-03], + [-1.1071e-01, -1.7562e-02, -1.5174e-02, ..., -1.2203e-01, + 3.5619e-02, -8.6844e-02]], device='cuda:0'), grad: tensor([[-4.3400e-07, 2.8163e-06, -3.2522e-06, ..., 5.5321e-06, + 1.9018e-06, 3.0641e-07], + [ 4.3176e-06, 7.3425e-06, 3.3882e-06, ..., 2.1324e-05, + 1.3120e-05, 2.2113e-05], + [-1.1474e-04, -1.4007e-04, -1.1456e-04, ..., -4.8542e-04, + -2.1672e-04, 4.4890e-07], + ..., + [ 4.1164e-07, 3.8184e-06, -5.4110e-07, ..., -1.3150e-05, + 7.2867e-06, -2.3261e-05], + [ 1.0735e-04, 1.0407e-04, 1.0937e-04, ..., 4.2748e-04, + 1.6594e-04, 4.6194e-07], + [ 1.0133e-06, 9.9614e-06, 1.4491e-06, ..., 2.4617e-05, + 1.1943e-05, -4.2431e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0064, -0.0293, 0.0091, -0.0190, 0.0118, 0.0079, 0.0194, -0.0039, + -0.0316, -0.0034], device='cuda:0'), grad: tensor([-1.1832e-05, 7.6413e-05, -1.0090e-03, 5.0902e-05, -3.0696e-06, + -1.4491e-05, 1.0915e-05, 2.9802e-06, 8.9979e-04, -1.5302e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 262.56, cls_loss 0.0043 cls_loss_mapping 0.0065 cls_loss_causal 0.5482 re_mapping 0.0088 re_causal 0.0264 /// teacc 98.74 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 1.5297e-02, -1.2388e-01, -7.4985e-02, ..., -1.3730e-01, + -6.7543e-02, -1.2019e-01], + [ 2.7546e-02, -3.8949e-02, 1.4371e-02, ..., 4.7277e-02, + 7.6496e-02, -2.7339e-02], + [-5.3680e-02, 7.9598e-02, -1.0889e-01, ..., 5.2872e-02, + 4.3715e-02, -2.0307e-02], + ..., + [-6.7722e-02, -5.8542e-02, -6.3357e-02, ..., 5.5455e-05, + -8.9460e-02, 1.0228e-01], + [ 6.4626e-02, -1.1799e-02, 4.0416e-02, ..., -4.3290e-03, + -1.1950e-01, -8.8830e-03], + [-1.1105e-01, -1.7790e-02, -1.5079e-02, ..., -1.2232e-01, + 3.4556e-02, -8.7034e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, 1.6671e-07, ..., 1.6484e-07, + 2.6263e-07, 9.3132e-10], + [ 0.0000e+00, 1.4901e-08, -4.8727e-06, ..., -3.4645e-06, + -8.3968e-06, 4.9360e-08], + [ 3.7253e-09, -9.1270e-08, 5.9977e-07, ..., 4.6100e-07, + 2.0675e-07, 3.2596e-08], + ..., + [ 9.3132e-10, 5.9605e-08, 7.0967e-06, ..., 5.5768e-06, + 8.2403e-06, -8.2329e-07], + [ 9.3132e-10, 3.7253e-09, -5.3421e-06, ..., -5.5470e-06, + 1.5832e-06, 2.3283e-08], + [ 9.3132e-10, 9.3132e-10, 2.5630e-06, ..., 3.0380e-06, + 3.8445e-05, 6.2492e-07]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0064, -0.0298, 0.0091, -0.0192, 0.0133, 0.0073, 0.0198, -0.0038, + -0.0312, -0.0043], device='cuda:0'), grad: tensor([ 5.1782e-07, -1.2487e-05, 1.6596e-06, -5.1968e-07, -1.4269e-04, + 5.9698e-07, 6.2678e-07, 2.4781e-05, -1.5661e-05, 1.4329e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 262.86, cls_loss 0.0043 cls_loss_mapping 0.0057 cls_loss_causal 0.5525 re_mapping 0.0090 re_causal 0.0266 /// teacc 98.87 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0153, -0.1240, -0.0753, ..., -0.1385, -0.0678, -0.1209], + [ 0.0276, -0.0393, 0.0146, ..., 0.0469, 0.0764, -0.0279], + [-0.0537, 0.0799, -0.1095, ..., 0.0521, 0.0433, -0.0206], + ..., + [-0.0677, -0.0585, -0.0636, ..., 0.0008, -0.0892, 0.1026], + [ 0.0647, -0.0119, 0.0406, ..., -0.0044, -0.1200, -0.0092], + [-0.1110, -0.0179, -0.0153, ..., -0.1220, 0.0349, -0.0870]], + device='cuda:0'), grad: tensor([[ 8.0094e-08, 6.6124e-08, 1.8254e-07, ..., 1.0049e-06, + 3.5483e-07, 2.1718e-06], + [ 2.8498e-07, 2.3935e-06, -3.2261e-06, ..., 5.8934e-06, + 8.9873e-07, 8.3968e-06], + [ 4.8429e-08, -5.5730e-05, 8.2515e-07, ..., -2.2352e-04, + -1.4579e-04, -6.8665e-05], + ..., + [-1.1194e-06, 5.2840e-05, 1.2992e-06, ..., 1.8287e-04, + 1.4400e-04, -2.3961e-05], + [ 6.5193e-08, 1.0710e-07, 1.3784e-06, ..., 3.1423e-06, + 1.0747e-06, 2.7344e-06], + [ 6.7428e-07, 4.0047e-08, 2.0117e-06, ..., 5.5730e-06, + 4.8615e-07, 9.4101e-06]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0065, -0.0302, 0.0084, -0.0192, 0.0131, 0.0076, 0.0199, -0.0032, + -0.0314, -0.0040], device='cuda:0'), grad: tensor([ 5.1111e-06, 1.1943e-05, -2.8992e-04, 4.1962e-05, 9.3281e-05, + -2.7269e-05, 2.4587e-07, 1.2636e-04, 1.2904e-05, 2.5585e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 262.88, cls_loss 0.0039 cls_loss_mapping 0.0067 cls_loss_causal 0.5704 re_mapping 0.0091 re_causal 0.0260 /// teacc 98.83 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0152, -0.1247, -0.0749, ..., -0.1390, -0.0677, -0.1212], + [ 0.0280, -0.0393, 0.0148, ..., 0.0466, 0.0766, -0.0285], + [-0.0539, 0.0801, -0.1098, ..., 0.0520, 0.0435, -0.0208], + ..., + [-0.0680, -0.0584, -0.0637, ..., 0.0012, -0.0896, 0.1033], + [ 0.0647, -0.0119, 0.0410, ..., -0.0043, -0.1206, -0.0093], + [-0.1112, -0.0180, -0.0158, ..., -0.1227, 0.0349, -0.0874]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 4.7497e-08, 5.0440e-06, ..., 2.3767e-06, + 3.2410e-07, 2.4121e-07], + [ 2.2724e-07, 4.9360e-08, -3.9116e-08, ..., 4.1813e-05, + 1.1876e-05, 2.2203e-05], + [ 6.7987e-08, -3.8333e-06, 2.1886e-06, ..., 2.1338e-04, + 6.9499e-05, 1.1361e-04], + ..., + [ 4.6566e-07, 1.9465e-06, 3.4645e-07, ..., -2.7823e-04, + -7.0512e-05, -1.4997e-04], + [ 1.3039e-08, 7.7300e-08, -2.6613e-05, ..., -1.0528e-05, + 4.0606e-07, 6.8638e-07], + [ 1.1548e-07, 8.1956e-08, 5.9418e-07, ..., 3.1218e-06, + 2.1622e-05, 2.0340e-06]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0063, -0.0305, 0.0081, -0.0191, 0.0131, 0.0072, 0.0201, -0.0027, + -0.0316, -0.0042], device='cuda:0'), grad: tensor([-9.4250e-06, 7.3433e-05, 3.6597e-04, 4.1336e-05, -1.6761e-04, + -7.7039e-06, 2.1219e-05, -4.4012e-04, -5.8353e-05, 1.8096e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 119---------------------------------------------------- +epoch 119, time 279.51, cls_loss 0.0038 cls_loss_mapping 0.0056 cls_loss_causal 0.5365 re_mapping 0.0088 re_causal 0.0250 /// teacc 98.98 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0152, -0.1248, -0.0753, ..., -0.1399, -0.0679, -0.1213], + [ 0.0279, -0.0394, 0.0150, ..., 0.0466, 0.0768, -0.0287], + [-0.0538, 0.0801, -0.1104, ..., 0.0520, 0.0442, -0.0219], + ..., + [-0.0681, -0.0580, -0.0636, ..., 0.0018, -0.0900, 0.1042], + [ 0.0646, -0.0121, 0.0412, ..., -0.0041, -0.1215, -0.0094], + [-0.1113, -0.0182, -0.0162, ..., -0.1234, 0.0349, -0.0877]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 2.7940e-08, 6.2026e-07, ..., 4.7497e-07, + 1.2303e-06, 5.2620e-08], + [ 1.2107e-08, 9.4529e-08, -4.3586e-06, ..., -1.4529e-06, + -6.8694e-06, 1.7555e-07], + [ 1.4901e-08, -8.3260e-07, 7.8045e-07, ..., -2.2678e-07, + -4.5123e-07, 6.4261e-08], + ..., + [ 2.4680e-08, 4.1630e-07, 2.4326e-06, ..., 1.7025e-06, + 6.8592e-07, -1.6661e-06], + [ 2.7707e-07, 7.0781e-08, -9.9614e-06, ..., -2.0750e-06, + 5.2452e-06, 7.0781e-08], + [ 2.4680e-08, 1.5367e-08, 1.4016e-06, ..., 1.6419e-06, + -6.7521e-08, 1.1139e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0065, -0.0305, 0.0081, -0.0200, 0.0138, 0.0074, 0.0201, -0.0020, + -0.0320, -0.0048], device='cuda:0'), grad: tensor([-3.7146e-04, -7.7263e-06, 3.6415e-06, -8.4281e-05, 2.6003e-06, + 1.0717e-04, 6.1154e-05, 4.0159e-06, 9.1968e-07, 2.8396e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 262.94, cls_loss 0.0037 cls_loss_mapping 0.0086 cls_loss_causal 0.5859 re_mapping 0.0085 re_causal 0.0252 /// teacc 98.82 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0152, -0.1253, -0.0756, ..., -0.1406, -0.0681, -0.1215], + [ 0.0278, -0.0395, 0.0155, ..., 0.0466, 0.0770, -0.0290], + [-0.0539, 0.0805, -0.1111, ..., 0.0519, 0.0446, -0.0221], + ..., + [-0.0680, -0.0581, -0.0641, ..., 0.0020, -0.0904, 0.1046], + [ 0.0651, -0.0123, 0.0411, ..., -0.0042, -0.1222, -0.0094], + [-0.1115, -0.0184, -0.0167, ..., -0.1239, 0.0351, -0.0878]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.9232e-07, 4.6194e-06, ..., 6.0126e-06, + 1.0595e-05, 4.1910e-09], + [ 0.0000e+00, -1.4961e-05, -3.9196e-04, ..., -5.1212e-04, + -9.1648e-04, 5.4948e-08], + [ 0.0000e+00, 3.4235e-06, 1.0359e-04, ..., 1.3423e-04, + 2.3806e-04, 1.2573e-08], + ..., + [-0.0000e+00, 1.0379e-05, 2.6608e-04, ..., 3.4738e-04, + 6.1893e-04, -1.8487e-07], + [ 0.0000e+00, 1.2200e-07, 1.2629e-06, ..., 1.6112e-06, + 5.6848e-06, 8.3819e-09], + [ 0.0000e+00, 3.4692e-07, 9.4473e-06, ..., 1.1854e-05, + 2.2918e-05, 7.9162e-08]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0064, -0.0305, 0.0076, -0.0197, 0.0140, 0.0082, 0.0194, -0.0017, + -0.0326, -0.0049], device='cuda:0'), grad: tensor([-5.3257e-05, -2.1286e-03, 5.5838e-04, 1.4320e-05, 3.7134e-05, + -2.5138e-05, 7.0035e-05, 1.4410e-03, 2.0355e-05, 6.7770e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 262.65, cls_loss 0.0039 cls_loss_mapping 0.0061 cls_loss_causal 0.5597 re_mapping 0.0080 re_causal 0.0246 /// teacc 98.89 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0152, -0.1253, -0.0756, ..., -0.1409, -0.0682, -0.1215], + [ 0.0277, -0.0396, 0.0134, ..., 0.0451, 0.0758, -0.0292], + [-0.0539, 0.0809, -0.1117, ..., 0.0519, 0.0448, -0.0223], + ..., + [-0.0680, -0.0578, -0.0627, ..., 0.0030, -0.0888, 0.1045], + [ 0.0651, -0.0125, 0.0425, ..., -0.0029, -0.1226, -0.0073], + [-0.1115, -0.0195, -0.0171, ..., -0.1243, 0.0350, -0.0881]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, -1.5367e-08, 5.8115e-06, ..., 1.7146e-06, + 8.4341e-06, 5.3085e-08], + [ 8.3819e-09, 1.4389e-07, -2.0817e-05, ..., -3.2261e-06, + -2.5943e-05, 2.4419e-06], + [ 1.5832e-08, -3.3788e-06, 3.9861e-06, ..., -3.9674e-06, + 2.1141e-06, -3.7253e-09], + ..., + [ 1.8626e-09, 3.0715e-06, -2.8033e-06, ..., 1.2107e-06, + 8.8140e-06, -6.6385e-06], + [ 1.0710e-08, 7.2643e-08, -7.4878e-07, ..., 1.3337e-06, + 1.4760e-05, 2.2259e-07], + [ 2.7940e-09, 2.3749e-08, -1.8999e-06, ..., -5.8375e-06, + -2.6271e-05, 3.4980e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0064, -0.0325, 0.0075, -0.0198, 0.0138, 0.0079, 0.0196, -0.0001, + -0.0315, -0.0052], device='cuda:0'), grad: tensor([ 1.4268e-05, -3.3289e-05, 6.3777e-06, 1.2860e-05, 8.2105e-06, + 2.7493e-06, 6.4820e-06, -5.3719e-06, 6.9216e-06, -1.9088e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 262.75, cls_loss 0.0042 cls_loss_mapping 0.0079 cls_loss_causal 0.5629 re_mapping 0.0084 re_causal 0.0239 /// teacc 98.83 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0149, -0.1258, -0.0764, ..., -0.1417, -0.0684, -0.1216], + [ 0.0285, -0.0397, 0.0139, ..., 0.0453, 0.0763, -0.0293], + [-0.0562, 0.0811, -0.1113, ..., 0.0534, 0.0445, -0.0199], + ..., + [-0.0681, -0.0580, -0.0628, ..., 0.0030, -0.0893, 0.1048], + [ 0.0653, -0.0146, 0.0412, ..., -0.0054, -0.1231, -0.0098], + [-0.1117, -0.0199, -0.0181, ..., -0.1247, 0.0342, -0.0883]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 8.3167e-07, 2.4606e-06, ..., 3.2634e-06, + 1.8822e-06, 1.0012e-07], + [ 3.7253e-09, 6.2631e-07, 8.0287e-05, ..., 1.0532e-04, + -3.1918e-05, 1.3553e-05], + [ 6.8918e-08, -7.4804e-05, -1.6332e-05, ..., -9.3341e-05, + -4.8548e-05, 3.8045e-07], + ..., + [ 4.2375e-08, 7.5297e-07, -1.1450e-04, ..., -1.4174e-04, + 1.1183e-05, -1.6630e-05], + [ 3.6322e-08, 7.0691e-05, 4.3690e-05, ..., 1.1563e-04, + 6.1691e-05, 1.0310e-06], + [ 2.7940e-09, 1.3690e-07, 3.6377e-06, ..., 3.3863e-06, + -1.8626e-08, 9.1502e-07]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0059, -0.0322, 0.0080, -0.0190, 0.0150, 0.0079, 0.0194, -0.0003, + -0.0330, -0.0064], device='cuda:0'), grad: tensor([ 6.5081e-06, 2.1958e-04, -1.5879e-04, -6.5148e-05, 6.5491e-06, + 7.1764e-05, 3.1590e-06, -2.9874e-04, 2.0647e-04, 8.4043e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 262.75, cls_loss 0.0037 cls_loss_mapping 0.0068 cls_loss_causal 0.5826 re_mapping 0.0083 re_causal 0.0266 /// teacc 98.84 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0149, -0.1260, -0.0769, ..., -0.1426, -0.0686, -0.1218], + [ 0.0286, -0.0398, 0.0151, ..., 0.0460, 0.0769, -0.0295], + [-0.0563, 0.0817, -0.1123, ..., 0.0531, 0.0446, -0.0200], + ..., + [-0.0689, -0.0585, -0.0643, ..., 0.0025, -0.0904, 0.1052], + [ 0.0653, -0.0144, 0.0416, ..., -0.0054, -0.1243, -0.0099], + [-0.1121, -0.0213, -0.0191, ..., -0.1247, 0.0344, -0.0879]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4040e-08, 2.5006e-07, ..., 2.3656e-07, + 8.9034e-07, 9.7789e-09], + [ 4.6566e-10, 6.5006e-07, -1.9651e-07, ..., 9.6578e-07, + 3.7700e-06, 3.4366e-07], + [ 9.3132e-10, -9.9167e-06, 3.9162e-07, ..., -1.9446e-05, + -1.7896e-05, 1.3504e-08], + ..., + [ 1.8626e-09, 5.2620e-07, 1.0822e-06, ..., 1.2470e-06, + 1.3476e-06, -4.6566e-09], + [ 9.3132e-10, 1.3970e-07, -7.0184e-06, ..., -2.3209e-06, + 1.3737e-06, 7.6368e-08], + [ 9.3132e-10, 1.5367e-08, 5.2080e-06, ..., 1.1073e-06, + 7.6108e-06, 2.1085e-06]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0059, -0.0314, 0.0075, -0.0187, 0.0148, 0.0082, 0.0201, -0.0013, + -0.0332, -0.0060], device='cuda:0'), grad: tensor([ 1.2107e-06, 7.5959e-06, -1.9804e-05, 9.3877e-07, -4.6670e-05, + 1.2498e-06, 3.3706e-05, 4.0680e-06, -1.0453e-05, 2.8104e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 262.92, cls_loss 0.0038 cls_loss_mapping 0.0056 cls_loss_causal 0.5601 re_mapping 0.0077 re_causal 0.0240 /// teacc 98.85 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0146, -0.1266, -0.0774, ..., -0.1439, -0.0701, -0.1223], + [ 0.0287, -0.0401, 0.0152, ..., 0.0460, 0.0769, -0.0297], + [-0.0572, 0.0812, -0.1128, ..., 0.0527, 0.0446, -0.0202], + ..., + [-0.0689, -0.0580, -0.0644, ..., 0.0027, -0.0905, 0.1056], + [ 0.0655, -0.0145, 0.0420, ..., -0.0054, -0.1249, -0.0100], + [-0.1123, -0.0220, -0.0196, ..., -0.1252, 0.0357, -0.0881]], + device='cuda:0'), grad: tensor([[ 1.7602e-07, 3.1013e-07, 1.0272e-06, ..., 8.0746e-07, + 2.7306e-06, 5.0291e-08], + [ 2.7940e-07, 2.8070e-06, 3.1322e-05, ..., 2.6003e-05, + 1.4760e-05, 4.4517e-06], + [ 2.9206e-06, 5.3830e-07, 1.7777e-05, ..., 7.3202e-06, + -7.6741e-06, 9.9372e-07], + ..., + [ 1.1586e-06, 5.1968e-07, 6.7204e-06, ..., 4.2394e-06, + 5.6587e-06, -1.0431e-06], + [ 1.0664e-06, -2.9162e-05, -1.0455e-04, ..., -6.7055e-05, + -1.0557e-05, -6.1244e-06], + [ 3.7439e-07, 2.9895e-07, 2.9113e-06, ..., 2.6021e-06, + -1.0490e-05, 1.1967e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0065, -0.0316, 0.0067, -0.0182, 0.0147, 0.0087, 0.0195, -0.0010, + -0.0331, -0.0057], device='cuda:0'), grad: tensor([ 4.6417e-06, 7.8857e-05, 3.5077e-05, -2.8342e-05, 9.5069e-06, + 1.4091e-04, 3.0510e-06, 1.8656e-05, -2.5129e-04, -1.1116e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 262.43, cls_loss 0.0044 cls_loss_mapping 0.0060 cls_loss_causal 0.5895 re_mapping 0.0083 re_causal 0.0249 /// teacc 98.86 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0147, -0.1269, -0.0781, ..., -0.1447, -0.0706, -0.1230], + [ 0.0287, -0.0403, 0.0152, ..., 0.0458, 0.0768, -0.0302], + [-0.0590, 0.0816, -0.1131, ..., 0.0526, 0.0456, -0.0203], + ..., + [-0.0696, -0.0584, -0.0646, ..., 0.0027, -0.0905, 0.1060], + [ 0.0677, -0.0137, 0.0423, ..., -0.0049, -0.1262, -0.0101], + [-0.1131, -0.0238, -0.0202, ..., -0.1262, 0.0352, -0.0884]], + device='cuda:0'), grad: tensor([[-1.1556e-05, 2.7940e-09, -2.8256e-06, ..., 1.4715e-07, + -4.4733e-05, 1.1176e-08], + [ 2.9746e-06, 2.7940e-09, -9.3691e-07, ..., -1.3253e-06, + 8.7246e-06, 3.5390e-08], + [ 1.2480e-07, -3.3528e-08, 1.9073e-06, ..., 1.1642e-06, + 3.1665e-07, 1.7695e-08], + ..., + [ 1.4063e-07, 1.8626e-08, 8.2608e-07, ..., 5.4762e-07, + 1.1306e-06, -7.8231e-08], + [ 6.5193e-07, 2.7940e-09, -1.6447e-06, ..., -6.3330e-07, + 2.9691e-06, 4.3306e-07], + [ 4.6380e-07, 9.3132e-10, 1.4799e-06, ..., 1.1986e-06, + -8.1509e-06, 2.9989e-07]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0067, -0.0319, 0.0065, -0.0190, 0.0142, 0.0098, 0.0201, -0.0009, + -0.0327, -0.0055], device='cuda:0'), grad: tensor([-1.7989e-04, 4.0650e-05, 4.6343e-06, -4.9621e-06, 2.1219e-05, + 1.2629e-05, 1.0788e-04, 3.0641e-06, 8.2180e-06, -1.3731e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 262.54, cls_loss 0.0038 cls_loss_mapping 0.0060 cls_loss_causal 0.5580 re_mapping 0.0085 re_causal 0.0245 /// teacc 98.87 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0146, -0.1274, -0.0781, ..., -0.1462, -0.0706, -0.1231], + [ 0.0299, -0.0406, 0.0151, ..., 0.0459, 0.0769, -0.0303], + [-0.0594, 0.0827, -0.1137, ..., 0.0535, 0.0465, -0.0203], + ..., + [-0.0693, -0.0591, -0.0645, ..., 0.0025, -0.0913, 0.1062], + [ 0.0677, -0.0142, 0.0426, ..., -0.0052, -0.1270, -0.0103], + [-0.1135, -0.0240, -0.0195, ..., -0.1264, 0.0358, -0.0883]], + device='cuda:0'), grad: tensor([[-7.8510e-07, 1.2107e-08, 2.3190e-07, ..., 2.1048e-07, + -7.1805e-07, 8.3819e-09], + [ 7.7300e-08, 4.6566e-08, -4.6164e-05, ..., -2.8953e-05, + -5.2363e-05, 2.8498e-07], + [ 2.0582e-07, -1.0785e-06, 7.4413e-07, ..., -4.3772e-07, + -4.2934e-07, 4.6566e-08], + ..., + [ 3.4459e-08, 5.1688e-07, 4.0263e-05, ..., 2.4840e-05, + 4.6700e-05, -1.2470e-06], + [ 6.9849e-08, 1.5832e-08, -1.5274e-07, ..., 1.8962e-06, + 1.9372e-06, 5.2154e-08], + [ 7.5437e-08, 8.3819e-09, 3.8408e-06, ..., 3.3267e-06, + -5.9232e-06, 4.7404e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0061, -0.0321, 0.0076, -0.0191, 0.0138, 0.0089, 0.0200, -0.0009, + -0.0332, -0.0051], device='cuda:0'), grad: tensor([-9.7156e-06, -9.2208e-05, 3.1292e-06, 1.6928e-05, 2.9340e-05, + -6.8009e-05, 1.9550e-05, 8.2791e-05, 3.7104e-05, -1.8746e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 262.06, cls_loss 0.0039 cls_loss_mapping 0.0057 cls_loss_causal 0.5335 re_mapping 0.0084 re_causal 0.0237 /// teacc 98.91 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0149, -0.1277, -0.0786, ..., -0.1468, -0.0720, -0.1233], + [ 0.0300, -0.0410, 0.0157, ..., 0.0465, 0.0777, -0.0299], + [-0.0596, 0.0832, -0.1142, ..., 0.0536, 0.0467, -0.0205], + ..., + [-0.0694, -0.0592, -0.0650, ..., 0.0020, -0.0922, 0.1061], + [ 0.0673, -0.0144, 0.0430, ..., -0.0055, -0.1275, -0.0104], + [-0.1140, -0.0242, -0.0203, ..., -0.1269, 0.0365, -0.0885]], + device='cuda:0'), grad: tensor([[ 9.1270e-08, 6.5193e-08, 2.1607e-07, ..., 3.0361e-07, + 4.7404e-07, 1.5460e-07], + [ 2.0675e-06, 2.6822e-07, -1.1492e-04, ..., -8.7142e-05, + -1.9991e-04, -3.4243e-05], + [-4.5169e-07, -1.6289e-06, 7.1526e-07, ..., -1.4016e-06, + -1.1874e-06, 5.6718e-07], + ..., + [-8.5682e-06, 7.2736e-07, 5.1767e-05, ..., 3.7521e-05, + 9.1195e-05, 8.8587e-06], + [ 9.7696e-07, 3.4180e-07, 1.7453e-06, ..., 1.7844e-06, + 4.9621e-06, 1.4091e-06], + [ 2.4997e-06, 1.3039e-08, 5.0306e-05, ..., 3.9607e-05, + 8.6784e-05, 1.7703e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0069, -0.0316, 0.0075, -0.0189, 0.0141, 0.0091, 0.0200, -0.0011, + -0.0334, -0.0050], device='cuda:0'), grad: tensor([-5.9128e-05, -3.4523e-04, 4.4890e-07, 7.9200e-06, 3.8326e-05, + 5.3458e-06, 9.0301e-06, 1.3351e-04, 1.6257e-05, 1.9348e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 257.48, cls_loss 0.0024 cls_loss_mapping 0.0046 cls_loss_causal 0.5360 re_mapping 0.0085 re_causal 0.0249 /// teacc 98.93 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0146, -0.1278, -0.0788, ..., -0.1475, -0.0730, -0.1237], + [ 0.0300, -0.0409, 0.0160, ..., 0.0467, 0.0780, -0.0300], + [-0.0597, 0.0834, -0.1144, ..., 0.0540, 0.0470, -0.0203], + ..., + [-0.0691, -0.0593, -0.0652, ..., 0.0018, -0.0926, 0.1063], + [ 0.0677, -0.0146, 0.0435, ..., -0.0056, -0.1279, -0.0106], + [-0.1146, -0.0243, -0.0208, ..., -0.1272, 0.0369, -0.0886]], + device='cuda:0'), grad: tensor([[ 1.2843e-06, 3.4180e-07, 6.9439e-06, ..., 4.0755e-06, + 1.0341e-05, 1.2107e-08], + [-4.4703e-05, 5.4017e-08, -2.3305e-04, ..., -1.1259e-04, + -1.8966e-04, 1.1735e-07], + [ 2.3469e-07, -2.2072e-06, 2.2911e-06, ..., -3.6042e-06, + -3.6750e-06, 1.3970e-08], + ..., + [ 1.5181e-07, 8.6334e-07, 1.1530e-06, ..., 2.3730e-06, + 4.2431e-06, -5.6345e-07], + [ 2.8536e-05, 1.3597e-07, 1.4174e-04, ..., 6.6280e-05, + 1.2565e-04, 1.8626e-08], + [ 1.8347e-07, 7.6368e-08, 6.9812e-06, ..., 4.2915e-06, + 6.7018e-06, 2.9150e-07]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0073, -0.0314, 0.0079, -0.0191, 0.0146, 0.0091, 0.0200, -0.0013, + -0.0334, -0.0051], device='cuda:0'), grad: tensor([ 2.7090e-05, -4.4799e-04, 2.2948e-06, 7.6368e-06, 9.1672e-05, + 1.3137e-04, -1.4877e-04, 8.0764e-06, 2.9230e-04, 3.5822e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 261.35, cls_loss 0.0042 cls_loss_mapping 0.0062 cls_loss_causal 0.5495 re_mapping 0.0081 re_causal 0.0240 /// teacc 98.82 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0142, -0.1284, -0.0798, ..., -0.1485, -0.0731, -0.1240], + [ 0.0301, -0.0414, 0.0162, ..., 0.0467, 0.0779, -0.0306], + [-0.0594, 0.0840, -0.1148, ..., 0.0547, 0.0479, -0.0200], + ..., + [-0.0691, -0.0593, -0.0655, ..., 0.0014, -0.0930, 0.1063], + [ 0.0674, -0.0151, 0.0441, ..., -0.0058, -0.1285, -0.0106], + [-0.1152, -0.0244, -0.0213, ..., -0.1277, 0.0369, -0.0890]], + device='cuda:0'), grad: tensor([[ 7.4040e-07, 4.9360e-08, 7.2680e-06, ..., 3.0417e-06, + 6.2771e-06, 4.0978e-08], + [-2.4904e-06, 2.7940e-08, -2.6003e-05, ..., -1.2584e-05, + -2.3827e-05, -9.0431e-07], + [-1.2647e-06, -5.4501e-06, 3.2987e-06, ..., -1.4871e-05, + -8.9034e-07, -1.6019e-07], + ..., + [ 4.7125e-07, 2.3562e-07, 7.4580e-06, ..., 4.9882e-06, + 6.2063e-06, 1.9651e-07], + [ 2.3656e-06, 5.0701e-06, 2.1309e-05, ..., 2.6047e-05, + 1.0610e-05, 5.6531e-07], + [ 9.6858e-08, 2.7940e-09, 1.6512e-06, ..., 8.5216e-07, + 9.2760e-07, 3.1851e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0076, -0.0316, 0.0082, -0.0189, 0.0144, 0.0094, 0.0203, -0.0012, + -0.0335, -0.0053], device='cuda:0'), grad: tensor([ 1.4633e-05, -4.9263e-05, -1.1452e-05, -1.9327e-05, -2.5425e-07, + -1.7628e-05, -8.9873e-07, 1.5989e-05, 6.3479e-05, 4.6901e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 262.81, cls_loss 0.0033 cls_loss_mapping 0.0065 cls_loss_causal 0.5523 re_mapping 0.0080 re_causal 0.0249 /// teacc 98.85 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0141, -0.1288, -0.0804, ..., -0.1492, -0.0732, -0.1241], + [ 0.0305, -0.0414, 0.0172, ..., 0.0475, 0.0786, -0.0295], + [-0.0596, 0.0844, -0.1157, ..., 0.0546, 0.0482, -0.0203], + ..., + [-0.0691, -0.0594, -0.0663, ..., 0.0009, -0.0938, 0.1062], + [ 0.0676, -0.0152, 0.0446, ..., -0.0057, -0.1292, -0.0107], + [-0.1157, -0.0247, -0.0222, ..., -0.1281, 0.0357, -0.0892]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, 1.1828e-07, 1.7881e-07, ..., 4.5542e-07, + 1.8533e-06, 3.1665e-08], + [ 4.6566e-08, 5.4017e-08, -5.8301e-06, ..., -1.1539e-06, + -4.4294e-06, 9.6764e-07], + [ 2.3283e-08, -1.3541e-06, 2.9150e-07, ..., -1.1139e-05, + -1.4238e-05, 1.0245e-07], + ..., + [ 4.3772e-08, 4.8243e-07, 4.7497e-07, ..., 2.2948e-06, + 8.4490e-06, -3.8035e-06], + [ 9.6858e-08, 7.4506e-08, 3.1758e-06, ..., 1.4389e-06, + 5.5097e-06, 2.9337e-07], + [ 2.1979e-07, 8.8476e-08, -7.9069e-07, ..., 8.7637e-07, + -3.5781e-06, 2.6729e-07]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0079, -0.0308, 0.0080, -0.0192, 0.0156, 0.0091, 0.0209, -0.0018, + -0.0334, -0.0061], device='cuda:0'), grad: tensor([ 3.0436e-06, -6.1132e-06, -2.4229e-05, -2.3365e-05, 7.2159e-06, + 4.0561e-05, -8.2701e-06, 8.3745e-06, 1.1414e-05, -8.7470e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 262.47, cls_loss 0.0056 cls_loss_mapping 0.0095 cls_loss_causal 0.6045 re_mapping 0.0077 re_causal 0.0241 /// teacc 98.84 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0139, -0.1291, -0.0806, ..., -0.1500, -0.0752, -0.1269], + [ 0.0322, -0.0417, 0.0175, ..., 0.0474, 0.0792, -0.0297], + [-0.0595, 0.0851, -0.1162, ..., 0.0551, 0.0497, -0.0202], + ..., + [-0.0697, -0.0599, -0.0665, ..., 0.0010, -0.0950, 0.1067], + [ 0.0676, -0.0153, 0.0461, ..., -0.0054, -0.1294, -0.0105], + [-0.1171, -0.0254, -0.0230, ..., -0.1295, 0.0378, -0.0887]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 1.5181e-07, 4.3586e-07, ..., 5.2433e-07, + -4.1008e-05, 1.7416e-07], + [ 1.5832e-08, 1.8217e-06, 9.7826e-06, ..., 1.1675e-05, + 1.4296e-06, -7.2829e-06], + [ 8.3819e-08, -1.3515e-05, -2.6263e-07, ..., -1.9699e-05, + 1.8328e-05, 3.7998e-07], + ..., + [ 4.8429e-08, 2.9430e-06, 2.1812e-06, ..., 5.7220e-06, + 1.0215e-05, 3.5204e-07], + [ 5.0105e-07, 6.8210e-06, -3.4839e-05, ..., -1.8626e-05, + -2.6785e-06, -3.9302e-07], + [ 1.6205e-07, 2.4959e-07, 2.3656e-06, ..., 2.1383e-06, + 1.0234e-04, 6.0257e-07]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0089, -0.0307, 0.0088, -0.0183, 0.0148, 0.0073, 0.0207, -0.0020, + -0.0330, -0.0045], device='cuda:0'), grad: tensor([-1.1772e-04, 2.8506e-05, 1.3426e-05, 4.5709e-06, -1.8716e-04, + 1.7315e-05, 4.3005e-05, 1.9714e-05, -3.8981e-05, 2.1744e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 262.58, cls_loss 0.0037 cls_loss_mapping 0.0062 cls_loss_causal 0.5709 re_mapping 0.0083 re_causal 0.0248 /// teacc 98.82 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0137, -0.1291, -0.0807, ..., -0.1512, -0.0746, -0.1273], + [ 0.0333, -0.0422, 0.0180, ..., 0.0473, 0.0791, -0.0302], + [-0.0591, 0.0861, -0.1167, ..., 0.0560, 0.0507, -0.0204], + ..., + [-0.0693, -0.0609, -0.0664, ..., 0.0012, -0.0956, 0.1075], + [ 0.0667, -0.0154, 0.0462, ..., -0.0057, -0.1307, -0.0108], + [-0.1174, -0.0256, -0.0239, ..., -0.1303, 0.0372, -0.0889]], + device='cuda:0'), grad: tensor([[ 6.7055e-08, 2.5146e-08, 3.7160e-07, ..., 4.5542e-07, + -3.0920e-07, 2.8871e-08], + [ 9.4716e-07, 1.3970e-08, 4.6007e-07, ..., 8.1360e-06, + 1.8664e-06, 3.4105e-06], + [ 6.2995e-06, -3.3528e-07, 3.2037e-05, ..., 3.4213e-05, + 4.8336e-07, 9.8906e-07], + ..., + [-1.0198e-06, 1.9744e-07, 1.3150e-06, ..., -9.4175e-06, + -8.0839e-07, -5.9679e-06], + [-7.1302e-06, 1.0245e-08, -3.0637e-05, ..., -2.8715e-05, + 9.8906e-07, 3.6415e-07], + [ 1.0524e-07, 2.4214e-08, 3.2131e-07, ..., 1.0198e-06, + 3.8370e-07, 3.5390e-07]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0079, -0.0309, 0.0095, -0.0185, 0.0159, 0.0072, 0.0194, -0.0019, + -0.0336, -0.0053], device='cuda:0'), grad: tensor([-3.9279e-05, 2.1279e-05, 8.9884e-05, -7.1302e-06, 5.3123e-06, + 2.3067e-05, 9.8124e-06, -3.6597e-05, -8.2672e-05, 1.6287e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 262.32, cls_loss 0.0037 cls_loss_mapping 0.0054 cls_loss_causal 0.5429 re_mapping 0.0085 re_causal 0.0246 /// teacc 98.80 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0134, -0.1298, -0.0810, ..., -0.1519, -0.0747, -0.1275], + [ 0.0346, -0.0438, 0.0180, ..., 0.0466, 0.0781, -0.0306], + [-0.0590, 0.0870, -0.1155, ..., 0.0572, 0.0533, -0.0207], + ..., + [-0.0693, -0.0611, -0.0666, ..., 0.0011, -0.0962, 0.1080], + [ 0.0662, -0.0154, 0.0461, ..., -0.0057, -0.1315, -0.0108], + [-0.1178, -0.0257, -0.0243, ..., -0.1311, 0.0371, -0.0891]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 2.7940e-08, 1.7416e-07, ..., 5.2899e-07, + 3.3993e-07, 4.9453e-07], + [ 4.2003e-07, 1.1921e-07, 1.6801e-06, ..., 4.7609e-06, + 4.6529e-06, 4.2170e-06], + [ 5.3737e-07, -4.2561e-07, 2.1011e-06, ..., 1.1541e-05, + 1.2070e-05, 1.1690e-05], + ..., + [ 2.7008e-08, 1.3877e-07, 5.6624e-07, ..., -2.7716e-05, + -3.0205e-05, -3.0637e-05], + [-2.1271e-06, -3.6508e-07, 5.9530e-06, ..., 7.1898e-07, + 3.5614e-06, 2.7008e-06], + [ 2.3469e-07, 2.6077e-08, 2.9169e-06, ..., 5.5246e-06, + 5.0478e-06, 5.2787e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0078, -0.0314, 0.0110, -0.0192, 0.0158, 0.0077, 0.0197, -0.0019, + -0.0341, -0.0054], device='cuda:0'), grad: tensor([ 9.9279e-07, 2.0266e-05, 4.7535e-05, 1.4853e-04, 3.9749e-06, + -2.1172e-04, 1.4491e-05, -1.0723e-04, 5.0455e-05, 3.2634e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 262.72, cls_loss 0.0042 cls_loss_mapping 0.0074 cls_loss_causal 0.5633 re_mapping 0.0084 re_causal 0.0238 /// teacc 98.89 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0135, -0.1301, -0.0817, ..., -0.1526, -0.0746, -0.1276], + [ 0.0350, -0.0440, 0.0190, ..., 0.0473, 0.0790, -0.0299], + [-0.0592, 0.0874, -0.1165, ..., 0.0569, 0.0532, -0.0210], + ..., + [-0.0696, -0.0608, -0.0671, ..., 0.0012, -0.0968, 0.1087], + [ 0.0669, -0.0157, 0.0468, ..., -0.0058, -0.1323, -0.0109], + [-0.1184, -0.0261, -0.0258, ..., -0.1321, 0.0368, -0.0899]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, 2.7008e-08, 5.0291e-07, ..., 1.9930e-07, + 3.9954e-07, 2.0489e-08], + [ 5.9605e-08, 1.0664e-06, -3.7737e-06, ..., -2.6133e-06, + -6.5751e-07, 2.7847e-07], + [ 2.5425e-07, -1.1735e-07, 4.3437e-06, ..., 1.8757e-06, + 3.1032e-06, 5.8673e-08], + ..., + [ 9.3132e-08, 1.0710e-07, 1.5823e-06, ..., -1.0328e-06, + 1.7276e-06, -1.9092e-06], + [-1.1111e-06, 2.5146e-08, -1.1533e-05, ..., -1.9688e-06, + 7.8790e-07, 4.0978e-07], + [ 3.6694e-07, 2.3283e-08, 5.3123e-06, ..., 1.7360e-06, + -9.9745e-07, 6.2957e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0071, -0.0307, 0.0106, -0.0212, 0.0159, 0.0091, 0.0198, -0.0020, + -0.0333, -0.0063], device='cuda:0'), grad: tensor([ 5.6699e-06, 5.6662e-06, 1.0975e-05, 1.7762e-05, 3.4153e-05, + -5.2595e-04, 2.5439e-04, -2.2817e-07, 1.9395e-04, 3.4403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 262.38, cls_loss 0.0032 cls_loss_mapping 0.0053 cls_loss_causal 0.5619 re_mapping 0.0081 re_causal 0.0248 /// teacc 98.97 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0134, -0.1307, -0.0825, ..., -0.1533, -0.0747, -0.1276], + [ 0.0352, -0.0435, 0.0194, ..., 0.0472, 0.0793, -0.0310], + [-0.0593, 0.0874, -0.1177, ..., 0.0562, 0.0527, -0.0213], + ..., + [-0.0696, -0.0609, -0.0671, ..., 0.0019, -0.0964, 0.1097], + [ 0.0669, -0.0155, 0.0485, ..., -0.0054, -0.1319, -0.0110], + [-0.1189, -0.0263, -0.0273, ..., -0.1331, 0.0369, -0.0901]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.2359e-06, 1.0338e-07, ..., 6.3237e-07, + 4.8578e-06, 2.2352e-08], + [ 5.5879e-09, 3.7074e-05, -6.6031e-07, ..., 1.2860e-05, + 1.2779e-04, 3.8650e-07], + [ 1.1269e-07, -4.1313e-06, 5.8953e-07, ..., -5.4687e-05, + -4.4644e-05, -5.9791e-07], + ..., + [ 6.0536e-08, 5.6438e-06, 4.0699e-07, ..., 3.4988e-05, + 3.8952e-05, -8.1025e-08], + [ 1.3970e-08, 1.1921e-06, 2.3935e-07, ..., 3.3919e-06, + 6.2138e-06, 2.2165e-07], + [ 3.7253e-09, 3.7253e-07, -2.7940e-07, ..., 1.9874e-06, + 1.6643e-06, -2.3935e-07]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0071, -0.0307, 0.0096, -0.0207, 0.0160, 0.0087, 0.0189, -0.0013, + -0.0325, -0.0066], device='cuda:0'), grad: tensor([ 5.9679e-06, 2.2578e-04, -8.8811e-05, 4.6846e-07, 4.9584e-06, + 8.8196e-07, -2.4033e-04, 7.4804e-05, 1.4342e-05, 2.4177e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 262.85, cls_loss 0.0032 cls_loss_mapping 0.0053 cls_loss_causal 0.5493 re_mapping 0.0080 re_causal 0.0241 /// teacc 98.86 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0137, -0.1309, -0.0833, ..., -0.1538, -0.0747, -0.1276], + [ 0.0356, -0.0433, 0.0198, ..., 0.0474, 0.0799, -0.0314], + [-0.0597, 0.0878, -0.1185, ..., 0.0558, 0.0525, -0.0215], + ..., + [-0.0696, -0.0613, -0.0671, ..., 0.0023, -0.0968, 0.1103], + [ 0.0670, -0.0155, 0.0485, ..., -0.0054, -0.1327, -0.0111], + [-0.1198, -0.0264, -0.0277, ..., -0.1335, 0.0373, -0.0904]], + device='cuda:0'), grad: tensor([[ 9.6858e-08, 6.0536e-08, 2.2911e-07, ..., 2.2911e-07, + 2.2396e-05, 5.6811e-08], + [ 1.0850e-06, 6.8545e-07, -1.2619e-06, ..., 1.7304e-06, + 1.1157e-06, 9.4622e-07], + [ 1.4760e-05, 8.2552e-06, 6.5658e-07, ..., 3.0443e-05, + 3.7253e-08, 1.1526e-05], + ..., + [-2.9102e-05, -1.7062e-05, 8.9686e-07, ..., -6.0946e-05, + 3.8482e-06, -2.2084e-05], + [ 7.9796e-06, 4.7535e-06, -2.4121e-07, ..., 1.6600e-05, + 2.4978e-06, 6.1691e-06], + [ 1.2387e-07, 3.1665e-08, 2.8312e-07, ..., 2.5984e-07, + 1.7462e-06, 3.7160e-07]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0071, -0.0305, 0.0092, -0.0209, 0.0152, 0.0089, 0.0189, -0.0010, + -0.0329, -0.0060], device='cuda:0'), grad: tensor([ 8.1301e-05, 8.3223e-06, 4.4435e-05, 3.2987e-06, 4.7386e-05, + 1.5244e-05, -1.6594e-04, -7.3433e-05, 3.1978e-05, 7.4282e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 262.28, cls_loss 0.0026 cls_loss_mapping 0.0046 cls_loss_causal 0.5232 re_mapping 0.0080 re_causal 0.0233 /// teacc 98.93 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 0.0137, -0.1315, -0.0843, ..., -0.1546, -0.0747, -0.1277], + [ 0.0357, -0.0431, 0.0203, ..., 0.0474, 0.0802, -0.0320], + [-0.0600, 0.0878, -0.1191, ..., 0.0555, 0.0525, -0.0218], + ..., + [-0.0692, -0.0611, -0.0674, ..., 0.0026, -0.0971, 0.1118], + [ 0.0671, -0.0154, 0.0488, ..., -0.0053, -0.1329, -0.0112], + [-0.1202, -0.0270, -0.0282, ..., -0.1342, 0.0370, -0.0907]], + device='cuda:0'), grad: tensor([[ 2.4900e-05, 9.9652e-08, 5.7742e-08, ..., 2.1141e-07, + 7.2457e-06, 5.5879e-09], + [ 8.5309e-07, 1.3411e-07, -4.7497e-08, ..., 2.7660e-07, + 2.1514e-07, 1.1269e-07], + [ 1.5125e-06, -1.2284e-06, 1.1269e-07, ..., -1.6252e-06, + -2.0824e-06, -8.3819e-09], + ..., + [ 2.6170e-07, 6.0722e-07, 1.1828e-07, ..., 4.7497e-07, + 1.3048e-06, -4.0419e-07], + [ 8.1807e-06, 1.1735e-07, 1.0412e-06, ..., 1.2452e-06, + 2.4103e-06, 2.4214e-08], + [ 1.4724e-06, 2.2352e-08, 2.1979e-07, ..., 3.3714e-07, + 4.1816e-07, 7.0781e-08]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0071, -0.0305, 0.0087, -0.0206, 0.0153, 0.0088, 0.0187, -0.0002, + -0.0328, -0.0065], device='cuda:0'), grad: tensor([ 8.3387e-05, 3.2671e-06, 1.4296e-06, 3.9190e-05, 1.0125e-05, + -2.3991e-05, -1.5378e-04, 2.5034e-06, 3.1650e-05, 6.0536e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 262.23, cls_loss 0.0035 cls_loss_mapping 0.0069 cls_loss_causal 0.5560 re_mapping 0.0080 re_causal 0.0235 /// teacc 98.81 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 0.0133, -0.1317, -0.0849, ..., -0.1555, -0.0754, -0.1289], + [ 0.0355, -0.0436, 0.0212, ..., 0.0481, 0.0807, -0.0311], + [-0.0602, 0.0884, -0.1196, ..., 0.0559, 0.0529, -0.0214], + ..., + [-0.0691, -0.0618, -0.0685, ..., 0.0016, -0.0982, 0.1108], + [ 0.0673, -0.0153, 0.0495, ..., -0.0048, -0.1335, -0.0101], + [-0.1217, -0.0272, -0.0293, ..., -0.1349, 0.0378, -0.0901]], + device='cuda:0'), grad: tensor([[ 5.0943e-07, 3.7253e-09, 7.0818e-06, ..., 2.1942e-06, + 6.1095e-07, 5.5879e-09], + [ 1.2666e-07, 8.3819e-09, 5.7407e-06, ..., 4.2170e-06, + -8.6240e-07, 6.1467e-08], + [ 4.6287e-07, -1.0058e-07, 6.5230e-06, ..., 3.1032e-06, + 1.1642e-07, 5.5879e-08], + ..., + [ 6.6031e-07, 5.5879e-09, 3.0454e-06, ..., 1.3411e-06, + 2.1905e-06, -1.8347e-07], + [ 2.3432e-06, 1.4901e-08, -7.4469e-06, ..., -1.4931e-05, + 9.1456e-07, 6.1467e-08], + [ 2.5965e-06, 0.0000e+00, 4.8041e-05, ..., 1.9431e-05, + 2.4494e-07, -9.6858e-08]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0078, -0.0301, 0.0090, -0.0204, 0.0149, 0.0089, 0.0190, -0.0010, + -0.0327, -0.0061], device='cuda:0'), grad: tensor([ 9.6783e-06, 1.5780e-05, 1.4447e-05, -1.2058e-04, -1.6531e-06, + 2.7806e-05, -4.6566e-07, 1.0915e-05, -4.4912e-05, 8.8930e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 139---------------------------------------------------- +epoch 139, time 279.28, cls_loss 0.0034 cls_loss_mapping 0.0054 cls_loss_causal 0.5502 re_mapping 0.0077 re_causal 0.0229 /// teacc 99.05 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 0.0132, -0.1332, -0.0850, ..., -0.1564, -0.0756, -0.1290], + [ 0.0353, -0.0437, 0.0216, ..., 0.0484, 0.0810, -0.0308], + [-0.0603, 0.0889, -0.1200, ..., 0.0560, 0.0531, -0.0215], + ..., + [-0.0692, -0.0619, -0.0689, ..., 0.0011, -0.0989, 0.1108], + [ 0.0670, -0.0151, 0.0501, ..., -0.0048, -0.1342, -0.0100], + [-0.1227, -0.0277, -0.0302, ..., -0.1356, 0.0370, -0.0905]], + device='cuda:0'), grad: tensor([[-1.6484e-07, -6.4634e-07, 2.1327e-07, ..., 2.3283e-07, + 1.8533e-07, 4.2189e-07], + [ 4.3772e-08, 1.3970e-08, 1.3672e-06, ..., 2.5202e-06, + 3.4459e-07, 2.8703e-06], + [ 4.1910e-08, 9.2201e-08, 3.8743e-07, ..., 5.1688e-07, + 1.6764e-08, 8.0187e-07], + ..., + [ 5.5879e-09, 4.3772e-08, 2.2259e-07, ..., -3.8929e-06, + 2.1048e-07, -2.2829e-05], + [-6.5193e-08, 2.8871e-08, 5.2303e-05, ..., 3.1888e-05, + 1.5691e-05, 3.5577e-07], + [ 9.2201e-08, 3.5577e-07, 7.7579e-07, ..., 1.7453e-06, + -6.1374e-07, 1.4842e-05]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0081, -0.0300, 0.0089, -0.0199, 0.0161, 0.0089, 0.0186, -0.0012, + -0.0323, -0.0070], device='cuda:0'), grad: tensor([-1.4946e-05, 1.1973e-05, 5.5172e-06, 8.5950e-05, 5.6103e-06, + -3.5739e-04, 2.8476e-05, -6.0767e-05, 2.4414e-04, 5.1558e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 262.24, cls_loss 0.0035 cls_loss_mapping 0.0054 cls_loss_causal 0.5304 re_mapping 0.0076 re_causal 0.0227 /// teacc 98.93 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 0.0127, -0.1335, -0.0848, ..., -0.1576, -0.0749, -0.1291], + [ 0.0354, -0.0446, 0.0217, ..., 0.0481, 0.0807, -0.0312], + [-0.0604, 0.0908, -0.1203, ..., 0.0562, 0.0546, -0.0216], + ..., + [-0.0693, -0.0627, -0.0690, ..., 0.0016, -0.0994, 0.1113], + [ 0.0684, -0.0152, 0.0507, ..., -0.0050, -0.1337, -0.0101], + [-0.1231, -0.0282, -0.0306, ..., -0.1364, 0.0370, -0.0908]], + device='cuda:0'), grad: tensor([[ 2.2259e-07, 6.8266e-07, 9.0152e-07, ..., 2.1476e-06, + 9.7416e-07, 7.4971e-07], + [ 2.1700e-07, 6.6962e-07, 1.0617e-05, ..., 1.7762e-05, + 7.0930e-06, 1.0721e-05], + [-3.5077e-05, -1.1516e-04, 3.1650e-05, ..., -9.9838e-05, + -8.8513e-06, 3.1501e-05], + ..., + [ 3.4511e-05, 1.1301e-04, -4.8429e-05, ..., 6.9261e-05, + -1.8021e-06, -4.9382e-05], + [ 5.4948e-08, 1.2480e-07, -8.5216e-07, ..., 2.0582e-07, + 1.5097e-06, 3.1758e-07], + [ 5.0291e-08, 3.5390e-08, 2.3562e-06, ..., 2.4885e-06, + -6.2883e-06, 1.2266e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0070, -0.0307, 0.0095, -0.0200, 0.0158, 0.0089, 0.0166, -0.0008, + -0.0307, -0.0072], device='cuda:0'), grad: tensor([ 5.5879e-06, 4.9204e-05, -1.0353e-04, 1.7226e-05, 1.1444e-05, + 2.5705e-06, -1.8012e-06, 2.4587e-05, 4.0643e-06, -9.9540e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 260.17, cls_loss 0.0027 cls_loss_mapping 0.0052 cls_loss_causal 0.5522 re_mapping 0.0077 re_causal 0.0232 /// teacc 98.95 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 0.0122, -0.1361, -0.0850, ..., -0.1582, -0.0750, -0.1291], + [ 0.0354, -0.0451, 0.0219, ..., 0.0481, 0.0809, -0.0314], + [-0.0604, 0.0916, -0.1209, ..., 0.0564, 0.0548, -0.0216], + ..., + [-0.0695, -0.0631, -0.0691, ..., 0.0016, -0.0997, 0.1118], + [ 0.0685, -0.0155, 0.0510, ..., -0.0049, -0.1347, -0.0101], + [-0.1232, -0.0289, -0.0318, ..., -0.1371, 0.0369, -0.0909]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.8720e-08, 2.7101e-07, ..., 2.8778e-07, + -1.4501e-06, 1.0245e-08], + [ 0.0000e+00, -1.1906e-05, -3.2932e-05, ..., -3.4243e-05, + -5.1677e-05, -2.0768e-07], + [ 1.8626e-09, 1.0833e-05, 2.6390e-05, ..., 2.8923e-05, + 4.1842e-05, 5.2154e-08], + ..., + [ 9.3132e-10, 6.9290e-07, 6.3255e-06, ..., 4.7795e-06, + 8.4713e-06, -2.0675e-07], + [ 9.3132e-10, 7.1712e-08, -1.8878e-06, ..., -1.8599e-06, + 5.1595e-07, 2.3283e-08], + [ 0.0000e+00, 2.2352e-07, 7.7393e-07, ..., 8.6147e-07, + 1.9595e-06, 1.1176e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0072, -0.0307, 0.0094, -0.0198, 0.0150, 0.0085, 0.0182, -0.0001, + -0.0309, -0.0077], device='cuda:0'), grad: tensor([-4.1962e-05, -6.8843e-05, 6.0886e-05, 1.4678e-05, -2.8592e-06, + -1.0870e-05, 1.2055e-05, 1.4715e-05, 6.5938e-07, 2.1502e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 259.68, cls_loss 0.0035 cls_loss_mapping 0.0055 cls_loss_causal 0.5485 re_mapping 0.0074 re_causal 0.0217 /// teacc 98.90 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0122, -0.1363, -0.0856, ..., -0.1591, -0.0753, -0.1293], + [ 0.0354, -0.0465, 0.0222, ..., 0.0481, 0.0808, -0.0316], + [-0.0601, 0.0939, -0.1214, ..., 0.0568, 0.0557, -0.0216], + ..., + [-0.0701, -0.0646, -0.0692, ..., 0.0016, -0.1004, 0.1123], + [ 0.0687, -0.0165, 0.0516, ..., -0.0050, -0.1358, -0.0103], + [-0.1233, -0.0308, -0.0325, ..., -0.1380, 0.0365, -0.0913]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.2399e-07, 2.8312e-07, ..., 8.4378e-07, + 6.5230e-06, 6.3330e-08], + [ 0.0000e+00, 2.6263e-07, -7.4506e-08, ..., 1.4007e-06, + 1.1157e-06, 8.4564e-07], + [ 0.0000e+00, -3.8669e-06, 1.1008e-06, ..., -6.5565e-06, + -2.9188e-06, 4.5821e-07], + ..., + [ 0.0000e+00, 1.7080e-06, 2.1979e-07, ..., 1.1902e-06, + 3.4925e-06, -2.1514e-06], + [ 0.0000e+00, 2.7753e-07, -2.6990e-06, ..., 9.5554e-07, + 2.9430e-05, -2.3283e-07], + [ 0.0000e+00, 1.3784e-07, 5.5879e-07, ..., 5.6997e-07, + -5.1931e-06, 2.5891e-07]], device='cuda:0') +Epoch 144, bias, value: tensor([-6.8900e-03, -3.0989e-02, 1.0346e-02, -1.9950e-02, 1.5448e-02, + 8.5766e-03, 1.8889e-02, -2.1528e-05, -3.1354e-02, -8.7578e-03], + device='cuda:0'), grad: tensor([ 1.0245e-05, 4.6715e-06, -7.0632e-06, 4.2021e-06, 4.9658e-06, + 1.3500e-05, -6.9261e-05, 4.8876e-06, 4.2766e-05, -8.9258e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 259.23, cls_loss 0.0023 cls_loss_mapping 0.0038 cls_loss_causal 0.5416 re_mapping 0.0075 re_causal 0.0230 /// teacc 98.98 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0122, -0.1364, -0.0860, ..., -0.1598, -0.0754, -0.1295], + [ 0.0353, -0.0466, 0.0225, ..., 0.0480, 0.0809, -0.0317], + [-0.0596, 0.0941, -0.1216, ..., 0.0569, 0.0558, -0.0216], + ..., + [-0.0704, -0.0646, -0.0696, ..., 0.0016, -0.1009, 0.1125], + [ 0.0685, -0.0167, 0.0514, ..., -0.0053, -0.1364, -0.0105], + [-0.1234, -0.0309, -0.0329, ..., -0.1381, 0.0366, -0.0913]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, 2.3656e-07, ..., 3.7253e-08, + 3.7439e-07, 9.3132e-09], + [ 2.6077e-08, 0.0000e+00, -7.4506e-07, ..., -1.1548e-07, + -1.2666e-06, 3.1665e-08], + [ 6.8918e-08, 0.0000e+00, 4.1910e-07, ..., 1.3411e-07, + 8.0466e-07, 2.0489e-08], + ..., + [ 2.7940e-08, 0.0000e+00, 3.5018e-07, ..., 7.0781e-08, + 3.5577e-07, -2.0862e-07], + [-1.1120e-06, 0.0000e+00, -6.4448e-06, ..., -2.3376e-06, + 2.0303e-07, 1.6764e-08], + [ 2.0489e-08, 0.0000e+00, 2.5332e-07, ..., 1.0431e-07, + 1.9558e-07, 9.3132e-08]], device='cuda:0') +Epoch 145, bias, value: tensor([-6.8792e-03, -3.1122e-02, 1.0350e-02, -1.9841e-02, 1.5689e-02, + 9.0812e-03, 1.8215e-02, 9.5557e-05, -3.2026e-02, -8.7079e-03], + device='cuda:0'), grad: tensor([ 1.0766e-06, -1.5292e-06, 1.6876e-06, 5.9493e-06, -6.8359e-07, + 3.2205e-06, -1.0468e-06, 4.2841e-07, -9.8124e-06, 6.9290e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 261.48, cls_loss 0.0033 cls_loss_mapping 0.0056 cls_loss_causal 0.5869 re_mapping 0.0074 re_causal 0.0234 /// teacc 98.97 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0122, -0.1366, -0.0868, ..., -0.1602, -0.0756, -0.1296], + [ 0.0355, -0.0467, 0.0240, ..., 0.0486, 0.0819, -0.0297], + [-0.0597, 0.0943, -0.1221, ..., 0.0567, 0.0558, -0.0221], + ..., + [-0.0707, -0.0645, -0.0714, ..., 0.0009, -0.1019, 0.1110], + [ 0.0683, -0.0167, 0.0520, ..., -0.0049, -0.1374, -0.0093], + [-0.1235, -0.0310, -0.0332, ..., -0.1383, 0.0367, -0.0917]], + device='cuda:0'), grad: tensor([[ 4.5821e-07, 1.3001e-06, 3.5390e-08, ..., 2.6282e-06, + 5.3085e-06, 1.8626e-08], + [ 5.7742e-08, 2.1994e-04, 8.7544e-08, ..., 5.4550e-04, + 8.5545e-04, 5.7742e-08], + [ 3.7253e-08, -2.2948e-04, -1.1921e-07, ..., -5.6696e-04, + -8.8787e-04, 8.3819e-08], + ..., + [ 9.3132e-09, 2.7046e-06, 1.0915e-05, ..., 1.8120e-05, + 9.6858e-06, 9.4324e-06], + [ 4.9733e-07, 2.7176e-06, 1.2927e-06, ..., 7.7263e-06, + 1.1042e-05, 1.0952e-06], + [ 2.2352e-08, 1.5087e-07, 1.1362e-07, ..., 3.7439e-07, + 3.9488e-07, 8.9407e-08]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0071, -0.0296, 0.0099, -0.0197, 0.0156, 0.0104, 0.0169, -0.0014, + -0.0320, -0.0085], device='cuda:0'), grad: tensor([ 2.1942e-06, 1.4658e-03, -1.5230e-03, -1.3635e-05, 7.4022e-06, + 2.0061e-06, -2.3693e-06, 3.4213e-05, 2.3752e-05, 3.9339e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 261.94, cls_loss 0.0036 cls_loss_mapping 0.0055 cls_loss_causal 0.5715 re_mapping 0.0072 re_causal 0.0226 /// teacc 98.85 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 0.0117, -0.1369, -0.0873, ..., -0.1610, -0.0758, -0.1297], + [ 0.0358, -0.0473, 0.0228, ..., 0.0475, 0.0818, -0.0312], + [-0.0599, 0.0931, -0.1228, ..., 0.0565, 0.0558, -0.0235], + ..., + [-0.0708, -0.0627, -0.0699, ..., 0.0019, -0.1025, 0.1125], + [ 0.0687, -0.0166, 0.0543, ..., -0.0040, -0.1360, -0.0095], + [-0.1239, -0.0313, -0.0337, ..., -0.1386, 0.0368, -0.0918]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.7486e-07, 5.9605e-08, ..., 9.0711e-07, + 3.4682e-06, -6.6422e-06], + [ 0.0000e+00, 5.8822e-06, 5.9269e-06, ..., 8.6352e-06, + 1.6764e-05, 9.9279e-07], + [ 0.0000e+00, -1.4283e-05, 3.8370e-07, ..., -1.5102e-05, + -2.5719e-05, 5.9046e-07], + ..., + [ 0.0000e+00, 5.4948e-07, 1.0487e-06, ..., -6.7614e-07, + 3.5111e-06, -2.6170e-06], + [ 0.0000e+00, 8.4750e-07, -7.5959e-06, ..., -1.1381e-06, + 3.0026e-06, 7.2643e-07], + [ 0.0000e+00, 2.5518e-07, 1.3150e-06, ..., 1.2554e-06, + 9.8944e-06, 1.0058e-06]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0074, -0.0311, 0.0089, -0.0193, 0.0155, 0.0103, 0.0161, -0.0001, + -0.0302, -0.0086], device='cuda:0'), grad: tensor([-1.5795e-04, 4.2975e-05, -4.0531e-05, 2.9430e-06, -1.0051e-05, + 2.2352e-05, 7.8022e-05, 4.3586e-06, 1.7226e-05, 4.0472e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 262.12, cls_loss 0.0029 cls_loss_mapping 0.0047 cls_loss_causal 0.5570 re_mapping 0.0077 re_causal 0.0225 /// teacc 98.88 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0118, -0.1377, -0.0878, ..., -0.1619, -0.0757, -0.1297], + [ 0.0360, -0.0476, 0.0230, ..., 0.0474, 0.0824, -0.0317], + [-0.0601, 0.0937, -0.1236, ..., 0.0565, 0.0558, -0.0240], + ..., + [-0.0710, -0.0638, -0.0700, ..., 0.0018, -0.1032, 0.1138], + [ 0.0687, -0.0163, 0.0537, ..., -0.0043, -0.1369, -0.0098], + [-0.1240, -0.0316, -0.0337, ..., -0.1389, 0.0370, -0.0926]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.0489e-08, 2.6077e-08, ..., 5.7742e-08, + 3.5278e-06, 2.4214e-08], + [ 1.8626e-09, 2.9616e-07, -1.9576e-06, ..., 2.2352e-07, + -1.5944e-06, 5.1409e-07], + [-6.1467e-08, -7.3761e-07, 2.7940e-08, ..., -9.0711e-07, + -1.1511e-06, 1.0617e-07], + ..., + [ 5.0291e-08, 2.2538e-07, 6.6869e-07, ..., -5.4576e-07, + 9.3132e-07, -1.6205e-06], + [ 4.6566e-08, 1.5460e-07, 6.5193e-08, ..., 4.2282e-07, + 3.0577e-05, 1.8440e-07], + [ 1.8626e-09, 7.4506e-09, 1.2107e-07, ..., 3.6694e-07, + 2.4978e-06, 3.8184e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0074, -0.0311, 0.0087, -0.0191, 0.0152, 0.0110, 0.0167, 0.0001, + -0.0316, -0.0087], device='cuda:0'), grad: tensor([ 7.0073e-06, -1.6186e-06, -1.3709e-06, 2.8741e-06, 3.2093e-06, + 3.4666e-04, -4.2248e-04, 2.3507e-06, 6.1005e-05, 1.8980e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 262.20, cls_loss 0.0035 cls_loss_mapping 0.0057 cls_loss_causal 0.5805 re_mapping 0.0073 re_causal 0.0221 /// teacc 98.96 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0122, -0.1381, -0.0880, ..., -0.1625, -0.0759, -0.1299], + [ 0.0359, -0.0478, 0.0230, ..., 0.0472, 0.0822, -0.0318], + [-0.0602, 0.0935, -0.1239, ..., 0.0568, 0.0564, -0.0243], + ..., + [-0.0710, -0.0638, -0.0702, ..., 0.0020, -0.1037, 0.1142], + [ 0.0688, -0.0156, 0.0542, ..., -0.0039, -0.1369, -0.0100], + [-0.1240, -0.0307, -0.0345, ..., -0.1394, 0.0365, -0.0926]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 4.7684e-07, 6.7055e-08, ..., 7.3016e-07, + 1.5981e-06, 1.1735e-07], + [ 1.8626e-09, 1.2852e-06, -1.0170e-06, ..., 1.3020e-06, + 1.2312e-06, 1.3970e-07], + [-2.4214e-08, -3.0383e-05, 3.8929e-07, ..., -3.6597e-05, + -3.6329e-05, -5.1148e-06], + ..., + [ 1.6764e-08, 1.3225e-06, 3.0920e-07, ..., 2.1365e-06, + 3.9935e-06, -7.2457e-07], + [-2.4214e-08, 2.5123e-05, -5.6066e-07, ..., 2.8700e-05, + 3.1233e-05, 5.1223e-06], + [ 1.8626e-09, 3.5577e-07, 7.0781e-08, ..., 5.8301e-07, + -5.6297e-05, 2.9802e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([-6.8160e-03, -3.1481e-02, 8.5887e-03, -1.8734e-02, 1.6498e-02, + 1.0562e-02, 1.6535e-02, -5.8479e-05, -3.1007e-02, -9.7955e-03], + device='cuda:0'), grad: tensor([ 1.6354e-06, 4.8950e-06, -6.9380e-05, 9.1419e-06, 1.3995e-04, + 2.5742e-06, 3.4124e-06, 1.0177e-05, 6.1989e-05, -1.6427e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 262.13, cls_loss 0.0025 cls_loss_mapping 0.0041 cls_loss_causal 0.5323 re_mapping 0.0075 re_causal 0.0218 /// teacc 98.96 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0119, -0.1379, -0.0884, ..., -0.1635, -0.0759, -0.1299], + [ 0.0362, -0.0480, 0.0233, ..., 0.0476, 0.0824, -0.0317], + [-0.0603, 0.0938, -0.1243, ..., 0.0571, 0.0574, -0.0247], + ..., + [-0.0711, -0.0631, -0.0704, ..., 0.0018, -0.1046, 0.1147], + [ 0.0687, -0.0160, 0.0541, ..., -0.0041, -0.1378, -0.0102], + [-0.1243, -0.0317, -0.0348, ..., -0.1399, 0.0366, -0.0929]], + device='cuda:0'), grad: tensor([[ 1.8865e-05, 2.1998e-06, 8.3074e-06, ..., 1.6615e-05, + 1.3307e-05, 5.5879e-09], + [ 2.9057e-06, 3.3062e-06, 1.6764e-07, ..., 1.8895e-05, + 1.0587e-05, 9.3132e-09], + [-2.2918e-05, -4.2945e-05, -3.3583e-06, ..., -1.9729e-04, + -1.1426e-04, -3.7253e-09], + ..., + [ 3.2913e-06, 1.0200e-05, 1.6391e-06, ..., 3.3170e-05, + 2.0683e-05, 1.3970e-07], + [ 1.5855e-05, 9.6783e-06, 9.4771e-06, ..., 1.4700e-05, + 1.4625e-05, -4.8429e-08], + [ 1.2573e-06, 7.9535e-07, 4.9919e-07, ..., 5.0142e-06, + 9.6485e-07, -2.0117e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0065, -0.0312, 0.0089, -0.0186, 0.0163, 0.0100, 0.0169, -0.0002, + -0.0314, -0.0098], device='cuda:0'), grad: tensor([-6.6233e-04, 3.6657e-05, -6.5327e-05, 2.2352e-04, 1.8060e-04, + 8.6427e-05, -6.5684e-05, 6.7711e-05, 1.6117e-04, 3.8803e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 261.49, cls_loss 0.0028 cls_loss_mapping 0.0043 cls_loss_causal 0.5381 re_mapping 0.0073 re_causal 0.0212 /// teacc 98.93 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0114, -0.1382, -0.0885, ..., -0.1649, -0.0765, -0.1300], + [ 0.0365, -0.0482, 0.0236, ..., 0.0479, 0.0830, -0.0319], + [-0.0605, 0.0946, -0.1249, ..., 0.0571, 0.0575, -0.0245], + ..., + [-0.0713, -0.0640, -0.0705, ..., 0.0017, -0.1054, 0.1150], + [ 0.0687, -0.0157, 0.0542, ..., -0.0040, -0.1385, -0.0102], + [-0.1249, -0.0328, -0.0353, ..., -0.1404, 0.0371, -0.0930]], + device='cuda:0'), grad: tensor([[-1.1921e-07, 3.1665e-08, 2.0489e-08, ..., 2.9989e-07, + -8.0094e-08, 9.3132e-09], + [ 6.7055e-08, 6.5565e-07, -2.9802e-08, ..., 2.1067e-06, + 1.3113e-06, 6.5565e-07], + [-2.4773e-07, -1.0915e-06, 8.3819e-08, ..., -7.7635e-06, + -3.2485e-06, 1.8068e-07], + ..., + [ 2.1048e-07, 2.1420e-07, 6.2399e-07, ..., 3.1572e-06, + 1.3225e-06, -1.0133e-06], + [ 3.3528e-08, 5.5879e-08, -4.3772e-07, ..., 2.8871e-07, + 5.1968e-07, 3.5390e-08], + [ 5.2154e-08, 1.6764e-08, -1.7919e-06, ..., 1.1362e-07, + -7.0781e-08, 4.2841e-08]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0068, -0.0311, 0.0088, -0.0175, 0.0160, 0.0086, 0.0174, -0.0003, + -0.0316, -0.0092], device='cuda:0'), grad: tensor([-1.5255e-06, 3.4403e-06, -1.0356e-05, 2.3264e-06, 3.1710e-05, + 9.7416e-07, -1.4734e-06, 1.4901e-05, 1.2461e-06, -4.1306e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 260.62, cls_loss 0.0036 cls_loss_mapping 0.0053 cls_loss_causal 0.5569 re_mapping 0.0073 re_causal 0.0212 /// teacc 98.93 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0151, -0.1384, -0.0888, ..., -0.1664, -0.0765, -0.1300], + [ 0.0365, -0.0486, 0.0237, ..., 0.0479, 0.0832, -0.0319], + [-0.0613, 0.0953, -0.1252, ..., 0.0572, 0.0576, -0.0247], + ..., + [-0.0714, -0.0638, -0.0716, ..., 0.0006, -0.1060, 0.1136], + [ 0.0686, -0.0156, 0.0552, ..., -0.0024, -0.1391, -0.0086], + [-0.1260, -0.0352, -0.0368, ..., -0.1421, 0.0384, -0.0932]], + device='cuda:0'), grad: tensor([[ 1.0580e-06, 7.4506e-09, 4.2841e-08, ..., 5.7481e-06, + 2.5518e-07, 1.0304e-05], + [ 9.1642e-07, 1.8626e-09, -2.0489e-07, ..., 5.5768e-06, + 1.7136e-07, 9.4548e-06], + [ 7.3388e-07, -5.0291e-08, 7.2643e-08, ..., 6.5938e-06, + 7.8976e-07, 9.0003e-06], + ..., + [-3.9712e-06, 1.1176e-08, 6.1467e-08, ..., -4.2260e-05, + 2.9430e-07, -5.4806e-05], + [ 2.2724e-07, 1.4901e-08, 1.2852e-07, ..., 6.5751e-06, + 1.6466e-06, 6.2324e-06], + [ 4.3586e-07, 0.0000e+00, 9.5926e-07, ..., 2.4829e-06, + 7.3835e-06, 4.3176e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0056, -0.0312, 0.0084, -0.0167, 0.0148, 0.0084, 0.0174, -0.0014, + -0.0306, -0.0088], device='cuda:0'), grad: tensor([ 3.0324e-05, 2.7210e-05, 2.5466e-05, 3.2365e-05, -3.4690e-05, + 1.4119e-06, 2.9989e-07, -1.4138e-04, 1.8120e-05, 4.0859e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 151---------------------------------------------------- +epoch 151, time 276.40, cls_loss 0.0032 cls_loss_mapping 0.0052 cls_loss_causal 0.5551 re_mapping 0.0074 re_causal 0.0213 /// teacc 99.06 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 1.5786e-02, -1.3908e-01, -8.8877e-02, ..., -1.6768e-01, + -7.6487e-02, -1.3008e-01], + [ 3.8513e-02, -4.9525e-02, 2.4310e-02, ..., 4.8156e-02, + 8.2941e-02, -3.2473e-02], + [-6.1386e-02, 9.6836e-02, -1.2500e-01, ..., 5.8504e-02, + 5.9178e-02, -2.4917e-02], + ..., + [-7.1543e-02, -6.3689e-02, -7.2168e-02, ..., -4.9144e-05, + -1.0638e-01, 1.1298e-01], + [ 6.6605e-02, -1.6713e-02, 5.5319e-02, ..., -1.9923e-03, + -1.4126e-01, -7.3466e-03], + [-1.2637e-01, -3.5514e-02, -3.7453e-02, ..., -1.4300e-01, + 3.8315e-02, -9.3554e-02]], device='cuda:0'), grad: tensor([[-4.3027e-07, 3.9116e-08, 1.0803e-07, ..., 2.4959e-07, + 3.6694e-07, 1.0058e-07], + [ 2.2165e-07, 1.4901e-08, -2.6077e-08, ..., 2.2016e-06, + -9.8720e-08, 2.4103e-06], + [ 1.2107e-06, -3.6694e-07, 9.1456e-07, ..., 5.6624e-07, + -2.7753e-07, 1.4529e-07], + ..., + [ 4.2841e-08, 5.4017e-08, -1.6481e-05, ..., -1.5700e-04, + 3.0547e-07, -1.1820e-04], + [-1.4976e-06, 1.7509e-07, -1.0058e-06, ..., 2.0955e-06, + 5.0478e-07, 2.4643e-06], + [ 5.4017e-08, 1.1176e-08, 5.2527e-07, ..., 4.6380e-06, + 1.8440e-07, 4.1686e-06]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0054, -0.0313, 0.0098, -0.0168, 0.0150, 0.0081, 0.0173, -0.0017, + -0.0307, -0.0089], device='cuda:0'), grad: tensor([-1.2740e-06, 1.1489e-05, 4.4554e-06, 4.7851e-04, -2.9430e-07, + 2.2873e-06, -3.4887e-06, -5.1546e-04, 6.6012e-06, 1.7866e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 258.60, cls_loss 0.0030 cls_loss_mapping 0.0050 cls_loss_causal 0.5384 re_mapping 0.0079 re_causal 0.0220 /// teacc 98.96 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0160, -0.1394, -0.0895, ..., -0.1694, -0.0769, -0.1301], + [ 0.0385, -0.0487, 0.0246, ..., 0.0485, 0.0843, -0.0339], + [-0.0613, 0.0975, -0.1268, ..., 0.0576, 0.0584, -0.0253], + ..., + [-0.0717, -0.0635, -0.0714, ..., 0.0019, -0.1066, 0.1146], + [ 0.0645, -0.0183, 0.0538, ..., -0.0035, -0.1420, -0.0077], + [-0.1266, -0.0353, -0.0379, ..., -0.1435, 0.0380, -0.0937]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 2.7940e-08, 1.7695e-07, ..., 3.7998e-07, + 3.8557e-07, 2.2352e-08], + [ 1.6764e-08, 3.1665e-08, -2.5518e-07, ..., 5.5879e-08, + -4.5821e-07, 4.4703e-08], + [ 3.0547e-07, -1.8068e-07, 2.4792e-06, ..., 5.3458e-06, + 5.7742e-08, 9.3132e-09], + ..., + [ 2.2352e-08, 6.7055e-08, 3.3341e-07, ..., 4.4331e-07, + 1.7509e-07, -2.9244e-07], + [ 6.3330e-08, 2.4214e-08, 8.9779e-07, ..., 1.2536e-06, + 1.3206e-06, 3.1665e-08], + [ 5.5879e-09, 1.8626e-09, 1.1921e-07, ..., 1.3411e-07, + -2.9802e-07, 1.0431e-07]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0056, -0.0312, 0.0090, -0.0164, 0.0154, 0.0080, 0.0172, -0.0003, + -0.0320, -0.0092], device='cuda:0'), grad: tensor([ 1.4473e-06, 2.2724e-07, 1.1653e-05, -8.5175e-05, 7.7300e-07, + 6.9678e-05, -4.2543e-06, 3.7439e-07, 6.3255e-06, -1.1344e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 259.12, cls_loss 0.0033 cls_loss_mapping 0.0061 cls_loss_causal 0.5324 re_mapping 0.0072 re_causal 0.0211 /// teacc 98.94 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 0.0157, -0.1398, -0.0901, ..., -0.1711, -0.0770, -0.1302], + [ 0.0384, -0.0489, 0.0249, ..., 0.0491, 0.0850, -0.0348], + [-0.0616, 0.0982, -0.1279, ..., 0.0570, 0.0581, -0.0255], + ..., + [-0.0717, -0.0640, -0.0713, ..., 0.0022, -0.1074, 0.1153], + [ 0.0648, -0.0184, 0.0537, ..., -0.0037, -0.1425, -0.0077], + [-0.1270, -0.0349, -0.0388, ..., -0.1440, 0.0382, -0.0949]], + device='cuda:0'), grad: tensor([[ 4.6752e-07, 3.7812e-07, 9.6858e-08, ..., 7.6555e-07, + 9.3132e-09, 1.8626e-09], + [ 3.2820e-06, 1.0971e-06, -3.1106e-07, ..., 2.7306e-06, + -9.2387e-07, 4.8429e-08], + [ 6.3255e-06, 8.0466e-07, 7.8045e-07, ..., 3.9749e-06, + -2.1607e-06, 9.3132e-09], + ..., + [ 4.3586e-07, 4.7870e-07, 2.5146e-07, ..., 6.0722e-07, + 1.2442e-06, -3.3528e-07], + [-2.4334e-05, -7.6815e-06, -2.6766e-06, ..., -2.1532e-05, + 4.2841e-07, 1.6950e-07], + [ 1.4715e-07, 2.8126e-07, 2.2538e-07, ..., 6.1095e-07, + 3.7812e-07, 1.3039e-08]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0058, -0.0313, 0.0086, -0.0166, 0.0154, 0.0083, 0.0173, 0.0013, + -0.0323, -0.0104], device='cuda:0'), grad: tensor([ 1.3430e-06, 1.4499e-05, 2.6450e-05, 7.6443e-06, 4.1053e-06, + 3.0816e-05, 1.9863e-05, 4.6007e-06, -1.1301e-04, 3.8818e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 259.47, cls_loss 0.0032 cls_loss_mapping 0.0045 cls_loss_causal 0.5273 re_mapping 0.0074 re_causal 0.0212 /// teacc 98.96 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 0.0158, -0.1404, -0.0892, ..., -0.1712, -0.0766, -0.1302], + [ 0.0384, -0.0494, 0.0257, ..., 0.0492, 0.0851, -0.0350], + [-0.0617, 0.1002, -0.1281, ..., 0.0578, 0.0586, -0.0254], + ..., + [-0.0718, -0.0645, -0.0720, ..., 0.0018, -0.1085, 0.1153], + [ 0.0643, -0.0204, 0.0534, ..., -0.0045, -0.1428, -0.0077], + [-0.1277, -0.0357, -0.0395, ..., -0.1448, 0.0382, -0.0944]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 5.5879e-09, 1.5274e-07, ..., 5.7742e-08, + 1.1250e-06, 5.2154e-08], + [ 7.6368e-08, 7.4506e-09, 2.3097e-07, ..., 1.3784e-07, + 1.3673e-04, 3.1665e-07], + [ 5.6997e-07, -6.7055e-08, 2.9411e-06, ..., 7.4133e-07, + 4.5002e-06, 6.1467e-08], + ..., + [ 3.0175e-07, 3.3528e-08, 1.5479e-06, ..., 3.3528e-08, + 8.0824e-05, -9.6112e-07], + [ 4.8801e-07, 9.3132e-09, 2.3991e-06, ..., 5.8487e-07, + 4.3809e-06, 7.0781e-08], + [ 1.1921e-07, 1.8626e-09, 6.0536e-07, ..., 2.3842e-07, + -2.4366e-04, 4.0606e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0050, -0.0311, 0.0091, -0.0165, 0.0155, 0.0087, 0.0167, 0.0008, + -0.0328, -0.0103], device='cuda:0'), grad: tensor([ 3.5409e-06, 2.8348e-04, 1.5087e-05, -9.9391e-06, 2.5034e-05, + 3.9116e-06, 1.3039e-07, 1.4687e-04, 1.4730e-05, -4.8327e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 259.40, cls_loss 0.0030 cls_loss_mapping 0.0035 cls_loss_causal 0.5644 re_mapping 0.0067 re_causal 0.0212 /// teacc 98.97 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 0.0158, -0.1405, -0.0896, ..., -0.1734, -0.0769, -0.1302], + [ 0.0383, -0.0501, 0.0254, ..., 0.0484, 0.0844, -0.0351], + [-0.0618, 0.1009, -0.1277, ..., 0.0587, 0.0596, -0.0256], + ..., + [-0.0719, -0.0653, -0.0722, ..., 0.0020, -0.1095, 0.1154], + [ 0.0642, -0.0201, 0.0538, ..., -0.0042, -0.1430, -0.0078], + [-0.1279, -0.0358, -0.0401, ..., -0.1454, 0.0384, -0.0945]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 1.8626e-09, -1.2666e-07, ..., 5.2154e-08, + 1.0803e-07, 1.1176e-08], + [ 1.4529e-07, 1.8626e-09, 5.6252e-07, ..., 5.4762e-07, + 2.2538e-07, 1.5087e-07], + [ 2.7940e-07, -1.4901e-08, 1.2051e-06, ..., 8.1398e-07, + 7.7672e-07, 5.9605e-08], + ..., + [ 1.5832e-07, 7.4506e-09, 8.9966e-07, ..., 4.4703e-08, + 5.3644e-07, -5.1782e-07], + [ 3.7253e-08, 0.0000e+00, -3.4571e-06, ..., -2.7996e-06, + 3.4776e-06, 3.1665e-08], + [ 2.0489e-08, 0.0000e+00, 2.0433e-06, ..., 1.7043e-06, + 2.9989e-07, 9.8720e-08]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0052, -0.0321, 0.0099, -0.0167, 0.0153, 0.0079, 0.0182, 0.0009, + -0.0325, -0.0102], device='cuda:0'), grad: tensor([-3.5297e-06, 2.2482e-06, 3.7495e-06, -2.0489e-07, 6.1691e-06, + 3.0786e-05, -4.9949e-05, 1.6596e-06, 5.9605e-08, 8.9854e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 259.57, cls_loss 0.0028 cls_loss_mapping 0.0048 cls_loss_causal 0.5464 re_mapping 0.0070 re_causal 0.0205 /// teacc 98.99 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 0.0159, -0.1407, -0.0903, ..., -0.1743, -0.0769, -0.1303], + [ 0.0383, -0.0506, 0.0243, ..., 0.0472, 0.0841, -0.0359], + [-0.0617, 0.1013, -0.1280, ..., 0.0588, 0.0599, -0.0257], + ..., + [-0.0725, -0.0656, -0.0712, ..., 0.0034, -0.1095, 0.1159], + [ 0.0641, -0.0198, 0.0542, ..., -0.0040, -0.1427, -0.0078], + [-0.1281, -0.0358, -0.0406, ..., -0.1468, 0.0387, -0.0950]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 3.7253e-09, 2.6077e-08, ..., 3.5390e-08, + 2.1793e-07, 9.3132e-09], + [ 5.5879e-09, 2.9244e-07, 1.4156e-07, ..., 1.7453e-06, + 6.6496e-07, 6.2585e-07], + [ 5.5879e-09, -4.4145e-07, 2.8685e-07, ..., -7.5437e-07, + -1.3132e-06, 3.7439e-07], + ..., + [ 2.6077e-08, 8.3819e-08, -7.0222e-07, ..., -3.7309e-06, + 4.8056e-07, -3.0510e-06], + [ 4.2841e-08, 2.6077e-08, -4.0978e-08, ..., 2.6077e-07, + 7.5810e-07, 1.6764e-07], + [ 3.7253e-09, 3.7253e-09, 1.5274e-07, ..., 3.7812e-07, + -2.7940e-08, 2.7381e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0050, -0.0333, 0.0099, -0.0169, 0.0153, 0.0081, 0.0179, 0.0016, + -0.0321, -0.0104], device='cuda:0'), grad: tensor([-5.1782e-07, 3.5986e-06, 3.1292e-07, 4.4107e-06, 2.7195e-07, + 1.0543e-06, -3.7309e-06, -9.9689e-06, 3.2745e-06, 1.2629e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 259.83, cls_loss 0.0024 cls_loss_mapping 0.0036 cls_loss_causal 0.5456 re_mapping 0.0071 re_causal 0.0216 /// teacc 98.90 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 0.0163, -0.1408, -0.0910, ..., -0.1752, -0.0769, -0.1303], + [ 0.0383, -0.0508, 0.0246, ..., 0.0473, 0.0845, -0.0363], + [-0.0616, 0.1016, -0.1284, ..., 0.0588, 0.0602, -0.0260], + ..., + [-0.0726, -0.0657, -0.0712, ..., 0.0030, -0.1111, 0.1155], + [ 0.0641, -0.0198, 0.0533, ..., -0.0046, -0.1436, -0.0078], + [-0.1282, -0.0362, -0.0413, ..., -0.1452, 0.0389, -0.0929]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3877e-07, 2.6077e-07, ..., 2.1234e-07, + 5.0142e-06, 3.7253e-09], + [ 3.7253e-09, 5.4389e-07, 2.1979e-07, ..., 1.3374e-06, + 1.0058e-06, 2.2352e-08], + [ 0.0000e+00, -6.2212e-07, 1.0431e-06, ..., -3.5763e-07, + -6.0350e-07, 1.4901e-08], + ..., + [ 3.7253e-09, 1.7509e-07, 5.5134e-07, ..., 6.7800e-07, + 3.8259e-06, -1.4901e-07], + [-1.8626e-08, 3.2037e-07, -2.0973e-06, ..., -1.6093e-06, + 1.1250e-06, 1.8626e-08], + [ 3.7253e-09, 3.7253e-08, 7.9721e-07, ..., 5.2527e-07, + -8.4862e-06, 8.1956e-08]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0046, -0.0332, 0.0098, -0.0167, 0.0153, 0.0083, 0.0182, 0.0007, + -0.0329, -0.0096], device='cuda:0'), grad: tensor([ 1.9446e-05, 4.8518e-05, 4.1313e-06, 2.2471e-05, 4.0054e-05, + -1.2502e-05, -3.4541e-05, -8.0585e-04, 6.6943e-06, 7.1192e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 259.74, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5588 re_mapping 0.0070 re_causal 0.0216 /// teacc 98.98 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0162, -0.1408, -0.0913, ..., -0.1771, -0.0770, -0.1303], + [ 0.0383, -0.0512, 0.0249, ..., 0.0472, 0.0844, -0.0364], + [-0.0616, 0.1022, -0.1288, ..., 0.0593, 0.0607, -0.0262], + ..., + [-0.0725, -0.0660, -0.0713, ..., 0.0030, -0.1118, 0.1156], + [ 0.0640, -0.0202, 0.0533, ..., -0.0050, -0.1442, -0.0079], + [-0.1283, -0.0363, -0.0418, ..., -0.1452, 0.0390, -0.0925]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.4820e-07, 3.0547e-07, ..., 1.4938e-06, + 1.5646e-06, 1.0803e-07], + [ 0.0000e+00, 4.5076e-07, -1.8999e-05, ..., -1.4119e-05, + -1.9029e-05, 8.0466e-07], + [ 0.0000e+00, -5.3525e-05, 1.7397e-06, ..., -2.6360e-05, + -4.9531e-05, 2.9191e-05], + ..., + [ 0.0000e+00, 4.8056e-07, 1.1623e-05, ..., -7.1108e-05, + 1.2904e-05, -3.2932e-05], + [ 3.7253e-09, 5.0247e-05, 4.6790e-06, ..., 1.0079e-04, + 5.2989e-05, 9.5740e-07], + [ 0.0000e+00, 2.2352e-07, 3.1814e-06, ..., 2.7865e-06, + 2.3730e-06, 4.0233e-07]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0042, -0.0332, 0.0101, -0.0169, 0.0153, 0.0085, 0.0182, 0.0004, + -0.0334, -0.0092], device='cuda:0'), grad: tensor([ 4.2394e-06, -4.0442e-05, -4.6432e-05, 1.2457e-05, 1.1474e-06, + -2.4199e-05, -5.9158e-06, -8.5235e-05, 1.6856e-04, 1.5602e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 259.59, cls_loss 0.0026 cls_loss_mapping 0.0040 cls_loss_causal 0.5550 re_mapping 0.0067 re_causal 0.0209 /// teacc 98.89 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0163, -0.1410, -0.0917, ..., -0.1788, -0.0792, -0.1304], + [ 0.0384, -0.0527, 0.0261, ..., 0.0473, 0.0843, -0.0365], + [-0.0618, 0.1030, -0.1293, ..., 0.0597, 0.0618, -0.0267], + ..., + [-0.0720, -0.0662, -0.0715, ..., 0.0031, -0.1125, 0.1158], + [ 0.0639, -0.0205, 0.0525, ..., -0.0056, -0.1461, -0.0080], + [-0.1285, -0.0362, -0.0423, ..., -0.1459, 0.0402, -0.0927]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-08, ..., 1.8626e-08, + -1.1027e-06, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -1.1548e-07, ..., -6.3330e-08, + 2.0117e-07, 4.0978e-08], + [ 0.0000e+00, -7.4506e-09, 1.3784e-07, ..., 7.0781e-08, + 2.8685e-07, 1.1176e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6019e-07, ..., 2.9802e-08, + 6.6683e-07, -1.1176e-07], + [ 0.0000e+00, 0.0000e+00, -1.6019e-07, ..., -1.1176e-07, + 1.4603e-06, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 2.8685e-07, ..., 1.1176e-07, + -2.2613e-06, 2.2352e-08]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0056, -0.0330, 0.0105, -0.0167, 0.0152, 0.0082, 0.0182, 0.0005, + -0.0340, -0.0084], device='cuda:0'), grad: tensor([ 5.2750e-05, 2.5705e-06, 3.2075e-06, 2.1458e-05, 5.9493e-06, + 7.6666e-06, -1.1367e-04, 3.8408e-06, 2.2113e-05, -5.8413e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 259.55, cls_loss 0.0030 cls_loss_mapping 0.0051 cls_loss_causal 0.5415 re_mapping 0.0071 re_causal 0.0208 /// teacc 98.91 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0161, -0.1412, -0.0925, ..., -0.1802, -0.0794, -0.1304], + [ 0.0384, -0.0528, 0.0270, ..., 0.0485, 0.0854, -0.0355], + [-0.0624, 0.1034, -0.1302, ..., 0.0593, 0.0622, -0.0288], + ..., + [-0.0720, -0.0665, -0.0721, ..., 0.0024, -0.1149, 0.1159], + [ 0.0639, -0.0205, 0.0528, ..., -0.0054, -0.1464, -0.0080], + [-0.1287, -0.0363, -0.0429, ..., -0.1465, 0.0409, -0.0928]], + device='cuda:0'), grad: tensor([[-7.4506e-09, 3.2410e-07, 9.9465e-07, ..., 1.2517e-06, + 1.4268e-06, 2.6077e-08], + [ 0.0000e+00, 2.6077e-08, -7.6532e-05, ..., -7.6950e-05, + -1.8024e-04, 1.4156e-07], + [ 0.0000e+00, -1.7099e-06, 7.4983e-05, ..., 7.3493e-05, + 1.7118e-04, 7.8231e-08], + ..., + [ 0.0000e+00, 4.5449e-07, 2.1420e-06, ..., 2.0526e-06, + 3.3379e-06, -6.3702e-07], + [ 0.0000e+00, 1.6019e-07, -2.2836e-06, ..., -1.8068e-06, + 1.6205e-06, 7.4506e-08], + [ 3.7253e-09, 3.7253e-08, 4.9248e-06, ..., 2.8722e-06, + 8.5682e-08, 1.3039e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([-6.1018e-03, -3.1724e-02, 9.8041e-03, -1.7036e-02, 1.3636e-02, + 8.5496e-03, 1.9008e-02, -8.1377e-05, -3.4139e-02, -7.2731e-03], + device='cuda:0'), grad: tensor([-1.8537e-05, -2.3723e-04, 2.2566e-04, -1.6987e-05, 1.0729e-06, + 4.2543e-06, 2.1338e-05, 6.3628e-06, 1.7621e-06, 1.2405e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 259.22, cls_loss 0.0024 cls_loss_mapping 0.0046 cls_loss_causal 0.5087 re_mapping 0.0067 re_causal 0.0202 /// teacc 98.87 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0160, -0.1414, -0.0940, ..., -0.1823, -0.0801, -0.1305], + [ 0.0385, -0.0524, 0.0276, ..., 0.0490, 0.0865, -0.0357], + [-0.0627, 0.1046, -0.1312, ..., 0.0606, 0.0626, -0.0289], + ..., + [-0.0722, -0.0681, -0.0721, ..., 0.0010, -0.1168, 0.1160], + [ 0.0642, -0.0210, 0.0531, ..., -0.0053, -0.1480, -0.0081], + [-0.1290, -0.0364, -0.0437, ..., -0.1470, 0.0411, -0.0928]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-08, 4.0978e-08, ..., 2.7567e-07, + 1.7881e-07, 0.0000e+00], + [ 0.0000e+00, 2.6822e-07, -1.2740e-06, ..., 7.3388e-07, + -9.6485e-07, 1.4901e-08], + [ 0.0000e+00, 2.3097e-06, 1.0580e-06, ..., 1.5043e-05, + 6.6906e-06, 2.9802e-08], + ..., + [ 0.0000e+00, -1.7628e-05, 2.1607e-07, ..., -1.0777e-04, + -4.0919e-05, -5.9605e-08], + [ 0.0000e+00, 1.3411e-07, 1.0431e-07, ..., 9.8348e-07, + 1.2405e-06, 3.7253e-09], + [ 0.0000e+00, 1.3165e-05, 7.8231e-08, ..., 8.0764e-05, + 2.9743e-05, 3.7253e-09]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0071, -0.0313, 0.0108, -0.0172, 0.0136, 0.0084, 0.0198, -0.0010, + -0.0344, -0.0069], device='cuda:0'), grad: tensor([ 8.8662e-07, 2.0824e-06, 3.2604e-05, 3.6359e-06, 2.0325e-05, + -5.3011e-06, 4.9546e-07, -2.3234e-04, 4.7497e-06, 1.7250e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 259.38, cls_loss 0.0029 cls_loss_mapping 0.0034 cls_loss_causal 0.5456 re_mapping 0.0069 re_causal 0.0198 /// teacc 98.93 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0160, -0.1416, -0.0962, ..., -0.1848, -0.0805, -0.1305], + [ 0.0385, -0.0524, 0.0281, ..., 0.0495, 0.0872, -0.0359], + [-0.0628, 0.1053, -0.1322, ..., 0.0610, 0.0626, -0.0291], + ..., + [-0.0720, -0.0689, -0.0723, ..., 0.0003, -0.1176, 0.1163], + [ 0.0642, -0.0213, 0.0537, ..., -0.0050, -0.1487, -0.0081], + [-0.1292, -0.0366, -0.0451, ..., -0.1481, 0.0414, -0.0930]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 7.0781e-08, 1.1176e-06, ..., 3.7625e-07, + 1.3001e-06, 0.0000e+00], + [ 3.7253e-09, 3.7253e-09, 1.6764e-07, ..., 1.4901e-07, + 4.2841e-07, 2.2352e-08], + [ 3.7253e-09, -1.8254e-07, 7.6741e-07, ..., 3.4273e-07, + 1.1176e-08, 9.3132e-08], + ..., + [ 0.0000e+00, 1.8626e-08, 3.3900e-07, ..., -3.0547e-07, + 3.4645e-07, -1.6391e-07], + [ 5.9605e-08, 1.1176e-08, 7.7188e-06, ..., 7.2643e-07, + 6.1542e-06, 1.4901e-08], + [ 0.0000e+00, 7.4506e-09, -1.0908e-05, ..., 1.4529e-07, + -8.1024e-03, 7.4506e-09]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0073, -0.0312, 0.0110, -0.0172, 0.0134, 0.0088, 0.0196, -0.0011, + -0.0343, -0.0069], device='cuda:0'), grad: tensor([ 6.7614e-06, 1.4380e-06, 2.9355e-06, 4.3511e-06, 1.3863e-02, + 2.2799e-06, -2.1942e-06, 3.6880e-07, 4.6223e-05, -1.3924e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 259.35, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5339 re_mapping 0.0066 re_causal 0.0204 /// teacc 98.90 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0158, -0.1417, -0.0970, ..., -0.1860, -0.0814, -0.1306], + [ 0.0384, -0.0526, 0.0283, ..., 0.0492, 0.0874, -0.0360], + [-0.0632, 0.1058, -0.1329, ..., 0.0608, 0.0627, -0.0292], + ..., + [-0.0719, -0.0691, -0.0720, ..., 0.0010, -0.1176, 0.1164], + [ 0.0646, -0.0215, 0.0538, ..., -0.0049, -0.1494, -0.0081], + [-0.1294, -0.0372, -0.0461, ..., -0.1492, 0.0415, -0.0932]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.0978e-08, 6.7055e-07, ..., 6.4075e-07, + 5.6997e-07, 0.0000e+00], + [ 0.0000e+00, 1.0803e-07, -6.0350e-07, ..., -7.4506e-08, + 1.1921e-07, 2.9802e-08], + [ 0.0000e+00, -2.5369e-06, 1.0744e-05, ..., 9.4771e-06, + -1.7993e-06, 4.0978e-08], + ..., + [ 0.0000e+00, 1.9409e-06, 2.7008e-06, ..., 3.6806e-06, + 3.3714e-06, -1.6019e-07], + [ 0.0000e+00, 2.1234e-07, 4.3549e-06, ..., 5.1670e-06, + 2.1495e-06, 3.7253e-08], + [ 0.0000e+00, 3.7253e-08, 1.0729e-06, ..., 7.2271e-07, + 6.4708e-06, 2.2352e-08]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0078, -0.0315, 0.0107, -0.0175, 0.0131, 0.0086, 0.0203, -0.0004, + -0.0342, -0.0069], device='cuda:0'), grad: tensor([ 6.5677e-06, 1.5944e-06, 2.3916e-05, -5.3346e-05, -2.5988e-05, + 5.4277e-06, -1.2755e-05, 1.3381e-05, 1.7464e-05, 2.3723e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 259.44, cls_loss 0.0024 cls_loss_mapping 0.0039 cls_loss_causal 0.5173 re_mapping 0.0070 re_causal 0.0209 /// teacc 98.87 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0168, -0.1418, -0.0975, ..., -0.1867, -0.0814, -0.1306], + [ 0.0384, -0.0526, 0.0291, ..., 0.0494, 0.0874, -0.0360], + [-0.0635, 0.1059, -0.1335, ..., 0.0608, 0.0631, -0.0296], + ..., + [-0.0719, -0.0691, -0.0723, ..., 0.0010, -0.1178, 0.1167], + [ 0.0646, -0.0216, 0.0538, ..., -0.0048, -0.1501, -0.0081], + [-0.1297, -0.0374, -0.0471, ..., -0.1504, 0.0414, -0.0935]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 1.4156e-07, 7.8231e-08, ..., 2.1979e-07, + 2.3842e-07, 2.9802e-08], + [ 7.4506e-09, 7.4506e-09, 1.3001e-06, ..., 4.5113e-06, + -1.6391e-07, 7.1265e-06], + [-1.8626e-08, -2.6450e-07, 1.1548e-07, ..., -1.1548e-07, + -3.6508e-07, 5.9605e-08], + ..., + [ 1.1176e-08, 4.0978e-08, -1.3635e-06, ..., -4.7572e-06, + 1.7136e-07, -7.7263e-06], + [ 1.1176e-08, 1.4901e-08, 6.7055e-08, ..., 9.6858e-08, + 2.6450e-07, 2.9802e-08], + [ 7.4506e-09, 2.2352e-08, 2.0117e-07, ..., 3.6880e-07, + 2.8685e-07, 4.0978e-07]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0067, -0.0313, 0.0105, -0.0176, 0.0130, 0.0082, 0.0210, -0.0002, + -0.0341, -0.0075], device='cuda:0'), grad: tensor([-4.0904e-06, 1.4752e-05, 1.0803e-06, -2.1495e-06, 2.6822e-07, + 1.2144e-06, -1.0356e-06, -1.5289e-05, 9.3877e-07, 4.2543e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 259.25, cls_loss 0.0027 cls_loss_mapping 0.0048 cls_loss_causal 0.5357 re_mapping 0.0066 re_causal 0.0202 /// teacc 99.02 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0166, -0.1420, -0.0987, ..., -0.1880, -0.0815, -0.1306], + [ 0.0389, -0.0528, 0.0272, ..., 0.0479, 0.0876, -0.0362], + [-0.0637, 0.1060, -0.1338, ..., 0.0607, 0.0633, -0.0297], + ..., + [-0.0723, -0.0689, -0.0704, ..., 0.0027, -0.1181, 0.1169], + [ 0.0648, -0.0216, 0.0540, ..., -0.0047, -0.1511, -0.0081], + [-0.1325, -0.0375, -0.0499, ..., -0.1533, 0.0411, -0.0937]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.7253e-09, 1.0058e-07, ..., 5.9605e-08, + 1.5274e-07, 7.4506e-09], + [ 3.7253e-09, 2.6077e-08, -1.9744e-07, ..., 2.1979e-07, + -2.2352e-07, 1.1921e-07], + [ 1.1548e-07, 7.4506e-09, 2.7977e-06, ..., 1.8589e-06, + 2.1607e-07, 2.2352e-08], + ..., + [ 7.4506e-09, -5.5879e-08, 1.9744e-07, ..., -5.4389e-07, + 1.7136e-07, -3.3155e-07], + [-3.3528e-08, 3.7253e-09, 3.0547e-07, ..., 2.3097e-07, + 9.7975e-07, 7.4506e-08], + [ 1.1176e-08, 7.4506e-09, 1.1548e-07, ..., 1.1176e-07, + -1.0282e-06, 4.8429e-08]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0070, -0.0331, 0.0103, -0.0167, 0.0134, 0.0091, 0.0211, 0.0016, + -0.0344, -0.0086], device='cuda:0'), grad: tensor([ 4.9174e-07, 8.1956e-07, 5.1558e-06, -4.6082e-06, 2.0005e-06, + 5.2154e-08, -3.1888e-06, -1.4417e-06, 2.7120e-06, -2.0415e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 259.50, cls_loss 0.0020 cls_loss_mapping 0.0032 cls_loss_causal 0.5391 re_mapping 0.0065 re_causal 0.0202 /// teacc 98.92 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0156, -0.1421, -0.0998, ..., -0.1893, -0.0815, -0.1307], + [ 0.0395, -0.0531, 0.0279, ..., 0.0478, 0.0880, -0.0363], + [-0.0624, 0.1078, -0.1331, ..., 0.0615, 0.0644, -0.0298], + ..., + [-0.0733, -0.0693, -0.0707, ..., 0.0028, -0.1186, 0.1170], + [ 0.0636, -0.0239, 0.0537, ..., -0.0057, -0.1536, -0.0082], + [-0.1337, -0.0377, -0.0503, ..., -0.1537, 0.0408, -0.0937]], + device='cuda:0'), grad: tensor([[-8.1956e-08, 4.4703e-08, -1.2293e-07, ..., 2.3097e-07, + -7.1526e-07, 1.0058e-07], + [ 1.4901e-08, 5.2154e-08, 6.3144e-06, ..., 1.0572e-05, + 3.1292e-07, 2.9542e-06], + [ 3.3528e-08, -4.0606e-07, 8.4564e-07, ..., 3.1367e-06, + 1.5646e-07, 2.3581e-06], + ..., + [ 1.8626e-08, 1.6019e-07, -7.0184e-06, ..., -1.7792e-05, + -6.0350e-07, -8.1137e-06], + [ 7.4506e-09, 4.0978e-08, 1.8999e-07, ..., 1.2405e-06, + 3.7998e-07, 7.8604e-07], + [ 7.4506e-09, 7.4506e-09, 3.9116e-07, ..., 1.6801e-06, + -4.4331e-07, 9.0897e-07]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0069, -0.0329, 0.0112, -0.0169, 0.0138, 0.0096, 0.0209, 0.0015, + -0.0355, -0.0091], device='cuda:0'), grad: tensor([-4.6268e-06, 2.5183e-05, 1.2159e-05, 1.4231e-06, 3.4384e-06, + 8.7842e-06, -6.8508e-06, -4.8846e-05, 4.9248e-06, 4.3176e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 259.18, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.4920 re_mapping 0.0067 re_causal 0.0189 /// teacc 98.98 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0169, -0.1425, -0.1029, ..., -0.1932, -0.0816, -0.1307], + [ 0.0394, -0.0535, 0.0278, ..., 0.0477, 0.0880, -0.0364], + [-0.0621, 0.1088, -0.1337, ..., 0.0616, 0.0648, -0.0301], + ..., + [-0.0739, -0.0696, -0.0709, ..., 0.0029, -0.1191, 0.1172], + [ 0.0632, -0.0244, 0.0545, ..., -0.0058, -0.1537, -0.0082], + [-0.1344, -0.0380, -0.0511, ..., -0.1541, 0.0409, -0.0937]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.7509e-07, 1.0431e-07, ..., 2.3097e-07, + 5.4762e-07, 0.0000e+00], + [ 0.0000e+00, 1.9819e-06, 1.1511e-06, ..., 2.9132e-06, + 3.4794e-06, 1.8626e-08], + [ 0.0000e+00, 3.7551e-06, 1.3746e-06, ..., 4.2468e-06, + 2.9802e-07, 7.4506e-09], + ..., + [ 0.0000e+00, -1.4067e-05, -3.0920e-06, ..., -1.4327e-05, + 1.0692e-06, -1.0431e-07], + [ 0.0000e+00, 3.3341e-06, 4.1910e-06, ..., 4.7684e-06, + 1.1586e-06, 1.1176e-08], + [ 0.0000e+00, 4.2468e-07, 2.7195e-07, ..., 6.4075e-07, + 5.3756e-06, 7.4506e-09]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0073, -0.0331, 0.0113, -0.0167, 0.0136, 0.0103, 0.0207, 0.0016, + -0.0356, -0.0091], device='cuda:0'), grad: tensor([ 1.8328e-06, 1.7971e-05, 1.2122e-05, 1.1146e-05, -5.5909e-05, + -1.6347e-05, 1.7732e-05, -3.3319e-05, 2.0564e-05, 2.4125e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 259.53, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.4963 re_mapping 0.0073 re_causal 0.0196 /// teacc 98.92 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0181, -0.1442, -0.1036, ..., -0.1941, -0.0818, -0.1307], + [ 0.0394, -0.0537, 0.0277, ..., 0.0476, 0.0879, -0.0367], + [-0.0627, 0.1095, -0.1343, ..., 0.0617, 0.0652, -0.0298], + ..., + [-0.0740, -0.0706, -0.0711, ..., 0.0028, -0.1197, 0.1173], + [ 0.0636, -0.0240, 0.0549, ..., -0.0058, -0.1541, -0.0082], + [-0.1351, -0.0383, -0.0513, ..., -0.1544, 0.0409, -0.0939]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.6858e-08, 7.0781e-08, ..., 3.8743e-07, + 2.1607e-07, 2.6077e-08], + [ 0.0000e+00, 9.3132e-08, -2.3954e-06, ..., -5.9977e-07, + -1.4305e-06, 1.7136e-07], + [ 0.0000e+00, -3.0175e-07, 2.1607e-07, ..., 2.1607e-07, + -8.7172e-07, 6.4448e-07], + ..., + [ 0.0000e+00, -3.2037e-07, 1.2890e-06, ..., -2.9691e-06, + 1.1735e-06, -1.9707e-06], + [-1.4901e-08, 8.1956e-08, -9.4622e-07, ..., -3.7998e-07, + 3.6508e-07, 4.0978e-08], + [ 3.7253e-09, 2.9802e-08, 8.9407e-08, ..., 4.2468e-07, + -3.0175e-07, 1.4529e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0068, -0.0335, 0.0115, -0.0168, 0.0135, 0.0109, 0.0204, 0.0013, + -0.0356, -0.0090], device='cuda:0'), grad: tensor([-2.2724e-07, -2.1495e-06, 2.8647e-06, 7.8604e-06, 2.3469e-06, + -9.6485e-07, 8.2329e-07, -9.1046e-06, -4.6194e-07, -1.0058e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 259.59, cls_loss 0.0026 cls_loss_mapping 0.0048 cls_loss_causal 0.5769 re_mapping 0.0068 re_causal 0.0208 /// teacc 98.96 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0182, -0.1447, -0.1044, ..., -0.1965, -0.0820, -0.1308], + [ 0.0393, -0.0545, 0.0281, ..., 0.0475, 0.0881, -0.0370], + [-0.0631, 0.1077, -0.1357, ..., 0.0607, 0.0647, -0.0323], + ..., + [-0.0737, -0.0678, -0.0710, ..., 0.0034, -0.1203, 0.1182], + [ 0.0639, -0.0245, 0.0550, ..., -0.0059, -0.1551, -0.0083], + [-0.1355, -0.0385, -0.0516, ..., -0.1548, 0.0410, -0.0940]], + device='cuda:0'), grad: tensor([[-1.0952e-06, 1.0103e-05, 2.8349e-06, ..., 3.3155e-07, + 8.7917e-07, 6.7055e-08], + [ 7.4506e-09, 1.8626e-07, -3.1292e-07, ..., 2.3469e-07, + 8.4117e-06, 3.7625e-07], + [ 2.2352e-08, -1.8813e-06, 1.5795e-06, ..., -1.1697e-06, + -1.1176e-06, 1.5832e-06], + ..., + [ 0.0000e+00, 2.4475e-06, 2.1309e-06, ..., -1.3486e-06, + 4.4703e-06, -1.0893e-05], + [ 1.8626e-08, -1.4067e-05, 2.3097e-06, ..., 6.4708e-06, + 1.2442e-06, 1.7397e-06], + [ 1.7881e-07, 1.0096e-06, 8.6054e-07, ..., 5.3272e-07, + 6.7830e-05, 1.4156e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0069, -0.0335, 0.0097, -0.0175, 0.0136, 0.0118, 0.0214, 0.0020, + -0.0360, -0.0088], device='cuda:0'), grad: tensor([ 2.7508e-05, 2.0057e-05, 5.1744e-06, -2.3752e-05, -2.5749e-04, + 7.2896e-05, 5.8234e-05, -1.1533e-05, -2.6345e-05, 1.3518e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 259.64, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.5362 re_mapping 0.0068 re_causal 0.0209 /// teacc 98.96 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0189, -0.1451, -0.1049, ..., -0.1977, -0.0820, -0.1297], + [ 0.0403, -0.0547, 0.0285, ..., 0.0477, 0.0884, -0.0373], + [-0.0636, 0.1083, -0.1362, ..., 0.0608, 0.0646, -0.0323], + ..., + [-0.0742, -0.0683, -0.0715, ..., 0.0026, -0.1209, 0.1176], + [ 0.0640, -0.0245, 0.0549, ..., -0.0062, -0.1559, -0.0083], + [-0.1359, -0.0387, -0.0519, ..., -0.1550, 0.0410, -0.0942]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4506e-09, 8.9407e-08, ..., 7.0781e-08, + 3.7253e-07, 1.1176e-08], + [ 0.0000e+00, 2.6077e-08, -1.2703e-06, ..., -5.2154e-07, + -1.5050e-06, -1.9744e-07], + [ 0.0000e+00, -3.1292e-07, 1.2293e-07, ..., -7.0781e-07, + -3.5390e-07, 7.4506e-09], + ..., + [ 0.0000e+00, 1.8626e-07, 5.1036e-07, ..., 7.1898e-07, + 1.0356e-06, -1.2293e-07], + [ 0.0000e+00, 5.2154e-08, -4.6045e-06, ..., -2.4214e-06, + 2.5705e-06, 3.3528e-08], + [ 0.0000e+00, 3.7253e-09, 2.4214e-07, ..., 1.6764e-07, + -4.9546e-07, 1.4529e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0067, -0.0333, 0.0096, -0.0177, 0.0139, 0.0125, 0.0214, 0.0014, + -0.0363, -0.0089], device='cuda:0'), grad: tensor([ 1.1250e-06, -2.7269e-06, -1.1958e-06, 4.7162e-06, 1.1064e-06, + 8.5533e-06, -1.2018e-05, 2.3320e-06, -2.0377e-06, 1.0431e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 258.83, cls_loss 0.0020 cls_loss_mapping 0.0029 cls_loss_causal 0.5241 re_mapping 0.0069 re_causal 0.0210 /// teacc 98.82 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0189, -0.1453, -0.1056, ..., -0.1995, -0.0822, -0.1297], + [ 0.0403, -0.0551, 0.0287, ..., 0.0477, 0.0885, -0.0377], + [-0.0638, 0.1085, -0.1366, ..., 0.0608, 0.0649, -0.0323], + ..., + [-0.0744, -0.0685, -0.0716, ..., 0.0021, -0.1223, 0.1168], + [ 0.0641, -0.0242, 0.0562, ..., -0.0053, -0.1561, -0.0083], + [-0.1363, -0.0388, -0.0522, ..., -0.1531, 0.0411, -0.0918]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 1.9744e-07, ..., 3.3528e-08, + -5.3272e-07, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, -2.3469e-07, ..., -1.3411e-07, + -2.7195e-07, -2.6077e-08], + [ 7.4506e-09, 0.0000e+00, 7.8231e-08, ..., 2.6077e-08, + 3.2410e-07, 0.0000e+00], + ..., + [ 2.6077e-08, 0.0000e+00, 2.1979e-07, ..., 3.3528e-08, + 3.5763e-07, -1.1176e-08], + [ 7.4506e-09, 0.0000e+00, 1.5482e-05, ..., 2.2352e-06, + 1.1124e-05, 3.7253e-09], + [ 2.9802e-08, 0.0000e+00, -1.7256e-05, ..., -2.4736e-06, + -1.0982e-05, 2.2352e-08]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0067, -0.0334, 0.0096, -0.0176, 0.0143, 0.0105, 0.0212, 0.0007, + -0.0342, -0.0084], device='cuda:0'), grad: tensor([-3.8072e-06, -2.0862e-07, 1.5572e-06, 6.0350e-06, -1.4529e-06, + -5.2229e-06, 3.5763e-06, 2.2724e-07, 5.3048e-05, -5.3823e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 259.04, cls_loss 0.0021 cls_loss_mapping 0.0023 cls_loss_causal 0.5276 re_mapping 0.0069 re_causal 0.0209 /// teacc 98.83 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0192, -0.1455, -0.1062, ..., -0.2012, -0.0823, -0.1297], + [ 0.0401, -0.0552, 0.0287, ..., 0.0475, 0.0880, -0.0377], + [-0.0639, 0.1088, -0.1366, ..., 0.0614, 0.0658, -0.0323], + ..., + [-0.0749, -0.0686, -0.0717, ..., 0.0020, -0.1226, 0.1169], + [ 0.0643, -0.0242, 0.0567, ..., -0.0052, -0.1561, -0.0083], + [-0.1368, -0.0388, -0.0523, ..., -0.1534, 0.0406, -0.0919]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 3.7253e-09, 4.5449e-07, ..., 1.8254e-07, + 7.5623e-07, 7.4506e-09], + [ 5.9605e-08, 7.4506e-09, 2.3022e-06, ..., 9.1270e-07, + 5.5879e-07, 2.9802e-08], + [ 1.0431e-07, -7.0781e-08, 1.1884e-06, ..., 6.7055e-07, + 3.0175e-07, 4.8429e-08], + ..., + [ 3.7253e-08, 2.9802e-08, 1.2293e-06, ..., 4.0978e-07, + 6.2436e-06, -8.5682e-08], + [-4.9174e-07, 1.1176e-08, -2.1420e-06, ..., -1.7397e-06, + 2.3767e-06, -2.2724e-07], + [ 2.2352e-08, 3.7253e-09, 2.1867e-06, ..., 8.0466e-07, + -3.6120e-05, 4.0978e-08]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0065, -0.0338, 0.0103, -0.0177, 0.0147, 0.0105, 0.0208, 0.0006, + -0.0340, -0.0089], device='cuda:0'), grad: tensor([-2.5332e-07, 5.5656e-06, 3.3453e-06, -5.0291e-06, 6.0171e-05, + 7.1004e-06, 3.1702e-06, 1.8746e-05, 1.5236e-06, -9.4354e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 258.99, cls_loss 0.0022 cls_loss_mapping 0.0048 cls_loss_causal 0.5648 re_mapping 0.0068 re_causal 0.0207 /// teacc 98.95 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0200, -0.1457, -0.1065, ..., -0.2020, -0.0822, -0.1295], + [ 0.0400, -0.0558, 0.0307, ..., 0.0492, 0.0879, -0.0350], + [-0.0640, 0.1092, -0.1369, ..., 0.0616, 0.0663, -0.0325], + ..., + [-0.0752, -0.0685, -0.0738, ..., 0.0004, -0.1230, 0.1155], + [ 0.0643, -0.0244, 0.0574, ..., -0.0049, -0.1562, -0.0084], + [-0.1372, -0.0390, -0.0524, ..., -0.1537, 0.0405, -0.0921]], + device='cuda:0'), grad: tensor([[-3.7253e-09, 0.0000e+00, 8.5682e-08, ..., 7.4506e-09, + 1.4901e-07, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 3.9488e-07, ..., 2.9802e-08, + 8.2329e-07, 1.4529e-07], + [ 0.0000e+00, -2.9802e-08, 7.0781e-08, ..., -5.9605e-08, + 6.3330e-08, 2.2352e-08], + ..., + [ 0.0000e+00, 1.1176e-08, 7.5251e-07, ..., -1.0431e-07, + 1.4082e-06, -5.8487e-07], + [ 0.0000e+00, 3.7253e-09, 3.9898e-06, ..., 1.1176e-08, + 2.4177e-06, 1.9372e-07], + [ 0.0000e+00, 0.0000e+00, -2.1905e-06, ..., 6.7055e-08, + -3.5092e-06, 3.1292e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0059, -0.0321, 0.0104, -0.0169, 0.0149, 0.0099, 0.0203, -0.0008, + -0.0339, -0.0092], device='cuda:0'), grad: tensor([-4.1723e-07, 5.0180e-06, 9.2015e-07, 1.3724e-05, 1.0088e-05, + -8.0347e-05, 1.3649e-05, 2.3060e-06, 4.7922e-05, -1.2867e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 259.53, cls_loss 0.0021 cls_loss_mapping 0.0043 cls_loss_causal 0.5473 re_mapping 0.0066 re_causal 0.0201 /// teacc 98.86 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0212, -0.1464, -0.1068, ..., -0.2032, -0.0824, -0.1296], + [ 0.0398, -0.0559, 0.0296, ..., 0.0483, 0.0878, -0.0353], + [-0.0641, 0.1096, -0.1373, ..., 0.0616, 0.0664, -0.0328], + ..., + [-0.0765, -0.0685, -0.0730, ..., 0.0013, -0.1234, 0.1158], + [ 0.0643, -0.0246, 0.0577, ..., -0.0048, -0.1565, -0.0084], + [-0.1381, -0.0401, -0.0527, ..., -0.1541, 0.0404, -0.0923]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 8.1956e-08, 2.9802e-08, ..., 1.0803e-07, + 3.3528e-07, 3.7253e-09], + [ 1.4901e-08, 1.1176e-08, -2.3209e-06, ..., -1.7136e-07, + 3.1702e-06, 1.8626e-07], + [ 7.4506e-09, -3.9861e-07, 4.0978e-08, ..., -3.2037e-07, + -4.7684e-07, 1.8626e-08], + ..., + [ 2.7567e-07, 1.3784e-07, 2.6487e-06, ..., 1.1288e-06, + 3.7923e-06, -3.3528e-08], + [ 3.7253e-09, 1.1176e-08, 8.1956e-08, ..., 3.7625e-07, + 2.8312e-07, 1.1176e-08], + [ 1.4901e-08, 2.6077e-08, 6.3330e-08, ..., 1.2293e-07, + 5.9605e-07, 4.0978e-08]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0059, -0.0333, 0.0103, -0.0169, 0.0152, 0.0105, 0.0205, -0.0001, + -0.0341, -0.0095], device='cuda:0'), grad: tensor([-5.5134e-07, 3.2745e-06, -2.0862e-07, 1.2890e-06, -2.8417e-05, + -2.7828e-06, 1.1489e-05, 1.0177e-05, 2.0191e-06, 3.6918e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 260.02, cls_loss 0.0026 cls_loss_mapping 0.0041 cls_loss_causal 0.5126 re_mapping 0.0069 re_causal 0.0190 /// teacc 98.80 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0213, -0.1476, -0.1078, ..., -0.2050, -0.0827, -0.1297], + [ 0.0399, -0.0560, 0.0302, ..., 0.0483, 0.0889, -0.0354], + [-0.0642, 0.1102, -0.1382, ..., 0.0616, 0.0668, -0.0333], + ..., + [-0.0767, -0.0685, -0.0727, ..., 0.0017, -0.1241, 0.1162], + [ 0.0643, -0.0249, 0.0573, ..., -0.0052, -0.1588, -0.0085], + [-0.1384, -0.0416, -0.0530, ..., -0.1542, 0.0405, -0.0926]], + device='cuda:0'), grad: tensor([[-2.5444e-06, -1.3784e-07, 3.5018e-07, ..., 7.0408e-07, + 7.8231e-08, 4.0978e-08], + [ 7.2271e-07, 1.8626e-08, 2.3395e-06, ..., 1.0766e-05, + -3.3826e-06, 5.5134e-07], + [ 2.6077e-08, 3.0175e-07, 2.5466e-05, ..., 5.0962e-05, + 2.6077e-08, 1.5870e-06], + ..., + [ 1.1176e-08, -4.8429e-08, -3.4392e-05, ..., -7.5400e-05, + 3.3379e-06, -2.8275e-06], + [ 1.3039e-07, -2.7940e-07, 9.3132e-08, ..., 5.9232e-07, + 1.1548e-07, 7.8231e-08], + [ 3.7253e-08, 3.7253e-08, 8.4192e-07, ..., 1.7174e-06, + 7.0781e-08, 1.7881e-07]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0057, -0.0331, 0.0102, -0.0180, 0.0150, 0.0111, 0.0205, 0.0003, + -0.0347, -0.0093], device='cuda:0'), grad: tensor([ 3.3170e-05, 2.7597e-05, 1.1897e-04, 2.4214e-05, -2.6077e-08, + 1.7472e-06, -3.8892e-05, -1.7452e-04, 2.7455e-06, 5.1335e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 259.00, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5356 re_mapping 0.0064 re_causal 0.0197 /// teacc 98.96 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0212, -0.1478, -0.1084, ..., -0.2062, -0.0829, -0.1297], + [ 0.0401, -0.0564, 0.0309, ..., 0.0493, 0.0893, -0.0347], + [-0.0641, 0.1105, -0.1386, ..., 0.0616, 0.0669, -0.0334], + ..., + [-0.0768, -0.0685, -0.0734, ..., 0.0010, -0.1249, 0.1159], + [ 0.0642, -0.0251, 0.0579, ..., -0.0048, -0.1586, -0.0085], + [-0.1385, -0.0420, -0.0532, ..., -0.1545, 0.0406, -0.0929]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 0.0000e+00, 5.5879e-08, ..., 8.1956e-08, + 2.9802e-07, 5.2154e-08], + [ 2.2352e-08, 0.0000e+00, -3.5763e-07, ..., 1.1288e-06, + 3.5390e-06, 1.1511e-06], + [ 4.4703e-08, 2.7195e-07, 2.8312e-07, ..., 8.5682e-07, + 7.4133e-07, 5.4389e-07], + ..., + [ 2.2352e-08, -2.7567e-07, 3.3155e-07, ..., 1.0394e-05, + 3.4541e-05, 9.2313e-06], + [ 3.3900e-07, 0.0000e+00, 1.3001e-06, ..., 7.4878e-07, + 1.4380e-06, 3.0175e-07], + [ 1.4901e-08, 0.0000e+00, 1.2293e-07, ..., -1.6853e-05, + -5.4359e-05, -1.5073e-05]], device='cuda:0') +Epoch 178, bias, value: tensor([-5.7890e-03, -3.2410e-02, 1.0068e-02, -1.8432e-02, 1.5083e-02, + 1.1284e-02, 2.0358e-02, -9.5728e-05, -3.4629e-02, -9.4174e-03], + device='cuda:0'), grad: tensor([ 9.2387e-07, 2.1249e-05, 4.7125e-06, -8.4564e-07, 7.1406e-05, + 2.3097e-06, 3.1665e-07, 1.8418e-04, 1.1019e-05, -2.9564e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 259.13, cls_loss 0.0022 cls_loss_mapping 0.0043 cls_loss_causal 0.5291 re_mapping 0.0062 re_causal 0.0191 /// teacc 98.93 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0216, -0.1481, -0.1092, ..., -0.2078, -0.0829, -0.1294], + [ 0.0408, -0.0566, 0.0303, ..., 0.0490, 0.0910, -0.0363], + [-0.0644, 0.1107, -0.1414, ..., 0.0595, 0.0645, -0.0336], + ..., + [-0.0769, -0.0686, -0.0719, ..., 0.0028, -0.1238, 0.1171], + [ 0.0641, -0.0250, 0.0578, ..., -0.0049, -0.1598, -0.0086], + [-0.1389, -0.0423, -0.0535, ..., -0.1554, 0.0404, -0.0928]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, 1.9558e-07, ..., 8.1398e-07, + 1.0114e-06, 9.3132e-08], + [-1.1548e-07, 3.7253e-09, 7.5027e-06, ..., 1.0930e-05, + 1.8895e-05, 1.5207e-05], + [ 3.7253e-09, -1.2480e-07, 2.0191e-06, ..., 2.5705e-07, + 2.1085e-06, 3.4403e-06], + ..., + [ 3.7253e-09, 1.0990e-07, -1.3940e-05, ..., -2.5406e-05, + -2.3097e-05, -4.1425e-05], + [ 3.7253e-09, 0.0000e+00, 1.7837e-05, ..., 2.4274e-05, + 3.0786e-05, 5.7966e-06], + [ 1.8626e-09, 0.0000e+00, 5.6997e-07, ..., 8.2254e-06, + 1.2033e-05, 1.4469e-05]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0055, -0.0331, 0.0080, -0.0183, 0.0155, 0.0115, 0.0204, 0.0017, + -0.0350, -0.0098], device='cuda:0'), grad: tensor([ 2.5537e-06, 5.8472e-05, 8.2403e-06, 3.3937e-06, -3.2544e-05, + -1.7858e-04, 3.2812e-05, -1.8406e-04, 1.6129e-04, 1.2815e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 259.47, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5339 re_mapping 0.0061 re_causal 0.0190 /// teacc 98.85 lr 0.00010000 +Epoch 180, weight, value: tensor([[ 0.0217, -0.1482, -0.1083, ..., -0.2091, -0.0856, -0.1292], + [ 0.0407, -0.0574, 0.0323, ..., 0.0514, 0.0923, -0.0335], + [-0.0645, 0.1109, -0.1420, ..., 0.0594, 0.0646, -0.0339], + ..., + [-0.0771, -0.0686, -0.0737, ..., 0.0005, -0.1261, 0.1147], + [ 0.0641, -0.0251, 0.0578, ..., -0.0049, -0.1608, -0.0086], + [-0.1390, -0.0424, -0.0539, ..., -0.1556, 0.0419, -0.0929]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4506e-09, 5.5879e-09, ..., 1.3039e-08, + 8.0094e-07, 1.8626e-09], + [ 1.8626e-09, 7.4506e-09, -1.0990e-07, ..., -9.3132e-09, + -5.9605e-08, 2.9802e-08], + [ 5.5879e-09, -1.5274e-07, 4.4703e-08, ..., -8.3819e-08, + 1.1176e-08, 3.7253e-09], + ..., + [ 0.0000e+00, 1.1548e-07, 6.5193e-08, ..., 8.1956e-08, + 2.0675e-07, -7.0781e-08], + [ 6.7055e-08, 7.4506e-09, 4.0978e-08, ..., -8.1956e-08, + 1.5628e-06, -1.8626e-09], + [ 0.0000e+00, 3.7253e-09, 7.4506e-09, ..., 1.6764e-08, + 1.0617e-07, 2.0489e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0073, -0.0310, 0.0079, -0.0184, 0.0154, 0.0117, 0.0204, -0.0004, + -0.0354, -0.0085], device='cuda:0'), grad: tensor([ 5.1185e-06, 2.1607e-07, 7.0781e-08, 3.2037e-07, -2.6077e-08, + 1.5646e-05, -3.1263e-05, 2.2911e-07, 9.0078e-06, 7.2643e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 259.08, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5158 re_mapping 0.0063 re_causal 0.0193 /// teacc 98.93 lr 0.00010000 +Epoch 181, weight, value: tensor([[ 0.0217, -0.1478, -0.1103, ..., -0.2113, -0.0856, -0.1293], + [ 0.0408, -0.0578, 0.0324, ..., 0.0514, 0.0924, -0.0335], + [-0.0645, 0.1112, -0.1418, ..., 0.0598, 0.0650, -0.0341], + ..., + [-0.0775, -0.0687, -0.0739, ..., 0.0005, -0.1268, 0.1147], + [ 0.0640, -0.0252, 0.0579, ..., -0.0050, -0.1614, -0.0085], + [-0.1394, -0.0428, -0.0544, ..., -0.1556, 0.0420, -0.0925]], + device='cuda:0'), grad: tensor([[-1.1176e-08, -3.3528e-08, 1.0058e-07, ..., 2.9802e-08, + 2.0303e-07, 5.5879e-09], + [ 1.8626e-09, 5.5879e-09, -6.1654e-07, ..., 4.4703e-08, + -7.2457e-07, 2.6450e-07], + [ 1.8626e-09, 1.8626e-09, 5.0105e-07, ..., 3.8743e-07, + 2.2165e-07, 2.4214e-08], + ..., + [ 0.0000e+00, 1.8626e-09, 7.0781e-08, ..., -1.9614e-06, + 1.9185e-07, -1.2945e-06], + [ 0.0000e+00, -9.3132e-09, 2.6785e-06, ..., 1.5777e-06, + 3.5893e-06, 9.1270e-08], + [ 3.7253e-09, 3.1665e-08, -1.8068e-07, ..., 5.1223e-07, + -2.5518e-07, 3.5390e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0071, -0.0310, 0.0083, -0.0186, 0.0154, 0.0118, 0.0204, -0.0007, + -0.0355, -0.0083], device='cuda:0'), grad: tensor([-3.1721e-06, -1.7881e-07, 1.2685e-06, -2.9802e-07, 1.1735e-07, + 1.5646e-06, -7.4878e-06, -4.2915e-06, 9.2387e-06, 3.2224e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 259.12, cls_loss 0.0018 cls_loss_mapping 0.0041 cls_loss_causal 0.5276 re_mapping 0.0064 re_causal 0.0190 /// teacc 98.80 lr 0.00010000 +Epoch 182, weight, value: tensor([[ 0.0217, -0.1479, -0.1107, ..., -0.2126, -0.0857, -0.1293], + [ 0.0409, -0.0583, 0.0324, ..., 0.0512, 0.0922, -0.0336], + [-0.0653, 0.1112, -0.1423, ..., 0.0598, 0.0652, -0.0337], + ..., + [-0.0777, -0.0689, -0.0739, ..., 0.0005, -0.1271, 0.1148], + [ 0.0647, -0.0244, 0.0582, ..., -0.0046, -0.1624, -0.0086], + [-0.1397, -0.0430, -0.0542, ..., -0.1551, 0.0421, -0.0925]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 4.2841e-08, ..., 8.9407e-08, + 3.0175e-07, 6.1467e-08], + [ 2.0489e-08, 1.8626e-09, -3.5949e-07, ..., 1.8254e-07, + -7.9907e-07, 4.1537e-07], + [ 1.1176e-08, -1.6764e-08, 2.7753e-07, ..., 4.2841e-08, + 8.3819e-08, 1.2480e-07], + ..., + [-6.3330e-08, 9.3132e-09, -1.5646e-07, ..., -1.5777e-06, + 3.9861e-07, -1.6782e-06], + [ 1.8626e-09, 3.7253e-09, 9.1270e-08, ..., 2.0675e-07, + 6.5006e-07, 1.5087e-07], + [ 1.4901e-08, 0.0000e+00, 9.4995e-08, ..., 8.1584e-07, + -1.0245e-07, 8.0653e-07]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0071, -0.0312, 0.0082, -0.0186, 0.0151, 0.0117, 0.0208, -0.0007, + -0.0354, -0.0079], device='cuda:0'), grad: tensor([ 9.5740e-07, 7.6368e-07, 5.7556e-07, 6.7055e-07, 1.6224e-06, + -2.9244e-07, -2.1104e-06, -9.0674e-06, 2.5388e-06, 4.3213e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 259.28, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5218 re_mapping 0.0062 re_causal 0.0194 /// teacc 98.82 lr 0.00010000 +Epoch 183, weight, value: tensor([[ 0.0220, -0.1479, -0.1110, ..., -0.2138, -0.0857, -0.1293], + [ 0.0409, -0.0584, 0.0324, ..., 0.0511, 0.0922, -0.0336], + [-0.0654, 0.1112, -0.1426, ..., 0.0594, 0.0655, -0.0338], + ..., + [-0.0778, -0.0689, -0.0739, ..., 0.0008, -0.1272, 0.1149], + [ 0.0648, -0.0242, 0.0584, ..., -0.0044, -0.1629, -0.0086], + [-0.1399, -0.0430, -0.0547, ..., -0.1556, 0.0411, -0.0926]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.8626e-09, 2.9560e-06, ..., 2.7940e-08, + 8.0094e-08, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, -3.6228e-06, ..., -3.3230e-06, + -4.4741e-06, 1.4901e-08], + [ 0.0000e+00, -4.2282e-07, 1.5777e-06, ..., 6.3330e-07, + 1.1306e-06, 1.8626e-09], + ..., + [ 0.0000e+00, 2.4214e-08, 1.2200e-06, ..., 1.0449e-06, + 1.3579e-06, -4.2841e-08], + [ 3.7253e-09, 3.7439e-07, 1.9073e-05, ..., 8.2143e-07, + 1.0412e-06, 5.5879e-09], + [ 0.0000e+00, 1.8626e-09, 1.2293e-06, ..., 5.9977e-07, + 4.5076e-07, 1.1176e-08]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0069, -0.0314, 0.0079, -0.0186, 0.0164, 0.0119, 0.0206, -0.0004, + -0.0355, -0.0091], device='cuda:0'), grad: tensor([ 1.2495e-05, -5.1372e-06, 1.9763e-06, 4.0978e-07, -1.6205e-07, + -7.6914e-04, 6.5899e-04, 2.0899e-06, 9.4533e-05, 4.2319e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 259.44, cls_loss 0.0018 cls_loss_mapping 0.0035 cls_loss_causal 0.5203 re_mapping 0.0064 re_causal 0.0189 /// teacc 98.82 lr 0.00010000 +Epoch 184, weight, value: tensor([[ 0.0220, -0.1479, -0.1128, ..., -0.2164, -0.0858, -0.1294], + [ 0.0409, -0.0586, 0.0324, ..., 0.0510, 0.0921, -0.0338], + [-0.0655, 0.1109, -0.1431, ..., 0.0594, 0.0656, -0.0338], + ..., + [-0.0778, -0.0689, -0.0739, ..., 0.0008, -0.1281, 0.1151], + [ 0.0649, -0.0234, 0.0577, ..., -0.0046, -0.1633, -0.0087], + [-0.1400, -0.0430, -0.0551, ..., -0.1561, 0.0412, -0.0928]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.9481e-06, ..., 2.5574e-06, + 1.9558e-07, 7.4506e-09], + [ 1.4901e-08, 0.0000e+00, 1.7509e-07, ..., 1.1735e-07, + 1.1362e-07, 1.8626e-08], + [ 5.5879e-09, -3.7253e-09, 2.9244e-07, ..., 1.0058e-07, + 1.7509e-07, 3.7253e-09], + ..., + [ 2.0489e-08, 0.0000e+00, 3.7998e-07, ..., 1.0990e-07, + 8.7544e-08, -8.5682e-08], + [-6.3330e-08, 1.8626e-09, -8.3447e-07, ..., -5.2899e-07, + 3.6694e-07, 9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 2.2165e-07, ..., 6.3330e-08, + 5.8115e-07, 5.4017e-08]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0069, -0.0315, 0.0076, -0.0183, 0.0174, 0.0126, 0.0197, -0.0006, + -0.0359, -0.0091], device='cuda:0'), grad: tensor([ 1.6466e-05, 9.4064e-07, 1.6484e-06, -1.3098e-05, -4.6976e-06, + -6.7568e-04, 6.6710e-04, 9.2387e-07, 3.0436e-06, 3.7756e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 259.27, cls_loss 0.0019 cls_loss_mapping 0.0032 cls_loss_causal 0.5229 re_mapping 0.0063 re_causal 0.0180 /// teacc 98.93 lr 0.00010000 +Epoch 185, weight, value: tensor([[ 0.0229, -0.1480, -0.1130, ..., -0.2173, -0.0861, -0.1294], + [ 0.0409, -0.0587, 0.0327, ..., 0.0511, 0.0924, -0.0338], + [-0.0646, 0.1112, -0.1435, ..., 0.0596, 0.0658, -0.0339], + ..., + [-0.0779, -0.0690, -0.0741, ..., 0.0007, -0.1284, 0.1152], + [ 0.0650, -0.0237, 0.0581, ..., -0.0044, -0.1641, -0.0087], + [-0.1403, -0.0432, -0.0562, ..., -0.1567, 0.0411, -0.0929]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, 0.0000e+00, 7.2643e-08, ..., 2.7940e-08, + -1.2107e-07, 9.3132e-09], + [ 3.7253e-09, 0.0000e+00, -3.4831e-07, ..., 2.2352e-07, + -2.9616e-07, 1.6391e-07], + [ 2.0489e-08, -3.7253e-09, 1.3411e-07, ..., 1.3597e-07, + 5.8115e-07, 6.5193e-08], + ..., + [ 0.0000e+00, 1.8626e-09, 3.5390e-08, ..., -1.9222e-06, + 9.6858e-08, -9.8348e-07], + [ 3.7253e-09, 0.0000e+00, 3.1106e-07, ..., 1.0617e-07, + 5.6066e-07, 4.8429e-08], + [ 3.7253e-09, 0.0000e+00, 2.4214e-08, ..., 1.0580e-06, + 2.4773e-07, 4.8615e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0069, -0.0314, 0.0079, -0.0182, 0.0173, 0.0125, 0.0195, -0.0006, + -0.0358, -0.0094], device='cuda:0'), grad: tensor([-6.8061e-06, 7.8231e-07, 3.2075e-06, 2.1756e-06, -5.3793e-06, + -8.4378e-07, 2.5574e-06, -5.5656e-06, 4.0457e-06, 5.8077e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 259.21, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5133 re_mapping 0.0066 re_causal 0.0186 /// teacc 98.92 lr 0.00010000 +Epoch 186, weight, value: tensor([[ 0.0232, -0.1474, -0.1136, ..., -0.2174, -0.0866, -0.1294], + [ 0.0410, -0.0589, 0.0328, ..., 0.0509, 0.0926, -0.0339], + [-0.0644, 0.1114, -0.1439, ..., 0.0593, 0.0659, -0.0340], + ..., + [-0.0780, -0.0691, -0.0741, ..., 0.0011, -0.1285, 0.1153], + [ 0.0651, -0.0238, 0.0588, ..., -0.0042, -0.1643, -0.0087], + [-0.1406, -0.0433, -0.0574, ..., -0.1571, 0.0420, -0.0929]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 2.4214e-08, ..., 2.6077e-08, + 7.2643e-08, 2.2352e-08], + [ 1.8626e-09, 0.0000e+00, 3.4362e-05, ..., 3.9309e-05, + -4.6566e-08, 3.9279e-05], + [ 1.8626e-09, -2.2352e-08, 1.0990e-07, ..., 3.5577e-07, + 2.6077e-08, 1.8626e-07], + ..., + [ 5.5879e-09, 1.1176e-08, -3.6359e-05, ..., -4.1842e-05, + 9.8720e-08, -4.1664e-05], + [ 1.8626e-09, 3.7253e-09, 1.2349e-06, ..., 1.4398e-06, + 5.5879e-08, 1.4026e-06], + [ 1.8626e-09, 0.0000e+00, 1.0058e-07, ..., 6.8918e-08, + -7.1526e-07, 1.1548e-07]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0067, -0.0315, 0.0075, -0.0182, 0.0164, 0.0126, 0.0194, -0.0003, + -0.0355, -0.0090], device='cuda:0'), grad: tensor([-4.4703e-07, 1.6320e-04, 1.0375e-06, 2.5127e-06, 2.4103e-06, + 8.8662e-07, 1.2480e-07, -1.7273e-04, 6.4149e-06, -3.2596e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 259.06, cls_loss 0.0025 cls_loss_mapping 0.0032 cls_loss_causal 0.5355 re_mapping 0.0070 re_causal 0.0197 /// teacc 98.98 lr 0.00010000 +Epoch 187, weight, value: tensor([[ 0.0231, -0.1476, -0.1143, ..., -0.2190, -0.0868, -0.1295], + [ 0.0410, -0.0591, 0.0318, ..., 0.0500, 0.0933, -0.0347], + [-0.0644, 0.1119, -0.1447, ..., 0.0596, 0.0660, -0.0342], + ..., + [-0.0781, -0.0699, -0.0729, ..., 0.0016, -0.1298, 0.1160], + [ 0.0654, -0.0233, 0.0595, ..., -0.0038, -0.1648, -0.0088], + [-0.1407, -0.0435, -0.0583, ..., -0.1556, 0.0422, -0.0920]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.3528e-08, 4.4145e-07, ..., 4.9919e-07, + 9.7416e-07, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, -2.1243e-04, ..., -1.7107e-04, + -1.2660e-04, 1.1176e-08], + [ 0.0000e+00, -4.0978e-08, 3.6154e-06, ..., 4.5933e-06, + 1.4529e-07, 9.3132e-09], + ..., + [ 0.0000e+00, 2.4959e-07, 2.0516e-04, ..., 1.6248e-04, + 1.2290e-04, -9.3132e-08], + [-1.8626e-09, 5.5879e-09, 1.3504e-06, ..., 2.7753e-06, + 2.5313e-06, 3.7253e-09], + [ 0.0000e+00, 7.4506e-09, 2.5611e-06, ..., 2.5518e-06, + 1.2070e-06, 5.9605e-08]], device='cuda:0') +Epoch 187, bias, value: tensor([-6.6125e-03, -3.2429e-02, 7.3708e-03, -1.7977e-02, 1.6774e-02, + 1.2034e-02, 1.9091e-02, 8.3569e-05, -3.5019e-02, -8.5028e-03], + device='cuda:0'), grad: tensor([ 4.2915e-06, -5.3453e-04, 1.5005e-05, -1.5259e-05, 5.8301e-07, + 1.9416e-05, -1.2219e-05, 5.0640e-04, 5.4613e-06, 1.1928e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 259.11, cls_loss 0.0022 cls_loss_mapping 0.0054 cls_loss_causal 0.5129 re_mapping 0.0064 re_causal 0.0194 /// teacc 98.94 lr 0.00010000 +Epoch 188, weight, value: tensor([[ 0.0230, -0.1478, -0.1150, ..., -0.2208, -0.0865, -0.1296], + [ 0.0407, -0.0586, 0.0329, ..., 0.0514, 0.0960, -0.0348], + [-0.0644, 0.1123, -0.1474, ..., 0.0578, 0.0642, -0.0349], + ..., + [-0.0782, -0.0703, -0.0735, ..., 0.0009, -0.1311, 0.1157], + [ 0.0663, -0.0234, 0.0613, ..., -0.0017, -0.1648, -0.0075], + [-0.1408, -0.0440, -0.0587, ..., -0.1562, 0.0417, -0.0922]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.5390e-08, 3.7998e-07, ..., 1.5665e-06, + 7.2643e-07, 1.8626e-08], + [ 0.0000e+00, 3.7253e-09, -4.0010e-06, ..., -6.5006e-07, + -5.4501e-06, 8.0094e-08], + [ 0.0000e+00, 4.2841e-08, 3.1535e-06, ..., 8.1733e-06, + 3.8557e-06, 4.6790e-06], + ..., + [ 0.0000e+00, 3.7253e-09, 6.7241e-07, ..., -1.8403e-05, + 9.6112e-07, -5.0403e-06], + [ 0.0000e+00, -2.0117e-07, -1.6168e-06, ..., 7.3574e-07, + 2.5749e-05, 5.0291e-08], + [ 0.0000e+00, 1.4901e-08, 2.8685e-07, ..., 9.1456e-07, + -2.6077e-08, 7.4506e-08]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0059, -0.0314, 0.0057, -0.0181, 0.0174, 0.0110, 0.0195, -0.0005, + -0.0335, -0.0092], device='cuda:0'), grad: tensor([ 7.6890e-06, 2.6748e-05, 2.0906e-05, 2.8208e-05, 3.2913e-06, + 7.6443e-06, -6.6400e-05, -1.5712e-04, 6.8724e-05, 6.0290e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 258.87, cls_loss 0.0018 cls_loss_mapping 0.0045 cls_loss_causal 0.5228 re_mapping 0.0062 re_causal 0.0198 /// teacc 98.99 lr 0.00010000 +Epoch 189, weight, value: tensor([[ 0.0226, -0.1480, -0.1171, ..., -0.2237, -0.0862, -0.1296], + [ 0.0409, -0.0587, 0.0328, ..., 0.0510, 0.0961, -0.0351], + [-0.0644, 0.1126, -0.1479, ..., 0.0578, 0.0641, -0.0353], + ..., + [-0.0783, -0.0704, -0.0731, ..., 0.0015, -0.1305, 0.1160], + [ 0.0666, -0.0234, 0.0612, ..., -0.0019, -0.1655, -0.0075], + [-0.1408, -0.0441, -0.0591, ..., -0.1565, 0.0420, -0.0922]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.9360e-07, ..., 2.7381e-07, + 6.1840e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.7031e-05, ..., 2.0742e-05, + 1.2051e-06, 9.3132e-09], + [ 0.0000e+00, -5.5879e-09, 1.0896e-06, ..., 7.7300e-07, + 2.8498e-07, 5.5879e-09], + ..., + [ 0.0000e+00, 1.8626e-09, -5.1379e-05, ..., -4.0084e-05, + 2.3656e-07, -3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 7.2159e-06, ..., 7.2978e-06, + -5.6624e-06, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.5553e-06, ..., 1.1791e-06, + 5.9605e-08, 9.3132e-09]], device='cuda:0') +Epoch 189, bias, value: tensor([-4.1264e-03, -3.1662e-02, 5.4680e-03, -1.8104e-02, 1.7119e-02, + 9.1500e-03, 2.0023e-02, 5.7972e-05, -3.3936e-02, -8.9590e-03], + device='cuda:0'), grad: tensor([ 3.6489e-06, 1.0389e-04, 3.0901e-06, 9.2089e-06, 1.3195e-05, + 6.4597e-06, -8.6501e-06, -2.1636e-04, 7.9274e-05, 5.8189e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 258.95, cls_loss 0.0015 cls_loss_mapping 0.0037 cls_loss_causal 0.5149 re_mapping 0.0064 re_causal 0.0190 /// teacc 99.06 lr 0.00010000 +Epoch 190, weight, value: tensor([[ 0.0223, -0.1482, -0.1188, ..., -0.2263, -0.0863, -0.1296], + [ 0.0409, -0.0590, 0.0331, ..., 0.0511, 0.0963, -0.0352], + [-0.0644, 0.1129, -0.1480, ..., 0.0578, 0.0643, -0.0354], + ..., + [-0.0783, -0.0707, -0.0733, ..., 0.0013, -0.1309, 0.1160], + [ 0.0665, -0.0235, 0.0609, ..., -0.0022, -0.1660, -0.0075], + [-0.1409, -0.0442, -0.0593, ..., -0.1566, 0.0420, -0.0922]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 5.5879e-09, 2.9802e-07, ..., 1.1548e-07, + 8.5682e-08, 5.5879e-09], + [ 4.2841e-08, 1.8626e-09, -1.0673e-06, ..., -4.2841e-08, + -1.4082e-06, 8.7544e-08], + [ 9.3132e-09, -1.6764e-08, 2.3842e-07, ..., 1.2293e-07, + 2.4214e-08, 1.1176e-08], + ..., + [ 2.6077e-08, 5.5879e-09, 3.7812e-07, ..., -7.6182e-07, + 1.0990e-07, -4.8988e-07], + [ 2.3842e-07, 1.8626e-09, 2.1160e-06, ..., 7.0594e-07, + 3.5949e-07, 4.2841e-08], + [ 5.5879e-08, 1.8626e-09, 5.3458e-07, ..., 2.9616e-07, + 1.6764e-08, 5.7742e-08]], device='cuda:0') +Epoch 190, bias, value: tensor([-4.3177e-03, -3.1596e-02, 5.5106e-03, -1.8221e-02, 1.7231e-02, + 9.6759e-03, 2.0172e-02, 4.5741e-06, -3.4630e-02, -8.9778e-03], + device='cuda:0'), grad: tensor([ 1.5050e-06, -1.3411e-06, 5.1409e-07, -2.5779e-05, 1.3672e-06, + 2.4036e-05, 1.6950e-06, -1.1232e-06, 6.0201e-06, -6.9812e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 258.74, cls_loss 0.0017 cls_loss_mapping 0.0034 cls_loss_causal 0.5449 re_mapping 0.0062 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 +Epoch 191, weight, value: tensor([[ 0.0248, -0.1483, -0.1191, ..., -0.2285, -0.0864, -0.1296], + [ 0.0408, -0.0593, 0.0334, ..., 0.0512, 0.0963, -0.0351], + [-0.0645, 0.1130, -0.1483, ..., 0.0581, 0.0649, -0.0357], + ..., + [-0.0784, -0.0708, -0.0735, ..., 0.0011, -0.1316, 0.1160], + [ 0.0665, -0.0233, 0.0611, ..., -0.0022, -0.1666, -0.0075], + [-0.1415, -0.0445, -0.0592, ..., -0.1568, 0.0421, -0.0921]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -3.4831e-07, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6380e-07, ..., -2.7940e-08, + -3.5390e-07, 1.8626e-09], + [ 0.0000e+00, -5.5879e-09, 1.1362e-07, ..., 1.1548e-07, + 1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 9.8720e-08, ..., 6.5193e-08, + 9.8720e-08, -9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 2.2724e-07, ..., 1.0431e-07, + 1.2293e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 2.5146e-07, + -3.2224e-07, 3.7253e-09]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0041, -0.0315, 0.0059, -0.0188, 0.0167, 0.0098, 0.0199, -0.0001, + -0.0348, -0.0084], device='cuda:0'), grad: tensor([-3.5781e-06, -8.9407e-07, 2.9057e-07, 8.7544e-08, 1.0822e-06, + 4.9919e-07, 8.1956e-08, 2.8498e-07, 5.4576e-07, 1.6075e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 258.78, cls_loss 0.0016 cls_loss_mapping 0.0031 cls_loss_causal 0.5125 re_mapping 0.0062 re_causal 0.0186 /// teacc 98.97 lr 0.00010000 +Epoch 192, weight, value: tensor([[ 0.0248, -0.1489, -0.1196, ..., -0.2293, -0.0865, -0.1296], + [ 0.0409, -0.0593, 0.0335, ..., 0.0512, 0.0964, -0.0352], + [-0.0646, 0.1133, -0.1484, ..., 0.0584, 0.0653, -0.0358], + ..., + [-0.0784, -0.0710, -0.0736, ..., 0.0008, -0.1326, 0.1160], + [ 0.0666, -0.0233, 0.0613, ..., -0.0021, -0.1669, -0.0075], + [-0.1417, -0.0446, -0.0594, ..., -0.1564, 0.0422, -0.0923]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.7881e-07, 1.6764e-07, ..., 4.4703e-07, + 5.9046e-07, 3.7253e-09], + [ 0.0000e+00, 3.3528e-08, -1.0997e-05, ..., -6.5118e-06, + -8.8364e-06, 5.7742e-08], + [ 0.0000e+00, -4.7497e-07, 2.5667e-06, ..., 1.7770e-06, + 8.1584e-07, 1.1176e-08], + ..., + [ 3.7253e-09, 9.8720e-08, 3.0920e-06, ..., -7.1600e-06, + 2.4065e-06, -2.3656e-07], + [-0.0000e+00, 1.6764e-08, 1.8440e-06, ..., 1.2983e-06, + 1.3653e-06, 3.7253e-08], + [ 0.0000e+00, 4.0978e-08, 2.3656e-07, ..., 1.3318e-06, + -1.1120e-06, 5.0291e-08]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0042, -0.0317, 0.0062, -0.0187, 0.0166, 0.0100, 0.0200, -0.0005, + -0.0348, -0.0081], device='cuda:0'), grad: tensor([-7.2904e-06, -2.1070e-05, 5.3123e-06, 1.1235e-05, 1.0148e-05, + 7.4431e-06, 1.6559e-06, -1.3798e-05, 4.8503e-06, 1.4734e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 259.28, cls_loss 0.0020 cls_loss_mapping 0.0039 cls_loss_causal 0.5306 re_mapping 0.0062 re_causal 0.0184 /// teacc 98.88 lr 0.00010000 +Epoch 193, weight, value: tensor([[ 0.0249, -0.1493, -0.1206, ..., -0.2304, -0.0867, -0.1297], + [ 0.0406, -0.0604, 0.0335, ..., 0.0508, 0.0956, -0.0352], + [-0.0647, 0.1140, -0.1479, ..., 0.0585, 0.0662, -0.0359], + ..., + [-0.0785, -0.0712, -0.0737, ..., 0.0013, -0.1327, 0.1161], + [ 0.0669, -0.0233, 0.0622, ..., -0.0020, -0.1673, -0.0076], + [-0.1418, -0.0453, -0.0598, ..., -0.1566, 0.0425, -0.0924]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-09, 3.9265e-06, ..., 2.2352e-07, + 2.1420e-07, 1.4901e-08], + [ 1.8626e-09, 0.0000e+00, -2.2680e-05, ..., -1.4670e-05, + -1.5572e-05, -1.1362e-06], + [-1.8626e-08, -6.8918e-08, 3.1777e-06, ..., 1.7677e-06, + 1.6801e-06, 2.0303e-07], + ..., + [ 7.4506e-09, 5.5879e-09, 8.0615e-06, ..., 4.1835e-06, + 5.0999e-06, -1.0431e-06], + [ 2.4214e-08, 5.5879e-09, -5.1260e-06, ..., 2.8443e-06, + 1.8924e-06, 5.2154e-07], + [ 1.8626e-09, 9.3132e-09, 2.3339e-06, ..., 2.2911e-06, + 1.7695e-07, 1.3970e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0048, -0.0324, 0.0068, -0.0192, 0.0163, 0.0099, 0.0203, -0.0001, + -0.0342, -0.0077], device='cuda:0'), grad: tensor([ 8.5682e-06, -4.9800e-05, 5.9083e-06, 7.7546e-05, 1.4544e-05, + -8.2731e-05, 4.9807e-06, 1.5542e-05, -6.7763e-06, 1.2189e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 259.50, cls_loss 0.0021 cls_loss_mapping 0.0045 cls_loss_causal 0.5395 re_mapping 0.0063 re_causal 0.0180 /// teacc 99.04 lr 0.00010000 +Epoch 194, weight, value: tensor([[ 0.0250, -0.1494, -0.1207, ..., -0.2320, -0.0869, -0.1297], + [ 0.0405, -0.0606, 0.0337, ..., 0.0507, 0.0958, -0.0353], + [-0.0647, 0.1141, -0.1484, ..., 0.0584, 0.0662, -0.0363], + ..., + [-0.0785, -0.0712, -0.0736, ..., 0.0017, -0.1335, 0.1167], + [ 0.0670, -0.0230, 0.0622, ..., -0.0020, -0.1685, -0.0077], + [-0.1418, -0.0456, -0.0617, ..., -0.1575, 0.0426, -0.0931]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 6.1467e-08, ..., 4.6566e-08, + 3.5390e-07, 2.9802e-08], + [-2.6077e-08, 5.5879e-09, -1.4491e-06, ..., 1.0170e-06, + -1.4286e-06, 1.0300e-06], + [ 1.8626e-09, -6.1467e-08, 8.5682e-08, ..., 1.8999e-07, + -3.5390e-08, 2.4214e-07], + ..., + [ 1.8626e-09, 2.7940e-08, 2.4214e-07, ..., -6.5751e-06, + 4.4145e-07, -6.1579e-06], + [-3.7253e-09, 7.4506e-09, 2.5891e-07, ..., 9.6858e-08, + 5.6438e-07, 3.9116e-08], + [ 3.7253e-09, 1.8626e-09, 1.2852e-07, ..., 5.1335e-06, + 1.1995e-06, 4.6715e-06]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0046, -0.0325, 0.0063, -0.0186, 0.0164, 0.0093, 0.0207, 0.0004, + -0.0344, -0.0080], device='cuda:0'), grad: tensor([ 4.7684e-07, 2.3358e-06, 1.0971e-06, 1.6950e-07, -2.1961e-06, + 1.8310e-06, -1.5013e-06, -2.8744e-05, 1.4771e-06, 2.5049e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 259.52, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5461 re_mapping 0.0061 re_causal 0.0184 /// teacc 98.99 lr 0.00010000 +Epoch 195, weight, value: tensor([[ 0.0249, -0.1507, -0.1219, ..., -0.2343, -0.0882, -0.1299], + [ 0.0404, -0.0609, 0.0340, ..., 0.0507, 0.0959, -0.0354], + [-0.0647, 0.1148, -0.1488, ..., 0.0586, 0.0665, -0.0364], + ..., + [-0.0785, -0.0717, -0.0737, ..., 0.0017, -0.1340, 0.1170], + [ 0.0670, -0.0227, 0.0625, ..., -0.0019, -0.1693, -0.0078], + [-0.1419, -0.0468, -0.0624, ..., -0.1582, 0.0433, -0.0937]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.1665e-08, 4.0419e-07, ..., 7.2457e-07, + 1.6261e-06, 3.7253e-09], + [ 0.0000e+00, 7.4506e-09, 1.4052e-05, ..., 2.5839e-05, + 5.3018e-05, -5.0291e-08], + [ 0.0000e+00, -1.7323e-07, -2.0504e-05, ..., -3.6865e-05, + -7.6711e-05, 2.6077e-08], + ..., + [ 0.0000e+00, 2.2352e-08, 4.2468e-07, ..., 4.5635e-07, + 1.2163e-06, -1.8254e-07], + [ 0.0000e+00, 6.5193e-08, 2.3097e-07, ..., 4.4331e-07, + 8.9221e-07, 1.8626e-08], + [ 0.0000e+00, 1.3039e-08, 1.1362e-06, ..., 2.1402e-06, + 3.8855e-06, 8.7544e-08]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0053, -0.0326, 0.0067, -0.0200, 0.0162, 0.0099, 0.0206, 0.0006, + -0.0342, -0.0076], device='cuda:0'), grad: tensor([ 2.8349e-06, 9.0539e-05, -1.3006e-04, 3.1710e-05, 4.4107e-06, + -1.7226e-05, 7.6592e-06, 1.6354e-06, 2.3991e-06, 5.9977e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 259.17, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.5419 re_mapping 0.0060 re_causal 0.0183 /// teacc 98.97 lr 0.00010000 +Epoch 196, weight, value: tensor([[ 0.0249, -0.1510, -0.1221, ..., -0.2356, -0.0883, -0.1299], + [ 0.0404, -0.0612, 0.0337, ..., 0.0501, 0.0951, -0.0358], + [-0.0648, 0.1153, -0.1482, ..., 0.0592, 0.0675, -0.0366], + ..., + [-0.0786, -0.0734, -0.0735, ..., 0.0025, -0.1339, 0.1177], + [ 0.0670, -0.0207, 0.0627, ..., -0.0017, -0.1699, -0.0078], + [-0.1419, -0.0475, -0.0627, ..., -0.1589, 0.0435, -0.0940]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.5390e-07, 1.0803e-07, ..., 5.1968e-07, + 8.8476e-07, 0.0000e+00], + [ 0.0000e+00, 5.3085e-07, -4.0978e-07, ..., 1.2405e-06, + 1.5926e-06, 0.0000e+00], + [ 0.0000e+00, -1.9893e-06, 2.7567e-07, ..., -1.9222e-06, + -5.4687e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 2.4587e-07, 3.8370e-07, ..., -2.6431e-06, + 1.0002e-06, 0.0000e+00], + [ 0.0000e+00, 3.5390e-07, 1.2107e-07, ..., 1.5125e-06, + 9.4064e-07, 0.0000e+00], + [ 0.0000e+00, 1.5274e-07, 1.3970e-07, ..., 4.5262e-07, + -5.0664e-07, 0.0000e+00]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0053, -0.0334, 0.0075, -0.0213, 0.0159, 0.0104, 0.0205, 0.0013, + -0.0340, -0.0077], device='cuda:0'), grad: tensor([ 1.9893e-06, 3.0417e-06, -6.1244e-06, 9.4250e-07, 7.7859e-07, + 4.5858e-06, 5.5879e-08, -8.9854e-06, 5.5209e-06, -1.8738e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 259.59, cls_loss 0.0020 cls_loss_mapping 0.0039 cls_loss_causal 0.5290 re_mapping 0.0065 re_causal 0.0182 /// teacc 98.89 lr 0.00010000 +Epoch 197, weight, value: tensor([[ 0.0249, -0.1512, -0.1224, ..., -0.2365, -0.0884, -0.1299], + [ 0.0405, -0.0617, 0.0339, ..., 0.0498, 0.0952, -0.0361], + [-0.0649, 0.1203, -0.1483, ..., 0.0617, 0.0687, -0.0367], + ..., + [-0.0786, -0.0776, -0.0735, ..., 0.0019, -0.1357, 0.1183], + [ 0.0669, -0.0218, 0.0628, ..., -0.0021, -0.1708, -0.0078], + [-0.1421, -0.0522, -0.0635, ..., -0.1602, 0.0437, -0.0942]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.7055e-08, 8.5682e-08, ..., 2.3469e-07, + 3.6880e-07, 7.4506e-09], + [ 1.8626e-09, -9.8050e-06, 5.1409e-07, ..., -1.6585e-05, + -4.8041e-05, 6.3330e-08], + [ 1.8626e-09, 1.1474e-06, 3.4142e-06, ..., 1.2502e-05, + 2.7612e-05, 1.1735e-07], + ..., + [ 3.7253e-09, 6.6832e-06, -1.1347e-05, ..., -1.1094e-05, + 1.0811e-05, -3.8743e-07], + [ 1.1176e-08, 5.5879e-08, 3.4813e-06, ..., 4.6007e-06, + 3.2596e-07, 4.2841e-08], + [ 0.0000e+00, 2.4959e-07, 3.0696e-06, ..., 5.9716e-06, + 1.3690e-06, 1.2852e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0053, -0.0336, 0.0103, -0.0251, 0.0156, 0.0126, 0.0205, 0.0005, + -0.0344, -0.0073], device='cuda:0'), grad: tensor([ 8.2701e-07, -4.8578e-05, 4.1723e-05, 2.4691e-05, 1.5888e-06, + 1.3411e-07, -1.3269e-05, -4.3452e-05, 1.5959e-05, 2.0355e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 259.47, cls_loss 0.0022 cls_loss_mapping 0.0038 cls_loss_causal 0.5298 re_mapping 0.0062 re_causal 0.0184 /// teacc 98.93 lr 0.00010000 +Epoch 198, weight, value: tensor([[ 0.0248, -0.1513, -0.1221, ..., -0.2379, -0.0885, -0.1300], + [ 0.0405, -0.0614, 0.0332, ..., 0.0491, 0.0948, -0.0363], + [-0.0646, 0.1226, -0.1484, ..., 0.0626, 0.0700, -0.0369], + ..., + [-0.0780, -0.0787, -0.0727, ..., 0.0025, -0.1357, 0.1186], + [ 0.0664, -0.0242, 0.0622, ..., -0.0035, -0.1737, -0.0080], + [-0.1424, -0.0565, -0.0657, ..., -0.1617, 0.0434, -0.0944]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0617e-07, ..., 8.5682e-08, + 1.2666e-07, 0.0000e+00], + [-1.0245e-07, 1.4901e-08, -5.2601e-05, ..., -8.1956e-05, + -6.4850e-05, 5.5879e-09], + [ 4.6566e-08, 1.4715e-07, 2.2247e-05, ..., 3.5316e-05, + 2.7165e-05, 4.0419e-07], + ..., + [ 5.5879e-08, -1.6764e-07, 2.8744e-05, ..., 4.3303e-05, + 3.4988e-05, -4.2096e-07], + [ 3.7253e-09, 1.8626e-09, 1.6838e-06, ..., 1.9185e-07, + 3.2224e-06, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 3.5204e-07, ..., 3.5949e-07, + -5.6624e-07, 3.7253e-09]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0043, -0.0345, 0.0116, -0.0244, 0.0160, 0.0124, 0.0200, 0.0010, + -0.0361, -0.0078], device='cuda:0'), grad: tensor([ 1.4715e-07, -1.4603e-04, 6.3419e-05, -2.7940e-08, 2.6450e-06, + 2.4401e-06, -6.5118e-06, 7.8022e-05, 6.1542e-06, -2.8126e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 259.81, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5273 re_mapping 0.0059 re_causal 0.0173 /// teacc 99.02 lr 0.00010000 +Epoch 199, weight, value: tensor([[ 0.0249, -0.1513, -0.1221, ..., -0.2391, -0.0890, -0.1300], + [ 0.0406, -0.0616, 0.0325, ..., 0.0483, 0.0949, -0.0364], + [-0.0647, 0.1228, -0.1491, ..., 0.0623, 0.0700, -0.0372], + ..., + [-0.0781, -0.0788, -0.0717, ..., 0.0035, -0.1353, 0.1188], + [ 0.0665, -0.0242, 0.0623, ..., -0.0037, -0.1745, -0.0081], + [-0.1425, -0.0569, -0.0682, ..., -0.1626, 0.0441, -0.0948]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.7497e-08, 1.8999e-07, ..., 4.9546e-07, + 4.0419e-07, 7.4506e-09], + [ 0.0000e+00, 4.2841e-08, -3.1665e-08, ..., 3.9116e-07, + 1.3690e-07, 1.4994e-07], + [ 0.0000e+00, -5.1875e-07, 1.6550e-06, ..., -2.8498e-06, + -2.5518e-06, 4.7497e-08], + ..., + [ 0.0000e+00, 2.0768e-07, 1.2107e-07, ..., -2.5146e-08, + 6.3144e-07, -4.6659e-07], + [ 0.0000e+00, 1.0524e-07, -2.8461e-06, ..., 5.2061e-07, + 1.6857e-07, 4.9360e-08], + [ 0.0000e+00, 1.2107e-08, 1.1083e-07, ..., 2.5332e-07, + 1.8068e-07, 1.6112e-07]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0046, -0.0352, 0.0112, -0.0214, 0.0147, 0.0098, 0.0200, 0.0018, + -0.0366, -0.0069], device='cuda:0'), grad: tensor([ 1.4361e-06, 1.2908e-06, -3.5688e-06, 2.0862e-06, 2.1793e-07, + -1.5181e-06, 2.7046e-06, -4.5914e-07, -2.7362e-06, 4.7591e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 259.66, cls_loss 0.0018 cls_loss_mapping 0.0028 cls_loss_causal 0.4975 re_mapping 0.0058 re_causal 0.0169 /// teacc 98.90 lr 0.00010000 +Epoch 200, weight, value: tensor([[ 0.0249, -0.1515, -0.1227, ..., -0.2402, -0.0921, -0.1300], + [ 0.0406, -0.0618, 0.0329, ..., 0.0488, 0.0955, -0.0363], + [-0.0652, 0.1227, -0.1497, ..., 0.0621, 0.0700, -0.0374], + ..., + [-0.0781, -0.0789, -0.0721, ..., 0.0029, -0.1361, 0.1187], + [ 0.0668, -0.0239, 0.0630, ..., -0.0032, -0.1750, -0.0081], + [-0.1425, -0.0571, -0.0688, ..., -0.1630, 0.0455, -0.0950]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7008e-08, ..., 8.6706e-07, + 1.5413e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3935e-07, ..., 4.2468e-07, + 4.6659e-07, 7.4506e-09], + [ 0.0000e+00, -4.0978e-08, 9.3225e-07, ..., -7.6711e-05, + -1.1706e-04, 1.5832e-08], + ..., + [ 0.0000e+00, 3.5390e-08, 3.4645e-07, ..., 2.2873e-06, + 3.3602e-06, -2.8871e-08], + [ 0.0000e+00, 0.0000e+00, -8.4192e-07, ..., -8.5682e-08, + 8.1025e-07, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.2003e-07, ..., 2.0768e-07, + 1.1222e-06, 1.8626e-09]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0073, -0.0347, 0.0108, -0.0213, 0.0147, 0.0098, 0.0208, 0.0014, + -0.0363, -0.0055], device='cuda:0'), grad: tensor([-4.5672e-06, 1.5246e-06, -1.9717e-04, 1.8859e-04, -4.2841e-06, + 5.3085e-07, 1.0999e-06, 6.5118e-06, -1.4622e-07, 7.6182e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 256.34, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5360 re_mapping 0.0058 re_causal 0.0177 /// teacc 98.95 lr 0.00010000 +Epoch 201, weight, value: tensor([[ 0.0253, -0.1518, -0.1234, ..., -0.2424, -0.0924, -0.1300], + [ 0.0407, -0.0622, 0.0332, ..., 0.0488, 0.0957, -0.0364], + [-0.0657, 0.1230, -0.1511, ..., 0.0628, 0.0702, -0.0375], + ..., + [-0.0773, -0.0794, -0.0722, ..., 0.0029, -0.1373, 0.1187], + [ 0.0668, -0.0229, 0.0637, ..., -0.0027, -0.1748, -0.0081], + [-0.1431, -0.0583, -0.0690, ..., -0.1640, 0.0463, -0.0951]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.1327e-07, 1.3039e-07, ..., 4.0885e-07, + 2.8368e-06, 0.0000e+00], + [ 0.0000e+00, 1.4901e-08, 1.5832e-08, ..., 9.5926e-08, + -1.2359e-06, 6.5193e-09], + [ 0.0000e+00, -6.7521e-07, 1.2172e-06, ..., -3.0641e-07, + -1.2126e-06, 2.7940e-09], + ..., + [ 0.0000e+00, 1.0245e-07, 1.5339e-06, ..., 1.1278e-06, + 1.7034e-06, -1.7695e-08], + [ 0.0000e+00, 1.4249e-07, 7.3574e-08, ..., 3.9395e-07, + 1.0151e-06, 9.3132e-10], + [ 0.0000e+00, 1.2107e-08, 6.5006e-07, ..., 4.5914e-07, + -1.8915e-06, 5.5879e-09]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0074, -0.0346, 0.0109, -0.0216, 0.0145, 0.0097, 0.0212, 0.0011, + -0.0358, -0.0046], device='cuda:0'), grad: tensor([ 6.6124e-06, -7.9162e-07, 9.9838e-07, -1.4104e-05, 6.6049e-06, + 3.7216e-06, -9.1270e-06, 5.9903e-06, 2.2873e-06, -2.2091e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 253.65, cls_loss 0.0025 cls_loss_mapping 0.0035 cls_loss_causal 0.5354 re_mapping 0.0059 re_causal 0.0181 /// teacc 98.97 lr 0.00010000 +Epoch 202, weight, value: tensor([[ 0.0256, -0.1517, -0.1248, ..., -0.2433, -0.0955, -0.1300], + [ 0.0408, -0.0623, 0.0336, ..., 0.0491, 0.0962, -0.0364], + [-0.0659, 0.1230, -0.1517, ..., 0.0626, 0.0701, -0.0378], + ..., + [-0.0771, -0.0795, -0.0724, ..., 0.0027, -0.1380, 0.1188], + [ 0.0668, -0.0223, 0.0641, ..., -0.0025, -0.1749, -0.0081], + [-0.1438, -0.0585, -0.0696, ..., -0.1643, 0.0484, -0.0952]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 1.3039e-08, -2.8200e-06, ..., 3.4180e-07, + 2.9709e-07, 9.3132e-10], + [-9.3132e-10, 3.7253e-09, 1.6028e-06, ..., 9.5274e-07, + 3.9488e-07, 1.3411e-07], + [ 3.4831e-07, -1.0245e-07, 1.5043e-05, ..., 5.8301e-06, + 4.1313e-06, 1.7695e-08], + ..., + [ 8.0094e-08, 3.2596e-08, 3.4682e-06, ..., 1.2862e-06, + 1.5106e-06, -2.6356e-07], + [ 9.3132e-08, 3.4459e-08, 3.6489e-06, ..., 1.5637e-06, + 1.3849e-06, 4.8429e-08], + [ 1.2107e-08, 6.5193e-09, 1.2908e-06, ..., 3.3807e-07, + -2.6524e-05, 2.7008e-08]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0101, -0.0344, 0.0106, -0.0217, 0.0170, 0.0099, 0.0212, 0.0009, + -0.0356, -0.0028], device='cuda:0'), grad: tensor([-3.3945e-05, 5.1968e-06, 3.9399e-05, -1.2422e-04, 9.8050e-05, + 5.9277e-05, 2.2858e-05, 9.4622e-06, 1.0803e-05, -8.6725e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 253.69, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.5475 re_mapping 0.0057 re_causal 0.0178 /// teacc 99.04 lr 0.00010000 +Epoch 203, weight, value: tensor([[ 0.0256, -0.1498, -0.1251, ..., -0.2430, -0.0955, -0.1301], + [ 0.0410, -0.0624, 0.0336, ..., 0.0490, 0.0960, -0.0365], + [-0.0659, 0.1232, -0.1519, ..., 0.0629, 0.0705, -0.0379], + ..., + [-0.0771, -0.0797, -0.0725, ..., 0.0027, -0.1384, 0.1190], + [ 0.0668, -0.0225, 0.0645, ..., -0.0025, -0.1753, -0.0082], + [-0.1440, -0.0591, -0.0700, ..., -0.1647, 0.0484, -0.0954]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.5274e-07, 9.3132e-09, ..., 1.2573e-07, + 1.3504e-07, 9.3132e-10], + [-2.1420e-08, 1.0151e-07, -6.5193e-09, ..., 3.1106e-07, + 2.7567e-07, 2.0489e-08], + [-1.5832e-08, -4.9919e-07, 4.2841e-08, ..., -1.1241e-06, + -1.2303e-06, 6.5193e-09], + ..., + [ 1.4901e-08, 1.4342e-07, 7.1712e-08, ..., 2.6356e-07, + 3.2224e-07, -1.7323e-07], + [-8.3819e-09, 1.8626e-08, -2.8592e-07, ..., -1.1735e-07, + 1.1176e-08, 5.5879e-09], + [ 2.7940e-09, 7.4506e-09, 8.1956e-08, ..., 5.8673e-08, + 2.6077e-08, 7.5437e-08]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0099, -0.0346, 0.0108, -0.0215, 0.0168, 0.0097, 0.0212, 0.0010, + -0.0354, -0.0029], device='cuda:0'), grad: tensor([ 2.2650e-06, 7.4599e-07, -2.9206e-06, 7.5437e-07, 3.1013e-07, + 2.6356e-07, -2.1458e-06, 3.0175e-07, -4.6194e-07, 8.8289e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 253.96, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.4946 re_mapping 0.0059 re_causal 0.0184 /// teacc 98.92 lr 0.00010000 +Epoch 204, weight, value: tensor([[ 0.0259, -0.1494, -0.1253, ..., -0.2434, -0.0955, -0.1301], + [ 0.0411, -0.0630, 0.0338, ..., 0.0491, 0.0962, -0.0365], + [-0.0660, 0.1235, -0.1523, ..., 0.0630, 0.0709, -0.0384], + ..., + [-0.0771, -0.0800, -0.0727, ..., 0.0022, -0.1392, 0.1187], + [ 0.0668, -0.0228, 0.0647, ..., -0.0026, -0.1757, -0.0082], + [-0.1443, -0.0592, -0.0705, ..., -0.1649, 0.0484, -0.0955]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 3.7253e-09, + 4.0978e-08, 0.0000e+00], + [-4.6566e-09, 0.0000e+00, -3.5856e-07, ..., -4.8429e-08, + -3.5856e-07, 1.9558e-08], + [ 0.0000e+00, -9.3132e-10, 2.6077e-08, ..., 2.5146e-08, + 4.8708e-07, 6.5193e-09], + ..., + [ 9.3132e-10, 9.3132e-10, 1.3877e-07, ..., -5.1223e-08, + 1.0496e-06, -5.6811e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 1.7695e-08, + 1.0617e-07, 3.7253e-09], + [ 2.7940e-09, 0.0000e+00, 1.4715e-07, ..., 5.5879e-08, + 8.4471e-07, 1.8626e-08]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0097, -0.0345, 0.0108, -0.0212, 0.0169, 0.0097, 0.0210, 0.0006, + -0.0355, -0.0031], device='cuda:0'), grad: tensor([ 1.1269e-07, -7.1619e-07, 1.5404e-06, -6.5193e-09, -7.4767e-06, + 1.1735e-07, 1.0990e-07, 3.0473e-06, 4.0233e-07, 2.8610e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 253.63, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.5163 re_mapping 0.0055 re_causal 0.0173 /// teacc 98.96 lr 0.00010000 +Epoch 205, weight, value: tensor([[ 0.0259, -0.1495, -0.1251, ..., -0.2439, -0.0954, -0.1301], + [ 0.0413, -0.0632, 0.0336, ..., 0.0488, 0.0951, -0.0366], + [-0.0660, 0.1237, -0.1508, ..., 0.0639, 0.0730, -0.0385], + ..., + [-0.0771, -0.0801, -0.0730, ..., 0.0020, -0.1402, 0.1189], + [ 0.0668, -0.0229, 0.0653, ..., -0.0026, -0.1761, -0.0082], + [-0.1445, -0.0594, -0.0715, ..., -0.1654, 0.0484, -0.0956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1176e-07, 4.5635e-08, ..., 1.1642e-07, + 2.6170e-07, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, -2.1793e-07, ..., -5.4017e-08, + -2.4587e-07, 1.3970e-08], + [ 0.0000e+00, -1.4808e-07, 8.8476e-08, ..., -3.2596e-08, + -2.4494e-07, 2.4214e-08], + ..., + [ 0.0000e+00, 1.9558e-08, 1.0617e-07, ..., -8.6613e-08, + 1.4435e-07, -7.6368e-08], + [ 0.0000e+00, 4.6566e-09, 6.0536e-08, ..., 3.3528e-08, + 2.8871e-08, 3.7253e-09], + [ 0.0000e+00, 2.7940e-09, 5.9605e-08, ..., 6.8918e-08, + 1.1921e-07, 2.7940e-08]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0094, -0.0350, 0.0124, -0.0214, 0.0166, 0.0099, 0.0204, 0.0004, + -0.0355, -0.0032], device='cuda:0'), grad: tensor([ 3.2131e-07, -2.0582e-07, -8.5682e-08, -1.0589e-06, -2.9523e-07, + 8.0466e-07, 1.3039e-07, -4.7591e-07, 1.6578e-07, 7.0874e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 253.58, cls_loss 0.0012 cls_loss_mapping 0.0029 cls_loss_causal 0.5072 re_mapping 0.0058 re_causal 0.0181 /// teacc 98.95 lr 0.00010000 +Epoch 206, weight, value: tensor([[ 0.0261, -0.1495, -0.1240, ..., -0.2443, -0.0954, -0.1301], + [ 0.0412, -0.0634, 0.0329, ..., 0.0480, 0.0940, -0.0369], + [-0.0664, 0.1239, -0.1506, ..., 0.0641, 0.0737, -0.0387], + ..., + [-0.0772, -0.0801, -0.0724, ..., 0.0028, -0.1392, 0.1192], + [ 0.0671, -0.0230, 0.0654, ..., -0.0024, -0.1767, -0.0082], + [-0.1447, -0.0596, -0.0722, ..., -0.1662, 0.0482, -0.0959]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 2.3283e-08, + 5.2899e-07, 5.4948e-08], + [ 0.0000e+00, 0.0000e+00, -6.1020e-06, ..., -6.1207e-06, + -1.3359e-05, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 1.3243e-06, ..., 1.2349e-06, + 2.1905e-06, 1.5832e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 5.1558e-06, ..., 4.8876e-06, + 1.0632e-05, -4.3772e-08], + [ 0.0000e+00, 0.0000e+00, 3.1590e-06, ..., 1.6410e-06, + 1.5367e-07, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 2.3190e-07, ..., 2.4587e-07, + 5.9642e-06, 1.0245e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0093, -0.0360, 0.0128, -0.0214, 0.0170, 0.0100, 0.0201, 0.0013, + -0.0356, -0.0036], device='cuda:0'), grad: tensor([ 2.0918e-06, -2.7314e-05, 5.2191e-06, -8.4117e-06, -1.2018e-05, + 2.6375e-06, -2.7567e-06, 2.1830e-05, 5.3570e-06, 1.3292e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 253.65, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.5155 re_mapping 0.0058 re_causal 0.0173 /// teacc 98.95 lr 0.00010000 +Epoch 207, weight, value: tensor([[ 0.0268, -0.1496, -0.1241, ..., -0.2451, -0.0954, -0.1301], + [ 0.0411, -0.0636, 0.0333, ..., 0.0482, 0.0945, -0.0369], + [-0.0680, 0.1240, -0.1518, ..., 0.0635, 0.0734, -0.0388], + ..., + [-0.0776, -0.0801, -0.0726, ..., 0.0027, -0.1394, 0.1190], + [ 0.0687, -0.0230, 0.0656, ..., -0.0020, -0.1773, -0.0082], + [-0.1455, -0.0598, -0.0720, ..., -0.1665, 0.0481, -0.0960]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, 6.2399e-08, ..., 1.7695e-08, + 1.0524e-07, 4.6566e-09], + [ 1.8347e-07, 2.7940e-09, 4.0326e-07, ..., 1.9558e-07, + 1.3597e-07, 4.3772e-08], + [ 1.2759e-07, -1.0245e-08, 4.3027e-07, ..., 1.8813e-07, + 2.5798e-07, 3.2596e-08], + ..., + [ 8.1770e-07, 5.5879e-09, 3.9563e-06, ..., 1.8440e-07, + 3.8370e-06, -2.8312e-07], + [-4.3027e-07, 1.8626e-09, -8.6054e-07, ..., -8.3353e-07, + -8.3819e-08, 4.7497e-08], + [ 4.2096e-07, 0.0000e+00, 2.1402e-06, ..., 7.0781e-08, + 9.0990e-07, 8.7544e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0091, -0.0359, 0.0123, -0.0212, 0.0172, 0.0099, 0.0204, 0.0010, + -0.0355, -0.0037], device='cuda:0'), grad: tensor([-2.9150e-07, 2.0713e-06, 1.7416e-06, 2.3916e-06, -2.7180e-05, + -8.8569e-07, 1.2591e-06, 1.5631e-05, -1.3839e-06, 6.6161e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 253.54, cls_loss 0.0016 cls_loss_mapping 0.0031 cls_loss_causal 0.5150 re_mapping 0.0058 re_causal 0.0176 /// teacc 98.95 lr 0.00010000 +Epoch 208, weight, value: tensor([[ 0.0277, -0.1500, -0.1247, ..., -0.2464, -0.0954, -0.1300], + [ 0.0410, -0.0638, 0.0340, ..., 0.0483, 0.0958, -0.0369], + [-0.0691, 0.1240, -0.1524, ..., 0.0634, 0.0734, -0.0389], + ..., + [-0.0777, -0.0802, -0.0732, ..., 0.0025, -0.1418, 0.1189], + [ 0.0698, -0.0227, 0.0672, ..., -0.0017, -0.1769, -0.0083], + [-0.1470, -0.0599, -0.0726, ..., -0.1668, 0.0481, -0.0960]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -4.6819e-05, ..., 2.7940e-08, + -7.8261e-05, 7.4506e-09], + [ 1.8626e-09, 0.0000e+00, 2.6211e-05, ..., -2.6356e-07, + 4.4256e-05, 6.7055e-08], + [ 1.1176e-08, -1.8626e-09, 9.5293e-06, ..., 1.0990e-07, + 1.5929e-05, 5.4948e-08], + ..., + [ 6.5193e-09, 0.0000e+00, 7.9069e-07, ..., 9.0338e-08, + 7.4320e-07, -1.6093e-06], + [ 8.3819e-09, 0.0000e+00, -8.0466e-07, ..., -1.5832e-07, + 3.0547e-07, 5.6811e-08], + [ 2.7940e-09, 0.0000e+00, 1.1884e-06, ..., 1.9185e-07, + 1.6578e-06, 8.8476e-08]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0091, -0.0346, 0.0121, -0.0211, 0.0173, 0.0100, 0.0196, -0.0002, + -0.0344, -0.0038], device='cuda:0'), grad: tensor([-2.4748e-04, 1.4210e-04, 5.6684e-05, 5.7071e-06, 7.9125e-06, + 3.3323e-06, 3.7611e-05, -1.3642e-05, 3.3341e-07, 7.1228e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 253.76, cls_loss 0.0019 cls_loss_mapping 0.0029 cls_loss_causal 0.5128 re_mapping 0.0055 re_causal 0.0169 /// teacc 98.91 lr 0.00010000 +Epoch 209, weight, value: tensor([[ 0.0282, -0.1497, -0.1230, ..., -0.2485, -0.0956, -0.1300], + [ 0.0378, -0.0638, 0.0341, ..., 0.0478, 0.0948, -0.0370], + [-0.0661, 0.1241, -0.1523, ..., 0.0643, 0.0747, -0.0390], + ..., + [-0.0779, -0.0802, -0.0733, ..., 0.0025, -0.1423, 0.1191], + [ 0.0700, -0.0227, 0.0673, ..., -0.0017, -0.1777, -0.0083], + [-0.1490, -0.0601, -0.0733, ..., -0.1668, 0.0484, -0.0962]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.6322e-08, 6.2399e-08, ..., 7.5437e-08, + 1.7192e-06, 0.0000e+00], + [-2.1420e-08, 7.4506e-09, -8.7544e-08, ..., 7.4506e-09, + -1.9558e-08, 5.5879e-09], + [ 6.5193e-09, -1.2573e-07, 1.7416e-07, ..., -1.2200e-07, + -1.8813e-07, 1.8626e-09], + ..., + [ 4.6566e-09, 3.8184e-08, 1.0431e-07, ..., 7.6368e-08, + 3.5204e-07, -1.3970e-08], + [ 1.8626e-09, -2.7381e-07, -1.4510e-06, ..., -8.2888e-07, + 1.2569e-05, 9.3132e-10], + [ 3.7253e-09, 2.4494e-07, 3.8650e-07, ..., 2.0675e-07, + 2.2128e-06, 5.5879e-09]], device='cuda:0') +Epoch 209, bias, value: tensor([-9.1916e-03, -3.5346e-02, 1.3071e-02, -2.1062e-02, 1.6832e-02, + 1.0018e-02, 1.9548e-02, -2.7881e-05, -3.4831e-02, -3.4739e-03], + device='cuda:0'), grad: tensor([ 3.9265e-06, 4.4703e-07, 5.5972e-07, 9.7230e-07, -5.8636e-06, + 3.3714e-07, -3.5644e-05, 4.2003e-07, 2.6599e-05, 8.1882e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 253.82, cls_loss 0.0019 cls_loss_mapping 0.0027 cls_loss_causal 0.5115 re_mapping 0.0059 re_causal 0.0167 /// teacc 98.97 lr 0.00010000 +Epoch 210, weight, value: tensor([[ 0.0281, -0.1496, -0.1244, ..., -0.2506, -0.0958, -0.1300], + [ 0.0381, -0.0639, 0.0346, ..., 0.0480, 0.0952, -0.0370], + [-0.0660, 0.1241, -0.1527, ..., 0.0642, 0.0747, -0.0391], + ..., + [-0.0781, -0.0802, -0.0735, ..., 0.0023, -0.1430, 0.1191], + [ 0.0706, -0.0228, 0.0677, ..., -0.0015, -0.1789, -0.0083], + [-0.1510, -0.0604, -0.0739, ..., -0.1670, 0.0486, -0.0962]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.5404e-06, ..., 1.7043e-07, + 4.4703e-08, 2.7940e-09], + [-2.7940e-09, 9.3132e-10, 3.2224e-07, ..., 1.0524e-07, + -1.0245e-08, 2.1420e-08], + [ 0.0000e+00, -1.0245e-08, 1.3867e-06, ..., 4.7684e-07, + -1.0245e-08, 1.3132e-07], + ..., + [ 2.7940e-09, 7.4506e-09, 4.0326e-07, ..., -7.1712e-07, + 5.9605e-08, -2.2538e-07], + [ 5.5879e-09, 9.3132e-10, 1.8761e-05, ..., 2.0545e-06, + 8.1025e-08, 5.5879e-09], + [ 4.6566e-09, 0.0000e+00, 8.8960e-06, ..., 1.0412e-06, + -9.0152e-07, 2.5146e-08]], device='cuda:0') +Epoch 210, bias, value: tensor([-9.2448e-03, -3.5078e-02, 1.2827e-02, -2.1133e-02, 1.7217e-02, + 1.0109e-02, 1.9483e-02, -8.8837e-05, -3.5133e-02, -3.5224e-03], + device='cuda:0'), grad: tensor([ 7.3090e-06, 6.8769e-06, 4.2692e-06, -5.5611e-05, 5.4501e-06, + 9.9018e-06, 3.0827e-07, -4.2468e-05, 3.4660e-05, 2.9385e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 253.45, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5425 re_mapping 0.0058 re_causal 0.0173 /// teacc 99.01 lr 0.00010000 +Epoch 211, weight, value: tensor([[ 0.0280, -0.1497, -0.1247, ..., -0.2520, -0.0958, -0.1300], + [ 0.0381, -0.0640, 0.0349, ..., 0.0481, 0.0954, -0.0370], + [-0.0661, 0.1241, -0.1533, ..., 0.0640, 0.0746, -0.0390], + ..., + [-0.0781, -0.0803, -0.0736, ..., 0.0027, -0.1431, 0.1192], + [ 0.0707, -0.0224, 0.0676, ..., -0.0015, -0.1805, -0.0083], + [-0.1515, -0.0604, -0.0745, ..., -0.1688, 0.0483, -0.0963]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 3.6322e-08, ..., 9.3132e-09, + 2.0955e-07, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, -3.4086e-07, ..., -1.0896e-07, + -4.0233e-07, 0.0000e+00], + [ 1.0245e-08, 1.0245e-08, 1.1548e-07, ..., 5.0291e-08, + 1.4901e-07, 0.0000e+00], + ..., + [ 2.4214e-08, 3.3528e-08, 2.0489e-07, ..., -4.6566e-09, + 3.5949e-07, -0.0000e+00], + [ 3.8184e-08, -5.4017e-08, 1.0151e-07, ..., -1.3318e-07, + 1.7509e-06, 0.0000e+00], + [ 1.0990e-07, 9.3132e-10, 5.4017e-08, ..., 1.2852e-07, + 5.2899e-07, 0.0000e+00]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0092, -0.0351, 0.0126, -0.0212, 0.0180, 0.0104, 0.0195, 0.0018, + -0.0360, -0.0050], device='cuda:0'), grad: tensor([-4.8522e-07, -3.6135e-07, 5.9605e-07, 1.9465e-07, -3.7923e-06, + -1.2703e-06, -2.1830e-06, 8.1956e-08, 3.3751e-06, 3.8221e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 253.09, cls_loss 0.0022 cls_loss_mapping 0.0033 cls_loss_causal 0.5148 re_mapping 0.0057 re_causal 0.0170 /// teacc 99.05 lr 0.00010000 +Epoch 212, weight, value: tensor([[ 0.0276, -0.1492, -0.1254, ..., -0.2536, -0.0970, -0.1300], + [ 0.0382, -0.0642, 0.0356, ..., 0.0482, 0.0958, -0.0370], + [-0.0662, 0.1244, -0.1537, ..., 0.0634, 0.0742, -0.0391], + ..., + [-0.0777, -0.0805, -0.0741, ..., 0.0029, -0.1433, 0.1192], + [ 0.0706, -0.0224, 0.0673, ..., -0.0017, -0.1820, -0.0083], + [-0.1523, -0.0610, -0.0750, ..., -0.1694, 0.0492, -0.0963]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-08, 7.6368e-08, ..., 6.2399e-08, + 1.6671e-07, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, -6.5193e-09, ..., 3.0734e-08, + 1.8720e-07, 4.6566e-09], + [ 0.0000e+00, 4.4145e-07, 1.5674e-06, ..., 1.6075e-06, + 2.7008e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 8.3819e-09, 3.5390e-08, ..., 4.3772e-08, + 6.7167e-06, -1.4901e-08], + [ 0.0000e+00, -6.2212e-07, -2.1588e-06, ..., -2.1625e-06, + 7.3574e-08, 1.8626e-09], + [ 0.0000e+00, 1.7695e-08, 7.5437e-08, ..., 7.1712e-08, + 7.2177e-07, 5.5879e-09]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0100, -0.0349, 0.0114, -0.0209, 0.0155, 0.0101, 0.0232, 0.0021, + -0.0367, -0.0045], device='cuda:0'), grad: tensor([ 7.8510e-07, 9.3039e-07, 4.7907e-06, 4.2841e-08, -3.2216e-05, + -9.7789e-08, 5.4855e-07, 2.7880e-05, -6.0089e-06, 3.3490e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 253.06, cls_loss 0.0023 cls_loss_mapping 0.0038 cls_loss_causal 0.5333 re_mapping 0.0058 re_causal 0.0169 /// teacc 98.96 lr 0.00010000 +Epoch 213, weight, value: tensor([[ 0.0280, -0.1492, -0.1266, ..., -0.2554, -0.0970, -0.1302], + [ 0.0384, -0.0645, 0.0344, ..., 0.0481, 0.0943, -0.0371], + [-0.0663, 0.1237, -0.1542, ..., 0.0625, 0.0742, -0.0392], + ..., + [-0.0803, -0.0794, -0.0744, ..., 0.0033, -0.1438, 0.1193], + [ 0.0698, -0.0228, 0.0668, ..., -0.0025, -0.1834, -0.0083], + [-0.1541, -0.0613, -0.0757, ..., -0.1700, 0.0490, -0.0966]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.6764e-08, 9.3132e-09, ..., 4.2841e-08, + 1.1791e-06, 3.7253e-09], + [ 0.0000e+00, 6.5193e-09, -1.0896e-07, ..., 3.4925e-07, + 6.4075e-07, 1.0617e-07], + [ 0.0000e+00, -1.8720e-07, 3.0734e-08, ..., -1.1260e-06, + -1.6941e-06, -2.6077e-07], + ..., + [ 0.0000e+00, 9.8720e-08, 2.9802e-08, ..., 5.9605e-08, + -1.2536e-06, 1.3039e-08], + [-1.8626e-09, 2.1420e-08, -8.8476e-08, ..., -1.1176e-08, + 1.3411e-07, 4.6566e-09], + [ 0.0000e+00, 1.4901e-08, 1.1176e-08, ..., 1.5832e-07, + 1.6941e-06, 2.7940e-09]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0099, -0.0363, 0.0103, -0.0205, 0.0162, 0.0102, 0.0250, 0.0026, + -0.0380, -0.0050], device='cuda:0'), grad: tensor([ 7.8138e-07, 2.2911e-06, -2.1048e-06, 8.7731e-07, 7.4040e-07, + -6.3851e-06, 3.0827e-07, -8.8587e-06, 1.9893e-06, 1.0341e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 253.74, cls_loss 0.0021 cls_loss_mapping 0.0028 cls_loss_causal 0.5151 re_mapping 0.0060 re_causal 0.0173 /// teacc 98.93 lr 0.00010000 +Epoch 214, weight, value: tensor([[ 0.0282, -0.1492, -0.1269, ..., -0.2568, -0.0970, -0.1302], + [ 0.0385, -0.0654, 0.0352, ..., 0.0501, 0.0966, -0.0371], + [-0.0662, 0.1222, -0.1547, ..., 0.0595, 0.0726, -0.0392], + ..., + [-0.0809, -0.0804, -0.0751, ..., 0.0028, -0.1451, 0.1194], + [ 0.0697, -0.0232, 0.0674, ..., -0.0027, -0.1845, -0.0083], + [-0.1547, -0.0615, -0.0762, ..., -0.1700, 0.0490, -0.0966]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.0361e-07, 2.7940e-09, ..., 8.2422e-07, + 6.9477e-07, 0.0000e+00], + [-4.6566e-09, 2.6431e-06, -1.0990e-07, ..., 9.0674e-06, + 1.1034e-05, 0.0000e+00], + [ 0.0000e+00, -5.0068e-06, 1.3039e-08, ..., -1.6674e-05, + -1.9565e-05, 0.0000e+00], + ..., + [ 1.8626e-09, 3.7067e-07, 4.1910e-08, ..., 9.1828e-07, + 7.9535e-07, -9.3132e-10], + [-8.8476e-08, 1.2340e-06, -3.7812e-07, ..., 3.2410e-06, + 2.8275e-06, 0.0000e+00], + [ 9.3132e-10, 3.0734e-08, 1.0245e-08, ..., 8.8755e-07, + 2.2743e-06, 0.0000e+00]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0098, -0.0344, 0.0077, -0.0198, 0.0160, 0.0102, 0.0250, 0.0018, + -0.0379, -0.0049], device='cuda:0'), grad: tensor([ 2.2445e-06, 2.5034e-05, -4.6551e-05, 1.4035e-06, 7.0687e-07, + 1.0338e-06, 2.6338e-06, 2.5015e-06, 8.2627e-06, 2.7586e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 253.45, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.5323 re_mapping 0.0057 re_causal 0.0170 /// teacc 98.94 lr 0.00010000 +Epoch 215, weight, value: tensor([[ 0.0282, -0.1489, -0.1271, ..., -0.2588, -0.0970, -0.1303], + [ 0.0386, -0.0660, 0.0355, ..., 0.0509, 0.0975, -0.0372], + [-0.0662, 0.1229, -0.1550, ..., 0.0591, 0.0721, -0.0392], + ..., + [-0.0811, -0.0810, -0.0754, ..., 0.0025, -0.1458, 0.1195], + [ 0.0697, -0.0239, 0.0682, ..., -0.0022, -0.1851, -0.0083], + [-0.1549, -0.0621, -0.0768, ..., -0.1711, 0.0489, -0.0967]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 5.8673e-08, ..., 5.9605e-08, + -1.0975e-05, 1.8626e-09], + [ 0.0000e+00, 1.1176e-08, 1.7229e-07, ..., 9.7696e-07, + -1.0151e-07, 1.2573e-07], + [ 0.0000e+00, 5.2154e-08, 1.3690e-07, ..., 1.2010e-05, + -8.2888e-07, 1.6904e-06], + ..., + [ 0.0000e+00, -8.5682e-08, 3.9395e-07, ..., -1.5453e-05, + 2.3562e-07, -2.0918e-06], + [ 0.0000e+00, 1.2107e-08, 2.9989e-07, ..., 1.4128e-06, + 1.1176e-07, 1.6950e-07], + [ 9.3132e-10, 9.3132e-10, 5.7742e-08, ..., 3.2783e-07, + 1.1034e-05, 1.7695e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0097, -0.0336, 0.0073, -0.0201, 0.0160, 0.0102, 0.0251, 0.0015, + -0.0377, -0.0052], device='cuda:0'), grad: tensor([-5.4240e-05, 1.3195e-05, 2.4050e-05, -4.5933e-06, 4.4592e-06, + 7.1377e-06, -3.9395e-07, -7.1347e-05, 3.7886e-06, 7.7903e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 251.07, cls_loss 0.0029 cls_loss_mapping 0.0046 cls_loss_causal 0.5478 re_mapping 0.0059 re_causal 0.0169 /// teacc 99.03 lr 0.00010000 +Epoch 216, weight, value: tensor([[ 0.0281, -0.1490, -0.1273, ..., -0.2607, -0.0971, -0.1307], + [ 0.0387, -0.0678, 0.0328, ..., 0.0488, 0.0948, -0.0373], + [-0.0662, 0.1232, -0.1556, ..., 0.0589, 0.0720, -0.0394], + ..., + [-0.0814, -0.0810, -0.0754, ..., 0.0027, -0.1460, 0.1197], + [ 0.0696, -0.0239, 0.0687, ..., -0.0017, -0.1859, -0.0083], + [-0.1556, -0.0624, -0.0781, ..., -0.1716, 0.0489, -0.0967]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 2.7940e-09, 3.8780e-06, ..., 1.8887e-06, + 8.4750e-08, 7.4506e-09], + [-2.6580e-06, 1.8626e-09, -6.9700e-06, ..., -5.3719e-06, + -3.2663e-05, 2.0210e-07], + [ 1.2945e-07, -6.2399e-08, 1.0161e-06, ..., 1.4864e-06, + 1.4771e-06, 2.9895e-07], + ..., + [ 9.3132e-10, 1.5832e-08, 1.2964e-06, ..., -1.5814e-06, + 5.0291e-08, -9.5833e-07], + [ 1.0245e-08, 3.0734e-08, 1.3560e-06, ..., 1.2387e-06, + 1.9278e-07, 2.6543e-07], + [ 9.3132e-10, 1.8626e-09, 2.0228e-06, ..., 9.8441e-07, + 8.0094e-08, 5.3085e-08]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0095, -0.0364, 0.0069, -0.0198, 0.0162, 0.0101, 0.0280, 0.0017, + -0.0377, -0.0055], device='cuda:0'), grad: tensor([ 9.2387e-06, -4.0948e-05, 6.8434e-06, -2.4185e-05, 3.9116e-08, + -9.2015e-06, 5.2214e-05, -6.4746e-06, 6.4895e-06, 5.9344e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 248.37, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.4911 re_mapping 0.0057 re_causal 0.0167 /// teacc 99.01 lr 0.00010000 +Epoch 217, weight, value: tensor([[ 0.0282, -0.1488, -0.1279, ..., -0.2624, -0.0971, -0.1308], + [ 0.0389, -0.0680, 0.0330, ..., 0.0487, 0.0950, -0.0374], + [-0.0662, 0.1234, -0.1561, ..., 0.0590, 0.0721, -0.0394], + ..., + [-0.0813, -0.0812, -0.0756, ..., 0.0027, -0.1465, 0.1198], + [ 0.0695, -0.0239, 0.0702, ..., -0.0019, -0.1877, -0.0084], + [-0.1558, -0.0627, -0.0785, ..., -0.1717, 0.0489, -0.0968]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 5.5414e-07, 5.5879e-09, ..., 6.3144e-07, + 8.8289e-07, 3.7253e-09], + [ 9.3132e-10, 8.3819e-09, -3.3993e-07, ..., 9.4064e-08, + -2.8219e-07, 3.6322e-08], + [-6.8918e-08, -8.3540e-07, 1.8626e-08, ..., -1.3271e-06, + -1.8943e-06, 6.5193e-09], + ..., + [ 1.8626e-09, 1.2107e-08, 1.5926e-07, ..., -1.1735e-07, + 3.2876e-07, -5.4017e-08], + [ 2.7940e-09, 1.3039e-08, 1.1176e-08, ..., 2.8871e-08, + 8.3819e-08, 8.3819e-09], + [ 1.8626e-09, 3.6322e-08, 9.4064e-08, ..., 8.1956e-08, + 2.0117e-07, 1.1176e-08]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0094, -0.0363, 0.0069, -0.0193, 0.0165, 0.0099, 0.0277, 0.0017, + -0.0372, -0.0057], device='cuda:0'), grad: tensor([ 1.8487e-06, -1.1735e-07, -4.7348e-06, 1.9018e-06, -4.9826e-07, + -3.7067e-07, 2.1234e-07, -1.4622e-07, 3.0827e-07, 1.5954e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 216---------------------------------------------------- +epoch 216, time 264.90, cls_loss 0.0017 cls_loss_mapping 0.0027 cls_loss_causal 0.4842 re_mapping 0.0053 re_causal 0.0157 /// teacc 99.10 lr 0.00010000 +Epoch 218, weight, value: tensor([[ 0.0281, -0.1492, -0.1261, ..., -0.2637, -0.0967, -0.1289], + [ 0.0397, -0.0710, 0.0330, ..., 0.0487, 0.0949, -0.0375], + [-0.0662, 0.1243, -0.1560, ..., 0.0591, 0.0725, -0.0387], + ..., + [-0.0816, -0.0813, -0.0757, ..., 0.0027, -0.1468, 0.1200], + [ 0.0676, -0.0242, 0.0696, ..., -0.0019, -0.1895, -0.0085], + [-0.1567, -0.0630, -0.0790, ..., -0.1720, 0.0488, -0.0970]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.7300e-08, 1.6764e-08, ..., 9.5926e-08, + 9.7789e-08, 2.7940e-09], + [ 0.0000e+00, 1.6671e-07, -6.0536e-08, ..., 2.1607e-07, + -1.3970e-08, 2.6077e-08], + [ 9.3132e-10, 4.2934e-07, 1.7323e-07, ..., 5.3365e-07, + -7.5437e-08, 2.7940e-09], + ..., + [ 0.0000e+00, -3.5614e-06, 3.9116e-08, ..., -3.3341e-06, + -1.5637e-06, -7.3574e-08], + [ 0.0000e+00, 1.3039e-08, -5.9325e-07, ..., -4.9733e-07, + 4.1910e-08, 4.6566e-09], + [ 9.3132e-10, 1.9558e-06, 1.8161e-07, ..., 1.9204e-06, + 6.5565e-07, 2.9802e-08]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0086, -0.0364, 0.0073, -0.0192, 0.0173, 0.0101, 0.0274, 0.0017, + -0.0383, -0.0060], device='cuda:0'), grad: tensor([-5.8394e-07, 8.5216e-07, 2.3246e-06, 3.9116e-07, 5.0291e-06, + 5.8394e-07, 3.0920e-07, -1.3016e-05, -1.5125e-06, 5.6364e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 247.69, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5140 re_mapping 0.0055 re_causal 0.0173 /// teacc 98.97 lr 0.00010000 +Epoch 219, weight, value: tensor([[ 0.0281, -0.1465, -0.1256, ..., -0.2647, -0.0967, -0.1291], + [ 0.0397, -0.0712, 0.0331, ..., 0.0487, 0.0951, -0.0377], + [-0.0662, 0.1244, -0.1563, ..., 0.0592, 0.0725, -0.0387], + ..., + [-0.0816, -0.0814, -0.0758, ..., 0.0027, -0.1472, 0.1202], + [ 0.0675, -0.0241, 0.0697, ..., -0.0020, -0.1901, -0.0086], + [-0.1567, -0.0654, -0.0793, ..., -0.1719, 0.0487, -0.0972]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8487e-07, 2.0955e-08, ..., 8.1956e-08, + 1.8999e-07, 0.0000e+00], + [-2.7940e-09, 6.5193e-08, -1.7034e-06, ..., -8.3400e-07, + -3.2447e-06, 4.6566e-09], + [ 9.3132e-10, -3.4161e-06, 1.0608e-06, ..., -7.7812e-07, + -7.1013e-07, 4.6566e-10], + ..., + [ 4.6566e-10, 2.8666e-06, 2.3935e-07, ..., 1.1884e-06, + 2.7660e-06, -2.9802e-08], + [ 0.0000e+00, 1.8161e-08, 6.4261e-08, ..., 8.8941e-08, + 1.7509e-07, 0.0000e+00], + [ 0.0000e+00, 1.1921e-07, 6.8918e-08, ..., 2.1094e-07, + 2.1420e-07, 1.8161e-08]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0078, -0.0363, 0.0073, -0.0192, 0.0171, 0.0095, 0.0273, 0.0016, + -0.0385, -0.0062], device='cuda:0'), grad: tensor([-6.0648e-06, -4.2692e-06, -2.9672e-06, 5.3756e-06, 6.7521e-08, + -6.2957e-06, 6.0871e-06, 5.3644e-06, 6.2305e-07, 2.0582e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 247.39, cls_loss 0.0019 cls_loss_mapping 0.0027 cls_loss_causal 0.4991 re_mapping 0.0054 re_causal 0.0165 /// teacc 98.92 lr 0.00010000 +Epoch 220, weight, value: tensor([[ 0.0278, -0.1466, -0.1260, ..., -0.2663, -0.0965, -0.1292], + [ 0.0398, -0.0715, 0.0331, ..., 0.0486, 0.0951, -0.0379], + [-0.0662, 0.1247, -0.1563, ..., 0.0595, 0.0727, -0.0389], + ..., + [-0.0816, -0.0819, -0.0758, ..., 0.0026, -0.1478, 0.1206], + [ 0.0676, -0.0242, 0.0699, ..., -0.0019, -0.1902, -0.0086], + [-0.1568, -0.0656, -0.0798, ..., -0.1728, 0.0485, -0.0981]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 9.3132e-09, + 3.1013e-07, -1.5832e-08], + [ 9.3132e-10, 0.0000e+00, 6.0536e-08, ..., 4.0047e-08, + 1.2107e-08, 1.1176e-08], + [ 0.0000e+00, 2.7940e-09, 5.6811e-08, ..., 5.6811e-08, + 1.2107e-08, 9.3132e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 2.9802e-08, ..., -3.4459e-08, + 3.5390e-08, -2.0489e-08], + [ 0.0000e+00, -3.7253e-09, -9.2480e-07, ..., -3.1851e-07, + 2.1420e-08, 5.5879e-09], + [ 1.8626e-09, 0.0000e+00, 6.1467e-08, ..., 4.6566e-08, + -1.8626e-09, 2.0489e-08]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0073, -0.0363, 0.0077, -0.0194, 0.0171, 0.0104, 0.0271, 0.0019, + -0.0387, -0.0071], device='cuda:0'), grad: tensor([-3.5688e-06, 4.3400e-07, 4.6380e-07, 7.3276e-06, 2.6096e-06, + 1.0729e-05, 6.2399e-08, 5.5879e-07, -1.2433e-06, -1.7434e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 247.62, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5294 re_mapping 0.0056 re_causal 0.0173 /// teacc 98.97 lr 0.00010000 +Epoch 221, weight, value: tensor([[ 0.0276, -0.1466, -0.1262, ..., -0.2672, -0.0965, -0.1292], + [ 0.0398, -0.0717, 0.0332, ..., 0.0486, 0.0950, -0.0380], + [-0.0662, 0.1248, -0.1562, ..., 0.0595, 0.0734, -0.0389], + ..., + [-0.0817, -0.0819, -0.0760, ..., 0.0026, -0.1482, 0.1207], + [ 0.0676, -0.0243, 0.0747, ..., 0.0027, -0.1896, -0.0086], + [-0.1570, -0.0656, -0.0805, ..., -0.1732, 0.0484, -0.0983]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.3819e-09, 1.9558e-08, ..., 1.0524e-07, + 1.3690e-07, 9.3132e-10], + [ 0.0000e+00, 1.3970e-08, -1.1753e-06, ..., -5.6531e-07, + -9.4902e-07, 5.5879e-09], + [ 0.0000e+00, -1.3504e-07, 2.1886e-07, ..., -2.4773e-06, + -2.8647e-06, -5.5879e-09], + ..., + [ 0.0000e+00, 1.0245e-08, 9.4716e-07, ..., 6.5751e-07, + 7.0687e-07, -0.0000e+00], + [ 0.0000e+00, 8.0094e-08, 3.3248e-07, ..., 2.4829e-06, + 2.3674e-06, 9.3132e-10], + [ 0.0000e+00, 8.3819e-09, 5.8487e-07, ..., 6.8452e-07, + 8.5868e-07, 0.0000e+00]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0070, -0.0364, 0.0082, -0.0192, 0.0173, 0.0099, 0.0261, 0.0019, + -0.0344, -0.0073], device='cuda:0'), grad: tensor([ 2.6822e-07, -2.3395e-06, -5.6326e-06, 1.2666e-06, -1.1828e-06, + -5.5581e-06, 4.2003e-07, 1.9595e-06, 6.3293e-06, 4.4741e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 247.67, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5269 re_mapping 0.0057 re_causal 0.0167 /// teacc 98.97 lr 0.00010000 +Epoch 222, weight, value: tensor([[ 0.0276, -0.1466, -0.1264, ..., -0.2684, -0.0965, -0.1289], + [ 0.0399, -0.0718, 0.0333, ..., 0.0487, 0.0951, -0.0383], + [-0.0662, 0.1250, -0.1566, ..., 0.0595, 0.0734, -0.0390], + ..., + [-0.0820, -0.0820, -0.0762, ..., 0.0024, -0.1487, 0.1208], + [ 0.0676, -0.0244, 0.0747, ..., 0.0027, -0.1897, -0.0087], + [-0.1581, -0.0660, -0.0812, ..., -0.1739, 0.0491, -0.0984]], + device='cuda:0'), grad: tensor([[-1.6298e-07, 1.8626e-09, 3.7253e-09, ..., 9.3132e-09, + 6.0238e-06, 9.3132e-10], + [ 2.3283e-08, 8.3819e-09, -7.0781e-08, ..., 4.4703e-08, + 2.5425e-07, 2.6077e-08], + [ 3.2596e-08, -2.7008e-08, 2.2352e-08, ..., -3.6322e-08, + 2.7753e-06, 1.8626e-09], + ..., + [ 3.7253e-09, 1.0245e-08, 7.4506e-08, ..., -2.8871e-08, + 3.4925e-07, -5.5879e-08], + [ 1.3039e-08, 9.3132e-10, -9.3132e-08, ..., -3.0734e-08, + 2.4028e-07, 7.4506e-09], + [ 6.2399e-08, 0.0000e+00, 7.4506e-09, ..., 4.0978e-08, + 7.0967e-07, 1.6764e-08]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0069, -0.0364, 0.0082, -0.0194, 0.0163, 0.0101, 0.0260, 0.0019, + -0.0345, -0.0067], device='cuda:0'), grad: tensor([ 8.9481e-06, 1.3625e-06, 6.2175e-06, 4.5598e-06, 1.7453e-06, + -4.2245e-06, -2.3276e-05, 4.8056e-07, 8.7637e-07, 3.2894e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 247.90, cls_loss 0.0020 cls_loss_mapping 0.0029 cls_loss_causal 0.5174 re_mapping 0.0061 re_causal 0.0169 /// teacc 99.03 lr 0.00010000 +Epoch 223, weight, value: tensor([[ 0.0274, -0.1468, -0.1271, ..., -0.2694, -0.0966, -0.1289], + [ 0.0398, -0.0721, 0.0333, ..., 0.0485, 0.0951, -0.0384], + [-0.0662, 0.1248, -0.1568, ..., 0.0589, 0.0730, -0.0390], + ..., + [-0.0821, -0.0814, -0.0762, ..., 0.0033, -0.1479, 0.1209], + [ 0.0678, -0.0248, 0.0749, ..., 0.0028, -0.1899, -0.0087], + [-0.1594, -0.0661, -0.0819, ..., -0.1745, 0.0500, -0.0985]], + device='cuda:0'), grad: tensor([[ 4.7404e-07, 9.3132e-10, 3.3062e-07, ..., 2.5146e-08, + 7.8231e-08, 3.7253e-09], + [ 6.6124e-08, 0.0000e+00, 1.6671e-07, ..., 3.6601e-07, + -5.1223e-08, 2.2352e-08], + [ 6.4541e-07, -7.4506e-09, 1.4240e-06, ..., 1.0710e-06, + 5.1968e-07, 1.1176e-08], + ..., + [ 2.3637e-06, 3.7253e-09, 1.6997e-06, ..., -2.8871e-07, + 4.2189e-07, -5.6811e-08], + [ 1.0254e-06, 9.3132e-10, -3.9395e-07, ..., -1.1669e-06, + -4.5355e-07, 5.5879e-09], + [ 1.0217e-06, 0.0000e+00, 6.9011e-07, ..., 6.7055e-08, + 1.3690e-07, 6.5193e-09]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0070, -0.0365, 0.0073, -0.0193, 0.0158, 0.0103, 0.0259, 0.0026, + -0.0345, -0.0058], device='cuda:0'), grad: tensor([ 1.0245e-05, 2.6599e-06, 1.6510e-05, 1.8969e-05, 2.3037e-05, + -1.7011e-04, 6.1877e-06, 5.0038e-05, 1.9729e-05, 2.2829e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 247.73, cls_loss 0.0019 cls_loss_mapping 0.0027 cls_loss_causal 0.5130 re_mapping 0.0061 re_causal 0.0175 /// teacc 98.91 lr 0.00010000 +Epoch 224, weight, value: tensor([[ 0.0273, -0.1463, -0.1275, ..., -0.2728, -0.0966, -0.1290], + [ 0.0400, -0.0732, 0.0335, ..., 0.0486, 0.0952, -0.0381], + [-0.0667, 0.1252, -0.1571, ..., 0.0600, 0.0742, -0.0390], + ..., + [-0.0791, -0.0811, -0.0768, ..., 0.0032, -0.1492, 0.1207], + [ 0.0656, -0.0262, 0.0748, ..., 0.0025, -0.1902, -0.0088], + [-0.1619, -0.0667, -0.0843, ..., -0.1751, 0.0493, -0.0987]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 3.7253e-09, 3.8184e-08, ..., 5.6811e-08, + 4.7497e-08, 2.6077e-08], + [ 3.4459e-08, 3.7253e-09, 3.8221e-06, ..., 8.8662e-06, + -7.9256e-07, 5.5283e-06], + [ 2.7940e-09, -5.8673e-08, 3.1479e-07, ..., 1.8254e-07, + -9.9652e-08, 1.4715e-07], + ..., + [ 2.7381e-07, 1.9558e-08, -6.9924e-06, ..., -1.4499e-05, + 1.1111e-06, -9.1866e-06], + [ 8.9407e-08, 4.6566e-09, -2.1681e-06, ..., -1.2284e-06, + 2.2724e-07, 7.5437e-07], + [ 1.6764e-07, 1.8626e-09, 3.0268e-07, ..., 5.3924e-07, + 1.5590e-06, 1.6857e-07]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0073, -0.0363, 0.0082, -0.0199, 0.0166, 0.0105, 0.0260, 0.0023, + -0.0348, -0.0066], device='cuda:0'), grad: tensor([-3.8370e-07, 4.2349e-05, 1.3914e-06, 2.4125e-05, -7.8231e-06, + 4.8578e-06, 6.3423e-07, -6.8069e-05, -2.0303e-07, 3.0696e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 247.47, cls_loss 0.0029 cls_loss_mapping 0.0046 cls_loss_causal 0.5051 re_mapping 0.0059 re_causal 0.0167 /// teacc 98.92 lr 0.00010000 +Epoch 225, weight, value: tensor([[ 0.0272, -0.1466, -0.1277, ..., -0.2748, -0.0967, -0.1290], + [ 0.0401, -0.0740, 0.0336, ..., 0.0487, 0.0953, -0.0381], + [-0.0668, 0.1257, -0.1548, ..., 0.0631, 0.0774, -0.0392], + ..., + [-0.0791, -0.0814, -0.0773, ..., 0.0028, -0.1504, 0.1208], + [ 0.0656, -0.0261, 0.0749, ..., 0.0026, -0.1903, -0.0088], + [-0.1623, -0.0672, -0.0852, ..., -0.1779, 0.0492, -0.0988]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 2.7008e-08, + 1.0245e-08, 0.0000e+00], + [-2.7940e-09, 0.0000e+00, 2.7940e-08, ..., 7.9162e-08, + -5.4017e-08, 1.8626e-09], + [ 0.0000e+00, -1.8626e-09, 2.5332e-07, ..., 6.1840e-07, + 1.1176e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, 5.0385e-07, ..., 5.0757e-07, + 4.1910e-08, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -3.4459e-08, ..., -9.3132e-10, + 5.1223e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 8.1956e-08, ..., 1.0245e-07, + -3.2783e-07, 9.3132e-10]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0071, -0.0363, 0.0112, -0.0222, 0.0167, 0.0101, 0.0260, 0.0021, + -0.0347, -0.0070], device='cuda:0'), grad: tensor([ 1.0710e-07, 1.9930e-07, 1.3197e-06, -4.1462e-06, 8.5495e-07, + 5.0850e-07, 2.3190e-07, 1.5562e-06, 1.7416e-07, -8.0280e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 248.40, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5237 re_mapping 0.0057 re_causal 0.0166 /// teacc 98.93 lr 0.00010000 +Epoch 226, weight, value: tensor([[ 0.0269, -0.1458, -0.1299, ..., -0.2768, -0.0967, -0.1291], + [ 0.0405, -0.0744, 0.0341, ..., 0.0496, 0.0958, -0.0382], + [-0.0669, 0.1255, -0.1547, ..., 0.0633, 0.0776, -0.0394], + ..., + [-0.0791, -0.0818, -0.0789, ..., 0.0016, -0.1533, 0.1210], + [ 0.0654, -0.0254, 0.0749, ..., 0.0027, -0.1906, -0.0088], + [-0.1630, -0.0680, -0.0856, ..., -0.1784, 0.0492, -0.0988]], + device='cuda:0'), grad: tensor([[ 7.5437e-08, 4.6566e-09, 1.6503e-06, ..., 9.3132e-08, + 2.4606e-06, 9.3132e-10], + [-2.0582e-07, 3.5390e-08, -5.4538e-06, ..., -4.4703e-07, + -7.8008e-06, 1.8626e-08], + [ 1.5832e-08, 6.8732e-07, 4.4238e-07, ..., 5.2340e-07, + 6.9011e-07, 8.3819e-09], + ..., + [ 2.5146e-08, -9.2853e-07, 1.0207e-06, ..., -4.3400e-07, + 1.0459e-06, -4.6566e-08], + [ 3.9116e-08, 3.2596e-08, 6.3237e-07, ..., 7.0781e-08, + 1.2955e-06, 5.5879e-09], + [ 6.2399e-08, 3.5390e-08, 2.9150e-07, ..., 1.6484e-07, + 6.0070e-07, 5.5879e-09]], device='cuda:0') +Epoch 226, bias, value: tensor([-7.2164e-03, -3.5470e-02, 1.1452e-02, -2.2439e-02, 1.6789e-02, + 1.0644e-02, 2.5866e-02, 7.7172e-05, -3.5012e-02, -7.2658e-03], + device='cuda:0'), grad: tensor([ 5.4911e-06, -1.5825e-05, 3.1665e-06, -1.2815e-06, 6.7614e-07, + 2.1327e-06, 1.8505e-06, -1.5972e-06, 2.7716e-06, 2.5779e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 248.04, cls_loss 0.0019 cls_loss_mapping 0.0028 cls_loss_causal 0.5010 re_mapping 0.0055 re_causal 0.0161 /// teacc 98.94 lr 0.00010000 +Epoch 227, weight, value: tensor([[ 0.0268, -0.1460, -0.1300, ..., -0.2787, -0.0970, -0.1297], + [ 0.0406, -0.0731, 0.0343, ..., 0.0502, 0.0962, -0.0387], + [-0.0670, 0.1247, -0.1549, ..., 0.0632, 0.0774, -0.0395], + ..., + [-0.0782, -0.0814, -0.0794, ..., 0.0012, -0.1542, 0.1214], + [ 0.0650, -0.0253, 0.0750, ..., 0.0027, -0.1908, -0.0089], + [-0.1634, -0.0692, -0.0857, ..., -0.1798, 0.0493, -0.0990]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.1665e-08, 5.5879e-08, ..., 1.1921e-07, + 1.3039e-07, 0.0000e+00], + [-1.8626e-09, -1.8431e-06, -2.0061e-06, ..., -3.8408e-06, + -4.8876e-06, -3.7253e-09], + [ 5.5879e-09, 1.5376e-06, 1.2135e-06, ..., 3.6396e-06, + 5.0738e-06, 9.3132e-10], + ..., + [ 4.6566e-09, -6.8080e-07, 4.4517e-07, ..., -8.5495e-07, + 1.8794e-06, 9.3132e-10], + [ 1.8626e-09, 3.1013e-07, 1.9744e-07, ..., 7.2643e-07, + 1.3402e-06, 0.0000e+00], + [ 1.8626e-09, 2.1514e-07, 4.3772e-08, ..., -8.1211e-07, + -7.8455e-06, 0.0000e+00]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0073, -0.0354, 0.0113, -0.0224, 0.0168, 0.0105, 0.0259, 0.0006, + -0.0351, -0.0075], device='cuda:0'), grad: tensor([ 4.9546e-07, 3.4552e-07, 1.0885e-05, -6.3702e-07, 7.4506e-06, + 2.0787e-06, 1.4994e-07, -3.4235e-06, 3.4608e-06, -2.0847e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 248.15, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5182 re_mapping 0.0057 re_causal 0.0172 /// teacc 98.88 lr 0.00010000 +Epoch 228, weight, value: tensor([[ 0.0267, -0.1462, -0.1302, ..., -0.2814, -0.0971, -0.1298], + [ 0.0409, -0.0713, 0.0346, ..., 0.0507, 0.0966, -0.0388], + [-0.0670, 0.1250, -0.1552, ..., 0.0632, 0.0774, -0.0396], + ..., + [-0.0781, -0.0821, -0.0798, ..., 0.0009, -0.1553, 0.1216], + [ 0.0646, -0.0252, 0.0750, ..., 0.0026, -0.1911, -0.0092], + [-0.1636, -0.0700, -0.0861, ..., -0.1801, 0.0495, -0.0990]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 1.1176e-08, ..., 8.3819e-09, + 1.3970e-08, 0.0000e+00], + [-0.0000e+00, 0.0000e+00, -2.6785e-06, ..., -2.2575e-06, + -4.2282e-06, -5.5879e-09], + [ 0.0000e+00, -1.2107e-08, 8.6706e-07, ..., 7.3668e-07, + 1.2992e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 6.5193e-09, 1.6736e-06, ..., 1.4082e-06, + 2.6934e-06, 3.7253e-09], + [ 0.0000e+00, 9.3132e-10, -8.1956e-08, ..., -2.5146e-08, + 7.0781e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.7987e-08, ..., 5.2154e-08, + 2.9895e-07, 0.0000e+00]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0073, -0.0351, 0.0112, -0.0224, 0.0165, 0.0110, 0.0257, 0.0003, + -0.0353, -0.0072], device='cuda:0'), grad: tensor([-8.6613e-08, -5.9716e-06, 2.0489e-06, -2.3600e-06, -1.2405e-06, + 2.0824e-06, 2.2072e-07, 4.1276e-06, 2.0396e-07, 9.4529e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 248.30, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5173 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.93 lr 0.00010000 +Epoch 229, weight, value: tensor([[ 0.0268, -0.1464, -0.1275, ..., -0.2792, -0.0976, -0.1299], + [ 0.0410, -0.0715, 0.0346, ..., 0.0506, 0.0964, -0.0390], + [-0.0671, 0.1247, -0.1556, ..., 0.0631, 0.0771, -0.0398], + ..., + [-0.0782, -0.0818, -0.0800, ..., 0.0010, -0.1563, 0.1217], + [ 0.0646, -0.0250, 0.0750, ..., 0.0024, -0.1913, -0.0093], + [-0.1639, -0.0698, -0.0886, ..., -0.1793, 0.0487, -0.0991]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 5.5879e-09, ..., 6.5193e-09, + 2.5779e-06, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -5.1223e-08, ..., -9.3132e-10, + -4.8429e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-08, 5.4017e-08, ..., 2.6077e-08, + -7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 8.8476e-08, ..., 9.7789e-08, + 2.2352e-08, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, -4.6566e-09, ..., 2.7940e-09, + 7.6368e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 6.5193e-09, + 1.1642e-07, 0.0000e+00]], device='cuda:0') +Epoch 229, bias, value: tensor([-7.1250e-03, -3.5381e-02, 1.1009e-02, -2.2298e-02, 1.9389e-02, + 1.0911e-02, 2.5526e-02, -9.1543e-05, -3.5612e-02, -7.4394e-03], + device='cuda:0'), grad: tensor([ 6.6608e-06, -4.4703e-08, 9.5926e-08, 6.8918e-07, 1.5367e-07, + -4.5806e-05, 3.4899e-05, 2.7381e-07, 1.2424e-06, 1.8552e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 248.12, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.5292 re_mapping 0.0058 re_causal 0.0171 /// teacc 98.99 lr 0.00010000 +Epoch 230, weight, value: tensor([[ 0.0269, -0.1468, -0.1274, ..., -0.2797, -0.0972, -0.1304], + [ 0.0410, -0.0716, 0.0347, ..., 0.0506, 0.0965, -0.0393], + [-0.0671, 0.1247, -0.1559, ..., 0.0631, 0.0771, -0.0407], + ..., + [-0.0782, -0.0819, -0.0801, ..., 0.0013, -0.1564, 0.1222], + [ 0.0647, -0.0245, 0.0751, ..., 0.0025, -0.1915, -0.0093], + [-0.1641, -0.0700, -0.0891, ..., -0.1801, 0.0488, -0.0994]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.1176e-08, ..., 1.2107e-08, + 9.5926e-08, 6.5193e-09], + [-2.0489e-08, 1.2107e-08, -5.4669e-07, ..., -2.1607e-07, + -4.8615e-07, 1.5832e-08], + [ 2.7940e-09, -5.5879e-09, 6.8918e-08, ..., 5.4017e-08, + 6.4261e-08, 9.3132e-09], + ..., + [ 1.0245e-08, -4.6566e-09, 2.0117e-07, ..., -1.6578e-07, + 2.3749e-07, -8.7544e-08], + [ 3.7253e-09, 2.7940e-09, -1.5087e-07, ..., -1.0058e-07, + 2.6170e-07, 1.0245e-08], + [ 0.0000e+00, 8.3819e-09, 8.8476e-08, ..., 1.6950e-07, + 7.1712e-08, 1.0245e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0062, -0.0353, 0.0110, -0.0223, 0.0190, 0.0111, 0.0255, 0.0001, + -0.0357, -0.0079], device='cuda:0'), grad: tensor([-6.8098e-06, -6.9942e-07, 4.3679e-07, 2.9430e-07, 1.1735e-07, + -2.1048e-07, -7.4692e-07, 4.7833e-06, 6.7707e-07, 2.1495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 248.03, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.4792 re_mapping 0.0054 re_causal 0.0157 /// teacc 98.96 lr 0.00010000 +Epoch 231, weight, value: tensor([[ 0.0268, -0.1470, -0.1277, ..., -0.2810, -0.0973, -0.1309], + [ 0.0410, -0.0694, 0.0356, ..., 0.0523, 0.0980, -0.0399], + [-0.0671, 0.1238, -0.1580, ..., 0.0621, 0.0754, -0.0410], + ..., + [-0.0782, -0.0815, -0.0795, ..., 0.0030, -0.1553, 0.1229], + [ 0.0647, -0.0253, 0.0749, ..., 0.0022, -0.1921, -0.0094], + [-0.1639, -0.0702, -0.0893, ..., -0.1806, 0.0486, -0.0995]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 1.5832e-07, ..., 1.9837e-07, + -1.8626e-08, 1.8626e-09], + [ 9.3132e-10, 2.7940e-09, 2.8405e-07, ..., 4.4703e-07, + 4.6007e-07, 9.3132e-10], + [ 9.3132e-10, -3.6322e-08, 2.5444e-06, ..., 3.2093e-06, + 1.1837e-06, 9.3132e-10], + ..., + [ 0.0000e+00, 1.7695e-08, 1.1986e-06, ..., 1.3933e-06, + 4.2096e-07, -9.3132e-10], + [-0.0000e+00, 7.4506e-09, 4.4517e-07, ..., 6.0722e-07, + 2.7381e-07, 1.8626e-09], + [ 0.0000e+00, 9.3132e-10, 3.8650e-07, ..., 4.9919e-07, + 7.3295e-07, 0.0000e+00]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0063, -0.0346, 0.0103, -0.0224, 0.0192, 0.0113, 0.0255, 0.0016, + -0.0362, -0.0083], device='cuda:0'), grad: tensor([-3.0827e-07, 2.0675e-06, 9.6709e-06, -2.0385e-05, -8.4564e-07, + 3.8296e-06, -3.4291e-06, 4.1872e-06, 2.0433e-06, 3.1367e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 248.03, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.5028 re_mapping 0.0054 re_causal 0.0156 /// teacc 99.08 lr 0.00010000 +Epoch 232, weight, value: tensor([[ 0.0260, -0.1472, -0.1278, ..., -0.2813, -0.0974, -0.1311], + [ 0.0409, -0.0711, 0.0359, ..., 0.0521, 0.0979, -0.0400], + [-0.0674, 0.1229, -0.1581, ..., 0.0621, 0.0756, -0.0410], + ..., + [-0.0765, -0.0823, -0.0802, ..., 0.0027, -0.1561, 0.1230], + [ 0.0647, -0.0256, 0.0749, ..., 0.0022, -0.1922, -0.0094], + [-0.1647, -0.0713, -0.0898, ..., -0.1818, 0.0486, -0.0997]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1176e-08, 5.8673e-08, ..., 1.5832e-07, + 1.0338e-07, 0.0000e+00], + [ 9.3132e-10, 1.3039e-08, 2.8685e-07, ..., 9.1270e-07, + 2.0210e-07, 0.0000e+00], + [ 0.0000e+00, -1.7881e-07, 1.0012e-06, ..., 5.3085e-06, + 1.9539e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 9.6858e-08, -8.8289e-07, ..., -7.9721e-06, + -2.9933e-06, -1.8626e-09], + [ 0.0000e+00, 1.5832e-08, -2.2985e-06, ..., -1.7490e-06, + -6.5472e-07, 0.0000e+00], + [ 1.8626e-09, 4.6566e-09, 3.2689e-07, ..., 1.4165e-06, + 7.2923e-07, 1.8626e-09]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0062, -0.0346, 0.0102, -0.0221, 0.0193, 0.0107, 0.0255, 0.0012, + -0.0363, -0.0084], device='cuda:0'), grad: tensor([ 6.0908e-07, 3.0398e-06, 2.1413e-05, 2.2873e-06, -9.3319e-07, + 8.7637e-07, 1.8561e-06, -3.2753e-05, -2.7008e-06, 6.3255e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 248.27, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.5090 re_mapping 0.0052 re_causal 0.0162 /// teacc 99.02 lr 0.00010000 +Epoch 233, weight, value: tensor([[ 0.0261, -0.1485, -0.1280, ..., -0.2819, -0.0974, -0.1311], + [ 0.0412, -0.0725, 0.0361, ..., 0.0520, 0.0979, -0.0401], + [-0.0675, 0.1241, -0.1582, ..., 0.0621, 0.0758, -0.0410], + ..., + [-0.0760, -0.0829, -0.0804, ..., 0.0027, -0.1566, 0.1231], + [ 0.0642, -0.0255, 0.0750, ..., 0.0023, -0.1924, -0.0095], + [-0.1659, -0.0723, -0.0905, ..., -0.1823, 0.0486, -0.0997]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.7253e-09, 1.0012e-06, ..., 4.6566e-07, + -1.4901e-08, 0.0000e+00], + [ 5.5879e-09, 4.0978e-08, -4.1537e-07, ..., -2.6915e-07, + -6.9384e-07, 9.3132e-10], + [ 9.3132e-10, -1.4622e-07, -1.2852e-07, ..., -5.1595e-07, + -1.0263e-06, 0.0000e+00], + ..., + [ 9.3132e-10, 6.3330e-08, 2.6450e-07, ..., 3.5670e-07, + 6.0722e-07, -2.7940e-09], + [ 1.2107e-08, 3.7253e-09, -2.4643e-06, ..., -1.0645e-06, + 1.0431e-07, 0.0000e+00], + [ 1.3970e-08, 1.7695e-08, 2.5705e-07, ..., 2.1048e-07, + 5.9232e-07, 9.3132e-10]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0062, -0.0346, 0.0103, -0.0222, 0.0194, 0.0107, 0.0255, 0.0012, + -0.0363, -0.0085], device='cuda:0'), grad: tensor([ 1.8785e-06, -6.6310e-07, -2.4475e-06, 7.4320e-07, -2.5146e-07, + 1.1176e-08, 1.5721e-06, 1.5749e-06, -5.0776e-06, 2.6487e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 248.30, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5144 re_mapping 0.0055 re_causal 0.0172 /// teacc 98.97 lr 0.00010000 +Epoch 234, weight, value: tensor([[ 0.0286, -0.1482, -0.1280, ..., -0.2822, -0.0974, -0.1312], + [ 0.0410, -0.0700, 0.0367, ..., 0.0525, 0.0982, -0.0399], + [-0.0684, 0.1229, -0.1588, ..., 0.0621, 0.0756, -0.0410], + ..., + [-0.0755, -0.0836, -0.0815, ..., 0.0020, -0.1574, 0.1230], + [ 0.0643, -0.0257, 0.0750, ..., 0.0022, -0.1926, -0.0096], + [-0.1672, -0.0731, -0.0910, ..., -0.1831, 0.0486, -0.1000]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-08, 2.5146e-08, ..., 4.4703e-08, + -3.2596e-08, 1.8626e-09], + [ 9.3132e-10, 1.9558e-08, -1.0617e-07, ..., 1.8440e-07, + -1.1269e-07, 6.6124e-08], + [ 9.3132e-10, -8.7544e-08, 4.1910e-08, ..., -4.2841e-08, + -6.3330e-08, 4.6566e-09], + ..., + [ 9.3132e-10, 1.4901e-08, 1.5832e-08, ..., -9.8068e-07, + 4.6566e-08, -5.6252e-07], + [ 0.0000e+00, -2.7940e-09, -3.5390e-08, ..., -1.7695e-08, + 8.4750e-08, 4.6566e-09], + [ 0.0000e+00, 7.4506e-09, 2.4214e-08, ..., 2.1234e-07, + 2.5146e-07, 9.4064e-08]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0058, -0.0343, 0.0102, -0.0222, 0.0194, 0.0112, 0.0253, 0.0007, + -0.0364, -0.0086], device='cuda:0'), grad: tensor([-4.6566e-07, 4.1816e-07, 2.0117e-07, 2.6748e-06, 7.5437e-07, + -6.9011e-07, -1.2154e-06, -3.8296e-06, 3.0175e-07, 1.8599e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 247.67, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.4887 re_mapping 0.0055 re_causal 0.0156 /// teacc 99.02 lr 0.00010000 +Epoch 235, weight, value: tensor([[ 0.0288, -0.1485, -0.1281, ..., -0.2827, -0.0974, -0.1312], + [ 0.0439, -0.0708, 0.0371, ..., 0.0521, 0.0986, -0.0407], + [-0.0689, 0.1234, -0.1590, ..., 0.0621, 0.0757, -0.0411], + ..., + [-0.0744, -0.0834, -0.0804, ..., 0.0038, -0.1578, 0.1240], + [ 0.0611, -0.0257, 0.0740, ..., 0.0014, -0.1937, -0.0097], + [-0.1684, -0.0733, -0.0917, ..., -0.1841, 0.0486, -0.1005]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 4.6566e-09, 1.4808e-07, ..., 1.1362e-07, + 1.6578e-07, 9.3132e-10], + [ 9.8161e-07, 2.5146e-08, -4.9099e-06, ..., -1.3541e-06, + -1.2949e-05, -2.1420e-08], + [ 2.0396e-07, -3.7253e-08, 3.1069e-06, ..., 2.2054e-06, + 3.2987e-06, 2.7940e-09], + ..., + [-2.2631e-07, -2.5891e-07, 9.3039e-07, ..., 3.6135e-07, + 1.0487e-06, 3.7253e-09], + [-1.1027e-06, 1.4342e-07, -5.1484e-06, ..., -5.2489e-06, + 1.4342e-07, 1.8626e-09], + [ 5.1223e-08, 5.1223e-08, -4.9453e-07, ..., 6.8918e-08, + -3.2783e-07, 0.0000e+00]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0062, -0.0341, 0.0102, -0.0221, 0.0193, 0.0119, 0.0254, 0.0018, + -0.0382, -0.0088], device='cuda:0'), grad: tensor([ 7.0874e-07, -1.0408e-05, 8.2776e-06, 2.5369e-06, 1.0520e-05, + 8.3726e-07, 2.9095e-06, 4.0196e-06, -1.3731e-05, -5.6773e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 248.08, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.5181 re_mapping 0.0056 re_causal 0.0162 /// teacc 98.99 lr 0.00010000 +Epoch 236, weight, value: tensor([[ 0.0286, -0.1491, -0.1284, ..., -0.2834, -0.0977, -0.1314], + [ 0.0439, -0.0708, 0.0374, ..., 0.0518, 0.0986, -0.0410], + [-0.0690, 0.1238, -0.1592, ..., 0.0621, 0.0758, -0.0412], + ..., + [-0.0748, -0.0840, -0.0806, ..., 0.0045, -0.1584, 0.1245], + [ 0.0615, -0.0250, 0.0740, ..., 0.0013, -0.1941, -0.0098], + [-0.1685, -0.0737, -0.0923, ..., -0.1840, 0.0487, -0.1006]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.9802e-08, -1.6391e-07, ..., 1.5832e-08, + -1.3132e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 0.0000e+00, ..., 1.5274e-07, + -4.1910e-08, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, 4.3772e-08, ..., -2.0489e-08, + -4.0047e-08, 0.0000e+00], + ..., + [-9.3132e-10, 1.7695e-08, 3.5390e-08, ..., -4.8988e-07, + 5.4017e-08, -0.0000e+00], + [ 0.0000e+00, -1.9372e-07, -1.7323e-07, ..., -1.3970e-08, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 6.3330e-08, 1.5087e-07, ..., 1.7229e-07, + 1.4063e-07, 0.0000e+00]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0064, -0.0343, 0.0102, -0.0221, 0.0193, 0.0131, 0.0252, 0.0022, + -0.0386, -0.0087], device='cuda:0'), grad: tensor([-1.4938e-06, 6.8173e-07, 1.7509e-07, 7.1060e-07, -1.0431e-07, + 1.1548e-07, 5.9884e-07, -1.8496e-06, -8.4378e-07, 1.9986e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 247.83, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.4963 re_mapping 0.0053 re_causal 0.0157 /// teacc 99.01 lr 0.00010000 +Epoch 237, weight, value: tensor([[ 0.0282, -0.1496, -0.1286, ..., -0.2842, -0.0978, -0.1314], + [ 0.0439, -0.0693, 0.0380, ..., 0.0522, 0.0990, -0.0410], + [-0.0693, 0.1230, -0.1601, ..., 0.0620, 0.0756, -0.0412], + ..., + [-0.0733, -0.0848, -0.0811, ..., 0.0045, -0.1596, 0.1246], + [ 0.0617, -0.0252, 0.0742, ..., 0.0015, -0.1942, -0.0098], + [-0.1685, -0.0736, -0.0929, ..., -0.1844, 0.0490, -0.1006]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 4.2003e-07, 9.3132e-09, ..., 3.1199e-07, + 6.6124e-08, 0.0000e+00], + [ 4.6566e-09, 7.4506e-08, -5.2154e-08, ..., 7.4506e-09, + -1.8161e-07, 0.0000e+00], + [-2.9802e-08, -2.7329e-05, 2.5146e-07, ..., -1.9029e-05, + -2.0489e-08, 0.0000e+00], + ..., + [-7.4506e-09, 2.3171e-05, 3.0734e-08, ..., 1.6242e-05, + 3.9116e-08, 0.0000e+00], + [ 1.8626e-08, 1.9558e-08, -4.0885e-07, ..., -1.7416e-07, + 1.8720e-07, 0.0000e+00], + [ 2.7940e-09, 2.7474e-07, 1.2107e-08, ..., 2.0768e-07, + 2.5146e-08, 0.0000e+00]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0065, -0.0339, 0.0100, -0.0221, 0.0188, 0.0131, 0.0252, 0.0017, + -0.0385, -0.0082], device='cuda:0'), grad: tensor([ 9.2015e-07, 1.0990e-07, -5.2929e-05, 6.5379e-06, 1.5739e-07, + -7.5281e-05, 7.4148e-05, 4.5210e-05, 4.2841e-07, 7.8697e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 247.54, cls_loss 0.0015 cls_loss_mapping 0.0032 cls_loss_causal 0.4941 re_mapping 0.0050 re_causal 0.0153 /// teacc 99.00 lr 0.00010000 +Epoch 238, weight, value: tensor([[ 0.0286, -0.1500, -0.1287, ..., -0.2848, -0.0979, -0.1319], + [ 0.0437, -0.0693, 0.0381, ..., 0.0522, 0.0990, -0.0411], + [-0.0693, 0.1232, -0.1602, ..., 0.0620, 0.0757, -0.0412], + ..., + [-0.0732, -0.0852, -0.0815, ..., 0.0042, -0.1605, 0.1246], + [ 0.0616, -0.0250, 0.0748, ..., 0.0024, -0.1940, -0.0098], + [-0.1692, -0.0738, -0.0932, ..., -0.1835, 0.0493, -0.1007]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 2.8871e-08, 6.0815e-07, ..., 4.4890e-07, + 6.8545e-07, 2.7940e-09], + [ 1.6298e-07, 8.3819e-09, -1.5823e-06, ..., 1.1176e-07, + -1.5469e-06, 1.8626e-09], + [ 9.3132e-09, -9.5926e-08, 4.7870e-07, ..., 6.4261e-07, + -1.3970e-08, -1.4901e-08], + ..., + [-3.2783e-07, 2.7940e-09, 1.4901e-07, ..., -1.5842e-06, + 9.9652e-08, 0.0000e+00], + [ 3.7253e-09, 1.7695e-08, -3.8650e-07, ..., 7.9162e-08, + 2.5146e-07, 2.7940e-09], + [ 2.6077e-08, 9.3132e-10, 7.5437e-08, ..., 1.7136e-07, + 9.5926e-07, 0.0000e+00]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0065, -0.0339, 0.0100, -0.0220, 0.0186, 0.0131, 0.0249, 0.0013, + -0.0378, -0.0078], device='cuda:0'), grad: tensor([ 3.0436e-06, -9.8627e-07, 2.1234e-06, -1.8366e-06, -2.2743e-06, + 5.1260e-06, -3.1665e-07, -8.5905e-06, 5.7742e-07, 3.1237e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 248.08, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.5226 re_mapping 0.0052 re_causal 0.0161 /// teacc 99.00 lr 0.00010000 +Epoch 239, weight, value: tensor([[ 0.0286, -0.1506, -0.1289, ..., -0.2857, -0.0980, -0.1317], + [ 0.0438, -0.0692, 0.0383, ..., 0.0523, 0.0991, -0.0410], + [-0.0694, 0.1230, -0.1603, ..., 0.0620, 0.0757, -0.0412], + ..., + [-0.0731, -0.0857, -0.0820, ..., 0.0041, -0.1611, 0.1247], + [ 0.0616, -0.0249, 0.0748, ..., 0.0023, -0.1942, -0.0098], + [-0.1697, -0.0741, -0.0935, ..., -0.1849, 0.0494, -0.1007]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 8.3819e-09, 1.2480e-07, ..., 1.5087e-07, + 8.8476e-08, 0.0000e+00], + [ 6.5193e-09, 1.3970e-08, -6.5193e-09, ..., 4.4703e-08, + 1.6578e-07, -9.3132e-10], + [ 1.2014e-07, -4.9360e-08, 8.9221e-07, ..., 1.0254e-06, + 3.4552e-07, 0.0000e+00], + ..., + [ 2.1420e-08, 1.0245e-08, 1.8254e-07, ..., 2.0675e-07, + 1.3411e-07, -1.8626e-09], + [ 9.3132e-10, 6.5193e-09, -2.8871e-08, ..., 4.6566e-09, + 5.4948e-08, 9.3132e-10], + [ 1.8626e-09, 1.8626e-09, 8.5682e-08, ..., 5.2154e-08, + 1.4836e-06, 0.0000e+00]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0063, -0.0337, 0.0100, -0.0221, 0.0187, 0.0136, 0.0249, 0.0012, + -0.0380, -0.0081], device='cuda:0'), grad: tensor([-4.3139e-06, 9.5088e-07, 3.4049e-06, -2.3935e-07, -4.6641e-06, + -4.7125e-06, 4.6417e-06, 7.9256e-07, 2.0955e-07, 3.9116e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 247.90, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4855 re_mapping 0.0053 re_causal 0.0153 /// teacc 99.01 lr 0.00010000 +Epoch 240, weight, value: tensor([[ 0.0285, -0.1508, -0.1290, ..., -0.2861, -0.0985, -0.1321], + [ 0.0437, -0.0693, 0.0383, ..., 0.0522, 0.0992, -0.0411], + [-0.0694, 0.1228, -0.1606, ..., 0.0618, 0.0757, -0.0412], + ..., + [-0.0731, -0.0857, -0.0820, ..., 0.0043, -0.1613, 0.1250], + [ 0.0619, -0.0242, 0.0751, ..., 0.0035, -0.1942, -0.0100], + [-0.1705, -0.0748, -0.0942, ..., -0.1856, 0.0498, -0.1009]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 4.6566e-09, ..., -1.2107e-08, + 1.0990e-07, 9.3132e-10], + [ 0.0000e+00, 2.7940e-09, 1.8999e-07, ..., 4.5914e-07, + 3.3528e-08, 1.8626e-09], + [ 0.0000e+00, -2.5146e-08, 2.4214e-08, ..., 2.1420e-08, + 2.1420e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3039e-08, 5.5879e-09, ..., -9.0897e-07, + 4.7497e-08, -8.3819e-09], + [ 0.0000e+00, 1.8626e-09, 2.0694e-06, ..., 2.2631e-06, + 1.2862e-06, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 5.5879e-09, ..., 5.7183e-07, + -8.3819e-09, 1.8626e-09]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0065, -0.0338, 0.0097, -0.0220, 0.0180, 0.0136, 0.0249, 0.0014, + -0.0370, -0.0077], device='cuda:0'), grad: tensor([-2.3283e-07, 3.1292e-06, 3.7253e-07, 4.7963e-07, -3.3155e-07, + -7.2002e-05, 4.4286e-05, -4.0680e-06, 2.5406e-05, 2.8033e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 249.65, cls_loss 0.0014 cls_loss_mapping 0.0021 cls_loss_causal 0.5108 re_mapping 0.0050 re_causal 0.0152 /// teacc 99.06 lr 0.00010000 +Epoch 241, weight, value: tensor([[ 0.0283, -0.1511, -0.1291, ..., -0.2866, -0.0986, -0.1323], + [ 0.0438, -0.0695, 0.0383, ..., 0.0522, 0.0992, -0.0411], + [-0.0695, 0.1232, -0.1608, ..., 0.0618, 0.0758, -0.0420], + ..., + [-0.0730, -0.0867, -0.0821, ..., 0.0042, -0.1619, 0.1252], + [ 0.0619, -0.0234, 0.0753, ..., 0.0041, -0.1943, -0.0100], + [-0.1708, -0.0750, -0.0949, ..., -0.1866, 0.0496, -0.1010]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.3819e-09, 3.9954e-07, ..., 7.4506e-08, + 4.2841e-08, 0.0000e+00], + [-1.3970e-08, 5.2154e-08, 1.5181e-07, ..., 5.9605e-08, + -2.5425e-07, 0.0000e+00], + [ 0.0000e+00, -8.4750e-08, 5.6624e-07, ..., 1.9651e-07, + 4.0978e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -1.1921e-07, 1.9539e-06, ..., 1.7695e-08, + 9.2201e-08, 0.0000e+00], + [ 1.2107e-08, 7.7300e-08, 1.1576e-06, ..., 3.1479e-07, + 1.8440e-07, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 7.9721e-06, ..., 7.8790e-07, + -1.5181e-07, 0.0000e+00]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0060, -0.0338, 0.0098, -0.0222, 0.0188, 0.0147, 0.0247, 0.0010, + -0.0367, -0.0084], device='cuda:0'), grad: tensor([-1.7416e-07, 1.6699e-06, 1.5227e-06, -5.9456e-05, 2.1178e-06, + 2.8029e-05, -4.3772e-07, 3.7700e-06, 3.4086e-06, 1.9550e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 240---------------------------------------------------- +epoch 240, time 266.03, cls_loss 0.0017 cls_loss_mapping 0.0032 cls_loss_causal 0.5200 re_mapping 0.0053 re_causal 0.0160 /// teacc 99.14 lr 0.00010000 +Epoch 242, weight, value: tensor([[ 0.0283, -0.1516, -0.1292, ..., -0.2870, -0.0985, -0.1326], + [ 0.0439, -0.0694, 0.0381, ..., 0.0518, 0.0993, -0.0417], + [-0.0696, 0.1237, -0.1610, ..., 0.0618, 0.0759, -0.0421], + ..., + [-0.0728, -0.0869, -0.0813, ..., 0.0049, -0.1622, 0.1256], + [ 0.0619, -0.0239, 0.0753, ..., 0.0041, -0.1945, -0.0101], + [-0.1714, -0.0756, -0.0958, ..., -0.1878, 0.0494, -0.1013]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.4506e-09, 3.9116e-08, ..., 2.6077e-08, + -3.6135e-07, 9.3132e-10], + [-6.0536e-08, 6.5193e-09, 8.9966e-07, ..., 5.2247e-07, + -2.4121e-07, 0.0000e+00], + [ 1.4901e-08, -9.3691e-07, 1.3206e-06, ..., 5.1130e-07, + -1.5274e-07, 0.0000e+00], + ..., + [ 1.3970e-08, 1.1176e-08, 2.0955e-07, ..., 1.1828e-07, + 7.3574e-08, 0.0000e+00], + [-1.5739e-07, 8.9500e-07, -3.1795e-06, ..., -1.5181e-06, + 1.6857e-07, 9.3132e-10], + [ 2.7008e-08, 9.3132e-10, 2.4308e-07, ..., 1.1921e-07, + -1.3039e-07, 0.0000e+00]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0053, -0.0342, 0.0098, -0.0239, 0.0189, 0.0175, 0.0247, 0.0020, + -0.0368, -0.0093], device='cuda:0'), grad: tensor([-9.4473e-06, 1.7779e-06, 8.2236e-07, -6.7130e-06, 1.5460e-06, + -3.0510e-06, 1.7405e-05, 6.3889e-07, -4.2841e-06, 1.2629e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 248.46, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4842 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.12 lr 0.00010000 +Epoch 243, weight, value: tensor([[ 0.0284, -0.1519, -0.1294, ..., -0.2874, -0.0986, -0.1326], + [ 0.0439, -0.0692, 0.0382, ..., 0.0520, 0.0994, -0.0417], + [-0.0696, 0.1239, -0.1613, ..., 0.0618, 0.0759, -0.0421], + ..., + [-0.0728, -0.0875, -0.0814, ..., 0.0048, -0.1628, 0.1258], + [ 0.0618, -0.0240, 0.0754, ..., 0.0041, -0.1945, -0.0101], + [-0.1713, -0.0768, -0.0964, ..., -0.1886, 0.0494, -0.1013]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.8476e-08, 5.7742e-08, ..., 2.4959e-07, + 3.5111e-07, 0.0000e+00], + [ 0.0000e+00, 2.7008e-08, -7.5437e-08, ..., 1.4063e-07, + 8.0094e-08, 2.7940e-09], + [ 0.0000e+00, -3.9022e-07, 8.9407e-08, ..., -6.7428e-07, + -1.3094e-06, 2.7940e-09], + ..., + [ 0.0000e+00, 2.6077e-08, 3.6322e-08, ..., -4.7870e-07, + 5.4948e-08, -8.3819e-09], + [ 0.0000e+00, 1.3597e-07, 4.0978e-08, ..., 2.5798e-07, + 4.8429e-07, 1.8626e-09], + [ 0.0000e+00, 4.2841e-08, 5.5879e-09, ..., 2.6822e-07, + 4.6380e-07, 0.0000e+00]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0051, -0.0341, 0.0098, -0.0240, 0.0187, 0.0177, 0.0247, 0.0019, + -0.0368, -0.0095], device='cuda:0'), grad: tensor([ 1.0198e-06, 5.9046e-07, -3.6843e-06, -2.3004e-07, -1.2536e-06, + 7.7207e-07, 7.0408e-07, -1.4622e-06, 1.6419e-06, 1.8943e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 247.77, cls_loss 0.0020 cls_loss_mapping 0.0032 cls_loss_causal 0.5290 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.09 lr 0.00010000 +Epoch 244, weight, value: tensor([[ 0.0282, -0.1522, -0.1298, ..., -0.2882, -0.0987, -0.1329], + [ 0.0441, -0.0690, 0.0381, ..., 0.0520, 0.0994, -0.0418], + [-0.0697, 0.1244, -0.1616, ..., 0.0618, 0.0760, -0.0434], + ..., + [-0.0728, -0.0878, -0.0818, ..., 0.0045, -0.1631, 0.1262], + [ 0.0618, -0.0250, 0.0753, ..., 0.0038, -0.1951, -0.0103], + [-0.1722, -0.0772, -0.0966, ..., -0.1897, 0.0494, -0.1014]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 3.2596e-08, ..., 1.2107e-08, + 1.0245e-08, 9.3132e-10], + [ 0.0000e+00, -3.0734e-08, -5.0850e-07, ..., -2.1420e-07, + -3.8836e-07, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 3.3528e-07, ..., 1.4435e-07, + 1.1642e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 3.1758e-07, ..., 1.4622e-07, + 1.7136e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -1.0384e-06, ..., -2.7101e-07, + 2.5146e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.6205e-07, ..., 5.4948e-08, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0052, -0.0343, 0.0098, -0.0239, 0.0189, 0.0175, 0.0249, 0.0016, + -0.0378, -0.0093], device='cuda:0'), grad: tensor([-1.9837e-07, -9.8161e-07, 7.3854e-07, -1.5460e-07, 5.7742e-08, + 7.6089e-07, 5.2527e-07, 6.5658e-07, -1.8878e-06, 4.6939e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 247.59, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4788 re_mapping 0.0053 re_causal 0.0155 /// teacc 98.96 lr 0.00010000 +Epoch 245, weight, value: tensor([[ 0.0284, -0.1524, -0.1298, ..., -0.2885, -0.0986, -0.1330], + [ 0.0442, -0.0673, 0.0374, ..., 0.0515, 0.0997, -0.0420], + [-0.0698, 0.1230, -0.1629, ..., 0.0615, 0.0756, -0.0428], + ..., + [-0.0728, -0.0878, -0.0810, ..., 0.0054, -0.1629, 0.1263], + [ 0.0618, -0.0245, 0.0755, ..., 0.0043, -0.1952, -0.0103], + [-0.1713, -0.0774, -0.0969, ..., -0.1901, 0.0495, -0.1015]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.3970e-08, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.5739e-07, ..., -2.5705e-07, + -6.2957e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.9639e-07, ..., 7.7020e-07, + 4.0606e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3621e-07, ..., 5.8580e-07, + 3.0361e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -6.8918e-08, ..., -1.8626e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.9360e-08, ..., 6.5193e-08, + 3.9116e-08, 0.0000e+00]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0048, -0.0350, 0.0094, -0.0238, 0.0191, 0.0173, 0.0248, 0.0025, + -0.0376, -0.0092], device='cuda:0'), grad: tensor([ 4.9360e-08, -6.6217e-07, 2.4624e-06, -4.3735e-06, -9.1270e-08, + 9.7603e-07, 3.2783e-07, 1.2843e-06, -2.3656e-07, 2.4959e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 247.62, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.4438 re_mapping 0.0052 re_causal 0.0147 /// teacc 99.05 lr 0.00010000 +Epoch 246, weight, value: tensor([[ 0.0284, -0.1521, -0.1300, ..., -0.2886, -0.0986, -0.1331], + [ 0.0442, -0.0673, 0.0373, ..., 0.0511, 0.0998, -0.0440], + [-0.0701, 0.1230, -0.1630, ..., 0.0614, 0.0756, -0.0434], + ..., + [-0.0723, -0.0880, -0.0809, ..., 0.0058, -0.1631, 0.1282], + [ 0.0618, -0.0246, 0.0781, ..., 0.0044, -0.1954, -0.0103], + [-0.1716, -0.0785, -0.0973, ..., -0.1907, 0.0494, -0.1017]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-08, 1.2107e-08, ..., 1.3039e-08, + 4.3772e-08, 0.0000e+00], + [ 0.0000e+00, -4.5449e-07, -1.2936e-06, ..., -1.7146e-06, + -2.6803e-06, 0.0000e+00], + [ 0.0000e+00, 1.7695e-07, 7.3481e-07, ..., 8.9779e-07, + 1.3504e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3749e-07, 3.5390e-07, ..., 5.9698e-07, + 8.1398e-07, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, -2.2352e-08, ..., 1.4901e-08, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, 8.3819e-09, ..., 6.5193e-09, + -5.0589e-06, 0.0000e+00]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0050, -0.0354, 0.0093, -0.0238, 0.0193, 0.0169, 0.0244, 0.0029, + -0.0354, -0.0095], device='cuda:0'), grad: tensor([-4.6007e-06, -5.2266e-06, 2.9169e-06, 7.1712e-08, 9.1493e-06, + 1.7975e-07, 1.3448e-06, 1.6689e-06, 3.1777e-06, -8.7246e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 247.87, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5163 re_mapping 0.0052 re_causal 0.0154 /// teacc 98.97 lr 0.00010000 +Epoch 247, weight, value: tensor([[ 0.0283, -0.1525, -0.1310, ..., -0.2893, -0.0987, -0.1333], + [ 0.0446, -0.0677, 0.0373, ..., 0.0510, 0.0998, -0.0440], + [-0.0702, 0.1256, -0.1631, ..., 0.0617, 0.0761, -0.0435], + ..., + [-0.0706, -0.0913, -0.0809, ..., 0.0056, -0.1645, 0.1282], + [ 0.0614, -0.0251, 0.0780, ..., 0.0043, -0.1959, -0.0103], + [-0.1737, -0.0789, -0.0994, ..., -0.1920, 0.0494, -0.1017]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 5.2154e-08, ..., 7.0781e-08, + 8.1025e-08, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, 2.7940e-08, ..., 5.3085e-08, + 2.8871e-08, 0.0000e+00], + [ 0.0000e+00, -8.3912e-07, 1.0524e-07, ..., -9.2294e-07, + -1.3364e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 1.4901e-08, 4.8429e-08, ..., 1.7509e-07, + 2.2259e-07, 0.0000e+00], + [ 0.0000e+00, 7.7207e-07, -1.1176e-08, ..., 5.7742e-07, + 8.4843e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.3411e-07, ..., 5.1223e-08, + 1.2107e-08, 0.0000e+00]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0050, -0.0354, 0.0096, -0.0236, 0.0195, 0.0168, 0.0243, 0.0028, + -0.0357, -0.0098], device='cuda:0'), grad: tensor([ 3.7067e-07, 2.1327e-07, -3.2298e-06, -1.7546e-06, 1.8626e-09, + 4.2003e-07, 3.0920e-07, 4.8615e-07, 2.6524e-06, 5.3458e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 247.66, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4771 re_mapping 0.0054 re_causal 0.0150 /// teacc 98.99 lr 0.00010000 +Epoch 248, weight, value: tensor([[ 0.0286, -0.1526, -0.1312, ..., -0.2896, -0.0989, -0.1334], + [ 0.0446, -0.0679, 0.0374, ..., 0.0510, 0.0998, -0.0443], + [-0.0703, 0.1257, -0.1633, ..., 0.0617, 0.0761, -0.0434], + ..., + [-0.0705, -0.0912, -0.0809, ..., 0.0057, -0.1645, 0.1284], + [ 0.0614, -0.0253, 0.0780, ..., 0.0043, -0.1961, -0.0104], + [-0.1739, -0.0790, -0.0997, ..., -0.1924, 0.0495, -0.1018]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 4.6566e-09, 5.5879e-09, ..., 1.7695e-08, + 2.9802e-08, 9.3132e-10], + [ 9.3132e-10, 2.9709e-07, -7.7561e-06, ..., -1.7762e-05, + -2.1294e-05, 0.0000e+00], + [-6.7987e-08, -5.0757e-07, 1.7537e-06, ..., 3.6787e-06, + 3.8594e-06, 0.0000e+00], + ..., + [ 7.4506e-09, 1.3039e-08, 5.8077e-06, ..., 1.3448e-05, + 1.6436e-05, 0.0000e+00], + [ 9.3132e-10, 4.6566e-09, -4.3772e-08, ..., -1.3039e-08, + 2.7008e-08, 0.0000e+00], + [ 0.0000e+00, 1.3597e-07, 3.3528e-08, ..., 1.0338e-07, + 3.3807e-07, 0.0000e+00]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0052, -0.0355, 0.0095, -0.0235, 0.0196, 0.0167, 0.0244, 0.0029, + -0.0358, -0.0099], device='cuda:0'), grad: tensor([ 6.3330e-08, -5.2005e-05, 1.0028e-05, 8.7451e-07, 1.5553e-07, + -7.1898e-07, 2.6356e-07, 3.9786e-05, 2.8592e-07, 1.2787e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 247.79, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5515 re_mapping 0.0053 re_causal 0.0161 /// teacc 99.01 lr 0.00010000 +Epoch 249, weight, value: tensor([[ 0.0291, -0.1526, -0.1314, ..., -0.2914, -0.0991, -0.1335], + [ 0.0446, -0.0679, 0.0375, ..., 0.0510, 0.1000, -0.0443], + [-0.0706, 0.1256, -0.1635, ..., 0.0616, 0.0761, -0.0435], + ..., + [-0.0701, -0.0910, -0.0810, ..., 0.0057, -0.1652, 0.1284], + [ 0.0615, -0.0251, 0.0782, ..., 0.0046, -0.1962, -0.0103], + [-0.1760, -0.0803, -0.1014, ..., -0.1938, 0.0495, -0.1018]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1176e-08, 6.4261e-08, ..., 3.1665e-08, + 2.7008e-08, 0.0000e+00], + [ 0.0000e+00, 4.1444e-07, -1.0632e-05, ..., -2.0154e-06, + -4.8429e-07, 0.0000e+00], + [ 0.0000e+00, -3.5483e-07, 2.7288e-07, ..., 3.2596e-08, + -8.5961e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.6764e-08, 2.7381e-07, ..., 1.3039e-08, + 9.3132e-08, 0.0000e+00], + [ 0.0000e+00, -1.1828e-07, 5.0887e-06, ..., 1.1893e-06, + 5.3924e-07, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 2.9802e-07, ..., 1.3690e-07, + -8.9407e-08, 0.0000e+00]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0054, -0.0354, 0.0093, -0.0234, 0.0200, 0.0166, 0.0243, 0.0030, + -0.0357, -0.0101], device='cuda:0'), grad: tensor([ 2.0489e-07, -2.9474e-05, -2.4587e-07, -5.1484e-06, 8.3912e-07, + 1.3739e-05, 5.6997e-06, 5.0385e-07, 1.3284e-05, 5.5693e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 247.63, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4958 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 250, weight, value: tensor([[ 0.0292, -0.1528, -0.1314, ..., -0.2916, -0.0991, -0.1335], + [ 0.0447, -0.0679, 0.0375, ..., 0.0510, 0.1001, -0.0443], + [-0.0706, 0.1257, -0.1637, ..., 0.0616, 0.0761, -0.0435], + ..., + [-0.0701, -0.0912, -0.0810, ..., 0.0058, -0.1655, 0.1285], + [ 0.0615, -0.0253, 0.0782, ..., 0.0044, -0.1965, -0.0103], + [-0.1762, -0.0805, -0.1019, ..., -0.1946, 0.0496, -0.1018]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.3819e-09, 1.0245e-08, ..., 1.5832e-08, + 8.5682e-08, 0.0000e+00], + [ 0.0000e+00, 2.1420e-07, -3.5483e-07, ..., 2.9244e-07, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, -2.7381e-07, 3.3528e-08, ..., -4.5076e-07, + -4.4517e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 3.1665e-08, -2.7008e-08, ..., -3.7253e-08, + 7.3574e-08, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 2.2352e-08, ..., 2.0489e-08, + 8.0094e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 4.4703e-08, ..., 8.4750e-08, + 5.4855e-07, 0.0000e+00]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0052, -0.0354, 0.0093, -0.0232, 0.0200, 0.0163, 0.0242, 0.0031, + -0.0359, -0.0105], device='cuda:0'), grad: tensor([ 2.4028e-07, 2.0396e-07, -1.0524e-06, 4.3213e-07, -2.7157e-06, + -5.9083e-06, 7.5772e-06, -4.3660e-06, 7.8417e-07, 4.7982e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 247.53, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4918 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.07 lr 0.00010000 +Epoch 251, weight, value: tensor([[ 0.0292, -0.1529, -0.1317, ..., -0.2918, -0.0992, -0.1335], + [ 0.0449, -0.0678, 0.0375, ..., 0.0509, 0.1005, -0.0443], + [-0.0707, 0.1258, -0.1639, ..., 0.0615, 0.0761, -0.0444], + ..., + [-0.0701, -0.0913, -0.0809, ..., 0.0060, -0.1662, 0.1286], + [ 0.0615, -0.0253, 0.0783, ..., 0.0045, -0.1966, -0.0103], + [-0.1763, -0.0808, -0.1025, ..., -0.1951, 0.0495, -0.1018]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 3.7253e-09, ..., 3.7253e-09, + 1.3039e-08, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, -7.2177e-07, ..., -4.0140e-07, + -1.4137e-06, 9.3132e-10], + [ 0.0000e+00, -3.4459e-08, 5.2806e-07, ..., 2.9523e-07, + 1.0217e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 2.1420e-08, 4.3772e-08, ..., 2.5146e-08, + 1.0990e-07, -9.3132e-10], + [ 0.0000e+00, 5.5879e-09, -1.4901e-08, ..., -6.5193e-09, + 2.8871e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.5832e-08, ..., 1.0245e-08, + -7.4506e-08, 0.0000e+00]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0054, -0.0354, 0.0092, -0.0232, 0.0201, 0.0164, 0.0241, 0.0031, + -0.0359, -0.0107], device='cuda:0'), grad: tensor([-4.2990e-06, -2.1532e-06, 1.5777e-06, 5.7742e-08, 3.9581e-07, + 3.6731e-06, 7.5158e-07, 2.6450e-07, 1.3039e-08, -2.8498e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 247.54, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.5049 re_mapping 0.0051 re_causal 0.0160 /// teacc 99.02 lr 0.00010000 +Epoch 252, weight, value: tensor([[ 0.0290, -0.1535, -0.1328, ..., -0.2924, -0.0993, -0.1335], + [ 0.0449, -0.0679, 0.0376, ..., 0.0509, 0.1006, -0.0443], + [-0.0706, 0.1260, -0.1639, ..., 0.0615, 0.0761, -0.0444], + ..., + [-0.0701, -0.0914, -0.0809, ..., 0.0060, -0.1667, 0.1287], + [ 0.0615, -0.0255, 0.0784, ..., 0.0044, -0.1966, -0.0103], + [-0.1764, -0.0811, -0.1035, ..., -0.1954, 0.0499, -0.1019]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 1.0245e-08, + 2.4382e-06, 9.3132e-10], + [-0.0000e+00, 0.0000e+00, 2.2445e-06, ..., 2.1663e-06, + -1.7323e-07, 2.4214e-07], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 2.7940e-08, + 1.4901e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -3.4440e-06, ..., -3.1460e-06, + 7.1712e-08, -3.4273e-07], + [ 0.0000e+00, 0.0000e+00, 7.0315e-07, ..., 6.2492e-07, + 3.4459e-08, 6.7055e-08], + [ 0.0000e+00, 0.0000e+00, 1.1921e-07, ..., 1.0245e-07, + -2.4885e-06, 1.0245e-08]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0057, -0.0354, 0.0092, -0.0230, 0.0196, 0.0163, 0.0242, 0.0031, + -0.0359, -0.0102], device='cuda:0'), grad: tensor([ 5.0105e-06, 1.2785e-05, 1.9185e-07, 7.5623e-07, 3.7719e-07, + 5.9605e-08, 2.6636e-07, -1.8418e-05, 3.8333e-06, -4.8354e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 247.69, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5013 re_mapping 0.0049 re_causal 0.0151 /// teacc 99.04 lr 0.00010000 +Epoch 253, weight, value: tensor([[ 0.0290, -0.1532, -0.1328, ..., -0.2925, -0.0994, -0.1337], + [ 0.0449, -0.0673, 0.0377, ..., 0.0510, 0.1009, -0.0443], + [-0.0706, 0.1258, -0.1643, ..., 0.0615, 0.0760, -0.0444], + ..., + [-0.0701, -0.0915, -0.0810, ..., 0.0059, -0.1672, 0.1288], + [ 0.0616, -0.0254, 0.0786, ..., 0.0045, -0.1967, -0.0104], + [-0.1765, -0.0815, -0.1040, ..., -0.1957, 0.0499, -0.1019]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 9.3132e-09, ..., 9.3132e-09, + 3.6322e-08, 1.8626e-09], + [ 1.4901e-08, 1.8626e-09, -1.1828e-07, ..., -5.3085e-08, + -1.4994e-07, 1.1176e-08], + [ 2.7940e-09, -5.0291e-08, 5.5879e-09, ..., -2.7940e-08, + 2.7940e-09, 1.8626e-09], + ..., + [-4.0047e-08, 3.9116e-08, 5.5879e-08, ..., -4.1910e-08, + 1.4156e-07, -2.9802e-08], + [ 9.3132e-10, 2.7940e-09, -0.0000e+00, ..., 4.0978e-08, + 5.4017e-08, 9.3132e-10], + [ 1.7695e-08, 9.3132e-10, 1.2107e-08, ..., 2.7008e-08, + 8.8383e-07, 1.3039e-08]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0051, -0.0353, 0.0091, -0.0230, 0.0197, 0.0162, 0.0240, 0.0030, + -0.0356, -0.0105], device='cuda:0'), grad: tensor([-4.6909e-05, 5.0664e-07, 4.6827e-06, 4.9844e-06, -6.1616e-06, + 6.3106e-06, 6.6124e-07, 9.4343e-07, 6.8732e-07, 3.4273e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 247.50, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.4568 re_mapping 0.0050 re_causal 0.0146 /// teacc 99.00 lr 0.00010000 +Epoch 254, weight, value: tensor([[ 0.0289, -0.1536, -0.1329, ..., -0.2928, -0.0997, -0.1338], + [ 0.0449, -0.0674, 0.0381, ..., 0.0516, 0.1014, -0.0444], + [-0.0706, 0.1260, -0.1644, ..., 0.0615, 0.0761, -0.0445], + ..., + [-0.0701, -0.0917, -0.0814, ..., 0.0053, -0.1693, 0.1310], + [ 0.0619, -0.0256, 0.0790, ..., 0.0046, -0.1968, -0.0104], + [-0.1793, -0.0833, -0.1070, ..., -0.1970, 0.0503, -0.1048]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.1106e-07, 5.5879e-09, ..., 3.9488e-07, + 4.5635e-08, 0.0000e+00], + [ 0.0000e+00, 6.5193e-09, -9.6858e-08, ..., -3.2596e-08, + -6.7987e-08, 0.0000e+00], + [ 0.0000e+00, -1.0841e-06, 4.0978e-08, ..., -1.3346e-06, + -2.9709e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 4.9081e-07, 6.3330e-08, ..., 6.4354e-07, + 1.5460e-07, 0.0000e+00], + [ 0.0000e+00, 3.3528e-08, -8.3819e-08, ..., -3.7253e-09, + 1.2293e-07, 0.0000e+00], + [ 0.0000e+00, 5.6811e-08, 4.6566e-09, ..., 6.2399e-08, + 6.9849e-08, 0.0000e+00]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0043, -0.0348, 0.0091, -0.0225, 0.0192, 0.0155, 0.0239, 0.0031, + -0.0353, -0.0131], device='cuda:0'), grad: tensor([ 1.1222e-06, -1.0338e-07, -3.7383e-06, 6.7893e-07, -1.2387e-07, + 1.7695e-08, -5.0291e-08, 1.8533e-06, 1.0151e-07, 2.4494e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 247.61, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4813 re_mapping 0.0050 re_causal 0.0153 /// teacc 98.97 lr 0.00010000 +Epoch 255, weight, value: tensor([[ 0.0289, -0.1546, -0.1331, ..., -0.2933, -0.0999, -0.1338], + [ 0.0454, -0.0674, 0.0382, ..., 0.0516, 0.1015, -0.0445], + [-0.0707, 0.1262, -0.1644, ..., 0.0615, 0.0765, -0.0448], + ..., + [-0.0702, -0.0916, -0.0815, ..., 0.0052, -0.1698, 0.1311], + [ 0.0619, -0.0260, 0.0790, ..., 0.0045, -0.1972, -0.0104], + [-0.1794, -0.0840, -0.1076, ..., -0.1982, 0.0504, -0.1048]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 5.8673e-08, ..., 5.7742e-08, + 9.4064e-08, 0.0000e+00], + [-1.8626e-09, 1.8626e-09, 3.5111e-07, ..., 3.5577e-07, + -4.8708e-07, 5.5879e-09], + [ 0.0000e+00, -5.4948e-08, 3.5577e-07, ..., 1.0058e-07, + -1.4249e-07, 9.3132e-10], + ..., + [ 9.3132e-10, 7.4506e-09, 1.7695e-07, ..., 5.1223e-08, + 1.4808e-07, -1.3039e-08], + [ 0.0000e+00, 2.7008e-08, -1.1437e-06, ..., -7.7486e-07, + 6.8918e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 2.5146e-08, + 2.3283e-08, 4.6566e-09]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0046, -0.0348, 0.0093, -0.0225, 0.0175, 0.0156, 0.0240, 0.0031, + -0.0356, -0.0125], device='cuda:0'), grad: tensor([ 2.1234e-07, 2.1886e-06, 8.2236e-07, 1.2293e-07, 9.0338e-08, + 9.0338e-08, 3.3900e-07, 2.5798e-07, -4.2766e-06, 1.3504e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 247.86, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.5293 re_mapping 0.0049 re_causal 0.0152 /// teacc 99.01 lr 0.00010000 +Epoch 256, weight, value: tensor([[ 0.0291, -0.1551, -0.1341, ..., -0.2936, -0.1000, -0.1341], + [ 0.0456, -0.0673, 0.0385, ..., 0.0521, 0.1020, -0.0447], + [-0.0707, 0.1261, -0.1648, ..., 0.0614, 0.0765, -0.0457], + ..., + [-0.0702, -0.0918, -0.0818, ..., 0.0048, -0.1712, 0.1310], + [ 0.0619, -0.0261, 0.0792, ..., 0.0048, -0.1975, -0.0096], + [-0.1797, -0.0845, -0.1080, ..., -0.1992, 0.0511, -0.1049]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.5832e-08, 4.6566e-09, ..., 2.3283e-08, + 2.9802e-07, 0.0000e+00], + [-4.6566e-09, 2.2352e-08, -1.1176e-07, ..., 3.7253e-08, + -2.4214e-08, 7.4506e-09], + [ 9.3132e-10, -1.6019e-07, 2.4214e-08, ..., -2.0023e-07, + -4.1816e-07, 9.3132e-10], + ..., + [ 1.8626e-09, 2.2352e-08, 3.7253e-08, ..., -2.9802e-08, + 9.6858e-08, -1.8626e-08], + [ 0.0000e+00, 1.9558e-08, 1.8626e-09, ..., 2.1420e-08, + 1.8161e-07, 9.3132e-10], + [ 9.3132e-10, 1.1176e-08, 1.1176e-08, ..., 2.1420e-08, + -8.4750e-08, 9.3132e-10]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0049, -0.0345, 0.0091, -0.0223, 0.0162, 0.0158, 0.0238, 0.0028, + -0.0355, -0.0121], device='cuda:0'), grad: tensor([ 6.2771e-07, 1.0524e-07, -7.0408e-07, 6.3144e-07, 5.4855e-07, + 2.9989e-07, -1.6121e-06, 1.2852e-07, 6.1654e-07, -6.2305e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 247.95, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.4694 re_mapping 0.0052 re_causal 0.0152 /// teacc 98.99 lr 0.00010000 +Epoch 257, weight, value: tensor([[ 0.0292, -0.1557, -0.1342, ..., -0.2941, -0.1024, -0.1342], + [ 0.0457, -0.0673, 0.0385, ..., 0.0516, 0.1015, -0.0448], + [-0.0705, 0.1264, -0.1650, ..., 0.0613, 0.0765, -0.0463], + ..., + [-0.0703, -0.0919, -0.0818, ..., 0.0055, -0.1708, 0.1312], + [ 0.0619, -0.0264, 0.0794, ..., 0.0047, -0.1977, -0.0097], + [-0.1799, -0.0850, -0.1081, ..., -0.1997, 0.0531, -0.1049]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 1.3970e-08, ..., 1.0245e-08, + 3.0734e-08, 0.0000e+00], + [-2.0489e-08, 3.8184e-08, -1.6391e-07, ..., 6.4261e-08, + -1.0803e-07, 4.6566e-09], + [ 9.3132e-10, -1.2666e-07, 1.7695e-07, ..., -5.5879e-08, + -1.8626e-07, 0.0000e+00], + ..., + [ 5.5879e-09, 1.0245e-08, 1.3225e-07, ..., 2.6077e-08, + 1.0990e-07, -1.2107e-08], + [ 9.3132e-10, 5.8673e-08, -1.3970e-08, ..., 7.9162e-08, + 1.5926e-07, 9.3132e-10], + [ 1.1176e-08, 9.3132e-10, 1.6112e-07, ..., 4.5635e-08, + -2.3525e-06, 3.7253e-09]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0072, -0.0348, 0.0090, -0.0223, 0.0164, 0.0157, 0.0241, 0.0032, + -0.0356, -0.0103], device='cuda:0'), grad: tensor([-1.5711e-06, -1.4901e-08, 4.2841e-08, 2.4252e-06, 8.5831e-06, + -2.6524e-06, -1.5441e-06, 3.6135e-07, 1.1018e-06, -6.7130e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 247.97, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4910 re_mapping 0.0053 re_causal 0.0151 /// teacc 99.06 lr 0.00010000 +Epoch 258, weight, value: tensor([[ 0.0290, -0.1585, -0.1346, ..., -0.2946, -0.1028, -0.1343], + [ 0.0457, -0.0676, 0.0382, ..., 0.0512, 0.1015, -0.0450], + [-0.0705, 0.1270, -0.1652, ..., 0.0614, 0.0768, -0.0465], + ..., + [-0.0704, -0.0925, -0.0814, ..., 0.0059, -0.1715, 0.1314], + [ 0.0617, -0.0251, 0.0798, ..., 0.0045, -0.1985, -0.0100], + [-0.1800, -0.0856, -0.1086, ..., -0.2001, 0.0531, -0.1049]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.3970e-08, 7.6368e-08, ..., 7.3574e-08, + 4.7497e-08, 0.0000e+00], + [ 9.3132e-09, 6.5193e-09, -1.9651e-07, ..., 1.7695e-08, + -1.5832e-07, 0.0000e+00], + [ 1.8626e-09, -2.2259e-07, 2.5798e-07, ..., 1.4901e-08, + -6.6124e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 4.0978e-08, 1.5367e-07, ..., 9.3132e-09, + 2.5053e-07, -0.0000e+00], + [ 2.7940e-09, -7.3574e-08, -2.4773e-07, ..., 2.3283e-08, + -7.9162e-08, 0.0000e+00], + [ 1.8626e-09, 7.5437e-08, 2.3842e-07, ..., 4.9360e-08, + 1.6298e-07, 0.0000e+00]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0075, -0.0353, 0.0091, -0.0222, 0.0158, 0.0152, 0.0252, 0.0035, + -0.0356, -0.0104], device='cuda:0'), grad: tensor([ 2.3749e-07, -1.6578e-07, -1.7425e-06, -8.7358e-07, 5.0571e-07, + -3.4459e-08, 1.0561e-06, 6.6590e-07, -7.8604e-07, 1.1306e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 247.83, cls_loss 0.0017 cls_loss_mapping 0.0036 cls_loss_causal 0.4669 re_mapping 0.0053 re_causal 0.0149 /// teacc 99.09 lr 0.00010000 +Epoch 259, weight, value: tensor([[ 0.0289, -0.1581, -0.1336, ..., -0.2953, -0.1034, -0.1355], + [ 0.0462, -0.0677, 0.0382, ..., 0.0512, 0.1015, -0.0452], + [-0.0706, 0.1275, -0.1655, ..., 0.0614, 0.0772, -0.0458], + ..., + [-0.0705, -0.0928, -0.0814, ..., 0.0059, -0.1720, 0.1316], + [ 0.0612, -0.0256, 0.0776, ..., 0.0042, -0.2022, -0.0102], + [-0.1802, -0.0868, -0.1100, ..., -0.2014, 0.0532, -0.1049]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.7695e-08, -4.2506e-06, ..., 4.1910e-08, + -4.9397e-06, 0.0000e+00], + [ 8.3819e-09, 1.2852e-07, 9.0897e-07, ..., 1.6019e-07, + 1.3076e-06, 0.0000e+00], + [ 0.0000e+00, -1.1083e-07, 1.2154e-06, ..., -1.3504e-07, + 5.8021e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 6.7055e-08, ..., 3.2596e-08, + 6.6124e-08, -0.0000e+00], + [ 0.0000e+00, -1.1828e-07, -1.8664e-06, ..., -2.7567e-07, + 5.1223e-08, 0.0000e+00], + [ 9.3132e-10, 7.4506e-09, 1.3225e-07, ..., 1.9558e-08, + 3.8184e-08, 0.0000e+00]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0076, -0.0353, 0.0091, -0.0219, 0.0157, 0.0147, 0.0268, 0.0035, + -0.0385, -0.0103], device='cuda:0'), grad: tensor([-3.5793e-05, 8.3074e-06, 6.6459e-06, 1.3746e-06, -1.4417e-06, + 5.8673e-08, 2.3410e-05, 1.7136e-07, -3.1590e-06, 4.2748e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 247.53, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5049 re_mapping 0.0050 re_causal 0.0156 /// teacc 99.10 lr 0.00010000 +Epoch 260, weight, value: tensor([[ 0.0289, -0.1609, -0.1341, ..., -0.2957, -0.1035, -0.1356], + [ 0.0466, -0.0679, 0.0385, ..., 0.0517, 0.1018, -0.0461], + [-0.0706, 0.1285, -0.1657, ..., 0.0616, 0.0773, -0.0461], + ..., + [-0.0707, -0.0941, -0.0817, ..., 0.0052, -0.1728, 0.1322], + [ 0.0612, -0.0231, 0.0776, ..., 0.0040, -0.2024, -0.0102], + [-0.1804, -0.0883, -0.1106, ..., -0.2026, 0.0529, -0.1050]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 6.5193e-09, ..., 7.4506e-09, + 6.0499e-06, 0.0000e+00], + [ 0.0000e+00, 2.0023e-08, 9.3132e-09, ..., 4.5169e-08, + 1.6065e-07, 0.0000e+00], + [ 4.6566e-10, -1.8943e-06, 9.0804e-08, ..., -3.1739e-06, + -2.0415e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8282e-06, 2.8871e-08, ..., 3.1497e-06, + 2.3134e-06, 0.0000e+00], + [ 4.6566e-10, 1.8626e-08, 4.9826e-08, ..., 5.1223e-08, + 4.2701e-07, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 2.7940e-09, ..., 3.7253e-09, + 3.1013e-07, 0.0000e+00]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0074, -0.0350, 0.0092, -0.0218, 0.0161, 0.0148, 0.0267, 0.0032, + -0.0387, -0.0107], device='cuda:0'), grad: tensor([ 2.2292e-05, 6.7893e-07, -9.0152e-06, -3.6741e-07, 4.8392e-06, + 1.1921e-06, -3.2991e-05, 1.0371e-05, 1.7099e-06, 1.2629e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 247.95, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.5100 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.04 lr 0.00010000 +Epoch 261, weight, value: tensor([[ 0.0294, -0.1612, -0.1341, ..., -0.2969, -0.1035, -0.1360], + [ 0.0467, -0.0683, 0.0387, ..., 0.0518, 0.1019, -0.0461], + [-0.0710, 0.1289, -0.1662, ..., 0.0616, 0.0775, -0.0463], + ..., + [-0.0708, -0.0945, -0.0819, ..., 0.0049, -0.1738, 0.1322], + [ 0.0612, -0.0232, 0.0777, ..., 0.0039, -0.2026, -0.0103], + [-0.1807, -0.0888, -0.1111, ..., -0.2027, 0.0528, -0.1050]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 1.2293e-07, ..., 8.0559e-08, + 1.3225e-07, 0.0000e+00], + [ 4.6566e-10, 1.3970e-09, -1.2564e-06, ..., -8.5449e-07, + -1.3309e-06, 0.0000e+00], + [ 3.7253e-09, -1.7229e-08, 5.4995e-07, ..., 4.2748e-07, + 6.2492e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 1.2107e-08, 2.7521e-07, ..., 9.6858e-08, + 2.6869e-07, 0.0000e+00], + [ 3.2596e-09, 1.8626e-09, 6.0536e-08, ..., 5.4948e-08, + 1.2433e-07, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 5.0291e-08, ..., 4.0047e-08, + 4.9360e-08, 0.0000e+00]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0072, -0.0349, 0.0092, -0.0205, 0.0165, 0.0131, 0.0267, 0.0030, + -0.0388, -0.0109], device='cuda:0'), grad: tensor([ 3.1386e-07, -2.6729e-06, 1.3392e-06, -5.2620e-08, -1.7695e-08, + 3.4180e-07, -3.8464e-07, 4.3120e-07, 4.4284e-07, 2.7148e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 247.75, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.5037 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.12 lr 0.00010000 +Epoch 262, weight, value: tensor([[ 0.0304, -0.1614, -0.1340, ..., -0.2977, -0.1038, -0.1361], + [ 0.0466, -0.0685, 0.0389, ..., 0.0520, 0.1021, -0.0461], + [-0.0713, 0.1299, -0.1665, ..., 0.0617, 0.0777, -0.0464], + ..., + [-0.0702, -0.0949, -0.0820, ..., 0.0049, -0.1744, 0.1323], + [ 0.0612, -0.0247, 0.0789, ..., 0.0042, -0.2028, -0.0103], + [-0.1820, -0.0892, -0.1131, ..., -0.2041, 0.0529, -0.1050]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 2.8871e-08, ..., 3.8184e-08, + 2.6077e-08, 9.3132e-10], + [ 0.0000e+00, 1.6764e-08, 3.8836e-07, ..., 1.8347e-07, + -4.8894e-07, -7.5437e-08], + [ 0.0000e+00, -8.4750e-08, 2.4028e-07, ..., 1.8347e-07, + -9.3132e-10, 4.6566e-09], + ..., + [ 0.0000e+00, 5.5879e-09, 1.0151e-07, ..., -2.5090e-06, + 4.0978e-08, 1.8626e-09], + [ 0.0000e+00, 4.6566e-09, -4.0196e-06, ..., -2.2911e-06, + -5.4389e-07, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 2.3283e-08, ..., 2.1979e-06, + 3.4124e-06, 0.0000e+00]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0069, -0.0348, 0.0093, -0.0201, 0.0169, 0.0125, 0.0263, 0.0028, + -0.0379, -0.0113], device='cuda:0'), grad: tensor([ 1.4342e-07, 1.0394e-06, 6.2119e-07, 2.2259e-06, -8.8662e-06, + -1.0207e-06, 3.8259e-06, -9.7603e-06, -6.4634e-06, 1.8224e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 247.75, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.4844 re_mapping 0.0049 re_causal 0.0145 /// teacc 99.00 lr 0.00010000 +Epoch 263, weight, value: tensor([[ 0.0303, -0.1613, -0.1343, ..., -0.2987, -0.1038, -0.1367], + [ 0.0464, -0.0685, 0.0392, ..., 0.0529, 0.1027, -0.0461], + [-0.0714, 0.1301, -0.1674, ..., 0.0613, 0.0775, -0.0464], + ..., + [-0.0700, -0.0950, -0.0822, ..., 0.0043, -0.1758, 0.1323], + [ 0.0612, -0.0251, 0.0791, ..., 0.0043, -0.2030, -0.0103], + [-0.1822, -0.0909, -0.1141, ..., -0.2047, 0.0530, -0.1049]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 6.5193e-09, ..., 4.6566e-08, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 1.0245e-08, -1.1493e-06, ..., -8.1304e-07, + -6.7335e-07, 0.0000e+00], + [ 0.0000e+00, 6.2212e-07, 2.6077e-08, ..., 1.9278e-06, + -1.7136e-07, 0.0000e+00], + ..., + [ 0.0000e+00, -8.3074e-07, 1.0813e-06, ..., -1.4659e-06, + 6.2771e-07, 0.0000e+00], + [ 0.0000e+00, 1.7323e-07, -1.5832e-08, ..., 3.7253e-07, + 1.0431e-07, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 1.7695e-08, ..., -2.4214e-07, + -6.7055e-08, 0.0000e+00]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0063, -0.0345, 0.0090, -0.0198, 0.0171, 0.0126, 0.0257, 0.0026, + -0.0379, -0.0116], device='cuda:0'), grad: tensor([ 1.6950e-07, -2.3656e-06, 5.0515e-06, 4.7963e-07, 2.4308e-07, + -2.7008e-08, 2.1048e-07, -3.4422e-06, 3.5726e-06, -3.9190e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 247.56, cls_loss 0.0015 cls_loss_mapping 0.0021 cls_loss_causal 0.4889 re_mapping 0.0051 re_causal 0.0147 /// teacc 99.02 lr 0.00010000 +Epoch 264, weight, value: tensor([[ 0.0302, -0.1614, -0.1345, ..., -0.2995, -0.1039, -0.1368], + [ 0.0469, -0.0686, 0.0372, ..., 0.0504, 0.1031, -0.0463], + [-0.0714, 0.1304, -0.1678, ..., 0.0613, 0.0776, -0.0465], + ..., + [-0.0700, -0.0951, -0.0801, ..., 0.0069, -0.1763, 0.1326], + [ 0.0614, -0.0253, 0.0794, ..., 0.0046, -0.2033, -0.0103], + [-0.1825, -0.0913, -0.1146, ..., -0.2060, 0.0537, -0.1050]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.1176e-08, ..., 9.3132e-09, + 5.7276e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -4.4238e-07, ..., -5.2061e-07, + -2.9802e-08, -9.3132e-10], + [ 0.0000e+00, -8.3819e-09, 1.0058e-07, ..., 1.0058e-07, + 4.9267e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 2.7940e-09, 3.4180e-07, ..., 4.1537e-07, + 9.8161e-07, -3.7253e-09], + [ 0.0000e+00, -9.3132e-10, -9.3132e-10, ..., -0.0000e+00, + 2.6077e-08, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 1.2107e-08, ..., 1.5832e-08, + 1.0906e-06, 9.3132e-10]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0064, -0.0367, 0.0089, -0.0198, 0.0152, 0.0124, 0.0256, 0.0049, + -0.0379, -0.0107], device='cuda:0'), grad: tensor([ 1.4137e-06, 6.6031e-07, 1.2554e-06, -2.0117e-07, -1.3329e-05, + 4.6566e-08, 5.4352e-06, 1.8254e-06, 1.2480e-07, 2.7604e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 247.68, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.5213 re_mapping 0.0051 re_causal 0.0153 /// teacc 99.11 lr 0.00010000 +Epoch 265, weight, value: tensor([[ 0.0301, -0.1618, -0.1347, ..., -0.3004, -0.1041, -0.1386], + [ 0.0485, -0.0689, 0.0373, ..., 0.0503, 0.1043, -0.0461], + [-0.0714, 0.1311, -0.1680, ..., 0.0615, 0.0780, -0.0447], + ..., + [-0.0701, -0.0955, -0.0802, ..., 0.0068, -0.1790, 0.1324], + [ 0.0613, -0.0252, 0.0803, ..., 0.0060, -0.2035, -0.0104], + [-0.1828, -0.0940, -0.1163, ..., -0.2076, 0.0535, -0.1051]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.6391e-07, 4.3772e-08, ..., 1.5274e-07, + 5.8394e-07, 1.8626e-09], + [-9.3132e-10, 3.7253e-08, 9.4064e-08, ..., 1.4063e-07, + 4.5635e-08, 1.8626e-09], + [ 0.0000e+00, -1.0217e-06, 3.3434e-07, ..., -4.6752e-07, + -9.3970e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 6.2957e-07, 4.3772e-08, ..., 2.4866e-07, + 6.3796e-07, 1.8626e-09], + [ 0.0000e+00, 6.0536e-08, -8.6427e-07, ..., -3.2317e-07, + 3.6694e-07, 9.3132e-09], + [ 0.0000e+00, 3.2596e-08, 4.5635e-08, ..., 1.6298e-07, + -1.2014e-07, -3.7253e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0064, -0.0364, 0.0092, -0.0195, 0.0152, 0.0119, 0.0252, 0.0047, + -0.0373, -0.0110], device='cuda:0'), grad: tensor([ 4.1015e-06, 6.3796e-07, -2.2948e-06, -7.2829e-07, 1.6950e-07, + 2.1651e-05, -2.6315e-05, 1.2731e-06, 1.0785e-06, 4.1258e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 247.88, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4720 re_mapping 0.0050 re_causal 0.0142 /// teacc 99.01 lr 0.00010000 +Epoch 266, weight, value: tensor([[ 0.0303, -0.1610, -0.1355, ..., -0.3045, -0.1055, -0.1413], + [ 0.0486, -0.0692, 0.0373, ..., 0.0503, 0.1045, -0.0463], + [-0.0714, 0.1313, -0.1687, ..., 0.0613, 0.0779, -0.0435], + ..., + [-0.0701, -0.0957, -0.0802, ..., 0.0069, -0.1794, 0.1336], + [ 0.0613, -0.0251, 0.0809, ..., 0.0071, -0.2034, -0.0108], + [-0.1830, -0.0969, -0.1198, ..., -0.2090, 0.0545, -0.1050]], + device='cuda:0'), grad: tensor([[-1.0617e-06, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, + -1.6764e-07, 5.5879e-09], + [ 3.7253e-09, 2.7940e-09, -3.0734e-08, ..., 1.4901e-08, + -7.4506e-09, 1.8626e-09], + [ 7.8231e-08, -3.7253e-09, 1.8626e-09, ..., 2.7940e-09, + 3.3528e-08, 1.8626e-09], + ..., + [ 9.3132e-10, 9.3132e-10, 1.0245e-08, ..., -5.6811e-08, + 2.8871e-08, -5.5879e-09], + [ 1.7695e-08, 0.0000e+00, 4.6566e-09, ..., 1.8626e-09, + 9.3132e-09, 9.3132e-10], + [ 2.4028e-07, 0.0000e+00, 2.7940e-09, ..., 2.9802e-08, + 4.5635e-08, 1.8626e-09]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0064, -0.0364, 0.0090, -0.0198, 0.0155, 0.0120, 0.0251, 0.0047, + -0.0366, -0.0109], device='cuda:0'), grad: tensor([-9.5218e-06, 1.2200e-07, 8.8196e-07, 2.8312e-07, 1.1269e-07, + 5.6997e-07, 4.9025e-06, -3.4459e-08, 1.9651e-07, 2.4661e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 247.55, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.4972 re_mapping 0.0050 re_causal 0.0150 /// teacc 99.00 lr 0.00010000 +Epoch 267, weight, value: tensor([[ 0.0307, -0.1611, -0.1375, ..., -0.3055, -0.1054, -0.1417], + [ 0.0490, -0.0694, 0.0377, ..., 0.0509, 0.1051, -0.0464], + [-0.0715, 0.1319, -0.1691, ..., 0.0613, 0.0780, -0.0434], + ..., + [-0.0701, -0.0964, -0.0805, ..., 0.0064, -0.1810, 0.1338], + [ 0.0613, -0.0248, 0.0810, ..., 0.0060, -0.2037, -0.0110], + [-0.1832, -0.0979, -0.1203, ..., -0.2099, 0.0544, -0.1051]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, -6.0908e-07, ..., 2.5146e-08, + 2.5146e-08, 6.5193e-09], + [ 0.0000e+00, 9.3132e-10, 3.3993e-07, ..., 8.9128e-07, + -3.8464e-07, -2.2445e-07], + [ 0.0000e+00, -1.2107e-08, 8.3819e-08, ..., 1.4249e-07, + -1.9558e-08, 4.6566e-09], + ..., + [ 0.0000e+00, 9.3132e-10, -4.9826e-07, ..., -1.7183e-06, + 4.3772e-08, -5.1223e-08], + [ 0.0000e+00, 4.6566e-09, -7.8604e-07, ..., -6.5286e-07, + 7.3574e-08, 3.0734e-08], + [ 0.0000e+00, 0.0000e+00, 4.0792e-07, ..., 6.1747e-07, + 9.3132e-09, 6.7987e-08]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0058, -0.0360, 0.0090, -0.0198, 0.0154, 0.0123, 0.0254, 0.0044, + -0.0373, -0.0117], device='cuda:0'), grad: tensor([-2.9877e-05, 1.3020e-06, 8.1304e-07, 2.8815e-06, 9.5926e-08, + 5.9139e-07, 1.6123e-05, -3.8743e-06, -1.1371e-06, 1.3024e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 247.66, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4969 re_mapping 0.0046 re_causal 0.0144 /// teacc 98.99 lr 0.00010000 +Epoch 268, weight, value: tensor([[ 0.0307, -0.1611, -0.1375, ..., -0.3057, -0.1052, -0.1420], + [ 0.0487, -0.0695, 0.0378, ..., 0.0509, 0.1052, -0.0467], + [-0.0716, 0.1319, -0.1693, ..., 0.0612, 0.0781, -0.0460], + ..., + [-0.0701, -0.0962, -0.0806, ..., 0.0064, -0.1814, 0.1348], + [ 0.0612, -0.0249, 0.0812, ..., 0.0061, -0.2040, -0.0118], + [-0.1834, -0.0980, -0.1210, ..., -0.2107, 0.0540, -0.1052]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.4703e-08, ..., 5.6811e-08, + 3.0734e-08, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, -1.2014e-07, ..., -7.3574e-08, + -1.6578e-07, 9.3132e-10], + [ 0.0000e+00, -7.4506e-09, 1.5274e-07, ..., 1.7043e-07, + 3.3528e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7940e-09, 5.9605e-08, ..., -2.7940e-08, + 6.9849e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -2.3656e-07, ..., -3.1106e-07, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0617e-07, ..., 1.1921e-07, + -2.7940e-08, -1.0245e-08]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0052, -0.0361, 0.0089, -0.0198, 0.0161, 0.0124, 0.0255, 0.0044, + -0.0372, -0.0127], device='cuda:0'), grad: tensor([ 5.1875e-07, -1.5646e-07, 9.7137e-07, -2.4401e-07, -9.3132e-10, + 2.2911e-07, 8.2515e-07, -8.2888e-08, -2.6226e-06, 5.5786e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 247.67, cls_loss 0.0018 cls_loss_mapping 0.0026 cls_loss_causal 0.5013 re_mapping 0.0048 re_causal 0.0143 /// teacc 99.01 lr 0.00010000 +Epoch 269, weight, value: tensor([[ 0.0307, -0.1612, -0.1380, ..., -0.3065, -0.1056, -0.1424], + [ 0.0486, -0.0694, 0.0354, ..., 0.0485, 0.1042, -0.0477], + [-0.0716, 0.1325, -0.1703, ..., 0.0612, 0.0781, -0.0460], + ..., + [-0.0701, -0.0963, -0.0780, ..., 0.0090, -0.1789, 0.1354], + [ 0.0612, -0.0259, 0.0811, ..., 0.0047, -0.2055, -0.0115], + [-0.1836, -0.0982, -0.1217, ..., -0.2115, 0.0535, -0.1051]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.7253e-09, + 9.1270e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -9.6858e-08, ..., 4.6194e-07, + 1.4435e-07, 3.6415e-07], + [ 0.0000e+00, -3.7253e-09, 3.8184e-08, ..., 3.4459e-08, + 9.0338e-08, 4.6566e-09], + ..., + [ 0.0000e+00, 3.7253e-09, 4.0978e-08, ..., -5.3179e-07, + 1.7788e-07, -3.8277e-07], + [ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., -2.0489e-08, + 1.7323e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.2107e-08, + 2.2855e-06, 7.4506e-09]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0051, -0.0381, 0.0089, -0.0198, 0.0173, 0.0122, 0.0243, 0.0067, + -0.0377, -0.0136], device='cuda:0'), grad: tensor([ 7.6368e-08, 2.1458e-06, 3.0547e-07, 1.2387e-07, -7.6815e-06, + 1.5460e-07, 1.1763e-06, -1.1129e-06, 8.8569e-07, 3.9339e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 247.61, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4616 re_mapping 0.0050 re_causal 0.0143 /// teacc 99.05 lr 0.00010000 +Epoch 270, weight, value: tensor([[ 0.0309, -0.1613, -0.1381, ..., -0.3068, -0.1057, -0.1444], + [ 0.0487, -0.0695, 0.0354, ..., 0.0485, 0.1042, -0.0485], + [-0.0716, 0.1326, -0.1706, ..., 0.0613, 0.0785, -0.0457], + ..., + [-0.0701, -0.0964, -0.0780, ..., 0.0090, -0.1790, 0.1359], + [ 0.0612, -0.0259, 0.0813, ..., 0.0051, -0.2060, -0.0119], + [-0.1839, -0.0980, -0.1219, ..., -0.2117, 0.0534, -0.1052]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 4.2841e-08, ..., 5.0291e-08, + 8.6613e-08, 1.8626e-09], + [ 0.0000e+00, 9.3132e-10, -3.3528e-08, ..., -3.9581e-07, + -1.8375e-06, 5.5879e-09], + [ 0.0000e+00, -3.7253e-09, 4.3586e-07, ..., 5.0105e-07, + 8.7637e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 9.3132e-10, 2.8033e-07, ..., 2.2165e-07, + 3.9395e-07, -1.3970e-08], + [ 0.0000e+00, 0.0000e+00, -1.6689e-06, ..., -1.2377e-06, + -2.7474e-07, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 4.8429e-08, + 9.3132e-09, 3.7253e-09]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0049, -0.0381, 0.0090, -0.0198, 0.0182, 0.0122, 0.0240, 0.0067, + -0.0380, -0.0137], device='cuda:0'), grad: tensor([ 1.9930e-07, -1.2396e-06, 1.4463e-06, 5.7556e-07, 1.9465e-07, + 7.0501e-07, 7.3109e-07, 5.5414e-07, -3.2820e-06, 1.0524e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 247.88, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4982 re_mapping 0.0047 re_causal 0.0148 /// teacc 99.00 lr 0.00010000 +Epoch 271, weight, value: tensor([[ 0.0309, -0.1613, -0.1388, ..., -0.3082, -0.1059, -0.1456], + [ 0.0488, -0.0696, 0.0354, ..., 0.0485, 0.1046, -0.0496], + [-0.0716, 0.1331, -0.1715, ..., 0.0613, 0.0786, -0.0466], + ..., + [-0.0701, -0.0966, -0.0780, ..., 0.0088, -0.1793, 0.1362], + [ 0.0612, -0.0265, 0.0822, ..., 0.0044, -0.2075, -0.0143], + [-0.1842, -0.0982, -0.1247, ..., -0.2136, 0.0534, -0.1057]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 9.3132e-09, + 5.4110e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -9.1195e-06, ..., -1.3039e-05, + -1.2264e-05, 3.6880e-07], + [ 0.0000e+00, -9.3132e-10, 1.1586e-06, ..., 1.6829e-06, + 1.6605e-06, 2.4214e-08], + ..., + [ 0.0000e+00, 9.3132e-10, 7.6219e-06, ..., 1.0870e-05, + 1.0565e-05, -4.0792e-07], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 4.9360e-08, + 5.0291e-08, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.7253e-09, + -4.4145e-07, -3.0734e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0049, -0.0381, 0.0091, -0.0192, 0.0182, 0.0120, 0.0243, 0.0066, + -0.0380, -0.0143], device='cuda:0'), grad: tensor([ 2.0303e-06, -2.0206e-05, 3.2037e-06, 1.2256e-06, 7.6741e-06, + 1.7202e-06, -1.1981e-05, 1.7717e-05, 3.6135e-07, -1.7500e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 247.81, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.5313 re_mapping 0.0049 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 272, weight, value: tensor([[ 0.0307, -0.1614, -0.1397, ..., -0.3086, -0.1061, -0.1457], + [ 0.0496, -0.0696, 0.0355, ..., 0.0486, 0.1049, -0.0499], + [-0.0717, 0.1333, -0.1723, ..., 0.0612, 0.0786, -0.0467], + ..., + [-0.0706, -0.0967, -0.0780, ..., 0.0085, -0.1796, 0.1371], + [ 0.0603, -0.0265, 0.0820, ..., 0.0039, -0.2084, -0.0143], + [-0.1841, -0.0983, -0.1249, ..., -0.2137, 0.0536, -0.1057]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.7253e-09, + 2.8871e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -9.5926e-08, ..., -8.4750e-08, + -3.8184e-08, -5.5879e-09], + [ 0.0000e+00, -3.7253e-09, 2.6077e-08, ..., -8.5682e-08, + -1.1828e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 1.3877e-07, ..., 7.9162e-08, + 1.4249e-07, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -1.7583e-06, ..., -4.6007e-07, + 5.4017e-08, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 1.1176e-08, ..., 4.6566e-09, + -1.4715e-07, -9.3132e-10]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0052, -0.0380, 0.0090, -0.0187, 0.0183, 0.0119, 0.0241, 0.0065, + -0.0382, -0.0141], device='cuda:0'), grad: tensor([ 8.3819e-09, 3.4459e-08, -2.6543e-07, 1.6410e-06, 1.9278e-07, + -2.1700e-07, 1.2862e-06, 3.0547e-07, -2.5332e-06, -4.5914e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 247.84, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4822 re_mapping 0.0050 re_causal 0.0140 /// teacc 99.01 lr 0.00010000 +Epoch 273, weight, value: tensor([[ 0.0305, -0.1615, -0.1401, ..., -0.3092, -0.1064, -0.1462], + [ 0.0497, -0.0696, 0.0355, ..., 0.0486, 0.1054, -0.0502], + [-0.0718, 0.1333, -0.1732, ..., 0.0610, 0.0783, -0.0474], + ..., + [-0.0708, -0.0967, -0.0780, ..., 0.0086, -0.1798, 0.1373], + [ 0.0604, -0.0265, 0.0833, ..., 0.0050, -0.2088, -0.0144], + [-0.1844, -0.0985, -0.1253, ..., -0.2138, 0.0538, -0.1057]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 2.4214e-08, ..., 2.7008e-08, + 1.0245e-08, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 2.7940e-08, ..., 4.4703e-08, + -8.3819e-09, 2.7940e-09], + [ 9.3132e-10, -5.0291e-08, 3.7793e-06, ..., 4.0159e-06, + 6.9849e-07, 2.6077e-08], + ..., + [ 9.3132e-10, 4.2841e-08, 1.6019e-07, ..., 7.4506e-08, + 9.1270e-08, -3.8184e-08], + [ 1.4901e-08, 2.7940e-09, 8.6613e-08, ..., 9.1270e-08, + 3.0734e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 2.4214e-08, ..., 3.9116e-08, + -5.8301e-07, 4.6566e-09]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0069, -0.0380, 0.0088, -0.0188, 0.0180, 0.0124, 0.0249, 0.0065, + -0.0374, -0.0141], device='cuda:0'), grad: tensor([-9.6764e-07, 1.6298e-07, 9.0078e-06, -9.3281e-06, 3.3230e-06, + -3.3248e-07, 8.8010e-07, 1.8440e-07, 3.5483e-07, -3.3118e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 247.38, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4888 re_mapping 0.0048 re_causal 0.0146 /// teacc 98.98 lr 0.00010000 +Epoch 274, weight, value: tensor([[ 0.0307, -0.1616, -0.1408, ..., -0.3095, -0.1065, -0.1463], + [ 0.0500, -0.0696, 0.0356, ..., 0.0486, 0.1055, -0.0503], + [-0.0718, 0.1337, -0.1737, ..., 0.0612, 0.0788, -0.0477], + ..., + [-0.0710, -0.0971, -0.0781, ..., 0.0085, -0.1801, 0.1374], + [ 0.0600, -0.0265, 0.0835, ..., 0.0036, -0.2092, -0.0143], + [-0.1848, -0.0986, -0.1255, ..., -0.2140, 0.0537, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 5.5879e-09, + 2.0955e-07, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 8.3819e-09, ..., 5.3085e-08, + 4.6566e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., -4.6566e-08, + 3.8184e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 4.2841e-08, 2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 1.1176e-08, ..., 1.0245e-08, + -4.4238e-07, 0.0000e+00]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0086, -0.0379, 0.0089, -0.0187, 0.0184, 0.0125, 0.0265, 0.0064, + -0.0385, -0.0143], device='cuda:0'), grad: tensor([-1.2945e-07, 6.8638e-07, 1.6019e-07, -2.9895e-07, 4.3772e-07, + 2.6450e-07, -1.0896e-07, -1.3039e-08, 1.2759e-07, -1.1260e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 247.36, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4913 re_mapping 0.0049 re_causal 0.0141 /// teacc 98.91 lr 0.00010000 +Epoch 275, weight, value: tensor([[ 0.0307, -0.1615, -0.1403, ..., -0.3097, -0.1063, -0.1464], + [ 0.0503, -0.0697, 0.0357, ..., 0.0486, 0.1056, -0.0505], + [-0.0719, 0.1339, -0.1740, ..., 0.0613, 0.0788, -0.0481], + ..., + [-0.0712, -0.0973, -0.0781, ..., 0.0084, -0.1803, 0.1394], + [ 0.0600, -0.0265, 0.0833, ..., 0.0034, -0.2094, -0.0145], + [-0.1846, -0.0990, -0.1259, ..., -0.2144, 0.0534, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 5.5879e-09, ..., 7.4506e-09, + 7.6368e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-08, ..., 1.6764e-08, + 1.4901e-08, 3.7253e-09], + [ 0.0000e+00, -1.3039e-08, 1.6950e-07, ..., 1.0058e-07, + 7.4506e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 1.8626e-09, 6.9663e-07, ..., 5.9232e-07, + 3.1851e-07, -1.6764e-08], + [ 0.0000e+00, 1.8626e-09, -1.1176e-07, ..., 5.2154e-08, + 2.0489e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 5.5879e-09, + -7.3947e-07, 1.8626e-09]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0084, -0.0379, 0.0089, -0.0186, 0.0180, 0.0127, 0.0267, 0.0065, + -0.0389, -0.0146], device='cuda:0'), grad: tensor([ 2.0489e-07, 2.3469e-07, 3.5949e-07, -2.6375e-06, 6.1095e-07, + 4.1164e-07, 1.4342e-07, 2.5239e-06, -2.0489e-08, -1.8403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 247.19, cls_loss 0.0011 cls_loss_mapping 0.0026 cls_loss_causal 0.4767 re_mapping 0.0047 re_causal 0.0141 /// teacc 98.98 lr 0.00010000 +Epoch 276, weight, value: tensor([[ 0.0306, -0.1616, -0.1407, ..., -0.3099, -0.1064, -0.1465], + [ 0.0507, -0.0697, 0.0358, ..., 0.0487, 0.1059, -0.0505], + [-0.0721, 0.1343, -0.1747, ..., 0.0613, 0.0790, -0.0481], + ..., + [-0.0718, -0.0976, -0.0783, ..., 0.0084, -0.1807, 0.1394], + [ 0.0599, -0.0265, 0.0836, ..., 0.0037, -0.2097, -0.0145], + [-0.1847, -0.0996, -0.1260, ..., -0.2146, 0.0539, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.3039e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -6.9104e-07, ..., 3.7253e-09, + -7.4133e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.3819e-08, ..., 1.7881e-07, + 8.7544e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.8860e-07, ..., -3.5204e-07, + 6.0908e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.9616e-07, ..., -5.7742e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 3.7253e-08, + -3.0175e-07, 0.0000e+00]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0085, -0.0378, 0.0089, -0.0185, 0.0172, 0.0129, 0.0267, 0.0063, + -0.0386, -0.0141], device='cuda:0'), grad: tensor([ 7.4506e-09, -5.3458e-07, 5.5321e-07, 4.2096e-07, 1.6019e-07, + 5.5879e-07, 2.0675e-07, -1.0803e-07, -4.9546e-07, -8.0653e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 247.02, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4892 re_mapping 0.0048 re_causal 0.0141 /// teacc 99.05 lr 0.00010000 +Epoch 277, weight, value: tensor([[ 0.0305, -0.1617, -0.1410, ..., -0.3101, -0.1066, -0.1466], + [ 0.0507, -0.0698, 0.0360, ..., 0.0488, 0.1062, -0.0505], + [-0.0722, 0.1348, -0.1752, ..., 0.0613, 0.0791, -0.0482], + ..., + [-0.0717, -0.0980, -0.0784, ..., 0.0083, -0.1811, 0.1394], + [ 0.0599, -0.0265, 0.0838, ..., 0.0038, -0.2098, -0.0146], + [-0.1847, -0.0998, -0.1265, ..., -0.2152, 0.0542, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 5.5879e-09, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.2498e-06, ..., -8.3074e-07, + -1.3076e-06, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 4.5449e-07, ..., 7.0781e-07, + 1.4901e-08, 9.8720e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 5.3458e-07, ..., -2.6450e-07, + 1.2387e-06, -1.6391e-07], + [ 0.0000e+00, 0.0000e+00, 6.5193e-08, ..., 1.2293e-07, + 1.8626e-08, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 1.8626e-08, + -2.2314e-06, 0.0000e+00]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0085, -0.0377, 0.0089, -0.0185, 0.0167, 0.0129, 0.0269, 0.0062, + -0.0387, -0.0140], device='cuda:0'), grad: tensor([ 6.3330e-08, -2.2221e-06, 1.4286e-06, 4.4517e-07, 6.4895e-06, + 6.1467e-08, -9.4995e-08, 9.4995e-08, 2.7940e-07, -6.5565e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 247.08, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5006 re_mapping 0.0046 re_causal 0.0139 /// teacc 99.00 lr 0.00010000 +Epoch 278, weight, value: tensor([[ 0.0305, -0.1617, -0.1415, ..., -0.3103, -0.1066, -0.1466], + [ 0.0507, -0.0699, 0.0361, ..., 0.0488, 0.1064, -0.0505], + [-0.0722, 0.1352, -0.1757, ..., 0.0613, 0.0793, -0.0482], + ..., + [-0.0717, -0.0982, -0.0786, ..., 0.0082, -0.1814, 0.1394], + [ 0.0599, -0.0266, 0.0844, ..., 0.0048, -0.2105, -0.0146], + [-0.1851, -0.1013, -0.1268, ..., -0.2169, 0.0555, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 3.7253e-09, ..., 3.7253e-09, + 2.7940e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, -1.8626e-09, ..., 2.9802e-08, + 9.1270e-06, 0.0000e+00], + [ 0.0000e+00, -6.7055e-08, 3.1665e-08, ..., -2.9802e-08, + -5.2154e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.9802e-08, 1.5832e-07, ..., 1.0245e-07, + 2.7008e-07, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -2.9430e-07, ..., -4.0978e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 3.9116e-08, ..., 2.4214e-08, + 2.6077e-05, 0.0000e+00]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0086, -0.0376, 0.0089, -0.0169, 0.0151, 0.0113, 0.0266, 0.0060, + -0.0384, -0.0127], device='cuda:0'), grad: tensor([-1.7863e-06, 2.0921e-05, 1.1362e-07, 1.5087e-07, -7.8917e-05, + 3.7067e-07, 1.3020e-06, 7.5810e-07, -3.6880e-07, 5.7399e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 247.09, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.5450 re_mapping 0.0050 re_causal 0.0151 /// teacc 98.97 lr 0.00010000 +Epoch 279, weight, value: tensor([[ 0.0311, -0.1617, -0.1414, ..., -0.3104, -0.1067, -0.1467], + [ 0.0508, -0.0703, 0.0361, ..., 0.0488, 0.1065, -0.0508], + [-0.0737, 0.1359, -0.1764, ..., 0.0612, 0.0793, -0.0482], + ..., + [-0.0700, -0.0985, -0.0786, ..., 0.0082, -0.1823, 0.1395], + [ 0.0596, -0.0258, 0.0849, ..., 0.0055, -0.2107, -0.0144], + [-0.1860, -0.1055, -0.1270, ..., -0.2173, 0.0577, -0.1058]], + device='cuda:0'), grad: tensor([[-5.5879e-09, 2.9802e-08, 0.0000e+00, ..., 2.2352e-08, + 8.6613e-07, 0.0000e+00], + [ 0.0000e+00, 2.7753e-07, -3.5390e-08, ..., 1.2480e-07, + 2.7008e-07, 0.0000e+00], + [ 0.0000e+00, -8.4750e-07, 2.2352e-08, ..., -4.8243e-07, + -8.2888e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 3.0175e-07, 2.0489e-08, ..., 1.9930e-07, + 4.0419e-07, 0.0000e+00], + [ 0.0000e+00, 5.4017e-08, 5.5879e-09, ..., 3.7253e-08, + 2.4959e-07, 0.0000e+00], + [ 1.8626e-09, 1.8626e-08, 0.0000e+00, ..., 1.3039e-08, + -3.2783e-07, 0.0000e+00]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0083, -0.0377, 0.0087, -0.0169, 0.0135, 0.0112, 0.0267, 0.0058, + -0.0382, -0.0107], device='cuda:0'), grad: tensor([ 5.0142e-06, 8.6240e-07, -2.0061e-06, 2.3469e-07, 1.1548e-06, + 4.4890e-07, -7.3388e-06, 1.1604e-06, 1.3877e-06, -9.2201e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 247.05, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4781 re_mapping 0.0048 re_causal 0.0140 /// teacc 98.93 lr 0.00010000 +Epoch 280, weight, value: tensor([[ 0.0338, -0.1618, -0.1417, ..., -0.3106, -0.1067, -0.1472], + [ 0.0514, -0.0704, 0.0363, ..., 0.0491, 0.1071, -0.0508], + [-0.0749, 0.1367, -0.1769, ..., 0.0612, 0.0792, -0.0483], + ..., + [-0.0722, -0.0995, -0.0787, ..., 0.0079, -0.1828, 0.1395], + [ 0.0592, -0.0253, 0.0850, ..., 0.0056, -0.2111, -0.0146], + [-0.1879, -0.1072, -0.1274, ..., -0.2178, 0.0581, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 3.7253e-09, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.5423e-06, ..., 1.1530e-06, + -1.8440e-07, 0.0000e+00], + [ 0.0000e+00, -1.4901e-08, 2.7940e-08, ..., 1.1176e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 7.4506e-09, -2.0731e-06, ..., -1.5199e-06, + 6.1467e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0431e-07, ..., 5.5879e-08, + 1.0058e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.6508e-07, ..., 2.6636e-07, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0080, -0.0375, 0.0085, -0.0168, 0.0127, 0.0112, 0.0268, 0.0057, + -0.0383, -0.0106], device='cuda:0'), grad: tensor([ 2.0117e-07, 6.9402e-06, 6.5193e-08, 2.0489e-08, 6.3330e-08, + -5.5879e-09, -2.4214e-07, -8.8662e-06, 2.6077e-07, 1.5572e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 247.31, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4898 re_mapping 0.0045 re_causal 0.0137 /// teacc 98.97 lr 0.00010000 +Epoch 281, weight, value: tensor([[ 0.0338, -0.1620, -0.1419, ..., -0.3108, -0.1067, -0.1482], + [ 0.0514, -0.0706, 0.0371, ..., 0.0502, 0.1093, -0.0508], + [-0.0751, 0.1375, -0.1774, ..., 0.0611, 0.0790, -0.0483], + ..., + [-0.0726, -0.1000, -0.0795, ..., 0.0068, -0.1853, 0.1395], + [ 0.0590, -0.0252, 0.0853, ..., 0.0057, -0.2116, -0.0146], + [-0.1885, -0.1076, -0.1276, ..., -0.2181, 0.0597, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 3.7253e-08, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, -3.9116e-08, ..., 1.8626e-09, + 1.7695e-07, 0.0000e+00], + [ 1.8626e-09, -5.9605e-08, 5.9605e-08, ..., 1.1176e-08, + 1.6764e-08, 0.0000e+00], + ..., + [-1.8626e-09, 1.8626e-09, 2.7940e-08, ..., -1.7881e-07, + 1.0431e-07, 0.0000e+00], + [ 0.0000e+00, 5.4017e-08, 1.8626e-09, ..., 8.1956e-08, + 7.0781e-08, 0.0000e+00], + [ 3.9116e-08, 0.0000e+00, 1.8626e-09, ..., 9.4995e-08, + 6.3218e-06, 0.0000e+00]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0078, -0.0363, 0.0084, -0.0167, 0.0109, 0.0112, 0.0268, 0.0044, + -0.0382, -0.0091], device='cuda:0'), grad: tensor([ 1.7881e-07, 6.8173e-07, 4.8988e-07, -4.6566e-08, 4.8205e-06, + 3.5595e-06, 5.4576e-07, -2.5928e-05, 5.8487e-07, 1.5102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 247.47, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.5090 re_mapping 0.0048 re_causal 0.0139 /// teacc 98.94 lr 0.00010000 +Epoch 282, weight, value: tensor([[ 0.0339, -0.1621, -0.1435, ..., -0.3110, -0.1068, -0.1499], + [ 0.0517, -0.0710, 0.0372, ..., 0.0502, 0.1094, -0.0510], + [-0.0753, 0.1380, -0.1770, ..., 0.0616, 0.0801, -0.0481], + ..., + [-0.0729, -0.1001, -0.0796, ..., 0.0067, -0.1857, 0.1395], + [ 0.0584, -0.0255, 0.0855, ..., 0.0060, -0.2122, -0.0149], + [-0.1892, -0.1074, -0.1279, ..., -0.2187, 0.0603, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.9802e-08, 5.7556e-07, ..., 4.4703e-08, + 1.0617e-07, 5.5879e-08], + [ 0.0000e+00, 2.7940e-08, 3.1665e-08, ..., -6.5193e-08, + -4.8429e-08, 1.3225e-07], + [ 0.0000e+00, -7.2643e-07, 5.6066e-07, ..., -7.2829e-07, + -3.8184e-07, 8.7544e-08], + ..., + [ 0.0000e+00, 4.1910e-07, 3.8743e-07, ..., 4.5262e-07, + 3.9302e-07, 3.7253e-09], + [ 0.0000e+00, 1.5460e-07, 1.5963e-06, ..., 3.3714e-07, + 6.4075e-07, 1.2871e-06], + [-3.7253e-09, 7.4506e-09, 6.2026e-07, ..., 2.2352e-08, + 7.8231e-08, 2.6077e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0076, -0.0363, 0.0090, -0.0166, 0.0101, 0.0111, 0.0266, 0.0043, + -0.0383, -0.0086], device='cuda:0'), grad: tensor([ 1.5143e-06, 9.1456e-07, -6.8918e-07, -7.8157e-06, -3.2037e-07, + -1.4044e-06, -4.5225e-06, 2.2557e-06, 6.7167e-06, 3.3230e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 247.04, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.4929 re_mapping 0.0047 re_causal 0.0136 /// teacc 98.97 lr 0.00010000 +Epoch 283, weight, value: tensor([[ 0.0339, -0.1623, -0.1422, ..., -0.3112, -0.1069, -0.1508], + [ 0.0530, -0.0715, 0.0373, ..., 0.0502, 0.1094, -0.0511], + [-0.0756, 0.1388, -0.1781, ..., 0.0618, 0.0804, -0.0483], + ..., + [-0.0731, -0.1003, -0.0797, ..., 0.0066, -0.1858, 0.1397], + [ 0.0578, -0.0259, 0.0862, ..., 0.0056, -0.2132, -0.0153], + [-0.1897, -0.1078, -0.1283, ..., -0.2194, 0.0608, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7998e-07, ..., 4.2841e-08, + 3.7067e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-06, ..., -3.9302e-07, + -1.0524e-06, 0.0000e+00], + [ 0.0000e+00, -1.6764e-08, 1.3970e-07, ..., 1.3039e-08, + 8.0094e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3039e-08, 3.4831e-07, ..., 8.5682e-08, + 2.2352e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1921e-07, ..., 3.7253e-08, + 1.0990e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-08, ..., 1.1176e-08, + -1.1362e-07, 0.0000e+00]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0074, -0.0364, 0.0088, -0.0162, 0.0100, 0.0112, 0.0262, 0.0047, + -0.0384, -0.0097], device='cuda:0'), grad: tensor([ 5.3868e-06, -4.5449e-06, 6.5938e-07, 5.9605e-08, -1.0610e-05, + 1.0412e-06, 1.0077e-06, 7.0594e-07, 3.3714e-07, 5.9418e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 246.91, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4882 re_mapping 0.0048 re_causal 0.0140 /// teacc 99.00 lr 0.00010000 +Epoch 284, weight, value: tensor([[ 0.0338, -0.1624, -0.1438, ..., -0.3113, -0.1069, -0.1511], + [ 0.0529, -0.0717, 0.0375, ..., 0.0502, 0.1095, -0.0511], + [-0.0757, 0.1393, -0.1787, ..., 0.0618, 0.0806, -0.0484], + ..., + [-0.0731, -0.1006, -0.0798, ..., 0.0065, -0.1889, 0.1397], + [ 0.0576, -0.0260, 0.0868, ..., 0.0057, -0.2138, -0.0154], + [-0.1922, -0.1079, -0.1289, ..., -0.2196, 0.0645, -0.1059]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 3.7253e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.5146e-07, ..., -9.6858e-08, + -2.1793e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.4703e-08, ..., 2.2352e-08, + 5.0291e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.8720e-08, ..., 3.5390e-08, + 8.7544e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 1.1735e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 8.5495e-07, 0.0000e+00]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0079, -0.0363, 0.0089, -0.0159, 0.0100, 0.0110, 0.0259, 0.0031, + -0.0383, -0.0062], device='cuda:0'), grad: tensor([ 2.2352e-08, -3.5949e-07, 1.1921e-07, 3.1050e-06, 1.0245e-07, + -2.3201e-05, 4.4703e-07, 1.4342e-07, 1.9334e-06, 1.7658e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 246.67, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.4903 re_mapping 0.0051 re_causal 0.0147 /// teacc 98.93 lr 0.00010000 +Epoch 285, weight, value: tensor([[ 0.0343, -0.1630, -0.1441, ..., -0.3115, -0.1093, -0.1515], + [ 0.0524, -0.0717, 0.0376, ..., 0.0503, 0.1100, -0.0511], + [-0.0760, 0.1404, -0.1800, ..., 0.0615, 0.0801, -0.0486], + ..., + [-0.0750, -0.1015, -0.0799, ..., 0.0064, -0.1891, 0.1400], + [ 0.0584, -0.0265, 0.0871, ..., 0.0056, -0.2142, -0.0155], + [-0.1941, -0.1081, -0.1299, ..., -0.2206, 0.0645, -0.1059]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.9802e-08, 5.5879e-09, ..., 2.6077e-08, + 2.0489e-07, 0.0000e+00], + [ 1.8626e-09, 7.8231e-08, -2.0489e-08, ..., 3.0361e-07, + 1.4529e-07, 0.0000e+00], + [ 0.0000e+00, -2.6263e-07, -4.6566e-08, ..., -2.0489e-07, + -2.8871e-07, 0.0000e+00], + ..., + [ 1.8626e-09, 3.7253e-09, 1.3039e-08, ..., -2.0042e-06, + 6.3330e-08, -0.0000e+00], + [ 9.3132e-08, 6.1467e-08, 1.4901e-08, ..., 6.7055e-08, + 1.2666e-07, 0.0000e+00], + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 3.5390e-08, + 5.7966e-06, 0.0000e+00]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0099, -0.0362, 0.0088, -0.0163, 0.0097, 0.0117, 0.0282, 0.0030, + -0.0384, -0.0064], device='cuda:0'), grad: tensor([ 5.8115e-07, 1.1139e-06, -1.1791e-06, 4.5300e-06, -1.2137e-05, + -2.7642e-06, 1.3988e-06, -4.5560e-06, 1.1008e-06, 1.1876e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 247.06, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.4645 re_mapping 0.0049 re_causal 0.0140 /// teacc 99.04 lr 0.00010000 +Epoch 286, weight, value: tensor([[ 0.0339, -0.1634, -0.1450, ..., -0.3116, -0.1096, -0.1519], + [ 0.0529, -0.0718, 0.0377, ..., 0.0505, 0.1107, -0.0511], + [-0.0765, 0.1409, -0.1806, ..., 0.0610, 0.0796, -0.0486], + ..., + [-0.0713, -0.1019, -0.0800, ..., 0.0064, -0.1894, 0.1400], + [ 0.0569, -0.0292, 0.0866, ..., 0.0054, -0.2152, -0.0155], + [-0.1961, -0.1055, -0.1278, ..., -0.2213, 0.0646, -0.1059]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.4506e-09, 1.4156e-07, ..., 2.7940e-08, + 1.9185e-07, 0.0000e+00], + [-6.3330e-08, 3.3528e-08, -2.1048e-06, ..., -2.2724e-07, + -3.3155e-06, 0.0000e+00], + [ 1.8626e-09, -7.8231e-08, 5.5879e-08, ..., -1.4901e-07, + -2.1048e-07, 0.0000e+00], + ..., + [ 2.4214e-08, 1.4901e-08, 1.7248e-06, ..., 2.0862e-07, + 2.5891e-06, 0.0000e+00], + [ 5.5879e-09, 5.5879e-09, 1.2666e-07, ..., 4.4703e-08, + 1.4529e-07, 0.0000e+00], + [ 1.4901e-08, 1.8626e-09, 1.9316e-06, ..., 3.5390e-08, + 3.8557e-07, 0.0000e+00]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0099, -0.0361, 0.0084, -0.0163, 0.0093, 0.0118, 0.0281, 0.0031, + -0.0408, -0.0058], device='cuda:0'), grad: tensor([-1.0710e-06, -6.4075e-06, -4.0792e-07, -5.0813e-06, 4.6566e-08, + 1.4286e-06, 1.4231e-06, 5.2452e-06, 4.5449e-07, 4.3139e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 247.22, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4891 re_mapping 0.0050 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +Epoch 287, weight, value: tensor([[ 0.0333, -0.1639, -0.1457, ..., -0.3119, -0.1097, -0.1519], + [ 0.0544, -0.0721, 0.0373, ..., 0.0498, 0.1104, -0.0511], + [-0.0775, 0.1389, -0.1810, ..., 0.0598, 0.0797, -0.0486], + ..., + [-0.0716, -0.1027, -0.0795, ..., 0.0071, -0.1890, 0.1400], + [ 0.0585, -0.0294, 0.0871, ..., 0.0054, -0.2156, -0.0156], + [-0.1961, -0.1056, -0.1279, ..., -0.2215, 0.0646, -0.1059]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 7.4506e-09, 0.0000e+00], + [ 7.4506e-09, -0.0000e+00, -3.1665e-08, ..., -1.4901e-08, + -2.2352e-08, 0.0000e+00], + [ 3.7253e-09, -5.2154e-08, 7.4506e-09, ..., -1.0990e-07, + -7.6368e-08, 0.0000e+00], + ..., + [-2.4214e-08, 4.8429e-08, 1.3039e-08, ..., 9.6858e-08, + 1.9558e-07, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -9.3132e-09, ..., -0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + -2.6077e-07, 0.0000e+00]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0099, -0.0368, 0.0077, -0.0157, 0.0093, 0.0114, 0.0280, 0.0037, + -0.0406, -0.0058], device='cuda:0'), grad: tensor([-2.3842e-07, 2.7940e-08, -1.4715e-07, 3.3528e-08, 3.9488e-07, + -6.3330e-08, 1.8626e-07, 4.6194e-07, 8.9407e-08, -7.7114e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 247.14, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4769 re_mapping 0.0047 re_causal 0.0137 /// teacc 99.04 lr 0.00010000 +Epoch 288, weight, value: tensor([[ 0.0332, -0.1641, -0.1458, ..., -0.3121, -0.1097, -0.1519], + [ 0.0562, -0.0722, 0.0375, ..., 0.0499, 0.1108, -0.0511], + [-0.0774, 0.1382, -0.1816, ..., 0.0584, 0.0798, -0.0486], + ..., + [-0.0716, -0.1011, -0.0796, ..., 0.0072, -0.1893, 0.1400], + [ 0.0583, -0.0301, 0.0872, ..., 0.0051, -0.2170, -0.0156], + [-0.1984, -0.1056, -0.1280, ..., -0.2221, 0.0646, -0.1059]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.0617e-07, 1.3597e-07, ..., 5.5879e-08, + 4.0047e-07, 0.0000e+00], + [ 0.0000e+00, 5.7742e-08, -3.4600e-05, ..., -2.6543e-06, + -6.9022e-05, 0.0000e+00], + [ 0.0000e+00, -2.0787e-06, 4.2841e-07, ..., -1.8030e-06, + -1.4175e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 4.2841e-08, 4.4703e-07, ..., 5.7742e-08, + 9.1456e-07, 0.0000e+00], + [-5.5879e-09, 9.8720e-08, 5.5283e-06, ..., 4.5821e-07, + 1.1377e-05, 0.0000e+00], + [ 0.0000e+00, 1.3039e-08, 1.0245e-07, ..., 1.6764e-08, + 2.1793e-07, 0.0000e+00]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0097, -0.0366, 0.0068, -0.0156, 0.0092, 0.0114, 0.0279, 0.0038, + -0.0410, -0.0057], device='cuda:0'), grad: tensor([ 1.4622e-06, -2.6011e-04, -3.6526e-06, 5.2489e-06, 1.6999e-04, + 3.1143e-06, 3.6836e-05, 3.3863e-06, 4.2677e-05, 8.1770e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 247.35, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4911 re_mapping 0.0047 re_causal 0.0143 /// teacc 98.99 lr 0.00010000 +Epoch 289, weight, value: tensor([[ 0.0325, -0.1642, -0.1462, ..., -0.3128, -0.1101, -0.1526], + [ 0.0577, -0.0724, 0.0376, ..., 0.0499, 0.1114, -0.0511], + [-0.0769, 0.1389, -0.1819, ..., 0.0584, 0.0802, -0.0480], + ..., + [-0.0716, -0.1013, -0.0796, ..., 0.0073, -0.1894, 0.1400], + [ 0.0583, -0.0307, 0.0878, ..., 0.0050, -0.2180, -0.0156], + [-0.1993, -0.1058, -0.1282, ..., -0.2224, 0.0645, -0.1059]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-6.0350e-07, 0.0000e+00, -1.7229e-06, ..., -1.4380e-06, + -2.1812e-06, 0.0000e+00], + [ 1.2293e-07, 1.8626e-09, 3.6508e-07, ..., 3.0547e-07, + 4.4145e-07, 0.0000e+00], + ..., + [ 2.7940e-08, 0.0000e+00, 1.0617e-07, ..., 6.3330e-08, + 1.1921e-07, 0.0000e+00], + [ 2.4401e-07, -3.7253e-09, 6.1281e-07, ..., 5.3085e-07, + 9.7044e-07, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 1.1176e-08, ..., 1.1176e-08, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0098, -0.0364, 0.0070, -0.0157, 0.0089, 0.0120, 0.0266, 0.0038, + -0.0410, -0.0059], device='cuda:0'), grad: tensor([-5.9605e-08, -2.8890e-06, 7.0781e-07, 1.1176e-07, 7.6368e-07, + 8.0094e-08, -2.1420e-07, 1.4715e-07, 1.2275e-06, 1.2107e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 247.12, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.5122 re_mapping 0.0045 re_causal 0.0139 /// teacc 99.09 lr 0.00010000 +Epoch 290, weight, value: tensor([[ 0.0345, -0.1635, -0.1463, ..., -0.3133, -0.1101, -0.1543], + [ 0.0576, -0.0725, 0.0376, ..., 0.0498, 0.1115, -0.0510], + [-0.0764, 0.1394, -0.1825, ..., 0.0584, 0.0805, -0.0480], + ..., + [-0.0719, -0.1019, -0.0795, ..., 0.0074, -0.1896, 0.1401], + [ 0.0574, -0.0307, 0.0879, ..., 0.0050, -0.2185, -0.0158], + [-0.2012, -0.1061, -0.1284, ..., -0.2231, 0.0646, -0.1059]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, 5.5879e-09, ..., 1.4901e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 2.7940e-08, -5.5879e-08, ..., 1.0990e-07, + -4.8429e-08, 0.0000e+00], + [ 0.0000e+00, 7.4506e-08, 3.1665e-08, ..., 4.5821e-07, + -3.9116e-08, 0.0000e+00], + ..., + [-9.3132e-09, -1.6019e-07, 1.3039e-08, ..., -2.2296e-06, + 2.7940e-08, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, -8.7544e-08, ..., 3.5949e-07, + 2.4214e-08, 8.7544e-08], + [ 5.5879e-09, 5.5879e-09, 2.4214e-08, ..., 1.7136e-07, + 5.4017e-08, 0.0000e+00]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0093, -0.0365, 0.0071, -0.0156, 0.0089, 0.0121, 0.0265, 0.0039, + -0.0412, -0.0060], device='cuda:0'), grad: tensor([ 6.5193e-08, 3.4831e-07, 1.2126e-06, 3.1963e-06, -2.1420e-07, + 2.8498e-07, -1.0617e-07, -6.9477e-06, 1.1139e-06, 1.0189e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 246.93, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.4815 re_mapping 0.0048 re_causal 0.0140 /// teacc 98.99 lr 0.00010000 +Epoch 291, weight, value: tensor([[ 0.0346, -0.1638, -0.1471, ..., -0.3139, -0.1103, -0.1553], + [ 0.0576, -0.0728, 0.0378, ..., 0.0499, 0.1117, -0.0510], + [-0.0762, 0.1403, -0.1830, ..., 0.0586, 0.0809, -0.0478], + ..., + [-0.0718, -0.1028, -0.0796, ..., 0.0073, -0.1899, 0.1401], + [ 0.0573, -0.0309, 0.0895, ..., 0.0049, -0.2189, -0.0161], + [-0.2016, -0.1063, -0.1287, ..., -0.2250, 0.0649, -0.1059]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -6.8918e-08, ..., -1.8626e-08, + -5.4017e-08, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 5.7742e-08, ..., 3.7253e-08, + -5.5879e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 3.5390e-08, ..., -4.6566e-08, + 4.6566e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, -9.6858e-08, ..., -7.6368e-08, + 2.2352e-08, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 3.7253e-09, ..., 5.5879e-08, + -6.7055e-08, 0.0000e+00]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0094, -0.0364, 0.0077, -0.0158, 0.0088, 0.0117, 0.0261, 0.0039, + -0.0398, -0.0060], device='cuda:0'), grad: tensor([ 5.4017e-08, -3.3528e-08, 1.0058e-07, 2.9616e-07, -1.3039e-08, + 3.9488e-07, -4.3027e-07, -5.0291e-07, -1.1735e-07, 2.2352e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 247.29, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5377 re_mapping 0.0048 re_causal 0.0138 /// teacc 99.10 lr 0.00010000 +Epoch 292, weight, value: tensor([[ 0.0345, -0.1639, -0.1473, ..., -0.3141, -0.1103, -0.1556], + [ 0.0576, -0.0730, 0.0379, ..., 0.0500, 0.1118, -0.0510], + [-0.0764, 0.1406, -0.1837, ..., 0.0586, 0.0809, -0.0479], + ..., + [-0.0719, -0.1029, -0.0797, ..., 0.0073, -0.1902, 0.1400], + [ 0.0568, -0.0310, 0.0897, ..., 0.0047, -0.2198, -0.0166], + [-0.2035, -0.1063, -0.1290, ..., -0.2255, 0.0648, -0.1058]], + device='cuda:0'), grad: tensor([[-3.7253e-09, 5.5879e-08, 3.1665e-08, ..., 5.9605e-08, + -3.1665e-08, 0.0000e+00], + [-3.7253e-08, 1.5460e-07, -3.3155e-07, ..., -1.4156e-07, + -4.9360e-07, 0.0000e+00], + [ 9.3132e-09, -1.4324e-06, 4.2841e-08, ..., -1.2256e-06, + -1.4976e-06, 0.0000e+00], + ..., + [ 1.8626e-09, 3.2410e-07, 9.4995e-08, ..., 3.4831e-07, + 5.1409e-07, 0.0000e+00], + [ 3.7253e-09, 1.0617e-07, 1.4901e-08, ..., 1.0245e-07, + 1.8254e-07, 0.0000e+00], + [ 1.8626e-08, 9.3132e-09, 2.0489e-08, ..., 1.8626e-08, + 3.7998e-07, 0.0000e+00]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0092, -0.0364, 0.0075, -0.0160, 0.0099, 0.0119, 0.0261, 0.0037, + -0.0397, -0.0063], device='cuda:0'), grad: tensor([-6.6459e-06, -4.6007e-07, -3.6061e-06, 2.2203e-06, -1.3672e-06, + 2.4214e-07, 3.1218e-06, 1.3411e-06, 6.0350e-07, 4.5374e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 246.99, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4871 re_mapping 0.0047 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 +Epoch 293, weight, value: tensor([[ 0.0346, -0.1643, -0.1474, ..., -0.3143, -0.1104, -0.1559], + [ 0.0578, -0.0734, 0.0382, ..., 0.0505, 0.1126, -0.0513], + [-0.0775, 0.1402, -0.1841, ..., 0.0583, 0.0809, -0.0480], + ..., + [-0.0714, -0.1023, -0.0799, ..., 0.0068, -0.1909, 0.1401], + [ 0.0567, -0.0310, 0.0899, ..., 0.0047, -0.2205, -0.0167], + [-0.2055, -0.1068, -0.1293, ..., -0.2271, 0.0647, -0.1061]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.8626e-09, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., -1.8626e-09, + -4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 1.8626e-08, ..., -3.7253e-09, + 5.2154e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -1.8626e-08, -1.8626e-09]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0090, -0.0359, 0.0073, -0.0160, 0.0102, 0.0119, 0.0259, 0.0033, + -0.0398, -0.0066], device='cuda:0'), grad: tensor([-1.9930e-07, 2.0489e-08, 3.5390e-08, -3.0808e-06, -4.3400e-07, + 3.0566e-06, -5.5879e-08, 4.6752e-07, 8.0094e-08, 1.0058e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 246.77, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.5066 re_mapping 0.0047 re_causal 0.0136 /// teacc 99.03 lr 0.00010000 +Epoch 294, weight, value: tensor([[ 0.0343, -0.1645, -0.1477, ..., -0.3148, -0.1110, -0.1575], + [ 0.0585, -0.0735, 0.0383, ..., 0.0506, 0.1128, -0.0515], + [-0.0789, 0.1406, -0.1845, ..., 0.0583, 0.0812, -0.0483], + ..., + [-0.0708, -0.1022, -0.0800, ..., 0.0069, -0.1911, 0.1405], + [ 0.0574, -0.0315, 0.0906, ..., 0.0050, -0.2214, -0.0178], + [-0.2061, -0.1069, -0.1292, ..., -0.2275, 0.0650, -0.1060]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 9.3132e-09, + 3.1665e-08, 0.0000e+00], + [-1.8626e-09, 4.0419e-07, -0.0000e+00, ..., 6.8545e-07, + 5.2527e-07, 0.0000e+00], + [ 0.0000e+00, -4.7497e-07, 7.4506e-09, ..., -7.8790e-07, + -1.1735e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 4.0978e-08, 4.6566e-08, ..., 9.8720e-08, + 6.7055e-08, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 3.7253e-09, ..., 9.3132e-09, + 3.9116e-08, 3.7253e-09], + [ 3.7253e-09, 5.5879e-09, 1.6764e-08, ..., 2.4214e-08, + 1.0245e-07, 0.0000e+00]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0095, -0.0358, 0.0073, -0.0162, 0.0099, 0.0118, 0.0265, 0.0033, + -0.0396, -0.0064], device='cuda:0'), grad: tensor([-1.6112e-06, 1.0170e-06, -2.1849e-06, -3.8743e-07, -3.2783e-07, + 1.0207e-06, 1.0021e-06, 2.8871e-07, 2.5891e-07, 9.3877e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 246.86, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4621 re_mapping 0.0045 re_causal 0.0131 /// teacc 99.05 lr 0.00010000 +Epoch 295, weight, value: tensor([[ 0.0342, -0.1649, -0.1478, ..., -0.3151, -0.1112, -0.1580], + [ 0.0588, -0.0739, 0.0382, ..., 0.0505, 0.1121, -0.0521], + [-0.0793, 0.1399, -0.1851, ..., 0.0572, 0.0819, -0.0487], + ..., + [-0.0707, -0.1000, -0.0800, ..., 0.0072, -0.1911, 0.1410], + [ 0.0592, -0.0313, 0.0912, ..., 0.0060, -0.2221, -0.0176], + [-0.2077, -0.1084, -0.1287, ..., -0.2283, 0.0653, -0.1061]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 3.3807e-07, 1.3039e-08, ..., 8.3819e-09, + 1.0747e-06, 0.0000e+00], + [-3.9116e-08, 1.8626e-09, -1.0312e-05, ..., -1.0416e-05, + -1.5229e-05, 0.0000e+00], + [ 0.0000e+00, 2.2352e-08, 4.3772e-08, ..., 3.7253e-08, + 1.5181e-07, 0.0000e+00], + ..., + [ 4.6566e-09, 3.7253e-09, 7.9870e-06, ..., 8.1435e-06, + 1.1772e-05, 0.0000e+00], + [ 1.1176e-08, 5.1595e-07, 3.4459e-08, ..., 1.5832e-08, + 1.6298e-06, 9.3132e-10], + [ 1.8626e-09, -8.9966e-07, 2.0415e-06, ..., 2.0880e-06, + 2.4866e-07, 0.0000e+00]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0092, -0.0364, 0.0067, -0.0162, 0.0097, 0.0117, 0.0267, 0.0037, + -0.0393, -0.0061], device='cuda:0'), grad: tensor([ 4.4517e-06, -6.3181e-05, 5.0385e-07, 8.2888e-08, 6.2771e-07, + 1.2852e-07, -1.7136e-07, 4.9442e-05, 6.6943e-06, 1.4333e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 246.92, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4742 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.09 lr 0.00010000 +Epoch 296, weight, value: tensor([[ 0.0341, -0.1652, -0.1483, ..., -0.3154, -0.1114, -0.1589], + [ 0.0595, -0.0741, 0.0379, ..., 0.0501, 0.1123, -0.0524], + [-0.0794, 0.1402, -0.1856, ..., 0.0572, 0.0821, -0.0489], + ..., + [-0.0708, -0.1005, -0.0796, ..., 0.0076, -0.1912, 0.1412], + [ 0.0588, -0.0306, 0.0915, ..., 0.0065, -0.2230, -0.0184], + [-0.2084, -0.1084, -0.1292, ..., -0.2293, 0.0653, -0.1062]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-09, -2.0489e-08, ..., 3.7253e-09, + 6.5193e-09, -1.1176e-07], + [ 2.7940e-09, 3.7253e-09, 2.5146e-08, ..., 5.5879e-09, + -1.2107e-08, 0.0000e+00], + [ 2.7940e-09, -2.2259e-07, 4.0047e-08, ..., -7.2643e-08, + -1.7416e-07, 1.3970e-08], + ..., + [ 1.8626e-09, 6.5193e-09, 2.0489e-08, ..., -1.3039e-08, + 1.3039e-08, 0.0000e+00], + [ 1.1176e-08, 1.7975e-07, 3.4459e-08, ..., 6.7055e-08, + 1.4994e-07, 1.8626e-08], + [ 3.3528e-08, 1.8626e-09, 1.9558e-08, ..., 3.7253e-09, + 1.9558e-08, 2.3283e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0092, -0.0369, 0.0066, -0.0168, 0.0098, 0.0124, 0.0268, 0.0042, + -0.0392, -0.0063], device='cuda:0'), grad: tensor([-1.2415e-06, 1.1455e-07, -3.7905e-07, -2.2538e-07, -3.0827e-07, + -1.5926e-07, 8.1398e-07, -1.0245e-08, 9.9558e-07, 4.1537e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 246.98, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.5013 re_mapping 0.0046 re_causal 0.0141 /// teacc 98.99 lr 0.00010000 +Epoch 297, weight, value: tensor([[ 0.0339, -0.1655, -0.1488, ..., -0.3157, -0.1112, -0.1591], + [ 0.0605, -0.0739, 0.0380, ..., 0.0501, 0.1126, -0.0524], + [-0.0800, 0.1404, -0.1861, ..., 0.0570, 0.0823, -0.0489], + ..., + [-0.0708, -0.1001, -0.0796, ..., 0.0077, -0.1913, 0.1413], + [ 0.0582, -0.0308, 0.0918, ..., 0.0064, -0.2238, -0.0187], + [-0.2103, -0.1091, -0.1297, ..., -0.2298, 0.0651, -0.1061]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.0151e-07, 4.0978e-08, ..., 6.1467e-08, + 1.3970e-07, 0.0000e+00], + [-5.5879e-09, 2.8871e-08, -2.1420e-08, ..., 5.2154e-08, + 2.9802e-08, 0.0000e+00], + [ 9.3132e-10, -3.3341e-07, -8.4750e-08, ..., -2.6729e-07, + -6.9756e-07, 0.0000e+00], + ..., + [ 1.8626e-09, 1.4901e-08, -3.7253e-09, ..., -2.4214e-08, + 2.9802e-08, 0.0000e+00], + [ 9.3132e-10, 8.9407e-08, -2.1420e-08, ..., 4.0047e-08, + 1.1642e-07, 0.0000e+00], + [ 0.0000e+00, 2.5146e-08, 2.2352e-08, ..., 2.5146e-08, + 2.7008e-08, -0.0000e+00]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0090, -0.0369, 0.0065, -0.0167, 0.0100, 0.0119, 0.0270, 0.0043, + -0.0393, -0.0066], device='cuda:0'), grad: tensor([ 5.4669e-07, 2.4401e-07, -1.9204e-06, 2.2911e-07, 1.2107e-08, + -3.5297e-07, 8.5216e-07, -1.5274e-07, 3.7625e-07, 1.6391e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 246.83, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.5075 re_mapping 0.0047 re_causal 0.0138 /// teacc 99.01 lr 0.00010000 +Epoch 298, weight, value: tensor([[ 0.0339, -0.1648, -0.1488, ..., -0.3158, -0.1112, -0.1600], + [ 0.0607, -0.0737, 0.0380, ..., 0.0501, 0.1133, -0.0527], + [-0.0801, 0.1403, -0.1875, ..., 0.0565, 0.0812, -0.0491], + ..., + [-0.0752, -0.1001, -0.0795, ..., 0.0076, -0.1917, 0.1423], + [ 0.0569, -0.0310, 0.0914, ..., 0.0065, -0.2253, -0.0191], + [-0.2108, -0.1096, -0.1298, ..., -0.2306, 0.0652, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 3.0734e-08, ..., 1.5832e-08, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 4.9360e-08, ..., 2.4214e-08, + -5.1223e-08, 0.0000e+00], + [ 0.0000e+00, -2.7847e-07, 6.1840e-07, ..., 7.6368e-08, + -1.8720e-07, -2.7940e-09], + ..., + [ 0.0000e+00, 8.0094e-08, 6.1188e-07, ..., 3.2596e-07, + 7.3574e-08, 9.3132e-10], + [ 0.0000e+00, 6.2399e-08, 1.8906e-07, ..., 1.7323e-07, + 2.2352e-08, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 9.9652e-08, ..., 4.6566e-09, + 1.5367e-07, 0.0000e+00]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0085, -0.0368, 0.0055, -0.0163, 0.0110, 0.0127, 0.0261, 0.0042, + -0.0404, -0.0067], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.2852e-07, 8.1211e-07, -3.5875e-06, -7.7486e-07, + 4.1071e-07, 1.8813e-07, 1.4259e-06, 4.6473e-07, 9.3039e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 247.10, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4632 re_mapping 0.0049 re_causal 0.0138 /// teacc 99.06 lr 0.00010000 +Epoch 299, weight, value: tensor([[ 0.0338, -0.1652, -0.1492, ..., -0.3160, -0.1114, -0.1602], + [ 0.0617, -0.0738, 0.0399, ..., 0.0508, 0.1138, -0.0528], + [-0.0803, 0.1403, -0.1885, ..., 0.0562, 0.0814, -0.0491], + ..., + [-0.0765, -0.1002, -0.0813, ..., 0.0071, -0.1920, 0.1427], + [ 0.0552, -0.0313, 0.0914, ..., 0.0065, -0.2264, -0.0192], + [-0.2114, -0.1098, -0.1304, ..., -0.2309, 0.0651, -0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.3819e-09, 9.3132e-09, ..., 5.5879e-09, + 1.3039e-08, 0.0000e+00], + [ 4.6566e-09, 3.7253e-09, 2.7940e-09, ..., 3.0734e-08, + -5.5879e-09, 3.7253e-09], + [ 0.0000e+00, -4.7497e-08, 1.0245e-08, ..., -2.3283e-08, + -5.6811e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.0245e-08, 2.1420e-08, ..., -1.3970e-08, + 1.5832e-08, 1.8626e-09], + [ 0.0000e+00, 3.7253e-09, -5.0291e-08, ..., 3.7253e-09, + 1.0245e-08, 0.0000e+00], + [ 9.3132e-10, 3.7253e-09, 2.4214e-08, ..., 1.2107e-08, + 3.7253e-09, -1.8626e-09]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0074, -0.0352, 0.0054, -0.0165, 0.0104, 0.0116, 0.0270, 0.0030, + -0.0407, -0.0069], device='cuda:0'), grad: tensor([ 5.4948e-08, 1.3225e-07, -9.7789e-08, -3.5297e-07, -4.8429e-08, + 3.7160e-07, -3.5390e-08, 0.0000e+00, -7.3574e-08, 7.4506e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 247.31, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4625 re_mapping 0.0047 re_causal 0.0134 /// teacc 99.10 lr 0.00010000 +Epoch 300, weight, value: tensor([[ 0.0338, -0.1664, -0.1499, ..., -0.3171, -0.1116, -0.1636], + [ 0.0617, -0.0748, 0.0399, ..., 0.0507, 0.1137, -0.0569], + [-0.0802, 0.1414, -0.1889, ..., 0.0564, 0.0820, -0.0477], + ..., + [-0.0762, -0.1006, -0.0813, ..., 0.0071, -0.1921, 0.1439], + [ 0.0552, -0.0317, 0.0921, ..., 0.0065, -0.2269, -0.0198], + [-0.2114, -0.1101, -0.1309, ..., -0.2320, 0.0652, -0.1057]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.4214e-08, 3.7253e-09, ..., 1.1176e-08, + 1.8626e-08, 0.0000e+00], + [-0.0000e+00, 1.3039e-08, -7.3947e-07, ..., -6.4448e-07, + -1.3821e-06, -1.6764e-08], + [ 0.0000e+00, -1.9185e-07, 8.1956e-08, ..., 1.6764e-08, + 5.2154e-08, 1.8626e-09], + ..., + [ 0.0000e+00, -2.7940e-08, 5.1595e-07, ..., 3.7253e-07, + 1.0114e-06, -9.3132e-09], + [ 0.0000e+00, 7.6368e-08, 0.0000e+00, ..., 3.3528e-08, + 4.8429e-08, 1.8626e-09], + [ 0.0000e+00, 9.3132e-09, 5.5879e-09, ..., 8.5682e-08, + 9.4995e-08, 1.8626e-09]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0075, -0.0354, 0.0058, -0.0163, 0.0103, 0.0114, 0.0272, 0.0032, + -0.0406, -0.0070], device='cuda:0'), grad: tensor([-1.3728e-06, -2.3898e-06, 3.5390e-08, 3.8929e-07, -1.4491e-06, + 1.6950e-07, 1.2964e-06, 9.8348e-07, 3.0361e-07, 2.0359e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 247.14, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4949 re_mapping 0.0047 re_causal 0.0139 /// teacc 99.11 lr 0.00010000 +Epoch 301, weight, value: tensor([[ 0.0338, -0.1667, -0.1504, ..., -0.3179, -0.1118, -0.1643], + [ 0.0619, -0.0754, 0.0400, ..., 0.0506, 0.1140, -0.0576], + [-0.0803, 0.1418, -0.1893, ..., 0.0564, 0.0824, -0.0478], + ..., + [-0.0762, -0.1003, -0.0814, ..., 0.0072, -0.1923, 0.1451], + [ 0.0552, -0.0318, 0.0935, ..., 0.0067, -0.2266, -0.0195], + [-0.2116, -0.1104, -0.1322, ..., -0.2339, 0.0651, -0.1073]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.2352e-08, 6.3330e-07, ..., 7.4506e-09, + 4.6566e-08, 1.8626e-09], + [-3.7253e-09, 9.1270e-08, 3.7253e-09, ..., 3.1665e-08, + 1.2666e-07, 0.0000e+00], + [ 0.0000e+00, -2.0862e-07, 5.9605e-08, ..., -2.0489e-08, + -3.6880e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.4901e-08, 3.7253e-08, ..., 2.4214e-08, + 2.9802e-08, 0.0000e+00], + [ 0.0000e+00, 6.3330e-08, -1.1995e-06, ..., 4.4703e-08, + 1.4529e-07, 5.5879e-09], + [ 0.0000e+00, 3.7253e-09, 4.2282e-07, ..., 1.8626e-09, + 3.4459e-07, 0.0000e+00]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0075, -0.0355, 0.0059, -0.0164, 0.0102, 0.0107, 0.0277, 0.0034, + -0.0399, -0.0073], device='cuda:0'), grad: tensor([ 1.2126e-06, 2.6263e-07, -3.9116e-07, -3.1665e-07, -2.6301e-06, + 2.2911e-07, 2.0303e-07, 1.7323e-07, -2.2743e-06, 3.5148e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 246.97, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4880 re_mapping 0.0046 re_causal 0.0143 /// teacc 99.04 lr 0.00010000 +Epoch 302, weight, value: tensor([[ 0.0338, -0.1673, -0.1507, ..., -0.3197, -0.1119, -0.1623], + [ 0.0619, -0.0758, 0.0404, ..., 0.0511, 0.1145, -0.0585], + [-0.0803, 0.1429, -0.1896, ..., 0.0568, 0.0831, -0.0467], + ..., + [-0.0762, -0.1004, -0.0818, ..., 0.0068, -0.1929, 0.1453], + [ 0.0552, -0.0328, 0.0947, ..., 0.0063, -0.2275, -0.0206], + [-0.2117, -0.1107, -0.1324, ..., -0.2357, 0.0652, -0.1093]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.1176e-08, ..., 3.7253e-09, + 1.8626e-08, 1.8626e-09], + [ 0.0000e+00, 3.7253e-08, -3.9116e-08, ..., 3.5390e-08, + -5.4017e-08, 0.0000e+00], + [ 0.0000e+00, -6.7428e-07, 5.4017e-08, ..., -8.1398e-07, + -4.5262e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 5.9605e-07, 2.7940e-08, ..., 7.2271e-07, + 4.4890e-07, -0.0000e+00], + [ 0.0000e+00, 9.3132e-09, -2.5705e-07, ..., 1.1176e-08, + -1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 7.6368e-08, ..., 7.4506e-09, + 7.2643e-08, 0.0000e+00]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0071, -0.0347, 0.0066, -0.0166, 0.0102, 0.0115, 0.0264, 0.0026, + -0.0396, -0.0076], device='cuda:0'), grad: tensor([ 5.1409e-07, 1.2666e-07, -2.7679e-06, 3.4422e-06, -1.3039e-07, + -7.2904e-06, 4.6752e-07, 2.6356e-06, -2.8685e-07, 3.2987e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 247.06, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4844 re_mapping 0.0045 re_causal 0.0139 /// teacc 99.08 lr 0.00010000 +Epoch 303, weight, value: tensor([[ 0.0337, -0.1678, -0.1513, ..., -0.3198, -0.1120, -0.1624], + [ 0.0619, -0.0760, 0.0401, ..., 0.0510, 0.1147, -0.0585], + [-0.0803, 0.1439, -0.1903, ..., 0.0568, 0.0831, -0.0469], + ..., + [-0.0763, -0.1006, -0.0815, ..., 0.0070, -0.1930, 0.1458], + [ 0.0552, -0.0338, 0.0949, ..., 0.0060, -0.2279, -0.0207], + [-0.2118, -0.1109, -0.1327, ..., -0.2364, 0.0650, -0.1093]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 1.8626e-09, ..., 1.8626e-09, + 7.4506e-09, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, -2.0489e-08, ..., 2.0489e-08, + -5.0291e-08, 0.0000e+00], + [ 0.0000e+00, -1.4901e-08, 2.3842e-07, ..., 8.3819e-08, + -1.1735e-07, -2.9802e-08], + ..., + [ 0.0000e+00, 3.7253e-09, 6.5193e-08, ..., -2.7940e-08, + 2.7940e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.0489e-08, ..., 1.3039e-08, + 7.4506e-07, 0.0000e+00]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0070, -0.0351, 0.0067, -0.0167, 0.0105, 0.0114, 0.0265, 0.0030, + -0.0397, -0.0078], device='cuda:0'), grad: tensor([-1.3039e-08, -2.0489e-08, 3.5577e-07, -6.6124e-07, -3.0734e-06, + 3.7253e-08, 1.6019e-07, 3.1665e-08, 1.1176e-08, 3.1553e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 246.98, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4887 re_mapping 0.0045 re_causal 0.0139 /// teacc 98.97 lr 0.00010000 +Epoch 304, weight, value: tensor([[ 0.0338, -0.1680, -0.1514, ..., -0.3199, -0.1121, -0.1625], + [ 0.0620, -0.0757, 0.0401, ..., 0.0507, 0.1149, -0.0586], + [-0.0804, 0.1445, -0.1915, ..., 0.0567, 0.0826, -0.0471], + ..., + [-0.0763, -0.1006, -0.0814, ..., 0.0074, -0.1930, 0.1463], + [ 0.0552, -0.0346, 0.0951, ..., 0.0055, -0.2284, -0.0209], + [-0.2119, -0.1109, -0.1330, ..., -0.2378, 0.0650, -0.1093]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 2.9802e-08, ..., 1.1176e-08, + 3.3528e-08, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, -7.4878e-07, ..., -2.3097e-07, + -2.0470e-06, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 1.5460e-07, ..., 5.0291e-08, + 2.8498e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-08, ..., 1.3039e-08, + 6.1467e-08, -0.0000e+00], + [-1.1176e-08, -7.4506e-09, -5.8301e-07, ..., -2.0675e-07, + -9.6858e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0070, -0.0354, 0.0064, -0.0166, 0.0105, 0.0112, 0.0264, 0.0035, + -0.0398, -0.0080], device='cuda:0'), grad: tensor([ 6.7055e-08, -3.1833e-06, 5.5507e-07, 3.5204e-07, 2.4959e-07, + -2.7195e-07, 3.0268e-06, 1.5460e-07, -1.0785e-06, 1.1548e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 246.84, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4568 re_mapping 0.0044 re_causal 0.0127 /// teacc 99.03 lr 0.00010000 +Epoch 305, weight, value: tensor([[ 0.0337, -0.1685, -0.1525, ..., -0.3203, -0.1123, -0.1626], + [ 0.0621, -0.0762, 0.0407, ..., 0.0508, 0.1154, -0.0586], + [-0.0805, 0.1445, -0.1922, ..., 0.0567, 0.0831, -0.0473], + ..., + [-0.0763, -0.1000, -0.0820, ..., 0.0073, -0.1937, 0.1470], + [ 0.0553, -0.0348, 0.0954, ..., 0.0054, -0.2292, -0.0210], + [-0.2120, -0.1111, -0.1334, ..., -0.2400, 0.0649, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7998e-07, ..., 2.4587e-07, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 8.0094e-08, ..., 7.0781e-08, + -5.4017e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-08, 1.8254e-07, ..., 8.7544e-08, + -1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 1.0245e-07, ..., 5.7742e-08, + 3.7253e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 4.6380e-07, ..., 3.0361e-07, + 2.7940e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.7055e-08, ..., 4.2841e-08, + -2.0675e-07, 0.0000e+00]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0068, -0.0351, 0.0063, -0.0163, 0.0106, 0.0111, 0.0266, 0.0034, + -0.0402, -0.0084], device='cuda:0'), grad: tensor([-1.1027e-05, 2.8685e-07, 1.5721e-06, -5.3272e-06, 1.6019e-07, + 4.0568e-06, 6.8024e-06, 2.5146e-07, 2.5406e-06, 6.2399e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 246.99, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4811 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.02 lr 0.00010000 +Epoch 306, weight, value: tensor([[ 0.0335, -0.1687, -0.1532, ..., -0.3206, -0.1124, -0.1626], + [ 0.0621, -0.0764, 0.0409, ..., 0.0508, 0.1160, -0.0586], + [-0.0807, 0.1447, -0.1928, ..., 0.0567, 0.0829, -0.0474], + ..., + [-0.0764, -0.1001, -0.0822, ..., 0.0073, -0.1940, 0.1478], + [ 0.0553, -0.0349, 0.0962, ..., 0.0059, -0.2297, -0.0211], + [-0.2127, -0.1111, -0.1338, ..., -0.2403, 0.0643, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, -7.4506e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 4.1351e-07, 5.5879e-09, ..., 5.1782e-07, + 1.8626e-09, 0.0000e+00], + ..., + [-1.8626e-09, 5.1036e-07, 1.1176e-08, ..., 6.3702e-07, + 8.5682e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 7.4506e-09, ..., 7.4506e-09, + 4.8429e-08, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + -7.8790e-07, 0.0000e+00]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0067, -0.0350, 0.0060, -0.0163, 0.0111, 0.0110, 0.0265, 0.0035, + -0.0402, -0.0090], device='cuda:0'), grad: tensor([ 2.4214e-08, 6.5193e-08, 8.1956e-07, -1.8831e-06, 1.8161e-06, + 1.1362e-07, -8.7544e-08, 1.1399e-06, 1.1176e-07, -2.1067e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 247.88, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4920 re_mapping 0.0044 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +Epoch 307, weight, value: tensor([[ 0.0335, -0.1691, -0.1533, ..., -0.3209, -0.1124, -0.1626], + [ 0.0625, -0.0775, 0.0410, ..., 0.0510, 0.1165, -0.0586], + [-0.0805, 0.1465, -0.1934, ..., 0.0571, 0.0843, -0.0474], + ..., + [-0.0764, -0.1021, -0.0824, ..., 0.0070, -0.1948, 0.1477], + [ 0.0550, -0.0349, 0.0964, ..., 0.0060, -0.2307, -0.0215], + [-0.2129, -0.1116, -0.1340, ..., -0.2420, 0.0638, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, -1.8626e-09, ..., 5.5879e-09, + -2.6077e-08, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 1.6764e-08, ..., 1.8626e-09, + -9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 2.4214e-08, ..., 2.4214e-08, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.8429e-08, ..., -1.6764e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0062, -0.0348, 0.0068, -0.0162, 0.0118, 0.0110, 0.0266, 0.0032, + -0.0403, -0.0097], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.8626e-08, 9.3132e-09, -5.1893e-06, -7.4506e-09, + 5.1111e-06, 2.7940e-08, 7.2643e-08, -8.9407e-08, 2.9802e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 247.89, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4841 re_mapping 0.0046 re_causal 0.0139 /// teacc 98.97 lr 0.00010000 +Epoch 308, weight, value: tensor([[ 0.0335, -0.1691, -0.1533, ..., -0.3220, -0.1128, -0.1626], + [ 0.0625, -0.0772, 0.0411, ..., 0.0510, 0.1166, -0.0586], + [-0.0807, 0.1464, -0.1940, ..., 0.0572, 0.0847, -0.0474], + ..., + [-0.0764, -0.1022, -0.0824, ..., 0.0070, -0.1950, 0.1477], + [ 0.0550, -0.0349, 0.0965, ..., 0.0059, -0.2316, -0.0215], + [-0.2130, -0.1116, -0.1342, ..., -0.2424, 0.0638, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 7.4506e-09, ..., 7.4506e-09, + 6.3330e-08, 0.0000e+00], + [ 5.5879e-09, 1.8626e-09, 4.0792e-07, ..., 3.2410e-07, + -4.2841e-08, 0.0000e+00], + [ 0.0000e+00, -1.6764e-08, 2.2352e-07, ..., 1.4529e-07, + 6.8918e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -9.3132e-09, 3.7253e-08, ..., -2.3283e-07, + 1.3970e-07, 0.0000e+00], + [-1.4901e-08, 1.1176e-08, -1.1642e-06, ..., -6.3889e-07, + -1.2107e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.0291e-08, ..., 3.5390e-08, + 4.7684e-07, 0.0000e+00]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0067, -0.0348, 0.0070, -0.0161, 0.0117, 0.0109, 0.0273, 0.0032, + -0.0404, -0.0098], device='cuda:0'), grad: tensor([ 1.7509e-07, 1.5777e-06, 5.8115e-07, -5.5879e-08, -2.1439e-06, + 1.2573e-06, 3.0175e-07, -3.9116e-07, -3.0529e-06, 1.7565e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 247.48, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4957 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.04 lr 0.00010000 +Epoch 309, weight, value: tensor([[ 0.0333, -0.1693, -0.1541, ..., -0.3229, -0.1160, -0.1629], + [ 0.0624, -0.0772, 0.0414, ..., 0.0510, 0.1177, -0.0585], + [-0.0809, 0.1466, -0.1963, ..., 0.0566, 0.0834, -0.0476], + ..., + [-0.0765, -0.1023, -0.0826, ..., 0.0072, -0.1952, 0.1482], + [ 0.0549, -0.0349, 0.0972, ..., 0.0061, -0.2318, -0.0216], + [-0.2132, -0.1116, -0.1349, ..., -0.2431, 0.0665, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.3330e-08, ..., 8.3819e-08, + 7.0781e-08, 0.0000e+00], + [ 1.8626e-09, -2.2352e-08, -3.3528e-08, ..., 5.0664e-07, + 3.8929e-07, 0.0000e+00], + [ 5.5879e-09, 1.8626e-09, 1.4156e-07, ..., -2.1346e-06, + -2.1551e-06, 0.0000e+00], + ..., + [ 1.1176e-08, 1.8626e-09, 1.3970e-07, ..., 1.1921e-06, + 1.2275e-06, 0.0000e+00], + [-2.9802e-08, 0.0000e+00, -1.8254e-07, ..., 9.8720e-08, + 1.1176e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 1.4901e-08, + 2.3656e-07, -0.0000e+00]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0097, -0.0348, 0.0064, -0.0159, 0.0117, 0.0105, 0.0269, 0.0034, + -0.0399, -0.0074], device='cuda:0'), grad: tensor([ 3.8743e-07, 1.7621e-06, -6.1691e-06, -6.6236e-06, -5.7183e-07, + 6.5267e-06, 2.4214e-07, 3.9116e-06, -1.1921e-07, 6.4634e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 247.75, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4888 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.05 lr 0.00010000 +Epoch 310, weight, value: tensor([[ 0.0331, -0.1693, -0.1543, ..., -0.3231, -0.1169, -0.1629], + [ 0.0626, -0.0772, 0.0414, ..., 0.0510, 0.1179, -0.0585], + [-0.0812, 0.1466, -0.1969, ..., 0.0564, 0.0835, -0.0476], + ..., + [-0.0765, -0.1024, -0.0826, ..., 0.0074, -0.1954, 0.1484], + [ 0.0542, -0.0349, 0.0980, ..., 0.0070, -0.2319, -0.0218], + [-0.2147, -0.1117, -0.1355, ..., -0.2465, 0.0672, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., 5.0291e-08, + -5.9605e-08, 0.0000e+00], + [ 0.0000e+00, -9.3132e-09, 1.3597e-07, ..., -5.5879e-09, + -1.6019e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 7.4506e-09, 1.6764e-08, ..., -6.1467e-08, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.1176e-08, ..., 9.3132e-09, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 9.3132e-09, + -3.3528e-08, 0.0000e+00]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0102, -0.0348, 0.0061, -0.0160, 0.0119, 0.0106, 0.0265, 0.0035, + -0.0397, -0.0070], device='cuda:0'), grad: tensor([-3.3528e-08, 8.9407e-08, 3.3528e-08, -2.4028e-07, 2.6636e-07, + 7.4506e-08, -2.6077e-08, -1.1921e-07, 2.0489e-08, -9.3132e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 247.43, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4760 re_mapping 0.0044 re_causal 0.0136 /// teacc 98.99 lr 0.00010000 +Epoch 311, weight, value: tensor([[ 0.0326, -0.1694, -0.1548, ..., -0.3232, -0.1183, -0.1629], + [ 0.0629, -0.0778, 0.0414, ..., 0.0508, 0.1179, -0.0582], + [-0.0813, 0.1467, -0.1979, ..., 0.0556, 0.0834, -0.0476], + ..., + [-0.0766, -0.1024, -0.0826, ..., 0.0076, -0.1954, 0.1485], + [ 0.0541, -0.0354, 0.0982, ..., 0.0069, -0.2324, -0.0219], + [-0.2166, -0.1117, -0.1358, ..., -0.2473, 0.0683, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.6368e-08, ..., 2.9802e-08, + 4.2282e-06, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, -9.9652e-07, ..., -4.1164e-07, + -1.2983e-06, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, 4.9174e-07, ..., 1.8999e-07, + 6.8769e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 8.1956e-08, ..., 4.0978e-08, + 1.0058e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 9.1270e-08, ..., 4.4703e-08, + 2.2911e-07, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + 1.0245e-07, 0.0000e+00]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0114, -0.0349, 0.0053, -0.0151, 0.0127, 0.0105, 0.0271, 0.0037, + -0.0400, -0.0065], device='cuda:0'), grad: tensor([ 2.1532e-05, -2.0266e-06, 3.4511e-05, 4.8243e-07, 5.5134e-06, + 4.4890e-07, -6.2585e-05, 2.0489e-07, 6.0722e-07, 1.1604e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 247.63, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4844 re_mapping 0.0044 re_causal 0.0129 /// teacc 99.10 lr 0.00010000 +Epoch 312, weight, value: tensor([[ 0.0326, -0.1696, -0.1557, ..., -0.3235, -0.1183, -0.1631], + [ 0.0619, -0.0779, 0.0417, ..., 0.0509, 0.1185, -0.0581], + [-0.0814, 0.1471, -0.1985, ..., 0.0554, 0.0831, -0.0477], + ..., + [-0.0753, -0.1029, -0.0828, ..., 0.0073, -0.1956, 0.1487], + [ 0.0541, -0.0355, 0.0986, ..., 0.0068, -0.2331, -0.0220], + [-0.2187, -0.1119, -0.1370, ..., -0.2484, 0.0682, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 5.5879e-08, ..., 6.2212e-07, + 1.2759e-06, 1.6764e-08], + [ 0.0000e+00, 1.8626e-09, -1.1493e-06, ..., -2.0117e-07, + -1.7844e-06, -5.2154e-07], + [ 0.0000e+00, -1.1176e-08, 1.1921e-07, ..., -7.2084e-07, + -1.5572e-06, -1.1176e-08], + ..., + [ 0.0000e+00, 5.5879e-09, 6.3330e-08, ..., 1.4901e-08, + 4.4703e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -8.6799e-06, ..., 1.0058e-07, + 1.0617e-07, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 7.4506e-09, + -2.2352e-08, 0.0000e+00]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0115, -0.0348, 0.0049, -0.0147, 0.0131, 0.0103, 0.0277, 0.0036, + -0.0401, -0.0068], device='cuda:0'), grad: tensor([ 3.7141e-06, -3.6955e-06, -4.2915e-06, -6.3330e-08, 6.7614e-07, + 1.9930e-07, 1.7002e-05, 1.2852e-07, -1.3843e-05, 1.4342e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 247.52, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4585 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.08 lr 0.00010000 +Epoch 313, weight, value: tensor([[ 0.0323, -0.1699, -0.1579, ..., -0.3243, -0.1183, -0.1634], + [ 0.0618, -0.0775, 0.0417, ..., 0.0510, 0.1191, -0.0581], + [-0.0815, 0.1473, -0.2002, ..., 0.0552, 0.0832, -0.0471], + ..., + [-0.0752, -0.1031, -0.0828, ..., 0.0072, -0.1962, 0.1489], + [ 0.0542, -0.0365, 0.0988, ..., 0.0055, -0.2347, -0.0224], + [-0.2188, -0.1119, -0.1376, ..., -0.2492, 0.0682, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 2.4214e-08, + 1.1176e-08, 1.8626e-09], + [-0.0000e+00, 1.8626e-09, -9.3132e-09, ..., 2.7120e-06, + -5.5879e-08, 1.8626e-09], + [ 0.0000e+00, -3.9116e-08, 5.7742e-08, ..., -2.8498e-07, + -3.3341e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., -3.7681e-06, + 7.4506e-09, -1.1735e-07], + [ 0.0000e+00, 3.7253e-08, 1.1362e-07, ..., 4.4331e-07, + 3.6508e-07, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 6.7055e-08, + -4.6566e-08, 1.0058e-07]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0116, -0.0348, 0.0047, -0.0136, 0.0132, 0.0104, 0.0275, 0.0036, + -0.0410, -0.0069], device='cuda:0'), grad: tensor([ 1.1735e-07, 6.6832e-06, -5.0850e-07, 9.1270e-07, 2.4028e-07, + 9.1642e-07, -2.1048e-07, -1.0625e-05, 1.2368e-06, 1.2219e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 248.01, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.5104 re_mapping 0.0044 re_causal 0.0129 /// teacc 99.07 lr 0.00010000 +Epoch 314, weight, value: tensor([[ 0.0322, -0.1705, -0.1580, ..., -0.3249, -0.1183, -0.1632], + [ 0.0622, -0.0783, 0.0403, ..., 0.0498, 0.1192, -0.0585], + [-0.0818, 0.1473, -0.2017, ..., 0.0548, 0.0832, -0.0472], + ..., + [-0.0753, -0.1014, -0.0815, ..., 0.0085, -0.1968, 0.1492], + [ 0.0542, -0.0375, 0.0987, ..., 0.0045, -0.2365, -0.0228], + [-0.2189, -0.1120, -0.1382, ..., -0.2504, 0.0685, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 2.0489e-08, + 3.3528e-08, 0.0000e+00], + [ 0.0000e+00, 3.5763e-07, 7.4506e-09, ..., 7.7486e-07, + 1.4491e-06, 0.0000e+00], + [ 0.0000e+00, -5.3085e-07, 2.6077e-08, ..., -1.0375e-06, + -1.8850e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 9.6858e-08, 5.4017e-08, ..., 1.9185e-07, + 6.2212e-07, -1.8626e-09], + [ 0.0000e+00, 1.3039e-08, 9.3132e-09, ..., 2.6077e-08, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 5.5879e-09, + -1.1381e-06, 0.0000e+00]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0115, -0.0363, 0.0041, -0.0132, 0.0128, 0.0100, 0.0287, 0.0048, + -0.0424, -0.0066], device='cuda:0'), grad: tensor([ 5.0291e-08, 2.5444e-06, -3.4533e-06, -1.1176e-08, 4.9546e-06, + 8.5682e-08, 7.8231e-08, 3.1702e-06, 1.8440e-07, -7.6145e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 247.84, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4694 re_mapping 0.0046 re_causal 0.0134 /// teacc 98.97 lr 0.00010000 +Epoch 315, weight, value: tensor([[ 0.0318, -0.1703, -0.1576, ..., -0.3253, -0.1184, -0.1632], + [ 0.0624, -0.0783, 0.0403, ..., 0.0496, 0.1190, -0.0585], + [-0.0798, 0.1475, -0.2013, ..., 0.0563, 0.0849, -0.0472], + ..., + [-0.0746, -0.1014, -0.0815, ..., 0.0087, -0.1972, 0.1492], + [ 0.0541, -0.0378, 0.0990, ..., 0.0042, -0.2372, -0.0228], + [-0.2189, -0.1124, -0.1391, ..., -0.2529, 0.0682, -0.1097]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 5.5879e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -3.1665e-08, ..., -0.0000e+00, + -2.6077e-08, 0.0000e+00], + [ 1.8626e-09, -1.8626e-09, 3.7253e-09, ..., -0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, 1.1176e-08, ..., -2.2352e-08, + 2.9802e-08, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 1.6764e-08, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0115, -0.0366, 0.0060, -0.0156, 0.0126, 0.0116, 0.0290, 0.0051, + -0.0428, -0.0069], device='cuda:0'), grad: tensor([-5.5879e-09, -1.4901e-08, 2.6077e-08, 1.3039e-08, -4.1537e-07, + 2.6636e-07, -2.0303e-07, 4.6566e-08, 8.5682e-08, 1.9372e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 247.60, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.4749 re_mapping 0.0044 re_causal 0.0129 /// teacc 99.13 lr 0.00010000 +Epoch 316, weight, value: tensor([[ 0.0313, -0.1705, -0.1582, ..., -0.3256, -0.1184, -0.1632], + [ 0.0623, -0.0786, 0.0402, ..., 0.0494, 0.1193, -0.0585], + [-0.0799, 0.1480, -0.2022, ..., 0.0565, 0.0854, -0.0472], + ..., + [-0.0739, -0.1018, -0.0815, ..., 0.0088, -0.1977, 0.1493], + [ 0.0544, -0.0378, 0.1017, ..., 0.0060, -0.2361, -0.0228], + [-0.2199, -0.1124, -0.1415, ..., -0.2548, 0.0684, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 5.5879e-09, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, -1.8626e-09, ..., 1.1176e-08, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, -5.5879e-08, 3.7253e-09, ..., -1.3039e-07, + -8.7544e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.6764e-08, 1.8626e-09, ..., 3.7253e-08, + 3.1665e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0112, -0.0367, 0.0061, -0.0160, 0.0125, 0.0117, 0.0291, 0.0049, + -0.0412, -0.0070], device='cuda:0'), grad: tensor([-7.7337e-06, 8.5682e-08, 6.5193e-08, 6.5193e-07, 2.8498e-07, + -2.0675e-07, 5.4166e-06, 1.5646e-07, 3.5949e-07, 8.9407e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 315---------------------------------------------------- +epoch 315, time 262.52, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4901 re_mapping 0.0043 re_causal 0.0132 /// teacc 99.15 lr 0.00010000 +Epoch 317, weight, value: tensor([[ 0.0311, -0.1708, -0.1592, ..., -0.3269, -0.1189, -0.1632], + [ 0.0639, -0.0789, 0.0411, ..., 0.0502, 0.1209, -0.0586], + [-0.0801, 0.1491, -0.2031, ..., 0.0564, 0.0860, -0.0472], + ..., + [-0.0741, -0.1035, -0.0823, ..., 0.0081, -0.1994, 0.1496], + [ 0.0532, -0.0378, 0.1018, ..., 0.0059, -0.2368, -0.0229], + [-0.2202, -0.1127, -0.1421, ..., -0.2574, 0.0688, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.9605e-08, ..., 2.4214e-08, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, 3.7253e-08, ..., 9.3132e-09, + 2.7940e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 3.3528e-08, ..., 7.4506e-09, + 1.4901e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.5635e-07, ..., -1.1735e-07, + -3.7253e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4529e-07, ..., 7.4506e-09, + 2.9802e-08, 0.0000e+00]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0117, -0.0357, 0.0062, -0.0156, 0.0125, 0.0116, 0.0286, 0.0042, + -0.0415, -0.0068], device='cuda:0'), grad: tensor([-1.0058e-07, 1.8254e-07, 1.9372e-07, -5.7742e-08, -1.7509e-07, + 2.7940e-07, 2.2724e-07, 9.1270e-08, -1.1902e-06, 5.2899e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 247.74, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4621 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.07 lr 0.00010000 +Epoch 318, weight, value: tensor([[ 0.0304, -0.1709, -0.1607, ..., -0.3272, -0.1195, -0.1633], + [ 0.0653, -0.0792, 0.0415, ..., 0.0504, 0.1214, -0.0586], + [-0.0802, 0.1492, -0.2039, ..., 0.0563, 0.0860, -0.0472], + ..., + [-0.0742, -0.1037, -0.0826, ..., 0.0080, -0.1999, 0.1497], + [ 0.0524, -0.0374, 0.1025, ..., 0.0059, -0.2371, -0.0228], + [-0.2210, -0.1128, -0.1428, ..., -0.2589, 0.0692, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 2.4214e-08, 0.0000e+00], + [-1.3039e-08, 1.8626e-09, -8.3819e-08, ..., -1.8626e-09, + -1.0431e-07, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 8.5682e-08, ..., 7.0781e-08, + 1.8626e-08, 0.0000e+00], + ..., + [ 1.8626e-09, -5.5879e-09, 4.0978e-08, ..., 9.3132e-09, + 4.6566e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 7.4506e-09, ..., 7.4506e-09, + 5.8450e-06, 0.0000e+00]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0123, -0.0355, 0.0060, -0.0154, 0.0129, 0.0116, 0.0283, 0.0041, + -0.0413, -0.0066], device='cuda:0'), grad: tensor([ 2.7940e-08, -1.9930e-07, 3.7812e-07, -7.8045e-07, -1.8775e-05, + 2.8312e-07, 1.0617e-07, 1.7695e-07, 8.5682e-08, 1.8686e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 247.55, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4965 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.07 lr 0.00010000 +Epoch 319, weight, value: tensor([[ 0.0300, -0.1713, -0.1606, ..., -0.3284, -0.1196, -0.1633], + [ 0.0662, -0.0795, 0.0415, ..., 0.0503, 0.1212, -0.0586], + [-0.0804, 0.1498, -0.2047, ..., 0.0565, 0.0866, -0.0471], + ..., + [-0.0751, -0.1037, -0.0826, ..., 0.0082, -0.2002, 0.1497], + [ 0.0519, -0.0376, 0.1026, ..., 0.0053, -0.2384, -0.0229], + [-0.2212, -0.1134, -0.1435, ..., -0.2603, 0.0692, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7567e-07, 5.5879e-09, ..., 2.8312e-07, + 3.4459e-07, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, -2.2911e-07, ..., 1.0058e-07, + -1.7136e-07, 0.0000e+00], + [ 0.0000e+00, -8.5831e-06, 3.7253e-09, ..., -8.0168e-06, + -6.9477e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 6.1318e-06, -1.3039e-08, ..., 5.8971e-06, + 6.2138e-06, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 1.1176e-08, ..., 1.3039e-08, + 9.1456e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 1.8626e-09, ..., 2.0489e-08, + -3.3658e-06, 0.0000e+00]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0123, -0.0358, 0.0064, -0.0153, 0.0134, 0.0113, 0.0290, 0.0043, + -0.0415, -0.0067], device='cuda:0'), grad: tensor([ 1.3504e-06, 2.4959e-07, -3.3647e-05, 4.5002e-06, 8.0615e-06, + 1.2852e-07, 6.4075e-07, 2.8253e-05, 3.5781e-06, -1.3113e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 247.68, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4978 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.14 lr 0.00010000 +Epoch 320, weight, value: tensor([[ 0.0304, -0.1716, -0.1609, ..., -0.3288, -0.1196, -0.1634], + [ 0.0667, -0.0795, 0.0390, ..., 0.0481, 0.1201, -0.0581], + [-0.0821, 0.1531, -0.2047, ..., 0.0579, 0.0881, -0.0467], + ..., + [-0.0746, -0.1044, -0.0800, ..., 0.0104, -0.1993, 0.1497], + [ 0.0511, -0.0418, 0.1029, ..., 0.0021, -0.2418, -0.0229], + [-0.2218, -0.1136, -0.1450, ..., -0.2606, 0.0686, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.4622e-07, ..., -5.5879e-09, + -1.4994e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 9.3132e-09, + 3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.2841e-08, ..., 1.1176e-08, + 3.5390e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -3.7253e-09, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0122, -0.0384, 0.0077, -0.0151, 0.0154, 0.0111, 0.0286, 0.0066, + -0.0425, -0.0074], device='cuda:0'), grad: tensor([-1.6550e-06, -3.9302e-07, 1.2293e-07, -1.6950e-07, 1.1083e-07, + 2.3283e-07, 1.1893e-06, 1.0990e-07, 7.0781e-08, 3.7253e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 247.65, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4931 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.12 lr 0.00010000 +Epoch 321, weight, value: tensor([[ 3.0378e-02, -1.7106e-01, -1.6109e-01, ..., -3.2938e-01, + -1.1967e-01, -1.6352e-01], + [ 6.6724e-02, -7.9626e-02, 4.1876e-02, ..., 5.1007e-02, + 1.2290e-01, -5.8057e-02], + [-8.2117e-02, 1.5526e-01, -2.0492e-01, ..., 5.9575e-02, + 9.0294e-02, -4.6785e-02], + ..., + [-7.4841e-02, -1.0425e-01, -8.2905e-02, ..., 7.3642e-03, + -2.0240e-01, 1.4974e-01], + [ 5.1117e-02, -4.4042e-02, 1.0306e-01, ..., -2.9579e-04, + -2.4396e-01, -2.3038e-02], + [-2.2142e-01, -1.1395e-01, -1.4675e-01, ..., -2.6159e-01, + 6.8663e-02, -1.0971e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 5.5879e-09, + -2.6077e-08, 0.0000e+00], + [ 0.0000e+00, -1.4529e-07, 2.2352e-08, ..., -1.9744e-07, + -2.8312e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 7.4506e-09, ..., -1.4901e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0123, -0.0355, 0.0092, -0.0147, 0.0154, 0.0119, 0.0273, 0.0037, + -0.0439, -0.0076], device='cuda:0'), grad: tensor([-9.1828e-07, -1.1176e-08, -5.4017e-07, 5.1968e-07, 2.4214e-08, + 1.0245e-07, 3.1292e-07, -1.0058e-07, 3.1665e-08, 5.7183e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 247.58, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4941 re_mapping 0.0044 re_causal 0.0133 /// teacc 99.10 lr 0.00010000 +Epoch 322, weight, value: tensor([[ 0.0303, -0.1707, -0.1615, ..., -0.3296, -0.1197, -0.1635], + [ 0.0668, -0.0798, 0.0397, ..., 0.0495, 0.1212, -0.0581], + [-0.0821, 0.1547, -0.2017, ..., 0.0601, 0.0934, -0.0468], + ..., + [-0.0749, -0.1017, -0.0811, ..., 0.0089, -0.2018, 0.1496], + [ 0.0511, -0.0453, 0.1035, ..., -0.0017, -0.2454, -0.0231], + [-0.2214, -0.1142, -0.1477, ..., -0.2626, 0.0687, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.5682e-08, ..., 3.7253e-09, + 5.7742e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 5.5879e-08, + -1.6764e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 5.4017e-08, ..., -1.8626e-09, + -4.8429e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -1.0990e-07, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.4273e-07, ..., 1.4901e-08, + -5.0291e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-08, ..., 1.1176e-08, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0121, -0.0371, 0.0104, -0.0149, 0.0151, 0.0117, 0.0272, 0.0052, + -0.0447, -0.0077], device='cuda:0'), grad: tensor([ 2.3842e-07, 1.6764e-07, -2.2352e-08, 1.0990e-07, 6.1467e-08, + 2.4214e-07, 3.1665e-08, -5.6624e-07, -5.9232e-07, 3.2224e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 247.45, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4571 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.02 lr 0.00010000 +Epoch 323, weight, value: tensor([[ 0.0305, -0.1709, -0.1618, ..., -0.3300, -0.1198, -0.1639], + [ 0.0668, -0.0798, 0.0399, ..., 0.0497, 0.1215, -0.0579], + [-0.0822, 0.1548, -0.2023, ..., 0.0594, 0.0930, -0.0468], + ..., + [-0.0749, -0.1017, -0.0812, ..., 0.0088, -0.2019, 0.1497], + [ 0.0511, -0.0453, 0.1039, ..., -0.0019, -0.2455, -0.0231], + [-0.2216, -0.1146, -0.1483, ..., -0.2638, 0.0687, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 2.4214e-08, + 3.1665e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.5542e-05, ..., -7.2084e-06, + -1.3866e-05, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.0952e-05, ..., 5.8115e-06, + 9.7528e-06, 1.6764e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.9823e-06, ..., 9.2387e-07, + 3.5185e-06, -1.4156e-07], + [ 0.0000e+00, 0.0000e+00, -2.5332e-07, ..., -2.1234e-07, + 1.8626e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 5.9605e-08, + 4.2841e-08, 1.8626e-09]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0122, -0.0369, 0.0100, -0.0149, 0.0150, 0.0116, 0.0275, 0.0051, + -0.0449, -0.0078], device='cuda:0'), grad: tensor([ 1.6019e-07, -2.7955e-05, 2.3350e-05, 1.8291e-06, 2.4997e-06, + 8.1956e-08, 3.3155e-07, -6.5193e-08, -6.5006e-07, 3.8370e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 247.49, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4862 re_mapping 0.0045 re_causal 0.0134 /// teacc 98.97 lr 0.00010000 +Epoch 324, weight, value: tensor([[ 0.0301, -0.1711, -0.1623, ..., -0.3303, -0.1198, -0.1643], + [ 0.0668, -0.0794, 0.0399, ..., 0.0497, 0.1216, -0.0580], + [-0.0834, 0.1548, -0.2025, ..., 0.0593, 0.0929, -0.0466], + ..., + [-0.0751, -0.1018, -0.0812, ..., 0.0089, -0.2019, 0.1504], + [ 0.0497, -0.0453, 0.1045, ..., -0.0019, -0.2456, -0.0233], + [-0.2219, -0.1146, -0.1486, ..., -0.2642, 0.0688, -0.1098]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.8626e-09, ..., 1.8626e-09, + 2.9802e-07, 0.0000e+00], + [ 0.0000e+00, -1.6764e-08, 3.7253e-09, ..., -1.8626e-08, + -1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-09, 5.5879e-09, ..., 1.1176e-08, + 1.2144e-06, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -1.5441e-06, 0.0000e+00]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0122, -0.0370, 0.0098, -0.0149, 0.0130, 0.0117, 0.0277, 0.0054, + -0.0450, -0.0077], device='cuda:0'), grad: tensor([-7.4506e-09, 8.6799e-07, -1.4901e-08, -3.7253e-08, 0.0000e+00, + 6.1467e-08, 2.0489e-08, 3.4962e-06, 3.7253e-08, -4.4480e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 247.87, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4634 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 +Epoch 325, weight, value: tensor([[ 0.0301, -0.1712, -0.1625, ..., -0.3311, -0.1199, -0.1643], + [ 0.0668, -0.0795, 0.0400, ..., 0.0497, 0.1217, -0.0580], + [-0.0835, 0.1551, -0.2026, ..., 0.0597, 0.0931, -0.0468], + ..., + [-0.0753, -0.1017, -0.0813, ..., 0.0089, -0.2021, 0.1505], + [ 0.0495, -0.0456, 0.1062, ..., -0.0021, -0.2462, -0.0235], + [-0.2219, -0.1150, -0.1502, ..., -0.2641, 0.0690, -0.1098]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 1.8626e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.3039e-08, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 1.3039e-08, ..., -1.1176e-08, + -2.6077e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 1.8626e-08, ..., 1.3039e-08, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.1234e-07, ..., 1.1176e-08, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0122, -0.0369, 0.0100, -0.0154, 0.0131, 0.0119, 0.0273, 0.0053, + -0.0447, -0.0074], device='cuda:0'), grad: tensor([-4.0978e-08, 1.8626e-08, -4.2841e-08, 1.4026e-06, 2.4214e-08, + -1.7621e-06, -5.9791e-07, 4.8429e-08, 8.9407e-07, 5.5879e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 247.76, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4645 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.10 lr 0.00010000 +Epoch 326, weight, value: tensor([[ 0.0301, -0.1713, -0.1628, ..., -0.3337, -0.1213, -0.1644], + [ 0.0668, -0.0793, 0.0400, ..., 0.0498, 0.1222, -0.0577], + [-0.0836, 0.1550, -0.2030, ..., 0.0586, 0.0924, -0.0468], + ..., + [-0.0753, -0.1018, -0.0812, ..., 0.0089, -0.2023, 0.1506], + [ 0.0493, -0.0456, 0.1069, ..., -0.0022, -0.2463, -0.0238], + [-0.2220, -0.1150, -0.1504, ..., -0.2633, 0.0703, -0.1098]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -5.6811e-06, ..., -1.5736e-05, + -1.6168e-05, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 5.2527e-06, ..., 1.4417e-05, + 1.4797e-05, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 4.6939e-07, ..., 1.2703e-06, + 1.3225e-06, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 9.3132e-09, + 1.8626e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.1176e-08, + -5.0291e-08, 0.0000e+00]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0135, -0.0369, 0.0093, -0.0155, 0.0132, 0.0124, 0.0264, 0.0053, + -0.0447, -0.0063], device='cuda:0'), grad: tensor([-4.1202e-06, -2.6137e-05, 2.5034e-05, 1.0245e-06, 2.0862e-07, + 1.5087e-07, 3.1106e-07, 2.1588e-06, 3.9116e-07, 9.8348e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 247.57, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4830 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.10 lr 0.00010000 +Epoch 327, weight, value: tensor([[ 0.0301, -0.1714, -0.1630, ..., -0.3341, -0.1214, -0.1647], + [ 0.0668, -0.0794, 0.0400, ..., 0.0498, 0.1222, -0.0577], + [-0.0839, 0.1550, -0.2032, ..., 0.0587, 0.0925, -0.0469], + ..., + [-0.0753, -0.1018, -0.0813, ..., 0.0089, -0.2024, 0.1506], + [ 0.0493, -0.0456, 0.1079, ..., -0.0018, -0.2464, -0.0240], + [-0.2221, -0.1152, -0.1521, ..., -0.2643, 0.0703, -0.1098]], + device='cuda:0'), grad: tensor([[-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -3.1665e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-08, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, -1.3039e-08, 3.7253e-09, ..., -2.4214e-07, + -2.5518e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-09, -7.4506e-09, ..., 1.5274e-07, + 2.0303e-07, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -3.7253e-09, ..., 3.3528e-08, + 4.6566e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0132, -0.0369, 0.0093, -0.0153, 0.0132, 0.0120, 0.0269, 0.0053, + -0.0445, -0.0064], device='cuda:0'), grad: tensor([-3.1851e-07, 3.2969e-07, -5.9232e-07, 9.4995e-08, -3.5390e-08, + -5.9605e-08, 6.3330e-08, 3.8370e-07, 1.3597e-07, -1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 247.67, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4919 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.01 lr 0.00010000 +Epoch 328, weight, value: tensor([[ 0.0299, -0.1713, -0.1635, ..., -0.3345, -0.1214, -0.1647], + [ 0.0678, -0.0796, 0.0401, ..., 0.0497, 0.1224, -0.0575], + [-0.0846, 0.1548, -0.2032, ..., 0.0585, 0.0925, -0.0469], + ..., + [-0.0750, -0.1014, -0.0813, ..., 0.0089, -0.2024, 0.1506], + [ 0.0490, -0.0456, 0.1082, ..., -0.0016, -0.2468, -0.0241], + [-0.2221, -0.1156, -0.1530, ..., -0.2667, 0.0704, -0.1098]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.5682e-08, 3.7253e-09, ..., 8.9407e-08, + 8.1956e-08, 0.0000e+00], + [ 0.0000e+00, 4.6380e-07, 3.3528e-08, ..., 5.1595e-07, + 3.7253e-07, 0.0000e+00], + [ 0.0000e+00, -2.3190e-06, 1.1176e-08, ..., -2.4345e-06, + -1.8235e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 4.8988e-07, 9.3132e-09, ..., 5.0850e-07, + 4.0606e-07, 0.0000e+00], + [ 0.0000e+00, 1.1679e-06, -7.0781e-08, ..., 1.1530e-06, + 1.1232e-06, 0.0000e+00], + [ 0.0000e+00, 1.3039e-08, 3.7253e-09, ..., 1.8626e-08, + -5.0291e-08, 0.0000e+00]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0130, -0.0369, 0.0091, -0.0151, 0.0130, 0.0115, 0.0264, 0.0055, + -0.0446, -0.0070], device='cuda:0'), grad: tensor([ 2.7753e-07, 1.4435e-06, -6.6496e-06, 1.6205e-07, 2.0489e-07, + 8.0094e-08, -6.3330e-07, 1.3858e-06, 3.8445e-06, -1.0617e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 248.28, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4816 re_mapping 0.0044 re_causal 0.0131 /// teacc 98.96 lr 0.00010000 +Epoch 329, weight, value: tensor([[ 0.0298, -0.1717, -0.1640, ..., -0.3348, -0.1214, -0.1647], + [ 0.0678, -0.0800, 0.0401, ..., 0.0496, 0.1224, -0.0575], + [-0.0846, 0.1559, -0.2033, ..., 0.0590, 0.0928, -0.0471], + ..., + [-0.0750, -0.1019, -0.0815, ..., 0.0086, -0.2025, 0.1506], + [ 0.0486, -0.0465, 0.1087, ..., -0.0024, -0.2478, -0.0242], + [-0.2222, -0.1158, -0.1533, ..., -0.2672, 0.0704, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.4214e-08, 0.0000e+00, ..., 1.8626e-08, + 1.8626e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -3.5390e-08, ..., -1.4901e-08, + -4.0978e-08, 0.0000e+00], + [ 0.0000e+00, -1.1548e-07, 7.4506e-09, ..., -8.7544e-08, + -8.1956e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.0489e-08, 1.4901e-08, ..., 2.4214e-08, + 2.9802e-08, 0.0000e+00], + [ 0.0000e+00, 5.9605e-08, 1.1176e-08, ..., 5.0291e-08, + 5.2154e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0129, -0.0370, 0.0093, -0.0123, 0.0129, 0.0115, 0.0262, 0.0052, + -0.0452, -0.0069], device='cuda:0'), grad: tensor([ 1.4342e-07, -6.5193e-08, -6.5751e-07, -8.5682e-08, -8.0094e-08, + 9.6858e-08, 3.3528e-08, 1.5646e-07, 4.0047e-07, 5.5879e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 248.18, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4756 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.05 lr 0.00010000 +Epoch 330, weight, value: tensor([[ 0.0298, -0.1719, -0.1645, ..., -0.3352, -0.1214, -0.1647], + [ 0.0679, -0.0801, 0.0403, ..., 0.0497, 0.1226, -0.0575], + [-0.0854, 0.1559, -0.2035, ..., 0.0591, 0.0930, -0.0471], + ..., + [-0.0752, -0.1019, -0.0816, ..., 0.0085, -0.2027, 0.1506], + [ 0.0490, -0.0465, 0.1093, ..., -0.0023, -0.2480, -0.0242], + [-0.2222, -0.1163, -0.1533, ..., -0.2697, 0.0707, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.1176e-08, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., -9.3132e-09, + -3.1665e-08, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 9.3132e-09, ..., -4.0978e-08, + -1.0803e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -6.1467e-08, ..., 7.4506e-09, + 8.5682e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0128, -0.0369, 0.0093, -0.0124, 0.0125, 0.0109, 0.0261, 0.0054, + -0.0453, -0.0081], device='cuda:0'), grad: tensor([ 8.1956e-08, -2.9802e-08, -1.3039e-07, 0.0000e+00, -1.4901e-07, + 2.2352e-07, -2.6450e-07, 2.6077e-08, 5.4017e-08, 1.8068e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 247.70, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4785 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.07 lr 0.00010000 +Epoch 331, weight, value: tensor([[ 0.0289, -0.1723, -0.1648, ..., -0.3361, -0.1215, -0.1649], + [ 0.0680, -0.0809, 0.0403, ..., 0.0496, 0.1225, -0.0575], + [-0.0852, 0.1561, -0.2034, ..., 0.0592, 0.0934, -0.0471], + ..., + [-0.0754, -0.1019, -0.0817, ..., 0.0086, -0.2027, 0.1506], + [ 0.0494, -0.0466, 0.1099, ..., -0.0021, -0.2481, -0.0242], + [-0.2226, -0.1167, -0.1533, ..., -0.2703, 0.0707, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.0781e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 2.5369e-06, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 7.4506e-09, ..., 1.1176e-08, + 1.8626e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -3.6657e-06, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 5.5879e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.7055e-08, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0129, -0.0370, 0.0098, -0.0124, 0.0126, 0.0109, 0.0262, 0.0055, + -0.0452, -0.0082], device='cuda:0'), grad: tensor([ 5.4389e-07, 1.8984e-05, 1.6950e-07, 5.5283e-06, 1.3784e-07, + 1.6708e-06, -1.2107e-07, -2.7448e-05, 7.8231e-08, 4.5076e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 247.69, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4553 re_mapping 0.0044 re_causal 0.0124 /// teacc 99.02 lr 0.00010000 +Epoch 332, weight, value: tensor([[ 0.0288, -0.1724, -0.1652, ..., -0.3365, -0.1217, -0.1649], + [ 0.0684, -0.0815, 0.0403, ..., 0.0496, 0.1223, -0.0574], + [-0.0856, 0.1565, -0.2034, ..., 0.0594, 0.0937, -0.0472], + ..., + [-0.0753, -0.1022, -0.0817, ..., 0.0086, -0.2028, 0.1507], + [ 0.0494, -0.0466, 0.1101, ..., -0.0021, -0.2490, -0.0243], + [-0.2227, -0.1176, -0.1538, ..., -0.2714, 0.0708, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 7.6368e-08, ..., 2.9802e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 7.6368e-08, ..., 5.9605e-08, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 3.7625e-07, ..., 1.0058e-07, + -2.7940e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 7.2643e-08, 2.5369e-06, ..., 9.3691e-07, + 3.1665e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 7.0781e-08, ..., 4.0978e-08, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 1.1176e-07, ..., 6.5193e-08, + -1.0431e-07, 0.0000e+00]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0133, -0.0371, 0.0101, -0.0125, 0.0121, 0.0113, 0.0279, 0.0055, + -0.0459, -0.0082], device='cuda:0'), grad: tensor([ 1.8813e-07, 3.0734e-07, 7.6741e-07, -8.1584e-06, 4.2841e-08, + 1.3895e-06, 1.0617e-07, 5.5283e-06, 4.1164e-07, -6.1095e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 247.70, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4906 re_mapping 0.0043 re_causal 0.0135 /// teacc 99.07 lr 0.00010000 +Epoch 333, weight, value: tensor([[ 0.0287, -0.1727, -0.1651, ..., -0.3369, -0.1218, -0.1650], + [ 0.0686, -0.0817, 0.0404, ..., 0.0497, 0.1225, -0.0573], + [-0.0863, 0.1565, -0.2035, ..., 0.0594, 0.0937, -0.0472], + ..., + [-0.0754, -0.1025, -0.0818, ..., 0.0084, -0.2031, 0.1507], + [ 0.0494, -0.0466, 0.1103, ..., -0.0020, -0.2495, -0.0243], + [-0.2229, -0.1176, -0.1542, ..., -0.2718, 0.0709, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.4715e-07, 1.8626e-09, ..., 1.6764e-08, + 1.1176e-08, 0.0000e+00], + [-0.0000e+00, 1.1176e-08, 2.9244e-07, ..., 1.0431e-06, + -2.1793e-07, 0.0000e+00], + [ 7.0781e-08, -0.0000e+00, 9.9093e-07, ..., 1.2629e-06, + 2.4587e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, -4.2096e-07, ..., -1.1344e-06, + 1.1176e-08, 0.0000e+00], + [-7.8231e-08, 7.4506e-09, -1.0822e-06, ..., -1.4864e-06, + -3.6694e-07, 0.0000e+00], + [ 0.0000e+00, 1.0431e-07, 1.8626e-09, ..., 1.4901e-08, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0133, -0.0370, 0.0101, -0.0114, 0.0121, 0.0107, 0.0280, 0.0054, + -0.0462, -0.0082], device='cuda:0'), grad: tensor([-1.5367e-06, 3.3341e-06, 2.5984e-06, 5.1409e-07, 3.9116e-08, + 9.6858e-08, 4.3400e-07, -3.6657e-06, -3.0231e-06, 1.1958e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 247.57, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4700 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.04 lr 0.00010000 +Epoch 334, weight, value: tensor([[ 0.0287, -0.1726, -0.1654, ..., -0.3372, -0.1218, -0.1650], + [ 0.0687, -0.0825, 0.0404, ..., 0.0497, 0.1227, -0.0568], + [-0.0868, 0.1565, -0.2037, ..., 0.0590, 0.0935, -0.0473], + ..., + [-0.0747, -0.1023, -0.0818, ..., 0.0085, -0.2031, 0.1508], + [ 0.0495, -0.0466, 0.1108, ..., -0.0018, -0.2494, -0.0243], + [-0.2236, -0.1180, -0.1547, ..., -0.2728, 0.0709, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, + 9.3132e-09, 0.0000e+00], + [-1.8626e-08, 0.0000e+00, -1.9185e-07, ..., -6.3330e-08, + -2.1048e-07, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 2.4214e-08, ..., 1.6764e-08, + 1.3039e-08, 0.0000e+00], + ..., + [ 1.3039e-08, 0.0000e+00, 1.1921e-07, ..., 2.2352e-08, + 9.8720e-08, -0.0000e+00], + [ 0.0000e+00, -0.0000e+00, -4.6566e-08, ..., -2.4214e-08, + -5.5879e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., 2.0489e-08, + -2.2352e-08, 0.0000e+00]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0133, -0.0370, 0.0098, -0.0113, 0.0124, 0.0108, 0.0275, 0.0055, + -0.0459, -0.0086], device='cuda:0'), grad: tensor([-3.7253e-09, -4.3027e-07, 8.1956e-08, -5.2154e-07, 1.0617e-07, + 4.8615e-07, 2.2352e-07, 0.0000e+00, -1.3411e-07, 1.6950e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 247.68, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4812 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.08 lr 0.00010000 +Epoch 335, weight, value: tensor([[ 0.0285, -0.1736, -0.1660, ..., -0.3377, -0.1218, -0.1650], + [ 0.0687, -0.0835, 0.0404, ..., 0.0496, 0.1226, -0.0565], + [-0.0871, 0.1572, -0.2039, ..., 0.0589, 0.0936, -0.0474], + ..., + [-0.0750, -0.1025, -0.0818, ..., 0.0086, -0.2032, 0.1509], + [ 0.0497, -0.0466, 0.1123, ..., -0.0017, -0.2494, -0.0243], + [-0.2250, -0.1178, -0.1556, ..., -0.2735, 0.0707, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 6.3330e-08, ..., 1.3039e-08, + -1.3039e-07, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, 1.4901e-07, ..., 2.4214e-08, + 5.0291e-08, 0.0000e+00], + [ 0.0000e+00, -6.6869e-07, 3.1665e-08, ..., -7.5437e-07, + -1.4026e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7998e-07, 1.9558e-07, ..., 4.2841e-07, + 8.0094e-07, 0.0000e+00], + [ 0.0000e+00, 1.5832e-07, -7.5623e-07, ..., 1.7509e-07, + 3.3900e-07, 0.0000e+00], + [ 0.0000e+00, 8.3819e-08, 2.1793e-07, ..., 8.5682e-08, + 1.6205e-07, 0.0000e+00]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0129, -0.0371, 0.0099, -0.0109, 0.0130, 0.0094, 0.0267, 0.0056, + -0.0453, -0.0090], device='cuda:0'), grad: tensor([-7.5065e-07, 8.9966e-07, -3.6377e-06, 2.3842e-07, 5.4017e-08, + 1.9185e-07, 8.5495e-07, 3.0808e-06, -2.6468e-06, 1.7136e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 248.19, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4751 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.09 lr 0.00010000 +Epoch 336, weight, value: tensor([[ 0.0285, -0.1735, -0.1657, ..., -0.3378, -0.1217, -0.1650], + [ 0.0688, -0.0837, 0.0402, ..., 0.0495, 0.1216, -0.0561], + [-0.0874, 0.1573, -0.2030, ..., 0.0597, 0.0957, -0.0475], + ..., + [-0.0751, -0.1026, -0.0817, ..., 0.0086, -0.2032, 0.1509], + [ 0.0499, -0.0468, 0.1127, ..., -0.0018, -0.2498, -0.0243], + [-0.2261, -0.1178, -0.1560, ..., -0.2738, 0.0712, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 6.5193e-08, ..., 2.0303e-07, + -1.3039e-08, 0.0000e+00], + [ 3.7253e-09, -9.3132e-09, 1.1176e-08, ..., -1.1176e-08, + -1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -7.4506e-08, ..., -2.2352e-07, + 1.4901e-08, 0.0000e+00], + [-3.7253e-09, 5.5879e-09, -1.8626e-09, ..., 1.1176e-08, + 2.2352e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.1176e-08, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0124, -0.0376, 0.0118, -0.0109, 0.0123, 0.0095, 0.0263, 0.0057, + -0.0454, -0.0088], device='cuda:0'), grad: tensor([-3.9116e-08, 5.1595e-07, 7.4506e-09, 2.9802e-08, 8.3819e-08, + 2.4214e-08, -2.0675e-07, -5.3830e-07, 8.0094e-08, 5.2154e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 247.49, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4872 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.01 lr 0.00010000 +Epoch 337, weight, value: tensor([[ 0.0283, -0.1739, -0.1665, ..., -0.3381, -0.1217, -0.1651], + [ 0.0690, -0.0839, 0.0403, ..., 0.0495, 0.1218, -0.0561], + [-0.0879, 0.1572, -0.2034, ..., 0.0594, 0.0956, -0.0476], + ..., + [-0.0756, -0.1026, -0.0818, ..., 0.0086, -0.2034, 0.1509], + [ 0.0499, -0.0468, 0.1133, ..., -0.0017, -0.2501, -0.0244], + [-0.2271, -0.1183, -0.1563, ..., -0.2740, 0.0717, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.0781e-08, ..., 3.7253e-09, + 2.3469e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.4890e-07, ..., 1.3597e-07, + -1.4994e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0990e-07, ..., 1.9930e-07, + 2.4773e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., -3.7625e-07, + -5.4017e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 7.6368e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.7253e-09, + -4.2841e-08, 0.0000e+00]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0123, -0.0375, 0.0115, -0.0109, 0.0117, 0.0095, 0.0260, 0.0057, + -0.0452, -0.0085], device='cuda:0'), grad: tensor([ 4.4890e-07, -2.7865e-06, 9.7416e-07, 2.4214e-08, 1.0431e-07, + 4.4703e-08, 1.9632e-06, -8.9034e-07, 2.3469e-07, -1.3784e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 247.78, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4638 re_mapping 0.0042 re_causal 0.0123 /// teacc 98.98 lr 0.00010000 +Epoch 338, weight, value: tensor([[ 0.0282, -0.1741, -0.1666, ..., -0.3389, -0.1218, -0.1651], + [ 0.0689, -0.0841, 0.0399, ..., 0.0493, 0.1213, -0.0561], + [-0.0879, 0.1573, -0.2036, ..., 0.0595, 0.0957, -0.0471], + ..., + [-0.0755, -0.1026, -0.0818, ..., 0.0086, -0.2035, 0.1509], + [ 0.0498, -0.0469, 0.1166, ..., 0.0012, -0.2472, -0.0244], + [-0.2274, -0.1186, -0.1569, ..., -0.2744, 0.0715, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, -1.0841e-05, ..., 3.7253e-09, + 2.2724e-07, 3.1665e-08], + [ 0.0000e+00, 1.3039e-08, 1.8626e-08, ..., 1.6764e-08, + 2.2352e-08, 7.4506e-09], + [ 0.0000e+00, -1.4417e-06, 2.6077e-08, ..., -2.2687e-06, + -2.5071e-06, 9.3132e-09], + ..., + [ 0.0000e+00, 2.7567e-07, 1.4901e-08, ..., 5.1782e-07, + 4.4331e-07, 0.0000e+00], + [ 0.0000e+00, 1.1418e-06, 1.8664e-06, ..., 1.7192e-06, + 2.1867e-06, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 8.0392e-06, ..., 1.8626e-09, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0119, -0.0379, 0.0115, -0.0114, 0.0122, 0.0094, 0.0254, 0.0057, + -0.0423, -0.0088], device='cuda:0'), grad: tensor([-5.8383e-05, 2.7381e-07, -4.8615e-06, 4.8615e-07, 1.2107e-07, + 6.1095e-06, -3.1628e-06, 1.1288e-06, 1.4454e-05, 4.3780e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 247.75, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.4852 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.02 lr 0.00010000 +Epoch 339, weight, value: tensor([[ 0.0288, -0.1755, -0.1649, ..., -0.3397, -0.1218, -0.1652], + [ 0.0696, -0.0852, 0.0383, ..., 0.0478, 0.1205, -0.0562], + [-0.0898, 0.1575, -0.2040, ..., 0.0595, 0.0956, -0.0471], + ..., + [-0.0760, -0.1028, -0.0802, ..., 0.0101, -0.2029, 0.1511], + [ 0.0494, -0.0469, 0.1166, ..., 0.0012, -0.2474, -0.0246], + [-0.2279, -0.1178, -0.1591, ..., -0.2750, 0.0720, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 3.7253e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, -7.4506e-09, ..., -1.3039e-08, + -2.6077e-08, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 1.4156e-07, ..., 5.4017e-08, + 1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -2.8126e-07, 2.9802e-08, ..., -0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-1.8626e-09, 5.5879e-09, -8.3260e-07, ..., -2.7753e-07, + -7.4506e-08, 0.0000e+00], + [ 0.0000e+00, 2.4401e-07, 1.6391e-07, ..., 5.7742e-08, + -2.2352e-08, 0.0000e+00]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0109, -0.0395, 0.0113, -0.0112, 0.0136, 0.0093, 0.0249, 0.0072, + -0.0424, -0.0093], device='cuda:0'), grad: tensor([ 4.2841e-08, 8.9407e-08, 4.5821e-07, 3.4459e-07, 4.8243e-07, + 8.5682e-07, 5.9605e-08, -5.7966e-06, -2.0396e-06, 5.4799e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 247.59, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4913 re_mapping 0.0045 re_causal 0.0138 /// teacc 99.06 lr 0.00010000 +Epoch 340, weight, value: tensor([[ 0.0287, -0.1782, -0.1650, ..., -0.3403, -0.1219, -0.1667], + [ 0.0699, -0.0862, 0.0372, ..., 0.0468, 0.1203, -0.0556], + [-0.0903, 0.1573, -0.2042, ..., 0.0591, 0.0956, -0.0473], + ..., + [-0.0763, -0.1021, -0.0790, ..., 0.0112, -0.2027, 0.1511], + [ 0.0498, -0.0469, 0.1167, ..., 0.0012, -0.2476, -0.0250], + [-0.2281, -0.1174, -0.1601, ..., -0.2761, 0.0727, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -5.5879e-09, ..., 3.7253e-09, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 6.4075e-07, 0.0000e+00, ..., 1.4231e-06, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -6.5193e-07, 0.0000e+00, ..., -1.4491e-06, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.2352e-08, ..., 1.3039e-08, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0111, -0.0406, 0.0109, -0.0127, 0.0133, 0.0113, 0.0252, 0.0083, + -0.0425, -0.0092], device='cuda:0'), grad: tensor([ 2.4214e-08, 1.4901e-08, 5.6624e-06, 1.5218e-06, 3.9116e-08, + -1.7174e-06, 1.3039e-08, -5.7817e-06, 1.7881e-07, 4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 247.49, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4607 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.12 lr 0.00010000 +Epoch 341, weight, value: tensor([[ 0.0275, -0.1787, -0.1656, ..., -0.3419, -0.1220, -0.1667], + [ 0.0699, -0.0863, 0.0370, ..., 0.0467, 0.1203, -0.0555], + [-0.0921, 0.1575, -0.2045, ..., 0.0592, 0.0958, -0.0473], + ..., + [-0.0766, -0.1022, -0.0789, ..., 0.0113, -0.2028, 0.1512], + [ 0.0519, -0.0469, 0.1171, ..., 0.0016, -0.2476, -0.0257], + [-0.2287, -0.1176, -0.1607, ..., -0.2764, 0.0730, -0.1099]], + device='cuda:0'), grad: tensor([[ 4.0047e-07, 9.3132e-10, 2.7940e-09, ..., 7.6815e-06, + 9.0972e-06, 0.0000e+00], + [ 5.5879e-09, 9.3132e-10, -6.5193e-09, ..., 1.1921e-07, + 1.2014e-07, 0.0000e+00], + [-5.7463e-07, -1.3039e-08, 1.6764e-08, ..., -1.1042e-05, + -1.3083e-05, 0.0000e+00], + ..., + [ 1.8626e-08, 1.0245e-08, 1.1176e-08, ..., 3.5390e-07, + 5.0385e-07, 0.0000e+00], + [ 1.2387e-07, 0.0000e+00, -5.8673e-08, ..., 2.3749e-06, + 2.8145e-06, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 4.0978e-08, ..., 7.2643e-08, + 8.0094e-08, 0.0000e+00]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0112, -0.0407, 0.0108, -0.0134, 0.0123, 0.0118, 0.0255, 0.0085, + -0.0422, -0.0089], device='cuda:0'), grad: tensor([ 2.4214e-05, 3.9767e-07, -3.4779e-05, 5.5507e-07, -6.5006e-07, + 1.1455e-07, 5.9512e-07, 1.2182e-06, 7.5623e-06, 7.7114e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 247.54, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4804 re_mapping 0.0042 re_causal 0.0124 /// teacc 98.99 lr 0.00010000 +Epoch 342, weight, value: tensor([[ 0.0272, -0.1797, -0.1650, ..., -0.3447, -0.1222, -0.1672], + [ 0.0701, -0.0871, 0.0370, ..., 0.0466, 0.1203, -0.0555], + [-0.0923, 0.1576, -0.2049, ..., 0.0592, 0.0960, -0.0475], + ..., + [-0.0778, -0.1024, -0.0789, ..., 0.0113, -0.2029, 0.1512], + [ 0.0520, -0.0468, 0.1173, ..., 0.0019, -0.2477, -0.0256], + [-0.2294, -0.1176, -0.1622, ..., -0.2771, 0.0712, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -4.6566e-09, + -2.8871e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., -7.9162e-08, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 2.7940e-09, + -8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0110, -0.0408, 0.0108, -0.0133, 0.0147, 0.0119, 0.0259, 0.0086, + -0.0422, -0.0106], device='cuda:0'), grad: tensor([ 3.8184e-08, 9.1270e-08, 2.1420e-08, -2.5146e-08, 5.9605e-08, + 9.0338e-08, 6.0536e-08, -2.8964e-07, 8.9407e-08, -1.4249e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 247.67, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4559 re_mapping 0.0044 re_causal 0.0120 /// teacc 98.98 lr 0.00010000 +Epoch 343, weight, value: tensor([[ 0.0273, -0.1804, -0.1650, ..., -0.3453, -0.1222, -0.1672], + [ 0.0710, -0.0882, 0.0371, ..., 0.0466, 0.1206, -0.0557], + [-0.0928, 0.1559, -0.2053, ..., 0.0576, 0.0957, -0.0474], + ..., + [-0.0778, -0.1000, -0.0789, ..., 0.0116, -0.2030, 0.1515], + [ 0.0522, -0.0471, 0.1173, ..., 0.0017, -0.2478, -0.0257], + [-0.2295, -0.1178, -0.1627, ..., -0.2784, 0.0717, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 3.7253e-09, + 5.9605e-08, 0.0000e+00], + [ 1.8626e-09, 7.8231e-08, -2.0675e-07, ..., 3.9116e-08, + -2.7008e-08, 0.0000e+00], + [ 0.0000e+00, -9.9652e-08, 4.9360e-08, ..., -5.5879e-08, + -4.9360e-08, 0.0000e+00], + ..., + [-3.7253e-09, 1.6764e-08, 1.5926e-07, ..., -6.6124e-08, + 1.7323e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 5.5879e-09, ..., 1.0245e-08, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -0.0000e+00, + -4.9360e-07, 0.0000e+00]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0109, -0.0408, 0.0093, -0.0133, 0.0139, 0.0114, 0.0257, 0.0089, + -0.0423, -0.0105], device='cuda:0'), grad: tensor([ 1.7975e-07, 2.3283e-08, 1.3504e-07, 2.8219e-07, 1.3132e-07, + 7.1339e-07, 2.3283e-08, -1.0338e-07, 8.6613e-08, -1.4836e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 247.61, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4983 re_mapping 0.0042 re_causal 0.0127 /// teacc 98.98 lr 0.00010000 +Epoch 344, weight, value: tensor([[ 0.0267, -0.1816, -0.1667, ..., -0.3469, -0.1224, -0.1672], + [ 0.0716, -0.0902, 0.0372, ..., 0.0467, 0.1210, -0.0551], + [-0.0929, 0.1550, -0.2057, ..., 0.0566, 0.0952, -0.0475], + ..., + [-0.0783, -0.0986, -0.0789, ..., 0.0122, -0.2028, 0.1521], + [ 0.0517, -0.0475, 0.1170, ..., 0.0013, -0.2482, -0.0258], + [-0.2296, -0.1179, -0.1630, ..., -0.2792, 0.0736, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 1.8626e-09, ..., 7.4506e-09, + 1.0245e-08, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, -4.5635e-07, ..., -3.0175e-07, + -3.2969e-07, 0.0000e+00], + [ 0.0000e+00, -5.4017e-08, 1.2107e-08, ..., -1.3411e-07, + -1.6019e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 3.0734e-08, 4.2748e-07, ..., 3.8836e-07, + 4.1910e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -5.0291e-08, ..., -4.6566e-09, + 2.9802e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 1.1176e-08, + -9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0121, -0.0406, 0.0082, -0.0135, 0.0117, 0.0122, 0.0258, 0.0096, + -0.0429, -0.0091], device='cuda:0'), grad: tensor([ 2.9802e-08, -8.4657e-07, -4.2468e-07, 3.6322e-08, 4.7497e-08, + 1.0524e-07, 4.0047e-08, 1.1185e-06, 6.5193e-09, -9.9652e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 247.60, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4735 re_mapping 0.0043 re_causal 0.0126 /// teacc 98.97 lr 0.00010000 +Epoch 345, weight, value: tensor([[ 0.0270, -0.1826, -0.1668, ..., -0.3477, -0.1225, -0.1673], + [ 0.0719, -0.0913, 0.0381, ..., 0.0478, 0.1230, -0.0552], + [-0.0930, 0.1529, -0.2058, ..., 0.0545, 0.0939, -0.0476], + ..., + [-0.0783, -0.0964, -0.0798, ..., 0.0127, -0.2036, 0.1529], + [ 0.0518, -0.0477, 0.1170, ..., 0.0012, -0.2484, -0.0258], + [-0.2298, -0.1180, -0.1636, ..., -0.2794, 0.0736, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., 4.6566e-09, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -1.5832e-08, 1.8626e-09, ..., 4.3772e-08, + -1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.5832e-08, 4.6566e-09, ..., -4.3772e-08, + 2.3283e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.2107e-08, 0.0000e+00]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0110, -0.0398, 0.0062, -0.0134, 0.0103, 0.0124, 0.0253, 0.0106, + -0.0431, -0.0101], device='cuda:0'), grad: tensor([-1.2480e-07, 2.5146e-08, 1.3690e-07, -6.6124e-08, -8.3819e-08, + 6.0536e-08, 1.6764e-08, -1.3318e-07, 4.6566e-09, 1.5926e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 247.18, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4851 re_mapping 0.0043 re_causal 0.0130 /// teacc 99.02 lr 0.00010000 +Epoch 346, weight, value: tensor([[ 0.0273, -0.1828, -0.1671, ..., -0.3479, -0.1221, -0.1673], + [ 0.0720, -0.0910, 0.0383, ..., 0.0480, 0.1233, -0.0554], + [-0.0943, 0.1530, -0.2063, ..., 0.0545, 0.0939, -0.0476], + ..., + [-0.0794, -0.0965, -0.0799, ..., 0.0126, -0.2040, 0.1530], + [ 0.0514, -0.0479, 0.1172, ..., 0.0010, -0.2485, -0.0261], + [-0.2305, -0.1180, -0.1650, ..., -0.2805, 0.0734, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.2200e-07, ..., 1.3039e-08, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 4.1537e-07, ..., 4.5635e-08, + -9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., -7.4506e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8841e-06, ..., -1.9744e-07, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0245e-07, ..., 1.2107e-08, + 1.0692e-06, 0.0000e+00]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0093, -0.0397, 0.0062, -0.0132, 0.0104, 0.0119, 0.0258, 0.0105, + -0.0430, -0.0114], device='cuda:0'), grad: tensor([ 3.9581e-07, 3.4459e-08, 1.4314e-06, 3.3639e-06, -2.0918e-06, + 5.8115e-07, 3.2410e-07, 9.3132e-09, -6.5416e-06, 2.4941e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 247.94, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4803 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.01 lr 0.00010000 +Epoch 347, weight, value: tensor([[ 0.0265, -0.1829, -0.1682, ..., -0.3503, -0.1223, -0.1673], + [ 0.0721, -0.0909, 0.0383, ..., 0.0480, 0.1233, -0.0555], + [-0.0937, 0.1530, -0.2065, ..., 0.0546, 0.0941, -0.0476], + ..., + [-0.0795, -0.0965, -0.0799, ..., 0.0126, -0.2041, 0.1531], + [ 0.0511, -0.0480, 0.1173, ..., 0.0008, -0.2489, -0.0262], + [-0.2311, -0.1181, -0.1686, ..., -0.2817, 0.0740, -0.1100]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 3.7253e-09, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 1.5832e-08, ..., 5.1223e-08, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -2.0489e-08, 0.0000e+00, ..., -6.1467e-08, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.1176e-08, -4.0978e-08, ..., -2.7008e-08, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.0489e-08, ..., 2.3283e-08, + 2.2352e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 3.7253e-09, + -6.3330e-08, 0.0000e+00]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0095, -0.0397, 0.0062, -0.0124, 0.0097, 0.0128, 0.0251, 0.0105, + -0.0433, -0.0116], device='cuda:0'), grad: tensor([-4.6473e-07, 1.7975e-07, 1.7136e-07, 4.8429e-08, 3.7253e-08, + 1.0338e-07, -7.2643e-08, -3.6322e-08, 1.4622e-07, -8.3819e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 247.55, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4830 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.06 lr 0.00010000 +Epoch 348, weight, value: tensor([[ 0.0265, -0.1830, -0.1689, ..., -0.3521, -0.1225, -0.1673], + [ 0.0722, -0.0910, 0.0381, ..., 0.0478, 0.1235, -0.0555], + [-0.0938, 0.1530, -0.2068, ..., 0.0546, 0.0941, -0.0475], + ..., + [-0.0795, -0.0965, -0.0797, ..., 0.0127, -0.2043, 0.1533], + [ 0.0512, -0.0481, 0.1178, ..., 0.0012, -0.2493, -0.0263], + [-0.2315, -0.1181, -0.1689, ..., -0.2831, 0.0748, -0.1100]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.9360e-08, ..., 1.0617e-07, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 2.7940e-09, ..., 1.8626e-09, + -4.6566e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.4214e-08, ..., -9.5926e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.6950e-07, ..., 9.5926e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0096, -0.0399, 0.0062, -0.0123, 0.0092, 0.0128, 0.0251, 0.0106, + -0.0433, -0.0110], device='cuda:0'), grad: tensor([ 7.4506e-09, 3.1479e-07, 3.7253e-09, -7.6089e-07, 1.8626e-09, + 3.6228e-07, 4.6566e-09, -2.7660e-07, 4.6566e-09, 3.4925e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 247.63, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4785 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 +Epoch 349, weight, value: tensor([[ 0.0264, -0.1830, -0.1689, ..., -0.3527, -0.1224, -0.1674], + [ 0.0748, -0.0925, 0.0382, ..., 0.0477, 0.1234, -0.0555], + [-0.0939, 0.1530, -0.2070, ..., 0.0547, 0.0946, -0.0476], + ..., + [-0.0796, -0.0965, -0.0798, ..., 0.0127, -0.2044, 0.1545], + [ 0.0511, -0.0481, 0.1181, ..., 0.0015, -0.2494, -0.0262], + [-0.2316, -0.1184, -0.1691, ..., -0.2844, 0.0749, -0.1100]], + device='cuda:0'), grad: tensor([[-5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 4.6566e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, -1.3039e-08, ..., 2.1420e-08, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, -1.3970e-08, 9.3132e-10, ..., -7.6368e-08, + -6.7055e-08, 0.0000e+00], + ..., + [ 1.8626e-09, -2.7940e-09, 2.7940e-09, ..., -6.0536e-08, + 1.4901e-08, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 9.3132e-10, ..., 1.1176e-08, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 7.4506e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0076, -0.0400, 0.0063, -0.0126, 0.0089, 0.0131, 0.0220, 0.0107, + -0.0431, -0.0112], device='cuda:0'), grad: tensor([-1.1176e-07, 4.0047e-08, -1.5087e-07, 1.2107e-07, 8.8476e-08, + -7.4506e-09, 2.7008e-08, -1.7323e-07, 2.9802e-08, 1.3225e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 247.74, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4736 re_mapping 0.0043 re_causal 0.0127 /// teacc 99.07 lr 0.00010000 +Epoch 350, weight, value: tensor([[ 0.0262, -0.1831, -0.1695, ..., -0.3545, -0.1225, -0.1674], + [ 0.0771, -0.0929, 0.0384, ..., 0.0482, 0.1249, -0.0554], + [-0.0962, 0.1530, -0.2073, ..., 0.0544, 0.0935, -0.0475], + ..., + [-0.0799, -0.0965, -0.0799, ..., 0.0126, -0.2053, 0.1546], + [ 0.0509, -0.0481, 0.1181, ..., 0.0012, -0.2497, -0.0264], + [-0.2334, -0.1184, -0.1694, ..., -0.2852, 0.0757, -0.1100]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.6764e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 4.4797e-07, + 1.5702e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.6764e-08, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., -9.6299e-07, + 1.1362e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.7043e-07, ..., -2.8871e-08, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 8.4750e-08, + 3.9022e-07, 0.0000e+00]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0074, -0.0396, 0.0062, -0.0126, 0.0088, 0.0137, 0.0207, 0.0105, + -0.0436, -0.0105], device='cuda:0'), grad: tensor([ 7.1712e-08, 5.4650e-06, 8.6613e-08, 1.2452e-06, -5.2527e-06, + 2.6450e-07, 4.9360e-08, -2.8610e-06, -3.9861e-07, 1.3364e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 247.37, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.5140 re_mapping 0.0042 re_causal 0.0137 /// teacc 99.02 lr 0.00010000 +Epoch 351, weight, value: tensor([[ 0.0265, -0.1834, -0.1697, ..., -0.3553, -0.1227, -0.1674], + [ 0.0781, -0.0933, 0.0385, ..., 0.0482, 0.1251, -0.0554], + [-0.0965, 0.1531, -0.2075, ..., 0.0544, 0.0936, -0.0475], + ..., + [-0.0792, -0.0965, -0.0800, ..., 0.0126, -0.2055, 0.1546], + [ 0.0508, -0.0482, 0.1197, ..., 0.0010, -0.2500, -0.0264], + [-0.2340, -0.1185, -0.1695, ..., -0.2854, 0.0761, -0.1100]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 1.8626e-09, ..., 8.3819e-09, + 3.7253e-08, 0.0000e+00], + [ 0.0000e+00, 2.1420e-08, 2.9802e-08, ..., 3.4459e-08, + 9.5926e-08, 0.0000e+00], + [ 0.0000e+00, -1.5926e-07, 1.3970e-08, ..., -1.5832e-07, + -2.9057e-07, -2.7940e-09], + ..., + [-0.0000e+00, 2.7008e-08, 8.3819e-09, ..., 2.5146e-08, + 4.9360e-08, 0.0000e+00], + [ 0.0000e+00, 7.0781e-08, -2.3562e-07, ..., -4.7497e-08, + 1.5739e-07, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 8.6613e-08, ..., 4.0978e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0079, -0.0395, 0.0062, -0.0129, 0.0087, 0.0127, 0.0215, 0.0104, + -0.0427, -0.0102], device='cuda:0'), grad: tensor([ 3.1665e-08, 2.8405e-07, -6.6124e-07, 2.3749e-07, 2.4214e-08, + 5.2340e-07, -6.1281e-07, 1.1548e-07, -1.8533e-07, 2.5146e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 247.48, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4610 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.01 lr 0.00010000 +Epoch 352, weight, value: tensor([[ 0.0262, -0.1839, -0.1704, ..., -0.3556, -0.1228, -0.1674], + [ 0.0785, -0.0920, 0.0385, ..., 0.0482, 0.1254, -0.0554], + [-0.0966, 0.1530, -0.2080, ..., 0.0544, 0.0936, -0.0474], + ..., + [-0.0792, -0.0965, -0.0801, ..., 0.0126, -0.2057, 0.1547], + [ 0.0499, -0.0482, 0.1195, ..., 0.0009, -0.2504, -0.0264], + [-0.2347, -0.1187, -0.1697, ..., -0.2857, 0.0761, -0.1100]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.2107e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 2.7940e-09, 4.8429e-08, 4.8429e-08, ..., 9.4995e-08, + 4.1910e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, -5.1223e-08, -5.4017e-08, ..., -1.0338e-07, + 6.8918e-08, 0.0000e+00], + [ 2.1420e-08, 1.8626e-09, 9.3132e-09, ..., 2.7940e-09, + 1.1176e-08, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, + -1.1269e-07, 0.0000e+00]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0080, -0.0395, 0.0061, -0.0128, 0.0088, 0.0130, 0.0215, 0.0104, + -0.0431, -0.0103], device='cuda:0'), grad: tensor([-4.3772e-08, 5.1409e-07, 1.5832e-08, 6.4634e-07, -2.7940e-09, + -1.0561e-06, 1.5832e-08, -1.2387e-07, 3.3714e-07, -2.9709e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 247.80, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4736 re_mapping 0.0043 re_causal 0.0127 /// teacc 99.03 lr 0.00010000 +Epoch 353, weight, value: tensor([[ 0.0258, -0.1842, -0.1707, ..., -0.3559, -0.1228, -0.1674], + [ 0.0785, -0.0920, 0.0386, ..., 0.0483, 0.1257, -0.0554], + [-0.0991, 0.1531, -0.2087, ..., 0.0544, 0.0937, -0.0474], + ..., + [-0.0819, -0.0966, -0.0803, ..., 0.0124, -0.2060, 0.1547], + [ 0.0530, -0.0485, 0.1208, ..., 0.0029, -0.2507, -0.0264], + [-0.2345, -0.1189, -0.1698, ..., -0.2861, 0.0761, -0.1100]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -7.4506e-09, ..., 5.5879e-09, + -2.3283e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -1.8626e-09, + 9.3132e-09, 0.0000e+00], + [-0.0000e+00, 0.0000e+00, -8.1025e-08, ..., -6.9849e-08, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + -1.0245e-08, 0.0000e+00]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0077, -0.0394, 0.0061, -0.0127, 0.0087, 0.0129, 0.0216, 0.0103, + -0.0419, -0.0104], device='cuda:0'), grad: tensor([ 0.0000e+00, -2.1420e-08, 2.5146e-08, 9.4064e-08, 8.7544e-08, + 5.8673e-08, 1.8626e-08, 1.3970e-08, -1.8347e-07, -9.1270e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 247.33, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4683 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.13 lr 0.00010000 +Epoch 354, weight, value: tensor([[ 0.0252, -0.1843, -0.1714, ..., -0.3561, -0.1230, -0.1674], + [ 0.0790, -0.0928, 0.0387, ..., 0.0483, 0.1257, -0.0554], + [-0.0992, 0.1531, -0.2088, ..., 0.0544, 0.0938, -0.0474], + ..., + [-0.0819, -0.0966, -0.0803, ..., 0.0124, -0.2063, 0.1548], + [ 0.0530, -0.0485, 0.1207, ..., 0.0029, -0.2509, -0.0264], + [-0.2353, -0.1163, -0.1698, ..., -0.2868, 0.0768, -0.1100]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 1.8626e-09, ..., 6.5193e-09, + 1.2107e-08, 0.0000e+00], + [-1.4901e-08, 7.7300e-08, -7.1712e-08, ..., 2.0489e-07, + 4.1910e-08, 0.0000e+00], + [ 3.6322e-08, -2.9150e-07, 2.2445e-07, ..., -4.5635e-07, + -8.8010e-07, 0.0000e+00], + ..., + [-7.5344e-07, 1.3411e-07, 1.3970e-07, ..., 5.0664e-07, + 5.8208e-07, 0.0000e+00], + [-5.3085e-08, 4.1910e-08, -2.3469e-07, ..., -3.1479e-07, + 1.3411e-07, 0.0000e+00], + [ 3.7253e-09, 3.4459e-08, 3.7253e-09, ..., 9.4064e-08, + -1.7695e-08, 0.0000e+00]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0079, -0.0395, 0.0061, -0.0128, 0.0087, 0.0127, 0.0222, 0.0103, + -0.0422, -0.0098], device='cuda:0'), grad: tensor([ 1.7695e-08, 4.7870e-07, -1.5935e-06, -1.1083e-07, 3.4980e-06, + 4.0513e-07, 1.1176e-08, -1.8356e-06, -5.0291e-07, -3.5856e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 248.13, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4637 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.09 lr 0.00010000 +Epoch 355, weight, value: tensor([[ 0.0248, -0.1847, -0.1719, ..., -0.3566, -0.1231, -0.1674], + [ 0.0794, -0.0928, 0.0388, ..., 0.0484, 0.1260, -0.0554], + [-0.0998, 0.1531, -0.2092, ..., 0.0544, 0.0938, -0.0474], + ..., + [-0.0814, -0.0966, -0.0804, ..., 0.0123, -0.2066, 0.1548], + [ 0.0529, -0.0486, 0.1207, ..., 0.0028, -0.2511, -0.0265], + [-0.2360, -0.1160, -0.1699, ..., -0.2870, 0.0769, -0.1100]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 3.7253e-09, 1.8626e-09, ..., 1.2107e-08, + 1.0245e-08, 0.0000e+00], + [ 1.4994e-07, 6.3330e-08, 2.0489e-08, ..., 2.0768e-07, + 1.6857e-07, 0.0000e+00], + [ 9.0078e-06, -1.1735e-07, 1.1781e-06, ..., 9.3803e-06, + -3.2689e-07, 0.0000e+00], + ..., + [-9.4026e-06, 1.2107e-08, -1.1725e-06, ..., -9.8795e-06, + 4.0047e-08, 0.0000e+00], + [ 1.9092e-07, 2.9802e-08, 2.8871e-08, ..., 2.4121e-07, + 8.8476e-08, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 9.3132e-09, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0077, -0.0394, 0.0061, -0.0129, 0.0088, 0.0128, 0.0220, 0.0102, + -0.0423, -0.0098], device='cuda:0'), grad: tensor([ 4.1910e-08, 6.8825e-07, 2.6852e-05, -5.0291e-08, 2.7008e-08, + 1.3970e-07, -1.1642e-07, -2.8387e-05, 7.5251e-07, 3.0734e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 247.42, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4957 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.11 lr 0.00010000 +Epoch 356, weight, value: tensor([[ 0.0248, -0.1853, -0.1728, ..., -0.3568, -0.1235, -0.1674], + [ 0.0794, -0.0940, 0.0389, ..., 0.0483, 0.1265, -0.0555], + [-0.1022, 0.1532, -0.2092, ..., 0.0544, 0.0941, -0.0474], + ..., + [-0.0793, -0.0967, -0.0804, ..., 0.0123, -0.2069, 0.1550], + [ 0.0528, -0.0487, 0.1207, ..., 0.0028, -0.2513, -0.0265], + [-0.2350, -0.1162, -0.1700, ..., -0.2873, 0.0767, -0.1100]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, 4.4703e-08, ..., 9.3132e-10, + 6.3330e-08, 0.0000e+00], + [-4.0978e-08, 0.0000e+00, -2.3004e-07, ..., 3.7253e-09, + -2.2352e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 2.2352e-08, ..., 2.3283e-08, + -3.0734e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 6.6124e-08, + 1.0245e-08, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 3.0734e-08, ..., 5.5879e-09, + 2.7008e-08, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 2.4214e-08, 0.0000e+00]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0079, -0.0392, 0.0061, -0.0127, 0.0086, 0.0126, 0.0220, 0.0102, + -0.0424, -0.0101], device='cuda:0'), grad: tensor([ 2.0303e-07, -6.4448e-07, -3.4459e-08, -2.3935e-07, -3.7253e-08, + 5.2154e-08, 3.9022e-07, -1.4901e-08, 1.0524e-07, 2.1141e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 355---------------------------------------------------- +epoch 355, time 263.92, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4485 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.16 lr 0.00010000 +Epoch 357, weight, value: tensor([[ 0.0257, -0.1855, -0.1752, ..., -0.3571, -0.1236, -0.1676], + [ 0.0795, -0.0932, 0.0391, ..., 0.0485, 0.1270, -0.0569], + [-0.1025, 0.1532, -0.2096, ..., 0.0544, 0.0939, -0.0468], + ..., + [-0.0791, -0.0967, -0.0806, ..., 0.0123, -0.2073, 0.1561], + [ 0.0527, -0.0487, 0.1207, ..., 0.0028, -0.2514, -0.0272], + [-0.2356, -0.1162, -0.1703, ..., -0.2877, 0.0770, -0.1101]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 1.4901e-08, + 6.3702e-07, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, -1.8626e-09, ..., 3.7253e-09, + 1.7695e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 8.3819e-09, + 1.1660e-06, 0.0000e+00], + ..., + [ 0.0000e+00, -2.7940e-09, 4.6566e-09, ..., -3.2596e-08, + 4.9174e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.2841e-08, + 3.5390e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -2.3469e-07, ..., -1.0617e-07, + -9.2909e-06, 0.0000e+00]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0080, -0.0390, 0.0061, -0.0123, 0.0081, 0.0127, 0.0220, 0.0101, + -0.0426, -0.0098], device='cuda:0'), grad: tensor([ 2.6990e-06, 1.4063e-07, 3.8743e-06, 2.8908e-06, 8.2180e-06, + 3.0641e-07, 4.8429e-08, 1.8016e-05, 1.7071e-06, -3.7879e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 247.80, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4717 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.15 lr 0.00010000 +Epoch 358, weight, value: tensor([[ 0.0266, -0.1860, -0.1756, ..., -0.3573, -0.1236, -0.1674], + [ 0.0799, -0.0934, 0.0391, ..., 0.0485, 0.1271, -0.0570], + [-0.1026, 0.1533, -0.2098, ..., 0.0544, 0.0940, -0.0460], + ..., + [-0.0787, -0.0967, -0.0806, ..., 0.0123, -0.2075, 0.1565], + [ 0.0527, -0.0488, 0.1208, ..., 0.0028, -0.2515, -0.0277], + [-0.2360, -0.1163, -0.1706, ..., -0.2880, 0.0772, -0.1102]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -2.7940e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 2.7940e-08, ..., 2.7008e-08, + 4.0978e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -9.3132e-10, 4.6566e-09, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.2107e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 2.0489e-08, 0.0000e+00]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0071, -0.0390, 0.0061, -0.0122, 0.0079, 0.0127, 0.0216, 0.0101, + -0.0427, -0.0100], device='cuda:0'), grad: tensor([ 1.3970e-08, 1.0245e-08, 1.9278e-07, -9.4995e-08, -2.1607e-07, + -1.8626e-09, 2.6077e-08, -1.8626e-09, 2.3283e-08, 5.7742e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 247.58, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4910 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.11 lr 0.00010000 +Epoch 359, weight, value: tensor([[ 0.0270, -0.1873, -0.1759, ..., -0.3578, -0.1236, -0.1676], + [ 0.0799, -0.0954, 0.0391, ..., 0.0483, 0.1270, -0.0569], + [-0.1035, 0.1536, -0.2099, ..., 0.0546, 0.0945, -0.0453], + ..., + [-0.0779, -0.0969, -0.0806, ..., 0.0123, -0.2076, 0.1565], + [ 0.0526, -0.0496, 0.1207, ..., 0.0026, -0.2519, -0.0285], + [-0.2363, -0.1166, -0.1708, ..., -0.2897, 0.0773, -0.1102]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 1.8626e-09, ..., 3.7253e-09, + 1.9558e-08, 0.0000e+00], + [-4.6566e-09, -2.7940e-09, -9.8720e-08, ..., -2.5146e-08, + -2.6636e-07, 0.0000e+00], + [ 1.8626e-09, -5.8953e-07, 1.6764e-08, ..., -6.4354e-07, + -1.9558e-08, 0.0000e+00], + ..., + [ 8.3819e-09, 5.4389e-07, 3.7253e-08, ..., 6.1467e-07, + 1.8533e-07, 0.0000e+00], + [ 9.3132e-10, 2.8871e-08, 8.3819e-09, ..., 3.1665e-08, + 2.4214e-08, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, + -6.2399e-08, 0.0000e+00]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0065, -0.0392, 0.0063, -0.0121, 0.0079, 0.0124, 0.0212, 0.0101, + -0.0431, -0.0102], device='cuda:0'), grad: tensor([ 4.5635e-08, -3.4645e-07, -2.4065e-06, -9.3132e-10, -1.5832e-07, + 3.3528e-08, 1.9558e-07, 2.5816e-06, 1.5181e-07, -9.4064e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 247.93, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4405 re_mapping 0.0040 re_causal 0.0115 /// teacc 99.03 lr 0.00010000 +Epoch 360, weight, value: tensor([[ 0.0243, -0.1877, -0.1761, ..., -0.3596, -0.1241, -0.1684], + [ 0.0799, -0.0961, 0.0393, ..., 0.0485, 0.1274, -0.0551], + [-0.1025, 0.1538, -0.2101, ..., 0.0547, 0.0947, -0.0439], + ..., + [-0.0779, -0.0970, -0.0807, ..., 0.0121, -0.2081, 0.1564], + [ 0.0524, -0.0504, 0.1207, ..., 0.0025, -0.2530, -0.0294], + [-0.2369, -0.1171, -0.1709, ..., -0.2911, 0.0790, -0.1102]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 1.8626e-09, ..., 5.5879e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 1.7695e-08, -1.4901e-08, ..., 1.6764e-08, + -9.3132e-10, 4.6566e-09], + [ 0.0000e+00, -1.2945e-07, 7.5437e-08, ..., -5.1223e-08, + -1.2480e-07, 2.7940e-09], + ..., + [ 0.0000e+00, 6.6124e-08, 2.7940e-09, ..., 4.2841e-08, + 7.5437e-08, -1.3970e-08], + [ 0.0000e+00, 2.6077e-08, 9.3132e-09, ..., 2.6077e-08, + 3.2596e-08, 2.7940e-09], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 1.8626e-09, + 2.8871e-08, 0.0000e+00]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0067, -0.0390, 0.0063, -0.0119, 0.0064, 0.0122, 0.0216, 0.0099, + -0.0439, -0.0088], device='cuda:0'), grad: tensor([ 4.6566e-08, 8.1025e-08, -4.3772e-07, -1.4249e-07, -2.6263e-07, + 2.7008e-08, 1.2200e-07, 3.1013e-07, 1.6298e-07, 1.0710e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 247.31, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4962 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.10 lr 0.00010000 +Epoch 361, weight, value: tensor([[ 0.0242, -0.1881, -0.1764, ..., -0.3601, -0.1245, -0.1686], + [ 0.0800, -0.0965, 0.0394, ..., 0.0487, 0.1278, -0.0566], + [-0.1029, 0.1537, -0.2104, ..., 0.0547, 0.0948, -0.0439], + ..., + [-0.0776, -0.0968, -0.0809, ..., 0.0120, -0.2087, 0.1581], + [ 0.0523, -0.0505, 0.1209, ..., 0.0028, -0.2535, -0.0318], + [-0.2371, -0.1173, -0.1710, ..., -0.2914, 0.0798, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 2.5611e-08, + 4.0047e-08, 3.2596e-09], + [ 0.0000e+00, -2.4680e-08, 1.3970e-09, ..., -8.8010e-08, + -2.7660e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.2817e-08, + 2.3283e-09, -6.0536e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 5.5879e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -6.9849e-09, 4.6566e-10]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0072, -0.0389, 0.0062, -0.0117, 0.0063, 0.0117, 0.0228, 0.0098, + -0.0440, -0.0080], device='cuda:0'), grad: tensor([-8.7079e-08, 1.0571e-07, -4.1304e-07, 2.7008e-08, 1.4901e-08, + 8.7544e-08, 2.8592e-07, -7.1712e-08, 2.4680e-08, 3.4925e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 247.71, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4861 re_mapping 0.0044 re_causal 0.0122 /// teacc 99.15 lr 0.00010000 +Epoch 362, weight, value: tensor([[ 0.0255, -0.1894, -0.1766, ..., -0.3612, -0.1249, -0.1687], + [ 0.0797, -0.0971, 0.0396, ..., 0.0490, 0.1284, -0.0568], + [-0.1029, 0.1539, -0.2108, ..., 0.0548, 0.0952, -0.0436], + ..., + [-0.0777, -0.0970, -0.0811, ..., 0.0117, -0.2095, 0.1582], + [ 0.0522, -0.0504, 0.1226, ..., 0.0029, -0.2535, -0.0322], + [-0.2379, -0.1183, -0.1741, ..., -0.2919, 0.0805, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 6.9849e-09, ..., 1.3970e-09, + -3.2596e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 9.3132e-10, ..., -9.3132e-10, + 5.1223e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 2.3283e-09, + 6.0536e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 8.8476e-09, 1.8626e-09]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0072, -0.0387, 0.0063, -0.0119, 0.0055, 0.0123, 0.0226, 0.0098, + -0.0413, -0.0102], device='cuda:0'), grad: tensor([-5.1148e-06, 1.6298e-08, 1.0896e-07, 1.6876e-06, -3.5390e-08, + -1.8431e-06, 1.7416e-07, 1.8626e-08, 9.0338e-08, 4.9211e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 247.34, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4475 re_mapping 0.0043 re_causal 0.0120 /// teacc 99.08 lr 0.00010000 +Epoch 363, weight, value: tensor([[ 0.0256, -0.1906, -0.1771, ..., -0.3615, -0.1255, -0.1687], + [ 0.0796, -0.0968, 0.0400, ..., 0.0494, 0.1300, -0.0568], + [-0.1029, 0.1538, -0.2122, ..., 0.0539, 0.0952, -0.0436], + ..., + [-0.0777, -0.0968, -0.0814, ..., 0.0120, -0.2111, 0.1584], + [ 0.0522, -0.0510, 0.1220, ..., 0.0027, -0.2553, -0.0323], + [-0.2380, -0.1183, -0.1742, ..., -0.2921, 0.0807, -0.1107]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 9.3132e-10, 0.0000e+00], + [ 6.9849e-09, 1.3970e-09, -2.7474e-08, ..., -4.1910e-09, + 1.7695e-08, 0.0000e+00], + [ 4.6566e-10, -5.1223e-09, 3.2596e-09, ..., -1.8626e-09, + -3.7253e-09, 0.0000e+00], + ..., + [ 1.5972e-07, 4.6566e-10, 7.9162e-09, ..., 2.3283e-09, + 1.3132e-07, 0.0000e+00], + [ 1.8626e-09, 1.3970e-09, 9.3132e-09, ..., 6.9849e-09, + 4.6566e-09, 0.0000e+00], + [ 1.8626e-09, 4.6566e-10, 4.6566e-10, ..., 9.3132e-10, + 1.0710e-08, 0.0000e+00]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0074, -0.0381, 0.0059, -0.0110, 0.0056, 0.0121, 0.0245, 0.0097, + -0.0427, -0.0102], device='cuda:0'), grad: tensor([ 1.0245e-08, 2.0396e-07, 2.7940e-09, -5.4501e-06, -9.3225e-07, + 5.1968e-06, 7.8231e-08, 6.2212e-07, 2.0908e-07, 5.8673e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 247.69, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4720 re_mapping 0.0041 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 364, weight, value: tensor([[ 0.0254, -0.1910, -0.1773, ..., -0.3617, -0.1256, -0.1687], + [ 0.0794, -0.0953, 0.0398, ..., 0.0492, 0.1303, -0.0568], + [-0.1029, 0.1540, -0.2133, ..., 0.0539, 0.0952, -0.0436], + ..., + [-0.0794, -0.0971, -0.0813, ..., 0.0121, -0.2114, 0.1584], + [ 0.0520, -0.0513, 0.1218, ..., 0.0025, -0.2558, -0.0323], + [-0.2405, -0.1184, -0.1744, ..., -0.2925, 0.0829, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 3.0361e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 5.0291e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 1.8626e-09, + 1.4901e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 5.5879e-09, + 1.8626e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.1176e-07, ..., -2.2352e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 1.8626e-09, + -1.0878e-06, 0.0000e+00]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0075, -0.0383, 0.0059, -0.0113, 0.0034, 0.0125, 0.0247, 0.0098, + -0.0432, -0.0082], device='cuda:0'), grad: tensor([ 1.6801e-06, 1.6019e-07, 1.2107e-07, 1.1176e-07, 1.7900e-06, + 2.3469e-07, -1.8254e-07, 7.6368e-08, -4.3958e-07, -3.5539e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 247.79, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4819 re_mapping 0.0041 re_causal 0.0119 /// teacc 98.98 lr 0.00010000 +Epoch 365, weight, value: tensor([[ 0.0253, -0.1912, -0.1775, ..., -0.3619, -0.1258, -0.1687], + [ 0.0796, -0.0947, 0.0400, ..., 0.0492, 0.1313, -0.0577], + [-0.1028, 0.1546, -0.2136, ..., 0.0543, 0.0957, -0.0434], + ..., + [-0.0793, -0.0977, -0.0814, ..., 0.0118, -0.2123, 0.1591], + [ 0.0519, -0.0515, 0.1217, ..., 0.0024, -0.2563, -0.0325], + [-0.2427, -0.1202, -0.1750, ..., -0.2941, 0.0822, -0.1107]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 1.8626e-09, + 7.4506e-09, 0.0000e+00], + [ 7.4506e-09, 2.4214e-08, 7.4506e-09, ..., 1.3039e-08, + 4.2841e-08, 0.0000e+00], + [ 1.8626e-09, -1.4529e-07, 3.7253e-09, ..., -4.4703e-08, + -2.5146e-07, 0.0000e+00], + ..., + [-2.2352e-08, 3.7253e-09, -0.0000e+00, ..., -5.5879e-09, + 1.3039e-08, 0.0000e+00], + [ 3.7253e-09, 8.9407e-08, -3.7253e-08, ..., 1.3039e-08, + 1.5832e-07, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 1.8626e-08, ..., 9.3132e-09, + -1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0074, -0.0380, 0.0062, -0.0103, 0.0036, 0.0109, 0.0260, 0.0095, + -0.0435, -0.0094], device='cuda:0'), grad: tensor([-1.1384e-05, 1.9558e-07, -6.8732e-07, 9.4995e-08, 2.0489e-08, + 1.0170e-06, 1.0118e-05, -5.5879e-09, 5.3458e-07, 1.3225e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 248.00, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4976 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.04 lr 0.00010000 +Epoch 366, weight, value: tensor([[ 0.0252, -0.1885, -0.1750, ..., -0.3622, -0.1254, -0.1688], + [ 0.0794, -0.0954, 0.0401, ..., 0.0492, 0.1313, -0.0585], + [-0.1028, 0.1547, -0.2140, ..., 0.0544, 0.0960, -0.0431], + ..., + [-0.0808, -0.0978, -0.0814, ..., 0.0118, -0.2127, 0.1599], + [ 0.0517, -0.0518, 0.1216, ..., 0.0022, -0.2569, -0.0327], + [-0.2459, -0.1217, -0.1751, ..., -0.2945, 0.0824, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 2.2352e-08, ..., 2.9802e-08, + 4.2841e-08, 0.0000e+00], + [ 0.0000e+00, 1.2107e-07, -4.8429e-07, ..., 2.7567e-07, + 1.9558e-07, 0.0000e+00], + [ 0.0000e+00, -3.9302e-07, 2.6077e-08, ..., -1.4044e-06, + -1.7099e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3097e-07, 7.2643e-08, ..., 7.0408e-07, + 8.5123e-07, 0.0000e+00], + [ 0.0000e+00, 2.2352e-08, 3.3714e-07, ..., 3.1851e-07, + 5.1036e-07, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, 1.6764e-08, ..., 6.3330e-08, + 8.3819e-08, 0.0000e+00]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0062, -0.0381, 0.0062, -0.0098, 0.0041, 0.0108, 0.0264, 0.0094, + -0.0440, -0.0095], device='cuda:0'), grad: tensor([ 1.0431e-07, -3.1665e-08, -3.5577e-06, 0.0000e+00, -5.5879e-09, + 1.6764e-08, 3.3528e-08, 1.8664e-06, 1.3616e-06, 1.9558e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 247.84, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.5032 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.03 lr 0.00010000 +Epoch 367, weight, value: tensor([[ 0.0249, -0.1905, -0.1755, ..., -0.3636, -0.1277, -0.1688], + [ 0.0793, -0.0967, 0.0399, ..., 0.0487, 0.1313, -0.0583], + [-0.1030, 0.1554, -0.2151, ..., 0.0547, 0.0972, -0.0431], + ..., + [-0.0807, -0.0981, -0.0813, ..., 0.0120, -0.2129, 0.1599], + [ 0.0514, -0.0506, 0.1214, ..., 0.0023, -0.2592, -0.0328], + [-0.2469, -0.1227, -0.1743, ..., -0.2955, 0.0839, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 7.4506e-09, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 4.2841e-08, -1.4901e-08, ..., 5.4017e-08, + 9.8720e-08, 0.0000e+00], + [ 0.0000e+00, -1.9856e-06, 0.0000e+00, ..., -2.6245e-06, + -5.7556e-06, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3039e-08, 1.8626e-09, ..., 1.8626e-08, + 5.4017e-08, 0.0000e+00], + [ 0.0000e+00, 1.8831e-06, 0.0000e+00, ..., 2.4922e-06, + 5.4613e-06, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 2.2352e-08, + 4.8429e-08, 0.0000e+00]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0060, -0.0385, 0.0066, -0.0108, 0.0039, 0.0107, 0.0266, 0.0097, + -0.0450, -0.0094], device='cuda:0'), grad: tensor([ 2.9802e-08, 1.9372e-07, -1.0870e-05, 1.1176e-08, 3.1665e-08, + 2.2352e-08, 6.3330e-08, 7.4506e-08, 1.0327e-05, 9.8720e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 247.76, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4834 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.13 lr 0.00010000 +Epoch 368, weight, value: tensor([[ 0.0233, -0.1923, -0.1757, ..., -0.3666, -0.1281, -0.1689], + [ 0.0791, -0.0947, 0.0400, ..., 0.0493, 0.1327, -0.0580], + [-0.1030, 0.1554, -0.2160, ..., 0.0544, 0.0961, -0.0432], + ..., + [-0.0802, -0.0982, -0.0813, ..., 0.0119, -0.2134, 0.1599], + [ 0.0508, -0.0509, 0.1215, ..., 0.0021, -0.2601, -0.0330], + [-0.2491, -0.1232, -0.1746, ..., -0.2963, 0.0838, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-08, ..., -3.7253e-09, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 1.8626e-09, ..., -1.8626e-09, + -3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., -1.8626e-09, + 2.7940e-08, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 9.3132e-07, 0.0000e+00]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0060, -0.0382, 0.0064, -0.0112, 0.0041, 0.0109, 0.0269, 0.0096, + -0.0453, -0.0099], device='cuda:0'), grad: tensor([ 3.7253e-09, -2.2352e-08, 1.8626e-09, 1.8626e-09, -1.4137e-06, + 3.7253e-09, 9.3132e-09, 1.6764e-08, -2.6077e-08, 1.4305e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 247.71, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4856 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.06 lr 0.00010000 +Epoch 369, weight, value: tensor([[ 0.0234, -0.1927, -0.1756, ..., -0.3668, -0.1281, -0.1689], + [ 0.0789, -0.0947, 0.0402, ..., 0.0495, 0.1334, -0.0580], + [-0.1031, 0.1555, -0.2163, ..., 0.0544, 0.0964, -0.0432], + ..., + [-0.0801, -0.0982, -0.0815, ..., 0.0118, -0.2141, 0.1599], + [ 0.0506, -0.0509, 0.1217, ..., 0.0020, -0.2603, -0.0329], + [-0.2506, -0.1233, -0.1751, ..., -0.2984, 0.0837, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4506e-09, 1.8626e-09, ..., 9.3132e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, -1.7323e-07, ..., -6.5193e-08, + -1.8813e-07, 0.0000e+00], + [ 1.8626e-09, -4.3586e-07, 1.1548e-07, ..., -4.0233e-07, + -4.0978e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7439e-07, 4.0978e-08, ..., 3.9488e-07, + 1.5274e-07, 0.0000e+00], + [-3.7253e-09, 2.9802e-08, -1.3858e-06, ..., -8.2888e-07, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 3.7253e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0059, -0.0380, 0.0065, -0.0122, 0.0044, 0.0114, 0.0268, 0.0096, + -0.0454, -0.0103], device='cuda:0'), grad: tensor([ 4.8429e-08, -3.6508e-07, -2.2203e-06, 2.9244e-06, -5.2154e-08, + 2.9802e-08, 1.8254e-07, 2.1476e-06, -2.7027e-06, -1.8626e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 247.53, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4961 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.07 lr 0.00010000 +Epoch 370, weight, value: tensor([[ 0.0231, -0.1941, -0.1760, ..., -0.3676, -0.1285, -0.1689], + [ 0.0791, -0.0949, 0.0408, ..., 0.0499, 0.1343, -0.0577], + [-0.1036, 0.1558, -0.2164, ..., 0.0546, 0.0967, -0.0436], + ..., + [-0.0807, -0.0985, -0.0820, ..., 0.0114, -0.2154, 0.1600], + [ 0.0506, -0.0509, 0.1225, ..., 0.0022, -0.2605, -0.0330], + [-0.2508, -0.1234, -0.1757, ..., -0.3000, 0.0836, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4506e-09, 1.8626e-09, ..., 9.3132e-09, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, -2.6077e-08, ..., 5.5879e-09, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, -2.4214e-08, 5.5879e-09, ..., -3.1665e-08, + -8.5682e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 7.4506e-09, 1.8626e-08, ..., 1.6764e-08, + 3.1665e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0065, -0.0375, 0.0067, -0.0116, 0.0044, 0.0095, 0.0285, 0.0091, + -0.0452, -0.0104], device='cuda:0'), grad: tensor([ 8.2143e-07, -1.1176e-08, -2.3842e-07, 7.4506e-09, 7.4506e-09, + 2.6636e-07, -1.0263e-06, 8.1956e-08, 8.0094e-08, 1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 248.01, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4763 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.06 lr 0.00010000 +Epoch 371, weight, value: tensor([[ 0.0230, -0.1949, -0.1765, ..., -0.3679, -0.1286, -0.1689], + [ 0.0791, -0.0950, 0.0404, ..., 0.0493, 0.1347, -0.0576], + [-0.1040, 0.1558, -0.2168, ..., 0.0547, 0.0970, -0.0442], + ..., + [-0.0809, -0.0985, -0.0816, ..., 0.0119, -0.2157, 0.1602], + [ 0.0505, -0.0509, 0.1228, ..., 0.0023, -0.2607, -0.0331], + [-0.2510, -0.1234, -0.1759, ..., -0.3011, 0.0834, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.7544e-08, + -2.8871e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 9.3132e-09, + 6.3330e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.4901e-08, + 4.4703e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.1176e-08, + 4.2841e-08, 0.0000e+00], + [ 1.1176e-08, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-08, + 1.8254e-07, 0.0000e+00]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0068, -0.0379, 0.0068, -0.0111, 0.0048, 0.0087, 0.0287, 0.0094, + -0.0452, -0.0108], device='cuda:0'), grad: tensor([-2.0266e-06, 3.2037e-07, 3.0175e-07, 5.4948e-07, -5.7369e-07, + -7.4692e-07, 6.4448e-07, 2.9244e-07, 3.2037e-07, 9.2015e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 247.77, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4467 re_mapping 0.0040 re_causal 0.0113 /// teacc 99.01 lr 0.00010000 +Epoch 372, weight, value: tensor([[ 0.0230, -0.1948, -0.1767, ..., -0.3684, -0.1286, -0.1690], + [ 0.0790, -0.0950, 0.0411, ..., 0.0499, 0.1362, -0.0572], + [-0.1041, 0.1559, -0.2177, ..., 0.0547, 0.0971, -0.0446], + ..., + [-0.0809, -0.0986, -0.0823, ..., 0.0115, -0.2172, 0.1606], + [ 0.0505, -0.0509, 0.1229, ..., 0.0023, -0.2608, -0.0332], + [-0.2511, -0.1238, -0.1764, ..., -0.3022, 0.0833, -0.1107]], + device='cuda:0'), grad: tensor([[-1.8626e-09, 2.7940e-08, 1.8626e-09, ..., 1.1176e-08, + 6.3330e-08, 0.0000e+00], + [ 0.0000e+00, 1.4901e-08, 1.0058e-07, ..., 4.8429e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, -3.1851e-07, 9.3132e-09, ..., -5.5321e-07, + -6.4075e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3225e-07, 1.8626e-09, ..., 2.6077e-07, + 2.8312e-07, 0.0000e+00], + [ 0.0000e+00, 1.3039e-07, -2.9989e-07, ..., 1.8254e-07, + 2.7381e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 3.5390e-08, 0.0000e+00]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0064, -0.0373, 0.0068, -0.0110, 0.0047, 0.0093, 0.0285, 0.0089, + -0.0453, -0.0111], device='cuda:0'), grad: tensor([ 1.4529e-07, 3.0920e-07, -1.7062e-06, 1.0245e-07, -5.1968e-07, + 7.2643e-08, 4.6194e-07, 7.0781e-07, 2.7195e-07, 1.6019e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 247.36, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4779 re_mapping 0.0041 re_causal 0.0115 /// teacc 99.05 lr 0.00010000 +Epoch 373, weight, value: tensor([[ 0.0235, -0.1953, -0.1771, ..., -0.3690, -0.1288, -0.1690], + [ 0.0790, -0.0957, 0.0415, ..., 0.0498, 0.1370, -0.0562], + [-0.1041, 0.1557, -0.2184, ..., 0.0546, 0.0974, -0.0447], + ..., + [-0.0809, -0.0983, -0.0824, ..., 0.0116, -0.2177, 0.1610], + [ 0.0506, -0.0511, 0.1229, ..., 0.0023, -0.2615, -0.0336], + [-0.2512, -0.1242, -0.1766, ..., -0.3035, 0.0830, -0.1108]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, -7.2643e-08, ..., -2.6077e-08, + -6.3330e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 3.7253e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -1.8626e-09, + 1.8626e-08, 0.0000e+00], + [-0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -5.5879e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.5390e-08, 0.0000e+00]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0062, -0.0372, 0.0067, -0.0109, 0.0052, 0.0103, 0.0259, 0.0092, + -0.0461, -0.0118], device='cuda:0'), grad: tensor([-1.5274e-07, -1.0431e-07, 5.0291e-08, 4.8429e-08, -1.1921e-07, + 1.8999e-07, 5.2154e-08, 5.9605e-08, 2.7940e-08, -6.7055e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 372---------------------------------------------------- +epoch 372, time 264.53, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4962 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.20 lr 0.00010000 +Epoch 374, weight, value: tensor([[ 0.0234, -0.1956, -0.1773, ..., -0.3695, -0.1291, -0.1702], + [ 0.0788, -0.0963, 0.0420, ..., 0.0502, 0.1381, -0.0565], + [-0.1042, 0.1557, -0.2188, ..., 0.0546, 0.0974, -0.0435], + ..., + [-0.0816, -0.0983, -0.0829, ..., 0.0114, -0.2186, 0.1616], + [ 0.0505, -0.0513, 0.1224, ..., 0.0019, -0.2623, -0.0338], + [-0.2518, -0.1243, -0.1768, ..., -0.3045, 0.0827, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.6764e-08, 3.7253e-09, ..., 1.6764e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-08, 1.8626e-09, ..., 5.0291e-08, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, -1.2498e-06, 1.2480e-07, ..., -1.0543e-06, + -6.4634e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.0300e-06, 0.0000e+00, ..., 8.1770e-07, + 5.0105e-07, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.3243e-06, ..., -3.6694e-07, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 5.7742e-08, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0062, -0.0367, 0.0067, -0.0109, 0.0058, 0.0106, 0.0258, 0.0089, + -0.0467, -0.0123], device='cuda:0'), grad: tensor([ 1.8626e-09, 2.0303e-07, -3.5781e-06, 9.7416e-07, -2.4773e-07, + -3.7253e-09, 1.4640e-06, 2.9057e-06, -1.9502e-06, 2.0117e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 247.45, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4440 re_mapping 0.0041 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +Epoch 375, weight, value: tensor([[ 0.0235, -0.1955, -0.1776, ..., -0.3702, -0.1269, -0.1700], + [ 0.0823, -0.0963, 0.0427, ..., 0.0496, 0.1381, -0.0566], + [-0.1050, 0.1569, -0.2203, ..., 0.0560, 0.0992, -0.0435], + ..., + [-0.0820, -0.0984, -0.0834, ..., 0.0110, -0.2199, 0.1618], + [ 0.0503, -0.0514, 0.1220, ..., 0.0014, -0.2633, -0.0338], + [-0.2521, -0.1246, -0.1770, ..., -0.3058, 0.0813, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + -5.5879e-09, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 2.5891e-07, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0027, -0.0369, 0.0077, -0.0107, 0.0055, 0.0102, 0.0250, 0.0085, + -0.0473, -0.0142], device='cuda:0'), grad: tensor([ 0.0000e+00, 2.7940e-08, 2.0489e-08, -6.6496e-07, 1.8626e-08, + 1.3039e-08, -1.3039e-08, 6.2957e-07, -1.0245e-07, 5.9605e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 247.21, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4622 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +Epoch 376, weight, value: tensor([[ 0.0234, -0.1957, -0.1780, ..., -0.3708, -0.1272, -0.1702], + [ 0.0830, -0.0978, 0.0449, ..., 0.0501, 0.1385, -0.0566], + [-0.1051, 0.1579, -0.2217, ..., 0.0571, 0.1006, -0.0434], + ..., + [-0.0823, -0.0985, -0.0854, ..., 0.0092, -0.2230, 0.1629], + [ 0.0504, -0.0514, 0.1223, ..., 0.0016, -0.2633, -0.0339], + [-0.2527, -0.1247, -0.1773, ..., -0.3064, 0.0814, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 7.4506e-08, -1.8626e-09, ..., 1.7323e-07, + 3.7625e-07, 0.0000e+00], + [ 0.0000e+00, -8.5682e-08, 2.4214e-08, ..., -1.9930e-07, + -4.1351e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-09, 1.8626e-09, ..., 2.0489e-08, + 4.6566e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0028, -0.0354, 0.0084, -0.0105, 0.0055, 0.0098, 0.0257, 0.0064, + -0.0470, -0.0143], device='cuda:0'), grad: tensor([ 4.6566e-08, 5.3272e-07, -4.1537e-07, 2.9802e-08, -2.4214e-08, + 1.4901e-08, 2.2352e-08, 7.6368e-08, -2.4214e-07, -4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 247.16, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4730 re_mapping 0.0041 re_causal 0.0120 /// teacc 99.06 lr 0.00010000 +Epoch 377, weight, value: tensor([[ 0.0231, -0.1960, -0.1784, ..., -0.3713, -0.1274, -0.1702], + [ 0.0831, -0.0978, 0.0448, ..., 0.0498, 0.1386, -0.0567], + [-0.1052, 0.1579, -0.2224, ..., 0.0571, 0.1006, -0.0435], + ..., + [-0.0824, -0.0985, -0.0852, ..., 0.0096, -0.2233, 0.1630], + [ 0.0506, -0.0514, 0.1220, ..., 0.0015, -0.2656, -0.0340], + [-0.2528, -0.1248, -0.1762, ..., -0.3069, 0.0837, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 1.0803e-07, 0.0000e+00], + [-2.8312e-07, 5.5879e-09, -2.1607e-07, ..., -1.8254e-07, + -4.8615e-07, 0.0000e+00], + [ 4.6566e-08, -2.9802e-08, 8.3819e-08, ..., 3.5390e-08, + 2.0862e-07, 0.0000e+00], + ..., + [ 1.8626e-09, 2.4214e-08, 1.4901e-08, ..., 7.4506e-09, + 8.7544e-08, 0.0000e+00], + [ 2.0489e-08, 0.0000e+00, 1.8626e-08, ..., 1.4901e-08, + 5.9605e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 2.2352e-07, 0.0000e+00]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0029, -0.0357, 0.0084, -0.0108, 0.0042, 0.0097, 0.0260, 0.0066, + -0.0482, -0.0125], device='cuda:0'), grad: tensor([ 2.2911e-07, -2.6636e-07, 3.4459e-07, -4.1537e-07, -1.5870e-06, + -3.9116e-08, 8.4564e-07, 2.2911e-07, 8.5682e-08, 5.7369e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 247.40, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4671 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.19 lr 0.00010000 +Epoch 378, weight, value: tensor([[ 0.0231, -0.1960, -0.1791, ..., -0.3717, -0.1276, -0.1704], + [ 0.0834, -0.0978, 0.0447, ..., 0.0497, 0.1387, -0.0568], + [-0.1056, 0.1579, -0.2229, ..., 0.0571, 0.1006, -0.0437], + ..., + [-0.0823, -0.0985, -0.0852, ..., 0.0098, -0.2234, 0.1630], + [ 0.0508, -0.0511, 0.1222, ..., 0.0016, -0.2660, -0.0341], + [-0.2528, -0.1248, -0.1761, ..., -0.3073, 0.0841, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4773e-07, + 1.8999e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -2.4959e-07, + -2.0675e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.5879e-09, + 2.2352e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0029, -0.0358, 0.0084, -0.0109, 0.0042, 0.0098, 0.0259, 0.0067, + -0.0483, -0.0122], device='cuda:0'), grad: tensor([-1.0058e-07, 4.7311e-07, -4.0606e-07, 2.0489e-08, 1.2480e-07, + 0.0000e+00, 7.4506e-09, -1.5087e-07, 3.7253e-09, 2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 247.36, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4977 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.10 lr 0.00010000 +Epoch 379, weight, value: tensor([[ 0.0231, -0.1964, -0.1818, ..., -0.3730, -0.1280, -0.1707], + [ 0.0835, -0.0978, 0.0450, ..., 0.0497, 0.1387, -0.0578], + [-0.1058, 0.1579, -0.2232, ..., 0.0571, 0.1006, -0.0459], + ..., + [-0.0824, -0.0985, -0.0853, ..., 0.0098, -0.2235, 0.1652], + [ 0.0510, -0.0511, 0.1222, ..., 0.0014, -0.2666, -0.0346], + [-0.2528, -0.1248, -0.1765, ..., -0.3086, 0.0853, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.8626e-09, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -0.0000e+00, + 4.2841e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 1.8626e-08, ..., 9.3132e-09, + -0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-09, ..., -9.3132e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.3448e-06, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 1.6764e-08, + -2.7940e-08, 0.0000e+00]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0027, -0.0358, 0.0083, -0.0109, 0.0034, 0.0099, 0.0254, 0.0068, + -0.0485, -0.0118], device='cuda:0'), grad: tensor([ 1.8440e-07, 1.5274e-07, 6.1467e-08, -2.7940e-08, -1.6205e-07, + 1.5032e-06, 1.0170e-06, -2.4214e-07, -2.6450e-06, 1.6764e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 247.07, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4802 re_mapping 0.0038 re_causal 0.0116 /// teacc 99.12 lr 0.00010000 +Epoch 380, weight, value: tensor([[ 0.0231, -0.1968, -0.1826, ..., -0.3735, -0.1282, -0.1711], + [ 0.0835, -0.0978, 0.0450, ..., 0.0497, 0.1388, -0.0591], + [-0.1058, 0.1579, -0.2240, ..., 0.0571, 0.1006, -0.0459], + ..., + [-0.0825, -0.0985, -0.0853, ..., 0.0098, -0.2236, 0.1658], + [ 0.0509, -0.0511, 0.1228, ..., 0.0015, -0.2667, -0.0327], + [-0.2529, -0.1249, -0.1769, ..., -0.3093, 0.0852, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0004, -0.0359, 0.0082, -0.0091, 0.0037, 0.0088, 0.0220, 0.0068, + -0.0483, -0.0120], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.4901e-08, 9.3132e-09, -9.3132e-09, -7.4506e-09, + -3.1665e-08, -2.9802e-08, -2.9989e-07, 2.0489e-08, 3.2410e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 247.88, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4619 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.08 lr 0.00010000 +Epoch 381, weight, value: tensor([[ 0.0230, -0.1969, -0.1828, ..., -0.3738, -0.1283, -0.1700], + [ 0.0835, -0.0978, 0.0450, ..., 0.0494, 0.1386, -0.0607], + [-0.1059, 0.1579, -0.2242, ..., 0.0572, 0.1010, -0.0460], + ..., + [-0.0825, -0.0985, -0.0853, ..., 0.0101, -0.2237, 0.1672], + [ 0.0509, -0.0511, 0.1236, ..., 0.0019, -0.2667, -0.0329], + [-0.2529, -0.1250, -0.1772, ..., -0.3108, 0.0851, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, -5.5879e-08, ..., 5.5879e-09, + -4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [-9.3132e-09, 3.7253e-09, 3.7253e-09, ..., 5.5879e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.4703e-08, ..., 9.3132e-09, + 7.4506e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0004, -0.0362, 0.0085, -0.0102, 0.0039, 0.0088, 0.0219, 0.0070, + -0.0477, -0.0122], device='cuda:0'), grad: tensor([ 3.7253e-08, -9.8720e-08, 3.7253e-09, -1.9372e-07, 1.8626e-09, + 1.8440e-07, 1.6764e-08, -5.7742e-08, 8.0094e-08, 2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 247.92, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4815 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.02 lr 0.00010000 +Epoch 382, weight, value: tensor([[ 0.0226, -0.1998, -0.1833, ..., -0.3756, -0.1285, -0.1700], + [ 0.0834, -0.0978, 0.0457, ..., 0.0500, 0.1390, -0.0618], + [-0.1060, 0.1584, -0.2251, ..., 0.0573, 0.1011, -0.0459], + ..., + [-0.0824, -0.0987, -0.0859, ..., 0.0096, -0.2245, 0.1692], + [ 0.0509, -0.0531, 0.1243, ..., 0.0010, -0.2672, -0.0351], + [-0.2530, -0.1251, -0.1775, ..., -0.3118, 0.0853, -0.1113]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 4.0978e-08, ..., 2.4214e-08, + 3.7253e-08, 0.0000e+00], + [ 5.0291e-08, 1.3039e-08, -1.8068e-07, ..., -5.0291e-08, + 2.0489e-08, -0.0000e+00], + [ 0.0000e+00, 3.3528e-08, 1.4901e-08, ..., 2.0489e-08, + -8.0094e-08, 0.0000e+00], + ..., + [ 5.5879e-09, -6.1467e-08, 2.2352e-08, ..., -8.5682e-08, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 8.3819e-08, ..., 4.8429e-08, + 7.2643e-08, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 3.7253e-09, ..., 5.5879e-09, + -0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0001, -0.0357, 0.0087, -0.0092, 0.0040, 0.0076, 0.0220, 0.0065, + -0.0478, -0.0122], device='cuda:0'), grad: tensor([ 1.6578e-07, -2.7567e-07, 1.1921e-07, 1.2480e-07, -2.5518e-07, + -4.9919e-07, 4.2096e-07, -2.2165e-07, 3.9861e-07, 2.4214e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 247.09, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4733 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.06 lr 0.00010000 +Epoch 383, weight, value: tensor([[ 0.0220, -0.2000, -0.1843, ..., -0.3761, -0.1276, -0.1701], + [ 0.0840, -0.0980, 0.0433, ..., 0.0478, 0.1391, -0.0629], + [-0.1061, 0.1590, -0.2259, ..., 0.0575, 0.1012, -0.0457], + ..., + [-0.0824, -0.0993, -0.0835, ..., 0.0117, -0.2246, 0.1700], + [ 0.0507, -0.0531, 0.1259, ..., 0.0020, -0.2673, -0.0350], + [-0.2531, -0.1254, -0.1779, ..., -0.3128, 0.0848, -0.1115]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 5.4017e-08, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 1.3039e-08, + -1.8626e-09, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.2107e-07, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9116e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0007, -0.0381, 0.0089, -0.0094, 0.0040, 0.0075, 0.0219, 0.0089, + -0.0467, -0.0135], device='cuda:0'), grad: tensor([ 3.7253e-08, 2.1420e-07, 5.0291e-08, 2.4214e-08, 0.0000e+00, + 3.7253e-09, -2.9802e-08, -4.7125e-07, 1.4342e-07, 2.4214e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 247.47, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4627 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.11 lr 0.00010000 +Epoch 384, weight, value: tensor([[ 0.0212, -0.2004, -0.1846, ..., -0.3768, -0.1277, -0.1706], + [ 0.0847, -0.0979, 0.0433, ..., 0.0478, 0.1392, -0.0630], + [-0.1064, 0.1603, -0.2265, ..., 0.0580, 0.1014, -0.0457], + ..., + [-0.0821, -0.1011, -0.0835, ..., 0.0114, -0.2249, 0.1700], + [ 0.0497, -0.0533, 0.1257, ..., 0.0017, -0.2681, -0.0351], + [-0.2531, -0.1255, -0.1775, ..., -0.3120, 0.0848, -0.1113]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.5635e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 8.3819e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 8.3819e-09, + 1.3039e-08, 0.0000e+00], + ..., + [ 1.3039e-08, 0.0000e+00, 3.7253e-09, ..., -6.5193e-09, + 4.1910e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.5146e-08, 0.0000e+00]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0007, -0.0382, 0.0096, -0.0094, 0.0033, 0.0075, 0.0219, 0.0086, + -0.0475, -0.0130], device='cuda:0'), grad: tensor([ 2.2817e-07, 4.2841e-08, 5.7742e-08, 4.5635e-08, -2.7474e-07, + -6.9663e-07, -2.8219e-07, 2.2631e-07, 1.6578e-07, 4.9546e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 247.51, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4707 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.12 lr 0.00010000 +Epoch 385, weight, value: tensor([[ 0.0211, -0.2003, -0.1852, ..., -0.3771, -0.1284, -0.1731], + [ 0.0860, -0.0979, 0.0433, ..., 0.0478, 0.1393, -0.0639], + [-0.1068, 0.1607, -0.2274, ..., 0.0582, 0.1013, -0.0454], + ..., + [-0.0823, -0.1016, -0.0835, ..., 0.0112, -0.2251, 0.1741], + [ 0.0495, -0.0531, 0.1265, ..., 0.0020, -0.2686, -0.0352], + [-0.2534, -0.1257, -0.1776, ..., -0.3126, 0.0852, -0.1139]], + device='cuda:0'), grad: tensor([[-0.0000e+00, 1.4901e-08, 4.3772e-08, ..., 4.4703e-08, + 5.7742e-08, 0.0000e+00], + [ 0.0000e+00, 1.7695e-08, -1.6820e-06, ..., -9.6578e-07, + -1.5656e-06, 0.0000e+00], + [ 0.0000e+00, -1.9372e-07, 2.5332e-07, ..., -8.0094e-08, + 1.2107e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-08, 1.1753e-06, ..., 7.8604e-07, + 1.2126e-06, 0.0000e+00], + [ 0.0000e+00, 3.5390e-08, 1.4622e-07, ..., 1.2945e-07, + 1.8533e-07, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, 5.4017e-08, ..., 4.0047e-08, + 5.7742e-08, -0.0000e+00]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0005, -0.0381, 0.0098, -0.0091, 0.0032, 0.0068, 0.0220, 0.0085, + -0.0472, -0.0127], device='cuda:0'), grad: tensor([ 2.5798e-07, -5.8934e-06, -4.6939e-07, 1.4435e-07, 4.0978e-08, + 5.0291e-08, -6.7987e-08, 4.8652e-06, 8.0373e-07, 2.4587e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 247.50, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4674 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 386, weight, value: tensor([[ 0.0211, -0.2006, -0.1854, ..., -0.3776, -0.1280, -0.1732], + [ 0.0862, -0.0979, 0.0434, ..., 0.0479, 0.1395, -0.0647], + [-0.1069, 0.1606, -0.2279, ..., 0.0580, 0.1013, -0.0457], + ..., + [-0.0823, -0.1014, -0.0836, ..., 0.0113, -0.2255, 0.1752], + [ 0.0494, -0.0533, 0.1269, ..., 0.0020, -0.2688, -0.0354], + [-0.2535, -0.1259, -0.1782, ..., -0.3140, 0.0851, -0.1147]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.8626e-09, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, -1.6764e-08, ..., 5.5879e-09, + -9.3132e-09, -0.0000e+00], + [ 0.0000e+00, 3.3528e-08, 2.7940e-09, ..., 3.0734e-08, + 2.0489e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -2.0862e-07, 5.5879e-09, ..., -1.8347e-07, + -9.6858e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -1.8626e-09, + 8.3819e-09, 9.3132e-10], + [ 0.0000e+00, 1.5460e-07, 9.3132e-10, ..., 1.3784e-07, + 8.5682e-08, 0.0000e+00]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0024, -0.0381, 0.0094, -0.0088, 0.0029, 0.0061, 0.0215, 0.0087, + -0.0473, -0.0151], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.7695e-08, 1.5274e-07, 9.3132e-09, -2.0489e-08, + -9.3132e-10, 3.7253e-09, -8.5775e-07, 1.2107e-08, 6.8918e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 247.60, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4823 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.08 lr 0.00010000 +Epoch 387, weight, value: tensor([[ 0.0210, -0.2007, -0.1859, ..., -0.3778, -0.1281, -0.1732], + [ 0.0860, -0.0980, 0.0440, ..., 0.0482, 0.1400, -0.0656], + [-0.1071, 0.1606, -0.2283, ..., 0.0580, 0.1014, -0.0458], + ..., + [-0.0805, -0.1014, -0.0840, ..., 0.0111, -0.2262, 0.1757], + [ 0.0492, -0.0532, 0.1269, ..., 0.0020, -0.2690, -0.0354], + [-0.2538, -0.1262, -0.1784, ..., -0.3150, 0.0850, -0.1150]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.4261e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0025, -0.0377, 0.0094, -0.0088, 0.0031, 0.0057, 0.0216, 0.0084, + -0.0475, -0.0153], device='cuda:0'), grad: tensor([ 1.8626e-07, 1.8626e-09, 9.3132e-09, 5.5879e-09, -7.4506e-09, + 1.3039e-07, -3.3434e-07, -9.3132e-10, 2.7940e-09, 1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 247.40, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4873 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.08 lr 0.00010000 +Epoch 388, weight, value: tensor([[ 0.0209, -0.2010, -0.1864, ..., -0.3781, -0.1282, -0.1732], + [ 0.0890, -0.0980, 0.0441, ..., 0.0482, 0.1401, -0.0674], + [-0.1083, 0.1607, -0.2287, ..., 0.0580, 0.1015, -0.0461], + ..., + [-0.0808, -0.1014, -0.0842, ..., 0.0111, -0.2265, 0.1754], + [ 0.0487, -0.0530, 0.1273, ..., 0.0021, -0.2694, -0.0367], + [-0.2553, -0.1263, -0.1786, ..., -0.3159, 0.0833, -0.1119]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 4.1910e-08, + 1.8626e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 6.4261e-08, ..., 3.5390e-08, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.3772e-08, ..., -2.0489e-08, + 2.7940e-09, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 5.5879e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0025, -0.0377, 0.0094, -0.0090, 0.0065, 0.0057, 0.0215, 0.0083, + -0.0475, -0.0172], device='cuda:0'), grad: tensor([ 1.3039e-08, 1.3318e-07, 1.7416e-07, -7.6089e-07, 0.0000e+00, + 3.7998e-07, 1.8626e-09, 3.7253e-09, 3.5390e-08, 2.3283e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 247.61, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4902 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.07 lr 0.00010000 +Epoch 389, weight, value: tensor([[ 0.0209, -0.2012, -0.1870, ..., -0.3788, -0.1281, -0.1734], + [ 0.0891, -0.0980, 0.0454, ..., 0.0491, 0.1414, -0.0641], + [-0.1088, 0.1607, -0.2293, ..., 0.0580, 0.1015, -0.0457], + ..., + [-0.0806, -0.1015, -0.0854, ..., 0.0102, -0.2285, 0.1721], + [ 0.0486, -0.0530, 0.1274, ..., 0.0021, -0.2697, -0.0371], + [-0.2555, -0.1265, -0.1788, ..., -0.3166, 0.0833, -0.1119]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 1.8626e-09, ..., 1.4901e-08, + 1.3039e-07, 0.0000e+00], + [ 0.0000e+00, 1.3039e-08, 1.1809e-06, ..., 2.0545e-06, + 1.3970e-08, 2.9802e-08], + [ 0.0000e+00, -2.7381e-07, 1.6764e-08, ..., -4.2282e-07, + -2.5425e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 2.4494e-07, -1.2247e-06, ..., -1.7080e-06, + 2.1979e-07, -3.0734e-08], + [ 0.0000e+00, 4.6566e-09, 9.3132e-10, ..., 1.5832e-08, + 1.7323e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 4.6566e-09, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0021, -0.0367, 0.0093, -0.0095, 0.0067, 0.0090, 0.0185, 0.0074, + -0.0485, -0.0173], device='cuda:0'), grad: tensor([ 4.9826e-07, 6.5938e-06, -1.4734e-06, 1.7136e-07, 5.3085e-08, + -7.4506e-09, -1.1334e-06, -5.3570e-06, 6.4075e-07, 2.7940e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 247.18, cls_loss 0.0007 cls_loss_mapping 0.0007 cls_loss_causal 0.4472 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.04 lr 0.00010000 +Epoch 390, weight, value: tensor([[ 0.0209, -0.2018, -0.1879, ..., -0.3796, -0.1284, -0.1737], + [ 0.0891, -0.0980, 0.0448, ..., 0.0484, 0.1413, -0.0642], + [-0.1088, 0.1608, -0.2298, ..., 0.0580, 0.1016, -0.0448], + ..., + [-0.0807, -0.1015, -0.0848, ..., 0.0108, -0.2298, 0.1718], + [ 0.0487, -0.0529, 0.1275, ..., 0.0020, -0.2701, -0.0384], + [-0.2563, -0.1269, -0.1791, ..., -0.3150, 0.0861, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -6.5193e-09, ..., -2.7940e-09, + -6.5193e-09, -0.0000e+00], + [ 0.0000e+00, -2.7940e-09, 1.8626e-09, ..., -9.3132e-10, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 2.7940e-09, ..., 1.8626e-09, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.2596e-08, ..., -2.7940e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0019, -0.0374, 0.0093, -0.0096, 0.0068, 0.0090, 0.0185, 0.0073, + -0.0487, -0.0151], device='cuda:0'), grad: tensor([ 9.6858e-08, -9.3132e-09, -7.4506e-09, 7.2643e-08, 4.6566e-09, + 8.3819e-09, -2.0489e-08, 1.3039e-08, -6.9849e-08, -7.9162e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 247.48, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4463 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.15 lr 0.00010000 +Epoch 391, weight, value: tensor([[ 0.0209, -0.2020, -0.1885, ..., -0.3801, -0.1285, -0.1739], + [ 0.0891, -0.0980, 0.0449, ..., 0.0484, 0.1415, -0.0642], + [-0.1088, 0.1609, -0.2300, ..., 0.0581, 0.1016, -0.0448], + ..., + [-0.0807, -0.1016, -0.0849, ..., 0.0108, -0.2301, 0.1719], + [ 0.0487, -0.0531, 0.1281, ..., 0.0021, -0.2703, -0.0389], + [-0.2564, -0.1272, -0.1793, ..., -0.3154, 0.0861, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.5832e-08, 0.0000e+00, ..., 5.0291e-08, + 2.7008e-08, 0.0000e+00], + [ 0.0000e+00, -2.0489e-08, 0.0000e+00, ..., -4.5635e-08, + -3.3528e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., -6.5193e-09, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0019, -0.0374, 0.0094, -0.0097, 0.0069, 0.0091, 0.0185, 0.0073, + -0.0486, -0.0152], device='cuda:0'), grad: tensor([-1.0543e-06, 9.5926e-08, 7.8231e-08, 1.9558e-08, 5.0291e-08, + 4.9360e-08, 5.2713e-07, -2.4214e-08, 2.7940e-08, 2.3656e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 247.55, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4804 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.10 lr 0.00010000 +Epoch 392, weight, value: tensor([[ 0.0213, -0.2019, -0.1890, ..., -0.3803, -0.1286, -0.1739], + [ 0.0890, -0.0981, 0.0450, ..., 0.0483, 0.1424, -0.0643], + [-0.1095, 0.1611, -0.2302, ..., 0.0582, 0.1017, -0.0446], + ..., + [-0.0803, -0.1017, -0.0851, ..., 0.0109, -0.2316, 0.1720], + [ 0.0488, -0.0534, 0.1285, ..., 0.0022, -0.2708, -0.0386], + [-0.2566, -0.1277, -0.1797, ..., -0.3159, 0.0857, -0.1095]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -2.7940e-09, ..., 7.4506e-09, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, -1.1269e-07, 1.8626e-09, ..., -1.1362e-07, + -7.0781e-08, -9.3132e-10], + ..., + [ 0.0000e+00, 6.1467e-08, 2.7940e-09, ..., 5.1223e-08, + 4.2841e-08, 0.0000e+00], + [ 0.0000e+00, 4.7497e-08, -9.3132e-10, ..., 4.7497e-08, + 2.9802e-08, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 2.7940e-09, ..., 2.7940e-09, + 7.5437e-08, 0.0000e+00]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0022, -0.0373, 0.0095, -0.0097, 0.0083, 0.0091, 0.0185, 0.0070, + -0.0487, -0.0156], device='cuda:0'), grad: tensor([-1.5832e-08, 5.7742e-08, -2.5984e-07, 5.5879e-08, -2.2817e-07, + -1.0896e-07, -8.3819e-09, 1.0431e-07, 1.6112e-07, 2.4401e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 247.49, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4647 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.05 lr 0.00010000 +Epoch 393, weight, value: tensor([[ 0.0212, -0.2014, -0.1893, ..., -0.3807, -0.1292, -0.1746], + [ 0.0890, -0.0987, 0.0450, ..., 0.0481, 0.1415, -0.0644], + [-0.1097, 0.1617, -0.2287, ..., 0.0586, 0.1026, -0.0444], + ..., + [-0.0802, -0.1020, -0.0851, ..., 0.0109, -0.2319, 0.1721], + [ 0.0487, -0.0535, 0.1288, ..., 0.0021, -0.2711, -0.0387], + [-0.2567, -0.1291, -0.1794, ..., -0.3167, 0.0871, -0.1096]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.7940e-09, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, -6.5193e-09, 9.3132e-10, ..., -1.9558e-08, + -7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 2.7940e-09, ..., 1.5832e-08, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0021, -0.0378, 0.0101, -0.0098, 0.0081, 0.0090, 0.0186, 0.0070, + -0.0487, -0.0148], device='cuda:0'), grad: tensor([-3.3528e-08, 3.7253e-09, -3.0734e-08, -2.7940e-09, 3.7253e-09, + 1.4901e-08, 1.0245e-08, 2.7940e-08, 9.3132e-10, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 247.54, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4620 re_mapping 0.0037 re_causal 0.0103 /// teacc 99.07 lr 0.00010000 +Epoch 394, weight, value: tensor([[ 0.0211, -0.2013, -0.1906, ..., -0.3830, -0.1303, -0.1753], + [ 0.0889, -0.0991, 0.0450, ..., 0.0474, 0.1415, -0.0644], + [-0.1096, 0.1627, -0.2282, ..., 0.0596, 0.1031, -0.0438], + ..., + [-0.0804, -0.1027, -0.0852, ..., 0.0112, -0.2323, 0.1721], + [ 0.0485, -0.0549, 0.1296, ..., 0.0020, -0.2714, -0.0389], + [-0.2567, -0.1306, -0.1808, ..., -0.3194, 0.0872, -0.1096]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, -2.8200e-06, ..., -1.0049e-06, + -2.1216e-06, 0.0000e+00], + [ 0.0000e+00, -1.4901e-08, 1.5832e-08, ..., 3.0734e-08, + -6.5193e-09, 0.0000e+00], + ..., + [ 1.8626e-09, -1.8626e-09, 2.7996e-06, ..., 9.3225e-07, + 2.1197e-06, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, -1.1176e-08, ..., -3.7253e-09, + 5.5879e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 2.1420e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0019, -0.0381, 0.0111, -0.0103, 0.0083, 0.0091, 0.0186, 0.0070, + -0.0489, -0.0149], device='cuda:0'), grad: tensor([ 1.3970e-08, -5.7071e-06, 5.0291e-08, 6.8918e-08, -1.4901e-08, + 1.2107e-08, 8.3819e-09, 5.5060e-06, -4.6566e-09, 7.3574e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 247.60, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4395 re_mapping 0.0039 re_causal 0.0114 /// teacc 99.07 lr 0.00010000 +Epoch 395, weight, value: tensor([[ 0.0211, -0.2014, -0.1909, ..., -0.3833, -0.1304, -0.1753], + [ 0.0890, -0.0986, 0.0451, ..., 0.0475, 0.1419, -0.0644], + [-0.1096, 0.1626, -0.2289, ..., 0.0593, 0.1030, -0.0437], + ..., + [-0.0805, -0.1028, -0.0852, ..., 0.0114, -0.2327, 0.1721], + [ 0.0484, -0.0550, 0.1297, ..., 0.0019, -0.2716, -0.0389], + [-0.2568, -0.1313, -0.1813, ..., -0.3209, 0.0871, -0.1096]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.1420e-08, 1.1176e-08, ..., 0.0000e+00, + 1.3225e-06, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 1.2107e-08, ..., -1.8626e-09, + 3.4459e-08, 0.0000e+00], + [ 0.0000e+00, -8.3819e-09, 9.3132e-10, ..., -2.2352e-08, + -1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 7.4506e-09, 2.7940e-09, ..., 2.0489e-08, + 1.6764e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 3.1665e-08, ..., 9.3132e-10, + 7.0781e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0019, -0.0380, 0.0109, -0.0103, 0.0083, 0.0090, 0.0186, 0.0071, + -0.0490, -0.0150], device='cuda:0'), grad: tensor([ 4.6566e-06, 1.3504e-07, -4.4703e-08, -6.5193e-09, 7.4506e-09, + 2.7381e-07, -5.3123e-06, 4.9360e-08, 2.2911e-07, 1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 247.39, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4676 re_mapping 0.0037 re_causal 0.0111 /// teacc 98.99 lr 0.00010000 +Epoch 396, weight, value: tensor([[ 0.0211, -0.2015, -0.1915, ..., -0.3836, -0.1306, -0.1754], + [ 0.0890, -0.0985, 0.0451, ..., 0.0474, 0.1419, -0.0649], + [-0.1097, 0.1627, -0.2298, ..., 0.0592, 0.1030, -0.0437], + ..., + [-0.0803, -0.1029, -0.0852, ..., 0.0116, -0.2330, 0.1727], + [ 0.0484, -0.0550, 0.1299, ..., 0.0016, -0.2720, -0.0401], + [-0.2571, -0.1316, -0.1804, ..., -0.3220, 0.0875, -0.1096]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 8.7544e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.3039e-08, ..., 1.6764e-07, + 3.5390e-08, 2.4214e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -3.9116e-07, + -1.9930e-07, -3.7253e-08], + ..., + [-1.8626e-09, -0.0000e+00, -1.3039e-08, ..., 9.8720e-08, + 1.8999e-07, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 8.0094e-08, + 1.1735e-07, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 3.7253e-08, 0.0000e+00]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0019, -0.0382, 0.0107, -0.0105, 0.0088, 0.0091, 0.0186, 0.0072, + -0.0493, -0.0149], device='cuda:0'), grad: tensor([ 3.8184e-07, 5.9977e-07, -5.5321e-07, 0.0000e+00, -6.1654e-07, + 9.2760e-07, -1.6037e-06, 1.1176e-07, 4.9733e-07, 2.5146e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 247.31, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4620 re_mapping 0.0038 re_causal 0.0110 /// teacc 98.98 lr 0.00010000 +Epoch 397, weight, value: tensor([[ 0.0205, -0.2015, -0.1925, ..., -0.3846, -0.1312, -0.1758], + [ 0.0888, -0.0984, 0.0451, ..., 0.0474, 0.1422, -0.0652], + [-0.1101, 0.1626, -0.2312, ..., 0.0589, 0.1028, -0.0435], + ..., + [-0.0799, -0.1029, -0.0852, ..., 0.0117, -0.2332, 0.1730], + [ 0.0483, -0.0550, 0.1319, ..., 0.0025, -0.2722, -0.0404], + [-0.2574, -0.1321, -0.1811, ..., -0.3231, 0.0876, -0.1097]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -3.3528e-08, 7.4506e-09, ..., -8.3819e-08, + -1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.6077e-08, 8.9407e-08, ..., 9.8720e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.5146e-07, ..., -8.7544e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-08, ..., 1.6764e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0018, -0.0383, 0.0105, -0.0106, 0.0091, 0.0091, 0.0187, 0.0073, + -0.0482, -0.0150], device='cuda:0'), grad: tensor([ 2.5891e-07, 3.1665e-08, -1.2852e-07, 1.8440e-07, 2.9802e-08, + 7.0781e-08, -2.9430e-07, 3.4645e-07, -6.0722e-07, 1.1921e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 247.50, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4554 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.02 lr 0.00010000 +Epoch 398, weight, value: tensor([[ 0.0205, -0.2016, -0.1934, ..., -0.3854, -0.1314, -0.1759], + [ 0.0888, -0.0985, 0.0450, ..., 0.0472, 0.1421, -0.0653], + [-0.1101, 0.1627, -0.2312, ..., 0.0590, 0.1028, -0.0418], + ..., + [-0.0799, -0.1030, -0.0853, ..., 0.0119, -0.2333, 0.1730], + [ 0.0483, -0.0551, 0.1345, ..., 0.0043, -0.2719, -0.0405], + [-0.2576, -0.1322, -0.1826, ..., -0.3242, 0.0879, -0.1098]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.0617e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 5.4017e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.0117e-07, 0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0016, -0.0386, 0.0106, -0.0071, 0.0092, 0.0062, 0.0187, 0.0073, + -0.0466, -0.0147], device='cuda:0'), grad: tensor([ 4.6194e-07, 2.0489e-08, 4.0978e-08, 2.0489e-08, 8.3819e-08, + 7.4506e-09, -1.1176e-08, 1.8440e-07, 4.2841e-08, -8.5123e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 247.44, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4848 re_mapping 0.0036 re_causal 0.0112 /// teacc 99.08 lr 0.00010000 +Epoch 399, weight, value: tensor([[ 0.0204, -0.2017, -0.1954, ..., -0.3862, -0.1318, -0.1759], + [ 0.0888, -0.0984, 0.0451, ..., 0.0472, 0.1422, -0.0655], + [-0.1102, 0.1627, -0.2318, ..., 0.0589, 0.1028, -0.0417], + ..., + [-0.0800, -0.1030, -0.0854, ..., 0.0119, -0.2335, 0.1732], + [ 0.0483, -0.0552, 0.1352, ..., 0.0046, -0.2721, -0.0405], + [-0.2580, -0.1323, -0.1829, ..., -0.3254, 0.0880, -0.1099]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 1.6764e-08, + -7.4506e-09, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 6.1467e-08, ..., 4.8056e-07, + 0.0000e+00, 3.5390e-08], + ..., + [ 0.0000e+00, -0.0000e+00, -7.2643e-08, ..., -5.5507e-07, + 3.7253e-09, -4.4703e-08], + [ 0.0000e+00, -1.8626e-09, -1.8626e-09, ..., 9.3132e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0015, -0.0386, 0.0104, -0.0071, 0.0092, 0.0062, 0.0187, 0.0074, + -0.0464, -0.0147], device='cuda:0'), grad: tensor([-2.2352e-08, 3.5390e-08, 1.1567e-06, 8.1956e-08, -5.3272e-07, + 1.4901e-08, 3.7253e-09, -1.0859e-06, 4.4703e-08, 3.0175e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 247.42, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4680 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.06 lr 0.00010000 +Epoch 400, weight, value: tensor([[ 0.0204, -0.2016, -0.1959, ..., -0.3869, -0.1325, -0.1778], + [ 0.0888, -0.0984, 0.0451, ..., 0.0471, 0.1424, -0.0662], + [-0.1102, 0.1628, -0.2326, ..., 0.0589, 0.1028, -0.0419], + ..., + [-0.0800, -0.1031, -0.0854, ..., 0.0120, -0.2336, 0.1741], + [ 0.0483, -0.0553, 0.1354, ..., 0.0044, -0.2725, -0.0413], + [-0.2581, -0.1324, -0.1830, ..., -0.3268, 0.0881, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.1176e-08, + 1.3039e-08, 0.0000e+00], + [ 3.1665e-08, 0.0000e+00, -6.5193e-07, ..., -4.7684e-07, + -5.5507e-07, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 2.7940e-08, ..., 3.1665e-08, + 5.0291e-08, 0.0000e+00], + ..., + [ 2.4214e-08, 0.0000e+00, 2.7940e-08, ..., -7.6368e-08, + 9.3132e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.7497e-07, ..., 3.5018e-07, + 4.0233e-07, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 9.3132e-09, ..., 3.9116e-08, + 8.7544e-08, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0010, -0.0387, 0.0104, -0.0070, 0.0089, 0.0064, 0.0184, 0.0076, + -0.0470, -0.0147], device='cuda:0'), grad: tensor([ 9.8720e-08, -1.1809e-06, 1.6205e-07, 3.9674e-07, -5.4203e-07, + 6.7055e-07, 1.5646e-07, -2.5332e-06, 1.6131e-06, 1.1455e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 247.53, cls_loss 0.0009 cls_loss_mapping 0.0009 cls_loss_causal 0.4632 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.05 lr 0.00001000 +Epoch 401, weight, value: tensor([[ 0.0202, -0.2017, -0.1966, ..., -0.3876, -0.1330, -0.1780], + [ 0.0886, -0.0985, 0.0440, ..., 0.0461, 0.1423, -0.0671], + [-0.1109, 0.1629, -0.2332, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0794, -0.1032, -0.0850, ..., 0.0125, -0.2337, 0.1753], + [ 0.0482, -0.0553, 0.1390, ..., 0.0082, -0.2701, -0.0413], + [-0.2582, -0.1325, -0.1833, ..., -0.3275, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[-7.4506e-08, 0.0000e+00, 3.3528e-08, ..., 1.8626e-09, + 1.7323e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-09, + 4.8429e-08, 5.5879e-09], + [ 9.3132e-09, 0.0000e+00, 5.5879e-08, ..., 3.3528e-08, + 3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., -1.1176e-08, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 3.5204e-07, ..., 1.8626e-09, + 2.0117e-06, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0006, -0.0397, 0.0104, -0.0067, 0.0086, 0.0064, 0.0179, 0.0079, + -0.0435, -0.0147], device='cuda:0'), grad: tensor([ 2.5705e-07, 1.6391e-07, 2.1048e-07, -2.0303e-07, 3.7253e-09, + 1.4499e-05, -2.2262e-05, 0.0000e+00, 7.3314e-06, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 247.49, cls_loss 0.0006 cls_loss_mapping 0.0006 cls_loss_causal 0.4681 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.07 lr 0.00001000 +Epoch 402, weight, value: tensor([[ 0.0202, -0.2017, -0.1967, ..., -0.3876, -0.1330, -0.1780], + [ 0.0886, -0.0985, 0.0440, ..., 0.0461, 0.1423, -0.0672], + [-0.1109, 0.1629, -0.2334, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0794, -0.1032, -0.0850, ..., 0.0125, -0.2337, 0.1753], + [ 0.0482, -0.0553, 0.1390, ..., 0.0082, -0.2702, -0.0414], + [-0.2583, -0.1325, -0.1833, ..., -0.3277, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 7.4506e-08, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 9.3132e-09, + 5.5879e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.1176e-07, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.1176e-08, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.3528e-08, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0006, -0.0397, 0.0104, -0.0067, 0.0086, 0.0064, 0.0179, 0.0080, + -0.0435, -0.0148], device='cuda:0'), grad: tensor([ 5.5879e-09, 5.6252e-07, 3.3528e-08, 7.4506e-09, 9.3132e-09, + 4.6566e-08, -3.7253e-09, -8.2888e-07, -5.9605e-08, 2.2165e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 247.57, cls_loss 0.0006 cls_loss_mapping 0.0006 cls_loss_causal 0.4178 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.08 lr 0.00001000 +Epoch 403, weight, value: tensor([[ 0.0202, -0.2017, -0.1967, ..., -0.3877, -0.1331, -0.1780], + [ 0.0886, -0.0985, 0.0440, ..., 0.0461, 0.1423, -0.0672], + [-0.1109, 0.1629, -0.2335, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0794, -0.1032, -0.0850, ..., 0.0126, -0.2337, 0.1753], + [ 0.0482, -0.0553, 0.1390, ..., 0.0082, -0.2702, -0.0414], + [-0.2583, -0.1325, -0.1834, ..., -0.3278, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[-1.3039e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -1.4156e-06, ..., -6.6869e-07, + -1.1977e-06, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 5.5879e-09, ..., 3.7253e-09, + 3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3951e-06, ..., 6.5751e-07, + 1.1828e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 9.3132e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0086, 0.0064, 0.0179, 0.0080, + -0.0435, -0.0147], device='cuda:0'), grad: tensor([-4.8429e-08, -2.9001e-06, 3.5390e-08, 9.3132e-09, 7.4506e-09, + 2.0489e-08, -4.8429e-08, 2.9039e-06, 3.3528e-08, -1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 247.68, cls_loss 0.0007 cls_loss_mapping 0.0006 cls_loss_causal 0.4467 re_mapping 0.0035 re_causal 0.0109 /// teacc 99.10 lr 0.00001000 +Epoch 404, weight, value: tensor([[ 0.0202, -0.2017, -0.1968, ..., -0.3878, -0.1331, -0.1780], + [ 0.0886, -0.0985, 0.0440, ..., 0.0461, 0.1424, -0.0672], + [-0.1110, 0.1629, -0.2336, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0794, -0.1032, -0.0850, ..., 0.0126, -0.2338, 0.1754], + [ 0.0482, -0.0553, 0.1390, ..., 0.0082, -0.2702, -0.0414], + [-0.2583, -0.1325, -0.1834, ..., -0.3279, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [-0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 1.3039e-08, + -4.4703e-08, -1.8626e-09], + [ 0.0000e+00, -3.7253e-09, 7.4506e-09, ..., -1.8626e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 3.3528e-08, ..., 2.4214e-08, + 6.1467e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -7.8231e-08, ..., -4.6566e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -2.0489e-08, 0.0000e+00]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0086, 0.0064, 0.0179, 0.0080, + -0.0435, -0.0147], device='cuda:0'), grad: tensor([ 5.5879e-09, 8.9407e-08, 0.0000e+00, 1.5832e-07, -3.7253e-08, + -1.5832e-07, 3.5390e-08, 1.4342e-07, -1.9185e-07, -5.7742e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 247.22, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4539 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.11 lr 0.00001000 +Epoch 405, weight, value: tensor([[ 0.0203, -0.2018, -0.1968, ..., -0.3879, -0.1331, -0.1780], + [ 0.0886, -0.0984, 0.0440, ..., 0.0460, 0.1424, -0.0672], + [-0.1110, 0.1629, -0.2337, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0794, -0.1032, -0.0850, ..., 0.0126, -0.2338, 0.1754], + [ 0.0482, -0.0553, 0.1390, ..., 0.0082, -0.2703, -0.0414], + [-0.2583, -0.1325, -0.1835, ..., -0.3280, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.3039e-08, ..., 2.6077e-08, + -1.3039e-08, 0.0000e+00], + [ 0.0000e+00, -2.9802e-08, 1.8626e-09, ..., -4.6939e-07, + -8.7544e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.4214e-08, 1.1176e-08, ..., 4.0047e-07, + 9.1270e-08, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, -1.6764e-08, ..., 2.9802e-08, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0086, 0.0064, 0.0179, 0.0080, + -0.0435, -0.0147], device='cuda:0'), grad: tensor([ 7.4506e-08, 5.5879e-08, -1.2238e-06, 1.6764e-08, 2.4214e-08, + -2.4214e-08, -3.5390e-08, 1.0747e-06, 7.4506e-08, -2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 247.12, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4462 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.13 lr 0.00001000 +Epoch 406, weight, value: tensor([[ 0.0203, -0.2018, -0.1969, ..., -0.3879, -0.1332, -0.1780], + [ 0.0886, -0.0984, 0.0440, ..., 0.0461, 0.1424, -0.0673], + [-0.1110, 0.1629, -0.2337, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0126, -0.2339, 0.1755], + [ 0.0481, -0.0553, 0.1390, ..., 0.0082, -0.2703, -0.0414], + [-0.2583, -0.1326, -0.1835, ..., -0.3280, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.4214e-08, ..., -1.8626e-09, + -3.9116e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0086, 0.0064, 0.0179, 0.0080, + -0.0435, -0.0147], device='cuda:0'), grad: tensor([ 8.0094e-08, -1.1176e-07, 3.5390e-08, 3.1665e-08, 1.4901e-08, + -5.9605e-08, -4.6566e-08, 1.1176e-08, 3.5390e-08, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 246.96, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4351 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.14 lr 0.00001000 +Epoch 407, weight, value: tensor([[ 0.0203, -0.2018, -0.1969, ..., -0.3880, -0.1332, -0.1780], + [ 0.0886, -0.0984, 0.0441, ..., 0.0461, 0.1424, -0.0673], + [-0.1110, 0.1629, -0.2338, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0126, -0.2339, 0.1755], + [ 0.0481, -0.0553, 0.1390, ..., 0.0082, -0.2703, -0.0415], + [-0.2584, -0.1326, -0.1835, ..., -0.3281, 0.0881, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., -1.8626e-09, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., -3.7253e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0086, 0.0064, 0.0179, 0.0080, + -0.0435, -0.0147], device='cuda:0'), grad: tensor([-4.6566e-08, 9.3132e-09, 1.6764e-08, -1.1176e-08, -8.0094e-08, + 2.2352e-08, -5.5879e-09, 2.6077e-08, 2.9802e-08, 3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 247.48, cls_loss 0.0006 cls_loss_mapping 0.0006 cls_loss_causal 0.4322 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.12 lr 0.00001000 +Epoch 408, weight, value: tensor([[ 0.0203, -0.2018, -0.1969, ..., -0.3880, -0.1332, -0.1780], + [ 0.0886, -0.0984, 0.0441, ..., 0.0461, 0.1425, -0.0673], + [-0.1110, 0.1629, -0.2338, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0126, -0.2339, 0.1756], + [ 0.0481, -0.0553, 0.1390, ..., 0.0082, -0.2704, -0.0415], + [-0.2584, -0.1326, -0.1835, ..., -0.3282, 0.0882, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -5.5879e-09, + -2.4214e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., -7.4506e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0084, 0.0064, 0.0179, 0.0080, + -0.0436, -0.0146], device='cuda:0'), grad: tensor([-6.7055e-07, 2.9802e-08, -3.5390e-08, -1.9446e-06, 3.7253e-08, + 2.1625e-06, 3.7439e-07, -2.0489e-08, 6.3330e-08, 5.5879e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 247.60, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4629 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.13 lr 0.00001000 +Epoch 409, weight, value: tensor([[ 0.0204, -0.2018, -0.1970, ..., -0.3881, -0.1332, -0.1780], + [ 0.0886, -0.0984, 0.0441, ..., 0.0461, 0.1425, -0.0674], + [-0.1110, 0.1629, -0.2340, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0126, -0.2340, 0.1756], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2704, -0.0414], + [-0.2584, -0.1326, -0.1836, ..., -0.3284, 0.0882, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.1793e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 9.3132e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -4.0978e-08, 0.0000e+00, ..., -5.4017e-08, + -4.6566e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 3.5390e-08, 0.0000e+00, ..., 4.6566e-08, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -1.1176e-08, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0006, -0.0397, 0.0103, -0.0067, 0.0084, 0.0064, 0.0179, 0.0080, + -0.0436, -0.0146], device='cuda:0'), grad: tensor([ 8.6613e-07, 4.6566e-08, -1.2293e-07, 1.6764e-08, 5.7742e-08, + 8.3819e-08, -9.6112e-07, 1.2293e-07, 3.7253e-08, -1.5087e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 247.94, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4565 re_mapping 0.0032 re_causal 0.0106 /// teacc 99.11 lr 0.00001000 +Epoch 410, weight, value: tensor([[ 0.0204, -0.2018, -0.1970, ..., -0.3882, -0.1333, -0.1780], + [ 0.0886, -0.0984, 0.0440, ..., 0.0460, 0.1425, -0.0675], + [-0.1110, 0.1629, -0.2340, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0849, ..., 0.0126, -0.2340, 0.1757], + [ 0.0481, -0.0553, 0.1390, ..., 0.0082, -0.2704, -0.0414], + [-0.2584, -0.1326, -0.1836, ..., -0.3284, 0.0883, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -7.4506e-09, ..., 2.2352e-08, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, -2.0489e-08, -5.5879e-09, ..., -9.6858e-08, + -3.1665e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.6764e-08, 9.3132e-09, ..., 6.1467e-08, + 3.3528e-08, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, -0.0000e+00, ..., -0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -1.5460e-07, 0.0000e+00]], device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0006, -0.0398, 0.0104, -0.0067, 0.0084, 0.0064, 0.0179, 0.0081, + -0.0436, -0.0146], device='cuda:0'), grad: tensor([-1.3039e-08, 1.0245e-07, -2.0303e-07, 1.1176e-08, 3.3528e-07, + 8.0094e-08, -1.8626e-08, 1.0431e-07, 9.3132e-09, -4.3958e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 247.42, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4551 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.13 lr 0.00001000 +Epoch 411, weight, value: tensor([[ 0.0204, -0.2019, -0.1971, ..., -0.3883, -0.1333, -0.1780], + [ 0.0886, -0.0984, 0.0440, ..., 0.0460, 0.1425, -0.0675], + [-0.1111, 0.1629, -0.2340, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0126, -0.2340, 0.1757], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2704, -0.0414], + [-0.2584, -0.1326, -0.1836, ..., -0.3285, 0.0883, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, 7.4506e-09, ..., 5.9605e-08, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, -7.8231e-08, 2.0489e-08, ..., -1.1176e-07, + -9.6858e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 5.0291e-08, 3.7253e-09, ..., 1.8626e-08, + 6.8918e-08, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, 1.8626e-09, ..., 4.2841e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.3039e-08, + -8.5309e-07, 0.0000e+00]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0006, -0.0398, 0.0104, -0.0067, 0.0084, 0.0064, 0.0179, 0.0081, + -0.0436, -0.0146], device='cuda:0'), grad: tensor([ 2.2352e-08, 2.0117e-07, -3.4831e-07, 6.8918e-07, 1.8626e-09, + 2.8275e-06, -9.3132e-09, 6.7055e-08, 2.9802e-07, -3.7700e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 247.35, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4413 re_mapping 0.0031 re_causal 0.0104 /// teacc 99.15 lr 0.00001000 +Epoch 412, weight, value: tensor([[ 0.0204, -0.2019, -0.1971, ..., -0.3883, -0.1333, -0.1780], + [ 0.0886, -0.0984, 0.0441, ..., 0.0460, 0.1425, -0.0675], + [-0.1111, 0.1630, -0.2341, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0126, -0.2340, 0.1758], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2705, -0.0415], + [-0.2584, -0.1326, -0.1837, ..., -0.3286, 0.0883, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., -5.5879e-09, + -1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -1.8626e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0006, -0.0398, 0.0104, -0.0067, 0.0084, 0.0064, 0.0179, 0.0081, + -0.0436, -0.0146], device='cuda:0'), grad: tensor([ 7.4506e-09, -2.4214e-08, 0.0000e+00, 0.0000e+00, 9.3132e-09, + 1.1176e-08, -2.7940e-08, 2.0489e-08, 2.0489e-08, -2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 246.97, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4405 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.16 lr 0.00001000 +Epoch 413, weight, value: tensor([[ 0.0205, -0.2019, -0.1971, ..., -0.3884, -0.1334, -0.1780], + [ 0.0886, -0.0984, 0.0441, ..., 0.0460, 0.1426, -0.0675], + [-0.1111, 0.1630, -0.2341, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0793, -0.1033, -0.0850, ..., 0.0127, -0.2341, 0.1759], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2705, -0.0414], + [-0.2585, -0.1326, -0.1837, ..., -0.3287, 0.0883, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.6950e-07, ..., -1.4529e-07, + -1.6205e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -0.0000e+00, + -1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6205e-07, ..., 1.3784e-07, + 1.4342e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -7.4506e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 7.4506e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0006, -0.0398, 0.0104, -0.0067, 0.0083, 0.0064, 0.0179, 0.0081, + -0.0436, -0.0145], device='cuda:0'), grad: tensor([ 3.3528e-08, -5.0478e-07, 1.8626e-08, 4.8429e-08, -8.7544e-08, + -4.6566e-08, 2.7940e-08, 4.8243e-07, -9.3132e-09, 3.9116e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 247.43, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4170 re_mapping 0.0030 re_causal 0.0101 /// teacc 99.12 lr 0.00001000 +Epoch 414, weight, value: tensor([[ 0.0206, -0.2019, -0.1972, ..., -0.3884, -0.1334, -0.1780], + [ 0.0886, -0.0984, 0.0440, ..., 0.0460, 0.1426, -0.0676], + [-0.1111, 0.1630, -0.2343, ..., 0.0589, 0.1028, -0.0419], + ..., + [-0.0793, -0.1033, -0.0849, ..., 0.0127, -0.2341, 0.1759], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2706, -0.0414], + [-0.2585, -0.1326, -0.1837, ..., -0.3288, 0.0883, -0.1104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.3330e-08, ..., 1.1176e-08, + 5.2154e-08, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, -1.1362e-07, ..., 1.0803e-07, + -7.8231e-08, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 1.8626e-09, ..., 2.2352e-08, + -1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 7.4506e-09, ..., -1.7136e-07, + 9.3132e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 3.1665e-08, ..., 7.4506e-09, + 2.6077e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0006, -0.0398, 0.0104, -0.0067, 0.0083, 0.0064, 0.0179, 0.0081, + -0.0436, -0.0145], device='cuda:0'), grad: tensor([ 2.4773e-07, -1.1176e-07, 8.0094e-08, 4.2841e-08, -3.7253e-09, + 4.8429e-08, 1.6764e-08, -4.6939e-07, 1.1362e-07, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 246.72, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4447 re_mapping 0.0030 re_causal 0.0102 /// teacc 99.15 lr 0.00001000 +Epoch 415, weight, value: tensor([[ 0.0207, -0.2020, -0.1972, ..., -0.3885, -0.1334, -0.1781], + [ 0.0886, -0.0985, 0.0440, ..., 0.0459, 0.1426, -0.0677], + [-0.1112, 0.1630, -0.2343, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1034, -0.0849, ..., 0.0127, -0.2341, 0.1760], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2706, -0.0414], + [-0.2586, -0.1326, -0.1837, ..., -0.3289, 0.0883, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 2.4214e-08, + 2.6077e-08, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, 3.7253e-09, ..., -1.8999e-07, + -2.3469e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 7.0781e-08, + 9.1270e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 3.3528e-08, 0.0000e+00]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0006, -0.0398, 0.0104, -0.0067, 0.0083, 0.0064, 0.0179, 0.0081, + -0.0437, -0.0145], device='cuda:0'), grad: tensor([-7.0781e-08, 7.2643e-08, -5.7369e-07, 1.8626e-07, 2.2352e-08, + 1.1176e-08, 1.4901e-08, 2.2724e-07, 5.5879e-09, 1.0803e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 247.38, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4374 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.17 lr 0.00001000 +Epoch 416, weight, value: tensor([[ 0.0207, -0.2020, -0.1972, ..., -0.3886, -0.1334, -0.1781], + [ 0.0886, -0.0985, 0.0440, ..., 0.0459, 0.1426, -0.0678], + [-0.1112, 0.1630, -0.2344, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1034, -0.0849, ..., 0.0128, -0.2342, 0.1761], + [ 0.0481, -0.0553, 0.1391, ..., 0.0082, -0.2706, -0.0415], + [-0.2586, -0.1327, -0.1838, ..., -0.3290, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, -5.5879e-09, ..., 1.1176e-08, + -0.0000e+00, -2.7940e-09], + [ 9.3132e-10, -7.5437e-08, 8.3819e-09, ..., -1.0431e-07, + -1.4715e-07, 0.0000e+00], + ..., + [ 2.7940e-09, 7.1712e-08, 5.5879e-09, ..., 8.8476e-08, + 1.3690e-07, 1.8626e-09], + [ 2.7940e-09, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 5.5879e-09, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0006, -0.0399, 0.0104, -0.0067, 0.0082, 0.0064, 0.0179, 0.0082, + -0.0437, -0.0145], device='cuda:0'), grad: tensor([ 1.9558e-08, 4.1910e-08, -2.9616e-07, -9.3132e-09, -6.4261e-08, + 2.8871e-08, -1.9558e-08, 2.4680e-07, 9.3132e-09, 4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 247.19, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4373 re_mapping 0.0030 re_causal 0.0100 /// teacc 99.15 lr 0.00001000 +Epoch 417, weight, value: tensor([[ 0.0207, -0.2020, -0.1973, ..., -0.3887, -0.1335, -0.1781], + [ 0.0886, -0.0985, 0.0440, ..., 0.0459, 0.1426, -0.0678], + [-0.1112, 0.1631, -0.2344, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1034, -0.0849, ..., 0.0128, -0.2342, 0.1762], + [ 0.0480, -0.0553, 0.1391, ..., 0.0082, -0.2707, -0.0415], + [-0.2586, -0.1327, -0.1838, ..., -0.3291, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 4.6566e-09, ..., 9.3132e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 3.6322e-08, -1.3970e-08, ..., 2.0582e-07, + -1.9558e-08, 0.0000e+00], + [ 0.0000e+00, 1.2107e-08, 4.6566e-09, ..., 6.9849e-08, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -5.4017e-08, 4.6566e-09, ..., -3.1106e-07, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., -6.5193e-09, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 5.5879e-09, ..., 1.2107e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0006, -0.0399, 0.0104, -0.0067, 0.0082, 0.0064, 0.0179, 0.0082, + -0.0437, -0.0145], device='cuda:0'), grad: tensor([ 3.8184e-08, 5.2527e-07, 1.9558e-07, 4.5635e-08, -9.3132e-10, + 3.7253e-09, 5.5879e-09, -8.3819e-07, -2.0489e-08, 4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 247.22, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4464 re_mapping 0.0029 re_causal 0.0102 /// teacc 99.15 lr 0.00001000 +Epoch 418, weight, value: tensor([[ 0.0207, -0.2021, -0.1973, ..., -0.3889, -0.1335, -0.1781], + [ 0.0886, -0.0985, 0.0440, ..., 0.0459, 0.1427, -0.0678], + [-0.1112, 0.1631, -0.2345, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1034, -0.0849, ..., 0.0128, -0.2342, 0.1762], + [ 0.0480, -0.0554, 0.1391, ..., 0.0082, -0.2707, -0.0415], + [-0.2586, -0.1327, -0.1839, ..., -0.3292, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -7.4506e-09, ..., -3.7253e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [-9.3132e-10, 1.8626e-09, 3.7253e-09, ..., 1.8626e-09, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 9.3132e-10, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, + -3.5390e-08, 0.0000e+00]], device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0006, -0.0399, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0082, + -0.0438, -0.0145], device='cuda:0'), grad: tensor([ 1.1176e-08, 3.5390e-08, -9.3132e-10, -1.8626e-09, 1.2107e-08, + 9.3132e-10, 1.8626e-08, 4.0978e-08, 7.0781e-08, -1.8068e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 247.08, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4177 re_mapping 0.0030 re_causal 0.0100 /// teacc 99.20 lr 0.00001000 +Epoch 419, weight, value: tensor([[ 0.0207, -0.2021, -0.1974, ..., -0.3889, -0.1336, -0.1781], + [ 0.0886, -0.0984, 0.0440, ..., 0.0459, 0.1427, -0.0679], + [-0.1112, 0.1631, -0.2346, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0792, -0.1034, -0.0849, ..., 0.0128, -0.2343, 0.1762], + [ 0.0480, -0.0554, 0.1391, ..., 0.0082, -0.2708, -0.0415], + [-0.2586, -0.1327, -0.1839, ..., -0.3293, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 4.6566e-09, + 2.7940e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6100e-07, ..., -7.2643e-08, + -6.1188e-07, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 3.7253e-09, ..., -2.7940e-09, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.3039e-08, + 4.0978e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-08, ..., 9.3132e-09, + 6.6124e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 1.0245e-08, + 4.8429e-08, 0.0000e+00]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0005, -0.0399, 0.0104, -0.0067, 0.0082, 0.0064, 0.0180, 0.0082, + -0.0438, -0.0145], device='cuda:0'), grad: tensor([ 5.7742e-08, -1.2070e-06, 4.6566e-09, 8.3819e-09, -2.6077e-07, + 1.2852e-07, 9.0338e-07, 1.1083e-07, 1.3597e-07, 1.2852e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 247.71, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4346 re_mapping 0.0029 re_causal 0.0101 /// teacc 99.18 lr 0.00001000 +Epoch 420, weight, value: tensor([[ 0.0207, -0.2021, -0.1974, ..., -0.3890, -0.1336, -0.1781], + [ 0.0886, -0.0984, 0.0440, ..., 0.0459, 0.1428, -0.0679], + [-0.1113, 0.1631, -0.2348, ..., 0.0589, 0.1028, -0.0420], + ..., + [-0.0792, -0.1035, -0.0848, ..., 0.0129, -0.2344, 0.1763], + [ 0.0480, -0.0554, 0.1391, ..., 0.0082, -0.2708, -0.0415], + [-0.2586, -0.1327, -0.1840, ..., -0.3294, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 0.0000e+00, ..., 5.0291e-08, + 1.9558e-08, 0.0000e+00], + [ 0.0000e+00, -1.6764e-08, 2.7940e-09, ..., -2.3283e-08, + -3.5390e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 2.7940e-09, ..., -4.6566e-08, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 3.7253e-09, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 1.8626e-08, + 4.0047e-08, 0.0000e+00]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0005, -0.0399, 0.0104, -0.0067, 0.0082, 0.0064, 0.0180, 0.0083, + -0.0438, -0.0145], device='cuda:0'), grad: tensor([ 4.6566e-09, 1.4901e-07, -7.5437e-08, -2.7940e-08, -1.0338e-07, + 4.6566e-09, 1.8626e-09, -1.8999e-07, 1.7695e-08, 2.2259e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 248.04, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4358 re_mapping 0.0029 re_causal 0.0101 /// teacc 99.18 lr 0.00001000 +Epoch 421, weight, value: tensor([[ 0.0207, -0.2021, -0.1974, ..., -0.3890, -0.1336, -0.1781], + [ 0.0885, -0.0984, 0.0440, ..., 0.0459, 0.1428, -0.0679], + [-0.1113, 0.1631, -0.2349, ..., 0.0588, 0.1028, -0.0419], + ..., + [-0.0792, -0.1035, -0.0848, ..., 0.0129, -0.2344, 0.1763], + [ 0.0480, -0.0554, 0.1391, ..., 0.0082, -0.2709, -0.0416], + [-0.2587, -0.1328, -0.1840, ..., -0.3295, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-08, + 8.3819e-09, 1.8626e-09], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, -3.7253e-09, ..., -5.6811e-08, + 2.7940e-09, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.9558e-08, + 8.3819e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0005, -0.0399, 0.0104, -0.0067, 0.0082, 0.0064, 0.0180, 0.0083, + -0.0438, -0.0145], device='cuda:0'), grad: tensor([ 1.6764e-08, 5.9605e-08, 3.7253e-09, 4.0047e-08, -1.1176e-08, + -4.6566e-09, -5.7742e-08, -1.2759e-07, 6.3330e-08, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 247.52, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4156 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.19 lr 0.00001000 +Epoch 422, weight, value: tensor([[ 0.0207, -0.2021, -0.1975, ..., -0.3891, -0.1336, -0.1781], + [ 0.0885, -0.0984, 0.0440, ..., 0.0459, 0.1429, -0.0680], + [-0.1113, 0.1631, -0.2350, ..., 0.0589, 0.1028, -0.0419], + ..., + [-0.0792, -0.1035, -0.0848, ..., 0.0129, -0.2344, 0.1764], + [ 0.0480, -0.0554, 0.1391, ..., 0.0082, -0.2709, -0.0416], + [-0.2587, -0.1328, -0.1840, ..., -0.3296, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 0.0000e+00, + 9.4995e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 6.5193e-09, + 3.4459e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -8.3819e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.0699e-07, ..., 9.3132e-10, + 1.1958e-06, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 4.6566e-09, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0005, -0.0399, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0083, + -0.0439, -0.0145], device='cuda:0'), grad: tensor([ 3.6974e-07, 1.4435e-07, 8.3819e-09, 9.3132e-10, 1.0245e-08, + 2.8051e-06, -7.8231e-06, -2.7008e-08, 4.4629e-06, 5.2154e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 247.34, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4317 re_mapping 0.0029 re_causal 0.0100 /// teacc 99.19 lr 0.00001000 +Epoch 423, weight, value: tensor([[ 0.0207, -0.2022, -0.1975, ..., -0.3893, -0.1337, -0.1781], + [ 0.0885, -0.0984, 0.0440, ..., 0.0458, 0.1428, -0.0681], + [-0.1113, 0.1632, -0.2350, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1035, -0.0848, ..., 0.0129, -0.2344, 0.1765], + [ 0.0480, -0.0554, 0.1391, ..., 0.0082, -0.2710, -0.0416], + [-0.2587, -0.1328, -0.1841, ..., -0.3297, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -7.3574e-08, ..., -9.3132e-10, + -9.9652e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 0.0000e+00, + 5.0291e-08, 0.0000e+00]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0005, -0.0399, 0.0104, -0.0067, 0.0082, 0.0064, 0.0180, 0.0083, + -0.0439, -0.0145], device='cuda:0'), grad: tensor([-4.6566e-09, -1.6484e-07, 2.2352e-08, -7.4506e-09, -7.6368e-08, + 2.3283e-08, 1.8626e-09, 4.4703e-08, 1.2107e-08, 1.5460e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 247.54, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4206 re_mapping 0.0029 re_causal 0.0101 /// teacc 99.18 lr 0.00001000 +Epoch 424, weight, value: tensor([[ 0.0207, -0.2023, -0.1975, ..., -0.3894, -0.1337, -0.1781], + [ 0.0885, -0.0984, 0.0440, ..., 0.0458, 0.1428, -0.0682], + [-0.1113, 0.1632, -0.2351, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1035, -0.0848, ..., 0.0130, -0.2344, 0.1766], + [ 0.0479, -0.0555, 0.1391, ..., 0.0081, -0.2710, -0.0416], + [-0.2587, -0.1328, -0.1841, ..., -0.3298, 0.0884, -0.1105]], + device='cuda:0'), grad: tensor([[-1.9558e-08, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, + 9.3132e-09, 9.3132e-10], + [ 1.1176e-08, 0.0000e+00, -5.5879e-09, ..., -6.5193e-09, + -1.5832e-08, -1.1176e-08], + [ 0.0000e+00, -9.3132e-10, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 0.0000e+00, 2.7940e-09, ..., -9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -1.5832e-08, ..., -1.8626e-09, + 1.0245e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0005, -0.0400, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0083, + -0.0439, -0.0146], device='cuda:0'), grad: tensor([-9.5926e-08, 5.2154e-08, 9.3132e-09, 8.3819e-09, 1.6764e-08, + 2.6077e-08, -8.5682e-08, 2.1420e-08, 7.4506e-09, 4.5635e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 247.56, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4277 re_mapping 0.0029 re_causal 0.0102 /// teacc 99.16 lr 0.00001000 +Epoch 425, weight, value: tensor([[ 0.0207, -0.2023, -0.1976, ..., -0.3895, -0.1338, -0.1781], + [ 0.0885, -0.0984, 0.0440, ..., 0.0458, 0.1429, -0.0682], + [-0.1113, 0.1632, -0.2351, ..., 0.0589, 0.1029, -0.0419], + ..., + [-0.0792, -0.1036, -0.0848, ..., 0.0130, -0.2345, 0.1766], + [ 0.0479, -0.0555, 0.1391, ..., 0.0082, -0.2711, -0.0416], + [-0.2587, -0.1329, -0.1842, ..., -0.3299, 0.0884, -0.1106]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, -4.6566e-09, ..., 2.6077e-08, + 8.3819e-09, 9.3132e-10], + [ 0.0000e+00, -1.3039e-08, 9.3132e-10, ..., -3.3528e-08, + -3.0734e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 6.5193e-09, -1.1176e-08, ..., -3.1572e-07, + 2.0489e-08, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.3970e-08, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0005, -0.0400, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0083, + -0.0439, -0.0146], device='cuda:0'), grad: tensor([ 2.7940e-09, 5.9605e-08, -8.2888e-08, 7.5996e-07, 6.5193e-09, + 2.7940e-09, 3.7253e-09, -7.8790e-07, 3.4459e-08, -1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 247.31, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4075 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.15 lr 0.00001000 +Epoch 426, weight, value: tensor([[ 0.0207, -0.2023, -0.1976, ..., -0.3895, -0.1338, -0.1781], + [ 0.0885, -0.0984, 0.0439, ..., 0.0457, 0.1429, -0.0682], + [-0.1113, 0.1632, -0.2352, ..., 0.0589, 0.1029, -0.0418], + ..., + [-0.0792, -0.1036, -0.0848, ..., 0.0130, -0.2345, 0.1766], + [ 0.0479, -0.0555, 0.1391, ..., 0.0082, -0.2711, -0.0416], + [-0.2587, -0.1329, -0.1842, ..., -0.3300, 0.0884, -0.1106]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -5.8673e-08, -3.1758e-07, ..., -2.4121e-07, + -3.8184e-07, 0.0000e+00], + [ 0.0000e+00, 1.0245e-08, 9.2201e-08, ..., 7.8231e-08, + 8.1956e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 4.8429e-08, 2.4773e-07, ..., 1.7695e-07, + 2.9802e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -5.1223e-08, ..., -3.9116e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0005, -0.0400, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0084, + -0.0440, -0.0146], device='cuda:0'), grad: tensor([ 7.4506e-09, -5.2620e-07, 2.3656e-07, 1.3039e-08, 8.3819e-09, + 1.4435e-07, -6.7987e-08, 3.7160e-07, -1.9744e-07, 2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 247.42, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4208 re_mapping 0.0029 re_causal 0.0101 /// teacc 99.16 lr 0.00001000 +Epoch 427, weight, value: tensor([[ 0.0207, -0.2023, -0.1976, ..., -0.3896, -0.1339, -0.1781], + [ 0.0885, -0.0984, 0.0439, ..., 0.0457, 0.1429, -0.0682], + [-0.1113, 0.1632, -0.2352, ..., 0.0589, 0.1029, -0.0418], + ..., + [-0.0792, -0.1036, -0.0847, ..., 0.0130, -0.2345, 0.1766], + [ 0.0479, -0.0555, 0.1392, ..., 0.0082, -0.2712, -0.0415], + [-0.2587, -0.1329, -0.1843, ..., -0.3302, 0.0884, -0.1106]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.6566e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7008e-08, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.4901e-08, 0.0000e+00]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0005, -0.0401, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0084, + -0.0440, -0.0146], device='cuda:0'), grad: tensor([ 9.3132e-10, 6.5193e-09, 9.3132e-09, 6.5193e-08, -4.0047e-08, + -1.9372e-07, 4.4703e-08, -8.3819e-09, 9.3132e-08, 3.7253e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 246.54, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4443 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.16 lr 0.00001000 +Epoch 428, weight, value: tensor([[ 0.0207, -0.2023, -0.1976, ..., -0.3896, -0.1339, -0.1781], + [ 0.0885, -0.0984, 0.0439, ..., 0.0457, 0.1429, -0.0682], + [-0.1114, 0.1633, -0.2353, ..., 0.0589, 0.1029, -0.0418], + ..., + [-0.0792, -0.1036, -0.0847, ..., 0.0131, -0.2346, 0.1766], + [ 0.0479, -0.0555, 0.1392, ..., 0.0082, -0.2713, -0.0415], + [-0.2587, -0.1329, -0.1844, ..., -0.3303, 0.0884, -0.1106]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, -1.0245e-08, 0.0000e+00, ..., -2.6077e-08, + -1.3970e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 6.5193e-09, 1.8626e-09, ..., 1.7695e-08, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -7.4506e-09, ..., -3.7253e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0005, -0.0401, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0084, + -0.0440, -0.0146], device='cuda:0'), grad: tensor([-1.8626e-09, 4.6566e-09, -7.4506e-08, 3.6322e-08, 9.3132e-10, + -6.5193e-09, -5.5879e-09, 4.8429e-08, -2.7940e-09, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 247.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4204 re_mapping 0.0028 re_causal 0.0102 /// teacc 99.19 lr 0.00001000 +Epoch 429, weight, value: tensor([[ 0.0207, -0.2023, -0.1976, ..., -0.3897, -0.1339, -0.1781], + [ 0.0885, -0.0984, 0.0439, ..., 0.0456, 0.1429, -0.0683], + [-0.1114, 0.1633, -0.2353, ..., 0.0589, 0.1029, -0.0418], + ..., + [-0.0792, -0.1036, -0.0847, ..., 0.0131, -0.2346, 0.1767], + [ 0.0479, -0.0555, 0.1392, ..., 0.0082, -0.2713, -0.0415], + [-0.2587, -0.1330, -0.1844, ..., -0.3304, 0.0884, -0.1106]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 2.7940e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-08, -3.4459e-08, ..., 2.5146e-08, + 2.5146e-08, 0.0000e+00], + [ 0.0000e+00, -4.0047e-08, 9.3132e-10, ..., -5.6811e-08, + -9.1270e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -0.0000e+00, 1.0245e-08, ..., -4.6566e-09, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 1.7695e-08, 6.5193e-09, ..., 2.4214e-08, + 2.9802e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + -2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0005, -0.0401, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0085, + -0.0440, -0.0146], device='cuda:0'), grad: tensor([ 2.7940e-08, 0.0000e+00, -1.8720e-07, 5.5879e-09, 2.7008e-08, + 6.5193e-09, 2.6077e-08, -5.5879e-09, 9.0338e-08, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 247.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4292 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.19 lr 0.00001000 +Epoch 430, weight, value: tensor([[ 0.0207, -0.2024, -0.1977, ..., -0.3898, -0.1340, -0.1782], + [ 0.0885, -0.0984, 0.0439, ..., 0.0456, 0.1429, -0.0683], + [-0.1114, 0.1633, -0.2354, ..., 0.0589, 0.1029, -0.0418], + ..., + [-0.0792, -0.1036, -0.0847, ..., 0.0131, -0.2346, 0.1767], + [ 0.0479, -0.0555, 0.1392, ..., 0.0082, -0.2713, -0.0415], + [-0.2587, -0.1330, -0.1844, ..., -0.3306, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -8.6613e-08, ..., -1.5460e-07, + -3.6135e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.0978e-08, ..., 7.3574e-08, + 1.6671e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.5635e-08, ..., 5.4017e-08, + 1.8533e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4214e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0005, -0.0401, 0.0104, -0.0067, 0.0083, 0.0064, 0.0180, 0.0085, + -0.0440, -0.0147], device='cuda:0'), grad: tensor([ 2.7940e-09, -6.2212e-07, 2.9616e-07, 1.3039e-08, 2.7940e-09, + -5.4017e-08, 7.4506e-09, 2.0675e-07, 3.9116e-08, 1.0617e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 247.57, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4407 re_mapping 0.0028 re_causal 0.0102 /// teacc 99.18 lr 0.00001000 +Epoch 431, weight, value: tensor([[ 0.0208, -0.2024, -0.1977, ..., -0.3898, -0.1340, -0.1782], + [ 0.0885, -0.0984, 0.0439, ..., 0.0456, 0.1430, -0.0683], + [-0.1114, 0.1633, -0.2355, ..., 0.0589, 0.1029, -0.0418], + ..., + [-0.0792, -0.1037, -0.0848, ..., 0.0131, -0.2346, 0.1767], + [ 0.0479, -0.0556, 0.1392, ..., 0.0082, -0.2713, -0.0415], + [-0.2587, -0.1330, -0.1845, ..., -0.3307, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-09, + 8.3819e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 6.5658e-07, + -4.7497e-08, 3.3900e-07], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 5.6811e-08, + 3.1665e-08, 1.3970e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -8.8569e-07, + 1.8626e-09, -4.4145e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 6.7055e-08, + 1.8626e-09, 3.2596e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.7300e-08, + 9.3132e-10, 3.8184e-08]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0005, -0.0401, 0.0104, -0.0067, 0.0083, 0.0064, 0.0181, 0.0085, + -0.0440, -0.0147], device='cuda:0'), grad: tensor([-4.6659e-07, 2.7977e-06, 2.5891e-07, 1.2387e-07, 2.0489e-08, + 3.9116e-08, 2.1420e-07, -3.7197e-06, 3.0082e-07, 4.3586e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 247.02, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4007 re_mapping 0.0028 re_causal 0.0097 /// teacc 99.17 lr 0.00001000 +Epoch 432, weight, value: tensor([[ 0.0208, -0.2024, -0.1977, ..., -0.3898, -0.1341, -0.1782], + [ 0.0885, -0.0985, 0.0440, ..., 0.0456, 0.1430, -0.0683], + [-0.1115, 0.1634, -0.2355, ..., 0.0589, 0.1030, -0.0418], + ..., + [-0.0791, -0.1037, -0.0848, ..., 0.0132, -0.2347, 0.1767], + [ 0.0478, -0.0556, 0.1393, ..., 0.0082, -0.2714, -0.0415], + [-0.2588, -0.1331, -0.1846, ..., -0.3309, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 0.0000e+00, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0005, -0.0401, 0.0105, -0.0067, 0.0084, 0.0064, 0.0181, 0.0085, + -0.0440, -0.0147], device='cuda:0'), grad: tensor([ 1.3039e-08, 1.1176e-08, 9.3132e-10, 2.7940e-09, -7.4506e-09, + -1.0245e-08, -1.5832e-08, -3.7253e-09, -5.0291e-08, 6.6124e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 247.14, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4138 re_mapping 0.0028 re_causal 0.0097 /// teacc 99.18 lr 0.00001000 +Epoch 433, weight, value: tensor([[ 0.0208, -0.2024, -0.1978, ..., -0.3899, -0.1341, -0.1782], + [ 0.0885, -0.0985, 0.0440, ..., 0.0456, 0.1429, -0.0684], + [-0.1115, 0.1634, -0.2355, ..., 0.0590, 0.1030, -0.0418], + ..., + [-0.0791, -0.1037, -0.0848, ..., 0.0132, -0.2347, 0.1768], + [ 0.0478, -0.0556, 0.1393, ..., 0.0082, -0.2714, -0.0415], + [-0.2588, -0.1331, -0.1846, ..., -0.3310, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, -0.0000e+00]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0005, -0.0402, 0.0105, -0.0067, 0.0084, 0.0064, 0.0181, 0.0085, + -0.0441, -0.0147], device='cuda:0'), grad: tensor([-1.1176e-08, 1.1176e-08, 3.5390e-08, 7.4506e-09, 1.2107e-08, + 4.6566e-09, -5.8673e-08, 4.6566e-09, 2.4214e-08, -2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 247.94, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4491 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.16 lr 0.00001000 +Epoch 434, weight, value: tensor([[ 0.0208, -0.2024, -0.1978, ..., -0.3899, -0.1342, -0.1782], + [ 0.0885, -0.0985, 0.0440, ..., 0.0456, 0.1430, -0.0684], + [-0.1115, 0.1634, -0.2355, ..., 0.0590, 0.1030, -0.0418], + ..., + [-0.0791, -0.1037, -0.0848, ..., 0.0132, -0.2347, 0.1768], + [ 0.0478, -0.0556, 0.1393, ..., 0.0082, -0.2714, -0.0415], + [-0.2588, -0.1331, -0.1847, ..., -0.3311, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.0978e-08, ..., 1.4156e-07, + 1.2107e-08, 9.3132e-10], + [ 0.0000e+00, -6.5193e-09, 2.7940e-09, ..., -4.9360e-08, + -9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 6.5193e-09, -4.0978e-08, ..., -1.0245e-07, + 1.6764e-08, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.0245e-08, + 5.9605e-08, 0.0000e+00]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0005, -0.0402, 0.0105, -0.0067, 0.0084, 0.0064, 0.0181, 0.0085, + -0.0441, -0.0147], device='cuda:0'), grad: tensor([ 2.2352e-08, 5.4203e-07, -2.0582e-07, -3.7253e-09, -2.8498e-07, + 1.3039e-08, -8.7544e-08, -2.9989e-07, 6.6124e-08, 2.4214e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 247.69, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4407 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.17 lr 0.00001000 +Epoch 435, weight, value: tensor([[ 0.0208, -0.2024, -0.1978, ..., -0.3900, -0.1342, -0.1782], + [ 0.0884, -0.0985, 0.0440, ..., 0.0456, 0.1430, -0.0685], + [-0.1115, 0.1635, -0.2356, ..., 0.0590, 0.1030, -0.0418], + ..., + [-0.0791, -0.1037, -0.0848, ..., 0.0131, -0.2348, 0.1769], + [ 0.0478, -0.0556, 0.1393, ..., 0.0082, -0.2715, -0.0415], + [-0.2588, -0.1332, -0.1847, ..., -0.3313, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.3411e-07, ..., -9.9652e-08, + -1.2107e-07, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, 1.8626e-09, ..., -6.5193e-09, + -1.3970e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 1.2666e-07, ..., 1.0151e-07, + 1.2293e-07, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0005, -0.0401, 0.0105, -0.0067, 0.0084, 0.0064, 0.0181, 0.0085, + -0.0441, -0.0147], device='cuda:0'), grad: tensor([ 2.6356e-07, -3.4552e-07, -2.2352e-08, 3.7253e-09, 2.2352e-08, + 1.8626e-09, -3.0454e-07, 3.3621e-07, 4.2841e-08, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 246.83, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3939 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.16 lr 0.00001000 +Epoch 436, weight, value: tensor([[ 0.0208, -0.2025, -0.1979, ..., -0.3901, -0.1343, -0.1782], + [ 0.0884, -0.0986, 0.0440, ..., 0.0456, 0.1430, -0.0685], + [-0.1115, 0.1635, -0.2357, ..., 0.0590, 0.1031, -0.0417], + ..., + [-0.0792, -0.1038, -0.0848, ..., 0.0131, -0.2348, 0.1769], + [ 0.0478, -0.0556, 0.1393, ..., 0.0082, -0.2715, -0.0415], + [-0.2588, -0.1332, -0.1848, ..., -0.3314, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 7.9162e-09, -2.7940e-09, ..., 1.2107e-08, + 1.5832e-08, -0.0000e+00], + [ 0.0000e+00, -1.3504e-08, 1.3970e-09, ..., -2.0023e-08, + -3.3528e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 3.2596e-09, 1.8626e-09, ..., 5.5879e-09, + 9.7789e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, -3.7253e-09, ..., -5.1223e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0004, -0.0402, 0.0106, -0.0067, 0.0084, 0.0064, 0.0181, 0.0085, + -0.0441, -0.0148], device='cuda:0'), grad: tensor([ 5.1223e-09, 4.2375e-08, -7.5903e-08, 6.0536e-09, 5.5879e-09, + 9.3132e-10, 0.0000e+00, 2.0489e-08, -6.0536e-09, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 246.99, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4225 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.18 lr 0.00001000 +Epoch 437, weight, value: tensor([[ 0.0208, -0.2025, -0.1980, ..., -0.3901, -0.1343, -0.1782], + [ 0.0884, -0.0985, 0.0441, ..., 0.0456, 0.1431, -0.0685], + [-0.1116, 0.1636, -0.2358, ..., 0.0590, 0.1031, -0.0417], + ..., + [-0.0791, -0.1038, -0.0849, ..., 0.0131, -0.2349, 0.1770], + [ 0.0478, -0.0557, 0.1393, ..., 0.0082, -0.2716, -0.0415], + [-0.2589, -0.1332, -0.1848, ..., -0.3316, 0.0884, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 7.4506e-09, 4.6566e-10, ..., 4.1910e-09, + 6.9849e-09, 0.0000e+00], + [ 2.2352e-08, 3.4925e-08, 3.2596e-09, ..., 3.1199e-08, + 8.4750e-08, 0.0000e+00], + [ 9.3132e-10, -1.8813e-07, 2.3283e-09, ..., -9.7789e-08, + -1.4715e-07, 0.0000e+00], + ..., + [ 4.6566e-09, 9.3132e-08, -4.6566e-10, ..., 3.1199e-08, + 9.5926e-08, 0.0000e+00], + [-4.6566e-10, 1.3970e-08, -1.3039e-08, ..., 4.6566e-10, + 1.0710e-08, 0.0000e+00], + [ 2.3283e-09, 5.1223e-09, 4.6566e-10, ..., 5.5879e-09, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0004, -0.0401, 0.0106, -0.0067, 0.0084, 0.0064, 0.0181, 0.0085, + -0.0442, -0.0148], device='cuda:0'), grad: tensor([ 2.3749e-08, 2.5099e-07, -4.4890e-07, 6.3330e-08, -1.6019e-07, + 1.2107e-08, 7.9162e-09, 2.1933e-07, 1.5367e-08, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 436---------------------------------------------------- +epoch 436, time 264.40, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4227 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.21 lr 0.00001000 +Epoch 438, weight, value: tensor([[ 0.0208, -0.2025, -0.1980, ..., -0.3902, -0.1344, -0.1782], + [ 0.0884, -0.0985, 0.0441, ..., 0.0456, 0.1432, -0.0685], + [-0.1116, 0.1636, -0.2360, ..., 0.0590, 0.1031, -0.0417], + ..., + [-0.0791, -0.1039, -0.0849, ..., 0.0132, -0.2350, 0.1770], + [ 0.0478, -0.0557, 0.1393, ..., 0.0082, -0.2716, -0.0415], + [-0.2589, -0.1332, -0.1849, ..., -0.3318, 0.0883, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., -8.8476e-09, + -1.9558e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., 4.6566e-09, + 2.3283e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 8.3819e-09, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 5.1223e-09, + 6.9849e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0004, -0.0402, 0.0105, -0.0067, 0.0085, 0.0064, 0.0181, 0.0085, + -0.0442, -0.0148], device='cuda:0'), grad: tensor([ 1.0710e-08, -2.9802e-08, 1.0291e-07, -4.5635e-08, 1.0943e-07, + 1.0245e-08, -2.1467e-07, 3.1665e-08, 2.6077e-08, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 437---------------------------------------------------- +epoch 437, time 263.05, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4222 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.22 lr 0.00001000 +Epoch 439, weight, value: tensor([[ 0.0208, -0.2026, -0.1981, ..., -0.3902, -0.1345, -0.1782], + [ 0.0884, -0.0986, 0.0441, ..., 0.0456, 0.1432, -0.0686], + [-0.1116, 0.1637, -0.2360, ..., 0.0590, 0.1032, -0.0417], + ..., + [-0.0791, -0.1039, -0.0849, ..., 0.0131, -0.2351, 0.1770], + [ 0.0478, -0.0557, 0.1393, ..., 0.0082, -0.2717, -0.0415], + [-0.2589, -0.1333, -0.1850, ..., -0.3319, 0.0883, -0.1107]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 2.5611e-08, 9.3132e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-09, + 1.6764e-08, 3.2596e-09], + [ 0.0000e+00, 1.8626e-09, 2.0023e-08, ..., 1.9092e-08, + 9.7789e-09, 3.2596e-09], + ..., + [ 0.0000e+00, -4.1910e-09, 2.8871e-08, ..., -2.7940e-09, + 4.0513e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.1910e-09, + 8.1491e-08, 5.1223e-09], + [ 0.0000e+00, 9.3132e-10, 1.3970e-09, ..., 6.0536e-09, + -1.1884e-06, 0.0000e+00]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0004, -0.0402, 0.0106, -0.0067, 0.0085, 0.0064, 0.0181, 0.0085, + -0.0442, -0.0149], device='cuda:0'), grad: tensor([ 8.9873e-08, 8.1025e-08, 9.7789e-08, -1.2154e-07, 3.9674e-06, + 6.9384e-08, -8.6613e-08, 1.4855e-07, 3.1758e-07, -4.5486e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 246.89, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4241 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.22 lr 0.00001000 +Epoch 440, weight, value: tensor([[ 0.0208, -0.2026, -0.1981, ..., -0.3904, -0.1345, -0.1783], + [ 0.0884, -0.0986, 0.0441, ..., 0.0456, 0.1432, -0.0686], + [-0.1116, 0.1637, -0.2361, ..., 0.0590, 0.1032, -0.0417], + ..., + [-0.0791, -0.1039, -0.0849, ..., 0.0132, -0.2351, 0.1771], + [ 0.0477, -0.0557, 0.1393, ..., 0.0082, -0.2718, -0.0416], + [-0.2589, -0.1333, -0.1851, ..., -0.3320, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 3.6787e-08, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 8.3819e-09, ..., 4.6566e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -2.7474e-08, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-09, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 3.2596e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0004, -0.0402, 0.0106, -0.0067, 0.0085, 0.0064, 0.0181, 0.0085, + -0.0443, -0.0149], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.4342e-07, 1.6764e-08, -7.0315e-08, 9.7789e-09, + 1.0710e-08, -1.8161e-08, -1.2526e-07, 2.3283e-08, 1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 247.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4265 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.19 lr 0.00001000 +Epoch 441, weight, value: tensor([[ 0.0208, -0.2026, -0.1981, ..., -0.3904, -0.1346, -0.1783], + [ 0.0884, -0.0986, 0.0442, ..., 0.0456, 0.1432, -0.0686], + [-0.1116, 0.1637, -0.2361, ..., 0.0590, 0.1032, -0.0417], + ..., + [-0.0791, -0.1039, -0.0849, ..., 0.0132, -0.2352, 0.1771], + [ 0.0477, -0.0558, 0.1393, ..., 0.0081, -0.2718, -0.0416], + [-0.2589, -0.1333, -0.1851, ..., -0.3322, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, -1.8626e-09, ..., -0.0000e+00, + -3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 1.3970e-09, + -4.6566e-10, 0.0000e+00], + ..., + [ 2.0955e-08, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 1.8626e-09, 0.0000e+00], + [ 1.1642e-08, -1.3970e-09, -3.2596e-09, ..., -2.3283e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -6.0536e-09, 0.0000e+00]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0004, -0.0402, 0.0107, -0.0067, 0.0086, 0.0064, 0.0181, 0.0085, + -0.0444, -0.0149], device='cuda:0'), grad: tensor([ 4.6566e-09, -2.7940e-09, 3.7253e-09, 3.2596e-08, 1.2573e-08, + -1.1036e-07, 6.5193e-09, 5.8673e-08, 2.2352e-08, -1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 247.15, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4143 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.18 lr 0.00001000 +Epoch 442, weight, value: tensor([[ 0.0208, -0.2027, -0.1982, ..., -0.3905, -0.1346, -0.1783], + [ 0.0884, -0.0986, 0.0441, ..., 0.0455, 0.1432, -0.0686], + [-0.1116, 0.1637, -0.2362, ..., 0.0591, 0.1033, -0.0417], + ..., + [-0.0791, -0.1040, -0.0849, ..., 0.0132, -0.2352, 0.1771], + [ 0.0477, -0.0558, 0.1393, ..., 0.0081, -0.2719, -0.0416], + [-0.2589, -0.1334, -0.1851, ..., -0.3324, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 3.7253e-09, ..., 1.3970e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, -1.3970e-09, 2.1420e-08, ..., -2.7940e-09, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3970e-09, 6.5193e-09, ..., 2.3283e-09, + 2.7008e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.2072e-07, ..., -2.1886e-08, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 4.6566e-10, + -2.5146e-08, 0.0000e+00]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0004, -0.0403, 0.0107, -0.0067, 0.0086, 0.0064, 0.0181, 0.0085, + -0.0444, -0.0150], device='cuda:0'), grad: tensor([ 2.7940e-09, 1.2107e-08, 3.9116e-08, 1.0664e-07, 5.0291e-08, + 6.9849e-09, 8.9407e-08, 1.1409e-07, -3.2410e-07, -8.2888e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 247.82, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4271 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.17 lr 0.00001000 +Epoch 443, weight, value: tensor([[ 0.0208, -0.2027, -0.1982, ..., -0.3905, -0.1347, -0.1783], + [ 0.0884, -0.0986, 0.0441, ..., 0.0455, 0.1432, -0.0686], + [-0.1117, 0.1638, -0.2362, ..., 0.0591, 0.1033, -0.0417], + ..., + [-0.0791, -0.1040, -0.0849, ..., 0.0132, -0.2353, 0.1771], + [ 0.0477, -0.0558, 0.1394, ..., 0.0082, -0.2720, -0.0416], + [-0.2589, -0.1334, -0.1852, ..., -0.3325, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 2.3283e-09, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 4.6566e-10, ..., -3.2596e-09, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-09, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0004, -0.0403, 0.0108, -0.0067, 0.0086, 0.0064, 0.0182, 0.0085, + -0.0444, -0.0150], device='cuda:0'), grad: tensor([ 4.0047e-08, 1.0710e-08, -6.0536e-09, 2.7940e-09, 7.9162e-09, + 1.3970e-08, -7.8231e-08, 1.0245e-08, 1.8626e-08, -1.6298e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 247.84, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4329 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.17 lr 0.00001000 +Epoch 444, weight, value: tensor([[ 0.0208, -0.2027, -0.1983, ..., -0.3906, -0.1348, -0.1783], + [ 0.0884, -0.0986, 0.0441, ..., 0.0455, 0.1432, -0.0686], + [-0.1117, 0.1638, -0.2364, ..., 0.0591, 0.1033, -0.0417], + ..., + [-0.0791, -0.1040, -0.0849, ..., 0.0132, -0.2353, 0.1772], + [ 0.0477, -0.0558, 0.1394, ..., 0.0081, -0.2721, -0.0416], + [-0.2589, -0.1334, -0.1852, ..., -0.3326, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.1910e-09, 0.0000e+00]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0004, -0.0403, 0.0108, -0.0067, 0.0086, 0.0064, 0.0182, 0.0086, + -0.0445, -0.0150], device='cuda:0'), grad: tensor([ 4.1910e-09, 6.9849e-09, 3.2596e-09, 7.9162e-09, -1.9558e-08, + 3.7253e-09, -1.4435e-08, -9.7789e-09, 4.1910e-09, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 246.88, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4280 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.15 lr 0.00001000 +Epoch 445, weight, value: tensor([[ 0.0209, -0.2027, -0.1983, ..., -0.3906, -0.1349, -0.1783], + [ 0.0884, -0.0988, 0.0441, ..., 0.0455, 0.1432, -0.0687], + [-0.1117, 0.1639, -0.2364, ..., 0.0591, 0.1035, -0.0417], + ..., + [-0.0792, -0.1041, -0.0849, ..., 0.0132, -0.2354, 0.1772], + [ 0.0477, -0.0558, 0.1394, ..., 0.0082, -0.2721, -0.0416], + [-0.2590, -0.1334, -0.1853, ..., -0.3327, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 5.5879e-09, + 2.5611e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.5856e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 4.6566e-10, ..., -1.3970e-09, + -1.8626e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, -4.6566e-10, ..., -7.4040e-08, + 1.5367e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + -2.8405e-08, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0004, -0.0404, 0.0109, -0.0067, 0.0086, 0.0064, 0.0182, 0.0086, + -0.0445, -0.0150], device='cuda:0'), grad: tensor([ 1.1502e-07, 1.4668e-07, -7.4506e-09, 1.1129e-07, 1.3504e-08, + -3.8650e-08, -2.3283e-09, -2.8964e-07, 1.1642e-08, -5.3085e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 247.07, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4224 re_mapping 0.0027 re_causal 0.0094 /// teacc 99.17 lr 0.00001000 +Epoch 446, weight, value: tensor([[ 0.0209, -0.2027, -0.1984, ..., -0.3907, -0.1349, -0.1783], + [ 0.0884, -0.0988, 0.0441, ..., 0.0455, 0.1432, -0.0687], + [-0.1117, 0.1640, -0.2365, ..., 0.0592, 0.1035, -0.0417], + ..., + [-0.0792, -0.1041, -0.0849, ..., 0.0132, -0.2356, 0.1772], + [ 0.0477, -0.0559, 0.1395, ..., 0.0082, -0.2722, -0.0415], + [-0.2590, -0.1335, -0.1854, ..., -0.3329, 0.0883, -0.1108]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.6566e-09, + 1.3970e-09, -2.5379e-07], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + -1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0004, -0.0404, 0.0109, -0.0067, 0.0086, 0.0064, 0.0182, 0.0086, + -0.0445, -0.0150], device='cuda:0'), grad: tensor([-2.7940e-09, 6.0536e-09, 5.1223e-09, -4.1444e-07, 6.3842e-07, + 3.9348e-07, 1.3970e-09, -6.1886e-07, -1.3970e-09, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 247.42, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4477 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.16 lr 0.00001000 +Epoch 447, weight, value: tensor([[ 0.0209, -0.2028, -0.1984, ..., -0.3908, -0.1350, -0.1783], + [ 0.0884, -0.0988, 0.0442, ..., 0.0455, 0.1433, -0.0687], + [-0.1117, 0.1641, -0.2366, ..., 0.0592, 0.1036, -0.0417], + ..., + [-0.0792, -0.1042, -0.0849, ..., 0.0132, -0.2357, 0.1773], + [ 0.0477, -0.0559, 0.1395, ..., 0.0082, -0.2723, -0.0415], + [-0.2590, -0.1335, -0.1855, ..., -0.3331, 0.0884, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, -1.3970e-09, ..., 2.7940e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, -9.7789e-09, 1.8626e-08, ..., -4.3772e-08, + -2.0489e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 7.9162e-09, 9.3132e-09, ..., 4.1444e-08, + 2.5611e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0004, -0.0404, 0.0110, -0.0068, 0.0085, 0.0064, 0.0182, 0.0085, + -0.0445, -0.0149], device='cuda:0'), grad: tensor([ 5.5879e-09, 2.5611e-08, -6.4261e-08, -6.8452e-08, -4.4238e-08, + 6.9849e-09, 9.3132e-10, 1.2619e-07, 6.5193e-09, 1.5367e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 247.53, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4031 re_mapping 0.0027 re_causal 0.0095 /// teacc 99.16 lr 0.00001000 +Epoch 448, weight, value: tensor([[ 0.0209, -0.2028, -0.1985, ..., -0.3908, -0.1351, -0.1783], + [ 0.0884, -0.0989, 0.0442, ..., 0.0454, 0.1434, -0.0687], + [-0.1117, 0.1642, -0.2366, ..., 0.0592, 0.1036, -0.0417], + ..., + [-0.0792, -0.1042, -0.0850, ..., 0.0132, -0.2359, 0.1773], + [ 0.0477, -0.0559, 0.1395, ..., 0.0082, -0.2724, -0.0415], + [-0.2590, -0.1335, -0.1856, ..., -0.3332, 0.0884, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 3.7253e-09, + 4.1910e-09, 0.0000e+00], + [ 0.0000e+00, -2.7008e-08, 4.6566e-10, ..., -7.9628e-08, + -8.6147e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.6077e-08, 9.3132e-10, ..., 7.7765e-08, + 8.4750e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0004, -0.0404, 0.0110, -0.0068, 0.0085, 0.0064, 0.0182, 0.0085, + -0.0446, -0.0149], device='cuda:0'), grad: tensor([ 2.3283e-09, 1.3504e-08, -2.6776e-07, 0.0000e+00, -3.7253e-09, + 0.0000e+00, -2.7940e-09, 2.6356e-07, 0.0000e+00, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 246.83, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4356 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.18 lr 0.00001000 +Epoch 449, weight, value: tensor([[ 0.0209, -0.2028, -0.1985, ..., -0.3909, -0.1351, -0.1783], + [ 0.0883, -0.0989, 0.0442, ..., 0.0454, 0.1434, -0.0687], + [-0.1117, 0.1642, -0.2368, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1043, -0.0849, ..., 0.0133, -0.2359, 0.1773], + [ 0.0476, -0.0559, 0.1396, ..., 0.0082, -0.2724, -0.0415], + [-0.2591, -0.1335, -0.1857, ..., -0.3333, 0.0885, -0.1109]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 3.2596e-09, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -5.5879e-09, + 4.6566e-10, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-09, 2.3283e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0004, -0.0404, 0.0110, -0.0068, 0.0085, 0.0064, 0.0182, 0.0085, + -0.0446, -0.0149], device='cuda:0'), grad: tensor([ 3.3062e-08, 2.1420e-08, 8.3819e-09, 5.5879e-09, 1.0245e-08, + 2.7940e-08, -1.2433e-07, -2.8871e-08, 3.7719e-08, 1.8161e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 247.39, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4368 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.18 lr 0.00001000 +Epoch 450, weight, value: tensor([[ 0.0210, -0.2028, -0.1986, ..., -0.3909, -0.1352, -0.1783], + [ 0.0883, -0.0989, 0.0443, ..., 0.0455, 0.1436, -0.0688], + [-0.1117, 0.1642, -0.2369, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1043, -0.0851, ..., 0.0131, -0.2362, 0.1773], + [ 0.0476, -0.0559, 0.1396, ..., 0.0082, -0.2725, -0.0415], + [-0.2591, -0.1336, -0.1858, ..., -0.3335, 0.0885, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 6.3796e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., -7.7300e-08, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.1176e-08, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0004, -0.0403, 0.0111, -0.0068, 0.0085, 0.0064, 0.0183, 0.0084, + -0.0446, -0.0149], device='cuda:0'), grad: tensor([ 3.2596e-09, 2.3562e-07, 9.7789e-09, 1.5832e-08, 1.3970e-09, + -1.3225e-07, 9.2667e-08, -2.8126e-07, 1.6298e-08, 3.9581e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 248.01, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4057 re_mapping 0.0027 re_causal 0.0095 /// teacc 99.17 lr 0.00001000 +Epoch 451, weight, value: tensor([[ 0.0210, -0.2028, -0.1986, ..., -0.3910, -0.1353, -0.1783], + [ 0.0883, -0.0989, 0.0444, ..., 0.0455, 0.1436, -0.0688], + [-0.1117, 0.1642, -0.2371, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1044, -0.0851, ..., 0.0132, -0.2362, 0.1774], + [ 0.0476, -0.0559, 0.1396, ..., 0.0082, -0.2726, -0.0415], + [-0.2591, -0.1336, -0.1858, ..., -0.3337, 0.0885, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 1.9092e-08, ..., 1.0012e-07, + 7.4506e-09, 3.3528e-08], + [ 0.0000e+00, -3.0268e-08, 9.3132e-10, ..., -4.8894e-08, + -4.5169e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.5611e-08, -1.9092e-08, ..., -8.7079e-08, + 3.9116e-08, -3.5856e-08], + [ 0.0000e+00, 1.8626e-09, -9.3132e-10, ..., 6.0536e-09, + 9.3132e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3749e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0004, -0.0403, 0.0110, -0.0068, 0.0085, 0.0064, 0.0183, 0.0084, + -0.0447, -0.0149], device='cuda:0'), grad: tensor([-2.4820e-07, 4.0932e-07, -1.4342e-07, 2.6077e-08, -8.3819e-09, + 5.6811e-08, 5.3085e-08, -3.2922e-07, 4.8894e-08, 1.4389e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 247.85, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4301 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.16 lr 0.00001000 +Epoch 452, weight, value: tensor([[ 0.0210, -0.2028, -0.1987, ..., -0.3911, -0.1353, -0.1783], + [ 0.0883, -0.0988, 0.0444, ..., 0.0455, 0.1437, -0.0689], + [-0.1117, 0.1643, -0.2373, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1045, -0.0851, ..., 0.0132, -0.2364, 0.1775], + [ 0.0476, -0.0560, 0.1396, ..., 0.0082, -0.2726, -0.0415], + [-0.2591, -0.1337, -0.1859, ..., -0.3339, 0.0885, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 4.6566e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 7.9162e-08, ..., 9.6392e-08, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, -1.3970e-09, 3.7253e-09, ..., 3.2596e-09, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -4.6566e-10, -8.8941e-08, ..., -1.1083e-07, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -6.0536e-09, ..., -4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 6.9849e-09, ..., 8.3819e-09, + -4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0004, -0.0404, 0.0110, -0.0068, 0.0086, 0.0064, 0.0183, 0.0085, + -0.0447, -0.0150], device='cuda:0'), grad: tensor([ 7.4506e-09, 3.5530e-07, 5.1223e-09, 2.6543e-08, 1.8626e-09, + -1.2573e-08, 3.2596e-09, -4.0419e-07, -9.7789e-09, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 247.26, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4272 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.14 lr 0.00001000 +Epoch 453, weight, value: tensor([[ 0.0210, -0.2028, -0.1987, ..., -0.3911, -0.1354, -0.1783], + [ 0.0883, -0.0988, 0.0443, ..., 0.0454, 0.1438, -0.0689], + [-0.1118, 0.1643, -0.2374, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1045, -0.0850, ..., 0.0133, -0.2364, 0.1775], + [ 0.0476, -0.0560, 0.1397, ..., 0.0082, -0.2727, -0.0415], + [-0.2591, -0.1337, -0.1860, ..., -0.3341, 0.0885, -0.1109]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 2.7940e-09, ..., 4.6566e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.1176e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.3504e-08, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.1420e-08, ..., 4.6566e-10, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 1.3970e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0004, -0.0404, 0.0110, -0.0068, 0.0086, 0.0064, 0.0183, 0.0085, + -0.0447, -0.0149], device='cuda:0'), grad: tensor([ 1.5367e-08, 5.4482e-08, 8.8476e-09, -1.6298e-08, -2.3283e-09, + 6.5193e-08, -8.0559e-08, -4.8894e-08, -8.1956e-08, 9.6858e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 247.84, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4112 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.14 lr 0.00001000 +Epoch 454, weight, value: tensor([[ 0.0210, -0.2028, -0.1988, ..., -0.3912, -0.1355, -0.1783], + [ 0.0882, -0.0988, 0.0444, ..., 0.0454, 0.1438, -0.0689], + [-0.1118, 0.1644, -0.2375, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1046, -0.0851, ..., 0.0132, -0.2365, 0.1775], + [ 0.0476, -0.0560, 0.1397, ..., 0.0082, -0.2727, -0.0415], + [-0.2591, -0.1337, -0.1861, ..., -0.3343, 0.0886, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 1.9558e-08, ..., 6.9849e-09, + 2.0023e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -1.2619e-07, ..., -1.6298e-08, + -1.2014e-07, 0.0000e+00], + [ 0.0000e+00, -6.0536e-09, 2.2817e-08, ..., -1.3039e-08, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-09, 6.2864e-08, ..., 7.4506e-09, + 7.4506e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 3.7253e-09, + 1.2573e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., 1.0245e-08, + 6.9849e-09, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0004, -0.0404, 0.0111, -0.0068, 0.0086, 0.0064, 0.0183, 0.0085, + -0.0447, -0.0149], device='cuda:0'), grad: tensor([-4.0978e-08, -3.2084e-07, 3.1665e-08, 1.1642e-08, 2.3283e-09, + -2.0489e-08, 2.4214e-08, 1.7183e-07, 6.8918e-08, 8.0094e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 247.17, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4440 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.17 lr 0.00001000 +Epoch 455, weight, value: tensor([[ 0.0211, -0.2028, -0.1990, ..., -0.3913, -0.1357, -0.1783], + [ 0.0882, -0.0988, 0.0444, ..., 0.0455, 0.1439, -0.0689], + [-0.1119, 0.1644, -0.2377, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0792, -0.1046, -0.0851, ..., 0.0133, -0.2366, 0.1775], + [ 0.0475, -0.0560, 0.1398, ..., 0.0082, -0.2729, -0.0415], + [-0.2592, -0.1338, -0.1862, ..., -0.3344, 0.0886, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.3970e-09, + 6.0536e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4435e-08, ..., 9.3132e-10, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 3.7253e-09, ..., 1.8626e-09, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 5.1223e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 3.3062e-08, 0.0000e+00]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0003, -0.0404, 0.0110, -0.0068, 0.0087, 0.0065, 0.0183, 0.0085, + -0.0448, -0.0149], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.9116e-08, 2.5611e-08, -5.4482e-08, -3.1758e-07, + 2.3283e-08, -3.2131e-08, 3.0268e-08, 2.6077e-08, 2.6589e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 247.38, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4366 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.15 lr 0.00001000 +Epoch 456, weight, value: tensor([[ 0.0211, -0.2028, -0.1991, ..., -0.3914, -0.1358, -0.1783], + [ 0.0882, -0.0988, 0.0445, ..., 0.0455, 0.1441, -0.0689], + [-0.1119, 0.1644, -0.2379, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0791, -0.1047, -0.0852, ..., 0.0132, -0.2369, 0.1775], + [ 0.0475, -0.0560, 0.1398, ..., 0.0083, -0.2731, -0.0415], + [-0.2592, -0.1338, -0.1863, ..., -0.3347, 0.0886, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 2.0955e-09, + 4.8894e-09, 0.0000e+00], + [ 0.0000e+00, 7.9162e-08, 2.4820e-07, ..., 2.1067e-06, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.7695e-08, ..., 7.3574e-08, + -8.1956e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -8.2888e-08, -2.8219e-07, ..., -2.2389e-06, + 3.0035e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.1642e-09, + 4.8894e-09, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, 6.9849e-10, ..., 1.2107e-08, + -8.4285e-08, 0.0000e+00]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0003, -0.0403, 0.0110, -0.0068, 0.0088, 0.0065, 0.0184, 0.0084, + -0.0449, -0.0150], device='cuda:0'), grad: tensor([ 1.4435e-08, 3.9116e-06, 5.6112e-08, 8.3819e-08, 2.6752e-07, + 6.0536e-09, 1.6997e-08, -4.1425e-06, 1.0477e-08, -2.2841e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 247.40, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4391 re_mapping 0.0026 re_causal 0.0097 /// teacc 99.15 lr 0.00001000 +Epoch 457, weight, value: tensor([[ 0.0211, -0.2028, -0.1992, ..., -0.3914, -0.1359, -0.1784], + [ 0.0881, -0.0987, 0.0445, ..., 0.0455, 0.1442, -0.0689], + [-0.1119, 0.1645, -0.2381, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0791, -0.1047, -0.0852, ..., 0.0133, -0.2370, 0.1775], + [ 0.0475, -0.0560, 0.1399, ..., 0.0083, -0.2731, -0.0415], + [-0.2592, -0.1338, -0.1864, ..., -0.3349, 0.0887, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 4.6566e-08, ..., 1.2480e-07, + -1.4435e-08, -4.6566e-10], + [ 0.0000e+00, -9.3132e-10, 6.0536e-09, ..., 7.9162e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, -8.3353e-08, ..., -1.9791e-07, + 1.6298e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 8.8476e-09, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.6298e-08, ..., 3.4925e-08, + -4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0003, -0.0404, 0.0110, -0.0068, 0.0087, 0.0065, 0.0184, 0.0085, + -0.0449, -0.0149], device='cuda:0'), grad: tensor([ 4.1910e-09, 4.6846e-07, 3.0268e-08, 9.4529e-08, 9.3132e-10, + -2.7940e-09, -9.3132e-10, -7.6089e-07, 3.7719e-08, 1.4063e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 247.29, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4462 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.19 lr 0.00001000 +Epoch 458, weight, value: tensor([[ 0.0211, -0.2029, -0.1992, ..., -0.3916, -0.1361, -0.1784], + [ 0.0881, -0.0987, 0.0444, ..., 0.0453, 0.1443, -0.0689], + [-0.1120, 0.1645, -0.2381, ..., 0.0593, 0.1037, -0.0417], + ..., + [-0.0790, -0.1048, -0.0850, ..., 0.0134, -0.2371, 0.1776], + [ 0.0475, -0.0561, 0.1399, ..., 0.0082, -0.2733, -0.0415], + [-0.2592, -0.1339, -0.1866, ..., -0.3352, 0.0887, -0.1109]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., -2.7940e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 1.8626e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0003, -0.0405, 0.0110, -0.0069, 0.0087, 0.0065, 0.0185, 0.0086, + -0.0450, -0.0149], device='cuda:0'), grad: tensor([ 5.5879e-08, 1.1642e-08, 6.0536e-09, 8.8476e-09, 1.1642e-08, + 3.3528e-08, -4.0513e-08, 4.7963e-08, 9.4529e-08, -2.2724e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 247.51, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4454 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.17 lr 0.00001000 +Epoch 459, weight, value: tensor([[ 0.0211, -0.2029, -0.1993, ..., -0.3917, -0.1361, -0.1784], + [ 0.0881, -0.0987, 0.0444, ..., 0.0453, 0.1444, -0.0689], + [-0.1121, 0.1646, -0.2382, ..., 0.0592, 0.1037, -0.0417], + ..., + [-0.0790, -0.1048, -0.0851, ..., 0.0134, -0.2372, 0.1776], + [ 0.0475, -0.0561, 0.1399, ..., 0.0082, -0.2734, -0.0415], + [-0.2592, -0.1339, -0.1866, ..., -0.3353, 0.0887, -0.1109]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 5.5879e-09, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0003, -0.0405, 0.0110, -0.0069, 0.0087, 0.0065, 0.0185, 0.0086, + -0.0451, -0.0150], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.6566e-09, 2.1420e-08, -3.4459e-08, 0.0000e+00, + 1.5832e-08, -2.0489e-08, 5.5879e-09, 9.3132e-10, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 247.28, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4235 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.16 lr 0.00001000 +Epoch 460, weight, value: tensor([[ 0.0211, -0.2029, -0.1994, ..., -0.3918, -0.1362, -0.1784], + [ 0.0881, -0.0987, 0.0445, ..., 0.0453, 0.1444, -0.0690], + [-0.1121, 0.1647, -0.2383, ..., 0.0593, 0.1038, -0.0417], + ..., + [-0.0790, -0.1049, -0.0851, ..., 0.0135, -0.2373, 0.1777], + [ 0.0475, -0.0561, 0.1400, ..., 0.0082, -0.2735, -0.0415], + [-0.2592, -0.1340, -0.1867, ..., -0.3356, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -1.8626e-09, + -1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -2.0489e-08, 0.0000e+00]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0003, -0.0405, 0.0110, -0.0069, 0.0087, 0.0065, 0.0185, 0.0086, + -0.0451, -0.0150], device='cuda:0'), grad: tensor([ 6.4261e-08, 5.5879e-09, 2.7940e-09, -2.7940e-09, 3.2596e-08, + -2.0489e-08, -3.6322e-08, 2.7008e-08, 1.5832e-08, -8.2888e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 246.92, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4179 re_mapping 0.0026 re_causal 0.0097 /// teacc 99.18 lr 0.00001000 +Epoch 461, weight, value: tensor([[ 0.0211, -0.2030, -0.1995, ..., -0.3919, -0.1363, -0.1784], + [ 0.0881, -0.0987, 0.0445, ..., 0.0453, 0.1444, -0.0691], + [-0.1121, 0.1647, -0.2385, ..., 0.0593, 0.1038, -0.0416], + ..., + [-0.0790, -0.1049, -0.0851, ..., 0.0135, -0.2374, 0.1777], + [ 0.0475, -0.0562, 0.1400, ..., 0.0082, -0.2736, -0.0415], + [-0.2592, -0.1340, -0.1868, ..., -0.3358, 0.0888, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -8.3819e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0003, -0.0406, 0.0110, -0.0069, 0.0088, 0.0065, 0.0185, 0.0087, + -0.0452, -0.0150], device='cuda:0'), grad: tensor([ 9.3132e-10, -2.7940e-09, 1.5832e-08, -3.7253e-09, 0.0000e+00, + 2.7940e-09, 0.0000e+00, -1.6764e-08, 4.6566e-09, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 247.06, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4392 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.17 lr 0.00001000 +Epoch 462, weight, value: tensor([[ 0.0211, -0.2030, -0.1996, ..., -0.3920, -0.1363, -0.1784], + [ 0.0881, -0.0987, 0.0445, ..., 0.0453, 0.1446, -0.0691], + [-0.1121, 0.1647, -0.2386, ..., 0.0593, 0.1038, -0.0416], + ..., + [-0.0790, -0.1050, -0.0851, ..., 0.0135, -0.2375, 0.1777], + [ 0.0475, -0.0562, 0.1401, ..., 0.0083, -0.2737, -0.0415], + [-0.2593, -0.1340, -0.1870, ..., -0.3360, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -2.0489e-08, ..., -0.0000e+00, + -1.3039e-08, 0.0000e+00], + [ 0.0000e+00, -2.4214e-08, 1.1176e-08, ..., -4.0978e-08, + -1.2107e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7008e-08, 2.7940e-08, ..., 5.5879e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, -5.5879e-09, -2.0489e-08, ..., -1.1176e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0003, -0.0406, 0.0110, -0.0069, 0.0088, 0.0065, 0.0185, 0.0087, + -0.0452, -0.0150], device='cuda:0'), grad: tensor([ 8.3819e-09, -5.4017e-08, -7.3574e-08, -3.0734e-08, 9.3132e-10, + 1.9558e-08, 4.6566e-09, 1.6391e-07, -3.7253e-08, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 247.09, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4306 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.18 lr 0.00001000 +Epoch 463, weight, value: tensor([[ 0.0211, -0.2030, -0.1997, ..., -0.3920, -0.1365, -0.1784], + [ 0.0881, -0.0986, 0.0445, ..., 0.0453, 0.1447, -0.0691], + [-0.1121, 0.1647, -0.2389, ..., 0.0592, 0.1037, -0.0416], + ..., + [-0.0790, -0.1050, -0.0851, ..., 0.0136, -0.2376, 0.1777], + [ 0.0475, -0.0562, 0.1402, ..., 0.0083, -0.2739, -0.0415], + [-0.2593, -0.1341, -0.1873, ..., -0.3364, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -4.6566e-09, 0.0000e+00, ..., -1.1176e-08, + 1.4901e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7940e-09, 1.8626e-09, ..., 8.3819e-09, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0003, -0.0405, 0.0109, -0.0069, 0.0087, 0.0065, 0.0185, 0.0087, + -0.0453, -0.0151], device='cuda:0'), grad: tensor([ 1.1176e-08, 6.5193e-09, 5.6811e-08, 1.5832e-08, 1.5832e-08, + -1.8626e-09, -1.3784e-07, 1.9558e-08, 1.2107e-08, 5.5879e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 246.92, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4380 re_mapping 0.0026 re_causal 0.0097 /// teacc 99.15 lr 0.00001000 +Epoch 464, weight, value: tensor([[ 0.0211, -0.2030, -0.1998, ..., -0.3921, -0.1367, -0.1784], + [ 0.0881, -0.0987, 0.0445, ..., 0.0451, 0.1447, -0.0691], + [-0.1121, 0.1649, -0.2389, ..., 0.0592, 0.1038, -0.0416], + ..., + [-0.0791, -0.1050, -0.0850, ..., 0.0137, -0.2377, 0.1777], + [ 0.0474, -0.0562, 0.1402, ..., 0.0083, -0.2741, -0.0415], + [-0.2593, -0.1341, -0.1873, ..., -0.3366, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, -7.4506e-07, -3.0324e-06, ..., -4.1611e-06, + -3.8631e-06, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 7.4506e-08, ..., 1.0151e-07, + 7.7300e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 7.3109e-07, 2.9393e-06, ..., 4.0308e-06, + 3.7681e-06, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, -3.7253e-09, ..., -1.8626e-09, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-09, + -3.1665e-08, 0.0000e+00]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0002, -0.0407, 0.0110, -0.0069, 0.0088, 0.0065, 0.0186, 0.0088, + -0.0454, -0.0152], device='cuda:0'), grad: tensor([ 4.7497e-08, -8.2180e-06, 2.0955e-07, 3.8184e-08, 4.1910e-08, + 1.4901e-08, -5.2154e-08, 7.9498e-06, 3.1665e-08, -6.9849e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 246.96, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4226 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.18 lr 0.00001000 +Epoch 465, weight, value: tensor([[ 0.0211, -0.2031, -0.1999, ..., -0.3922, -0.1368, -0.1784], + [ 0.0881, -0.0985, 0.0446, ..., 0.0453, 0.1449, -0.0691], + [-0.1122, 0.1649, -0.2391, ..., 0.0592, 0.1038, -0.0416], + ..., + [-0.0790, -0.1053, -0.0851, ..., 0.0136, -0.2380, 0.1778], + [ 0.0475, -0.0562, 0.1403, ..., 0.0083, -0.2742, -0.0415], + [-0.2593, -0.1342, -0.1874, ..., -0.3367, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, 9.3132e-10, ..., -9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.5832e-08, 2.7940e-09, ..., 2.0489e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, -2.7008e-08, -1.1176e-08, ..., -4.0047e-08, + -3.0734e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 9.3132e-10, ..., 4.6566e-09, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 9.3132e-10, ..., 7.4506e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0002, -0.0406, 0.0109, -0.0069, 0.0088, 0.0065, 0.0186, 0.0088, + -0.0454, -0.0152], device='cuda:0'), grad: tensor([-3.7253e-08, 6.7987e-08, -1.1176e-07, 2.6077e-08, 5.5879e-09, + 0.0000e+00, 2.7940e-09, 4.4703e-08, 2.1420e-08, -1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 247.19, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4305 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.16 lr 0.00001000 +Epoch 466, weight, value: tensor([[ 0.0211, -0.2031, -0.2000, ..., -0.3923, -0.1370, -0.1784], + [ 0.0881, -0.0984, 0.0446, ..., 0.0452, 0.1451, -0.0691], + [-0.1122, 0.1649, -0.2394, ..., 0.0591, 0.1038, -0.0416], + ..., + [-0.0790, -0.1054, -0.0851, ..., 0.0137, -0.2381, 0.1778], + [ 0.0474, -0.0563, 0.1403, ..., 0.0083, -0.2743, -0.0415], + [-0.2593, -0.1343, -0.1875, ..., -0.3370, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-0.0000e+00, -9.3132e-10, -1.5832e-08, ..., -1.1176e-08, + -2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 6.5193e-09, + -0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.8626e-09, 1.2107e-08, ..., 1.8626e-09, + 1.8626e-08, 0.0000e+00], + [-5.5879e-09, 0.0000e+00, -5.5879e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0001, -0.0406, 0.0109, -0.0069, 0.0088, 0.0065, 0.0186, 0.0088, + -0.0455, -0.0152], device='cuda:0'), grad: tensor([ 4.5635e-08, -2.0489e-08, 2.4214e-08, 4.6566e-09, 0.0000e+00, + 4.6566e-09, 4.6566e-09, -3.5390e-08, -5.2154e-08, 3.7253e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 247.01, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4238 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.16 lr 0.00001000 +Epoch 467, weight, value: tensor([[ 0.0211, -0.2031, -0.2001, ..., -0.3924, -0.1372, -0.1784], + [ 0.0881, -0.0982, 0.0447, ..., 0.0453, 0.1452, -0.0692], + [-0.1122, 0.1649, -0.2395, ..., 0.0591, 0.1038, -0.0416], + ..., + [-0.0790, -0.1056, -0.0852, ..., 0.0136, -0.2383, 0.1779], + [ 0.0474, -0.0563, 0.1405, ..., 0.0084, -0.2744, -0.0415], + [-0.2593, -0.1343, -0.1876, ..., -0.3371, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 3.7253e-09, ..., 2.0489e-08, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 9.3132e-10, ..., 2.4214e-08, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.7497e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 467, bias, value: tensor([ 7.9802e-05, -4.0509e-02, 1.0798e-02, -6.9628e-03, 8.8327e-03, + 6.5426e-03, 1.8661e-02, 8.7374e-03, -4.5480e-02, -1.5290e-02], + device='cuda:0'), grad: tensor([ 9.3132e-09, 1.1642e-07, 1.1735e-07, 4.6566e-09, -3.5390e-08, + -4.8429e-08, 2.3283e-08, -2.3749e-07, 1.1176e-08, 4.0047e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 248.17, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4230 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.17 lr 0.00001000 +Epoch 468, weight, value: tensor([[ 0.0212, -0.2031, -0.2001, ..., -0.3925, -0.1373, -0.1784], + [ 0.0881, -0.0982, 0.0449, ..., 0.0454, 0.1454, -0.0692], + [-0.1122, 0.1650, -0.2397, ..., 0.0591, 0.1038, -0.0415], + ..., + [-0.0790, -0.1057, -0.0854, ..., 0.0135, -0.2385, 0.1779], + [ 0.0474, -0.0563, 0.1406, ..., 0.0084, -0.2745, -0.0415], + [-0.2593, -0.1343, -0.1877, ..., -0.3374, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, -5.5879e-09, 9.3132e-10, ..., -1.5832e-08, + -1.9558e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 6.5193e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 468, bias, value: tensor([ 5.8610e-05, -4.0401e-02, 1.0800e-02, -7.0061e-03, 8.8884e-03, + 6.5748e-03, 1.8683e-02, 8.6304e-03, -4.5473e-02, -1.5358e-02], + device='cuda:0'), grad: tensor([ 4.6566e-09, 7.4506e-09, -5.3085e-08, 4.6566e-09, 1.8626e-09, + 0.0000e+00, 1.8626e-09, 1.6764e-08, 5.5879e-09, 1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 247.74, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4198 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.16 lr 0.00001000 +Epoch 469, weight, value: tensor([[ 0.0212, -0.2032, -0.2002, ..., -0.3927, -0.1375, -0.1784], + [ 0.0881, -0.0981, 0.0450, ..., 0.0455, 0.1455, -0.0693], + [-0.1122, 0.1650, -0.2398, ..., 0.0591, 0.1038, -0.0415], + ..., + [-0.0790, -0.1058, -0.0854, ..., 0.0135, -0.2386, 0.1779], + [ 0.0474, -0.0563, 0.1406, ..., 0.0085, -0.2746, -0.0415], + [-0.2593, -0.1344, -0.1879, ..., -0.3377, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, 1.8626e-09, ..., -3.7253e-09, + 6.4261e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7940e-09, 1.8626e-09, ..., 6.5193e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 1.3970e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 469, bias, value: tensor([ 5.1839e-05, -4.0382e-02, 1.0822e-02, -7.0342e-03, 8.8915e-03, + 6.6138e-03, 1.8662e-02, 8.6278e-03, -4.5544e-02, -1.5432e-02], + device='cuda:0'), grad: tensor([ 1.6764e-08, 2.8871e-08, 1.4063e-07, -7.4506e-09, 7.6368e-08, + 8.3819e-09, -2.9802e-07, 1.7695e-08, 2.9802e-08, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 247.51, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4395 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.15 lr 0.00001000 +Epoch 470, weight, value: tensor([[ 0.0212, -0.2032, -0.2003, ..., -0.3927, -0.1376, -0.1784], + [ 0.0881, -0.0982, 0.0449, ..., 0.0454, 0.1456, -0.0693], + [-0.1122, 0.1651, -0.2399, ..., 0.0591, 0.1039, -0.0415], + ..., + [-0.0790, -0.1058, -0.0854, ..., 0.0136, -0.2387, 0.1780], + [ 0.0474, -0.0564, 0.1407, ..., 0.0085, -0.2747, -0.0415], + [-0.2593, -0.1344, -0.1881, ..., -0.3379, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -5.5879e-09, + -1.2107e-08, 0.0000e+00], + [ 0.0000e+00, -2.7940e-09, 1.8626e-09, ..., -5.5879e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 1.1176e-08, ..., 9.3132e-09, + 1.1176e-08, 0.0000e+00], + [-0.0000e+00, -9.3132e-10, -1.3039e-08, ..., -5.5879e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([ 3.2656e-05, -4.0442e-02, 1.0837e-02, -7.0698e-03, 8.9484e-03, + 6.6815e-03, 1.8622e-02, 8.7021e-03, -4.5662e-02, -1.5547e-02], + device='cuda:0'), grad: tensor([ 7.4506e-09, -3.0734e-08, -6.5193e-09, 1.4901e-08, 4.6566e-09, + 1.4901e-08, -7.4506e-09, 3.4459e-08, -3.0734e-08, -2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 247.42, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.18 lr 0.00001000 +Epoch 471, weight, value: tensor([[ 0.0212, -0.2032, -0.2003, ..., -0.3928, -0.1376, -0.1784], + [ 0.0881, -0.0983, 0.0448, ..., 0.0452, 0.1457, -0.0694], + [-0.1122, 0.1652, -0.2400, ..., 0.0591, 0.1039, -0.0415], + ..., + [-0.0790, -0.1058, -0.0853, ..., 0.0138, -0.2389, 0.1782], + [ 0.0474, -0.0564, 0.1407, ..., 0.0085, -0.2748, -0.0415], + [-0.2594, -0.1345, -0.1882, ..., -0.3382, 0.0886, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 471, bias, value: tensor([ 6.4606e-05, -4.0588e-02, 1.0845e-02, -7.0959e-03, 8.9410e-03, + 6.6904e-03, 1.8641e-02, 8.8803e-03, -4.5703e-02, -1.5671e-02], + device='cuda:0'), grad: tensor([ 0.0000e+00, 9.3132e-09, 7.4506e-09, 5.3085e-08, -1.9558e-08, + -5.6811e-08, 1.0245e-08, 4.6566e-09, -2.6077e-08, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 247.23, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4382 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.18 lr 0.00001000 +Epoch 472, weight, value: tensor([[ 0.0212, -0.2033, -0.2004, ..., -0.3930, -0.1378, -0.1785], + [ 0.0881, -0.0983, 0.0449, ..., 0.0452, 0.1458, -0.0694], + [-0.1123, 0.1653, -0.2401, ..., 0.0592, 0.1040, -0.0415], + ..., + [-0.0789, -0.1058, -0.0853, ..., 0.0138, -0.2390, 0.1782], + [ 0.0474, -0.0564, 0.1408, ..., 0.0085, -0.2750, -0.0415], + [-0.2594, -0.1345, -0.1884, ..., -0.3385, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.9605e-08, ..., 8.3819e-09, + 3.0734e-08, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, -1.0245e-07, ..., 1.9558e-08, + -3.5390e-08, 0.0000e+00], + [ 0.0000e+00, -6.5193e-09, 7.4506e-09, ..., -4.1910e-08, + -2.6077e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-09, ..., 1.5832e-08, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.6322e-08, ..., 1.0245e-08, + 1.8626e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 472, bias, value: tensor([ 2.1036e-05, -4.0563e-02, 1.0875e-02, -7.1199e-03, 8.9023e-03, + 6.7147e-03, 1.8649e-02, 8.8817e-03, -4.5842e-02, -1.5727e-02], + device='cuda:0'), grad: tensor([ 1.3877e-07, -2.0955e-07, -5.1223e-08, -3.4459e-08, -3.1665e-08, + -6.4261e-08, 6.1467e-08, 4.4703e-08, 1.2387e-07, 3.5390e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 247.25, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4239 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.19 lr 0.00001000 +Epoch 473, weight, value: tensor([[ 0.0212, -0.2033, -0.2005, ..., -0.3931, -0.1380, -0.1785], + [ 0.0881, -0.0983, 0.0450, ..., 0.0453, 0.1459, -0.0694], + [-0.1123, 0.1654, -0.2401, ..., 0.0592, 0.1040, -0.0415], + ..., + [-0.0789, -0.1059, -0.0854, ..., 0.0137, -0.2392, 0.1782], + [ 0.0474, -0.0564, 0.1408, ..., 0.0085, -0.2751, -0.0415], + [-0.2594, -0.1346, -0.1885, ..., -0.3386, 0.0887, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 7.4506e-09, + 3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -1.0245e-08, + 2.7940e-09, -3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 2.1420e-08, ..., 1.1176e-08, + 9.3132e-10, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + -9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 473, bias, value: tensor([ 3.3948e-05, -4.0497e-02, 1.0873e-02, -7.1374e-03, 8.8745e-03, + 6.7238e-03, 1.8661e-02, 8.8219e-03, -4.5903e-02, -1.5720e-02], + device='cuda:0'), grad: tensor([-6.5193e-09, 6.5193e-08, 4.0978e-08, -7.1712e-08, -4.0978e-08, + 1.7695e-08, 1.8626e-09, -8.5682e-08, 5.4948e-08, 3.0734e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 247.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4149 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.17 lr 0.00001000 +Epoch 474, weight, value: tensor([[ 0.0212, -0.2033, -0.2006, ..., -0.3931, -0.1381, -0.1785], + [ 0.0880, -0.0983, 0.0451, ..., 0.0453, 0.1460, -0.0696], + [-0.1123, 0.1654, -0.2403, ..., 0.0592, 0.1040, -0.0415], + ..., + [-0.0789, -0.1059, -0.0855, ..., 0.0137, -0.2393, 0.1784], + [ 0.0474, -0.0565, 0.1409, ..., 0.0084, -0.2752, -0.0415], + [-0.2594, -0.1346, -0.1886, ..., -0.3389, 0.0888, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, -1.8626e-09, ..., 2.2352e-08, + 3.5390e-08, 0.0000e+00], + [ 0.0000e+00, -1.3970e-08, -9.3132e-10, ..., -2.4214e-08, + -5.4017e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -6.5193e-09, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([ 6.5861e-05, -4.0483e-02, 1.0831e-02, -7.1458e-03, 8.8257e-03, + 6.7321e-03, 1.8672e-02, 8.8335e-03, -4.5961e-02, -1.5763e-02], + device='cuda:0'), grad: tensor([ 3.7253e-09, 6.7055e-08, -7.4506e-08, 4.6566e-09, 1.0245e-08, + -7.4506e-09, -9.3132e-09, -2.2352e-08, 1.6764e-08, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 247.48, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4473 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.17 lr 0.00001000 +Epoch 475, weight, value: tensor([[ 0.0212, -0.2033, -0.2007, ..., -0.3932, -0.1382, -0.1785], + [ 0.0880, -0.0982, 0.0451, ..., 0.0453, 0.1461, -0.0696], + [-0.1123, 0.1655, -0.2404, ..., 0.0592, 0.1040, -0.0415], + ..., + [-0.0789, -0.1060, -0.0855, ..., 0.0137, -0.2394, 0.1784], + [ 0.0474, -0.0565, 0.1408, ..., 0.0084, -0.2753, -0.0415], + [-0.2594, -0.1346, -0.1886, ..., -0.3391, 0.0888, -0.1110]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, -6.5193e-09, ..., 2.7940e-09, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, -2.6077e-08, 0.0000e+00, ..., -4.3772e-08, + -1.0245e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.1420e-08, 6.5193e-09, ..., 3.8184e-08, + 1.5832e-08, -0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 475, bias, value: tensor([ 5.7248e-05, -4.0466e-02, 1.0816e-02, -7.1474e-03, 8.8152e-03, + 6.7775e-03, 1.8677e-02, 8.8251e-03, -4.6215e-02, -1.5813e-02], + device='cuda:0'), grad: tensor([-3.7253e-09, 1.8626e-09, -1.2387e-07, 1.8626e-09, -2.7940e-09, + 2.7940e-09, -2.7940e-09, 1.1642e-07, 1.2107e-08, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 247.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4202 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.21 lr 0.00001000 +Epoch 476, weight, value: tensor([[ 0.0213, -0.2033, -0.2007, ..., -0.3932, -0.1383, -0.1785], + [ 0.0880, -0.0982, 0.0452, ..., 0.0454, 0.1462, -0.0697], + [-0.1123, 0.1655, -0.2406, ..., 0.0591, 0.1039, -0.0415], + ..., + [-0.0789, -0.1061, -0.0855, ..., 0.0137, -0.2395, 0.1785], + [ 0.0474, -0.0565, 0.1409, ..., 0.0084, -0.2754, -0.0415], + [-0.2594, -0.1347, -0.1887, ..., -0.3393, 0.0888, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -3.7253e-09, + 4.6566e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.9558e-08, 0.0000e+00]], device='cuda:0') +Epoch 476, bias, value: tensor([ 3.3043e-05, -4.0450e-02, 1.0697e-02, -7.1433e-03, 8.8149e-03, + 6.7709e-03, 1.8719e-02, 8.8479e-03, -4.6248e-02, -1.5884e-02], + device='cuda:0'), grad: tensor([ 1.7695e-08, 1.8626e-09, 1.3970e-08, 6.5193e-09, -6.5193e-09, + -2.8871e-08, 2.9802e-08, -8.3819e-09, 3.7253e-09, -2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 246.71, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4327 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.17 lr 0.00001000 +Epoch 477, weight, value: tensor([[ 0.0213, -0.2034, -0.2008, ..., -0.3933, -0.1384, -0.1785], + [ 0.0880, -0.0981, 0.0453, ..., 0.0454, 0.1463, -0.0697], + [-0.1123, 0.1656, -0.2408, ..., 0.0591, 0.1040, -0.0415], + ..., + [-0.0789, -0.1062, -0.0856, ..., 0.0137, -0.2396, 0.1785], + [ 0.0474, -0.0565, 0.1410, ..., 0.0084, -0.2755, -0.0415], + [-0.2594, -0.1347, -0.1888, ..., -0.3394, 0.0888, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, -3.7253e-09, ..., -9.3132e-10, + -1.0245e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -2.7940e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 477, bias, value: tensor([ 5.6391e-06, -4.0398e-02, 1.0677e-02, -7.1491e-03, 8.8128e-03, + 6.7670e-03, 1.8741e-02, 8.8157e-03, -4.6258e-02, -1.5942e-02], + device='cuda:0'), grad: tensor([ 6.5193e-09, -9.3132e-09, 2.7008e-08, 1.8626e-09, 1.3970e-08, + 4.6566e-09, -3.5390e-08, -1.2107e-08, 7.4506e-09, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 247.46, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4215 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.14 lr 0.00001000 +Epoch 478, weight, value: tensor([[ 0.0213, -0.2034, -0.2008, ..., -0.3934, -0.1385, -0.1785], + [ 0.0880, -0.0982, 0.0453, ..., 0.0454, 0.1463, -0.0697], + [-0.1123, 0.1656, -0.2408, ..., 0.0591, 0.1040, -0.0415], + ..., + [-0.0789, -0.1062, -0.0856, ..., 0.0137, -0.2397, 0.1786], + [ 0.0474, -0.0566, 0.1410, ..., 0.0084, -0.2756, -0.0415], + [-0.2595, -0.1347, -0.1888, ..., -0.3396, 0.0888, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 2.7940e-09, ..., 2.7940e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, -8.3819e-09, ..., -3.7253e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 9.3132e-10, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 478, bias, value: tensor([-3.3555e-05, -4.0387e-02, 1.0684e-02, -7.1559e-03, 8.7509e-03, + 6.7723e-03, 1.8780e-02, 8.8098e-03, -4.6307e-02, -1.5952e-02], + device='cuda:0'), grad: tensor([-1.4063e-07, 3.7253e-09, 2.7940e-08, -4.6566e-09, 6.5193e-09, + -8.3819e-09, 1.3970e-08, 1.4901e-08, 3.8184e-08, 4.3772e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 248.42, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4203 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.16 lr 0.00001000 +Epoch 479, weight, value: tensor([[ 0.0212, -0.2034, -0.2009, ..., -0.3936, -0.1388, -0.1785], + [ 0.0880, -0.0982, 0.0454, ..., 0.0454, 0.1464, -0.0697], + [-0.1124, 0.1657, -0.2409, ..., 0.0591, 0.1040, -0.0415], + ..., + [-0.0789, -0.1063, -0.0857, ..., 0.0137, -0.2398, 0.1786], + [ 0.0474, -0.0566, 0.1410, ..., 0.0084, -0.2757, -0.0416], + [-0.2595, -0.1348, -0.1890, ..., -0.3399, 0.0888, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 8.8476e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., -1.7229e-08, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + -4.1910e-09, 0.0000e+00]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0001, -0.0404, 0.0107, -0.0072, 0.0088, 0.0068, 0.0188, 0.0088, + -0.0464, -0.0161], device='cuda:0'), grad: tensor([ 1.8626e-09, 6.0536e-09, 4.0047e-08, 1.8626e-09, 3.2596e-09, + 1.8161e-08, 1.0245e-08, 1.2573e-08, -6.7987e-08, -2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 248.71, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4185 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.14 lr 0.00001000 +Epoch 480, weight, value: tensor([[ 0.0212, -0.2034, -0.2010, ..., -0.3936, -0.1389, -0.1786], + [ 0.0880, -0.0983, 0.0455, ..., 0.0455, 0.1466, -0.0698], + [-0.1124, 0.1659, -0.2411, ..., 0.0591, 0.1041, -0.0415], + ..., + [-0.0789, -0.1063, -0.0858, ..., 0.0136, -0.2400, 0.1788], + [ 0.0474, -0.0566, 0.1411, ..., 0.0084, -0.2759, -0.0416], + [-0.2595, -0.1348, -0.1891, ..., -0.3401, 0.0888, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.3283e-09, ..., 4.6566e-10, + -1.3970e-09, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 9.3132e-10, ..., -3.3062e-08, + -2.0489e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.0955e-08, + 1.5367e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0002, -0.0403, 0.0108, -0.0072, 0.0089, 0.0068, 0.0189, 0.0087, + -0.0465, -0.0162], device='cuda:0'), grad: tensor([-7.4506e-09, -4.6566e-10, -5.7742e-08, 1.2573e-08, -2.3283e-09, + -3.7253e-09, 5.5879e-09, 3.5856e-08, 1.4435e-08, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 248.65, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4422 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.15 lr 0.00001000 +Epoch 481, weight, value: tensor([[ 0.0212, -0.2034, -0.2011, ..., -0.3937, -0.1390, -0.1786], + [ 0.0880, -0.0982, 0.0456, ..., 0.0455, 0.1466, -0.0700], + [-0.1124, 0.1660, -0.2411, ..., 0.0592, 0.1042, -0.0415], + ..., + [-0.0789, -0.1065, -0.0859, ..., 0.0136, -0.2401, 0.1790], + [ 0.0474, -0.0566, 0.1411, ..., 0.0083, -0.2760, -0.0417], + [-0.2595, -0.1349, -0.1891, ..., -0.3403, 0.0887, -0.1111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + 3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + 1.7695e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -1.3970e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 7.9162e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0002, -0.0403, 0.0108, -0.0072, 0.0090, 0.0068, 0.0189, 0.0087, + -0.0466, -0.0164], device='cuda:0'), grad: tensor([ 5.0757e-08, 2.0955e-08, 4.9826e-08, 0.0000e+00, -3.7253e-08, + 3.4925e-08, -2.2212e-07, -8.8476e-09, 1.1176e-07, 7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 248.59, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4238 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.13 lr 0.00001000 +Epoch 482, weight, value: tensor([[ 0.0212, -0.2035, -0.2012, ..., -0.3939, -0.1392, -0.1786], + [ 0.0880, -0.0982, 0.0456, ..., 0.0455, 0.1467, -0.0702], + [-0.1124, 0.1660, -0.2413, ..., 0.0593, 0.1042, -0.0415], + ..., + [-0.0789, -0.1065, -0.0859, ..., 0.0137, -0.2401, 0.1792], + [ 0.0474, -0.0566, 0.1411, ..., 0.0083, -0.2762, -0.0417], + [-0.2595, -0.1349, -0.1892, ..., -0.3406, 0.0886, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.3970e-09, ..., 7.4506e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 4.1444e-08, 2.3283e-09, ..., 1.6484e-07, + -1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 2.5146e-08, 1.5832e-08, ..., 1.0012e-07, + 1.3970e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -7.0315e-08, 2.3283e-09, ..., -2.8638e-07, + 2.7940e-09, -1.3970e-09], + [ 0.0000e+00, -5.5879e-09, -3.5856e-08, ..., -2.0955e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 1.3970e-09, ..., 6.0536e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0003, -0.0403, 0.0108, -0.0073, 0.0092, 0.0069, 0.0189, 0.0088, + -0.0467, -0.0166], device='cuda:0'), grad: tensor([ 2.9802e-08, 5.4995e-07, 2.5472e-07, 4.7032e-08, -8.8476e-09, + 2.8871e-08, 9.7789e-09, -9.3272e-07, -8.8010e-08, 1.1362e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 249.02, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4256 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.12 lr 0.00001000 +Epoch 483, weight, value: tensor([[ 0.0212, -0.2035, -0.2014, ..., -0.3941, -0.1394, -0.1786], + [ 0.0880, -0.0984, 0.0456, ..., 0.0453, 0.1466, -0.0702], + [-0.1124, 0.1663, -0.2413, ..., 0.0593, 0.1044, -0.0414], + ..., + [-0.0789, -0.1066, -0.0858, ..., 0.0139, -0.2402, 0.1792], + [ 0.0474, -0.0567, 0.1411, ..., 0.0082, -0.2765, -0.0417], + [-0.2595, -0.1349, -0.1893, ..., -0.3407, 0.0886, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.1176e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0004, -0.0405, 0.0109, -0.0073, 0.0092, 0.0069, 0.0190, 0.0090, + -0.0470, -0.0166], device='cuda:0'), grad: tensor([ 6.1467e-08, 3.2596e-08, 2.7940e-09, 3.7253e-09, 1.8626e-09, + 8.3819e-09, -7.3574e-08, -5.2154e-08, 1.2107e-08, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 249.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4217 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.15 lr 0.00001000 +Epoch 484, weight, value: tensor([[ 0.0212, -0.2036, -0.2015, ..., -0.3942, -0.1397, -0.1786], + [ 0.0880, -0.0984, 0.0456, ..., 0.0452, 0.1467, -0.0704], + [-0.1124, 0.1663, -0.2414, ..., 0.0593, 0.1045, -0.0414], + ..., + [-0.0789, -0.1066, -0.0858, ..., 0.0140, -0.2403, 0.1794], + [ 0.0474, -0.0567, 0.1411, ..., 0.0082, -0.2767, -0.0418], + [-0.2595, -0.1349, -0.1894, ..., -0.3410, 0.0886, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0006, -0.0406, 0.0109, -0.0073, 0.0092, 0.0069, 0.0191, 0.0091, + -0.0471, -0.0167], device='cuda:0'), grad: tensor([-2.7940e-09, 6.5193e-09, 3.7253e-09, 0.0000e+00, -2.7940e-09, + -9.3132e-10, -1.9558e-08, 3.7253e-09, 7.4506e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 248.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4317 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.18 lr 0.00001000 +Epoch 485, weight, value: tensor([[ 0.0212, -0.2036, -0.2016, ..., -0.3943, -0.1398, -0.1786], + [ 0.0880, -0.0986, 0.0456, ..., 0.0451, 0.1467, -0.0704], + [-0.1124, 0.1665, -0.2415, ..., 0.0593, 0.1046, -0.0414], + ..., + [-0.0789, -0.1066, -0.0858, ..., 0.0141, -0.2404, 0.1795], + [ 0.0473, -0.0567, 0.1412, ..., 0.0082, -0.2768, -0.0418], + [-0.2595, -0.1350, -0.1895, ..., -0.3411, 0.0886, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 2.4214e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.0245e-08, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.9360e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -8.6613e-08, ..., -4.9360e-08, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + -4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0007, -0.0407, 0.0110, -0.0074, 0.0092, 0.0069, 0.0191, 0.0092, + -0.0472, -0.0168], device='cuda:0'), grad: tensor([ 5.4017e-08, 8.5682e-08, 3.1665e-08, 4.6566e-08, 1.8626e-08, + -2.7940e-09, 9.1270e-08, -1.1269e-07, -8.6613e-08, -1.2200e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 248.67, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4355 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.15 lr 0.00001000 +Epoch 486, weight, value: tensor([[ 0.0213, -0.2036, -0.2017, ..., -0.3943, -0.1399, -0.1786], + [ 0.0880, -0.0985, 0.0456, ..., 0.0451, 0.1468, -0.0704], + [-0.1124, 0.1666, -0.2416, ..., 0.0593, 0.1047, -0.0414], + ..., + [-0.0790, -0.1067, -0.0858, ..., 0.0142, -0.2405, 0.1795], + [ 0.0473, -0.0567, 0.1413, ..., 0.0082, -0.2770, -0.0418], + [-0.2595, -0.1350, -0.1895, ..., -0.3413, 0.0886, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 3.0734e-08, 3.7253e-09, ..., 1.1828e-07, + 5.1223e-08, 0.0000e+00], + [ 0.0000e+00, -3.6322e-08, 9.3132e-10, ..., -8.9407e-08, + -6.1467e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-09, -3.7253e-09, ..., -2.9802e-08, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0007, -0.0408, 0.0110, -0.0074, 0.0092, 0.0070, 0.0192, 0.0092, + -0.0473, -0.0168], device='cuda:0'), grad: tensor([-1.3597e-07, 1.7975e-07, -8.2888e-08, 8.3819e-09, 1.8626e-09, + 1.6764e-08, 3.1665e-08, -4.6566e-08, 1.0245e-08, 2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 248.89, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4379 re_mapping 0.0025 re_causal 0.0096 /// teacc 99.17 lr 0.00001000 +Epoch 487, weight, value: tensor([[ 0.0212, -0.2037, -0.2017, ..., -0.3944, -0.1401, -0.1786], + [ 0.0880, -0.0986, 0.0457, ..., 0.0451, 0.1469, -0.0704], + [-0.1125, 0.1667, -0.2417, ..., 0.0594, 0.1047, -0.0414], + ..., + [-0.0789, -0.1067, -0.0859, ..., 0.0141, -0.2408, 0.1795], + [ 0.0473, -0.0567, 0.1414, ..., 0.0083, -0.2770, -0.0418], + [-0.2595, -0.1351, -0.1896, ..., -0.3414, 0.0886, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.2107e-08, 0.0000e+00]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0007, -0.0407, 0.0110, -0.0074, 0.0093, 0.0070, 0.0192, 0.0092, + -0.0473, -0.0169], device='cuda:0'), grad: tensor([-1.8626e-09, 6.5193e-09, 9.3132e-10, 2.7940e-09, 2.3283e-08, + -3.7253e-09, 0.0000e+00, 3.7253e-09, 9.3132e-10, -2.5146e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 248.67, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4054 re_mapping 0.0025 re_causal 0.0091 /// teacc 99.16 lr 0.00001000 +Epoch 488, weight, value: tensor([[ 0.0212, -0.2037, -0.2018, ..., -0.3945, -0.1403, -0.1787], + [ 0.0880, -0.0986, 0.0458, ..., 0.0452, 0.1470, -0.0704], + [-0.1125, 0.1668, -0.2418, ..., 0.0594, 0.1047, -0.0414], + ..., + [-0.0789, -0.1069, -0.0860, ..., 0.0141, -0.2409, 0.1795], + [ 0.0473, -0.0567, 0.1415, ..., 0.0084, -0.2772, -0.0418], + [-0.2595, -0.1351, -0.1897, ..., -0.3415, 0.0887, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 3.7253e-09, ..., -0.0000e+00, + -1.8626e-09, -0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0008, -0.0407, 0.0110, -0.0074, 0.0094, 0.0070, 0.0193, 0.0091, + -0.0473, -0.0169], device='cuda:0'), grad: tensor([ 6.3330e-08, 7.4506e-09, 5.5879e-09, -4.6566e-08, 1.1176e-08, + 3.7253e-08, -1.0803e-07, 0.0000e+00, 9.3132e-09, 1.1176e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 248.77, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4316 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.18 lr 0.00001000 +Epoch 489, weight, value: tensor([[ 0.0213, -0.2037, -0.2019, ..., -0.3946, -0.1405, -0.1787], + [ 0.0880, -0.0986, 0.0459, ..., 0.0452, 0.1471, -0.0705], + [-0.1125, 0.1668, -0.2418, ..., 0.0594, 0.1048, -0.0414], + ..., + [-0.0789, -0.1070, -0.0860, ..., 0.0141, -0.2411, 0.1796], + [ 0.0473, -0.0567, 0.1416, ..., 0.0084, -0.2773, -0.0418], + [-0.2595, -0.1351, -0.1898, ..., -0.3416, 0.0888, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -1.6764e-08, 0.0000e+00, ..., -4.2841e-08, + -3.9116e-08, -1.8626e-09], + ..., + [ 0.0000e+00, 1.4901e-08, 0.0000e+00, ..., 2.9802e-08, + 3.5390e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0008, -0.0407, 0.0110, -0.0074, 0.0093, 0.0070, 0.0194, 0.0091, + -0.0474, -0.0169], device='cuda:0'), grad: tensor([-2.2352e-08, 2.4214e-08, -1.2293e-07, 4.2841e-08, 0.0000e+00, + -7.2643e-08, 1.3039e-08, 9.1270e-08, 1.3039e-08, 1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 249.25, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3846 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.18 lr 0.00001000 +Epoch 490, weight, value: tensor([[ 0.0213, -0.2037, -0.2020, ..., -0.3946, -0.1406, -0.1787], + [ 0.0880, -0.0987, 0.0461, ..., 0.0453, 0.1472, -0.0705], + [-0.1125, 0.1671, -0.2420, ..., 0.0595, 0.1050, -0.0414], + ..., + [-0.0789, -0.1071, -0.0863, ..., 0.0139, -0.2413, 0.1796], + [ 0.0473, -0.0567, 0.1417, ..., 0.0085, -0.2775, -0.0418], + [-0.2596, -0.1352, -0.1899, ..., -0.3417, 0.0888, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 3.5390e-08, 1.0245e-07, ..., 2.5705e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 3.7253e-09, ..., 1.6764e-08, + -2.2352e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -3.7253e-08, -1.1548e-07, ..., -2.9430e-07, + 5.5879e-09, 0.0000e+00], + [-1.8626e-09, 1.8626e-09, -3.7253e-09, ..., 1.8626e-09, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0008, -0.0406, 0.0112, -0.0075, 0.0094, 0.0070, 0.0194, 0.0090, + -0.0474, -0.0170], device='cuda:0'), grad: tensor([ 1.4901e-08, 7.2829e-07, -3.5390e-08, 5.0291e-08, 0.0000e+00, + 7.4506e-09, -3.7253e-09, -8.2329e-07, 4.0978e-08, 2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 249.33, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4278 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.15 lr 0.00001000 +Epoch 491, weight, value: tensor([[ 0.0213, -0.2037, -0.2021, ..., -0.3947, -0.1408, -0.1787], + [ 0.0880, -0.0986, 0.0462, ..., 0.0453, 0.1473, -0.0705], + [-0.1125, 0.1671, -0.2423, ..., 0.0594, 0.1050, -0.0414], + ..., + [-0.0789, -0.1073, -0.0863, ..., 0.0139, -0.2415, 0.1797], + [ 0.0473, -0.0567, 0.1418, ..., 0.0085, -0.2776, -0.0418], + [-0.2596, -0.1352, -0.1900, ..., -0.3419, 0.0889, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 1.3039e-08, ..., -9.3132e-09, + 1.8626e-09, -1.8626e-09], + [ 0.0000e+00, -5.5879e-09, -1.4901e-08, ..., -5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0008, -0.0406, 0.0112, -0.0075, 0.0093, 0.0070, 0.0195, 0.0090, + -0.0474, -0.0170], device='cuda:0'), grad: tensor([ 1.1176e-08, 4.2841e-08, 1.8626e-08, -1.1176e-08, -2.0489e-08, + 5.5879e-09, 3.7253e-09, -7.4506e-09, -2.6077e-08, -1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 248.16, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0025 re_causal 0.0096 /// teacc 99.13 lr 0.00001000 +Epoch 492, weight, value: tensor([[ 0.0214, -0.2037, -0.2022, ..., -0.3947, -0.1410, -0.1787], + [ 0.0880, -0.0986, 0.0463, ..., 0.0454, 0.1474, -0.0705], + [-0.1126, 0.1671, -0.2425, ..., 0.0593, 0.1050, -0.0414], + ..., + [-0.0788, -0.1074, -0.0864, ..., 0.0139, -0.2416, 0.1797], + [ 0.0473, -0.0567, 0.1419, ..., 0.0086, -0.2776, -0.0417], + [-0.2596, -0.1352, -0.1902, ..., -0.3421, 0.0890, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0009, -0.0405, 0.0111, -0.0075, 0.0094, 0.0070, 0.0195, 0.0089, + -0.0474, -0.0170], device='cuda:0'), grad: tensor([ 0.0000e+00, 7.4506e-09, -1.8626e-09, 2.0489e-08, -2.4214e-08, + 1.8626e-09, 1.3039e-08, 3.7253e-09, 1.8626e-09, -2.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 247.40, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4269 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.15 lr 0.00001000 +Epoch 493, weight, value: tensor([[ 0.0214, -0.2037, -0.2023, ..., -0.3948, -0.1411, -0.1787], + [ 0.0880, -0.0986, 0.0464, ..., 0.0454, 0.1476, -0.0705], + [-0.1126, 0.1672, -0.2426, ..., 0.0592, 0.1050, -0.0414], + ..., + [-0.0789, -0.1075, -0.0865, ..., 0.0139, -0.2418, 0.1797], + [ 0.0473, -0.0567, 0.1421, ..., 0.0087, -0.2778, -0.0417], + [-0.2596, -0.1353, -0.1904, ..., -0.3423, 0.0889, -0.1112]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -0.0000e+00, 1.8626e-09, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0009, -0.0404, 0.0110, -0.0075, 0.0095, 0.0070, 0.0195, 0.0089, + -0.0474, -0.0172], device='cuda:0'), grad: tensor([ 2.4214e-08, 1.8626e-09, 0.0000e+00, -3.7253e-09, 1.8626e-08, + 5.5879e-09, -3.7253e-08, 0.0000e+00, 5.5879e-09, -1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 247.51, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4308 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.13 lr 0.00001000 +Epoch 494, weight, value: tensor([[ 0.0214, -0.2037, -0.2024, ..., -0.3948, -0.1412, -0.1787], + [ 0.0880, -0.0985, 0.0466, ..., 0.0455, 0.1479, -0.0706], + [-0.1127, 0.1673, -0.2427, ..., 0.0593, 0.1050, -0.0414], + ..., + [-0.0788, -0.1077, -0.0866, ..., 0.0138, -0.2421, 0.1798], + [ 0.0472, -0.0567, 0.1421, ..., 0.0087, -0.2779, -0.0418], + [-0.2596, -0.1353, -0.1905, ..., -0.3425, 0.0889, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.5460e-07, ..., -7.6368e-08, + -1.8254e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 7.4506e-09, + 1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2852e-07, ..., 6.1467e-08, + 1.5087e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0008, -0.0403, 0.0111, -0.0076, 0.0095, 0.0070, 0.0195, 0.0088, + -0.0475, -0.0173], device='cuda:0'), grad: tensor([ 8.0094e-08, -4.0419e-07, 4.2841e-08, 1.2666e-07, 2.2352e-08, + -9.6858e-08, -1.1548e-07, 3.2969e-07, 1.3039e-08, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 247.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4121 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.17 lr 0.00001000 +Epoch 495, weight, value: tensor([[ 0.0214, -0.2037, -0.2024, ..., -0.3949, -0.1413, -0.1787], + [ 0.0880, -0.0985, 0.0467, ..., 0.0455, 0.1480, -0.0707], + [-0.1127, 0.1674, -0.2428, ..., 0.0594, 0.1050, -0.0414], + ..., + [-0.0788, -0.1079, -0.0867, ..., 0.0138, -0.2423, 0.1799], + [ 0.0472, -0.0567, 0.1422, ..., 0.0088, -0.2780, -0.0418], + [-0.2597, -0.1354, -0.1907, ..., -0.3426, 0.0890, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0008, -0.0403, 0.0111, -0.0076, 0.0094, 0.0070, 0.0196, 0.0088, + -0.0476, -0.0173], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.8626e-09, 1.8626e-09, 7.4506e-09, 0.0000e+00, + -3.7253e-09, -9.3132e-09, -7.4506e-09, 0.0000e+00, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 247.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4444 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.18 lr 0.00001000 +Epoch 496, weight, value: tensor([[ 0.0214, -0.2037, -0.2025, ..., -0.3950, -0.1414, -0.1787], + [ 0.0880, -0.0985, 0.0468, ..., 0.0455, 0.1481, -0.0707], + [-0.1127, 0.1675, -0.2431, ..., 0.0593, 0.1051, -0.0414], + ..., + [-0.0788, -0.1080, -0.0867, ..., 0.0138, -0.2425, 0.1800], + [ 0.0472, -0.0568, 0.1422, ..., 0.0087, -0.2781, -0.0418], + [-0.2597, -0.1354, -0.1908, ..., -0.3427, 0.0892, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-08, 1.8626e-09, ..., 6.5193e-08, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, -6.3330e-08, 0.0000e+00, ..., -3.1665e-08, + -5.9605e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.4214e-08, -1.8626e-09, ..., -4.6566e-08, + 3.1665e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 1.8626e-09, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0008, -0.0403, 0.0112, -0.0076, 0.0093, 0.0070, 0.0196, 0.0088, + -0.0476, -0.0173], device='cuda:0'), grad: tensor([ 1.1176e-08, 2.0117e-07, -1.4901e-07, 6.5193e-08, 1.3039e-08, + -3.5390e-08, -3.5390e-08, -1.0617e-07, 3.1665e-08, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 247.41, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3879 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.14 lr 0.00001000 +Epoch 497, weight, value: tensor([[ 0.0214, -0.2038, -0.2026, ..., -0.3951, -0.1417, -0.1787], + [ 0.0880, -0.0985, 0.0468, ..., 0.0454, 0.1481, -0.0708], + [-0.1127, 0.1676, -0.2433, ..., 0.0593, 0.1052, -0.0413], + ..., + [-0.0788, -0.1082, -0.0867, ..., 0.0139, -0.2426, 0.1801], + [ 0.0472, -0.0568, 0.1423, ..., 0.0087, -0.2783, -0.0418], + [-0.2597, -0.1355, -0.1909, ..., -0.3429, 0.0892, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, -3.7253e-09, ..., -1.8626e-09, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, -0.0000e+00]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0009, -0.0404, 0.0112, -0.0076, 0.0093, 0.0070, 0.0198, 0.0089, + -0.0477, -0.0174], device='cuda:0'), grad: tensor([ 3.3528e-08, -5.5879e-09, 2.4214e-08, 0.0000e+00, 5.5879e-09, + 0.0000e+00, -6.5193e-08, 7.4506e-09, -3.7253e-09, -1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 247.41, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4189 re_mapping 0.0024 re_causal 0.0093 /// teacc 99.17 lr 0.00001000 +Epoch 498, weight, value: tensor([[ 0.0214, -0.2038, -0.2027, ..., -0.3951, -0.1418, -0.1787], + [ 0.0879, -0.0985, 0.0469, ..., 0.0454, 0.1482, -0.0708], + [-0.1127, 0.1677, -0.2433, ..., 0.0592, 0.1052, -0.0413], + ..., + [-0.0788, -0.1081, -0.0868, ..., 0.0140, -0.2427, 0.1801], + [ 0.0472, -0.0568, 0.1423, ..., 0.0087, -0.2785, -0.0417], + [-0.2597, -0.1355, -0.1909, ..., -0.3430, 0.0892, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.7253e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0009, -0.0404, 0.0112, -0.0076, 0.0094, 0.0070, 0.0199, 0.0089, + -0.0479, -0.0175], device='cuda:0'), grad: tensor([ 1.8626e-09, 2.0489e-08, 1.1176e-08, 1.1921e-07, -5.2154e-08, + -1.6205e-07, 2.6077e-08, -3.7253e-09, 2.0489e-08, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 247.19, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4471 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.18 lr 0.00001000 +Epoch 499, weight, value: tensor([[ 0.0214, -0.2038, -0.2027, ..., -0.3952, -0.1419, -0.1787], + [ 0.0879, -0.0985, 0.0469, ..., 0.0453, 0.1482, -0.0708], + [-0.1127, 0.1678, -0.2435, ..., 0.0592, 0.1054, -0.0413], + ..., + [-0.0788, -0.1082, -0.0868, ..., 0.0141, -0.2428, 0.1802], + [ 0.0472, -0.0568, 0.1425, ..., 0.0088, -0.2785, -0.0417], + [-0.2597, -0.1355, -0.1910, ..., -0.3432, 0.0893, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, -7.4506e-09, ..., -3.7253e-09, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.3469e-07, ..., -4.0978e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0008, -0.0405, 0.0112, -0.0076, 0.0094, 0.0071, 0.0199, 0.0090, + -0.0479, -0.0175], device='cuda:0'), grad: tensor([ 9.3132e-09, -1.8626e-08, 7.4506e-09, 1.1548e-07, 1.8626e-09, + 5.5879e-09, 1.6950e-07, 5.5879e-09, -2.9244e-07, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 247.22, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4335 re_mapping 0.0024 re_causal 0.0093 /// teacc 99.14 lr 0.00001000 +Epoch 500, weight, value: tensor([[ 0.0214, -0.2038, -0.2028, ..., -0.3952, -0.1420, -0.1787], + [ 0.0879, -0.0986, 0.0469, ..., 0.0453, 0.1482, -0.0708], + [-0.1128, 0.1679, -0.2436, ..., 0.0592, 0.1054, -0.0413], + ..., + [-0.0788, -0.1082, -0.0868, ..., 0.0141, -0.2429, 0.1802], + [ 0.0472, -0.0568, 0.1426, ..., 0.0089, -0.2786, -0.0417], + [-0.2597, -0.1355, -0.1911, ..., -0.3433, 0.0893, -0.1112]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.9802e-08, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, -3.9116e-08, ..., -1.6764e-08, + -4.2841e-08, 0.0000e+00], + [ 0.0000e+00, -3.7253e-08, 2.4214e-08, ..., -2.6077e-08, + -1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 4.4703e-08, 1.3039e-08, ..., 4.2841e-08, + 5.4017e-08, -0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0009, -0.0406, 0.0112, -0.0077, 0.0093, 0.0072, 0.0199, 0.0090, + -0.0480, -0.0176], device='cuda:0'), grad: tensor([ 1.7323e-07, -7.8231e-08, -3.1665e-08, 5.0291e-08, -7.4506e-09, + -2.3097e-07, -1.7509e-07, 2.0117e-07, 2.7940e-08, 7.2643e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 247.44, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4273 re_mapping 0.0024 re_causal 0.0091 /// teacc 99.16 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_adam', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_adam/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.040001 98.930000 ... 79.62133 68.275971 +ShearY 98.769997 98.699997 ... 79.62133 64.653491 +AutoContrast 99.199997 99.099998 ... 79.62133 59.755771 +Invert 98.860001 98.369995 ... 79.62133 64.192465 +Equalize 98.439995 98.229996 ... 79.62133 70.835657 +Solarize 98.239998 97.639999 ... 79.62133 59.579589 +SolarizeAdd 98.400002 97.779999 ... 79.62133 72.674644 +Posterize 98.909996 99.029999 ... 79.62133 72.055723 +Contrast 99.159996 99.180000 ... 79.62133 66.427597 +Color 99.119995 99.220001 ... 79.62133 59.084085 +Brightness 99.119995 99.229996 ... 79.62133 65.520764 +Sharpness 99.099998 99.150002 ... 79.62133 69.842453 +NoiseSalt 99.099998 99.169998 ... 79.62133 53.793901 +NoiseGaussian 99.080002 99.199997 ... 79.62133 55.708333 +w/o do (original x) 99.220000 0.000000 ... 0.00000 73.114814 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.13 69.283958 78.813465 77.546321 83.507723 77.287867 diff --git a/Meta-causal/code-withStyleAttack/66560.error b/Meta-causal/code-withStyleAttack/66560.error new file mode 100644 index 0000000000000000000000000000000000000000..57a075d0c4f7d6342343977072b7d558fa37ce15 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66560.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 35: oaug: command not found diff --git a/Meta-causal/code-withStyleAttack/66560.log b/Meta-causal/code-withStyleAttack/66560.log new file mode 100644 index 0000000000000000000000000000000000000000..d05ecaf7c3bed6dc29bcf39b1d6419d4d5ac8bd1 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66560.log @@ -0,0 +1,14047 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0094, -0.0300, -0.0202, ..., -0.0133, 0.0254, 0.0255], + [-0.0269, -0.0168, 0.0046, ..., -0.0113, -0.0025, -0.0234], + [ 0.0059, -0.0206, 0.0299, ..., -0.0036, 0.0285, -0.0219], + ..., + [-0.0253, -0.0004, 0.0165, ..., 0.0212, 0.0120, -0.0137], + [-0.0243, 0.0281, -0.0075, ..., 0.0071, -0.0178, -0.0153], + [ 0.0121, 0.0164, -0.0064, ..., 0.0142, -0.0213, 0.0214]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0146, -0.0232, -0.0161, -0.0271, -0.0302, -0.0007, 0.0295, -0.0105, + 0.0252, 0.0037], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 282.82, cls_loss 1.3336 cls_loss_mapping 1.8518 cls_loss_causal 2.2299 re_mapping 0.1420 re_causal 0.1531 /// teacc 87.82 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0055, -0.0290, -0.0202, ..., -0.0107, 0.0303, 0.0314], + [-0.0261, -0.0221, 0.0046, ..., -0.0192, -0.0076, -0.0313], + [ 0.0041, -0.0206, 0.0299, ..., -0.0090, 0.0275, -0.0207], + ..., + [-0.0252, 0.0030, 0.0165, ..., 0.0254, 0.0123, -0.0181], + [-0.0266, 0.0263, -0.0075, ..., 0.0078, -0.0190, -0.0177], + [ 0.0128, 0.0130, -0.0064, ..., 0.0154, -0.0234, 0.0160]], + device='cuda:0'), grad: tensor([[ 4.3144e-03, 1.3351e-03, 0.0000e+00, ..., 4.6462e-05, + -5.3711e-03, -9.2926e-03], + [ 7.6592e-05, 5.3520e-03, 0.0000e+00, ..., 7.4463e-03, + 4.2248e-04, 8.9073e-04], + [-2.7275e-03, -1.2817e-02, 0.0000e+00, ..., -9.4910e-03, + -9.2411e-04, 1.5173e-03], + ..., + [ 1.1215e-02, 1.5182e-02, 0.0000e+00, ..., 5.3375e-02, + 3.4084e-03, 3.1624e-03], + [ 8.5831e-03, 1.8021e-02, 0.0000e+00, ..., 5.5618e-03, + -1.3704e-03, 9.5901e-03], + [-1.7303e-02, -7.7362e-03, 0.0000e+00, ..., -8.3313e-02, + 3.7365e-03, 2.5520e-03]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0159, -0.0208, -0.0162, -0.0281, -0.0302, -0.0004, 0.0285, -0.0104, + 0.0253, 0.0029], device='cuda:0'), grad: tensor([ 0.0027, -0.0289, -0.0156, -0.0481, 0.0213, 0.0228, -0.0087, 0.0459, + 0.0480, -0.0393], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 281.67, cls_loss 0.4012 cls_loss_mapping 0.7534 cls_loss_causal 1.9334 re_mapping 0.2077 re_causal 0.2738 /// teacc 93.25 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0022, -0.0282, -0.0202, ..., -0.0099, 0.0362, 0.0349], + [-0.0259, -0.0246, 0.0046, ..., -0.0213, -0.0107, -0.0325], + [ 0.0010, -0.0205, 0.0299, ..., -0.0113, 0.0292, -0.0209], + ..., + [-0.0242, 0.0055, 0.0165, ..., 0.0269, 0.0072, -0.0198], + [-0.0320, 0.0258, -0.0075, ..., 0.0082, -0.0207, -0.0205], + [ 0.0134, 0.0127, -0.0064, ..., 0.0164, -0.0222, 0.0148]], + device='cuda:0'), grad: tensor([[ 2.1038e-03, 1.9836e-03, 0.0000e+00, ..., 1.3411e-05, + -1.7061e-03, -2.1172e-03], + [ 2.2125e-03, 5.4207e-03, 0.0000e+00, ..., 1.2369e-03, + 1.3618e-03, 2.9087e-03], + [ 3.6907e-03, -1.3664e-02, 0.0000e+00, ..., 2.2430e-03, + -2.1164e-02, -1.0399e-02], + ..., + [-3.5238e-04, -4.6272e-03, 0.0000e+00, ..., 1.6642e-03, + 4.7264e-03, 3.5000e-03], + [ 5.5885e-03, 7.3090e-03, 0.0000e+00, ..., 6.1798e-03, + 5.9509e-03, 8.5144e-03], + [ 3.1395e-03, 7.3671e-04, 0.0000e+00, ..., 1.4439e-03, + 1.4496e-03, 1.5640e-03]], device='cuda:0') +Epoch 3, bias, value: tensor([-1.5880e-02, -2.0302e-02, -1.6533e-02, -2.8419e-02, -3.0375e-02, + -7.7978e-05, 2.8361e-02, -1.0864e-02, 2.5489e-02, 3.3449e-03], + device='cuda:0'), grad: tensor([ 0.0057, 0.0006, -0.0258, 0.0253, -0.0084, -0.0065, -0.0296, 0.0029, + 0.0284, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 282.63, cls_loss 0.2555 cls_loss_mapping 0.4624 cls_loss_causal 1.7128 re_mapping 0.1559 re_causal 0.2582 /// teacc 94.25 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0010, -0.0277, -0.0203, ..., -0.0096, 0.0400, 0.0373], + [-0.0269, -0.0269, 0.0003, ..., -0.0233, -0.0158, -0.0343], + [-0.0026, -0.0202, 0.0260, ..., -0.0129, 0.0329, -0.0202], + ..., + [-0.0244, 0.0070, 0.0179, ..., 0.0281, 0.0037, -0.0209], + [-0.0364, 0.0257, -0.0115, ..., 0.0083, -0.0209, -0.0216], + [ 0.0141, 0.0133, -0.0103, ..., 0.0167, -0.0215, 0.0142]], + device='cuda:0'), grad: tensor([[-0.0026, -0.0088, 0.0000, ..., -0.0040, -0.0160, -0.0242], + [ 0.0007, 0.0026, 0.0000, ..., 0.0031, 0.0004, 0.0006], + [ 0.0015, 0.0109, 0.0000, ..., 0.0027, 0.0036, 0.0076], + ..., + [-0.0056, -0.0101, 0.0000, ..., -0.0206, 0.0016, 0.0023], + [-0.0090, -0.0121, 0.0000, ..., -0.0152, 0.0004, 0.0023], + [ 0.0074, 0.0182, 0.0000, ..., 0.0283, 0.0013, 0.0036]], + device='cuda:0') +Epoch 4, bias, value: tensor([-1.5505e-02, -1.9999e-02, -1.6656e-02, -2.8608e-02, -3.0272e-02, + -4.8942e-05, 2.8023e-02, -1.1061e-02, 2.5536e-02, 3.4134e-03], + device='cuda:0'), grad: tensor([-0.0162, 0.0015, 0.0200, -0.0070, 0.0192, 0.0060, 0.0029, -0.0074, + -0.0452, 0.0262], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 282.60, cls_loss 0.2090 cls_loss_mapping 0.3590 cls_loss_causal 1.5909 re_mapping 0.1200 re_causal 0.2258 /// teacc 94.39 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 2.5493e-04, -2.7750e-02, -2.0303e-02, ..., -8.9923e-03, + 4.3243e-02, 3.9449e-02], + [-2.9210e-02, -2.9602e-02, 3.2619e-04, ..., -2.5070e-02, + -1.9404e-02, -3.5510e-02], + [-5.9607e-03, -2.0000e-02, 2.5963e-02, ..., -1.4382e-02, + 3.6077e-02, -2.0003e-02], + ..., + [-2.5401e-02, 8.9050e-03, 1.7921e-02, ..., 2.9216e-02, + 4.4441e-05, -2.2393e-02], + [-3.9703e-02, 2.5462e-02, -1.1513e-02, ..., 8.7589e-03, + -2.1776e-02, -2.2173e-02], + [ 1.6515e-02, 1.3325e-02, -1.0273e-02, ..., 1.7216e-02, + -2.2674e-02, 1.2541e-02]], device='cuda:0'), grad: tensor([[ 0.0027, 0.0031, 0.0000, ..., 0.0023, 0.0094, 0.0074], + [ 0.0018, 0.0021, 0.0000, ..., 0.0005, 0.0012, 0.0016], + [ 0.0048, 0.0085, 0.0000, ..., -0.0031, 0.0103, -0.0066], + ..., + [-0.0050, -0.0013, 0.0000, ..., -0.0181, 0.0089, 0.0077], + [ 0.0069, -0.0212, 0.0000, ..., 0.0060, -0.0143, -0.0090], + [ 0.0175, 0.0194, 0.0000, ..., 0.0287, 0.0017, 0.0018]], + device='cuda:0') +Epoch 5, bias, value: tensor([-0.0152, -0.0199, -0.0166, -0.0291, -0.0305, -0.0003, 0.0278, -0.0110, + 0.0257, 0.0039], device='cuda:0'), grad: tensor([ 0.0084, 0.0034, 0.0093, -0.0302, -0.0279, 0.0112, 0.0061, -0.0065, + -0.0066, 0.0328], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 281.51, cls_loss 0.1658 cls_loss_mapping 0.2742 cls_loss_causal 1.4598 re_mapping 0.1014 re_causal 0.2046 /// teacc 96.27 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0021, -0.0277, -0.0203, ..., -0.0089, 0.0456, 0.0410], + [-0.0318, -0.0321, 0.0003, ..., -0.0259, -0.0237, -0.0371], + [-0.0077, -0.0207, 0.0260, ..., -0.0157, 0.0379, -0.0194], + ..., + [-0.0263, 0.0101, 0.0179, ..., 0.0304, -0.0027, -0.0232], + [-0.0431, 0.0261, -0.0115, ..., 0.0088, -0.0217, -0.0225], + [ 0.0183, 0.0129, -0.0103, ..., 0.0169, -0.0224, 0.0118]], + device='cuda:0'), grad: tensor([[ 0.0012, 0.0040, 0.0000, ..., 0.0147, 0.0457, 0.0405], + [ 0.0020, 0.0083, 0.0000, ..., 0.0069, 0.0022, 0.0052], + [ 0.0005, -0.0112, 0.0000, ..., 0.0024, -0.0245, -0.0163], + ..., + [ 0.0003, -0.0018, 0.0000, ..., 0.0006, 0.0016, -0.0008], + [ 0.0083, 0.0035, 0.0000, ..., 0.0126, -0.0011, 0.0004], + [ 0.0039, 0.0016, 0.0000, ..., -0.0119, -0.0455, -0.0388]], + device='cuda:0') +Epoch 6, bias, value: tensor([-0.0150, -0.0198, -0.0166, -0.0290, -0.0305, -0.0004, 0.0278, -0.0112, + 0.0257, 0.0039], device='cuda:0'), grad: tensor([ 0.0339, 0.0194, -0.0110, 0.0025, -0.0038, 0.0027, -0.0183, -0.0130, + 0.0092, -0.0214], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 282.87, cls_loss 0.1459 cls_loss_mapping 0.2310 cls_loss_causal 1.3923 re_mapping 0.0834 re_causal 0.1812 /// teacc 96.57 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0028, -0.0283, -0.0208, ..., -0.0088, 0.0478, 0.0425], + [-0.0341, -0.0343, -0.0055, ..., -0.0272, -0.0243, -0.0370], + [-0.0100, -0.0209, 0.0199, ..., -0.0167, 0.0402, -0.0190], + ..., + [-0.0270, 0.0116, 0.0221, ..., 0.0316, -0.0060, -0.0246], + [-0.0467, 0.0262, -0.0158, ..., 0.0088, -0.0228, -0.0234], + [ 0.0199, 0.0127, -0.0133, ..., 0.0170, -0.0218, 0.0120]], + device='cuda:0'), grad: tensor([[ 1.8060e-04, 9.7752e-04, 6.4773e-07, ..., 1.9608e-02, + 4.5349e-02, 3.0823e-02], + [ 1.3971e-04, 2.5082e-03, 4.6566e-06, ..., 1.7109e-03, + 3.0689e-03, 1.7014e-03], + [ 5.6934e-04, 2.4242e-03, 1.0632e-05, ..., 6.1369e-04, + -5.5145e-02, -1.5068e-02], + ..., + [ 9.0408e-04, 4.4937e-03, -4.3273e-05, ..., 1.0996e-03, + 1.3056e-03, 7.0095e-04], + [ 2.8095e-03, 2.6550e-03, 2.2054e-06, ..., -2.4994e-02, + -2.8015e-02, -3.0899e-02], + [-2.0809e-03, -1.1635e-03, 1.4558e-05, ..., -1.0240e-04, + 2.0123e-03, 9.8610e-04]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0147, -0.0197, -0.0168, -0.0292, -0.0306, -0.0006, 0.0275, -0.0111, + 0.0259, 0.0041], device='cuda:0'), grad: tensor([ 0.0331, 0.0101, -0.0394, -0.0025, 0.0005, 0.0057, 0.0027, 0.0099, + -0.0198, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 281.63, cls_loss 0.1317 cls_loss_mapping 0.1933 cls_loss_causal 1.3203 re_mapping 0.0716 re_causal 0.1593 /// teacc 96.81 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0048, -0.0286, -0.0268, ..., -0.0093, 0.0498, 0.0438], + [-0.0365, -0.0366, -0.0260, ..., -0.0284, -0.0273, -0.0377], + [-0.0119, -0.0213, 0.0253, ..., -0.0181, 0.0427, -0.0184], + ..., + [-0.0277, 0.0125, 0.0295, ..., 0.0327, -0.0083, -0.0260], + [-0.0498, 0.0266, -0.0332, ..., 0.0093, -0.0239, -0.0239], + [ 0.0209, 0.0119, -0.0229, ..., 0.0168, -0.0217, 0.0116]], + device='cuda:0'), grad: tensor([[ 9.6083e-05, 3.7003e-04, 4.8369e-05, ..., -9.7036e-05, + -1.8158e-03, -1.6947e-03], + [ 8.8692e-05, 5.4398e-03, 1.2779e-04, ..., 1.9608e-03, + 6.4909e-05, -1.0681e-04], + [ 1.9610e-04, 2.5043e-03, 9.3603e-04, ..., 1.2054e-03, + -2.2125e-03, -1.3580e-03], + ..., + [ 2.0561e-03, 1.9424e-02, 3.8795e-03, ..., 5.5237e-03, + 4.4560e-04, 3.5691e-04], + [ 8.5545e-04, 1.6947e-03, 7.8738e-05, ..., 1.2903e-03, + 1.2207e-03, 1.1091e-03], + [-4.1847e-03, -1.0818e-02, -6.6853e-04, ..., -8.7051e-03, + 2.7251e-04, 2.3580e-04]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0147, -0.0198, -0.0167, -0.0289, -0.0301, -0.0008, 0.0271, -0.0111, + 0.0261, 0.0038], device='cuda:0'), grad: tensor([-0.0005, 0.0265, 0.0027, -0.0231, 0.0046, 0.0014, 0.0007, 0.0280, + 0.0056, -0.0459], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 281.50, cls_loss 0.1119 cls_loss_mapping 0.1637 cls_loss_causal 1.2541 re_mapping 0.0646 re_causal 0.1477 /// teacc 97.21 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0055, -0.0287, -0.0292, ..., -0.0092, 0.0518, 0.0451], + [-0.0375, -0.0384, -0.0342, ..., -0.0291, -0.0285, -0.0377], + [-0.0135, -0.0212, 0.0253, ..., -0.0189, 0.0435, -0.0183], + ..., + [-0.0286, 0.0129, 0.0278, ..., 0.0331, -0.0106, -0.0276], + [-0.0521, 0.0269, -0.0451, ..., 0.0096, -0.0245, -0.0247], + [ 0.0218, 0.0119, -0.0259, ..., 0.0166, -0.0212, 0.0115]], + device='cuda:0'), grad: tensor([[ 3.9673e-04, 1.1339e-03, 2.7013e-04, ..., -4.6670e-05, + -2.2182e-03, -1.8110e-03], + [ 1.2434e-04, 6.1560e-04, 3.9268e-04, ..., 2.0015e-04, + 4.9019e-04, 6.7651e-05], + [ 3.3617e-04, 8.3542e-04, -3.4022e-04, ..., 5.9128e-04, + -2.4867e-04, 2.5988e-04], + ..., + [-2.2674e-04, -5.3310e-04, 1.7214e-04, ..., -1.3895e-03, + 7.6723e-04, 3.7551e-04], + [ 3.4008e-03, 1.4007e-04, 2.4676e-04, ..., 1.8578e-03, + 1.0824e-03, 1.4007e-04], + [-1.2541e-03, 1.7948e-03, 3.2425e-04, ..., -1.7605e-03, + 1.2903e-03, 8.0490e-04]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0142, -0.0195, -0.0172, -0.0290, -0.0302, -0.0011, 0.0270, -0.0117, + 0.0265, 0.0043], device='cuda:0'), grad: tensor([-0.0001, -0.0001, 0.0011, -0.0037, 0.0018, 0.0002, -0.0034, 0.0008, + 0.0037, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 281.69, cls_loss 0.0992 cls_loss_mapping 0.1439 cls_loss_causal 1.2044 re_mapping 0.0564 re_causal 0.1344 /// teacc 97.50 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0057, -0.0294, -0.0289, ..., -0.0092, 0.0536, 0.0461], + [-0.0388, -0.0407, -0.0449, ..., -0.0301, -0.0292, -0.0375], + [-0.0148, -0.0214, 0.0277, ..., -0.0197, 0.0453, -0.0176], + ..., + [-0.0293, 0.0140, 0.0274, ..., 0.0341, -0.0124, -0.0287], + [-0.0545, 0.0268, -0.0502, ..., 0.0095, -0.0253, -0.0255], + [ 0.0226, 0.0111, -0.0275, ..., 0.0163, -0.0221, 0.0109]], + device='cuda:0'), grad: tensor([[-0.0026, -0.0006, 0.0001, ..., 0.0009, -0.0057, -0.0052], + [ 0.0046, 0.0027, 0.0001, ..., 0.0015, 0.0008, 0.0007], + [ 0.0004, 0.0009, 0.0004, ..., 0.0001, -0.0009, -0.0006], + ..., + [ 0.0015, -0.0027, -0.0001, ..., -0.0011, 0.0005, 0.0003], + [ 0.0059, 0.0044, 0.0001, ..., 0.0035, 0.0008, 0.0010], + [ 0.0037, 0.0028, 0.0005, ..., 0.0029, 0.0013, 0.0016]], + device='cuda:0') +Epoch 10, bias, value: tensor([-0.0138, -0.0195, -0.0171, -0.0286, -0.0304, -0.0011, 0.0269, -0.0116, + 0.0263, 0.0039], device='cuda:0'), grad: tensor([-0.0038, 0.0088, -0.0031, 0.0168, -0.0009, -0.0331, 0.0011, 0.0007, + 0.0072, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 281.41, cls_loss 0.0994 cls_loss_mapping 0.1390 cls_loss_causal 1.1666 re_mapping 0.0513 re_causal 0.1208 /// teacc 97.74 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0060, -0.0299, -0.0298, ..., -0.0092, 0.0553, 0.0474], + [-0.0403, -0.0424, -0.0480, ..., -0.0307, -0.0322, -0.0386], + [-0.0162, -0.0218, 0.0302, ..., -0.0205, 0.0467, -0.0172], + ..., + [-0.0304, 0.0149, 0.0290, ..., 0.0350, -0.0137, -0.0298], + [-0.0559, 0.0270, -0.0550, ..., 0.0100, -0.0257, -0.0255], + [ 0.0228, 0.0103, -0.0300, ..., 0.0160, -0.0222, 0.0105]], + device='cuda:0'), grad: tensor([[ 1.1911e-03, 1.7395e-03, 6.1607e-04, ..., -7.6008e-04, + 2.2304e-04, -1.7090e-03], + [ 2.6441e-04, 7.8201e-04, 8.9526e-05, ..., 4.4799e-04, + -5.9426e-05, -2.5797e-04], + [ 1.3046e-03, -5.6152e-03, -7.3242e-04, ..., 1.2197e-03, + -1.3664e-02, 7.8773e-04], + ..., + [ 3.5739e-04, 5.3215e-04, -5.3942e-05, ..., -9.1982e-04, + 1.3084e-03, 3.5310e-04], + [ 5.1193e-03, 3.9749e-03, 1.5440e-03, ..., 1.4076e-03, + 1.6088e-03, 3.7241e-04], + [-2.4612e-02, -4.1509e-04, -7.1716e-03, ..., 1.3113e-03, + 2.7800e-04, 3.8576e-04]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0136, -0.0197, -0.0168, -0.0286, -0.0305, -0.0013, 0.0268, -0.0115, + 0.0263, 0.0038], device='cuda:0'), grad: tensor([ 0.0028, -0.0005, -0.0145, 0.0050, 0.0034, 0.0058, 0.0110, 0.0021, + 0.0101, -0.0251], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 265.14, cls_loss 0.0918 cls_loss_mapping 0.1268 cls_loss_causal 1.1186 re_mapping 0.0481 re_causal 0.1114 /// teacc 97.62 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0072, -0.0308, -0.0315, ..., -0.0098, 0.0561, 0.0478], + [-0.0419, -0.0449, -0.0511, ..., -0.0315, -0.0340, -0.0388], + [-0.0180, -0.0221, 0.0321, ..., -0.0212, 0.0479, -0.0170], + ..., + [-0.0307, 0.0159, 0.0284, ..., 0.0360, -0.0144, -0.0304], + [-0.0582, 0.0277, -0.0586, ..., 0.0104, -0.0271, -0.0261], + [ 0.0233, 0.0086, -0.0323, ..., 0.0156, -0.0222, 0.0103]], + device='cuda:0'), grad: tensor([[ 2.7752e-04, 4.7541e-04, 5.3734e-05, ..., 1.3363e-04, + -1.8448e-05, 1.7136e-05], + [ 1.5879e-04, 8.2111e-04, -1.7142e-04, ..., 4.9204e-05, + 2.1189e-05, 1.7691e-04], + [ 1.5056e-04, 7.0076e-03, 9.8586e-05, ..., 6.6996e-04, + 4.2367e-04, 8.3780e-04], + ..., + [ 6.7234e-04, 1.0826e-02, 2.0409e-04, ..., 2.4853e-03, + 7.4767e-06, 1.6813e-03], + [ 8.7023e-04, 2.4834e-03, 1.3018e-04, ..., 8.6641e-04, + 1.0425e-04, 3.4237e-04], + [-8.8587e-06, -9.2602e-04, -4.1628e-04, ..., -3.2673e-03, + 4.3035e-05, 1.3018e-04]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0142, -0.0200, -0.0168, -0.0285, -0.0304, -0.0012, 0.0268, -0.0111, + 0.0264, 0.0037], device='cuda:0'), grad: tensor([ 0.0007, -0.0018, 0.0068, -0.0216, 0.0044, 0.0007, -0.0002, 0.0138, + 0.0045, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 281.92, cls_loss 0.0749 cls_loss_mapping 0.1097 cls_loss_causal 1.1203 re_mapping 0.0461 re_causal 0.1116 /// teacc 97.76 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0080, -0.0312, -0.0324, ..., -0.0103, 0.0576, 0.0488], + [-0.0431, -0.0461, -0.0549, ..., -0.0321, -0.0366, -0.0395], + [-0.0196, -0.0220, 0.0354, ..., -0.0217, 0.0494, -0.0164], + ..., + [-0.0308, 0.0164, 0.0267, ..., 0.0368, -0.0167, -0.0316], + [-0.0593, 0.0284, -0.0611, ..., 0.0110, -0.0271, -0.0261], + [ 0.0238, 0.0077, -0.0325, ..., 0.0152, -0.0227, 0.0099]], + device='cuda:0'), grad: tensor([[ 1.4830e-04, -2.0361e-04, 6.0111e-05, ..., -2.1410e-04, + -2.2736e-03, -2.0027e-03], + [ 8.8654e-03, 4.7684e-03, 1.3847e-03, ..., 3.6073e-04, + 2.8286e-03, 1.9140e-03], + [ 2.4021e-04, 2.7275e-04, -2.1763e-03, ..., 8.3542e-04, + -3.2215e-03, -2.3804e-03], + ..., + [ 4.5156e-04, -2.9354e-03, -3.3927e-04, ..., -1.9722e-03, + 4.6611e-04, 3.6907e-04], + [ 1.0786e-03, -4.7684e-03, 1.6797e-04, ..., -2.7618e-03, + 3.1066e-04, 3.1757e-04], + [-5.4806e-05, 4.3907e-03, -5.7030e-04, ..., 2.4376e-03, + 2.5225e-04, 3.1352e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0141, -0.0198, -0.0166, -0.0284, -0.0303, -0.0012, 0.0266, -0.0115, + 0.0267, 0.0034], device='cuda:0'), grad: tensor([-0.0012, 0.0144, -0.0036, 0.0174, -0.0041, -0.0256, 0.0018, -0.0023, + -0.0058, 0.0091], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 282.95, cls_loss 0.0668 cls_loss_mapping 0.1031 cls_loss_causal 1.0754 re_mapping 0.0426 re_causal 0.1025 /// teacc 97.88 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0083, -0.0318, -0.0333, ..., -0.0105, 0.0589, 0.0497], + [-0.0443, -0.0477, -0.0565, ..., -0.0329, -0.0368, -0.0391], + [-0.0204, -0.0220, 0.0367, ..., -0.0222, 0.0498, -0.0164], + ..., + [-0.0313, 0.0173, 0.0275, ..., 0.0377, -0.0179, -0.0328], + [-0.0612, 0.0295, -0.0643, ..., 0.0112, -0.0270, -0.0261], + [ 0.0238, 0.0067, -0.0334, ..., 0.0147, -0.0218, 0.0100]], + device='cuda:0'), grad: tensor([[ 1.0672e-03, 8.1778e-04, 2.6298e-04, ..., 1.4839e-03, + 1.7238e-04, 1.6046e-04], + [ 2.6035e-04, 3.4332e-04, 1.2457e-04, ..., 3.6597e-04, + 3.1734e-04, 1.2732e-04], + [ 8.8036e-05, 7.3929e-03, 6.3276e-04, ..., 5.8937e-03, + 1.5364e-03, 5.2547e-04], + ..., + [-4.7708e-04, -1.0117e-02, -1.6041e-03, ..., -9.0637e-03, + -3.1185e-03, -9.1648e-04], + [ 6.3419e-04, -8.1539e-05, 1.8799e-04, ..., -4.4107e-04, + 4.7350e-04, -1.5363e-05], + [-1.7252e-03, 3.9864e-04, 1.1921e-04, ..., -5.5809e-03, + 2.3401e-04, 1.8728e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0142, -0.0196, -0.0165, -0.0288, -0.0305, -0.0015, 0.0266, -0.0113, + 0.0271, 0.0035], device='cuda:0'), grad: tensor([ 0.0017, -0.0028, 0.0081, 0.0054, 0.0240, -0.0030, -0.0044, -0.0093, + 0.0002, -0.0200], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 265.57, cls_loss 0.0709 cls_loss_mapping 0.0945 cls_loss_causal 1.0394 re_mapping 0.0408 re_causal 0.0975 /// teacc 97.69 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0091, -0.0324, -0.0323, ..., -0.0109, 0.0604, 0.0506], + [-0.0450, -0.0492, -0.0588, ..., -0.0338, -0.0384, -0.0395], + [-0.0216, -0.0226, 0.0385, ..., -0.0234, 0.0510, -0.0162], + ..., + [-0.0316, 0.0181, 0.0276, ..., 0.0387, -0.0195, -0.0336], + [-0.0627, 0.0293, -0.0656, ..., 0.0120, -0.0271, -0.0262], + [ 0.0238, 0.0054, -0.0350, ..., 0.0143, -0.0222, 0.0096]], + device='cuda:0'), grad: tensor([[ 3.2127e-05, 7.0453e-05, 3.2693e-05, ..., 1.7524e-05, + -1.4992e-03, -1.2417e-03], + [ 1.6749e-05, 1.5316e-03, 5.8079e-04, ..., 6.3019e-03, + 4.1294e-04, -1.6555e-05], + [ 2.8148e-05, -3.1567e-04, -9.8324e-04, ..., 3.6550e-04, + -9.9850e-04, 2.0236e-05], + ..., + [ 1.4067e-04, -4.0283e-03, 3.2353e-04, ..., -1.4124e-03, + 4.3273e-04, 1.1021e-04], + [ 3.8481e-04, -1.2112e-03, 6.7055e-05, ..., -7.7286e-03, + 3.2806e-04, 2.4021e-04], + [-1.4853e-04, -5.0402e-04, 2.7776e-04, ..., -5.8317e-04, + 3.4952e-04, 2.8372e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0140, -0.0199, -0.0164, -0.0286, -0.0305, -0.0016, 0.0263, -0.0111, + 0.0274, 0.0032], device='cuda:0'), grad: tensor([-0.0012, 0.0095, -0.0009, 0.0043, 0.0003, 0.0013, 0.0006, -0.0017, + -0.0113, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 265.06, cls_loss 0.0640 cls_loss_mapping 0.0867 cls_loss_causal 1.0318 re_mapping 0.0367 re_causal 0.0886 /// teacc 97.86 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0097, -0.0331, -0.0333, ..., -0.0108, 0.0616, 0.0515], + [-0.0463, -0.0507, -0.0607, ..., -0.0348, -0.0401, -0.0395], + [-0.0226, -0.0226, 0.0402, ..., -0.0242, 0.0522, -0.0159], + ..., + [-0.0321, 0.0188, 0.0285, ..., 0.0393, -0.0204, -0.0344], + [-0.0651, 0.0297, -0.0685, ..., 0.0120, -0.0280, -0.0273], + [ 0.0243, 0.0048, -0.0349, ..., 0.0140, -0.0229, 0.0089]], + device='cuda:0'), grad: tensor([[-2.9922e-04, -3.2187e-04, 8.2254e-05, ..., -5.1498e-04, + -4.7379e-03, -3.7441e-03], + [ 5.4747e-05, 8.0490e-04, 4.3058e-04, ..., 3.7622e-04, + 1.1069e-04, -7.5388e-04], + [ 6.7234e-05, 8.2550e-03, 3.0479e-03, ..., 3.6669e-04, + -9.0265e-04, -1.6117e-04], + ..., + [ 4.0102e-04, 6.3248e-03, 2.6817e-03, ..., -1.5306e-03, + 8.9645e-04, 8.5890e-05], + [ 7.3576e-04, 1.7080e-03, 1.7738e-04, ..., 1.0729e-03, + 2.7227e-04, 5.3644e-04], + [-7.4744e-05, 8.8549e-04, 3.7909e-04, ..., -8.3447e-04, + 3.6192e-04, 3.6955e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0141, -0.0198, -0.0164, -0.0287, -0.0308, -0.0009, 0.0264, -0.0111, + 0.0270, 0.0032], device='cuda:0'), grad: tensor([-0.0041, -0.0013, 0.0087, -0.0195, 0.0013, 0.0010, 0.0038, 0.0065, + 0.0032, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 265.54, cls_loss 0.0576 cls_loss_mapping 0.0810 cls_loss_causal 0.9884 re_mapping 0.0355 re_causal 0.0873 /// teacc 97.85 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0107, -0.0340, -0.0344, ..., -0.0114, 0.0624, 0.0521], + [-0.0473, -0.0520, -0.0621, ..., -0.0352, -0.0418, -0.0396], + [-0.0239, -0.0223, 0.0421, ..., -0.0247, 0.0536, -0.0153], + ..., + [-0.0323, 0.0190, 0.0280, ..., 0.0398, -0.0211, -0.0352], + [-0.0672, 0.0301, -0.0698, ..., 0.0126, -0.0284, -0.0276], + [ 0.0245, 0.0038, -0.0352, ..., 0.0135, -0.0228, 0.0087]], + device='cuda:0'), grad: tensor([[ 7.4148e-04, 1.5211e-04, 5.4479e-05, ..., 2.4509e-04, + -1.0252e-03, -8.2874e-04], + [ 8.4162e-05, 1.1683e-04, 4.3541e-05, ..., 1.3924e-04, + 3.9428e-05, 1.4946e-05], + [ 1.6582e-04, 5.5313e-04, 3.8218e-04, ..., 6.0225e-04, + 3.4833e-04, 3.8362e-04], + ..., + [-2.2717e-03, -3.6049e-03, -7.1335e-04, ..., -4.6654e-03, + -3.4451e-05, -2.2030e-04], + [ 2.2686e-04, -4.1544e-05, 7.2837e-05, ..., -2.5034e-04, + 2.3580e-04, 3.2783e-05], + [-7.3862e-04, 6.4468e-04, 6.2275e-04, ..., 1.6603e-03, + 4.7064e-04, 1.9741e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0146, -0.0200, -0.0159, -0.0287, -0.0307, -0.0006, 0.0262, -0.0115, + 0.0271, 0.0033], device='cuda:0'), grad: tensor([ 1.5342e-04, -5.4665e-03, 3.0155e-03, 7.4234e-03, -8.9169e-05, + -1.4706e-03, 1.3084e-03, -8.3694e-03, 3.7760e-05, 3.4618e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 282.44, cls_loss 0.0571 cls_loss_mapping 0.0798 cls_loss_causal 0.9809 re_mapping 0.0340 re_causal 0.0860 /// teacc 98.08 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0110, -0.0343, -0.0353, ..., -0.0113, 0.0640, 0.0532], + [-0.0481, -0.0529, -0.0629, ..., -0.0359, -0.0433, -0.0400], + [-0.0247, -0.0228, 0.0419, ..., -0.0257, 0.0541, -0.0153], + ..., + [-0.0325, 0.0199, 0.0305, ..., 0.0408, -0.0220, -0.0357], + [-0.0688, 0.0308, -0.0699, ..., 0.0129, -0.0285, -0.0277], + [ 0.0247, 0.0030, -0.0367, ..., 0.0130, -0.0229, 0.0085]], + device='cuda:0'), grad: tensor([[ 2.3353e-04, -1.3494e-04, 1.1712e-04, ..., 1.3173e-04, + -1.4181e-03, -9.3794e-04], + [ 1.2863e-04, 3.7813e-04, 7.9513e-05, ..., 1.4830e-04, + 2.9588e-04, -4.6968e-05], + [ 1.9350e-03, 3.5076e-03, 1.1645e-05, ..., 7.7820e-04, + 1.9779e-03, 3.0732e-04], + ..., + [ 1.3411e-04, -1.3866e-03, -3.0971e-04, ..., -7.6342e-04, + 2.0206e-04, 6.9559e-05], + [ 2.9230e-04, -9.2077e-04, 8.8096e-05, ..., -6.7043e-04, + -7.9060e-04, -6.0618e-05], + [ 1.8227e-04, -9.1493e-05, -4.8232e-04, ..., -1.4424e-04, + 5.1069e-04, 4.4274e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0139, -0.0200, -0.0165, -0.0290, -0.0306, -0.0008, 0.0257, -0.0112, + 0.0278, 0.0030], device='cuda:0'), grad: tensor([-0.0008, -0.0004, 0.0059, -0.0030, 0.0007, -0.0004, 0.0008, -0.0005, + -0.0017, -0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 265.80, cls_loss 0.0578 cls_loss_mapping 0.0747 cls_loss_causal 0.9653 re_mapping 0.0321 re_causal 0.0793 /// teacc 97.83 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0115, -0.0346, -0.0362, ..., -0.0118, 0.0646, 0.0537], + [-0.0493, -0.0542, -0.0661, ..., -0.0367, -0.0446, -0.0405], + [-0.0255, -0.0233, 0.0436, ..., -0.0266, 0.0558, -0.0149], + ..., + [-0.0331, 0.0203, 0.0304, ..., 0.0417, -0.0232, -0.0364], + [-0.0702, 0.0316, -0.0708, ..., 0.0133, -0.0289, -0.0282], + [ 0.0250, 0.0025, -0.0362, ..., 0.0127, -0.0235, 0.0081]], + device='cuda:0'), grad: tensor([[-1.0228e-04, 1.2040e-04, 2.4331e-04, ..., 2.0587e-04, + -1.6756e-03, -1.8559e-03], + [ 5.5164e-05, 4.7708e-04, 2.9135e-04, ..., 8.1491e-04, + 1.2362e-04, 3.7521e-05], + [-1.3649e-04, -1.3857e-03, -1.9798e-03, ..., -3.0975e-03, + -1.0994e-02, -3.7346e-03], + ..., + [-1.0055e-04, -5.2881e-04, -8.5640e-04, ..., 2.3499e-03, + 4.2419e-03, 3.4313e-03], + [ 9.1887e-04, 3.6812e-04, -1.4544e-04, ..., -2.7580e-03, + 4.5538e-04, 3.5357e-04], + [-4.2677e-04, -1.6677e-04, -2.3805e-06, ..., -1.2070e-04, + 3.7479e-04, 1.9300e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0140, -0.0204, -0.0158, -0.0290, -0.0310, -0.0008, 0.0260, -0.0112, + 0.0281, 0.0027], device='cuda:0'), grad: tensor([-0.0008, 0.0029, -0.0085, 0.0013, 0.0049, 0.0067, 0.0026, 0.0011, + -0.0097, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 265.19, cls_loss 0.0517 cls_loss_mapping 0.0738 cls_loss_causal 0.9729 re_mapping 0.0319 re_causal 0.0811 /// teacc 97.87 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0127, -0.0353, -0.0372, ..., -0.0121, 0.0654, 0.0541], + [-0.0512, -0.0553, -0.0675, ..., -0.0374, -0.0453, -0.0408], + [-0.0273, -0.0239, 0.0439, ..., -0.0278, 0.0565, -0.0148], + ..., + [-0.0339, 0.0207, 0.0295, ..., 0.0423, -0.0244, -0.0375], + [-0.0712, 0.0313, -0.0719, ..., 0.0136, -0.0297, -0.0284], + [ 0.0256, 0.0015, -0.0367, ..., 0.0123, -0.0232, 0.0084]], + device='cuda:0'), grad: tensor([[-1.0097e-04, 3.1972e-04, 3.2330e-04, ..., 3.2115e-04, + 1.9569e-03, 7.4768e-04], + [ 5.6833e-05, 6.4754e-04, 2.2784e-05, ..., 2.9445e-04, + 1.2469e-04, 9.7454e-05], + [ 8.6904e-05, -1.6081e-04, -3.7026e-04, ..., 2.0623e-04, + -3.1891e-03, -1.6003e-03], + ..., + [ 1.1134e-04, -5.0049e-03, 1.0639e-04, ..., -2.5291e-03, + 3.4809e-04, 1.1009e-04], + [ 4.6825e-04, -8.8751e-05, 9.2208e-05, ..., 2.4843e-04, + -1.6284e-04, -3.2926e-04], + [-6.8903e-04, -2.9254e-04, -2.4050e-05, ..., -5.8365e-04, + 1.9264e-04, 1.8930e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0145, -0.0205, -0.0162, -0.0283, -0.0308, -0.0008, 0.0258, -0.0115, + 0.0282, 0.0028], device='cuda:0'), grad: tensor([ 0.0017, 0.0004, -0.0016, 0.0063, 0.0006, 0.0006, -0.0002, -0.0034, + 0.0007, -0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 265.07, cls_loss 0.0492 cls_loss_mapping 0.0693 cls_loss_causal 0.9366 re_mapping 0.0311 re_causal 0.0800 /// teacc 98.00 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0122, -0.0357, -0.0349, ..., -0.0124, 0.0668, 0.0552], + [-0.0526, -0.0569, -0.0690, ..., -0.0380, -0.0468, -0.0414], + [-0.0284, -0.0242, 0.0451, ..., -0.0287, 0.0573, -0.0146], + ..., + [-0.0339, 0.0211, 0.0294, ..., 0.0430, -0.0256, -0.0381], + [-0.0725, 0.0322, -0.0730, ..., 0.0140, -0.0295, -0.0283], + [ 0.0254, 0.0007, -0.0377, ..., 0.0119, -0.0233, 0.0079]], + device='cuda:0'), grad: tensor([[-2.9106e-03, -1.7376e-03, 3.3639e-06, ..., 1.6674e-05, + -3.6526e-03, -2.6703e-03], + [ 5.2786e-04, 1.2608e-03, 3.9190e-06, ..., 6.0034e-04, + 3.3170e-05, 1.2316e-05], + [ 2.4796e-04, 1.0881e-03, 6.6459e-05, ..., 4.3178e-04, + 4.2582e-04, 3.3855e-04], + ..., + [ 3.0422e-04, -2.1725e-03, -2.0707e-04, ..., -1.1654e-03, + 1.4114e-04, 1.1230e-04], + [ 6.0749e-04, -2.4319e-03, 4.5411e-06, ..., -1.3628e-03, + 2.5415e-04, 2.1744e-04], + [-2.3937e-03, 1.9140e-03, 2.5839e-05, ..., 8.4114e-04, + 1.4324e-03, 9.8896e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0138, -0.0204, -0.0161, -0.0281, -0.0309, -0.0015, 0.0260, -0.0113, + 0.0285, 0.0024], device='cuda:0'), grad: tensor([-0.0040, 0.0057, 0.0019, 0.0030, 0.0070, -0.0004, 0.0004, -0.0007, + -0.0080, -0.0048], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 281.30, cls_loss 0.0417 cls_loss_mapping 0.0594 cls_loss_causal 0.8965 re_mapping 0.0301 re_causal 0.0776 /// teacc 98.20 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0130, -0.0363, -0.0351, ..., -0.0127, 0.0678, 0.0559], + [-0.0537, -0.0582, -0.0701, ..., -0.0384, -0.0479, -0.0414], + [-0.0289, -0.0244, 0.0459, ..., -0.0294, 0.0582, -0.0141], + ..., + [-0.0341, 0.0221, 0.0292, ..., 0.0437, -0.0262, -0.0391], + [-0.0737, 0.0329, -0.0729, ..., 0.0142, -0.0297, -0.0284], + [ 0.0262, -0.0002, -0.0380, ..., 0.0118, -0.0236, 0.0075]], + device='cuda:0'), grad: tensor([[ 9.3889e-04, 8.0645e-05, 3.8505e-05, ..., 7.6890e-05, + 2.7275e-04, 3.1447e-04], + [ 2.2352e-05, 4.8459e-05, 6.3896e-05, ..., 1.6361e-05, + 7.4744e-05, -1.4508e-04], + [ 5.1945e-05, -4.6206e-04, -1.5993e-03, ..., 1.2502e-05, + -1.0929e-03, 2.0349e-04], + ..., + [ 2.9027e-05, 6.9797e-05, 5.3358e-04, ..., -7.2062e-05, + 4.1699e-04, 2.9266e-05], + [ 1.6057e-04, 3.9846e-05, 4.0799e-05, ..., 4.6492e-05, + 1.4389e-04, 1.1462e-04], + [-6.3002e-05, 4.2081e-05, 3.9428e-05, ..., -5.2750e-05, + 1.5533e-04, 9.2089e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0138, -0.0205, -0.0160, -0.0285, -0.0311, -0.0014, 0.0256, -0.0109, + 0.0285, 0.0025], device='cuda:0'), grad: tensor([ 5.2500e-04, -2.5806e-03, -1.3514e-03, 1.0576e-03, 1.5154e-03, + 4.9973e-04, -1.0567e-03, 9.6035e-04, 4.4799e-04, -2.0161e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 265.82, cls_loss 0.0366 cls_loss_mapping 0.0546 cls_loss_causal 0.8627 re_mapping 0.0293 re_causal 0.0739 /// teacc 98.10 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0133, -0.0367, -0.0352, ..., -0.0130, 0.0688, 0.0566], + [-0.0547, -0.0592, -0.0716, ..., -0.0392, -0.0492, -0.0416], + [-0.0298, -0.0247, 0.0469, ..., -0.0302, 0.0594, -0.0133], + ..., + [-0.0347, 0.0225, 0.0289, ..., 0.0447, -0.0272, -0.0402], + [-0.0753, 0.0330, -0.0731, ..., 0.0142, -0.0299, -0.0290], + [ 0.0267, -0.0007, -0.0379, ..., 0.0115, -0.0239, 0.0069]], + device='cuda:0'), grad: tensor([[ 3.0851e-04, 1.6403e-04, 4.5896e-05, ..., 1.1301e-04, + -4.5228e-04, -3.0017e-04], + [ 7.6413e-05, 6.7472e-05, -1.1957e-04, ..., 3.5644e-05, + 3.4034e-05, -3.4869e-05], + [ 3.4142e-03, 2.5654e-04, -1.7395e-03, ..., 4.3106e-04, + -1.0562e-04, -7.8392e-04], + ..., + [ 7.8440e-05, 1.1196e-03, 3.4928e-04, ..., 7.2193e-04, + 1.5373e-03, 1.3628e-03], + [ 3.2406e-03, 3.1829e-04, 6.7759e-04, ..., 3.9291e-03, + 3.5048e-04, 9.1136e-05], + [-1.2231e-04, 1.0175e-04, 2.1219e-04, ..., -3.9876e-05, + 3.4165e-04, 1.0741e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0137, -0.0209, -0.0156, -0.0286, -0.0314, -0.0013, 0.0255, -0.0110, + 0.0287, 0.0026], device='cuda:0'), grad: tensor([ 0.0002, -0.0009, 0.0019, -0.0095, -0.0003, 0.0002, 0.0004, 0.0022, + 0.0052, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 282.33, cls_loss 0.0412 cls_loss_mapping 0.0602 cls_loss_causal 0.8785 re_mapping 0.0275 re_causal 0.0709 /// teacc 98.34 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0137, -0.0374, -0.0356, ..., -0.0132, 0.0696, 0.0573], + [-0.0556, -0.0607, -0.0726, ..., -0.0393, -0.0502, -0.0415], + [-0.0311, -0.0256, 0.0476, ..., -0.0314, 0.0602, -0.0130], + ..., + [-0.0358, 0.0234, 0.0304, ..., 0.0455, -0.0282, -0.0408], + [-0.0767, 0.0336, -0.0737, ..., 0.0144, -0.0300, -0.0294], + [ 0.0271, -0.0016, -0.0381, ..., 0.0111, -0.0233, 0.0071]], + device='cuda:0'), grad: tensor([[ 1.2565e-04, 4.5359e-05, 2.7642e-05, ..., 5.3756e-06, + 4.8786e-05, 4.8757e-05], + [ 1.1563e-04, 2.6584e-04, 7.7724e-04, ..., 1.0306e-04, + 3.6001e-04, 1.0133e-04], + [ 3.5167e-05, -1.5091e-02, -3.3989e-03, ..., 4.0978e-05, + -5.6124e-04, -2.2054e-04], + ..., + [-2.4605e-03, -5.5075e-04, 2.2221e-04, ..., -2.7828e-03, + 7.7665e-05, 3.8087e-05], + [ 1.3101e-04, 4.4250e-03, 7.4673e-04, ..., -4.8429e-06, + 9.8273e-06, -2.8953e-05], + [ 2.2945e-03, 1.0462e-03, -1.8269e-05, ..., 2.0466e-03, + 7.0691e-05, 6.5386e-05]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0139, -0.0201, -0.0162, -0.0283, -0.0315, -0.0014, 0.0254, -0.0110, + 0.0288, 0.0025], device='cuda:0'), grad: tensor([ 3.6669e-04, 4.7302e-04, -1.6617e-02, 7.6637e-03, 1.4343e-03, + 2.6588e-03, 1.0133e-05, -6.0463e-03, 3.9215e-03, 6.1378e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 265.34, cls_loss 0.0390 cls_loss_mapping 0.0519 cls_loss_causal 0.8758 re_mapping 0.0267 re_causal 0.0688 /// teacc 98.29 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0140, -0.0380, -0.0361, ..., -0.0132, 0.0703, 0.0578], + [-0.0564, -0.0619, -0.0744, ..., -0.0400, -0.0524, -0.0423], + [-0.0318, -0.0255, 0.0495, ..., -0.0321, 0.0614, -0.0122], + ..., + [-0.0358, 0.0238, 0.0305, ..., 0.0459, -0.0289, -0.0415], + [-0.0783, 0.0337, -0.0752, ..., 0.0146, -0.0306, -0.0296], + [ 0.0279, -0.0022, -0.0387, ..., 0.0108, -0.0239, 0.0065]], + device='cuda:0'), grad: tensor([[ 2.0683e-04, 5.2214e-04, 1.5587e-05, ..., 3.4839e-05, + 1.4524e-03, 1.2064e-03], + [ 1.7011e-04, 2.3162e-04, 6.7055e-05, ..., 6.6876e-05, + 3.8791e-04, 2.4235e-04], + [ 5.4598e-04, 1.1005e-03, 5.7489e-05, ..., 2.0778e-04, + 5.7840e-04, 4.3511e-04], + ..., + [ 1.8612e-05, -7.9012e-04, -4.6611e-05, ..., -7.7772e-04, + 6.2466e-05, 8.1301e-05], + [ 1.2541e-04, -2.3448e-04, 2.9221e-05, ..., -7.2658e-05, + -2.1038e-03, -1.7824e-03], + [ 8.1444e-04, 9.0599e-04, 1.3041e-04, ..., 2.6965e-04, + 1.8954e-04, 1.8978e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0140, -0.0199, -0.0163, -0.0284, -0.0313, -0.0012, 0.0250, -0.0110, + 0.0288, 0.0027], device='cuda:0'), grad: tensor([ 0.0036, 0.0008, 0.0023, -0.0024, 0.0004, 0.0007, -0.0018, -0.0008, + -0.0045, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 265.08, cls_loss 0.0397 cls_loss_mapping 0.0607 cls_loss_causal 0.8547 re_mapping 0.0276 re_causal 0.0704 /// teacc 98.30 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0142, -0.0382, -0.0359, ..., -0.0133, 0.0714, 0.0586], + [-0.0572, -0.0626, -0.0758, ..., -0.0400, -0.0542, -0.0422], + [-0.0328, -0.0261, 0.0500, ..., -0.0327, 0.0625, -0.0119], + ..., + [-0.0361, 0.0240, 0.0302, ..., 0.0469, -0.0300, -0.0427], + [-0.0795, 0.0342, -0.0759, ..., 0.0149, -0.0309, -0.0299], + [ 0.0280, -0.0030, -0.0394, ..., 0.0102, -0.0242, 0.0060]], + device='cuda:0'), grad: tensor([[ 2.7657e-04, -5.1558e-05, 9.2834e-06, ..., 1.2159e-04, + 2.9964e-03, 1.7128e-03], + [ 6.0380e-05, 1.0687e-04, 1.6674e-05, ..., 7.5042e-05, + 3.0541e-04, 1.8775e-04], + [ 3.5286e-05, -1.7357e-03, -1.6189e-04, ..., 3.2997e-04, + -7.0152e-03, -3.6488e-03], + ..., + [ 4.4775e-04, -8.9645e-04, 2.2912e-04, ..., -9.5129e-05, + 2.2030e-04, -1.0991e-04], + [ 6.8760e-04, 1.0710e-03, 1.6466e-05, ..., 5.8079e-04, + 1.0767e-03, 6.8378e-04], + [ 2.8305e-03, -1.2636e-04, 2.3305e-04, ..., 4.1351e-03, + 2.9683e-04, 2.1410e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0137, -0.0201, -0.0158, -0.0282, -0.0315, -0.0015, 0.0251, -0.0109, + 0.0288, 0.0022], device='cuda:0'), grad: tensor([ 0.0022, 0.0002, -0.0067, 0.0040, -0.0171, 0.0020, -0.0024, 0.0015, + 0.0034, 0.0129], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 264.81, cls_loss 0.0334 cls_loss_mapping 0.0535 cls_loss_causal 0.8335 re_mapping 0.0263 re_causal 0.0676 /// teacc 98.14 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0149, -0.0387, -0.0364, ..., -0.0136, 0.0719, 0.0590], + [-0.0579, -0.0646, -0.0763, ..., -0.0412, -0.0545, -0.0419], + [-0.0333, -0.0264, 0.0509, ..., -0.0331, 0.0633, -0.0114], + ..., + [-0.0367, 0.0249, 0.0300, ..., 0.0477, -0.0311, -0.0435], + [-0.0811, 0.0343, -0.0764, ..., 0.0151, -0.0313, -0.0304], + [ 0.0285, -0.0034, -0.0400, ..., 0.0100, -0.0243, 0.0056]], + device='cuda:0'), grad: tensor([[ 1.4558e-05, 4.0829e-05, 4.6998e-05, ..., 4.0144e-05, + 1.5080e-04, -3.7074e-05], + [ 1.2949e-05, 5.7101e-05, 1.0198e-04, ..., 1.4257e-04, + 3.4690e-05, -1.0028e-05], + [ 1.8001e-05, 6.5625e-05, 1.0714e-05, ..., 1.1075e-04, + 1.2577e-04, -2.1219e-05], + ..., + [-1.1347e-05, -1.3542e-04, 7.9036e-05, ..., -1.9908e-04, + 1.5008e-04, 2.7716e-05], + [ 5.8383e-05, -2.8396e-04, 5.3167e-05, ..., -2.9945e-04, + -6.1274e-05, -1.2733e-05], + [-2.2972e-04, -2.9254e-04, 6.8605e-05, ..., 1.4651e-04, + 3.1614e-04, 9.8944e-06]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0139, -0.0202, -0.0154, -0.0284, -0.0311, -0.0016, 0.0253, -0.0108, + 0.0284, 0.0023], device='cuda:0'), grad: tensor([ 0.0004, 0.0005, 0.0005, 0.0006, -0.0032, 0.0004, 0.0007, 0.0005, + -0.0002, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 282.30, cls_loss 0.0337 cls_loss_mapping 0.0485 cls_loss_causal 0.8751 re_mapping 0.0241 re_causal 0.0648 /// teacc 98.40 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0153, -0.0391, -0.0365, ..., -0.0137, 0.0729, 0.0597], + [-0.0588, -0.0654, -0.0776, ..., -0.0415, -0.0553, -0.0420], + [-0.0339, -0.0275, 0.0512, ..., -0.0344, 0.0642, -0.0114], + ..., + [-0.0372, 0.0256, 0.0307, ..., 0.0487, -0.0319, -0.0441], + [-0.0823, 0.0349, -0.0768, ..., 0.0152, -0.0313, -0.0302], + [ 0.0287, -0.0036, -0.0405, ..., 0.0095, -0.0249, 0.0051]], + device='cuda:0'), grad: tensor([[-1.0042e-03, -6.8235e-04, 2.0027e-05, ..., -2.8553e-03, + -5.8556e-04, -1.3151e-03], + [ 3.1680e-05, 6.6936e-05, 1.0423e-05, ..., 5.3495e-05, + 2.2098e-05, 1.3724e-05], + [ 2.7135e-05, 8.2552e-05, 4.6223e-05, ..., 9.2506e-05, + 8.1301e-05, 3.6657e-05], + ..., + [ 5.8222e-04, -7.0953e-04, 1.4722e-04, ..., 5.4884e-04, + 2.8133e-04, 5.4932e-04], + [ 8.3923e-05, 6.9618e-05, 1.8850e-05, ..., -8.5890e-05, + -7.0632e-05, 7.2896e-05], + [-1.8132e-04, -1.9491e-04, -1.2481e-04, ..., -4.5180e-05, + 1.1134e-04, 3.5673e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0140, -0.0202, -0.0159, -0.0286, -0.0313, -0.0015, 0.0254, -0.0108, + 0.0286, 0.0026], device='cuda:0'), grad: tensor([-2.5196e-03, -4.1223e-04, 2.9564e-04, 8.1158e-04, -1.4865e-04, + 1.4057e-03, 2.6494e-05, 1.0033e-03, -8.6278e-06, -4.5371e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 27---------------------------------------------------- +epoch 27, time 282.18, cls_loss 0.0329 cls_loss_mapping 0.0475 cls_loss_causal 0.8096 re_mapping 0.0249 re_causal 0.0653 /// teacc 98.48 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0153, -0.0395, -0.0373, ..., -0.0139, 0.0739, 0.0604], + [-0.0593, -0.0665, -0.0777, ..., -0.0422, -0.0566, -0.0420], + [-0.0345, -0.0277, 0.0530, ..., -0.0350, 0.0652, -0.0109], + ..., + [-0.0376, 0.0252, 0.0314, ..., 0.0491, -0.0330, -0.0449], + [-0.0827, 0.0351, -0.0782, ..., 0.0157, -0.0319, -0.0306], + [ 0.0290, -0.0046, -0.0413, ..., 0.0090, -0.0257, 0.0046]], + device='cuda:0'), grad: tensor([[ 6.5625e-05, 7.7605e-05, 7.0930e-05, ..., 4.1366e-05, + 4.6849e-04, 2.8062e-04], + [ 2.9340e-05, 9.6738e-05, 9.3654e-06, ..., 6.3181e-05, + 1.1581e-04, -4.3482e-05], + [ 3.2902e-05, 1.1826e-04, -3.9190e-06, ..., 1.5986e-04, + -1.2474e-03, -7.5722e-04], + ..., + [ 6.6161e-05, -9.0075e-04, 3.5405e-05, ..., -7.6342e-04, + 8.1897e-05, 5.1588e-05], + [ 1.9276e-04, 2.7609e-04, 4.1276e-05, ..., -7.2956e-05, + 4.0817e-04, 2.7704e-04], + [-1.6117e-04, 9.6202e-05, -8.8453e-05, ..., 1.2815e-04, + 2.0325e-04, 6.2466e-05]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0137, -0.0195, -0.0159, -0.0282, -0.0313, -0.0013, 0.0248, -0.0111, + 0.0288, 0.0020], device='cuda:0'), grad: tensor([ 0.0007, -0.0008, -0.0006, -0.0001, 0.0004, 0.0004, 0.0004, -0.0008, + 0.0007, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 281.63, cls_loss 0.0341 cls_loss_mapping 0.0507 cls_loss_causal 0.8270 re_mapping 0.0236 re_causal 0.0618 /// teacc 98.52 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0160, -0.0404, -0.0380, ..., -0.0142, 0.0742, 0.0606], + [-0.0602, -0.0682, -0.0791, ..., -0.0433, -0.0571, -0.0420], + [-0.0349, -0.0278, 0.0539, ..., -0.0359, 0.0665, -0.0105], + ..., + [-0.0382, 0.0259, 0.0313, ..., 0.0500, -0.0338, -0.0452], + [-0.0836, 0.0351, -0.0792, ..., 0.0160, -0.0325, -0.0309], + [ 0.0292, -0.0052, -0.0411, ..., 0.0088, -0.0261, 0.0042]], + device='cuda:0'), grad: tensor([[ 2.3997e-04, 4.9591e-05, -5.3978e-04, ..., 8.4102e-05, + -2.6817e-03, -1.6956e-03], + [ 8.3625e-05, 9.1493e-05, 1.1659e-04, ..., 8.0585e-05, + 2.1660e-04, 1.4651e-04], + [ 1.2316e-05, 7.2002e-05, 3.7527e-04, ..., 2.9117e-05, + 1.1606e-03, 6.8092e-04], + ..., + [ 2.4116e-04, -7.4744e-05, 9.3102e-05, ..., -7.4029e-05, + 1.5163e-04, 8.9824e-05], + [ 1.7557e-03, 4.7952e-05, 3.6812e-04, ..., 1.1921e-03, + 4.1866e-04, 4.9210e-04], + [-5.3024e-04, 2.4700e-03, 3.2578e-03, ..., -1.1110e-04, + 8.7976e-05, 5.6714e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0142, -0.0198, -0.0155, -0.0280, -0.0313, -0.0015, 0.0244, -0.0109, + 0.0291, 0.0017], device='cuda:0'), grad: tensor([-0.0012, 0.0001, 0.0010, -0.0148, 0.0014, 0.0131, -0.0146, 0.0005, + 0.0029, 0.0115], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 266.35, cls_loss 0.0318 cls_loss_mapping 0.0457 cls_loss_causal 0.8244 re_mapping 0.0239 re_causal 0.0629 /// teacc 98.47 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0168, -0.0412, -0.0380, ..., -0.0145, 0.0748, 0.0610], + [-0.0611, -0.0696, -0.0803, ..., -0.0443, -0.0587, -0.0429], + [-0.0356, -0.0275, 0.0551, ..., -0.0366, 0.0674, -0.0100], + ..., + [-0.0383, 0.0266, 0.0314, ..., 0.0510, -0.0349, -0.0459], + [-0.0844, 0.0354, -0.0795, ..., 0.0164, -0.0326, -0.0307], + [ 0.0297, -0.0057, -0.0415, ..., 0.0084, -0.0260, 0.0042]], + device='cuda:0'), grad: tensor([[ 4.9472e-05, 3.1531e-05, 1.8096e-06, ..., 3.0756e-05, + 1.5363e-05, 1.1697e-05], + [ 3.5942e-05, 1.2894e-03, 7.1712e-06, ..., 1.6241e-03, + 1.0824e-04, 1.9103e-05], + [ 3.3677e-05, 3.7003e-04, -4.9859e-05, ..., 6.6614e-04, + -3.5429e-04, -1.5402e-04], + ..., + [ 4.9263e-05, -2.3785e-03, 3.6601e-06, ..., -3.2692e-03, + 1.3876e-04, 1.0002e-04], + [-1.3924e-04, -1.7083e-04, 4.0196e-06, ..., 3.8409e-04, + 3.1978e-05, 2.5690e-05], + [ 7.0333e-04, 1.7369e-04, 2.5570e-05, ..., 2.3234e-04, + 8.6101e-07, 7.2550e-07]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0145, -0.0201, -0.0154, -0.0284, -0.0310, -0.0017, 0.0247, -0.0108, + 0.0296, 0.0018], device='cuda:0'), grad: tensor([ 0.0001, 0.0024, 0.0008, 0.0015, -0.0014, -0.0013, 0.0006, -0.0042, + -0.0002, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 281.97, cls_loss 0.0308 cls_loss_mapping 0.0474 cls_loss_causal 0.8714 re_mapping 0.0224 re_causal 0.0631 /// teacc 98.54 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0174, -0.0415, -0.0384, ..., -0.0150, 0.0757, 0.0616], + [-0.0623, -0.0710, -0.0813, ..., -0.0452, -0.0595, -0.0431], + [-0.0357, -0.0278, 0.0556, ..., -0.0373, 0.0680, -0.0098], + ..., + [-0.0391, 0.0273, 0.0316, ..., 0.0517, -0.0357, -0.0464], + [-0.0852, 0.0357, -0.0796, ..., 0.0164, -0.0327, -0.0308], + [ 0.0299, -0.0063, -0.0414, ..., 0.0080, -0.0263, 0.0040]], + device='cuda:0'), grad: tensor([[ 9.3997e-05, 5.1439e-05, 5.5172e-06, ..., -8.8871e-05, + -3.0351e-04, -3.2258e-04], + [ 5.2899e-05, 9.7990e-05, 1.9893e-05, ..., 3.9876e-05, + 1.5533e-04, 5.6535e-05], + [ 4.0948e-05, 6.6943e-06, -8.3029e-05, ..., 2.5094e-05, + -1.6057e-04, -4.2021e-05], + ..., + [ 2.3156e-05, -3.2806e-04, 1.4700e-05, ..., -2.3806e-04, + 9.1612e-05, 9.2685e-05], + [ 8.6260e-04, 6.9332e-04, 8.1286e-06, ..., 6.8724e-05, + 5.2309e-04, 4.6873e-04], + [-1.1384e-04, 3.3915e-05, 7.2271e-06, ..., 1.0300e-04, + 1.5974e-05, 4.2140e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0141, -0.0207, -0.0156, -0.0284, -0.0305, -0.0016, 0.0243, -0.0109, + 0.0300, 0.0015], device='cuda:0'), grad: tensor([ 3.9250e-05, -6.3956e-05, 3.9816e-04, 4.8971e-04, 7.4005e-04, + 1.0376e-03, -6.5918e-03, 9.9063e-05, 4.0779e-03, -2.2602e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 265.20, cls_loss 0.0348 cls_loss_mapping 0.0505 cls_loss_causal 0.7935 re_mapping 0.0221 re_causal 0.0547 /// teacc 98.52 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0179, -0.0424, -0.0384, ..., -0.0151, 0.0766, 0.0625], + [-0.0633, -0.0725, -0.0823, ..., -0.0457, -0.0602, -0.0426], + [-0.0365, -0.0283, 0.0560, ..., -0.0381, 0.0688, -0.0093], + ..., + [-0.0396, 0.0282, 0.0337, ..., 0.0527, -0.0363, -0.0470], + [-0.0865, 0.0354, -0.0806, ..., 0.0166, -0.0335, -0.0314], + [ 0.0300, -0.0068, -0.0430, ..., 0.0074, -0.0268, 0.0030]], + device='cuda:0'), grad: tensor([[ 3.0965e-05, 4.7445e-05, 2.0251e-05, ..., 3.3945e-05, + 1.8954e-05, 2.4170e-05], + [ 1.9565e-05, 2.8181e-04, 9.1076e-05, ..., 8.9765e-05, + 6.0856e-05, 8.1122e-05], + [ 1.9580e-05, 7.5459e-05, -2.4986e-04, ..., -1.3125e-04, + -2.0468e-04, -7.9751e-05], + ..., + [ 3.2902e-05, 8.1444e-04, 1.1039e-04, ..., -3.4070e-04, + 8.8155e-05, 6.7532e-05], + [ 1.0926e-04, 1.6165e-04, 2.0647e-04, ..., 4.6670e-05, + 3.5644e-05, -1.2457e-04], + [ 1.6773e-04, -3.7479e-04, -1.7059e-04, ..., 9.2208e-05, + 5.0329e-06, 1.5087e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0143, -0.0204, -0.0157, -0.0286, -0.0306, -0.0012, 0.0244, -0.0106, + 0.0296, 0.0016], device='cuda:0'), grad: tensor([ 0.0002, 0.0005, -0.0002, -0.0004, -0.0004, -0.0008, 0.0006, 0.0014, + -0.0017, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 265.23, cls_loss 0.0299 cls_loss_mapping 0.0459 cls_loss_causal 0.7943 re_mapping 0.0219 re_causal 0.0568 /// teacc 98.50 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0188, -0.0429, -0.0388, ..., -0.0157, 0.0772, 0.0631], + [-0.0643, -0.0739, -0.0834, ..., -0.0461, -0.0612, -0.0427], + [-0.0372, -0.0291, 0.0563, ..., -0.0391, 0.0690, -0.0095], + ..., + [-0.0402, 0.0282, 0.0340, ..., 0.0531, -0.0360, -0.0473], + [-0.0872, 0.0368, -0.0817, ..., 0.0173, -0.0336, -0.0315], + [ 0.0305, -0.0065, -0.0428, ..., 0.0071, -0.0271, 0.0026]], + device='cuda:0'), grad: tensor([[ 1.3962e-05, 2.5004e-05, 5.5507e-06, ..., 9.5293e-06, + -8.6844e-05, -7.7724e-05], + [ 7.3239e-06, 6.9380e-05, -4.4417e-04, ..., 4.1813e-05, + 4.5858e-06, 2.7642e-06], + [ 3.1680e-05, 3.6788e-04, 7.2241e-05, ..., 1.2141e-04, + 2.5228e-05, 4.1515e-05], + ..., + [-4.6045e-06, -4.7994e-04, -1.1250e-05, ..., -3.9077e-04, + 1.6153e-05, 2.9150e-06], + [ 3.7432e-05, 3.5167e-05, 9.6977e-05, ..., 9.1717e-06, + 1.6466e-05, 2.0489e-08], + [-9.0718e-05, -2.3592e-04, -1.2136e-04, ..., -1.6248e-04, + 4.2558e-05, 2.9862e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0142, -0.0206, -0.0159, -0.0286, -0.0308, -0.0015, 0.0245, -0.0107, + 0.0301, 0.0017], device='cuda:0'), grad: tensor([-2.1741e-05, -1.2646e-03, 6.4230e-04, 7.8201e-05, 1.2970e-03, + 3.6311e-04, 3.2711e-04, -3.7026e-04, 3.2926e-04, -1.3809e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 265.40, cls_loss 0.0240 cls_loss_mapping 0.0373 cls_loss_causal 0.7918 re_mapping 0.0217 re_causal 0.0574 /// teacc 98.42 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0193, -0.0436, -0.0391, ..., -0.0158, 0.0777, 0.0634], + [-0.0649, -0.0742, -0.0847, ..., -0.0467, -0.0625, -0.0430], + [-0.0379, -0.0294, 0.0572, ..., -0.0398, 0.0702, -0.0090], + ..., + [-0.0405, 0.0286, 0.0343, ..., 0.0538, -0.0367, -0.0477], + [-0.0882, 0.0375, -0.0823, ..., 0.0178, -0.0338, -0.0317], + [ 0.0307, -0.0072, -0.0431, ..., 0.0066, -0.0276, 0.0024]], + device='cuda:0'), grad: tensor([[ 1.0526e-04, 4.4733e-05, 6.3032e-06, ..., 8.2791e-05, + -2.2483e-04, -1.2565e-04], + [ 1.6347e-05, 6.2764e-05, 6.4075e-05, ..., 5.2601e-05, + 2.9027e-05, 6.0722e-06], + [ 1.2264e-05, -9.0778e-05, -3.4499e-04, ..., 5.6058e-05, + -1.7846e-04, 3.7458e-06], + ..., + [-1.6898e-05, -4.4298e-04, -4.7743e-05, ..., -5.7650e-04, + 1.8165e-05, 1.0081e-05], + [ 2.5320e-04, 1.3316e-04, 2.8443e-04, ..., 5.4389e-05, + 2.2817e-04, 1.3161e-04], + [ 4.9561e-05, 1.8060e-04, 1.5688e-04, ..., 2.6107e-04, + 3.5346e-05, 2.4989e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0143, -0.0207, -0.0155, -0.0283, -0.0309, -0.0022, 0.0249, -0.0106, + 0.0302, 0.0016], device='cuda:0'), grad: tensor([-2.0832e-05, 1.3697e-04, -9.7942e-04, 4.3631e-04, -2.7966e-04, + -4.7684e-04, -1.0949e-04, -4.3726e-04, 1.2121e-03, 5.2071e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 264.97, cls_loss 0.0260 cls_loss_mapping 0.0348 cls_loss_causal 0.7714 re_mapping 0.0205 re_causal 0.0515 /// teacc 98.53 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0198, -0.0440, -0.0394, ..., -0.0160, 0.0782, 0.0637], + [-0.0661, -0.0752, -0.0854, ..., -0.0474, -0.0628, -0.0436], + [-0.0387, -0.0300, 0.0578, ..., -0.0402, 0.0712, -0.0084], + ..., + [-0.0416, 0.0292, 0.0345, ..., 0.0544, -0.0377, -0.0484], + [-0.0895, 0.0375, -0.0831, ..., 0.0177, -0.0341, -0.0319], + [ 0.0316, -0.0075, -0.0436, ..., 0.0067, -0.0278, 0.0021]], + device='cuda:0'), grad: tensor([[ 9.4056e-05, 1.0654e-05, 2.7373e-05, ..., 4.7863e-05, + -3.2425e-05, -1.0514e-04], + [ 4.0740e-05, 9.2015e-06, 2.2855e-06, ..., 2.2352e-05, + 5.9679e-06, -1.2904e-05], + [ 3.1114e-05, 3.8445e-05, -2.5500e-06, ..., 2.0206e-05, + 4.5151e-05, 2.2292e-05], + ..., + [ 8.4098e-07, -5.1498e-05, -1.4165e-06, ..., -7.1049e-05, + 2.2501e-05, 1.6391e-05], + [ 3.7813e-04, -8.8140e-06, -3.6005e-06, ..., 4.9889e-05, + 1.4114e-04, 7.6354e-05], + [ 4.0919e-05, 4.7982e-05, 1.5825e-05, ..., 1.0145e-04, + 8.4162e-05, 4.8369e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0146, -0.0211, -0.0153, -0.0285, -0.0310, -0.0017, 0.0249, -0.0108, + 0.0298, 0.0024], device='cuda:0'), grad: tensor([ 1.8716e-04, -5.8317e-04, 3.1543e-04, 3.3200e-05, -9.3222e-04, + 7.7286e-03, -7.8735e-03, 2.0301e-04, 6.2799e-04, 2.9135e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 265.30, cls_loss 0.0236 cls_loss_mapping 0.0360 cls_loss_causal 0.7585 re_mapping 0.0217 re_causal 0.0573 /// teacc 98.45 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0200, -0.0444, -0.0395, ..., -0.0161, 0.0792, 0.0645], + [-0.0669, -0.0761, -0.0862, ..., -0.0478, -0.0634, -0.0436], + [-0.0394, -0.0306, 0.0580, ..., -0.0406, 0.0714, -0.0081], + ..., + [-0.0420, 0.0295, 0.0349, ..., 0.0546, -0.0374, -0.0487], + [-0.0903, 0.0380, -0.0832, ..., 0.0181, -0.0339, -0.0319], + [ 0.0315, -0.0081, -0.0434, ..., 0.0061, -0.0278, 0.0020]], + device='cuda:0'), grad: tensor([[ 3.0589e-04, 1.1230e-04, 5.1379e-05, ..., 2.3448e-04, + 1.0848e-04, 1.5402e-04], + [ 4.5031e-05, 1.1122e-04, 2.6608e-04, ..., 3.3051e-05, + 5.9700e-04, 3.6788e-04], + [ 1.9312e-05, -2.7895e-04, -6.9427e-04, ..., 6.3717e-05, + -1.9913e-03, -1.2560e-03], + ..., + [ 7.4863e-05, 8.3148e-05, 1.6761e-04, ..., 6.9030e-06, + 5.4073e-04, 3.6955e-04], + [ 2.8348e-04, -2.0194e-04, 1.0628e-04, ..., -9.2864e-05, + 2.4629e-04, -7.0594e-06], + [ 2.1386e-04, 1.6904e-04, -1.3128e-05, ..., 2.2113e-04, + 6.7890e-05, 1.6916e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0148, -0.0208, -0.0160, -0.0284, -0.0307, -0.0018, 0.0246, -0.0106, + 0.0300, 0.0022], device='cuda:0'), grad: tensor([ 7.8583e-04, 9.4080e-04, -2.8839e-03, 4.3201e-04, 1.3483e-04, + 9.2793e-04, -2.0351e-03, 1.0262e-03, 9.3222e-05, 5.7602e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 265.20, cls_loss 0.0250 cls_loss_mapping 0.0348 cls_loss_causal 0.7694 re_mapping 0.0214 re_causal 0.0553 /// teacc 98.33 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0204, -0.0447, -0.0399, ..., -0.0164, 0.0799, 0.0650], + [-0.0679, -0.0771, -0.0865, ..., -0.0486, -0.0652, -0.0443], + [-0.0399, -0.0310, 0.0591, ..., -0.0418, 0.0724, -0.0071], + ..., + [-0.0425, 0.0302, 0.0341, ..., 0.0553, -0.0378, -0.0496], + [-0.0914, 0.0373, -0.0850, ..., 0.0179, -0.0347, -0.0327], + [ 0.0324, -0.0088, -0.0441, ..., 0.0057, -0.0280, 0.0016]], + device='cuda:0'), grad: tensor([[ 7.4729e-06, 8.9347e-05, 1.1712e-04, ..., 5.2080e-06, + 1.6317e-05, -3.2216e-05], + [ 7.8157e-06, 1.7965e-04, 4.3333e-05, ..., 3.9011e-05, + 3.1739e-05, 7.5251e-06], + [ 9.3430e-06, -1.6665e-04, -1.6174e-03, ..., 2.9564e-05, + -1.0071e-03, -8.2612e-05], + ..., + [ 1.7747e-05, 1.0118e-03, 4.7994e-04, ..., 8.2910e-05, + 4.2415e-04, 5.7697e-05], + [ 3.3230e-05, 3.7575e-04, 9.0361e-05, ..., 3.4899e-05, + 8.1837e-05, 2.0474e-05], + [-5.6833e-05, 1.2951e-03, 3.6812e-04, ..., 2.0289e-04, + 1.4544e-04, 6.5938e-06]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0149, -0.0206, -0.0161, -0.0281, -0.0302, -0.0018, 0.0242, -0.0105, + 0.0297, 0.0020], device='cuda:0'), grad: tensor([ 0.0004, -0.0017, -0.0026, -0.0050, 0.0005, 0.0007, 0.0004, 0.0033, + 0.0009, 0.0031], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 265.27, cls_loss 0.0239 cls_loss_mapping 0.0318 cls_loss_causal 0.7645 re_mapping 0.0195 re_causal 0.0513 /// teacc 98.30 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0205, -0.0450, -0.0402, ..., -0.0165, 0.0807, 0.0656], + [-0.0701, -0.0777, -0.0873, ..., -0.0493, -0.0663, -0.0446], + [-0.0404, -0.0313, 0.0604, ..., -0.0423, 0.0733, -0.0066], + ..., + [-0.0432, 0.0301, 0.0342, ..., 0.0558, -0.0390, -0.0507], + [-0.0922, 0.0378, -0.0862, ..., 0.0183, -0.0353, -0.0331], + [ 0.0329, -0.0101, -0.0443, ..., 0.0053, -0.0285, 0.0012]], + device='cuda:0'), grad: tensor([[ 1.0449e-04, 1.4961e-04, 1.1927e-04, ..., 2.5201e-04, + -6.5386e-05, 4.2528e-05], + [ 2.1207e-04, 5.5790e-05, 8.4102e-05, ..., 2.5320e-04, + 8.8096e-05, 1.4111e-05], + [ 1.9833e-05, -5.8413e-04, -4.5896e-04, ..., -1.9002e-04, + -9.0218e-04, -2.9683e-04], + ..., + [-4.2534e-03, -7.0286e-04, -1.1044e-03, ..., -5.1727e-03, + 3.3283e-04, 1.5533e-04], + [ 7.5102e-05, 2.1636e-04, 4.5151e-05, ..., -1.5819e-04, + -3.7718e-04, -4.7326e-04], + [ 3.7079e-03, 7.0572e-04, 1.1330e-03, ..., 4.5128e-03, + 1.4257e-04, 1.5175e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0148, -0.0209, -0.0158, -0.0279, -0.0303, -0.0019, 0.0244, -0.0110, + 0.0303, 0.0017], device='cuda:0'), grad: tensor([ 0.0005, 0.0006, -0.0013, 0.0002, 0.0003, 0.0007, 0.0012, -0.0108, + -0.0015, 0.0101], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 283.68, cls_loss 0.0256 cls_loss_mapping 0.0387 cls_loss_causal 0.7818 re_mapping 0.0197 re_causal 0.0518 /// teacc 98.66 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0206, -0.0453, -0.0405, ..., -0.0168, 0.0813, 0.0661], + [-0.0708, -0.0778, -0.0882, ..., -0.0502, -0.0672, -0.0441], + [-0.0412, -0.0319, 0.0616, ..., -0.0430, 0.0742, -0.0064], + ..., + [-0.0426, 0.0309, 0.0349, ..., 0.0574, -0.0405, -0.0516], + [-0.0930, 0.0378, -0.0872, ..., 0.0184, -0.0358, -0.0337], + [ 0.0332, -0.0103, -0.0448, ..., 0.0045, -0.0288, 0.0009]], + device='cuda:0'), grad: tensor([[ 4.9174e-05, 4.4018e-05, 5.6922e-06, ..., 2.8297e-05, + -5.4628e-05, -2.3454e-05], + [ 1.3009e-05, 1.0109e-03, 7.2300e-05, ..., 3.6502e-04, + 4.2245e-06, 3.2689e-06], + [ 1.0364e-05, 1.1330e-03, 2.3499e-05, ..., 2.9898e-04, + -1.2733e-05, 8.3074e-06], + ..., + [ 9.1791e-05, 1.3695e-02, 1.5364e-03, ..., 4.7417e-03, + 4.0457e-06, 2.0098e-06], + [ 9.3162e-05, 3.6407e-04, 6.9082e-05, ..., 1.1909e-04, + 4.1544e-05, 3.7968e-05], + [-1.0008e-04, -1.5778e-02, -1.7748e-03, ..., -5.7297e-03, + 2.4721e-05, 1.4685e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0148, -0.0206, -0.0161, -0.0282, -0.0303, -0.0018, 0.0240, -0.0101, + 0.0301, 0.0017], device='cuda:0'), grad: tensor([ 8.9705e-05, 1.6785e-03, 1.7271e-03, 1.3173e-04, 5.9223e-04, + -2.3925e-04, -2.2364e-04, 2.5299e-02, 8.4734e-04, -2.9907e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 265.21, cls_loss 0.0182 cls_loss_mapping 0.0292 cls_loss_causal 0.7523 re_mapping 0.0198 re_causal 0.0513 /// teacc 98.63 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0210, -0.0459, -0.0410, ..., -0.0169, 0.0819, 0.0665], + [-0.0714, -0.0790, -0.0891, ..., -0.0509, -0.0680, -0.0440], + [-0.0418, -0.0324, 0.0628, ..., -0.0438, 0.0754, -0.0058], + ..., + [-0.0427, 0.0316, 0.0349, ..., 0.0584, -0.0418, -0.0524], + [-0.0939, 0.0375, -0.0882, ..., 0.0185, -0.0363, -0.0340], + [ 0.0337, -0.0103, -0.0453, ..., 0.0041, -0.0292, 0.0005]], + device='cuda:0'), grad: tensor([[-2.2388e-04, 6.4559e-06, 1.2144e-05, ..., -4.7803e-05, + -8.6021e-04, -6.6710e-04], + [ 5.0999e-06, 9.2506e-05, 3.3593e-04, ..., 6.1095e-05, + 2.2864e-04, 1.1742e-04], + [ 4.0308e-06, 2.7585e-04, -3.9840e-04, ..., 2.1303e-04, + -4.9496e-04, -2.6894e-04], + ..., + [ 8.4043e-06, -5.6171e-04, -4.3660e-05, ..., -3.5977e-04, + 1.5378e-04, 8.6665e-05], + [ 8.6844e-05, 9.5248e-05, 6.4552e-05, ..., 1.1641e-04, + 5.5313e-05, 3.8266e-05], + [ 7.0632e-05, 3.4070e-04, 6.4898e-04, ..., 3.4833e-04, + 2.1505e-04, 1.6367e-04]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0149, -0.0209, -0.0162, -0.0278, -0.0304, -0.0022, 0.0241, -0.0099, + 0.0300, 0.0018], device='cuda:0'), grad: tensor([-1.0538e-03, 4.2272e-04, -3.1710e-04, 1.6797e-04, -1.1644e-03, + 6.1631e-05, 6.5994e-04, -5.7316e-04, 2.8014e-04, 1.5144e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 265.16, cls_loss 0.0209 cls_loss_mapping 0.0304 cls_loss_causal 0.7307 re_mapping 0.0194 re_causal 0.0509 /// teacc 98.60 lr 0.00010000 +Epoch 42, weight, value: tensor([[-2.1314e-02, -4.6477e-02, -4.1340e-02, ..., -1.6991e-02, + 8.3152e-02, 6.7446e-02], + [-7.1942e-02, -8.0279e-02, -8.8827e-02, ..., -5.1931e-02, + -6.8565e-02, -4.3897e-02], + [-4.2324e-02, -3.2757e-02, 6.3314e-02, ..., -4.4658e-02, + 7.5998e-02, -5.7002e-03], + ..., + [-4.2829e-02, 3.1579e-02, 3.4212e-02, ..., 5.8817e-02, + -4.2757e-02, -5.3114e-02], + [-9.4988e-02, 3.7984e-02, -8.8170e-02, ..., 1.8525e-02, + -3.6656e-02, -3.4171e-02], + [ 3.3962e-02, -1.0141e-02, -4.5372e-02, ..., 3.8681e-03, + -2.9619e-02, 7.8126e-05]], device='cuda:0'), grad: tensor([[ 4.8548e-05, 1.9535e-05, 1.4283e-05, ..., 6.9067e-06, + 1.0145e-04, 5.6416e-05], + [ 3.2812e-05, 1.0103e-04, 3.7265e-04, ..., 3.4600e-05, + 7.2098e-04, 4.8065e-04], + [ 4.0323e-05, -5.0926e-04, -1.2369e-03, ..., -1.1313e-04, + -2.8515e-03, -1.7376e-03], + ..., + [ 1.0617e-05, 3.5429e-04, 7.0143e-04, ..., 5.8532e-05, + 1.4172e-03, 8.3542e-04], + [ 6.5231e-04, 6.0320e-05, 3.3796e-05, ..., 1.8448e-05, + 1.4079e-04, 9.6083e-05], + [ 2.2173e-05, 2.0623e-05, 5.2862e-06, ..., 1.3977e-05, + 9.5844e-05, 5.5552e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0144, -0.0207, -0.0162, -0.0285, -0.0307, -0.0012, 0.0239, -0.0103, + 0.0300, 0.0019], device='cuda:0'), grad: tensor([ 0.0003, 0.0015, -0.0056, 0.0004, 0.0005, 0.0026, -0.0036, 0.0028, + 0.0009, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 265.10, cls_loss 0.0260 cls_loss_mapping 0.0317 cls_loss_causal 0.7500 re_mapping 0.0191 re_causal 0.0476 /// teacc 98.43 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0223, -0.0473, -0.0405, ..., -0.0173, 0.0839, 0.0677], + [-0.0729, -0.0813, -0.0894, ..., -0.0523, -0.0698, -0.0448], + [-0.0434, -0.0326, 0.0638, ..., -0.0453, 0.0770, -0.0049], + ..., + [-0.0432, 0.0319, 0.0361, ..., 0.0598, -0.0433, -0.0541], + [-0.0957, 0.0382, -0.0890, ..., 0.0191, -0.0376, -0.0342], + [ 0.0343, -0.0108, -0.0454, ..., 0.0032, -0.0290, -0.0001]], + device='cuda:0'), grad: tensor([[-9.3555e-04, 2.1253e-06, 1.1034e-05, ..., -1.3285e-03, + -6.0415e-04, -1.6546e-03], + [ 1.0297e-05, 8.0109e-05, 2.3752e-05, ..., 1.0979e-04, + 1.0744e-05, 3.1680e-05], + [ 1.9431e-05, 1.2589e-04, 2.1827e-04, ..., 5.1641e-04, + 6.5744e-05, 3.0828e-04], + ..., + [ 6.7949e-05, -1.1320e-03, -2.8934e-03, ..., -2.5215e-03, + 4.9591e-05, 1.5581e-04], + [ 2.5630e-04, 1.1444e-04, 1.6525e-05, ..., 5.5075e-04, + 1.6630e-04, 5.5075e-04], + [ 2.4945e-05, 5.7191e-05, 1.1712e-05, ..., 1.3268e-04, + 2.6748e-05, 7.4863e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0145, -0.0211, -0.0160, -0.0287, -0.0313, -0.0017, 0.0244, -0.0100, + 0.0308, 0.0018], device='cuda:0'), grad: tensor([-0.0018, 0.0002, 0.0020, 0.0027, 0.0036, -0.0052, -0.0001, -0.0045, + 0.0027, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 282.04, cls_loss 0.0204 cls_loss_mapping 0.0293 cls_loss_causal 0.7493 re_mapping 0.0192 re_causal 0.0503 /// teacc 98.77 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0227, -0.0478, -0.0406, ..., -0.0176, 0.0846, 0.0683], + [-0.0742, -0.0830, -0.0902, ..., -0.0529, -0.0715, -0.0456], + [-0.0441, -0.0330, 0.0641, ..., -0.0460, 0.0781, -0.0046], + ..., + [-0.0431, 0.0329, 0.0370, ..., 0.0609, -0.0439, -0.0545], + [-0.0970, 0.0384, -0.0894, ..., 0.0194, -0.0376, -0.0340], + [ 0.0343, -0.0113, -0.0458, ..., 0.0026, -0.0292, -0.0006]], + device='cuda:0'), grad: tensor([[ 9.7826e-06, 2.5600e-05, 3.4627e-06, ..., -3.4403e-06, + -5.8532e-05, -1.1772e-04], + [ 2.1532e-06, 6.6280e-05, 7.3649e-06, ..., 4.1008e-05, + 9.2268e-05, 1.7032e-05], + [ 1.2696e-05, -1.3018e-03, -8.5473e-05, ..., 1.6630e-05, + -4.7417e-03, -1.6165e-04], + ..., + [ 5.7966e-06, -1.1736e-04, 1.0890e-04, ..., -3.1757e-04, + 7.1859e-04, 1.6284e-04], + [ 1.2554e-05, -2.6748e-06, 3.3733e-06, ..., -4.9949e-05, + 6.3419e-05, 5.4270e-05], + [-3.4243e-05, 1.6415e-04, 2.3484e-05, ..., 1.8907e-04, + 6.5804e-05, 1.2033e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0145, -0.0220, -0.0158, -0.0291, -0.0309, -0.0012, 0.0241, -0.0095, + 0.0310, 0.0014], device='cuda:0'), grad: tensor([-2.7359e-05, 7.6711e-05, -7.3357e-03, 6.1378e-03, -1.4865e-04, + 1.5962e-04, 2.1890e-05, 7.9727e-04, 2.9519e-05, 2.8968e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 260.04, cls_loss 0.0178 cls_loss_mapping 0.0247 cls_loss_causal 0.7020 re_mapping 0.0194 re_causal 0.0473 /// teacc 98.70 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0231, -0.0481, -0.0408, ..., -0.0176, 0.0853, 0.0689], + [-0.0748, -0.0841, -0.0903, ..., -0.0538, -0.0720, -0.0459], + [-0.0449, -0.0333, 0.0645, ..., -0.0467, 0.0789, -0.0042], + ..., + [-0.0437, 0.0327, 0.0373, ..., 0.0613, -0.0449, -0.0552], + [-0.0971, 0.0385, -0.0897, ..., 0.0202, -0.0376, -0.0337], + [ 0.0351, -0.0115, -0.0460, ..., 0.0023, -0.0293, -0.0010]], + device='cuda:0'), grad: tensor([[ 1.7002e-05, 1.1194e-04, 5.8860e-07, ..., 3.2634e-05, + 1.2517e-04, 2.3410e-05], + [ 9.2387e-06, 1.4342e-05, 5.4576e-07, ..., 4.7311e-06, + 7.6592e-06, 2.3581e-06], + [ 1.1981e-05, 1.0443e-04, -7.6592e-06, ..., 3.0667e-05, + 1.3268e-04, 5.1141e-05], + ..., + [ 1.0729e-05, -4.5776e-05, -2.7753e-07, ..., -4.5806e-05, + 1.2167e-05, 4.0792e-06], + [ 6.2764e-05, -2.6011e-04, 1.9725e-06, ..., -6.2168e-05, + -4.8637e-04, -1.4532e-04], + [ 2.2650e-05, 4.4405e-05, 1.5013e-06, ..., 1.3880e-05, + 2.3797e-05, 1.2212e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0145, -0.0218, -0.0156, -0.0289, -0.0311, -0.0012, 0.0236, -0.0098, + 0.0310, 0.0017], device='cuda:0'), grad: tensor([ 2.9135e-04, -5.2738e-04, 4.2129e-04, 1.1241e-04, -5.7250e-05, + 1.0651e-04, 3.5793e-05, 1.1969e-04, -7.1955e-04, 2.1541e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 263.58, cls_loss 0.0172 cls_loss_mapping 0.0285 cls_loss_causal 0.7269 re_mapping 0.0181 re_causal 0.0481 /// teacc 98.47 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0232, -0.0487, -0.0409, ..., -0.0175, 0.0860, 0.0695], + [-0.0766, -0.0850, -0.0906, ..., -0.0537, -0.0726, -0.0462], + [-0.0456, -0.0338, 0.0650, ..., -0.0475, 0.0797, -0.0040], + ..., + [-0.0442, 0.0335, 0.0375, ..., 0.0621, -0.0456, -0.0554], + [-0.0976, 0.0385, -0.0901, ..., 0.0205, -0.0380, -0.0338], + [ 0.0355, -0.0123, -0.0461, ..., 0.0019, -0.0293, -0.0015]], + device='cuda:0'), grad: tensor([[ 6.2585e-06, 2.1845e-05, 1.0453e-05, ..., 1.7136e-05, + 3.6389e-05, -8.2403e-06], + [ 1.4246e-05, 1.7047e-04, 2.4326e-06, ..., 2.2125e-04, + -1.4424e-05, -2.2948e-05], + [ 2.5690e-05, 2.8205e-04, -6.1631e-05, ..., 1.8418e-04, + -2.4271e-04, 1.0043e-05], + ..., + [ 8.5533e-06, -9.5665e-05, -2.1905e-06, ..., -1.9193e-04, + 2.1428e-05, 3.9265e-06], + [ 6.2168e-05, -6.0368e-04, 1.9055e-06, ..., -5.4169e-04, + -6.6042e-05, 7.0445e-06], + [ 1.3739e-05, 1.0091e-04, 3.1106e-07, ..., 8.4877e-05, + 9.7603e-06, 3.4422e-06]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0143, -0.0216, -0.0159, -0.0285, -0.0311, -0.0015, 0.0233, -0.0097, + 0.0311, 0.0016], device='cuda:0'), grad: tensor([ 1.2338e-04, -1.1482e-03, 1.2074e-03, 4.2486e-04, 1.4699e-04, + 3.2115e-04, -1.7679e-04, 6.6496e-06, -1.1892e-03, 2.8348e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 265.11, cls_loss 0.0174 cls_loss_mapping 0.0280 cls_loss_causal 0.7351 re_mapping 0.0175 re_causal 0.0477 /// teacc 98.69 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0239, -0.0492, -0.0408, ..., -0.0176, 0.0870, 0.0701], + [-0.0771, -0.0852, -0.0910, ..., -0.0542, -0.0739, -0.0465], + [-0.0461, -0.0342, 0.0661, ..., -0.0481, 0.0807, -0.0038], + ..., + [-0.0447, 0.0337, 0.0371, ..., 0.0626, -0.0465, -0.0558], + [-0.0989, 0.0383, -0.0906, ..., 0.0206, -0.0389, -0.0342], + [ 0.0361, -0.0124, -0.0460, ..., 0.0016, -0.0295, -0.0019]], + device='cuda:0'), grad: tensor([[ 1.3523e-05, 4.0221e-04, 2.0042e-06, ..., 6.1607e-04, + 8.2922e-04, 1.2177e-04], + [ 1.5222e-05, 6.3360e-05, 1.8459e-06, ..., 3.8952e-05, + 2.9415e-05, 8.6799e-06], + [ 8.6904e-05, 1.8096e-04, -2.9057e-05, ..., -1.0766e-05, + -7.3624e-04, -2.3091e-04], + ..., + [-2.2918e-05, -2.7800e-04, -1.3880e-05, ..., -9.6798e-04, + -4.1318e-04, 1.0478e-04], + [ 4.8995e-05, 1.4436e-04, 8.1286e-06, ..., -7.1786e-06, + 4.8190e-05, 2.7806e-05], + [ 1.1623e-04, 2.2912e-04, 1.8910e-05, ..., 5.2299e-03, + 1.3757e-04, 2.2545e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0142, -0.0211, -0.0155, -0.0284, -0.0316, -0.0016, 0.0239, -0.0104, + 0.0308, 0.0018], device='cuda:0'), grad: tensor([ 1.6327e-03, 1.2457e-04, -4.6039e-04, -7.7820e-04, -2.8305e-02, + 2.3651e-04, 4.7147e-05, -1.2856e-03, 7.5623e-06, 2.8763e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 265.52, cls_loss 0.0210 cls_loss_mapping 0.0260 cls_loss_causal 0.7203 re_mapping 0.0169 re_causal 0.0437 /// teacc 98.59 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0249, -0.0497, -0.0410, ..., -0.0179, 0.0879, 0.0707], + [-0.0775, -0.0862, -0.0917, ..., -0.0549, -0.0745, -0.0464], + [-0.0469, -0.0344, 0.0668, ..., -0.0484, 0.0817, -0.0033], + ..., + [-0.0452, 0.0330, 0.0372, ..., 0.0626, -0.0480, -0.0567], + [-0.1006, 0.0383, -0.0910, ..., 0.0204, -0.0399, -0.0351], + [ 0.0363, -0.0125, -0.0460, ..., 0.0014, -0.0294, -0.0025]], + device='cuda:0'), grad: tensor([[ 2.1112e-04, 4.4346e-05, 3.1352e-05, ..., 2.1443e-05, + 2.6464e-04, 2.6703e-04], + [ 1.3337e-05, 8.6963e-05, 3.2216e-05, ..., 5.9426e-05, + 8.0228e-05, 5.9396e-05], + [ 2.0057e-05, -5.0402e-04, -4.3321e-04, ..., -9.5814e-06, + -1.1053e-03, -7.6151e-04], + ..., + [ 6.4325e-04, 2.6588e-03, 5.3972e-05, ..., 4.9782e-04, + 3.2234e-04, 2.6441e-04], + [-5.0688e-04, -8.4352e-04, -1.3041e-04, ..., -2.7885e-03, + 2.1183e-04, 1.7905e-04], + [ 1.4865e-04, 7.1383e-04, 2.8566e-05, ..., 2.2089e-04, + 6.8009e-05, 6.2585e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0143, -0.0209, -0.0156, -0.0278, -0.0320, -0.0017, 0.0241, -0.0112, + 0.0306, 0.0025], device='cuda:0'), grad: tensor([ 4.2415e-04, 2.2626e-04, -2.2335e-03, -1.5144e-03, 3.1614e-04, + 3.2558e-03, -7.7128e-05, 2.1763e-03, -3.2520e-03, 6.8283e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 265.55, cls_loss 0.0164 cls_loss_mapping 0.0261 cls_loss_causal 0.7087 re_mapping 0.0170 re_causal 0.0462 /// teacc 98.74 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0251, -0.0502, -0.0412, ..., -0.0181, 0.0889, 0.0715], + [-0.0783, -0.0868, -0.0919, ..., -0.0558, -0.0750, -0.0466], + [-0.0480, -0.0346, 0.0668, ..., -0.0492, 0.0821, -0.0030], + ..., + [-0.0458, 0.0332, 0.0375, ..., 0.0640, -0.0483, -0.0572], + [-0.1015, 0.0383, -0.0912, ..., 0.0206, -0.0398, -0.0353], + [ 0.0365, -0.0129, -0.0462, ..., 0.0008, -0.0301, -0.0032]], + device='cuda:0'), grad: tensor([[ 7.5221e-05, 6.3181e-05, 9.3803e-06, ..., 1.1384e-04, + -6.1333e-05, -4.0770e-05], + [ 2.0489e-05, 1.2189e-04, 8.2254e-05, ..., 1.4174e-04, + 3.2783e-05, 5.6662e-06], + [ 1.9327e-05, 3.6049e-04, 1.8764e-04, ..., 2.8849e-04, + -1.2577e-05, 3.4440e-06], + ..., + [ 8.0490e-04, -2.3460e-04, -4.3678e-04, ..., 4.8923e-04, + -1.2469e-04, -1.8284e-05], + [ 2.1946e-04, 1.0526e-04, 2.0504e-05, ..., 2.6751e-04, + 4.1515e-05, 2.4922e-06], + [ 8.5771e-05, -8.1897e-05, 2.4602e-05, ..., 4.2295e-04, + 1.0997e-05, 4.0531e-06]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0140, -0.0206, -0.0159, -0.0276, -0.0318, -0.0019, 0.0232, -0.0109, + 0.0310, 0.0020], device='cuda:0'), grad: tensor([ 8.0347e-05, 7.2718e-04, 5.2881e-04, 4.1890e-04, -7.9803e-03, + -8.6403e-04, 5.7891e-06, 7.5722e-04, 2.4724e-04, 6.0806e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 265.56, cls_loss 0.0161 cls_loss_mapping 0.0223 cls_loss_causal 0.7041 re_mapping 0.0165 re_causal 0.0440 /// teacc 98.65 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0254, -0.0507, -0.0413, ..., -0.0182, 0.0894, 0.0719], + [-0.0801, -0.0874, -0.0925, ..., -0.0566, -0.0765, -0.0473], + [-0.0475, -0.0350, 0.0676, ..., -0.0501, 0.0831, -0.0024], + ..., + [-0.0464, 0.0335, 0.0376, ..., 0.0647, -0.0490, -0.0577], + [-0.1024, 0.0382, -0.0917, ..., 0.0212, -0.0405, -0.0356], + [ 0.0372, -0.0129, -0.0460, ..., 0.0004, -0.0308, -0.0036]], + device='cuda:0'), grad: tensor([[-8.2612e-05, -8.3506e-05, 5.6401e-06, ..., -3.4451e-05, + -9.3699e-04, -7.7105e-04], + [ 6.0126e-06, 1.9324e-04, 7.0967e-06, ..., 1.1557e-04, + 4.2409e-05, 1.4096e-05], + [ 1.4558e-05, 6.8247e-05, -2.1732e-04, ..., 6.9559e-05, + -3.7909e-04, 1.3940e-05], + ..., + [ 3.4012e-06, -1.0958e-03, 3.8929e-06, ..., -8.5163e-04, + 1.2994e-04, 5.9485e-05], + [ 4.1515e-05, 2.5600e-05, 4.2617e-06, ..., 3.4660e-05, + 5.3883e-05, 4.7237e-05], + [-2.5809e-05, 5.0879e-04, 9.7752e-05, ..., 4.6492e-04, + 1.8334e-04, 3.1412e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0139, -0.0211, -0.0151, -0.0274, -0.0314, -0.0024, 0.0236, -0.0114, + 0.0311, 0.0019], device='cuda:0'), grad: tensor([-1.2026e-03, 2.6727e-04, -7.3528e-04, 7.1096e-04, 2.9635e-04, + -1.5393e-05, 9.8419e-04, -1.4954e-03, 1.2910e-04, 1.0624e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 265.88, cls_loss 0.0199 cls_loss_mapping 0.0287 cls_loss_causal 0.6855 re_mapping 0.0175 re_causal 0.0426 /// teacc 98.66 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0259, -0.0510, -0.0415, ..., -0.0184, 0.0900, 0.0723], + [-0.0808, -0.0883, -0.0934, ..., -0.0572, -0.0772, -0.0473], + [-0.0482, -0.0355, 0.0680, ..., -0.0515, 0.0837, -0.0026], + ..., + [-0.0474, 0.0336, 0.0377, ..., 0.0655, -0.0494, -0.0580], + [-0.1034, 0.0385, -0.0913, ..., 0.0216, -0.0400, -0.0353], + [ 0.0375, -0.0136, -0.0458, ..., -0.0002, -0.0313, -0.0040]], + device='cuda:0'), grad: tensor([[ 2.4453e-05, 1.2286e-05, 5.9977e-06, ..., 1.1809e-05, + 1.9670e-05, 1.4350e-05], + [ 9.4697e-06, 8.0943e-05, 3.3259e-05, ..., 6.8843e-05, + 4.0948e-05, 2.0593e-05], + [ 1.5402e-04, 1.6892e-04, -4.1723e-05, ..., 4.5061e-05, + -6.3598e-05, -8.1122e-05], + ..., + [-1.8895e-04, -1.3275e-03, -4.7493e-04, ..., -1.1206e-03, + 3.7491e-05, 1.5222e-05], + [ 8.3685e-05, 1.4327e-05, 2.9743e-05, ..., 6.2466e-05, + 3.6925e-05, 1.4700e-05], + [ 1.3518e-04, 2.8253e-04, 3.6627e-05, ..., 3.2663e-04, + 1.1303e-05, 5.0813e-06]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0141, -0.0218, -0.0151, -0.0272, -0.0310, -0.0021, 0.0231, -0.0113, + 0.0320, 0.0013], device='cuda:0'), grad: tensor([ 5.9098e-05, 1.4687e-04, 9.7632e-05, -2.2495e-04, 1.2102e-03, + -2.7031e-05, 4.1991e-05, -2.0885e-03, 1.6928e-04, 6.1560e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 265.37, cls_loss 0.0193 cls_loss_mapping 0.0255 cls_loss_causal 0.6766 re_mapping 0.0170 re_causal 0.0436 /// teacc 98.60 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0264, -0.0515, -0.0417, ..., -0.0186, 0.0900, 0.0723], + [-0.0814, -0.0901, -0.0941, ..., -0.0576, -0.0780, -0.0476], + [-0.0487, -0.0361, 0.0685, ..., -0.0521, 0.0854, -0.0017], + ..., + [-0.0476, 0.0345, 0.0378, ..., 0.0661, -0.0507, -0.0591], + [-0.1049, 0.0388, -0.0914, ..., 0.0215, -0.0401, -0.0357], + [ 0.0375, -0.0144, -0.0459, ..., -0.0006, -0.0316, -0.0042]], + device='cuda:0'), grad: tensor([[ 8.5711e-05, 2.8342e-05, 2.2314e-06, ..., 2.0340e-05, + -4.4368e-06, 7.0594e-07], + [ 1.4193e-05, 2.1648e-04, 3.8557e-06, ..., 1.4615e-04, + 1.4659e-06, 1.6298e-06], + [ 2.4393e-05, 1.3566e-04, -3.2075e-06, ..., 8.6129e-05, + -7.8678e-06, -4.0419e-07], + ..., + [ 2.0355e-05, 2.2797e-02, -3.0324e-05, ..., 1.4809e-02, + 7.1339e-06, 3.7383e-06], + [ 1.1146e-04, -2.5055e-02, 1.1779e-05, ..., -1.6312e-02, + 8.2180e-06, 1.0990e-05], + [ 9.1910e-05, 3.1638e-04, 2.4680e-06, ..., 2.3460e-04, + 2.2519e-06, 5.3383e-06]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0146, -0.0223, -0.0153, -0.0272, -0.0312, -0.0024, 0.0238, -0.0106, + 0.0320, 0.0013], device='cuda:0'), grad: tensor([ 1.5092e-04, 1.6272e-04, 1.9479e-04, 3.7079e-03, 3.9667e-05, + -2.4738e-03, 5.1349e-05, 2.4918e-02, -2.7191e-02, 4.5109e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 51---------------------------------------------------- +epoch 51, time 281.80, cls_loss 0.0164 cls_loss_mapping 0.0217 cls_loss_causal 0.6797 re_mapping 0.0158 re_causal 0.0399 /// teacc 98.79 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0266, -0.0522, -0.0419, ..., -0.0188, 0.0915, 0.0736], + [-0.0819, -0.0912, -0.0936, ..., -0.0584, -0.0795, -0.0484], + [-0.0495, -0.0373, 0.0690, ..., -0.0536, 0.0857, -0.0016], + ..., + [-0.0477, 0.0356, 0.0379, ..., 0.0672, -0.0511, -0.0593], + [-0.1055, 0.0395, -0.0919, ..., 0.0223, -0.0408, -0.0362], + [ 0.0377, -0.0154, -0.0465, ..., -0.0010, -0.0320, -0.0046]], + device='cuda:0'), grad: tensor([[-1.0562e-04, 2.3972e-06, 4.6566e-07, ..., 1.2890e-06, + -1.5426e-04, -1.3697e-04], + [ 1.0669e-05, 1.1452e-05, 1.3541e-06, ..., 3.4831e-06, + 2.3115e-06, 1.3057e-06], + [ 5.2229e-06, 1.9759e-05, -3.4086e-06, ..., 3.2578e-06, + -2.4125e-05, -6.8508e-06], + ..., + [ 1.6585e-05, -6.9797e-05, -1.4484e-05, ..., -5.0634e-05, + 1.3985e-05, 4.7348e-06], + [ 1.2159e-04, 2.0787e-05, 2.0713e-06, ..., 7.6219e-06, + 4.7296e-05, 5.5254e-05], + [-1.7726e-04, 1.0855e-05, 7.1377e-06, ..., -2.4423e-05, + 7.4863e-05, 4.8019e-06]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0137, -0.0217, -0.0156, -0.0277, -0.0313, -0.0026, 0.0233, -0.0101, + 0.0321, 0.0009], device='cuda:0'), grad: tensor([-1.3626e-04, -1.8477e-05, -4.1425e-06, 1.5271e-04, -1.0985e-04, + -4.9919e-05, 1.0081e-05, -3.4213e-05, 2.7704e-04, -8.6546e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 266.08, cls_loss 0.0163 cls_loss_mapping 0.0217 cls_loss_causal 0.6746 re_mapping 0.0155 re_causal 0.0400 /// teacc 98.72 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0274, -0.0527, -0.0421, ..., -0.0190, 0.0928, 0.0745], + [-0.0826, -0.0909, -0.0941, ..., -0.0591, -0.0804, -0.0490], + [-0.0506, -0.0375, 0.0694, ..., -0.0543, 0.0863, -0.0012], + ..., + [-0.0486, 0.0355, 0.0381, ..., 0.0679, -0.0520, -0.0603], + [-0.1061, 0.0396, -0.0924, ..., 0.0225, -0.0413, -0.0360], + [ 0.0387, -0.0155, -0.0469, ..., -0.0015, -0.0324, -0.0050]], + device='cuda:0'), grad: tensor([[ 1.3851e-05, 2.1681e-05, 3.7253e-05, ..., 1.3702e-05, + 7.0333e-05, 2.9579e-05], + [ 5.8636e-06, 6.2659e-06, 2.2831e-03, ..., 5.5842e-06, + 4.3526e-03, 2.8000e-03], + [ 4.4703e-06, -1.0006e-05, -2.6588e-03, ..., 7.7486e-06, + -5.1041e-03, -3.2463e-03], + ..., + [-1.1511e-05, -6.2466e-05, 1.0121e-04, ..., -8.0347e-05, + 1.8632e-04, 1.1730e-04], + [ 6.4313e-05, 2.0877e-05, 7.8440e-05, ..., 1.7166e-05, + 1.4794e-04, 8.6725e-05], + [-1.3745e-04, 2.4542e-05, -1.5032e-06, ..., 5.4330e-05, + 1.4007e-05, 9.3952e-06]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0131, -0.0212, -0.0155, -0.0278, -0.0316, -0.0029, 0.0233, -0.0108, + 0.0323, 0.0013], device='cuda:0'), grad: tensor([ 0.0002, 0.0106, -0.0122, 0.0011, 0.0004, -0.0008, 0.0003, 0.0004, + 0.0004, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 266.08, cls_loss 0.0172 cls_loss_mapping 0.0232 cls_loss_causal 0.7259 re_mapping 0.0152 re_causal 0.0411 /// teacc 98.77 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0282, -0.0534, -0.0425, ..., -0.0193, 0.0931, 0.0748], + [-0.0834, -0.0923, -0.0949, ..., -0.0600, -0.0816, -0.0493], + [-0.0513, -0.0374, 0.0702, ..., -0.0553, 0.0874, 0.0001], + ..., + [-0.0501, 0.0356, 0.0382, ..., 0.0686, -0.0516, -0.0617], + [-0.1072, 0.0401, -0.0930, ..., 0.0222, -0.0417, -0.0363], + [ 0.0392, -0.0165, -0.0469, ..., -0.0017, -0.0325, -0.0051]], + device='cuda:0'), grad: tensor([[ 3.2635e-03, 5.1744e-06, 4.1910e-07, ..., 9.8991e-04, + 1.9703e-03, 2.4052e-03], + [ 3.3025e-06, 3.0249e-05, -1.2722e-06, ..., 1.7917e-04, + 1.6410e-06, -8.6613e-07], + [ 9.6112e-06, 4.6194e-05, -2.2594e-06, ..., 6.1274e-05, + -1.6261e-06, 4.1537e-06], + ..., + [-3.2298e-06, -1.9121e-04, 2.5611e-06, ..., -1.2045e-03, + 4.5262e-06, 2.6114e-06], + [ 3.1888e-05, -6.4731e-05, 1.2517e-06, ..., 2.3794e-04, + 1.9014e-05, 2.0936e-05], + [ 5.5879e-08, 6.5386e-05, -9.9093e-07, ..., 4.2892e-04, + 3.9861e-06, 2.6878e-06]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0138, -0.0217, -0.0151, -0.0283, -0.0319, -0.0022, 0.0230, -0.0104, + 0.0324, 0.0014], device='cuda:0'), grad: tensor([ 3.5305e-03, -4.2462e-04, 2.3603e-04, 2.3639e-04, -6.1274e-05, + 4.6921e-04, -3.7003e-03, -1.4229e-03, 4.6706e-04, 6.7234e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 265.52, cls_loss 0.0172 cls_loss_mapping 0.0235 cls_loss_causal 0.7089 re_mapping 0.0149 re_causal 0.0394 /// teacc 98.70 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0288, -0.0540, -0.0426, ..., -0.0198, 0.0940, 0.0754], + [-0.0854, -0.0928, -0.0951, ..., -0.0607, -0.0822, -0.0488], + [-0.0520, -0.0378, 0.0709, ..., -0.0561, 0.0881, 0.0002], + ..., + [-0.0505, 0.0355, 0.0383, ..., 0.0692, -0.0526, -0.0619], + [-0.1084, 0.0413, -0.0934, ..., 0.0232, -0.0418, -0.0366], + [ 0.0399, -0.0173, -0.0471, ..., -0.0023, -0.0325, -0.0052]], + device='cuda:0'), grad: tensor([[ 1.1787e-05, 3.2723e-05, 2.5239e-06, ..., 3.0845e-05, + -2.1076e-04, -1.7679e-04], + [ 3.6173e-06, 7.6175e-05, 1.1642e-06, ..., 7.1406e-05, + 9.0301e-06, 4.6901e-06], + [ 8.5458e-06, 5.3883e-04, 1.0077e-06, ..., 5.0259e-04, + 4.9472e-05, 5.2065e-05], + ..., + [ 8.8960e-06, -1.4877e-03, 9.3281e-06, ..., -1.3924e-03, + 3.0443e-05, 1.3448e-05], + [ 2.7835e-05, 1.2040e-04, 1.1977e-06, ..., 6.6638e-05, + 4.8429e-05, 3.9011e-05], + [-5.9795e-04, 3.6329e-05, 4.9919e-05, ..., 5.7071e-05, + 4.2617e-05, 3.0518e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0144, -0.0216, -0.0151, -0.0285, -0.0317, -0.0017, 0.0230, -0.0108, + 0.0326, 0.0016], device='cuda:0'), grad: tensor([-1.8311e-04, 1.2398e-04, 1.2779e-03, 1.6975e-03, -1.5497e-04, + 6.3133e-04, 4.2677e-05, -3.1357e-03, 3.0041e-04, -6.0081e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 265.32, cls_loss 0.0178 cls_loss_mapping 0.0219 cls_loss_causal 0.6829 re_mapping 0.0157 re_causal 0.0380 /// teacc 98.72 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0296, -0.0545, -0.0429, ..., -0.0199, 0.0946, 0.0759], + [-0.0863, -0.0937, -0.0953, ..., -0.0614, -0.0839, -0.0493], + [-0.0531, -0.0384, 0.0713, ..., -0.0569, 0.0888, 0.0002], + ..., + [-0.0518, 0.0354, 0.0382, ..., 0.0698, -0.0530, -0.0622], + [-0.1097, 0.0415, -0.0939, ..., 0.0232, -0.0426, -0.0369], + [ 0.0405, -0.0182, -0.0474, ..., -0.0031, -0.0327, -0.0057]], + device='cuda:0'), grad: tensor([[ 3.1721e-06, 8.3223e-06, 1.8291e-06, ..., 4.4405e-06, + -9.2015e-06, -5.0850e-06], + [ 1.6522e-06, 1.0267e-05, 5.8115e-06, ..., 4.3958e-06, + 5.6289e-06, 1.1511e-06], + [ 2.8498e-06, -1.1563e-04, -1.3304e-04, ..., -4.7199e-06, + -2.2531e-04, -1.0926e-04], + ..., + [-7.5221e-05, -5.6684e-05, 1.4341e-04, ..., -1.2195e-04, + 2.1470e-04, 9.9719e-05], + [ 1.3411e-05, -2.1172e-04, 6.9179e-06, ..., -1.2541e-04, + 1.1221e-05, 6.1691e-06], + [ 5.3942e-05, 1.3876e-04, 1.2910e-04, ..., 8.1956e-05, + 2.8253e-05, 2.9430e-06]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0144, -0.0221, -0.0153, -0.0282, -0.0314, -0.0014, 0.0235, -0.0112, + 0.0326, 0.0015], device='cuda:0'), grad: tensor([ 1.1526e-05, -2.8219e-06, -3.6502e-04, 2.0409e-04, -2.3308e-03, + 5.5730e-05, -2.7567e-07, 3.6979e-04, -1.8990e-04, 2.2469e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 265.73, cls_loss 0.0169 cls_loss_mapping 0.0210 cls_loss_causal 0.6573 re_mapping 0.0156 re_causal 0.0372 /// teacc 98.78 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0305, -0.0552, -0.0436, ..., -0.0202, 0.0948, 0.0761], + [-0.0870, -0.0946, -0.0955, ..., -0.0621, -0.0853, -0.0494], + [-0.0538, -0.0386, 0.0721, ..., -0.0575, 0.0896, 0.0013], + ..., + [-0.0522, 0.0356, 0.0381, ..., 0.0706, -0.0532, -0.0636], + [-0.1115, 0.0418, -0.0945, ..., 0.0232, -0.0430, -0.0372], + [ 0.0406, -0.0197, -0.0475, ..., -0.0038, -0.0332, -0.0062]], + device='cuda:0'), grad: tensor([[ 2.5202e-06, 9.0301e-06, 1.9558e-07, ..., -4.9055e-05, + -3.1590e-04, -2.5058e-04], + [ 1.5885e-05, 5.2959e-05, 1.2852e-06, ..., 2.3201e-05, + 3.5077e-05, 2.2799e-05], + [ 4.0308e-06, 1.6224e-04, -8.0243e-06, ..., 2.9713e-05, + 5.0366e-05, 1.3739e-05], + ..., + [-5.2787e-06, -7.3195e-04, -4.3772e-07, ..., -4.0197e-04, + 4.2498e-05, 2.7090e-05], + [ 4.2051e-05, 6.9559e-05, 1.0356e-06, ..., 1.5005e-05, + 1.2493e-04, 7.8499e-05], + [-6.0797e-05, 1.7130e-04, 1.9800e-06, ..., 1.1557e-04, + 3.6597e-05, 2.4036e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0151, -0.0218, -0.0156, -0.0274, -0.0312, -0.0013, 0.0237, -0.0106, + 0.0323, 0.0006], device='cuda:0'), grad: tensor([-4.1175e-04, 1.1873e-04, 2.3568e-04, 4.8733e-04, 1.4372e-05, + 1.7571e-04, -1.3101e-04, -9.3079e-04, 3.1281e-04, 1.2946e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 265.39, cls_loss 0.0164 cls_loss_mapping 0.0201 cls_loss_causal 0.7051 re_mapping 0.0146 re_causal 0.0384 /// teacc 98.74 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0314, -0.0558, -0.0437, ..., -0.0199, 0.0957, 0.0769], + [-0.0878, -0.0955, -0.0959, ..., -0.0627, -0.0867, -0.0500], + [-0.0544, -0.0384, 0.0719, ..., -0.0575, 0.0905, 0.0020], + ..., + [-0.0530, 0.0361, 0.0390, ..., 0.0717, -0.0539, -0.0646], + [-0.1127, 0.0416, -0.0945, ..., 0.0229, -0.0436, -0.0377], + [ 0.0406, -0.0206, -0.0479, ..., -0.0043, -0.0337, -0.0066]], + device='cuda:0'), grad: tensor([[ 3.8408e-06, 7.3493e-05, 1.5393e-05, ..., 2.8200e-06, + 1.5116e-04, 4.0829e-05], + [-9.5889e-06, 6.6347e-06, 3.1758e-06, ..., 4.4480e-06, + 1.2651e-05, -2.4214e-07], + [ 1.0543e-06, -1.0514e-04, 1.5140e-04, ..., 1.4675e-04, + 1.4052e-05, -7.9036e-05], + ..., + [ 7.6666e-06, -2.8563e-04, 3.6322e-06, ..., -1.8013e-04, + -9.6738e-05, -4.1097e-05], + [ 4.3958e-06, 2.1315e-04, 9.4064e-07, ..., 5.8487e-07, + 3.6860e-04, 1.4055e-04], + [ 3.6489e-06, 9.7156e-06, 1.5795e-05, ..., 6.8136e-06, + 8.1658e-06, 1.7863e-06]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0144, -0.0218, -0.0150, -0.0273, -0.0319, -0.0010, 0.0239, -0.0103, + 0.0319, -0.0002], device='cuda:0'), grad: tensor([ 2.1160e-04, -2.2840e-04, 3.7265e-04, 1.8775e-04, -2.6932e-03, + 7.9691e-05, 1.6966e-03, -2.0850e-04, -7.9930e-05, 6.6137e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 265.55, cls_loss 0.0151 cls_loss_mapping 0.0201 cls_loss_causal 0.6425 re_mapping 0.0153 re_causal 0.0380 /// teacc 98.72 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0313, -0.0564, -0.0439, ..., -0.0198, 0.0963, 0.0775], + [-0.0904, -0.0960, -0.0974, ..., -0.0634, -0.0885, -0.0507], + [-0.0540, -0.0381, 0.0731, ..., -0.0577, 0.0915, 0.0030], + ..., + [-0.0533, 0.0364, 0.0389, ..., 0.0724, -0.0554, -0.0661], + [-0.1130, 0.0416, -0.0949, ..., 0.0231, -0.0432, -0.0376], + [ 0.0409, -0.0209, -0.0478, ..., -0.0050, -0.0340, -0.0069]], + device='cuda:0'), grad: tensor([[ 2.2259e-06, 1.5020e-05, 4.8168e-06, ..., 1.9521e-05, + -6.5446e-05, -3.9816e-05], + [ 1.8720e-06, 7.5400e-05, 1.9580e-05, ..., 1.0425e-04, + 1.8552e-05, 4.0121e-06], + [ 3.9749e-06, 1.6308e-04, 4.2208e-06, ..., 1.8561e-04, + 2.8804e-05, 1.2256e-05], + ..., + [ 2.9169e-06, -4.2826e-05, -1.4400e-04, ..., -1.6558e-04, + 8.1360e-06, 4.1053e-06], + [ 7.2159e-06, -9.0456e-04, 5.0291e-07, ..., -1.0576e-03, + -1.3554e-04, -4.0710e-05], + [-7.6815e-06, 4.7028e-05, 1.0383e-04, ..., 1.4424e-04, + 8.2031e-06, 4.6492e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0144, -0.0227, -0.0138, -0.0277, -0.0315, -0.0008, 0.0231, -0.0105, + 0.0325, -0.0003], device='cuda:0'), grad: tensor([-7.8045e-07, 1.0937e-04, 5.3120e-04, 6.1321e-04, 1.9109e-04, + 1.2064e-03, 2.4056e-04, -3.5286e-04, -2.6760e-03, 1.3494e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 265.37, cls_loss 0.0142 cls_loss_mapping 0.0233 cls_loss_causal 0.6752 re_mapping 0.0155 re_causal 0.0393 /// teacc 98.77 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0316, -0.0568, -0.0441, ..., -0.0200, 0.0972, 0.0781], + [-0.0916, -0.0971, -0.0974, ..., -0.0647, -0.0899, -0.0511], + [-0.0536, -0.0385, 0.0735, ..., -0.0583, 0.0917, 0.0032], + ..., + [-0.0540, 0.0361, 0.0388, ..., 0.0723, -0.0554, -0.0668], + [-0.1138, 0.0429, -0.0956, ..., 0.0244, -0.0435, -0.0377], + [ 0.0417, -0.0212, -0.0482, ..., -0.0050, -0.0346, -0.0073]], + device='cuda:0'), grad: tensor([[ 3.1114e-05, 1.5569e-04, -3.4124e-06, ..., 1.6633e-06, + 6.7353e-05, -1.9148e-05], + [ 5.8673e-07, 3.0547e-05, 4.6305e-06, ..., 2.0891e-05, + -3.9101e-04, 4.2878e-06], + [ 1.1310e-05, 1.8501e-04, -4.8205e-06, ..., 1.3359e-05, + 5.9247e-05, 7.4089e-05], + ..., + [ 1.2461e-06, -1.0781e-05, -4.2357e-06, ..., -7.0572e-05, + 2.8938e-05, 6.6385e-06], + [ 2.8536e-05, 3.1531e-05, 6.2771e-06, ..., 2.0817e-05, + 3.1590e-04, 3.6597e-05], + [-6.4187e-06, 8.6576e-06, -3.1609e-06, ..., 1.2301e-05, + 5.9828e-06, 1.8273e-06]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0141, -0.0235, -0.0140, -0.0278, -0.0311, -0.0014, 0.0234, -0.0103, + 0.0329, -0.0003], device='cuda:0'), grad: tensor([ 0.0002, -0.0057, 0.0005, -0.0007, -0.0013, 0.0001, 0.0021, 0.0002, + 0.0043, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 265.16, cls_loss 0.0152 cls_loss_mapping 0.0236 cls_loss_causal 0.6916 re_mapping 0.0143 re_causal 0.0368 /// teacc 98.79 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0316, -0.0573, -0.0444, ..., -0.0200, 0.0976, 0.0789], + [-0.0921, -0.0977, -0.0979, ..., -0.0653, -0.0907, -0.0511], + [-0.0543, -0.0380, 0.0741, ..., -0.0591, 0.0925, 0.0034], + ..., + [-0.0545, 0.0366, 0.0389, ..., 0.0730, -0.0567, -0.0677], + [-0.1146, 0.0430, -0.0962, ..., 0.0250, -0.0442, -0.0381], + [ 0.0420, -0.0219, -0.0482, ..., -0.0058, -0.0336, -0.0078]], + device='cuda:0'), grad: tensor([[ 4.4294e-06, 3.5726e-06, 2.7940e-07, ..., 2.3413e-06, + -8.4996e-05, -6.3837e-05], + [ 5.3570e-06, 1.1273e-05, 1.5274e-07, ..., 1.6037e-06, + 1.1977e-06, 7.1153e-07], + [ 7.5176e-06, 7.5340e-05, 3.9898e-06, ..., 1.4335e-05, + 1.2621e-05, 3.5055e-06], + ..., + [ 8.0913e-06, 1.4283e-05, 8.4564e-07, ..., -1.8492e-05, + -2.9802e-08, 4.8056e-07], + [ 2.4691e-05, 2.6867e-05, -1.6652e-06, ..., -1.0736e-05, + -2.6841e-06, 3.4124e-06], + [-7.2233e-06, 1.6198e-05, 1.2703e-06, ..., 9.9689e-06, + 5.2571e-05, 4.0799e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0147, -0.0229, -0.0140, -0.0283, -0.0312, -0.0010, 0.0229, -0.0106, + 0.0328, 0.0005], device='cuda:0'), grad: tensor([-7.4863e-05, -3.1665e-06, 1.1635e-04, -2.3711e-04, 8.4937e-05, + 4.4405e-05, 2.5779e-05, 5.5015e-05, 4.6760e-05, -5.8174e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 61---------------------------------------------------- +epoch 61, time 282.08, cls_loss 0.0121 cls_loss_mapping 0.0156 cls_loss_causal 0.6431 re_mapping 0.0137 re_causal 0.0341 /// teacc 98.80 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0324, -0.0579, -0.0446, ..., -0.0209, 0.0980, 0.0792], + [-0.0925, -0.0981, -0.0977, ..., -0.0659, -0.0913, -0.0511], + [-0.0549, -0.0381, 0.0745, ..., -0.0598, 0.0930, 0.0035], + ..., + [-0.0554, 0.0370, 0.0389, ..., 0.0737, -0.0574, -0.0682], + [-0.1155, 0.0430, -0.0964, ..., 0.0253, -0.0441, -0.0383], + [ 0.0418, -0.0225, -0.0485, ..., -0.0066, -0.0339, -0.0083]], + device='cuda:0'), grad: tensor([[ 1.2860e-05, 6.2212e-06, 2.4810e-06, ..., 2.0996e-05, + 3.5968e-06, -1.2822e-05], + [ 4.0419e-06, 2.0728e-05, 1.3471e-05, ..., 7.8321e-05, + -6.1207e-06, -9.1344e-06], + [ 3.5428e-06, 5.6595e-05, -1.3866e-05, ..., 3.6418e-05, + -2.6941e-04, 8.7395e-06], + ..., + [ 5.2452e-06, -2.0409e-04, -3.0637e-05, ..., -2.7895e-04, + -1.7267e-06, 1.3281e-06], + [ 2.2218e-05, 1.8671e-05, 8.4564e-07, ..., 7.9945e-06, + 4.6603e-06, 6.1467e-08], + [-2.9683e-04, -5.3532e-06, 9.2015e-06, ..., -2.5481e-05, + -2.5168e-05, 3.1106e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0152, -0.0223, -0.0141, -0.0284, -0.0305, -0.0006, 0.0232, -0.0109, + 0.0329, -0.0003], device='cuda:0'), grad: tensor([ 8.5115e-05, 1.4925e-04, -1.7717e-05, 4.1318e-04, 4.7898e-04, + 5.2118e-04, 4.2528e-05, -6.6328e-04, 5.8353e-05, -1.0672e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 265.67, cls_loss 0.0129 cls_loss_mapping 0.0170 cls_loss_causal 0.6684 re_mapping 0.0140 re_causal 0.0348 /// teacc 98.73 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0328, -0.0583, -0.0447, ..., -0.0210, 0.0985, 0.0796], + [-0.0933, -0.0987, -0.0983, ..., -0.0669, -0.0929, -0.0520], + [-0.0551, -0.0383, 0.0750, ..., -0.0607, 0.0944, 0.0048], + ..., + [-0.0548, 0.0372, 0.0390, ..., 0.0748, -0.0580, -0.0691], + [-0.1163, 0.0430, -0.0965, ..., 0.0254, -0.0451, -0.0388], + [ 0.0421, -0.0229, -0.0486, ..., -0.0072, -0.0344, -0.0087]], + device='cuda:0'), grad: tensor([[ 2.5660e-05, 1.9699e-05, 3.0845e-06, ..., 2.9653e-05, + -3.3349e-05, -1.2346e-05], + [ 6.8918e-06, 8.8438e-06, 8.1770e-07, ..., 1.7285e-05, + 2.5425e-06, 2.3078e-06], + [ 5.1484e-06, 7.0520e-06, 5.1409e-07, ..., 9.7305e-06, + -5.8375e-06, 1.1362e-07], + ..., + [-1.6659e-05, -2.8893e-05, -3.9302e-06, ..., -6.3956e-05, + 6.2287e-06, 4.2394e-06], + [-3.4366e-06, -5.4538e-05, 1.3858e-06, ..., -1.6928e-04, + 8.4341e-06, -3.2365e-05], + [ 3.6269e-05, 5.9187e-05, -4.4182e-06, ..., 1.9073e-04, + 7.9349e-06, 3.2395e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0153, -0.0229, -0.0134, -0.0283, -0.0308, -0.0009, 0.0233, -0.0110, + 0.0327, 0.0001], device='cuda:0'), grad: tensor([ 4.9025e-05, 3.7760e-05, 3.7134e-05, 5.4836e-05, -3.6502e-04, + 7.9513e-05, -2.6956e-05, 1.5581e-04, -9.9468e-04, 9.7227e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 63---------------------------------------------------- +epoch 63, time 281.33, cls_loss 0.0141 cls_loss_mapping 0.0188 cls_loss_causal 0.6417 re_mapping 0.0137 re_causal 0.0341 /// teacc 98.81 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0334, -0.0586, -0.0449, ..., -0.0213, 0.0991, 0.0804], + [-0.0943, -0.0994, -0.0981, ..., -0.0681, -0.0942, -0.0522], + [-0.0560, -0.0390, 0.0750, ..., -0.0617, 0.0951, 0.0051], + ..., + [-0.0550, 0.0378, 0.0393, ..., 0.0761, -0.0587, -0.0695], + [-0.1184, 0.0430, -0.0966, ..., 0.0254, -0.0462, -0.0392], + [ 0.0427, -0.0235, -0.0488, ..., -0.0079, -0.0339, -0.0094]], + device='cuda:0'), grad: tensor([[ 1.7241e-05, 1.1958e-06, 7.6368e-08, ..., 2.8804e-05, + -3.6693e-04, -3.1424e-04], + [ 3.3733e-06, 4.1090e-06, 2.2165e-07, ..., 8.4937e-05, + 1.5414e-04, 8.7142e-05], + [ 1.9092e-06, 3.9674e-07, -1.0654e-06, ..., 1.2815e-04, + 1.2147e-04, 2.2173e-05], + ..., + [ 4.0866e-06, 5.2862e-06, -3.2037e-07, ..., 4.8243e-06, + 1.4864e-05, 4.3809e-06], + [ 2.7224e-05, 1.9297e-05, 3.9674e-07, ..., -7.5579e-04, + -3.7670e-04, 1.4675e-04], + [ 1.4454e-05, 3.6135e-06, 2.6263e-07, ..., 1.8507e-05, + 3.4213e-05, 2.5436e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0157, -0.0226, -0.0137, -0.0281, -0.0307, -0.0015, 0.0239, -0.0105, + 0.0321, 0.0002], device='cuda:0'), grad: tensor([-2.7585e-04, 5.7507e-04, 8.1921e-04, 1.6260e-04, 9.7394e-05, + 2.2106e-03, 6.7472e-04, 1.3077e-04, -4.4632e-03, 6.9380e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 265.17, cls_loss 0.0137 cls_loss_mapping 0.0175 cls_loss_causal 0.6738 re_mapping 0.0138 re_causal 0.0352 /// teacc 98.80 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0338, -0.0591, -0.0451, ..., -0.0216, 0.1002, 0.0811], + [-0.0950, -0.1000, -0.0984, ..., -0.0689, -0.0951, -0.0521], + [-0.0568, -0.0393, 0.0757, ..., -0.0624, 0.0956, 0.0055], + ..., + [-0.0554, 0.0380, 0.0392, ..., 0.0767, -0.0593, -0.0702], + [-0.1195, 0.0430, -0.0970, ..., 0.0254, -0.0468, -0.0397], + [ 0.0434, -0.0235, -0.0490, ..., -0.0078, -0.0344, -0.0099]], + device='cuda:0'), grad: tensor([[ 4.6007e-06, 3.8892e-05, 7.7561e-06, ..., 1.9222e-05, + 4.4443e-06, 4.1202e-06], + [ 2.9411e-06, 4.8727e-05, 1.2666e-05, ..., 3.9339e-05, + 1.7434e-05, 2.0992e-06], + [ 2.5928e-05, 4.7708e-04, 1.9026e-04, ..., 4.7827e-04, + 2.3854e-04, 2.4699e-06], + ..., + [-2.8029e-05, -6.7091e-04, -2.6512e-04, ..., -6.9523e-04, + -3.1710e-04, 9.7044e-07], + [ 8.2329e-06, -1.2326e-04, 9.9167e-06, ..., 2.4229e-05, + 3.0756e-05, -6.2644e-05], + [ 2.7772e-06, 1.2839e-04, 1.5102e-05, ..., 4.1336e-05, + 2.1592e-05, 3.9309e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0155, -0.0226, -0.0134, -0.0279, -0.0304, -0.0024, 0.0245, -0.0111, + 0.0320, 0.0006], device='cuda:0'), grad: tensor([ 1.1611e-04, 1.4460e-04, 1.5898e-03, 1.3793e-04, 4.5300e-05, + 5.8085e-05, -1.0245e-06, -2.1534e-03, -3.1590e-04, 3.7861e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 65---------------------------------------------------- +epoch 65, time 281.46, cls_loss 0.0119 cls_loss_mapping 0.0165 cls_loss_causal 0.6781 re_mapping 0.0136 re_causal 0.0363 /// teacc 98.84 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0344, -0.0600, -0.0453, ..., -0.0219, 0.1008, 0.0817], + [-0.0954, -0.1008, -0.0986, ..., -0.0690, -0.0957, -0.0524], + [-0.0574, -0.0397, 0.0759, ..., -0.0637, 0.0963, 0.0056], + ..., + [-0.0563, 0.0384, 0.0394, ..., 0.0771, -0.0594, -0.0705], + [-0.1201, 0.0430, -0.0969, ..., 0.0255, -0.0470, -0.0402], + [ 0.0431, -0.0244, -0.0492, ..., -0.0083, -0.0348, -0.0102]], + device='cuda:0'), grad: tensor([[ 2.6729e-06, 3.6471e-06, 1.5832e-07, ..., -2.9847e-05, + -1.3494e-04, -1.0055e-04], + [ 3.7048e-06, 4.3184e-05, 2.9001e-06, ..., 1.5259e-04, + 1.0431e-05, 9.1568e-06], + [ 4.3735e-06, 1.1438e-04, 5.9828e-06, ..., 1.3304e-04, + 1.5870e-05, 1.4454e-06], + ..., + [ 7.5735e-06, -1.0425e-04, -8.3223e-06, ..., -9.7513e-05, + -1.2755e-05, 1.6242e-06], + [ 5.2899e-07, -1.0860e-04, 3.5763e-07, ..., -8.2445e-04, + -1.3858e-05, -1.3582e-05], + [ 6.4559e-06, 1.0818e-05, 6.0536e-07, ..., 1.5974e-05, + 1.0528e-05, 5.7444e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([-1.5696e-02, -2.1927e-02, -1.3872e-02, -2.7807e-02, -3.0770e-02, + -1.5916e-03, 2.4542e-02, -1.0917e-02, 3.1704e-02, 9.6543e-05], + device='cuda:0'), grad: tensor([-1.5485e-04, 3.9387e-04, 4.3249e-04, -6.4909e-05, 6.3276e-04, + -1.4267e-03, 2.1305e-03, -1.4639e-04, -1.8282e-03, 2.6956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 265.34, cls_loss 0.0129 cls_loss_mapping 0.0197 cls_loss_causal 0.6607 re_mapping 0.0133 re_causal 0.0350 /// teacc 98.83 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0344, -0.0604, -0.0456, ..., -0.0219, 0.1017, 0.0826], + [-0.0974, -0.1012, -0.0994, ..., -0.0702, -0.0973, -0.0533], + [-0.0560, -0.0411, 0.0762, ..., -0.0656, 0.0973, 0.0066], + ..., + [-0.0571, 0.0382, 0.0399, ..., 0.0776, -0.0598, -0.0720], + [-0.1202, 0.0432, -0.0968, ..., 0.0258, -0.0472, -0.0405], + [ 0.0430, -0.0252, -0.0495, ..., -0.0088, -0.0350, -0.0105]], + device='cuda:0'), grad: tensor([[ 2.1592e-05, 3.2149e-06, 2.6263e-07, ..., 3.7253e-06, + 3.4031e-06, 5.0552e-06], + [ 4.9360e-06, 3.8958e-04, 9.2015e-06, ..., 6.3801e-04, + 3.5197e-05, 1.6689e-05], + [ 1.9401e-05, 3.3021e-05, -1.2182e-05, ..., 6.8136e-06, + -4.6968e-05, -1.7360e-05], + ..., + [ 5.4613e-06, -5.5027e-04, -9.5554e-07, ..., -8.8406e-04, + 4.6007e-06, 2.8089e-06], + [ 1.8284e-05, 4.7892e-05, 1.1344e-06, ..., 7.9453e-05, + 4.2506e-06, 2.6897e-06], + [-4.1574e-05, 7.1287e-05, 1.0524e-06, ..., 8.2076e-05, + 1.0841e-06, 1.1083e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0153, -0.0232, -0.0133, -0.0275, -0.0311, -0.0017, 0.0242, -0.0107, + 0.0326, -0.0003], device='cuda:0'), grad: tensor([ 3.2544e-05, 1.1053e-03, -1.9237e-05, -6.5982e-05, 1.8966e-04, + 7.2896e-05, -1.2957e-05, -1.3857e-03, 1.8251e-04, -1.0067e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 265.12, cls_loss 0.0109 cls_loss_mapping 0.0153 cls_loss_causal 0.6328 re_mapping 0.0137 re_causal 0.0351 /// teacc 98.78 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0347, -0.0609, -0.0457, ..., -0.0224, 0.1028, 0.0831], + [-0.0988, -0.1027, -0.0997, ..., -0.0715, -0.0984, -0.0539], + [-0.0564, -0.0414, 0.0766, ..., -0.0663, 0.0979, 0.0069], + ..., + [-0.0578, 0.0388, 0.0402, ..., 0.0786, -0.0601, -0.0724], + [-0.1210, 0.0431, -0.0972, ..., 0.0262, -0.0474, -0.0405], + [ 0.0433, -0.0259, -0.0498, ..., -0.0094, -0.0358, -0.0109]], + device='cuda:0'), grad: tensor([[ 4.6715e-06, 2.6412e-06, 1.9558e-08, ..., 2.6450e-07, + -3.0696e-05, -1.1936e-05], + [ 4.2934e-07, 2.2091e-06, 1.4249e-07, ..., 1.8319e-06, + 2.6878e-06, -2.1076e-04], + [ 3.2540e-06, 2.7314e-05, -2.8312e-07, ..., 2.0880e-06, + 1.6302e-05, 9.6321e-05], + ..., + [ 4.4331e-07, -1.7598e-05, 2.7195e-07, ..., -1.0997e-05, + 3.6508e-06, 7.2360e-05], + [ 2.6040e-06, 1.1750e-05, 7.6368e-08, ..., -3.3733e-06, + 8.4266e-06, 1.7956e-05], + [-2.5611e-06, 1.5851e-06, 3.5856e-07, ..., 3.5204e-07, + 1.1601e-05, 1.5020e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0152, -0.0242, -0.0136, -0.0276, -0.0316, -0.0012, 0.0248, -0.0102, + 0.0329, -0.0004], device='cuda:0'), grad: tensor([ 1.6332e-05, -9.7132e-04, 6.1607e-04, -5.9247e-05, -4.5753e-04, + 4.5002e-05, 9.5963e-05, 5.4121e-04, 8.3804e-05, 9.0599e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 68---------------------------------------------------- +epoch 68, time 281.68, cls_loss 0.0112 cls_loss_mapping 0.0139 cls_loss_causal 0.6863 re_mapping 0.0132 re_causal 0.0341 /// teacc 98.88 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0359, -0.0612, -0.0459, ..., -0.0227, 0.1039, 0.0837], + [-0.0996, -0.1031, -0.1000, ..., -0.0721, -0.0996, -0.0541], + [-0.0569, -0.0419, 0.0766, ..., -0.0670, 0.0984, 0.0069], + ..., + [-0.0581, 0.0395, 0.0407, ..., 0.0794, -0.0607, -0.0726], + [-0.1221, 0.0433, -0.0975, ..., 0.0266, -0.0476, -0.0407], + [ 0.0437, -0.0264, -0.0500, ..., -0.0098, -0.0360, -0.0114]], + device='cuda:0'), grad: tensor([[ 6.5416e-06, 3.4180e-07, 8.7544e-08, ..., -1.7524e-05, + -1.2887e-04, -8.5115e-05], + [ 1.1949e-06, 1.1489e-05, 5.2154e-08, ..., 6.4038e-06, + 6.7428e-07, 2.9244e-07], + [ 2.2035e-06, 1.9949e-06, -2.0023e-06, ..., 2.2855e-06, + -4.5188e-06, -3.8557e-07], + ..., + [ 8.0280e-07, -7.6830e-05, 2.4587e-07, ..., -3.7223e-05, + 1.1083e-06, 6.2771e-07], + [ 1.7181e-05, 2.5462e-06, 1.2033e-06, ..., -6.6310e-06, + 9.4399e-06, 9.5144e-06], + [-1.3635e-05, 3.5763e-07, 7.4506e-09, ..., 6.1989e-06, + 1.3225e-06, 9.7137e-07]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0149, -0.0238, -0.0133, -0.0276, -0.0317, -0.0018, 0.0250, -0.0103, + 0.0327, -0.0003], device='cuda:0'), grad: tensor([-1.3340e-04, 3.0637e-04, 1.5676e-05, 9.3877e-05, 5.6553e-04, + 3.6955e-05, 1.1021e-04, -5.9158e-05, 1.4222e-04, -1.0767e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 265.81, cls_loss 0.0110 cls_loss_mapping 0.0166 cls_loss_causal 0.6154 re_mapping 0.0130 re_causal 0.0328 /// teacc 98.72 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0361, -0.0616, -0.0460, ..., -0.0228, 0.1036, 0.0838], + [-0.0999, -0.1029, -0.1002, ..., -0.0714, -0.1001, -0.0538], + [-0.0577, -0.0432, 0.0770, ..., -0.0679, 0.0982, 0.0067], + ..., + [-0.0584, 0.0395, 0.0406, ..., 0.0794, -0.0610, -0.0733], + [-0.1231, 0.0432, -0.0978, ..., 0.0266, -0.0485, -0.0412], + [ 0.0437, -0.0263, -0.0501, ..., -0.0098, -0.0354, -0.0119]], + device='cuda:0'), grad: tensor([[ 1.6484e-06, -4.3884e-06, 2.0582e-07, ..., -6.5923e-05, + -1.0939e-03, -5.4312e-04], + [ 7.1432e-07, 6.6459e-05, 1.1921e-07, ..., 1.9908e-05, + 8.9481e-06, 4.7907e-06], + [ 4.6939e-07, 1.2165e-04, 1.4715e-07, ..., 3.9726e-05, + 1.8150e-05, 1.1243e-05], + ..., + [ 5.4110e-07, -1.0872e-03, 1.9930e-07, ..., -2.6488e-04, + 1.6928e-05, 1.0557e-05], + [ 4.7125e-06, 7.3649e-06, 4.3306e-07, ..., 3.1054e-05, + 1.4472e-04, 1.1861e-04], + [-7.3202e-07, 5.4407e-04, -3.0920e-07, ..., 1.2010e-04, + 1.5092e-04, 6.6638e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0158, -0.0223, -0.0146, -0.0275, -0.0308, -0.0016, 0.0252, -0.0113, + 0.0324, 0.0001], device='cuda:0'), grad: tensor([-1.5755e-03, 1.1009e-04, 2.0659e-04, 3.7217e-04, 2.0528e-04, + 6.6698e-05, 1.0099e-03, -1.4505e-03, 1.5497e-04, 8.9931e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 265.24, cls_loss 0.0120 cls_loss_mapping 0.0149 cls_loss_causal 0.6525 re_mapping 0.0127 re_causal 0.0332 /// teacc 98.72 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0365, -0.0619, -0.0462, ..., -0.0228, 0.1047, 0.0846], + [-0.1005, -0.1046, -0.1008, ..., -0.0728, -0.1006, -0.0539], + [-0.0581, -0.0438, 0.0773, ..., -0.0689, 0.0986, 0.0071], + ..., + [-0.0588, 0.0402, 0.0407, ..., 0.0805, -0.0615, -0.0738], + [-0.1241, 0.0435, -0.0965, ..., 0.0270, -0.0492, -0.0417], + [ 0.0437, -0.0270, -0.0503, ..., -0.0103, -0.0357, -0.0125]], + device='cuda:0'), grad: tensor([[ 4.6454e-06, 1.1988e-05, 3.9823e-06, ..., -9.5546e-05, + -1.1320e-03, -7.8726e-04], + [ 2.9299e-06, 3.0816e-05, 6.5342e-06, ..., 1.7643e-05, + 1.5891e-04, 1.1069e-04], + [ 4.0159e-06, 1.0371e-04, 1.7568e-05, ..., 8.6308e-05, + 3.4833e-04, 2.6512e-04], + ..., + [ 2.7232e-06, -2.3830e-04, -5.6446e-05, ..., -1.3375e-04, + 1.3936e-04, 6.4373e-05], + [ 7.6257e-06, 3.8683e-05, 6.0536e-06, ..., 2.6181e-05, + 6.5207e-05, 4.6849e-05], + [ 3.8631e-06, 3.9339e-05, 6.1132e-06, ..., 2.4676e-05, + 1.7747e-05, 1.1854e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0153, -0.0237, -0.0142, -0.0274, -0.0311, -0.0018, 0.0252, -0.0103, + 0.0329, -0.0002], device='cuda:0'), grad: tensor([-1.7014e-03, 2.7800e-04, 7.3862e-04, 3.5048e-05, 1.0097e-04, + 1.0300e-04, 4.4250e-04, -2.6035e-04, 1.9276e-04, 7.1406e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 265.35, cls_loss 0.0130 cls_loss_mapping 0.0142 cls_loss_causal 0.6438 re_mapping 0.0132 re_causal 0.0336 /// teacc 98.78 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0376, -0.0623, -0.0464, ..., -0.0232, 0.1055, 0.0852], + [-0.1010, -0.1062, -0.1010, ..., -0.0733, -0.1009, -0.0535], + [-0.0586, -0.0447, 0.0778, ..., -0.0699, 0.0996, 0.0069], + ..., + [-0.0594, 0.0405, 0.0408, ..., 0.0813, -0.0621, -0.0743], + [-0.1251, 0.0437, -0.0970, ..., 0.0260, -0.0500, -0.0421], + [ 0.0441, -0.0282, -0.0507, ..., -0.0116, -0.0360, -0.0130]], + device='cuda:0'), grad: tensor([[ 1.2159e-05, 6.6720e-06, 2.6077e-07, ..., 1.0356e-05, + 4.6082e-06, 7.4171e-06], + [-2.5272e-05, 1.6809e-05, 2.8592e-07, ..., 7.7933e-06, + 9.9465e-07, 7.4785e-07], + [ 8.8364e-06, 3.6210e-05, -2.4568e-06, ..., 1.4737e-05, + -5.7183e-06, -3.7365e-06], + ..., + [ 2.6878e-06, -2.3448e-04, 6.7241e-07, ..., -1.1295e-04, + 1.0524e-06, 5.3551e-07], + [ 1.1012e-05, 1.1530e-06, 7.8324e-07, ..., -4.4815e-06, + 5.6475e-06, 5.3383e-06], + [ 7.4208e-06, 8.8632e-05, 3.6601e-07, ..., 3.8713e-05, + 1.0813e-06, 8.1118e-07]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0151, -0.0232, -0.0144, -0.0272, -0.0307, -0.0009, 0.0251, -0.0102, + 0.0318, -0.0011], device='cuda:0'), grad: tensor([ 5.4985e-05, -3.3417e-03, 1.2469e-04, 7.8261e-05, -2.6274e-04, + 2.1970e-04, 8.8289e-07, -2.3389e-04, 2.5787e-03, 7.8058e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 264.95, cls_loss 0.0130 cls_loss_mapping 0.0179 cls_loss_causal 0.6449 re_mapping 0.0128 re_causal 0.0323 /// teacc 98.85 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0381, -0.0627, -0.0466, ..., -0.0230, 0.1059, 0.0855], + [-0.1015, -0.1078, -0.1010, ..., -0.0743, -0.1015, -0.0524], + [-0.0591, -0.0445, 0.0782, ..., -0.0707, 0.1001, 0.0073], + ..., + [-0.0593, 0.0404, 0.0409, ..., 0.0815, -0.0625, -0.0750], + [-0.1247, 0.0450, -0.0979, ..., 0.0283, -0.0494, -0.0424], + [ 0.0443, -0.0290, -0.0507, ..., -0.0121, -0.0361, -0.0132]], + device='cuda:0'), grad: tensor([[ 2.4401e-06, 7.0632e-05, 3.9302e-07, ..., -2.4959e-06, + 8.8751e-05, 6.2764e-05], + [ 7.1712e-06, 8.3521e-06, 1.5274e-07, ..., 8.7731e-07, + 1.0766e-05, 7.6108e-06], + [ 5.8860e-07, -6.8855e-04, -4.2878e-06, ..., 9.4436e-07, + -9.2840e-04, -6.8045e-04], + ..., + [ 5.0962e-06, 4.9305e-04, 2.3991e-06, ..., -3.4012e-06, + 6.7759e-04, 4.8876e-04], + [ 6.6981e-06, 3.0056e-05, 9.0711e-07, ..., 2.4159e-06, + 3.4720e-05, 2.3365e-05], + [-1.0949e-04, 5.7697e-05, 2.6226e-06, ..., -4.2841e-06, + 1.2493e-04, 7.4029e-05]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0153, -0.0233, -0.0146, -0.0273, -0.0307, -0.0012, 0.0238, -0.0103, + 0.0332, -0.0007], device='cuda:0'), grad: tensor([ 2.7442e-04, 1.2040e-05, -2.4643e-03, 7.1764e-05, 1.4467e-03, + 7.5698e-05, 8.5592e-05, 1.9588e-03, 1.9836e-04, -1.6613e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 265.31, cls_loss 0.0130 cls_loss_mapping 0.0172 cls_loss_causal 0.6246 re_mapping 0.0124 re_causal 0.0307 /// teacc 98.81 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0392, -0.0634, -0.0466, ..., -0.0235, 0.1063, 0.0858], + [-0.1023, -0.1074, -0.1013, ..., -0.0754, -0.1022, -0.0528], + [-0.0596, -0.0444, 0.0785, ..., -0.0714, 0.1006, 0.0078], + ..., + [-0.0591, 0.0405, 0.0408, ..., 0.0828, -0.0636, -0.0760], + [-0.1257, 0.0452, -0.0972, ..., 0.0285, -0.0499, -0.0428], + [ 0.0450, -0.0298, -0.0508, ..., -0.0132, -0.0364, -0.0142]], + device='cuda:0'), grad: tensor([[ 1.9744e-05, 7.4357e-06, 2.1309e-06, ..., 1.0997e-05, + -5.3123e-06, -2.7083e-06], + [ 2.0087e-05, 2.0340e-05, 4.2021e-06, ..., 2.3156e-05, + 1.6123e-05, 6.1058e-06], + [ 2.3127e-05, -2.5462e-06, -5.6595e-05, ..., 4.8578e-05, + -1.4114e-04, -7.2837e-05], + ..., + [ 3.4362e-05, -2.6762e-05, 4.1455e-05, ..., -3.0667e-05, + 1.0407e-04, 4.7088e-05], + [ 7.7188e-05, 1.1559e-03, 8.3447e-07, ..., -3.5930e-06, + -1.4409e-05, 2.7958e-06], + [ 2.7910e-05, -1.2131e-03, 1.0617e-07, ..., 2.1800e-05, + 4.3139e-06, 2.4773e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0156, -0.0228, -0.0144, -0.0276, -0.0302, -0.0009, 0.0238, -0.0104, + 0.0332, -0.0013], device='cuda:0'), grad: tensor([ 5.1051e-05, 1.6427e-04, 2.8181e-04, 8.0299e-04, -8.1348e-04, + -9.2506e-04, 2.3985e-04, 2.2018e-04, 2.3346e-03, -2.3518e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 265.16, cls_loss 0.0114 cls_loss_mapping 0.0146 cls_loss_causal 0.6592 re_mapping 0.0124 re_causal 0.0320 /// teacc 98.68 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0399, -0.0640, -0.0468, ..., -0.0238, 0.1068, 0.0861], + [-0.1028, -0.1070, -0.1019, ..., -0.0755, -0.1030, -0.0531], + [-0.0601, -0.0445, 0.0795, ..., -0.0720, 0.1014, 0.0084], + ..., + [-0.0595, 0.0407, 0.0407, ..., 0.0836, -0.0645, -0.0768], + [-0.1264, 0.0453, -0.0978, ..., 0.0283, -0.0503, -0.0426], + [ 0.0457, -0.0302, -0.0511, ..., -0.0136, -0.0364, -0.0148]], + device='cuda:0'), grad: tensor([[ 3.7611e-05, 4.2841e-07, 5.0329e-06, ..., 1.6883e-05, + -7.1786e-06, -2.5723e-06], + [ 8.2180e-06, 1.7695e-06, 1.5441e-06, ..., 6.2473e-06, + 3.8520e-06, 4.0457e-06], + [ 3.4515e-06, 2.9076e-06, -1.0028e-05, ..., 9.2983e-06, + -1.8850e-06, 3.7532e-06], + ..., + [ 5.5879e-06, 3.4645e-07, 8.5682e-06, ..., -5.2899e-07, + 9.6187e-06, 4.0345e-06], + [ 1.1891e-04, -7.4506e-06, 8.4750e-07, ..., 1.2434e-04, + 7.6964e-06, 7.7710e-06], + [ 1.4961e-05, 6.7428e-07, 6.2101e-06, ..., 1.1019e-05, + 1.8507e-05, 8.0094e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0158, -0.0225, -0.0141, -0.0278, -0.0303, -0.0008, 0.0234, -0.0108, + 0.0333, -0.0011], device='cuda:0'), grad: tensor([ 4.6164e-05, 1.4737e-05, 1.5348e-05, 6.1131e-04, 3.0994e-05, + -1.1139e-03, 1.1122e-04, 3.2812e-05, 2.0659e-04, 4.4107e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 265.37, cls_loss 0.0116 cls_loss_mapping 0.0156 cls_loss_causal 0.6792 re_mapping 0.0127 re_causal 0.0312 /// teacc 98.79 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0406, -0.0647, -0.0470, ..., -0.0241, 0.1071, 0.0863], + [-0.1032, -0.1074, -0.1025, ..., -0.0759, -0.1038, -0.0534], + [-0.0605, -0.0448, 0.0798, ..., -0.0728, 0.1022, 0.0086], + ..., + [-0.0595, 0.0412, 0.0412, ..., 0.0846, -0.0651, -0.0773], + [-0.1271, 0.0455, -0.0975, ..., 0.0287, -0.0504, -0.0428], + [ 0.0455, -0.0304, -0.0511, ..., -0.0144, -0.0367, -0.0152]], + device='cuda:0'), grad: tensor([[ 2.2128e-06, 6.7592e-05, 1.7751e-06, ..., 1.6280e-06, + -1.0781e-05, -1.5303e-05], + [ 3.4831e-07, 3.9577e-05, 5.8740e-05, ..., 2.7847e-06, + 4.6760e-05, 9.3356e-06], + [ 4.5449e-07, 4.5013e-04, -5.9187e-05, ..., 7.6443e-06, + -2.0370e-05, -1.3106e-05], + ..., + [ 3.1479e-07, 8.9943e-05, 9.9987e-06, ..., -2.8118e-05, + 1.4655e-05, 2.7902e-06], + [ 3.2075e-06, 3.1352e-05, 2.8703e-06, ..., 3.6899e-06, + 1.3627e-05, 8.1211e-06], + [ 8.4750e-07, 4.2647e-05, -1.0408e-05, ..., 8.7693e-06, + 2.9039e-06, 3.1982e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0163, -0.0227, -0.0139, -0.0271, -0.0303, -0.0020, 0.0240, -0.0110, + 0.0338, -0.0009], device='cuda:0'), grad: tensor([ 1.8418e-04, 3.4142e-04, 9.6321e-04, -1.9569e-03, 1.5378e-05, + 4.2796e-05, 1.0110e-05, 3.4118e-04, 1.0854e-04, -4.8429e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 265.20, cls_loss 0.0099 cls_loss_mapping 0.0129 cls_loss_causal 0.6261 re_mapping 0.0120 re_causal 0.0312 /// teacc 98.80 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0409, -0.0653, -0.0473, ..., -0.0246, 0.1067, 0.0863], + [-0.1034, -0.1080, -0.1029, ..., -0.0768, -0.1042, -0.0535], + [-0.0607, -0.0452, 0.0802, ..., -0.0737, 0.1029, 0.0093], + ..., + [-0.0595, 0.0419, 0.0414, ..., 0.0857, -0.0658, -0.0782], + [-0.1279, 0.0454, -0.0978, ..., 0.0285, -0.0506, -0.0433], + [ 0.0453, -0.0310, -0.0511, ..., -0.0147, -0.0358, -0.0150]], + device='cuda:0'), grad: tensor([[ 1.3649e-04, 1.2331e-05, 2.1979e-07, ..., 1.6928e-05, + 2.2840e-04, 2.4915e-04], + [ 2.8126e-07, 9.1046e-06, 1.3225e-07, ..., 1.6075e-06, + 1.1921e-07, -7.3791e-05], + [ 6.5565e-07, 8.2433e-05, -5.1223e-07, ..., 3.3788e-06, + -1.3225e-06, 3.7458e-06], + ..., + [-4.4703e-06, 3.5226e-05, -1.2554e-06, ..., -6.1274e-05, + 7.2457e-07, 3.5781e-06], + [ 2.4904e-06, 8.7991e-06, 1.8440e-07, ..., 3.5260e-06, + 5.6252e-07, 3.1412e-05], + [ 1.5069e-06, 2.2992e-05, 1.2945e-06, ..., 2.0966e-05, + 1.0803e-07, 4.0084e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0180, -0.0233, -0.0139, -0.0273, -0.0298, -0.0018, 0.0241, -0.0107, + 0.0337, -0.0002], device='cuda:0'), grad: tensor([ 3.6454e-04, -5.0783e-04, 1.5473e-04, -2.6679e-04, 4.1556e-04, + 4.8101e-05, -1.8764e-04, 7.4089e-05, 2.3997e-04, -3.3307e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 265.59, cls_loss 0.0114 cls_loss_mapping 0.0156 cls_loss_causal 0.6235 re_mapping 0.0122 re_causal 0.0312 /// teacc 98.82 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0414, -0.0660, -0.0475, ..., -0.0251, 0.1075, 0.0871], + [-0.1038, -0.1084, -0.1037, ..., -0.0768, -0.1053, -0.0538], + [-0.0611, -0.0456, 0.0807, ..., -0.0743, 0.1036, 0.0097], + ..., + [-0.0597, 0.0413, 0.0420, ..., 0.0860, -0.0663, -0.0786], + [-0.1290, 0.0455, -0.0986, ..., 0.0279, -0.0511, -0.0440], + [ 0.0461, -0.0319, -0.0515, ..., -0.0156, -0.0362, -0.0155]], + device='cuda:0'), grad: tensor([[ 1.0446e-05, 1.8496e-06, 4.3660e-06, ..., 5.0291e-07, + -7.1786e-06, 1.9930e-07], + [ 1.1548e-05, 4.2915e-06, 8.8103e-07, ..., 5.5693e-06, + 1.5482e-05, 1.0513e-05], + [ 4.5337e-06, 1.0997e-05, 3.5018e-07, ..., 4.6752e-06, + -1.1303e-05, -8.6203e-06], + ..., + [ 1.0375e-06, -7.1079e-06, -2.4848e-06, ..., -1.1571e-05, + 6.4559e-06, 4.1611e-06], + [ 5.2601e-06, 3.7216e-06, 5.6997e-07, ..., -4.4852e-06, + 1.7196e-05, 1.1578e-05], + [ 1.7714e-06, 6.1095e-06, 6.4634e-07, ..., 1.7416e-06, + 5.8599e-06, 3.4142e-06]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0178, -0.0230, -0.0137, -0.0273, -0.0293, -0.0010, 0.0239, -0.0110, + 0.0327, -0.0003], device='cuda:0'), grad: tensor([ 1.4096e-05, 5.9605e-05, 7.4320e-07, -3.4273e-05, 4.7803e-04, + 6.8188e-05, -5.5933e-04, 9.7975e-07, -4.1500e-06, -2.3589e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 265.79, cls_loss 0.0100 cls_loss_mapping 0.0138 cls_loss_causal 0.6601 re_mapping 0.0118 re_causal 0.0313 /// teacc 98.83 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0425, -0.0664, -0.0477, ..., -0.0255, 0.1084, 0.0878], + [-0.1043, -0.1088, -0.1046, ..., -0.0775, -0.1061, -0.0540], + [-0.0616, -0.0463, 0.0811, ..., -0.0759, 0.1042, 0.0100], + ..., + [-0.0602, 0.0417, 0.0428, ..., 0.0871, -0.0669, -0.0791], + [-0.1296, 0.0455, -0.0991, ..., 0.0279, -0.0516, -0.0443], + [ 0.0459, -0.0322, -0.0519, ..., -0.0161, -0.0364, -0.0158]], + device='cuda:0'), grad: tensor([[ 3.6657e-06, 5.9418e-07, 2.9430e-07, ..., -3.4273e-05, + -7.8976e-05, -3.2514e-05], + [ 2.0359e-06, 2.9039e-06, 3.7812e-07, ..., 1.4529e-06, + 3.1665e-06, 1.5572e-06], + [ 2.2054e-06, 2.9683e-05, -1.8887e-06, ..., 8.6501e-06, + -3.2932e-06, -5.3644e-07], + ..., + [ 7.0222e-07, 1.5154e-05, -1.6522e-06, ..., -2.5164e-06, + 7.7114e-06, 2.5798e-06], + [ 6.3896e-05, -9.1121e-06, 4.5821e-07, ..., -2.0921e-05, + 6.0238e-06, 3.3855e-05], + [ 1.6149e-06, 5.2862e-06, 9.5740e-07, ..., 3.5483e-06, + 4.0717e-06, 1.6484e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0174, -0.0238, -0.0140, -0.0277, -0.0291, -0.0005, 0.0240, -0.0106, + 0.0328, -0.0005], device='cuda:0'), grad: tensor([-6.0767e-05, 1.2591e-05, 3.4720e-05, -7.4506e-05, 4.5896e-05, + 5.5820e-05, -6.6578e-05, 2.5615e-05, 4.8757e-05, -2.1517e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 265.73, cls_loss 0.0095 cls_loss_mapping 0.0127 cls_loss_causal 0.6425 re_mapping 0.0121 re_causal 0.0319 /// teacc 98.72 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0429, -0.0667, -0.0479, ..., -0.0256, 0.1089, 0.0883], + [-0.1048, -0.1093, -0.1047, ..., -0.0782, -0.1069, -0.0541], + [-0.0622, -0.0470, 0.0812, ..., -0.0779, 0.1049, 0.0101], + ..., + [-0.0606, 0.0423, 0.0432, ..., 0.0884, -0.0676, -0.0796], + [-0.1300, 0.0455, -0.0992, ..., 0.0282, -0.0516, -0.0446], + [ 0.0458, -0.0326, -0.0521, ..., -0.0166, -0.0368, -0.0163]], + device='cuda:0'), grad: tensor([[ 1.5274e-06, 7.5623e-07, 5.6811e-07, ..., 1.0412e-06, + -2.8759e-06, -1.3560e-06], + [ 3.3155e-07, 3.9376e-06, 3.3248e-06, ..., 6.0759e-06, + 9.5367e-07, 5.4389e-07], + [ 5.0850e-07, 1.2398e-05, 7.0184e-06, ..., 1.0684e-05, + 2.8580e-05, -4.6007e-06], + ..., + [ 1.3281e-06, -9.1732e-05, -2.3041e-06, ..., -4.2677e-05, + -5.5879e-07, 1.0207e-06], + [ 1.5385e-06, 3.7104e-06, 3.0566e-06, ..., 2.4103e-06, + 2.8927e-06, 1.7472e-06], + [-1.4957e-06, 6.1654e-06, 1.3113e-06, ..., 4.6194e-06, + 7.2643e-07, 4.3027e-07]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0174, -0.0232, -0.0138, -0.0281, -0.0297, -0.0004, 0.0233, -0.0102, + 0.0327, -0.0005], device='cuda:0'), grad: tensor([ 2.8118e-05, 7.5758e-05, 5.8830e-05, 1.1235e-04, -8.7261e-04, + 3.7193e-05, 7.9215e-05, -4.5121e-05, 1.1826e-04, 4.0865e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 265.32, cls_loss 0.0094 cls_loss_mapping 0.0133 cls_loss_causal 0.6273 re_mapping 0.0118 re_causal 0.0303 /// teacc 98.81 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0432, -0.0670, -0.0480, ..., -0.0259, 0.1101, 0.0894], + [-0.1051, -0.1097, -0.1048, ..., -0.0786, -0.1080, -0.0544], + [-0.0627, -0.0472, 0.0814, ..., -0.0785, 0.1051, 0.0102], + ..., + [-0.0607, 0.0423, 0.0433, ..., 0.0889, -0.0681, -0.0801], + [-0.1308, 0.0453, -0.0994, ..., 0.0283, -0.0524, -0.0452], + [ 0.0458, -0.0317, -0.0523, ..., -0.0169, -0.0372, -0.0172]], + device='cuda:0'), grad: tensor([[-2.9117e-05, 1.7546e-06, 3.9116e-08, ..., -3.5822e-05, + -9.1553e-04, -7.3385e-04], + [ 1.3106e-05, 5.3421e-06, 1.3411e-07, ..., 2.1979e-06, + 1.4313e-05, 9.9540e-06], + [ 1.3523e-06, -1.8954e-05, -9.6671e-07, ..., 2.0675e-07, + -2.6468e-06, -3.1460e-06], + ..., + [ 2.0042e-06, -3.6322e-07, 1.4342e-07, ..., -9.7305e-06, + 8.9630e-06, 6.0797e-06], + [ 1.2338e-05, 2.0444e-05, 3.1851e-07, ..., 3.8706e-06, + 5.0813e-05, 3.5375e-05], + [ 9.0972e-06, 7.7039e-06, 7.4506e-09, ..., 9.1419e-06, + 7.4983e-05, 4.5300e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0169, -0.0232, -0.0141, -0.0280, -0.0299, -0.0006, 0.0237, -0.0106, + 0.0328, -0.0002], device='cuda:0'), grad: tensor([-1.2360e-03, -2.1553e-04, -2.9474e-05, 9.1732e-05, 9.0647e-04, + 7.4565e-05, 8.3590e-04, 8.2076e-05, 2.0885e-04, -7.1907e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 265.30, cls_loss 0.0116 cls_loss_mapping 0.0137 cls_loss_causal 0.6138 re_mapping 0.0110 re_causal 0.0274 /// teacc 98.81 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0437, -0.0680, -0.0482, ..., -0.0266, 0.1108, 0.0903], + [-0.1055, -0.1101, -0.1048, ..., -0.0791, -0.1103, -0.0547], + [-0.0632, -0.0476, 0.0819, ..., -0.0788, 0.1067, 0.0112], + ..., + [-0.0611, 0.0428, 0.0432, ..., 0.0894, -0.0681, -0.0814], + [-0.1319, 0.0454, -0.0994, ..., 0.0284, -0.0528, -0.0455], + [ 0.0461, -0.0326, -0.0525, ..., -0.0172, -0.0379, -0.0179]], + device='cuda:0'), grad: tensor([[ 1.0021e-06, -1.4588e-05, 1.6391e-07, ..., -6.2436e-06, + -1.6880e-04, -9.2447e-05], + [ 1.7881e-07, 3.9786e-06, 3.1665e-08, ..., 7.2829e-06, + 1.4640e-06, 8.1398e-07], + [ 9.4995e-07, 1.3225e-06, 4.3772e-07, ..., 2.5630e-06, + 2.1815e-05, 1.2539e-05], + ..., + [ 5.7556e-07, -8.6427e-06, -6.4820e-07, ..., -1.8910e-05, + 6.7614e-06, 3.3416e-06], + [ 3.4943e-06, 8.4937e-06, 5.2154e-08, ..., 3.9004e-06, + 7.7426e-05, 4.1723e-05], + [ 9.8422e-06, 4.4405e-06, 5.2154e-08, ..., 4.3251e-06, + 4.8488e-05, 2.2411e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([-1.6904e-02, -2.3820e-02, -1.3472e-02, -2.8182e-02, -3.0519e-02, + -6.4255e-04, 2.3336e-02, -1.0101e-02, 3.3028e-02, -3.5299e-05], + device='cuda:0'), grad: tensor([-2.2829e-04, 1.5110e-05, 3.4243e-05, 6.1840e-06, -7.7486e-05, + 1.5959e-05, 2.5079e-05, -4.7162e-06, 1.0866e-04, 1.0508e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 265.55, cls_loss 0.0096 cls_loss_mapping 0.0114 cls_loss_causal 0.5904 re_mapping 0.0124 re_causal 0.0290 /// teacc 98.84 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0442, -0.0686, -0.0483, ..., -0.0269, 0.1124, 0.0913], + [-0.1058, -0.1109, -0.1045, ..., -0.0798, -0.1112, -0.0547], + [-0.0637, -0.0488, 0.0823, ..., -0.0793, 0.1071, 0.0111], + ..., + [-0.0612, 0.0428, 0.0431, ..., 0.0900, -0.0688, -0.0819], + [-0.1328, 0.0451, -0.0997, ..., 0.0285, -0.0532, -0.0460], + [ 0.0467, -0.0320, -0.0526, ..., -0.0168, -0.0384, -0.0184]], + device='cuda:0'), grad: tensor([[ 4.7833e-05, 2.9817e-05, 3.1106e-07, ..., 1.8016e-05, + 6.6996e-05, 4.1306e-05], + [ 1.3057e-06, 1.1489e-05, 2.0489e-08, ..., 5.1707e-05, + 1.7151e-05, 6.0536e-07], + [ 2.0750e-06, 1.2919e-05, -1.9558e-07, ..., 7.2539e-05, + 2.5257e-05, 3.5409e-06], + ..., + [ 2.9281e-06, -1.4174e-04, 5.0291e-08, ..., -1.1188e-04, + 1.2983e-06, 2.5705e-07], + [ 1.5050e-05, 1.9312e-05, 2.9802e-08, ..., -8.3780e-04, + -3.1662e-04, 3.0771e-06], + [ 7.1898e-06, 8.2076e-05, 7.4506e-09, ..., 6.6996e-05, + 7.1675e-06, 3.6545e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0161, -0.0235, -0.0135, -0.0278, -0.0316, -0.0011, 0.0230, -0.0107, + 0.0328, 0.0012], device='cuda:0'), grad: tensor([ 1.6260e-04, 1.7190e-04, 2.9755e-04, -3.5810e-04, 5.7966e-05, + 2.8896e-03, 4.9543e-04, -1.6665e-04, -3.6869e-03, 1.3769e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 83---------------------------------------------------- +epoch 83, time 282.41, cls_loss 0.0095 cls_loss_mapping 0.0117 cls_loss_causal 0.6118 re_mapping 0.0116 re_causal 0.0281 /// teacc 98.90 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0452, -0.0691, -0.0485, ..., -0.0273, 0.1132, 0.0921], + [-0.1062, -0.1117, -0.1051, ..., -0.0801, -0.1122, -0.0549], + [-0.0641, -0.0496, 0.0831, ..., -0.0799, 0.1079, 0.0113], + ..., + [-0.0614, 0.0432, 0.0432, ..., 0.0909, -0.0692, -0.0823], + [-0.1335, 0.0454, -0.1000, ..., 0.0293, -0.0535, -0.0465], + [ 0.0468, -0.0324, -0.0531, ..., -0.0176, -0.0387, -0.0188]], + device='cuda:0'), grad: tensor([[ 4.4741e-06, 2.3264e-06, 2.6822e-07, ..., 1.7844e-06, + -7.0274e-05, -5.2124e-05], + [ 2.9262e-06, 3.7760e-05, 1.5404e-06, ..., 2.4587e-05, + 4.3996e-06, 2.6636e-06], + [ 2.7046e-06, 1.7434e-05, 3.9116e-07, ..., 1.2822e-05, + 3.3267e-06, 4.1239e-06], + ..., + [ 9.0152e-07, -2.2733e-04, -6.1058e-06, ..., -1.5843e-04, + 2.0806e-06, 1.2368e-06], + [ 4.7028e-05, 1.9535e-05, 1.1306e-06, ..., -1.4029e-05, + 4.0919e-05, 1.7986e-05], + [ 4.8392e-06, -8.2329e-06, 1.2033e-06, ..., 6.1505e-06, + 8.1658e-06, 5.1633e-06]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0162, -0.0238, -0.0133, -0.0276, -0.0316, -0.0018, 0.0236, -0.0106, + 0.0332, 0.0010], device='cuda:0'), grad: tensor([-6.9618e-05, 5.4449e-05, 3.6508e-05, 1.8847e-04, 1.3864e-04, + 2.4259e-04, -2.1017e-04, -2.5201e-04, 1.9282e-05, -1.4806e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 265.39, cls_loss 0.0088 cls_loss_mapping 0.0118 cls_loss_causal 0.6303 re_mapping 0.0112 re_causal 0.0290 /// teacc 98.82 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0454, -0.0695, -0.0486, ..., -0.0272, 0.1147, 0.0934], + [-0.1068, -0.1118, -0.1053, ..., -0.0814, -0.1131, -0.0552], + [-0.0651, -0.0491, 0.0839, ..., -0.0796, 0.1083, 0.0113], + ..., + [-0.0615, 0.0435, 0.0429, ..., 0.0919, -0.0701, -0.0828], + [-0.1344, 0.0457, -0.0999, ..., 0.0290, -0.0539, -0.0468], + [ 0.0468, -0.0332, -0.0533, ..., -0.0182, -0.0393, -0.0194]], + device='cuda:0'), grad: tensor([[ 6.9104e-07, 5.0105e-06, 7.2643e-08, ..., 4.8615e-07, + -1.7807e-05, -2.0023e-06], + [ 1.5777e-06, 1.5780e-05, 4.6566e-08, ..., 6.2697e-06, + 7.7069e-05, 3.6716e-05], + [ 2.7381e-07, -3.0875e-05, -2.5500e-06, ..., 1.2666e-07, + -7.1573e-04, -1.3328e-04], + ..., + [ 1.3243e-06, -1.4216e-05, 1.2480e-06, ..., -1.1697e-05, + 4.5359e-05, 2.0206e-05], + [ 7.0632e-06, -1.5676e-05, 9.1456e-07, ..., -5.9269e-06, + 1.0633e-04, 5.4896e-05], + [ 1.8962e-06, 1.1381e-06, 5.5879e-09, ..., 3.6396e-06, + 5.5504e-04, 1.9781e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0154, -0.0237, -0.0133, -0.0278, -0.0318, -0.0013, 0.0233, -0.0107, + 0.0329, 0.0007], device='cuda:0'), grad: tensor([ 4.3720e-05, -2.5153e-04, -1.6556e-03, 1.0830e-04, -8.1491e-04, + 1.3411e-07, 2.8133e-05, 1.2314e-04, 4.8804e-04, 1.9312e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 265.42, cls_loss 0.0109 cls_loss_mapping 0.0140 cls_loss_causal 0.6145 re_mapping 0.0112 re_causal 0.0287 /// teacc 98.82 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0459, -0.0702, -0.0487, ..., -0.0272, 0.1150, 0.0938], + [-0.1074, -0.1129, -0.1055, ..., -0.0818, -0.1158, -0.0564], + [-0.0657, -0.0497, 0.0842, ..., -0.0805, 0.1095, 0.0122], + ..., + [-0.0607, 0.0441, 0.0429, ..., 0.0927, -0.0697, -0.0833], + [-0.1358, 0.0458, -0.0996, ..., 0.0290, -0.0547, -0.0472], + [ 0.0467, -0.0339, -0.0534, ..., -0.0188, -0.0398, -0.0198]], + device='cuda:0'), grad: tensor([[ 4.1336e-05, 9.9778e-05, 7.9349e-07, ..., 1.7548e-04, + 5.2601e-05, -1.4249e-06], + [ 9.6411e-06, 7.4580e-06, 6.6683e-07, ..., -7.4320e-06, + 4.7311e-06, 7.4506e-08], + [ 1.0557e-05, 1.2055e-05, -4.8690e-06, ..., 1.7464e-05, + -2.2352e-08, -2.4308e-06], + ..., + [-2.2173e-05, -1.9646e-04, -5.1856e-06, ..., -2.7084e-04, + -4.8190e-05, 4.2655e-07], + [ 9.5591e-06, 1.0625e-05, 1.7416e-06, ..., 1.0267e-05, + 8.1882e-06, 6.7241e-07], + [ 2.7835e-05, 3.8713e-05, 9.4436e-07, ..., 6.8545e-05, + 2.6867e-05, 3.7439e-07]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0159, -0.0247, -0.0136, -0.0269, -0.0320, -0.0015, 0.0235, -0.0102, + 0.0329, 0.0013], device='cuda:0'), grad: tensor([ 3.6168e-04, -1.7250e-04, 4.5896e-05, 1.0097e-04, -1.5469e-03, + -1.3471e-04, 1.5545e-03, -3.8075e-04, 1.2612e-04, 4.4793e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 266.32, cls_loss 0.0095 cls_loss_mapping 0.0110 cls_loss_causal 0.6441 re_mapping 0.0113 re_causal 0.0292 /// teacc 98.74 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0466, -0.0708, -0.0488, ..., -0.0271, 0.1156, 0.0947], + [-0.1077, -0.1136, -0.1056, ..., -0.0825, -0.1165, -0.0567], + [-0.0661, -0.0502, 0.0847, ..., -0.0812, 0.1100, 0.0123], + ..., + [-0.0605, 0.0429, 0.0429, ..., 0.0931, -0.0701, -0.0836], + [-0.1367, 0.0460, -0.0999, ..., 0.0290, -0.0549, -0.0474], + [ 0.0466, -0.0347, -0.0536, ..., -0.0193, -0.0395, -0.0205]], + device='cuda:0'), grad: tensor([[ 1.5780e-05, 1.6764e-07, 1.7881e-07, ..., 3.6880e-07, + 5.4568e-05, 3.6687e-05], + [ 3.8184e-07, 1.3616e-06, 4.4703e-08, ..., 8.4750e-07, + 6.8732e-07, 3.9674e-07], + [ 1.4212e-06, 1.1362e-07, -1.1493e-06, ..., 2.1048e-07, + 1.5087e-07, 1.2126e-06], + ..., + [ 6.8173e-07, 3.2019e-06, 5.6997e-07, ..., 8.9779e-06, + 2.1942e-06, 8.6427e-07], + [ 8.2515e-07, 1.0803e-07, 5.2154e-08, ..., -2.5257e-06, + 1.3784e-07, 6.0722e-07], + [-2.7940e-08, -1.1019e-05, 1.3039e-08, ..., -1.5512e-05, + 2.0172e-06, 9.3132e-07]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0166, -0.0246, -0.0140, -0.0262, -0.0319, -0.0023, 0.0248, -0.0106, + 0.0330, 0.0013], device='cuda:0'), grad: tensor([ 7.7009e-05, -4.9472e-06, 3.8296e-06, 9.7454e-06, 4.1819e-04, + -1.9386e-05, -1.1283e-04, 3.7003e-04, -7.5921e-06, -7.3385e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 265.42, cls_loss 0.0091 cls_loss_mapping 0.0123 cls_loss_causal 0.5913 re_mapping 0.0118 re_causal 0.0286 /// teacc 98.85 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0472, -0.0711, -0.0490, ..., -0.0273, 0.1162, 0.0953], + [-0.1079, -0.1143, -0.1057, ..., -0.0832, -0.1168, -0.0567], + [-0.0665, -0.0503, 0.0850, ..., -0.0821, 0.1106, 0.0126], + ..., + [-0.0614, 0.0435, 0.0430, ..., 0.0941, -0.0710, -0.0844], + [-0.1380, 0.0459, -0.0999, ..., 0.0291, -0.0547, -0.0472], + [ 0.0472, -0.0353, -0.0537, ..., -0.0200, -0.0394, -0.0211]], + device='cuda:0'), grad: tensor([[-4.3005e-05, -3.0734e-06, 2.6394e-06, ..., -5.6267e-05, + -2.6798e-04, -1.7524e-04], + [ 4.4703e-07, 1.7047e-05, 6.1430e-06, ..., 1.5236e-05, + 1.3821e-06, 7.8417e-07], + [ 3.2037e-07, 2.9624e-05, 1.4827e-05, ..., 3.3021e-05, + -3.6806e-06, -1.4063e-06], + ..., + [ 1.9968e-06, -1.1170e-04, -5.9247e-05, ..., -1.1647e-04, + 5.5358e-06, 3.3248e-06], + [ 4.6790e-06, 1.3642e-05, 8.4564e-06, ..., 4.7199e-06, + 3.3323e-06, 2.7716e-06], + [-2.0191e-06, 1.1310e-05, 4.1910e-06, ..., 1.6078e-05, + 1.7494e-05, 1.1526e-05]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0169, -0.0252, -0.0132, -0.0265, -0.0334, -0.0019, 0.0246, -0.0103, + 0.0328, 0.0023], device='cuda:0'), grad: tensor([-4.9734e-04, 4.5687e-05, 8.1658e-05, 6.0439e-05, 5.4002e-05, + 2.4581e-04, 2.2781e-04, -3.0637e-04, 4.5151e-05, 4.3213e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 265.64, cls_loss 0.0098 cls_loss_mapping 0.0124 cls_loss_causal 0.5771 re_mapping 0.0111 re_causal 0.0267 /// teacc 98.81 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0486, -0.0715, -0.0491, ..., -0.0277, 0.1158, 0.0950], + [-0.1083, -0.1148, -0.1065, ..., -0.0840, -0.1183, -0.0578], + [-0.0669, -0.0495, 0.0861, ..., -0.0826, 0.1122, 0.0142], + ..., + [-0.0620, 0.0438, 0.0433, ..., 0.0950, -0.0719, -0.0849], + [-0.1391, 0.0452, -0.1012, ..., 0.0284, -0.0555, -0.0483], + [ 0.0469, -0.0352, -0.0540, ..., -0.0208, -0.0391, -0.0215]], + device='cuda:0'), grad: tensor([[ 2.6965e-04, 7.6368e-07, 3.3714e-07, ..., 9.7036e-05, + 2.6859e-06, -9.9279e-07], + [ 7.5661e-06, 1.0170e-06, 5.2154e-08, ..., 2.7735e-06, + 1.9222e-06, 2.8498e-07], + [ 1.8954e-05, 3.2596e-07, -7.8790e-07, ..., 8.5682e-06, + 1.1615e-05, -1.4342e-07], + ..., + [ 1.7524e-05, -3.3528e-08, 6.6124e-07, ..., 4.4219e-06, + 1.3541e-06, 2.7567e-07], + [ 1.3143e-05, 2.3469e-07, 4.3027e-07, ..., -1.2927e-06, + -5.0008e-05, 3.8482e-06], + [ 7.0691e-05, 1.4938e-06, 3.2280e-06, ..., 3.2276e-05, + 1.4342e-05, 1.3113e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0180, -0.0260, -0.0112, -0.0273, -0.0326, -0.0009, 0.0247, -0.0109, + 0.0321, 0.0022], device='cuda:0'), grad: tensor([ 6.5327e-04, 2.8223e-05, 1.0657e-04, 6.7568e-04, -1.4436e-04, + -1.6241e-03, 1.8215e-04, 5.8979e-05, -2.7204e-04, 3.3450e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 265.48, cls_loss 0.0098 cls_loss_mapping 0.0125 cls_loss_causal 0.6316 re_mapping 0.0106 re_causal 0.0271 /// teacc 98.77 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0488, -0.0719, -0.0495, ..., -0.0279, 0.1167, 0.0958], + [-0.1086, -0.1146, -0.1073, ..., -0.0844, -0.1191, -0.0580], + [-0.0674, -0.0497, 0.0872, ..., -0.0835, 0.1130, 0.0146], + ..., + [-0.0623, 0.0430, 0.0438, ..., 0.0948, -0.0724, -0.0853], + [-0.1395, 0.0462, -0.1021, ..., 0.0310, -0.0559, -0.0488], + [ 0.0466, -0.0360, -0.0544, ..., -0.0216, -0.0397, -0.0223]], + device='cuda:0'), grad: tensor([[ 4.0568e-06, 3.9823e-06, 1.8626e-07, ..., 3.0138e-06, + -1.2539e-05, -7.3165e-06], + [ 1.0580e-06, 1.5102e-05, 2.3283e-07, ..., 4.4927e-06, + 2.8498e-07, 1.2293e-07], + [ 1.0729e-06, 1.8969e-05, -4.2357e-06, ..., 1.0796e-05, + -1.1683e-05, -1.7211e-06], + ..., + [ 1.0990e-06, -8.6352e-06, 1.7732e-06, ..., -3.4869e-05, + 2.8424e-06, 1.0245e-06], + [ 6.6012e-06, 1.0088e-05, 4.0755e-06, ..., -1.0833e-05, + 1.0289e-05, 1.6801e-06], + [ 7.0453e-05, 2.4125e-05, 2.8070e-06, ..., 8.3745e-05, + 8.3297e-06, 4.0792e-06]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0176, -0.0249, -0.0112, -0.0273, -0.0319, -0.0019, 0.0245, -0.0116, + 0.0334, 0.0013], device='cuda:0'), grad: tensor([ 4.6939e-07, -8.4788e-06, 4.2796e-05, -1.0890e-04, -5.1588e-05, + -8.4460e-05, -1.0140e-05, 2.7239e-05, -2.6673e-06, 1.9562e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 265.76, cls_loss 0.0085 cls_loss_mapping 0.0113 cls_loss_causal 0.6149 re_mapping 0.0111 re_causal 0.0275 /// teacc 98.80 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0488, -0.0725, -0.0497, ..., -0.0281, 0.1181, 0.0973], + [-0.1087, -0.1150, -0.1072, ..., -0.0853, -0.1197, -0.0583], + [-0.0677, -0.0502, 0.0878, ..., -0.0847, 0.1142, 0.0151], + ..., + [-0.0624, 0.0435, 0.0439, ..., 0.0960, -0.0732, -0.0860], + [-0.1401, 0.0460, -0.1024, ..., 0.0309, -0.0569, -0.0497], + [ 0.0460, -0.0365, -0.0549, ..., -0.0227, -0.0408, -0.0237]], + device='cuda:0'), grad: tensor([[ 1.8571e-06, 8.3633e-07, 1.3765e-06, ..., 1.0617e-06, + -9.3460e-05, -3.9965e-05], + [ 1.3001e-06, 2.0619e-06, 1.1706e-04, ..., 2.2221e-06, + 1.6546e-04, 1.7043e-06], + [ 4.7125e-07, -1.9714e-05, -1.6987e-04, ..., -4.4703e-07, + -2.3985e-04, -2.4259e-05], + ..., + [ 1.1511e-06, 1.7643e-05, 3.8564e-05, ..., -5.3160e-06, + 7.8261e-05, 3.0354e-05], + [ 8.0094e-06, -4.9695e-06, 4.9658e-06, ..., -8.1770e-07, + 5.0634e-05, 1.7986e-05], + [ 2.2367e-05, 2.9374e-06, 1.9483e-06, ..., 1.6481e-05, + 1.3284e-05, 5.0925e-06]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0165, -0.0247, -0.0109, -0.0266, -0.0310, -0.0024, 0.0240, -0.0114, + 0.0328, 0.0002], device='cuda:0'), grad: tensor([-9.4533e-05, 3.5334e-04, -5.5456e-04, 8.2910e-05, 3.5524e-04, + -3.9071e-05, -6.2346e-05, 1.9276e-04, 1.1295e-04, -3.4738e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 265.15, cls_loss 0.0085 cls_loss_mapping 0.0108 cls_loss_causal 0.6061 re_mapping 0.0110 re_causal 0.0270 /// teacc 98.88 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0492, -0.0731, -0.0499, ..., -0.0282, 0.1185, 0.0977], + [-0.1096, -0.1152, -0.1076, ..., -0.0858, -0.1201, -0.0585], + [-0.0681, -0.0503, 0.0884, ..., -0.0851, 0.1149, 0.0155], + ..., + [-0.0626, 0.0434, 0.0440, ..., 0.0963, -0.0737, -0.0869], + [-0.1409, 0.0463, -0.1028, ..., 0.0312, -0.0570, -0.0496], + [ 0.0457, -0.0370, -0.0550, ..., -0.0234, -0.0410, -0.0239]], + device='cuda:0'), grad: tensor([[ 7.2271e-07, 5.5358e-06, 2.9802e-08, ..., 1.3467e-06, + 1.4909e-05, 1.3381e-05], + [ 3.5129e-06, 5.9381e-06, 2.6077e-08, ..., 1.4342e-06, + 2.1070e-05, 1.6108e-05], + [ 3.7774e-06, -3.5256e-05, -3.7253e-09, ..., -7.5996e-06, + -1.3602e-04, -1.2189e-04], + ..., + [ 2.7865e-06, 1.4491e-05, -4.2841e-08, ..., 5.7295e-06, + 2.2039e-05, 1.9535e-05], + [ 3.4925e-06, 2.0534e-05, 4.2841e-08, ..., 3.4887e-06, + 5.4091e-05, 4.4137e-05], + [ 8.5309e-06, -3.9972e-06, 3.9116e-08, ..., -1.7464e-05, + 1.3299e-05, 1.3322e-05]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0168, -0.0249, -0.0110, -0.0271, -0.0299, -0.0022, 0.0238, -0.0112, + 0.0331, -0.0004], device='cuda:0'), grad: tensor([ 5.5492e-05, -1.2445e-03, 7.1621e-04, -7.5459e-05, 5.6922e-05, + 1.9145e-04, -1.9744e-05, 1.7917e-04, 1.3220e-04, 7.2531e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 265.56, cls_loss 0.0091 cls_loss_mapping 0.0117 cls_loss_causal 0.5977 re_mapping 0.0106 re_causal 0.0272 /// teacc 98.85 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0496, -0.0735, -0.0500, ..., -0.0283, 0.1191, 0.0984], + [-0.1101, -0.1147, -0.1079, ..., -0.0851, -0.1210, -0.0591], + [-0.0684, -0.0506, 0.0889, ..., -0.0861, 0.1158, 0.0168], + ..., + [-0.0629, 0.0439, 0.0440, ..., 0.0974, -0.0741, -0.0876], + [-0.1416, 0.0469, -0.1029, ..., 0.0323, -0.0576, -0.0503], + [ 0.0454, -0.0393, -0.0553, ..., -0.0261, -0.0413, -0.0243]], + device='cuda:0'), grad: tensor([[ 7.7963e-05, 1.1828e-07, 1.4901e-08, ..., 1.2759e-07, + 8.6784e-05, 4.9740e-05], + [ 1.1586e-06, 2.0452e-06, 3.0734e-08, ..., 1.6801e-06, + 2.0526e-06, 1.4240e-06], + [ 4.0494e-06, 5.8021e-07, -2.1327e-07, ..., 4.1071e-07, + 6.7055e-06, 4.3288e-06], + ..., + [ 7.2177e-07, -1.3873e-05, 2.7008e-08, ..., -1.1660e-05, + 2.4252e-06, 1.4994e-06], + [ 2.0951e-05, 1.5786e-06, 3.0734e-08, ..., 1.8245e-06, + 1.7136e-05, 1.2979e-05], + [ 4.5896e-06, 1.7136e-07, 1.8626e-09, ..., 8.9500e-07, + 5.3719e-06, 1.9092e-06]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0168, -0.0253, -0.0095, -0.0284, -0.0299, -0.0018, 0.0243, -0.0111, + 0.0343, -0.0017], device='cuda:0'), grad: tensor([ 1.3137e-04, -6.1244e-06, 1.7643e-05, 3.4690e-05, 5.3674e-05, + -8.3268e-05, -1.6642e-04, -6.8322e-06, 6.7770e-05, -4.2856e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 265.70, cls_loss 0.0084 cls_loss_mapping 0.0099 cls_loss_causal 0.6133 re_mapping 0.0110 re_causal 0.0282 /// teacc 98.86 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0494, -0.0738, -0.0501, ..., -0.0280, 0.1202, 0.0996], + [-0.1105, -0.1153, -0.1080, ..., -0.0855, -0.1214, -0.0591], + [-0.0689, -0.0514, 0.0890, ..., -0.0864, 0.1167, 0.0178], + ..., + [-0.0631, 0.0434, 0.0441, ..., 0.0981, -0.0746, -0.0880], + [-0.1427, 0.0469, -0.1029, ..., 0.0320, -0.0591, -0.0522], + [ 0.0453, -0.0399, -0.0553, ..., -0.0267, -0.0416, -0.0248]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, 4.8429e-07, 3.7253e-09, ..., 2.9150e-07, + -3.4068e-06, -2.4643e-06], + [ 1.6671e-07, 4.7907e-06, 6.4261e-08, ..., 4.8578e-06, + 1.8068e-07, 1.4435e-07], + [ 2.1234e-06, 7.5549e-06, -2.2631e-07, ..., 1.5246e-06, + 5.5879e-08, -1.0245e-08], + ..., + [ 2.4308e-07, -2.2426e-05, 1.5832e-08, ..., -2.4453e-05, + 3.3248e-07, 2.5332e-07], + [ 4.6473e-07, 4.0978e-06, 8.8476e-08, ..., -1.1213e-06, + 2.1607e-07, 1.7043e-07], + [ 2.2631e-07, 1.3346e-06, 5.5879e-09, ..., 6.2808e-06, + 1.0673e-06, 7.7672e-07]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0162, -0.0252, -0.0095, -0.0280, -0.0301, -0.0014, 0.0240, -0.0113, + 0.0336, -0.0017], device='cuda:0'), grad: tensor([-2.5928e-06, 1.8045e-05, 1.3247e-05, 8.8736e-06, 7.1287e-05, + 1.5488e-06, 2.5127e-06, -2.9966e-05, 1.2636e-05, -9.5785e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 94---------------------------------------------------- +epoch 94, time 282.03, cls_loss 0.0072 cls_loss_mapping 0.0079 cls_loss_causal 0.5720 re_mapping 0.0103 re_causal 0.0265 /// teacc 99.00 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0496, -0.0742, -0.0501, ..., -0.0282, 0.1219, 0.1012], + [-0.1108, -0.1156, -0.1081, ..., -0.0859, -0.1217, -0.0594], + [-0.0694, -0.0513, 0.0893, ..., -0.0867, 0.1171, 0.0181], + ..., + [-0.0635, 0.0433, 0.0441, ..., 0.0986, -0.0752, -0.0888], + [-0.1431, 0.0469, -0.1031, ..., 0.0322, -0.0595, -0.0527], + [ 0.0450, -0.0403, -0.0554, ..., -0.0275, -0.0418, -0.0252]], + device='cuda:0'), grad: tensor([[ 1.5926e-07, 1.6205e-07, 3.1665e-08, ..., 2.8778e-07, + -3.8832e-05, -2.9862e-05], + [ 1.0338e-07, 9.6764e-07, 3.3528e-08, ..., 6.0163e-07, + 1.2387e-06, 8.5402e-07], + [ 8.0094e-08, 6.2399e-07, -9.6858e-08, ..., 1.1735e-07, + 8.6874e-06, 7.2680e-06], + ..., + [ 4.2468e-07, 1.6671e-07, 1.3411e-07, ..., 1.3663e-06, + 1.4082e-06, 9.7696e-07], + [ 1.9576e-06, 1.7686e-06, 1.8626e-08, ..., 3.2969e-06, + 3.0361e-06, 1.8906e-06], + [ 1.9185e-07, -2.1048e-06, 1.7136e-07, ..., -5.7407e-06, + 1.0259e-05, 7.6368e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0154, -0.0251, -0.0091, -0.0283, -0.0293, -0.0012, 0.0237, -0.0114, + 0.0335, -0.0025], device='cuda:0'), grad: tensor([-5.6624e-05, -1.7118e-06, 1.3165e-05, 4.7348e-06, 1.0282e-05, + -1.2554e-06, 1.0282e-05, 1.4961e-05, 1.2122e-05, -5.9418e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 265.58, cls_loss 0.0086 cls_loss_mapping 0.0103 cls_loss_causal 0.6327 re_mapping 0.0105 re_causal 0.0261 /// teacc 98.90 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0501, -0.0750, -0.0505, ..., -0.0287, 0.1227, 0.1020], + [-0.1111, -0.1162, -0.1083, ..., -0.0871, -0.1226, -0.0593], + [-0.0700, -0.0517, 0.0893, ..., -0.0878, 0.1176, 0.0181], + ..., + [-0.0635, 0.0441, 0.0442, ..., 0.1004, -0.0755, -0.0892], + [-0.1443, 0.0468, -0.1029, ..., 0.0330, -0.0599, -0.0537], + [ 0.0449, -0.0407, -0.0554, ..., -0.0283, -0.0420, -0.0258]], + device='cuda:0'), grad: tensor([[ 1.5218e-06, 2.1495e-06, 1.6978e-06, ..., 3.3136e-06, + -4.5113e-06, -2.0899e-06], + [ 1.7416e-07, 8.8289e-06, 9.1493e-06, ..., 1.2107e-05, + 5.2061e-07, 1.3029e-06], + [ 2.6263e-06, 8.1968e-04, 8.7786e-04, ..., 1.0061e-03, + 6.9626e-06, 1.0824e-04], + ..., + [ 2.6915e-07, -8.4925e-04, -8.9741e-04, ..., -1.0357e-03, + 2.0526e-06, -1.0645e-04], + [ 3.1888e-06, 3.4049e-06, 1.7472e-06, ..., -8.0094e-07, + 1.1791e-06, 8.0559e-07], + [ 1.0552e-06, 7.0818e-06, 9.4809e-07, ..., -1.0364e-05, + 1.4296e-06, 8.6520e-07]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0152, -0.0254, -0.0094, -0.0294, -0.0295, -0.0010, 0.0240, -0.0105, + 0.0337, -0.0028], device='cuda:0'), grad: tensor([ 7.8753e-06, 5.0068e-05, 2.7237e-03, 1.0423e-05, 3.3069e-04, + 1.7166e-05, -1.3359e-05, -2.7008e-03, -3.0212e-06, -4.2319e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 265.23, cls_loss 0.0089 cls_loss_mapping 0.0115 cls_loss_causal 0.6186 re_mapping 0.0100 re_causal 0.0260 /// teacc 98.79 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0525, -0.0755, -0.0506, ..., -0.0298, 0.1225, 0.1022], + [-0.1115, -0.1168, -0.1084, ..., -0.0869, -0.1237, -0.0597], + [-0.0708, -0.0526, 0.0888, ..., -0.0894, 0.1182, 0.0184], + ..., + [-0.0638, 0.0448, 0.0449, ..., 0.1013, -0.0756, -0.0896], + [-0.1452, 0.0470, -0.1031, ..., 0.0331, -0.0606, -0.0540], + [ 0.0447, -0.0410, -0.0555, ..., -0.0287, -0.0421, -0.0262]], + device='cuda:0'), grad: tensor([[ 1.2470e-06, 2.0918e-06, 1.1176e-08, ..., 2.6617e-06, + -1.0318e-04, -9.8050e-05], + [ 1.6093e-06, 4.8429e-06, 2.0489e-08, ..., 6.3591e-06, + 5.4296e-07, 4.3865e-07], + [ 4.8988e-06, 1.1027e-05, -2.1327e-07, ..., 5.0403e-06, + 3.9637e-06, 4.2394e-06], + ..., + [ 2.7381e-06, -1.7002e-05, 2.7008e-08, ..., -1.5423e-05, + 7.9814e-07, 6.1281e-07], + [ 6.9141e-06, 8.8662e-06, 6.3330e-08, ..., -1.1522e-04, + 3.2224e-06, 2.8871e-06], + [ 3.5986e-06, 1.1630e-05, 1.8626e-09, ..., 6.8069e-05, + 6.0499e-05, 5.7846e-05]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0159, -0.0247, -0.0100, -0.0281, -0.0299, -0.0019, 0.0244, -0.0100, + 0.0331, -0.0027], device='cuda:0'), grad: tensor([-1.5819e-04, 4.8697e-05, 3.8236e-05, -1.2279e-05, 3.7819e-05, + 2.2805e-04, 5.6863e-05, -8.7172e-06, -1.0262e-03, 7.9536e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 265.38, cls_loss 0.0099 cls_loss_mapping 0.0105 cls_loss_causal 0.6036 re_mapping 0.0098 re_causal 0.0251 /// teacc 98.90 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0533, -0.0761, -0.0510, ..., -0.0299, 0.1233, 0.1032], + [-0.1120, -0.1178, -0.1090, ..., -0.0881, -0.1245, -0.0600], + [-0.0713, -0.0526, 0.0895, ..., -0.0899, 0.1183, 0.0188], + ..., + [-0.0642, 0.0450, 0.0447, ..., 0.1021, -0.0760, -0.0909], + [-0.1461, 0.0471, -0.1033, ..., 0.0334, -0.0611, -0.0545], + [ 0.0447, -0.0412, -0.0555, ..., -0.0293, -0.0425, -0.0268]], + device='cuda:0'), grad: tensor([[-5.3868e-06, 8.5775e-07, 4.5635e-08, ..., -1.2553e-04, + -2.0456e-04, -1.8442e-04], + [ 2.5984e-07, 3.5286e-05, 2.3842e-07, ..., -2.7325e-06, + 3.7346e-07, 3.0827e-07], + [ 5.0198e-07, 2.0280e-05, 1.4501e-06, ..., 4.9204e-05, + 1.7527e-06, 1.5656e-06], + ..., + [ 8.8196e-07, 4.6659e-04, 5.9418e-07, ..., 1.8759e-03, + 5.0999e-06, 4.5635e-06], + [ 1.5702e-06, -7.4096e-06, -1.4920e-06, ..., 6.8285e-06, + 2.7027e-06, 2.6692e-06], + [-3.4962e-06, -5.2357e-04, 8.2888e-07, ..., -1.9798e-03, + 9.9652e-07, 8.5030e-07]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0159, -0.0252, -0.0105, -0.0281, -0.0303, -0.0024, 0.0251, -0.0101, + 0.0332, -0.0016], device='cuda:0'), grad: tensor([-2.5034e-04, -1.2245e-03, 2.1756e-04, 4.2289e-05, 2.2674e-04, + 9.1195e-05, 2.4605e-04, 9.1858e-03, 1.5900e-05, -8.5449e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 265.61, cls_loss 0.0072 cls_loss_mapping 0.0104 cls_loss_causal 0.6021 re_mapping 0.0102 re_causal 0.0270 /// teacc 98.89 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0535, -0.0764, -0.0513, ..., -0.0300, 0.1240, 0.1040], + [-0.1123, -0.1194, -0.1091, ..., -0.0890, -0.1247, -0.0603], + [-0.0717, -0.0531, 0.0898, ..., -0.0906, 0.1188, 0.0195], + ..., + [-0.0645, 0.0454, 0.0447, ..., 0.1026, -0.0764, -0.0913], + [-0.1468, 0.0471, -0.1035, ..., 0.0336, -0.0614, -0.0550], + [ 0.0449, -0.0415, -0.0556, ..., -0.0293, -0.0426, -0.0272]], + device='cuda:0'), grad: tensor([[ 1.6857e-07, 4.6007e-07, 3.9116e-08, ..., 1.3225e-06, + -3.2306e-05, -3.4183e-05], + [ 2.6450e-07, 1.5339e-06, 1.0245e-08, ..., 1.5914e-05, + 1.0155e-05, 1.2092e-05], + [ 5.8673e-08, 1.2293e-06, 3.9116e-08, ..., 2.5108e-06, + 1.9744e-07, 6.0908e-07], + ..., + [ 1.8440e-07, -7.5579e-05, 1.6764e-08, ..., -3.3706e-05, + 1.4585e-06, 1.5106e-06], + [ 7.4320e-07, -2.0526e-06, 9.3132e-09, ..., -1.2088e-04, + 2.8498e-06, 2.9802e-06], + [ 1.0803e-07, 4.0680e-05, 4.0047e-08, ..., 2.2173e-05, + 8.4192e-06, 7.6964e-06]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0157, -0.0268, -0.0097, -0.0279, -0.0298, -0.0025, 0.0246, -0.0101, + 0.0332, -0.0011], device='cuda:0'), grad: tensor([-5.0902e-05, 5.0515e-05, 7.7263e-06, 5.0366e-05, -5.3644e-06, + 1.9193e-04, 4.2409e-05, -1.0055e-04, -2.7609e-04, 8.9943e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 266.02, cls_loss 0.0080 cls_loss_mapping 0.0099 cls_loss_causal 0.6017 re_mapping 0.0103 re_causal 0.0263 /// teacc 98.98 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0542, -0.0770, -0.0514, ..., -0.0304, 0.1246, 0.1045], + [-0.1133, -0.1198, -0.1091, ..., -0.0893, -0.1250, -0.0597], + [-0.0723, -0.0538, 0.0900, ..., -0.0921, 0.1196, 0.0199], + ..., + [-0.0647, 0.0458, 0.0447, ..., 0.1033, -0.0773, -0.0927], + [-0.1478, 0.0482, -0.1047, ..., 0.0340, -0.0618, -0.0553], + [ 0.0445, -0.0433, -0.0556, ..., -0.0309, -0.0429, -0.0279]], + device='cuda:0'), grad: tensor([[ 8.4937e-07, -5.6252e-07, 6.1840e-07, ..., 3.1609e-06, + -1.6540e-05, -2.0027e-05], + [ 2.7530e-06, 1.2748e-05, 4.3288e-06, ..., -6.2305e-07, + 1.0356e-05, 1.1520e-06], + [ 2.0098e-06, 1.3679e-05, -3.2604e-05, ..., 8.4266e-06, + -7.1883e-05, -1.9465e-06], + ..., + [ 1.5303e-05, -3.5286e-05, 1.3690e-06, ..., -7.1406e-05, + 3.7327e-06, 8.7451e-07], + [ 1.4536e-05, 6.2704e-05, 6.4727e-07, ..., 2.8223e-05, + 7.8231e-06, 7.4059e-06], + [ 4.7117e-05, 1.6525e-05, 7.5437e-08, ..., 7.1049e-05, + 1.1241e-06, 1.0626e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0158, -0.0261, -0.0104, -0.0279, -0.0296, -0.0016, 0.0243, -0.0101, + 0.0338, -0.0022], device='cuda:0'), grad: tensor([-2.3559e-05, -1.2606e-05, -9.7156e-05, -1.4007e-04, 2.2292e-05, + -2.4945e-05, 9.8109e-05, -4.7415e-05, 1.0163e-04, 1.2350e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 265.60, cls_loss 0.0083 cls_loss_mapping 0.0109 cls_loss_causal 0.5971 re_mapping 0.0099 re_causal 0.0255 /// teacc 98.88 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0550, -0.0775, -0.0516, ..., -0.0307, 0.1254, 0.1061], + [-0.1137, -0.1191, -0.1093, ..., -0.0877, -0.1258, -0.0605], + [-0.0729, -0.0539, 0.0902, ..., -0.0938, 0.1205, 0.0205], + ..., + [-0.0648, 0.0459, 0.0450, ..., 0.1038, -0.0781, -0.0933], + [-0.1486, 0.0483, -0.1048, ..., 0.0340, -0.0623, -0.0560], + [ 0.0441, -0.0440, -0.0557, ..., -0.0321, -0.0413, -0.0286]], + device='cuda:0'), grad: tensor([[-2.2985e-06, -3.7253e-09, 5.1558e-06, ..., -7.5530e-07, + -2.4840e-05, -3.7909e-05], + [ 4.8522e-07, 1.2182e-06, 1.1576e-06, ..., -6.9477e-06, + 4.3996e-06, 1.7509e-06], + [ 1.1111e-06, -2.2724e-07, -4.6760e-05, ..., 1.2219e-06, + -1.2887e-04, -1.8582e-05], + ..., + [ 7.6257e-06, 2.3171e-05, 2.1141e-07, ..., -2.2855e-06, + 1.5562e-06, 9.7696e-07], + [ 6.2659e-06, 4.1164e-07, 1.4678e-05, ..., 9.3505e-06, + 4.9949e-05, 1.5870e-05], + [ 6.6049e-06, 2.3201e-05, 6.7204e-06, ..., 1.1036e-06, + 2.4229e-05, 8.5458e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0162, -0.0245, -0.0107, -0.0282, -0.0310, -0.0015, 0.0244, -0.0105, + 0.0336, -0.0016], device='cuda:0'), grad: tensor([-1.6749e-05, -1.0502e-04, -2.5845e-04, -5.7429e-05, 1.0741e-04, + -6.0648e-06, 3.3826e-05, 4.2826e-05, 1.7893e-04, 8.0407e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 265.63, cls_loss 0.0060 cls_loss_mapping 0.0082 cls_loss_causal 0.5821 re_mapping 0.0100 re_causal 0.0261 /// teacc 98.89 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0573, -0.0789, -0.0517, ..., -0.0308, 0.1258, 0.1065], + [-0.1141, -0.1193, -0.1096, ..., -0.0882, -0.1262, -0.0609], + [-0.0733, -0.0538, 0.0906, ..., -0.0940, 0.1212, 0.0210], + ..., + [-0.0652, 0.0461, 0.0449, ..., 0.1043, -0.0787, -0.0942], + [-0.1495, 0.0481, -0.1049, ..., 0.0342, -0.0630, -0.0563], + [ 0.0439, -0.0446, -0.0558, ..., -0.0327, -0.0416, -0.0293]], + device='cuda:0'), grad: tensor([[ 5.4017e-07, 1.5367e-07, 7.0781e-08, ..., 1.0477e-06, + -2.6785e-06, -3.8259e-06], + [ 2.2631e-07, 2.5835e-06, 2.1234e-07, ..., 2.2501e-06, + 2.6692e-06, 1.3821e-06], + [ 6.9384e-07, 1.9372e-05, 1.0058e-06, ..., -1.1539e-04, + -2.2376e-04, -1.2422e-04], + ..., + [ 3.6415e-07, -2.7940e-05, -2.1569e-06, ..., -1.1802e-05, + 2.2203e-05, 1.1474e-05], + [ 2.8554e-06, 1.1884e-05, 5.0198e-07, ..., 2.8443e-06, + 5.4240e-06, 3.1348e-06], + [ 1.6578e-07, 2.6915e-07, -2.7101e-07, ..., 1.5181e-06, + 2.2929e-06, 1.4352e-06]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0163, -0.0246, -0.0105, -0.0284, -0.0312, -0.0013, 0.0250, -0.0106, + 0.0335, -0.0017], device='cuda:0'), grad: tensor([ 6.1207e-06, 8.7246e-06, -6.6996e-04, 8.4996e-05, 1.2517e-05, + 5.2834e-04, 3.9041e-06, 3.6918e-06, 1.8179e-05, 3.8259e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 265.92, cls_loss 0.0068 cls_loss_mapping 0.0086 cls_loss_causal 0.5737 re_mapping 0.0095 re_causal 0.0248 /// teacc 98.92 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0577, -0.0795, -0.0518, ..., -0.0311, 0.1265, 0.1073], + [-0.1144, -0.1198, -0.1097, ..., -0.0897, -0.1266, -0.0609], + [-0.0738, -0.0544, 0.0909, ..., -0.0950, 0.1220, 0.0215], + ..., + [-0.0654, 0.0455, 0.0449, ..., 0.1051, -0.0792, -0.0947], + [-0.1502, 0.0480, -0.1051, ..., 0.0345, -0.0634, -0.0565], + [ 0.0443, -0.0449, -0.0558, ..., -0.0332, -0.0418, -0.0302]], + device='cuda:0'), grad: tensor([[ 2.5332e-07, 5.6904e-07, 1.9558e-08, ..., 3.5856e-07, + -3.9995e-05, -1.9863e-05], + [ 1.4156e-07, 1.3523e-06, 1.2107e-08, ..., 9.6299e-07, + 4.2468e-07, 1.4715e-07], + [ 3.6135e-07, 2.2668e-06, -1.8254e-07, ..., 1.5171e-06, + 5.4613e-06, 3.4906e-06], + ..., + [ 1.6112e-07, -2.7604e-06, 2.3283e-08, ..., -3.3863e-06, + 1.6792e-06, 6.3889e-07], + [-5.7966e-06, -1.5289e-05, 2.6077e-08, ..., -6.0722e-06, + 1.3664e-05, 6.5900e-06], + [-2.2352e-08, 5.4576e-07, 1.8626e-09, ..., 1.4780e-06, + 7.4171e-06, 3.6582e-06]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0161, -0.0252, -0.0106, -0.0276, -0.0303, -0.0016, 0.0247, -0.0106, + 0.0336, -0.0020], device='cuda:0'), grad: tensor([-4.0919e-05, -4.3303e-05, 2.3678e-05, 6.4254e-05, 1.9416e-05, + 2.5332e-05, 6.7800e-06, 3.2187e-05, -7.4685e-05, -1.2711e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 265.39, cls_loss 0.0101 cls_loss_mapping 0.0119 cls_loss_causal 0.6117 re_mapping 0.0095 re_causal 0.0239 /// teacc 98.91 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0586, -0.0814, -0.0519, ..., -0.0314, 0.1266, 0.1086], + [-0.1151, -0.1207, -0.1097, ..., -0.0908, -0.1273, -0.0618], + [-0.0750, -0.0551, 0.0909, ..., -0.0964, 0.1215, 0.0209], + ..., + [-0.0662, 0.0461, 0.0450, ..., 0.1058, -0.0796, -0.0952], + [-0.1512, 0.0483, -0.1052, ..., 0.0351, -0.0641, -0.0567], + [ 0.0440, -0.0459, -0.0559, ..., -0.0336, -0.0407, -0.0307]], + device='cuda:0'), grad: tensor([[ 7.6275e-07, 1.6456e-06, 2.7940e-09, ..., 1.0682e-06, + -1.0592e-04, -2.7940e-05], + [ 4.1723e-07, 7.0222e-06, 9.3132e-10, ..., 2.1588e-06, + 3.8520e-06, 4.1537e-07], + [ 1.1912e-06, 2.6017e-05, -2.1979e-07, ..., 8.6874e-06, + -4.8369e-05, 3.5763e-07], + ..., + [ 2.3469e-07, 8.6501e-06, 1.4249e-07, ..., -2.6319e-06, + 4.3154e-05, 2.2054e-06], + [ 3.1084e-05, 1.6451e-04, 2.7008e-08, ..., 7.6652e-05, + 9.9018e-06, 2.7083e-06], + [ 7.2457e-07, 3.5092e-06, 9.3132e-10, ..., -5.2378e-06, + 1.2673e-05, 2.9001e-06]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0171, -0.0269, -0.0116, -0.0274, -0.0300, -0.0010, 0.0249, -0.0103, + 0.0349, -0.0014], device='cuda:0'), grad: tensor([-1.7977e-04, -2.6250e-04, -1.4296e-06, -5.6362e-04, 7.5042e-05, + -2.3097e-05, 1.0091e-04, 2.8992e-04, 5.7125e-04, -7.1265e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 265.16, cls_loss 0.0064 cls_loss_mapping 0.0078 cls_loss_causal 0.5711 re_mapping 0.0104 re_causal 0.0260 /// teacc 98.88 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0590, -0.0820, -0.0520, ..., -0.0315, 0.1271, 0.1091], + [-0.1153, -0.1216, -0.1098, ..., -0.0918, -0.1277, -0.0619], + [-0.0754, -0.0561, 0.0913, ..., -0.0981, 0.1222, 0.0211], + ..., + [-0.0665, 0.0467, 0.0449, ..., 0.1070, -0.0803, -0.0956], + [-0.1518, 0.0484, -0.1054, ..., 0.0351, -0.0641, -0.0570], + [ 0.0449, -0.0462, -0.0559, ..., -0.0339, -0.0410, -0.0311]], + device='cuda:0'), grad: tensor([[ 3.9581e-07, 1.3039e-07, 1.9558e-08, ..., 3.7439e-07, + -2.0012e-05, -1.2428e-05], + [ 2.3749e-07, 7.5623e-07, 2.7940e-09, ..., 3.0268e-07, + 1.6112e-07, 7.8231e-08], + [ 8.6613e-07, 6.8806e-06, -1.5832e-08, ..., 6.7055e-07, + 2.3860e-06, 1.4622e-06], + ..., + [ 1.5441e-06, -3.9823e-06, 1.8626e-09, ..., -1.3657e-05, + 3.0920e-07, 1.4342e-07], + [ 3.2876e-06, 7.9442e-07, 7.4506e-09, ..., 4.9025e-06, + 1.0636e-06, 6.6124e-07], + [ 8.3726e-07, 7.1004e-06, 0.0000e+00, ..., 9.5740e-06, + 1.2293e-05, 7.6517e-06]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0171, -0.0271, -0.0120, -0.0281, -0.0297, -0.0007, 0.0247, -0.0097, + 0.0354, -0.0016], device='cuda:0'), grad: tensor([ 2.1350e-04, -1.7080e-03, 1.9044e-05, -4.8131e-06, 1.4529e-05, + 8.4341e-05, 9.0063e-05, 1.2308e-05, 1.1902e-03, 8.9705e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 265.16, cls_loss 0.0077 cls_loss_mapping 0.0103 cls_loss_causal 0.5827 re_mapping 0.0096 re_causal 0.0241 /// teacc 98.84 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0600, -0.0825, -0.0523, ..., -0.0318, 0.1272, 0.1092], + [-0.1156, -0.1219, -0.1098, ..., -0.0921, -0.1278, -0.0619], + [-0.0760, -0.0564, 0.0916, ..., -0.0993, 0.1232, 0.0215], + ..., + [-0.0671, 0.0471, 0.0448, ..., 0.1076, -0.0817, -0.0966], + [-0.1528, 0.0488, -0.1053, ..., 0.0364, -0.0646, -0.0574], + [ 0.0450, -0.0470, -0.0559, ..., -0.0348, -0.0410, -0.0318]], + device='cuda:0'), grad: tensor([[ 2.6464e-04, 1.4715e-07, 3.9116e-08, ..., 5.8937e-04, + 1.7679e-04, -9.9361e-05], + [ 2.6934e-06, 4.7777e-07, 3.7253e-09, ..., 6.2063e-06, + 6.6236e-06, 2.3674e-06], + [ 8.1118e-07, 2.6822e-07, 4.6566e-09, ..., 1.7751e-06, + 1.0185e-05, 4.3660e-06], + ..., + [ 1.2785e-05, -2.9802e-07, 3.7253e-09, ..., 2.6882e-05, + 1.9655e-05, 3.2634e-06], + [ 1.1700e-04, -7.4580e-06, 2.7940e-09, ..., 2.3901e-04, + 1.4186e-04, 9.0227e-06], + [ 8.3223e-06, 2.4401e-07, 1.8626e-09, ..., 1.8165e-05, + 3.7760e-05, 1.9625e-05]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0176, -0.0262, -0.0124, -0.0285, -0.0313, -0.0006, 0.0250, -0.0099, + 0.0355, -0.0009], device='cuda:0'), grad: tensor([ 7.2908e-04, 1.6987e-05, 7.5340e-05, 5.2929e-05, 3.5912e-05, + -5.5542e-03, 4.1580e-03, 6.1452e-05, 3.1543e-04, 1.0973e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 265.26, cls_loss 0.0077 cls_loss_mapping 0.0085 cls_loss_causal 0.5726 re_mapping 0.0093 re_causal 0.0231 /// teacc 98.94 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0610, -0.0831, -0.0529, ..., -0.0327, 0.1274, 0.1096], + [-0.1159, -0.1221, -0.1102, ..., -0.0924, -0.1283, -0.0623], + [-0.0763, -0.0568, 0.0927, ..., -0.1001, 0.1240, 0.0222], + ..., + [-0.0672, 0.0473, 0.0445, ..., 0.1085, -0.0824, -0.0975], + [-0.1539, 0.0486, -0.1059, ..., 0.0366, -0.0645, -0.0578], + [ 0.0447, -0.0474, -0.0562, ..., -0.0355, -0.0413, -0.0324]], + device='cuda:0'), grad: tensor([[ 1.3132e-07, 1.2973e-06, 8.0187e-07, ..., 3.5986e-06, + -2.5225e-04, -1.5152e-04], + [ 1.1455e-07, 2.5071e-06, 1.8161e-07, ..., 3.6180e-05, + 1.1092e-06, 4.9174e-07], + [ 1.0803e-07, 1.5004e-06, -2.6077e-06, ..., 3.8743e-06, + -4.8839e-06, -1.6745e-06], + ..., + [ 4.0047e-07, 3.7234e-06, 6.9011e-07, ..., -3.3051e-05, + 1.8273e-06, 7.4506e-07], + [ 4.0606e-07, -5.6297e-05, 1.0524e-07, ..., -1.7166e-04, + 9.0338e-07, 6.2212e-07], + [ 3.0454e-07, 5.3421e-06, 9.7509e-07, ..., 1.8403e-05, + 5.9940e-06, 1.6429e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0179, -0.0274, -0.0103, -0.0288, -0.0308, -0.0007, 0.0254, -0.0100, + 0.0353, -0.0011], device='cuda:0'), grad: tensor([-2.5392e-04, 7.6592e-05, -3.5092e-06, 8.2374e-05, 6.1877e-06, + 2.1541e-04, 2.7847e-04, -4.6939e-05, -4.0960e-04, 5.5015e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 265.03, cls_loss 0.0086 cls_loss_mapping 0.0097 cls_loss_causal 0.5773 re_mapping 0.0094 re_causal 0.0231 /// teacc 98.88 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0614, -0.0835, -0.0529, ..., -0.0331, 0.1274, 0.1105], + [-0.1171, -0.1224, -0.1090, ..., -0.0928, -0.1306, -0.0646], + [-0.0769, -0.0571, 0.0921, ..., -0.1008, 0.1250, 0.0238], + ..., + [-0.0677, 0.0475, 0.0446, ..., 0.1094, -0.0826, -0.0982], + [-0.1549, 0.0487, -0.1058, ..., 0.0365, -0.0652, -0.0583], + [ 0.0443, -0.0479, -0.0569, ..., -0.0360, -0.0405, -0.0332]], + device='cuda:0'), grad: tensor([[ 1.5404e-06, 2.0433e-06, 1.6885e-06, ..., 8.5682e-07, + 2.5462e-06, 8.9221e-07], + [ 5.7276e-07, 6.6645e-06, 2.4680e-07, ..., 2.5537e-06, + 4.0513e-07, 4.5542e-07], + [ 1.5395e-06, 9.8571e-06, -1.0476e-05, ..., 1.1856e-06, + -3.0566e-06, 1.9055e-06], + ..., + [ 2.7549e-06, 2.7046e-05, 4.0829e-06, ..., -1.2428e-05, + 4.9695e-06, 4.7535e-06], + [ 5.9754e-06, 1.4514e-05, 1.1874e-06, ..., 3.7532e-06, + 1.6624e-06, 3.6173e-06], + [ 4.1015e-06, -3.1233e-05, 4.7870e-07, ..., 6.6273e-06, + 5.6345e-07, 1.1222e-06]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0186, -0.0264, -0.0111, -0.0286, -0.0315, -0.0007, 0.0255, -0.0101, + 0.0351, -0.0006], device='cuda:0'), grad: tensor([ 2.0966e-05, 2.5302e-05, -5.3793e-05, 1.0163e-05, -1.2428e-05, + -6.8724e-05, -2.8983e-06, 2.2137e-04, 4.5180e-05, -1.8489e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 265.12, cls_loss 0.0079 cls_loss_mapping 0.0085 cls_loss_causal 0.6028 re_mapping 0.0098 re_causal 0.0233 /// teacc 98.91 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0617, -0.0841, -0.0530, ..., -0.0334, 0.1276, 0.1103], + [-0.1174, -0.1228, -0.1097, ..., -0.0930, -0.1331, -0.0650], + [-0.0774, -0.0569, 0.0926, ..., -0.1011, 0.1262, 0.0244], + ..., + [-0.0679, 0.0475, 0.0446, ..., 0.1102, -0.0837, -0.0994], + [-0.1565, 0.0484, -0.1059, ..., 0.0364, -0.0662, -0.0592], + [ 0.0444, -0.0486, -0.0570, ..., -0.0374, -0.0410, -0.0342]], + device='cuda:0'), grad: tensor([[ 5.7742e-08, 2.6710e-06, 1.8626e-09, ..., 6.9253e-06, + 2.3004e-06, 1.0189e-06], + [ 1.2107e-07, 7.3552e-05, 0.0000e+00, ..., 5.9038e-05, + 7.1805e-07, 4.2468e-07], + [ 1.1642e-07, 1.5509e-04, -1.3039e-08, ..., 1.3173e-04, + 7.8883e-07, 8.9686e-07], + ..., + [ 2.5518e-07, -7.0667e-04, 5.5879e-09, ..., -5.7697e-04, + 2.6748e-06, 1.5590e-06], + [ 2.6077e-07, 2.3782e-05, 9.3132e-10, ..., 5.2191e-06, + -2.5898e-05, -1.5587e-05], + [ 3.8184e-08, 3.8836e-07, 0.0000e+00, ..., 3.6675e-06, + 3.3733e-06, 2.0191e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0189, -0.0275, -0.0108, -0.0277, -0.0305, -0.0011, 0.0277, -0.0105, + 0.0349, -0.0015], device='cuda:0'), grad: tensor([ 3.2544e-05, 1.5080e-04, 3.4738e-04, 1.0128e-03, 7.0184e-06, + 4.4912e-05, 1.6049e-05, -1.5764e-03, -1.8761e-05, -1.6958e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 265.55, cls_loss 0.0074 cls_loss_mapping 0.0086 cls_loss_causal 0.5927 re_mapping 0.0088 re_causal 0.0226 /// teacc 98.87 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0624, -0.0849, -0.0532, ..., -0.0336, 0.1290, 0.1112], + [-0.1179, -0.1232, -0.1101, ..., -0.0932, -0.1332, -0.0649], + [-0.0778, -0.0570, 0.0933, ..., -0.1020, 0.1270, 0.0246], + ..., + [-0.0676, 0.0479, 0.0445, ..., 0.1111, -0.0848, -0.0999], + [-0.1575, 0.0484, -0.1062, ..., 0.0362, -0.0661, -0.0596], + [ 0.0439, -0.0494, -0.0571, ..., -0.0375, -0.0417, -0.0351]], + device='cuda:0'), grad: tensor([[ 2.9746e-06, 4.0419e-07, 4.6566e-09, ..., 1.7490e-06, + -3.9369e-05, -2.3171e-05], + [ 1.4454e-06, 1.0980e-06, 0.0000e+00, ..., 1.6307e-06, + 7.4320e-07, 4.0885e-07], + [ 7.6182e-07, 5.8077e-06, -8.3819e-09, ..., 7.1563e-06, + 1.9539e-06, 1.0813e-06], + ..., + [ 1.0338e-06, -7.2643e-06, 1.8626e-09, ..., -1.0528e-05, + 5.6159e-07, 2.9989e-07], + [ 2.1160e-05, 1.6112e-06, 9.3132e-10, ..., -2.3559e-05, + 9.2760e-06, 7.8455e-06], + [ 1.4370e-06, 1.2405e-06, 9.3132e-10, ..., 3.2395e-05, + 1.7002e-05, 7.7784e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0185, -0.0271, -0.0104, -0.0279, -0.0309, -0.0016, 0.0281, -0.0110, + 0.0349, -0.0016], device='cuda:0'), grad: tensor([-6.1333e-05, 6.8918e-06, 2.6241e-05, 1.4019e-04, 2.2382e-05, + -2.4509e-04, 7.3671e-05, -1.1474e-05, -2.0063e-04, 2.4843e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 265.84, cls_loss 0.0061 cls_loss_mapping 0.0077 cls_loss_causal 0.5713 re_mapping 0.0092 re_causal 0.0233 /// teacc 98.86 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0625, -0.0849, -0.0534, ..., -0.0336, 0.1296, 0.1119], + [-0.1183, -0.1230, -0.1089, ..., -0.0945, -0.1332, -0.0641], + [-0.0783, -0.0573, 0.0938, ..., -0.1026, 0.1274, 0.0245], + ..., + [-0.0677, 0.0481, 0.0438, ..., 0.1123, -0.0856, -0.1006], + [-0.1591, 0.0481, -0.1063, ..., 0.0358, -0.0667, -0.0603], + [ 0.0433, -0.0497, -0.0578, ..., -0.0381, -0.0421, -0.0357]], + device='cuda:0'), grad: tensor([[ 2.0321e-06, 3.8184e-08, 9.3319e-07, ..., -3.0827e-07, + 3.1646e-06, 1.7351e-06], + [ 6.2305e-07, 5.4110e-07, 6.6590e-07, ..., 5.1316e-07, + 1.4780e-06, 7.9069e-07], + [ 3.3993e-07, -2.6245e-06, -7.9051e-06, ..., 9.4064e-08, + -7.8306e-06, -1.4501e-06], + ..., + [ 2.2724e-07, -2.0415e-06, 6.4857e-06, ..., -3.6452e-06, + 6.2138e-06, 1.4333e-06], + [ 4.6901e-06, 4.4890e-07, 3.9767e-07, ..., 1.5721e-06, + 3.8520e-06, 3.1963e-06], + [-9.0152e-07, 6.0536e-08, 1.6019e-07, ..., -3.8743e-07, + 4.9733e-07, 1.9744e-07]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0182, -0.0257, -0.0114, -0.0279, -0.0313, -0.0010, 0.0276, -0.0111, + 0.0343, -0.0018], device='cuda:0'), grad: tensor([ 7.6145e-06, 7.6517e-06, -1.2375e-05, 1.4804e-05, -1.0557e-05, + 2.8074e-05, -5.9903e-05, 1.0207e-05, 4.2945e-05, -2.8446e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 265.53, cls_loss 0.0070 cls_loss_mapping 0.0071 cls_loss_causal 0.5595 re_mapping 0.0086 re_causal 0.0217 /// teacc 98.93 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0633, -0.0850, -0.0537, ..., -0.0345, 0.1311, 0.1135], + [-0.1193, -0.1228, -0.1095, ..., -0.0948, -0.1340, -0.0642], + [-0.0789, -0.0574, 0.0951, ..., -0.1029, 0.1280, 0.0249], + ..., + [-0.0680, 0.0483, 0.0432, ..., 0.1128, -0.0868, -0.1022], + [-0.1610, 0.0478, -0.1065, ..., 0.0352, -0.0675, -0.0607], + [ 0.0433, -0.0501, -0.0579, ..., -0.0390, -0.0424, -0.0365]], + device='cuda:0'), grad: tensor([[ 3.7178e-06, 1.1027e-06, 4.6566e-09, ..., 2.6356e-07, + -5.4091e-06, -3.9488e-06], + [ 3.9339e-06, 8.5473e-05, 3.7439e-07, ..., 9.6381e-05, + 2.2501e-06, 1.1269e-06], + [ 4.1798e-06, 6.1728e-06, -5.0291e-07, ..., 4.4629e-06, + 2.5015e-06, 2.2817e-06], + ..., + [ 2.5220e-06, -9.8586e-05, 3.1665e-08, ..., -1.1450e-04, + 8.0932e-07, 2.3656e-07], + [ 6.0238e-06, 1.0123e-06, 9.3132e-09, ..., -1.5432e-06, + 1.2303e-06, 6.7614e-07], + [ 1.4260e-05, 4.3623e-06, 9.3132e-10, ..., 2.3972e-06, + 1.3569e-06, 6.5938e-07]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0171, -0.0252, -0.0113, -0.0276, -0.0310, -0.0016, 0.0276, -0.0116, + 0.0335, -0.0020], device='cuda:0'), grad: tensor([ 2.0824e-06, 2.9755e-04, 2.3052e-05, 4.3416e-04, 1.9073e-05, + -5.2404e-04, 4.2409e-05, -3.2973e-04, 4.6901e-06, 3.0398e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 265.61, cls_loss 0.0068 cls_loss_mapping 0.0069 cls_loss_causal 0.5831 re_mapping 0.0088 re_causal 0.0225 /// teacc 98.76 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0636, -0.0855, -0.0538, ..., -0.0351, 0.1312, 0.1148], + [-0.1198, -0.1233, -0.1095, ..., -0.0957, -0.1344, -0.0644], + [-0.0792, -0.0585, 0.0952, ..., -0.1050, 0.1280, 0.0250], + ..., + [-0.0683, 0.0480, 0.0433, ..., 0.1121, -0.0875, -0.1027], + [-0.1618, 0.0491, -0.1066, ..., 0.0370, -0.0678, -0.0611], + [ 0.0430, -0.0497, -0.0580, ..., -0.0380, -0.0414, -0.0380]], + device='cuda:0'), grad: tensor([[ 1.1027e-05, 1.4063e-07, 5.1688e-08, ..., 5.3532e-06, + -2.7735e-06, -1.8906e-06], + [ 2.0117e-06, 8.6101e-07, 4.4238e-08, ..., 1.5143e-06, + 3.0966e-07, 1.2061e-07], + [ 2.1178e-06, 1.3635e-06, 1.7695e-08, ..., 1.1353e-06, + -3.8967e-06, -7.7765e-07], + ..., + [ 6.4373e-06, 2.2780e-06, 2.4540e-07, ..., -1.8487e-06, + 4.1462e-06, 1.0310e-06], + [ 5.2452e-05, 4.8429e-06, 6.8452e-08, ..., 2.2113e-05, + 6.4559e-06, 3.1479e-07], + [ 2.9996e-05, 1.1353e-06, 9.2201e-08, ..., 1.6868e-05, + 1.1567e-06, 6.7381e-07]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0180, -0.0253, -0.0121, -0.0280, -0.0309, -0.0013, 0.0275, -0.0124, + 0.0343, -0.0004], device='cuda:0'), grad: tensor([ 2.8715e-05, 4.3750e-05, 1.5959e-05, 4.3368e-04, -1.2684e-04, + -6.7759e-04, 9.4846e-06, 3.4899e-05, 1.5461e-04, 8.3804e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 265.41, cls_loss 0.0060 cls_loss_mapping 0.0064 cls_loss_causal 0.5533 re_mapping 0.0091 re_causal 0.0226 /// teacc 98.93 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0640, -0.0859, -0.0542, ..., -0.0353, 0.1319, 0.1158], + [-0.1215, -0.1236, -0.1095, ..., -0.0961, -0.1347, -0.0643], + [-0.0796, -0.0587, 0.0955, ..., -0.1054, 0.1282, 0.0253], + ..., + [-0.0686, 0.0485, 0.0432, ..., 0.1130, -0.0880, -0.1035], + [-0.1628, 0.0488, -0.1070, ..., 0.0367, -0.0682, -0.0614], + [ 0.0429, -0.0506, -0.0582, ..., -0.0389, -0.0417, -0.0387]], + device='cuda:0'), grad: tensor([[ 2.3469e-06, 4.5169e-08, 4.0047e-08, ..., 1.1781e-06, + -6.7521e-08, 4.0606e-06], + [ 2.0117e-06, 1.4203e-07, 2.4334e-05, ..., 1.1455e-06, + 2.7925e-05, -1.1303e-05], + [ 1.4147e-06, 2.5844e-07, -2.6867e-05, ..., 7.6788e-07, + -3.2365e-05, 1.1148e-06], + ..., + [ 1.9483e-06, -3.8370e-07, 2.1327e-06, ..., 4.4471e-07, + 2.7996e-06, 1.2117e-06], + [ 8.2329e-06, -5.9139e-07, 7.9628e-08, ..., 4.0159e-06, + 5.2806e-07, 1.9372e-06], + [ 2.9169e-06, 1.4016e-07, 6.5155e-06, ..., 1.8179e-06, + 6.1374e-07, 1.0403e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0177, -0.0250, -0.0125, -0.0288, -0.0305, -0.0006, 0.0275, -0.0122, + 0.0338, -0.0008], device='cuda:0'), grad: tensor([ 2.3663e-05, 3.5495e-05, -8.7559e-05, 1.1510e-04, -5.9932e-05, + -1.4782e-04, 1.3724e-05, 1.6600e-05, 2.0489e-05, 7.0274e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 265.37, cls_loss 0.0055 cls_loss_mapping 0.0065 cls_loss_causal 0.5722 re_mapping 0.0091 re_causal 0.0229 /// teacc 98.86 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0643, -0.0864, -0.0544, ..., -0.0357, 0.1323, 0.1163], + [-0.1220, -0.1240, -0.1098, ..., -0.0965, -0.1351, -0.0646], + [-0.0799, -0.0591, 0.0961, ..., -0.1059, 0.1290, 0.0260], + ..., + [-0.0688, 0.0474, 0.0429, ..., 0.1132, -0.0886, -0.1045], + [-0.1633, 0.0491, -0.1072, ..., 0.0370, -0.0686, -0.0617], + [ 0.0425, -0.0512, -0.0578, ..., -0.0395, -0.0419, -0.0394]], + device='cuda:0'), grad: tensor([[ 7.4506e-08, 4.2981e-07, 4.6566e-10, ..., 3.8417e-07, + -1.0310e-06, -1.2424e-06], + [ 5.4948e-08, 1.9418e-07, 4.6566e-10, ..., 1.7090e-07, + 2.0536e-07, 1.3178e-07], + [ 3.1991e-07, 3.6806e-06, -2.3283e-09, ..., 5.0925e-06, + 4.9286e-06, 2.9188e-06], + ..., + [ 5.2620e-08, -7.0874e-07, 2.7940e-09, ..., -8.2934e-07, + 4.1910e-07, 2.8266e-07], + [ 2.0443e-07, -6.2101e-06, 4.6566e-10, ..., -9.7454e-06, + -1.1772e-05, -6.7502e-06], + [ 2.2352e-07, 2.5239e-07, 4.6566e-10, ..., 2.0675e-07, + 5.8720e-07, 4.8103e-07]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0177, -0.0255, -0.0115, -0.0286, -0.0307, -0.0002, 0.0272, -0.0127, + 0.0340, -0.0009], device='cuda:0'), grad: tensor([ 7.7346e-07, 9.8161e-07, 2.4781e-05, 6.6236e-06, 2.6524e-06, + 1.7226e-05, 8.4564e-07, 4.5868e-07, -5.3018e-05, -1.3616e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 265.41, cls_loss 0.0063 cls_loss_mapping 0.0073 cls_loss_causal 0.5718 re_mapping 0.0092 re_causal 0.0221 /// teacc 98.93 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0649, -0.0869, -0.0546, ..., -0.0365, 0.1333, 0.1173], + [-0.1225, -0.1244, -0.1099, ..., -0.0971, -0.1352, -0.0645], + [-0.0804, -0.0603, 0.0967, ..., -0.1062, 0.1298, 0.0265], + ..., + [-0.0691, 0.0476, 0.0426, ..., 0.1141, -0.0896, -0.1058], + [-0.1642, 0.0487, -0.1075, ..., 0.0369, -0.0691, -0.0620], + [ 0.0420, -0.0520, -0.0578, ..., -0.0400, -0.0421, -0.0399]], + device='cuda:0'), grad: tensor([[ 2.5518e-07, 2.5891e-07, 4.6566e-10, ..., 3.5716e-07, + -4.9509e-06, -3.5148e-06], + [ 2.4633e-07, 1.1120e-06, -1.9278e-07, ..., 9.4902e-07, + 1.1781e-07, 5.4948e-08], + [ 1.8440e-07, 7.3295e-07, 1.2387e-07, ..., 6.4215e-07, + -1.5246e-06, 4.1723e-07], + ..., + [ 2.8824e-07, -7.9200e-06, 1.8626e-09, ..., -1.0408e-05, + 3.8883e-07, 1.1967e-07], + [ 6.0210e-07, 2.0396e-06, 3.2596e-09, ..., 1.7341e-06, + 4.6659e-07, 1.3411e-07], + [ 2.7046e-06, 3.0734e-06, 9.3132e-10, ..., 3.1069e-06, + 2.3376e-06, 1.3886e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0173, -0.0254, -0.0115, -0.0275, -0.0306, -0.0002, 0.0267, -0.0128, + 0.0335, -0.0012], device='cuda:0'), grad: tensor([-5.5209e-06, -1.7390e-05, -3.0510e-06, 4.9099e-06, 1.1921e-05, + -2.5630e-06, 4.3437e-06, -3.5316e-06, 8.9183e-06, 1.9483e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 265.55, cls_loss 0.0073 cls_loss_mapping 0.0079 cls_loss_causal 0.5915 re_mapping 0.0085 re_causal 0.0214 /// teacc 98.93 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0654, -0.0874, -0.0549, ..., -0.0365, 0.1334, 0.1174], + [-0.1229, -0.1268, -0.1098, ..., -0.0965, -0.1353, -0.0646], + [-0.0810, -0.0610, 0.0970, ..., -0.1073, 0.1301, 0.0269], + ..., + [-0.0693, 0.0489, 0.0430, ..., 0.1159, -0.0902, -0.1064], + [-0.1648, 0.0486, -0.1080, ..., 0.0370, -0.0692, -0.0622], + [ 0.0416, -0.0523, -0.0579, ..., -0.0414, -0.0422, -0.0405]], + device='cuda:0'), grad: tensor([[ 5.6392e-07, 2.3562e-07, 9.5926e-08, ..., -1.9204e-06, + -1.4119e-05, -8.2180e-06], + [ 5.5553e-07, 4.6752e-07, 5.3551e-08, ..., 1.0338e-06, + 1.2582e-06, 7.6462e-07], + [ 1.5823e-06, 3.3956e-06, 2.5611e-08, ..., 7.1991e-07, + -2.5146e-08, -3.1479e-07], + ..., + [ 4.7265e-07, 1.1437e-06, 2.4680e-08, ..., 1.8878e-06, + 2.8834e-06, 2.0470e-06], + [ 6.0834e-06, 1.3197e-06, 1.6764e-08, ..., 8.6799e-06, + 1.6317e-06, 9.0990e-07], + [ 2.3190e-07, 1.3923e-07, 1.3970e-09, ..., 9.4622e-06, + 4.3064e-06, 2.3022e-06]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0176, -0.0260, -0.0120, -0.0276, -0.0312, -0.0002, 0.0268, -0.0110, + 0.0334, -0.0015], device='cuda:0'), grad: tensor([-1.5162e-05, -1.2264e-03, 2.7806e-05, 5.8591e-05, -2.3067e-04, + -1.7807e-05, -6.2166e-07, 1.1492e-03, 3.2365e-05, 2.2197e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 265.24, cls_loss 0.0066 cls_loss_mapping 0.0085 cls_loss_causal 0.5924 re_mapping 0.0089 re_causal 0.0223 /// teacc 98.92 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0656, -0.0880, -0.0551, ..., -0.0360, 0.1342, 0.1181], + [-0.1232, -0.1274, -0.1095, ..., -0.0969, -0.1355, -0.0650], + [-0.0813, -0.0617, 0.0970, ..., -0.1086, 0.1309, 0.0275], + ..., + [-0.0698, 0.0482, 0.0429, ..., 0.1157, -0.0909, -0.1071], + [-0.1658, 0.0489, -0.1082, ..., 0.0373, -0.0697, -0.0625], + [ 0.0420, -0.0533, -0.0579, ..., -0.0414, -0.0429, -0.0411]], + device='cuda:0'), grad: tensor([[ 6.4401e-07, 2.5146e-08, 1.3644e-07, ..., 8.5682e-08, + 5.5879e-08, -5.2759e-07], + [ 2.7474e-07, 1.0291e-07, 1.8161e-08, ..., 1.6158e-07, + 1.0943e-06, 3.7579e-07], + [ 2.8173e-07, 1.2992e-07, 2.8871e-08, ..., 1.2014e-07, + 1.0394e-06, 1.9697e-07], + ..., + [ 7.2177e-08, 4.6100e-08, 4.6566e-10, ..., 7.9162e-08, + 4.0140e-07, 2.3749e-07], + [ 1.1353e-06, 2.2491e-07, 2.5146e-08, ..., 3.7486e-07, + -5.3644e-07, 7.0035e-07], + [ 2.8685e-07, -1.0990e-07, 2.7940e-09, ..., -2.3330e-07, + 2.5798e-06, 4.7917e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0172, -0.0255, -0.0123, -0.0271, -0.0309, -0.0002, 0.0265, -0.0121, + 0.0342, -0.0016], device='cuda:0'), grad: tensor([ 2.7977e-06, 7.4804e-06, 2.3004e-06, 1.1995e-05, -4.1425e-05, + -1.0125e-05, -1.3083e-05, 7.0594e-06, -5.8524e-06, 3.8743e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 265.30, cls_loss 0.0066 cls_loss_mapping 0.0076 cls_loss_causal 0.5834 re_mapping 0.0088 re_causal 0.0221 /// teacc 98.91 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0659, -0.0886, -0.0554, ..., -0.0360, 0.1348, 0.1188], + [-0.1240, -0.1281, -0.1100, ..., -0.0975, -0.1362, -0.0655], + [-0.0820, -0.0621, 0.0972, ..., -0.1093, 0.1312, 0.0275], + ..., + [-0.0701, 0.0490, 0.0428, ..., 0.1168, -0.0913, -0.1076], + [-0.1657, 0.0490, -0.1081, ..., 0.0385, -0.0673, -0.0629], + [ 0.0421, -0.0538, -0.0579, ..., -0.0416, -0.0432, -0.0415]], + device='cuda:0'), grad: tensor([[ 7.5437e-08, 1.6298e-07, 1.3970e-08, ..., 1.4110e-06, + -5.9232e-07, -3.8277e-07], + [ 5.8673e-08, 3.1143e-06, 9.3132e-10, ..., 4.8317e-06, + 6.7987e-08, 2.0489e-08], + [ 6.3330e-08, 3.4571e-06, 1.8626e-09, ..., 1.1332e-05, + -1.1146e-05, 1.0896e-07], + ..., + [ 6.4261e-08, -1.0543e-05, 0.0000e+00, ..., -5.7101e-05, + 5.1595e-07, 1.7695e-08], + [ 2.4587e-07, 8.8848e-07, 9.3132e-10, ..., 4.2841e-07, + 9.3691e-07, 7.5437e-08], + [ 1.4249e-07, 1.0384e-06, 0.0000e+00, ..., 2.2817e-06, + 2.3656e-07, 1.2014e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0170, -0.0257, -0.0118, -0.0278, -0.0317, 0.0002, 0.0256, -0.0118, + 0.0354, -0.0016], device='cuda:0'), grad: tensor([ 2.2613e-06, 5.0254e-06, -1.3141e-06, 8.0615e-06, 4.7028e-05, + 2.7016e-05, 1.3923e-06, -9.7692e-05, 6.0685e-06, 2.0899e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 265.39, cls_loss 0.0058 cls_loss_mapping 0.0057 cls_loss_causal 0.5625 re_mapping 0.0087 re_causal 0.0217 /// teacc 98.80 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0673, -0.0893, -0.0557, ..., -0.0370, 0.1342, 0.1183], + [-0.1248, -0.1284, -0.1102, ..., -0.0977, -0.1364, -0.0656], + [-0.0825, -0.0628, 0.0976, ..., -0.1104, 0.1317, 0.0275], + ..., + [-0.0704, 0.0497, 0.0427, ..., 0.1179, -0.0919, -0.1078], + [-0.1665, 0.0489, -0.1081, ..., 0.0385, -0.0676, -0.0634], + [ 0.0419, -0.0546, -0.0579, ..., -0.0425, -0.0435, -0.0418]], + device='cuda:0'), grad: tensor([[ 2.7381e-07, 6.1467e-08, 0.0000e+00, ..., -4.3474e-06, + -1.4052e-05, -8.6427e-06], + [ 8.3819e-08, 7.2364e-07, 2.0489e-08, ..., 3.8557e-07, + 6.4634e-07, 2.5891e-07], + [ 8.1025e-08, -2.4885e-06, -5.3551e-07, ..., 6.1002e-07, + -4.7497e-06, -5.6811e-07], + ..., + [ 2.1141e-07, 8.4750e-07, 5.1502e-07, ..., -1.6307e-06, + 4.7944e-06, 8.0001e-07], + [ 7.1824e-05, 2.7567e-06, 0.0000e+00, ..., 1.6224e-04, + 1.2098e-06, 9.8813e-07], + [ 2.1700e-07, 4.3306e-07, 0.0000e+00, ..., 1.8338e-06, + 3.0138e-06, 1.7975e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0177, -0.0259, -0.0120, -0.0282, -0.0312, 0.0004, 0.0263, -0.0114, + 0.0354, -0.0022], device='cuda:0'), grad: tensor([-3.1024e-05, 1.3180e-05, -2.5287e-05, 1.1124e-05, 2.1458e-05, + -2.1064e-04, 1.1973e-05, 1.5274e-05, 2.2864e-04, -3.4988e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 265.84, cls_loss 0.0067 cls_loss_mapping 0.0070 cls_loss_causal 0.5548 re_mapping 0.0088 re_causal 0.0220 /// teacc 98.98 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0676, -0.0898, -0.0558, ..., -0.0372, 0.1341, 0.1190], + [-0.1254, -0.1290, -0.1109, ..., -0.0984, -0.1372, -0.0663], + [-0.0830, -0.0635, 0.0983, ..., -0.1113, 0.1325, 0.0282], + ..., + [-0.0711, 0.0505, 0.0426, ..., 0.1194, -0.0925, -0.1083], + [-0.1669, 0.0487, -0.1074, ..., 0.0384, -0.0679, -0.0638], + [ 0.0421, -0.0554, -0.0579, ..., -0.0436, -0.0426, -0.0425]], + device='cuda:0'), grad: tensor([[ 3.5763e-07, 3.7439e-07, 9.3132e-10, ..., 1.7229e-07, + -3.6322e-08, 3.5670e-07], + [ 2.8126e-07, 4.6566e-07, 0.0000e+00, ..., 3.2224e-07, + 1.7866e-05, 1.1593e-05], + [ 1.7695e-07, 1.1548e-06, 9.3132e-10, ..., 2.0377e-06, + -4.4316e-05, -3.3647e-05], + ..., + [ 5.3365e-07, -7.5437e-07, 0.0000e+00, ..., -1.8561e-06, + 2.1607e-06, 1.2936e-06], + [ 1.3076e-06, 1.2711e-05, 0.0000e+00, ..., -1.7267e-06, + 1.0692e-06, 1.1930e-06], + [ 4.9360e-07, -1.1541e-05, 0.0000e+00, ..., 8.0187e-07, + 1.4920e-06, 9.2201e-07]], device='cuda:0') +Epoch 122, bias, value: tensor([-1.8815e-02, -2.6876e-02, -1.1588e-02, -2.8067e-02, -3.0534e-02, + 2.2794e-05, 2.5934e-02, -1.1187e-02, 3.5494e-02, -9.9249e-04], + device='cuda:0'), grad: tensor([ 9.5665e-06, -4.2468e-05, -1.1522e-04, -7.1287e-04, -1.0513e-05, + 7.2145e-04, 1.0979e-04, 9.1419e-06, 7.0155e-05, -3.9339e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 265.15, cls_loss 0.0055 cls_loss_mapping 0.0066 cls_loss_causal 0.5497 re_mapping 0.0087 re_causal 0.0221 /// teacc 98.93 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0679, -0.0903, -0.0559, ..., -0.0373, 0.1348, 0.1198], + [-0.1271, -0.1279, -0.1110, ..., -0.0987, -0.1375, -0.0664], + [-0.0836, -0.0635, 0.0984, ..., -0.1117, 0.1330, 0.0286], + ..., + [-0.0711, 0.0500, 0.0426, ..., 0.1199, -0.0932, -0.1091], + [-0.1673, 0.0485, -0.1074, ..., 0.0382, -0.0681, -0.0645], + [ 0.0423, -0.0549, -0.0579, ..., -0.0435, -0.0431, -0.0437]], + device='cuda:0'), grad: tensor([[ 1.7788e-07, 8.4378e-07, 0.0000e+00, ..., 4.5914e-07, + -6.1803e-06, -4.0755e-06], + [ 1.5087e-07, 6.7148e-07, 0.0000e+00, ..., 4.2841e-07, + 9.4157e-07, 3.1386e-07], + [ 3.0454e-07, 2.0433e-06, 0.0000e+00, ..., -2.7623e-06, + -9.0748e-06, -2.3454e-05], + ..., + [ 1.2293e-07, 1.9558e-08, 0.0000e+00, ..., -2.5984e-07, + 1.1288e-06, 5.1130e-07], + [ 1.0598e-06, -6.4895e-06, 0.0000e+00, ..., -7.3574e-08, + 2.3376e-07, 2.0996e-05], + [-2.7940e-09, 1.4491e-06, 0.0000e+00, ..., 3.9209e-07, + 4.7944e-06, 2.1998e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0184, -0.0260, -0.0115, -0.0280, -0.0306, -0.0006, 0.0261, -0.0127, + 0.0354, -0.0003], device='cuda:0'), grad: tensor([ 2.2985e-06, -8.1122e-05, -1.7017e-05, 3.8624e-05, -6.3241e-05, + 7.4618e-06, 8.1435e-06, 7.1764e-05, -9.0599e-06, 4.1991e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 265.36, cls_loss 0.0069 cls_loss_mapping 0.0074 cls_loss_causal 0.5992 re_mapping 0.0088 re_causal 0.0213 /// teacc 98.90 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0684, -0.0909, -0.0560, ..., -0.0376, 0.1358, 0.1210], + [-0.1283, -0.1284, -0.1121, ..., -0.0996, -0.1379, -0.0664], + [-0.0836, -0.0641, 0.0993, ..., -0.1124, 0.1334, 0.0289], + ..., + [-0.0716, 0.0507, 0.0428, ..., 0.1216, -0.0941, -0.1107], + [-0.1676, 0.0480, -0.1070, ..., 0.0385, -0.0673, -0.0656], + [ 0.0421, -0.0559, -0.0580, ..., -0.0448, -0.0437, -0.0450]], + device='cuda:0'), grad: tensor([[ 6.5379e-07, 4.0978e-07, 2.2352e-08, ..., 2.8592e-07, + -6.1467e-06, -3.0473e-06], + [ 7.0315e-07, 1.2890e-06, 2.9802e-08, ..., 5.6718e-07, + 1.3132e-07, 8.7544e-08], + [ 5.3924e-07, 4.5076e-07, -3.5297e-06, ..., 3.2969e-07, + 3.1758e-07, -1.1828e-07], + ..., + [ 2.7400e-06, 2.1420e-06, 1.2061e-06, ..., -1.8617e-06, + 7.6648e-07, 5.7835e-07], + [ 3.2708e-06, 1.8105e-06, 1.1204e-06, ..., 1.0859e-06, + 4.2841e-07, 3.6415e-07], + [ 4.3400e-06, 5.3979e-06, 1.8626e-09, ..., 1.4929e-06, + 2.0768e-06, 1.1083e-06]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0181, -0.0262, -0.0107, -0.0272, -0.0306, -0.0013, 0.0257, -0.0124, + 0.0351, -0.0007], device='cuda:0'), grad: tensor([-4.9397e-06, 2.9616e-07, -6.8367e-05, 7.5674e-04, 1.7866e-05, + -7.7915e-04, 2.9914e-06, 3.1024e-05, 2.9266e-05, 1.4521e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 265.50, cls_loss 0.0058 cls_loss_mapping 0.0068 cls_loss_causal 0.5907 re_mapping 0.0086 re_causal 0.0222 /// teacc 98.90 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0690, -0.0912, -0.0564, ..., -0.0384, 0.1364, 0.1215], + [-0.1286, -0.1285, -0.1127, ..., -0.0996, -0.1381, -0.0664], + [-0.0843, -0.0643, 0.0996, ..., -0.1122, 0.1335, 0.0295], + ..., + [-0.0716, 0.0510, 0.0427, ..., 0.1221, -0.0946, -0.1113], + [-0.1675, 0.0478, -0.1065, ..., 0.0388, -0.0671, -0.0668], + [ 0.0417, -0.0562, -0.0580, ..., -0.0453, -0.0441, -0.0459]], + device='cuda:0'), grad: tensor([[ 7.9349e-07, 1.0701e-06, 0.0000e+00, ..., -7.3109e-07, + -5.3756e-06, -6.4299e-06], + [ 1.6484e-07, 1.1828e-07, 0.0000e+00, ..., 2.9616e-07, + 1.5730e-06, 1.4845e-06], + [ 5.4948e-08, 3.9302e-07, 0.0000e+00, ..., 9.5926e-08, + 1.0580e-06, 4.7684e-07], + ..., + [ 3.2596e-07, 2.9430e-07, 0.0000e+00, ..., -2.3283e-08, + 1.7602e-07, 1.5553e-07], + [ 8.0280e-07, -6.2734e-06, 0.0000e+00, ..., 6.5751e-07, + -7.8082e-06, 1.0990e-06], + [ 2.7008e-06, 2.8182e-06, 0.0000e+00, ..., 1.1576e-06, + 5.6997e-06, 1.6699e-06]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0181, -0.0257, -0.0107, -0.0289, -0.0295, -0.0012, 0.0259, -0.0126, + 0.0358, -0.0013], device='cuda:0'), grad: tensor([-6.9067e-06, 3.4459e-06, 3.8035e-06, 1.4031e-04, -1.4268e-05, + -1.4412e-04, 8.2925e-06, 5.4240e-06, -1.6361e-05, 2.0251e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 265.47, cls_loss 0.0052 cls_loss_mapping 0.0069 cls_loss_causal 0.5725 re_mapping 0.0085 re_causal 0.0214 /// teacc 98.90 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0693, -0.0923, -0.0565, ..., -0.0376, 0.1365, 0.1218], + [-0.1289, -0.1290, -0.1130, ..., -0.1001, -0.1391, -0.0666], + [-0.0847, -0.0646, 0.0999, ..., -0.1125, 0.1325, 0.0295], + ..., + [-0.0720, 0.0515, 0.0428, ..., 0.1230, -0.0957, -0.1121], + [-0.1683, 0.0479, -0.1060, ..., 0.0387, -0.0671, -0.0676], + [ 0.0415, -0.0571, -0.0580, ..., -0.0460, -0.0443, -0.0464]], + device='cuda:0'), grad: tensor([[ 2.5611e-07, 8.7544e-07, 1.8626e-08, ..., 8.7079e-07, + -1.0908e-05, -1.4096e-05], + [ 5.4017e-08, 2.2445e-06, 9.3132e-10, ..., 4.5262e-06, + 6.8367e-05, 6.5506e-05], + [ 2.4214e-08, 9.8720e-08, -8.7544e-08, ..., 8.5682e-08, + -7.7486e-05, -7.3075e-05], + ..., + [ 1.3318e-07, -2.8804e-05, 1.9558e-08, ..., -6.2168e-05, + 4.8205e-06, 4.1872e-06], + [ 1.5516e-06, 1.8626e-08, 8.3819e-09, ..., 2.0918e-06, + 1.0431e-06, 1.0375e-06], + [ 1.3318e-06, 2.2769e-05, 2.7940e-09, ..., 4.7892e-05, + 4.8522e-07, 3.5297e-07]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0182, -0.0258, -0.0113, -0.0288, -0.0291, -0.0012, 0.0269, -0.0126, + 0.0355, -0.0014], device='cuda:0'), grad: tensor([-1.4178e-05, 2.4104e-04, -2.5868e-04, 2.8223e-05, 1.9222e-05, + -1.9252e-05, 2.8104e-05, -1.2624e-04, 6.4969e-06, 9.5367e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 261.93, cls_loss 0.0061 cls_loss_mapping 0.0086 cls_loss_causal 0.5695 re_mapping 0.0082 re_causal 0.0203 /// teacc 98.89 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0690, -0.0929, -0.0567, ..., -0.0374, 0.1378, 0.1230], + [-0.1293, -0.1299, -0.1127, ..., -0.1003, -0.1395, -0.0670], + [-0.0852, -0.0651, 0.1001, ..., -0.1138, 0.1332, 0.0302], + ..., + [-0.0719, 0.0522, 0.0427, ..., 0.1242, -0.0967, -0.1135], + [-0.1693, 0.0473, -0.1065, ..., 0.0383, -0.0676, -0.0680], + [ 0.0413, -0.0579, -0.0584, ..., -0.0472, -0.0449, -0.0477]], + device='cuda:0'), grad: tensor([[ 3.5856e-07, 1.4063e-07, 1.5832e-08, ..., 4.6156e-06, + 2.3395e-06, 4.0531e-06], + [ 4.8522e-07, 2.9337e-07, 1.3039e-08, ..., 4.5635e-07, + 6.6590e-07, 4.5262e-07], + [ 4.7684e-07, 4.5355e-07, -1.7695e-07, ..., 1.2964e-06, + -4.5635e-06, -5.3085e-08], + ..., + [ 4.4052e-07, 1.5553e-07, 3.4459e-08, ..., -1.0617e-07, + 1.4156e-06, 7.4599e-07], + [ 2.7381e-06, 7.8045e-07, 3.8184e-08, ..., -8.2925e-06, + -1.5244e-05, -1.5765e-05], + [ 3.7905e-07, 4.3958e-07, 2.4866e-07, ..., 1.4752e-06, + 8.4471e-07, 5.8208e-07]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0174, -0.0261, -0.0114, -0.0285, -0.0293, -0.0015, 0.0270, -0.0124, + 0.0350, -0.0014], device='cuda:0'), grad: tensor([ 2.7850e-05, 5.7463e-07, -3.1237e-06, 1.5342e-04, 5.4501e-06, + -1.2016e-04, 1.2271e-05, 5.7593e-06, -5.4181e-05, -2.8074e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 126---------------------------------------------------- +epoch 126, time 278.75, cls_loss 0.0070 cls_loss_mapping 0.0083 cls_loss_causal 0.5859 re_mapping 0.0089 re_causal 0.0215 /// teacc 99.01 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0692, -0.0935, -0.0571, ..., -0.0378, 0.1388, 0.1243], + [-0.1299, -0.1309, -0.1132, ..., -0.1021, -0.1404, -0.0673], + [-0.0865, -0.0671, 0.1005, ..., -0.1147, 0.1343, 0.0308], + ..., + [-0.0721, 0.0527, 0.0426, ..., 0.1259, -0.0978, -0.1150], + [-0.1709, 0.0469, -0.1069, ..., 0.0361, -0.0676, -0.0707], + [ 0.0406, -0.0589, -0.0587, ..., -0.0480, -0.0449, -0.0469]], + device='cuda:0'), grad: tensor([[-1.3225e-07, 2.7101e-07, 1.2107e-08, ..., -2.6561e-06, + -1.6034e-05, -1.3024e-05], + [ 1.3504e-07, 2.7660e-07, 0.0000e+00, ..., 1.9465e-07, + 2.3562e-07, -2.5984e-07], + [ 6.2399e-08, 1.3318e-07, -2.5146e-08, ..., 2.7847e-07, + 3.1665e-08, 1.6671e-07], + ..., + [ 8.0094e-08, 3.3639e-06, 0.0000e+00, ..., 4.0643e-06, + 7.1805e-07, 7.8045e-07], + [ 4.7795e-06, 7.0110e-06, 2.7940e-09, ..., 2.3302e-06, + 4.2319e-06, 3.6433e-06], + [ 3.2689e-06, -4.9621e-06, 0.0000e+00, ..., -1.1064e-05, + 4.6287e-07, 4.3586e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0173, -0.0263, -0.0115, -0.0281, -0.0294, 0.0002, 0.0269, -0.0119, + 0.0333, -0.0019], device='cuda:0'), grad: tensor([-1.6406e-05, -2.6658e-05, 1.7613e-05, -4.6968e-05, 2.1681e-05, + 4.8906e-05, 9.4697e-06, 4.3005e-05, 2.3589e-05, -7.4267e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 265.45, cls_loss 0.0057 cls_loss_mapping 0.0062 cls_loss_causal 0.5984 re_mapping 0.0084 re_causal 0.0213 /// teacc 98.83 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0702, -0.0941, -0.0573, ..., -0.0390, 0.1393, 0.1251], + [-0.1304, -0.1313, -0.1133, ..., -0.1026, -0.1406, -0.0674], + [-0.0872, -0.0679, 0.1007, ..., -0.1144, 0.1350, 0.0317], + ..., + [-0.0726, 0.0531, 0.0430, ..., 0.1269, -0.0984, -0.1159], + [-0.1716, 0.0469, -0.1071, ..., 0.0361, -0.0679, -0.0714], + [ 0.0404, -0.0597, -0.0594, ..., -0.0491, -0.0452, -0.0473]], + device='cuda:0'), grad: tensor([[ 2.0582e-07, 3.6974e-07, 4.1910e-08, ..., 4.5262e-07, + -1.7602e-07, -7.5437e-08], + [ 3.2596e-08, 8.7079e-07, 1.6764e-08, ..., 7.0222e-07, + 1.6391e-07, 8.0094e-08], + [ 6.3330e-08, 2.9020e-06, -3.9395e-07, ..., 3.8669e-06, + -3.4720e-06, -4.2282e-07], + ..., + [ 9.3132e-09, -8.8811e-05, 1.3970e-08, ..., -5.9038e-05, + 1.5926e-07, 8.2888e-08], + [ 2.1420e-07, 7.3202e-07, 3.1665e-08, ..., -2.2411e-05, + 2.6450e-07, -9.8255e-07], + [ 7.3574e-08, 1.1204e-06, 9.3132e-10, ..., 9.8813e-07, + 2.5798e-07, 1.3784e-07]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0175, -0.0260, -0.0107, -0.0275, -0.0295, 0.0005, 0.0260, -0.0119, + 0.0331, -0.0023], device='cuda:0'), grad: tensor([ 3.5744e-06, 2.0005e-06, -5.1469e-05, 1.0329e-04, 1.5378e-05, + 3.1501e-05, 5.4613e-06, -8.4579e-05, -3.9726e-05, 1.4544e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 265.56, cls_loss 0.0057 cls_loss_mapping 0.0069 cls_loss_causal 0.5453 re_mapping 0.0082 re_causal 0.0211 /// teacc 98.90 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0710, -0.0947, -0.0576, ..., -0.0389, 0.1381, 0.1256], + [-0.1309, -0.1319, -0.1111, ..., -0.1029, -0.1408, -0.0676], + [-0.0877, -0.0676, 0.1021, ..., -0.1145, 0.1360, 0.0327], + ..., + [-0.0729, 0.0530, 0.0408, ..., 0.1272, -0.1000, -0.1177], + [-0.1721, 0.0471, -0.1075, ..., 0.0364, -0.0681, -0.0719], + [ 0.0401, -0.0600, -0.0602, ..., -0.0498, -0.0435, -0.0482]], + device='cuda:0'), grad: tensor([[ 1.3225e-07, 1.2759e-07, 1.9185e-07, ..., 4.8336e-07, + 1.1213e-06, 1.1064e-06], + [ 1.2573e-07, 1.3173e-05, 7.4506e-08, ..., 1.4426e-06, + 3.2876e-07, 3.4366e-07], + [ 4.8429e-08, 2.2352e-07, -3.1833e-06, ..., 1.8906e-07, + -1.8865e-05, -1.8656e-05], + ..., + [ 3.4459e-08, 2.2575e-05, 2.7400e-06, ..., 2.9802e-06, + 1.0483e-05, 1.1310e-05], + [ 1.8552e-06, -1.6605e-06, 1.9558e-08, ..., -1.1772e-05, + 5.7258e-06, 5.1074e-06], + [ 1.1083e-07, -6.1840e-07, 3.7253e-09, ..., -2.0321e-06, + 2.5146e-08, 2.4214e-08]], device='cuda:0') +Epoch 130, bias, value: tensor([-1.9512e-02, -2.5615e-02, -1.0418e-02, -2.7191e-02, -2.9085e-02, + 7.2359e-05, 2.5716e-02, -1.2597e-02, 3.3429e-02, -1.1608e-03], + device='cuda:0'), grad: tensor([ 4.0084e-06, 1.9163e-05, -4.0025e-05, -5.7638e-05, -1.2862e-06, + 2.3425e-05, -6.2771e-07, 7.2002e-05, -1.5870e-06, -1.7509e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 265.23, cls_loss 0.0059 cls_loss_mapping 0.0060 cls_loss_causal 0.5636 re_mapping 0.0082 re_causal 0.0211 /// teacc 98.92 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0715, -0.0953, -0.0577, ..., -0.0389, 0.1384, 0.1260], + [-0.1347, -0.1323, -0.1111, ..., -0.1033, -0.1414, -0.0698], + [-0.0848, -0.0678, 0.1027, ..., -0.1144, 0.1367, 0.0349], + ..., + [-0.0732, 0.0524, 0.0402, ..., 0.1269, -0.1013, -0.1196], + [-0.1729, 0.0478, -0.1077, ..., 0.0367, -0.0684, -0.0730], + [ 0.0396, -0.0605, -0.0601, ..., -0.0500, -0.0437, -0.0487]], + device='cuda:0'), grad: tensor([[ 2.9746e-06, 4.4983e-07, 8.3819e-09, ..., 2.6077e-06, + -2.4904e-06, 4.4424e-07], + [ 8.7079e-07, 7.0520e-06, 5.4017e-08, ..., 7.0184e-06, + 3.2131e-07, 3.3528e-07], + [ 5.8021e-07, 2.2147e-06, -4.2841e-08, ..., 2.8536e-06, + -4.7777e-07, -5.9977e-07], + ..., + [ 1.8906e-07, -3.9124e-04, -7.6368e-08, ..., -3.2806e-04, + 5.2433e-07, 5.3830e-07], + [ 9.2506e-05, 1.4743e-06, 3.7253e-09, ..., 9.7871e-05, + 1.7852e-05, 2.6047e-05], + [ 8.1025e-07, 1.4313e-05, 1.7695e-08, ..., 1.2875e-05, + 1.1679e-06, 8.0653e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0196, -0.0266, -0.0089, -0.0265, -0.0290, -0.0003, 0.0264, -0.0134, + 0.0334, -0.0010], device='cuda:0'), grad: tensor([ 6.7651e-06, 1.0377e-04, 6.2324e-06, 6.8998e-04, -1.1033e-04, + -2.3341e-04, -1.2696e-04, -7.3147e-04, 3.4618e-04, 4.8310e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 265.15, cls_loss 0.0061 cls_loss_mapping 0.0064 cls_loss_causal 0.5444 re_mapping 0.0082 re_causal 0.0212 /// teacc 98.97 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0721, -0.0964, -0.0578, ..., -0.0393, 0.1387, 0.1262], + [-0.1354, -0.1326, -0.1111, ..., -0.1031, -0.1416, -0.0696], + [-0.0849, -0.0684, 0.1024, ..., -0.1152, 0.1369, 0.0353], + ..., + [-0.0731, 0.0532, 0.0413, ..., 0.1282, -0.1017, -0.1207], + [-0.1739, 0.0479, -0.1078, ..., 0.0368, -0.0686, -0.0737], + [ 0.0401, -0.0610, -0.0605, ..., -0.0503, -0.0438, -0.0492]], + device='cuda:0'), grad: tensor([[ 1.3383e-06, 5.4017e-08, 2.4214e-08, ..., 4.2841e-08, + -2.3283e-07, 7.6927e-07], + [ 9.4995e-08, 1.1548e-07, 1.2107e-08, ..., 6.7055e-08, + 5.6811e-08, 6.1467e-08], + [ 2.6263e-07, 7.8045e-07, 1.4529e-07, ..., 6.6124e-07, + -7.9069e-07, -1.1427e-06], + ..., + [ 3.4459e-08, -1.1837e-06, -1.7136e-07, ..., -1.0915e-06, + 3.0734e-07, 3.9488e-07], + [ 3.5111e-07, 2.5705e-07, 5.8673e-08, ..., 1.5926e-07, + 4.5914e-07, 5.5041e-07], + [ 4.7497e-08, -1.3504e-07, -1.1828e-07, ..., 1.8068e-07, + 1.1828e-07, 9.4064e-08]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0198, -0.0264, -0.0091, -0.0247, -0.0302, -0.0023, 0.0268, -0.0125, + 0.0331, -0.0006], device='cuda:0'), grad: tensor([ 2.9057e-06, -5.3458e-06, -8.2236e-07, 1.5311e-06, 3.0816e-05, + 2.1458e-06, -3.1918e-05, -2.3562e-07, 4.4852e-06, -3.5297e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 265.42, cls_loss 0.0058 cls_loss_mapping 0.0062 cls_loss_causal 0.5757 re_mapping 0.0080 re_causal 0.0205 /// teacc 98.94 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0743, -0.0968, -0.0584, ..., -0.0398, 0.1374, 0.1268], + [-0.1356, -0.1330, -0.1114, ..., -0.1039, -0.1417, -0.0697], + [-0.0852, -0.0688, 0.1025, ..., -0.1157, 0.1371, 0.0351], + ..., + [-0.0735, 0.0536, 0.0417, ..., 0.1286, -0.1020, -0.1210], + [-0.1752, 0.0486, -0.1082, ..., 0.0372, -0.0689, -0.0746], + [ 0.0400, -0.0618, -0.0611, ..., -0.0508, -0.0419, -0.0500]], + device='cuda:0'), grad: tensor([[ 1.4435e-07, 7.9162e-08, 1.3970e-08, ..., 1.9651e-07, + -1.8254e-05, -7.4767e-06], + [ 8.4750e-08, -6.1877e-06, 2.7940e-09, ..., 8.7172e-06, + 6.8173e-07, 6.9849e-08], + [ 5.5879e-08, 4.2561e-07, 6.5193e-09, ..., 1.7835e-06, + 1.3690e-07, 1.5646e-07], + ..., + [ 1.2759e-07, 5.2191e-06, -1.3039e-08, ..., 2.1141e-07, + 1.3225e-07, 3.3528e-08], + [ 8.7321e-06, -7.2420e-06, 3.7253e-09, ..., 1.2644e-05, + -3.4738e-07, 3.8091e-07], + [ 2.3786e-06, 5.3346e-06, 6.5193e-09, ..., 1.4380e-05, + 1.5467e-05, 6.1616e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0214, -0.0262, -0.0095, -0.0249, -0.0303, -0.0022, 0.0269, -0.0125, + 0.0332, 0.0008], device='cuda:0'), grad: tensor([-2.0459e-05, -7.8011e-04, 6.4850e-04, 2.9266e-05, -9.0361e-05, + -2.5094e-05, 1.4849e-05, 1.2887e-04, 1.8561e-06, 9.2268e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 264.84, cls_loss 0.0062 cls_loss_mapping 0.0054 cls_loss_causal 0.5834 re_mapping 0.0079 re_causal 0.0199 /// teacc 98.88 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0746, -0.0984, -0.0590, ..., -0.0400, 0.1378, 0.1284], + [-0.1360, -0.1339, -0.1115, ..., -0.1051, -0.1422, -0.0699], + [-0.0854, -0.0696, 0.1024, ..., -0.1176, 0.1372, 0.0349], + ..., + [-0.0736, 0.0539, 0.0419, ..., 0.1296, -0.1024, -0.1214], + [-0.1766, 0.0496, -0.1086, ..., 0.0381, -0.0686, -0.0754], + [ 0.0392, -0.0626, -0.0610, ..., -0.0519, -0.0418, -0.0521]], + device='cuda:0'), grad: tensor([[ 5.0385e-07, 3.4366e-07, 5.0850e-07, ..., 4.7125e-07, + -9.8720e-08, -7.4040e-07], + [ 1.0990e-07, 3.2037e-07, 7.1712e-08, ..., 2.9616e-07, + 4.9267e-07, 8.7544e-08], + [ 2.3190e-07, 3.2969e-07, 1.2666e-07, ..., 1.7788e-07, + 1.2275e-06, 7.2643e-08], + ..., + [-2.5053e-07, -3.9786e-06, 1.1176e-07, ..., -5.1521e-06, + 6.9570e-07, 8.4750e-08], + [ 2.0023e-06, 2.2352e-07, 3.1013e-07, ..., 1.0738e-06, + 1.4622e-06, 2.9057e-07], + [ 2.5705e-06, 2.4382e-06, 1.7788e-07, ..., 3.4589e-06, + 2.7865e-06, 5.3272e-07]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0216, -0.0265, -0.0100, -0.0250, -0.0295, -0.0016, 0.0259, -0.0121, + 0.0338, 0.0005], device='cuda:0'), grad: tensor([ 3.2894e-06, 3.1702e-06, 4.2617e-06, 2.0750e-06, -1.3888e-04, + -8.3074e-06, 1.9085e-04, -4.2319e-06, 9.3877e-06, -6.1631e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 265.03, cls_loss 0.0079 cls_loss_mapping 0.0069 cls_loss_causal 0.5588 re_mapping 0.0080 re_causal 0.0200 /// teacc 98.99 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0750, -0.0991, -0.0602, ..., -0.0405, 0.1383, 0.1310], + [-0.1367, -0.1344, -0.1117, ..., -0.1058, -0.1432, -0.0699], + [-0.0857, -0.0701, 0.1031, ..., -0.1186, 0.1370, 0.0341], + ..., + [-0.0738, 0.0532, 0.0417, ..., 0.1305, -0.1027, -0.1218], + [-0.1776, 0.0496, -0.1089, ..., 0.0380, -0.0689, -0.0764], + [ 0.0390, -0.0627, -0.0613, ..., -0.0520, -0.0410, -0.0535]], + device='cuda:0'), grad: tensor([[ 3.1851e-07, 2.8685e-07, 8.3819e-09, ..., 3.1106e-07, + -8.2422e-07, -7.3388e-07], + [ 5.2992e-07, 6.1654e-07, 1.6764e-08, ..., 6.0424e-06, + 3.1572e-07, 7.0781e-08], + [ 1.1735e-07, 3.3993e-07, -3.5111e-07, ..., 4.1798e-06, + -1.3234e-06, -8.1025e-08], + ..., + [ 1.1912e-06, -5.4389e-07, 1.6764e-08, ..., -1.5842e-06, + 1.3784e-07, 2.7008e-08], + [ 2.3246e-06, -4.0606e-06, 1.2107e-08, ..., -1.1906e-05, + -1.4687e-06, 4.4052e-07], + [ 1.4603e-06, 2.9821e-06, 9.3132e-10, ..., 1.6605e-06, + 5.9512e-07, 3.6042e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0223, -0.0257, -0.0110, -0.0256, -0.0304, -0.0011, 0.0256, -0.0119, + 0.0336, 0.0016], device='cuda:0'), grad: tensor([ 5.8860e-07, -4.8423e-04, 3.9673e-04, -8.1491e-04, 2.2426e-05, + 8.6403e-04, -2.6803e-06, 6.9141e-05, -6.3479e-05, 1.3307e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 265.59, cls_loss 0.0055 cls_loss_mapping 0.0058 cls_loss_causal 0.5690 re_mapping 0.0079 re_causal 0.0201 /// teacc 98.94 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0783, -0.0999, -0.0614, ..., -0.0413, 0.1361, 0.1287], + [-0.1371, -0.1349, -0.1118, ..., -0.1060, -0.1436, -0.0700], + [-0.0858, -0.0705, 0.1034, ..., -0.1189, 0.1376, 0.0344], + ..., + [-0.0741, 0.0541, 0.0419, ..., 0.1317, -0.1032, -0.1222], + [-0.1778, 0.0493, -0.1089, ..., 0.0381, -0.0690, -0.0770], + [ 0.0388, -0.0633, -0.0615, ..., -0.0526, -0.0411, -0.0546]], + device='cuda:0'), grad: tensor([[ 1.9185e-07, 1.2293e-07, 7.4506e-09, ..., 5.0291e-08, + -1.4333e-06, -7.1619e-07], + [ 6.0536e-07, 1.5842e-06, 0.0000e+00, ..., 8.2888e-07, + 1.1642e-07, 5.6811e-08], + [ 2.7567e-07, 6.5472e-07, 7.4506e-09, ..., 7.9162e-08, + -9.3132e-10, 5.5879e-08], + ..., + [ 3.2689e-07, -1.1995e-06, 0.0000e+00, ..., -2.1067e-06, + 6.6124e-08, 2.7940e-08], + [ 3.8967e-06, 5.9605e-06, 9.3132e-10, ..., 2.5891e-07, + 2.2817e-07, 1.0338e-07], + [ 1.8440e-06, 1.8505e-06, 0.0000e+00, ..., 1.0962e-06, + 9.5740e-07, 4.7963e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0242, -0.0256, -0.0110, -0.0265, -0.0302, -0.0005, 0.0276, -0.0116, + 0.0335, 0.0015], device='cuda:0'), grad: tensor([-1.1129e-06, -3.3110e-05, 5.6624e-06, -2.8372e-05, -5.6475e-06, + 1.5080e-05, 2.0824e-06, -4.0140e-07, 1.5363e-05, 3.0473e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 265.17, cls_loss 0.0063 cls_loss_mapping 0.0085 cls_loss_causal 0.5731 re_mapping 0.0081 re_causal 0.0200 /// teacc 98.97 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0784, -0.1008, -0.0619, ..., -0.0414, 0.1366, 0.1295], + [-0.1377, -0.1354, -0.1119, ..., -0.1075, -0.1439, -0.0705], + [-0.0861, -0.0712, 0.1034, ..., -0.1201, 0.1382, 0.0346], + ..., + [-0.0745, 0.0541, 0.0419, ..., 0.1329, -0.1044, -0.1231], + [-0.1785, 0.0494, -0.1089, ..., 0.0381, -0.0692, -0.0777], + [ 0.0383, -0.0635, -0.0618, ..., -0.0530, -0.0408, -0.0569]], + device='cuda:0'), grad: tensor([[ 2.1793e-07, 3.6508e-07, 0.0000e+00, ..., 1.5367e-07, + 9.2201e-08, 4.0978e-08], + [ 3.0175e-07, 5.1782e-07, 0.0000e+00, ..., 8.8569e-07, + 5.7742e-08, 2.2352e-08], + [ 8.0839e-07, 2.0657e-06, 0.0000e+00, ..., 1.1427e-06, + -8.2888e-08, -1.7881e-07], + ..., + [ 1.1269e-07, -4.9826e-07, 0.0000e+00, ..., -5.1688e-07, + 2.1420e-08, 1.3039e-08], + [-1.1317e-05, -2.6971e-05, 0.0000e+00, ..., -9.0301e-06, + -4.4927e-06, -1.2740e-06], + [ 4.8708e-07, 1.2862e-06, 0.0000e+00, ..., 1.8273e-06, + 9.6858e-08, 2.9802e-08]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0242, -0.0255, -0.0111, -0.0273, -0.0303, 0.0005, 0.0270, -0.0119, + 0.0333, 0.0022], device='cuda:0'), grad: tensor([ 2.7101e-06, 8.8394e-05, 6.1430e-06, 7.4059e-06, -2.7180e-04, + 2.4855e-05, 4.3988e-05, 1.5676e-05, -8.3447e-05, 1.6582e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 265.39, cls_loss 0.0054 cls_loss_mapping 0.0055 cls_loss_causal 0.5564 re_mapping 0.0080 re_causal 0.0199 /// teacc 98.94 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0787, -0.1014, -0.0622, ..., -0.0423, 0.1369, 0.1298], + [-0.1380, -0.1360, -0.1119, ..., -0.1080, -0.1443, -0.0707], + [-0.0864, -0.0718, 0.1036, ..., -0.1208, 0.1379, 0.0342], + ..., + [-0.0749, 0.0538, 0.0417, ..., 0.1335, -0.1048, -0.1236], + [-0.1791, 0.0493, -0.1091, ..., 0.0381, -0.0694, -0.0787], + [ 0.0382, -0.0642, -0.0613, ..., -0.0535, -0.0411, -0.0589]], + device='cuda:0'), grad: tensor([[ 1.7229e-06, 5.5879e-09, 3.7253e-09, ..., 1.3597e-07, + -9.9421e-05, -6.0499e-05], + [ 1.2014e-07, 7.6368e-08, 1.1176e-08, ..., 8.6613e-08, + 3.5495e-05, 2.1547e-05], + [ 6.4261e-08, 6.7987e-08, 5.3085e-08, ..., 4.1351e-07, + 9.1344e-06, 5.5693e-06], + ..., + [ 1.0058e-07, 6.5193e-09, 9.4064e-08, ..., -6.5193e-09, + 8.3260e-07, 5.0850e-07], + [ 5.0664e-07, 6.5193e-08, 1.8626e-08, ..., -1.3560e-06, + 7.6368e-06, 4.6454e-06], + [ 2.6729e-07, -4.9360e-08, -1.4156e-07, ..., 4.8149e-07, + 4.3847e-06, 2.6952e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0241, -0.0251, -0.0114, -0.0271, -0.0300, 0.0003, 0.0272, -0.0120, + 0.0333, 0.0018], device='cuda:0'), grad: tensor([-1.5879e-04, 5.0187e-05, 1.9193e-05, 4.9025e-06, -1.7295e-06, + 6.6943e-06, 5.7906e-05, 4.0904e-06, 5.0440e-06, 1.2323e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 137---------------------------------------------------- +epoch 137, time 281.77, cls_loss 0.0054 cls_loss_mapping 0.0053 cls_loss_causal 0.5253 re_mapping 0.0078 re_causal 0.0192 /// teacc 99.02 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0790, -0.1023, -0.0624, ..., -0.0425, 0.1374, 0.1303], + [-0.1394, -0.1363, -0.1120, ..., -0.1083, -0.1449, -0.0720], + [-0.0853, -0.0721, 0.1037, ..., -0.1210, 0.1374, 0.0347], + ..., + [-0.0751, 0.0539, 0.0417, ..., 0.1338, -0.1054, -0.1242], + [-0.1797, 0.0501, -0.1093, ..., 0.0388, -0.0690, -0.0795], + [ 0.0385, -0.0646, -0.0613, ..., -0.0540, -0.0412, -0.0600]], + device='cuda:0'), grad: tensor([[ 1.5181e-07, 8.3819e-08, 0.0000e+00, ..., 4.6566e-08, + 5.7183e-07, 1.0803e-07], + [ 6.0536e-08, 3.2503e-07, 0.0000e+00, ..., 1.5367e-07, + 1.9558e-08, 2.7940e-09], + [ 2.0117e-07, 5.6531e-07, 0.0000e+00, ..., 6.9849e-08, + -1.4782e-05, -3.6322e-08], + ..., + [ 7.4506e-08, -1.2927e-06, 9.3132e-10, ..., -1.0096e-06, + 1.0626e-06, 1.2107e-08], + [ 4.8615e-07, 6.4541e-07, 9.3132e-10, ..., 5.0757e-07, + 1.0710e-07, 1.8626e-08], + [-2.0172e-06, 2.1700e-07, 9.3132e-10, ..., 2.8778e-07, + 2.4214e-08, 1.8626e-09]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0240, -0.0255, -0.0106, -0.0277, -0.0312, 0.0007, 0.0274, -0.0123, + 0.0337, 0.0022], device='cuda:0'), grad: tensor([ 1.0617e-06, -1.2415e-06, -1.5259e-05, -3.8184e-08, 7.4565e-05, + 1.1874e-06, 1.1660e-06, 9.6764e-07, 2.4363e-06, -6.4850e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 263.83, cls_loss 0.0054 cls_loss_mapping 0.0053 cls_loss_causal 0.5331 re_mapping 0.0080 re_causal 0.0194 /// teacc 98.85 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0792, -0.1029, -0.0625, ..., -0.0428, 0.1377, 0.1305], + [-0.1396, -0.1371, -0.1119, ..., -0.1088, -0.1461, -0.0735], + [-0.0854, -0.0731, 0.1039, ..., -0.1226, 0.1384, 0.0358], + ..., + [-0.0754, 0.0543, 0.0418, ..., 0.1344, -0.1059, -0.1248], + [-0.1809, 0.0501, -0.1095, ..., 0.0384, -0.0690, -0.0802], + [ 0.0390, -0.0651, -0.0617, ..., -0.0543, -0.0411, -0.0608]], + device='cuda:0'), grad: tensor([[ 5.6550e-06, 2.5844e-07, 1.3132e-07, ..., -1.3344e-05, + -3.1322e-05, -2.1607e-05], + [ 2.3562e-07, 6.8033e-07, 3.2596e-07, ..., 7.7719e-07, + 4.7358e-07, 2.3004e-07], + [ 2.4168e-07, 1.9390e-06, 9.3272e-07, ..., 2.4233e-06, + 1.6056e-06, 1.1344e-06], + ..., + [ 1.2340e-07, -7.1898e-06, -3.2149e-06, ..., 4.6082e-06, + 2.9996e-05, 2.0266e-05], + [ 7.9442e-07, 1.2023e-06, 6.8033e-07, ..., 1.3625e-06, + 6.9803e-07, 3.9814e-07], + [ 7.3574e-07, 3.8510e-07, -4.7311e-07, ..., 7.8883e-07, + 1.6065e-06, 1.0477e-06]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0240, -0.0258, -0.0110, -0.0276, -0.0312, 0.0012, 0.0270, -0.0116, + 0.0332, 0.0023], device='cuda:0'), grad: tensor([-4.2975e-05, -1.8671e-05, 8.2105e-06, 1.3113e-05, 1.9208e-05, + 1.0371e-04, -1.1522e-04, 3.1233e-05, 1.2629e-05, -1.1258e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 263.27, cls_loss 0.0050 cls_loss_mapping 0.0070 cls_loss_causal 0.5478 re_mapping 0.0083 re_causal 0.0203 /// teacc 98.98 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0785, -0.1045, -0.0628, ..., -0.0431, 0.1382, 0.1314], + [-0.1400, -0.1376, -0.1116, ..., -0.1089, -0.1466, -0.0737], + [-0.0858, -0.0735, 0.1039, ..., -0.1228, 0.1398, 0.0380], + ..., + [-0.0756, 0.0528, 0.0420, ..., 0.1344, -0.1067, -0.1254], + [-0.1817, 0.0500, -0.1100, ..., 0.0379, -0.0690, -0.0824], + [ 0.0370, -0.0658, -0.0621, ..., -0.0549, -0.0422, -0.0654]], + device='cuda:0'), grad: tensor([[ 4.7218e-07, 4.0047e-08, 1.2107e-08, ..., 4.4703e-08, + -7.5437e-07, -3.4831e-07], + [ 5.6028e-06, 1.4156e-07, 1.6764e-08, ..., 1.3784e-07, + 4.8429e-08, 2.2352e-08], + [ 1.2089e-06, -7.0035e-07, -2.0023e-07, ..., 9.1270e-08, + -2.8387e-06, -1.6624e-06], + ..., + [ 6.0536e-07, 4.9360e-08, 1.3132e-07, ..., -2.0582e-07, + 2.2780e-06, 1.3821e-06], + [ 2.9709e-07, -1.4603e-05, 7.7300e-08, ..., -2.7671e-05, + 1.7602e-07, 8.4750e-08], + [ 9.1456e-07, 1.3597e-07, -3.4925e-07, ..., -1.4063e-07, + 6.2864e-07, 2.8778e-07]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0236, -0.0255, -0.0103, -0.0261, -0.0313, 0.0005, 0.0271, -0.0121, + 0.0323, 0.0016], device='cuda:0'), grad: tensor([ 2.3246e-06, 2.9668e-05, 3.7216e-06, 3.2112e-06, -1.4794e-04, + 3.1263e-05, 1.0329e-04, 1.3068e-05, -3.3408e-05, -5.4501e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 262.06, cls_loss 0.0055 cls_loss_mapping 0.0068 cls_loss_causal 0.5637 re_mapping 0.0077 re_causal 0.0199 /// teacc 98.94 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0787, -0.1056, -0.0632, ..., -0.0436, 0.1386, 0.1317], + [-0.1406, -0.1381, -0.1116, ..., -0.1097, -0.1470, -0.0738], + [-0.0862, -0.0738, 0.1052, ..., -0.1233, 0.1410, 0.0389], + ..., + [-0.0757, 0.0532, 0.0422, ..., 0.1354, -0.1071, -0.1261], + [-0.1825, 0.0500, -0.1103, ..., 0.0381, -0.0696, -0.0833], + [ 0.0358, -0.0669, -0.0630, ..., -0.0580, -0.0427, -0.0666]], + device='cuda:0'), grad: tensor([[ 8.3167e-07, 1.6764e-08, 4.8429e-08, ..., 9.1176e-07, + -3.0641e-07, -1.2480e-07], + [ 7.4133e-06, 1.5367e-07, 2.6543e-07, ..., 9.5069e-06, + 2.4680e-07, 3.4925e-07], + [ 4.7497e-08, 2.2799e-06, 1.9744e-07, ..., 1.4463e-06, + -1.5192e-05, -2.2367e-05], + ..., + [ 5.2191e-06, -3.0585e-06, 9.7752e-06, ..., 2.4632e-05, + 1.5087e-07, 2.0303e-07], + [ 3.6567e-05, -2.4214e-07, 5.4017e-08, ..., 3.8058e-05, + 1.3217e-05, 2.0444e-05], + [ 5.8524e-06, 1.1921e-07, -1.0528e-05, ..., -1.7256e-05, + 8.6334e-07, 3.3528e-07]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0234, -0.0255, -0.0101, -0.0266, -0.0293, 0.0006, 0.0273, -0.0117, + 0.0323, -0.0001], device='cuda:0'), grad: tensor([ 3.3565e-06, 1.9923e-05, -6.9261e-05, 2.8357e-05, -9.5427e-05, + -1.3566e-04, 7.6443e-06, 1.9360e-04, 1.3924e-04, -9.1791e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 263.44, cls_loss 0.0054 cls_loss_mapping 0.0048 cls_loss_causal 0.5554 re_mapping 0.0076 re_causal 0.0193 /// teacc 98.93 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0788, -0.1059, -0.0634, ..., -0.0438, 0.1388, 0.1321], + [-0.1418, -0.1388, -0.1116, ..., -0.1103, -0.1472, -0.0740], + [-0.0867, -0.0743, 0.1055, ..., -0.1239, 0.1417, 0.0388], + ..., + [-0.0765, 0.0533, 0.0422, ..., 0.1360, -0.1082, -0.1266], + [-0.1832, 0.0500, -0.1102, ..., 0.0382, -0.0698, -0.0835], + [ 0.0349, -0.0676, -0.0632, ..., -0.0583, -0.0430, -0.0674]], + device='cuda:0'), grad: tensor([[ 1.1194e-06, 2.1420e-08, 3.3528e-08, ..., 9.2201e-08, + -7.0315e-07, -3.4086e-07], + [ 6.2399e-08, 1.0524e-07, 2.4177e-06, ..., 3.7979e-06, + -2.0117e-07, 1.5832e-08], + [ 4.5635e-08, -1.1455e-07, 1.5832e-08, ..., 1.9372e-07, + -5.3644e-07, -3.6322e-07], + ..., + [ 4.0978e-08, -3.8091e-07, 7.9535e-07, ..., 7.5065e-07, + 1.2107e-07, 5.3085e-08], + [ 2.4199e-05, 1.2200e-07, 5.4576e-07, ..., 1.8962e-06, + 3.7998e-07, 2.0768e-07], + [-3.5435e-05, 1.0617e-07, -4.2617e-06, ..., -8.2031e-06, + 4.3679e-07, 2.1327e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0235, -0.0255, -0.0102, -0.0273, -0.0293, 0.0016, 0.0275, -0.0121, + 0.0326, -0.0002], device='cuda:0'), grad: tensor([ 4.9360e-06, 6.4194e-05, 3.2391e-06, 2.7895e-05, 1.6242e-05, + 1.5110e-05, 2.8238e-06, 2.6271e-05, 1.2279e-04, -2.8348e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 265.29, cls_loss 0.0048 cls_loss_mapping 0.0051 cls_loss_causal 0.5091 re_mapping 0.0074 re_causal 0.0185 /// teacc 98.97 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0791, -0.1051, -0.0635, ..., -0.0441, 0.1397, 0.1330], + [-0.1421, -0.1400, -0.1117, ..., -0.1116, -0.1486, -0.0756], + [-0.0868, -0.0748, 0.1056, ..., -0.1247, 0.1428, 0.0399], + ..., + [-0.0769, 0.0539, 0.0422, ..., 0.1372, -0.1088, -0.1271], + [-0.1835, 0.0492, -0.1104, ..., 0.0391, -0.0689, -0.0840], + [ 0.0347, -0.0681, -0.0630, ..., -0.0586, -0.0439, -0.0692]], + device='cuda:0'), grad: tensor([[ 3.2652e-06, 5.2154e-08, 1.8626e-09, ..., 1.7695e-08, + 3.2932e-06, 4.2990e-06], + [ 3.6974e-07, 1.7509e-07, 9.3132e-10, ..., 1.3039e-08, + 3.3062e-07, -1.2359e-06], + [ 2.7362e-06, 1.1735e-07, 0.0000e+00, ..., 1.2107e-08, + 3.5074e-06, 3.9078e-06], + ..., + [ 1.5926e-07, 1.6764e-07, 1.8626e-09, ..., -1.2107e-08, + 3.6322e-08, 8.2888e-08], + [ 2.7493e-05, 5.6438e-07, 9.3132e-10, ..., 2.0862e-07, + 3.5048e-05, 3.5971e-05], + [ 5.6345e-07, 2.2165e-07, -5.5879e-09, ..., 4.4703e-08, + 1.0505e-06, 7.5903e-07]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0228, -0.0258, -0.0097, -0.0274, -0.0297, 0.0017, 0.0268, -0.0120, + 0.0330, -0.0006], device='cuda:0'), grad: tensor([ 1.1384e-05, -7.9796e-06, 1.0796e-05, -1.0375e-06, 1.7453e-06, + 1.1690e-05, -1.1635e-04, 1.5255e-06, 8.8692e-05, -5.2433e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 265.37, cls_loss 0.0056 cls_loss_mapping 0.0065 cls_loss_causal 0.5403 re_mapping 0.0076 re_causal 0.0194 /// teacc 98.91 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0791, -0.1051, -0.0637, ..., -0.0442, 0.1398, 0.1331], + [-0.1422, -0.1379, -0.1118, ..., -0.1094, -0.1492, -0.0760], + [-0.0870, -0.0750, 0.1067, ..., -0.1256, 0.1432, 0.0401], + ..., + [-0.0771, 0.0532, 0.0414, ..., 0.1369, -0.1092, -0.1277], + [-0.1853, 0.0496, -0.1105, ..., 0.0386, -0.0692, -0.0845], + [ 0.0345, -0.0693, -0.0629, ..., -0.0600, -0.0441, -0.0697]], + device='cuda:0'), grad: tensor([[ 1.3309e-06, 5.6811e-08, 6.5193e-09, ..., 3.4459e-08, + -5.0701e-06, -3.3099e-06], + [ 1.1045e-06, 2.3935e-07, 1.7695e-08, ..., 2.9150e-07, + 2.5705e-07, -3.3528e-08], + [ 1.4929e-06, 3.2503e-07, 1.8626e-08, ..., 3.4645e-07, + 3.2596e-08, 5.1502e-07], + ..., + [ 2.5611e-06, -1.8235e-06, -7.2643e-08, ..., -1.8906e-06, + 3.6694e-07, 2.3097e-07], + [ 3.2246e-05, 2.1569e-06, 7.7672e-07, ..., 3.1199e-07, + 1.7444e-06, 1.1194e-06], + [-2.1744e-04, -1.6801e-06, -8.4471e-07, ..., 5.6159e-07, + 7.3854e-07, 4.8056e-07]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0227, -0.0229, -0.0097, -0.0278, -0.0294, 0.0021, 0.0269, -0.0146, + 0.0326, -0.0009], device='cuda:0'), grad: tensor([-4.9658e-06, -2.5854e-06, 1.1094e-05, 6.1952e-06, -4.8637e-05, + 3.7789e-04, 1.0258e-04, 9.7156e-06, 1.1081e-04, -5.6124e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 264.97, cls_loss 0.0058 cls_loss_mapping 0.0062 cls_loss_causal 0.5481 re_mapping 0.0075 re_causal 0.0196 /// teacc 99.02 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0791, -0.1058, -0.0643, ..., -0.0443, 0.1402, 0.1336], + [-0.1445, -0.1402, -0.1119, ..., -0.1122, -0.1496, -0.0755], + [-0.0852, -0.0749, 0.1072, ..., -0.1259, 0.1436, 0.0399], + ..., + [-0.0777, 0.0546, 0.0411, ..., 0.1393, -0.1100, -0.1286], + [-0.1866, 0.0493, -0.1108, ..., 0.0382, -0.0695, -0.0851], + [ 0.0344, -0.0698, -0.0630, ..., -0.0609, -0.0446, -0.0712]], + device='cuda:0'), grad: tensor([[ 2.2538e-07, 6.5193e-09, 0.0000e+00, ..., 4.7591e-07, + -3.9414e-06, -1.1176e-06], + [ 1.4156e-07, 4.3772e-08, 0.0000e+00, ..., 3.5204e-07, + 6.5491e-06, 1.1168e-05], + [ 6.7055e-08, 4.1910e-08, 0.0000e+00, ..., 1.3504e-07, + -1.4976e-05, -2.4632e-05], + ..., + [ 6.7335e-07, -5.0291e-08, 0.0000e+00, ..., 3.1404e-06, + 3.1050e-06, 4.6864e-06], + [ 1.6028e-06, -3.5390e-08, 0.0000e+00, ..., 3.2876e-06, + 1.9111e-06, 1.4473e-06], + [ 6.1747e-07, 8.8476e-08, 9.3132e-10, ..., -1.0822e-06, + 2.2780e-06, 2.1793e-06]], device='cuda:0') +Epoch 146, bias, value: tensor([-2.2472e-02, -2.4970e-02, -8.9712e-03, -2.8102e-02, -3.0582e-02, + 1.8346e-03, 2.6861e-02, -1.2489e-02, 3.1929e-02, -9.4486e-05], + device='cuda:0'), grad: tensor([ 5.2787e-06, 6.8784e-05, -1.8716e-04, 2.3007e-05, 2.3156e-05, + -1.1496e-05, 6.8173e-06, 5.8979e-05, 1.4797e-05, -2.1309e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 265.71, cls_loss 0.0063 cls_loss_mapping 0.0062 cls_loss_causal 0.5588 re_mapping 0.0075 re_causal 0.0187 /// teacc 98.99 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0790, -0.1061, -0.0614, ..., -0.0449, 0.1414, 0.1350], + [-0.1451, -0.1404, -0.1126, ..., -0.1130, -0.1500, -0.0757], + [-0.0855, -0.0754, 0.1072, ..., -0.1266, 0.1433, 0.0395], + ..., + [-0.0781, 0.0539, 0.0420, ..., 0.1386, -0.1108, -0.1294], + [-0.1877, 0.0519, -0.1119, ..., 0.0405, -0.0700, -0.0860], + [ 0.0332, -0.0708, -0.0633, ..., -0.0615, -0.0455, -0.0735]], + device='cuda:0'), grad: tensor([[ 3.9209e-07, 4.1444e-07, 1.8626e-09, ..., 1.2666e-07, + -3.9861e-07, -2.9244e-07], + [ 7.0874e-07, 8.5682e-07, 9.3132e-10, ..., 2.3935e-07, + 1.4808e-07, 5.7742e-08], + [ 1.6829e-06, 2.0582e-06, 0.0000e+00, ..., 5.7649e-07, + 8.2236e-07, 1.3225e-07], + ..., + [ 5.0571e-07, 5.3924e-07, 3.7253e-09, ..., -1.7509e-07, + 5.6811e-08, 2.3283e-08], + [ 1.4633e-05, 6.9663e-06, 5.4017e-08, ..., 1.0088e-05, + 5.7183e-06, 2.0992e-06], + [ 6.5006e-07, 9.6112e-07, -7.1712e-08, ..., 1.9465e-07, + 3.5204e-07, 2.1048e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0216, -0.0251, -0.0093, -0.0291, -0.0307, 0.0028, 0.0267, -0.0125, + 0.0328, -0.0006], device='cuda:0'), grad: tensor([ 5.1409e-07, 1.8897e-06, 5.8599e-06, -2.9176e-05, 1.9558e-07, + -8.9407e-05, 6.7174e-05, 1.2247e-06, 3.9548e-05, 2.0340e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 265.32, cls_loss 0.0049 cls_loss_mapping 0.0038 cls_loss_causal 0.5518 re_mapping 0.0077 re_causal 0.0193 /// teacc 98.96 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0795, -0.1072, -0.0614, ..., -0.0455, 0.1417, 0.1353], + [-0.1457, -0.1406, -0.1128, ..., -0.1133, -0.1505, -0.0757], + [-0.0857, -0.0762, 0.1073, ..., -0.1288, 0.1440, 0.0394], + ..., + [-0.0798, 0.0543, 0.0423, ..., 0.1397, -0.1108, -0.1298], + [-0.1882, 0.0518, -0.1126, ..., 0.0403, -0.0701, -0.0863], + [ 0.0335, -0.0713, -0.0626, ..., -0.0620, -0.0458, -0.0744]], + device='cuda:0'), grad: tensor([[ 1.1183e-05, 3.4999e-06, 3.7253e-09, ..., 1.5832e-08, + 1.7090e-06, 5.8711e-06], + [ 1.4873e-06, 5.7630e-06, 9.3132e-10, ..., 1.9558e-08, + 1.5553e-07, 4.6194e-07], + [ 5.6811e-07, 2.7284e-05, 1.8626e-09, ..., 3.3528e-08, + 2.1420e-08, 2.1886e-07], + ..., + [ 2.9244e-07, 1.3141e-06, 1.8626e-09, ..., 1.6764e-08, + 4.0978e-08, 1.0151e-07], + [ 3.0212e-06, 1.1034e-05, -1.8626e-09, ..., 5.1316e-07, + 4.6659e-07, 1.1595e-06], + [ 4.2655e-06, 1.5888e-06, 4.6566e-09, ..., -9.1922e-07, + 8.0373e-07, 2.6561e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0215, -0.0250, -0.0098, -0.0292, -0.0310, 0.0028, 0.0270, -0.0121, + 0.0320, -0.0005], device='cuda:0'), grad: tensor([ 4.3035e-05, 1.1697e-05, 4.2230e-05, -6.6638e-05, 3.9965e-05, + 1.0669e-05, -1.2279e-04, 2.8908e-06, 2.6852e-05, 1.2286e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 265.66, cls_loss 0.0044 cls_loss_mapping 0.0048 cls_loss_causal 0.5414 re_mapping 0.0081 re_causal 0.0198 /// teacc 98.90 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0797, -0.1078, -0.0613, ..., -0.0457, 0.1422, 0.1359], + [-0.1462, -0.1407, -0.1118, ..., -0.1136, -0.1508, -0.0760], + [-0.0858, -0.0773, 0.1081, ..., -0.1293, 0.1445, 0.0397], + ..., + [-0.0798, 0.0545, 0.0418, ..., 0.1401, -0.1112, -0.1303], + [-0.1888, 0.0525, -0.1128, ..., 0.0401, -0.0703, -0.0867], + [ 0.0327, -0.0733, -0.0627, ..., -0.0622, -0.0459, -0.0749]], + device='cuda:0'), grad: tensor([[ 4.1090e-06, 6.7428e-07, 0.0000e+00, ..., 6.8638e-07, + 2.9758e-05, 2.5779e-05], + [ 3.9116e-08, 3.0827e-07, 2.7940e-09, ..., 3.5949e-07, + 1.9185e-07, 1.4622e-07], + [ 3.1665e-08, 1.3039e-07, 4.6566e-09, ..., 1.3411e-07, + 3.5856e-07, 2.4494e-07], + ..., + [ 8.1398e-07, -3.2559e-06, 9.3132e-09, ..., -5.4874e-06, + 3.8184e-08, 2.4214e-08], + [ 2.5295e-06, 1.2852e-07, 0.0000e+00, ..., 3.7476e-06, + 2.7008e-07, 1.4156e-07], + [-2.4959e-06, -1.2293e-06, 2.7940e-09, ..., 2.6990e-06, + 2.8741e-06, 1.3411e-06]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0215, -0.0248, -0.0101, -0.0291, -0.0309, 0.0027, 0.0268, -0.0123, + 0.0331, -0.0010], device='cuda:0'), grad: tensor([ 6.9439e-05, 4.1574e-06, 1.2526e-06, 7.3537e-06, 1.1757e-05, + 4.5933e-06, -7.2062e-05, 9.6392e-07, 4.5970e-06, -3.2187e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 265.60, cls_loss 0.0034 cls_loss_mapping 0.0038 cls_loss_causal 0.5251 re_mapping 0.0077 re_causal 0.0195 /// teacc 98.88 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0799, -0.1082, -0.0613, ..., -0.0460, 0.1423, 0.1360], + [-0.1466, -0.1410, -0.1118, ..., -0.1147, -0.1514, -0.0761], + [-0.0860, -0.0779, 0.1084, ..., -0.1295, 0.1450, 0.0400], + ..., + [-0.0802, 0.0547, 0.0417, ..., 0.1410, -0.1119, -0.1310], + [-0.1893, 0.0525, -0.1129, ..., 0.0401, -0.0705, -0.0869], + [ 0.0324, -0.0737, -0.0627, ..., -0.0626, -0.0460, -0.0751]], + device='cuda:0'), grad: tensor([[ 8.0094e-08, 3.9116e-08, 1.8626e-09, ..., 1.5926e-07, + -1.1422e-05, -5.3421e-06], + [ 8.0094e-08, 9.9465e-07, 9.3132e-10, ..., 4.9509e-06, + 2.5872e-06, 1.2945e-06], + [ 8.6613e-08, 1.6112e-07, 0.0000e+00, ..., 6.1188e-07, + -1.0617e-06, -4.8988e-07], + ..., + [ 2.8871e-08, -6.0648e-06, 0.0000e+00, ..., -3.1352e-05, + 1.0077e-06, 5.1688e-07], + [ 2.6822e-07, 2.6841e-06, 2.7940e-09, ..., 1.4454e-05, + 4.3809e-06, 1.9204e-06], + [ 4.4703e-08, 1.5954e-06, 0.0000e+00, ..., 8.1956e-06, + 3.2745e-06, 1.6131e-06]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0215, -0.0253, -0.0102, -0.0290, -0.0307, 0.0028, 0.0266, -0.0118, + 0.0331, -0.0012], device='cuda:0'), grad: tensor([-1.7419e-05, 1.2405e-05, -2.9299e-06, 2.7288e-06, 9.1735e-07, + 3.2373e-06, -1.0412e-06, -3.7730e-05, 2.3752e-05, 1.6093e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 263.72, cls_loss 0.0047 cls_loss_mapping 0.0052 cls_loss_causal 0.5462 re_mapping 0.0073 re_causal 0.0191 /// teacc 98.95 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0801, -0.1088, -0.0617, ..., -0.0466, 0.1429, 0.1365], + [-0.1501, -0.1416, -0.1120, ..., -0.1155, -0.1518, -0.0771], + [-0.0836, -0.0789, 0.1085, ..., -0.1310, 0.1453, 0.0403], + ..., + [-0.0802, 0.0553, 0.0417, ..., 0.1419, -0.1126, -0.1313], + [-0.1907, 0.0524, -0.1131, ..., 0.0401, -0.0709, -0.0873], + [ 0.0325, -0.0742, -0.0628, ..., -0.0630, -0.0464, -0.0757]], + device='cuda:0'), grad: tensor([[ 3.8464e-07, 1.2191e-06, 1.8626e-09, ..., 1.4156e-07, + -2.8387e-06, -1.5032e-06], + [ 3.5483e-07, 4.2934e-07, 9.3132e-10, ..., 1.4994e-07, + 4.5635e-08, 2.3283e-08], + [ 1.6857e-07, 8.9407e-07, 0.0000e+00, ..., 1.4249e-07, + -2.1048e-07, -4.1910e-08], + ..., + [ 1.4994e-07, -2.3730e-06, 2.7940e-09, ..., -1.6326e-06, + 1.9185e-07, 7.4506e-08], + [ 7.0453e-05, 2.1324e-05, 2.7940e-09, ..., 2.5183e-05, + 2.6263e-07, 1.3877e-07], + [ 1.1595e-06, 1.1660e-06, -1.3039e-08, ..., 4.6007e-07, + 2.0545e-06, 1.0710e-06]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0211, -0.0265, -0.0095, -0.0290, -0.0307, 0.0025, 0.0268, -0.0108, + 0.0329, -0.0013], device='cuda:0'), grad: tensor([-1.3616e-06, -5.3905e-06, 4.8839e-06, 4.3297e-04, -2.5146e-07, + -5.4550e-04, 3.4664e-06, -6.6124e-07, 1.0383e-04, 7.0408e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 260.43, cls_loss 0.0048 cls_loss_mapping 0.0055 cls_loss_causal 0.5582 re_mapping 0.0077 re_causal 0.0192 /// teacc 98.95 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0805, -0.1110, -0.0618, ..., -0.0473, 0.1430, 0.1367], + [-0.1508, -0.1417, -0.1121, ..., -0.1160, -0.1525, -0.0775], + [-0.0833, -0.0798, 0.1092, ..., -0.1323, 0.1466, 0.0406], + ..., + [-0.0813, 0.0555, 0.0418, ..., 0.1425, -0.1142, -0.1319], + [-0.1921, 0.0523, -0.1132, ..., 0.0401, -0.0713, -0.0878], + [ 0.0323, -0.0746, -0.0628, ..., -0.0629, -0.0466, -0.0761]], + device='cuda:0'), grad: tensor([[ 1.0096e-05, 2.5146e-08, 1.6019e-07, ..., 1.3039e-07, + 3.8017e-06, 6.8955e-06], + [ 1.1679e-06, 4.3772e-07, 6.9849e-08, ..., -2.9296e-05, + 6.4541e-07, 9.1642e-07], + [ 6.4857e-06, 1.5646e-07, 5.9605e-08, ..., 9.5461e-07, + 2.9560e-06, 4.9025e-06], + ..., + [ 1.4529e-07, -1.2293e-06, 1.1642e-07, ..., -2.2855e-06, + 1.2666e-07, 1.0896e-07], + [ 2.9817e-05, 1.2666e-07, 3.2596e-08, ..., 1.6809e-05, + 1.2361e-05, 2.1785e-05], + [-4.0904e-06, 4.1816e-07, 1.8626e-07, ..., 1.2992e-06, + 3.4180e-07, 4.8708e-07]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0212, -0.0268, -0.0096, -0.0289, -0.0302, 0.0032, 0.0256, -0.0108, + 0.0325, -0.0013], device='cuda:0'), grad: tensor([ 2.9132e-05, -1.6201e-04, 2.1517e-05, 2.2203e-06, 1.8785e-06, + 2.2721e-04, -2.8348e-04, 2.7880e-05, 1.8239e-04, -4.6670e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 256.46, cls_loss 0.0055 cls_loss_mapping 0.0063 cls_loss_causal 0.4956 re_mapping 0.0075 re_causal 0.0180 /// teacc 98.95 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0809, -0.1117, -0.0620, ..., -0.0487, 0.1433, 0.1371], + [-0.1511, -0.1419, -0.1122, ..., -0.1169, -0.1532, -0.0777], + [-0.0836, -0.0804, 0.1094, ..., -0.1330, 0.1483, 0.0408], + ..., + [-0.0814, 0.0546, 0.0417, ..., 0.1420, -0.1159, -0.1329], + [-0.1929, 0.0542, -0.1134, ..., 0.0420, -0.0713, -0.0883], + [ 0.0322, -0.0757, -0.0636, ..., -0.0627, -0.0470, -0.0769]], + device='cuda:0'), grad: tensor([[1.3721e-04, 1.2293e-07, 2.7940e-09, ..., 9.6977e-05, 6.7532e-05, + 2.9922e-05], + [1.3672e-06, 2.6077e-08, 2.7940e-09, ..., 9.9093e-07, 5.0198e-07, + 1.6578e-07], + [2.7604e-06, 6.6124e-08, 3.7253e-09, ..., 1.4575e-06, 2.1681e-06, + 1.3160e-06], + ..., + [1.4566e-06, 3.3528e-08, 9.3132e-10, ..., 1.1260e-06, 3.0920e-07, + 1.3970e-08], + [7.8678e-06, 5.4948e-08, 3.7253e-09, ..., 2.3190e-07, 1.3709e-05, + 1.1079e-05], + [6.0350e-07, 3.8184e-08, 2.7940e-09, ..., 4.6603e-06, 1.6978e-06, + 1.8347e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0212, -0.0269, -0.0097, -0.0289, -0.0297, 0.0025, 0.0270, -0.0117, + 0.0344, -0.0015], device='cuda:0'), grad: tensor([ 5.3072e-04, 4.8690e-06, 1.1258e-05, 1.5211e-04, -1.6761e-04, + -6.4421e-04, -9.2268e-05, 5.2117e-06, 1.7786e-04, 2.1443e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 256.05, cls_loss 0.0060 cls_loss_mapping 0.0060 cls_loss_causal 0.5058 re_mapping 0.0073 re_causal 0.0180 /// teacc 99.01 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0814, -0.1127, -0.0621, ..., -0.0500, 0.1430, 0.1368], + [-0.1533, -0.1430, -0.1122, ..., -0.1173, -0.1543, -0.0780], + [-0.0843, -0.0812, 0.1094, ..., -0.1342, 0.1503, 0.0407], + ..., + [-0.0833, 0.0551, 0.0418, ..., 0.1415, -0.1173, -0.1334], + [-0.1908, 0.0552, -0.1134, ..., 0.0433, -0.0703, -0.0883], + [ 0.0318, -0.0783, -0.0636, ..., -0.0629, -0.0471, -0.0773]], + device='cuda:0'), grad: tensor([[ 3.5577e-07, 1.1176e-07, 0.0000e+00, ..., 5.0291e-08, + 1.1679e-06, 8.1956e-08], + [ 8.1286e-06, 4.3847e-06, 0.0000e+00, ..., -6.1374e-07, + 1.2666e-07, 4.2841e-08], + [ 7.5530e-07, 1.0906e-06, 0.0000e+00, ..., 1.8068e-07, + -1.0161e-06, -1.9744e-07], + ..., + [ 5.4855e-07, -5.1051e-05, 0.0000e+00, ..., -4.2319e-05, + 2.5798e-07, 4.5635e-08], + [ 5.3272e-07, 2.2892e-06, 0.0000e+00, ..., 1.5004e-06, + 3.1106e-07, 1.2945e-07], + [-7.4618e-06, 2.8372e-05, 0.0000e+00, ..., 2.3976e-05, + 1.8999e-07, 2.6077e-08]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0220, -0.0274, -0.0097, -0.0292, -0.0292, 0.0024, 0.0272, -0.0120, + 0.0367, -0.0018], device='cuda:0'), grad: tensor([ 3.6526e-06, 6.3717e-05, 1.8226e-06, 2.2203e-05, 1.6931e-06, + 5.4613e-06, -1.5926e-06, -7.3493e-05, 6.1356e-06, -2.9609e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 153---------------------------------------------------- +epoch 153, time 272.72, cls_loss 0.0050 cls_loss_mapping 0.0048 cls_loss_causal 0.5256 re_mapping 0.0072 re_causal 0.0188 /// teacc 99.03 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0816, -0.1130, -0.0621, ..., -0.0502, 0.1435, 0.1372], + [-0.1541, -0.1438, -0.1122, ..., -0.1168, -0.1550, -0.0782], + [-0.0840, -0.0817, 0.1089, ..., -0.1356, 0.1507, 0.0407], + ..., + [-0.0839, 0.0552, 0.0428, ..., 0.1418, -0.1182, -0.1339], + [-0.1917, 0.0552, -0.1142, ..., 0.0432, -0.0707, -0.0888], + [ 0.0320, -0.0790, -0.0639, ..., -0.0636, -0.0472, -0.0777]], + device='cuda:0'), grad: tensor([[ 1.9576e-06, 1.8468e-06, 0.0000e+00, ..., 2.7195e-06, + 1.6727e-06, 1.7127e-06], + [ 6.9849e-08, 4.4219e-06, 0.0000e+00, ..., 6.4224e-06, + 4.3772e-08, 4.0978e-08], + [ 1.4808e-07, 2.7195e-07, 0.0000e+00, ..., 3.4180e-07, + 8.3819e-08, 1.0245e-07], + ..., + [ 2.3283e-08, -1.0826e-05, 0.0000e+00, ..., -1.6212e-05, + 1.2107e-08, 9.3132e-09], + [ 2.0489e-07, 4.0419e-07, 0.0000e+00, ..., 2.6263e-07, + 9.4064e-08, 7.4506e-08], + [ 6.6496e-07, 2.5239e-06, 0.0000e+00, ..., 2.0601e-06, + 2.2538e-07, 1.6950e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0219, -0.0274, -0.0098, -0.0290, -0.0291, 0.0023, 0.0271, -0.0119, + 0.0365, -0.0019], device='cuda:0'), grad: tensor([ 1.6868e-05, 1.3704e-03, 4.3064e-06, 3.8370e-06, -1.6575e-03, + 1.2398e-05, -2.4945e-05, -2.5451e-05, 2.7582e-05, 2.7323e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 255.78, cls_loss 0.0056 cls_loss_mapping 0.0056 cls_loss_causal 0.5775 re_mapping 0.0071 re_causal 0.0182 /// teacc 98.95 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0819, -0.1132, -0.0622, ..., -0.0505, 0.1439, 0.1376], + [-0.1543, -0.1439, -0.1123, ..., -0.1167, -0.1555, -0.0784], + [-0.0847, -0.0844, 0.1091, ..., -0.1368, 0.1521, 0.0410], + ..., + [-0.0870, 0.0555, 0.0427, ..., 0.1418, -0.1204, -0.1350], + [-0.1930, 0.0550, -0.1143, ..., 0.0431, -0.0711, -0.0891], + [ 0.0319, -0.0794, -0.0639, ..., -0.0641, -0.0477, -0.0785]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 3.1665e-08, 1.1176e-08, ..., 3.9116e-08, + -1.7762e-05, -1.1437e-05], + [ 7.2643e-08, 2.9709e-07, 1.8626e-09, ..., 4.7777e-07, + 4.9435e-06, 3.1907e-06], + [ 2.0489e-08, -2.5146e-08, 9.3132e-10, ..., 1.1083e-07, + 1.7229e-06, 1.2247e-06], + ..., + [ 1.4901e-08, -2.5705e-07, 0.0000e+00, ..., -1.8347e-07, + 6.3423e-07, 3.9861e-07], + [ 2.2817e-07, -4.1723e-07, 3.7253e-09, ..., -2.3730e-06, + 3.1721e-06, 2.0117e-06], + [ 6.6124e-08, 1.8347e-07, 0.0000e+00, ..., 3.5018e-07, + 1.6531e-06, 1.0058e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0218, -0.0271, -0.0098, -0.0291, -0.0293, 0.0036, 0.0269, -0.0129, + 0.0360, -0.0020], device='cuda:0'), grad: tensor([-2.2978e-05, 1.0997e-05, 1.5255e-06, 2.6345e-05, 3.8296e-06, + 3.8110e-06, 7.9572e-06, 1.1846e-06, -4.2439e-05, 9.7156e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 255.86, cls_loss 0.0056 cls_loss_mapping 0.0051 cls_loss_causal 0.5195 re_mapping 0.0071 re_causal 0.0174 /// teacc 99.01 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0821, -0.1141, -0.0622, ..., -0.0513, 0.1441, 0.1383], + [-0.1548, -0.1439, -0.1126, ..., -0.1168, -0.1560, -0.0784], + [-0.0845, -0.0846, 0.1094, ..., -0.1377, 0.1527, 0.0409], + ..., + [-0.0874, 0.0554, 0.0427, ..., 0.1420, -0.1209, -0.1353], + [-0.1939, 0.0548, -0.1140, ..., 0.0431, -0.0715, -0.0898], + [ 0.0316, -0.0786, -0.0639, ..., -0.0640, -0.0473, -0.0791]], + device='cuda:0'), grad: tensor([[ 3.6974e-07, 4.0419e-07, 0.0000e+00, ..., 5.2154e-08, + -1.2442e-06, -8.6706e-07], + [ 4.7777e-07, 7.2457e-07, 0.0000e+00, ..., 4.4703e-08, + 1.6764e-08, 1.0245e-08], + [ 1.6168e-06, 2.4550e-06, 0.0000e+00, ..., 1.8626e-09, + 5.1223e-08, 2.6077e-08], + ..., + [ 1.2852e-07, 9.3132e-08, 0.0000e+00, ..., -5.4948e-08, + 5.5879e-09, 2.7940e-09], + [ 8.8848e-07, 2.5015e-06, 0.0000e+00, ..., 4.7497e-08, + 3.2224e-07, 1.5646e-07], + [ 1.4715e-07, 2.4959e-07, 0.0000e+00, ..., 9.9745e-07, + 2.4866e-07, 1.6578e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0221, -0.0263, -0.0112, -0.0289, -0.0292, 0.0033, 0.0267, -0.0132, + 0.0357, -0.0006], device='cuda:0'), grad: tensor([ 9.8348e-07, 3.6150e-05, 1.4305e-05, -2.6435e-05, -3.7789e-04, + 1.2722e-06, 1.8682e-06, 4.8801e-06, 6.0648e-06, 3.3832e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 255.52, cls_loss 0.0047 cls_loss_mapping 0.0048 cls_loss_causal 0.5536 re_mapping 0.0071 re_causal 0.0182 /// teacc 98.95 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0826, -0.1148, -0.0622, ..., -0.0516, 0.1441, 0.1384], + [-0.1550, -0.1442, -0.1125, ..., -0.1152, -0.1566, -0.0783], + [-0.0846, -0.0867, 0.1094, ..., -0.1411, 0.1531, 0.0411], + ..., + [-0.0874, 0.0562, 0.0427, ..., 0.1421, -0.1217, -0.1361], + [-0.1965, 0.0547, -0.1135, ..., 0.0428, -0.0723, -0.0901], + [ 0.0326, -0.0794, -0.0639, ..., -0.0650, -0.0474, -0.0795]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, 3.0734e-08, 0.0000e+00, ..., 5.2154e-08, + -5.7183e-06, -2.4606e-06], + [ 2.6077e-08, 2.7288e-07, 0.0000e+00, ..., 4.3493e-07, + 2.3190e-07, 1.1176e-07], + [ 5.2154e-08, 3.8184e-08, 0.0000e+00, ..., 7.8231e-08, + 2.8405e-07, 4.3772e-08], + ..., + [ 6.5193e-09, -2.7902e-06, 0.0000e+00, ..., -4.0568e-06, + 2.5611e-07, 1.5553e-07], + [ 1.3039e-07, 1.1716e-06, 0.0000e+00, ..., 1.2359e-06, + 7.1619e-07, 2.7101e-07], + [ 1.3318e-07, 9.3877e-07, 0.0000e+00, ..., 1.5814e-06, + 1.9167e-06, 8.4192e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0222, -0.0250, -0.0124, -0.0287, -0.0292, 0.0033, 0.0268, -0.0134, + 0.0350, -0.0009], device='cuda:0'), grad: tensor([-7.8753e-06, -2.6631e-04, 1.8287e-04, 6.1616e-06, -1.5143e-06, + 4.2170e-06, 3.4459e-06, 6.5982e-05, 3.9265e-06, 9.1791e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 255.83, cls_loss 0.0058 cls_loss_mapping 0.0050 cls_loss_causal 0.5710 re_mapping 0.0069 re_causal 0.0179 /// teacc 98.98 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0830, -0.1158, -0.0623, ..., -0.0489, 0.1454, 0.1395], + [-0.1553, -0.1447, -0.1126, ..., -0.1156, -0.1576, -0.0786], + [-0.0852, -0.0875, 0.1095, ..., -0.1420, 0.1535, 0.0408], + ..., + [-0.0877, 0.0561, 0.0426, ..., 0.1419, -0.1237, -0.1372], + [-0.1993, 0.0547, -0.1134, ..., 0.0423, -0.0737, -0.0921], + [ 0.0328, -0.0788, -0.0639, ..., -0.0644, -0.0480, -0.0803]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, 1.0245e-08, 3.7253e-09, ..., 5.7742e-08, + -2.2724e-05, -1.1079e-05], + [ 1.1176e-08, 3.0454e-07, 9.3132e-10, ..., 1.1455e-07, + 7.2736e-07, 2.6077e-07], + [ 1.0245e-08, 6.6683e-07, 9.3132e-10, ..., 9.9372e-07, + 4.5225e-06, 1.1427e-06], + ..., + [ 3.6322e-08, 5.9046e-06, 0.0000e+00, ..., -3.7812e-07, + 6.7893e-07, 3.1106e-07], + [ 5.5879e-08, 1.9278e-07, 6.5193e-09, ..., -1.1623e-06, + -3.0547e-06, -1.5646e-07], + [-1.4622e-07, 6.5193e-08, 0.0000e+00, ..., 2.4773e-07, + 1.4424e-05, 6.8322e-06]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0212, -0.0250, -0.0127, -0.0281, -0.0290, 0.0028, 0.0277, -0.0143, + 0.0341, -0.0004], device='cuda:0'), grad: tensor([-3.4481e-05, 1.4612e-06, 1.5467e-05, -4.1798e-06, -3.5092e-06, + 2.8126e-06, 6.0350e-06, 6.2026e-06, -1.3612e-05, 2.3767e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 255.68, cls_loss 0.0061 cls_loss_mapping 0.0055 cls_loss_causal 0.5288 re_mapping 0.0072 re_causal 0.0174 /// teacc 98.93 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0835, -0.1177, -0.0623, ..., -0.0496, 0.1461, 0.1402], + [-0.1579, -0.1449, -0.1128, ..., -0.1158, -0.1588, -0.0790], + [-0.0857, -0.0883, 0.1086, ..., -0.1439, 0.1539, 0.0405], + ..., + [-0.0885, 0.0563, 0.0436, ..., 0.1421, -0.1250, -0.1377], + [-0.1992, 0.0550, -0.1131, ..., 0.0424, -0.0742, -0.0931], + [ 0.0317, -0.0796, -0.0639, ..., -0.0640, -0.0484, -0.0808]], + device='cuda:0'), grad: tensor([[ 4.5635e-08, 3.6322e-08, 0.0000e+00, ..., 2.7940e-08, + -2.7847e-07, -1.1921e-07], + [ 3.3528e-08, 4.3213e-07, 0.0000e+00, ..., 3.9022e-07, + 2.7940e-08, -6.5193e-09], + [ 7.3574e-08, 1.6652e-06, 0.0000e+00, ..., 1.6764e-07, + -6.4448e-07, -1.6671e-07], + ..., + [ 2.8871e-08, 1.1465e-06, 0.0000e+00, ..., -7.5717e-07, + 8.3819e-08, 1.7695e-08], + [ 3.6228e-07, 2.3078e-06, 0.0000e+00, ..., -3.0827e-07, + 3.5204e-07, 1.2387e-07], + [ 2.9244e-07, 2.7847e-07, 0.0000e+00, ..., 5.3085e-07, + 2.4680e-07, 1.0245e-07]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0208, -0.0253, -0.0129, -0.0283, -0.0309, 0.0034, 0.0274, -0.0149, + 0.0352, 0.0008], device='cuda:0'), grad: tensor([-7.6368e-08, -2.8685e-07, 1.6475e-06, -6.1393e-06, -7.3135e-05, + -3.5353e-06, 1.3961e-06, 2.3171e-06, 2.8275e-06, 7.4804e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 255.88, cls_loss 0.0057 cls_loss_mapping 0.0057 cls_loss_causal 0.5416 re_mapping 0.0072 re_causal 0.0175 /// teacc 98.98 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0839, -0.1187, -0.0629, ..., -0.0500, 0.1464, 0.1404], + [-0.1596, -0.1462, -0.1130, ..., -0.1168, -0.1596, -0.0798], + [-0.0843, -0.0887, 0.1087, ..., -0.1447, 0.1546, 0.0413], + ..., + [-0.0888, 0.0568, 0.0437, ..., 0.1428, -0.1262, -0.1386], + [-0.1996, 0.0553, -0.1134, ..., 0.0425, -0.0745, -0.0938], + [ 0.0305, -0.0805, -0.0639, ..., -0.0648, -0.0486, -0.0812]], + device='cuda:0'), grad: tensor([[ 1.5544e-06, 3.1153e-07, 1.4901e-08, ..., 1.3867e-06, + -3.5390e-07, -1.8533e-07], + [ 1.4687e-06, 1.8501e-04, 1.8626e-09, ..., 1.1772e-04, + 3.9581e-08, 6.9849e-09], + [ 4.6492e-06, 6.0424e-06, -1.4063e-07, ..., 7.8455e-06, + -2.6682e-07, 6.1002e-08], + ..., + [ 3.5703e-05, -1.9813e-04, 1.3504e-08, ..., -9.3937e-05, + 2.9290e-07, 1.3039e-08], + [ 7.7039e-06, 5.4725e-06, 1.1455e-07, ..., 9.2164e-06, + 4.7125e-06, 1.2051e-06], + [ 8.4471e-07, -3.0063e-06, 0.0000e+00, ..., -5.2853e-07, + -3.2596e-09, 9.3598e-08]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0209, -0.0262, -0.0129, -0.0288, -0.0307, 0.0037, 0.0279, -0.0140, + 0.0355, 0.0002], device='cuda:0'), grad: tensor([ 6.5416e-06, 6.2370e-04, 3.4720e-05, 4.8757e-05, -1.6332e-05, + -2.2972e-04, 2.7642e-06, -5.4216e-04, 5.1916e-05, 2.0072e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 255.92, cls_loss 0.0057 cls_loss_mapping 0.0050 cls_loss_causal 0.5578 re_mapping 0.0069 re_causal 0.0171 /// teacc 98.96 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0843, -0.1192, -0.0631, ..., -0.0505, 0.1449, 0.1407], + [-0.1603, -0.1468, -0.1131, ..., -0.1180, -0.1600, -0.0800], + [-0.0839, -0.0890, 0.1088, ..., -0.1455, 0.1553, 0.0417], + ..., + [-0.0893, 0.0574, 0.0437, ..., 0.1436, -0.1278, -0.1399], + [-0.2000, 0.0551, -0.1135, ..., 0.0425, -0.0752, -0.0947], + [ 0.0296, -0.0808, -0.0641, ..., -0.0653, -0.0464, -0.0818]], + device='cuda:0'), grad: tensor([[-1.1567e-06, 2.3283e-09, 0.0000e+00, ..., 1.5320e-07, + -5.6848e-06, -1.3392e-06], + [ 9.3132e-09, 2.0023e-08, 0.0000e+00, ..., 1.7229e-07, + 1.2992e-07, -1.2126e-06], + [ 1.6298e-08, 9.3132e-09, 0.0000e+00, ..., 8.8755e-07, + -3.4552e-06, -9.7323e-07], + ..., + [ 6.0536e-09, -4.1816e-07, 0.0000e+00, ..., -8.3121e-07, + 1.9465e-06, 9.7696e-07], + [ 1.0943e-07, 2.6543e-08, 0.0000e+00, ..., -4.2245e-06, + 2.2911e-07, 2.8312e-07], + [ 3.5809e-07, 3.8231e-07, 0.0000e+00, ..., 2.9504e-06, + 2.8815e-06, 1.1837e-06]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0232, -0.0267, -0.0126, -0.0291, -0.0306, 0.0037, 0.0282, -0.0137, + 0.0354, 0.0016], device='cuda:0'), grad: tensor([-6.9961e-06, -1.2510e-05, 4.2170e-06, 2.5574e-06, 5.5414e-08, + 7.8380e-06, 2.4736e-06, 7.0147e-06, -2.7478e-05, 2.2799e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 255.91, cls_loss 0.0042 cls_loss_mapping 0.0045 cls_loss_causal 0.5471 re_mapping 0.0071 re_causal 0.0183 /// teacc 99.00 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0835, -0.1202, -0.0633, ..., -0.0489, 0.1457, 0.1420], + [-0.1605, -0.1468, -0.1131, ..., -0.1174, -0.1628, -0.0815], + [-0.0841, -0.0895, 0.1091, ..., -0.1460, 0.1563, 0.0424], + ..., + [-0.0894, 0.0577, 0.0437, ..., 0.1438, -0.1284, -0.1405], + [-0.2006, 0.0555, -0.1141, ..., 0.0425, -0.0759, -0.0958], + [ 0.0294, -0.0824, -0.0642, ..., -0.0660, -0.0464, -0.0829]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, -1.6764e-07, 0.0000e+00, ..., -3.7067e-07, + -2.5630e-06, -2.2110e-06], + [ 9.3132e-09, 4.9360e-08, 0.0000e+00, ..., -2.3842e-06, + 6.7055e-08, 5.6811e-08], + [ 2.9802e-08, 2.2352e-08, -9.3132e-10, ..., 1.3318e-07, + 2.0489e-07, 2.0862e-07], + ..., + [ 3.7253e-09, -1.3970e-07, 0.0000e+00, ..., 1.7965e-06, + 3.1106e-07, 2.7288e-07], + [ 8.3819e-08, 8.2888e-08, 0.0000e+00, ..., 3.6415e-07, + 9.0525e-07, 7.8790e-07], + [ 2.7940e-08, 7.8231e-08, 0.0000e+00, ..., 3.3993e-07, + 6.6869e-07, 5.4389e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0228, -0.0266, -0.0119, -0.0290, -0.0305, 0.0035, 0.0279, -0.0139, + 0.0360, 0.0011], device='cuda:0'), grad: tensor([-5.5805e-06, -2.0057e-05, 1.2843e-06, 8.4471e-07, -1.8001e-05, + 1.4435e-07, 3.8650e-07, 1.6779e-05, 5.2527e-06, 1.8895e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 255.75, cls_loss 0.0049 cls_loss_mapping 0.0050 cls_loss_causal 0.5682 re_mapping 0.0071 re_causal 0.0178 /// teacc 98.99 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0845, -0.1210, -0.0634, ..., -0.0505, 0.1466, 0.1432], + [-0.1607, -0.1483, -0.1132, ..., -0.1183, -0.1634, -0.0819], + [-0.0842, -0.0893, 0.1092, ..., -0.1463, 0.1568, 0.0428], + ..., + [-0.0898, 0.0584, 0.0437, ..., 0.1445, -0.1294, -0.1414], + [-0.2018, 0.0554, -0.1139, ..., 0.0424, -0.0773, -0.0975], + [ 0.0291, -0.0829, -0.0639, ..., -0.0665, -0.0470, -0.0854]], + device='cuda:0'), grad: tensor([[ 1.7881e-07, 8.2888e-08, 0.0000e+00, ..., -3.9116e-08, + -8.5458e-06, -3.8147e-06], + [ 2.4587e-07, 2.7288e-07, 0.0000e+00, ..., 1.0245e-08, + 7.8231e-08, 1.5832e-08], + [ 4.0140e-07, 3.9674e-07, 0.0000e+00, ..., 9.3132e-09, + 5.6103e-06, 2.2165e-06], + ..., + [ 4.1630e-07, 5.1782e-07, 0.0000e+00, ..., -1.8626e-08, + 1.0431e-07, 5.4948e-08], + [ 3.0361e-07, 3.2317e-07, 0.0000e+00, ..., 5.4948e-08, + 3.7346e-07, 2.3842e-07], + [ 1.2480e-07, 3.2596e-08, 0.0000e+00, ..., -6.3330e-08, + 1.8757e-06, 1.0207e-06]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0223, -0.0273, -0.0117, -0.0283, -0.0295, 0.0030, 0.0278, -0.0131, + 0.0355, 0.0001], device='cuda:0'), grad: tensor([-9.8795e-06, 2.3451e-06, 8.3745e-06, -4.4443e-06, -2.6852e-05, + 1.7481e-06, 1.3011e-06, 2.5611e-06, 4.1351e-06, 2.0683e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 255.80, cls_loss 0.0052 cls_loss_mapping 0.0050 cls_loss_causal 0.5613 re_mapping 0.0070 re_causal 0.0175 /// teacc 98.90 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0850, -0.1221, -0.0634, ..., -0.0515, 0.1469, 0.1435], + [-0.1609, -0.1488, -0.1132, ..., -0.1191, -0.1638, -0.0820], + [-0.0845, -0.0897, 0.1092, ..., -0.1467, 0.1571, 0.0428], + ..., + [-0.0902, 0.0589, 0.0437, ..., 0.1451, -0.1309, -0.1425], + [-0.2035, 0.0554, -0.1140, ..., 0.0425, -0.0770, -0.0984], + [ 0.0314, -0.0833, -0.0639, ..., -0.0668, -0.0472, -0.0861]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.1386e-07, 4.6566e-09, ..., 5.9791e-07, + 2.9057e-07, -1.6764e-08], + [ 2.3283e-08, 1.6922e-06, 1.8626e-09, ..., 1.2321e-06, + 2.5705e-07, 9.3132e-10], + [ 7.4506e-09, 4.1537e-07, 0.0000e+00, ..., 6.6217e-07, + -1.2256e-06, -5.5879e-09], + ..., + [ 1.2107e-08, -2.1338e-05, 5.5879e-09, ..., -1.2100e-05, + 7.3574e-08, 3.7253e-09], + [ 6.5193e-08, 2.3209e-06, 7.4506e-09, ..., 1.1921e-06, + -4.4145e-07, 2.7940e-09], + [ 1.6764e-08, -8.8662e-06, -3.7253e-08, ..., -1.1213e-05, + 2.0396e-07, 9.3132e-09]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0224, -0.0274, -0.0119, -0.0284, -0.0293, 0.0027, 0.0282, -0.0128, + 0.0350, 0.0003], device='cuda:0'), grad: tensor([ 3.7178e-06, 2.0504e-05, -1.6885e-06, 2.8998e-05, 4.6939e-05, + 3.2634e-06, 5.9530e-06, -2.6524e-06, -3.0279e-05, -7.4744e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 255.31, cls_loss 0.0049 cls_loss_mapping 0.0055 cls_loss_causal 0.5246 re_mapping 0.0072 re_causal 0.0173 /// teacc 98.93 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0863, -0.1224, -0.0634, ..., -0.0522, 0.1464, 0.1428], + [-0.1611, -0.1496, -0.1133, ..., -0.1199, -0.1643, -0.0821], + [-0.0847, -0.0897, 0.1093, ..., -0.1470, 0.1580, 0.0431], + ..., + [-0.0905, 0.0593, 0.0440, ..., 0.1459, -0.1325, -0.1435], + [-0.2049, 0.0552, -0.1141, ..., 0.0423, -0.0777, -0.0998], + [ 0.0314, -0.0836, -0.0644, ..., -0.0673, -0.0473, -0.0865]], + device='cuda:0'), grad: tensor([[ 3.1572e-07, 1.8626e-08, 1.8626e-09, ..., -8.2888e-08, + -9.4399e-06, -5.4389e-06], + [ 5.6345e-07, 7.2643e-08, 2.4214e-08, ..., -1.4104e-05, + 6.1654e-07, 2.1514e-07], + [ 3.7253e-08, 2.3283e-08, -1.1269e-07, ..., 1.1548e-07, + -2.5034e-06, -2.8219e-07], + ..., + [ 1.7695e-07, 2.5146e-08, 2.7940e-09, ..., 8.5980e-06, + 2.3451e-06, 1.2098e-06], + [ 1.4797e-05, 4.4778e-06, 2.7940e-09, ..., 7.8604e-06, + 1.0449e-06, 4.4797e-07], + [ 1.5711e-06, 4.8708e-07, 0.0000e+00, ..., 3.4776e-06, + 3.8706e-06, 2.0582e-06]], device='cuda:0') +Epoch 166, bias, value: tensor([-2.2879e-02, -2.8045e-02, -1.1545e-02, -2.8122e-02, -2.8689e-02, + 2.4492e-03, 2.9381e-02, -1.2313e-02, 3.4289e-02, -4.0254e-05], + device='cuda:0'), grad: tensor([-1.5199e-05, -6.5625e-05, -7.2345e-06, 2.0790e-04, 1.7911e-05, + -2.4724e-04, 2.2426e-05, 4.4525e-05, 2.2337e-05, 2.0340e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 255.64, cls_loss 0.0049 cls_loss_mapping 0.0060 cls_loss_causal 0.5394 re_mapping 0.0069 re_causal 0.0173 /// teacc 98.95 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0867, -0.1232, -0.0635, ..., -0.0524, 0.1461, 0.1432], + [-0.1620, -0.1499, -0.1133, ..., -0.1202, -0.1652, -0.0827], + [-0.0844, -0.0895, 0.1093, ..., -0.1483, 0.1586, 0.0432], + ..., + [-0.0907, 0.0593, 0.0440, ..., 0.1462, -0.1344, -0.1453], + [-0.2054, 0.0558, -0.1142, ..., 0.0424, -0.0777, -0.1002], + [ 0.0313, -0.0847, -0.0638, ..., -0.0675, -0.0467, -0.0872]], + device='cuda:0'), grad: tensor([[ 4.3064e-06, 7.3295e-07, 3.7253e-09, ..., 6.7055e-08, + -5.3365e-07, 5.0254e-06], + [ 4.1723e-07, 8.3148e-06, 4.6007e-06, ..., 2.3127e-05, + 3.9022e-07, -7.6964e-06], + [ 9.8627e-07, 4.1537e-07, 1.2945e-07, ..., 6.7335e-07, + -1.1958e-06, 2.5574e-06], + ..., + [ 1.0617e-07, -1.1459e-05, -6.1020e-06, ..., -3.1531e-05, + 3.9488e-07, 2.6822e-07], + [-1.5333e-05, -2.6189e-06, 9.2201e-08, ..., 1.5832e-07, + -3.6377e-06, -1.7270e-05], + [ 7.9349e-06, 3.5893e-06, 1.1418e-06, ..., 7.1265e-06, + 1.9874e-06, 1.0177e-05]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0236, -0.0279, -0.0124, -0.0284, -0.0289, 0.0027, 0.0296, -0.0124, + 0.0354, 0.0002], device='cuda:0'), grad: tensor([ 5.4359e-05, -2.3425e-05, 2.6256e-05, 1.7196e-05, 4.0978e-06, + -5.5820e-05, 1.0389e-04, -8.0049e-05, -1.6344e-04, 1.1677e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 255.56, cls_loss 0.0047 cls_loss_mapping 0.0044 cls_loss_causal 0.5702 re_mapping 0.0070 re_causal 0.0177 /// teacc 98.94 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0877, -0.1234, -0.0635, ..., -0.0526, 0.1465, 0.1435], + [-0.1622, -0.1501, -0.1135, ..., -0.1200, -0.1661, -0.0833], + [-0.0846, -0.0907, 0.1096, ..., -0.1492, 0.1585, 0.0430], + ..., + [-0.0910, 0.0594, 0.0441, ..., 0.1462, -0.1354, -0.1464], + [-0.2056, 0.0559, -0.1143, ..., 0.0424, -0.0785, -0.1007], + [ 0.0302, -0.0851, -0.0639, ..., -0.0676, -0.0469, -0.0894]], + device='cuda:0'), grad: tensor([[-1.0513e-05, 1.6205e-07, 0.0000e+00, ..., -3.2395e-05, + -1.7047e-04, -8.1122e-05], + [ 7.0222e-07, 6.7707e-07, -2.3749e-07, ..., 6.2734e-06, + 3.9767e-07, 1.9558e-07], + [ 1.1278e-06, -5.9567e-06, 6.5193e-08, ..., -4.2394e-06, + 2.0787e-06, -9.1456e-07], + ..., + [ 3.3621e-07, 3.0790e-06, 7.4506e-09, ..., 3.9823e-06, + 5.2825e-06, 3.7458e-06], + [ 1.0103e-05, 1.8347e-07, 2.2352e-08, ..., -4.4852e-06, + 1.1340e-05, 4.5411e-06], + [ 2.1812e-06, 4.0047e-07, 9.3132e-10, ..., 2.4028e-07, + 1.0602e-05, 4.1686e-06]], device='cuda:0') +Epoch 168, bias, value: tensor([-2.3524e-02, -2.7244e-02, -1.3227e-02, -2.8207e-02, -2.9483e-02, + 2.8143e-03, 3.0259e-02, -1.2617e-02, 3.5282e-02, 9.3513e-05], + device='cuda:0'), grad: tensor([-2.3413e-04, 1.7732e-05, -2.2426e-05, 1.7166e-05, 6.7502e-06, + -1.2064e-04, 2.9731e-04, 2.1189e-05, 1.7472e-06, 1.5303e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 255.64, cls_loss 0.0051 cls_loss_mapping 0.0053 cls_loss_causal 0.5374 re_mapping 0.0069 re_causal 0.0170 /// teacc 99.01 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0875, -0.1238, -0.0636, ..., -0.0515, 0.1478, 0.1450], + [-0.1638, -0.1498, -0.1114, ..., -0.1199, -0.1671, -0.0839], + [-0.0836, -0.0903, 0.1099, ..., -0.1492, 0.1596, 0.0437], + ..., + [-0.0916, 0.0593, 0.0430, ..., 0.1465, -0.1372, -0.1481], + [-0.2061, 0.0559, -0.1144, ..., 0.0423, -0.0791, -0.1012], + [ 0.0297, -0.0857, -0.0652, ..., -0.0680, -0.0472, -0.0902]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 2.7344e-06, 9.3132e-10, ..., 1.4435e-07, + 6.1803e-06, 2.7437e-06], + [ 9.2201e-08, 4.5449e-07, 1.0245e-08, ..., 9.9745e-07, + 1.1642e-07, 1.9558e-08], + [ 1.9558e-08, -6.9924e-06, 3.7253e-09, ..., 4.6659e-07, + -1.6779e-05, -7.6666e-06], + ..., + [ 1.8626e-08, -3.4198e-06, -8.2888e-08, ..., -1.2457e-05, + 5.4650e-06, 2.4363e-06], + [ 9.6019e-07, 2.1588e-06, 2.7940e-09, ..., 3.5018e-07, + 4.1537e-06, 1.6047e-06], + [ 1.2945e-07, 4.9695e-06, 6.2399e-08, ..., 1.0058e-05, + 1.5926e-07, 7.1712e-08]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0226, -0.0258, -0.0125, -0.0281, -0.0294, 0.0041, 0.0272, -0.0136, + 0.0351, -0.0005], device='cuda:0'), grad: tensor([ 1.6421e-05, 7.0687e-07, -4.2945e-05, -1.7509e-05, 1.2973e-06, + 2.2471e-05, -7.1712e-06, -8.9556e-06, 1.6347e-05, 1.9342e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 255.63, cls_loss 0.0045 cls_loss_mapping 0.0044 cls_loss_causal 0.5680 re_mapping 0.0069 re_causal 0.0175 /// teacc 98.94 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0875, -0.1242, -0.0640, ..., -0.0515, 0.1481, 0.1454], + [-0.1640, -0.1501, -0.1118, ..., -0.1201, -0.1675, -0.0835], + [-0.0838, -0.0909, 0.1098, ..., -0.1499, 0.1597, 0.0428], + ..., + [-0.0918, 0.0595, 0.0429, ..., 0.1479, -0.1385, -0.1488], + [-0.2064, 0.0561, -0.1128, ..., 0.0426, -0.0790, -0.1012], + [ 0.0287, -0.0867, -0.0651, ..., -0.0696, -0.0474, -0.0910]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 1.3970e-08, 0.0000e+00, ..., 4.1910e-08, + 1.3690e-07, -1.8813e-07], + [ 8.1956e-08, 8.8476e-08, 0.0000e+00, ..., 1.2014e-07, + 4.2468e-07, 1.5460e-07], + [ 8.3819e-09, 3.3993e-07, 0.0000e+00, ..., 5.0012e-07, + -3.3304e-06, -1.5274e-06], + ..., + [ 1.8626e-09, -9.2946e-07, 0.0000e+00, ..., -1.0803e-06, + 7.5810e-07, 3.6694e-07], + [ 4.8429e-08, 1.3039e-08, 0.0000e+00, ..., -2.1514e-07, + 8.0001e-07, 4.8988e-07], + [ 7.4506e-09, 6.4261e-08, 0.0000e+00, ..., 8.4750e-08, + 7.6834e-07, 4.9360e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0225, -0.0256, -0.0130, -0.0280, -0.0303, 0.0044, 0.0269, -0.0126, + 0.0356, -0.0014], device='cuda:0'), grad: tensor([ 8.7358e-07, -1.5618e-06, -3.2857e-06, 1.5786e-06, -3.0547e-07, + 5.7090e-07, -6.6310e-07, -6.3889e-07, 1.6028e-06, 1.7798e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 255.32, cls_loss 0.0036 cls_loss_mapping 0.0036 cls_loss_causal 0.5091 re_mapping 0.0068 re_causal 0.0176 /// teacc 99.01 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0877, -0.1244, -0.0640, ..., -0.0534, 0.1483, 0.1456], + [-0.1659, -0.1507, -0.1121, ..., -0.1208, -0.1682, -0.0848], + [-0.0816, -0.0912, 0.1098, ..., -0.1508, 0.1605, 0.0439], + ..., + [-0.0920, 0.0599, 0.0433, ..., 0.1485, -0.1382, -0.1495], + [-0.2068, 0.0561, -0.1129, ..., 0.0425, -0.0793, -0.1016], + [ 0.0285, -0.0868, -0.0651, ..., -0.0701, -0.0475, -0.0913]], + device='cuda:0'), grad: tensor([[ 4.0792e-07, 1.4901e-08, 0.0000e+00, ..., 1.0151e-07, + -9.0338e-07, -5.5972e-07], + [ 4.9360e-08, 5.8673e-08, 0.0000e+00, ..., 8.6613e-08, + 3.5390e-08, 1.9558e-08], + [ 2.2352e-08, 3.5390e-08, 0.0000e+00, ..., 1.0524e-07, + 8.7544e-08, 5.8673e-08], + ..., + [ 1.3039e-07, -1.7229e-07, 0.0000e+00, ..., -5.9605e-08, + 8.3819e-09, 3.7253e-09], + [ 5.5600e-07, -2.1420e-08, 0.0000e+00, ..., 5.5879e-08, + 4.6007e-07, 2.6077e-07], + [ 3.0268e-07, 2.6077e-08, 0.0000e+00, ..., -1.4650e-06, + 1.0030e-06, 6.0536e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0225, -0.0267, -0.0115, -0.0280, -0.0299, 0.0045, 0.0267, -0.0122, + 0.0354, -0.0016], device='cuda:0'), grad: tensor([ 2.4680e-07, -2.1982e-04, 1.2898e-04, 1.2383e-05, 3.9160e-05, + -1.9968e-06, -1.5032e-06, 8.0228e-05, 3.9339e-06, -4.1038e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 255.42, cls_loss 0.0040 cls_loss_mapping 0.0048 cls_loss_causal 0.5591 re_mapping 0.0067 re_causal 0.0173 /// teacc 98.95 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0878, -0.1247, -0.0642, ..., -0.0531, 0.1490, 0.1463], + [-0.1661, -0.1519, -0.1112, ..., -0.1220, -0.1685, -0.0853], + [-0.0816, -0.0913, 0.1096, ..., -0.1504, 0.1615, 0.0446], + ..., + [-0.0923, 0.0603, 0.0431, ..., 0.1491, -0.1409, -0.1515], + [-0.2075, 0.0560, -0.1123, ..., 0.0424, -0.0795, -0.1022], + [ 0.0276, -0.0873, -0.0666, ..., -0.0702, -0.0482, -0.0922]], + device='cuda:0'), grad: tensor([[ 3.1386e-07, 6.5193e-09, 9.3132e-10, ..., 8.1956e-08, + -1.8179e-06, -3.0492e-06], + [ 4.7497e-08, 7.3574e-08, 5.5879e-09, ..., 9.8720e-08, + 3.6322e-07, 1.6298e-07], + [ 2.4214e-08, 1.9558e-08, 0.0000e+00, ..., 1.5553e-07, + -2.8461e-06, 1.6950e-07], + ..., + [ 7.4506e-09, 1.0245e-08, 1.4901e-08, ..., 1.6158e-06, + 2.0582e-07, 1.1455e-07], + [ 1.4342e-07, 1.5832e-08, 1.8626e-09, ..., 6.2212e-07, + 8.1770e-07, 5.4669e-07], + [ 1.3970e-08, -3.6694e-07, -6.4261e-08, ..., -4.9882e-06, + 4.2841e-07, 2.8033e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0218, -0.0270, -0.0115, -0.0278, -0.0305, 0.0042, 0.0269, -0.0119, + 0.0352, -0.0019], device='cuda:0'), grad: tensor([ 6.2063e-06, 3.7365e-06, 3.2596e-06, 2.2888e-05, 1.1913e-05, + 7.4394e-06, -1.4855e-06, 1.0937e-05, -3.6329e-05, -2.8595e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 255.39, cls_loss 0.0045 cls_loss_mapping 0.0057 cls_loss_causal 0.5226 re_mapping 0.0067 re_causal 0.0173 /// teacc 98.88 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0880, -0.1250, -0.0643, ..., -0.0532, 0.1496, 0.1471], + [-0.1666, -0.1521, -0.1112, ..., -0.1225, -0.1692, -0.0857], + [-0.0819, -0.0918, 0.1096, ..., -0.1511, 0.1615, 0.0444], + ..., + [-0.0926, 0.0604, 0.0431, ..., 0.1495, -0.1413, -0.1519], + [-0.2083, 0.0560, -0.1115, ..., 0.0424, -0.0801, -0.1031], + [ 0.0265, -0.0867, -0.0663, ..., -0.0700, -0.0483, -0.0934]], + device='cuda:0'), grad: tensor([[ 3.6303e-06, 2.0489e-08, 0.0000e+00, ..., 8.1025e-07, + 2.4848e-06, 2.0843e-06], + [ 2.3283e-07, 9.9652e-08, 0.0000e+00, ..., 5.1223e-08, + 4.2748e-07, 1.0617e-07], + [ 2.6822e-07, 2.4866e-07, -0.0000e+00, ..., -6.7428e-07, + -8.7544e-07, -8.0187e-07], + ..., + [ 4.0978e-08, -3.7253e-08, 0.0000e+00, ..., 3.2783e-07, + 6.2026e-07, 4.8615e-07], + [ 2.6338e-06, 1.2945e-07, 0.0000e+00, ..., 7.4413e-07, + 1.3206e-06, 1.4920e-06], + [ 1.9111e-06, 2.2631e-07, 0.0000e+00, ..., 4.8708e-07, + 1.3411e-07, 8.8476e-08]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0215, -0.0271, -0.0120, -0.0264, -0.0313, 0.0030, 0.0275, -0.0120, + 0.0349, -0.0013], device='cuda:0'), grad: tensor([ 8.4639e-06, 4.7944e-06, -6.0908e-07, 2.5228e-05, -9.9614e-06, + -1.3523e-05, -2.5526e-05, 1.7723e-06, -5.6718e-07, 9.9391e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 255.89, cls_loss 0.0044 cls_loss_mapping 0.0045 cls_loss_causal 0.5408 re_mapping 0.0069 re_causal 0.0171 /// teacc 99.02 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0882, -0.1256, -0.0646, ..., -0.0530, 0.1499, 0.1474], + [-0.1668, -0.1525, -0.1104, ..., -0.1228, -0.1696, -0.0840], + [-0.0821, -0.0919, 0.1097, ..., -0.1514, 0.1620, 0.0439], + ..., + [-0.0928, 0.0605, 0.0440, ..., 0.1500, -0.1418, -0.1541], + [-0.2077, 0.0560, -0.1128, ..., 0.0430, -0.0790, -0.1037], + [ 0.0262, -0.0868, -0.0682, ..., -0.0704, -0.0484, -0.0941]], + device='cuda:0'), grad: tensor([[ 6.7987e-08, 2.9802e-08, 0.0000e+00, ..., 4.6566e-09, + -8.9407e-08, -6.2399e-08], + [ 4.0047e-08, 3.5390e-08, 0.0000e+00, ..., 1.3970e-08, + 2.7940e-08, 1.8626e-09], + [ 1.2200e-07, 8.8476e-08, -4.6566e-09, ..., 2.5146e-08, + 8.3819e-08, 3.7253e-09], + ..., + [ 5.7742e-08, -2.7940e-09, 9.3132e-10, ..., -6.6124e-08, + 8.3819e-09, 3.7253e-09], + [ 6.5304e-06, 3.6974e-07, 4.6566e-09, ..., 8.3819e-09, + -2.5891e-07, 9.3132e-10], + [-5.6997e-06, -3.3528e-08, -0.0000e+00, ..., 2.7940e-09, + 6.6124e-08, 3.6322e-08]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0215, -0.0263, -0.0125, -0.0269, -0.0314, 0.0031, 0.0275, -0.0119, + 0.0357, -0.0018], device='cuda:0'), grad: tensor([ 5.6811e-08, -1.0338e-07, 5.0105e-07, -5.4166e-06, 1.6456e-06, + 4.0829e-06, 5.0105e-07, 7.6741e-07, 3.1739e-05, -3.3736e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 255.60, cls_loss 0.0046 cls_loss_mapping 0.0043 cls_loss_causal 0.5481 re_mapping 0.0067 re_causal 0.0173 /// teacc 98.96 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0884, -0.1261, -0.0646, ..., -0.0531, 0.1503, 0.1478], + [-0.1670, -0.1529, -0.1105, ..., -0.1240, -0.1699, -0.0841], + [-0.0824, -0.0921, 0.1100, ..., -0.1514, 0.1629, 0.0446], + ..., + [-0.0930, 0.0597, 0.0442, ..., 0.1503, -0.1429, -0.1561], + [-0.2066, 0.0559, -0.1130, ..., 0.0436, -0.0777, -0.1043], + [ 0.0262, -0.0868, -0.0683, ..., -0.0703, -0.0485, -0.0948]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.7416e-07, 1.8626e-09, ..., 2.1700e-07, + -3.6418e-05, -2.0429e-05], + [ 6.5193e-09, 9.3598e-07, 1.8626e-09, ..., 1.2787e-06, + 3.5390e-07, 2.3935e-07], + [ 9.3132e-10, 2.2903e-05, -1.1455e-07, ..., 3.0905e-05, + -6.2771e-07, -1.0999e-06], + ..., + [ 5.5879e-09, -3.0175e-05, 9.3132e-09, ..., -4.0919e-05, + 4.1910e-07, 6.4727e-07], + [ 1.0245e-08, 2.5798e-06, 8.6613e-08, ..., 3.1982e-06, + 2.7940e-07, 1.9651e-07], + [ 3.7253e-09, 2.1532e-06, 0.0000e+00, ..., 3.1870e-06, + 9.4809e-07, 4.5262e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0212, -0.0265, -0.0125, -0.0264, -0.0314, 0.0029, 0.0267, -0.0123, + 0.0366, -0.0016], device='cuda:0'), grad: tensor([-5.9098e-05, -1.2025e-05, 6.0588e-05, 3.7402e-06, 1.1977e-06, + 4.0866e-06, 5.7340e-05, -7.1824e-05, 6.5491e-06, 9.3728e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 255.61, cls_loss 0.0062 cls_loss_mapping 0.0039 cls_loss_causal 0.5438 re_mapping 0.0069 re_causal 0.0169 /// teacc 98.95 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0892, -0.1270, -0.0638, ..., -0.0536, 0.1508, 0.1484], + [-0.1673, -0.1533, -0.1091, ..., -0.1241, -0.1717, -0.0849], + [-0.0826, -0.0932, 0.1103, ..., -0.1533, 0.1637, 0.0446], + ..., + [-0.0938, 0.0568, 0.0434, ..., 0.1497, -0.1436, -0.1565], + [-0.2078, 0.0567, -0.1132, ..., 0.0441, -0.0780, -0.1048], + [ 0.0255, -0.0844, -0.0689, ..., -0.0690, -0.0487, -0.0956]], + device='cuda:0'), grad: tensor([[ 4.7404e-07, 1.7695e-08, 9.3132e-10, ..., 7.4320e-07, + -2.3019e-04, -1.1796e-04], + [ 8.7544e-07, 1.4175e-06, 9.3132e-10, ..., 2.3153e-06, + 1.9725e-06, 1.0133e-06], + [ 8.5682e-07, 2.3562e-07, 9.3132e-10, ..., 1.4966e-06, + 4.6473e-07, 2.9150e-07], + ..., + [ 2.7567e-07, -7.8604e-06, 1.8626e-09, ..., -5.6438e-06, + 4.6939e-07, 2.3749e-07], + [ 5.6297e-05, 1.3718e-06, 9.3132e-10, ..., 8.3029e-05, + 1.1753e-06, 8.0001e-07], + [ 1.2945e-07, 8.0094e-08, 1.3039e-08, ..., 5.4389e-07, + 2.9784e-06, 1.6205e-06]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0211, -0.0265, -0.0130, -0.0257, -0.0318, 0.0038, 0.0261, -0.0149, + 0.0370, 0.0001], device='cuda:0'), grad: tensor([-3.0088e-04, -5.4762e-06, 3.9861e-06, 1.2815e-05, 4.5225e-06, + -4.0126e-04, 5.1498e-04, -1.3299e-05, 1.8358e-04, 1.6047e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 255.66, cls_loss 0.0048 cls_loss_mapping 0.0036 cls_loss_causal 0.5312 re_mapping 0.0067 re_causal 0.0167 /// teacc 98.94 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0902, -0.1277, -0.0638, ..., -0.0543, 0.1505, 0.1487], + [-0.1677, -0.1535, -0.1087, ..., -0.1242, -0.1723, -0.0853], + [-0.0827, -0.0934, 0.1109, ..., -0.1534, 0.1649, 0.0455], + ..., + [-0.0941, 0.0588, 0.0428, ..., 0.1518, -0.1443, -0.1573], + [-0.2084, 0.0567, -0.1133, ..., 0.0442, -0.0782, -0.1050], + [ 0.0249, -0.0881, -0.0701, ..., -0.0723, -0.0480, -0.0964]], + device='cuda:0'), grad: tensor([[ 9.4064e-08, 4.0978e-08, 0.0000e+00, ..., 1.6764e-08, + -8.1509e-06, -3.1441e-06], + [ 4.2841e-08, 6.4261e-08, 9.3132e-10, ..., 2.9802e-08, + 1.8254e-07, 3.9116e-08], + [ 1.0245e-07, 2.0489e-08, 0.0000e+00, ..., 1.3039e-08, + -3.1851e-07, -2.0675e-07], + ..., + [ 2.6077e-08, -2.9989e-07, 2.7940e-09, ..., -3.1106e-07, + 2.3935e-07, 1.4529e-07], + [ 3.3621e-07, 2.7195e-07, 9.3132e-10, ..., 6.7987e-08, + 9.7789e-08, 5.6811e-08], + [ 9.4064e-08, 6.7987e-08, -0.0000e+00, ..., 6.8918e-08, + 7.3910e-06, 2.8070e-06]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0218, -0.0262, -0.0127, -0.0262, -0.0317, 0.0041, 0.0262, -0.0132, + 0.0372, -0.0016], device='cuda:0'), grad: tensor([-1.5132e-05, 8.2143e-07, -3.2671e-06, 7.8417e-07, -1.0189e-06, + -1.1167e-06, 1.7295e-06, 4.6194e-07, 1.0598e-06, 1.5676e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 256.12, cls_loss 0.0052 cls_loss_mapping 0.0054 cls_loss_causal 0.5361 re_mapping 0.0068 re_causal 0.0167 /// teacc 99.00 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0909, -0.1283, -0.0642, ..., -0.0541, 0.1508, 0.1489], + [-0.1680, -0.1537, -0.1088, ..., -0.1240, -0.1744, -0.0859], + [-0.0829, -0.0933, 0.1124, ..., -0.1518, 0.1684, 0.0475], + ..., + [-0.0944, 0.0590, 0.0445, ..., 0.1520, -0.1485, -0.1613], + [-0.2066, 0.0566, -0.1138, ..., 0.0455, -0.0763, -0.1063], + [ 0.0248, -0.0885, -0.0729, ..., -0.0739, -0.0485, -0.0979]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, 1.6764e-08, 0.0000e+00, ..., 2.6077e-08, + -9.3132e-08, -4.9360e-08], + [ 2.6636e-07, 2.0768e-07, 1.8626e-09, ..., 6.3330e-08, + 1.5832e-08, 1.3039e-08], + [ 1.0990e-07, 8.0094e-08, -1.8626e-09, ..., 2.4214e-08, + -1.0505e-06, -8.5402e-07], + ..., + [ 9.8161e-07, 1.1539e-06, -2.7940e-09, ..., -2.5146e-08, + 1.0030e-06, 8.0653e-07], + [ 2.9951e-06, 2.4214e-07, 0.0000e+00, ..., 1.3681e-06, + 4.2841e-08, 2.5146e-08], + [ 1.6298e-07, 1.5646e-07, 9.3132e-10, ..., 4.8429e-08, + 5.3085e-08, 2.7940e-08]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0219, -0.0260, -0.0121, -0.0276, -0.0304, 0.0044, 0.0264, -0.0132, + 0.0387, -0.0026], device='cuda:0'), grad: tensor([ 1.2573e-07, 8.4285e-07, -1.9781e-06, -4.1053e-06, -8.4788e-06, + -2.2754e-05, 1.8626e-05, 6.0573e-06, 5.9009e-06, 5.7817e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 255.46, cls_loss 0.0047 cls_loss_mapping 0.0042 cls_loss_causal 0.5499 re_mapping 0.0068 re_causal 0.0168 /// teacc 99.02 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0911, -0.1290, -0.0641, ..., -0.0543, 0.1514, 0.1495], + [-0.1688, -0.1560, -0.1087, ..., -0.1256, -0.1756, -0.0869], + [-0.0822, -0.0944, 0.1112, ..., -0.1524, 0.1699, 0.0489], + ..., + [-0.0946, 0.0592, 0.0462, ..., 0.1525, -0.1497, -0.1625], + [-0.2068, 0.0565, -0.1143, ..., 0.0456, -0.0766, -0.1071], + [ 0.0250, -0.0879, -0.0737, ..., -0.0735, -0.0489, -0.0990]], + device='cuda:0'), grad: tensor([[ 1.6019e-07, 1.2387e-07, 8.3819e-09, ..., 1.2480e-07, + 2.3190e-07, 9.5926e-08], + [ 1.3877e-07, 1.2657e-06, 1.8626e-09, ..., 1.8775e-06, + -2.7940e-08, 1.0245e-08], + [ 1.5646e-07, 3.2131e-07, -2.9709e-07, ..., 3.6508e-07, + -2.2538e-07, -2.4214e-07], + ..., + [ 1.6764e-08, -3.5577e-06, 1.1176e-08, ..., -4.7833e-06, + 1.0617e-07, 8.7544e-08], + [ 1.7602e-07, 3.1199e-07, 2.6822e-07, ..., 4.4890e-07, + 4.1630e-07, 2.5053e-07], + [ 9.3132e-09, 1.2442e-06, -0.0000e+00, ..., 1.2415e-06, + 4.2841e-08, 1.8626e-08]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0215, -0.0280, -0.0111, -0.0275, -0.0305, 0.0043, 0.0261, -0.0134, + 0.0388, -0.0017], device='cuda:0'), grad: tensor([ 9.4809e-07, 2.0694e-06, -5.7742e-07, 2.9709e-07, 9.6671e-07, + 6.1654e-07, -1.7369e-06, -8.1211e-06, 3.0287e-06, 2.4959e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 256.06, cls_loss 0.0040 cls_loss_mapping 0.0037 cls_loss_causal 0.5314 re_mapping 0.0067 re_causal 0.0169 /// teacc 98.94 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0914, -0.1296, -0.0642, ..., -0.0546, 0.1516, 0.1497], + [-0.1688, -0.1562, -0.1086, ..., -0.1256, -0.1759, -0.0870], + [-0.0823, -0.0954, 0.1111, ..., -0.1534, 0.1704, 0.0492], + ..., + [-0.0947, 0.0596, 0.0463, ..., 0.1528, -0.1506, -0.1631], + [-0.2070, 0.0564, -0.1128, ..., 0.0456, -0.0763, -0.1073], + [ 0.0251, -0.0880, -0.0736, ..., -0.0736, -0.0491, -0.0996]], + device='cuda:0'), grad: tensor([[ 5.1223e-08, 3.8184e-08, 9.3132e-10, ..., 2.4214e-08, + -5.6624e-07, -1.4696e-06], + [ 1.6764e-08, 1.7397e-06, 1.1176e-08, ..., 1.1390e-06, + 9.1046e-06, 1.9781e-06], + [ 2.4214e-08, 4.3306e-07, 1.8626e-09, ..., 2.5798e-07, + -1.5467e-05, 6.6776e-07], + ..., + [ 1.1455e-07, -3.5558e-06, -3.3528e-08, ..., -2.4457e-06, + 4.9360e-07, 1.7695e-08], + [ 6.2399e-08, 9.0338e-08, 0.0000e+00, ..., 1.9558e-08, + 4.3213e-07, 1.6205e-07], + [-4.7963e-07, 7.9535e-07, 1.4901e-08, ..., 6.3144e-07, + 3.0175e-06, 1.2387e-06]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0215, -0.0275, -0.0116, -0.0273, -0.0307, 0.0041, 0.0260, -0.0135, + 0.0389, -0.0016], device='cuda:0'), grad: tensor([ 3.2671e-06, 2.6420e-05, -2.9445e-05, 3.6836e-05, 9.6951e-07, + 9.3654e-06, -2.4214e-05, -9.7826e-06, 1.1027e-06, -1.4655e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 255.98, cls_loss 0.0038 cls_loss_mapping 0.0049 cls_loss_causal 0.5086 re_mapping 0.0067 re_causal 0.0169 /// teacc 98.99 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0925, -0.1304, -0.0643, ..., -0.0568, 0.1515, 0.1495], + [-0.1691, -0.1566, -0.1087, ..., -0.1265, -0.1761, -0.0871], + [-0.0824, -0.0958, 0.1111, ..., -0.1538, 0.1707, 0.0492], + ..., + [-0.0952, 0.0598, 0.0463, ..., 0.1532, -0.1508, -0.1633], + [-0.2082, 0.0566, -0.1126, ..., 0.0456, -0.0762, -0.1081], + [ 0.0251, -0.0881, -0.0735, ..., -0.0739, -0.0492, -0.1000]], + device='cuda:0'), grad: tensor([[ 1.9558e-06, 8.3819e-08, 0.0000e+00, ..., 7.5903e-07, + 8.8476e-07, 1.0207e-06], + [ 1.2480e-07, 9.7454e-06, 0.0000e+00, ..., 1.2122e-05, + 2.2352e-08, 2.0489e-08], + [ 1.2200e-07, 1.0967e-05, 0.0000e+00, ..., 1.3471e-05, + 2.7940e-09, 1.8626e-09], + ..., + [ 8.5682e-08, -2.3380e-05, 0.0000e+00, ..., -3.0458e-05, + 1.8626e-09, 1.8626e-09], + [ 3.2634e-06, -6.5099e-07, 0.0000e+00, ..., 2.0731e-06, + 1.3569e-06, 1.9409e-06], + [ 1.6779e-05, 1.4016e-06, 0.0000e+00, ..., 7.7263e-06, + 2.3283e-08, 2.9802e-08]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0219, -0.0279, -0.0120, -0.0271, -0.0297, 0.0052, 0.0245, -0.0133, + 0.0389, -0.0019], device='cuda:0'), grad: tensor([ 5.0142e-06, 4.4912e-05, 5.0187e-05, 1.2830e-05, -4.8243e-06, + 1.7062e-05, -5.8055e-05, -1.0926e-04, 4.9062e-06, 3.7253e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 255.99, cls_loss 0.0045 cls_loss_mapping 0.0040 cls_loss_causal 0.5369 re_mapping 0.0066 re_causal 0.0166 /// teacc 98.98 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0938, -0.1317, -0.0645, ..., -0.0573, 0.1515, 0.1496], + [-0.1693, -0.1569, -0.1088, ..., -0.1267, -0.1789, -0.0896], + [-0.0826, -0.0963, 0.1114, ..., -0.1540, 0.1729, 0.0511], + ..., + [-0.0954, 0.0599, 0.0462, ..., 0.1534, -0.1516, -0.1641], + [-0.2089, 0.0564, -0.1127, ..., 0.0454, -0.0750, -0.1072], + [ 0.0247, -0.0882, -0.0735, ..., -0.0740, -0.0497, -0.1027]], + device='cuda:0'), grad: tensor([[ 4.6566e-08, 2.4308e-07, 0.0000e+00, ..., 1.3132e-07, + -2.6636e-07, -1.8254e-07], + [ 1.6205e-07, 1.6689e-06, 0.0000e+00, ..., 3.8464e-07, + 1.0245e-08, 5.5879e-09], + [ 1.1362e-07, 4.1902e-05, 0.0000e+00, ..., 1.7107e-05, + 3.8650e-07, 1.6671e-07], + ..., + [ 3.4459e-08, -9.4652e-05, 0.0000e+00, ..., -3.7432e-05, + 1.0245e-08, 2.7940e-09], + [ 8.2143e-07, 2.0117e-07, 0.0000e+00, ..., -2.1886e-07, + -6.5938e-07, -2.5705e-07], + [ 2.7809e-06, 5.2527e-07, 0.0000e+00, ..., 4.1164e-07, + 1.1642e-07, 7.6368e-08]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0223, -0.0289, -0.0106, -0.0270, -0.0295, 0.0054, 0.0245, -0.0135, + 0.0390, -0.0020], device='cuda:0'), grad: tensor([ 3.6880e-07, -6.5684e-05, 9.0659e-05, 2.3186e-04, 1.8338e-06, + -1.6010e-04, 1.5628e-06, -1.4436e-04, 3.7402e-05, 6.7167e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 255.48, cls_loss 0.0033 cls_loss_mapping 0.0034 cls_loss_causal 0.5572 re_mapping 0.0067 re_causal 0.0176 /// teacc 99.01 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0943, -0.1322, -0.0647, ..., -0.0576, 0.1517, 0.1498], + [-0.1693, -0.1559, -0.1089, ..., -0.1262, -0.1785, -0.0897], + [-0.0827, -0.0967, 0.1115, ..., -0.1544, 0.1730, 0.0513], + ..., + [-0.0955, 0.0600, 0.0476, ..., 0.1536, -0.1538, -0.1647], + [-0.2092, 0.0564, -0.1129, ..., 0.0454, -0.0747, -0.1075], + [ 0.0246, -0.0887, -0.0757, ..., -0.0745, -0.0501, -0.1040]], + device='cuda:0'), grad: tensor([[ 4.5933e-06, 1.0477e-07, 2.3283e-09, ..., 9.1270e-08, + 4.6892e-07, -4.0047e-07], + [ 1.7136e-07, 1.5413e-07, 2.7940e-09, ..., 2.0768e-07, + 7.1749e-06, 2.4214e-06], + [ 1.2293e-07, 1.9465e-07, 2.8871e-08, ..., 3.4459e-08, + 1.8077e-06, 8.9174e-07], + ..., + [ 2.0023e-08, -6.6981e-06, 3.7253e-09, ..., -9.6932e-06, + 2.0443e-07, 1.3830e-07], + [ 8.3214e-07, 6.4587e-07, 0.0000e+00, ..., -1.3970e-09, + 1.8058e-06, 1.6857e-06], + [ 1.1493e-06, 4.9137e-06, 3.2596e-09, ..., 7.1637e-06, + 1.7174e-06, 1.3979e-06]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0223, -0.0278, -0.0109, -0.0268, -0.0295, 0.0050, 0.0251, -0.0139, + 0.0391, -0.0023], device='cuda:0'), grad: tensor([ 9.2313e-06, 1.3158e-05, 5.4613e-06, -4.5784e-06, 5.9679e-06, + 7.9721e-06, -4.6432e-05, -1.8597e-05, 6.0350e-06, 2.1696e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 255.90, cls_loss 0.0044 cls_loss_mapping 0.0045 cls_loss_causal 0.5392 re_mapping 0.0066 re_causal 0.0167 /// teacc 98.98 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0945, -0.1331, -0.0653, ..., -0.0577, 0.1524, 0.1506], + [-0.1699, -0.1561, -0.1090, ..., -0.1264, -0.1786, -0.0898], + [-0.0829, -0.0978, 0.1117, ..., -0.1555, 0.1733, 0.0516], + ..., + [-0.0957, 0.0601, 0.0477, ..., 0.1542, -0.1541, -0.1651], + [-0.2109, 0.0562, -0.1136, ..., 0.0442, -0.0753, -0.1096], + [ 0.0238, -0.0891, -0.0758, ..., -0.0748, -0.0506, -0.1060]], + device='cuda:0'), grad: tensor([[ 2.2398e-07, 2.3469e-07, 0.0000e+00, ..., 0.0000e+00, + -2.8312e-07, -2.4308e-07], + [ 1.2619e-07, 1.2014e-07, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-08, 1.6298e-08], + [ 1.5926e-07, 1.5153e-06, 0.0000e+00, ..., 9.3132e-10, + 6.4261e-08, 2.3749e-08], + ..., + [ 1.0105e-07, 7.9628e-08, 0.0000e+00, ..., -1.3504e-08, + 1.3039e-08, 6.0536e-09], + [ 1.9185e-07, -1.4585e-06, 0.0000e+00, ..., 1.0245e-08, + 7.0315e-08, 3.9116e-08], + [ 4.3539e-07, 3.8231e-07, 0.0000e+00, ..., 9.7789e-09, + 1.6205e-07, 9.4064e-08]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0218, -0.0276, -0.0112, -0.0269, -0.0297, 0.0055, 0.0257, -0.0141, + 0.0372, -0.0021], device='cuda:0'), grad: tensor([ 3.1758e-07, 2.6962e-07, 7.2494e-06, -7.4148e-05, 3.9581e-08, + 7.2122e-05, 1.2666e-07, 2.6124e-07, -7.3761e-06, 1.2089e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 256.02, cls_loss 0.0047 cls_loss_mapping 0.0052 cls_loss_causal 0.5267 re_mapping 0.0065 re_causal 0.0155 /// teacc 99.00 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0948, -0.1335, -0.0655, ..., -0.0581, 0.1536, 0.1523], + [-0.1724, -0.1564, -0.1093, ..., -0.1271, -0.1789, -0.0901], + [-0.0806, -0.0986, 0.1123, ..., -0.1565, 0.1737, 0.0519], + ..., + [-0.0964, 0.0604, 0.0479, ..., 0.1548, -0.1545, -0.1659], + [-0.2117, 0.0561, -0.1137, ..., 0.0441, -0.0757, -0.1107], + [ 0.0233, -0.0892, -0.0760, ..., -0.0751, -0.0508, -0.1067]], + device='cuda:0'), grad: tensor([[ 4.2748e-07, 8.8476e-09, 0.0000e+00, ..., 2.3004e-07, + -2.3600e-06, -1.2983e-06], + [ 1.1129e-07, 2.0955e-08, 0.0000e+00, ..., 4.7032e-08, + 1.1828e-07, 8.7079e-08], + [ 7.4040e-08, 1.3039e-08, 0.0000e+00, ..., 2.6077e-08, + 1.2480e-07, 7.4971e-08], + ..., + [ 1.0524e-07, 2.1420e-08, 4.6566e-10, ..., 8.3353e-08, + 2.3283e-08, 1.5367e-08], + [ 2.4848e-06, 8.4983e-07, 0.0000e+00, ..., 1.6466e-06, + 7.7300e-08, 4.7963e-08], + [ 5.4576e-07, -8.4611e-07, 4.6566e-10, ..., 8.6147e-08, + 1.6633e-06, 8.9500e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0210, -0.0293, -0.0100, -0.0267, -0.0298, 0.0055, 0.0253, -0.0139, + 0.0368, -0.0016], device='cuda:0'), grad: tensor([-1.7751e-06, 8.7824e-07, 1.0179e-06, 1.7911e-05, -5.2229e-06, + -3.0398e-05, 4.5411e-06, 7.9349e-06, 2.0415e-05, -1.5303e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 255.91, cls_loss 0.0041 cls_loss_mapping 0.0042 cls_loss_causal 0.5307 re_mapping 0.0066 re_causal 0.0165 /// teacc 99.00 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0979, -0.1343, -0.0657, ..., -0.0608, 0.1528, 0.1506], + [-0.1726, -0.1565, -0.1093, ..., -0.1279, -0.1790, -0.0902], + [-0.0812, -0.0986, 0.1127, ..., -0.1569, 0.1737, 0.0520], + ..., + [-0.0965, 0.0605, 0.0479, ..., 0.1553, -0.1554, -0.1678], + [-0.2124, 0.0561, -0.1145, ..., 0.0437, -0.0759, -0.1114], + [ 0.0245, -0.0892, -0.0759, ..., -0.0752, -0.0509, -0.1073]], + device='cuda:0'), grad: tensor([[ 3.8603e-07, 6.9849e-08, 4.6566e-10, ..., 4.1118e-07, + 4.1071e-07, 3.7253e-07], + [ 8.9407e-08, 6.8266e-07, 1.3970e-09, ..., 7.6601e-07, + 6.6590e-08, 5.2154e-08], + [ 3.0780e-07, 7.2755e-06, 9.3132e-10, ..., 1.5810e-05, + -2.0303e-06, 4.0606e-07], + ..., + [ 2.8778e-07, -1.5050e-05, -7.3574e-08, ..., -3.0398e-05, + 5.0291e-08, 3.1199e-08], + [ 3.3248e-06, 2.9262e-06, 1.3970e-09, ..., -8.3968e-06, + 4.2617e-06, 2.1234e-06], + [ 2.2352e-07, 9.2685e-06, 1.5926e-07, ..., 1.7866e-05, + 6.3330e-08, 5.1688e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0224, -0.0293, -0.0103, -0.0270, -0.0299, 0.0056, 0.0268, -0.0139, + 0.0362, -0.0011], device='cuda:0'), grad: tensor([ 3.1013e-06, 2.3916e-06, 2.6837e-05, -4.7609e-06, -6.4895e-06, + 3.8058e-05, -4.0889e-05, -7.1883e-05, -3.0100e-06, 5.6475e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 255.66, cls_loss 0.0036 cls_loss_mapping 0.0039 cls_loss_causal 0.5679 re_mapping 0.0062 re_causal 0.0165 /// teacc 99.02 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0980, -0.1349, -0.0655, ..., -0.0607, 0.1531, 0.1509], + [-0.1729, -0.1568, -0.1075, ..., -0.1282, -0.1793, -0.0905], + [-0.0814, -0.0993, 0.1120, ..., -0.1574, 0.1747, 0.0535], + ..., + [-0.0967, 0.0610, 0.0490, ..., 0.1561, -0.1560, -0.1688], + [-0.2128, 0.0558, -0.1151, ..., 0.0435, -0.0769, -0.1130], + [ 0.0248, -0.0894, -0.0789, ..., -0.0759, -0.0511, -0.1079]], + device='cuda:0'), grad: tensor([[ 4.4517e-07, 2.4959e-07, 5.5879e-09, ..., 2.4401e-07, + 2.5705e-06, 9.7882e-07], + [ 7.1246e-08, 1.8813e-07, 8.8476e-09, ..., 1.3085e-07, + -2.1793e-06, 3.2131e-08], + [ 1.0245e-07, 2.6217e-07, 4.6566e-10, ..., 7.0781e-08, + -4.5635e-06, -2.6617e-06], + ..., + [ 7.3574e-08, -1.7975e-06, 2.1467e-07, ..., -1.8990e-06, + 9.1689e-07, 5.3784e-07], + [ 5.0431e-07, 7.2876e-07, 4.0047e-08, ..., 7.8836e-07, + 2.0079e-06, 8.4890e-07], + [ 3.9535e-07, 4.9453e-07, -1.9651e-07, ..., 1.1055e-06, + 5.7044e-07, 3.0361e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0223, -0.0292, -0.0097, -0.0270, -0.0298, 0.0055, 0.0270, -0.0136, + 0.0351, -0.0015], device='cuda:0'), grad: tensor([ 1.2361e-05, -3.6865e-05, -4.6194e-06, -2.7753e-07, 5.6345e-08, + 4.5151e-06, 6.7055e-06, 4.5029e-07, 1.2249e-05, 5.3719e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 255.65, cls_loss 0.0039 cls_loss_mapping 0.0043 cls_loss_causal 0.5232 re_mapping 0.0064 re_causal 0.0162 /// teacc 99.00 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0981, -0.1355, -0.0657, ..., -0.0607, 0.1535, 0.1512], + [-0.1734, -0.1569, -0.1082, ..., -0.1287, -0.1797, -0.0909], + [-0.0815, -0.1015, 0.1112, ..., -0.1594, 0.1753, 0.0541], + ..., + [-0.0969, 0.0615, 0.0512, ..., 0.1574, -0.1571, -0.1700], + [-0.2131, 0.0557, -0.1152, ..., 0.0435, -0.0770, -0.1131], + [ 0.0250, -0.0899, -0.0803, ..., -0.0767, -0.0514, -0.1087]], + device='cuda:0'), grad: tensor([[ 1.3830e-07, 1.0245e-08, 7.9628e-08, ..., 5.2154e-08, + 6.3190e-07, 5.6811e-08], + [ 9.9652e-08, 1.3364e-07, 1.8626e-09, ..., 2.2398e-07, + 2.0023e-08, 3.7253e-09], + [ 3.5856e-08, 2.6543e-08, 2.3283e-08, ..., 1.7695e-08, + -1.1986e-06, -9.3132e-09], + ..., + [ 2.7474e-08, -7.2364e-07, 7.9162e-09, ..., -1.8785e-06, + 6.4261e-08, 3.2596e-09], + [ 1.1912e-06, 1.4901e-08, 0.0000e+00, ..., 5.6345e-07, + 1.3728e-06, 1.1409e-07], + [ 1.3551e-07, 2.1979e-07, 9.7789e-09, ..., 3.3714e-07, + 1.0803e-07, 6.9849e-09]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0222, -0.0291, -0.0103, -0.0283, -0.0294, 0.0068, 0.0265, -0.0131, + 0.0350, -0.0019], device='cuda:0'), grad: tensor([ 1.4473e-06, -2.9653e-05, -1.2502e-05, 9.9558e-07, -1.8161e-07, + -3.9712e-06, 1.1828e-06, 1.5467e-05, 2.5943e-05, 1.3281e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 255.55, cls_loss 0.0045 cls_loss_mapping 0.0051 cls_loss_causal 0.5122 re_mapping 0.0061 re_causal 0.0153 /// teacc 98.95 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0981, -0.1362, -0.0659, ..., -0.0608, 0.1543, 0.1521], + [-0.1749, -0.1574, -0.1082, ..., -0.1276, -0.1799, -0.0918], + [-0.0803, -0.1034, 0.1112, ..., -0.1616, 0.1756, 0.0555], + ..., + [-0.0973, 0.0623, 0.0512, ..., 0.1575, -0.1574, -0.1707], + [-0.2134, 0.0557, -0.1153, ..., 0.0437, -0.0773, -0.1137], + [ 0.0240, -0.0903, -0.0803, ..., -0.0771, -0.0515, -0.1092]], + device='cuda:0'), grad: tensor([[ 4.4797e-07, 1.7714e-06, 0.0000e+00, ..., 1.4435e-08, + -8.4843e-07, -6.9663e-07], + [ 3.3714e-07, 6.8359e-07, 0.0000e+00, ..., 5.4017e-08, + 2.7940e-08, 1.7229e-08], + [ 2.9569e-07, 7.9069e-07, 0.0000e+00, ..., 2.1933e-07, + 4.1444e-08, 4.3772e-08], + ..., + [ 8.2422e-07, 9.9465e-07, 0.0000e+00, ..., -4.6426e-07, + 2.2817e-08, 1.3970e-08], + [-1.8537e-05, -8.0705e-05, 0.0000e+00, ..., -4.1723e-07, + 9.8255e-08, 8.0094e-08], + [ 1.7390e-05, 7.6234e-05, 0.0000e+00, ..., 4.1770e-07, + 2.5565e-07, 2.0396e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0217, -0.0298, -0.0098, -0.0270, -0.0301, 0.0057, 0.0260, -0.0132, + 0.0353, -0.0013], device='cuda:0'), grad: tensor([ 7.5400e-06, 3.3807e-06, 4.3660e-06, 2.6170e-06, -1.1250e-06, + 5.7146e-06, 2.9616e-06, 3.6471e-06, -3.8457e-04, 3.5548e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 255.71, cls_loss 0.0051 cls_loss_mapping 0.0055 cls_loss_causal 0.5581 re_mapping 0.0062 re_causal 0.0161 /// teacc 99.02 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0978, -0.1376, -0.0668, ..., -0.0615, 0.1559, 0.1538], + [-0.1757, -0.1577, -0.1082, ..., -0.1274, -0.1822, -0.0948], + [-0.0799, -0.1038, 0.1113, ..., -0.1620, 0.1760, 0.0558], + ..., + [-0.0976, 0.0625, 0.0512, ..., 0.1576, -0.1579, -0.1711], + [-0.2134, 0.0557, -0.1154, ..., 0.0439, -0.0772, -0.1141], + [ 0.0234, -0.0906, -0.0805, ..., -0.0775, -0.0517, -0.1098]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 1.3039e-08, 1.4435e-08, ..., 9.3132e-10, + 5.1688e-08, 2.6077e-08], + [ 6.6124e-08, 2.1653e-07, 4.1910e-09, ..., 1.4296e-07, + 1.6438e-07, 2.2352e-08], + [ 8.1956e-08, 1.3076e-06, -1.7043e-07, ..., 9.3132e-10, + -5.0385e-07, -1.6391e-07], + ..., + [ 2.0955e-08, -8.3819e-09, 1.6764e-08, ..., -1.9418e-07, + 7.6834e-08, 1.9558e-08], + [ 2.7521e-07, 6.1886e-07, 5.8208e-08, ..., 1.3039e-08, + 1.4622e-07, 8.8941e-08], + [ 1.6531e-07, 1.6671e-07, 2.3283e-09, ..., 6.2864e-08, + 6.0536e-09, 2.7940e-09]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0203, -0.0310, -0.0092, -0.0278, -0.0300, 0.0064, 0.0253, -0.0133, + 0.0362, -0.0013], device='cuda:0'), grad: tensor([ 3.6694e-07, -5.6531e-07, -1.1120e-06, -1.1966e-05, -6.0257e-07, + 9.6187e-06, 1.8487e-07, 6.5146e-07, 2.2650e-06, 1.1288e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 256.13, cls_loss 0.0040 cls_loss_mapping 0.0037 cls_loss_causal 0.5327 re_mapping 0.0062 re_causal 0.0156 /// teacc 98.95 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0980, -0.1382, -0.0675, ..., -0.0619, 0.1561, 0.1539], + [-0.1757, -0.1587, -0.1083, ..., -0.1281, -0.1834, -0.0948], + [-0.0800, -0.1040, 0.1117, ..., -0.1625, 0.1773, 0.0559], + ..., + [-0.0986, 0.0619, 0.0513, ..., 0.1580, -0.1575, -0.1715], + [-0.2138, 0.0559, -0.1156, ..., 0.0438, -0.0774, -0.1144], + [ 0.0230, -0.0911, -0.0807, ..., -0.0776, -0.0519, -0.1103]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 1.4016e-07, 0.0000e+00, ..., 8.1956e-08, + -5.3085e-08, 2.9337e-08], + [ 7.7300e-08, 3.5949e-07, 0.0000e+00, ..., 2.7195e-07, + 2.4214e-08, 4.0978e-08], + [ 2.1420e-08, 5.2527e-06, 0.0000e+00, ..., 3.1758e-06, + -3.7253e-09, 3.2596e-09], + ..., + [ 2.3283e-09, -6.9402e-06, 9.3132e-10, ..., -5.2825e-06, + 1.3970e-09, 1.3970e-09], + [ 1.1437e-06, 2.3600e-06, 0.0000e+00, ..., -8.4750e-08, + 3.0501e-07, 6.2073e-07], + [-2.5518e-07, 5.6159e-07, -9.3132e-10, ..., 4.9965e-07, + 5.8673e-08, 3.5390e-08]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0203, -0.0314, -0.0096, -0.0279, -0.0302, 0.0067, 0.0255, -0.0129, + 0.0366, -0.0015], device='cuda:0'), grad: tensor([ 6.7428e-07, 1.1604e-06, 1.4037e-05, -1.1683e-05, 2.6207e-06, + 6.1579e-06, -9.3356e-06, -1.6153e-05, 1.3173e-05, -6.7288e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 255.80, cls_loss 0.0035 cls_loss_mapping 0.0037 cls_loss_causal 0.5005 re_mapping 0.0064 re_causal 0.0163 /// teacc 98.93 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0982, -0.1385, -0.0689, ..., -0.0621, 0.1563, 0.1540], + [-0.1759, -0.1591, -0.1082, ..., -0.1286, -0.1835, -0.0948], + [-0.0801, -0.1045, 0.1125, ..., -0.1629, 0.1778, 0.0559], + ..., + [-0.0990, 0.0618, 0.0512, ..., 0.1583, -0.1586, -0.1720], + [-0.2140, 0.0559, -0.1162, ..., 0.0437, -0.0775, -0.1146], + [ 0.0224, -0.0912, -0.0804, ..., -0.0782, -0.0521, -0.1108]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 2.5146e-08, 3.7253e-09, ..., 2.0489e-08, + -1.6764e-08, -5.1223e-09], + [ 3.4925e-08, 6.4820e-07, 1.2992e-07, ..., 6.2771e-07, + 6.2399e-08, 2.3283e-09], + [ 6.0536e-09, 2.3749e-08, -4.9826e-08, ..., 2.7474e-08, + -7.8697e-08, 9.3132e-10], + ..., + [ 2.5146e-08, -8.3633e-07, -9.5461e-08, ..., -1.0524e-06, + 6.5193e-09, 9.3132e-10], + [ 7.8231e-08, 2.0489e-08, 1.8626e-09, ..., 8.3819e-09, + 2.0023e-08, 2.1420e-08], + [ 2.2491e-07, 1.5274e-07, -2.2352e-08, ..., 2.3516e-07, + 7.9162e-09, 5.5879e-09]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0203, -0.0316, -0.0095, -0.0277, -0.0291, 0.0065, 0.0257, -0.0129, + 0.0367, -0.0022], device='cuda:0'), grad: tensor([ 2.2165e-07, 1.7658e-06, 1.6158e-07, -3.3015e-07, 7.9162e-08, + 1.0077e-06, -7.1479e-07, -2.2352e-06, 5.4762e-07, -5.2154e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 255.98, cls_loss 0.0045 cls_loss_mapping 0.0031 cls_loss_causal 0.5416 re_mapping 0.0063 re_causal 0.0157 /// teacc 98.98 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0990, -0.1393, -0.0691, ..., -0.0625, 0.1571, 0.1546], + [-0.1761, -0.1593, -0.1072, ..., -0.1287, -0.1836, -0.0950], + [-0.0802, -0.1074, 0.1129, ..., -0.1660, 0.1778, 0.0557], + ..., + [-0.0996, 0.0627, 0.0508, ..., 0.1597, -0.1591, -0.1723], + [-0.2147, 0.0560, -0.1163, ..., 0.0442, -0.0780, -0.1157], + [ 0.0216, -0.0915, -0.0807, ..., -0.0789, -0.0522, -0.1115]], + device='cuda:0'), grad: tensor([[-4.0140e-07, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + -2.1935e-05, -1.1869e-05], + [ 2.9802e-08, 1.1176e-08, 0.0000e+00, ..., 1.5832e-08, + 1.4808e-07, 7.0781e-08], + [ 7.6368e-08, 5.5879e-09, -6.5193e-09, ..., 7.4506e-09, + -2.1055e-05, 1.8161e-07], + ..., + [ 2.1420e-08, -4.3772e-08, 0.0000e+00, ..., -6.1467e-08, + 4.9360e-08, 2.1420e-08], + [ 3.9600e-06, 9.3132e-10, 0.0000e+00, ..., -8.2888e-08, + 1.4016e-06, 5.6066e-07], + [-3.4552e-07, 1.9558e-08, 0.0000e+00, ..., 3.3528e-08, + 4.9546e-07, 2.5611e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0200, -0.0310, -0.0111, -0.0276, -0.0287, 0.0064, 0.0259, -0.0125, + 0.0366, -0.0027], device='cuda:0'), grad: tensor([-2.7746e-05, -6.5193e-07, -5.0664e-05, 9.4995e-07, 4.7445e-05, + 2.5034e-06, 8.2925e-06, 2.9709e-07, 1.4655e-05, 4.9062e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 255.48, cls_loss 0.0043 cls_loss_mapping 0.0058 cls_loss_causal 0.5499 re_mapping 0.0064 re_causal 0.0161 /// teacc 98.92 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0992, -0.1398, -0.0692, ..., -0.0629, 0.1574, 0.1549], + [-0.1763, -0.1577, -0.1072, ..., -0.1294, -0.1837, -0.0953], + [-0.0802, -0.1074, 0.1135, ..., -0.1661, 0.1783, 0.0562], + ..., + [-0.1007, 0.0614, 0.0507, ..., 0.1602, -0.1598, -0.1725], + [-0.2152, 0.0560, -0.1172, ..., 0.0441, -0.0782, -0.1160], + [ 0.0217, -0.0916, -0.0807, ..., -0.0793, -0.0524, -0.1123]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 4.6566e-09, 0.0000e+00, ..., -3.2596e-07, + -2.7325e-06, -2.4736e-06], + [ 2.2352e-08, 6.2399e-08, 0.0000e+00, ..., 4.6566e-08, + 1.3569e-06, 4.2282e-07], + [ 7.4506e-09, 9.3132e-09, 0.0000e+00, ..., 3.0734e-08, + -5.2340e-07, -8.0559e-07], + ..., + [ 1.4901e-08, -4.1630e-07, 0.0000e+00, ..., -3.4645e-07, + 5.4110e-07, 3.1665e-07], + [ 5.3085e-08, 5.4017e-08, 0.0000e+00, ..., -2.8778e-07, + -1.7369e-06, 8.4471e-07], + [ 2.0489e-08, 7.8231e-08, 0.0000e+00, ..., 3.5483e-07, + 1.0524e-06, 6.9477e-07]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0200, -0.0289, -0.0107, -0.0271, -0.0293, 0.0060, 0.0261, -0.0145, + 0.0362, -0.0025], device='cuda:0'), grad: tensor([-6.7391e-06, 1.2908e-06, 4.4741e-06, 2.4047e-06, 2.3916e-06, + 1.4612e-06, 2.6990e-06, 1.6252e-06, -1.2174e-05, 2.4885e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 255.66, cls_loss 0.0049 cls_loss_mapping 0.0052 cls_loss_causal 0.4969 re_mapping 0.0063 re_causal 0.0154 /// teacc 98.96 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0998, -0.1408, -0.0696, ..., -0.0632, 0.1573, 0.1548], + [-0.1773, -0.1577, -0.1072, ..., -0.1298, -0.1838, -0.0953], + [-0.0799, -0.1080, 0.1138, ..., -0.1666, 0.1784, 0.0562], + ..., + [-0.1013, 0.0602, 0.0507, ..., 0.1595, -0.1603, -0.1730], + [-0.2154, 0.0586, -0.1178, ..., 0.0464, -0.0784, -0.1166], + [ 0.0215, -0.0919, -0.0806, ..., -0.0796, -0.0524, -0.1127]], + device='cuda:0'), grad: tensor([[ 1.8813e-07, 3.3528e-08, 0.0000e+00, ..., 6.5193e-09, + -1.2815e-05, -4.0568e-06], + [ 1.6764e-08, 6.7055e-08, 0.0000e+00, ..., 5.4017e-08, + 1.0775e-06, 9.3132e-10], + [ 1.0245e-08, -1.0896e-07, 0.0000e+00, ..., 2.7940e-09, + -2.4438e-06, 4.4703e-08], + ..., + [ 2.1420e-08, -4.2878e-06, 0.0000e+00, ..., -6.5342e-06, + 1.1493e-06, 1.8626e-09], + [ 3.2596e-08, -2.5146e-08, 0.0000e+00, ..., 1.4901e-08, + 1.1086e-05, 3.5521e-06], + [ 9.9652e-08, 4.2319e-06, 0.0000e+00, ..., 6.2697e-06, + 2.2613e-06, 5.9232e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0205, -0.0288, -0.0103, -0.0260, -0.0293, 0.0046, 0.0266, -0.0150, + 0.0383, -0.0027], device='cuda:0'), grad: tensor([-1.9297e-05, 2.6137e-05, -6.9499e-05, 6.4075e-07, 6.6124e-07, + -1.6484e-07, -2.6077e-07, 1.5058e-05, 2.3097e-05, 2.3484e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 255.81, cls_loss 0.0039 cls_loss_mapping 0.0037 cls_loss_causal 0.5165 re_mapping 0.0060 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1000, -0.1412, -0.0698, ..., -0.0637, 0.1575, 0.1550], + [-0.1774, -0.1579, -0.1070, ..., -0.1300, -0.1842, -0.0959], + [-0.0800, -0.1086, 0.1138, ..., -0.1673, 0.1788, 0.0570], + ..., + [-0.1021, 0.0605, 0.0507, ..., 0.1599, -0.1607, -0.1735], + [-0.2159, 0.0584, -0.1179, ..., 0.0471, -0.0779, -0.1169], + [ 0.0213, -0.0920, -0.0812, ..., -0.0798, -0.0526, -0.1131]], + device='cuda:0'), grad: tensor([[-3.9767e-07, 2.0489e-08, 0.0000e+00, ..., 4.1910e-08, + -3.3844e-06, -1.8496e-06], + [ 1.3039e-08, 8.3819e-08, 0.0000e+00, ..., 1.6484e-07, + 3.9116e-08, 1.5832e-08], + [ 1.3970e-08, 7.7859e-07, 0.0000e+00, ..., 1.6941e-06, + 1.5367e-07, 7.3574e-08], + ..., + [ 5.2154e-08, -9.1735e-07, 0.0000e+00, ..., -1.9949e-06, + 1.8626e-07, 1.2107e-07], + [ 1.5832e-08, 1.8626e-08, 0.0000e+00, ..., -5.0291e-08, + 1.5181e-07, 6.3330e-08], + [ 7.1712e-08, 7.2643e-08, 0.0000e+00, ..., 7.4506e-08, + 1.3215e-06, 6.6962e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0205, -0.0289, -0.0103, -0.0261, -0.0293, 0.0046, 0.0261, -0.0147, + 0.0385, -0.0029], device='cuda:0'), grad: tensor([-5.0068e-06, 3.0734e-06, 1.0312e-05, -2.4945e-05, -6.2771e-06, + 2.6584e-05, 3.5018e-06, -9.8422e-06, 2.9150e-07, 2.3916e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 195---------------------------------------------------- +epoch 195, time 272.05, cls_loss 0.0038 cls_loss_mapping 0.0039 cls_loss_causal 0.5316 re_mapping 0.0060 re_causal 0.0156 /// teacc 99.05 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1003, -0.1416, -0.0703, ..., -0.0644, 0.1576, 0.1549], + [-0.1776, -0.1580, -0.1070, ..., -0.1303, -0.1844, -0.0961], + [-0.0801, -0.1099, 0.1140, ..., -0.1697, 0.1790, 0.0558], + ..., + [-0.1024, 0.0609, 0.0506, ..., 0.1606, -0.1610, -0.1739], + [-0.2165, 0.0584, -0.1179, ..., 0.0475, -0.0770, -0.1180], + [ 0.0214, -0.0922, -0.0812, ..., -0.0801, -0.0529, -0.1139]], + device='cuda:0'), grad: tensor([[ 1.7136e-07, 5.5879e-09, 0.0000e+00, ..., -3.1292e-07, + -3.6731e-06, -4.5598e-06], + [ 3.0175e-07, 2.1420e-08, 0.0000e+00, ..., 6.6124e-08, + 2.1420e-08, 1.7695e-08], + [ 1.8720e-07, 7.4506e-09, 0.0000e+00, ..., -3.0994e-05, + -9.6187e-06, -1.0775e-06], + ..., + [ 2.3097e-07, -4.7497e-08, 0.0000e+00, ..., 3.0845e-05, + 9.6411e-06, 1.1232e-06], + [ 5.6997e-06, 2.5146e-08, 0.0000e+00, ..., 2.0396e-07, + 2.8871e-08, 3.8184e-08], + [ 2.6450e-07, 4.2841e-08, 0.0000e+00, ..., 7.6368e-08, + 1.9185e-07, 1.9278e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0208, -0.0289, -0.0115, -0.0259, -0.0299, 0.0047, 0.0260, -0.0145, + 0.0389, -0.0025], device='cuda:0'), grad: tensor([-6.8992e-06, -1.0934e-06, -2.0278e-04, 7.3574e-06, 2.6356e-07, + -4.6104e-05, 4.1217e-05, 2.0480e-04, 7.1228e-06, -3.8072e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 255.76, cls_loss 0.0043 cls_loss_mapping 0.0040 cls_loss_causal 0.5381 re_mapping 0.0060 re_causal 0.0155 /// teacc 98.97 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1003, -0.1426, -0.0705, ..., -0.0637, 0.1577, 0.1554], + [-0.1780, -0.1599, -0.1071, ..., -0.1316, -0.1847, -0.0968], + [-0.0802, -0.1101, 0.1141, ..., -0.1698, 0.1800, 0.0572], + ..., + [-0.1027, 0.0623, 0.0506, ..., 0.1616, -0.1614, -0.1748], + [-0.2169, 0.0583, -0.1180, ..., 0.0474, -0.0775, -0.1190], + [ 0.0208, -0.0929, -0.0812, ..., -0.0808, -0.0529, -0.1153]], + device='cuda:0'), grad: tensor([[ 8.1025e-08, 3.7253e-09, 1.8626e-09, ..., 2.1420e-08, + -2.2724e-07, -1.7975e-07], + [ 8.5682e-08, 6.4727e-07, 1.8626e-09, ..., 1.8533e-07, + 2.2352e-08, 1.1176e-08], + [ 2.0489e-08, 3.8557e-07, -8.3819e-09, ..., 5.5134e-07, + -6.2399e-08, -1.8626e-08], + ..., + [ 5.4017e-08, 1.6335e-06, 1.8626e-09, ..., -3.6601e-07, + 3.2596e-08, 1.9558e-08], + [ 1.9465e-07, -2.6822e-07, 9.3132e-10, ..., -2.2724e-07, + 1.9558e-08, 1.2107e-08], + [ 2.8592e-07, 1.3039e-08, 0.0000e+00, ..., 7.3574e-08, + 7.4506e-08, 5.7742e-08]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0211, -0.0301, -0.0110, -0.0258, -0.0298, 0.0047, 0.0261, -0.0134, + 0.0384, -0.0023], device='cuda:0'), grad: tensor([-2.0489e-07, 2.5760e-06, 2.0899e-06, 3.3647e-05, 1.2405e-06, + -3.9309e-05, 2.8480e-06, 5.3868e-06, -4.1164e-06, -4.1015e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 251.39, cls_loss 0.0046 cls_loss_mapping 0.0046 cls_loss_causal 0.5418 re_mapping 0.0063 re_causal 0.0157 /// teacc 99.02 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1020, -0.1433, -0.0706, ..., -0.0640, 0.1583, 0.1549], + [-0.1783, -0.1600, -0.1071, ..., -0.1319, -0.1848, -0.0973], + [-0.0801, -0.1105, 0.1141, ..., -0.1701, 0.1803, 0.0580], + ..., + [-0.1030, 0.0622, 0.0506, ..., 0.1619, -0.1621, -0.1762], + [-0.2178, 0.0585, -0.1181, ..., 0.0454, -0.0780, -0.1212], + [ 0.0198, -0.0934, -0.0814, ..., -0.0810, -0.0550, -0.1189]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.4214e-08, + -3.2317e-07, -2.0303e-07], + [ 1.8626e-09, 6.5193e-09, 1.8626e-09, ..., 1.6403e-04, + 6.5193e-09, 4.6566e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 2.4494e-07, + -3.5390e-08, 5.5879e-09], + ..., + [ 1.8626e-09, -1.8626e-09, 7.4506e-09, ..., -1.6797e-04, + 8.3819e-09, 3.7253e-09], + [ 1.0710e-07, 1.3039e-08, 9.3132e-10, ..., -5.9791e-07, + 1.0803e-07, 6.9849e-08], + [ 9.3132e-10, 4.6566e-09, -2.2352e-08, ..., 3.8482e-06, + 1.3877e-07, 9.2201e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0209, -0.0302, -0.0106, -0.0250, -0.0300, 0.0054, 0.0267, -0.0134, + 0.0364, -0.0030], device='cuda:0'), grad: tensor([-1.0990e-07, 5.2547e-04, 8.3447e-07, 5.0142e-06, 3.7774e-06, + 3.9767e-07, -7.3574e-08, -5.3549e-04, 3.4738e-07, -1.2182e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 250.75, cls_loss 0.0040 cls_loss_mapping 0.0040 cls_loss_causal 0.5100 re_mapping 0.0061 re_causal 0.0155 /// teacc 98.99 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1022, -0.1439, -0.0709, ..., -0.0641, 0.1586, 0.1551], + [-0.1787, -0.1601, -0.1070, ..., -0.1328, -0.1850, -0.0974], + [-0.0802, -0.1109, 0.1141, ..., -0.1706, 0.1805, 0.0579], + ..., + [-0.1036, 0.0622, 0.0505, ..., 0.1626, -0.1626, -0.1765], + [-0.2184, 0.0585, -0.1177, ..., 0.0448, -0.0783, -0.1223], + [ 0.0204, -0.0935, -0.0815, ..., -0.0813, -0.0553, -0.1197]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 7.1712e-08, 0.0000e+00, ..., 1.8720e-07, + -1.1018e-06, -7.2457e-07], + [ 1.5832e-07, 9.2201e-08, 0.0000e+00, ..., 1.7229e-07, + 1.0990e-07, 7.7300e-08], + [ 1.5832e-08, 1.1111e-06, 0.0000e+00, ..., 1.4976e-06, + 1.8626e-08, 4.1910e-08], + ..., + [ 6.5193e-09, -2.6971e-06, 0.0000e+00, ..., -4.5598e-06, + 2.3283e-08, 2.7940e-09], + [ 1.0338e-07, 5.8394e-07, 0.0000e+00, ..., 7.7859e-07, + 1.1362e-07, 7.4506e-08], + [ 1.6764e-08, 1.5181e-07, 0.0000e+00, ..., 3.4273e-07, + 9.0338e-08, 5.5879e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0209, -0.0292, -0.0118, -0.0252, -0.0299, 0.0058, 0.0274, -0.0141, + 0.0358, -0.0030], device='cuda:0'), grad: tensor([-1.1390e-06, 6.9477e-07, 2.5630e-06, 1.1111e-06, 2.5276e-06, + 1.0535e-05, -1.0252e-05, -7.4394e-06, 1.3635e-06, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 251.15, cls_loss 0.0036 cls_loss_mapping 0.0031 cls_loss_causal 0.5240 re_mapping 0.0060 re_causal 0.0150 /// teacc 98.99 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1026, -0.1445, -0.0715, ..., -0.0644, 0.1588, 0.1551], + [-0.1792, -0.1601, -0.1070, ..., -0.1333, -0.1852, -0.0976], + [-0.0799, -0.1112, 0.1151, ..., -0.1711, 0.1806, 0.0582], + ..., + [-0.1038, 0.0622, 0.0504, ..., 0.1630, -0.1630, -0.1770], + [-0.2181, 0.0584, -0.1183, ..., 0.0454, -0.0777, -0.1226], + [ 0.0188, -0.0939, -0.0816, ..., -0.0816, -0.0556, -0.1202]], + device='cuda:0'), grad: tensor([[ 2.4159e-06, 9.3132e-10, 9.3132e-10, ..., -0.0000e+00, + -4.4964e-06, 1.4370e-06], + [ 1.5832e-08, 1.1735e-07, 4.6566e-09, ..., -1.6494e-06, + 5.3924e-07, 2.1234e-07], + [ 2.1420e-08, 3.7253e-09, 0.0000e+00, ..., 1.6764e-08, + -6.8285e-06, -2.8498e-06], + ..., + [ 5.5879e-09, -1.9185e-07, 1.4901e-08, ..., -2.9057e-07, + 2.0582e-06, 8.3260e-07], + [ 1.8626e-07, 1.8626e-09, 2.7940e-09, ..., -2.7940e-08, + 3.8110e-06, 1.7760e-06], + [ 4.5635e-08, 3.3528e-08, -4.0978e-08, ..., 9.9652e-08, + 5.1372e-06, 1.1846e-06]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0209, -0.0293, -0.0119, -0.0249, -0.0299, 0.0053, 0.0276, -0.0141, + 0.0362, -0.0032], device='cuda:0'), grad: tensor([-1.3337e-06, -1.3933e-05, -1.4499e-05, 2.3600e-06, 1.3039e-05, + 1.7798e-06, -1.1519e-05, 4.4033e-06, 9.7007e-06, 9.9838e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 250.88, cls_loss 0.0037 cls_loss_mapping 0.0031 cls_loss_causal 0.5298 re_mapping 0.0061 re_causal 0.0153 /// teacc 98.92 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1027, -0.1448, -0.0689, ..., -0.0644, 0.1598, 0.1561], + [-0.1793, -0.1602, -0.1067, ..., -0.1332, -0.1854, -0.0978], + [-0.0800, -0.1120, 0.1155, ..., -0.1715, 0.1813, 0.0584], + ..., + [-0.1040, 0.0624, 0.0499, ..., 0.1633, -0.1643, -0.1779], + [-0.2186, 0.0584, -0.1182, ..., 0.0455, -0.0779, -0.1231], + [ 0.0183, -0.0940, -0.0827, ..., -0.0819, -0.0567, -0.1229]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 5.5879e-09, 0.0000e+00, ..., 5.3085e-08, + 5.9698e-07, 1.8999e-07], + [ 1.8626e-08, 1.6578e-07, 0.0000e+00, ..., 6.1467e-06, + 1.4165e-06, 4.6659e-07], + [ 9.3132e-10, 4.3679e-07, 0.0000e+00, ..., 1.0049e-06, + -9.8050e-06, -3.2280e-06], + ..., + [ 6.3330e-08, -5.0850e-07, 0.0000e+00, ..., -1.6075e-06, + 2.5053e-06, 8.2515e-07], + [ 3.6322e-08, 1.9465e-07, 0.0000e+00, ..., -6.0536e-06, + 2.8722e-06, 9.4622e-07], + [ 3.1665e-08, 1.1828e-07, 0.0000e+00, ..., 6.5193e-08, + 1.3877e-07, 4.6566e-08]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0202, -0.0291, -0.0121, -0.0251, -0.0298, 0.0051, 0.0280, -0.0141, + 0.0361, -0.0032], device='cuda:0'), grad: tensor([ 2.4624e-06, 2.3261e-05, -3.4869e-05, 5.1446e-06, 9.0338e-07, + 5.8301e-07, 2.5406e-06, 7.7337e-06, -8.4788e-06, 6.9477e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 250.64, cls_loss 0.0037 cls_loss_mapping 0.0033 cls_loss_causal 0.5372 re_mapping 0.0060 re_causal 0.0155 /// teacc 99.01 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1019, -0.1452, -0.0665, ..., -0.0642, 0.1613, 0.1577], + [-0.1794, -0.1612, -0.1064, ..., -0.1365, -0.1855, -0.0979], + [-0.0802, -0.1125, 0.1150, ..., -0.1718, 0.1810, 0.0575], + ..., + [-0.1047, 0.0632, 0.0499, ..., 0.1652, -0.1648, -0.1783], + [-0.2189, 0.0583, -0.1190, ..., 0.0455, -0.0781, -0.1234], + [ 0.0179, -0.0943, -0.0830, ..., -0.0819, -0.0569, -0.1236]], + device='cuda:0'), grad: tensor([[ 2.6263e-07, 2.1886e-07, 1.4901e-08, ..., 1.5832e-08, + -1.1651e-06, -9.1270e-07], + [ 1.0803e-07, 4.0606e-07, 9.3132e-10, ..., 1.1362e-07, + 1.7695e-08, 1.5832e-08], + [ 5.3085e-07, 6.2585e-07, -1.9372e-07, ..., 1.0338e-07, + -6.4354e-07, -1.1176e-08], + ..., + [ 2.5146e-07, 1.9416e-05, 1.8720e-07, ..., 6.1058e-06, + 6.5938e-07, 5.3085e-08], + [ 4.6641e-06, 2.4401e-06, 2.7008e-08, ..., 3.7663e-06, + 1.0617e-07, 1.8999e-07], + [ 3.7253e-07, 5.4948e-07, 2.7940e-09, ..., 2.6636e-07, + 2.0023e-07, 1.5553e-07]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0190, -0.0300, -0.0123, -0.0254, -0.0305, 0.0053, 0.0276, -0.0134, + 0.0360, -0.0027], device='cuda:0'), grad: tensor([-1.2759e-06, 8.0746e-07, 1.0533e-06, -3.6240e-05, 1.5926e-07, + -3.6433e-06, -3.1851e-06, 3.0696e-05, 1.0252e-05, 1.3830e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 250.51, cls_loss 0.0034 cls_loss_mapping 0.0034 cls_loss_causal 0.5381 re_mapping 0.0060 re_causal 0.0153 /// teacc 98.97 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1021, -0.1462, -0.0663, ..., -0.0643, 0.1615, 0.1580], + [-0.1798, -0.1613, -0.1065, ..., -0.1370, -0.1856, -0.0982], + [-0.0803, -0.1129, 0.1160, ..., -0.1720, 0.1812, 0.0577], + ..., + [-0.1054, 0.0633, 0.0497, ..., 0.1656, -0.1658, -0.1791], + [-0.2199, 0.0584, -0.1185, ..., 0.0455, -0.0781, -0.1240], + [ 0.0162, -0.0945, -0.0830, ..., -0.0819, -0.0570, -0.1243]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 9.3132e-10, 0.0000e+00, ..., -2.7940e-09, + -2.3749e-07, -1.9278e-07], + [ 1.2107e-08, 1.7695e-08, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 3.7253e-09, 4.6566e-09, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 2.7940e-09], + ..., + [ 1.0245e-08, 2.7940e-09, 0.0000e+00, ..., -5.4017e-08, + 1.8626e-09, 9.3132e-10], + [ 2.4214e-08, 1.4901e-08, 0.0000e+00, ..., -3.5390e-08, + 2.9802e-08, 2.4214e-08], + [ 5.5879e-09, 1.2107e-08, 0.0000e+00, ..., 6.0536e-08, + 1.6019e-07, 1.3132e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0191, -0.0301, -0.0122, -0.0254, -0.0321, 0.0051, 0.0280, -0.0134, + 0.0361, -0.0013], device='cuda:0'), grad: tensor([-3.6415e-07, -4.2282e-07, 4.2841e-08, -1.2014e-07, 5.3085e-08, + 3.9116e-08, 7.1712e-08, 1.4249e-07, -1.1083e-07, 6.4541e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 250.91, cls_loss 0.0036 cls_loss_mapping 0.0028 cls_loss_causal 0.5382 re_mapping 0.0059 re_causal 0.0154 /// teacc 99.01 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1030, -0.1473, -0.0664, ..., -0.0647, 0.1611, 0.1577], + [-0.1800, -0.1614, -0.1065, ..., -0.1376, -0.1857, -0.0984], + [-0.0805, -0.1131, 0.1165, ..., -0.1726, 0.1815, 0.0577], + ..., + [-0.1066, 0.0634, 0.0496, ..., 0.1662, -0.1660, -0.1794], + [-0.2203, 0.0583, -0.1187, ..., 0.0456, -0.0782, -0.1241], + [ 0.0152, -0.0948, -0.0831, ..., -0.0824, -0.0567, -0.1247]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, 1.8626e-09, 0.0000e+00, ..., 7.1712e-08, + -1.4184e-06, -7.3947e-07], + [ 3.8184e-08, 4.3772e-08, 0.0000e+00, ..., 7.2643e-08, + 4.4052e-07, 3.4645e-07], + [ 2.7940e-09, 1.2480e-07, 0.0000e+00, ..., 1.1455e-07, + -3.9954e-07, -3.3807e-07], + ..., + [ 2.4214e-08, -7.9256e-07, 0.0000e+00, ..., -8.1304e-07, + 6.1467e-08, 3.8184e-08], + [ 1.6484e-07, 3.9116e-08, 0.0000e+00, ..., 1.8161e-07, + 7.3574e-08, 4.0978e-08], + [-1.5832e-08, 1.0151e-07, 0.0000e+00, ..., 1.6298e-07, + 7.5530e-07, 3.9488e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0199, -0.0299, -0.0130, -0.0256, -0.0311, 0.0053, 0.0285, -0.0134, + 0.0360, -0.0019], device='cuda:0'), grad: tensor([-2.5816e-06, 1.9819e-06, -1.3299e-06, 8.9221e-07, -5.1688e-07, + -2.0750e-06, 2.5332e-06, -9.8534e-07, 5.4669e-07, 1.5339e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 250.46, cls_loss 0.0045 cls_loss_mapping 0.0043 cls_loss_causal 0.5049 re_mapping 0.0056 re_causal 0.0134 /// teacc 98.98 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1024, -0.1488, -0.0665, ..., -0.0636, 0.1623, 0.1589], + [-0.1802, -0.1615, -0.1051, ..., -0.1382, -0.1859, -0.0987], + [-0.0806, -0.1129, 0.1169, ..., -0.1730, 0.1820, 0.0583], + ..., + [-0.1073, 0.0636, 0.0487, ..., 0.1671, -0.1670, -0.1813], + [-0.2210, 0.0582, -0.1189, ..., 0.0455, -0.0795, -0.1264], + [ 0.0149, -0.0952, -0.0835, ..., -0.0830, -0.0572, -0.1256]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 1.0431e-07, 9.3132e-10, ..., 1.6205e-07, + 7.5623e-07, 2.1793e-07], + [ 3.3528e-08, 1.6391e-07, 2.7940e-09, ..., 2.6356e-07, + 6.5938e-07, 4.3679e-07], + [ 1.0245e-08, -4.9658e-06, 0.0000e+00, ..., -1.8068e-07, + -1.0766e-05, -1.6671e-06], + ..., + [ 5.5879e-09, 5.4948e-07, 1.8626e-09, ..., -8.7824e-07, + 1.4482e-06, 3.0547e-07], + [ 9.5088e-07, 1.9744e-07, 1.8626e-09, ..., 1.1399e-06, + 4.7460e-06, 9.0804e-07], + [ 4.6566e-09, 7.1712e-08, -1.6764e-08, ..., 1.0245e-07, + 4.9360e-08, 2.2352e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0186, -0.0296, -0.0126, -0.0258, -0.0320, 0.0051, 0.0280, -0.0136, + 0.0354, -0.0013], device='cuda:0'), grad: tensor([ 2.2445e-06, 2.3115e-06, -3.6955e-05, 1.6183e-05, -3.2037e-07, + 6.7335e-07, -2.0396e-06, 4.4256e-06, 1.4037e-05, -5.2154e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 250.74, cls_loss 0.0034 cls_loss_mapping 0.0028 cls_loss_causal 0.4910 re_mapping 0.0060 re_causal 0.0146 /// teacc 98.90 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1027, -0.1493, -0.0666, ..., -0.0637, 0.1625, 0.1591], + [-0.1813, -0.1617, -0.1052, ..., -0.1387, -0.1864, -0.1006], + [-0.0797, -0.1128, 0.1173, ..., -0.1726, 0.1837, 0.0621], + ..., + [-0.1075, 0.0637, 0.0486, ..., 0.1673, -0.1697, -0.1856], + [-0.2195, 0.0581, -0.1194, ..., 0.0468, -0.0775, -0.1253], + [ 0.0146, -0.0952, -0.0831, ..., -0.0825, -0.0574, -0.1261]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.4506e-09, 3.4459e-08, ..., 0.0000e+00, + -1.1176e-07, -1.0245e-07], + [ 2.7008e-08, 8.9407e-08, 1.8626e-09, ..., 8.3819e-09, + 2.7940e-08, 1.8626e-08], + [ 2.7008e-08, 1.4715e-07, -1.9558e-07, ..., 5.4017e-08, + -4.4983e-07, -1.0990e-07], + ..., + [ 1.6764e-08, -1.2107e-07, 8.4750e-08, ..., 5.6811e-08, + 1.4435e-07, 1.1176e-08], + [ 1.3690e-07, 2.2259e-07, 9.3132e-09, ..., 1.5553e-07, + 1.1642e-07, 7.3574e-08], + [ 7.7300e-08, 4.4703e-08, -6.5193e-09, ..., 9.0338e-08, + 9.9652e-08, 6.4261e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0187, -0.0299, -0.0117, -0.0257, -0.0317, 0.0049, 0.0265, -0.0137, + 0.0363, -0.0009], device='cuda:0'), grad: tensor([ 6.2399e-08, 1.9744e-07, -1.4175e-06, -5.6997e-07, -1.0589e-06, + -2.5425e-07, 2.3469e-07, 1.6363e-06, 7.5623e-07, 4.2655e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 250.91, cls_loss 0.0043 cls_loss_mapping 0.0040 cls_loss_causal 0.5214 re_mapping 0.0058 re_causal 0.0146 /// teacc 98.98 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1032, -0.1508, -0.0664, ..., -0.0639, 0.1626, 0.1598], + [-0.1820, -0.1618, -0.1052, ..., -0.1392, -0.1875, -0.1021], + [-0.0798, -0.1135, 0.1172, ..., -0.1733, 0.1839, 0.0626], + ..., + [-0.1078, 0.0642, 0.0485, ..., 0.1682, -0.1708, -0.1867], + [-0.2192, 0.0582, -0.1187, ..., 0.0476, -0.0770, -0.1260], + [ 0.0146, -0.0967, -0.0829, ..., -0.0840, -0.0564, -0.1268]], + device='cuda:0'), grad: tensor([[-4.5635e-06, 3.7253e-09, 1.4901e-08, ..., -4.0740e-05, + -5.4479e-05, -5.6982e-05], + [ 1.1176e-08, 6.5193e-08, 1.3039e-08, ..., 2.3842e-07, + 1.4901e-07, 5.9605e-08], + [ 3.7253e-09, 1.4901e-08, -4.7032e-07, ..., 7.1712e-08, + -2.0918e-06, 2.7008e-08], + ..., + [ 1.3039e-08, -1.0710e-07, 3.1665e-07, ..., -1.6764e-08, + 1.6177e-06, 1.1269e-07], + [ 1.0710e-07, -2.1439e-06, 1.8626e-09, ..., -3.5435e-05, + 1.4994e-07, 1.2945e-07], + [ 4.8429e-08, 2.0489e-08, 9.3132e-10, ..., 3.3993e-07, + 2.4810e-06, 4.8894e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0192, -0.0299, -0.0115, -0.0257, -0.0327, 0.0047, 0.0263, -0.0136, + 0.0370, -0.0003], device='cuda:0'), grad: tensor([-1.4341e-04, 8.2236e-07, -4.9211e-06, 3.6061e-06, -3.3438e-05, + 1.5271e-04, 4.7296e-05, 4.8652e-06, -5.7936e-05, 3.0145e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 250.54, cls_loss 0.0039 cls_loss_mapping 0.0031 cls_loss_causal 0.5036 re_mapping 0.0062 re_causal 0.0148 /// teacc 98.92 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1037, -0.1520, -0.0665, ..., -0.0634, 0.1629, 0.1601], + [-0.1835, -0.1620, -0.1052, ..., -0.1399, -0.1876, -0.1024], + [-0.0800, -0.1140, 0.1173, ..., -0.1736, 0.1842, 0.0626], + ..., + [-0.1083, 0.0646, 0.0485, ..., 0.1685, -0.1712, -0.1868], + [-0.2201, 0.0591, -0.1187, ..., 0.0480, -0.0774, -0.1273], + [ 0.0141, -0.0974, -0.0830, ..., -0.0847, -0.0566, -0.1274]], + device='cuda:0'), grad: tensor([[ 7.1712e-08, 1.1828e-07, 0.0000e+00, ..., 2.7847e-07, + 5.9605e-08, 3.3528e-08], + [ 1.0245e-08, 3.4459e-08, 0.0000e+00, ..., 6.9849e-08, + 6.5193e-09, 3.7253e-09], + [ 1.3039e-08, -2.4773e-06, 0.0000e+00, ..., 2.8871e-08, + 1.3039e-08, -5.9512e-07], + ..., + [ 8.3819e-09, -1.0524e-06, 0.0000e+00, ..., -2.7157e-06, + 9.3132e-10, 2.5146e-08], + [ 4.1444e-07, 2.2352e-06, 0.0000e+00, ..., 1.5367e-07, + 5.3085e-08, 5.7276e-07], + [ 1.5832e-07, 9.3319e-07, -9.3132e-10, ..., 2.2892e-06, + 1.3039e-08, 8.3819e-09]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0192, -0.0300, -0.0116, -0.0258, -0.0328, 0.0043, 0.0278, -0.0136, + 0.0372, -0.0004], device='cuda:0'), grad: tensor([ 6.0163e-07, -4.1537e-07, -2.2709e-05, 5.6960e-06, -1.8626e-08, + -4.9248e-06, -5.8115e-07, -2.2035e-06, 2.1130e-05, 3.4403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 250.55, cls_loss 0.0046 cls_loss_mapping 0.0037 cls_loss_causal 0.4852 re_mapping 0.0061 re_causal 0.0144 /// teacc 99.00 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1041, -0.1541, -0.0666, ..., -0.0639, 0.1629, 0.1601], + [-0.1853, -0.1633, -0.1036, ..., -0.1427, -0.1887, -0.1038], + [-0.0786, -0.1143, 0.1172, ..., -0.1741, 0.1853, 0.0631], + ..., + [-0.1088, 0.0661, 0.0466, ..., 0.1707, -0.1719, -0.1870], + [-0.2210, 0.0592, -0.1185, ..., 0.0480, -0.0778, -0.1288], + [ 0.0135, -0.0987, -0.0834, ..., -0.0851, -0.0568, -0.1282]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + -8.0094e-08, -3.3528e-08], + [ 5.5879e-09, 3.6322e-08, 0.0000e+00, ..., 1.2293e-07, + 9.3132e-10, 0.0000e+00], + [ 1.8626e-09, 8.3819e-09, -0.0000e+00, ..., 2.7940e-08, + -2.8871e-08, 9.3132e-10], + ..., + [ 4.6566e-09, -1.3039e-07, 0.0000e+00, ..., -1.2107e-07, + 3.7253e-09, 9.3132e-10], + [ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., -2.3004e-07, + 5.5879e-09, 2.7940e-09], + [ 1.8626e-09, 6.6124e-08, 9.3132e-09, ..., 6.4261e-08, + 5.2154e-08, 2.1420e-08]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0197, -0.0312, -0.0115, -0.0260, -0.0325, 0.0041, 0.0292, -0.0126, + 0.0369, -0.0003], device='cuda:0'), grad: tensor([-9.2201e-08, -2.7940e-09, 1.1362e-07, -5.6811e-08, -1.5553e-07, + 2.3749e-07, 3.0082e-07, -5.7742e-08, -6.8266e-07, 3.9581e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 250.31, cls_loss 0.0034 cls_loss_mapping 0.0038 cls_loss_causal 0.5119 re_mapping 0.0058 re_causal 0.0149 /// teacc 98.92 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1051, -0.1564, -0.0667, ..., -0.0640, 0.1636, 0.1605], + [-0.1853, -0.1636, -0.1034, ..., -0.1431, -0.1904, -0.1047], + [-0.0788, -0.1149, 0.1194, ..., -0.1745, 0.1867, 0.0633], + ..., + [-0.1093, 0.0667, 0.0457, ..., 0.1721, -0.1735, -0.1880], + [-0.2212, 0.0592, -0.1188, ..., 0.0477, -0.0780, -0.1291], + [ 0.0135, -0.0997, -0.0840, ..., -0.0873, -0.0571, -0.1289]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 9.3132e-09, 0.0000e+00, ..., 7.4506e-09, + -1.4715e-07, -7.8231e-08], + [ 2.7940e-09, 1.9558e-08, 0.0000e+00, ..., 2.7940e-08, + 1.0058e-07, 1.9558e-08], + [ 1.1176e-08, 7.8976e-07, 0.0000e+00, ..., 5.7090e-07, + -1.3355e-06, -3.0454e-07], + ..., + [ 9.3132e-10, -6.4261e-08, 0.0000e+00, ..., -1.6298e-07, + 3.3155e-07, 6.3330e-08], + [ 4.0047e-08, -3.2093e-06, 0.0000e+00, ..., -2.2277e-06, + -1.2610e-06, 5.2154e-08], + [ 1.8626e-09, 2.9802e-08, -9.3132e-10, ..., 7.8231e-08, + 1.7323e-07, 6.7055e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0191, -0.0309, -0.0125, -0.0258, -0.0315, 0.0039, 0.0293, -0.0123, + 0.0368, -0.0016], device='cuda:0'), grad: tensor([-4.6566e-08, -1.1064e-05, 3.7998e-06, 1.2338e-05, 1.2666e-07, + 5.6252e-07, 1.8720e-07, 6.9551e-06, -1.3255e-05, 3.7253e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 250.50, cls_loss 0.0039 cls_loss_mapping 0.0039 cls_loss_causal 0.5498 re_mapping 0.0060 re_causal 0.0154 /// teacc 98.99 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1061, -0.1571, -0.0661, ..., -0.0641, 0.1639, 0.1606], + [-0.1855, -0.1638, -0.1031, ..., -0.1437, -0.1909, -0.1050], + [-0.0791, -0.1151, 0.1194, ..., -0.1747, 0.1874, 0.0632], + ..., + [-0.1105, 0.0648, 0.0456, ..., 0.1722, -0.1742, -0.1884], + [-0.2215, 0.0590, -0.1194, ..., 0.0478, -0.0785, -0.1296], + [ 0.0134, -0.1004, -0.0837, ..., -0.0881, -0.0575, -0.1301]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, 9.3132e-09, 0.0000e+00, ..., 3.7253e-09, + -1.1241e-06, -8.0559e-07], + [ 2.7940e-09, 1.7136e-07, 1.0245e-08, ..., 1.7136e-07, + 6.9756e-07, 8.5682e-08], + [ 1.2573e-07, 8.3819e-08, 0.0000e+00, ..., 4.6566e-09, + -1.7276e-06, -8.1956e-08], + ..., + [ 0.0000e+00, -7.6741e-07, -1.8626e-08, ..., -7.4971e-07, + 5.2061e-07, 8.6613e-08], + [ 1.1548e-07, 4.3213e-07, 2.7940e-09, ..., 8.7544e-08, + 9.5833e-07, 3.1758e-07], + [ 9.3132e-10, 1.1828e-07, 6.6124e-08, ..., 1.4156e-07, + 6.3237e-07, 3.5949e-07]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0190, -0.0309, -0.0124, -0.0241, -0.0320, 0.0035, 0.0296, -0.0128, + 0.0365, -0.0016], device='cuda:0'), grad: tensor([-9.5274e-07, 3.3788e-06, -4.1723e-06, 8.7358e-07, -6.3814e-06, + 3.7346e-07, 8.8103e-07, 1.8440e-07, 3.3714e-06, 2.4475e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 250.51, cls_loss 0.0039 cls_loss_mapping 0.0030 cls_loss_causal 0.5251 re_mapping 0.0058 re_causal 0.0148 /// teacc 98.96 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1071, -0.1584, -0.0661, ..., -0.0645, 0.1635, 0.1609], + [-0.1856, -0.1641, -0.1017, ..., -0.1434, -0.1911, -0.1051], + [-0.0792, -0.1155, 0.1197, ..., -0.1750, 0.1883, 0.0640], + ..., + [-0.1108, 0.0645, 0.0445, ..., 0.1724, -0.1756, -0.1896], + [-0.2222, 0.0590, -0.1199, ..., 0.0478, -0.0791, -0.1306], + [ 0.0131, -0.1005, -0.0846, ..., -0.0888, -0.0571, -0.1321]], + device='cuda:0'), grad: tensor([[ 3.7812e-07, 9.2667e-08, 0.0000e+00, ..., 6.5984e-07, + -1.2042e-06, -2.1402e-06], + [ 4.6520e-07, 1.1288e-06, 0.0000e+00, ..., 3.7439e-06, + 3.0873e-07, 1.0431e-07], + [ 2.5611e-07, 3.3900e-07, 0.0000e+00, ..., 9.1782e-07, + 8.8802e-07, 2.9802e-08], + ..., + [ 8.7218e-07, -4.6119e-06, 0.0000e+00, ..., -1.5207e-05, + 1.6252e-07, 2.4214e-08], + [ 2.2817e-07, 3.0315e-07, 0.0000e+00, ..., 7.9954e-07, + 2.3190e-07, 1.6112e-07], + [ 6.1560e-07, 7.5297e-07, 0.0000e+00, ..., 1.4966e-06, + 3.0966e-07, 2.0862e-07]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0201, -0.0311, -0.0134, -0.0242, -0.0316, 0.0035, 0.0304, -0.0120, + 0.0362, -0.0017], device='cuda:0'), grad: tensor([-6.1467e-07, 1.1489e-05, 3.7327e-06, 7.5549e-06, 3.0193e-06, + 7.8185e-07, 4.1686e-06, 5.2065e-05, 2.0247e-06, -8.4221e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 250.47, cls_loss 0.0037 cls_loss_mapping 0.0038 cls_loss_causal 0.5281 re_mapping 0.0059 re_causal 0.0149 /// teacc 98.98 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1040, -0.1601, -0.0660, ..., -0.0614, 0.1668, 0.1644], + [-0.1884, -0.1642, -0.1003, ..., -0.1431, -0.1916, -0.1063], + [-0.0777, -0.1163, 0.1198, ..., -0.1752, 0.1894, 0.0655], + ..., + [-0.1121, 0.0648, 0.0434, ..., 0.1729, -0.1774, -0.1917], + [-0.2229, 0.0592, -0.1200, ..., 0.0478, -0.0794, -0.1311], + [ 0.0119, -0.1010, -0.0855, ..., -0.0896, -0.0571, -0.1340]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 9.3132e-09, 1.3970e-09, ..., 2.3749e-08, + -1.2107e-08, -9.3132e-09], + [ 5.6345e-08, 7.9628e-08, 0.0000e+00, ..., 3.3062e-08, + 4.6566e-10, 4.6566e-10], + [ 4.2375e-08, 5.8208e-08, 0.0000e+00, ..., 2.3283e-09, + 1.3970e-09, 1.3970e-09], + ..., + [ 1.4808e-07, 2.1560e-07, 2.3283e-09, ..., 6.4820e-07, + 4.6566e-10, 4.6566e-10], + [ 7.7765e-08, -1.0384e-06, 9.4995e-08, ..., -1.2182e-05, + 4.1910e-09, 3.2596e-09], + [ 1.4808e-07, 1.2573e-06, -1.1036e-07, ..., 1.1526e-05, + 3.2596e-09, 2.3283e-09]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0169, -0.0310, -0.0138, -0.0267, -0.0314, 0.0052, 0.0274, -0.0117, + 0.0360, -0.0022], device='cuda:0'), grad: tensor([ 1.5553e-07, 3.3379e-06, 2.0582e-07, -1.6307e-06, -6.2622e-06, + 1.4575e-07, 3.4366e-07, 3.2336e-06, -2.5600e-05, 2.6107e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 250.22, cls_loss 0.0039 cls_loss_mapping 0.0037 cls_loss_causal 0.5058 re_mapping 0.0056 re_causal 0.0145 /// teacc 98.95 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1041, -0.1620, -0.0661, ..., -0.0614, 0.1668, 0.1644], + [-0.1884, -0.1643, -0.1003, ..., -0.1432, -0.1918, -0.1065], + [-0.0778, -0.1170, 0.1198, ..., -0.1760, 0.1899, 0.0658], + ..., + [-0.1132, 0.0650, 0.0435, ..., 0.1734, -0.1782, -0.1925], + [-0.2234, 0.0591, -0.1201, ..., 0.0481, -0.0796, -0.1317], + [ 0.0115, -0.1014, -0.0856, ..., -0.0934, -0.0572, -0.1344]], + device='cuda:0'), grad: tensor([[ 4.2841e-08, 7.4506e-09, 0.0000e+00, ..., 1.0151e-07, + -4.6566e-08, -2.7474e-08], + [ 5.1223e-09, 1.3784e-06, 0.0000e+00, ..., 2.1439e-06, + 1.8161e-08, 2.3283e-09], + [ 7.4506e-09, 1.1176e-08, 0.0000e+00, ..., 3.5577e-07, + -1.9535e-05, 4.6566e-09], + ..., + [ 1.3970e-09, -1.4370e-06, 0.0000e+00, ..., -2.2221e-06, + 1.9046e-07, 1.3970e-09], + [ 5.5879e-09, 2.2817e-08, 0.0000e+00, ..., -1.2415e-06, + 1.0710e-08, 6.9849e-09], + [ 2.3283e-09, -2.7800e-07, 0.0000e+00, ..., -6.2399e-07, + 2.8405e-08, 2.0023e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0170, -0.0305, -0.0140, -0.0269, -0.0277, 0.0054, 0.0275, -0.0120, + 0.0361, -0.0055], device='cuda:0'), grad: tensor([ 3.0780e-07, 3.5930e-06, -9.9361e-05, 6.1467e-08, 1.0514e-04, + 2.1271e-06, -3.1665e-07, -2.5071e-06, -3.7570e-06, -5.3719e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 250.35, cls_loss 0.0046 cls_loss_mapping 0.0041 cls_loss_causal 0.5307 re_mapping 0.0058 re_causal 0.0146 /// teacc 99.02 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1044, -0.1657, -0.0661, ..., -0.0616, 0.1667, 0.1643], + [-0.1888, -0.1647, -0.1003, ..., -0.1445, -0.1918, -0.1070], + [-0.0779, -0.1174, 0.1199, ..., -0.1770, 0.1921, 0.0668], + ..., + [-0.1141, 0.0648, 0.0435, ..., 0.1724, -0.1808, -0.1944], + [-0.2239, 0.0603, -0.1202, ..., 0.0494, -0.0798, -0.1314], + [ 0.0116, -0.1013, -0.0856, ..., -0.0934, -0.0574, -0.1350]], + device='cuda:0'), grad: tensor([[ 1.2144e-06, 2.9337e-08, 1.3970e-09, ..., 8.0094e-08, + 2.7735e-06, 1.9260e-06], + [ 1.0245e-08, 1.8300e-07, 5.5879e-09, ..., 4.8382e-07, + 1.3504e-08, 1.0710e-08], + [ 4.6566e-09, 9.3132e-10, 4.6566e-10, ..., 4.1910e-08, + 3.2596e-09, 3.2596e-09], + ..., + [ 4.6566e-10, -6.4028e-07, 4.6566e-10, ..., -1.7509e-06, + 2.3283e-09, 1.3970e-09], + [ 8.1630e-07, 2.3283e-09, 9.3132e-10, ..., -2.4540e-07, + 7.2876e-07, 6.7614e-07], + [ 6.0536e-09, 3.3900e-07, -9.3132e-09, ..., 9.6299e-07, + 1.1176e-08, 8.3819e-09]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0174, -0.0304, -0.0140, -0.0267, -0.0271, 0.0052, 0.0276, -0.0129, + 0.0373, -0.0056], device='cuda:0'), grad: tensor([ 7.5847e-06, -5.0617e-07, 1.8906e-07, 5.8673e-08, 6.7288e-07, + 4.2245e-06, -1.3456e-05, -3.0585e-06, 1.7080e-06, 2.5816e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 250.61, cls_loss 0.0038 cls_loss_mapping 0.0030 cls_loss_causal 0.5319 re_mapping 0.0056 re_causal 0.0147 /// teacc 99.03 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1047, -0.1677, -0.0661, ..., -0.0617, 0.1665, 0.1642], + [-0.1889, -0.1652, -0.1003, ..., -0.1452, -0.1921, -0.1074], + [-0.0780, -0.1185, 0.1204, ..., -0.1775, 0.1930, 0.0673], + ..., + [-0.1149, 0.0651, 0.0438, ..., 0.1730, -0.1817, -0.1952], + [-0.2242, 0.0604, -0.1208, ..., 0.0494, -0.0801, -0.1308], + [ 0.0113, -0.1020, -0.0861, ..., -0.0935, -0.0577, -0.1365]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 1.3970e-09, 4.6566e-10, ..., -3.4459e-08, + -7.3574e-08, -3.4925e-08], + [ 1.3970e-09, 2.3283e-09, 2.7940e-09, ..., 6.0536e-09, + 3.1432e-07, 2.3702e-07], + [ 1.3970e-09, 1.2573e-08, 4.6566e-10, ..., 1.7229e-08, + -8.5169e-07, -6.4867e-07], + ..., + [ 4.6566e-10, -2.2817e-08, 1.0924e-06, ..., 8.7963e-07, + 3.0501e-07, 1.9325e-07], + [ 8.3819e-09, 1.8626e-09, 4.6566e-10, ..., 1.4435e-08, + 1.0151e-07, 7.4506e-08], + [ 9.3132e-10, 3.7253e-09, -1.1036e-06, ..., -9.9652e-07, + 2.8871e-08, 1.6298e-08]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0175, -0.0304, -0.0143, -0.0264, -0.0271, 0.0051, 0.0278, -0.0128, + 0.0377, -0.0057], device='cuda:0'), grad: tensor([-2.8871e-08, -6.8903e-05, 3.8624e-05, 4.3726e-07, 3.3248e-07, + 7.4785e-07, 1.3430e-06, 3.4124e-05, 3.9674e-07, -7.0482e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 250.27, cls_loss 0.0037 cls_loss_mapping 0.0029 cls_loss_causal 0.5419 re_mapping 0.0059 re_causal 0.0153 /// teacc 99.05 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1049, -0.1695, -0.0661, ..., -0.0617, 0.1666, 0.1642], + [-0.1891, -0.1655, -0.1003, ..., -0.1457, -0.1925, -0.1078], + [-0.0784, -0.1193, 0.1204, ..., -0.1780, 0.1934, 0.0672], + ..., + [-0.1157, 0.0643, 0.0438, ..., 0.1717, -0.1827, -0.1958], + [-0.2245, 0.0623, -0.1208, ..., 0.0510, -0.0804, -0.1306], + [ 0.0108, -0.1025, -0.0861, ..., -0.0932, -0.0577, -0.1371]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.0245e-08, 0.0000e+00, ..., 6.5193e-09, + -3.8184e-08, -2.4680e-08], + [ 5.1223e-09, 5.4017e-08, 0.0000e+00, ..., 4.6100e-08, + 7.1712e-08, 1.3970e-09], + [ 2.3283e-09, 4.0000e-07, 0.0000e+00, ..., 4.2841e-07, + -1.2713e-07, 2.3283e-08], + ..., + [ 2.3749e-08, -4.8056e-07, 0.0000e+00, ..., -6.2212e-07, + 5.4482e-08, 3.7253e-09], + [ 2.7474e-08, 1.4948e-07, 0.0000e+00, ..., 5.9605e-08, + -4.5169e-08, -6.4261e-08], + [ 5.5879e-09, 7.6368e-08, 4.6566e-10, ..., 7.3574e-08, + 1.2573e-08, 8.3819e-09]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0176, -0.0295, -0.0161, -0.0260, -0.0273, 0.0050, 0.0278, -0.0137, + 0.0391, -0.0054], device='cuda:0'), grad: tensor([ 6.4727e-08, 2.8731e-07, 5.0198e-07, -5.8999e-07, -4.5635e-07, + 3.3621e-07, 1.5553e-07, -7.4459e-07, 7.4506e-09, 4.5449e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 250.37, cls_loss 0.0033 cls_loss_mapping 0.0025 cls_loss_causal 0.5043 re_mapping 0.0059 re_causal 0.0149 /// teacc 98.94 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1049, -0.1707, -0.0662, ..., -0.0617, 0.1666, 0.1642], + [-0.1905, -0.1655, -0.0998, ..., -0.1456, -0.1929, -0.1092], + [-0.0772, -0.1196, 0.1203, ..., -0.1782, 0.1943, 0.0688], + ..., + [-0.1162, 0.0644, 0.0436, ..., 0.1719, -0.1834, -0.1965], + [-0.2249, 0.0623, -0.1209, ..., 0.0511, -0.0807, -0.1311], + [ 0.0109, -0.1030, -0.0855, ..., -0.0935, -0.0576, -0.1370]], + device='cuda:0'), grad: tensor([[ 1.6689e-05, 1.1176e-08, 0.0000e+00, ..., 2.0955e-08, + 4.3064e-05, 2.6360e-05], + [ 6.0536e-09, 2.2864e-07, 0.0000e+00, ..., 4.0326e-07, + 5.8673e-08, 1.5367e-08], + [ 4.2375e-08, 5.1111e-06, 0.0000e+00, ..., 9.6783e-06, + 1.0515e-06, 8.4285e-08], + ..., + [ 1.3970e-09, -5.4948e-06, 4.6566e-10, ..., -1.0267e-05, + -9.5181e-07, 5.5879e-09], + [ 9.0804e-08, 1.6671e-07, 1.3970e-09, ..., -3.0734e-08, + 3.5902e-07, 2.2957e-07], + [ 5.1223e-08, 6.7987e-08, -3.2596e-09, ..., -1.4622e-07, + 5.7789e-07, 3.7719e-07]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0176, -0.0297, -0.0162, -0.0259, -0.0271, 0.0049, 0.0278, -0.0134, + 0.0386, -0.0055], device='cuda:0'), grad: tensor([ 7.8917e-05, 9.4483e-07, 2.2382e-05, -2.7847e-07, 6.5472e-07, + 6.3377e-07, -8.1599e-05, -2.3261e-05, 6.0257e-07, 8.8522e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 250.29, cls_loss 0.0034 cls_loss_mapping 0.0039 cls_loss_causal 0.5202 re_mapping 0.0056 re_causal 0.0142 /// teacc 98.96 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1052, -0.1720, -0.0657, ..., -0.0617, 0.1665, 0.1642], + [-0.1908, -0.1657, -0.0975, ..., -0.1454, -0.1930, -0.1095], + [-0.0773, -0.1196, 0.1201, ..., -0.1786, 0.1951, 0.0704], + ..., + [-0.1164, 0.0649, 0.0428, ..., 0.1729, -0.1843, -0.1977], + [-0.2252, 0.0621, -0.1214, ..., 0.0511, -0.0814, -0.1328], + [ 0.0102, -0.1044, -0.0879, ..., -0.0945, -0.0577, -0.1376]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 6.5193e-09, 1.0245e-08, ..., 1.3039e-08, + -6.5845e-07, -4.1351e-07], + [ 4.6566e-09, 1.9372e-07, 3.7253e-09, ..., 5.0291e-07, + 1.8161e-07, 1.9744e-07], + [ 4.6566e-09, 1.2107e-08, 0.0000e+00, ..., 1.7695e-08, + 4.6566e-09, 5.5879e-09], + ..., + [ 3.7253e-09, -1.0179e-06, 5.7742e-08, ..., -2.7418e-06, + 1.1176e-08, 4.6566e-09], + [ 8.5682e-08, 6.5193e-08, 1.5832e-08, ..., 1.1455e-07, + 4.0978e-08, 4.0047e-08], + [ 2.9802e-08, 7.4878e-07, 2.1420e-08, ..., 1.9483e-06, + 4.1071e-07, 1.4622e-07]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0177, -0.0293, -0.0160, -0.0260, -0.0272, 0.0049, 0.0279, -0.0132, + 0.0384, -0.0059], device='cuda:0'), grad: tensor([-1.0785e-06, 1.6987e-06, 6.2399e-08, 8.1025e-08, -1.0647e-05, + 6.3237e-07, 9.7975e-06, -5.2191e-06, 9.4622e-07, 3.6806e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 250.41, cls_loss 0.0032 cls_loss_mapping 0.0037 cls_loss_causal 0.5289 re_mapping 0.0055 re_causal 0.0147 /// teacc 99.05 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1053, -0.1730, -0.0658, ..., -0.0617, 0.1666, 0.1642], + [-0.1912, -0.1658, -0.0975, ..., -0.1449, -0.1933, -0.1093], + [-0.0773, -0.1197, 0.1205, ..., -0.1788, 0.1963, 0.0716], + ..., + [-0.1167, 0.0650, 0.0427, ..., 0.1728, -0.1856, -0.1998], + [-0.2253, 0.0623, -0.1217, ..., 0.0514, -0.0815, -0.1327], + [ 0.0105, -0.1052, -0.0879, ..., -0.0947, -0.0578, -0.1392]], + device='cuda:0'), grad: tensor([[-6.2399e-08, 1.0245e-08, 0.0000e+00, ..., 2.0489e-08, + -4.8708e-07, -3.1292e-07], + [ 5.5879e-09, 9.9652e-08, 0.0000e+00, ..., 7.7672e-07, + 1.3039e-08, 1.3039e-08], + [ 0.0000e+00, -1.1176e-08, -0.0000e+00, ..., -1.8626e-09, + -9.4716e-07, -9.3970e-07], + ..., + [ 4.6566e-09, -3.2410e-07, 0.0000e+00, ..., -6.1002e-07, + 8.8383e-07, 8.8848e-07], + [ 8.3819e-08, 3.5390e-08, 0.0000e+00, ..., 2.0117e-07, + 2.7008e-08, 1.8626e-08], + [ 1.8626e-08, 2.4494e-07, 0.0000e+00, ..., 8.7079e-07, + 8.3819e-09, 5.5879e-09]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0178, -0.0295, -0.0162, -0.0258, -0.0274, 0.0047, 0.0278, -0.0127, + 0.0389, -0.0060], device='cuda:0'), grad: tensor([-5.6159e-07, 4.5002e-06, -2.3805e-06, -4.3772e-08, -4.8429e-06, + -3.0268e-07, 7.9069e-07, 1.1912e-06, 9.7696e-07, 6.5938e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 250.50, cls_loss 0.0028 cls_loss_mapping 0.0032 cls_loss_causal 0.5151 re_mapping 0.0059 re_causal 0.0149 /// teacc 99.00 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1053, -0.1737, -0.0658, ..., -0.0617, 0.1665, 0.1643], + [-0.1916, -0.1660, -0.0975, ..., -0.1450, -0.1940, -0.1101], + [-0.0774, -0.1203, 0.1206, ..., -0.1793, 0.1984, 0.0731], + ..., + [-0.1173, 0.0650, 0.0422, ..., 0.1730, -0.1885, -0.2022], + [-0.2257, 0.0626, -0.1219, ..., 0.0521, -0.0816, -0.1335], + [ 0.0099, -0.1054, -0.0874, ..., -0.0949, -0.0568, -0.1401]], + device='cuda:0'), grad: tensor([[ 1.8766e-06, 6.5193e-09, 6.9849e-08, ..., 3.0361e-07, + 1.5041e-06, 1.1232e-06], + [ 2.7008e-08, 7.0781e-08, 7.8231e-08, ..., 3.5204e-07, + 1.0245e-08, 5.5879e-09], + [ 4.6566e-08, 1.0151e-07, -5.1372e-06, ..., -2.3201e-05, + 4.7497e-08, 2.4214e-08], + ..., + [ 4.8429e-08, 1.4715e-07, 1.0822e-06, ..., 4.8839e-06, + 1.5832e-08, 9.3132e-09], + [ 3.5204e-07, 3.4459e-08, 2.2557e-06, ..., 1.0185e-05, + 2.9057e-07, 2.1420e-07], + [ 8.3819e-09, 1.4901e-08, -0.0000e+00, ..., 3.7253e-09, + 8.3819e-09, 5.5879e-09]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0179, -0.0294, -0.0160, -0.0256, -0.0274, 0.0043, 0.0278, -0.0131, + 0.0395, -0.0057], device='cuda:0'), grad: tensor([ 5.7817e-06, 1.4259e-06, -1.0556e-04, -2.8778e-07, 2.4259e-05, + 8.4266e-06, -4.1388e-06, 2.2650e-05, 4.7475e-05, 6.3330e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 250.30, cls_loss 0.0040 cls_loss_mapping 0.0038 cls_loss_causal 0.5513 re_mapping 0.0054 re_causal 0.0140 /// teacc 99.00 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1053, -0.1761, -0.0658, ..., -0.0618, 0.1667, 0.1643], + [-0.1929, -0.1667, -0.0975, ..., -0.1476, -0.1941, -0.1101], + [-0.0775, -0.1208, 0.1210, ..., -0.1796, 0.1994, 0.0732], + ..., + [-0.1192, 0.0653, 0.0421, ..., 0.1754, -0.1894, -0.2028], + [-0.2257, 0.0625, -0.1227, ..., 0.0523, -0.0815, -0.1340], + [ 0.0097, -0.1059, -0.0874, ..., -0.0946, -0.0573, -0.1411]], + device='cuda:0'), grad: tensor([[ 3.7067e-07, 1.2014e-07, 0.0000e+00, ..., 2.7940e-09, + 4.5635e-07, 3.9581e-07], + [ 1.0040e-06, 2.6971e-06, 0.0000e+00, ..., 1.3970e-08, + 1.0710e-07, 8.9407e-08], + [ 3.9302e-07, 4.7870e-07, 0.0000e+00, ..., -9.3132e-09, + -3.7625e-07, -2.4773e-07], + ..., + [ 1.0105e-06, 2.8610e-06, 0.0000e+00, ..., 5.0291e-08, + 1.9744e-07, 1.4529e-07], + [ 2.6241e-05, 7.5530e-07, 0.0000e+00, ..., 1.2480e-07, + 2.7165e-05, 2.4363e-05], + [ 3.5390e-07, 8.2795e-07, 0.0000e+00, ..., 2.2352e-08, + 7.7300e-08, 6.5193e-08]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0179, -0.0308, -0.0160, -0.0259, -0.0272, 0.0045, 0.0277, -0.0121, + 0.0399, -0.0057], device='cuda:0'), grad: tensor([ 1.8040e-06, 6.7204e-06, 5.5786e-07, -2.1264e-05, 4.0047e-08, + 2.2814e-05, -1.1659e-04, 7.3835e-06, 9.6858e-05, 1.6643e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 250.41, cls_loss 0.0030 cls_loss_mapping 0.0029 cls_loss_causal 0.5339 re_mapping 0.0057 re_causal 0.0149 /// teacc 98.96 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1058, -0.1774, -0.0660, ..., -0.0618, 0.1660, 0.1642], + [-0.1934, -0.1669, -0.0974, ..., -0.1476, -0.1944, -0.1103], + [-0.0777, -0.1212, 0.1211, ..., -0.1799, 0.2002, 0.0739], + ..., + [-0.1199, 0.0656, 0.0421, ..., 0.1757, -0.1896, -0.2030], + [-0.2261, 0.0624, -0.1228, ..., 0.0522, -0.0826, -0.1357], + [ 0.0116, -0.1065, -0.0874, ..., -0.0949, -0.0553, -0.1420]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + -1.3284e-05, -8.0019e-06], + [ 0.0000e+00, 6.5193e-09, 0.0000e+00, ..., 1.8626e-08, + 1.0151e-07, 3.0734e-08], + [ 0.0000e+00, 1.5832e-08, 0.0000e+00, ..., 3.6322e-08, + -9.4622e-07, -2.2538e-07], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.7497e-07, + 2.4214e-08, 8.3819e-09], + [ 2.7940e-09, -7.6368e-08, 0.0000e+00, ..., -1.3318e-07, + -2.3283e-08, -6.4261e-08], + [ 5.5879e-09, -3.6322e-08, 0.0000e+00, ..., -5.3458e-07, + 1.1176e-07, 6.8918e-08]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0190, -0.0306, -0.0159, -0.0261, -0.0279, 0.0042, 0.0283, -0.0121, + 0.0395, -0.0044], device='cuda:0'), grad: tensor([-1.6391e-05, 3.1013e-07, -2.0731e-06, 1.3057e-06, -8.1770e-07, + -5.9605e-08, 1.6987e-05, 4.3772e-06, -3.5297e-07, -3.2559e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 250.44, cls_loss 0.0026 cls_loss_mapping 0.0025 cls_loss_causal 0.5000 re_mapping 0.0056 re_causal 0.0146 /// teacc 99.02 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1059, -0.1788, -0.0662, ..., -0.0618, 0.1659, 0.1642], + [-0.1948, -0.1671, -0.0975, ..., -0.1478, -0.1944, -0.1117], + [-0.0763, -0.1215, 0.1211, ..., -0.1807, 0.2003, 0.0751], + ..., + [-0.1205, 0.0646, 0.0421, ..., 0.1758, -0.1899, -0.2032], + [-0.2263, 0.0625, -0.1231, ..., 0.0524, -0.0827, -0.1359], + [ 0.0121, -0.1072, -0.0874, ..., -0.0951, -0.0549, -0.1432]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., -6.7055e-08, + -6.2305e-07, -6.6962e-07], + [ 2.7940e-09, 1.4901e-08, 0.0000e+00, ..., 1.1176e-08, + 4.0047e-08, 3.8184e-08], + [ 2.0489e-08, 5.0291e-08, 0.0000e+00, ..., 3.6322e-08, + 1.1362e-07, 4.6566e-08], + ..., + [ 4.6566e-09, 1.6764e-08, 0.0000e+00, ..., 5.5879e-09, + 1.1176e-08, 2.7940e-09], + [-3.5390e-08, 1.8626e-09, 0.0000e+00, ..., -2.2314e-06, + -1.9651e-07, -7.3574e-08], + [ 9.3132e-10, -1.3970e-08, 0.0000e+00, ..., -3.6322e-08, + 4.6566e-08, 2.0489e-08]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0192, -0.0306, -0.0159, -0.0247, -0.0280, 0.0031, 0.0284, -0.0122, + 0.0398, -0.0043], device='cuda:0'), grad: tensor([-1.4063e-06, 1.1828e-07, 8.3074e-07, 1.1269e-07, 4.1910e-08, + 2.4028e-06, 6.2771e-06, 1.3132e-07, -8.0392e-06, -4.7963e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 250.53, cls_loss 0.0027 cls_loss_mapping 0.0028 cls_loss_causal 0.4744 re_mapping 0.0058 re_causal 0.0143 /// teacc 98.98 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1059, -0.1802, -0.0663, ..., -0.0618, 0.1660, 0.1643], + [-0.1956, -0.1672, -0.0974, ..., -0.1479, -0.1946, -0.1120], + [-0.0762, -0.1221, 0.1209, ..., -0.1812, 0.2005, 0.0751], + ..., + [-0.1211, 0.0646, 0.0421, ..., 0.1761, -0.1904, -0.2034], + [-0.2270, 0.0624, -0.1232, ..., 0.0524, -0.0829, -0.1362], + [ 0.0116, -0.1075, -0.0874, ..., -0.0954, -0.0551, -0.1440]], + device='cuda:0'), grad: tensor([[ 7.3761e-06, 1.0245e-08, 0.0000e+00, ..., 8.0466e-06, + -7.5437e-08, -2.7008e-08], + [ 8.5682e-08, 1.6205e-07, 0.0000e+00, ..., 3.3993e-07, + 0.0000e+00, 0.0000e+00], + [ 9.0338e-08, 2.6356e-07, 0.0000e+00, ..., 5.0478e-07, + -1.8626e-09, -9.3132e-10], + ..., + [ 1.2293e-07, -1.0887e-06, 9.3132e-10, ..., -1.5302e-06, + 2.7940e-09, 9.3132e-10], + [ 2.7940e-08, 3.7253e-08, 0.0000e+00, ..., 5.8673e-08, + 2.7940e-09, 1.8626e-09], + [ 3.7625e-07, 5.6438e-07, -2.7940e-09, ..., 1.2740e-06, + 6.6124e-08, 2.4214e-08]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0191, -0.0304, -0.0161, -0.0246, -0.0281, 0.0028, 0.0287, -0.0123, + 0.0397, -0.0043], device='cuda:0'), grad: tensor([ 2.3037e-05, 9.0525e-07, 1.5497e-06, 8.5309e-06, -4.0978e-08, + -3.5048e-05, 1.5767e-06, -4.6566e-06, 1.5926e-07, 4.0308e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 250.57, cls_loss 0.0029 cls_loss_mapping 0.0029 cls_loss_causal 0.4978 re_mapping 0.0052 re_causal 0.0139 /// teacc 99.05 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1061, -0.1810, -0.0664, ..., -0.0618, 0.1661, 0.1643], + [-0.1960, -0.1675, -0.0973, ..., -0.1481, -0.1950, -0.1123], + [-0.0759, -0.1224, 0.1210, ..., -0.1813, 0.2019, 0.0757], + ..., + [-0.1218, 0.0643, 0.0420, ..., 0.1761, -0.1921, -0.2040], + [-0.2279, 0.0623, -0.1236, ..., 0.0522, -0.0832, -0.1368], + [ 0.0098, -0.1081, -0.0874, ..., -0.0954, -0.0553, -0.1450]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.5832e-08, 0.0000e+00, ..., 2.1420e-08, + -1.9930e-07, -1.3225e-07], + [ 4.6566e-08, 1.2666e-07, 1.8626e-09, ..., 2.9989e-07, + 8.2888e-08, 9.3132e-10], + [ 9.3132e-10, 7.4320e-07, 4.6566e-09, ..., 1.2787e-06, + -2.3842e-07, 6.5193e-09], + ..., + [ 1.2107e-08, -2.5220e-06, 0.0000e+00, ..., -4.1313e-06, + 1.4994e-07, 2.7940e-09], + [ 6.3144e-07, -2.2724e-07, 0.0000e+00, ..., -1.9744e-07, + 1.3970e-08, 7.4506e-09], + [ 3.0641e-07, 1.5534e-06, 0.0000e+00, ..., 2.7921e-06, + 5.2154e-08, 3.1665e-08]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0191, -0.0306, -0.0156, -0.0241, -0.0278, 0.0032, 0.0286, -0.0125, + 0.0395, -0.0045], device='cuda:0'), grad: tensor([-2.7474e-07, 1.6801e-06, 1.6131e-06, 4.8280e-06, 2.9430e-06, + -4.9658e-06, 4.7497e-07, -6.0201e-06, -1.1539e-06, 8.4750e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 226---------------------------------------------------- +epoch 226, time 267.53, cls_loss 0.0036 cls_loss_mapping 0.0027 cls_loss_causal 0.4922 re_mapping 0.0055 re_causal 0.0137 /// teacc 99.07 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.1067, -0.1835, -0.0664, ..., -0.0618, 0.1659, 0.1641], + [-0.1962, -0.1681, -0.0973, ..., -0.1486, -0.1955, -0.1128], + [-0.0761, -0.1223, 0.1212, ..., -0.1812, 0.2032, 0.0762], + ..., + [-0.1211, 0.0648, 0.0420, ..., 0.1770, -0.1934, -0.2051], + [-0.2271, 0.0621, -0.1261, ..., 0.0526, -0.0832, -0.1370], + [ 0.0075, -0.1097, -0.0874, ..., -0.0960, -0.0554, -0.1458]], + device='cuda:0'), grad: tensor([[ 8.2050e-07, 1.0245e-08, 0.0000e+00, ..., 2.3283e-08, + 5.6252e-07, 2.4494e-07], + [ 1.3039e-08, 1.5832e-08, 0.0000e+00, ..., 1.5274e-07, + 6.5193e-09, 6.5193e-09], + [ 1.1176e-08, -2.5705e-07, 0.0000e+00, ..., 6.5193e-09, + -1.1548e-07, -2.8405e-07], + ..., + [ 3.9022e-07, 2.7940e-09, 0.0000e+00, ..., 1.6754e-06, + 2.1420e-08, 2.8871e-08], + [ 3.3714e-07, 1.0710e-07, 0.0000e+00, ..., 4.7404e-07, + 1.0524e-07, 8.9407e-08], + [-2.6580e-06, 2.9802e-08, -0.0000e+00, ..., -1.3158e-05, + 5.5879e-08, 3.6322e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0194, -0.0308, -0.0152, -0.0237, -0.0275, 0.0028, 0.0284, -0.0121, + 0.0401, -0.0051], device='cuda:0'), grad: tensor([ 1.2778e-06, 6.5938e-07, -1.2312e-06, 9.7509e-07, 3.2127e-05, + 2.2620e-05, -8.7246e-06, 8.0094e-06, 3.1181e-06, -5.8800e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 250.63, cls_loss 0.0026 cls_loss_mapping 0.0030 cls_loss_causal 0.5020 re_mapping 0.0058 re_causal 0.0149 /// teacc 99.03 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.1067, -0.1840, -0.0663, ..., -0.0618, 0.1659, 0.1641], + [-0.1967, -0.1694, -0.0973, ..., -0.1502, -0.1959, -0.1139], + [-0.0759, -0.1226, 0.1212, ..., -0.1819, 0.2036, 0.0772], + ..., + [-0.1210, 0.0659, 0.0419, ..., 0.1790, -0.1935, -0.2053], + [-0.2272, 0.0616, -0.1262, ..., 0.0526, -0.0832, -0.1372], + [ 0.0067, -0.1114, -0.0874, ..., -0.0969, -0.0554, -0.1461]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.1176e-08, 1.1902e-06, ..., 1.3970e-08, + 7.5884e-06, 3.2689e-07], + [ 1.5832e-08, 2.0489e-08, 7.4506e-09, ..., 1.7695e-08, + 5.1223e-08, 6.5193e-09], + [ 7.4506e-09, 6.5193e-09, -1.5264e-06, ..., 1.6764e-08, + -9.8348e-06, -4.5821e-07], + ..., + [ 8.3819e-09, -8.0187e-07, 4.0047e-08, ..., -9.7230e-07, + 2.7847e-07, 2.8871e-08], + [ 2.2352e-08, 9.4995e-08, 4.3772e-08, ..., 7.2643e-08, + 2.7567e-07, 1.7695e-08], + [-5.2713e-07, 6.5006e-07, 5.0291e-08, ..., -1.7034e-06, + 3.6508e-07, 2.1420e-08]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0194, -0.0313, -0.0147, -0.0252, -0.0275, 0.0040, 0.0284, -0.0113, + 0.0397, -0.0055], device='cuda:0'), grad: tensor([ 1.4089e-05, 1.7323e-07, -1.8209e-05, -2.8796e-06, 2.7418e-06, + 2.0325e-05, 4.4703e-08, -1.6596e-06, 7.4506e-07, -1.5363e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 250.54, cls_loss 0.0029 cls_loss_mapping 0.0032 cls_loss_causal 0.4890 re_mapping 0.0055 re_causal 0.0140 /// teacc 98.99 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.1068, -0.1848, -0.0661, ..., -0.0618, 0.1660, 0.1642], + [-0.1974, -0.1695, -0.0973, ..., -0.1505, -0.1970, -0.1148], + [-0.0756, -0.1237, 0.1218, ..., -0.1832, 0.2048, 0.0780], + ..., + [-0.1213, 0.0661, 0.0419, ..., 0.1794, -0.1939, -0.2056], + [-0.2286, 0.0615, -0.1264, ..., 0.0524, -0.0835, -0.1377], + [ 0.0067, -0.1118, -0.0874, ..., -0.0970, -0.0556, -0.1465]], + device='cuda:0'), grad: tensor([[-3.0734e-08, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + -1.8533e-07, -6.4261e-08], + [ 1.7136e-07, 1.3690e-07, 0.0000e+00, ..., 3.5856e-07, + 9.7789e-08, 1.2107e-07], + [ 1.5832e-08, 3.7253e-09, 0.0000e+00, ..., -4.9658e-06, + -1.6868e-05, -1.4544e-05], + ..., + [ 1.8626e-09, -1.4715e-07, 0.0000e+00, ..., -3.8557e-07, + 1.6764e-08, 1.0245e-08], + [ 1.4342e-07, 1.8626e-09, 0.0000e+00, ..., 4.9621e-06, + 1.6913e-05, 1.4596e-05], + [ 1.2107e-08, 5.5879e-09, 0.0000e+00, ..., 1.4901e-08, + 5.9605e-08, 2.5146e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0193, -0.0312, -0.0146, -0.0252, -0.0285, 0.0040, 0.0286, -0.0113, + 0.0391, -0.0046], device='cuda:0'), grad: tensor([-2.8778e-07, 1.2117e-06, -7.1824e-05, 5.2527e-07, -4.6566e-08, + 4.5542e-07, -1.7127e-06, -6.5193e-07, 7.2122e-05, 1.2200e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 250.26, cls_loss 0.0025 cls_loss_mapping 0.0027 cls_loss_causal 0.5179 re_mapping 0.0053 re_causal 0.0144 /// teacc 98.97 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.1068, -0.1854, -0.0662, ..., -0.0618, 0.1661, 0.1642], + [-0.1995, -0.1696, -0.0962, ..., -0.1507, -0.1973, -0.1150], + [-0.0757, -0.1249, 0.1218, ..., -0.1834, 0.2049, 0.0780], + ..., + [-0.1226, 0.0659, 0.0406, ..., 0.1794, -0.1943, -0.2060], + [-0.2297, 0.0620, -0.1264, ..., 0.0520, -0.0841, -0.1400], + [ 0.0065, -0.1123, -0.0881, ..., -0.0973, -0.0557, -0.1470]], + device='cuda:0'), grad: tensor([[ 1.0338e-07, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + -6.7987e-08, -5.3085e-08], + [ 8.3819e-09, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-09, 2.7940e-09], + [ 6.5193e-09, -9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 5.5879e-09, 5.5879e-09], + ..., + [ 2.6077e-08, 1.8626e-09, 0.0000e+00, ..., 6.5193e-09, + 2.9802e-08, 2.1420e-08], + [ 3.8184e-08, 2.7940e-09, 0.0000e+00, ..., -4.6566e-09, + 9.3132e-09, 6.5193e-09], + [ 3.2037e-07, -1.8626e-09, 0.0000e+00, ..., 6.9849e-08, + 6.1467e-08, 4.2841e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0193, -0.0312, -0.0148, -0.0251, -0.0287, 0.0053, 0.0285, -0.0114, + 0.0384, -0.0047], device='cuda:0'), grad: tensor([ 3.6322e-08, -2.0295e-05, 2.7455e-06, 3.3051e-05, 8.5682e-08, + -2.9787e-05, -1.8626e-08, 8.6278e-06, 1.6950e-07, 5.4426e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 250.41, cls_loss 0.0030 cls_loss_mapping 0.0026 cls_loss_causal 0.5130 re_mapping 0.0056 re_causal 0.0139 /// teacc 98.96 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.1069, -0.1884, -0.0662, ..., -0.0619, 0.1664, 0.1643], + [-0.2000, -0.1698, -0.0962, ..., -0.1509, -0.1989, -0.1159], + [-0.0756, -0.1256, 0.1219, ..., -0.1842, 0.2050, 0.0780], + ..., + [-0.1243, 0.0664, 0.0407, ..., 0.1799, -0.1947, -0.2062], + [-0.2303, 0.0620, -0.1267, ..., 0.0522, -0.0840, -0.1400], + [ 0.0064, -0.1137, -0.0882, ..., -0.0983, -0.0558, -0.1478]], + device='cuda:0'), grad: tensor([[-3.7253e-09, 2.3283e-08, 0.0000e+00, ..., 1.0151e-07, + -2.9709e-07, -1.2945e-07], + [ 5.5879e-08, 2.9337e-07, 0.0000e+00, ..., 8.3819e-09, + 9.3132e-10, 0.0000e+00], + [ 3.0175e-07, 1.5795e-06, 0.0000e+00, ..., 6.4354e-07, + 2.2352e-07, 7.0781e-08], + ..., + [ 1.8626e-08, 7.3574e-08, 0.0000e+00, ..., -9.3132e-09, + 3.7253e-09, 1.8626e-09], + [ 2.8051e-06, 8.0839e-07, 0.0000e+00, ..., 1.1921e-06, + -8.1025e-07, -2.4959e-07], + [ 4.9360e-08, 5.4017e-08, 0.0000e+00, ..., 1.3970e-08, + 2.2352e-08, 1.0245e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0190, -0.0314, -0.0153, -0.0257, -0.0284, 0.0055, 0.0285, -0.0111, + 0.0386, -0.0051], device='cuda:0'), grad: tensor([ 2.2352e-08, 4.8708e-07, 4.4107e-06, 9.6709e-06, 1.2759e-07, + -2.2948e-05, 5.0291e-07, 1.7975e-07, 7.5437e-06, -6.5193e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 250.38, cls_loss 0.0027 cls_loss_mapping 0.0025 cls_loss_causal 0.5024 re_mapping 0.0056 re_causal 0.0144 /// teacc 98.99 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.1072, -0.1899, -0.0670, ..., -0.0621, 0.1663, 0.1643], + [-0.2010, -0.1699, -0.0962, ..., -0.1512, -0.1993, -0.1159], + [-0.0756, -0.1264, 0.1226, ..., -0.1849, 0.2061, 0.0784], + ..., + [-0.1249, 0.0639, 0.0405, ..., 0.1777, -0.1962, -0.2075], + [-0.2314, 0.0619, -0.1271, ..., 0.0520, -0.0843, -0.1405], + [ 0.0060, -0.1108, -0.0881, ..., -0.0955, -0.0559, -0.1481]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -3.7253e-08, 0.0000e+00, ..., 2.7940e-09, + -3.9767e-07, -2.4494e-07], + [ 7.4506e-09, 2.3283e-08, 0.0000e+00, ..., -4.2617e-05, + 1.0245e-08, 6.5193e-09], + [ 7.4506e-09, 3.1665e-08, 0.0000e+00, ..., 2.4214e-08, + 2.3562e-07, 1.5367e-07], + ..., + [ 9.3132e-10, -1.7509e-07, 0.0000e+00, ..., 3.8981e-05, + 5.5879e-09, 3.7253e-09], + [ 3.7253e-09, 1.7695e-08, 0.0000e+00, ..., 3.6322e-08, + 7.2643e-08, 4.6566e-08], + [ 9.3132e-10, 3.0734e-08, 0.0000e+00, ..., 3.1143e-06, + 1.0151e-07, 6.3330e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0192, -0.0316, -0.0151, -0.0256, -0.0282, 0.0055, 0.0290, -0.0134, + 0.0383, -0.0029], device='cuda:0'), grad: tensor([-4.8243e-07, -9.3997e-05, 4.3958e-07, 1.8720e-07, 7.9628e-07, + 4.4424e-07, -1.0617e-07, 8.6546e-05, 2.6543e-07, 6.0014e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 250.30, cls_loss 0.0035 cls_loss_mapping 0.0028 cls_loss_causal 0.5041 re_mapping 0.0056 re_causal 0.0132 /// teacc 99.01 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.1073, -0.1928, -0.0678, ..., -0.0621, 0.1668, 0.1645], + [-0.2013, -0.1705, -0.0956, ..., -0.1518, -0.2006, -0.1166], + [-0.0760, -0.1277, 0.1241, ..., -0.1850, 0.2080, 0.0780], + ..., + [-0.1278, 0.0640, 0.0398, ..., 0.1781, -0.1997, -0.2082], + [-0.2318, 0.0616, -0.1276, ..., 0.0517, -0.0848, -0.1410], + [ 0.0033, -0.1109, -0.0885, ..., -0.0956, -0.0563, -0.1505]], + device='cuda:0'), grad: tensor([[-2.2836e-06, 1.8626e-08, 6.5193e-09, ..., 3.3528e-08, + -7.5139e-06, -6.9998e-06], + [ 6.5193e-09, 7.4506e-09, 1.8626e-09, ..., 2.3283e-08, + 1.4994e-07, 7.6368e-08], + [ 2.0023e-07, 5.4948e-08, 0.0000e+00, ..., 7.2643e-08, + -1.0375e-06, -2.0489e-07], + ..., + [ 2.7940e-08, -1.6764e-07, 1.5832e-08, ..., -2.0210e-07, + 2.5705e-07, 1.6112e-07], + [ 2.1793e-07, 3.3528e-08, 1.8626e-08, ..., -4.7591e-07, + 1.1511e-06, 8.3726e-07], + [ 4.0326e-07, 3.9116e-08, -5.4948e-08, ..., 4.8056e-07, + 1.3923e-06, 1.2573e-06]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0187, -0.0314, -0.0144, -0.0242, -0.0280, 0.0050, 0.0286, -0.0137, + 0.0379, -0.0032], device='cuda:0'), grad: tensor([-1.1846e-05, 8.7265e-07, -7.6219e-06, 9.9745e-07, 1.5339e-06, + 1.2163e-06, 8.6948e-06, 9.2387e-07, 1.8915e-06, 3.3379e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 250.42, cls_loss 0.0025 cls_loss_mapping 0.0021 cls_loss_causal 0.5273 re_mapping 0.0057 re_causal 0.0150 /// teacc 99.01 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.1074, -0.1952, -0.0680, ..., -0.0622, 0.1669, 0.1646], + [-0.2013, -0.1707, -0.0954, ..., -0.1534, -0.2009, -0.1167], + [-0.0761, -0.1279, 0.1242, ..., -0.1839, 0.2101, 0.0779], + ..., + [-0.1277, 0.0641, 0.0397, ..., 0.1789, -0.2018, -0.2082], + [-0.2322, 0.0616, -0.1276, ..., 0.0518, -0.0847, -0.1410], + [ 0.0030, -0.1110, -0.0890, ..., -0.0957, -0.0564, -0.1509]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 9.3132e-09, + -2.0489e-08, -1.4901e-08], + [ 0.0000e+00, 1.3690e-07, 0.0000e+00, ..., 2.2817e-07, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 1.4901e-07, 0.0000e+00, ..., 1.6671e-07, + -2.9802e-08, -3.7253e-09], + ..., + [ 9.3132e-10, -9.7230e-07, 0.0000e+00, ..., -1.3439e-06, + 2.2352e-08, 3.7253e-09], + [ 4.0978e-08, 7.4506e-09, 0.0000e+00, ..., 7.9162e-08, + 1.3039e-08, 1.8626e-09], + [ 4.6566e-09, 6.5193e-07, 0.0000e+00, ..., 9.2015e-07, + -3.7253e-09, 3.7253e-09]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0186, -0.0324, -0.0136, -0.0231, -0.0280, 0.0039, 0.0285, -0.0133, + 0.0380, -0.0032], device='cuda:0'), grad: tensor([ 3.7253e-08, 6.4634e-07, 3.9488e-07, 2.0489e-07, -8.8476e-08, + -2.6729e-07, 4.7497e-08, -3.4552e-06, 2.4959e-07, 2.2221e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 250.56, cls_loss 0.0030 cls_loss_mapping 0.0031 cls_loss_causal 0.5277 re_mapping 0.0053 re_causal 0.0142 /// teacc 99.03 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.1075, -0.1960, -0.0684, ..., -0.0622, 0.1670, 0.1647], + [-0.2016, -0.1708, -0.0954, ..., -0.1547, -0.2012, -0.1167], + [-0.0762, -0.1280, 0.1260, ..., -0.1840, 0.2108, 0.0780], + ..., + [-0.1280, 0.0645, 0.0398, ..., 0.1797, -0.2022, -0.2085], + [-0.2327, 0.0615, -0.1280, ..., 0.0519, -0.0850, -0.1415], + [ 0.0026, -0.1114, -0.0890, ..., -0.0962, -0.0565, -0.1515]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 4.6566e-09, 3.2596e-08, ..., 1.5832e-08, + 1.6298e-07, 9.3132e-10], + [ 3.7253e-09, 8.3819e-08, 0.0000e+00, ..., 1.1362e-07, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.0245e-08, 0.0000e+00, ..., 1.4901e-08, + -2.7940e-09, -9.3132e-10], + ..., + [ 4.6566e-09, -8.1584e-07, 0.0000e+00, ..., -1.1642e-06, + 1.8626e-09, 9.3132e-10], + [ 1.2107e-08, 1.0990e-07, 4.6566e-09, ..., 1.5926e-07, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 4.9081e-07, 2.8871e-08, ..., 7.1060e-07, + 1.7136e-07, 0.0000e+00]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0186, -0.0338, -0.0137, -0.0249, -0.0271, 0.0054, 0.0285, -0.0121, + 0.0381, -0.0038], device='cuda:0'), grad: tensor([ 9.5926e-07, -4.8149e-07, 4.6566e-08, 2.6897e-06, -6.2305e-07, + -2.4606e-06, 1.8068e-07, -2.5034e-06, 5.0385e-07, 1.7183e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 250.36, cls_loss 0.0026 cls_loss_mapping 0.0033 cls_loss_causal 0.4962 re_mapping 0.0052 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.1079, -0.1971, -0.0691, ..., -0.0623, 0.1671, 0.1647], + [-0.2016, -0.1709, -0.0956, ..., -0.1548, -0.2016, -0.1169], + [-0.0763, -0.1280, 0.1269, ..., -0.1841, 0.2114, 0.0781], + ..., + [-0.1281, 0.0647, 0.0405, ..., 0.1800, -0.2026, -0.2089], + [-0.2331, 0.0614, -0.1285, ..., 0.0519, -0.0853, -0.1417], + [ 0.0031, -0.1114, -0.0922, ..., -0.0963, -0.0567, -0.1523]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 2.7940e-09, + -4.2375e-07, -2.0210e-07], + [ 0.0000e+00, 6.4261e-08, -5.3085e-08, ..., 1.2759e-07, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 1.0245e-08, 0.0000e+00, ..., 1.7695e-08, + 3.2596e-08, 2.3283e-08], + ..., + [ 0.0000e+00, -2.5518e-07, 1.4901e-08, ..., -5.0478e-07, + 4.6566e-09, 3.7253e-09], + [ 0.0000e+00, 5.5879e-09, 7.4506e-09, ..., 2.7940e-09, + 1.3411e-07, 5.8673e-08], + [ 0.0000e+00, 1.6391e-07, -9.8720e-08, ..., 3.4459e-07, + 2.3190e-07, 1.0431e-07]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0187, -0.0338, -0.0137, -0.0246, -0.0260, 0.0050, 0.0283, -0.0120, + 0.0379, -0.0044], device='cuda:0'), grad: tensor([-6.4634e-07, -1.0151e-06, 1.1083e-07, 3.5390e-08, 1.5125e-06, + 1.3970e-08, 2.8871e-08, -1.2992e-06, 2.6543e-07, 9.9000e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 250.41, cls_loss 0.0023 cls_loss_mapping 0.0022 cls_loss_causal 0.5031 re_mapping 0.0054 re_causal 0.0142 /// teacc 98.97 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.1081, -0.1988, -0.0697, ..., -0.0625, 0.1671, 0.1647], + [-0.2020, -0.1713, -0.0954, ..., -0.1552, -0.2016, -0.1173], + [-0.0760, -0.1292, 0.1274, ..., -0.1850, 0.2120, 0.0783], + ..., + [-0.1284, 0.0650, 0.0408, ..., 0.1805, -0.2034, -0.2091], + [-0.2334, 0.0613, -0.1300, ..., 0.0520, -0.0853, -0.1417], + [ 0.0030, -0.1116, -0.0913, ..., -0.0964, -0.0568, -0.1531]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + -4.4703e-08, -4.5635e-08], + [ 1.8626e-09, 4.6566e-09, -9.3132e-10, ..., -3.6322e-08, + 7.0781e-08, 2.1420e-08], + [ 0.0000e+00, 9.3132e-10, -3.3528e-08, ..., 2.7940e-09, + -2.0303e-07, -6.1467e-08], + ..., + [ 9.3132e-10, -1.5832e-08, 1.3970e-08, ..., -1.4901e-08, + 1.6764e-08, 4.6566e-09], + [ 9.3132e-10, 1.8626e-09, 1.8626e-09, ..., -6.5193e-09, + 7.4506e-09, 2.7940e-09], + [ 1.8626e-09, 1.9558e-08, -1.0524e-07, ..., 3.8184e-08, + 1.0245e-08, 7.4506e-09]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0188, -0.0340, -0.0137, -0.0235, -0.0265, 0.0038, 0.0285, -0.0117, + 0.0380, -0.0042], device='cuda:0'), grad: tensor([-5.4948e-08, -1.6764e-07, -2.8405e-07, -8.3819e-09, 5.1875e-07, + 1.4901e-08, 1.8068e-07, 1.0617e-07, -6.5193e-09, -2.9989e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 250.57, cls_loss 0.0036 cls_loss_mapping 0.0035 cls_loss_causal 0.5428 re_mapping 0.0052 re_causal 0.0137 /// teacc 99.01 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.1084, -0.2000, -0.0701, ..., -0.0625, 0.1671, 0.1647], + [-0.2031, -0.1717, -0.0950, ..., -0.1554, -0.2018, -0.1174], + [-0.0761, -0.1299, 0.1276, ..., -0.1862, 0.2125, 0.0785], + ..., + [-0.1286, 0.0650, 0.0425, ..., 0.1798, -0.2037, -0.2097], + [-0.2327, 0.0630, -0.1297, ..., 0.0561, -0.0847, -0.1421], + [ 0.0017, -0.1121, -0.0924, ..., -0.0967, -0.0570, -0.1543]], + device='cuda:0'), grad: tensor([[ 1.2945e-07, 1.7509e-07, 5.1223e-08, ..., 2.7008e-07, + -4.0326e-07, -2.4959e-07], + [ 3.7253e-08, 3.0175e-07, 8.5682e-08, ..., 2.9244e-07, + 1.8626e-09, 1.8626e-09], + [ 7.4506e-09, 9.3132e-09, 2.7940e-09, ..., 1.4901e-08, + 4.4703e-08, 2.6077e-08], + ..., + [ 5.1223e-08, -6.5565e-05, -1.9163e-05, ..., -5.6177e-05, + 1.8626e-08, 1.1176e-08], + [ 1.6764e-08, 2.1048e-07, 5.9605e-08, ..., 1.9185e-07, + 9.2201e-08, 5.7742e-08], + [ 3.7514e-06, 6.4611e-05, 1.8880e-05, ..., 5.8383e-05, + 1.0896e-07, 6.7987e-08]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0189, -0.0339, -0.0142, -0.0232, -0.0277, 0.0029, 0.0276, -0.0121, + 0.0416, -0.0036], device='cuda:0'), grad: tensor([ 3.0361e-07, 8.0653e-07, 7.6368e-08, 3.2689e-06, 1.2703e-06, + -1.0841e-05, 2.8592e-07, -2.0337e-04, 9.4157e-07, 2.0742e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 250.18, cls_loss 0.0036 cls_loss_mapping 0.0030 cls_loss_causal 0.5197 re_mapping 0.0055 re_causal 0.0139 /// teacc 98.97 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.1085, -0.2013, -0.0703, ..., -0.0626, 0.1651, 0.1649], + [-0.2038, -0.1721, -0.0945, ..., -0.1560, -0.2025, -0.1181], + [-0.0753, -0.1302, 0.1287, ..., -0.1865, 0.2132, 0.0791], + ..., + [-0.1294, 0.0658, 0.0439, ..., 0.1807, -0.2041, -0.2101], + [-0.2348, 0.0629, -0.1307, ..., 0.0558, -0.0859, -0.1451], + [ 0.0014, -0.1128, -0.0949, ..., -0.0973, -0.0541, -0.1565]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, 1.3970e-09, 0.0000e+00, ..., 1.2573e-08, + -1.1157e-06, -5.4482e-07], + [ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 3.2596e-09, + -9.3132e-10, 1.8626e-09], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 4.6566e-09, + 4.6799e-07, 2.2678e-07], + ..., + [ 5.5879e-09, -2.4214e-08, 4.6566e-10, ..., -2.2817e-08, + 1.1176e-08, 5.1223e-09], + [ 5.5879e-08, 9.3132e-09, 2.7940e-09, ..., 4.9826e-08, + 1.4435e-08, 6.9849e-09], + [ 4.0047e-08, 1.9092e-08, -3.7253e-09, ..., 6.3330e-08, + 5.8953e-07, 2.8871e-07]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0215, -0.0343, -0.0149, -0.0233, -0.0272, 0.0029, 0.0278, -0.0111, + 0.0410, -0.0021], device='cuda:0'), grad: tensor([-1.9576e-06, -3.8696e-07, 9.1782e-07, 3.9339e-06, 1.6112e-07, + -4.0121e-06, 2.2165e-07, 2.4214e-08, 2.5379e-07, 8.4611e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 252.06, cls_loss 0.0036 cls_loss_mapping 0.0034 cls_loss_causal 0.5169 re_mapping 0.0052 re_causal 0.0135 /// teacc 99.01 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.1085, -0.2028, -0.0711, ..., -0.0626, 0.1652, 0.1650], + [-0.2040, -0.1724, -0.0947, ..., -0.1564, -0.2023, -0.1168], + [-0.0754, -0.1306, 0.1297, ..., -0.1870, 0.2130, 0.0775], + ..., + [-0.1309, 0.0645, 0.0443, ..., 0.1809, -0.2043, -0.2105], + [-0.2359, 0.0627, -0.1328, ..., 0.0552, -0.0866, -0.1462], + [ 0.0010, -0.1131, -0.0948, ..., -0.0976, -0.0542, -0.1585]], + device='cuda:0'), grad: tensor([[ 3.4645e-07, 5.5693e-07, 0.0000e+00, ..., 6.1467e-08, + 6.7055e-08, 1.4165e-06], + [ 3.5390e-08, 3.5809e-07, 0.0000e+00, ..., 3.9162e-07, + 1.6298e-08, 9.2667e-08], + [ 3.6787e-08, 6.0070e-08, 0.0000e+00, ..., 1.8859e-07, + -2.7381e-07, -1.9418e-07], + ..., + [ 4.4238e-08, -1.1073e-06, 0.0000e+00, ..., -1.4696e-06, + 2.2678e-07, 2.9476e-07], + [ 8.8476e-07, 1.3132e-06, 0.0000e+00, ..., -2.1653e-07, + 4.6287e-07, 3.5204e-06], + [ 1.5832e-06, 3.4608e-06, -0.0000e+00, ..., 1.0002e-06, + 4.0978e-08, 1.2293e-07]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0215, -0.0336, -0.0173, -0.0211, -0.0272, 0.0026, 0.0279, -0.0111, + 0.0403, -0.0023], device='cuda:0'), grad: tensor([ 8.1062e-06, 1.8766e-07, 8.0140e-07, -5.0291e-06, 5.0813e-06, + 6.1989e-06, -3.9846e-05, -1.9409e-06, 1.8939e-05, 7.5139e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 252.34, cls_loss 0.0028 cls_loss_mapping 0.0022 cls_loss_causal 0.4799 re_mapping 0.0054 re_causal 0.0136 /// teacc 98.96 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.1086, -0.2040, -0.0713, ..., -0.0627, 0.1653, 0.1651], + [-0.2045, -0.1730, -0.0927, ..., -0.1562, -0.2031, -0.1169], + [-0.0751, -0.1311, 0.1301, ..., -0.1873, 0.2140, 0.0778], + ..., + [-0.1310, 0.0645, 0.0419, ..., 0.1812, -0.2049, -0.2109], + [-0.2363, 0.0626, -0.1331, ..., 0.0552, -0.0873, -0.1472], + [ 0.0010, -0.1134, -0.0951, ..., -0.0979, -0.0542, -0.1608]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0943e-07, ..., 4.0093e-07, + -2.7940e-09, -3.7253e-09], + [ 1.3970e-09, 7.9162e-09, 9.9745e-07, ..., 2.7288e-07, + 6.5230e-06, 0.0000e+00], + [ 9.3132e-10, -9.3132e-09, -1.1943e-05, ..., -4.5985e-05, + -9.9421e-05, 0.0000e+00], + ..., + [ 3.2596e-09, -7.4506e-09, 1.1548e-06, ..., 9.8199e-06, + 9.2864e-05, 0.0000e+00], + [ 1.3970e-08, 1.7695e-08, 7.3090e-06, ..., 2.6777e-05, + 2.5611e-08, 0.0000e+00], + [ 5.5879e-09, -2.4214e-08, -9.3132e-09, ..., 1.2573e-08, + 8.3819e-09, 4.6566e-09]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0214, -0.0339, -0.0166, -0.0200, -0.0272, 0.0019, 0.0277, -0.0110, + 0.0401, -0.0026], device='cuda:0'), grad: tensor([ 1.3746e-06, 1.8075e-05, -4.4513e-04, 2.7521e-07, 2.8446e-05, + 2.9169e-06, 6.3190e-07, 3.0541e-04, 9.1851e-05, -3.9712e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 252.42, cls_loss 0.0027 cls_loss_mapping 0.0024 cls_loss_causal 0.4987 re_mapping 0.0054 re_causal 0.0141 /// teacc 99.03 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.1086, -0.2043, -0.0713, ..., -0.0627, 0.1654, 0.1653], + [-0.2053, -0.1731, -0.0918, ..., -0.1566, -0.2034, -0.1170], + [-0.0743, -0.1323, 0.1317, ..., -0.1871, 0.2161, 0.0781], + ..., + [-0.1317, 0.0645, 0.0412, ..., 0.1815, -0.2082, -0.2122], + [-0.2365, 0.0624, -0.1351, ..., 0.0552, -0.0875, -0.1474], + [ 0.0020, -0.1134, -0.0957, ..., -0.0980, -0.0543, -0.1618]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 1.1409e-07, + -2.9942e-07, -2.0582e-07], + [ 9.3132e-10, 7.4506e-09, 0.0000e+00, ..., 1.3970e-08, + 2.9802e-08, 1.3970e-09], + [ 4.6566e-10, 1.7229e-08, 0.0000e+00, ..., 2.6496e-07, + -2.1234e-07, -2.7474e-08], + ..., + [ 3.2596e-09, -4.8894e-08, 0.0000e+00, ..., -6.2864e-08, + 7.9162e-08, 1.3970e-09], + [ 2.7940e-09, 2.3283e-09, 0.0000e+00, ..., -1.4184e-06, + -3.3295e-07, 4.0047e-08], + [ 9.3132e-10, 2.1420e-08, 0.0000e+00, ..., 2.5611e-08, + 3.0082e-07, 1.7881e-07]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0214, -0.0344, -0.0146, -0.0192, -0.0274, 0.0009, 0.0279, -0.0111, + 0.0400, -0.0026], device='cuda:0'), grad: tensor([-1.9977e-07, -3.7253e-09, 1.0757e-07, 2.6217e-07, 1.9232e-07, + 2.3562e-06, 9.1409e-07, 1.7975e-07, -4.4070e-06, 6.1234e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 252.48, cls_loss 0.0027 cls_loss_mapping 0.0033 cls_loss_causal 0.5178 re_mapping 0.0052 re_causal 0.0138 /// teacc 98.96 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.1087, -0.2045, -0.0714, ..., -0.0629, 0.1656, 0.1653], + [-0.2054, -0.1734, -0.0917, ..., -0.1567, -0.2035, -0.1170], + [-0.0744, -0.1342, 0.1321, ..., -0.1884, 0.2168, 0.0781], + ..., + [-0.1345, 0.0647, 0.0412, ..., 0.1818, -0.2087, -0.2124], + [-0.2370, 0.0623, -0.1351, ..., 0.0551, -0.0876, -0.1475], + [ 0.0019, -0.1136, -0.0956, ..., -0.0985, -0.0544, -0.1633]], + device='cuda:0'), grad: tensor([[ 1.0710e-07, 3.7253e-09, 6.0536e-09, ..., 6.0536e-08, + 8.3819e-09, -1.1176e-08], + [ 1.0198e-07, 2.0489e-08, 1.3970e-09, ..., 1.4808e-07, + 3.7253e-09, 0.0000e+00], + [ 2.6077e-08, -3.2596e-09, -2.5891e-07, ..., 2.4214e-08, + -9.7137e-07, -1.3970e-09], + ..., + [ 1.2433e-07, 1.7695e-08, 4.6566e-09, ..., 2.1514e-07, + 1.7229e-08, 0.0000e+00], + [ 3.4004e-05, 2.1979e-06, 4.6566e-10, ..., 3.3349e-05, + 4.6566e-09, 4.6566e-10], + [ 1.0449e-06, 2.9802e-08, 0.0000e+00, ..., -3.1982e-06, + 9.7789e-09, 6.9849e-09]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0213, -0.0342, -0.0150, -0.0191, -0.0271, 0.0012, 0.0277, -0.0109, + 0.0399, -0.0029], device='cuda:0'), grad: tensor([ 2.3702e-07, 3.4459e-07, -9.7789e-07, 4.8089e-04, 1.4514e-05, + -5.5122e-04, 3.9563e-06, 1.2424e-06, 6.4135e-05, -1.3404e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 250.57, cls_loss 0.0031 cls_loss_mapping 0.0030 cls_loss_causal 0.5225 re_mapping 0.0050 re_causal 0.0138 /// teacc 98.99 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.1087, -0.2054, -0.0702, ..., -0.0629, 0.1659, 0.1655], + [-0.2055, -0.1736, -0.0908, ..., -0.1569, -0.2046, -0.1178], + [-0.0744, -0.1355, 0.1303, ..., -0.1895, 0.2171, 0.0782], + ..., + [-0.1358, 0.0632, 0.0409, ..., 0.1795, -0.2090, -0.2128], + [-0.2383, 0.0653, -0.1360, ..., 0.0577, -0.0883, -0.1486], + [ 0.0010, -0.1138, -0.0955, ..., -0.0987, -0.0544, -0.1648]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, 2.1886e-08, 0.0000e+00, ..., 6.5193e-09, + -1.8962e-06, -1.5860e-06], + [ 8.7544e-08, 1.0105e-07, 0.0000e+00, ..., 3.9116e-08, + 1.1176e-08, 8.8476e-09], + [ 1.1083e-07, 1.2666e-07, 0.0000e+00, ..., 4.6566e-10, + 9.4995e-08, 7.9162e-08], + ..., + [ 8.3353e-08, 1.2107e-07, 0.0000e+00, ..., 1.2573e-08, + 6.9849e-09, 5.5879e-09], + [ 1.8626e-07, 2.6310e-07, 0.0000e+00, ..., 4.0047e-08, + 1.9185e-07, 1.6019e-07], + [ 2.8871e-08, 3.4925e-08, 0.0000e+00, ..., -7.2643e-08, + 1.5479e-06, 1.2945e-06]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0211, -0.0331, -0.0170, -0.0199, -0.0277, 0.0024, 0.0275, -0.0124, + 0.0421, -0.0028], device='cuda:0'), grad: tensor([-3.8557e-06, 5.6764e-07, 7.0455e-07, 3.5018e-06, 1.7043e-07, + -5.5432e-06, -3.9814e-07, 5.6112e-07, 1.4091e-06, 2.8685e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 250.29, cls_loss 0.0028 cls_loss_mapping 0.0025 cls_loss_causal 0.4792 re_mapping 0.0052 re_causal 0.0131 /// teacc 99.03 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.1087, -0.2065, -0.0709, ..., -0.0629, 0.1663, 0.1658], + [-0.2056, -0.1739, -0.0908, ..., -0.1573, -0.2064, -0.1193], + [-0.0745, -0.1357, 0.1306, ..., -0.1900, 0.2177, 0.0783], + ..., + [-0.1361, 0.0634, 0.0408, ..., 0.1797, -0.2095, -0.2133], + [-0.2387, 0.0652, -0.1370, ..., 0.0577, -0.0886, -0.1490], + [ 0.0010, -0.1141, -0.0956, ..., -0.0987, -0.0546, -0.1687]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.1921e-07, 0.0000e+00, ..., 0.0000e+00, + -6.5146e-07, -3.3295e-07], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., -4.6566e-09, + 3.9116e-08, 2.7940e-09], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + -1.0738e-06, 5.5879e-09], + ..., + [ 1.3970e-09, 1.9558e-08, 0.0000e+00, ..., 1.8626e-09, + 5.4017e-08, 2.7474e-08], + [ 4.0978e-08, 2.5611e-08, 0.0000e+00, ..., 2.4214e-08, + 4.1910e-08, 2.0955e-08], + [ 2.1420e-08, 6.5193e-08, 0.0000e+00, ..., 2.2817e-08, + 2.7660e-07, 1.4110e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0208, -0.0334, -0.0172, -0.0204, -0.0279, 0.0026, 0.0278, -0.0123, + 0.0420, -0.0028], device='cuda:0'), grad: tensor([-1.5479e-06, -5.2676e-06, 1.5693e-07, 2.0908e-07, 4.7125e-07, + 1.7183e-07, 2.3264e-06, 3.3583e-06, 1.8626e-07, -5.7742e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 250.25, cls_loss 0.0027 cls_loss_mapping 0.0021 cls_loss_causal 0.5151 re_mapping 0.0053 re_causal 0.0140 /// teacc 98.96 lr 0.00010000 +Epoch 247, weight, value: tensor([[-1.0872e-01, -2.0688e-01, -7.0241e-02, ..., -6.2884e-02, + 1.6696e-01, 1.6604e-01], + [-2.0644e-01, -1.7410e-01, -9.0777e-02, ..., -1.5660e-01, + -2.0770e-01, -1.1931e-01], + [-7.4634e-02, -1.3569e-01, 1.3060e-01, ..., -1.9005e-01, + 2.1911e-01, 7.8251e-02], + ..., + [-1.3666e-01, 6.3550e-02, 4.0760e-02, ..., 1.7963e-01, + -2.0993e-01, -2.1364e-01], + [-2.3937e-01, 6.5157e-02, -1.3700e-01, ..., 5.7704e-02, + -8.9496e-02, -1.4954e-01], + [ 4.4920e-05, -1.1426e-01, -9.5559e-02, ..., -9.8916e-02, + -5.4780e-02, -1.7133e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.5879e-07, 0.0000e+00, ..., 2.7940e-09, + -1.9222e-06, -1.1241e-06], + [ 0.0000e+00, 3.5390e-08, 2.0489e-08, ..., 2.5705e-07, + 2.1420e-08, 1.2573e-08], + [ 0.0000e+00, 2.7008e-08, 2.2817e-08, ..., 2.9197e-07, + 8.8010e-08, 4.7963e-08], + ..., + [ 5.1223e-09, -1.9697e-07, 4.6566e-10, ..., -3.1618e-07, + 2.0023e-08, 1.1176e-08], + [ 5.0757e-08, -2.1514e-07, -6.3330e-08, ..., -1.2368e-06, + 7.2084e-07, 4.3400e-07], + [ 9.3132e-10, 1.6904e-07, 0.0000e+00, ..., 2.1188e-07, + 3.0827e-07, 1.7928e-07]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0205, -0.0331, -0.0168, -0.0206, -0.0286, 0.0028, 0.0278, -0.0126, + 0.0419, -0.0028], device='cuda:0'), grad: tensor([-4.0643e-06, 1.2163e-06, 1.4771e-06, 1.3597e-06, -7.5698e-06, + 1.7472e-06, 1.1101e-06, -5.6857e-07, -3.3714e-06, 8.6427e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 249.85, cls_loss 0.0035 cls_loss_mapping 0.0025 cls_loss_causal 0.5216 re_mapping 0.0053 re_causal 0.0134 /// teacc 99.07 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.1088, -0.2084, -0.0706, ..., -0.0630, 0.1672, 0.1664], + [-0.2067, -0.1747, -0.0908, ..., -0.1563, -0.2080, -0.1194], + [-0.0747, -0.1360, 0.1307, ..., -0.1903, 0.2199, 0.0784], + ..., + [-0.1371, 0.0639, 0.0407, ..., 0.1797, -0.2107, -0.2139], + [-0.2397, 0.0654, -0.1370, ..., 0.0577, -0.0886, -0.1497], + [-0.0006, -0.1145, -0.0956, ..., -0.0991, -0.0547, -0.1730]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 3.2596e-09, 0.0000e+00, ..., 9.3132e-09, + -1.7341e-06, -1.6410e-06], + [ 2.7940e-09, 1.0524e-07, 0.0000e+00, ..., 3.5018e-07, + 4.9360e-08, 1.7229e-08], + [ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 4.6566e-09, + -1.4855e-07, -5.6345e-08], + ..., + [ 4.6566e-09, -9.8068e-07, 0.0000e+00, ..., -1.2834e-06, + 1.9092e-08, 9.7789e-09], + [ 2.0955e-08, 2.7940e-08, 0.0000e+00, ..., 2.2352e-08, + 4.6799e-07, 4.0559e-07], + [ 1.2107e-08, 6.9384e-08, 0.0000e+00, ..., 1.9465e-07, + 1.1781e-06, 1.1129e-06]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0205, -0.0325, -0.0170, -0.0209, -0.0298, 0.0027, 0.0271, -0.0130, + 0.0424, -0.0019], device='cuda:0'), grad: tensor([-5.1558e-06, 6.5658e-07, -3.1479e-07, 6.5425e-07, 4.1910e-08, + 7.6462e-07, 3.3388e-07, -2.6114e-06, 1.4473e-06, 4.1761e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 250.19, cls_loss 0.0028 cls_loss_mapping 0.0028 cls_loss_causal 0.4910 re_mapping 0.0051 re_causal 0.0130 /// teacc 98.99 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.1110, -0.2091, -0.0706, ..., -0.0630, 0.1663, 0.1658], + [-0.2069, -0.1753, -0.0908, ..., -0.1577, -0.2091, -0.1195], + [-0.0749, -0.1363, 0.1307, ..., -0.1908, 0.2217, 0.0785], + ..., + [-0.1389, 0.0641, 0.0407, ..., 0.1802, -0.2110, -0.2142], + [-0.2400, 0.0654, -0.1371, ..., 0.0577, -0.0887, -0.1500], + [-0.0008, -0.1146, -0.0956, ..., -0.0992, -0.0549, -0.1748]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + -1.4575e-07, -5.9605e-08], + [ 5.1223e-09, 5.1223e-09, 0.0000e+00, ..., 3.2596e-09, + -1.6019e-07, 9.3132e-10], + [ 1.3970e-09, 2.3283e-09, 0.0000e+00, ..., 5.5879e-09, + 4.3306e-08, -1.8626e-09], + ..., + [ 1.1642e-08, -8.7544e-08, 0.0000e+00, ..., -1.2992e-07, + 2.0489e-08, 2.3283e-09], + [ 8.0094e-08, 6.5193e-09, 0.0000e+00, ..., 4.7963e-08, + 4.5169e-08, 9.3132e-09], + [ 8.8476e-09, 5.4017e-08, 0.0000e+00, ..., 9.7323e-08, + 1.3877e-07, 5.2620e-08]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0211, -0.0335, -0.0168, -0.0205, -0.0297, 0.0020, 0.0285, -0.0124, + 0.0425, -0.0021], device='cuda:0'), grad: tensor([-1.9139e-07, -8.5495e-07, 3.0873e-07, 3.3528e-08, -1.2247e-07, + -2.4075e-07, 4.2329e-07, -6.6590e-08, 2.3330e-07, 4.7032e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 248---------------------------------------------------- +epoch 248, time 267.07, cls_loss 0.0026 cls_loss_mapping 0.0032 cls_loss_causal 0.5209 re_mapping 0.0051 re_causal 0.0138 /// teacc 99.14 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.1111, -0.2104, -0.0707, ..., -0.0631, 0.1666, 0.1661], + [-0.2073, -0.1757, -0.0907, ..., -0.1579, -0.2104, -0.1201], + [-0.0744, -0.1365, 0.1307, ..., -0.1911, 0.2236, 0.0791], + ..., + [-0.1389, 0.0650, 0.0408, ..., 0.1808, -0.2112, -0.2145], + [-0.2404, 0.0654, -0.1373, ..., 0.0578, -0.0887, -0.1504], + [-0.0008, -0.1159, -0.0956, ..., -0.1003, -0.0551, -0.1771]], + device='cuda:0'), grad: tensor([[ 5.7276e-08, 4.3306e-08, 0.0000e+00, ..., 1.3970e-09, + 1.3113e-05, -1.3970e-08], + [ 1.0245e-08, 1.8626e-08, 0.0000e+00, ..., 1.3504e-08, + 1.1642e-08, 4.6566e-10], + [ 7.7765e-08, 6.5658e-08, 0.0000e+00, ..., 6.0536e-09, + 8.2422e-08, -0.0000e+00], + ..., + [ 1.8626e-08, -9.4064e-08, 0.0000e+00, ..., -1.3504e-07, + 1.3132e-07, 9.3132e-10], + [ 1.4901e-08, 6.0536e-09, 0.0000e+00, ..., 7.4506e-09, + 5.1223e-09, 9.3132e-10], + [ 1.1176e-08, 8.8476e-08, 0.0000e+00, ..., 9.8720e-08, + -1.3344e-05, 9.3132e-09]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0209, -0.0344, -0.0152, -0.0204, -0.0301, 0.0014, 0.0282, -0.0119, + 0.0428, -0.0025], device='cuda:0'), grad: tensor([ 6.8963e-05, 1.0850e-07, 5.7649e-07, -8.6520e-07, 1.0524e-07, + 5.6438e-07, 4.5169e-08, 4.1304e-07, 5.5414e-08, -6.9857e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 250.41, cls_loss 0.0033 cls_loss_mapping 0.0029 cls_loss_causal 0.5149 re_mapping 0.0049 re_causal 0.0130 /// teacc 98.99 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.1111, -0.2124, -0.0707, ..., -0.0634, 0.1668, 0.1663], + [-0.2079, -0.1761, -0.0906, ..., -0.1583, -0.2106, -0.1202], + [-0.0745, -0.1369, 0.1308, ..., -0.1914, 0.2241, 0.0791], + ..., + [-0.1392, 0.0653, 0.0407, ..., 0.1812, -0.2116, -0.2147], + [-0.2409, 0.0654, -0.1383, ..., 0.0577, -0.0889, -0.1507], + [-0.0010, -0.1164, -0.0956, ..., -0.1010, -0.0551, -0.1788]], + device='cuda:0'), grad: tensor([[ 9.5461e-08, 4.9174e-07, 0.0000e+00, ..., 2.0489e-08, + 1.8626e-08, -7.5903e-08], + [ 7.0315e-08, 1.0012e-06, 0.0000e+00, ..., 5.5414e-08, + 4.9500e-07, 2.3283e-09], + [ 1.5926e-07, 4.1090e-06, 4.6566e-10, ..., 1.6950e-07, + 2.9653e-06, 0.0000e+00], + ..., + [ 5.3085e-08, -1.5218e-06, 2.3283e-09, ..., -5.2992e-07, + 1.4016e-07, 3.7253e-09], + [ 1.9139e-07, -5.3406e-05, 4.6566e-10, ..., 3.9116e-08, + -4.5627e-05, 9.3132e-09], + [ 1.8161e-07, 1.0151e-06, -4.6566e-09, ..., 1.7136e-07, + 1.2061e-07, 4.4703e-08]], device='cuda:0') +Epoch 251, bias, value: tensor([-2.0832e-02, -3.6316e-02, -1.5020e-02, -1.8726e-02, -2.9570e-02, + 8.0183e-05, 2.7928e-02, -1.1584e-02, 4.2911e-02, -1.6742e-03], + device='cuda:0'), grad: tensor([ 2.2277e-06, 5.7891e-06, 2.9624e-05, 7.8753e-06, 5.0571e-07, + 2.0843e-06, 3.6812e-04, -2.9393e-06, -4.1604e-04, 3.1721e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 250.33, cls_loss 0.0029 cls_loss_mapping 0.0030 cls_loss_causal 0.5065 re_mapping 0.0055 re_causal 0.0132 /// teacc 98.97 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.1111, -0.2119, -0.0713, ..., -0.0635, 0.1673, 0.1667], + [-0.2080, -0.1763, -0.0907, ..., -0.1583, -0.2111, -0.1203], + [-0.0746, -0.1378, 0.1311, ..., -0.1920, 0.2251, 0.0789], + ..., + [-0.1395, 0.0656, 0.0407, ..., 0.1815, -0.2127, -0.2152], + [-0.2424, 0.0655, -0.1386, ..., 0.0575, -0.0888, -0.1527], + [-0.0013, -0.1167, -0.0956, ..., -0.1013, -0.0553, -0.1814]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8626e-09, 2.2352e-08, ..., 0.0000e+00, + -2.7567e-07, -2.9569e-07], + [ 2.9383e-07, 1.5367e-07, 1.0245e-08, ..., 1.3970e-09, + 3.3993e-08, 8.3819e-09], + [ 9.3132e-09, 8.3819e-09, -2.0582e-06, ..., 4.6566e-10, + 1.6764e-08, 2.5146e-08], + ..., + [ 2.0489e-08, -4.6566e-08, 1.9511e-07, ..., -5.8673e-08, + 3.7253e-09, 1.3970e-09], + [ 5.4482e-08, 1.1176e-08, 1.5162e-06, ..., -0.0000e+00, + 1.5367e-08, 1.0710e-08], + [ 1.8207e-07, 1.4622e-07, 9.3132e-10, ..., 4.0047e-08, + 2.5099e-07, 3.1199e-08]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0205, -0.0365, -0.0148, -0.0192, -0.0295, 0.0006, 0.0272, -0.0116, + 0.0427, -0.0016], device='cuda:0'), grad: tensor([-1.1735e-07, 1.0757e-06, -1.4283e-05, -7.8883e-07, -1.3802e-06, + -5.5041e-07, 7.5158e-07, 1.3374e-06, 1.0692e-05, 3.2708e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 250.50, cls_loss 0.0023 cls_loss_mapping 0.0020 cls_loss_causal 0.4796 re_mapping 0.0054 re_causal 0.0132 /// teacc 99.02 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.1111, -0.2125, -0.0713, ..., -0.0636, 0.1674, 0.1668], + [-0.2082, -0.1767, -0.0907, ..., -0.1585, -0.2112, -0.1203], + [-0.0747, -0.1383, 0.1315, ..., -0.1921, 0.2263, 0.0791], + ..., + [-0.1404, 0.0665, 0.0411, ..., 0.1818, -0.2142, -0.2160], + [-0.2441, 0.0654, -0.1393, ..., 0.0574, -0.0891, -0.1543], + [-0.0015, -0.1169, -0.0955, ..., -0.1015, -0.0554, -0.1820]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + -6.0536e-09, -5.1223e-09], + [ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-09, 3.2596e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + -6.9849e-09, -2.3283e-09], + ..., + [ 4.6566e-10, -6.7987e-08, 0.0000e+00, ..., -7.1712e-08, + 1.3970e-09, 4.6566e-10], + [ 6.0536e-09, 6.5193e-09, 0.0000e+00, ..., -3.7253e-09, + 7.9162e-09, 3.2596e-09], + [ 4.6566e-10, 4.6100e-08, 0.0000e+00, ..., 4.8894e-08, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0204, -0.0365, -0.0144, -0.0198, -0.0295, 0.0008, 0.0274, -0.0115, + 0.0424, -0.0016], device='cuda:0'), grad: tensor([ 2.6543e-08, -2.7940e-08, 1.5367e-08, -2.7474e-08, -1.0766e-06, + 8.1491e-08, 2.1420e-08, 7.5437e-08, 1.4435e-08, 9.1502e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 250.19, cls_loss 0.0026 cls_loss_mapping 0.0023 cls_loss_causal 0.5083 re_mapping 0.0052 re_causal 0.0130 /// teacc 99.02 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.1112, -0.2147, -0.0714, ..., -0.0636, 0.1674, 0.1669], + [-0.2083, -0.1768, -0.0906, ..., -0.1586, -0.2114, -0.1204], + [-0.0747, -0.1389, 0.1316, ..., -0.1927, 0.2294, 0.0794], + ..., + [-0.1406, 0.0666, 0.0410, ..., 0.1819, -0.2175, -0.2167], + [-0.2456, 0.0654, -0.1395, ..., 0.0572, -0.0895, -0.1546], + [-0.0009, -0.1169, -0.0955, ..., -0.1014, -0.0553, -0.1825]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 1.3970e-09, + -1.0151e-07, -5.5414e-08], + [ 9.3132e-10, 2.3749e-08, 0.0000e+00, ..., 1.3970e-08, + 9.3132e-10, 4.6566e-10], + [ 7.4506e-09, 1.3970e-08, 0.0000e+00, ..., 8.3819e-09, + 7.9162e-09, 3.7253e-09], + ..., + [ 5.1223e-09, -3.9628e-07, 0.0000e+00, ..., -2.8685e-07, + 3.7253e-09, 2.3283e-09], + [ 4.1910e-09, 1.4435e-08, 0.0000e+00, ..., -5.4017e-08, + 1.4901e-08, 8.8476e-09], + [ 2.3283e-09, 1.0664e-07, -0.0000e+00, ..., 6.7987e-08, + 5.7276e-08, 3.0268e-08]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0205, -0.0363, -0.0135, -0.0200, -0.0302, 0.0023, 0.0265, -0.0122, + 0.0418, -0.0010], device='cuda:0'), grad: tensor([-1.9139e-07, -1.4808e-07, 8.8476e-08, 4.2375e-07, -5.1223e-08, + 1.8300e-07, 2.9802e-08, -5.6205e-07, -9.0804e-08, 3.4738e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 250.25, cls_loss 0.0025 cls_loss_mapping 0.0026 cls_loss_causal 0.4969 re_mapping 0.0051 re_causal 0.0128 /// teacc 99.00 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.1112, -0.2158, -0.0715, ..., -0.0636, 0.1681, 0.1673], + [-0.2118, -0.1777, -0.0906, ..., -0.1589, -0.2117, -0.1210], + [-0.0740, -0.1394, 0.1319, ..., -0.1930, 0.2296, 0.0800], + ..., + [-0.1405, 0.0669, 0.0410, ..., 0.1821, -0.2176, -0.2169], + [-0.2470, 0.0653, -0.1399, ..., 0.0572, -0.0910, -0.1559], + [-0.0023, -0.1171, -0.0954, ..., -0.1015, -0.0560, -0.1871]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 2.6077e-08, 0.0000e+00, ..., 4.9826e-08, + -5.8860e-06, -5.6736e-06], + [ 9.3132e-10, 6.5193e-09, 0.0000e+00, ..., 6.9849e-09, + 5.4948e-08, 5.2154e-08], + [ 9.3132e-10, 4.2841e-08, 0.0000e+00, ..., 8.1025e-08, + 3.2643e-07, 3.1013e-07], + ..., + [ 4.6566e-10, -2.4261e-07, 0.0000e+00, ..., -2.5332e-07, + 1.1502e-07, 1.2992e-07], + [ 2.9150e-07, 8.3819e-09, 0.0000e+00, ..., 3.6834e-07, + 2.9374e-06, 2.8331e-06], + [ 4.6566e-10, 1.4435e-07, 0.0000e+00, ..., 1.5926e-07, + 1.1502e-06, 1.1045e-06]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0199, -0.0367, -0.0134, -0.0210, -0.0307, 0.0040, 0.0264, -0.0122, + 0.0415, -0.0013], device='cuda:0'), grad: tensor([-2.0012e-05, 2.2352e-07, 1.4110e-06, 4.5868e-07, -4.9826e-07, + 1.2796e-06, 2.2054e-06, -1.3085e-07, 1.0245e-05, 4.7982e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 250.41, cls_loss 0.0030 cls_loss_mapping 0.0024 cls_loss_causal 0.5248 re_mapping 0.0050 re_causal 0.0130 /// teacc 99.02 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.1113, -0.2163, -0.0716, ..., -0.0639, 0.1684, 0.1675], + [-0.2122, -0.1791, -0.0902, ..., -0.1596, -0.2123, -0.1218], + [-0.0724, -0.1415, 0.1321, ..., -0.1944, 0.2302, 0.0813], + ..., + [-0.1413, 0.0666, 0.0410, ..., 0.1815, -0.2178, -0.2177], + [-0.2482, 0.0666, -0.1425, ..., 0.0582, -0.0916, -0.1568], + [-0.0031, -0.1174, -0.0954, ..., -0.1018, -0.0562, -0.1884]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.9626e-06, -2.6282e-06], + [ 4.6566e-10, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + 9.8255e-08, 3.7719e-08], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -9.1735e-08, -5.7276e-08], + ..., + [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 2.5285e-07, 9.6392e-08], + [ 2.9197e-07, 8.2189e-07, 0.0000e+00, ..., 1.6158e-07, + 4.9919e-07, 1.8813e-07], + [ 7.4506e-09, 2.0023e-08, 0.0000e+00, ..., 3.2596e-09, + 3.4012e-06, 1.3011e-06]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0197, -0.0369, -0.0130, -0.0213, -0.0308, 0.0042, 0.0262, -0.0126, + 0.0427, -0.0015], device='cuda:0'), grad: tensor([-1.8612e-05, 2.7101e-07, -2.5844e-07, -9.7509e-07, -1.4948e-07, + 2.2538e-06, 4.7423e-06, 6.9523e-07, 2.8498e-06, 9.1717e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 250.27, cls_loss 0.0022 cls_loss_mapping 0.0017 cls_loss_causal 0.5129 re_mapping 0.0049 re_causal 0.0131 /// teacc 99.05 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.1115, -0.2171, -0.0717, ..., -0.0640, 0.1685, 0.1676], + [-0.2122, -0.1795, -0.0902, ..., -0.1598, -0.2124, -0.1219], + [-0.0726, -0.1424, 0.1332, ..., -0.1961, 0.2307, 0.0811], + ..., + [-0.1427, 0.0668, 0.0409, ..., 0.1816, -0.2179, -0.2182], + [-0.2487, 0.0666, -0.1431, ..., 0.0582, -0.0922, -0.1574], + [-0.0029, -0.1175, -0.0954, ..., -0.1019, -0.0563, -0.1892]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 6.0536e-08, 3.2596e-09], + [ 4.1910e-09, 6.9849e-09, 0.0000e+00, ..., 0.0000e+00, + 2.6543e-08, 4.1910e-09], + [ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 9.3132e-10, + -5.5768e-06, -8.1584e-07], + ..., + [ 7.4506e-09, 1.3970e-08, 0.0000e+00, ..., 1.8626e-09, + 5.3570e-06, 7.8417e-07], + [ 2.1420e-08, 3.4925e-08, 0.0000e+00, ..., 4.6566e-10, + 7.9162e-08, 1.2107e-08], + [ 6.9849e-09, 1.3504e-08, 0.0000e+00, ..., -4.6566e-09, + 1.8626e-08, 4.1910e-09]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0196, -0.0369, -0.0132, -0.0214, -0.0309, 0.0045, 0.0262, -0.0125, + 0.0426, -0.0016], device='cuda:0'), grad: tensor([ 1.5274e-07, -1.1502e-07, -1.2666e-05, -1.9837e-07, 1.0803e-07, + 8.1491e-08, 3.3528e-08, 1.2361e-05, 2.8545e-07, -3.5856e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 250.44, cls_loss 0.0023 cls_loss_mapping 0.0026 cls_loss_causal 0.4881 re_mapping 0.0048 re_causal 0.0128 /// teacc 99.00 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.1118, -0.2176, -0.0721, ..., -0.0641, 0.1688, 0.1678], + [-0.2123, -0.1786, -0.0903, ..., -0.1600, -0.2128, -0.1220], + [-0.0728, -0.1429, 0.1333, ..., -0.1971, 0.2310, 0.0811], + ..., + [-0.1431, 0.0665, 0.0409, ..., 0.1815, -0.2180, -0.2188], + [-0.2495, 0.0670, -0.1432, ..., 0.0586, -0.0928, -0.1583], + [-0.0030, -0.1176, -0.0954, ..., -0.1021, -0.0564, -0.1902]], + device='cuda:0'), grad: tensor([[-3.6787e-08, 4.6566e-10, 0.0000e+00, ..., 2.3283e-08, + -5.2713e-07, -2.0815e-07], + [ 2.2352e-08, 1.8626e-09, 0.0000e+00, ..., 2.3749e-08, + 6.0070e-08, 2.3283e-09], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + -4.5635e-07, 2.4214e-08], + ..., + [ 5.8673e-08, -9.2667e-08, 0.0000e+00, ..., -2.7940e-08, + 4.5355e-07, 2.3283e-09], + [ 2.8405e-08, 1.2573e-08, 0.0000e+00, ..., 2.0955e-08, + 4.5169e-08, 1.3970e-08], + [ 5.4017e-08, 7.4971e-08, 0.0000e+00, ..., 1.2713e-07, + 1.6065e-07, 5.6345e-08]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0194, -0.0358, -0.0133, -0.0216, -0.0303, 0.0047, 0.0262, -0.0136, + 0.0425, -0.0017], device='cuda:0'), grad: tensor([-6.5332e-07, -1.4994e-07, -1.2508e-06, 2.2491e-07, 5.6904e-07, + 7.1712e-08, -3.2736e-07, 1.4473e-06, 3.7253e-08, 4.2375e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 250.58, cls_loss 0.0029 cls_loss_mapping 0.0028 cls_loss_causal 0.5121 re_mapping 0.0050 re_causal 0.0130 /// teacc 99.04 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.1119, -0.2187, -0.0721, ..., -0.0641, 0.1690, 0.1679], + [-0.2123, -0.1818, -0.0903, ..., -0.1604, -0.2159, -0.1221], + [-0.0730, -0.1439, 0.1334, ..., -0.1978, 0.2312, 0.0811], + ..., + [-0.1441, 0.0685, 0.0409, ..., 0.1817, -0.2152, -0.2189], + [-0.2501, 0.0669, -0.1432, ..., 0.0585, -0.0933, -0.1587], + [-0.0034, -0.1177, -0.0954, ..., -0.1022, -0.0565, -0.1911]], + device='cuda:0'), grad: tensor([[ 2.8405e-08, 4.6566e-10, 0.0000e+00, ..., 4.9360e-08, + 1.7555e-07, 1.1967e-07], + [ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 5.5879e-09, + 2.3283e-09, 1.3970e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -2.7707e-07, + -8.4704e-07, -6.0815e-07], + ..., + [ 0.0000e+00, -1.8161e-08, 0.0000e+00, ..., 2.7940e-09, + 7.4971e-08, 5.3551e-08], + [ 2.7940e-08, 2.3283e-09, 0.0000e+00, ..., 1.7975e-07, + 5.1409e-07, 3.6741e-07], + [ 1.3039e-08, 7.4506e-09, 0.0000e+00, ..., 2.0489e-08, + 3.7253e-09, 2.3283e-09]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0193, -0.0387, -0.0141, -0.0219, -0.0300, 0.0052, 0.0261, -0.0105, + 0.0424, -0.0020], device='cuda:0'), grad: tensor([ 5.6485e-07, 1.7276e-07, -2.7586e-06, 4.2515e-07, -2.7027e-06, + -8.5682e-08, -1.1548e-07, 2.2491e-07, 1.7053e-06, 2.5705e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 250.64, cls_loss 0.0027 cls_loss_mapping 0.0030 cls_loss_causal 0.4839 re_mapping 0.0050 re_causal 0.0130 /// teacc 99.04 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.1121, -0.2219, -0.0725, ..., -0.0642, 0.1694, 0.1684], + [-0.2125, -0.1819, -0.0903, ..., -0.1608, -0.2159, -0.1221], + [-0.0740, -0.1463, 0.1334, ..., -0.1998, 0.2307, 0.0811], + ..., + [-0.1445, 0.0687, 0.0409, ..., 0.1821, -0.2153, -0.2191], + [-0.2498, 0.0668, -0.1432, ..., 0.0586, -0.0925, -0.1590], + [-0.0035, -0.1179, -0.0954, ..., -0.1023, -0.0566, -0.1915]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 9.3132e-10, 0.0000e+00, ..., 4.1910e-09, + -4.6566e-10, -7.9162e-09], + [ 9.3132e-10, 1.4435e-08, 0.0000e+00, ..., 2.2817e-08, + 1.1176e-08, 2.3283e-09], + [ 2.7940e-09, 1.3970e-09, 0.0000e+00, ..., 9.3132e-10, + -1.6868e-05, -3.0808e-06], + ..., + [ 1.3970e-09, -1.1036e-07, 0.0000e+00, ..., -1.2200e-07, + 1.6809e-05, 3.0715e-06], + [ 6.3330e-08, 4.6566e-09, 0.0000e+00, ..., 3.0268e-08, + 2.0955e-08, 6.9849e-09], + [ 1.8626e-09, 6.4261e-08, 0.0000e+00, ..., 4.3772e-08, + 1.0710e-08, 7.4506e-09]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0190, -0.0386, -0.0148, -0.0223, -0.0304, 0.0054, 0.0254, -0.0105, + 0.0428, -0.0020], device='cuda:0'), grad: tensor([ 3.9116e-08, 1.5926e-07, -7.0214e-05, 1.0524e-07, 4.0932e-07, + -4.0373e-07, 1.9697e-07, 6.9737e-05, 1.8254e-07, -3.4599e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 250.55, cls_loss 0.0023 cls_loss_mapping 0.0016 cls_loss_causal 0.5076 re_mapping 0.0049 re_causal 0.0129 /// teacc 98.96 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.1124, -0.2233, -0.0733, ..., -0.0643, 0.1695, 0.1686], + [-0.2125, -0.1820, -0.0903, ..., -0.1610, -0.2159, -0.1222], + [-0.0740, -0.1475, 0.1334, ..., -0.2003, 0.2313, 0.0807], + ..., + [-0.1447, 0.0689, 0.0409, ..., 0.1823, -0.2155, -0.2202], + [-0.2512, 0.0667, -0.1436, ..., 0.0585, -0.0933, -0.1596], + [-0.0036, -0.1180, -0.0954, ..., -0.1024, -0.0567, -0.1934]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.1910e-09, 0.0000e+00, ..., 1.8626e-09, + -1.8626e-08, -3.3528e-08], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 1.0245e-08, 4.6566e-10], + [ 9.3132e-10, -6.5193e-09, 0.0000e+00, ..., -9.7789e-09, + -4.9500e-07, -5.8208e-08], + ..., + [ 1.3970e-09, 2.3283e-09, 0.0000e+00, ..., 8.3819e-09, + 2.5379e-07, 1.9092e-08], + [ 5.5879e-09, -4.0513e-08, 0.0000e+00, ..., -3.2596e-08, + 9.6392e-08, 1.7695e-08], + [ 2.7940e-09, 2.6543e-08, 0.0000e+00, ..., 1.0245e-08, + 1.1083e-07, 4.8429e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0191, -0.0386, -0.0147, -0.0230, -0.0307, 0.0065, 0.0251, -0.0105, + 0.0424, -0.0019], device='cuda:0'), grad: tensor([ 1.2154e-07, 4.3772e-08, -1.5050e-06, -8.4750e-08, -6.9514e-06, + 5.4203e-07, 5.9139e-08, 8.3726e-07, 2.3283e-08, 6.9514e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 250.62, cls_loss 0.0023 cls_loss_mapping 0.0024 cls_loss_causal 0.4879 re_mapping 0.0051 re_causal 0.0132 /// teacc 99.01 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.1125, -0.2249, -0.0733, ..., -0.0644, 0.1699, 0.1689], + [-0.2126, -0.1820, -0.0903, ..., -0.1611, -0.2159, -0.1223], + [-0.0741, -0.1479, 0.1334, ..., -0.2008, 0.2314, 0.0804], + ..., + [-0.1459, 0.0686, 0.0409, ..., 0.1821, -0.2156, -0.2206], + [-0.2519, 0.0669, -0.1436, ..., 0.0586, -0.0939, -0.1601], + [-0.0030, -0.1175, -0.0954, ..., -0.1020, -0.0570, -0.1956]], + device='cuda:0'), grad: tensor([[-1.2115e-05, -9.0599e-06, 0.0000e+00, ..., 9.3132e-10, + -4.5866e-05, -1.9342e-05], + [ 5.0757e-08, 1.0664e-07, 0.0000e+00, ..., 1.8766e-07, + 2.0955e-08, 1.2573e-08], + [ 3.1991e-07, 2.4028e-07, 0.0000e+00, ..., 1.8626e-09, + 1.1949e-06, 5.0385e-07], + ..., + [ 7.8231e-08, 9.1735e-08, 0.0000e+00, ..., -2.4633e-07, + 2.3283e-08, 1.1642e-08], + [ 1.8021e-06, 1.3513e-06, 0.0000e+00, ..., 9.7789e-09, + 6.7726e-06, 2.8573e-06], + [ 3.4105e-06, 2.7269e-06, 0.0000e+00, ..., 4.2375e-08, + 1.0811e-05, 4.5560e-06]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0188, -0.0386, -0.0150, -0.0223, -0.0292, 0.0057, 0.0251, -0.0107, + 0.0428, -0.0023], device='cuda:0'), grad: tensor([-6.0827e-05, 7.1852e-07, 1.6019e-06, -1.0826e-05, 2.0955e-07, + 1.1206e-05, 3.3349e-05, -1.9697e-07, 9.2462e-06, 1.5602e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 250.56, cls_loss 0.0022 cls_loss_mapping 0.0019 cls_loss_causal 0.4995 re_mapping 0.0051 re_causal 0.0136 /// teacc 99.04 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.1126, -0.2220, -0.0733, ..., -0.0647, 0.1701, 0.1690], + [-0.2128, -0.1821, -0.0903, ..., -0.1619, -0.2159, -0.1223], + [-0.0734, -0.1481, 0.1334, ..., -0.2008, 0.2320, 0.0806], + ..., + [-0.1466, 0.0687, 0.0409, ..., 0.1824, -0.2158, -0.2211], + [-0.2531, 0.0668, -0.1436, ..., 0.0585, -0.0947, -0.1610], + [-0.0050, -0.1176, -0.0954, ..., -0.1021, -0.0573, -0.1971]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 5.5879e-09, 0.0000e+00, ..., 4.7497e-08, + 4.2841e-08, 4.1910e-08], + [ 1.2573e-08, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 4.1910e-09, 5.5879e-09], + [ 2.3283e-09, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-09, 1.3970e-09], + ..., + [ 3.7253e-09, -1.7323e-07, 0.0000e+00, ..., -1.5805e-06, + 3.2596e-09, 0.0000e+00], + [ 2.1793e-07, 1.8626e-09, 0.0000e+00, ..., 1.3970e-09, + 7.8836e-07, 6.1002e-07], + [ 5.5879e-09, 4.6566e-08, 0.0000e+00, ..., 3.3388e-07, + 1.8626e-09, 1.3970e-09]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0188, -0.0387, -0.0146, -0.0221, -0.0290, 0.0058, 0.0251, -0.0106, + 0.0427, -0.0021], device='cuda:0'), grad: tensor([ 4.7497e-07, 9.7323e-08, 2.4214e-08, -7.4506e-08, 3.4645e-07, + 2.2836e-06, -7.3276e-06, -2.2035e-06, 6.7614e-06, -3.8231e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 250.55, cls_loss 0.0020 cls_loss_mapping 0.0020 cls_loss_causal 0.4693 re_mapping 0.0048 re_causal 0.0131 /// teacc 98.96 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.1130, -0.2240, -0.0733, ..., -0.0648, 0.1700, 0.1690], + [-0.2128, -0.1821, -0.0903, ..., -0.1618, -0.2159, -0.1223], + [-0.0735, -0.1484, 0.1334, ..., -0.2013, 0.2322, 0.0808], + ..., + [-0.1468, 0.0687, 0.0409, ..., 0.1822, -0.2159, -0.2213], + [-0.2542, 0.0671, -0.1436, ..., 0.0586, -0.0950, -0.1614], + [-0.0050, -0.1177, -0.0954, ..., -0.1022, -0.0573, -0.1974]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 5.5879e-09, + -1.6801e-06, -8.1863e-07], + [ 2.1420e-08, 2.9802e-08, 0.0000e+00, ..., 3.9395e-07, + 4.6566e-09, 1.8626e-09], + [ 5.5879e-09, 8.3819e-09, 0.0000e+00, ..., 4.4703e-08, + -1.3495e-06, -1.3970e-08], + ..., + [ 3.4459e-08, 3.3528e-08, 0.0000e+00, ..., -4.4703e-08, + 1.3225e-06, 2.7940e-09], + [ 1.9558e-08, 2.7008e-08, 0.0000e+00, ..., -2.2557e-06, + 9.0338e-08, 4.2841e-08], + [ 3.5390e-08, 5.5879e-08, 0.0000e+00, ..., 4.7311e-07, + 1.3877e-06, 6.7987e-07]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0189, -0.0386, -0.0147, -0.0218, -0.0305, 0.0060, 0.0251, -0.0107, + 0.0427, -0.0015], device='cuda:0'), grad: tensor([-2.8629e-06, 7.0371e-06, -3.5353e-06, 8.6054e-06, 5.8766e-07, + 2.5347e-05, 2.3320e-06, 4.5523e-06, -5.2512e-05, 1.0416e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 250.15, cls_loss 0.0024 cls_loss_mapping 0.0020 cls_loss_causal 0.4943 re_mapping 0.0048 re_causal 0.0127 /// teacc 98.91 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.1132, -0.2247, -0.0734, ..., -0.0654, 0.1703, 0.1691], + [-0.2130, -0.1822, -0.0902, ..., -0.1619, -0.2162, -0.1228], + [-0.0737, -0.1490, 0.1334, ..., -0.2019, 0.2323, 0.0807], + ..., + [-0.1470, 0.0688, 0.0409, ..., 0.1826, -0.2159, -0.2215], + [-0.2571, 0.0670, -0.1437, ..., 0.0575, -0.0980, -0.1619], + [-0.0054, -0.1178, -0.0954, ..., -0.1026, -0.0574, -0.1981]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -3.8184e-08, -2.1420e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 6.6124e-08, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 8.3819e-09, 4.6566e-09], + ..., + [ 1.8626e-09, -0.0000e+00, 0.0000e+00, ..., 9.4064e-08, + 9.3132e-10, 0.0000e+00], + [ 1.7695e-08, 1.8626e-09, 0.0000e+00, ..., -1.7881e-07, + 6.5193e-09, 7.4506e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 2.7940e-09, + 6.5193e-09, 3.7253e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0188, -0.0386, -0.0148, -0.0218, -0.0309, 0.0070, 0.0255, -0.0107, + 0.0412, -0.0016], device='cuda:0'), grad: tensor([-7.1712e-08, 1.1083e-07, 2.6077e-08, 4.4424e-07, 4.6566e-09, + -4.3027e-07, 2.4214e-08, 2.2352e-07, -3.2689e-07, -1.2107e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 250.62, cls_loss 0.0026 cls_loss_mapping 0.0028 cls_loss_causal 0.4877 re_mapping 0.0049 re_causal 0.0127 /// teacc 99.06 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.1133, -0.2252, -0.0734, ..., -0.0656, 0.1706, 0.1693], + [-0.2132, -0.1822, -0.0902, ..., -0.1605, -0.2164, -0.1229], + [-0.0738, -0.1493, 0.1334, ..., -0.2011, 0.2334, 0.0811], + ..., + [-0.1472, 0.0689, 0.0409, ..., 0.1818, -0.2163, -0.2230], + [-0.2575, 0.0669, -0.1437, ..., 0.0576, -0.0984, -0.1619], + [-0.0055, -0.1177, -0.0954, ..., -0.1022, -0.0575, -0.1990]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 1.6764e-08, 7.4506e-09], + [ 2.5146e-08, 1.3970e-08, 0.0000e+00, ..., 6.5193e-09, + 1.3039e-08, 6.5193e-09], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 5.3085e-08, + -6.7055e-08, -3.2596e-08], + ..., + [ 9.3132e-09, -2.0489e-08, 0.0000e+00, ..., -2.5146e-08, + 1.0245e-08, 4.6566e-09], + [ 8.3819e-08, 4.0978e-08, 0.0000e+00, ..., -7.1712e-08, + 2.1420e-08, 1.0245e-08], + [ 3.2596e-08, 3.2596e-08, 0.0000e+00, ..., 2.1420e-08, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0186, -0.0381, -0.0143, -0.0222, -0.0304, 0.0073, 0.0255, -0.0111, + 0.0411, -0.0021], device='cuda:0'), grad: tensor([ 9.7789e-08, 5.3085e-08, 1.8347e-07, -3.6359e-05, -4.1574e-06, + 3.6180e-05, 1.2107e-08, 1.1828e-07, -2.6170e-07, 4.1537e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 250.71, cls_loss 0.0028 cls_loss_mapping 0.0034 cls_loss_causal 0.4800 re_mapping 0.0052 re_causal 0.0134 /// teacc 99.01 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.1135, -0.2265, -0.0740, ..., -0.0658, 0.1708, 0.1695], + [-0.2148, -0.1824, -0.0901, ..., -0.1611, -0.2170, -0.1240], + [-0.0710, -0.1500, 0.1340, ..., -0.2002, 0.2356, 0.0821], + ..., + [-0.1474, 0.0692, 0.0408, ..., 0.1815, -0.2167, -0.2249], + [-0.2586, 0.0670, -0.1440, ..., 0.0578, -0.0996, -0.1625], + [-0.0058, -0.1178, -0.0954, ..., -0.1012, -0.0575, -0.1995]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.7695e-08, -9.3132e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 3.7253e-09]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0186, -0.0383, -0.0119, -0.0225, -0.0306, 0.0076, 0.0249, -0.0116, + 0.0411, -0.0009], device='cuda:0'), grad: tensor([ 9.1791e-06, -2.7463e-05, 2.8033e-07, -2.5146e-07, 9.3132e-10, + 2.3916e-06, 2.7046e-06, 8.1211e-07, 1.2577e-05, -2.3935e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 250.48, cls_loss 0.0023 cls_loss_mapping 0.0020 cls_loss_causal 0.5130 re_mapping 0.0048 re_causal 0.0132 /// teacc 99.04 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.1135, -0.2267, -0.0741, ..., -0.0658, 0.1714, 0.1699], + [-0.2150, -0.1827, -0.0901, ..., -0.1618, -0.2172, -0.1242], + [-0.0710, -0.1507, 0.1355, ..., -0.2002, 0.2369, 0.0827], + ..., + [-0.1475, 0.0699, 0.0408, ..., 0.1820, -0.2171, -0.2268], + [-0.2590, 0.0669, -0.1440, ..., 0.0577, -0.0997, -0.1630], + [-0.0059, -0.1180, -0.0954, ..., -0.1014, -0.0576, -0.2002]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.7695e-08, 0.0000e+00, ..., 2.7940e-08, + -2.0675e-07, -8.1025e-08], + [ 9.3132e-10, 1.8626e-07, 0.0000e+00, ..., 3.2037e-07, + 6.5193e-09, 2.7940e-09], + [ 0.0000e+00, 4.5635e-08, 0.0000e+00, ..., 5.6997e-07, + 1.3970e-07, 5.3085e-08], + ..., + [ 9.3132e-10, -5.9046e-07, 0.0000e+00, ..., -8.5123e-07, + 5.5879e-09, 1.8626e-09], + [ 3.7253e-09, 3.9116e-08, 0.0000e+00, ..., -3.2838e-06, + 6.5193e-09, 2.7940e-09], + [ 4.6566e-09, 1.9930e-07, 0.0000e+00, ..., 3.0454e-07, + 3.5390e-08, 1.4901e-08]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0183, -0.0383, -0.0118, -0.0223, -0.0305, 0.0072, 0.0245, -0.0116, + 0.0412, -0.0005], device='cuda:0'), grad: tensor([-3.1758e-07, 9.8720e-07, 1.4640e-06, 3.1292e-07, 1.4901e-08, + 5.2378e-06, 5.2154e-08, -2.5816e-06, -6.2101e-06, 1.0263e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 250.28, cls_loss 0.0021 cls_loss_mapping 0.0024 cls_loss_causal 0.5097 re_mapping 0.0047 re_causal 0.0133 /// teacc 99.09 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.1137, -0.2269, -0.0741, ..., -0.0658, 0.1714, 0.1699], + [-0.2150, -0.1829, -0.0901, ..., -0.1626, -0.2173, -0.1243], + [-0.0710, -0.1511, 0.1356, ..., -0.2016, 0.2370, 0.0827], + ..., + [-0.1475, 0.0703, 0.0408, ..., 0.1827, -0.2172, -0.2270], + [-0.2593, 0.0668, -0.1440, ..., 0.0577, -0.0999, -0.1630], + [-0.0063, -0.1183, -0.0954, ..., -0.1018, -0.0577, -0.2007]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, -2.7940e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.2352e-08, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., -0.0000e+00, + 6.5193e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.2107e-08, 0.0000e+00], + [ 1.7695e-08, 9.3132e-10, 0.0000e+00, ..., 8.3819e-09, + 3.7253e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, + 7.4506e-09, 1.8626e-09]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0185, -0.0384, -0.0121, -0.0222, -0.0304, 0.0068, 0.0245, -0.0115, + 0.0417, -0.0007], device='cuda:0'), grad: tensor([ 8.1025e-08, -2.4680e-07, 6.2399e-08, 1.1642e-07, 7.4506e-08, + -8.6613e-08, 6.0070e-07, 8.5402e-07, 3.4459e-08, -1.4892e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 250.35, cls_loss 0.0029 cls_loss_mapping 0.0027 cls_loss_causal 0.5217 re_mapping 0.0048 re_causal 0.0127 /// teacc 99.04 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.1127, -0.2279, -0.0741, ..., -0.0661, 0.1740, 0.1715], + [-0.2151, -0.1829, -0.0901, ..., -0.1626, -0.2173, -0.1244], + [-0.0711, -0.1517, 0.1357, ..., -0.2018, 0.2372, 0.0827], + ..., + [-0.1477, 0.0699, 0.0408, ..., 0.1830, -0.2172, -0.2275], + [-0.2602, 0.0670, -0.1440, ..., 0.0577, -0.1007, -0.1638], + [-0.0061, -0.1187, -0.0954, ..., -0.1019, -0.0595, -0.2045]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.7940e-09, 0.0000e+00, ..., 1.0245e-08, + -3.7253e-08, -3.0734e-08], + [ 1.1176e-08, 8.1956e-08, 0.0000e+00, ..., 1.8906e-07, + 1.7695e-08, 0.0000e+00], + [ 9.3132e-10, 3.5390e-08, 0.0000e+00, ..., 4.5076e-07, + 1.8626e-09, 9.3132e-10], + ..., + [ 1.8626e-09, -3.0156e-06, 0.0000e+00, ..., -3.2149e-06, + 1.8626e-09, 0.0000e+00], + [ 8.9407e-08, 2.4214e-08, 0.0000e+00, ..., 9.2201e-08, + 9.3132e-10, 9.3132e-10], + [ 5.5879e-09, 2.0117e-07, 0.0000e+00, ..., 7.1619e-07, + 4.0978e-08, 2.4214e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0154, -0.0384, -0.0129, -0.0216, -0.0306, 0.0067, 0.0231, -0.0114, + 0.0417, -0.0020], device='cuda:0'), grad: tensor([-1.8626e-09, 5.5134e-07, 1.0002e-06, 5.9232e-06, -6.5006e-07, + -4.7684e-07, 8.1863e-07, -9.0897e-06, 2.4959e-07, 1.6792e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 250.44, cls_loss 0.0021 cls_loss_mapping 0.0027 cls_loss_causal 0.5090 re_mapping 0.0051 re_causal 0.0130 /// teacc 99.09 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.1127, -0.2283, -0.0741, ..., -0.0662, 0.1742, 0.1717], + [-0.2151, -0.1832, -0.0901, ..., -0.1631, -0.2173, -0.1246], + [-0.0711, -0.1548, 0.1357, ..., -0.2047, 0.2375, 0.0830], + ..., + [-0.1477, 0.0710, 0.0408, ..., 0.1840, -0.2173, -0.2277], + [-0.2603, 0.0668, -0.1440, ..., 0.0576, -0.1010, -0.1641], + [-0.0065, -0.1191, -0.0954, ..., -0.1023, -0.0596, -0.2047]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.6822e-07, 0.0000e+00, ..., 2.4121e-07, + 6.5193e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -2.5984e-07, 0.0000e+00, ..., -2.5146e-07, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 2.5146e-08, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.9360e-08, 0.0000e+00, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0153, -0.0384, -0.0133, -0.0219, -0.0302, 0.0070, 0.0224, -0.0113, + 0.0414, -0.0023], device='cuda:0'), grad: tensor([ 1.0245e-08, 4.9360e-08, 9.3598e-07, 1.8626e-08, -7.1991e-07, + 8.3819e-09, 1.0245e-08, -6.0443e-07, 1.8999e-07, 1.0245e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 250.16, cls_loss 0.0018 cls_loss_mapping 0.0020 cls_loss_causal 0.5004 re_mapping 0.0049 re_causal 0.0132 /// teacc 99.05 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.1129, -0.2283, -0.0741, ..., -0.0663, 0.1743, 0.1718], + [-0.2151, -0.1834, -0.0901, ..., -0.1632, -0.2174, -0.1244], + [-0.0711, -0.1551, 0.1357, ..., -0.2049, 0.2377, 0.0829], + ..., + [-0.1488, 0.0712, 0.0408, ..., 0.1841, -0.2173, -0.2283], + [-0.2608, 0.0667, -0.1440, ..., 0.0576, -0.1013, -0.1648], + [-0.0067, -0.1192, -0.0954, ..., -0.1025, -0.0596, -0.2049]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.4214e-08, -1.7695e-08], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 1.2107e-08, 0.0000e+00, ..., 1.1176e-08, + -1.5926e-07, 0.0000e+00], + ..., + [ 0.0000e+00, -3.4459e-08, 0.0000e+00, ..., -3.4459e-08, + 9.4064e-08, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-09, 2.7940e-09], + [ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 6.5193e-09, + 1.2107e-08, 6.5193e-09]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0152, -0.0382, -0.0137, -0.0220, -0.0303, 0.0068, 0.0228, -0.0113, + 0.0410, -0.0024], device='cuda:0'), grad: tensor([ 3.7253e-08, -2.3283e-07, -2.0340e-06, 3.5390e-08, 6.7614e-07, + 2.4214e-08, 4.8429e-08, 1.2834e-06, 6.9849e-08, 9.1270e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 250.59, cls_loss 0.0020 cls_loss_mapping 0.0020 cls_loss_causal 0.4910 re_mapping 0.0047 re_causal 0.0129 /// teacc 99.07 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.1152, -0.2285, -0.0743, ..., -0.0663, 0.1737, 0.1713], + [-0.2152, -0.1834, -0.0902, ..., -0.1634, -0.2173, -0.1243], + [-0.0712, -0.1560, 0.1358, ..., -0.2055, 0.2378, 0.0828], + ..., + [-0.1491, 0.0715, 0.0408, ..., 0.1845, -0.2174, -0.2285], + [-0.2613, 0.0666, -0.1441, ..., 0.0576, -0.1015, -0.1651], + [-0.0071, -0.1193, -0.0954, ..., -0.1025, -0.0596, -0.2050]], + device='cuda:0'), grad: tensor([[ 1.2266e-06, 9.3132e-09, 0.0000e+00, ..., 2.7940e-09, + 8.0746e-07, 7.2364e-07], + [ 1.8626e-09, 8.3819e-08, 0.0000e+00, ..., 2.9802e-08, + 1.8626e-09, 1.8626e-09], + [ 1.8626e-09, 3.7439e-07, 0.0000e+00, ..., 1.2293e-07, + -0.0000e+00, 9.3132e-10], + ..., + [ 9.3132e-10, -5.7928e-07, 0.0000e+00, ..., -2.0768e-07, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, 1.3970e-08, 0.0000e+00, ..., -5.0291e-08, + 3.5390e-08, 2.7940e-08], + [ 2.4214e-08, 2.8871e-08, 0.0000e+00, ..., 2.7008e-08, + 4.7497e-08, 3.8184e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0158, -0.0381, -0.0141, -0.0221, -0.0301, 0.0069, 0.0238, -0.0113, + 0.0410, -0.0026], device='cuda:0'), grad: tensor([ 2.2650e-06, 3.8557e-07, 1.2014e-06, 2.1234e-07, -3.8072e-06, + 3.0547e-07, -2.3339e-06, -1.6000e-06, 1.0151e-07, 3.2671e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 250.35, cls_loss 0.0019 cls_loss_mapping 0.0017 cls_loss_causal 0.5351 re_mapping 0.0047 re_causal 0.0129 /// teacc 99.03 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.1152, -0.2284, -0.0745, ..., -0.0664, 0.1737, 0.1713], + [-0.2152, -0.1835, -0.0901, ..., -0.1635, -0.2173, -0.1245], + [-0.0712, -0.1563, 0.1361, ..., -0.2058, 0.2382, 0.0838], + ..., + [-0.1498, 0.0719, 0.0408, ..., 0.1849, -0.2175, -0.2295], + [-0.2618, 0.0665, -0.1442, ..., 0.0574, -0.1023, -0.1681], + [-0.0083, -0.1197, -0.0954, ..., -0.1028, -0.0597, -0.2051]], + device='cuda:0'), grad: tensor([[ 2.8778e-06, 5.5879e-09, 0.0000e+00, ..., 9.3132e-10, + 1.0198e-06, 1.6335e-06], + [ 7.4506e-09, 4.9360e-08, 0.0000e+00, ..., 4.9360e-08, + 9.3132e-10, 1.8626e-09], + [ 6.5193e-09, 7.4506e-09, 0.0000e+00, ..., 2.7940e-09, + -9.3132e-10, 1.8626e-09], + ..., + [ 8.3819e-09, -1.2852e-07, 0.0000e+00, ..., -1.4994e-07, + 2.7940e-09, 0.0000e+00], + [ 1.4529e-07, 1.3039e-08, 0.0000e+00, ..., 4.6566e-09, + 5.0291e-08, 7.9162e-08], + [ 6.4261e-08, 1.3504e-07, 0.0000e+00, ..., 7.7300e-08, + 5.5879e-09, 5.5879e-09]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0158, -0.0380, -0.0141, -0.0222, -0.0297, 0.0068, 0.0242, -0.0113, + 0.0405, -0.0028], device='cuda:0'), grad: tensor([ 9.4920e-06, 6.5751e-07, 2.7008e-08, -5.6159e-07, -6.3516e-07, + 7.1302e-06, -1.6823e-05, -1.1921e-07, 5.0571e-07, 2.9150e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 250.29, cls_loss 0.0032 cls_loss_mapping 0.0023 cls_loss_causal 0.4844 re_mapping 0.0048 re_causal 0.0117 /// teacc 99.00 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.1153, -0.2287, -0.0748, ..., -0.0664, 0.1744, 0.1721], + [-0.2153, -0.1836, -0.0898, ..., -0.1637, -0.2175, -0.1247], + [-0.0713, -0.1565, 0.1360, ..., -0.2061, 0.2388, 0.0845], + ..., + [-0.1505, 0.0719, 0.0408, ..., 0.1854, -0.2177, -0.2318], + [-0.2621, 0.0664, -0.1445, ..., 0.0575, -0.1026, -0.1689], + [-0.0095, -0.1207, -0.0955, ..., -0.1033, -0.0597, -0.2053]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + -0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 3.7253e-09, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.5832e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, -7.0781e-08, 0.0000e+00, ..., -9.0338e-08, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 9.3132e-10, 0.0000e+00, ..., -3.3528e-08, + 0.0000e+00, 0.0000e+00], + [ 2.9802e-08, 6.5193e-08, 0.0000e+00, ..., 1.2014e-07, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0153, -0.0379, -0.0140, -0.0218, -0.0300, 0.0065, 0.0236, -0.0114, + 0.0404, -0.0027], device='cuda:0'), grad: tensor([ 3.6322e-08, 7.4506e-09, 1.1176e-07, 3.9116e-07, 3.8184e-08, + -4.3586e-07, -2.5146e-08, -2.3656e-07, -2.2259e-07, 3.4645e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 250.49, cls_loss 0.0019 cls_loss_mapping 0.0017 cls_loss_causal 0.4741 re_mapping 0.0049 re_causal 0.0124 /// teacc 98.88 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.1157, -0.2296, -0.0768, ..., -0.0664, 0.1742, 0.1720], + [-0.2153, -0.1839, -0.0899, ..., -0.1638, -0.2176, -0.1248], + [-0.0715, -0.1575, 0.1361, ..., -0.2068, 0.2390, 0.0845], + ..., + [-0.1510, 0.0725, 0.0408, ..., 0.1860, -0.2178, -0.2323], + [-0.2627, 0.0663, -0.1445, ..., 0.0574, -0.1028, -0.1694], + [-0.0102, -0.1214, -0.0955, ..., -0.1039, -0.0598, -0.2054]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 7.4506e-09, + -5.8487e-07, -2.3935e-07], + [ 9.3132e-10, 1.9651e-06, 0.0000e+00, ..., 1.3933e-06, + 4.6566e-08, 1.8626e-09], + [ 0.0000e+00, 2.1420e-08, 0.0000e+00, ..., 4.4703e-08, + 6.2399e-08, -1.7695e-08], + ..., + [ 0.0000e+00, -2.6356e-06, 0.0000e+00, ..., -2.0657e-06, + 6.5193e-09, 9.3132e-10], + [ 2.4214e-08, 3.1665e-07, 0.0000e+00, ..., 1.9651e-07, + -1.4026e-06, 7.4506e-09], + [ 0.0000e+00, 1.1548e-07, 0.0000e+00, ..., 2.8312e-07, + 9.1083e-07, 2.2072e-07]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0156, -0.0379, -0.0142, -0.0223, -0.0301, 0.0068, 0.0244, -0.0113, + 0.0405, -0.0032], device='cuda:0'), grad: tensor([ 5.4017e-08, 4.3400e-06, 1.9316e-06, 7.0781e-07, 1.2256e-06, + 1.3085e-06, 1.5810e-05, -7.1861e-06, -2.6003e-05, 7.8082e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 250.35, cls_loss 0.0026 cls_loss_mapping 0.0020 cls_loss_causal 0.4862 re_mapping 0.0049 re_causal 0.0125 /// teacc 99.01 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.1164, -0.2323, -0.0774, ..., -0.0665, 0.1742, 0.1720], + [-0.2154, -0.1841, -0.0899, ..., -0.1641, -0.2179, -0.1252], + [-0.0716, -0.1571, 0.1369, ..., -0.2063, 0.2405, 0.0861], + ..., + [-0.1518, 0.0727, 0.0407, ..., 0.1863, -0.2182, -0.2355], + [-0.2636, 0.0665, -0.1448, ..., 0.0574, -0.1038, -0.1725], + [-0.0104, -0.1219, -0.0955, ..., -0.1042, -0.0598, -0.2056]], + device='cuda:0'), grad: tensor([[ 9.5088e-07, 2.7940e-09, 0.0000e+00, ..., 3.6675e-06, + -5.5879e-09, -4.6566e-09], + [ 2.6077e-08, 1.1176e-08, 0.0000e+00, ..., 8.5682e-08, + 9.3132e-10, 9.3132e-10], + [ 1.3970e-08, 4.6566e-09, 0.0000e+00, ..., 3.8184e-08, + -9.0338e-08, -0.0000e+00], + ..., + [ 4.2468e-07, 2.4214e-08, 0.0000e+00, ..., 1.5972e-06, + 1.8626e-09, 0.0000e+00], + [ 1.0394e-06, 1.0151e-07, 0.0000e+00, ..., 3.9637e-06, + 1.8626e-08, 1.8626e-09], + [ 4.4703e-08, 1.0245e-08, 0.0000e+00, ..., 1.6857e-07, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0158, -0.0376, -0.0135, -0.0240, -0.0304, 0.0072, 0.0258, -0.0116, + 0.0407, -0.0030], device='cuda:0'), grad: tensor([ 7.6219e-06, 2.1514e-07, -2.3190e-07, 2.2873e-06, 1.0896e-07, + -2.4199e-05, 2.3060e-06, 3.4086e-06, 8.5682e-06, -9.8720e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 250.32, cls_loss 0.0027 cls_loss_mapping 0.0030 cls_loss_causal 0.4916 re_mapping 0.0051 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.1170, -0.2368, -0.0778, ..., -0.0689, 0.1744, 0.1722], + [-0.2155, -0.1842, -0.0898, ..., -0.1649, -0.2180, -0.1253], + [-0.0716, -0.1573, 0.1371, ..., -0.2066, 0.2412, 0.0870], + ..., + [-0.1490, 0.0734, 0.0407, ..., 0.1871, -0.2183, -0.2365], + [-0.2645, 0.0664, -0.1448, ..., 0.0573, -0.1044, -0.1743], + [-0.0108, -0.1224, -0.0955, ..., -0.1045, -0.0599, -0.2060]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -8.6520e-07, -5.2713e-07], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-09, 9.3132e-10], + [ 2.7940e-09, -1.0058e-07, 0.0000e+00, ..., 1.3039e-08, + -4.2841e-07, -1.3784e-07], + ..., + [ 0.0000e+00, 1.0151e-07, 0.0000e+00, ..., -1.8626e-09, + 4.7311e-07, 1.6578e-07], + [ 9.3132e-09, 1.8626e-09, 0.0000e+00, ..., -1.4901e-08, + 2.1420e-08, 1.3039e-08], + [ 9.3132e-10, -1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 7.3202e-07, 4.4610e-07]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0162, -0.0372, -0.0135, -0.0243, -0.0293, 0.0066, 0.0269, -0.0119, + 0.0404, -0.0036], device='cuda:0'), grad: tensor([-1.3905e-06, 5.4017e-08, -1.0028e-05, 1.9558e-08, 5.8766e-07, + 3.4086e-07, 6.4261e-08, 3.3677e-06, 5.8413e-06, 1.1548e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 250.40, cls_loss 0.0026 cls_loss_mapping 0.0031 cls_loss_causal 0.4941 re_mapping 0.0048 re_causal 0.0120 /// teacc 99.04 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.1171, -0.2372, -0.0779, ..., -0.0692, 0.1751, 0.1727], + [-0.2155, -0.1840, -0.0898, ..., -0.1627, -0.2180, -0.1253], + [-0.0718, -0.1573, 0.1377, ..., -0.2080, 0.2416, 0.0871], + ..., + [-0.1497, 0.0738, 0.0407, ..., 0.1865, -0.2185, -0.2381], + [-0.2676, 0.0662, -0.1449, ..., 0.0563, -0.1046, -0.1746], + [-0.0115, -0.1231, -0.0955, ..., -0.1054, -0.0603, -0.2070]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 8.3819e-09, 0.0000e+00, ..., 8.3819e-09, + 7.4506e-09, 6.5193e-09], + [ 5.5879e-09, 3.0734e-08, 0.0000e+00, ..., 4.6566e-08, + 2.7940e-09, 9.3132e-10], + [ 1.8626e-09, 1.4156e-07, 0.0000e+00, ..., 5.3085e-08, + -1.0245e-07, -4.6566e-08], + ..., + [ 9.3132e-09, -3.8836e-07, 0.0000e+00, ..., -1.7416e-07, + 2.6077e-08, 1.2107e-08], + [ 6.2585e-07, 9.3132e-09, 0.0000e+00, ..., 3.8221e-06, + 7.9162e-08, 3.5390e-08], + [ 2.7940e-08, 5.4948e-08, 0.0000e+00, ..., 3.7253e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0160, -0.0363, -0.0137, -0.0243, -0.0293, 0.0080, 0.0263, -0.0126, + 0.0389, -0.0041], device='cuda:0'), grad: tensor([ 5.9605e-08, 1.5646e-07, 1.7788e-07, 1.4594e-06, 1.6410e-06, + -1.0870e-05, 4.2375e-07, -7.8138e-07, 9.1717e-06, -1.4342e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 250.48, cls_loss 0.0023 cls_loss_mapping 0.0021 cls_loss_causal 0.4885 re_mapping 0.0048 re_causal 0.0123 /// teacc 99.02 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.1168, -0.2373, -0.0780, ..., -0.0683, 0.1755, 0.1733], + [-0.2156, -0.1842, -0.0896, ..., -0.1629, -0.2181, -0.1254], + [-0.0719, -0.1574, 0.1376, ..., -0.2081, 0.2424, 0.0875], + ..., + [-0.1510, 0.0744, 0.0407, ..., 0.1870, -0.2188, -0.2402], + [-0.2690, 0.0661, -0.1451, ..., 0.0559, -0.1049, -0.1747], + [-0.0120, -0.1241, -0.0955, ..., -0.1061, -0.0604, -0.2075]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09], + [ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 3.7253e-09, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + -1.2107e-08, 0.0000e+00], + ..., + [ 9.3132e-10, -5.3085e-08, 0.0000e+00, ..., -8.1956e-08, + 1.8626e-09, 0.0000e+00], + [ 1.3039e-08, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 3.7253e-09], + [ 7.4506e-09, 5.7742e-08, 0.0000e+00, ..., 7.5437e-08, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0158, -0.0362, -0.0139, -0.0237, -0.0290, 0.0079, 0.0262, -0.0125, + 0.0380, -0.0046], device='cuda:0'), grad: tensor([ 3.3528e-08, -2.2817e-07, -2.6077e-08, -1.3132e-07, 2.0582e-07, + 2.7567e-07, -2.5425e-07, -2.0768e-07, 5.4948e-08, 2.8592e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 250.32, cls_loss 0.0022 cls_loss_mapping 0.0026 cls_loss_causal 0.5018 re_mapping 0.0047 re_causal 0.0123 /// teacc 98.97 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.1173, -0.2380, -0.0790, ..., -0.0682, 0.1755, 0.1735], + [-0.2158, -0.1843, -0.0896, ..., -0.1631, -0.2183, -0.1257], + [-0.0720, -0.1574, 0.1375, ..., -0.2080, 0.2435, 0.0886], + ..., + [-0.1514, 0.0746, 0.0406, ..., 0.1873, -0.2190, -0.2412], + [-0.2691, 0.0660, -0.1452, ..., 0.0559, -0.1052, -0.1752], + [-0.0154, -0.1246, -0.0955, ..., -0.1064, -0.0605, -0.2081]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 1.7695e-08, + 1.2759e-07, 9.4995e-08], + [ 1.8626e-09, 1.5832e-08, 2.3562e-07, ..., 2.7940e-08, + 3.9227e-06, 1.1269e-07], + [ 9.3132e-10, 3.5390e-08, -2.3935e-07, ..., -1.1269e-07, + -4.9695e-06, -8.6706e-07], + ..., + [ 9.3132e-10, 1.1828e-07, 1.8626e-09, ..., 4.5635e-08, + 6.5099e-07, 4.7218e-07], + [ 7.4506e-09, 9.3132e-09, 0.0000e+00, ..., 2.9802e-08, + 2.1700e-07, 1.6484e-07], + [ 2.7940e-09, 1.0245e-08, 0.0000e+00, ..., 6.5193e-09, + 1.2107e-08, 9.3132e-09]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0160, -0.0360, -0.0135, -0.0235, -0.0288, 0.0078, 0.0266, -0.0127, + 0.0382, -0.0053], device='cuda:0'), grad: tensor([ 4.0885e-07, 6.2101e-06, -9.3505e-06, -3.4086e-07, 2.3283e-08, + 3.6135e-07, -2.9244e-07, 2.2184e-06, 7.1619e-07, 4.6566e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 250.12, cls_loss 0.0022 cls_loss_mapping 0.0021 cls_loss_causal 0.4760 re_mapping 0.0049 re_causal 0.0122 /// teacc 98.99 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.1175, -0.2381, -0.0792, ..., -0.0690, 0.1757, 0.1736], + [-0.2158, -0.1849, -0.0893, ..., -0.1634, -0.2185, -0.1262], + [-0.0722, -0.1577, 0.1375, ..., -0.2084, 0.2438, 0.0894], + ..., + [-0.1521, 0.0752, 0.0405, ..., 0.1877, -0.2191, -0.2418], + [-0.2695, 0.0659, -0.1454, ..., 0.0560, -0.1046, -0.1770], + [-0.0168, -0.1248, -0.0956, ..., -0.1067, -0.0605, -0.2082]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 8.3819e-09, 3.7253e-09], + [ 9.3132e-10, 2.7940e-08, 4.2841e-08, ..., 1.7695e-08, + 6.7614e-07, 3.3528e-07], + [ 9.3132e-10, 9.3132e-10, -6.2399e-08, ..., 9.3132e-10, + -9.8627e-07, -4.9081e-07], + ..., + [ 3.7253e-09, -1.0896e-07, 4.6566e-09, ..., -8.6613e-08, + 7.7300e-08, 3.8184e-08], + [ 1.1176e-08, 7.4506e-09, 9.3132e-09, ..., -0.0000e+00, + 1.3784e-07, 6.8918e-08], + [ 1.8626e-09, 8.2888e-08, 0.0000e+00, ..., 6.5193e-08, + 4.6566e-09, 1.8626e-09]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0161, -0.0359, -0.0135, -0.0233, -0.0292, 0.0077, 0.0264, -0.0126, + 0.0381, -0.0051], device='cuda:0'), grad: tensor([ 4.2841e-08, 2.8405e-06, -4.0159e-06, -1.2293e-07, 2.6543e-07, + 2.0023e-07, 2.1141e-07, 1.7695e-07, 5.8208e-07, -1.8347e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 250.82, cls_loss 0.0025 cls_loss_mapping 0.0034 cls_loss_causal 0.5122 re_mapping 0.0048 re_causal 0.0123 /// teacc 99.01 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.1176, -0.2382, -0.0796, ..., -0.0694, 0.1760, 0.1740], + [-0.2159, -0.1850, -0.0890, ..., -0.1636, -0.2187, -0.1263], + [-0.0728, -0.1597, 0.1380, ..., -0.2089, 0.2447, 0.0898], + ..., + [-0.1511, 0.0752, 0.0402, ..., 0.1874, -0.2195, -0.2434], + [-0.2707, 0.0670, -0.1457, ..., 0.0564, -0.1051, -0.1786], + [-0.0174, -0.1252, -0.0951, ..., -0.1069, -0.0607, -0.2087]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + -1.0896e-07, -7.7300e-08], + [ 2.7940e-09, 5.5879e-09, 2.3283e-08, ..., 0.0000e+00, + 5.8673e-08, 1.8626e-09], + [ 9.3132e-10, 1.8626e-09, -9.4995e-08, ..., 0.0000e+00, + -2.1141e-07, 2.1420e-08], + ..., + [ 1.5832e-07, 3.7346e-07, 1.3970e-08, ..., 0.0000e+00, + 3.7253e-08, 9.3132e-10], + [ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 1.1176e-08], + [ 2.3283e-08, 3.7253e-09, 0.0000e+00, ..., 1.3039e-08, + 4.9360e-08, 3.4459e-08]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0160, -0.0360, -0.0137, -0.0235, -0.0292, 0.0082, 0.0265, -0.0126, + 0.0381, -0.0053], device='cuda:0'), grad: tensor([-8.4750e-08, -3.7812e-07, -3.3248e-07, -4.5076e-07, 2.0005e-06, + -6.5193e-08, 2.3935e-07, 6.5099e-07, 8.1956e-08, -1.6605e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 250.51, cls_loss 0.0022 cls_loss_mapping 0.0022 cls_loss_causal 0.4795 re_mapping 0.0047 re_causal 0.0122 /// teacc 99.00 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.1174, -0.2384, -0.0798, ..., -0.0683, 0.1762, 0.1744], + [-0.2160, -0.1854, -0.0877, ..., -0.1640, -0.2188, -0.1264], + [-0.0729, -0.1610, 0.1383, ..., -0.2121, 0.2420, 0.0901], + ..., + [-0.1529, 0.0757, 0.0395, ..., 0.1892, -0.2169, -0.2447], + [-0.2711, 0.0669, -0.1475, ..., 0.0562, -0.1055, -0.1790], + [-0.0177, -0.1253, -0.0955, ..., -0.1072, -0.0607, -0.2092]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -5.7742e-08, -3.0734e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 7.6368e-08, 2.6543e-08], + [ 0.0000e+00, -5.1223e-09, 0.0000e+00, ..., 3.2596e-09, + -1.8999e-07, -6.1933e-08], + ..., + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 1.0710e-07, 3.5856e-08], + [ 1.1176e-08, 1.8626e-09, 0.0000e+00, ..., -4.6566e-10, + 1.3504e-08, 6.9849e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.9802e-08, 1.5367e-08]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0159, -0.0361, -0.0165, -0.0237, -0.0293, 0.0088, 0.0262, -0.0113, + 0.0378, -0.0052], device='cuda:0'), grad: tensor([-1.0384e-07, 1.7555e-07, -4.1863e-07, 1.8161e-08, -2.6962e-07, + 1.4482e-07, -1.3597e-07, 3.2037e-07, 1.2340e-07, 1.6298e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 250.73, cls_loss 0.0021 cls_loss_mapping 0.0021 cls_loss_causal 0.4700 re_mapping 0.0050 re_causal 0.0126 /// teacc 99.04 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.1175, -0.2385, -0.0799, ..., -0.0683, 0.1764, 0.1746], + [-0.2161, -0.1856, -0.0871, ..., -0.1642, -0.2188, -0.1265], + [-0.0729, -0.1608, 0.1382, ..., -0.2121, 0.2422, 0.0903], + ..., + [-0.1530, 0.0758, 0.0393, ..., 0.1897, -0.2170, -0.2452], + [-0.2713, 0.0668, -0.1482, ..., 0.0563, -0.1060, -0.1798], + [-0.0186, -0.1256, -0.0957, ..., -0.1080, -0.0610, -0.2098]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.5611e-08, 4.6566e-10, ..., 9.3132e-09, + -1.4901e-08, -9.3132e-09], + [ 1.3970e-09, 2.6962e-07, 0.0000e+00, ..., 5.2946e-07, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-10, 2.7940e-08, 0.0000e+00, ..., 4.5169e-08, + 6.0536e-09, 4.1910e-09], + ..., + [ 1.3970e-09, -1.3215e-06, 0.0000e+00, ..., -1.3579e-06, + 4.6566e-10, 0.0000e+00], + [ 9.3132e-10, 1.2945e-07, 4.6566e-10, ..., 1.4901e-07, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 3.6275e-07, -9.3132e-10, ..., 2.9057e-07, + 6.0536e-09, 3.7253e-09]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0158, -0.0356, -0.0165, -0.0234, -0.0294, 0.0086, 0.0262, -0.0117, + 0.0375, -0.0055], device='cuda:0'), grad: tensor([ 3.6787e-08, 7.8464e-07, 1.8254e-07, 4.4098e-07, 4.7358e-07, + 2.3656e-07, 3.5390e-08, -4.1686e-06, 9.1596e-07, 1.0598e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 250.32, cls_loss 0.0021 cls_loss_mapping 0.0016 cls_loss_causal 0.4917 re_mapping 0.0050 re_causal 0.0130 /// teacc 99.05 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.1172, -0.2387, -0.0801, ..., -0.0672, 0.1768, 0.1751], + [-0.2161, -0.1858, -0.0862, ..., -0.1645, -0.2193, -0.1267], + [-0.0731, -0.1643, 0.1380, ..., -0.2128, 0.2424, 0.0901], + ..., + [-0.1546, 0.0769, 0.0391, ..., 0.1903, -0.2170, -0.2458], + [-0.2731, 0.0660, -0.1485, ..., 0.0557, -0.1060, -0.1799], + [-0.0189, -0.1262, -0.0960, ..., -0.1086, -0.0611, -0.2104]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 4.6566e-10, 4.6566e-10, ..., 9.3132e-10, + -4.0978e-08, -2.7008e-08], + [ 9.3132e-10, 4.8662e-07, 0.0000e+00, ..., 9.4529e-07, + 1.3970e-09, 9.3132e-10], + [ 2.7940e-09, 2.3283e-09, 0.0000e+00, ..., 3.7253e-09, + 6.0536e-09, 4.1910e-09], + ..., + [ 0.0000e+00, -5.3924e-07, 0.0000e+00, ..., -1.0468e-06, + 1.8626e-09, 1.3970e-09], + [ 3.2596e-09, -0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.7789e-09, 6.5193e-09], + [ 0.0000e+00, 4.3306e-08, 0.0000e+00, ..., 8.0094e-08, + 3.5390e-08, 2.4214e-08]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0156, -0.0355, -0.0167, -0.0216, -0.0296, 0.0072, 0.0260, -0.0116, + 0.0372, -0.0057], device='cuda:0'), grad: tensor([-2.7474e-08, 2.8275e-06, 5.4017e-08, 2.9337e-08, 6.2399e-08, + 2.0023e-08, -6.7055e-08, -3.1292e-06, -4.3772e-08, 2.6450e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 250.27, cls_loss 0.0022 cls_loss_mapping 0.0017 cls_loss_causal 0.4996 re_mapping 0.0047 re_causal 0.0122 /// teacc 99.06 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.1174, -0.2388, -0.0802, ..., -0.0671, 0.1769, 0.1752], + [-0.2162, -0.1861, -0.0858, ..., -0.1647, -0.2191, -0.1267], + [-0.0758, -0.1644, 0.1378, ..., -0.2128, 0.2413, 0.0883], + ..., + [-0.1550, 0.0772, 0.0391, ..., 0.1908, -0.2171, -0.2469], + [-0.2735, 0.0659, -0.1487, ..., 0.0557, -0.1059, -0.1805], + [-0.0206, -0.1268, -0.0960, ..., -0.1091, -0.0612, -0.2106]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 2.5611e-08, 0.0000e+00, ..., 4.6566e-10, + -4.2208e-06, -2.7847e-06], + [ 3.2596e-09, 6.9849e-09, 0.0000e+00, ..., 2.7940e-09, + 2.0489e-08, 2.0955e-08], + [ 9.3132e-09, 1.6764e-08, 0.0000e+00, ..., 3.7253e-09, + 1.2992e-06, 8.5216e-07], + ..., + [ 7.4506e-09, 9.3132e-09, 0.0000e+00, ..., -1.8626e-09, + 5.4482e-08, 3.6787e-08], + [ 1.9558e-08, 2.7474e-08, 0.0000e+00, ..., -3.1199e-08, + 1.8114e-07, 1.1828e-07], + [ 2.2817e-08, 4.2841e-08, 0.0000e+00, ..., 3.2596e-09, + 3.1432e-07, 2.0675e-07]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0156, -0.0353, -0.0177, -0.0215, -0.0298, 0.0063, 0.0280, -0.0115, + 0.0383, -0.0060], device='cuda:0'), grad: tensor([-6.3442e-06, -7.9628e-08, 2.1420e-06, -5.0664e-07, 3.6787e-08, + 5.3551e-07, 3.4608e-06, 2.0536e-07, -5.5879e-09, 5.6624e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 250.32, cls_loss 0.0028 cls_loss_mapping 0.0028 cls_loss_causal 0.4976 re_mapping 0.0047 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.1185, -0.2389, -0.0802, ..., -0.0669, 0.1769, 0.1752], + [-0.2175, -0.1864, -0.0858, ..., -0.1656, -0.2192, -0.1276], + [-0.0757, -0.1648, 0.1378, ..., -0.2130, 0.2414, 0.0893], + ..., + [-0.1561, 0.0791, 0.0391, ..., 0.1922, -0.2172, -0.2494], + [-0.2737, 0.0658, -0.1487, ..., 0.0555, -0.1061, -0.1812], + [-0.0208, -0.1296, -0.0960, ..., -0.1112, -0.0615, -0.2119]], + device='cuda:0'), grad: tensor([[ 1.6438e-07, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 3.7253e-09], + [ 6.6496e-06, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 9.3132e-10, 9.3132e-10], + [ 7.9162e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -8.3819e-09, -6.0536e-09], + ..., + [ 7.9162e-09, -5.3551e-08, 0.0000e+00, ..., -6.0536e-08, + 8.3819e-09, 6.0536e-09], + [ 2.0973e-06, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 1.3970e-09, 1.1083e-07], + [ 9.3272e-07, 3.7253e-08, 0.0000e+00, ..., 4.2841e-08, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0157, -0.0347, -0.0176, -0.0212, -0.0306, 0.0063, 0.0288, -0.0117, + 0.0382, -0.0069], device='cuda:0'), grad: tensor([ 5.0897e-07, 3.7521e-05, 3.9302e-06, 3.4213e-04, 1.0338e-07, + -3.7241e-04, -3.5344e-07, -2.1443e-05, 6.5528e-06, 3.1218e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 250.30, cls_loss 0.0020 cls_loss_mapping 0.0014 cls_loss_causal 0.4994 re_mapping 0.0051 re_causal 0.0129 /// teacc 99.09 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.1182, -0.2390, -0.0802, ..., -0.0657, 0.1771, 0.1757], + [-0.2183, -0.1875, -0.0857, ..., -0.1670, -0.2193, -0.1281], + [-0.0757, -0.1649, 0.1378, ..., -0.2129, 0.2417, 0.0902], + ..., + [-0.1562, 0.0810, 0.0391, ..., 0.1941, -0.2175, -0.2517], + [-0.2739, 0.0657, -0.1487, ..., 0.0553, -0.1064, -0.1816], + [-0.0213, -0.1316, -0.0960, ..., -0.1130, -0.0616, -0.2122]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-09, 5.5879e-09], + [ 4.6566e-09, 2.7940e-09, 0.0000e+00, ..., 4.6566e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, -1.1642e-08, 0.0000e+00, ..., -1.3504e-08, + 0.0000e+00, 0.0000e+00], + [ 1.9558e-07, 9.3132e-10, 0.0000e+00, ..., 7.9162e-09, + 5.2620e-08, 1.3877e-07], + [ 8.8476e-09, 1.0245e-08, 0.0000e+00, ..., 1.1642e-08, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0155, -0.0347, -0.0174, -0.0216, -0.0308, 0.0067, 0.0282, -0.0112, + 0.0377, -0.0080], device='cuda:0'), grad: tensor([ 3.4459e-08, -4.7497e-07, 7.9162e-09, 1.8813e-07, 1.2526e-07, + -3.8650e-08, -6.4401e-07, -1.1176e-08, 7.5018e-07, 5.5414e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 250.45, cls_loss 0.0020 cls_loss_mapping 0.0021 cls_loss_causal 0.4964 re_mapping 0.0050 re_causal 0.0127 /// teacc 99.06 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.1185, -0.2392, -0.0802, ..., -0.0657, 0.1775, 0.1760], + [-0.2187, -0.1878, -0.0854, ..., -0.1672, -0.2195, -0.1285], + [-0.0757, -0.1651, 0.1377, ..., -0.2130, 0.2418, 0.0902], + ..., + [-0.1563, 0.0813, 0.0391, ..., 0.1944, -0.2175, -0.2525], + [-0.2741, 0.0655, -0.1488, ..., 0.0553, -0.1068, -0.1824], + [-0.0218, -0.1319, -0.0960, ..., -0.1133, -0.0617, -0.2128]], + device='cuda:0'), grad: tensor([[-6.9849e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -3.3528e-08, -2.3283e-08], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 4.6566e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.0571e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.0896e-07, 1.8626e-09]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0154, -0.0347, -0.0175, -0.0231, -0.0308, 0.0076, 0.0290, -0.0112, + 0.0376, -0.0081], device='cuda:0'), grad: tensor([ 2.6077e-08, -1.8785e-06, -4.0699e-07, 1.0598e-06, 1.3914e-06, + -6.4587e-07, 2.8173e-07, 4.2142e-07, 7.6508e-07, -1.0068e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 250.33, cls_loss 0.0025 cls_loss_mapping 0.0022 cls_loss_causal 0.4829 re_mapping 0.0049 re_causal 0.0123 /// teacc 99.04 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.1174, -0.2402, -0.0802, ..., -0.0629, 0.1803, 0.1783], + [-0.2190, -0.1881, -0.0854, ..., -0.1677, -0.2200, -0.1304], + [-0.0757, -0.1654, 0.1377, ..., -0.2131, 0.2419, 0.0897], + ..., + [-0.1571, 0.0815, 0.0391, ..., 0.1948, -0.2177, -0.2541], + [-0.2743, 0.0658, -0.1488, ..., 0.0553, -0.1076, -0.1833], + [-0.0222, -0.1320, -0.0960, ..., -0.1136, -0.0619, -0.2135]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 1.3039e-08, 0.0000e+00, ..., 5.2154e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 2.3283e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3970e-09, -1.7229e-08, 0.0000e+00, ..., -9.1735e-08, + 0.0000e+00, 0.0000e+00], + [ 4.2841e-08, 6.5193e-09, 0.0000e+00, ..., 1.7695e-08, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 9.7789e-09, 0.0000e+00, ..., 3.1665e-08, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0132, -0.0346, -0.0176, -0.0234, -0.0315, 0.0076, 0.0262, -0.0112, + 0.0386, -0.0080], device='cuda:0'), grad: tensor([ 1.5832e-08, 8.0559e-08, 2.2398e-07, -2.6636e-07, -1.1642e-08, + 1.3132e-07, 2.8405e-08, -1.4389e-07, -1.3970e-07, 9.2201e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 250.43, cls_loss 0.0021 cls_loss_mapping 0.0020 cls_loss_causal 0.4857 re_mapping 0.0048 re_causal 0.0125 /// teacc 99.09 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.1142, -0.2401, -0.0804, ..., -0.0615, 0.1838, 0.1819], + [-0.2191, -0.1883, -0.0854, ..., -0.1679, -0.2201, -0.1312], + [-0.0759, -0.1655, 0.1381, ..., -0.2131, 0.2422, 0.0896], + ..., + [-0.1578, 0.0820, 0.0390, ..., 0.1953, -0.2180, -0.2577], + [-0.2748, 0.0658, -0.1491, ..., 0.0550, -0.1086, -0.1861], + [-0.0225, -0.1326, -0.0960, ..., -0.1141, -0.0619, -0.2135]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 4.6566e-10], + [ 2.7940e-09, 2.8405e-08, 3.2596e-09, ..., 3.7719e-08, + 3.6322e-08, 1.0245e-08], + [ 1.8626e-09, 4.1910e-09, -7.4506e-09, ..., 1.3970e-09, + -8.8941e-08, -2.5611e-08], + ..., + [ 1.8626e-09, -1.3923e-07, 1.8626e-09, ..., -2.0768e-07, + 2.0023e-08, 5.5879e-09], + [ 1.6298e-08, 1.4435e-08, 2.3283e-09, ..., 3.2596e-09, + 2.7474e-08, 7.9162e-09], + [ 2.4214e-08, 1.4203e-07, 0.0000e+00, ..., 1.6764e-07, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0096, -0.0346, -0.0175, -0.0234, -0.0315, 0.0074, 0.0226, -0.0111, + 0.0384, -0.0082], device='cuda:0'), grad: tensor([ 2.2817e-08, 1.9418e-07, -1.6438e-07, -2.4633e-07, 1.1176e-08, + 7.7300e-08, 3.2131e-08, -6.4960e-07, 7.7300e-08, 6.4960e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 250.14, cls_loss 0.0019 cls_loss_mapping 0.0023 cls_loss_causal 0.4776 re_mapping 0.0046 re_causal 0.0120 /// teacc 99.07 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.1142, -0.2403, -0.0822, ..., -0.0616, 0.1839, 0.1820], + [-0.2191, -0.1888, -0.0855, ..., -0.1683, -0.2202, -0.1314], + [-0.0760, -0.1654, 0.1389, ..., -0.2132, 0.2424, 0.0897], + ..., + [-0.1578, 0.0828, 0.0390, ..., 0.1960, -0.2181, -0.2583], + [-0.2751, 0.0657, -0.1488, ..., 0.0549, -0.1092, -0.1874], + [-0.0228, -0.1333, -0.0958, ..., -0.1149, -0.0619, -0.2138]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + -6.6170e-07, -3.9674e-07], + [ 4.1910e-09, 1.3039e-08, 0.0000e+00, ..., 2.7940e-08, + 3.5856e-08, 2.1886e-08], + [ 2.7940e-09, 4.6566e-09, 0.0000e+00, ..., 2.8405e-08, + -4.4703e-08, 3.7253e-09], + ..., + [ 1.8626e-09, -2.1886e-08, 0.0000e+00, ..., -2.3749e-08, + 4.6566e-09, 1.8626e-09], + [ 4.6566e-09, 6.0536e-09, 0.0000e+00, ..., -7.4506e-08, + 1.9558e-08, 1.0710e-08], + [ 2.3283e-09, 1.7229e-08, -0.0000e+00, ..., 1.8161e-08, + 1.8300e-07, 1.0896e-07]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0096, -0.0343, -0.0175, -0.0235, -0.0308, 0.0074, 0.0226, -0.0111, + 0.0383, -0.0092], device='cuda:0'), grad: tensor([-1.4864e-06, 2.8452e-07, 2.5611e-08, 9.4576e-07, -2.6077e-08, + 1.3411e-07, 5.4948e-08, -1.4435e-08, -4.1537e-07, 5.0664e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 250.26, cls_loss 0.0021 cls_loss_mapping 0.0020 cls_loss_causal 0.4935 re_mapping 0.0045 re_causal 0.0118 /// teacc 99.03 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.1142, -0.2409, -0.0824, ..., -0.0616, 0.1840, 0.1821], + [-0.2195, -0.1889, -0.0855, ..., -0.1687, -0.2207, -0.1349], + [-0.0756, -0.1656, 0.1390, ..., -0.2132, 0.2430, 0.0927], + ..., + [-0.1583, 0.0829, 0.0389, ..., 0.1962, -0.2184, -0.2614], + [-0.2753, 0.0661, -0.1481, ..., 0.0553, -0.1096, -0.1886], + [-0.0235, -0.1335, -0.0957, ..., -0.1151, -0.0620, -0.2141]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-09, 1.3970e-09], + [ 1.3970e-09, 2.5146e-08, 0.0000e+00, ..., 2.2352e-08, + 2.3283e-09, 1.3970e-09], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., -6.0536e-09, + -4.0047e-08, -3.2131e-08], + ..., + [ 1.3970e-09, -4.0513e-08, 0.0000e+00, ..., -4.0978e-08, + 3.0268e-08, 2.6077e-08], + [ 1.3970e-09, 6.5193e-09, 0.0000e+00, ..., 4.1910e-09, + 3.2596e-09, 2.3283e-09], + [ 2.3283e-09, 5.1223e-09, 0.0000e+00, ..., 6.9849e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0095, -0.0344, -0.0172, -0.0236, -0.0307, 0.0076, 0.0225, -0.0112, + 0.0385, -0.0091], device='cuda:0'), grad: tensor([ 1.0245e-08, 1.3784e-07, -1.1036e-07, 3.3528e-08, 9.0748e-06, + 9.3132e-10, 6.0536e-09, -1.5832e-08, 3.1851e-07, -9.4324e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 250.54, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.4848 re_mapping 0.0046 re_causal 0.0125 /// teacc 98.98 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.1143, -0.2411, -0.0826, ..., -0.0616, 0.1840, 0.1821], + [-0.2196, -0.1896, -0.0855, ..., -0.1692, -0.2209, -0.1350], + [-0.0757, -0.1657, 0.1390, ..., -0.2132, 0.2434, 0.0933], + ..., + [-0.1586, 0.0831, 0.0387, ..., 0.1968, -0.2186, -0.2626], + [-0.2755, 0.0661, -0.1480, ..., 0.0553, -0.1098, -0.1888], + [-0.0241, -0.1330, -0.0949, ..., -0.1156, -0.0618, -0.2147]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 1.3970e-09, 0.0000e+00, ..., -2.0955e-08, + -2.6077e-07, -2.7521e-07], + [ 9.3132e-10, 1.2387e-07, 0.0000e+00, ..., 1.9278e-07, + 2.0489e-08, 2.1886e-08], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 5.1223e-09, + 1.3970e-08, 1.4901e-08], + ..., + [ 0.0000e+00, -4.3865e-07, 0.0000e+00, ..., -6.2259e-07, + 1.7229e-08, 1.7695e-08], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 1.8161e-08, + 1.9046e-07, 2.0163e-07], + [ 0.0000e+00, 3.0268e-07, 0.0000e+00, ..., 4.1444e-07, + 2.0955e-08, 2.0023e-08]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0096, -0.0345, -0.0172, -0.0239, -0.0306, 0.0079, 0.0225, -0.0120, + 0.0386, -0.0071], device='cuda:0'), grad: tensor([-6.4075e-07, 7.9675e-07, 4.6566e-08, 1.9558e-08, 1.6298e-08, + 6.1933e-08, -5.2620e-08, -1.9316e-06, 4.7917e-07, 1.2126e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 250.52, cls_loss 0.0025 cls_loss_mapping 0.0022 cls_loss_causal 0.5022 re_mapping 0.0045 re_causal 0.0115 /// teacc 99.07 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.1143, -0.2416, -0.0826, ..., -0.0611, 0.1840, 0.1822], + [-0.2197, -0.1927, -0.0855, ..., -0.1720, -0.2244, -0.1371], + [-0.0759, -0.1663, 0.1395, ..., -0.2134, 0.2460, 0.0941], + ..., + [-0.1594, 0.0856, 0.0386, ..., 0.1987, -0.2187, -0.2639], + [-0.2768, 0.0662, -0.1483, ..., 0.0548, -0.1090, -0.1914], + [-0.0242, -0.1333, -0.0946, ..., -0.1162, -0.0620, -0.2153]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 1.3970e-09], + [ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-09, 4.1910e-09], + [ 0.0000e+00, -5.9139e-08, 0.0000e+00, ..., -1.9977e-07, + -3.3155e-07, -2.6310e-07], + ..., + [ 4.6566e-10, 5.0757e-08, 0.0000e+00, ..., 1.8300e-07, + 3.1991e-07, 2.5472e-07], + [ 3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 6.5193e-09, + 1.3970e-08, 1.1176e-08], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 3.7253e-09, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0096, -0.0361, -0.0159, -0.0240, -0.0301, 0.0082, 0.0224, -0.0110, + 0.0381, -0.0073], device='cuda:0'), grad: tensor([ 4.0047e-08, 6.6124e-08, -8.8289e-07, 9.3132e-10, 7.6462e-07, + 1.0198e-07, -3.9581e-08, 8.9267e-07, 6.5044e-06, -7.4357e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 250.28, cls_loss 0.0019 cls_loss_mapping 0.0023 cls_loss_causal 0.4658 re_mapping 0.0048 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.1143, -0.2417, -0.0828, ..., -0.0608, 0.1840, 0.1822], + [-0.2197, -0.1928, -0.0855, ..., -0.1721, -0.2249, -0.1363], + [-0.0759, -0.1667, 0.1396, ..., -0.2135, 0.2467, 0.0946], + ..., + [-0.1598, 0.0856, 0.0383, ..., 0.1989, -0.2190, -0.2663], + [-0.2770, 0.0661, -0.1476, ..., 0.0549, -0.1091, -0.1915], + [-0.0242, -0.1335, -0.0927, ..., -0.1166, -0.0618, -0.2161]], + device='cuda:0'), grad: tensor([[ 1.5600e-07, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.0617e-07, 9.4064e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.5832e-08, 4.6566e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.1176e-08, -2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 5.1223e-09, 9.3132e-10], + [ 1.1642e-08, 0.0000e+00, 0.0000e+00, ..., -1.3793e-06, + 1.6298e-08, 1.3039e-08], + [ 4.6566e-10, 0.0000e+00, -2.3283e-09, ..., 0.0000e+00, + 4.0513e-08, 2.9337e-08]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0096, -0.0359, -0.0158, -0.0236, -0.0295, 0.0079, 0.0224, -0.0112, + 0.0382, -0.0075], device='cuda:0'), grad: tensor([ 2.8405e-07, -2.0675e-07, 1.6578e-07, 1.3970e-08, -1.1154e-05, + 4.4703e-07, 6.9514e-06, 1.0982e-05, -7.6890e-06, 2.2771e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 249.90, cls_loss 0.0020 cls_loss_mapping 0.0015 cls_loss_causal 0.4873 re_mapping 0.0047 re_causal 0.0120 /// teacc 99.10 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.1143, -0.2419, -0.0820, ..., -0.0608, 0.1841, 0.1823], + [-0.2198, -0.1929, -0.0855, ..., -0.1723, -0.2251, -0.1368], + [-0.0759, -0.1675, 0.1394, ..., -0.2135, 0.2472, 0.0956], + ..., + [-0.1601, 0.0857, 0.0382, ..., 0.1991, -0.2194, -0.2688], + [-0.2772, 0.0657, -0.1475, ..., 0.0552, -0.1083, -0.1917], + [-0.0244, -0.1337, -0.0922, ..., -0.1168, -0.0620, -0.2167]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -1.0943e-07, -4.9360e-08], + [ 2.7940e-09, 4.1910e-09, 0.0000e+00, ..., 6.0536e-09, + 1.7229e-08, 1.3504e-08], + [ 3.6787e-08, 6.0536e-09, 0.0000e+00, ..., 6.0536e-09, + 1.3970e-09, 3.2596e-09], + ..., + [ 4.6566e-10, -3.3528e-08, 0.0000e+00, ..., -5.4482e-08, + 2.3283e-09, 9.3132e-10], + [ 7.2503e-07, 2.1420e-08, 4.6566e-10, ..., 2.3749e-08, + 1.1511e-06, 1.1623e-06], + [-3.0873e-07, 8.3819e-09, -1.8626e-09, ..., 1.3039e-08, + 4.4238e-08, 2.8405e-08]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0096, -0.0357, -0.0157, -0.0235, -0.0281, 0.0079, 0.0223, -0.0112, + 0.0380, -0.0083], device='cuda:0'), grad: tensor([ 4.1444e-08, 1.6624e-07, 2.5751e-07, 5.5414e-08, 3.9488e-06, + 6.8732e-07, -8.1137e-06, -1.2852e-07, 9.9167e-06, -6.8322e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 250.88, cls_loss 0.0022 cls_loss_mapping 0.0020 cls_loss_causal 0.5134 re_mapping 0.0046 re_causal 0.0122 /// teacc 99.06 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.1145, -0.2421, -0.0798, ..., -0.0608, 0.1842, 0.1824], + [-0.2217, -0.1931, -0.0855, ..., -0.1727, -0.2253, -0.1384], + [-0.0757, -0.1679, 0.1392, ..., -0.2137, 0.2472, 0.0968], + ..., + [-0.1603, 0.0866, 0.0382, ..., 0.1999, -0.2196, -0.2698], + [-0.2779, 0.0658, -0.1476, ..., 0.0553, -0.1088, -0.1926], + [-0.0247, -0.1351, -0.0919, ..., -0.1182, -0.0622, -0.2177]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.5926e-08, 0.0000e+00, ..., -3.0175e-07, + -6.7567e-07, -6.6776e-07], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 1.0710e-08, + 6.9849e-09, 6.9849e-09], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 2.9802e-08, + 2.5611e-08, 2.8871e-08], + ..., + [ 0.0000e+00, -9.2667e-08, 0.0000e+00, ..., -1.0198e-07, + 4.7032e-08, 4.5169e-08], + [ 1.3970e-09, 5.4948e-08, 0.0000e+00, ..., 6.7987e-08, + 7.4506e-09, 7.4506e-09], + [-2.8871e-08, 7.9162e-09, 0.0000e+00, ..., 1.6764e-08, + 2.8871e-08, 2.7940e-08]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0096, -0.0358, -0.0158, -0.0231, -0.0294, 0.0075, 0.0224, -0.0108, + 0.0376, -0.0087], device='cuda:0'), grad: tensor([-1.9297e-06, 4.1444e-08, 1.1316e-07, 1.1176e-07, 2.3749e-08, + 6.7428e-07, 1.0571e-06, -1.8813e-07, 1.9139e-07, -9.1735e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 250.68, cls_loss 0.0017 cls_loss_mapping 0.0015 cls_loss_causal 0.4881 re_mapping 0.0044 re_causal 0.0119 /// teacc 98.94 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.1145, -0.2424, -0.0797, ..., -0.0604, 0.1843, 0.1825], + [-0.2223, -0.1932, -0.0855, ..., -0.1728, -0.2254, -0.1387], + [-0.0758, -0.1684, 0.1393, ..., -0.2139, 0.2473, 0.0972], + ..., + [-0.1612, 0.0853, 0.0382, ..., 0.1998, -0.2197, -0.2706], + [-0.2781, 0.0665, -0.1477, ..., 0.0558, -0.1089, -0.1930], + [-0.0247, -0.1353, -0.0918, ..., -0.1184, -0.0623, -0.2178]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 2.7940e-09, 0.0000e+00, ..., 6.0536e-09, + -1.0896e-07, -4.7032e-08], + [ 9.3132e-10, 3.0594e-07, 0.0000e+00, ..., 7.3761e-07, + 3.2596e-09, 8.8476e-09], + [ 2.3283e-09, 2.8312e-07, 0.0000e+00, ..., 6.7148e-07, + 8.3819e-08, 4.3306e-08], + ..., + [ 7.9162e-09, -6.2305e-07, 0.0000e+00, ..., -1.5581e-06, + 4.6566e-09, -9.7789e-09], + [ 1.2573e-08, 2.2352e-08, 0.0000e+00, ..., -3.6322e-08, + 2.3283e-09, 4.6566e-09], + [ 1.3970e-09, 1.5367e-08, 0.0000e+00, ..., 3.7253e-08, + 1.6764e-08, 8.3819e-09]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0096, -0.0358, -0.0158, -0.0225, -0.0295, 0.0075, 0.0224, -0.0110, + 0.0379, -0.0086], device='cuda:0'), grad: tensor([-2.0070e-07, 1.8794e-06, 1.8999e-06, 1.8859e-07, -1.8859e-07, + 1.4622e-07, -3.8184e-08, -3.7216e-06, -8.3353e-08, 1.3784e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 250.59, cls_loss 0.0025 cls_loss_mapping 0.0021 cls_loss_causal 0.4816 re_mapping 0.0047 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.1148, -0.2426, -0.0800, ..., -0.0604, 0.1855, 0.1836], + [-0.2229, -0.1933, -0.0856, ..., -0.1731, -0.2256, -0.1397], + [-0.0753, -0.1691, 0.1415, ..., -0.2142, 0.2478, 0.0989], + ..., + [-0.1618, 0.0847, 0.0366, ..., 0.1994, -0.2200, -0.2733], + [-0.2787, 0.0680, -0.1478, ..., 0.0570, -0.1076, -0.1914], + [-0.0254, -0.1354, -0.0918, ..., -0.1188, -0.0645, -0.2216]], + device='cuda:0'), grad: tensor([[ 2.2491e-07, 3.2131e-07, 4.6566e-10, ..., 0.0000e+00, + 1.0394e-06, 5.3085e-07], + [ 1.3970e-09, 5.5879e-09, 9.3132e-10, ..., 5.1223e-09, + 5.1200e-05, 4.1947e-06], + [ 1.4435e-08, 2.0955e-08, 0.0000e+00, ..., 1.3970e-09, + -5.2840e-05, -4.2766e-06], + ..., + [ 4.6566e-10, -3.3993e-08, 3.2596e-09, ..., -3.9116e-08, + 8.4145e-07, 6.9384e-08], + [ 2.1420e-08, 2.5146e-08, 9.3132e-10, ..., 2.4680e-08, + 1.6345e-07, -5.1223e-09], + [ 1.3970e-09, 4.6566e-09, -6.9849e-09, ..., 5.1223e-09, + 6.3796e-08, 5.4017e-08]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0085, -0.0359, -0.0161, -0.0223, -0.0295, 0.0075, 0.0225, -0.0110, + 0.0388, -0.0104], device='cuda:0'), grad: tensor([ 3.0845e-06, 1.6296e-04, -1.6797e-04, -1.7006e-06, 7.9162e-07, + -7.4832e-07, 6.1607e-07, 2.8666e-06, 4.1770e-07, -2.9802e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 250.30, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4767 re_mapping 0.0046 re_causal 0.0118 /// teacc 98.94 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.1151, -0.2428, -0.0800, ..., -0.0605, 0.1857, 0.1839], + [-0.2236, -0.1965, -0.0857, ..., -0.1769, -0.2267, -0.1409], + [-0.0754, -0.1698, 0.1420, ..., -0.2148, 0.2484, 0.0994], + ..., + [-0.1621, 0.0873, 0.0361, ..., 0.2022, -0.2201, -0.2743], + [-0.2789, 0.0679, -0.1479, ..., 0.0571, -0.1078, -0.1917], + [-0.0250, -0.1356, -0.0913, ..., -0.1191, -0.0651, -0.2222]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -1.9092e-08, -1.6298e-08], + [ 9.3132e-10, 3.0268e-08, 0.0000e+00, ..., 3.5390e-08, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 1.7229e-08, 0.0000e+00, ..., 3.5390e-08, + -4.1910e-08, -1.2107e-08], + ..., + [ 1.3970e-09, -1.3364e-07, 0.0000e+00, ..., -1.5367e-07, + 1.8626e-09, 1.8626e-09], + [ 1.3970e-08, 2.3283e-08, 4.6566e-10, ..., -3.7253e-09, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 3.7719e-08, -9.3132e-10, ..., 4.3772e-08, + 2.0023e-08, 2.0489e-08]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0082, -0.0379, -0.0164, -0.0222, -0.0307, 0.0074, 0.0226, -0.0090, + 0.0389, -0.0107], device='cuda:0'), grad: tensor([-4.3772e-08, 9.0804e-08, 1.6531e-07, 7.4971e-08, 1.0757e-07, + 2.5611e-08, -4.1910e-09, -3.8370e-07, -1.3597e-07, 1.0990e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 250.54, cls_loss 0.0024 cls_loss_mapping 0.0026 cls_loss_causal 0.4864 re_mapping 0.0048 re_causal 0.0115 /// teacc 99.09 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.1151, -0.2430, -0.0801, ..., -0.0603, 0.1858, 0.1839], + [-0.2236, -0.1971, -0.0855, ..., -0.1778, -0.2271, -0.1410], + [-0.0755, -0.1700, 0.1425, ..., -0.2150, 0.2489, 0.0996], + ..., + [-0.1631, 0.0882, 0.0352, ..., 0.2031, -0.2203, -0.2748], + [-0.2794, 0.0681, -0.1486, ..., 0.0572, -0.1078, -0.1923], + [-0.0256, -0.1361, -0.0907, ..., -0.1197, -0.0651, -0.2222]], + device='cuda:0'), grad: tensor([[ 2.5611e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, 4.1910e-09], + [ 1.4901e-08, 7.4506e-09, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 9.3132e-10], + [ 3.7253e-09, 2.7474e-08, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 1.6764e-08, -2.3283e-09, 0.0000e+00, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 1.7369e-07, -6.0536e-08, 0.0000e+00, ..., -6.9849e-09, + 1.8626e-09, 2.3283e-09], + [ 7.8231e-08, 5.4948e-08, 0.0000e+00, ..., 8.8476e-09, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0083, -0.0379, -0.0163, -0.0212, -0.0300, 0.0063, 0.0226, -0.0090, + 0.0394, -0.0107], device='cuda:0'), grad: tensor([ 2.9337e-08, 3.0920e-07, 1.1642e-07, 3.7253e-07, -4.9081e-07, + -1.1958e-06, 5.7369e-07, 1.0012e-07, -9.5461e-08, 2.9802e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 250.18, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4380 re_mapping 0.0047 re_causal 0.0112 /// teacc 99.08 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.1152, -0.2432, -0.0801, ..., -0.0604, 0.1858, 0.1840], + [-0.2272, -0.1972, -0.0855, ..., -0.1781, -0.2275, -0.1448], + [-0.0720, -0.1703, 0.1425, ..., -0.2151, 0.2491, 0.1027], + ..., + [-0.1642, 0.0887, 0.0352, ..., 0.2034, -0.2203, -0.2752], + [-0.2800, 0.0680, -0.1487, ..., 0.0573, -0.1082, -0.1935], + [-0.0265, -0.1364, -0.0905, ..., -0.1200, -0.0651, -0.2222]], + device='cuda:0'), grad: tensor([[ 6.9384e-08, 4.4703e-08, 0.0000e+00, ..., 3.4925e-09, + 1.5437e-07, 6.7987e-08], + [ 1.6298e-09, 1.2573e-08, 0.0000e+00, ..., 2.6077e-08, + 2.8871e-08, 2.3283e-10], + [ 1.7928e-08, 1.1642e-08, 0.0000e+00, ..., 6.9849e-10, + 6.9849e-08, 1.5367e-08], + ..., + [ 2.5611e-09, -2.5844e-08, 0.0000e+00, ..., -6.3796e-08, + 2.3982e-08, 4.6566e-10], + [ 4.4238e-09, 3.2596e-09, 0.0000e+00, ..., 2.3283e-10, + 2.7940e-09, 1.1642e-09], + [ 6.9849e-10, 1.3504e-08, 0.0000e+00, ..., 3.0501e-08, + 1.6554e-07, 2.3283e-10]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0083, -0.0385, -0.0147, -0.0213, -0.0299, 0.0060, 0.0227, -0.0089, + 0.0395, -0.0110], device='cuda:0'), grad: tensor([ 6.2119e-07, 2.2468e-07, 4.7521e-07, -5.7416e-07, -3.2056e-06, + 6.9616e-08, 9.4762e-08, 1.4831e-07, 2.7707e-08, 2.1309e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 250.42, cls_loss 0.0018 cls_loss_mapping 0.0019 cls_loss_causal 0.4951 re_mapping 0.0045 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.1152, -0.2435, -0.0801, ..., -0.0605, 0.1859, 0.1840], + [-0.2274, -0.1974, -0.0855, ..., -0.1785, -0.2276, -0.1447], + [-0.0721, -0.1705, 0.1425, ..., -0.2152, 0.2490, 0.1027], + ..., + [-0.1645, 0.0887, 0.0352, ..., 0.2034, -0.2204, -0.2764], + [-0.2804, 0.0685, -0.1486, ..., 0.0580, -0.1086, -0.1940], + [-0.0276, -0.1366, -0.0904, ..., -0.1201, -0.0651, -0.2222]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.6298e-09, 2.3283e-10, ..., 1.8626e-09, + -7.7672e-07, -5.6252e-07], + [ 4.1910e-09, 1.0477e-08, 1.1642e-09, ..., 1.1409e-08, + 1.9558e-08, 1.2107e-08], + [ 6.9849e-10, 2.5611e-09, 0.0000e+00, ..., 2.7940e-09, + 1.0012e-08, 6.9849e-09], + ..., + [ 4.6566e-10, -2.3516e-08, 3.4925e-09, ..., -2.5611e-08, + 1.0477e-08, 7.4506e-09], + [ 1.3970e-09, 1.8626e-09, 2.3283e-10, ..., 1.8626e-09, + 3.9116e-08, 2.9337e-08], + [ 4.6566e-10, 5.3551e-09, -9.3132e-09, ..., 5.8208e-09, + 3.3434e-07, 2.4191e-07]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0083, -0.0380, -0.0151, -0.0218, -0.0300, 0.0061, 0.0228, -0.0091, + 0.0395, -0.0110], device='cuda:0'), grad: tensor([-2.2426e-06, 9.5461e-08, 4.2841e-08, 4.4005e-08, -7.2177e-07, + 2.1793e-07, 1.5469e-06, -1.6531e-08, 1.2713e-07, 9.1270e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 250.51, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.5029 re_mapping 0.0048 re_causal 0.0123 /// teacc 99.11 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.1152, -0.2437, -0.0801, ..., -0.0606, 0.1859, 0.1841], + [-0.2274, -0.1976, -0.0855, ..., -0.1784, -0.2276, -0.1448], + [-0.0721, -0.1706, 0.1425, ..., -0.2152, 0.2492, 0.1033], + ..., + [-0.1646, 0.0890, 0.0352, ..., 0.2039, -0.2206, -0.2784], + [-0.2809, 0.0684, -0.1487, ..., 0.0581, -0.1091, -0.1950], + [-0.0283, -0.1370, -0.0903, ..., -0.1207, -0.0651, -0.2223]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 2.5379e-08, 0.0000e+00, ..., 6.4028e-08, + -6.3796e-08, -4.7032e-08], + [ 6.0536e-09, 9.0338e-08, 0.0000e+00, ..., 1.5250e-07, + 3.9581e-09, 2.3283e-09], + [ 1.6298e-09, 1.3015e-07, 0.0000e+00, ..., 1.1059e-07, + -3.8417e-08, -1.3504e-08], + ..., + [ 9.3132e-10, -1.2284e-06, 0.0000e+00, ..., -1.5274e-06, + 1.6298e-09, -1.5600e-08], + [ 1.4203e-08, 4.7730e-08, 0.0000e+00, ..., 4.0978e-08, + 3.0268e-09, 2.7940e-09], + [ 1.6764e-08, 1.6973e-07, 0.0000e+00, ..., 4.8103e-07, + 5.2620e-08, 3.8184e-08]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0083, -0.0380, -0.0151, -0.0220, -0.0298, 0.0061, 0.0229, -0.0092, + 0.0398, -0.0110], device='cuda:0'), grad: tensor([ 1.0873e-07, -3.0398e-06, 2.3795e-07, 3.9153e-06, -1.0524e-06, + -2.3190e-06, 8.1258e-07, -8.3493e-07, 1.6624e-07, 2.0079e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 251.16, cls_loss 0.0024 cls_loss_mapping 0.0026 cls_loss_causal 0.4824 re_mapping 0.0046 re_causal 0.0110 /// teacc 99.08 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.1153, -0.2441, -0.0804, ..., -0.0606, 0.1860, 0.1841], + [-0.2276, -0.1977, -0.0857, ..., -0.1789, -0.2277, -0.1453], + [-0.0721, -0.1708, 0.1426, ..., -0.2153, 0.2495, 0.1042], + ..., + [-0.1658, 0.0894, 0.0333, ..., 0.2051, -0.2209, -0.2794], + [-0.2812, 0.0683, -0.1490, ..., 0.0585, -0.1097, -0.1960], + [-0.0279, -0.1374, -0.0865, ..., -0.1220, -0.0651, -0.2223]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.7509e-08, -3.4925e-08], + [ 0.0000e+00, 3.0268e-09, 0.0000e+00, ..., 4.4238e-09, + 3.0268e-09, 1.6298e-09], + [ 0.0000e+00, 3.3993e-08, 0.0000e+00, ..., 4.9826e-08, + -3.0268e-08, -1.7928e-08], + ..., + [ 0.0000e+00, -1.8300e-07, 0.0000e+00, ..., -2.1793e-07, + 6.9849e-09, 3.4925e-09], + [ 1.3271e-08, 1.4203e-07, 0.0000e+00, ..., 1.5926e-07, + 2.5611e-08, 1.6531e-08], + [ 4.6566e-10, 2.3283e-09, 0.0000e+00, ..., 6.5193e-09, + 3.5856e-08, 2.0023e-08]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0083, -0.0381, -0.0152, -0.0215, -0.0315, 0.0050, 0.0230, -0.0089, + 0.0399, -0.0105], device='cuda:0'), grad: tensor([-1.1781e-07, -1.7928e-08, -9.0804e-09, 1.3039e-08, -6.4960e-08, + -5.6578e-08, 4.9593e-08, -3.0617e-07, 3.4692e-07, 1.6228e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 250.91, cls_loss 0.0018 cls_loss_mapping 0.0019 cls_loss_causal 0.4890 re_mapping 0.0046 re_causal 0.0118 /// teacc 99.12 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.1154, -0.2448, -0.0804, ..., -0.0607, 0.1860, 0.1842], + [-0.2294, -0.1977, -0.0857, ..., -0.1788, -0.2278, -0.1456], + [-0.0720, -0.1707, 0.1427, ..., -0.2154, 0.2495, 0.1044], + ..., + [-0.1676, 0.0893, 0.0332, ..., 0.2050, -0.2209, -0.2800], + [-0.2818, 0.0691, -0.1490, ..., 0.0595, -0.1097, -0.1970], + [-0.0288, -0.1377, -0.0863, ..., -0.1229, -0.0652, -0.2225]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + -2.2096e-07, -2.5402e-07], + [ 6.9849e-10, 9.3132e-10, 0.0000e+00, ..., 5.5879e-09, + -4.6566e-10, 1.3970e-09], + [ 0.0000e+00, -2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + -8.7311e-08, -6.0769e-08], + ..., + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.0710e-08, + 5.5879e-09, 4.4238e-09], + [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., -2.0722e-08, + 3.2596e-09, 2.7940e-09], + [ 9.3132e-10, 1.1642e-09, 0.0000e+00, ..., 2.3283e-09, + 1.9325e-08, 2.1886e-08]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0084, -0.0382, -0.0151, -0.0203, -0.0316, 0.0041, 0.0231, -0.0090, + 0.0406, -0.0106], device='cuda:0'), grad: tensor([-4.9360e-07, -1.1409e-08, -1.8999e-07, 2.6543e-08, 2.7474e-07, + 1.0920e-07, 3.0617e-07, 8.3121e-08, -4.7963e-08, -4.0047e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 250.64, cls_loss 0.0015 cls_loss_mapping 0.0016 cls_loss_causal 0.4640 re_mapping 0.0044 re_causal 0.0114 /// teacc 99.10 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.1154, -0.2460, -0.0805, ..., -0.0606, 0.1861, 0.1843], + [-0.2301, -0.1979, -0.0857, ..., -0.1786, -0.2281, -0.1458], + [-0.0720, -0.1704, 0.1427, ..., -0.2155, 0.2498, 0.1057], + ..., + [-0.1688, 0.0893, 0.0330, ..., 0.2050, -0.2213, -0.2814], + [-0.2824, 0.0690, -0.1491, ..., 0.0595, -0.1109, -0.1984], + [-0.0299, -0.1377, -0.0860, ..., -0.1230, -0.0654, -0.2227]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 2.3283e-10], + [-8.8476e-09, 1.6298e-09, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 7.2177e-09, -2.7940e-09, 0.0000e+00, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 5.3551e-09, 3.9581e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.0955e-09, 2.3283e-09, 0.0000e+00, ..., 2.0955e-09, + 4.6566e-10, 2.3283e-10]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0083, -0.0374, -0.0151, -0.0202, -0.0315, 0.0040, 0.0232, -0.0099, + 0.0404, -0.0106], device='cuda:0'), grad: tensor([ 1.6298e-09, -1.3411e-07, 3.7253e-09, -4.4238e-08, 7.6834e-09, + 5.7975e-08, 2.3283e-09, 1.2824e-06, 2.1420e-08, -1.1958e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 250.31, cls_loss 0.0020 cls_loss_mapping 0.0018 cls_loss_causal 0.4983 re_mapping 0.0043 re_causal 0.0112 /// teacc 99.11 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.1156, -0.2465, -0.0806, ..., -0.0606, 0.1860, 0.1844], + [-0.2302, -0.1983, -0.0858, ..., -0.1797, -0.2287, -0.1462], + [-0.0721, -0.1703, 0.1428, ..., -0.2157, 0.2501, 0.1068], + ..., + [-0.1702, 0.0894, 0.0328, ..., 0.2058, -0.2212, -0.2829], + [-0.2827, 0.0690, -0.1491, ..., 0.0595, -0.1105, -0.1993], + [-0.0311, -0.1380, -0.0855, ..., -0.1232, -0.0651, -0.2228]], + device='cuda:0'), grad: tensor([[ 5.3551e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 3.0035e-08, 4.4238e-09], + [ 5.1223e-09, 3.0268e-09, 0.0000e+00, ..., 2.7940e-09, + 3.2596e-08, 0.0000e+00], + [ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 8.6147e-09, + 1.7928e-08, 2.3283e-10], + ..., + [ 2.3283e-09, -8.3121e-08, 0.0000e+00, ..., -1.2922e-07, + 6.2864e-09, 0.0000e+00], + [ 2.7940e-09, 6.1002e-08, 0.0000e+00, ..., 9.0338e-08, + 6.9849e-10, 2.3283e-10], + [ 2.5611e-09, 5.3551e-09, 0.0000e+00, ..., 9.5461e-09, + 6.2864e-09, 6.9849e-10]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0086, -0.0379, -0.0150, -0.0194, -0.0318, 0.0036, 0.0232, -0.0097, + 0.0408, -0.0101], device='cuda:0'), grad: tensor([ 9.2201e-08, 1.2224e-07, 7.6368e-08, 1.1409e-08, -1.2070e-06, + 2.0256e-08, 7.4133e-07, -1.3621e-07, 1.3551e-07, 1.5902e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 250.37, cls_loss 0.0023 cls_loss_mapping 0.0022 cls_loss_causal 0.4856 re_mapping 0.0043 re_causal 0.0115 /// teacc 99.07 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.1151, -0.2467, -0.0807, ..., -0.0606, 0.1860, 0.1851], + [-0.2302, -0.1984, -0.0858, ..., -0.1801, -0.2298, -0.1462], + [-0.0722, -0.1706, 0.1428, ..., -0.2160, 0.2513, 0.1064], + ..., + [-0.1707, 0.0894, 0.0328, ..., 0.2057, -0.2222, -0.2836], + [-0.2830, 0.0694, -0.1492, ..., 0.0610, -0.1104, -0.1996], + [-0.0309, -0.1381, -0.0855, ..., -0.1237, -0.0643, -0.2229]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.0245e-08, 0.0000e+00, ..., 2.0722e-08, + 2.0955e-09, 0.0000e+00], + [ 0.0000e+00, 2.0955e-09, 0.0000e+00, ..., 4.1910e-09, + -2.5379e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -4.8662e-08, 0.0000e+00, ..., -9.5461e-08, + 6.7521e-09, 4.6566e-10], + [ 1.3970e-09, 2.0489e-08, 0.0000e+00, ..., 3.7253e-08, + 1.0012e-08, 0.0000e+00], + [ 4.6566e-10, 2.3283e-09, -0.0000e+00, ..., 3.9581e-09, + 9.3132e-10, 2.3283e-10]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0091, -0.0376, -0.0138, -0.0189, -0.0326, 0.0028, 0.0228, -0.0105, + 0.0426, -0.0089], device='cuda:0'), grad: tensor([ 5.3551e-09, 9.8255e-08, -6.0536e-08, 8.2422e-08, -2.0373e-07, + 2.2817e-08, 1.0943e-08, -1.2713e-07, 7.4506e-08, 1.0594e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 250.34, cls_loss 0.0020 cls_loss_mapping 0.0016 cls_loss_causal 0.4781 re_mapping 0.0048 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.1158, -0.2469, -0.0807, ..., -0.0613, 0.1859, 0.1851], + [-0.2304, -0.1985, -0.0858, ..., -0.1803, -0.2301, -0.1465], + [-0.0723, -0.1708, 0.1436, ..., -0.2162, 0.2519, 0.1069], + ..., + [-0.1712, 0.0895, 0.0328, ..., 0.2059, -0.2226, -0.2850], + [-0.2846, 0.0692, -0.1492, ..., 0.0613, -0.1105, -0.2001], + [-0.0332, -0.1382, -0.0855, ..., -0.1236, -0.0645, -0.2231]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 3.0268e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 1.1409e-08, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 1.6298e-09, 0.0000e+00, ..., 7.4506e-09, + -9.0804e-09, -4.6566e-09], + ..., + [ 2.3283e-10, -3.6322e-08, 0.0000e+00, ..., -8.4285e-08, + 4.8894e-09, 2.5611e-09], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10], + [ 4.4238e-09, 2.2817e-08, 0.0000e+00, ..., 5.1921e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0093, -0.0374, -0.0136, -0.0194, -0.0338, 0.0032, 0.0229, -0.0110, + 0.0428, -0.0082], device='cuda:0'), grad: tensor([ 8.6147e-09, -1.2526e-07, 1.3039e-08, 4.2375e-08, -6.8918e-08, + -2.2817e-08, 2.7940e-09, -4.0978e-08, 5.1223e-09, 1.9115e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 250.39, cls_loss 0.0020 cls_loss_mapping 0.0016 cls_loss_causal 0.4989 re_mapping 0.0043 re_causal 0.0120 /// teacc 99.03 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.1161, -0.2470, -0.0807, ..., -0.0613, 0.1857, 0.1850], + [-0.2306, -0.1986, -0.0859, ..., -0.1822, -0.2303, -0.1469], + [-0.0723, -0.1716, 0.1437, ..., -0.2162, 0.2522, 0.1080], + ..., + [-0.1714, 0.0897, 0.0328, ..., 0.2073, -0.2230, -0.2873], + [-0.2850, 0.0692, -0.1476, ..., 0.0605, -0.1109, -0.2009], + [-0.0326, -0.1383, -0.0857, ..., -0.1242, -0.0642, -0.2232]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + -3.0268e-09, -1.1642e-08], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 1.8161e-08, + -1.8557e-07, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + 1.4668e-08, 4.6566e-10], + ..., + [ 2.3283e-10, -3.0268e-09, 0.0000e+00, ..., 8.6147e-09, + 2.7707e-08, 2.3283e-10], + [ 9.7789e-09, 1.1642e-09, 0.0000e+00, ..., -3.7486e-08, + 8.1491e-08, 6.9849e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.0804e-09, + 2.5379e-08, 1.2806e-08]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0098, -0.0373, -0.0135, -0.0174, -0.0339, 0.0017, 0.0229, -0.0109, + 0.0421, -0.0079], device='cuda:0'), grad: tensor([ 1.5670e-07, -2.4159e-06, 2.7148e-07, 4.5868e-08, 3.0105e-07, + 6.4261e-08, 3.9721e-07, 4.1653e-07, 8.2841e-07, -5.6578e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 250.52, cls_loss 0.0021 cls_loss_mapping 0.0028 cls_loss_causal 0.4984 re_mapping 0.0043 re_causal 0.0113 /// teacc 99.10 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.1162, -0.2470, -0.0807, ..., -0.0614, 0.1857, 0.1851], + [-0.2307, -0.1986, -0.0859, ..., -0.1828, -0.2303, -0.1469], + [-0.0723, -0.1721, 0.1436, ..., -0.2163, 0.2524, 0.1084], + ..., + [-0.1721, 0.0895, 0.0327, ..., 0.2078, -0.2232, -0.2881], + [-0.2852, 0.0691, -0.1477, ..., 0.0608, -0.1096, -0.2023], + [-0.0326, -0.1384, -0.0857, ..., -0.1245, -0.0641, -0.2233]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -1.4435e-08, -8.6147e-09], + [ 0.0000e+00, 8.0327e-08, 0.0000e+00, ..., 1.5460e-07, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 4.6566e-09, + 4.6566e-10, 2.3283e-10], + ..., + [ 0.0000e+00, -1.1665e-07, 0.0000e+00, ..., -2.2771e-07, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 3.1898e-08, -2.3283e-10, ..., 6.4494e-08, + 8.3819e-09, 5.1223e-09]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0100, -0.0377, -0.0137, -0.0173, -0.0359, 0.0015, 0.0226, -0.0105, + 0.0426, -0.0068], device='cuda:0'), grad: tensor([-9.0804e-09, 7.4878e-07, 2.1188e-08, 5.1223e-09, -9.3691e-07, + 1.7695e-08, 3.6531e-07, -5.9418e-07, 1.7928e-08, 3.7695e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 250.66, cls_loss 0.0020 cls_loss_mapping 0.0019 cls_loss_causal 0.4778 re_mapping 0.0046 re_causal 0.0115 /// teacc 99.06 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.1163, -0.2472, -0.0808, ..., -0.0613, 0.1858, 0.1852], + [-0.2308, -0.1987, -0.0862, ..., -0.1830, -0.2305, -0.1461], + [-0.0723, -0.1730, 0.1464, ..., -0.2162, 0.2533, 0.1096], + ..., + [-0.1725, 0.0896, 0.0327, ..., 0.2081, -0.2240, -0.2908], + [-0.2856, 0.0690, -0.1477, ..., 0.0607, -0.1099, -0.2026], + [-0.0329, -0.1381, -0.0856, ..., -0.1247, -0.0642, -0.2235]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.0955e-09, 0.0000e+00, ..., 2.5611e-09, + -4.0233e-06, -3.5595e-06], + [ 2.3283e-10, 6.7521e-09, 0.0000e+00, ..., 5.3551e-09, + 5.9372e-08, 4.9360e-08], + [ 0.0000e+00, 6.4727e-08, 0.0000e+00, ..., 7.5903e-08, + 2.3982e-08, 3.1432e-08], + ..., + [ 4.6566e-10, -5.6811e-08, 0.0000e+00, ..., -3.4226e-08, + 3.4692e-08, 2.2817e-08], + [ 4.6566e-10, -4.4471e-08, 0.0000e+00, ..., -1.5344e-07, + 1.2526e-07, 1.1479e-07], + [ 0.0000e+00, 2.1886e-08, 0.0000e+00, ..., 9.8487e-08, + 1.2759e-06, 1.1278e-06]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0099, -0.0368, -0.0138, -0.0173, -0.0353, 0.0016, 0.0225, -0.0117, + 0.0426, -0.0064], device='cuda:0'), grad: tensor([-1.0304e-05, 1.4459e-07, 2.8731e-07, 1.8184e-07, 3.0268e-08, + 3.3015e-07, 5.9418e-06, 2.3004e-07, -6.9849e-07, 3.8594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 250.55, cls_loss 0.0020 cls_loss_mapping 0.0019 cls_loss_causal 0.4872 re_mapping 0.0044 re_causal 0.0114 /// teacc 99.02 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.1163, -0.2474, -0.0808, ..., -0.0611, 0.1861, 0.1857], + [-0.2311, -0.2000, -0.0862, ..., -0.1845, -0.2310, -0.1476], + [-0.0723, -0.1731, 0.1466, ..., -0.2158, 0.2541, 0.1111], + ..., + [-0.1728, 0.0907, 0.0327, ..., 0.2089, -0.2250, -0.2933], + [-0.2869, 0.0688, -0.1478, ..., 0.0605, -0.1108, -0.2054], + [-0.0331, -0.1385, -0.0856, ..., -0.1253, -0.0644, -0.2238]], + device='cuda:0'), grad: tensor([[ 1.1874e-08, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-08, -7.4506e-09], + [ 4.6566e-09, 4.1910e-09, 0.0000e+00, ..., 2.3283e-10, + -1.6089e-07, 7.9162e-09], + [ 2.4191e-07, 1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + 1.7998e-07, -1.1642e-08], + ..., + [ 6.6590e-08, 6.3097e-08, 0.0000e+00, ..., 4.6566e-10, + 6.0536e-09, 1.6298e-08], + [ 2.0256e-07, 4.6566e-10, 0.0000e+00, ..., 4.8429e-08, + 2.6077e-08, 4.6566e-10], + [-3.3039e-07, 6.9849e-10, 0.0000e+00, ..., 1.8161e-08, + -8.2189e-08, 6.0536e-09]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0097, -0.0375, -0.0135, -0.0174, -0.0353, 0.0014, 0.0227, -0.0108, + 0.0423, -0.0070], device='cuda:0'), grad: tensor([-4.6566e-09, -1.0999e-06, 1.6754e-06, -2.3982e-07, -8.1817e-07, + -2.0699e-07, 2.9337e-07, 3.0128e-07, 4.7497e-07, -3.8208e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 250.12, cls_loss 0.0020 cls_loss_mapping 0.0020 cls_loss_causal 0.5071 re_mapping 0.0043 re_causal 0.0119 /// teacc 99.00 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.1163, -0.2481, -0.0808, ..., -0.0612, 0.1857, 0.1858], + [-0.2311, -0.2001, -0.0862, ..., -0.1847, -0.2310, -0.1469], + [-0.0724, -0.1730, 0.1468, ..., -0.2161, 0.2544, 0.1115], + ..., + [-0.1734, 0.0896, 0.0327, ..., 0.2070, -0.2253, -0.2943], + [-0.2873, 0.0688, -0.1478, ..., 0.0606, -0.1111, -0.2053], + [-0.0335, -0.1363, -0.0855, ..., -0.1226, -0.0637, -0.2239]], + device='cuda:0'), grad: tensor([[ 1.4668e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, -9.7789e-09], + [ 6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 2.3283e-10, -4.6566e-10], + [ 1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 6.9849e-10, 6.9849e-10], + ..., + [ 6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 6.9849e-10], + [ 6.5193e-09, 6.9849e-10, 0.0000e+00, ..., -8.1491e-09, + 1.3970e-09, 1.1642e-09], + [-2.9802e-08, 2.3283e-10, -0.0000e+00, ..., -4.6566e-10, + 2.1188e-08, 1.7695e-08]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0103, -0.0372, -0.0138, -0.0173, -0.0340, 0.0013, 0.0227, -0.0133, + 0.0423, -0.0042], device='cuda:0'), grad: tensor([ 2.2352e-08, 1.6531e-08, 1.0594e-07, 3.0105e-07, 4.0280e-08, + 4.9593e-08, -2.9104e-08, 4.5402e-08, -1.1642e-08, -5.3458e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 250.05, cls_loss 0.0019 cls_loss_mapping 0.0019 cls_loss_causal 0.4770 re_mapping 0.0044 re_causal 0.0113 /// teacc 98.96 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.1164, -0.2485, -0.0808, ..., -0.0612, 0.1858, 0.1858], + [-0.2312, -0.2001, -0.0862, ..., -0.1849, -0.2311, -0.1460], + [-0.0728, -0.1737, 0.1468, ..., -0.2164, 0.2543, 0.1111], + ..., + [-0.1739, 0.0891, 0.0326, ..., 0.2048, -0.2253, -0.2947], + [-0.2879, 0.0718, -0.1479, ..., 0.0646, -0.1115, -0.2064], + [-0.0339, -0.1369, -0.0853, ..., -0.1229, -0.0638, -0.2240]], + device='cuda:0'), grad: tensor([[ 5.8208e-09, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + -3.4925e-09, -3.0268e-09], + [ 1.1874e-08, 1.0012e-08, 0.0000e+00, ..., 1.1642e-09, + 3.0268e-09, 1.1642e-09], + [ 1.8626e-09, 5.3551e-09, 0.0000e+00, ..., 2.5611e-09, + -1.0012e-08, -4.8894e-09], + ..., + [ 1.8394e-08, -1.1642e-09, 0.0000e+00, ..., -1.1874e-08, + 6.2864e-09, 3.0268e-09], + [ 5.6112e-08, 3.6322e-08, 0.0000e+00, ..., 1.2107e-08, + 2.3283e-10, 2.3283e-10], + [ 3.3760e-08, 2.8405e-08, 0.0000e+00, ..., 2.5611e-09, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0103, -0.0368, -0.0141, -0.0171, -0.0339, 0.0011, 0.0229, -0.0138, + 0.0446, -0.0044], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.1273e-05, -1.4668e-08, -1.7476e-04, -1.1563e-05, + 1.7452e-04, 1.5600e-08, 4.6799e-08, 2.3213e-07, 2.3632e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 250.44, cls_loss 0.0025 cls_loss_mapping 0.0022 cls_loss_causal 0.4825 re_mapping 0.0041 re_causal 0.0107 /// teacc 99.04 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.1168, -0.2487, -0.0809, ..., -0.0614, 0.1863, 0.1867], + [-0.2323, -0.2003, -0.0862, ..., -0.1831, -0.2312, -0.1468], + [-0.0717, -0.1743, 0.1468, ..., -0.2167, 0.2547, 0.1119], + ..., + [-0.1748, 0.0891, 0.0326, ..., 0.2034, -0.2258, -0.2951], + [-0.2925, 0.0741, -0.1482, ..., 0.0641, -0.1121, -0.2079], + [-0.0353, -0.1379, -0.0852, ..., -0.1235, -0.0639, -0.2245]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -3.2783e-07, -2.1420e-07], + [ 0.0000e+00, 2.1094e-07, 0.0000e+00, ..., 4.3400e-07, + 2.6543e-08, 1.8161e-08], + [ 0.0000e+00, 3.7253e-08, 0.0000e+00, ..., 7.6368e-08, + 9.7789e-09, 6.0536e-09], + ..., + [ 0.0000e+00, -3.0315e-07, 0.0000e+00, ..., -6.1560e-07, + 1.1176e-08, 6.5193e-09], + [ 3.2596e-09, 1.2107e-08, 0.0000e+00, ..., 2.4680e-08, + 8.3819e-09, 5.5879e-09], + [ 0.0000e+00, 1.2573e-08, -4.6566e-10, ..., 1.7229e-08, + 1.1735e-07, 8.0094e-08]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0102, -0.0362, -0.0141, -0.0168, -0.0327, 0.0021, 0.0225, -0.0143, + 0.0444, -0.0047], device='cuda:0'), grad: tensor([-6.9803e-07, 1.2238e-06, 2.2445e-07, 1.6065e-07, 4.5169e-08, + -1.5832e-08, 3.2084e-07, -1.6149e-06, 8.7079e-08, 2.7614e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 250.40, cls_loss 0.0016 cls_loss_mapping 0.0011 cls_loss_causal 0.4987 re_mapping 0.0044 re_causal 0.0119 /// teacc 98.99 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.1170, -0.2488, -0.0810, ..., -0.0616, 0.1863, 0.1868], + [-0.2329, -0.2004, -0.0862, ..., -0.1835, -0.2315, -0.1476], + [-0.0714, -0.1728, 0.1468, ..., -0.2166, 0.2550, 0.1127], + ..., + [-0.1755, 0.0892, 0.0326, ..., 0.2036, -0.2262, -0.2961], + [-0.2927, 0.0746, -0.1484, ..., 0.0644, -0.1110, -0.2077], + [-0.0353, -0.1386, -0.0851, ..., -0.1237, -0.0640, -0.2246]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.4110e-07, + -1.7695e-08, 6.9849e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.4435e-08, + 1.3970e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8636e-06, + -5.1223e-09, 2.8824e-07], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + 4.1910e-09, 5.5879e-09], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -5.9940e-06, + -9.3132e-10, -9.3831e-07], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + 8.8476e-09, 9.7789e-09]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0102, -0.0362, -0.0141, -0.0169, -0.0324, 0.0014, 0.0228, -0.0144, + 0.0456, -0.0049], device='cuda:0'), grad: tensor([ 4.2235e-07, -2.3283e-08, 5.8860e-06, 3.9116e-08, 9.7789e-09, + 1.0595e-05, 1.8859e-06, 1.4016e-07, -1.8999e-05, 4.5169e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 250.35, cls_loss 0.0019 cls_loss_mapping 0.0020 cls_loss_causal 0.4904 re_mapping 0.0044 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.1171, -0.2504, -0.0831, ..., -0.0617, 0.1864, 0.1868], + [-0.2333, -0.2006, -0.0862, ..., -0.1840, -0.2317, -0.1479], + [-0.0714, -0.1730, 0.1468, ..., -0.2171, 0.2555, 0.1132], + ..., + [-0.1757, 0.0894, 0.0326, ..., 0.2040, -0.2263, -0.2962], + [-0.2929, 0.0745, -0.1485, ..., 0.0646, -0.1111, -0.2079], + [-0.0358, -0.1388, -0.0850, ..., -0.1236, -0.0640, -0.2247]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + -2.6263e-07, -1.4761e-07], + [ 9.5461e-08, 5.5879e-09, 0.0000e+00, ..., -3.0873e-07, + 4.5635e-08, 2.5611e-08], + [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 5.5879e-09, 4.6566e-09], + ..., + [ 5.1223e-09, -2.4820e-07, 0.0000e+00, ..., -1.3411e-07, + 4.1910e-09, 2.3283e-09], + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 1.4901e-08, 8.3819e-09], + [-1.6298e-08, 2.3749e-07, 0.0000e+00, ..., 4.3446e-07, + 1.8114e-07, 1.0012e-07]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0102, -0.0347, -0.0140, -0.0169, -0.0330, 0.0011, 0.0229, -0.0156, + 0.0458, -0.0046], device='cuda:0'), grad: tensor([-5.5321e-07, -2.5649e-06, 3.0268e-08, 6.5193e-09, -9.2480e-07, + 9.8255e-08, 1.1921e-07, 1.7481e-06, 4.2841e-08, 2.0154e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 250.19, cls_loss 0.0023 cls_loss_mapping 0.0023 cls_loss_causal 0.4827 re_mapping 0.0041 re_causal 0.0106 /// teacc 99.03 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.1176, -0.2513, -0.0834, ..., -0.0619, 0.1865, 0.1870], + [-0.2333, -0.2008, -0.0861, ..., -0.1844, -0.2317, -0.1480], + [-0.0717, -0.1733, 0.1470, ..., -0.2191, 0.2557, 0.1135], + ..., + [-0.1758, 0.0896, 0.0325, ..., 0.2044, -0.2267, -0.2968], + [-0.2934, 0.0748, -0.1486, ..., 0.0657, -0.1111, -0.2081], + [-0.0377, -0.1397, -0.0852, ..., -0.1243, -0.0643, -0.2251]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -8.6613e-08, -6.4261e-08], + [ 6.9849e-09, 1.3970e-09, 0.0000e+00, ..., 4.6566e-09, + 8.3819e-09, 3.7253e-09], + [ 2.3283e-09, 3.2596e-09, 0.0000e+00, ..., 1.3970e-09, + -6.7055e-08, 4.6566e-09], + ..., + [ 8.8476e-09, 7.9162e-09, 0.0000e+00, ..., 5.5879e-09, + 6.1467e-08, 4.6566e-10], + [ 5.5209e-06, 5.7276e-08, 0.0000e+00, ..., 3.4384e-06, + 2.2817e-08, 1.9092e-08], + [ 3.7719e-08, 4.6566e-10, 0.0000e+00, ..., 2.3749e-08, + 1.3504e-08, 1.0710e-08]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0102, -0.0346, -0.0145, -0.0172, -0.0328, 0.0015, 0.0228, -0.0155, + 0.0472, -0.0050], device='cuda:0'), grad: tensor([-8.9407e-08, 5.2363e-05, -1.3923e-07, 2.7522e-05, 3.0454e-07, + -3.8564e-05, 9.4529e-08, 1.0945e-05, 1.1168e-05, -6.3658e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 250.39, cls_loss 0.0019 cls_loss_mapping 0.0015 cls_loss_causal 0.4700 re_mapping 0.0044 re_causal 0.0114 /// teacc 99.07 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.1176, -0.2516, -0.0847, ..., -0.0620, 0.1866, 0.1871], + [-0.2338, -0.2008, -0.0849, ..., -0.1853, -0.2321, -0.1485], + [-0.0717, -0.1753, 0.1475, ..., -0.2205, 0.2565, 0.1141], + ..., + [-0.1754, 0.0900, 0.0316, ..., 0.2054, -0.2269, -0.2972], + [-0.2939, 0.0748, -0.1495, ..., 0.0657, -0.1114, -0.2083], + [-0.0391, -0.1399, -0.0862, ..., -0.1248, -0.0643, -0.2252]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + -2.7008e-08, -1.5367e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 7.6219e-06, + 5.5879e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 5.1223e-09, 1.3970e-09], + ..., + [ 0.0000e+00, -4.1910e-09, 0.0000e+00, ..., -8.0615e-06, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.9360e-08, + 1.3970e-09, 9.3132e-10], + [ 0.0000e+00, 2.7940e-09, -4.6566e-10, ..., 3.5577e-07, + 2.2817e-08, 1.2107e-08]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0102, -0.0348, -0.0145, -0.0174, -0.0321, 0.0025, 0.0224, -0.0153, + 0.0472, -0.0052], device='cuda:0'), grad: tensor([-2.8405e-08, 2.4199e-05, 1.0896e-07, 2.3283e-08, 3.1665e-08, + -5.1223e-09, 1.1642e-08, -2.6479e-05, 8.3726e-07, 1.3541e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 250.26, cls_loss 0.0022 cls_loss_mapping 0.0020 cls_loss_causal 0.5031 re_mapping 0.0043 re_causal 0.0115 /// teacc 99.10 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.1176, -0.2518, -0.0843, ..., -0.0622, 0.1868, 0.1874], + [-0.2350, -0.2010, -0.0834, ..., -0.1859, -0.2332, -0.1491], + [-0.0717, -0.1755, 0.1476, ..., -0.2215, 0.2567, 0.1142], + ..., + [-0.1762, 0.0912, 0.0307, ..., 0.2070, -0.2271, -0.2974], + [-0.2944, 0.0745, -0.1511, ..., 0.0656, -0.1107, -0.2098], + [-0.0396, -0.1421, -0.0877, ..., -0.1264, -0.0645, -0.2254]], + device='cuda:0'), grad: tensor([[-1.2061e-07, 4.6566e-10, 0.0000e+00, ..., 3.6787e-08, + -4.2794e-07, -4.8196e-07], + [ 1.3970e-09, 8.8476e-09, 0.0000e+00, ..., 1.3970e-08, + -1.8626e-09, 9.3132e-10], + [ 4.6566e-10, 6.6590e-08, 0.0000e+00, ..., 1.0757e-07, + -8.3819e-09, -9.3132e-09], + ..., + [ 1.2713e-07, -3.7579e-07, 4.6566e-10, ..., -5.7416e-07, + 7.4506e-09, 6.9849e-09], + [ 6.2399e-08, 2.9523e-07, 4.6566e-10, ..., 4.9965e-07, + 9.3132e-10, 4.6566e-10], + [ 1.1176e-08, 1.0245e-08, -4.6566e-10, ..., 1.7695e-08, + 1.0710e-08, 1.0710e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0101, -0.0328, -0.0149, -0.0178, -0.0316, 0.0035, 0.0219, -0.0165, + 0.0473, -0.0060], device='cuda:0'), grad: tensor([-7.0501e-07, 2.5146e-08, 1.4622e-07, 5.1223e-09, 2.8452e-07, + -1.2144e-06, 1.6009e-06, -7.1246e-07, 8.5682e-07, -2.7986e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 250.21, cls_loss 0.0022 cls_loss_mapping 0.0015 cls_loss_causal 0.4906 re_mapping 0.0042 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.1183, -0.2522, -0.0869, ..., -0.0624, 0.1865, 0.1871], + [-0.2353, -0.2014, -0.0835, ..., -0.1870, -0.2341, -0.1497], + [-0.0719, -0.1759, 0.1476, ..., -0.2218, 0.2572, 0.1147], + ..., + [-0.1766, 0.0921, 0.0302, ..., 0.2079, -0.2272, -0.2981], + [-0.2950, 0.0743, -0.1521, ..., 0.0658, -0.1101, -0.2107], + [-0.0400, -0.1431, -0.0873, ..., -0.1268, -0.0645, -0.2254]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 4.6566e-10, 0.0000e+00, ..., 6.7055e-08, + -9.7789e-09, -8.8476e-09], + [ 2.1420e-08, 5.8208e-08, 0.0000e+00, ..., 4.4517e-07, + 1.8626e-09, 1.3970e-09], + [ 4.6566e-10, 1.1176e-08, 0.0000e+00, ..., 6.6543e-07, + 1.8626e-09, 9.3132e-10], + ..., + [ 1.3970e-09, -7.4971e-08, 0.0000e+00, ..., -3.0175e-07, + 0.0000e+00, 0.0000e+00], + [ 1.1548e-07, 0.0000e+00, 0.0000e+00, ..., -8.0792e-07, + 3.2596e-09, 2.3283e-09], + [ 5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 1.3504e-08, + 3.6787e-08, 2.5611e-08]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0104, -0.0331, -0.0146, -0.0154, -0.0315, 0.0015, 0.0222, -0.0156, + 0.0473, -0.0069], device='cuda:0'), grad: tensor([ 4.1071e-07, 1.6429e-06, 4.6603e-06, 5.1782e-07, 4.8429e-08, + -1.8291e-06, 1.3085e-06, -4.9081e-07, -6.4038e-06, 1.4110e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 250.23, cls_loss 0.0019 cls_loss_mapping 0.0017 cls_loss_causal 0.4962 re_mapping 0.0043 re_causal 0.0112 /// teacc 99.00 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.1192, -0.2528, -0.0872, ..., -0.0621, 0.1859, 0.1868], + [-0.2357, -0.2020, -0.0836, ..., -0.1877, -0.2343, -0.1498], + [-0.0721, -0.1771, 0.1480, ..., -0.2226, 0.2577, 0.1155], + ..., + [-0.1760, 0.0932, 0.0298, ..., 0.2088, -0.2276, -0.2985], + [-0.2954, 0.0742, -0.1537, ..., 0.0658, -0.1113, -0.2130], + [-0.0402, -0.1442, -0.0862, ..., -0.1272, -0.0641, -0.2257]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 4.6566e-10, 4.6566e-10, ..., 1.3970e-09, + 2.1420e-08, -0.0000e+00], + [ 5.8394e-07, 1.8161e-08, 9.3132e-10, ..., 5.4482e-08, + 8.8802e-07, 1.7136e-07], + [ 2.7940e-09, -2.5146e-08, -1.3970e-08, ..., 2.9802e-08, + -2.5611e-08, 9.3132e-10], + ..., + [ 0.0000e+00, -3.4971e-07, 9.3132e-10, ..., -1.0906e-06, + 9.3132e-10, 0.0000e+00], + [ 1.1781e-07, 2.2678e-07, 4.1910e-09, ..., 6.9011e-07, + 1.8533e-07, 3.4925e-08], + [ 0.0000e+00, 9.5461e-08, -4.6566e-10, ..., 2.9895e-07, + 7.4506e-09, 6.5193e-09]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0112, -0.0327, -0.0148, -0.0153, -0.0334, 0.0014, 0.0223, -0.0152, + 0.0463, -0.0066], device='cuda:0'), grad: tensor([ 1.2852e-07, 7.9162e-09, 1.8738e-06, 1.5413e-07, 1.6810e-07, + 4.5933e-06, -9.9167e-06, -1.7183e-07, 2.4866e-06, 6.7521e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 250.31, cls_loss 0.0020 cls_loss_mapping 0.0016 cls_loss_causal 0.4430 re_mapping 0.0044 re_causal 0.0110 /// teacc 99.03 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.1196, -0.2521, -0.0873, ..., -0.0620, 0.1861, 0.1870], + [-0.2359, -0.2029, -0.0836, ..., -0.1886, -0.2347, -0.1500], + [-0.0722, -0.1772, 0.1486, ..., -0.2221, 0.2586, 0.1166], + ..., + [-0.1763, 0.0939, 0.0297, ..., 0.2092, -0.2279, -0.2987], + [-0.2959, 0.0743, -0.1551, ..., 0.0657, -0.1129, -0.2159], + [-0.0406, -0.1444, -0.0860, ..., -0.1274, -0.0646, -0.2265]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.0508e-06, 1.8170e-06], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 2.3283e-09, + 2.1420e-08, 1.8161e-08], + [ 0.0000e+00, 1.3970e-09, 9.3132e-10, ..., 1.8626e-09, + -4.8280e-06, -4.2506e-06], + ..., + [ 4.6566e-10, -6.9849e-09, 1.8626e-09, ..., -8.8476e-09, + 4.2375e-07, 3.7299e-07], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 1.0664e-06, 9.3784e-07], + [ 0.0000e+00, 3.7253e-09, -2.7940e-09, ..., 4.6566e-09, + 1.1548e-07, 9.2667e-08]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0111, -0.0329, -0.0123, -0.0155, -0.0316, 0.0023, 0.0220, -0.0156, + 0.0450, -0.0071], device='cuda:0'), grad: tensor([ 4.8466e-06, 1.0245e-08, -1.1273e-05, 2.5742e-06, 4.6678e-06, + -5.4017e-08, 2.0489e-07, 1.1362e-06, 4.4927e-06, -6.6087e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 250.23, cls_loss 0.0023 cls_loss_mapping 0.0019 cls_loss_causal 0.4618 re_mapping 0.0043 re_causal 0.0107 /// teacc 99.07 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.1200, -0.2524, -0.0875, ..., -0.0620, 0.1863, 0.1872], + [-0.2360, -0.2030, -0.0835, ..., -0.1888, -0.2356, -0.1504], + [-0.0723, -0.1776, 0.1485, ..., -0.2225, 0.2596, 0.1176], + ..., + [-0.1766, 0.0945, 0.0270, ..., 0.2097, -0.2281, -0.2990], + [-0.2961, 0.0743, -0.1556, ..., 0.0661, -0.1127, -0.2172], + [-0.0411, -0.1452, -0.0837, ..., -0.1280, -0.0650, -0.2271]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.0245e-08, + -4.0932e-07, -4.4098e-07], + [ 8.3819e-09, 2.7474e-08, 0.0000e+00, ..., 1.0151e-07, + 1.1176e-08, 1.3970e-09], + [ 3.4925e-08, 5.5414e-08, 0.0000e+00, ..., 2.2212e-07, + 1.6298e-08, 0.0000e+00], + ..., + [ 1.0384e-07, 1.1967e-07, 0.0000e+00, ..., -3.2643e-07, + 4.1910e-09, 4.6566e-10], + [ 1.1642e-08, 9.3132e-10, 0.0000e+00, ..., -2.4606e-06, + -8.0466e-06, 1.3970e-09], + [ 2.8405e-08, 4.6566e-08, 0.0000e+00, ..., 9.3132e-09, + 2.3749e-08, 1.3504e-08]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0112, -0.0330, -0.0119, -0.0155, -0.0305, 0.0021, 0.0218, -0.0158, + 0.0453, -0.0072], device='cuda:0'), grad: tensor([-6.3702e-07, 3.1758e-07, 6.7474e-07, -8.9174e-07, 9.3132e-10, + 2.3674e-06, 3.6895e-05, -2.9057e-07, -3.8743e-05, 2.1234e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 250.51, cls_loss 0.0019 cls_loss_mapping 0.0015 cls_loss_causal 0.4779 re_mapping 0.0044 re_causal 0.0112 /// teacc 99.09 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.1205, -0.2528, -0.0878, ..., -0.0620, 0.1863, 0.1872], + [-0.2361, -0.2031, -0.0825, ..., -0.1892, -0.2356, -0.1502], + [-0.0724, -0.1777, 0.1484, ..., -0.2227, 0.2597, 0.1176], + ..., + [-0.1795, 0.0931, 0.0257, ..., 0.2091, -0.2284, -0.2997], + [-0.2963, 0.0742, -0.1558, ..., 0.0663, -0.1124, -0.2177], + [-0.0411, -0.1453, -0.0835, ..., -0.1281, -0.0651, -0.2273]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, 4.7032e-08, 0.0000e+00, ..., 1.1548e-07, + -1.1642e-08, -8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + 6.5193e-09, 1.8626e-09], + [ 0.0000e+00, 4.6566e-10, -1.6298e-08, ..., 9.3132e-10, + -2.5611e-08, -7.9162e-09], + ..., + [-9.7789e-09, -5.6345e-08, 1.3970e-09, ..., -1.3877e-07, + 2.7940e-09, 9.3132e-10], + [ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 4.1910e-09, 3.2596e-09], + [ 1.3970e-09, 4.6566e-09, 0.0000e+00, ..., 1.1642e-08, + 6.0536e-09, 4.6566e-09]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0112, -0.0330, -0.0121, -0.0149, -0.0307, 0.0022, 0.0218, -0.0168, + 0.0459, -0.0061], device='cuda:0'), grad: tensor([ 2.2212e-07, -1.3970e-09, -5.1223e-08, 3.7253e-09, -1.7229e-07, + -3.1199e-08, 7.9628e-08, -2.7847e-07, 2.9802e-08, 2.0303e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 250.46, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.4780 re_mapping 0.0044 re_causal 0.0116 /// teacc 99.01 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.1205, -0.2531, -0.0875, ..., -0.0621, 0.1863, 0.1877], + [-0.2361, -0.2031, -0.0825, ..., -0.1889, -0.2361, -0.1507], + [-0.0724, -0.1781, 0.1484, ..., -0.2227, 0.2599, 0.1177], + ..., + [-0.1800, 0.0931, 0.0255, ..., 0.2091, -0.2285, -0.3000], + [-0.2968, 0.0740, -0.1571, ..., 0.0663, -0.1128, -0.2185], + [-0.0408, -0.1454, -0.0833, ..., -0.1283, -0.0646, -0.2274]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.8405e-08, 4.6566e-10, ..., 6.8452e-08, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.1222e-07, -1.7695e-08, ..., -2.2817e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.9558e-08, 0.0000e+00, ..., 4.7497e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.3330e-08, 1.7229e-08, ..., 1.0943e-07, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0114, -0.0329, -0.0120, -0.0150, -0.0309, 0.0022, 0.0216, -0.0170, + 0.0465, -0.0058], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.2619e-07, 9.3132e-10, 9.3132e-10, -6.0536e-08, + 4.6566e-10, 1.3970e-09, -4.4517e-07, 8.6147e-08, 2.8731e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 250.83, cls_loss 0.0020 cls_loss_mapping 0.0022 cls_loss_causal 0.5074 re_mapping 0.0046 re_causal 0.0120 /// teacc 99.05 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.1218, -0.2535, -0.0876, ..., -0.0622, 0.1852, 0.1860], + [-0.2370, -0.2032, -0.0826, ..., -0.1895, -0.2362, -0.1509], + [-0.0726, -0.1790, 0.1487, ..., -0.2234, 0.2600, 0.1176], + ..., + [-0.1810, 0.0921, 0.0251, ..., 0.2076, -0.2286, -0.3001], + [-0.2970, 0.0763, -0.1572, ..., 0.0691, -0.1130, -0.2189], + [-0.0413, -0.1459, -0.0832, ..., -0.1292, -0.0646, -0.2275]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + -2.1886e-08, -2.4680e-08], + [ 1.3970e-09, 3.0268e-08, 0.0000e+00, ..., 6.4261e-08, + 7.9162e-09, 0.0000e+00], + [ 4.6566e-10, 1.1176e-08, 0.0000e+00, ..., -1.8300e-07, + -2.0163e-07, 0.0000e+00], + ..., + [ 3.2596e-09, -6.8452e-08, 0.0000e+00, ..., -1.4063e-07, + -2.3283e-09, 0.0000e+00], + [ 6.9849e-09, 7.4506e-09, 0.0000e+00, ..., 1.6764e-08, + 6.5193e-09, 0.0000e+00], + [ 1.0245e-08, 3.4459e-08, 0.0000e+00, ..., 5.0291e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0128, -0.0329, -0.0122, -0.0153, -0.0310, 0.0028, 0.0233, -0.0178, + 0.0489, -0.0055], device='cuda:0'), grad: tensor([-3.8184e-08, 1.3178e-07, -5.4343e-07, -1.6391e-07, 9.3132e-10, + 6.2445e-07, 7.5903e-08, -2.6030e-07, 4.1910e-08, 1.4435e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 250.89, cls_loss 0.0017 cls_loss_mapping 0.0018 cls_loss_causal 0.5001 re_mapping 0.0043 re_causal 0.0117 /// teacc 99.02 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.1220, -0.2538, -0.0877, ..., -0.0623, 0.1830, 0.1861], + [-0.2371, -0.2039, -0.0826, ..., -0.1896, -0.2367, -0.1511], + [-0.0726, -0.1797, 0.1493, ..., -0.2237, 0.2602, 0.1177], + ..., + [-0.1815, 0.0928, 0.0251, ..., 0.2080, -0.2285, -0.3002], + [-0.2974, 0.0761, -0.1577, ..., 0.0689, -0.1132, -0.2202], + [-0.0418, -0.1461, -0.0831, ..., -0.1294, -0.0617, -0.2276]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + -4.6100e-08, -3.9581e-08], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 2.8461e-06, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -2.9299e-06, 9.3132e-10], + ..., + [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 3.8184e-08, 0.0000e+00], + [ 2.3283e-08, 1.3970e-09, 0.0000e+00, ..., 1.2107e-08, + 2.3283e-09, 4.6566e-10], + [ 1.0710e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 2.0955e-08, 1.3970e-08]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0149, -0.0332, -0.0122, -0.0150, -0.0313, 0.0022, 0.0237, -0.0178, + 0.0486, -0.0036], device='cuda:0'), grad: tensor([-6.3330e-08, 8.4043e-06, -8.6650e-06, 1.1688e-07, -2.7008e-08, + -3.2736e-07, 2.7195e-07, 1.2945e-07, 7.3109e-08, 7.7765e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 250.78, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4626 re_mapping 0.0044 re_causal 0.0113 /// teacc 99.06 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.1220, -0.2541, -0.0878, ..., -0.0623, 0.1841, 0.1875], + [-0.2373, -0.2043, -0.0827, ..., -0.1910, -0.2375, -0.1513], + [-0.0728, -0.1805, 0.1493, ..., -0.2238, 0.2610, 0.1179], + ..., + [-0.1824, 0.0932, 0.0251, ..., 0.2088, -0.2288, -0.3006], + [-0.2979, 0.0759, -0.1575, ..., 0.0689, -0.1137, -0.2214], + [-0.0420, -0.1464, -0.0830, ..., -0.1296, -0.0618, -0.2279]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 6.5193e-09, 4.6566e-09, 0.0000e+00, ..., 1.3039e-08, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 1.3970e-09, -7.4506e-09, 0.0000e+00, ..., -6.9849e-09, + 4.6566e-10, 4.6566e-10], + [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 8.3819e-09, 1.7229e-08], + [ 1.1642e-08, 9.3132e-10, 0.0000e+00, ..., 1.4435e-08, + 1.8626e-09, 1.3970e-09]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0142, -0.0338, -0.0117, -0.0130, -0.0317, 0.0007, 0.0225, -0.0176, + 0.0485, -0.0033], device='cuda:0'), grad: tensor([ 9.3132e-09, 4.5635e-08, 4.1910e-09, 1.3039e-08, 1.8626e-09, + -1.0431e-07, -6.6590e-08, -9.3132e-09, 5.2620e-08, 6.1933e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 250.56, cls_loss 0.0016 cls_loss_mapping 0.0020 cls_loss_causal 0.4698 re_mapping 0.0045 re_causal 0.0112 /// teacc 99.03 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.1221, -0.2546, -0.0879, ..., -0.0628, 0.1841, 0.1875], + [-0.2377, -0.2044, -0.0828, ..., -0.1912, -0.2385, -0.1514], + [-0.0729, -0.1812, 0.1500, ..., -0.2242, 0.2618, 0.1182], + ..., + [-0.1829, 0.0935, 0.0246, ..., 0.2090, -0.2290, -0.3008], + [-0.2981, 0.0759, -0.1569, ..., 0.0691, -0.1139, -0.2217], + [-0.0422, -0.1465, -0.0829, ..., -0.1299, -0.0618, -0.2279]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 8.7544e-08, 2.0955e-08], + [ 1.3970e-09, 2.3283e-09, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 4.6566e-10], + [ 2.7940e-09, 3.2596e-09, 4.6566e-10, ..., 0.0000e+00, + 2.2165e-07, 1.3970e-09], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.1642e-08, 1.8626e-09, 0.0000e+00, ..., -4.6566e-10, + 4.6100e-08, 4.6100e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 9.7789e-09, 1.8626e-09]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0142, -0.0346, -0.0114, -0.0131, -0.0306, 0.0007, 0.0226, -0.0176, + 0.0486, -0.0032], device='cuda:0'), grad: tensor([ 2.9849e-07, 1.6298e-08, 7.2550e-07, 2.0489e-08, -1.1390e-06, + 1.5832e-08, -2.1840e-07, 2.7008e-08, 2.8312e-07, -1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 250.30, cls_loss 0.0017 cls_loss_mapping 0.0017 cls_loss_causal 0.4903 re_mapping 0.0043 re_causal 0.0114 /// teacc 98.99 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.1222, -0.2549, -0.0882, ..., -0.0629, 0.1841, 0.1875], + [-0.2378, -0.2037, -0.0828, ..., -0.1916, -0.2390, -0.1516], + [-0.0733, -0.1813, 0.1501, ..., -0.2242, 0.2628, 0.1187], + ..., + [-0.1833, 0.0932, 0.0246, ..., 0.2099, -0.2296, -0.3014], + [-0.2984, 0.0761, -0.1570, ..., 0.0694, -0.1138, -0.2220], + [-0.0417, -0.1471, -0.0838, ..., -0.1312, -0.0619, -0.2279]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.5461e-08, 3.2131e-08], + [ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 2.8871e-08, + 2.3283e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.4715e-07, -4.1910e-08], + ..., + [ 0.0000e+00, -8.8476e-09, 0.0000e+00, ..., -5.4482e-08, + 6.5193e-09, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-09, 2.3283e-09], + [ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 2.4214e-08, + -6.5193e-09, 4.6566e-10]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0143, -0.0330, -0.0116, -0.0131, -0.0303, 0.0003, 0.0227, -0.0184, + 0.0488, -0.0035], device='cuda:0'), grad: tensor([ 3.1339e-07, 1.2666e-07, -3.3714e-07, 1.2014e-07, 7.1246e-08, + 8.6613e-08, -1.0990e-07, 8.7079e-08, 1.8626e-08, -3.7719e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 250.56, cls_loss 0.0018 cls_loss_mapping 0.0018 cls_loss_causal 0.4799 re_mapping 0.0042 re_causal 0.0111 /// teacc 99.03 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.1222, -0.2555, -0.0882, ..., -0.0631, 0.1841, 0.1876], + [-0.2380, -0.2044, -0.0828, ..., -0.1925, -0.2391, -0.1516], + [-0.0736, -0.1805, 0.1502, ..., -0.2238, 0.2636, 0.1194], + ..., + [-0.1837, 0.0936, 0.0245, ..., 0.2104, -0.2307, -0.3026], + [-0.2987, 0.0760, -0.1572, ..., 0.0694, -0.1140, -0.2224], + [-0.0396, -0.1473, -0.0838, ..., -0.1315, -0.0619, -0.2280]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.5903e-08, -3.5390e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.0536e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 3.7253e-09], + ..., + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., -3.7253e-09, + 6.9849e-09, 3.2596e-09], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 3.2596e-09, 7.9162e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 3.2596e-09, + 3.4925e-08, 1.9558e-08]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0143, -0.0331, -0.0113, -0.0136, -0.0293, 0.0006, 0.0228, -0.0187, + 0.0488, -0.0036], device='cuda:0'), grad: tensor([-3.0966e-07, -2.0210e-06, 1.1725e-06, -6.9849e-09, 2.2864e-07, + 5.0291e-08, 7.9162e-09, 9.6858e-07, 3.9581e-08, -1.3923e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 250.08, cls_loss 0.0019 cls_loss_mapping 0.0018 cls_loss_causal 0.4542 re_mapping 0.0042 re_causal 0.0107 /// teacc 99.10 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.1222, -0.2561, -0.0882, ..., -0.0632, 0.1842, 0.1877], + [-0.2381, -0.2048, -0.0826, ..., -0.1915, -0.2393, -0.1518], + [-0.0738, -0.1805, 0.1505, ..., -0.2242, 0.2645, 0.1203], + ..., + [-0.1840, 0.0940, 0.0241, ..., 0.2107, -0.2318, -0.3037], + [-0.2991, 0.0759, -0.1574, ..., 0.0694, -0.1142, -0.2227], + [-0.0393, -0.1480, -0.0837, ..., -0.1328, -0.0619, -0.2281]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 4.3772e-08, 2.5146e-08], + [ 0.0000e+00, 6.5193e-09, 0.0000e+00, ..., 5.7742e-08, + 4.6566e-09, 3.7253e-09], + [ 0.0000e+00, 3.0827e-07, 0.0000e+00, ..., 2.3469e-07, + -2.6077e-07, -2.2352e-07], + ..., + [ 0.0000e+00, -3.3528e-07, 0.0000e+00, ..., -2.5518e-07, + 2.2072e-07, 1.9837e-07], + [ 9.3132e-10, 1.1176e-08, 9.3132e-10, ..., -1.6019e-07, + 1.2107e-08, 6.5193e-09], + [ 0.0000e+00, 3.7253e-09, -1.8626e-09, ..., 4.6566e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0142, -0.0322, -0.0111, -0.0137, -0.0303, -0.0002, 0.0237, -0.0192, + 0.0487, -0.0035], device='cuda:0'), grad: tensor([ 1.2759e-07, -2.4587e-07, -2.0489e-07, 1.3970e-08, 5.5879e-09, + 2.5239e-07, -4.5635e-08, 4.3400e-07, -3.2037e-07, -2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 250.70, cls_loss 0.0018 cls_loss_mapping 0.0027 cls_loss_causal 0.4963 re_mapping 0.0041 re_causal 0.0113 /// teacc 99.04 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.1223, -0.2564, -0.0888, ..., -0.0633, 0.1849, 0.1889], + [-0.2390, -0.2058, -0.0826, ..., -0.1928, -0.2402, -0.1519], + [-0.0734, -0.1808, 0.1520, ..., -0.2246, 0.2657, 0.1205], + ..., + [-0.1843, 0.0949, 0.0235, ..., 0.2118, -0.2323, -0.3040], + [-0.2997, 0.0756, -0.1577, ..., 0.0690, -0.1143, -0.2230], + [-0.0400, -0.1483, -0.0838, ..., -0.1332, -0.0620, -0.2284]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, 1.0151e-07, 0.0000e+00, ..., 2.1327e-07, + 1.8626e-09, -0.0000e+00], + [ 0.0000e+00, 1.9558e-08, 0.0000e+00, ..., 3.0734e-08, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 6.5193e-09, 0.0000e+00, ..., 1.1176e-08, + -2.7940e-09, -1.8626e-09], + ..., + [-4.8429e-08, -3.0175e-07, 0.0000e+00, ..., -5.5786e-07, + 9.3132e-10, 0.0000e+00], + [ 1.6764e-08, 1.8626e-08, 0.0000e+00, ..., 4.0978e-08, + -4.3772e-08, -1.5832e-08], + [ 8.3819e-09, 6.8918e-08, 0.0000e+00, ..., 1.2200e-07, + 4.6566e-09, 2.7940e-09]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0138, -0.0328, -0.0095, -0.0120, -0.0304, -0.0016, 0.0230, -0.0190, + 0.0481, -0.0036], device='cuda:0'), grad: tensor([ 5.9325e-07, 1.1269e-07, 4.8429e-08, 4.6100e-07, 1.1241e-06, + -3.3528e-08, 3.4925e-07, -1.4240e-06, -3.4925e-07, -8.8476e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 250.45, cls_loss 0.0019 cls_loss_mapping 0.0017 cls_loss_causal 0.4810 re_mapping 0.0041 re_causal 0.0109 /// teacc 99.04 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.1240, -0.2569, -0.0889, ..., -0.0633, 0.1844, 0.1883], + [-0.2392, -0.2074, -0.0826, ..., -0.1944, -0.2404, -0.1522], + [-0.0735, -0.1813, 0.1521, ..., -0.2259, 0.2662, 0.1212], + ..., + [-0.1846, 0.0963, 0.0235, ..., 0.2134, -0.2328, -0.3049], + [-0.3004, 0.0753, -0.1577, ..., 0.0691, -0.1147, -0.2234], + [-0.0414, -0.1487, -0.0837, ..., -0.1339, -0.0620, -0.2289]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, -9.3132e-10], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.9558e-08, 0.0000e+00, 0.0000e+00, ..., 1.5832e-08, + 0.0000e+00, 0.0000e+00], + [ 6.8918e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0146, -0.0333, -0.0095, -0.0114, -0.0298, -0.0024, 0.0245, -0.0184, + 0.0479, -0.0040], device='cuda:0'), grad: tensor([ 2.7940e-09, 6.5193e-09, 0.0000e+00, 1.4901e-07, 6.7987e-08, + -5.5134e-07, 8.1956e-08, 1.2107e-08, 6.8918e-08, 1.6671e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 250.39, cls_loss 0.0018 cls_loss_mapping 0.0017 cls_loss_causal 0.4754 re_mapping 0.0043 re_causal 0.0112 /// teacc 99.12 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.1246, -0.2575, -0.0879, ..., -0.0634, 0.1844, 0.1882], + [-0.2394, -0.2076, -0.0828, ..., -0.1948, -0.2407, -0.1524], + [-0.0735, -0.1812, 0.1526, ..., -0.2263, 0.2668, 0.1217], + ..., + [-0.1848, 0.0965, 0.0233, ..., 0.2141, -0.2336, -0.3060], + [-0.3007, 0.0752, -0.1586, ..., 0.0695, -0.1150, -0.2245], + [-0.0418, -0.1490, -0.0836, ..., -0.1342, -0.0621, -0.2291]], + device='cuda:0'), grad: tensor([[-4.0885e-07, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -5.4315e-06, -4.4666e-06], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 2.4214e-08, 3.2596e-08, 0.0000e+00, ..., 4.7497e-08, + 3.1944e-07, 2.6263e-07], + ..., + [ 0.0000e+00, -4.4703e-08, 0.0000e+00, ..., -6.6124e-08, + 9.3132e-10, 9.3132e-10], + [ 1.3039e-08, 4.6566e-09, 0.0000e+00, ..., 5.0291e-08, + 2.0489e-08, 1.6764e-08], + [ 1.0524e-07, 5.5879e-09, -9.3132e-10, ..., -0.0000e+00, + 1.3951e-06, 1.1474e-06]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0148, -0.0341, -0.0094, -0.0114, -0.0314, -0.0027, 0.0254, -0.0185, + 0.0484, -0.0029], device='cuda:0'), grad: tensor([-7.8231e-06, 3.7253e-09, 5.4017e-07, 4.6566e-09, 5.4948e-08, + -1.2014e-07, 5.3234e-06, -9.9652e-08, 3.1572e-07, 1.7993e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 250.32, cls_loss 0.0021 cls_loss_mapping 0.0021 cls_loss_causal 0.4624 re_mapping 0.0042 re_causal 0.0108 /// teacc 99.10 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.1255, -0.2583, -0.0883, ..., -0.0634, 0.1832, 0.1874], + [-0.2394, -0.2084, -0.0823, ..., -0.1956, -0.2408, -0.1534], + [-0.0737, -0.1816, 0.1528, ..., -0.2265, 0.2671, 0.1222], + ..., + [-0.1849, 0.0972, 0.0230, ..., 0.2141, -0.2340, -0.3066], + [-0.3010, 0.0753, -0.1591, ..., 0.0694, -0.1152, -0.2244], + [-0.0423, -0.1494, -0.0843, ..., -0.1331, -0.0614, -0.2293]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.4141e-05, -1.8813e-07], + [ 9.3132e-10, 8.3819e-09, 0.0000e+00, ..., 1.7695e-08, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.5832e-08, 0.0000e+00, ..., 3.8184e-08, + -4.6566e-09, -5.5879e-09], + ..., + [ 9.3132e-10, -3.2596e-08, 0.0000e+00, ..., -8.1025e-08, + 2.6077e-08, 2.7940e-09], + [ 1.3970e-08, 3.3528e-08, 0.0000e+00, ..., -1.6764e-08, + 2.7940e-09, 9.3132e-10], + [ 4.6566e-09, 1.4901e-08, 0.0000e+00, ..., 2.2352e-08, + -1.4201e-05, 1.8813e-07]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0161, -0.0344, -0.0097, -0.0112, -0.0319, -0.0027, 0.0264, -0.0194, + 0.0485, -0.0014], device='cuda:0'), grad: tensor([ 4.1544e-05, 4.1910e-08, 5.9605e-08, -1.0710e-07, 4.8429e-08, + 1.5832e-08, 9.3132e-10, -6.0536e-08, -4.8429e-08, -4.1574e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 250.12, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4616 re_mapping 0.0041 re_causal 0.0109 /// teacc 99.08 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.1256, -0.2595, -0.0883, ..., -0.0632, 0.1834, 0.1877], + [-0.2395, -0.2093, -0.0825, ..., -0.1968, -0.2411, -0.1548], + [-0.0737, -0.1822, 0.1533, ..., -0.2268, 0.2676, 0.1223], + ..., + [-0.1851, 0.0981, 0.0229, ..., 0.2150, -0.2343, -0.3071], + [-0.3014, 0.0751, -0.1594, ..., 0.0691, -0.1156, -0.2231], + [-0.0430, -0.1500, -0.0842, ..., -0.1332, -0.0614, -0.2298]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 6.5193e-09, 4.6566e-09], + [ 0.0000e+00, 2.3842e-07, 0.0000e+00, ..., 3.7998e-07, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.2945e-07, 0.0000e+00, ..., 2.7195e-07, + -8.3819e-08, -1.3039e-08], + ..., + [ 0.0000e+00, -1.6876e-06, 0.0000e+00, ..., -2.7493e-06, + 6.7055e-08, 1.0245e-08], + [ 0.0000e+00, 1.2834e-06, 0.0000e+00, ..., 2.0508e-06, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 9.3132e-09, 0.0000e+00, ..., 1.4901e-08, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0161, -0.0351, -0.0096, -0.0112, -0.0326, -0.0025, 0.0263, -0.0191, + 0.0495, -0.0011], device='cuda:0'), grad: tensor([ 2.1420e-08, -4.3139e-06, 3.4086e-06, 1.0524e-07, 6.2399e-08, + 4.0047e-08, -1.3970e-08, -3.1386e-06, 3.7588e-06, 4.0978e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 250.32, cls_loss 0.0019 cls_loss_mapping 0.0019 cls_loss_causal 0.4969 re_mapping 0.0041 re_causal 0.0111 /// teacc 98.97 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.1257, -0.2601, -0.0884, ..., -0.0639, 0.1823, 0.1878], + [-0.2422, -0.2095, -0.0832, ..., -0.1971, -0.2416, -0.1564], + [-0.0709, -0.1821, 0.1568, ..., -0.2269, 0.2706, 0.1250], + ..., + [-0.1853, 0.0984, 0.0229, ..., 0.2154, -0.2362, -0.3076], + [-0.3032, 0.0748, -0.1595, ..., 0.0675, -0.1154, -0.2259], + [-0.0425, -0.1504, -0.0842, ..., -0.1337, -0.0599, -0.2300]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.2107e-08, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, 0.0000e+00, ..., 1.8626e-09, + -3.7253e-09, 0.0000e+00], + ..., + [ 1.3039e-08, 1.9558e-08, 0.0000e+00, ..., -1.8626e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 1.1176e-08, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0171, -0.0365, -0.0080, -0.0113, -0.0328, -0.0021, 0.0265, -0.0182, + 0.0475, -0.0003], device='cuda:0'), grad: tensor([ 5.4948e-08, -4.4610e-07, 4.0978e-08, -1.8626e-09, 8.0746e-07, + 2.2762e-06, 1.7881e-07, 2.0489e-06, 8.3819e-08, -5.0664e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 250.40, cls_loss 0.0017 cls_loss_mapping 0.0018 cls_loss_causal 0.4793 re_mapping 0.0040 re_causal 0.0108 /// teacc 99.00 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.1262, -0.2612, -0.0885, ..., -0.0640, 0.1821, 0.1879], + [-0.2422, -0.2111, -0.0834, ..., -0.1986, -0.2427, -0.1566], + [-0.0709, -0.1827, 0.1591, ..., -0.2272, 0.2734, 0.1269], + ..., + [-0.1856, 0.0996, 0.0229, ..., 0.2153, -0.2368, -0.3079], + [-0.3030, 0.0760, -0.1594, ..., 0.0695, -0.1162, -0.2265], + [-0.0427, -0.1518, -0.0842, ..., -0.1345, -0.0597, -0.2301]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -8.3819e-09, -5.5879e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 5.5879e-09]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0174, -0.0372, -0.0056, -0.0118, -0.0328, -0.0018, 0.0262, -0.0184, + 0.0492, -0.0001], device='cuda:0'), grad: tensor([-1.4901e-08, -2.2158e-05, 3.2261e-06, 5.8673e-08, 1.3039e-08, + 3.9116e-08, -3.3528e-08, 1.8850e-05, 4.0978e-08, -2.8871e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 250.37, cls_loss 0.0016 cls_loss_mapping 0.0011 cls_loss_causal 0.4720 re_mapping 0.0039 re_causal 0.0107 /// teacc 99.03 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.1262, -0.2613, -0.0887, ..., -0.0641, 0.1823, 0.1881], + [-0.2422, -0.2114, -0.0836, ..., -0.1983, -0.2438, -0.1574], + [-0.0710, -0.1836, 0.1597, ..., -0.2281, 0.2741, 0.1275], + ..., + [-0.1860, 0.1004, 0.0229, ..., 0.2155, -0.2371, -0.3082], + [-0.3036, 0.0758, -0.1591, ..., 0.0693, -0.1164, -0.2268], + [-0.0428, -0.1527, -0.0844, ..., -0.1346, -0.0598, -0.2303]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -3.0734e-08, -2.1420e-08], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, -2.7940e-09], + [ 0.0000e+00, 9.3132e-10, -6.1467e-08, ..., 0.0000e+00, + -4.4703e-08, -1.0245e-08], + ..., + [ 2.7940e-09, 3.7253e-09, 5.7742e-08, ..., 2.7940e-09, + 4.1910e-08, 9.3132e-09], + [ 1.4901e-08, 1.9558e-08, 0.0000e+00, ..., 1.6764e-08, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-08, 1.7695e-08]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0173, -0.0367, -0.0055, -0.0123, -0.0319, -0.0014, 0.0262, -0.0191, + 0.0490, -0.0001], device='cuda:0'), grad: tensor([-3.8184e-08, -1.6205e-07, -5.8487e-07, -3.9116e-08, -6.6031e-07, + -8.5682e-08, 6.8266e-07, 6.0629e-07, 1.9092e-07, 7.7300e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 250.14, cls_loss 0.0018 cls_loss_mapping 0.0013 cls_loss_causal 0.4554 re_mapping 0.0042 re_causal 0.0107 /// teacc 99.13 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.1263, -0.2612, -0.0888, ..., -0.0659, 0.1826, 0.1885], + [-0.2430, -0.2116, -0.0837, ..., -0.1986, -0.2439, -0.1571], + [-0.0712, -0.1858, 0.1597, ..., -0.2302, 0.2741, 0.1272], + ..., + [-0.1859, 0.1006, 0.0227, ..., 0.2155, -0.2371, -0.3083], + [-0.3043, 0.0768, -0.1586, ..., 0.0703, -0.1167, -0.2270], + [-0.0448, -0.1539, -0.0853, ..., -0.1358, -0.0601, -0.2313]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.7253e-09, 0.0000e+00, ..., -4.7497e-08, + -2.4308e-06, -1.1157e-06], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 3.7253e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.0489e-08, 9.3132e-09], + ..., + [ 9.3132e-10, -1.8626e-09, 1.2107e-08, ..., 8.5682e-08, + 6.5193e-09, 3.7253e-09], + [ 4.0978e-08, 9.3132e-10, 0.0000e+00, ..., 3.1665e-08, + 1.1129e-06, 5.1130e-07], + [ 3.7253e-09, 4.6566e-09, -1.3970e-08, ..., -9.1270e-08, + 1.0859e-06, 5.0012e-07]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0171, -0.0365, -0.0062, -0.0132, -0.0323, -0.0005, 0.0263, -0.0194, + 0.0502, -0.0002], device='cuda:0'), grad: tensor([-4.7311e-06, 2.4214e-08, 3.9116e-08, 1.8999e-07, 4.2841e-08, + -9.2909e-06, 9.3058e-06, 4.5914e-07, 2.2873e-06, 1.6494e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 250.29, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4955 re_mapping 0.0041 re_causal 0.0112 /// teacc 99.06 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.1282, -0.2616, -0.0888, ..., -0.0659, 0.1811, 0.1868], + [-0.2431, -0.2116, -0.0837, ..., -0.1956, -0.2443, -0.1571], + [-0.0712, -0.1867, 0.1597, ..., -0.2308, 0.2747, 0.1274], + ..., + [-0.1859, 0.1010, 0.0226, ..., 0.2134, -0.2378, -0.3086], + [-0.3045, 0.0768, -0.1589, ..., 0.0703, -0.1177, -0.2277], + [-0.0452, -0.1548, -0.0852, ..., -0.1367, -0.0602, -0.2315]], + device='cuda:0'), grad: tensor([[ 5.7966e-05, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.0842e-04, 6.5684e-05], + [ 4.6566e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 5.5879e-09], + [ 3.1665e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.8673e-08, 3.5390e-08], + ..., + [ 5.5879e-09, -9.3132e-10, 0.0000e+00, ..., -1.8626e-09, + 1.1176e-08, 6.5193e-09], + [ 2.6077e-08, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 3.8184e-08, 2.2352e-08], + [ 1.7695e-08, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 3.4459e-08, 2.0489e-08]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0181, -0.0339, -0.0059, -0.0150, -0.0322, 0.0012, 0.0277, -0.0219, + 0.0501, -0.0004], device='cuda:0'), grad: tensor([ 2.3305e-04, 2.0489e-08, 1.2666e-07, 0.0000e+00, 4.1910e-08, + 6.4541e-07, -2.3448e-04, 2.1420e-08, 9.4064e-08, 7.5437e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 250.31, cls_loss 0.0015 cls_loss_mapping 0.0016 cls_loss_causal 0.4503 re_mapping 0.0040 re_causal 0.0105 /// teacc 99.11 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.1295, -0.2624, -0.0888, ..., -0.0661, 0.1806, 0.1863], + [-0.2450, -0.2124, -0.0837, ..., -0.1961, -0.2457, -0.1583], + [-0.0694, -0.1837, 0.1597, ..., -0.2301, 0.2761, 0.1280], + ..., + [-0.1864, 0.1014, 0.0226, ..., 0.2141, -0.2398, -0.3088], + [-0.3046, 0.0759, -0.1590, ..., 0.0702, -0.1185, -0.2281], + [-0.0459, -0.1558, -0.0851, ..., -0.1373, -0.0604, -0.2322]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.7940e-08, 0.0000e+00, ..., 2.0489e-08, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 3.5949e-07, 0.0000e+00, ..., 2.8126e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.4459e-08, 0.0000e+00, ..., 3.5390e-08, + -1.8626e-09, -9.3132e-10], + ..., + [ 0.0000e+00, -1.2117e-06, 0.0000e+00, ..., -9.3784e-07, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.2480e-07, 0.0000e+00, ..., 8.7544e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.8243e-07, 0.0000e+00, ..., 3.5018e-07, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0186, -0.0341, -0.0044, -0.0150, -0.0330, 0.0015, 0.0283, -0.0218, + 0.0499, -0.0005], device='cuda:0'), grad: tensor([ 8.3819e-08, 1.0040e-06, 9.1270e-08, 4.7963e-07, -2.0489e-08, + 4.0047e-08, 3.5390e-08, -3.4273e-06, 3.4645e-07, 1.3690e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 250.07, cls_loss 0.0016 cls_loss_mapping 0.0013 cls_loss_causal 0.4882 re_mapping 0.0040 re_causal 0.0110 /// teacc 99.04 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.1295, -0.2633, -0.0888, ..., -0.0665, 0.1807, 0.1866], + [-0.2455, -0.2126, -0.0837, ..., -0.1963, -0.2468, -0.1586], + [-0.0694, -0.1840, 0.1598, ..., -0.2307, 0.2765, 0.1280], + ..., + [-0.1866, 0.1020, 0.0222, ..., 0.2148, -0.2398, -0.3090], + [-0.3051, 0.0758, -0.1590, ..., 0.0702, -0.1194, -0.2289], + [-0.0449, -0.1567, -0.0847, ..., -0.1389, -0.0603, -0.2330]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-08, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -1.8626e-09, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, -6.5193e-09, 0.0000e+00, ..., -1.8626e-09, + -1.6112e-07, -1.8626e-08], + ..., + [ 0.0000e+00, 9.3132e-10, 2.7940e-09, ..., 1.8626e-09, + 6.5193e-09, 3.7253e-09], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 9.3132e-10, + 2.2352e-08, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 3.7253e-09, 1.8626e-09]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0186, -0.0343, -0.0046, -0.0149, -0.0333, 0.0016, 0.0281, -0.0216, + 0.0502, -0.0005], device='cuda:0'), grad: tensor([ 1.2573e-07, -1.5460e-07, -3.0082e-07, 5.9605e-08, 8.3819e-09, + 2.7940e-09, 2.2352e-08, 1.1828e-07, 5.8673e-08, 7.1712e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 250.39, cls_loss 0.0014 cls_loss_mapping 0.0011 cls_loss_causal 0.4805 re_mapping 0.0041 re_causal 0.0111 /// teacc 99.09 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.1296, -0.2645, -0.0888, ..., -0.0674, 0.1808, 0.1867], + [-0.2456, -0.2159, -0.0836, ..., -0.1983, -0.2475, -0.1578], + [-0.0695, -0.1844, 0.1598, ..., -0.2312, 0.2767, 0.1275], + ..., + [-0.1869, 0.1052, 0.0224, ..., 0.2167, -0.2400, -0.3095], + [-0.3062, 0.0760, -0.1593, ..., 0.0711, -0.1196, -0.2296], + [-0.0443, -0.1574, -0.0846, ..., -0.1396, -0.0605, -0.2334]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.7497e-08, -2.8871e-08], + [ 4.6566e-09, 1.2107e-08, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 2.6077e-08, 0.0000e+00, ..., 9.3132e-10, + 3.7253e-09, 1.8626e-09], + [ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + 2.2352e-08, 1.4901e-08]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0185, -0.0356, -0.0051, -0.0148, -0.0328, 0.0011, 0.0280, -0.0204, + 0.0526, -0.0008], device='cuda:0'), grad: tensor([-6.5193e-08, -4.3772e-08, 1.1176e-08, -1.4342e-07, 1.8626e-09, + 3.9116e-08, 3.5390e-08, 8.3819e-09, 1.0431e-07, 5.5879e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 250.15, cls_loss 0.0017 cls_loss_mapping 0.0021 cls_loss_causal 0.4706 re_mapping 0.0042 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.1299, -0.2653, -0.0891, ..., -0.0677, 0.1808, 0.1867], + [-0.2460, -0.2162, -0.0854, ..., -0.1987, -0.2480, -0.1578], + [-0.0696, -0.1853, 0.1598, ..., -0.2327, 0.2768, 0.1274], + ..., + [-0.1873, 0.1082, 0.0263, ..., 0.2199, -0.2402, -0.3096], + [-0.3076, 0.0757, -0.1595, ..., 0.0713, -0.1182, -0.2303], + [-0.0442, -0.1612, -0.0882, ..., -0.1435, -0.0607, -0.2339]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.9116e-08, -1.7695e-08], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 9.3132e-10], + [ 0.0000e+00, -2.6077e-08, 0.0000e+00, ..., 0.0000e+00, + -6.4261e-08, -1.4901e-08], + ..., + [ 0.0000e+00, 2.1420e-08, 0.0000e+00, ..., 0.0000e+00, + 5.8673e-08, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 3.5390e-08, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 5.5879e-09, 2.7940e-09]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0185, -0.0356, -0.0056, -0.0148, -0.0335, 0.0011, 0.0278, -0.0185, + 0.0530, -0.0031], device='cuda:0'), grad: tensor([-6.4261e-08, 7.4506e-09, -1.7509e-07, 5.7742e-08, 1.8626e-08, + 4.6566e-09, 3.7253e-09, 1.5646e-07, -2.1420e-08, 1.9558e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 350---------------------------------------------------- +epoch 350, time 268.03, cls_loss 0.0017 cls_loss_mapping 0.0017 cls_loss_causal 0.4852 re_mapping 0.0041 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.1301, -0.2672, -0.0890, ..., -0.0681, 0.1812, 0.1872], + [-0.2460, -0.2163, -0.0847, ..., -0.1991, -0.2483, -0.1581], + [-0.0697, -0.1861, 0.1598, ..., -0.2355, 0.2772, 0.1285], + ..., + [-0.1876, 0.1085, 0.0263, ..., 0.2208, -0.2404, -0.3096], + [-0.3088, 0.0754, -0.1596, ..., 0.0712, -0.1189, -0.2326], + [-0.0448, -0.1613, -0.0882, ..., -0.1439, -0.0611, -0.2348]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, 0.0000e+00, ..., -9.3132e-10, + 1.8626e-09, 2.7940e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, + 3.7253e-09, 1.8626e-09]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0183, -0.0357, -0.0059, -0.0152, -0.0329, 0.0015, 0.0277, -0.0182, + 0.0528, -0.0035], device='cuda:0'), grad: tensor([ 9.3132e-10, 2.7940e-09, 2.7940e-09, -1.1176e-08, -3.2596e-08, + 4.7125e-07, -4.5542e-07, -6.5193e-09, 1.1176e-08, 1.9558e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 250.51, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4895 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.01 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.1301, -0.2667, -0.0890, ..., -0.0682, 0.1818, 0.1877], + [-0.2460, -0.2164, -0.0847, ..., -0.1992, -0.2485, -0.1582], + [-0.0697, -0.1866, 0.1600, ..., -0.2365, 0.2773, 0.1289], + ..., + [-0.1886, 0.1092, 0.0263, ..., 0.2219, -0.2405, -0.3099], + [-0.3109, 0.0752, -0.1603, ..., 0.0709, -0.1199, -0.2337], + [-0.0452, -0.1622, -0.0882, ..., -0.1449, -0.0613, -0.2353]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, -2.7940e-09], + ..., + [ 1.8626e-08, -0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + 1.8626e-09, 9.3132e-10], + [ 1.5832e-08, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0179, -0.0357, -0.0060, -0.0150, -0.0329, 0.0006, 0.0282, -0.0178, + 0.0520, -0.0039], device='cuda:0'), grad: tensor([ 9.3132e-09, -2.4121e-07, 1.8347e-07, 0.0000e+00, 5.5879e-09, + -8.1956e-08, 9.3132e-09, 6.9849e-08, 2.7008e-08, 1.7695e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 250.26, cls_loss 0.0020 cls_loss_mapping 0.0020 cls_loss_causal 0.4979 re_mapping 0.0039 re_causal 0.0106 /// teacc 98.93 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.1301, -0.2674, -0.0890, ..., -0.0685, 0.1816, 0.1878], + [-0.2460, -0.2165, -0.0848, ..., -0.1990, -0.2486, -0.1583], + [-0.0698, -0.1868, 0.1600, ..., -0.2353, 0.2786, 0.1306], + ..., + [-0.1895, 0.1093, 0.0264, ..., 0.2219, -0.2422, -0.3127], + [-0.3133, 0.0750, -0.1602, ..., 0.0701, -0.1210, -0.2345], + [-0.0450, -0.1624, -0.0882, ..., -0.1453, -0.0612, -0.2357]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 2.7940e-09], + [ 2.7940e-09, 1.8626e-08, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-09, 9.3132e-10], + [ 2.7940e-09, 5.4017e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 1.9558e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, -9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0182, -0.0354, -0.0057, -0.0150, -0.0333, 0.0014, 0.0277, -0.0181, + 0.0508, -0.0036], device='cuda:0'), grad: tensor([ 3.5390e-08, -6.4075e-07, 1.0058e-07, -1.3784e-07, 6.4820e-07, + 5.4017e-08, -7.5437e-08, 4.6566e-08, 3.0734e-08, -5.2154e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 250.79, cls_loss 0.0014 cls_loss_mapping 0.0012 cls_loss_causal 0.4762 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.02 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.1302, -0.2681, -0.0890, ..., -0.0694, 0.1818, 0.1880], + [-0.2464, -0.2165, -0.0848, ..., -0.1991, -0.2507, -0.1583], + [-0.0698, -0.1870, 0.1600, ..., -0.2354, 0.2800, 0.1311], + ..., + [-0.1896, 0.1093, 0.0263, ..., 0.2223, -0.2426, -0.3131], + [-0.3136, 0.0749, -0.1608, ..., 0.0702, -0.1216, -0.2351], + [-0.0453, -0.1625, -0.0882, ..., -0.1456, -0.0615, -0.2363]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-09, -9.3132e-10], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 1.3039e-08, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 5.4948e-08, 0.0000e+00, ..., 6.0536e-08, + -6.5193e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -7.0781e-08, 0.0000e+00, ..., -8.7544e-08, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 8.3819e-09, 0.0000e+00, ..., 1.1176e-08, + 1.8626e-09, 0.0000e+00], + [ 4.6566e-09, 7.4506e-09, 0.0000e+00, ..., 7.4506e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0182, -0.0354, -0.0050, -0.0149, -0.0332, 0.0012, 0.0280, -0.0181, + 0.0506, -0.0037], device='cuda:0'), grad: tensor([ 2.3283e-08, -3.6228e-07, 1.4063e-07, -1.3970e-08, 1.0710e-07, + -5.8673e-08, 1.3039e-07, -1.5553e-07, 1.0710e-07, 8.4750e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 250.57, cls_loss 0.0015 cls_loss_mapping 0.0015 cls_loss_causal 0.4948 re_mapping 0.0040 re_causal 0.0108 /// teacc 99.08 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.1304, -0.2683, -0.0890, ..., -0.0697, 0.1818, 0.1882], + [-0.2476, -0.2166, -0.0849, ..., -0.1993, -0.2508, -0.1591], + [-0.0687, -0.1872, 0.1600, ..., -0.2356, 0.2803, 0.1316], + ..., + [-0.1900, 0.1094, 0.0263, ..., 0.2230, -0.2428, -0.3132], + [-0.3156, 0.0748, -0.1608, ..., 0.0693, -0.1227, -0.2356], + [-0.0465, -0.1626, -0.0882, ..., -0.1464, -0.0616, -0.2366]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 2.7940e-09, -0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-08, 1.9558e-08, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 1.6764e-08, 7.4506e-09, 0.0000e+00, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0184, -0.0354, -0.0046, -0.0149, -0.0345, 0.0017, 0.0282, -0.0178, + 0.0497, -0.0038], device='cuda:0'), grad: tensor([ 7.4506e-09, 2.0489e-08, 1.3039e-08, -1.0431e-07, 9.3132e-10, + -1.8999e-07, 1.3411e-07, 9.3132e-09, -1.1083e-07, 2.2724e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 250.49, cls_loss 0.0014 cls_loss_mapping 0.0010 cls_loss_causal 0.4571 re_mapping 0.0039 re_causal 0.0106 /// teacc 99.03 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.1304, -0.2693, -0.0879, ..., -0.0704, 0.1820, 0.1884], + [-0.2477, -0.2182, -0.0847, ..., -0.2001, -0.2509, -0.1591], + [-0.0687, -0.1872, 0.1600, ..., -0.2358, 0.2804, 0.1318], + ..., + [-0.1904, 0.1106, 0.0263, ..., 0.2236, -0.2431, -0.3136], + [-0.3173, 0.0748, -0.1603, ..., 0.0690, -0.1226, -0.2356], + [-0.0466, -0.1627, -0.0882, ..., -0.1465, -0.0616, -0.2367]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + -3.6322e-08, -1.8161e-08], + [ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 4.6566e-10], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, 9.3132e-10], + ..., + [ 1.3970e-09, -1.3970e-09, 0.0000e+00, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 9.3132e-10], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 2.4214e-08, 1.2573e-08]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0183, -0.0360, -0.0045, -0.0152, -0.0363, 0.0021, 0.0280, -0.0173, + 0.0493, -0.0033], device='cuda:0'), grad: tensor([-6.7055e-08, -3.3528e-08, 1.2573e-08, -9.0804e-08, -9.3132e-09, + 7.4971e-08, 9.3132e-09, 3.4459e-08, 1.4435e-08, 5.6811e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 250.74, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.4599 re_mapping 0.0040 re_causal 0.0109 /// teacc 99.07 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.1304, -0.2704, -0.0885, ..., -0.0704, 0.1820, 0.1885], + [-0.2497, -0.2182, -0.0846, ..., -0.2002, -0.2512, -0.1592], + [-0.0667, -0.1880, 0.1601, ..., -0.2375, 0.2808, 0.1319], + ..., + [-0.1908, 0.1099, 0.0263, ..., 0.2239, -0.2435, -0.3138], + [-0.3176, 0.0745, -0.1590, ..., 0.0689, -0.1225, -0.2354], + [-0.0469, -0.1628, -0.0882, ..., -0.1467, -0.0617, -0.2369]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + -6.0536e-09, -3.7253e-09], + [ 3.7253e-09, 1.3504e-08, 0.0000e+00, ..., 8.8476e-09, + 4.6566e-10, 0.0000e+00], + [ 3.2596e-09, 1.4435e-08, 0.0000e+00, ..., 1.0245e-08, + 4.6566e-10, 4.6566e-10], + ..., + [ 2.7940e-09, -3.2131e-08, 0.0000e+00, ..., -4.6566e-08, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 7.9162e-09, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 4.6566e-10], + [-9.3132e-10, 6.0536e-09, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-09, 3.2596e-09]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0183, -0.0362, -0.0037, -0.0146, -0.0338, 0.0021, 0.0280, -0.0174, + 0.0491, -0.0046], device='cuda:0'), grad: tensor([-7.4506e-09, 3.9116e-08, 4.3306e-08, -6.0536e-09, -1.1176e-08, + 2.5146e-08, 6.5193e-09, -8.2422e-08, 1.8626e-08, -2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 250.33, cls_loss 0.0018 cls_loss_mapping 0.0020 cls_loss_causal 0.4706 re_mapping 0.0040 re_causal 0.0104 /// teacc 99.13 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.1306, -0.2713, -0.0887, ..., -0.0704, 0.1821, 0.1886], + [-0.2499, -0.2183, -0.0853, ..., -0.2003, -0.2516, -0.1594], + [-0.0666, -0.1888, 0.1603, ..., -0.2395, 0.2810, 0.1320], + ..., + [-0.1915, 0.1093, 0.0263, ..., 0.2228, -0.2436, -0.3140], + [-0.3179, 0.0771, -0.1584, ..., 0.0718, -0.1229, -0.2359], + [-0.0476, -0.1629, -0.0882, ..., -0.1468, -0.0618, -0.2373]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 5.5879e-09, + 5.1921e-07, -2.6077e-08], + [ 0.0000e+00, 8.8476e-09, 0.0000e+00, ..., 1.0664e-07, + 3.9581e-08, 6.0536e-09], + [ 0.0000e+00, 5.1223e-09, -0.0000e+00, ..., 6.0536e-08, + -8.2795e-07, -9.3598e-08], + ..., + [ 4.6566e-10, 2.6543e-08, 0.0000e+00, ..., -1.9884e-07, + 6.9849e-09, 1.3970e-09], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 8.3819e-09, + 2.3283e-09, 9.3132e-10], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.9558e-08, + 8.3819e-08, 4.0978e-08]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0183, -0.0361, -0.0041, -0.0136, -0.0319, 0.0009, 0.0282, -0.0179, + 0.0515, -0.0056], device='cuda:0'), grad: tensor([ 9.7416e-07, 3.7951e-07, -1.3150e-06, 6.4727e-08, 7.8697e-08, + 1.2647e-06, 1.8114e-07, -5.0105e-07, 3.1665e-08, -1.1306e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 250.50, cls_loss 0.0019 cls_loss_mapping 0.0020 cls_loss_causal 0.4925 re_mapping 0.0039 re_causal 0.0105 /// teacc 99.15 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.1306, -0.2718, -0.0888, ..., -0.0703, 0.1831, 0.1897], + [-0.2500, -0.2185, -0.0825, ..., -0.2003, -0.2547, -0.1598], + [-0.0669, -0.1906, 0.1604, ..., -0.2399, 0.2840, 0.1323], + ..., + [-0.1920, 0.1096, 0.0261, ..., 0.2231, -0.2442, -0.3145], + [-0.3185, 0.0766, -0.1587, ..., 0.0714, -0.1237, -0.2368], + [-0.0478, -0.1630, -0.0882, ..., -0.1469, -0.0621, -0.2382]], + device='cuda:0'), grad: tensor([[ 1.4435e-08, 5.5879e-09, 0.0000e+00, ..., -2.6636e-07, + -1.1437e-06, -7.0501e-07], + [ 4.6566e-10, 5.4948e-08, 0.0000e+00, ..., 4.4238e-08, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-10, 2.7986e-07, 0.0000e+00, ..., 1.3504e-08, + -1.0245e-08, 4.6566e-10], + ..., + [ 0.0000e+00, -1.2480e-07, 0.0000e+00, ..., -4.4471e-07, + 1.8626e-09, 1.3970e-09], + [ 5.8673e-08, 1.5367e-07, 0.0000e+00, ..., 3.3993e-08, + 3.8650e-08, 7.4971e-08], + [ 4.6566e-10, 1.9697e-07, 0.0000e+00, ..., 2.7195e-07, + 1.9558e-08, 1.2107e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0176, -0.0357, -0.0016, -0.0135, -0.0319, 0.0013, 0.0272, -0.0192, + 0.0509, -0.0057], device='cuda:0'), grad: tensor([-2.1737e-06, 2.9150e-07, 4.5961e-07, -8.7824e-07, 1.7229e-08, + 3.3807e-06, -1.1995e-06, -1.5795e-06, 5.1875e-07, 1.1828e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 250.73, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4487 re_mapping 0.0039 re_causal 0.0104 /// teacc 99.00 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.1307, -0.2721, -0.0888, ..., -0.0701, 0.1839, 0.1906], + [-0.2501, -0.2186, -0.0829, ..., -0.2006, -0.2547, -0.1585], + [-0.0674, -0.1922, 0.1604, ..., -0.2406, 0.2839, 0.1302], + ..., + [-0.1929, 0.1100, 0.0262, ..., 0.2236, -0.2457, -0.3149], + [-0.3192, 0.0761, -0.1588, ..., 0.0711, -0.1280, -0.2406], + [-0.0464, -0.1631, -0.0882, ..., -0.1470, -0.0623, -0.2390]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.8255e-08, -8.4285e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + -3.2596e-09, -4.6566e-10], + ..., + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, -2.7940e-09], + [ 1.3970e-09, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + 1.2573e-08, 1.0710e-08]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0170, -0.0355, -0.0024, -0.0134, -0.0322, 0.0015, 0.0272, -0.0190, + 0.0498, -0.0056], device='cuda:0'), grad: tensor([-1.4808e-07, 8.5682e-08, -5.5879e-09, -1.6438e-07, 7.5549e-06, + 1.6764e-07, 1.5274e-07, 5.3272e-07, 1.2619e-07, -8.2925e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 250.61, cls_loss 0.0015 cls_loss_mapping 0.0014 cls_loss_causal 0.4910 re_mapping 0.0038 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.1308, -0.2733, -0.0888, ..., -0.0706, 0.1841, 0.1909], + [-0.2508, -0.2188, -0.0837, ..., -0.2008, -0.2547, -0.1583], + [-0.0670, -0.1922, 0.1606, ..., -0.2407, 0.2840, 0.1300], + ..., + [-0.1930, 0.1106, 0.0262, ..., 0.2241, -0.2459, -0.3152], + [-0.3203, 0.0760, -0.1595, ..., 0.0709, -0.1283, -0.2409], + [-0.0465, -0.1641, -0.0882, ..., -0.1475, -0.0624, -0.2392]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, -2.3283e-09], + ..., + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 9.3132e-10], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0171, -0.0352, -0.0028, -0.0134, -0.0321, 0.0015, 0.0272, -0.0189, + 0.0492, -0.0059], device='cuda:0'), grad: tensor([ 6.5193e-09, -2.8405e-08, -9.7789e-09, 3.7253e-09, 6.5193e-09, + 1.8626e-09, 6.5193e-09, 2.3283e-09, 5.5879e-09, 1.8161e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 250.51, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4686 re_mapping 0.0038 re_causal 0.0100 /// teacc 99.09 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.1308, -0.2761, -0.0888, ..., -0.0717, 0.1846, 0.1915], + [-0.2510, -0.2190, -0.0837, ..., -0.2010, -0.2548, -0.1585], + [-0.0671, -0.1924, 0.1607, ..., -0.2408, 0.2841, 0.1301], + ..., + [-0.1936, 0.1117, 0.0262, ..., 0.2253, -0.2461, -0.3154], + [-0.3207, 0.0759, -0.1595, ..., 0.0708, -0.1281, -0.2404], + [-0.0464, -0.1655, -0.0882, ..., -0.1490, -0.0624, -0.2395]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 7.8697e-08, 0.0000e+00, ..., 1.4855e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.5367e-08, 0.0000e+00, ..., 2.9802e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.6578e-07, 0.0000e+00, ..., -3.1525e-07, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 1.3970e-09, 0.0000e+00, ..., 3.2596e-09, + 4.1910e-09, 4.6566e-09], + [ 0.0000e+00, 6.8452e-08, 0.0000e+00, ..., 1.3085e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0168, -0.0353, -0.0028, -0.0133, -0.0321, 0.0011, 0.0271, -0.0183, + 0.0493, -0.0065], device='cuda:0'), grad: tensor([ 6.3796e-08, -9.9186e-08, 8.5216e-08, 1.3039e-08, 1.6205e-07, + 7.9162e-09, -2.0955e-08, -6.9337e-07, 3.8184e-08, 4.5123e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 250.65, cls_loss 0.0016 cls_loss_mapping 0.0019 cls_loss_causal 0.4497 re_mapping 0.0040 re_causal 0.0104 /// teacc 99.04 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.1309, -0.2780, -0.0886, ..., -0.0727, 0.1846, 0.1915], + [-0.2531, -0.2192, -0.0839, ..., -0.2013, -0.2548, -0.1593], + [-0.0650, -0.1928, 0.1607, ..., -0.2409, 0.2841, 0.1307], + ..., + [-0.1942, 0.1119, 0.0262, ..., 0.2257, -0.2461, -0.3155], + [-0.3212, 0.0759, -0.1596, ..., 0.0711, -0.1273, -0.2394], + [-0.0469, -0.1657, -0.0882, ..., -0.1493, -0.0625, -0.2396]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -6.6124e-08, -4.9360e-08], + [ 5.1223e-09, 1.4575e-07, 0.0000e+00, ..., 4.0606e-07, + 3.9116e-08, 2.3749e-08], + [ 6.9849e-09, 1.3970e-08, 0.0000e+00, ..., 1.0710e-08, + 9.3132e-09, 5.1223e-09], + ..., + [ 4.7497e-08, -6.4261e-08, 0.0000e+00, ..., -4.0093e-07, + 0.0000e+00, 0.0000e+00], + [ 7.5903e-08, 1.2200e-07, 0.0000e+00, ..., 1.8626e-08, + -1.4435e-08, 2.9337e-08], + [ 2.7474e-08, 4.6566e-08, 0.0000e+00, ..., 1.0710e-08, + 3.2596e-09, -2.7940e-08]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0169, -0.0357, -0.0020, -0.0133, -0.0324, 0.0042, 0.0238, -0.0182, + 0.0501, -0.0065], device='cuda:0'), grad: tensor([ 1.8626e-07, 1.1101e-06, 8.6613e-08, -5.0701e-06, 1.1176e-08, + 4.5672e-06, 4.1444e-08, -8.8196e-07, 5.6718e-07, -6.2119e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 250.85, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4321 re_mapping 0.0041 re_causal 0.0102 /// teacc 99.05 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.1310, -0.2816, -0.0886, ..., -0.0741, 0.1846, 0.1914], + [-0.2534, -0.2195, -0.0839, ..., -0.2016, -0.2549, -0.1594], + [-0.0649, -0.1931, 0.1608, ..., -0.2418, 0.2845, 0.1319], + ..., + [-0.1966, 0.1124, 0.0262, ..., 0.2262, -0.2466, -0.3160], + [-0.3217, 0.0755, -0.1601, ..., 0.0711, -0.1279, -0.2402], + [-0.0484, -0.1659, -0.0882, ..., -0.1494, -0.0628, -0.2407]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., -5.2154e-08, + -6.4261e-08, -4.8429e-08], + [ 0.0000e+00, 9.3132e-10, -0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + -3.7253e-09, -1.3970e-09], + ..., + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., -2.3283e-09, + 1.8626e-09, 9.3132e-10], + [ 3.5390e-08, -1.3970e-09, 0.0000e+00, ..., 1.2759e-07, + 1.3970e-09, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, -0.0000e+00, ..., 1.8626e-09, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0172, -0.0358, -0.0019, -0.0132, -0.0320, 0.0042, 0.0237, -0.0181, + 0.0497, -0.0066], device='cuda:0'), grad: tensor([-1.4994e-07, -4.6566e-10, -9.7789e-09, 9.3132e-09, 3.2596e-09, + -8.8476e-08, -1.3970e-09, -9.3132e-10, 2.3376e-07, 3.2596e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 250.81, cls_loss 0.0013 cls_loss_mapping 0.0012 cls_loss_causal 0.4638 re_mapping 0.0039 re_causal 0.0104 /// teacc 99.02 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.1310, -0.2819, -0.0887, ..., -0.0741, 0.1848, 0.1917], + [-0.2534, -0.2199, -0.0841, ..., -0.2024, -0.2550, -0.1595], + [-0.0650, -0.1934, 0.1608, ..., -0.2421, 0.2847, 0.1320], + ..., + [-0.1972, 0.1127, 0.0262, ..., 0.2267, -0.2471, -0.3163], + [-0.3221, 0.0754, -0.1603, ..., 0.0712, -0.1278, -0.2404], + [-0.0478, -0.1659, -0.0882, ..., -0.1495, -0.0629, -0.2409]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 2.7940e-09, 0.0000e+00, ..., -5.3085e-08, + -3.6368e-07, -2.1746e-07], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 9.3132e-10], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + [ 1.3970e-08, 1.4435e-08, 4.6566e-10, ..., 8.8476e-09, + 3.2596e-09, 3.2596e-09], + [ 4.1910e-09, -3.3528e-08, -1.3970e-09, ..., 3.2596e-09, + -3.2596e-09, -5.1223e-09]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0171, -0.0362, -0.0017, -0.0132, -0.0322, 0.0041, 0.0237, -0.0179, + 0.0500, -0.0064], device='cuda:0'), grad: tensor([-6.8173e-07, -2.1886e-08, 1.0245e-08, 3.1292e-07, 1.2107e-08, + 4.7032e-08, 4.4936e-07, 2.4680e-08, 1.7229e-07, -3.1246e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 250.80, cls_loss 0.0012 cls_loss_mapping 0.0012 cls_loss_causal 0.4917 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.08 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.1311, -0.2821, -0.0887, ..., -0.0741, 0.1848, 0.1919], + [-0.2534, -0.2200, -0.0841, ..., -0.2025, -0.2550, -0.1595], + [-0.0651, -0.1944, 0.1608, ..., -0.2434, 0.2847, 0.1321], + ..., + [-0.1974, 0.1130, 0.0262, ..., 0.2269, -0.2477, -0.3172], + [-0.3226, 0.0750, -0.1603, ..., 0.0709, -0.1281, -0.2408], + [-0.0480, -0.1660, -0.0882, ..., -0.1493, -0.0630, -0.2414]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -4.6566e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0268e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 8.0280e-07, + -1.3970e-09, -4.6566e-10], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + 1.3970e-09, 4.6566e-10], + [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., -4.6119e-06, + 1.3970e-09, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 6.0536e-09, 3.2596e-09]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0171, -0.0362, -0.0018, -0.0131, -0.0324, 0.0040, 0.0239, -0.0183, + 0.0496, -0.0056], device='cuda:0'), grad: tensor([ 2.7940e-09, -3.8650e-08, 2.3097e-06, 6.5193e-09, 2.6543e-08, + 9.6187e-06, 8.5123e-07, 5.4948e-08, -1.2830e-05, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 250.82, cls_loss 0.0017 cls_loss_mapping 0.0013 cls_loss_causal 0.4533 re_mapping 0.0037 re_causal 0.0097 /// teacc 98.97 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.1312, -0.2824, -0.0887, ..., -0.0741, 0.1848, 0.1919], + [-0.2562, -0.2202, -0.0842, ..., -0.2026, -0.2551, -0.1597], + [-0.0637, -0.1944, 0.1608, ..., -0.2433, 0.2852, 0.1326], + ..., + [-0.1982, 0.1132, 0.0262, ..., 0.2270, -0.2486, -0.3183], + [-0.3233, 0.0749, -0.1604, ..., 0.0709, -0.1286, -0.2414], + [-0.0489, -0.1663, -0.0882, ..., -0.1494, -0.0631, -0.2420]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.7183e-07, 9.7323e-08], + [ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 2.7940e-09, 1.3970e-09], + ..., + [ 0.0000e+00, -8.8476e-09, 0.0000e+00, ..., -1.4901e-08, + 1.3970e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 6.5193e-09, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0172, -0.0360, -0.0011, -0.0132, -0.0324, 0.0045, 0.0236, -0.0189, + 0.0494, -0.0055], device='cuda:0'), grad: tensor([ 3.3295e-07, -3.0734e-08, 8.8476e-09, 5.1223e-09, -2.2352e-08, + 2.7940e-09, -3.0408e-07, -1.2107e-08, 1.3970e-09, 2.4680e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 250.95, cls_loss 0.0017 cls_loss_mapping 0.0016 cls_loss_causal 0.4845 re_mapping 0.0036 re_causal 0.0098 /// teacc 99.00 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.1313, -0.2835, -0.0887, ..., -0.0751, 0.1842, 0.1921], + [-0.2565, -0.2203, -0.0842, ..., -0.2026, -0.2552, -0.1603], + [-0.0635, -0.1943, 0.1608, ..., -0.2434, 0.2858, 0.1333], + ..., + [-0.1988, 0.1136, 0.0262, ..., 0.2273, -0.2496, -0.3192], + [-0.3237, 0.0749, -0.1604, ..., 0.0725, -0.1238, -0.2386], + [-0.0491, -0.1668, -0.0882, ..., -0.1497, -0.0624, -0.2424]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.3970e-09, -9.3132e-10], + [ 0.0000e+00, 1.3392e-06, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.4249e-07, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, -1.5963e-06, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., -3.3528e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.2596e-09, 0.0000e+00, ..., 6.0536e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0187, -0.0360, -0.0007, -0.0133, -0.0323, 0.0041, 0.0230, -0.0188, + 0.0530, -0.0054], device='cuda:0'), grad: tensor([ 1.1642e-08, 1.2740e-05, 1.3625e-06, 1.0543e-06, 3.2596e-09, + 6.9384e-08, 1.4901e-08, -1.5177e-05, -2.8778e-07, 1.9837e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 250.81, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4749 re_mapping 0.0038 re_causal 0.0101 /// teacc 98.95 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.1313, -0.2851, -0.0887, ..., -0.0761, 0.1848, 0.1929], + [-0.2567, -0.2224, -0.0798, ..., -0.2016, -0.2553, -0.1604], + [-0.0636, -0.1952, 0.1608, ..., -0.2442, 0.2857, 0.1329], + ..., + [-0.1987, 0.1180, 0.0258, ..., 0.2302, -0.2498, -0.3194], + [-0.3259, 0.0748, -0.1604, ..., 0.0723, -0.1240, -0.2389], + [-0.0494, -0.1706, -0.0883, ..., -0.1535, -0.0628, -0.2438]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -5.7276e-08, -4.1444e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-09, 4.6566e-09], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 5.0757e-08, 3.7253e-08]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0184, -0.0353, -0.0009, -0.0135, -0.0324, 0.0036, 0.0235, -0.0163, + 0.0526, -0.0088], device='cuda:0'), grad: tensor([-1.1828e-07, -5.0943e-07, 3.6787e-08, 4.6566e-10, -5.5414e-08, + -1.3504e-08, 6.9849e-09, 2.2026e-07, 1.2759e-07, 3.1758e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 250.51, cls_loss 0.0012 cls_loss_mapping 0.0013 cls_loss_causal 0.4410 re_mapping 0.0038 re_causal 0.0103 /// teacc 99.06 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.1314, -0.2858, -0.0887, ..., -0.0770, 0.1851, 0.1932], + [-0.2567, -0.2239, -0.0804, ..., -0.2032, -0.2553, -0.1605], + [-0.0636, -0.1954, 0.1608, ..., -0.2447, 0.2858, 0.1330], + ..., + [-0.1989, 0.1193, 0.0259, ..., 0.2314, -0.2503, -0.3200], + [-0.3274, 0.0748, -0.1605, ..., 0.0726, -0.1238, -0.2395], + [-0.0498, -0.1711, -0.0883, ..., -0.1540, -0.0629, -0.2443]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -3.2596e-08, -2.6077e-08], + [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + 5.1223e-09, 3.7253e-09], + ..., + [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 6.0536e-09, + 0.0000e+00, 0.0000e+00], + [-2.6748e-06, -1.9744e-07, 0.0000e+00, ..., -4.3772e-06, + 7.9162e-09, 6.5193e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 6.0536e-09, 4.6566e-09]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0183, -0.0358, -0.0009, -0.0137, -0.0323, 0.0035, 0.0236, -0.0156, + 0.0528, -0.0091], device='cuda:0'), grad: tensor([-6.1467e-08, -2.3805e-06, 1.3327e-06, 1.2247e-07, -1.4901e-07, + 5.2378e-06, 9.6858e-06, 1.1409e-06, -1.4968e-05, 1.9558e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 250.28, cls_loss 0.0015 cls_loss_mapping 0.0015 cls_loss_causal 0.4550 re_mapping 0.0040 re_causal 0.0102 /// teacc 99.03 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.1314, -0.2856, -0.0887, ..., -0.0773, 0.1862, 0.1946], + [-0.2567, -0.2239, -0.0804, ..., -0.2033, -0.2554, -0.1607], + [-0.0636, -0.1965, 0.1608, ..., -0.2457, 0.2859, 0.1329], + ..., + [-0.1997, 0.1195, 0.0259, ..., 0.2317, -0.2505, -0.3202], + [-0.3273, 0.0745, -0.1629, ..., 0.0727, -0.1238, -0.2396], + [-0.0508, -0.1712, -0.0882, ..., -0.1542, -0.0631, -0.2449]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.7229e-08, -1.5367e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.6566e-09, 1.3970e-08, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 4.6566e-10, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.7695e-08, 1.5367e-08]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0175, -0.0366, -0.0010, -0.0136, -0.0324, 0.0032, 0.0235, -0.0155, + 0.0528, -0.0084], device='cuda:0'), grad: tensor([-4.0978e-08, 9.3132e-10, 2.7940e-09, -2.7940e-08, -2.7940e-09, + -8.8476e-09, 2.7940e-09, 1.8161e-08, 8.8476e-09, 4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 250.45, cls_loss 0.0017 cls_loss_mapping 0.0016 cls_loss_causal 0.4816 re_mapping 0.0039 re_causal 0.0100 /// teacc 98.95 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.1317, -0.2862, -0.0887, ..., -0.0780, 0.1865, 0.1949], + [-0.2575, -0.2241, -0.0804, ..., -0.2034, -0.2556, -0.1617], + [-0.0637, -0.1993, 0.1608, ..., -0.2493, 0.2856, 0.1321], + ..., + [-0.2004, 0.1196, 0.0259, ..., 0.2323, -0.2507, -0.3193], + [-0.3284, 0.0744, -0.1629, ..., 0.0729, -0.1239, -0.2398], + [-0.0509, -0.1713, -0.0882, ..., -0.1544, -0.0632, -0.2452]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.1642e-08, -7.9162e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 2.3283e-09, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-09, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0174, -0.0364, -0.0024, -0.0131, -0.0329, 0.0030, 0.0237, -0.0155, + 0.0529, -0.0082], device='cuda:0'), grad: tensor([-2.0489e-08, 1.0245e-08, 1.8626e-09, 4.1910e-09, -6.6124e-08, + 7.0315e-08, -6.6590e-08, 2.3283e-09, 1.1642e-08, 5.6345e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 250.31, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4776 re_mapping 0.0038 re_causal 0.0102 /// teacc 99.02 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.1322, -0.2844, -0.0887, ..., -0.0782, 0.1866, 0.1953], + [-0.2580, -0.2242, -0.0804, ..., -0.2036, -0.2557, -0.1623], + [-0.0634, -0.1994, 0.1608, ..., -0.2495, 0.2857, 0.1321], + ..., + [-0.2006, 0.1197, 0.0259, ..., 0.2326, -0.2516, -0.3195], + [-0.3292, 0.0739, -0.1629, ..., 0.0728, -0.1235, -0.2397], + [-0.0489, -0.1713, -0.0882, ..., -0.1545, -0.0633, -0.2454]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.6322e-08, -2.4214e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + -1.3970e-09, -4.6566e-10], + ..., + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., -1.3970e-09, + 1.3970e-09, 4.6566e-10], + [ 6.9849e-09, 9.3132e-10, 0.0000e+00, ..., -3.1665e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 1.3970e-09]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0174, -0.0365, -0.0022, -0.0134, -0.0324, 0.0030, 0.0238, -0.0156, + 0.0533, -0.0084], device='cuda:0'), grad: tensor([-6.2399e-08, 5.8673e-08, 3.6322e-08, 1.3504e-08, -2.0070e-07, + 4.6566e-08, 6.8918e-08, 9.3132e-09, -4.9826e-08, 9.3132e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 250.19, cls_loss 0.0018 cls_loss_mapping 0.0014 cls_loss_causal 0.4658 re_mapping 0.0036 re_causal 0.0095 /// teacc 99.09 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.1323, -0.2841, -0.0887, ..., -0.0772, 0.1868, 0.1955], + [-0.2594, -0.2243, -0.0804, ..., -0.2039, -0.2558, -0.1624], + [-0.0637, -0.2012, 0.1608, ..., -0.2517, 0.2862, 0.1330], + ..., + [-0.2027, 0.1201, 0.0259, ..., 0.2334, -0.2525, -0.3206], + [-0.3273, 0.0729, -0.1629, ..., 0.0732, -0.1234, -0.2399], + [-0.0494, -0.1714, -0.0882, ..., -0.1546, -0.0634, -0.2458]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -5.5879e-09, 0.0000e+00, ..., -1.2107e-08, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 7.4506e-09, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0173, -0.0363, -0.0029, -0.0124, -0.0325, 0.0022, 0.0237, -0.0158, + 0.0554, -0.0085], device='cuda:0'), grad: tensor([ 1.3970e-09, 1.1176e-08, 1.8626e-09, -5.1223e-09, 4.1910e-09, + 1.0710e-08, 4.6566e-10, -2.5611e-08, -1.2107e-08, 2.0023e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 250.26, cls_loss 0.0016 cls_loss_mapping 0.0013 cls_loss_causal 0.4651 re_mapping 0.0037 re_causal 0.0101 /// teacc 99.00 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.1326, -0.2842, -0.0887, ..., -0.0770, 0.1895, 0.1983], + [-0.2595, -0.2246, -0.0784, ..., -0.2041, -0.2560, -0.1625], + [-0.0638, -0.2017, 0.1609, ..., -0.2525, 0.2864, 0.1330], + ..., + [-0.2033, 0.1204, 0.0252, ..., 0.2338, -0.2531, -0.3209], + [-0.3280, 0.0723, -0.1630, ..., 0.0729, -0.1236, -0.2401], + [-0.0513, -0.1714, -0.0882, ..., -0.1547, -0.0666, -0.2506]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 6.2399e-08, 0.0000e+00, ..., 2.0862e-07, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.9912e-06, 0.0000e+00, ..., 6.6012e-06, + 9.7789e-09, 4.6566e-09], + ..., + [ 0.0000e+00, -2.0564e-06, 0.0000e+00, ..., -6.8285e-06, + -1.0245e-08, -5.1223e-09], + [ 4.6566e-10, 3.7253e-09, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -4.6566e-10, ..., 6.0536e-09, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0151, -0.0364, -0.0028, -0.0125, -0.0328, 0.0022, 0.0240, -0.0157, + 0.0551, -0.0094], device='cuda:0'), grad: tensor([ 1.6298e-08, 4.2003e-07, 1.3255e-05, -4.6566e-09, -2.7940e-08, + 1.3970e-09, 9.3132e-10, -1.3687e-05, 1.7229e-08, 2.7474e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 250.73, cls_loss 0.0015 cls_loss_mapping 0.0016 cls_loss_causal 0.4958 re_mapping 0.0038 re_causal 0.0104 /// teacc 98.99 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.1341, -0.2843, -0.0888, ..., -0.0771, 0.1897, 0.1984], + [-0.2597, -0.2248, -0.0776, ..., -0.2045, -0.2561, -0.1623], + [-0.0655, -0.2036, 0.1608, ..., -0.2540, 0.2863, 0.1323], + ..., + [-0.2039, 0.1213, 0.0249, ..., 0.2349, -0.2537, -0.3212], + [-0.3279, 0.0699, -0.1630, ..., 0.0721, -0.1233, -0.2401], + [-0.0513, -0.1714, -0.0882, ..., -0.1548, -0.0671, -0.2511]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 4.1910e-09, + -4.6566e-10, -1.8626e-09], + [ 5.3551e-09, 2.5844e-08, 0.0000e+00, ..., 1.2107e-07, + 2.8638e-08, 0.0000e+00], + [ 2.3283e-10, 1.0477e-08, 0.0000e+00, ..., 1.6065e-08, + -7.9162e-09, -3.9581e-09], + ..., + [ 6.9849e-10, -4.4005e-08, 0.0000e+00, ..., -9.1735e-08, + 1.1642e-09, 2.3283e-10], + [ 6.2166e-08, 2.5611e-09, 0.0000e+00, ..., -1.4016e-07, + -7.3342e-08, 4.1910e-09], + [ 2.0955e-09, 6.9849e-09, 0.0000e+00, ..., 1.3271e-08, + 4.8894e-09, 3.2596e-09]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0151, -0.0364, -0.0034, -0.0127, -0.0349, 0.0020, 0.0244, -0.0154, + 0.0546, -0.0086], device='cuda:0'), grad: tensor([ 2.1420e-08, 5.2061e-07, 2.9337e-08, 4.1281e-07, -3.7486e-08, + 3.5856e-08, 7.6601e-08, -2.1257e-07, -9.3039e-07, 9.4064e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 250.65, cls_loss 0.0013 cls_loss_mapping 0.0013 cls_loss_causal 0.4490 re_mapping 0.0036 re_causal 0.0098 /// teacc 99.00 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.1341, -0.2845, -0.0888, ..., -0.0775, 0.1897, 0.1985], + [-0.2601, -0.2262, -0.0774, ..., -0.2057, -0.2566, -0.1625], + [-0.0656, -0.2045, 0.1608, ..., -0.2570, 0.2866, 0.1322], + ..., + [-0.2041, 0.1221, 0.0249, ..., 0.2357, -0.2521, -0.3210], + [-0.3299, 0.0698, -0.1630, ..., 0.0721, -0.1237, -0.2406], + [-0.0513, -0.1715, -0.0882, ..., -0.1549, -0.0670, -0.2511]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 3.2596e-09, + 1.5600e-08, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.0745e-08, + 1.0012e-08, -9.3132e-10], + [ 0.0000e+00, 3.2596e-09, 0.0000e+00, ..., -8.6147e-09, + -1.8440e-07, -1.1292e-07], + ..., + [ 0.0000e+00, -4.1910e-09, 0.0000e+00, ..., 1.6997e-08, + 1.9092e-08, 1.5832e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 6.0536e-09, + 5.4948e-08, 3.3993e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.7462e-08, + 9.7789e-09, 9.7789e-09]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0152, -0.0371, -0.0036, -0.0131, -0.0345, 0.0020, 0.0248, -0.0149, + 0.0546, -0.0088], device='cuda:0'), grad: tensor([ 8.8708e-08, -3.3295e-07, -7.2457e-07, 6.5193e-09, 2.3446e-07, + 3.1199e-08, 6.2631e-08, 2.8801e-07, 2.2841e-07, 1.2922e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 250.60, cls_loss 0.0014 cls_loss_mapping 0.0010 cls_loss_causal 0.4627 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.03 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.1343, -0.2848, -0.0888, ..., -0.0775, 0.1902, 0.1992], + [-0.2602, -0.2263, -0.0769, ..., -0.2049, -0.2566, -0.1627], + [-0.0656, -0.2047, 0.1608, ..., -0.2574, 0.2865, 0.1318], + ..., + [-0.2047, 0.1223, 0.0248, ..., 0.2358, -0.2526, -0.3217], + [-0.3306, 0.0696, -0.1630, ..., 0.0727, -0.1240, -0.2408], + [-0.0515, -0.1716, -0.0883, ..., -0.1553, -0.0672, -0.2514]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 6.9849e-10, 0.0000e+00, ..., 6.9849e-10, + -9.0804e-09, -5.8208e-09], + [ 9.3132e-10, 2.6310e-08, 0.0000e+00, ..., 7.3807e-08, + 1.1642e-09, 4.6566e-09], + [-4.6566e-10, 5.3318e-08, 0.0000e+00, ..., 1.5623e-07, + -3.4925e-09, 8.3819e-09], + ..., + [ 3.7253e-09, -7.6601e-08, 0.0000e+00, ..., -2.3423e-07, + 3.2596e-09, -1.2806e-08], + [ 6.9849e-10, 9.3132e-10, 0.0000e+00, ..., -0.0000e+00, + -6.9849e-10, 2.3283e-10], + [ 2.0955e-09, 3.2596e-09, 0.0000e+00, ..., 2.0955e-09, + 5.3551e-09, 3.2596e-09]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0149, -0.0362, -0.0037, -0.0132, -0.0344, 0.0017, 0.0249, -0.0154, + 0.0549, -0.0090], device='cuda:0'), grad: tensor([ 1.1874e-08, 1.6205e-07, 3.2200e-07, -2.0955e-08, 1.0221e-07, + 2.7940e-09, 6.2864e-09, -4.6892e-07, 2.5146e-08, -1.3434e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 250.53, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4871 re_mapping 0.0039 re_causal 0.0103 /// teacc 99.10 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.1354, -0.2849, -0.0888, ..., -0.0777, 0.1903, 0.1992], + [-0.2604, -0.2267, -0.0767, ..., -0.2065, -0.2567, -0.1627], + [-0.0657, -0.2049, 0.1608, ..., -0.2577, 0.2868, 0.1324], + ..., + [-0.2053, 0.1226, 0.0247, ..., 0.2368, -0.2535, -0.3239], + [-0.3342, 0.0694, -0.1630, ..., 0.0710, -0.1255, -0.2413], + [-0.0515, -0.1716, -0.0883, ..., -0.1554, -0.0676, -0.2517]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.8417e-08, -2.4913e-08], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 2.3283e-10, 0.0000e+00, ..., 1.4901e-08, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 7.6834e-09, ..., 8.6147e-09, + 2.0955e-09, 1.3970e-09]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0148, -0.0366, -0.0039, -0.0129, -0.0344, 0.0027, 0.0245, -0.0151, + 0.0533, -0.0090], device='cuda:0'), grad: tensor([-5.5879e-08, 2.0955e-09, 4.6566e-10, 4.6566e-10, -5.6345e-08, + -6.9616e-08, 9.9884e-08, 3.0268e-09, 2.5844e-08, 5.7509e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 250.61, cls_loss 0.0014 cls_loss_mapping 0.0010 cls_loss_causal 0.4825 re_mapping 0.0038 re_causal 0.0102 /// teacc 99.04 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.1329, -0.2851, -0.0888, ..., -0.0780, 0.1923, 0.2018], + [-0.2605, -0.2269, -0.0767, ..., -0.2067, -0.2568, -0.1633], + [-0.0658, -0.2081, 0.1608, ..., -0.2612, 0.2867, 0.1314], + ..., + [-0.2060, 0.1233, 0.0247, ..., 0.2382, -0.2528, -0.3219], + [-0.3351, 0.0682, -0.1630, ..., 0.0700, -0.1259, -0.2416], + [-0.0514, -0.1716, -0.0883, ..., -0.1556, -0.0672, -0.2518]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 6.9849e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0133, -0.0368, -0.0049, -0.0129, -0.0346, 0.0031, 0.0229, -0.0145, + 0.0524, -0.0087], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.8626e-09, 4.6566e-10, 2.5611e-09, 4.4238e-09, + 4.6566e-09, 6.9849e-10, 1.2573e-08, -6.2864e-09, -1.3737e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 250.41, cls_loss 0.0014 cls_loss_mapping 0.0013 cls_loss_causal 0.4781 re_mapping 0.0039 re_causal 0.0104 /// teacc 99.00 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.1331, -0.2857, -0.0888, ..., -0.0785, 0.1928, 0.2026], + [-0.2605, -0.2270, -0.0767, ..., -0.2068, -0.2569, -0.1640], + [-0.0659, -0.2082, 0.1608, ..., -0.2617, 0.2868, 0.1311], + ..., + [-0.2063, 0.1233, 0.0247, ..., 0.2386, -0.2537, -0.3227], + [-0.3356, 0.0680, -0.1630, ..., 0.0699, -0.1263, -0.2422], + [-0.0523, -0.1718, -0.0883, ..., -0.1559, -0.0674, -0.2522]], + device='cuda:0'), grad: tensor([[-9.0804e-09, 6.9849e-10, 0.0000e+00, ..., 5.1223e-09, + -1.3085e-07, -8.6613e-08], + [ 2.3283e-10, 1.6065e-08, 0.0000e+00, ..., 1.1735e-07, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 4.6333e-08, 0.0000e+00, ..., 3.4133e-07, + 2.3283e-10, 2.3283e-10], + ..., + [ 2.3283e-10, -1.8021e-07, 0.0000e+00, ..., -1.3141e-06, + 0.0000e+00, 0.0000e+00], + [ 1.6298e-09, 6.0536e-09, 0.0000e+00, ..., 2.8405e-08, + 4.6566e-10, 6.9849e-10], + [ 2.3283e-10, 1.1479e-07, 0.0000e+00, ..., 8.2050e-07, + 9.3132e-10, 6.9849e-10]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0128, -0.0368, -0.0050, -0.0135, -0.0343, 0.0034, 0.0227, -0.0145, + 0.0522, -0.0090], device='cuda:0'), grad: tensor([-1.6345e-07, 2.8289e-07, 8.1584e-07, -1.4203e-08, 8.3819e-09, + 4.8894e-08, 1.1595e-07, -3.1237e-06, 7.7765e-08, 1.9558e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 250.39, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4738 re_mapping 0.0038 re_causal 0.0102 /// teacc 98.96 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.1332, -0.2858, -0.0888, ..., -0.0786, 0.1928, 0.2026], + [-0.2606, -0.2270, -0.0767, ..., -0.2070, -0.2569, -0.1641], + [-0.0662, -0.2090, 0.1608, ..., -0.2622, 0.2869, 0.1312], + ..., + [-0.2044, 0.1235, 0.0247, ..., 0.2394, -0.2538, -0.3228], + [-0.3364, 0.0680, -0.1631, ..., 0.0697, -0.1264, -0.2423], + [-0.0517, -0.1720, -0.0883, ..., -0.1564, -0.0674, -0.2523]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -1.6764e-08, -1.0245e-08], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 2.5611e-09, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 1.1642e-09, 6.9849e-10], + ..., + [ 0.0000e+00, 1.6103e-06, 0.0000e+00, ..., 3.6806e-06, + 2.3283e-10, 2.3283e-10], + [ 2.3283e-10, -1.6727e-06, 0.0000e+00, ..., -3.8296e-06, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 1.8626e-09, + 1.1176e-08, 6.7521e-09]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0129, -0.0369, -0.0051, -0.0134, -0.0349, 0.0009, 0.0252, -0.0143, + 0.0521, -0.0089], device='cuda:0'), grad: tensor([-2.9104e-08, 6.0536e-09, 4.8894e-09, 1.9628e-07, -1.0384e-07, + 1.7928e-08, -2.0955e-09, 5.9642e-06, -6.1654e-06, 9.5693e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 250.37, cls_loss 0.0022 cls_loss_mapping 0.0018 cls_loss_causal 0.4499 re_mapping 0.0039 re_causal 0.0096 /// teacc 98.97 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.1334, -0.2866, -0.0888, ..., -0.0799, 0.1929, 0.2027], + [-0.2609, -0.2283, -0.0757, ..., -0.2074, -0.2573, -0.1645], + [-0.0663, -0.2091, 0.1610, ..., -0.2622, 0.2875, 0.1316], + ..., + [-0.2050, 0.1242, 0.0247, ..., 0.2402, -0.2548, -0.3232], + [-0.3370, 0.0696, -0.1631, ..., 0.0701, -0.1266, -0.2405], + [-0.0522, -0.1728, -0.0887, ..., -0.1571, -0.0677, -0.2535]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.1409e-08, -6.7521e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 3.9581e-09, + 2.3283e-10, 0.0000e+00], + ..., + [ 2.3283e-10, 1.6298e-09, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -5.4017e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.0804e-07, + 9.7789e-09, 5.8208e-09]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0131, -0.0377, -0.0048, -0.0128, -0.0346, 0.0006, 0.0254, -0.0136, + 0.0552, -0.0112], device='cuda:0'), grad: tensor([-1.3737e-08, 3.0501e-08, 2.1420e-08, 6.5193e-09, 1.0245e-08, + 5.7509e-08, 2.1886e-08, 1.7695e-08, -3.7737e-06, 3.6303e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 250.59, cls_loss 0.0017 cls_loss_mapping 0.0018 cls_loss_causal 0.4628 re_mapping 0.0038 re_causal 0.0099 /// teacc 98.99 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.1336, -0.2868, -0.0888, ..., -0.0831, 0.1935, 0.2034], + [-0.2610, -0.2284, -0.0725, ..., -0.2047, -0.2574, -0.1647], + [-0.0663, -0.2092, 0.1617, ..., -0.2624, 0.2884, 0.1324], + ..., + [-0.2057, 0.1243, 0.0215, ..., 0.2376, -0.2551, -0.3237], + [-0.3412, 0.0695, -0.1632, ..., 0.0715, -0.1248, -0.2425], + [-0.0531, -0.1728, -0.0887, ..., -0.1574, -0.0687, -0.2548]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4971e-08, -4.9593e-08], + [ 6.9849e-10, 1.5832e-08, 0.0000e+00, ..., 1.5064e-07, + 4.4238e-09, 2.7940e-09], + [ 0.0000e+00, -2.3283e-10, 0.0000e+00, ..., -5.1223e-09, + -5.3551e-09, -2.3283e-09], + ..., + [ 6.9849e-10, -1.7695e-08, 0.0000e+00, ..., -1.7346e-07, + 1.6298e-09, 9.3132e-10], + [ 3.5623e-08, 2.3283e-10, 0.0000e+00, ..., 4.1677e-08, + 9.3132e-09, 5.3551e-09], + [-9.1270e-08, 1.6298e-09, -2.3283e-10, ..., 1.6531e-08, + 4.7497e-08, 3.1665e-08]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0127, -0.0353, -0.0044, -0.0131, -0.0339, 0.0007, 0.0250, -0.0158, + 0.0556, -0.0118], device='cuda:0'), grad: tensor([-2.2235e-07, 1.1805e-07, -1.6997e-08, 4.0443e-07, 7.4506e-09, + 1.2573e-08, 4.6333e-08, -2.8359e-07, 3.0361e-07, -3.6648e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 251.15, cls_loss 0.0017 cls_loss_mapping 0.0014 cls_loss_causal 0.4993 re_mapping 0.0037 re_causal 0.0100 /// teacc 99.00 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.1364, -0.2870, -0.0889, ..., -0.0861, 0.1908, 0.2009], + [-0.2620, -0.2286, -0.0724, ..., -0.2048, -0.2583, -0.1655], + [-0.0658, -0.2089, 0.1618, ..., -0.2622, 0.2894, 0.1333], + ..., + [-0.2059, 0.1245, 0.0213, ..., 0.2376, -0.2552, -0.3239], + [-0.3429, 0.0693, -0.1634, ..., 0.0704, -0.1251, -0.2438], + [-0.0533, -0.1729, -0.0878, ..., -0.1573, -0.0687, -0.2549]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8394e-08, + -6.7288e-08, -3.1665e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 3.0268e-09, + 6.9849e-09, 3.0268e-09], + [ 0.0000e+00, -1.6298e-09, 0.0000e+00, ..., -1.0245e-08, + -2.3516e-08, -1.0943e-08], + ..., + [ 2.3283e-10, 1.6298e-09, 0.0000e+00, ..., 1.3737e-08, + 3.3062e-08, 1.3737e-08], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., -1.0943e-08, + 1.3970e-09, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 6.9849e-09, + 2.0256e-08, 8.1491e-09]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0154, -0.0355, -0.0034, -0.0134, -0.0345, -0.0029, 0.0307, -0.0159, + 0.0548, -0.0114], device='cuda:0'), grad: tensor([-1.4808e-07, 3.4226e-08, -8.5915e-08, 1.8510e-07, -3.0734e-08, + -1.2177e-07, 5.8440e-08, 1.0966e-07, -3.6089e-08, 4.4238e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 250.69, cls_loss 0.0011 cls_loss_mapping 0.0011 cls_loss_causal 0.4491 re_mapping 0.0038 re_causal 0.0102 /// teacc 99.02 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.1367, -0.2872, -0.0890, ..., -0.0863, 0.1906, 0.2007], + [-0.2629, -0.2292, -0.0725, ..., -0.2050, -0.2583, -0.1662], + [-0.0655, -0.2114, 0.1618, ..., -0.2639, 0.2895, 0.1322], + ..., + [-0.2061, 0.1254, 0.0213, ..., 0.2382, -0.2551, -0.3218], + [-0.3431, 0.0691, -0.1634, ..., 0.0703, -0.1252, -0.2438], + [-0.0541, -0.1731, -0.0878, ..., -0.1577, -0.0688, -0.2549]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.2573e-08, 9.3132e-10], + [ 2.3283e-10, 9.3132e-09, 0.0000e+00, ..., 9.3132e-10, + 8.6147e-09, 6.9849e-10], + [ 0.0000e+00, -8.7963e-07, 0.0000e+00, ..., 0.0000e+00, + -1.0971e-06, -1.0873e-07], + ..., + [ 0.0000e+00, 7.1572e-07, 0.0000e+00, ..., 4.6566e-10, + 6.3749e-07, 5.9605e-08], + [ 4.6566e-10, 8.6846e-08, 0.0000e+00, ..., -6.9849e-10, + 8.2189e-08, 8.1491e-09], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-09, 4.6566e-10]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0156, -0.0356, -0.0037, -0.0137, -0.0343, -0.0028, 0.0308, -0.0155, + 0.0546, -0.0116], device='cuda:0'), grad: tensor([ 3.1665e-08, 3.4692e-08, -4.8093e-06, 2.9197e-07, -1.7020e-07, + 1.1642e-09, 5.1223e-07, 3.4999e-06, 4.2818e-07, 1.8440e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 250.30, cls_loss 0.0014 cls_loss_mapping 0.0012 cls_loss_causal 0.4691 re_mapping 0.0036 re_causal 0.0101 /// teacc 99.08 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.1367, -0.2874, -0.0890, ..., -0.0865, 0.1907, 0.2008], + [-0.2635, -0.2319, -0.0724, ..., -0.2053, -0.2597, -0.1679], + [-0.0655, -0.2118, 0.1628, ..., -0.2644, 0.2916, 0.1337], + ..., + [-0.2066, 0.1271, 0.0213, ..., 0.2385, -0.2566, -0.3236], + [-0.3438, 0.0689, -0.1634, ..., 0.0698, -0.1256, -0.2440], + [-0.0544, -0.1731, -0.0879, ..., -0.1577, -0.0689, -0.2550]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 9.7789e-09, + 6.9849e-10, 1.6298e-09], + [ 4.6566e-10, 2.0955e-09, 0.0000e+00, ..., 1.5600e-08, + 2.7940e-09, -5.3551e-08], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 3.3993e-08, + 7.4506e-09, 2.7008e-08], + ..., + [ 4.6566e-10, -5.5879e-09, 0.0000e+00, ..., -8.6147e-09, + 6.9849e-10, 4.0280e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., -7.2829e-06, + -1.5404e-06, -2.0787e-06], + [ 9.3132e-09, 1.9791e-08, 0.0000e+00, ..., 1.0477e-08, + 3.0268e-09, 2.7940e-09]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0156, -0.0362, -0.0027, -0.0163, -0.0339, -0.0012, 0.0302, -0.0149, + 0.0541, -0.0117], device='cuda:0'), grad: tensor([ 4.3306e-08, -2.2892e-06, 8.2515e-07, -2.6543e-08, 1.3039e-08, + 2.7213e-06, 2.6658e-05, 1.5935e-06, -2.9609e-05, 9.9884e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 250.48, cls_loss 0.0018 cls_loss_mapping 0.0017 cls_loss_causal 0.4650 re_mapping 0.0038 re_causal 0.0097 /// teacc 99.04 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.1377, -0.2871, -0.0890, ..., -0.0892, 0.1888, 0.1991], + [-0.2644, -0.2325, -0.0724, ..., -0.2059, -0.2599, -0.1677], + [-0.0656, -0.2116, 0.1628, ..., -0.2652, 0.2922, 0.1344], + ..., + [-0.2071, 0.1285, 0.0213, ..., 0.2401, -0.2583, -0.3248], + [-0.3440, 0.0685, -0.1634, ..., 0.0696, -0.1256, -0.2441], + [-0.0548, -0.1748, -0.0879, ..., -0.1604, -0.0690, -0.2551]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 2.7940e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.5832e-08, 0.0000e+00, ..., -6.0536e-09, + -8.3819e-09, -4.6566e-10], + ..., + [ 0.0000e+00, 5.1223e-09, 0.0000e+00, ..., 3.2596e-09, + 3.7253e-09, 4.6566e-10], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 2.3283e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0175, -0.0363, -0.0024, -0.0159, -0.0334, -0.0013, 0.0316, -0.0141, + 0.0538, -0.0128], device='cuda:0'), grad: tensor([ 1.1176e-08, 6.9849e-09, -9.8255e-08, 6.2399e-08, -1.0431e-07, + 9.3132e-09, 1.1176e-08, 2.9337e-08, -2.1886e-08, 1.0431e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 250.34, cls_loss 0.0013 cls_loss_mapping 0.0011 cls_loss_causal 0.4639 re_mapping 0.0037 re_causal 0.0101 /// teacc 99.11 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.1377, -0.2872, -0.0892, ..., -0.0891, 0.1890, 0.1993], + [-0.2647, -0.2339, -0.0724, ..., -0.2060, -0.2614, -0.1701], + [-0.0658, -0.2119, 0.1643, ..., -0.2653, 0.2929, 0.1349], + ..., + [-0.2072, 0.1295, 0.0213, ..., 0.2404, -0.2587, -0.3250], + [-0.3443, 0.0684, -0.1639, ..., 0.0696, -0.1257, -0.2441], + [-0.0549, -0.1750, -0.0879, ..., -0.1606, -0.0690, -0.2551]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.3504e-08, 0.0000e+00, ..., -1.8626e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 2.7940e-09, 0.0000e+00, ..., 8.3819e-09, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 7.9162e-09, 0.0000e+00, ..., 1.3039e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0172, -0.0370, -0.0020, -0.0162, -0.0333, -0.0013, 0.0316, -0.0137, + 0.0537, -0.0129], device='cuda:0'), grad: tensor([ 4.1910e-09, 5.1223e-09, 1.8626e-09, -6.5193e-09, -1.3039e-08, + -7.4506e-09, -6.9849e-09, -2.7940e-08, 2.7474e-08, 3.2596e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 250.07, cls_loss 0.0012 cls_loss_mapping 0.0012 cls_loss_causal 0.4674 re_mapping 0.0038 re_causal 0.0101 /// teacc 99.08 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.1377, -0.2883, -0.0892, ..., -0.0891, 0.1890, 0.1994], + [-0.2648, -0.2338, -0.0724, ..., -0.2060, -0.2615, -0.1703], + [-0.0659, -0.2119, 0.1644, ..., -0.2654, 0.2929, 0.1349], + ..., + [-0.2073, 0.1297, 0.0213, ..., 0.2406, -0.2592, -0.3254], + [-0.3444, 0.0684, -0.1639, ..., 0.0696, -0.1259, -0.2441], + [-0.0550, -0.1752, -0.0879, ..., -0.1609, -0.0691, -0.2553]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.5390e-08, -2.0955e-08], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 3.7253e-09, + 4.6566e-10, 0.0000e+00], + [ 6.9849e-09, 1.6298e-08, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.6566e-10, -1.8626e-09, -0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 4.6566e-10], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 9.7789e-09, 6.5193e-09]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0172, -0.0344, -0.0023, -0.0169, -0.0335, -0.0012, 0.0316, -0.0157, + 0.0537, -0.0132], device='cuda:0'), grad: tensor([-6.5193e-08, 4.6566e-09, 3.6322e-08, -3.6787e-08, -2.3283e-09, + 1.3970e-09, 3.9116e-08, -3.2596e-09, 5.1223e-09, 2.7474e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 250.44, cls_loss 0.0012 cls_loss_mapping 0.0012 cls_loss_causal 0.4607 re_mapping 0.0037 re_causal 0.0102 /// teacc 99.07 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.1379, -0.2883, -0.0892, ..., -0.0887, 0.1893, 0.1998], + [-0.2649, -0.2370, -0.0724, ..., -0.2067, -0.2616, -0.1704], + [-0.0659, -0.2132, 0.1644, ..., -0.2665, 0.2930, 0.1349], + ..., + [-0.2083, 0.1329, 0.0213, ..., 0.2414, -0.2595, -0.3254], + [-0.3447, 0.0681, -0.1639, ..., 0.0696, -0.1255, -0.2440], + [-0.0549, -0.1753, -0.0879, ..., -0.1610, -0.0692, -0.2554]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.4727e-08, -4.4238e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.6345e-07, 1.3970e-09], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.4808e-07, 9.3132e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.0536e-09, 3.2596e-09], + [ 8.3819e-09, 1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 3.3528e-08, 2.2817e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 8.8476e-09]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0170, -0.0367, -0.0028, -0.0168, -0.0336, -0.0012, 0.0314, -0.0133, + 0.0537, -0.0132], device='cuda:0'), grad: tensor([-1.3737e-07, 4.6566e-07, -4.2515e-07, -2.8871e-08, 1.0710e-07, + 5.1223e-09, -1.2806e-07, 1.5367e-08, 1.0151e-07, 3.2131e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 250.09, cls_loss 0.0014 cls_loss_mapping 0.0012 cls_loss_causal 0.4496 re_mapping 0.0038 re_causal 0.0098 /// teacc 99.06 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.1380, -0.2884, -0.0892, ..., -0.0887, 0.1894, 0.1998], + [-0.2651, -0.2371, -0.0724, ..., -0.2068, -0.2617, -0.1704], + [-0.0658, -0.2150, 0.1644, ..., -0.2678, 0.2933, 0.1355], + ..., + [-0.2085, 0.1331, 0.0212, ..., 0.2417, -0.2605, -0.3260], + [-0.3468, 0.0677, -0.1640, ..., 0.0683, -0.1264, -0.2466], + [-0.0550, -0.1754, -0.0879, ..., -0.1615, -0.0693, -0.2555]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.7695e-08, -1.3504e-08], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 9.7789e-09, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0170, -0.0368, -0.0036, -0.0168, -0.0336, -0.0011, 0.0321, -0.0132, + 0.0521, -0.0132], device='cuda:0'), grad: tensor([-3.2131e-08, 4.1910e-09, 4.6566e-10, 5.5879e-09, 1.1642e-08, + -4.0047e-08, 3.2131e-08, 1.4901e-07, 1.8626e-09, -1.2852e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 250.17, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4686 re_mapping 0.0037 re_causal 0.0099 /// teacc 99.05 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.1380, -0.2888, -0.0899, ..., -0.0887, 0.1893, 0.1999], + [-0.2651, -0.2371, -0.0724, ..., -0.2069, -0.2617, -0.1705], + [-0.0658, -0.2150, 0.1653, ..., -0.2678, 0.2936, 0.1364], + ..., + [-0.2090, 0.1314, 0.0188, ..., 0.2393, -0.2611, -0.3262], + [-0.3470, 0.0676, -0.1661, ..., 0.0682, -0.1266, -0.2467], + [-0.0556, -0.1724, -0.0848, ..., -0.1585, -0.0691, -0.2555]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-09, 6.9849e-09], + [ 0.0000e+00, 1.3690e-07, 0.0000e+00, ..., 7.5437e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.1002e-08, 0.0000e+00, ..., 3.3993e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.5355e-07, 0.0000e+00, ..., -2.5146e-07, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.1910e-09, 0.0000e+00, ..., 5.1223e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.4156e-07, -4.6566e-10, ..., 7.5437e-08, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0171, -0.0368, -0.0033, -0.0162, -0.0364, -0.0012, 0.0321, -0.0150, + 0.0520, -0.0096], device='cuda:0'), grad: tensor([ 3.3062e-08, 9.0804e-07, 4.0000e-07, 6.1654e-07, 8.2888e-08, + 4.6566e-10, -3.2131e-08, -2.9616e-06, 3.7719e-08, 9.1316e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 250.55, cls_loss 0.0016 cls_loss_mapping 0.0014 cls_loss_causal 0.4581 re_mapping 0.0037 re_causal 0.0096 /// teacc 99.01 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.1380, -0.2894, -0.0901, ..., -0.0888, 0.1893, 0.1999], + [-0.2683, -0.2372, -0.0724, ..., -0.2070, -0.2618, -0.1707], + [-0.0660, -0.2150, 0.1655, ..., -0.2680, 0.2939, 0.1369], + ..., + [-0.2094, 0.1314, 0.0188, ..., 0.2394, -0.2615, -0.3266], + [-0.3469, 0.0686, -0.1664, ..., 0.0687, -0.1267, -0.2468], + [-0.0585, -0.1726, -0.0848, ..., -0.1588, -0.0690, -0.2556]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, -1.4016e-07, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.0536e-09, 4.6566e-10, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -6.9849e-09, 1.1921e-07, ..., -1.1176e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 5.5879e-09, ..., 1.8626e-09, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0172, -0.0366, -0.0070, -0.0164, -0.0372, -0.0006, 0.0321, -0.0150, + 0.0525, -0.0094], device='cuda:0'), grad: tensor([ 6.9849e-09, -1.5497e-06, 2.9802e-08, 2.3749e-08, 1.5330e-06, + -1.6764e-08, 2.7940e-09, 1.3169e-06, 3.2596e-09, -1.3560e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 250.92, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4421 re_mapping 0.0038 re_causal 0.0101 /// teacc 99.07 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.1380, -0.2897, -0.0901, ..., -0.0888, 0.1893, 0.1999], + [-0.2682, -0.2372, -0.0724, ..., -0.2070, -0.2618, -0.1707], + [-0.0664, -0.2177, 0.1655, ..., -0.2683, 0.2946, 0.1379], + ..., + [-0.2097, 0.1317, 0.0188, ..., 0.2395, -0.2631, -0.3271], + [-0.3470, 0.0684, -0.1665, ..., 0.0687, -0.1268, -0.2468], + [-0.0588, -0.1726, -0.0848, ..., -0.1588, -0.0689, -0.2556]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0172, -0.0364, -0.0089, -0.0166, -0.0375, -0.0005, 0.0320, -0.0148, + 0.0524, -0.0092], device='cuda:0'), grad: tensor([-7.4506e-09, 0.0000e+00, 0.0000e+00, 0.0000e+00, -3.7253e-09, + 1.3970e-09, 2.3283e-09, 1.8626e-09, 1.3970e-09, 4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 250.94, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4844 re_mapping 0.0039 re_causal 0.0103 /// teacc 99.08 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.1380, -0.2898, -0.0901, ..., -0.0888, 0.1893, 0.2000], + [-0.2683, -0.2372, -0.0724, ..., -0.2071, -0.2618, -0.1707], + [-0.0664, -0.2179, 0.1655, ..., -0.2686, 0.2947, 0.1380], + ..., + [-0.2094, 0.1318, 0.0188, ..., 0.2396, -0.2632, -0.3271], + [-0.3470, 0.0682, -0.1665, ..., 0.0687, -0.1269, -0.2469], + [-0.0596, -0.1727, -0.0848, ..., -0.1589, -0.0689, -0.2556]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + -1.3970e-09, -1.8626e-09], + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -5.5879e-09, 0.0000e+00, ..., -8.3819e-09, + 0.0000e+00, -0.0000e+00], + [ 4.6566e-09, 5.1223e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 1.3970e-09]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0172, -0.0363, -0.0090, -0.0166, -0.0373, -0.0006, 0.0321, -0.0148, + 0.0526, -0.0093], device='cuda:0'), grad: tensor([-2.3283e-09, 4.1910e-09, 8.8476e-09, 2.3749e-08, 0.0000e+00, + -3.9116e-08, 6.0536e-09, -1.6298e-08, 1.0710e-08, 8.8476e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 250.64, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.4638 re_mapping 0.0035 re_causal 0.0095 /// teacc 99.07 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.1382, -0.2898, -0.0899, ..., -0.0889, 0.1893, 0.2000], + [-0.2683, -0.2373, -0.0724, ..., -0.2082, -0.2619, -0.1708], + [-0.0665, -0.2179, 0.1655, ..., -0.2686, 0.2951, 0.1387], + ..., + [-0.2098, 0.1314, 0.0188, ..., 0.2403, -0.2640, -0.3277], + [-0.3471, 0.0679, -0.1671, ..., 0.0687, -0.1270, -0.2470], + [-0.0597, -0.1728, -0.0848, ..., -0.1590, -0.0690, -0.2557]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3504e-08, 1.3039e-08], + [ 0.0000e+00, 7.9162e-09, 0.0000e+00, ..., 6.0536e-09, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -8.3819e-09, 4.6566e-10], + ..., + [ 0.0000e+00, -3.7253e-08, 0.0000e+00, ..., -2.9337e-08, + 1.3970e-09, 0.0000e+00], + [ 3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 2.3283e-09], + [ 0.0000e+00, 2.8871e-08, 0.0000e+00, ..., 2.2817e-08, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0173, -0.0368, -0.0090, -0.0151, -0.0375, -0.0005, 0.0321, -0.0146, + 0.0525, -0.0094], device='cuda:0'), grad: tensor([ 4.9826e-08, 1.3132e-07, -2.4214e-08, 5.1223e-09, 1.8161e-08, + 3.7486e-07, -4.3819e-07, 3.4459e-08, 2.5146e-08, -1.7462e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 250.74, cls_loss 0.0016 cls_loss_mapping 0.0010 cls_loss_causal 0.4310 re_mapping 0.0036 re_causal 0.0096 /// teacc 99.03 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.1385, -0.2898, -0.0899, ..., -0.0889, 0.1896, 0.2004], + [-0.2684, -0.2374, -0.0708, ..., -0.2072, -0.2620, -0.1709], + [-0.0665, -0.2181, 0.1656, ..., -0.2694, 0.2954, 0.1393], + ..., + [-0.2102, 0.1315, 0.0177, ..., 0.2400, -0.2645, -0.3280], + [-0.3472, 0.0678, -0.1672, ..., 0.0685, -0.1270, -0.2470], + [-0.0601, -0.1729, -0.0848, ..., -0.1593, -0.0691, -0.2559]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.6811e-08, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 2.4214e-08], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 2.3283e-09], + [ 0.0000e+00, 5.5879e-09, -4.6566e-10, ..., 0.0000e+00, + -1.8626e-09, 2.3283e-09], + ..., + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 4.6566e-10, -9.9186e-07, 0.0000e+00, ..., -5.1223e-09, + 0.0000e+00, -4.4517e-07], + [ 0.0000e+00, 8.6660e-07, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 3.8929e-07]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0171, -0.0358, -0.0102, -0.0153, -0.0364, -0.0003, 0.0319, -0.0151, + 0.0523, -0.0098], device='cuda:0'), grad: tensor([ 6.8219e-07, 4.0513e-08, 6.3330e-08, 2.5006e-07, 3.0734e-08, + 3.4133e-07, 1.0105e-07, 3.4925e-08, -1.2033e-05, 1.0513e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 250.70, cls_loss 0.0015 cls_loss_mapping 0.0011 cls_loss_causal 0.4728 re_mapping 0.0036 re_causal 0.0096 /// teacc 99.09 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.1386, -0.2893, -0.0899, ..., -0.0888, 0.1901, 0.2008], + [-0.2684, -0.2375, -0.0708, ..., -0.2074, -0.2620, -0.1710], + [-0.0667, -0.2184, 0.1657, ..., -0.2695, 0.2969, 0.1395], + ..., + [-0.2104, 0.1320, 0.0177, ..., 0.2407, -0.2667, -0.3291], + [-0.3472, 0.0678, -0.1672, ..., 0.0681, -0.1278, -0.2470], + [-0.0608, -0.1736, -0.0848, ..., -0.1598, -0.0708, -0.2570]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.5390e-08, 0.0000e+00, ..., 1.3039e-08, + 1.8626e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 1.5367e-08, 0.0000e+00, ..., 9.7789e-09, + 9.3132e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, -1.6112e-07, 0.0000e+00, ..., -3.5297e-07, + -1.7602e-07, -7.1712e-08], + [ 0.0000e+00, 1.1642e-08, 0.0000e+00, ..., 4.1910e-09, + 6.0536e-09, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0165, -0.0358, -0.0097, -0.0152, -0.0367, -0.0001, 0.0319, -0.0147, + 0.0518, -0.0106], device='cuda:0'), grad: tensor([ 2.6403e-07, 1.5832e-08, 1.0803e-07, 3.0594e-07, 4.6566e-10, + 1.0664e-07, 1.6401e-06, 5.1223e-09, -2.5239e-06, 6.3330e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 250.24, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4713 re_mapping 0.0037 re_causal 0.0103 /// teacc 98.98 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.1386, -0.2896, -0.0899, ..., -0.0887, 0.1903, 0.2011], + [-0.2687, -0.2375, -0.0708, ..., -0.2075, -0.2627, -0.1726], + [-0.0672, -0.2186, 0.1657, ..., -0.2700, 0.2969, 0.1392], + ..., + [-0.2108, 0.1315, 0.0177, ..., 0.2400, -0.2668, -0.3279], + [-0.3473, 0.0707, -0.1672, ..., 0.0701, -0.1278, -0.2470], + [-0.0612, -0.1736, -0.0848, ..., -0.1600, -0.0713, -0.2575]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + -9.3132e-09, -1.0245e-08], + [ 0.0000e+00, 1.3504e-08, 0.0000e+00, ..., 1.3504e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, -2.0023e-08, 0.0000e+00, ..., -2.0023e-08, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 3.2596e-09, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 8.3819e-09, 7.9162e-09]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0161, -0.0358, -0.0099, -0.0135, -0.0369, -0.0008, 0.0318, -0.0150, + 0.0531, -0.0110], device='cuda:0'), grad: tensor([-3.3062e-08, 4.4238e-08, 3.2596e-09, 1.3970e-09, 9.0897e-07, + 1.3970e-09, 5.5879e-09, -1.4901e-08, 1.2107e-08, -9.2341e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 250.40, cls_loss 0.0013 cls_loss_mapping 0.0013 cls_loss_causal 0.4725 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.06 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.1386, -0.2896, -0.0899, ..., -0.0887, 0.1904, 0.2011], + [-0.2687, -0.2376, -0.0705, ..., -0.2073, -0.2627, -0.1726], + [-0.0672, -0.2186, 0.1657, ..., -0.2700, 0.2969, 0.1392], + ..., + [-0.2108, 0.1315, 0.0175, ..., 0.2399, -0.2668, -0.3280], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1278, -0.2470], + [-0.0614, -0.1736, -0.0848, ..., -0.1600, -0.0714, -0.2576]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -6.5193e-08, -4.2375e-08], + [ 4.6566e-10, 8.8476e-09, 0.0000e+00, ..., 1.3225e-06, + 4.6566e-10, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-08, + 1.8626e-09, 4.6566e-10], + ..., + [ 9.3132e-10, -2.0955e-08, 0.0000e+00, ..., -1.4743e-06, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 3.7253e-09, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 7.9162e-09, 0.0000e+00, ..., 9.8255e-08, + 7.4506e-09, 5.1223e-09]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0161, -0.0357, -0.0099, -0.0135, -0.0367, -0.0008, 0.0318, -0.0151, + 0.0531, -0.0110], device='cuda:0'), grad: tensor([-8.2888e-08, 9.9689e-06, 1.1548e-07, 2.9337e-07, 1.0710e-08, + 8.8476e-09, 8.2888e-08, -1.2241e-05, 9.4110e-07, 9.0059e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 250.64, cls_loss 0.0013 cls_loss_mapping 0.0009 cls_loss_causal 0.4504 re_mapping 0.0035 re_causal 0.0098 /// teacc 99.07 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.1386, -0.2896, -0.0899, ..., -0.0887, 0.1904, 0.2011], + [-0.2687, -0.2376, -0.0703, ..., -0.2071, -0.2628, -0.1726], + [-0.0672, -0.2186, 0.1657, ..., -0.2700, 0.2970, 0.1393], + ..., + [-0.2108, 0.1314, 0.0173, ..., 0.2397, -0.2670, -0.3280], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1278, -0.2470], + [-0.0615, -0.1736, -0.0848, ..., -0.1600, -0.0714, -0.2576]], + device='cuda:0'), grad: tensor([[ 1.1688e-07, 5.9605e-08, 0.0000e+00, ..., 1.3271e-07, + 4.6566e-10, 4.6566e-10], + [ 2.3283e-09, 2.3283e-09, 0.0000e+00, ..., 1.8626e-09, + -2.7940e-09, 0.0000e+00], + [ 3.2596e-09, 3.7253e-09, 0.0000e+00, ..., 2.3283e-08, + 4.6566e-10, 4.6566e-10], + ..., + [ 3.1199e-08, 1.3970e-08, 0.0000e+00, ..., 3.1665e-08, + 0.0000e+00, 0.0000e+00], + [ 5.7742e-08, 4.0047e-08, 0.0000e+00, ..., 1.4761e-07, + 0.0000e+00, 0.0000e+00], + [ 1.2107e-07, 6.6124e-08, 0.0000e+00, ..., 1.2992e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0161, -0.0355, -0.0099, -0.0135, -0.0366, -0.0008, 0.0318, -0.0153, + 0.0531, -0.0110], device='cuda:0'), grad: tensor([ 3.8324e-07, -3.0641e-07, 5.9605e-08, 2.5742e-06, 1.5367e-08, + -3.8520e-06, 2.6496e-07, 9.6858e-08, 3.4738e-07, 4.0093e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 250.50, cls_loss 0.0011 cls_loss_mapping 0.0007 cls_loss_causal 0.4296 re_mapping 0.0033 re_causal 0.0093 /// teacc 99.08 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.1386, -0.2896, -0.0899, ..., -0.0887, 0.1904, 0.2011], + [-0.2687, -0.2376, -0.0703, ..., -0.2071, -0.2628, -0.1726], + [-0.0672, -0.2185, 0.1659, ..., -0.2700, 0.2973, 0.1393], + ..., + [-0.2109, 0.1315, 0.0173, ..., 0.2397, -0.2672, -0.3280], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1278, -0.2470], + [-0.0615, -0.1737, -0.0848, ..., -0.1601, -0.0714, -0.2576]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + -2.3283e-09, -2.3283e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7474e-08, 1.2107e-08, 0.0000e+00, ..., 6.5193e-08, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 5.1223e-09, 0.0000e+00, ..., 2.3749e-08, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 4.1910e-09, 0.0000e+00, ..., 2.4214e-08, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0161, -0.0355, -0.0098, -0.0135, -0.0366, -0.0007, 0.0317, -0.0152, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([-9.3132e-10, 5.5879e-09, 3.2596e-09, 1.0990e-07, 1.8626e-09, + -3.8184e-07, 3.0734e-08, 1.3318e-07, 4.7032e-08, 5.3551e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 250.80, cls_loss 0.0012 cls_loss_mapping 0.0007 cls_loss_causal 0.4800 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.09 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.1386, -0.2897, -0.0899, ..., -0.0887, 0.1904, 0.2012], + [-0.2687, -0.2376, -0.0703, ..., -0.2071, -0.2628, -0.1726], + [-0.0672, -0.2184, 0.1659, ..., -0.2701, 0.2973, 0.1392], + ..., + [-0.2109, 0.1315, 0.0173, ..., 0.2397, -0.2673, -0.3280], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1279, -0.2470], + [-0.0615, -0.1737, -0.0848, ..., -0.1601, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 2.7940e-09, + 5.5879e-09, 0.0000e+00], + [ 1.0710e-08, 3.7253e-09, 0.0000e+00, ..., 6.9849e-09, + 2.6077e-08, 9.3132e-10], + [ 2.6543e-08, -9.3132e-10, 0.0000e+00, ..., 6.9849e-09, + -4.5635e-08, -3.1199e-08], + ..., + [ 0.0000e+00, -1.3039e-08, 0.0000e+00, ..., -3.4459e-08, + 5.8208e-08, 3.0268e-08], + [ 2.3283e-09, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 6.0536e-09, 0.0000e+00], + [ 4.6566e-10, 7.4506e-09, 0.0000e+00, ..., 1.5367e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0161, -0.0355, -0.0097, -0.0135, -0.0366, -0.0006, 0.0316, -0.0152, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([ 2.5146e-08, 1.4249e-07, -1.3132e-07, 4.1910e-09, 2.1886e-08, + 2.3283e-09, -2.0256e-07, 8.2422e-08, 1.7695e-08, 4.5169e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 250.41, cls_loss 0.0010 cls_loss_mapping 0.0006 cls_loss_causal 0.4403 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.1386, -0.2897, -0.0899, ..., -0.0887, 0.1904, 0.2012], + [-0.2687, -0.2376, -0.0703, ..., -0.2071, -0.2628, -0.1726], + [-0.0672, -0.2185, 0.1659, ..., -0.2701, 0.2974, 0.1392], + ..., + [-0.2109, 0.1315, 0.0173, ..., 0.2397, -0.2673, -0.3281], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0616, -0.1737, -0.0848, ..., -0.1601, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -6.5193e-09, -2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3970e-09, -9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 1.5367e-08, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0161, -0.0355, -0.0097, -0.0135, -0.0365, -0.0006, 0.0316, -0.0152, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([-1.8626e-09, -1.4668e-07, -2.7940e-09, 3.7253e-09, 1.9418e-07, + -4.5635e-08, -1.8952e-07, 1.1828e-07, 4.5635e-08, 2.1886e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 250.44, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4012 re_mapping 0.0032 re_causal 0.0091 /// teacc 99.12 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.1386, -0.2897, -0.0899, ..., -0.0887, 0.1904, 0.2012], + [-0.2687, -0.2376, -0.0703, ..., -0.2071, -0.2628, -0.1726], + [-0.0672, -0.2185, 0.1659, ..., -0.2701, 0.2974, 0.1392], + ..., + [-0.2110, 0.1315, 0.0173, ..., 0.2397, -0.2673, -0.3281], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0616, -0.1737, -0.0848, ..., -0.1601, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., -3.2596e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0161, -0.0355, -0.0097, -0.0135, -0.0365, -0.0006, 0.0316, -0.0152, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([ 9.3132e-10, -2.7940e-09, 1.0245e-08, 9.3132e-10, 2.7157e-06, + 9.3132e-10, 1.3970e-09, 1.2154e-07, 4.6566e-10, -2.8424e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 250.35, cls_loss 0.0013 cls_loss_mapping 0.0005 cls_loss_causal 0.4361 re_mapping 0.0032 re_causal 0.0092 /// teacc 99.09 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.1386, -0.2897, -0.0899, ..., -0.0886, 0.1904, 0.2012], + [-0.2687, -0.2376, -0.0701, ..., -0.2069, -0.2628, -0.1726], + [-0.0672, -0.2185, 0.1660, ..., -0.2701, 0.2974, 0.1392], + ..., + [-0.2110, 0.1315, 0.0171, ..., 0.2396, -0.2674, -0.3281], + [-0.3474, 0.0707, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0617, -0.1737, -0.0848, ..., -0.1602, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, + 4.6566e-10]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0160, -0.0355, -0.0097, -0.0134, -0.0365, -0.0006, 0.0316, -0.0153, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([ 4.6566e-10, 4.6566e-10, 0.0000e+00, 1.3504e-08, 9.3132e-10, + -2.2352e-08, 4.6566e-09, 2.3283e-09, 2.3283e-09, 9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 250.25, cls_loss 0.0012 cls_loss_mapping 0.0006 cls_loss_causal 0.4462 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.1386, -0.2897, -0.0899, ..., -0.0886, 0.1905, 0.2012], + [-0.2687, -0.2378, -0.0701, ..., -0.2070, -0.2628, -0.1727], + [-0.0672, -0.2185, 0.1660, ..., -0.2701, 0.2974, 0.1393], + ..., + [-0.2111, 0.1315, 0.0171, ..., 0.2397, -0.2674, -0.3281], + [-0.3474, 0.0706, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0618, -0.1737, -0.0848, ..., -0.1602, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, -0.0000e+00, 0.0000e+00, ..., -2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9791e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0160, -0.0355, -0.0097, -0.0132, -0.0364, -0.0007, 0.0316, -0.0153, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([ 5.1223e-09, -6.9849e-09, 1.1642e-08, 4.4238e-09, -6.5938e-07, + 1.8626e-09, 7.6834e-09, 8.3819e-08, 4.6566e-10, 5.5879e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 250.37, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4277 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.11 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.1386, -0.2898, -0.0899, ..., -0.0886, 0.1905, 0.2012], + [-0.2687, -0.2380, -0.0701, ..., -0.2071, -0.2628, -0.1726], + [-0.0672, -0.2185, 0.1660, ..., -0.2701, 0.2975, 0.1392], + ..., + [-0.2111, 0.1317, 0.0171, ..., 0.2397, -0.2674, -0.3281], + [-0.3474, 0.0706, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0618, -0.1737, -0.0848, ..., -0.1602, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 4.1910e-09, 0.0000e+00, ..., 5.3551e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, 0.0000e+00, ..., 1.4203e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -2.6077e-08, 0.0000e+00, ..., -4.4471e-08, + 0.0000e+00, 0.0000e+00], + [ 1.2340e-08, 1.4435e-08, 0.0000e+00, ..., 3.3528e-08, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 4.6566e-09, 0.0000e+00, ..., 6.0536e-09, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0160, -0.0357, -0.0098, -0.0131, -0.0364, -0.0007, 0.0316, -0.0151, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.6531e-08, 3.3295e-08, -1.1176e-08, 8.3819e-09, + -6.8918e-08, 2.4214e-08, -9.9884e-08, 8.1724e-08, 1.6298e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 250.48, cls_loss 0.0011 cls_loss_mapping 0.0005 cls_loss_causal 0.4435 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.09 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.1386, -0.2898, -0.0899, ..., -0.0886, 0.1905, 0.2012], + [-0.2687, -0.2380, -0.0701, ..., -0.2071, -0.2629, -0.1726], + [-0.0672, -0.2186, 0.1660, ..., -0.2702, 0.2975, 0.1393], + ..., + [-0.2112, 0.1317, 0.0171, ..., 0.2398, -0.2674, -0.3282], + [-0.3475, 0.0706, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0619, -0.1737, -0.0848, ..., -0.1602, -0.0714, -0.2577]], + device='cuda:0'), grad: tensor([[ 3.9581e-09, 1.3970e-09, 0.0000e+00, ..., -5.8347e-07, + -4.2059e-06, -3.0324e-06], + [ 0.0000e+00, 1.6298e-09, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 4.1910e-09, + 7.6834e-09, 5.5879e-09], + ..., + [ 0.0000e+00, -8.8476e-09, 0.0000e+00, ..., -1.0710e-08, + 4.6566e-10, 2.3283e-10], + [ 5.1223e-09, 9.3132e-10, 0.0000e+00, ..., 2.3283e-09, + 1.1642e-09, 1.8626e-09], + [ 0.0000e+00, 3.4925e-09, 0.0000e+00, ..., 5.5879e-09, + 1.0710e-08, 7.6834e-09]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0160, -0.0357, -0.0098, -0.0131, -0.0364, -0.0006, 0.0316, -0.0151, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([-7.2643e-06, -1.6820e-06, 1.2945e-06, 2.1188e-08, 3.2596e-08, + -1.5367e-08, 7.2382e-06, 3.7905e-07, 2.4680e-08, -2.6310e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 250.19, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4328 re_mapping 0.0030 re_causal 0.0092 /// teacc 99.11 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.1386, -0.2899, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2381, -0.0701, ..., -0.2072, -0.2629, -0.1726], + [-0.0672, -0.2186, 0.1660, ..., -0.2702, 0.2975, 0.1393], + ..., + [-0.2112, 0.1317, 0.0171, ..., 0.2398, -0.2675, -0.3282], + [-0.3475, 0.0706, -0.1673, ..., 0.0701, -0.1279, -0.2471], + [-0.0619, -0.1737, -0.0848, ..., -0.1602, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.8894e-09, 0.0000e+00, ..., 2.3283e-09, + -2.9569e-08, -2.3516e-08], + [ 0.0000e+00, 6.2166e-08, 0.0000e+00, ..., 3.0966e-08, + -0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.0571e-07, 0.0000e+00, ..., 5.2620e-08, + -1.0012e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -1.8999e-07, 0.0000e+00, ..., -9.4762e-08, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 1.5600e-08, 0.0000e+00, ..., 7.9162e-09, + 2.3516e-08, 1.8626e-08]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0160, -0.0357, -0.0099, -0.0131, -0.0364, -0.0006, 0.0316, -0.0151, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([-6.5425e-08, 2.9244e-07, 4.8988e-07, 1.1874e-08, 3.7253e-09, + 5.5879e-09, 2.1188e-08, -9.0059e-07, 2.5611e-09, 1.4203e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 250.40, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4309 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.13 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.1386, -0.2899, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2381, -0.0701, ..., -0.2072, -0.2629, -0.1726], + [-0.0672, -0.2187, 0.1660, ..., -0.2702, 0.2975, 0.1393], + ..., + [-0.2112, 0.1318, 0.0171, ..., 0.2398, -0.2675, -0.3282], + [-0.3475, 0.0706, -0.1673, ..., 0.0701, -0.1280, -0.2471], + [-0.0619, -0.1738, -0.0848, ..., -0.1602, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + -1.0608e-06, -5.8580e-07], + [ 0.0000e+00, 5.3551e-09, 0.0000e+00, ..., 6.2864e-09, + 6.7521e-09, 3.7253e-09], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.1642e-09, + 2.5425e-07, 1.4319e-07], + ..., + [ 0.0000e+00, -2.8871e-08, 0.0000e+00, ..., -3.1665e-08, + 4.8894e-09, 2.5611e-09], + [ 2.3283e-10, 3.9581e-09, 0.0000e+00, ..., 4.1910e-09, + 2.8405e-08, 1.8394e-08], + [ 0.0000e+00, 1.3271e-08, 0.0000e+00, ..., 1.5832e-08, + 1.0361e-07, 6.3097e-08]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0160, -0.0358, -0.0099, -0.0129, -0.0364, -0.0007, 0.0316, -0.0150, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([-1.9632e-06, 2.9802e-08, 4.7870e-07, 2.3749e-08, 8.1491e-09, + 5.1223e-09, 1.1846e-06, -7.6601e-08, 6.6357e-08, 2.4354e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 249.93, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4508 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.10 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.1386, -0.2899, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2381, -0.0701, ..., -0.2072, -0.2629, -0.1727], + [-0.0672, -0.2187, 0.1661, ..., -0.2702, 0.2975, 0.1393], + ..., + [-0.2113, 0.1318, 0.0171, ..., 0.2399, -0.2675, -0.3282], + [-0.3475, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0620, -0.1738, -0.0848, ..., -0.1602, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + [ 3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, -2.3283e-10, 0.0000e+00, ..., -1.1642e-09, + -1.8626e-09, -1.3970e-09], + ..., + [ 2.0955e-09, 1.1642e-09, 0.0000e+00, ..., 2.5611e-09, + 1.3970e-09, 1.1642e-09], + [ 5.7044e-08, 3.2596e-09, 0.0000e+00, ..., 4.5868e-08, + 1.3970e-09, 1.3970e-09], + [ 3.1898e-08, 3.7253e-08, 0.0000e+00, ..., 2.0256e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0160, -0.0358, -0.0099, -0.0129, -0.0364, -0.0006, 0.0315, -0.0150, + 0.0531, -0.0111], device='cuda:0'), grad: tensor([ 1.3970e-09, 1.4435e-08, -3.4925e-09, 3.4552e-07, -7.2177e-09, + -6.7893e-07, -4.6566e-10, 1.3504e-08, 1.4110e-07, 1.8254e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 250.24, cls_loss 0.0010 cls_loss_mapping 0.0004 cls_loss_causal 0.4607 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.1386, -0.2900, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2381, -0.0701, ..., -0.2072, -0.2629, -0.1727], + [-0.0672, -0.2188, 0.1661, ..., -0.2702, 0.2976, 0.1393], + ..., + [-0.2113, 0.1318, 0.0171, ..., 0.2399, -0.2676, -0.3282], + [-0.3475, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0620, -0.1738, -0.0848, ..., -0.1603, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -3.2596e-09, -3.4925e-09], + [ 2.3283e-10, 4.1910e-09, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-09, -5.1223e-09, 0.0000e+00, ..., -6.2864e-09, + 0.0000e+00, 2.3283e-10], + [ 2.0955e-09, 8.8476e-09, 0.0000e+00, ..., -2.3283e-10, + 0.0000e+00, 9.3132e-10], + [ 2.3283e-10, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 2.0955e-09, 2.3283e-09]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0160, -0.0358, -0.0099, -0.0129, -0.0363, -0.0006, 0.0315, -0.0150, + 0.0531, -0.0112], device='cuda:0'), grad: tensor([-1.0477e-08, 1.2107e-08, 2.5611e-09, -2.6077e-08, 0.0000e+00, + -1.3970e-09, 3.4925e-09, -1.2806e-08, 1.7229e-08, 1.2107e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 250.17, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4290 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.11 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.1386, -0.2900, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2381, -0.0701, ..., -0.2072, -0.2629, -0.1727], + [-0.0672, -0.2188, 0.1661, ..., -0.2702, 0.2976, 0.1393], + ..., + [-0.2113, 0.1318, 0.0171, ..., 0.2399, -0.2676, -0.3283], + [-0.3475, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0621, -0.1738, -0.0848, ..., -0.1603, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., -4.4238e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 3.9581e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0160, -0.0358, -0.0099, -0.0129, -0.0363, -0.0006, 0.0315, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 9.3132e-10, 9.3132e-10, 2.3283e-10, 1.3970e-09, 5.8208e-09, + 1.8626e-09, 2.3283e-10, -6.0536e-09, 3.8883e-08, -4.4936e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 250.23, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4450 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.12 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.1386, -0.2901, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2381, -0.0701, ..., -0.2072, -0.2629, -0.1727], + [-0.0672, -0.2188, 0.1661, ..., -0.2702, 0.2977, 0.1394], + ..., + [-0.2114, 0.1318, 0.0171, ..., 0.2399, -0.2676, -0.3283], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0621, -0.1738, -0.0848, ..., -0.1603, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.8825e-07, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 2.3283e-10]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0160, -0.0358, -0.0099, -0.0129, -0.0363, -0.0005, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 3.9581e-09, 3.9581e-09, 0.0000e+00, 3.7253e-09, 8.0978e-07, + 2.7311e-07, -1.0710e-08, -9.9279e-07, 5.1223e-09, -9.2201e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 250.38, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4438 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.15 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.1386, -0.2901, -0.0899, ..., -0.0886, 0.1905, 0.2013], + [-0.2688, -0.2382, -0.0700, ..., -0.2072, -0.2630, -0.1727], + [-0.0672, -0.2188, 0.1663, ..., -0.2703, 0.2978, 0.1394], + ..., + [-0.2114, 0.1319, 0.0170, ..., 0.2399, -0.2676, -0.3283], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0621, -0.1738, -0.0848, ..., -0.1603, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.8894e-09, 0.0000e+00, ..., -2.7940e-09, + -1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.6298e-09, 1.1642e-09]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0160, -0.0358, -0.0099, -0.0128, -0.0363, -0.0005, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 2.3283e-10, -1.8626e-09, 2.0955e-09, 4.6566e-09, 9.3132e-10, + 1.1642e-09, 4.1211e-08, 1.6298e-09, -4.4238e-08, 3.9581e-09], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 417---------------------------------------------------- +epoch 417, time 267.27, cls_loss 0.0010 cls_loss_mapping 0.0006 cls_loss_causal 0.4379 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.17 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.1386, -0.2902, -0.0899, ..., -0.0887, 0.1905, 0.2014], + [-0.2688, -0.2383, -0.0700, ..., -0.2072, -0.2630, -0.1727], + [-0.0673, -0.2187, 0.1663, ..., -0.2702, 0.2979, 0.1394], + ..., + [-0.2115, 0.1320, 0.0170, ..., 0.2399, -0.2679, -0.3284], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0621, -0.1738, -0.0848, ..., -0.1603, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3504e-08, -1.0710e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.2107e-08, 4.6566e-10], + ..., + [ 4.6566e-10, -4.6566e-10, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 6.9849e-09, 1.3970e-09, 0.0000e+00, ..., 6.5193e-09, + 2.7940e-09, 1.8626e-09]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0160, -0.0359, -0.0098, -0.0127, -0.0363, -0.0005, 0.0314, -0.0149, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([-3.4925e-08, -4.1444e-08, -7.6834e-08, 9.3132e-10, 2.5611e-08, + -2.3283e-08, 2.7474e-08, 1.6298e-08, 1.1316e-07, -1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 250.41, cls_loss 0.0011 cls_loss_mapping 0.0005 cls_loss_causal 0.4268 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.15 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.1386, -0.2902, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2688, -0.2383, -0.0699, ..., -0.2071, -0.2630, -0.1727], + [-0.0673, -0.2187, 0.1663, ..., -0.2702, 0.2979, 0.1394], + ..., + [-0.2115, 0.1320, 0.0169, ..., 0.2399, -0.2679, -0.3284], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0621, -0.1739, -0.0848, ..., -0.1603, -0.0715, -0.2578]], + device='cuda:0'), grad: tensor([[4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0160, -0.0358, -0.0098, -0.0126, -0.0363, -0.0006, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 1.3970e-09, -4.9826e-08, 2.7940e-09, -2.9337e-08, 5.3085e-08, + -1.0245e-08, 1.3504e-08, 6.9849e-09, 4.1910e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 250.29, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4327 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.12 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.1386, -0.2903, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2688, -0.2383, -0.0699, ..., -0.2071, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1663, ..., -0.2702, 0.2980, 0.1395], + ..., + [-0.2115, 0.1320, 0.0169, ..., 0.2399, -0.2679, -0.3284], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0621, -0.1739, -0.0848, ..., -0.1603, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -3.7253e-08, 0.0000e+00, ..., -9.5926e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.6787e-08, 0.0000e+00, ..., 9.4529e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0160, -0.0358, -0.0098, -0.0126, -0.0363, -0.0006, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 6.0536e-09, -4.6566e-10, 1.7695e-08, 6.5193e-09, 1.3970e-09, + 3.2596e-09, 4.1910e-09, -1.7462e-07, -4.0047e-08, 1.8161e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 250.61, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.3993 re_mapping 0.0028 re_causal 0.0084 /// teacc 99.09 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.1386, -0.2903, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2688, -0.2384, -0.0699, ..., -0.2071, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1663, ..., -0.2702, 0.2980, 0.1395], + ..., + [-0.2116, 0.1320, 0.0169, ..., 0.2399, -0.2680, -0.3284], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0622, -0.1739, -0.0848, ..., -0.1604, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, -0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0160, -0.0358, -0.0097, -0.0126, -0.0363, -0.0006, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 1.3970e-09, 9.3132e-10, -3.7253e-09, 5.1223e-09, 1.8626e-09, + -8.2888e-08, 5.4948e-08, 2.7940e-09, 1.0245e-08, 1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 250.81, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4276 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.09 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.1386, -0.2904, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2689, -0.2384, -0.0699, ..., -0.2072, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1663, ..., -0.2703, 0.2980, 0.1395], + ..., + [-0.2116, 0.1320, 0.0169, ..., 0.2399, -0.2680, -0.3285], + [-0.3476, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0622, -0.1739, -0.0848, ..., -0.1604, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, -2.3283e-09], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 7.4506e-08, 0.0000e+00, ..., 1.0571e-07, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -8.5682e-08, 0.0000e+00, ..., -1.2200e-07, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.2573e-08, 0.0000e+00, ..., 1.4901e-08, + 4.6566e-10, 4.6566e-10], + [ 2.3283e-09, 2.3283e-09, 0.0000e+00, ..., -0.0000e+00, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0160, -0.0358, -0.0097, -0.0125, -0.0363, -0.0006, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([-6.5193e-09, 5.5879e-09, 1.7742e-07, -1.9092e-08, 9.3132e-10, + 9.3132e-09, 4.1910e-09, -1.9930e-07, 3.1665e-08, 6.9849e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 250.69, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4694 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.10 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.1386, -0.2904, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2689, -0.2384, -0.0699, ..., -0.2072, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1664, ..., -0.2703, 0.2981, 0.1396], + ..., + [-0.2117, 0.1321, 0.0169, ..., 0.2399, -0.2681, -0.3285], + [-0.3477, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0623, -0.1739, -0.0848, ..., -0.1604, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-09, + 0.0000e+00, 1.3970e-09], + ..., + [ 0.0000e+00, -6.5193e-09, 0.0000e+00, ..., -9.1270e-08, + 0.0000e+00, -1.1642e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -0.0000e+00, -0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0160, -0.0358, -0.0097, -0.0124, -0.0363, -0.0007, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 1.8626e-09, 2.3283e-09, 1.6764e-08, 9.3132e-10, 2.7940e-08, + 9.8255e-08, 9.3132e-10, -1.5926e-07, 1.4901e-08, -4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 250.91, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4292 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.12 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.1386, -0.2904, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2689, -0.2385, -0.0699, ..., -0.2072, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1664, ..., -0.2703, 0.2981, 0.1396], + ..., + [-0.2117, 0.1321, 0.0169, ..., 0.2399, -0.2681, -0.3286], + [-0.3477, 0.0706, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0623, -0.1739, -0.0848, ..., -0.1604, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, -0.0000e+00, 0.0000e+00, ..., -3.2596e-09, + 9.3132e-10, 4.6566e-10], + [ 2.7940e-09, 4.6566e-10, 0.0000e+00, ..., 4.1910e-09, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0160, -0.0358, -0.0097, -0.0123, -0.0362, -0.0007, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([-3.7253e-09, -3.5716e-07, 3.0128e-07, 2.2352e-08, 3.2596e-09, + -2.3283e-08, 6.5193e-09, 2.6543e-08, -6.9849e-09, 3.9581e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 250.69, cls_loss 0.0010 cls_loss_mapping 0.0004 cls_loss_causal 0.4186 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.10 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.1386, -0.2904, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2689, -0.2385, -0.0699, ..., -0.2072, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1664, ..., -0.2703, 0.2982, 0.1398], + ..., + [-0.2117, 0.1321, 0.0169, ..., 0.2399, -0.2683, -0.3287], + [-0.3477, 0.0705, -0.1673, ..., 0.0700, -0.1280, -0.2471], + [-0.0623, -0.1739, -0.0848, ..., -0.1604, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.1910e-09, -3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.0245e-08, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.6077e-08, -1.5832e-08], + ..., + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., -1.8626e-09, + 1.4901e-08, 8.8476e-09], + [-3.7253e-09, -4.6566e-09, 0.0000e+00, ..., -2.3749e-08, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 4.1910e-09, 3.2596e-09]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0160, -0.0358, -0.0096, -0.0123, -0.0362, -0.0007, 0.0314, -0.0150, + 0.0530, -0.0112], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.4398e-06, -1.3364e-07, 7.8231e-08, -3.2596e-09, + 1.3364e-07, 4.0093e-07, 9.0804e-08, -2.0433e-06, 3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 250.71, cls_loss 0.0010 cls_loss_mapping 0.0004 cls_loss_causal 0.4549 re_mapping 0.0028 re_causal 0.0088 /// teacc 99.10 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.1387, -0.2904, -0.0899, ..., -0.0887, 0.1906, 0.2014], + [-0.2689, -0.2385, -0.0698, ..., -0.2071, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1664, ..., -0.2703, 0.2982, 0.1398], + ..., + [-0.2118, 0.1321, 0.0168, ..., 0.2399, -0.2683, -0.3287], + [-0.3477, 0.0705, -0.1674, ..., 0.0701, -0.1280, -0.2471], + [-0.0623, -0.1739, -0.0848, ..., -0.1605, -0.0714, -0.2579]], + device='cuda:0'), grad: tensor([[4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, + 4.6566e-10], + [0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, + 4.6566e-10], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0161, -0.0358, -0.0097, -0.0122, -0.0361, -0.0007, 0.0314, -0.0150, + 0.0530, -0.0113], device='cuda:0'), grad: tensor([ 2.3283e-09, 2.3283e-09, 1.3970e-09, 1.8626e-09, 7.9162e-09, + 1.4435e-08, -2.3749e-08, 1.8626e-09, 4.6566e-10, -6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 250.25, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4291 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.12 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.1387, -0.2905, -0.0899, ..., -0.0886, 0.1906, 0.2015], + [-0.2689, -0.2385, -0.0698, ..., -0.2071, -0.2630, -0.1728], + [-0.0673, -0.2187, 0.1664, ..., -0.2704, 0.2982, 0.1398], + ..., + [-0.2118, 0.1321, 0.0168, ..., 0.2399, -0.2683, -0.3288], + [-0.3477, 0.0705, -0.1674, ..., 0.0700, -0.1281, -0.2472], + [-0.0624, -0.1739, -0.0848, ..., -0.1605, -0.0714, -0.2579]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, 0.0000e+00, + 0.0000e+00], + [9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, 0.0000e+00, + 0.0000e+00], + [1.3504e-08, 0.0000e+00, 0.0000e+00, ..., 7.4971e-08, 4.6566e-10, + 4.6566e-10]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0161, -0.0358, -0.0097, -0.0122, -0.0362, -0.0008, 0.0314, -0.0150, + 0.0530, -0.0113], device='cuda:0'), grad: tensor([ 5.5879e-09, 9.3132e-09, 2.7940e-09, 3.2131e-08, -7.9395e-07, + -8.9407e-08, 1.5832e-08, 1.9372e-07, 1.7229e-08, 6.1281e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 250.32, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4183 re_mapping 0.0027 re_causal 0.0086 /// teacc 99.12 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.1387, -0.2905, -0.0899, ..., -0.0886, 0.1906, 0.2015], + [-0.2689, -0.2385, -0.0698, ..., -0.2071, -0.2631, -0.1728], + [-0.0673, -0.2187, 0.1664, ..., -0.2704, 0.2982, 0.1398], + ..., + [-0.2119, 0.1321, 0.0168, ..., 0.2399, -0.2684, -0.3288], + [-0.3477, 0.0705, -0.1674, ..., 0.0700, -0.1281, -0.2472], + [-0.0624, -0.1739, -0.0848, ..., -0.1605, -0.0715, -0.2579]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., -4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., -9.3132e-10, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0160, -0.0358, -0.0096, -0.0121, -0.0361, -0.0008, 0.0314, -0.0150, + 0.0530, -0.0113], device='cuda:0'), grad: tensor([ 4.6566e-10, -2.0443e-07, 1.3970e-09, -6.0536e-09, 0.0000e+00, + 7.9162e-09, -3.7253e-09, 1.6252e-07, 2.3283e-09, 4.3772e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 250.72, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4242 re_mapping 0.0027 re_causal 0.0086 /// teacc 99.13 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.1387, -0.2906, -0.0899, ..., -0.0886, 0.1906, 0.2015], + [-0.2689, -0.2385, -0.0698, ..., -0.2072, -0.2631, -0.1728], + [-0.0673, -0.2187, 0.1665, ..., -0.2704, 0.2982, 0.1399], + ..., + [-0.2119, 0.1321, 0.0168, ..., 0.2400, -0.2684, -0.3288], + [-0.3478, 0.0705, -0.1674, ..., 0.0700, -0.1281, -0.2472], + [-0.0624, -0.1740, -0.0848, ..., -0.1605, -0.0715, -0.2580]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, -4.6566e-10, ..., -4.6566e-10, + -5.1223e-09, -3.7253e-09], + ..., + [ 4.6566e-10, 2.7940e-09, 4.6566e-10, ..., -3.6322e-08, + 4.6566e-09, 3.2596e-09], + [ 8.3819e-09, 9.3132e-10, 0.0000e+00, ..., 7.9162e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0160, -0.0358, -0.0096, -0.0120, -0.0361, -0.0008, 0.0314, -0.0150, + 0.0530, -0.0113], device='cuda:0'), grad: tensor([ 9.3132e-10, -7.4506e-09, -2.2352e-08, 1.3970e-09, 1.1874e-07, + -6.2864e-08, 6.2864e-08, -6.2864e-08, 2.9337e-08, -5.1688e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 250.20, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4689 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.12 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.1387, -0.2906, -0.0899, ..., -0.0886, 0.1906, 0.2015], + [-0.2690, -0.2385, -0.0698, ..., -0.2072, -0.2631, -0.1728], + [-0.0673, -0.2187, 0.1665, ..., -0.2705, 0.2983, 0.1399], + ..., + [-0.2120, 0.1321, 0.0168, ..., 0.2400, -0.2684, -0.3289], + [-0.3478, 0.0705, -0.1674, ..., 0.0700, -0.1281, -0.2472], + [-0.0624, -0.1740, -0.0848, ..., -0.1606, -0.0714, -0.2580]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -1.0431e-07, -5.3551e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + -2.7940e-09, -3.2596e-09], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -1.8626e-09, + 3.7253e-09, 3.7253e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.0477e-07, 5.4017e-08]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0161, -0.0358, -0.0096, -0.0120, -0.0361, -0.0008, 0.0314, -0.0150, + 0.0530, -0.0113], device='cuda:0'), grad: tensor([-2.5751e-07, 2.3283e-09, -9.3132e-09, 5.1223e-09, -5.9046e-07, + 1.8161e-08, 1.8626e-09, 2.2817e-08, -2.1420e-08, 8.3586e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 250.30, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4265 re_mapping 0.0027 re_causal 0.0085 /// teacc 99.13 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.1387, -0.2907, -0.0899, ..., -0.0886, 0.1906, 0.2015], + [-0.2690, -0.2385, -0.0698, ..., -0.2072, -0.2631, -0.1728], + [-0.0673, -0.2187, 0.1665, ..., -0.2705, 0.2984, 0.1400], + ..., + [-0.2121, 0.1321, 0.0168, ..., 0.2400, -0.2686, -0.3289], + [-0.3478, 0.0705, -0.1674, ..., 0.0701, -0.1281, -0.2472], + [-0.0624, -0.1740, -0.0848, ..., -0.1606, -0.0714, -0.2580]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -3.8557e-07, -3.5483e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -3.2596e-09, -2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 3.7253e-09, 3.7253e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 5.4948e-08, 5.1223e-08]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0161, -0.0358, -0.0096, -0.0120, -0.0360, -0.0009, 0.0314, -0.0150, + 0.0530, -0.0113], device='cuda:0'), grad: tensor([-9.2713e-07, 7.9162e-09, -2.9802e-08, 1.3970e-09, 4.7171e-07, + 6.5193e-09, 7.7626e-07, 5.1223e-08, 2.8871e-08, -3.8324e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 250.24, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4369 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.14 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.1388, -0.2907, -0.0899, ..., -0.0886, 0.1906, 0.2015], + [-0.2690, -0.2386, -0.0698, ..., -0.2073, -0.2631, -0.1729], + [-0.0673, -0.2187, 0.1666, ..., -0.2705, 0.2984, 0.1400], + ..., + [-0.2121, 0.1322, 0.0168, ..., 0.2401, -0.2686, -0.3290], + [-0.3478, 0.0705, -0.1674, ..., 0.0701, -0.1281, -0.2472], + [-0.0625, -0.1741, -0.0848, ..., -0.1606, -0.0714, -0.2580]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, -4.6566e-10], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 1.0710e-08, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -6.9849e-09, 0.0000e+00, ..., -1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.1910e-09, 3.2596e-09], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 2.3283e-09, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0162, -0.0359, -0.0095, -0.0119, -0.0360, -0.0009, 0.0314, -0.0149, + 0.0531, -0.0113], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.3528e-08, 0.0000e+00, 4.6566e-10, 0.0000e+00, + 5.7742e-08, -8.0559e-08, -4.0513e-08, 1.9558e-08, 1.0710e-08], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 432---------------------------------------------------- +epoch 432, time 266.64, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4147 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.18 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.1388, -0.2907, -0.0899, ..., -0.0886, 0.1906, 0.2016], + [-0.2690, -0.2386, -0.0698, ..., -0.2073, -0.2631, -0.1729], + [-0.0673, -0.2187, 0.1666, ..., -0.2706, 0.2984, 0.1400], + ..., + [-0.2122, 0.1322, 0.0168, ..., 0.2401, -0.2686, -0.3290], + [-0.3479, 0.0705, -0.1674, ..., 0.0701, -0.1281, -0.2472], + [-0.0625, -0.1741, -0.0848, ..., -0.1607, -0.0714, -0.2581]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.1176e-08, -6.5193e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 7.9162e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 2.1886e-08, 7.9162e-09, 0.0000e+00, ..., 4.6566e-10, + 2.8405e-08, 2.7940e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.1910e-09, 2.7940e-09]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0161, -0.0359, -0.0095, -0.0119, -0.0360, -0.0009, 0.0314, -0.0149, + 0.0531, -0.0113], device='cuda:0'), grad: tensor([-2.5611e-08, 3.2596e-09, 3.2131e-08, -6.9849e-09, -9.3132e-10, + 1.8626e-09, -1.3271e-07, 3.4925e-08, 1.1967e-07, -2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 250.32, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4450 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.1388, -0.2907, -0.0899, ..., -0.0886, 0.1907, 0.2016], + [-0.2690, -0.2386, -0.0698, ..., -0.2073, -0.2631, -0.1729], + [-0.0673, -0.2187, 0.1666, ..., -0.2706, 0.2985, 0.1401], + ..., + [-0.2122, 0.1322, 0.0168, ..., 0.2401, -0.2687, -0.3290], + [-0.3479, 0.0705, -0.1674, ..., 0.0701, -0.1282, -0.2472], + [-0.0625, -0.1741, -0.0848, ..., -0.1607, -0.0715, -0.2581]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + -6.7987e-08, -6.7987e-08], + [ 4.6566e-10, 1.2573e-08, 0.0000e+00, ..., 2.4214e-08, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 5.1223e-09, 0.0000e+00, ..., 9.7789e-09, + -7.4506e-09, -3.2596e-09], + ..., + [ 4.6566e-10, -2.6869e-07, 0.0000e+00, ..., -5.0897e-07, + 2.3283e-09, 9.3132e-10], + [ 4.6566e-09, 2.0023e-08, 0.0000e+00, ..., 1.7229e-08, + 9.3132e-10, 4.6566e-10], + [ 1.0245e-08, 2.3423e-07, 0.0000e+00, ..., 4.5029e-07, + 5.1223e-09, 3.7253e-09]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0161, -0.0359, -0.0094, -0.0119, -0.0360, -0.0009, 0.0314, -0.0149, + 0.0530, -0.0114], device='cuda:0'), grad: tensor([-1.2526e-07, 2.0629e-07, -2.7940e-09, 3.7253e-09, -1.7639e-06, + -1.5832e-08, 1.4063e-07, -1.3355e-06, 7.5903e-08, 2.8126e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 250.70, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4049 re_mapping 0.0027 re_causal 0.0086 /// teacc 99.14 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.1388, -0.2907, -0.0899, ..., -0.0885, 0.1907, 0.2017], + [-0.2690, -0.2386, -0.0698, ..., -0.2073, -0.2631, -0.1729], + [-0.0674, -0.2186, 0.1666, ..., -0.2706, 0.2986, 0.1401], + ..., + [-0.2122, 0.1322, 0.0168, ..., 0.2402, -0.2689, -0.3290], + [-0.3479, 0.0704, -0.1674, ..., 0.0701, -0.1282, -0.2472], + [-0.0625, -0.1741, -0.0848, ..., -0.1607, -0.0715, -0.2582]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -7.9162e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -7.4506e-09, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0161, -0.0359, -0.0094, -0.0119, -0.0360, -0.0009, 0.0314, -0.0149, + 0.0531, -0.0114], device='cuda:0'), grad: tensor([ 2.3283e-09, -1.2200e-07, 2.7940e-09, 3.2596e-09, 1.8626e-09, + 4.6566e-10, -3.7253e-09, 9.2667e-08, 7.4506e-09, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 250.94, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4288 re_mapping 0.0028 re_causal 0.0088 /// teacc 99.12 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.1388, -0.2907, -0.0899, ..., -0.0885, 0.1907, 0.2017], + [-0.2691, -0.2386, -0.0698, ..., -0.2073, -0.2632, -0.1729], + [-0.0674, -0.2187, 0.1667, ..., -0.2707, 0.2986, 0.1401], + ..., + [-0.2123, 0.1322, 0.0168, ..., 0.2402, -0.2689, -0.3290], + [-0.3480, 0.0704, -0.1674, ..., 0.0701, -0.1282, -0.2472], + [-0.0624, -0.1741, -0.0848, ..., -0.1607, -0.0715, -0.2582]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.7323e-07, -1.4203e-07], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.6298e-08, + -2.2817e-08, -2.9802e-08], + ..., + [ 0.0000e+00, -1.2573e-08, 0.0000e+00, ..., 3.2596e-09, + 2.1886e-08, 2.8871e-08], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 5.5879e-09, + 2.7940e-09, 2.7940e-09]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0161, -0.0359, -0.0094, -0.0119, -0.0360, -0.0009, 0.0314, -0.0149, + 0.0530, -0.0114], device='cuda:0'), grad: tensor([-3.0175e-07, 7.4506e-09, -2.5611e-07, 1.3970e-09, 2.7940e-09, + 6.5193e-09, 3.0920e-07, 2.5798e-07, 2.7940e-09, -1.8161e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 250.68, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4422 re_mapping 0.0027 re_causal 0.0089 /// teacc 99.13 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.1388, -0.2908, -0.0899, ..., -0.0885, 0.1907, 0.2018], + [-0.2691, -0.2387, -0.0698, ..., -0.2073, -0.2632, -0.1729], + [-0.0674, -0.2187, 0.1667, ..., -0.2707, 0.2987, 0.1401], + ..., + [-0.2124, 0.1322, 0.0168, ..., 0.2402, -0.2689, -0.3291], + [-0.3480, 0.0704, -0.1674, ..., 0.0700, -0.1282, -0.2472], + [-0.0625, -0.1742, -0.0848, ..., -0.1607, -0.0714, -0.2582]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -9.3132e-10, 0.0000e+00], + ..., + [ 4.6566e-10, -4.6566e-10, 0.0000e+00, ..., 7.4506e-09, + 9.3132e-10, 4.6566e-10], + [ 1.0710e-08, 4.6566e-10, 0.0000e+00, ..., 7.9162e-09, + -4.6566e-10, -1.8626e-09], + [ 2.7940e-09, 4.6566e-10, 0.0000e+00, ..., 1.7229e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0161, -0.0359, -0.0094, -0.0117, -0.0360, -0.0009, 0.0313, -0.0149, + 0.0530, -0.0114], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.3504e-08, 4.1910e-09, 1.1642e-08, -2.6403e-07, + -4.2375e-08, 5.5879e-09, 5.8673e-08, 1.3039e-08, 1.9697e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 250.53, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4196 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.11 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.1389, -0.2909, -0.0899, ..., -0.0886, 0.1907, 0.2018], + [-0.2691, -0.2387, -0.0697, ..., -0.2072, -0.2632, -0.1729], + [-0.0674, -0.2187, 0.1667, ..., -0.2707, 0.2987, 0.1401], + ..., + [-0.2125, 0.1322, 0.0167, ..., 0.2401, -0.2689, -0.3291], + [-0.3481, 0.0704, -0.1674, ..., 0.0700, -0.1283, -0.2473], + [-0.0625, -0.1742, -0.0848, ..., -0.1608, -0.0715, -0.2583]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 1.3970e-09], + [ 6.9849e-10, 9.3132e-10, 0.0000e+00, ..., -3.0268e-09, + 2.3283e-10, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -2.0256e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -3.0268e-09, 0.0000e+00, ..., 1.3970e-09, + 1.6298e-09, 0.0000e+00], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + [ 2.3283e-10, 1.6298e-09, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0161, -0.0359, -0.0094, -0.0117, -0.0360, -0.0008, 0.0313, -0.0150, + 0.0529, -0.0114], device='cuda:0'), grad: tensor([ 1.2806e-08, -5.1921e-08, -2.6776e-08, 2.4447e-08, 1.5134e-08, + 5.8906e-08, -9.0571e-08, 6.1933e-08, 3.0268e-09, -6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 250.50, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4287 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.13 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.1389, -0.2910, -0.0899, ..., -0.0886, 0.1908, 0.2018], + [-0.2691, -0.2387, -0.0697, ..., -0.2073, -0.2632, -0.1730], + [-0.0674, -0.2187, 0.1667, ..., -0.2707, 0.2987, 0.1402], + ..., + [-0.2125, 0.1322, 0.0167, ..., 0.2401, -0.2690, -0.3292], + [-0.3481, 0.0704, -0.1674, ..., 0.0700, -0.1283, -0.2473], + [-0.0625, -0.1742, -0.0848, ..., -0.1608, -0.0715, -0.2583]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 4.6566e-10], + [ 0.0000e+00, 6.7521e-09, 0.0000e+00, ..., 8.3819e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 4.6566e-10, -1.8161e-08, 0.0000e+00, ..., -2.0489e-08, + 0.0000e+00, -6.9849e-10], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 1.2573e-08, 0.0000e+00, ..., 8.8476e-09, + 4.6566e-10, 2.3283e-10]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0161, -0.0359, -0.0094, -0.0117, -0.0360, -0.0007, 0.0312, -0.0150, + 0.0529, -0.0114], device='cuda:0'), grad: tensor([ 2.3283e-10, 5.3551e-09, 2.2352e-08, -1.7649e-07, 4.6566e-10, + 1.6764e-07, 6.9849e-10, -4.4005e-08, 9.3132e-10, 2.8638e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 250.44, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4035 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.12 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.1389, -0.2910, -0.0899, ..., -0.0885, 0.1909, 0.2019], + [-0.2691, -0.2387, -0.0697, ..., -0.2073, -0.2632, -0.1730], + [-0.0674, -0.2187, 0.1667, ..., -0.2708, 0.2987, 0.1402], + ..., + [-0.2126, 0.1322, 0.0167, ..., 0.2402, -0.2690, -0.3292], + [-0.3482, 0.0704, -0.1674, ..., 0.0700, -0.1283, -0.2473], + [-0.0624, -0.1742, -0.0848, ..., -0.1608, -0.0715, -0.2583]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.9162e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.2002e-07, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 9.7789e-09, 2.3283e-10, 0.0000e+00, ..., -1.6298e-09, + 2.3283e-10, 2.3283e-10], + [ 4.1910e-09, 2.5611e-09, 0.0000e+00, ..., -4.6566e-10, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0160, -0.0359, -0.0093, -0.0116, -0.0360, -0.0007, 0.0312, -0.0150, + 0.0529, -0.0114], device='cuda:0'), grad: tensor([ 1.2806e-08, 1.4855e-06, -1.7090e-06, -1.5600e-08, 7.5670e-08, + 4.2841e-08, 2.3749e-08, 9.0804e-09, 6.1700e-08, 1.2573e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 250.95, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4243 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.15 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.1389, -0.2910, -0.0899, ..., -0.0884, 0.1909, 0.2020], + [-0.2692, -0.2387, -0.0697, ..., -0.2073, -0.2632, -0.1730], + [-0.0674, -0.2188, 0.1667, ..., -0.2708, 0.2988, 0.1403], + ..., + [-0.2126, 0.1322, 0.0167, ..., 0.2402, -0.2691, -0.3293], + [-0.3482, 0.0704, -0.1674, ..., 0.0699, -0.1283, -0.2473], + [-0.0625, -0.1742, -0.0848, ..., -0.1608, -0.0715, -0.2584]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.4040e-07, -1.3551e-07], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 4.1910e-09], + ..., + [ 4.6566e-10, 2.0955e-09, 0.0000e+00, ..., 6.9849e-10, + 1.8626e-09, 1.6298e-09], + [ 4.6566e-10, -2.5611e-09, 0.0000e+00, ..., -1.5832e-08, + 3.5157e-08, 3.0501e-08], + [ 0.0000e+00, 3.0268e-09, 0.0000e+00, ..., 4.8894e-09, + 2.3749e-08, 2.2119e-08]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0160, -0.0359, -0.0094, -0.0116, -0.0358, -0.0008, 0.0312, -0.0150, + 0.0529, -0.0114], device='cuda:0'), grad: tensor([-2.3795e-07, 7.2177e-09, 1.0710e-08, -2.5611e-09, -2.6310e-08, + 2.8173e-08, 9.7556e-08, 1.1642e-08, 4.3772e-08, 7.8930e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 250.66, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4282 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.14 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.1389, -0.2910, -0.0899, ..., -0.0884, 0.1909, 0.2020], + [-0.2692, -0.2387, -0.0697, ..., -0.2073, -0.2633, -0.1730], + [-0.0674, -0.2188, 0.1667, ..., -0.2708, 0.2988, 0.1403], + ..., + [-0.2127, 0.1322, 0.0167, ..., 0.2402, -0.2691, -0.3293], + [-0.3483, 0.0704, -0.1674, ..., 0.0700, -0.1283, -0.2473], + [-0.0625, -0.1742, -0.0848, ..., -0.1608, -0.0715, -0.2584]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-09, 1.1642e-09], + [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., 1.1642e-09, + 6.9849e-10, 4.6566e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 4.6566e-10], + ..., + [ 2.3283e-10, -8.1491e-09, 4.6566e-10, ..., -1.8859e-08, + 0.0000e+00, 0.0000e+00], + [ 6.7521e-09, 0.0000e+00, 0.0000e+00, ..., 3.9581e-09, + 6.9849e-10, 4.6566e-10], + [ 4.6566e-10, 7.9162e-09, 1.6298e-09, ..., 2.6077e-08, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0160, -0.0359, -0.0094, -0.0116, -0.0358, -0.0008, 0.0312, -0.0150, + 0.0529, -0.0114], device='cuda:0'), grad: tensor([ 1.5134e-08, 4.8894e-09, 1.8626e-09, 4.2934e-07, -1.9092e-08, + -3.9325e-07, -8.0792e-08, -2.3283e-08, 1.5367e-08, 5.1688e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 251.06, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.4097 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.13 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.1390, -0.2910, -0.0899, ..., -0.0884, 0.1909, 0.2020], + [-0.2692, -0.2389, -0.0697, ..., -0.2073, -0.2633, -0.1729], + [-0.0674, -0.2188, 0.1667, ..., -0.2709, 0.2989, 0.1402], + ..., + [-0.2128, 0.1323, 0.0167, ..., 0.2403, -0.2691, -0.3294], + [-0.3483, 0.0704, -0.1674, ..., 0.0700, -0.1284, -0.2473], + [-0.0625, -0.1742, -0.0848, ..., -0.1609, -0.0715, -0.2584]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3504e-08, -9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.1642e-09, 0.0000e+00, ..., -1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 4.6566e-10]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0160, -0.0360, -0.0094, -0.0116, -0.0358, -0.0008, 0.0312, -0.0149, + 0.0529, -0.0115], device='cuda:0'), grad: tensor([-2.2119e-08, 1.1642e-09, 2.3283e-10, 3.7253e-09, 1.3970e-09, + 1.3970e-09, 2.4680e-08, 1.1642e-09, 9.3132e-10, -4.4238e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 250.96, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.3950 re_mapping 0.0026 re_causal 0.0083 /// teacc 99.13 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.1390, -0.2910, -0.0899, ..., -0.0883, 0.1911, 0.2023], + [-0.2692, -0.2391, -0.0697, ..., -0.2074, -0.2633, -0.1729], + [-0.0674, -0.2188, 0.1667, ..., -0.2709, 0.2989, 0.1403], + ..., + [-0.2129, 0.1325, 0.0167, ..., 0.2404, -0.2692, -0.3294], + [-0.3483, 0.0703, -0.1674, ..., 0.0700, -0.1284, -0.2474], + [-0.0625, -0.1742, -0.0848, ..., -0.1609, -0.0716, -0.2584]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.1642e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.4238e-09, + -8.6147e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 2.0955e-09, 0.0000e+00, ..., 1.1642e-09, + 6.5193e-09, 0.0000e+00], + [ 1.1642e-09, 2.5611e-09, 0.0000e+00, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00], + [-1.6298e-09, 0.0000e+00, 0.0000e+00, ..., -2.3283e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0158, -0.0361, -0.0094, -0.0114, -0.0357, -0.0008, 0.0311, -0.0148, + 0.0529, -0.0115], device='cuda:0'), grad: tensor([ 4.1910e-09, -1.3970e-09, -3.4925e-09, -6.2864e-09, 3.0268e-09, + 6.0536e-09, 8.3819e-09, 3.3062e-08, -1.9791e-08, -1.6531e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 250.95, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4197 re_mapping 0.0025 re_causal 0.0086 /// teacc 99.13 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.1390, -0.2911, -0.0900, ..., -0.0883, 0.1912, 0.2024], + [-0.2693, -0.2391, -0.0697, ..., -0.2075, -0.2633, -0.1729], + [-0.0674, -0.2188, 0.1667, ..., -0.2710, 0.2989, 0.1402], + ..., + [-0.2130, 0.1325, 0.0167, ..., 0.2404, -0.2692, -0.3294], + [-0.3484, 0.0703, -0.1674, ..., 0.0700, -0.1284, -0.2474], + [-0.0625, -0.1742, -0.0848, ..., -0.1609, -0.0716, -0.2585]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., -1.8394e-08, + -2.9569e-08, -3.3295e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 6.9849e-10], + ..., + [ 2.3283e-10, -4.4238e-09, 0.0000e+00, ..., -2.0955e-09, + 6.9849e-10, 6.9849e-10], + [ 6.9849e-10, 1.1642e-09, 0.0000e+00, ..., 2.0955e-09, + 2.3283e-10, 2.3283e-10], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 3.0268e-09, + 4.4238e-09, 4.8894e-09]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0158, -0.0361, -0.0095, -0.0115, -0.0357, -0.0008, 0.0311, -0.0147, + 0.0529, -0.0115], device='cuda:0'), grad: tensor([-9.2201e-08, 1.6298e-09, 1.8626e-09, 4.6566e-09, -2.5611e-09, + -1.6298e-09, 7.5903e-08, 3.7486e-08, 6.0536e-09, -2.9569e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 250.65, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.4285 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.13 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.1390, -0.2911, -0.0900, ..., -0.0883, 0.1912, 0.2024], + [-0.2694, -0.2392, -0.0697, ..., -0.2075, -0.2634, -0.1731], + [-0.0675, -0.2189, 0.1667, ..., -0.2711, 0.2990, 0.1404], + ..., + [-0.2130, 0.1326, 0.0167, ..., 0.2404, -0.2693, -0.3295], + [-0.3484, 0.0703, -0.1674, ..., 0.0699, -0.1284, -0.2474], + [-0.0626, -0.1743, -0.0848, ..., -0.1609, -0.0716, -0.2585]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, + -6.9849e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0158, -0.0361, -0.0094, -0.0115, -0.0357, -0.0008, 0.0311, -0.0147, + 0.0529, -0.0115], device='cuda:0'), grad: tensor([ 9.3132e-10, 4.6566e-10, -3.7253e-09, 3.2596e-09, -1.5064e-07, + -3.4925e-09, 1.6298e-09, 2.3283e-09, 6.9849e-10, 1.5320e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 250.35, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4023 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.11 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.1390, -0.2911, -0.0900, ..., -0.0883, 0.1912, 0.2024], + [-0.2694, -0.2392, -0.0697, ..., -0.2075, -0.2634, -0.1731], + [-0.0675, -0.2189, 0.1668, ..., -0.2712, 0.2990, 0.1405], + ..., + [-0.2130, 0.1326, 0.0167, ..., 0.2404, -0.2693, -0.3297], + [-0.3485, 0.0703, -0.1674, ..., 0.0700, -0.1285, -0.2474], + [-0.0626, -0.1743, -0.0848, ..., -0.1610, -0.0716, -0.2585]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -3.4925e-09, -2.7940e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.6298e-09, 2.3283e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + -6.9849e-10, 2.3283e-10], + ..., + [ 0.0000e+00, -2.0955e-09, 0.0000e+00, ..., -4.8894e-09, + 0.0000e+00, 0.0000e+00], + [ 2.0955e-09, 0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 2.5611e-09, 2.0955e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 5.1223e-09, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0158, -0.0361, -0.0094, -0.0115, -0.0357, -0.0008, 0.0311, -0.0147, + 0.0529, -0.0116], device='cuda:0'), grad: tensor([-3.4925e-09, -1.0058e-07, 1.1874e-08, 4.6566e-09, 4.4238e-09, + 1.1176e-08, 9.3132e-10, -3.4925e-09, 6.3097e-08, 1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 250.52, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4229 re_mapping 0.0026 re_causal 0.0087 /// teacc 99.10 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.1390, -0.2912, -0.0900, ..., -0.0883, 0.1912, 0.2024], + [-0.2694, -0.2392, -0.0697, ..., -0.2075, -0.2635, -0.1732], + [-0.0675, -0.2189, 0.1668, ..., -0.2712, 0.2991, 0.1407], + ..., + [-0.2130, 0.1327, 0.0167, ..., 0.2405, -0.2694, -0.3298], + [-0.3485, 0.0703, -0.1674, ..., 0.0698, -0.1285, -0.2475], + [-0.0626, -0.1743, -0.0848, ..., -0.1610, -0.0716, -0.2586]], + device='cuda:0'), grad: tensor([[6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, 4.8894e-09, + 4.4238e-09], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [3.4925e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 2.7940e-09, + 2.5611e-09], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0157, -0.0362, -0.0094, -0.0114, -0.0357, -0.0007, 0.0310, -0.0147, + 0.0528, -0.0116], device='cuda:0'), grad: tensor([ 2.3516e-08, 4.6566e-10, 4.6566e-10, 0.0000e+00, 1.8626e-09, + 2.7344e-06, -2.7716e-06, 6.9849e-10, 1.0012e-08, 2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 250.32, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4334 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.10 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.1391, -0.2912, -0.0900, ..., -0.0882, 0.1913, 0.2025], + [-0.2694, -0.2392, -0.0697, ..., -0.2075, -0.2635, -0.1732], + [-0.0675, -0.2189, 0.1668, ..., -0.2712, 0.2992, 0.1407], + ..., + [-0.2131, 0.1327, 0.0167, ..., 0.2405, -0.2695, -0.3298], + [-0.3486, 0.0702, -0.1674, ..., 0.0698, -0.1286, -0.2475], + [-0.0627, -0.1743, -0.0848, ..., -0.1611, -0.0717, -0.2586]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 6.9849e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + ..., + [ 2.3283e-10, -2.3283e-09, 0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 3.0268e-09, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0157, -0.0362, -0.0093, -0.0113, -0.0357, -0.0006, 0.0310, -0.0147, + 0.0527, -0.0116], device='cuda:0'), grad: tensor([ 2.3283e-10, 3.0268e-09, 9.3132e-10, -3.0268e-09, 0.0000e+00, + 3.7253e-09, 9.3132e-10, -4.6566e-09, -1.3970e-09, 7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 250.43, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4232 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.08 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.1391, -0.2912, -0.0900, ..., -0.0882, 0.1914, 0.2026], + [-0.2695, -0.2392, -0.0697, ..., -0.2075, -0.2636, -0.1733], + [-0.0675, -0.2189, 0.1668, ..., -0.2713, 0.2993, 0.1408], + ..., + [-0.2131, 0.1327, 0.0167, ..., 0.2405, -0.2695, -0.3299], + [-0.3486, 0.0702, -0.1674, ..., 0.0697, -0.1286, -0.2475], + [-0.0627, -0.1744, -0.0848, ..., -0.1611, -0.0717, -0.2586]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-08, -1.6298e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., -0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7474e-08, 1.5832e-08]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0156, -0.0362, -0.0093, -0.0112, -0.0357, -0.0005, 0.0308, -0.0147, + 0.0525, -0.0116], device='cuda:0'), grad: tensor([-6.7055e-08, 4.6566e-09, 2.7940e-09, -8.8476e-09, 1.6717e-07, + 4.6566e-10, 1.3970e-09, 8.8476e-08, 9.3132e-10, -1.9278e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 250.62, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4244 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.1391, -0.2913, -0.0900, ..., -0.0881, 0.1914, 0.2026], + [-0.2696, -0.2393, -0.0697, ..., -0.2076, -0.2637, -0.1734], + [-0.0676, -0.2189, 0.1668, ..., -0.2714, 0.2992, 0.1407], + ..., + [-0.2132, 0.1327, 0.0167, ..., 0.2406, -0.2695, -0.3298], + [-0.3487, 0.0702, -0.1674, ..., 0.0696, -0.1286, -0.2475], + [-0.0628, -0.1744, -0.0848, ..., -0.1611, -0.0717, -0.2587]], + device='cuda:0'), grad: tensor([[ 1.0710e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.0489e-08, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., -4.1910e-09, + -4.6566e-10, -3.2596e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.2596e-09, + 9.3132e-10, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0156, -0.0363, -0.0093, -0.0112, -0.0357, -0.0005, 0.0308, -0.0146, + 0.0525, -0.0116], device='cuda:0'), grad: tensor([ 8.6147e-08, 1.3970e-08, -2.8405e-08, 4.6566e-10, -2.3283e-08, + 2.7940e-09, -7.9162e-08, 4.6566e-09, 2.4680e-08, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 250.63, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4081 re_mapping 0.0025 re_causal 0.0082 /// teacc 99.09 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.1392, -0.2913, -0.0900, ..., -0.0881, 0.1914, 0.2026], + [-0.2696, -0.2393, -0.0696, ..., -0.2076, -0.2637, -0.1734], + [-0.0676, -0.2190, 0.1668, ..., -0.2714, 0.2993, 0.1408], + ..., + [-0.2133, 0.1327, 0.0166, ..., 0.2406, -0.2696, -0.3299], + [-0.3488, 0.0702, -0.1674, ..., 0.0696, -0.1287, -0.2475], + [-0.0628, -0.1744, -0.0848, ..., -0.1612, -0.0717, -0.2587]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 1.3970e-09, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -3.2596e-09, 0.0000e+00, ..., -3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0157, -0.0362, -0.0093, -0.0112, -0.0357, -0.0005, 0.0309, -0.0147, + 0.0525, -0.0116], device='cuda:0'), grad: tensor([ 9.3132e-10, -7.4506e-09, 1.8626e-09, 1.3970e-09, 5.1223e-09, + 4.6566e-10, 7.4506e-09, -5.5879e-09, 3.7253e-09, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 250.79, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4297 re_mapping 0.0026 re_causal 0.0083 /// teacc 99.09 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.1392, -0.2914, -0.0900, ..., -0.0882, 0.1914, 0.2026], + [-0.2696, -0.2394, -0.0696, ..., -0.2077, -0.2637, -0.1733], + [-0.0676, -0.2190, 0.1668, ..., -0.2715, 0.2993, 0.1407], + ..., + [-0.2133, 0.1328, 0.0166, ..., 0.2407, -0.2697, -0.3300], + [-0.3489, 0.0702, -0.1674, ..., 0.0696, -0.1287, -0.2476], + [-0.0628, -0.1745, -0.0848, ..., -0.1612, -0.0717, -0.2587]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 1.3970e-09], + [ 4.6566e-10, 2.3283e-09, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.3283e-09, 0.0000e+00, ..., -2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., -4.6566e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0157, -0.0363, -0.0095, -0.0112, -0.0358, -0.0006, 0.0309, -0.0146, + 0.0525, -0.0116], device='cuda:0'), grad: tensor([ 1.1176e-08, 6.9849e-09, 2.3283e-09, 5.5879e-09, 1.2573e-08, + 3.8650e-08, -6.3796e-08, -6.0536e-09, -8.8476e-09, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 250.50, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4328 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.09 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.1392, -0.2914, -0.0900, ..., -0.0881, 0.1915, 0.2027], + [-0.2696, -0.2394, -0.0696, ..., -0.2077, -0.2637, -0.1733], + [-0.0676, -0.2190, 0.1668, ..., -0.2715, 0.2994, 0.1407], + ..., + [-0.2134, 0.1328, 0.0166, ..., 0.2407, -0.2698, -0.3300], + [-0.3489, 0.0702, -0.1674, ..., 0.0697, -0.1287, -0.2476], + [-0.0628, -0.1745, -0.0848, ..., -0.1613, -0.0718, -0.2588]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 0.0000e+00, ..., 6.9849e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -9.3132e-10, -0.0000e+00], + ..., + [ 0.0000e+00, -6.1467e-08, 0.0000e+00, ..., -5.0757e-08, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 2.3283e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 3.7719e-08, 0.0000e+00, ..., 3.2596e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0156, -0.0363, -0.0094, -0.0112, -0.0357, -0.0006, 0.0309, -0.0146, + 0.0526, -0.0117], device='cuda:0'), grad: tensor([ 2.3283e-09, 3.2596e-08, -2.3283e-09, 5.5879e-08, 1.8626e-09, + -6.0536e-09, 3.2596e-09, -2.2864e-07, 4.6566e-09, 1.4529e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 250.83, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4313 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.11 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.1392, -0.2914, -0.0900, ..., -0.0881, 0.1916, 0.2028], + [-0.2697, -0.2394, -0.0696, ..., -0.2077, -0.2637, -0.1733], + [-0.0677, -0.2190, 0.1668, ..., -0.2717, 0.2994, 0.1407], + ..., + [-0.2134, 0.1329, 0.0166, ..., 0.2408, -0.2699, -0.3301], + [-0.3490, 0.0701, -0.1675, ..., 0.0696, -0.1287, -0.2476], + [-0.0629, -0.1746, -0.0848, ..., -0.1614, -0.0718, -0.2588]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 5.1223e-09, + 1.3970e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., -1.0245e-08, + -1.3970e-09, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0156, -0.0363, -0.0094, -0.0112, -0.0356, -0.0005, 0.0308, -0.0146, + 0.0525, -0.0118], device='cuda:0'), grad: tensor([ 5.5414e-08, -9.9186e-08, 3.7253e-09, 1.3970e-09, -9.5926e-08, + 1.4435e-08, 3.5390e-08, -5.5879e-09, 8.8476e-09, 8.2422e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 250.91, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4448 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.10 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.1392, -0.2915, -0.0900, ..., -0.0881, 0.1916, 0.2029], + [-0.2697, -0.2395, -0.0695, ..., -0.2076, -0.2638, -0.1734], + [-0.0677, -0.2191, 0.1668, ..., -0.2717, 0.2995, 0.1410], + ..., + [-0.2134, 0.1330, 0.0165, ..., 0.2407, -0.2700, -0.3303], + [-0.3490, 0.0701, -0.1675, ..., 0.0696, -0.1288, -0.2477], + [-0.0629, -0.1746, -0.0848, ..., -0.1615, -0.0718, -0.2589]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + -4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, -6.5193e-09, 0.0000e+00, ..., -9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 4.6566e-10], + [ 2.3283e-09, 6.0536e-09, 0.0000e+00, ..., 1.3504e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0155, -0.0363, -0.0093, -0.0111, -0.0355, -0.0005, 0.0307, -0.0146, + 0.0525, -0.0118], device='cuda:0'), grad: tensor([ 1.8626e-09, 7.9162e-09, -2.7940e-09, 8.4750e-08, -4.8429e-08, + -9.0804e-08, -1.8626e-09, -9.3132e-09, 1.3504e-08, 5.4948e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 250.29, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4220 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.11 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.1393, -0.2915, -0.0900, ..., -0.0881, 0.1916, 0.2029], + [-0.2698, -0.2395, -0.0695, ..., -0.2076, -0.2638, -0.1734], + [-0.0677, -0.2190, 0.1668, ..., -0.2718, 0.2996, 0.1410], + ..., + [-0.2135, 0.1330, 0.0165, ..., 0.2407, -0.2701, -0.3304], + [-0.3491, 0.0701, -0.1675, ..., 0.0696, -0.1288, -0.2477], + [-0.0629, -0.1747, -0.0848, ..., -0.1615, -0.0719, -0.2590]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.7229e-08, -1.3970e-08], + [ 1.5460e-07, 4.6566e-10, 0.0000e+00, ..., 6.4727e-08, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.8871e-08, 0.0000e+00, 0.0000e+00, ..., 1.3504e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 6.5193e-09, 4.6566e-09]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0156, -0.0363, -0.0092, -0.0111, -0.0355, -0.0005, 0.0308, -0.0146, + 0.0525, -0.0118], device='cuda:0'), grad: tensor([-2.9802e-08, 3.4133e-07, 1.3970e-09, 4.6566e-10, 5.1223e-09, + 1.0189e-06, -1.4026e-06, 2.7940e-09, 6.4261e-08, 5.5879e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 250.44, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4257 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.1393, -0.2916, -0.0900, ..., -0.0881, 0.1916, 0.2029], + [-0.2699, -0.2395, -0.0695, ..., -0.2076, -0.2639, -0.1734], + [-0.0677, -0.2191, 0.1669, ..., -0.2719, 0.2997, 0.1410], + ..., + [-0.2137, 0.1330, 0.0165, ..., 0.2408, -0.2702, -0.3305], + [-0.3492, 0.0701, -0.1675, ..., 0.0697, -0.1288, -0.2478], + [-0.0629, -0.1748, -0.0848, ..., -0.1616, -0.0719, -0.2591]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -4.1910e-09, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0156, -0.0363, -0.0093, -0.0110, -0.0354, -0.0005, 0.0308, -0.0146, + 0.0525, -0.0119], device='cuda:0'), grad: tensor([ 9.3132e-10, 1.0896e-07, -3.3062e-08, 6.2399e-08, -3.8510e-07, + -6.7055e-08, 5.1223e-09, 2.6543e-08, 3.2596e-09, 2.8545e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 251.08, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4383 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.06 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.1394, -0.2916, -0.0900, ..., -0.0881, 0.1916, 0.2029], + [-0.2699, -0.2396, -0.0695, ..., -0.2076, -0.2639, -0.1734], + [-0.0677, -0.2190, 0.1669, ..., -0.2719, 0.2998, 0.1411], + ..., + [-0.2137, 0.1330, 0.0165, ..., 0.2408, -0.2703, -0.3305], + [-0.3493, 0.0701, -0.1675, ..., 0.0697, -0.1289, -0.2478], + [-0.0630, -0.1748, -0.0848, ..., -0.1617, -0.0720, -0.2592]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, 4.6566e-10], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.3504e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-09, 1.8626e-09]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0156, -0.0362, -0.0092, -0.0109, -0.0354, -0.0006, 0.0308, -0.0147, + 0.0525, -0.0120], device='cuda:0'), grad: tensor([ 4.6566e-10, 3.4738e-07, -1.3504e-08, 2.1886e-08, 1.0710e-08, + 4.4238e-08, 1.1967e-07, 2.3283e-09, -5.5321e-07, 2.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 250.52, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4059 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.11 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.1394, -0.2916, -0.0900, ..., -0.0880, 0.1918, 0.2031], + [-0.2699, -0.2396, -0.0695, ..., -0.2076, -0.2640, -0.1734], + [-0.0677, -0.2190, 0.1669, ..., -0.2719, 0.2998, 0.1410], + ..., + [-0.2137, 0.1330, 0.0165, ..., 0.2408, -0.2704, -0.3306], + [-0.3494, 0.0700, -0.1675, ..., 0.0697, -0.1289, -0.2478], + [-0.0630, -0.1748, -0.0848, ..., -0.1617, -0.0720, -0.2592]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., -4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 3.2596e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0154, -0.0362, -0.0092, -0.0110, -0.0353, -0.0006, 0.0307, -0.0147, + 0.0525, -0.0120], device='cuda:0'), grad: tensor([ 0.0000e+00, 2.3283e-09, 4.6566e-10, 6.9849e-09, 9.3132e-10, + -1.1176e-08, 3.7253e-09, 1.3970e-09, 3.2596e-09, -4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 251.02, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4459 re_mapping 0.0025 re_causal 0.0086 /// teacc 99.11 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.1395, -0.2917, -0.0900, ..., -0.0880, 0.1918, 0.2031], + [-0.2700, -0.2397, -0.0695, ..., -0.2077, -0.2640, -0.1734], + [-0.0677, -0.2190, 0.1669, ..., -0.2720, 0.2999, 0.1411], + ..., + [-0.2138, 0.1331, 0.0165, ..., 0.2409, -0.2705, -0.3307], + [-0.3494, 0.0700, -0.1675, ..., 0.0697, -0.1290, -0.2479], + [-0.0630, -0.1749, -0.0848, ..., -0.1617, -0.0720, -0.2592]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 5.1223e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-09, -4.1910e-09], + ..., + [ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., -4.1910e-09, + 1.3970e-09, 1.3970e-09], + [ 8.8476e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.5146e-08, 1.5367e-08], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 5.1223e-09, 4.6566e-10]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0154, -0.0363, -0.0091, -0.0109, -0.0353, -0.0006, 0.0307, -0.0146, + 0.0525, -0.0121], device='cuda:0'), grad: tensor([ 2.4214e-08, 1.8626e-09, -3.5390e-08, 9.3132e-10, -3.2596e-09, + 2.7940e-09, -1.0803e-07, -4.1910e-09, 9.7789e-08, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 250.80, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4110 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.09 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.1395, -0.2917, -0.0900, ..., -0.0878, 0.1920, 0.2034], + [-0.2700, -0.2397, -0.0695, ..., -0.2077, -0.2640, -0.1734], + [-0.0678, -0.2190, 0.1670, ..., -0.2720, 0.3000, 0.1412], + ..., + [-0.2138, 0.1332, 0.0165, ..., 0.2410, -0.2706, -0.3308], + [-0.3495, 0.0700, -0.1675, ..., 0.0697, -0.1290, -0.2479], + [-0.0631, -0.1750, -0.0848, ..., -0.1619, -0.0721, -0.2593]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.6298e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0152, -0.0363, -0.0090, -0.0109, -0.0349, -0.0006, 0.0306, -0.0146, + 0.0525, -0.0123], device='cuda:0'), grad: tensor([ 4.6566e-10, 9.3132e-10, -1.0803e-07, 5.1223e-09, 1.5367e-08, + -1.8626e-09, 1.8626e-09, 3.7253e-09, 1.1455e-07, -2.7474e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 250.68, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4331 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.09 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.1396, -0.2917, -0.0900, ..., -0.0877, 0.1921, 0.2034], + [-0.2700, -0.2397, -0.0695, ..., -0.2077, -0.2640, -0.1733], + [-0.0678, -0.2190, 0.1670, ..., -0.2721, 0.3001, 0.1411], + ..., + [-0.2139, 0.1333, 0.0165, ..., 0.2410, -0.2706, -0.3308], + [-0.3495, 0.0700, -0.1675, ..., 0.0697, -0.1291, -0.2479], + [-0.0631, -0.1752, -0.0848, ..., -0.1621, -0.0721, -0.2593]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 4.6566e-10], + ..., + [ 0.0000e+00, -1.2107e-08, 0.0000e+00, ..., -1.2573e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.8476e-09, 0.0000e+00, ..., 9.3132e-09, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0152, -0.0363, -0.0091, -0.0109, -0.0348, -0.0006, 0.0306, -0.0146, + 0.0525, -0.0125], device='cuda:0'), grad: tensor([ 4.6566e-10, 4.6566e-09, 6.9849e-09, 2.3283e-09, 1.8626e-09, + 2.3283e-09, -1.4435e-08, -3.3528e-08, -5.1223e-09, 4.0047e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 250.68, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4468 re_mapping 0.0025 re_causal 0.0086 /// teacc 99.08 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.1396, -0.2917, -0.0900, ..., -0.0877, 0.1921, 0.2034], + [-0.2700, -0.2398, -0.0695, ..., -0.2078, -0.2641, -0.1732], + [-0.0678, -0.2190, 0.1670, ..., -0.2721, 0.3003, 0.1411], + ..., + [-0.2139, 0.1333, 0.0165, ..., 0.2411, -0.2708, -0.3309], + [-0.3496, 0.0699, -0.1675, ..., 0.0696, -0.1291, -0.2480], + [-0.0631, -0.1752, -0.0848, ..., -0.1622, -0.0719, -0.2594]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + -1.3970e-09, -9.3132e-10], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.8405e-08, 0.0000e+00, ..., -4.1910e-08, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 2.1886e-08, 0.0000e+00, ..., 2.5611e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 1.1176e-08, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0153, -0.0363, -0.0090, -0.0109, -0.0347, -0.0005, 0.0305, -0.0146, + 0.0523, -0.0125], device='cuda:0'), grad: tensor([-2.7940e-09, 6.5193e-09, 0.0000e+00, 1.3970e-09, 9.3132e-10, + 2.3283e-09, 9.3132e-10, -8.6613e-08, 5.7742e-08, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 250.29, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4373 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.11 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.1397, -0.2917, -0.0900, ..., -0.0876, 0.1921, 0.2035], + [-0.2701, -0.2398, -0.0695, ..., -0.2078, -0.2641, -0.1732], + [-0.0678, -0.2190, 0.1670, ..., -0.2721, 0.3003, 0.1412], + ..., + [-0.2140, 0.1333, 0.0165, ..., 0.2411, -0.2709, -0.3310], + [-0.3498, 0.0699, -0.1675, ..., 0.0695, -0.1292, -0.2480], + [-0.0632, -0.1753, -0.0848, ..., -0.1623, -0.0720, -0.2596]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + -5.5879e-09, -6.0536e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -2.4680e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., -3.2596e-09, + 6.5193e-09, 4.1910e-09], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 3.2596e-09, + 1.3970e-09, 1.3970e-09]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0153, -0.0363, -0.0090, -0.0108, -0.0345, -0.0006, 0.0306, -0.0146, + 0.0523, -0.0126], device='cuda:0'), grad: tensor([-1.2107e-08, 4.6566e-09, -3.5390e-08, 3.1665e-08, 3.2596e-09, + -6.5193e-09, 4.6566e-09, 6.0536e-09, 7.9162e-09, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 250.99, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4509 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.11 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.1397, -0.2918, -0.0900, ..., -0.0876, 0.1922, 0.2036], + [-0.2701, -0.2399, -0.0695, ..., -0.2077, -0.2641, -0.1733], + [-0.0678, -0.2191, 0.1670, ..., -0.2722, 0.3004, 0.1413], + ..., + [-0.2141, 0.1334, 0.0165, ..., 0.2410, -0.2709, -0.3312], + [-0.3499, 0.0698, -0.1675, ..., 0.0695, -0.1293, -0.2481], + [-0.0632, -0.1753, -0.0848, ..., -0.1624, -0.0719, -0.2596]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -3.7253e-09, -4.6566e-10], + ..., + [ 0.0000e+00, -3.7253e-08, -9.3132e-10, ..., -4.7963e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 3.7253e-09, 4.6566e-10], + [ 0.0000e+00, 3.7253e-08, 9.3132e-10, ..., 6.8452e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0153, -0.0362, -0.0090, -0.0104, -0.0344, -0.0007, 0.0305, -0.0146, + 0.0522, -0.0127], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.3039e-08, -2.1420e-08, 9.3132e-10, -1.4482e-07, + 9.3132e-10, 4.1910e-09, -8.3819e-08, 2.5146e-08, 2.0955e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 250.31, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4187 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.13 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.1397, -0.2918, -0.0900, ..., -0.0875, 0.1922, 0.2036], + [-0.2702, -0.2400, -0.0695, ..., -0.2077, -0.2642, -0.1733], + [-0.0678, -0.2191, 0.1670, ..., -0.2722, 0.3005, 0.1414], + ..., + [-0.2142, 0.1335, 0.0165, ..., 0.2411, -0.2710, -0.3312], + [-0.3500, 0.0698, -0.1675, ..., 0.0695, -0.1294, -0.2481], + [-0.0632, -0.1754, -0.0848, ..., -0.1626, -0.0720, -0.2597]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, -0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0152, -0.0363, -0.0089, -0.0104, -0.0342, -0.0007, 0.0305, -0.0146, + 0.0522, -0.0129], device='cuda:0'), grad: tensor([ 4.6566e-10, 6.0536e-09, 4.6566e-10, 1.8626e-09, -5.5414e-08, + 1.3970e-09, -5.5879e-09, -1.3970e-09, 5.1223e-09, 4.6566e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 250.15, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4478 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.10 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.1398, -0.2918, -0.0900, ..., -0.0875, 0.1922, 0.2036], + [-0.2702, -0.2400, -0.0695, ..., -0.2078, -0.2642, -0.1733], + [-0.0678, -0.2190, 0.1670, ..., -0.2722, 0.3006, 0.1414], + ..., + [-0.2144, 0.1335, 0.0165, ..., 0.2411, -0.2711, -0.3313], + [-0.3500, 0.0697, -0.1675, ..., 0.0695, -0.1294, -0.2482], + [-0.0632, -0.1755, -0.0848, ..., -0.1627, -0.0718, -0.2598]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0154, -0.0363, -0.0089, -0.0103, -0.0342, -0.0007, 0.0306, -0.0146, + 0.0521, -0.0128], device='cuda:0'), grad: tensor([ 4.6566e-10, 9.3132e-10, 4.6566e-10, 9.3132e-10, 3.2596e-09, + 9.3132e-10, 0.0000e+00, 2.4680e-08, -2.7940e-09, -2.5611e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 250.51, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4293 re_mapping 0.0024 re_causal 0.0084 /// teacc 99.12 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.1399, -0.2918, -0.0900, ..., -0.0875, 0.1922, 0.2036], + [-0.2702, -0.2400, -0.0695, ..., -0.2078, -0.2643, -0.1733], + [-0.0679, -0.2190, 0.1670, ..., -0.2722, 0.3007, 0.1414], + ..., + [-0.2144, 0.1335, 0.0165, ..., 0.2412, -0.2712, -0.3314], + [-0.3501, 0.0697, -0.1675, ..., 0.0694, -0.1295, -0.2482], + [-0.0633, -0.1755, -0.0848, ..., -0.1627, -0.0719, -0.2599]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 3.1013e-07, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, -2.7940e-08, 0.0000e+00, ..., 0.0000e+00, + -2.8405e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -3.2829e-07, + 9.3132e-10, 0.0000e+00], + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 1.7229e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0154, -0.0364, -0.0088, -0.0103, -0.0342, -0.0007, 0.0306, -0.0145, + 0.0521, -0.0128], device='cuda:0'), grad: tensor([ 9.3132e-10, 8.9081e-07, -2.4401e-07, 2.5891e-07, -8.3819e-09, + -3.2596e-08, 9.3132e-10, -9.3039e-07, 6.9849e-09, 6.0070e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 250.28, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4113 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.11 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.1399, -0.2919, -0.0900, ..., -0.0873, 0.1924, 0.2039], + [-0.2703, -0.2400, -0.0695, ..., -0.2078, -0.2644, -0.1734], + [-0.0679, -0.2190, 0.1670, ..., -0.2723, 0.3008, 0.1416], + ..., + [-0.2145, 0.1335, 0.0165, ..., 0.2413, -0.2713, -0.3315], + [-0.3502, 0.0696, -0.1675, ..., 0.0694, -0.1296, -0.2483], + [-0.0634, -0.1755, -0.0848, ..., -0.1627, -0.0719, -0.2601]], + device='cuda:0'), grad: tensor([[ 1.1409e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.8638e-08, 1.0477e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., -2.3283e-09, + -1.3970e-09, -9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 6.9849e-10, 4.6566e-10], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -2.3283e-10, + 6.2864e-09, 2.3283e-09], + [ 1.6298e-08, 2.3283e-10, 0.0000e+00, ..., 7.9162e-09, + 2.3283e-10, 2.3283e-10]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0152, -0.0364, -0.0088, -0.0102, -0.0342, -0.0006, 0.0304, -0.0145, + 0.0520, -0.0128], device='cuda:0'), grad: tensor([ 5.7509e-08, 6.2864e-09, -9.0804e-09, 4.9826e-07, 2.2817e-08, + -5.4250e-07, -7.4040e-08, 1.1874e-08, 2.4913e-08, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 250.33, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.3858 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.10 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.1400, -0.2919, -0.0900, ..., -0.0873, 0.1924, 0.2039], + [-0.2703, -0.2402, -0.0694, ..., -0.2078, -0.2644, -0.1734], + [-0.0680, -0.2191, 0.1671, ..., -0.2724, 0.3008, 0.1416], + ..., + [-0.2145, 0.1337, 0.0164, ..., 0.2413, -0.2713, -0.3316], + [-0.3503, 0.0696, -0.1675, ..., 0.0694, -0.1297, -0.2483], + [-0.0635, -0.1756, -0.0848, ..., -0.1628, -0.0720, -0.2602]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 3.4925e-09, 0.0000e+00, ..., 6.9849e-10, + 2.5611e-09, 2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-09, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-1.3970e-09, -2.1886e-08, 0.0000e+00, ..., -4.8894e-09, + -1.5134e-08, -9.3132e-10], + [ 2.3283e-10, 2.5611e-09, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-09, 4.6566e-10]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0152, -0.0364, -0.0088, -0.0103, -0.0342, -0.0006, 0.0304, -0.0145, + 0.0520, -0.0128], device='cuda:0'), grad: tensor([ 2.2119e-08, 9.3132e-10, 8.1491e-09, 1.3504e-08, -6.7521e-09, + -1.6298e-09, 7.7765e-08, 1.3970e-09, -1.3039e-07, 2.1653e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 250.21, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4249 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.1400, -0.2919, -0.0900, ..., -0.0873, 0.1924, 0.2039], + [-0.2704, -0.2403, -0.0694, ..., -0.2078, -0.2644, -0.1734], + [-0.0681, -0.2192, 0.1671, ..., -0.2725, 0.3008, 0.1416], + ..., + [-0.2145, 0.1338, 0.0164, ..., 0.2413, -0.2714, -0.3316], + [-0.3504, 0.0695, -0.1675, ..., 0.0694, -0.1297, -0.2483], + [-0.0635, -0.1756, -0.0848, ..., -0.1629, -0.0721, -0.2605]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -9.3132e-10, 0.0000e+00, ..., 1.6298e-09, + -1.2573e-08, -1.2107e-08], + [ 0.0000e+00, 4.6566e-10, -2.0955e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.6298e-09, 0.0000e+00, ..., -0.0000e+00, + -8.3819e-09, -7.2177e-09], + ..., + [ 2.3283e-10, 6.9849e-10, 6.9849e-10, ..., 5.3551e-09, + 1.3970e-09, 1.6298e-09], + [ 5.1223e-09, 4.6566e-10, 0.0000e+00, ..., -9.0804e-09, + 4.4238e-09, 4.6566e-09], + [ 3.7253e-09, 6.9849e-10, -1.1642e-09, ..., 1.2573e-08, + 1.4435e-08, 1.4435e-08]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0152, -0.0364, -0.0089, -0.0103, -0.0341, -0.0006, 0.0304, -0.0145, + 0.0520, -0.0130], device='cuda:0'), grad: tensor([-3.6554e-08, -1.7695e-08, -2.9569e-08, 1.5600e-08, 1.2340e-08, + -4.4005e-08, 1.7695e-08, 3.6554e-08, 6.7521e-09, 4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 250.58, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4545 re_mapping 0.0024 re_causal 0.0087 /// teacc 99.14 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.1400, -0.2919, -0.0900, ..., -0.0872, 0.1925, 0.2040], + [-0.2705, -0.2404, -0.0694, ..., -0.2078, -0.2645, -0.1735], + [-0.0682, -0.2191, 0.1670, ..., -0.2726, 0.3009, 0.1416], + ..., + [-0.2146, 0.1339, 0.0164, ..., 0.2413, -0.2714, -0.3316], + [-0.3505, 0.0695, -0.1675, ..., 0.0693, -0.1298, -0.2484], + [-0.0635, -0.1756, -0.0848, ..., -0.1630, -0.0722, -0.2606]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0152, -0.0364, -0.0088, -0.0103, -0.0340, -0.0005, 0.0303, -0.0145, + 0.0519, -0.0130], device='cuda:0'), grad: tensor([ 6.9849e-10, 0.0000e+00, 1.7928e-08, 0.0000e+00, -1.0477e-08, + 3.7253e-09, 3.4925e-09, 2.7940e-09, -2.9802e-08, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 250.31, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4233 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.13 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.1401, -0.2919, -0.0900, ..., -0.0872, 0.1925, 0.2040], + [-0.2705, -0.2405, -0.0694, ..., -0.2079, -0.2646, -0.1736], + [-0.0682, -0.2192, 0.1671, ..., -0.2726, 0.3010, 0.1416], + ..., + [-0.2146, 0.1340, 0.0164, ..., 0.2414, -0.2715, -0.3316], + [-0.3507, 0.0694, -0.1675, ..., 0.0693, -0.1299, -0.2485], + [-0.0636, -0.1758, -0.0848, ..., -0.1631, -0.0723, -0.2608]], + device='cuda:0'), grad: tensor([[ 1.2610e-06, 2.3283e-10, 0.0000e+00, ..., 1.4226e-07, + 1.6764e-06, 1.4352e-06], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 2.3283e-10], + [ 9.3132e-10, -2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + 1.8626e-09, 1.3970e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 6.9849e-10]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0152, -0.0365, -0.0088, -0.0100, -0.0340, -0.0007, 0.0304, -0.0144, + 0.0519, -0.0132], device='cuda:0'), grad: tensor([ 4.5486e-06, 2.7940e-09, -4.6566e-10, 2.0955e-09, 3.5390e-08, + 9.4762e-08, -4.6715e-06, 3.2596e-09, 9.5461e-09, -3.8883e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 250.34, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4213 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.14 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.1404, -0.2919, -0.0900, ..., -0.0870, 0.1927, 0.2042], + [-0.2705, -0.2405, -0.0694, ..., -0.2079, -0.2646, -0.1736], + [-0.0682, -0.2192, 0.1671, ..., -0.2727, 0.3010, 0.1417], + ..., + [-0.2148, 0.1340, 0.0164, ..., 0.2415, -0.2715, -0.3317], + [-0.3508, 0.0694, -0.1675, ..., 0.0693, -0.1300, -0.2485], + [-0.0637, -0.1758, -0.0848, ..., -0.1632, -0.0723, -0.2610]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.1002e-08, -3.6554e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 3.4925e-09, 2.0955e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 2.3283e-10], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.4680e-08, 1.4901e-08], + [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.9104e-08, 1.7462e-08]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0151, -0.0365, -0.0088, -0.0100, -0.0340, -0.0007, 0.0303, -0.0144, + 0.0518, -0.0132], device='cuda:0'), grad: tensor([-1.2224e-07, 7.4506e-09, 1.3271e-08, 2.3283e-09, -3.1898e-08, + 6.5193e-09, -5.1223e-09, 5.3551e-09, 7.1712e-08, 5.0524e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 250.30, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3898 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.13 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.1404, -0.2920, -0.0900, ..., -0.0869, 0.1928, 0.2044], + [-0.2705, -0.2405, -0.0694, ..., -0.2079, -0.2646, -0.1736], + [-0.0682, -0.2192, 0.1671, ..., -0.2728, 0.3011, 0.1417], + ..., + [-0.2148, 0.1341, 0.0164, ..., 0.2415, -0.2716, -0.3317], + [-0.3509, 0.0693, -0.1675, ..., 0.0692, -0.1300, -0.2486], + [-0.0638, -0.1759, -0.0848, ..., -0.1632, -0.0725, -0.2612]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.4925e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -3.4925e-09, 0.0000e+00, ..., 0.0000e+00, + -9.5461e-09, -5.1223e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0149, -0.0365, -0.0088, -0.0100, -0.0339, -0.0006, 0.0302, -0.0144, + 0.0518, -0.0133], device='cuda:0'), grad: tensor([ 2.2585e-08, 2.3283e-10, -2.2352e-08, 1.1642e-09, 1.3970e-09, + 0.0000e+00, 0.0000e+00, 1.6298e-09, 1.8626e-09, 9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 250.12, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4434 re_mapping 0.0025 re_causal 0.0086 /// teacc 99.13 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.1404, -0.2920, -0.0900, ..., -0.0869, 0.1928, 0.2044], + [-0.2705, -0.2405, -0.0694, ..., -0.2079, -0.2647, -0.1736], + [-0.0683, -0.2192, 0.1671, ..., -0.2729, 0.3011, 0.1417], + ..., + [-0.2148, 0.1341, 0.0163, ..., 0.2415, -0.2716, -0.3317], + [-0.3510, 0.0692, -0.1675, ..., 0.0692, -0.1300, -0.2486], + [-0.0638, -0.1759, -0.0848, ..., -0.1632, -0.0725, -0.2613]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + -2.5611e-09, -1.1642e-09], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 4.4238e-09, 2.2352e-08, 0.0000e+00, ..., -9.0804e-09, + 0.0000e+00, 0.0000e+00], + [ 3.0268e-09, 1.2573e-08, 0.0000e+00, ..., 7.4506e-09, + -1.1642e-09, -6.9849e-10], + [ 2.3283e-09, 9.5461e-09, 0.0000e+00, ..., 2.4913e-08, + 6.9849e-10, 4.6566e-10]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0149, -0.0364, -0.0087, -0.0100, -0.0339, -0.0006, 0.0302, -0.0145, + 0.0517, -0.0133], device='cuda:0'), grad: tensor([-1.6298e-09, 3.0268e-09, 1.1642e-09, -2.7008e-08, 2.3283e-10, + -8.6846e-08, 2.0256e-08, 1.1874e-08, 2.0256e-08, 6.2166e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 251.66, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4234 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.14 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.1404, -0.2920, -0.0900, ..., -0.0868, 0.1929, 0.2044], + [-0.2705, -0.2406, -0.0693, ..., -0.2079, -0.2647, -0.1736], + [-0.0683, -0.2193, 0.1671, ..., -0.2730, 0.3012, 0.1418], + ..., + [-0.2148, 0.1341, 0.0163, ..., 0.2415, -0.2717, -0.3318], + [-0.3511, 0.0692, -0.1676, ..., 0.0692, -0.1301, -0.2487], + [-0.0638, -0.1759, -0.0848, ..., -0.1633, -0.0725, -0.2614]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., -2.3283e-10, + -1.3970e-09, -1.3970e-09], + [ 0.0000e+00, 7.9162e-09, 0.0000e+00, ..., 7.9162e-09, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 2.3283e-10], + ..., + [ 0.0000e+00, -2.1420e-08, 0.0000e+00, ..., -1.8161e-08, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + -3.2596e-09, -1.3970e-09], + [ 2.3283e-10, 1.0943e-08, 0.0000e+00, ..., 9.0804e-09, + 9.3132e-10, 6.9849e-10]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0149, -0.0364, -0.0088, -0.0100, -0.0339, -0.0007, 0.0302, -0.0145, + 0.0517, -0.0133], device='cuda:0'), grad: tensor([-1.6298e-09, 5.8440e-08, 7.4506e-09, 3.9581e-09, -4.2841e-08, + 4.4238e-09, 3.1665e-08, -6.8219e-08, -3.1898e-08, 4.6100e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 251.93, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4380 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.14 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.1404, -0.2920, -0.0900, ..., -0.0868, 0.1929, 0.2045], + [-0.2706, -0.2407, -0.0693, ..., -0.2079, -0.2647, -0.1736], + [-0.0683, -0.2193, 0.1672, ..., -0.2730, 0.3012, 0.1418], + ..., + [-0.2149, 0.1342, 0.0163, ..., 0.2416, -0.2717, -0.3318], + [-0.3513, 0.0692, -0.1676, ..., 0.0692, -0.1301, -0.2487], + [-0.0639, -0.1761, -0.0848, ..., -0.1635, -0.0726, -0.2617]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -2.0955e-09, 0.0000e+00, ..., 2.3283e-10, + -1.6764e-08, -1.7229e-08], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + -3.2596e-09, -4.6566e-10], + ..., + [ 0.0000e+00, -2.0955e-09, 0.0000e+00, ..., -3.0268e-09, + 2.3283e-10, 0.0000e+00], + [ 1.2340e-08, 1.6298e-09, 0.0000e+00, ..., 1.4435e-08, + 4.6566e-10, 0.0000e+00], + [ 1.8626e-09, 3.2596e-09, 0.0000e+00, ..., 4.4238e-09, + 1.4435e-08, 1.4435e-08]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0149, -0.0364, -0.0088, -0.0100, -0.0339, -0.0006, 0.0301, -0.0144, + 0.0517, -0.0134], device='cuda:0'), grad: tensor([-5.5414e-08, 3.0268e-09, -9.0804e-09, 2.6077e-08, 6.0536e-09, + -6.8685e-08, 5.3551e-09, -5.5879e-09, 4.6566e-08, 4.5868e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 251.87, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4438 re_mapping 0.0024 re_causal 0.0085 /// teacc 99.12 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.1405, -0.2920, -0.0900, ..., -0.0868, 0.1929, 0.2045], + [-0.2706, -0.2407, -0.0693, ..., -0.2079, -0.2647, -0.1736], + [-0.0684, -0.2194, 0.1671, ..., -0.2732, 0.3012, 0.1417], + ..., + [-0.2149, 0.1343, 0.0163, ..., 0.2416, -0.2718, -0.3319], + [-0.3514, 0.0691, -0.1676, ..., 0.0692, -0.1302, -0.2488], + [-0.0639, -0.1761, -0.0848, ..., -0.1635, -0.0727, -0.2618]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 6.9849e-10], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 6.9849e-10, + -1.6298e-09, 0.0000e+00], + [ 6.9849e-10, 1.3970e-09, 0.0000e+00, ..., -1.6298e-09, + -3.9581e-09, -3.4925e-09], + ..., + [ 2.3283e-10, -1.3039e-08, 0.0000e+00, ..., -1.6065e-08, + 2.5611e-09, 1.8626e-09], + [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 4.6566e-10, 2.3283e-10], + [ 2.3283e-10, 1.3970e-08, -0.0000e+00, ..., 1.7462e-08, + 6.9849e-10, 4.6566e-10]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0149, -0.0364, -0.0089, -0.0100, -0.0339, -0.0006, 0.0302, -0.0144, + 0.0518, -0.0134], device='cuda:0'), grad: tensor([ 9.7789e-09, -7.8464e-08, -4.8894e-09, -1.3970e-09, 1.2573e-08, + -8.1491e-09, 1.6764e-08, -4.1910e-09, 2.0955e-08, 4.4005e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 251.82, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4136 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.13 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.1405, -0.2920, -0.0900, ..., -0.0868, 0.1930, 0.2045], + [-0.2707, -0.2407, -0.0693, ..., -0.2079, -0.2648, -0.1736], + [-0.0685, -0.2194, 0.1672, ..., -0.2733, 0.3012, 0.1417], + ..., + [-0.2149, 0.1343, 0.0163, ..., 0.2417, -0.2718, -0.3319], + [-0.3514, 0.0691, -0.1676, ..., 0.0692, -0.1302, -0.2488], + [-0.0639, -0.1762, -0.0848, ..., -0.1636, -0.0727, -0.2618]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, + 0.0000e+00]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0149, -0.0364, -0.0089, -0.0100, -0.0339, -0.0006, 0.0302, -0.0145, + 0.0518, -0.0134], device='cuda:0'), grad: tensor([0.0000e+00, 2.3283e-10, 0.0000e+00, 0.0000e+00, 0.0000e+00, 6.9849e-10, + 1.6298e-09, 0.0000e+00, 4.6566e-10, 2.3283e-10], device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 251.88, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4248 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.11 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.1407, -0.2921, -0.0900, ..., -0.0868, 0.1929, 0.2045], + [-0.2707, -0.2407, -0.0693, ..., -0.2079, -0.2648, -0.1736], + [-0.0685, -0.2195, 0.1672, ..., -0.2734, 0.3012, 0.1417], + ..., + [-0.2149, 0.1344, 0.0163, ..., 0.2417, -0.2719, -0.3320], + [-0.3515, 0.0690, -0.1676, ..., 0.0693, -0.1303, -0.2489], + [-0.0639, -0.1763, -0.0848, ..., -0.1637, -0.0728, -0.2620]], + device='cuda:0'), grad: tensor([[-2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.0955e-09, -2.0955e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.5611e-09, 1.6298e-09], + [ 0.0000e+00, -1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + -1.5832e-08, -6.9849e-09], + ..., + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 5.3551e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 2.3283e-10]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0149, -0.0363, -0.0090, -0.0100, -0.0340, -0.0007, 0.0303, -0.0145, + 0.0518, -0.0134], device='cuda:0'), grad: tensor([-2.0955e-09, 1.2573e-08, -7.6136e-08, 2.5611e-09, 1.3970e-09, + 1.1642e-09, 3.4925e-09, 5.8440e-08, 1.3970e-09, 1.6298e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 251.80, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4271 re_mapping 0.0024 re_causal 0.0084 /// teacc 99.13 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.1407, -0.2921, -0.0900, ..., -0.0869, 0.1930, 0.2046], + [-0.2708, -0.2408, -0.0693, ..., -0.2079, -0.2648, -0.1736], + [-0.0686, -0.2195, 0.1672, ..., -0.2734, 0.3012, 0.1417], + ..., + [-0.2150, 0.1344, 0.0163, ..., 0.2417, -0.2719, -0.3320], + [-0.3517, 0.0690, -0.1676, ..., 0.0692, -0.1304, -0.2489], + [-0.0640, -0.1763, -0.0848, ..., -0.1638, -0.0729, -0.2621]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.6997e-08, -1.5367e-08], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, -0.0000e+00], + ..., + [ 0.0000e+00, -9.5461e-09, 0.0000e+00, ..., -1.2573e-08, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 4.6566e-10], + [ 0.0000e+00, 9.0804e-09, 0.0000e+00, ..., 1.1642e-08, + 4.8894e-09, 3.9581e-09]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0149, -0.0364, -0.0090, -0.0098, -0.0340, -0.0006, 0.0302, -0.0145, + 0.0517, -0.0134], device='cuda:0'), grad: tensor([-6.1234e-08, 4.6566e-09, -5.3551e-09, 5.3551e-09, 2.5611e-09, + 3.1665e-08, 1.1642e-08, -2.0256e-08, 6.0536e-09, 3.7486e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 252.38, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4296 re_mapping 0.0024 re_causal 0.0084 /// teacc 99.13 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.1408, -0.2922, -0.0900, ..., -0.0869, 0.1930, 0.2046], + [-0.2709, -0.2413, -0.0693, ..., -0.2083, -0.2650, -0.1736], + [-0.0687, -0.2196, 0.1672, ..., -0.2735, 0.3013, 0.1417], + ..., + [-0.2150, 0.1348, 0.0163, ..., 0.2422, -0.2720, -0.3321], + [-0.3518, 0.0690, -0.1676, ..., 0.0692, -0.1304, -0.2489], + [-0.0640, -0.1764, -0.0848, ..., -0.1638, -0.0730, -0.2623]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 3.2596e-09, 6.9849e-10], + [ 0.0000e+00, 3.5856e-08, 0.0000e+00, ..., 4.2608e-08, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.2608e-08, 0.0000e+00, ..., -5.0291e-08, + 2.3283e-10, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 4.6566e-10, 2.3283e-10], + [-1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + -3.4925e-09, -6.9849e-10]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0149, -0.0368, -0.0092, -0.0096, -0.0341, -0.0007, 0.0302, -0.0140, + 0.0517, -0.0135], device='cuda:0'), grad: tensor([ 2.7707e-08, 1.4668e-07, 2.1420e-08, 4.1910e-09, 4.1910e-09, + 2.5611e-09, 2.3283e-09, -1.7229e-07, -4.4238e-09, -2.7241e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 252.02, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4207 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.13 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.1408, -0.2922, -0.0900, ..., -0.0869, 0.1931, 0.2047], + [-0.2709, -0.2414, -0.0693, ..., -0.2085, -0.2650, -0.1737], + [-0.0687, -0.2195, 0.1673, ..., -0.2736, 0.3016, 0.1418], + ..., + [-0.2151, 0.1350, 0.0163, ..., 0.2423, -0.2722, -0.3322], + [-0.3519, 0.0689, -0.1676, ..., 0.0692, -0.1305, -0.2490], + [-0.0640, -0.1765, -0.0848, ..., -0.1639, -0.0731, -0.2623]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -1.4761e-07, -1.1991e-07], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 2.3283e-09, -1.3271e-08, 0.0000e+00, ..., -1.5832e-08, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [-3.2596e-09, 1.3039e-08, 0.0000e+00, ..., 1.4901e-08, + 6.9849e-10, 4.6566e-10]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0149, -0.0369, -0.0090, -0.0094, -0.0340, -0.0008, 0.0302, -0.0139, + 0.0517, -0.0135], device='cuda:0'), grad: tensor([-2.4331e-07, 4.6566e-09, 4.6566e-10, 1.1642e-09, -7.0315e-08, + 2.5611e-09, 2.5076e-07, -2.0955e-08, 2.0955e-09, 8.7544e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 252.11, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.4364 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.14 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.1410, -0.2922, -0.0900, ..., -0.0869, 0.1931, 0.2046], + [-0.2710, -0.2415, -0.0692, ..., -0.2085, -0.2651, -0.1737], + [-0.0687, -0.2196, 0.1673, ..., -0.2737, 0.3017, 0.1419], + ..., + [-0.2151, 0.1351, 0.0162, ..., 0.2424, -0.2723, -0.3323], + [-0.3520, 0.0689, -0.1676, ..., 0.0691, -0.1306, -0.2490], + [-0.0641, -0.1766, -0.0848, ..., -0.1640, -0.0732, -0.2625]], + device='cuda:0'), grad: tensor([[6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.8894e-09, + 1.3970e-09], + [2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 1.3970e-09, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 6.9849e-10, + 0.0000e+00], + ..., + [2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 1.1642e-09, + 0.0000e+00], + [2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, + 0.0000e+00], + [2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 1.3970e-09, + 0.0000e+00]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0149, -0.0370, -0.0089, -0.0094, -0.0341, -0.0008, 0.0303, -0.0139, + 0.0517, -0.0135], device='cuda:0'), grad: tensor([ 1.9092e-08, 3.4925e-09, 7.2177e-09, -1.1642e-09, -5.4017e-08, + 2.7940e-09, 1.3970e-08, 5.8208e-09, 9.3132e-10, 6.7521e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 252.29, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.4286 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.13 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.1410, -0.2922, -0.0900, ..., -0.0869, 0.1931, 0.2047], + [-0.2711, -0.2415, -0.0691, ..., -0.2085, -0.2651, -0.1737], + [-0.0687, -0.2197, 0.1674, ..., -0.2737, 0.3019, 0.1420], + ..., + [-0.2152, 0.1351, 0.0161, ..., 0.2424, -0.2723, -0.3324], + [-0.3522, 0.0688, -0.1676, ..., 0.0691, -0.1307, -0.2490], + [-0.0641, -0.1767, -0.0848, ..., -0.1642, -0.0732, -0.2626]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, -3.4925e-09], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, -1.1642e-09, 0.0000e+00, ..., -2.3283e-09, + -6.2864e-09, -3.4925e-09], + ..., + [ 2.3283e-10, -1.6601e-07, 0.0000e+00, ..., -1.5902e-07, + 5.3551e-09, 3.2596e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 2.3283e-10], + [ 0.0000e+00, 1.6042e-07, 0.0000e+00, ..., 1.5460e-07, + 4.6566e-09, 3.4925e-09]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0149, -0.0370, -0.0088, -0.0094, -0.0341, -0.0008, 0.0303, -0.0139, + 0.0516, -0.0136], device='cuda:0'), grad: tensor([-1.0710e-08, 4.8894e-09, -2.7474e-08, -5.8208e-09, -3.7253e-09, + 2.1886e-08, 1.1642e-09, -4.2585e-07, 3.2596e-09, 4.4378e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 252.33, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4353 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.1411, -0.2922, -0.0900, ..., -0.0868, 0.1932, 0.2048], + [-0.2712, -0.2416, -0.0691, ..., -0.2086, -0.2652, -0.1737], + [-0.0688, -0.2197, 0.1674, ..., -0.2737, 0.3019, 0.1420], + ..., + [-0.2152, 0.1352, 0.0161, ..., 0.2425, -0.2724, -0.3325], + [-0.3523, 0.0688, -0.1676, ..., 0.0690, -0.1308, -0.2491], + [-0.0641, -0.1768, -0.0848, ..., -0.1643, -0.0733, -0.2627]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, -2.5611e-09], + [ 0.0000e+00, 1.1409e-08, 0.0000e+00, ..., 2.9337e-08, + 1.1642e-09, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -4.6566e-09, -1.3970e-09], + ..., + [ 0.0000e+00, -1.2340e-08, 0.0000e+00, ..., -3.1898e-08, + 9.3132e-10, 2.3283e-10], + [-4.6566e-10, -1.3970e-09, 0.0000e+00, ..., -1.4901e-08, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.1642e-09, + 3.7253e-09, 2.7940e-09]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0148, -0.0370, -0.0088, -0.0094, -0.0341, -0.0009, 0.0303, -0.0138, + 0.0515, -0.0137], device='cuda:0'), grad: tensor([-3.9581e-09, -3.0571e-07, 5.9139e-08, 1.1409e-08, 1.2759e-07, + 6.0536e-09, 4.9826e-08, 6.4261e-08, -4.4238e-08, 3.5390e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 252.07, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4126 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.12 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.1411, -0.2922, -0.0900, ..., -0.0867, 0.1933, 0.2049], + [-0.2712, -0.2416, -0.0689, ..., -0.2085, -0.2653, -0.1738], + [-0.0688, -0.2197, 0.1674, ..., -0.2738, 0.3021, 0.1422], + ..., + [-0.2152, 0.1352, 0.0159, ..., 0.2424, -0.2725, -0.3326], + [-0.3524, 0.0688, -0.1676, ..., 0.0691, -0.1309, -0.2492], + [-0.0641, -0.1769, -0.0848, ..., -0.1644, -0.0734, -0.2629]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-09, 4.6566e-10], + [ 4.6566e-10, 1.1642e-09, 0.0000e+00, ..., 6.9849e-10, + 6.9849e-10, 2.3283e-10], + [ 2.3283e-10, -2.5611e-09, 0.0000e+00, ..., 0.0000e+00, + -9.0804e-09, -3.0268e-09], + ..., + [ 4.6566e-10, -2.0722e-08, 0.0000e+00, ..., -5.0524e-08, + 2.5611e-09, 6.9849e-10], + [ 6.9849e-10, 1.8626e-09, 0.0000e+00, ..., 2.3283e-10, + 2.5611e-09, 9.3132e-10], + [ 2.7940e-09, 2.5379e-08, 0.0000e+00, ..., 4.8429e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0147, -0.0369, -0.0087, -0.0094, -0.0342, -0.0009, 0.0303, -0.0139, + 0.0516, -0.0137], device='cuda:0'), grad: tensor([ 5.3551e-09, 6.9849e-09, -3.0268e-08, -1.6298e-08, 3.2363e-08, + 5.1223e-09, -6.9849e-10, -7.1246e-08, 1.1642e-08, 5.8906e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 250.27, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.3918 re_mapping 0.0025 re_causal 0.0080 /// teacc 99.15 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.1412, -0.2923, -0.0900, ..., -0.0867, 0.1933, 0.2050], + [-0.2713, -0.2416, -0.0689, ..., -0.2086, -0.2653, -0.1738], + [-0.0688, -0.2197, 0.1674, ..., -0.2738, 0.3021, 0.1422], + ..., + [-0.2152, 0.1352, 0.0159, ..., 0.2425, -0.2726, -0.3326], + [-0.3526, 0.0688, -0.1676, ..., 0.0691, -0.1310, -0.2493], + [-0.0641, -0.1769, -0.0848, ..., -0.1646, -0.0733, -0.2629]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -2.0955e-09, -1.5134e-09], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 2.2119e-09, + 1.1642e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -2.0955e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.2387e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., -2.3167e-08, + 1.0477e-09, 6.9849e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 3.4925e-10, + 3.2596e-09, 6.9849e-10]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0148, -0.0369, -0.0086, -0.0093, -0.0341, -0.0010, 0.0303, -0.0139, + 0.0515, -0.0137], device='cuda:0'), grad: tensor([-2.7940e-09, 5.8208e-09, -2.7823e-08, 4.0745e-09, 3.4925e-10, + 2.8987e-08, 4.7730e-09, 1.1292e-08, -3.6671e-08, 2.0256e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 250.50, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4094 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.15 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.1416, -0.2923, -0.0900, ..., -0.0867, 0.1932, 0.2048], + [-0.2714, -0.2416, -0.0689, ..., -0.2086, -0.2653, -0.1738], + [-0.0688, -0.2197, 0.1674, ..., -0.2739, 0.3023, 0.1423], + ..., + [-0.2152, 0.1352, 0.0159, ..., 0.2426, -0.2727, -0.3328], + [-0.3526, 0.0688, -0.1676, ..., 0.0693, -0.1310, -0.2493], + [-0.0642, -0.1770, -0.0848, ..., -0.1646, -0.0734, -0.2631]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., -1.1642e-09, + -3.2713e-08, -2.9686e-08], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 4.4238e-09, + -2.4447e-09, 4.6566e-10], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 2.3283e-10, + 4.3074e-09, 1.9791e-09], + ..., + [ 0.0000e+00, -4.3074e-09, 0.0000e+00, ..., -2.6776e-09, + 1.2806e-09, 8.1491e-10], + [ 1.1642e-10, 2.5611e-09, 0.0000e+00, ..., 1.7462e-09, + 1.4901e-08, 1.4086e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 6.7521e-09, 5.9372e-09]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0150, -0.0369, -0.0084, -0.0094, -0.0342, -0.0010, 0.0304, -0.0139, + 0.0517, -0.0138], device='cuda:0'), grad: tensor([-8.0210e-08, -1.2806e-09, 7.3807e-08, 7.4506e-09, -8.5915e-08, + 7.9162e-09, 2.6543e-08, 1.1409e-08, 4.0629e-08, 1.0477e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 250.49, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4578 re_mapping 0.0024 re_causal 0.0085 /// teacc 99.18 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.1416, -0.2923, -0.0900, ..., -0.0865, 0.1935, 0.2051], + [-0.2714, -0.2416, -0.0688, ..., -0.2086, -0.2654, -0.1739], + [-0.0688, -0.2198, 0.1674, ..., -0.2740, 0.3023, 0.1424], + ..., + [-0.2153, 0.1352, 0.0158, ..., 0.2426, -0.2728, -0.3329], + [-0.3528, 0.0688, -0.1676, ..., 0.0692, -0.1311, -0.2494], + [-0.0642, -0.1771, -0.0848, ..., -0.1648, -0.0735, -0.2632]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 8.1491e-10, 0.0000e+00, ..., 1.3970e-09, + -2.8056e-08, -2.2817e-08], + [ 2.3283e-10, 5.3085e-08, 0.0000e+00, ..., 9.3714e-08, + 5.8208e-10, 4.6566e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + -6.9849e-10, 1.1642e-10], + ..., + [ 2.3283e-10, -6.2049e-08, 0.0000e+00, ..., -1.1094e-07, + 9.3132e-10, 1.1642e-10], + [ 4.6566e-10, 6.9849e-10, 0.0000e+00, ..., 5.8208e-10, + 4.6566e-10, 3.4925e-10], + [ 1.1642e-10, 3.7253e-09, 0.0000e+00, ..., 6.4028e-09, + 6.4028e-09, 5.2387e-09]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0147, -0.0369, -0.0084, -0.0094, -0.0341, -0.0009, 0.0302, -0.0139, + 0.0516, -0.0139], device='cuda:0'), grad: tensor([-6.1002e-08, 2.0408e-07, -2.3283e-10, 1.9558e-08, -2.3283e-09, + 9.3132e-10, 4.2375e-08, -2.3213e-07, 3.9581e-09, 3.4110e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 250.41, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4151 re_mapping 0.0025 re_causal 0.0082 /// teacc 99.14 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.1416, -0.2923, -0.0900, ..., -0.0866, 0.1935, 0.2052], + [-0.2716, -0.2416, -0.0688, ..., -0.2087, -0.2655, -0.1739], + [-0.0688, -0.2197, 0.1674, ..., -0.2740, 0.3024, 0.1424], + ..., + [-0.2153, 0.1353, 0.0158, ..., 0.2427, -0.2729, -0.3330], + [-0.3530, 0.0687, -0.1676, ..., 0.0692, -0.1313, -0.2495], + [-0.0642, -0.1772, -0.0848, ..., -0.1649, -0.0733, -0.2632]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, -1.1642e-10, 0.0000e+00, ..., 5.8208e-10, + -9.4296e-09, -7.4506e-09], + [ 1.3458e-07, 1.1642e-10, 0.0000e+00, ..., 4.4471e-08, + 1.1642e-10, 1.1642e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 4.6566e-10, 3.4925e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 1.1642e-10], + [ 3.2480e-08, 1.1642e-10, 0.0000e+00, ..., 1.0710e-08, + 4.6566e-10, 2.3283e-10], + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 5.8208e-10, + 7.9162e-09, 6.2864e-09]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0148, -0.0369, -0.0083, -0.0094, -0.0340, -0.0010, 0.0303, -0.0139, + 0.0515, -0.0140], device='cuda:0'), grad: tensor([-1.7812e-08, 1.8859e-08, 6.7404e-08, 2.3283e-09, -4.6566e-10, + 1.0170e-06, -1.3569e-06, 1.6915e-07, 7.9046e-08, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 250.37, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4489 re_mapping 0.0024 re_causal 0.0085 /// teacc 99.13 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.1417, -0.2924, -0.0900, ..., -0.0866, 0.1935, 0.2052], + [-0.2717, -0.2416, -0.0688, ..., -0.2087, -0.2655, -0.1740], + [-0.0689, -0.2198, 0.1674, ..., -0.2742, 0.3024, 0.1424], + ..., + [-0.2153, 0.1353, 0.0158, ..., 0.2427, -0.2730, -0.3330], + [-0.3531, 0.0687, -0.1676, ..., 0.0693, -0.1314, -0.2495], + [-0.0643, -0.1773, -0.0848, ..., -0.1651, -0.0734, -0.2634]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + -8.1491e-10, -0.0000e+00], + [ 3.4925e-10, 1.1642e-10, -1.1642e-10, ..., 1.1642e-10, + 4.6566e-10, 6.9849e-10], + [ 3.4925e-10, 2.3283e-10, 1.1642e-10, ..., 3.4925e-10, + 3.4925e-10, 5.8208e-10], + ..., + [ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 5.8208e-10], + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.1642e-10, -1.6298e-09], + [ 2.8522e-08, 4.6566e-10, 0.0000e+00, ..., 1.2806e-09, + 3.5390e-08, 5.0175e-08]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0148, -0.0369, -0.0084, -0.0092, -0.0338, -0.0011, 0.0303, -0.0139, + 0.0516, -0.0142], device='cuda:0'), grad: tensor([ 2.4098e-08, -4.8894e-09, 1.8394e-08, 2.5611e-09, -5.0291e-07, + 1.5134e-09, 4.0047e-08, 5.1223e-09, -1.9674e-08, 4.4215e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 250.46, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4220 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.13 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.1417, -0.2924, -0.0900, ..., -0.0865, 0.1936, 0.2053], + [-0.2717, -0.2416, -0.0688, ..., -0.2087, -0.2656, -0.1740], + [-0.0689, -0.2198, 0.1674, ..., -0.2742, 0.3025, 0.1424], + ..., + [-0.2154, 0.1353, 0.0158, ..., 0.2427, -0.2730, -0.3331], + [-0.3532, 0.0686, -0.1676, ..., 0.0693, -0.1314, -0.2495], + [-0.0643, -0.1774, -0.0849, ..., -0.1653, -0.0735, -0.2635]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 1.6298e-09, 0.0000e+00, ..., 6.6357e-09, + 0.0000e+00, 1.5134e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.3074e-09, + 0.0000e+00, 1.1642e-09], + ..., + [ 0.0000e+00, -3.0268e-09, 0.0000e+00, ..., -1.2689e-08, + 0.0000e+00, -3.1432e-09], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 5.8208e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 3.4925e-10]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0147, -0.0369, -0.0084, -0.0091, -0.0338, -0.0012, 0.0303, -0.0139, + 0.0517, -0.0142], device='cuda:0'), grad: tensor([ 1.0477e-09, 1.6764e-08, 1.1642e-08, 1.3970e-09, 9.3132e-10, + 0.0000e+00, 1.1642e-10, -3.0035e-08, 9.3132e-10, 5.2387e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 249.88, cls_loss 0.0008 cls_loss_mapping 0.0003 cls_loss_causal 0.4342 re_mapping 0.0024 re_causal 0.0083 /// teacc 99.13 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.1417, -0.2924, -0.0900, ..., -0.0864, 0.1937, 0.2054], + [-0.2718, -0.2416, -0.0687, ..., -0.2086, -0.2657, -0.1740], + [-0.0690, -0.2199, 0.1675, ..., -0.2744, 0.3026, 0.1424], + ..., + [-0.2154, 0.1353, 0.0157, ..., 0.2427, -0.2731, -0.3331], + [-0.3534, 0.0686, -0.1676, ..., 0.0693, -0.1315, -0.2496], + [-0.0643, -0.1775, -0.0849, ..., -0.1654, -0.0736, -0.2637]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.5461e-09, -7.3342e-09], + [ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, + 1.1642e-10, 1.1642e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-10, 1.1642e-10, 0.0000e+00, ..., -2.0955e-09, + 1.1642e-10, 1.1642e-10], + [ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 1.6298e-09, 1.2806e-09]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0146, -0.0369, -0.0082, -0.0090, -0.0337, -0.0013, 0.0302, -0.0139, + 0.0516, -0.0143], device='cuda:0'), grad: tensor([-1.7928e-08, -3.0501e-08, 4.5402e-09, 1.6741e-07, 4.6566e-10, + -1.5786e-07, 1.5600e-08, 2.5495e-08, -9.3132e-09, 6.6357e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 249.75, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4177 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.14 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.1420, -0.2924, -0.0900, ..., -0.0865, 0.1936, 0.2053], + [-0.2719, -0.2416, -0.0687, ..., -0.2087, -0.2657, -0.1739], + [-0.0690, -0.2199, 0.1675, ..., -0.2745, 0.3027, 0.1424], + ..., + [-0.2155, 0.1353, 0.0157, ..., 0.2428, -0.2731, -0.3331], + [-0.3537, 0.0686, -0.1676, ..., 0.0691, -0.1316, -0.2497], + [-0.0642, -0.1776, -0.0849, ..., -0.1655, -0.0736, -0.2638]], + device='cuda:0'), grad: tensor([[ 8.1491e-09, 2.3283e-10, 0.0000e+00, ..., 4.8894e-09, + 1.4203e-08, 5.8208e-09], + [ 1.2806e-08, 2.5611e-09, 0.0000e+00, ..., 4.6566e-10, + 1.6764e-08, 7.9162e-09], + [ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + -6.0536e-09, -4.6566e-10], + ..., + [ 5.5879e-09, 3.4925e-09, 0.0000e+00, ..., 1.3970e-09, + 6.9849e-10, 0.0000e+00], + [ 6.7521e-08, 1.1874e-08, 0.0000e+00, ..., 3.2829e-08, + -1.6531e-08, -8.1491e-09], + [ 2.2585e-08, 4.1910e-09, 0.0000e+00, ..., 7.2177e-09, + 3.4925e-09, 1.6298e-09]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0148, -0.0369, -0.0083, -0.0090, -0.0337, -0.0011, 0.0302, -0.0139, + 0.0513, -0.0144], device='cuda:0'), grad: tensor([ 8.8708e-08, 7.6834e-08, -7.4506e-09, 6.8452e-07, -7.2177e-09, + -1.1483e-06, 1.5204e-07, 2.4680e-08, 6.0536e-08, 8.2189e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 250.23, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4486 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.12 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.1420, -0.2925, -0.0900, ..., -0.0865, 0.1937, 0.2053], + [-0.2720, -0.2416, -0.0687, ..., -0.2088, -0.2659, -0.1740], + [-0.0690, -0.2199, 0.1675, ..., -0.2745, 0.3029, 0.1424], + ..., + [-0.2156, 0.1353, 0.0157, ..., 0.2429, -0.2732, -0.3332], + [-0.3538, 0.0685, -0.1676, ..., 0.0691, -0.1317, -0.2498], + [-0.0643, -0.1777, -0.0849, ..., -0.1656, -0.0737, -0.2640]], + device='cuda:0'), grad: tensor([[-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -2.8871e-08, -2.0489e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 4.6566e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 4.6566e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 8.1956e-08, 0.0000e+00, 0.0000e+00, ..., 5.3318e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 1.8394e-08, 1.1642e-08]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0147, -0.0369, -0.0083, -0.0088, -0.0337, -0.0012, 0.0302, -0.0139, + 0.0513, -0.0145], device='cuda:0'), grad: tensor([-8.4983e-08, 3.7253e-09, 2.0955e-09, 2.5937e-07, -1.7462e-08, + -6.7800e-07, 2.9942e-07, 1.3970e-09, 1.5483e-07, 7.0781e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 249.84, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4336 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.12 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.1422, -0.2925, -0.0900, ..., -0.0865, 0.1936, 0.2053], + [-0.2721, -0.2417, -0.0687, ..., -0.2088, -0.2660, -0.1740], + [-0.0690, -0.2199, 0.1676, ..., -0.2745, 0.3031, 0.1425], + ..., + [-0.2157, 0.1353, 0.0157, ..., 0.2430, -0.2734, -0.3333], + [-0.3539, 0.0685, -0.1676, ..., 0.0691, -0.1318, -0.2499], + [-0.0643, -0.1778, -0.0849, ..., -0.1657, -0.0736, -0.2641]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, + 2.3283e-10], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 6.9849e-10, + 4.6566e-10]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0149, -0.0369, -0.0083, -0.0088, -0.0336, -0.0012, 0.0302, -0.0139, + 0.0514, -0.0145], device='cuda:0'), grad: tensor([ 2.3283e-10, 4.6566e-10, 9.3132e-10, 4.6566e-10, -6.9849e-09, + 4.6566e-10, 2.3283e-10, 1.6298e-09, 1.6298e-09, 6.7521e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 250.06, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4281 re_mapping 0.0024 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps2', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps2/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.940002 98.900002 ... 81.913307 71.451877 +ShearY 98.830002 98.769997 ... 81.913307 70.115474 +AutoContrast 99.010002 99.110001 ... 81.913307 62.938438 +Invert 98.889999 81.290001 ... 81.913307 56.000196 +Equalize 98.400002 97.909996 ... 81.913307 71.951622 +Solarize 98.379997 96.630005 ... 81.913307 62.680078 +SolarizeAdd 98.529999 96.599998 ... 81.913307 68.729369 +Posterize 98.959999 99.029999 ... 81.913307 76.495037 +Contrast 99.119995 99.129997 ... 81.913307 69.983360 +Color 99.059998 99.180000 ... 81.913307 61.149043 +Brightness 99.040001 99.139999 ... 81.913307 70.621150 +Sharpness 99.029999 99.049995 ... 81.913307 71.066257 +NoiseSalt 98.900002 99.139999 ... 81.913307 60.974798 +NoiseGaussian 98.959999 99.180000 ... 81.913307 59.648846 +w/o do (original x) 99.180000 0.000000 ... 0.000000 65.803656 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.12 66.318377 78.657927 77.336962 86.596911 77.227544 diff --git a/Meta-causal/code-withStyleAttack/66565.error b/Meta-causal/code-withStyleAttack/66565.error new file mode 100644 index 0000000000000000000000000000000000000000..5b09d3d759b1a068653d824f401cf6b8e10ef88f --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66565.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 37: eduler: command not found diff --git a/Meta-causal/code-withStyleAttack/66565.log b/Meta-causal/code-withStyleAttack/66565.log new file mode 100644 index 0000000000000000000000000000000000000000..74162796e55ad743cfe7e8ef93c30c5c10cc7942 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66565.log @@ -0,0 +1,14131 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0094, -0.0244, 0.0308, ..., 0.0211, 0.0115, 0.0051], + [-0.0293, 0.0002, 0.0169, ..., 0.0102, -0.0280, 0.0046], + [ 0.0241, 0.0062, 0.0289, ..., 0.0189, -0.0016, -0.0282], + ..., + [ 0.0229, -0.0107, -0.0058, ..., 0.0187, 0.0176, -0.0031], + [ 0.0023, -0.0012, 0.0263, ..., -0.0122, -0.0144, 0.0034], + [-0.0090, -0.0070, 0.0110, ..., 0.0217, 0.0004, -0.0025]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0112, -0.0106, -0.0244, -0.0266, -0.0124, 0.0053, 0.0089, -0.0239, + -0.0148, 0.0090], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 223.10, cls_loss 1.1838 cls_loss_mapping 1.7620 cls_loss_causal 2.2054 re_mapping 0.1758 re_causal 0.1904 /// teacc 87.96 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0102, -0.0302, 0.0347, ..., 0.0147, 0.0101, 0.0058], + [-0.0288, 0.0037, 0.0112, ..., 0.0028, -0.0296, 0.0049], + [ 0.0232, 0.0029, 0.0223, ..., 0.0115, -0.0036, -0.0281], + ..., + [ 0.0221, -0.0058, -0.0028, ..., 0.0245, 0.0157, -0.0037], + [ 0.0014, -0.0071, 0.0280, ..., -0.0174, -0.0159, 0.0027], + [-0.0098, -0.0099, 0.0150, ..., 0.0273, -0.0014, -0.0019]], + device='cuda:0'), grad: tensor([[ 0.0000, -0.0006, -0.0244, ..., -0.0132, 0.0000, 0.0000], + [ 0.0000, -0.0006, 0.0034, ..., 0.0035, 0.0000, 0.0000], + [ 0.0000, 0.0057, 0.0125, ..., 0.0074, 0.0000, 0.0000], + ..., + [ 0.0000, -0.0370, -0.0328, ..., -0.0488, 0.0000, 0.0000], + [ 0.0000, 0.0048, -0.0117, ..., 0.0083, 0.0000, 0.0000], + [ 0.0000, 0.0233, 0.0095, ..., 0.0180, 0.0000, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0133, -0.0088, -0.0258, -0.0262, -0.0123, 0.0054, 0.0089, -0.0243, + -0.0152, 0.0095], device='cuda:0'), grad: tensor([-0.0223, 0.0052, -0.0102, 0.0014, 0.0158, 0.0041, 0.0057, -0.0280, + 0.0059, 0.0224], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 222.14, cls_loss 0.3420 cls_loss_mapping 0.6896 cls_loss_causal 1.8949 re_mapping 0.2106 re_causal 0.2836 /// teacc 93.35 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0102, -0.0333, 0.0370, ..., 0.0127, 0.0100, 0.0040], + [-0.0288, 0.0042, 0.0085, ..., -0.0009, -0.0296, 0.0020], + [ 0.0232, 0.0022, 0.0201, ..., 0.0104, -0.0037, -0.0317], + ..., + [ 0.0221, -0.0019, -0.0017, ..., 0.0269, 0.0157, -0.0065], + [ 0.0014, -0.0097, 0.0313, ..., -0.0202, -0.0159, 0.0017], + [-0.0098, -0.0135, 0.0188, ..., 0.0291, -0.0014, -0.0024]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.3562e-04, -1.3481e-02, ..., 1.4420e-03, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.4663e-04, -1.0521e-02, ..., 3.3236e-04, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.9564e-04, 4.7150e-03, ..., 2.2984e-03, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -5.7411e-03, 7.5054e-04, ..., -4.2439e-05, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.5196e-03, 2.0004e-02, ..., 9.8267e-03, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.3160e-03, -1.0891e-03, ..., 8.2626e-03, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0136, -0.0088, -0.0263, -0.0266, -0.0123, 0.0060, 0.0084, -0.0246, + -0.0148, 0.0101], device='cuda:0'), grad: tensor([-0.0090, -0.0100, 0.0161, -0.0004, -0.0300, -0.0151, 0.0059, 0.0083, + 0.0104, 0.0238], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 222.14, cls_loss 0.2206 cls_loss_mapping 0.4132 cls_loss_causal 1.6410 re_mapping 0.1505 re_causal 0.2434 /// teacc 95.27 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0102, -0.0355, 0.0397, ..., 0.0110, 0.0100, 0.0043], + [-0.0288, 0.0033, 0.0065, ..., -0.0033, -0.0296, 0.0016], + [ 0.0232, 0.0019, 0.0186, ..., 0.0091, -0.0037, -0.0322], + ..., + [ 0.0221, 0.0012, -0.0015, ..., 0.0283, 0.0157, -0.0069], + [ 0.0014, -0.0113, 0.0332, ..., -0.0215, -0.0160, 0.0015], + [-0.0098, -0.0165, 0.0203, ..., 0.0301, -0.0014, -0.0025]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.3035e-04, -4.9553e-03, ..., 7.9489e-04, + 0.0000e+00, 1.5929e-05], + [ 0.0000e+00, 3.3784e-04, -2.4147e-03, ..., 1.3103e-03, + 0.0000e+00, 4.8950e-06], + [ 0.0000e+00, 2.2662e-04, -6.0081e-03, ..., 1.0223e-03, + 0.0000e+00, 4.8205e-06], + ..., + [ 0.0000e+00, -4.8876e-04, 2.9030e-03, ..., -3.5572e-04, + 0.0000e+00, 2.2855e-06], + [ 0.0000e+00, 1.3504e-03, 7.3395e-03, ..., 5.1498e-03, + 0.0000e+00, -1.3342e-03], + [ 0.0000e+00, 1.2960e-03, -1.7517e-02, ..., -1.7960e-02, + 0.0000e+00, 1.3202e-05]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0129, -0.0087, -0.0266, -0.0261, -0.0122, 0.0059, 0.0081, -0.0248, + -0.0148, 0.0098], device='cuda:0'), grad: tensor([ 0.0034, -0.0030, -0.0192, 0.0032, 0.0040, -0.0036, 0.0095, 0.0063, + 0.0117, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 222.10, cls_loss 0.1628 cls_loss_mapping 0.2806 cls_loss_causal 1.4610 re_mapping 0.1172 re_causal 0.2083 /// teacc 96.23 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0102, -0.0371, 0.0407, ..., 0.0088, 0.0085, 0.0027], + [-0.0288, 0.0017, 0.0046, ..., -0.0058, -0.0314, -0.0096], + [ 0.0232, 0.0019, 0.0177, ..., 0.0077, -0.0053, -0.0411], + ..., + [ 0.0221, 0.0029, -0.0011, ..., 0.0296, 0.0140, -0.0111], + [ 0.0014, -0.0124, 0.0343, ..., -0.0232, -0.0198, 0.0009], + [-0.0098, -0.0180, 0.0221, ..., 0.0316, -0.0029, -0.0077]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.4312e-04, 1.6203e-03, ..., 1.5535e-03, + 0.0000e+00, 3.6502e-04], + [ 0.0000e+00, 2.3639e-04, -1.1816e-03, ..., 5.6601e-04, + 0.0000e+00, 8.8155e-05], + [ 0.0000e+00, 1.7805e-03, 2.1248e-03, ..., 1.9646e-03, + 0.0000e+00, 1.6415e-04], + ..., + [ 0.0000e+00, -2.6798e-03, 1.2650e-02, ..., 4.8103e-03, + 0.0000e+00, 1.2474e-03], + [ 0.0000e+00, 1.2999e-03, 7.7782e-03, ..., 5.3101e-03, + 0.0000e+00, 5.9986e-04], + [ 0.0000e+00, 2.3403e-03, -2.6749e-02, ..., -2.0981e-02, + 0.0000e+00, -1.9569e-03]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0128, -0.0091, -0.0265, -0.0258, -0.0123, 0.0058, 0.0079, -0.0247, + -0.0145, 0.0098], device='cuda:0'), grad: tensor([ 0.0049, -0.0069, 0.0021, -0.0152, 0.0052, 0.0053, 0.0014, 0.0087, + 0.0106, -0.0161], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 222.46, cls_loss 0.1178 cls_loss_mapping 0.2042 cls_loss_causal 1.3494 re_mapping 0.0931 re_causal 0.1854 /// teacc 96.63 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0102, -0.0377, 0.0420, ..., 0.0074, 0.0108, 0.0013], + [-0.0288, 0.0008, 0.0034, ..., -0.0066, -0.0332, -0.0179], + [ 0.0232, 0.0011, 0.0173, ..., 0.0065, -0.0113, -0.0469], + ..., + [ 0.0221, 0.0053, -0.0012, ..., 0.0306, 0.0100, -0.0171], + [ 0.0014, -0.0133, 0.0352, ..., -0.0243, -0.0199, 0.0053], + [-0.0098, -0.0200, 0.0240, ..., 0.0326, -0.0083, -0.0096]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.0282e-04, -5.1117e-04, ..., 1.9741e-04, + -3.6335e-04, 3.5167e-05], + [ 0.0000e+00, 5.9986e-04, 1.0345e-02, ..., 1.1072e-03, + 4.6417e-06, 1.5659e-03], + [ 0.0000e+00, 1.9169e-03, 7.7591e-03, ..., 2.5730e-03, + 9.7036e-05, 7.6771e-05], + ..., + [ 0.0000e+00, -3.6316e-03, -2.2903e-02, ..., -7.1907e-03, + 5.6893e-05, -9.1743e-04], + [ 0.0000e+00, 1.2264e-03, -1.7147e-03, ..., 2.7752e-03, + 5.6714e-05, -1.0956e-02], + [ 0.0000e+00, 1.1806e-03, 1.5430e-03, ..., -3.5839e-03, + 6.1214e-05, 1.1034e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0128, -0.0090, -0.0265, -0.0258, -0.0120, 0.0053, 0.0076, -0.0248, + -0.0144, 0.0101], device='cuda:0'), grad: tensor([ 0.0010, 0.0238, -0.0146, -0.0152, 0.0044, 0.0085, 0.0042, -0.0163, + 0.0046, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 222.53, cls_loss 0.1038 cls_loss_mapping 0.1739 cls_loss_causal 1.2566 re_mapping 0.0784 re_causal 0.1645 /// teacc 97.31 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0102, -0.0384, 0.0434, ..., 0.0062, 0.0153, 0.0021], + [-0.0288, 0.0005, 0.0026, ..., -0.0087, -0.0386, -0.0227], + [ 0.0232, 0.0002, 0.0165, ..., 0.0048, -0.0176, -0.0535], + ..., + [ 0.0221, 0.0068, -0.0014, ..., 0.0317, 0.0040, -0.0231], + [ 0.0014, -0.0133, 0.0366, ..., -0.0250, -0.0232, 0.0117], + [-0.0098, -0.0219, 0.0251, ..., 0.0333, -0.0145, -0.0131]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.7881e-03, -4.2877e-03, ..., 2.2125e-04, + 1.0414e-02, 4.4227e-05], + [ 0.0000e+00, 2.7919e-04, 1.3857e-03, ..., 1.9670e-04, + 2.4247e-04, 2.6569e-03], + [ 0.0000e+00, 4.4327e-03, 4.9744e-03, ..., 4.5853e-03, + 7.8964e-04, 2.3270e-03], + ..., + [ 0.0000e+00, -7.0381e-03, -3.9024e-03, ..., -8.3694e-03, + 3.3855e-04, 8.6784e-04], + [ 0.0000e+00, 4.2295e-04, -3.5324e-03, ..., -4.2677e-04, + 2.3890e-04, 1.1196e-03], + [ 0.0000e+00, 1.8501e-03, 2.3727e-03, ..., 2.0180e-03, + 2.3007e-04, 1.9274e-03]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0128, -0.0089, -0.0266, -0.0257, -0.0118, 0.0051, 0.0076, -0.0250, + -0.0140, 0.0098], device='cuda:0'), grad: tensor([ 0.0036, 0.0059, 0.0101, 0.0082, 0.0014, -0.0275, -0.0049, -0.0062, + 0.0035, 0.0059], device='cuda:0') +100 +0.0001 +changing lr +epoch 6, time 221.76, cls_loss 0.1073 cls_loss_mapping 0.1726 cls_loss_causal 1.2434 re_mapping 0.0640 re_causal 0.1467 /// teacc 97.28 lr 0.00010000 +Epoch 8, weight, value: tensor([[-1.0243e-02, -3.9061e-02, 4.4652e-02, ..., 5.0099e-03, + 1.7334e-02, 2.3563e-03], + [-2.8781e-02, 4.5420e-05, 1.5794e-03, ..., -9.1317e-03, + -4.5100e-02, -2.4899e-02], + [ 2.3198e-02, 4.5113e-05, 1.5047e-02, ..., 3.7007e-03, + -2.2945e-02, -6.0259e-02], + ..., + [ 2.2054e-02, 8.2435e-03, -1.5907e-03, ..., 3.2357e-02, + -3.4719e-03, -2.6924e-02], + [ 1.4234e-03, -1.4464e-02, 3.8041e-02, ..., -2.5661e-02, + -2.9706e-02, 1.5350e-02], + [-9.8281e-03, -2.2710e-02, 2.5992e-02, ..., 3.4042e-02, + -2.3881e-02, -1.5346e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3997e-05, 6.1493e-03, ..., 3.3417e-03, + -9.2983e-06, 3.4695e-03], + [ 0.0000e+00, 2.1029e-04, 4.6468e-04, ..., 3.7789e-04, + 1.6636e-07, 1.3638e-04], + [ 0.0000e+00, 4.2892e-04, 3.8929e-03, ..., 1.2379e-03, + 3.1888e-06, 1.6861e-03], + ..., + [ 0.0000e+00, -7.6628e-04, 7.6218e-03, ..., 2.9945e-03, + 2.5518e-06, 4.9782e-03], + [ 0.0000e+00, 1.6892e-04, -4.7607e-02, ..., -1.8585e-02, + 3.1339e-07, -3.2349e-02], + [ 0.0000e+00, 8.4877e-04, 1.2276e-02, ..., 8.6746e-03, + 4.1490e-07, 9.2850e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0126, -0.0092, -0.0264, -0.0255, -0.0119, 0.0051, 0.0074, -0.0256, + -0.0137, 0.0099], device='cuda:0'), grad: tensor([ 0.0132, 0.0012, 0.0076, -0.0029, -0.0009, 0.0133, -0.0075, 0.0056, + -0.0480, 0.0183], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 222.34, cls_loss 0.0913 cls_loss_mapping 0.1452 cls_loss_causal 1.1394 re_mapping 0.0562 re_causal 0.1284 /// teacc 97.83 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0102, -0.0393, 0.0454, ..., 0.0038, 0.0197, 0.0022], + [-0.0288, -0.0001, 0.0007, ..., -0.0098, -0.0513, -0.0270], + [ 0.0232, -0.0011, 0.0141, ..., 0.0022, -0.0221, -0.0638], + ..., + [ 0.0221, 0.0100, -0.0016, ..., 0.0340, -0.0119, -0.0300], + [ 0.0014, -0.0152, 0.0391, ..., -0.0262, -0.0419, 0.0188], + [-0.0098, -0.0246, 0.0267, ..., 0.0341, -0.0353, -0.0173]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8358e-04, -7.5722e-03, ..., 1.1629e-04, + 6.6519e-05, -1.9970e-03], + [ 0.0000e+00, 4.9263e-05, 6.8521e-04, ..., 8.8751e-05, + 1.5154e-05, 3.6860e-04], + [ 0.0000e+00, -1.7989e-04, 2.6560e-04, ..., 6.4969e-05, + 3.0234e-05, 1.6956e-03], + ..., + [ 0.0000e+00, -6.5374e-04, -3.3045e-04, ..., -1.7166e-03, + 3.1859e-05, 3.7551e-04], + [ 0.0000e+00, 1.2016e-04, -5.6419e-03, ..., -9.7215e-05, + 8.0729e-04, -4.3335e-03], + [ 0.0000e+00, 5.5838e-04, 3.0613e-03, ..., 2.4567e-03, + 1.7536e-04, 2.4929e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0129, -0.0092, -0.0264, -0.0250, -0.0117, 0.0045, 0.0074, -0.0252, + -0.0135, 0.0094], device='cuda:0'), grad: tensor([-0.0085, -0.0006, -0.0001, 0.0066, -0.0010, -0.0028, 0.0025, -0.0002, + -0.0042, 0.0083], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 222.29, cls_loss 0.0680 cls_loss_mapping 0.1142 cls_loss_causal 1.1269 re_mapping 0.0524 re_causal 0.1266 /// teacc 98.02 lr 0.00010000 +Epoch 10, weight, value: tensor([[-1.0243e-02, -4.0390e-02, 4.6042e-02, ..., 2.5107e-03, + 2.0465e-02, 3.1441e-03], + [-2.8781e-02, -3.7975e-04, -5.2456e-06, ..., -9.3388e-03, + -5.1943e-02, -2.9051e-02], + [ 2.3198e-02, -8.8327e-04, 1.3205e-02, ..., 1.3086e-03, + -2.2454e-02, -6.7880e-02], + ..., + [ 2.2054e-02, 1.0954e-02, -1.4651e-03, ..., 3.4792e-02, + -1.3332e-02, -3.1514e-02], + [ 1.4234e-03, -1.5573e-02, 4.0345e-02, ..., -2.6844e-02, + -4.4691e-02, 2.1960e-02], + [-9.8281e-03, -2.5898e-02, 2.7957e-02, ..., 3.4514e-02, + -3.8534e-02, -1.8399e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.5465e-05, -1.1673e-03, ..., 1.3840e-04, + -9.9480e-05, -8.4257e-04], + [ 0.0000e+00, -6.3610e-04, 7.6056e-05, ..., 1.6302e-05, + 6.7428e-07, 4.5300e-05], + [ 0.0000e+00, 1.3542e-04, 4.6039e-04, ..., 1.9436e-03, + 7.3351e-06, 2.2686e-04], + ..., + [ 0.0000e+00, -4.0233e-05, 1.3661e-04, ..., 1.3895e-03, + 1.9968e-06, 1.1212e-04], + [ 0.0000e+00, 7.7367e-05, 5.9962e-05, ..., 3.4189e-04, + 5.2825e-06, 4.2152e-04], + [ 0.0000e+00, 2.6083e-04, 2.4348e-05, ..., 2.0161e-03, + 1.7941e-05, 1.7071e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0131, -0.0091, -0.0261, -0.0250, -0.0117, 0.0044, 0.0071, -0.0253, + -0.0132, 0.0094], device='cuda:0'), grad: tensor([-0.0010, -0.0046, 0.0032, 0.0024, -0.0084, 0.0006, 0.0001, 0.0027, + 0.0013, 0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 221.74, cls_loss 0.0747 cls_loss_mapping 0.1213 cls_loss_causal 1.0785 re_mapping 0.0462 re_causal 0.1112 /// teacc 97.36 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0102, -0.0406, 0.0467, ..., 0.0018, 0.0202, 0.0046], + [-0.0288, -0.0011, -0.0010, ..., -0.0100, -0.0526, -0.0331], + [ 0.0232, -0.0017, 0.0123, ..., -0.0002, -0.0234, -0.0721], + ..., + [ 0.0221, 0.0119, -0.0016, ..., 0.0356, -0.0156, -0.0328], + [ 0.0014, -0.0156, 0.0417, ..., -0.0263, -0.0482, 0.0246], + [-0.0098, -0.0269, 0.0285, ..., 0.0349, -0.0431, -0.0213]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.4687e-04, 3.8457e-04, ..., 5.7316e-04, + -3.8981e-04, -2.1148e-04], + [ 0.0000e+00, 9.1267e-04, 3.9077e-04, ..., 3.1400e-04, + 2.1607e-05, 4.4227e-04], + [ 0.0000e+00, -1.0710e-03, 3.6383e-04, ..., 3.7408e-04, + 2.8744e-05, -3.4380e-04], + ..., + [ 0.0000e+00, -6.2895e-04, -2.0733e-03, ..., -2.6417e-03, + 1.9297e-05, -2.7657e-04], + [ 0.0000e+00, 4.4417e-04, 2.3518e-03, ..., 1.2255e-03, + 3.2067e-04, 1.2140e-03], + [ 0.0000e+00, 9.9754e-04, -1.2360e-03, ..., -5.3978e-04, + 9.8825e-05, 8.8155e-05]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0130, -0.0093, -0.0265, -0.0248, -0.0120, 0.0043, 0.0072, -0.0252, + -0.0127, 0.0092], device='cuda:0'), grad: tensor([ 0.0005, 0.0089, -0.0119, -0.0398, -0.0009, 0.0359, 0.0014, -0.0014, + 0.0058, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 221.48, cls_loss 0.0622 cls_loss_mapping 0.1063 cls_loss_causal 1.0394 re_mapping 0.0428 re_causal 0.1072 /// teacc 98.01 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0102, -0.0417, 0.0473, ..., 0.0008, 0.0199, 0.0058], + [-0.0288, -0.0018, -0.0016, ..., -0.0113, -0.0525, -0.0337], + [ 0.0232, -0.0021, 0.0113, ..., -0.0007, -0.0249, -0.0759], + ..., + [ 0.0221, 0.0130, -0.0017, ..., 0.0364, -0.0178, -0.0331], + [ 0.0014, -0.0160, 0.0422, ..., -0.0266, -0.0512, 0.0266], + [-0.0098, -0.0281, 0.0297, ..., 0.0354, -0.0478, -0.0228]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1873e-04, 2.4378e-04, ..., 8.0109e-04, + 1.6270e-03, 2.8782e-03], + [ 0.0000e+00, 6.4325e-04, 1.1129e-03, ..., 8.1253e-04, + 6.8963e-05, 4.5037e-04], + [ 0.0000e+00, 4.9782e-04, 1.1263e-03, ..., 6.9571e-04, + 2.4700e-04, 1.0624e-03], + ..., + [ 0.0000e+00, -1.7910e-03, -1.0061e-03, ..., -1.2445e-03, + 4.2677e-04, 1.3103e-03], + [ 0.0000e+00, 1.3471e-04, -1.2875e-03, ..., 4.4084e-04, + 5.3883e-04, -3.1395e-03], + [ 0.0000e+00, 3.3593e-04, 3.3140e-05, ..., -8.6832e-04, + 1.4770e-04, 1.7357e-03]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0129, -0.0096, -0.0263, -0.0248, -0.0120, 0.0045, 0.0070, -0.0253, + -0.0127, 0.0093], device='cuda:0'), grad: tensor([ 0.0106, -0.0064, 0.0107, -0.0079, -0.0087, -0.0131, 0.0101, 0.0008, + -0.0002, 0.0042], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 222.29, cls_loss 0.0581 cls_loss_mapping 0.0946 cls_loss_causal 1.0517 re_mapping 0.0386 re_causal 0.1013 /// teacc 98.12 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0102, -0.0425, 0.0478, ..., -0.0002, 0.0206, 0.0062], + [-0.0288, -0.0027, -0.0025, ..., -0.0111, -0.0523, -0.0354], + [ 0.0232, -0.0030, 0.0104, ..., -0.0019, -0.0256, -0.0799], + ..., + [ 0.0221, 0.0142, -0.0018, ..., 0.0373, -0.0201, -0.0333], + [ 0.0014, -0.0163, 0.0431, ..., -0.0273, -0.0562, 0.0287], + [-0.0098, -0.0293, 0.0304, ..., 0.0358, -0.0521, -0.0235]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.0327e-04, -2.8992e-04, ..., 2.6178e-04, + 4.5955e-05, -5.0008e-05], + [ 0.0000e+00, 2.4772e-04, 2.6894e-04, ..., 3.1757e-04, + 9.8068e-07, 5.8293e-05], + [ 0.0000e+00, -3.2091e-04, 6.8426e-04, ..., 4.2105e-04, + 8.8988e-07, 2.6679e-04], + ..., + [ 0.0000e+00, -2.7485e-03, -1.0834e-03, ..., -3.9902e-03, + 2.5332e-06, 1.6785e-04], + [ 0.0000e+00, 2.2590e-04, -3.7861e-03, ..., -6.0081e-04, + 4.6194e-06, -2.9564e-03], + [ 0.0000e+00, 8.8596e-04, 2.4223e-03, ..., 1.1959e-03, + 1.3411e-05, 2.3785e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0133, -0.0098, -0.0264, -0.0244, -0.0123, 0.0046, 0.0073, -0.0250, + -0.0126, 0.0090], device='cuda:0'), grad: tensor([ 0.0007, 0.0013, -0.0017, -0.0024, -0.0036, 0.0010, 0.0030, -0.0036, + -0.0003, 0.0055], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 222.24, cls_loss 0.0468 cls_loss_mapping 0.0808 cls_loss_causal 0.9970 re_mapping 0.0358 re_causal 0.0954 /// teacc 98.38 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0102, -0.0435, 0.0486, ..., -0.0006, 0.0206, 0.0072], + [-0.0288, -0.0035, -0.0030, ..., -0.0124, -0.0525, -0.0352], + [ 0.0232, -0.0044, 0.0095, ..., -0.0031, -0.0261, -0.0835], + ..., + [ 0.0221, 0.0157, -0.0019, ..., 0.0379, -0.0215, -0.0345], + [ 0.0014, -0.0169, 0.0440, ..., -0.0278, -0.0580, 0.0307], + [-0.0098, -0.0304, 0.0311, ..., 0.0363, -0.0545, -0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3723e-05, -2.0707e-04, ..., 9.2506e-05, + 1.0991e-04, 1.0672e-03], + [ 0.0000e+00, 6.0320e-05, 3.0851e-04, ..., 3.9291e-04, + 6.9924e-06, 6.1274e-05], + [ 0.0000e+00, 7.6711e-05, 2.6870e-04, ..., 1.1873e-04, + 1.7613e-05, 2.1064e-04], + ..., + [ 0.0000e+00, -1.1902e-03, 7.3147e-04, ..., 1.2236e-03, + 2.6867e-05, 2.1708e-04], + [ 0.0000e+00, 8.5354e-05, 1.2941e-03, ..., 2.3389e-04, + 4.4250e-04, 4.0321e-03], + [ 0.0000e+00, 2.5368e-04, -4.0627e-04, ..., 1.2279e-04, + 6.9320e-05, 8.9693e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0131, -0.0099, -0.0267, -0.0243, -0.0121, 0.0046, 0.0069, -0.0249, + -0.0123, 0.0088], device='cuda:0'), grad: tensor([ 0.0024, 0.0007, -0.0007, 0.0017, -0.0054, -0.0066, -0.0025, 0.0021, + 0.0051, 0.0031], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 222.28, cls_loss 0.0430 cls_loss_mapping 0.0745 cls_loss_causal 0.9732 re_mapping 0.0335 re_causal 0.0909 /// teacc 98.40 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0102, -0.0444, 0.0491, ..., -0.0013, 0.0218, 0.0075], + [-0.0288, -0.0042, -0.0037, ..., -0.0124, -0.0522, -0.0362], + [ 0.0232, -0.0046, 0.0087, ..., -0.0032, -0.0263, -0.0867], + ..., + [ 0.0221, 0.0172, -0.0019, ..., 0.0384, -0.0242, -0.0355], + [ 0.0014, -0.0177, 0.0452, ..., -0.0284, -0.0609, 0.0333], + [-0.0098, -0.0316, 0.0318, ..., 0.0367, -0.0590, -0.0250]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.8052e-04, -8.1863e-03, ..., 2.5582e-04, + -5.2631e-05, -7.6141e-03], + [ 0.0000e+00, 8.3542e-04, 8.0299e-04, ..., 6.4850e-05, + 9.5926e-07, 3.5310e-04], + [ 0.0000e+00, 1.4954e-03, 1.4963e-03, ..., 2.3413e-04, + 1.2465e-05, 2.9230e-04], + ..., + [ 0.0000e+00, 4.3945e-03, 1.9350e-03, ..., -5.7125e-04, + 2.4829e-06, 1.5748e-04], + [ 0.0000e+00, 1.2960e-03, 1.8108e-04, ..., -2.8163e-05, + 1.7472e-06, -3.0541e-04], + [ 0.0000e+00, 2.2564e-03, 1.3342e-03, ..., -2.6393e-04, + 4.8243e-06, 7.5626e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0132, -0.0101, -0.0264, -0.0247, -0.0122, 0.0043, 0.0070, -0.0249, + -0.0117, 0.0089], device='cuda:0'), grad: tensor([-0.0105, 0.0027, 0.0030, -0.0222, 0.0007, -0.0007, 0.0106, 0.0087, + 0.0029, 0.0048], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 221.47, cls_loss 0.0342 cls_loss_mapping 0.0605 cls_loss_causal 0.9607 re_mapping 0.0322 re_causal 0.0886 /// teacc 98.11 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0102, -0.0449, 0.0498, ..., -0.0018, 0.0219, 0.0077], + [-0.0288, -0.0049, -0.0041, ..., -0.0133, -0.0520, -0.0359], + [ 0.0232, -0.0049, 0.0076, ..., -0.0037, -0.0258, -0.0894], + ..., + [ 0.0221, 0.0181, -0.0017, ..., 0.0391, -0.0293, -0.0375], + [ 0.0014, -0.0184, 0.0457, ..., -0.0286, -0.0664, 0.0346], + [-0.0098, -0.0327, 0.0324, ..., 0.0369, -0.0652, -0.0253]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.3745e-05, 2.6202e-04, ..., 3.7241e-04, + 1.4625e-05, 7.4387e-05], + [ 0.0000e+00, 5.4687e-05, 1.7703e-04, ..., 2.3663e-04, + 2.6412e-06, 2.8238e-06], + [ 0.0000e+00, 9.6381e-05, 3.3587e-05, ..., 5.7077e-04, + -7.0989e-05, 1.8865e-05], + ..., + [ 0.0000e+00, 3.5477e-04, 1.6785e-03, ..., 1.3418e-03, + 9.2462e-06, 1.6689e-04], + [ 0.0000e+00, 1.4234e-04, 1.0500e-03, ..., 7.0190e-04, + 4.6223e-05, 2.5535e-04], + [ 0.0000e+00, 1.8969e-03, 3.7651e-03, ..., 3.4885e-03, + 8.2850e-06, -1.5825e-05]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0131, -0.0101, -0.0264, -0.0244, -0.0120, 0.0041, 0.0070, -0.0251, + -0.0120, 0.0089], device='cuda:0'), grad: tensor([ 0.0012, 0.0006, 0.0004, -0.0017, -0.0162, -0.0011, 0.0003, 0.0035, + 0.0018, 0.0113], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 222.11, cls_loss 0.0443 cls_loss_mapping 0.0782 cls_loss_causal 0.9163 re_mapping 0.0300 re_causal 0.0841 /// teacc 98.44 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0102, -0.0455, 0.0501, ..., -0.0025, 0.0221, 0.0078], + [-0.0288, -0.0056, -0.0051, ..., -0.0138, -0.0538, -0.0371], + [ 0.0232, -0.0057, 0.0074, ..., -0.0044, -0.0257, -0.0922], + ..., + [ 0.0221, 0.0192, -0.0020, ..., 0.0397, -0.0285, -0.0393], + [ 0.0014, -0.0189, 0.0464, ..., -0.0289, -0.0722, 0.0369], + [-0.0098, -0.0340, 0.0333, ..., 0.0372, -0.0695, -0.0258]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.4150e-04, -3.8409e-04, ..., 1.5032e-04, + 9.3281e-06, -3.8886e-04], + [ 0.0000e+00, 1.0180e-04, 1.0270e-04, ..., 1.2255e-04, + 1.0476e-05, 2.9042e-05], + [ 0.0000e+00, 5.6458e-04, 5.9462e-04, ..., 5.7316e-04, + -8.3208e-05, 3.2616e-04], + ..., + [ 0.0000e+00, -1.4238e-03, -9.2649e-04, ..., -1.4534e-03, + 1.3828e-05, 1.0973e-04], + [ 0.0000e+00, 9.9719e-05, -1.1759e-03, ..., 1.1027e-04, + 4.4070e-06, -2.2488e-03], + [ 0.0000e+00, 2.6369e-04, 2.3103e-04, ..., 2.5463e-04, + 1.7453e-06, 1.2076e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0136, -0.0107, -0.0264, -0.0245, -0.0119, 0.0044, 0.0071, -0.0249, + -0.0117, 0.0090], device='cuda:0'), grad: tensor([ 6.9253e-06, 2.5582e-04, 3.7146e-04, 4.9257e-04, 1.6689e-04, + 1.1673e-03, 1.0824e-03, -2.4261e-03, -1.8377e-03, 7.2098e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 221.91, cls_loss 0.0319 cls_loss_mapping 0.0588 cls_loss_causal 0.8806 re_mapping 0.0283 re_causal 0.0804 /// teacc 98.46 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0102, -0.0465, 0.0506, ..., -0.0028, 0.0226, 0.0079], + [-0.0288, -0.0058, -0.0056, ..., -0.0141, -0.0553, -0.0380], + [ 0.0232, -0.0061, 0.0066, ..., -0.0050, -0.0243, -0.0947], + ..., + [ 0.0221, 0.0203, -0.0017, ..., 0.0405, -0.0273, -0.0393], + [ 0.0014, -0.0192, 0.0472, ..., -0.0290, -0.0772, 0.0380], + [-0.0098, -0.0353, 0.0337, ..., 0.0375, -0.0725, -0.0267]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.0206e-05, -5.5504e-04, ..., 6.8724e-05, + -1.5691e-05, -1.1492e-04], + [ 0.0000e+00, 4.6015e-05, 9.7096e-05, ..., 3.9756e-05, + 7.0892e-06, 6.7616e-04], + [ 0.0000e+00, 8.9288e-05, 1.1814e-04, ..., -2.0134e-04, + -1.7738e-04, 3.2926e-04], + ..., + [ 0.0000e+00, -5.9652e-04, -1.0383e-04, ..., -3.8409e-04, + 4.5002e-05, 1.9348e-04], + [ 0.0000e+00, 3.7491e-05, 6.1989e-06, ..., 7.7426e-05, + 1.3299e-05, 5.4884e-04], + [ 0.0000e+00, 1.5652e-04, -1.1599e-04, ..., -3.4198e-06, + 2.8864e-05, 6.5041e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0134, -0.0107, -0.0265, -0.0244, -0.0119, 0.0043, 0.0071, -0.0249, + -0.0116, 0.0088], device='cuda:0'), grad: tensor([-9.1255e-05, 1.1120e-03, -7.4911e-04, 4.8599e-03, 2.0962e-03, + -1.9485e-02, 9.4376e-03, 9.1076e-05, 1.5030e-03, 1.2341e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 222.01, cls_loss 0.0265 cls_loss_mapping 0.0527 cls_loss_causal 0.8679 re_mapping 0.0277 re_causal 0.0802 /// teacc 98.50 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0102, -0.0473, 0.0511, ..., -0.0034, 0.0230, 0.0084], + [-0.0288, -0.0067, -0.0060, ..., -0.0148, -0.0554, -0.0373], + [ 0.0232, -0.0068, 0.0060, ..., -0.0057, -0.0252, -0.0978], + ..., + [ 0.0221, 0.0218, -0.0015, ..., 0.0411, -0.0286, -0.0400], + [ 0.0014, -0.0197, 0.0477, ..., -0.0296, -0.0798, 0.0395], + [-0.0098, -0.0363, 0.0342, ..., 0.0377, -0.0779, -0.0278]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.1723e-06, 3.1024e-05, ..., 1.6451e-05, + 5.0128e-05, 3.0565e-04], + [ 0.0000e+00, 2.3067e-05, -5.0105e-07, ..., 3.1412e-05, + 1.4268e-05, -8.7976e-05], + [ 0.0000e+00, 1.1903e-04, 7.6115e-05, ..., 3.2097e-05, + -8.6904e-05, 9.2924e-05], + ..., + [ 0.0000e+00, -2.3037e-05, 1.9670e-04, ..., 1.9526e-04, + 3.1829e-05, 1.2803e-04], + [ 0.0000e+00, 8.6248e-05, 1.4961e-04, ..., 4.9496e-04, + 5.8591e-05, 1.2565e-04], + [ 0.0000e+00, 3.8385e-05, -7.0238e-04, ..., -3.6860e-04, + 2.2307e-05, -2.0492e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0133, -0.0110, -0.0264, -0.0244, -0.0117, 0.0040, 0.0073, -0.0247, + -0.0115, 0.0084], device='cuda:0'), grad: tensor([ 5.4359e-04, 6.8069e-05, -2.5749e-04, -2.0370e-03, -9.8324e-04, + 6.1464e-04, -3.4690e-04, 5.2881e-04, 1.2436e-03, 6.2704e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 222.27, cls_loss 0.0275 cls_loss_mapping 0.0500 cls_loss_causal 0.8878 re_mapping 0.0265 re_causal 0.0817 /// teacc 98.56 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0102, -0.0482, 0.0515, ..., -0.0039, 0.0231, 0.0086], + [-0.0288, -0.0073, -0.0069, ..., -0.0155, -0.0566, -0.0388], + [ 0.0232, -0.0073, 0.0057, ..., -0.0063, -0.0241, -0.1003], + ..., + [ 0.0221, 0.0227, -0.0018, ..., 0.0415, -0.0269, -0.0404], + [ 0.0014, -0.0200, 0.0482, ..., -0.0298, -0.0831, 0.0404], + [-0.0098, -0.0370, 0.0350, ..., 0.0381, -0.0803, -0.0277]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.8965e-05, -8.6784e-04, ..., 7.6413e-05, + -4.4167e-05, -7.4863e-04], + [ 0.0000e+00, 3.3170e-05, 1.5628e-04, ..., 8.8513e-05, + 3.1516e-06, 1.4079e-04], + [ 0.0000e+00, 1.0902e-04, 5.0926e-04, ..., 1.2493e-04, + -8.2552e-06, 5.9223e-04], + ..., + [ 0.0000e+00, -9.5797e-04, -5.8222e-04, ..., -8.4066e-04, + 4.0308e-06, 2.9349e-04], + [ 0.0000e+00, 4.3184e-05, 2.7218e-03, ..., 1.4601e-03, + 5.0254e-06, 1.4830e-03], + [ 0.0000e+00, 1.4985e-04, -4.3488e-03, ..., -1.7738e-03, + 3.0752e-06, -2.4090e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0136, -0.0113, -0.0261, -0.0245, -0.0117, 0.0042, 0.0075, -0.0250, + -0.0116, 0.0088], device='cuda:0'), grad: tensor([-2.1019e-03, -1.4037e-05, 1.1406e-03, 2.3365e-03, -2.6321e-04, + 5.6219e-04, 1.2922e-04, -6.2561e-04, 2.7504e-03, -3.9139e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 221.37, cls_loss 0.0294 cls_loss_mapping 0.0511 cls_loss_causal 0.8758 re_mapping 0.0254 re_causal 0.0775 /// teacc 98.38 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0102, -0.0490, 0.0513, ..., -0.0043, 0.0226, 0.0078], + [-0.0288, -0.0076, -0.0074, ..., -0.0160, -0.0544, -0.0402], + [ 0.0232, -0.0075, 0.0051, ..., -0.0069, -0.0235, -0.1032], + ..., + [ 0.0221, 0.0236, -0.0018, ..., 0.0423, -0.0275, -0.0407], + [ 0.0014, -0.0207, 0.0489, ..., -0.0300, -0.0842, 0.0419], + [-0.0098, -0.0383, 0.0359, ..., 0.0382, -0.0802, -0.0268]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3575e-05, -8.9931e-04, ..., 1.6764e-05, + 2.3270e-04, -1.3380e-03], + [ 0.0000e+00, 4.0650e-05, 1.3466e-03, ..., 3.7938e-05, + 1.2722e-06, 1.7948e-03], + [ 0.0000e+00, 1.1492e-04, 2.7037e-04, ..., 8.3089e-05, + -8.6021e-04, 3.0541e-04], + ..., + [ 0.0000e+00, -3.6716e-04, 1.6463e-04, ..., -3.9577e-05, + 8.7261e-05, 1.5914e-04], + [ 0.0000e+00, -2.3134e-06, -2.0008e-03, ..., 8.0109e-05, + 5.5408e-04, -2.4719e-03], + [ 0.0000e+00, 1.4555e-04, -4.9210e-04, ..., 1.8239e-04, + 1.2815e-04, 6.6090e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0140, -0.0114, -0.0263, -0.0247, -0.0120, 0.0045, 0.0074, -0.0246, + -0.0112, 0.0089], device='cuda:0'), grad: tensor([-0.0014, 0.0018, -0.0033, 0.0011, 0.0004, -0.0018, 0.0030, 0.0004, + -0.0015, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 221.26, cls_loss 0.0270 cls_loss_mapping 0.0525 cls_loss_causal 0.8307 re_mapping 0.0250 re_causal 0.0737 /// teacc 98.31 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0102, -0.0498, 0.0518, ..., -0.0049, 0.0233, 0.0084], + [-0.0288, -0.0079, -0.0076, ..., -0.0164, -0.0557, -0.0399], + [ 0.0232, -0.0085, 0.0044, ..., -0.0078, -0.0234, -0.1054], + ..., + [ 0.0221, 0.0244, -0.0019, ..., 0.0431, -0.0264, -0.0416], + [ 0.0014, -0.0205, 0.0493, ..., -0.0300, -0.0867, 0.0425], + [-0.0098, -0.0404, 0.0363, ..., 0.0382, -0.0836, -0.0279]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.1837e-05, -4.0829e-05, ..., 9.5129e-05, + -3.0577e-05, 9.2447e-05], + [ 0.0000e+00, 6.0022e-05, 3.1173e-05, ..., 8.6486e-05, + 5.2527e-06, -6.1691e-05], + [ 0.0000e+00, 1.8489e-04, -1.3180e-05, ..., 1.5008e-04, + -1.6296e-04, 4.0650e-05], + ..., + [ 0.0000e+00, -5.5695e-04, -2.2018e-04, ..., -4.5252e-04, + 7.9155e-05, 2.0489e-06], + [ 0.0000e+00, 4.2409e-05, 3.7217e-04, ..., 2.9445e-04, + 1.9014e-05, -4.4405e-05], + [ 0.0000e+00, 1.0085e-04, -6.4898e-04, ..., -1.7512e-04, + 1.1630e-05, 4.3750e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0138, -0.0113, -0.0263, -0.0243, -0.0119, 0.0046, 0.0073, -0.0245, + -0.0115, 0.0085], device='cuda:0'), grad: tensor([ 2.8992e-04, -4.5031e-05, -8.3733e-04, -2.7802e-02, -5.4502e-04, + 2.8412e-02, -3.0071e-05, -6.5863e-05, 4.6039e-04, 1.6022e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 221.62, cls_loss 0.0261 cls_loss_mapping 0.0515 cls_loss_causal 0.8687 re_mapping 0.0234 re_causal 0.0715 /// teacc 98.51 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0102, -0.0506, 0.0523, ..., -0.0053, 0.0250, 0.0095], + [-0.0288, -0.0085, -0.0084, ..., -0.0166, -0.0559, -0.0399], + [ 0.0232, -0.0094, 0.0036, ..., -0.0082, -0.0229, -0.1096], + ..., + [ 0.0221, 0.0255, -0.0018, ..., 0.0435, -0.0270, -0.0424], + [ 0.0014, -0.0209, 0.0500, ..., -0.0303, -0.0877, 0.0441], + [-0.0098, -0.0414, 0.0370, ..., 0.0384, -0.0868, -0.0282]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.4938e-05, -2.0289e-04, ..., 1.0766e-05, + 1.3031e-05, -1.7858e-04], + [ 0.0000e+00, 5.3942e-05, 8.0824e-05, ..., 5.1022e-05, + 3.4384e-06, -1.8632e-04], + [ 0.0000e+00, -3.3826e-06, 1.1587e-04, ..., 1.2445e-04, + -1.2016e-04, 9.3937e-05], + ..., + [ 0.0000e+00, -1.8501e-04, -9.9063e-05, ..., -1.2553e-04, + 5.1558e-05, 1.3196e-04], + [ 0.0000e+00, 1.0586e-04, 1.0729e-05, ..., 4.8280e-05, + 4.2707e-05, 1.0508e-04], + [ 0.0000e+00, 5.3085e-08, -3.1495e-04, ..., -2.6751e-04, + 2.4717e-06, 1.0109e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0137, -0.0111, -0.0266, -0.0245, -0.0119, 0.0046, 0.0070, -0.0244, + -0.0112, 0.0085], device='cuda:0'), grad: tensor([-1.9997e-05, -3.1829e-04, -5.3644e-04, 1.9932e-04, 2.1565e-04, + 3.0422e-04, -3.6550e-04, 4.3797e-04, 4.6587e-04, -3.8385e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 221.51, cls_loss 0.0279 cls_loss_mapping 0.0520 cls_loss_causal 0.8301 re_mapping 0.0219 re_causal 0.0648 /// teacc 98.57 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0102, -0.0518, 0.0532, ..., -0.0057, 0.0253, 0.0104], + [-0.0288, -0.0090, -0.0089, ..., -0.0166, -0.0560, -0.0405], + [ 0.0232, -0.0099, 0.0031, ..., -0.0088, -0.0231, -0.1120], + ..., + [ 0.0221, 0.0258, -0.0017, ..., 0.0438, -0.0273, -0.0424], + [ 0.0014, -0.0205, 0.0506, ..., -0.0303, -0.0886, 0.0453], + [-0.0098, -0.0423, 0.0371, ..., 0.0386, -0.0881, -0.0294]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1705e-05, -1.3173e-04, ..., 7.6056e-05, + -1.0140e-05, -8.1122e-05], + [ 0.0000e+00, 1.8048e-04, 3.1447e-04, ..., 3.6120e-04, + 5.3411e-07, -1.6272e-04], + [ 0.0000e+00, 7.5579e-05, 6.5684e-05, ..., 2.2805e-04, + 4.2394e-06, 8.4758e-05], + ..., + [ 0.0000e+00, 2.4853e-03, 6.3171e-03, ..., 7.3166e-03, + 1.2470e-06, 3.6359e-05], + [ 0.0000e+00, 2.5940e-04, 3.8362e-04, ..., 3.0899e-04, + 1.2089e-06, 3.7217e-04], + [ 0.0000e+00, -3.5915e-03, -8.5449e-03, ..., -9.8877e-03, + 1.5832e-06, -8.9705e-06]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0133, -0.0114, -0.0265, -0.0245, -0.0113, 0.0046, 0.0069, -0.0247, + -0.0110, 0.0079], device='cuda:0'), grad: tensor([-7.3314e-05, 3.2139e-04, -4.4799e-04, 7.0095e-04, 1.7262e-03, + 2.1040e-04, -2.3746e-04, 1.2276e-02, 1.2541e-03, -1.5732e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 222.21, cls_loss 0.0224 cls_loss_mapping 0.0467 cls_loss_causal 0.8327 re_mapping 0.0222 re_causal 0.0680 /// teacc 98.68 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0102, -0.0529, 0.0533, ..., -0.0063, 0.0254, 0.0104], + [-0.0288, -0.0090, -0.0095, ..., -0.0177, -0.0568, -0.0403], + [ 0.0232, -0.0107, 0.0025, ..., -0.0095, -0.0227, -0.1139], + ..., + [ 0.0221, 0.0266, -0.0019, ..., 0.0443, -0.0268, -0.0427], + [ 0.0014, -0.0206, 0.0513, ..., -0.0304, -0.0897, 0.0461], + [-0.0098, -0.0429, 0.0380, ..., 0.0390, -0.0892, -0.0295]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3007e-05, -2.3861e-03, ..., 1.4015e-05, + -8.4639e-06, -1.1292e-03], + [ 0.0000e+00, 2.9892e-05, 5.7310e-05, ..., 2.1964e-05, + 3.5688e-06, -1.2865e-03], + [ 0.0000e+00, 3.1352e-05, 3.3283e-03, ..., 2.8238e-05, + 8.7023e-04, 2.1267e-03], + ..., + [ 0.0000e+00, -6.5804e-05, -1.8626e-05, ..., -1.5366e-04, + 3.3796e-05, 1.3363e-04], + [ 0.0000e+00, 5.9843e-04, -2.0199e-03, ..., 3.1829e-05, + -9.1505e-04, -1.3609e-03], + [ 0.0000e+00, 1.5478e-03, 3.5038e-03, ..., 1.3399e-03, + 1.7047e-05, 1.0071e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0139, -0.0118, -0.0264, -0.0243, -0.0112, 0.0045, 0.0070, -0.0245, + -0.0109, 0.0081], device='cuda:0'), grad: tensor([-0.0035, -0.0016, 0.0030, -0.0037, -0.0021, -0.0012, 0.0013, 0.0002, + -0.0005, 0.0080], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 221.01, cls_loss 0.0255 cls_loss_mapping 0.0496 cls_loss_causal 0.8450 re_mapping 0.0209 re_causal 0.0660 /// teacc 98.47 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0102, -0.0539, 0.0537, ..., -0.0067, 0.0249, 0.0106], + [-0.0288, -0.0092, -0.0100, ..., -0.0185, -0.0586, -0.0400], + [ 0.0232, -0.0121, 0.0020, ..., -0.0100, -0.0213, -0.1154], + ..., + [ 0.0221, 0.0273, -0.0019, ..., 0.0450, -0.0254, -0.0427], + [ 0.0014, -0.0216, 0.0516, ..., -0.0308, -0.0923, 0.0471], + [-0.0098, -0.0443, 0.0384, ..., 0.0392, -0.0927, -0.0295]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6819e-05, 3.4332e-04, ..., 6.5506e-05, + 1.9884e-04, 6.8045e-04], + [ 0.0000e+00, 2.9549e-05, 1.7715e-04, ..., 8.5950e-05, + 3.0100e-05, 2.0909e-04], + [ 0.0000e+00, 1.4603e-05, 1.0929e-03, ..., 1.8132e-04, + 3.8266e-04, 1.8244e-03], + ..., + [ 0.0000e+00, -2.6560e-04, 7.7534e-04, ..., 4.4227e-04, + 8.2135e-05, 2.4557e-04], + [ 0.0000e+00, -9.1456e-07, -2.6932e-03, ..., -2.5129e-04, + -1.1272e-03, -5.0163e-03], + [ 0.0000e+00, -1.8370e-04, -1.5631e-03, ..., -1.4019e-03, + 7.7367e-05, 3.6216e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0141, -0.0120, -0.0265, -0.0236, -0.0116, 0.0041, 0.0066, -0.0243, + -0.0109, 0.0084], device='cuda:0'), grad: tensor([ 0.0012, 0.0004, 0.0029, 0.0019, 0.0005, -0.0004, 0.0018, 0.0011, + -0.0079, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 221.25, cls_loss 0.0216 cls_loss_mapping 0.0436 cls_loss_causal 0.8224 re_mapping 0.0205 re_causal 0.0617 /// teacc 98.66 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0102, -0.0549, 0.0541, ..., -0.0074, 0.0250, 0.0110], + [-0.0288, -0.0085, -0.0102, ..., -0.0186, -0.0591, -0.0400], + [ 0.0232, -0.0131, 0.0015, ..., -0.0109, -0.0205, -0.1181], + ..., + [ 0.0221, 0.0281, -0.0018, ..., 0.0462, -0.0253, -0.0435], + [ 0.0014, -0.0223, 0.0521, ..., -0.0311, -0.0936, 0.0477], + [-0.0098, -0.0450, 0.0393, ..., 0.0395, -0.0941, -0.0288]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.0115e-05, -3.4261e-04, ..., -4.3988e-05, + -9.0718e-05, 5.3883e-05], + [ 0.0000e+00, 8.3685e-05, 1.1092e-04, ..., 7.2718e-05, + 1.6056e-06, 3.2008e-05], + [ 0.0000e+00, 1.7560e-04, 2.5487e-04, ..., 1.7416e-04, + 2.3004e-06, 6.8247e-05], + ..., + [ 0.0000e+00, 2.6779e-03, 1.9569e-03, ..., -6.3562e-04, + 3.0659e-06, 3.5048e-05], + [ 0.0000e+00, 1.1474e-04, -1.1069e-04, ..., 2.5436e-05, + 8.8215e-06, -9.9778e-05], + [ 0.0000e+00, 1.5569e-04, -1.4976e-05, ..., 7.6473e-05, + 5.0254e-06, 5.3793e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0143, -0.0117, -0.0266, -0.0240, -0.0115, 0.0045, 0.0063, -0.0243, + -0.0110, 0.0087], device='cuda:0'), grad: tensor([-0.0002, 0.0002, 0.0002, -0.0031, 0.0002, 0.0004, -0.0005, 0.0025, + 0.0002, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 221.30, cls_loss 0.0206 cls_loss_mapping 0.0410 cls_loss_causal 0.7796 re_mapping 0.0205 re_causal 0.0605 /// teacc 98.52 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0102, -0.0561, 0.0545, ..., -0.0078, 0.0252, 0.0111], + [-0.0288, -0.0082, -0.0103, ..., -0.0184, -0.0592, -0.0402], + [ 0.0232, -0.0138, 0.0012, ..., -0.0113, -0.0202, -0.1188], + ..., + [ 0.0221, 0.0291, -0.0018, ..., 0.0465, -0.0255, -0.0441], + [ 0.0014, -0.0231, 0.0521, ..., -0.0313, -0.0942, 0.0485], + [-0.0098, -0.0468, 0.0397, ..., 0.0396, -0.0956, -0.0293]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.9174e-06, -3.0255e-04, ..., 3.3528e-07, + -2.2009e-05, -1.6284e-04], + [ 0.0000e+00, 6.2771e-06, 2.7522e-05, ..., 2.8729e-05, + 5.6531e-07, -9.9850e-04], + [ 0.0000e+00, 2.4378e-05, 1.0425e-04, ..., 4.8250e-05, + -4.5672e-06, 1.1075e-04], + ..., + [ 0.0000e+00, 3.0823e-03, 4.6730e-03, ..., 5.3825e-03, + 4.8839e-06, 1.1694e-04], + [ 0.0000e+00, 6.6519e-05, -2.9016e-04, ..., 9.7230e-06, + 5.6513e-06, -1.3411e-05], + [ 0.0000e+00, -3.2902e-03, -5.0278e-03, ..., -6.0005e-03, + 4.2506e-06, 4.1246e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0144, -0.0121, -0.0260, -0.0239, -0.0113, 0.0045, 0.0065, -0.0246, + -0.0111, 0.0086], device='cuda:0'), grad: tensor([-0.0003, -0.0039, 0.0002, 0.0009, 0.0009, 0.0006, -0.0001, 0.0070, + 0.0013, -0.0066], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 221.17, cls_loss 0.0174 cls_loss_mapping 0.0388 cls_loss_causal 0.7790 re_mapping 0.0193 re_causal 0.0599 /// teacc 98.65 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0194, -0.0569, 0.0551, ..., -0.0084, 0.0259, 0.0119], + [-0.0262, -0.0091, -0.0108, ..., -0.0192, -0.0599, -0.0397], + [ 0.0264, -0.0149, 0.0007, ..., -0.0119, -0.0198, -0.1209], + ..., + [ 0.0160, 0.0306, -0.0018, ..., 0.0473, -0.0254, -0.0444], + [-0.0017, -0.0234, 0.0524, ..., -0.0315, -0.0951, 0.0492], + [-0.0144, -0.0475, 0.0401, ..., 0.0398, -0.0979, -0.0293]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3210e-05, -1.9586e-04, ..., 1.0684e-05, + -3.8385e-05, -2.1803e-04], + [ 0.0000e+00, 2.9996e-05, 2.1845e-05, ..., 9.4324e-06, + 3.4384e-06, -1.0200e-05], + [ 0.0000e+00, 3.3230e-05, 4.0293e-05, ..., 9.6858e-06, + -2.2352e-05, 7.6473e-05], + ..., + [ 0.0000e+00, -8.5533e-05, 1.1072e-05, ..., -5.1409e-05, + 1.1429e-05, 4.3631e-05], + [ 0.0000e+00, 2.4557e-04, 3.1680e-05, ..., 1.1630e-05, + 1.4156e-05, 1.6510e-04], + [ 0.0000e+00, 1.0693e-04, -9.4235e-05, ..., -7.0512e-05, + 9.6932e-06, 1.0651e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0141, -0.0119, -0.0263, -0.0243, -0.0113, 0.0044, 0.0066, -0.0241, + -0.0111, 0.0085], device='cuda:0'), grad: tensor([-2.4962e-04, -4.2289e-05, 8.6010e-05, -4.1313e-03, 1.5509e-04, + 3.3150e-03, -1.5783e-04, 9.6083e-05, 7.7343e-04, 1.5187e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 221.43, cls_loss 0.0179 cls_loss_mapping 0.0371 cls_loss_causal 0.7640 re_mapping 0.0196 re_causal 0.0604 /// teacc 98.64 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0195, -0.0580, 0.0552, ..., -0.0094, 0.0258, 0.0122], + [-0.0264, -0.0085, -0.0109, ..., -0.0195, -0.0601, -0.0395], + [ 0.0266, -0.0157, 0.0001, ..., -0.0118, -0.0169, -0.1236], + ..., + [ 0.0160, 0.0312, -0.0019, ..., 0.0477, -0.0267, -0.0453], + [-0.0017, -0.0238, 0.0532, ..., -0.0314, -0.0963, 0.0502], + [-0.0144, -0.0483, 0.0406, ..., 0.0399, -0.1018, -0.0300]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.5423e-05, -9.0957e-05, ..., 2.9832e-05, + -7.8306e-06, 2.0489e-05], + [ 0.0000e+00, 3.0017e-04, 1.1319e-04, ..., 2.1625e-04, + 9.5427e-05, -1.2469e-04], + [ 0.0000e+00, 9.4354e-05, 3.4404e-04, ..., 3.4785e-04, + -1.7583e-04, 3.4142e-04], + ..., + [ 0.0000e+00, -4.3845e-04, 1.3075e-03, ..., 3.6573e-04, + 9.4622e-06, 8.6129e-05], + [ 0.0000e+00, 6.3241e-05, -8.0681e-04, ..., -3.8981e-04, + 2.3216e-05, -2.5773e-04], + [ 0.0000e+00, -6.6805e-04, -1.7977e-03, ..., -1.4868e-03, + 6.3553e-06, 1.1146e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0143, -0.0117, -0.0256, -0.0243, -0.0114, 0.0044, 0.0064, -0.0245, + -0.0108, 0.0081], device='cuda:0'), grad: tensor([ 8.3029e-05, 6.9523e-04, -1.8454e-04, 1.5488e-03, 7.8773e-04, + 1.6317e-03, -2.0504e-03, 8.2731e-04, 1.0705e-04, -3.4447e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 221.19, cls_loss 0.0218 cls_loss_mapping 0.0440 cls_loss_causal 0.7997 re_mapping 0.0187 re_causal 0.0597 /// teacc 98.61 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0197, -0.0586, 0.0550, ..., -0.0101, 0.0277, 0.0113], + [-0.0286, -0.0090, -0.0115, ..., -0.0206, -0.0612, -0.0401], + [ 0.0293, -0.0160, -0.0002, ..., -0.0122, -0.0164, -0.1254], + ..., + [ 0.0157, 0.0321, -0.0015, ..., 0.0486, -0.0265, -0.0452], + [-0.0015, -0.0248, 0.0535, ..., -0.0316, -0.0991, 0.0514], + [-0.0144, -0.0494, 0.0415, ..., 0.0401, -0.1082, -0.0300]], + device='cuda:0'), grad: tensor([[ 2.5309e-07, 1.9300e-04, 3.3557e-05, ..., 2.2495e-04, + -7.7784e-06, -3.0786e-05], + [ 1.0850e-07, 1.4603e-04, 5.0813e-05, ..., 1.2219e-04, + 8.9128e-07, -1.1069e-04], + [ 1.0547e-07, -2.9774e-03, -1.0437e-04, ..., -2.0905e-03, + -1.8045e-05, 1.3649e-04], + ..., + [ 2.8010e-07, 3.1815e-03, 1.0748e-03, ..., 1.6909e-03, + 6.7838e-06, 9.0981e-04], + [ 1.0831e-06, -1.7805e-03, -1.5593e-03, ..., -3.0828e-04, + 3.1888e-06, -1.4696e-03], + [ 2.4959e-06, 2.4939e-04, -4.9543e-04, ..., -3.3641e-04, + 2.2165e-06, 1.6916e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0147, -0.0120, -0.0258, -0.0243, -0.0113, 0.0042, 0.0065, -0.0244, + -0.0105, 0.0083], device='cuda:0'), grad: tensor([ 0.0005, -0.0006, -0.0063, 0.0019, 0.0007, 0.0001, -0.0003, 0.0067, + -0.0025, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 221.83, cls_loss 0.0148 cls_loss_mapping 0.0338 cls_loss_causal 0.7667 re_mapping 0.0195 re_causal 0.0595 /// teacc 98.69 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0205, -0.0593, 0.0555, ..., -0.0108, 0.0282, 0.0117], + [-0.0316, -0.0090, -0.0120, ..., -0.0208, -0.0615, -0.0402], + [ 0.0339, -0.0166, -0.0006, ..., -0.0129, -0.0152, -0.1265], + ..., + [ 0.0146, 0.0330, -0.0016, ..., 0.0492, -0.0267, -0.0460], + [-0.0008, -0.0250, 0.0543, ..., -0.0310, -0.0998, 0.0525], + [-0.0145, -0.0512, 0.0416, ..., 0.0399, -0.1103, -0.0310]], + device='cuda:0'), grad: tensor([[ 1.4901e-07, 1.1846e-05, 1.8626e-05, ..., 7.5065e-06, + 5.4054e-06, -6.5327e-05], + [ 8.1956e-08, 3.2157e-05, 8.8513e-05, ..., 1.0088e-05, + 3.2037e-06, -5.9038e-05], + [-2.1998e-06, 5.4985e-05, -3.3855e-04, ..., 8.8215e-06, + -8.2612e-05, 6.9737e-05], + ..., + [ 1.6950e-06, -1.8597e-04, 2.1291e-04, ..., -1.9014e-04, + 1.7837e-05, 3.3069e-04], + [-6.5425e-07, -6.8426e-04, -4.3631e-04, ..., 1.7673e-05, + 5.4687e-05, -5.0497e-04], + [ 1.9302e-07, 3.4511e-05, -2.6369e-04, ..., -1.0091e-04, + 6.5863e-06, -7.7784e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0147, -0.0121, -0.0256, -0.0242, -0.0112, 0.0039, 0.0064, -0.0245, + -0.0099, 0.0081], device='cuda:0'), grad: tensor([ 2.8992e-04, 7.9572e-05, -1.1978e-03, 9.0361e-04, 1.4985e-04, + 3.5048e-05, 4.2766e-06, 3.1257e-04, -4.1127e-04, -1.6403e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 221.61, cls_loss 0.0135 cls_loss_mapping 0.0301 cls_loss_causal 0.7681 re_mapping 0.0186 re_causal 0.0595 /// teacc 98.60 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0221, -0.0600, 0.0557, ..., -0.0113, 0.0289, 0.0119], + [-0.0324, -0.0081, -0.0120, ..., -0.0212, -0.0619, -0.0406], + [ 0.0352, -0.0180, -0.0012, ..., -0.0140, -0.0147, -0.1286], + ..., + [ 0.0129, 0.0341, -0.0014, ..., 0.0498, -0.0270, -0.0466], + [ 0.0021, -0.0252, 0.0547, ..., -0.0312, -0.1003, 0.0530], + [-0.0148, -0.0524, 0.0422, ..., 0.0401, -0.1115, -0.0306]], + device='cuda:0'), grad: tensor([[ 1.0012e-07, 1.0692e-05, -4.0078e-04, ..., -1.3016e-05, + -5.0813e-05, -5.8651e-04], + [ 3.5437e-07, 1.3433e-05, 3.0577e-05, ..., -3.1441e-05, + 8.7097e-06, 6.0806e-03], + [-3.2820e-06, -1.8883e-04, 3.9607e-05, ..., 2.4691e-05, + -4.6879e-05, 1.2016e-04], + ..., + [ 9.9000e-07, -2.8515e-04, -1.3471e-04, ..., -1.7047e-04, + 1.5453e-05, 1.3578e-04], + [ 1.4286e-06, 4.4674e-05, -1.3602e-04, ..., -4.8429e-05, + 1.5095e-05, 2.6488e-04], + [ 2.3050e-08, 1.4186e-04, 1.3971e-04, ..., 9.7275e-05, + 7.1861e-06, 3.9601e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0148, -0.0120, -0.0259, -0.0245, -0.0114, 0.0041, 0.0066, -0.0243, + -0.0100, 0.0083], device='cuda:0'), grad: tensor([-0.0011, 0.0077, -0.0008, 0.0033, 0.0002, -0.0115, 0.0005, 0.0001, + 0.0007, 0.0008], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 221.79, cls_loss 0.0131 cls_loss_mapping 0.0304 cls_loss_causal 0.7708 re_mapping 0.0173 re_causal 0.0567 /// teacc 98.74 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0246, -0.0612, 0.0562, ..., -0.0124, 0.0300, 0.0126], + [-0.0339, -0.0087, -0.0124, ..., -0.0215, -0.0620, -0.0413], + [ 0.0370, -0.0185, -0.0018, ..., -0.0144, -0.0147, -0.1306], + ..., + [ 0.0116, 0.0345, -0.0017, ..., 0.0502, -0.0274, -0.0473], + [ 0.0024, -0.0258, 0.0551, ..., -0.0313, -0.1008, 0.0538], + [-0.0150, -0.0533, 0.0425, ..., 0.0403, -0.1137, -0.0309]], + device='cuda:0'), grad: tensor([[ 1.2051e-06, 6.5118e-06, -1.9896e-04, ..., 4.5076e-06, + -3.6359e-05, -1.5211e-04], + [ 3.1829e-04, 2.6420e-05, -1.7136e-05, ..., 8.0094e-06, + 4.3511e-06, -1.4496e-04], + [-3.3617e-04, 1.5640e-04, 2.0623e-05, ..., -2.1920e-05, + -9.6500e-05, 3.1352e-05], + ..., + [ 7.5288e-06, 1.3721e-04, -6.4850e-05, ..., -2.4843e-04, + 9.7573e-05, 5.3227e-05], + [ 3.5986e-06, 1.0744e-05, -2.4632e-05, ..., 1.2591e-05, + 3.4422e-05, 1.2122e-05], + [ 4.4331e-07, 1.2201e-04, 1.3971e-04, ..., 2.5153e-04, + 5.5321e-06, 4.3541e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0147, -0.0123, -0.0259, -0.0242, -0.0114, 0.0040, 0.0066, -0.0245, + -0.0098, 0.0082], device='cuda:0'), grad: tensor([-1.8144e-04, 7.2384e-04, -1.2417e-03, -3.3712e-04, -4.0650e-05, + -7.5400e-06, 1.5974e-04, 3.8600e-04, 1.5426e-04, 3.8528e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 221.33, cls_loss 0.0133 cls_loss_mapping 0.0301 cls_loss_causal 0.7677 re_mapping 0.0168 re_causal 0.0551 /// teacc 98.72 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0267, -0.0620, 0.0567, ..., -0.0130, 0.0312, 0.0131], + [-0.0323, -0.0095, -0.0131, ..., -0.0220, -0.0620, -0.0417], + [ 0.0375, -0.0190, -0.0023, ..., -0.0151, -0.0143, -0.1320], + ..., + [ 0.0099, 0.0358, -0.0015, ..., 0.0512, -0.0279, -0.0480], + [ 0.0020, -0.0264, 0.0555, ..., -0.0315, -0.1017, 0.0549], + [-0.0152, -0.0546, 0.0429, ..., 0.0403, -0.1121, -0.0308]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 6.5982e-05, -5.9223e-04, ..., 1.4104e-05, + 0.0000e+00, -9.1600e-04], + [ 4.4238e-09, 9.1076e-05, 1.2800e-05, ..., 1.1414e-04, + 0.0000e+00, -1.9157e-04], + [ 1.4435e-08, 4.8599e-03, 1.7393e-04, ..., 5.9128e-03, + 0.0000e+00, 1.9312e-04], + ..., + [ 5.8208e-09, -4.2458e-03, -9.8825e-05, ..., -5.8784e-03, + 0.0000e+00, 1.6546e-04], + [-6.5193e-08, 1.0853e-03, 2.3115e-04, ..., 1.7846e-04, + 0.0000e+00, 5.3930e-04], + [ 1.1874e-08, 4.1270e-04, 2.0742e-04, ..., 1.4269e-04, + 0.0000e+00, 2.5964e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0146, -0.0124, -0.0256, -0.0247, -0.0110, 0.0040, 0.0067, -0.0242, + -0.0097, 0.0079], device='cuda:0'), grad: tensor([-1.1101e-03, -2.8968e-04, 1.0834e-02, -4.5624e-03, -9.6321e-04, + 8.5533e-05, 1.5430e-03, -9.3384e-03, 2.6493e-03, 1.1597e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 221.31, cls_loss 0.0132 cls_loss_mapping 0.0335 cls_loss_causal 0.7453 re_mapping 0.0172 re_causal 0.0540 /// teacc 98.62 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0301, -0.0628, 0.0568, ..., -0.0136, 0.0319, 0.0132], + [-0.0326, -0.0096, -0.0135, ..., -0.0226, -0.0622, -0.0414], + [ 0.0382, -0.0195, -0.0027, ..., -0.0158, -0.0139, -0.1334], + ..., + [ 0.0091, 0.0362, -0.0013, ..., 0.0521, -0.0283, -0.0488], + [ 0.0017, -0.0278, 0.0558, ..., -0.0316, -0.1026, 0.0558], + [-0.0154, -0.0559, 0.0431, ..., 0.0403, -0.1131, -0.0317]], + device='cuda:0'), grad: tensor([[ 6.9151e-08, 1.2517e-05, -1.9759e-05, ..., 1.5184e-05, + -1.0252e-05, -4.7654e-05], + [ 1.1176e-08, 6.8605e-05, 5.4777e-05, ..., 4.6581e-05, + 4.0070e-07, -3.3408e-05], + [-6.5146e-07, 5.4121e-05, 6.7234e-05, ..., 3.4422e-05, + 1.4203e-07, 2.2784e-05], + ..., + [ 1.2410e-07, -2.3246e-04, -8.3089e-05, ..., -1.7607e-04, + 6.2399e-07, 3.4243e-05], + [ 3.7043e-07, 2.7940e-05, 5.3406e-05, ..., 3.7342e-05, + 1.0617e-06, 1.2986e-05], + [ 1.6298e-09, 8.9586e-05, -1.4734e-04, ..., -8.4817e-05, + 6.7148e-07, -4.2111e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0148, -0.0123, -0.0257, -0.0245, -0.0111, 0.0042, 0.0067, -0.0242, + -0.0096, 0.0075], device='cuda:0'), grad: tensor([-1.8492e-05, 1.1659e-04, 1.3888e-04, -3.0732e-04, 7.0989e-05, + 1.7130e-04, -3.7760e-05, -1.8656e-04, 1.2314e-04, -7.1228e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 222.16, cls_loss 0.0109 cls_loss_mapping 0.0317 cls_loss_causal 0.7748 re_mapping 0.0169 re_causal 0.0552 /// teacc 98.78 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0307, -0.0636, 0.0572, ..., -0.0140, 0.0322, 0.0139], + [-0.0327, -0.0098, -0.0138, ..., -0.0234, -0.0622, -0.0410], + [ 0.0384, -0.0204, -0.0033, ..., -0.0163, -0.0137, -0.1348], + ..., + [ 0.0087, 0.0373, -0.0012, ..., 0.0525, -0.0284, -0.0497], + [ 0.0013, -0.0273, 0.0563, ..., -0.0317, -0.1028, 0.0562], + [-0.0156, -0.0574, 0.0441, ..., 0.0410, -0.1135, -0.0314]], + device='cuda:0'), grad: tensor([[ 1.2562e-05, 2.2352e-05, -9.6083e-05, ..., 2.3186e-05, + -4.6287e-07, 1.7989e-04], + [ 3.6269e-05, -8.1778e-04, -3.7742e-04, ..., 3.6210e-05, + -7.4208e-05, 4.9770e-05], + [ 9.0599e-05, -1.4639e-03, -7.6532e-05, ..., -1.0023e-03, + 1.9986e-06, 1.3771e-03], + ..., + [ 1.5691e-05, 1.8358e-03, 4.3106e-04, ..., 7.2050e-04, + 3.6247e-06, 5.5742e-04], + [-4.2367e-04, 5.1379e-05, 1.8871e-04, ..., 5.0187e-05, + 8.7358e-07, -5.3825e-03], + [ 5.9493e-06, 2.2531e-04, -6.4659e-04, ..., -6.4611e-04, + 6.0834e-06, 2.2399e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0146, -0.0123, -0.0259, -0.0249, -0.0113, 0.0042, 0.0069, -0.0240, + -0.0096, 0.0078], device='cuda:0'), grad: tensor([ 0.0003, -0.0019, -0.0008, 0.0008, 0.0011, 0.0051, -0.0004, 0.0050, + -0.0089, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 221.58, cls_loss 0.0102 cls_loss_mapping 0.0251 cls_loss_causal 0.7310 re_mapping 0.0160 re_causal 0.0516 /// teacc 98.78 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0343, -0.0640, 0.0577, ..., -0.0146, 0.0328, 0.0146], + [-0.0330, -0.0106, -0.0142, ..., -0.0236, -0.0623, -0.0407], + [ 0.0390, -0.0212, -0.0037, ..., -0.0166, -0.0134, -0.1356], + ..., + [ 0.0043, 0.0385, -0.0016, ..., 0.0525, -0.0285, -0.0517], + [ 0.0042, -0.0271, 0.0569, ..., -0.0314, -0.1034, 0.0572], + [-0.0161, -0.0583, 0.0447, ..., 0.0413, -0.1137, -0.0320]], + device='cuda:0'), grad: tensor([[ 1.8477e-05, 1.2711e-05, -4.5896e-05, ..., 1.6704e-05, + 8.0466e-06, 2.1124e-04], + [ 1.2433e-07, 4.7356e-05, -3.8207e-05, ..., 8.4341e-05, + 3.5507e-07, -1.9467e-04], + [ 3.4668e-07, 4.9382e-05, 7.3075e-05, ..., 4.1306e-05, + 2.6952e-06, 8.8573e-05], + ..., + [ 1.1958e-06, -3.1400e-04, -1.0407e-04, ..., -3.1900e-04, + 1.9139e-07, 3.8773e-05], + [ 1.3765e-06, 2.3797e-05, 3.7462e-05, ..., 2.6777e-05, + 9.0618e-07, -3.6061e-06], + [ 4.3958e-06, -1.3304e-04, -1.4811e-03, ..., 5.5456e-04, + 2.7637e-07, -3.5095e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0143, -0.0126, -0.0257, -0.0252, -0.0107, 0.0039, 0.0067, -0.0241, + -0.0093, 0.0076], device='cuda:0'), grad: tensor([ 4.0197e-04, -5.9795e-04, 3.4046e-04, 2.3384e-03, -4.4775e-04, + -8.0943e-05, -1.0216e-04, -3.0184e-04, 2.4116e-04, -1.7891e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 221.47, cls_loss 0.0107 cls_loss_mapping 0.0243 cls_loss_causal 0.7298 re_mapping 0.0164 re_causal 0.0516 /// teacc 98.73 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0412, -0.0649, 0.0580, ..., -0.0154, 0.0333, 0.0150], + [-0.0333, -0.0113, -0.0148, ..., -0.0250, -0.0623, -0.0410], + [ 0.0397, -0.0221, -0.0041, ..., -0.0174, -0.0132, -0.1370], + ..., + [-0.0011, 0.0398, -0.0015, ..., 0.0536, -0.0287, -0.0524], + [ 0.0039, -0.0278, 0.0573, ..., -0.0312, -0.1036, 0.0572], + [-0.0191, -0.0596, 0.0450, ..., 0.0413, -0.1142, -0.0323]], + device='cuda:0'), grad: tensor([[ 1.0328e-06, 2.6560e-04, 9.8705e-04, ..., 6.0606e-04, + 6.2864e-09, 1.2302e-04], + [ 3.1129e-07, -4.2394e-06, 2.9039e-04, ..., 4.0054e-04, + 3.1432e-08, 1.3447e-04], + [ 2.5006e-07, 1.8883e-04, 3.2473e-04, ..., 2.0611e-04, + 5.3551e-08, 4.4793e-05], + ..., + [ 3.3621e-06, 3.4630e-05, 2.3804e-03, ..., 1.3609e-03, + 4.9826e-08, 4.5276e-04], + [ 6.0610e-06, 1.1253e-04, 2.6894e-04, ..., 2.9969e-04, + 8.7544e-08, 2.3246e-04], + [ 2.0862e-06, -8.6641e-04, -3.8643e-03, ..., -1.7633e-03, + 1.7695e-07, 6.5446e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0147, -0.0129, -0.0261, -0.0249, -0.0108, 0.0039, 0.0072, -0.0235, + -0.0095, 0.0074], device='cuda:0'), grad: tensor([ 2.0428e-03, 3.9744e-04, 7.0000e-04, 2.5463e-04, -1.6680e-03, + -3.3170e-05, 1.2058e-04, 5.0659e-03, 7.4530e-04, -7.6294e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 38---------------------------------------------------- +epoch 38, time 222.39, cls_loss 0.0096 cls_loss_mapping 0.0223 cls_loss_causal 0.7305 re_mapping 0.0154 re_causal 0.0492 /// teacc 98.80 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0438, -0.0652, 0.0586, ..., -0.0160, 0.0338, 0.0153], + [-0.0333, -0.0111, -0.0151, ..., -0.0257, -0.0626, -0.0414], + [ 0.0399, -0.0227, -0.0047, ..., -0.0179, -0.0131, -0.1387], + ..., + [-0.0018, 0.0404, -0.0015, ..., 0.0544, -0.0286, -0.0529], + [ 0.0036, -0.0284, 0.0580, ..., -0.0312, -0.1040, 0.0578], + [-0.0197, -0.0610, 0.0453, ..., 0.0414, -0.1147, -0.0331]], + device='cuda:0'), grad: tensor([[ 2.8163e-06, 7.1637e-06, -1.2100e-04, ..., 7.0035e-05, + -1.5339e-06, -5.6028e-05], + [ 4.4610e-07, 2.3186e-05, 1.5177e-05, ..., 9.8109e-05, + 2.8871e-08, -3.6168e-04], + [ 3.8883e-07, 5.2512e-05, 3.1054e-05, ..., 1.5700e-04, + 1.3318e-07, 5.9307e-05], + ..., + [ 3.6345e-07, -9.5308e-05, 1.7834e-04, ..., 4.9496e-04, + 1.0710e-07, 4.5329e-05], + [-7.3537e-06, 1.4693e-05, 5.0783e-05, ..., 1.3006e-04, + 1.9628e-07, 2.7478e-05], + [ 1.5311e-06, -3.4183e-05, -4.9210e-04, ..., -1.8740e-04, + 3.1199e-07, -1.9252e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0144, -0.0129, -0.0260, -0.0249, -0.0106, 0.0039, 0.0074, -0.0236, + -0.0094, 0.0070], device='cuda:0'), grad: tensor([ 0.0002, -0.0002, 0.0005, 0.0001, -0.0027, 0.0001, 0.0006, 0.0011, + 0.0003, -0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 221.18, cls_loss 0.0095 cls_loss_mapping 0.0264 cls_loss_causal 0.7367 re_mapping 0.0145 re_causal 0.0485 /// teacc 98.74 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0500, -0.0665, 0.0590, ..., -0.0164, 0.0340, 0.0156], + [-0.0335, -0.0115, -0.0160, ..., -0.0263, -0.0626, -0.0413], + [ 0.0404, -0.0230, -0.0048, ..., -0.0181, -0.0129, -0.1402], + ..., + [-0.0066, 0.0414, -0.0016, ..., 0.0547, -0.0286, -0.0538], + [ 0.0023, -0.0291, 0.0585, ..., -0.0314, -0.1043, 0.0589], + [-0.0222, -0.0621, 0.0461, ..., 0.0424, -0.1148, -0.0334]], + device='cuda:0'), grad: tensor([[ 5.7705e-06, 3.2037e-06, -4.7636e-04, ..., 1.1912e-06, + -1.4508e-04, -3.5238e-04], + [ 1.3988e-06, -4.5848e-04, -5.0068e-05, ..., 1.3717e-05, + 6.2399e-08, -2.5797e-04], + [ 2.9653e-06, 2.2817e-04, 2.4959e-05, ..., 5.0068e-06, + 4.4750e-07, 6.7472e-05], + ..., + [ 2.4643e-06, -2.5377e-05, -3.0100e-05, ..., -6.4313e-05, + 1.5576e-07, 9.4414e-05], + [ 7.4804e-06, 3.6925e-05, -1.6546e-04, ..., 9.8944e-06, + 8.1491e-07, 1.9535e-05], + [ 2.2456e-05, 1.9163e-05, 3.1412e-05, ..., -6.5342e-06, + 1.6391e-06, 1.4639e-04]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0144, -0.0131, -0.0257, -0.0252, -0.0110, 0.0040, 0.0072, -0.0236, + -0.0092, 0.0073], device='cuda:0'), grad: tensor([-0.0007, -0.0036, 0.0016, 0.0003, 0.0001, -0.0002, 0.0009, 0.0013, + 0.0002, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 221.97, cls_loss 0.0080 cls_loss_mapping 0.0236 cls_loss_causal 0.7013 re_mapping 0.0157 re_causal 0.0509 /// teacc 98.86 lr 0.00010000 +Epoch 42, weight, value: tensor([[-5.6239e-02, -6.7219e-02, 5.9017e-02, ..., -1.6961e-02, + 3.3974e-02, 1.5501e-02], + [-3.3698e-02, -1.2168e-02, -1.6550e-02, ..., -2.6520e-02, + -6.2712e-02, -4.0966e-02], + [ 3.9960e-02, -2.3497e-02, -5.2284e-03, ..., -1.8055e-02, + -1.2972e-02, -1.4215e-01], + ..., + [-8.7664e-03, 4.2172e-02, -1.7169e-03, ..., 5.4585e-02, + -2.8594e-02, -5.4580e-02], + [ 6.7347e-05, -2.9791e-02, 5.8638e-02, ..., -3.1699e-02, + -1.0404e-01, 5.9027e-02], + [-2.0229e-02, -6.2979e-02, 4.6852e-02, ..., 4.2775e-02, + -1.1516e-01, -3.3228e-02]], device='cuda:0'), grad: tensor([[ 9.9316e-06, 1.2390e-05, -8.3089e-05, ..., 2.5090e-06, + -1.2731e-06, -3.7253e-05], + [ 1.8235e-06, -1.6105e-04, 1.2957e-05, ..., 4.0419e-06, + 2.7008e-08, -1.0687e-04], + [-2.8275e-06, 7.4029e-05, 6.0827e-05, ..., 5.0738e-06, + 3.1409e-07, 2.4632e-05], + ..., + [ 6.2995e-06, 5.6553e-04, 9.5248e-05, ..., 5.8487e-07, + 1.7905e-07, 3.0145e-05], + [ 1.0751e-05, 3.1114e-05, 1.8984e-05, ..., 8.6799e-06, + 6.8918e-08, 3.1739e-05], + [ 4.0442e-05, 2.0057e-05, -1.1593e-04, ..., 1.4877e-04, + 4.0396e-07, 1.0341e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0147, -0.0128, -0.0259, -0.0248, -0.0108, 0.0039, 0.0073, -0.0238, + -0.0097, 0.0075], device='cuda:0'), grad: tensor([-8.8930e-05, -6.5756e-04, -6.7997e-04, -4.8542e-04, -1.6618e-04, + -3.7026e-04, 3.7861e-04, 1.4973e-03, 1.1820e-04, 4.5419e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 221.03, cls_loss 0.0079 cls_loss_mapping 0.0219 cls_loss_causal 0.6764 re_mapping 0.0151 re_causal 0.0472 /// teacc 98.72 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0547, -0.0680, 0.0595, ..., -0.0174, 0.0343, 0.0161], + [-0.0339, -0.0122, -0.0167, ..., -0.0272, -0.0631, -0.0407], + [ 0.0400, -0.0242, -0.0057, ..., -0.0184, -0.0128, -0.1433], + ..., + [-0.0137, 0.0430, -0.0013, ..., 0.0552, -0.0283, -0.0548], + [ 0.0032, -0.0302, 0.0591, ..., -0.0319, -0.1047, 0.0598], + [-0.0207, -0.0640, 0.0471, ..., 0.0428, -0.1158, -0.0337]], + device='cuda:0'), grad: tensor([[ 1.0361e-07, 4.7609e-06, 6.1244e-06, ..., 6.6720e-06, + -5.5246e-06, -8.9109e-06], + [ 1.3737e-08, 1.3578e-04, 1.1951e-04, ..., 1.9741e-04, + 2.7241e-07, -1.6205e-07], + [ 1.1642e-08, 1.5020e-05, 1.4198e-04, ..., 1.5482e-05, + 2.5555e-06, 7.3075e-05], + ..., + [ 4.6799e-08, -4.2653e-04, -2.1100e-04, ..., -6.1893e-04, + 4.0047e-07, 2.1055e-05], + [ 2.5658e-07, 6.4038e-06, -6.2101e-06, ..., 1.1571e-05, + 1.3039e-06, -3.5822e-05], + [ 4.1001e-07, 7.2122e-05, -9.2220e-04, ..., 9.1717e-06, + 4.8429e-07, -4.0770e-04]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0144, -0.0129, -0.0257, -0.0253, -0.0108, 0.0038, 0.0072, -0.0234, + -0.0094, 0.0071], device='cuda:0'), grad: tensor([ 0.0001, -0.0012, 0.0008, 0.0008, 0.0023, 0.0006, 0.0002, -0.0009, + 0.0001, -0.0026], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 221.86, cls_loss 0.0076 cls_loss_mapping 0.0227 cls_loss_causal 0.7138 re_mapping 0.0146 re_causal 0.0502 /// teacc 98.90 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0561, -0.0691, 0.0599, ..., -0.0177, 0.0343, 0.0166], + [-0.0341, -0.0121, -0.0170, ..., -0.0279, -0.0632, -0.0409], + [ 0.0400, -0.0245, -0.0062, ..., -0.0188, -0.0130, -0.1446], + ..., + [-0.0158, 0.0435, -0.0015, ..., 0.0556, -0.0284, -0.0553], + [ 0.0019, -0.0300, 0.0598, ..., -0.0322, -0.1049, 0.0601], + [-0.0203, -0.0646, 0.0473, ..., 0.0427, -0.1161, -0.0340]], + device='cuda:0'), grad: tensor([[ 8.7544e-06, 1.1437e-05, 1.5154e-05, ..., 8.8662e-06, + 7.4622e-08, 3.0458e-05], + [ 2.1197e-06, 4.9882e-06, 5.0850e-06, ..., 5.0366e-06, + 2.8056e-08, -4.8392e-06], + [ 1.4855e-06, 4.6268e-06, 1.5080e-05, ..., 3.2280e-06, + -1.1567e-06, 3.6091e-05], + ..., + [ 4.4592e-06, -5.9903e-05, -2.8312e-05, ..., -4.3750e-05, + 8.8941e-07, -2.4941e-06], + [ 3.4064e-05, 1.1800e-06, -8.2731e-05, ..., 5.6252e-06, + 7.1479e-08, -3.9846e-05], + [ 1.0304e-05, 3.6448e-05, -2.7612e-05, ..., -5.2154e-06, + 4.6566e-09, 3.6687e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0142, -0.0126, -0.0259, -0.0252, -0.0101, 0.0038, 0.0071, -0.0235, + -0.0093, 0.0065], device='cuda:0'), grad: tensor([ 9.9182e-05, -8.5384e-06, -2.4462e-04, 5.5599e-04, 1.9297e-05, + -6.1417e-04, 1.1289e-04, -8.4490e-06, 5.8383e-05, 2.9683e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 221.11, cls_loss 0.0085 cls_loss_mapping 0.0243 cls_loss_causal 0.7035 re_mapping 0.0144 re_causal 0.0463 /// teacc 98.69 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0580, -0.0699, 0.0598, ..., -0.0184, 0.0345, 0.0167], + [-0.0345, -0.0123, -0.0172, ..., -0.0285, -0.0632, -0.0405], + [ 0.0399, -0.0245, -0.0065, ..., -0.0192, -0.0127, -0.1460], + ..., + [-0.0206, 0.0436, -0.0020, ..., 0.0561, -0.0285, -0.0568], + [ 0.0002, -0.0292, 0.0604, ..., -0.0325, -0.1051, 0.0610], + [-0.0167, -0.0659, 0.0487, ..., 0.0435, -0.1163, -0.0332]], + device='cuda:0'), grad: tensor([[ 3.8370e-07, 1.9401e-05, -7.5638e-05, ..., 2.1800e-05, + -4.1467e-07, -5.0783e-05], + [ 5.7276e-07, 9.9540e-06, 2.1040e-05, ..., 1.2621e-05, + 2.2002e-07, 5.1372e-06], + [ 2.3609e-07, -1.1221e-05, 5.6267e-05, ..., -8.0466e-06, + -5.2154e-06, 4.3571e-05], + ..., + [ 9.0525e-07, -1.4317e-04, -7.0572e-05, ..., -1.2660e-04, + 3.8892e-06, 1.2442e-05], + [ 3.6322e-06, 2.8893e-05, -3.2276e-05, ..., 3.7760e-05, + 2.1956e-07, -3.0220e-05], + [-3.5726e-06, 6.0141e-05, -5.8532e-05, ..., -1.6108e-05, + 7.8231e-08, 3.0264e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0145, -0.0126, -0.0254, -0.0252, -0.0112, 0.0040, 0.0069, -0.0239, + -0.0091, 0.0073], device='cuda:0'), grad: tensor([-8.3864e-05, 4.0621e-05, -4.4322e-04, 1.7130e-04, 1.1468e-04, + 7.6368e-06, -9.5963e-05, -1.2362e-04, 4.1008e-04, 3.3677e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 221.22, cls_loss 0.0085 cls_loss_mapping 0.0249 cls_loss_causal 0.7140 re_mapping 0.0140 re_causal 0.0458 /// teacc 98.78 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0626, -0.0710, 0.0602, ..., -0.0188, 0.0345, 0.0169], + [-0.0350, -0.0125, -0.0175, ..., -0.0293, -0.0637, -0.0408], + [ 0.0395, -0.0257, -0.0069, ..., -0.0198, -0.0123, -0.1469], + ..., + [-0.0248, 0.0445, -0.0018, ..., 0.0570, -0.0287, -0.0571], + [ 0.0007, -0.0296, 0.0607, ..., -0.0328, -0.1053, 0.0614], + [-0.0198, -0.0673, 0.0491, ..., 0.0431, -0.1166, -0.0335]], + device='cuda:0'), grad: tensor([[ 7.2002e-05, 1.0945e-05, -1.5516e-03, ..., 3.4332e-05, + 6.8080e-07, -3.1948e-03], + [ 5.8115e-05, 1.9699e-05, 3.2872e-05, ..., 2.1517e-05, + 3.0012e-07, -1.4365e-04], + [ 1.6582e-04, 1.3006e-04, 2.7609e-04, ..., 9.2328e-05, + 1.2480e-05, 5.9128e-04], + ..., + [ 1.7837e-05, -4.5359e-05, 6.7241e-07, ..., -3.6806e-05, + 1.5218e-06, 4.0501e-05], + [ 2.2066e-04, 2.8205e-04, -1.7595e-04, ..., -4.7982e-05, + 7.7859e-07, -5.4741e-04], + [ 1.1355e-04, 1.5736e-04, 5.5361e-04, ..., 4.1537e-06, + 2.0210e-06, 7.2050e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0144, -0.0127, -0.0255, -0.0253, -0.0107, 0.0041, 0.0072, -0.0236, + -0.0094, 0.0067], device='cuda:0'), grad: tensor([-5.1613e-03, -3.4833e-04, 1.4801e-03, -1.9684e-03, 1.8816e-03, + 8.9455e-04, 2.0638e-03, 8.7678e-05, -2.2757e-04, 1.2980e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 45---------------------------------------------------- +epoch 45, time 221.83, cls_loss 0.0078 cls_loss_mapping 0.0193 cls_loss_causal 0.6629 re_mapping 0.0132 re_causal 0.0425 /// teacc 98.94 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0678, -0.0724, 0.0607, ..., -0.0192, 0.0343, 0.0166], + [-0.0351, -0.0126, -0.0178, ..., -0.0296, -0.0649, -0.0417], + [ 0.0378, -0.0266, -0.0071, ..., -0.0202, -0.0119, -0.1493], + ..., + [-0.0261, 0.0456, -0.0017, ..., 0.0574, -0.0285, -0.0575], + [-0.0024, -0.0301, 0.0609, ..., -0.0331, -0.1059, 0.0619], + [-0.0193, -0.0684, 0.0494, ..., 0.0433, -0.1168, -0.0336]], + device='cuda:0'), grad: tensor([[ 4.8101e-05, 6.6422e-06, 1.0334e-05, ..., 1.9580e-05, + 1.3737e-08, 1.6904e-04], + [ 1.4074e-05, 1.1846e-05, 1.8448e-05, ..., 1.5616e-05, + 9.3132e-10, 1.0677e-05], + [ 5.7429e-05, 3.5822e-05, 3.1859e-05, ..., 4.7803e-05, + 1.3970e-09, 7.0214e-05], + ..., + [ 5.9716e-06, -9.5665e-05, -1.7807e-05, ..., -6.8188e-05, + 0.0000e+00, 3.1441e-05], + [ 3.7163e-05, -4.2468e-06, 2.6798e-04, ..., 9.7826e-06, + 1.5367e-08, 1.0481e-03], + [ 2.0862e-05, 5.1498e-05, 2.7448e-05, ..., 2.4140e-06, + 1.1642e-09, 6.8009e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0146, -0.0124, -0.0260, -0.0253, -0.0105, 0.0043, 0.0070, -0.0235, + -0.0093, 0.0068], device='cuda:0'), grad: tensor([ 5.0163e-04, -1.2755e-04, 5.4836e-04, 5.7757e-05, 2.6817e-03, + 4.1237e-03, -9.1476e-03, 4.8250e-05, 1.0233e-03, 2.8729e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 220.87, cls_loss 0.0079 cls_loss_mapping 0.0207 cls_loss_causal 0.7199 re_mapping 0.0135 re_causal 0.0434 /// teacc 98.84 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0696, -0.0736, 0.0604, ..., -0.0205, 0.0344, 0.0164], + [-0.0355, -0.0130, -0.0184, ..., -0.0307, -0.0651, -0.0419], + [ 0.0380, -0.0273, -0.0077, ..., -0.0213, -0.0116, -0.1505], + ..., + [-0.0290, 0.0459, -0.0019, ..., 0.0580, -0.0285, -0.0583], + [-0.0040, -0.0303, 0.0614, ..., -0.0330, -0.1061, 0.0620], + [-0.0209, -0.0694, 0.0503, ..., 0.0439, -0.1170, -0.0340]], + device='cuda:0'), grad: tensor([[ 2.4289e-06, 3.0901e-06, -1.2469e-04, ..., 3.3975e-06, + -2.2456e-05, -5.7071e-06], + [ 1.6704e-05, 1.8567e-05, 1.5640e-04, ..., 1.4506e-05, + 2.6263e-07, -3.3545e-04], + [ 5.3167e-05, 6.2346e-05, 3.9011e-05, ..., 4.2319e-05, + 3.2932e-06, 8.2314e-05], + ..., + [-2.2054e-05, -2.6369e-04, -2.0742e-04, ..., -3.1447e-04, + 1.4724e-06, 2.9594e-05], + [-3.7819e-05, 1.2353e-05, -2.7227e-04, ..., 2.3901e-05, + 7.6089e-07, -4.1389e-04], + [ 2.6274e-04, 1.5259e-04, 1.2529e-04, ..., 2.6393e-04, + 9.2387e-06, 4.4227e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0154, -0.0126, -0.0259, -0.0249, -0.0107, 0.0043, 0.0078, -0.0238, + -0.0096, 0.0071], device='cuda:0'), grad: tensor([-0.0001, -0.0008, 0.0001, 0.0002, -0.0004, 0.0002, 0.0005, -0.0004, + -0.0002, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 221.62, cls_loss 0.0068 cls_loss_mapping 0.0191 cls_loss_causal 0.6630 re_mapping 0.0129 re_causal 0.0418 /// teacc 98.72 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0697, -0.0746, 0.0609, ..., -0.0207, 0.0346, 0.0165], + [-0.0357, -0.0139, -0.0192, ..., -0.0320, -0.0656, -0.0420], + [ 0.0381, -0.0281, -0.0081, ..., -0.0216, -0.0115, -0.1518], + ..., + [-0.0306, 0.0473, -0.0018, ..., 0.0587, -0.0282, -0.0589], + [-0.0043, -0.0305, 0.0620, ..., -0.0331, -0.1068, 0.0637], + [-0.0205, -0.0706, 0.0506, ..., 0.0439, -0.1174, -0.0345]], + device='cuda:0'), grad: tensor([[ 3.7136e-07, 3.8415e-05, 1.1377e-05, ..., 1.5646e-06, + 1.6913e-06, 2.9162e-05], + [ 5.2992e-07, -1.9705e-04, 2.1309e-05, ..., 3.1870e-06, + 1.6764e-07, -1.5318e-04], + [ 1.2293e-07, -6.8426e-04, -3.9554e-04, ..., -6.5416e-06, + -6.5327e-05, 2.4021e-05], + ..., + [ 4.5402e-07, 1.5783e-04, 7.7665e-05, ..., -5.3644e-06, + 7.1060e-07, 2.5883e-05], + [ 4.4629e-06, 2.9898e-04, 1.1462e-04, ..., 1.0826e-05, + 5.8174e-05, -2.1780e-04], + [ 2.9150e-07, 6.1914e-06, -8.9228e-05, ..., -7.3016e-05, + 1.5786e-07, 4.2021e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0153, -0.0131, -0.0263, -0.0253, -0.0104, 0.0042, 0.0074, -0.0233, + -0.0087, 0.0069], device='cuda:0'), grad: tensor([ 2.0373e-04, -1.4105e-03, -1.3838e-03, 1.2455e-03, 2.2590e-04, + -8.2374e-05, 1.2457e-04, 5.1308e-04, 5.7220e-04, -6.8992e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 221.23, cls_loss 0.0086 cls_loss_mapping 0.0210 cls_loss_causal 0.7020 re_mapping 0.0132 re_causal 0.0422 /// teacc 98.91 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0701, -0.0758, 0.0608, ..., -0.0218, 0.0349, 0.0169], + [-0.0358, -0.0138, -0.0204, ..., -0.0329, -0.0659, -0.0426], + [ 0.0380, -0.0284, -0.0087, ..., -0.0223, -0.0106, -0.1529], + ..., + [-0.0311, 0.0484, -0.0013, ..., 0.0595, -0.0283, -0.0594], + [-0.0048, -0.0307, 0.0626, ..., -0.0331, -0.1078, 0.0641], + [-0.0204, -0.0721, 0.0513, ..., 0.0441, -0.1177, -0.0345]], + device='cuda:0'), grad: tensor([[ 7.4692e-07, 2.1979e-05, 6.2287e-06, ..., 9.5293e-06, + 6.3479e-06, 4.3213e-05], + [ 1.8207e-07, 9.4399e-06, 9.7081e-06, ..., 1.0863e-05, + 1.3970e-05, 4.0233e-05], + [ 2.1770e-07, 6.8545e-05, 5.6505e-05, ..., 5.6803e-05, + 2.1622e-05, 1.4174e-04], + ..., + [ 1.2312e-06, -1.1814e-04, -6.1274e-05, ..., -1.1331e-04, + 8.9966e-07, 2.1562e-05], + [ 5.6215e-06, 1.1957e-04, 3.1382e-05, ..., -1.6540e-05, + 1.7554e-05, 2.4557e-04], + [ 9.3728e-06, 4.0740e-05, 5.7727e-05, ..., 5.6237e-05, + 5.6392e-07, 4.8637e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0151, -0.0135, -0.0256, -0.0256, -0.0105, 0.0038, 0.0072, -0.0231, + -0.0087, 0.0071], device='cuda:0'), grad: tensor([ 9.6619e-05, 8.7440e-05, 3.7432e-04, -8.1253e-04, -3.7819e-05, + 1.8501e-04, -5.6791e-04, -1.0949e-04, 6.1655e-04, 1.6749e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 220.86, cls_loss 0.0061 cls_loss_mapping 0.0193 cls_loss_causal 0.6482 re_mapping 0.0129 re_causal 0.0405 /// teacc 98.86 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0708, -0.0765, 0.0611, ..., -0.0224, 0.0350, 0.0173], + [-0.0358, -0.0142, -0.0208, ..., -0.0334, -0.0659, -0.0419], + [ 0.0381, -0.0291, -0.0092, ..., -0.0229, -0.0105, -0.1544], + ..., + [-0.0313, 0.0493, -0.0012, ..., 0.0599, -0.0284, -0.0600], + [-0.0056, -0.0314, 0.0628, ..., -0.0332, -0.1082, 0.0642], + [-0.0208, -0.0730, 0.0518, ..., 0.0443, -0.1179, -0.0351]], + device='cuda:0'), grad: tensor([[ 2.6114e-06, 2.5332e-05, 2.6852e-05, ..., 2.0102e-05, + 7.0259e-06, 7.3731e-05], + [ 4.4750e-07, 5.3614e-05, 1.2793e-05, ..., 1.9699e-05, + 2.4997e-06, -1.1808e-04], + [ 8.4843e-07, -1.6940e-04, 3.6836e-05, ..., -7.2718e-05, + 3.5763e-06, 1.1784e-04], + ..., + [ 4.3102e-06, 2.5058e-04, 7.5698e-05, ..., -8.0228e-05, + 3.3178e-07, 4.3273e-05], + [ 4.9397e-06, 1.1081e-04, 1.2290e-04, ..., 4.9502e-05, + 1.4722e-05, 1.0192e-04], + [-2.3823e-06, 7.5996e-05, -1.5700e-04, ..., 2.2721e-04, + 1.2117e-06, -1.1140e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0150, -0.0135, -0.0257, -0.0255, -0.0103, 0.0036, 0.0076, -0.0231, + -0.0090, 0.0070], device='cuda:0'), grad: tensor([ 0.0002, -0.0002, -0.0004, -0.0007, -0.0005, 0.0002, -0.0003, 0.0007, + 0.0004, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 221.53, cls_loss 0.0066 cls_loss_mapping 0.0181 cls_loss_causal 0.7141 re_mapping 0.0126 re_causal 0.0411 /// teacc 98.62 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0731, -0.0775, 0.0612, ..., -0.0229, 0.0350, 0.0169], + [-0.0359, -0.0136, -0.0207, ..., -0.0326, -0.0658, -0.0405], + [ 0.0382, -0.0291, -0.0096, ..., -0.0231, -0.0105, -0.1560], + ..., + [-0.0319, 0.0495, -0.0014, ..., 0.0597, -0.0284, -0.0614], + [-0.0062, -0.0321, 0.0636, ..., -0.0331, -0.1085, 0.0655], + [-0.0216, -0.0742, 0.0522, ..., 0.0447, -0.1183, -0.0363]], + device='cuda:0'), grad: tensor([[ 5.1036e-07, 2.3186e-05, 1.4916e-05, ..., 1.1228e-05, + 6.6757e-06, 4.2409e-05], + [ 5.3318e-08, 2.5749e-04, 2.0051e-04, ..., 1.0580e-04, + 1.6196e-06, 7.2002e-05], + [ 8.0327e-08, 4.4316e-05, 4.6521e-05, ..., 1.8224e-05, + -1.0453e-05, 5.2452e-05], + ..., + [ 3.2969e-07, -6.3562e-04, -4.3464e-04, ..., -2.1458e-04, + 4.2096e-06, -2.2173e-04], + [ 8.6334e-07, 6.1572e-05, 2.8722e-06, ..., 3.4034e-05, + 1.2200e-06, -7.4387e-05], + [ 1.0580e-06, 1.4544e-04, 2.2650e-05, ..., 2.2680e-05, + 1.4645e-07, 4.2558e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0156, -0.0128, -0.0252, -0.0252, -0.0106, 0.0033, 0.0074, -0.0238, + -0.0084, 0.0070], device='cuda:0'), grad: tensor([ 1.1647e-04, 1.9860e-04, 7.9274e-05, 1.6809e-04, 3.7909e-05, + 6.8367e-05, -8.9169e-05, -7.4673e-04, 1.1258e-05, 1.5557e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 221.36, cls_loss 0.0068 cls_loss_mapping 0.0206 cls_loss_causal 0.6648 re_mapping 0.0126 re_causal 0.0403 /// teacc 98.77 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0744, -0.0780, 0.0614, ..., -0.0235, 0.0350, 0.0165], + [-0.0360, -0.0141, -0.0212, ..., -0.0335, -0.0665, -0.0406], + [ 0.0381, -0.0298, -0.0103, ..., -0.0235, -0.0105, -0.1582], + ..., + [-0.0331, 0.0504, -0.0013, ..., 0.0605, -0.0278, -0.0617], + [-0.0073, -0.0324, 0.0645, ..., -0.0332, -0.1087, 0.0663], + [-0.0214, -0.0752, 0.0527, ..., 0.0448, -0.1185, -0.0368]], + device='cuda:0'), grad: tensor([[ 1.3024e-05, 4.5523e-06, 7.9349e-07, ..., 4.4852e-06, + 0.0000e+00, 2.9296e-05], + [ 1.7844e-06, 3.6471e-06, 3.2801e-06, ..., 4.4554e-06, + 0.0000e+00, -2.7686e-05], + [ 1.2189e-05, 1.4424e-05, 1.2815e-05, ..., 1.2532e-05, + 0.0000e+00, 2.2545e-05], + ..., + [ 1.1615e-05, -5.3287e-05, 4.0233e-06, ..., -1.5929e-05, + 0.0000e+00, 1.6630e-05], + [ 1.8984e-05, 5.2825e-06, -4.6194e-05, ..., -4.3362e-06, + 0.0000e+00, -2.4408e-05], + [ 1.4710e-04, 2.9117e-05, -6.2764e-05, ..., -6.5029e-05, + 0.0000e+00, 1.6117e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0161, -0.0131, -0.0257, -0.0253, -0.0105, 0.0043, 0.0065, -0.0235, + -0.0080, 0.0070], device='cuda:0'), grad: tensor([ 7.5221e-05, -7.5817e-05, 2.4855e-05, 9.7752e-04, 8.0466e-05, + -1.5793e-03, -3.8266e-05, 3.4332e-05, 7.9393e-05, 4.2200e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 221.39, cls_loss 0.0060 cls_loss_mapping 0.0156 cls_loss_causal 0.6599 re_mapping 0.0127 re_causal 0.0395 /// teacc 98.84 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0755, -0.0786, 0.0617, ..., -0.0239, 0.0350, 0.0169], + [-0.0362, -0.0144, -0.0216, ..., -0.0342, -0.0666, -0.0406], + [ 0.0384, -0.0305, -0.0108, ..., -0.0239, -0.0105, -0.1609], + ..., + [-0.0343, 0.0511, -0.0013, ..., 0.0605, -0.0279, -0.0622], + [-0.0082, -0.0329, 0.0651, ..., -0.0328, -0.1089, 0.0666], + [-0.0210, -0.0760, 0.0529, ..., 0.0449, -0.1185, -0.0377]], + device='cuda:0'), grad: tensor([[ 1.0327e-05, 1.8477e-05, 4.9407e-07, ..., 1.0826e-05, + 4.5411e-06, 3.1084e-05], + [ 4.5784e-06, 2.2531e-05, 1.4871e-05, ..., 1.0774e-05, + 2.0489e-06, 1.2107e-05], + [ 8.4750e-07, -4.7088e-05, 5.2273e-05, ..., 1.4409e-05, + -8.2612e-05, 3.3736e-05], + ..., + [ 8.9779e-06, 1.2732e-04, 1.1295e-04, ..., 5.0992e-05, + 2.6196e-05, 1.0198e-04], + [ 9.0450e-06, 8.4519e-05, -2.0540e-04, ..., -1.3030e-04, + 9.7379e-06, -1.7512e-04], + [ 1.8024e-04, 4.5121e-05, 7.5996e-05, ..., 1.1820e-04, + 9.6858e-07, 4.2057e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0156, -0.0134, -0.0262, -0.0251, -0.0101, 0.0049, 0.0066, -0.0237, + -0.0080, 0.0066], device='cuda:0'), grad: tensor([ 8.8871e-05, 4.5061e-05, -3.6025e-04, -1.7996e-03, 5.6028e-06, + 4.1699e-04, 4.1199e-04, 4.4823e-04, -1.3120e-05, 7.5626e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 221.18, cls_loss 0.0048 cls_loss_mapping 0.0144 cls_loss_causal 0.6426 re_mapping 0.0127 re_causal 0.0399 /// teacc 98.82 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0772, -0.0785, 0.0622, ..., -0.0243, 0.0358, 0.0177], + [-0.0364, -0.0147, -0.0218, ..., -0.0349, -0.0668, -0.0401], + [ 0.0381, -0.0309, -0.0113, ..., -0.0244, -0.0101, -0.1619], + ..., + [-0.0360, 0.0516, -0.0015, ..., 0.0606, -0.0279, -0.0626], + [-0.0095, -0.0333, 0.0650, ..., -0.0332, -0.1092, 0.0667], + [-0.0197, -0.0766, 0.0537, ..., 0.0454, -0.1189, -0.0377]], + device='cuda:0'), grad: tensor([[ 3.0478e-07, 3.0249e-06, -6.8367e-05, ..., 3.2801e-06, + 6.3842e-07, -2.8387e-05], + [ 1.0841e-06, 1.8343e-05, 8.5905e-06, ..., 1.2703e-05, + 5.3691e-07, -1.4491e-05], + [ 6.7614e-07, 6.9201e-05, 5.6088e-05, ..., 2.5347e-05, + 1.2971e-05, 8.8751e-05], + ..., + [ 3.5703e-05, -6.9618e-05, -2.5213e-05, ..., 1.2457e-04, + 1.1129e-07, 1.4164e-05], + [ 3.3248e-06, 1.0207e-05, -4.9472e-05, ..., 1.4957e-06, + 9.2667e-07, 2.1029e-06], + [ 1.0245e-05, 1.9163e-05, -3.7588e-06, ..., 3.6985e-05, + 1.3923e-07, 2.6509e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0153, -0.0132, -0.0263, -0.0251, -0.0102, 0.0045, 0.0071, -0.0240, + -0.0087, 0.0070], device='cuda:0'), grad: tensor([-8.7321e-05, 1.7032e-05, 6.2609e-04, -3.2616e-04, -2.7552e-05, + 3.3766e-05, -7.9727e-04, 3.1352e-04, 8.8632e-05, 1.5903e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 54---------------------------------------------------- +epoch 54, time 222.17, cls_loss 0.0061 cls_loss_mapping 0.0194 cls_loss_causal 0.6924 re_mapping 0.0119 re_causal 0.0388 /// teacc 99.00 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0803, -0.0795, 0.0623, ..., -0.0250, 0.0358, 0.0177], + [-0.0370, -0.0156, -0.0228, ..., -0.0361, -0.0668, -0.0400], + [ 0.0383, -0.0315, -0.0119, ..., -0.0247, -0.0101, -0.1634], + ..., + [-0.0371, 0.0523, -0.0009, ..., 0.0612, -0.0279, -0.0627], + [-0.0116, -0.0337, 0.0655, ..., -0.0334, -0.1094, 0.0670], + [-0.0191, -0.0782, 0.0541, ..., 0.0457, -0.1190, -0.0377]], + device='cuda:0'), grad: tensor([[ 8.8587e-06, 5.5581e-05, 1.6764e-05, ..., 3.4004e-05, + 0.0000e+00, 2.5421e-05], + [ 8.5682e-07, 4.1202e-06, 3.5577e-06, ..., 9.3162e-05, + 0.0000e+00, 9.2387e-06], + [ 6.0583e-07, -4.1723e-05, -7.5519e-05, ..., 2.2814e-05, + 0.0000e+00, 2.0444e-05], + ..., + [-1.9029e-05, -1.2058e-04, -3.8743e-05, ..., -5.4479e-05, + 0.0000e+00, -1.1343e-04], + [ 8.4117e-06, 2.3007e-05, 3.7670e-05, ..., 2.6405e-05, + 0.0000e+00, 3.3706e-05], + [ 1.2759e-06, 6.9812e-06, -1.2413e-05, ..., 3.6657e-05, + 0.0000e+00, 2.2486e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0154, -0.0137, -0.0265, -0.0245, -0.0101, 0.0042, 0.0076, -0.0235, + -0.0088, 0.0067], device='cuda:0'), grad: tensor([ 1.8227e-04, 6.0177e-04, -3.6645e-04, 1.3912e-04, -1.5125e-03, + 1.1253e-04, 1.8454e-04, -9.2983e-06, 4.1962e-04, 2.4819e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 220.79, cls_loss 0.0069 cls_loss_mapping 0.0173 cls_loss_causal 0.6373 re_mapping 0.0120 re_causal 0.0365 /// teacc 98.92 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0815, -0.0810, 0.0626, ..., -0.0255, 0.0357, 0.0180], + [-0.0374, -0.0161, -0.0234, ..., -0.0368, -0.0670, -0.0407], + [ 0.0393, -0.0317, -0.0126, ..., -0.0246, -0.0099, -0.1650], + ..., + [-0.0398, 0.0525, -0.0007, ..., 0.0619, -0.0278, -0.0635], + [-0.0128, -0.0341, 0.0660, ..., -0.0335, -0.1096, 0.0679], + [-0.0179, -0.0793, 0.0545, ..., 0.0460, -0.1191, -0.0382]], + device='cuda:0'), grad: tensor([[ 4.3074e-07, 8.4098e-07, -2.1473e-05, ..., 1.0896e-06, + 0.0000e+00, -3.6927e-07], + [ 7.9395e-08, -1.4696e-06, 7.8231e-06, ..., 2.0657e-06, + 0.0000e+00, 2.2247e-05], + [ 5.9139e-08, 4.9844e-06, 2.0877e-05, ..., 4.5076e-06, + 0.0000e+00, 6.8665e-05], + ..., + [ 7.2271e-07, -1.7881e-05, -4.5262e-06, ..., -1.2711e-05, + 0.0000e+00, 2.7612e-05], + [ 2.3656e-07, 5.2759e-07, -1.8525e-04, ..., -2.6152e-06, + 0.0000e+00, -7.7152e-04], + [-3.9153e-06, 1.8366e-06, 1.4573e-05, ..., -3.7942e-06, + 0.0000e+00, 3.3885e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0152, -0.0144, -0.0259, -0.0242, -0.0104, 0.0042, 0.0079, -0.0235, + -0.0086, 0.0065], device='cuda:0'), grad: tensor([ 6.7689e-06, 3.1501e-05, 1.0920e-04, 4.6444e-04, 5.6362e-04, + 1.0377e-04, -2.2873e-05, 6.6578e-05, -1.3742e-03, 5.2333e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 221.14, cls_loss 0.0050 cls_loss_mapping 0.0183 cls_loss_causal 0.6524 re_mapping 0.0121 re_causal 0.0392 /// teacc 98.84 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0823, -0.0824, 0.0629, ..., -0.0265, 0.0357, 0.0182], + [-0.0376, -0.0164, -0.0238, ..., -0.0372, -0.0671, -0.0407], + [ 0.0394, -0.0321, -0.0132, ..., -0.0251, -0.0097, -0.1664], + ..., + [-0.0412, 0.0533, -0.0006, ..., 0.0621, -0.0278, -0.0639], + [-0.0136, -0.0344, 0.0665, ..., -0.0336, -0.1100, 0.0689], + [-0.0170, -0.0802, 0.0553, ..., 0.0465, -0.1191, -0.0386]], + device='cuda:0'), grad: tensor([[ 1.2806e-06, 2.6263e-06, 4.1217e-05, ..., 4.7386e-06, + 1.6997e-08, 7.3195e-05], + [ 5.3085e-07, 5.6922e-06, 9.4473e-06, ..., 1.1161e-05, + 1.0245e-08, 7.4469e-06], + [-5.8450e-06, -2.5168e-05, 1.1511e-05, ..., -2.6956e-05, + -3.3760e-07, 3.8415e-05], + ..., + [ 5.3719e-06, -8.2478e-06, 2.8029e-05, ..., 1.5333e-05, + 1.9162e-07, 1.5140e-05], + [ 4.2886e-05, 4.9084e-05, 7.9041e-03, ..., 3.2902e-04, + 7.8697e-08, 1.2505e-02], + [-9.0957e-05, -1.6121e-06, -5.7077e-04, ..., -4.2486e-04, + 3.4925e-09, -2.3305e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0154, -0.0144, -0.0261, -0.0246, -0.0106, 0.0043, 0.0081, -0.0234, + -0.0083, 0.0066], device='cuda:0'), grad: tensor([ 6.4433e-05, 9.7930e-05, -2.0325e-04, 6.9514e-06, -1.7977e-04, + 1.7905e-04, -1.4740e-02, 1.0967e-04, 1.5236e-02, -5.7220e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 220.81, cls_loss 0.0049 cls_loss_mapping 0.0150 cls_loss_causal 0.6467 re_mapping 0.0116 re_causal 0.0371 /// teacc 98.95 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0838, -0.0838, 0.0632, ..., -0.0271, 0.0356, 0.0186], + [-0.0379, -0.0166, -0.0240, ..., -0.0377, -0.0672, -0.0405], + [ 0.0396, -0.0322, -0.0136, ..., -0.0251, -0.0095, -0.1681], + ..., + [-0.0442, 0.0538, -0.0006, ..., 0.0624, -0.0278, -0.0647], + [-0.0158, -0.0348, 0.0666, ..., -0.0338, -0.1103, 0.0686], + [-0.0178, -0.0811, 0.0555, ..., 0.0463, -0.1192, -0.0396]], + device='cuda:0'), grad: tensor([[-5.0012e-07, 6.9104e-07, -4.0047e-06, ..., 1.6000e-06, + 3.9185e-07, 4.9740e-05], + [-2.6021e-06, 6.6981e-06, 1.3918e-05, ..., 1.4246e-05, + 2.7940e-09, -1.4722e-04], + [ 1.1288e-06, 1.1377e-05, 1.2279e-05, ..., -1.8347e-07, + 4.1910e-09, 2.5630e-05], + ..., + [ 6.7707e-07, -1.0744e-05, -3.2075e-06, ..., -7.8678e-06, + 6.9849e-10, 1.2897e-05], + [-2.7061e-05, 7.0445e-06, -9.9599e-05, ..., -8.6203e-06, + 8.6147e-09, -2.0063e-04], + [ 1.0487e-06, 1.0669e-05, 2.6897e-05, ..., 2.0638e-05, + 9.3132e-10, 2.5347e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0155, -0.0147, -0.0256, -0.0248, -0.0101, 0.0049, 0.0085, -0.0238, + -0.0089, 0.0063], device='cuda:0'), grad: tensor([ 7.3791e-05, -1.1742e-04, 4.5806e-05, 3.7719e-06, -2.3782e-04, + 1.1724e-04, 5.0694e-05, 6.2764e-05, -1.0830e-04, 1.0967e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 58---------------------------------------------------- +epoch 58, time 221.73, cls_loss 0.0053 cls_loss_mapping 0.0166 cls_loss_causal 0.6714 re_mapping 0.0116 re_causal 0.0375 /// teacc 99.04 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0848, -0.0845, 0.0634, ..., -0.0282, 0.0356, 0.0191], + [-0.0379, -0.0171, -0.0245, ..., -0.0381, -0.0674, -0.0398], + [ 0.0395, -0.0342, -0.0141, ..., -0.0254, -0.0095, -0.1694], + ..., + [-0.0457, 0.0555, -0.0003, ..., 0.0629, -0.0275, -0.0654], + [-0.0163, -0.0351, 0.0671, ..., -0.0338, -0.1104, 0.0692], + [-0.0173, -0.0821, 0.0559, ..., 0.0464, -0.1192, -0.0399]], + device='cuda:0'), grad: tensor([[ 2.4810e-06, 1.0990e-06, -1.1295e-04, ..., 9.3412e-07, + 0.0000e+00, -1.7166e-05], + [ 6.0536e-06, 1.7732e-06, 7.7635e-06, ..., 2.5518e-06, + 0.0000e+00, 3.2485e-06], + [ 6.5472e-07, 2.5213e-05, 4.6551e-05, ..., 4.3400e-06, + 0.0000e+00, 1.4655e-05], + ..., + [ 6.0797e-06, -9.8795e-06, 9.4026e-06, ..., -1.2688e-05, + 0.0000e+00, 1.4648e-05], + [ 1.2688e-05, 5.6550e-06, 7.6368e-06, ..., 7.6890e-06, + 0.0000e+00, 1.7881e-05], + [-6.6683e-06, 1.2808e-05, -1.1817e-05, ..., -2.2754e-05, + 0.0000e+00, 1.3858e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0153, -0.0143, -0.0266, -0.0250, -0.0101, 0.0046, 0.0084, -0.0230, + -0.0087, 0.0062], device='cuda:0'), grad: tensor([-1.9073e-04, -8.6248e-05, 1.9014e-04, -5.2840e-05, 6.4611e-05, + -7.3254e-05, -1.1716e-06, 6.6340e-05, 6.0350e-05, 2.2903e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 221.07, cls_loss 0.0050 cls_loss_mapping 0.0152 cls_loss_causal 0.6401 re_mapping 0.0115 re_causal 0.0357 /// teacc 98.70 lr 0.00010000 +Epoch 61, weight, value: tensor([[-8.7586e-02, -8.5348e-02, 6.3669e-02, ..., -2.8630e-02, + 3.5521e-02, 1.8728e-02], + [-3.7998e-02, -1.7688e-02, -2.5157e-02, ..., -3.8419e-02, + -6.7636e-02, -3.9765e-02], + [ 3.9595e-02, -3.4645e-02, -1.4802e-02, ..., -2.5841e-02, + -9.1996e-03, -1.7051e-01], + ..., + [-4.7898e-02, 5.6211e-02, -9.4824e-05, ..., 6.3397e-02, + -2.7647e-02, -6.6495e-02], + [-1.7915e-02, -3.5388e-02, 6.7867e-02, ..., -3.3921e-02, + -1.1050e-01, 7.0252e-02], + [-1.6809e-02, -8.3604e-02, 5.5888e-02, ..., 4.6394e-02, + -1.1937e-01, -4.0386e-02]], device='cuda:0'), grad: tensor([[ 4.2804e-06, 7.5027e-06, 6.4075e-05, ..., 2.4103e-06, + -1.3653e-06, 9.1970e-05], + [ 3.5129e-06, 2.2724e-06, 1.1182e-04, ..., 4.1462e-06, + 9.6159e-08, 9.9897e-05], + [ 8.4788e-06, 2.5898e-05, 2.7871e-04, ..., 3.2112e-06, + -1.2163e-06, 3.5381e-04], + ..., + [ 3.9004e-06, -2.0638e-05, 7.6294e-05, ..., -6.0722e-07, + 8.5495e-07, 1.2088e-04], + [-1.5885e-05, -3.9965e-05, -6.2704e-04, ..., 3.5949e-06, + 4.1490e-07, -7.7486e-04], + [ 6.9849e-06, 1.3098e-05, 4.3549e-06, ..., 2.2918e-05, + 1.2829e-07, 2.0817e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0151, -0.0139, -0.0270, -0.0251, -0.0104, 0.0051, 0.0078, -0.0229, + -0.0079, 0.0057], device='cuda:0'), grad: tensor([ 1.4710e-04, -1.8030e-05, 5.7602e-04, -8.5950e-05, -7.0274e-05, + 1.9622e-04, 5.2750e-05, 2.5415e-04, -1.1501e-03, 9.7215e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 221.05, cls_loss 0.0050 cls_loss_mapping 0.0139 cls_loss_causal 0.6380 re_mapping 0.0114 re_causal 0.0361 /// teacc 98.90 lr 0.00010000 +Epoch 62, weight, value: tensor([[-8.9935e-02, -8.6741e-02, 6.3638e-02, ..., -2.9091e-02, + 3.5424e-02, 1.8451e-02], + [-3.8314e-02, -1.7649e-02, -2.5083e-02, ..., -3.7885e-02, + -6.7814e-02, -3.9270e-02], + [ 3.9670e-02, -3.5146e-02, -1.5428e-02, ..., -2.6617e-02, + -8.9856e-03, -1.7238e-01], + ..., + [-4.8691e-02, 5.6915e-02, -1.2902e-04, ..., 6.3803e-02, + -2.7468e-02, -6.6767e-02], + [-1.8037e-02, -3.5685e-02, 6.8185e-02, ..., -3.4353e-02, + -1.1063e-01, 7.1167e-02], + [-1.5477e-02, -8.4803e-02, 5.6596e-02, ..., 4.6860e-02, + -1.1947e-01, -4.0722e-02]], device='cuda:0'), grad: tensor([[ 1.5497e-05, 6.2473e-06, 1.5497e-04, ..., 6.3255e-06, + 1.0765e-04, 4.2009e-04], + [ 5.0664e-07, 3.4511e-05, 2.4676e-05, ..., 3.4511e-05, + 7.1526e-07, -3.3528e-05], + [ 3.2061e-07, 1.7226e-05, 1.2994e-05, ..., 1.3366e-05, + -6.7949e-06, 6.2026e-06], + ..., + [ 8.8383e-07, -1.0099e-03, -1.0519e-03, ..., -1.0061e-03, + 1.6466e-06, -7.7039e-06], + [ 8.3894e-06, 1.6272e-05, 3.2097e-05, ..., 2.1622e-05, + 8.3745e-06, 5.2243e-05], + [ 5.7667e-06, 8.7309e-04, 9.2268e-04, ..., 9.9754e-04, + 2.7735e-06, 4.3213e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0156, -0.0130, -0.0269, -0.0251, -0.0110, 0.0048, 0.0078, -0.0231, + -0.0077, 0.0059], device='cuda:0'), grad: tensor([ 5.1689e-04, -6.9976e-05, 6.9058e-07, 1.4412e-04, -3.5620e-04, + 1.1826e-04, -6.5422e-04, -2.2984e-03, 1.2827e-04, 2.4681e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 220.74, cls_loss 0.0050 cls_loss_mapping 0.0175 cls_loss_causal 0.6167 re_mapping 0.0113 re_causal 0.0350 /// teacc 98.81 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0911, -0.0882, 0.0642, ..., -0.0294, 0.0358, 0.0187], + [-0.0388, -0.0173, -0.0250, ..., -0.0378, -0.0683, -0.0389], + [ 0.0396, -0.0360, -0.0157, ..., -0.0269, -0.0084, -0.1735], + ..., + [-0.0487, 0.0578, -0.0003, ..., 0.0639, -0.0268, -0.0672], + [-0.0201, -0.0362, 0.0682, ..., -0.0346, -0.1119, 0.0715], + [-0.0134, -0.0861, 0.0572, ..., 0.0472, -0.1200, -0.0402]], + device='cuda:0'), grad: tensor([[ 3.4249e-07, 1.5190e-06, 1.9418e-07, ..., 7.9162e-07, + 2.8359e-07, 2.5053e-06], + [ 6.3796e-08, 8.6576e-06, 3.4738e-06, ..., 3.9898e-06, + 7.1246e-08, -3.4384e-06], + [ 1.1967e-07, 6.7130e-06, 1.8880e-05, ..., 6.3609e-07, + 3.8883e-08, 2.2277e-05], + ..., + [ 4.8522e-07, -2.5436e-05, -6.9439e-06, ..., -1.3456e-05, + 4.4238e-09, 3.6173e-06], + [ 3.4040e-07, 1.9576e-06, -3.1948e-05, ..., 4.6147e-07, + 2.3562e-07, -4.1008e-05], + [-1.2922e-07, 8.4192e-06, 2.8294e-06, ..., 4.1048e-07, + 2.3982e-08, 7.6182e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0154, -0.0130, -0.0267, -0.0252, -0.0111, 0.0043, 0.0079, -0.0232, + -0.0077, 0.0064], device='cuda:0'), grad: tensor([ 3.9786e-06, 1.4156e-06, 3.8862e-05, 3.9488e-06, 3.1795e-06, + -3.8929e-06, 2.8480e-06, -1.7554e-05, -4.9859e-05, 1.7032e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 220.62, cls_loss 0.0035 cls_loss_mapping 0.0118 cls_loss_causal 0.6114 re_mapping 0.0110 re_causal 0.0341 /// teacc 98.94 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0920, -0.0888, 0.0641, ..., -0.0308, 0.0357, 0.0191], + [-0.0395, -0.0178, -0.0255, ..., -0.0386, -0.0689, -0.0389], + [ 0.0403, -0.0365, -0.0161, ..., -0.0271, -0.0082, -0.1743], + ..., + [-0.0491, 0.0585, -0.0003, ..., 0.0644, -0.0270, -0.0677], + [-0.0213, -0.0367, 0.0684, ..., -0.0347, -0.1121, 0.0718], + [-0.0142, -0.0870, 0.0577, ..., 0.0471, -0.1202, -0.0407]], + device='cuda:0'), grad: tensor([[ 9.9614e-06, 9.5963e-06, -4.3452e-05, ..., 9.2909e-06, + -7.4739e-07, -3.3885e-05], + [ 4.6581e-05, 6.9812e-06, 6.6049e-06, ..., 1.1690e-05, + 8.7917e-06, 2.6878e-06], + [-5.1111e-05, 1.7792e-05, 1.9968e-05, ..., 5.5492e-05, + 2.2799e-05, 1.1109e-05], + ..., + [-2.7232e-06, -5.7817e-05, -7.0557e-06, ..., -2.2173e-04, + 7.8753e-06, -1.0513e-05], + [ 1.5736e-05, 2.2426e-05, 2.2545e-05, ..., 6.7204e-06, + 1.5087e-06, 1.1288e-05], + [ 7.4804e-05, 1.6272e-05, 3.7700e-05, ..., -3.9250e-05, + 8.2329e-07, 9.6500e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0157, -0.0132, -0.0268, -0.0252, -0.0099, 0.0046, 0.0076, -0.0231, + -0.0079, 0.0059], device='cuda:0'), grad: tensor([-9.4414e-05, 5.0545e-04, -2.8753e-04, 4.5300e-05, 1.7357e-04, + -1.8716e-04, 5.1796e-05, -4.3631e-04, 7.7307e-05, 1.5175e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 221.45, cls_loss 0.0045 cls_loss_mapping 0.0141 cls_loss_causal 0.6705 re_mapping 0.0113 re_causal 0.0358 /// teacc 98.92 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0931, -0.0889, 0.0644, ..., -0.0311, 0.0357, 0.0196], + [-0.0399, -0.0187, -0.0261, ..., -0.0393, -0.0693, -0.0387], + [ 0.0401, -0.0367, -0.0165, ..., -0.0274, -0.0079, -0.1755], + ..., + [-0.0498, 0.0582, -0.0005, ..., 0.0646, -0.0266, -0.0690], + [-0.0223, -0.0371, 0.0692, ..., -0.0341, -0.1128, 0.0726], + [-0.0140, -0.0877, 0.0581, ..., 0.0474, -0.1207, -0.0415]], + device='cuda:0'), grad: tensor([[ 1.1288e-06, -2.1867e-06, -6.5804e-05, ..., 8.2701e-07, + 1.4983e-07, -5.1796e-05], + [ 7.2364e-07, 1.4678e-06, 4.3251e-06, ..., 2.2091e-06, + 8.2422e-08, -8.7768e-06], + [ 3.3062e-07, 2.3749e-06, 1.3240e-05, ..., 1.2387e-06, + -1.1250e-05, 1.4104e-05], + ..., + [ 2.7437e-06, 2.8331e-06, 2.4378e-05, ..., 1.5929e-05, + 3.1409e-07, 1.8209e-05], + [ 7.1004e-06, 2.2501e-05, 4.2409e-05, ..., 9.0227e-06, + 9.7901e-06, 2.6941e-05], + [-1.2711e-05, 2.6420e-05, -5.5403e-05, ..., -4.9025e-05, + 6.5193e-09, -1.6555e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0155, -0.0140, -0.0261, -0.0248, -0.0098, 0.0046, 0.0072, -0.0236, + -0.0076, 0.0059], device='cuda:0'), grad: tensor([-1.0228e-04, -1.3299e-06, -7.1287e-05, -7.7128e-05, 1.5363e-05, + 3.8669e-06, 1.2077e-05, 7.1824e-05, 1.7154e-04, -2.2575e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 220.71, cls_loss 0.0052 cls_loss_mapping 0.0170 cls_loss_causal 0.6611 re_mapping 0.0108 re_causal 0.0334 /// teacc 98.82 lr 0.00010000 +Epoch 66, weight, value: tensor([[-9.5910e-02, -8.9588e-02, 6.4716e-02, ..., -3.1470e-02, + 3.5619e-02, 1.9589e-02], + [-4.0356e-02, -1.8216e-02, -2.6165e-02, ..., -4.0121e-02, + -6.9409e-02, -3.8593e-02], + [ 3.9682e-02, -3.8073e-02, -1.7227e-02, ..., -2.8051e-02, + -7.5368e-03, -1.7700e-01], + ..., + [-5.1391e-02, 5.9070e-02, -1.7051e-04, ..., 6.5523e-02, + -2.6714e-02, -6.9117e-02], + [-2.0215e-02, -3.7696e-02, 6.9631e-02, ..., -3.4256e-02, + -1.1316e-01, 7.4453e-02], + [-1.4236e-02, -8.9063e-02, 5.8154e-02, ..., 4.7135e-02, + -1.2092e-01, -4.2342e-02]], device='cuda:0'), grad: tensor([[ 3.0547e-06, 3.5763e-06, -9.6336e-06, ..., 3.5707e-06, + 3.8766e-07, -3.9414e-06], + [ 5.4911e-06, 4.1649e-06, 2.3320e-06, ..., 3.2745e-06, + -5.2527e-07, -6.2771e-06], + [-7.6175e-05, 7.6182e-06, 4.7088e-06, ..., 2.1923e-06, + -1.2922e-08, 2.9299e-06], + ..., + [ 8.0690e-06, -2.3261e-05, -1.2346e-05, ..., -1.5885e-05, + 3.5274e-07, 3.5539e-06], + [ 3.1024e-05, 9.6709e-06, -6.4857e-06, ..., 3.6396e-06, + 6.8638e-07, -9.5069e-06], + [ 2.4401e-06, 3.6597e-05, 3.1382e-05, ..., 2.9945e-04, + 1.2200e-07, 9.7603e-06]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0157, -0.0138, -0.0266, -0.0249, -0.0094, 0.0043, 0.0070, -0.0229, + -0.0068, 0.0050], device='cuda:0'), grad: tensor([ 1.5691e-05, 3.3647e-05, -6.1035e-04, -1.0931e-04, -5.5122e-04, + 2.1446e-04, 6.3121e-05, 8.3566e-05, 2.6560e-04, 5.9366e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 220.94, cls_loss 0.0043 cls_loss_mapping 0.0141 cls_loss_causal 0.6327 re_mapping 0.0108 re_causal 0.0334 /// teacc 98.86 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0966, -0.0902, 0.0650, ..., -0.0322, 0.0358, 0.0194], + [-0.0405, -0.0189, -0.0266, ..., -0.0406, -0.0700, -0.0385], + [ 0.0396, -0.0382, -0.0176, ..., -0.0281, -0.0068, -0.1782], + ..., + [-0.0520, 0.0601, 0.0002, ..., 0.0662, -0.0267, -0.0695], + [-0.0206, -0.0385, 0.0702, ..., -0.0340, -0.1151, 0.0751], + [-0.0144, -0.0909, 0.0583, ..., 0.0473, -0.1216, -0.0436]], + device='cuda:0'), grad: tensor([[ 1.6214e-06, 9.0674e-06, 2.1085e-06, ..., 1.3962e-05, + 1.1467e-07, 1.3992e-05], + [ 3.0436e-06, 1.9348e-04, 6.0201e-05, ..., 2.5797e-04, + 7.6182e-07, -4.0698e-04], + [ 5.6997e-07, 4.9919e-05, 1.6585e-05, ..., 1.5900e-05, + -2.5537e-06, 1.9521e-05], + ..., + [ 2.2314e-06, -3.5954e-04, -1.4818e-04, ..., -4.5300e-04, + 4.4424e-07, -2.6727e-04], + [ 7.9572e-06, 1.0258e-04, 8.2195e-05, ..., 1.4198e-04, + 7.3947e-07, 4.2200e-04], + [-5.2415e-06, 4.5002e-05, -1.7476e-04, ..., 1.5945e-03, + 2.3283e-08, 5.4884e-04]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0158, -0.0139, -0.0265, -0.0249, -0.0097, 0.0043, 0.0075, -0.0227, + -0.0065, 0.0045], device='cuda:0'), grad: tensor([ 3.9697e-05, -2.0936e-05, 1.4150e-04, -2.2328e-04, -4.0092e-03, + 1.4412e-04, 1.1539e-04, -9.1124e-04, 7.5865e-04, 3.9635e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 220.89, cls_loss 0.0036 cls_loss_mapping 0.0115 cls_loss_causal 0.6132 re_mapping 0.0105 re_causal 0.0332 /// teacc 98.82 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0973, -0.0910, 0.0652, ..., -0.0326, 0.0356, 0.0193], + [-0.0416, -0.0193, -0.0271, ..., -0.0413, -0.0700, -0.0387], + [ 0.0395, -0.0389, -0.0180, ..., -0.0285, -0.0057, -0.1788], + ..., + [-0.0528, 0.0606, -0.0002, ..., 0.0661, -0.0270, -0.0699], + [-0.0213, -0.0392, 0.0701, ..., -0.0348, -0.1161, 0.0754], + [-0.0131, -0.0914, 0.0597, ..., 0.0483, -0.1223, -0.0436]], + device='cuda:0'), grad: tensor([[ 1.4054e-06, 1.2666e-05, -1.0245e-05, ..., 4.8392e-06, + -1.3718e-06, -1.1474e-05], + [ 5.8208e-07, 9.3281e-06, 5.0403e-06, ..., 2.3210e-04, + 4.3074e-08, -5.4762e-06], + [-7.4580e-06, 6.6662e-04, 6.8471e-06, ..., 3.0339e-05, + 2.5658e-07, 1.4514e-05], + ..., + [ 2.2277e-06, -1.2562e-05, -1.6555e-05, ..., -1.7017e-05, + 1.5192e-07, 4.7833e-06], + [ 2.6226e-06, 3.7223e-05, -1.2740e-05, ..., 1.2107e-05, + 9.0455e-08, -1.8924e-05], + [-3.1367e-06, 2.9102e-05, 1.8124e-06, ..., 3.7104e-05, + 2.0128e-07, 8.2254e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0161, -0.0140, -0.0266, -0.0247, -0.0098, 0.0045, 0.0075, -0.0230, + -0.0070, 0.0053], device='cuda:0'), grad: tensor([ 2.7135e-05, 1.0939e-03, 1.6813e-03, -2.0046e-03, -1.3571e-03, + 7.3731e-05, 3.6329e-05, 1.1641e-04, 1.6308e-04, 1.6844e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 220.91, cls_loss 0.0052 cls_loss_mapping 0.0138 cls_loss_causal 0.6095 re_mapping 0.0107 re_causal 0.0315 /// teacc 98.89 lr 0.00010000 +Epoch 69, weight, value: tensor([[-9.9328e-02, -9.2960e-02, 6.5479e-02, ..., -3.3219e-02, + 3.5898e-02, 1.9430e-02], + [-4.2223e-02, -1.9435e-02, -2.7322e-02, ..., -4.2476e-02, + -7.0213e-02, -3.7225e-02], + [ 3.8645e-02, -3.9803e-02, -1.8525e-02, ..., -2.8716e-02, + -5.1369e-03, -1.7998e-01], + ..., + [-5.4270e-02, 6.1442e-02, 7.7591e-05, ..., 6.6688e-02, + -2.7054e-02, -7.0322e-02], + [-2.2508e-02, -3.9859e-02, 6.9951e-02, ..., -3.5444e-02, + -1.1677e-01, 7.5236e-02], + [-1.1081e-02, -9.2710e-02, 6.0678e-02, ..., 4.8831e-02, + -1.2263e-01, -4.2861e-02]], device='cuda:0'), grad: tensor([[ 1.4659e-06, 7.4245e-06, -1.0222e-04, ..., 6.0163e-06, + -2.5183e-06, -4.1306e-05], + [ 4.2003e-07, 1.2025e-05, 1.6406e-05, ..., 1.8757e-06, + -8.2329e-06, -2.2724e-06], + [ 8.0233e-07, 2.1607e-05, 2.3246e-05, ..., 7.9945e-06, + 1.7844e-06, 1.2398e-05], + ..., + [ 1.6708e-06, -8.1360e-05, -3.8087e-05, ..., -9.9182e-05, + 3.5577e-06, 9.6187e-06], + [ 6.3470e-07, 7.2047e-06, -9.1121e-06, ..., 9.1502e-08, + 6.6916e-07, -1.4260e-05], + [ 1.2539e-05, 7.4625e-05, 5.9277e-05, ..., 4.8429e-05, + 1.4780e-06, 2.9534e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0162, -0.0134, -0.0271, -0.0241, -0.0100, 0.0040, 0.0072, -0.0230, + -0.0078, 0.0062], device='cuda:0'), grad: tensor([-1.9825e-04, -1.3798e-05, 7.6771e-05, -7.8380e-05, 3.5167e-05, + 1.5199e-05, 2.6479e-05, -9.0897e-05, 2.3872e-05, 2.0397e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 221.14, cls_loss 0.0041 cls_loss_mapping 0.0124 cls_loss_causal 0.6249 re_mapping 0.0107 re_causal 0.0326 /// teacc 98.93 lr 0.00010000 +Epoch 70, weight, value: tensor([[-1.0028e-01, -9.4065e-02, 6.5735e-02, ..., -3.4036e-02, + 3.5948e-02, 1.9471e-02], + [-4.3422e-02, -1.9520e-02, -2.7693e-02, ..., -4.3268e-02, + -7.0333e-02, -3.7810e-02], + [ 3.8321e-02, -4.0124e-02, -1.8931e-02, ..., -2.8981e-02, + -4.3879e-03, -1.8022e-01], + ..., + [-5.5846e-02, 6.1808e-02, -2.8262e-05, ..., 6.7153e-02, + -2.7870e-02, -7.1521e-02], + [-2.5063e-02, -4.0118e-02, 7.0309e-02, ..., -3.5769e-02, + -1.1718e-01, 7.5464e-02], + [-1.0742e-02, -9.3598e-02, 6.1329e-02, ..., 4.9233e-02, + -1.2344e-01, -4.3237e-02]], device='cuda:0'), grad: tensor([[ 9.6634e-06, 7.7533e-08, 1.5721e-05, ..., 1.4164e-05, + 9.7556e-08, 1.2398e-05], + [ 2.0526e-06, 3.9837e-07, 1.7062e-05, ..., 1.5780e-05, + -1.3085e-06, 3.6228e-06], + [ 8.8364e-06, 7.2550e-07, -6.6519e-05, ..., -9.9897e-05, + 3.7206e-07, 1.9699e-05], + ..., + [ 3.5353e-06, -1.7118e-06, 4.2021e-05, ..., 4.0144e-05, + 1.8510e-07, 1.3083e-05], + [ 3.1125e-06, 8.7777e-08, 5.0575e-05, ..., 7.3612e-05, + 1.8766e-07, -4.8503e-06], + [-4.8280e-05, 9.4669e-07, -1.5652e-04, ..., -1.0288e-04, + 3.8417e-08, -1.0014e-04]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0163, -0.0143, -0.0261, -0.0243, -0.0101, 0.0049, 0.0071, -0.0233, + -0.0081, 0.0062], device='cuda:0'), grad: tensor([ 1.4114e-04, 2.3353e-04, -1.7042e-03, 2.2817e-04, 1.6201e-04, + 4.4554e-05, 7.4744e-05, 4.4107e-04, 7.7963e-04, -3.9983e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 221.15, cls_loss 0.0042 cls_loss_mapping 0.0139 cls_loss_causal 0.6302 re_mapping 0.0103 re_causal 0.0303 /// teacc 99.03 lr 0.00010000 +Epoch 71, weight, value: tensor([[-1.0128e-01, -9.5025e-02, 6.6271e-02, ..., -3.4453e-02, + 3.5453e-02, 1.9490e-02], + [-4.4875e-02, -1.9944e-02, -2.8387e-02, ..., -4.3862e-02, + -7.0604e-02, -3.8275e-02], + [ 3.9023e-02, -4.0689e-02, -1.9194e-02, ..., -2.9701e-02, + -4.1942e-03, -1.8142e-01], + ..., + [-5.8773e-02, 6.2668e-02, 1.5886e-04, ..., 6.7803e-02, + -2.7503e-02, -7.1914e-02], + [-2.5152e-02, -4.0960e-02, 7.0609e-02, ..., -3.6206e-02, + -1.1760e-01, 7.6072e-02], + [-1.1733e-02, -9.4554e-02, 6.1799e-02, ..., 4.9529e-02, + -1.2386e-01, -4.4147e-02]], device='cuda:0'), grad: tensor([[-3.9667e-05, 4.1761e-06, -2.1553e-04, ..., -9.2527e-07, + 0.0000e+00, -1.1730e-04], + [ 1.1874e-06, 7.5363e-06, 9.7752e-06, ..., 5.8077e-06, + 0.0000e+00, -1.0312e-05], + [ 8.1509e-06, 5.5432e-06, 3.7372e-05, ..., 7.1637e-06, + 0.0000e+00, 3.2157e-05], + ..., + [ 5.1782e-06, -1.0139e-04, -8.8155e-05, ..., -5.9903e-05, + 0.0000e+00, 6.8098e-06], + [-1.6034e-05, 2.6226e-05, -4.5270e-05, ..., -5.3048e-06, + 0.0000e+00, -7.4148e-05], + [ 3.0905e-05, 2.3305e-05, 2.0158e-04, ..., 1.3709e-06, + 0.0000e+00, 1.5104e-04]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0158, -0.0146, -0.0259, -0.0252, -0.0101, 0.0058, 0.0073, -0.0232, + -0.0081, 0.0058], device='cuda:0'), grad: tensor([-3.5024e-04, -4.1485e-05, 9.1970e-05, 5.7399e-05, 3.3557e-05, + 7.9155e-05, -5.0098e-05, -1.3900e-04, -2.5943e-05, 3.4428e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 221.02, cls_loss 0.0040 cls_loss_mapping 0.0122 cls_loss_causal 0.6702 re_mapping 0.0098 re_causal 0.0307 /// teacc 98.85 lr 0.00010000 +Epoch 72, weight, value: tensor([[-1.0383e-01, -9.6180e-02, 6.5927e-02, ..., -3.6494e-02, + 3.5432e-02, 1.9657e-02], + [-4.5761e-02, -1.9981e-02, -2.8816e-02, ..., -4.4818e-02, + -7.0526e-02, -3.8264e-02], + [ 3.9723e-02, -4.1471e-02, -1.9756e-02, ..., -3.0688e-02, + -4.1340e-03, -1.8225e-01], + ..., + [-6.2494e-02, 6.3408e-02, -7.6233e-05, ..., 6.8104e-02, + -2.7516e-02, -7.2993e-02], + [-2.5614e-02, -4.1368e-02, 7.1568e-02, ..., -3.5893e-02, + -1.1780e-01, 7.6823e-02], + [-1.1123e-02, -9.5848e-02, 6.2688e-02, ..., 5.0189e-02, + -1.2414e-01, -4.4614e-02]], device='cuda:0'), grad: tensor([[ 9.2201e-07, 8.3074e-06, -7.1347e-05, ..., 1.1548e-06, + -4.1956e-07, -4.2945e-05], + [ 1.0841e-06, 1.7822e-05, 8.2627e-06, ..., 4.9025e-05, + 1.6834e-07, 5.4613e-06], + [ 1.8198e-06, 3.8087e-05, 1.8492e-05, ..., 3.3639e-06, + -1.1781e-07, 1.5005e-05], + ..., + [ 2.1011e-06, 3.0965e-05, 1.5914e-05, ..., 4.3958e-06, + 1.9884e-07, 6.2399e-06], + [-2.2948e-05, 8.3387e-05, -8.0690e-06, ..., 2.7679e-06, + 2.8871e-08, -1.7154e-04], + [-2.0877e-05, 3.9607e-05, 1.0163e-05, ..., -5.3287e-05, + 1.0314e-07, 2.7716e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0165, -0.0139, -0.0266, -0.0252, -0.0098, 0.0054, 0.0076, -0.0235, + -0.0077, 0.0060], device='cuda:0'), grad: tensor([-1.1939e-04, 3.4308e-04, 7.6294e-05, -4.5586e-04, -2.9516e-04, + 2.3723e-04, 4.7684e-05, 1.1444e-04, -2.1517e-05, 7.2956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 221.38, cls_loss 0.0029 cls_loss_mapping 0.0105 cls_loss_causal 0.6300 re_mapping 0.0102 re_causal 0.0316 /// teacc 98.96 lr 0.00010000 +Epoch 73, weight, value: tensor([[-1.0472e-01, -9.7094e-02, 6.6779e-02, ..., -3.6808e-02, + 3.5866e-02, 2.0704e-02], + [-4.6260e-02, -1.9846e-02, -2.9030e-02, ..., -4.5041e-02, + -7.0563e-02, -3.8531e-02], + [ 3.9378e-02, -4.2009e-02, -2.0163e-02, ..., -3.0945e-02, + -3.9697e-03, -1.8269e-01], + ..., + [-6.5326e-02, 6.3799e-02, 7.3742e-05, ..., 6.8669e-02, + -2.7593e-02, -7.3399e-02], + [-2.5925e-02, -4.1601e-02, 7.1880e-02, ..., -3.6257e-02, + -1.1793e-01, 7.7332e-02], + [-1.1195e-02, -9.6822e-02, 6.2974e-02, ..., 5.0153e-02, + -1.2440e-01, -4.4980e-02]], device='cuda:0'), grad: tensor([[ 5.5879e-08, 3.8147e-06, 9.4948e-07, ..., 9.6038e-06, + 4.5868e-08, 4.9211e-06], + [ 5.3411e-07, 9.9242e-05, 4.6790e-05, ..., 8.5711e-05, + 1.3597e-07, 6.7353e-06], + [ 4.3074e-08, 8.4221e-05, 5.1677e-05, ..., 2.9095e-06, + -4.2794e-07, 9.9316e-06], + ..., + [ 3.5367e-07, -2.2817e-04, -1.3697e-04, ..., -1.2994e-04, + 7.9162e-08, 3.1348e-06], + [ 7.7765e-08, 2.4110e-05, 1.8850e-05, ..., 2.2829e-05, + 3.5623e-08, -1.1146e-05], + [-1.1623e-06, 1.1042e-05, -5.0757e-07, ..., 1.4216e-05, + 1.0943e-08, 1.6475e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0158, -0.0137, -0.0265, -0.0254, -0.0096, 0.0055, 0.0076, -0.0236, + -0.0077, 0.0056], device='cuda:0'), grad: tensor([ 5.1349e-05, 1.9932e-04, -1.0353e-04, 9.8422e-06, 4.2510e-04, + 1.0258e-04, -6.3467e-04, -1.4627e-04, 4.9919e-05, 4.5598e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 220.79, cls_loss 0.0034 cls_loss_mapping 0.0105 cls_loss_causal 0.6402 re_mapping 0.0099 re_causal 0.0318 /// teacc 98.96 lr 0.00010000 +Epoch 74, weight, value: tensor([[-1.0622e-01, -9.8084e-02, 6.7103e-02, ..., -3.7010e-02, + 3.5898e-02, 2.0848e-02], + [-4.6850e-02, -2.0325e-02, -2.9660e-02, ..., -4.5662e-02, + -7.0578e-02, -3.8688e-02], + [ 3.9122e-02, -4.2322e-02, -2.0700e-02, ..., -3.1431e-02, + -3.8751e-03, -1.8379e-01], + ..., + [-6.5980e-02, 6.4523e-02, 7.6562e-05, ..., 6.9139e-02, + -2.7647e-02, -7.3702e-02], + [-2.5934e-02, -4.2124e-02, 7.2241e-02, ..., -3.6740e-02, + -1.1800e-01, 7.8442e-02], + [-1.0924e-02, -9.7798e-02, 6.3615e-02, ..., 5.0389e-02, + -1.2446e-01, -4.5077e-02]], device='cuda:0'), grad: tensor([[ 8.1491e-07, 8.2701e-07, -1.5783e-04, ..., -1.2696e-05, + -4.8429e-05, -1.0395e-04], + [ 3.1125e-06, 6.3479e-06, 4.0419e-06, ..., 7.9125e-06, + 9.8161e-07, 3.0212e-06], + [ 1.7155e-06, 2.2545e-05, 9.9838e-05, ..., 3.0667e-05, + 2.3216e-05, 7.4863e-05], + ..., + [ 7.8371e-07, -4.6939e-05, 9.9763e-06, ..., -3.7044e-05, + 1.0990e-05, 1.5289e-05], + [ 2.9262e-06, 2.6915e-06, 5.1707e-06, ..., 1.0645e-06, + 9.1270e-07, 8.4341e-06], + [ 4.3884e-06, 8.9854e-06, 1.1161e-05, ..., 1.0498e-05, + 3.7495e-06, 9.6858e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0154, -0.0140, -0.0266, -0.0257, -0.0094, 0.0054, 0.0070, -0.0235, + -0.0070, 0.0055], device='cuda:0'), grad: tensor([-3.8552e-04, 7.0572e-05, 3.4356e-04, 4.8071e-05, -5.1558e-05, + -6.3926e-06, -1.1837e-04, -5.2191e-06, 3.0965e-05, 7.3433e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 221.18, cls_loss 0.0044 cls_loss_mapping 0.0136 cls_loss_causal 0.5990 re_mapping 0.0093 re_causal 0.0275 /// teacc 98.93 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.1078, -0.0998, 0.0673, ..., -0.0377, 0.0358, 0.0214], + [-0.0468, -0.0204, -0.0301, ..., -0.0460, -0.0706, -0.0382], + [ 0.0393, -0.0429, -0.0208, ..., -0.0320, -0.0035, -0.1850], + ..., + [-0.0673, 0.0662, 0.0021, ..., 0.0711, -0.0280, -0.0727], + [-0.0272, -0.0434, 0.0721, ..., -0.0375, -0.1182, 0.0786], + [-0.0099, -0.1006, 0.0633, ..., 0.0499, -0.1247, -0.0460]], + device='cuda:0'), grad: tensor([[ 7.0967e-07, 1.2731e-06, -1.2591e-05, ..., 1.9465e-06, + 1.0710e-08, -1.3702e-05], + [ 1.4789e-06, 8.8215e-06, 9.3365e-07, ..., 5.6326e-06, + 1.7276e-07, 3.6359e-06], + [ 6.0536e-07, 1.3039e-05, 1.1511e-05, ..., 7.0781e-06, + 1.5530e-07, 1.5438e-05], + ..., + [ 7.0296e-06, -2.0951e-05, 5.7012e-05, ..., 1.9029e-05, + -8.7777e-07, 8.5756e-06], + [ 1.7826e-06, 6.4448e-07, -1.7896e-05, ..., -4.1015e-06, + 2.0256e-08, -1.7926e-05], + [ 7.0743e-06, -3.0547e-05, -3.0637e-05, ..., 9.0837e-05, + 1.3039e-08, 1.0520e-05]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0158, -0.0137, -0.0263, -0.0247, -0.0099, 0.0040, 0.0076, -0.0221, + -0.0077, 0.0046], device='cuda:0'), grad: tensor([-1.9521e-05, -3.2485e-06, 4.9204e-05, 4.3303e-05, -2.5463e-04, + 3.3945e-05, -7.7307e-05, 4.6074e-05, -1.0580e-06, 1.8299e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 220.59, cls_loss 0.0033 cls_loss_mapping 0.0101 cls_loss_causal 0.6326 re_mapping 0.0092 re_causal 0.0293 /// teacc 98.90 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.1085, -0.1008, 0.0676, ..., -0.0383, 0.0359, 0.0224], + [-0.0479, -0.0204, -0.0307, ..., -0.0469, -0.0703, -0.0379], + [ 0.0403, -0.0425, -0.0214, ..., -0.0323, -0.0036, -0.1860], + ..., + [-0.0693, 0.0664, 0.0020, ..., 0.0716, -0.0282, -0.0735], + [-0.0285, -0.0440, 0.0722, ..., -0.0380, -0.1184, 0.0786], + [-0.0101, -0.1014, 0.0641, ..., 0.0501, -0.1250, -0.0460]], + device='cuda:0'), grad: tensor([[-8.6799e-06, 9.8627e-07, -1.7092e-05, ..., 3.2037e-06, + 6.6403e-07, -7.4446e-05], + [ 8.3586e-07, -1.8165e-05, 1.4165e-06, ..., -1.4706e-06, + -1.2070e-05, -2.9318e-06], + [ 1.1036e-06, 5.0887e-06, 4.5411e-06, ..., 5.0776e-06, + -2.0256e-07, 7.2867e-06], + ..., + [ 3.6340e-06, -2.0862e-07, -9.2248e-07, ..., 4.6864e-06, + 6.3367e-06, 5.9046e-06], + [ 7.5437e-06, 1.4715e-06, -9.9540e-06, ..., 1.1377e-05, + 2.1197e-06, -4.0978e-06], + [-3.3919e-06, 3.6247e-06, -1.3530e-05, ..., -9.7007e-06, + 1.8161e-07, 8.3596e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0151, -0.0135, -0.0259, -0.0249, -0.0095, 0.0043, 0.0072, -0.0226, + -0.0081, 0.0045], device='cuda:0'), grad: tensor([-8.6725e-05, -1.6594e-04, 2.5973e-05, 2.6539e-05, -9.2626e-05, + 8.2493e-05, 3.9995e-05, 1.2612e-04, 4.2826e-05, 1.4910e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 220.28, cls_loss 0.0050 cls_loss_mapping 0.0130 cls_loss_causal 0.6295 re_mapping 0.0091 re_causal 0.0292 /// teacc 98.90 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.1096, -0.1016, 0.0675, ..., -0.0389, 0.0359, 0.0207], + [-0.0496, -0.0199, -0.0293, ..., -0.0449, -0.0701, -0.0381], + [ 0.0401, -0.0447, -0.0237, ..., -0.0329, -0.0036, -0.1892], + ..., + [-0.0706, 0.0669, 0.0015, ..., 0.0710, -0.0280, -0.0729], + [-0.0316, -0.0421, 0.0733, ..., -0.0385, -0.1186, 0.0775], + [-0.0090, -0.1022, 0.0648, ..., 0.0503, -0.1251, -0.0461]], + device='cuda:0'), grad: tensor([[ 4.4182e-06, 3.7029e-06, 8.2478e-06, ..., 7.6666e-06, + 5.0757e-08, -1.9111e-06], + [ 1.0014e-05, 4.2580e-06, 2.6941e-05, ..., 1.4380e-05, + 4.7497e-08, 2.7958e-06], + [ 8.9360e-07, 3.0398e-06, 3.6862e-06, ..., 1.2899e-06, + -1.7053e-06, 4.4517e-06], + ..., + [ 4.1842e-05, 2.7001e-05, 1.4150e-04, ..., 7.5519e-05, + 9.0711e-07, 5.6848e-06], + [-2.2769e-05, 2.4159e-06, 7.2941e-06, ..., 1.2539e-05, + 4.5355e-07, -1.0157e-04], + [-8.5831e-05, -6.8247e-05, -3.2806e-04, ..., -1.8466e-04, + 4.8894e-09, 6.0536e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0164, -0.0131, -0.0273, -0.0258, -0.0093, 0.0058, 0.0086, -0.0227, + -0.0082, 0.0045], device='cuda:0'), grad: tensor([ 2.4438e-05, 2.2888e-05, 3.5968e-06, 1.2350e-04, 3.1292e-05, + 1.7726e-04, 3.6448e-05, 3.1805e-04, -8.8692e-05, -6.4898e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 220.71, cls_loss 0.0032 cls_loss_mapping 0.0093 cls_loss_causal 0.5902 re_mapping 0.0098 re_causal 0.0301 /// teacc 98.88 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.1101, -0.1025, 0.0678, ..., -0.0393, 0.0359, 0.0209], + [-0.0504, -0.0203, -0.0302, ..., -0.0458, -0.0703, -0.0385], + [ 0.0400, -0.0452, -0.0241, ..., -0.0333, -0.0030, -0.1895], + ..., + [-0.0708, 0.0671, 0.0018, ..., 0.0717, -0.0282, -0.0731], + [-0.0319, -0.0426, 0.0739, ..., -0.0386, -0.1188, 0.0782], + [-0.0082, -0.1031, 0.0653, ..., 0.0506, -0.1255, -0.0462]], + device='cuda:0'), grad: tensor([[ 2.7404e-07, 1.5542e-05, 2.7299e-05, ..., 2.3078e-06, + 1.1824e-05, 3.3796e-05], + [ 4.6240e-07, 6.9499e-05, 3.1710e-05, ..., 3.4243e-05, + 2.6412e-06, 6.3330e-06], + [ 5.9605e-08, -5.7876e-05, 4.9710e-05, ..., 4.5329e-05, + -1.5199e-04, 1.8016e-05], + ..., + [ 9.3225e-07, -2.6345e-04, -1.0407e-04, ..., -1.3864e-04, + 1.8045e-05, 4.5523e-06], + [ 3.2000e-06, 2.8908e-05, -1.1230e-04, ..., 4.5598e-06, + 2.1726e-05, -1.4293e-04], + [-5.8487e-06, 1.3582e-05, -1.7196e-05, ..., -1.0751e-05, + 9.7416e-07, 8.6948e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0162, -0.0137, -0.0267, -0.0253, -0.0092, 0.0050, 0.0086, -0.0227, + -0.0081, 0.0046], device='cuda:0'), grad: tensor([ 1.1289e-04, 1.2720e-04, -3.6693e-04, 5.8460e-04, 5.5879e-05, + 8.1956e-06, 1.9133e-05, -3.6001e-04, -1.7643e-04, -4.9211e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 220.51, cls_loss 0.0033 cls_loss_mapping 0.0126 cls_loss_causal 0.6020 re_mapping 0.0096 re_causal 0.0293 /// teacc 99.01 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.1123, -0.1038, 0.0674, ..., -0.0405, 0.0361, 0.0211], + [-0.0514, -0.0207, -0.0307, ..., -0.0463, -0.0712, -0.0389], + [ 0.0400, -0.0453, -0.0245, ..., -0.0336, -0.0024, -0.1901], + ..., + [-0.0723, 0.0681, 0.0023, ..., 0.0724, -0.0277, -0.0734], + [-0.0331, -0.0432, 0.0741, ..., -0.0389, -0.1194, 0.0784], + [-0.0074, -0.1048, 0.0661, ..., 0.0507, -0.1259, -0.0464]], + device='cuda:0'), grad: tensor([[ 1.4296e-07, 7.6648e-07, -4.8317e-06, ..., 4.9686e-07, + -8.2422e-07, -2.5239e-06], + [ 3.8533e-07, 6.2361e-06, 2.0377e-06, ..., 3.8706e-06, + 9.5926e-08, -5.0589e-06], + [ 4.0862e-07, 1.0297e-05, 4.1053e-06, ..., 2.8946e-06, + 3.8673e-07, 3.0342e-06], + ..., + [ 2.7614e-07, -1.5222e-05, -1.1273e-05, ..., -2.3320e-05, + 4.2166e-07, 2.6282e-06], + [ 2.6282e-06, 6.1169e-06, 6.7465e-06, ..., 6.1952e-06, + 7.6322e-07, 6.9328e-06], + [-1.8403e-06, 1.3672e-05, -2.7977e-06, ..., -4.3213e-07, + 2.0326e-07, 1.0408e-07]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0168, -0.0141, -0.0265, -0.0252, -0.0093, 0.0052, 0.0087, -0.0223, + -0.0083, 0.0047], device='cuda:0'), grad: tensor([-6.4410e-06, -1.7518e-06, 1.7568e-05, -5.4896e-05, 6.5900e-06, + 2.9638e-05, -1.4208e-05, -1.6540e-05, 2.5243e-05, 1.4775e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 220.98, cls_loss 0.0034 cls_loss_mapping 0.0095 cls_loss_causal 0.5775 re_mapping 0.0090 re_causal 0.0276 /// teacc 98.96 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.1131, -0.1047, 0.0676, ..., -0.0410, 0.0362, 0.0214], + [-0.0526, -0.0206, -0.0312, ..., -0.0467, -0.0718, -0.0380], + [ 0.0396, -0.0456, -0.0247, ..., -0.0337, -0.0021, -0.1907], + ..., + [-0.0735, 0.0678, 0.0023, ..., 0.0726, -0.0278, -0.0738], + [-0.0338, -0.0437, 0.0742, ..., -0.0392, -0.1197, 0.0780], + [-0.0063, -0.1059, 0.0668, ..., 0.0513, -0.1266, -0.0465]], + device='cuda:0'), grad: tensor([[ 3.0873e-07, 1.0915e-06, -5.1521e-06, ..., 9.9279e-07, + -8.2422e-08, -1.1675e-05], + [ 6.1933e-08, 1.3098e-05, 5.5432e-06, ..., 1.0014e-05, + 1.8775e-06, 7.2300e-05], + [-7.8185e-07, 2.6319e-06, 1.1791e-06, ..., 8.0094e-07, + -1.6809e-05, -9.1493e-05], + ..., + [ 5.7556e-07, -3.5077e-05, -1.2867e-05, ..., -2.6822e-05, + 1.0477e-06, 7.6182e-07], + [ 7.7765e-07, 4.3809e-06, 4.3772e-08, ..., 3.1982e-06, + 9.2387e-07, 2.1860e-05], + [-2.7590e-07, 5.6624e-06, -3.7104e-06, ..., -6.2771e-07, + 6.3935e-07, 2.9355e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0171, -0.0139, -0.0266, -0.0241, -0.0097, 0.0051, 0.0091, -0.0226, + -0.0091, 0.0050], device='cuda:0'), grad: tensor([-4.3325e-06, 6.6519e-04, -8.3733e-04, 5.3979e-06, 3.1330e-06, + 4.4823e-05, 1.6749e-05, -4.3303e-05, 1.3638e-04, 1.3039e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 220.47, cls_loss 0.0042 cls_loss_mapping 0.0111 cls_loss_causal 0.6112 re_mapping 0.0090 re_causal 0.0272 /// teacc 98.86 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.1132, -0.1052, 0.0684, ..., -0.0415, 0.0368, 0.0224], + [-0.0532, -0.0213, -0.0320, ..., -0.0472, -0.0717, -0.0381], + [ 0.0394, -0.0465, -0.0255, ..., -0.0335, -0.0010, -0.1917], + ..., + [-0.0745, 0.0693, 0.0024, ..., 0.0732, -0.0282, -0.0744], + [-0.0342, -0.0441, 0.0746, ..., -0.0396, -0.1205, 0.0785], + [-0.0064, -0.1063, 0.0678, ..., 0.0515, -0.1269, -0.0473]], + device='cuda:0'), grad: tensor([[-8.4098e-07, 1.5721e-05, -2.5943e-05, ..., 3.3174e-06, + -4.2990e-06, -2.4796e-05], + [-1.5569e-04, -3.0708e-03, -1.5659e-03, ..., -4.8518e-04, + 6.9756e-07, -1.0565e-05], + [ 1.6373e-06, 5.8556e-04, 1.8179e-05, ..., 1.8328e-05, + -4.0568e-06, 1.1146e-05], + ..., + [ 8.8274e-05, 8.5354e-04, 8.8406e-04, ..., 2.1017e-04, + 1.1111e-06, 9.1419e-06], + [-3.0901e-06, 1.0394e-05, -2.0549e-05, ..., -3.6545e-06, + 2.5742e-06, -2.8953e-05], + [ 6.4850e-05, 1.4420e-03, 6.4659e-04, ..., 2.3103e-04, + 2.3888e-07, 2.5004e-05]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0165, -0.0153, -0.0253, -0.0249, -0.0095, 0.0047, 0.0088, -0.0222, + -0.0089, 0.0054], device='cuda:0'), grad: tensor([-1.9699e-05, -8.4839e-03, 7.6628e-04, 2.4486e-04, 1.1927e-04, + 3.3647e-05, 1.3381e-05, 3.5477e-03, 1.2200e-06, 3.7804e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 220.65, cls_loss 0.0032 cls_loss_mapping 0.0105 cls_loss_causal 0.6311 re_mapping 0.0091 re_causal 0.0287 /// teacc 98.89 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.1150, -0.1059, 0.0685, ..., -0.0421, 0.0365, 0.0224], + [-0.0537, -0.0215, -0.0321, ..., -0.0477, -0.0719, -0.0380], + [ 0.0381, -0.0472, -0.0264, ..., -0.0341, -0.0010, -0.1939], + ..., + [-0.0756, 0.0704, 0.0025, ..., 0.0740, -0.0288, -0.0750], + [-0.0335, -0.0443, 0.0760, ..., -0.0390, -0.1198, 0.0802], + [-0.0059, -0.1076, 0.0679, ..., 0.0515, -0.1292, -0.0484]], + device='cuda:0'), grad: tensor([[ 3.9525e-06, 6.5565e-07, -2.9981e-05, ..., 1.6224e-06, + -5.3179e-07, -2.2411e-05], + [ 7.5996e-07, 1.3364e-07, 2.3358e-06, ..., 8.4490e-06, + -3.0827e-07, -4.8019e-06], + [-2.3887e-05, 3.8370e-06, 1.2465e-05, ..., 4.0121e-06, + 1.6275e-07, 1.0967e-05], + ..., + [ 9.1922e-07, -1.7017e-05, 4.9397e-06, ..., 2.4736e-06, + 2.9011e-07, 7.9051e-06], + [ 1.4894e-05, 1.0021e-06, -7.0371e-06, ..., 5.0366e-06, + 5.2154e-08, -1.0237e-05], + [ 7.9302e-07, 3.7700e-06, -5.5879e-06, ..., 1.5751e-05, + 7.8231e-08, 6.1505e-06]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0168, -0.0151, -0.0261, -0.0258, -0.0094, 0.0049, 0.0084, -0.0218, + -0.0071, 0.0048], device='cuda:0'), grad: tensor([-6.1035e-05, 1.6466e-05, -1.0848e-04, 3.3677e-05, -1.4806e-04, + 2.8566e-05, 5.1185e-06, 4.4912e-05, 8.6963e-05, 1.0163e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 220.60, cls_loss 0.0034 cls_loss_mapping 0.0094 cls_loss_causal 0.5551 re_mapping 0.0092 re_causal 0.0262 /// teacc 98.89 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.1156, -0.1067, 0.0689, ..., -0.0425, 0.0365, 0.0226], + [-0.0541, -0.0216, -0.0323, ..., -0.0480, -0.0716, -0.0379], + [ 0.0379, -0.0476, -0.0267, ..., -0.0345, -0.0008, -0.1944], + ..., + [-0.0766, 0.0708, 0.0025, ..., 0.0742, -0.0289, -0.0755], + [-0.0340, -0.0447, 0.0762, ..., -0.0394, -0.1200, 0.0804], + [-0.0057, -0.1083, 0.0688, ..., 0.0523, -0.1293, -0.0487]], + device='cuda:0'), grad: tensor([[ 1.6671e-07, 1.3607e-06, -1.2882e-05, ..., 9.5740e-07, + 2.9765e-06, 2.6412e-06], + [ 1.1385e-07, 2.5108e-06, -2.1402e-06, ..., 2.1085e-06, + 1.3621e-07, -7.6368e-06], + [ 1.4715e-07, 6.6273e-06, 3.7700e-06, ..., 2.5574e-06, + -5.6848e-06, 1.2018e-05], + ..., + [ 2.8731e-07, -2.7671e-05, -8.6874e-06, ..., -2.3350e-05, + 4.9919e-07, 3.4850e-06], + [ 1.0490e-05, 4.9844e-06, -5.9903e-06, ..., 9.3598e-07, + 1.7509e-07, 2.6435e-05], + [-1.1530e-06, 6.8247e-06, -4.7009e-07, ..., -1.3970e-07, + 1.4668e-07, 9.8944e-06]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0167, -0.0149, -0.0260, -0.0253, -0.0094, 0.0042, 0.0090, -0.0221, + -0.0075, 0.0050], device='cuda:0'), grad: tensor([ 6.5207e-05, -3.8326e-05, 4.5486e-06, 8.8587e-06, 5.0485e-05, + 1.4162e-04, -2.9874e-04, -2.6450e-05, 6.7890e-05, 2.4796e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 220.69, cls_loss 0.0029 cls_loss_mapping 0.0110 cls_loss_causal 0.5669 re_mapping 0.0090 re_causal 0.0272 /// teacc 98.92 lr 0.00010000 +Epoch 84, weight, value: tensor([[-1.1605e-01, -1.0887e-01, 6.9048e-02, ..., -4.3175e-02, + 3.6572e-02, 2.2682e-02], + [-5.4631e-02, -2.0724e-02, -3.2177e-02, ..., -4.8267e-02, + -7.1158e-02, -3.7393e-02], + [ 3.7713e-02, -4.9078e-02, -2.7272e-02, ..., -3.5733e-02, + 1.1970e-04, -1.9499e-01], + ..., + [-7.6891e-02, 7.1293e-02, 2.4593e-03, ..., 7.5045e-02, + -2.8202e-02, -7.6081e-02], + [-3.4615e-02, -4.5182e-02, 7.6742e-02, ..., -3.9170e-02, + -1.2020e-01, 8.0557e-02], + [-6.3753e-03, -1.0903e-01, 6.9221e-02, ..., 5.2737e-02, + -1.2944e-01, -4.9051e-02]], device='cuda:0'), grad: tensor([[ 5.1223e-09, 1.0128e-07, -5.8254e-07, ..., 4.7428e-07, + 1.8161e-08, -3.5460e-07], + [ 1.0943e-08, 7.7579e-07, 7.7020e-07, ..., 8.7358e-07, + 4.6566e-10, 9.3132e-08], + [ 2.3283e-09, 3.2373e-06, 1.5274e-06, ..., 1.1809e-06, + -2.4913e-08, 1.4529e-06], + ..., + [ 8.8941e-08, -2.2873e-06, 5.2201e-07, ..., -1.4026e-06, + 5.1223e-09, 7.0501e-07], + [ 4.7963e-08, -9.5461e-07, -1.1483e-06, ..., 2.3972e-06, + 2.0023e-08, -4.8578e-06], + [-2.8848e-07, 7.1619e-07, -5.2482e-05, ..., -4.5419e-05, + 2.5611e-09, 8.8615e-07]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0166, -0.0146, -0.0263, -0.0255, -0.0099, 0.0044, 0.0091, -0.0220, + -0.0076, 0.0052], device='cuda:0'), grad: tensor([ 3.3062e-06, 1.9115e-07, 3.2876e-06, 1.1688e-06, 1.0478e-04, + 9.1866e-06, -1.1154e-05, 1.9409e-06, 2.3842e-06, -1.1533e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 220.94, cls_loss 0.0027 cls_loss_mapping 0.0101 cls_loss_causal 0.5791 re_mapping 0.0091 re_causal 0.0271 /// teacc 98.86 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.1167, -0.1092, 0.0692, ..., -0.0437, 0.0365, 0.0232], + [-0.0547, -0.0210, -0.0323, ..., -0.0486, -0.0710, -0.0373], + [ 0.0379, -0.0496, -0.0278, ..., -0.0362, 0.0006, -0.1962], + ..., + [-0.0783, 0.0720, 0.0025, ..., 0.0755, -0.0284, -0.0767], + [-0.0345, -0.0454, 0.0777, ..., -0.0388, -0.1204, 0.0815], + [-0.0060, -0.1096, 0.0694, ..., 0.0524, -0.1298, -0.0499]], + device='cuda:0'), grad: tensor([[-1.0366e-06, 1.2212e-05, -3.5852e-05, ..., 8.6501e-06, + 2.3283e-10, -2.0161e-05], + [ 5.5833e-07, 4.6529e-06, 2.7753e-06, ..., 4.0866e-06, + 6.9849e-10, -4.3027e-06], + [ 1.3458e-06, 3.3770e-06, 1.5497e-05, ..., 3.7402e-06, + 6.9849e-10, 1.2554e-05], + ..., + [-3.7160e-06, -4.1574e-05, -1.4625e-05, ..., -1.7956e-05, + 1.3970e-09, -6.8620e-06], + [ 1.6466e-05, 1.1623e-06, 2.0787e-05, ..., 2.1607e-05, + 4.6566e-10, 1.0937e-05], + [-2.1815e-05, 4.1015e-06, -2.7984e-05, ..., -5.5432e-06, + 1.6298e-09, -1.2308e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0167, -0.0147, -0.0259, -0.0260, -0.0095, 0.0039, 0.0098, -0.0221, + -0.0069, 0.0046], device='cuda:0'), grad: tensor([-8.1837e-05, 5.9009e-06, 4.3511e-05, 5.8591e-05, -6.6638e-05, + 1.3880e-05, 4.1239e-06, -4.7386e-05, 5.7936e-05, 1.1690e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 221.02, cls_loss 0.0030 cls_loss_mapping 0.0090 cls_loss_causal 0.5993 re_mapping 0.0089 re_causal 0.0276 /// teacc 99.00 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.1174, -0.1091, 0.0698, ..., -0.0441, 0.0365, 0.0243], + [-0.0548, -0.0211, -0.0325, ..., -0.0490, -0.0709, -0.0372], + [ 0.0377, -0.0490, -0.0267, ..., -0.0368, 0.0032, -0.1969], + ..., + [-0.0772, 0.0727, 0.0029, ..., 0.0773, -0.0282, -0.0773], + [-0.0351, -0.0467, 0.0771, ..., -0.0394, -0.1234, 0.0819], + [-0.0053, -0.1113, 0.0701, ..., 0.0525, -0.1300, -0.0502]], + device='cuda:0'), grad: tensor([[ 1.5171e-06, 2.0233e-07, -4.9211e-06, ..., 1.3143e-05, + 4.9137e-06, -7.3388e-06], + [ 6.8126e-07, 1.4575e-07, 4.4033e-06, ..., 2.3600e-06, + 7.5810e-07, 1.6680e-06], + [-1.8124e-06, 2.0079e-06, -1.5453e-05, ..., -4.4435e-05, + -1.5065e-05, 2.9802e-06], + ..., + [ 7.9675e-07, -4.2617e-06, 8.5607e-06, ..., 1.9386e-05, + 9.5554e-07, 3.2559e-06], + [ 2.7735e-06, 3.1665e-07, 2.9281e-06, ..., 9.0525e-06, + 3.6750e-06, -6.9775e-06], + [-3.2876e-06, 1.6009e-06, -2.0579e-05, ..., -9.8944e-06, + 8.1398e-07, 1.9595e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0155, -0.0145, -0.0256, -0.0265, -0.0104, 0.0034, 0.0097, -0.0213, + -0.0073, 0.0046], device='cuda:0'), grad: tensor([ 4.3243e-05, 4.2133e-06, -1.8573e-04, 9.9540e-06, 6.0610e-06, + 9.6619e-05, -6.8367e-05, 7.0512e-05, 3.4630e-05, -1.1630e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 220.61, cls_loss 0.0032 cls_loss_mapping 0.0099 cls_loss_causal 0.5822 re_mapping 0.0090 re_causal 0.0263 /// teacc 98.91 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.1201, -0.1115, 0.0699, ..., -0.0444, 0.0363, 0.0239], + [-0.0552, -0.0207, -0.0326, ..., -0.0489, -0.0713, -0.0377], + [ 0.0368, -0.0493, -0.0271, ..., -0.0373, 0.0034, -0.1981], + ..., + [-0.0781, 0.0730, 0.0033, ..., 0.0780, -0.0276, -0.0780], + [-0.0350, -0.0471, 0.0777, ..., -0.0395, -0.1236, 0.0828], + [-0.0036, -0.1131, 0.0700, ..., 0.0524, -0.1303, -0.0505]], + device='cuda:0'), grad: tensor([[-3.3760e-08, 4.9686e-07, 1.5823e-06, ..., 6.5565e-07, + 4.0680e-06, 2.6166e-05], + [ 1.3388e-07, 3.0287e-06, 4.6372e-04, ..., 2.6915e-06, + 3.3691e-07, 1.1606e-03], + [ 7.0315e-08, 2.0768e-06, 3.7503e-04, ..., 1.0068e-06, + 5.6438e-07, 9.3889e-04], + ..., + [ 3.9442e-07, -2.7075e-05, 1.8165e-05, ..., -2.3142e-05, + 1.2876e-07, 6.4552e-05], + [ 1.1805e-07, 4.6217e-07, -9.0790e-04, ..., 6.3144e-07, + 3.9823e-06, -2.2640e-03], + [-7.2643e-07, 1.4991e-05, 4.8019e-06, ..., 7.5623e-06, + 6.6403e-07, 9.7156e-06]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0161, -0.0139, -0.0260, -0.0264, -0.0104, 0.0032, 0.0102, -0.0215, + -0.0069, 0.0045], device='cuda:0'), grad: tensor([ 7.2777e-05, 2.1763e-03, 1.7605e-03, 4.0680e-05, 8.7321e-06, + 7.3910e-05, -6.0111e-05, 8.4400e-05, -4.2000e-03, 4.3720e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 220.47, cls_loss 0.0029 cls_loss_mapping 0.0108 cls_loss_causal 0.5926 re_mapping 0.0082 re_causal 0.0251 /// teacc 98.97 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.1209, -0.1122, 0.0702, ..., -0.0446, 0.0363, 0.0235], + [-0.0557, -0.0203, -0.0332, ..., -0.0496, -0.0714, -0.0383], + [ 0.0364, -0.0498, -0.0279, ..., -0.0366, 0.0034, -0.2001], + ..., + [-0.0790, 0.0734, 0.0033, ..., 0.0784, -0.0268, -0.0788], + [-0.0355, -0.0473, 0.0783, ..., -0.0401, -0.1237, 0.0837], + [-0.0036, -0.1143, 0.0706, ..., 0.0524, -0.1307, -0.0511]], + device='cuda:0'), grad: tensor([[ 1.5949e-07, 1.0364e-05, -7.0818e-06, ..., 3.9581e-07, + 5.5647e-08, -8.9705e-06], + [ 1.0221e-07, 5.4911e-06, 2.4401e-06, ..., 3.3788e-06, + 1.3104e-06, 2.3525e-06], + [ 9.3598e-07, 9.4175e-05, 2.9653e-06, ..., 2.2724e-06, + 3.4668e-07, 4.8131e-06], + ..., + [ 4.9081e-07, 8.4639e-06, 2.7847e-07, ..., 1.8626e-08, + 1.0878e-06, 1.8394e-06], + [ 4.9062e-06, 1.8589e-06, -1.5214e-05, ..., -6.5267e-06, + 1.1083e-06, -1.3843e-05], + [ 5.5879e-07, 3.5577e-06, 6.4373e-06, ..., 2.6040e-06, + 3.1688e-07, 9.8869e-06]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0167, -0.0140, -0.0263, -0.0265, -0.0100, 0.0038, 0.0106, -0.0218, + -0.0064, 0.0040], device='cuda:0'), grad: tensor([ 7.9125e-06, 2.7239e-05, 2.7418e-04, -3.8600e-04, -2.4721e-05, + 4.4465e-05, -4.5355e-07, 3.5971e-05, -6.1765e-06, 2.6971e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 220.13, cls_loss 0.0029 cls_loss_mapping 0.0103 cls_loss_causal 0.6152 re_mapping 0.0084 re_causal 0.0263 /// teacc 98.98 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.1219, -0.1127, 0.0708, ..., -0.0450, 0.0364, 0.0236], + [-0.0564, -0.0209, -0.0336, ..., -0.0506, -0.0722, -0.0383], + [ 0.0363, -0.0505, -0.0281, ..., -0.0373, 0.0040, -0.2011], + ..., + [-0.0812, 0.0744, 0.0033, ..., 0.0791, -0.0269, -0.0793], + [-0.0365, -0.0477, 0.0785, ..., -0.0405, -0.1239, 0.0846], + [-0.0019, -0.1149, 0.0713, ..., 0.0528, -0.1313, -0.0512]], + device='cuda:0'), grad: tensor([[ 8.6846e-07, 1.7649e-06, -8.9556e-06, ..., 1.5087e-06, + -1.6382e-06, -1.1049e-05], + [ 3.0641e-07, 4.0904e-06, 3.3416e-06, ..., 2.9039e-06, + -3.4436e-07, -2.6566e-07], + [ 1.2647e-06, -2.6077e-05, 4.6641e-05, ..., -2.1160e-05, + -3.2480e-07, 4.2677e-05], + ..., + [ 1.1679e-06, 1.1019e-05, 2.7427e-07, ..., 1.9431e-05, + 4.4028e-07, 5.0180e-06], + [ 6.0052e-06, 3.6377e-06, -8.9586e-05, ..., -4.5836e-05, + 1.5064e-07, -8.2076e-05], + [-3.7309e-06, 9.2201e-07, 2.6584e-05, ..., 3.2216e-05, + 2.5611e-07, 4.8012e-05]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0166, -0.0148, -0.0258, -0.0265, -0.0101, 0.0036, 0.0100, -0.0213, + -0.0060, 0.0042], device='cuda:0'), grad: tensor([-9.0897e-06, 9.0376e-06, -1.2887e-04, 2.2635e-05, 7.3969e-05, + -9.9540e-06, -4.4674e-05, 1.7655e-04, -1.4770e-04, 5.7995e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 220.83, cls_loss 0.0035 cls_loss_mapping 0.0114 cls_loss_causal 0.5983 re_mapping 0.0087 re_causal 0.0249 /// teacc 98.83 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.1236, -0.1138, 0.0712, ..., -0.0454, 0.0359, 0.0236], + [-0.0578, -0.0208, -0.0340, ..., -0.0510, -0.0715, -0.0386], + [ 0.0367, -0.0500, -0.0284, ..., -0.0379, 0.0039, -0.2017], + ..., + [-0.0809, 0.0744, 0.0036, ..., 0.0802, -0.0271, -0.0802], + [-0.0382, -0.0479, 0.0792, ..., -0.0405, -0.1239, 0.0848], + [-0.0036, -0.1166, 0.0711, ..., 0.0525, -0.1328, -0.0530]], + device='cuda:0'), grad: tensor([[ 1.3423e-07, 4.8662e-07, -7.0818e-06, ..., 3.0966e-07, + -5.9837e-08, -6.3777e-06], + [ 1.1292e-07, 1.4510e-06, 1.2200e-06, ..., 1.3532e-06, + 4.6566e-09, -1.2256e-06], + [ 8.2538e-08, 1.4044e-06, 3.0659e-06, ..., 7.8417e-07, + -5.4482e-08, 4.2394e-06], + ..., + [ 5.7183e-06, -8.2701e-06, 3.0771e-06, ..., -1.2377e-06, + 1.5716e-08, 2.2501e-06], + [ 2.0731e-06, 3.7141e-06, -4.0792e-06, ..., 2.4401e-06, + 3.6671e-08, -4.2915e-06], + [-1.1295e-05, 4.5933e-06, -3.3807e-06, ..., -1.2033e-05, + 7.6834e-09, 3.0342e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0169, -0.0146, -0.0250, -0.0257, -0.0100, 0.0041, 0.0101, -0.0220, + -0.0063, 0.0033], device='cuda:0'), grad: tensor([-1.2539e-05, -2.7604e-06, 4.7237e-06, -1.0610e-05, 7.8976e-06, + 4.3422e-05, -3.4451e-05, 6.3181e-06, 4.1388e-06, -6.1728e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 220.52, cls_loss 0.0024 cls_loss_mapping 0.0062 cls_loss_causal 0.6230 re_mapping 0.0084 re_causal 0.0251 /// teacc 98.76 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.1243, -0.1148, 0.0716, ..., -0.0458, 0.0360, 0.0240], + [-0.0583, -0.0212, -0.0344, ..., -0.0515, -0.0712, -0.0387], + [ 0.0366, -0.0508, -0.0291, ..., -0.0386, 0.0041, -0.2028], + ..., + [-0.0813, 0.0751, 0.0036, ..., 0.0804, -0.0273, -0.0807], + [-0.0393, -0.0480, 0.0797, ..., -0.0411, -0.1241, 0.0851], + [-0.0043, -0.1172, 0.0715, ..., 0.0526, -0.1331, -0.0539]], + device='cuda:0'), grad: tensor([[ 7.0548e-08, 1.8150e-05, 5.6960e-06, ..., 1.3694e-05, + 2.9220e-08, 5.9744e-07], + [ 6.2166e-08, 5.5432e-05, 2.3142e-05, ..., 2.0489e-05, + -1.9907e-08, -1.0347e-06], + [ 4.6217e-08, 2.8923e-05, 9.2313e-06, ..., 1.4469e-05, + -1.4703e-07, 1.2843e-06], + ..., + [ 5.8487e-07, -1.6117e-04, -4.3929e-05, ..., 3.5167e-05, + 1.4005e-07, 1.2359e-06], + [ 3.6415e-07, 2.2724e-06, 2.0757e-05, ..., 2.1219e-05, + 6.1118e-08, -4.0941e-06], + [-3.6918e-06, 3.1799e-05, -5.8025e-05, ..., 4.0323e-05, + 7.4506e-09, 9.8534e-07]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0166, -0.0149, -0.0253, -0.0255, -0.0093, 0.0042, 0.0101, -0.0221, + -0.0063, 0.0030], device='cuda:0'), grad: tensor([ 4.6045e-05, 1.0115e-04, 5.1498e-05, 7.3195e-05, -5.2500e-04, + 2.9907e-05, 2.8014e-06, 1.9267e-05, 7.1824e-05, 1.2887e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 220.76, cls_loss 0.0024 cls_loss_mapping 0.0091 cls_loss_causal 0.5834 re_mapping 0.0085 re_causal 0.0252 /// teacc 98.78 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.1248, -0.1154, 0.0722, ..., -0.0463, 0.0360, 0.0247], + [-0.0591, -0.0214, -0.0348, ..., -0.0521, -0.0712, -0.0385], + [ 0.0362, -0.0515, -0.0299, ..., -0.0392, 0.0042, -0.2037], + ..., + [-0.0818, 0.0758, 0.0039, ..., 0.0810, -0.0274, -0.0810], + [-0.0396, -0.0483, 0.0804, ..., -0.0413, -0.1241, 0.0858], + [-0.0038, -0.1182, 0.0721, ..., 0.0529, -0.1335, -0.0544]], + device='cuda:0'), grad: tensor([[-9.4296e-08, 3.5018e-07, -6.7055e-06, ..., 2.6077e-08, + 6.4494e-08, 6.7115e-05], + [ 3.1525e-07, -1.3232e-04, -1.2022e-04, ..., -2.4334e-05, + -3.3667e-07, -2.0361e-04], + [-3.9376e-06, 6.2920e-06, 2.8554e-06, ..., 3.0585e-06, + -7.0315e-08, 3.1590e-05], + ..., + [ 3.3551e-07, 8.0884e-05, 7.7367e-05, ..., 1.1928e-05, + 9.9186e-08, 1.2982e-04], + [ 5.6531e-07, 1.4447e-05, 1.3553e-05, ..., 3.3323e-06, + 1.0361e-07, 2.0039e-04], + [ 2.3283e-07, 2.7820e-05, 2.1935e-05, ..., 2.8051e-06, + 2.4913e-08, 4.3392e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0147, -0.0154, -0.0257, -0.0258, -0.0094, 0.0041, 0.0093, -0.0218, + -0.0059, 0.0031], device='cuda:0'), grad: tensor([ 1.0765e-04, -5.0068e-04, 1.6108e-05, 2.0534e-05, -1.4174e-04, + 1.2231e-04, -5.4121e-04, 4.0603e-04, 3.5095e-04, 1.6022e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 220.52, cls_loss 0.0033 cls_loss_mapping 0.0102 cls_loss_causal 0.5862 re_mapping 0.0085 re_causal 0.0253 /// teacc 98.87 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.1253, -0.1170, 0.0728, ..., -0.0466, 0.0359, 0.0227], + [-0.0597, -0.0219, -0.0355, ..., -0.0526, -0.0714, -0.0371], + [ 0.0355, -0.0520, -0.0307, ..., -0.0388, 0.0041, -0.2055], + ..., + [-0.0805, 0.0772, 0.0048, ..., 0.0826, -0.0268, -0.0816], + [-0.0398, -0.0488, 0.0815, ..., -0.0412, -0.1242, 0.0863], + [-0.0029, -0.1205, 0.0717, ..., 0.0523, -0.1339, -0.0555]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 1.7593e-06, -1.0960e-05, ..., 9.4529e-07, + 1.1642e-09, -2.8014e-06], + [ 4.4005e-08, -3.6001e-05, 2.4643e-06, ..., 1.8105e-06, + 3.4925e-09, -1.6659e-05], + [ 4.4471e-08, 4.4197e-05, 7.8455e-06, ..., 2.8089e-06, + 3.4925e-09, 1.6212e-05], + ..., + [-7.7672e-07, -1.3538e-05, -1.2070e-06, ..., -9.6262e-06, + 1.6764e-08, 1.7164e-06], + [ 4.5798e-07, 1.2390e-05, 4.7870e-06, ..., 1.7304e-06, + 6.9849e-10, -4.6194e-06], + [ 3.7532e-07, 7.3016e-06, -9.7528e-06, ..., -5.1260e-06, + 1.3970e-09, 1.7444e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0160, -0.0150, -0.0259, -0.0263, -0.0095, 0.0038, 0.0108, -0.0208, + -0.0059, 0.0022], device='cuda:0'), grad: tensor([-1.7121e-05, -5.1022e-04, 4.6182e-04, -1.7539e-05, 1.5814e-06, + 1.5885e-05, 6.4485e-06, 3.2693e-05, 3.3200e-05, -7.1153e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 220.44, cls_loss 0.0034 cls_loss_mapping 0.0092 cls_loss_causal 0.6051 re_mapping 0.0083 re_causal 0.0247 /// teacc 98.78 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.1259, -0.1175, 0.0730, ..., -0.0471, 0.0359, 0.0226], + [-0.0603, -0.0218, -0.0362, ..., -0.0530, -0.0712, -0.0373], + [ 0.0362, -0.0525, -0.0309, ..., -0.0392, 0.0043, -0.2061], + ..., + [-0.0814, 0.0775, 0.0046, ..., 0.0829, -0.0269, -0.0820], + [-0.0401, -0.0491, 0.0822, ..., -0.0415, -0.1243, 0.0871], + [-0.0013, -0.1215, 0.0733, ..., 0.0533, -0.1343, -0.0557]], + device='cuda:0'), grad: tensor([[ 1.5832e-06, 2.6412e-06, 5.4799e-06, ..., 4.5151e-06, + 7.2177e-09, 3.7141e-06], + [ 2.6897e-06, 2.6841e-06, 5.0329e-06, ..., 7.6741e-06, + 3.7253e-09, 4.3795e-07], + [ 5.2992e-07, 4.4256e-06, 5.4426e-06, ..., 6.8471e-06, + -5.5879e-08, 2.3469e-06], + ..., + [ 4.4778e-06, -5.5581e-05, -6.7592e-05, ..., -8.7857e-05, + 3.1199e-08, -2.3663e-05], + [ 1.2226e-05, 1.0140e-05, 1.2942e-05, ..., 2.2337e-05, + 8.8476e-09, 4.5868e-07], + [ 1.3806e-05, 2.2173e-05, -7.3195e-05, ..., 6.7592e-05, + 6.9849e-10, -9.1866e-06]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0160, -0.0156, -0.0252, -0.0260, -0.0105, 0.0031, 0.0101, -0.0212, + -0.0057, 0.0040], device='cuda:0'), grad: tensor([ 1.9506e-05, 3.5286e-05, 2.0623e-05, 7.3135e-05, -2.2054e-04, + 8.5354e-05, 2.5854e-05, -1.3101e-04, 6.1512e-05, 3.0428e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 220.77, cls_loss 0.0027 cls_loss_mapping 0.0080 cls_loss_causal 0.5983 re_mapping 0.0082 re_causal 0.0238 /// teacc 98.98 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.1266, -0.1188, 0.0727, ..., -0.0479, 0.0358, 0.0223], + [-0.0603, -0.0230, -0.0365, ..., -0.0538, -0.0716, -0.0368], + [ 0.0365, -0.0533, -0.0311, ..., -0.0398, 0.0045, -0.2069], + ..., + [-0.0822, 0.0791, 0.0048, ..., 0.0837, -0.0270, -0.0824], + [-0.0406, -0.0497, 0.0825, ..., -0.0421, -0.1245, 0.0875], + [-0.0013, -0.1227, 0.0746, ..., 0.0539, -0.1347, -0.0561]], + device='cuda:0'), grad: tensor([[ 1.1735e-07, 7.2457e-07, -1.0859e-06, ..., 6.4541e-07, + 2.6077e-08, 3.9488e-07], + [ 6.5193e-08, 9.3356e-06, 8.4424e-07, ..., 6.4857e-06, + -1.8114e-07, -2.2184e-06], + [ 3.6974e-07, 1.3590e-05, 1.2644e-05, ..., 9.4920e-06, + -4.4331e-07, 1.4879e-05], + ..., + [ 1.6880e-07, -2.3127e-05, -5.8021e-07, ..., -1.8939e-05, + 3.1036e-07, -3.0920e-07], + [-2.6692e-06, 1.5981e-06, -3.4988e-05, ..., -1.8533e-06, + 5.4250e-08, -5.2482e-05], + [ 1.0780e-07, 1.7602e-06, 6.5193e-06, ..., 5.3365e-07, + 2.6310e-08, 1.5393e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0167, -0.0160, -0.0249, -0.0260, -0.0109, 0.0027, 0.0105, -0.0206, + -0.0058, 0.0043], device='cuda:0'), grad: tensor([ 2.1625e-06, 7.1079e-06, 4.4078e-05, -9.9540e-06, 2.1383e-06, + 2.2113e-05, 1.4469e-05, -2.6688e-05, -7.7248e-05, 2.1830e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 220.64, cls_loss 0.0037 cls_loss_mapping 0.0085 cls_loss_causal 0.5913 re_mapping 0.0077 re_causal 0.0233 /// teacc 98.89 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.1273, -0.1200, 0.0718, ..., -0.0497, 0.0366, 0.0224], + [-0.0624, -0.0230, -0.0367, ..., -0.0541, -0.0717, -0.0369], + [ 0.0372, -0.0542, -0.0316, ..., -0.0404, 0.0047, -0.2077], + ..., + [-0.0826, 0.0798, 0.0043, ..., 0.0839, -0.0269, -0.0836], + [-0.0412, -0.0503, 0.0826, ..., -0.0425, -0.1247, 0.0872], + [-0.0007, -0.1244, 0.0765, ..., 0.0546, -0.1352, -0.0558]], + device='cuda:0'), grad: tensor([[ 3.9279e-07, 9.1270e-07, -4.8503e-06, ..., 9.5181e-07, + 3.0268e-09, -3.2503e-06], + [ 3.3737e-07, 4.6380e-06, 1.2619e-06, ..., 3.9414e-06, + 4.1211e-08, 5.0385e-07], + [ 7.0967e-07, 5.7817e-06, 1.7863e-06, ..., 2.5146e-06, + -7.1479e-08, 2.1271e-06], + ..., + [ 4.6683e-07, -2.5794e-05, -5.7183e-06, ..., -4.7237e-05, + 1.3039e-08, 1.8086e-06], + [ 1.4395e-05, 4.3958e-06, -5.5507e-07, ..., 9.9372e-07, + 6.7521e-09, 1.6913e-05], + [ 9.3952e-06, 1.6674e-05, 2.2706e-06, ..., 1.7136e-05, + 1.3970e-09, 1.4402e-05]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0173, -0.0156, -0.0252, -0.0241, -0.0109, 0.0010, 0.0113, -0.0213, + -0.0067, 0.0052], device='cuda:0'), grad: tensor([ 4.5309e-07, 1.3232e-05, -3.4600e-05, 4.0740e-05, 3.0249e-05, + -8.5235e-05, -5.1968e-06, -4.1544e-05, 2.8446e-05, 5.3346e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 220.42, cls_loss 0.0026 cls_loss_mapping 0.0079 cls_loss_causal 0.5762 re_mapping 0.0079 re_causal 0.0240 /// teacc 98.91 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.1276, -0.1210, 0.0721, ..., -0.0499, 0.0365, 0.0227], + [-0.0636, -0.0231, -0.0373, ..., -0.0547, -0.0717, -0.0368], + [ 0.0369, -0.0544, -0.0319, ..., -0.0409, 0.0048, -0.2085], + ..., + [-0.0827, 0.0801, 0.0037, ..., 0.0841, -0.0269, -0.0846], + [-0.0416, -0.0507, 0.0830, ..., -0.0422, -0.1248, 0.0877], + [-0.0008, -0.1254, 0.0772, ..., 0.0549, -0.1353, -0.0560]], + device='cuda:0'), grad: tensor([[ 1.6931e-06, 5.0198e-07, -1.4892e-06, ..., 1.0980e-06, + 0.0000e+00, 2.3004e-06], + [ 5.9232e-07, 1.9539e-06, 1.6242e-06, ..., 1.5171e-06, + 0.0000e+00, 1.0151e-06], + [ 2.0582e-06, 1.0923e-05, 8.2776e-06, ..., 8.1509e-06, + 0.0000e+00, 1.7546e-06], + ..., + [ 7.4394e-06, -1.6600e-05, -6.9588e-06, ..., -1.0639e-05, + 0.0000e+00, 1.7568e-05], + [ 3.9876e-05, 6.4820e-06, 4.4674e-05, ..., 2.6584e-05, + 0.0000e+00, 3.9607e-05], + [-5.9038e-05, 2.6319e-06, -9.1732e-05, ..., -5.0157e-05, + 0.0000e+00, -1.6615e-05]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0172, -0.0158, -0.0251, -0.0245, -0.0106, 0.0022, 0.0106, -0.0220, + -0.0066, 0.0054], device='cuda:0'), grad: tensor([ 4.3213e-06, 8.2701e-06, 2.7850e-05, 4.6104e-05, 4.4256e-05, + -1.4156e-05, -6.8605e-05, 6.1095e-06, 1.4162e-04, -1.9574e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 220.47, cls_loss 0.0027 cls_loss_mapping 0.0080 cls_loss_causal 0.5928 re_mapping 0.0080 re_causal 0.0239 /// teacc 98.83 lr 0.00010000 +Epoch 98, weight, value: tensor([[-1.2902e-01, -1.2323e-01, 7.1964e-02, ..., -5.0926e-02, + 3.6483e-02, 2.2723e-02], + [-6.5255e-02, -2.3615e-02, -3.7695e-02, ..., -5.5340e-02, + -7.1735e-02, -3.7051e-02], + [ 3.6466e-02, -5.5314e-02, -3.2395e-02, ..., -4.1621e-02, + 4.8801e-03, -2.0956e-01], + ..., + [-8.3021e-02, 8.1728e-02, 4.1624e-03, ..., 8.5183e-02, + -2.6903e-02, -8.5333e-02], + [-4.1732e-02, -5.1501e-02, 8.3464e-02, ..., -4.2543e-02, + -1.2482e-01, 8.8447e-02], + [-1.8180e-04, -1.2707e-01, 7.7528e-02, ..., 5.4682e-02, + -1.3581e-01, -5.6596e-02]], device='cuda:0'), grad: tensor([[ 1.7099e-06, 4.5518e-07, -8.6650e-06, ..., 2.0899e-06, + 2.3283e-10, -3.5316e-06], + [ 7.2876e-07, 8.6101e-07, 1.8170e-06, ..., 3.6061e-06, + 4.6566e-10, 2.5192e-07], + [ 4.2804e-06, 1.2591e-06, 7.4469e-06, ..., 4.8801e-06, + 2.0955e-09, 3.0901e-06], + ..., + [ 3.5286e-05, -1.0263e-06, 4.8727e-05, ..., 4.1187e-05, + 4.6566e-09, 1.9342e-05], + [ 1.8273e-06, 3.1441e-06, -1.3955e-05, ..., 3.5446e-06, + 4.6566e-10, -1.2614e-05], + [-5.5999e-05, 1.7453e-06, -6.1870e-05, ..., -1.2249e-05, + 9.3132e-10, -1.9819e-05]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0173, -0.0163, -0.0256, -0.0250, -0.0104, 0.0022, 0.0108, -0.0210, + -0.0063, 0.0050], device='cuda:0'), grad: tensor([-7.8008e-06, 9.8869e-06, 2.3887e-05, 2.5313e-06, -1.3268e-04, + 1.7166e-05, 5.5172e-06, 1.3983e-04, -1.0185e-05, -4.8161e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 220.78, cls_loss 0.0025 cls_loss_mapping 0.0076 cls_loss_causal 0.5469 re_mapping 0.0079 re_causal 0.0223 /// teacc 98.95 lr 0.00010000 +Epoch 99, weight, value: tensor([[-1.2927e-01, -1.2396e-01, 7.2616e-02, ..., -5.1338e-02, + 3.6392e-02, 2.2933e-02], + [-6.5970e-02, -2.4011e-02, -3.8223e-02, ..., -5.6125e-02, + -7.1555e-02, -3.7248e-02], + [ 3.6582e-02, -5.5851e-02, -3.2798e-02, ..., -4.2062e-02, + 5.0595e-03, -2.1055e-01], + ..., + [-8.4358e-02, 8.2614e-02, 4.2962e-03, ..., 8.5965e-02, + -2.6940e-02, -8.5672e-02], + [-4.2361e-02, -5.1977e-02, 8.4030e-02, ..., -4.2685e-02, + -1.2489e-01, 8.9035e-02], + [ 9.0722e-05, -1.2806e-01, 7.7720e-02, ..., 5.4486e-02, + -1.3632e-01, -5.7351e-02]], device='cuda:0'), grad: tensor([[ 7.6648e-07, 1.9674e-07, -3.6731e-06, ..., 9.8487e-08, + 1.1898e-07, -1.1930e-06], + [ 5.0198e-07, 7.4599e-07, 6.4261e-07, ..., 1.8789e-07, + 1.2596e-07, -3.0408e-07], + [-6.3956e-05, -1.1243e-05, 8.3726e-07, ..., -1.3476e-06, + -1.3784e-06, 1.7984e-06], + ..., + [ 6.6264e-07, 4.3698e-06, 5.4296e-07, ..., 1.3616e-06, + 2.4471e-07, 9.3877e-07], + [ 1.1018e-06, 1.9856e-06, 1.2806e-06, ..., 5.5972e-07, + 5.3830e-07, 8.4341e-06], + [ 6.2346e-05, 6.3591e-06, 8.4378e-07, ..., -9.7509e-07, + 2.2585e-08, 3.7104e-06]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0172, -0.0164, -0.0257, -0.0251, -0.0102, 0.0023, 0.0107, -0.0208, + -0.0060, 0.0047], device='cuda:0'), grad: tensor([ 4.9826e-08, 4.0904e-06, -3.4523e-04, 3.0577e-05, 3.5204e-06, + -9.1195e-05, 4.9025e-05, 2.1890e-05, 2.9325e-05, 2.9826e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 221.06, cls_loss 0.0027 cls_loss_mapping 0.0059 cls_loss_causal 0.5309 re_mapping 0.0075 re_causal 0.0221 /// teacc 98.88 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.1297, -0.1245, 0.0729, ..., -0.0516, 0.0367, 0.0231], + [-0.0667, -0.0243, -0.0388, ..., -0.0565, -0.0719, -0.0367], + [ 0.0366, -0.0562, -0.0330, ..., -0.0419, 0.0055, -0.2112], + ..., + [-0.0848, 0.0833, 0.0046, ..., 0.0864, -0.0259, -0.0862], + [-0.0432, -0.0528, 0.0843, ..., -0.0430, -0.1251, 0.0892], + [-0.0024, -0.1292, 0.0778, ..., 0.0533, -0.1371, -0.0586]], + device='cuda:0'), grad: tensor([[ 2.5821e-07, 5.2666e-07, -1.5604e-04, ..., 1.4110e-07, + -2.4736e-06, -1.6475e-04], + [ 6.9849e-08, 1.3849e-06, 2.7213e-06, ..., 2.4028e-07, + 6.1700e-08, -6.1020e-06], + [ 5.3318e-08, 4.8056e-07, 7.6517e-06, ..., 1.6298e-07, + 4.3050e-07, 1.4357e-05], + ..., + [ 3.3784e-07, 7.9069e-07, 3.8184e-06, ..., 8.5495e-07, + 2.1094e-07, 4.1425e-06], + [ 1.7462e-06, 4.1723e-06, -2.9773e-05, ..., 5.8580e-07, + 1.9185e-07, -4.8429e-05], + [-4.2818e-07, 2.0433e-06, 1.1936e-05, ..., -4.5970e-06, + 5.4482e-07, 1.9372e-05]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0172, -0.0161, -0.0255, -0.0248, -0.0087, 0.0030, 0.0097, -0.0207, + -0.0064, 0.0031], device='cuda:0'), grad: tensor([-3.7503e-04, -3.2872e-05, 2.2814e-05, 2.9393e-06, 1.2517e-05, + 9.8571e-06, 3.7479e-04, 1.5616e-05, -5.6922e-05, 2.6435e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 220.53, cls_loss 0.0023 cls_loss_mapping 0.0060 cls_loss_causal 0.5899 re_mapping 0.0072 re_causal 0.0225 /// teacc 98.98 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.1291, -0.1247, 0.0739, ..., -0.0519, 0.0368, 0.0237], + [-0.0690, -0.0246, -0.0392, ..., -0.0569, -0.0719, -0.0367], + [ 0.0361, -0.0566, -0.0335, ..., -0.0424, 0.0056, -0.2124], + ..., + [-0.0856, 0.0837, 0.0046, ..., 0.0869, -0.0259, -0.0868], + [-0.0439, -0.0532, 0.0847, ..., -0.0432, -0.1251, 0.0899], + [-0.0015, -0.1301, 0.0783, ..., 0.0537, -0.1375, -0.0587]], + device='cuda:0'), grad: tensor([[ 3.1702e-06, 3.5405e-05, -8.6240e-07, ..., 8.7321e-06, + 6.9849e-10, 3.7737e-06], + [ 2.3637e-06, -7.8797e-05, 7.0315e-07, ..., 6.2399e-06, + 4.6566e-10, -1.4752e-05], + [ 8.0280e-07, 1.0036e-05, 4.5076e-06, ..., 1.7593e-06, + -7.6834e-09, 1.6436e-05], + ..., + [-8.7693e-06, -5.9515e-05, 1.9595e-06, ..., -3.5614e-05, + 5.3551e-09, 1.2875e-05], + [ 1.2033e-05, 5.2787e-06, 2.1420e-06, ..., 1.1828e-06, + 9.3132e-10, 2.2709e-05], + [-3.9220e-05, 1.4715e-05, -1.9550e-05, ..., -3.4750e-05, + 0.0000e+00, -1.7151e-05]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0166, -0.0163, -0.0255, -0.0250, -0.0091, 0.0027, 0.0101, -0.0208, + -0.0062, 0.0035], device='cuda:0'), grad: tensor([ 6.2048e-05, -1.9860e-04, 9.4473e-05, 4.3929e-05, 4.7565e-05, + 7.1526e-05, -1.8048e-04, 3.7283e-05, 6.6042e-05, -4.3392e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 220.87, cls_loss 0.0027 cls_loss_mapping 0.0089 cls_loss_causal 0.5637 re_mapping 0.0078 re_causal 0.0227 /// teacc 98.88 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.1293, -0.1255, 0.0745, ..., -0.0523, 0.0368, 0.0246], + [-0.0698, -0.0251, -0.0397, ..., -0.0577, -0.0722, -0.0367], + [ 0.0361, -0.0569, -0.0339, ..., -0.0423, 0.0057, -0.2128], + ..., + [-0.0878, 0.0847, 0.0047, ..., 0.0869, -0.0253, -0.0869], + [-0.0454, -0.0541, 0.0847, ..., -0.0439, -0.1255, 0.0910], + [-0.0011, -0.1309, 0.0790, ..., 0.0539, -0.1376, -0.0590]], + device='cuda:0'), grad: tensor([[-2.6617e-06, 1.8580e-07, -6.1467e-06, ..., 2.1397e-07, + 3.9022e-07, -1.8954e-05], + [ 4.9826e-07, 1.7639e-06, 5.5553e-07, ..., 5.4808e-07, + -2.7986e-07, 4.2804e-06], + [ 1.4622e-06, 9.7692e-05, 2.0087e-05, ..., 1.0079e-04, + -4.2818e-07, -4.0643e-06], + ..., + [ 8.4657e-07, -1.0109e-04, -1.7703e-05, ..., -1.0222e-04, + 2.1607e-07, 6.6943e-06], + [ 1.2163e-06, 2.4238e-07, 2.0787e-06, ..., 1.4603e-06, + 2.7195e-07, 5.5283e-06], + [ 5.7369e-06, 1.8142e-06, -2.7250e-06, ..., -1.8254e-06, + 3.0035e-08, 5.3234e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0153, -0.0167, -0.0249, -0.0256, -0.0088, 0.0027, 0.0087, -0.0208, + -0.0058, 0.0034], device='cuda:0'), grad: tensor([-3.1769e-05, 3.6716e-05, 1.6963e-04, 8.7693e-06, 2.0135e-06, + 1.1154e-05, -7.6666e-06, -2.1636e-04, 1.5870e-05, 1.1511e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 220.33, cls_loss 0.0028 cls_loss_mapping 0.0085 cls_loss_causal 0.5693 re_mapping 0.0075 re_causal 0.0222 /// teacc 98.86 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.1311, -0.1249, 0.0747, ..., -0.0527, 0.0361, 0.0244], + [-0.0710, -0.0256, -0.0403, ..., -0.0587, -0.0727, -0.0365], + [ 0.0356, -0.0576, -0.0343, ..., -0.0428, 0.0057, -0.2135], + ..., + [-0.0900, 0.0850, 0.0034, ..., 0.0870, -0.0246, -0.0897], + [-0.0463, -0.0533, 0.0864, ..., -0.0425, -0.1254, 0.0912], + [-0.0005, -0.1321, 0.0796, ..., 0.0542, -0.1385, -0.0594]], + device='cuda:0'), grad: tensor([[ 2.1383e-06, 1.0803e-06, 3.4254e-06, ..., 3.3919e-06, + 4.8894e-08, 2.6003e-06], + [ 1.0189e-06, 3.1926e-06, 2.3786e-06, ..., 4.2766e-06, + -3.2596e-08, -6.7689e-06], + [ 8.5542e-07, 3.0156e-06, 2.9523e-06, ..., 3.3360e-06, + 4.8894e-08, 3.6284e-06], + ..., + [ 2.0433e-06, -2.1309e-05, -7.4394e-06, ..., -1.8463e-05, + 1.1409e-08, -1.6075e-06], + [ 1.4767e-05, 2.1998e-06, 2.7686e-05, ..., 1.8761e-05, + 1.0012e-08, 9.0301e-06], + [ 1.3137e-04, 7.9051e-06, -5.6624e-05, ..., 3.9244e-04, + 6.0536e-09, -1.8954e-05]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0157, -0.0170, -0.0249, -0.0258, -0.0087, 0.0022, 0.0103, -0.0218, + -0.0054, 0.0037], device='cuda:0'), grad: tensor([ 1.2659e-05, -1.2286e-05, 1.5959e-05, 3.8296e-05, -9.6655e-04, + 7.0706e-06, 1.5870e-06, -2.8253e-05, 6.3777e-05, 8.6880e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 220.72, cls_loss 0.0025 cls_loss_mapping 0.0070 cls_loss_causal 0.5962 re_mapping 0.0073 re_causal 0.0225 /// teacc 98.85 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.1334, -0.1256, 0.0752, ..., -0.0538, 0.0365, 0.0242], + [-0.0717, -0.0259, -0.0407, ..., -0.0593, -0.0728, -0.0363], + [ 0.0345, -0.0584, -0.0349, ..., -0.0435, 0.0059, -0.2148], + ..., + [-0.0906, 0.0859, 0.0034, ..., 0.0876, -0.0247, -0.0902], + [-0.0478, -0.0538, 0.0864, ..., -0.0428, -0.1254, 0.0911], + [-0.0004, -0.1333, 0.0804, ..., 0.0541, -0.1389, -0.0594]], + device='cuda:0'), grad: tensor([[ 5.4110e-07, 3.0305e-06, -2.1188e-08, ..., 3.2485e-06, + -1.1222e-07, 5.2247e-07], + [ 5.7369e-07, 6.5304e-06, 6.6347e-06, ..., 7.2159e-06, + 3.8417e-08, 2.3507e-06], + [ 1.3807e-07, 1.0885e-05, 1.0714e-05, ..., 1.1757e-05, + 6.5193e-08, 4.4256e-06], + ..., + [ 5.4762e-07, -6.7234e-05, -6.3300e-05, ..., -7.5340e-05, + 1.3271e-08, -1.8835e-05], + [ 7.0333e-06, 1.4901e-05, 1.1653e-05, ..., 1.2338e-05, + 2.2352e-08, 1.0997e-05], + [ 8.1817e-07, 3.3945e-05, 3.1859e-05, ..., 3.6329e-05, + 7.7067e-08, 1.1928e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0162, -0.0171, -0.0252, -0.0261, -0.0082, 0.0033, 0.0102, -0.0218, + -0.0059, 0.0036], device='cuda:0'), grad: tensor([ 5.8375e-06, 1.6510e-05, 3.0547e-05, -2.7884e-06, 1.5870e-05, + -1.3076e-05, -1.9103e-05, -1.6904e-04, 4.6581e-05, 8.8573e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 220.80, cls_loss 0.0027 cls_loss_mapping 0.0076 cls_loss_causal 0.5501 re_mapping 0.0077 re_causal 0.0210 /// teacc 98.92 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.1337, -0.1268, 0.0750, ..., -0.0559, 0.0380, 0.0256], + [-0.0717, -0.0271, -0.0412, ..., -0.0605, -0.0728, -0.0363], + [ 0.0341, -0.0592, -0.0356, ..., -0.0445, 0.0059, -0.2156], + ..., + [-0.0909, 0.0876, 0.0033, ..., 0.0884, -0.0247, -0.0908], + [-0.0484, -0.0546, 0.0866, ..., -0.0432, -0.1256, 0.0915], + [-0.0003, -0.1346, 0.0818, ..., 0.0548, -0.1399, -0.0597]], + device='cuda:0'), grad: tensor([[ 7.9582e-07, 1.3532e-06, -1.3387e-04, ..., 2.6990e-06, + -9.0659e-05, -2.0897e-04], + [ 9.0718e-05, 4.3698e-06, 4.3184e-05, ..., 7.6219e-06, + 1.3886e-06, 3.5852e-05], + [ 1.0908e-05, -1.4789e-05, 1.6734e-05, ..., -4.9114e-05, + 6.0536e-06, 2.2054e-05], + ..., + [ 2.3358e-06, 7.6322e-07, 3.6359e-06, ..., 2.0489e-05, + 2.5835e-06, 8.1137e-06], + [ 1.9232e-07, 9.6206e-07, -9.3654e-06, ..., 1.7751e-06, + 4.5337e-06, -3.4094e-05], + [-1.4532e-04, 4.5113e-06, -6.1095e-05, ..., 4.9546e-06, + 2.8592e-06, -3.9846e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0154, -0.0181, -0.0251, -0.0265, -0.0083, 0.0030, 0.0098, -0.0212, + -0.0059, 0.0041], device='cuda:0'), grad: tensor([-5.8270e-04, 4.1127e-04, -1.3196e-04, 1.8656e-04, 1.1760e-04, + 3.9339e-04, 2.8729e-05, 1.4770e-04, -2.7940e-05, -5.4264e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 220.28, cls_loss 0.0037 cls_loss_mapping 0.0084 cls_loss_causal 0.5590 re_mapping 0.0073 re_causal 0.0208 /// teacc 98.98 lr 0.00010000 +Epoch 106, weight, value: tensor([[-1.3398e-01, -1.2755e-01, 7.5774e-02, ..., -5.6171e-02, + 3.8864e-02, 2.6090e-02], + [-7.2445e-02, -2.8048e-02, -4.2173e-02, ..., -5.8669e-02, + -7.2861e-02, -3.6340e-02], + [ 3.3669e-02, -6.1068e-02, -3.6191e-02, ..., -4.6640e-02, + 5.8702e-03, -2.1739e-01], + ..., + [-9.1139e-02, 8.9187e-02, 3.6625e-03, ..., 8.8551e-02, + -2.4946e-02, -9.1034e-02], + [-4.9131e-02, -5.5057e-02, 8.6511e-02, ..., -4.3750e-02, + -1.2588e-01, 9.2437e-02], + [-8.9680e-06, -1.3590e-01, 8.2141e-02, ..., 5.4388e-02, + -1.4077e-01, -6.0066e-02]], device='cuda:0'), grad: tensor([[ 6.5658e-08, 1.3877e-06, 1.0636e-06, ..., 1.4761e-06, + 2.0210e-07, 1.6531e-06], + [ 1.8859e-08, 7.0259e-06, 4.1611e-06, ..., 5.5432e-06, + -9.1828e-07, -1.7434e-05], + [ 2.0722e-08, 9.1493e-06, 4.2319e-06, ..., 3.5800e-06, + -1.1390e-06, 2.1253e-06], + ..., + [ 1.1339e-07, -5.7131e-05, -2.9117e-05, ..., -3.2395e-05, + 2.3656e-07, -9.5740e-07], + [ 4.1421e-07, 1.9073e-06, 3.7551e-05, ..., 2.1830e-05, + 8.2981e-07, 2.6926e-05], + [-5.0664e-07, 1.1228e-05, -6.6161e-05, ..., -2.3931e-05, + 4.8894e-08, -3.4034e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0150, -0.0183, -0.0267, -0.0272, -0.0077, 0.0031, 0.0101, -0.0203, + -0.0051, 0.0028], device='cuda:0'), grad: tensor([ 8.2701e-06, -1.7747e-05, 1.1280e-05, 4.1336e-05, 2.5123e-05, + 1.1891e-05, 9.8161e-07, -9.0599e-05, 7.8738e-05, -6.9439e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 220.72, cls_loss 0.0020 cls_loss_mapping 0.0057 cls_loss_causal 0.5607 re_mapping 0.0075 re_causal 0.0216 /// teacc 98.87 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.1342, -0.1283, 0.0762, ..., -0.0564, 0.0403, 0.0264], + [-0.0730, -0.0282, -0.0425, ..., -0.0589, -0.0733, -0.0364], + [ 0.0334, -0.0617, -0.0366, ..., -0.0471, 0.0060, -0.2175], + ..., + [-0.0920, 0.0896, 0.0037, ..., 0.0887, -0.0250, -0.0915], + [-0.0499, -0.0557, 0.0864, ..., -0.0443, -0.1260, 0.0924], + [ 0.0003, -0.1364, 0.0830, ..., 0.0549, -0.1414, -0.0601]], + device='cuda:0'), grad: tensor([[ 1.9986e-06, 6.0303e-08, 6.9616e-07, ..., 7.0874e-07, + 2.3283e-10, 3.3416e-06], + [ 1.1995e-06, 1.5181e-07, 1.0720e-06, ..., 1.7146e-06, + 0.0000e+00, -2.2296e-06], + [ 1.1260e-06, -8.6380e-07, 5.5879e-07, ..., 5.5134e-07, + 0.0000e+00, 2.3227e-06], + ..., + [ 6.8471e-06, 5.2061e-07, 3.8110e-06, ..., 5.4911e-06, + 0.0000e+00, 6.4075e-06], + [-2.1644e-06, 1.2643e-07, -2.0146e-05, ..., 1.5777e-06, + 2.3283e-10, -5.7995e-05], + [-2.9225e-06, 1.5553e-07, -1.6838e-05, ..., 1.8239e-05, + 0.0000e+00, 3.4291e-06]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0148, -0.0186, -0.0261, -0.0268, -0.0080, 0.0032, 0.0101, -0.0203, + -0.0059, 0.0031], device='cuda:0'), grad: tensor([ 1.0282e-05, -1.7080e-06, -7.6443e-06, 2.3961e-05, -8.6904e-05, + 2.2590e-05, 2.9862e-05, 3.4809e-05, -7.7605e-05, 5.2303e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 220.60, cls_loss 0.0024 cls_loss_mapping 0.0074 cls_loss_causal 0.5928 re_mapping 0.0072 re_causal 0.0220 /// teacc 98.80 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.1346, -0.1290, 0.0763, ..., -0.0567, 0.0405, 0.0262], + [-0.0735, -0.0275, -0.0426, ..., -0.0593, -0.0739, -0.0354], + [ 0.0331, -0.0622, -0.0370, ..., -0.0478, 0.0061, -0.2185], + ..., + [-0.0927, 0.0894, 0.0037, ..., 0.0892, -0.0239, -0.0917], + [-0.0510, -0.0565, 0.0866, ..., -0.0453, -0.1268, 0.0923], + [ 0.0005, -0.1371, 0.0836, ..., 0.0554, -0.1416, -0.0600]], + device='cuda:0'), grad: tensor([[ 1.6456e-06, 2.0210e-06, 6.1393e-05, ..., 1.6287e-05, + 2.3283e-10, 5.7518e-05], + [ 2.8824e-07, 1.3374e-06, 4.1351e-06, ..., 1.6978e-06, + 0.0000e+00, -4.1239e-06], + [ 6.0350e-07, 2.7977e-06, 2.0698e-05, ..., 5.6438e-06, + -9.3132e-09, 1.8343e-05], + ..., + [-6.2957e-06, -5.0634e-05, 6.9151e-07, ..., -7.2062e-05, + 6.9849e-10, 1.8492e-05], + [-2.2754e-05, 1.4612e-06, -1.4019e-03, ..., -3.4213e-04, + 2.3283e-09, -1.2980e-03], + [ 2.2531e-05, 2.6807e-05, 1.2207e-03, ..., 3.0398e-04, + 0.0000e+00, 1.1215e-03]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0152, -0.0177, -0.0265, -0.0264, -0.0084, 0.0025, 0.0112, -0.0207, + -0.0064, 0.0035], device='cuda:0'), grad: tensor([ 1.2338e-04, -9.8422e-06, 4.9233e-05, 1.4186e-04, 1.0002e-04, + 2.8327e-05, 9.6858e-06, -1.0961e-04, -2.7370e-03, 2.4071e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 220.69, cls_loss 0.0026 cls_loss_mapping 0.0068 cls_loss_causal 0.5484 re_mapping 0.0077 re_causal 0.0215 /// teacc 99.01 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.1346, -0.1293, 0.0764, ..., -0.0571, 0.0406, 0.0260], + [-0.0739, -0.0285, -0.0431, ..., -0.0596, -0.0741, -0.0356], + [ 0.0329, -0.0635, -0.0377, ..., -0.0483, 0.0061, -0.2195], + ..., + [-0.0933, 0.0909, 0.0029, ..., 0.0892, -0.0237, -0.0925], + [-0.0529, -0.0569, 0.0876, ..., -0.0451, -0.1269, 0.0930], + [ 0.0005, -0.1374, 0.0844, ..., 0.0559, -0.1417, -0.0613]], + device='cuda:0'), grad: tensor([[ 8.1491e-09, 6.1560e-07, -6.1980e-07, ..., 6.4541e-07, + 6.9849e-10, -1.1013e-07], + [ 9.5461e-09, 2.1290e-06, 1.1548e-06, ..., 2.7679e-06, + 1.1642e-09, -2.9579e-06], + [ 4.6566e-09, 1.8552e-06, 1.7704e-06, ..., 3.7509e-07, + -5.2387e-08, 2.1439e-06], + ..., + [ 3.0268e-08, -2.1175e-05, -7.1786e-06, ..., -2.0087e-05, + 5.1223e-09, 2.1420e-06], + [ 3.0734e-08, 1.0766e-06, -3.0510e-06, ..., 2.5285e-07, + 4.0745e-08, -5.3346e-06], + [-3.2550e-07, 8.1211e-06, 4.2580e-06, ..., 7.8529e-06, + 2.3283e-10, 2.9337e-06]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0159, -0.0186, -0.0269, -0.0274, -0.0087, 0.0029, 0.0126, -0.0204, + -0.0059, 0.0039], device='cuda:0'), grad: tensor([ 2.3209e-06, -6.3963e-06, -2.0459e-05, 5.4315e-06, 1.3359e-05, + 6.4149e-06, -3.6433e-06, -1.4745e-05, -3.9041e-06, 2.1636e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 220.86, cls_loss 0.0025 cls_loss_mapping 0.0077 cls_loss_causal 0.5597 re_mapping 0.0070 re_causal 0.0205 /// teacc 98.94 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.1347, -0.1303, 0.0768, ..., -0.0574, 0.0406, 0.0260], + [-0.0742, -0.0280, -0.0432, ..., -0.0591, -0.0741, -0.0346], + [ 0.0324, -0.0644, -0.0384, ..., -0.0491, 0.0063, -0.2210], + ..., + [-0.0937, 0.0912, 0.0028, ..., 0.0893, -0.0238, -0.0941], + [-0.0532, -0.0573, 0.0878, ..., -0.0456, -0.1269, 0.0944], + [ 0.0009, -0.1383, 0.0849, ..., 0.0562, -0.1418, -0.0617]], + device='cuda:0'), grad: tensor([[ 8.1062e-06, 3.0897e-07, 4.7013e-06, ..., 6.7335e-07, + 1.0105e-07, 1.6969e-06], + [ 4.8988e-06, 1.8813e-07, 3.2336e-06, ..., 8.8057e-07, + 2.1188e-08, -5.3737e-07], + [ 6.3121e-05, -5.3272e-07, 3.8564e-05, ..., 6.5286e-07, + -1.6037e-06, 1.0550e-05], + ..., + [ 2.9467e-06, -2.1197e-06, 8.5495e-07, ..., -2.4699e-06, + 3.9837e-07, 1.0515e-06], + [ 9.8944e-06, 3.3062e-07, 5.8115e-06, ..., 1.3513e-06, + 2.9593e-07, 1.0338e-06], + [-1.2884e-03, 1.0692e-06, -4.4823e-04, ..., -1.4181e-03, + 3.2829e-08, -5.7173e-04]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0160, -0.0178, -0.0272, -0.0278, -0.0090, 0.0025, 0.0125, -0.0209, + -0.0049, 0.0039], device='cuda:0'), grad: tensor([ 2.9787e-05, 1.3575e-05, 2.2233e-04, 1.7416e-04, 4.1466e-03, + 7.6652e-05, 5.7936e-05, 8.9854e-06, 3.5733e-05, -4.7684e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 220.86, cls_loss 0.0023 cls_loss_mapping 0.0071 cls_loss_causal 0.5800 re_mapping 0.0070 re_causal 0.0206 /// teacc 98.85 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.1357, -0.1313, 0.0767, ..., -0.0580, 0.0407, 0.0260], + [-0.0762, -0.0277, -0.0437, ..., -0.0597, -0.0743, -0.0348], + [ 0.0336, -0.0649, -0.0389, ..., -0.0496, 0.0065, -0.2220], + ..., + [-0.0945, 0.0911, 0.0030, ..., 0.0901, -0.0239, -0.0946], + [-0.0540, -0.0576, 0.0883, ..., -0.0459, -0.1270, 0.0950], + [ 0.0012, -0.1395, 0.0853, ..., 0.0561, -0.1421, -0.0618]], + device='cuda:0'), grad: tensor([[ 4.2724e-07, 3.2526e-07, -1.4789e-06, ..., 1.7565e-06, + -4.0978e-08, -1.4296e-06], + [ 5.8766e-07, 1.4305e-06, 1.6503e-06, ..., 3.5204e-06, + 3.2596e-09, 3.0152e-07], + [ 3.3760e-07, 1.2526e-06, 2.8498e-06, ..., 2.2165e-06, + 4.8894e-09, 2.0619e-06], + ..., + [ 4.7311e-06, -6.1877e-06, 7.5474e-06, ..., 9.6485e-06, + 5.1223e-09, 6.4820e-07], + [ 2.2799e-06, 2.9802e-08, 3.5875e-06, ..., 8.0690e-06, + 3.2596e-09, -1.6149e-06], + [-4.4107e-05, 1.6075e-06, -8.7619e-05, ..., -1.4758e-04, + 3.4925e-09, 3.7742e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0160, -0.0174, -0.0270, -0.0279, -0.0086, 0.0030, 0.0122, -0.0214, + -0.0049, 0.0037], device='cuda:0'), grad: tensor([-3.0501e-08, 8.5384e-06, 1.9316e-06, 2.0146e-05, 2.6441e-04, + 5.4091e-06, -3.1609e-06, 3.5763e-05, 1.7524e-05, -3.5095e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 220.83, cls_loss 0.0026 cls_loss_mapping 0.0070 cls_loss_causal 0.5335 re_mapping 0.0072 re_causal 0.0197 /// teacc 98.94 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.1359, -0.1307, 0.0757, ..., -0.0606, 0.0407, 0.0270], + [-0.0766, -0.0280, -0.0443, ..., -0.0603, -0.0750, -0.0350], + [ 0.0332, -0.0656, -0.0401, ..., -0.0500, 0.0072, -0.2230], + ..., + [-0.0952, 0.0918, 0.0030, ..., 0.0906, -0.0240, -0.0951], + [-0.0551, -0.0583, 0.0888, ..., -0.0462, -0.1271, 0.0955], + [ 0.0018, -0.1416, 0.0869, ..., 0.0569, -0.1422, -0.0625]], + device='cuda:0'), grad: tensor([[ 1.2908e-06, 1.2224e-07, -7.8678e-06, ..., 7.6951e-08, + -2.1153e-07, -5.8375e-06], + [ 1.5628e-06, 3.3434e-07, 6.4000e-06, ..., 2.2934e-07, + 1.2002e-07, 3.2634e-06], + [ 8.7991e-06, -5.8971e-06, 2.4393e-05, ..., 1.8964e-07, + -2.3097e-06, 7.7933e-06], + ..., + [ 9.1316e-07, 1.1399e-06, 2.5537e-06, ..., -6.1560e-07, + 6.6031e-07, 1.4147e-06], + [-8.7261e-05, 7.2643e-07, -2.5654e-04, ..., 1.0477e-07, + 4.4843e-07, -4.8667e-05], + [ 4.8637e-05, 8.5030e-07, 1.2082e-04, ..., -3.7858e-07, + 3.4226e-08, 3.7223e-05]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0166, -0.0175, -0.0273, -0.0276, -0.0084, 0.0030, 0.0114, -0.0214, + -0.0048, 0.0043], device='cuda:0'), grad: tensor([-6.2734e-06, 1.5676e-05, 4.5806e-05, 1.8620e-04, -3.5972e-08, + 2.8148e-05, -1.3344e-05, 1.1258e-05, -5.5408e-04, 2.8658e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 220.65, cls_loss 0.0020 cls_loss_mapping 0.0050 cls_loss_causal 0.5534 re_mapping 0.0067 re_causal 0.0204 /// teacc 98.93 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.1359, -0.1312, 0.0761, ..., -0.0607, 0.0427, 0.0277], + [-0.0773, -0.0287, -0.0453, ..., -0.0605, -0.0752, -0.0350], + [ 0.0323, -0.0662, -0.0405, ..., -0.0504, 0.0074, -0.2239], + ..., + [-0.0957, 0.0922, 0.0031, ..., 0.0909, -0.0242, -0.0958], + [-0.0564, -0.0583, 0.0897, ..., -0.0462, -0.1273, 0.0956], + [ 0.0011, -0.1427, 0.0869, ..., 0.0571, -0.1432, -0.0637]], + device='cuda:0'), grad: tensor([[ 5.3644e-07, 1.1770e-07, 3.7774e-06, ..., 3.4962e-06, + -6.4145e-08, 4.2492e-07], + [ 8.7544e-08, 5.8580e-07, 7.6462e-07, ..., 6.2305e-07, + 1.9989e-07, -1.0328e-06], + [ 7.9861e-08, -5.6997e-07, 2.7958e-06, ..., 1.1884e-06, + -9.5461e-07, 2.2706e-06], + ..., + [ 1.3039e-06, 3.8510e-07, 1.7434e-05, ..., 1.3441e-05, + 5.9232e-07, 2.4159e-06], + [ 7.7765e-07, -3.1013e-07, -5.9456e-06, ..., 4.4284e-07, + 6.8219e-08, -9.3505e-06], + [ 3.9116e-06, 1.1902e-06, -3.4809e-05, ..., -2.7657e-05, + 1.1874e-08, 1.4432e-05]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0161, -0.0181, -0.0275, -0.0274, -0.0082, 0.0039, 0.0109, -0.0212, + -0.0047, 0.0038], device='cuda:0'), grad: tensor([ 8.9407e-06, -1.6391e-06, -1.6838e-06, 1.2994e-05, 1.1757e-05, + -1.7166e-05, 5.1484e-06, 4.2319e-05, -1.0118e-05, -5.0694e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 221.31, cls_loss 0.0021 cls_loss_mapping 0.0055 cls_loss_causal 0.5629 re_mapping 0.0070 re_causal 0.0204 /// teacc 98.92 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.1365, -0.1316, 0.0765, ..., -0.0609, 0.0428, 0.0280], + [-0.0777, -0.0284, -0.0454, ..., -0.0611, -0.0754, -0.0350], + [ 0.0319, -0.0667, -0.0411, ..., -0.0509, 0.0079, -0.2247], + ..., + [-0.0960, 0.0923, 0.0028, ..., 0.0915, -0.0243, -0.0962], + [-0.0574, -0.0587, 0.0904, ..., -0.0463, -0.1277, 0.0960], + [ 0.0013, -0.1440, 0.0870, ..., 0.0568, -0.1434, -0.0644]], + device='cuda:0'), grad: tensor([[-8.5915e-08, 1.7113e-07, -1.6140e-06, ..., 4.8243e-07, + 6.8266e-07, 1.1269e-07], + [ 2.5611e-08, 7.7765e-08, 1.2601e-06, ..., 4.7544e-07, + 2.2687e-06, 5.2117e-06], + [ 3.8650e-08, -8.3074e-07, 1.6242e-06, ..., 7.9814e-07, + 1.6531e-07, 2.0098e-06], + ..., + [ 1.6904e-07, -3.9372e-07, 2.7586e-06, ..., 1.0459e-06, + 4.5169e-08, 2.2575e-06], + [ 3.0850e-07, -1.0384e-07, -9.1866e-06, ..., -2.1607e-06, + 8.9873e-08, -8.7991e-06], + [ 1.9395e-07, 4.7125e-07, 1.1898e-07, ..., 2.1257e-07, + 3.2829e-08, 2.6207e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0159, -0.0183, -0.0270, -0.0272, -0.0078, 0.0038, 0.0110, -0.0215, + -0.0045, 0.0035], device='cuda:0'), grad: tensor([ 5.3123e-06, -4.6998e-05, -7.7486e-06, 4.7460e-06, 3.2663e-05, + 4.6976e-06, -2.4855e-05, 1.8731e-05, 3.3788e-06, 9.9838e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 220.79, cls_loss 0.0021 cls_loss_mapping 0.0060 cls_loss_causal 0.5457 re_mapping 0.0067 re_causal 0.0196 /// teacc 99.00 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.1368, -0.1324, 0.0763, ..., -0.0614, 0.0427, 0.0278], + [-0.0783, -0.0285, -0.0455, ..., -0.0614, -0.0757, -0.0352], + [ 0.0324, -0.0673, -0.0417, ..., -0.0516, 0.0091, -0.2256], + ..., + [-0.0964, 0.0925, 0.0024, ..., 0.0919, -0.0244, -0.0976], + [-0.0585, -0.0587, 0.0906, ..., -0.0472, -0.1289, 0.0959], + [ 0.0012, -0.1448, 0.0881, ..., 0.0573, -0.1437, -0.0639]], + device='cuda:0'), grad: tensor([[ 6.8359e-06, -6.5705e-07, 2.6543e-08, ..., 1.3644e-06, + -3.1665e-08, 3.3919e-06], + [ 2.6124e-07, 5.4343e-07, 2.3637e-06, ..., 1.3094e-06, + 2.8405e-08, 9.5367e-06], + [-1.3493e-05, 1.2433e-06, 1.6525e-05, ..., 7.2196e-06, + -1.5199e-06, 1.5914e-05], + ..., + [ 5.0059e-07, -5.2080e-06, 8.2701e-07, ..., -9.6112e-07, + 3.3481e-07, 2.5183e-06], + [ 5.9009e-06, 1.6540e-06, -1.4448e-04, ..., -6.7532e-05, + 2.2165e-07, -2.4486e-04], + [ 1.0416e-05, 7.7160e-07, 9.3699e-05, ..., 6.0827e-05, + 5.9139e-08, 1.0049e-04]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0164, -0.0183, -0.0268, -0.0272, -0.0079, 0.0042, 0.0114, -0.0217, + -0.0053, 0.0040], device='cuda:0'), grad: tensor([ 5.4002e-05, 1.7673e-05, -8.1062e-05, 9.0361e-05, -2.9892e-05, + 6.5923e-05, 4.0889e-05, 4.5970e-06, -4.2152e-04, 2.5916e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 220.71, cls_loss 0.0017 cls_loss_mapping 0.0052 cls_loss_causal 0.5714 re_mapping 0.0067 re_causal 0.0207 /// teacc 98.98 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.1372, -0.1338, 0.0762, ..., -0.0618, 0.0426, 0.0279], + [-0.0785, -0.0290, -0.0458, ..., -0.0617, -0.0761, -0.0352], + [ 0.0323, -0.0676, -0.0418, ..., -0.0514, 0.0094, -0.2261], + ..., + [-0.0967, 0.0932, 0.0020, ..., 0.0920, -0.0245, -0.0982], + [-0.0591, -0.0589, 0.0911, ..., -0.0475, -0.1292, 0.0962], + [ 0.0013, -0.1453, 0.0887, ..., 0.0577, -0.1440, -0.0642]], + device='cuda:0'), grad: tensor([[ 5.3877e-07, 2.8331e-06, 3.9348e-07, ..., 2.5406e-06, + 1.1176e-07, -6.9384e-08], + [ 2.0117e-07, 1.1876e-05, 5.9754e-06, ..., 6.5751e-06, + 1.3039e-08, -6.0815e-07], + [ 1.7369e-07, 1.1131e-05, 5.5879e-06, ..., 6.6087e-06, + -1.3281e-06, 1.1856e-06], + ..., + [ 1.0058e-07, -7.0274e-05, -3.1501e-05, ..., -3.6180e-05, + 1.5367e-07, 2.6636e-07], + [ 2.2668e-06, 1.7956e-05, 1.0803e-05, ..., 2.8647e-06, + 5.1083e-07, 8.7917e-07], + [-1.9185e-06, 1.2331e-05, 7.8930e-07, ..., 9.5293e-06, + 6.1002e-08, -8.8476e-08]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0169, -0.0185, -0.0265, -0.0272, -0.0083, 0.0042, 0.0115, -0.0215, + -0.0053, 0.0042], device='cuda:0'), grad: tensor([ 1.0833e-05, 2.0668e-05, 2.9132e-06, 1.8612e-05, -3.7163e-05, + 1.4529e-06, 2.0042e-05, -1.1253e-04, 5.0515e-05, 2.4676e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 220.72, cls_loss 0.0020 cls_loss_mapping 0.0063 cls_loss_causal 0.5355 re_mapping 0.0072 re_causal 0.0199 /// teacc 98.99 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.1375, -0.1366, 0.0763, ..., -0.0620, 0.0425, 0.0279], + [-0.0792, -0.0300, -0.0462, ..., -0.0621, -0.0765, -0.0351], + [ 0.0321, -0.0683, -0.0421, ..., -0.0521, 0.0107, -0.2265], + ..., + [-0.0965, 0.0960, 0.0023, ..., 0.0931, -0.0238, -0.0986], + [-0.0605, -0.0608, 0.0912, ..., -0.0478, -0.1309, 0.0964], + [ 0.0014, -0.1474, 0.0893, ..., 0.0580, -0.1442, -0.0644]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 9.2667e-08, -1.6764e-06, ..., 8.3353e-08, + 4.1910e-09, -2.2016e-06], + [ 1.0245e-08, -8.2403e-06, 3.0827e-07, ..., 1.2666e-07, + -2.3283e-09, 1.0524e-07], + [ 2.5146e-08, 3.6974e-07, 2.4866e-07, ..., 2.0955e-07, + 1.3970e-08, 3.5390e-07], + ..., + [ 5.1688e-08, -9.3644e-07, -8.4285e-08, ..., -1.2182e-06, + 1.7695e-08, 4.0140e-07], + [ 5.7789e-07, 1.2573e-07, 2.1327e-07, ..., 1.2899e-07, + 2.2817e-08, 1.7863e-06], + [-4.7171e-07, 6.8638e-07, -1.4678e-06, ..., -1.0058e-06, + 7.9162e-09, 4.2189e-07]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0177, -0.0191, -0.0266, -0.0278, -0.0091, 0.0040, 0.0114, -0.0195, + -0.0057, 0.0042], device='cuda:0'), grad: tensor([-4.7013e-06, -4.8488e-05, 1.2554e-06, 7.5735e-06, 4.7415e-05, + -1.4365e-05, 7.4059e-06, 1.3374e-06, 3.2037e-06, -6.5146e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 220.67, cls_loss 0.0024 cls_loss_mapping 0.0073 cls_loss_causal 0.5962 re_mapping 0.0067 re_causal 0.0200 /// teacc 98.98 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.1378, -0.1368, 0.0759, ..., -0.0634, 0.0424, 0.0282], + [-0.0799, -0.0303, -0.0468, ..., -0.0624, -0.0749, -0.0341], + [ 0.0321, -0.0684, -0.0429, ..., -0.0526, 0.0107, -0.2282], + ..., + [-0.0969, 0.0965, 0.0024, ..., 0.0936, -0.0240, -0.0994], + [-0.0614, -0.0613, 0.0919, ..., -0.0482, -0.1312, 0.0970], + [ 0.0014, -0.1480, 0.0901, ..., 0.0584, -0.1447, -0.0651]], + device='cuda:0'), grad: tensor([[ 4.6380e-07, 2.8173e-07, -1.3039e-08, ..., 6.5193e-07, + 4.6566e-10, 1.6866e-06], + [ 9.4110e-07, 1.3281e-06, 1.1884e-05, ..., 1.3215e-06, + 4.6566e-10, 2.7642e-05], + [ 4.6287e-07, 1.3541e-06, 1.0610e-04, ..., 8.7125e-07, + -8.3819e-09, 2.7204e-04], + ..., + [ 1.6615e-06, -1.4901e-07, 4.6268e-06, ..., 2.8871e-06, + 2.7940e-09, 3.4459e-06], + [ 1.7229e-06, 8.0978e-07, -1.2279e-04, ..., 2.0359e-06, + 2.3283e-09, -3.2043e-04], + [-6.6400e-05, 3.2969e-07, -4.1336e-05, ..., -5.0604e-05, + 0.0000e+00, -3.7737e-06]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0178, -0.0190, -0.0267, -0.0283, -0.0090, 0.0042, 0.0107, -0.0195, + -0.0055, 0.0043], device='cuda:0'), grad: tensor([ 7.1339e-06, 8.3745e-05, 6.9618e-04, 1.7583e-05, 1.2875e-04, + 3.2842e-05, 1.0893e-05, 2.2724e-05, -8.2970e-04, -1.7047e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 220.75, cls_loss 0.0022 cls_loss_mapping 0.0058 cls_loss_causal 0.5815 re_mapping 0.0065 re_causal 0.0194 /// teacc 98.94 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.1382, -0.1370, 0.0761, ..., -0.0636, 0.0433, 0.0283], + [-0.0800, -0.0303, -0.0472, ..., -0.0628, -0.0753, -0.0346], + [ 0.0316, -0.0688, -0.0434, ..., -0.0530, 0.0111, -0.2306], + ..., + [-0.0976, 0.0968, 0.0022, ..., 0.0939, -0.0238, -0.0999], + [-0.0618, -0.0616, 0.0926, ..., -0.0488, -0.1315, 0.0984], + [ 0.0020, -0.1488, 0.0907, ..., 0.0588, -0.1461, -0.0659]], + device='cuda:0'), grad: tensor([[ 1.5926e-07, 1.5050e-06, -3.4692e-07, ..., 5.8860e-07, + 1.7136e-07, -9.9931e-07], + [ 1.1921e-07, 5.3048e-06, 1.7900e-06, ..., 1.1967e-06, + 5.2154e-08, 4.1910e-09], + [ 4.0000e-07, 2.9728e-06, 1.0747e-06, ..., -6.9803e-07, + -7.7719e-07, 1.8813e-07], + ..., + [ 8.3679e-07, -1.6436e-05, -5.4128e-06, ..., -1.7788e-06, + 4.1444e-07, 3.4971e-07], + [ 1.1586e-06, 1.2480e-06, 3.5437e-07, ..., 7.2736e-07, + 6.3796e-08, 1.4976e-06], + [ 2.8759e-06, 5.5358e-06, 2.0489e-08, ..., 4.6305e-06, + 1.0245e-08, 8.4937e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0179, -0.0188, -0.0272, -0.0286, -0.0090, 0.0042, 0.0107, -0.0198, + -0.0042, 0.0043], device='cuda:0'), grad: tensor([ 1.0423e-05, 1.0565e-05, -2.4550e-06, -1.2163e-06, -1.6376e-05, + -3.2149e-06, -1.1325e-05, -1.9982e-05, 7.5586e-06, 2.6077e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 221.05, cls_loss 0.0028 cls_loss_mapping 0.0061 cls_loss_causal 0.5739 re_mapping 0.0072 re_causal 0.0197 /// teacc 99.02 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.1392, -0.1375, 0.0762, ..., -0.0636, 0.0431, 0.0279], + [-0.0807, -0.0303, -0.0475, ..., -0.0633, -0.0762, -0.0337], + [ 0.0305, -0.0695, -0.0439, ..., -0.0535, 0.0119, -0.2326], + ..., + [-0.0984, 0.0972, 0.0022, ..., 0.0943, -0.0246, -0.1009], + [-0.0622, -0.0622, 0.0938, ..., -0.0490, -0.1321, 0.0997], + [ 0.0025, -0.1503, 0.0908, ..., 0.0586, -0.1477, -0.0672]], + device='cuda:0'), grad: tensor([[ 2.7008e-07, 1.5292e-06, -5.0431e-07, ..., 2.3656e-07, + 1.5320e-07, 6.3423e-07], + [ 3.0221e-07, 1.6298e-06, 3.8045e-07, ..., 3.6554e-07, + 1.6410e-06, 9.9279e-07], + [ 5.5274e-07, 3.8221e-06, 2.9383e-07, ..., 2.1420e-07, + 3.9581e-07, 1.3858e-06], + ..., + [ 8.4843e-07, 2.9728e-06, 1.7341e-06, ..., 8.1630e-07, + 3.8138e-07, 1.6363e-06], + [ 4.9509e-06, 2.7433e-05, 1.2470e-06, ..., 2.9225e-06, + 1.0151e-07, -5.0098e-05], + [-1.1511e-06, 3.2634e-06, -9.5293e-06, ..., -6.1691e-06, + 6.0070e-07, -1.1642e-06]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0182, -0.0184, -0.0277, -0.0287, -0.0085, 0.0037, 0.0113, -0.0199, + -0.0034, 0.0036], device='cuda:0'), grad: tensor([ 5.1931e-06, 2.2113e-05, 1.0520e-05, -9.5749e-04, -4.3201e-04, + 9.4175e-04, 4.4227e-04, 1.2778e-05, -4.0859e-05, -5.5581e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 220.53, cls_loss 0.0027 cls_loss_mapping 0.0078 cls_loss_causal 0.5718 re_mapping 0.0068 re_causal 0.0192 /// teacc 98.86 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.1391, -0.1383, 0.0761, ..., -0.0644, 0.0428, 0.0279], + [-0.0816, -0.0308, -0.0482, ..., -0.0639, -0.0765, -0.0340], + [ 0.0327, -0.0711, -0.0444, ..., -0.0541, 0.0118, -0.2337], + ..., + [-0.1003, 0.0976, 0.0022, ..., 0.0949, -0.0246, -0.1016], + [-0.0637, -0.0640, 0.0938, ..., -0.0495, -0.1322, 0.1002], + [ 0.0026, -0.1519, 0.0915, ..., 0.0590, -0.1481, -0.0673]], + device='cuda:0'), grad: tensor([[-5.5879e-09, 7.7533e-07, -6.0583e-07, ..., 1.2433e-07, + 0.0000e+00, -1.3132e-06], + [ 4.1910e-09, 2.2706e-06, 2.5891e-07, ..., 2.2678e-07, + 4.6566e-10, -1.0496e-06], + [ 1.8626e-09, 3.4533e-06, 9.6299e-07, ..., 2.9653e-06, + -2.1886e-08, 2.1979e-07], + ..., + [ 1.6764e-08, 1.6773e-04, 7.2062e-05, ..., -7.0371e-06, + 7.9162e-09, 3.3155e-07], + [ 1.7881e-07, 2.0321e-06, 9.8906e-07, ..., 4.4284e-07, + 6.5193e-09, 4.4517e-07], + [-6.8918e-08, 1.6969e-06, -1.1958e-06, ..., -1.0654e-06, + 9.3132e-10, 2.1560e-07]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0188, -0.0193, -0.0269, -0.0262, -0.0080, 0.0030, 0.0115, -0.0199, + -0.0038, 0.0033], device='cuda:0'), grad: tensor([-1.8124e-06, 1.1519e-05, -1.7017e-05, -1.8764e-04, 2.1476e-06, + 1.4631e-06, 2.1830e-06, 1.8442e-04, 4.8317e-06, -3.4506e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 220.98, cls_loss 0.0032 cls_loss_mapping 0.0063 cls_loss_causal 0.5652 re_mapping 0.0070 re_causal 0.0196 /// teacc 99.01 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.1396, -0.1387, 0.0763, ..., -0.0647, 0.0428, 0.0281], + [-0.0798, -0.0312, -0.0481, ..., -0.0634, -0.0766, -0.0303], + [ 0.0323, -0.0711, -0.0452, ..., -0.0554, 0.0120, -0.2358], + ..., + [-0.1007, 0.0988, 0.0023, ..., 0.0952, -0.0247, -0.1032], + [-0.0648, -0.0660, 0.0944, ..., -0.0497, -0.1325, 0.1003], + [ 0.0033, -0.1534, 0.0919, ..., 0.0591, -0.1485, -0.0678]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 2.5611e-08, -6.2399e-07, ..., 6.5193e-08, + 1.3970e-09, -2.6124e-07], + [ 2.5611e-08, 1.5460e-07, 1.2713e-07, ..., 1.1688e-07, + -6.0070e-08, -7.0455e-07], + [-2.1420e-07, 3.7272e-06, 1.4249e-06, ..., 2.3693e-06, + -1.9558e-08, 1.3085e-07], + ..., + [ 7.4692e-07, -4.5188e-06, -1.5199e-06, ..., -2.9635e-06, + 5.1223e-09, 2.8126e-06], + [ 2.4252e-06, 4.2003e-07, 3.6275e-07, ..., 3.3947e-07, + 5.7276e-08, 8.3819e-06], + [-1.2619e-07, 1.0477e-07, -7.5530e-07, ..., -3.2736e-07, + 4.6566e-10, 4.8894e-08]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0189, -0.0181, -0.0275, -0.0271, -0.0079, 0.0028, 0.0110, -0.0197, + -0.0048, 0.0034], device='cuda:0'), grad: tensor([-5.3365e-07, -9.1409e-07, -6.5565e-06, 3.3397e-06, 8.9454e-07, + -9.9838e-06, 6.5984e-07, -3.1404e-06, 1.6302e-05, -1.0384e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 220.59, cls_loss 0.0024 cls_loss_mapping 0.0043 cls_loss_causal 0.5396 re_mapping 0.0068 re_causal 0.0193 /// teacc 98.93 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.1400, -0.1389, 0.0766, ..., -0.0648, 0.0427, 0.0282], + [-0.0802, -0.0306, -0.0482, ..., -0.0638, -0.0769, -0.0301], + [ 0.0332, -0.0694, -0.0445, ..., -0.0534, 0.0121, -0.2362], + ..., + [-0.1015, 0.0976, 0.0018, ..., 0.0950, -0.0244, -0.1037], + [-0.0652, -0.0671, 0.0949, ..., -0.0501, -0.1328, 0.1008], + [ 0.0035, -0.1544, 0.0920, ..., 0.0589, -0.1487, -0.0678]], + device='cuda:0'), grad: tensor([[-1.6717e-07, 9.4855e-07, -1.2815e-05, ..., 4.2422e-07, + -4.3772e-08, -8.6650e-06], + [ 2.2352e-08, 6.9523e-07, 1.2182e-06, ..., 4.5775e-07, + 1.3970e-09, 3.2177e-07], + [ 5.2620e-08, 2.0508e-06, 2.7362e-06, ..., 1.0887e-06, + 3.7253e-09, 2.1011e-06], + ..., + [-4.3306e-08, -1.2465e-05, -2.7623e-06, ..., -6.3740e-06, + 7.9162e-09, 1.6745e-06], + [ 1.2945e-07, -3.4785e-07, -8.2180e-06, ..., 7.6974e-07, + 1.8626e-09, -1.2189e-05], + [ 3.4925e-08, 6.0610e-06, 3.8594e-06, ..., 1.3635e-06, + 1.2107e-08, 5.1782e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0188, -0.0176, -0.0254, -0.0272, -0.0071, 0.0023, 0.0110, -0.0210, + -0.0048, 0.0027], device='cuda:0'), grad: tensor([-5.2214e-05, 2.6990e-06, 1.0259e-05, 3.9071e-05, -1.2685e-06, + 9.4324e-06, 6.8061e-06, -1.2174e-05, -2.1979e-05, 1.9401e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 220.15, cls_loss 0.0022 cls_loss_mapping 0.0060 cls_loss_causal 0.5607 re_mapping 0.0067 re_causal 0.0192 /// teacc 98.93 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.1400, -0.1389, 0.0768, ..., -0.0650, 0.0427, 0.0280], + [-0.0807, -0.0309, -0.0486, ..., -0.0644, -0.0769, -0.0302], + [ 0.0330, -0.0698, -0.0449, ..., -0.0541, 0.0122, -0.2367], + ..., + [-0.1022, 0.0983, 0.0025, ..., 0.0960, -0.0245, -0.1038], + [-0.0659, -0.0676, 0.0953, ..., -0.0504, -0.1328, 0.1013], + [ 0.0028, -0.1573, 0.0922, ..., 0.0585, -0.1488, -0.0693]], + device='cuda:0'), grad: tensor([[ 2.7381e-06, 1.1455e-07, 1.5035e-05, ..., 1.6406e-05, + 7.9162e-09, -2.8824e-07], + [ 1.2266e-06, 4.8941e-07, 4.7162e-06, ..., 1.0863e-05, + 4.1910e-09, -1.3644e-07], + [ 3.2708e-06, 2.5239e-07, 3.6694e-06, ..., 1.2062e-05, + -4.6100e-08, 4.5635e-08], + ..., + [ 2.2184e-06, -1.0114e-06, 1.4342e-06, ..., 1.4889e-04, + 5.1223e-09, 1.3690e-07], + [ 5.2117e-06, 9.5041e-07, 2.4270e-06, ..., 1.8358e-05, + 1.3039e-08, 3.5856e-08], + [-9.4175e-05, 9.0385e-07, -6.2883e-05, ..., -6.7592e-05, + 1.3970e-09, 1.9791e-07]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0188, -0.0180, -0.0255, -0.0293, -0.0068, 0.0045, 0.0111, -0.0205, + -0.0046, 0.0019], device='cuda:0'), grad: tensor([ 6.7294e-05, 3.3557e-05, 3.6687e-05, -2.6628e-05, -3.7074e-04, + 9.0182e-05, 1.4067e-04, 3.3498e-04, 5.2303e-05, -3.5763e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 220.41, cls_loss 0.0023 cls_loss_mapping 0.0063 cls_loss_causal 0.5327 re_mapping 0.0066 re_causal 0.0184 /// teacc 99.00 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.1401, -0.1391, 0.0771, ..., -0.0651, 0.0427, 0.0280], + [-0.0812, -0.0309, -0.0494, ..., -0.0652, -0.0771, -0.0308], + [ 0.0332, -0.0700, -0.0454, ..., -0.0545, 0.0125, -0.2376], + ..., + [-0.1029, 0.0989, 0.0034, ..., 0.0971, -0.0246, -0.1039], + [-0.0673, -0.0680, 0.0957, ..., -0.0508, -0.1330, 0.1030], + [ 0.0026, -0.1599, 0.0921, ..., 0.0582, -0.1488, -0.0701]], + device='cuda:0'), grad: tensor([[-3.4599e-07, 5.0897e-07, -4.7162e-06, ..., 4.4703e-08, + -5.2573e-07, -6.1980e-07], + [ 1.0990e-06, 4.6678e-06, 5.2974e-06, ..., 8.9873e-07, + 1.8626e-09, 4.2468e-06], + [ 1.0058e-07, 8.7768e-06, 1.8820e-05, ..., -3.0268e-08, + 8.3353e-08, 1.3389e-05], + ..., + [ 9.2201e-08, 2.2221e-06, 3.7998e-06, ..., 2.7148e-07, + 9.3132e-10, 2.5928e-06], + [ 1.6997e-07, -1.5251e-05, -3.4034e-05, ..., 1.4203e-07, + 2.5611e-08, -2.4691e-05], + [ 2.9942e-07, 1.0077e-06, 1.1940e-06, ..., 1.4063e-07, + 2.1886e-08, 9.3924e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0188, -0.0180, -0.0257, -0.0288, -0.0069, 0.0039, 0.0103, -0.0203, + -0.0032, 0.0015], device='cuda:0'), grad: tensor([-1.2077e-05, 3.7879e-05, 4.5300e-05, -2.1756e-06, -1.9580e-05, + 1.6987e-05, -4.2170e-06, 1.6004e-05, -8.4937e-05, 6.8657e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 220.41, cls_loss 0.0022 cls_loss_mapping 0.0056 cls_loss_causal 0.5346 re_mapping 0.0066 re_causal 0.0180 /// teacc 99.03 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.1399, -0.1392, 0.0796, ..., -0.0640, 0.0429, 0.0300], + [-0.0817, -0.0315, -0.0500, ..., -0.0660, -0.0784, -0.0315], + [ 0.0336, -0.0703, -0.0460, ..., -0.0548, 0.0124, -0.2386], + ..., + [-0.1035, 0.0993, 0.0023, ..., 0.0976, -0.0246, -0.1050], + [-0.0697, -0.0685, 0.0954, ..., -0.0516, -0.1330, 0.1021], + [ 0.0032, -0.1606, 0.0923, ..., 0.0586, -0.1490, -0.0714]], + device='cuda:0'), grad: tensor([[ 3.5856e-07, 2.2119e-07, -1.3970e-08, ..., 2.0955e-08, + 0.0000e+00, 4.6473e-07], + [ 1.9791e-07, 1.3551e-07, 2.2817e-08, ..., 2.5146e-08, + 0.0000e+00, 2.8033e-07], + [ 5.2899e-07, -3.9628e-07, 9.4529e-08, ..., 7.9628e-08, + 0.0000e+00, 7.3621e-07], + ..., + [ 3.1665e-07, 9.4995e-08, -7.9628e-08, ..., -1.1362e-07, + 0.0000e+00, 3.9535e-07], + [ 1.7229e-06, 6.8499e-07, 1.2061e-07, ..., 8.5682e-08, + 4.6566e-10, 2.6375e-06], + [ 2.3469e-06, 8.2236e-07, -3.3015e-07, ..., -1.4063e-07, + 0.0000e+00, 2.8722e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0160, -0.0186, -0.0258, -0.0281, -0.0072, 0.0037, 0.0112, -0.0204, + -0.0044, 0.0010], device='cuda:0'), grad: tensor([ 1.9427e-06, 1.0775e-06, -4.8243e-06, 1.7047e-05, -4.6473e-07, + -2.3663e-05, -4.8652e-06, 2.1737e-06, 6.4485e-06, 5.1297e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 220.53, cls_loss 0.0021 cls_loss_mapping 0.0046 cls_loss_causal 0.5821 re_mapping 0.0065 re_causal 0.0189 /// teacc 98.86 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.1414, -0.1394, 0.0796, ..., -0.0642, 0.0432, 0.0299], + [-0.0821, -0.0316, -0.0515, ..., -0.0669, -0.0788, -0.0319], + [ 0.0338, -0.0706, -0.0467, ..., -0.0551, 0.0126, -0.2397], + ..., + [-0.1041, 0.0998, 0.0023, ..., 0.0982, -0.0242, -0.1053], + [-0.0708, -0.0696, 0.0951, ..., -0.0528, -0.1332, 0.1024], + [ 0.0041, -0.1612, 0.0938, ..., 0.0595, -0.1494, -0.0704]], + device='cuda:0'), grad: tensor([[ 3.7858e-07, 1.2107e-07, 6.0983e-06, ..., 9.4995e-08, + 0.0000e+00, 2.7418e-05], + [ 1.8682e-06, 5.2527e-07, -4.1962e-05, ..., 3.6601e-07, + 0.0000e+00, -1.0389e-04], + [ 4.3027e-06, 3.5763e-06, 1.6615e-06, ..., 3.0994e-06, + 0.0000e+00, 1.1063e-04], + ..., + [ 3.2270e-07, -4.6045e-06, 1.5991e-06, ..., -4.0457e-06, + 0.0000e+00, 1.3940e-05], + [-2.1830e-05, 1.8766e-07, 2.1547e-05, ..., 1.2806e-07, + 0.0000e+00, -3.7503e-04], + [ 1.4575e-07, 4.4238e-07, 8.6520e-07, ..., 2.9337e-08, + 0.0000e+00, 5.5358e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0162, -0.0191, -0.0257, -0.0275, -0.0079, 0.0031, 0.0112, -0.0202, + -0.0047, 0.0023], device='cuda:0'), grad: tensor([ 7.9274e-05, -3.8099e-04, 2.2626e-04, 1.9819e-05, 3.0011e-05, + 3.6454e-04, 2.5702e-04, 3.2127e-05, -6.4230e-04, 1.4432e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 220.99, cls_loss 0.0021 cls_loss_mapping 0.0065 cls_loss_causal 0.5581 re_mapping 0.0064 re_causal 0.0186 /// teacc 98.88 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.1416, -0.1397, 0.0798, ..., -0.0645, 0.0431, 0.0299], + [-0.0823, -0.0316, -0.0517, ..., -0.0673, -0.0789, -0.0315], + [ 0.0339, -0.0713, -0.0474, ..., -0.0553, 0.0127, -0.2407], + ..., + [-0.1045, 0.1000, 0.0006, ..., 0.0967, -0.0242, -0.1057], + [-0.0711, -0.0699, 0.0954, ..., -0.0531, -0.1333, 0.1034], + [ 0.0041, -0.1627, 0.0961, ..., 0.0623, -0.1496, -0.0709]], + device='cuda:0'), grad: tensor([[-2.5611e-08, 2.5239e-07, -3.3993e-08, ..., 2.3562e-07, + 1.3039e-08, 3.2596e-09], + [ 1.8626e-09, 1.3165e-05, 2.9374e-06, ..., 5.4091e-06, + 6.6124e-08, -7.7300e-08], + [ 2.3283e-09, 1.2042e-06, 5.4296e-07, ..., 7.1293e-07, + 7.5437e-08, 4.2235e-07], + ..., + [ 5.1223e-09, -3.7462e-05, -1.1541e-05, ..., -2.1189e-05, + 1.3039e-08, 1.9511e-07], + [ 1.8626e-09, 2.2352e-06, -4.2608e-07, ..., 1.0431e-06, + 1.5367e-08, -4.8522e-07], + [ 4.6566e-09, 8.3596e-06, 4.8354e-06, ..., 8.9034e-06, + 4.6566e-09, 2.1094e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0166, -0.0189, -0.0258, -0.0271, -0.0088, 0.0031, 0.0104, -0.0214, + -0.0040, 0.0043], device='cuda:0'), grad: tensor([ 9.2899e-07, 2.6315e-05, -2.0303e-07, 1.5691e-05, 8.2105e-06, + 4.0159e-06, -8.1733e-06, -7.6354e-05, 6.7353e-06, 2.2933e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 220.50, cls_loss 0.0025 cls_loss_mapping 0.0060 cls_loss_causal 0.5463 re_mapping 0.0065 re_causal 0.0183 /// teacc 99.00 lr 0.00010000 +Epoch 129, weight, value: tensor([[-1.4290e-01, -1.4000e-01, 7.9997e-02, ..., -6.4629e-02, + 4.3052e-02, 2.9482e-02], + [-8.2983e-02, -3.2171e-02, -5.2553e-02, ..., -6.8332e-02, + -7.9693e-02, -3.1947e-02], + [ 3.3566e-02, -7.1794e-02, -4.8039e-02, ..., -5.5922e-02, + 1.2678e-02, -2.4167e-01], + ..., + [-1.0481e-01, 9.9471e-02, 1.1697e-04, ..., 9.7235e-02, + -2.3901e-02, -1.0618e-01], + [-7.4304e-02, -7.0558e-02, 9.6268e-02, ..., -5.3398e-02, + -1.3333e-01, 1.0218e-01], + [ 4.1234e-03, -1.6389e-01, 9.6092e-02, ..., 6.2147e-02, + -1.4965e-01, -7.1429e-02]], device='cuda:0'), grad: tensor([[ 2.8592e-07, 1.4119e-06, 1.5860e-06, ..., 2.3097e-06, + 0.0000e+00, 7.5251e-07], + [ 8.0559e-08, 1.9576e-06, 3.0287e-06, ..., 4.8727e-06, + 0.0000e+00, 1.6708e-06], + [ 2.0955e-07, 2.5406e-06, 2.9355e-06, ..., 4.5225e-06, + -4.6566e-10, 4.0345e-06], + ..., + [ 3.3388e-07, -1.2064e-04, -1.9801e-04, ..., -3.2663e-04, + 0.0000e+00, 5.7230e-07], + [ 1.7472e-06, 7.2382e-06, 7.6890e-06, ..., 8.5011e-06, + 0.0000e+00, 2.7418e-06], + [ 1.8580e-07, 1.0020e-04, 1.7166e-04, ..., 2.9206e-04, + 0.0000e+00, 4.8848e-07]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0164, -0.0193, -0.0261, -0.0253, -0.0086, 0.0023, 0.0124, -0.0216, + -0.0049, 0.0038], device='cuda:0'), grad: tensor([ 1.2085e-05, 1.5177e-05, 6.7651e-05, 2.0057e-05, 5.3018e-05, + -1.8980e-06, -1.4222e-04, -6.8712e-04, 3.8147e-05, 6.2513e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 221.14, cls_loss 0.0021 cls_loss_mapping 0.0053 cls_loss_causal 0.5017 re_mapping 0.0067 re_causal 0.0181 /// teacc 99.01 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.1436, -0.1402, 0.0802, ..., -0.0650, 0.0431, 0.0297], + [-0.0840, -0.0324, -0.0543, ..., -0.0698, -0.0798, -0.0345], + [ 0.0319, -0.0723, -0.0486, ..., -0.0567, 0.0126, -0.2436], + ..., + [-0.1057, 0.1000, 0.0011, ..., 0.0982, -0.0238, -0.1038], + [-0.0756, -0.0709, 0.0965, ..., -0.0538, -0.1334, 0.1021], + [ 0.0043, -0.1648, 0.0963, ..., 0.0621, -0.1498, -0.0718]], + device='cuda:0'), grad: tensor([[ 2.4028e-07, 1.2107e-07, -5.9046e-07, ..., 1.0012e-07, + 1.0245e-08, -5.6904e-07], + [ 1.2806e-07, 4.9919e-07, 6.1747e-07, ..., 2.4028e-07, + 9.7789e-09, 5.0757e-08], + [ 1.6717e-07, 1.6727e-06, 1.5348e-06, ..., 1.8580e-07, + 4.3772e-08, 9.2899e-07], + ..., + [ 1.9837e-07, -4.0680e-06, -3.2652e-06, ..., -6.0424e-06, + 2.4214e-08, 8.5169e-07], + [ 2.4633e-07, 3.0156e-06, -2.7463e-05, ..., -1.3798e-05, + 9.7789e-09, -3.2693e-05], + [ 5.9046e-07, 4.9174e-06, 2.9474e-05, ..., 1.7807e-05, + 3.6322e-08, 2.9176e-05]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0165, -0.0199, -0.0266, -0.0263, -0.0085, 0.0033, 0.0130, -0.0209, + -0.0053, 0.0036], device='cuda:0'), grad: tensor([-8.0699e-07, -9.3319e-07, 6.0424e-06, -1.1459e-05, 1.5423e-06, + -2.3432e-06, 2.8852e-06, -6.4336e-06, -6.4611e-05, 7.6115e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 221.12, cls_loss 0.0020 cls_loss_mapping 0.0049 cls_loss_causal 0.5496 re_mapping 0.0064 re_causal 0.0181 /// teacc 98.96 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.1440, -0.1407, 0.0803, ..., -0.0652, 0.0431, 0.0298], + [-0.0847, -0.0327, -0.0550, ..., -0.0703, -0.0798, -0.0345], + [ 0.0318, -0.0721, -0.0487, ..., -0.0572, 0.0127, -0.2441], + ..., + [-0.1063, 0.1007, 0.0020, ..., 0.0989, -0.0239, -0.1034], + [-0.0763, -0.0724, 0.0963, ..., -0.0548, -0.1334, 0.1021], + [ 0.0046, -0.1664, 0.0964, ..., 0.0619, -0.1499, -0.0722]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 2.8238e-06, 2.2147e-06, ..., 2.9709e-07, + 0.0000e+00, 3.5614e-06], + [ 2.3283e-09, 4.6492e-06, 8.6473e-07, ..., 7.1386e-07, + 0.0000e+00, -3.1339e-07], + [ 1.8626e-09, 5.4352e-06, 1.2644e-05, ..., 2.7288e-07, + -9.3132e-10, 2.1085e-05], + ..., + [ 1.0710e-08, 1.2359e-06, 3.1758e-06, ..., -2.6561e-06, + 0.0000e+00, 5.4836e-06], + [ 6.9849e-09, 3.4105e-06, -4.3988e-05, ..., 6.1467e-08, + 0.0000e+00, -7.0572e-05], + [-5.1223e-09, 4.8913e-06, 1.8030e-06, ..., 2.5667e-06, + 0.0000e+00, 1.6317e-06]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0164, -0.0199, -0.0265, -0.0266, -0.0081, 0.0036, 0.0128, -0.0206, + -0.0058, 0.0033], device='cuda:0'), grad: tensor([ 1.0528e-05, 5.5991e-06, 3.7044e-05, -2.8038e-04, 8.8150e-07, + 2.7895e-04, 7.7253e-07, 1.5587e-05, -8.4937e-05, 1.5676e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 221.11, cls_loss 0.0024 cls_loss_mapping 0.0056 cls_loss_causal 0.5521 re_mapping 0.0066 re_causal 0.0179 /// teacc 99.00 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.1463, -0.1413, 0.0804, ..., -0.0655, 0.0430, 0.0294], + [-0.0852, -0.0330, -0.0555, ..., -0.0711, -0.0798, -0.0348], + [ 0.0317, -0.0727, -0.0494, ..., -0.0578, 0.0127, -0.2452], + ..., + [-0.1068, 0.1015, 0.0025, ..., 0.0997, -0.0239, -0.1034], + [-0.0772, -0.0734, 0.0968, ..., -0.0553, -0.1335, 0.1027], + [ 0.0041, -0.1687, 0.0964, ..., 0.0619, -0.1499, -0.0733]], + device='cuda:0'), grad: tensor([[ 4.5933e-06, 3.6741e-07, 1.7896e-05, ..., 1.1340e-05, + 0.0000e+00, 7.1898e-06], + [ 2.5611e-08, -1.6158e-07, 7.5949e-07, ..., 1.7947e-06, + 0.0000e+00, -1.2875e-05], + [ 2.0070e-07, 1.1064e-06, 9.7975e-07, ..., 2.3320e-06, + -4.6566e-10, 5.4855e-07], + ..., + [ 1.4994e-07, -4.9733e-06, 3.3733e-06, ..., 5.1931e-06, + 0.0000e+00, 1.3877e-06], + [ 5.5833e-07, 9.4017e-07, 1.2387e-06, ..., 2.1402e-06, + 4.6566e-10, 3.3658e-06], + [-1.3568e-05, 1.2778e-06, -6.6936e-05, ..., -6.0916e-05, + 0.0000e+00, -2.7176e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0168, -0.0200, -0.0269, -0.0259, -0.0081, 0.0049, 0.0120, -0.0202, + -0.0055, 0.0025], device='cuda:0'), grad: tensor([ 4.8548e-05, -3.4213e-05, 9.7007e-06, 8.3959e-07, 6.9916e-05, + -2.3559e-05, 2.4229e-05, 2.1845e-05, 1.9372e-05, -1.3673e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 131---------------------------------------------------- +epoch 131, time 221.44, cls_loss 0.0022 cls_loss_mapping 0.0061 cls_loss_causal 0.5876 re_mapping 0.0070 re_causal 0.0194 /// teacc 99.06 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.1459, -0.1404, 0.0813, ..., -0.0659, 0.0430, 0.0296], + [-0.0860, -0.0331, -0.0559, ..., -0.0715, -0.0819, -0.0356], + [ 0.0314, -0.0731, -0.0504, ..., -0.0582, 0.0128, -0.2465], + ..., + [-0.1076, 0.1018, 0.0027, ..., 0.1002, -0.0237, -0.1038], + [-0.0776, -0.0730, 0.0981, ..., -0.0557, -0.1335, 0.1042], + [ 0.0022, -0.1716, 0.0944, ..., 0.0616, -0.1500, -0.0746]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, -1.1362e-06, -4.0010e-06, ..., 1.0943e-07, + 0.0000e+00, -3.7570e-06], + [ 2.7940e-09, 3.4645e-07, 5.4622e-07, ..., 8.8476e-08, + 0.0000e+00, -9.3132e-08], + [ 4.6566e-10, 4.7917e-07, 5.6112e-07, ..., -5.2620e-08, + -9.3132e-10, 6.3237e-07], + ..., + [ 1.7229e-08, -2.6030e-07, 5.7556e-07, ..., 1.1595e-07, + 0.0000e+00, 4.5495e-07], + [ 5.5879e-09, 2.4959e-07, 2.8405e-07, ..., 3.8790e-07, + 0.0000e+00, -4.9546e-07], + [-1.8952e-07, 4.8196e-07, -2.4587e-06, ..., 3.2466e-06, + 0.0000e+00, 3.7719e-07]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0159, -0.0204, -0.0272, -0.0242, -0.0069, 0.0041, 0.0122, -0.0202, + -0.0042, 0.0003], device='cuda:0'), grad: tensor([-8.0541e-06, -8.6520e-07, 1.4007e-06, 2.1923e-06, -2.0474e-05, + 3.0342e-06, -7.6788e-07, 2.2538e-06, 1.2536e-06, 2.0042e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 220.42, cls_loss 0.0021 cls_loss_mapping 0.0059 cls_loss_causal 0.5416 re_mapping 0.0064 re_causal 0.0173 /// teacc 98.90 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.1462, -0.1409, 0.0816, ..., -0.0663, 0.0430, 0.0300], + [-0.0867, -0.0329, -0.0556, ..., -0.0721, -0.0825, -0.0354], + [ 0.0309, -0.0731, -0.0511, ..., -0.0586, 0.0132, -0.2478], + ..., + [-0.1077, 0.1020, 0.0026, ..., 0.1007, -0.0233, -0.1046], + [-0.0787, -0.0731, 0.0986, ..., -0.0560, -0.1337, 0.1046], + [ 0.0021, -0.1721, 0.0946, ..., 0.0613, -0.1502, -0.0750]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 2.3004e-07, 5.9837e-07, ..., 9.0944e-07, + 1.3970e-09, 8.5728e-07], + [ 1.3039e-08, 5.3644e-07, 2.0023e-07, ..., 7.2969e-07, + -1.9558e-08, -1.4314e-06], + [ 2.7008e-08, 2.5388e-06, 3.6154e-06, ..., -4.5123e-07, + 3.7253e-09, 1.2638e-06], + ..., + [ 1.9558e-08, 3.0100e-06, 1.5832e-07, ..., 9.5228e-07, + 4.1910e-09, 6.3889e-07], + [ 4.6100e-08, 7.5297e-07, 5.3644e-06, ..., 4.0717e-06, + 3.7253e-09, 2.1867e-06], + [ 1.7509e-07, 5.4715e-07, -1.2748e-05, ..., -1.6028e-06, + 9.3132e-10, -3.0287e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([-1.5717e-02, -2.0013e-02, -2.7000e-02, -2.4380e-02, -6.2750e-03, + 4.3196e-03, 1.1450e-02, -2.0538e-02, -3.7213e-03, -8.6620e-05], + device='cuda:0'), grad: tensor([ 5.7071e-06, -4.2701e-07, 1.6596e-06, -1.9744e-05, -1.9029e-05, + 1.2740e-05, -5.3160e-06, 1.2703e-05, 2.0817e-05, -9.1568e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 220.73, cls_loss 0.0023 cls_loss_mapping 0.0052 cls_loss_causal 0.5491 re_mapping 0.0063 re_causal 0.0178 /// teacc 98.97 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.1465, -0.1417, 0.0815, ..., -0.0667, 0.0433, 0.0301], + [-0.0868, -0.0336, -0.0563, ..., -0.0731, -0.0832, -0.0364], + [ 0.0318, -0.0752, -0.0527, ..., -0.0600, 0.0140, -0.2490], + ..., + [-0.1080, 0.1041, 0.0040, ..., 0.1026, -0.0225, -0.1036], + [-0.0788, -0.0751, 0.0986, ..., -0.0573, -0.1342, 0.1050], + [ 0.0023, -0.1735, 0.0949, ..., 0.0612, -0.1507, -0.0758]], + device='cuda:0'), grad: tensor([[ 4.4238e-09, 2.0233e-07, 1.0058e-06, ..., 2.4773e-07, + 3.9139e-07, 3.2168e-06], + [ 4.4238e-09, 1.0291e-07, 1.1618e-07, ..., 8.7311e-08, + -5.4250e-08, -1.4831e-07], + [ 9.3132e-10, -2.1812e-06, 4.3749e-07, ..., -1.4938e-06, + 2.1630e-07, 1.5264e-06], + ..., + [ 8.3819e-09, 1.3001e-06, 9.4995e-08, ..., 9.7137e-07, + 5.6578e-08, 2.8964e-07], + [ 1.6228e-07, -2.4983e-07, -2.9076e-06, ..., -3.6415e-07, + -1.3434e-07, -4.8615e-06], + [-2.9337e-08, 1.0803e-07, -1.6540e-06, ..., -1.1744e-06, + 4.2375e-08, 7.7393e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0160, -0.0204, -0.0276, -0.0248, -0.0073, 0.0054, 0.0108, -0.0191, + -0.0039, -0.0002], device='cuda:0'), grad: tensor([ 1.1310e-05, 6.4261e-08, -1.0446e-05, 6.2585e-06, 2.4289e-05, + 1.4799e-06, -3.2395e-05, 1.1675e-05, -1.0476e-05, -1.8254e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 220.67, cls_loss 0.0024 cls_loss_mapping 0.0056 cls_loss_causal 0.5407 re_mapping 0.0064 re_causal 0.0177 /// teacc 98.97 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.1467, -0.1420, 0.0815, ..., -0.0671, 0.0433, 0.0301], + [-0.0874, -0.0355, -0.0584, ..., -0.0734, -0.0835, -0.0368], + [ 0.0314, -0.0755, -0.0534, ..., -0.0603, 0.0142, -0.2500], + ..., + [-0.1086, 0.1063, 0.0060, ..., 0.1032, -0.0227, -0.1017], + [-0.0799, -0.0755, 0.0984, ..., -0.0580, -0.1343, 0.1050], + [ 0.0030, -0.1743, 0.0956, ..., 0.0620, -0.1510, -0.0757]], + device='cuda:0'), grad: tensor([[ 1.4529e-07, 2.4145e-07, 6.3796e-07, ..., 7.3295e-07, + 0.0000e+00, 5.7975e-08], + [ 1.7965e-06, 6.5472e-07, 9.2015e-06, ..., 1.6913e-05, + 0.0000e+00, -4.1351e-07], + [-1.1874e-07, 1.7667e-06, -1.8068e-07, ..., 1.5739e-06, + -1.1642e-09, -3.0734e-08], + ..., + [ 6.4783e-06, -5.0701e-06, 3.2037e-05, ..., 2.2113e-05, + 4.6566e-10, 3.0291e-07], + [ 4.6985e-07, 3.1246e-07, 2.5202e-06, ..., 3.7905e-06, + 2.3283e-10, 2.4540e-07], + [-9.2685e-06, -9.1083e-07, -4.8965e-05, ..., -1.8314e-05, + 0.0000e+00, -2.1560e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0163, -0.0216, -0.0278, -0.0254, -0.0084, 0.0057, 0.0111, -0.0172, + -0.0046, 0.0004], device='cuda:0'), grad: tensor([ 3.3528e-06, 5.3912e-05, -4.7944e-06, 8.6278e-06, -8.5950e-05, + 4.1127e-06, 1.3774e-06, 9.3877e-05, 1.4573e-05, -8.8930e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 220.55, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5660 re_mapping 0.0060 re_causal 0.0177 /// teacc 99.01 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.1471, -0.1424, 0.0818, ..., -0.0673, 0.0434, 0.0304], + [-0.0877, -0.0352, -0.0581, ..., -0.0739, -0.0836, -0.0363], + [ 0.0315, -0.0753, -0.0539, ..., -0.0590, 0.0143, -0.2507], + ..., + [-0.1090, 0.1063, 0.0058, ..., 0.1032, -0.0229, -0.1021], + [-0.0803, -0.0762, 0.0987, ..., -0.0584, -0.1343, 0.1055], + [ 0.0033, -0.1749, 0.0959, ..., 0.0614, -0.1514, -0.0763]], + device='cuda:0'), grad: tensor([[ 7.0082e-08, 1.6764e-08, 6.5193e-09, ..., 8.3819e-09, + 2.4028e-06, 3.5539e-06], + [ 1.5832e-08, 1.3807e-07, 4.6566e-08, ..., 8.7544e-08, + 1.7742e-07, -4.9081e-07], + [ 1.0943e-08, 5.1921e-08, 1.2806e-08, ..., 2.2585e-08, + 3.2131e-07, 5.1688e-07], + ..., + [ 6.7754e-08, -3.6834e-07, -1.2456e-07, ..., -2.9383e-07, + 3.8650e-08, 6.6403e-07], + [ 2.4363e-06, 2.2305e-07, 7.0781e-08, ..., 3.7020e-08, + 3.4049e-06, 9.3505e-06], + [ 2.0140e-07, 1.6321e-07, -2.3306e-07, ..., 1.3970e-09, + 7.8417e-07, 1.6401e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0159, -0.0208, -0.0279, -0.0255, -0.0077, 0.0056, 0.0108, -0.0178, + -0.0043, -0.0001], device='cuda:0'), grad: tensor([ 2.0191e-05, -1.3532e-06, 3.0957e-06, 1.3568e-05, 3.9749e-06, + 3.5353e-06, -8.6725e-05, 2.0638e-06, 3.4332e-05, 7.2829e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 220.22, cls_loss 0.0020 cls_loss_mapping 0.0059 cls_loss_causal 0.5620 re_mapping 0.0060 re_causal 0.0180 /// teacc 98.98 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.1478, -0.1426, 0.0818, ..., -0.0675, 0.0433, 0.0301], + [-0.0885, -0.0353, -0.0583, ..., -0.0743, -0.0834, -0.0363], + [ 0.0315, -0.0754, -0.0540, ..., -0.0592, 0.0144, -0.2513], + ..., + [-0.1096, 0.1060, 0.0052, ..., 0.1036, -0.0227, -0.1023], + [-0.0829, -0.0766, 0.0992, ..., -0.0589, -0.1346, 0.1050], + [ 0.0037, -0.1752, 0.0962, ..., 0.0613, -0.1518, -0.0763]], + device='cuda:0'), grad: tensor([[ 3.3062e-08, 2.4680e-07, -9.5181e-07, ..., 2.7940e-07, + 9.7789e-09, -9.7090e-08], + [ 4.8894e-09, -8.6240e-07, 7.5437e-08, ..., 5.7276e-07, + -3.7951e-08, -1.0341e-05], + [ 4.4238e-09, 1.5143e-06, 2.0047e-07, ..., 1.0580e-06, + -1.8766e-07, 5.6392e-07], + ..., + [ 1.9558e-08, -1.9576e-06, 3.1060e-07, ..., -2.1271e-06, + 2.5146e-08, 4.3847e-06], + [ 2.0000e-07, 2.7567e-07, 8.5449e-08, ..., 1.4203e-07, + 1.5181e-07, 2.2687e-06], + [ 4.3074e-08, 4.2631e-07, -5.3179e-07, ..., -6.1281e-07, + 2.7940e-09, 1.3635e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([-1.6052e-02, -2.1034e-02, -2.7399e-02, -2.5140e-02, -7.1120e-03, + 6.0711e-03, 1.0604e-02, -1.8083e-02, -5.0000e-03, -7.7302e-05], + device='cuda:0'), grad: tensor([-1.2033e-06, -4.0114e-05, 2.1365e-06, 2.4773e-06, 2.8964e-06, + 5.3085e-07, 2.8647e-06, 1.5132e-05, 9.7752e-06, 5.5395e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 220.39, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5559 re_mapping 0.0061 re_causal 0.0178 /// teacc 98.95 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.1479, -0.1430, 0.0819, ..., -0.0677, 0.0433, 0.0300], + [-0.0888, -0.0353, -0.0584, ..., -0.0747, -0.0833, -0.0359], + [ 0.0314, -0.0756, -0.0542, ..., -0.0593, 0.0145, -0.2519], + ..., + [-0.1103, 0.1063, 0.0053, ..., 0.1042, -0.0227, -0.1024], + [-0.0832, -0.0768, 0.0994, ..., -0.0596, -0.1349, 0.1053], + [ 0.0040, -0.1757, 0.0964, ..., 0.0612, -0.1520, -0.0765]], + device='cuda:0'), grad: tensor([[ 3.0035e-08, 3.5390e-08, -1.2098e-06, ..., 5.6112e-08, + 0.0000e+00, -4.7721e-06], + [ 1.7695e-08, 9.4762e-08, 5.6345e-08, ..., 9.4762e-08, + 0.0000e+00, -1.6997e-07], + [ 8.6147e-09, 1.1967e-07, 3.8720e-07, ..., 8.1491e-08, + 0.0000e+00, 6.0257e-07], + ..., + [ 7.2410e-08, -4.9081e-07, 2.6077e-07, ..., -1.4040e-07, + 0.0000e+00, 5.6997e-07], + [ 1.7649e-07, -6.1467e-08, -7.2364e-07, ..., 1.0198e-07, + 0.0000e+00, -5.0617e-07], + [-2.6869e-07, 1.7113e-07, -6.5984e-07, ..., -6.6496e-07, + 0.0000e+00, 5.0664e-07]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0163, -0.0210, -0.0270, -0.0253, -0.0068, 0.0058, 0.0107, -0.0180, + -0.0050, -0.0003], device='cuda:0'), grad: tensor([-1.5691e-05, -2.4028e-07, 2.2892e-06, 4.9695e-06, -9.2108e-07, + -8.1165e-07, 8.3596e-06, 2.1122e-06, -1.5320e-07, 4.6566e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 220.44, cls_loss 0.0018 cls_loss_mapping 0.0051 cls_loss_causal 0.5168 re_mapping 0.0063 re_causal 0.0178 /// teacc 98.98 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1481, -0.1432, 0.0820, ..., -0.0680, 0.0433, 0.0305], + [-0.0885, -0.0334, -0.0586, ..., -0.0756, -0.0838, -0.0357], + [ 0.0311, -0.0761, -0.0553, ..., -0.0596, 0.0145, -0.2537], + ..., + [-0.1107, 0.1048, 0.0054, ..., 0.1048, -0.0226, -0.1025], + [-0.0835, -0.0759, 0.0999, ..., -0.0600, -0.1350, 0.1056], + [ 0.0045, -0.1763, 0.0968, ..., 0.0613, -0.1522, -0.0763]], + device='cuda:0'), grad: tensor([[ 3.2387e-07, 3.4808e-07, -4.2794e-07, ..., 9.9419e-08, + 6.2864e-09, 7.1526e-07], + [-3.0771e-06, 3.9786e-06, 1.5134e-07, ..., 3.1311e-06, + 6.9849e-10, -7.1302e-06], + [ 6.6357e-08, 2.5816e-06, 4.4820e-07, ..., -2.3516e-08, + -1.1642e-08, 2.8312e-07], + ..., + [ 2.8801e-07, -2.8629e-06, 1.8161e-07, ..., -3.2391e-06, + 3.9581e-09, 2.0838e-07], + [ 4.5635e-07, 2.1234e-06, 3.9116e-07, ..., 1.9209e-07, + 1.8626e-09, 8.8196e-07], + [-1.2722e-06, 6.3796e-07, -2.0657e-06, ..., -1.4137e-06, + 2.3283e-10, 1.5739e-07]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0161, -0.0194, -0.0275, -0.0260, -0.0068, 0.0060, 0.0112, -0.0193, + -0.0048, -0.0003], device='cuda:0'), grad: tensor([ 3.8482e-06, -5.5969e-05, 3.9607e-05, -3.6925e-05, 1.1854e-05, + 2.5257e-05, 7.2159e-06, -1.6354e-06, 8.4266e-06, -1.7677e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 220.35, cls_loss 0.0020 cls_loss_mapping 0.0052 cls_loss_causal 0.5619 re_mapping 0.0057 re_causal 0.0173 /// teacc 98.92 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.1484, -0.1436, 0.0820, ..., -0.0684, 0.0433, 0.0306], + [-0.0890, -0.0336, -0.0588, ..., -0.0763, -0.0837, -0.0356], + [ 0.0309, -0.0754, -0.0557, ..., -0.0596, 0.0147, -0.2545], + ..., + [-0.1112, 0.1050, 0.0054, ..., 0.1051, -0.0225, -0.1025], + [-0.0833, -0.0763, 0.1004, ..., -0.0611, -0.1351, 0.1063], + [ 0.0051, -0.1767, 0.0976, ..., 0.0622, -0.1523, -0.0749]], + device='cuda:0'), grad: tensor([[ 4.8894e-09, 1.0049e-06, -2.1211e-07, ..., 6.7102e-07, + 1.0245e-08, -1.0291e-07], + [ 6.5193e-09, 1.0841e-05, 4.7265e-08, ..., 7.4282e-06, + -1.4435e-08, -1.1194e-06], + [ 6.9849e-10, 4.0770e-05, 2.1886e-07, ..., 2.8208e-05, + -1.3853e-07, 3.5926e-07], + ..., + [ 5.3551e-09, -6.1393e-05, 7.4506e-08, ..., -4.2617e-05, + 8.8476e-09, 3.4249e-07], + [ 2.4214e-08, 1.2554e-06, 5.0012e-07, ..., 1.1120e-06, + 3.1199e-08, 1.1176e-07], + [-1.0524e-07, 4.7125e-06, -1.0896e-06, ..., 2.4997e-06, + 3.4925e-09, 3.6787e-08]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0161, -0.0197, -0.0266, -0.0258, -0.0074, 0.0048, 0.0109, -0.0193, + -0.0044, 0.0003], device='cuda:0'), grad: tensor([ 2.3413e-06, 1.9997e-05, 8.8215e-05, 5.7034e-06, 2.5891e-06, + 1.3337e-06, 4.1979e-07, -1.3518e-04, 5.2452e-06, 9.1866e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 220.56, cls_loss 0.0018 cls_loss_mapping 0.0042 cls_loss_causal 0.5496 re_mapping 0.0060 re_causal 0.0174 /// teacc 98.97 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.1488, -0.1438, 0.0823, ..., -0.0685, 0.0432, 0.0307], + [-0.0892, -0.0338, -0.0593, ..., -0.0769, -0.0836, -0.0354], + [ 0.0308, -0.0758, -0.0563, ..., -0.0601, 0.0146, -0.2555], + ..., + [-0.1118, 0.1054, 0.0052, ..., 0.1055, -0.0223, -0.1028], + [-0.0836, -0.0765, 0.1012, ..., -0.0616, -0.1352, 0.1068], + [ 0.0054, -0.1770, 0.0980, ..., 0.0622, -0.1524, -0.0754]], + device='cuda:0'), grad: tensor([[ 8.6846e-08, 2.9919e-07, -4.2003e-07, ..., 7.9861e-08, + 3.0035e-08, -5.3830e-07], + [ 5.9139e-08, 3.7788e-07, 1.7486e-07, ..., 6.6776e-07, + 4.6566e-09, -2.0415e-06], + [ 6.0303e-08, 6.7614e-07, 1.6554e-07, ..., 4.6100e-08, + 5.1921e-08, 1.7229e-06], + ..., + [ 1.0966e-07, 1.2806e-08, 4.1444e-07, ..., 3.3900e-07, + 9.0804e-09, 1.5786e-07], + [ 5.0105e-07, 4.7660e-07, 3.4412e-07, ..., 8.3726e-07, + 8.8476e-09, 1.7649e-07], + [-1.8626e-09, 2.0070e-07, -1.8701e-06, ..., -1.1660e-06, + 1.2806e-08, 2.9500e-07]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0160, -0.0197, -0.0264, -0.0260, -0.0072, 0.0046, 0.0110, -0.0193, + -0.0043, 0.0004], device='cuda:0'), grad: tensor([-1.2387e-06, -4.2394e-06, 3.8818e-06, -1.9401e-05, -3.5502e-06, + 2.5600e-05, -7.0259e-06, 3.0994e-06, 3.5334e-06, -6.5099e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 220.51, cls_loss 0.0022 cls_loss_mapping 0.0053 cls_loss_causal 0.5694 re_mapping 0.0057 re_causal 0.0169 /// teacc 99.00 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1484, -0.1441, 0.0825, ..., -0.0693, 0.0432, 0.0315], + [-0.0892, -0.0338, -0.0596, ..., -0.0773, -0.0830, -0.0351], + [ 0.0305, -0.0760, -0.0572, ..., -0.0605, 0.0147, -0.2572], + ..., + [-0.1129, 0.1056, 0.0052, ..., 0.1060, -0.0225, -0.1030], + [-0.0842, -0.0769, 0.1018, ..., -0.0621, -0.1354, 0.1072], + [ 0.0059, -0.1775, 0.0986, ..., 0.0624, -0.1528, -0.0759]], + device='cuda:0'), grad: tensor([[ 4.5449e-06, 5.8208e-08, -1.7986e-05, ..., 1.3039e-08, + 3.0757e-07, -3.5278e-06], + [ 9.6299e-07, 3.1758e-07, 2.8941e-07, ..., 1.6554e-07, + 1.3760e-07, 1.3057e-06], + [ 6.8806e-06, 1.7323e-07, 3.3923e-07, ..., 2.6776e-08, + 4.4191e-07, 9.1940e-06], + ..., + [ 8.0653e-07, 3.7951e-08, 2.7427e-07, ..., 5.9139e-08, + 1.7323e-07, 1.1912e-06], + [ 6.6273e-06, 4.7963e-07, -4.4703e-07, ..., 5.7276e-08, + 3.4738e-07, 7.1190e-06], + [ 3.9418e-07, 2.3586e-07, 1.6645e-05, ..., -9.9884e-08, + 5.9372e-08, 9.6187e-06]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0155, -0.0186, -0.0284, -0.0257, -0.0074, 0.0039, 0.0104, -0.0193, + -0.0042, 0.0006], device='cuda:0'), grad: tensor([-2.9385e-05, 6.9253e-06, 1.8850e-05, 1.3113e-05, 7.6219e-06, + 5.0992e-05, -1.4865e-04, 4.0233e-06, 2.8133e-05, 4.8429e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 142---------------------------------------------------- +epoch 142, time 221.18, cls_loss 0.0023 cls_loss_mapping 0.0047 cls_loss_causal 0.5552 re_mapping 0.0061 re_causal 0.0165 /// teacc 99.08 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1486, -0.1444, 0.0836, ..., -0.0695, 0.0435, 0.0335], + [-0.0907, -0.0339, -0.0598, ..., -0.0777, -0.0836, -0.0350], + [ 0.0307, -0.0762, -0.0579, ..., -0.0608, 0.0153, -0.2580], + ..., + [-0.1140, 0.1058, 0.0052, ..., 0.1062, -0.0223, -0.1033], + [-0.0847, -0.0778, 0.1021, ..., -0.0624, -0.1357, 0.1077], + [ 0.0061, -0.1776, 0.0988, ..., 0.0627, -0.1536, -0.0768]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 8.9640e-08, -2.4051e-07, ..., 4.5868e-08, + 1.3504e-08, -4.4773e-07], + [ 9.3132e-10, 2.4796e-07, 1.4366e-07, ..., 9.8022e-08, + 9.3132e-09, -8.0280e-07], + [ 2.3283e-10, 1.8510e-07, 1.1316e-07, ..., 6.1700e-08, + -5.8440e-08, 7.4971e-08], + ..., + [ 7.4506e-09, -8.2422e-08, 1.4878e-07, ..., 1.0454e-07, + 1.2573e-08, 2.4354e-07], + [ 4.1910e-09, 1.0012e-06, 3.5088e-07, ..., 3.3528e-08, + 2.5611e-09, -5.6811e-08], + [-9.3132e-09, 1.5059e-06, 3.4343e-07, ..., -2.2002e-07, + 3.2596e-09, 5.3784e-08]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0133, -0.0182, -0.0286, -0.0243, -0.0076, 0.0024, 0.0089, -0.0195, + -0.0043, 0.0005], device='cuda:0'), grad: tensor([-1.9115e-07, -1.6093e-05, -8.3167e-07, -6.1207e-06, 1.2666e-05, + 1.4547e-06, -3.5251e-07, 4.0904e-06, 2.3544e-06, 3.0380e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 220.33, cls_loss 0.0020 cls_loss_mapping 0.0051 cls_loss_causal 0.5599 re_mapping 0.0061 re_causal 0.0173 /// teacc 98.94 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1490, -0.1447, 0.0836, ..., -0.0698, 0.0434, 0.0327], + [-0.0919, -0.0336, -0.0596, ..., -0.0789, -0.0838, -0.0346], + [ 0.0307, -0.0766, -0.0584, ..., -0.0612, 0.0155, -0.2588], + ..., + [-0.1148, 0.1062, 0.0047, ..., 0.1060, -0.0224, -0.1036], + [-0.0851, -0.0784, 0.1022, ..., -0.0631, -0.1358, 0.1079], + [ 0.0063, -0.1787, 0.0996, ..., 0.0632, -0.1540, -0.0769]], + device='cuda:0'), grad: tensor([[-5.3085e-08, 6.0769e-07, -9.1851e-05, ..., -9.5591e-06, + 6.9849e-09, -1.9923e-05], + [ 8.3819e-09, 2.8722e-06, 1.6559e-06, ..., 1.4286e-06, + 6.0536e-09, 3.5670e-07], + [ 1.1642e-08, 1.4370e-06, 2.2948e-06, ..., 2.0629e-07, + -3.6787e-08, 1.3607e-06], + ..., + [ 1.6298e-08, -1.8269e-05, -6.9886e-06, ..., -1.3068e-05, + 6.0536e-09, 3.6834e-07], + [ 1.2107e-07, 4.0559e-07, 4.0233e-06, ..., 1.5181e-07, + 3.2596e-09, 3.7495e-06], + [ 7.5903e-08, 6.7502e-06, 8.3506e-05, ..., 1.3299e-05, + 3.2596e-09, 1.2971e-05]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0143, -0.0177, -0.0286, -0.0249, -0.0075, 0.0027, 0.0095, -0.0199, + -0.0046, 0.0008], device='cuda:0'), grad: tensor([-1.8263e-04, 6.3144e-06, 7.5623e-06, 9.8944e-06, 2.9113e-06, + 6.7167e-06, -6.3553e-06, -2.8014e-05, 1.1906e-05, 1.7190e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 220.51, cls_loss 0.0018 cls_loss_mapping 0.0045 cls_loss_causal 0.5484 re_mapping 0.0064 re_causal 0.0172 /// teacc 98.94 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1493, -0.1450, 0.0839, ..., -0.0700, 0.0434, 0.0328], + [-0.0935, -0.0332, -0.0590, ..., -0.0794, -0.0839, -0.0341], + [ 0.0304, -0.0771, -0.0590, ..., -0.0618, 0.0155, -0.2597], + ..., + [-0.1170, 0.1063, 0.0043, ..., 0.1065, -0.0225, -0.1045], + [-0.0856, -0.0785, 0.1025, ..., -0.0633, -0.1359, 0.1082], + [ 0.0072, -0.1797, 0.1001, ..., 0.0632, -0.1542, -0.0770]], + device='cuda:0'), grad: tensor([[ 1.0710e-08, 7.8185e-07, 3.5577e-07, ..., 6.2445e-07, + 4.6566e-10, 2.1840e-07], + [ 3.2596e-09, 1.2191e-06, 1.0617e-06, ..., 1.0058e-06, + 0.0000e+00, 1.7267e-06], + [ 2.7940e-09, 1.2526e-07, 1.0394e-06, ..., 6.0536e-08, + -7.4506e-09, 3.1926e-06], + ..., + [ 5.1223e-09, -3.5428e-06, -1.5981e-06, ..., -2.9095e-06, + 4.1910e-09, -4.4517e-07], + [ 2.7008e-08, 8.2888e-08, -1.9670e-06, ..., 4.3306e-08, + 1.3970e-09, -6.0275e-06], + [ 5.1223e-09, 7.8743e-07, 3.6415e-07, ..., 6.3982e-07, + 0.0000e+00, 1.8114e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0143, -0.0173, -0.0287, -0.0251, -0.0074, 0.0026, 0.0099, -0.0204, + -0.0044, 0.0009], device='cuda:0'), grad: tensor([ 1.7118e-06, 5.2936e-06, 7.0184e-06, 1.9837e-06, 3.8929e-07, + 8.5309e-07, -3.4738e-07, -5.8711e-06, -1.2733e-05, 1.6540e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 220.16, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5361 re_mapping 0.0061 re_causal 0.0176 /// teacc 99.04 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.1492, -0.1454, 0.0840, ..., -0.0704, 0.0435, 0.0333], + [-0.0935, -0.0333, -0.0593, ..., -0.0800, -0.0841, -0.0338], + [ 0.0303, -0.0776, -0.0595, ..., -0.0625, 0.0158, -0.2602], + ..., + [-0.1183, 0.1068, 0.0043, ..., 0.1070, -0.0224, -0.1046], + [-0.0863, -0.0789, 0.1021, ..., -0.0642, -0.1360, 0.1082], + [ 0.0077, -0.1802, 0.1007, ..., 0.0633, -0.1545, -0.0770]], + device='cuda:0'), grad: tensor([[ 3.0780e-07, 1.1874e-07, 1.4575e-07, ..., 9.4902e-07, + 0.0000e+00, 1.8766e-07], + [ 2.3469e-06, 7.5623e-07, 5.6578e-07, ..., 7.5549e-06, + 0.0000e+00, 2.1840e-07], + [ 3.6834e-07, 1.3895e-06, 5.3085e-07, ..., 2.7996e-06, + 0.0000e+00, 1.1362e-07], + ..., + [ 8.8215e-06, -3.6210e-06, -1.9372e-07, ..., 2.0310e-05, + 0.0000e+00, 1.3085e-07], + [ 3.4319e-07, 3.3993e-08, -8.2003e-07, ..., 9.2294e-07, + 0.0000e+00, -2.6524e-06], + [ 1.8311e-04, 9.4157e-07, 1.8403e-05, ..., 5.1737e-04, + 0.0000e+00, 1.3411e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0141, -0.0173, -0.0288, -0.0251, -0.0073, 0.0024, 0.0097, -0.0202, + -0.0048, 0.0011], device='cuda:0'), grad: tensor([ 2.9057e-06, 1.9938e-05, 7.3835e-06, 1.5832e-06, -1.4210e-03, + 2.0768e-06, 3.5018e-06, 5.3018e-05, -2.0731e-06, 1.3332e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 220.19, cls_loss 0.0021 cls_loss_mapping 0.0047 cls_loss_causal 0.5633 re_mapping 0.0056 re_causal 0.0161 /// teacc 99.02 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.1511, -0.1460, 0.0841, ..., -0.0708, 0.0435, 0.0329], + [-0.0925, -0.0338, -0.0599, ..., -0.0818, -0.0841, -0.0339], + [ 0.0298, -0.0782, -0.0603, ..., -0.0633, 0.0159, -0.2619], + ..., + [-0.1189, 0.1077, 0.0045, ..., 0.1081, -0.0225, -0.1045], + [-0.0874, -0.0793, 0.1025, ..., -0.0648, -0.1362, 0.1085], + [ 0.0076, -0.1812, 0.1013, ..., 0.0629, -0.1547, -0.0776]], + device='cuda:0'), grad: tensor([[ 2.2817e-07, 1.1083e-06, 4.8801e-07, ..., 2.4680e-07, + -8.3819e-09, 7.7020e-07], + [ 1.8626e-07, 2.1920e-05, 2.2016e-06, ..., 3.2969e-06, + 9.3132e-10, 2.4252e-06], + [ 3.2131e-08, 5.8711e-06, 1.0803e-06, ..., 1.1874e-06, + 4.6566e-10, 6.0583e-07], + ..., + [ 1.2759e-07, -6.0320e-05, -2.5444e-06, ..., -1.0878e-05, + 4.6566e-10, 5.2527e-07], + [ 1.4901e-06, 1.4175e-06, -5.3495e-06, ..., 1.4948e-07, + 4.6566e-10, -5.7369e-06], + [ 1.1874e-07, 8.0094e-06, 1.6484e-06, ..., 1.7565e-06, + 1.8626e-09, 1.4650e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0147, -0.0176, -0.0287, -0.0255, -0.0071, 0.0030, 0.0095, -0.0197, + -0.0051, 0.0007], device='cuda:0'), grad: tensor([ 1.1586e-05, 4.5478e-05, -1.0423e-05, 3.0845e-05, 8.0690e-06, + -1.3575e-05, 2.0996e-05, -1.0127e-04, -9.5367e-06, 1.7703e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 220.29, cls_loss 0.0018 cls_loss_mapping 0.0046 cls_loss_causal 0.5142 re_mapping 0.0059 re_causal 0.0163 /// teacc 99.02 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1518, -0.1464, 0.0845, ..., -0.0710, 0.0438, 0.0326], + [-0.0910, -0.0340, -0.0603, ..., -0.0825, -0.0845, -0.0333], + [ 0.0308, -0.0785, -0.0605, ..., -0.0639, 0.0159, -0.2603], + ..., + [-0.1195, 0.1079, 0.0042, ..., 0.1084, -0.0224, -0.1047], + [-0.0881, -0.0796, 0.1026, ..., -0.0654, -0.1363, 0.1083], + [ 0.0085, -0.1816, 0.1024, ..., 0.0634, -0.1552, -0.0768]], + device='cuda:0'), grad: tensor([[ 2.1933e-07, 2.0955e-08, -1.8626e-09, ..., 4.2841e-08, + 0.0000e+00, 1.9511e-07], + [ 5.9139e-07, 3.1712e-07, 4.1956e-07, ..., 3.0082e-07, + 0.0000e+00, 9.9186e-07], + [ 1.0012e-07, 2.6263e-07, 7.3109e-08, ..., -1.7462e-07, + 0.0000e+00, 1.3597e-07], + ..., + [ 4.5542e-07, -3.9348e-07, 3.9395e-07, ..., -1.4901e-08, + 0.0000e+00, 5.3365e-07], + [ 1.6205e-06, 2.8359e-07, 5.3458e-07, ..., 6.4215e-07, + 0.0000e+00, 6.6310e-07], + [ 3.0287e-06, 2.4959e-07, -5.5097e-06, ..., -2.5257e-06, + 0.0000e+00, 3.4589e-06]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0153, -0.0176, -0.0278, -0.0251, -0.0080, 0.0026, 0.0096, -0.0197, + -0.0060, 0.0014], device='cuda:0'), grad: tensor([ 6.4820e-07, 2.8387e-06, -2.0899e-06, 7.2010e-06, 4.8056e-06, + -1.8701e-05, 7.9861e-07, 2.4028e-06, 4.7348e-06, -2.6729e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 220.19, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5536 re_mapping 0.0058 re_causal 0.0165 /// teacc 98.97 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1520, -0.1466, 0.0847, ..., -0.0713, 0.0452, 0.0330], + [-0.0904, -0.0346, -0.0622, ..., -0.0831, -0.0845, -0.0341], + [ 0.0310, -0.0790, -0.0612, ..., -0.0649, 0.0158, -0.2610], + ..., + [-0.1209, 0.1087, 0.0051, ..., 0.1090, -0.0225, -0.1042], + [-0.0885, -0.0797, 0.1033, ..., -0.0652, -0.1366, 0.1089], + [ 0.0086, -0.1823, 0.1027, ..., 0.0629, -0.1559, -0.0770]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 7.5437e-08, -1.7695e-08, ..., 4.2841e-08, + 5.1223e-09, -3.4925e-08], + [ 8.3819e-09, 3.3667e-07, 1.0105e-07, ..., 9.3598e-08, + 1.9092e-08, -3.9116e-08], + [ 1.3970e-09, 6.2305e-07, 3.6322e-07, ..., 1.0245e-07, + 1.9558e-08, 7.0315e-07], + ..., + [ 5.5414e-08, 2.3004e-07, 4.3819e-07, ..., 2.2491e-07, + 3.5390e-08, 2.8126e-07], + [ 4.4703e-08, 2.5239e-07, -5.8580e-07, ..., 1.4901e-07, + 2.6077e-08, -1.7844e-06], + [-2.8824e-07, 3.3807e-07, -2.4419e-06, ..., -1.8105e-06, + 7.9162e-09, -2.5658e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0149, -0.0184, -0.0278, -0.0247, -0.0076, 0.0017, 0.0099, -0.0189, + -0.0056, 0.0009], device='cuda:0'), grad: tensor([ 1.1642e-07, 4.9267e-07, 2.0750e-06, -1.0878e-05, 2.9895e-06, + 8.0168e-06, 2.2631e-07, 2.0079e-06, -1.1558e-06, -3.8892e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 220.13, cls_loss 0.0020 cls_loss_mapping 0.0046 cls_loss_causal 0.5281 re_mapping 0.0056 re_causal 0.0164 /// teacc 99.08 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1529, -0.1473, 0.0848, ..., -0.0715, 0.0452, 0.0330], + [-0.0906, -0.0356, -0.0637, ..., -0.0850, -0.0846, -0.0360], + [ 0.0310, -0.0797, -0.0616, ..., -0.0654, 0.0158, -0.2620], + ..., + [-0.1232, 0.1100, 0.0063, ..., 0.1101, -0.0225, -0.1026], + [-0.0892, -0.0796, 0.1039, ..., -0.0650, -0.1366, 0.1098], + [ 0.0087, -0.1830, 0.1029, ..., 0.0626, -0.1561, -0.0770]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 6.7055e-08, -1.3784e-07, ..., 4.0978e-08, + 1.3970e-09, -1.3830e-07], + [ 2.1141e-07, 2.5034e-06, 2.0824e-06, ..., 1.0571e-06, + -8.5682e-08, 4.8429e-08], + [ 7.9162e-09, -2.7986e-07, 1.1455e-07, ..., 2.3749e-08, + 2.7940e-09, 9.4064e-08], + ..., + [ 6.2399e-08, -4.7348e-06, -3.0156e-06, ..., -1.4286e-06, + 7.4040e-08, 2.1886e-07], + [ 1.9930e-07, 2.9290e-07, 1.4622e-07, ..., 2.7241e-07, + 4.6566e-10, -4.7637e-07], + [-7.4366e-07, 2.4913e-07, -1.4212e-06, ..., -9.6764e-07, + 2.3283e-09, -2.7847e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0149, -0.0192, -0.0279, -0.0252, -0.0075, 0.0023, 0.0094, -0.0179, + -0.0049, 0.0005], device='cuda:0'), grad: tensor([-1.8161e-08, 5.5842e-06, -1.0118e-05, 6.9551e-06, 8.8662e-07, + 1.8999e-07, -4.0326e-07, -6.2957e-06, 5.1409e-06, -1.9558e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 220.57, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5114 re_mapping 0.0057 re_causal 0.0162 /// teacc 98.97 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1540, -0.1478, 0.0851, ..., -0.0716, 0.0458, 0.0327], + [-0.0914, -0.0357, -0.0648, ..., -0.0854, -0.0845, -0.0373], + [ 0.0307, -0.0800, -0.0622, ..., -0.0658, 0.0158, -0.2627], + ..., + [-0.1242, 0.1101, 0.0064, ..., 0.1104, -0.0226, -0.1028], + [-0.0898, -0.0794, 0.1056, ..., -0.0653, -0.1367, 0.1113], + [ 0.0089, -0.1834, 0.1034, ..., 0.0624, -0.1566, -0.0770]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 1.4435e-08, 1.8254e-07, ..., 1.9558e-07, + 0.0000e+00, 2.0815e-07], + [ 4.6566e-09, 2.2398e-07, -8.5160e-06, ..., 2.1048e-07, + 0.0000e+00, -6.6422e-06], + [ 1.8626e-09, 1.8813e-07, 1.7742e-07, ..., 1.3597e-07, + 0.0000e+00, 1.5413e-07], + ..., + [ 1.8626e-08, -3.7206e-07, 4.4405e-06, ..., 5.9791e-07, + 0.0000e+00, 3.4012e-06], + [ 1.5832e-08, 5.8673e-08, -3.3760e-07, ..., 2.2678e-07, + 0.0000e+00, -1.0012e-06], + [-1.5367e-08, 9.4995e-08, 2.9095e-06, ..., 3.7365e-06, + 0.0000e+00, 2.8089e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0150, -0.0200, -0.0280, -0.0255, -0.0077, 0.0027, 0.0096, -0.0180, + -0.0030, 0.0008], device='cuda:0'), grad: tensor([ 2.8647e-06, -7.4625e-05, -4.6752e-07, 1.3327e-06, -1.1809e-05, + 3.3760e-07, 4.8010e-07, 4.0233e-05, 1.1846e-06, 4.0561e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 220.41, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5545 re_mapping 0.0057 re_causal 0.0165 /// teacc 98.97 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1542, -0.1483, 0.0851, ..., -0.0719, 0.0458, 0.0310], + [-0.0919, -0.0357, -0.0652, ..., -0.0860, -0.0845, -0.0376], + [ 0.0310, -0.0803, -0.0627, ..., -0.0662, 0.0162, -0.2633], + ..., + [-0.1249, 0.1101, 0.0060, ..., 0.1116, -0.0227, -0.1030], + [-0.0911, -0.0798, 0.1048, ..., -0.0657, -0.1370, 0.1114], + [ 0.0096, -0.1839, 0.1047, ..., 0.0626, -0.1567, -0.0758]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.9558e-08, -3.3993e-08, ..., 9.3132e-09, + 0.0000e+00, 4.3772e-08], + [ 0.0000e+00, 2.7381e-07, 2.2352e-08, ..., 1.0384e-07, + 0.0000e+00, -3.6787e-08], + [ 0.0000e+00, 5.6811e-08, 8.8476e-09, ..., 3.7253e-09, + -9.3132e-10, 8.2888e-08], + ..., + [ 4.6566e-10, -2.9244e-07, -4.2375e-08, ..., -2.4633e-07, + 4.6566e-10, 4.7497e-08], + [ 4.6566e-10, 2.7148e-07, 3.4459e-08, ..., 1.0710e-08, + 0.0000e+00, 1.7229e-07], + [-6.9849e-09, 4.6007e-07, 3.2596e-09, ..., 1.3784e-07, + 0.0000e+00, 4.7497e-08]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0166, -0.0200, -0.0280, -0.0253, -0.0083, 0.0026, 0.0105, -0.0179, + -0.0032, 0.0012], device='cuda:0'), grad: tensor([ 3.7206e-07, 5.7369e-07, 3.7672e-07, -3.3025e-06, 1.0312e-05, + 2.5108e-06, -1.3031e-05, -2.1979e-07, 1.0477e-06, 1.3197e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 220.43, cls_loss 0.0017 cls_loss_mapping 0.0041 cls_loss_causal 0.5196 re_mapping 0.0057 re_causal 0.0156 /// teacc 99.08 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.1550, -0.1499, 0.0854, ..., -0.0722, 0.0460, 0.0310], + [-0.0917, -0.0357, -0.0648, ..., -0.0868, -0.0844, -0.0375], + [ 0.0308, -0.0807, -0.0632, ..., -0.0667, 0.0166, -0.2637], + ..., + [-0.1257, 0.1105, 0.0057, ..., 0.1124, -0.0227, -0.1034], + [-0.0921, -0.0804, 0.1049, ..., -0.0662, -0.1374, 0.1115], + [ 0.0092, -0.1848, 0.1048, ..., 0.0625, -0.1572, -0.0763]], + device='cuda:0'), grad: tensor([[ 1.3085e-07, 1.2983e-06, -4.8662e-07, ..., 7.1200e-07, + -1.3970e-08, -9.2667e-07], + [ 8.9873e-08, 5.5246e-06, 8.1956e-07, ..., 1.9027e-06, + 9.3132e-10, 4.6566e-08], + [ 3.2596e-08, 1.7256e-05, 2.0303e-06, ..., 5.1223e-06, + 4.6566e-10, 1.3877e-07], + ..., + [ 4.6333e-07, -3.1447e-04, -3.4004e-05, ..., -8.9347e-05, + 4.6566e-10, -3.7998e-07], + [ 1.8161e-07, 8.8587e-06, 1.0077e-06, ..., 2.6189e-06, + 4.6566e-10, 1.7462e-07], + [-6.8545e-07, 2.0918e-06, -2.2929e-06, ..., -3.1348e-06, + 4.1910e-09, 4.7544e-07]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0168, -0.0198, -0.0278, -0.0254, -0.0084, 0.0033, 0.0103, -0.0180, + -0.0035, 0.0009], device='cuda:0'), grad: tensor([-1.6354e-06, 9.9167e-06, 2.1890e-05, 3.9911e-04, 2.1234e-06, + 3.5465e-06, 4.7311e-06, -4.5085e-04, 1.3858e-05, -3.3360e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 220.69, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5296 re_mapping 0.0059 re_causal 0.0163 /// teacc 98.97 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1563, -0.1501, 0.0859, ..., -0.0724, 0.0462, 0.0326], + [-0.0920, -0.0360, -0.0650, ..., -0.0875, -0.0846, -0.0372], + [ 0.0303, -0.0805, -0.0636, ..., -0.0672, 0.0168, -0.2646], + ..., + [-0.1267, 0.1110, 0.0056, ..., 0.1128, -0.0228, -0.1038], + [-0.0937, -0.0812, 0.1052, ..., -0.0669, -0.1375, 0.1121], + [ 0.0089, -0.1852, 0.1054, ..., 0.0622, -0.1577, -0.0771]], + device='cuda:0'), grad: tensor([[ 6.6590e-08, 4.0652e-07, 2.7008e-08, ..., 9.4529e-08, + 3.4459e-08, 4.6566e-08], + [ 2.3283e-08, 1.2815e-06, 2.1793e-07, ..., 5.9558e-07, + 6.8452e-08, -2.9383e-07], + [-1.1409e-07, 1.4119e-06, 9.6392e-08, ..., -4.3400e-07, + -1.9046e-07, 4.2841e-08], + ..., + [ 8.9407e-08, -1.8999e-06, -9.4529e-07, ..., -2.4885e-06, + 3.5623e-07, 2.1607e-07], + [ 9.0804e-08, 5.7742e-07, 6.1002e-08, ..., 2.6030e-07, + 8.8476e-08, 9.3598e-08], + [ 1.2945e-07, 2.1774e-06, 3.5251e-07, ..., 8.9640e-07, + 8.8476e-09, 1.2200e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0150, -0.0200, -0.0271, -0.0258, -0.0079, 0.0056, 0.0063, -0.0181, + -0.0031, 0.0004], device='cuda:0'), grad: tensor([ 8.8096e-05, 2.1607e-06, 4.9263e-05, -1.7500e-04, 1.3448e-06, + 8.6054e-06, 1.3314e-05, 1.6280e-06, 4.6678e-06, 5.8115e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 220.77, cls_loss 0.0014 cls_loss_mapping 0.0042 cls_loss_causal 0.5198 re_mapping 0.0057 re_causal 0.0163 /// teacc 98.86 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1566, -0.1506, 0.0857, ..., -0.0725, 0.0461, 0.0323], + [-0.0920, -0.0361, -0.0651, ..., -0.0878, -0.0845, -0.0372], + [ 0.0302, -0.0801, -0.0639, ..., -0.0671, 0.0171, -0.2650], + ..., + [-0.1271, 0.1111, 0.0057, ..., 0.1129, -0.0232, -0.1039], + [-0.0950, -0.0824, 0.1054, ..., -0.0673, -0.1375, 0.1122], + [ 0.0089, -0.1857, 0.1058, ..., 0.0622, -0.1578, -0.0772]], + device='cuda:0'), grad: tensor([[ 8.0373e-07, 2.3888e-07, 2.7381e-06, ..., 2.7344e-06, + -1.1176e-08, 3.3248e-07], + [ 2.8033e-07, 3.5623e-07, 1.0515e-06, ..., 1.0645e-06, + 5.5879e-09, -6.6077e-07], + [ 6.7335e-07, -2.0862e-07, 2.4289e-06, ..., 4.0904e-06, + -7.4506e-09, 1.4855e-07], + ..., + [ 2.5593e-06, -9.5926e-08, 9.0897e-06, ..., -6.0238e-06, + 3.7253e-09, 2.4633e-07], + [ 7.9023e-07, 2.4959e-07, 1.2200e-06, ..., 1.1511e-06, + 9.3132e-10, 9.4855e-07], + [-7.1637e-06, 4.8801e-07, -2.6867e-05, ..., -2.4378e-05, + 2.3283e-09, 1.4529e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0154, -0.0199, -0.0267, -0.0252, -0.0076, 0.0057, 0.0062, -0.0183, + -0.0034, 0.0001], device='cuda:0'), grad: tensor([ 1.4357e-05, 3.8669e-06, 1.0364e-05, 4.8280e-06, 5.8591e-05, + 4.6007e-06, -3.4813e-06, 8.0317e-06, 7.1526e-06, -1.0818e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 220.52, cls_loss 0.0026 cls_loss_mapping 0.0048 cls_loss_causal 0.5196 re_mapping 0.0057 re_causal 0.0152 /// teacc 99.08 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1570, -0.1512, 0.0850, ..., -0.0739, 0.0460, 0.0322], + [-0.0922, -0.0363, -0.0662, ..., -0.0890, -0.0849, -0.0372], + [ 0.0303, -0.0827, -0.0652, ..., -0.0689, 0.0168, -0.2659], + ..., + [-0.1279, 0.1119, 0.0033, ..., 0.1126, -0.0234, -0.1041], + [-0.0957, -0.0831, 0.1071, ..., -0.0667, -0.1377, 0.1127], + [ 0.0097, -0.1869, 0.1087, ..., 0.0632, -0.1584, -0.0790]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 7.1712e-08, -1.3951e-06, ..., 1.2852e-07, + 0.0000e+00, -1.5711e-06], + [ 1.3039e-08, 7.9442e-07, 3.4273e-07, ..., 3.2503e-07, + 9.3132e-10, -2.8126e-07], + [ 2.7940e-09, 5.1595e-07, 1.7602e-07, ..., 2.9802e-07, + -5.5879e-09, 1.1269e-07], + ..., + [ 5.8673e-08, -1.5749e-06, 4.8708e-07, ..., -2.0396e-07, + 2.7940e-09, 2.5053e-07], + [ 4.0978e-08, 4.0699e-07, -3.2503e-07, ..., 3.5204e-07, + 9.3132e-10, -9.9279e-07], + [-1.7975e-07, 2.7101e-07, -5.5850e-05, ..., -7.2002e-05, + 0.0000e+00, 7.1526e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0159, -0.0203, -0.0279, -0.0244, -0.0077, 0.0053, 0.0067, -0.0193, + -0.0033, 0.0021], device='cuda:0'), grad: tensor([-4.9621e-06, 1.1409e-06, -1.6075e-06, 7.1712e-07, 1.5318e-04, + 2.9989e-06, 2.2687e-06, 1.0040e-06, 4.2468e-07, -1.5497e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 220.57, cls_loss 0.0015 cls_loss_mapping 0.0032 cls_loss_causal 0.5212 re_mapping 0.0055 re_causal 0.0158 /// teacc 98.98 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.1573, -0.1516, 0.0852, ..., -0.0742, 0.0461, 0.0323], + [-0.0924, -0.0365, -0.0664, ..., -0.0894, -0.0850, -0.0372], + [ 0.0302, -0.0833, -0.0663, ..., -0.0700, 0.0170, -0.2665], + ..., + [-0.1283, 0.1125, 0.0035, ..., 0.1136, -0.0235, -0.1041], + [-0.0960, -0.0838, 0.1077, ..., -0.0669, -0.1378, 0.1129], + [ 0.0090, -0.1875, 0.1087, ..., 0.0625, -0.1587, -0.0797]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 1.5832e-08, -7.4506e-08, ..., 9.2201e-08, + -6.5193e-09, 2.0310e-05], + [ 2.2352e-08, 7.2643e-08, 1.5553e-07, ..., 1.8999e-07, + 7.4506e-09, 1.2666e-07], + [ 1.8626e-09, -1.5553e-07, 2.7008e-08, ..., 1.3039e-07, + -4.3027e-07, 7.9162e-08], + ..., + [ 5.6811e-08, 2.6822e-07, 1.1986e-06, ..., 1.2424e-06, + 4.8429e-08, 7.0781e-08], + [ 8.6613e-08, 2.3376e-07, 3.3807e-07, ..., 3.2131e-07, + 2.2631e-07, 1.6019e-07], + [-4.1537e-07, -3.3155e-07, -3.4701e-06, ..., 5.8115e-07, + 1.8626e-09, 1.2200e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0160, -0.0204, -0.0279, -0.0244, -0.0073, 0.0053, 0.0070, -0.0189, + -0.0034, 0.0015], device='cuda:0'), grad: tensor([ 4.4823e-05, 1.0058e-06, -4.2319e-06, 1.6624e-06, -9.1344e-06, + 1.5572e-05, -5.9336e-05, 4.4592e-06, 3.3639e-06, 1.8319e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 220.31, cls_loss 0.0020 cls_loss_mapping 0.0043 cls_loss_causal 0.5309 re_mapping 0.0055 re_causal 0.0152 /// teacc 99.05 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1576, -0.1520, 0.0853, ..., -0.0745, 0.0462, 0.0324], + [-0.0928, -0.0365, -0.0665, ..., -0.0901, -0.0849, -0.0368], + [ 0.0301, -0.0843, -0.0670, ..., -0.0715, 0.0171, -0.2671], + ..., + [-0.1294, 0.1132, 0.0026, ..., 0.1139, -0.0236, -0.1046], + [-0.0972, -0.0843, 0.1078, ..., -0.0686, -0.1379, 0.1129], + [ 0.0092, -0.1890, 0.1100, ..., 0.0635, -0.1589, -0.0806]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.4249e-07, -1.0710e-07, ..., 1.5832e-08, + 0.0000e+00, 1.3784e-07], + [ 1.9651e-07, 8.4843e-07, 2.2631e-07, ..., 4.2003e-07, + 0.0000e+00, 4.1910e-07], + [ 6.5193e-09, 1.7416e-07, 9.7789e-08, ..., 5.8673e-08, + -1.3039e-08, 1.2014e-07], + ..., + [ 4.0047e-08, -1.9651e-07, -3.1013e-07, ..., -8.5309e-07, + 3.7253e-09, 1.9558e-07], + [ 2.0973e-06, 1.9260e-06, -3.3062e-07, ..., 2.8871e-08, + 9.3132e-10, 3.4831e-06], + [ 7.4320e-07, 3.9209e-07, 7.5437e-08, ..., 3.2876e-07, + 9.3132e-10, 8.8010e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0161, -0.0201, -0.0281, -0.0248, -0.0080, 0.0062, 0.0066, -0.0193, + -0.0040, 0.0023], device='cuda:0'), grad: tensor([ 4.4145e-07, 2.1625e-06, 2.5146e-07, -6.0499e-05, 6.1467e-07, + 4.5002e-05, 3.5018e-07, -1.5646e-07, 9.4622e-06, 2.2389e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 220.26, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5131 re_mapping 0.0060 re_causal 0.0155 /// teacc 98.99 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.1578, -0.1525, 0.0854, ..., -0.0748, 0.0462, 0.0324], + [-0.0931, -0.0365, -0.0664, ..., -0.0905, -0.0848, -0.0367], + [ 0.0299, -0.0846, -0.0674, ..., -0.0721, 0.0173, -0.2674], + ..., + [-0.1297, 0.1136, 0.0030, ..., 0.1148, -0.0236, -0.1047], + [-0.0976, -0.0860, 0.1079, ..., -0.0688, -0.1380, 0.1130], + [ 0.0090, -0.1905, 0.1099, ..., 0.0629, -0.1589, -0.0811]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.1642e-07, -7.4506e-09, ..., 3.1665e-08, + -9.3132e-10, 1.3690e-07], + [ 3.7253e-09, 6.7707e-07, 1.2387e-07, ..., 1.8161e-07, + 9.3132e-10, 1.2480e-07], + [ 0.0000e+00, 3.5111e-07, 4.7497e-08, ..., 8.3819e-08, + -2.7940e-09, 4.1910e-08], + ..., + [ 5.5879e-09, -1.8952e-06, -2.3097e-07, ..., -6.5845e-07, + 1.8626e-09, 1.9558e-08], + [ 2.5146e-08, 1.7788e-07, 6.7055e-08, ..., 9.5926e-08, + 0.0000e+00, 1.2387e-07], + [-6.2399e-08, 1.1642e-07, -2.4866e-07, ..., -6.2399e-08, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0159, -0.0200, -0.0282, -0.0246, -0.0077, 0.0061, 0.0068, -0.0192, + -0.0042, 0.0018], device='cuda:0'), grad: tensor([ 7.7672e-07, 1.6438e-06, 6.8825e-07, 3.7532e-07, 1.5246e-06, + 1.3607e-06, -4.0568e-06, -3.1888e-06, 1.0645e-06, -1.9465e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 220.70, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5278 re_mapping 0.0057 re_causal 0.0156 /// teacc 99.07 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1582, -0.1529, 0.0855, ..., -0.0750, 0.0463, 0.0323], + [-0.0934, -0.0367, -0.0667, ..., -0.0910, -0.0848, -0.0367], + [ 0.0296, -0.0850, -0.0677, ..., -0.0726, 0.0175, -0.2679], + ..., + [-0.1302, 0.1140, 0.0031, ..., 0.1158, -0.0237, -0.1049], + [-0.0979, -0.0865, 0.1082, ..., -0.0691, -0.1381, 0.1133], + [ 0.0092, -0.1909, 0.1104, ..., 0.0629, -0.1591, -0.0812]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.7008e-08, -8.3819e-09, ..., 3.0734e-08, + 4.6566e-09, 8.6665e-05], + [ 1.8626e-09, 3.5390e-08, 1.6764e-08, ..., 1.1642e-07, + -2.7474e-07, 1.6540e-06], + [ 0.0000e+00, 2.6077e-07, 6.3330e-08, ..., 1.2666e-07, + 1.8626e-08, 1.7602e-07], + ..., + [ 5.5879e-09, -3.7253e-07, -8.7544e-08, ..., -1.8347e-07, + 1.1083e-07, 3.2410e-07], + [ 2.7940e-09, 1.9558e-08, 1.0245e-08, ..., 3.7253e-08, + 1.3970e-08, 7.7579e-07], + [-2.3283e-08, 5.0291e-08, -5.8673e-08, ..., 5.9325e-07, + 5.5879e-09, 1.6671e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0163, -0.0201, -0.0283, -0.0240, -0.0083, 0.0055, 0.0068, -0.0189, + -0.0043, 0.0023], device='cuda:0'), grad: tensor([ 3.8171e-04, 7.1526e-06, 7.6834e-07, 2.3469e-07, 3.0756e-05, + 1.3616e-06, -4.3201e-04, 1.6661e-06, 3.7253e-06, 5.2191e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 220.06, cls_loss 0.0016 cls_loss_mapping 0.0034 cls_loss_causal 0.5431 re_mapping 0.0058 re_causal 0.0156 /// teacc 99.01 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1594, -0.1534, 0.0857, ..., -0.0753, 0.0465, 0.0318], + [-0.0937, -0.0365, -0.0667, ..., -0.0914, -0.0877, -0.0375], + [ 0.0294, -0.0855, -0.0686, ..., -0.0731, 0.0176, -0.2687], + ..., + [-0.1308, 0.1142, 0.0033, ..., 0.1164, -0.0238, -0.1049], + [-0.0982, -0.0874, 0.1083, ..., -0.0707, -0.1383, 0.1136], + [ 0.0093, -0.1917, 0.1105, ..., 0.0626, -0.1598, -0.0819]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 1.0710e-07, -2.8126e-07, ..., 8.3260e-07, + -1.8813e-07, -1.3942e-06], + [ 8.3819e-09, 3.6228e-07, 6.6776e-07, ..., 1.0487e-06, + 3.7253e-09, 1.1176e-07], + [ 1.2107e-08, 2.8498e-07, 3.4459e-07, ..., 2.4121e-07, + 3.3528e-08, 4.9639e-07], + ..., + [ 2.4214e-08, -2.0154e-06, 5.9605e-06, ..., 1.0334e-05, + 2.2352e-08, 2.5705e-07], + [-5.3085e-08, 8.8476e-08, 6.6217e-07, ..., 1.5842e-06, + 1.9558e-08, -1.8552e-06], + [-1.7602e-07, 4.0419e-07, -2.1160e-05, ..., -3.6895e-05, + 1.3970e-08, 4.8988e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0170, -0.0201, -0.0284, -0.0241, -0.0079, 0.0055, 0.0079, -0.0190, + -0.0046, 0.0019], device='cuda:0'), grad: tensor([-1.2591e-06, 3.7104e-06, 2.3227e-06, 1.4521e-05, 4.9084e-05, + 6.1542e-06, 1.8105e-06, 3.2723e-05, 2.3171e-06, -1.1140e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 220.08, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5141 re_mapping 0.0056 re_causal 0.0154 /// teacc 98.99 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1596, -0.1532, 0.0860, ..., -0.0755, 0.0467, 0.0315], + [-0.0937, -0.0367, -0.0668, ..., -0.0922, -0.0876, -0.0373], + [ 0.0292, -0.0856, -0.0692, ..., -0.0731, 0.0178, -0.2691], + ..., + [-0.1313, 0.1146, 0.0034, ..., 0.1166, -0.0243, -0.1051], + [-0.0987, -0.0876, 0.1090, ..., -0.0711, -0.1386, 0.1138], + [ 0.0091, -0.1925, 0.1105, ..., 0.0615, -0.1604, -0.0829]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 1.3225e-07, -9.3132e-08, ..., 8.5682e-08, + 0.0000e+00, -4.3772e-08], + [ 6.0536e-08, 2.6785e-06, 8.4843e-07, ..., 1.8151e-06, + 0.0000e+00, 4.1910e-08], + [ 3.7253e-09, 4.1816e-07, 9.7789e-08, ..., 2.5891e-07, + 0.0000e+00, 1.1269e-07], + ..., + [ 5.9605e-08, -1.0610e-05, -1.8794e-06, ..., -6.7018e-06, + 0.0000e+00, 8.3819e-08], + [ 5.3085e-08, 8.2329e-07, 2.6915e-07, ..., 2.5798e-07, + 0.0000e+00, 9.4343e-07], + [-1.5814e-06, 1.4808e-06, -7.5325e-06, ..., -4.2431e-06, + 0.0000e+00, -2.0899e-06]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0175, -0.0200, -0.0280, -0.0245, -0.0067, 0.0055, 0.0086, -0.0191, + -0.0050, 0.0009], device='cuda:0'), grad: tensor([ 8.2981e-07, 5.7705e-06, -3.1386e-07, 8.8587e-06, 2.5243e-05, + 3.1330e-06, -3.7067e-06, -1.8701e-05, 4.3325e-06, -2.5481e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 220.18, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.5428 re_mapping 0.0056 re_causal 0.0157 /// teacc 99.06 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1605, -0.1540, 0.0862, ..., -0.0757, 0.0478, 0.0316], + [-0.0938, -0.0367, -0.0670, ..., -0.0927, -0.0878, -0.0370], + [ 0.0291, -0.0858, -0.0699, ..., -0.0736, 0.0185, -0.2712], + ..., + [-0.1320, 0.1149, 0.0035, ..., 0.1168, -0.0245, -0.1052], + [-0.0991, -0.0885, 0.1091, ..., -0.0719, -0.1396, 0.1141], + [ 0.0093, -0.1929, 0.1108, ..., 0.0608, -0.1624, -0.0832]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -2.9337e-07, -1.5786e-06, ..., 1.1176e-08, + -1.0617e-07, -5.9325e-07], + [ 1.3039e-08, 3.9116e-08, 6.2399e-08, ..., 3.9116e-08, + -3.7253e-09, -2.1420e-08], + [ 2.7940e-09, 7.5437e-08, 4.0047e-07, ..., 4.4703e-08, + 1.8626e-08, 2.0862e-07], + ..., + [ 3.0734e-08, -6.7055e-08, 1.3970e-07, ..., -1.4901e-08, + 8.3819e-09, 9.6858e-08], + [ 7.2643e-08, 9.6858e-08, -3.7998e-06, ..., -1.3830e-06, + 1.3970e-08, -2.7083e-06], + [-7.4506e-09, 4.5635e-08, 3.3304e-06, ..., 1.0142e-06, + 4.6566e-09, 2.6077e-06]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0175, -0.0196, -0.0288, -0.0252, -0.0057, 0.0058, 0.0088, -0.0192, + -0.0051, 0.0004], device='cuda:0'), grad: tensor([-2.7679e-06, -3.1479e-07, 1.0096e-06, 2.1607e-06, 5.9325e-07, + -6.4354e-07, 1.6019e-07, 3.9209e-07, -8.0168e-06, 7.4357e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 163---------------------------------------------------- +epoch 163, time 220.89, cls_loss 0.0016 cls_loss_mapping 0.0031 cls_loss_causal 0.5121 re_mapping 0.0056 re_causal 0.0154 /// teacc 99.11 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1603, -0.1545, 0.0868, ..., -0.0760, 0.0480, 0.0318], + [-0.0940, -0.0344, -0.0644, ..., -0.0933, -0.0878, -0.0371], + [ 0.0289, -0.0859, -0.0704, ..., -0.0736, 0.0187, -0.2724], + ..., + [-0.1326, 0.1129, 0.0016, ..., 0.1171, -0.0246, -0.1053], + [-0.0997, -0.0891, 0.1093, ..., -0.0728, -0.1399, 0.1144], + [ 0.0102, -0.1938, 0.1112, ..., 0.0611, -0.1628, -0.0834]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, 2.8685e-07, -4.5542e-07, ..., 1.0896e-07, + 0.0000e+00, -3.9767e-07], + [ 5.4110e-07, 1.1260e-06, 1.2452e-06, ..., 1.0412e-06, + -9.3132e-10, 1.2107e-07], + [-6.1560e-07, 4.7721e-06, 8.3819e-08, ..., -3.1926e-06, + 0.0000e+00, 5.5879e-08], + ..., + [ 1.7695e-07, -2.6077e-07, 4.4890e-07, ..., -8.2422e-07, + 9.3132e-10, -2.6077e-08], + [ 2.6263e-07, 8.1211e-07, 2.8778e-07, ..., 2.0675e-07, + 0.0000e+00, 6.9849e-08], + [-1.6496e-05, 8.4341e-06, -3.4302e-05, ..., -2.3678e-05, + 0.0000e+00, -1.6261e-06]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0174, -0.0177, -0.0287, -0.0253, -0.0059, 0.0057, 0.0093, -0.0212, + -0.0053, 0.0006], device='cuda:0'), grad: tensor([-5.6066e-07, 6.6385e-06, -2.7746e-05, -2.7463e-05, 1.3375e-04, + 3.7514e-06, 7.0594e-07, 7.1526e-07, 2.2184e-06, -9.1970e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 220.31, cls_loss 0.0014 cls_loss_mapping 0.0031 cls_loss_causal 0.5319 re_mapping 0.0055 re_causal 0.0158 /// teacc 99.01 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1605, -0.1551, 0.0871, ..., -0.0761, 0.0483, 0.0320], + [-0.0943, -0.0347, -0.0649, ..., -0.0942, -0.0876, -0.0370], + [ 0.0287, -0.0860, -0.0711, ..., -0.0739, 0.0195, -0.2728], + ..., + [-0.1333, 0.1135, 0.0020, ..., 0.1175, -0.0249, -0.1054], + [-0.1003, -0.0907, 0.1093, ..., -0.0734, -0.1407, 0.1147], + [ 0.0103, -0.1945, 0.1116, ..., 0.0613, -0.1642, -0.0837]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.3039e-08, -5.7742e-08, ..., 5.5879e-09, + 2.8871e-08, -9.3132e-09], + [ 8.3819e-09, -8.7395e-06, 3.8184e-08, ..., 4.9360e-08, + 3.7253e-08, -7.1712e-08], + [ 1.8626e-09, 1.7034e-06, 4.5635e-08, ..., 7.2643e-08, + -1.4342e-07, 4.6566e-08], + ..., + [ 1.2107e-08, 6.7428e-06, -5.5879e-08, ..., -1.4156e-07, + 1.2107e-08, 3.6322e-08], + [ 2.1793e-07, 2.5146e-08, 8.5682e-08, ..., 2.1607e-07, + 3.8184e-08, -3.1665e-08], + [-3.9861e-07, 3.4459e-08, -4.6100e-07, ..., -4.7404e-07, + 2.7940e-09, -1.2387e-07]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0174, -0.0179, -0.0280, -0.0258, -0.0060, 0.0057, 0.0093, -0.0210, + -0.0056, 0.0008], device='cuda:0'), grad: tensor([ 8.2608e-07, -3.1263e-05, 9.8161e-07, 1.2079e-06, 7.1712e-07, + 2.9150e-07, -7.0781e-08, 2.5630e-05, 1.7928e-06, -1.9185e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 220.15, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5153 re_mapping 0.0055 re_causal 0.0154 /// teacc 98.96 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1606, -0.1553, 0.0874, ..., -0.0763, 0.0486, 0.0320], + [-0.0945, -0.0351, -0.0653, ..., -0.0950, -0.0875, -0.0366], + [ 0.0285, -0.0867, -0.0728, ..., -0.0743, 0.0194, -0.2755], + ..., + [-0.1337, 0.1144, 0.0025, ..., 0.1182, -0.0249, -0.1056], + [-0.1007, -0.0906, 0.1103, ..., -0.0737, -0.1407, 0.1153], + [ 0.0104, -0.1958, 0.1116, ..., 0.0610, -0.1649, -0.0843]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3784e-07, 1.3709e-06, ..., 7.3761e-07, + 0.0000e+00, 5.5507e-07], + [ 0.0000e+00, 7.7859e-07, 6.5751e-07, ..., 6.1933e-07, + -9.3132e-09, 3.8184e-08], + [ 0.0000e+00, 5.9605e-07, 2.4959e-07, ..., 6.2864e-07, + 0.0000e+00, 1.4529e-07], + ..., + [ 0.0000e+00, -1.7220e-06, 2.3283e-07, ..., -8.2143e-07, + 6.5193e-09, -1.0524e-07], + [ 1.8626e-09, 6.1467e-08, 2.9709e-07, ..., 2.4959e-07, + 0.0000e+00, -2.4494e-07], + [ 9.3132e-10, 8.1956e-08, -5.7481e-06, ..., -2.7586e-06, + 0.0000e+00, 1.9092e-07]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0173, -0.0180, -0.0282, -0.0260, -0.0060, 0.0053, 0.0100, -0.0206, + -0.0055, 0.0004], device='cuda:0'), grad: tensor([ 5.6326e-06, 3.9600e-06, -7.0967e-06, 1.5963e-06, 5.7332e-06, + 1.6335e-06, -2.3842e-06, 2.6301e-06, 9.6019e-07, -1.2696e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 220.15, cls_loss 0.0016 cls_loss_mapping 0.0033 cls_loss_causal 0.5337 re_mapping 0.0055 re_causal 0.0159 /// teacc 99.04 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1608, -0.1556, 0.0876, ..., -0.0766, 0.0486, 0.0321], + [-0.0948, -0.0351, -0.0652, ..., -0.0958, -0.0874, -0.0366], + [ 0.0283, -0.0887, -0.0735, ..., -0.0762, 0.0194, -0.2759], + ..., + [-0.1340, 0.1152, 0.0025, ..., 0.1189, -0.0250, -0.1057], + [-0.1028, -0.0910, 0.1104, ..., -0.0745, -0.1409, 0.1143], + [ 0.0109, -0.1975, 0.1119, ..., 0.0610, -0.1653, -0.0843]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 8.5682e-08, 9.4995e-08, ..., 6.5193e-08, + -1.8626e-09, 7.7300e-08], + [ 1.1735e-07, 3.5390e-07, 4.4703e-07, ..., 5.0664e-07, + -9.3132e-10, 7.5437e-07], + [ 8.3819e-09, 3.9581e-07, 2.4214e-07, ..., 1.2573e-07, + -9.3132e-10, 2.1886e-07], + ..., + [ 6.4261e-08, -8.3633e-07, 1.3132e-07, ..., -2.5146e-07, + 0.0000e+00, 2.3562e-07], + [-4.3772e-07, -1.0990e-07, -3.3118e-06, ..., -1.2759e-07, + 0.0000e+00, -9.0450e-06], + [ 6.0052e-06, 2.4773e-07, 1.0822e-06, ..., 1.1466e-05, + 0.0000e+00, 4.0755e-06]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0173, -0.0181, -0.0280, -0.0261, -0.0058, 0.0051, 0.0104, -0.0204, + -0.0062, 0.0003], device='cuda:0'), grad: tensor([ 4.5635e-07, 4.0233e-06, 1.0030e-06, 2.5108e-06, -4.2439e-05, + 2.9020e-06, 1.4538e-06, 1.4342e-07, -1.6093e-05, 4.6074e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 220.36, cls_loss 0.0020 cls_loss_mapping 0.0038 cls_loss_causal 0.5034 re_mapping 0.0053 re_causal 0.0146 /// teacc 99.11 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1611, -0.1570, 0.0876, ..., -0.0774, 0.0489, 0.0321], + [-0.0960, -0.0357, -0.0658, ..., -0.0977, -0.0873, -0.0367], + [ 0.0278, -0.0916, -0.0740, ..., -0.0791, 0.0196, -0.2763], + ..., + [-0.1360, 0.1171, 0.0031, ..., 0.1213, -0.0254, -0.1060], + [-0.1034, -0.0916, 0.1107, ..., -0.0756, -0.1409, 0.1145], + [ 0.0103, -0.2001, 0.1120, ..., 0.0607, -0.1655, -0.0849]], + device='cuda:0'), grad: tensor([[-5.5879e-09, 1.2666e-07, -1.0803e-07, ..., 1.2107e-08, + 9.3132e-10, -1.1642e-07], + [ 1.8626e-09, 2.3283e-06, 6.0536e-08, ..., 3.0454e-07, + 1.4901e-08, -1.0272e-06], + [ 9.3132e-10, 2.3544e-06, 3.7253e-08, ..., 9.2201e-08, + -2.1420e-08, 4.1910e-08], + ..., + [ 2.7940e-09, 6.8955e-06, -3.3341e-07, ..., -1.1139e-06, + 2.7940e-09, 3.4831e-07], + [-9.4064e-08, 4.3027e-07, -4.6380e-07, ..., 1.2107e-08, + 9.3132e-10, -8.6240e-07], + [ 3.7253e-09, 6.2771e-07, 2.7847e-07, ..., 2.8871e-07, + 0.0000e+00, 4.3493e-07]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0174, -0.0187, -0.0288, -0.0257, -0.0056, 0.0046, 0.0100, -0.0191, + -0.0063, 0.0001], device='cuda:0'), grad: tensor([ 1.2014e-07, 1.8636e-06, 2.3469e-06, -2.4214e-05, 1.7555e-06, + 3.6061e-06, -7.6927e-07, 1.2845e-05, -1.6484e-07, 2.6114e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 220.34, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5102 re_mapping 0.0052 re_causal 0.0144 /// teacc 99.06 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1618, -0.1575, 0.0875, ..., -0.0777, 0.0494, 0.0320], + [-0.0963, -0.0353, -0.0655, ..., -0.0986, -0.0874, -0.0360], + [ 0.0271, -0.0918, -0.0746, ..., -0.0792, 0.0201, -0.2783], + ..., + [-0.1370, 0.1168, 0.0022, ..., 0.1206, -0.0257, -0.1073], + [-0.1042, -0.0915, 0.1118, ..., -0.0743, -0.1413, 0.1154], + [ 0.0104, -0.2013, 0.1129, ..., 0.0610, -0.1660, -0.0857]], + device='cuda:0'), grad: tensor([[ 3.7253e-07, 1.9707e-06, 9.6206e-07, ..., 3.3621e-07, + 0.0000e+00, 5.1223e-07], + [ 1.8068e-07, 1.4398e-06, 7.1786e-06, ..., 5.3830e-07, + -1.8626e-09, 9.0897e-06], + [ 1.5739e-07, 2.1420e-06, 6.9663e-07, ..., 3.1386e-07, + 0.0000e+00, 6.2771e-07], + ..., + [ 1.4249e-06, 8.1062e-06, 3.0957e-06, ..., -6.0815e-07, + 9.3132e-10, 1.9036e-06], + [ 1.6233e-06, 2.9802e-07, -1.0133e-05, ..., 2.6375e-06, + 0.0000e+00, -1.7554e-05], + [-1.9222e-06, 7.8380e-06, -2.5406e-06, ..., -2.9393e-06, + 0.0000e+00, 1.6317e-06]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0179, -0.0184, -0.0281, -0.0260, -0.0056, 0.0050, 0.0104, -0.0202, + -0.0060, 0.0003], device='cuda:0'), grad: tensor([ 6.2324e-06, 2.7701e-05, 8.7693e-06, -4.1485e-05, -1.7241e-05, + 7.5139e-06, 3.6843e-06, 2.0072e-05, -3.0145e-05, 1.4886e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 220.43, cls_loss 0.0020 cls_loss_mapping 0.0041 cls_loss_causal 0.5085 re_mapping 0.0050 re_causal 0.0137 /// teacc 98.94 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1622, -0.1582, 0.0868, ..., -0.0791, 0.0495, 0.0321], + [-0.0966, -0.0359, -0.0670, ..., -0.1018, -0.0873, -0.0358], + [ 0.0262, -0.0927, -0.0754, ..., -0.0800, 0.0202, -0.2794], + ..., + [-0.1378, 0.1176, 0.0030, ..., 0.1219, -0.0257, -0.1076], + [-0.1035, -0.0919, 0.1127, ..., -0.0749, -0.1414, 0.1162], + [ 0.0106, -0.2022, 0.1136, ..., 0.0618, -0.1663, -0.0866]], + device='cuda:0'), grad: tensor([[ 7.3574e-08, 6.5193e-09, -5.8487e-07, ..., 8.5682e-08, + -1.9558e-08, -4.4517e-07], + [ 8.1956e-08, 9.9652e-08, 2.4214e-07, ..., 1.8813e-07, + 1.8626e-09, 5.0291e-08], + [ 8.8476e-08, -1.9558e-08, 3.4459e-07, ..., 2.3376e-07, + 1.8626e-09, 1.4994e-07], + ..., + [ 1.0524e-07, 9.2201e-08, 7.5698e-06, ..., 6.3777e-06, + 9.3132e-10, 2.0396e-07], + [ 6.8825e-07, 2.1420e-08, 4.9826e-07, ..., 4.8056e-07, + 9.3132e-10, 5.3924e-07], + [ 5.9977e-07, 1.3970e-08, -1.0386e-05, ..., -9.0599e-06, + 5.5879e-09, 5.5041e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0184, -0.0193, -0.0281, -0.0257, -0.0061, 0.0048, 0.0105, -0.0194, + -0.0055, 0.0007], device='cuda:0'), grad: tensor([-1.7565e-06, 1.4110e-06, -2.3395e-06, 2.3767e-06, 5.5358e-06, + -6.7316e-06, 2.3972e-06, 2.1666e-05, 2.7195e-06, -2.5257e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 220.26, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5368 re_mapping 0.0052 re_causal 0.0145 /// teacc 98.96 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1634, -0.1588, 0.0863, ..., -0.0804, 0.0496, 0.0319], + [-0.0983, -0.0361, -0.0674, ..., -0.1036, -0.0875, -0.0358], + [ 0.0263, -0.0937, -0.0761, ..., -0.0813, 0.0204, -0.2802], + ..., + [-0.1391, 0.1181, 0.0027, ..., 0.1224, -0.0259, -0.1080], + [-0.1046, -0.0929, 0.1127, ..., -0.0773, -0.1416, 0.1165], + [ 0.0122, -0.2031, 0.1149, ..., 0.0638, -0.1667, -0.0845]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 3.7253e-09, 1.6764e-08, ..., 1.8440e-07, + -9.3132e-10, 2.9802e-08], + [ 5.5879e-09, -4.6566e-09, -5.1968e-07, ..., 5.1223e-07, + -5.5879e-09, -7.4878e-07], + [ 3.7253e-09, 1.0245e-08, 2.1420e-08, ..., 1.6298e-07, + 0.0000e+00, 3.1665e-08], + ..., + [ 3.9116e-08, 2.3283e-08, 2.6915e-07, ..., 1.2787e-06, + 3.7253e-09, 3.3341e-07], + [ 1.0338e-07, 1.3970e-08, 1.3318e-07, ..., 7.6927e-07, + 0.0000e+00, 1.6857e-07], + [-5.1502e-07, 3.3528e-08, -8.0746e-07, ..., 6.2063e-06, + 0.0000e+00, 3.7160e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0192, -0.0195, -0.0286, -0.0256, -0.0071, 0.0043, 0.0106, -0.0196, + -0.0061, 0.0025], device='cuda:0'), grad: tensor([ 7.3668e-07, -3.5577e-07, 7.6182e-07, -5.2154e-08, -3.8475e-05, + 1.3951e-06, 2.4028e-06, 5.6885e-06, 3.3639e-06, 2.4572e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 220.23, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.5273 re_mapping 0.0054 re_causal 0.0148 /// teacc 99.09 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1653, -0.1593, 0.0865, ..., -0.0811, 0.0499, 0.0315], + [-0.0989, -0.0363, -0.0675, ..., -0.1046, -0.0876, -0.0358], + [ 0.0259, -0.0941, -0.0773, ..., -0.0828, 0.0221, -0.2812], + ..., + [-0.1401, 0.1189, 0.0031, ..., 0.1243, -0.0261, -0.1084], + [-0.1056, -0.0942, 0.1129, ..., -0.0780, -0.1434, 0.1166], + [ 0.0125, -0.2055, 0.1151, ..., 0.0636, -0.1670, -0.0836]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, 1.0058e-07, 1.8626e-08, ..., 9.1270e-08, + 0.0000e+00, 9.3132e-09], + [ 1.3411e-07, -4.7125e-06, 3.7253e-07, ..., 5.6811e-07, + -1.3039e-08, 2.0489e-07], + [ 2.0489e-08, 3.1553e-06, 7.1116e-06, ..., 3.3900e-07, + 1.8626e-09, 9.3579e-06], + ..., + [ 6.4168e-07, -4.7348e-06, 1.1409e-06, ..., -5.9903e-06, + 1.8626e-09, 6.3051e-07], + [-1.2200e-07, -3.0417e-06, -8.9109e-06, ..., -9.7789e-08, + 0.0000e+00, -1.2539e-05], + [-1.5860e-06, 8.8103e-07, -1.8720e-06, ..., -4.6939e-07, + 0.0000e+00, 2.1420e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0196, -0.0198, -0.0284, -0.0258, -0.0072, 0.0044, 0.0106, -0.0190, + -0.0067, 0.0026], device='cuda:0'), grad: tensor([ 2.1830e-06, -1.8597e-05, -4.6939e-07, 2.8610e-06, 3.1471e-05, + 4.2319e-06, -6.4448e-07, 1.4998e-05, -3.7193e-05, 1.2564e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 220.21, cls_loss 0.0014 cls_loss_mapping 0.0032 cls_loss_causal 0.5246 re_mapping 0.0050 re_causal 0.0149 /// teacc 98.99 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1658, -0.1600, 0.0868, ..., -0.0814, 0.0504, 0.0319], + [-0.0992, -0.0366, -0.0676, ..., -0.1056, -0.0875, -0.0357], + [ 0.0255, -0.0943, -0.0777, ..., -0.0832, 0.0226, -0.2817], + ..., + [-0.1416, 0.1196, 0.0033, ..., 0.1253, -0.0264, -0.1086], + [-0.1066, -0.0947, 0.1134, ..., -0.0783, -0.1439, 0.1168], + [ 0.0123, -0.2065, 0.1152, ..., 0.0635, -0.1676, -0.0840]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.9558e-08, -4.1071e-07, ..., 9.3132e-10, + 0.0000e+00, -3.4925e-07], + [ 2.5146e-08, -1.0246e-04, -5.2959e-05, ..., 4.2841e-08, + 0.0000e+00, 3.9116e-08], + [ 1.2107e-08, 1.2387e-07, 1.9372e-07, ..., 5.5879e-09, + 1.8626e-09, 2.5332e-07], + ..., + [ 1.2107e-08, 1.0133e-04, 5.2512e-05, ..., -2.2352e-08, + 9.3132e-10, 9.5926e-08], + [ 1.3597e-07, -6.0629e-07, -5.6252e-07, ..., -4.3772e-08, + 0.0000e+00, -9.9372e-07], + [ 5.1223e-08, 4.7404e-07, 3.2410e-07, ..., 1.9558e-08, + 0.0000e+00, 1.9837e-07]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0186, -0.0199, -0.0281, -0.0261, -0.0073, 0.0045, 0.0103, -0.0187, + -0.0070, 0.0024], device='cuda:0'), grad: tensor([-1.4454e-06, -2.8515e-04, 9.1176e-07, 5.3048e-06, -3.4831e-07, + -3.0212e-06, 1.7276e-06, 2.8253e-04, -2.8014e-06, 1.8403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 220.45, cls_loss 0.0014 cls_loss_mapping 0.0035 cls_loss_causal 0.4998 re_mapping 0.0054 re_causal 0.0148 /// teacc 99.08 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1662, -0.1594, 0.0873, ..., -0.0817, 0.0506, 0.0327], + [-0.0996, -0.0372, -0.0684, ..., -0.1067, -0.0875, -0.0356], + [ 0.0249, -0.0945, -0.0782, ..., -0.0832, 0.0226, -0.2826], + ..., + [-0.1428, 0.1203, 0.0037, ..., 0.1257, -0.0267, -0.1091], + [-0.1069, -0.0951, 0.1138, ..., -0.0787, -0.1440, 0.1171], + [ 0.0122, -0.2071, 0.1154, ..., 0.0636, -0.1682, -0.0846]], + device='cuda:0'), grad: tensor([[ 8.1025e-08, 1.6112e-07, -3.9116e-08, ..., 4.4703e-08, + 2.5146e-08, 7.3574e-08], + [ 3.4459e-08, 6.4261e-08, 2.7940e-08, ..., 5.9605e-08, + 1.8626e-09, 2.7008e-08], + [ 3.6322e-08, 4.9639e-07, 2.4121e-07, ..., 1.5832e-08, + -1.6764e-07, 9.2387e-07], + ..., + [ 3.6322e-08, 2.0564e-06, 1.1455e-06, ..., -1.8626e-08, + 4.9360e-08, 4.5486e-06], + [ 5.4576e-07, -3.0082e-07, -1.7472e-06, ..., 2.0768e-07, + 1.4901e-08, -6.5491e-06], + [-1.7323e-07, 1.7229e-07, -3.2783e-07, ..., 1.8720e-06, + 6.5193e-09, 6.0536e-08]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0176, -0.0205, -0.0279, -0.0265, -0.0075, 0.0049, 0.0103, -0.0183, + -0.0069, 0.0025], device='cuda:0'), grad: tensor([ 1.3746e-06, 5.1223e-07, -1.6391e-07, -1.1757e-05, -7.9423e-06, + 4.4778e-06, 1.1176e-06, 8.8960e-06, -2.4326e-06, 5.8897e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 220.33, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.5040 re_mapping 0.0052 re_causal 0.0143 /// teacc 99.08 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1666, -0.1606, 0.0892, ..., -0.0812, 0.0511, 0.0359], + [-0.0999, -0.0376, -0.0688, ..., -0.1077, -0.0874, -0.0355], + [ 0.0246, -0.0949, -0.0790, ..., -0.0840, 0.0223, -0.2836], + ..., + [-0.1436, 0.1209, 0.0042, ..., 0.1273, -0.0270, -0.1093], + [-0.1083, -0.0955, 0.1139, ..., -0.0797, -0.1441, 0.1171], + [ 0.0118, -0.2083, 0.1152, ..., 0.0633, -0.1688, -0.0853]], + device='cuda:0'), grad: tensor([[ 2.7101e-07, 9.3132e-10, -1.1288e-06, ..., 4.0159e-06, + 2.1141e-07, 2.5220e-06], + [ 4.4703e-08, 5.5879e-09, 1.8626e-08, ..., 2.8219e-07, + 6.9849e-08, -3.5297e-07], + [ 2.4214e-08, 1.8626e-09, 5.8115e-07, ..., 6.8452e-07, + 6.1747e-07, 6.7055e-07], + ..., + [ 1.8720e-07, -6.5193e-09, 6.7055e-08, ..., 2.1979e-07, + 5.4948e-08, 5.7090e-07], + [ 5.8021e-07, 1.8626e-09, 6.2399e-08, ..., 6.9663e-07, + 1.8626e-07, 1.0626e-06], + [ 4.3865e-07, 3.7253e-09, -8.9407e-08, ..., 8.8476e-07, + 2.7288e-07, 1.2321e-06]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0144, -0.0206, -0.0281, -0.0265, -0.0075, 0.0052, 0.0088, -0.0177, + -0.0075, 0.0019], device='cuda:0'), grad: tensor([ 1.0081e-05, -2.1085e-06, 7.7933e-06, 2.5295e-06, 2.0951e-05, + 1.4871e-05, -6.7234e-05, 2.6412e-06, 4.6417e-06, 5.7928e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 220.37, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.4922 re_mapping 0.0051 re_causal 0.0143 /// teacc 99.08 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1668, -0.1614, 0.0895, ..., -0.0817, 0.0510, 0.0360], + [-0.1003, -0.0372, -0.0687, ..., -0.1084, -0.0873, -0.0355], + [ 0.0241, -0.0952, -0.0794, ..., -0.0845, 0.0222, -0.2841], + ..., + [-0.1446, 0.1210, 0.0041, ..., 0.1275, -0.0272, -0.1096], + [-0.1100, -0.0961, 0.1144, ..., -0.0799, -0.1443, 0.1172], + [ 0.0117, -0.2102, 0.1153, ..., 0.0626, -0.1692, -0.0858]], + device='cuda:0'), grad: tensor([[ 4.7684e-07, 1.0571e-07, 2.3749e-08, ..., 3.7719e-08, + 0.0000e+00, 2.0973e-06], + [ 6.4727e-08, 2.8126e-07, 9.8255e-08, ..., 1.1828e-07, + -1.2573e-08, 7.5391e-07], + [ 1.4901e-08, 1.9884e-07, 1.5832e-07, ..., 9.4995e-08, + 9.3132e-10, 2.1933e-07], + ..., + [ 1.1642e-08, -9.0757e-07, -1.2899e-07, ..., -4.9919e-07, + -3.3993e-08, 8.8476e-08], + [ 1.8021e-07, 3.6322e-08, -2.5015e-06, ..., 1.3970e-09, + 2.3283e-09, -2.7586e-06], + [ 1.6205e-07, 3.0361e-07, 1.2852e-06, ..., 9.0338e-08, + 4.6566e-10, 1.9502e-06]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0143, -0.0202, -0.0279, -0.0273, -0.0066, 0.0065, 0.0085, -0.0182, + -0.0077, 0.0012], device='cuda:0'), grad: tensor([ 9.0450e-06, 4.0047e-06, 9.4296e-07, 4.3446e-07, 6.6496e-06, + 3.2634e-05, -5.2214e-05, -1.0943e-06, -5.6773e-06, 5.2713e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 220.14, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5403 re_mapping 0.0050 re_causal 0.0146 /// teacc 98.96 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1677, -0.1623, 0.0890, ..., -0.0823, 0.0508, 0.0351], + [-0.1012, -0.0374, -0.0689, ..., -0.1096, -0.0867, -0.0353], + [ 0.0239, -0.0953, -0.0797, ..., -0.0841, 0.0225, -0.2845], + ..., + [-0.1459, 0.1215, 0.0040, ..., 0.1278, -0.0275, -0.1098], + [-0.1102, -0.0966, 0.1149, ..., -0.0805, -0.1446, 0.1176], + [ 0.0124, -0.2117, 0.1159, ..., 0.0628, -0.1694, -0.0860]], + device='cuda:0'), grad: tensor([[ 4.0047e-08, 2.0582e-07, 1.8626e-09, ..., 1.6298e-08, + 0.0000e+00, 2.8871e-08], + [ 4.3306e-08, 5.5414e-07, 2.2305e-07, ..., 7.1246e-08, + 0.0000e+00, -4.7404e-07], + [ 1.0710e-08, 3.1199e-08, 8.6147e-08, ..., 1.7695e-08, + 0.0000e+00, 2.1840e-07], + ..., + [ 2.5146e-08, 2.3749e-07, 5.0757e-08, ..., -1.4994e-07, + 0.0000e+00, 6.0070e-08], + [ 3.3062e-08, 2.5239e-07, -7.2224e-07, ..., 4.1910e-09, + 0.0000e+00, -8.1770e-07], + [-6.5984e-07, 1.9651e-06, -1.1176e-08, ..., 3.7253e-08, + 0.0000e+00, 3.0594e-07]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0154, -0.0205, -0.0274, -0.0272, -0.0067, 0.0065, 0.0088, -0.0182, + -0.0077, 0.0014], device='cuda:0'), grad: tensor([ 1.2033e-06, 2.8852e-06, -7.5847e-06, -5.0925e-06, -6.2399e-08, + 2.2035e-06, 1.1539e-06, 3.6024e-06, -6.7800e-07, 2.3264e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 219.98, cls_loss 0.0016 cls_loss_mapping 0.0041 cls_loss_causal 0.5235 re_mapping 0.0049 re_causal 0.0141 /// teacc 99.09 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1681, -0.1632, 0.0884, ..., -0.0833, 0.0508, 0.0351], + [-0.1016, -0.0386, -0.0690, ..., -0.1120, -0.0866, -0.0350], + [ 0.0238, -0.0959, -0.0805, ..., -0.0847, 0.0226, -0.2853], + ..., + [-0.1465, 0.1231, 0.0043, ..., 0.1292, -0.0277, -0.1101], + [-0.1109, -0.0980, 0.1153, ..., -0.0813, -0.1446, 0.1184], + [ 0.0128, -0.2131, 0.1162, ..., 0.0629, -0.1696, -0.0861]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 2.9337e-08, 2.7940e-09, ..., 2.5611e-08, + 0.0000e+00, 3.0734e-08], + [ 3.5856e-08, 1.4622e-07, 1.1642e-08, ..., 3.1665e-08, + -3.2596e-09, -1.2759e-07], + [ 1.2713e-07, -1.2089e-06, 8.8476e-09, ..., -2.1011e-06, + 0.0000e+00, 3.3528e-08], + ..., + [ 9.7789e-08, 2.1867e-06, 2.4214e-08, ..., 2.0172e-06, + 9.3132e-10, 8.1025e-08], + [ 8.8010e-08, 1.1921e-07, -6.3330e-08, ..., 1.0105e-07, + 9.3132e-10, -3.7253e-08], + [-2.0489e-08, 2.4121e-07, -1.8394e-07, ..., 1.9073e-06, + 0.0000e+00, 4.5635e-08]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0157, -0.0212, -0.0275, -0.0272, -0.0070, 0.0064, 0.0088, -0.0171, + -0.0073, 0.0014], device='cuda:0'), grad: tensor([ 4.3260e-07, -2.1234e-06, -7.1898e-06, -3.2242e-06, -7.8082e-06, + -6.9849e-08, -2.8545e-07, 1.0483e-05, 1.0580e-06, 8.7023e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 220.54, cls_loss 0.0011 cls_loss_mapping 0.0029 cls_loss_causal 0.4874 re_mapping 0.0053 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.1682, -0.1636, 0.0891, ..., -0.0836, 0.0508, 0.0352], + [-0.1018, -0.0388, -0.0692, ..., -0.1127, -0.0866, -0.0350], + [ 0.0239, -0.0963, -0.0811, ..., -0.0850, 0.0228, -0.2857], + ..., + [-0.1469, 0.1236, 0.0045, ..., 0.1300, -0.0278, -0.1103], + [-0.1111, -0.0980, 0.1158, ..., -0.0815, -0.1446, 0.1188], + [ 0.0131, -0.2140, 0.1161, ..., 0.0629, -0.1698, -0.0867]], + device='cuda:0'), grad: tensor([[-4.0978e-08, 8.3819e-09, -1.6913e-06, ..., 6.3330e-08, + 0.0000e+00, -1.9372e-06], + [ 3.7253e-08, 5.4948e-08, 7.5437e-08, ..., 3.0734e-08, + 0.0000e+00, 4.5635e-08], + [-3.1348e-06, 1.2945e-07, -2.1607e-07, ..., -5.1176e-07, + 0.0000e+00, 3.7765e-07], + ..., + [ 2.6077e-08, -1.0012e-07, 1.2200e-06, ..., -1.2014e-07, + 0.0000e+00, 1.3486e-06], + [ 1.6810e-07, 8.3819e-09, 1.6298e-08, ..., 1.0896e-07, + 0.0000e+00, -1.5041e-07], + [ 2.6822e-06, 4.3772e-08, -8.4937e-07, ..., -1.0282e-06, + 0.0000e+00, 6.0070e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0157, -0.0213, -0.0272, -0.0274, -0.0071, 0.0061, 0.0088, -0.0169, + -0.0068, 0.0011], device='cuda:0'), grad: tensor([-4.7944e-06, 2.9523e-07, -1.5795e-05, 1.7257e-06, 3.9116e-06, + -2.2836e-06, 1.8701e-06, 4.0978e-06, 1.4212e-06, 9.5442e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 220.06, cls_loss 0.0016 cls_loss_mapping 0.0036 cls_loss_causal 0.5433 re_mapping 0.0055 re_causal 0.0148 /// teacc 98.98 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1683, -0.1644, 0.0897, ..., -0.0838, 0.0508, 0.0353], + [-0.1022, -0.0392, -0.0693, ..., -0.1141, -0.0869, -0.0351], + [ 0.0239, -0.0969, -0.0821, ..., -0.0857, 0.0229, -0.2864], + ..., + [-0.1473, 0.1243, 0.0044, ..., 0.1306, -0.0282, -0.1107], + [-0.1112, -0.0986, 0.1167, ..., -0.0817, -0.1448, 0.1202], + [ 0.0132, -0.2148, 0.1164, ..., 0.0628, -0.1700, -0.0870]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.7695e-08, -1.9073e-06, ..., 2.4680e-08, + 4.6566e-10, -7.1712e-08], + [ 0.0000e+00, 5.7742e-08, 1.0012e-07, ..., 3.9581e-08, + -4.6566e-10, -2.8871e-08], + [ 0.0000e+00, -1.7956e-06, 1.1176e-07, ..., -6.8545e-07, + -6.9849e-09, 2.0489e-08], + ..., + [ 4.6566e-10, 1.5013e-06, 7.4971e-08, ..., 6.6543e-07, + 2.3283e-09, 3.8650e-08], + [ 1.3970e-09, 1.2247e-07, 3.7253e-08, ..., 5.0291e-08, + 2.3283e-09, -5.9139e-08], + [ 4.6566e-10, 2.2352e-08, 1.0617e-06, ..., 7.3574e-08, + 0.0000e+00, 3.4925e-08]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0156, -0.0218, -0.0271, -0.0272, -0.0069, 0.0031, 0.0116, -0.0166, + -0.0061, 0.0010], device='cuda:0'), grad: tensor([-3.7979e-06, -2.6077e-08, -7.4022e-06, 8.7265e-07, -1.2293e-06, + 2.7055e-07, 1.7975e-07, 7.8529e-06, 6.4261e-07, 2.6319e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 220.54, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5467 re_mapping 0.0049 re_causal 0.0141 /// teacc 99.04 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.1684, -0.1648, 0.0900, ..., -0.0844, 0.0508, 0.0355], + [-0.1027, -0.0398, -0.0699, ..., -0.1152, -0.0869, -0.0347], + [ 0.0235, -0.0984, -0.0828, ..., -0.0869, 0.0230, -0.2869], + ..., + [-0.1486, 0.1253, 0.0040, ..., 0.1309, -0.0283, -0.1110], + [-0.1120, -0.0991, 0.1170, ..., -0.0821, -0.1449, 0.1202], + [ 0.0149, -0.2157, 0.1175, ..., 0.0642, -0.1702, -0.0874]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 8.9407e-08, 1.0943e-07, ..., 4.6985e-07, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 8.1537e-07, 5.1130e-07, ..., 1.0831e-06, + 0.0000e+00, 1.5832e-08], + [ 4.6566e-10, 3.8091e-06, 9.1270e-08, ..., 1.2945e-07, + -2.3283e-09, 2.7940e-08], + ..., + [ 8.8476e-09, 7.2382e-06, 4.6790e-06, ..., 1.0841e-05, + 4.6566e-10, 2.4214e-08], + [ 4.2841e-08, 2.1188e-07, 2.2585e-07, ..., 8.5169e-07, + 9.3132e-10, -1.5181e-07], + [-6.5193e-09, -5.9791e-07, -1.2323e-05, ..., -2.8566e-05, + 0.0000e+00, 8.5216e-08]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0149, -0.0223, -0.0281, -0.0280, -0.0084, 0.0037, 0.0114, -0.0161, + -0.0064, 0.0023], device='cuda:0'), grad: tensor([ 1.6578e-06, 5.2825e-06, 8.7470e-06, -1.8477e-05, 4.2856e-05, + 4.3996e-06, 8.9360e-07, 5.5552e-05, 3.2466e-06, -1.0425e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 220.20, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.5216 re_mapping 0.0050 re_causal 0.0139 /// teacc 98.96 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1688, -0.1661, 0.0906, ..., -0.0846, 0.0508, 0.0356], + [-0.1028, -0.0405, -0.0703, ..., -0.1156, -0.0869, -0.0342], + [ 0.0237, -0.0971, -0.0831, ..., -0.0866, 0.0232, -0.2874], + ..., + [-0.1496, 0.1254, 0.0041, ..., 0.1309, -0.0286, -0.1114], + [-0.1126, -0.0996, 0.1176, ..., -0.0823, -0.1449, 0.1203], + [ 0.0147, -0.2162, 0.1176, ..., 0.0640, -0.1704, -0.0879]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.7963e-08, -7.8697e-08, ..., 2.1420e-08, + 1.8626e-09, -3.8184e-08], + [ 4.6566e-10, 4.0000e-07, 2.0349e-07, ..., 2.3423e-07, + 4.6566e-10, -9.4064e-08], + [ 0.0000e+00, -7.4040e-08, 2.7940e-08, ..., 3.5856e-08, + -1.9092e-08, 1.7229e-08], + ..., + [ 6.5193e-09, -1.9092e-07, -2.3143e-07, ..., -5.0478e-07, + 1.8626e-09, 5.3551e-08], + [ 5.5879e-09, 4.6566e-08, -1.1036e-07, ..., 1.7695e-08, + 9.3132e-10, -1.0990e-07], + [ 6.3330e-08, 1.8114e-07, 5.3085e-08, ..., 9.6392e-08, + 4.6566e-10, 7.5903e-08]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0148, -0.0225, -0.0266, -0.0279, -0.0080, 0.0040, 0.0111, -0.0164, + -0.0068, 0.0020], device='cuda:0'), grad: tensor([ 3.2596e-08, 4.6706e-07, -1.1446e-06, -4.0419e-07, 9.1270e-08, + 8.7079e-08, 1.3271e-07, 3.4925e-07, -3.5856e-08, 4.3679e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 220.31, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.4984 re_mapping 0.0050 re_causal 0.0143 /// teacc 99.03 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1694, -0.1665, 0.0915, ..., -0.0848, 0.0507, 0.0359], + [-0.1038, -0.0400, -0.0699, ..., -0.1162, -0.0869, -0.0336], + [ 0.0234, -0.0960, -0.0830, ..., -0.0865, 0.0264, -0.2879], + ..., + [-0.1505, 0.1251, 0.0039, ..., 0.1316, -0.0299, -0.1121], + [-0.1129, -0.1028, 0.1181, ..., -0.0826, -0.1481, 0.1209], + [ 0.0150, -0.2174, 0.1174, ..., 0.0637, -0.1707, -0.0882]], + device='cuda:0'), grad: tensor([[-2.3749e-08, 1.2433e-07, 1.2107e-08, ..., 1.9930e-07, + 0.0000e+00, -4.9360e-08], + [ 1.9558e-08, 1.7649e-07, 8.5682e-08, ..., 1.4622e-07, + 0.0000e+00, -2.9290e-07], + [ 7.9162e-09, 2.8312e-07, 1.2945e-07, ..., 1.3784e-07, + 0.0000e+00, 8.6613e-08], + ..., + [ 1.0245e-08, -2.4438e-06, -4.1090e-06, ..., -4.4107e-06, + 0.0000e+00, 7.4040e-08], + [ 3.2596e-09, 8.6613e-08, 1.0291e-07, ..., 9.5461e-08, + 0.0000e+00, 1.2852e-07], + [ 9.2667e-08, 2.0172e-06, 3.3174e-06, ..., 3.6918e-06, + 0.0000e+00, 3.2596e-08]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0144, -0.0218, -0.0255, -0.0279, -0.0078, 0.0037, 0.0111, -0.0171, + -0.0075, 0.0018], device='cuda:0'), grad: tensor([ 5.5181e-07, -1.9073e-06, 1.2238e-06, -7.7626e-07, -2.9383e-07, + 6.0443e-07, 4.1910e-09, -7.3574e-06, 7.1339e-07, 7.2494e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 220.41, cls_loss 0.0016 cls_loss_mapping 0.0031 cls_loss_causal 0.5063 re_mapping 0.0046 re_causal 0.0133 /// teacc 99.04 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1697, -0.1677, 0.0937, ..., -0.0845, 0.0507, 0.0372], + [-0.1037, -0.0395, -0.0694, ..., -0.1169, -0.0869, -0.0333], + [ 0.0229, -0.0962, -0.0842, ..., -0.0868, 0.0264, -0.2887], + ..., + [-0.1517, 0.1248, 0.0036, ..., 0.1322, -0.0301, -0.1131], + [-0.1143, -0.1034, 0.1197, ..., -0.0827, -0.1481, 0.1217], + [ 0.0149, -0.2180, 0.1174, ..., 0.0636, -0.1710, -0.0895]], + device='cuda:0'), grad: tensor([[ 6.2073e-07, 1.1846e-06, 7.4320e-07, ..., 7.5065e-07, + -8.8476e-09, 7.1200e-07], + [ 1.9604e-07, 3.1432e-07, 2.3516e-07, ..., 2.1001e-07, + 9.3132e-10, 1.5227e-07], + [ 4.9267e-07, 2.1420e-07, 2.0955e-07, ..., 1.3877e-07, + 0.0000e+00, 3.9302e-07], + ..., + [ 2.6962e-07, -1.1474e-05, -8.2403e-06, ..., -7.4022e-06, + 0.0000e+00, 1.4901e-07], + [ 5.0366e-06, 5.0385e-07, -6.4261e-08, ..., 6.7987e-08, + 0.0000e+00, 2.5239e-06], + [ 6.7800e-07, 9.3803e-06, 6.1579e-06, ..., 6.3665e-06, + 0.0000e+00, 8.3679e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0119, -0.0212, -0.0253, -0.0282, -0.0079, 0.0040, 0.0103, -0.0177, + -0.0073, 0.0015], device='cuda:0'), grad: tensor([ 5.7705e-06, 1.3970e-06, 1.6959e-06, 7.2062e-05, 5.5879e-06, + -9.5844e-05, -3.0994e-06, -2.0236e-05, 1.2651e-05, 1.9923e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 220.00, cls_loss 0.0014 cls_loss_mapping 0.0032 cls_loss_causal 0.5273 re_mapping 0.0048 re_causal 0.0147 /// teacc 99.05 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1703, -0.1684, 0.0937, ..., -0.0850, 0.0507, 0.0371], + [-0.1058, -0.0394, -0.0694, ..., -0.1174, -0.0870, -0.0337], + [ 0.0245, -0.0964, -0.0852, ..., -0.0870, 0.0265, -0.2893], + ..., + [-0.1529, 0.1248, 0.0032, ..., 0.1320, -0.0304, -0.1132], + [-0.1174, -0.1038, 0.1207, ..., -0.0834, -0.1481, 0.1201], + [ 0.0154, -0.2188, 0.1186, ..., 0.0652, -0.1712, -0.0900]], + device='cuda:0'), grad: tensor([[-7.3574e-08, 1.6298e-08, -5.0105e-06, ..., 2.1653e-07, + 0.0000e+00, -2.3823e-06], + [ 4.4238e-08, 5.2201e-07, 1.3690e-07, ..., 2.1840e-07, + 0.0000e+00, -3.6368e-07], + [ 1.1176e-08, 1.2340e-07, 1.6345e-07, ..., -1.3821e-06, + -1.3970e-09, 1.3411e-07], + ..., + [ 2.1886e-08, -7.6229e-07, -1.1129e-07, ..., 1.0654e-06, + 1.3970e-09, 1.6438e-07], + [ 2.0564e-06, -1.4855e-07, 1.8682e-06, ..., 1.5516e-06, + 0.0000e+00, 1.4110e-06], + [-5.4762e-07, 1.7602e-07, 2.2240e-06, ..., -1.9334e-06, + 0.0000e+00, 1.4473e-06]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0121, -0.0210, -0.0254, -0.0284, -0.0093, 0.0051, 0.0102, -0.0181, + -0.0085, 0.0027], device='cuda:0'), grad: tensor([-8.3074e-06, 1.9092e-08, -7.7263e-06, 1.0438e-05, 4.9360e-07, + -1.2636e-05, -1.0785e-06, 6.6645e-06, 8.5384e-06, 3.5726e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 220.50, cls_loss 0.0013 cls_loss_mapping 0.0036 cls_loss_causal 0.5169 re_mapping 0.0050 re_causal 0.0146 /// teacc 98.96 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1709, -0.1701, 0.0939, ..., -0.0853, 0.0507, 0.0374], + [-0.1077, -0.0393, -0.0695, ..., -0.1183, -0.0871, -0.0337], + [ 0.0247, -0.0965, -0.0859, ..., -0.0872, 0.0265, -0.2899], + ..., + [-0.1549, 0.1249, 0.0031, ..., 0.1322, -0.0306, -0.1134], + [-0.1180, -0.1043, 0.1213, ..., -0.0837, -0.1481, 0.1205], + [ 0.0147, -0.2200, 0.1188, ..., 0.0648, -0.1713, -0.0904]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 1.5367e-08, -3.3062e-08, ..., 9.3132e-09, + 9.3132e-10, -1.4901e-08], + [ 3.8184e-08, 1.4529e-07, 2.6077e-08, ..., 8.8010e-08, + 4.6566e-10, -2.1886e-08], + [ 6.0536e-09, 2.9104e-07, 1.2573e-08, ..., 1.9092e-08, + -1.4435e-08, 1.6764e-08], + ..., + [ 1.2573e-08, 1.2573e-07, 1.5367e-08, ..., 1.3039e-08, + 2.7940e-09, 2.1886e-08], + [ 3.7719e-08, 2.3749e-08, -4.1910e-09, ..., 2.7008e-08, + 9.3132e-10, 3.2596e-09], + [ 2.6077e-08, 3.7253e-08, -1.1129e-07, ..., -3.3528e-08, + 4.6566e-10, 1.6298e-08]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0121, -0.0210, -0.0250, -0.0282, -0.0086, 0.0054, 0.0100, -0.0185, + -0.0086, 0.0023], device='cuda:0'), grad: tensor([ 9.9186e-08, 6.6310e-07, 8.7870e-07, -1.3169e-06, -1.8645e-06, + -1.7807e-06, 1.9334e-06, 7.5949e-07, 2.1746e-07, 3.9861e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 220.75, cls_loss 0.0020 cls_loss_mapping 0.0038 cls_loss_causal 0.4810 re_mapping 0.0054 re_causal 0.0143 /// teacc 99.05 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1718, -0.1717, 0.0940, ..., -0.0857, 0.0507, 0.0374], + [-0.1084, -0.0395, -0.0697, ..., -0.1191, -0.0871, -0.0337], + [ 0.0239, -0.0971, -0.0869, ..., -0.0872, 0.0265, -0.2909], + ..., + [-0.1567, 0.1252, 0.0034, ..., 0.1330, -0.0308, -0.1135], + [-0.1184, -0.1050, 0.1219, ..., -0.0840, -0.1482, 0.1209], + [ 0.0152, -0.2223, 0.1195, ..., 0.0646, -0.1714, -0.0909]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 8.8476e-09, 1.2806e-07, ..., 9.3132e-09, + 0.0000e+00, 1.9092e-08], + [ 3.5856e-08, 2.1420e-08, -1.2875e-05, ..., 1.1642e-08, + 0.0000e+00, 3.1991e-07], + [ 1.3039e-08, 1.2852e-07, 6.1467e-07, ..., 9.6392e-08, + -4.6566e-10, 1.5041e-07], + ..., + [ 1.8626e-09, -1.9139e-07, 1.8384e-06, ..., -3.2503e-07, + 0.0000e+00, 3.4925e-08], + [-4.4797e-07, 1.1036e-07, -8.1817e-07, ..., 3.2596e-09, + 1.8626e-09, -5.0105e-06], + [ 2.3283e-09, 1.1967e-07, 9.8497e-06, ..., 1.5646e-07, + 0.0000e+00, 2.5146e-08]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0126, -0.0205, -0.0256, -0.0297, -0.0087, 0.0069, 0.0099, -0.0185, + -0.0088, 0.0019], device='cuda:0'), grad: tensor([ 1.0710e-06, -8.1122e-05, 3.3490e-06, -2.3143e-07, 3.3416e-06, + -7.1637e-06, 1.1764e-05, 1.1951e-05, -4.8429e-06, 6.1810e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 220.54, cls_loss 0.0020 cls_loss_mapping 0.0036 cls_loss_causal 0.4965 re_mapping 0.0047 re_causal 0.0134 /// teacc 98.97 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1724, -0.1724, 0.0946, ..., -0.0864, 0.0507, 0.0376], + [-0.1088, -0.0397, -0.0702, ..., -0.1205, -0.0871, -0.0337], + [ 0.0232, -0.0976, -0.0888, ..., -0.0881, 0.0265, -0.2917], + ..., + [-0.1580, 0.1260, 0.0033, ..., 0.1332, -0.0308, -0.1139], + [-0.1190, -0.1054, 0.1232, ..., -0.0848, -0.1482, 0.1221], + [ 0.0153, -0.2255, 0.1204, ..., 0.0652, -0.1714, -0.0916]], + device='cuda:0'), grad: tensor([[-5.0245e-07, 2.3283e-09, -3.1888e-06, ..., 1.1967e-07, + 0.0000e+00, -4.3176e-06], + [ 2.6450e-07, 1.1642e-08, 8.3819e-08, ..., 5.0990e-07, + 0.0000e+00, 8.1956e-08], + [ 2.0117e-07, 2.8871e-08, 9.3179e-07, ..., 6.4261e-08, + 0.0000e+00, 1.3364e-06], + ..., + [ 8.8010e-08, 4.1910e-08, 4.0932e-07, ..., 2.7567e-07, + 0.0000e+00, 3.3900e-07], + [ 3.0221e-07, 1.8626e-09, 3.2131e-08, ..., 1.0896e-07, + 0.0000e+00, 6.7428e-07], + [ 2.8126e-07, 1.4901e-08, 9.6858e-08, ..., 2.0163e-07, + 0.0000e+00, 1.0245e-06]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0126, -0.0191, -0.0281, -0.0295, -0.0089, 0.0068, 0.0095, -0.0186, + -0.0084, 0.0021], device='cuda:0'), grad: tensor([-1.1392e-05, 5.4352e-06, 4.0568e-06, 1.8831e-06, -1.2696e-05, + -5.5619e-06, 8.3670e-06, 2.4177e-06, 2.1737e-06, 5.2750e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 220.38, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.5036 re_mapping 0.0048 re_causal 0.0137 /// teacc 99.05 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1728, -0.1735, 0.0946, ..., -0.0868, 0.0507, 0.0375], + [-0.1097, -0.0398, -0.0705, ..., -0.1213, -0.0871, -0.0338], + [ 0.0231, -0.0983, -0.0900, ..., -0.0890, 0.0265, -0.2923], + ..., + [-0.1594, 0.1265, 0.0029, ..., 0.1334, -0.0309, -0.1141], + [-0.1197, -0.1058, 0.1238, ..., -0.0851, -0.1482, 0.1221], + [ 0.0161, -0.2266, 0.1212, ..., 0.0658, -0.1715, -0.0919]], + device='cuda:0'), grad: tensor([[ 1.9930e-07, 1.3504e-08, -4.7963e-08, ..., 2.6543e-08, + 0.0000e+00, 1.8394e-07], + [ 3.4925e-08, 1.6578e-07, 2.2352e-08, ..., 6.4727e-08, + 0.0000e+00, -2.3283e-09], + [ 6.1467e-08, -4.4703e-08, 2.2352e-08, ..., -7.7253e-07, + 0.0000e+00, 5.5879e-08], + ..., + [ 1.0990e-07, 3.3062e-08, -7.8231e-08, ..., 6.4867e-07, + 0.0000e+00, 1.4482e-07], + [ 8.7917e-07, 1.0384e-07, 2.8405e-08, ..., 6.9849e-08, + 0.0000e+00, 7.9582e-07], + [ 2.4168e-07, 4.0513e-08, 5.5879e-09, ..., 4.6566e-10, + 0.0000e+00, 2.1793e-07]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0128, -0.0193, -0.0283, -0.0293, -0.0094, 0.0067, 0.0099, -0.0187, + -0.0088, 0.0027], device='cuda:0'), grad: tensor([ 1.6801e-06, 2.6748e-06, -1.3679e-05, 6.0312e-06, 2.1420e-07, + -1.3903e-05, 2.8629e-06, 9.4697e-06, 3.5428e-06, 1.0589e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 220.51, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5046 re_mapping 0.0048 re_causal 0.0135 /// teacc 99.07 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1731, -0.1744, 0.0948, ..., -0.0875, 0.0507, 0.0374], + [-0.1099, -0.0404, -0.0712, ..., -0.1226, -0.0871, -0.0337], + [ 0.0231, -0.0993, -0.0917, ..., -0.0901, 0.0265, -0.2932], + ..., + [-0.1602, 0.1280, 0.0045, ..., 0.1354, -0.0309, -0.1144], + [-0.1205, -0.1060, 0.1244, ..., -0.0855, -0.1482, 0.1223], + [ 0.0143, -0.2294, 0.1203, ..., 0.0645, -0.1715, -0.0923]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 8.8057e-07, 1.9418e-07, ..., 1.2573e-08, + 0.0000e+00, 1.3644e-07], + [ 2.3283e-09, -6.1467e-07, -1.4286e-06, ..., 8.8476e-09, + 0.0000e+00, -2.5406e-06], + [ 9.3132e-10, 1.5041e-06, 3.4459e-07, ..., 3.7253e-09, + 0.0000e+00, 3.3341e-07], + ..., + [ 1.3039e-08, 1.1465e-06, 9.0804e-07, ..., 9.3132e-10, + 0.0000e+00, 1.4175e-06], + [ 4.6566e-09, 6.0676e-07, -5.0291e-07, ..., 1.0710e-08, + 0.0000e+00, -1.4044e-06], + [-1.5600e-07, 1.1399e-06, 1.0850e-07, ..., -1.1455e-07, + 0.0000e+00, 2.4773e-07]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0132, -0.0198, -0.0285, -0.0294, -0.0081, 0.0067, 0.0101, -0.0175, + -0.0087, 0.0011], device='cuda:0'), grad: tensor([ 4.8652e-06, -1.6347e-05, 8.7321e-06, -2.3827e-05, 2.2613e-06, + 2.2966e-06, 4.4294e-06, 1.2971e-05, -1.9167e-06, 6.4746e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 190---------------------------------------------------- +epoch 190, time 221.48, cls_loss 0.0018 cls_loss_mapping 0.0038 cls_loss_causal 0.5507 re_mapping 0.0048 re_causal 0.0140 /// teacc 99.12 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1735, -0.1753, 0.0951, ..., -0.0883, 0.0506, 0.0373], + [-0.1106, -0.0400, -0.0704, ..., -0.1249, -0.0871, -0.0334], + [ 0.0231, -0.0997, -0.0925, ..., -0.0904, 0.0265, -0.2935], + ..., + [-0.1607, 0.1282, 0.0041, ..., 0.1367, -0.0310, -0.1146], + [-0.1211, -0.1065, 0.1242, ..., -0.0868, -0.1482, 0.1224], + [ 0.0142, -0.2317, 0.1206, ..., 0.0641, -0.1715, -0.0924]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.1712e-07, 1.3504e-07, ..., 7.2643e-08, + 0.0000e+00, -1.9558e-08], + [ 1.1176e-08, 1.1735e-06, 5.0850e-07, ..., 3.4645e-07, + 0.0000e+00, -1.8626e-08], + [ 1.8626e-09, 1.5059e-06, 3.9861e-07, ..., 8.3819e-08, + 0.0000e+00, 1.3039e-08], + ..., + [ 2.0489e-08, -1.0049e-06, -8.0839e-07, ..., -9.9093e-07, + 0.0000e+00, 2.3283e-08], + [ 3.0734e-08, 8.0373e-07, 4.3679e-07, ..., 3.1292e-07, + 0.0000e+00, 1.0710e-07], + [-2.2631e-07, 1.1148e-06, -2.6729e-07, ..., -2.3376e-07, + 0.0000e+00, 1.0245e-08]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0128, -0.0196, -0.0281, -0.0290, -0.0077, 0.0063, 0.0104, -0.0177, + -0.0093, 0.0005], device='cuda:0'), grad: tensor([ 2.6245e-06, 3.9153e-06, 5.6960e-06, -1.9446e-05, 1.0971e-06, + 1.8273e-06, -3.2503e-07, -1.1874e-06, 3.0771e-06, 2.7530e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 220.54, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5186 re_mapping 0.0047 re_causal 0.0138 /// teacc 99.08 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.1751, -0.1760, 0.0952, ..., -0.0892, 0.0507, 0.0371], + [-0.1118, -0.0406, -0.0713, ..., -0.1268, -0.0872, -0.0330], + [ 0.0229, -0.0999, -0.0931, ..., -0.0903, 0.0265, -0.2940], + ..., + [-0.1615, 0.1292, 0.0049, ..., 0.1379, -0.0310, -0.1151], + [-0.1213, -0.1071, 0.1253, ..., -0.0874, -0.1482, 0.1233], + [ 0.0169, -0.2331, 0.1214, ..., 0.0646, -0.1716, -0.0921]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 6.5193e-08, ..., 2.2352e-08, + 0.0000e+00, 4.0792e-07], + [ 1.8626e-09, 5.5879e-09, 4.0345e-06, ..., 4.1910e-08, + 0.0000e+00, 5.8599e-06], + [ 0.0000e+00, 9.3132e-10, 4.2934e-07, ..., 3.3528e-08, + 0.0000e+00, 5.8580e-07], + ..., + [ 1.4901e-08, -9.3132e-09, 1.2778e-06, ..., 2.8033e-07, + 0.0000e+00, 1.5087e-06], + [ 1.5832e-08, 0.0000e+00, -6.6794e-06, ..., -2.0675e-07, + 0.0000e+00, -9.5367e-06], + [-4.2841e-08, 9.3132e-10, -6.2399e-08, ..., -4.1910e-08, + 0.0000e+00, 7.4506e-08]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0133, -0.0199, -0.0282, -0.0291, -0.0093, 0.0064, 0.0102, -0.0170, + -0.0090, 0.0016], device='cuda:0'), grad: tensor([ 2.3842e-07, 9.5442e-06, 1.1232e-06, 4.4703e-07, -7.6089e-07, + 3.0641e-07, 1.4575e-06, 2.6394e-06, -1.5222e-05, 1.8068e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 220.28, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.4918 re_mapping 0.0050 re_causal 0.0134 /// teacc 99.03 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1754, -0.1770, 0.0949, ..., -0.0898, 0.0507, 0.0369], + [-0.1123, -0.0403, -0.0711, ..., -0.1275, -0.0871, -0.0331], + [ 0.0224, -0.1001, -0.0948, ..., -0.0900, 0.0266, -0.2952], + ..., + [-0.1617, 0.1291, 0.0048, ..., 0.1378, -0.0314, -0.1155], + [-0.1216, -0.1073, 0.1271, ..., -0.0880, -0.1482, 0.1248], + [ 0.0168, -0.2338, 0.1215, ..., 0.0647, -0.1717, -0.0926]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 8.3819e-09, -1.1176e-07, ..., 5.5879e-09, + 0.0000e+00, -1.9651e-06], + [ 2.9802e-08, 2.4959e-07, 9.3132e-10, ..., 1.7881e-07, + 0.0000e+00, -4.6566e-09], + [ 8.3819e-09, 2.4028e-07, 2.4214e-08, ..., 1.7509e-07, + 0.0000e+00, 2.2352e-08], + ..., + [ 8.3819e-09, -5.4948e-07, -3.4459e-08, ..., -4.1537e-07, + 0.0000e+00, 3.6322e-08], + [ 5.4017e-08, 1.1176e-08, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 3.1665e-08], + [ 2.2352e-08, 2.7940e-08, 5.5879e-09, ..., 1.4901e-08, + 0.0000e+00, 3.1665e-08]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0138, -0.0196, -0.0280, -0.0291, -0.0092, 0.0064, 0.0097, -0.0174, + -0.0076, 0.0015], device='cuda:0'), grad: tensor([-2.8480e-06, 1.2647e-06, -1.5469e-06, 3.0734e-06, 7.4506e-08, + -2.2147e-06, 2.2575e-06, -7.1526e-07, 3.1479e-07, 3.0082e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 220.17, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5275 re_mapping 0.0047 re_causal 0.0130 /// teacc 99.06 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1757, -0.1778, 0.0952, ..., -0.0908, 0.0507, 0.0352], + [-0.1130, -0.0400, -0.0710, ..., -0.1296, -0.0875, -0.0335], + [ 0.0224, -0.1010, -0.0960, ..., -0.0910, 0.0266, -0.2963], + ..., + [-0.1641, 0.1293, 0.0052, ..., 0.1384, -0.0311, -0.1158], + [-0.1226, -0.1077, 0.1285, ..., -0.0886, -0.1483, 0.1254], + [ 0.0169, -0.2355, 0.1212, ..., 0.0648, -0.1720, -0.0927]], + device='cuda:0'), grad: tensor([[-2.2352e-08, 2.7940e-09, -1.7323e-07, ..., 3.7253e-09, + 0.0000e+00, -1.4063e-07], + [ 1.8626e-09, -1.2349e-06, 1.2107e-08, ..., -9.5647e-07, + 0.0000e+00, -2.8871e-07], + [ 9.3132e-10, 6.5193e-08, 2.6077e-08, ..., 2.7940e-08, + 0.0000e+00, 1.7695e-08], + ..., + [ 5.5879e-09, 1.1763e-06, -6.5193e-09, ..., 9.4622e-07, + 0.0000e+00, 2.7567e-07], + [ 3.5390e-08, 2.8871e-08, 2.3283e-08, ..., 4.8429e-08, + 0.0000e+00, 2.5146e-08], + [-1.9558e-08, 1.0245e-08, -5.9605e-08, ..., -3.2596e-08, + 0.0000e+00, 1.5832e-08]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0162, -0.0196, -0.0283, -0.0287, -0.0090, 0.0064, 0.0105, -0.0177, + -0.0075, 0.0012], device='cuda:0'), grad: tensor([-5.9512e-07, -5.2340e-06, 1.9930e-07, -1.2387e-07, -2.4401e-07, + 1.9185e-07, 4.4238e-07, 5.1782e-06, 1.9651e-07, 3.0734e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 194---------------------------------------------------- +epoch 194, time 221.37, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.5006 re_mapping 0.0049 re_causal 0.0135 /// teacc 99.15 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1757, -0.1783, 0.0964, ..., -0.0912, 0.0508, 0.0359], + [-0.1132, -0.0402, -0.0712, ..., -0.1305, -0.0877, -0.0335], + [ 0.0222, -0.1013, -0.0965, ..., -0.0913, 0.0267, -0.2968], + ..., + [-0.1649, 0.1297, 0.0053, ..., 0.1389, -0.0314, -0.1165], + [-0.1231, -0.1078, 0.1290, ..., -0.0896, -0.1483, 0.1258], + [ 0.0173, -0.2360, 0.1213, ..., 0.0648, -0.1724, -0.0933]], + device='cuda:0'), grad: tensor([[-2.1514e-06, 3.4459e-08, 4.6566e-09, ..., 5.4948e-08, + 4.6566e-09, -3.9600e-06], + [ 5.5879e-09, 4.9919e-07, -3.9116e-08, ..., 7.5717e-07, + 4.4703e-08, -1.0803e-07], + [ 1.5832e-08, 1.6335e-06, 2.8871e-08, ..., 2.6841e-06, + 3.3528e-08, 6.5193e-08], + ..., + [ 3.1758e-07, -8.4490e-06, -1.6578e-07, ..., -1.4067e-05, + -3.1665e-08, 6.9570e-07], + [ 6.5193e-09, 1.0896e-07, 2.7940e-09, ..., 1.3784e-07, + 5.0571e-07, 1.4184e-06], + [ 1.3970e-08, 3.3155e-06, 5.6811e-08, ..., 5.6252e-06, + 4.6566e-09, 7.4506e-08]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0153, -0.0198, -0.0282, -0.0288, -0.0090, 0.0064, 0.0104, -0.0176, + -0.0076, 0.0012], device='cuda:0'), grad: tensor([-9.5740e-06, 1.3402e-06, 6.1169e-06, 5.1130e-07, 1.0043e-05, + 7.9870e-06, -5.7593e-06, -2.9400e-05, 5.9642e-06, 1.2733e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 220.40, cls_loss 0.0011 cls_loss_mapping 0.0028 cls_loss_causal 0.4992 re_mapping 0.0048 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1758, -0.1787, 0.0965, ..., -0.0914, 0.0510, 0.0359], + [-0.1133, -0.0401, -0.0711, ..., -0.1308, -0.0882, -0.0334], + [ 0.0221, -0.1016, -0.0967, ..., -0.0914, 0.0268, -0.2977], + ..., + [-0.1654, 0.1297, 0.0051, ..., 0.1391, -0.0324, -0.1169], + [-0.1233, -0.1078, 0.1297, ..., -0.0898, -0.1484, 0.1264], + [ 0.0172, -0.2363, 0.1219, ..., 0.0656, -0.1732, -0.0940]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.9802e-08, 8.3819e-09, ..., 1.0245e-08, + 0.0000e+00, 3.7253e-09], + [ 4.6566e-09, 9.4101e-06, 1.8207e-06, ..., 7.9814e-07, + 1.8626e-09, -5.4017e-08], + [ 9.3132e-10, 5.6345e-07, 1.1083e-07, ..., 5.6811e-08, + 1.8626e-09, 8.3819e-09], + ..., + [-1.7975e-07, -1.1459e-05, -2.1271e-06, ..., -1.5628e-06, + 1.8626e-09, 1.6764e-08], + [ 2.7940e-09, 4.0047e-08, -6.5193e-09, ..., 1.6764e-08, + 3.7253e-09, -2.7940e-08], + [-3.9116e-08, 5.1223e-08, -9.9652e-08, ..., -1.6764e-08, + 1.8626e-09, 2.5146e-08]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0154, -0.0197, -0.0276, -0.0289, -0.0100, 0.0065, 0.0104, -0.0179, + -0.0073, 0.0018], device='cuda:0'), grad: tensor([ 9.1270e-08, 1.6272e-05, 8.5589e-07, 1.6922e-06, 9.6019e-07, + 2.7567e-07, -2.7940e-08, -2.0236e-05, 1.1921e-07, -1.7695e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 220.05, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.4959 re_mapping 0.0047 re_causal 0.0135 /// teacc 99.07 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1760, -0.1791, 0.0967, ..., -0.0917, 0.0509, 0.0358], + [-0.1133, -0.0406, -0.0716, ..., -0.1319, -0.0885, -0.0335], + [ 0.0220, -0.1018, -0.0974, ..., -0.0913, 0.0269, -0.2984], + ..., + [-0.1659, 0.1302, 0.0053, ..., 0.1393, -0.0341, -0.1168], + [-0.1237, -0.1080, 0.1306, ..., -0.0904, -0.1484, 0.1279], + [ 0.0174, -0.2367, 0.1222, ..., 0.0657, -0.1741, -0.0942]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 1.1176e-08], + [ 9.3132e-09, 7.4506e-09, 1.5832e-08, ..., 7.4506e-09, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, -5.5879e-09, 1.5832e-08, ..., 1.8626e-09, + -2.2352e-08, 1.9558e-08], + ..., + [ 7.4506e-09, 1.6037e-06, 4.6492e-06, ..., 7.6741e-07, + 5.5879e-09, 3.6806e-06], + [ 1.2666e-07, -1.6242e-06, -4.7460e-06, ..., -7.7952e-07, + 1.7695e-08, -3.6061e-06], + [ 1.9558e-08, 7.4506e-09, 8.3819e-09, ..., 3.3528e-08, + 9.3132e-10, 4.0978e-08]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0155, -0.0201, -0.0271, -0.0289, -0.0100, 0.0063, 0.0104, -0.0179, + -0.0063, 0.0019], device='cuda:0'), grad: tensor([ 3.8184e-08, -5.4948e-08, -7.5437e-08, 1.0710e-07, -1.0617e-07, + -3.5763e-07, 7.2643e-08, 7.4878e-06, -7.2792e-06, 1.8906e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 220.58, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.5047 re_mapping 0.0049 re_causal 0.0134 /// teacc 99.06 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1763, -0.1790, 0.0970, ..., -0.0921, 0.0511, 0.0357], + [-0.1144, -0.0403, -0.0714, ..., -0.1338, -0.0891, -0.0340], + [ 0.0216, -0.1021, -0.0987, ..., -0.0920, 0.0270, -0.3006], + ..., + [-0.1665, 0.1304, 0.0055, ..., 0.1408, -0.0326, -0.1171], + [-0.1247, -0.1081, 0.1315, ..., -0.0910, -0.1484, 0.1284], + [ 0.0175, -0.2378, 0.1222, ..., 0.0651, -0.1755, -0.0947]], + device='cuda:0'), grad: tensor([[ 2.4214e-08, 1.0245e-08, 2.7940e-08, ..., 1.0896e-07, + 3.7253e-09, 1.0179e-06], + [ 3.8184e-08, 1.6671e-07, 5.4948e-08, ..., 2.4866e-07, + 6.5193e-09, 1.9372e-07], + [ 4.6566e-09, 3.9563e-06, 6.9756e-07, ..., 2.5295e-06, + 7.4506e-09, 6.9290e-07], + ..., + [ 5.1223e-08, -4.1537e-06, -3.0734e-07, ..., -2.5071e-06, + 6.5193e-09, 5.7835e-07], + [ 1.1269e-07, -1.7695e-08, -3.9767e-07, ..., 2.2072e-07, + 1.3690e-07, -1.2275e-06], + [ 5.7556e-06, 6.5193e-08, -1.4054e-06, ..., 1.8612e-05, + 1.8626e-09, 2.0489e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0156, -0.0201, -0.0264, -0.0290, -0.0097, 0.0064, 0.0104, -0.0177, + -0.0058, 0.0011], device='cuda:0'), grad: tensor([ 5.9828e-06, 1.8701e-06, 1.0327e-05, 2.6822e-07, -6.0886e-05, + -2.1420e-06, -9.1717e-06, -8.4713e-06, -6.3796e-07, 6.2943e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 220.51, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.5130 re_mapping 0.0049 re_causal 0.0139 /// teacc 99.10 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1766, -0.1798, 0.0970, ..., -0.0927, 0.0511, 0.0357], + [-0.1145, -0.0419, -0.0727, ..., -0.1348, -0.0887, -0.0339], + [ 0.0218, -0.1022, -0.0996, ..., -0.0924, 0.0270, -0.3018], + ..., + [-0.1677, 0.1319, 0.0060, ..., 0.1409, -0.0332, -0.1175], + [-0.1250, -0.1081, 0.1323, ..., -0.0916, -0.1484, 0.1291], + [ 0.0175, -0.2380, 0.1226, ..., 0.0649, -0.1762, -0.0951]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, -3.9954e-07, ..., 9.3132e-09, + 9.3132e-10, 1.0245e-08], + [ 0.0000e+00, 1.7695e-08, 2.5146e-08, ..., 2.4214e-08, + 1.8626e-09, -6.5193e-08], + [ 0.0000e+00, -2.1420e-08, 6.1467e-08, ..., -1.1828e-07, + -2.2352e-08, 5.5879e-08], + ..., + [ 3.7253e-09, -4.9360e-08, 1.8626e-09, ..., 4.4703e-08, + 1.6764e-08, 5.6811e-08], + [ 2.4214e-08, 3.7253e-09, -1.0617e-07, ..., -1.9558e-08, + 9.3132e-10, -1.1921e-07], + [ 3.7253e-09, 2.7008e-08, 3.3248e-07, ..., 8.6613e-08, + 9.3132e-10, 6.5193e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0157, -0.0212, -0.0263, -0.0291, -0.0093, 0.0064, 0.0104, -0.0168, + -0.0053, 0.0009], device='cuda:0'), grad: tensor([-9.3225e-07, -1.2871e-06, -6.8452e-07, 1.6391e-07, 1.5542e-05, + -7.5437e-08, -1.5125e-05, 1.3607e-06, -2.3562e-07, 1.2768e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 220.45, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.4862 re_mapping 0.0050 re_causal 0.0133 /// teacc 99.01 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1766, -0.1800, 0.0973, ..., -0.0931, 0.0509, 0.0358], + [-0.1165, -0.0454, -0.0765, ..., -0.1356, -0.0892, -0.0345], + [ 0.0216, -0.1013, -0.1013, ..., -0.0935, 0.0272, -0.3036], + ..., + [-0.1678, 0.1355, 0.0092, ..., 0.1417, -0.0353, -0.1177], + [-0.1260, -0.1088, 0.1330, ..., -0.0927, -0.1485, 0.1296], + [ 0.0175, -0.2385, 0.1228, ..., 0.0645, -0.1792, -0.0955]], + device='cuda:0'), grad: tensor([[-5.5879e-09, 9.3132e-10, -4.7497e-08, ..., 9.8720e-08, + 9.3132e-10, -2.0675e-07], + [-1.9278e-07, -1.3970e-08, -1.9185e-07, ..., -2.2855e-06, + 9.3132e-10, -7.4506e-09], + [ 3.7253e-09, 3.1665e-08, 8.5682e-08, ..., 6.5193e-09, + 9.3132e-10, 6.2473e-06], + ..., + [ 1.2293e-07, 2.7008e-08, 2.6543e-07, ..., 2.2780e-06, + 0.0000e+00, 1.7360e-06], + [ 7.2643e-08, -1.3039e-08, 2.4214e-08, ..., 3.2596e-08, + 1.8626e-09, -1.2547e-05], + [-8.4750e-08, 4.6566e-09, -8.0746e-07, ..., -4.5728e-07, + 0.0000e+00, -7.4506e-08]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0158, -0.0244, -0.0243, -0.0306, -0.0088, 0.0065, 0.0102, -0.0135, + -0.0044, 0.0006], device='cuda:0'), grad: tensor([ 4.8429e-08, -4.5657e-05, 1.8120e-05, 4.0978e-06, 1.1735e-06, + 7.9498e-06, 2.3842e-06, 4.9621e-05, -3.5912e-05, -1.8990e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 220.82, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.5121 re_mapping 0.0048 re_causal 0.0137 /// teacc 99.00 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1768, -0.1800, 0.0972, ..., -0.0936, 0.0512, 0.0357], + [-0.1169, -0.0450, -0.0762, ..., -0.1368, -0.0873, -0.0344], + [ 0.0233, -0.1020, -0.1033, ..., -0.0947, 0.0271, -0.3051], + ..., + [-0.1687, 0.1353, 0.0089, ..., 0.1424, -0.0369, -0.1182], + [-0.1262, -0.1091, 0.1354, ..., -0.0927, -0.1484, 0.1306], + [ 0.0173, -0.2396, 0.1227, ..., 0.0642, -0.1816, -0.0970]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.3283e-08, 6.5193e-09, ..., 1.4901e-08, + 0.0000e+00, 7.4506e-09], + [ 1.8626e-09, 4.2841e-08, 8.9407e-08, ..., 1.2293e-07, + 0.0000e+00, -6.6124e-08], + [ 1.8626e-09, 2.0768e-07, 3.4459e-08, ..., 1.1735e-07, + 0.0000e+00, 1.6764e-08], + ..., + [ 1.3039e-08, -5.2620e-07, -2.0862e-07, ..., -3.5670e-07, + 0.0000e+00, 6.7055e-08], + [ 3.7253e-08, 3.3528e-08, -8.3819e-09, ..., 2.9802e-08, + 0.0000e+00, 1.8626e-09], + [ 2.5146e-08, 1.2666e-07, -4.8429e-08, ..., -3.9116e-08, + 0.0000e+00, 2.3283e-08]], device='cuda:0') +Epoch 202, bias, value: tensor([-1.6109e-02, -2.3727e-02, -2.4842e-02, -3.0751e-02, -8.5017e-03, + 6.5119e-03, 1.0222e-02, -1.3897e-02, -3.2987e-03, -4.6578e-06], + device='cuda:0'), grad: tensor([ 1.3970e-07, -9.1828e-07, 2.0396e-07, 4.2655e-07, 3.3900e-07, + -3.5204e-07, 1.8626e-09, 1.4435e-07, 1.2945e-07, -1.0151e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 220.28, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4771 re_mapping 0.0048 re_causal 0.0131 /// teacc 98.98 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1768, -0.1804, 0.0975, ..., -0.0940, 0.0510, 0.0358], + [-0.1172, -0.0448, -0.0761, ..., -0.1376, -0.0884, -0.0343], + [ 0.0236, -0.1025, -0.1050, ..., -0.0957, 0.0273, -0.3059], + ..., + [-0.1694, 0.1349, 0.0082, ..., 0.1426, -0.0377, -0.1187], + [-0.1265, -0.1093, 0.1364, ..., -0.0934, -0.1485, 0.1314], + [ 0.0174, -0.2400, 0.1236, ..., 0.0641, -0.1824, -0.0972]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 3.7253e-09, 1.7695e-08, ..., 1.3970e-08, + 0.0000e+00, 4.6566e-09], + [ 7.4506e-09, 4.8131e-06, 5.0291e-08, ..., 3.9712e-06, + 9.3132e-10, -3.1386e-07], + [ 6.6124e-08, 8.0094e-08, 8.4750e-08, ..., 1.1362e-07, + 9.3132e-10, 6.5193e-09], + ..., + [ 3.7625e-07, -4.9546e-06, 3.7346e-07, ..., -3.8259e-06, + -4.6566e-09, 9.5926e-08], + [ 1.0245e-08, 7.4506e-09, 1.8626e-09, ..., 1.3970e-08, + 0.0000e+00, -2.7940e-09], + [-1.0990e-06, -5.4017e-08, -1.2470e-06, ..., -8.0187e-07, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 203, bias, value: tensor([-1.6049e-02, -2.3541e-02, -2.4778e-02, -3.0594e-02, -8.1404e-03, + 6.4418e-03, 1.0046e-02, -1.4275e-02, -2.7867e-03, -3.0324e-05], + device='cuda:0'), grad: tensor([ 8.9407e-08, 9.5889e-06, 3.5856e-07, 7.2084e-07, 6.3051e-07, + 1.2023e-06, 8.1398e-07, -9.3430e-06, 7.5437e-08, -4.1425e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 220.70, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.5164 re_mapping 0.0048 re_causal 0.0136 /// teacc 99.09 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1772, -0.1814, 0.0977, ..., -0.0946, 0.0509, 0.0356], + [-0.1164, -0.0448, -0.0762, ..., -0.1389, -0.0889, -0.0336], + [ 0.0235, -0.1029, -0.1058, ..., -0.0964, 0.0273, -0.3078], + ..., + [-0.1701, 0.1347, 0.0081, ..., 0.1439, -0.0381, -0.1197], + [-0.1270, -0.1093, 0.1369, ..., -0.0944, -0.1485, 0.1320], + [ 0.0165, -0.2419, 0.1238, ..., 0.0640, -0.1828, -0.0988]], + device='cuda:0'), grad: tensor([[ 3.8184e-08, 3.7532e-07, -2.0210e-07, ..., 1.5832e-08, + 0.0000e+00, -1.1083e-07], + [ 2.2911e-07, -2.4587e-05, -7.2159e-06, ..., 3.8184e-08, + 0.0000e+00, -1.0066e-05], + [-7.6815e-06, 1.6484e-07, 6.3330e-08, ..., -1.8347e-07, + 0.0000e+00, 7.5437e-08], + ..., + [ 3.1292e-07, 2.2799e-05, 6.6906e-06, ..., -1.1176e-07, + 0.0000e+00, 9.3430e-06], + [ 1.0338e-07, 9.6858e-08, 1.5739e-07, ..., 1.0245e-08, + 0.0000e+00, 1.5832e-07], + [ 5.7705e-06, 1.0636e-06, 2.6077e-07, ..., 1.8068e-07, + 0.0000e+00, 4.6007e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0163, -0.0235, -0.0248, -0.0303, -0.0080, 0.0064, 0.0101, -0.0143, + -0.0026, -0.0005], device='cuda:0'), grad: tensor([ 9.5833e-07, -8.6904e-05, -5.7727e-05, 8.2701e-06, 1.2806e-06, + 3.0734e-07, 2.6077e-07, 8.7678e-05, 2.4550e-06, 4.3601e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 220.46, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5090 re_mapping 0.0045 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1774, -0.1819, 0.0990, ..., -0.0950, 0.0507, 0.0365], + [-0.1169, -0.0445, -0.0760, ..., -0.1397, -0.0918, -0.0331], + [ 0.0239, -0.1035, -0.1066, ..., -0.0967, 0.0283, -0.3086], + ..., + [-0.1705, 0.1345, 0.0078, ..., 0.1445, -0.0401, -0.1207], + [-0.1273, -0.1097, 0.1373, ..., -0.0955, -0.1486, 0.1323], + [ 0.0165, -0.2428, 0.1238, ..., 0.0640, -0.1834, -0.0994]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 5.5879e-09, -3.7253e-09, ..., 7.4506e-09, + -1.8626e-09, 3.0734e-08], + [-2.7288e-07, 4.2841e-08, 2.4214e-08, ..., 4.7497e-08, + 0.0000e+00, -1.1204e-06], + [ 4.6566e-09, 1.6764e-07, 2.9802e-08, ..., 2.2352e-07, + 0.0000e+00, 4.8429e-08], + ..., + [ 8.3819e-09, -2.6822e-07, -1.9558e-08, ..., -3.0920e-07, + 0.0000e+00, 8.1956e-08], + [ 2.0955e-07, 3.7253e-09, -2.7660e-07, ..., -1.6764e-08, + 0.0000e+00, 6.5193e-08], + [ 3.7253e-09, 2.9802e-08, -5.3085e-08, ..., -2.7008e-08, + 0.0000e+00, 3.5390e-08]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0149, -0.0232, -0.0241, -0.0303, -0.0082, 0.0065, 0.0098, -0.0148, + -0.0029, -0.0010], device='cuda:0'), grad: tensor([ 1.1735e-07, -2.9206e-06, 5.1502e-07, 9.3132e-08, 1.6484e-07, + 8.1398e-07, 6.4075e-07, -2.2538e-07, 8.2236e-07, -1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 220.84, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4972 re_mapping 0.0046 re_causal 0.0131 /// teacc 99.01 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1778, -0.1825, 0.0999, ..., -0.0955, 0.0507, 0.0367], + [-0.1174, -0.0445, -0.0760, ..., -0.1416, -0.0919, -0.0331], + [ 0.0238, -0.1040, -0.1071, ..., -0.0978, 0.0284, -0.3091], + ..., + [-0.1713, 0.1347, 0.0078, ..., 0.1461, -0.0405, -0.1210], + [-0.1274, -0.1098, 0.1374, ..., -0.0960, -0.1486, 0.1327], + [ 0.0166, -0.2433, 0.1238, ..., 0.0639, -0.1834, -0.0996]], + device='cuda:0'), grad: tensor([[-4.0978e-08, 2.9895e-07, -1.1455e-07, ..., 1.6764e-08, + 0.0000e+00, -5.0664e-07], + [ 3.7253e-09, 5.5134e-07, 1.2945e-07, ..., 1.9558e-07, + 1.8626e-09, 1.6578e-07], + [ 1.8626e-09, 3.9004e-06, 1.6764e-07, ..., 1.2107e-07, + -1.8626e-09, 2.5425e-07], + ..., + [ 1.0245e-08, 1.1042e-05, 1.3784e-07, ..., -1.3039e-08, + 0.0000e+00, 2.3842e-07], + [ 1.8626e-09, 3.1162e-06, -1.8915e-06, ..., 1.0431e-07, + 0.0000e+00, -2.8443e-06], + [ 9.3132e-10, 2.8685e-07, 5.8115e-07, ..., -9.3132e-08, + 0.0000e+00, 6.0629e-07]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0145, -0.0232, -0.0242, -0.0303, -0.0084, 0.0064, 0.0097, -0.0147, + -0.0028, -0.0011], device='cuda:0'), grad: tensor([-6.5938e-07, 1.1129e-06, 5.5917e-06, -2.1607e-05, 7.6927e-07, + 3.4906e-06, -1.2834e-06, 1.3351e-05, -3.5241e-06, 2.7344e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 220.53, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4775 re_mapping 0.0048 re_causal 0.0132 /// teacc 99.03 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1780, -0.1831, 0.0987, ..., -0.0980, 0.0507, 0.0368], + [-0.1169, -0.0447, -0.0761, ..., -0.1428, -0.0920, -0.0329], + [ 0.0236, -0.1045, -0.1076, ..., -0.0989, 0.0286, -0.3097], + ..., + [-0.1718, 0.1350, 0.0079, ..., 0.1459, -0.0414, -0.1213], + [-0.1276, -0.1102, 0.1377, ..., -0.0963, -0.1486, 0.1331], + [ 0.0167, -0.2437, 0.1243, ..., 0.0623, -0.1837, -0.1000]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, -3.2969e-07, ..., 2.4214e-08, + 0.0000e+00, -3.7812e-07], + [ 0.0000e+00, 1.0245e-08, 9.4995e-08, ..., 4.3772e-08, + 0.0000e+00, 4.4703e-08], + [ 0.0000e+00, 5.5879e-09, 1.8068e-07, ..., 5.5879e-09, + -1.8626e-09, 2.7101e-07], + ..., + [ 0.0000e+00, -4.1910e-08, 1.3001e-06, ..., 8.4378e-07, + 0.0000e+00, 1.9185e-07], + [ 0.0000e+00, 9.3132e-10, -5.5600e-07, ..., 7.1712e-08, + 0.0000e+00, -9.7603e-07], + [ 0.0000e+00, 3.7253e-09, -1.4408e-06, ..., -1.2470e-06, + 0.0000e+00, 3.7905e-07]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0151, -0.0232, -0.0242, -0.0302, -0.0067, 0.0061, 0.0105, -0.0147, + -0.0028, -0.0021], device='cuda:0'), grad: tensor([-6.9011e-07, 8.9407e-08, 7.3761e-07, 5.5321e-07, 5.6066e-07, + 9.0525e-07, -2.7753e-07, 3.0454e-06, -2.2911e-06, -2.6375e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 220.70, cls_loss 0.0014 cls_loss_mapping 0.0031 cls_loss_causal 0.5256 re_mapping 0.0048 re_causal 0.0132 /// teacc 98.96 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1786, -0.1836, 0.0990, ..., -0.0984, 0.0509, 0.0368], + [-0.1168, -0.0447, -0.0762, ..., -0.1435, -0.0926, -0.0328], + [ 0.0231, -0.1052, -0.1093, ..., -0.0998, 0.0286, -0.3109], + ..., + [-0.1723, 0.1350, 0.0080, ..., 0.1461, -0.0412, -0.1219], + [-0.1279, -0.1106, 0.1384, ..., -0.0967, -0.1487, 0.1338], + [ 0.0174, -0.2442, 0.1249, ..., 0.0623, -0.1842, -0.1004]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, -5.2154e-08, ..., 2.5146e-08, + 1.8626e-09, 1.1548e-07], + [ 0.0000e+00, 1.9558e-08, 7.4506e-09, ..., 9.5926e-08, + 3.5390e-08, 2.7008e-08], + [ 0.0000e+00, 2.7381e-07, 3.0734e-08, ..., 1.9558e-08, + -1.5832e-08, 1.1176e-07], + ..., + [ 0.0000e+00, 4.0047e-08, 8.3819e-09, ..., 6.5193e-09, + 5.5879e-09, 4.5635e-08], + [ 0.0000e+00, 2.1420e-08, -4.7497e-08, ..., 3.2596e-08, + 0.0000e+00, -1.9558e-08], + [-9.3132e-10, 1.0245e-08, 1.8626e-09, ..., 6.4261e-08, + 3.7253e-09, 7.7300e-08]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0153, -0.0231, -0.0242, -0.0302, -0.0070, 0.0063, 0.0098, -0.0149, + -0.0027, -0.0018], device='cuda:0'), grad: tensor([ 1.9651e-06, 1.7127e-06, 1.6307e-06, -2.9895e-07, 9.1270e-06, + 1.3262e-06, -1.7822e-05, 6.4168e-07, 1.0123e-06, 6.9104e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 220.48, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4879 re_mapping 0.0046 re_causal 0.0127 /// teacc 99.08 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1790, -0.1840, 0.0987, ..., -0.0992, 0.0510, 0.0365], + [-0.1169, -0.0447, -0.0764, ..., -0.1454, -0.0927, -0.0326], + [ 0.0229, -0.1055, -0.1106, ..., -0.1001, 0.0286, -0.3112], + ..., + [-0.1727, 0.1351, 0.0079, ..., 0.1463, -0.0412, -0.1228], + [-0.1284, -0.1113, 0.1387, ..., -0.0976, -0.1487, 0.1338], + [ 0.0180, -0.2445, 0.1260, ..., 0.0627, -0.1845, -0.1008]], + device='cuda:0'), grad: tensor([[ 2.7008e-08, 3.1665e-08, -5.0943e-07, ..., 4.4703e-08, + -2.7940e-09, -1.5274e-07], + [ 7.4506e-09, 6.3330e-08, 5.0291e-08, ..., 8.6613e-08, + 0.0000e+00, -4.4703e-08], + [ 5.5879e-09, 4.1910e-08, 3.0827e-07, ..., 5.3085e-08, + 0.0000e+00, 1.4901e-07], + ..., + [ 2.3283e-08, -5.6997e-07, -4.3772e-08, ..., -7.2177e-07, + 0.0000e+00, 5.7742e-08], + [ 8.4750e-08, 2.9802e-08, 2.5146e-08, ..., 8.0094e-08, + 0.0000e+00, 1.1176e-08], + [-1.4622e-07, 1.2759e-07, -1.1362e-07, ..., 3.8184e-08, + 0.0000e+00, -5.8673e-08]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0158, -0.0230, -0.0239, -0.0302, -0.0072, 0.0061, 0.0103, -0.0151, + -0.0034, -0.0010], device='cuda:0'), grad: tensor([-1.4110e-06, 5.0291e-08, 1.1306e-06, 2.5518e-07, 9.2853e-07, + -1.0245e-06, 1.3765e-06, -1.6363e-06, 1.5646e-07, 1.5087e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 220.01, cls_loss 0.0014 cls_loss_mapping 0.0032 cls_loss_causal 0.5218 re_mapping 0.0046 re_causal 0.0128 /// teacc 99.04 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1804, -0.1852, 0.0988, ..., -0.0995, 0.0510, 0.0353], + [-0.1149, -0.0448, -0.0764, ..., -0.1461, -0.0927, -0.0310], + [ 0.0224, -0.1066, -0.1119, ..., -0.1009, 0.0286, -0.3126], + ..., + [-0.1739, 0.1352, 0.0083, ..., 0.1474, -0.0413, -0.1234], + [-0.1295, -0.1133, 0.1398, ..., -0.0984, -0.1487, 0.1339], + [ 0.0179, -0.2465, 0.1253, ..., 0.0612, -0.1851, -0.1016]], + device='cuda:0'), grad: tensor([[ 6.3237e-07, 6.5193e-09, 7.4506e-09, ..., 7.4506e-09, + 0.0000e+00, 1.0319e-06], + [ 3.8091e-07, 7.5437e-08, 8.3819e-09, ..., 4.6566e-08, + 0.0000e+00, 5.8208e-07], + [ 1.1269e-07, 3.8836e-07, 1.3039e-08, ..., 1.2573e-07, + 0.0000e+00, 1.9092e-07], + ..., + [ 1.5739e-07, -2.2724e-07, -9.3132e-09, ..., -9.2201e-08, + 0.0000e+00, 2.8126e-07], + [ 1.1563e-05, 3.0734e-08, 6.1467e-08, ..., 2.7847e-07, + 0.0000e+00, 2.5123e-05], + [ 1.4994e-07, 3.2596e-08, -1.6764e-07, ..., 3.4459e-07, + 0.0000e+00, 3.1479e-07]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0168, -0.0228, -0.0242, -0.0301, -0.0062, 0.0063, 0.0105, -0.0151, + -0.0039, -0.0023], device='cuda:0'), grad: tensor([ 2.2128e-06, 1.8142e-06, 1.1278e-06, 8.1807e-06, -1.9576e-06, + -5.0217e-05, -3.0957e-06, 4.7684e-07, 3.8892e-05, 2.5742e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 220.47, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4678 re_mapping 0.0046 re_causal 0.0131 /// teacc 99.07 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1812, -0.1856, 0.0983, ..., -0.1009, 0.0510, 0.0356], + [-0.1161, -0.0448, -0.0765, ..., -0.1473, -0.0927, -0.0312], + [ 0.0220, -0.1071, -0.1125, ..., -0.1014, 0.0286, -0.3144], + ..., + [-0.1747, 0.1356, 0.0085, ..., 0.1487, -0.0414, -0.1234], + [-0.1331, -0.1138, 0.1403, ..., -0.0996, -0.1487, 0.1319], + [ 0.0179, -0.2471, 0.1256, ..., 0.0613, -0.1852, -0.1022]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, -4.6566e-08, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-08], + [ 0.0000e+00, 2.7008e-08, 5.5879e-09, ..., 4.6566e-09, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, 3.0734e-08, 3.3528e-08, ..., 1.8626e-09, + 0.0000e+00, 1.5832e-08], + ..., + [ 9.3132e-10, 3.7253e-08, 3.7253e-09, ..., -7.4506e-09, + 0.0000e+00, 6.5193e-09], + [ 1.2107e-08, 6.5193e-09, 2.8871e-08, ..., 2.5146e-08, + 0.0000e+00, 1.5832e-08], + [-5.5879e-09, 6.5193e-09, -5.0291e-08, ..., -2.9802e-08, + 0.0000e+00, -1.2107e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0170, -0.0228, -0.0245, -0.0305, -0.0063, 0.0071, 0.0108, -0.0150, + -0.0064, -0.0024], device='cuda:0'), grad: tensor([-8.9407e-08, 8.3819e-08, 1.3970e-08, -1.9278e-07, 5.7742e-08, + 1.2107e-08, -6.7987e-08, 9.7789e-08, 1.6764e-07, -8.0094e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 220.49, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.5046 re_mapping 0.0045 re_causal 0.0126 /// teacc 99.08 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1815, -0.1871, 0.0955, ..., -0.1044, 0.0510, 0.0360], + [-0.1161, -0.0449, -0.0767, ..., -0.1488, -0.0927, -0.0309], + [ 0.0218, -0.1078, -0.1137, ..., -0.1035, 0.0286, -0.3156], + ..., + [-0.1755, 0.1358, 0.0079, ..., 0.1489, -0.0415, -0.1255], + [-0.1335, -0.1143, 0.1406, ..., -0.1012, -0.1487, 0.1326], + [ 0.0179, -0.2476, 0.1288, ..., 0.0626, -0.1853, -0.1006]], + device='cuda:0'), grad: tensor([[ 5.1223e-08, 6.9849e-09, 1.1967e-07, ..., 1.7695e-08, + 0.0000e+00, -4.6566e-10], + [ 3.8650e-08, 5.6345e-08, 1.0477e-07, ..., 3.0734e-08, + -5.1223e-09, -4.6566e-08], + [ 3.8883e-07, 4.1444e-08, 9.2946e-07, ..., 1.1222e-07, + -9.3132e-10, 2.3283e-09], + ..., + [ 1.3039e-08, -6.2399e-08, 4.6566e-10, ..., -4.1910e-08, + 6.5193e-09, 1.2573e-08], + [ 7.5437e-08, 1.7695e-08, 1.8766e-07, ..., 3.6787e-08, + 9.3132e-10, 8.8476e-09], + [-1.0598e-06, 2.3283e-08, -2.5276e-06, ..., -3.0175e-07, + 4.6566e-10, -4.1910e-09]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0188, -0.0228, -0.0247, -0.0307, -0.0063, 0.0072, 0.0106, -0.0152, + -0.0060, -0.0007], device='cuda:0'), grad: tensor([ 4.9500e-07, 6.0955e-07, 2.4494e-06, 9.9000e-07, 1.0021e-06, + 1.3448e-06, 8.3214e-07, 2.3050e-07, 6.9616e-07, -8.6576e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 220.59, cls_loss 0.0012 cls_loss_mapping 0.0029 cls_loss_causal 0.5070 re_mapping 0.0047 re_causal 0.0129 /// teacc 99.08 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1818, -0.1886, 0.0958, ..., -0.1049, 0.0482, 0.0362], + [-0.1162, -0.0447, -0.0768, ..., -0.1499, -0.0930, -0.0312], + [ 0.0217, -0.1083, -0.1147, ..., -0.1035, 0.0290, -0.3172], + ..., + [-0.1756, 0.1358, 0.0079, ..., 0.1502, -0.0407, -0.1256], + [-0.1341, -0.1149, 0.1411, ..., -0.1034, -0.1489, 0.1333], + [ 0.0181, -0.2483, 0.1290, ..., 0.0624, -0.1859, -0.1010]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 3.2596e-09, 1.1753e-06, ..., 9.3924e-07, + 4.6566e-10, -5.5879e-09], + [ 7.4506e-09, 2.0629e-07, 1.3504e-07, ..., 1.0198e-07, + 4.6566e-10, -2.6077e-08], + [ 0.0000e+00, 1.1176e-08, 6.8918e-08, ..., 5.3085e-08, + -4.6566e-10, 1.9558e-08], + ..., + [ 6.5193e-09, 1.0729e-06, 3.2876e-07, ..., 2.1560e-07, + 1.3970e-09, 4.6566e-08], + [ 1.1176e-08, 3.7253e-09, -9.7789e-09, ..., 2.3283e-09, + 0.0000e+00, 2.3283e-09], + [ 8.8476e-09, 9.5461e-08, -2.1867e-06, ..., -1.7043e-06, + 0.0000e+00, 1.3970e-08]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0183, -0.0226, -0.0247, -0.0307, -0.0064, 0.0071, 0.0107, -0.0153, + -0.0056, -0.0009], device='cuda:0'), grad: tensor([ 2.2613e-06, 2.8498e-07, 9.2667e-08, -1.8645e-06, 1.3560e-06, + -1.2061e-07, -8.2888e-08, 2.0452e-06, 5.6345e-08, -4.0270e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 220.41, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4699 re_mapping 0.0045 re_causal 0.0125 /// teacc 99.08 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1820, -0.1893, 0.0960, ..., -0.1050, 0.0482, 0.0363], + [-0.1159, -0.0448, -0.0768, ..., -0.1505, -0.0932, -0.0311], + [ 0.0214, -0.1088, -0.1159, ..., -0.1039, 0.0290, -0.3181], + ..., + [-0.1762, 0.1359, 0.0079, ..., 0.1508, -0.0408, -0.1258], + [-0.1343, -0.1156, 0.1422, ..., -0.1042, -0.1489, 0.1339], + [ 0.0181, -0.2485, 0.1291, ..., 0.0625, -0.1862, -0.1018]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.1886e-08, 5.5879e-09, ..., 1.8626e-08, + 0.0000e+00, 3.2596e-09], + [ 4.6566e-10, 3.3528e-08, 7.4506e-09, ..., 2.7008e-08, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 3.3528e-08, 7.9162e-09, ..., 1.8626e-08, + 0.0000e+00, 2.3283e-09], + ..., + [ 9.3132e-10, -1.0291e-07, -1.9092e-08, ..., -7.3574e-08, + 0.0000e+00, 2.7940e-09], + [ 6.0536e-09, -4.6566e-10, -1.9558e-08, ..., -4.6566e-10, + 0.0000e+00, 1.8626e-08], + [ 2.3283e-09, 1.2107e-08, -1.3039e-08, ..., -1.2573e-08, + 0.0000e+00, 6.0536e-09]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0181, -0.0226, -0.0248, -0.0306, -0.0060, 0.0069, 0.0101, -0.0153, + -0.0053, -0.0009], device='cuda:0'), grad: tensor([ 7.3574e-08, 7.3109e-08, 3.8650e-08, 3.2131e-08, 1.1642e-08, + 1.2247e-07, -2.9942e-07, -1.4203e-07, 1.1735e-07, -1.3504e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 220.43, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4958 re_mapping 0.0045 re_causal 0.0126 /// teacc 99.03 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1838, -0.1901, 0.0958, ..., -0.1050, 0.0482, 0.0359], + [-0.1163, -0.0448, -0.0769, ..., -0.1511, -0.0933, -0.0312], + [ 0.0223, -0.1098, -0.1174, ..., -0.1046, 0.0290, -0.3199], + ..., + [-0.1771, 0.1362, 0.0080, ..., 0.1513, -0.0409, -0.1260], + [-0.1339, -0.1158, 0.1443, ..., -0.1032, -0.1489, 0.1361], + [ 0.0169, -0.2497, 0.1288, ..., 0.0625, -0.1862, -0.1039]], + device='cuda:0'), grad: tensor([[ 1.0710e-08, 1.0710e-08, 1.2573e-08, ..., 3.2131e-08, + 0.0000e+00, 9.3132e-10], + [ 3.4925e-08, 6.9849e-08, 3.2131e-08, ..., 1.2992e-07, + 0.0000e+00, 1.8626e-09], + [ 1.2573e-08, 6.2399e-08, 1.4901e-08, ..., 2.1420e-08, + 0.0000e+00, 5.1223e-09], + ..., + [ 2.5611e-08, 6.4727e-08, 5.8208e-08, ..., 9.2201e-08, + 0.0000e+00, 2.3283e-09], + [ 6.3330e-08, 5.1223e-08, 1.2014e-07, ..., 1.2293e-07, + 0.0000e+00, 2.0489e-08], + [-7.0315e-08, 2.5611e-08, -5.8953e-07, ..., -1.6624e-07, + 0.0000e+00, 4.6566e-09]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0184, -0.0227, -0.0251, -0.0308, -0.0061, 0.0071, 0.0104, -0.0152, + -0.0035, -0.0014], device='cuda:0'), grad: tensor([ 1.6904e-07, 1.0002e-06, -5.9605e-08, 6.4587e-07, -2.6561e-06, + -3.2783e-07, 6.7241e-07, 5.7882e-07, 7.3528e-07, -7.3435e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 220.73, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.5174 re_mapping 0.0044 re_causal 0.0124 /// teacc 98.97 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1842, -0.1905, 0.0960, ..., -0.1052, 0.0470, 0.0359], + [-0.1166, -0.0449, -0.0776, ..., -0.1529, -0.0934, -0.0314], + [ 0.0221, -0.1120, -0.1182, ..., -0.1067, 0.0293, -0.3207], + ..., + [-0.1781, 0.1366, 0.0076, ..., 0.1514, -0.0418, -0.1263], + [-0.1341, -0.1162, 0.1450, ..., -0.1042, -0.1489, 0.1366], + [ 0.0145, -0.2502, 0.1303, ..., 0.0602, -0.1865, -0.1043]], + device='cuda:0'), grad: tensor([[ 1.8161e-08, 1.0710e-08, 2.3283e-08, ..., 5.4017e-08, + 0.0000e+00, 3.8184e-08], + [ 5.1223e-09, 4.3912e-07, 4.7497e-08, ..., 7.1013e-07, + 0.0000e+00, 3.2596e-09], + [ 2.7940e-09, 8.8429e-07, 8.8476e-09, ..., 1.3709e-06, + 0.0000e+00, 5.1223e-09], + ..., + [ 1.5832e-08, -1.5404e-06, 8.1863e-07, ..., -1.6093e-06, + 0.0000e+00, 1.0245e-08], + [ 2.2352e-08, 3.3993e-08, 9.7323e-08, ..., 1.2526e-07, + 0.0000e+00, 2.5611e-08], + [-1.7695e-08, 5.5414e-08, -1.6121e-06, ..., -1.4864e-06, + 0.0000e+00, 2.3749e-08]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0183, -0.0230, -0.0252, -0.0307, -0.0034, 0.0071, 0.0103, -0.0151, + -0.0034, -0.0035], device='cuda:0'), grad: tensor([ 1.9558e-07, 1.5311e-06, 2.8815e-06, 6.9477e-07, 1.6764e-06, + -1.5926e-07, -1.9092e-07, -2.4550e-06, 4.5402e-07, -4.6268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 220.75, cls_loss 0.0011 cls_loss_mapping 0.0029 cls_loss_causal 0.4941 re_mapping 0.0045 re_causal 0.0125 /// teacc 98.97 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1847, -0.1914, 0.0958, ..., -0.1060, 0.0461, 0.0361], + [-0.1169, -0.0450, -0.0777, ..., -0.1554, -0.0936, -0.0314], + [ 0.0222, -0.1124, -0.1192, ..., -0.1071, 0.0297, -0.3214], + ..., + [-0.1791, 0.1368, 0.0075, ..., 0.1520, -0.0428, -0.1265], + [-0.1341, -0.1169, 0.1460, ..., -0.1053, -0.1490, 0.1381], + [ 0.0148, -0.2505, 0.1309, ..., 0.0606, -0.1866, -0.1052]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 4.2375e-08, ..., 4.3772e-08, + 0.0000e+00, -6.0536e-09], + [ 5.1223e-09, 3.7253e-09, -1.8766e-07, ..., 2.5798e-07, + 0.0000e+00, -2.1933e-07], + [ 0.0000e+00, 3.7253e-09, 4.0047e-08, ..., 1.1176e-08, + 4.6566e-10, 6.3796e-08], + ..., + [ 1.8626e-09, 4.1910e-09, 3.4217e-06, ..., 2.6431e-06, + 4.6566e-10, 1.0710e-07], + [ 1.1642e-08, 1.2107e-08, 8.3353e-08, ..., 4.7637e-07, + 0.0000e+00, -2.6915e-07], + [-1.5181e-07, 1.3970e-09, -6.0722e-06, ..., -5.7146e-06, + 0.0000e+00, 1.4575e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0184, -0.0232, -0.0252, -0.0310, -0.0037, 0.0073, 0.0098, -0.0151, + -0.0023, -0.0031], device='cuda:0'), grad: tensor([ 9.4064e-08, -2.9579e-06, 5.4296e-07, 1.2144e-06, 5.4725e-06, + 2.1374e-07, 5.2294e-07, 9.3803e-06, 1.1967e-07, -1.4633e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 220.31, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.5040 re_mapping 0.0046 re_causal 0.0122 /// teacc 99.09 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1850, -0.1924, 0.0958, ..., -0.1063, 0.0472, 0.0360], + [-0.1169, -0.0452, -0.0792, ..., -0.1592, -0.0937, -0.0316], + [ 0.0220, -0.1127, -0.1219, ..., -0.1075, 0.0299, -0.3226], + ..., + [-0.1797, 0.1371, 0.0095, ..., 0.1548, -0.0431, -0.1266], + [-0.1346, -0.1189, 0.1461, ..., -0.1110, -0.1492, 0.1387], + [ 0.0155, -0.2519, 0.1309, ..., 0.0613, -0.1876, -0.1048]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.2352e-08, 2.6077e-08, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09], + [ 4.6566e-10, 8.3819e-09, 9.3132e-10, ..., 1.2107e-08, + 0.0000e+00, -2.4214e-08], + [ 0.0000e+00, 4.5169e-08, 5.4017e-08, ..., 4.7963e-08, + 0.0000e+00, 5.5879e-09], + ..., + [ 3.7253e-09, -3.8650e-08, -2.4680e-08, ..., -3.8650e-08, + 0.0000e+00, 1.7695e-08], + [ 6.5193e-09, 2.3795e-07, 3.3900e-07, ..., 1.3039e-08, + 4.6566e-10, -2.7940e-08], + [ 2.7940e-09, 4.1910e-09, -4.3772e-08, ..., -4.1444e-08, + 0.0000e+00, 1.2573e-08]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0186, -0.0233, -0.0259, -0.0309, -0.0048, 0.0073, 0.0105, -0.0144, + -0.0029, -0.0026], device='cuda:0'), grad: tensor([ 1.7462e-07, -4.7963e-08, 3.0734e-08, -3.0007e-06, -1.3970e-09, + 1.0785e-06, 2.4680e-08, 5.9139e-08, 1.7481e-06, -5.5414e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 220.42, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.4942 re_mapping 0.0043 re_causal 0.0122 /// teacc 99.05 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1855, -0.1929, 0.0962, ..., -0.1064, 0.0472, 0.0363], + [-0.1166, -0.0453, -0.0792, ..., -0.1607, -0.0938, -0.0315], + [ 0.0217, -0.1130, -0.1235, ..., -0.1082, 0.0300, -0.3232], + ..., + [-0.1808, 0.1374, 0.0096, ..., 0.1563, -0.0426, -0.1268], + [-0.1350, -0.1196, 0.1467, ..., -0.1120, -0.1493, 0.1397], + [ 0.0153, -0.2530, 0.1309, ..., 0.0616, -0.1879, -0.1059]], + device='cuda:0'), grad: tensor([[ 9.2667e-08, 6.0536e-09, -1.2107e-08, ..., 8.8476e-09, + 0.0000e+00, 3.7625e-07], + [ 3.5996e-07, 4.0047e-08, 1.8161e-08, ..., 5.5414e-08, + 0.0000e+00, 1.6065e-06], + [ 7.4506e-09, 3.7253e-08, 2.4680e-08, ..., 3.6322e-08, + -2.7940e-09, 3.4925e-08], + ..., + [ 5.1223e-09, -1.7788e-07, -4.8429e-08, ..., -2.3562e-07, + 1.8626e-09, 2.0489e-08], + [ 2.2352e-08, 1.3970e-09, -9.0804e-08, ..., 4.1910e-09, + 0.0000e+00, 9.5461e-08], + [ 2.9802e-08, 7.0781e-08, -6.5193e-09, ..., 4.4703e-08, + 0.0000e+00, 4.0047e-08]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0183, -0.0232, -0.0259, -0.0308, -0.0052, 0.0073, 0.0102, -0.0143, + -0.0027, -0.0026], device='cuda:0'), grad: tensor([ 1.2480e-06, 5.1484e-06, -4.8755e-07, 7.4646e-07, 5.0664e-07, + 1.3970e-06, -8.7470e-06, -3.6787e-07, 3.5809e-07, 2.1886e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 219.89, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4738 re_mapping 0.0044 re_causal 0.0119 /// teacc 99.11 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1859, -0.1933, 0.0955, ..., -0.1072, 0.0472, 0.0365], + [-0.1168, -0.0454, -0.0795, ..., -0.1616, -0.0939, -0.0314], + [ 0.0217, -0.1132, -0.1243, ..., -0.1084, 0.0300, -0.3237], + ..., + [-0.1820, 0.1376, 0.0098, ..., 0.1568, -0.0424, -0.1270], + [-0.1358, -0.1200, 0.1473, ..., -0.1124, -0.1493, 0.1381], + [ 0.0156, -0.2533, 0.1315, ..., 0.0617, -0.1882, -0.1062]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 6.9849e-09, -7.9162e-09, ..., 6.5193e-09, + 2.0955e-08, 2.1029e-06], + [ 1.3970e-09, 6.9384e-08, 3.6322e-08, ..., 6.1933e-08, + 1.3644e-07, 3.7253e-08], + [ 4.6566e-10, 3.0734e-08, 8.3819e-09, ..., 7.9162e-09, + 2.4680e-08, 3.0734e-08], + ..., + [ 1.8626e-09, -1.6019e-07, -9.9186e-08, ..., -1.7183e-07, + 2.8405e-08, 3.4459e-08], + [ 8.4285e-08, 4.7963e-08, 2.9663e-07, ..., 1.9325e-07, + 4.0978e-08, 3.7579e-07], + [-6.2864e-08, 1.0803e-07, -2.8173e-07, ..., -1.1409e-07, + 2.7008e-08, 4.6100e-08]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0188, -0.0233, -0.0259, -0.0307, -0.0053, 0.0072, 0.0112, -0.0143, + -0.0040, -0.0024], device='cuda:0'), grad: tensor([ 1.0267e-05, 1.6084e-06, 1.0571e-07, -1.0757e-07, -2.6338e-06, + 1.4864e-06, -1.3664e-05, -4.6100e-08, 2.8424e-06, 1.4948e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 220.35, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4763 re_mapping 0.0046 re_causal 0.0123 /// teacc 99.08 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1885, -0.1950, 0.0960, ..., -0.1072, 0.0473, 0.0365], + [-0.1171, -0.0455, -0.0798, ..., -0.1626, -0.0941, -0.0311], + [ 0.0212, -0.1135, -0.1264, ..., -0.1088, 0.0297, -0.3254], + ..., + [-0.1837, 0.1378, 0.0098, ..., 0.1568, -0.0429, -0.1273], + [-0.1355, -0.1206, 0.1475, ..., -0.1129, -0.1493, 0.1389], + [ 0.0159, -0.2538, 0.1324, ..., 0.0619, -0.1890, -0.1064]], + device='cuda:0'), grad: tensor([[ 1.1409e-07, 4.1910e-09, 0.0000e+00, ..., 6.9849e-09, + 0.0000e+00, 3.3528e-07], + [ 7.0781e-08, 3.4925e-08, 9.3132e-09, ..., 2.8871e-08, + 0.0000e+00, 9.9186e-08], + [ 1.7229e-08, 3.2131e-08, 9.3132e-09, ..., 2.4214e-08, + 0.0000e+00, 2.5146e-08], + ..., + [ 4.1444e-08, -1.1595e-07, -4.8894e-08, ..., -8.5682e-08, + 0.0000e+00, 3.9581e-08], + [ 6.7521e-07, 9.7789e-09, -4.1910e-09, ..., 1.8626e-09, + 0.0000e+00, 6.9337e-07], + [ 1.8021e-07, 3.8650e-08, 2.7940e-09, ..., 1.8859e-07, + 0.0000e+00, 1.3039e-07]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0182, -0.0233, -0.0260, -0.0311, -0.0052, 0.0072, 0.0109, -0.0144, + -0.0035, -0.0020], device='cuda:0'), grad: tensor([ 1.1493e-06, 3.6601e-07, -3.7253e-09, 2.0452e-06, 7.7393e-07, + -3.0845e-06, -3.6974e-06, -1.1129e-07, 1.6838e-06, 8.7684e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 219.98, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.4747 re_mapping 0.0046 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1907, -0.1962, 0.0963, ..., -0.1074, 0.0473, 0.0363], + [-0.1176, -0.0457, -0.0801, ..., -0.1640, -0.0943, -0.0308], + [ 0.0210, -0.1144, -0.1292, ..., -0.1099, 0.0298, -0.3272], + ..., + [-0.1845, 0.1381, 0.0100, ..., 0.1575, -0.0432, -0.1276], + [-0.1359, -0.1218, 0.1477, ..., -0.1136, -0.1493, 0.1394], + [ 0.0156, -0.2553, 0.1326, ..., 0.0618, -0.1893, -0.1069]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, -6.5193e-09, ..., 1.4901e-08, + 0.0000e+00, -2.7940e-09], + [ 2.3283e-09, 3.6322e-08, 2.5611e-08, ..., 3.0641e-07, + 0.0000e+00, 8.3819e-09], + [ 0.0000e+00, 4.6566e-09, 2.7940e-09, ..., 1.6764e-08, + 0.0000e+00, 1.8626e-09], + ..., + [ 7.4506e-09, -3.3993e-08, -1.3970e-08, ..., 3.3062e-08, + 0.0000e+00, 5.5879e-09], + [ 2.7940e-09, 9.3132e-09, -2.2212e-07, ..., -8.7544e-08, + 0.0000e+00, -2.1653e-07], + [ 1.1502e-07, 4.6566e-09, 1.9651e-07, ..., 1.1967e-06, + 0.0000e+00, 2.0629e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0182, -0.0234, -0.0261, -0.0310, -0.0049, 0.0072, 0.0105, -0.0144, + -0.0025, -0.0023], device='cuda:0'), grad: tensor([ 5.1688e-08, 1.8226e-06, -9.8720e-07, -8.8476e-09, -7.0482e-06, + -1.3039e-08, 4.0699e-07, 1.4715e-07, 4.6054e-07, 5.1856e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 220.05, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4874 re_mapping 0.0043 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1909, -0.1967, 0.0964, ..., -0.1075, 0.0472, 0.0366], + [-0.1175, -0.0458, -0.0801, ..., -0.1654, -0.0944, -0.0307], + [ 0.0212, -0.1154, -0.1300, ..., -0.1115, 0.0297, -0.3276], + ..., + [-0.1850, 0.1385, 0.0099, ..., 0.1580, -0.0421, -0.1279], + [-0.1362, -0.1225, 0.1474, ..., -0.1140, -0.1494, 0.1395], + [ 0.0156, -0.2555, 0.1330, ..., 0.0618, -0.1896, -0.1071]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 6.9849e-09, -3.2131e-08, ..., 7.4506e-09, + 4.6566e-10, -1.9558e-08], + [ 1.3970e-09, 4.6100e-08, 1.0710e-08, ..., 2.3283e-08, + 0.0000e+00, 2.7940e-09], + [ 4.6566e-10, 5.0291e-08, 1.6298e-08, ..., 3.3993e-08, + -1.3970e-09, 1.5832e-08], + ..., + [ 3.7253e-09, -6.7055e-08, -9.3132e-09, ..., -5.9605e-08, + 4.6566e-10, 1.3504e-08], + [ 5.5879e-09, 9.0338e-08, -1.4901e-08, ..., 9.3132e-10, + 4.6566e-10, -1.5367e-08], + [ 5.5879e-09, 1.4901e-08, -7.4506e-09, ..., 1.0710e-08, + 0.0000e+00, 1.9092e-08]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0182, -0.0236, -0.0265, -0.0310, -0.0047, 0.0071, 0.0107, -0.0140, + -0.0028, -0.0024], device='cuda:0'), grad: tensor([-4.1910e-08, 1.2852e-07, 1.5087e-07, -7.5921e-06, -1.4156e-07, + 6.8061e-06, 3.9814e-07, -6.9849e-08, 2.3143e-07, 1.4529e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 220.28, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.5020 re_mapping 0.0044 re_causal 0.0127 /// teacc 99.07 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1912, -0.1973, 0.0963, ..., -0.1079, 0.0471, 0.0348], + [-0.1176, -0.0459, -0.0802, ..., -0.1662, -0.0936, -0.0308], + [ 0.0215, -0.1160, -0.1305, ..., -0.1123, 0.0299, -0.3281], + ..., + [-0.1856, 0.1387, 0.0098, ..., 0.1585, -0.0438, -0.1282], + [-0.1361, -0.1234, 0.1474, ..., -0.1144, -0.1495, 0.1412], + [ 0.0157, -0.2563, 0.1335, ..., 0.0618, -0.1899, -0.1073]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-09, -1.1176e-08, ..., 1.3970e-09, + 0.0000e+00, 9.7789e-09], + [ 1.3970e-09, 8.3353e-08, -2.3982e-07, ..., 7.9162e-08, + 1.9558e-08, -6.8732e-07], + [ 9.3132e-10, 1.6764e-08, 6.9849e-09, ..., 1.4435e-08, + 4.1910e-09, 1.0710e-08], + ..., + [ 6.9849e-09, -8.9873e-08, 8.6613e-08, ..., -1.0105e-07, + -3.3062e-08, 2.6356e-07], + [ 1.3970e-09, 9.4064e-08, 1.6717e-07, ..., 2.3283e-09, + 4.6566e-10, 3.7299e-07], + [ 1.8626e-08, 2.1420e-08, -1.1176e-08, ..., 3.5856e-08, + 9.3132e-10, 3.3528e-08]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0193, -0.0236, -0.0266, -0.0309, -0.0047, 0.0067, 0.0110, -0.0141, + -0.0017, -0.0023], device='cuda:0'), grad: tensor([ 5.7230e-07, -3.2596e-09, -4.5486e-06, 1.7555e-07, -1.6391e-07, + -6.1467e-08, 7.9162e-08, 1.5236e-06, 1.4007e-06, 1.0021e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 220.42, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.5024 re_mapping 0.0043 re_causal 0.0121 /// teacc 99.11 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1917, -0.1984, 0.0965, ..., -0.1080, 0.0471, 0.0349], + [-0.1173, -0.0461, -0.0804, ..., -0.1672, -0.0938, -0.0306], + [ 0.0214, -0.1163, -0.1314, ..., -0.1125, 0.0301, -0.3287], + ..., + [-0.1863, 0.1389, 0.0099, ..., 0.1592, -0.0433, -0.1285], + [-0.1364, -0.1241, 0.1482, ..., -0.1141, -0.1496, 0.1417], + [ 0.0159, -0.2570, 0.1340, ..., 0.0620, -0.1902, -0.1075]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 1.3970e-09, -3.1665e-07, ..., 1.5832e-08, + -4.6566e-10, -3.7532e-07], + [ 4.6566e-10, 2.5611e-08, 1.7229e-08, ..., 4.7032e-08, + 0.0000e+00, -3.9116e-07], + [ 0.0000e+00, 3.3062e-08, 2.4214e-08, ..., 3.9116e-08, + 0.0000e+00, 2.3283e-08], + ..., + [ 4.1910e-09, -2.0210e-07, -5.7276e-08, ..., -1.9837e-07, + 0.0000e+00, 2.1048e-07], + [ 8.8476e-09, 1.4901e-08, 2.1746e-07, ..., 1.3970e-08, + 0.0000e+00, 2.8266e-07], + [-2.3749e-08, 1.1222e-07, 2.9802e-08, ..., 3.5912e-06, + 0.0000e+00, 1.8161e-08]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0193, -0.0236, -0.0266, -0.0307, -0.0049, 0.0066, 0.0107, -0.0141, + -0.0014, -0.0022], device='cuda:0'), grad: tensor([-9.6485e-07, -4.3213e-06, 2.5099e-07, 5.5879e-08, -5.9456e-06, + -1.2247e-06, 1.5851e-06, 2.0452e-06, 8.0932e-07, 7.6815e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 221.00, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.4924 re_mapping 0.0043 re_causal 0.0122 /// teacc 99.05 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1923, -0.1990, 0.0968, ..., -0.1081, 0.0470, 0.0352], + [-0.1180, -0.0464, -0.0805, ..., -0.1694, -0.0939, -0.0297], + [ 0.0220, -0.1170, -0.1319, ..., -0.1133, 0.0302, -0.3297], + ..., + [-0.1868, 0.1394, 0.0100, ..., 0.1608, -0.0435, -0.1291], + [-0.1366, -0.1246, 0.1483, ..., -0.1146, -0.1496, 0.1418], + [ 0.0161, -0.2576, 0.1342, ..., 0.0620, -0.1904, -0.1077]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -4.1910e-09, ..., 2.7940e-09, + 4.6566e-10, -7.9162e-09], + [ 3.2596e-09, 1.8626e-09, 1.9558e-08, ..., 1.3039e-08, + 1.3970e-09, -2.0023e-08], + [ 0.0000e+00, 4.6566e-10, 3.3528e-08, ..., 0.0000e+00, + -4.0513e-08, 1.9185e-07], + ..., + [ 5.5879e-09, 1.8626e-09, 2.7008e-08, ..., 1.8161e-08, + 3.8184e-08, 2.1886e-08], + [ 5.1223e-09, 4.6566e-10, -3.3993e-08, ..., 5.1223e-09, + 9.3132e-10, -2.3050e-07], + [-9.7789e-09, 4.6566e-10, -1.0477e-07, ..., -6.9849e-08, + 0.0000e+00, 8.8476e-09]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0192, -0.0237, -0.0259, -0.0309, -0.0051, 0.0068, 0.0106, -0.0140, + -0.0018, -0.0022], device='cuda:0'), grad: tensor([ 2.6543e-08, -5.5879e-09, -1.8347e-06, 2.4959e-07, 3.2596e-09, + -4.2841e-08, 1.7509e-07, 3.8743e-07, 1.2610e-06, -1.9977e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 220.00, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.5211 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.01 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1925, -0.1997, 0.0971, ..., -0.1085, 0.0466, 0.0357], + [-0.1185, -0.0465, -0.0807, ..., -0.1701, -0.0940, -0.0301], + [ 0.0216, -0.1178, -0.1325, ..., -0.1140, 0.0312, -0.3312], + ..., + [-0.1874, 0.1396, 0.0101, ..., 0.1616, -0.0464, -0.1295], + [-0.1368, -0.1262, 0.1485, ..., -0.1148, -0.1497, 0.1415], + [ 0.0172, -0.2584, 0.1348, ..., 0.0623, -0.1906, -0.1081]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.9558e-08, 6.7567e-07, ..., 1.0245e-08, + 0.0000e+00, 8.2981e-07], + [ 1.3970e-09, 8.7544e-08, 3.3993e-08, ..., 6.5193e-09, + 0.0000e+00, -3.1665e-08], + [ 0.0000e+00, 4.2375e-08, 1.0245e-07, ..., 9.3132e-10, + 0.0000e+00, 1.2387e-07], + ..., + [ 9.3132e-10, 4.0047e-08, 9.7789e-09, ..., -3.7253e-09, + 0.0000e+00, 1.8626e-08], + [ 2.3283e-09, 7.0315e-08, -1.0170e-06, ..., 4.6566e-10, + 0.0000e+00, -1.2685e-06], + [-1.9558e-08, 2.0489e-08, 1.4994e-07, ..., -5.8208e-08, + 0.0000e+00, 2.8871e-07]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0190, -0.0239, -0.0259, -0.0301, -0.0055, 0.0065, 0.0102, -0.0139, + -0.0023, -0.0019], device='cuda:0'), grad: tensor([ 1.5870e-06, 1.2526e-07, -2.0210e-07, -6.1281e-07, 1.7229e-07, + 1.0477e-07, -7.1712e-08, 1.4622e-07, -1.6391e-06, 3.8836e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 220.54, cls_loss 0.0011 cls_loss_mapping 0.0024 cls_loss_causal 0.5220 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.1926, -0.2009, 0.0977, ..., -0.1087, 0.0465, 0.0361], + [-0.1191, -0.0456, -0.0798, ..., -0.1719, -0.0944, -0.0294], + [ 0.0217, -0.1183, -0.1314, ..., -0.1147, 0.0312, -0.3316], + ..., + [-0.1893, 0.1390, 0.0094, ..., 0.1631, -0.0452, -0.1288], + [-0.1373, -0.1278, 0.1476, ..., -0.1176, -0.1499, 0.1410], + [ 0.0170, -0.2592, 0.1349, ..., 0.0622, -0.1907, -0.1093]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 1.7229e-08, 3.7253e-09, ..., 1.3504e-08, + 0.0000e+00, 1.2107e-08], + [ 9.3132e-10, 1.6205e-07, 3.7719e-08, ..., 1.3271e-07, + 0.0000e+00, -2.0955e-08], + [ 0.0000e+00, 1.0896e-07, 1.3970e-08, ..., 1.0105e-07, + 0.0000e+00, 9.3132e-09], + ..., + [ 1.8626e-09, -3.7206e-07, -5.7742e-08, ..., -3.3481e-07, + 0.0000e+00, 2.0489e-08], + [ 6.0536e-09, 1.0245e-07, 1.7695e-08, ..., 4.1444e-08, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.2806e-07, -3.6787e-08, ..., -4.6566e-09, + 0.0000e+00, 5.1223e-09]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0186, -0.0229, -0.0257, -0.0302, -0.0054, 0.0068, 0.0100, -0.0148, + -0.0034, -0.0021], device='cuda:0'), grad: tensor([ 1.0477e-07, 3.1153e-07, 8.9873e-08, -4.9919e-07, 6.4727e-08, + 2.2445e-07, -7.6368e-08, -6.4494e-07, 2.6962e-07, 1.6578e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 220.18, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5015 re_mapping 0.0043 re_causal 0.0122 /// teacc 99.09 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.1927, -0.2022, 0.0979, ..., -0.1089, 0.0465, 0.0333], + [-0.1196, -0.0465, -0.0816, ..., -0.1729, -0.0944, -0.0313], + [ 0.0219, -0.1198, -0.1332, ..., -0.1154, 0.0313, -0.3320], + ..., + [-0.1899, 0.1400, 0.0110, ..., 0.1638, -0.0454, -0.1272], + [-0.1379, -0.1288, 0.1475, ..., -0.1186, -0.1499, 0.1408], + [ 0.0171, -0.2599, 0.1352, ..., 0.0623, -0.1909, -0.1095]], + device='cuda:0'), grad: tensor([[ 1.8161e-08, 9.3132e-10, -4.3958e-07, ..., 9.3132e-10, + 4.6566e-10, -4.0838e-07], + [ 7.9162e-09, 6.0536e-09, 6.0536e-09, ..., 5.1223e-09, + 4.6566e-10, -5.5879e-08], + [ 4.6566e-09, 1.3970e-09, 3.4319e-07, ..., 4.6566e-10, + -5.4482e-08, 3.2783e-07], + ..., + [ 1.5367e-08, 4.6566e-10, 2.9337e-08, ..., 1.6764e-08, + 4.6566e-10, 4.4238e-08], + [ 7.1712e-08, 4.1910e-09, 2.0023e-08, ..., 8.3819e-09, + 0.0000e+00, 8.0094e-08], + [ 2.5099e-07, 2.3283e-09, 5.1223e-09, ..., -2.2352e-08, + 0.0000e+00, 2.6217e-07]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0206, -0.0238, -0.0251, -0.0297, -0.0053, 0.0061, 0.0130, -0.0143, + -0.0045, -0.0021], device='cuda:0'), grad: tensor([-1.7285e-06, -1.4389e-07, -2.3236e-07, 8.2422e-08, 1.5693e-06, + -1.6363e-06, 1.0338e-06, 2.1979e-07, 2.3749e-07, 5.9837e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 228---------------------------------------------------- +epoch 228, time 220.68, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4948 re_mapping 0.0046 re_causal 0.0124 /// teacc 99.16 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.1932, -0.2027, 0.0976, ..., -0.1092, 0.0464, 0.0334], + [-0.1200, -0.0466, -0.0817, ..., -0.1735, -0.0943, -0.0315], + [ 0.0216, -0.1197, -0.1337, ..., -0.1151, 0.0313, -0.3324], + ..., + [-0.1949, 0.1400, 0.0110, ..., 0.1636, -0.0455, -0.1277], + [-0.1389, -0.1294, 0.1478, ..., -0.1194, -0.1500, 0.1406], + [ 0.0170, -0.2606, 0.1358, ..., 0.0625, -0.1910, -0.1102]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -1.0245e-08, -9.7323e-08, ..., 3.7253e-09, + 0.0000e+00, -5.2107e-07], + [ 5.1223e-09, 6.0536e-09, 4.6566e-09, ..., 9.3132e-09, + 0.0000e+00, 1.0245e-08], + [ 4.6566e-10, 7.9162e-09, 2.3283e-08, ..., 8.8476e-09, + 0.0000e+00, 2.7008e-08], + ..., + [ 5.5879e-09, -7.9162e-09, 8.8476e-09, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 7.9162e-09, 7.5903e-08, 7.9162e-09, ..., 2.7474e-08, + 0.0000e+00, 3.7253e-09], + [ 2.2817e-08, 1.3039e-08, -7.7300e-08, ..., -5.4482e-08, + 0.0000e+00, 3.2596e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0207, -0.0238, -0.0248, -0.0299, -0.0052, 0.0066, 0.0129, -0.0146, + -0.0049, -0.0020], device='cuda:0'), grad: tensor([-6.0117e-07, 6.6590e-08, 1.0896e-07, 3.7253e-08, 3.5763e-07, + -3.3854e-07, 7.9628e-08, 5.8673e-08, 2.1653e-07, 1.4435e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 220.43, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.5204 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.15 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.1931, -0.2032, 0.0983, ..., -0.1090, 0.0464, 0.0343], + [-0.1194, -0.0467, -0.0819, ..., -0.1745, -0.0944, -0.0312], + [ 0.0220, -0.1202, -0.1344, ..., -0.1157, 0.0313, -0.3330], + ..., + [-0.1955, 0.1404, 0.0114, ..., 0.1646, -0.0456, -0.1277], + [-0.1395, -0.1298, 0.1476, ..., -0.1201, -0.1500, 0.1409], + [ 0.0168, -0.2612, 0.1358, ..., 0.0624, -0.1914, -0.1108]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, 7.5903e-08, 7.9162e-09, ..., 4.1910e-09, + 0.0000e+00, 2.8405e-08], + [ 3.4785e-07, 8.3819e-08, 2.9802e-08, ..., 1.1642e-08, + 0.0000e+00, 5.4576e-07], + [-1.1176e-08, 8.8010e-08, 5.2154e-08, ..., 4.6566e-09, + 0.0000e+00, 3.6787e-08], + ..., + [ 7.2177e-08, 1.1688e-07, 1.9558e-08, ..., -1.5367e-08, + 0.0000e+00, 7.0315e-08], + [ 1.2871e-06, 1.0384e-07, -1.1269e-07, ..., 1.1642e-08, + 0.0000e+00, 1.8924e-06], + [ 7.2224e-07, 4.9779e-07, 7.9162e-09, ..., -1.2387e-07, + 0.0000e+00, 6.4448e-07]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0197, -0.0238, -0.0249, -0.0300, -0.0052, 0.0066, 0.0124, -0.0144, + -0.0050, -0.0021], device='cuda:0'), grad: tensor([ 3.6508e-07, 1.3243e-06, -3.7253e-09, -3.6154e-06, 7.4925e-07, + -7.8157e-06, 8.8057e-07, 6.7055e-07, 3.7700e-06, 3.6787e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 220.56, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4989 re_mapping 0.0044 re_causal 0.0124 /// teacc 99.14 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.1931, -0.2033, 0.0989, ..., -0.1088, 0.0462, 0.0341], + [-0.1198, -0.0469, -0.0822, ..., -0.1753, -0.0940, -0.0315], + [ 0.0220, -0.1201, -0.1351, ..., -0.1156, 0.0316, -0.3335], + ..., + [-0.1961, 0.1408, 0.0116, ..., 0.1650, -0.0460, -0.1277], + [-0.1398, -0.1300, 0.1478, ..., -0.1202, -0.1502, 0.1410], + [ 0.0168, -0.2619, 0.1362, ..., 0.0625, -0.1921, -0.1111]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.6298e-09, -4.1910e-08, ..., 1.6298e-09, + 0.0000e+00, -1.1642e-08], + [ 1.1409e-08, 8.6147e-09, 3.7253e-09, ..., 6.7521e-09, + -1.1642e-09, 1.4668e-08], + [ 2.3283e-10, 1.7928e-08, 3.9581e-09, ..., -1.4901e-08, + 2.3283e-10, 3.4925e-09], + ..., + [ 1.1642e-09, 2.1886e-08, 4.6566e-10, ..., 2.3283e-09, + 4.6566e-10, 4.4238e-09], + [ 1.2573e-08, 6.9849e-10, -6.0536e-09, ..., 3.4925e-09, + 0.0000e+00, -4.6566e-10], + [ 1.8626e-09, 4.6566e-09, 1.9092e-08, ..., 6.9849e-09, + 4.6566e-10, 1.0943e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0196, -0.0239, -0.0247, -0.0301, -0.0053, 0.0067, 0.0127, -0.0144, + -0.0051, -0.0021], device='cuda:0'), grad: tensor([-4.1444e-08, 1.3504e-07, -7.3388e-07, 2.7241e-08, 7.3109e-08, + -5.9372e-08, 7.7998e-08, 3.3388e-07, 7.9861e-08, 1.2270e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 220.56, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4844 re_mapping 0.0043 re_causal 0.0120 /// teacc 99.11 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.1924, -0.2036, 0.1000, ..., -0.1088, 0.0476, 0.0344], + [-0.1206, -0.0482, -0.0837, ..., -0.1783, -0.0949, -0.0324], + [ 0.0224, -0.1201, -0.1363, ..., -0.1158, 0.0317, -0.3346], + ..., + [-0.1964, 0.1423, 0.0133, ..., 0.1679, -0.0463, -0.1265], + [-0.1403, -0.1302, 0.1478, ..., -0.1205, -0.1502, 0.1409], + [ 0.0168, -0.2631, 0.1361, ..., 0.0625, -0.1949, -0.1116]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.7940e-09, -2.3283e-09, ..., 2.7940e-09, + 4.6566e-10, 2.3283e-10], + [ 2.3283e-10, 1.0710e-08, 2.0955e-09, ..., 3.2363e-08, + 8.6147e-09, -1.4668e-08], + [ 0.0000e+00, 2.4913e-08, 6.7521e-09, ..., 1.3737e-08, + 0.0000e+00, 2.5611e-09], + ..., + [ 1.3970e-09, -6.9151e-08, -1.3039e-08, ..., -2.6543e-08, + 5.5879e-09, 4.1910e-09], + [ 3.7253e-09, 2.3283e-09, 2.0955e-09, ..., 7.2876e-08, + 2.2817e-08, 6.9849e-09], + [-4.6566e-10, 7.2177e-09, -4.8894e-09, ..., 2.0722e-08, + 6.0536e-09, 4.1910e-09]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0188, -0.0249, -0.0245, -0.0306, -0.0049, 0.0067, 0.0124, -0.0134, + -0.0055, -0.0022], device='cuda:0'), grad: tensor([ 2.2119e-08, 1.6484e-07, 4.5169e-08, 2.5146e-08, -1.1446e-06, + 2.6776e-08, -1.1409e-08, 6.3330e-08, 6.2678e-07, 1.7812e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 220.37, cls_loss 0.0010 cls_loss_mapping 0.0028 cls_loss_causal 0.5050 re_mapping 0.0043 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.1919, -0.2045, 0.1010, ..., -0.1089, 0.0474, 0.0344], + [-0.1208, -0.0485, -0.0840, ..., -0.1788, -0.0931, -0.0315], + [ 0.0225, -0.1203, -0.1370, ..., -0.1159, 0.0314, -0.3357], + ..., + [-0.1967, 0.1426, 0.0134, ..., 0.1682, -0.0467, -0.1265], + [-0.1411, -0.1307, 0.1477, ..., -0.1207, -0.1502, 0.1406], + [ 0.0168, -0.2636, 0.1364, ..., 0.0625, -0.1951, -0.1122]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, 2.2352e-08, 2.3283e-10, ..., 1.7928e-08, + 0.0000e+00, -3.7253e-09], + [ 6.7521e-09, 9.9884e-08, 3.2829e-08, ..., 7.2410e-08, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 7.3388e-07, 6.7521e-09, ..., 1.0477e-08, + 0.0000e+00, 6.7521e-09], + ..., + [ 6.9849e-09, 7.1479e-08, 7.6834e-09, ..., 7.2177e-08, + 0.0000e+00, 3.7253e-09], + [ 2.0023e-08, 4.4238e-08, 7.2177e-09, ..., 3.0268e-08, + 0.0000e+00, 1.1176e-08], + [-6.8452e-08, 3.3528e-08, -2.3353e-07, ..., 2.8890e-06, + 0.0000e+00, 4.6566e-09]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0182, -0.0249, -0.0239, -0.0307, -0.0048, 0.0063, 0.0130, -0.0135, + -0.0063, -0.0024], device='cuda:0'), grad: tensor([ 1.3807e-07, 8.2236e-07, 1.7807e-06, -5.8599e-06, -1.1623e-05, + 2.2389e-06, 7.7998e-07, 6.9616e-07, 2.7148e-07, 1.0736e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 220.09, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4984 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.02 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.1921, -0.2056, 0.1019, ..., -0.1090, 0.0474, 0.0345], + [-0.1214, -0.0486, -0.0842, ..., -0.1790, -0.0929, -0.0316], + [ 0.0221, -0.1206, -0.1378, ..., -0.1162, 0.0314, -0.3362], + ..., + [-0.1981, 0.1427, 0.0134, ..., 0.1684, -0.0469, -0.1266], + [-0.1415, -0.1310, 0.1482, ..., -0.1210, -0.1503, 0.1409], + [ 0.0168, -0.2639, 0.1365, ..., 0.0625, -0.1953, -0.1134]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 3.9581e-09, 1.6531e-08, ..., 2.3283e-10, + 0.0000e+00, -5.7742e-08], + [ 1.6298e-09, 2.8173e-08, 9.0804e-09, ..., 1.2573e-08, + 0.0000e+00, -6.2864e-09], + [ 1.3970e-09, 1.6065e-08, -8.6240e-07, ..., 1.6298e-09, + -4.6566e-10, 6.0536e-09], + ..., + [ 2.2585e-08, 1.7975e-07, 1.5600e-08, ..., -1.1409e-08, + 4.6566e-10, 4.4238e-09], + [ 4.4238e-09, 2.7474e-08, -5.1316e-07, ..., 1.3970e-09, + 0.0000e+00, -3.1702e-06], + [-3.4925e-09, 3.7253e-09, -7.6834e-09, ..., -8.6147e-09, + 0.0000e+00, 4.4238e-09]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0179, -0.0240, -0.0254, -0.0305, -0.0047, 0.0062, 0.0131, -0.0135, + -0.0063, -0.0027], device='cuda:0'), grad: tensor([ 1.4598e-07, -7.7765e-08, -4.4778e-06, -2.7497e-07, 3.7951e-08, + 3.2689e-07, 6.7204e-06, 4.3376e-07, -2.8424e-06, 1.1409e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 220.59, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.4984 re_mapping 0.0044 re_causal 0.0123 /// teacc 98.97 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.1922, -0.2063, 0.1023, ..., -0.1091, 0.0474, 0.0346], + [-0.1225, -0.0487, -0.0844, ..., -0.1793, -0.0929, -0.0318], + [ 0.0222, -0.1216, -0.1403, ..., -0.1183, 0.0314, -0.3365], + ..., + [-0.1988, 0.1431, 0.0137, ..., 0.1692, -0.0468, -0.1269], + [-0.1418, -0.1312, 0.1488, ..., -0.1213, -0.1503, 0.1412], + [ 0.0169, -0.2657, 0.1367, ..., 0.0622, -0.1954, -0.1137]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 1.1642e-09, 2.0722e-08, ..., 3.6089e-08, + 0.0000e+00, 1.2224e-07], + [ 2.3586e-07, 4.1910e-09, 1.5879e-06, ..., 3.4925e-09, + 0.0000e+00, 2.3060e-06], + [ 3.9581e-09, 5.8208e-09, 2.9104e-08, ..., 2.7940e-09, + 0.0000e+00, 4.0280e-08], + ..., + [ 1.7928e-08, -8.3819e-09, 1.1618e-07, ..., -1.3039e-08, + 0.0000e+00, 1.7509e-07], + [-2.2165e-07, 4.6566e-10, -2.2929e-06, ..., 9.3132e-10, + 0.0000e+00, -3.1907e-06], + [ 2.7940e-08, 5.1223e-09, 1.9325e-07, ..., 5.3551e-09, + 0.0000e+00, 2.8475e-07]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0177, -0.0235, -0.0266, -0.0304, -0.0048, 0.0064, 0.0123, -0.0132, + -0.0061, -0.0027], device='cuda:0'), grad: tensor([ 7.2923e-07, 2.5257e-06, 1.0040e-06, 2.5821e-07, 2.1677e-07, + 1.8114e-07, -3.6368e-07, 2.4606e-06, -7.6964e-06, 7.0035e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 220.87, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5046 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.02 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.1925, -0.2072, 0.1041, ..., -0.1092, 0.0476, 0.0347], + [-0.1229, -0.0492, -0.0846, ..., -0.1801, -0.0929, -0.0321], + [ 0.0226, -0.1226, -0.1426, ..., -0.1204, 0.0315, -0.3369], + ..., + [-0.1994, 0.1439, 0.0140, ..., 0.1705, -0.0464, -0.1270], + [-0.1433, -0.1316, 0.1497, ..., -0.1218, -0.1503, 0.1412], + [ 0.0169, -0.2675, 0.1363, ..., 0.0621, -0.1955, -0.1140]], + device='cuda:0'), grad: tensor([[ 5.3551e-09, 2.8638e-08, 8.6147e-09, ..., 2.1886e-08, + 0.0000e+00, 1.2107e-08], + [ 4.1910e-09, 1.8794e-06, 1.7928e-08, ..., 1.3029e-06, + 0.0000e+00, 5.6112e-08], + [ 6.9849e-10, 2.1979e-06, 1.8626e-09, ..., 1.5162e-06, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.6065e-08, -4.5858e-06, 5.2853e-08, ..., -3.1460e-06, + 0.0000e+00, 6.2864e-09], + [ 8.6846e-08, 1.8626e-08, 1.5274e-07, ..., 1.1013e-07, + 0.0000e+00, 6.2631e-08], + [-6.6124e-08, 8.6846e-08, -3.5646e-07, ..., -1.6182e-07, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0168, -0.0240, -0.0267, -0.0297, -0.0048, 0.0058, 0.0123, -0.0126, + -0.0062, -0.0031], device='cuda:0'), grad: tensor([ 2.5355e-07, 6.1989e-06, 4.6752e-06, 6.8359e-07, 5.3179e-07, + -1.7579e-07, -3.4403e-06, -9.3207e-06, 1.1427e-06, -5.3737e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 220.48, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4970 re_mapping 0.0040 re_causal 0.0115 /// teacc 99.16 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.1926, -0.2083, 0.1044, ..., -0.1095, 0.0475, 0.0350], + [-0.1236, -0.0494, -0.0849, ..., -0.1807, -0.0901, -0.0322], + [ 0.0223, -0.1229, -0.1432, ..., -0.1207, 0.0304, -0.3375], + ..., + [-0.1997, 0.1443, 0.0143, ..., 0.1712, -0.0471, -0.1272], + [-0.1475, -0.1320, 0.1501, ..., -0.1222, -0.1503, 0.1400], + [ 0.0172, -0.2683, 0.1368, ..., 0.0620, -0.1960, -0.1144]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3982e-08, 8.3819e-09, ..., 1.6531e-08, + 0.0000e+00, 7.6834e-09], + [ 4.6566e-10, 4.0699e-07, 1.3807e-07, ..., 4.0024e-07, + 2.3283e-10, 5.5879e-09], + [ 2.3283e-10, 2.4494e-07, 2.4214e-08, ..., 5.5647e-08, + -9.3132e-10, 1.1176e-08], + ..., + [ 2.3283e-10, -6.5379e-07, -3.0245e-07, ..., -9.7416e-07, + 4.6566e-10, 3.2596e-09], + [ 0.0000e+00, 1.3853e-07, -4.0745e-08, ..., 1.1409e-08, + 2.3283e-10, -6.1933e-08], + [ 2.3283e-10, 2.0186e-07, 6.7521e-08, ..., 2.6426e-07, + 0.0000e+00, 6.9849e-10]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0163, -0.0240, -0.0268, -0.0304, -0.0048, 0.0075, 0.0118, -0.0125, + -0.0084, -0.0032], device='cuda:0'), grad: tensor([ 1.0594e-07, 1.1530e-06, 6.2585e-07, -2.1681e-06, 3.2014e-07, + 9.9838e-07, 1.5879e-07, -2.2072e-06, 1.1805e-07, 9.0245e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 220.48, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.5089 re_mapping 0.0041 re_causal 0.0120 /// teacc 98.99 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.1917, -0.2098, 0.1051, ..., -0.1100, 0.0477, 0.0363], + [-0.1239, -0.0495, -0.0850, ..., -0.1808, -0.0901, -0.0325], + [ 0.0254, -0.1232, -0.1436, ..., -0.1204, 0.0306, -0.3379], + ..., + [-0.2002, 0.1444, 0.0143, ..., 0.1715, -0.0473, -0.1275], + [-0.1477, -0.1329, 0.1510, ..., -0.1226, -0.1503, 0.1410], + [ 0.0157, -0.2698, 0.1373, ..., 0.0620, -0.1960, -0.1147]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 6.9849e-09, -9.9419e-08, ..., 1.5832e-08, + -1.0477e-07, -9.5228e-08], + [ 3.9581e-09, 2.8638e-08, 7.9162e-09, ..., 1.6205e-07, + 6.9849e-09, 1.7462e-08], + [ 2.3283e-10, 7.2410e-08, 4.7497e-08, ..., 7.2177e-09, + 4.9826e-08, 5.6345e-08], + ..., + [ 2.3283e-09, 7.0781e-08, 4.8894e-09, ..., 3.4692e-08, + 2.0955e-09, 4.6566e-09], + [ 5.0990e-08, 3.4925e-09, 3.2596e-09, ..., 2.1118e-07, + 3.2596e-09, 2.7381e-07], + [ 8.6147e-09, 1.0012e-08, 4.6566e-10, ..., 5.6252e-07, + 2.2817e-08, 2.8173e-08]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0150, -0.0240, -0.0261, -0.0302, -0.0048, 0.0072, 0.0109, -0.0125, + -0.0079, -0.0040], device='cuda:0'), grad: tensor([-5.5879e-07, 1.0636e-06, 5.2433e-07, -3.3830e-07, -5.6922e-06, + 1.8328e-05, -1.8597e-05, 4.0140e-07, 1.6149e-06, 3.2485e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 220.56, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5179 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.1919, -0.2102, 0.1055, ..., -0.1101, 0.0479, 0.0364], + [-0.1243, -0.0495, -0.0851, ..., -0.1811, -0.0903, -0.0332], + [ 0.0253, -0.1238, -0.1459, ..., -0.1208, 0.0313, -0.3415], + ..., + [-0.2011, 0.1445, 0.0143, ..., 0.1718, -0.0477, -0.1284], + [-0.1486, -0.1332, 0.1542, ..., -0.1232, -0.1508, 0.1426], + [ 0.0158, -0.2707, 0.1375, ..., 0.0617, -0.1964, -0.1153]], + device='cuda:0'), grad: tensor([[ 8.6147e-09, 3.7253e-09, 1.0361e-07, ..., 5.8208e-08, + 6.5193e-09, 2.2585e-08], + [ 1.5134e-08, 1.9558e-08, 3.2596e-08, ..., 4.8894e-08, + 1.1642e-09, -9.5461e-09], + [ 2.5611e-09, 1.0012e-08, 1.2573e-08, ..., 1.2107e-08, + -1.5832e-08, 4.8894e-09], + ..., + [ 2.7008e-08, 8.8476e-09, 1.5018e-07, ..., 1.1642e-07, + 2.3283e-09, 2.7707e-08], + [ 5.1921e-08, 1.6298e-09, 1.0924e-06, ..., 5.1968e-07, + 2.3283e-09, 1.9697e-07], + [ 1.1642e-07, 1.2573e-08, -3.6657e-06, ..., -1.1167e-06, + 6.9849e-10, -6.3796e-07]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0148, -0.0241, -0.0262, -0.0300, -0.0044, 0.0070, 0.0111, -0.0126, + -0.0068, -0.0043], device='cuda:0'), grad: tensor([ 4.2864e-07, 2.6217e-07, -2.8173e-08, 5.1297e-06, -7.2233e-06, + 6.4913e-07, 4.4773e-07, 8.8103e-07, 2.8647e-06, -3.4012e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 221.02, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4587 re_mapping 0.0043 re_causal 0.0118 /// teacc 99.07 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.1920, -0.2122, 0.1056, ..., -0.1102, 0.0478, 0.0371], + [-0.1246, -0.0500, -0.0855, ..., -0.1814, -0.0905, -0.0339], + [ 0.0253, -0.1239, -0.1477, ..., -0.1207, 0.0322, -0.3435], + ..., + [-0.2014, 0.1451, 0.0144, ..., 0.1720, -0.0495, -0.1290], + [-0.1490, -0.1333, 0.1562, ..., -0.1241, -0.1510, 0.1435], + [ 0.0160, -0.2710, 0.1383, ..., 0.0616, -0.1967, -0.1153]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1642e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 6.9849e-10, 2.5611e-09, ..., 0.0000e+00, + -1.8626e-09, -3.2596e-09], + [ 0.0000e+00, 1.8626e-09, 4.2841e-08, ..., 0.0000e+00, + 1.1642e-09, 4.0513e-08], + ..., + [ 6.9849e-10, 1.3970e-09, 1.5367e-08, ..., 7.2177e-09, + 4.6566e-10, 8.8476e-09], + [ 2.3283e-10, 3.2596e-09, -7.4971e-08, ..., 0.0000e+00, + 0.0000e+00, -6.7754e-08], + [-1.1642e-09, 4.6566e-10, -1.0245e-08, ..., -1.1642e-08, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0143, -0.0245, -0.0261, -0.0301, -0.0041, 0.0069, 0.0107, -0.0124, + -0.0063, -0.0044], device='cuda:0'), grad: tensor([ 8.1491e-09, -3.9116e-08, 1.2363e-07, 2.1188e-08, 1.2340e-08, + 1.8859e-08, -4.6566e-10, 6.3563e-08, -1.7765e-07, -1.8859e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 220.68, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4840 re_mapping 0.0044 re_causal 0.0122 /// teacc 99.09 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.1922, -0.2132, 0.1058, ..., -0.1103, 0.0478, 0.0371], + [-0.1252, -0.0500, -0.0856, ..., -0.1815, -0.0906, -0.0337], + [ 0.0253, -0.1239, -0.1473, ..., -0.1208, 0.0326, -0.3435], + ..., + [-0.2021, 0.1452, 0.0142, ..., 0.1719, -0.0502, -0.1292], + [-0.1493, -0.1341, 0.1561, ..., -0.1245, -0.1511, 0.1437], + [ 0.0162, -0.2710, 0.1395, ..., 0.0618, -0.1967, -0.1154]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, 1.5832e-08, -1.3015e-07, ..., 3.0268e-09, + 0.0000e+00, 2.4214e-06], + [ 4.9826e-08, 8.6147e-08, 5.8208e-09, ..., 4.1910e-09, + 0.0000e+00, 3.0664e-07], + [ 3.7253e-09, 6.9151e-07, 1.0012e-08, ..., 1.8626e-09, + -4.6566e-10, 2.6776e-08], + ..., + [ 1.8394e-08, 3.1362e-07, 5.3551e-09, ..., -1.1595e-07, + 0.0000e+00, 2.9104e-08], + [ 8.2422e-07, 3.5390e-08, -4.4703e-08, ..., 2.3283e-10, + 0.0000e+00, 9.2154e-07], + [ 1.9325e-08, 2.0256e-08, -6.5193e-09, ..., 6.2864e-09, + 0.0000e+00, 3.0268e-08]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0142, -0.0245, -0.0260, -0.0297, -0.0040, 0.0065, 0.0105, -0.0125, + -0.0065, -0.0042], device='cuda:0'), grad: tensor([ 1.7077e-05, 2.0638e-06, 1.8394e-06, -3.3285e-06, 6.6962e-07, + -1.1390e-06, -1.9938e-05, 9.8534e-07, 1.6233e-06, 1.2456e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 220.23, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5155 re_mapping 0.0043 re_causal 0.0121 /// teacc 99.16 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.1928, -0.2141, 0.1061, ..., -0.1107, 0.0476, 0.0371], + [-0.1258, -0.0505, -0.0857, ..., -0.1820, -0.0901, -0.0335], + [ 0.0252, -0.1244, -0.1473, ..., -0.1212, 0.0328, -0.3439], + ..., + [-0.2025, 0.1457, 0.0141, ..., 0.1724, -0.0510, -0.1296], + [-0.1503, -0.1345, 0.1564, ..., -0.1249, -0.1513, 0.1438], + [ 0.0162, -0.2715, 0.1405, ..., 0.0621, -0.1968, -0.1158]], + device='cuda:0'), grad: tensor([[-1.3970e-09, 4.6566e-10, -1.7229e-08, ..., 1.1642e-09, + 0.0000e+00, -2.2585e-08], + [ 0.0000e+00, 2.5611e-09, 1.3970e-09, ..., 3.4925e-09, + 2.3283e-10, -4.6566e-08], + [ 2.3283e-10, 6.9849e-10, -1.6298e-09, ..., 0.0000e+00, + -1.1642e-09, 7.9162e-09], + ..., + [ 0.0000e+00, 2.3283e-10, 1.0012e-08, ..., 6.5193e-09, + 6.9849e-10, 3.0268e-09], + [ 4.6566e-10, 3.7253e-09, 2.0955e-09, ..., 2.3283e-10, + 0.0000e+00, 2.7707e-08], + [ 0.0000e+00, 1.1642e-09, -1.9791e-08, ..., 2.2817e-08, + 0.0000e+00, 3.4925e-09]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0141, -0.0248, -0.0259, -0.0299, -0.0042, 0.0066, 0.0108, -0.0122, + -0.0069, -0.0040], device='cuda:0'), grad: tensor([ 1.4435e-08, -1.3993e-07, -3.5344e-07, 6.4960e-08, -2.1770e-07, + 2.9802e-08, 7.1712e-08, 1.2969e-07, 2.0908e-07, 2.0349e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 220.14, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4903 re_mapping 0.0041 re_causal 0.0116 /// teacc 99.06 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.1927, -0.2157, 0.1067, ..., -0.1108, 0.0486, 0.0372], + [-0.1261, -0.0506, -0.0858, ..., -0.1821, -0.0905, -0.0334], + [ 0.0252, -0.1240, -0.1475, ..., -0.1212, 0.0337, -0.3442], + ..., + [-0.2029, 0.1458, 0.0142, ..., 0.1724, -0.0520, -0.1298], + [-0.1508, -0.1354, 0.1554, ..., -0.1274, -0.1514, 0.1434], + [ 0.0163, -0.2721, 0.1418, ..., 0.0624, -0.1975, -0.1157]], + device='cuda:0'), grad: tensor([[ 3.9116e-08, 1.3970e-09, -3.9116e-08, ..., 9.3132e-10, + 0.0000e+00, -3.7253e-09], + [ 1.4901e-08, 1.7695e-08, 1.4901e-08, ..., 1.8626e-08, + 0.0000e+00, 1.0710e-08], + [ 3.7719e-08, 5.1223e-09, 1.0710e-08, ..., 2.7940e-09, + 0.0000e+00, 2.4214e-08], + ..., + [ 8.5589e-07, -2.7474e-08, -3.2596e-09, ..., -2.5611e-08, + 4.6566e-10, 4.0932e-07], + [ 1.2619e-07, 6.5193e-09, -4.1910e-09, ..., 3.7253e-09, + 0.0000e+00, 3.6787e-08], + [ 4.9360e-08, 8.8476e-09, -4.0513e-08, ..., -2.3283e-08, + 0.0000e+00, 2.7474e-08]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0141, -0.0249, -0.0255, -0.0299, -0.0042, 0.0066, 0.0112, -0.0124, + -0.0077, -0.0039], device='cuda:0'), grad: tensor([ 1.7229e-08, 9.5926e-08, 1.7881e-07, 4.9500e-07, 7.5111e-07, + -5.5507e-06, 3.1432e-07, 3.1479e-06, 4.5821e-07, 1.0803e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 220.42, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4849 re_mapping 0.0041 re_causal 0.0115 /// teacc 99.04 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.1930, -0.2167, 0.1067, ..., -0.1112, 0.0485, 0.0370], + [-0.1264, -0.0507, -0.0860, ..., -0.1823, -0.0900, -0.0335], + [ 0.0252, -0.1244, -0.1479, ..., -0.1216, 0.0335, -0.3445], + ..., + [-0.2033, 0.1460, 0.0141, ..., 0.1727, -0.0519, -0.1300], + [-0.1510, -0.1362, 0.1559, ..., -0.1282, -0.1515, 0.1440], + [ 0.0163, -0.2728, 0.1429, ..., 0.0627, -0.1977, -0.1166]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 8.3819e-09, -1.2573e-08, ..., 9.3132e-10, + 4.6566e-10, 1.3970e-09], + [-5.3551e-08, 1.3970e-08, -6.8452e-08, ..., 2.7940e-09, + -1.0245e-08, -3.7858e-07], + [ 9.3132e-10, 4.6100e-08, 2.7940e-09, ..., 4.6566e-10, + 5.5879e-09, 1.5832e-08], + ..., + [ 1.4435e-08, 4.6566e-09, 2.9337e-08, ..., -2.7940e-09, + 1.3970e-09, 1.0058e-07], + [ 1.8161e-08, 9.3132e-10, -4.8429e-08, ..., -6.0536e-09, + 0.0000e+00, 4.6566e-09], + [ 5.1223e-09, 4.6566e-09, 3.2131e-08, ..., 1.8626e-09, + 0.0000e+00, 6.1933e-08]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0145, -0.0249, -0.0256, -0.0298, -0.0042, 0.0066, 0.0112, -0.0124, + -0.0075, -0.0037], device='cuda:0'), grad: tensor([ 1.4901e-08, -1.3057e-06, 2.1979e-07, -6.7754e-07, 1.9558e-08, + 8.9360e-07, 2.2631e-07, 3.9209e-07, 6.9849e-09, 2.0396e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 220.04, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4995 re_mapping 0.0043 re_causal 0.0117 /// teacc 99.01 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.1957, -0.2184, 0.1069, ..., -0.1115, 0.0483, 0.0367], + [-0.1269, -0.0511, -0.0866, ..., -0.1829, -0.0889, -0.0335], + [ 0.0252, -0.1249, -0.1484, ..., -0.1219, 0.0333, -0.3450], + ..., + [-0.2037, 0.1465, 0.0141, ..., 0.1731, -0.0523, -0.1309], + [-0.1523, -0.1372, 0.1571, ..., -0.1287, -0.1516, 0.1437], + [ 0.0165, -0.2731, 0.1448, ..., 0.0632, -0.1979, -0.1175]], + device='cuda:0'), grad: tensor([[-1.7602e-07, 4.9826e-08, -9.2201e-07, ..., 4.2375e-08, + 0.0000e+00, -8.6287e-07], + [ 6.0536e-09, 2.3143e-07, 9.4995e-08, ..., 1.7835e-07, + 0.0000e+00, -4.0978e-08], + [ 3.2596e-09, 2.0349e-07, 6.7987e-08, ..., 2.3004e-07, + 0.0000e+00, 4.0047e-08], + ..., + [ 1.1176e-08, 6.0536e-09, -1.0245e-07, ..., -8.2841e-07, + 0.0000e+00, 1.0757e-07], + [ 5.2620e-08, 5.4017e-08, 9.0804e-08, ..., 1.5832e-08, + 0.0000e+00, 1.3784e-07], + [ 3.8184e-08, 2.4401e-07, 1.7229e-07, ..., 1.0291e-07, + 0.0000e+00, 1.8533e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0151, -0.0245, -0.0263, -0.0297, -0.0048, 0.0071, 0.0107, -0.0125, + -0.0081, -0.0030], device='cuda:0'), grad: tensor([-2.7493e-06, 7.5810e-07, -2.4214e-07, -1.6475e-06, -1.0170e-06, + 1.5479e-06, 1.5795e-06, 6.0536e-08, 5.7137e-07, 1.1390e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 220.49, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.4990 re_mapping 0.0041 re_causal 0.0115 /// teacc 99.08 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.1962, -0.2215, 0.1068, ..., -0.1122, 0.0480, 0.0368], + [-0.1271, -0.0513, -0.0868, ..., -0.1833, -0.0880, -0.0313], + [ 0.0252, -0.1272, -0.1498, ..., -0.1242, 0.0328, -0.3460], + ..., + [-0.2039, 0.1473, 0.0144, ..., 0.1744, -0.0525, -0.1320], + [-0.1528, -0.1383, 0.1581, ..., -0.1302, -0.1516, 0.1446], + [ 0.0166, -0.2744, 0.1452, ..., 0.0633, -0.1980, -0.1181]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 2.3283e-09, 8.8476e-09, ..., 1.3970e-09, + 0.0000e+00, 8.8476e-09], + [ 3.7253e-09, 5.2107e-07, 1.5926e-07, ..., 4.1444e-07, + 0.0000e+00, 6.5193e-09], + [ 4.6566e-10, 8.1956e-08, 7.8697e-08, ..., 6.2864e-08, + 0.0000e+00, 5.5879e-08], + ..., + [ 1.8626e-09, -6.7567e-07, -1.9837e-07, ..., -5.3644e-07, + 0.0000e+00, 8.3819e-09], + [ 4.7963e-08, 8.3819e-09, -9.4064e-08, ..., 2.7940e-09, + 0.0000e+00, -9.2201e-08], + [-2.0629e-07, 9.3132e-09, -3.3574e-07, ..., 2.1420e-08, + 0.0000e+00, -8.3819e-08]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0157, -0.0245, -0.0260, -0.0276, -0.0049, 0.0048, 0.0094, -0.0120, + -0.0080, -0.0032], device='cuda:0'), grad: tensor([ 3.9116e-08, 1.2061e-06, 3.3714e-07, 5.3132e-07, 1.6391e-07, + 1.0105e-07, 3.0734e-08, -1.4789e-06, -2.0396e-07, -7.2364e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 220.24, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4755 re_mapping 0.0041 re_causal 0.0116 /// teacc 99.07 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.1963, -0.2227, 0.1073, ..., -0.1124, 0.0481, 0.0367], + [-0.1272, -0.0513, -0.0869, ..., -0.1835, -0.0880, -0.0313], + [ 0.0252, -0.1273, -0.1501, ..., -0.1243, 0.0329, -0.3462], + ..., + [-0.2051, 0.1474, 0.0145, ..., 0.1747, -0.0525, -0.1329], + [-0.1532, -0.1405, 0.1580, ..., -0.1311, -0.1518, 0.1445], + [ 0.0165, -0.2753, 0.1454, ..., 0.0633, -0.1983, -0.1186]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, -6.9384e-08, ..., 1.3970e-09, + 0.0000e+00, -4.3772e-08], + [ 5.5414e-08, 3.3062e-08, 3.6322e-08, ..., 2.1886e-08, + 0.0000e+00, -1.7695e-08], + [ 1.1642e-08, 4.6566e-09, 3.9581e-08, ..., 3.7253e-09, + 0.0000e+00, 2.2352e-08], + ..., + [ 1.8626e-09, -6.5658e-08, -2.5146e-08, ..., -5.1223e-08, + 0.0000e+00, 8.3819e-09], + [ 6.9849e-09, 2.7940e-09, 2.0023e-08, ..., 2.7940e-09, + 0.0000e+00, 1.0710e-08], + [-3.5064e-07, 1.0245e-08, -2.2631e-07, ..., 7.9162e-09, + 4.6566e-10, 2.0489e-08]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0155, -0.0245, -0.0259, -0.0275, -0.0050, 0.0050, 0.0097, -0.0121, + -0.0087, -0.0033], device='cuda:0'), grad: tensor([-1.7881e-07, 8.0559e-08, 1.7043e-07, 5.1782e-07, 4.2049e-07, + 2.7148e-07, -2.7753e-07, -4.7032e-08, 7.9628e-08, -1.0207e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 220.08, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4633 re_mapping 0.0042 re_causal 0.0118 /// teacc 99.11 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.1963, -0.2231, 0.1078, ..., -0.1125, 0.0478, 0.0369], + [-0.1280, -0.0508, -0.0860, ..., -0.1836, -0.0880, -0.0311], + [ 0.0252, -0.1277, -0.1508, ..., -0.1246, 0.0332, -0.3466], + ..., + [-0.2063, 0.1471, 0.0138, ..., 0.1751, -0.0526, -0.1335], + [-0.1535, -0.1411, 0.1588, ..., -0.1314, -0.1518, 0.1447], + [ 0.0168, -0.2771, 0.1458, ..., 0.0633, -0.1985, -0.1196]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 4.6566e-09, 3.5390e-08, ..., 6.5193e-09, + 0.0000e+00, 5.2154e-08], + [ 1.3970e-09, 3.7253e-09, 8.3819e-09, ..., 9.3132e-09, + 0.0000e+00, 9.3132e-10], + [ 1.3970e-09, 1.3970e-09, 1.1176e-08, ..., -1.0710e-08, + 0.0000e+00, 1.0245e-08], + ..., + [ 1.3970e-09, -1.6298e-08, 7.4506e-09, ..., -4.1910e-09, + 0.0000e+00, 1.0710e-08], + [ 1.3504e-08, -1.8626e-08, -7.9162e-08, ..., 2.3749e-08, + 0.0000e+00, -1.5367e-07], + [-1.4063e-07, 9.3132e-09, -6.1141e-07, ..., -3.3574e-07, + 0.0000e+00, 4.1910e-09]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0153, -0.0238, -0.0261, -0.0259, -0.0054, 0.0035, 0.0097, -0.0127, + -0.0087, -0.0033], device='cuda:0'), grad: tensor([ 1.9232e-07, 4.9826e-08, -1.4063e-07, 3.2829e-07, 1.4957e-06, + 3.7905e-07, -4.5355e-07, 6.3330e-08, -3.2829e-07, -1.5721e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 220.76, cls_loss 0.0011 cls_loss_mapping 0.0038 cls_loss_causal 0.5015 re_mapping 0.0040 re_causal 0.0113 /// teacc 99.16 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.1964, -0.2242, 0.1084, ..., -0.1132, 0.0477, 0.0371], + [-0.1284, -0.0509, -0.0862, ..., -0.1841, -0.0884, -0.0309], + [ 0.0252, -0.1281, -0.1509, ..., -0.1249, 0.0344, -0.3472], + ..., + [-0.2066, 0.1478, 0.0145, ..., 0.1763, -0.0526, -0.1339], + [-0.1538, -0.1415, 0.1602, ..., -0.1316, -0.1519, 0.1457], + [ 0.0175, -0.2790, 0.1490, ..., 0.0650, -0.1987, -0.1213]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.3504e-08, ..., 0.0000e+00, + -3.7253e-09, -4.6566e-10], + [ 0.0000e+00, 6.5193e-09, 1.3970e-09, ..., 6.0536e-09, + 9.3132e-10, 1.3970e-09], + [ 0.0000e+00, 2.1420e-08, 1.2573e-08, ..., 2.0489e-08, + 5.1223e-09, 2.7940e-09], + ..., + [ 4.6566e-10, -2.9337e-08, 9.3132e-10, ..., -2.7940e-08, + -3.7253e-09, 3.7253e-09], + [ 2.3283e-09, 4.6566e-10, -1.6298e-08, ..., 0.0000e+00, + 0.0000e+00, -1.3970e-08], + [ 4.6566e-10, 9.3132e-10, 4.6566e-10, ..., -9.3132e-10, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0153, -0.0240, -0.0254, -0.0261, -0.0071, 0.0035, 0.0093, -0.0125, + -0.0080, -0.0022], device='cuda:0'), grad: tensor([-4.3772e-08, 1.8626e-08, 6.3796e-08, 5.4482e-08, 1.8161e-08, + -4.4703e-08, 5.1223e-09, -4.0978e-08, -2.9802e-08, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 249---------------------------------------------------- +epoch 249, time 221.21, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4744 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.18 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.1968, -0.2256, 0.1079, ..., -0.1139, 0.0475, 0.0367], + [-0.1286, -0.0520, -0.0879, ..., -0.1844, -0.0885, -0.0320], + [ 0.0252, -0.1284, -0.1516, ..., -0.1250, 0.0344, -0.3480], + ..., + [-0.2071, 0.1491, 0.0163, ..., 0.1769, -0.0523, -0.1326], + [-0.1537, -0.1419, 0.1621, ..., -0.1318, -0.1518, 0.1469], + [ 0.0175, -0.2797, 0.1488, ..., 0.0650, -0.1989, -0.1231]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 3.0268e-08, 1.3970e-09, ..., 2.7940e-09, + 1.8626e-09, -3.7253e-09], + [ 3.7253e-09, 2.1840e-07, 7.9628e-08, ..., 2.2817e-08, + 3.2596e-09, 4.6566e-10], + [ 4.6566e-10, 2.8964e-07, 1.1316e-07, ..., -2.3283e-09, + -2.4214e-08, 3.2596e-09], + ..., + [ 8.3819e-09, -5.9418e-07, -2.1374e-07, ..., 5.0291e-08, + 1.3039e-08, 0.0000e+00], + [ 9.3132e-10, 2.3283e-09, 6.0536e-09, ..., 6.5193e-09, + 2.7940e-09, 3.7253e-09], + [ 2.0489e-08, 2.8405e-08, -3.2596e-09, ..., 9.9652e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0159, -0.0247, -0.0252, -0.0263, -0.0071, 0.0033, 0.0095, -0.0117, + -0.0069, -0.0025], device='cuda:0'), grad: tensor([ 6.9384e-08, 5.0059e-07, 3.0221e-07, 5.2154e-08, -7.1758e-07, + 5.0757e-08, 1.8906e-07, -8.4843e-07, 7.5903e-08, 3.3202e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 220.36, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4910 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.10 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.1970, -0.2264, 0.1080, ..., -0.1142, 0.0482, 0.0368], + [-0.1293, -0.0523, -0.0883, ..., -0.1849, -0.0889, -0.0322], + [ 0.0253, -0.1282, -0.1522, ..., -0.1250, 0.0345, -0.3483], + ..., + [-0.2084, 0.1494, 0.0166, ..., 0.1773, -0.0516, -0.1330], + [-0.1539, -0.1423, 0.1627, ..., -0.1327, -0.1519, 0.1473], + [ 0.0187, -0.2803, 0.1499, ..., 0.0649, -0.1995, -0.1220]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 4.4238e-08, ..., 1.8626e-09, + 0.0000e+00, 1.8161e-08], + [ 1.6298e-08, 1.3970e-09, 6.0536e-09, ..., 3.3528e-08, + 0.0000e+00, 2.3283e-09], + [ 1.0245e-08, 9.3132e-10, -5.0291e-08, ..., 5.1223e-09, + 0.0000e+00, 8.8476e-09], + ..., + [-2.0023e-08, -1.7555e-07, 2.5146e-08, ..., -1.6997e-07, + 0.0000e+00, 1.4901e-08], + [ 9.3132e-10, 0.0000e+00, -1.5274e-07, ..., 9.3132e-10, + 0.0000e+00, -1.3411e-07], + [ 4.5868e-07, 1.3970e-09, -6.0536e-08, ..., 1.2117e-06, + 0.0000e+00, 3.3528e-08]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0160, -0.0249, -0.0248, -0.0263, -0.0061, 0.0032, 0.0088, -0.0118, + -0.0070, -0.0027], device='cuda:0'), grad: tensor([ 3.3760e-07, 5.8487e-07, -1.3020e-06, 7.5437e-08, -3.7551e-06, + 5.6578e-07, 5.1223e-08, -2.1141e-07, 9.6858e-08, 3.5539e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 220.51, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4835 re_mapping 0.0038 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.1971, -0.2270, 0.1088, ..., -0.1144, 0.0483, 0.0369], + [-0.1296, -0.0524, -0.0883, ..., -0.1852, -0.0891, -0.0320], + [ 0.0254, -0.1285, -0.1532, ..., -0.1252, 0.0349, -0.3491], + ..., + [-0.2090, 0.1496, 0.0167, ..., 0.1777, -0.0526, -0.1332], + [-0.1543, -0.1427, 0.1630, ..., -0.1332, -0.1519, 0.1476], + [ 0.0187, -0.2810, 0.1499, ..., 0.0646, -0.1996, -0.1226]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 6.4261e-08, -1.0878e-06, ..., 5.4482e-08, + 0.0000e+00, -2.3516e-07], + [ 1.0245e-08, -1.4100e-06, -8.8243e-07, ..., 1.0449e-06, + 0.0000e+00, 3.1665e-08], + [ 1.4435e-08, 1.8813e-07, 6.7055e-07, ..., 1.4249e-07, + 0.0000e+00, 1.7369e-07], + ..., + [ 3.5856e-08, 2.4820e-07, 5.8394e-07, ..., -1.6689e-06, + 0.0000e+00, 1.1455e-07], + [-4.1677e-07, 8.7544e-08, -1.0533e-06, ..., 6.9849e-08, + 0.0000e+00, -1.3635e-06], + [ 6.3796e-08, 4.4284e-07, 2.7055e-07, ..., 3.6135e-07, + 0.0000e+00, 2.3609e-07]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0158, -0.0245, -0.0249, -0.0264, -0.0063, 0.0033, 0.0087, -0.0117, + -0.0070, -0.0033], device='cuda:0'), grad: tensor([-3.9972e-06, -6.5528e-06, 2.7642e-06, 1.4342e-06, -1.1642e-07, + 1.7975e-06, 6.2445e-07, 3.6415e-06, -3.1069e-06, 3.5092e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 220.31, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.5040 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.07 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.1972, -0.2276, 0.1087, ..., -0.1152, 0.0483, 0.0367], + [-0.1310, -0.0524, -0.0885, ..., -0.1854, -0.0891, -0.0323], + [ 0.0253, -0.1286, -0.1537, ..., -0.1252, 0.0351, -0.3496], + ..., + [-0.2095, 0.1497, 0.0163, ..., 0.1770, -0.0528, -0.1337], + [-0.1548, -0.1427, 0.1634, ..., -0.1337, -0.1520, 0.1478], + [ 0.0193, -0.2816, 0.1525, ..., 0.0665, -0.1997, -0.1234]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 0.0000e+00, -3.2876e-07, ..., 0.0000e+00, + 0.0000e+00, -1.1781e-07], + [ 0.0000e+00, 1.3970e-09, 2.2817e-08, ..., 2.1420e-08, + 0.0000e+00, 1.3039e-08], + [ 4.6566e-10, 1.8626e-09, 3.2131e-08, ..., 2.3283e-09, + 0.0000e+00, 2.2352e-08], + ..., + [ 4.6566e-10, -3.7253e-09, 4.6566e-09, ..., -1.3970e-09, + 0.0000e+00, 5.1223e-09], + [ 4.6566e-10, 0.0000e+00, -3.7719e-08, ..., 0.0000e+00, + 0.0000e+00, -5.8673e-08], + [ 9.3132e-10, 9.3132e-10, 2.3283e-08, ..., 6.3796e-08, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0161, -0.0247, -0.0241, -0.0272, -0.0077, 0.0042, 0.0089, -0.0125, + -0.0071, -0.0015], device='cuda:0'), grad: tensor([-1.0105e-06, 1.3970e-07, 1.1595e-07, 5.0757e-08, -2.8918e-07, + 6.5193e-08, 7.1805e-07, 2.6543e-08, -1.2899e-07, 3.1572e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 220.02, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4937 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.11 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.1976, -0.2281, 0.1077, ..., -0.1163, 0.0483, 0.0364], + [-0.1315, -0.0524, -0.0885, ..., -0.1856, -0.0892, -0.0322], + [ 0.0253, -0.1290, -0.1546, ..., -0.1254, 0.0351, -0.3501], + ..., + [-0.2065, 0.1497, 0.0163, ..., 0.1775, -0.0527, -0.1343], + [-0.1553, -0.1429, 0.1635, ..., -0.1345, -0.1520, 0.1480], + [ 0.0194, -0.2818, 0.1534, ..., 0.0668, -0.1997, -0.1232]], + device='cuda:0'), grad: tensor([[ 4.2841e-08, 8.7544e-08, 3.6322e-08, ..., 7.8697e-08, + 0.0000e+00, 2.3283e-09], + [ 4.8429e-08, 2.5565e-07, 1.0664e-07, ..., 1.9651e-07, + 0.0000e+00, 1.3504e-08], + [ 4.9360e-08, 1.1642e-07, 4.3772e-08, ..., 1.0105e-07, + 0.0000e+00, 2.7940e-09], + ..., + [-2.8266e-07, -7.9395e-07, -2.8685e-07, ..., -6.6636e-07, + 0.0000e+00, 5.1223e-09], + [ 3.6787e-08, 1.3970e-09, -1.0990e-07, ..., 2.3283e-09, + 0.0000e+00, -2.0443e-07], + [ 2.2817e-08, 4.7963e-08, 7.9162e-09, ..., 1.0990e-07, + 0.0000e+00, 1.0710e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0173, -0.0247, -0.0241, -0.0277, -0.0080, 0.0050, 0.0095, -0.0125, + -0.0072, -0.0013], device='cuda:0'), grad: tensor([ 2.8638e-07, 6.4308e-07, 3.6089e-07, 8.5356e-07, -2.3423e-07, + -4.0513e-08, 3.0734e-07, -2.2650e-06, -3.3062e-07, 4.1397e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 254---------------------------------------------------- +epoch 254, time 221.06, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4680 re_mapping 0.0042 re_causal 0.0112 /// teacc 99.19 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.1977, -0.2293, 0.1080, ..., -0.1165, 0.0483, 0.0365], + [-0.1318, -0.0525, -0.0886, ..., -0.1861, -0.0892, -0.0319], + [ 0.0252, -0.1295, -0.1551, ..., -0.1259, 0.0351, -0.3505], + ..., + [-0.2037, 0.1500, 0.0163, ..., 0.1785, -0.0527, -0.1347], + [-0.1559, -0.1432, 0.1637, ..., -0.1350, -0.1520, 0.1479], + [ 0.0193, -0.2824, 0.1536, ..., 0.0667, -0.1997, -0.1235]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.3970e-09, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, -2.5146e-08], + [ 0.0000e+00, 1.3970e-09, 1.3970e-09, ..., 2.3283e-09, + 0.0000e+00, 9.3132e-10], + ..., + [ 4.6566e-10, -6.5193e-09, -5.1223e-09, ..., -7.9162e-09, + 0.0000e+00, 1.3504e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 1.8626e-09, 1.3970e-09, ..., 4.1910e-09, + 0.0000e+00, 4.6566e-09]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0168, -0.0257, -0.0226, -0.0278, -0.0082, 0.0050, 0.0096, -0.0123, + -0.0075, -0.0014], device='cuda:0'), grad: tensor([ 3.3993e-08, -1.4016e-07, -3.0268e-08, 2.0955e-08, 3.9116e-08, + -1.3970e-09, -4.7963e-08, 7.4040e-08, 1.7695e-08, 3.9116e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 220.65, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4391 re_mapping 0.0040 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.1977, -0.2301, 0.1081, ..., -0.1166, 0.0483, 0.0364], + [-0.1328, -0.0526, -0.0886, ..., -0.1866, -0.0892, -0.0312], + [ 0.0255, -0.1298, -0.1555, ..., -0.1261, 0.0351, -0.3509], + ..., + [-0.2041, 0.1503, 0.0165, ..., 0.1787, -0.0526, -0.1354], + [-0.1566, -0.1436, 0.1639, ..., -0.1355, -0.1520, 0.1477], + [ 0.0189, -0.2842, 0.1536, ..., 0.0662, -0.1997, -0.1244]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 4.6566e-10, -8.8476e-09, ..., 4.6566e-10, + 0.0000e+00, -2.0955e-08], + [ 0.0000e+00, 7.1712e-08, 1.5832e-08, ..., 5.4017e-08, + 1.1176e-08, -5.1223e-09], + [ 0.0000e+00, 4.7963e-08, 1.0710e-08, ..., 3.5856e-08, + 7.4506e-09, 5.5879e-09], + ..., + [ 0.0000e+00, -1.4715e-07, -2.5146e-08, ..., -1.0896e-07, + -2.2817e-08, 6.0536e-09], + [ 0.0000e+00, 9.3132e-10, -1.4901e-08, ..., 9.3132e-10, + 0.0000e+00, -2.0023e-08], + [-1.3970e-09, 3.7253e-09, -3.7253e-09, ..., -9.3132e-10, + 4.6566e-10, 1.3970e-09]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0168, -0.0258, -0.0224, -0.0285, -0.0073, 0.0058, 0.0095, -0.0124, + -0.0080, -0.0022], device='cuda:0'), grad: tensor([-3.4925e-08, 3.4412e-07, -2.7148e-07, 3.4459e-08, 3.4459e-08, + 5.4482e-08, 1.5367e-08, -1.7881e-07, 4.6566e-10, 4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 256---------------------------------------------------- +epoch 256, time 221.31, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.5087 re_mapping 0.0041 re_causal 0.0115 /// teacc 99.20 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.1979, -0.2308, 0.1083, ..., -0.1168, 0.0483, 0.0364], + [-0.1328, -0.0523, -0.0885, ..., -0.1868, -0.0891, -0.0312], + [ 0.0254, -0.1300, -0.1557, ..., -0.1261, 0.0351, -0.3512], + ..., + [-0.2046, 0.1502, 0.0164, ..., 0.1788, -0.0527, -0.1357], + [-0.1576, -0.1439, 0.1642, ..., -0.1358, -0.1519, 0.1473], + [ 0.0190, -0.2856, 0.1540, ..., 0.0662, -0.1997, -0.1247]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.2596e-08, -9.4064e-08, ..., 2.1886e-08, + 0.0000e+00, 1.9465e-07], + [ 2.7940e-09, 1.4249e-07, 8.7079e-08, ..., 9.7323e-08, + 0.0000e+00, -4.0978e-08], + [ 9.3132e-10, 2.1933e-07, 1.7090e-07, ..., 1.4994e-07, + 0.0000e+00, 1.5367e-08], + ..., + [ 4.6566e-10, -6.7055e-07, -3.9069e-07, ..., -4.6333e-07, + 0.0000e+00, 1.0710e-08], + [ 1.8626e-09, 6.0536e-08, 5.6811e-08, ..., 4.0978e-08, + 0.0000e+00, 3.3993e-08], + [ 1.7649e-07, 2.2259e-07, 8.6613e-08, ..., 9.4995e-08, + 0.0000e+00, 1.0151e-07]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0170, -0.0247, -0.0229, -0.0289, -0.0077, 0.0067, 0.0083, -0.0128, + -0.0084, -0.0024], device='cuda:0'), grad: tensor([ 6.7707e-07, 4.0047e-07, 9.4576e-07, 3.1050e-06, 5.9372e-07, + -2.7996e-06, -2.5705e-06, -1.5814e-06, 3.7672e-07, 8.4611e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 220.23, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4767 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.09 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.1981, -0.2315, 0.1080, ..., -0.1173, 0.0483, 0.0361], + [-0.1330, -0.0525, -0.0887, ..., -0.1872, -0.0891, -0.0311], + [ 0.0254, -0.1304, -0.1564, ..., -0.1269, 0.0350, -0.3517], + ..., + [-0.2048, 0.1510, 0.0176, ..., 0.1808, -0.0527, -0.1359], + [-0.1578, -0.1441, 0.1643, ..., -0.1368, -0.1519, 0.1475], + [ 0.0199, -0.2892, 0.1534, ..., 0.0652, -0.1997, -0.1247]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-09, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 2.2352e-08, 8.3819e-09, ..., 8.8476e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.2596e-09, 1.8626e-09, ..., 1.3970e-09, + 0.0000e+00, 1.3970e-09], + ..., + [ 4.6566e-10, -3.6322e-08, -1.3970e-08, ..., -1.5367e-08, + 0.0000e+00, 9.3132e-10], + [-4.6566e-09, 4.6566e-10, -4.5169e-08, ..., 0.0000e+00, + 0.0000e+00, -5.2154e-08], + [ 2.7940e-09, 2.7940e-09, 6.5193e-09, ..., 1.8626e-09, + 0.0000e+00, 8.3819e-09]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0176, -0.0247, -0.0231, -0.0290, -0.0079, 0.0063, 0.0096, -0.0119, + -0.0087, -0.0031], device='cuda:0'), grad: tensor([ 3.9116e-08, 3.2363e-07, -4.5821e-07, 4.7032e-08, 1.2107e-08, + 4.0513e-08, 3.3528e-08, 2.7940e-09, -7.2177e-08, 3.4459e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 220.70, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.4658 re_mapping 0.0041 re_causal 0.0111 /// teacc 99.19 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.1983, -0.2329, 0.1087, ..., -0.1176, 0.0483, 0.0363], + [-0.1332, -0.0525, -0.0888, ..., -0.1874, -0.0891, -0.0307], + [ 0.0254, -0.1308, -0.1571, ..., -0.1273, 0.0351, -0.3524], + ..., + [-0.2056, 0.1511, 0.0176, ..., 0.1810, -0.0527, -0.1363], + [-0.1585, -0.1443, 0.1644, ..., -0.1376, -0.1520, 0.1477], + [ 0.0201, -0.2899, 0.1539, ..., 0.0653, -0.1997, -0.1254]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.8161e-08, 9.3132e-09, ..., 1.6764e-08, + 0.0000e+00, -1.3970e-09], + [ 0.0000e+00, 3.7253e-09, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10], + ..., + [ 4.6566e-10, -3.9116e-08, -1.5367e-08, ..., -5.1688e-08, + 0.0000e+00, 4.6566e-10], + [ 1.3970e-09, 4.1910e-09, 5.5879e-09, ..., 3.7253e-09, + 0.0000e+00, -1.3504e-08], + [-5.5879e-09, 3.8184e-08, -2.8871e-08, ..., 1.5367e-08, + 0.0000e+00, -1.3970e-09]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0172, -0.0245, -0.0233, -0.0291, -0.0082, 0.0063, 0.0106, -0.0120, + -0.0089, -0.0032], device='cuda:0'), grad: tensor([ 1.3970e-08, 6.2864e-08, -1.1642e-07, -1.6857e-07, 5.6345e-08, + 1.9372e-07, 6.5193e-09, -7.1246e-08, 1.7695e-08, 2.2817e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 220.39, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4810 re_mapping 0.0040 re_causal 0.0113 /// teacc 99.12 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.1986, -0.2335, 0.1090, ..., -0.1180, 0.0483, 0.0351], + [-0.1337, -0.0526, -0.0887, ..., -0.1877, -0.0891, -0.0299], + [ 0.0253, -0.1310, -0.1577, ..., -0.1275, 0.0351, -0.3528], + ..., + [-0.2059, 0.1514, 0.0178, ..., 0.1817, -0.0527, -0.1370], + [-0.1589, -0.1446, 0.1643, ..., -0.1383, -0.1520, 0.1477], + [ 0.0194, -0.2913, 0.1539, ..., 0.0642, -0.1998, -0.1258]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, -9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 7.4506e-08, 4.8429e-08, ..., 5.6345e-08, + 0.0000e+00, 4.2841e-08], + [ 0.0000e+00, 1.1642e-08, 8.7544e-08, ..., 9.3132e-09, + 0.0000e+00, 1.2014e-07], + ..., + [ 0.0000e+00, -1.9930e-07, -4.3772e-08, ..., -1.5181e-07, + 0.0000e+00, 9.3132e-09], + [ 0.0000e+00, 2.3283e-09, -1.6578e-07, ..., 1.8626e-09, + 0.0000e+00, -2.6030e-07], + [ 9.3132e-10, 3.6787e-08, 1.1176e-08, ..., 2.7940e-08, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0187, -0.0241, -0.0233, -0.0291, -0.0073, 0.0065, 0.0114, -0.0121, + -0.0093, -0.0043], device='cuda:0'), grad: tensor([-8.6799e-06, 2.9011e-07, 1.1306e-06, 3.0361e-07, 1.7975e-06, + 4.1910e-09, 5.9418e-06, -3.7486e-07, -8.3260e-07, 4.2003e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 221.00, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4950 re_mapping 0.0039 re_causal 0.0111 /// teacc 99.05 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.1983, -0.2345, 0.1106, ..., -0.1183, 0.0483, 0.0360], + [-0.1345, -0.0527, -0.0890, ..., -0.1880, -0.0891, -0.0303], + [ 0.0252, -0.1315, -0.1588, ..., -0.1277, 0.0351, -0.3581], + ..., + [-0.2062, 0.1516, 0.0179, ..., 0.1819, -0.0527, -0.1379], + [-0.1593, -0.1449, 0.1652, ..., -0.1391, -0.1520, 0.1508], + [ 0.0197, -0.2918, 0.1542, ..., 0.0641, -0.1998, -0.1264]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.1223e-09, 2.3283e-09, ..., 4.6566e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 2.7474e-08, 3.2596e-09, ..., 1.7695e-08, + 0.0000e+00, -3.2596e-09], + [-4.6566e-10, 1.0664e-07, 4.6566e-09, ..., 3.9581e-08, + 0.0000e+00, 2.7940e-09], + ..., + [ 9.3132e-10, -1.7416e-07, 7.9162e-09, ..., -5.4948e-08, + 0.0000e+00, 3.2596e-09], + [ 9.3132e-10, 2.3283e-09, -1.8626e-09, ..., 6.9849e-09, + 0.0000e+00, -3.7719e-08], + [-3.7253e-09, 5.1223e-09, -3.6787e-08, ..., 8.4611e-07, + 0.0000e+00, 2.6543e-08]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0167, -0.0240, -0.0237, -0.0293, -0.0070, 0.0061, 0.0110, -0.0124, + -0.0069, -0.0046], device='cuda:0'), grad: tensor([ 2.7940e-08, 6.5658e-08, 1.4994e-07, 5.5879e-08, -2.4308e-06, + 2.7474e-08, -5.5879e-09, -2.5379e-07, -8.3819e-09, 2.3842e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 220.38, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4957 re_mapping 0.0039 re_causal 0.0111 /// teacc 99.13 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.1987, -0.2357, 0.1099, ..., -0.1185, 0.0483, 0.0354], + [-0.1348, -0.0527, -0.0890, ..., -0.1883, -0.0891, -0.0294], + [ 0.0252, -0.1324, -0.1593, ..., -0.1287, 0.0351, -0.3585], + ..., + [-0.2063, 0.1518, 0.0180, ..., 0.1824, -0.0527, -0.1386], + [-0.1600, -0.1455, 0.1649, ..., -0.1404, -0.1520, 0.1509], + [ 0.0195, -0.2921, 0.1547, ..., 0.0639, -0.1998, -0.1273]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.7323e-08, ..., 2.7008e-08, + 0.0000e+00, -2.5658e-07], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 1.3132e-07, + 0.0000e+00, -2.3283e-09], + [ 0.0000e+00, 7.9162e-09, 4.6566e-10, ..., 2.1420e-08, + 0.0000e+00, 4.6566e-10], + ..., + [ 1.8626e-09, -2.3283e-09, 4.6566e-10, ..., 1.9046e-07, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.9558e-08, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-08, 1.3970e-09, 0.0000e+00, ..., 5.5805e-06, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0181, -0.0237, -0.0239, -0.0295, -0.0068, 0.0067, 0.0111, -0.0125, + -0.0073, -0.0048], device='cuda:0'), grad: tensor([-6.6590e-07, 4.2142e-07, -3.8650e-08, -1.2619e-07, -1.8418e-05, + 2.7660e-07, 8.7172e-07, 6.7474e-07, 7.8231e-08, 1.6928e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 220.25, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4853 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.12 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.1992, -0.2373, 0.1104, ..., -0.1187, 0.0484, 0.0355], + [-0.1351, -0.0545, -0.0901, ..., -0.1891, -0.0891, -0.0290], + [ 0.0251, -0.1332, -0.1600, ..., -0.1290, 0.0351, -0.3588], + ..., + [-0.2064, 0.1538, 0.0190, ..., 0.1836, -0.0528, -0.1391], + [-0.1604, -0.1460, 0.1653, ..., -0.1405, -0.1520, 0.1511], + [ 0.0194, -0.2936, 0.1549, ..., 0.0636, -0.1998, -0.1278]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7020e-08, 2.3283e-10, ..., 1.6298e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 7.0548e-08, 5.8208e-09, ..., 1.1409e-08, + 0.0000e+00, -7.4506e-09], + [ 6.9849e-10, 1.1805e-07, 8.8476e-09, ..., 1.7462e-08, + 0.0000e+00, 4.6566e-09], + ..., + [ 0.0000e+00, 8.1491e-09, -9.5461e-09, ..., -2.0955e-08, + 0.0000e+00, 5.1223e-09], + [-6.9849e-10, 1.8626e-08, -3.0268e-09, ..., -1.3970e-09, + 0.0000e+00, -6.9849e-10], + [ 2.3283e-10, 5.2154e-08, 2.3283e-09, ..., 1.2107e-08, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0180, -0.0249, -0.0238, -0.0294, -0.0070, 0.0078, 0.0090, -0.0109, + -0.0075, -0.0052], device='cuda:0'), grad: tensor([ 1.1991e-07, 1.6065e-07, 3.5926e-07, -9.4250e-07, -1.0757e-07, + 9.8487e-08, -1.6997e-08, 7.6601e-08, 6.1467e-08, 1.9674e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 220.24, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4835 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.13 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.1992, -0.2366, 0.1107, ..., -0.1187, 0.0484, 0.0360], + [-0.1352, -0.0546, -0.0902, ..., -0.1892, -0.0891, -0.0290], + [ 0.0251, -0.1334, -0.1603, ..., -0.1292, 0.0351, -0.3590], + ..., + [-0.2065, 0.1538, 0.0190, ..., 0.1837, -0.0528, -0.1393], + [-0.1605, -0.1464, 0.1656, ..., -0.1407, -0.1520, 0.1513], + [ 0.0194, -0.2938, 0.1551, ..., 0.0635, -0.1998, -0.1281]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 1.1642e-09, 6.9849e-09, 1.3970e-09, ..., 6.5193e-09, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 9.3132e-10, 2.3283e-10, ..., 6.9849e-10, + -9.3132e-10, 0.0000e+00], + ..., + [ 2.3283e-10, -6.0536e-09, 4.6566e-10, ..., -5.3551e-09, + 6.9849e-10, 2.3283e-10], + [ 3.7253e-09, 2.3283e-10, 4.6566e-10, ..., 2.3283e-10, + 2.3283e-10, 3.0268e-09], + [-1.6298e-09, 2.3283e-10, -2.2352e-08, ..., -1.2573e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0176, -0.0249, -0.0237, -0.0291, -0.0068, 0.0072, 0.0092, -0.0110, + -0.0074, -0.0053], device='cuda:0'), grad: tensor([ 5.5879e-09, -1.0198e-06, 1.5134e-08, 4.8662e-08, 8.3819e-07, + -5.4715e-08, 3.9581e-09, 2.0163e-07, 1.6065e-08, -3.2829e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 220.79, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4789 re_mapping 0.0041 re_causal 0.0112 /// teacc 99.12 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.1994, -0.2375, 0.1111, ..., -0.1188, 0.0483, 0.0363], + [-0.1354, -0.0546, -0.0902, ..., -0.1893, -0.0891, -0.0287], + [ 0.0251, -0.1338, -0.1612, ..., -0.1294, 0.0353, -0.3594], + ..., + [-0.2067, 0.1539, 0.0191, ..., 0.1838, -0.0531, -0.1399], + [-0.1609, -0.1470, 0.1673, ..., -0.1410, -0.1522, 0.1518], + [ 0.0197, -0.2940, 0.1555, ..., 0.0637, -0.1998, -0.1288]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 5.1223e-09, -8.2422e-08, ..., 9.3132e-09, + 6.9849e-10, -5.5181e-08], + [ 1.3039e-08, 1.2526e-07, 3.9348e-08, ..., 1.3527e-07, + 2.0955e-09, -2.3283e-10], + [ 9.3132e-10, 4.4238e-08, 6.2864e-09, ..., 3.5390e-08, + 7.2177e-08, 2.0955e-09], + ..., + [ 6.9849e-09, -4.5449e-07, -3.3062e-08, ..., -3.8906e-07, + 0.0000e+00, 1.6298e-09], + [ 2.2119e-08, 1.6298e-09, 1.1595e-07, ..., 5.1688e-08, + 1.3970e-09, 4.6333e-08], + [-1.3672e-06, 1.0361e-07, -2.9337e-06, ..., -2.7716e-06, + 0.0000e+00, 3.0268e-09]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0173, -0.0245, -0.0239, -0.0291, -0.0069, 0.0066, 0.0104, -0.0113, + -0.0072, -0.0053], device='cuda:0'), grad: tensor([-3.6554e-08, 7.6322e-07, 1.0222e-05, 3.5553e-07, 1.0110e-05, + 1.3458e-07, -1.1295e-05, -1.1409e-06, 5.7090e-07, -9.7305e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 220.71, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4769 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.10 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.1995, -0.2388, 0.1114, ..., -0.1189, 0.0483, 0.0363], + [-0.1364, -0.0550, -0.0909, ..., -0.1899, -0.0891, -0.0289], + [ 0.0250, -0.1343, -0.1618, ..., -0.1295, 0.0353, -0.3597], + ..., + [-0.2069, 0.1542, 0.0195, ..., 0.1842, -0.0531, -0.1395], + [-0.1626, -0.1483, 0.1707, ..., -0.1412, -0.1522, 0.1529], + [ 0.0184, -0.2949, 0.1548, ..., 0.0632, -0.1998, -0.1317]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.9849e-10, 6.9849e-10, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 8.3819e-09, 1.6298e-09, ..., 6.9849e-10, + 0.0000e+00, -7.9162e-09], + [ 0.0000e+00, 3.9581e-09, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 1.5134e-08, 2.7940e-09, ..., 4.6566e-10, + 0.0000e+00, 2.5611e-09], + [ 4.4703e-08, 9.3132e-10, 2.1886e-08, ..., 8.6147e-09, + 0.0000e+00, 6.3563e-08], + [-6.2864e-09, 2.0955e-09, -6.0070e-08, ..., -2.1188e-08, + 0.0000e+00, -3.4925e-08]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0173, -0.0247, -0.0240, -0.0287, -0.0066, 0.0088, 0.0075, -0.0111, + -0.0065, -0.0062], device='cuda:0'), grad: tensor([ 5.5879e-09, -7.6834e-09, -4.7730e-08, -3.3993e-08, -7.6834e-09, + -4.4238e-09, 1.9558e-08, 5.7044e-08, 1.7928e-07, -1.5437e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 220.05, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4829 re_mapping 0.0041 re_causal 0.0111 /// teacc 99.08 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.1995, -0.2406, 0.1116, ..., -0.1191, 0.0483, 0.0364], + [-0.1368, -0.0552, -0.0910, ..., -0.1902, -0.0891, -0.0290], + [ 0.0250, -0.1341, -0.1626, ..., -0.1298, 0.0355, -0.3604], + ..., + [-0.2069, 0.1544, 0.0196, ..., 0.1846, -0.0531, -0.1398], + [-0.1630, -0.1507, 0.1713, ..., -0.1425, -0.1524, 0.1542], + [ 0.0183, -0.2957, 0.1551, ..., 0.0631, -0.1998, -0.1321]], + device='cuda:0'), grad: tensor([[ 5.3551e-09, 6.9849e-10, 3.4925e-09, ..., 1.6065e-08, + 0.0000e+00, 1.0617e-07], + [ 4.8894e-09, 8.6147e-09, 2.0349e-07, ..., 4.1444e-08, + 0.0000e+00, 2.3586e-07], + [ 2.3283e-09, -9.5461e-09, 2.0489e-08, ..., 9.5461e-09, + 0.0000e+00, 2.5611e-08], + ..., + [ 3.5623e-08, 4.4238e-09, 1.8161e-08, ..., 9.0804e-09, + 0.0000e+00, 2.9104e-08], + [ 1.8626e-08, 9.3132e-10, -3.9325e-07, ..., 2.0955e-09, + 0.0000e+00, -1.1874e-08], + [ 1.0477e-08, 2.0955e-09, 3.2829e-08, ..., 1.1059e-07, + 0.0000e+00, 2.0117e-07]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0174, -0.0247, -0.0239, -0.0280, -0.0066, 0.0084, 0.0072, -0.0111, + -0.0060, -0.0063], device='cuda:0'), grad: tensor([ 4.4634e-07, 6.4913e-07, 1.3132e-07, 1.1828e-07, -7.2271e-07, + 1.6958e-05, -1.9118e-05, 6.0443e-07, -3.4459e-07, 1.2498e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 220.31, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4768 re_mapping 0.0040 re_causal 0.0104 /// teacc 99.15 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.1997, -0.2422, 0.1115, ..., -0.1193, 0.0483, 0.0361], + [-0.1375, -0.0553, -0.0912, ..., -0.1905, -0.0892, -0.0292], + [ 0.0248, -0.1349, -0.1632, ..., -0.1302, 0.0355, -0.3613], + ..., + [-0.2061, 0.1546, 0.0195, ..., 0.1851, -0.0532, -0.1407], + [-0.1643, -0.1510, 0.1732, ..., -0.1428, -0.1524, 0.1560], + [ 0.0177, -0.2962, 0.1553, ..., 0.0631, -0.1999, -0.1333]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, -3.5390e-08, ..., 7.4506e-09, + 0.0000e+00, 4.0513e-08], + [ 1.5832e-08, 1.3970e-09, 1.4016e-07, ..., 2.4680e-08, + 0.0000e+00, 1.0757e-07], + [ 4.6566e-10, 4.6566e-10, 6.0536e-09, ..., 2.3283e-09, + 0.0000e+00, -6.0583e-07], + ..., + [ 9.3132e-10, -5.5879e-09, 2.7940e-09, ..., -4.1910e-09, + 0.0000e+00, 1.5832e-07], + [ 1.3970e-09, 4.6566e-10, 4.0513e-08, ..., 3.0268e-08, + 0.0000e+00, 1.3225e-07], + [-4.1910e-09, 9.3132e-10, -2.4494e-07, ..., -7.4971e-08, + 0.0000e+00, 1.1316e-07]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0183, -0.0247, -0.0240, -0.0284, -0.0067, 0.0085, 0.0078, -0.0112, + -0.0047, -0.0068], device='cuda:0'), grad: tensor([ 9.3644e-07, 1.4734e-06, -8.3297e-06, 1.7043e-07, 3.3528e-08, + 1.3970e-07, 1.9325e-07, 2.0210e-06, 2.3134e-06, 1.0468e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 220.81, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.4686 re_mapping 0.0040 re_causal 0.0109 /// teacc 99.18 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.1998, -0.2435, 0.1120, ..., -0.1196, 0.0482, 0.0364], + [-0.1377, -0.0559, -0.0915, ..., -0.1911, -0.0893, -0.0306], + [ 0.0248, -0.1357, -0.1636, ..., -0.1305, 0.0355, -0.3616], + ..., + [-0.2052, 0.1551, 0.0196, ..., 0.1858, -0.0536, -0.1410], + [-0.1646, -0.1514, 0.1741, ..., -0.1432, -0.1525, 0.1573], + [ 0.0177, -0.2968, 0.1555, ..., 0.0631, -0.1999, -0.1336]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 2.3283e-09, -1.2573e-08, ..., 4.6566e-10, + 0.0000e+00, -4.6566e-10], + [ 8.2422e-08, 1.9558e-07, 4.1910e-09, ..., 1.2852e-07, + 0.0000e+00, 4.1910e-08], + [ 1.0245e-08, 8.4750e-08, 4.6566e-10, ..., 1.3039e-08, + 0.0000e+00, 5.1223e-09], + ..., + [ 9.7789e-09, 4.5123e-07, -1.3970e-09, ..., -1.5041e-07, + 0.0000e+00, 4.6566e-09], + [ 8.0559e-08, 1.0245e-08, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 4.0978e-08], + [ 1.6810e-07, 1.2573e-08, -2.0955e-08, ..., -7.9162e-09, + 0.0000e+00, 8.4285e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0180, -0.0253, -0.0236, -0.0283, -0.0067, 0.0085, 0.0075, -0.0109, + -0.0041, -0.0069], device='cuda:0'), grad: tensor([-1.8161e-08, 6.5612e-07, 1.4761e-07, 6.3740e-06, 7.3109e-08, + -8.1584e-06, 1.2247e-07, 2.5798e-07, 1.9930e-07, 3.6508e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 220.54, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4827 re_mapping 0.0039 re_causal 0.0113 /// teacc 98.95 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.1999, -0.2440, 0.1113, ..., -0.1202, 0.0483, 0.0354], + [-0.1380, -0.0558, -0.0915, ..., -0.1913, -0.0893, -0.0306], + [ 0.0248, -0.1357, -0.1636, ..., -0.1308, 0.0356, -0.3623], + ..., + [-0.2055, 0.1550, 0.0195, ..., 0.1860, -0.0536, -0.1411], + [-0.1648, -0.1523, 0.1754, ..., -0.1448, -0.1525, 0.1586], + [ 0.0176, -0.2970, 0.1557, ..., 0.0632, -0.1999, -0.1360]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 0.0000e+00, -1.9232e-07, ..., 0.0000e+00, + 0.0000e+00, -1.2107e-07], + [ 2.7940e-09, 1.8161e-08, 1.0245e-08, ..., 1.1176e-08, + 0.0000e+00, -5.1223e-09], + [ 1.1176e-08, 8.3819e-09, 1.2107e-08, ..., 5.1223e-09, + 0.0000e+00, 2.5611e-08], + ..., + [ 2.5611e-08, -2.9802e-08, -7.4506e-09, ..., 1.3970e-08, + 5.5879e-09, 4.7497e-08], + [ 4.2841e-08, 4.6566e-10, 1.5739e-07, ..., 4.6566e-10, + 0.0000e+00, 1.8161e-07], + [ 8.8802e-07, 2.3283e-09, 9.3132e-10, ..., 2.8405e-08, + 4.6566e-10, 1.3690e-06]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0189, -0.0252, -0.0233, -0.0282, -0.0064, 0.0083, 0.0077, -0.0112, + -0.0034, -0.0073], device='cuda:0'), grad: tensor([-4.2655e-07, -1.0338e-06, 5.6578e-07, 5.2759e-07, -7.4506e-09, + -5.4576e-06, 5.7276e-07, 4.2887e-07, 6.4820e-07, 4.1537e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 220.29, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4727 re_mapping 0.0040 re_causal 0.0111 /// teacc 99.08 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.2004, -0.2452, 0.1112, ..., -0.1205, 0.0483, 0.0352], + [-0.1394, -0.0557, -0.0914, ..., -0.1916, -0.0893, -0.0303], + [ 0.0245, -0.1380, -0.1644, ..., -0.1328, 0.0355, -0.3629], + ..., + [-0.2079, 0.1555, 0.0195, ..., 0.1865, -0.0541, -0.1416], + [-0.1652, -0.1529, 0.1764, ..., -0.1456, -0.1525, 0.1592], + [ 0.0138, -0.2973, 0.1558, ..., 0.0617, -0.2000, -0.1378]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.9802e-08, ..., 4.6566e-10, + 0.0000e+00, -1.3970e-08], + [ 0.0000e+00, 1.1642e-08, 5.5879e-09, ..., 9.7789e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, -9.8255e-08, 4.6566e-10, ..., -1.1781e-07, + 0.0000e+00, -1.9418e-07], + ..., + [ 0.0000e+00, 4.8429e-08, -1.9558e-08, ..., 7.4971e-08, + 0.0000e+00, 1.9325e-07], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 2.1886e-08, 3.2131e-08, ..., 1.6298e-08, + 0.0000e+00, 1.3039e-08]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0196, -0.0250, -0.0234, -0.0286, -0.0052, 0.0091, 0.0075, -0.0115, + -0.0030, -0.0090], device='cuda:0'), grad: tensor([ 9.3132e-09, 4.6613e-07, -5.5134e-06, 1.5181e-07, 9.7789e-08, + 4.9826e-08, -8.4285e-08, 4.4852e-06, 7.2177e-08, 2.6869e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 220.48, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4775 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.05 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.2006, -0.2479, 0.1115, ..., -0.1208, 0.0483, 0.0352], + [-0.1397, -0.0560, -0.0918, ..., -0.1921, -0.0893, -0.0306], + [ 0.0242, -0.1389, -0.1653, ..., -0.1338, 0.0355, -0.3639], + ..., + [-0.2083, 0.1565, 0.0197, ..., 0.1875, -0.0541, -0.1426], + [-0.1657, -0.1532, 0.1777, ..., -0.1438, -0.1525, 0.1601], + [ 0.0139, -0.2976, 0.1560, ..., 0.0615, -0.2000, -0.1388]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, -2.6589e-07, ..., 0.0000e+00, + 0.0000e+00, -8.7079e-08], + [ 0.0000e+00, 6.1933e-08, 2.7474e-08, ..., 3.7719e-08, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 5.0291e-08, 6.0536e-09, ..., -3.0268e-08, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, -2.1840e-07, -2.9802e-08, ..., -6.1002e-08, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 2.3283e-09, 3.7253e-09, ..., 1.3970e-09, + 0.0000e+00, 1.3970e-09], + [-4.6566e-10, 4.1910e-09, 1.0710e-08, ..., 9.3132e-10, + 0.0000e+00, 4.1910e-09]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0190, -0.0234, -0.0255, -0.0295, -0.0051, 0.0094, 0.0076, -0.0111, + -0.0022, -0.0094], device='cuda:0'), grad: tensor([-8.4937e-07, 8.8476e-09, -1.1129e-06, 1.5274e-07, -9.7789e-08, + 2.5146e-08, 8.0559e-07, 8.7824e-07, 1.4622e-07, 4.8894e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 220.50, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4886 re_mapping 0.0040 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.2006, -0.2485, 0.1124, ..., -0.1211, 0.0483, 0.0355], + [-0.1409, -0.0561, -0.0923, ..., -0.1926, -0.0893, -0.0306], + [ 0.0241, -0.1387, -0.1657, ..., -0.1339, 0.0355, -0.3642], + ..., + [-0.2088, 0.1572, 0.0209, ..., 0.1890, -0.0541, -0.1434], + [-0.1662, -0.1535, 0.1783, ..., -0.1441, -0.1525, 0.1604], + [ 0.0139, -0.3005, 0.1550, ..., 0.0611, -0.2000, -0.1395]], + device='cuda:0'), grad: tensor([[-7.4506e-09, -1.1642e-08, -7.4971e-08, ..., 4.6566e-10, + -4.6566e-10, -4.5635e-08], + [ 0.0000e+00, 2.3283e-09, 8.3819e-09, ..., 2.3283e-09, + 0.0000e+00, 7.4506e-09], + [ 9.3132e-10, 5.5879e-09, 1.2107e-08, ..., 9.3132e-10, + 0.0000e+00, 7.9162e-09], + ..., + [ 1.8626e-09, 3.7253e-09, 3.7253e-08, ..., 2.1420e-08, + 0.0000e+00, 1.1642e-08], + [ 7.4506e-09, 1.8626e-09, 1.2247e-07, ..., 8.8476e-08, + 0.0000e+00, 2.2817e-08], + [-6.0536e-09, 0.0000e+00, -1.6531e-07, ..., -1.3690e-07, + 0.0000e+00, 8.8476e-09]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0186, -0.0233, -0.0254, -0.0302, -0.0053, 0.0098, 0.0076, -0.0107, + -0.0022, -0.0100], device='cuda:0'), grad: tensor([-9.1270e-08, 4.0978e-08, -6.0070e-08, 2.0489e-08, 7.9162e-08, + 6.6357e-07, -7.8836e-07, 1.2480e-07, 3.9162e-07, -3.8370e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 220.60, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4632 re_mapping 0.0038 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.2007, -0.2492, 0.1120, ..., -0.1224, 0.0483, 0.0354], + [-0.1410, -0.0559, -0.0924, ..., -0.1928, -0.0893, -0.0300], + [ 0.0241, -0.1393, -0.1662, ..., -0.1346, 0.0355, -0.3652], + ..., + [-0.2089, 0.1572, 0.0208, ..., 0.1901, -0.0541, -0.1442], + [-0.1666, -0.1541, 0.1786, ..., -0.1443, -0.1525, 0.1606], + [ 0.0140, -0.3009, 0.1556, ..., 0.0611, -0.2000, -0.1394]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.7940e-09, 4.6566e-10, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 7.7300e-08, 0.0000e+00, ..., 8.1491e-08, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 3.5465e-06, 0.0000e+00, ..., 3.7942e-06, + 0.0000e+00, 4.6566e-10], + ..., + [ 4.6566e-10, -3.6471e-06, 0.0000e+00, ..., -3.9078e-06, + 0.0000e+00, 1.8626e-09], + [ 9.7789e-09, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 7.9162e-09], + [ 9.3132e-10, 1.7695e-08, 9.3132e-10, ..., 7.4506e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0187, -0.0234, -0.0252, -0.0301, -0.0059, 0.0098, 0.0078, -0.0106, + -0.0021, -0.0101], device='cuda:0'), grad: tensor([ 9.7789e-09, 1.6764e-07, 8.9109e-06, 3.0408e-07, 5.1223e-09, + -3.3574e-07, -9.3132e-10, -9.1344e-06, 3.3528e-08, 4.1444e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 220.14, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4683 re_mapping 0.0040 re_causal 0.0108 /// teacc 99.16 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.2010, -0.2502, 0.1114, ..., -0.1249, 0.0483, 0.0356], + [-0.1410, -0.0564, -0.0931, ..., -0.1937, -0.0894, -0.0307], + [ 0.0241, -0.1406, -0.1670, ..., -0.1369, 0.0355, -0.3656], + ..., + [-0.2092, 0.1579, 0.0213, ..., 0.1912, -0.0541, -0.1437], + [-0.1669, -0.1549, 0.1801, ..., -0.1446, -0.1525, 0.1626], + [ 0.0140, -0.3017, 0.1563, ..., 0.0613, -0.2000, -0.1395]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.0536e-09, 4.6566e-09, ..., 6.0536e-09, + 0.0000e+00, 1.2107e-08], + [ 0.0000e+00, 9.5926e-08, 2.8405e-08, ..., 0.0000e+00, + 0.0000e+00, 1.3411e-07], + ..., + [ 4.6566e-10, -2.3283e-09, -5.5879e-09, ..., -1.4901e-08, + 0.0000e+00, 4.1910e-09], + [ 4.6566e-10, 4.6566e-10, -8.2422e-08, ..., 0.0000e+00, + 0.0000e+00, -3.8976e-07], + [ 1.8626e-09, 9.7789e-09, 4.1910e-09, ..., 1.0245e-08, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0192, -0.0238, -0.0250, -0.0299, -0.0059, 0.0095, 0.0071, -0.0103, + -0.0003, -0.0101], device='cuda:0'), grad: tensor([ 4.6566e-10, 3.6787e-08, 4.6473e-07, -1.8626e-09, -2.3283e-09, + 4.2003e-07, 2.1420e-08, 1.0710e-08, -9.7137e-07, 2.3283e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 221.13, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4923 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.07 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.2012, -0.2507, 0.1125, ..., -0.1251, 0.0487, 0.0363], + [-0.1415, -0.0579, -0.0943, ..., -0.1972, -0.0894, -0.0308], + [ 0.0241, -0.1408, -0.1675, ..., -0.1371, 0.0354, -0.3659], + ..., + [-0.2087, 0.1598, 0.0226, ..., 0.1949, -0.0541, -0.1439], + [-0.1690, -0.1558, 0.1801, ..., -0.1456, -0.1526, 0.1621], + [ 0.0140, -0.3025, 0.1563, ..., 0.0609, -0.2000, -0.1398]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 1.8720e-07, ..., 1.1129e-07, + 0.0000e+00, -4.6566e-09], + [ 4.6566e-10, -3.2596e-09, 1.3784e-07, ..., 7.4971e-08, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 9.3132e-10, 1.4901e-08, ..., -6.5193e-09, + 0.0000e+00, 1.3970e-09], + ..., + [ 4.6566e-10, 7.4506e-09, 2.3786e-06, ..., 1.3039e-06, + 0.0000e+00, 4.1910e-09], + [ 4.6566e-10, 0.0000e+00, 2.4214e-08, ..., 1.5367e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, -3.2019e-06, ..., -1.7490e-06, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0183, -0.0252, -0.0250, -0.0300, -0.0059, 0.0081, 0.0087, -0.0081, + -0.0010, -0.0106], device='cuda:0'), grad: tensor([ 5.1968e-07, 3.1665e-07, -3.0268e-08, 6.9151e-07, 1.0990e-07, + 4.4284e-07, -1.8161e-08, 6.4969e-06, 6.6124e-08, -8.5607e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 220.53, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4916 re_mapping 0.0040 re_causal 0.0111 /// teacc 99.05 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.2014, -0.2514, 0.1135, ..., -0.1255, 0.0489, 0.0369], + [-0.1420, -0.0583, -0.0944, ..., -0.1976, -0.0894, -0.0307], + [ 0.0246, -0.1414, -0.1680, ..., -0.1379, 0.0352, -0.3664], + ..., + [-0.2092, 0.1604, 0.0226, ..., 0.1953, -0.0542, -0.1443], + [-0.1696, -0.1569, 0.1804, ..., -0.1463, -0.1528, 0.1622], + [ 0.0139, -0.3028, 0.1568, ..., 0.0608, -0.2002, -0.1400]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.0245e-08, ..., 8.3819e-09, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 1.6904e-07, 1.7649e-07, ..., 4.7730e-07, + 0.0000e+00, -6.0536e-09], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 3.2596e-09, + 0.0000e+00, 2.7940e-09], + ..., + [ 1.5832e-08, -1.8114e-07, -1.8720e-07, ..., -4.5542e-07, + 0.0000e+00, 2.4214e-08], + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 2.3283e-09, + 0.0000e+00, 4.1910e-09], + [ 1.9558e-08, 1.2573e-08, 1.6298e-08, ..., 2.2575e-06, + 0.0000e+00, 2.5611e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0176, -0.0254, -0.0249, -0.0298, -0.0054, 0.0080, 0.0087, -0.0079, + -0.0014, -0.0108], device='cuda:0'), grad: tensor([ 6.5193e-09, 9.4529e-07, 4.3772e-08, 1.2293e-07, -7.8753e-06, + -2.5984e-07, 5.5414e-08, -7.7765e-07, 2.2817e-08, 7.7263e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 220.52, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.5067 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.13 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.2016, -0.2514, 0.1139, ..., -0.1263, 0.0490, 0.0372], + [-0.1425, -0.0583, -0.0947, ..., -0.1976, -0.0892, -0.0307], + [ 0.0245, -0.1418, -0.1686, ..., -0.1386, 0.0351, -0.3670], + ..., + [-0.2095, 0.1606, 0.0225, ..., 0.1954, -0.0539, -0.1446], + [-0.1698, -0.1576, 0.1807, ..., -0.1471, -0.1529, 0.1624], + [ 0.0140, -0.3030, 0.1574, ..., 0.0608, -0.2002, -0.1405]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6298e-08, ..., 7.4506e-09, + -4.6566e-10, -1.0245e-08], + [ 0.0000e+00, 1.8626e-09, 4.1910e-09, ..., 4.6566e-09, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 3.7253e-09, ..., 4.6566e-10, + -9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, -9.3132e-09, 4.6566e-09, ..., -8.3819e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 9.7789e-09, ..., 8.8476e-09, + 9.3132e-10, -3.7253e-09], + [ 0.0000e+00, 9.7789e-09, -4.6100e-08, ..., -4.1444e-08, + 4.6566e-10, 4.1910e-09]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0171, -0.0252, -0.0249, -0.0300, -0.0053, 0.0079, 0.0088, -0.0083, + -0.0014, -0.0107], device='cuda:0'), grad: tensor([-4.4703e-08, 1.1176e-08, -4.1910e-09, -3.8184e-08, 5.2154e-08, + 6.3796e-08, 1.6298e-08, 3.7253e-09, 2.7940e-08, -8.7079e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 220.88, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4800 re_mapping 0.0036 re_causal 0.0106 /// teacc 99.07 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.2023, -0.2525, 0.1146, ..., -0.1268, 0.0490, 0.0373], + [-0.1433, -0.0584, -0.0948, ..., -0.1977, -0.0893, -0.0306], + [ 0.0244, -0.1416, -0.1692, ..., -0.1389, 0.0351, -0.3677], + ..., + [-0.2108, 0.1606, 0.0224, ..., 0.1954, -0.0538, -0.1451], + [-0.1704, -0.1585, 0.1808, ..., -0.1475, -0.1529, 0.1626], + [ 0.0160, -0.3031, 0.1587, ..., 0.0614, -0.2003, -0.1407]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-09, 3.2596e-09, ..., 3.2596e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 5.1782e-07, 5.5879e-09, ..., 5.9139e-07, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, 1.4622e-07, 6.0536e-09, ..., 1.5739e-07, + 0.0000e+00, 1.3970e-09], + ..., + [ 4.6566e-10, -6.8592e-07, -2.0023e-08, ..., -7.7393e-07, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 1.2573e-08, 6.0536e-09, ..., 3.2596e-09, + 0.0000e+00, 1.3970e-09], + [ 1.8626e-09, 4.0513e-08, 2.7940e-09, ..., 2.3283e-09, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0166, -0.0252, -0.0247, -0.0300, -0.0050, 0.0078, 0.0085, -0.0086, + -0.0014, -0.0102], device='cuda:0'), grad: tensor([ 1.9558e-08, 1.2107e-06, 3.6135e-07, -1.3690e-07, 2.1420e-08, + 2.6543e-08, -5.1223e-09, -1.6121e-06, 4.1910e-08, 8.2422e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 220.50, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.5053 re_mapping 0.0037 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.2037, -0.2532, 0.1143, ..., -0.1276, 0.0484, 0.0369], + [-0.1457, -0.0569, -0.0946, ..., -0.1979, -0.0891, -0.0300], + [ 0.0216, -0.1422, -0.1713, ..., -0.1396, 0.0339, -0.3710], + ..., + [-0.2113, 0.1596, 0.0220, ..., 0.1954, -0.0538, -0.1465], + [-0.1721, -0.1594, 0.1809, ..., -0.1484, -0.1531, 0.1626], + [ 0.0164, -0.3033, 0.1614, ..., 0.0637, -0.2003, -0.1406]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.8626e-09, -1.1642e-07, ..., 6.5193e-09, + 0.0000e+00, -5.8673e-08], + [ 9.3132e-10, 3.1199e-08, 2.3749e-08, ..., 2.8405e-08, + 0.0000e+00, 1.1176e-08], + [ 0.0000e+00, -7.2643e-08, 8.3819e-09, ..., -7.2177e-08, + 0.0000e+00, -7.9162e-09], + ..., + [ 1.3970e-09, 4.9360e-08, 5.8208e-08, ..., 8.3353e-08, + 0.0000e+00, 2.7008e-08], + [ 7.4506e-09, 5.5879e-09, 3.3993e-08, ..., 2.5146e-08, + 0.0000e+00, 2.2352e-08], + [-9.3132e-10, 9.3132e-10, -2.5099e-07, ..., -3.6322e-07, + 0.0000e+00, -1.2573e-08]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0163, -0.0245, -0.0253, -0.0294, -0.0065, 0.0080, 0.0083, -0.0099, + -0.0018, -0.0081], device='cuda:0'), grad: tensor([-3.0873e-07, 3.7532e-07, -1.2759e-06, 2.7940e-08, 5.5647e-07, + 1.2573e-08, 4.0047e-08, 1.0990e-06, 1.4668e-07, -6.7474e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 221.02, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4744 re_mapping 0.0037 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.2039, -0.2538, 0.1131, ..., -0.1302, 0.0484, 0.0364], + [-0.1459, -0.0569, -0.0946, ..., -0.1980, -0.0891, -0.0297], + [ 0.0216, -0.1422, -0.1721, ..., -0.1388, 0.0341, -0.3708], + ..., + [-0.2117, 0.1596, 0.0220, ..., 0.1953, -0.0542, -0.1477], + [-0.1722, -0.1597, 0.1811, ..., -0.1490, -0.1531, 0.1629], + [ 0.0164, -0.3035, 0.1622, ..., 0.0631, -0.2010, -0.1409]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-09, -1.9092e-08, ..., 1.4901e-08, + 0.0000e+00, -9.3132e-09], + [ 1.3970e-09, 1.0245e-08, 3.7253e-09, ..., 5.4482e-08, + 0.0000e+00, -1.8626e-09], + [ 1.3970e-09, 1.8161e-08, 6.0536e-09, ..., 2.9337e-08, + 0.0000e+00, 1.8626e-09], + ..., + [ 4.6566e-09, 1.0245e-08, 1.3504e-08, ..., 3.0734e-08, + 0.0000e+00, 5.1223e-09], + [ 2.3283e-09, 4.1910e-09, 1.1735e-07, ..., 9.8720e-08, + 0.0000e+00, 2.7474e-08], + [-2.2352e-08, 3.7253e-09, -2.2585e-07, ..., 1.3039e-07, + 0.0000e+00, -5.1223e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0177, -0.0244, -0.0250, -0.0295, -0.0058, 0.0079, 0.0086, -0.0103, + -0.0017, -0.0085], device='cuda:0'), grad: tensor([-1.1176e-08, 1.6671e-07, 1.0943e-07, -2.5937e-07, -1.2312e-06, + 2.5472e-07, 6.4261e-08, 1.4668e-07, 2.6124e-07, 5.0152e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 220.64, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4430 re_mapping 0.0039 re_causal 0.0105 /// teacc 99.13 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.2044, -0.2545, 0.1155, ..., -0.1310, 0.0485, 0.0372], + [-0.1462, -0.0569, -0.0948, ..., -0.1982, -0.0888, -0.0294], + [ 0.0216, -0.1426, -0.1755, ..., -0.1394, 0.0338, -0.3711], + ..., + [-0.2118, 0.1597, 0.0221, ..., 0.1956, -0.0544, -0.1483], + [-0.1726, -0.1601, 0.1813, ..., -0.1496, -0.1532, 0.1630], + [ 0.0162, -0.3040, 0.1618, ..., 0.0623, -0.2018, -0.1422]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, -6.3563e-07, ..., 1.3970e-09, + 0.0000e+00, -1.7276e-07], + [ 0.0000e+00, 2.7008e-08, 1.7695e-08, ..., 1.8626e-08, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 1.3085e-07, 6.2585e-07, ..., 5.9139e-08, + 0.0000e+00, 1.2480e-07], + ..., + [ 0.0000e+00, -2.4494e-07, -1.0803e-07, ..., -1.3458e-07, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, 5.3085e-08, 2.7008e-08, ..., 2.3283e-08, + 0.0000e+00, 2.4214e-08], + [-6.9849e-09, 2.5146e-08, 2.6543e-08, ..., 2.4214e-08, + 0.0000e+00, 6.0536e-09]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0162, -0.0243, -0.0251, -0.0294, -0.0059, 0.0095, 0.0078, -0.0105, + -0.0018, -0.0098], device='cuda:0'), grad: tensor([-1.8757e-06, 8.7079e-08, 1.5115e-06, 1.1828e-07, 1.6298e-08, + 3.2596e-08, 1.9558e-08, -5.2946e-07, 4.9733e-07, 1.2713e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 220.32, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4963 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.12 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.2056, -0.2549, 0.1167, ..., -0.1311, 0.0487, 0.0374], + [-0.1464, -0.0569, -0.0948, ..., -0.1983, -0.0887, -0.0290], + [ 0.0215, -0.1428, -0.1773, ..., -0.1394, 0.0338, -0.3716], + ..., + [-0.2119, 0.1598, 0.0222, ..., 0.1957, -0.0547, -0.1486], + [-0.1733, -0.1610, 0.1812, ..., -0.1511, -0.1532, 0.1631], + [ 0.0160, -0.3043, 0.1618, ..., 0.0622, -0.2020, -0.1426]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.9162e-09, ..., 4.6566e-10, + 0.0000e+00, 2.6971e-06], + [ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 1.8626e-09, + 0.0000e+00, -1.3039e-08], + [ 0.0000e+00, -4.6566e-09, 0.0000e+00, ..., -9.7789e-09, + 0.0000e+00, 2.7940e-09], + ..., + [ 0.0000e+00, 2.3283e-09, 2.1420e-08, ..., 1.9092e-08, + 0.0000e+00, 1.3970e-09], + [-9.3132e-10, 0.0000e+00, -4.1910e-09, ..., 0.0000e+00, + 0.0000e+00, 6.0536e-09], + [ 4.6566e-10, 0.0000e+00, -2.4680e-08, ..., -9.7789e-09, + 0.0000e+00, 4.6566e-09]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0156, -0.0243, -0.0250, -0.0296, -0.0056, 0.0096, 0.0077, -0.0107, + -0.0023, -0.0102], device='cuda:0'), grad: tensor([ 9.8869e-06, -4.1910e-08, -7.2643e-08, 1.2107e-08, 2.7940e-08, + -4.6100e-08, -9.9093e-06, 1.0291e-07, 4.1444e-08, -1.1642e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 220.40, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4982 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.05 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.2060, -0.2568, 0.1177, ..., -0.1315, 0.0489, 0.0379], + [-0.1469, -0.0570, -0.0949, ..., -0.1984, -0.0887, -0.0287], + [ 0.0217, -0.1436, -0.1778, ..., -0.1398, 0.0339, -0.3719], + ..., + [-0.2123, 0.1601, 0.0226, ..., 0.1960, -0.0548, -0.1490], + [-0.1741, -0.1621, 0.1808, ..., -0.1523, -0.1532, 0.1630], + [ 0.0156, -0.3054, 0.1618, ..., 0.0606, -0.2020, -0.1430]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 9.3132e-10, -4.6566e-10, ..., 2.7940e-09, + 0.0000e+00, 2.7940e-09], + [ 3.2596e-09, 1.1642e-08, 3.7253e-09, ..., 1.5832e-08, + 0.0000e+00, 2.3283e-09], + [ 3.3062e-08, 7.2643e-08, 1.8626e-09, ..., 8.8941e-08, + 0.0000e+00, 2.3283e-09], + ..., + [-3.9581e-08, -1.2107e-07, -9.3132e-09, ..., -1.5972e-07, + 0.0000e+00, 1.3970e-09], + [ 1.1176e-08, 4.6566e-10, -3.5390e-08, ..., 9.3132e-10, + 0.0000e+00, -2.5146e-08], + [ 2.8266e-07, 1.6764e-08, 3.2596e-08, ..., 1.3970e-07, + 0.0000e+00, 1.8626e-07]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0150, -0.0243, -0.0250, -0.0290, -0.0045, 0.0096, 0.0077, -0.0106, + -0.0027, -0.0118], device='cuda:0'), grad: tensor([ 1.5832e-08, 6.3330e-08, 3.1479e-07, 1.0151e-07, -4.1956e-07, + -6.6776e-07, 3.8184e-08, -5.0524e-07, -3.7719e-08, 1.1064e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 220.21, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.4685 re_mapping 0.0036 re_causal 0.0104 /// teacc 99.08 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.2062, -0.2578, 0.1156, ..., -0.1341, 0.0487, 0.0380], + [-0.1473, -0.0574, -0.0953, ..., -0.1986, -0.0887, -0.0287], + [ 0.0217, -0.1445, -0.1786, ..., -0.1408, 0.0341, -0.3721], + ..., + [-0.2123, 0.1606, 0.0228, ..., 0.1964, -0.0548, -0.1493], + [-0.1746, -0.1627, 0.1810, ..., -0.1526, -0.1532, 0.1628], + [ 0.0156, -0.3057, 0.1637, ..., 0.0612, -0.2020, -0.1431]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, -9.8441e-07, ..., 9.7789e-09, + 0.0000e+00, -1.3905e-06], + [ 1.3970e-09, 2.3283e-09, 1.0710e-08, ..., 2.7940e-09, + 0.0000e+00, 1.3970e-08], + [ 4.6566e-10, 9.3132e-10, 4.6566e-09, ..., 4.1910e-09, + 0.0000e+00, 6.5193e-09], + ..., + [ 4.6566e-09, 2.3283e-09, 3.0268e-08, ..., 4.1910e-09, + 0.0000e+00, 3.7253e-08], + [ 9.3132e-10, 9.3132e-10, 3.9442e-07, ..., 1.3970e-09, + 0.0000e+00, 5.5460e-07], + [-5.0291e-08, 9.3132e-09, -6.9849e-08, ..., 3.5064e-07, + 0.0000e+00, 1.7229e-08]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0164, -0.0245, -0.0247, -0.0292, -0.0053, 0.0096, 0.0077, -0.0102, + -0.0033, -0.0111], device='cuda:0'), grad: tensor([-3.2671e-06, 5.0291e-08, 3.1199e-08, -1.4529e-07, -1.1008e-06, + 7.9349e-07, 1.2573e-06, 1.1222e-07, 1.3318e-06, 9.4250e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 220.38, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4817 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.04 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.2064, -0.2580, 0.1160, ..., -0.1342, 0.0488, 0.0386], + [-0.1492, -0.0577, -0.0959, ..., -0.1991, -0.0879, -0.0289], + [ 0.0218, -0.1449, -0.1793, ..., -0.1409, 0.0337, -0.3720], + ..., + [-0.2131, 0.1608, 0.0228, ..., 0.1969, -0.0549, -0.1501], + [-0.1748, -0.1629, 0.1816, ..., -0.1524, -0.1532, 0.1631], + [ 0.0159, -0.3061, 0.1643, ..., 0.0613, -0.2020, -0.1437]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.7987e-08, 4.9826e-08, ..., 7.1246e-08, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 4.6566e-10, -1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -8.8941e-08, -6.4261e-08, ..., -9.2667e-08, + 0.0000e+00, -1.3970e-09], + [ 4.6566e-10, 5.5879e-09, 5.1223e-09, ..., 6.0536e-09, + 0.0000e+00, 4.6566e-10], + [-4.1910e-09, 9.7789e-09, 1.3970e-09, ..., 9.7789e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0161, -0.0249, -0.0244, -0.0293, -0.0055, 0.0096, 0.0078, -0.0101, + -0.0034, -0.0110], device='cuda:0'), grad: tensor([ 3.7719e-08, 1.8300e-07, -3.7253e-08, 9.3132e-09, 1.0096e-06, + 1.5134e-07, -1.1791e-06, -2.5611e-07, 4.0047e-08, 4.0047e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 220.25, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4385 re_mapping 0.0038 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.2072, -0.2590, 0.1158, ..., -0.1346, 0.0487, 0.0382], + [-0.1502, -0.0578, -0.0961, ..., -0.1992, -0.0877, -0.0287], + [ 0.0218, -0.1453, -0.1797, ..., -0.1413, 0.0336, -0.3722], + ..., + [-0.2133, 0.1609, 0.0228, ..., 0.1970, -0.0550, -0.1504], + [-0.1756, -0.1632, 0.1817, ..., -0.1528, -0.1532, 0.1630], + [ 0.0156, -0.3065, 0.1647, ..., 0.0610, -0.2021, -0.1440]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 3.2596e-09, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 3.4925e-08, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, -3.7253e-09, ..., -2.7940e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0166, -0.0248, -0.0244, -0.0293, -0.0048, 0.0095, 0.0078, -0.0103, + -0.0037, -0.0115], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.1409e-07, -1.1735e-07, -7.6368e-08, 3.7253e-09, + 2.3749e-08, -1.8626e-09, 5.8208e-08, 6.9849e-09, -3.2596e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 220.48, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4865 re_mapping 0.0036 re_causal 0.0105 /// teacc 99.09 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.2088, -0.2603, 0.1156, ..., -0.1349, 0.0487, 0.0378], + [-0.1507, -0.0580, -0.0962, ..., -0.1993, -0.0876, -0.0293], + [ 0.0214, -0.1450, -0.1808, ..., -0.1418, 0.0336, -0.3727], + ..., + [-0.2137, 0.1610, 0.0228, ..., 0.1972, -0.0551, -0.1506], + [-0.1760, -0.1636, 0.1819, ..., -0.1533, -0.1532, 0.1635], + [ 0.0161, -0.3068, 0.1663, ..., 0.0619, -0.2021, -0.1442]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.5611e-09, 6.7521e-09, ..., 9.3132e-10, + 0.0000e+00, 2.3516e-08], + [ 9.3132e-10, 9.5228e-08, 6.9384e-08, ..., 6.4960e-08, + 0.0000e+00, 2.4005e-07], + [ 0.0000e+00, 3.5856e-08, 3.4925e-09, ..., 2.5146e-08, + -1.1642e-09, 1.1874e-08], + ..., + [ 2.3283e-10, -1.1944e-07, 1.0710e-08, ..., -9.1502e-08, + 9.3132e-10, 3.7951e-08], + [ 1.6298e-09, -8.0559e-08, -2.3865e-07, ..., 9.3132e-10, + 0.0000e+00, -8.3260e-07], + [ 1.2340e-08, 4.6566e-09, -1.5367e-08, ..., 2.5914e-07, + 0.0000e+00, 2.3749e-08]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0172, -0.0250, -0.0239, -0.0291, -0.0051, 0.0095, 0.0078, -0.0104, + -0.0041, -0.0111], device='cuda:0'), grad: tensor([ 8.3121e-08, 8.5169e-07, 9.8255e-08, 5.4948e-07, -1.3374e-06, + 8.1584e-07, 2.3586e-07, -2.9337e-08, -2.6934e-06, 1.4259e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 220.84, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4893 re_mapping 0.0036 re_causal 0.0103 /// teacc 99.07 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.2095, -0.2627, 0.1153, ..., -0.1352, 0.0486, 0.0372], + [-0.1515, -0.0579, -0.0963, ..., -0.1995, -0.0876, -0.0296], + [ 0.0214, -0.1452, -0.1812, ..., -0.1426, 0.0337, -0.3729], + ..., + [-0.2139, 0.1611, 0.0229, ..., 0.1976, -0.0551, -0.1509], + [-0.1770, -0.1649, 0.1834, ..., -0.1538, -0.1532, 0.1641], + [ 0.0162, -0.3078, 0.1666, ..., 0.0620, -0.2021, -0.1461]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 6.9849e-10, 2.7940e-09, ..., 2.2352e-08, + 0.0000e+00, 3.7253e-09], + [ 6.3796e-08, 1.6997e-08, 8.6147e-09, ..., 2.0629e-07, + 0.0000e+00, 6.5193e-09], + [ 3.2596e-09, 4.1910e-09, 1.0012e-08, ..., 2.3283e-08, + 0.0000e+00, 2.0955e-09], + ..., + [ 6.1234e-08, 1.0012e-07, 1.2503e-07, ..., 2.7358e-07, + 0.0000e+00, 3.7253e-09], + [ 8.3819e-09, 4.6566e-10, 3.7253e-09, ..., 2.9337e-08, + 0.0000e+00, -1.8626e-09], + [ 3.3760e-08, 5.8208e-09, -1.9278e-07, ..., -4.4471e-08, + 0.0000e+00, 3.4925e-09]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0184, -0.0250, -0.0237, -0.0293, -0.0054, 0.0095, 0.0079, -0.0105, + -0.0037, -0.0116], device='cuda:0'), grad: tensor([ 1.1176e-07, 9.7323e-07, 7.5903e-08, -1.5576e-07, -3.4329e-06, + 2.3632e-07, 6.7707e-07, 1.2554e-06, 1.2689e-07, 1.4156e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 220.83, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4774 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.04 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.2100, -0.2657, 0.1152, ..., -0.1356, 0.0487, 0.0372], + [-0.1486, -0.0604, -0.0975, ..., -0.2011, -0.0875, -0.0299], + [ 0.0207, -0.1461, -0.1830, ..., -0.1442, 0.0337, -0.3734], + ..., + [-0.2174, 0.1636, 0.0234, ..., 0.1992, -0.0552, -0.1520], + [-0.1780, -0.1656, 0.1854, ..., -0.1535, -0.1532, 0.1655], + [ 0.0171, -0.3081, 0.1677, ..., 0.0617, -0.2021, -0.1477]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3271e-08, 4.1910e-09, ..., 9.3132e-10, + 0.0000e+00, 3.2596e-09], + [ 1.1642e-09, 1.8161e-08, 3.0268e-09, ..., 4.6566e-10, + 0.0000e+00, 2.4680e-07], + [ 4.6566e-10, 3.9581e-09, 3.4925e-09, ..., 2.3283e-10, + 0.0000e+00, 1.3504e-08], + ..., + [ 4.6566e-10, 1.9092e-08, 1.8626e-09, ..., 6.9849e-10, + 0.0000e+00, 1.1874e-08], + [ 2.0955e-09, 2.8173e-08, -9.7789e-09, ..., 4.6566e-10, + 0.0000e+00, -1.5022e-06], + [-1.9791e-08, 4.1910e-09, -4.4703e-08, ..., -1.0477e-08, + 0.0000e+00, 7.5204e-08]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0188, -0.0263, -0.0238, -0.0300, -0.0050, 0.0095, 0.0079, -0.0092, + -0.0024, -0.0114], device='cuda:0'), grad: tensor([ 1.4063e-07, 7.9582e-07, -1.3150e-06, -1.5041e-06, 1.3458e-07, + 3.1125e-06, 2.2016e-06, 1.2806e-07, -3.7737e-06, 8.8708e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 221.39, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4808 re_mapping 0.0037 re_causal 0.0101 /// teacc 99.06 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.2113, -0.2670, 0.1153, ..., -0.1358, 0.0487, 0.0372], + [-0.1482, -0.0607, -0.0981, ..., -0.2013, -0.0873, -0.0304], + [ 0.0207, -0.1466, -0.1838, ..., -0.1446, 0.0336, -0.3737], + ..., + [-0.2187, 0.1638, 0.0236, ..., 0.1993, -0.0553, -0.1526], + [-0.1828, -0.1670, 0.1858, ..., -0.1539, -0.1533, 0.1623], + [ 0.0172, -0.3087, 0.1686, ..., 0.0620, -0.2021, -0.1482]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, -1.2806e-08, ..., 0.0000e+00, + -2.5611e-09, -1.2806e-08], + [ 0.0000e+00, 4.4238e-09, 2.3283e-10, ..., 6.9849e-10, + 0.0000e+00, -1.8859e-08], + [ 0.0000e+00, -1.6764e-08, 5.5879e-09, ..., -1.7462e-08, + 9.3132e-10, 1.2573e-08], + ..., + [ 0.0000e+00, 2.0256e-08, -2.3283e-10, ..., 1.6065e-08, + 0.0000e+00, 5.1223e-09], + [ 0.0000e+00, 9.3132e-10, -1.0245e-08, ..., -3.0268e-09, + 2.3283e-10, -1.1874e-08], + [ 2.3283e-10, 2.7940e-09, 7.9162e-09, ..., 2.5611e-09, + 2.3283e-10, 1.0012e-08]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0191, -0.0259, -0.0238, -0.0299, -0.0062, 0.0099, 0.0078, -0.0100, + -0.0059, -0.0113], device='cuda:0'), grad: tensor([-4.1677e-08, -5.4715e-08, -7.4273e-08, -5.4017e-08, 2.6776e-08, + 5.4948e-08, -4.4238e-09, 1.4040e-07, -1.4668e-08, 2.8405e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 220.68, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4471 re_mapping 0.0036 re_causal 0.0101 /// teacc 99.19 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.2115, -0.2677, 0.1153, ..., -0.1359, 0.0486, 0.0371], + [-0.1483, -0.0606, -0.0979, ..., -0.2014, -0.0872, -0.0290], + [ 0.0206, -0.1469, -0.1845, ..., -0.1452, 0.0335, -0.3738], + ..., + [-0.2186, 0.1638, 0.0233, ..., 0.1995, -0.0550, -0.1542], + [-0.1829, -0.1678, 0.1862, ..., -0.1542, -0.1533, 0.1623], + [ 0.0175, -0.3089, 0.1691, ..., 0.0623, -0.2022, -0.1485]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 8.4750e-08, 6.9849e-10, ..., 1.3970e-09, + 0.0000e+00, 9.3132e-10], + [ 6.9849e-10, 1.2317e-07, 2.3283e-10, ..., 1.3970e-09, + 0.0000e+00, -1.9558e-08], + [ 6.9849e-10, 2.8242e-07, 1.1642e-09, ..., 2.3283e-09, + 0.0000e+00, 6.9849e-10], + ..., + [ 6.5193e-09, 9.8255e-08, 2.3283e-09, ..., 5.1223e-09, + 0.0000e+00, 2.2585e-08], + [ 1.3970e-09, 2.3050e-08, 1.3970e-09, ..., 3.0268e-09, + 0.0000e+00, -1.6298e-09], + [ 3.2596e-08, 2.7707e-08, -2.7008e-08, ..., 4.0513e-08, + 0.0000e+00, 1.2340e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0193, -0.0255, -0.0236, -0.0295, -0.0066, 0.0099, 0.0078, -0.0104, + -0.0060, -0.0111], device='cuda:0'), grad: tensor([ 3.3597e-07, 3.0082e-07, 9.7696e-07, -2.6394e-06, -1.9395e-07, + 1.8626e-09, 1.6973e-07, 6.6590e-07, 1.0105e-07, 2.7963e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 220.97, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4761 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.10 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.2116, -0.2683, 0.1154, ..., -0.1363, 0.0487, 0.0377], + [-0.1490, -0.0607, -0.0980, ..., -0.2015, -0.0872, -0.0289], + [ 0.0204, -0.1478, -0.1855, ..., -0.1461, 0.0334, -0.3742], + ..., + [-0.2184, 0.1641, 0.0231, ..., 0.1998, -0.0548, -0.1548], + [-0.1829, -0.1683, 0.1865, ..., -0.1547, -0.1534, 0.1624], + [ 0.0177, -0.3092, 0.1698, ..., 0.0625, -0.2022, -0.1486]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 4.6566e-10, + 0.0000e+00, 7.2177e-09], + [ 6.9849e-10, 1.7229e-08, 4.4238e-09, ..., 1.4901e-08, + -2.3283e-10, -4.6566e-09], + [ 2.3283e-10, 1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 1.6298e-09], + ..., + [ 6.9849e-10, -1.7229e-08, -2.3283e-10, ..., -1.3737e-08, + 0.0000e+00, 2.5611e-09], + [ 9.3132e-10, 2.3283e-10, -6.9849e-10, ..., 2.0955e-09, + 0.0000e+00, -7.4506e-09], + [-1.7928e-08, 9.3132e-10, -5.4482e-08, ..., -1.5367e-08, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0189, -0.0255, -0.0240, -0.0300, -0.0068, 0.0099, 0.0078, -0.0102, + -0.0060, -0.0110], device='cuda:0'), grad: tensor([ 1.5134e-08, 1.2806e-08, 1.2573e-08, 7.0548e-08, 4.5169e-08, + -9.3132e-10, -1.9092e-08, -2.3283e-08, 1.8626e-09, -9.6159e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 220.61, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.5011 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.02 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.2120, -0.2694, 0.1156, ..., -0.1368, 0.0487, 0.0380], + [-0.1491, -0.0618, -0.0981, ..., -0.2028, -0.0871, -0.0278], + [ 0.0203, -0.1486, -0.1865, ..., -0.1469, 0.0334, -0.3746], + ..., + [-0.2189, 0.1654, 0.0232, ..., 0.2012, -0.0548, -0.1562], + [-0.1829, -0.1684, 0.1873, ..., -0.1550, -0.1530, 0.1624], + [ 0.0179, -0.3096, 0.1700, ..., 0.0625, -0.2023, -0.1490]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 5.8208e-09, ..., 3.4925e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 2.0955e-09, 2.0955e-09, ..., 1.8626e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 9.5461e-09, 2.6776e-08, ..., 1.3039e-08, + 0.0000e+00, 2.4447e-08], + ..., + [-2.5611e-09, -1.6298e-08, 3.0268e-09, ..., -1.7229e-08, + 0.0000e+00, 2.0955e-09], + [ 0.0000e+00, 6.9849e-10, -5.7975e-08, ..., -1.3039e-08, + 0.0000e+00, -5.4250e-08], + [ 2.5611e-09, 6.5193e-09, -1.9791e-08, ..., -3.7253e-09, + 0.0000e+00, 4.1910e-09]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0189, -0.0250, -0.0249, -0.0303, -0.0069, 0.0098, 0.0078, -0.0098, + -0.0059, -0.0111], device='cuda:0'), grad: tensor([ 1.5134e-08, 7.6834e-09, 1.0221e-07, 3.8883e-08, 6.5193e-09, + 1.5367e-08, 7.1712e-08, -2.3516e-08, -1.7881e-07, -2.9337e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 220.69, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4783 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.11 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.2123, -0.2695, 0.1160, ..., -0.1369, 0.0485, 0.0386], + [-0.1493, -0.0624, -0.0990, ..., -0.2028, -0.0869, -0.0285], + [ 0.0203, -0.1492, -0.1872, ..., -0.1474, 0.0333, -0.3750], + ..., + [-0.2189, 0.1660, 0.0239, ..., 0.2013, -0.0549, -0.1556], + [-0.1830, -0.1688, 0.1877, ..., -0.1551, -0.1530, 0.1625], + [ 0.0175, -0.3098, 0.1700, ..., 0.0625, -0.2023, -0.1498]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 2.3283e-09, ..., 1.3970e-09, + 0.0000e+00, 6.0536e-09], + [ 0.0000e+00, 8.1491e-09, -9.3132e-09, ..., 6.2864e-09, + 0.0000e+00, -1.9209e-07], + [ 0.0000e+00, 3.9581e-09, 4.6566e-10, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-09], + ..., + [ 6.9849e-10, -1.6298e-09, 3.4459e-08, ..., 2.6310e-08, + 0.0000e+00, 6.2864e-09], + [ 4.1910e-09, 4.8894e-09, 3.0035e-08, ..., 2.5611e-08, + 0.0000e+00, 7.5437e-08], + [-6.9849e-09, 6.7521e-09, -9.2434e-08, ..., -8.4052e-08, + 0.0000e+00, -2.4447e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0186, -0.0257, -0.0242, -0.0307, -0.0070, 0.0099, 0.0078, -0.0096, + -0.0059, -0.0114], device='cuda:0'), grad: tensor([ 3.5623e-08, -7.6834e-07, -4.2608e-08, -7.1898e-07, 1.9325e-08, + 7.3155e-07, 4.8522e-07, 1.3295e-07, 3.8696e-07, -2.5542e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 220.73, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5134 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.2125, -0.2697, 0.1161, ..., -0.1370, 0.0487, 0.0384], + [-0.1493, -0.0624, -0.0989, ..., -0.2030, -0.0865, -0.0268], + [ 0.0200, -0.1495, -0.1885, ..., -0.1481, 0.0333, -0.3753], + ..., + [-0.2190, 0.1662, 0.0239, ..., 0.2016, -0.0550, -0.1561], + [-0.1832, -0.1698, 0.1875, ..., -0.1557, -0.1534, 0.1624], + [ 0.0173, -0.3101, 0.1706, ..., 0.0624, -0.2024, -0.1503]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.3283e-09, ..., 6.9849e-10, + 0.0000e+00, 1.1874e-08], + [ 6.9849e-10, 0.0000e+00, 5.3551e-09, ..., 2.3283e-10, + 0.0000e+00, 9.0804e-09], + [-1.2806e-08, 0.0000e+00, 1.0245e-08, ..., 4.6566e-10, + 0.0000e+00, 1.6997e-08], + ..., + [ 1.1642e-09, 0.0000e+00, 3.9581e-09, ..., 2.7940e-09, + 0.0000e+00, 2.3283e-09], + [ 3.2596e-09, 0.0000e+00, -1.2573e-07, ..., 6.9849e-10, + 0.0000e+00, -3.0175e-07], + [ 1.3039e-08, 0.0000e+00, -6.9849e-09, ..., -7.2177e-09, + 0.0000e+00, 5.8208e-09]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0189, -0.0248, -0.0249, -0.0323, -0.0067, 0.0100, 0.0078, -0.0099, + -0.0063, -0.0116], device='cuda:0'), grad: tensor([ 5.5647e-08, 3.3062e-08, -9.8022e-08, 5.5879e-08, 2.2119e-08, + 9.3365e-08, 5.1269e-07, 1.6531e-08, -7.9721e-07, 1.0571e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 220.81, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4788 re_mapping 0.0037 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.2130, -0.2704, 0.1161, ..., -0.1377, 0.0487, 0.0385], + [-0.1497, -0.0628, -0.0991, ..., -0.2042, -0.0863, -0.0270], + [ 0.0199, -0.1500, -0.1890, ..., -0.1493, 0.0333, -0.3756], + ..., + [-0.2194, 0.1666, 0.0237, ..., 0.2027, -0.0548, -0.1571], + [-0.1833, -0.1699, 0.1881, ..., -0.1558, -0.1536, 0.1623], + [ 0.0180, -0.3105, 0.1715, ..., 0.0629, -0.2026, -0.1503]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 2.3283e-10, 3.9581e-09, ..., 2.0955e-09, + 4.6566e-10, 4.6566e-09], + [ 3.4925e-09, 4.1910e-09, 1.2573e-08, ..., 4.4238e-09, + -3.0268e-09, 7.2177e-09], + [ 6.9849e-10, 4.6566e-10, 2.5611e-09, ..., 4.6566e-10, + 0.0000e+00, 3.7253e-09], + ..., + [ 4.4238e-08, 4.1910e-09, 1.2596e-07, ..., 8.5915e-08, + 2.3283e-10, 3.2596e-08], + [-5.4482e-08, 2.3283e-10, -1.7020e-07, ..., 2.5611e-09, + 2.3283e-10, -2.6333e-07], + [-5.0757e-08, -1.2107e-08, -1.4366e-07, ..., -1.4668e-07, + 0.0000e+00, 4.6100e-08]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0192, -0.0252, -0.0249, -0.0323, -0.0069, 0.0100, 0.0078, -0.0097, + -0.0064, -0.0110], device='cuda:0'), grad: tensor([ 2.0955e-08, 2.6543e-08, 4.8894e-09, 9.4762e-08, 1.6228e-07, + 4.2398e-07, -5.2387e-08, 4.0722e-07, -6.7893e-07, -4.0350e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 220.38, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4921 re_mapping 0.0038 re_causal 0.0104 /// teacc 99.13 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.2133, -0.2713, 0.1166, ..., -0.1381, 0.0491, 0.0388], + [-0.1497, -0.0627, -0.0992, ..., -0.2042, -0.0862, -0.0260], + [ 0.0199, -0.1502, -0.1904, ..., -0.1497, 0.0334, -0.3759], + ..., + [-0.2195, 0.1666, 0.0236, ..., 0.2027, -0.0552, -0.1585], + [-0.1834, -0.1699, 0.1885, ..., -0.1560, -0.1536, 0.1623], + [ 0.0174, -0.3109, 0.1721, ..., 0.0621, -0.2027, -0.1506]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -3.7253e-09, ..., 4.6566e-10, + 0.0000e+00, -4.1910e-09], + [ 2.3283e-09, 2.3283e-09, 2.3283e-09, ..., 3.2596e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10], + ..., + [ 4.6566e-09, -2.7940e-09, 8.3819e-09, ..., 2.3283e-09, + 0.0000e+00, 1.3970e-09], + [ 4.6566e-10, -4.6566e-10, -2.7940e-09, ..., -4.6566e-10, + 0.0000e+00, -9.3132e-10], + [-1.3970e-08, 4.6566e-10, -2.0023e-08, ..., -1.2107e-08, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0190, -0.0250, -0.0246, -0.0323, -0.0058, 0.0100, 0.0078, -0.0100, + -0.0064, -0.0118], device='cuda:0'), grad: tensor([-1.8161e-08, 1.6764e-08, 3.2596e-09, 1.0245e-08, 2.4680e-08, + 7.9162e-08, -7.9162e-08, 1.7695e-08, -2.3283e-09, -5.4482e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 220.30, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4586 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.08 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.2136, -0.2722, 0.1184, ..., -0.1376, 0.0491, 0.0407], + [-0.1509, -0.0625, -0.1001, ..., -0.2042, -0.0862, -0.0264], + [ 0.0199, -0.1518, -0.1923, ..., -0.1517, 0.0337, -0.3764], + ..., + [-0.2192, 0.1667, 0.0234, ..., 0.2030, -0.0555, -0.1598], + [-0.1834, -0.1700, 0.1905, ..., -0.1565, -0.1539, 0.1625], + [ 0.0179, -0.3111, 0.1714, ..., 0.0621, -0.2027, -0.1534]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, -9.7789e-09, -2.4680e-08, ..., 0.0000e+00, + 0.0000e+00, -3.4459e-08], + [-9.3132e-10, 1.8626e-09, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, -4.6566e-10], + ..., + [ 4.6566e-10, 9.3132e-09, 1.9092e-08, ..., 1.3970e-09, + 0.0000e+00, 2.3749e-08], + [ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, -2.3283e-09], + [-4.6566e-10, 3.2596e-09, 3.2596e-09, ..., -1.8626e-09, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0172, -0.0251, -0.0249, -0.0322, -0.0059, 0.0100, 0.0078, -0.0101, + -0.0063, -0.0128], device='cuda:0'), grad: tensor([ 4.6566e-09, -7.1712e-08, -2.3469e-07, 1.2107e-08, 1.3504e-08, + 3.3528e-08, 6.1933e-08, 1.4622e-07, 2.3283e-09, 4.2841e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 220.57, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4561 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.08 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.2141, -0.2731, 0.1176, ..., -0.1399, 0.0498, 0.0411], + [-0.1509, -0.0625, -0.1001, ..., -0.2043, -0.0862, -0.0240], + [ 0.0197, -0.1527, -0.1936, ..., -0.1524, 0.0337, -0.3770], + ..., + [-0.2194, 0.1671, 0.0233, ..., 0.2032, -0.0555, -0.1616], + [-0.1836, -0.1705, 0.1907, ..., -0.1573, -0.1539, 0.1624], + [ 0.0174, -0.3115, 0.1726, ..., 0.0628, -0.2028, -0.1549]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -1.7975e-07, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09], + [ 2.3283e-09, 4.6566e-10, 8.4285e-08, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-08], + [ 9.3132e-10, 0.0000e+00, 5.1223e-09, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + ..., + [ 2.3283e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.4435e-08], + [ 1.0291e-07, 0.0000e+00, 6.0536e-09, ..., 1.8626e-09, + 0.0000e+00, 8.3353e-08], + [ 3.2596e-09, 0.0000e+00, -9.3132e-09, ..., -6.5193e-09, + 0.0000e+00, 1.3970e-08]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0178, -0.0244, -0.0250, -0.0326, -0.0059, 0.0100, 0.0078, -0.0107, + -0.0066, -0.0123], device='cuda:0'), grad: tensor([-1.1269e-06, 2.6077e-07, 7.9162e-08, 2.0815e-07, 5.5414e-08, + -4.4983e-07, 5.9512e-07, 7.8231e-08, 2.4866e-07, 4.9826e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 220.59, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5051 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.08 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.2160, -0.2781, 0.1176, ..., -0.1403, 0.0488, 0.0409], + [-0.1510, -0.0632, -0.0995, ..., -0.2051, -0.0841, -0.0206], + [ 0.0195, -0.1567, -0.1957, ..., -0.1547, 0.0328, -0.3782], + ..., + [-0.2197, 0.1692, 0.0245, ..., 0.2048, -0.0568, -0.1623], + [-0.1839, -0.1742, 0.1886, ..., -0.1581, -0.1545, 0.1619], + [ 0.0173, -0.3146, 0.1728, ..., 0.0622, -0.2033, -0.1556]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.3970e-09, -2.1886e-08, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, -2.2398e-07, -4.6100e-08, ..., 2.0955e-08, + -2.3283e-09, -4.0047e-08], + [ 0.0000e+00, 5.1223e-09, 5.1223e-09, ..., -2.7940e-09, + 4.6566e-10, 1.6764e-08], + ..., + [ 4.6566e-10, 1.8673e-07, 2.7008e-08, ..., -3.1665e-08, + 1.3970e-09, 1.1176e-08], + [ 3.2596e-09, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 9.3132e-10, 1.1176e-08], + [-4.6566e-09, 2.1420e-08, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0181, -0.0231, -0.0267, -0.0321, -0.0052, 0.0100, 0.0078, -0.0099, + -0.0081, -0.0130], device='cuda:0'), grad: tensor([-2.9802e-08, -9.8161e-07, -5.2527e-07, 9.0804e-08, 9.3132e-08, + -4.6194e-06, 4.9770e-06, 8.8103e-07, 4.0513e-08, 6.3330e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 220.28, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4722 re_mapping 0.0035 re_causal 0.0097 /// teacc 99.09 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.2162, -0.2793, 0.1177, ..., -0.1405, 0.0490, 0.0412], + [-0.1511, -0.0633, -0.0997, ..., -0.2052, -0.0841, -0.0206], + [ 0.0196, -0.1570, -0.1957, ..., -0.1547, 0.0324, -0.3771], + ..., + [-0.2196, 0.1696, 0.0250, ..., 0.2052, -0.0557, -0.1628], + [-0.1838, -0.1748, 0.1895, ..., -0.1580, -0.1546, 0.1620], + [ 0.0176, -0.3155, 0.1727, ..., 0.0614, -0.2035, -0.1566]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.2573e-08, 4.4238e-08, ..., 2.2817e-08, + 0.0000e+00, 5.9605e-08], + [ 5.5879e-09, 9.3132e-09, 3.5856e-08, ..., 1.3504e-08, + 0.0000e+00, 3.9581e-08], + [ 2.3283e-09, 1.7695e-08, 4.8894e-08, ..., 1.3970e-08, + 0.0000e+00, 4.6566e-08], + ..., + [ 6.9849e-09, -1.1967e-07, -1.8673e-07, ..., -1.6345e-07, + 0.0000e+00, 3.7719e-08], + [ 3.0268e-08, 4.6566e-09, -7.7579e-07, ..., 2.6543e-08, + 0.0000e+00, -1.0431e-06], + [-9.8255e-08, 4.1910e-08, 2.0023e-08, ..., -1.5879e-07, + 0.0000e+00, 1.2759e-07]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0181, -0.0231, -0.0259, -0.0321, -0.0044, 0.0100, 0.0078, -0.0099, + -0.0082, -0.0141], device='cuda:0'), grad: tensor([ 2.9104e-07, 1.5227e-07, 1.8300e-07, 7.5437e-08, 7.4320e-07, + 2.2314e-06, -8.5309e-07, -4.8755e-07, -2.2687e-06, -6.8452e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 220.39, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4841 re_mapping 0.0037 re_causal 0.0103 /// teacc 99.06 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.2164, -0.2822, 0.1176, ..., -0.1408, 0.0488, 0.0410], + [-0.1513, -0.0633, -0.0997, ..., -0.2054, -0.0838, -0.0208], + [ 0.0196, -0.1576, -0.1973, ..., -0.1556, 0.0327, -0.3778], + ..., + [-0.2197, 0.1700, 0.0252, ..., 0.2057, -0.0560, -0.1628], + [-0.1839, -0.1749, 0.1916, ..., -0.1584, -0.1546, 0.1623], + [ 0.0176, -0.3163, 0.1727, ..., 0.0609, -0.2039, -0.1581]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.1910e-09], + ..., + [ 9.3132e-10, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 5.5879e-09], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [-1.8626e-09, 0.0000e+00, -3.2596e-09, ..., -1.3970e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0185, -0.0231, -0.0259, -0.0321, -0.0037, 0.0100, 0.0078, -0.0098, + -0.0076, -0.0149], device='cuda:0'), grad: tensor([ 1.6298e-08, -7.0874e-07, 1.5832e-07, 5.2620e-08, 3.5437e-07, + -2.2817e-08, 2.3283e-09, 1.2992e-07, 1.2573e-08, 1.4435e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 220.54, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4616 re_mapping 0.0037 re_causal 0.0101 /// teacc 99.07 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.2167, -0.2837, 0.1176, ..., -0.1409, 0.0484, 0.0409], + [-0.1512, -0.0634, -0.0998, ..., -0.2054, -0.0833, -0.0207], + [ 0.0196, -0.1576, -0.1981, ..., -0.1556, 0.0325, -0.3787], + ..., + [-0.2200, 0.1702, 0.0247, ..., 0.2056, -0.0562, -0.1634], + [-0.1840, -0.1750, 0.1919, ..., -0.1588, -0.1546, 0.1624], + [ 0.0178, -0.3169, 0.1731, ..., 0.0612, -0.2040, -0.1583]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, 1.3970e-09, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.2596e-09, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -9.3132e-10, ..., -3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0188, -0.0230, -0.0262, -0.0305, -0.0039, 0.0099, 0.0078, -0.0101, + -0.0076, -0.0150], device='cuda:0'), grad: tensor([ 6.9849e-09, 2.8871e-08, 6.5193e-09, -4.0978e-08, 4.7032e-08, + 6.0536e-09, -5.9605e-08, 6.0536e-09, 2.3283e-09, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 220.62, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4944 re_mapping 0.0035 re_causal 0.0102 /// teacc 99.00 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.2170, -0.2869, 0.1176, ..., -0.1411, 0.0484, 0.0411], + [-0.1512, -0.0655, -0.1013, ..., -0.2058, -0.0835, -0.0206], + [ 0.0196, -0.1579, -0.1993, ..., -0.1557, 0.0309, -0.3792], + ..., + [-0.2203, 0.1726, 0.0261, ..., 0.2059, -0.0564, -0.1636], + [-0.1840, -0.1752, 0.1924, ..., -0.1587, -0.1547, 0.1625], + [ 0.0178, -0.3178, 0.1733, ..., 0.0609, -0.2040, -0.1588]], + device='cuda:0'), grad: tensor([[ 4.1444e-08, 1.8626e-09, 4.6566e-08, ..., 9.7789e-09, + 0.0000e+00, 8.8941e-08], + [ 1.8626e-09, 1.8161e-08, 2.0023e-08, ..., 9.3132e-09, + 0.0000e+00, 1.6764e-08], + [-6.0536e-09, 5.9139e-08, 2.4727e-07, ..., 3.6787e-08, + -4.6566e-10, 2.7008e-07], + ..., + [ 2.3283e-09, 1.0710e-08, 1.5460e-06, ..., 2.7986e-07, + 0.0000e+00, 1.7043e-06], + [ 4.6566e-09, 4.6566e-10, -2.3581e-06, ..., -4.6706e-07, + 0.0000e+00, -2.5630e-06], + [ 1.1176e-08, 1.5832e-08, 3.9302e-07, ..., 1.0291e-07, + 0.0000e+00, 4.1490e-07]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0190, -0.0241, -0.0263, -0.0307, -0.0034, 0.0099, 0.0079, -0.0085, + -0.0075, -0.0155], device='cuda:0'), grad: tensor([ 2.1188e-07, 1.3830e-07, 4.0140e-07, 3.4273e-07, 3.1665e-08, + -5.1921e-07, 1.3690e-07, 3.5018e-06, -5.1931e-06, 9.4110e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 220.55, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4690 re_mapping 0.0035 re_causal 0.0093 /// teacc 99.08 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.2206, -0.2879, 0.1178, ..., -0.1413, 0.0465, 0.0394], + [-0.1520, -0.0657, -0.1016, ..., -0.2059, -0.0837, -0.0210], + [ 0.0182, -0.1587, -0.2003, ..., -0.1564, 0.0299, -0.3815], + ..., + [-0.2179, 0.1730, 0.0260, ..., 0.2083, -0.0564, -0.1675], + [-0.1851, -0.1755, 0.1938, ..., -0.1582, -0.1557, 0.1623], + [ 0.0166, -0.3188, 0.1738, ..., 0.0604, -0.2045, -0.1604]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 6.5658e-08, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.0710e-08, 0.0000e+00, ..., -1.3970e-08, + 0.0000e+00, 1.3970e-09], + ..., + [ 4.6566e-10, -1.1269e-07, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 4.6566e-10], + [ 3.2596e-09, 0.0000e+00, -1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 1.8626e-09], + [ 1.3504e-08, 2.1886e-08, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.0710e-08]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0196, -0.0251, -0.0246, -0.0315, -0.0050, 0.0101, 0.0079, -0.0073, + -0.0083, -0.0166], device='cuda:0'), grad: tensor([ 1.2200e-07, 8.0047e-07, -2.1718e-06, 1.5181e-07, 7.2643e-08, + 1.0151e-07, -9.3132e-10, 3.6135e-07, 4.0792e-07, 1.6531e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 220.36, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4795 re_mapping 0.0034 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.2209, -0.2887, 0.1178, ..., -0.1415, 0.0465, 0.0392], + [-0.1514, -0.0659, -0.1018, ..., -0.2062, -0.0837, -0.0208], + [ 0.0182, -0.1591, -0.2008, ..., -0.1568, 0.0299, -0.3815], + ..., + [-0.2181, 0.1733, 0.0258, ..., 0.2085, -0.0563, -0.1683], + [-0.1851, -0.1758, 0.1939, ..., -0.1587, -0.1558, 0.1625], + [ 0.0166, -0.3191, 0.1743, ..., 0.0608, -0.2045, -0.1602]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 4.6566e-10, -1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0198, -0.0249, -0.0245, -0.0314, -0.0050, 0.0100, 0.0079, -0.0076, + -0.0083, -0.0164], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.8626e-09, 9.3132e-10, 1.8626e-09, -2.7940e-09, + 1.3970e-09, -6.0536e-09, 4.6566e-10, 1.1642e-08, 1.3970e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 220.87, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4396 re_mapping 0.0034 re_causal 0.0098 /// teacc 99.05 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.2211, -0.2892, 0.1178, ..., -0.1416, 0.0464, 0.0391], + [-0.1516, -0.0665, -0.1023, ..., -0.2065, -0.0838, -0.0213], + [ 0.0182, -0.1593, -0.2011, ..., -0.1571, 0.0298, -0.3816], + ..., + [-0.2183, 0.1737, 0.0255, ..., 0.2085, -0.0560, -0.1684], + [-0.1853, -0.1762, 0.1945, ..., -0.1598, -0.1558, 0.1627], + [ 0.0170, -0.3194, 0.1748, ..., 0.0613, -0.2046, -0.1604]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3283e-09, ..., 9.3132e-10, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.1642e-08, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.0291e-08, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.2352e-07, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0200, -0.0252, -0.0244, -0.0310, -0.0051, 0.0100, 0.0079, -0.0075, + -0.0081, -0.0160], device='cuda:0'), grad: tensor([-8.3819e-09, -4.8429e-08, -5.1223e-09, 7.4506e-09, -5.9186e-07, + 3.7253e-09, 2.0489e-08, 1.6997e-07, -5.1223e-09, 4.7358e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 220.49, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.4588 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.16 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.2211, -0.2901, 0.1184, ..., -0.1418, 0.0464, 0.0415], + [-0.1517, -0.0668, -0.1026, ..., -0.2082, -0.0838, -0.0198], + [ 0.0181, -0.1601, -0.2015, ..., -0.1577, 0.0297, -0.3819], + ..., + [-0.2183, 0.1741, 0.0252, ..., 0.2093, -0.0562, -0.1716], + [-0.1854, -0.1796, 0.1945, ..., -0.1610, -0.1560, 0.1628], + [ 0.0170, -0.3199, 0.1753, ..., 0.0608, -0.2046, -0.1609]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.2201e-07, 4.6566e-10, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.2375e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -6.6264e-07, 1.3970e-09, ..., -9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 9.3132e-10, ..., 1.3970e-09, + 0.0000e+00, -9.3132e-10], + [-2.7940e-09, 1.0384e-07, -1.1642e-08, ..., -9.3132e-09, + 0.0000e+00, -9.3132e-10]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0182, -0.0252, -0.0244, -0.0306, -0.0041, 0.0099, 0.0079, -0.0083, + -0.0085, -0.0163], device='cuda:0'), grad: tensor([ 3.3062e-08, 1.7677e-06, 8.7544e-08, -1.4110e-07, 1.0291e-07, + 1.1595e-07, -9.2201e-08, -2.1812e-06, 6.9849e-09, 3.0501e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 220.90, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4638 re_mapping 0.0036 re_causal 0.0100 /// teacc 99.15 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.2213, -0.2903, 0.1183, ..., -0.1419, 0.0464, 0.0411], + [-0.1517, -0.0676, -0.1039, ..., -0.2091, -0.0838, -0.0214], + [ 0.0180, -0.1605, -0.2021, ..., -0.1580, 0.0299, -0.3822], + ..., + [-0.2183, 0.1748, 0.0252, ..., 0.2097, -0.0564, -0.1713], + [-0.1854, -0.1798, 0.1959, ..., -0.1607, -0.1560, 0.1632], + [ 0.0169, -0.3201, 0.1757, ..., 0.0611, -0.2046, -0.1615]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-08, 3.0268e-08, ..., 5.3551e-08, + -4.6566e-10, -2.7940e-09], + [ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + 4.6566e-10, 1.3970e-09], + ..., + [ 0.0000e+00, -1.2061e-07, -6.2864e-08, ..., -1.2154e-07, + 4.6566e-10, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.1910e-09], + [ 4.6566e-10, 4.6100e-08, 2.4214e-08, ..., 5.8673e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0184, -0.0259, -0.0244, -0.0304, -0.0041, 0.0098, 0.0080, -0.0081, + -0.0080, -0.0158], device='cuda:0'), grad: tensor([ 2.1886e-08, 1.7742e-07, -1.7323e-07, 3.5390e-08, 1.3411e-07, + -3.8184e-08, -2.4214e-08, -3.2550e-07, 4.8429e-08, 1.3690e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 220.13, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4751 re_mapping 0.0034 re_causal 0.0098 /// teacc 99.05 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.2215, -0.2905, 0.1183, ..., -0.1420, 0.0464, 0.0409], + [-0.1517, -0.0685, -0.1052, ..., -0.2100, -0.0838, -0.0214], + [ 0.0180, -0.1610, -0.2026, ..., -0.1586, 0.0300, -0.3825], + ..., + [-0.2184, 0.1757, 0.0262, ..., 0.2104, -0.0566, -0.1711], + [-0.1855, -0.1799, 0.1961, ..., -0.1609, -0.1560, 0.1631], + [ 0.0171, -0.3204, 0.1761, ..., 0.0615, -0.2047, -0.1619]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.3970e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 1.4435e-08, 1.0710e-08, ..., 2.4680e-08, + 0.0000e+00, 2.2817e-08], + [ 0.0000e+00, 4.1910e-09, 2.3283e-09, ..., 5.5879e-09, + 0.0000e+00, 4.1910e-09], + ..., + [ 9.3132e-10, -2.2817e-08, 6.9849e-09, ..., -3.3528e-08, + 0.0000e+00, 4.6566e-10], + [ 3.2596e-09, 0.0000e+00, -7.1712e-08, ..., 1.3970e-09, + 0.0000e+00, -1.8347e-07], + [ 1.8626e-09, 1.8626e-09, -2.7474e-08, ..., -2.0489e-08, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0187, -0.0261, -0.0244, -0.0304, -0.0045, 0.0098, 0.0080, -0.0080, + -0.0082, -0.0156], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.5507e-07, 2.7474e-08, 1.6205e-07, 4.2375e-08, + 2.0955e-08, 3.7625e-07, -9.5461e-08, -6.4168e-07, -4.9360e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 220.66, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4662 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.06 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.2217, -0.2906, 0.1182, ..., -0.1424, 0.0464, 0.0409], + [-0.1518, -0.0688, -0.1054, ..., -0.2104, -0.0832, -0.0215], + [ 0.0179, -0.1634, -0.2033, ..., -0.1618, 0.0297, -0.3832], + ..., + [-0.2184, 0.1764, 0.0262, ..., 0.2111, -0.0566, -0.1713], + [-0.1855, -0.1800, 0.1966, ..., -0.1613, -0.1561, 0.1633], + [ 0.0181, -0.3207, 0.1770, ..., 0.0620, -0.2047, -0.1622]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.3283e-09, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 1.3970e-09, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 5.1223e-09, 3.2596e-09, ..., 4.6566e-09, + 0.0000e+00, 9.3132e-10], + ..., + [ 4.6566e-10, -6.9849e-09, -1.8626e-09, ..., -6.5193e-09, + 0.0000e+00, 1.3970e-09], + [ 4.1910e-09, 4.6566e-10, -3.5856e-08, ..., 9.3132e-10, + 0.0000e+00, -7.0781e-08], + [ 5.5879e-09, 1.8626e-09, -1.3970e-08, ..., -1.0710e-08, + 0.0000e+00, 6.9849e-09]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0188, -0.0263, -0.0247, -0.0304, -0.0051, 0.0098, 0.0081, -0.0077, + -0.0079, -0.0150], device='cuda:0'), grad: tensor([ 9.7789e-09, 9.3132e-10, 5.5879e-09, -4.1910e-09, 3.9581e-08, + -7.4878e-07, 9.0478e-07, 8.8476e-09, -1.9372e-07, -1.1642e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 220.35, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.5017 re_mapping 0.0035 re_causal 0.0102 /// teacc 99.16 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.2218, -0.2908, 0.1183, ..., -0.1425, 0.0463, 0.0409], + [-0.1519, -0.0690, -0.1056, ..., -0.2107, -0.0833, -0.0215], + [ 0.0179, -0.1639, -0.2039, ..., -0.1622, 0.0306, -0.3838], + ..., + [-0.2184, 0.1767, 0.0262, ..., 0.2113, -0.0587, -0.1715], + [-0.1856, -0.1801, 0.1974, ..., -0.1614, -0.1568, 0.1635], + [ 0.0182, -0.3209, 0.1772, ..., 0.0622, -0.2047, -0.1626]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.3283e-09, 5.5879e-09, ..., 5.1223e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, 1.3970e-09, ..., 2.1886e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 3.7719e-08, 1.3970e-09, ..., 4.0513e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, -5.2620e-08, 1.5832e-08, ..., -5.2620e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-5.5879e-09, 9.3132e-10, -3.3528e-08, ..., -2.2817e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0189, -0.0265, -0.0245, -0.0303, -0.0052, 0.0098, 0.0081, -0.0078, + -0.0076, -0.0150], device='cuda:0'), grad: tensor([ 2.1886e-08, 3.0734e-08, 1.2247e-07, -2.6878e-06, 2.2817e-08, + 2.6505e-06, 9.3132e-10, -9.2667e-08, 1.2573e-08, -7.0315e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 220.82, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4663 re_mapping 0.0036 re_causal 0.0098 /// teacc 99.07 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.2219, -0.2911, 0.1184, ..., -0.1426, 0.0464, 0.0412], + [-0.1520, -0.0694, -0.1060, ..., -0.2111, -0.0833, -0.0214], + [ 0.0179, -0.1644, -0.2045, ..., -0.1626, 0.0306, -0.3841], + ..., + [-0.2183, 0.1771, 0.0264, ..., 0.2118, -0.0588, -0.1716], + [-0.1858, -0.1805, 0.1978, ..., -0.1617, -0.1571, 0.1635], + [ 0.0184, -0.3212, 0.1775, ..., 0.0623, -0.2048, -0.1630]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-09, 8.3819e-09, ..., 3.2596e-09, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 1.4296e-07, ..., 3.7253e-09, + 0.0000e+00, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.2596e-09, -7.4506e-09, 4.6566e-10, ..., -1.3970e-09, + 0.0000e+00, -1.3970e-09], + [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 1.3970e-09, + 0.0000e+00, 9.3132e-10], + [-2.4680e-08, 3.2596e-09, -3.4925e-07, ..., -5.4948e-08, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0187, -0.0264, -0.0249, -0.0303, -0.0053, 0.0098, 0.0081, -0.0076, + -0.0078, -0.0149], device='cuda:0'), grad: tensor([ 3.7253e-08, 5.7695e-07, -1.7695e-08, 1.3784e-07, 6.7009e-07, + -4.5169e-08, 9.3132e-09, 1.4901e-08, 1.9092e-08, -1.3970e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 220.43, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4788 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.14 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.2221, -0.2913, 0.1186, ..., -0.1427, 0.0465, 0.0416], + [-0.1520, -0.0697, -0.1062, ..., -0.2114, -0.0833, -0.0213], + [ 0.0179, -0.1648, -0.2039, ..., -0.1631, 0.0307, -0.3842], + ..., + [-0.2183, 0.1775, 0.0265, ..., 0.2122, -0.0588, -0.1717], + [-0.1858, -0.1806, 0.1978, ..., -0.1618, -0.1571, 0.1636], + [ 0.0184, -0.3216, 0.1777, ..., 0.0618, -0.2049, -0.1633]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 1.8626e-09, 4.6566e-10, ..., 9.3132e-10, + 2.3283e-09, 0.0000e+00], + [-3.7253e-09, -1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, -5.5879e-09, -2.3283e-09, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 3.7253e-09, 5.1223e-09, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0185, -0.0262, -0.0246, -0.0305, -0.0045, 0.0098, 0.0080, -0.0079, + -0.0078, -0.0155], device='cuda:0'), grad: tensor([ 1.9558e-08, 3.0268e-08, -1.9325e-07, 6.9849e-09, -1.2852e-07, + 3.1665e-08, 2.9802e-08, -4.6566e-09, 1.8626e-09, 2.1094e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 220.70, cls_loss 0.0009 cls_loss_mapping 0.0028 cls_loss_causal 0.4688 re_mapping 0.0033 re_causal 0.0093 /// teacc 99.06 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.2223, -0.2917, 0.1187, ..., -0.1428, 0.0466, 0.0417], + [-0.1522, -0.0734, -0.1098, ..., -0.2118, -0.0834, -0.0242], + [ 0.0178, -0.1659, -0.2053, ..., -0.1639, 0.0311, -0.3848], + ..., + [-0.2184, 0.1814, 0.0299, ..., 0.2126, -0.0590, -0.1689], + [-0.1872, -0.1808, 0.1986, ..., -0.1619, -0.1571, 0.1630], + [ 0.0182, -0.3230, 0.1777, ..., 0.0613, -0.2052, -0.1642]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.9337e-08, ..., -3.2596e-09, + 0.0000e+00, -1.0245e-08], + [ 1.8626e-09, 4.6566e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -6.9849e-09], + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 4.6566e-09, 0.0000e+00, 3.7253e-09, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-09], + [ 3.7253e-09, 0.0000e+00, 1.8626e-08, ..., 1.3970e-09, + 0.0000e+00, 9.7789e-09]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0185, -0.0291, -0.0251, -0.0298, -0.0039, 0.0097, 0.0082, -0.0047, + -0.0085, -0.0164], device='cuda:0'), grad: tensor([-6.5658e-08, -1.5972e-07, 1.2573e-08, 1.1176e-08, -4.2375e-08, + -1.8859e-07, 1.6717e-07, 1.8394e-07, 1.9092e-08, 6.4727e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 220.64, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4742 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.13 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.2228, -0.2923, 0.1188, ..., -0.1429, 0.0466, 0.0416], + [-0.1526, -0.0736, -0.1099, ..., -0.2119, -0.0836, -0.0244], + [ 0.0175, -0.1664, -0.2075, ..., -0.1645, 0.0311, -0.3856], + ..., + [-0.2186, 0.1816, 0.0300, ..., 0.2130, -0.0592, -0.1688], + [-0.1881, -0.1810, 0.1993, ..., -0.1621, -0.1575, 0.1625], + [ 0.0178, -0.3244, 0.1778, ..., 0.0611, -0.2054, -0.1648]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 4.6566e-10, + 0.0000e+00, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + -2.7940e-09, -8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -9.3132e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 1.3504e-08, + 1.3970e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, 9.3132e-10], + [-9.3132e-10, 0.0000e+00, -3.3062e-08, ..., -2.9337e-08, + 0.0000e+00, 6.9849e-09]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0185, -0.0292, -0.0255, -0.0297, -0.0039, 0.0098, 0.0081, -0.0046, + -0.0089, -0.0168], device='cuda:0'), grad: tensor([-3.3993e-08, -4.5169e-08, -4.6566e-10, 6.0536e-09, 5.1688e-08, + 1.1176e-08, -7.9162e-09, 8.3819e-08, 1.3504e-08, -6.7521e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 220.74, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4557 re_mapping 0.0035 re_causal 0.0094 /// teacc 99.13 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.2234, -0.2930, 0.1191, ..., -0.1430, 0.0461, 0.0416], + [-0.1530, -0.0736, -0.1099, ..., -0.2120, -0.0809, -0.0244], + [ 0.0172, -0.1671, -0.2095, ..., -0.1648, 0.0303, -0.3862], + ..., + [-0.2188, 0.1817, 0.0300, ..., 0.2131, -0.0621, -0.1689], + [-0.1883, -0.1818, 0.1996, ..., -0.1621, -0.1581, 0.1625], + [ 0.0174, -0.3252, 0.1785, ..., 0.0615, -0.2055, -0.1660]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.4960e-08, ..., 0.0000e+00, + 0.0000e+00, -1.7462e-08], + [ 0.0000e+00, -1.0920e-07, 8.1491e-09, ..., 0.0000e+00, + 0.0000e+00, 7.2177e-09], + [ 0.0000e+00, 0.0000e+00, 1.9814e-07, ..., 0.0000e+00, + 0.0000e+00, 1.6089e-07], + ..., + [ 0.0000e+00, 1.0384e-07, 8.8476e-09, ..., 2.3283e-10, + 0.0000e+00, 7.6834e-09], + [ 0.0000e+00, 2.3283e-10, -1.6205e-07, ..., 0.0000e+00, + 0.0000e+00, -1.7066e-07], + [ 0.0000e+00, 4.1910e-09, 2.5611e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0177, -0.0291, -0.0260, -0.0297, -0.0042, 0.0099, 0.0081, -0.0047, + -0.0091, -0.0167], device='cuda:0'), grad: tensor([-2.0303e-07, -1.0105e-06, 7.0781e-07, 1.8626e-08, -1.4855e-07, + 1.3039e-08, 2.3982e-08, 1.0738e-06, -5.8766e-07, 1.2759e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 220.36, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4661 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.12 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.2245, -0.2933, 0.1182, ..., -0.1431, 0.0462, 0.0406], + [-0.1538, -0.0736, -0.1100, ..., -0.2122, -0.0809, -0.0245], + [ 0.0170, -0.1692, -0.2120, ..., -0.1662, 0.0303, -0.3872], + ..., + [-0.2188, 0.1817, 0.0300, ..., 0.2136, -0.0621, -0.1689], + [-0.1884, -0.1822, 0.2000, ..., -0.1620, -0.1581, 0.1627], + [ 0.0182, -0.3265, 0.1807, ..., 0.0614, -0.2056, -0.1635]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 1.1642e-09, 6.9849e-10, ..., 1.1642e-09, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 2.3283e-09, 1.3970e-09, ..., 2.3283e-09, + 0.0000e+00, 2.3283e-10], + ..., + [ 4.6566e-10, -9.0804e-09, -3.7253e-09, ..., -6.9849e-09, + 0.0000e+00, 2.3283e-10], + [ 8.1491e-09, 0.0000e+00, -2.3283e-09, ..., 2.3283e-10, + 0.0000e+00, 5.3551e-09], + [ 3.7253e-09, 2.5611e-09, -6.9849e-10, ..., 2.3283e-10, + 0.0000e+00, 5.1223e-09]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0191, -0.0291, -0.0269, -0.0313, -0.0044, 0.0107, 0.0081, -0.0047, + -0.0090, -0.0147], device='cuda:0'), grad: tensor([ 3.9581e-09, 4.6566e-09, 7.9162e-09, 3.3993e-08, 1.0012e-08, + -6.2631e-08, 1.1642e-09, -1.8161e-08, 1.3504e-08, 1.2806e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 220.61, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4883 re_mapping 0.0033 re_causal 0.0092 /// teacc 99.10 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.2249, -0.2936, 0.1186, ..., -0.1431, 0.0463, 0.0408], + [-0.1533, -0.0737, -0.1101, ..., -0.2125, -0.0810, -0.0245], + [ 0.0169, -0.1701, -0.2136, ..., -0.1668, 0.0303, -0.3877], + ..., + [-0.2189, 0.1818, 0.0301, ..., 0.2142, -0.0620, -0.1689], + [-0.1886, -0.1831, 0.2006, ..., -0.1626, -0.1582, 0.1628], + [ 0.0185, -0.3281, 0.1811, ..., 0.0610, -0.2056, -0.1638]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 7.2177e-09, 2.0955e-09, ..., 4.4238e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 1.1642e-09, + 2.3283e-10, 2.0955e-09], + ..., + [ 0.0000e+00, -9.5461e-09, -2.3283e-09, ..., -5.8208e-09, + -6.9849e-10, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -1.8161e-08], + [ 0.0000e+00, 6.9849e-10, 9.3132e-10, ..., -1.3970e-09, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0188, -0.0289, -0.0277, -0.0332, -0.0050, 0.0119, 0.0081, -0.0047, + -0.0090, -0.0149], device='cuda:0'), grad: tensor([ 1.6298e-09, 1.1176e-08, 9.5461e-09, 5.1223e-09, 1.6065e-08, + -2.0955e-08, 4.4703e-08, -1.4668e-08, -3.7020e-08, 7.6834e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 220.23, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4644 re_mapping 0.0034 re_causal 0.0097 /// teacc 99.17 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.2254, -0.2938, 0.1186, ..., -0.1432, 0.0463, 0.0407], + [-0.1512, -0.0737, -0.1100, ..., -0.2125, -0.0810, -0.0245], + [ 0.0168, -0.1706, -0.2142, ..., -0.1673, 0.0304, -0.3881], + ..., + [-0.2192, 0.1819, 0.0301, ..., 0.2144, -0.0621, -0.1689], + [-0.1887, -0.1833, 0.2009, ..., -0.1634, -0.1582, 0.1629], + [ 0.0211, -0.3284, 0.1824, ..., 0.0616, -0.2056, -0.1639]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.5611e-09, 1.8626e-09, ..., 1.6298e-09, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 1.1176e-08, 8.1491e-09, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.7940e-09, 1.1642e-09, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, -7.6834e-08, -6.7055e-08, ..., -7.9861e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 1.3970e-08, 3.7253e-09, ..., 2.5611e-09, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 6.3796e-08, -1.7439e-07, ..., -6.4960e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0188, -0.0288, -0.0279, -0.0332, -0.0073, 0.0120, 0.0081, -0.0047, + -0.0090, -0.0133], device='cuda:0'), grad: tensor([ 1.1642e-08, 3.7719e-08, 8.6147e-09, -8.7544e-07, 7.0315e-07, + 8.5495e-07, -6.0536e-09, -2.6100e-07, 4.5402e-08, -5.0105e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 220.59, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4961 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.17 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.2260, -0.2941, 0.1186, ..., -0.1433, 0.0463, 0.0405], + [-0.1513, -0.0736, -0.1100, ..., -0.2126, -0.0810, -0.0245], + [ 0.0169, -0.1707, -0.2132, ..., -0.1674, 0.0305, -0.3881], + ..., + [-0.2193, 0.1819, 0.0301, ..., 0.2147, -0.0621, -0.1689], + [-0.1890, -0.1836, 0.2007, ..., -0.1644, -0.1582, 0.1630], + [ 0.0213, -0.3297, 0.1828, ..., 0.0618, -0.2056, -0.1638]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 9.7789e-09, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 1.0245e-08, 5.5879e-09, ..., 2.3283e-08, + 0.0000e+00, 4.6566e-10], + [-4.1910e-09, 2.3283e-09, 9.3132e-10, ..., -3.5856e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.3283e-08, 2.2352e-08, ..., 3.4459e-08, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -6.9849e-09, ..., 9.3132e-10, + 0.0000e+00, -1.1642e-08], + [ 0.0000e+00, 6.5193e-09, -5.6345e-08, ..., -7.2177e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0190, -0.0286, -0.0274, -0.0332, -0.0073, 0.0120, 0.0081, -0.0049, + -0.0093, -0.0135], device='cuda:0'), grad: tensor([ 2.3376e-07, 4.8196e-07, -1.1176e-06, 3.0268e-08, 9.8255e-08, + 3.5856e-08, 4.9826e-08, 3.5809e-07, 3.2596e-09, -1.7369e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 220.35, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4623 re_mapping 0.0033 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.2266, -0.2951, 0.1186, ..., -0.1442, 0.0463, 0.0407], + [-0.1476, -0.0736, -0.1100, ..., -0.2127, -0.0811, -0.0246], + [ 0.0171, -0.1712, -0.2141, ..., -0.1678, 0.0306, -0.3886], + ..., + [-0.2200, 0.1819, 0.0300, ..., 0.2150, -0.0621, -0.1690], + [-0.1890, -0.1841, 0.2020, ..., -0.1648, -0.1583, 0.1639], + [ 0.0179, -0.3307, 0.1826, ..., 0.0618, -0.2057, -0.1667]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 0.0000e+00, 4.6566e-10, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, -6.0070e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 8.8476e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.1176e-08, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0190, -0.0283, -0.0275, -0.0332, -0.0076, 0.0120, 0.0080, -0.0051, + -0.0077, -0.0155], device='cuda:0'), grad: tensor([-1.1642e-08, -4.9174e-07, 8.6147e-08, 2.7940e-09, 7.4506e-08, + 1.2573e-08, 2.7474e-07, 1.5367e-08, 1.3970e-08, 2.7008e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 220.65, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4841 re_mapping 0.0032 re_causal 0.0092 /// teacc 99.18 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.2274, -0.2958, 0.1185, ..., -0.1449, 0.0465, 0.0406], + [-0.1476, -0.0736, -0.1100, ..., -0.2131, -0.0814, -0.0245], + [ 0.0171, -0.1727, -0.2152, ..., -0.1690, 0.0358, -0.3898], + ..., + [-0.2203, 0.1819, 0.0300, ..., 0.2154, -0.0663, -0.1691], + [-0.1899, -0.1849, 0.2022, ..., -0.1667, -0.1583, 0.1640], + [ 0.0184, -0.3314, 0.1837, ..., 0.0595, -0.2058, -0.1668]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.8859e-07, ..., 0.0000e+00, + 4.6566e-10, -2.4820e-07], + [ 4.6566e-10, 4.6566e-10, 2.0955e-08, ..., 4.6566e-10, + 2.3283e-09, 3.2131e-08], + [ 4.6566e-10, 4.6566e-10, 8.8476e-09, ..., -4.1910e-09, + -1.4435e-08, 1.2107e-08], + ..., + [ 4.6566e-10, 9.3132e-10, 5.2620e-08, ..., 2.7940e-09, + 1.0245e-08, 7.0315e-08], + [-2.3283e-09, 2.3283e-09, -3.0268e-08, ..., -2.3283e-09, + 0.0000e+00, -1.7602e-07], + [ 1.3970e-09, 4.6566e-10, 3.7719e-08, ..., 0.0000e+00, + 0.0000e+00, 5.1688e-08]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0194, -0.0281, -0.0269, -0.0330, -0.0050, 0.0120, 0.0078, -0.0053, + -0.0083, -0.0172], device='cuda:0'), grad: tensor([-1.8142e-06, 1.7229e-07, 2.7474e-08, 2.0536e-07, 6.1002e-08, + 5.4017e-07, 1.9511e-07, 5.6019e-07, -3.2876e-07, 3.6974e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 220.78, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4363 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.15 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.2287, -0.2960, 0.1192, ..., -0.1450, 0.0463, 0.0415], + [-0.1478, -0.0737, -0.1100, ..., -0.2134, -0.0820, -0.0245], + [ 0.0175, -0.1727, -0.2164, ..., -0.1694, 0.0395, -0.3906], + ..., + [-0.2204, 0.1819, 0.0300, ..., 0.2161, -0.0694, -0.1691], + [-0.1906, -0.1854, 0.2028, ..., -0.1675, -0.1586, 0.1640], + [ 0.0184, -0.3337, 0.1837, ..., 0.0592, -0.2063, -0.1669]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 6.0536e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -1.3970e-09], + [ 3.8650e-08, 1.0151e-07, 3.0268e-08, ..., 7.3574e-08, + 0.0000e+00, 4.6566e-08], + [ 4.6566e-10, 1.8626e-08, 4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, 4.1910e-09], + ..., + [ 1.3970e-09, -1.8440e-07, -6.0070e-08, ..., -1.7183e-07, + 0.0000e+00, 1.3970e-09], + [ 2.1653e-07, 1.4901e-08, 3.7253e-09, ..., 8.3819e-09, + 0.0000e+00, 2.3842e-07], + [ 5.7276e-08, 4.0513e-08, 6.9849e-09, ..., 2.3283e-09, + 0.0000e+00, 3.6322e-08]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0175, -0.0281, -0.0259, -0.0330, -0.0051, 0.0121, 0.0077, -0.0054, + -0.0086, -0.0175], device='cuda:0'), grad: tensor([ 1.2107e-08, 4.1071e-07, 1.0431e-07, -4.0233e-07, 4.0699e-07, + -1.0263e-06, 4.1816e-07, -6.4727e-07, 5.0385e-07, 2.3982e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 220.55, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4543 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.12 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.2289, -0.2963, 0.1192, ..., -0.1451, 0.0461, 0.0413], + [-0.1479, -0.0737, -0.1100, ..., -0.2138, -0.0821, -0.0245], + [ 0.0176, -0.1734, -0.2169, ..., -0.1704, 0.0399, -0.3913], + ..., + [-0.2205, 0.1820, 0.0300, ..., 0.2170, -0.0696, -0.1692], + [-0.1920, -0.1875, 0.2028, ..., -0.1682, -0.1587, 0.1634], + [ 0.0184, -0.3363, 0.1838, ..., 0.0590, -0.2064, -0.1670]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.1002e-08, ..., 0.0000e+00, + 0.0000e+00, -6.1467e-08], + [ 0.0000e+00, 1.8626e-09, 4.6566e-09, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09], + ..., + [ 0.0000e+00, -1.8626e-09, 4.6566e-10, ..., -4.6566e-10, + 0.0000e+00, 2.3283e-09], + [ 3.7253e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0178, -0.0281, -0.0263, -0.0330, -0.0051, 0.0122, 0.0077, -0.0054, + -0.0095, -0.0178], device='cuda:0'), grad: tensor([-5.6624e-07, 2.1420e-08, 2.1420e-08, 3.1199e-08, 6.0536e-09, + 3.7719e-08, 4.1351e-07, 1.3504e-08, 1.2573e-08, 2.2352e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 220.86, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4539 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.15 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.2293, -0.2964, 0.1192, ..., -0.1454, 0.0456, 0.0412], + [-0.1480, -0.0737, -0.1101, ..., -0.2139, -0.0821, -0.0245], + [ 0.0176, -0.1735, -0.2184, ..., -0.1704, 0.0400, -0.3921], + ..., + [-0.2206, 0.1820, 0.0300, ..., 0.2170, -0.0696, -0.1692], + [-0.1935, -0.1877, 0.2040, ..., -0.1689, -0.1587, 0.1629], + [ 0.0185, -0.3367, 0.1844, ..., 0.0592, -0.2064, -0.1671]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-2.3283e-09, 5.5879e-09, 0.0000e+00, ..., 5.1223e-09, + 0.0000e+00, -4.6566e-09], + [ 0.0000e+00, 1.0245e-08, 1.3970e-09, ..., 9.7789e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 1.3970e-09, -1.8161e-08, -1.3970e-09, ..., -1.6764e-08, + 0.0000e+00, 3.7253e-09], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 2.7940e-09, -2.7940e-09, ..., -9.3132e-10, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0177, -0.0281, -0.0262, -0.0330, -0.0052, 0.0123, 0.0075, -0.0054, + -0.0101, -0.0175], device='cuda:0'), grad: tensor([ 1.3970e-09, -9.0059e-07, 6.9849e-07, 2.4214e-08, 2.1886e-08, + -3.3993e-08, 1.0245e-08, 1.8068e-07, 4.1910e-09, 4.6566e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 220.27, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4823 re_mapping 0.0033 re_causal 0.0096 /// teacc 99.15 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.2302, -0.2976, 0.1189, ..., -0.1464, 0.0449, 0.0410], + [-0.1481, -0.0740, -0.1103, ..., -0.2157, -0.0824, -0.0245], + [ 0.0177, -0.1742, -0.2185, ..., -0.1712, 0.0403, -0.3924], + ..., + [-0.2207, 0.1824, 0.0302, ..., 0.2185, -0.0697, -0.1692], + [-0.1935, -0.1879, 0.2058, ..., -0.1697, -0.1589, 0.1636], + [ 0.0185, -0.3372, 0.1850, ..., 0.0596, -0.2066, -0.1673]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, -1.3504e-08], + [ 4.6566e-10, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 1.2573e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 2.8871e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0184, -0.0283, -0.0259, -0.0330, -0.0053, 0.0123, 0.0075, -0.0052, + -0.0091, -0.0174], device='cuda:0'), grad: tensor([-5.4017e-08, 7.6834e-08, -1.2061e-07, -9.2201e-08, 3.2596e-09, + 1.9558e-08, 2.9802e-08, 1.0710e-07, 1.9092e-08, 1.7229e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 220.16, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4913 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.10 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.2306, -0.2987, 0.1187, ..., -0.1470, 0.0449, 0.0407], + [-0.1481, -0.0741, -0.1103, ..., -0.2160, -0.0826, -0.0245], + [ 0.0180, -0.1762, -0.2199, ..., -0.1729, 0.0404, -0.3929], + ..., + [-0.2208, 0.1825, 0.0302, ..., 0.2189, -0.0698, -0.1692], + [-0.1936, -0.1879, 0.2063, ..., -0.1708, -0.1590, 0.1639], + [ 0.0186, -0.3377, 0.1868, ..., 0.0609, -0.2072, -0.1674]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + -4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0189, -0.0283, -0.0265, -0.0330, -0.0061, 0.0123, 0.0076, -0.0052, + -0.0089, -0.0165], device='cuda:0'), grad: tensor([ 4.1910e-09, 3.9581e-08, -9.2201e-08, 5.4482e-08, -4.3772e-08, + -7.9162e-09, -2.9802e-08, 1.7229e-08, -1.3970e-09, 6.4261e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 220.89, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4467 re_mapping 0.0034 re_causal 0.0095 /// teacc 99.08 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.2325, -0.2988, 0.1184, ..., -0.1479, 0.0449, 0.0401], + [-0.1481, -0.0741, -0.1103, ..., -0.2161, -0.0826, -0.0244], + [ 0.0179, -0.1759, -0.2201, ..., -0.1726, 0.0404, -0.3931], + ..., + [-0.2208, 0.1825, 0.0302, ..., 0.2189, -0.0698, -0.1693], + [-0.1939, -0.1883, 0.2052, ..., -0.1737, -0.1592, 0.1637], + [ 0.0183, -0.3382, 0.1878, ..., 0.0614, -0.2073, -0.1673]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + -1.3970e-09, -4.6566e-10], + [ 0.0000e+00, 1.7229e-08, 5.5879e-09, ..., 1.3970e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, -1.9092e-08, -4.1910e-09, ..., -1.4901e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 4.6566e-10, -4.6566e-09], + [ 0.0000e+00, 9.3132e-10, -8.3819e-09, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0194, -0.0283, -0.0258, -0.0331, -0.0063, 0.0124, 0.0075, -0.0052, + -0.0095, -0.0162], device='cuda:0'), grad: tensor([-2.5611e-08, 4.7032e-08, 2.0023e-08, 1.4435e-08, 1.3504e-08, + 6.0536e-09, 3.7253e-09, -4.4703e-08, -5.5879e-09, -1.8626e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 220.46, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4762 re_mapping 0.0033 re_causal 0.0093 /// teacc 99.07 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.2327, -0.2991, 0.1183, ..., -0.1485, 0.0445, 0.0401], + [-0.1482, -0.0741, -0.1103, ..., -0.2163, -0.0827, -0.0245], + [ 0.0176, -0.1783, -0.2220, ..., -0.1745, 0.0406, -0.3937], + ..., + [-0.2210, 0.1827, 0.0302, ..., 0.2197, -0.0699, -0.1693], + [-0.1940, -0.1886, 0.2080, ..., -0.1741, -0.1596, 0.1642], + [ 0.0183, -0.3387, 0.1884, ..., 0.0614, -0.2073, -0.1675]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 1.3970e-09, 9.3132e-10, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.1910e-09, -2.7940e-09, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 3.2596e-09], + [ 4.6566e-10, 4.1910e-09, 1.3970e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0198, -0.0283, -0.0264, -0.0331, -0.0069, 0.0124, 0.0076, -0.0051, + -0.0086, -0.0162], device='cuda:0'), grad: tensor([ 1.3970e-09, 3.8650e-08, -1.8161e-08, 6.0536e-09, 5.0291e-08, + -4.4703e-08, -2.7474e-08, -8.3819e-09, 7.4506e-09, 7.4506e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 220.32, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4814 re_mapping 0.0032 re_causal 0.0092 /// teacc 99.05 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.2331, -0.2996, 0.1188, ..., -0.1487, 0.0445, 0.0407], + [-0.1483, -0.0744, -0.1103, ..., -0.2190, -0.0827, -0.0246], + [ 0.0175, -0.1783, -0.2224, ..., -0.1748, 0.0407, -0.3940], + ..., + [-0.2211, 0.1831, 0.0302, ..., 0.2224, -0.0699, -0.1693], + [-0.1942, -0.1888, 0.2084, ..., -0.1744, -0.1597, 0.1644], + [ 0.0180, -0.3405, 0.1882, ..., 0.0611, -0.2073, -0.1677]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.9558e-08, 0.0000e+00, ..., 5.1223e-09, + 0.0000e+00, 9.3132e-10], + [-9.3132e-10, 1.0338e-07, 0.0000e+00, ..., 2.8871e-08, + 0.0000e+00, -3.2596e-09], + [ 0.0000e+00, 5.1223e-09, 1.8626e-09, ..., 1.3970e-09, + 0.0000e+00, 1.8626e-09], + ..., + [ 4.6566e-10, -2.2771e-07, 0.0000e+00, ..., -6.3330e-08, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.3970e-09, -3.7253e-09, ..., 4.6566e-10, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 9.7789e-09, 4.6566e-10, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0189, -0.0288, -0.0262, -0.0332, -0.0068, 0.0125, 0.0075, -0.0047, + -0.0086, -0.0166], device='cuda:0'), grad: tensor([ 5.0757e-08, 2.0117e-07, 1.9092e-08, 3.8184e-08, 5.5879e-08, + 1.5320e-07, -5.0757e-08, -4.9407e-07, -9.3132e-10, 2.4214e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 220.32, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4914 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.13 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.2333, -0.2998, 0.1189, ..., -0.1490, 0.0445, 0.0409], + [-0.1484, -0.0750, -0.1110, ..., -0.2192, -0.0828, -0.0243], + [ 0.0175, -0.1785, -0.2226, ..., -0.1750, 0.0407, -0.3942], + ..., + [-0.2211, 0.1837, 0.0309, ..., 0.2225, -0.0700, -0.1696], + [-0.1943, -0.1890, 0.2089, ..., -0.1749, -0.1597, 0.1645], + [ 0.0180, -0.3411, 0.1886, ..., 0.0614, -0.2074, -0.1679]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 0.0000e+00, 4.6566e-10, -2.0070e-07, ..., 9.3132e-10, + 0.0000e+00, -2.9476e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 1.2573e-08], + ..., + [ 0.0000e+00, 2.7940e-09, 3.8184e-08, ..., 1.8626e-09, + 0.0000e+00, 5.3551e-08], + [ 4.6566e-10, 4.6566e-10, 1.3644e-07, ..., 0.0000e+00, + 0.0000e+00, 2.0023e-07], + [-9.3132e-10, 4.6566e-10, 3.7253e-09, ..., -8.3819e-09, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0191, -0.0294, -0.0257, -0.0332, -0.0068, 0.0125, 0.0075, -0.0042, + -0.0086, -0.0165], device='cuda:0'), grad: tensor([ 5.6345e-08, -3.3844e-06, 1.3970e-07, -2.7940e-09, 5.7276e-08, + 4.4703e-08, 2.8405e-08, 6.2771e-07, 2.3041e-06, 1.2247e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 220.64, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4603 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.07 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.2334, -0.3015, 0.1190, ..., -0.1495, 0.0445, 0.0409], + [-0.1487, -0.0750, -0.1111, ..., -0.2198, -0.0828, -0.0243], + [ 0.0174, -0.1787, -0.2232, ..., -0.1755, 0.0407, -0.3946], + ..., + [-0.2212, 0.1838, 0.0309, ..., 0.2231, -0.0700, -0.1696], + [-0.1943, -0.1893, 0.2095, ..., -0.1764, -0.1598, 0.1648], + [ 0.0186, -0.3424, 0.1892, ..., 0.0618, -0.2074, -0.1677]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 2.3283e-09, 1.0245e-08, ..., 1.3970e-09, + 0.0000e+00, 1.8626e-08], + ..., + [ 0.0000e+00, -5.1223e-09, -4.6566e-10, ..., -6.0536e-09, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, -1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, -2.5611e-08], + [ 0.0000e+00, 9.3132e-10, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0193, -0.0295, -0.0259, -0.0332, -0.0063, 0.0125, 0.0075, -0.0041, + -0.0088, -0.0163], device='cuda:0'), grad: tensor([ 3.2596e-08, -6.6590e-08, 1.2666e-07, 9.3132e-10, 3.0734e-08, + 3.5856e-08, -1.1828e-07, 7.4506e-09, -6.1467e-08, 1.8161e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 220.45, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4741 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.14 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.2340, -0.3024, 0.1190, ..., -0.1497, 0.0445, 0.0403], + [-0.1488, -0.0751, -0.1111, ..., -0.2199, -0.0828, -0.0242], + [ 0.0158, -0.1793, -0.2242, ..., -0.1762, 0.0407, -0.3960], + ..., + [-0.2213, 0.1839, 0.0310, ..., 0.2235, -0.0700, -0.1697], + [-0.1945, -0.1896, 0.2098, ..., -0.1771, -0.1598, 0.1648], + [ 0.0182, -0.3439, 0.1896, ..., 0.0620, -0.2074, -0.1682]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 3.7253e-09, 4.6566e-10, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -5.5879e-09, 2.7940e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, -6.5193e-09, ..., -3.2596e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0199, -0.0294, -0.0264, -0.0332, -0.0065, 0.0125, 0.0076, -0.0041, + -0.0089, -0.0164], device='cuda:0'), grad: tensor([ 6.0536e-09, 9.3132e-09, 4.1910e-09, 1.8626e-09, 3.7253e-09, + 8.3819e-09, -1.4901e-08, -4.6566e-09, 2.7940e-09, -1.3039e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 220.35, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4657 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.12 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.2341, -0.3032, 0.1190, ..., -0.1501, 0.0445, 0.0401], + [-0.1487, -0.0751, -0.1111, ..., -0.2201, -0.0829, -0.0242], + [ 0.0153, -0.1794, -0.2251, ..., -0.1762, 0.0407, -0.3968], + ..., + [-0.2213, 0.1840, 0.0309, ..., 0.2235, -0.0700, -0.1697], + [-0.1946, -0.1916, 0.2104, ..., -0.1773, -0.1598, 0.1651], + [ 0.0182, -0.3448, 0.1899, ..., 0.0622, -0.2075, -0.1683]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.1420e-08, ..., 0.0000e+00, + 0.0000e+00, -8.3819e-09], + [ 0.0000e+00, 4.6566e-10, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + ..., + [ 0.0000e+00, 4.6566e-10, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09], + [-4.6566e-10, 0.0000e+00, 2.7940e-09, ..., -2.3283e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0203, -0.0294, -0.0266, -0.0332, -0.0064, 0.0125, 0.0076, -0.0042, + -0.0087, -0.0164], device='cuda:0'), grad: tensor([ 4.6566e-09, 2.9337e-08, -2.6263e-07, 1.5181e-07, 9.7789e-09, + 3.2596e-09, 8.3819e-09, 2.4680e-08, 1.4901e-08, 1.9558e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 220.33, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4826 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.19 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.2342, -0.3021, 0.1195, ..., -0.1503, 0.0444, 0.0402], + [-0.1488, -0.0751, -0.1111, ..., -0.2205, -0.0828, -0.0241], + [ 0.0149, -0.1803, -0.2271, ..., -0.1765, 0.0407, -0.3978], + ..., + [-0.2214, 0.1840, 0.0309, ..., 0.2240, -0.0700, -0.1698], + [-0.1946, -0.1920, 0.2113, ..., -0.1770, -0.1598, 0.1655], + [ 0.0183, -0.3475, 0.1899, ..., 0.0622, -0.2076, -0.1685]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -1.1642e-08, ..., 0.0000e+00, + 0.0000e+00, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0199, -0.0292, -0.0270, -0.0331, -0.0064, 0.0125, 0.0076, -0.0043, + -0.0082, -0.0167], device='cuda:0'), grad: tensor([-3.7253e-09, 3.2596e-09, 1.8626e-09, 2.5146e-08, 1.3970e-09, + -1.2107e-08, 7.9162e-09, 1.8626e-09, -2.6077e-08, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 221.02, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4641 re_mapping 0.0033 re_causal 0.0096 /// teacc 99.12 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.2344, -0.3023, 0.1197, ..., -0.1504, 0.0444, 0.0403], + [-0.1488, -0.0751, -0.1111, ..., -0.2207, -0.0825, -0.0241], + [ 0.0145, -0.1811, -0.2280, ..., -0.1775, 0.0404, -0.3992], + ..., + [-0.2214, 0.1842, 0.0310, ..., 0.2249, -0.0701, -0.1698], + [-0.1947, -0.1923, 0.2119, ..., -0.1763, -0.1602, 0.1656], + [ 0.0184, -0.3496, 0.1900, ..., 0.0625, -0.2076, -0.1687]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.8476e-09, 1.8626e-09, ..., 8.8476e-09, + 0.0000e+00, -1.3970e-09], + [ 0.0000e+00, 9.3132e-10, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, -2.2817e-08, -5.1223e-09, ..., -2.3283e-08, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 1.0710e-08, 2.7940e-09, ..., 1.0710e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0198, -0.0293, -0.0277, -0.0331, -0.0071, 0.0124, 0.0078, -0.0042, + -0.0084, -0.0166], device='cuda:0'), grad: tensor([ 2.7940e-09, -9.1735e-08, -2.3050e-07, 3.2596e-09, 1.3318e-07, + 6.0536e-09, 4.6566e-09, -5.2620e-08, 1.8999e-07, 4.3306e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 221.01, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4713 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.20 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.2347, -0.3029, 0.1197, ..., -0.1511, 0.0441, 0.0403], + [-0.1489, -0.0751, -0.1111, ..., -0.2209, -0.0826, -0.0241], + [ 0.0115, -0.1846, -0.2290, ..., -0.1782, 0.0403, -0.3999], + ..., + [-0.2215, 0.1842, 0.0310, ..., 0.2253, -0.0701, -0.1698], + [-0.1948, -0.1929, 0.2127, ..., -0.1767, -0.1603, 0.1663], + [ 0.0183, -0.3512, 0.1904, ..., 0.0619, -0.2076, -0.1689]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -5.5879e-09, 4.6566e-10, ..., -2.7940e-09, + 0.0000e+00, 4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 4.6566e-10, -9.3132e-10, ..., -1.3970e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0200, -0.0293, -0.0304, -0.0319, -0.0067, 0.0122, 0.0078, -0.0043, + -0.0079, -0.0172], device='cuda:0'), grad: tensor([ 3.2596e-09, 2.6776e-07, -2.6822e-07, -9.3132e-10, 3.7253e-09, + -2.3283e-09, 5.1223e-09, -6.0536e-09, 5.1223e-09, -4.6566e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 220.47, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4786 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.12 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.2349, -0.3036, 0.1200, ..., -0.1514, 0.0438, 0.0400], + [-0.1490, -0.0752, -0.1111, ..., -0.2211, -0.0790, -0.0239], + [ 0.0115, -0.1855, -0.2303, ..., -0.1798, 0.0379, -0.4018], + ..., + [-0.2215, 0.1844, 0.0310, ..., 0.2257, -0.0711, -0.1699], + [-0.1949, -0.1945, 0.2124, ..., -0.1774, -0.1615, 0.1667], + [ 0.0182, -0.3522, 0.1906, ..., 0.0618, -0.2077, -0.1690]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 8.8476e-09, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -4.1910e-09], + ..., + [ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., -3.2596e-09, + 0.0000e+00, 1.3970e-09], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09], + [ 4.6566e-10, 4.6566e-10, -1.3970e-09, ..., -4.6566e-10, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0203, -0.0290, -0.0316, -0.0318, -0.0063, 0.0121, 0.0079, -0.0043, + -0.0078, -0.0175], device='cuda:0'), grad: tensor([ 2.9337e-08, 2.9337e-08, -1.0105e-07, -1.2573e-08, 3.2596e-09, + 1.0710e-08, -1.1176e-08, 2.0023e-08, 3.7719e-08, 4.1910e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 220.69, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4767 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.13 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.2361, -0.3021, 0.1205, ..., -0.1515, 0.0438, 0.0394], + [-0.1490, -0.0753, -0.1112, ..., -0.2213, -0.0790, -0.0238], + [ 0.0116, -0.1852, -0.2307, ..., -0.1790, 0.0378, -0.4024], + ..., + [-0.2216, 0.1844, 0.0310, ..., 0.2257, -0.0711, -0.1700], + [-0.1950, -0.1949, 0.2136, ..., -0.1774, -0.1616, 0.1671], + [ 0.0181, -0.3531, 0.1909, ..., 0.0621, -0.2077, -0.1692]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.5832e-08, 2.7940e-09, ..., 1.1642e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 1.2107e-08, 9.3132e-10, ..., 8.3819e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, -2.1979e-07, -3.0734e-08, ..., -1.9278e-07, + 0.0000e+00, -1.8626e-09], + [ 7.9162e-09, 9.3132e-10, 5.5879e-09, ..., 4.1910e-09, + 0.0000e+00, -9.3132e-10], + [-2.7008e-08, 9.3132e-10, -2.7474e-08, ..., -1.2107e-08, + 0.0000e+00, -3.2596e-09]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0207, -0.0300, -0.0285, -0.0318, -0.0063, 0.0121, 0.0079, -0.0044, + -0.0077, -0.0175], device='cuda:0'), grad: tensor([ 9.3132e-09, 3.8184e-08, 4.6566e-08, 5.5879e-09, 4.8662e-07, + 6.6590e-08, 3.2596e-09, -5.4762e-07, 3.3993e-08, -1.3364e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 220.76, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4716 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.11 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.2375, -0.3028, 0.1204, ..., -0.1518, 0.0432, 0.0386], + [-0.1491, -0.0753, -0.1112, ..., -0.2215, -0.0760, -0.0238], + [ 0.0112, -0.1852, -0.2314, ..., -0.1793, 0.0348, -0.4036], + ..., + [-0.2217, 0.1845, 0.0310, ..., 0.2261, -0.0716, -0.1700], + [-0.1952, -0.1952, 0.2147, ..., -0.1778, -0.1619, 0.1673], + [ 0.0165, -0.3543, 0.1913, ..., 0.0597, -0.2078, -0.1695]], + device='cuda:0'), grad: tensor([[ 4.8894e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.3551e-09], + [ 4.6566e-10, 1.6298e-09, 2.3283e-10, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 9.3132e-10, -1.5832e-08, -2.7940e-09, ..., -1.4668e-08, + 0.0000e+00, 9.3132e-10], + [ 1.0012e-08, -9.3132e-10, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.9581e-09], + [ 1.1642e-09, 2.3283e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0215, -0.0298, -0.0293, -0.0318, -0.0036, 0.0121, 0.0079, -0.0044, + -0.0077, -0.0202], device='cuda:0'), grad: tensor([ 1.4668e-08, 6.7521e-09, -4.6566e-10, 1.5367e-08, 3.2596e-08, + -3.5111e-07, 3.0315e-07, -2.5844e-08, 9.3132e-09, 4.8894e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 220.81, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4470 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.14 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.2380, -0.3031, 0.1194, ..., -0.1542, 0.0432, 0.0364], + [-0.1491, -0.0754, -0.1112, ..., -0.2216, -0.0760, -0.0238], + [ 0.0113, -0.1849, -0.2315, ..., -0.1788, 0.0349, -0.4036], + ..., + [-0.2218, 0.1846, 0.0310, ..., 0.2265, -0.0716, -0.1700], + [-0.1954, -0.1972, 0.2145, ..., -0.1782, -0.1625, 0.1672], + [ 0.0165, -0.3549, 0.1922, ..., 0.0601, -0.2078, -0.1695]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, -1.8626e-09, ..., 2.3283e-10, + -2.3283e-10, -1.8626e-09], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 2.3283e-10, 2.3283e-10, ..., 9.3132e-10, + 0.0000e+00, 6.9849e-10], + [ 2.3283e-10, 6.9849e-10, -4.1910e-09, ..., -8.3819e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0246, -0.0298, -0.0290, -0.0317, -0.0041, 0.0120, 0.0084, -0.0043, + -0.0083, -0.0200], device='cuda:0'), grad: tensor([-5.8208e-09, 1.2340e-08, -5.8208e-09, -7.6834e-09, 1.2107e-08, + 1.1176e-08, -1.2573e-08, 4.6566e-09, 8.1491e-09, -1.5134e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 220.47, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4618 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.15 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.2381, -0.3032, 0.1199, ..., -0.1541, 0.0432, 0.0370], + [-0.1492, -0.0754, -0.1112, ..., -0.2218, -0.0760, -0.0238], + [ 0.0115, -0.1850, -0.2325, ..., -0.1789, 0.0349, -0.4042], + ..., + [-0.2218, 0.1846, 0.0310, ..., 0.2264, -0.0716, -0.1701], + [-0.1956, -0.1975, 0.2148, ..., -0.1791, -0.1628, 0.1674], + [ 0.0165, -0.3553, 0.1925, ..., 0.0604, -0.2078, -0.1700]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.7940e-09, 6.9849e-10, ..., 1.1642e-09, + 0.0000e+00, -6.9849e-10], + [ 9.3132e-10, 2.0955e-09, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 9.3132e-10, -1.1176e-08, -5.3551e-09, ..., -7.4506e-09, + 0.0000e+00, 2.3283e-10], + [ 2.5611e-09, 2.0955e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.6298e-09, 1.4668e-08, -9.5228e-08, ..., -2.2375e-07, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0244, -0.0298, -0.0290, -0.0317, -0.0041, 0.0120, 0.0084, -0.0044, + -0.0084, -0.0199], device='cuda:0'), grad: tensor([ 1.1642e-09, -5.1223e-09, 1.0710e-08, -8.5915e-08, 4.4028e-07, + 5.3551e-08, 2.7940e-09, -1.4668e-08, 8.8476e-09, -4.0978e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 220.82, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4953 re_mapping 0.0030 re_causal 0.0092 /// teacc 99.10 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.2382, -0.3038, 0.1179, ..., -0.1566, 0.0430, 0.0368], + [-0.1493, -0.0754, -0.1113, ..., -0.2219, -0.0758, -0.0242], + [ 0.0115, -0.1851, -0.2328, ..., -0.1791, 0.0347, -0.4044], + ..., + [-0.2219, 0.1846, 0.0309, ..., 0.2263, -0.0718, -0.1702], + [-0.1967, -0.1982, 0.2147, ..., -0.1801, -0.1629, 0.1668], + [ 0.0167, -0.3560, 0.1951, ..., 0.0617, -0.2079, -0.1699]], + device='cuda:0'), grad: tensor([[ 6.2864e-09, 1.1642e-09, 4.6566e-10, ..., 6.9849e-10, + 0.0000e+00, 2.7940e-09], + [ 1.2806e-08, 2.8638e-08, -1.8626e-08, ..., 2.7940e-08, + 0.0000e+00, -3.8184e-08], + [ 2.1886e-08, 6.2864e-08, 4.4238e-09, ..., 6.1002e-08, + 0.0000e+00, 7.6834e-09], + ..., + [ 1.0012e-08, -2.0862e-07, -1.1409e-08, ..., -2.1211e-07, + 0.0000e+00, 7.2177e-09], + [ 6.8918e-08, 6.7521e-09, 1.6531e-08, ..., 1.3970e-09, + 0.0000e+00, 6.7288e-08], + [ 3.1921e-07, 1.4435e-07, -1.5600e-08, ..., 7.9861e-08, + 0.0000e+00, 1.1176e-07]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0258, -0.0299, -0.0290, -0.0316, -0.0041, 0.0120, 0.0086, -0.0045, + -0.0096, -0.0189], device='cuda:0'), grad: tensor([ 2.2817e-08, -1.7905e-07, 2.2654e-07, 3.0603e-06, 1.5856e-07, + -3.9935e-06, 8.8243e-08, -6.2864e-07, 3.6834e-07, 8.7172e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 220.68, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4507 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.2385, -0.3045, 0.1175, ..., -0.1572, 0.0429, 0.0367], + [-0.1494, -0.0754, -0.1113, ..., -0.2221, -0.0758, -0.0241], + [ 0.0118, -0.1852, -0.2333, ..., -0.1796, 0.0347, -0.4052], + ..., + [-0.2221, 0.1846, 0.0309, ..., 0.2265, -0.0718, -0.1702], + [-0.1981, -0.1994, 0.2145, ..., -0.1811, -0.1629, 0.1664], + [ 0.0164, -0.3571, 0.1960, ..., 0.0620, -0.2079, -0.1701]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.6298e-09, -1.4203e-08, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 6.9849e-10, 6.0536e-09, 5.8208e-09, ..., 1.1642e-09, + 0.0000e+00, -6.9849e-10], + [-1.1642e-09, 1.6298e-09, 2.5611e-09, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 2.3283e-10, -6.5193e-09, 9.3132e-10, ..., -2.0955e-09, + 0.0000e+00, 1.3970e-09], + [ 2.0955e-09, 2.3283e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 1.3970e-09], + [ 8.3819e-09, 9.3132e-10, -1.1409e-08, ..., -1.4203e-08, + 0.0000e+00, 6.2864e-09]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0262, -0.0297, -0.0291, -0.0315, -0.0040, 0.0121, 0.0086, -0.0045, + -0.0109, -0.0189], device='cuda:0'), grad: tensor([-3.6322e-08, 6.9384e-08, -8.9966e-07, 8.3121e-08, 2.3912e-07, + -6.7521e-08, 1.4203e-08, 7.1246e-08, 5.5414e-07, -1.8161e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 220.28, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4360 re_mapping 0.0030 re_causal 0.0084 /// teacc 99.09 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.2386, -0.3047, 0.1178, ..., -0.1573, 0.0429, 0.0369], + [-0.1497, -0.0754, -0.1113, ..., -0.2223, -0.0758, -0.0241], + [ 0.0117, -0.1853, -0.2340, ..., -0.1797, 0.0347, -0.4055], + ..., + [-0.2224, 0.1849, 0.0311, ..., 0.2278, -0.0719, -0.1703], + [-0.1986, -0.1997, 0.2148, ..., -0.1817, -0.1629, 0.1663], + [ 0.0163, -0.3582, 0.1975, ..., 0.0636, -0.2079, -0.1699]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 0.0000e+00, 6.9849e-10, 2.3283e-10, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -9.7789e-09, -7.4506e-09, ..., -1.4901e-08, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 6.5193e-09, ..., 1.3504e-08, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0262, -0.0297, -0.0289, -0.0317, -0.0053, 0.0121, 0.0086, -0.0044, + -0.0116, -0.0179], device='cuda:0'), grad: tensor([ 1.2084e-07, 4.4238e-09, 2.0955e-09, 6.9849e-10, 1.6997e-08, + 6.0536e-08, -2.0256e-07, -3.1432e-08, 2.7940e-09, 3.3062e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 220.47, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4739 re_mapping 0.0030 re_causal 0.0085 /// teacc 99.13 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.2390, -0.3049, 0.1181, ..., -0.1574, 0.0430, 0.0366], + [-0.1491, -0.0756, -0.1113, ..., -0.2230, -0.0733, -0.0237], + [ 0.0118, -0.1857, -0.2351, ..., -0.1804, 0.0323, -0.4055], + ..., + [-0.2227, 0.1851, 0.0311, ..., 0.2287, -0.0722, -0.1704], + [-0.1995, -0.1999, 0.2155, ..., -0.1819, -0.1634, 0.1663], + [ 0.0163, -0.3587, 0.1995, ..., 0.0654, -0.2080, -0.1704]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, -1.6764e-08], + [ 7.6834e-09, 2.3283e-10, 1.1176e-08, ..., 2.3283e-10, + 0.0000e+00, 3.7719e-08], + [ 6.9849e-09, 0.0000e+00, 5.8208e-09, ..., 0.0000e+00, + 0.0000e+00, 3.0268e-08], + ..., + [ 1.8626e-09, 2.3283e-10, 3.9581e-09, ..., 6.2864e-09, + 0.0000e+00, 9.3132e-09], + [-3.5041e-07, 0.0000e+00, -2.5774e-07, ..., 4.4238e-09, + 0.0000e+00, -1.4678e-06], + [-2.3283e-09, 2.3283e-10, -5.8208e-09, ..., -6.9849e-09, + 0.0000e+00, -4.8894e-09]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0265, -0.0290, -0.0300, -0.0317, -0.0071, 0.0121, 0.0085, -0.0043, + -0.0118, -0.0165], device='cuda:0'), grad: tensor([-6.9151e-08, 1.2456e-07, 9.5461e-08, -4.5868e-08, -2.7940e-08, + 3.9414e-06, 3.1106e-07, 5.9837e-08, -4.3511e-06, -1.5600e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 220.19, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4619 re_mapping 0.0031 re_causal 0.0087 /// teacc 98.99 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.2392, -0.3050, 0.1184, ..., -0.1573, 0.0430, 0.0377], + [-0.1493, -0.0766, -0.1129, ..., -0.2234, -0.0732, -0.0266], + [ 0.0120, -0.1861, -0.2358, ..., -0.1813, 0.0323, -0.4071], + ..., + [-0.2228, 0.1862, 0.0326, ..., 0.2291, -0.0726, -0.1679], + [-0.1996, -0.2023, 0.2165, ..., -0.1829, -0.1640, 0.1677], + [ 0.0163, -0.3592, 0.2004, ..., 0.0659, -0.2080, -0.1706]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.4925e-09, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, -1.2573e-08], + [ 0.0000e+00, 3.0268e-09, 2.3283e-10, ..., 4.1910e-09, + 0.0000e+00, 1.6298e-09], + ..., + [ 0.0000e+00, 3.3760e-07, 1.4435e-08, ..., 4.6566e-10, + 0.0000e+00, 1.0477e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 6.0536e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0260, -0.0307, -0.0298, -0.0317, -0.0071, 0.0120, 0.0084, -0.0027, + -0.0107, -0.0162], device='cuda:0'), grad: tensor([ 1.0012e-08, -1.6182e-07, 5.8440e-08, -5.7649e-07, -1.5530e-07, + 3.0966e-08, 6.0536e-09, 7.2038e-07, 1.3271e-08, 6.2864e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 220.15, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4501 re_mapping 0.0032 re_causal 0.0090 /// teacc 99.10 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.2394, -0.3046, 0.1185, ..., -0.1573, 0.0431, 0.0375], + [-0.1494, -0.0766, -0.1129, ..., -0.2236, -0.0733, -0.0267], + [ 0.0119, -0.1857, -0.2363, ..., -0.1819, 0.0324, -0.4093], + ..., + [-0.2229, 0.1863, 0.0326, ..., 0.2294, -0.0726, -0.1680], + [-0.1996, -0.2027, 0.2167, ..., -0.1830, -0.1642, 0.1682], + [ 0.0163, -0.3595, 0.2005, ..., 0.0654, -0.2080, -0.1708]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 4.1910e-09, 2.7940e-09, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, -6.3563e-08, 7.2177e-09, ..., 3.7253e-09, + 1.3970e-09, 2.3283e-10], + [ 4.6566e-10, 5.4250e-08, 2.5146e-08, ..., 9.3132e-10, + -1.6298e-09, 2.3283e-10], + ..., + [ 2.3283e-10, 5.8208e-08, 1.8626e-09, ..., -6.9849e-09, + 0.0000e+00, 2.3283e-10], + [ 1.9325e-08, 4.7265e-08, 2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, 4.1910e-09], + [ 3.0734e-08, 2.3283e-09, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 1.7462e-08]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0263, -0.0305, -0.0296, -0.0318, -0.0067, 0.0116, 0.0092, -0.0031, + -0.0107, -0.0167], device='cuda:0'), grad: tensor([ 1.7229e-08, -2.9709e-07, 9.1502e-08, -2.6356e-07, 3.7253e-09, + -8.5682e-08, 1.3504e-08, 3.3691e-07, 1.4203e-07, 4.8196e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 220.33, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4870 re_mapping 0.0031 re_causal 0.0086 /// teacc 99.16 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.2406, -0.3048, 0.1183, ..., -0.1583, 0.0418, 0.0375], + [-0.1498, -0.0765, -0.1131, ..., -0.2239, -0.0733, -0.0267], + [ 0.0115, -0.1858, -0.2370, ..., -0.1824, 0.0322, -0.4108], + ..., + [-0.2230, 0.1862, 0.0327, ..., 0.2298, -0.0727, -0.1681], + [-0.2005, -0.2032, 0.2169, ..., -0.1830, -0.1647, 0.1681], + [ 0.0175, -0.3603, 0.2017, ..., 0.0656, -0.2082, -0.1715]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.4238e-09, 1.6298e-09, ..., 6.0536e-09, + 0.0000e+00, -6.9849e-10], + [ 3.2596e-09, 1.7136e-07, 6.1700e-08, ..., 2.0722e-07, + 0.0000e+00, 1.6298e-09], + [ 1.3970e-09, 2.4214e-08, 6.2864e-09, ..., 2.8173e-08, + 0.0000e+00, 2.3283e-10], + ..., + [ 4.8894e-09, -4.5472e-07, -1.2666e-07, ..., -6.7102e-07, + 0.0000e+00, 2.0955e-09], + [ 8.8476e-09, 8.6147e-09, 2.3283e-09, ..., 5.1223e-09, + 0.0000e+00, 7.9162e-09], + [ 6.0536e-09, 7.6136e-08, -3.8417e-08, ..., 6.6590e-08, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0272, -0.0302, -0.0288, -0.0319, -0.0082, 0.0116, 0.0094, -0.0037, + -0.0118, -0.0150], device='cuda:0'), grad: tensor([ 4.3306e-08, 1.1642e-06, -1.0533e-06, 3.8324e-07, 1.7192e-06, + 4.6031e-07, 7.5204e-08, -3.2559e-06, 6.5425e-08, 4.0815e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 220.59, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4645 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.07 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.2408, -0.3050, 0.1191, ..., -0.1583, 0.0418, 0.0383], + [-0.1499, -0.0769, -0.1133, ..., -0.2254, -0.0733, -0.0269], + [ 0.0115, -0.1868, -0.2382, ..., -0.1849, 0.0322, -0.4116], + ..., + [-0.2231, 0.1868, 0.0330, ..., 0.2320, -0.0727, -0.1681], + [-0.2008, -0.2050, 0.2175, ..., -0.1848, -0.1647, 0.1688], + [ 0.0188, -0.3615, 0.2024, ..., 0.0655, -0.2082, -0.1718]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.3283e-10, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 1.7928e-08, 6.0536e-09, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 3.4925e-09, 1.1642e-09, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, -2.1420e-08, -8.6147e-09, ..., -1.1176e-08, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 6.9849e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 7.6834e-09, 3.9581e-09, 1.3970e-09, ..., 1.1642e-09, + 0.0000e+00, 1.0245e-08]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0269, -0.0306, -0.0285, -0.0319, -0.0097, 0.0114, 0.0096, -0.0035, + -0.0120, -0.0135], device='cuda:0'), grad: tensor([-3.9581e-09, 4.2375e-08, 9.0804e-09, 1.8394e-08, 6.5193e-09, + -5.9372e-08, 1.0943e-08, -4.9593e-08, 6.0536e-09, 2.7707e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 220.33, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4427 re_mapping 0.0031 re_causal 0.0088 /// teacc 99.09 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.2404, -0.3051, 0.1193, ..., -0.1583, 0.0418, 0.0391], + [-0.1500, -0.0771, -0.1133, ..., -0.2258, -0.0733, -0.0269], + [ 0.0118, -0.1868, -0.2382, ..., -0.1847, 0.0322, -0.4119], + ..., + [-0.2231, 0.1870, 0.0330, ..., 0.2323, -0.0727, -0.1681], + [-0.2010, -0.2052, 0.2176, ..., -0.1852, -0.1647, 0.1690], + [ 0.0188, -0.3620, 0.2024, ..., 0.0655, -0.2082, -0.1720]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 6.2864e-09, 3.0268e-09, ..., 5.3551e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 8.3819e-09, 2.3283e-10, ..., 7.6834e-09, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, -2.1188e-08, 5.1223e-09, ..., 2.3097e-07, + 0.0000e+00, 4.4238e-09], + [ 9.3132e-10, 1.6298e-09, -1.3970e-09, ..., 6.9849e-10, + 0.0000e+00, -6.9849e-10], + [-1.7229e-08, 4.6566e-10, -7.6368e-08, ..., -2.0955e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0265, -0.0315, -0.0266, -0.0319, -0.0096, 0.0114, 0.0096, -0.0035, + -0.0123, -0.0135], device='cuda:0'), grad: tensor([ 7.6834e-09, 2.4913e-08, 2.4680e-08, 1.3970e-08, -4.3330e-07, + 4.7497e-08, -4.7265e-08, 7.5251e-07, 4.1910e-08, -4.4378e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 220.72, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4593 re_mapping 0.0032 re_causal 0.0090 /// teacc 99.17 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.2404, -0.3051, 0.1194, ..., -0.1583, 0.0418, 0.0393], + [-0.1500, -0.0771, -0.1134, ..., -0.2259, -0.0732, -0.0269], + [ 0.0118, -0.1869, -0.2394, ..., -0.1848, 0.0322, -0.4125], + ..., + [-0.2233, 0.1870, 0.0330, ..., 0.2323, -0.0727, -0.1682], + [-0.2028, -0.2055, 0.2182, ..., -0.1857, -0.1648, 0.1682], + [ 0.0188, -0.3622, 0.2025, ..., 0.0656, -0.2083, -0.1722]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 4.0559e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 1.6298e-08, 5.8208e-09, ..., 2.3283e-09, + 0.0000e+00, -4.6566e-10], + [-9.3132e-10, 2.3283e-09, 3.4925e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, -1.9092e-08, -7.4506e-09, ..., -3.2596e-09, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 2.3283e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 1.1642e-09, 2.0955e-09, -4.1374e-07, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0265, -0.0315, -0.0272, -0.0319, -0.0098, 0.0115, 0.0101, -0.0036, + -0.0130, -0.0135], device='cuda:0'), grad: tensor([ 8.6147e-07, 2.1420e-08, 1.0710e-08, -2.3050e-08, -1.3271e-08, + 1.6997e-08, 1.3039e-08, -2.4447e-08, 2.5611e-09, -8.6613e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 221.33, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4350 re_mapping 0.0032 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.2412, -0.3053, 0.1195, ..., -0.1583, 0.0417, 0.0393], + [-0.1500, -0.0771, -0.1134, ..., -0.2261, -0.0732, -0.0269], + [ 0.0113, -0.1870, -0.2398, ..., -0.1849, 0.0322, -0.4137], + ..., + [-0.2239, 0.1870, 0.0330, ..., 0.2325, -0.0727, -0.1682], + [-0.2030, -0.2057, 0.2193, ..., -0.1852, -0.1649, 0.1686], + [ 0.0192, -0.3635, 0.2026, ..., 0.0657, -0.2083, -0.1729]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -1.6764e-08, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-5.0757e-08, 1.2806e-08, 1.1642e-09, ..., 1.0245e-08, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 1.1642e-09, 3.0268e-09, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 1.3504e-08, -1.9558e-08, 2.0955e-09, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, -1.1642e-09], + [ 2.5611e-08, 2.3283e-09, 1.8626e-09, ..., 2.1863e-07, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0266, -0.0315, -0.0270, -0.0315, -0.0098, 0.0113, 0.0101, -0.0037, + -0.0126, -0.0136], device='cuda:0'), grad: tensor([-4.4005e-08, -6.4494e-07, 2.3749e-08, 5.1223e-08, -9.7696e-07, + 1.0012e-07, -1.8370e-07, 1.9139e-07, 1.4668e-08, 1.4687e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 220.53, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4610 re_mapping 0.0033 re_causal 0.0093 /// teacc 99.15 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.2424, -0.3053, 0.1198, ..., -0.1583, 0.0417, 0.0393], + [-0.1497, -0.0771, -0.1134, ..., -0.2262, -0.0732, -0.0268], + [ 0.0112, -0.1872, -0.2408, ..., -0.1851, 0.0322, -0.4140], + ..., + [-0.2243, 0.1871, 0.0330, ..., 0.2327, -0.0727, -0.1683], + [-0.2039, -0.2057, 0.2203, ..., -0.1852, -0.1649, 0.1685], + [ 0.0192, -0.3645, 0.2026, ..., 0.0655, -0.2083, -0.1739]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.3970e-09, -9.3132e-10, ..., 4.4238e-09, + 0.0000e+00, -2.3283e-10], + [ 0.0000e+00, 1.3970e-09, -1.6298e-09, ..., 9.5461e-09, + 0.0000e+00, -3.9581e-09], + [ 2.3283e-10, 1.4901e-08, 4.4238e-09, ..., 2.7940e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 4.4238e-09, 2.0955e-09, ..., 6.0536e-09, + 0.0000e+00, 1.1642e-09], + [ 4.1910e-09, 1.8626e-09, 1.1642e-09, ..., 1.3970e-09, + 0.0000e+00, 3.2596e-09], + [ 1.4435e-08, 1.6298e-09, -3.0268e-09, ..., 9.7323e-08, + 0.0000e+00, 9.5461e-09]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0265, -0.0315, -0.0270, -0.0315, -0.0097, 0.0113, 0.0102, -0.0038, + -0.0128, -0.0137], device='cuda:0'), grad: tensor([ 2.6543e-08, 1.6065e-08, 5.9837e-08, -8.1491e-08, -8.1351e-07, + 9.5461e-09, 1.2433e-07, 5.6578e-08, 2.9337e-08, 5.7276e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 220.38, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4666 re_mapping 0.0032 re_causal 0.0090 /// teacc 99.15 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.2427, -0.3061, 0.1199, ..., -0.1584, 0.0417, 0.0391], + [-0.1491, -0.0776, -0.1135, ..., -0.2273, -0.0732, -0.0268], + [ 0.0112, -0.1876, -0.2412, ..., -0.1859, 0.0322, -0.4141], + ..., + [-0.2260, 0.1878, 0.0332, ..., 0.2346, -0.0728, -0.1684], + [-0.2047, -0.2059, 0.2211, ..., -0.1854, -0.1649, 0.1685], + [ 0.0184, -0.3685, 0.2026, ..., 0.0651, -0.2083, -0.1756]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, 1.8626e-09, 6.9849e-10, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., -1.6298e-09, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 6.9849e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0268, -0.0317, -0.0269, -0.0315, -0.0096, 0.0114, 0.0103, -0.0035, + -0.0131, -0.0139], device='cuda:0'), grad: tensor([-1.0245e-08, -4.6566e-09, -5.5879e-09, -2.1188e-08, 2.0955e-09, + 1.7229e-08, -5.3551e-09, 2.6543e-08, 1.8626e-09, 4.1910e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 220.24, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4926 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.12 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.2426, -0.3061, 0.1201, ..., -0.1584, 0.0417, 0.0395], + [-0.1496, -0.0777, -0.1136, ..., -0.2281, -0.0732, -0.0269], + [ 0.0108, -0.1878, -0.2414, ..., -0.1861, 0.0322, -0.4142], + ..., + [-0.2279, 0.1879, 0.0332, ..., 0.2354, -0.0728, -0.1685], + [-0.2054, -0.2063, 0.2215, ..., -0.1850, -0.1649, 0.1686], + [ 0.0183, -0.3696, 0.2026, ..., 0.0650, -0.2084, -0.1763]], + device='cuda:0'), grad: tensor([[-2.3283e-10, 0.0000e+00, -1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, -3.4459e-08], + [ 0.0000e+00, 3.4925e-09, 9.3132e-10, ..., 3.0268e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, -6.9849e-10, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.3551e-09], + [ 2.3283e-10, 1.8626e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0266, -0.0319, -0.0267, -0.0313, -0.0096, 0.0112, 0.0105, -0.0035, + -0.0133, -0.0141], device='cuda:0'), grad: tensor([-9.1502e-08, 2.1188e-08, -5.1223e-09, -6.0536e-09, 1.6298e-09, + 2.9569e-08, 2.5146e-08, 2.3283e-10, 3.2363e-08, 6.2864e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 220.48, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4444 re_mapping 0.0030 re_causal 0.0085 /// teacc 99.16 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.2429, -0.3061, 0.1231, ..., -0.1585, 0.0417, 0.0418], + [-0.1499, -0.0779, -0.1137, ..., -0.2287, -0.0733, -0.0271], + [ 0.0107, -0.1879, -0.2422, ..., -0.1863, 0.0321, -0.4144], + ..., + [-0.2281, 0.1881, 0.0333, ..., 0.2361, -0.0728, -0.1686], + [-0.2058, -0.2066, 0.2216, ..., -0.1854, -0.1650, 0.1687], + [ 0.0183, -0.3706, 0.2006, ..., 0.0649, -0.2084, -0.1778]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 4.4238e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3271e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, -1.1642e-09, ..., 0.0000e+00, + 0.0000e+00, -1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0238, -0.0319, -0.0266, -0.0312, -0.0095, 0.0111, 0.0108, -0.0036, + -0.0136, -0.0146], device='cuda:0'), grad: tensor([ 1.0477e-08, 5.3551e-09, 1.6298e-09, -4.5868e-08, 4.6566e-09, + 3.9814e-08, -5.3551e-08, 4.1444e-08, 9.3132e-10, 2.5611e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 220.50, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4619 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.2431, -0.3063, 0.1232, ..., -0.1585, 0.0417, 0.0420], + [-0.1503, -0.0776, -0.1137, ..., -0.2291, -0.0733, -0.0273], + [ 0.0108, -0.1878, -0.2424, ..., -0.1859, 0.0321, -0.4145], + ..., + [-0.2283, 0.1878, 0.0333, ..., 0.2365, -0.0728, -0.1686], + [-0.2068, -0.2068, 0.2228, ..., -0.1854, -0.1650, 0.1654], + [ 0.0184, -0.3712, 0.2007, ..., 0.0649, -0.2084, -0.1784]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 6.9849e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, -6.0536e-09, 1.6298e-09, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 1.1176e-08, 6.9849e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 7.4506e-09, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 1.8626e-09, 9.5461e-09, 3.2596e-08, ..., 1.9092e-08, + 0.0000e+00, 1.0012e-08], + [ 9.3132e-10, 0.0000e+00, -3.7253e-08, ..., -2.6310e-08, + 0.0000e+00, -1.1874e-08]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0236, -0.0317, -0.0262, -0.0312, -0.0095, 0.0110, 0.0124, -0.0042, + -0.0186, -0.0146], device='cuda:0'), grad: tensor([ 6.7521e-09, -1.7835e-07, 3.7020e-08, -6.9384e-08, 1.3062e-07, + -3.9581e-09, 2.3516e-08, 6.2864e-08, 1.3947e-07, -1.4110e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 220.66, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4665 re_mapping 0.0032 re_causal 0.0091 /// teacc 99.15 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.2433, -0.3065, 0.1232, ..., -0.1587, 0.0417, 0.0421], + [-0.1500, -0.0777, -0.1138, ..., -0.2299, -0.0733, -0.0273], + [ 0.0108, -0.1884, -0.2445, ..., -0.1867, 0.0321, -0.4154], + ..., + [-0.2283, 0.1880, 0.0334, ..., 0.2376, -0.0728, -0.1686], + [-0.2077, -0.2069, 0.2239, ..., -0.1860, -0.1651, 0.1656], + [ 0.0183, -0.3722, 0.2008, ..., 0.0649, -0.2084, -0.1786]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 7.6834e-09, 1.6298e-09, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -7.6834e-09, -1.6298e-09, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0235, -0.0318, -0.0264, -0.0311, -0.0096, 0.0110, 0.0124, -0.0040, + -0.0186, -0.0146], device='cuda:0'), grad: tensor([ 2.7940e-09, -9.3132e-10, 3.0501e-08, 2.3283e-10, -1.8626e-09, + 1.6298e-08, -4.0513e-08, -1.2107e-08, 4.6566e-09, 1.4435e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 220.63, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4606 re_mapping 0.0033 re_causal 0.0091 /// teacc 99.16 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.2435, -0.3067, 0.1242, ..., -0.1587, 0.0419, 0.0423], + [-0.1500, -0.0777, -0.1139, ..., -0.2303, -0.0733, -0.0270], + [ 0.0107, -0.1889, -0.2473, ..., -0.1875, 0.0321, -0.4177], + ..., + [-0.2284, 0.1878, 0.0334, ..., 0.2381, -0.0728, -0.1686], + [-0.2082, -0.2070, 0.2248, ..., -0.1860, -0.1651, 0.1657], + [ 0.0185, -0.3726, 0.2005, ..., 0.0650, -0.2084, -0.1790]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.5134e-09, 5.8208e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 7.6834e-09, 3.0268e-09, ..., 5.1223e-09, + 0.0000e+00, 1.0477e-09], + [ 0.0000e+00, 1.8626e-08, 3.8417e-09, ..., 2.2817e-08, + 0.0000e+00, 1.1642e-10], + ..., + [ 1.1642e-09, -3.5623e-08, -5.2387e-09, ..., -4.8196e-08, + 0.0000e+00, 0.0000e+00], + [ 1.6298e-09, 6.4028e-09, 1.1642e-10, ..., 1.1642e-10, + 0.0000e+00, -3.2596e-09], + [ 5.1223e-09, 2.3516e-08, 8.6147e-09, ..., 1.0477e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0227, -0.0303, -0.0282, -0.0311, -0.0097, 0.0111, 0.0124, -0.0048, + -0.0185, -0.0147], device='cuda:0'), grad: tensor([ 3.9581e-09, 2.5379e-08, 4.9360e-08, -8.9174e-08, 1.6065e-08, + 2.7008e-08, 8.1491e-10, -1.1094e-07, 1.3039e-08, 6.6590e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 220.50, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4738 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.16 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.2437, -0.3069, 0.1243, ..., -0.1587, 0.0419, 0.0422], + [-0.1501, -0.0776, -0.1139, ..., -0.2306, -0.0733, -0.0271], + [ 0.0108, -0.1888, -0.2471, ..., -0.1871, 0.0321, -0.4179], + ..., + [-0.2285, 0.1878, 0.0334, ..., 0.2385, -0.0728, -0.1688], + [-0.2083, -0.2076, 0.2257, ..., -0.1872, -0.1651, 0.1659], + [ 0.0182, -0.3732, 0.2006, ..., 0.0651, -0.2084, -0.1797]], + device='cuda:0'), grad: tensor([[2.3283e-10, 9.1968e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 2.3283e-10], + [8.1491e-10, 2.8173e-08, 0.0000e+00, ..., 5.8208e-10, 0.0000e+00, + 6.9849e-10], + [1.1642e-10, 1.9209e-08, 1.1642e-10, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + ..., + [1.1642e-10, 9.5461e-09, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 1.1642e-10], + [6.5193e-09, 4.3074e-09, 0.0000e+00, ..., 0.0000e+00, 1.1642e-10, + 5.7044e-09], + [1.2806e-09, 2.2119e-09, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 8.1491e-10]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0228, -0.0301, -0.0277, -0.0311, -0.0097, 0.0110, 0.0126, -0.0051, + -0.0187, -0.0147], device='cuda:0'), grad: tensor([ 3.5274e-08, 1.0617e-07, 6.8103e-08, -2.7521e-07, 7.2177e-09, + -7.3691e-08, 6.2864e-08, 3.7951e-08, 2.9220e-08, 1.2806e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 220.52, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4351 re_mapping 0.0031 re_causal 0.0087 /// teacc 99.06 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.2438, -0.3077, 0.1243, ..., -0.1588, 0.0419, 0.0421], + [-0.1502, -0.0779, -0.1139, ..., -0.2312, -0.0732, -0.0271], + [ 0.0108, -0.1891, -0.2475, ..., -0.1877, 0.0320, -0.4181], + ..., + [-0.2285, 0.1880, 0.0334, ..., 0.2394, -0.0728, -0.1688], + [-0.2092, -0.2084, 0.2257, ..., -0.1873, -0.1651, 0.1656], + [ 0.0186, -0.3746, 0.2008, ..., 0.0646, -0.2084, -0.1799]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.9849e-10, 1.0477e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 2.6776e-09, 9.3132e-10, ..., 1.1642e-10, + 0.0000e+00, 1.0477e-09], + [ 0.0000e+00, 6.9849e-10, 2.2119e-09, ..., 0.0000e+00, + 0.0000e+00, 5.1223e-09], + ..., + [ 1.1642e-10, 7.7998e-09, 8.1491e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 1.1642e-10, 2.4447e-09, -1.5483e-08, ..., 0.0000e+00, + 0.0000e+00, -2.0955e-08], + [ 0.0000e+00, 4.6566e-10, 3.3760e-09, ..., 0.0000e+00, + 0.0000e+00, 4.5402e-09]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0231, -0.0302, -0.0277, -0.0310, -0.0096, 0.0110, 0.0127, -0.0050, + -0.0190, -0.0148], device='cuda:0'), grad: tensor([ 8.6147e-09, 1.1642e-09, 5.7044e-09, -4.9546e-07, 8.1491e-10, + 4.6776e-07, 1.4086e-08, 3.5274e-08, -5.6112e-08, 1.5716e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 220.20, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4496 re_mapping 0.0030 re_causal 0.0085 /// teacc 98.98 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.2441, -0.3080, 0.1242, ..., -0.1588, 0.0419, 0.0419], + [-0.1505, -0.0779, -0.1140, ..., -0.2314, -0.0732, -0.0271], + [ 0.0109, -0.1898, -0.2483, ..., -0.1882, 0.0320, -0.4185], + ..., + [-0.2288, 0.1878, 0.0334, ..., 0.2395, -0.0728, -0.1689], + [-0.2096, -0.2092, 0.2262, ..., -0.1879, -0.1651, 0.1657], + [ 0.0185, -0.3750, 0.2012, ..., 0.0631, -0.2084, -0.1799]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.0955e-09, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 6.2864e-09, 6.9849e-10, ..., 2.5611e-09, + 0.0000e+00, 3.4925e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, -1.8626e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0233, -0.0302, -0.0278, -0.0306, -0.0088, 0.0109, 0.0128, -0.0054, + -0.0192, -0.0155], device='cuda:0'), grad: tensor([ 1.8626e-09, -1.0640e-07, 9.3132e-09, -2.7241e-08, -3.9581e-09, + 8.1491e-09, 1.1642e-09, 1.2876e-07, 2.3283e-10, -6.7521e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 220.52, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4382 re_mapping 0.0031 re_causal 0.0087 /// teacc 99.18 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.2444, -0.3082, 0.1243, ..., -0.1588, 0.0419, 0.0420], + [-0.1506, -0.0779, -0.1140, ..., -0.2315, -0.0732, -0.0271], + [ 0.0111, -0.1899, -0.2482, ..., -0.1884, 0.0320, -0.4188], + ..., + [-0.2293, 0.1877, 0.0334, ..., 0.2395, -0.0728, -0.1691], + [-0.2098, -0.2095, 0.2266, ..., -0.1879, -0.1651, 0.1659], + [ 0.0184, -0.3753, 0.2014, ..., 0.0631, -0.2085, -0.1803]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, -2.5611e-09, ..., 0.0000e+00, + 0.0000e+00, -1.1642e-09], + [ 0.0000e+00, 1.3970e-09, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.5832e-08, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 1.1642e-09, 6.9849e-10, ..., 6.9849e-10, + 0.0000e+00, 4.6566e-10], + [-3.7253e-09, 2.3283e-10, -1.5367e-08, ..., -2.3749e-08, + 0.0000e+00, -4.6566e-10]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0233, -0.0301, -0.0274, -0.0306, -0.0088, 0.0111, 0.0126, -0.0056, + -0.0192, -0.0155], device='cuda:0'), grad: tensor([-6.5193e-09, 4.8894e-09, 3.4925e-09, -4.3772e-08, 7.2410e-08, + 4.6566e-09, 6.9849e-09, 3.7719e-08, 5.5879e-09, -8.2655e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 220.54, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4675 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.2446, -0.3087, 0.1267, ..., -0.1574, 0.0420, 0.0420], + [-0.1507, -0.0783, -0.1144, ..., -0.2323, -0.0732, -0.0272], + [ 0.0111, -0.1902, -0.2487, ..., -0.1888, 0.0320, -0.4189], + ..., + [-0.2293, 0.1880, 0.0337, ..., 0.2404, -0.0728, -0.1690], + [-0.2102, -0.2102, 0.2271, ..., -0.1881, -0.1651, 0.1660], + [ 0.0180, -0.3764, 0.1994, ..., 0.0627, -0.2085, -0.1811]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 5.8906e-08, 4.6566e-09, ..., 7.2643e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.2340e-08, -4.6566e-10, ..., 1.1642e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.0547e-07, -7.9162e-09, ..., -1.2619e-07, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.0955e-09, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 2.3283e-10], + [-2.3283e-10, 6.9849e-10, -2.3283e-09, ..., -6.9849e-10, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0212, -0.0305, -0.0274, -0.0303, -0.0087, 0.0110, 0.0126, -0.0054, + -0.0192, -0.0162], device='cuda:0'), grad: tensor([ 5.5879e-09, 2.9383e-07, 2.4913e-08, 1.3970e-09, 1.6671e-07, + -8.9221e-07, 8.9779e-07, -5.0012e-07, 2.0023e-08, -2.0955e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 220.65, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4804 re_mapping 0.0028 re_causal 0.0084 /// teacc 99.16 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.2450, -0.3089, 0.1268, ..., -0.1574, 0.0420, 0.0424], + [-0.1511, -0.0784, -0.1144, ..., -0.2331, -0.0732, -0.0272], + [ 0.0110, -0.1927, -0.2521, ..., -0.1922, 0.0320, -0.4203], + ..., + [-0.2295, 0.1885, 0.0337, ..., 0.2423, -0.0728, -0.1691], + [-0.2100, -0.2105, 0.2308, ..., -0.1875, -0.1651, 0.1673], + [ 0.0180, -0.3774, 0.1995, ..., 0.0623, -0.2085, -0.1818]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 2.3283e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 4.1910e-09, 1.6298e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, -3.0268e-09, ..., 6.9849e-10, + 0.0000e+00, -3.2596e-09], + ..., + [ 0.0000e+00, -6.0536e-09, 2.3283e-10, ..., -5.3551e-09, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 2.3283e-10, -4.8894e-08, ..., -5.4250e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0211, -0.0306, -0.0282, -0.0303, -0.0085, 0.0110, 0.0123, -0.0049, + -0.0183, -0.0164], device='cuda:0'), grad: tensor([ 2.9802e-08, 1.1944e-07, -1.7090e-07, 1.1642e-09, 1.8859e-07, + 9.3132e-09, 2.3283e-09, -4.8894e-09, 7.6834e-09, -1.8789e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 220.42, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4754 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.19 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.2452, -0.3093, 0.1269, ..., -0.1575, 0.0420, 0.0426], + [-0.1513, -0.0792, -0.1146, ..., -0.2364, -0.0732, -0.0273], + [ 0.0110, -0.1935, -0.2533, ..., -0.1936, 0.0320, -0.4209], + ..., + [-0.2296, 0.1894, 0.0338, ..., 0.2456, -0.0728, -0.1691], + [-0.2101, -0.2111, 0.2333, ..., -0.1878, -0.1651, 0.1683], + [ 0.0180, -0.3804, 0.1997, ..., 0.0625, -0.2085, -0.1829]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.3551e-09, -1.3970e-09, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 2.3283e-10, ..., 5.1223e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.1828e-07, 0.0000e+00, ..., -7.3807e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.9419e-08, 0.0000e+00, ..., 6.2166e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0211, -0.0305, -0.0289, -0.0301, -0.0086, 0.0104, 0.0128, -0.0043, + -0.0179, -0.0164], device='cuda:0'), grad: tensor([ 1.1735e-07, 2.0931e-07, -3.0119e-06, 1.0291e-06, 3.5623e-08, + -9.6485e-07, 2.8173e-08, 2.1830e-06, 3.5577e-07, 2.3982e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 220.21, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4676 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.09 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.2454, -0.3096, 0.1269, ..., -0.1576, 0.0421, 0.0427], + [-0.1514, -0.0793, -0.1147, ..., -0.2370, -0.0732, -0.0274], + [ 0.0107, -0.1939, -0.2537, ..., -0.1947, 0.0320, -0.4210], + ..., + [-0.2297, 0.1896, 0.0339, ..., 0.2477, -0.0728, -0.1691], + [-0.2101, -0.2113, 0.2348, ..., -0.1882, -0.1651, 0.1692], + [ 0.0182, -0.3826, 0.1997, ..., 0.0620, -0.2085, -0.1838]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0212, -0.0307, -0.0288, -0.0300, -0.0087, 0.0102, 0.0129, -0.0041, + -0.0174, -0.0166], device='cuda:0'), grad: tensor([ 5.5414e-08, -3.6787e-08, 4.1979e-07, -1.0710e-08, -5.2899e-07, + 1.0477e-08, 2.9569e-08, 5.1688e-08, 9.3132e-10, 5.3551e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 220.79, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4774 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.20 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.2458, -0.3102, 0.1270, ..., -0.1578, 0.0424, 0.0429], + [-0.1520, -0.0793, -0.1148, ..., -0.2374, -0.0732, -0.0275], + [ 0.0106, -0.1942, -0.2556, ..., -0.1951, 0.0319, -0.4228], + ..., + [-0.2299, 0.1896, 0.0339, ..., 0.2481, -0.0728, -0.1693], + [-0.2102, -0.2111, 0.2368, ..., -0.1897, -0.1651, 0.1707], + [ 0.0183, -0.3832, 0.2000, ..., 0.0602, -0.2085, -0.1844]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 3.9581e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.1642e-08, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 6.0536e-09, 0.0000e+00, -1.7695e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0211, -0.0312, -0.0279, -0.0304, -0.0069, 0.0106, 0.0124, -0.0043, + -0.0166, -0.0182], device='cuda:0'), grad: tensor([ 8.8476e-09, 1.0477e-08, -4.8894e-09, 3.0734e-07, 3.2596e-09, + -3.2643e-07, 4.1910e-09, 3.4925e-09, 1.8626e-08, -1.9092e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 220.54, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4394 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.14 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.2472, -0.3102, 0.1271, ..., -0.1579, 0.0425, 0.0426], + [-0.1522, -0.0794, -0.1149, ..., -0.2379, -0.0733, -0.0275], + [ 0.0104, -0.1935, -0.2561, ..., -0.1935, 0.0320, -0.4232], + ..., + [-0.2301, 0.1897, 0.0339, ..., 0.2485, -0.0729, -0.1693], + [-0.2115, -0.2113, 0.2361, ..., -0.1918, -0.1651, 0.1707], + [ 0.0186, -0.3845, 0.2006, ..., 0.0604, -0.2085, -0.1847]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.0733e-07, 6.9849e-10, ..., 1.6298e-09, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 2.0023e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-09], + ..., + [ 0.0000e+00, 8.7777e-08, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, -3.0268e-09, ..., 2.3283e-10, + 0.0000e+00, -5.1223e-09], + [ 2.3283e-10, 4.6566e-10, -2.0955e-09, ..., -2.0955e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0212, -0.0316, -0.0263, -0.0306, -0.0069, 0.0103, 0.0129, -0.0047, + -0.0172, -0.0181], device='cuda:0'), grad: tensor([ 3.0268e-09, -9.2667e-07, 1.7532e-07, 1.2340e-08, 4.4238e-09, + 5.8208e-09, 3.0268e-09, 7.5111e-07, -9.5461e-09, -5.8208e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 220.71, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4515 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.15 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.2485, -0.3106, 0.1258, ..., -0.1580, 0.0425, 0.0395], + [-0.1521, -0.0794, -0.1150, ..., -0.2384, -0.0733, -0.0271], + [ 0.0102, -0.1939, -0.2580, ..., -0.1939, 0.0320, -0.4250], + ..., + [-0.2303, 0.1898, 0.0340, ..., 0.2491, -0.0729, -0.1694], + [-0.2120, -0.2115, 0.2401, ..., -0.1926, -0.1651, 0.1731], + [ 0.0159, -0.3848, 0.2010, ..., 0.0605, -0.2085, -0.1858]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.9849e-10, -4.8894e-09, ..., 0.0000e+00, + 0.0000e+00, -1.1642e-09], + [ 0.0000e+00, 2.3283e-10, 1.6298e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -2.5611e-09, ..., 0.0000e+00, + 0.0000e+00, -3.2596e-09], + [-4.6566e-10, 2.3283e-10, -2.7940e-09, ..., -2.3283e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0225, -0.0311, -0.0267, -0.0306, -0.0069, 0.0105, 0.0126, -0.0049, + -0.0152, -0.0184], device='cuda:0'), grad: tensor([-1.3039e-08, -1.3970e-09, 8.6147e-09, 5.3551e-09, 5.8208e-09, + 6.5193e-09, 6.9849e-09, 4.1910e-09, -1.3737e-08, -2.3283e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 221.14, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4618 re_mapping 0.0029 re_causal 0.0080 /// teacc 99.13 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.2505, -0.3110, 0.1266, ..., -0.1595, 0.0425, 0.0418], + [-0.1497, -0.0797, -0.1151, ..., -0.2394, -0.0727, -0.0262], + [ 0.0102, -0.1942, -0.2588, ..., -0.1935, 0.0313, -0.4266], + ..., + [-0.2307, 0.1902, 0.0340, ..., 0.2502, -0.0731, -0.1695], + [-0.2128, -0.2121, 0.2405, ..., -0.1926, -0.1652, 0.1731], + [ 0.0156, -0.3866, 0.2016, ..., 0.0596, -0.2085, -0.1864]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -1.1874e-08, ..., 0.0000e+00, + 0.0000e+00, -1.1409e-08], + [ 2.3283e-10, 1.3970e-09, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 2.0955e-09, ..., 0.0000e+00, + 0.0000e+00, 5.3551e-09], + ..., + [ 4.4238e-09, -4.6566e-10, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 4.6566e-09], + [ 4.6566e-10, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-08], + [ 2.0955e-09, 6.9849e-10, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0212, -0.0305, -0.0271, -0.0303, -0.0067, 0.0101, 0.0121, -0.0048, + -0.0154, -0.0187], device='cuda:0'), grad: tensor([-4.4238e-08, 1.9092e-08, 2.7940e-09, 1.8626e-09, 1.8626e-09, + 5.8208e-09, 4.1444e-08, 1.3970e-08, -4.1444e-08, 8.6147e-09], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 374---------------------------------------------------- +epoch 374, time 221.15, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4212 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.25 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.2510, -0.3111, 0.1266, ..., -0.1597, 0.0423, 0.0417], + [-0.1499, -0.0797, -0.1149, ..., -0.2410, -0.0728, -0.0262], + [ 0.0100, -0.1944, -0.2592, ..., -0.1936, 0.0313, -0.4273], + ..., + [-0.2312, 0.1903, 0.0338, ..., 0.2514, -0.0732, -0.1696], + [-0.2133, -0.2122, 0.2409, ..., -0.1927, -0.1652, 0.1735], + [ 0.0158, -0.3870, 0.2018, ..., 0.0595, -0.2085, -0.1867]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.9162e-09, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8324e-07, 0.0000e+00, 2.0955e-09, ..., 1.1642e-09, + 0.0000e+00, 5.9372e-08]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0214, -0.0299, -0.0270, -0.0304, -0.0073, 0.0101, 0.0132, -0.0055, + -0.0154, -0.0188], device='cuda:0'), grad: tensor([-2.2119e-08, 6.9849e-10, 3.0268e-09, -4.4238e-09, -3.9581e-09, + -2.5542e-07, 6.2864e-09, 3.4925e-09, 2.3283e-10, 2.7288e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 220.24, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4595 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.20 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.2519, -0.3115, 0.1266, ..., -0.1598, 0.0425, 0.0417], + [-0.1501, -0.0807, -0.1152, ..., -0.2436, -0.0728, -0.0263], + [ 0.0101, -0.1948, -0.2604, ..., -0.1941, 0.0313, -0.4285], + ..., + [-0.2313, 0.1911, 0.0339, ..., 0.2534, -0.0733, -0.1697], + [-0.2146, -0.2127, 0.2415, ..., -0.1927, -0.1652, 0.1706], + [ 0.0156, -0.3873, 0.2022, ..., 0.0559, -0.2086, -0.1873]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 7.7067e-08, ..., 0.0000e+00, + 0.0000e+00, 8.8476e-08], + [ 2.3283e-10, 0.0000e+00, -7.8231e-08, ..., 0.0000e+00, + 0.0000e+00, -8.9640e-08], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0207, -0.0307, -0.0272, -0.0303, -0.0058, 0.0125, 0.0142, -0.0051, + -0.0184, -0.0203], device='cuda:0'), grad: tensor([ 2.8405e-08, 3.2596e-09, 3.4925e-09, -2.3283e-09, 7.5903e-08, + 6.2864e-09, -1.1385e-07, 1.4040e-07, -1.3388e-07, 9.3132e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 220.44, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4587 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.20 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.2523, -0.3120, 0.1268, ..., -0.1599, 0.0429, 0.0420], + [-0.1500, -0.0810, -0.1157, ..., -0.2444, -0.0728, -0.0272], + [ 0.0099, -0.1950, -0.2621, ..., -0.1941, 0.0313, -0.4304], + ..., + [-0.2314, 0.1916, 0.0340, ..., 0.2542, -0.0734, -0.1698], + [-0.2150, -0.2129, 0.2438, ..., -0.1928, -0.1653, 0.1711], + [ 0.0156, -0.3880, 0.2021, ..., 0.0556, -0.2086, -0.1890]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, 6.9849e-10, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 2.7940e-09, ..., 2.3283e-10, + 0.0000e+00, 5.1223e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, -7.6834e-09, ..., 0.0000e+00, + 0.0000e+00, -1.3504e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0206, -0.0309, -0.0276, -0.0305, -0.0057, 0.0131, 0.0136, -0.0049, + -0.0179, -0.0205], device='cuda:0'), grad: tensor([ 4.6566e-10, 3.2596e-09, 1.3737e-08, -4.8894e-09, -2.7940e-09, + 1.9092e-08, 2.3283e-10, 1.3970e-09, -3.1432e-08, 2.3283e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 220.11, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4538 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.14 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.2535, -0.3121, 0.1268, ..., -0.1600, 0.0430, 0.0416], + [-0.1501, -0.0816, -0.1161, ..., -0.2459, -0.0728, -0.0273], + [ 0.0098, -0.1952, -0.2624, ..., -0.1944, 0.0313, -0.4306], + ..., + [-0.2315, 0.1922, 0.0344, ..., 0.2556, -0.0733, -0.1698], + [-0.2151, -0.2132, 0.2443, ..., -0.1928, -0.1653, 0.1713], + [ 0.0156, -0.3882, 0.2023, ..., 0.0556, -0.2086, -0.1895]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.3283e-09, 4.6566e-09, ..., 6.5193e-09, + 0.0000e+00, -2.0955e-09], + [ 3.4925e-09, 2.7241e-08, 7.5204e-08, ..., 8.8010e-08, + 0.0000e+00, 1.8626e-09], + [ 2.0955e-09, 2.3283e-09, 9.0804e-09, ..., 4.6566e-09, + 0.0000e+00, 2.3283e-10], + ..., + [-2.0210e-07, -1.8976e-07, -9.2480e-07, ..., -5.6066e-07, + 0.0000e+00, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, -2.5611e-09, ..., 0.0000e+00, + 0.0000e+00, -5.1223e-09], + [ 1.6298e-07, 1.3364e-07, 7.0129e-07, ..., 3.8790e-07, + 0.0000e+00, 2.0955e-09]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0209, -0.0314, -0.0274, -0.0305, -0.0057, 0.0131, 0.0136, -0.0045, + -0.0179, -0.0205], device='cuda:0'), grad: tensor([ 1.1176e-08, 2.5635e-07, 3.0268e-08, 1.8859e-08, 4.1910e-09, + 3.2550e-07, 1.4668e-08, -2.4457e-06, -1.4203e-08, 1.8040e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 220.42, cls_loss 0.0007 cls_loss_mapping 0.0009 cls_loss_causal 0.4214 re_mapping 0.0029 re_causal 0.0081 /// teacc 99.08 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.2552, -0.3124, 0.1269, ..., -0.1601, 0.0432, 0.0411], + [-0.1501, -0.0817, -0.1164, ..., -0.2463, -0.0728, -0.0271], + [ 0.0104, -0.1956, -0.2636, ..., -0.1950, 0.0313, -0.4317], + ..., + [-0.2314, 0.1925, 0.0346, ..., 0.2563, -0.0728, -0.1699], + [-0.2153, -0.2134, 0.2448, ..., -0.1929, -0.1653, 0.1714], + [ 0.0156, -0.3894, 0.2025, ..., 0.0555, -0.2086, -0.1896]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.0722e-07, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -6.3330e-08], + [ 0.0000e+00, 2.5611e-09, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 0.0000e+00, 4.0000e-07, 2.3283e-10, ..., -1.1642e-09, + -6.9849e-10, 6.2399e-08], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0211, -0.0313, -0.0274, -0.0308, -0.0057, 0.0132, 0.0135, -0.0044, + -0.0177, -0.0206], device='cuda:0'), grad: tensor([ 4.4238e-09, -2.8554e-06, 1.8859e-08, 3.5157e-08, 6.7521e-09, + -4.0513e-08, 1.5367e-08, 2.8107e-06, 1.6298e-09, 3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 220.83, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4788 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.12 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.2553, -0.3129, 0.1270, ..., -0.1603, 0.0442, 0.0414], + [-0.1516, -0.0820, -0.1193, ..., -0.2468, -0.0727, -0.0273], + [ 0.0101, -0.1968, -0.2665, ..., -0.1962, 0.0312, -0.4341], + ..., + [-0.2315, 0.1929, 0.0320, ..., 0.2539, -0.0728, -0.1700], + [-0.2158, -0.2139, 0.2458, ..., -0.1926, -0.1653, 0.1716], + [ 0.0161, -0.3900, 0.2073, ..., 0.0574, -0.2087, -0.1905]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.1910e-09, -1.8626e-09, ..., -4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0210, -0.0334, -0.0287, -0.0302, -0.0058, 0.0134, 0.0132, -0.0065, + -0.0175, -0.0175], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.7253e-09, -1.8626e-09, 2.3283e-09, 1.3970e-09, + 2.7940e-09, -2.7940e-09, -6.9849e-09, 0.0000e+00, 3.2596e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 221.17, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4408 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.12 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.2553, -0.3131, 0.1273, ..., -0.1605, 0.0461, 0.0419], + [-0.1520, -0.0820, -0.1194, ..., -0.2472, -0.0727, -0.0284], + [ 0.0101, -0.1972, -0.2678, ..., -0.1968, 0.0312, -0.4351], + ..., + [-0.2316, 0.1930, 0.0321, ..., 0.2543, -0.0730, -0.1701], + [-0.2161, -0.2137, 0.2471, ..., -0.1928, -0.1654, 0.1718], + [ 0.0160, -0.3907, 0.2074, ..., 0.0573, -0.2088, -0.1910]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 7.1712e-08, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 7.1712e-08, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0209, -0.0336, -0.0289, -0.0303, -0.0058, 0.0135, 0.0133, -0.0065, + -0.0173, -0.0176], device='cuda:0'), grad: tensor([ 4.0047e-08, 7.4506e-09, 5.1223e-09, -7.6834e-08, -1.3597e-06, + 8.9407e-08, 5.1688e-08, 1.0245e-08, 6.1793e-07, 6.1700e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 220.93, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4756 re_mapping 0.0029 re_causal 0.0081 /// teacc 99.11 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.2554, -0.3136, 0.1284, ..., -0.1607, 0.0468, 0.0473], + [-0.1521, -0.0824, -0.1195, ..., -0.2486, -0.0727, -0.0283], + [ 0.0100, -0.1976, -0.2698, ..., -0.1973, 0.0311, -0.4362], + ..., + [-0.2320, 0.1936, 0.0323, ..., 0.2551, -0.0730, -0.1704], + [-0.2163, -0.2143, 0.2473, ..., -0.1934, -0.1654, 0.1720], + [ 0.0161, -0.3930, 0.2073, ..., 0.0572, -0.2089, -0.1915]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [-7.9162e-09, 1.3970e-09, 4.6566e-10, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -3.6322e-08, 9.3132e-10, ..., -4.1910e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., 1.8626e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, -1.3504e-08, ..., -4.1910e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0163, -0.0336, -0.0295, -0.0302, -0.0059, 0.0132, 0.0121, -0.0064, + -0.0172, -0.0177], device='cuda:0'), grad: tensor([ 2.4214e-08, -3.4925e-08, 7.4506e-09, 8.3819e-09, 1.0291e-07, + 1.6764e-08, -4.4703e-08, -1.0571e-07, 3.6322e-08, -1.4435e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 220.95, cls_loss 0.0006 cls_loss_mapping 0.0008 cls_loss_causal 0.4514 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.16 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.2576, -0.3140, 0.1285, ..., -0.1609, 0.0468, 0.0472], + [-0.1522, -0.0827, -0.1195, ..., -0.2492, -0.0727, -0.0283], + [ 0.0081, -0.1992, -0.2704, ..., -0.1982, 0.0311, -0.4379], + ..., + [-0.2321, 0.1939, 0.0323, ..., 0.2554, -0.0730, -0.1704], + [-0.2166, -0.2146, 0.2474, ..., -0.1936, -0.1654, 0.1720], + [ 0.0160, -0.3931, 0.2074, ..., 0.0573, -0.2089, -0.1917]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0165, -0.0336, -0.0300, -0.0312, -0.0060, 0.0140, 0.0121, -0.0065, + -0.0172, -0.0177], device='cuda:0'), grad: tensor([ 6.0536e-09, 9.3132e-10, -2.1420e-08, 4.1910e-09, 1.0245e-08, + 2.3283e-09, -6.0536e-09, 6.5193e-09, 4.6566e-09, -4.6566e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 220.64, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4698 re_mapping 0.0031 re_causal 0.0083 /// teacc 99.06 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.2576, -0.3142, 0.1286, ..., -0.1611, 0.0468, 0.0472], + [-0.1524, -0.0830, -0.1196, ..., -0.2507, -0.0727, -0.0281], + [ 0.0082, -0.2025, -0.2730, ..., -0.2019, 0.0310, -0.4381], + ..., + [-0.2323, 0.1958, 0.0327, ..., 0.2578, -0.0730, -0.1705], + [-0.2170, -0.2149, 0.2475, ..., -0.1940, -0.1655, 0.1720], + [ 0.0159, -0.3937, 0.2075, ..., 0.0573, -0.2089, -0.1921]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 4.2841e-08, ..., 2.2817e-08, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 1.3039e-08, ..., 6.0536e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 3.5530e-07, ..., 2.7940e-09, + 0.0000e+00, 3.0641e-07], + ..., + [ 4.6566e-10, 0.0000e+00, 1.0338e-07, ..., 9.3132e-09, + 0.0000e+00, 7.6368e-08], + [ 1.3970e-09, 0.0000e+00, -4.0140e-07, ..., 4.0513e-08, + 0.0000e+00, -3.9442e-07], + [-7.9162e-09, 0.0000e+00, -2.9150e-07, ..., -5.3085e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0165, -0.0336, -0.0326, -0.0314, -0.0061, 0.0139, 0.0123, -0.0047, + -0.0173, -0.0178], device='cuda:0'), grad: tensor([ 1.7462e-07, 4.8429e-08, 1.0114e-06, 4.4843e-07, -5.9418e-07, + 9.6858e-08, 3.2596e-08, 3.1944e-07, -9.8720e-07, -5.3877e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 220.59, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4586 re_mapping 0.0028 re_causal 0.0080 /// teacc 99.12 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.2577, -0.3146, 0.1286, ..., -0.1612, 0.0467, 0.0473], + [-0.1524, -0.0849, -0.1202, ..., -0.2527, -0.0727, -0.0281], + [ 0.0082, -0.2028, -0.2738, ..., -0.2022, 0.0309, -0.4389], + ..., + [-0.2324, 0.1976, 0.0334, ..., 0.2585, -0.0729, -0.1706], + [-0.2173, -0.2153, 0.2481, ..., -0.1940, -0.1655, 0.1721], + [ 0.0159, -0.3939, 0.2076, ..., 0.0573, -0.2089, -0.1923]], + device='cuda:0'), grad: tensor([[0.0000e+00, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 8.3819e-09, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 2.6543e-08, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 2.8871e-08, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 1.1642e-08, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0165, -0.0347, -0.0328, -0.0314, -0.0060, 0.0139, 0.0123, -0.0033, + -0.0173, -0.0178], device='cuda:0'), grad: tensor([ 1.1642e-08, 3.1199e-08, 9.3132e-08, -5.2527e-07, 4.1910e-09, + 2.5379e-07, -1.8626e-09, 9.9652e-08, 3.9116e-08, 4.1910e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 220.74, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4737 re_mapping 0.0028 re_causal 0.0081 /// teacc 99.07 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.2579, -0.3151, 0.1286, ..., -0.1615, 0.0468, 0.0471], + [-0.1526, -0.0859, -0.1207, ..., -0.2540, -0.0727, -0.0283], + [ 0.0094, -0.2034, -0.2741, ..., -0.2023, 0.0309, -0.4392], + ..., + [-0.2326, 0.1986, 0.0339, ..., 0.2591, -0.0728, -0.1707], + [-0.2175, -0.2159, 0.2491, ..., -0.1942, -0.1655, 0.1723], + [ 0.0157, -0.3944, 0.2076, ..., 0.0573, -0.2089, -0.1933]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.0536e-08, 2.4680e-08, ..., 5.4948e-08, + 0.0000e+00, 0.0000e+00], + [ 2.9802e-08, 1.6764e-08, 6.0536e-09, ..., 1.4435e-08, + 0.0000e+00, 3.0268e-08], + [ 0.0000e+00, 5.0757e-08, 2.0489e-08, ..., 4.5635e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -1.8580e-07, -6.8452e-08, ..., -1.5553e-07, + 0.0000e+00, 4.6566e-10], + [ 7.4506e-09, 0.0000e+00, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 6.0536e-09], + [ 9.3132e-10, 2.0489e-08, 8.3819e-09, ..., 1.8161e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0172, -0.0354, -0.0329, -0.0311, -0.0061, 0.0138, 0.0126, -0.0025, + -0.0172, -0.0179], device='cuda:0'), grad: tensor([ 2.3143e-07, 1.3877e-07, 1.9465e-07, 7.1712e-08, 5.1223e-09, + -2.6962e-07, 1.9046e-07, -6.6357e-07, 1.5832e-08, 8.1491e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 221.28, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4724 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.18 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.2580, -0.3160, 0.1286, ..., -0.1617, 0.0468, 0.0471], + [-0.1521, -0.0862, -0.1207, ..., -0.2548, -0.0727, -0.0282], + [ 0.0097, -0.2041, -0.2744, ..., -0.2020, 0.0309, -0.4396], + ..., + [-0.2326, 0.1988, 0.0339, ..., 0.2593, -0.0728, -0.1708], + [-0.2180, -0.2163, 0.2496, ..., -0.1942, -0.1655, 0.1725], + [ 0.0157, -0.3947, 0.2077, ..., 0.0574, -0.2089, -0.1935]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 4.6566e-10, 2.5565e-07, 2.3283e-09, ..., 1.7975e-07, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, -2.7567e-07, -4.6566e-10, ..., -1.9372e-07, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, -2.3283e-08, ..., 0.0000e+00, + 0.0000e+00, -3.2131e-08], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0173, -0.0351, -0.0331, -0.0305, -0.0063, 0.0137, 0.0125, -0.0026, + -0.0171, -0.0179], device='cuda:0'), grad: tensor([ 7.4506e-09, 6.1747e-07, 1.2107e-08, 7.9628e-08, -4.1910e-09, + 1.5367e-08, 6.1933e-08, -6.4820e-07, -1.4948e-07, 1.2107e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 220.49, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4601 re_mapping 0.0028 re_causal 0.0081 /// teacc 99.18 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.2585, -0.3183, 0.1283, ..., -0.1622, 0.0468, 0.0469], + [-0.1522, -0.0866, -0.1211, ..., -0.2559, -0.0727, -0.0288], + [ 0.0095, -0.2042, -0.2718, ..., -0.2020, 0.0309, -0.4377], + ..., + [-0.2328, 0.1992, 0.0341, ..., 0.2598, -0.0729, -0.1709], + [-0.2191, -0.2186, 0.2501, ..., -0.1943, -0.1655, 0.1726], + [ 0.0158, -0.3953, 0.2078, ..., 0.0574, -0.2089, -0.1938]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.2107e-08, 0.0000e+00, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.3062e-08, 0.0000e+00, ..., 2.0955e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.4703e-08, 4.6566e-10, ..., -2.7940e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, -1.8626e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0179, -0.0361, -0.0313, -0.0274, -0.0062, 0.0109, 0.0123, -0.0024, + -0.0171, -0.0180], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.2131e-08, 8.5682e-08, 5.5879e-09, 6.9849e-09, + 6.0536e-09, -6.9849e-09, -1.1269e-07, 9.3132e-10, -6.5193e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 221.15, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4542 re_mapping 0.0030 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.2588, -0.3186, 0.1283, ..., -0.1623, 0.0468, 0.0469], + [-0.1523, -0.0866, -0.1211, ..., -0.2562, -0.0727, -0.0287], + [ 0.0094, -0.2044, -0.2727, ..., -0.2021, 0.0309, -0.4392], + ..., + [-0.2331, 0.1988, 0.0340, ..., 0.2599, -0.0729, -0.1711], + [-0.2196, -0.2190, 0.2513, ..., -0.1944, -0.1655, 0.1728], + [ 0.0138, -0.3956, 0.2079, ..., 0.0558, -0.2089, -0.1961]], + device='cuda:0'), grad: tensor([[9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 4.6566e-10], + [0.0000e+00, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 9.3132e-10], + ..., + [0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.7695e-08, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.2107e-08], + [5.1223e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 3.2596e-09]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0180, -0.0358, -0.0316, -0.0275, -0.0051, 0.0109, 0.0123, -0.0028, + -0.0170, -0.0191], device='cuda:0'), grad: tensor([ 2.7940e-09, 5.5879e-09, 5.1223e-09, 1.6019e-06, 4.6566e-10, + -1.6885e-06, 4.1910e-09, 4.1910e-09, 5.1688e-08, 1.2107e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 220.79, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4765 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.15 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.2589, -0.3190, 0.1283, ..., -0.1625, 0.0468, 0.0469], + [-0.1523, -0.0864, -0.1211, ..., -0.2566, -0.0728, -0.0286], + [ 0.0094, -0.2044, -0.2728, ..., -0.2020, 0.0310, -0.4393], + ..., + [-0.2332, 0.1988, 0.0341, ..., 0.2604, -0.0730, -0.1712], + [-0.2202, -0.2188, 0.2516, ..., -0.1946, -0.1655, 0.1728], + [ 0.0138, -0.3969, 0.2079, ..., 0.0557, -0.2090, -0.1963]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.6764e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.0955e-08], + [ 5.5879e-09, 4.6566e-10, -5.1223e-09, ..., -2.3283e-09, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0181, -0.0355, -0.0315, -0.0275, -0.0051, 0.0109, 0.0122, -0.0030, + -0.0169, -0.0191], device='cuda:0'), grad: tensor([ 3.7253e-09, 8.8476e-09, -2.5146e-08, 7.4506e-09, 4.1910e-09, + -6.6217e-07, 5.8720e-07, 1.1176e-08, 8.3819e-08, -6.9849e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 220.81, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4527 re_mapping 0.0029 re_causal 0.0080 /// teacc 99.17 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.2590, -0.3190, 0.1285, ..., -0.1625, 0.0468, 0.0469], + [-0.1535, -0.0865, -0.1211, ..., -0.2567, -0.0728, -0.0289], + [ 0.0093, -0.2045, -0.2729, ..., -0.2020, 0.0310, -0.4395], + ..., + [-0.2334, 0.1988, 0.0341, ..., 0.2604, -0.0730, -0.1714], + [-0.2212, -0.2189, 0.2521, ..., -0.1947, -0.1655, 0.1730], + [ 0.0137, -0.3972, 0.2080, ..., 0.0559, -0.2090, -0.1965]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.7742e-08, ..., 0.0000e+00, + 0.0000e+00, -3.2596e-09], + [ 0.0000e+00, 4.6566e-10, 3.2596e-09, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.3993e-08, ..., -1.8626e-09, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0181, -0.0356, -0.0312, -0.0275, -0.0053, 0.0110, 0.0121, -0.0032, + -0.0168, -0.0191], device='cuda:0'), grad: tensor([-1.7323e-07, -6.4261e-08, 7.4506e-09, 3.4459e-08, -3.7253e-09, + 1.8626e-09, 1.2107e-08, 8.0094e-08, 4.1910e-09, 1.1269e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 220.55, cls_loss 0.0007 cls_loss_mapping 0.0008 cls_loss_causal 0.4605 re_mapping 0.0028 re_causal 0.0081 /// teacc 99.13 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.2591, -0.3192, 0.1287, ..., -0.1626, 0.0468, 0.0469], + [-0.1536, -0.0858, -0.1212, ..., -0.2574, -0.0728, -0.0287], + [ 0.0094, -0.2051, -0.2738, ..., -0.2027, 0.0310, -0.4400], + ..., + [-0.2335, 0.1982, 0.0342, ..., 0.2610, -0.0730, -0.1716], + [-0.2214, -0.2189, 0.2529, ..., -0.1948, -0.1655, 0.1731], + [ 0.0137, -0.3976, 0.2080, ..., 0.0559, -0.2090, -0.1967]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 2.7940e-09, + 0.0000e+00, 4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 1.3970e-09, ..., 5.1223e-09, + 0.0000e+00, -9.3132e-09], + [-4.1910e-09, 0.0000e+00, -2.2352e-08, ..., -2.0489e-08, + 0.0000e+00, -3.7253e-09]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0182, -0.0345, -0.0314, -0.0275, -0.0053, 0.0110, 0.0121, -0.0044, + -0.0167, -0.0191], device='cuda:0'), grad: tensor([ 1.8626e-09, 4.6566e-09, -6.5193e-09, 1.0710e-08, 3.7253e-08, + 2.2352e-08, 2.7940e-09, 1.2107e-08, -4.6566e-10, -8.2422e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 220.99, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4416 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.16 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.2591, -0.3194, 0.1287, ..., -0.1627, 0.0468, 0.0468], + [-0.1547, -0.0859, -0.1213, ..., -0.2581, -0.0728, -0.0291], + [ 0.0094, -0.2055, -0.2743, ..., -0.2029, 0.0310, -0.4414], + ..., + [-0.2335, 0.1984, 0.0344, ..., 0.2615, -0.0730, -0.1717], + [-0.2217, -0.2193, 0.2538, ..., -0.1948, -0.1655, 0.1734], + [ 0.0137, -0.3978, 0.2082, ..., 0.0559, -0.2090, -0.1969]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 0.0000e+00, 8.3819e-09, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.1444e-08, 0.0000e+00, -2.8685e-07, ..., -1.1083e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0183, -0.0347, -0.0316, -0.0275, -0.0053, 0.0110, 0.0122, -0.0043, + -0.0165, -0.0191], device='cuda:0'), grad: tensor([ 1.9558e-08, 3.2596e-09, -6.9849e-09, 3.7253e-09, 6.0350e-07, + 3.4459e-08, 1.3970e-08, 4.1910e-09, 0.0000e+00, -6.7428e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 220.70, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4572 re_mapping 0.0028 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.2593, -0.3193, 0.1287, ..., -0.1628, 0.0468, 0.0468], + [-0.1547, -0.0860, -0.1213, ..., -0.2591, -0.0728, -0.0289], + [ 0.0093, -0.2061, -0.2746, ..., -0.2033, 0.0310, -0.4426], + ..., + [-0.2336, 0.1988, 0.0344, ..., 0.2622, -0.0730, -0.1719], + [-0.2221, -0.2196, 0.2538, ..., -0.1948, -0.1655, 0.1736], + [ 0.0141, -0.3988, 0.2085, ..., 0.0560, -0.2090, -0.1969]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, -7.9162e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 9.3132e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 4.6566e-10, 0.0000e+00, -1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, -1.1642e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0184, -0.0347, -0.0319, -0.0275, -0.0055, 0.0109, 0.0122, -0.0042, + -0.0163, -0.0189], device='cuda:0'), grad: tensor([-9.3132e-09, 1.1642e-08, 3.7253e-09, 3.2596e-09, 2.7008e-08, + 4.2841e-08, -4.3772e-08, 9.3132e-09, -3.3993e-08, 4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 220.47, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4652 re_mapping 0.0028 re_causal 0.0083 /// teacc 99.01 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.2595, -0.3195, 0.1285, ..., -0.1629, 0.0468, 0.0467], + [-0.1548, -0.0860, -0.1213, ..., -0.2594, -0.0728, -0.0289], + [ 0.0092, -0.2063, -0.2749, ..., -0.2034, 0.0309, -0.4431], + ..., + [-0.2337, 0.1988, 0.0344, ..., 0.2625, -0.0729, -0.1719], + [-0.2228, -0.2202, 0.2539, ..., -0.1951, -0.1658, 0.1737], + [ 0.0139, -0.3997, 0.2089, ..., 0.0561, -0.2090, -0.1976]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0188, -0.0346, -0.0320, -0.0275, -0.0057, 0.0111, 0.0116, -0.0043, + -0.0163, -0.0188], device='cuda:0'), grad: tensor([ 9.3132e-10, -2.1979e-07, 5.9605e-08, -6.5193e-09, 0.0000e+00, + 6.9849e-09, 4.6566e-10, 1.6298e-07, 0.0000e+00, 4.6566e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 220.72, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4588 re_mapping 0.0028 re_causal 0.0080 /// teacc 99.14 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.2596, -0.3209, 0.1285, ..., -0.1632, 0.0469, 0.0467], + [-0.1550, -0.0865, -0.1214, ..., -0.2602, -0.0728, -0.0287], + [ 0.0091, -0.2064, -0.2752, ..., -0.2035, 0.0310, -0.4436], + ..., + [-0.2340, 0.1995, 0.0343, ..., 0.2628, -0.0729, -0.1723], + [-0.2259, -0.2205, 0.2545, ..., -0.1952, -0.1658, 0.1736], + [ 0.0138, -0.4008, 0.2093, ..., 0.0561, -0.2090, -0.1984]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -1.8626e-09, 2.5611e-08, ..., 0.0000e+00, + 0.0000e+00, -1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-2.7940e-09, 0.0000e+00, -3.8650e-08, ..., 0.0000e+00, + 0.0000e+00, -1.3970e-09]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0189, -0.0348, -0.0319, -0.0275, -0.0057, 0.0111, 0.0116, -0.0042, + -0.0164, -0.0188], device='cuda:0'), grad: tensor([ 5.1223e-09, -2.5239e-07, 2.6543e-08, 1.0710e-08, 4.3306e-08, + 6.0536e-09, 1.8673e-07, 6.8918e-08, 2.0023e-08, -1.1316e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 220.72, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4805 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.09 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.2596, -0.3212, 0.1268, ..., -0.1653, 0.0469, 0.0468], + [-0.1551, -0.0862, -0.1213, ..., -0.2606, -0.0728, -0.0261], + [ 0.0084, -0.2065, -0.2757, ..., -0.2036, 0.0310, -0.4442], + ..., + [-0.2340, 0.1994, 0.0343, ..., 0.2632, -0.0729, -0.1750], + [-0.2262, -0.2206, 0.2546, ..., -0.1954, -0.1658, 0.1737], + [ 0.0146, -0.4014, 0.2101, ..., 0.0555, -0.2090, -0.1988]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-08], + [ 0.0000e+00, 1.3970e-09, 9.3132e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.7789e-09, 0.0000e+00, ..., -2.3283e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.2107e-08, 1.8626e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-09, 0.0000e+00, -2.0955e-08, ..., -2.0955e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0194, -0.0336, -0.0319, -0.0275, -0.0043, 0.0110, 0.0117, -0.0053, + -0.0164, -0.0196], device='cuda:0'), grad: tensor([ 1.3877e-07, 2.3283e-08, -6.4727e-08, 5.8673e-08, 6.5658e-08, + 7.8743e-07, -1.2629e-06, 8.7544e-08, 2.3097e-07, -6.7521e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 220.75, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4498 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.10 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.2598, -0.3214, 0.1269, ..., -0.1655, 0.0468, 0.0469], + [-0.1551, -0.0857, -0.1213, ..., -0.2606, -0.0728, -0.0248], + [ 0.0080, -0.2065, -0.2758, ..., -0.2036, 0.0308, -0.4455], + ..., + [-0.2342, 0.1991, 0.0342, ..., 0.2634, -0.0729, -0.1763], + [-0.2266, -0.2209, 0.2553, ..., -0.1955, -0.1663, 0.1738], + [ 0.0161, -0.4032, 0.2108, ..., 0.0572, -0.2090, -0.1996]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.3970e-09, -1.7229e-08, ..., 9.3132e-10, + 0.0000e+00, -2.3283e-08], + [ 4.6566e-10, 9.3132e-10, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 5.5879e-09, 4.6566e-10, ..., 5.1223e-09, + 0.0000e+00, 9.3132e-10], + ..., + [ 4.6566e-10, -9.3132e-09, -1.3970e-09, ..., -8.3819e-09, + 0.0000e+00, -9.3132e-10], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.8626e-09], + [ 1.3970e-09, 4.6566e-10, 1.8626e-09, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0194, -0.0329, -0.0320, -0.0275, -0.0056, 0.0111, 0.0117, -0.0060, + -0.0164, -0.0186], device='cuda:0'), grad: tensor([-6.5658e-08, 3.7253e-09, 1.5367e-08, 9.1270e-08, 2.7940e-09, + -1.0198e-07, 5.5414e-08, -1.9092e-08, 5.5879e-09, 8.8476e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 220.77, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4224 re_mapping 0.0027 re_causal 0.0077 /// teacc 99.18 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.2599, -0.3215, 0.1272, ..., -0.1655, 0.0468, 0.0471], + [-0.1551, -0.0858, -0.1213, ..., -0.2610, -0.0728, -0.0249], + [ 0.0082, -0.2065, -0.2759, ..., -0.2036, 0.0308, -0.4456], + ..., + [-0.2346, 0.1994, 0.0343, ..., 0.2640, -0.0729, -0.1763], + [-0.2269, -0.2204, 0.2557, ..., -0.1955, -0.1663, 0.1738], + [ 0.0161, -0.4048, 0.2108, ..., 0.0570, -0.2090, -0.1998]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 0.0000e+00, -3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 4.6566e-10, 0.0000e+00, 2.7940e-09, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0192, -0.0328, -0.0320, -0.0275, -0.0055, 0.0111, 0.0115, -0.0059, + -0.0164, -0.0188], device='cuda:0'), grad: tensor([ 1.8626e-09, 6.0070e-08, -1.5832e-08, 3.7253e-09, 1.3113e-06, + 6.9849e-09, -1.3448e-06, 2.7940e-09, -2.0955e-08, 9.7789e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 220.62, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4706 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.14 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.2608, -0.3219, 0.1270, ..., -0.1658, 0.0469, 0.0470], + [-0.1562, -0.0858, -0.1216, ..., -0.2614, -0.0728, -0.0249], + [ 0.0081, -0.2066, -0.2763, ..., -0.2037, 0.0308, -0.4459], + ..., + [-0.2349, 0.1994, 0.0342, ..., 0.2641, -0.0729, -0.1763], + [-0.2272, -0.2207, 0.2557, ..., -0.1958, -0.1663, 0.1739], + [ 0.0168, -0.4057, 0.2114, ..., 0.0571, -0.2090, -0.1999]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.6298e-09, 2.5611e-09, ..., 1.6298e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.1910e-09, 2.5611e-09, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, -1.2573e-08, -8.3819e-09, ..., -1.0477e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 6.7521e-09, 3.0268e-09, ..., 3.9581e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0194, -0.0332, -0.0319, -0.0275, -0.0057, 0.0111, 0.0116, -0.0061, + -0.0164, -0.0184], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.0012e-08, 3.2596e-09, 1.6298e-09, 2.5611e-09, + -6.5193e-09, 2.5611e-09, -3.0734e-08, 2.3283e-09, 1.3737e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 220.37, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4582 re_mapping 0.0028 re_causal 0.0084 /// teacc 99.19 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.2608, -0.3219, 0.1270, ..., -0.1658, 0.0469, 0.0470], + [-0.1562, -0.0857, -0.1215, ..., -0.2614, -0.0728, -0.0249], + [ 0.0081, -0.2066, -0.2763, ..., -0.2037, 0.0308, -0.4459], + ..., + [-0.2349, 0.1993, 0.0342, ..., 0.2641, -0.0729, -0.1764], + [-0.2273, -0.2207, 0.2557, ..., -0.1958, -0.1663, 0.1739], + [ 0.0167, -0.4057, 0.2114, ..., 0.0571, -0.2090, -0.1999]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -1.2340e-08, ..., 0.0000e+00, + -4.6566e-10, -6.7521e-09], + [ 0.0000e+00, 1.9791e-08, 9.3132e-09, ..., 1.3970e-08, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.1420e-08, -1.0012e-08, ..., -1.5600e-08, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.8161e-08, 2.3283e-10, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 9.5461e-09]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0194, -0.0331, -0.0319, -0.0275, -0.0056, 0.0111, 0.0116, -0.0061, + -0.0164, -0.0184], device='cuda:0'), grad: tensor([-3.7020e-08, 4.5635e-08, 2.3283e-10, 3.4925e-09, -1.3970e-09, + -3.7951e-08, 4.0745e-08, -4.7497e-08, 9.3132e-09, 3.3062e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 220.67, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4141 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.21 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.2608, -0.3219, 0.1270, ..., -0.1658, 0.0469, 0.0470], + [-0.1563, -0.0857, -0.1215, ..., -0.2615, -0.0728, -0.0249], + [ 0.0081, -0.2066, -0.2763, ..., -0.2037, 0.0308, -0.4459], + ..., + [-0.2350, 0.1994, 0.0342, ..., 0.2642, -0.0729, -0.1764], + [-0.2273, -0.2207, 0.2558, ..., -0.1958, -0.1663, 0.1739], + [ 0.0165, -0.4058, 0.2114, ..., 0.0570, -0.2090, -0.2001]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 5.1223e-09, ..., 2.7940e-09, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.6566e-09, -9.3132e-10, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0194, -0.0331, -0.0319, -0.0275, -0.0055, 0.0111, 0.0116, -0.0061, + -0.0164, -0.0185], device='cuda:0'), grad: tensor([-1.0710e-08, 8.3819e-09, 1.8626e-09, 6.0536e-09, 2.7940e-09, + -8.3819e-09, 1.4901e-08, -5.5879e-09, -9.3132e-10, -4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 220.85, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4405 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.22 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.2608, -0.3219, 0.1270, ..., -0.1658, 0.0468, 0.0470], + [-0.1564, -0.0857, -0.1216, ..., -0.2616, -0.0728, -0.0249], + [ 0.0081, -0.2066, -0.2763, ..., -0.2037, 0.0308, -0.4460], + ..., + [-0.2350, 0.1994, 0.0342, ..., 0.2642, -0.0729, -0.1764], + [-0.2276, -0.2207, 0.2558, ..., -0.1959, -0.1663, 0.1739], + [ 0.0164, -0.4058, 0.2114, ..., 0.0570, -0.2090, -0.2002]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3039e-08, 4.1910e-09, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, -2.7474e-08, -8.8476e-09, ..., -1.6298e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.1223e-09, 3.7253e-09, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0194, -0.0331, -0.0319, -0.0275, -0.0055, 0.0111, 0.0116, -0.0061, + -0.0164, -0.0185], device='cuda:0'), grad: tensor([-3.7253e-09, 1.0571e-07, 3.7253e-09, 1.6764e-08, -7.5903e-08, + 2.3283e-09, 2.7940e-09, -5.9139e-08, 4.6566e-10, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 220.28, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4493 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.21 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.2609, -0.3219, 0.1270, ..., -0.1658, 0.0468, 0.0470], + [-0.1564, -0.0857, -0.1215, ..., -0.2616, -0.0728, -0.0249], + [ 0.0081, -0.2066, -0.2763, ..., -0.2037, 0.0308, -0.4460], + ..., + [-0.2350, 0.1994, 0.0342, ..., 0.2642, -0.0729, -0.1764], + [-0.2276, -0.2207, 0.2558, ..., -0.1959, -0.1663, 0.1739], + [ 0.0163, -0.4059, 0.2114, ..., 0.0570, -0.2090, -0.2003]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, 5.1223e-09, ..., 2.0955e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0195, -0.0331, -0.0319, -0.0275, -0.0055, 0.0111, 0.0115, -0.0061, + -0.0164, -0.0185], device='cuda:0'), grad: tensor([-8.8476e-09, 7.9162e-09, -4.6566e-10, -1.6764e-08, -1.3784e-07, + 8.3819e-09, 4.1910e-09, 6.0536e-09, 1.8626e-09, 1.3877e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 220.43, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4240 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.19 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.2609, -0.3220, 0.1270, ..., -0.1658, 0.0468, 0.0470], + [-0.1565, -0.0858, -0.1216, ..., -0.2617, -0.0728, -0.0249], + [ 0.0080, -0.2066, -0.2764, ..., -0.2037, 0.0308, -0.4461], + ..., + [-0.2350, 0.1994, 0.0342, ..., 0.2643, -0.0729, -0.1764], + [-0.2278, -0.2208, 0.2558, ..., -0.1959, -0.1664, 0.1738], + [ 0.0162, -0.4059, 0.2114, ..., 0.0570, -0.2090, -0.2005]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, -4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.3970e-09, 0.0000e+00, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, -6.9849e-09], + [-9.3132e-10, 9.3132e-10, -4.1910e-09, ..., 1.2573e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0055, 0.0111, 0.0115, -0.0061, + -0.0164, -0.0185], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.0710e-08, 1.5367e-08, 9.3132e-09, -7.6368e-08, + -3.0501e-07, 3.2037e-07, 1.6298e-08, -2.7008e-08, 5.9605e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 220.71, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4448 re_mapping 0.0024 re_causal 0.0078 /// teacc 99.21 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.2609, -0.3220, 0.1270, ..., -0.1658, 0.0468, 0.0470], + [-0.1565, -0.0858, -0.1216, ..., -0.2618, -0.0728, -0.0249], + [ 0.0080, -0.2067, -0.2764, ..., -0.2037, 0.0308, -0.4461], + ..., + [-0.2350, 0.1994, 0.0342, ..., 0.2643, -0.0729, -0.1764], + [-0.2278, -0.2208, 0.2559, ..., -0.1959, -0.1664, 0.1739], + [ 0.0161, -0.4060, 0.2114, ..., 0.0570, -0.2090, -0.2006]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 3.2596e-09, 1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -4.6566e-09, ..., -4.6566e-10, + 0.0000e+00, -3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0055, 0.0111, 0.0115, -0.0061, + -0.0164, -0.0186], device='cuda:0'), grad: tensor([ 9.3132e-10, 4.6566e-10, 3.7253e-09, -7.4506e-09, -5.2620e-08, + -1.8626e-08, 2.8871e-08, 6.5193e-09, -1.5832e-08, 4.6566e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 220.57, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4385 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.20 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.2609, -0.3220, 0.1270, ..., -0.1658, 0.0468, 0.0470], + [-0.1566, -0.0858, -0.1216, ..., -0.2619, -0.0728, -0.0249], + [ 0.0080, -0.2067, -0.2764, ..., -0.2037, 0.0308, -0.4461], + ..., + [-0.2351, 0.1995, 0.0342, ..., 0.2644, -0.0729, -0.1764], + [-0.2278, -0.2208, 0.2559, ..., -0.1959, -0.1664, 0.1739], + [ 0.0161, -0.4061, 0.2114, ..., 0.0569, -0.2090, -0.2006]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + -4.6566e-10, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3970e-09, 4.6566e-10, 1.8626e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, -5.1223e-09], + [-3.7253e-09, 0.0000e+00, -4.6566e-09, ..., -1.3970e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0055, 0.0111, 0.0115, -0.0061, + -0.0164, -0.0186], device='cuda:0'), grad: tensor([-1.4435e-08, 1.3970e-09, -3.7253e-09, 3.2596e-09, -3.2596e-09, + 1.6764e-08, 1.0710e-08, 1.3039e-08, -1.2573e-08, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 220.66, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4437 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.20 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.2609, -0.3220, 0.1270, ..., -0.1659, 0.0468, 0.0470], + [-0.1566, -0.0858, -0.1216, ..., -0.2619, -0.0728, -0.0250], + [ 0.0080, -0.2067, -0.2764, ..., -0.2038, 0.0308, -0.4461], + ..., + [-0.2351, 0.1995, 0.0342, ..., 0.2644, -0.0729, -0.1764], + [-0.2279, -0.2208, 0.2559, ..., -0.1959, -0.1664, 0.1739], + [ 0.0160, -0.4061, 0.2115, ..., 0.0569, -0.2090, -0.2007]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.2107e-08, 4.6566e-09, ..., 1.0245e-08, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, -1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, -1.4901e-08, -6.0536e-09, ..., -1.3504e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 1.8626e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0115, -0.0061, + -0.0164, -0.0186], device='cuda:0'), grad: tensor([ 1.3970e-09, 3.7253e-08, -1.3504e-08, 1.3970e-08, -1.8626e-09, + -9.7789e-09, 1.8626e-09, -3.3993e-08, 9.3132e-10, 1.4435e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 220.53, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4458 re_mapping 0.0023 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.2609, -0.3220, 0.1270, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0858, -0.1216, ..., -0.2620, -0.0728, -0.0250], + [ 0.0080, -0.2067, -0.2765, ..., -0.2038, 0.0308, -0.4462], + ..., + [-0.2351, 0.1995, 0.0342, ..., 0.2645, -0.0729, -0.1764], + [-0.2280, -0.2208, 0.2559, ..., -0.1960, -0.1664, 0.1739], + [ 0.0160, -0.4062, 0.2115, ..., 0.0569, -0.2090, -0.2007]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 9.3132e-10, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 4.6566e-10, 1.3970e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, -4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, -4.6566e-10], + [ 9.3132e-10, 4.6566e-10, -6.1467e-08, ..., -7.4971e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0164, -0.0186], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.1502e-07, -1.4575e-07, 6.9849e-09, 2.7847e-07, + 3.4459e-08, -4.9826e-08, 2.3749e-08, 2.7008e-08, -2.9523e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 220.40, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4504 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.19 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.2610, -0.3220, 0.1270, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0858, -0.1216, ..., -0.2621, -0.0728, -0.0250], + [ 0.0079, -0.2067, -0.2765, ..., -0.2038, 0.0308, -0.4462], + ..., + [-0.2351, 0.1996, 0.0342, ..., 0.2646, -0.0729, -0.1764], + [-0.2281, -0.2208, 0.2559, ..., -0.1960, -0.1664, 0.1738], + [ 0.0160, -0.4062, 0.2115, ..., 0.0569, -0.2090, -0.2007]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 0.0000e+00, -1.3970e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0165, -0.0186], device='cuda:0'), grad: tensor([-1.4901e-08, -2.0489e-08, -6.9849e-09, -1.4435e-08, 9.3132e-10, + 1.8161e-08, 1.1642e-08, 2.7474e-08, 2.3283e-09, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 220.34, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4415 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.18 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.2610, -0.3220, 0.1271, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0859, -0.1216, ..., -0.2622, -0.0728, -0.0250], + [ 0.0079, -0.2068, -0.2765, ..., -0.2038, 0.0308, -0.4463], + ..., + [-0.2351, 0.1996, 0.0342, ..., 0.2646, -0.0729, -0.1764], + [-0.2281, -0.2208, 0.2559, ..., -0.1960, -0.1664, 0.1739], + [ 0.0160, -0.4062, 0.2115, ..., 0.0569, -0.2090, -0.2008]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, -3.1665e-08, 2.3283e-09, ..., -3.6275e-07, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, -4.6566e-10, -5.1223e-09, ..., -1.8626e-09, + 0.0000e+00, -3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0195, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0165, -0.0186], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.2596e-09, 6.0536e-09, 2.7940e-09, 9.9093e-07, + 1.1642e-08, 6.9849e-09, -1.0002e-06, -1.3039e-08, 5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 220.51, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4385 re_mapping 0.0023 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.2610, -0.3221, 0.1271, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0858, -0.1216, ..., -0.2622, -0.0728, -0.0250], + [ 0.0079, -0.2068, -0.2765, ..., -0.2039, 0.0308, -0.4463], + ..., + [-0.2351, 0.1996, 0.0342, ..., 0.2647, -0.0729, -0.1764], + [-0.2281, -0.2208, 0.2560, ..., -0.1960, -0.1664, 0.1739], + [ 0.0159, -0.4063, 0.2115, ..., 0.0569, -0.2090, -0.2009]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.7940e-09, -2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 6.0536e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0194, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 9.3132e-09, 3.7253e-09, 2.7940e-09, 4.6566e-10, -6.7055e-08, + -2.5053e-07, 2.4540e-07, -1.8626e-09, 1.1176e-08, 5.1688e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 220.16, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4277 re_mapping 0.0023 re_causal 0.0075 /// teacc 99.19 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.2610, -0.3221, 0.1271, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0859, -0.1216, ..., -0.2624, -0.0728, -0.0250], + [ 0.0079, -0.2068, -0.2766, ..., -0.2039, 0.0308, -0.4464], + ..., + [-0.2352, 0.1997, 0.0342, ..., 0.2648, -0.0729, -0.1764], + [-0.2282, -0.2208, 0.2560, ..., -0.1960, -0.1664, 0.1739], + [ 0.0159, -0.4063, 0.2116, ..., 0.0569, -0.2090, -0.2009]], + device='cuda:0'), grad: tensor([[1.3970e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, + 4.6566e-10], + [0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 2.3283e-09, 1.8626e-09, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0194, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 8.3819e-09, 4.1910e-09, 2.3283e-09, -5.3551e-08, 1.3970e-09, + 2.8871e-08, 9.3132e-10, 0.0000e+00, 1.3504e-08, 1.3970e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 220.50, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4228 re_mapping 0.0022 re_causal 0.0073 /// teacc 99.19 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.2610, -0.3221, 0.1271, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0859, -0.1216, ..., -0.2625, -0.0728, -0.0250], + [ 0.0078, -0.2068, -0.2766, ..., -0.2039, 0.0307, -0.4465], + ..., + [-0.2352, 0.1997, 0.0342, ..., 0.2649, -0.0729, -0.1764], + [-0.2282, -0.2208, 0.2560, ..., -0.1961, -0.1665, 0.1739], + [ 0.0159, -0.4063, 0.2116, ..., 0.0569, -0.2090, -0.2009]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-4.6566e-10, 0.0000e+00, -2.7940e-09, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0194, -0.0332, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 4.1910e-09, 1.2899e-07, -2.1793e-07, 8.3819e-09, 8.3819e-08, + -1.1642e-08, -7.9162e-09, 3.2596e-09, 1.3970e-08, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 220.49, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4249 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.2610, -0.3221, 0.1271, ..., -0.1659, 0.0468, 0.0470], + [-0.1567, -0.0860, -0.1217, ..., -0.2626, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2766, ..., -0.2039, 0.0307, -0.4465], + ..., + [-0.2352, 0.1998, 0.0342, ..., 0.2649, -0.0729, -0.1764], + [-0.2283, -0.2208, 0.2560, ..., -0.1961, -0.1665, 0.1739], + [ 0.0159, -0.4064, 0.2116, ..., 0.0569, -0.2090, -0.2010]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-09, -3.7253e-09, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3516e-07, 4.2375e-08, ..., 3.8184e-08, + 0.0000e+00, -1.7229e-08], + [ 0.0000e+00, 2.7940e-09, 4.1910e-09, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.8219e-07, -5.2620e-08, ..., -4.5635e-08, + 0.0000e+00, 1.6298e-08], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 2.7940e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0194, -0.0333, -0.0319, -0.0275, -0.0054, 0.0111, 0.0114, -0.0061, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([-1.0710e-08, 4.0978e-07, 2.5611e-08, 2.5146e-08, 8.8476e-09, + 3.3993e-08, 4.6566e-09, -5.1083e-07, 4.1910e-09, 7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 220.76, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4488 re_mapping 0.0022 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.2611, -0.3221, 0.1272, ..., -0.1659, 0.0468, 0.0470], + [-0.1568, -0.0860, -0.1217, ..., -0.2626, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2766, ..., -0.2039, 0.0307, -0.4465], + ..., + [-0.2352, 0.1998, 0.0342, ..., 0.2649, -0.0729, -0.1764], + [-0.2284, -0.2208, 0.2560, ..., -0.1961, -0.1665, 0.1739], + [ 0.0158, -0.4064, 0.2116, ..., 0.0569, -0.2090, -0.2011]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 9.3132e-10, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -4.1910e-09, -1.3970e-09, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0194, -0.0333, -0.0319, -0.0275, -0.0054, 0.0111, 0.0113, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 1.3970e-09, -2.0815e-07, 1.0990e-07, 1.3970e-09, 1.1176e-08, + 1.8626e-09, 2.7940e-09, 7.4040e-08, 4.6566e-09, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 220.01, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4483 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.2611, -0.3221, 0.1272, ..., -0.1659, 0.0468, 0.0470], + [-0.1569, -0.0860, -0.1217, ..., -0.2627, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2766, ..., -0.2039, 0.0307, -0.4465], + ..., + [-0.2352, 0.1998, 0.0342, ..., 0.2650, -0.0729, -0.1764], + [-0.2285, -0.2209, 0.2561, ..., -0.1961, -0.1665, 0.1739], + [ 0.0158, -0.4065, 0.2116, ..., 0.0569, -0.2090, -0.2011]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 9.3132e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 6.9849e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 9.7789e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0194, -0.0333, -0.0318, -0.0275, -0.0054, 0.0111, 0.0113, -0.0061, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([-2.7474e-08, 3.8650e-08, 7.9162e-09, 3.4506e-07, 1.4435e-08, + -5.4715e-07, 1.0664e-07, 1.7695e-08, 1.3039e-08, 3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 220.58, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4427 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.17 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.2611, -0.3222, 0.1272, ..., -0.1659, 0.0468, 0.0470], + [-0.1570, -0.0860, -0.1217, ..., -0.2628, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2767, ..., -0.2039, 0.0307, -0.4466], + ..., + [-0.2353, 0.1999, 0.0343, ..., 0.2651, -0.0729, -0.1764], + [-0.2286, -0.2209, 0.2561, ..., -0.1961, -0.1665, 0.1738], + [ 0.0157, -0.4067, 0.2116, ..., 0.0569, -0.2090, -0.2012]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, -1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0194, -0.0333, -0.0319, -0.0275, -0.0054, 0.0111, 0.0112, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 4.6566e-10, 6.5193e-09, -4.6566e-09, 9.3132e-10, 2.3283e-09, + 1.8626e-09, 4.6566e-09, 0.0000e+00, -3.2596e-09, 9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 220.49, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4269 re_mapping 0.0022 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.2611, -0.3222, 0.1272, ..., -0.1660, 0.0468, 0.0470], + [-0.1570, -0.0861, -0.1217, ..., -0.2629, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2767, ..., -0.2040, 0.0307, -0.4466], + ..., + [-0.2353, 0.2000, 0.0343, ..., 0.2652, -0.0729, -0.1764], + [-0.2286, -0.2209, 0.2561, ..., -0.1962, -0.1665, 0.1739], + [ 0.0157, -0.4068, 0.2117, ..., 0.0569, -0.2090, -0.2014]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0194, -0.0333, -0.0318, -0.0275, -0.0054, 0.0111, 0.0111, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 3.2596e-09, 6.0536e-09, -4.3306e-08, 2.7940e-08, -4.7963e-08, + -8.3819e-09, 2.7940e-09, 9.7789e-09, 5.5879e-09, 3.6322e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 221.21, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.3942 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.17 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.2611, -0.3222, 0.1272, ..., -0.1660, 0.0468, 0.0470], + [-0.1570, -0.0861, -0.1217, ..., -0.2629, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2767, ..., -0.2040, 0.0307, -0.4467], + ..., + [-0.2353, 0.2000, 0.0343, ..., 0.2652, -0.0729, -0.1764], + [-0.2287, -0.2209, 0.2561, ..., -0.1962, -0.1665, 0.1739], + [ 0.0156, -0.4069, 0.2117, ..., 0.0569, -0.2090, -0.2015]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 4.6566e-10, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -5.1223e-09, -1.3970e-09, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0194, -0.0333, -0.0318, -0.0275, -0.0054, 0.0111, 0.0111, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 4.6566e-10, 2.2817e-08, 1.3970e-09, 9.3132e-10, -2.2817e-08, + 3.2596e-09, 1.3970e-09, -1.8626e-08, 9.3132e-10, 1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 221.22, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4547 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.17 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.2612, -0.3222, 0.1271, ..., -0.1661, 0.0468, 0.0470], + [-0.1570, -0.0861, -0.1217, ..., -0.2630, -0.0728, -0.0250], + [ 0.0078, -0.2069, -0.2767, ..., -0.2040, 0.0307, -0.4468], + ..., + [-0.2353, 0.2000, 0.0343, ..., 0.2653, -0.0729, -0.1764], + [-0.2287, -0.2209, 0.2561, ..., -0.1962, -0.1665, 0.1739], + [ 0.0156, -0.4069, 0.2118, ..., 0.0569, -0.2090, -0.2015]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.3283e-09, 0.0000e+00, -7.4506e-09, ..., -6.5193e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0195, -0.0333, -0.0318, -0.0275, -0.0053, 0.0112, 0.0111, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 8.8476e-09, 9.3132e-09, -5.1223e-09, 1.4435e-08, -5.6345e-08, + -9.3132e-10, 2.6077e-08, 4.6566e-09, 4.6566e-10, 2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 220.41, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4403 re_mapping 0.0022 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.2612, -0.3223, 0.1271, ..., -0.1661, 0.0467, 0.0470], + [-0.1570, -0.0861, -0.1218, ..., -0.2631, -0.0728, -0.0250], + [ 0.0077, -0.2070, -0.2768, ..., -0.2040, 0.0307, -0.4469], + ..., + [-0.2354, 0.2001, 0.0343, ..., 0.2653, -0.0729, -0.1764], + [-0.2288, -0.2209, 0.2562, ..., -0.1962, -0.1665, 0.1739], + [ 0.0155, -0.4069, 0.2118, ..., 0.0569, -0.2090, -0.2016]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -6.9849e-09, -2.5611e-09, ..., -7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [-1.1642e-09, 6.5193e-09, -1.3970e-09, ..., 3.4925e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0195, -0.0333, -0.0319, -0.0275, -0.0054, 0.0112, 0.0111, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 1.3970e-09, 4.4238e-09, -6.7521e-09, 6.9849e-09, 7.9162e-09, + 6.9849e-09, -6.7521e-09, -1.2573e-08, 5.1223e-09, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 220.23, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4176 re_mapping 0.0021 re_causal 0.0071 /// teacc 99.14 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.2613, -0.3223, 0.1271, ..., -0.1661, 0.0467, 0.0470], + [-0.1571, -0.0862, -0.1218, ..., -0.2632, -0.0728, -0.0250], + [ 0.0077, -0.2069, -0.2768, ..., -0.2040, 0.0307, -0.4469], + ..., + [-0.2354, 0.2001, 0.0343, ..., 0.2654, -0.0729, -0.1764], + [-0.2288, -0.2209, 0.2562, ..., -0.1962, -0.1665, 0.1739], + [ 0.0153, -0.4070, 0.2118, ..., 0.0569, -0.2090, -0.2018]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, -1.8626e-09, ..., 2.3283e-10, + -6.9849e-10, -1.3970e-09], + [ 2.0955e-09, 7.6368e-08, 6.4727e-08, ..., 3.0966e-08, + 0.0000e+00, 5.1223e-09], + [ 2.3283e-10, 3.4925e-09, 3.0268e-09, ..., 1.1642e-09, + 0.0000e+00, 6.9849e-10], + ..., + [ 4.6566e-10, -7.6136e-08, -6.4727e-08, ..., -2.9337e-08, + 0.0000e+00, 9.3132e-10], + [ 1.1642e-09, -1.1642e-08, -1.5600e-08, ..., 1.3970e-09, + 0.0000e+00, -3.0035e-08], + [ 1.8626e-09, 4.6566e-10, 2.3283e-10, ..., 2.0955e-09, + 2.3283e-10, 1.3970e-09]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0195, -0.0334, -0.0318, -0.0275, -0.0054, 0.0112, 0.0111, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([-4.6566e-09, 2.3772e-07, 1.3039e-08, 1.3039e-08, -1.3295e-07, + 6.0303e-08, 8.6613e-08, -2.0000e-07, -8.7079e-08, 2.0023e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 220.89, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4180 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.15 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.2613, -0.3223, 0.1271, ..., -0.1661, 0.0467, 0.0470], + [-0.1571, -0.0862, -0.1218, ..., -0.2632, -0.0728, -0.0251], + [ 0.0077, -0.2070, -0.2768, ..., -0.2040, 0.0307, -0.4469], + ..., + [-0.2354, 0.2002, 0.0343, ..., 0.2654, -0.0729, -0.1764], + [-0.2288, -0.2209, 0.2562, ..., -0.1962, -0.1665, 0.1739], + [ 0.0153, -0.4071, 0.2119, ..., 0.0569, -0.2090, -0.2018]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.7521e-09, 2.3283e-09, ..., 4.8894e-09, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 2.0955e-08, 1.3970e-09, ..., 1.3970e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.7707e-08, -3.4925e-09, ..., -2.0256e-08, + 0.0000e+00, -2.3283e-10], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0195, -0.0334, -0.0318, -0.0275, -0.0054, 0.0112, 0.0111, -0.0060, + -0.0165, -0.0187], device='cuda:0'), grad: tensor([ 6.9849e-10, 1.5832e-08, 4.9360e-08, -6.9849e-10, 2.3283e-09, + -8.8476e-09, 1.0012e-08, -6.2631e-08, 2.0955e-09, 2.3283e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 220.38, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4032 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.14 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.2614, -0.3223, 0.1272, ..., -0.1661, 0.0467, 0.0470], + [-0.1571, -0.0862, -0.1218, ..., -0.2633, -0.0728, -0.0251], + [ 0.0076, -0.2070, -0.2769, ..., -0.2040, 0.0307, -0.4470], + ..., + [-0.2355, 0.2002, 0.0343, ..., 0.2655, -0.0729, -0.1764], + [-0.2289, -0.2209, 0.2563, ..., -0.1963, -0.1665, 0.1739], + [ 0.0152, -0.4071, 0.2119, ..., 0.0569, -0.2090, -0.2019]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 1.3970e-09, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, -2.6310e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 2.4680e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.3283e-10, 1.1642e-09, ..., 6.9849e-10, + 0.0000e+00, 6.9849e-10], + [-6.9849e-09, 0.0000e+00, -9.5461e-09, ..., -8.6147e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0195, -0.0334, -0.0318, -0.0275, -0.0054, 0.0112, 0.0111, -0.0060, + -0.0166, -0.0187], device='cuda:0'), grad: tensor([ 3.4925e-09, 9.5461e-09, -1.2456e-07, 2.6543e-08, 2.7241e-08, + -1.4901e-08, -9.3132e-10, 1.1781e-07, 9.0804e-09, -4.4005e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 220.71, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4473 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.2614, -0.3223, 0.1272, ..., -0.1662, 0.0467, 0.0470], + [-0.1571, -0.0863, -0.1218, ..., -0.2633, -0.0728, -0.0251], + [ 0.0076, -0.2070, -0.2769, ..., -0.2040, 0.0306, -0.4471], + ..., + [-0.2355, 0.2002, 0.0343, ..., 0.2655, -0.0729, -0.1764], + [-0.2290, -0.2209, 0.2563, ..., -0.1963, -0.1666, 0.1739], + [ 0.0151, -0.4072, 0.2120, ..., 0.0570, -0.2090, -0.2020]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 9.3132e-10], + [ 1.1642e-09, 2.3283e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., -1.6298e-09, + 0.0000e+00, -1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-10], + [-4.6566e-10, 0.0000e+00, -9.3132e-10, ..., -2.3283e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0195, -0.0334, -0.0318, -0.0275, -0.0054, 0.0112, 0.0110, -0.0060, + -0.0166, -0.0187], device='cuda:0'), grad: tensor([ 2.5146e-08, 1.6531e-08, 7.9162e-09, 2.0023e-08, -1.8859e-08, + -3.9348e-08, -1.4435e-08, 1.5367e-08, 1.8626e-09, -1.6298e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 220.27, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.3990 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.2615, -0.3224, 0.1271, ..., -0.1663, 0.0467, 0.0470], + [-0.1571, -0.0863, -0.1218, ..., -0.2634, -0.0728, -0.0250], + [ 0.0075, -0.2070, -0.2769, ..., -0.2040, 0.0306, -0.4471], + ..., + [-0.2355, 0.2003, 0.0344, ..., 0.2656, -0.0729, -0.1765], + [-0.2290, -0.2210, 0.2563, ..., -0.1963, -0.1666, 0.1739], + [ 0.0151, -0.4072, 0.2120, ..., 0.0570, -0.2090, -0.2021]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0196, -0.0334, -0.0318, -0.0275, -0.0054, 0.0112, 0.0110, -0.0059, + -0.0166, -0.0187], device='cuda:0'), grad: tensor([-3.0268e-09, 1.3970e-09, 1.8626e-09, 2.5611e-09, 1.3970e-09, + 1.6298e-09, 1.6298e-09, 1.6298e-09, 1.1642e-09, 5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 220.36, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4038 re_mapping 0.0021 re_causal 0.0071 /// teacc 99.18 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.2616, -0.3224, 0.1269, ..., -0.1665, 0.0467, 0.0470], + [-0.1571, -0.0863, -0.1218, ..., -0.2635, -0.0728, -0.0250], + [ 0.0075, -0.2070, -0.2770, ..., -0.2040, 0.0306, -0.4472], + ..., + [-0.2356, 0.2004, 0.0344, ..., 0.2657, -0.0729, -0.1765], + [-0.2290, -0.2210, 0.2563, ..., -0.1963, -0.1666, 0.1739], + [ 0.0151, -0.4073, 0.2122, ..., 0.0570, -0.2090, -0.2021]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.3970e-09, 2.3283e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-09, 3.0268e-09, 6.9849e-10, ..., 3.2596e-09, + 0.0000e+00, -2.3283e-10], + [-1.0710e-08, 1.8626e-09, 4.6566e-10, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, -9.7789e-09, -9.3132e-10, ..., -9.3132e-09, + 0.0000e+00, 0.0000e+00], + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, -6.9849e-10, ..., 1.6298e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0196, -0.0334, -0.0318, -0.0275, -0.0054, 0.0112, 0.0110, -0.0059, + -0.0166, -0.0187], device='cuda:0'), grad: tensor([ 6.2864e-09, 1.0943e-08, -4.2375e-08, 2.7940e-09, 1.0943e-08, + 2.3283e-09, 3.7253e-09, -1.9558e-08, 2.3749e-08, 4.4238e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 220.39, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4428 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.17 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.2616, -0.3224, 0.1269, ..., -0.1665, 0.0467, 0.0470], + [-0.1572, -0.0864, -0.1219, ..., -0.2635, -0.0728, -0.0251], + [ 0.0075, -0.2070, -0.2770, ..., -0.2041, 0.0306, -0.4472], + ..., + [-0.2356, 0.2004, 0.0344, ..., 0.2658, -0.0729, -0.1765], + [-0.2291, -0.2210, 0.2563, ..., -0.1963, -0.1666, 0.1739], + [ 0.0151, -0.4074, 0.2122, ..., 0.0570, -0.2090, -0.2022]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, -1.3970e-09, ..., 2.3283e-10, + 0.0000e+00, -2.3283e-10], + [ 2.3283e-10, -1.8626e-09, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, 6.9849e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, -4.6566e-10, 2.3283e-10, ..., -2.0955e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3970e-09, 2.3283e-10, -3.2596e-09, ..., -1.6298e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0196, -0.0335, -0.0318, -0.0275, -0.0055, 0.0112, 0.0110, -0.0059, + -0.0166, -0.0187], device='cuda:0'), grad: tensor([-2.0955e-09, -7.9162e-09, 4.4238e-09, 1.3970e-09, 1.0245e-08, + 1.3970e-09, 4.6566e-10, 3.9581e-09, 2.3283e-10, -9.7789e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 219.93, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4230 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.16 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.2616, -0.3224, 0.1269, ..., -0.1665, 0.0467, 0.0470], + [-0.1573, -0.0864, -0.1219, ..., -0.2636, -0.0728, -0.0251], + [ 0.0075, -0.2071, -0.2770, ..., -0.2041, 0.0306, -0.4472], + ..., + [-0.2356, 0.2004, 0.0344, ..., 0.2658, -0.0729, -0.1765], + [-0.2292, -0.2210, 0.2563, ..., -0.1964, -0.1666, 0.1739], + [ 0.0151, -0.4074, 0.2122, ..., 0.0570, -0.2090, -0.2022]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0012e-08, ..., 0.0000e+00, + 0.0000e+00, -6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 4.4238e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1642e-09, 0.0000e+00, -1.1642e-09, ..., 0.0000e+00, + 0.0000e+00, -3.0268e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0196, -0.0335, -0.0318, -0.0275, -0.0054, 0.0112, 0.0110, -0.0059, + -0.0166, -0.0187], device='cuda:0'), grad: tensor([-1.0431e-07, 5.1223e-09, 4.9826e-08, 4.1910e-09, 1.2806e-08, + 2.0955e-09, 2.2817e-08, 3.2596e-09, -6.7521e-09, 5.5879e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 220.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4234 re_mapping 0.0021 re_causal 0.0071 /// teacc 99.16 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.2616, -0.3224, 0.1269, ..., -0.1665, 0.0467, 0.0471], + [-0.1573, -0.0864, -0.1219, ..., -0.2636, -0.0728, -0.0251], + [ 0.0075, -0.2071, -0.2771, ..., -0.2041, 0.0306, -0.4472], + ..., + [-0.2356, 0.2004, 0.0344, ..., 0.2658, -0.0729, -0.1765], + [-0.2293, -0.2210, 0.2564, ..., -0.1964, -0.1666, 0.1739], + [ 0.0151, -0.4074, 0.2123, ..., 0.0571, -0.2090, -0.2023]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0196, -0.0335, -0.0318, -0.0275, -0.0054, 0.0112, 0.0109, -0.0060, + -0.0166, -0.0186], device='cuda:0'), grad: tensor([ 5.8208e-09, -3.9814e-08, 3.0268e-09, -1.9325e-08, -3.0268e-09, + 2.2585e-08, -7.4506e-09, 3.4925e-09, 2.6776e-08, 1.1874e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 220.41, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4287 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.17 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.2616, -0.3224, 0.1269, ..., -0.1666, 0.0467, 0.0471], + [-0.1573, -0.0864, -0.1219, ..., -0.2637, -0.0728, -0.0251], + [ 0.0076, -0.2071, -0.2771, ..., -0.2041, 0.0306, -0.4473], + ..., + [-0.2356, 0.2005, 0.0344, ..., 0.2659, -0.0729, -0.1765], + [-0.2294, -0.2210, 0.2564, ..., -0.1964, -0.1666, 0.1739], + [ 0.0151, -0.4076, 0.2123, ..., 0.0571, -0.2090, -0.2023]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 7.2177e-08, 6.4727e-08, ..., 2.8173e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.1910e-09, 3.2596e-09, ..., 2.0955e-09, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, -7.8231e-08, -4.4936e-08, ..., -1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 6.9849e-10, 3.0268e-09, ..., 2.0955e-09, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 4.6566e-10, -6.1467e-08, ..., -4.7497e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0196, -0.0335, -0.0318, -0.0275, -0.0054, 0.0112, 0.0109, -0.0059, + -0.0167, -0.0187], device='cuda:0'), grad: tensor([ 3.4925e-09, 2.0117e-07, 1.1642e-08, 3.1898e-08, 8.8708e-08, + -2.1653e-08, 2.8173e-08, -1.3178e-07, 1.0245e-08, -2.1514e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 220.36, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4437 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.2616, -0.3224, 0.1269, ..., -0.1666, 0.0467, 0.0471], + [-0.1574, -0.0865, -0.1219, ..., -0.2638, -0.0728, -0.0251], + [ 0.0076, -0.2071, -0.2771, ..., -0.2041, 0.0306, -0.4473], + ..., + [-0.2356, 0.2005, 0.0344, ..., 0.2659, -0.0729, -0.1765], + [-0.2294, -0.2210, 0.2564, ..., -0.1964, -0.1666, 0.1739], + [ 0.0150, -0.4076, 0.2124, ..., 0.0571, -0.2090, -0.2024]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5367e-08, ..., 0.0000e+00, + 0.0000e+00, -3.4925e-09], + [ 0.0000e+00, 0.0000e+00, 2.0955e-09, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, 6.7521e-09, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 6.9849e-10, ..., 1.6298e-09, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0196, -0.0335, -0.0317, -0.0275, -0.0054, 0.0112, 0.0109, -0.0060, + -0.0167, -0.0186], device='cuda:0'), grad: tensor([-3.9814e-08, 9.5461e-09, 1.4435e-08, 8.8476e-09, -4.4238e-09, + 1.4203e-08, -1.7695e-08, 3.0268e-09, 8.8476e-09, 9.0804e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 220.22, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4170 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.17 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.2617, -0.3224, 0.1270, ..., -0.1666, 0.0467, 0.0471], + [-0.1574, -0.0865, -0.1219, ..., -0.2639, -0.0728, -0.0251], + [ 0.0076, -0.2071, -0.2772, ..., -0.2041, 0.0306, -0.4473], + ..., + [-0.2356, 0.2005, 0.0344, ..., 0.2660, -0.0729, -0.1765], + [-0.2295, -0.2211, 0.2564, ..., -0.1964, -0.1666, 0.1739], + [ 0.0150, -0.4076, 0.2124, ..., 0.0571, -0.2090, -0.2025]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 3.2596e-09, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, -5.5879e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.8626e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.4238e-09, + 0.0000e+00, 4.6566e-10], + [ 2.0955e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.4226e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0196, -0.0336, -0.0316, -0.0275, -0.0055, 0.0112, 0.0109, -0.0060, + -0.0167, -0.0187], device='cuda:0'), grad: tensor([ 3.7253e-09, -8.3121e-08, 2.3516e-08, 1.8161e-08, -1.8044e-07, + -3.3062e-08, 8.6147e-09, 2.4680e-08, 6.4494e-08, 1.5832e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 220.74, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4147 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.2617, -0.3225, 0.1270, ..., -0.1666, 0.0467, 0.0471], + [-0.1575, -0.0866, -0.1220, ..., -0.2640, -0.0728, -0.0251], + [ 0.0076, -0.2071, -0.2772, ..., -0.2041, 0.0306, -0.4473], + ..., + [-0.2357, 0.2006, 0.0345, ..., 0.2661, -0.0729, -0.1765], + [-0.2296, -0.2211, 0.2564, ..., -0.1964, -0.1666, 0.1739], + [ 0.0150, -0.4077, 0.2124, ..., 0.0570, -0.2090, -0.2025]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 1.6298e-09, 1.1642e-09, ..., 2.0955e-09, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.3970e-09, -6.9849e-10, ..., -1.6298e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0196, -0.0336, -0.0316, -0.0275, -0.0055, 0.0112, 0.0108, -0.0060, + -0.0167, -0.0187], device='cuda:0'), grad: tensor([ 2.3283e-09, 6.2864e-09, -1.0477e-08, 6.5193e-09, -1.1642e-09, + -7.5437e-08, 7.4739e-08, -3.2596e-09, 3.0268e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 220.66, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.2617, -0.3225, 0.1270, ..., -0.1667, 0.0467, 0.0471], + [-0.1575, -0.0866, -0.1220, ..., -0.2641, -0.0728, -0.0252], + [ 0.0076, -0.2071, -0.2772, ..., -0.2041, 0.0306, -0.4473], + ..., + [-0.2357, 0.2006, 0.0345, ..., 0.2661, -0.0729, -0.1765], + [-0.2297, -0.2211, 0.2565, ..., -0.1964, -0.1666, 0.1739], + [ 0.0150, -0.4078, 0.2124, ..., 0.0570, -0.2090, -0.2026]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 3.9581e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -3.9581e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0196, -0.0336, -0.0316, -0.0275, -0.0055, 0.0112, 0.0108, -0.0060, + -0.0167, -0.0187], device='cuda:0'), grad: tensor([ 0.0000e+00, -3.7253e-09, 1.5134e-08, 4.6566e-10, 6.9849e-10, + 2.7940e-09, 1.1642e-09, 3.4925e-09, -1.4901e-08, 6.9849e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 220.66, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4060 re_mapping 0.0020 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.2617, -0.3225, 0.1270, ..., -0.1667, 0.0467, 0.0471], + [-0.1575, -0.0868, -0.1221, ..., -0.2644, -0.0728, -0.0251], + [ 0.0077, -0.2071, -0.2773, ..., -0.2041, 0.0306, -0.4474], + ..., + [-0.2357, 0.2008, 0.0346, ..., 0.2663, -0.0729, -0.1765], + [-0.2297, -0.2211, 0.2566, ..., -0.1965, -0.1666, 0.1739], + [ 0.0150, -0.4080, 0.2125, ..., 0.0570, -0.2090, -0.2027]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, -1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -2.0955e-09], + [ 0.0000e+00, -2.3283e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 2.3283e-10, 0.0000e+00, -2.7940e-09, ..., 2.3283e-10, + 0.0000e+00, -7.6834e-09], + [ 2.3283e-10, 4.6566e-10, -1.6298e-09, ..., -9.3132e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0196, -0.0337, -0.0315, -0.0275, -0.0054, 0.0112, 0.0108, -0.0059, + -0.0168, -0.0187], device='cuda:0'), grad: tensor([ 4.8894e-09, -1.7928e-08, -3.5157e-08, 1.3039e-08, 3.7253e-09, + -2.3283e-10, 3.1898e-08, 3.7486e-08, -3.0268e-08, 1.3970e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 220.97, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4003 re_mapping 0.0020 re_causal 0.0069 /// teacc 99.18 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.2618, -0.3225, 0.1270, ..., -0.1667, 0.0467, 0.0471], + [-0.1576, -0.0869, -0.1221, ..., -0.2645, -0.0728, -0.0251], + [ 0.0076, -0.2071, -0.2773, ..., -0.2041, 0.0306, -0.4475], + ..., + [-0.2357, 0.2009, 0.0346, ..., 0.2664, -0.0729, -0.1765], + [-0.2298, -0.2212, 0.2566, ..., -0.1965, -0.1666, 0.1739], + [ 0.0150, -0.4080, 0.2125, ..., 0.0571, -0.2090, -0.2027]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, -3.2596e-09, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 2.3283e-10, -3.9581e-09, 2.7940e-09, ..., -9.5461e-09, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.0710e-08, 2.3283e-09, -1.8859e-08, ..., -2.2119e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0196, -0.0337, -0.0315, -0.0275, -0.0055, 0.0112, 0.0107, -0.0059, + -0.0168, -0.0187], device='cuda:0'), grad: tensor([ 1.6298e-09, -5.1921e-08, 3.0268e-09, 3.7253e-09, 7.3807e-08, + 8.1491e-09, 4.6566e-10, 4.9593e-08, 1.3970e-09, -8.0094e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 220.34, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4363 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.18 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.2619, -0.3225, 0.1271, ..., -0.1667, 0.0467, 0.0471], + [-0.1576, -0.0869, -0.1221, ..., -0.2646, -0.0728, -0.0251], + [ 0.0074, -0.2072, -0.2773, ..., -0.2041, 0.0306, -0.4477], + ..., + [-0.2357, 0.2009, 0.0346, ..., 0.2665, -0.0729, -0.1765], + [-0.2299, -0.2212, 0.2566, ..., -0.1965, -0.1666, 0.1738], + [ 0.0150, -0.4081, 0.2125, ..., 0.0571, -0.2090, -0.2027]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, -2.5611e-09], + [ 0.0000e+00, -2.6310e-08, 0.0000e+00, ..., -1.8161e-08, + 0.0000e+00, 2.3283e-10], + ..., + [ 6.9849e-10, 2.5844e-08, 1.1642e-09, ..., 1.8161e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 1.1642e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.0268e-09, 0.0000e+00, -5.1223e-09, ..., -1.6298e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0196, -0.0337, -0.0315, -0.0275, -0.0055, 0.0113, 0.0107, -0.0059, + -0.0168, -0.0187], device='cuda:0'), grad: tensor([ 2.0955e-09, -1.0245e-08, -1.7462e-07, 6.9849e-10, 2.5611e-09, + 5.8208e-09, 1.6997e-08, 1.7462e-07, 4.1910e-09, -1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 220.84, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4151 re_mapping 0.0020 re_causal 0.0069 /// teacc 99.19 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.2619, -0.3226, 0.1271, ..., -0.1667, 0.0467, 0.0471], + [-0.1576, -0.0869, -0.1221, ..., -0.2646, -0.0728, -0.0251], + [ 0.0074, -0.2072, -0.2774, ..., -0.2042, 0.0306, -0.4478], + ..., + [-0.2358, 0.2010, 0.0346, ..., 0.2666, -0.0729, -0.1765], + [-0.2300, -0.2212, 0.2566, ..., -0.1965, -0.1666, 0.1738], + [ 0.0150, -0.4082, 0.2126, ..., 0.0571, -0.2090, -0.2028]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 4.6566e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 9.3132e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 6.9849e-10, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 2.0955e-09], + [ 6.9849e-10, 4.6566e-10, -2.3283e-09, ..., -1.1642e-09, + 0.0000e+00, -2.5611e-09], + [ 1.3970e-09, 1.1642e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 1.1642e-09]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0196, -0.0337, -0.0314, -0.0276, -0.0055, 0.0113, 0.0106, -0.0059, + -0.0169, -0.0187], device='cuda:0'), grad: tensor([ 2.5611e-09, 3.9581e-09, 3.2596e-09, -1.2107e-08, 2.5611e-09, + -2.3283e-10, -1.1176e-08, 9.3132e-09, -3.9581e-09, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 220.74, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4120 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.16 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.2619, -0.3226, 0.1271, ..., -0.1667, 0.0467, 0.0471], + [-0.1577, -0.0870, -0.1222, ..., -0.2648, -0.0728, -0.0251], + [ 0.0074, -0.2072, -0.2774, ..., -0.2042, 0.0306, -0.4478], + ..., + [-0.2358, 0.2011, 0.0346, ..., 0.2667, -0.0729, -0.1765], + [-0.2301, -0.2212, 0.2566, ..., -0.1965, -0.1666, 0.1738], + [ 0.0150, -0.4082, 0.2126, ..., 0.0571, -0.2090, -0.2028]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-1.3970e-09, 0.0000e+00, -1.8626e-09, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0196, -0.0337, -0.0314, -0.0276, -0.0056, 0.0113, 0.0106, -0.0059, + -0.0169, -0.0186], device='cuda:0'), grad: tensor([ 6.9849e-10, 1.3970e-09, -3.4925e-09, 3.7951e-08, 6.7521e-09, + -4.7265e-08, 1.1176e-08, 2.3283e-09, 2.3283e-09, -6.7521e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 220.58, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4422 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.18 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.2619, -0.3226, 0.1272, ..., -0.1667, 0.0467, 0.0471], + [-0.1578, -0.0871, -0.1222, ..., -0.2649, -0.0728, -0.0252], + [ 0.0074, -0.2072, -0.2774, ..., -0.2042, 0.0306, -0.4478], + ..., + [-0.2358, 0.2011, 0.0347, ..., 0.2667, -0.0729, -0.1766], + [-0.2302, -0.2213, 0.2566, ..., -0.1966, -0.1666, 0.1738], + [ 0.0150, -0.4082, 0.2127, ..., 0.0572, -0.2090, -0.2028]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 1.1642e-10, 4.6566e-10, ..., 1.1642e-10, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + ..., + [ 0.0000e+00, 1.1642e-10, 5.8208e-10, ..., 2.3283e-10, + 0.0000e+00, 1.0477e-09], + [ 0.0000e+00, 1.1642e-10, -2.5611e-09, ..., 0.0000e+00, + 0.0000e+00, -4.0745e-09], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1525e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0196, -0.0338, -0.0313, -0.0276, -0.0056, 0.0113, 0.0106, -0.0059, + -0.0170, -0.0186], device='cuda:0'), grad: tensor([ 5.8208e-10, 3.4925e-09, -1.4086e-08, -4.3074e-09, -3.2480e-08, + 2.1886e-08, 3.2596e-09, 5.1223e-09, -1.4552e-08, 3.3411e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 220.64, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4424 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.17 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.2620, -0.3226, 0.1271, ..., -0.1668, 0.0467, 0.0471], + [-0.1578, -0.0872, -0.1223, ..., -0.2650, -0.0728, -0.0252], + [ 0.0074, -0.2072, -0.2775, ..., -0.2042, 0.0306, -0.4478], + ..., + [-0.2358, 0.2012, 0.0347, ..., 0.2668, -0.0729, -0.1766], + [-0.2303, -0.2213, 0.2565, ..., -0.1966, -0.1666, 0.1738], + [ 0.0151, -0.4083, 0.2128, ..., 0.0572, -0.2090, -0.2028]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 8.1491e-10, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 2.9104e-09, ..., 1.7462e-09, + 0.0000e+00, 9.3132e-10], + [ 3.9581e-09, 0.0000e+00, -4.5402e-09, ..., -2.6776e-09, + 0.0000e+00, 1.6298e-09]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0197, -0.0338, -0.0313, -0.0276, -0.0056, 0.0113, 0.0106, -0.0059, + -0.0170, -0.0185], device='cuda:0'), grad: tensor([ 2.2119e-09, 5.8208e-10, 1.1642e-10, 1.7462e-09, 4.4238e-09, + -1.0245e-08, 1.9791e-09, 4.6566e-10, 7.5670e-09, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 220.45, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4196 re_mapping 0.0020 re_causal 0.0070 /// teacc 99.22 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.2620, -0.3227, 0.1271, ..., -0.1668, 0.0467, 0.0471], + [-0.1578, -0.0873, -0.1223, ..., -0.2652, -0.0728, -0.0252], + [ 0.0074, -0.2072, -0.2775, ..., -0.2042, 0.0306, -0.4479], + ..., + [-0.2358, 0.2013, 0.0348, ..., 0.2669, -0.0729, -0.1766], + [-0.2303, -0.2213, 0.2566, ..., -0.1967, -0.1666, 0.1738], + [ 0.0151, -0.4083, 0.2129, ..., 0.0573, -0.2090, -0.2029]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 1.1642e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.1642e-10, -3.4925e-10, ..., 0.0000e+00, + 0.0000e+00, -8.1491e-10], + [ 9.3132e-10, 2.3283e-10, -2.7940e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0197, -0.0339, -0.0312, -0.0276, -0.0057, 0.0113, 0.0106, -0.0059, + -0.0170, -0.0185], device='cuda:0'), grad: tensor([ 1.5134e-09, 3.4925e-09, 3.0268e-09, -5.0059e-09, -2.5495e-08, + 6.8685e-09, 4.6566e-09, 2.2119e-09, -1.3970e-09, 9.4296e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 220.32, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4099 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.19 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.2620, -0.3227, 0.1272, ..., -0.1668, 0.0467, 0.0471], + [-0.1578, -0.0873, -0.1224, ..., -0.2652, -0.0728, -0.0252], + [ 0.0074, -0.2072, -0.2775, ..., -0.2042, 0.0305, -0.4479], + ..., + [-0.2359, 0.2013, 0.0348, ..., 0.2670, -0.0729, -0.1766], + [-0.2304, -0.2213, 0.2566, ..., -0.1967, -0.1666, 0.1738], + [ 0.0151, -0.4084, 0.2129, ..., 0.0573, -0.2090, -0.2030]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.1491e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.1642e-09, 0.0000e+00, ..., -1.7462e-09, + 0.0000e+00, 0.0000e+00], + [ 8.1491e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 3.4925e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0197, -0.0339, -0.0312, -0.0276, -0.0057, 0.0113, 0.0106, -0.0059, + -0.0170, -0.0185], device='cuda:0'), grad: tensor([ 1.2806e-09, 1.8626e-09, 2.2119e-09, 3.6089e-09, -4.7730e-09, + -2.5611e-09, -1.5134e-09, -2.7940e-09, 2.0955e-09, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 220.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4225 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.21 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.2620, -0.3227, 0.1272, ..., -0.1668, 0.0467, 0.0472], + [-0.1579, -0.0873, -0.1224, ..., -0.2653, -0.0728, -0.0253], + [ 0.0074, -0.2073, -0.2776, ..., -0.2042, 0.0305, -0.4480], + ..., + [-0.2359, 0.2014, 0.0348, ..., 0.2670, -0.0729, -0.1766], + [-0.2305, -0.2213, 0.2566, ..., -0.1967, -0.1666, 0.1738], + [ 0.0151, -0.4084, 0.2129, ..., 0.0573, -0.2090, -0.2030]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-10, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 3.4925e-10, ..., 4.6566e-10, + 0.0000e+00, 1.1642e-10], + [-6.9849e-10, 0.0000e+00, -1.6298e-09, ..., -2.2119e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0196, -0.0339, -0.0312, -0.0276, -0.0056, 0.0113, 0.0105, -0.0059, + -0.0171, -0.0185], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.5134e-09, 1.0477e-09, 2.0955e-09, 8.0327e-09, + -8.1491e-10, -1.6298e-09, 2.9104e-09, 2.3283e-09, -6.7521e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 220.14, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4263 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.18 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.2621, -0.3227, 0.1273, ..., -0.1668, 0.0467, 0.0472], + [-0.1579, -0.0873, -0.1224, ..., -0.2654, -0.0728, -0.0252], + [ 0.0074, -0.2073, -0.2776, ..., -0.2042, 0.0305, -0.4481], + ..., + [-0.2359, 0.2014, 0.0348, ..., 0.2670, -0.0729, -0.1766], + [-0.2306, -0.2213, 0.2567, ..., -0.1967, -0.1666, 0.1738], + [ 0.0151, -0.4085, 0.2130, ..., 0.0573, -0.2090, -0.2031]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 0.0000e+00, 3.4925e-10, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + ..., + [ 5.8208e-10, 1.1642e-10, 1.8626e-09, ..., 8.1491e-10, + 0.0000e+00, 1.1642e-10], + [ 3.9581e-09, 1.1642e-10, 1.5134e-09, ..., 1.1642e-09, + 0.0000e+00, 2.2119e-09], + [ 1.1409e-08, 0.0000e+00, -5.0059e-09, ..., -3.0268e-09, + 0.0000e+00, 8.6147e-09]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0196, -0.0339, -0.0311, -0.0276, -0.0057, 0.0113, 0.0105, -0.0059, + -0.0171, -0.0185], device='cuda:0'), grad: tensor([ 1.3970e-09, 2.9104e-09, 4.6566e-10, 8.0559e-08, 5.8208e-09, + -1.1746e-07, 9.0804e-09, 4.6566e-09, 1.1292e-08, 1.5250e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 220.69, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3927 re_mapping 0.0020 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.2621, -0.3226, 0.1273, ..., -0.1668, 0.0467, 0.0472], + [-0.1580, -0.0874, -0.1224, ..., -0.2655, -0.0728, -0.0252], + [ 0.0074, -0.2073, -0.2776, ..., -0.2042, 0.0305, -0.4481], + ..., + [-0.2359, 0.2015, 0.0348, ..., 0.2671, -0.0729, -0.1766], + [-0.2306, -0.2214, 0.2567, ..., -0.1967, -0.1666, 0.1738], + [ 0.0151, -0.4085, 0.2131, ..., 0.0573, -0.2090, -0.2031]], + device='cuda:0'), grad: tensor([[1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [1.1642e-10, 1.1642e-10, 1.1642e-10, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [2.3283e-10, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [3.4925e-09, 1.1642e-10, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, + 1.9791e-09], + [4.0745e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 2.3283e-09]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0196, -0.0340, -0.0311, -0.0276, -0.0057, 0.0113, 0.0105, -0.0059, + -0.0171, -0.0185], device='cuda:0'), grad: tensor([ 6.0536e-09, 1.9791e-09, 1.1642e-09, 1.5832e-08, -2.7940e-09, + -2.9569e-08, -3.2596e-09, 1.3970e-09, 8.7311e-09, 9.4296e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 220.50, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4387 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.2621, -0.3227, 0.1273, ..., -0.1668, 0.0467, 0.0472], + [-0.1580, -0.0875, -0.1225, ..., -0.2655, -0.0728, -0.0252], + [ 0.0073, -0.2074, -0.2777, ..., -0.2043, 0.0305, -0.4482], + ..., + [-0.2360, 0.2016, 0.0348, ..., 0.2672, -0.0729, -0.1766], + [-0.2307, -0.2214, 0.2567, ..., -0.1967, -0.1666, 0.1738], + [ 0.0151, -0.4085, 0.2131, ..., 0.0573, -0.2090, -0.2032]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0196, -0.0340, -0.0311, -0.0276, -0.0057, 0.0113, 0.0105, -0.0059, + -0.0171, -0.0185], device='cuda:0'), grad: tensor([ 2.3283e-10, 1.2806e-09, -1.1642e-09, -7.4506e-09, 1.1642e-10, + 2.4447e-09, 1.1642e-10, 1.8626e-09, 1.1642e-09, 8.1491e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 220.58, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4178 re_mapping 0.0020 re_causal 0.0070 /// teacc 99.19 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.2621, -0.3227, 0.1274, ..., -0.1668, 0.0467, 0.0472], + [-0.1580, -0.0875, -0.1225, ..., -0.2656, -0.0728, -0.0252], + [ 0.0073, -0.2074, -0.2777, ..., -0.2043, 0.0305, -0.4482], + ..., + [-0.2360, 0.2016, 0.0349, ..., 0.2673, -0.0729, -0.1766], + [-0.2307, -0.2214, 0.2567, ..., -0.1968, -0.1666, 0.1738], + [ 0.0152, -0.4086, 0.2132, ..., 0.0573, -0.2090, -0.2032]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00], + [1.7462e-09, 0.0000e+00, 0.0000e+00, ..., 6.4028e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, 0.0000e+00, + 0.0000e+00], + [2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [2.5611e-09, 0.0000e+00, 0.0000e+00, ..., 9.4296e-09, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0195, -0.0341, -0.0311, -0.0276, -0.0056, 0.0114, 0.0105, -0.0059, + -0.0171, -0.0184], device='cuda:0'), grad: tensor([ 9.1968e-09, 4.5053e-08, -1.1059e-08, 1.2806e-09, -1.1432e-07, + 1.2806e-09, 4.0745e-09, 6.7521e-09, 4.6566e-10, 6.6822e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 220.50, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4171 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.21 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.2621, -0.3227, 0.1274, ..., -0.1669, 0.0467, 0.0473], + [-0.1580, -0.0875, -0.1225, ..., -0.2657, -0.0728, -0.0252], + [ 0.0073, -0.2074, -0.2778, ..., -0.2043, 0.0305, -0.4482], + ..., + [-0.2360, 0.2016, 0.0349, ..., 0.2673, -0.0729, -0.1767], + [-0.2308, -0.2214, 0.2567, ..., -0.1968, -0.1666, 0.1739], + [ 0.0152, -0.4086, 0.2133, ..., 0.0574, -0.2090, -0.2033]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, -5.2387e-09, 1.1642e-10, ..., 3.4925e-10, + 0.0000e+00, -1.9791e-09], + [ 0.0000e+00, 1.1642e-10, 3.4925e-10, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 1.1642e-10, 1.1642e-09, -2.6776e-09, ..., -3.8417e-09, + 0.0000e+00, 1.3970e-09], + [ 1.0012e-08, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 9.3132e-10, 3.9581e-09, 3.1432e-09, ..., 4.6566e-09, + 0.0000e+00, 8.1491e-10]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0195, -0.0340, -0.0310, -0.0276, -0.0056, 0.0113, 0.0105, -0.0060, + -0.0172, -0.0184], device='cuda:0'), grad: tensor([ 1.0477e-09, -2.8638e-08, 1.5134e-09, 5.9372e-09, 3.7253e-09, + -2.6193e-08, 4.7730e-09, 1.5716e-08, 1.6298e-08, 1.2922e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 220.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4111 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.2621, -0.3227, 0.1275, ..., -0.1669, 0.0467, 0.0473], + [-0.1581, -0.0876, -0.1226, ..., -0.2658, -0.0728, -0.0252], + [ 0.0073, -0.2075, -0.2778, ..., -0.2044, 0.0305, -0.4483], + ..., + [-0.2360, 0.2017, 0.0349, ..., 0.2674, -0.0729, -0.1767], + [-0.2309, -0.2214, 0.2567, ..., -0.1968, -0.1666, 0.1738], + [ 0.0152, -0.4087, 0.2134, ..., 0.0574, -0.2090, -0.2033]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.2119e-09, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 3.4925e-10], + [ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, -1.1642e-09, ..., 1.1642e-10, + 0.0000e+00, -2.6776e-09], + [ 1.1642e-10, 0.0000e+00, 5.8208e-10, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-09]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0195, -0.0341, -0.0310, -0.0276, -0.0056, 0.0114, 0.0105, -0.0060, + -0.0172, -0.0184], device='cuda:0'), grad: tensor([ 2.5728e-08, 3.7253e-09, -1.3737e-08, 5.9372e-09, -2.4447e-09, + 1.6764e-08, -4.5518e-08, 4.1910e-09, -2.9104e-09, 1.4319e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 220.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4163 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.18 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.2621, -0.3227, 0.1275, ..., -0.1669, 0.0467, 0.0473], + [-0.1581, -0.0876, -0.1226, ..., -0.2659, -0.0728, -0.0252], + [ 0.0073, -0.2075, -0.2779, ..., -0.2044, 0.0305, -0.4483], + ..., + [-0.2360, 0.2018, 0.0349, ..., 0.2675, -0.0729, -0.1767], + [-0.2309, -0.2214, 0.2567, ..., -0.1968, -0.1666, 0.1738], + [ 0.0152, -0.4087, 0.2134, ..., 0.0574, -0.2090, -0.2034]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 8.1491e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0195, -0.0341, -0.0309, -0.0276, -0.0057, 0.0114, 0.0104, -0.0060, + -0.0173, -0.0184], device='cuda:0'), grad: tensor([ 1.5134e-09, -6.5193e-09, 1.1642e-09, 2.3283e-10, -1.0477e-09, + 1.9791e-09, 7.3342e-09, 5.8208e-10, 6.9849e-10, 3.9581e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 220.47, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4152 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.2621, -0.3227, 0.1276, ..., -0.1669, 0.0467, 0.0473], + [-0.1581, -0.0877, -0.1226, ..., -0.2660, -0.0728, -0.0252], + [ 0.0073, -0.2075, -0.2779, ..., -0.2044, 0.0305, -0.4483], + ..., + [-0.2361, 0.2018, 0.0349, ..., 0.2676, -0.0729, -0.1767], + [-0.2310, -0.2214, 0.2567, ..., -0.1968, -0.1666, 0.1739], + [ 0.0153, -0.4088, 0.2135, ..., 0.0575, -0.2090, -0.2034]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1642e-10, -1.8626e-09, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.4447e-09, 2.5611e-09, ..., 3.3760e-09, + 0.0000e+00, 1.8626e-09], + [ 1.1642e-10, 5.8208e-10, 4.1910e-09, ..., 5.8208e-10, + 0.0000e+00, 6.8685e-09], + ..., + [ 6.9849e-10, -5.1223e-09, 5.8208e-10, ..., -1.9791e-09, + 0.0000e+00, 3.4925e-10], + [ 6.4028e-09, 0.0000e+00, -5.8208e-09, ..., 2.3283e-10, + 0.0000e+00, -7.9162e-09], + [-2.0955e-09, 2.3283e-10, -8.7311e-09, ..., -1.3737e-08, + 0.0000e+00, 6.9849e-10]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0194, -0.0341, -0.0310, -0.0276, -0.0057, 0.0114, 0.0104, -0.0060, + -0.0173, -0.0183], device='cuda:0'), grad: tensor([-4.3772e-08, 4.3889e-08, 3.9116e-08, 3.4692e-08, 1.6764e-08, + -4.1095e-08, 1.9209e-08, -5.0059e-09, -2.0722e-08, -3.3877e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 220.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4294 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.2622, -0.3227, 0.1276, ..., -0.1669, 0.0467, 0.0473], + [-0.1581, -0.0877, -0.1227, ..., -0.2661, -0.0727, -0.0253], + [ 0.0073, -0.2075, -0.2780, ..., -0.2044, 0.0305, -0.4484], + ..., + [-0.2361, 0.2018, 0.0350, ..., 0.2676, -0.0729, -0.1767], + [-0.2311, -0.2215, 0.2568, ..., -0.1969, -0.1666, 0.1738], + [ 0.0153, -0.4089, 0.2136, ..., 0.0575, -0.2090, -0.2034]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 5.8208e-10], + [-5.8208e-10, 8.1491e-10, 3.4925e-10, ..., 6.9849e-10, + 0.0000e+00, -5.5879e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + ..., + [ 0.0000e+00, 1.1642e-10, 6.9849e-10, ..., 8.1491e-10, + 0.0000e+00, 4.6566e-10], + [ 3.4925e-10, 0.0000e+00, 1.1642e-10, ..., 1.1642e-10, + 0.0000e+00, 3.2596e-09], + [ 1.1642e-10, 8.1491e-10, -1.1874e-08, ..., -1.4435e-08, + 0.0000e+00, 1.2806e-09]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0194, -0.0341, -0.0309, -0.0276, -0.0057, 0.0114, 0.0104, -0.0061, + -0.0173, -0.0183], device='cuda:0'), grad: tensor([ 3.7253e-09, -1.9441e-08, 1.6298e-09, -1.1642e-10, 4.8778e-08, + 5.2387e-09, 9.3132e-10, 6.9849e-09, 1.3504e-08, -4.7032e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 220.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4543 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.19 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.2622, -0.3227, 0.1277, ..., -0.1669, 0.0467, 0.0473], + [-0.1582, -0.0878, -0.1228, ..., -0.2662, -0.0727, -0.0253], + [ 0.0073, -0.2075, -0.2780, ..., -0.2044, 0.0305, -0.4485], + ..., + [-0.2361, 0.2019, 0.0350, ..., 0.2678, -0.0729, -0.1767], + [-0.2311, -0.2215, 0.2568, ..., -0.1969, -0.1666, 0.1738], + [ 0.0153, -0.4090, 0.2136, ..., 0.0574, -0.2090, -0.2035]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, 3.4925e-10, ..., 3.1432e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 2.0955e-09, ..., 2.3283e-10, + 0.0000e+00, 1.5134e-09], + ..., + [ 0.0000e+00, 1.1642e-10, 5.8208e-10, ..., 1.6298e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 8.1491e-10, -3.1432e-09, ..., 0.0000e+00, + 0.0000e+00, -2.2119e-09], + [ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 8.1491e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0194, -0.0343, -0.0308, -0.0276, -0.0057, 0.0114, 0.0106, -0.0061, + -0.0174, -0.0184], device='cuda:0'), grad: tensor([ 3.4925e-10, 1.4086e-08, 6.1700e-09, -2.9104e-09, -3.7020e-08, + 2.2119e-09, 1.5716e-08, 8.4983e-09, -4.3074e-09, 3.8417e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 220.33, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4151 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.2622, -0.3227, 0.1277, ..., -0.1669, 0.0467, 0.0474], + [-0.1582, -0.0879, -0.1228, ..., -0.2662, -0.0727, -0.0253], + [ 0.0073, -0.2076, -0.2781, ..., -0.2045, 0.0305, -0.4485], + ..., + [-0.2361, 0.2020, 0.0350, ..., 0.2678, -0.0729, -0.1767], + [-0.2312, -0.2215, 0.2568, ..., -0.1969, -0.1666, 0.1739], + [ 0.0153, -0.4090, 0.2137, ..., 0.0574, -0.2090, -0.2035]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 5.8208e-10, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 2.3283e-10, 3.2596e-09, 5.8208e-10, ..., 1.0477e-09, + 0.0000e+00, 0.0000e+00], + [-2.3283e-10, 2.4447e-09, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, -1.1642e-10], + ..., + [ 1.1642e-10, 2.2375e-07, 0.0000e+00, ..., 3.4808e-08, + 0.0000e+00, 1.1642e-10], + [ 5.8208e-10, 1.1642e-10, 4.6566e-10, ..., 6.9849e-10, + 0.0000e+00, 1.1642e-10], + [-6.0536e-09, 6.6357e-09, -5.8208e-09, ..., -5.5879e-09, + 0.0000e+00, -1.6298e-09]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0194, -0.0343, -0.0307, -0.0276, -0.0057, 0.0113, 0.0106, -0.0061, + -0.0174, -0.0183], device='cuda:0'), grad: tensor([ 3.4925e-09, 9.6625e-09, -5.2387e-09, -4.5076e-07, 1.4668e-08, + 2.2119e-08, 1.3970e-09, 4.1630e-07, 4.5402e-09, -1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 220.56, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4032 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.2622, -0.3227, 0.1278, ..., -0.1670, 0.0467, 0.0474], + [-0.1582, -0.0879, -0.1228, ..., -0.2664, -0.0727, -0.0253], + [ 0.0073, -0.2076, -0.2781, ..., -0.2045, 0.0305, -0.4485], + ..., + [-0.2362, 0.2020, 0.0350, ..., 0.2679, -0.0729, -0.1768], + [-0.2312, -0.2215, 0.2569, ..., -0.1969, -0.1666, 0.1739], + [ 0.0154, -0.4091, 0.2138, ..., 0.0575, -0.2090, -0.2035]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.4925e-10, -7.6834e-09, ..., 1.1642e-10, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 3.4925e-10], + [ 0.0000e+00, 1.1642e-10, 3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, 2.6776e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 1.1642e-10, 0.0000e+00, -5.8208e-10, ..., 1.1642e-10, + 0.0000e+00, -3.9581e-09], + [ 0.0000e+00, 0.0000e+00, -5.8208e-10, ..., -6.9849e-10, + 0.0000e+00, 8.1491e-10]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0193, -0.0343, -0.0306, -0.0276, -0.0057, 0.0113, 0.0106, -0.0062, + -0.0174, -0.0183], device='cuda:0'), grad: tensor([-1.9441e-08, 2.2119e-09, 1.2806e-08, 7.7998e-09, 2.5611e-09, + -5.8208e-10, 2.3283e-09, 3.1432e-09, -5.7044e-09, -2.3283e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 220.77, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3960 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.2622, -0.3228, 0.1279, ..., -0.1670, 0.0467, 0.0474], + [-0.1582, -0.0879, -0.1228, ..., -0.2664, -0.0727, -0.0253], + [ 0.0073, -0.2076, -0.2781, ..., -0.2045, 0.0305, -0.4486], + ..., + [-0.2362, 0.2020, 0.0350, ..., 0.2679, -0.0729, -0.1768], + [-0.2312, -0.2215, 0.2569, ..., -0.1969, -0.1666, 0.1739], + [ 0.0153, -0.4092, 0.2138, ..., 0.0575, -0.2090, -0.2036]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.0477e-09, ..., 3.4925e-10, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -3.2596e-09, 2.3283e-10, ..., -6.5193e-09, + 0.0000e+00, 6.9849e-10], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.5134e-09, 2.3283e-10, ..., 3.1432e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0193, -0.0344, -0.0305, -0.0276, -0.0057, 0.0114, 0.0106, -0.0062, + -0.0174, -0.0183], device='cuda:0'), grad: tensor([ 2.3283e-10, -6.8685e-09, 5.8208e-10, 3.4925e-10, 1.6298e-09, + 5.0059e-09, 1.6298e-09, -7.5670e-09, 3.4925e-10, 7.2177e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 220.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4120 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.2623, -0.3228, 0.1279, ..., -0.1670, 0.0467, 0.0475], + [-0.1582, -0.0879, -0.1228, ..., -0.2664, -0.0727, -0.0253], + [ 0.0073, -0.2076, -0.2782, ..., -0.2045, 0.0305, -0.4486], + ..., + [-0.2362, 0.2019, 0.0350, ..., 0.2679, -0.0729, -0.1768], + [-0.2313, -0.2215, 0.2569, ..., -0.1969, -0.1666, 0.1739], + [ 0.0153, -0.4092, 0.2139, ..., 0.0575, -0.2090, -0.2036]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 9.3132e-10, ..., 8.1491e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.1491e-10, 3.4925e-10, ..., 8.1491e-10, + 0.0000e+00, 0.0000e+00], + [-3.3760e-09, -1.1642e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-10, -5.0059e-09, -1.6298e-09, ..., -3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 1.8626e-09, -1.0477e-09, ..., 1.0827e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0192, -0.0344, -0.0305, -0.0276, -0.0057, 0.0114, 0.0105, -0.0063, + -0.0174, -0.0183], device='cuda:0'), grad: tensor([ 6.6357e-09, 3.0268e-09, -2.6659e-08, 7.4506e-09, -6.2981e-08, + 1.7462e-09, 2.5611e-09, -8.4983e-09, 1.1642e-09, 7.8930e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 220.64, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4124 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.2623, -0.3228, 0.1279, ..., -0.1670, 0.0467, 0.0475], + [-0.1582, -0.0880, -0.1229, ..., -0.2665, -0.0727, -0.0253], + [ 0.0073, -0.2076, -0.2782, ..., -0.2045, 0.0305, -0.4486], + ..., + [-0.2362, 0.2020, 0.0350, ..., 0.2680, -0.0729, -0.1768], + [-0.2313, -0.2215, 0.2570, ..., -0.1969, -0.1666, 0.1740], + [ 0.0153, -0.4093, 0.2140, ..., 0.0575, -0.2090, -0.2037]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3388e-08, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09], + [ 0.0000e+00, 2.3283e-10, 1.1642e-10, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-1.5134e-09, 0.0000e+00, -4.5402e-09, ..., 0.0000e+00, + 0.0000e+00, -4.7730e-09], + [ 1.1642e-10, 1.1642e-10, 2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0192, -0.0344, -0.0302, -0.0277, -0.0057, 0.0114, 0.0105, -0.0063, + -0.0175, -0.0183], device='cuda:0'), grad: tensor([-6.0769e-08, 2.7940e-09, 2.3283e-10, 6.9849e-10, -3.0268e-09, + 2.4447e-09, 7.6485e-08, 3.0268e-09, -2.2352e-08, 2.2119e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 220.19, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4267 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.2623, -0.3228, 0.1280, ..., -0.1670, 0.0467, 0.0475], + [-0.1582, -0.0881, -0.1229, ..., -0.2667, -0.0727, -0.0253], + [ 0.0073, -0.2076, -0.2783, ..., -0.2045, 0.0305, -0.4487], + ..., + [-0.2362, 0.2020, 0.0350, ..., 0.2681, -0.0729, -0.1768], + [-0.2314, -0.2216, 0.2571, ..., -0.1970, -0.1666, 0.1740], + [ 0.0153, -0.4093, 0.2140, ..., 0.0575, -0.2090, -0.2037]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.1491e-10, 2.3283e-10, ..., 9.3132e-10, + 3.0268e-09, 0.0000e+00], + [ 0.0000e+00, 2.2701e-09, 5.2387e-10, ..., 2.9104e-09, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 8.0909e-09, 2.9104e-10, ..., 1.1059e-09, + 1.7462e-10, 0.0000e+00], + ..., + [ 0.0000e+00, -5.1223e-09, -1.3388e-09, ..., -5.4133e-09, + 0.0000e+00, 0.0000e+00], + [ 8.1491e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.8208e-11, 6.9849e-10], + [ 5.8208e-11, 2.1537e-09, 6.4028e-10, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0192, -0.0345, -0.0302, -0.0277, -0.0057, 0.0114, 0.0105, -0.0064, + -0.0175, -0.0183], device='cuda:0'), grad: tensor([ 1.1787e-07, 1.3446e-08, 3.3644e-08, -3.9523e-08, -6.1700e-09, + 3.0443e-08, -1.4773e-07, -1.3446e-08, 2.7358e-09, 1.2049e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 220.34, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4351 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.2623, -0.3228, 0.1281, ..., -0.1670, 0.0467, 0.0476], + [-0.1583, -0.0881, -0.1229, ..., -0.2667, -0.0727, -0.0253], + [ 0.0074, -0.2077, -0.2783, ..., -0.2045, 0.0305, -0.4487], + ..., + [-0.2363, 0.2021, 0.0350, ..., 0.2681, -0.0729, -0.1768], + [-0.2314, -0.2216, 0.2571, ..., -0.1970, -0.1666, 0.1741], + [ 0.0153, -0.4094, 0.2141, ..., 0.0575, -0.2090, -0.2038]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.0373e-09, 0.0000e+00, ..., -1.5134e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.9104e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.7462e-10, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0191, -0.0345, -0.0300, -0.0277, -0.0057, 0.0114, 0.0104, -0.0064, + -0.0175, -0.0183], device='cuda:0'), grad: tensor([ 1.7462e-10, 8.7311e-10, 8.7311e-10, -3.3120e-08, 2.6193e-09, + 3.5157e-08, 4.6566e-10, -3.2014e-09, 6.9849e-10, 4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 220.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4037 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.22 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.2623, -0.3228, 0.1281, ..., -0.1670, 0.0467, 0.0476], + [-0.1583, -0.0882, -0.1230, ..., -0.2669, -0.0727, -0.0253], + [ 0.0074, -0.2077, -0.2783, ..., -0.2045, 0.0305, -0.4488], + ..., + [-0.2363, 0.2022, 0.0351, ..., 0.2682, -0.0729, -0.1768], + [-0.2314, -0.2216, 0.2571, ..., -0.1970, -0.1666, 0.1741], + [ 0.0153, -0.4094, 0.2141, ..., 0.0575, -0.2090, -0.2039]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 1.2224e-09, 2.9104e-10, ..., 1.1642e-09, + 0.0000e+00, 5.8208e-11], + [ 0.0000e+00, -1.4552e-09, 0.0000e+00, ..., -1.4435e-08, + 0.0000e+00, -4.7148e-09], + ..., + [ 5.8208e-11, -4.0745e-10, -4.0745e-10, ..., 1.2456e-08, + 0.0000e+00, 4.6566e-09], + [ 5.8208e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 4.0745e-10], + [ 1.5134e-09, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0191, -0.0346, -0.0299, -0.0277, -0.0057, 0.0114, 0.0104, -0.0064, + -0.0175, -0.0184], device='cuda:0'), grad: tensor([ 5.2387e-10, 3.9581e-09, -6.1293e-08, 4.0163e-09, 1.0477e-09, + -4.0745e-09, -2.0955e-09, 5.0291e-08, 4.1910e-09, 3.4925e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 220.67, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4162 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.21 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.2623, -0.3228, 0.1281, ..., -0.1671, 0.0467, 0.0476], + [-0.1583, -0.0883, -0.1231, ..., -0.2670, -0.0727, -0.0253], + [ 0.0074, -0.2077, -0.2784, ..., -0.2046, 0.0305, -0.4488], + ..., + [-0.2363, 0.2023, 0.0351, ..., 0.2683, -0.0729, -0.1768], + [-0.2315, -0.2216, 0.2572, ..., -0.1970, -0.1666, 0.1741], + [ 0.0153, -0.4095, 0.2143, ..., 0.0576, -0.2090, -0.2039]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -9.7207e-09, ..., 5.8208e-11, + 0.0000e+00, -2.9104e-09], + [ 2.3283e-10, 1.7462e-10, 7.5670e-09, ..., 2.3283e-10, + 0.0000e+00, 1.9791e-09], + [ 2.2119e-09, 5.8208e-11, 2.6193e-09, ..., 5.8208e-10, + 0.0000e+00, 1.1642e-10], + ..., + [ 4.6566e-10, -1.1642e-09, 2.3283e-10, ..., -1.3970e-09, + 0.0000e+00, 5.8208e-11], + [ 9.3132e-10, 0.0000e+00, 3.4925e-10, ..., 1.1642e-10, + 0.0000e+00, 6.4028e-10], + [-9.0222e-09, 5.2387e-10, -8.3237e-09, ..., -2.0955e-09, + 0.0000e+00, 3.4925e-10]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0191, -0.0347, -0.0298, -0.0277, -0.0057, 0.0114, 0.0103, -0.0064, + -0.0176, -0.0183], device='cuda:0'), grad: tensor([-2.4971e-08, 2.0082e-08, 1.0128e-08, 1.8685e-08, 8.9058e-09, + 1.1642e-10, 5.4715e-09, -8.1491e-10, 3.7835e-09, -3.2713e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 220.47, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4149 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.2623, -0.3228, 0.1282, ..., -0.1671, 0.0467, 0.0477], + [-0.1584, -0.0884, -0.1231, ..., -0.2671, -0.0727, -0.0253], + [ 0.0074, -0.2077, -0.2784, ..., -0.2045, 0.0305, -0.4488], + ..., + [-0.2363, 0.2023, 0.0350, ..., 0.2683, -0.0729, -0.1769], + [-0.2316, -0.2216, 0.2573, ..., -0.1970, -0.1666, 0.1742], + [ 0.0152, -0.4095, 0.2144, ..., 0.0576, -0.2090, -0.2040]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2224e-09, ..., 5.8208e-11, + 0.0000e+00, 5.8208e-11], + [-5.8208e-11, 1.1642e-10, 1.1642e-10, ..., 3.4925e-10, + 0.0000e+00, -6.4028e-10], + [ 0.0000e+00, 0.0000e+00, 3.0850e-09, ..., -2.4447e-09, + 0.0000e+00, 5.3551e-09], + ..., + [ 0.0000e+00, 2.9104e-10, 3.4925e-10, ..., 1.9209e-09, + 0.0000e+00, 5.2387e-10], + [ 0.0000e+00, 0.0000e+00, -4.3074e-09, ..., 0.0000e+00, + 0.0000e+00, -7.5088e-09], + [ 2.3283e-10, 5.8208e-11, 1.1642e-10, ..., 5.2387e-10, + 0.0000e+00, 5.8208e-11]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0190, -0.0347, -0.0296, -0.0277, -0.0057, 0.0114, 0.0101, -0.0065, + -0.0175, -0.0183], device='cuda:0'), grad: tensor([-2.9104e-09, -2.3283e-10, -1.3039e-08, 6.0536e-09, -2.0373e-09, + 2.8522e-09, 6.5193e-09, 2.4796e-08, -2.0082e-08, 3.4925e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 220.11, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4157 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.2623, -0.3228, 0.1282, ..., -0.1671, 0.0467, 0.0477], + [-0.1584, -0.0884, -0.1232, ..., -0.2671, -0.0727, -0.0253], + [ 0.0074, -0.2077, -0.2785, ..., -0.2045, 0.0304, -0.4489], + ..., + [-0.2363, 0.2023, 0.0351, ..., 0.2684, -0.0729, -0.1769], + [-0.2316, -0.2216, 0.2574, ..., -0.1970, -0.1666, 0.1743], + [ 0.0152, -0.4095, 0.2144, ..., 0.0576, -0.2090, -0.2041]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, 2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 1.7462e-10, 1.1642e-10, ..., 5.8208e-11, + 0.0000e+00, 1.1642e-10], + [ 5.8208e-11, -1.7462e-10, 5.8208e-11, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + ..., + [ 5.8208e-11, 2.3283e-10, 7.5670e-10, ..., 4.6566e-10, + 0.0000e+00, 1.1642e-10], + [-1.6880e-09, 1.1642e-10, -3.1432e-09, ..., 5.8208e-11, + 0.0000e+00, -4.5984e-09], + [ 1.2224e-09, 0.0000e+00, 7.5670e-10, ..., -9.8953e-10, + 0.0000e+00, 2.9104e-09]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0190, -0.0348, -0.0295, -0.0277, -0.0057, 0.0114, 0.0101, -0.0066, + -0.0174, -0.0183], device='cuda:0'), grad: tensor([ 1.3970e-09, 4.9477e-09, -3.2596e-09, 2.7358e-09, 2.4447e-09, + 3.7719e-08, -3.1025e-08, 5.5879e-09, -9.6625e-09, 3.0268e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 220.04, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4119 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.2623, -0.3228, 0.1283, ..., -0.1672, 0.0467, 0.0478], + [-0.1584, -0.0885, -0.1232, ..., -0.2672, -0.0727, -0.0253], + [ 0.0075, -0.2077, -0.2785, ..., -0.2046, 0.0304, -0.4489], + ..., + [-0.2364, 0.2024, 0.0351, ..., 0.2685, -0.0729, -0.1769], + [-0.2316, -0.2216, 0.2575, ..., -0.1970, -0.1666, 0.1743], + [ 0.0152, -0.4096, 0.2145, ..., 0.0576, -0.2090, -0.2042]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 5.8208e-11], + [ 5.8208e-11, -3.7835e-09, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, -1.1642e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + ..., + [ 5.8208e-11, 3.5507e-09, 6.1700e-09, ..., 9.3132e-10, + 0.0000e+00, 1.1001e-08], + [ 2.9104e-10, 5.8208e-11, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 1.6880e-09], + [ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0190, -0.0348, -0.0294, -0.0277, -0.0057, 0.0115, 0.0101, -0.0066, + -0.0174, -0.0184], device='cuda:0'), grad: tensor([ 1.1642e-09, -6.0827e-08, 1.3388e-09, 4.3656e-09, -9.0222e-09, + 2.8522e-09, -1.0710e-08, 6.1700e-08, 1.7288e-08, 4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 220.57, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4249 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.21 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.2623, -0.3228, 0.1283, ..., -0.1672, 0.0467, 0.0478], + [-0.1584, -0.0885, -0.1232, ..., -0.2673, -0.0727, -0.0253], + [ 0.0074, -0.2077, -0.2786, ..., -0.2046, 0.0304, -0.4490], + ..., + [-0.2364, 0.2024, 0.0351, ..., 0.2686, -0.0729, -0.1769], + [-0.2316, -0.2216, 0.2575, ..., -0.1970, -0.1666, 0.1744], + [ 0.0151, -0.4097, 0.2146, ..., 0.0577, -0.2090, -0.2043]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 1.1642e-10], + [ 1.2806e-09, 7.7998e-09, 3.4925e-09, ..., 1.1292e-08, + 1.1642e-10, 1.7462e-10], + [ 5.8208e-11, 2.3283e-10, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 1.7462e-10], + ..., + [ 1.7462e-10, -8.2655e-09, -3.5507e-09, ..., -6.9267e-09, + 0.0000e+00, 0.0000e+00], + [ 6.4028e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.1642e-10, 8.7311e-10], + [ 1.5716e-09, 1.1642e-10, 5.8208e-11, ..., 5.4133e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0189, -0.0349, -0.0294, -0.0277, -0.0057, 0.0114, 0.0102, -0.0067, + -0.0174, -0.0184], device='cuda:0'), grad: tensor([ 1.5134e-09, 4.3074e-08, 2.2701e-09, 1.9791e-09, -6.1525e-08, + 1.1793e-07, -1.1799e-07, -1.7812e-08, 9.6625e-09, 2.7474e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 220.27, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4142 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.2624, -0.3229, 0.1283, ..., -0.1672, 0.0467, 0.0479], + [-0.1585, -0.0886, -0.1233, ..., -0.2674, -0.0727, -0.0253], + [ 0.0075, -0.2078, -0.2786, ..., -0.2046, 0.0304, -0.4490], + ..., + [-0.2364, 0.2024, 0.0351, ..., 0.2686, -0.0729, -0.1770], + [-0.2317, -0.2216, 0.2576, ..., -0.1970, -0.1666, 0.1745], + [ 0.0152, -0.4097, 0.2147, ..., 0.0577, -0.2090, -0.2043]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, -4.4238e-09, ..., 5.8208e-10, + 0.0000e+00, -1.8626e-09], + [ 1.1642e-10, 3.9581e-09, 1.1059e-09, ..., 3.4925e-09, + 0.0000e+00, 5.8208e-11], + [ 5.8208e-11, 0.0000e+00, 2.9104e-10, ..., 5.8208e-11, + 0.0000e+00, 1.1642e-10], + ..., + [ 5.8208e-11, -4.4820e-09, -1.1059e-09, ..., -4.0163e-09, + 0.0000e+00, 5.8208e-11], + [ 4.0745e-10, 0.0000e+00, 4.8894e-09, ..., 0.0000e+00, + 0.0000e+00, 2.4447e-09], + [ 2.3283e-10, 5.8208e-11, 8.1491e-10, ..., 1.7462e-10, + 0.0000e+00, 5.2387e-10]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0189, -0.0350, -0.0292, -0.0277, -0.0057, 0.0114, 0.0102, -0.0067, + -0.0174, -0.0183], device='cuda:0'), grad: tensor([-5.3551e-09, 1.0827e-08, 1.6880e-09, 7.9162e-09, -2.9686e-09, + -6.4028e-09, 3.4343e-09, -9.4878e-09, 1.3446e-08, 3.2014e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 220.83, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4075 re_mapping 0.0018 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.2624, -0.3229, 0.1284, ..., -0.1672, 0.0467, 0.0479], + [-0.1586, -0.0887, -0.1234, ..., -0.2675, -0.0727, -0.0254], + [ 0.0075, -0.2078, -0.2787, ..., -0.2046, 0.0304, -0.4491], + ..., + [-0.2365, 0.2025, 0.0351, ..., 0.2687, -0.0729, -0.1770], + [-0.2318, -0.2216, 0.2577, ..., -0.1971, -0.1666, 0.1745], + [ 0.0152, -0.4097, 0.2149, ..., 0.0579, -0.2090, -0.2044]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 5.2387e-10, -5.0641e-09, ..., 6.9849e-10, + 0.0000e+00, -4.1327e-09], + [ 1.7462e-10, 2.0373e-09, 4.4238e-09, ..., 1.4552e-09, + 0.0000e+00, 4.0745e-09], + [ 5.8208e-11, 2.6193e-09, 1.3970e-09, ..., 1.3970e-09, + 0.0000e+00, 6.9849e-10], + ..., + [ 1.1642e-10, -5.6461e-09, -5.1805e-09, ..., -6.0536e-09, + 0.0000e+00, 1.7462e-10], + [ 4.3074e-09, 1.7462e-10, -4.8196e-08, ..., 0.0000e+00, + 0.0000e+00, -5.9488e-08], + [ 2.9104e-10, 1.4552e-09, 4.0978e-08, ..., 7.5670e-10, + 0.0000e+00, 5.3435e-08]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0188, -0.0352, -0.0292, -0.0277, -0.0058, 0.0115, 0.0099, -0.0067, + -0.0174, -0.0182], device='cuda:0'), grad: tensor([-2.4738e-08, 1.2573e-08, 9.5461e-09, 1.7462e-09, 1.5716e-09, + 1.2224e-08, 2.8347e-08, -1.5600e-08, -1.3434e-07, 1.2270e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 220.43, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4303 re_mapping 0.0018 re_causal 0.0069 /// teacc 99.22 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.2624, -0.3229, 0.1284, ..., -0.1673, 0.0467, 0.0480], + [-0.1586, -0.0888, -0.1235, ..., -0.2677, -0.0727, -0.0254], + [ 0.0075, -0.2079, -0.2787, ..., -0.2047, 0.0304, -0.4492], + ..., + [-0.2365, 0.2027, 0.0351, ..., 0.2687, -0.0729, -0.1770], + [-0.2319, -0.2217, 0.2578, ..., -0.1971, -0.1666, 0.1746], + [ 0.0153, -0.4098, 0.2151, ..., 0.0580, -0.2090, -0.2045]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 5.8208e-11, ..., 0.0000e+00, 0.0000e+00, + 5.8208e-11], + [0.0000e+00, 1.7462e-10, 5.8208e-11, ..., 2.9104e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 1.7462e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [1.7462e-10, 5.8208e-11, 8.7311e-10, ..., 6.4028e-10, 0.0000e+00, + 5.8208e-11], + [5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0189, -0.0353, -0.0292, -0.0277, -0.0058, 0.0115, 0.0099, -0.0067, + -0.0174, -0.0181], device='cuda:0'), grad: tensor([ 3.7253e-09, 4.3248e-08, 1.3970e-09, 2.3865e-09, -5.3493e-08, + 5.4133e-09, -5.5297e-09, 4.7730e-09, 4.2492e-09, 4.8894e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 220.39, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4139 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.20 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.2624, -0.3229, 0.1285, ..., -0.1673, 0.0467, 0.0480], + [-0.1586, -0.0889, -0.1235, ..., -0.2678, -0.0727, -0.0254], + [ 0.0075, -0.2079, -0.2788, ..., -0.2047, 0.0304, -0.4493], + ..., + [-0.2365, 0.2027, 0.0350, ..., 0.2688, -0.0729, -0.1770], + [-0.2320, -0.2217, 0.2579, ..., -0.1971, -0.1666, 0.1746], + [ 0.0152, -0.4099, 0.2153, ..., 0.0580, -0.2090, -0.2046]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + [ 1.7462e-10, 8.1491e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 1.7462e-10, 5.8208e-11, 5.8208e-11, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + ..., + [ 0.0000e+00, -2.7940e-09, 5.8208e-11, ..., -1.2515e-08, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 5.8208e-11, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 5.8208e-11, 5.8208e-11, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0188, -0.0354, -0.0292, -0.0277, -0.0058, 0.0115, 0.0098, -0.0068, + -0.0175, -0.0180], device='cuda:0'), grad: tensor([ 1.5716e-09, 4.1327e-09, 1.7462e-09, -3.4925e-09, 3.2713e-08, + 1.1059e-08, -1.0768e-08, -2.9162e-08, 3.2014e-09, 6.4028e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 220.19, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4239 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.2624, -0.3229, 0.1285, ..., -0.1673, 0.0467, 0.0480], + [-0.1587, -0.0890, -0.1236, ..., -0.2679, -0.0727, -0.0255], + [ 0.0074, -0.2080, -0.2788, ..., -0.2047, 0.0304, -0.4493], + ..., + [-0.2365, 0.2029, 0.0351, ..., 0.2689, -0.0729, -0.1770], + [-0.2321, -0.2217, 0.2580, ..., -0.1971, -0.1666, 0.1746], + [ 0.0153, -0.4099, 0.2154, ..., 0.0580, -0.2090, -0.2047]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + [ 5.8208e-11, 9.8953e-10, 2.9104e-10, ..., 1.0477e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.3388e-09, 5.8208e-11, ..., 5.8208e-11, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 5.8208e-11], + [ 1.1642e-10, 1.2806e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 1.7462e-10, 5.8208e-11, ..., 1.7462e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0188, -0.0355, -0.0293, -0.0277, -0.0058, 0.0116, 0.0097, -0.0068, + -0.0175, -0.0179], device='cuda:0'), grad: tensor([ 2.3283e-09, 3.2596e-09, -1.4494e-08, 1.3388e-09, 2.1537e-09, + 6.5193e-09, -5.0059e-09, 7.5670e-10, 1.4319e-08, 9.8953e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 220.24, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4219 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.2624, -0.3229, 0.1286, ..., -0.1673, 0.0467, 0.0481], + [-0.1587, -0.0892, -0.1237, ..., -0.2680, -0.0727, -0.0254], + [ 0.0074, -0.2080, -0.2789, ..., -0.2048, 0.0304, -0.4494], + ..., + [-0.2366, 0.2031, 0.0352, ..., 0.2691, -0.0729, -0.1770], + [-0.2322, -0.2217, 0.2580, ..., -0.1971, -0.1666, 0.1746], + [ 0.0153, -0.4100, 0.2155, ..., 0.0581, -0.2090, -0.2047]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.4925e-10, 2.3283e-10, 1.7462e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-10, 0.0000e+00, 2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 1.7462e-10], + [ 1.7462e-10, 0.0000e+00, 5.8208e-11, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + [-6.9849e-10, 1.1642e-10, -9.8953e-10, ..., -4.6566e-10, + 0.0000e+00, 5.8208e-11]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0188, -0.0356, -0.0293, -0.0277, -0.0058, 0.0116, 0.0096, -0.0066, + -0.0175, -0.0179], device='cuda:0'), grad: tensor([ 6.4028e-10, 1.0070e-08, -1.0477e-08, 1.2806e-09, -2.9686e-09, + -9.8953e-10, 3.9581e-09, 4.2492e-09, 3.4925e-10, -2.7358e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 220.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0018 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.2625, -0.3229, 0.1286, ..., -0.1673, 0.0467, 0.0481], + [-0.1588, -0.0893, -0.1238, ..., -0.2681, -0.0727, -0.0254], + [ 0.0074, -0.2081, -0.2789, ..., -0.2048, 0.0304, -0.4495], + ..., + [-0.2366, 0.2031, 0.0352, ..., 0.2692, -0.0729, -0.1771], + [-0.2322, -0.2218, 0.2581, ..., -0.1971, -0.1666, 0.1747], + [ 0.0153, -0.4101, 0.2156, ..., 0.0581, -0.2090, -0.2047]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.2806e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 8.1491e-10], + [1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0187, -0.0357, -0.0294, -0.0277, -0.0059, 0.0116, 0.0095, -0.0066, + -0.0175, -0.0179], device='cuda:0'), grad: tensor([ 2.3283e-10, 6.9849e-10, 1.1642e-10, 8.1491e-10, 1.1642e-10, + -2.3283e-09, 2.0955e-09, 8.1491e-10, 1.9791e-09, 8.1491e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 220.66, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4191 re_mapping 0.0018 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.2625, -0.3230, 0.1287, ..., -0.1673, 0.0467, 0.0481], + [-0.1588, -0.0894, -0.1238, ..., -0.2682, -0.0727, -0.0254], + [ 0.0074, -0.2081, -0.2790, ..., -0.2049, 0.0304, -0.4495], + ..., + [-0.2366, 0.2032, 0.0352, ..., 0.2693, -0.0729, -0.1771], + [-0.2322, -0.2218, 0.2581, ..., -0.1971, -0.1666, 0.1747], + [ 0.0152, -0.4102, 0.2156, ..., 0.0580, -0.2090, -0.2049]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, 5.8208e-10, ..., 1.1642e-10, + 1.1642e-10, 0.0000e+00], + [ 0.0000e+00, 1.0477e-09, 3.4925e-10, ..., 9.3132e-10, + 1.0477e-09, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.1642e-10, + 9.1968e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -1.3970e-09, -5.8208e-10, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 6.9849e-10, 3.4925e-10, ..., 0.0000e+00, + 6.9849e-10, 1.1642e-10], + [ 1.1642e-10, 8.1491e-10, 3.4925e-10, ..., 8.1491e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0188, -0.0358, -0.0293, -0.0277, -0.0059, 0.0116, 0.0096, -0.0066, + -0.0175, -0.0180], device='cuda:0'), grad: tensor([ 4.5402e-09, 8.4983e-09, 5.0291e-08, -1.7462e-09, 1.0710e-08, + 9.1968e-09, -7.1712e-08, -3.2596e-09, 5.9372e-09, 2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 220.67, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4027 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.2625, -0.3230, 0.1287, ..., -0.1674, 0.0467, 0.0481], + [-0.1588, -0.0894, -0.1239, ..., -0.2684, -0.0727, -0.0254], + [ 0.0074, -0.2082, -0.2790, ..., -0.2049, 0.0303, -0.4496], + ..., + [-0.2366, 0.2034, 0.0353, ..., 0.2695, -0.0729, -0.1771], + [-0.2323, -0.2218, 0.2582, ..., -0.1972, -0.1666, 0.1747], + [ 0.0153, -0.4103, 0.2157, ..., 0.0581, -0.2090, -0.2049]], + device='cuda:0'), grad: tensor([[-2.4447e-09, 0.0000e+00, -1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, -2.1188e-08], + [ 5.8208e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 2.4447e-09], + [ 1.9791e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.5134e-09]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0188, -0.0358, -0.0294, -0.0277, -0.0059, 0.0117, 0.0094, -0.0066, + -0.0176, -0.0179], device='cuda:0'), grad: tensor([-7.0548e-08, 1.9791e-09, 2.3283e-10, 1.0827e-07, -3.1432e-09, + -1.1781e-07, 6.6124e-08, 9.3132e-10, 6.4028e-09, 7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 220.68, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4252 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.17 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.2625, -0.3230, 0.1288, ..., -0.1674, 0.0467, 0.0481], + [-0.1589, -0.0895, -0.1239, ..., -0.2684, -0.0727, -0.0254], + [ 0.0074, -0.2082, -0.2791, ..., -0.2049, 0.0303, -0.4496], + ..., + [-0.2366, 0.2034, 0.0353, ..., 0.2695, -0.0729, -0.1771], + [-0.2324, -0.2218, 0.2582, ..., -0.1972, -0.1666, 0.1747], + [ 0.0154, -0.4103, 0.2158, ..., 0.0581, -0.2090, -0.2049]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 2.3283e-10], + [0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0187, -0.0358, -0.0293, -0.0277, -0.0060, 0.0117, 0.0093, -0.0066, + -0.0176, -0.0179], device='cuda:0'), grad: tensor([ 2.4447e-09, 1.8626e-09, 1.1642e-09, 6.9849e-10, -3.4925e-09, + 1.3970e-09, 3.2596e-09, 6.9849e-10, 4.6566e-10, 8.1491e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 220.64, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4246 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.18 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.2625, -0.3230, 0.1288, ..., -0.1674, 0.0467, 0.0482], + [-0.1589, -0.0895, -0.1240, ..., -0.2685, -0.0727, -0.0255], + [ 0.0075, -0.2083, -0.2791, ..., -0.2050, 0.0303, -0.4497], + ..., + [-0.2367, 0.2034, 0.0353, ..., 0.2696, -0.0729, -0.1771], + [-0.2324, -0.2218, 0.2583, ..., -0.1972, -0.1666, 0.1748], + [ 0.0153, -0.4103, 0.2159, ..., 0.0581, -0.2090, -0.2050]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.4925e-10], + [ 2.3283e-10, -2.5611e-09, -3.4925e-10, ..., 2.3283e-10, + 0.0000e+00, -1.2806e-09], + [ 1.1642e-10, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [ 5.8208e-10, 6.9849e-10, 2.3283e-10, ..., -1.2806e-09, + 0.0000e+00, 1.3970e-09], + [ 1.9791e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.6298e-09], + [ 9.3132e-10, 5.8208e-10, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.5134e-09]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0187, -0.0358, -0.0292, -0.0278, -0.0060, 0.0117, 0.0093, -0.0067, + -0.0176, -0.0179], device='cuda:0'), grad: tensor([ 2.0955e-09, -1.3155e-08, 2.3283e-09, 8.1491e-09, 6.9849e-10, + -1.0675e-07, 8.7428e-08, 1.2456e-08, 6.8685e-09, 3.9581e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 220.62, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4308 re_mapping 0.0018 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.2626, -0.3230, 0.1288, ..., -0.1674, 0.0467, 0.0482], + [-0.1589, -0.0895, -0.1240, ..., -0.2686, -0.0727, -0.0254], + [ 0.0074, -0.2083, -0.2791, ..., -0.2050, 0.0303, -0.4498], + ..., + [-0.2367, 0.2034, 0.0353, ..., 0.2696, -0.0729, -0.1772], + [-0.2325, -0.2218, 0.2583, ..., -0.1972, -0.1666, 0.1748], + [ 0.0153, -0.4103, 0.2160, ..., 0.0581, -0.2090, -0.2050]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.1642e-10, -1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, -5.8208e-10], + [ 0.0000e+00, 3.4925e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 8.1491e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 5.8208e-10], + [ 0.0000e+00, 3.4925e-10, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-6.9849e-10, 3.4925e-10, -2.2119e-09, ..., -3.6089e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0188, -0.0358, -0.0291, -0.0278, -0.0061, 0.0118, 0.0093, -0.0069, + -0.0177, -0.0179], device='cuda:0'), grad: tensor([ 5.8208e-10, -5.1223e-09, 1.7462e-09, -2.9104e-09, 9.8953e-09, + 3.7253e-09, 1.7462e-09, 8.0327e-09, 1.7462e-09, -1.0594e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 220.69, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4029 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.16 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.2626, -0.3230, 0.1289, ..., -0.1674, 0.0467, 0.0482], + [-0.1590, -0.0897, -0.1240, ..., -0.2687, -0.0727, -0.0255], + [ 0.0074, -0.2083, -0.2792, ..., -0.2050, 0.0303, -0.4498], + ..., + [-0.2367, 0.2035, 0.0353, ..., 0.2697, -0.0729, -0.1772], + [-0.2326, -0.2219, 0.2584, ..., -0.1972, -0.1666, 0.1748], + [ 0.0153, -0.4104, 0.2161, ..., 0.0582, -0.2090, -0.2051]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0827e-08, ..., 0.0000e+00, + -2.3283e-10, -1.4203e-08], + [ 1.1642e-10, 3.4925e-10, 6.9849e-10, ..., 1.1642e-10, + 0.0000e+00, 3.4925e-10], + [-2.3283e-09, -4.6566e-10, 2.6776e-09, ..., 0.0000e+00, + 0.0000e+00, 2.9104e-09], + ..., + [ 0.0000e+00, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 8.1491e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 4.6566e-10, 5.8208e-10, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-10]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0188, -0.0359, -0.0290, -0.0278, -0.0061, 0.0119, 0.0092, -0.0069, + -0.0179, -0.0179], device='cuda:0'), grad: tensor([-5.5530e-08, 4.3074e-09, -9.8953e-09, 3.1432e-09, 1.5134e-09, + 2.0838e-08, 2.4564e-08, 3.2596e-09, 3.8417e-09, 1.2340e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 220.32, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4415 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.2626, -0.3230, 0.1289, ..., -0.1674, 0.0467, 0.0482], + [-0.1590, -0.0898, -0.1241, ..., -0.2688, -0.0727, -0.0255], + [ 0.0074, -0.2084, -0.2792, ..., -0.2050, 0.0303, -0.4499], + ..., + [-0.2367, 0.2036, 0.0353, ..., 0.2698, -0.0729, -0.1772], + [-0.2327, -0.2219, 0.2584, ..., -0.1972, -0.1666, 0.1748], + [ 0.0154, -0.4104, 0.2163, ..., 0.0582, -0.2090, -0.2051]], + device='cuda:0'), grad: tensor([[2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 1.1642e-10], + [1.1642e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [2.3283e-10, 1.1642e-10, 1.1642e-10, ..., 2.3283e-10, 0.0000e+00, + 1.1642e-10], + [1.5134e-09, 3.4925e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 9.3132e-10], + [6.0536e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 3.8417e-09]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0188, -0.0361, -0.0289, -0.0278, -0.0063, 0.0119, 0.0092, -0.0069, + -0.0178, -0.0177], device='cuda:0'), grad: tensor([ 6.9849e-10, 1.2806e-09, 1.0477e-09, -3.1083e-08, 2.0955e-09, + 8.0327e-09, 6.4028e-09, 1.7462e-09, 3.4925e-09, 1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 220.57, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4020 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.19 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.2626, -0.3230, 0.1290, ..., -0.1674, 0.0467, 0.0482], + [-0.1591, -0.0899, -0.1243, ..., -0.2690, -0.0727, -0.0255], + [ 0.0074, -0.2085, -0.2793, ..., -0.2051, 0.0303, -0.4499], + ..., + [-0.2368, 0.2037, 0.0354, ..., 0.2700, -0.0729, -0.1772], + [-0.2328, -0.2219, 0.2584, ..., -0.1972, -0.1666, 0.1748], + [ 0.0154, -0.4105, 0.2164, ..., 0.0583, -0.2090, -0.2052]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.8953e-09, ..., -2.0955e-09, + 0.0000e+00, -8.1491e-09], + [ 0.0000e+00, 6.9849e-10, 2.3283e-09, ..., 1.5134e-09, + 0.0000e+00, 8.1491e-10], + [ 0.0000e+00, -8.1491e-10, 5.8208e-10, ..., 1.1642e-10, + 0.0000e+00, 8.1491e-10], + ..., + [ 0.0000e+00, 6.9849e-10, 1.6531e-08, ..., 1.3621e-08, + 0.0000e+00, 1.2806e-09], + [ 2.3283e-10, 2.6776e-09, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, -1.5134e-09], + [ 2.3283e-10, 9.3132e-10, -3.0152e-08, ..., -2.8522e-08, + 0.0000e+00, 2.5611e-09]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0188, -0.0363, -0.0289, -0.0278, -0.0063, 0.0120, 0.0090, -0.0068, + -0.0180, -0.0176], device='cuda:0'), grad: tensor([-3.5157e-08, 1.0477e-08, -2.0606e-08, -7.4506e-09, 4.6799e-08, + 5.5181e-08, -2.1770e-08, 6.4028e-08, 2.2934e-08, -1.0629e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 220.50, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4091 re_mapping 0.0018 re_causal 0.0068 /// teacc 99.18 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.2626, -0.3230, 0.1291, ..., -0.1675, 0.0467, 0.0483], + [-0.1591, -0.0900, -0.1243, ..., -0.2691, -0.0727, -0.0256], + [ 0.0074, -0.2085, -0.2793, ..., -0.2051, 0.0303, -0.4500], + ..., + [-0.2368, 0.2038, 0.0354, ..., 0.2701, -0.0729, -0.1772], + [-0.2329, -0.2219, 0.2585, ..., -0.1972, -0.1666, 0.1748], + [ 0.0154, -0.4106, 0.2166, ..., 0.0583, -0.2090, -0.2052]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, -1.2806e-09, -1.5134e-09, ..., -1.1642e-09, + 0.0000e+00, -1.1642e-09], + ..., + [ 0.0000e+00, 2.3283e-10, 5.8208e-10, ..., 4.6566e-10, + 0.0000e+00, 3.4925e-10], + [ 2.3283e-10, 3.4925e-10, 6.9849e-10, ..., 5.8208e-10, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0187, -0.0364, -0.0289, -0.0278, -0.0063, 0.0119, 0.0091, -0.0069, + -0.0181, -0.0176], device='cuda:0'), grad: tensor([ 2.0955e-09, 2.1886e-08, -3.5157e-08, 2.4447e-09, 1.8626e-09, + 2.7940e-09, -2.7940e-09, 5.0059e-09, 6.4028e-09, 5.8208e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 220.61, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4117 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.2627, -0.3230, 0.1291, ..., -0.1675, 0.0467, 0.0484], + [-0.1591, -0.0900, -0.1244, ..., -0.2691, -0.0727, -0.0256], + [ 0.0074, -0.2085, -0.2793, ..., -0.2051, 0.0303, -0.4500], + ..., + [-0.2368, 0.2038, 0.0354, ..., 0.2702, -0.0729, -0.1773], + [-0.2329, -0.2220, 0.2585, ..., -0.1972, -0.1666, 0.1748], + [ 0.0155, -0.4106, 0.2167, ..., 0.0584, -0.2090, -0.2053]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0477e-09, ..., 1.5134e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 1.1642e-10, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, -2.4098e-08, ..., -3.3993e-08, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0186, -0.0365, -0.0288, -0.0278, -0.0065, 0.0119, 0.0090, -0.0070, + -0.0181, -0.0175], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.9791e-09, 1.8626e-09, 1.9791e-09, 8.8592e-08, + 1.1642e-09, 9.1968e-09, 6.5193e-09, 9.3132e-10, -9.9652e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 220.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4382 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.20 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.2627, -0.3230, 0.1292, ..., -0.1675, 0.0467, 0.0484], + [-0.1592, -0.0901, -0.1245, ..., -0.2692, -0.0727, -0.0255], + [ 0.0074, -0.2086, -0.2794, ..., -0.2052, 0.0303, -0.4500], + ..., + [-0.2368, 0.2039, 0.0354, ..., 0.2702, -0.0729, -0.1773], + [-0.2329, -0.2220, 0.2585, ..., -0.1972, -0.1666, 0.1748], + [ 0.0155, -0.4107, 0.2169, ..., 0.0585, -0.2090, -0.2053]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 1.2806e-09, 4.6566e-10, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -3.4925e-10, -6.9849e-10, ..., 0.0000e+00, + 1.7462e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -2.4447e-09, -8.1491e-10, ..., -1.0477e-09, + 0.0000e+00, -3.4925e-10], + [ 0.0000e+00, 1.6298e-09, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-09, -1.2806e-09], + [ 0.0000e+00, 3.4925e-10, -3.9581e-09, ..., -3.7253e-09, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0185, -0.0365, -0.0287, -0.0278, -0.0067, 0.0119, 0.0091, -0.0071, + -0.0182, -0.0174], device='cuda:0'), grad: tensor([ 2.4447e-09, 1.4086e-08, 2.4913e-08, 6.7055e-08, 2.0140e-08, + 5.1572e-08, -1.6892e-07, -1.9791e-09, 6.9849e-09, -9.5461e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 220.21, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4160 re_mapping 0.0018 re_causal 0.0066 /// teacc 99.18 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.2627, -0.3231, 0.1293, ..., -0.1675, 0.0467, 0.0485], + [-0.1592, -0.0901, -0.1245, ..., -0.2694, -0.0727, -0.0256], + [ 0.0075, -0.2086, -0.2794, ..., -0.2052, 0.0303, -0.4501], + ..., + [-0.2369, 0.2040, 0.0353, ..., 0.2704, -0.0729, -0.1773], + [-0.2330, -0.2220, 0.2586, ..., -0.1973, -0.1666, 0.1749], + [ 0.0155, -0.4108, 0.2171, ..., 0.0586, -0.2090, -0.2054]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -6.9849e-10], + [ 0.0000e+00, 2.3283e-10, 1.1642e-10, ..., 3.4925e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 1.8626e-09, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0185, -0.0366, -0.0286, -0.0279, -0.0068, 0.0120, 0.0090, -0.0071, + -0.0182, -0.0173], device='cuda:0'), grad: tensor([ 3.3760e-09, -4.6566e-09, 9.1968e-09, 3.4925e-10, -8.3819e-09, + 5.1223e-09, -9.3132e-09, 2.6776e-09, 1.8626e-09, 9.4296e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 220.51, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4301 re_mapping 0.0018 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.2627, -0.3231, 0.1294, ..., -0.1675, 0.0467, 0.0486], + [-0.1592, -0.0902, -0.1246, ..., -0.2695, -0.0727, -0.0255], + [ 0.0074, -0.2087, -0.2795, ..., -0.2053, 0.0303, -0.4502], + ..., + [-0.2369, 0.2041, 0.0353, ..., 0.2705, -0.0729, -0.1773], + [-0.2331, -0.2220, 0.2586, ..., -0.1973, -0.1666, 0.1748], + [ 0.0155, -0.4109, 0.2173, ..., 0.0587, -0.2090, -0.2055]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.8208e-10, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, -8.4983e-09, 2.6776e-09, ..., 5.8208e-09, + 0.0000e+00, -2.3283e-09], + [-3.4925e-10, 2.4447e-09, 5.8208e-10, ..., 1.2806e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.0477e-09, -3.6089e-09, ..., -8.1491e-09, + 0.0000e+00, 1.9791e-09], + [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 3.4925e-10, + 0.0000e+00, 0.0000e+00], + [-5.8208e-10, 1.2806e-09, -9.3132e-10, ..., -4.6566e-10, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0183, -0.0366, -0.0286, -0.0279, -0.0069, 0.0120, 0.0089, -0.0071, + -0.0184, -0.0172], device='cuda:0'), grad: tensor([ 2.9104e-09, -7.7649e-08, 3.4925e-09, 6.0536e-09, 5.4715e-09, + 4.5402e-09, 2.7940e-09, 4.9127e-08, 3.6089e-09, 2.6776e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 220.47, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4347 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.2627, -0.3231, 0.1296, ..., -0.1675, 0.0467, 0.0487], + [-0.1592, -0.0902, -0.1247, ..., -0.2696, -0.0727, -0.0256], + [ 0.0075, -0.2087, -0.2795, ..., -0.2053, 0.0303, -0.4503], + ..., + [-0.2369, 0.2041, 0.0354, ..., 0.2706, -0.0729, -0.1774], + [-0.2332, -0.2220, 0.2586, ..., -0.1973, -0.1666, 0.1748], + [ 0.0154, -0.4109, 0.2174, ..., 0.0588, -0.2090, -0.2056]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 1.1642e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + -9.0804e-09, -7.9162e-09], + [ 0.0000e+00, 1.9791e-09, 0.0000e+00, ..., 0.0000e+00, + 4.8894e-09, 7.7998e-09], + ..., + [ 1.1642e-10, 1.1642e-10, 1.1642e-10, ..., 1.1642e-10, + 3.7253e-09, 1.1642e-10], + [ 1.0477e-09, 1.5134e-09, 6.9849e-10, ..., 0.0000e+00, + 1.1642e-10, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0181, -0.0367, -0.0283, -0.0280, -0.0071, 0.0121, 0.0090, -0.0072, + -0.0186, -0.0172], device='cuda:0'), grad: tensor([ 2.9104e-09, -2.4168e-07, 1.6158e-07, 4.5402e-09, 2.7940e-09, + -5.1921e-08, 3.3178e-08, 8.2888e-08, 9.4296e-09, 4.0745e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 220.82, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4055 re_mapping 0.0018 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.2627, -0.3231, 0.1297, ..., -0.1675, 0.0467, 0.0487], + [-0.1593, -0.0903, -0.1247, ..., -0.2696, -0.0727, -0.0255], + [ 0.0075, -0.2087, -0.2796, ..., -0.2053, 0.0302, -0.4504], + ..., + [-0.2369, 0.2042, 0.0354, ..., 0.2707, -0.0729, -0.1774], + [-0.2333, -0.2221, 0.2587, ..., -0.1973, -0.1666, 0.1748], + [ 0.0154, -0.4110, 0.2176, ..., 0.0589, -0.2090, -0.2056]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.2806e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 0.0000e+00, 1.1642e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., -3.4925e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 1.0012e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 1.8626e-09, 1.1642e-10, -1.2573e-08, ..., 3.0268e-09, + 0.0000e+00, -1.3970e-09]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0180, -0.0367, -0.0283, -0.0280, -0.0073, 0.0121, 0.0089, -0.0073, + -0.0188, -0.0171], device='cuda:0'), grad: tensor([ 3.3760e-09, 2.5611e-09, -2.6776e-09, 2.5611e-09, -2.0838e-08, + 2.7940e-09, 5.9372e-09, 3.9581e-09, 2.3632e-08, -1.2224e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 220.64, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4043 re_mapping 0.0018 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.2627, -0.3231, 0.1298, ..., -0.1675, 0.0467, 0.0488], + [-0.1593, -0.0903, -0.1248, ..., -0.2697, -0.0727, -0.0255], + [ 0.0075, -0.2088, -0.2797, ..., -0.2053, 0.0302, -0.4505], + ..., + [-0.2370, 0.2042, 0.0354, ..., 0.2707, -0.0729, -0.1775], + [-0.2334, -0.2221, 0.2587, ..., -0.1973, -0.1666, 0.1748], + [ 0.0154, -0.4110, 0.2177, ..., 0.0590, -0.2090, -0.2057]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 1.7462e-09, ..., 1.3970e-09, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 1.1642e-10, 3.4925e-10, ..., 1.1642e-10, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 2.3283e-10, 6.9849e-10, ..., 2.3283e-10, + 1.1642e-10, 1.3970e-09], + ..., + [ 0.0000e+00, 1.1642e-10, 1.6298e-09, ..., 1.0477e-09, + 0.0000e+00, 5.8208e-10], + [ 0.0000e+00, -3.4925e-10, -2.6776e-09, ..., -9.3132e-10, + 0.0000e+00, -1.5134e-09], + [-8.1491e-10, 0.0000e+00, -4.0745e-09, ..., -3.3760e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0178, -0.0365, -0.0284, -0.0280, -0.0074, 0.0122, 0.0088, -0.0075, + -0.0189, -0.0170], device='cuda:0'), grad: tensor([ 6.0536e-09, -2.8173e-08, 1.9441e-08, 1.1642e-09, 2.6776e-09, + 2.5611e-09, 5.2387e-09, 7.3342e-09, -3.3760e-09, -1.0827e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 220.60, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4123 re_mapping 0.0018 re_causal 0.0067 /// teacc 99.20 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.2628, -0.3231, 0.1299, ..., -0.1675, 0.0467, 0.0489], + [-0.1594, -0.0904, -0.1249, ..., -0.2698, -0.0727, -0.0255], + [ 0.0075, -0.2089, -0.2797, ..., -0.2054, 0.0302, -0.4505], + ..., + [-0.2370, 0.2043, 0.0354, ..., 0.2708, -0.0729, -0.1775], + [-0.2335, -0.2221, 0.2587, ..., -0.1974, -0.1666, 0.1748], + [ 0.0155, -0.4111, 0.2179, ..., 0.0592, -0.2090, -0.2058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.8208e-10, -6.9849e-10, ..., 1.1642e-10, + -1.1642e-10, -8.1491e-10], + [ 0.0000e+00, -2.0256e-08, 4.6566e-10, ..., 1.0477e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 1.1642e-10], + ..., + [ 0.0000e+00, 1.7462e-08, -1.7462e-09, ..., -3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 3.4925e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 1.1642e-10, 3.3760e-09, 1.6298e-09, ..., 2.4447e-09, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0177, -0.0366, -0.0284, -0.0280, -0.0075, 0.0122, 0.0089, -0.0075, + -0.0190, -0.0168], device='cuda:0'), grad: tensor([-2.2119e-09, -1.3667e-07, 4.3074e-09, 8.6147e-09, 8.1491e-10, + -5.4715e-09, 3.3760e-09, 1.2596e-07, 1.3970e-09, 9.1968e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 220.59, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4101 re_mapping 0.0018 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.2628, -0.3231, 0.1299, ..., -0.1676, 0.0467, 0.0489], + [-0.1594, -0.0904, -0.1249, ..., -0.2698, -0.0727, -0.0255], + [ 0.0075, -0.2089, -0.2797, ..., -0.2054, 0.0302, -0.4505], + ..., + [-0.2370, 0.2043, 0.0354, ..., 0.2709, -0.0729, -0.1775], + [-0.2335, -0.2221, 0.2588, ..., -0.1974, -0.1666, 0.1749], + [ 0.0155, -0.4111, 0.2180, ..., 0.0592, -0.2090, -0.2058]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0176, -0.0367, -0.0283, -0.0280, -0.0076, 0.0122, 0.0089, -0.0076, + -0.0191, -0.0168], device='cuda:0'), grad: tensor([ 1.1642e-09, 2.9104e-09, -3.7253e-09, 4.6566e-10, 3.4925e-10, + 5.8208e-10, 2.3283e-10, 9.3132e-10, 1.1642e-10, 2.3283e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 221.21, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.2628, -0.3231, 0.1301, ..., -0.1676, 0.0467, 0.0490], + [-0.1594, -0.0905, -0.1250, ..., -0.2699, -0.0727, -0.0256], + [ 0.0075, -0.2090, -0.2798, ..., -0.2055, 0.0302, -0.4505], + ..., + [-0.2370, 0.2043, 0.0354, ..., 0.2709, -0.0729, -0.1776], + [-0.2336, -0.2221, 0.2588, ..., -0.1974, -0.1666, 0.1749], + [ 0.0155, -0.4112, 0.2181, ..., 0.0593, -0.2090, -0.2059]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-09, -3.2596e-09, ..., -5.8208e-10, + 0.0000e+00, -4.3074e-09], + [ 0.0000e+00, 1.3970e-09, 3.4925e-10, ..., 6.9849e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 3.7253e-09, 9.3132e-10, ..., 1.3970e-09, + 0.0000e+00, 6.9849e-10], + ..., + [ 4.6566e-10, -5.8208e-10, 2.3283e-09, ..., -1.1642e-09, + 0.0000e+00, 5.8208e-10], + [ 0.0000e+00, 2.3283e-10, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.9791e-09], + [-5.8208e-10, 3.4925e-10, -2.4447e-09, ..., -1.9791e-09, + 0.0000e+00, 6.9849e-10]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0174, -0.0367, -0.0282, -0.0280, -0.0076, 0.0121, 0.0089, -0.0077, + -0.0193, -0.0168], device='cuda:0'), grad: tensor([-1.0827e-08, 4.8894e-09, 1.3504e-08, -2.6892e-08, 3.6089e-09, + 1.5716e-08, 4.8894e-09, 8.2655e-09, -6.6357e-09, -2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 220.80, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4348 re_mapping 0.0018 re_causal 0.0068 /// teacc 99.18 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.2628, -0.3231, 0.1302, ..., -0.1676, 0.0467, 0.0491], + [-0.1594, -0.0905, -0.1250, ..., -0.2700, -0.0727, -0.0256], + [ 0.0075, -0.2090, -0.2798, ..., -0.2055, 0.0302, -0.4506], + ..., + [-0.2370, 0.2043, 0.0354, ..., 0.2710, -0.0729, -0.1776], + [-0.2337, -0.2222, 0.2589, ..., -0.1974, -0.1666, 0.1749], + [ 0.0155, -0.4112, 0.2182, ..., 0.0593, -0.2090, -0.2060]], + device='cuda:0'), grad: tensor([[ 4.0745e-10, 0.0000e+00, 5.2387e-10, ..., 5.8208e-11, + 0.0000e+00, 4.3074e-09], + [ 5.0641e-09, 5.8208e-11, 5.8208e-09, ..., 2.3283e-10, + 0.0000e+00, 5.8208e-11], + [ 1.1642e-10, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 1.7462e-10, 0.0000e+00], + ..., + [ 5.8208e-11, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 1.7462e-10, 1.1642e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + [-1.0710e-08, 0.0000e+00, -1.2398e-08, ..., 4.6566e-10, + 0.0000e+00, 5.8208e-11]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0173, -0.0367, -0.0282, -0.0280, -0.0077, 0.0122, 0.0089, -0.0078, + -0.0193, -0.0168], device='cuda:0'), grad: tensor([ 3.8242e-08, 4.0105e-08, 2.7940e-09, 1.3446e-08, 1.5774e-08, + 3.0268e-08, -5.2736e-08, 2.2701e-09, 1.9209e-09, -8.1898e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 220.10, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4339 re_mapping 0.0018 re_causal 0.0069 /// teacc 99.21 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.2628, -0.3231, 0.1302, ..., -0.1676, 0.0467, 0.0491], + [-0.1595, -0.0905, -0.1250, ..., -0.2700, -0.0727, -0.0256], + [ 0.0075, -0.2090, -0.2798, ..., -0.2055, 0.0302, -0.4506], + ..., + [-0.2371, 0.2044, 0.0354, ..., 0.2710, -0.0729, -0.1776], + [-0.2337, -0.2222, 0.2589, ..., -0.1974, -0.1666, 0.1749], + [ 0.0154, -0.4113, 0.2182, ..., 0.0593, -0.2090, -0.2061]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.7462e-10], + [-5.8208e-11, 1.1642e-10, 5.8208e-11, ..., 1.1642e-10, + 0.0000e+00, -4.0745e-10], + [ 0.0000e+00, 5.8208e-11, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 5.8208e-11], + ..., + [ 5.8208e-11, 0.0000e+00, 3.4925e-10, ..., 1.7462e-10, + 0.0000e+00, 5.8208e-11], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 5.8208e-11, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0173, -0.0367, -0.0281, -0.0280, -0.0076, 0.0122, 0.0088, -0.0078, + -0.0193, -0.0169], device='cuda:0'), grad: tensor([ 1.8626e-09, -2.8522e-09, 5.2387e-10, 1.5716e-09, 1.6298e-09, + 1.3970e-09, 5.8208e-10, 1.9209e-09, 5.8208e-10, 3.4925e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 219.92, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4140 re_mapping 0.0018 re_causal 0.0066 /// teacc 99.21 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.2628, -0.3231, 0.1303, ..., -0.1676, 0.0467, 0.0492], + [-0.1595, -0.0905, -0.1251, ..., -0.2701, -0.0727, -0.0256], + [ 0.0075, -0.2091, -0.2799, ..., -0.2055, 0.0302, -0.4507], + ..., + [-0.2371, 0.2044, 0.0354, ..., 0.2711, -0.0729, -0.1776], + [-0.2338, -0.2222, 0.2590, ..., -0.1974, -0.1666, 0.1750], + [ 0.0155, -0.4114, 0.2183, ..., 0.0594, -0.2090, -0.2062]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 5.8208e-11, ..., 5.8208e-11, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.8208e-11, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + ..., + [ 0.0000e+00, 1.1642e-10, 5.8208e-11, ..., 1.7462e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1525e-08, 0.0000e+00, -8.2073e-09, ..., -2.6193e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0171, -0.0369, -0.0280, -0.0281, -0.0075, 0.0123, 0.0087, -0.0079, + -0.0194, -0.0169], device='cuda:0'), grad: tensor([ 2.3283e-10, 7.9744e-09, -7.1013e-09, 2.9104e-10, 3.9756e-08, + 5.2387e-10, 8.7311e-10, 2.2701e-09, 1.1642e-10, -3.6962e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 219.92, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4347 re_mapping 0.0018 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.2628, -0.3231, 0.1304, ..., -0.1676, 0.0467, 0.0492], + [-0.1595, -0.0905, -0.1251, ..., -0.2702, -0.0727, -0.0256], + [ 0.0075, -0.2091, -0.2799, ..., -0.2055, 0.0302, -0.4507], + ..., + [-0.2371, 0.2044, 0.0354, ..., 0.2712, -0.0729, -0.1777], + [-0.2339, -0.2222, 0.2590, ..., -0.1974, -0.1666, 0.1750], + [ 0.0154, -0.4114, 0.2185, ..., 0.0595, -0.2090, -0.2062]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.2387e-10, ..., 0.0000e+00, + 0.0000e+00, -1.2806e-09], + [ 1.1642e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.9104e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-11], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 1.7462e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0170, -0.0369, -0.0279, -0.0281, -0.0076, 0.0123, 0.0087, -0.0081, + -0.0195, -0.0169], device='cuda:0'), grad: tensor([-3.2596e-09, 2.6776e-09, 1.0477e-09, 9.8953e-10, -5.8790e-09, + 5.4715e-09, 5.8208e-09, 1.6298e-09, 5.8208e-10, 2.0955e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 219.98, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4091 re_mapping 0.0018 re_causal 0.0065 /// teacc 99.19 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps3', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps3/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.790001 99.040001 ... 87.045341 69.751637 +ShearY 98.809998 98.830002 ... 87.045341 65.730553 +AutoContrast 98.799995 99.190002 ... 87.045341 56.790876 +Invert 98.699997 95.540001 ... 87.045341 55.481846 +Equalize 98.000000 98.369995 ... 87.045341 68.588643 +Solarize 98.150002 98.220001 ... 87.045341 56.272267 +SolarizeAdd 98.449997 98.119995 ... 87.045341 66.457808 +Posterize 98.930000 99.049995 ... 87.045341 73.027880 +Contrast 98.979996 99.209999 ... 87.045341 68.391936 +Color 99.010002 99.250000 ... 87.045341 63.124621 +Brightness 98.940002 99.239998 ... 87.045341 67.506813 +Sharpness 99.059998 99.129997 ... 87.045341 71.598598 +NoiseSalt 99.150002 99.180000 ... 87.045341 60.771899 +NoiseGaussian 99.070000 99.250000 ... 87.045341 58.463491 +w/o do (original x) 99.250000 0.000000 ... 0.000000 71.863251 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.08 66.345267 78.791245 75.379462 85.749875 76.566463 diff --git a/Meta-causal/code-withStyleAttack/66566.error b/Meta-causal/code-withStyleAttack/66566.error new file mode 100644 index 0000000000000000000000000000000000000000..5b09d3d759b1a068653d824f401cf6b8e10ef88f --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66566.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 37: eduler: command not found diff --git a/Meta-causal/code-withStyleAttack/66566.log b/Meta-causal/code-withStyleAttack/66566.log new file mode 100644 index 0000000000000000000000000000000000000000..ce0c6762af4e08207b8326369013efffcd186858 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66566.log @@ -0,0 +1,14134 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps4', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0040, -0.0094, 0.0302, ..., -0.0206, -0.0094, -0.0237], + [-0.0110, 0.0086, -0.0225, ..., -0.0102, 0.0209, -0.0099], + [ 0.0143, 0.0205, 0.0170, ..., 0.0053, 0.0092, -0.0274], + ..., + [-0.0212, -0.0267, 0.0230, ..., -0.0150, -0.0215, 0.0058], + [ 0.0261, 0.0148, -0.0223, ..., -0.0261, -0.0155, -0.0265], + [-0.0272, 0.0137, -0.0300, ..., -0.0267, -0.0201, 0.0062]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0209, -0.0213, -0.0202, -0.0011, -0.0063, 0.0021, 0.0005, -0.0190, + 0.0223, -0.0304], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 221.49, cls_loss 1.1811 cls_loss_mapping 1.7556 cls_loss_causal 2.2091 re_mapping 0.1709 re_causal 0.1857 /// teacc 86.97 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0059, -0.0149, 0.0355, ..., -0.0193, -0.0173, -0.0243], + [-0.0201, 0.0044, -0.0295, ..., -0.0047, 0.0274, -0.0105], + [ 0.0085, 0.0153, 0.0148, ..., 0.0009, 0.0133, -0.0280], + ..., + [-0.0297, -0.0292, 0.0219, ..., -0.0223, -0.0262, 0.0052], + [ 0.0289, 0.0163, -0.0248, ..., -0.0320, -0.0158, -0.0271], + [-0.0289, 0.0160, -0.0292, ..., -0.0333, -0.0233, 0.0056]], + device='cuda:0'), grad: tensor([[ 2.6306e-02, 1.6357e-02, 8.7128e-03, ..., 1.2924e-02, + 8.2092e-03, 0.0000e+00], + [ 1.0490e-02, 7.7400e-03, 9.6416e-04, ..., -1.1187e-03, + -3.4065e-03, 0.0000e+00], + [ 1.1681e-02, 8.8882e-03, 5.1231e-03, ..., 7.8278e-03, + -1.0071e-02, 0.0000e+00], + ..., + [ 4.7607e-03, 2.8610e-03, -4.9114e-05, ..., 2.0638e-03, + 9.2621e-03, 0.0000e+00], + [-2.2049e-02, -2.0157e-02, 1.7044e-02, ..., 1.4984e-02, + -9.9792e-03, 0.0000e+00], + [ 3.1281e-02, 3.7262e-02, 1.1818e-02, ..., 3.8681e-03, + 4.2236e-02, 0.0000e+00]], device='cuda:0') +Epoch 2, bias, value: tensor([-0.0229, -0.0193, -0.0201, -0.0019, -0.0060, 0.0019, 0.0006, -0.0186, + 0.0216, -0.0302], device='cuda:0'), grad: tensor([ 0.0220, 0.0046, 0.0021, -0.0116, -0.0584, -0.0240, 0.0173, 0.0075, + -0.0140, 0.0545], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 221.15, cls_loss 0.3725 cls_loss_mapping 0.7320 cls_loss_causal 1.9172 re_mapping 0.2041 re_causal 0.2720 /// teacc 92.86 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0051, -0.0169, 0.0385, ..., -0.0202, -0.0192, -0.0243], + [-0.0251, 0.0003, -0.0315, ..., -0.0027, 0.0308, -0.0105], + [ 0.0056, 0.0118, 0.0123, ..., -0.0026, 0.0137, -0.0280], + ..., + [-0.0343, -0.0323, 0.0211, ..., -0.0256, -0.0267, 0.0052], + [ 0.0316, 0.0192, -0.0261, ..., -0.0355, -0.0161, -0.0271], + [-0.0273, 0.0172, -0.0305, ..., -0.0338, -0.0289, 0.0056]], + device='cuda:0'), grad: tensor([[ 0.0027, 0.0011, -0.0009, ..., 0.0028, 0.0048, 0.0000], + [ 0.0039, 0.0006, 0.0018, ..., -0.0023, 0.0010, 0.0000], + [ 0.0022, 0.0011, -0.0193, ..., -0.0172, -0.0197, 0.0000], + ..., + [-0.0198, -0.0021, -0.0063, ..., -0.0084, -0.0246, 0.0000], + [-0.0043, -0.0004, -0.0155, ..., 0.0024, 0.0054, 0.0000], + [ 0.0059, 0.0017, 0.0037, ..., 0.0040, 0.0193, 0.0000]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0236, -0.0189, -0.0203, -0.0018, -0.0057, 0.0022, 0.0002, -0.0187, + 0.0214, -0.0298], device='cuda:0'), grad: tensor([ 0.0042, 0.0059, -0.0139, 0.0147, -0.0131, 0.0050, 0.0282, -0.0587, + 0.0051, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 220.46, cls_loss 0.2282 cls_loss_mapping 0.4342 cls_loss_causal 1.7166 re_mapping 0.1476 re_causal 0.2397 /// teacc 95.14 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0039, -0.0186, 0.0403, ..., -0.0204, -0.0206, -0.0226], + [-0.0289, -0.0026, -0.0339, ..., -0.0020, 0.0322, -0.0111], + [ 0.0035, 0.0092, 0.0097, ..., -0.0051, 0.0145, -0.0295], + ..., + [-0.0380, -0.0350, 0.0210, ..., -0.0295, -0.0281, 0.0046], + [ 0.0334, 0.0209, -0.0272, ..., -0.0378, -0.0154, -0.0286], + [-0.0253, 0.0186, -0.0309, ..., -0.0327, -0.0322, 0.0049]], + device='cuda:0'), grad: tensor([[ 0.0005, -0.0003, -0.0033, ..., -0.0003, 0.0016, 0.0000], + [ 0.0042, 0.0006, 0.0002, ..., -0.0022, -0.0005, 0.0000], + [ 0.0077, 0.0024, 0.0012, ..., 0.0001, -0.0306, 0.0000], + ..., + [-0.0072, 0.0011, -0.0024, ..., -0.0040, -0.0245, 0.0000], + [ 0.0044, 0.0008, 0.0014, ..., 0.0057, 0.0084, 0.0000], + [ 0.0082, 0.0037, 0.0030, ..., 0.0020, 0.0066, 0.0000]], + device='cuda:0') +Epoch 4, bias, value: tensor([-0.0234, -0.0191, -0.0204, -0.0016, -0.0054, 0.0018, -0.0001, -0.0190, + 0.0216, -0.0293], device='cuda:0'), grad: tensor([ 0.0003, 0.0037, -0.0182, 0.0059, 0.0154, 0.0054, 0.0099, -0.0401, + 0.0058, 0.0119], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 220.48, cls_loss 0.1618 cls_loss_mapping 0.2946 cls_loss_causal 1.5369 re_mapping 0.1186 re_causal 0.2086 /// teacc 96.21 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 0.0028, -0.0204, 0.0418, ..., -0.0201, -0.0215, -0.0226], + [-0.0315, -0.0050, -0.0356, ..., -0.0015, 0.0337, -0.0111], + [ 0.0019, 0.0073, 0.0071, ..., -0.0070, 0.0153, -0.0295], + ..., + [-0.0411, -0.0379, 0.0215, ..., -0.0324, -0.0293, 0.0046], + [ 0.0343, 0.0216, -0.0282, ..., -0.0405, -0.0148, -0.0286], + [-0.0241, 0.0204, -0.0316, ..., -0.0315, -0.0349, 0.0049]], + device='cuda:0'), grad: tensor([[ 0.0016, 0.0008, -0.0007, ..., 0.0004, 0.0015, 0.0000], + [ 0.0161, 0.0047, 0.0002, ..., 0.0205, 0.0262, 0.0000], + [ 0.0045, 0.0022, 0.0008, ..., 0.0025, -0.0004, 0.0000], + ..., + [ 0.0033, 0.0022, -0.0006, ..., 0.0015, 0.0033, 0.0000], + [-0.0151, -0.0074, 0.0010, ..., 0.0020, -0.0120, 0.0000], + [-0.0119, -0.0036, -0.0021, ..., -0.0256, -0.0225, 0.0000]], + device='cuda:0') +Epoch 5, bias, value: tensor([-0.0233, -0.0191, -0.0205, -0.0011, -0.0055, 0.0013, -0.0003, -0.0192, + 0.0218, -0.0292], device='cuda:0'), grad: tensor([ 0.0020, 0.0303, 0.0024, 0.0002, 0.0002, 0.0045, 0.0012, 0.0024, + -0.0155, -0.0276], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 220.71, cls_loss 0.1348 cls_loss_mapping 0.2192 cls_loss_causal 1.3486 re_mapping 0.0959 re_causal 0.1773 /// teacc 96.58 lr 0.00010000 +Epoch 6, weight, value: tensor([[ 0.0018, -0.0221, 0.0435, ..., -0.0193, -0.0221, -0.0226], + [-0.0335, -0.0073, -0.0372, ..., -0.0017, 0.0345, -0.0111], + [ 0.0004, 0.0058, 0.0045, ..., -0.0090, 0.0159, -0.0295], + ..., + [-0.0435, -0.0400, 0.0219, ..., -0.0360, -0.0299, 0.0046], + [ 0.0352, 0.0222, -0.0287, ..., -0.0425, -0.0146, -0.0286], + [-0.0227, 0.0220, -0.0320, ..., -0.0301, -0.0373, 0.0049]], + device='cuda:0'), grad: tensor([[-2.3353e-04, -4.9734e-04, -5.7411e-03, ..., -4.4746e-03, + 3.0708e-03, 0.0000e+00], + [ 3.3379e-04, -6.4492e-05, 1.0462e-03, ..., -2.3537e-03, + -6.3400e-03, 0.0000e+00], + [-9.6893e-04, 1.0271e-03, 2.2221e-03, ..., 2.7828e-03, + -9.8953e-03, 0.0000e+00], + ..., + [ 1.4925e-03, 1.2484e-03, 1.0242e-03, ..., 2.5997e-03, + 3.2539e-03, 0.0000e+00], + [ 6.2408e-03, 4.3259e-03, 7.3586e-03, ..., 7.3776e-03, + 4.5242e-03, 0.0000e+00], + [ 4.3602e-03, 3.0956e-03, 4.9934e-03, ..., 4.8294e-03, + 3.3417e-03, 0.0000e+00]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0229, -0.0194, -0.0205, -0.0011, -0.0053, 0.0009, -0.0006, -0.0191, + 0.0219, -0.0290], device='cuda:0'), grad: tensor([-0.0039, -0.0034, -0.0064, 0.0164, 0.0041, -0.0250, -0.0048, 0.0037, + 0.0114, 0.0078], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 220.42, cls_loss 0.1034 cls_loss_mapping 0.1745 cls_loss_causal 1.3033 re_mapping 0.0769 re_causal 0.1612 /// teacc 97.17 lr 0.00010000 +Epoch 7, weight, value: tensor([[ 0.0011, -0.0231, 0.0447, ..., -0.0190, -0.0228, -0.0226], + [-0.0352, -0.0092, -0.0395, ..., -0.0013, 0.0356, -0.0111], + [-0.0004, 0.0045, 0.0025, ..., -0.0106, 0.0161, -0.0295], + ..., + [-0.0454, -0.0416, 0.0218, ..., -0.0386, -0.0314, 0.0046], + [ 0.0360, 0.0229, -0.0293, ..., -0.0445, -0.0144, -0.0286], + [-0.0218, 0.0229, -0.0326, ..., -0.0288, -0.0391, 0.0049]], + device='cuda:0'), grad: tensor([[-0.0028, -0.0014, -0.0095, ..., -0.0036, 0.0013, 0.0000], + [ 0.0022, 0.0013, 0.0005, ..., -0.0169, -0.0372, 0.0000], + [ 0.0011, 0.0008, 0.0011, ..., 0.0135, 0.0279, 0.0000], + ..., + [ 0.0044, 0.0025, 0.0029, ..., 0.0035, 0.0023, 0.0000], + [ 0.0002, -0.0003, 0.0021, ..., 0.0011, -0.0003, 0.0000], + [-0.0098, -0.0056, 0.0032, ..., -0.0070, -0.0045, 0.0000]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0230, -0.0192, -0.0205, -0.0010, -0.0052, 0.0009, -0.0008, -0.0191, + 0.0219, -0.0291], device='cuda:0'), grad: tensor([-6.0272e-03, -2.5040e-02, 2.4506e-02, 8.4229e-03, 3.5381e-03, + 2.7802e-02, -2.3956e-02, 7.2122e-05, 1.7719e-03, -1.1093e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 220.70, cls_loss 0.0983 cls_loss_mapping 0.1655 cls_loss_causal 1.2166 re_mapping 0.0664 re_causal 0.1403 /// teacc 97.26 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0001, -0.0240, 0.0463, ..., -0.0182, -0.0236, -0.0233], + [-0.0376, -0.0115, -0.0411, ..., -0.0013, 0.0364, -0.0123], + [-0.0013, 0.0034, 0.0009, ..., -0.0117, 0.0166, -0.0289], + ..., + [-0.0473, -0.0434, 0.0220, ..., -0.0411, -0.0325, 0.0017], + [ 0.0367, 0.0229, -0.0305, ..., -0.0462, -0.0136, -0.0293], + [-0.0206, 0.0243, -0.0334, ..., -0.0282, -0.0404, 0.0024]], + device='cuda:0'), grad: tensor([[-1.0185e-03, 4.8542e-04, -1.6375e-03, ..., -4.5151e-05, + -4.9591e-04, 0.0000e+00], + [ 2.7275e-04, 1.0467e-04, 2.3186e-04, ..., -5.3263e-04, + -1.2283e-03, 0.0000e+00], + [-1.3094e-03, -1.0309e-03, 1.9112e-03, ..., 1.8978e-03, + -2.8658e-04, 0.0000e+00], + ..., + [ 1.3523e-03, 7.9155e-04, 4.6563e-04, ..., 1.1129e-03, + 3.7193e-04, 0.0000e+00], + [-1.6205e-02, -4.5633e-04, -2.5970e-02, ..., -2.4063e-02, + -9.8495e-03, 0.0000e+00], + [-5.8794e-04, -6.4421e-04, -1.4693e-05, ..., -3.3498e-04, + 2.4104e-04, 0.0000e+00]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0229, -0.0194, -0.0202, -0.0007, -0.0054, 0.0007, -0.0011, -0.0190, + 0.0219, -0.0291], device='cuda:0'), grad: tensor([-0.0032, -0.0006, -0.0001, 0.0026, 0.0008, -0.0003, 0.0241, 0.0014, + -0.0242, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 220.54, cls_loss 0.0731 cls_loss_mapping 0.1247 cls_loss_causal 1.1419 re_mapping 0.0608 re_causal 0.1358 /// teacc 97.77 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0005, -0.0251, 0.0478, ..., -0.0176, -0.0243, -0.0233], + [-0.0392, -0.0131, -0.0422, ..., -0.0003, 0.0371, -0.0123], + [-0.0023, 0.0021, -0.0004, ..., -0.0131, 0.0170, -0.0289], + ..., + [-0.0492, -0.0449, 0.0222, ..., -0.0434, -0.0330, 0.0017], + [ 0.0375, 0.0234, -0.0311, ..., -0.0480, -0.0135, -0.0293], + [-0.0202, 0.0248, -0.0345, ..., -0.0276, -0.0414, 0.0024]], + device='cuda:0'), grad: tensor([[ 5.4407e-04, 5.0306e-04, 2.9011e-03, ..., 2.4033e-03, + 8.0252e-04, 0.0000e+00], + [ 3.1757e-04, 3.6359e-04, 4.2772e-04, ..., 7.0152e-03, + 1.4442e-02, 0.0000e+00], + [ 3.7026e-04, 2.8443e-04, 3.6454e-04, ..., 3.0975e-03, + 5.7449e-03, 0.0000e+00], + ..., + [ 9.2697e-04, 7.9441e-04, 8.6352e-06, ..., 8.2588e-04, + 1.1635e-03, 0.0000e+00], + [ 4.6425e-03, 4.4708e-03, 7.5722e-04, ..., 2.7981e-03, + 1.9407e-03, 0.0000e+00], + [-1.7151e-02, -1.6342e-02, 5.4550e-04, ..., -6.4163e-03, + 1.0757e-03, 0.0000e+00]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0226, -0.0195, -0.0201, -0.0007, -0.0056, 0.0008, -0.0014, -0.0189, + 0.0219, -0.0291], device='cuda:0'), grad: tensor([ 0.0022, 0.0116, 0.0048, 0.0102, -0.0271, 0.0025, 0.0018, 0.0009, + 0.0057, -0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 8, time 220.08, cls_loss 0.0763 cls_loss_mapping 0.1265 cls_loss_causal 1.1270 re_mapping 0.0519 re_causal 0.1186 /// teacc 97.55 lr 0.00010000 +Epoch 10, weight, value: tensor([[-9.4165e-04, -2.5567e-02, 4.8835e-02, ..., -1.7152e-02, + -2.4587e-02, -2.3264e-02], + [-4.0173e-02, -1.4588e-02, -4.3807e-02, ..., 4.4619e-05, + 3.7624e-02, -1.2302e-02], + [-3.2320e-03, 8.7407e-04, -1.5643e-03, ..., -1.4620e-02, + 1.7067e-02, -2.8891e-02], + ..., + [-5.0981e-02, -4.6098e-02, 2.2313e-02, ..., -4.5784e-02, + -3.3669e-02, 1.6602e-03], + [ 3.8122e-02, 2.3588e-02, -3.1482e-02, ..., -4.9262e-02, + -1.3223e-02, -2.9330e-02], + [-1.9152e-02, 2.5907e-02, -3.5265e-02, ..., -2.6991e-02, + -4.2645e-02, 2.4479e-03]], device='cuda:0'), grad: tensor([[ 3.4690e-04, 8.0645e-05, -6.7234e-05, ..., 1.3304e-04, + 1.0986e-03, 0.0000e+00], + [ 4.7541e-04, 2.3961e-04, 8.1122e-05, ..., -2.2626e-04, + 5.9724e-05, 0.0000e+00], + [-6.6996e-05, 1.3657e-03, -7.4339e-04, ..., -1.0529e-03, + -8.4229e-03, 0.0000e+00], + ..., + [ 3.8218e-04, 1.6356e-04, 7.1406e-05, ..., 3.2926e-04, + 1.6918e-03, 0.0000e+00], + [-2.2769e-05, -3.2234e-04, 4.1485e-04, ..., 9.8991e-04, + 2.2984e-03, 0.0000e+00], + [ 9.8896e-04, 5.9509e-04, 1.8859e-04, ..., 4.8280e-04, + 1.1196e-03, 0.0000e+00]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0225, -0.0193, -0.0200, -0.0006, -0.0056, 0.0005, -0.0015, -0.0190, + 0.0220, -0.0291], device='cuda:0'), grad: tensor([ 0.0009, 0.0004, -0.0106, 0.0010, -0.0003, -0.0001, 0.0006, 0.0043, + 0.0021, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 220.80, cls_loss 0.0687 cls_loss_mapping 0.1145 cls_loss_causal 1.0487 re_mapping 0.0474 re_causal 0.1112 /// teacc 97.84 lr 0.00010000 +Epoch 11, weight, value: tensor([[-1.2253e-03, -2.6097e-02, 4.9519e-02, ..., -1.6645e-02, + -2.5348e-02, -2.3871e-02], + [-4.1476e-02, -1.6130e-02, -4.5295e-02, ..., -5.3146e-05, + 3.7956e-02, -2.1434e-02], + [-4.0241e-03, 3.6048e-04, -2.7722e-03, ..., -1.5528e-02, + 1.7524e-02, -3.3136e-02], + ..., + [-5.2076e-02, -4.6983e-02, 2.3109e-02, ..., -4.7417e-02, + -3.4181e-02, -1.5165e-02], + [ 3.9169e-02, 2.4468e-02, -3.1636e-02, ..., -4.9980e-02, + -1.3242e-02, -3.5135e-02], + [-1.9152e-02, 2.5997e-02, -3.5879e-02, ..., -2.6443e-02, + -4.3814e-02, -4.1340e-03]], device='cuda:0'), grad: tensor([[ 2.4152e-04, 2.7752e-04, 1.4806e-04, ..., 1.7405e-04, + 1.4257e-04, 0.0000e+00], + [ 1.1766e-04, 5.6326e-05, 1.2493e-04, ..., -4.5896e-04, + -8.8215e-04, 0.0000e+00], + [-2.7704e-04, -1.1015e-04, 1.5104e-04, ..., 1.5128e-04, + -5.6028e-04, 0.0000e+00], + ..., + [ 1.3933e-03, 7.6342e-04, -8.6367e-05, ..., 8.2207e-04, + 6.4945e-04, 0.0000e+00], + [ 1.2293e-03, 7.7343e-04, 1.6136e-03, ..., 1.1549e-03, + 1.0061e-03, 0.0000e+00], + [-2.7409e-03, -1.3304e-03, 6.4421e-04, ..., -1.3552e-03, + 1.0281e-03, 0.0000e+00]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0226, -0.0195, -0.0199, -0.0005, -0.0054, 0.0003, -0.0019, -0.0188, + 0.0223, -0.0294], device='cuda:0'), grad: tensor([ 0.0004, -0.0006, -0.0010, 0.0026, -0.0018, 0.0035, -0.0063, 0.0014, + 0.0031, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 220.68, cls_loss 0.0652 cls_loss_mapping 0.1110 cls_loss_causal 1.0764 re_mapping 0.0428 re_causal 0.1062 /// teacc 98.16 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0013, -0.0264, 0.0505, ..., -0.0159, -0.0257, -0.0224], + [-0.0427, -0.0177, -0.0466, ..., 0.0002, 0.0382, -0.0246], + [-0.0044, -0.0005, -0.0040, ..., -0.0162, 0.0179, -0.0361], + ..., + [-0.0534, -0.0480, 0.0234, ..., -0.0489, -0.0349, -0.0202], + [ 0.0396, 0.0245, -0.0319, ..., -0.0514, -0.0129, -0.0334], + [-0.0187, 0.0266, -0.0362, ..., -0.0255, -0.0449, -0.0064]], + device='cuda:0'), grad: tensor([[ 1.3609e-03, 3.6192e-04, 9.6893e-04, ..., 1.0834e-03, + 8.5640e-04, 2.7637e-07], + [ 3.5238e-04, 2.3127e-04, 8.7619e-05, ..., -3.3879e-04, + -5.9175e-04, 1.2945e-07], + [-7.9117e-03, -1.4210e-03, -3.1910e-03, ..., -3.7098e-03, + -4.2458e-03, 3.2550e-07], + ..., + [ 8.2159e-04, 3.5214e-04, 3.5048e-04, ..., 5.7364e-04, + 6.9666e-04, 5.9418e-07], + [-8.5640e-04, -6.0844e-04, 4.7159e-04, ..., 6.2656e-04, + 1.6820e-04, 6.2108e-08], + [ 3.0842e-03, 1.5202e-03, 1.3056e-03, ..., 1.5068e-03, + 1.5974e-03, 6.6161e-06]], device='cuda:0') +Epoch 12, bias, value: tensor([-2.2334e-02, -1.9536e-02, -2.0071e-02, -3.8515e-04, -5.6846e-03, + 3.5607e-05, -1.7969e-03, -1.8514e-02, 2.2573e-02, -2.9588e-02], + device='cuda:0'), grad: tensor([ 2.3251e-03, -6.2180e-04, -1.4931e-02, 1.2760e-03, -3.3998e-04, + 3.7785e-03, -3.5429e-04, 5.4970e-03, -6.6698e-05, 3.4428e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 220.95, cls_loss 0.0462 cls_loss_mapping 0.0811 cls_loss_causal 1.0166 re_mapping 0.0395 re_causal 0.1034 /// teacc 98.17 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0019, -0.0270, 0.0510, ..., -0.0158, -0.0261, -0.0226], + [-0.0436, -0.0188, -0.0481, ..., 0.0002, 0.0384, -0.0248], + [-0.0050, -0.0010, -0.0049, ..., -0.0170, 0.0181, -0.0367], + ..., + [-0.0541, -0.0491, 0.0235, ..., -0.0509, -0.0355, -0.0226], + [ 0.0400, 0.0246, -0.0313, ..., -0.0525, -0.0127, -0.0327], + [-0.0183, 0.0273, -0.0364, ..., -0.0242, -0.0459, -0.0077]], + device='cuda:0'), grad: tensor([[-6.1560e-04, -1.0815e-03, -1.6575e-03, ..., -1.7757e-03, + 3.1877e-04, 0.0000e+00], + [ 5.0831e-04, 4.1842e-04, 2.2471e-04, ..., 9.8705e-04, + 2.8648e-03, 0.0000e+00], + [ 2.9540e-04, 2.4724e-04, 4.6229e-04, ..., 6.0701e-04, + 5.4264e-04, 0.0000e+00], + ..., + [ 4.2725e-04, 3.5930e-04, 9.1195e-05, ..., -1.2445e-03, + -3.0689e-03, 0.0000e+00], + [-4.6539e-03, -2.1992e-03, -2.2621e-03, ..., 5.4657e-05, + -2.2011e-03, 0.0000e+00], + [ 1.2264e-03, 2.2488e-03, 5.3263e-04, ..., 1.2379e-03, + 6.3858e-03, 0.0000e+00]], device='cuda:0') +Epoch 13, bias, value: tensor([-2.2753e-02, -1.9538e-02, -2.0149e-02, -2.3593e-04, -5.7608e-03, + -5.2678e-05, -2.1927e-03, -1.8378e-02, 2.2796e-02, -2.9327e-02], + device='cuda:0'), grad: tensor([-0.0018, 0.0146, 0.0014, 0.0028, -0.0070, 0.0043, -0.0011, -0.0174, + -0.0048, 0.0091], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 220.04, cls_loss 0.0487 cls_loss_mapping 0.0823 cls_loss_causal 0.9818 re_mapping 0.0359 re_causal 0.0944 /// teacc 98.11 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0021, -0.0273, 0.0519, ..., -0.0150, -0.0268, -0.0229], + [-0.0439, -0.0198, -0.0489, ..., 0.0006, 0.0385, -0.0236], + [-0.0053, -0.0005, -0.0059, ..., -0.0176, 0.0183, -0.0347], + ..., + [-0.0552, -0.0504, 0.0238, ..., -0.0524, -0.0365, -0.0247], + [ 0.0405, 0.0245, -0.0318, ..., -0.0535, -0.0124, -0.0332], + [-0.0180, 0.0276, -0.0374, ..., -0.0239, -0.0469, -0.0091]], + device='cuda:0'), grad: tensor([[ 2.0540e-04, 1.4937e-04, -2.9421e-04, ..., -1.9979e-04, + 8.6904e-05, 0.0000e+00], + [ 5.5075e-04, 3.9172e-04, 1.5998e-04, ..., -1.0079e-04, + -3.5548e-04, 0.0000e+00], + [ 5.3101e-03, 2.7905e-03, 3.1447e-04, ..., 9.9373e-04, + 2.2316e-03, 0.0000e+00], + ..., + [ 1.8330e-03, 1.7328e-03, -1.2436e-03, ..., 2.5511e-04, + -1.5473e-04, 0.0000e+00], + [ 3.3493e-03, 1.9817e-03, 3.0947e-04, ..., 8.8692e-04, + 1.4973e-03, 0.0000e+00], + [-2.2221e-03, -2.3289e-03, 3.8052e-04, ..., -1.3566e-04, + 3.1090e-04, 0.0000e+00]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0228, -0.0195, -0.0197, -0.0004, -0.0058, -0.0001, -0.0020, -0.0187, + 0.0230, -0.0295], device='cuda:0'), grad: tensor([ 0.0002, 0.0008, 0.0065, -0.0099, 0.0015, -0.0001, 0.0006, 0.0042, + 0.0054, -0.0092], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 220.79, cls_loss 0.0490 cls_loss_mapping 0.0800 cls_loss_causal 0.9607 re_mapping 0.0330 re_causal 0.0881 /// teacc 98.24 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0024, -0.0276, 0.0527, ..., -0.0144, -0.0273, -0.0230], + [-0.0445, -0.0208, -0.0498, ..., 0.0008, 0.0387, -0.0167], + [-0.0062, -0.0016, -0.0067, ..., -0.0187, 0.0187, -0.0350], + ..., + [-0.0562, -0.0513, 0.0236, ..., -0.0539, -0.0371, -0.0278], + [ 0.0412, 0.0248, -0.0319, ..., -0.0546, -0.0125, -0.0327], + [-0.0176, 0.0281, -0.0379, ..., -0.0234, -0.0477, -0.0097]], + device='cuda:0'), grad: tensor([[-6.3229e-04, -4.2343e-04, -1.3518e-04, ..., -8.2588e-04, + 6.1941e-04, 7.3791e-05], + [ 4.0245e-04, 4.0317e-04, 2.9516e-04, ..., 3.5620e-04, + 1.1177e-03, 8.9258e-06], + [ 3.8218e-04, 2.9707e-04, 4.6372e-04, ..., 2.4152e-04, + -1.7433e-03, 1.0580e-04], + ..., + [ 8.4925e-04, 6.9046e-04, 5.5313e-04, ..., 2.7800e-04, + 2.0847e-03, 7.0155e-05], + [-6.2609e-04, -1.7929e-04, 3.7074e-04, ..., 4.6468e-04, + 4.5955e-05, 7.7426e-05], + [-7.5388e-04, -5.0354e-04, 7.4673e-04, ..., -1.5080e-05, + 9.8228e-04, 2.7001e-05]], device='cuda:0') +Epoch 15, bias, value: tensor([-2.2635e-02, -1.9713e-02, -1.9907e-02, 6.2496e-05, -5.5963e-03, + -4.7145e-05, -2.5151e-03, -1.8588e-02, 2.3083e-02, -2.9806e-02], + device='cuda:0'), grad: tensor([-0.0006, 0.0018, -0.0047, -0.0016, -0.0053, 0.0035, 0.0018, 0.0038, + 0.0005, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 220.77, cls_loss 0.0470 cls_loss_mapping 0.0782 cls_loss_causal 0.9508 re_mapping 0.0305 re_causal 0.0856 /// teacc 98.27 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0028, -0.0280, 0.0537, ..., -0.0136, -0.0275, -0.0224], + [-0.0451, -0.0222, -0.0504, ..., 0.0013, 0.0392, -0.0074], + [-0.0065, -0.0018, -0.0081, ..., -0.0193, 0.0185, -0.0370], + ..., + [-0.0570, -0.0522, 0.0235, ..., -0.0549, -0.0376, -0.0314], + [ 0.0416, 0.0249, -0.0328, ..., -0.0556, -0.0124, -0.0336], + [-0.0171, 0.0287, -0.0378, ..., -0.0229, -0.0485, -0.0103]], + device='cuda:0'), grad: tensor([[ 3.2574e-05, -1.2531e-03, -4.5052e-03, ..., -1.3885e-03, + 1.8859e-04, 2.9430e-06], + [ 5.1022e-05, -3.6073e-04, -3.5524e-04, ..., -1.3313e-03, + -1.1539e-03, -2.3091e-04], + [-8.1182e-05, 6.4194e-05, 1.6189e-04, ..., 1.2732e-04, + -8.9407e-04, 5.5097e-06], + ..., + [-1.5044e-04, 9.6858e-05, 1.7798e-04, ..., 1.2159e-04, + 4.6039e-04, 5.8748e-06], + [ 7.4029e-05, 2.7657e-04, 5.4932e-04, ..., 1.0929e-03, + 1.0138e-03, 1.5080e-04], + [ 1.4961e-04, 1.1057e-04, 2.7895e-04, ..., 1.5247e-04, + 1.7750e-04, 3.8221e-06]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0222, -0.0196, -0.0200, -0.0001, -0.0058, 0.0003, -0.0031, -0.0183, + 0.0230, -0.0299], device='cuda:0'), grad: tensor([-0.0025, -0.0011, -0.0007, 0.0014, 0.0014, 0.0013, 0.0004, -0.0044, + 0.0024, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 220.62, cls_loss 0.0423 cls_loss_mapping 0.0752 cls_loss_causal 0.9295 re_mapping 0.0297 re_causal 0.0820 /// teacc 98.43 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0033, -0.0283, 0.0536, ..., -0.0134, -0.0277, -0.0228], + [-0.0458, -0.0238, -0.0515, ..., 0.0009, 0.0393, 0.0002], + [-0.0070, -0.0025, -0.0090, ..., -0.0199, 0.0188, -0.0384], + ..., + [-0.0579, -0.0534, 0.0234, ..., -0.0561, -0.0381, -0.0344], + [ 0.0422, 0.0254, -0.0332, ..., -0.0563, -0.0123, -0.0354], + [-0.0164, 0.0295, -0.0376, ..., -0.0216, -0.0489, -0.0108]], + device='cuda:0'), grad: tensor([[ 1.9574e-04, -5.7220e-05, -7.7295e-04, ..., -1.5283e-04, + 2.3949e-04, 2.9523e-06], + [ 1.7571e-04, 1.3041e-04, 9.1553e-05, ..., -2.4724e-04, + -1.8585e-04, -1.4400e-04], + [-4.2114e-03, -2.6073e-03, 1.0580e-04, ..., -1.6279e-03, + -2.8687e-03, 1.4134e-05], + ..., + [ 7.8392e-04, 6.0129e-04, 9.9301e-05, ..., 6.1131e-04, + 6.0558e-04, 4.4972e-05], + [-4.2801e-03, -2.9488e-03, -8.9979e-04, ..., -7.1812e-04, + -3.3951e-03, 2.4438e-05], + [ 3.4599e-03, 2.5139e-03, 6.5422e-04, ..., 1.3523e-03, + 3.0022e-03, 1.3083e-05]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0223, -0.0199, -0.0198, -0.0004, -0.0059, 0.0002, -0.0027, -0.0182, + 0.0232, -0.0297], device='cuda:0'), grad: tensor([-6.4671e-06, -8.0884e-05, -7.5874e-03, 4.0512e-03, 2.1858e-03, + 1.1444e-03, 5.2500e-04, 1.0595e-03, -7.4692e-03, 6.1760e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 219.99, cls_loss 0.0341 cls_loss_mapping 0.0590 cls_loss_causal 0.9120 re_mapping 0.0280 re_causal 0.0806 /// teacc 98.41 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0035, -0.0284, 0.0545, ..., -0.0128, -0.0281, -0.0224], + [-0.0466, -0.0249, -0.0524, ..., 0.0011, 0.0397, 0.0093], + [-0.0077, -0.0034, -0.0099, ..., -0.0208, 0.0189, -0.0410], + ..., + [-0.0586, -0.0541, 0.0235, ..., -0.0576, -0.0387, -0.0353], + [ 0.0428, 0.0259, -0.0335, ..., -0.0569, -0.0123, -0.0379], + [-0.0161, 0.0296, -0.0385, ..., -0.0211, -0.0495, -0.0116]], + device='cuda:0'), grad: tensor([[ 4.4656e-04, 9.0170e-04, 4.4327e-03, ..., 1.8530e-03, + 2.5368e-04, 2.2560e-05], + [ 3.0231e-04, 2.5463e-04, 1.2863e-04, ..., 1.9407e-04, + 1.2267e-04, -2.4095e-05], + [ 7.1096e-04, 4.8757e-04, 4.0531e-04, ..., 6.1417e-04, + 3.5858e-04, 1.5959e-05], + ..., + [ 5.8212e-03, 4.0741e-03, 2.4676e-04, ..., 4.3259e-03, + 2.7990e-04, 9.2164e-06], + [ 5.1594e-04, 1.3838e-03, 2.2984e-03, ..., 8.2684e-04, + -1.7703e-05, 1.4871e-05], + [-9.9468e-04, -8.5497e-04, 8.9526e-05, ..., -1.0443e-03, + -2.0874e-04, 6.7912e-06]], device='cuda:0') +Epoch 18, bias, value: tensor([-2.2324e-02, -1.9855e-02, -1.9607e-02, -5.1715e-04, -5.8483e-03, + 5.3230e-05, -2.8812e-03, -1.8256e-02, 2.3400e-02, -2.9867e-02], + device='cuda:0'), grad: tensor([ 0.0038, 0.0006, 0.0007, -0.0001, 0.0012, -0.0161, -0.0005, 0.0093, + 0.0025, -0.0013], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 220.41, cls_loss 0.0330 cls_loss_mapping 0.0570 cls_loss_causal 0.8765 re_mapping 0.0265 re_causal 0.0769 /// teacc 98.55 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0036, -0.0283, 0.0553, ..., -0.0122, -0.0285, -0.0219], + [-0.0473, -0.0259, -0.0529, ..., 0.0012, 0.0397, 0.0109], + [-0.0079, -0.0040, -0.0107, ..., -0.0216, 0.0191, -0.0432], + ..., + [-0.0597, -0.0551, 0.0240, ..., -0.0588, -0.0388, -0.0367], + [ 0.0431, 0.0258, -0.0339, ..., -0.0581, -0.0121, -0.0362], + [-0.0154, 0.0303, -0.0393, ..., -0.0202, -0.0502, -0.0112]], + device='cuda:0'), grad: tensor([[ 4.0207e-03, 4.6778e-04, 1.1911e-03, ..., 2.6302e-03, + 8.7357e-04, 5.7742e-06], + [ 5.0974e-04, 5.4264e-04, 2.1696e-04, ..., 4.5586e-04, + 6.9666e-04, -1.3316e-04], + [ 6.1684e-03, 6.3133e-04, 1.1387e-03, ..., 3.3836e-03, + 2.8725e-03, 1.8016e-05], + ..., + [ 2.2352e-04, 2.2471e-04, 3.6091e-05, ..., 2.3472e-04, + 1.4007e-04, 1.0751e-05], + [-1.3565e-02, -9.7275e-05, -2.6245e-03, ..., -6.8474e-03, + -5.4073e-04, 4.1872e-05], + [ 6.5851e-04, -4.4179e-04, 8.1658e-05, ..., 3.5524e-05, + 5.1498e-04, 1.0662e-05]], device='cuda:0') +Epoch 19, bias, value: tensor([-2.2243e-02, -1.9986e-02, -1.9706e-02, -3.3146e-04, -6.0938e-03, + 6.2519e-06, -2.8947e-03, -1.7964e-02, 2.3210e-02, -2.9736e-02], + device='cuda:0'), grad: tensor([ 0.0036, 0.0016, 0.0067, -0.0080, 0.0006, 0.0063, -0.0033, -0.0003, + -0.0074, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 220.31, cls_loss 0.0276 cls_loss_mapping 0.0525 cls_loss_causal 0.8589 re_mapping 0.0260 re_causal 0.0777 /// teacc 98.57 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0043, -0.0283, 0.0560, ..., -0.0118, -0.0289, -0.0216], + [-0.0481, -0.0269, -0.0534, ..., 0.0012, 0.0398, 0.0136], + [-0.0087, -0.0047, -0.0115, ..., -0.0223, 0.0190, -0.0449], + ..., + [-0.0608, -0.0560, 0.0236, ..., -0.0601, -0.0395, -0.0381], + [ 0.0438, 0.0256, -0.0344, ..., -0.0589, -0.0118, -0.0373], + [-0.0149, 0.0309, -0.0394, ..., -0.0192, -0.0507, -0.0104]], + device='cuda:0'), grad: tensor([[-8.4734e-04, -1.8625e-03, -1.2718e-02, ..., -1.0666e-02, + -2.7504e-03, 5.1595e-06], + [ 1.5211e-04, 7.9513e-05, 7.7248e-04, ..., -5.1677e-05, + 7.5054e-04, -1.1213e-05], + [ 5.1212e-04, 2.4378e-04, 1.1797e-03, ..., 2.9135e-04, + 1.6251e-03, 2.0996e-05], + ..., + [ 4.0412e-04, -5.2564e-06, 7.8738e-05, ..., 2.1493e-04, + 1.4858e-03, 5.0068e-05], + [ 4.8071e-05, 7.1955e-04, 1.0986e-03, ..., 1.0004e-03, + 7.5483e-04, 3.4302e-05], + [ 7.5758e-05, 4.0412e-05, 5.2881e-04, ..., 2.1040e-04, + 9.8705e-04, 8.7172e-06]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0225, -0.0202, -0.0199, -0.0001, -0.0059, 0.0001, -0.0029, -0.0181, + 0.0234, -0.0297], device='cuda:0'), grad: tensor([-0.0110, 0.0011, 0.0021, -0.0034, -0.0045, 0.0026, 0.0109, -0.0005, + 0.0019, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 219.58, cls_loss 0.0282 cls_loss_mapping 0.0559 cls_loss_causal 0.8187 re_mapping 0.0248 re_causal 0.0710 /// teacc 98.40 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0045, -0.0282, 0.0564, ..., -0.0115, -0.0293, -0.0220], + [-0.0490, -0.0280, -0.0543, ..., 0.0014, 0.0399, 0.0188], + [-0.0094, -0.0054, -0.0124, ..., -0.0231, 0.0194, -0.0452], + ..., + [-0.0613, -0.0568, 0.0236, ..., -0.0611, -0.0396, -0.0402], + [ 0.0443, 0.0258, -0.0346, ..., -0.0598, -0.0117, -0.0382], + [-0.0148, 0.0309, -0.0401, ..., -0.0191, -0.0517, -0.0112]], + device='cuda:0'), grad: tensor([[ 5.8460e-04, 2.7776e-04, 7.3016e-05, ..., 1.1557e-04, + 6.0940e-04, 8.5980e-06], + [ 2.3282e-04, 9.1553e-05, 2.3067e-04, ..., 1.1605e-04, + 3.4595e-04, -5.6237e-05], + [ 1.0309e-03, 4.0483e-04, 5.7793e-04, ..., 7.0190e-04, + -3.5143e-04, 2.4870e-05], + ..., + [ 6.2895e-04, 2.0528e-04, -4.3631e-05, ..., 2.7680e-04, + 4.9925e-04, 7.3463e-06], + [-2.6646e-03, -2.3270e-03, -1.9875e-03, ..., 6.3705e-04, + -4.1885e-03, 1.5073e-05], + [-6.2408e-03, -1.1301e-03, 3.0780e-04, ..., -2.5444e-03, + -1.0719e-03, 1.0021e-06]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0227, -0.0206, -0.0194, -0.0003, -0.0059, 0.0007, -0.0029, -0.0178, + 0.0232, -0.0301], device='cuda:0'), grad: tensor([ 0.0012, 0.0008, 0.0032, 0.0101, -0.0003, -0.0016, 0.0008, -0.0026, + -0.0039, -0.0076], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 220.37, cls_loss 0.0258 cls_loss_mapping 0.0491 cls_loss_causal 0.8721 re_mapping 0.0231 re_causal 0.0714 /// teacc 98.66 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0044, -0.0281, 0.0571, ..., -0.0103, -0.0293, -0.0218], + [-0.0495, -0.0286, -0.0551, ..., 0.0016, 0.0399, 0.0250], + [-0.0100, -0.0062, -0.0132, ..., -0.0237, 0.0196, -0.0467], + ..., + [-0.0619, -0.0575, 0.0233, ..., -0.0627, -0.0400, -0.0404], + [ 0.0447, 0.0260, -0.0350, ..., -0.0607, -0.0115, -0.0408], + [-0.0147, 0.0312, -0.0403, ..., -0.0184, -0.0524, -0.0121]], + device='cuda:0'), grad: tensor([[ 3.5071e-04, 2.1338e-04, 3.4976e-04, ..., 3.7336e-04, + 2.1362e-04, 6.2212e-06], + [ 2.9397e-04, 2.0523e-03, 5.1379e-05, ..., 3.9139e-03, + 3.5210e-03, -1.1873e-04], + [ 1.0691e-03, 8.0681e-04, -1.5542e-05, ..., 1.7290e-03, + 1.1644e-03, 4.1693e-05], + ..., + [ 1.1438e-04, 2.4700e-04, 1.2386e-04, ..., 5.2595e-04, + 5.2118e-04, 2.7344e-05], + [-2.2244e-04, -2.4509e-04, -8.8692e-05, ..., 4.8113e-04, + 3.9744e-04, 5.5641e-05], + [-5.2166e-04, -2.7771e-03, -8.4579e-05, ..., -5.1155e-03, + -4.5853e-03, -7.6532e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0223, -0.0207, -0.0192, -0.0005, -0.0055, 0.0003, -0.0031, -0.0179, + 0.0231, -0.0300], device='cuda:0'), grad: tensor([ 0.0010, 0.0183, 0.0026, -0.0029, 0.0024, 0.0006, -0.0012, 0.0020, + 0.0008, -0.0236], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 219.77, cls_loss 0.0251 cls_loss_mapping 0.0487 cls_loss_causal 0.8708 re_mapping 0.0228 re_causal 0.0673 /// teacc 98.60 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0044, -0.0281, 0.0576, ..., -0.0104, -0.0298, -0.0221], + [-0.0500, -0.0291, -0.0558, ..., 0.0020, 0.0403, 0.0306], + [-0.0106, -0.0071, -0.0140, ..., -0.0248, 0.0193, -0.0485], + ..., + [-0.0628, -0.0584, 0.0235, ..., -0.0639, -0.0403, -0.0415], + [ 0.0453, 0.0264, -0.0354, ..., -0.0612, -0.0115, -0.0426], + [-0.0143, 0.0317, -0.0407, ..., -0.0175, -0.0530, -0.0128]], + device='cuda:0'), grad: tensor([[ 9.3460e-05, 1.4055e-04, 1.0866e-04, ..., 6.3181e-05, + 9.5844e-05, 1.9476e-05], + [ 7.1287e-05, 6.5684e-05, -7.5054e-04, ..., -1.6508e-03, + -2.8992e-03, -1.1339e-03], + [ 9.6226e-04, 6.8331e-04, 1.5152e-04, ..., 5.2738e-04, + 7.9250e-04, 1.0085e-04], + ..., + [ 1.1511e-06, -4.3392e-04, 1.0639e-05, ..., 5.2071e-04, + 1.0157e-03, 2.5582e-04], + [ 4.6730e-04, 3.4666e-04, 1.5450e-04, ..., 5.4216e-04, + 4.7803e-04, 2.1064e-04], + [-2.3282e-04, 9.0957e-05, 2.7728e-04, ..., 2.7943e-04, + 5.2500e-04, 1.1879e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0227, -0.0208, -0.0195, -0.0002, -0.0057, 0.0005, -0.0032, -0.0179, + 0.0232, -0.0298], device='cuda:0'), grad: tensor([ 0.0005, -0.0045, 0.0022, -0.0001, 0.0019, -0.0008, -0.0005, -0.0014, + 0.0017, 0.0010], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 220.33, cls_loss 0.0211 cls_loss_mapping 0.0446 cls_loss_causal 0.7895 re_mapping 0.0229 re_causal 0.0676 /// teacc 98.67 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0043, -0.0278, 0.0582, ..., -0.0103, -0.0301, -0.0222], + [-0.0504, -0.0302, -0.0564, ..., 0.0027, 0.0404, 0.0342], + [-0.0112, -0.0081, -0.0146, ..., -0.0255, 0.0196, -0.0505], + ..., + [-0.0633, -0.0586, 0.0235, ..., -0.0648, -0.0408, -0.0428], + [ 0.0456, 0.0266, -0.0353, ..., -0.0621, -0.0113, -0.0422], + [-0.0141, 0.0321, -0.0411, ..., -0.0165, -0.0536, -0.0135]], + device='cuda:0'), grad: tensor([[ 6.8188e-05, 1.6257e-05, -6.4552e-05, ..., -9.7156e-05, + 5.8830e-05, 2.5630e-06], + [ 7.0393e-05, 8.8453e-05, 5.4419e-05, ..., 7.2956e-05, + 7.0930e-05, -3.2336e-05], + [ 9.5069e-05, 6.6996e-05, 8.1301e-05, ..., 1.4818e-04, + -2.8181e-04, 5.8487e-06], + ..., + [ 9.3031e-04, 8.4114e-04, 5.3120e-04, ..., 1.6403e-03, + 1.4868e-03, 7.1786e-06], + [ 1.1110e-03, 1.1539e-03, 3.7837e-04, ..., 1.0824e-03, + 1.7929e-04, 1.6689e-05], + [-1.0214e-03, -8.4209e-04, 1.0431e-04, ..., -4.0078e-04, + 3.0208e-04, 2.9039e-06]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0229, -0.0209, -0.0194, -0.0001, -0.0060, 0.0004, -0.0033, -0.0176, + 0.0233, -0.0297], device='cuda:0'), grad: tensor([-7.1414e-06, 3.7909e-04, -1.7309e-03, -1.6060e-03, -1.7719e-03, + -9.1314e-04, 2.7671e-05, 4.1695e-03, 1.9064e-03, -4.5228e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 219.60, cls_loss 0.0213 cls_loss_mapping 0.0442 cls_loss_causal 0.8028 re_mapping 0.0213 re_causal 0.0642 /// teacc 98.54 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0046, -0.0275, 0.0588, ..., -0.0096, -0.0305, -0.0225], + [-0.0507, -0.0307, -0.0569, ..., 0.0033, 0.0405, 0.0384], + [-0.0116, -0.0089, -0.0152, ..., -0.0261, 0.0196, -0.0516], + ..., + [-0.0640, -0.0592, 0.0237, ..., -0.0660, -0.0407, -0.0443], + [ 0.0457, 0.0264, -0.0359, ..., -0.0632, -0.0114, -0.0422], + [-0.0136, 0.0325, -0.0416, ..., -0.0162, -0.0542, -0.0143]], + device='cuda:0'), grad: tensor([[ 2.2459e-04, 1.7607e-04, 7.1669e-04, ..., 5.5045e-05, + 5.7793e-04, 2.4056e-04], + [ 2.5773e-04, 2.8324e-04, 2.0528e-04, ..., 2.0039e-04, + 2.1303e-04, 3.0875e-05], + [-5.8222e-04, -2.6250e-04, 4.2295e-04, ..., 4.8310e-05, + -1.3638e-04, 1.6975e-04], + ..., + [ 2.5511e-04, 2.4104e-04, 1.3120e-05, ..., 4.0591e-05, + -2.3007e-04, 7.2896e-05], + [ 7.0691e-05, 1.5891e-04, 1.7822e-04, ..., 2.7800e-04, + 4.0501e-05, 3.5763e-05], + [ 9.9468e-04, 1.2684e-03, 5.1975e-04, ..., 8.3542e-04, + 1.6069e-04, 1.9044e-05]], device='cuda:0') +Epoch 25, bias, value: tensor([-2.2794e-02, -2.0575e-02, -1.9383e-02, -9.0776e-05, -5.9636e-03, + 3.1688e-04, -3.0130e-03, -1.7527e-02, 2.2929e-02, -2.9919e-02], + device='cuda:0'), grad: tensor([ 0.0015, 0.0011, -0.0031, -0.0008, -0.0005, -0.0003, 0.0005, -0.0023, + 0.0009, 0.0030], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 219.45, cls_loss 0.0211 cls_loss_mapping 0.0444 cls_loss_causal 0.8315 re_mapping 0.0210 re_causal 0.0647 /// teacc 98.65 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0048, -0.0277, 0.0594, ..., -0.0092, -0.0310, -0.0228], + [-0.0517, -0.0320, -0.0583, ..., 0.0026, 0.0405, 0.0415], + [-0.0123, -0.0097, -0.0163, ..., -0.0263, 0.0198, -0.0539], + ..., + [-0.0645, -0.0597, 0.0240, ..., -0.0667, -0.0413, -0.0456], + [ 0.0462, 0.0267, -0.0361, ..., -0.0634, -0.0109, -0.0426], + [-0.0129, 0.0333, -0.0422, ..., -0.0158, -0.0545, -0.0132]], + device='cuda:0'), grad: tensor([[ 6.3372e-04, 4.0698e-04, 3.4595e-04, ..., 1.0711e-04, + 4.9639e-04, 1.9316e-06], + [ 1.4091e-04, 7.8082e-05, 1.0413e-04, ..., 2.2322e-05, + 1.4794e-04, -3.0309e-05], + [ 3.4928e-04, 2.1315e-04, 2.0659e-04, ..., 6.9559e-05, + 2.7466e-04, 7.0632e-06], + ..., + [ 3.5810e-04, -5.5939e-05, -4.5180e-04, ..., 2.5320e-04, + 1.5748e-04, 7.8678e-06], + [-2.6970e-03, -1.2741e-03, -5.4121e-04, ..., -1.5688e-04, + -1.9646e-03, 6.7987e-06], + [-3.4618e-04, -5.6028e-05, 3.5644e-04, ..., -5.8746e-04, + 4.7469e-04, 2.2128e-06]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0227, -0.0211, -0.0194, -0.0001, -0.0060, 0.0003, -0.0031, -0.0174, + 0.0231, -0.0298], device='cuda:0'), grad: tensor([ 0.0011, 0.0005, 0.0006, 0.0012, 0.0002, 0.0008, 0.0005, -0.0004, + -0.0043, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 220.15, cls_loss 0.0184 cls_loss_mapping 0.0395 cls_loss_causal 0.8051 re_mapping 0.0204 re_causal 0.0641 /// teacc 98.91 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0050, -0.0274, 0.0600, ..., -0.0090, -0.0312, -0.0232], + [-0.0521, -0.0325, -0.0597, ..., 0.0028, 0.0406, 0.0441], + [-0.0128, -0.0106, -0.0166, ..., -0.0271, 0.0199, -0.0555], + ..., + [-0.0650, -0.0601, 0.0242, ..., -0.0675, -0.0418, -0.0462], + [ 0.0465, 0.0265, -0.0366, ..., -0.0643, -0.0108, -0.0431], + [-0.0124, 0.0336, -0.0426, ..., -0.0150, -0.0553, -0.0140]], + device='cuda:0'), grad: tensor([[ 2.7597e-05, -7.8619e-05, -1.1152e-04, ..., -1.0216e-04, + 1.3769e-04, 1.9580e-05], + [ 5.5611e-05, 2.0385e-05, 2.3155e-03, ..., -1.6105e-04, + 9.9792e-03, 1.9484e-03], + [ 3.3641e-04, 1.1474e-04, -2.5139e-03, ..., 1.2636e-04, + -1.1322e-02, -2.1725e-03], + ..., + [ 1.1301e-04, 7.5340e-05, -3.6880e-06, ..., 1.1563e-04, + 6.0225e-04, 1.0520e-04], + [-1.1587e-03, -4.5061e-04, 1.0006e-05, ..., 1.9878e-05, + -6.5851e-04, -6.6102e-05], + [-6.8903e-05, -4.5747e-05, 1.3173e-04, ..., -9.3281e-05, + 1.1122e-04, 8.9630e-06]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0229, -0.0215, -0.0194, -0.0003, -0.0061, 0.0003, -0.0029, -0.0170, + 0.0230, -0.0296], device='cuda:0'), grad: tensor([ 0.0001, 0.0122, -0.0141, 0.0012, 0.0005, 0.0003, 0.0002, 0.0010, + -0.0016, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 219.39, cls_loss 0.0229 cls_loss_mapping 0.0422 cls_loss_causal 0.8035 re_mapping 0.0193 re_causal 0.0598 /// teacc 98.72 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0053, -0.0275, 0.0601, ..., -0.0090, -0.0317, -0.0235], + [-0.0527, -0.0334, -0.0605, ..., 0.0031, 0.0407, 0.0484], + [-0.0136, -0.0110, -0.0173, ..., -0.0276, 0.0197, -0.0573], + ..., + [-0.0660, -0.0611, 0.0246, ..., -0.0687, -0.0418, -0.0482], + [ 0.0471, 0.0265, -0.0370, ..., -0.0648, -0.0107, -0.0433], + [-0.0124, 0.0337, -0.0433, ..., -0.0149, -0.0558, -0.0150]], + device='cuda:0'), grad: tensor([[ 1.7619e-04, 1.4818e-04, -1.1528e-04, ..., -2.9302e-04, + 1.0538e-04, 3.5251e-07], + [ 4.1485e-05, 5.5015e-05, 3.7134e-05, ..., -1.5616e-05, + -1.0610e-05, -2.3365e-05], + [ 9.0480e-05, 1.2827e-04, 1.1098e-04, ..., 6.5565e-05, + 2.1800e-05, 4.0233e-06], + ..., + [ 5.1689e-04, 5.6505e-04, 3.6448e-05, ..., 2.6441e-04, + 2.9892e-05, 8.1360e-06], + [-3.5787e-04, -6.3229e-04, -6.5327e-04, ..., 1.2445e-04, + -2.4092e-04, 4.0531e-06], + [-8.6308e-04, -8.1062e-04, 1.3554e-04, ..., -1.9586e-04, + 4.2289e-05, 1.4938e-06]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0233, -0.0217, -0.0195, -0.0001, -0.0059, 0.0007, -0.0026, -0.0170, + 0.0231, -0.0299], device='cuda:0'), grad: tensor([ 1.8454e-04, 7.2181e-05, 2.1756e-04, 8.5235e-06, 8.0442e-04, + 4.2796e-04, 1.9372e-04, 9.3842e-04, -9.4748e-04, -1.9026e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 219.55, cls_loss 0.0165 cls_loss_mapping 0.0374 cls_loss_causal 0.8007 re_mapping 0.0192 re_causal 0.0601 /// teacc 98.76 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0057, -0.0275, 0.0604, ..., -0.0092, -0.0322, -0.0237], + [-0.0528, -0.0336, -0.0607, ..., 0.0040, 0.0410, 0.0518], + [-0.0140, -0.0118, -0.0179, ..., -0.0281, 0.0199, -0.0586], + ..., + [-0.0666, -0.0616, 0.0249, ..., -0.0701, -0.0423, -0.0490], + [ 0.0476, 0.0266, -0.0370, ..., -0.0653, -0.0106, -0.0445], + [-0.0121, 0.0340, -0.0440, ..., -0.0143, -0.0565, -0.0154]], + device='cuda:0'), grad: tensor([[ 2.2128e-05, -4.6253e-05, -8.6188e-05, ..., -8.3208e-05, + 2.4229e-05, 3.8594e-06], + [ 1.1742e-05, 2.0623e-05, 1.0081e-05, ..., -7.2420e-05, + -6.7890e-05, -6.5088e-05], + [ 6.6698e-05, 3.8505e-05, 6.8918e-06, ..., 6.3419e-05, + 3.9816e-05, 2.4348e-05], + ..., + [ 6.2287e-05, 1.4871e-05, 3.2067e-04, ..., 2.6703e-05, + 3.8218e-04, 1.4883e-06], + [-4.2081e-04, -3.8981e-04, 3.8713e-05, ..., -5.2959e-05, + 1.1720e-05, 2.0772e-05], + [ 4.1533e-04, 4.1389e-04, 3.8099e-04, ..., 2.3055e-04, + 2.1338e-04, 9.0897e-07]], device='cuda:0') +Epoch 29, bias, value: tensor([-2.3827e-02, -2.1447e-02, -1.9447e-02, -2.9188e-05, -5.9508e-03, + 8.6953e-04, -2.9948e-03, -1.7191e-02, 2.3226e-02, -2.9735e-02], + device='cuda:0'), grad: tensor([-2.3227e-06, -2.5406e-06, 7.0512e-05, -1.5676e-04, -9.9659e-04, + 2.1207e-04, 1.5885e-05, -4.3654e-04, -3.3069e-04, 1.6270e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 219.66, cls_loss 0.0150 cls_loss_mapping 0.0309 cls_loss_causal 0.7713 re_mapping 0.0187 re_causal 0.0585 /// teacc 98.90 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0058, -0.0273, 0.0609, ..., -0.0087, -0.0323, -0.0240], + [-0.0533, -0.0340, -0.0611, ..., 0.0042, 0.0410, 0.0560], + [-0.0146, -0.0127, -0.0186, ..., -0.0288, 0.0200, -0.0600], + ..., + [-0.0672, -0.0625, 0.0253, ..., -0.0709, -0.0420, -0.0505], + [ 0.0480, 0.0268, -0.0371, ..., -0.0658, -0.0106, -0.0456], + [-0.0117, 0.0341, -0.0442, ..., -0.0141, -0.0568, -0.0160]], + device='cuda:0'), grad: tensor([[-8.7619e-05, -4.9171e-03, -1.0834e-02, ..., -8.1711e-03, + 3.7014e-05, 6.9737e-06], + [ 1.0747e-04, 7.9691e-05, 5.2661e-05, ..., -1.8537e-04, + -1.3137e-04, -2.5439e-04], + [ 2.5526e-05, 3.5524e-05, -3.2449e-04, ..., 4.6909e-05, + -1.5993e-03, 2.7508e-05], + ..., + [ 1.4693e-05, 3.5286e-05, 1.5724e-04, ..., 9.3341e-05, + 5.6982e-04, 7.9453e-05], + [-2.9397e-04, -1.1957e-04, 2.6369e-04, ..., 1.1671e-04, + 7.7391e-04, 1.4827e-05], + [ 3.0637e-05, 1.4293e-04, 2.4557e-04, ..., 1.3173e-04, + 8.1062e-05, 2.8223e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([-2.3563e-02, -2.1745e-02, -1.9390e-02, -7.9219e-05, -5.9359e-03, + 6.2990e-04, -3.2675e-03, -1.6970e-02, 2.3242e-02, -2.9532e-02], + device='cuda:0'), grad: tensor([-0.0090, -0.0001, -0.0028, 0.0006, 0.0004, 0.0070, 0.0015, 0.0008, + 0.0012, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 219.67, cls_loss 0.0179 cls_loss_mapping 0.0366 cls_loss_causal 0.7344 re_mapping 0.0181 re_causal 0.0555 /// teacc 98.84 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0059, -0.0270, 0.0618, ..., -0.0081, -0.0325, -0.0242], + [-0.0536, -0.0348, -0.0621, ..., 0.0046, 0.0409, 0.0583], + [-0.0152, -0.0133, -0.0192, ..., -0.0295, 0.0200, -0.0614], + ..., + [-0.0678, -0.0630, 0.0249, ..., -0.0717, -0.0425, -0.0517], + [ 0.0485, 0.0273, -0.0372, ..., -0.0662, -0.0104, -0.0463], + [-0.0114, 0.0341, -0.0447, ..., -0.0137, -0.0576, -0.0169]], + device='cuda:0'), grad: tensor([[-1.0071e-03, -1.8978e-03, -2.2717e-03, ..., -1.6336e-03, + -5.6887e-04, 1.5097e-06], + [ 1.1533e-04, 1.8656e-04, 2.1601e-04, ..., 9.8348e-05, + -1.0163e-05, -9.6440e-05], + [-3.0413e-05, 5.8144e-05, 5.4300e-05, ..., 1.1373e-04, + -1.3816e-04, 3.0205e-05], + ..., + [ 4.0591e-05, 3.7551e-05, -9.0152e-07, ..., 4.5329e-05, + 5.2959e-05, 1.5318e-05], + [ 3.4881e-04, 5.8889e-04, 6.3848e-04, ..., 5.0640e-04, + 2.3019e-04, 9.1940e-06], + [ 6.0916e-05, 8.7440e-05, 2.4986e-04, ..., 8.5652e-05, + 8.9526e-05, 7.6964e-06]], device='cuda:0') +Epoch 31, bias, value: tensor([-2.3240e-02, -2.2042e-02, -1.9281e-02, -3.8417e-05, -6.0768e-03, + 8.6962e-04, -3.0596e-03, -1.7329e-02, 2.3506e-02, -2.9813e-02], + device='cuda:0'), grad: tensor([-0.0029, 0.0002, -0.0001, 0.0011, -0.0006, -0.0002, 0.0012, -0.0001, + 0.0010, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 219.78, cls_loss 0.0143 cls_loss_mapping 0.0308 cls_loss_causal 0.7170 re_mapping 0.0183 re_causal 0.0555 /// teacc 98.89 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0057, -0.0268, 0.0618, ..., -0.0083, -0.0328, -0.0245], + [-0.0539, -0.0353, -0.0626, ..., 0.0052, 0.0410, 0.0625], + [-0.0156, -0.0137, -0.0198, ..., -0.0301, 0.0199, -0.0644], + ..., + [-0.0686, -0.0638, 0.0248, ..., -0.0725, -0.0429, -0.0527], + [ 0.0487, 0.0273, -0.0372, ..., -0.0671, -0.0102, -0.0465], + [-0.0114, 0.0342, -0.0450, ..., -0.0132, -0.0582, -0.0176]], + device='cuda:0'), grad: tensor([[ 2.4080e-05, -3.0957e-06, 2.8431e-05, ..., 6.5982e-05, + 4.3362e-05, 7.3202e-07], + [ 1.9833e-05, -9.4950e-05, 5.5619e-06, ..., -4.4441e-04, + -5.4073e-04, -4.6968e-05], + [ 7.1824e-05, 8.1599e-05, 7.6890e-06, ..., 2.8539e-04, + 2.4843e-04, 1.5765e-05], + ..., + [-7.8753e-06, 2.7746e-05, 3.1553e-06, ..., 8.5235e-05, + 1.3781e-04, 4.9509e-06], + [ 5.4747e-05, 5.5999e-05, 4.3988e-05, ..., 1.3542e-04, + 1.1408e-04, 7.7933e-06], + [-4.6420e-04, -4.6587e-04, -1.4029e-05, ..., -9.4175e-05, + 4.1938e-04, 3.7849e-06]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0236, -0.0222, -0.0195, 0.0003, -0.0058, 0.0006, -0.0028, -0.0174, + 0.0238, -0.0300], device='cuda:0'), grad: tensor([ 3.9649e-04, -1.3590e-03, -1.5726e-03, 4.8709e-04, -1.3161e-04, + -3.2544e-05, 9.9182e-05, 6.4135e-04, 5.5456e-04, 9.1887e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 219.60, cls_loss 0.0142 cls_loss_mapping 0.0343 cls_loss_causal 0.7776 re_mapping 0.0173 re_causal 0.0552 /// teacc 98.67 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0056, -0.0265, 0.0624, ..., -0.0079, -0.0331, -0.0246], + [-0.0543, -0.0360, -0.0633, ..., 0.0057, 0.0411, 0.0647], + [-0.0162, -0.0147, -0.0205, ..., -0.0311, 0.0200, -0.0648], + ..., + [-0.0691, -0.0644, 0.0249, ..., -0.0729, -0.0430, -0.0542], + [ 0.0492, 0.0276, -0.0376, ..., -0.0679, -0.0101, -0.0479], + [-0.0110, 0.0345, -0.0453, ..., -0.0128, -0.0588, -0.0182]], + device='cuda:0'), grad: tensor([[ 8.8140e-06, -9.6202e-05, -2.4223e-04, ..., -9.9123e-05, + 4.3005e-05, 1.7881e-05], + [-3.4422e-05, 9.2238e-06, -7.6532e-05, ..., -4.4751e-04, + -1.0719e-03, -6.0320e-04], + [ 3.2455e-05, 4.5091e-05, 5.1260e-05, ..., 7.9155e-05, + 2.4378e-04, 1.1843e-04], + ..., + [ 7.3075e-05, 7.7128e-05, 3.9726e-05, ..., 1.1468e-04, + 1.7250e-04, 9.3818e-05], + [ 4.1509e-04, 7.4053e-04, 2.1374e-04, ..., 3.3164e-04, + 4.5347e-04, 2.1505e-04], + [-2.4652e-04, -1.0949e-04, 4.3929e-05, ..., -1.2529e-04, + 1.2958e-04, 1.0483e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0235, -0.0223, -0.0197, 0.0001, -0.0060, 0.0012, -0.0033, -0.0171, + 0.0239, -0.0301], device='cuda:0'), grad: tensor([-1.7071e-04, -1.7376e-03, 5.1308e-04, 2.5673e-03, 1.2338e-05, + -2.7905e-03, 4.8995e-05, 1.4496e-04, 1.5249e-03, -1.1241e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 219.52, cls_loss 0.0121 cls_loss_mapping 0.0304 cls_loss_causal 0.7599 re_mapping 0.0174 re_causal 0.0545 /// teacc 98.76 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0059, -0.0263, 0.0629, ..., -0.0076, -0.0335, -0.0248], + [-0.0545, -0.0364, -0.0642, ..., 0.0063, 0.0410, 0.0680], + [-0.0167, -0.0154, -0.0209, ..., -0.0317, 0.0201, -0.0666], + ..., + [-0.0696, -0.0650, 0.0253, ..., -0.0740, -0.0426, -0.0556], + [ 0.0493, 0.0272, -0.0376, ..., -0.0684, -0.0099, -0.0488], + [-0.0106, 0.0349, -0.0461, ..., -0.0122, -0.0595, -0.0188]], + device='cuda:0'), grad: tensor([[-2.3365e-05, 3.4422e-05, 2.2829e-04, ..., 5.1945e-05, + 1.2875e-04, 2.1998e-06], + [ 2.7359e-05, 8.1182e-05, 1.1384e-04, ..., 5.9754e-05, + 5.6416e-05, -4.9472e-05], + [-2.2516e-05, -3.1829e-05, 1.1718e-04, ..., -5.0738e-06, + -2.7442e-04, 1.6734e-05], + ..., + [ 8.9467e-05, 1.1533e-04, 5.9605e-05, ..., 7.7546e-05, + 1.5438e-04, 8.4043e-06], + [ 1.3959e-04, 3.3283e-03, 2.7046e-03, ..., 1.5135e-03, + 1.3280e-04, 1.5870e-05], + [-3.4714e-04, -1.5879e-04, 1.4651e-04, ..., -2.7251e-04, + 2.8086e-04, 2.2277e-06]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0235, -0.0223, -0.0195, 0.0005, -0.0060, 0.0010, -0.0034, -0.0169, + 0.0240, -0.0304], device='cuda:0'), grad: tensor([ 0.0003, 0.0004, -0.0006, 0.0010, -0.0011, -0.0051, 0.0003, -0.0004, + 0.0045, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 219.54, cls_loss 0.0118 cls_loss_mapping 0.0292 cls_loss_causal 0.7482 re_mapping 0.0163 re_causal 0.0520 /// teacc 98.88 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0061, -0.0262, 0.0633, ..., -0.0072, -0.0339, -0.0250], + [-0.0550, -0.0369, -0.0645, ..., 0.0063, 0.0409, 0.0710], + [-0.0168, -0.0155, -0.0214, ..., -0.0315, 0.0202, -0.0673], + ..., + [-0.0702, -0.0657, 0.0252, ..., -0.0751, -0.0425, -0.0567], + [ 0.0498, 0.0271, -0.0382, ..., -0.0694, -0.0095, -0.0499], + [-0.0102, 0.0351, -0.0465, ..., -0.0115, -0.0602, -0.0193]], + device='cuda:0'), grad: tensor([[ 7.9775e-04, 7.4339e-04, 4.0674e-04, ..., 7.0047e-04, + 4.0674e-04, 2.9728e-06], + [ 3.2395e-05, 5.9605e-05, 9.5189e-05, ..., -9.4652e-05, + -1.0169e-04, -2.1827e-04], + [ 2.7299e-04, 3.4571e-04, 2.8753e-04, ..., 2.0874e-04, + 3.9697e-04, 3.4064e-05], + ..., + [ 8.7857e-05, 1.3828e-04, 1.5938e-04, ..., 1.0002e-04, + 2.5392e-04, 1.2286e-05], + [ 3.1891e-03, 2.8286e-03, 1.1873e-03, ..., 3.1681e-03, + -1.1492e-03, 8.3745e-05], + [-6.1836e-03, -5.9204e-03, -2.9697e-03, ..., -5.1880e-03, + -9.5272e-04, 1.1668e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0234, -0.0226, -0.0196, 0.0003, -0.0061, 0.0008, -0.0032, -0.0166, + 0.0241, -0.0304], device='cuda:0'), grad: tensor([ 0.0013, 0.0003, 0.0003, 0.0011, 0.0039, 0.0011, -0.0010, -0.0002, + 0.0020, -0.0088], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 219.71, cls_loss 0.0122 cls_loss_mapping 0.0291 cls_loss_causal 0.7240 re_mapping 0.0159 re_causal 0.0498 /// teacc 98.82 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0064, -0.0264, 0.0636, ..., -0.0072, -0.0343, -0.0255], + [-0.0553, -0.0372, -0.0649, ..., 0.0068, 0.0410, 0.0758], + [-0.0171, -0.0163, -0.0222, ..., -0.0322, 0.0203, -0.0692], + ..., + [-0.0707, -0.0663, 0.0251, ..., -0.0757, -0.0427, -0.0591], + [ 0.0502, 0.0273, -0.0383, ..., -0.0701, -0.0095, -0.0522], + [-0.0098, 0.0354, -0.0469, ..., -0.0111, -0.0608, -0.0204]], + device='cuda:0'), grad: tensor([[ 2.6539e-05, -9.7632e-05, -3.0470e-04, ..., -2.1553e-04, + 3.0361e-06, 7.8324e-07], + [ 3.5912e-05, 6.0022e-05, 7.3671e-05, ..., 5.4628e-05, + -6.7616e-04, -3.1412e-05], + [ 1.1247e-04, 1.3542e-04, -5.8293e-05, ..., 1.2779e-04, + 3.2616e-04, 2.7772e-06], + ..., + [ 6.1274e-05, 6.5088e-05, 2.1911e-04, ..., 5.9098e-05, + 2.2125e-04, 5.6997e-06], + [ 2.5296e-04, 3.0470e-04, 1.3626e-04, ..., 2.2697e-04, + 2.7013e-04, 5.5507e-06], + [-2.9230e-04, -2.1541e-04, 1.4460e-04, ..., -1.0276e-04, + 1.4174e-04, 1.0403e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([-2.3718e-02, -2.2638e-02, -1.9088e-02, 7.9559e-05, -6.2840e-03, + 1.0021e-03, -2.9039e-03, -1.6538e-02, 2.4004e-02, -3.0642e-02], + device='cuda:0'), grad: tensor([-3.4428e-04, -8.6832e-04, -5.1260e-05, -5.9652e-04, -1.6332e-04, + 2.2745e-04, 7.6354e-05, 9.3937e-04, 8.7929e-04, -9.7752e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 219.72, cls_loss 0.0134 cls_loss_mapping 0.0288 cls_loss_causal 0.7445 re_mapping 0.0157 re_causal 0.0493 /// teacc 98.74 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0066, -0.0263, 0.0641, ..., -0.0072, -0.0346, -0.0256], + [-0.0556, -0.0382, -0.0657, ..., 0.0072, 0.0411, 0.0780], + [-0.0176, -0.0171, -0.0228, ..., -0.0330, 0.0202, -0.0689], + ..., + [-0.0717, -0.0673, 0.0254, ..., -0.0768, -0.0430, -0.0615], + [ 0.0504, 0.0271, -0.0390, ..., -0.0708, -0.0094, -0.0527], + [-0.0093, 0.0362, -0.0469, ..., -0.0102, -0.0613, -0.0210]], + device='cuda:0'), grad: tensor([[ 4.6670e-05, 1.7628e-05, -4.1991e-05, ..., -3.6597e-05, + 2.8357e-05, 1.3690e-06], + [ 2.0519e-05, 2.1830e-05, 1.7926e-05, ..., -2.6390e-05, + -1.3947e-05, -5.6565e-05], + [ 1.3983e-04, 1.5152e-04, 2.9907e-05, ..., 8.7321e-05, + 4.2468e-06, 5.1968e-06], + ..., + [ 4.2468e-05, 4.6730e-05, 9.7007e-06, ..., 3.6776e-05, + 7.0572e-05, 1.4454e-05], + [-3.9887e-04, -3.1853e-04, -2.2006e-04, ..., 4.3869e-05, + -1.5402e-04, 1.4074e-05], + [ 3.4428e-04, 3.3355e-04, 2.9206e-04, ..., 1.2255e-04, + 2.9445e-04, 4.5411e-06]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0239, -0.0229, -0.0192, 0.0004, -0.0061, 0.0007, -0.0025, -0.0165, + 0.0237, -0.0304], device='cuda:0'), grad: tensor([ 1.2827e-04, 4.0591e-05, -3.2330e-04, -5.3940e-03, -3.0541e-04, + 5.2986e-03, 1.5962e-04, 1.3554e-04, -6.2799e-04, 8.9121e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 36---------------------------------------------------- +epoch 36, time 220.66, cls_loss 0.0093 cls_loss_mapping 0.0249 cls_loss_causal 0.7233 re_mapping 0.0157 re_causal 0.0507 /// teacc 98.99 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0067, -0.0261, 0.0644, ..., -0.0069, -0.0349, -0.0257], + [-0.0561, -0.0388, -0.0665, ..., 0.0075, 0.0413, 0.0812], + [-0.0182, -0.0178, -0.0234, ..., -0.0335, 0.0201, -0.0703], + ..., + [-0.0723, -0.0682, 0.0252, ..., -0.0778, -0.0435, -0.0636], + [ 0.0509, 0.0271, -0.0393, ..., -0.0713, -0.0093, -0.0537], + [-0.0091, 0.0364, -0.0474, ..., -0.0099, -0.0620, -0.0219]], + device='cuda:0'), grad: tensor([[ 2.9653e-05, 2.4945e-05, -1.0061e-04, ..., -6.4969e-05, + 1.7345e-05, 2.4922e-06], + [ 2.2113e-05, 3.2097e-05, 7.1414e-06, ..., -4.0126e-04, + -2.8110e-04, -2.6560e-04], + [ 4.5031e-05, 6.8426e-05, 3.3081e-05, ..., 5.4926e-05, + 7.5027e-06, 2.2471e-05], + ..., + [ 3.0828e-04, 3.8075e-04, 4.2498e-05, ..., 3.7670e-04, + 1.7595e-04, 1.5032e-04], + [ 6.2823e-05, 1.1790e-04, 1.2732e-04, ..., 1.9217e-04, + -6.8963e-05, 2.8059e-05], + [-1.6251e-03, -1.7433e-03, -1.3554e-04, ..., -5.6028e-04, + 2.6762e-05, 1.7852e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0240, -0.0228, -0.0196, 0.0005, -0.0058, 0.0008, -0.0025, -0.0164, + 0.0238, -0.0307], device='cuda:0'), grad: tensor([ 3.8855e-06, -8.8358e-04, 2.1696e-04, 1.2817e-03, 9.3126e-04, + 1.0204e-03, -4.3917e-04, 1.1654e-03, 2.5797e-04, -3.5534e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 219.74, cls_loss 0.0085 cls_loss_mapping 0.0231 cls_loss_causal 0.7267 re_mapping 0.0151 re_causal 0.0494 /// teacc 98.87 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0070, -0.0262, 0.0646, ..., -0.0070, -0.0350, -0.0256], + [-0.0564, -0.0392, -0.0667, ..., 0.0080, 0.0417, 0.0850], + [-0.0187, -0.0183, -0.0237, ..., -0.0336, 0.0201, -0.0724], + ..., + [-0.0727, -0.0688, 0.0252, ..., -0.0787, -0.0439, -0.0649], + [ 0.0514, 0.0273, -0.0396, ..., -0.0719, -0.0092, -0.0548], + [-0.0085, 0.0368, -0.0477, ..., -0.0089, -0.0625, -0.0224]], + device='cuda:0'), grad: tensor([[-4.6104e-05, -1.4997e-04, -1.3387e-04, ..., -1.1921e-04, + 3.7313e-05, 5.6298e-07], + [ 2.3752e-05, 2.7329e-05, 1.6344e-04, ..., 1.8203e-04, + 3.4285e-04, -3.4839e-05], + [ 1.7017e-05, 4.0948e-05, 5.0735e-04, ..., 6.2323e-04, + 1.1282e-03, 1.4082e-06], + ..., + [ 2.6613e-05, 2.5973e-05, 1.9446e-05, ..., 3.7819e-05, + 4.8250e-05, 1.3806e-05], + [-1.6558e-04, -1.1182e-04, 1.5128e-04, ..., 2.3282e-04, + 2.6441e-04, 3.7365e-06], + [ 2.4989e-05, 1.3125e-04, 1.5569e-04, ..., 3.7640e-05, + 1.0353e-04, 1.2498e-06]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0242, -0.0226, -0.0192, 0.0004, -0.0060, 0.0005, -0.0023, -0.0166, + 0.0239, -0.0305], device='cuda:0'), grad: tensor([-2.0146e-04, 5.2786e-04, 1.6470e-03, 4.1032e-04, 3.5095e-04, + -6.2704e-05, -3.2597e-03, -3.2902e-05, 2.8157e-04, 3.4022e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 219.31, cls_loss 0.0117 cls_loss_mapping 0.0286 cls_loss_causal 0.7104 re_mapping 0.0150 re_causal 0.0477 /// teacc 98.89 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0074, -0.0259, 0.0650, ..., -0.0068, -0.0354, -0.0253], + [-0.0574, -0.0405, -0.0672, ..., 0.0071, 0.0413, 0.0873], + [-0.0191, -0.0186, -0.0242, ..., -0.0338, 0.0204, -0.0732], + ..., + [-0.0731, -0.0694, 0.0256, ..., -0.0792, -0.0438, -0.0666], + [ 0.0516, 0.0273, -0.0399, ..., -0.0725, -0.0093, -0.0559], + [-0.0074, 0.0375, -0.0479, ..., -0.0077, -0.0629, -0.0220]], + device='cuda:0'), grad: tensor([[-1.2684e-04, -4.8780e-04, -8.0538e-04, ..., -5.9462e-04, + 3.3677e-05, 7.5623e-07], + [-1.4808e-07, 2.4959e-05, 2.6021e-06, ..., -3.5906e-04, + -6.6471e-04, -2.0921e-04], + [ 1.9252e-05, 5.0783e-05, 4.9412e-05, ..., 1.6046e-04, + 1.8811e-04, 2.7627e-05], + ..., + [ 4.8846e-05, 7.1526e-05, 4.8935e-05, ..., 1.2982e-04, + 2.0134e-04, 6.3479e-05], + [-1.2922e-04, 1.2082e-04, 1.2970e-04, ..., 1.6093e-04, + 2.3916e-05, 4.3184e-05], + [ 2.2203e-05, 2.4247e-04, 1.7309e-04, ..., 1.7333e-04, + 4.7058e-05, 2.0027e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-2.4196e-02, -2.3602e-02, -1.9108e-02, -5.6704e-05, -6.4744e-03, + 8.1751e-04, -2.4925e-03, -1.5650e-02, 2.3544e-02, -2.9902e-02], + device='cuda:0'), grad: tensor([-0.0012, -0.0012, 0.0003, 0.0004, 0.0002, -0.0006, 0.0008, 0.0005, + 0.0003, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 219.56, cls_loss 0.0090 cls_loss_mapping 0.0260 cls_loss_causal 0.7121 re_mapping 0.0149 re_causal 0.0469 /// teacc 98.83 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0073, -0.0254, 0.0656, ..., -0.0062, -0.0358, -0.0253], + [-0.0577, -0.0409, -0.0675, ..., 0.0076, 0.0420, 0.0914], + [-0.0195, -0.0193, -0.0246, ..., -0.0344, 0.0203, -0.0747], + ..., + [-0.0735, -0.0700, 0.0257, ..., -0.0801, -0.0446, -0.0693], + [ 0.0519, 0.0271, -0.0402, ..., -0.0730, -0.0092, -0.0579], + [-0.0072, 0.0377, -0.0484, ..., -0.0071, -0.0637, -0.0221]], + device='cuda:0'), grad: tensor([[-2.7530e-06, -5.1260e-05, -1.0639e-04, ..., -5.5462e-05, + 1.6108e-05, 4.7460e-06], + [ 4.8578e-06, -1.1586e-05, -3.3498e-04, ..., -3.3349e-05, + -6.1369e-04, -4.5753e-04], + [ 1.4558e-05, 3.6746e-05, 5.5373e-05, ..., 4.0084e-05, + 5.4449e-05, 5.7310e-05], + ..., + [ 3.0413e-05, 2.9698e-05, 1.7002e-05, ..., 5.7071e-05, + 1.3483e-04, 6.6459e-05], + [ 2.6435e-05, 6.0469e-05, 4.6700e-05, ..., 4.5896e-05, + 5.6475e-05, 3.1948e-05], + [-7.9751e-05, -2.5406e-05, 2.7120e-05, ..., -8.3148e-05, + 1.3493e-05, 5.2974e-06]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0240, -0.0227, -0.0192, -0.0002, -0.0059, 0.0007, -0.0025, -0.0160, + 0.0233, -0.0302], device='cuda:0'), grad: tensor([ 4.6581e-05, -1.4086e-03, -4.2272e-04, 3.0327e-04, 5.0402e-04, + 4.9531e-05, 2.0516e-04, 4.5252e-04, 2.7418e-04, -4.3996e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 219.77, cls_loss 0.0098 cls_loss_mapping 0.0239 cls_loss_causal 0.7020 re_mapping 0.0145 re_causal 0.0449 /// teacc 98.86 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0073, -0.0253, 0.0661, ..., -0.0056, -0.0353, -0.0257], + [-0.0580, -0.0412, -0.0681, ..., 0.0077, 0.0421, 0.0948], + [-0.0200, -0.0200, -0.0252, ..., -0.0351, 0.0204, -0.0760], + ..., + [-0.0740, -0.0708, 0.0259, ..., -0.0812, -0.0450, -0.0706], + [ 0.0521, 0.0273, -0.0398, ..., -0.0734, -0.0091, -0.0594], + [-0.0070, 0.0379, -0.0489, ..., -0.0064, -0.0641, -0.0209]], + device='cuda:0'), grad: tensor([[ 1.8179e-04, 2.5344e-04, 1.0359e-04, ..., 1.0389e-04, + 2.3678e-05, 5.3830e-07], + [ 1.8075e-05, 1.7777e-05, 1.4879e-05, ..., -7.1079e-06, + 5.9158e-06, -2.4885e-05], + [ 1.2118e-04, 1.5748e-04, 6.2585e-05, ..., 6.9022e-05, + -7.0477e-04, 4.2208e-06], + ..., + [ 4.7028e-05, 5.5581e-05, 3.2455e-05, ..., 3.5822e-05, + 8.1062e-05, 6.4000e-06], + [ 6.3372e-04, 1.4086e-03, 8.5449e-04, ..., 7.7963e-04, + -9.0659e-05, 4.3474e-06], + [-1.6870e-03, -2.7752e-03, -1.2255e-03, ..., -1.2960e-03, + 1.5378e-04, 1.1250e-06]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0238, -0.0231, -0.0194, 0.0003, -0.0056, 0.0003, -0.0030, -0.0159, + 0.0236, -0.0303], device='cuda:0'), grad: tensor([ 3.9768e-04, 4.2528e-05, -6.1131e-04, 3.8218e-04, 6.8140e-04, + 6.2752e-04, -3.8505e-05, 2.0266e-04, 2.1381e-03, -3.8242e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 219.62, cls_loss 0.0099 cls_loss_mapping 0.0267 cls_loss_causal 0.6996 re_mapping 0.0137 re_causal 0.0455 /// teacc 98.93 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0077, -0.0249, 0.0661, ..., -0.0057, -0.0356, -0.0261], + [-0.0583, -0.0416, -0.0683, ..., 0.0082, 0.0423, 0.0969], + [-0.0207, -0.0214, -0.0257, ..., -0.0358, 0.0202, -0.0766], + ..., + [-0.0749, -0.0713, 0.0258, ..., -0.0819, -0.0453, -0.0722], + [ 0.0522, 0.0269, -0.0404, ..., -0.0743, -0.0089, -0.0600], + [-0.0065, 0.0380, -0.0492, ..., -0.0061, -0.0648, -0.0212]], + device='cuda:0'), grad: tensor([[ 5.8264e-05, -7.8506e-03, -8.6594e-03, ..., -8.7280e-03, + 1.8388e-05, 3.6526e-06], + [ 6.2399e-06, 3.8117e-05, 3.8058e-05, ..., -1.4079e-04, + -2.2316e-04, -2.7204e-04], + [ 3.0130e-05, 1.5330e-04, 1.3733e-04, ..., 1.1647e-04, + -1.5187e-04, 8.1480e-05], + ..., + [-1.9765e-04, -3.4899e-05, 5.3078e-05, ..., 1.2290e-04, + 1.7452e-04, 6.9141e-05], + [ 4.5776e-04, 6.8903e-04, 1.5402e-04, ..., 4.1986e-04, + 5.4240e-05, 2.7359e-05], + [-6.7091e-04, -8.3780e-04, -4.8205e-06, ..., -4.0078e-04, + 8.5115e-05, 2.3380e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0244, -0.0230, -0.0196, 0.0006, -0.0057, 0.0009, -0.0026, -0.0155, + 0.0232, -0.0307], device='cuda:0'), grad: tensor([-0.0114, -0.0004, -0.0009, 0.0012, 0.0002, 0.0062, 0.0042, 0.0003, + 0.0010, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 219.93, cls_loss 0.0073 cls_loss_mapping 0.0192 cls_loss_causal 0.6782 re_mapping 0.0142 re_causal 0.0451 /// teacc 98.85 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0079, -0.0239, 0.0668, ..., -0.0047, -0.0359, -0.0261], + [-0.0585, -0.0420, -0.0689, ..., 0.0086, 0.0423, 0.0991], + [-0.0210, -0.0221, -0.0262, ..., -0.0364, 0.0203, -0.0781], + ..., + [-0.0756, -0.0719, 0.0256, ..., -0.0827, -0.0456, -0.0749], + [ 0.0527, 0.0270, -0.0406, ..., -0.0747, -0.0089, -0.0605], + [-0.0062, 0.0379, -0.0496, ..., -0.0057, -0.0654, -0.0219]], + device='cuda:0'), grad: tensor([[ 1.0179e-06, -2.1070e-05, 1.3351e-05, ..., -8.9109e-06, + 7.2084e-06, 2.4326e-06], + [ 4.5151e-06, 4.5709e-06, 2.3931e-05, ..., -3.1614e-04, + -9.8944e-06, -3.0661e-04], + [ 2.3589e-05, 2.0772e-05, 1.2791e-04, ..., 7.0333e-05, + 1.4350e-05, 2.8923e-05], + ..., + [ 1.0885e-05, 1.9297e-05, 4.4733e-05, ..., 2.5606e-04, + 3.5077e-05, 2.2399e-04], + [-3.2540e-06, 1.2085e-05, 9.6440e-05, ..., 3.8326e-05, + 2.2024e-05, 1.1459e-05], + [ 1.2359e-06, 9.0674e-06, 2.5630e-05, ..., 2.1651e-05, + 1.5646e-05, 1.4581e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0240, -0.0231, -0.0197, 0.0006, -0.0053, 0.0008, -0.0025, -0.0158, + 0.0234, -0.0310], device='cuda:0'), grad: tensor([ 8.2850e-05, -1.6165e-03, 2.8849e-04, 5.9456e-05, 4.1455e-05, + 6.0320e-05, -6.1941e-04, 1.3695e-03, 1.9467e-04, 1.3888e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 219.54, cls_loss 0.0090 cls_loss_mapping 0.0255 cls_loss_causal 0.7274 re_mapping 0.0135 re_causal 0.0438 /// teacc 98.95 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0080, -0.0236, 0.0678, ..., -0.0041, -0.0360, -0.0269], + [-0.0590, -0.0429, -0.0691, ..., 0.0089, 0.0425, 0.1015], + [-0.0215, -0.0229, -0.0266, ..., -0.0369, 0.0203, -0.0794], + ..., + [-0.0766, -0.0726, 0.0260, ..., -0.0843, -0.0461, -0.0779], + [ 0.0531, 0.0272, -0.0412, ..., -0.0751, -0.0086, -0.0605], + [-0.0054, 0.0384, -0.0500, ..., -0.0049, -0.0657, -0.0216]], + device='cuda:0'), grad: tensor([[ 2.0698e-05, 9.7692e-05, 6.9916e-05, ..., 7.4148e-05, + 3.9697e-05, 4.5449e-06], + [ 9.9018e-06, 1.1176e-05, -5.7407e-06, ..., -1.5092e-04, + 3.1739e-05, -1.7297e-04], + [ 1.8939e-05, 1.7151e-05, 1.1109e-05, ..., 2.8476e-05, + -1.9753e-04, 1.1832e-05], + ..., + [ 1.6198e-05, 2.2545e-05, 9.9018e-06, ..., 3.7014e-05, + -7.7724e-05, 1.7866e-05], + [-2.0251e-05, 1.6630e-05, 5.3257e-05, ..., 1.0252e-04, + 9.8169e-05, 2.5272e-05], + [-1.4734e-04, -1.1712e-04, 4.1425e-05, ..., -4.8429e-05, + 1.5974e-04, 1.3500e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([-2.3655e-02, -2.2926e-02, -1.9637e-02, 8.4637e-04, -5.6999e-03, + 6.9748e-05, -2.5480e-03, -1.6004e-02, 2.3440e-02, -3.0521e-02], + device='cuda:0'), grad: tensor([ 2.8539e-04, 5.7125e-04, -7.5388e-04, 3.0327e-04, 2.2328e-04, + -1.9205e-04, -2.8431e-05, -1.0834e-03, 3.9792e-04, 2.7776e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 219.69, cls_loss 0.0064 cls_loss_mapping 0.0209 cls_loss_causal 0.7112 re_mapping 0.0135 re_causal 0.0442 /// teacc 98.93 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0081, -0.0233, 0.0680, ..., -0.0039, -0.0363, -0.0270], + [-0.0591, -0.0434, -0.0694, ..., 0.0098, 0.0425, 0.1034], + [-0.0219, -0.0234, -0.0271, ..., -0.0373, 0.0206, -0.0795], + ..., + [-0.0771, -0.0732, 0.0258, ..., -0.0854, -0.0464, -0.0793], + [ 0.0533, 0.0273, -0.0415, ..., -0.0757, -0.0086, -0.0610], + [-0.0051, 0.0385, -0.0505, ..., -0.0044, -0.0663, -0.0220]], + device='cuda:0'), grad: tensor([[ 8.3148e-06, -2.2069e-05, -7.3612e-05, ..., -6.2168e-05, + 4.7050e-06, -2.1979e-06], + [ 1.7658e-05, 3.3557e-05, 3.0398e-06, ..., 6.3479e-06, + 9.0972e-06, -1.2264e-05], + [ 3.3617e-05, 4.1902e-05, 4.7646e-06, ..., 1.6302e-05, + 1.3866e-05, 4.3102e-06], + ..., + [ 7.1824e-06, 1.9774e-05, 6.3814e-06, ..., 6.5044e-06, + 1.1310e-05, 1.5805e-06], + [-3.5524e-05, -3.3714e-06, 1.0997e-05, ..., 3.5763e-05, + -4.5091e-05, 2.4643e-06], + [ 1.2644e-05, 9.2447e-05, 2.9027e-05, ..., 2.5600e-05, + 3.4153e-05, 7.7952e-07]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0240, -0.0230, -0.0194, 0.0009, -0.0058, 0.0002, -0.0023, -0.0160, + 0.0233, -0.0306], device='cuda:0'), grad: tensor([-6.3300e-05, 4.5836e-05, 3.9309e-05, -2.5654e-04, 5.1141e-05, + 4.5925e-05, 1.4448e-04, -6.2346e-05, -4.8339e-05, 1.0449e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 219.65, cls_loss 0.0076 cls_loss_mapping 0.0190 cls_loss_causal 0.6688 re_mapping 0.0134 re_causal 0.0424 /// teacc 98.93 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0081, -0.0230, 0.0684, ..., -0.0036, -0.0364, -0.0272], + [-0.0595, -0.0443, -0.0708, ..., 0.0098, 0.0424, 0.1044], + [-0.0225, -0.0247, -0.0280, ..., -0.0377, 0.0207, -0.0810], + ..., + [-0.0776, -0.0739, 0.0259, ..., -0.0863, -0.0467, -0.0809], + [ 0.0537, 0.0275, -0.0419, ..., -0.0761, -0.0085, -0.0632], + [-0.0047, 0.0386, -0.0509, ..., -0.0039, -0.0668, -0.0223]], + device='cuda:0'), grad: tensor([[ 3.8934e-04, 7.6675e-04, 3.7122e-04, ..., 4.4203e-04, + 1.4961e-05, 1.9278e-06], + [ 1.1468e-04, 1.5199e-04, 5.5544e-06, ..., -8.2552e-06, + 9.3132e-06, -7.4685e-05], + [ 1.6046e-04, 3.0398e-04, 1.2815e-04, ..., 1.7822e-04, + -3.8624e-05, 1.2219e-05], + ..., + [ 8.2135e-05, 1.4842e-04, 5.4121e-05, ..., 9.3222e-05, + 2.5243e-05, 8.9929e-06], + [ 9.7871e-05, 1.5128e-04, 3.8177e-05, ..., 1.0896e-04, + 3.1412e-05, 1.2599e-05], + [-7.2908e-04, -1.4830e-03, -7.5436e-04, ..., -8.4925e-04, + 4.0233e-05, 1.5393e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0235, -0.0234, -0.0194, 0.0004, -0.0058, 0.0011, -0.0021, -0.0161, + 0.0232, -0.0308], device='cuda:0'), grad: tensor([ 1.8826e-03, 4.1461e-04, 4.4489e-04, -6.3944e-04, 5.1022e-04, + 5.3740e-04, -9.5427e-05, 2.0289e-04, 3.4857e-04, -3.6087e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 219.79, cls_loss 0.0065 cls_loss_mapping 0.0208 cls_loss_causal 0.6651 re_mapping 0.0138 re_causal 0.0421 /// teacc 98.90 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0083, -0.0230, 0.0686, ..., -0.0036, -0.0367, -0.0272], + [-0.0597, -0.0448, -0.0714, ..., 0.0103, 0.0423, 0.1062], + [-0.0231, -0.0254, -0.0286, ..., -0.0387, 0.0206, -0.0816], + ..., + [-0.0781, -0.0745, 0.0258, ..., -0.0872, -0.0469, -0.0833], + [ 0.0538, 0.0272, -0.0422, ..., -0.0767, -0.0084, -0.0636], + [-0.0042, 0.0389, -0.0511, ..., -0.0032, -0.0671, -0.0227]], + device='cuda:0'), grad: tensor([[-8.6501e-06, -6.0380e-05, -9.2983e-05, ..., -9.6381e-05, + -3.2177e-07, 5.5740e-07], + [ 7.7486e-06, 2.0117e-05, 3.1948e-05, ..., 2.3142e-05, + -1.3977e-05, -2.9653e-05], + [ 1.3582e-05, 2.2814e-05, 4.0144e-05, ..., 4.5240e-05, + 1.3165e-05, 1.0341e-05], + ..., + [ 1.8090e-05, 1.8492e-05, 9.2536e-06, ..., 2.5123e-05, + 1.4573e-05, 8.7917e-06], + [-1.4432e-05, 5.9694e-05, 4.9442e-05, ..., 3.2246e-05, + 8.0541e-06, 4.0047e-06], + [-1.3217e-05, 5.6028e-06, 1.4469e-05, ..., -1.7099e-06, + 1.4879e-05, 1.3448e-06]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0238, -0.0237, -0.0196, 0.0006, -0.0056, 0.0013, -0.0018, -0.0159, + 0.0230, -0.0310], device='cuda:0'), grad: tensor([-1.3018e-04, 8.2493e-05, 3.5501e-04, 2.4629e-04, 2.2411e-04, + 1.6987e-05, -1.8871e-04, -8.1348e-04, 1.3316e-04, 7.4863e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 219.67, cls_loss 0.0075 cls_loss_mapping 0.0210 cls_loss_causal 0.6638 re_mapping 0.0132 re_causal 0.0412 /// teacc 98.72 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0083, -0.0227, 0.0685, ..., -0.0036, -0.0369, -0.0275], + [-0.0599, -0.0458, -0.0725, ..., 0.0107, 0.0424, 0.1094], + [-0.0234, -0.0260, -0.0292, ..., -0.0394, 0.0204, -0.0840], + ..., + [-0.0785, -0.0750, 0.0256, ..., -0.0882, -0.0472, -0.0866], + [ 0.0541, 0.0276, -0.0424, ..., -0.0772, -0.0080, -0.0647], + [-0.0037, 0.0392, -0.0516, ..., -0.0028, -0.0675, -0.0232]], + device='cuda:0'), grad: tensor([[-2.4751e-05, -6.3539e-05, -1.0300e-04, ..., -9.5367e-05, + 7.7933e-06, 7.7561e-06], + [ 1.7956e-06, 5.4747e-05, 2.9713e-05, ..., -5.4836e-05, + -5.7220e-05, -8.0466e-05], + [ 1.0401e-05, 9.8884e-05, 6.3181e-05, ..., 6.1512e-05, + 2.4408e-05, 2.0415e-05], + ..., + [ 1.0543e-05, 7.6056e-05, 5.6356e-05, ..., 3.4064e-05, + 1.8284e-05, 7.4208e-06], + [ 4.7497e-06, 3.7611e-05, 2.7165e-05, ..., 4.6164e-05, + 1.5244e-05, 2.4289e-05], + [-1.3053e-05, 4.0144e-05, 5.2720e-05, ..., 2.1353e-05, + 2.7299e-05, 3.1162e-06]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0240, -0.0238, -0.0200, 0.0010, -0.0057, 0.0005, -0.0009, -0.0155, + 0.0234, -0.0313], device='cuda:0'), grad: tensor([-1.0997e-04, 3.5077e-05, -3.1292e-05, 4.8256e-04, 2.5049e-05, + -1.0834e-03, 7.8201e-05, 3.2711e-04, 1.3494e-04, 1.4162e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 219.55, cls_loss 0.0064 cls_loss_mapping 0.0165 cls_loss_causal 0.6787 re_mapping 0.0123 re_causal 0.0389 /// teacc 98.86 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0083, -0.0219, 0.0692, ..., -0.0031, -0.0372, -0.0277], + [-0.0601, -0.0464, -0.0731, ..., 0.0113, 0.0426, 0.1110], + [-0.0239, -0.0270, -0.0295, ..., -0.0401, 0.0207, -0.0850], + ..., + [-0.0790, -0.0756, 0.0258, ..., -0.0892, -0.0475, -0.0880], + [ 0.0545, 0.0279, -0.0426, ..., -0.0775, -0.0076, -0.0649], + [-0.0032, 0.0395, -0.0521, ..., -0.0019, -0.0680, -0.0234]], + device='cuda:0'), grad: tensor([[ 4.3996e-06, 9.2834e-06, 2.2173e-05, ..., 1.4424e-05, + 1.1712e-05, 1.0937e-05], + [ 7.8082e-06, 5.3719e-06, -2.4140e-05, ..., -1.2016e-04, + -1.0955e-04, -1.4472e-04], + [ 6.3740e-06, 1.0520e-05, 2.5108e-05, ..., 3.3468e-05, + 3.5495e-05, 2.5719e-05], + ..., + [ 2.0161e-05, -8.0228e-05, -3.0589e-04, ..., -4.0674e-04, + 1.7807e-05, 1.2010e-05], + [ 6.4790e-05, 9.7513e-05, 4.8637e-05, ..., 9.4652e-05, + 2.2411e-05, 3.2157e-05], + [-2.1040e-04, -4.1246e-05, 2.9945e-04, ..., 2.4652e-04, + 1.1928e-05, 8.0243e-06]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0238, -0.0236, -0.0198, 0.0007, -0.0063, 0.0009, -0.0013, -0.0158, + 0.0236, -0.0311], device='cuda:0'), grad: tensor([ 9.1970e-05, -8.4937e-05, 1.2815e-04, 2.5439e-04, 2.7943e-04, + -4.4823e-04, 3.5071e-04, -2.3460e-03, 2.9993e-04, 1.4753e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 219.52, cls_loss 0.0088 cls_loss_mapping 0.0211 cls_loss_causal 0.6786 re_mapping 0.0133 re_causal 0.0410 /// teacc 98.94 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0087, -0.0220, 0.0692, ..., -0.0032, -0.0375, -0.0279], + [-0.0605, -0.0474, -0.0739, ..., 0.0112, 0.0427, 0.1140], + [-0.0243, -0.0276, -0.0298, ..., -0.0405, 0.0209, -0.0864], + ..., + [-0.0796, -0.0763, 0.0258, ..., -0.0902, -0.0478, -0.0902], + [ 0.0549, 0.0281, -0.0430, ..., -0.0780, -0.0075, -0.0663], + [-0.0029, 0.0397, -0.0525, ..., -0.0009, -0.0685, -0.0237]], + device='cuda:0'), grad: tensor([[ 3.7458e-06, -1.6880e-04, -1.3943e-03, ..., -8.2588e-04, + 1.5751e-05, 3.8883e-07], + [ 1.8522e-05, 2.9713e-05, 4.1366e-05, ..., 3.9011e-05, + 2.2292e-05, -2.1741e-05], + [ 2.8983e-05, 5.1439e-05, 4.9353e-05, ..., 7.3314e-05, + 3.9220e-05, 4.3772e-06], + ..., + [ 2.0891e-05, 3.4213e-05, 6.2883e-05, ..., 5.7310e-05, + 4.8667e-05, 4.6864e-06], + [-1.5944e-06, 2.1234e-05, 1.0943e-04, ..., 6.9141e-05, + 1.1407e-05, 2.0154e-06], + [-6.4000e-06, 1.2510e-05, 3.4153e-05, ..., 1.9103e-05, + 4.3213e-05, 3.6675e-06]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0242, -0.0239, -0.0198, 0.0009, -0.0060, 0.0006, -0.0009, -0.0164, + 0.0237, -0.0305], device='cuda:0'), grad: tensor([-1.8091e-03, 1.5533e-04, -1.1140e-04, -2.7251e-04, -9.3877e-05, + 6.8760e-04, 8.5878e-04, 3.0589e-04, 1.7691e-04, 1.0586e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 219.32, cls_loss 0.0068 cls_loss_mapping 0.0144 cls_loss_causal 0.6500 re_mapping 0.0123 re_causal 0.0372 /// teacc 98.80 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0087, -0.0216, 0.0695, ..., -0.0029, -0.0377, -0.0283], + [-0.0606, -0.0479, -0.0743, ..., 0.0116, 0.0428, 0.1183], + [-0.0247, -0.0283, -0.0303, ..., -0.0410, 0.0211, -0.0884], + ..., + [-0.0801, -0.0771, 0.0258, ..., -0.0905, -0.0478, -0.0925], + [ 0.0551, 0.0281, -0.0435, ..., -0.0788, -0.0076, -0.0681], + [-0.0023, 0.0402, -0.0529, ..., -0.0005, -0.0691, -0.0246]], + device='cuda:0'), grad: tensor([[ 2.3991e-06, -9.7081e-06, -3.1620e-05, ..., -2.0593e-05, + 7.9572e-06, -2.2911e-06], + [ 2.7642e-06, 4.5113e-06, 6.5900e-06, ..., -5.8293e-05, + -7.1883e-05, -1.1563e-04], + [ 7.0520e-06, 1.1429e-05, 1.1750e-05, ..., 2.0593e-05, + -7.6056e-05, 2.5019e-05], + ..., + [ 6.9916e-05, 7.9513e-05, 1.5363e-05, ..., 6.3896e-05, + 1.1706e-04, 1.2800e-05], + [ 5.2601e-05, 7.9334e-05, 9.4593e-05, ..., 6.4075e-05, + 1.2648e-04, 4.5598e-05], + [-7.3493e-05, -5.9098e-05, 1.6773e-04, ..., -8.7142e-05, + 1.7667e-04, 4.5560e-06]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0241, -0.0243, -0.0196, 0.0006, -0.0062, 0.0006, -0.0007, -0.0158, + 0.0235, -0.0306], device='cuda:0'), grad: tensor([-1.6123e-05, -7.8917e-05, 1.3332e-03, 8.8692e-05, -8.4639e-04, + -4.9919e-06, 7.1406e-05, -1.1978e-03, 4.3273e-04, 2.2030e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 219.75, cls_loss 0.0063 cls_loss_mapping 0.0187 cls_loss_causal 0.6361 re_mapping 0.0121 re_causal 0.0385 /// teacc 98.93 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0088, -0.0209, 0.0701, ..., -0.0025, -0.0379, -0.0284], + [-0.0609, -0.0485, -0.0751, ..., 0.0121, 0.0425, 0.1199], + [-0.0251, -0.0290, -0.0307, ..., -0.0415, 0.0213, -0.0895], + ..., + [-0.0808, -0.0780, 0.0257, ..., -0.0917, -0.0481, -0.0942], + [ 0.0556, 0.0282, -0.0438, ..., -0.0796, -0.0072, -0.0700], + [-0.0021, 0.0402, -0.0535, ..., -0.0001, -0.0699, -0.0249]], + device='cuda:0'), grad: tensor([[ 1.6093e-04, 1.1915e-04, 1.2165e-04, ..., 2.4176e-04, + 9.0122e-05, 1.0859e-06], + [ 9.4995e-06, 1.0520e-05, 2.5347e-05, ..., 2.0564e-06, + -1.8865e-05, -4.1395e-05], + [ 1.1392e-05, 2.1368e-05, 4.3750e-05, ..., 6.2168e-05, + -1.9044e-05, 8.8215e-06], + ..., + [ 2.8163e-05, 3.7491e-05, 2.7537e-05, ..., 8.3148e-06, + -3.5852e-05, 7.1377e-06], + [ 9.6917e-05, 9.7811e-05, 8.9109e-05, ..., 1.2189e-04, + 4.4972e-05, 1.2957e-05], + [-3.7146e-04, -3.1996e-04, -2.1636e-05, ..., -2.2292e-04, + -3.9607e-05, 2.4699e-06]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0237, -0.0245, -0.0194, 0.0004, -0.0060, 0.0010, -0.0008, -0.0162, + 0.0238, -0.0309], device='cuda:0'), grad: tensor([ 4.6015e-04, 7.4208e-05, -2.4056e-04, 4.8065e-04, 7.8964e-04, + 3.2878e-04, -1.3685e-03, -9.1600e-04, 3.1495e-04, 7.5161e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 52---------------------------------------------------- +epoch 52, time 220.30, cls_loss 0.0051 cls_loss_mapping 0.0162 cls_loss_causal 0.6490 re_mapping 0.0119 re_causal 0.0383 /// teacc 99.02 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0090, -0.0206, 0.0707, ..., -0.0020, -0.0382, -0.0287], + [-0.0612, -0.0490, -0.0754, ..., 0.0126, 0.0427, 0.1226], + [-0.0255, -0.0296, -0.0311, ..., -0.0420, 0.0215, -0.0909], + ..., + [-0.0812, -0.0787, 0.0253, ..., -0.0922, -0.0483, -0.0960], + [ 0.0560, 0.0284, -0.0441, ..., -0.0799, -0.0071, -0.0707], + [-0.0017, 0.0404, -0.0538, ..., 0.0007, -0.0705, -0.0256]], + device='cuda:0'), grad: tensor([[ 5.7817e-06, -2.0117e-06, 3.0696e-06, ..., 1.4432e-05, + 3.9302e-06, 2.7493e-06], + [ 7.1190e-06, 6.9849e-06, 3.3081e-06, ..., -5.3495e-05, + -4.4137e-05, -1.1235e-04], + [ 1.9923e-05, 1.8567e-05, 2.0236e-05, ..., 2.3380e-05, + 1.8463e-05, 9.5814e-06], + ..., + [ 1.9222e-05, 1.9565e-05, 9.7305e-06, ..., 3.0547e-05, + 2.0176e-05, 2.6956e-05], + [-8.4341e-06, -4.7572e-06, 1.1437e-05, ..., 3.1084e-05, + 3.8967e-06, 2.1651e-05], + [-7.5293e-04, -7.4911e-04, -1.9133e-04, ..., -6.4373e-04, + -1.5557e-05, 1.7926e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0237, -0.0246, -0.0193, 0.0004, -0.0056, 0.0010, -0.0013, -0.0161, + 0.0238, -0.0310], device='cuda:0'), grad: tensor([ 3.3587e-05, -1.2922e-04, 1.0979e-04, 3.9726e-05, 1.7500e-03, + 9.1612e-05, -1.0628e-04, 7.8857e-05, 7.9051e-06, -1.8768e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 220.04, cls_loss 0.0042 cls_loss_mapping 0.0147 cls_loss_causal 0.6571 re_mapping 0.0121 re_causal 0.0385 /// teacc 98.86 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0092, -0.0207, 0.0706, ..., -0.0021, -0.0384, -0.0291], + [-0.0615, -0.0496, -0.0758, ..., 0.0133, 0.0430, 0.1258], + [-0.0258, -0.0298, -0.0313, ..., -0.0424, 0.0217, -0.0924], + ..., + [-0.0815, -0.0792, 0.0252, ..., -0.0931, -0.0490, -0.0983], + [ 0.0563, 0.0284, -0.0444, ..., -0.0802, -0.0070, -0.0724], + [-0.0009, 0.0409, -0.0539, ..., 0.0013, -0.0710, -0.0259]], + device='cuda:0'), grad: tensor([[ 6.0312e-06, -1.7196e-05, -2.4885e-05, ..., -2.3425e-05, + 6.1169e-06, 2.4168e-07], + [ 7.5400e-06, 1.1981e-05, 9.0450e-06, ..., 6.2585e-07, + 7.5959e-06, -1.2323e-05], + [ 1.5199e-04, 2.9087e-04, 1.1757e-05, ..., 1.9407e-04, + 1.5020e-04, 1.3644e-06], + ..., + [ 1.4409e-05, 1.7107e-05, -1.5393e-05, ..., 1.4067e-05, + 2.1368e-05, 1.4547e-06], + [-2.1666e-05, 1.0204e-04, 8.1062e-05, ..., 5.7906e-05, + 1.0706e-05, 4.7050e-06], + [-3.4511e-05, 2.1458e-06, 2.3186e-05, ..., -7.5921e-06, + 4.7386e-05, 1.3011e-06]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0241, -0.0241, -0.0190, 0.0005, -0.0060, 0.0006, -0.0012, -0.0160, + 0.0236, -0.0309], device='cuda:0'), grad: tensor([-1.7628e-05, 6.3121e-05, 6.0749e-04, -6.3705e-04, -9.1374e-05, + -2.6393e-04, 1.8847e-04, -6.5565e-05, 1.2082e-04, 9.4950e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 219.87, cls_loss 0.0055 cls_loss_mapping 0.0173 cls_loss_causal 0.6553 re_mapping 0.0121 re_causal 0.0389 /// teacc 98.96 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0093, -0.0204, 0.0708, ..., -0.0020, -0.0387, -0.0294], + [-0.0617, -0.0498, -0.0759, ..., 0.0139, 0.0433, 0.1283], + [-0.0263, -0.0310, -0.0315, ..., -0.0435, 0.0215, -0.0939], + ..., + [-0.0822, -0.0798, 0.0252, ..., -0.0935, -0.0489, -0.0995], + [ 0.0568, 0.0286, -0.0447, ..., -0.0807, -0.0071, -0.0742], + [-0.0009, 0.0406, -0.0547, ..., 0.0012, -0.0716, -0.0263]], + device='cuda:0'), grad: tensor([[ 3.3099e-06, -1.0893e-05, -1.7539e-05, ..., -1.1802e-05, + 4.4741e-06, 1.3085e-06], + [-1.7494e-05, 9.0431e-07, 5.2527e-06, ..., -4.5776e-05, + -3.5226e-05, -5.9813e-05], + [-1.3269e-05, 6.7875e-06, 9.2328e-05, ..., 5.8115e-06, + 1.9133e-04, 4.5858e-06], + ..., + [ 7.0632e-06, -1.0900e-05, -7.8157e-06, ..., 8.1956e-06, + 1.3940e-05, 4.2357e-06], + [ 1.1735e-06, -2.0675e-06, 1.2860e-05, ..., 3.6716e-05, + 2.2292e-05, 3.3468e-05], + [-6.4075e-05, -3.9905e-05, 3.4720e-05, ..., -3.9726e-05, + 2.4468e-05, 1.1899e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0243, -0.0239, -0.0190, 0.0010, -0.0056, 0.0005, -0.0011, -0.0162, + 0.0237, -0.0316], device='cuda:0'), grad: tensor([ 4.2394e-06, -1.0872e-04, 1.7130e-04, 9.3341e-05, -3.5048e-04, + 4.8488e-05, 3.8117e-05, -1.0741e-04, 1.1951e-04, 9.1851e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 219.69, cls_loss 0.0049 cls_loss_mapping 0.0164 cls_loss_causal 0.6450 re_mapping 0.0117 re_causal 0.0386 /// teacc 98.74 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0097, -0.0204, 0.0715, ..., -0.0011, -0.0384, -0.0279], + [-0.0620, -0.0503, -0.0767, ..., 0.0142, 0.0435, 0.1308], + [-0.0269, -0.0315, -0.0320, ..., -0.0441, 0.0213, -0.0957], + ..., + [-0.0826, -0.0803, 0.0254, ..., -0.0948, -0.0493, -0.1010], + [ 0.0573, 0.0292, -0.0452, ..., -0.0814, -0.0068, -0.0752], + [-0.0002, 0.0410, -0.0551, ..., 0.0021, -0.0721, -0.0267]], + device='cuda:0'), grad: tensor([[ 1.8135e-05, -6.1274e-05, -9.1374e-05, ..., -8.1718e-05, + 2.3514e-05, 1.2433e-06], + [ 3.1590e-05, 2.2665e-05, 9.8795e-06, ..., 1.4544e-05, + 3.4273e-05, 5.2750e-06], + [ 2.0278e-04, 9.6798e-05, 5.1111e-05, ..., 5.3138e-05, + 2.1362e-04, 1.5438e-05], + ..., + [ 1.4782e-05, 2.3350e-05, 9.4026e-06, ..., 2.6599e-05, + 6.2287e-06, 8.3307e-07], + [-3.0041e-04, -1.2040e-04, -1.8850e-05, ..., -1.2167e-05, + -3.4547e-04, -2.8268e-05], + [-4.3201e-04, -2.2590e-04, 1.4037e-05, ..., -3.5501e-04, + -1.7929e-04, 8.3633e-07]], device='cuda:0') +Epoch 57, bias, value: tensor([-2.3827e-02, -2.2864e-02, -1.9049e-02, 9.7030e-04, -5.4276e-03, + -8.1272e-05, -1.7233e-03, -1.6652e-02, 2.3911e-02, -3.1519e-02], + device='cuda:0'), grad: tensor([-1.9705e-04, 1.5044e-04, 4.8542e-04, 1.2708e-04, 8.3303e-04, + 1.5393e-05, 1.0276e-04, -5.1588e-05, -6.6566e-04, -8.0013e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 219.96, cls_loss 0.0065 cls_loss_mapping 0.0180 cls_loss_causal 0.6375 re_mapping 0.0119 re_causal 0.0373 /// teacc 98.92 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0097, -0.0202, 0.0715, ..., -0.0015, -0.0388, -0.0288], + [-0.0623, -0.0509, -0.0770, ..., 0.0146, 0.0437, 0.1328], + [-0.0273, -0.0323, -0.0328, ..., -0.0446, 0.0212, -0.0965], + ..., + [-0.0833, -0.0807, 0.0255, ..., -0.0948, -0.0494, -0.1019], + [ 0.0577, 0.0292, -0.0455, ..., -0.0820, -0.0067, -0.0762], + [ 0.0004, 0.0411, -0.0553, ..., 0.0031, -0.0724, -0.0273]], + device='cuda:0'), grad: tensor([[ 1.8300e-06, -1.8835e-05, -5.9456e-05, ..., -4.9740e-05, + 5.0925e-06, 7.2364e-07], + [ 1.2470e-06, 2.9784e-06, 5.1856e-06, ..., -9.8348e-06, + -1.1191e-05, -2.6301e-05], + [ 6.0685e-06, 1.2673e-05, 3.0130e-05, ..., 2.4915e-05, + 1.5631e-05, 4.8019e-06], + ..., + [ 1.8813e-06, 5.6811e-06, 7.5847e-06, ..., 3.8445e-06, + 1.0788e-05, 3.6675e-06], + [ 2.1413e-05, 9.6262e-05, 4.8339e-05, ..., 2.0385e-05, + 3.5882e-05, 7.3463e-06], + [ 1.0103e-05, 2.8223e-05, 2.6897e-05, ..., 2.3395e-05, + 9.6485e-06, 2.1663e-06]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0243, -0.0232, -0.0197, 0.0009, -0.0056, 0.0002, -0.0017, -0.0167, + 0.0242, -0.0307], device='cuda:0'), grad: tensor([-9.8348e-05, -1.3828e-05, 6.1095e-05, 7.8559e-05, 9.1344e-06, + -2.7823e-04, -3.4839e-05, -1.2434e-04, 2.1279e-04, 1.8799e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 219.55, cls_loss 0.0061 cls_loss_mapping 0.0173 cls_loss_causal 0.6703 re_mapping 0.0111 re_causal 0.0355 /// teacc 98.87 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0099, -0.0197, 0.0718, ..., -0.0013, -0.0390, -0.0289], + [-0.0631, -0.0527, -0.0780, ..., 0.0141, 0.0437, 0.1344], + [-0.0278, -0.0332, -0.0335, ..., -0.0451, 0.0218, -0.0972], + ..., + [-0.0837, -0.0814, 0.0256, ..., -0.0957, -0.0500, -0.1036], + [ 0.0581, 0.0292, -0.0456, ..., -0.0823, -0.0069, -0.0768], + [ 0.0009, 0.0416, -0.0554, ..., 0.0045, -0.0725, -0.0278]], + device='cuda:0'), grad: tensor([[ 2.4781e-05, 4.1813e-05, -1.5348e-05, ..., 1.9409e-06, + 2.0899e-06, 1.7881e-07], + [ 1.2651e-05, 2.6017e-05, 8.7768e-06, ..., -2.1793e-06, + -1.1623e-05, -2.1905e-05], + [ 3.5524e-05, 7.5996e-05, 1.1832e-05, ..., 2.3752e-05, + -5.2974e-06, 1.3728e-06], + ..., + [ 5.7280e-05, 1.0252e-04, 2.5809e-05, ..., 3.4779e-05, + 2.1219e-05, 8.1584e-06], + [ 1.3363e-04, 2.7490e-04, 6.0171e-05, ..., 9.3460e-05, + 5.6103e-06, 1.8915e-06], + [-5.7012e-05, 5.2482e-05, 7.7844e-05, ..., -4.1693e-05, + 4.4465e-05, 3.8594e-06]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0244, -0.0233, -0.0194, 0.0010, -0.0058, 0.0003, -0.0023, -0.0161, + 0.0237, -0.0306], device='cuda:0'), grad: tensor([ 6.7174e-05, 2.3156e-05, 1.0192e-04, -1.9588e-03, -3.3081e-05, + 9.0742e-04, 4.9591e-05, 1.8549e-04, 4.8208e-04, 1.7452e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 219.55, cls_loss 0.0051 cls_loss_mapping 0.0149 cls_loss_causal 0.6408 re_mapping 0.0113 re_causal 0.0358 /// teacc 98.91 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0099, -0.0194, 0.0723, ..., -0.0011, -0.0393, -0.0290], + [-0.0635, -0.0534, -0.0787, ..., 0.0143, 0.0438, 0.1361], + [-0.0284, -0.0339, -0.0340, ..., -0.0458, 0.0221, -0.0986], + ..., + [-0.0841, -0.0821, 0.0258, ..., -0.0961, -0.0504, -0.1045], + [ 0.0584, 0.0293, -0.0459, ..., -0.0826, -0.0066, -0.0777], + [ 0.0013, 0.0415, -0.0559, ..., 0.0049, -0.0729, -0.0284]], + device='cuda:0'), grad: tensor([[ 5.3197e-06, -9.8646e-06, -1.0699e-05, ..., 5.4017e-06, + 1.2174e-05, 8.9221e-07], + [ 2.8778e-06, 3.0939e-06, 3.8892e-06, ..., -4.9174e-05, + -3.0458e-05, -5.9962e-05], + [ 6.9961e-06, 8.9779e-06, 1.1444e-05, ..., 1.1474e-05, + -1.1325e-04, 9.5442e-06], + ..., + [ 2.0117e-05, 1.7807e-05, 4.3884e-06, ..., 2.8327e-05, + 2.6435e-05, 1.3337e-05], + [ 4.8935e-05, 5.4628e-05, 4.7460e-06, ..., 6.6817e-05, + 3.6120e-05, 8.3894e-06], + [-1.2046e-04, -9.4652e-05, 1.4618e-05, ..., -7.5102e-05, + 1.7583e-05, 1.4409e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0241, -0.0236, -0.0192, 0.0012, -0.0062, 0.0006, -0.0030, -0.0162, + 0.0239, -0.0304], device='cuda:0'), grad: tensor([ 3.3617e-05, -1.1122e-04, -3.3259e-04, 1.1188e-04, 2.4050e-05, + 4.9695e-06, 4.8876e-05, 6.7830e-05, 2.7442e-04, -1.2165e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 219.59, cls_loss 0.0045 cls_loss_mapping 0.0145 cls_loss_causal 0.6248 re_mapping 0.0115 re_causal 0.0359 /// teacc 98.83 lr 0.00010000 +Epoch 61, weight, value: tensor([[-1.0150e-02, -1.9187e-02, 7.2744e-02, ..., -7.7151e-05, + -3.8802e-02, -2.9051e-02], + [-6.3748e-02, -5.3834e-02, -7.9327e-02, ..., 1.4305e-02, + 4.3806e-02, 1.3715e-01], + [-2.9081e-02, -3.4738e-02, -3.4402e-02, ..., -4.6660e-02, + 2.1635e-02, -9.9023e-02], + ..., + [-8.4705e-02, -8.2390e-02, 2.5934e-02, ..., -9.6079e-02, + -5.0727e-02, -1.0554e-01], + [ 5.9385e-02, 3.0106e-02, -4.6109e-02, ..., -8.2789e-02, + -6.0182e-03, -7.8373e-02], + [ 1.4368e-03, 4.1156e-02, -5.6534e-02, ..., 4.7414e-03, + -7.3724e-02, -2.8866e-02]], device='cuda:0'), grad: tensor([[ 1.6958e-05, 2.7958e-06, 1.0151e-04, ..., 8.7321e-05, + 6.8359e-06, 2.1653e-07], + [ 5.9456e-06, 3.4571e-06, 1.6004e-05, ..., 2.3693e-06, + -2.1547e-05, -1.0937e-05], + [ 2.8476e-05, 1.7628e-05, 1.8492e-05, ..., 3.0294e-05, + 3.0115e-05, 1.7434e-06], + ..., + [ 2.5764e-05, 1.3396e-05, -8.6380e-07, ..., 2.2173e-05, + 2.1741e-05, 6.5304e-06], + [-1.5125e-05, 2.8405e-06, 6.2346e-05, ..., 9.5665e-05, + -3.3915e-05, 6.2166e-07], + [-1.9819e-05, -8.4043e-06, 1.0915e-05, ..., -1.7643e-05, + 2.0206e-05, 3.8906e-07]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0234, -0.0239, -0.0199, 0.0006, -0.0060, 0.0003, -0.0025, -0.0152, + 0.0245, -0.0313], device='cuda:0'), grad: tensor([ 3.1590e-04, 1.6810e-06, 1.1373e-04, 2.1718e-06, 1.7500e-04, + 1.4007e-04, -9.5177e-04, -2.3827e-05, 2.0874e-04, 1.8686e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 219.64, cls_loss 0.0068 cls_loss_mapping 0.0181 cls_loss_causal 0.6501 re_mapping 0.0114 re_causal 0.0360 /// teacc 98.86 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0104, -0.0194, 0.0726, ..., -0.0002, -0.0391, -0.0294], + [-0.0640, -0.0538, -0.0799, ..., 0.0149, 0.0439, 0.1385], + [-0.0295, -0.0352, -0.0352, ..., -0.0472, 0.0218, -0.0996], + ..., + [-0.0865, -0.0843, 0.0264, ..., -0.0980, -0.0513, -0.1073], + [ 0.0597, 0.0301, -0.0464, ..., -0.0835, -0.0058, -0.0790], + [ 0.0023, 0.0415, -0.0568, ..., 0.0053, -0.0743, -0.0296]], + device='cuda:0'), grad: tensor([[-1.0766e-06, -2.2948e-05, 5.8442e-05, ..., 1.4052e-05, + 1.6063e-05, -1.2964e-06], + [ 3.0585e-06, 3.8631e-06, 1.2867e-05, ..., -3.0294e-05, + -4.0650e-05, -5.8323e-05], + [ 4.1500e-06, 5.7518e-06, 1.5600e-06, ..., 2.5928e-05, + 1.9178e-05, 3.0994e-05], + ..., + [-4.4368e-06, 4.3437e-06, 3.5949e-06, ..., -3.5465e-06, + 7.9721e-06, 7.0520e-06], + [ 3.6985e-05, 7.2002e-05, 1.8373e-05, ..., 3.4928e-05, + 2.6554e-05, 7.5065e-06], + [ 4.9025e-06, 2.5421e-05, 2.1398e-05, ..., 1.3366e-05, + 1.5318e-05, 3.2634e-06]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0240, -0.0249, -0.0194, 0.0007, -0.0063, 0.0006, -0.0019, -0.0155, + 0.0242, -0.0306], device='cuda:0'), grad: tensor([ 5.6267e-05, -7.1049e-05, 5.9038e-05, 4.6819e-05, 1.0145e-04, + -2.3413e-06, -2.5606e-04, -1.5080e-04, 1.3149e-04, 8.4519e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 219.45, cls_loss 0.0043 cls_loss_mapping 0.0120 cls_loss_causal 0.6512 re_mapping 0.0106 re_causal 0.0352 /// teacc 98.87 lr 0.00010000 +Epoch 63, weight, value: tensor([[-1.0480e-02, -1.8722e-02, 7.2897e-02, ..., -1.2312e-04, + -3.9482e-02, -2.9536e-02], + [-6.4445e-02, -5.4326e-02, -8.0617e-02, ..., 1.5488e-02, + 4.4137e-02, 1.4105e-01], + [-3.0154e-02, -3.6561e-02, -3.6052e-02, ..., -4.7672e-02, + 2.1526e-02, -1.0123e-01], + ..., + [-8.6611e-02, -8.4553e-02, 2.6363e-02, ..., -9.8868e-02, + -5.1532e-02, -1.0894e-01], + [ 5.9679e-02, 2.9708e-02, -4.6802e-02, ..., -8.4162e-02, + -5.8878e-03, -7.9723e-02], + [ 2.6605e-03, 4.1653e-02, -5.7320e-02, ..., 5.7699e-03, + -7.4748e-02, -3.0532e-02]], device='cuda:0'), grad: tensor([[ 2.0146e-05, 2.1979e-05, -1.8450e-06, ..., 1.8522e-05, + 1.4655e-05, 2.7707e-08], + [ 3.0175e-05, 2.4721e-05, 1.0446e-05, ..., 2.7269e-05, + 6.6340e-05, -1.6168e-06], + [ 2.0355e-05, 2.2128e-05, 9.8646e-06, ..., 2.1935e-05, + 1.8388e-05, 1.7392e-07], + ..., + [ 9.5814e-06, 1.0535e-05, 5.7280e-05, ..., 3.0905e-05, + 8.2135e-05, 4.5449e-07], + [-2.7418e-04, -1.8358e-04, 5.0031e-06, ..., -2.1136e-04, + -6.2847e-04, 2.2619e-07], + [ 1.0192e-04, 1.6069e-04, 1.4889e-04, ..., 1.5461e-04, + 2.3055e-04, 2.9616e-07]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0242, -0.0246, -0.0204, 0.0005, -0.0064, 0.0010, -0.0011, -0.0153, + 0.0238, -0.0303], device='cuda:0'), grad: tensor([ 4.8727e-05, 1.3649e-04, 5.0873e-05, 4.8727e-05, -7.4387e-04, + 3.1400e-04, 3.1352e-04, 2.0897e-04, -1.0977e-03, 7.2002e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 219.33, cls_loss 0.0046 cls_loss_mapping 0.0156 cls_loss_causal 0.6318 re_mapping 0.0109 re_causal 0.0342 /// teacc 99.01 lr 0.00010000 +Epoch 64, weight, value: tensor([[-9.6093e-03, -1.8601e-02, 7.3532e-02, ..., -4.1594e-05, + -3.9237e-02, -2.9668e-02], + [-6.4791e-02, -5.4968e-02, -8.1079e-02, ..., 1.6164e-02, + 4.4140e-02, 1.4255e-01], + [-3.0405e-02, -3.6393e-02, -3.6875e-02, ..., -4.8027e-02, + 2.1701e-02, -1.0115e-01], + ..., + [-8.6852e-02, -8.4881e-02, 2.6272e-02, ..., -9.9533e-02, + -5.2002e-02, -1.1082e-01], + [ 6.0135e-02, 2.9950e-02, -4.7212e-02, ..., -8.4199e-02, + -5.4413e-03, -8.0637e-02], + [ 3.7669e-03, 4.2496e-02, -5.7546e-02, ..., 6.1397e-03, + -7.5217e-02, -3.1091e-02]], device='cuda:0'), grad: tensor([[ 1.3374e-05, -2.8163e-06, -9.7752e-06, ..., 2.0206e-05, + 2.6137e-05, 1.5810e-05], + [ 2.5377e-05, -1.1504e-04, -1.2422e-04, ..., -7.8487e-04, + -2.6321e-04, -2.9802e-04], + [ 6.7532e-05, 5.7817e-05, 1.7178e-04, ..., 1.0896e-04, + 1.7679e-04, 4.5180e-05], + ..., + [ 2.1085e-05, 1.2755e-04, -1.0535e-05, ..., 3.7932e-04, + 2.4348e-05, 1.0222e-05], + [ 8.1539e-05, 8.3685e-05, 4.7922e-05, ..., 1.2898e-04, + 1.1778e-04, 7.0453e-05], + [ 5.1670e-06, 1.5587e-05, 2.8566e-05, ..., 3.7014e-05, + 3.1769e-05, 1.2644e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0234, -0.0244, -0.0202, 0.0005, -0.0070, 0.0004, -0.0012, -0.0153, + 0.0240, -0.0302], device='cuda:0'), grad: tensor([ 7.4267e-05, -2.5921e-03, 4.7541e-04, -3.6693e-04, 9.0241e-05, + 9.5010e-05, -8.3745e-06, 1.5745e-03, 4.4703e-04, 2.0981e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 219.60, cls_loss 0.0042 cls_loss_mapping 0.0135 cls_loss_causal 0.6259 re_mapping 0.0106 re_causal 0.0337 /// teacc 98.90 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0098, -0.0184, 0.0737, ..., -0.0003, -0.0394, -0.0298], + [-0.0650, -0.0554, -0.0815, ..., 0.0166, 0.0442, 0.1439], + [-0.0307, -0.0369, -0.0371, ..., -0.0482, 0.0217, -0.1014], + ..., + [-0.0873, -0.0856, 0.0264, ..., -0.1002, -0.0519, -0.1124], + [ 0.0602, 0.0296, -0.0477, ..., -0.0848, -0.0054, -0.0813], + [ 0.0042, 0.0431, -0.0580, ..., 0.0069, -0.0759, -0.0316]], + device='cuda:0'), grad: tensor([[ 1.2800e-05, 1.6615e-06, -2.6673e-06, ..., -4.8019e-06, + 3.0063e-06, 5.5833e-07], + [ 1.2189e-05, 3.6974e-06, 1.0952e-06, ..., -1.5043e-05, + -8.0019e-06, -2.7850e-05], + [ 3.0398e-05, 1.8835e-05, 1.0729e-05, ..., 5.1782e-06, + -3.4943e-06, 3.6936e-06], + ..., + [ 1.0023e-03, 4.3035e-05, -1.6373e-06, ..., 5.9940e-06, + 4.1090e-06, 4.4480e-06], + [ 6.2883e-06, -3.5852e-05, 1.0484e-04, ..., 3.5197e-05, + 5.0157e-05, 9.1866e-06], + [-1.2302e-03, -9.3043e-05, -1.2922e-04, ..., -5.3734e-05, + -1.8418e-05, 2.8666e-06]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0237, -0.0245, -0.0203, 0.0003, -0.0065, 0.0014, -0.0022, -0.0150, + 0.0237, -0.0304], device='cuda:0'), grad: tensor([ 2.5764e-05, 1.7032e-05, 8.5890e-05, 1.7607e-04, 2.0051e-04, + 3.7700e-05, 2.8890e-06, 3.6964e-03, 8.3327e-05, -4.3297e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 219.58, cls_loss 0.0045 cls_loss_mapping 0.0142 cls_loss_causal 0.6020 re_mapping 0.0101 re_causal 0.0313 /// teacc 98.96 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0100, -0.0175, 0.0744, ..., 0.0007, -0.0397, -0.0299], + [-0.0651, -0.0555, -0.0821, ..., 0.0170, 0.0443, 0.1457], + [-0.0310, -0.0373, -0.0376, ..., -0.0486, 0.0217, -0.1021], + ..., + [-0.0885, -0.0863, 0.0266, ..., -0.1012, -0.0523, -0.1141], + [ 0.0604, 0.0296, -0.0483, ..., -0.0853, -0.0051, -0.0823], + [ 0.0054, 0.0435, -0.0588, ..., 0.0072, -0.0768, -0.0322]], + device='cuda:0'), grad: tensor([[ 2.2173e-05, 9.9167e-06, -2.8580e-05, ..., -3.7514e-06, + 2.7157e-06, 7.3016e-07], + [ 2.3946e-05, 1.4909e-05, 6.6608e-06, ..., 3.9488e-06, + -1.9863e-05, -3.9428e-05], + [ 1.3918e-05, 1.4231e-05, 5.7295e-06, ..., 1.3031e-05, + 1.7807e-05, 1.5929e-05], + ..., + [ 2.8849e-05, 2.0012e-05, 1.3441e-05, ..., 2.3514e-05, + 1.2383e-05, 3.4217e-06], + [ 2.4581e-04, 1.5080e-04, 4.5359e-05, ..., 8.6546e-05, + 4.0084e-05, 9.7528e-06], + [-8.2159e-04, -4.8518e-04, -1.9515e-04, ..., -4.8828e-04, + -2.4414e-04, 2.5891e-06]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0232, -0.0245, -0.0200, -0.0003, -0.0063, 0.0014, -0.0024, -0.0153, + 0.0236, -0.0302], device='cuda:0'), grad: tensor([-8.5905e-06, -6.9365e-06, 1.3697e-04, -4.4525e-05, 1.0767e-03, + 1.0121e-04, 1.8775e-05, -2.5898e-05, 5.2738e-04, -1.7748e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 219.59, cls_loss 0.0032 cls_loss_mapping 0.0095 cls_loss_causal 0.6176 re_mapping 0.0101 re_causal 0.0327 /// teacc 99.01 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0103, -0.0172, 0.0747, ..., 0.0010, -0.0400, -0.0300], + [-0.0656, -0.0560, -0.0825, ..., 0.0173, 0.0442, 0.1472], + [-0.0314, -0.0379, -0.0380, ..., -0.0490, 0.0218, -0.1029], + ..., + [-0.0887, -0.0868, 0.0264, ..., -0.1017, -0.0524, -0.1154], + [ 0.0606, 0.0296, -0.0485, ..., -0.0857, -0.0048, -0.0829], + [ 0.0061, 0.0439, -0.0594, ..., 0.0074, -0.0775, -0.0328]], + device='cuda:0'), grad: tensor([[ 2.0247e-06, -4.5943e-04, -5.4693e-04, ..., -2.5249e-04, + 1.5367e-06, 2.1141e-07], + [ 2.5239e-06, 2.9281e-06, 1.7732e-06, ..., -2.7604e-06, + -4.9919e-06, -1.2398e-05], + [ 3.3006e-06, 4.4852e-06, 2.5723e-06, ..., 5.2191e-06, + 3.8818e-06, 3.5726e-06], + ..., + [ 6.8843e-06, 5.9158e-06, 3.6322e-06, ..., 7.4431e-06, + 2.5406e-06, 1.3513e-06], + [ 1.1966e-05, 1.0364e-05, 6.4932e-06, ..., 1.5855e-05, + 4.9770e-06, 3.6955e-06], + [-9.5844e-05, -2.8253e-05, 2.7418e-06, ..., -7.3195e-05, + -3.0790e-06, 8.4238e-07]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0231, -0.0248, -0.0200, -0.0005, -0.0060, 0.0012, -0.0022, -0.0150, + 0.0236, -0.0303], device='cuda:0'), grad: tensor([-7.3433e-04, -9.7379e-06, 1.7077e-05, -2.6271e-05, 1.4675e-04, + 5.8842e-04, 1.3316e-04, 1.4707e-05, 3.6240e-05, -1.6546e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 66---------------------------------------------------- +epoch 66, time 220.32, cls_loss 0.0037 cls_loss_mapping 0.0118 cls_loss_causal 0.6276 re_mapping 0.0101 re_causal 0.0325 /// teacc 99.03 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0104, -0.0171, 0.0749, ..., 0.0007, -0.0402, -0.0304], + [-0.0660, -0.0566, -0.0826, ..., 0.0177, 0.0445, 0.1508], + [-0.0319, -0.0387, -0.0385, ..., -0.0492, 0.0218, -0.1063], + ..., + [-0.0892, -0.0873, 0.0262, ..., -0.1022, -0.0528, -0.1174], + [ 0.0613, 0.0303, -0.0487, ..., -0.0858, -0.0048, -0.0834], + [ 0.0064, 0.0437, -0.0599, ..., 0.0084, -0.0781, -0.0338]], + device='cuda:0'), grad: tensor([[ 5.4128e-06, -1.2375e-05, -1.0155e-05, ..., -1.3947e-05, + 1.0014e-05, 2.5518e-07], + [ 8.9183e-06, 4.8354e-06, 6.7949e-06, ..., 2.4904e-06, + 2.3060e-06, -8.8364e-06], + [ 1.1466e-05, 1.2420e-05, -2.1785e-05, ..., 5.4985e-06, + -6.6042e-05, 1.2573e-06], + ..., + [ 2.1189e-05, 7.4804e-06, 1.2182e-05, ..., 1.8090e-05, + 1.0684e-05, 9.2480e-07], + [-2.6450e-07, -2.8178e-05, 3.1382e-05, ..., 3.0443e-05, + 1.1012e-05, 1.2582e-06], + [-1.3721e-04, -3.4750e-05, -6.0707e-05, ..., -1.1539e-04, + -1.9908e-05, 9.7603e-07]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0235, -0.0246, -0.0200, -0.0007, -0.0061, 0.0012, -0.0020, -0.0148, + 0.0237, -0.0302], device='cuda:0'), grad: tensor([ 4.0606e-06, 3.4988e-05, -1.1039e-04, 6.3002e-05, 1.1218e-04, + 8.4758e-05, 7.4565e-05, -4.0650e-05, 1.0513e-05, -2.3293e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 219.36, cls_loss 0.0053 cls_loss_mapping 0.0145 cls_loss_causal 0.5877 re_mapping 0.0100 re_causal 0.0311 /// teacc 98.97 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0101, -0.0159, 0.0760, ..., 0.0019, -0.0399, -0.0307], + [-0.0664, -0.0571, -0.0839, ..., 0.0176, 0.0444, 0.1521], + [-0.0323, -0.0392, -0.0391, ..., -0.0497, 0.0218, -0.1063], + ..., + [-0.0892, -0.0874, 0.0263, ..., -0.1019, -0.0532, -0.1186], + [ 0.0616, 0.0304, -0.0491, ..., -0.0862, -0.0046, -0.0839], + [ 0.0065, 0.0434, -0.0611, ..., 0.0085, -0.0790, -0.0342]], + device='cuda:0'), grad: tensor([[ 4.2208e-06, 3.3733e-06, 1.2480e-05, ..., 6.9030e-06, + 1.3031e-05, 1.7637e-07], + [ 8.8960e-06, 1.0915e-05, 2.5462e-06, ..., 4.1611e-06, + 8.3596e-06, -6.4336e-06], + [-2.8685e-05, 9.8273e-06, 7.3165e-06, ..., 1.2346e-05, + -5.0128e-05, 8.0094e-07], + ..., + [ 1.2606e-05, 1.1727e-05, 3.4831e-06, ..., 1.0051e-05, + 1.3016e-05, 1.9204e-06], + [ 9.8825e-05, 1.1951e-04, 6.8322e-06, ..., 6.0767e-05, + 1.0437e-04, 6.5519e-07], + [-9.3039e-07, 1.2204e-05, 3.0901e-06, ..., -3.8184e-08, + 1.4335e-05, 1.0431e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0222, -0.0250, -0.0200, 0.0004, -0.0058, 0.0003, -0.0017, -0.0157, + 0.0237, -0.0303], device='cuda:0'), grad: tensor([ 4.3213e-05, 3.3647e-05, -3.5238e-04, -2.2316e-04, 6.0014e-06, + 3.1412e-05, -4.6968e-05, 4.4733e-05, 4.1962e-04, 4.3482e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 219.40, cls_loss 0.0042 cls_loss_mapping 0.0137 cls_loss_causal 0.6517 re_mapping 0.0096 re_causal 0.0313 /// teacc 98.99 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0104, -0.0154, 0.0764, ..., 0.0024, -0.0402, -0.0312], + [-0.0668, -0.0577, -0.0846, ..., 0.0178, 0.0441, 0.1543], + [-0.0327, -0.0396, -0.0396, ..., -0.0503, 0.0224, -0.1069], + ..., + [-0.0897, -0.0877, 0.0257, ..., -0.1028, -0.0538, -0.1214], + [ 0.0618, 0.0311, -0.0488, ..., -0.0865, -0.0042, -0.0845], + [ 0.0076, 0.0440, -0.0616, ..., 0.0091, -0.0795, -0.0350]], + device='cuda:0'), grad: tensor([[ 6.7204e-06, 2.0694e-06, 2.2724e-05, ..., 2.9169e-06, + 1.6689e-05, 1.5935e-06], + [ 4.6566e-06, 1.4538e-06, 8.4862e-06, ..., -6.4746e-06, + 3.2857e-06, -3.5733e-05], + [ 9.3132e-06, 8.3223e-06, 2.3380e-05, ..., 8.6948e-06, + 8.9765e-05, 1.7360e-05], + ..., + [ 6.3777e-06, 2.8498e-06, 6.5625e-05, ..., 5.9828e-06, + 1.9282e-05, 3.1572e-06], + [ 2.0772e-05, 1.9461e-05, 1.7077e-05, ..., 2.8655e-05, + 2.5049e-05, 2.2519e-06], + [ 6.7532e-05, -3.1218e-06, 2.3985e-04, ..., 5.8293e-05, + 3.4976e-04, 1.3644e-06]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0222, -0.0259, -0.0193, -0.0003, -0.0053, 0.0004, -0.0025, -0.0155, + 0.0242, -0.0301], device='cuda:0'), grad: tensor([ 1.0842e-04, 4.6730e-05, 4.6682e-04, 1.5819e-04, -1.4696e-03, + -7.5519e-05, 6.7830e-05, -4.0984e-04, 1.6093e-04, 9.4604e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 219.42, cls_loss 0.0034 cls_loss_mapping 0.0113 cls_loss_causal 0.6086 re_mapping 0.0092 re_causal 0.0301 /// teacc 98.87 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0105, -0.0151, 0.0763, ..., 0.0023, -0.0408, -0.0316], + [-0.0677, -0.0597, -0.0862, ..., 0.0166, 0.0442, 0.1545], + [-0.0332, -0.0405, -0.0402, ..., -0.0510, 0.0222, -0.1075], + ..., + [-0.0901, -0.0885, 0.0254, ..., -0.1034, -0.0541, -0.1238], + [ 0.0622, 0.0310, -0.0497, ..., -0.0872, -0.0039, -0.0871], + [ 0.0080, 0.0445, -0.0617, ..., 0.0108, -0.0799, -0.0330]], + device='cuda:0'), grad: tensor([[ 1.0114e-06, 7.0706e-06, 1.0625e-05, ..., -3.4925e-08, + 1.2666e-05, 1.2584e-07], + [ 6.3842e-07, 1.0192e-05, 1.7598e-05, ..., -1.6447e-06, + 1.9312e-05, -4.2245e-06], + [ 1.0595e-05, 2.3186e-05, 3.2037e-05, ..., 7.8185e-07, + 3.9101e-05, 7.1293e-07], + ..., + [ 2.3656e-06, 5.3570e-06, -4.4465e-05, ..., 1.7500e-06, + -1.2137e-05, 7.0734e-07], + [-1.6108e-05, -1.7107e-04, -2.8110e-04, ..., 1.0952e-06, + -3.4118e-04, 4.9779e-07], + [-2.4829e-06, 4.9174e-06, 4.1157e-05, ..., -2.5742e-06, + 3.9488e-05, 8.4052e-07]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0224, -0.0266, -0.0196, 0.0002, -0.0054, 0.0004, -0.0019, -0.0153, + 0.0239, -0.0298], device='cuda:0'), grad: tensor([ 4.8667e-05, 7.1883e-05, 1.7166e-04, 3.8296e-05, 2.2799e-05, + 7.1001e-04, 1.5748e-04, -1.1432e-04, -1.2217e-03, 1.1402e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 219.51, cls_loss 0.0031 cls_loss_mapping 0.0116 cls_loss_causal 0.6094 re_mapping 0.0097 re_causal 0.0317 /// teacc 99.00 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0105, -0.0150, 0.0765, ..., 0.0023, -0.0410, -0.0318], + [-0.0679, -0.0600, -0.0865, ..., 0.0168, 0.0444, 0.1564], + [-0.0337, -0.0412, -0.0408, ..., -0.0514, 0.0222, -0.1084], + ..., + [-0.0905, -0.0893, 0.0255, ..., -0.1039, -0.0546, -0.1262], + [ 0.0625, 0.0310, -0.0499, ..., -0.0876, -0.0034, -0.0873], + [ 0.0085, 0.0450, -0.0623, ..., 0.0113, -0.0808, -0.0334]], + device='cuda:0'), grad: tensor([[ 9.1419e-06, 4.1313e-06, 7.3947e-07, ..., 7.2792e-06, + 1.5222e-05, 1.8720e-06], + [ 2.0154e-06, 1.5097e-06, -4.1202e-06, ..., -1.6019e-05, + -1.1280e-05, -2.8640e-05], + [ 1.0604e-04, 6.2585e-05, 9.6709e-06, ..., 7.0274e-05, + 1.6654e-04, 6.7875e-06], + ..., + [ 8.7246e-06, 5.5954e-06, 3.4738e-06, ..., 7.4059e-06, + -6.9067e-06, 3.3323e-06], + [-6.5193e-06, -5.1260e-06, 7.1377e-06, ..., 1.5080e-05, + 1.1139e-05, 4.2468e-06], + [ 8.8895e-07, 2.9728e-06, 1.4804e-05, ..., -1.1805e-07, + 2.5526e-05, 3.3528e-06]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0226, -0.0267, -0.0193, -0.0003, -0.0049, 0.0004, -0.0016, -0.0153, + 0.0240, -0.0302], device='cuda:0'), grad: tensor([ 3.2127e-05, -2.2262e-05, 4.0698e-04, -2.8205e-04, -2.0385e-05, + 3.9816e-05, -2.9147e-05, -2.3210e-04, 3.4839e-05, 7.2539e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 219.22, cls_loss 0.0038 cls_loss_mapping 0.0124 cls_loss_causal 0.6211 re_mapping 0.0096 re_causal 0.0305 /// teacc 98.97 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0109, -0.0152, 0.0765, ..., 0.0021, -0.0414, -0.0322], + [-0.0686, -0.0605, -0.0863, ..., 0.0174, 0.0446, 0.1595], + [-0.0342, -0.0417, -0.0413, ..., -0.0519, 0.0219, -0.1096], + ..., + [-0.0909, -0.0896, 0.0262, ..., -0.1044, -0.0540, -0.1312], + [ 0.0628, 0.0311, -0.0503, ..., -0.0881, -0.0035, -0.0881], + [ 0.0102, 0.0459, -0.0625, ..., 0.0123, -0.0806, -0.0340]], + device='cuda:0'), grad: tensor([[ 2.7269e-06, -1.3839e-06, -8.0243e-06, ..., -2.5257e-06, + 5.7220e-06, 7.0734e-07], + [ 1.9409e-06, 3.6135e-06, 5.6345e-07, ..., -1.6168e-06, + -1.0395e-04, -1.4198e-04], + [ 2.4512e-05, 1.3985e-05, 4.9509e-06, ..., 4.0084e-06, + 1.4961e-05, 1.1474e-05], + ..., + [ 8.1137e-06, 7.2718e-06, 3.2037e-06, ..., 8.8885e-06, + 2.2292e-05, 1.7509e-05], + [-3.7819e-05, -6.5714e-06, 7.8902e-06, ..., 5.1148e-06, + -3.9876e-05, 1.5106e-06], + [-2.7522e-05, -6.2212e-06, 7.3761e-06, ..., -3.1292e-05, + 1.0341e-05, 2.9299e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0231, -0.0266, -0.0195, -0.0007, -0.0060, 0.0006, -0.0013, -0.0145, + 0.0237, -0.0296], device='cuda:0'), grad: tensor([ 9.7975e-07, -1.8156e-04, 6.4850e-05, 5.0664e-05, 1.8597e-04, + -9.4175e-05, 5.2661e-05, 5.9187e-05, -1.1480e-04, -2.3872e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 72---------------------------------------------------- +epoch 72, time 220.25, cls_loss 0.0035 cls_loss_mapping 0.0106 cls_loss_causal 0.6097 re_mapping 0.0098 re_causal 0.0305 /// teacc 99.06 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0109, -0.0150, 0.0765, ..., 0.0021, -0.0419, -0.0327], + [-0.0692, -0.0609, -0.0869, ..., 0.0175, 0.0444, 0.1613], + [-0.0349, -0.0423, -0.0419, ..., -0.0524, 0.0217, -0.1105], + ..., + [-0.0918, -0.0901, 0.0264, ..., -0.1055, -0.0542, -0.1336], + [ 0.0631, 0.0309, -0.0508, ..., -0.0886, -0.0031, -0.0881], + [ 0.0102, 0.0453, -0.0637, ..., 0.0125, -0.0818, -0.0350]], + device='cuda:0'), grad: tensor([[ 6.3330e-06, 1.2204e-05, 2.5006e-07, ..., 3.7104e-06, + 4.3064e-06, 3.4086e-06], + [ 3.3826e-06, 4.7497e-06, 7.2457e-07, ..., -3.0637e-05, + -2.5973e-05, -4.7207e-05], + [ 2.8014e-05, 2.4900e-05, 7.1246e-07, ..., 8.3819e-06, + 1.6347e-05, 9.4324e-06], + ..., + [ 2.7418e-05, 2.7463e-05, 5.3877e-07, ..., 1.7479e-05, + 8.0466e-06, 5.9009e-06], + [-8.8736e-06, 3.8981e-05, 3.0305e-06, ..., 1.4231e-05, + -6.7241e-06, 1.1310e-05], + [-2.3823e-06, 9.7036e-05, 3.3379e-06, ..., -3.0488e-05, + 1.0133e-05, 5.8524e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0236, -0.0272, -0.0197, -0.0006, -0.0048, 0.0011, -0.0010, -0.0144, + 0.0232, -0.0301], device='cuda:0'), grad: tensor([ 3.5197e-05, -5.5641e-05, 8.7798e-05, 8.2350e-04, 5.1647e-05, + -1.2798e-03, 8.8394e-05, 2.0340e-05, 4.5568e-05, 1.8334e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 219.01, cls_loss 0.0029 cls_loss_mapping 0.0098 cls_loss_causal 0.6053 re_mapping 0.0094 re_causal 0.0294 /// teacc 98.91 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0110, -0.0149, 0.0767, ..., 0.0022, -0.0423, -0.0330], + [-0.0699, -0.0616, -0.0875, ..., 0.0181, 0.0448, 0.1629], + [-0.0355, -0.0429, -0.0427, ..., -0.0532, 0.0212, -0.1113], + ..., + [-0.0923, -0.0906, 0.0263, ..., -0.1066, -0.0549, -0.1354], + [ 0.0636, 0.0308, -0.0511, ..., -0.0888, -0.0027, -0.0887], + [ 0.0105, 0.0454, -0.0641, ..., 0.0129, -0.0823, -0.0353]], + device='cuda:0'), grad: tensor([[ 2.5406e-05, -7.6145e-06, -4.4256e-05, ..., -6.0141e-05, + 5.3234e-06, 2.8498e-07], + [ 2.0768e-06, 7.3984e-06, 2.1949e-05, ..., 1.5676e-05, + -1.2107e-05, -1.6287e-05], + [ 1.2144e-05, 8.3521e-06, 1.2346e-05, ..., 1.1794e-05, + 1.1355e-05, 3.4161e-06], + ..., + [ 1.0133e-06, 1.5050e-06, 1.0289e-05, ..., 5.1446e-06, + 1.5721e-05, 4.9546e-06], + [-9.2685e-05, -5.3495e-05, -4.4405e-06, ..., -1.4365e-05, + -1.7613e-05, 1.9046e-06], + [ 2.8968e-05, 2.0772e-05, 1.0276e-04, ..., 1.8418e-05, + 5.8830e-05, 1.8487e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0237, -0.0269, -0.0202, -0.0005, -0.0048, 0.0011, -0.0008, -0.0145, + 0.0238, -0.0304], device='cuda:0'), grad: tensor([-1.0139e-04, 2.3499e-05, 4.3184e-05, 2.2545e-05, -3.1376e-04, + 6.5923e-05, 8.1480e-05, 4.4376e-05, -1.2815e-04, 2.6250e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 218.28, cls_loss 0.0034 cls_loss_mapping 0.0107 cls_loss_causal 0.6381 re_mapping 0.0092 re_causal 0.0294 /// teacc 99.02 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0114, -0.0153, 0.0771, ..., 0.0023, -0.0424, -0.0333], + [-0.0704, -0.0624, -0.0881, ..., 0.0186, 0.0451, 0.1651], + [-0.0360, -0.0436, -0.0433, ..., -0.0539, 0.0210, -0.1121], + ..., + [-0.0927, -0.0912, 0.0260, ..., -0.1078, -0.0549, -0.1385], + [ 0.0641, 0.0310, -0.0514, ..., -0.0893, -0.0028, -0.0893], + [ 0.0108, 0.0457, -0.0646, ..., 0.0137, -0.0829, -0.0357]], + device='cuda:0'), grad: tensor([[ 3.0138e-06, -1.0217e-06, -5.0738e-06, ..., -3.2261e-06, + 1.7788e-06, 2.2743e-06], + [ 4.1015e-06, 3.8259e-06, -2.1383e-06, ..., -4.0144e-05, + -1.9789e-05, -4.4823e-05], + [ 3.9563e-06, 4.3213e-06, 1.1474e-06, ..., 5.5954e-06, + 3.1777e-06, 4.9993e-06], + ..., + [ 3.1199e-06, 2.9784e-06, 7.5204e-07, ..., 1.2085e-05, + 6.3665e-06, 1.1615e-05], + [-9.2149e-05, -8.1718e-05, 1.4501e-06, ..., -6.3702e-07, + -1.3344e-05, 6.5975e-06], + [ 5.0753e-05, 5.4985e-05, 2.9691e-06, ..., 1.3277e-05, + 1.3240e-05, 5.9754e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0237, -0.0268, -0.0208, -0.0007, -0.0047, 0.0009, -0.0006, -0.0139, + 0.0236, -0.0304], device='cuda:0'), grad: tensor([ 3.3081e-06, -6.9141e-05, 2.3142e-05, 3.4243e-05, 2.1413e-05, + 3.0220e-05, 1.8239e-05, -1.4268e-06, -2.1636e-04, 1.5616e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 218.41, cls_loss 0.0025 cls_loss_mapping 0.0080 cls_loss_causal 0.6022 re_mapping 0.0091 re_causal 0.0289 /// teacc 98.97 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0118, -0.0155, 0.0773, ..., 0.0024, -0.0428, -0.0338], + [-0.0707, -0.0627, -0.0885, ..., 0.0190, 0.0450, 0.1666], + [-0.0364, -0.0441, -0.0438, ..., -0.0542, 0.0212, -0.1118], + ..., + [-0.0932, -0.0917, 0.0260, ..., -0.1085, -0.0553, -0.1411], + [ 0.0641, 0.0308, -0.0515, ..., -0.0899, -0.0028, -0.0899], + [ 0.0107, 0.0452, -0.0650, ..., 0.0137, -0.0836, -0.0362]], + device='cuda:0'), grad: tensor([[ 2.2650e-06, -3.6269e-05, -4.7714e-05, ..., -7.8678e-06, + 6.1747e-07, -4.2841e-06], + [ 1.0878e-06, 1.2750e-06, 8.2189e-07, ..., -4.2245e-06, + -1.1511e-06, -9.9689e-06], + [ 3.8631e-06, 1.5154e-05, 1.6958e-05, ..., 5.6885e-06, + 2.0508e-06, 2.7753e-06], + ..., + [ 3.5968e-06, 4.0121e-06, 1.7760e-06, ..., 4.4666e-06, + -5.3458e-07, 2.6152e-06], + [-3.5428e-06, 1.1265e-05, 2.1532e-05, ..., 9.7379e-06, + -2.3395e-06, 3.1553e-06], + [-4.0263e-05, -2.6658e-05, 8.7023e-06, ..., -3.1203e-05, + 4.8764e-06, 2.3097e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0239, -0.0266, -0.0203, 0.0004, -0.0046, 0.0005, -0.0006, -0.0145, + 0.0234, -0.0308], device='cuda:0'), grad: tensor([-9.0599e-05, 1.8761e-05, 8.8036e-05, 2.7061e-05, 5.6356e-05, + 3.2008e-05, -9.9167e-06, -1.3113e-04, 4.4823e-05, -3.5226e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 218.50, cls_loss 0.0030 cls_loss_mapping 0.0096 cls_loss_causal 0.6168 re_mapping 0.0091 re_causal 0.0283 /// teacc 98.90 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0120, -0.0154, 0.0774, ..., 0.0024, -0.0432, -0.0341], + [-0.0713, -0.0630, -0.0890, ..., 0.0193, 0.0453, 0.1683], + [-0.0369, -0.0446, -0.0445, ..., -0.0548, 0.0210, -0.1124], + ..., + [-0.0939, -0.0925, 0.0263, ..., -0.1096, -0.0555, -0.1430], + [ 0.0643, 0.0308, -0.0518, ..., -0.0906, -0.0029, -0.0905], + [ 0.0114, 0.0453, -0.0655, ..., 0.0146, -0.0844, -0.0367]], + device='cuda:0'), grad: tensor([[ 1.8608e-06, -4.4852e-06, -3.4999e-06, ..., -5.5023e-06, + 5.1446e-06, 5.9139e-07], + [ 6.9514e-06, 3.0976e-06, 3.4988e-05, ..., 7.5735e-06, + 4.3511e-05, -2.3380e-05], + [ 2.3562e-06, 2.6524e-06, 1.1742e-05, ..., 1.1131e-05, + 1.3530e-05, 4.4443e-06], + ..., + [ 4.9807e-06, 2.6524e-06, 1.0375e-06, ..., 8.8438e-06, + 2.0415e-05, 5.6587e-06], + [-3.1311e-06, 1.6196e-06, 2.3872e-05, ..., 3.2842e-05, + 2.9668e-05, 3.1814e-06], + [-1.3745e-04, -6.6936e-05, 1.5885e-05, ..., -1.5771e-04, + 9.0078e-06, 3.6135e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0244, -0.0265, -0.0204, 0.0004, -0.0047, 0.0006, -0.0002, -0.0146, + 0.0232, -0.0304], device='cuda:0'), grad: tensor([ 1.2890e-05, 1.4067e-04, 6.0350e-05, 2.4974e-05, 4.3869e-05, + 1.2141e-04, -2.6774e-04, -1.7428e-04, 1.3769e-04, -9.9897e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 218.10, cls_loss 0.0034 cls_loss_mapping 0.0096 cls_loss_causal 0.5961 re_mapping 0.0089 re_causal 0.0276 /// teacc 98.96 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0117, -0.0155, 0.0783, ..., 0.0029, -0.0428, -0.0341], + [-0.0717, -0.0635, -0.0898, ..., 0.0194, 0.0453, 0.1694], + [-0.0376, -0.0456, -0.0453, ..., -0.0554, 0.0207, -0.1127], + ..., + [-0.0945, -0.0933, 0.0271, ..., -0.1104, -0.0556, -0.1452], + [ 0.0649, 0.0310, -0.0524, ..., -0.0911, -0.0025, -0.0907], + [ 0.0117, 0.0452, -0.0662, ..., 0.0149, -0.0850, -0.0370]], + device='cuda:0'), grad: tensor([[ 5.8830e-05, 5.8442e-05, -2.6003e-06, ..., 2.8208e-05, + 2.3693e-05, 5.5181e-08], + [ 4.9382e-05, 5.1737e-05, 9.7379e-06, ..., 7.8797e-05, + 1.4501e-06, -2.0675e-06], + [ 1.0757e-06, 1.4212e-06, 1.0338e-06, ..., 1.6820e-06, + -9.6709e-06, 1.0850e-07], + ..., + [ 9.5963e-06, 6.5006e-06, 2.6692e-06, ..., 8.9854e-06, + 9.6858e-06, 5.3504e-07], + [-8.0168e-05, -7.7844e-05, 8.1807e-06, ..., -2.5466e-05, + -3.2485e-05, 5.1502e-07], + [-7.8082e-05, -7.6294e-05, -1.4663e-05, ..., -1.3113e-04, + 4.2729e-06, 2.8964e-07]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0236, -0.0268, -0.0211, 0.0004, -0.0058, 0.0014, -0.0001, -0.0142, + 0.0233, -0.0305], device='cuda:0'), grad: tensor([ 1.0777e-04, 2.1660e-04, -2.8744e-05, 2.2739e-05, 7.2062e-05, + 2.1949e-05, -8.1882e-06, 3.7819e-05, -1.1396e-04, -3.2830e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 218.42, cls_loss 0.0027 cls_loss_mapping 0.0099 cls_loss_causal 0.6210 re_mapping 0.0093 re_causal 0.0287 /// teacc 99.01 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0124, -0.0169, 0.0783, ..., 0.0025, -0.0431, -0.0344], + [-0.0720, -0.0639, -0.0904, ..., 0.0198, 0.0453, 0.1714], + [-0.0375, -0.0451, -0.0456, ..., -0.0556, 0.0206, -0.1132], + ..., + [-0.0947, -0.0936, 0.0266, ..., -0.1112, -0.0557, -0.1482], + [ 0.0657, 0.0313, -0.0526, ..., -0.0910, -0.0023, -0.0915], + [ 0.0121, 0.0452, -0.0665, ..., 0.0153, -0.0855, -0.0375]], + device='cuda:0'), grad: tensor([[ 6.0163e-06, 3.5409e-06, -6.2101e-06, ..., 2.3022e-06, + 1.5311e-06, 1.6764e-08], + [ 2.2873e-06, 2.5574e-06, 9.8627e-07, ..., 1.2927e-06, + 5.8766e-07, -1.3234e-06], + [ 1.4134e-05, 8.4862e-06, 3.0622e-06, ..., 3.6620e-06, + 5.9828e-06, 9.4529e-08], + ..., + [ 1.9550e-05, 1.3269e-05, 1.7677e-06, ..., 8.8960e-06, + 6.7614e-06, 2.1141e-07], + [-3.1680e-05, -1.2536e-06, 2.5947e-06, ..., 8.9854e-06, + -1.1697e-05, 1.9558e-07], + [-7.2122e-06, -1.7546e-06, 2.3525e-06, ..., -1.5959e-05, + 4.7497e-06, 2.5891e-07]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0246, -0.0272, -0.0210, 0.0003, -0.0057, 0.0014, 0.0003, -0.0138, + 0.0237, -0.0309], device='cuda:0'), grad: tensor([ 1.1787e-05, 6.9067e-06, -5.6103e-06, 2.1100e-05, 2.1592e-05, + -8.5354e-05, -4.1276e-06, 4.1574e-05, 8.0839e-06, -1.6049e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 218.22, cls_loss 0.0027 cls_loss_mapping 0.0098 cls_loss_causal 0.6117 re_mapping 0.0091 re_causal 0.0280 /// teacc 99.02 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0125, -0.0167, 0.0788, ..., 0.0028, -0.0433, -0.0354], + [-0.0723, -0.0649, -0.0905, ..., 0.0205, 0.0445, 0.1736], + [-0.0378, -0.0448, -0.0460, ..., -0.0558, 0.0218, -0.1127], + ..., + [-0.0956, -0.0941, 0.0263, ..., -0.1122, -0.0560, -0.1520], + [ 0.0658, 0.0311, -0.0530, ..., -0.0916, -0.0023, -0.0924], + [ 0.0125, 0.0452, -0.0672, ..., 0.0156, -0.0863, -0.0382]], + device='cuda:0'), grad: tensor([[ 9.3039e-07, -1.7872e-06, -1.0114e-06, ..., -1.0775e-06, + 1.5907e-06, 1.2759e-07], + [ 6.9803e-07, 1.5609e-06, 1.8477e-06, ..., -4.2398e-07, + 2.6124e-07, -3.6694e-06], + [ 5.0217e-06, 8.0913e-06, 1.3866e-05, ..., 1.1340e-05, + 1.2383e-05, 4.8336e-07], + ..., + [ 3.8482e-06, 5.1260e-06, 1.4286e-06, ..., 2.7940e-06, + 2.0303e-06, 8.5309e-07], + [-2.4363e-06, 8.1062e-06, 1.4812e-05, ..., 1.0066e-05, + 7.2047e-06, 7.9442e-07], + [-2.7232e-06, 2.1979e-06, 2.4382e-06, ..., -2.1663e-06, + 1.8450e-06, 3.0361e-07]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0244, -0.0282, -0.0196, 0.0002, -0.0050, 0.0017, -0.0007, -0.0137, + 0.0232, -0.0310], device='cuda:0'), grad: tensor([ 6.7130e-06, 4.9025e-06, 5.0813e-05, -1.8728e-04, 1.5423e-05, + 9.2566e-05, -5.9277e-05, 1.2666e-05, 5.6654e-05, 6.5193e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 218.50, cls_loss 0.0033 cls_loss_mapping 0.0120 cls_loss_causal 0.6396 re_mapping 0.0087 re_causal 0.0274 /// teacc 99.03 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0127, -0.0166, 0.0793, ..., 0.0031, -0.0436, -0.0357], + [-0.0724, -0.0653, -0.0913, ..., 0.0213, 0.0450, 0.1759], + [-0.0383, -0.0446, -0.0463, ..., -0.0563, 0.0217, -0.1133], + ..., + [-0.0957, -0.0949, 0.0259, ..., -0.1134, -0.0567, -0.1558], + [ 0.0657, 0.0305, -0.0536, ..., -0.0924, -0.0024, -0.0929], + [ 0.0128, 0.0452, -0.0675, ..., 0.0160, -0.0870, -0.0387]], + device='cuda:0'), grad: tensor([[ 7.5903e-07, -3.7695e-07, 4.1910e-06, ..., 4.2506e-06, + 7.7114e-07, 2.0396e-07], + [ 6.1048e-07, 7.2923e-07, 8.1491e-07, ..., -4.8876e-06, + -6.1430e-06, -1.3024e-05], + [ 2.5947e-06, 2.5015e-06, 1.7742e-06, ..., 3.9339e-06, + -3.5353e-06, 3.0417e-06], + ..., + [ 3.3855e-05, 2.4930e-05, 2.7437e-06, ..., 3.2276e-05, + 1.2591e-05, 5.6103e-06], + [ 5.3830e-06, 1.1377e-05, 2.5071e-06, ..., 7.2382e-06, + 1.5479e-06, 1.0999e-06], + [-5.9128e-05, -3.3289e-05, 2.7586e-06, ..., -5.0485e-05, + 4.6603e-06, 9.8255e-07]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0243, -0.0278, -0.0194, 0.0002, -0.0047, 0.0024, -0.0011, -0.0136, + 0.0225, -0.0316], device='cuda:0'), grad: tensor([ 1.0520e-05, -1.5408e-05, -2.0549e-05, 6.1154e-05, 1.4678e-05, + -7.2122e-05, -1.3322e-05, 1.2827e-04, 2.7135e-05, -1.2016e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 218.53, cls_loss 0.0027 cls_loss_mapping 0.0103 cls_loss_causal 0.5931 re_mapping 0.0093 re_causal 0.0288 /// teacc 98.95 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0130, -0.0166, 0.0795, ..., 0.0031, -0.0441, -0.0361], + [-0.0727, -0.0656, -0.0918, ..., 0.0218, 0.0455, 0.1787], + [-0.0389, -0.0452, -0.0467, ..., -0.0569, 0.0216, -0.1145], + ..., + [-0.0961, -0.0956, 0.0259, ..., -0.1141, -0.0572, -0.1600], + [ 0.0660, 0.0305, -0.0539, ..., -0.0930, -0.0023, -0.0935], + [ 0.0136, 0.0453, -0.0677, ..., 0.0170, -0.0873, -0.0392]], + device='cuda:0'), grad: tensor([[ 9.9838e-06, 1.5959e-05, -1.0403e-06, ..., 1.9431e-05, + 1.6177e-06, 2.2119e-08], + [ 6.4373e-05, 1.3673e-04, 4.2543e-06, ..., 1.7786e-04, + 4.0680e-06, -7.1246e-07], + [ 8.7544e-06, 1.1854e-05, 1.3066e-06, ..., 1.1258e-05, + 1.8971e-06, 1.1385e-07], + ..., + [ 1.0617e-05, 1.3649e-05, 1.5572e-06, ..., 1.5140e-05, + 3.0082e-06, 1.6834e-07], + [ 1.7107e-05, 1.9044e-05, 6.1058e-06, ..., 2.1011e-05, + 5.1856e-06, 1.3644e-07], + [-4.5276e-04, -4.8876e-04, -8.7082e-05, ..., -8.0538e-04, + -8.1778e-05, 5.7276e-08]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0247, -0.0274, -0.0192, 0.0003, -0.0048, 0.0019, -0.0004, -0.0138, + 0.0222, -0.0315], device='cuda:0'), grad: tensor([ 5.5552e-05, 4.5061e-04, 3.4928e-05, 6.8605e-05, 1.1301e-03, + 1.5819e-04, 4.1008e-05, 1.6257e-05, 6.3658e-05, -2.0199e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 218.57, cls_loss 0.0045 cls_loss_mapping 0.0131 cls_loss_causal 0.6193 re_mapping 0.0088 re_causal 0.0277 /// teacc 98.89 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0130, -0.0164, 0.0797, ..., 0.0030, -0.0443, -0.0368], + [-0.0730, -0.0660, -0.0928, ..., 0.0214, 0.0458, 0.1810], + [-0.0396, -0.0459, -0.0473, ..., -0.0579, 0.0210, -0.1151], + ..., + [-0.0975, -0.0963, 0.0255, ..., -0.1171, -0.0580, -0.1634], + [ 0.0664, 0.0307, -0.0543, ..., -0.0937, -0.0022, -0.0946], + [ 0.0155, 0.0455, -0.0677, ..., 0.0185, -0.0872, -0.0399]], + device='cuda:0'), grad: tensor([[ 7.7635e-06, 3.2425e-05, -7.0706e-06, ..., 1.2629e-05, + 8.8662e-06, 5.2620e-08], + [ 6.8210e-06, 1.3091e-05, 1.2824e-06, ..., 7.0557e-06, + -1.8489e-04, -5.7230e-07], + [ 2.1398e-05, 3.4034e-05, 1.6605e-06, ..., 1.5751e-05, + 1.7190e-04, 4.9593e-08], + ..., + [ 4.8541e-06, 3.0309e-05, 6.3963e-06, ..., 1.9088e-05, + 3.0875e-05, 8.0559e-08], + [ 1.0836e-04, 1.6725e-04, 9.2387e-07, ..., 9.2804e-05, + 5.0396e-05, 8.1956e-08], + [ 1.9819e-05, 2.9758e-05, 9.0182e-05, ..., 8.1420e-05, + 7.6234e-05, 6.6124e-08]], device='cuda:0') +Epoch 84, bias, value: tensor([-2.4774e-02, -2.7211e-02, -1.9935e-02, 4.0391e-05, -4.0641e-03, + 2.2523e-03, -2.0202e-04, -1.5410e-02, 2.2319e-02, -3.0677e-02], + device='cuda:0'), grad: tensor([ 1.6427e-04, -4.7755e-04, 2.6188e-03, -1.2674e-03, -2.4581e-04, + 5.5170e-04, 6.9261e-05, -2.0771e-03, 3.5453e-04, 3.1042e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 218.57, cls_loss 0.0039 cls_loss_mapping 0.0114 cls_loss_causal 0.6168 re_mapping 0.0084 re_causal 0.0271 /// teacc 99.03 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0131, -0.0167, 0.0796, ..., 0.0017, -0.0445, -0.0375], + [-0.0742, -0.0672, -0.0932, ..., 0.0219, 0.0461, 0.1833], + [-0.0402, -0.0466, -0.0477, ..., -0.0587, 0.0206, -0.1159], + ..., + [-0.0973, -0.0966, 0.0259, ..., -0.1172, -0.0583, -0.1660], + [ 0.0668, 0.0302, -0.0545, ..., -0.0935, -0.0016, -0.0949], + [ 0.0156, 0.0460, -0.0677, ..., 0.0198, -0.0882, -0.0408]], + device='cuda:0'), grad: tensor([[ 2.6636e-07, -6.9499e-05, -4.2021e-05, ..., -8.6844e-05, + 1.0505e-06, 3.6159e-07], + [ 3.0990e-07, 9.5926e-07, 8.1584e-07, ..., -1.0684e-05, + -7.9945e-06, -1.6838e-05], + [ 8.7451e-07, 2.6479e-05, 1.5162e-05, ..., 3.3230e-05, + -4.8894e-07, 6.0350e-07], + ..., + [ 1.3169e-06, 3.4384e-06, 1.9055e-06, ..., 4.3400e-06, + 1.9185e-06, 1.4035e-06], + [ 3.0827e-06, 4.6864e-06, 1.7937e-06, ..., 4.7833e-06, + 3.4925e-06, 1.2293e-06], + [-8.1677e-07, 2.4721e-05, 1.5318e-05, ..., 3.1531e-05, + 2.9020e-06, 2.4326e-06]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0260, -0.0272, -0.0207, 0.0003, -0.0035, 0.0019, -0.0002, -0.0141, + 0.0224, -0.0314], device='cuda:0'), grad: tensor([-1.5426e-04, -1.9118e-05, 5.2601e-05, 1.9029e-05, 3.5129e-06, + 1.0230e-05, 1.6212e-05, -1.5855e-05, 1.4246e-05, 7.3373e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 218.59, cls_loss 0.0025 cls_loss_mapping 0.0077 cls_loss_causal 0.5625 re_mapping 0.0088 re_causal 0.0259 /// teacc 99.01 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0131, -0.0164, 0.0798, ..., 0.0017, -0.0446, -0.0386], + [-0.0745, -0.0675, -0.0934, ..., 0.0233, 0.0463, 0.1856], + [-0.0408, -0.0471, -0.0482, ..., -0.0593, 0.0206, -0.1164], + ..., + [-0.0976, -0.0969, 0.0260, ..., -0.1183, -0.0588, -0.1676], + [ 0.0671, 0.0303, -0.0550, ..., -0.0943, -0.0013, -0.0957], + [ 0.0161, 0.0460, -0.0681, ..., 0.0203, -0.0889, -0.0417]], + device='cuda:0'), grad: tensor([[ 4.2543e-06, 4.7162e-06, -3.4049e-06, ..., 1.9427e-06, + 8.1584e-07, 1.0198e-07], + [ 1.3784e-06, -6.7167e-06, 3.9418e-07, ..., -2.3097e-06, + -1.4767e-05, -1.2279e-05], + [ 1.8757e-06, 5.4464e-06, 6.3283e-07, ..., 2.7549e-06, + 6.9626e-06, 5.3346e-06], + ..., + [ 3.5539e-06, 4.5635e-06, 1.2768e-06, ..., 5.1409e-06, + 2.8424e-06, 2.0806e-06], + [ 8.0913e-06, 9.3356e-06, 5.3458e-07, ..., 8.1584e-06, + 1.2759e-06, 3.4086e-07], + [-7.9572e-05, -6.7830e-05, 3.5334e-06, ..., -6.7770e-05, + -5.1260e-06, 2.1998e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0260, -0.0266, -0.0208, -0.0002, -0.0035, 0.0018, -0.0003, -0.0140, + 0.0225, -0.0315], device='cuda:0'), grad: tensor([ 4.0606e-06, -5.5730e-05, 3.0786e-05, 3.2485e-05, 8.5652e-05, + -8.9407e-04, 9.1124e-04, 1.9729e-05, 2.1994e-05, -1.5664e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 218.46, cls_loss 0.0033 cls_loss_mapping 0.0095 cls_loss_causal 0.6054 re_mapping 0.0077 re_causal 0.0253 /// teacc 98.97 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0134, -0.0162, 0.0798, ..., 0.0018, -0.0450, -0.0390], + [-0.0751, -0.0678, -0.0939, ..., 0.0242, 0.0469, 0.1875], + [-0.0416, -0.0475, -0.0486, ..., -0.0599, 0.0200, -0.1169], + ..., + [-0.0978, -0.0973, 0.0259, ..., -0.1190, -0.0587, -0.1692], + [ 0.0674, 0.0303, -0.0555, ..., -0.0949, -0.0009, -0.0963], + [ 0.0168, 0.0465, -0.0684, ..., 0.0207, -0.0896, -0.0424]], + device='cuda:0'), grad: tensor([[ 2.6897e-06, -5.4270e-05, -7.4565e-05, ..., -8.3089e-05, + 2.2389e-06, 1.4273e-07], + [ 5.2489e-06, 7.8082e-06, 1.9222e-06, ..., -1.2051e-06, + 2.7437e-06, -7.4618e-06], + [ 3.8631e-06, 1.7211e-05, 1.9893e-05, ..., 2.1383e-05, + 2.0079e-06, 6.7614e-07], + ..., + [ 5.9493e-06, 5.2527e-06, 1.1094e-05, ..., 5.4091e-06, + 1.2763e-05, 2.5164e-06], + [ 8.8215e-06, 3.6657e-05, 3.3230e-06, ..., 2.9147e-05, + 2.7925e-05, 1.1334e-06], + [ 3.6368e-07, 3.7968e-05, 5.1439e-05, ..., 4.5210e-05, + 1.4700e-05, 1.2610e-06]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0264, -0.0264, -0.0216, -0.0005, -0.0038, 0.0019, 0.0004, -0.0132, + 0.0226, -0.0317], device='cuda:0'), grad: tensor([-1.6677e-04, 1.4558e-05, 7.8499e-05, -7.8821e-04, -6.9916e-05, + 7.5054e-04, 1.8269e-05, -1.6615e-05, 3.7581e-05, 1.4210e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 218.82, cls_loss 0.0019 cls_loss_mapping 0.0068 cls_loss_causal 0.5700 re_mapping 0.0080 re_causal 0.0257 /// teacc 98.95 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0136, -0.0159, 0.0802, ..., 0.0021, -0.0452, -0.0402], + [-0.0752, -0.0681, -0.0940, ..., 0.0249, 0.0470, 0.1898], + [-0.0424, -0.0484, -0.0491, ..., -0.0605, 0.0201, -0.1174], + ..., + [-0.0982, -0.0979, 0.0259, ..., -0.1192, -0.0589, -0.1712], + [ 0.0671, 0.0299, -0.0560, ..., -0.0960, -0.0011, -0.0978], + [ 0.0176, 0.0470, -0.0680, ..., 0.0211, -0.0900, -0.0433]], + device='cuda:0'), grad: tensor([[ 5.0059e-07, 4.3819e-07, 6.2585e-06, ..., 7.0184e-06, + 1.3504e-06, 1.0710e-08], + [ 2.8615e-07, 6.6310e-07, 1.2536e-06, ..., 1.1809e-06, + 2.3711e-06, -3.5460e-07], + [ 5.9418e-07, 1.3364e-06, 4.8764e-06, ..., 4.0978e-06, + -1.7032e-05, 7.3342e-08], + ..., + [ 1.0002e-06, 1.3504e-06, 8.7172e-07, ..., 1.3988e-06, + 1.4029e-05, 6.2399e-08], + [ 1.0058e-06, 1.7844e-06, 3.9227e-06, ..., 4.9435e-06, + 1.2126e-06, 5.6345e-08], + [-1.5376e-06, 6.3702e-06, 2.9150e-06, ..., 2.4959e-06, + 1.3765e-06, 5.5879e-08]], device='cuda:0') +Epoch 88, bias, value: tensor([-2.6400e-02, -2.6197e-02, -2.1541e-02, 2.2103e-05, -4.1389e-03, + 1.6026e-03, 6.5071e-04, -1.3013e-02, 2.1647e-02, -3.1433e-02], + device='cuda:0'), grad: tensor([ 3.2514e-05, 1.3456e-05, -6.2883e-05, 8.5384e-06, 7.2271e-06, + 4.0717e-06, -8.9943e-05, 4.9591e-05, 1.7434e-05, 1.9893e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 218.42, cls_loss 0.0018 cls_loss_mapping 0.0081 cls_loss_causal 0.5945 re_mapping 0.0081 re_causal 0.0268 /// teacc 99.05 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0141, -0.0158, 0.0807, ..., 0.0024, -0.0456, -0.0403], + [-0.0755, -0.0684, -0.0938, ..., 0.0255, 0.0473, 0.1931], + [-0.0427, -0.0488, -0.0495, ..., -0.0608, 0.0201, -0.1177], + ..., + [-0.0996, -0.0998, 0.0257, ..., -0.1201, -0.0593, -0.1764], + [ 0.0682, 0.0307, -0.0561, ..., -0.0960, -0.0009, -0.0987], + [ 0.0175, 0.0468, -0.0684, ..., 0.0212, -0.0906, -0.0446]], + device='cuda:0'), grad: tensor([[ 2.2948e-06, -1.1316e-06, -4.7833e-06, ..., -3.7625e-07, + 7.2196e-06, 4.5747e-06], + [ 6.6217e-07, 8.7172e-07, 5.1688e-07, ..., -1.8179e-06, + 1.0815e-03, 6.8188e-04], + [ 4.8522e-07, 1.0533e-06, 9.2573e-07, ..., 1.2275e-06, + -1.1349e-03, -7.2050e-04], + ..., + [ 4.0717e-06, 5.1670e-06, 8.9407e-07, ..., 5.2936e-06, + 1.2405e-05, 9.5889e-06], + [ 1.6699e-06, 2.2575e-06, 1.0505e-06, ..., 2.8536e-06, + 6.1318e-06, 4.2059e-06], + [-1.6421e-05, -1.1250e-05, 2.5760e-06, ..., -1.4342e-05, + 2.5779e-06, 1.4920e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0264, -0.0257, -0.0213, 0.0003, -0.0042, 0.0018, 0.0003, -0.0136, + 0.0222, -0.0318], device='cuda:0'), grad: tensor([ 1.7703e-05, 2.9564e-03, -3.0956e-03, 2.0504e-05, 5.1439e-05, + -6.2920e-06, 1.5408e-05, 3.9786e-05, 2.3022e-05, -2.4796e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 218.49, cls_loss 0.0031 cls_loss_mapping 0.0130 cls_loss_causal 0.5684 re_mapping 0.0081 re_causal 0.0256 /// teacc 98.90 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0149, -0.0160, 0.0812, ..., 0.0021, -0.0459, -0.0396], + [-0.0760, -0.0696, -0.0945, ..., 0.0256, 0.0477, 0.1961], + [-0.0434, -0.0491, -0.0507, ..., -0.0617, 0.0195, -0.1193], + ..., + [-0.1006, -0.1012, 0.0259, ..., -0.1207, -0.0595, -0.1798], + [ 0.0688, 0.0308, -0.0566, ..., -0.0966, -0.0003, -0.0995], + [ 0.0178, 0.0473, -0.0688, ..., 0.0221, -0.0919, -0.0452]], + device='cuda:0'), grad: tensor([[ 1.8459e-06, 2.3730e-06, -2.6450e-06, ..., 1.8533e-07, + 2.7120e-06, 1.0133e-06], + [ 1.3644e-06, 2.2687e-06, 9.7975e-07, ..., -7.7263e-06, + -2.2501e-06, -1.4767e-05], + [ 2.4617e-05, 6.3777e-05, 1.1679e-06, ..., 7.7784e-06, + 3.0413e-05, 1.0617e-06], + ..., + [ 3.3583e-06, 5.8971e-06, 1.2722e-06, ..., 5.2415e-06, + 4.3586e-06, 4.0345e-06], + [-1.3061e-05, -1.3687e-05, 7.5996e-07, ..., -1.0692e-05, + -2.1547e-05, 3.3788e-06], + [-1.4879e-05, -1.2353e-05, 1.8179e-05, ..., -7.8306e-06, + 1.3508e-05, 1.9334e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([-2.6562e-02, -2.5616e-02, -2.1955e-02, 3.1365e-05, -3.6911e-03, + 2.2127e-03, 2.7310e-04, -1.3563e-02, 2.2562e-02, -3.2157e-02], + device='cuda:0'), grad: tensor([ 1.0148e-05, -8.7991e-06, 1.2922e-04, -1.6809e-04, -5.1558e-05, + 3.9220e-05, 5.2869e-05, 1.5765e-05, -5.3048e-05, 3.4511e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 218.34, cls_loss 0.0027 cls_loss_mapping 0.0088 cls_loss_causal 0.5911 re_mapping 0.0077 re_causal 0.0254 /// teacc 98.98 lr 0.00010000 +Epoch 91, weight, value: tensor([[-1.4960e-02, -1.6695e-02, 8.0342e-02, ..., 1.4680e-03, + -4.6905e-02, -3.9990e-02], + [-7.6850e-02, -7.0190e-02, -9.5295e-02, ..., 2.5631e-02, + 4.7845e-02, 1.9733e-01], + [-4.3648e-02, -4.9553e-02, -5.1229e-02, ..., -6.2035e-02, + 1.9511e-02, -1.1917e-01], + ..., + [-1.0151e-01, -1.0165e-01, 2.5926e-02, ..., -1.2189e-01, + -5.9759e-02, -1.8154e-01], + [ 6.9167e-02, 3.0689e-02, -5.7051e-02, ..., -9.7042e-02, + -1.2919e-04, -1.0082e-01], + [ 1.8712e-02, 4.7687e-02, -6.9359e-02, ..., 2.2813e-02, + -9.2588e-02, -4.5859e-02]], device='cuda:0'), grad: tensor([[-2.9588e-04, 1.3493e-05, -9.9277e-04, ..., -1.5736e-04, + -4.2820e-04, 1.6298e-09], + [ 4.9546e-06, 3.5912e-06, 6.7316e-06, ..., 4.9919e-06, + 6.1952e-06, -2.2841e-07], + [ 1.4925e-04, 3.6895e-05, 3.7479e-04, ..., 1.0657e-04, + 1.8406e-04, 1.1176e-08], + ..., + [ 4.0196e-06, 3.2242e-06, 4.1164e-06, ..., 3.5819e-06, + 5.7556e-06, 8.6147e-09], + [ 6.5088e-05, 1.0379e-05, 1.9300e-04, ..., 5.4657e-05, + 9.0182e-05, 3.0966e-08], + [ 3.7283e-05, 2.6692e-06, 1.4222e-04, ..., 6.1877e-06, + 9.1791e-05, 5.2387e-08]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0276, -0.0259, -0.0214, -0.0001, -0.0036, 0.0026, 0.0009, -0.0141, + 0.0221, -0.0317], device='cuda:0'), grad: tensor([-2.7504e-03, 2.3484e-05, 1.0977e-03, -1.1629e-04, 6.7139e-04, + 6.7532e-05, 1.3709e-04, 1.6645e-05, 5.3358e-04, 3.2020e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 218.28, cls_loss 0.0024 cls_loss_mapping 0.0091 cls_loss_causal 0.5988 re_mapping 0.0079 re_causal 0.0248 /// teacc 98.91 lr 0.00010000 +Epoch 92, weight, value: tensor([[-1.5069e-02, -1.6743e-02, 8.0770e-02, ..., 1.3270e-03, + -4.6842e-02, -4.0025e-02], + [-7.7238e-02, -7.0436e-02, -9.5869e-02, ..., 2.5897e-02, + 4.7847e-02, 1.9869e-01], + [-4.4231e-02, -5.0242e-02, -5.2391e-02, ..., -6.2678e-02, + 1.9534e-02, -1.1905e-01], + ..., + [-1.0191e-01, -1.0237e-01, 2.5854e-02, ..., -1.2242e-01, + -5.9829e-02, -1.8323e-01], + [ 6.9597e-02, 3.1021e-02, -5.7602e-02, ..., -9.7787e-02, + 1.0676e-05, -1.0185e-01], + [ 1.9195e-02, 4.7898e-02, -7.0084e-02, ..., 2.2987e-02, + -9.3154e-02, -4.6787e-02]], device='cuda:0'), grad: tensor([[ 4.7009e-07, 3.3667e-07, 1.2713e-06, ..., 6.8359e-07, + 1.3430e-06, 1.1432e-07], + [ 7.3249e-07, 4.4121e-07, 3.9637e-06, ..., -9.6485e-07, + 2.8685e-06, -4.3660e-06], + [ 3.1069e-06, 1.7434e-06, 1.2415e-06, ..., 2.7604e-06, + 2.5257e-06, 9.8906e-07], + ..., + [ 2.1402e-06, 1.2191e-06, 6.7689e-06, ..., 1.9781e-06, + 2.7902e-06, 8.3772e-07], + [ 7.9628e-07, 5.2201e-07, 1.0356e-06, ..., 1.3905e-06, + 1.4640e-06, 8.1584e-07], + [ 5.6345e-08, 1.1199e-07, 1.0245e-05, ..., -8.4424e-07, + 3.1386e-06, 2.1956e-07]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0275, -0.0260, -0.0210, -0.0002, -0.0041, 0.0022, 0.0028, -0.0140, + 0.0217, -0.0321], device='cuda:0'), grad: tensor([ 4.3176e-06, 5.5656e-06, 8.2254e-06, -1.5602e-05, -6.2048e-05, + 1.7080e-06, 7.7263e-06, 1.9923e-05, 5.0664e-06, 2.5094e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 218.29, cls_loss 0.0028 cls_loss_mapping 0.0090 cls_loss_causal 0.5810 re_mapping 0.0078 re_causal 0.0249 /// teacc 98.90 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0153, -0.0168, 0.0806, ..., 0.0016, -0.0466, -0.0375], + [-0.0778, -0.0710, -0.0974, ..., 0.0256, 0.0477, 0.1992], + [-0.0448, -0.0510, -0.0530, ..., -0.0632, 0.0199, -0.1194], + ..., + [-0.1032, -0.1032, 0.0257, ..., -0.1227, -0.0604, -0.1839], + [ 0.0704, 0.0313, -0.0579, ..., -0.0982, 0.0008, -0.1023], + [ 0.0196, 0.0479, -0.0705, ..., 0.0233, -0.0937, -0.0472]], + device='cuda:0'), grad: tensor([[ 1.9949e-06, -3.4478e-06, -5.7369e-06, ..., -2.2855e-06, + 1.0310e-06, 1.9139e-07], + [ 1.2785e-05, 7.9274e-06, 1.1027e-06, ..., 8.9183e-06, + 9.3952e-06, 1.1008e-06], + [ 1.8269e-05, 9.5665e-06, 2.5444e-06, ..., 4.1947e-06, + 2.0373e-07, -3.5986e-06], + ..., + [ 1.0413e-04, 1.5533e-04, 2.3507e-06, ..., 2.1648e-04, + -9.8348e-07, 7.6508e-07], + [-3.6489e-06, 2.0787e-06, 1.6969e-06, ..., 3.3434e-06, + -8.2701e-06, 6.3563e-07], + [ 8.0705e-05, 6.3419e-05, 1.6406e-05, ..., 8.9884e-05, + 2.5719e-05, 1.0058e-07]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0276, -0.0265, -0.0197, 0.0007, -0.0040, 0.0031, 0.0019, -0.0157, + 0.0221, -0.0324], device='cuda:0'), grad: tensor([-4.4554e-06, 7.8440e-05, 1.7206e-07, -5.1737e-04, -6.2048e-05, + 5.4181e-05, 5.8813e-07, 1.9550e-04, -4.4517e-07, 2.5535e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 218.09, cls_loss 0.0021 cls_loss_mapping 0.0059 cls_loss_causal 0.5659 re_mapping 0.0079 re_causal 0.0250 /// teacc 98.99 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0158, -0.0171, 0.0809, ..., 0.0015, -0.0469, -0.0379], + [-0.0790, -0.0717, -0.0976, ..., 0.0257, 0.0478, 0.2017], + [-0.0455, -0.0516, -0.0534, ..., -0.0638, 0.0198, -0.1203], + ..., + [-0.1031, -0.1038, 0.0257, ..., -0.1231, -0.0603, -0.1865], + [ 0.0708, 0.0317, -0.0582, ..., -0.0987, 0.0006, -0.1047], + [ 0.0203, 0.0484, -0.0706, ..., 0.0241, -0.0940, -0.0481]], + device='cuda:0'), grad: tensor([[ 5.4054e-06, -3.6299e-05, -3.5793e-05, ..., -1.5542e-05, + 2.1309e-06, 1.2340e-08], + [ 1.9483e-06, 2.8480e-06, 1.6214e-06, ..., 4.5355e-07, + 8.5402e-07, -5.5972e-07], + [ 9.3639e-05, 3.7670e-05, 3.6471e-06, ..., 2.3246e-06, + 1.7002e-05, 5.3085e-08], + ..., + [ 2.0474e-05, 1.1288e-05, 2.4233e-06, ..., 9.0152e-07, + 5.4725e-06, 8.3586e-08], + [-2.4581e-04, -7.7367e-05, 1.9923e-05, ..., 7.6741e-06, + -5.8860e-05, 1.5972e-07], + [ 2.9318e-06, 3.8564e-05, 4.5419e-05, ..., 2.3730e-06, + 2.2024e-05, 4.9826e-08]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0279, -0.0270, -0.0198, 0.0006, -0.0045, 0.0032, 0.0021, -0.0150, + 0.0219, -0.0321], device='cuda:0'), grad: tensor([-6.3360e-05, 2.6509e-05, 2.1672e-04, 4.1342e-04, -1.2159e-04, + -1.8030e-06, 3.1013e-06, -7.2956e-05, -5.4932e-04, 1.4901e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 218.56, cls_loss 0.0024 cls_loss_mapping 0.0088 cls_loss_causal 0.5621 re_mapping 0.0081 re_causal 0.0246 /// teacc 98.92 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0136, -0.0157, 0.0826, ..., 0.0034, -0.0472, -0.0382], + [-0.0791, -0.0721, -0.0978, ..., 0.0261, 0.0480, 0.2033], + [-0.0457, -0.0518, -0.0541, ..., -0.0644, 0.0196, -0.1207], + ..., + [-0.1035, -0.1047, 0.0262, ..., -0.1236, -0.0604, -0.1892], + [ 0.0710, 0.0323, -0.0580, ..., -0.0997, 0.0009, -0.1063], + [ 0.0196, 0.0483, -0.0725, ..., 0.0233, -0.0946, -0.0487]], + device='cuda:0'), grad: tensor([[ 5.3784e-07, -4.8913e-06, -7.2643e-06, ..., -2.3991e-06, + 6.2957e-07, 5.5647e-08], + [ 1.9055e-06, 1.1018e-06, 4.3237e-07, ..., 3.1665e-08, + 4.9360e-07, -1.1241e-06], + [ 6.6422e-06, 2.8498e-06, 1.6242e-06, ..., 1.1520e-06, + -2.2277e-06, 1.1222e-07], + ..., + [ 7.3731e-05, 6.5453e-06, 8.3959e-07, ..., 1.9930e-06, + 9.6764e-07, 3.6275e-07], + [ 1.2098e-06, 2.0817e-05, 3.9339e-06, ..., 5.5544e-06, + 3.2294e-07, 1.9558e-07], + [-7.5437e-06, -1.1353e-06, 2.2426e-06, ..., -5.9344e-06, + 7.0501e-07, 1.0058e-07]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0256, -0.0268, -0.0199, 0.0006, -0.0047, 0.0030, 0.0017, -0.0148, + 0.0221, -0.0332], device='cuda:0'), grad: tensor([-6.3330e-06, 1.4231e-05, 1.6794e-05, -3.1185e-04, 1.0505e-05, + -8.2433e-05, 9.0972e-06, 3.0661e-04, 4.2796e-05, -8.4052e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 218.91, cls_loss 0.0022 cls_loss_mapping 0.0084 cls_loss_causal 0.5578 re_mapping 0.0080 re_causal 0.0240 /// teacc 98.95 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0137, -0.0158, 0.0830, ..., 0.0035, -0.0474, -0.0384], + [-0.0798, -0.0730, -0.0989, ..., 0.0258, 0.0478, 0.2037], + [-0.0462, -0.0523, -0.0547, ..., -0.0648, 0.0198, -0.1210], + ..., + [-0.1043, -0.1054, 0.0261, ..., -0.1245, -0.0607, -0.1908], + [ 0.0711, 0.0322, -0.0583, ..., -0.1005, 0.0012, -0.1066], + [ 0.0203, 0.0485, -0.0732, ..., 0.0242, -0.0954, -0.0492]], + device='cuda:0'), grad: tensor([[ 1.4738e-07, 3.4925e-07, 1.4547e-06, ..., -7.2550e-07, + 3.1702e-06, 2.6589e-07], + [ 2.5192e-07, 4.0755e-06, 4.9397e-06, ..., -2.8126e-07, + 5.7667e-06, -1.9241e-06], + [ 4.8941e-07, 9.7696e-07, 4.9137e-06, ..., 5.5786e-07, + 5.3048e-06, 1.0487e-06], + ..., + [ 4.4261e-07, 8.0690e-06, 1.7762e-05, ..., 3.1129e-07, + 1.5259e-05, 1.2536e-06], + [-1.0226e-06, -6.1747e-07, 1.5408e-05, ..., 7.1619e-07, + -7.1451e-06, 2.8289e-07], + [ 2.7772e-06, 9.7826e-06, 1.5236e-05, ..., 2.9919e-07, + 2.3678e-05, 9.2108e-07]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0257, -0.0274, -0.0192, 0.0005, -0.0042, 0.0028, 0.0023, -0.0153, + 0.0218, -0.0331], device='cuda:0'), grad: tensor([ 1.3694e-05, 3.4690e-05, 1.2279e-05, 1.1139e-05, -4.5681e-04, + 1.6257e-05, 2.4986e-04, 6.5804e-05, -5.3227e-05, 1.0592e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 218.17, cls_loss 0.0023 cls_loss_mapping 0.0077 cls_loss_causal 0.5739 re_mapping 0.0074 re_causal 0.0236 /// teacc 98.95 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0134, -0.0148, 0.0839, ..., 0.0040, -0.0476, -0.0387], + [-0.0803, -0.0737, -0.0990, ..., 0.0263, 0.0481, 0.2056], + [-0.0468, -0.0532, -0.0551, ..., -0.0654, 0.0196, -0.1216], + ..., + [-0.1053, -0.1060, 0.0267, ..., -0.1245, -0.0609, -0.1923], + [ 0.0713, 0.0322, -0.0589, ..., -0.1013, 0.0016, -0.1077], + [ 0.0210, 0.0484, -0.0742, ..., 0.0243, -0.0959, -0.0498]], + device='cuda:0'), grad: tensor([[ 2.4326e-06, -1.1921e-06, -3.3323e-06, ..., -8.0559e-07, + 1.6298e-06, 0.0000e+00], + [ 2.1309e-06, 2.6990e-06, 5.5553e-07, ..., 8.7917e-06, + 4.2915e-06, 0.0000e+00], + [-9.6858e-06, 1.3877e-06, 1.4370e-06, ..., 1.6699e-06, + -1.3268e-04, 0.0000e+00], + ..., + [ 1.7155e-06, 4.8950e-06, -3.8780e-06, ..., -8.1658e-06, + 1.1206e-04, 0.0000e+00], + [ 5.2661e-05, 6.0797e-05, 4.8950e-06, ..., 3.8892e-05, + 8.9407e-06, 0.0000e+00], + [-7.2777e-05, -6.2406e-05, 2.7157e-06, ..., -4.0799e-05, + 1.2247e-06, 0.0000e+00]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0252, -0.0272, -0.0194, 0.0004, -0.0041, 0.0028, 0.0018, -0.0149, + 0.0219, -0.0334], device='cuda:0'), grad: tensor([ 3.3677e-05, 5.4693e-04, -9.6321e-04, 4.1819e-04, 1.9014e-05, + -7.1287e-05, -1.1183e-05, -5.7727e-05, 2.0182e-04, -1.1563e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 218.28, cls_loss 0.0018 cls_loss_mapping 0.0078 cls_loss_causal 0.5552 re_mapping 0.0076 re_causal 0.0237 /// teacc 99.05 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0136, -0.0145, 0.0844, ..., 0.0041, -0.0478, -0.0396], + [-0.0810, -0.0740, -0.0985, ..., 0.0270, 0.0483, 0.2079], + [-0.0475, -0.0537, -0.0555, ..., -0.0658, 0.0197, -0.1220], + ..., + [-0.1061, -0.1065, 0.0265, ..., -0.1252, -0.0613, -0.1957], + [ 0.0714, 0.0317, -0.0591, ..., -0.1020, 0.0018, -0.1084], + [ 0.0213, 0.0483, -0.0745, ..., 0.0248, -0.0965, -0.0510]], + device='cuda:0'), grad: tensor([[ 5.1549e-07, -3.5726e-06, -5.0152e-07, ..., 1.1614e-06, + 2.7120e-06, 5.9139e-08], + [ 3.9376e-06, 3.8813e-07, 6.5744e-05, ..., 7.1883e-05, + 8.2143e-07, -1.4659e-06], + [ 4.6082e-06, 1.2051e-06, 5.5805e-06, ..., 3.2559e-06, + 4.7758e-06, 1.8417e-07], + ..., + [-8.1241e-05, 5.1130e-07, -1.0657e-04, ..., -1.1939e-04, + -5.5321e-06, 1.7160e-07], + [-3.2596e-06, -1.2880e-06, 5.2117e-06, ..., 2.9709e-06, + -7.3574e-07, 3.5157e-07], + [ 9.3728e-06, 4.7125e-07, 2.6181e-05, ..., 2.4751e-05, + 2.0117e-06, 6.7754e-08]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0251, -0.0271, -0.0194, 0.0007, -0.0037, 0.0025, 0.0010, -0.0148, + 0.0219, -0.0336], device='cuda:0'), grad: tensor([ 1.2055e-05, 4.9591e-04, 3.4243e-05, 3.0303e-04, 1.8656e-04, + 5.1379e-05, -1.5616e-04, -1.1559e-03, 7.0557e-06, 2.2185e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 218.54, cls_loss 0.0023 cls_loss_mapping 0.0078 cls_loss_causal 0.5886 re_mapping 0.0072 re_causal 0.0230 /// teacc 98.80 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0141, -0.0146, 0.0844, ..., 0.0040, -0.0488, -0.0401], + [-0.0815, -0.0752, -0.0991, ..., 0.0270, 0.0483, 0.2093], + [-0.0485, -0.0560, -0.0564, ..., -0.0664, 0.0195, -0.1227], + ..., + [-0.1073, -0.1069, 0.0276, ..., -0.1258, -0.0615, -0.1970], + [ 0.0720, 0.0314, -0.0593, ..., -0.1029, 0.0024, -0.1087], + [ 0.0222, 0.0488, -0.0742, ..., 0.0256, -0.0971, -0.0515]], + device='cuda:0'), grad: tensor([[ 7.9162e-07, 1.8207e-07, -7.3249e-07, ..., 5.7742e-08, + 2.8289e-07, 1.9255e-07], + [ 2.4121e-07, 4.0047e-07, 3.2922e-07, ..., -1.8366e-06, + -3.9227e-06, -8.3521e-06], + [ 2.6496e-07, 7.0408e-07, 5.0012e-07, ..., 7.9535e-07, + -2.6375e-06, 1.6438e-06], + ..., + [ 1.9632e-06, 1.9372e-06, -1.0729e-06, ..., 2.3786e-06, + 3.9637e-06, 2.7902e-06], + [ 4.5775e-07, 2.6710e-06, 4.3050e-07, ..., 8.6892e-07, + 5.2480e-07, 7.0408e-07], + [-1.0237e-05, -6.3777e-06, 8.0001e-07, ..., -8.2701e-06, + 4.8708e-07, 4.7614e-07]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0256, -0.0275, -0.0197, 0.0004, -0.0044, 0.0030, 0.0010, -0.0142, + 0.0226, -0.0333], device='cuda:0'), grad: tensor([ 4.2506e-06, -5.3160e-06, -5.7876e-05, 5.4240e-05, 2.0251e-05, + -1.8430e-04, 1.3793e-04, 3.4332e-05, 9.0450e-06, -1.2808e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 218.45, cls_loss 0.0023 cls_loss_mapping 0.0073 cls_loss_causal 0.5788 re_mapping 0.0074 re_causal 0.0236 /// teacc 98.94 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0143, -0.0145, 0.0847, ..., 0.0039, -0.0490, -0.0402], + [-0.0822, -0.0762, -0.0996, ..., 0.0270, 0.0479, 0.2104], + [-0.0494, -0.0568, -0.0568, ..., -0.0670, 0.0201, -0.1235], + ..., + [-0.1083, -0.1074, 0.0274, ..., -0.1266, -0.0618, -0.1985], + [ 0.0721, 0.0318, -0.0597, ..., -0.1035, 0.0027, -0.1091], + [ 0.0225, 0.0487, -0.0745, ..., 0.0260, -0.0979, -0.0520]], + device='cuda:0'), grad: tensor([[ 1.6391e-07, -4.5076e-07, -8.6706e-07, ..., -4.2235e-07, + 7.4971e-07, 5.1269e-07], + [ 1.1846e-06, 1.0822e-06, 1.9581e-07, ..., 3.3411e-07, + -3.2163e-04, -3.1066e-04], + [ 1.0617e-06, 1.1455e-06, 4.0093e-07, ..., 1.0841e-06, + 2.8801e-04, 2.7752e-04], + ..., + [ 4.6603e-06, 3.8296e-06, -1.4819e-05, ..., -7.8157e-06, + 2.0221e-05, 1.9729e-05], + [ 1.0962e-06, 2.2762e-06, 2.5658e-07, ..., 8.1025e-07, + 2.2314e-06, 1.7155e-06], + [-1.4730e-05, -9.8646e-06, 1.2569e-05, ..., -9.5088e-07, + 1.4119e-06, 1.0133e-06]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0256, -0.0283, -0.0189, 0.0009, -0.0042, 0.0025, 0.0013, -0.0144, + 0.0230, -0.0336], device='cuda:0'), grad: tensor([ 5.4985e-06, -1.5726e-03, 1.4219e-03, 3.4124e-05, 3.1024e-05, + 1.3418e-05, 6.3740e-06, -8.9556e-06, 1.5274e-05, 5.3108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 218.46, cls_loss 0.0023 cls_loss_mapping 0.0074 cls_loss_causal 0.5747 re_mapping 0.0073 re_causal 0.0233 /// teacc 99.02 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0145, -0.0142, 0.0852, ..., 0.0042, -0.0496, -0.0406], + [-0.0829, -0.0771, -0.1001, ..., 0.0287, 0.0487, 0.2125], + [-0.0498, -0.0575, -0.0576, ..., -0.0676, 0.0197, -0.1252], + ..., + [-0.1088, -0.1080, 0.0274, ..., -0.1289, -0.0627, -0.2002], + [ 0.0721, 0.0319, -0.0596, ..., -0.1043, 0.0032, -0.1092], + [ 0.0235, 0.0491, -0.0744, ..., 0.0271, -0.0983, -0.0529]], + device='cuda:0'), grad: tensor([[ 6.8918e-07, -3.3155e-06, 3.2365e-05, ..., 1.1437e-05, + 7.8008e-06, 2.3693e-05], + [ 3.2457e-07, 6.1793e-07, 3.0082e-06, ..., 7.5437e-07, + -1.4281e-04, -1.7822e-04], + [ 2.1346e-06, 2.2724e-06, 4.0829e-06, ..., 2.0079e-06, + 1.1730e-04, 1.1623e-04], + ..., + [-1.0412e-06, 7.7160e-07, 2.0419e-07, ..., 8.6986e-07, + 1.5542e-05, 1.8612e-05], + [-4.7907e-06, -9.3225e-07, 5.2005e-06, ..., 3.5297e-06, + -2.3469e-06, 1.5395e-06], + [-3.3919e-06, 2.3805e-06, 6.4112e-06, ..., -9.4110e-07, + 4.4927e-06, 2.3283e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0257, -0.0279, -0.0189, 0.0003, -0.0043, 0.0029, 0.0009, -0.0147, + 0.0235, -0.0332], device='cuda:0'), grad: tensor([ 2.7633e-04, -3.6192e-04, 9.4533e-05, 4.5627e-05, 6.1154e-05, + 9.5427e-05, -2.3544e-04, -4.9286e-06, 7.0594e-06, 2.1890e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 218.34, cls_loss 0.0025 cls_loss_mapping 0.0093 cls_loss_causal 0.5774 re_mapping 0.0075 re_causal 0.0236 /// teacc 98.97 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0144, -0.0134, 0.0861, ..., 0.0054, -0.0486, -0.0382], + [-0.0821, -0.0768, -0.1023, ..., 0.0287, 0.0491, 0.2144], + [-0.0505, -0.0593, -0.0595, ..., -0.0694, 0.0198, -0.1260], + ..., + [-0.1095, -0.1088, 0.0282, ..., -0.1298, -0.0636, -0.2038], + [ 0.0731, 0.0329, -0.0601, ..., -0.1039, 0.0035, -0.1107], + [ 0.0230, 0.0478, -0.0755, ..., 0.0266, -0.1000, -0.0540]], + device='cuda:0'), grad: tensor([[ 5.2573e-07, -7.4320e-07, -3.6526e-06, ..., -1.3979e-06, + 1.1548e-06, 3.8161e-07], + [ 6.6031e-07, 4.9546e-07, 2.2165e-06, ..., -3.0585e-06, + 4.6231e-06, -7.9498e-06], + [ 2.6617e-06, 3.4142e-06, 4.6417e-06, ..., 4.5225e-06, + 4.5113e-06, 1.4920e-06], + ..., + [-1.3830e-06, 7.9442e-07, 1.2182e-06, ..., 1.9893e-06, + -2.0906e-05, 1.7928e-06], + [ 3.3025e-06, 3.1795e-06, 3.4515e-06, ..., 5.6587e-06, + 4.1090e-06, 9.8161e-07], + [-6.7838e-06, -2.2259e-06, 6.1691e-06, ..., -6.0387e-06, + 9.5814e-06, 1.2415e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0247, -0.0278, -0.0190, -0.0002, -0.0039, 0.0033, 0.0007, -0.0143, + 0.0238, -0.0346], device='cuda:0'), grad: tensor([-5.4948e-06, 2.2039e-05, 2.8387e-05, 4.1217e-05, 1.8841e-06, + 3.9935e-05, -4.5896e-05, -1.4496e-04, 3.3498e-05, 2.9519e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 218.67, cls_loss 0.0020 cls_loss_mapping 0.0079 cls_loss_causal 0.5706 re_mapping 0.0074 re_causal 0.0236 /// teacc 98.99 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0145, -0.0131, 0.0862, ..., 0.0055, -0.0490, -0.0384], + [-0.0825, -0.0770, -0.1030, ..., 0.0293, 0.0487, 0.2155], + [-0.0511, -0.0600, -0.0605, ..., -0.0701, 0.0202, -0.1262], + ..., + [-0.1103, -0.1092, 0.0296, ..., -0.1305, -0.0633, -0.2053], + [ 0.0734, 0.0327, -0.0604, ..., -0.1044, 0.0038, -0.1117], + [ 0.0237, 0.0477, -0.0762, ..., 0.0272, -0.1007, -0.0545]], + device='cuda:0'), grad: tensor([[ 8.6613e-07, -2.0694e-06, -4.0010e-06, ..., -2.6859e-06, + 2.9267e-07, 3.8883e-08], + [ 2.2650e-06, 1.1958e-06, 4.1001e-07, ..., 6.0536e-07, + -9.2853e-07, -3.1814e-06], + [ 7.3798e-06, 3.7402e-06, 8.7218e-07, ..., 6.0722e-06, + 1.5274e-06, 1.2517e-06], + ..., + [ 9.0748e-06, 4.5858e-06, 2.0466e-07, ..., 7.4357e-06, + 1.9483e-06, 1.0747e-06], + [ 1.3839e-06, 1.2349e-06, 2.8573e-06, ..., 4.0159e-06, + -3.5157e-07, 2.0722e-07], + [-2.7552e-05, -8.5961e-07, -1.4886e-05, ..., -2.1935e-05, + -4.1202e-06, 1.8603e-07]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0248, -0.0281, -0.0192, -0.0004, -0.0039, 0.0032, 0.0009, -0.0137, + 0.0236, -0.0347], device='cuda:0'), grad: tensor([-4.4256e-06, 4.8243e-06, 1.6719e-05, -6.0111e-05, 1.1218e-04, + 2.4483e-05, -2.4989e-05, 2.4512e-06, 1.1079e-05, -8.2195e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 218.58, cls_loss 0.0019 cls_loss_mapping 0.0076 cls_loss_causal 0.5683 re_mapping 0.0073 re_causal 0.0229 /// teacc 99.01 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0147, -0.0126, 0.0865, ..., 0.0057, -0.0496, -0.0386], + [-0.0828, -0.0774, -0.1037, ..., 0.0294, 0.0491, 0.2177], + [-0.0526, -0.0604, -0.0611, ..., -0.0706, 0.0200, -0.1277], + ..., + [-0.1111, -0.1096, 0.0296, ..., -0.1314, -0.0635, -0.2069], + [ 0.0739, 0.0325, -0.0609, ..., -0.1050, 0.0042, -0.1125], + [ 0.0243, 0.0478, -0.0766, ..., 0.0281, -0.1013, -0.0553]], + device='cuda:0'), grad: tensor([[ 7.5717e-07, -2.0340e-05, -4.4912e-05, ..., -3.0875e-05, + 1.3057e-06, -9.5367e-07], + [ 5.1707e-06, 9.7230e-07, 8.8941e-07, ..., -1.1295e-05, + 5.8524e-06, -1.6853e-05], + [ 2.0377e-06, 1.3895e-06, 2.0582e-06, ..., 2.9393e-06, + 2.8368e-06, 1.0850e-06], + ..., + [-2.5630e-05, 2.0377e-06, 7.0734e-07, ..., 5.3979e-06, + -7.0930e-05, 4.1761e-06], + [ 6.0052e-06, 9.4064e-07, 1.7453e-06, ..., 4.4554e-06, + 1.2986e-05, 2.2780e-06], + [-1.3359e-05, -3.0156e-06, 2.7269e-06, ..., -4.8131e-06, + 5.3756e-06, 4.2953e-06]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0248, -0.0281, -0.0198, 0.0003, -0.0040, 0.0022, 0.0019, -0.0137, + 0.0237, -0.0344], device='cuda:0'), grad: tensor([-7.3850e-05, 4.0293e-05, -1.1986e-06, 1.8609e-04, 4.1962e-05, + 2.2158e-05, 4.9323e-05, -3.6383e-04, 6.9082e-05, 2.9907e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 218.14, cls_loss 0.0019 cls_loss_mapping 0.0061 cls_loss_causal 0.5615 re_mapping 0.0074 re_causal 0.0229 /// teacc 98.86 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0149, -0.0126, 0.0867, ..., 0.0057, -0.0500, -0.0388], + [-0.0832, -0.0788, -0.1040, ..., 0.0298, 0.0495, 0.2187], + [-0.0532, -0.0609, -0.0619, ..., -0.0712, 0.0198, -0.1282], + ..., + [-0.1113, -0.1102, 0.0282, ..., -0.1324, -0.0641, -0.2086], + [ 0.0742, 0.0325, -0.0612, ..., -0.1052, 0.0043, -0.1130], + [ 0.0251, 0.0482, -0.0763, ..., 0.0292, -0.1018, -0.0560]], + device='cuda:0'), grad: tensor([[ 1.3933e-06, 9.4902e-07, -4.6799e-08, ..., 5.9791e-07, + 1.0608e-06, 1.8673e-07], + [ 9.7696e-07, 9.8255e-07, 9.6858e-08, ..., -1.6978e-06, + -1.0952e-06, -6.0163e-06], + [ 2.3432e-06, 7.2690e-07, 3.6554e-08, ..., 9.3505e-07, + 9.5461e-07, 8.0140e-07], + ..., + [ 8.2254e-06, 1.3612e-05, 2.9756e-07, ..., 1.5676e-05, + 1.3486e-06, 2.6356e-06], + [-5.1074e-06, 1.1548e-06, 2.2817e-08, ..., 5.1223e-07, + -3.4813e-06, 5.0478e-07], + [-9.6634e-06, -1.3322e-05, 3.4347e-06, ..., -1.5289e-05, + 3.6992e-06, 8.8615e-07]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0251, -0.0278, -0.0205, 0.0002, -0.0041, 0.0018, 0.0027, -0.0136, + 0.0237, -0.0340], device='cuda:0'), grad: tensor([ 5.1409e-06, -3.8743e-06, 6.0238e-06, -1.5721e-05, -4.5635e-06, + 1.2964e-05, 5.6922e-06, -4.5598e-06, -3.9116e-06, 2.7996e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 218.37, cls_loss 0.0027 cls_loss_mapping 0.0089 cls_loss_causal 0.5741 re_mapping 0.0069 re_causal 0.0217 /// teacc 98.89 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0164, -0.0141, 0.0866, ..., 0.0054, -0.0498, -0.0375], + [-0.0836, -0.0790, -0.1059, ..., 0.0292, 0.0490, 0.2197], + [-0.0543, -0.0616, -0.0627, ..., -0.0721, 0.0203, -0.1287], + ..., + [-0.1119, -0.1107, 0.0277, ..., -0.1331, -0.0649, -0.2112], + [ 0.0741, 0.0322, -0.0621, ..., -0.1060, 0.0043, -0.1136], + [ 0.0261, 0.0493, -0.0767, ..., 0.0304, -0.1034, -0.0565]], + device='cuda:0'), grad: tensor([[-6.0201e-06, -3.0234e-05, -5.4955e-05, ..., -3.5375e-05, + 1.1688e-06, 1.8161e-07], + [ 5.9325e-07, 2.1812e-06, 3.2466e-06, ..., 2.3516e-07, + -9.3784e-07, -4.4852e-06], + [ 1.5916e-06, 4.8466e-06, 9.6858e-06, ..., 6.5714e-06, + -3.0249e-06, 2.6985e-07], + ..., + [ 3.0063e-06, 2.1961e-06, 2.3730e-06, ..., 4.2990e-06, + 1.6484e-06, 8.1770e-07], + [-4.6380e-06, 4.8652e-06, 1.3612e-05, ..., 1.0222e-05, + 1.1269e-06, 3.9535e-07], + [-2.4140e-06, 3.4999e-06, 6.9030e-06, ..., -4.0717e-06, + 1.7229e-06, 9.0431e-07]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0259, -0.0286, -0.0204, 0.0013, -0.0031, 0.0013, 0.0023, -0.0133, + 0.0231, -0.0339], device='cuda:0'), grad: tensor([-1.3685e-04, 7.1861e-06, -6.8955e-06, 4.3601e-05, 1.6615e-05, + -6.4597e-06, 1.6347e-05, 2.0161e-05, 3.5435e-05, 1.0729e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 218.47, cls_loss 0.0021 cls_loss_mapping 0.0069 cls_loss_causal 0.5703 re_mapping 0.0072 re_causal 0.0223 /// teacc 98.98 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0167, -0.0142, 0.0864, ..., 0.0051, -0.0502, -0.0376], + [-0.0843, -0.0801, -0.1066, ..., 0.0297, 0.0494, 0.2209], + [-0.0551, -0.0632, -0.0637, ..., -0.0728, 0.0202, -0.1290], + ..., + [-0.1123, -0.1114, 0.0275, ..., -0.1345, -0.0658, -0.2141], + [ 0.0742, 0.0318, -0.0631, ..., -0.1069, 0.0044, -0.1142], + [ 0.0267, 0.0491, -0.0768, ..., 0.0314, -0.1038, -0.0571]], + device='cuda:0'), grad: tensor([[-6.2585e-05, -2.1541e-04, -1.8501e-04, ..., -1.8167e-04, + 4.0536e-07, 1.7299e-07], + [ 1.5385e-06, 2.8517e-06, 2.5276e-06, ..., 6.4680e-07, + -6.7800e-07, -4.3847e-06], + [ 4.6194e-06, 9.1940e-06, 1.1437e-05, ..., 1.1548e-05, + -4.9584e-06, 3.5367e-07], + ..., + [ 1.1787e-05, 1.2584e-05, 2.3976e-05, ..., 1.5363e-05, + -1.9521e-06, 6.5099e-07], + [ 5.1782e-06, 1.2808e-05, 1.0498e-05, ..., 1.1355e-05, + 2.3805e-06, 1.0626e-06], + [-1.6883e-05, 8.3074e-06, 5.8651e-05, ..., -3.2075e-06, + 4.5449e-07, 5.5972e-07]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0265, -0.0287, -0.0203, 0.0009, -0.0034, 0.0027, 0.0024, -0.0137, + 0.0229, -0.0338], device='cuda:0'), grad: tensor([-5.3549e-04, 7.1116e-06, 1.4737e-05, 1.2875e-04, -1.8430e-04, + 2.3985e-04, 5.4806e-05, 7.9393e-05, 4.8727e-05, 1.4663e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 218.48, cls_loss 0.0019 cls_loss_mapping 0.0068 cls_loss_causal 0.5598 re_mapping 0.0070 re_causal 0.0217 /// teacc 98.96 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0167, -0.0131, 0.0875, ..., 0.0059, -0.0504, -0.0378], + [-0.0841, -0.0808, -0.1074, ..., 0.0306, 0.0493, 0.2218], + [-0.0557, -0.0639, -0.0642, ..., -0.0736, 0.0204, -0.1294], + ..., + [-0.1135, -0.1123, 0.0271, ..., -0.1358, -0.0665, -0.2162], + [ 0.0746, 0.0323, -0.0631, ..., -0.1079, 0.0045, -0.1146], + [ 0.0274, 0.0489, -0.0772, ..., 0.0314, -0.1044, -0.0591]], + device='cuda:0'), grad: tensor([[ 6.2678e-07, -4.2305e-07, -1.5236e-06, ..., -6.0303e-08, + 8.8755e-07, 1.4491e-06], + [ 6.4960e-07, 9.1735e-07, -9.0152e-07, ..., -4.6976e-06, + -1.0833e-05, -1.8403e-05], + [ 1.6382e-06, 2.3320e-06, 1.0505e-06, ..., 9.4576e-07, + 5.9493e-06, 3.3937e-06], + ..., + [ 2.5053e-06, 3.0939e-06, 7.7905e-07, ..., 1.6578e-06, + -4.3772e-06, 2.5146e-06], + [ 5.2191e-06, 9.2238e-06, 3.9767e-07, ..., 1.9874e-06, + -4.9314e-07, 2.7195e-06], + [-2.8126e-06, 2.3730e-06, 7.9209e-07, ..., -8.6054e-06, + 1.3886e-06, 1.0710e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0259, -0.0285, -0.0198, 0.0006, -0.0032, 0.0026, 0.0025, -0.0139, + 0.0230, -0.0343], device='cuda:0'), grad: tensor([ 2.5798e-06, -5.5507e-06, 2.7478e-05, 3.7432e-04, 2.1547e-05, + -3.8338e-04, 6.2026e-06, -6.8843e-05, 2.0221e-05, 4.7274e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 218.34, cls_loss 0.0015 cls_loss_mapping 0.0060 cls_loss_causal 0.5577 re_mapping 0.0073 re_causal 0.0218 /// teacc 98.86 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0168, -0.0126, 0.0880, ..., 0.0062, -0.0505, -0.0381], + [-0.0846, -0.0812, -0.1080, ..., 0.0312, 0.0495, 0.2234], + [-0.0560, -0.0645, -0.0650, ..., -0.0744, 0.0207, -0.1297], + ..., + [-0.1138, -0.1134, 0.0269, ..., -0.1365, -0.0670, -0.2184], + [ 0.0748, 0.0323, -0.0638, ..., -0.1087, 0.0043, -0.1157], + [ 0.0279, 0.0493, -0.0776, ..., 0.0319, -0.1050, -0.0601]], + device='cuda:0'), grad: tensor([[ 6.9570e-07, -7.3854e-07, -3.4068e-06, ..., -1.1455e-07, + 1.0403e-06, 5.8021e-07], + [ 1.6866e-06, 3.1316e-07, 8.0932e-07, ..., 4.5402e-08, + 3.1274e-06, -3.9376e-06], + [ 1.0822e-06, 1.5376e-06, 1.5600e-06, ..., 1.3020e-06, + -2.6405e-05, 1.1232e-06], + ..., + [ 5.3421e-06, 1.0068e-06, 1.0934e-06, ..., 8.1658e-06, + 3.2391e-06, 6.0722e-07], + [-3.9749e-06, -6.7689e-06, 4.9639e-07, ..., 1.2778e-06, + 1.4022e-05, 6.7148e-07], + [-1.1504e-05, 1.4557e-06, 2.9579e-06, ..., -1.6838e-05, + 2.8666e-06, 2.0047e-07]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0256, -0.0284, -0.0193, 0.0003, -0.0032, 0.0024, 0.0032, -0.0140, + 0.0225, -0.0345], device='cuda:0'), grad: tensor([-2.2836e-06, 1.4707e-05, -6.6280e-05, 1.3150e-05, -5.2573e-07, + 3.7979e-06, 5.0366e-06, 2.8059e-05, 2.3380e-05, -1.9014e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 218.35, cls_loss 0.0023 cls_loss_mapping 0.0086 cls_loss_causal 0.5702 re_mapping 0.0069 re_causal 0.0214 /// teacc 98.92 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0170, -0.0122, 0.0885, ..., 0.0063, -0.0510, -0.0382], + [-0.0854, -0.0818, -0.1084, ..., 0.0313, 0.0491, 0.2242], + [-0.0563, -0.0650, -0.0660, ..., -0.0752, 0.0211, -0.1290], + ..., + [-0.1158, -0.1141, 0.0270, ..., -0.1374, -0.0672, -0.2197], + [ 0.0755, 0.0327, -0.0642, ..., -0.1089, 0.0046, -0.1159], + [ 0.0299, 0.0502, -0.0788, ..., 0.0312, -0.1072, -0.0607]], + device='cuda:0'), grad: tensor([[ 7.2122e-06, 1.1865e-06, 5.9512e-07, ..., 6.1542e-06, + 3.1572e-06, 4.0815e-07], + [ 5.9992e-05, 3.5428e-06, 6.5677e-06, ..., 5.1826e-05, + 1.9506e-05, -9.6411e-06], + [ 5.9865e-06, 2.5723e-06, 7.4413e-07, ..., 4.5411e-06, + 8.4639e-06, 2.3488e-06], + ..., + [ 1.2554e-05, 1.7658e-06, 1.3737e-06, ..., 1.1846e-05, + -6.0014e-06, 1.9837e-06], + [ 8.5607e-06, 3.0119e-06, 1.1595e-06, ..., 9.0525e-06, + 2.1234e-06, 8.9128e-07], + [-5.4073e-04, -4.6849e-05, -4.9204e-05, ..., -4.8184e-04, + -1.9991e-04, 9.2294e-07]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0257, -0.0290, -0.0189, 0.0005, -0.0025, 0.0024, 0.0034, -0.0144, + 0.0227, -0.0349], device='cuda:0'), grad: tensor([ 3.4928e-05, 2.9159e-04, 1.0175e-04, 5.1916e-05, 1.9398e-03, + 5.8413e-05, 6.1616e-06, -1.4853e-04, 3.2395e-05, -2.3708e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 218.40, cls_loss 0.0019 cls_loss_mapping 0.0057 cls_loss_causal 0.5719 re_mapping 0.0071 re_causal 0.0220 /// teacc 98.93 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0177, -0.0128, 0.0881, ..., 0.0058, -0.0513, -0.0385], + [-0.0858, -0.0827, -0.1088, ..., 0.0325, 0.0496, 0.2258], + [-0.0560, -0.0651, -0.0668, ..., -0.0756, 0.0210, -0.1294], + ..., + [-0.1156, -0.1146, 0.0267, ..., -0.1382, -0.0677, -0.2222], + [ 0.0754, 0.0331, -0.0640, ..., -0.1098, 0.0047, -0.1151], + [ 0.0310, 0.0509, -0.0788, ..., 0.0321, -0.1076, -0.0622]], + device='cuda:0'), grad: tensor([[ 3.1339e-07, -3.2759e-07, 1.6093e-06, ..., 8.0606e-07, + 1.9046e-07, 8.1956e-08], + [ 4.1276e-06, 6.1542e-06, 1.0291e-06, ..., 1.5683e-06, + 8.6706e-07, -1.5935e-06], + [ 1.8943e-06, 2.7101e-06, 1.4957e-06, ..., 1.7453e-06, + 7.3388e-07, 2.7893e-07], + ..., + [ 1.0096e-06, 1.5413e-06, 3.1013e-07, ..., 8.2375e-07, + 5.1828e-07, 4.2701e-07], + [ 7.8678e-06, 1.4484e-05, 3.2280e-06, ..., 6.3628e-06, + 1.8571e-06, 2.4750e-07], + [ 1.6727e-06, 2.5295e-06, 3.4831e-06, ..., 1.1036e-06, + 2.4494e-06, 1.9465e-07]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0263, -0.0287, -0.0185, 0.0002, -0.0027, 0.0018, 0.0042, -0.0143, + 0.0228, -0.0348], device='cuda:0'), grad: tensor([ 5.7891e-06, 1.3970e-05, 1.0177e-05, -5.8174e-04, -6.2492e-07, + 5.2547e-04, -2.1651e-05, 5.7556e-07, 3.6001e-05, 1.2085e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 110---------------------------------------------------- +epoch 110, time 219.32, cls_loss 0.0019 cls_loss_mapping 0.0058 cls_loss_causal 0.5679 re_mapping 0.0069 re_causal 0.0207 /// teacc 99.08 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0183, -0.0128, 0.0885, ..., 0.0058, -0.0520, -0.0387], + [-0.0866, -0.0839, -0.1096, ..., 0.0325, 0.0498, 0.2287], + [-0.0568, -0.0660, -0.0673, ..., -0.0765, 0.0207, -0.1323], + ..., + [-0.1166, -0.1158, 0.0267, ..., -0.1391, -0.0681, -0.2241], + [ 0.0761, 0.0332, -0.0642, ..., -0.1103, 0.0056, -0.1153], + [ 0.0317, 0.0515, -0.0791, ..., 0.0331, -0.1081, -0.0631]], + device='cuda:0'), grad: tensor([[ 8.1025e-07, 1.5218e-06, 3.2913e-06, ..., -1.6931e-06, + 1.3169e-06, 1.3481e-07], + [ 6.3237e-07, 4.2431e-06, 5.4650e-06, ..., -1.8831e-06, + -5.3179e-07, -5.0925e-06], + [ 9.7882e-07, 2.8629e-06, 2.1290e-06, ..., 1.1930e-06, + -1.1742e-05, 9.3412e-07], + ..., + [ 9.8255e-07, -9.5367e-06, -2.8461e-05, ..., 1.4286e-06, + 7.9209e-07, 8.7079e-07], + [-7.4804e-05, -3.3307e-04, 7.8371e-07, ..., 1.6233e-06, + -8.1301e-05, 1.4491e-06], + [-4.5449e-06, 6.2957e-07, 3.5614e-06, ..., -4.6417e-06, + 1.6252e-06, 5.8161e-07]], device='cuda:0') +Epoch 112, bias, value: tensor([-2.6519e-02, -2.8234e-02, -1.9693e-02, 8.9110e-05, -2.9717e-03, + 1.9723e-03, 3.9479e-03, -1.4073e-02, 2.3597e-02, -3.4553e-02], + device='cuda:0'), grad: tensor([ 7.3791e-05, 7.4029e-05, 3.2395e-05, 6.1572e-05, 6.5386e-05, + 6.6042e-04, 3.4857e-04, -4.1938e-04, -9.4032e-04, 4.2856e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 218.46, cls_loss 0.0021 cls_loss_mapping 0.0072 cls_loss_causal 0.5975 re_mapping 0.0067 re_causal 0.0216 /// teacc 99.00 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0182, -0.0122, 0.0888, ..., 0.0060, -0.0523, -0.0387], + [-0.0872, -0.0848, -0.1100, ..., 0.0328, 0.0499, 0.2302], + [-0.0575, -0.0666, -0.0678, ..., -0.0774, 0.0208, -0.1330], + ..., + [-0.1177, -0.1173, 0.0267, ..., -0.1405, -0.0682, -0.2259], + [ 0.0760, 0.0330, -0.0647, ..., -0.1113, 0.0055, -0.1162], + [ 0.0328, 0.0521, -0.0793, ..., 0.0343, -0.1081, -0.0641]], + device='cuda:0'), grad: tensor([[ 1.3122e-06, 1.1073e-06, -9.9093e-07, ..., 1.2219e-06, + 4.1798e-06, 1.5674e-06], + [ 1.9800e-06, 3.2280e-06, 1.3586e-07, ..., 3.9816e-04, + 1.7624e-03, 1.3285e-03], + [ 9.8497e-06, 1.7628e-05, 2.3621e-07, ..., 1.8016e-05, + 4.1604e-05, 2.1428e-05], + ..., + [ 2.9784e-06, 4.3325e-06, 1.3458e-07, ..., -4.1413e-04, + -1.8501e-03, -1.3981e-03], + [ 2.3432e-06, 4.6343e-06, 4.0606e-07, ..., 1.0364e-05, + 3.3319e-05, 2.2784e-05], + [-8.8103e-07, 9.8906e-07, 1.9546e-07, ..., 6.7241e-07, + 7.2382e-06, 4.0457e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0264, -0.0283, -0.0198, -0.0002, -0.0031, 0.0021, 0.0037, -0.0138, + 0.0230, -0.0341], device='cuda:0'), grad: tensor([ 1.3843e-05, 8.2779e-03, 1.7273e-04, 2.8834e-05, 1.2986e-05, + 1.5929e-05, -2.3860e-06, -8.6975e-03, 1.5306e-04, 2.7463e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 218.24, cls_loss 0.0015 cls_loss_mapping 0.0050 cls_loss_causal 0.5790 re_mapping 0.0065 re_causal 0.0209 /// teacc 98.99 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0180, -0.0120, 0.0893, ..., 0.0061, -0.0526, -0.0387], + [-0.0877, -0.0849, -0.1104, ..., 0.0351, 0.0494, 0.2313], + [-0.0576, -0.0668, -0.0685, ..., -0.0782, 0.0210, -0.1332], + ..., + [-0.1179, -0.1183, 0.0264, ..., -0.1430, -0.0676, -0.2263], + [ 0.0765, 0.0332, -0.0650, ..., -0.1115, 0.0060, -0.1167], + [ 0.0328, 0.0518, -0.0799, ..., 0.0343, -0.1089, -0.0649]], + device='cuda:0'), grad: tensor([[ 8.2282e-07, -4.7497e-06, -4.4554e-06, ..., -5.1968e-06, + 6.7614e-07, 1.7288e-07], + [ 9.3691e-07, 2.4531e-06, 5.8208e-09, ..., -6.0014e-06, + -1.4804e-05, -3.0786e-05], + [ 3.4943e-06, 6.0610e-06, 1.3802e-06, ..., 6.8657e-06, + 8.4266e-06, 1.5020e-05], + ..., + [ 1.8692e-06, 2.3264e-06, 5.6624e-07, ..., 2.8647e-06, + 3.0398e-06, 2.4214e-06], + [ 7.0482e-06, 8.5384e-06, 4.2748e-07, ..., 5.9120e-06, + 5.1595e-06, 1.0487e-06], + [-1.7295e-06, 3.2000e-06, 1.8626e-06, ..., -2.8871e-07, + 3.3602e-06, 1.6792e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0264, -0.0281, -0.0200, -0.0003, -0.0028, 0.0021, 0.0038, -0.0135, + 0.0230, -0.0346], device='cuda:0'), grad: tensor([-1.6332e-05, -4.8012e-05, 3.4004e-05, -4.6343e-05, 1.5192e-05, + -1.1288e-06, 1.4625e-05, 1.4797e-05, 2.8402e-05, 4.6417e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 218.39, cls_loss 0.0022 cls_loss_mapping 0.0066 cls_loss_causal 0.5701 re_mapping 0.0063 re_causal 0.0196 /// teacc 99.02 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0183, -0.0118, 0.0897, ..., 0.0060, -0.0529, -0.0390], + [-0.0885, -0.0856, -0.1108, ..., 0.0369, 0.0502, 0.2332], + [-0.0584, -0.0678, -0.0694, ..., -0.0793, 0.0211, -0.1335], + ..., + [-0.1191, -0.1205, 0.0253, ..., -0.1453, -0.0689, -0.2274], + [ 0.0765, 0.0332, -0.0651, ..., -0.1130, 0.0063, -0.1178], + [ 0.0337, 0.0524, -0.0798, ..., 0.0354, -0.1106, -0.0656]], + device='cuda:0'), grad: tensor([[ 1.0070e-07, -1.9252e-05, -1.5244e-05, ..., -1.7643e-05, + 5.3830e-07, 6.5309e-08], + [ 6.5472e-07, 8.4983e-07, 3.6927e-07, ..., 1.4808e-07, + 3.5530e-07, -1.9614e-06], + [ 6.6776e-07, 6.9290e-07, 4.7125e-07, ..., 8.1817e-07, + 1.0924e-06, 3.0780e-07], + ..., + [ 3.9814e-07, 2.8173e-07, -9.4296e-09, ..., 5.2946e-07, + 5.9977e-07, 5.7463e-07], + [ 2.3842e-07, 1.4966e-06, 4.7358e-07, ..., 1.8459e-06, + 5.9605e-07, 3.6904e-07], + [ 5.2713e-07, 1.4603e-06, 1.1623e-06, ..., 1.1194e-06, + 1.2591e-06, 1.5146e-07]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0265, -0.0277, -0.0197, -0.0007, -0.0021, 0.0025, 0.0032, -0.0138, + 0.0225, -0.0349], device='cuda:0'), grad: tensor([-3.7223e-05, 4.8708e-07, 3.2559e-06, -6.0908e-06, -4.7148e-07, + 2.2382e-05, 1.1511e-05, -6.8964e-07, 1.4333e-06, 5.4687e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 114---------------------------------------------------- +epoch 114, time 219.08, cls_loss 0.0017 cls_loss_mapping 0.0061 cls_loss_causal 0.5646 re_mapping 0.0067 re_causal 0.0204 /// teacc 99.11 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0183, -0.0113, 0.0900, ..., 0.0062, -0.0535, -0.0391], + [-0.0892, -0.0863, -0.1109, ..., 0.0381, 0.0505, 0.2348], + [-0.0587, -0.0683, -0.0702, ..., -0.0808, 0.0210, -0.1341], + ..., + [-0.1195, -0.1206, 0.0250, ..., -0.1463, -0.0693, -0.2285], + [ 0.0768, 0.0330, -0.0654, ..., -0.1135, 0.0068, -0.1184], + [ 0.0343, 0.0525, -0.0801, ..., 0.0359, -0.1110, -0.0665]], + device='cuda:0'), grad: tensor([[ 1.1455e-06, -1.6928e-05, -2.8297e-05, ..., -1.4283e-05, + -3.6992e-06, 7.7765e-08], + [ 9.6485e-07, 1.6019e-06, 8.1491e-07, ..., -1.2442e-06, + 1.3663e-06, -6.1095e-06], + [ 4.8690e-06, 8.9630e-06, 4.7088e-06, ..., 4.6194e-06, + -4.7505e-05, 4.2422e-07], + ..., + [ 1.4100e-06, 3.2280e-06, 2.7008e-06, ..., 2.4624e-06, + 4.6790e-05, 9.1735e-07], + [-6.2138e-06, -2.3749e-06, 2.4624e-06, ..., 5.7667e-06, + -4.7497e-06, 5.3551e-07], + [ 6.6049e-06, 9.9763e-06, 5.7928e-06, ..., 9.6112e-07, + 6.0052e-06, 2.8256e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0265, -0.0271, -0.0200, -0.0006, -0.0025, 0.0026, 0.0031, -0.0138, + 0.0223, -0.0348], device='cuda:0'), grad: tensor([-5.3883e-05, 9.5665e-06, -3.8815e-04, -4.4376e-05, 1.5512e-05, + 3.1561e-05, 9.3058e-06, 3.9029e-04, -2.9877e-05, 5.9187e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 218.46, cls_loss 0.0022 cls_loss_mapping 0.0083 cls_loss_causal 0.5426 re_mapping 0.0069 re_causal 0.0191 /// teacc 98.97 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0185, -0.0110, 0.0902, ..., 0.0062, -0.0540, -0.0392], + [-0.0906, -0.0872, -0.1119, ..., 0.0382, 0.0503, 0.2360], + [-0.0595, -0.0696, -0.0711, ..., -0.0824, 0.0214, -0.1339], + ..., + [-0.1201, -0.1215, 0.0247, ..., -0.1470, -0.0697, -0.2291], + [ 0.0768, 0.0324, -0.0659, ..., -0.1143, 0.0068, -0.1200], + [ 0.0350, 0.0528, -0.0810, ..., 0.0365, -0.1122, -0.0674]], + device='cuda:0'), grad: tensor([[ 2.9448e-06, -1.6876e-06, -2.0508e-06, ..., 6.1095e-07, + 3.7067e-06, 2.2911e-06], + [ 3.6396e-06, 4.6007e-06, 2.4475e-06, ..., -5.8077e-06, + -1.5208e-06, -1.4812e-05], + [ 1.4317e-04, 1.6010e-04, 1.0151e-06, ..., 1.1945e-04, + 8.3923e-05, 4.8615e-06], + ..., + [ 5.5619e-06, 5.9195e-06, 1.8310e-06, ..., 5.3681e-06, + 4.6380e-06, 1.0906e-06], + [-1.2852e-07, 5.0198e-07, 1.7118e-06, ..., 3.6694e-06, + 4.9826e-07, 1.7006e-06], + [ 1.0058e-05, 1.2040e-05, 1.1796e-04, ..., 2.0206e-05, + 8.8871e-05, 1.7518e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0268, -0.0280, -0.0192, -0.0011, -0.0020, 0.0031, 0.0032, -0.0135, + 0.0215, -0.0349], device='cuda:0'), grad: tensor([ 1.6928e-05, 2.8157e-04, 7.5960e-04, -8.3685e-04, -5.4264e-04, + 4.7922e-05, 2.8744e-05, -3.0303e-04, 1.7658e-05, 5.2977e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 218.35, cls_loss 0.0019 cls_loss_mapping 0.0067 cls_loss_causal 0.5883 re_mapping 0.0062 re_causal 0.0203 /// teacc 99.08 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0185, -0.0107, 0.0911, ..., 0.0072, -0.0533, -0.0378], + [-0.0910, -0.0876, -0.1132, ..., 0.0386, 0.0501, 0.2378], + [-0.0601, -0.0705, -0.0725, ..., -0.0838, 0.0216, -0.1344], + ..., + [-0.1217, -0.1220, 0.0252, ..., -0.1489, -0.0700, -0.2300], + [ 0.0775, 0.0339, -0.0657, ..., -0.1149, 0.0084, -0.1201], + [ 0.0363, 0.0532, -0.0813, ..., 0.0379, -0.1127, -0.0687]], + device='cuda:0'), grad: tensor([[ 7.2992e-08, -2.5937e-07, 9.2783e-08, ..., 1.1874e-08, + 2.9523e-07, -1.6880e-08], + [ 5.8440e-08, 1.0512e-07, 1.9767e-07, ..., -3.4808e-08, + -1.1188e-07, -8.4005e-07], + [ 1.3690e-07, 2.0198e-07, 6.8778e-07, ..., 4.3889e-07, + 5.9698e-07, 2.0140e-07], + ..., + [ 2.3039e-07, 2.3842e-07, 2.3982e-08, ..., 3.1153e-07, + -1.2340e-07, 3.6927e-07], + [ 9.3970e-07, 1.1846e-06, 5.0850e-07, ..., 1.0906e-06, + 2.9895e-07, 5.7626e-08], + [-2.8182e-06, -2.7958e-06, 1.0291e-07, ..., -2.5705e-06, + 7.4506e-08, 6.6590e-08]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0260, -0.0285, -0.0188, -0.0012, -0.0023, 0.0028, 0.0016, -0.0139, + 0.0230, -0.0342], device='cuda:0'), grad: tensor([ 1.3057e-06, 1.5413e-06, 4.5598e-06, 3.5278e-06, 6.5565e-06, + 1.3011e-06, -6.8732e-06, -1.2137e-05, 4.2841e-06, -4.1127e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 218.67, cls_loss 0.0013 cls_loss_mapping 0.0048 cls_loss_causal 0.5145 re_mapping 0.0064 re_causal 0.0198 /// teacc 99.02 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0186, -0.0107, 0.0912, ..., 0.0070, -0.0537, -0.0380], + [-0.0923, -0.0888, -0.1139, ..., 0.0386, 0.0500, 0.2387], + [-0.0607, -0.0716, -0.0732, ..., -0.0848, 0.0214, -0.1349], + ..., + [-0.1221, -0.1227, 0.0247, ..., -0.1492, -0.0704, -0.2305], + [ 0.0787, 0.0356, -0.0658, ..., -0.1144, 0.0094, -0.1205], + [ 0.0361, 0.0518, -0.0818, ..., 0.0379, -0.1135, -0.0693]], + device='cuda:0'), grad: tensor([[ 1.9781e-06, 1.5870e-06, -3.3854e-07, ..., 3.0035e-07, + 1.2554e-06, 2.6077e-08], + [ 7.3109e-07, 6.5891e-07, 6.5099e-07, ..., 3.5809e-07, + 1.3579e-06, -5.8208e-07], + [ 7.3388e-06, 7.2606e-06, 3.2783e-07, ..., 7.5996e-07, + 1.2922e-07, 8.3703e-08], + ..., + [ 3.6340e-06, 2.4959e-06, 1.2899e-06, ..., 3.1497e-06, + 2.1383e-06, 8.9640e-08], + [-6.8881e-06, -8.8513e-06, 2.1455e-07, ..., 5.2750e-06, + -5.5917e-06, 1.7066e-07], + [-1.9327e-05, -9.8273e-06, 1.0496e-06, ..., -1.9222e-05, + 2.6282e-06, 4.1793e-08]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0262, -0.0286, -0.0191, -0.0014, -0.0021, 0.0030, 0.0019, -0.0139, + 0.0240, -0.0348], device='cuda:0'), grad: tensor([ 3.0026e-05, 4.9353e-05, -1.3208e-04, 4.9025e-05, 9.4771e-06, + 2.4632e-05, 6.8173e-06, 1.7643e-05, -2.4885e-05, -3.0220e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 218.49, cls_loss 0.0016 cls_loss_mapping 0.0056 cls_loss_causal 0.5525 re_mapping 0.0066 re_causal 0.0204 /// teacc 99.01 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0187, -0.0120, 0.0904, ..., 0.0052, -0.0539, -0.0380], + [-0.0932, -0.0899, -0.1140, ..., 0.0387, 0.0500, 0.2403], + [-0.0616, -0.0733, -0.0736, ..., -0.0856, 0.0216, -0.1354], + ..., + [-0.1226, -0.1240, 0.0245, ..., -0.1497, -0.0707, -0.2317], + [ 0.0779, 0.0340, -0.0666, ..., -0.1157, 0.0088, -0.1212], + [ 0.0371, 0.0535, -0.0807, ..., 0.0403, -0.1137, -0.0697]], + device='cuda:0'), grad: tensor([[-1.1042e-05, -3.5673e-05, -3.2663e-05, ..., -6.0886e-05, + 3.4459e-07, 1.8347e-07], + [ 1.2210e-06, 2.2445e-06, 1.7164e-06, ..., 1.8142e-06, + 3.4925e-07, -2.3842e-06], + [ 1.6158e-06, 2.6692e-06, 1.9390e-06, ..., 3.9339e-06, + 1.4063e-06, 4.9733e-07], + ..., + [ 2.0824e-06, 3.5353e-06, 4.0345e-06, ..., 5.4203e-06, + 2.8331e-06, 6.0676e-07], + [ 9.5144e-06, 1.1295e-05, 1.8729e-06, ..., 8.8066e-06, + 1.0021e-05, 2.2224e-07], + [ 5.8152e-06, 2.0385e-05, 1.9908e-05, ..., 3.3975e-05, + 1.3001e-06, 2.6450e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0277, -0.0287, -0.0191, -0.0008, -0.0024, 0.0035, 0.0013, -0.0139, + 0.0228, -0.0335], device='cuda:0'), grad: tensor([-1.4484e-04, 5.5023e-06, 1.1921e-05, -2.8729e-05, -1.6866e-06, + 5.1148e-06, 1.6943e-05, 1.8597e-05, 3.2276e-05, 8.4758e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 218.41, cls_loss 0.0018 cls_loss_mapping 0.0049 cls_loss_causal 0.5617 re_mapping 0.0067 re_causal 0.0208 /// teacc 99.01 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0183, -0.0114, 0.0908, ..., 0.0055, -0.0539, -0.0382], + [-0.0939, -0.0910, -0.1146, ..., 0.0391, 0.0500, 0.2412], + [-0.0621, -0.0743, -0.0744, ..., -0.0863, 0.0217, -0.1357], + ..., + [-0.1232, -0.1252, 0.0238, ..., -0.1502, -0.0710, -0.2324], + [ 0.0777, 0.0337, -0.0678, ..., -0.1166, 0.0086, -0.1217], + [ 0.0391, 0.0546, -0.0797, ..., 0.0407, -0.1150, -0.0705]], + device='cuda:0'), grad: tensor([[ 3.5996e-07, 5.3197e-06, -5.3644e-06, ..., -2.4308e-06, + 1.0066e-05, 2.7940e-09], + [ 3.5996e-07, 2.2184e-06, 1.7881e-06, ..., 7.2643e-07, + 1.2135e-06, -1.0955e-07], + [ 1.3243e-06, -1.5885e-05, 1.0943e-06, ..., -9.5069e-06, + -1.5825e-05, 1.3039e-08], + ..., + [ 7.6508e-07, 2.3454e-05, 6.4587e-07, ..., 5.2061e-07, + 8.0094e-07, 2.9919e-08], + [ 5.7835e-07, 3.2395e-05, 2.9001e-06, ..., 3.3136e-06, + 3.3472e-06, 2.0140e-08], + [ 7.7393e-07, 2.5500e-06, 2.8223e-05, ..., 3.5483e-06, + 1.1243e-05, 1.4552e-08]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0274, -0.0287, -0.0190, -0.0009, -0.0036, 0.0036, 0.0013, -0.0142, + 0.0220, -0.0322], device='cuda:0'), grad: tensor([ 4.2260e-05, 1.0610e-05, -9.0837e-05, 2.3887e-05, -8.7380e-05, + -1.3673e-04, 5.0247e-05, 5.3674e-05, 7.4983e-05, 5.9336e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 218.30, cls_loss 0.0017 cls_loss_mapping 0.0046 cls_loss_causal 0.5455 re_mapping 0.0063 re_causal 0.0193 /// teacc 98.98 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0185, -0.0113, 0.0913, ..., 0.0058, -0.0539, -0.0380], + [-0.0943, -0.0917, -0.1168, ..., 0.0400, 0.0497, 0.2422], + [-0.0628, -0.0755, -0.0756, ..., -0.0872, 0.0213, -0.1360], + ..., + [-0.1242, -0.1264, 0.0231, ..., -0.1517, -0.0716, -0.2336], + [ 0.0779, 0.0332, -0.0684, ..., -0.1171, 0.0088, -0.1213], + [ 0.0396, 0.0548, -0.0802, ..., 0.0412, -0.1157, -0.0712]], + device='cuda:0'), grad: tensor([[ 2.3730e-06, 5.9954e-08, 3.2425e-05, ..., 1.9390e-06, + 4.8503e-06, 4.8778e-08], + [ 8.7265e-07, 1.8661e-07, 1.1645e-05, ..., -6.7148e-07, + 1.3486e-06, -2.2780e-06], + [ 2.1718e-06, 2.1199e-07, 3.1888e-05, ..., 3.5251e-07, + 4.8876e-06, 4.8522e-07], + ..., + [ 5.3737e-07, 8.8941e-07, 2.9560e-06, ..., 7.0361e-07, + 9.6485e-07, 6.3283e-07], + [ 3.7961e-06, 1.9725e-06, 2.1815e-05, ..., 1.9725e-06, + 2.7940e-06, 1.5693e-07], + [-5.7779e-06, -2.0172e-06, 1.4797e-05, ..., -2.7791e-06, + 7.1190e-06, 2.5844e-07]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0270, -0.0280, -0.0195, -0.0005, -0.0030, 0.0036, 0.0013, -0.0155, + 0.0220, -0.0323], device='cuda:0'), grad: tensor([ 8.6248e-05, 2.8476e-05, 8.3864e-05, 8.6948e-06, 4.1747e-04, + 1.0476e-05, -7.3385e-04, 7.3612e-06, 6.3956e-05, 2.6435e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 218.52, cls_loss 0.0018 cls_loss_mapping 0.0052 cls_loss_causal 0.5506 re_mapping 0.0063 re_causal 0.0195 /// teacc 99.02 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0189, -0.0112, 0.0918, ..., 0.0064, -0.0533, -0.0371], + [-0.0951, -0.0928, -0.1188, ..., 0.0403, 0.0507, 0.2437], + [-0.0637, -0.0766, -0.0765, ..., -0.0897, 0.0200, -0.1376], + ..., + [-0.1249, -0.1275, 0.0229, ..., -0.1522, -0.0718, -0.2340], + [ 0.0785, 0.0335, -0.0689, ..., -0.1178, 0.0093, -0.1217], + [ 0.0394, 0.0537, -0.0805, ..., 0.0412, -0.1164, -0.0719]], + device='cuda:0'), grad: tensor([[ 1.0654e-06, -1.1906e-05, -1.8403e-05, ..., -2.3529e-05, + 2.2422e-07, 5.0059e-08], + [ 6.5565e-07, 9.7696e-07, 2.5565e-07, ..., 4.1956e-07, + -7.1479e-08, -9.9279e-07], + [ 8.1817e-07, 1.4519e-06, 4.6473e-07, ..., 1.4342e-06, + 2.8033e-07, 1.9372e-07], + ..., + [ 1.0217e-06, 1.2275e-06, 1.4598e-07, ..., 1.2275e-06, + 2.5099e-07, 2.8359e-07], + [ 5.7071e-06, 7.1637e-06, 1.3795e-07, ..., 6.2995e-06, + 1.9209e-08, 1.6857e-07], + [-1.7714e-06, 1.0759e-05, 2.4319e-05, ..., 1.9789e-05, + 8.7023e-06, 1.4959e-07]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0265, -0.0272, -0.0214, 0.0006, -0.0033, 0.0029, 0.0023, -0.0152, + 0.0225, -0.0329], device='cuda:0'), grad: tensor([-3.8326e-05, 2.4065e-06, 1.1958e-06, -1.4499e-05, -1.8910e-05, + -4.8615e-06, 4.2357e-06, 1.2284e-06, 1.5661e-05, 5.1796e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 218.25, cls_loss 0.0015 cls_loss_mapping 0.0061 cls_loss_causal 0.5462 re_mapping 0.0062 re_causal 0.0191 /// teacc 99.02 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0189, -0.0093, 0.0932, ..., 0.0079, -0.0536, -0.0372], + [-0.0955, -0.0937, -0.1193, ..., 0.0407, 0.0508, 0.2451], + [-0.0645, -0.0780, -0.0776, ..., -0.0908, 0.0199, -0.1378], + ..., + [-0.1254, -0.1285, 0.0227, ..., -0.1529, -0.0720, -0.2351], + [ 0.0787, 0.0335, -0.0693, ..., -0.1184, 0.0096, -0.1222], + [ 0.0397, 0.0531, -0.0818, ..., 0.0406, -0.1174, -0.0741]], + device='cuda:0'), grad: tensor([[ 3.2014e-08, 9.5461e-09, 5.4017e-08, ..., -4.2026e-08, + 3.3062e-07, 6.5309e-08], + [ 4.2492e-08, 1.4761e-07, 1.4913e-07, ..., -2.5379e-07, + -3.1549e-08, -6.9616e-07], + [ 1.1234e-07, -9.0618e-07, 3.7905e-07, ..., 6.4611e-08, + 1.7439e-07, 5.6112e-08], + ..., + [ 1.1176e-07, 2.8568e-07, 1.6494e-06, ..., 1.5332e-07, + 3.5334e-06, 1.4016e-07], + [-3.1386e-07, -7.0431e-08, 6.7055e-08, ..., 1.1385e-07, + 3.2014e-08, 9.9186e-08], + [-5.1642e-07, -1.6380e-07, 6.1274e-05, ..., -7.1991e-07, + 1.2529e-04, 9.4529e-08]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0252, -0.0271, -0.0213, 0.0006, -0.0030, 0.0032, 0.0021, -0.0155, + 0.0225, -0.0339], device='cuda:0'), grad: tensor([ 1.1912e-06, 4.0932e-07, -1.5935e-06, 3.8408e-06, -4.7445e-04, + -1.3411e-07, 8.6799e-07, 1.3575e-05, -6.7637e-08, 4.5633e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 218.19, cls_loss 0.0012 cls_loss_mapping 0.0053 cls_loss_causal 0.5370 re_mapping 0.0063 re_causal 0.0202 /// teacc 99.06 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0192, -0.0094, 0.0935, ..., 0.0078, -0.0538, -0.0377], + [-0.0968, -0.0959, -0.1207, ..., 0.0402, 0.0506, 0.2451], + [-0.0649, -0.0786, -0.0782, ..., -0.0906, 0.0202, -0.1379], + ..., + [-0.1256, -0.1289, 0.0228, ..., -0.1531, -0.0721, -0.2354], + [ 0.0788, 0.0335, -0.0698, ..., -0.1193, 0.0099, -0.1225], + [ 0.0401, 0.0535, -0.0819, ..., 0.0412, -0.1180, -0.0724]], + device='cuda:0'), grad: tensor([[-1.8729e-06, -4.5449e-06, -4.7050e-06, ..., -4.7199e-06, + 1.6149e-06, 1.6298e-08], + [ 1.1045e-06, 1.5534e-06, 3.8347e-07, ..., 6.2585e-07, + 1.8887e-06, -6.9989e-07], + [ 4.7963e-07, 1.2293e-06, 1.0757e-06, ..., 1.7062e-06, + 2.1495e-06, 1.0838e-07], + ..., + [ 2.1316e-07, 3.0897e-07, 1.7730e-07, ..., 2.1246e-07, + 4.8708e-07, 1.9930e-07], + [-2.1547e-05, -2.8968e-05, -2.5220e-06, ..., -1.1943e-05, + -4.4614e-05, 6.6939e-08], + [ 1.5041e-06, 2.8946e-06, 1.4445e-06, ..., 1.0096e-06, + 6.0871e-06, 5.9372e-08]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0253, -0.0275, -0.0210, 0.0010, -0.0031, 0.0030, 0.0024, -0.0154, + 0.0224, -0.0338], device='cuda:0'), grad: tensor([-7.9945e-06, 5.3160e-06, -1.1688e-06, 5.5194e-05, 7.3202e-06, + 8.0094e-06, 1.1496e-05, -6.7893e-07, -8.9586e-05, 1.1988e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 218.17, cls_loss 0.0014 cls_loss_mapping 0.0060 cls_loss_causal 0.5401 re_mapping 0.0065 re_causal 0.0199 /// teacc 99.02 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0194, -0.0092, 0.0936, ..., 0.0078, -0.0544, -0.0379], + [-0.0976, -0.0973, -0.1215, ..., 0.0405, 0.0500, 0.2461], + [-0.0652, -0.0795, -0.0786, ..., -0.0911, 0.0200, -0.1386], + ..., + [-0.1256, -0.1294, 0.0224, ..., -0.1536, -0.0715, -0.2353], + [ 0.0794, 0.0341, -0.0711, ..., -0.1202, 0.0091, -0.1228], + [ 0.0402, 0.0537, -0.0827, ..., 0.0413, -0.1193, -0.0732]], + device='cuda:0'), grad: tensor([[ 9.1968e-08, -8.5402e-07, -1.4920e-06, ..., -7.0315e-07, + 1.3399e-07, 1.9767e-07], + [ 3.0221e-07, 5.4203e-07, 5.0059e-08, ..., -1.4352e-06, + -1.8403e-06, -4.8764e-06], + [ 8.7079e-08, -7.2643e-08, 1.5914e-07, ..., 3.4575e-07, + -1.9942e-07, 4.0047e-07], + ..., + [ 4.5146e-07, 8.3214e-07, 7.9512e-08, ..., 6.5472e-07, + 1.7136e-07, 3.5367e-07], + [ 1.9986e-06, 3.4589e-06, 2.4494e-07, ..., 2.5015e-06, + 2.3807e-07, 3.9302e-07], + [-8.1778e-05, -1.3292e-04, 3.0827e-07, ..., -8.6427e-05, + 3.7020e-07, 8.8848e-07]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0254, -0.0283, -0.0211, 0.0003, -0.0023, 0.0031, 0.0035, -0.0147, + 0.0222, -0.0344], device='cuda:0'), grad: tensor([-9.5810e-08, 4.6864e-06, -3.4943e-06, 8.3372e-06, 1.5438e-05, + 2.6393e-04, 1.6913e-06, -1.4767e-05, 1.0595e-05, -2.8634e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 125---------------------------------------------------- +epoch 125, time 219.04, cls_loss 0.0013 cls_loss_mapping 0.0047 cls_loss_causal 0.5400 re_mapping 0.0062 re_causal 0.0196 /// teacc 99.12 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0195, -0.0090, 0.0937, ..., 0.0078, -0.0547, -0.0383], + [-0.0992, -0.0992, -0.1217, ..., 0.0399, 0.0501, 0.2469], + [-0.0656, -0.0801, -0.0792, ..., -0.0916, 0.0201, -0.1385], + ..., + [-0.1263, -0.1297, 0.0222, ..., -0.1538, -0.0717, -0.2358], + [ 0.0795, 0.0339, -0.0716, ..., -0.1209, 0.0091, -0.1240], + [ 0.0411, 0.0541, -0.0831, ..., 0.0422, -0.1196, -0.0726]], + device='cuda:0'), grad: tensor([[ 1.6275e-07, -2.2817e-06, -2.6058e-06, ..., -1.0245e-06, + 3.9442e-07, 8.5915e-07], + [ 3.0478e-07, -3.9861e-07, 1.4249e-07, ..., -4.7460e-06, + -4.4890e-06, -1.3597e-05], + [ 3.6228e-07, 2.0582e-06, 1.5330e-06, ..., 2.0359e-06, + 1.7118e-06, 3.8184e-06], + ..., + [ 4.1677e-07, -8.0885e-07, 2.6636e-07, ..., 1.3206e-06, + 5.3551e-07, 1.8775e-06], + [-1.0664e-07, 5.4110e-07, 3.9628e-07, ..., 1.0021e-06, + 3.6927e-07, 1.6829e-06], + [-1.1760e-04, 4.7311e-07, -5.3257e-05, ..., -1.2982e-04, + -2.7701e-05, 2.5667e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([-2.5533e-02, -2.9178e-02, -2.0428e-02, 5.9478e-06, -2.1568e-03, + 3.1594e-03, 3.7151e-03, -1.4561e-02, 2.1754e-02, -3.3942e-02], + device='cuda:0'), grad: tensor([-1.0887e-06, -1.7494e-05, -6.5775e-08, 1.7956e-05, 5.2404e-04, + -4.3097e-07, 3.1460e-06, -1.8641e-05, 8.1211e-06, -5.1594e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 126---------------------------------------------------- +epoch 126, time 219.12, cls_loss 0.0017 cls_loss_mapping 0.0042 cls_loss_causal 0.5461 re_mapping 0.0064 re_causal 0.0190 /// teacc 99.15 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0196, -0.0092, 0.0932, ..., 0.0075, -0.0551, -0.0388], + [-0.0996, -0.1000, -0.1218, ..., 0.0400, 0.0502, 0.2483], + [-0.0660, -0.0794, -0.0798, ..., -0.0920, 0.0205, -0.1384], + ..., + [-0.1264, -0.1298, 0.0244, ..., -0.1540, -0.0722, -0.2364], + [ 0.0795, 0.0333, -0.0721, ..., -0.1216, 0.0087, -0.1249], + [ 0.0416, 0.0543, -0.0833, ..., 0.0429, -0.1200, -0.0737]], + device='cuda:0'), grad: tensor([[ 4.6827e-06, -1.7518e-06, -5.6401e-06, ..., 8.1137e-06, + 5.0152e-07, 7.2760e-08], + [ 8.7544e-08, 1.6147e-07, 2.8871e-07, ..., -2.6170e-06, + -5.3905e-06, -5.9940e-06], + [ 2.1583e-07, 3.3900e-07, 4.7032e-07, ..., 2.1718e-06, + -6.2846e-06, 3.3174e-06], + ..., + [ 1.9278e-07, 2.7008e-07, 5.5414e-08, ..., 8.4890e-07, + 1.3793e-06, 9.0664e-07], + [-3.7835e-08, 3.3760e-07, 5.7705e-06, ..., 3.8259e-06, + 4.1462e-06, 5.4156e-07], + [-8.1360e-06, -7.9945e-06, -7.3314e-06, ..., -2.4274e-05, + 4.4890e-07, 3.0454e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0260, -0.0294, -0.0189, -0.0002, -0.0028, 0.0033, 0.0039, -0.0137, + 0.0206, -0.0340], device='cuda:0'), grad: tensor([ 1.3664e-05, -2.2367e-05, -2.0713e-05, 4.2915e-06, 2.6390e-05, + 8.5473e-05, -7.6115e-05, -2.3358e-06, 3.7462e-05, -4.5657e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 218.35, cls_loss 0.0012 cls_loss_mapping 0.0053 cls_loss_causal 0.5596 re_mapping 0.0064 re_causal 0.0195 /// teacc 99.06 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0198, -0.0083, 0.0942, ..., 0.0076, -0.0554, -0.0391], + [-0.0995, -0.1000, -0.1223, ..., 0.0406, 0.0503, 0.2496], + [-0.0665, -0.0793, -0.0804, ..., -0.0923, 0.0206, -0.1386], + ..., + [-0.1271, -0.1311, 0.0243, ..., -0.1546, -0.0725, -0.2372], + [ 0.0799, 0.0327, -0.0738, ..., -0.1224, 0.0087, -0.1255], + [ 0.0420, 0.0545, -0.0835, ..., 0.0432, -0.1204, -0.0742]], + device='cuda:0'), grad: tensor([[ 1.6287e-05, 1.1273e-05, -5.6950e-07, ..., -3.3970e-07, + 1.1951e-05, 1.0477e-08], + [ 1.7006e-06, 1.4491e-06, 1.2340e-07, ..., 9.2899e-08, + 9.6671e-07, -3.7299e-07], + [ 2.4382e-06, 2.0880e-06, 2.6589e-07, ..., 5.7416e-07, + 1.3271e-06, 3.6554e-08], + ..., + [ 1.0431e-05, 1.6198e-05, -1.3933e-06, ..., 4.5402e-07, + 6.9700e-06, 7.3109e-08], + [-5.6952e-05, -5.3674e-05, 2.5076e-07, ..., 4.7428e-07, + -4.1693e-05, 2.4214e-08], + [ 2.9411e-06, 2.8759e-06, 8.5402e-07, ..., -3.6950e-07, + 2.7064e-06, 7.3342e-08]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0257, -0.0290, -0.0187, -0.0001, -0.0029, 0.0034, 0.0038, -0.0138, + 0.0197, -0.0341], device='cuda:0'), grad: tensor([ 4.1872e-05, 4.6901e-06, 8.1286e-06, 4.6074e-05, 1.1176e-06, + 1.6823e-05, 5.0589e-06, 4.0859e-05, -1.7619e-04, 1.1578e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 217.96, cls_loss 0.0013 cls_loss_mapping 0.0037 cls_loss_causal 0.5197 re_mapping 0.0061 re_causal 0.0188 /// teacc 99.04 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0201, -0.0080, 0.0944, ..., 0.0077, -0.0557, -0.0393], + [-0.0999, -0.1004, -0.1228, ..., 0.0408, 0.0504, 0.2503], + [-0.0680, -0.0810, -0.0812, ..., -0.0932, 0.0205, -0.1390], + ..., + [-0.1269, -0.1307, 0.0240, ..., -0.1544, -0.0728, -0.2375], + [ 0.0804, 0.0332, -0.0742, ..., -0.1228, 0.0090, -0.1258], + [ 0.0423, 0.0544, -0.0839, ..., 0.0433, -0.1212, -0.0746]], + device='cuda:0'), grad: tensor([[ 1.1437e-06, 2.2314e-06, -9.4716e-07, ..., 1.0990e-07, + 1.8463e-07, 3.4925e-09], + [ 3.9041e-05, 1.5485e-04, 7.4739e-08, ..., 5.9634e-05, + 1.7555e-07, -5.9605e-08], + [ 3.9414e-06, 6.2585e-06, 9.2899e-08, ..., 6.7055e-07, + 9.5926e-08, 9.3132e-09], + ..., + [-1.3983e-04, -3.2282e-04, 1.0221e-07, ..., -8.1658e-05, + 1.8231e-07, 1.2107e-08], + [ 1.4831e-07, 2.8685e-06, 1.1409e-07, ..., 2.1476e-06, + -1.0822e-06, 4.1910e-09], + [ 5.7489e-05, 8.7619e-05, 6.5472e-07, ..., 6.3218e-06, + 8.3074e-07, 1.6298e-08]], device='cuda:0') +Epoch 130, bias, value: tensor([-2.5805e-02, -2.9070e-02, -1.9362e-02, -4.7532e-05, -2.6411e-03, + 3.0547e-03, 4.2037e-03, -1.3054e-02, 1.9995e-02, -3.4561e-02], + device='cuda:0'), grad: tensor([ 1.7449e-05, 6.8855e-04, 4.7475e-05, 3.8576e-04, 2.4900e-05, + -1.1120e-06, 6.5155e-06, -1.7347e-03, 2.1905e-05, 5.4312e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 218.40, cls_loss 0.0020 cls_loss_mapping 0.0049 cls_loss_causal 0.5606 re_mapping 0.0060 re_causal 0.0180 /// teacc 99.07 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0205, -0.0072, 0.0951, ..., 0.0080, -0.0560, -0.0395], + [-0.1006, -0.1014, -0.1231, ..., 0.0413, 0.0509, 0.2516], + [-0.0695, -0.0835, -0.0822, ..., -0.0939, 0.0204, -0.1393], + ..., + [-0.1283, -0.1305, 0.0234, ..., -0.1555, -0.0735, -0.2383], + [ 0.0806, 0.0332, -0.0745, ..., -0.1234, 0.0092, -0.1262], + [ 0.0440, 0.0544, -0.0842, ..., 0.0439, -0.1217, -0.0752]], + device='cuda:0'), grad: tensor([[ 1.3970e-07, -3.1330e-06, -3.4422e-06, ..., -1.3951e-06, + 3.7719e-07, 1.0235e-06], + [ 1.3574e-07, 2.0908e-07, -1.9465e-07, ..., -7.5437e-06, + -3.2447e-06, -1.3165e-05], + [ 1.8673e-07, -8.4518e-08, 1.3653e-06, ..., 1.0962e-06, + -4.4797e-07, 3.9418e-07], + ..., + [ 3.0827e-07, 3.5553e-07, 8.3866e-07, ..., 5.3737e-07, + 7.9814e-07, 4.9826e-07], + [ 3.5367e-07, 1.9055e-06, 5.0804e-07, ..., 2.4829e-06, + 1.3011e-06, 3.0063e-06], + [-5.6718e-07, 5.0338e-07, 1.5795e-06, ..., 8.2841e-07, + 9.8720e-07, 6.7148e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0256, -0.0288, -0.0204, 0.0011, -0.0027, 0.0025, 0.0045, -0.0138, + 0.0200, -0.0340], device='cuda:0'), grad: tensor([-5.0291e-06, -2.1189e-05, -1.6302e-05, 2.2119e-07, -1.5926e-06, + 7.7412e-06, 3.3081e-06, 3.6731e-06, 2.5123e-05, 4.0010e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 218.27, cls_loss 0.0017 cls_loss_mapping 0.0066 cls_loss_causal 0.5321 re_mapping 0.0063 re_causal 0.0183 /// teacc 99.06 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0207, -0.0070, 0.0954, ..., 0.0080, -0.0563, -0.0397], + [-0.1011, -0.1024, -0.1236, ..., 0.0415, 0.0509, 0.2526], + [-0.0700, -0.0842, -0.0828, ..., -0.0943, 0.0205, -0.1397], + ..., + [-0.1286, -0.1313, 0.0232, ..., -0.1558, -0.0739, -0.2388], + [ 0.0807, 0.0332, -0.0748, ..., -0.1242, 0.0094, -0.1268], + [ 0.0449, 0.0551, -0.0846, ..., 0.0444, -0.1231, -0.0759]], + device='cuda:0'), grad: tensor([[ 1.6205e-06, 7.5437e-06, 6.7195e-07, ..., 3.6769e-06, + 9.4436e-07, 5.5367e-07], + [ 7.1200e-07, -1.4398e-06, 1.6810e-07, ..., -1.1653e-05, + -7.9125e-06, -1.1079e-05], + [ 8.9640e-07, 1.5572e-06, 3.5297e-07, ..., 1.6484e-06, + -4.9770e-06, 1.5311e-06], + ..., + [ 2.1290e-06, 2.6114e-06, 2.1653e-07, ..., 3.8221e-06, + 1.6689e-06, 1.0878e-06], + [ 2.0131e-05, 1.9863e-05, 7.4059e-06, ..., 1.0170e-05, + 8.0001e-07, 7.7300e-07], + [-7.6517e-06, 2.6155e-04, -4.9621e-06, ..., 7.0512e-05, + -2.4587e-06, 2.6673e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0257, -0.0290, -0.0194, 0.0006, -0.0026, 0.0034, 0.0041, -0.0145, + 0.0200, -0.0341], device='cuda:0'), grad: tensor([ 1.8179e-05, -2.0131e-05, -1.8552e-05, 2.8551e-05, 1.1760e-04, + -5.2261e-04, -4.3511e-05, 6.3255e-06, 6.4790e-05, 3.6979e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 218.39, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5189 re_mapping 0.0058 re_causal 0.0168 /// teacc 98.96 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0214, -0.0079, 0.0949, ..., 0.0077, -0.0568, -0.0398], + [-0.1016, -0.1030, -0.1242, ..., 0.0421, 0.0516, 0.2539], + [-0.0705, -0.0854, -0.0835, ..., -0.0951, 0.0204, -0.1408], + ..., + [-0.1294, -0.1324, 0.0232, ..., -0.1568, -0.0749, -0.2392], + [ 0.0814, 0.0338, -0.0751, ..., -0.1245, 0.0098, -0.1274], + [ 0.0463, 0.0552, -0.0846, ..., 0.0454, -0.1245, -0.0767]], + device='cuda:0'), grad: tensor([[ 1.4370e-06, 5.7649e-07, -2.6189e-06, ..., -1.0096e-06, + 1.9185e-06, 1.3039e-08], + [ 2.1067e-06, 2.7381e-06, 3.8883e-07, ..., 8.5356e-07, + 1.9930e-06, -7.2224e-07], + [ 2.4363e-06, 2.1905e-06, 4.0233e-07, ..., 1.4566e-06, + 3.4235e-06, 6.6590e-08], + ..., + [ 4.5151e-06, 4.8615e-06, 7.1758e-07, ..., 2.9579e-06, + 3.2168e-06, 2.7940e-07], + [-4.6015e-05, -8.6486e-05, -2.5164e-06, ..., 1.5302e-06, + -8.0884e-05, 6.9849e-08], + [-2.5108e-05, -2.8744e-05, 2.0191e-06, ..., -2.2694e-05, + -3.4869e-06, 9.7789e-08]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0267, -0.0285, -0.0220, 0.0013, -0.0029, 0.0028, 0.0039, -0.0121, + 0.0203, -0.0338], device='cuda:0'), grad: tensor([ 1.9418e-07, 1.1094e-05, 1.0878e-05, 4.1313e-06, 4.8608e-05, + 9.8169e-05, 2.7132e-04, 1.9357e-05, -3.8075e-04, -8.2552e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 218.21, cls_loss 0.0016 cls_loss_mapping 0.0036 cls_loss_causal 0.5773 re_mapping 0.0063 re_causal 0.0182 /// teacc 98.99 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0212, -0.0075, 0.0951, ..., 0.0078, -0.0573, -0.0401], + [-0.1022, -0.1037, -0.1242, ..., 0.0427, 0.0525, 0.2551], + [-0.0709, -0.0855, -0.0843, ..., -0.0956, 0.0211, -0.1414], + ..., + [-0.1297, -0.1334, 0.0225, ..., -0.1580, -0.0761, -0.2397], + [ 0.0820, 0.0340, -0.0754, ..., -0.1249, 0.0100, -0.1279], + [ 0.0466, 0.0554, -0.0853, ..., 0.0457, -0.1251, -0.0772]], + device='cuda:0'), grad: tensor([[ 1.0145e-04, 1.2314e-04, 7.2181e-05, ..., 2.7013e-04, + 2.1048e-07, 6.1933e-08], + [ 2.8359e-07, 3.2224e-07, 5.3598e-07, ..., 3.0687e-07, + 4.8429e-08, -1.1306e-06], + [ 3.1898e-07, 3.9255e-07, 4.3027e-07, ..., 8.1258e-07, + 1.7276e-07, 1.3271e-07], + ..., + [ 8.1072e-07, 8.2562e-07, 7.0874e-07, ..., 1.8077e-06, + 3.5809e-07, 2.1607e-07], + [ 6.7800e-07, 9.1875e-07, 4.0494e-06, ..., 2.2519e-06, + 4.7777e-07, 1.5507e-07], + [-1.0824e-04, -1.3101e-04, -7.1347e-05, ..., -2.8729e-04, + 5.3644e-06, 1.6205e-07]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0267, -0.0278, -0.0223, 0.0014, -0.0029, 0.0022, 0.0039, -0.0117, + 0.0202, -0.0341], device='cuda:0'), grad: tensor([ 6.1846e-04, 1.5516e-06, 3.8445e-06, 3.5912e-06, 7.8790e-07, + 4.1686e-06, -1.1124e-05, -2.3516e-07, 1.7077e-05, -6.3753e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 218.37, cls_loss 0.0018 cls_loss_mapping 0.0051 cls_loss_causal 0.5867 re_mapping 0.0060 re_causal 0.0187 /// teacc 99.00 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0217, -0.0074, 0.0953, ..., 0.0076, -0.0576, -0.0403], + [-0.1031, -0.1042, -0.1251, ..., 0.0431, 0.0528, 0.2575], + [-0.0714, -0.0860, -0.0848, ..., -0.0960, 0.0208, -0.1426], + ..., + [-0.1303, -0.1341, 0.0226, ..., -0.1587, -0.0766, -0.2410], + [ 0.0820, 0.0337, -0.0761, ..., -0.1256, 0.0102, -0.1286], + [ 0.0467, 0.0547, -0.0869, ..., 0.0457, -0.1266, -0.0783]], + device='cuda:0'), grad: tensor([[ 9.8161e-07, 5.2294e-07, -4.0419e-06, ..., -4.5868e-07, + 8.2795e-07, 2.1048e-07], + [ 2.6710e-06, 3.4980e-06, -1.5022e-06, ..., -3.0864e-06, + -2.7753e-07, -7.1079e-06], + [ 1.2837e-05, 1.6555e-05, 3.8370e-07, ..., 8.8811e-06, + 1.0282e-05, -3.8743e-07], + ..., + [ 2.8498e-06, 3.7719e-06, 3.6834e-07, ..., 2.4308e-06, + 2.2128e-06, 2.9244e-07], + [-3.7113e-07, -9.0571e-07, 8.2888e-07, ..., 2.0787e-06, + 1.0747e-06, 2.6673e-06], + [-1.2629e-06, -8.3400e-07, 3.7812e-07, ..., -2.5183e-06, + 6.4680e-07, 2.8918e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0271, -0.0268, -0.0222, 0.0011, -0.0009, 0.0029, 0.0034, -0.0124, + 0.0197, -0.0354], device='cuda:0'), grad: tensor([ 7.0333e-06, 2.1160e-05, 1.0544e-04, -2.2578e-04, 9.6560e-06, + 2.4900e-05, 9.3654e-06, 4.4376e-05, 1.7127e-06, 2.3451e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 218.33, cls_loss 0.0018 cls_loss_mapping 0.0043 cls_loss_causal 0.5328 re_mapping 0.0060 re_causal 0.0175 /// teacc 98.84 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0224, -0.0079, 0.0951, ..., 0.0072, -0.0586, -0.0405], + [-0.1040, -0.1050, -0.1260, ..., 0.0430, 0.0522, 0.2569], + [-0.0722, -0.0871, -0.0860, ..., -0.0971, 0.0213, -0.1409], + ..., + [-0.1306, -0.1348, 0.0218, ..., -0.1590, -0.0768, -0.2415], + [ 0.0822, 0.0335, -0.0761, ..., -0.1262, 0.0104, -0.1284], + [ 0.0471, 0.0545, -0.0881, ..., 0.0459, -0.1280, -0.0787]], + device='cuda:0'), grad: tensor([[ 8.4285e-08, -5.2759e-07, -4.5821e-07, ..., -3.3574e-07, + 7.4971e-07, 1.1688e-07], + [ 1.1874e-07, 1.0896e-07, 5.3085e-08, ..., -6.9290e-07, + -7.2122e-05, -4.2692e-06], + [ 3.5996e-07, 4.4052e-07, 7.8510e-07, ..., 5.1083e-07, + 2.5958e-05, 1.2573e-06], + ..., + [ 5.3318e-07, 3.6648e-07, 1.0571e-07, ..., 5.3365e-07, + 3.1531e-05, 9.1875e-07], + [-4.1584e-07, 8.2515e-07, 4.0885e-07, ..., 2.7195e-07, + 1.0282e-06, 3.3015e-07], + [-2.5472e-07, 5.1223e-08, 2.3330e-07, ..., -1.7695e-07, + 2.3246e-06, 3.3155e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0279, -0.0278, -0.0218, 0.0015, 0.0002, 0.0031, 0.0032, -0.0123, + 0.0197, -0.0361], device='cuda:0'), grad: tensor([ 1.6689e-06, -5.1975e-04, 1.8525e-04, 7.2658e-05, 8.9034e-06, + 3.9265e-06, -3.8333e-06, 2.2769e-04, 9.4101e-06, 1.4760e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 218.33, cls_loss 0.0022 cls_loss_mapping 0.0053 cls_loss_causal 0.5484 re_mapping 0.0060 re_causal 0.0172 /// teacc 99.02 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0225, -0.0074, 0.0957, ..., 0.0073, -0.0595, -0.0408], + [-0.1049, -0.1057, -0.1267, ..., 0.0434, 0.0520, 0.2585], + [-0.0729, -0.0874, -0.0871, ..., -0.0976, 0.0220, -0.1406], + ..., + [-0.1317, -0.1360, 0.0208, ..., -0.1601, -0.0764, -0.2420], + [ 0.0829, 0.0337, -0.0769, ..., -0.1268, 0.0102, -0.1290], + [ 0.0495, 0.0550, -0.0876, ..., 0.0473, -0.1287, -0.0795]], + device='cuda:0'), grad: tensor([[ 1.6764e-07, 1.0757e-07, 2.0489e-08, ..., 8.8476e-08, + 7.4506e-08, 3.2596e-09], + [ 1.4398e-06, 1.6904e-07, 2.0955e-08, ..., 1.0282e-06, + 9.4995e-08, -1.2433e-07], + [ 5.8766e-07, 4.7963e-07, 2.3749e-08, ..., 7.9628e-08, + 1.4948e-07, 3.2131e-08], + ..., + [-3.2745e-06, 2.7195e-07, 2.4214e-08, ..., -2.7232e-06, + 2.2911e-07, 1.2107e-08], + [-5.2452e-06, -3.3565e-06, 7.0781e-08, ..., 9.4064e-08, + -2.5891e-06, 1.3970e-08], + [ 3.0883e-06, 9.4390e-07, -1.5832e-08, ..., 6.0070e-07, + 1.0133e-06, 2.0023e-08]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0278, -0.0259, -0.0214, 0.0008, -0.0014, 0.0032, 0.0031, -0.0144, + 0.0196, -0.0345], device='cuda:0'), grad: tensor([ 1.2945e-06, 4.1276e-05, -9.9465e-07, 3.7141e-06, 4.5821e-06, + 3.1572e-07, 4.9314e-07, -1.1593e-04, -1.0490e-05, 7.5817e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 218.30, cls_loss 0.0014 cls_loss_mapping 0.0048 cls_loss_causal 0.5448 re_mapping 0.0061 re_causal 0.0186 /// teacc 98.95 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0226, -0.0072, 0.0957, ..., 0.0072, -0.0596, -0.0408], + [-0.1048, -0.1063, -0.1271, ..., 0.0439, 0.0522, 0.2596], + [-0.0737, -0.0887, -0.0877, ..., -0.0981, 0.0218, -0.1409], + ..., + [-0.1318, -0.1364, 0.0205, ..., -0.1603, -0.0768, -0.2424], + [ 0.0837, 0.0342, -0.0775, ..., -0.1274, 0.0107, -0.1297], + [ 0.0495, 0.0550, -0.0882, ..., 0.0474, -0.1296, -0.0805]], + device='cuda:0'), grad: tensor([[ 3.6508e-07, 3.7579e-07, 4.4657e-07, ..., 5.0850e-07, + 3.8277e-07, 2.3283e-09], + [ 5.6811e-07, 7.5903e-07, 3.7393e-07, ..., 5.2014e-07, + 1.3094e-06, -1.9604e-07], + [ 8.7265e-07, 1.1101e-06, 8.3027e-07, ..., 1.0673e-06, + -4.1164e-07, 1.7695e-08], + ..., + [ 6.6496e-07, 5.4296e-07, 1.5274e-07, ..., 5.7695e-07, + 4.7823e-07, 1.2107e-08], + [-3.6303e-06, -5.8375e-06, 6.5612e-07, ..., -1.0580e-06, + -9.0674e-06, 3.4925e-08], + [-2.4773e-06, -8.1537e-07, 1.2852e-07, ..., -2.3916e-06, + 1.3951e-06, 3.1665e-08]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0281, -0.0257, -0.0216, 0.0012, -0.0008, 0.0032, 0.0029, -0.0145, + 0.0201, -0.0350], device='cuda:0'), grad: tensor([ 3.7272e-06, 1.1377e-05, 4.6566e-06, 4.3549e-06, 7.5251e-06, + 1.4290e-05, -6.4597e-06, -1.9282e-05, -2.0564e-05, 2.8173e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 218.34, cls_loss 0.0014 cls_loss_mapping 0.0048 cls_loss_causal 0.5380 re_mapping 0.0060 re_causal 0.0179 /// teacc 99.09 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0228, -0.0070, 0.0959, ..., 0.0072, -0.0598, -0.0410], + [-0.1056, -0.1070, -0.1274, ..., 0.0441, 0.0523, 0.2609], + [-0.0745, -0.0903, -0.0886, ..., -0.0994, 0.0216, -0.1414], + ..., + [-0.1320, -0.1368, 0.0205, ..., -0.1605, -0.0766, -0.2429], + [ 0.0841, 0.0340, -0.0785, ..., -0.1282, 0.0108, -0.1307], + [ 0.0500, 0.0554, -0.0884, ..., 0.0480, -0.1301, -0.0807]], + device='cuda:0'), grad: tensor([[ 4.3306e-08, 4.4070e-06, 4.3921e-06, ..., 4.7721e-06, + 4.8801e-07, 3.5390e-08], + [ 6.9384e-08, 1.8626e-07, 1.0803e-07, ..., -3.3854e-07, + -2.5891e-07, -1.7351e-06], + [ 1.2293e-07, 1.9046e-07, 1.1129e-07, ..., 2.0396e-07, + 2.0396e-07, 2.3982e-07], + ..., + [ 7.1712e-08, 7.0781e-08, 2.6077e-08, ..., 1.8114e-07, + 1.3364e-07, 4.8941e-07], + [-5.4250e-07, 2.0489e-06, 2.5667e-06, ..., 2.8238e-06, + -4.7125e-07, 1.6065e-07], + [-1.8394e-07, 1.1642e-07, 1.9511e-07, ..., 8.6147e-08, + 3.8417e-07, 4.8243e-07]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0281, -0.0259, -0.0218, 0.0009, -0.0010, 0.0033, 0.0034, -0.0142, + 0.0197, -0.0349], device='cuda:0'), grad: tensor([ 1.5020e-05, -1.6540e-06, 1.0990e-06, 2.7940e-06, 1.6652e-06, + 4.2200e-05, -7.0155e-05, 6.1654e-07, 6.7204e-06, 1.6354e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 218.45, cls_loss 0.0019 cls_loss_mapping 0.0051 cls_loss_causal 0.5193 re_mapping 0.0058 re_causal 0.0174 /// teacc 98.98 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0230, -0.0067, 0.0964, ..., 0.0069, -0.0601, -0.0410], + [-0.1061, -0.1079, -0.1286, ..., 0.0441, 0.0525, 0.2615], + [-0.0756, -0.0909, -0.0904, ..., -0.1004, 0.0208, -0.1421], + ..., + [-0.1345, -0.1375, 0.0200, ..., -0.1625, -0.0761, -0.2428], + [ 0.0840, 0.0333, -0.0790, ..., -0.1291, 0.0108, -0.1311], + [ 0.0522, 0.0559, -0.0884, ..., 0.0497, -0.1303, -0.0811]], + device='cuda:0'), grad: tensor([[ 5.7369e-07, 1.5646e-07, 1.2204e-05, ..., 4.5784e-06, + 8.6147e-08, -1.3970e-08], + [ 1.7341e-06, 6.5332e-07, 2.4848e-06, ..., 1.1269e-06, + 5.9092e-07, -2.5239e-07], + [ 1.5181e-06, 3.1013e-07, 1.1437e-06, ..., 5.1921e-07, + 1.6391e-07, 2.4680e-08], + ..., + [-7.1943e-05, 7.6136e-07, 2.3143e-07, ..., 1.1567e-06, + 3.1758e-07, 4.0513e-08], + [-7.9628e-07, -1.2172e-06, 1.6153e-05, ..., 6.0871e-06, + -2.0787e-06, 4.6100e-08], + [ 6.2943e-05, -9.9093e-07, 1.3104e-06, ..., -3.7458e-06, + 1.7239e-06, 6.0070e-08]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0282, -0.0261, -0.0222, 0.0019, -0.0012, 0.0032, 0.0026, -0.0153, + 0.0193, -0.0327], device='cuda:0'), grad: tensor([ 3.8177e-05, 2.1681e-05, 1.3620e-05, 6.5088e-05, 1.1109e-05, + 4.1991e-05, -1.6677e-04, -7.6437e-04, 6.1333e-05, 6.7854e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 218.41, cls_loss 0.0016 cls_loss_mapping 0.0047 cls_loss_causal 0.5181 re_mapping 0.0057 re_causal 0.0173 /// teacc 99.03 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0234, -0.0065, 0.0974, ..., 0.0075, -0.0585, -0.0415], + [-0.1069, -0.1091, -0.1313, ..., 0.0433, 0.0509, 0.2616], + [-0.0763, -0.0910, -0.0921, ..., -0.1014, 0.0220, -0.1403], + ..., + [-0.1350, -0.1382, 0.0196, ..., -0.1633, -0.0757, -0.2437], + [ 0.0851, 0.0341, -0.0798, ..., -0.1296, 0.0108, -0.1322], + [ 0.0524, 0.0557, -0.0893, ..., 0.0504, -0.1309, -0.0826]], + device='cuda:0'), grad: tensor([[ 2.1495e-06, 1.6447e-06, 1.2238e-06, ..., 1.9260e-06, + 1.6717e-07, 2.0023e-08], + [ 9.4995e-08, 7.4040e-08, 9.1270e-08, ..., -3.3062e-08, + -3.2596e-09, -1.9232e-07], + [-6.9290e-06, -4.8913e-06, 1.3113e-06, ..., 1.3784e-07, + 7.4413e-07, 2.6543e-08], + ..., + [ 3.3760e-07, 2.8079e-07, 4.0606e-07, ..., 3.1292e-07, + 1.7136e-07, 4.8429e-08], + [ 5.8394e-07, 5.0012e-07, 3.7253e-07, ..., 6.4634e-07, + -1.9092e-08, 2.4680e-08], + [ 4.8429e-08, -2.3469e-07, -2.9225e-06, ..., -5.9381e-06, + 5.9605e-08, 3.8184e-08]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0276, -0.0270, -0.0216, 0.0013, -0.0011, 0.0039, 0.0029, -0.0150, + 0.0197, -0.0333], device='cuda:0'), grad: tensor([ 9.2164e-06, 8.2981e-07, -3.7134e-05, 8.3297e-06, -1.4622e-06, + -2.6077e-06, 2.6394e-06, 2.5947e-06, 2.8871e-06, 1.4745e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 218.22, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5323 re_mapping 0.0057 re_causal 0.0179 /// teacc 99.07 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0234, -0.0058, 0.0985, ..., 0.0080, -0.0585, -0.0421], + [-0.1073, -0.1094, -0.1322, ..., 0.0437, 0.0514, 0.2641], + [-0.0767, -0.0914, -0.0930, ..., -0.1022, 0.0214, -0.1425], + ..., + [-0.1353, -0.1387, 0.0193, ..., -0.1639, -0.0760, -0.2451], + [ 0.0854, 0.0337, -0.0814, ..., -0.1306, 0.0109, -0.1347], + [ 0.0526, 0.0555, -0.0895, ..., 0.0505, -0.1313, -0.0837]], + device='cuda:0'), grad: tensor([[ 4.7218e-07, -1.0626e-06, -1.1418e-06, ..., -9.1642e-07, + 1.2526e-07, 2.0489e-08], + [ 5.1875e-07, 1.3690e-07, 6.8452e-08, ..., 2.7148e-07, + 2.9150e-07, -1.9651e-07], + [ 1.0841e-06, 5.1828e-07, 8.9873e-08, ..., 8.2236e-07, + 2.1048e-07, 9.3132e-09], + ..., + [-4.3167e-07, 5.8953e-07, 1.3597e-07, ..., 3.8650e-08, + -1.3262e-06, 2.2352e-08], + [ 2.1141e-07, 1.0571e-07, 6.9849e-08, ..., 5.6811e-07, + -4.6566e-08, 1.6298e-08], + [ 4.0885e-07, 6.9849e-09, 2.6776e-07, ..., -1.1735e-07, + 1.0030e-06, 3.7253e-08]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0269, -0.0266, -0.0221, 0.0033, -0.0014, 0.0033, 0.0022, -0.0150, + 0.0188, -0.0334], device='cuda:0'), grad: tensor([-1.0021e-06, 7.7263e-06, 3.0417e-06, 5.9567e-06, 1.3243e-06, + 2.9579e-06, 1.4054e-06, -3.8177e-05, 1.4398e-06, 1.5303e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 218.45, cls_loss 0.0013 cls_loss_mapping 0.0040 cls_loss_causal 0.5376 re_mapping 0.0061 re_causal 0.0176 /// teacc 99.07 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0235, -0.0050, 0.0991, ..., 0.0083, -0.0588, -0.0424], + [-0.1089, -0.1096, -0.1323, ..., 0.0421, 0.0503, 0.2648], + [-0.0772, -0.0918, -0.0940, ..., -0.1028, 0.0212, -0.1432], + ..., + [-0.1352, -0.1395, 0.0220, ..., -0.1624, -0.0735, -0.2445], + [ 0.0859, 0.0339, -0.0818, ..., -0.1310, 0.0111, -0.1349], + [ 0.0529, 0.0557, -0.0899, ..., 0.0506, -0.1317, -0.0842]], + device='cuda:0'), grad: tensor([[ 9.2108e-07, 4.6985e-07, -5.1223e-08, ..., 6.5751e-07, + 1.6764e-08, 1.9092e-08], + [ 4.5728e-07, 2.1746e-07, -4.6566e-09, ..., 9.5461e-08, + -9.6392e-08, -5.8347e-07], + [ 2.4633e-07, 1.1502e-07, 3.0734e-08, ..., 1.6950e-07, + 5.9139e-08, 6.7521e-08], + ..., + [ 5.0142e-06, 2.7344e-06, 4.3306e-08, ..., 3.7290e-06, + 8.9407e-08, 5.6811e-08], + [ 1.9725e-06, 5.4436e-07, 1.7276e-07, ..., 1.2591e-06, + -1.9092e-07, 5.3551e-08], + [-5.3406e-05, -1.0528e-05, -4.1313e-06, ..., -2.6420e-05, + 9.6392e-08, 1.4063e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0266, -0.0277, -0.0223, 0.0028, -0.0022, 0.0033, 0.0023, -0.0136, + 0.0192, -0.0335], device='cuda:0'), grad: tensor([ 2.4829e-06, 6.7521e-07, -1.4137e-06, 8.6650e-06, 8.1062e-05, + 6.0908e-06, 5.3132e-07, 1.2778e-05, 4.7162e-06, -1.1569e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 218.57, cls_loss 0.0018 cls_loss_mapping 0.0046 cls_loss_causal 0.5203 re_mapping 0.0055 re_causal 0.0161 /// teacc 99.05 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0242, -0.0061, 0.0990, ..., 0.0077, -0.0588, -0.0430], + [-0.1114, -0.1127, -0.1327, ..., 0.0426, 0.0498, 0.2662], + [-0.0781, -0.0926, -0.0950, ..., -0.1035, 0.0211, -0.1435], + ..., + [-0.1353, -0.1402, 0.0220, ..., -0.1627, -0.0737, -0.2458], + [ 0.0878, 0.0361, -0.0821, ..., -0.1302, 0.0129, -0.1338], + [ 0.0530, 0.0562, -0.0903, ..., 0.0512, -0.1326, -0.0865]], + device='cuda:0'), grad: tensor([[ 1.3737e-07, -4.5868e-07, -7.2131e-07, ..., -5.1130e-07, + 6.5193e-08, 1.3039e-08], + [ 1.2293e-07, 2.4820e-07, 3.3062e-08, ..., -2.2817e-08, + -4.8429e-08, -2.9616e-07], + [ 1.0207e-06, 8.4378e-07, 1.1874e-07, ..., 1.5274e-07, + 5.8301e-07, 5.1223e-08], + ..., + [ 1.1995e-06, 2.0508e-06, 2.9802e-08, ..., 7.9395e-07, + 7.2690e-07, 7.2643e-08], + [-1.6373e-06, 7.0315e-08, 9.4529e-08, ..., 1.4435e-07, + -9.2853e-07, 2.1420e-08], + [-2.5332e-06, -1.5385e-06, 2.4214e-07, ..., -2.0657e-06, + 9.3598e-08, 2.2817e-08]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0275, -0.0280, -0.0223, 0.0022, -0.0020, 0.0031, 0.0026, -0.0132, + 0.0213, -0.0341], device='cuda:0'), grad: tensor([-5.2247e-07, 9.1596e-07, -8.9929e-06, -1.1083e-06, 2.3935e-06, + -1.1642e-06, 2.3544e-06, 7.9423e-06, 4.4554e-06, -6.3106e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 218.50, cls_loss 0.0014 cls_loss_mapping 0.0044 cls_loss_causal 0.5492 re_mapping 0.0057 re_causal 0.0175 /// teacc 99.13 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0243, -0.0052, 0.0995, ..., 0.0080, -0.0591, -0.0432], + [-0.1117, -0.1130, -0.1332, ..., 0.0436, 0.0498, 0.2671], + [-0.0787, -0.0932, -0.0956, ..., -0.1040, 0.0212, -0.1437], + ..., + [-0.1357, -0.1413, 0.0220, ..., -0.1636, -0.0740, -0.2467], + [ 0.0879, 0.0358, -0.0833, ..., -0.1312, 0.0130, -0.1340], + [ 0.0534, 0.0563, -0.0907, ..., 0.0514, -0.1332, -0.0872]], + device='cuda:0'), grad: tensor([[ 9.0338e-08, -1.5959e-05, -2.1264e-05, ..., -2.1502e-05, + -4.4033e-06, 1.8580e-07], + [ 6.6590e-08, 2.7800e-07, 2.7753e-07, ..., -2.1746e-07, + -1.8013e-04, -1.0282e-04], + [ 1.1735e-07, 2.3544e-06, 3.0454e-06, ..., 3.2000e-06, + 1.5891e-04, 8.9467e-05], + ..., + [ 3.3062e-07, 3.7858e-07, 9.8255e-08, ..., 4.7637e-07, + 1.5453e-05, 8.8140e-06], + [ 4.6473e-07, 1.4566e-06, 1.1846e-06, ..., 1.6112e-06, + 5.5507e-07, 2.1514e-07], + [-8.4238e-07, -4.0047e-07, 5.7556e-07, ..., -4.4378e-07, + 1.1623e-06, 5.2853e-07]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0273, -0.0278, -0.0221, 0.0017, -0.0017, 0.0025, 0.0047, -0.0132, + 0.0208, -0.0346], device='cuda:0'), grad: tensor([-7.0691e-05, -4.3440e-04, 3.9053e-04, 6.6683e-06, 3.7644e-06, + 3.8482e-06, 5.3227e-05, 3.8534e-05, 6.1467e-06, 1.8748e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 218.09, cls_loss 0.0013 cls_loss_mapping 0.0028 cls_loss_causal 0.5235 re_mapping 0.0056 re_causal 0.0162 /// teacc 99.12 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0242, -0.0039, 0.1004, ..., 0.0087, -0.0593, -0.0435], + [-0.1119, -0.1130, -0.1330, ..., 0.0443, 0.0503, 0.2687], + [-0.0790, -0.0934, -0.0961, ..., -0.1046, 0.0214, -0.1441], + ..., + [-0.1354, -0.1415, 0.0218, ..., -0.1639, -0.0743, -0.2474], + [ 0.0881, 0.0358, -0.0839, ..., -0.1319, 0.0125, -0.1349], + [ 0.0533, 0.0562, -0.0911, ..., 0.0516, -0.1337, -0.0880]], + device='cuda:0'), grad: tensor([[ 2.7008e-08, -7.1293e-07, -8.4611e-07, ..., -6.7102e-07, + 5.9139e-08, 1.5832e-08], + [ 6.7521e-08, 5.8673e-08, 1.2154e-07, ..., 1.5832e-08, + 3.3528e-08, -2.3469e-07], + [ 1.4668e-07, 1.3504e-07, 2.1094e-07, ..., 2.5518e-07, + -1.6950e-07, 1.7229e-08], + ..., + [ 1.5814e-06, 1.1167e-06, 4.5169e-08, ..., 1.8626e-07, + 8.3819e-08, 4.7497e-08], + [-1.7602e-07, -2.6263e-07, 7.5297e-07, ..., 9.9093e-07, + -7.5437e-08, 4.4238e-08], + [-2.8219e-07, -4.7497e-08, 1.0524e-07, ..., -2.8405e-07, + 9.9652e-08, 3.0268e-08]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0267, -0.0276, -0.0218, 0.0012, -0.0014, 0.0026, 0.0038, -0.0128, + 0.0201, -0.0351], device='cuda:0'), grad: tensor([ 3.5912e-06, 8.2105e-06, -4.0382e-05, -3.0138e-06, 5.0031e-06, + 3.9965e-05, -3.8713e-05, 8.0764e-06, 5.6997e-06, 1.1563e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 217.39, cls_loss 0.0013 cls_loss_mapping 0.0047 cls_loss_causal 0.5464 re_mapping 0.0054 re_causal 0.0166 /// teacc 99.04 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0254, -0.0038, 0.1010, ..., 0.0092, -0.0602, -0.0437], + [-0.1124, -0.1134, -0.1330, ..., 0.0446, 0.0500, 0.2700], + [-0.0796, -0.0940, -0.0968, ..., -0.1052, 0.0218, -0.1443], + ..., + [-0.1357, -0.1422, 0.0216, ..., -0.1641, -0.0741, -0.2482], + [ 0.0881, 0.0355, -0.0850, ..., -0.1331, 0.0123, -0.1355], + [ 0.0536, 0.0561, -0.0913, ..., 0.0519, -0.1343, -0.0885]], + device='cuda:0'), grad: tensor([[ 7.4357e-06, -1.3374e-06, -1.6391e-05, ..., -1.0274e-05, + 1.1120e-06, 8.6147e-08], + [ 8.7358e-07, 1.3663e-06, -1.6205e-07, ..., -1.2713e-07, + -1.7397e-06, -4.6678e-06], + [ 1.8552e-06, 1.8496e-06, 4.2561e-07, ..., 8.8476e-07, + 1.4585e-06, 4.5681e-07], + ..., + [ 4.9733e-07, 4.4610e-07, 1.7695e-07, ..., 4.7451e-07, + 3.2410e-07, 1.8021e-07], + [-6.5923e-05, -6.4850e-05, -4.0308e-06, ..., -2.1592e-05, + -4.1097e-05, 9.8534e-07], + [ 1.9729e-05, 1.9729e-05, 1.7099e-06, ..., 5.9418e-06, + 1.3173e-05, 6.8918e-08]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0267, -0.0278, -0.0212, 0.0022, -0.0009, 0.0021, 0.0024, -0.0128, + 0.0197, -0.0352], device='cuda:0'), grad: tensor([-1.3307e-05, -9.5647e-07, 7.2420e-06, 5.7518e-06, 5.6848e-06, + 7.7128e-05, 3.4600e-05, -4.4964e-06, -1.6057e-04, 4.8727e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 217.66, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.5539 re_mapping 0.0054 re_causal 0.0168 /// teacc 99.09 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0259, -0.0036, 0.1014, ..., 0.0093, -0.0608, -0.0439], + [-0.1129, -0.1138, -0.1333, ..., 0.0446, 0.0505, 0.2709], + [-0.0802, -0.0946, -0.0984, ..., -0.1063, 0.0216, -0.1448], + ..., + [-0.1359, -0.1434, 0.0214, ..., -0.1643, -0.0745, -0.2486], + [ 0.0885, 0.0356, -0.0856, ..., -0.1337, 0.0123, -0.1358], + [ 0.0539, 0.0562, -0.0916, ..., 0.0524, -0.1351, -0.0889]], + device='cuda:0'), grad: tensor([[ 2.0117e-07, -1.5413e-07, -3.0827e-07, ..., 7.4971e-08, + 2.1281e-07, 7.8697e-08], + [ 6.3796e-08, 1.2247e-07, -2.1933e-07, ..., -7.2829e-07, + -1.7323e-06, -3.4068e-06], + [ 5.1083e-07, 1.1753e-06, 5.8115e-07, ..., 8.7917e-07, + -5.4389e-06, 1.9111e-06], + ..., + [ 2.3516e-07, 4.0280e-07, 7.5437e-08, ..., 2.4401e-07, + 5.3942e-06, 3.7067e-07], + [-8.3260e-07, -1.9837e-06, 2.0117e-07, ..., 3.5856e-07, + -1.2061e-07, 2.2538e-07], + [-6.4075e-07, 3.2485e-06, 2.9057e-07, ..., -2.5565e-07, + 1.8161e-07, 9.7789e-08]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0267, -0.0276, -0.0214, 0.0023, -0.0009, 0.0021, 0.0025, -0.0128, + 0.0196, -0.0352], device='cuda:0'), grad: tensor([ 9.6299e-07, -4.6752e-06, -9.0301e-06, 2.4326e-06, 3.7029e-06, + -3.7327e-06, -3.8929e-06, 1.3031e-05, -3.3416e-06, 4.5300e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 217.17, cls_loss 0.0012 cls_loss_mapping 0.0033 cls_loss_causal 0.5354 re_mapping 0.0055 re_causal 0.0160 /// teacc 98.93 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0260, -0.0032, 0.1017, ..., 0.0096, -0.0612, -0.0442], + [-0.1132, -0.1141, -0.1339, ..., 0.0448, 0.0507, 0.2716], + [-0.0801, -0.0931, -0.0999, ..., -0.1074, 0.0228, -0.1453], + ..., + [-0.1360, -0.1445, 0.0214, ..., -0.1645, -0.0751, -0.2491], + [ 0.0889, 0.0358, -0.0867, ..., -0.1342, 0.0126, -0.1358], + [ 0.0540, 0.0560, -0.0923, ..., 0.0525, -0.1362, -0.0893]], + device='cuda:0'), grad: tensor([[ 1.3923e-06, 1.7453e-06, -5.0757e-08, ..., 8.6147e-07, + 9.6206e-07, 1.4761e-07], + [ 2.4252e-06, 2.4047e-06, -2.2212e-07, ..., 4.7265e-07, + -4.3446e-07, -2.9318e-06], + [ 9.1195e-06, 9.4622e-06, 1.0896e-07, ..., 5.0701e-06, + 3.1888e-06, 5.9698e-07], + ..., + [ 2.7698e-06, 2.5257e-06, 3.4738e-07, ..., 2.6375e-06, + 2.1532e-06, 8.6986e-07], + [ 7.2941e-06, 8.3372e-06, 9.7323e-08, ..., 4.9435e-06, + 3.1069e-06, 3.9022e-07], + [-6.9737e-06, -3.5837e-06, -6.9197e-07, ..., -7.7710e-06, + -1.2089e-06, 1.7835e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0267, -0.0276, -0.0208, 0.0014, -0.0004, 0.0021, 0.0032, -0.0129, + 0.0194, -0.0357], device='cuda:0'), grad: tensor([ 7.9423e-06, 3.5726e-06, 1.2890e-05, -4.8548e-05, 1.4417e-05, + -1.8388e-05, 5.3793e-06, 1.5259e-05, 2.4155e-05, -1.6764e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 217.14, cls_loss 0.0013 cls_loss_mapping 0.0045 cls_loss_causal 0.5349 re_mapping 0.0054 re_causal 0.0160 /// teacc 98.98 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0265, -0.0031, 0.1021, ..., 0.0097, -0.0617, -0.0446], + [-0.1136, -0.1146, -0.1342, ..., 0.0450, 0.0507, 0.2722], + [-0.0816, -0.0945, -0.1009, ..., -0.1085, 0.0228, -0.1456], + ..., + [-0.1367, -0.1476, 0.0203, ..., -0.1652, -0.0754, -0.2494], + [ 0.0896, 0.0366, -0.0870, ..., -0.1348, 0.0133, -0.1361], + [ 0.0547, 0.0570, -0.0919, ..., 0.0532, -0.1368, -0.0895]], + device='cuda:0'), grad: tensor([[ 1.9837e-07, -2.9849e-07, -1.1502e-06, ..., -6.5053e-07, + 1.4901e-07, -6.6124e-08], + [ 2.1467e-07, 1.8114e-07, 1.7416e-07, ..., -1.0552e-06, + -9.9372e-07, -3.2224e-06], + [ 1.1222e-07, 2.0955e-07, 4.1956e-07, ..., 5.9325e-07, + 2.4633e-07, 6.2352e-07], + ..., + [ 1.6652e-06, 1.2536e-06, 1.2852e-07, ..., 1.9409e-06, + 3.6601e-07, 8.3353e-07], + [ 9.0804e-08, 1.1595e-07, 4.1071e-07, ..., 7.0594e-07, + 5.6345e-08, 3.6648e-07], + [-4.3213e-06, -3.2708e-06, 4.2990e-06, ..., -3.0007e-06, + 1.7453e-06, 5.8068e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0269, -0.0276, -0.0208, 0.0016, -0.0006, 0.0022, 0.0029, -0.0130, + 0.0198, -0.0355], device='cuda:0'), grad: tensor([-2.1625e-06, -4.5076e-06, 1.6047e-06, 2.4755e-06, -1.1519e-05, + 6.1803e-06, -2.7567e-06, 6.2846e-06, 2.0657e-06, 2.3097e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.34, cls_loss 0.0012 cls_loss_mapping 0.0045 cls_loss_causal 0.5261 re_mapping 0.0055 re_causal 0.0164 /// teacc 99.03 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0265, -0.0014, 0.1033, ..., 0.0102, -0.0619, -0.0448], + [-0.1140, -0.1149, -0.1348, ..., 0.0453, 0.0508, 0.2730], + [-0.0830, -0.0957, -0.1018, ..., -0.1096, 0.0226, -0.1458], + ..., + [-0.1369, -0.1484, 0.0205, ..., -0.1654, -0.0756, -0.2498], + [ 0.0896, 0.0364, -0.0880, ..., -0.1358, 0.0134, -0.1367], + [ 0.0549, 0.0572, -0.0924, ..., 0.0535, -0.1373, -0.0900]], + device='cuda:0'), grad: tensor([[ 4.0233e-07, 2.9616e-07, 4.9919e-07, ..., 2.1094e-07, + 1.2973e-06, 3.3993e-08], + [ 2.5379e-07, 3.8603e-07, 1.1483e-06, ..., 6.5193e-08, + 5.9716e-06, -7.3155e-07], + [ 8.8848e-07, -2.0582e-07, 4.6054e-07, ..., 4.1351e-07, + -2.9400e-05, 1.2061e-07], + ..., + [ 5.8254e-07, 2.8079e-07, 4.0699e-07, ..., 1.6810e-07, + 8.6194e-07, 9.3598e-08], + [ 2.7195e-06, 3.0044e-06, 1.4901e-07, ..., 2.0918e-06, + 3.7029e-06, 1.6298e-07], + [-5.2936e-06, -3.2522e-06, -1.9139e-07, ..., -3.6117e-06, + 6.1607e-07, 5.6345e-08]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0260, -0.0277, -0.0206, 0.0018, -0.0006, 0.0012, 0.0040, -0.0130, + 0.0194, -0.0357], device='cuda:0'), grad: tensor([ 4.9956e-06, 2.7761e-05, -1.4067e-04, -2.5988e-05, 1.0423e-05, + 2.1815e-05, 8.7202e-05, 3.6266e-06, 1.7911e-05, -7.1861e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 217.07, cls_loss 0.0011 cls_loss_mapping 0.0033 cls_loss_causal 0.5081 re_mapping 0.0054 re_causal 0.0162 /// teacc 99.06 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0267, -0.0012, 0.1039, ..., 0.0102, -0.0625, -0.0452], + [-0.1144, -0.1153, -0.1336, ..., 0.0461, 0.0514, 0.2749], + [-0.0840, -0.0968, -0.1037, ..., -0.1108, 0.0224, -0.1467], + ..., + [-0.1371, -0.1490, 0.0203, ..., -0.1657, -0.0759, -0.2503], + [ 0.0902, 0.0367, -0.0901, ..., -0.1366, 0.0139, -0.1373], + [ 0.0553, 0.0577, -0.0927, ..., 0.0541, -0.1380, -0.0907]], + device='cuda:0'), grad: tensor([[ 8.1956e-08, -8.2236e-07, -9.2015e-07, ..., -6.8406e-07, + 1.2573e-08, 2.0722e-08], + [ 5.0990e-08, 2.0210e-07, 6.7754e-08, ..., -9.5693e-08, + -9.8022e-08, -6.4913e-07], + [ 3.9581e-08, 1.3341e-07, 1.2200e-07, ..., 1.3155e-07, + 2.6077e-08, 3.2131e-08], + ..., + [ 1.8883e-07, 2.0349e-07, 5.9605e-08, ..., 2.2259e-07, + 3.5157e-08, 9.6625e-08], + [ 1.4156e-07, 2.3958e-07, 2.0466e-07, ..., 3.4040e-07, + 1.1176e-08, 1.1572e-07], + [-5.9698e-07, 1.1548e-07, 5.5041e-07, ..., -1.7742e-07, + 1.7718e-07, 9.6392e-08]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0258, -0.0273, -0.0208, 0.0016, -0.0010, 0.0013, 0.0043, -0.0130, + 0.0191, -0.0357], device='cuda:0'), grad: tensor([-1.9036e-06, -3.7299e-07, 4.2934e-07, 3.0617e-07, 6.1467e-07, + 2.1085e-06, -2.9895e-06, 6.5193e-07, 9.7603e-07, 1.7090e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 217.49, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5292 re_mapping 0.0053 re_causal 0.0166 /// teacc 99.05 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0270, -0.0010, 0.1044, ..., 0.0105, -0.0631, -0.0454], + [-0.1120, -0.1130, -0.1344, ..., 0.0456, 0.0485, 0.2756], + [-0.0850, -0.0976, -0.1047, ..., -0.1119, 0.0225, -0.1470], + ..., + [-0.1384, -0.1520, 0.0202, ..., -0.1652, -0.0730, -0.2510], + [ 0.0914, 0.0373, -0.0908, ..., -0.1372, 0.0142, -0.1386], + [ 0.0548, 0.0567, -0.0938, ..., 0.0535, -0.1397, -0.0917]], + device='cuda:0'), grad: tensor([[ 1.5297e-07, 7.0082e-08, 3.1665e-08, ..., 8.7544e-08, + 4.3074e-08, 8.8476e-09], + [ 2.2049e-07, 2.0606e-07, 2.4447e-08, ..., 3.1199e-08, + -2.3982e-08, -3.4086e-07], + [ 1.7392e-07, 1.9092e-07, 6.2631e-08, ..., 5.8673e-08, + 1.1036e-07, 1.0128e-07], + ..., + [ 5.0478e-07, 4.1490e-07, 1.1874e-08, ..., 3.3062e-07, + 1.4808e-07, 1.1013e-07], + [ 5.7137e-07, 6.0350e-07, 1.3271e-07, ..., 1.9744e-07, + 4.0187e-07, 2.9337e-08], + [-2.2240e-06, -1.5879e-06, 9.4529e-08, ..., -1.7090e-06, + 6.3097e-08, 2.5379e-08]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0258, -0.0289, -0.0207, 0.0015, -0.0005, 0.0015, 0.0050, -0.0114, + 0.0194, -0.0370], device='cuda:0'), grad: tensor([ 1.0505e-06, 5.8562e-05, 3.0324e-06, 2.7604e-06, 3.2689e-06, + 3.9116e-06, -7.5623e-06, -7.0810e-05, 4.1015e-06, 1.7481e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 217.53, cls_loss 0.0011 cls_loss_mapping 0.0039 cls_loss_causal 0.5151 re_mapping 0.0053 re_causal 0.0165 /// teacc 99.02 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0273, -0.0012, 0.1045, ..., 0.0102, -0.0635, -0.0458], + [-0.1122, -0.1132, -0.1349, ..., 0.0459, 0.0486, 0.2764], + [-0.0855, -0.0981, -0.1055, ..., -0.1125, 0.0228, -0.1470], + ..., + [-0.1388, -0.1526, 0.0203, ..., -0.1655, -0.0731, -0.2515], + [ 0.0914, 0.0365, -0.0916, ..., -0.1384, 0.0141, -0.1394], + [ 0.0559, 0.0580, -0.0936, ..., 0.0544, -0.1401, -0.0923]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, 9.3132e-09, -7.0082e-08, ..., 7.1013e-08, + 1.6438e-07, 2.6776e-07], + [ 2.2724e-07, -1.7537e-06, -1.0505e-06, ..., -4.9286e-06, + -2.9262e-06, -9.8944e-06], + [ 7.3807e-08, 1.3178e-07, 1.2200e-07, ..., 1.7718e-07, + 3.8813e-07, 4.5123e-07], + ..., + [ 4.1872e-06, 4.0270e-06, 6.6869e-07, ..., 5.8301e-06, + 1.6987e-06, 5.7034e-06], + [-2.7614e-07, -4.6566e-09, 3.9022e-07, ..., 9.7882e-07, + 7.4273e-08, 1.1101e-06], + [-5.2564e-06, -3.0026e-06, 1.5390e-07, ..., -2.9411e-06, + 4.2352e-07, 1.0757e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0261, -0.0288, -0.0204, 0.0007, -0.0010, 0.0025, 0.0050, -0.0115, + 0.0187, -0.0365], device='cuda:0'), grad: tensor([ 8.0001e-07, -2.5213e-05, -6.6776e-07, 2.4531e-06, 3.9265e-06, + -9.7696e-07, -5.3551e-07, 2.9236e-05, 2.5406e-06, -1.1571e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 217.26, cls_loss 0.0013 cls_loss_mapping 0.0048 cls_loss_causal 0.5365 re_mapping 0.0054 re_causal 0.0162 /// teacc 98.95 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0282, -0.0014, 0.1044, ..., 0.0098, -0.0639, -0.0461], + [-0.1124, -0.1133, -0.1350, ..., 0.0467, 0.0490, 0.2773], + [-0.0866, -0.0991, -0.1068, ..., -0.1135, 0.0227, -0.1472], + ..., + [-0.1397, -0.1531, 0.0202, ..., -0.1661, -0.0735, -0.2520], + [ 0.0913, 0.0355, -0.0920, ..., -0.1408, 0.0140, -0.1402], + [ 0.0576, 0.0601, -0.0934, ..., 0.0557, -0.1407, -0.0927]], + device='cuda:0'), grad: tensor([[ 1.3737e-07, -7.9162e-09, -2.3004e-07, ..., -1.6647e-07, + 5.3830e-07, 9.0804e-09], + [ 3.1620e-05, 7.9628e-08, 1.0803e-07, ..., -9.4995e-08, + 5.4911e-06, -6.3609e-07], + [ 6.6124e-07, 6.1607e-07, 7.0082e-08, ..., 9.0804e-08, + -3.0249e-06, 4.7358e-07], + ..., + [-4.3273e-05, 4.0345e-06, 1.8976e-07, ..., 3.1898e-08, + -5.9791e-06, 5.2154e-08], + [ 3.6135e-06, -5.9977e-06, 2.9942e-07, ..., 3.9814e-08, + 2.4475e-06, 3.8417e-08], + [ 5.7518e-06, 1.3439e-06, 3.4235e-06, ..., 1.5576e-07, + 2.9225e-06, 1.4203e-08]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0265, -0.0286, -0.0206, 0.0026, -0.0012, 0.0007, 0.0050, -0.0119, + 0.0193, -0.0360], device='cuda:0'), grad: tensor([ 3.1833e-06, 1.7524e-04, -1.4298e-05, 1.3366e-05, -9.0972e-06, + -1.6112e-06, 4.0010e-06, -2.2757e-04, 2.3589e-05, 3.2932e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 217.34, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.5169 re_mapping 0.0054 re_causal 0.0158 /// teacc 99.04 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0288, -0.0017, 0.1044, ..., 0.0098, -0.0639, -0.0462], + [-0.1129, -0.1135, -0.1355, ..., 0.0466, 0.0489, 0.2785], + [-0.0896, -0.1016, -0.1082, ..., -0.1144, 0.0214, -0.1476], + ..., + [-0.1398, -0.1534, 0.0201, ..., -0.1663, -0.0735, -0.2523], + [ 0.0911, 0.0354, -0.0926, ..., -0.1417, 0.0139, -0.1413], + [ 0.0582, 0.0605, -0.0938, ..., 0.0563, -0.1416, -0.0930]], + device='cuda:0'), grad: tensor([[ 2.7427e-07, 1.1828e-07, 1.2228e-06, ..., 3.0035e-07, + 1.6904e-07, 4.7730e-08], + [ 1.2722e-06, 1.3616e-06, 1.3341e-07, ..., 8.8476e-07, + 1.4296e-06, -1.0505e-06], + [ 1.1045e-06, 1.2303e-06, 3.4156e-07, ..., 9.7882e-07, + 1.6745e-06, 1.1944e-07], + ..., + [ 2.0973e-06, 1.3346e-06, -3.0119e-06, ..., 1.7593e-06, + 9.2108e-07, 2.9523e-07], + [ 2.4657e-07, 1.6298e-09, 9.3598e-08, ..., 5.5972e-07, + -9.9884e-08, 8.9174e-08], + [-5.1893e-06, -2.6226e-06, 1.4631e-06, ..., -3.9861e-06, + 8.4192e-07, 8.1258e-08]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0268, -0.0287, -0.0211, 0.0029, -0.0012, 0.0011, 0.0049, -0.0118, + 0.0190, -0.0360], device='cuda:0'), grad: tensor([ 7.1488e-06, 5.0701e-06, 6.7465e-06, -1.1049e-05, 4.0010e-06, + 3.7570e-06, 1.0477e-08, -1.1064e-05, 4.1351e-07, -5.0440e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 217.35, cls_loss 0.0013 cls_loss_mapping 0.0039 cls_loss_causal 0.5047 re_mapping 0.0052 re_causal 0.0154 /// teacc 99.08 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0285, -0.0017, 0.1043, ..., 0.0099, -0.0639, -0.0464], + [-0.1137, -0.1140, -0.1360, ..., 0.0465, 0.0488, 0.2792], + [-0.0908, -0.1029, -0.1092, ..., -0.1164, 0.0208, -0.1478], + ..., + [-0.1403, -0.1539, 0.0202, ..., -0.1665, -0.0734, -0.2531], + [ 0.0914, 0.0356, -0.0930, ..., -0.1422, 0.0138, -0.1416], + [ 0.0595, 0.0614, -0.0941, ..., 0.0576, -0.1421, -0.0937]], + device='cuda:0'), grad: tensor([[ 1.7136e-07, -7.4971e-08, -1.9092e-07, ..., -9.2899e-08, + 2.3074e-07, 2.8173e-08], + [ 2.4098e-07, 2.6124e-07, 2.5146e-08, ..., -2.7614e-07, + -8.3353e-08, -9.4809e-07], + [ 7.1898e-07, 9.5740e-07, 8.4285e-08, ..., 4.4121e-07, + -2.9318e-06, 6.8685e-08], + ..., + [ 8.2422e-07, 8.7405e-07, 2.1886e-08, ..., 5.6997e-07, + 1.3970e-06, 2.6543e-07], + [ 5.2433e-07, 5.6904e-07, 2.8405e-08, ..., 3.4203e-07, + 1.2526e-06, 6.1700e-08], + [-1.1781e-06, -8.0327e-08, 4.0792e-07, ..., -1.2107e-06, + 4.2119e-07, 2.0792e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0266, -0.0289, -0.0215, 0.0029, -0.0018, 0.0015, 0.0051, -0.0117, + 0.0189, -0.0355], device='cuda:0'), grad: tensor([ 1.2191e-06, 2.9262e-06, -1.5512e-05, -3.7290e-06, 2.5760e-06, + 4.0419e-06, 4.8429e-07, -7.1339e-07, 8.4937e-06, 1.7812e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 217.17, cls_loss 0.0012 cls_loss_mapping 0.0042 cls_loss_causal 0.5141 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.04 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0291, -0.0017, 0.1046, ..., 0.0098, -0.0642, -0.0466], + [-0.1146, -0.1148, -0.1366, ..., 0.0465, 0.0490, 0.2804], + [-0.0919, -0.1044, -0.1099, ..., -0.1187, 0.0204, -0.1481], + ..., + [-0.1399, -0.1544, 0.0205, ..., -0.1664, -0.0736, -0.2545], + [ 0.0915, 0.0355, -0.0935, ..., -0.1430, 0.0142, -0.1418], + [ 0.0602, 0.0620, -0.0939, ..., 0.0583, -0.1426, -0.0947]], + device='cuda:0'), grad: tensor([[ 3.9348e-08, 1.1874e-08, 4.6566e-09, ..., 5.8208e-08, + 3.9558e-07, 1.6997e-08], + [ 4.4936e-08, 4.4703e-08, 2.7707e-08, ..., -1.7323e-07, + 6.9151e-08, -4.8708e-07], + [ 5.8906e-08, 6.7521e-08, 2.3283e-08, ..., 6.4028e-08, + -2.9076e-06, 6.2631e-08], + ..., + [ 1.3947e-07, 1.2782e-07, 7.6834e-09, ..., 1.5600e-07, + 2.0172e-06, 1.1805e-07], + [ 2.6589e-07, 2.3935e-07, 2.9802e-08, ..., 2.4005e-07, + 3.2969e-07, 5.3318e-08], + [-4.6287e-07, -4.0443e-07, 7.8464e-08, ..., -3.9767e-07, + 1.3225e-07, 7.9395e-08]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0269, -0.0289, -0.0218, 0.0031, -0.0029, 0.0021, 0.0048, -0.0115, + 0.0187, -0.0357], device='cuda:0'), grad: tensor([ 4.8243e-06, 1.0535e-05, -3.1888e-05, 1.1697e-06, 1.8086e-06, + 5.0105e-06, -1.2969e-07, -3.6717e-07, 3.8445e-06, 5.1446e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 217.19, cls_loss 0.0011 cls_loss_mapping 0.0048 cls_loss_causal 0.5373 re_mapping 0.0051 re_causal 0.0161 /// teacc 99.04 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0294, -0.0016, 0.1047, ..., 0.0098, -0.0644, -0.0469], + [-0.1149, -0.1151, -0.1371, ..., 0.0466, 0.0476, 0.2821], + [-0.0922, -0.1048, -0.1105, ..., -0.1194, 0.0227, -0.1487], + ..., + [-0.1404, -0.1548, 0.0211, ..., -0.1665, -0.0737, -0.2555], + [ 0.0914, 0.0351, -0.0938, ..., -0.1435, 0.0146, -0.1422], + [ 0.0607, 0.0623, -0.0949, ..., 0.0585, -0.1435, -0.0953]], + device='cuda:0'), grad: tensor([[ 9.8487e-08, -3.0734e-07, -1.1344e-06, ..., -6.5146e-07, + 1.8906e-07, 1.8231e-07], + [ 9.3132e-08, 1.8906e-07, 1.0431e-07, ..., -1.1530e-06, + -8.3912e-07, -2.2519e-06], + [ 1.8370e-07, 2.7474e-07, 2.7218e-07, ..., 2.2934e-07, + 9.3132e-09, 3.0431e-07], + ..., + [ 1.3178e-07, 2.3120e-07, -1.1073e-06, ..., 3.7136e-07, + 3.5740e-07, 4.8522e-07], + [-7.3295e-07, 1.3807e-07, 3.9651e-07, ..., 2.3539e-07, + -6.4075e-07, 2.5658e-07], + [-1.1036e-07, 3.9139e-07, 6.2026e-07, ..., 3.7160e-07, + 2.6310e-07, 3.6554e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0271, -0.0298, -0.0199, 0.0030, -0.0028, 0.0024, 0.0046, -0.0114, + 0.0186, -0.0360], device='cuda:0'), grad: tensor([-1.2172e-06, -3.4459e-06, 2.3991e-06, 7.1106e-07, 8.3074e-06, + -3.8296e-06, 2.6301e-06, -6.6385e-06, -1.6466e-06, 2.7437e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 216.96, cls_loss 0.0017 cls_loss_mapping 0.0038 cls_loss_causal 0.5520 re_mapping 0.0050 re_causal 0.0157 /// teacc 98.96 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0298, -0.0015, 0.1052, ..., 0.0098, -0.0648, -0.0471], + [-0.1159, -0.1163, -0.1374, ..., 0.0472, 0.0481, 0.2842], + [-0.0925, -0.1052, -0.1119, ..., -0.1202, 0.0228, -0.1491], + ..., + [-0.1408, -0.1559, 0.0202, ..., -0.1680, -0.0743, -0.2575], + [ 0.0913, 0.0348, -0.0942, ..., -0.1440, 0.0148, -0.1425], + [ 0.0621, 0.0639, -0.0963, ..., 0.0609, -0.1444, -0.0959]], + device='cuda:0'), grad: tensor([[ 2.2119e-08, -3.1441e-06, -8.1956e-06, ..., -8.8215e-06, + 1.4133e-07, 1.1618e-07], + [ 3.9814e-08, 1.8650e-07, 2.3260e-07, ..., -2.3432e-06, + -2.6971e-06, -6.4597e-06], + [ 1.8138e-07, 3.0454e-07, 2.2282e-07, ..., 3.3970e-07, + 4.6100e-08, 1.9511e-07], + ..., + [ 9.0105e-08, 4.5728e-07, 2.8452e-07, ..., 3.0361e-07, + 2.2864e-07, 7.4971e-08], + [ 9.6159e-08, 1.3448e-06, 1.1805e-07, ..., 2.2948e-06, + 2.5574e-06, 5.5358e-06], + [ 2.1122e-06, 6.4969e-05, 3.9637e-05, ..., 3.2514e-05, + 2.2739e-05, 6.8918e-08]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0272, -0.0300, -0.0197, 0.0026, -0.0022, 0.0024, 0.0049, -0.0112, + 0.0184, -0.0359], device='cuda:0'), grad: tensor([-1.8343e-05, -7.2978e-06, 1.3714e-07, 5.6863e-05, -2.0421e-04, + -5.7459e-05, 1.7047e-05, 1.5432e-06, 9.6634e-06, 2.0170e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 217.10, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5430 re_mapping 0.0053 re_causal 0.0164 /// teacc 99.03 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0300, -0.0013, 0.1056, ..., 0.0098, -0.0650, -0.0476], + [-0.1162, -0.1165, -0.1393, ..., 0.0469, 0.0480, 0.2835], + [-0.0927, -0.1054, -0.1136, ..., -0.1209, 0.0227, -0.1494], + ..., + [-0.1410, -0.1562, 0.0202, ..., -0.1683, -0.0744, -0.2578], + [ 0.0918, 0.0351, -0.0937, ..., -0.1436, 0.0152, -0.1433], + [ 0.0625, 0.0638, -0.0974, ..., 0.0610, -0.1457, -0.0966]], + device='cuda:0'), grad: tensor([[ 2.7707e-08, -5.9372e-08, -2.8033e-07, ..., -2.1700e-07, + 3.5623e-08, 2.5611e-09], + [ 4.8429e-08, 4.9360e-08, 7.9628e-08, ..., -6.0303e-08, + -1.6298e-07, -3.6997e-07], + [ 5.4017e-08, 6.7288e-08, 1.4529e-07, ..., 6.7288e-08, + 6.1933e-08, 2.8173e-08], + ..., + [ 2.5518e-07, 2.2654e-07, -1.5460e-07, ..., 2.5658e-07, + 2.8173e-07, 1.0035e-07], + [ 1.0058e-07, 3.0268e-08, -1.3970e-09, ..., 1.6904e-07, + 1.7462e-08, 4.2841e-08], + [-1.0314e-07, 8.0792e-08, 1.4426e-06, ..., -3.0966e-08, + 9.2620e-07, 1.3248e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0272, -0.0303, -0.0196, 0.0016, -0.0016, 0.0032, 0.0052, -0.0112, + 0.0188, -0.0363], device='cuda:0'), grad: tensor([-4.1141e-07, -1.5693e-07, 1.4026e-06, 8.6240e-07, -4.3102e-06, + 3.7812e-07, 1.9819e-06, -3.0678e-06, 3.9674e-07, 2.9318e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 217.33, cls_loss 0.0014 cls_loss_mapping 0.0049 cls_loss_causal 0.5002 re_mapping 0.0052 re_causal 0.0157 /// teacc 99.03 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0299, -0.0009, 0.1065, ..., 0.0100, -0.0649, -0.0477], + [-0.1164, -0.1166, -0.1400, ..., 0.0471, 0.0481, 0.2843], + [-0.0929, -0.1058, -0.1150, ..., -0.1222, 0.0226, -0.1498], + ..., + [-0.1411, -0.1566, 0.0197, ..., -0.1685, -0.0747, -0.2585], + [ 0.0920, 0.0370, -0.0947, ..., -0.1445, 0.0153, -0.1445], + [ 0.0644, 0.0651, -0.0967, ..., 0.0630, -0.1473, -0.0973]], + device='cuda:0'), grad: tensor([[ 1.6065e-08, 3.7951e-08, 1.1176e-08, ..., 2.7474e-08, + 3.3528e-08, 1.1874e-08], + [ 4.7497e-08, 5.5181e-08, 1.2107e-08, ..., -3.8184e-08, + -6.3563e-08, -2.2189e-07], + [ 2.1909e-07, 2.0931e-07, 4.8894e-08, ..., 2.8498e-07, + 2.7637e-07, 5.7742e-08], + ..., + [ 8.0792e-08, 1.2410e-07, 2.4913e-08, ..., 1.2014e-07, + 1.7253e-07, 7.2177e-08], + [ 3.8417e-08, 6.4494e-08, 6.7521e-09, ..., 5.9372e-08, + 6.2399e-08, 2.2585e-08], + [-4.4936e-08, 3.6089e-08, 3.5623e-08, ..., -2.3283e-08, + 9.2899e-08, 1.8394e-08]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0267, -0.0303, -0.0199, 0.0026, -0.0025, 0.0012, 0.0051, -0.0110, + 0.0202, -0.0353], device='cuda:0'), grad: tensor([ 1.7183e-07, 4.7730e-08, 6.9896e-07, -7.4506e-08, -1.9488e-07, + -1.4426e-06, 2.8871e-08, 7.8231e-08, 3.6927e-07, 3.1572e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 217.32, cls_loss 0.0011 cls_loss_mapping 0.0033 cls_loss_causal 0.5166 re_mapping 0.0054 re_causal 0.0160 /// teacc 99.07 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0307, -0.0010, 0.1066, ..., 0.0100, -0.0652, -0.0480], + [-0.1166, -0.1169, -0.1402, ..., 0.0471, 0.0480, 0.2852], + [-0.0933, -0.1073, -0.1162, ..., -0.1232, 0.0229, -0.1501], + ..., + [-0.1412, -0.1569, 0.0195, ..., -0.1688, -0.0747, -0.2597], + [ 0.0930, 0.0377, -0.0946, ..., -0.1445, 0.0160, -0.1446], + [ 0.0646, 0.0652, -0.0970, ..., 0.0635, -0.1479, -0.0979]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, -5.6345e-08, -5.6345e-08, ..., 6.0629e-07, + 4.5658e-07, 8.7917e-07], + [ 8.3586e-08, 8.4518e-08, 1.5367e-08, ..., -3.0175e-06, + -2.4177e-06, -5.0738e-06], + [ 2.8405e-08, 4.8894e-08, 4.0280e-08, ..., 3.2340e-07, + 2.1746e-07, 6.1188e-07], + ..., + [ 2.5076e-07, 2.4145e-07, 9.0804e-09, ..., 4.2887e-07, + 2.8638e-07, 4.6613e-07], + [-1.3504e-08, -8.6147e-09, 2.7008e-08, ..., 1.5190e-06, + 9.9558e-07, 2.0992e-06], + [-8.0001e-07, -6.5472e-07, 5.5647e-08, ..., -5.8068e-07, + -3.2363e-08, 1.6810e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0271, -0.0304, -0.0195, 0.0007, -0.0024, 0.0025, 0.0045, -0.0108, + 0.0209, -0.0355], device='cuda:0'), grad: tensor([ 1.6708e-06, -8.3968e-06, -5.6531e-07, 5.2340e-07, 3.1926e-06, + -9.3132e-10, -9.0990e-07, 2.0824e-06, 4.0606e-06, -1.6652e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 217.22, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.5361 re_mapping 0.0050 re_causal 0.0155 /// teacc 99.06 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0301, -0.0006, 0.1068, ..., 0.0101, -0.0657, -0.0483], + [-0.1167, -0.1170, -0.1405, ..., 0.0473, 0.0482, 0.2866], + [-0.0936, -0.1078, -0.1175, ..., -0.1241, 0.0228, -0.1504], + ..., + [-0.1414, -0.1573, 0.0199, ..., -0.1690, -0.0748, -0.2611], + [ 0.0936, 0.0380, -0.0952, ..., -0.1452, 0.0161, -0.1453], + [ 0.0647, 0.0657, -0.0983, ..., 0.0635, -0.1496, -0.0987]], + device='cuda:0'), grad: tensor([[ 7.0035e-07, 6.8778e-07, 1.7434e-06, ..., 1.6168e-06, + 2.2119e-08, 1.8626e-09], + [ 3.2596e-09, 5.1223e-09, 8.2655e-08, ..., -4.2375e-08, + 1.7183e-07, -1.4575e-07], + [ 1.8626e-09, 9.7789e-09, 9.4762e-08, ..., 4.5169e-08, + -2.3586e-07, 6.7521e-09], + ..., + [ 1.9558e-08, 1.4203e-08, 3.8184e-08, ..., 3.5157e-08, + -1.5716e-07, 4.3772e-08], + [ 1.4203e-08, 1.1176e-08, 3.9581e-08, ..., 5.5181e-08, + 9.6159e-08, 2.3283e-08], + [-3.1199e-08, -7.9162e-09, 1.2526e-07, ..., -1.0710e-08, + 1.3364e-07, 2.5146e-08]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0271, -0.0304, -0.0191, 0.0007, -0.0013, 0.0025, 0.0048, -0.0110, + 0.0210, -0.0365], device='cuda:0'), grad: tensor([ 2.9486e-06, 1.5292e-06, -1.4622e-06, 1.3337e-06, -1.8477e-06, + 2.9821e-06, -4.4741e-06, -2.7642e-06, 9.1083e-07, 8.4564e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 217.32, cls_loss 0.0011 cls_loss_mapping 0.0038 cls_loss_causal 0.5410 re_mapping 0.0053 re_causal 0.0162 /// teacc 99.00 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0301, -0.0007, 0.1069, ..., 0.0102, -0.0660, -0.0488], + [-0.1169, -0.1171, -0.1411, ..., 0.0476, 0.0487, 0.2890], + [-0.0939, -0.1084, -0.1180, ..., -0.1247, 0.0224, -0.1526], + ..., + [-0.1416, -0.1577, 0.0200, ..., -0.1693, -0.0750, -0.2624], + [ 0.0932, 0.0376, -0.0959, ..., -0.1463, 0.0160, -0.1461], + [ 0.0649, 0.0660, -0.0990, ..., 0.0636, -0.1505, -0.0996]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -4.4773e-07, -5.3551e-07, ..., -4.5169e-07, + -7.4273e-08, 4.0047e-08], + [ 2.1188e-08, 5.3085e-08, 2.2515e-07, ..., -1.0338e-06, + -1.2405e-06, -9.0059e-07], + [ 5.8208e-08, 1.1898e-07, 5.6718e-07, ..., 3.2620e-07, + 4.6240e-07, 1.1572e-07], + ..., + [ 3.9116e-08, 8.7079e-08, -1.5043e-05, ..., 3.3132e-07, + 4.1095e-07, 2.0838e-07], + [ 2.0256e-08, 6.9849e-07, 5.9046e-07, ..., 3.9884e-07, + 1.5809e-07, 1.6089e-07], + [-2.6915e-07, -1.5227e-07, 1.1399e-06, ..., -9.3831e-08, + 1.1665e-07, 9.2434e-08]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0274, -0.0301, -0.0195, 0.0008, -0.0006, 0.0023, 0.0054, -0.0110, + 0.0205, -0.0371], device='cuda:0'), grad: tensor([-5.6159e-07, -3.5353e-06, 3.9823e-06, 2.4252e-06, 4.2170e-05, + 3.4422e-05, 7.7579e-07, -9.0957e-05, 4.5672e-06, 6.6683e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 217.29, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5108 re_mapping 0.0050 re_causal 0.0152 /// teacc 99.06 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0306, -0.0008, 0.1069, ..., 0.0101, -0.0667, -0.0492], + [-0.1175, -0.1177, -0.1417, ..., 0.0479, 0.0490, 0.2906], + [-0.0940, -0.1087, -0.1186, ..., -0.1262, 0.0221, -0.1531], + ..., + [-0.1419, -0.1573, 0.0209, ..., -0.1693, -0.0752, -0.2635], + [ 0.0929, 0.0377, -0.0968, ..., -0.1476, 0.0159, -0.1480], + [ 0.0660, 0.0667, -0.0994, ..., 0.0643, -0.1509, -0.1005]], + device='cuda:0'), grad: tensor([[ 9.5461e-07, 8.1072e-07, -2.1188e-08, ..., 9.5228e-07, + 2.0396e-07, 2.3749e-08], + [ 9.4809e-07, 7.3202e-07, 5.7928e-07, ..., 1.8626e-08, + 6.8452e-07, -3.1781e-07], + [ 4.2701e-07, 4.1747e-07, 9.8255e-08, ..., 3.6694e-07, + -6.1188e-07, 1.9860e-07], + ..., + [ 2.7586e-06, 2.2985e-06, 1.2596e-07, ..., 1.4259e-06, + 4.9826e-07, 4.2305e-07], + [-1.3746e-06, 1.1711e-07, 4.7497e-08, ..., 3.6578e-07, + 3.2037e-07, -1.2154e-06], + [-3.7681e-06, -1.1511e-06, 3.8650e-07, ..., -7.0222e-06, + 1.0831e-06, 6.3283e-07]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0278, -0.0301, -0.0195, 0.0007, -0.0010, 0.0025, 0.0049, -0.0106, + 0.0201, -0.0371], device='cuda:0'), grad: tensor([ 4.9546e-06, 7.3835e-06, -2.9076e-06, -5.2936e-06, 5.2638e-06, + -2.1644e-06, 1.2154e-06, 6.1840e-06, -3.6433e-06, -1.1057e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 217.21, cls_loss 0.0010 cls_loss_mapping 0.0029 cls_loss_causal 0.5349 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0300, 0.0004, 0.1073, ..., 0.0105, -0.0670, -0.0491], + [-0.1179, -0.1180, -0.1420, ..., 0.0483, 0.0495, 0.2916], + [-0.0943, -0.1092, -0.1193, ..., -0.1280, 0.0216, -0.1536], + ..., + [-0.1426, -0.1580, 0.0210, ..., -0.1698, -0.0753, -0.2646], + [ 0.0930, 0.0375, -0.0972, ..., -0.1482, 0.0163, -0.1481], + [ 0.0664, 0.0667, -0.0996, ..., 0.0648, -0.1512, -0.1011]], + device='cuda:0'), grad: tensor([[ 4.4261e-07, 9.7789e-08, 4.6566e-08, ..., 2.6845e-07, + 8.5449e-08, 6.0536e-08], + [ 5.7556e-07, 6.6683e-07, 1.3644e-07, ..., -1.3388e-07, + 3.2187e-06, 1.4855e-06], + [ 6.6357e-08, 1.1339e-07, 2.1840e-07, ..., 2.8382e-07, + 8.0140e-07, 6.0489e-07], + ..., + [ 4.2468e-06, 8.7591e-07, 6.6496e-07, ..., 2.6915e-06, + 1.6717e-07, 1.6927e-07], + [ 8.4192e-07, -4.4750e-07, 1.6927e-07, ..., 6.5984e-07, + -4.4145e-06, -2.7250e-06], + [-1.2621e-05, -2.3153e-06, -1.7956e-06, ..., -7.7933e-06, + 1.8766e-07, 5.8673e-08]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0272, -0.0298, -0.0196, 0.0009, -0.0012, 0.0027, 0.0048, -0.0112, + 0.0200, -0.0370], device='cuda:0'), grad: tensor([ 1.3020e-06, 1.8030e-05, 3.4738e-06, 8.5542e-07, 9.9540e-06, + 3.7067e-06, 7.6275e-07, 1.0282e-05, -2.0534e-05, -2.7820e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 217.44, cls_loss 0.0010 cls_loss_mapping 0.0035 cls_loss_causal 0.4965 re_mapping 0.0050 re_causal 0.0151 /// teacc 99.01 lr 0.00010000 +Epoch 168, weight, value: tensor([[-3.0801e-02, 1.3499e-04, 1.0699e-01, ..., 1.0014e-02, + -6.7260e-02, -4.9806e-02], + [-1.1822e-01, -1.1830e-01, -1.4235e-01, ..., 4.8629e-02, + 4.9542e-02, 2.9262e-01], + [-9.4628e-02, -1.0964e-01, -1.1994e-01, ..., -1.2914e-01, + 2.1490e-02, -1.5402e-01], + ..., + [-1.4281e-01, -1.5838e-01, 2.0631e-02, ..., -1.7008e-01, + -7.5419e-02, -2.6556e-01], + [ 9.3824e-02, 3.8050e-02, -9.7013e-02, ..., -1.4806e-01, + 1.7467e-02, -1.4819e-01], + [ 6.6645e-02, 6.6955e-02, -1.0032e-01, ..., 6.5119e-02, + -1.5165e-01, -1.0206e-01]], device='cuda:0'), grad: tensor([[ 1.1129e-07, 1.3411e-07, 1.0850e-07, ..., 8.5682e-08, + 3.8650e-07, 5.8673e-08], + [ 1.3318e-07, -2.4214e-08, 7.7300e-08, ..., -3.3062e-07, + 6.8918e-08, -1.2731e-06], + [ 4.6100e-08, -5.6811e-08, 3.4459e-08, ..., 1.3551e-07, + -1.7993e-06, 1.9697e-07], + ..., + [ 2.9290e-07, 3.9907e-07, 1.5507e-07, ..., 3.2177e-07, + 2.6915e-07, 3.8417e-07], + [ 8.7731e-07, 1.1846e-06, 6.9384e-08, ..., 6.3190e-07, + 6.8452e-07, 2.0768e-07], + [ 2.6058e-06, 3.0827e-06, 1.7248e-06, ..., 1.5618e-06, + 1.1800e-06, 2.2165e-07]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0281, -0.0298, -0.0195, 0.0008, -0.0007, 0.0028, 0.0041, -0.0112, + 0.0210, -0.0372], device='cuda:0'), grad: tensor([ 2.4736e-06, 9.4529e-08, -9.4771e-06, -7.5363e-06, -1.1697e-05, + 3.7979e-06, 4.6566e-06, 1.3383e-06, 5.4277e-06, 1.0893e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 217.52, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.5187 re_mapping 0.0054 re_causal 0.0154 /// teacc 99.04 lr 0.00010000 +Epoch 169, weight, value: tensor([[-3.1158e-02, 1.6366e-04, 1.0749e-01, ..., 1.0017e-02, + -6.7811e-02, -5.0016e-02], + [-1.1861e-01, -1.1818e-01, -1.4260e-01, ..., 4.9119e-02, + 5.0074e-02, 2.9500e-01], + [-9.4975e-02, -1.1021e-01, -1.2065e-01, ..., -1.3008e-01, + 2.1012e-02, -1.5457e-01], + ..., + [-1.4308e-01, -1.5753e-01, 2.1615e-02, ..., -1.6993e-01, + -7.5690e-02, -2.6698e-01], + [ 9.3649e-02, 3.7825e-02, -9.7406e-02, ..., -1.4897e-01, + 1.7187e-02, -1.4933e-01], + [ 6.7027e-02, 6.6550e-02, -1.0126e-01, ..., 6.5002e-02, + -1.5232e-01, -1.0406e-01]], device='cuda:0'), grad: tensor([[ 6.9849e-09, 5.3551e-08, 3.2410e-07, ..., 2.5611e-07, + 1.2061e-07, 1.3364e-07], + [ 2.2352e-08, -2.8145e-06, -3.6992e-06, ..., -2.0429e-05, + -1.1951e-05, -1.5825e-05], + [ 1.8161e-08, 4.0047e-08, 1.8673e-07, ..., 2.9104e-07, + 1.3784e-07, 1.9092e-07], + ..., + [ 5.1549e-07, 2.5034e-06, 3.2410e-06, ..., 1.7345e-05, + 9.6932e-06, 1.2688e-05], + [ 1.0710e-08, 3.5530e-07, 3.5241e-06, ..., 1.8366e-06, + 9.8255e-07, 8.8941e-08], + [-5.7369e-07, 2.5891e-07, 3.6927e-07, ..., 1.0841e-06, + 1.0282e-06, 1.3728e-06]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0280, -0.0295, -0.0200, 0.0010, -0.0006, 0.0026, 0.0041, -0.0108, + 0.0207, -0.0379], device='cuda:0'), grad: tensor([ 1.8971e-06, -1.0198e-04, 2.1867e-06, 2.3358e-06, 7.0371e-06, + -3.0501e-07, -1.2055e-05, 8.3029e-05, 1.0625e-05, 7.1451e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 217.22, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.4690 re_mapping 0.0052 re_causal 0.0143 /// teacc 99.09 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0320, 0.0005, 0.1082, ..., 0.0102, -0.0681, -0.0501], + [-0.1189, -0.1178, -0.1429, ..., 0.0503, 0.0503, 0.2987], + [-0.0955, -0.1089, -0.1212, ..., -0.1305, 0.0213, -0.1543], + ..., + [-0.1433, -0.1579, 0.0216, ..., -0.1703, -0.0758, -0.2708], + [ 0.0924, 0.0364, -0.0986, ..., -0.1533, 0.0164, -0.1519], + [ 0.0681, 0.0682, -0.1013, ..., 0.0659, -0.1555, -0.1055]], + device='cuda:0'), grad: tensor([[ 1.3178e-07, 3.5763e-07, -7.4785e-07, ..., -8.4285e-08, + 1.5134e-07, 9.0804e-08], + [ 1.5413e-07, 3.9861e-07, 3.3993e-08, ..., -2.8722e-06, + -2.4810e-06, -4.3400e-06], + [ 4.4797e-07, 7.3621e-07, 1.1548e-07, ..., 1.6838e-06, + 1.3309e-06, 1.6121e-06], + ..., + [ 7.4599e-07, 1.2815e-06, 5.9605e-08, ..., 2.0228e-06, + 1.4976e-06, 1.5441e-06], + [-1.7555e-07, 1.4203e-07, 1.4435e-08, ..., 4.1677e-07, + 2.1886e-08, 4.7544e-07], + [ 4.1304e-07, 3.5278e-06, 1.1921e-07, ..., 8.7637e-07, + 7.3528e-07, 3.8370e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([-2.8087e-02, -2.9013e-02, -1.8642e-02, -5.7026e-05, 7.7173e-04, + 2.6446e-03, 3.7380e-03, -1.0654e-02, 1.8685e-02, -3.9012e-02], + device='cuda:0'), grad: tensor([ 2.2389e-06, -1.5542e-05, -5.0701e-06, -1.3635e-05, 1.3653e-06, + -8.0233e-07, 1.5805e-06, 1.6272e-05, 2.9355e-06, 1.0528e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 216.94, cls_loss 0.0011 cls_loss_mapping 0.0035 cls_loss_causal 0.5261 re_mapping 0.0052 re_causal 0.0157 /// teacc 99.01 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0329, 0.0013, 0.1086, ..., 0.0104, -0.0688, -0.0502], + [-0.1193, -0.1184, -0.1434, ..., 0.0505, 0.0494, 0.2994], + [-0.0958, -0.1093, -0.1233, ..., -0.1318, 0.0226, -0.1548], + ..., + [-0.1439, -0.1585, 0.0215, ..., -0.1707, -0.0759, -0.2714], + [ 0.0923, 0.0360, -0.0993, ..., -0.1542, 0.0160, -0.1522], + [ 0.0684, 0.0681, -0.1018, ..., 0.0662, -0.1561, -0.1060]], + device='cuda:0'), grad: tensor([[ 1.8720e-07, 5.8906e-07, 1.4547e-06, ..., 9.2667e-08, + 5.1223e-09, 1.0245e-08], + [ 6.9477e-07, 8.9733e-07, 1.3271e-07, ..., 4.0000e-07, + -1.6205e-07, -5.6345e-07], + [ 5.8673e-08, 1.2247e-07, 6.1467e-08, ..., 4.6566e-08, + 3.6787e-08, 1.0151e-07], + ..., + [ 1.1940e-06, 1.0822e-06, -2.3693e-06, ..., 9.9558e-07, + 7.2643e-08, 2.0396e-07], + [ 2.5891e-07, 9.8813e-07, 1.1176e-08, ..., 9.9186e-08, + 1.9092e-08, 5.4482e-08], + [-3.4962e-06, -3.8464e-07, 5.4808e-07, ..., -3.4217e-06, + 2.6543e-08, 6.9849e-09]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0279, -0.0301, -0.0175, 0.0005, 0.0009, 0.0023, 0.0042, -0.0103, + 0.0183, -0.0393], device='cuda:0'), grad: tensor([ 1.3478e-05, 2.9337e-06, -7.5579e-05, 1.1384e-04, 1.4268e-06, + -4.9859e-05, 1.2694e-06, -8.3148e-06, 2.3842e-06, -1.6559e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 217.32, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5364 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.05 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0333, 0.0009, 0.1081, ..., 0.0098, -0.0691, -0.0503], + [-0.1200, -0.1187, -0.1438, ..., 0.0506, 0.0493, 0.3003], + [-0.0963, -0.1096, -0.1238, ..., -0.1326, 0.0227, -0.1551], + ..., + [-0.1449, -0.1592, 0.0212, ..., -0.1712, -0.0761, -0.2719], + [ 0.0931, 0.0362, -0.0995, ..., -0.1545, 0.0167, -0.1523], + [ 0.0690, 0.0674, -0.1024, ..., 0.0667, -0.1567, -0.1063]], + device='cuda:0'), grad: tensor([[ 2.9244e-07, 4.6287e-07, 2.7241e-07, ..., 2.3004e-07, + 1.5646e-07, 1.4435e-08], + [ 3.4226e-07, 7.6788e-07, 2.5751e-07, ..., 8.1956e-08, + 2.1420e-08, -4.4703e-07], + [ 6.5193e-08, 8.1118e-07, 8.1817e-07, ..., 2.6030e-07, + 5.6066e-07, 9.7323e-08], + ..., + [ 2.8703e-06, 4.7572e-06, 5.4482e-07, ..., 1.0077e-06, + 7.7346e-07, 1.3458e-07], + [ 6.8918e-07, 1.1753e-06, 3.1758e-07, ..., 3.9442e-07, + 6.0070e-08, 9.1270e-08], + [-1.2539e-05, -1.2800e-05, -1.2759e-07, ..., -2.4736e-06, + 1.4808e-07, 2.8405e-08]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0287, -0.0302, -0.0173, 0.0003, 0.0013, 0.0027, 0.0045, -0.0105, + 0.0186, -0.0395], device='cuda:0'), grad: tensor([ 3.6620e-06, 2.8685e-06, 9.7156e-06, -2.7597e-05, 1.9759e-05, + 2.4781e-05, -9.6709e-06, 7.7039e-06, 4.8093e-06, -3.6061e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 171---------------------------------------------------- +epoch 171, time 218.11, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.4955 re_mapping 0.0050 re_causal 0.0152 /// teacc 99.21 lr 0.00010000 +Epoch 173, weight, value: tensor([[-3.3891e-02, -1.8215e-05, 1.0613e-01, ..., 8.0183e-03, + -6.9624e-02, -5.0109e-02], + [-1.2083e-01, -1.1909e-01, -1.4411e-01, ..., 5.0624e-02, + 4.9432e-02, 3.0119e-01], + [-9.6735e-02, -1.1008e-01, -1.2465e-01, ..., -1.3337e-01, + 2.2600e-02, -1.5586e-01], + ..., + [-1.4582e-01, -1.5980e-01, 2.0835e-02, ..., -1.7159e-01, + -7.6243e-02, -2.7247e-01], + [ 9.3707e-02, 3.6453e-02, -9.9995e-02, ..., -1.5472e-01, + 1.7073e-02, -1.5259e-01], + [ 7.0433e-02, 6.7883e-02, -1.0297e-01, ..., 6.7227e-02, + -1.5724e-01, -1.0673e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-09, -2.1961e-06, -2.6934e-06, ..., -1.5646e-06, + 6.6590e-08, -2.3935e-07], + [ 2.1420e-08, 2.3562e-07, 1.3737e-07, ..., -9.9838e-07, + -2.0936e-06, -2.2277e-06], + [ 6.2864e-08, 6.4168e-07, 7.5391e-07, ..., 4.9779e-07, + 1.0245e-07, 2.4261e-07], + ..., + [ 2.1420e-08, 1.8720e-07, 3.4086e-07, ..., 3.4925e-07, + 5.3458e-07, 5.0385e-07], + [ 4.6100e-08, 3.9488e-07, 2.2491e-07, ..., 5.7416e-07, + 9.4855e-07, 1.0077e-06], + [-5.1688e-08, 6.1793e-07, 9.4436e-07, ..., 3.2969e-07, + 3.5996e-07, 1.9697e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0306, -0.0303, -0.0178, 0.0001, 0.0011, 0.0026, 0.0065, -0.0105, + 0.0187, -0.0391], device='cuda:0'), grad: tensor([-7.8380e-06, -7.4506e-09, -4.5836e-05, 5.0366e-06, 8.1072e-07, + 7.4552e-07, 1.6559e-06, 2.5421e-05, 6.3106e-06, 1.3754e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 217.17, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.5017 re_mapping 0.0048 re_causal 0.0145 /// teacc 99.05 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0339, 0.0012, 0.1071, ..., 0.0087, -0.0697, -0.0485], + [-0.1211, -0.1194, -0.1446, ..., 0.0507, 0.0496, 0.3020], + [-0.0972, -0.1107, -0.1266, ..., -0.1342, 0.0224, -0.1563], + ..., + [-0.1463, -0.1605, 0.0213, ..., -0.1718, -0.0765, -0.2732], + [ 0.0941, 0.0367, -0.1005, ..., -0.1550, 0.0174, -0.1529], + [ 0.0709, 0.0678, -0.1033, ..., 0.0677, -0.1576, -0.1075]], + device='cuda:0'), grad: tensor([[ 1.5767e-06, 1.1846e-06, 3.0315e-07, ..., 1.9744e-06, + 3.6554e-07, 2.8871e-08], + [ 4.3819e-07, 2.2259e-07, 5.3085e-08, ..., -5.0105e-06, + -2.0266e-06, -1.1146e-05], + [ 2.6636e-07, 2.5565e-07, 3.8184e-08, ..., 2.8592e-07, + 1.0617e-07, 8.1956e-08], + ..., + [ 8.4611e-07, 7.2876e-07, 9.5926e-08, ..., 5.9493e-06, + 2.0899e-06, 1.0334e-05], + [-5.6904e-07, 8.3353e-07, 1.1642e-08, ..., 3.3481e-07, + -9.6485e-07, 2.2817e-08], + [-2.3805e-06, 8.5905e-06, -7.8976e-07, ..., -8.3670e-06, + -5.1316e-07, 3.5577e-07]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0296, -0.0303, -0.0179, 0.0003, 0.0005, 0.0027, 0.0063, -0.0104, + 0.0189, -0.0391], device='cuda:0'), grad: tensor([ 6.2026e-06, -2.5198e-05, 9.2993e-07, -2.5272e-05, 1.1109e-05, + 5.4538e-06, 1.6708e-06, 2.7284e-05, -3.3667e-07, -1.8552e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 217.58, cls_loss 0.0009 cls_loss_mapping 0.0036 cls_loss_causal 0.5210 re_mapping 0.0048 re_causal 0.0146 /// teacc 99.12 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.0341, 0.0013, 0.1072, ..., 0.0086, -0.0701, -0.0486], + [-0.1214, -0.1195, -0.1448, ..., 0.0509, 0.0499, 0.3041], + [-0.0979, -0.1112, -0.1274, ..., -0.1354, 0.0222, -0.1571], + ..., + [-0.1468, -0.1611, 0.0210, ..., -0.1723, -0.0769, -0.2755], + [ 0.0942, 0.0367, -0.1009, ..., -0.1554, 0.0177, -0.1532], + [ 0.0712, 0.0680, -0.1045, ..., 0.0681, -0.1592, -0.1083]], + device='cuda:0'), grad: tensor([[ 1.0757e-07, -4.6566e-10, 9.7789e-09, ..., 1.3504e-07, + 1.4389e-07, 1.6950e-07], + [ 8.3074e-07, 4.5216e-07, 4.6100e-08, ..., -2.8871e-08, + -1.3150e-06, -2.9244e-06], + [ 4.2561e-07, 2.1048e-07, 9.8255e-08, ..., 3.1712e-07, + 7.9069e-07, 1.0906e-06], + ..., + [ 7.7719e-07, 4.7963e-07, 1.0477e-07, ..., 6.5519e-07, + 3.7998e-07, 5.3598e-07], + [ 1.3150e-06, 7.8464e-07, 1.0524e-07, ..., 1.1176e-06, + 1.5367e-08, 2.4447e-07], + [-1.6138e-05, -8.6129e-06, 7.5437e-08, ..., -1.0200e-05, + -4.3632e-07, 7.2643e-08]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0297, -0.0302, -0.0181, 0.0003, 0.0011, 0.0028, 0.0064, -0.0104, + 0.0188, -0.0396], device='cuda:0'), grad: tensor([ 8.7311e-07, -1.5255e-06, 3.7514e-06, 1.0803e-07, 4.1991e-05, + 1.8813e-07, -1.2256e-06, 1.7900e-06, 3.8780e-06, -4.9889e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 217.32, cls_loss 0.0010 cls_loss_mapping 0.0033 cls_loss_causal 0.5173 re_mapping 0.0045 re_causal 0.0142 /// teacc 99.15 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0345, 0.0015, 0.1072, ..., 0.0084, -0.0709, -0.0494], + [-0.1226, -0.1203, -0.1453, ..., 0.0512, 0.0503, 0.3060], + [-0.0984, -0.1115, -0.1282, ..., -0.1361, 0.0221, -0.1576], + ..., + [-0.1474, -0.1617, 0.0209, ..., -0.1726, -0.0778, -0.2772], + [ 0.0946, 0.0369, -0.1012, ..., -0.1553, 0.0189, -0.1527], + [ 0.0723, 0.0688, -0.1045, ..., 0.0689, -0.1595, -0.1091]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -9.7789e-09, 1.4948e-07, ..., 1.5413e-07, + 2.8405e-08, 8.3819e-09], + [ 3.2131e-08, 3.2596e-08, 4.7497e-08, ..., -8.5542e-07, + -8.5402e-07, -1.7257e-06], + [ 5.6811e-08, 4.9360e-08, 4.7032e-08, ..., 5.1223e-08, + 4.4703e-08, 1.9092e-08], + ..., + [ 5.3085e-08, 4.2841e-08, 3.2596e-08, ..., 1.0477e-07, + 1.2992e-07, 2.0303e-07], + [ 2.4587e-07, 2.2957e-07, 4.3306e-07, ..., 4.1118e-07, + 1.7835e-07, 1.2107e-08], + [ 0.0000e+00, 2.8871e-08, 3.3528e-08, ..., 3.8184e-08, + 9.5926e-08, 1.1548e-07]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0300, -0.0299, -0.0182, 0.0003, 0.0006, 0.0028, 0.0066, -0.0107, + 0.0190, -0.0393], device='cuda:0'), grad: tensor([ 5.5041e-07, -2.6673e-06, 2.9569e-07, -8.9314e-07, 2.6729e-06, + 1.2130e-05, -1.4447e-05, 5.1642e-07, 1.5236e-06, 3.0454e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 217.27, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5231 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.08 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0333, 0.0030, 0.1086, ..., 0.0091, -0.0713, -0.0501], + [-0.1229, -0.1205, -0.1456, ..., 0.0516, 0.0506, 0.3074], + [-0.0989, -0.1120, -0.1302, ..., -0.1373, 0.0220, -0.1580], + ..., + [-0.1480, -0.1627, 0.0207, ..., -0.1730, -0.0780, -0.2781], + [ 0.0949, 0.0369, -0.1023, ..., -0.1561, 0.0190, -0.1531], + [ 0.0726, 0.0684, -0.1052, ..., 0.0692, -0.1597, -0.1100]], + device='cuda:0'), grad: tensor([[ 9.3598e-08, 1.3364e-07, 3.7253e-09, ..., 3.8184e-08, + 2.3749e-08, 1.0245e-08], + [ 1.6112e-07, 2.8079e-07, 5.3551e-08, ..., -1.0617e-07, + -4.7963e-08, -4.5169e-07], + [ 1.8440e-07, 1.5507e-07, 5.7276e-08, ..., 7.0315e-08, + -1.5460e-07, 3.5390e-08], + ..., + [ 3.7014e-05, 2.4289e-05, 1.2713e-07, ..., 1.3784e-07, + 2.3050e-07, 1.8394e-07], + [ 1.0068e-06, 1.4007e-06, 2.9383e-07, ..., 5.7975e-07, + 3.2131e-08, 3.2596e-08], + [ 4.5914e-07, 1.0841e-06, 4.4890e-07, ..., 2.9942e-07, + 4.0233e-07, 8.1956e-08]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0293, -0.0298, -0.0184, 0.0006, 0.0003, 0.0025, 0.0070, -0.0108, + 0.0190, -0.0393], device='cuda:0'), grad: tensor([ 9.9182e-05, 3.3565e-06, -1.5664e-04, -1.1086e-04, -1.1269e-06, + -2.4512e-05, 2.1979e-07, 1.6773e-04, 1.7762e-05, 5.0701e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 217.27, cls_loss 0.0010 cls_loss_mapping 0.0030 cls_loss_causal 0.5254 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.12 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0339, 0.0027, 0.1084, ..., 0.0086, -0.0715, -0.0508], + [-0.1229, -0.1207, -0.1466, ..., 0.0523, 0.0502, 0.3090], + [-0.0992, -0.1122, -0.1310, ..., -0.1368, 0.0228, -0.1582], + ..., + [-0.1489, -0.1635, 0.0205, ..., -0.1735, -0.0783, -0.2789], + [ 0.0949, 0.0369, -0.1029, ..., -0.1570, 0.0189, -0.1543], + [ 0.0734, 0.0690, -0.1052, ..., 0.0698, -0.1599, -0.1111]], + device='cuda:0'), grad: tensor([[ 3.9116e-08, -3.6228e-07, -1.3057e-06, ..., -6.4960e-07, + 4.3772e-08, -8.3353e-08], + [ 2.2771e-07, 1.8347e-07, 6.6124e-08, ..., -6.0536e-09, + -3.5530e-07, -8.7917e-07], + [ 1.9604e-07, 3.3714e-07, 6.2352e-07, ..., 5.7789e-07, + 1.9046e-07, 1.3411e-07], + ..., + [ 2.3330e-07, 1.7416e-07, 3.3062e-07, ..., 4.0419e-07, + 7.7253e-07, 4.3213e-07], + [ 9.4064e-08, 1.0664e-07, 7.7765e-08, ..., 2.1281e-07, + 1.6112e-07, 1.8533e-07], + [ 1.0710e-07, 1.4715e-07, 7.1526e-07, ..., 3.1386e-07, + 4.3213e-07, 5.5879e-08]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0300, -0.0299, -0.0176, 0.0003, 0.0003, 0.0028, 0.0070, -0.0109, + 0.0186, -0.0392], device='cuda:0'), grad: tensor([-2.7195e-06, -9.5554e-07, 2.4401e-06, -4.2468e-06, -1.8794e-06, + 1.4417e-06, 4.4797e-07, -6.1281e-07, 1.0338e-06, 5.0291e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 217.28, cls_loss 0.0018 cls_loss_mapping 0.0057 cls_loss_causal 0.5192 re_mapping 0.0047 re_causal 0.0140 /// teacc 99.11 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0334, 0.0005, 0.1078, ..., 0.0081, -0.0688, -0.0514], + [-0.1234, -0.1211, -0.1495, ..., 0.0547, 0.0473, 0.3089], + [-0.0996, -0.1127, -0.1348, ..., -0.1396, 0.0257, -0.1561], + ..., + [-0.1493, -0.1640, 0.0205, ..., -0.1765, -0.0789, -0.2794], + [ 0.0951, 0.0368, -0.1043, ..., -0.1579, 0.0185, -0.1560], + [ 0.0735, 0.0715, -0.1030, ..., 0.0726, -0.1604, -0.1118]], + device='cuda:0'), grad: tensor([[ 1.2806e-07, -9.0431e-07, -1.0841e-06, ..., -1.4761e-06, + 3.2317e-07, 1.3039e-08], + [ 1.7304e-06, 2.3358e-06, 6.6217e-07, ..., 2.0042e-06, + 3.9265e-06, -4.4424e-07], + [ 1.2591e-06, 1.7919e-06, 3.5251e-07, ..., 1.6596e-06, + 2.6058e-06, 1.3411e-07], + ..., + [ 4.0745e-07, 5.4203e-07, 8.5123e-07, ..., 5.0571e-07, + 1.4920e-06, 1.3132e-07], + [ 1.4435e-08, 1.0524e-07, 1.7369e-07, ..., 2.7660e-07, + 4.0606e-07, 1.7229e-08], + [ 2.3283e-08, 4.3539e-07, 3.1963e-06, ..., 4.5635e-07, + 2.1718e-06, 2.2352e-08]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0308, -0.0302, -0.0148, 0.0002, 0.0004, 0.0030, 0.0058, -0.0126, + 0.0180, -0.0367], device='cuda:0'), grad: tensor([-1.3039e-08, 3.1114e-05, 4.9055e-05, -1.8150e-05, -1.0736e-05, + 8.0094e-06, 2.0415e-06, -7.4565e-05, 2.0284e-06, 1.1012e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.43, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.5248 re_mapping 0.0049 re_causal 0.0153 /// teacc 99.06 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0337, 0.0007, 0.1078, ..., 0.0082, -0.0689, -0.0515], + [-0.1236, -0.1214, -0.1494, ..., 0.0548, 0.0474, 0.3096], + [-0.1000, -0.1133, -0.1367, ..., -0.1415, 0.0257, -0.1563], + ..., + [-0.1496, -0.1645, 0.0211, ..., -0.1767, -0.0790, -0.2798], + [ 0.0955, 0.0369, -0.1045, ..., -0.1583, 0.0186, -0.1565], + [ 0.0742, 0.0715, -0.1030, ..., 0.0726, -0.1607, -0.1124]], + device='cuda:0'), grad: tensor([[ 1.0710e-08, -3.0342e-06, -2.0228e-06, ..., -5.1074e-06, + 6.9849e-09, 2.3283e-09], + [ 2.9802e-08, 7.7300e-08, 6.4261e-08, ..., 5.6345e-08, + -3.2596e-08, -1.6019e-07], + [ 1.1921e-07, 3.5902e-07, 1.7975e-07, ..., 3.7625e-07, + 5.6345e-08, 4.2375e-08], + ..., + [ 7.4506e-08, 6.4261e-08, 1.7695e-08, ..., 1.0710e-07, + 4.0513e-08, 6.2399e-08], + [-4.8429e-08, -8.2422e-08, 2.5146e-06, ..., 7.1712e-07, + 2.7940e-09, 5.1223e-09], + [-3.8091e-07, 1.9725e-06, 1.6913e-06, ..., 3.5986e-06, + 2.2817e-07, 1.4435e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0307, -0.0302, -0.0149, 0.0004, 0.0001, 0.0027, 0.0056, -0.0125, + 0.0180, -0.0367], device='cuda:0'), grad: tensor([-1.0282e-05, 1.0291e-07, 1.3988e-06, -1.6484e-07, 6.9384e-08, + 5.0627e-06, -1.3173e-05, 1.3970e-07, 8.8662e-06, 8.0243e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 217.25, cls_loss 0.0010 cls_loss_mapping 0.0032 cls_loss_causal 0.5406 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.06 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0337, 0.0008, 0.1079, ..., 0.0083, -0.0689, -0.0512], + [-0.1239, -0.1214, -0.1495, ..., 0.0555, 0.0479, 0.3126], + [-0.1006, -0.1148, -0.1391, ..., -0.1452, 0.0255, -0.1580], + ..., + [-0.1500, -0.1650, 0.0209, ..., -0.1768, -0.0793, -0.2806], + [ 0.0954, 0.0365, -0.1050, ..., -0.1605, 0.0170, -0.1594], + [ 0.0747, 0.0715, -0.1031, ..., 0.0726, -0.1613, -0.1137]], + device='cuda:0'), grad: tensor([[ 1.6717e-07, -3.2363e-07, -6.9849e-07, ..., -2.2817e-07, + 7.7765e-08, 3.7253e-09], + [ 1.0654e-06, 7.6508e-07, 3.7672e-07, ..., 7.1246e-08, + 8.5542e-07, -8.4750e-08], + [ 1.2154e-07, 1.0850e-07, 6.1002e-08, ..., 5.6811e-08, + -2.0303e-07, 1.8161e-08], + ..., + [ 9.0385e-07, 4.7497e-07, 2.8312e-07, ..., 1.0859e-06, + 2.0396e-07, 2.4680e-08], + [-2.9523e-06, -2.2892e-06, -4.6566e-07, ..., 2.6077e-08, + -1.4026e-06, 8.8476e-09], + [-1.5059e-06, -4.3167e-07, 1.3821e-06, ..., -2.2464e-06, + 1.6522e-06, 7.4506e-09]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0306, -0.0297, -0.0153, 0.0001, 0.0001, 0.0032, 0.0055, -0.0125, + 0.0166, -0.0368], device='cuda:0'), grad: tensor([-1.8394e-07, 6.3032e-06, -3.8221e-06, 1.4827e-06, -1.6596e-06, + 4.9807e-06, 2.1327e-06, 3.2485e-06, -1.4417e-05, 1.8757e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 217.16, cls_loss 0.0011 cls_loss_mapping 0.0030 cls_loss_causal 0.5346 re_mapping 0.0052 re_causal 0.0148 /// teacc 99.13 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0343, 0.0008, 0.1079, ..., 0.0083, -0.0690, -0.0517], + [-0.1245, -0.1218, -0.1498, ..., 0.0556, 0.0479, 0.3132], + [-0.1016, -0.1155, -0.1405, ..., -0.1468, 0.0254, -0.1583], + ..., + [-0.1514, -0.1653, 0.0210, ..., -0.1769, -0.0794, -0.2810], + [ 0.0970, 0.0367, -0.1052, ..., -0.1607, 0.0172, -0.1597], + [ 0.0754, 0.0715, -0.1032, ..., 0.0726, -0.1622, -0.1145]], + device='cuda:0'), grad: tensor([[ 2.2817e-08, 1.7229e-08, 3.9116e-08, ..., 7.1246e-08, + 4.2375e-08, 8.1956e-08], + [ 4.5169e-08, 4.7497e-08, 6.9058e-07, ..., -1.0505e-06, + 1.1642e-06, -2.6934e-06], + [ 3.4925e-08, 4.6100e-08, 3.3528e-08, ..., 8.0792e-07, + 3.8976e-07, 1.8720e-06], + ..., + [ 2.4308e-07, 8.7079e-08, 4.5635e-08, ..., 3.7998e-07, + 1.9185e-07, 4.8801e-07], + [-6.2445e-07, -1.4221e-06, 2.1420e-08, ..., 4.1910e-08, + -9.1642e-07, 2.3283e-08], + [-3.7812e-07, -6.8452e-08, 4.4797e-07, ..., -4.0513e-07, + 8.5216e-07, 5.0291e-08]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0306, -0.0297, -0.0154, 0.0005, 0.0001, 0.0034, 0.0051, -0.0127, + 0.0177, -0.0369], device='cuda:0'), grad: tensor([ 4.4703e-07, -1.1027e-06, 3.6377e-06, 4.3120e-07, -7.2978e-06, + 5.9009e-06, 7.2271e-07, 2.2613e-06, -6.4597e-06, 1.3933e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 217.30, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5359 re_mapping 0.0051 re_causal 0.0150 /// teacc 99.11 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0347, 0.0009, 0.1079, ..., 0.0083, -0.0691, -0.0523], + [-0.1249, -0.1220, -0.1501, ..., 0.0556, 0.0480, 0.3137], + [-0.1021, -0.1161, -0.1424, ..., -0.1479, 0.0254, -0.1586], + ..., + [-0.1523, -0.1659, 0.0206, ..., -0.1770, -0.0797, -0.2814], + [ 0.0977, 0.0365, -0.1056, ..., -0.1611, 0.0173, -0.1599], + [ 0.0766, 0.0717, -0.1033, ..., 0.0727, -0.1634, -0.1152]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -1.6904e-07, -2.9569e-07, ..., -2.1420e-07, + 1.3504e-08, -1.3970e-08], + [ 4.2375e-08, 4.0047e-08, 1.2713e-07, ..., 4.6566e-09, + 1.4203e-07, -3.7719e-08], + [ 2.8824e-07, 2.4727e-07, 1.1176e-07, ..., 8.7079e-08, + 2.8592e-07, 9.3132e-09], + ..., + [ 4.4703e-08, 3.6322e-08, 5.4482e-08, ..., 1.9558e-08, + 9.2201e-08, 1.6764e-08], + [ 4.6566e-10, 4.9826e-08, 1.3504e-08, ..., 1.3970e-08, + -1.5832e-08, 1.3970e-09], + [-8.3819e-08, 5.7742e-08, 1.9791e-07, ..., 2.6077e-08, + 6.7987e-08, 1.1176e-08]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0306, -0.0297, -0.0155, 0.0003, 0.0015, 0.0034, 0.0052, -0.0129, + 0.0181, -0.0370], device='cuda:0'), grad: tensor([-6.5006e-07, 6.3842e-07, 2.3730e-06, -1.4249e-06, -3.8836e-07, + -5.6159e-07, 5.8534e-07, -1.3690e-06, 2.0070e-07, 6.0024e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 217.19, cls_loss 0.0010 cls_loss_mapping 0.0037 cls_loss_causal 0.5023 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.06 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0361, 0.0008, 0.1079, ..., 0.0083, -0.0697, -0.0524], + [-0.1252, -0.1222, -0.1503, ..., 0.0557, 0.0480, 0.3145], + [-0.1026, -0.1165, -0.1434, ..., -0.1485, 0.0255, -0.1589], + ..., + [-0.1525, -0.1662, 0.0213, ..., -0.1771, -0.0800, -0.2822], + [ 0.0977, 0.0384, -0.1035, ..., -0.1616, 0.0176, -0.1600], + [ 0.0779, 0.0718, -0.1033, ..., 0.0728, -0.1636, -0.1159]], + device='cuda:0'), grad: tensor([[ 9.4995e-08, 1.1735e-07, 3.5390e-08, ..., 4.9360e-08, + 1.6252e-07, 9.3132e-09], + [ 2.3702e-07, 2.9523e-07, 2.5611e-07, ..., 2.4680e-08, + 6.2771e-07, -6.0350e-07], + [ 4.8522e-07, 6.0489e-07, 4.6100e-08, ..., 2.3656e-07, + 5.1502e-07, 5.6345e-08], + ..., + [ 2.8452e-07, 3.4971e-07, 5.4948e-08, ..., 1.4948e-07, + 4.1351e-07, 4.0513e-08], + [-2.4680e-08, 2.0489e-08, 1.2107e-07, ..., 1.2806e-07, + 1.9977e-07, 2.9942e-07], + [-2.7427e-07, -9.7789e-08, 7.6136e-07, ..., -2.4214e-07, + 1.0077e-06, 1.3039e-08]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0306, -0.0297, -0.0154, 0.0008, 0.0009, 0.0032, 0.0026, -0.0128, + 0.0206, -0.0370], device='cuda:0'), grad: tensor([ 6.3330e-07, 1.6429e-06, 8.5728e-07, -6.0052e-06, -3.7849e-06, + 2.0228e-06, 7.9162e-09, 1.5423e-06, 9.1363e-07, 2.1495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 217.27, cls_loss 0.0015 cls_loss_mapping 0.0039 cls_loss_causal 0.5211 re_mapping 0.0046 re_causal 0.0143 /// teacc 99.06 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0374, 0.0006, 0.1079, ..., 0.0082, -0.0700, -0.0527], + [-0.1258, -0.1228, -0.1506, ..., 0.0565, 0.0481, 0.3154], + [-0.1034, -0.1173, -0.1448, ..., -0.1500, 0.0255, -0.1593], + ..., + [-0.1532, -0.1669, 0.0206, ..., -0.1780, -0.0804, -0.2829], + [ 0.0982, 0.0373, -0.1054, ..., -0.1623, 0.0177, -0.1628], + [ 0.0791, 0.0720, -0.1035, ..., 0.0729, -0.1657, -0.1171]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, -2.3283e-07, -5.1036e-07, ..., -2.6915e-07, + -1.2806e-07, -1.8114e-07], + [ 7.9162e-09, 1.4435e-08, 1.2852e-07, ..., -7.2643e-08, + -5.2750e-06, -3.7253e-07], + [ 9.3132e-09, 1.3039e-08, 2.2817e-08, ..., 3.9116e-08, + 4.3176e-06, 1.6950e-07], + ..., + [ 1.3039e-08, 1.2573e-08, 1.8021e-07, ..., 2.4680e-08, + 1.2405e-06, 5.9139e-08], + [-4.0047e-08, 7.0315e-08, 1.4296e-07, ..., 9.4529e-08, + 6.0536e-09, 9.0804e-08], + [-7.9162e-08, -3.2131e-08, 1.1362e-07, ..., -2.8871e-08, + 1.3085e-07, 2.6543e-08]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0307, -0.0290, -0.0153, 0.0008, 0.0025, 0.0047, 0.0024, -0.0137, + 0.0196, -0.0372], device='cuda:0'), grad: tensor([ 1.5879e-07, -1.6257e-05, 9.9540e-06, 4.4238e-07, -2.2491e-07, + 6.1002e-07, 9.8720e-07, 3.3993e-07, 9.8348e-07, 2.9691e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 217.26, cls_loss 0.0010 cls_loss_mapping 0.0036 cls_loss_causal 0.5467 re_mapping 0.0046 re_causal 0.0144 /// teacc 99.09 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0377, 0.0006, 0.1078, ..., 0.0082, -0.0702, -0.0531], + [-0.1266, -0.1234, -0.1509, ..., 0.0571, 0.0482, 0.3167], + [-0.1036, -0.1176, -0.1455, ..., -0.1504, 0.0256, -0.1599], + ..., + [-0.1534, -0.1672, 0.0206, ..., -0.1785, -0.0807, -0.2834], + [ 0.0988, 0.0376, -0.1057, ..., -0.1628, 0.0178, -0.1634], + [ 0.0800, 0.0723, -0.1035, ..., 0.0729, -0.1662, -0.1179]], + device='cuda:0'), grad: tensor([[ 6.8452e-08, 1.0664e-07, 5.6345e-08, ..., 0.0000e+00, + 1.4994e-07, 7.4506e-09], + [ 3.3388e-07, 7.1293e-07, 1.2200e-07, ..., 3.3528e-08, + 3.2783e-07, -3.8277e-07], + [ 6.7195e-07, 1.0747e-06, 1.2899e-07, ..., 2.0862e-07, + 6.8219e-07, 4.7032e-08], + ..., + [-1.4640e-06, -1.4082e-06, 2.8592e-07, ..., 1.1688e-07, + 5.6112e-07, 1.6764e-07], + [ 2.2165e-07, 3.8417e-07, 8.6613e-08, ..., 8.5682e-08, + 1.3085e-07, 3.3993e-08], + [ 5.6205e-07, 1.5255e-06, 3.1944e-07, ..., -3.7253e-08, + 4.6007e-07, 8.9407e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0308, -0.0279, -0.0153, 0.0008, 0.0014, 0.0044, 0.0029, -0.0147, + 0.0195, -0.0371], device='cuda:0'), grad: tensor([ 1.3039e-06, 5.5283e-06, 1.0796e-05, -6.3889e-06, -7.5530e-07, + 1.0403e-06, 8.4704e-07, -3.2604e-05, 4.0978e-06, 1.6108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 217.46, cls_loss 0.0011 cls_loss_mapping 0.0026 cls_loss_causal 0.5242 re_mapping 0.0046 re_causal 0.0141 /// teacc 99.08 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0379, 0.0006, 0.1078, ..., 0.0082, -0.0705, -0.0539], + [-0.1263, -0.1239, -0.1514, ..., 0.0557, 0.0468, 0.3146], + [-0.1043, -0.1181, -0.1469, ..., -0.1515, 0.0255, -0.1603], + ..., + [-0.1540, -0.1676, 0.0204, ..., -0.1788, -0.0810, -0.2845], + [ 0.0991, 0.0376, -0.1065, ..., -0.1645, 0.0166, -0.1647], + [ 0.0811, 0.0725, -0.1036, ..., 0.0730, -0.1667, -0.1200]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, -4.1910e-09, -1.2107e-08, ..., -3.2596e-09, + 5.5879e-09, 6.9849e-09], + [ 9.7789e-09, 7.9162e-09, 1.8626e-09, ..., -4.7963e-08, + -1.8626e-08, -1.7881e-07], + [ 2.8871e-08, 1.7695e-08, 7.4506e-09, ..., 2.3283e-08, + -5.9139e-08, 3.4459e-08], + ..., + [ 6.8452e-08, 4.4703e-08, 0.0000e+00, ..., 6.4727e-08, + 2.2817e-08, 5.1688e-08], + [ 9.4064e-08, 9.4995e-08, 4.1910e-09, ..., 7.5437e-08, + 1.3039e-08, 2.1886e-08], + [-2.2352e-07, -1.9465e-07, 2.7940e-09, ..., -1.6810e-07, + 1.0245e-08, 1.8161e-08]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0309, -0.0288, -0.0152, 0.0008, 0.0009, 0.0038, 0.0068, -0.0147, + 0.0188, -0.0371], device='cuda:0'), grad: tensor([ 9.2201e-08, 9.7416e-07, 7.2224e-07, 6.8452e-08, 2.5285e-07, + 1.4110e-07, 4.9360e-08, -2.6003e-06, 3.8603e-07, -8.1956e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 217.21, cls_loss 0.0009 cls_loss_mapping 0.0028 cls_loss_causal 0.5164 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.14 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0380, 0.0008, 0.1079, ..., 0.0082, -0.0707, -0.0543], + [-0.1272, -0.1244, -0.1517, ..., 0.0557, 0.0468, 0.3149], + [-0.1049, -0.1186, -0.1477, ..., -0.1521, 0.0255, -0.1606], + ..., + [-0.1555, -0.1679, 0.0205, ..., -0.1789, -0.0812, -0.2850], + [ 0.0994, 0.0374, -0.1067, ..., -0.1651, 0.0170, -0.1649], + [ 0.0819, 0.0727, -0.1036, ..., 0.0731, -0.1668, -0.1210]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, -2.7614e-07, -3.1665e-07, ..., -2.1630e-07, + 2.4447e-08, 3.7253e-09], + [ 3.5856e-08, 3.4226e-08, 6.2166e-08, ..., -8.9873e-08, + -1.1977e-06, -3.0808e-06], + [ 1.8161e-08, 1.9697e-07, 2.9174e-07, ..., 2.3260e-07, + 8.0420e-07, 1.9055e-06], + ..., + [-6.8685e-08, 2.5844e-08, 4.2608e-08, ..., 5.2154e-08, + 4.0885e-07, 9.5647e-07], + [-1.3108e-07, -9.4995e-08, 4.7497e-08, ..., 4.5169e-08, + 2.0955e-08, 6.2864e-09], + [ 4.4703e-08, 1.1339e-07, 2.3516e-07, ..., -3.9116e-08, + 3.6322e-07, 3.7020e-08]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0308, -0.0289, -0.0150, 0.0015, 0.0011, 0.0036, 0.0066, -0.0148, + 0.0187, -0.0371], device='cuda:0'), grad: tensor([-4.5565e-07, 1.0028e-05, 7.1265e-06, 2.3525e-06, -2.6412e-06, + -2.4773e-07, 1.6647e-07, -1.9625e-05, 9.1502e-08, 3.1926e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 217.05, cls_loss 0.0011 cls_loss_mapping 0.0027 cls_loss_causal 0.4863 re_mapping 0.0047 re_causal 0.0135 /// teacc 99.12 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0382, 0.0008, 0.1081, ..., 0.0083, -0.0687, -0.0550], + [-0.1277, -0.1248, -0.1524, ..., 0.0559, 0.0470, 0.3157], + [-0.1054, -0.1191, -0.1516, ..., -0.1559, 0.0250, -0.1616], + ..., + [-0.1561, -0.1683, 0.0205, ..., -0.1790, -0.0817, -0.2862], + [ 0.0996, 0.0372, -0.1070, ..., -0.1658, 0.0171, -0.1653], + [ 0.0831, 0.0727, -0.1037, ..., 0.0731, -0.1674, -0.1231]], + device='cuda:0'), grad: tensor([[ 1.5367e-08, -1.0501e-07, -1.4040e-07, ..., -1.2247e-07, + 9.7789e-09, 1.2806e-08], + [ 1.8859e-08, 1.7928e-08, 2.1420e-08, ..., -3.4319e-07, + -2.2119e-07, -8.5216e-07], + [ 1.4901e-08, 2.0256e-08, 5.4250e-08, ..., 6.4494e-08, + 1.0245e-08, 4.3772e-08], + ..., + [ 5.1456e-08, 4.1677e-08, 9.3132e-09, ..., 2.5751e-07, + 1.3434e-07, 4.5262e-07], + [-6.0536e-09, 1.5367e-08, 4.7265e-08, ..., 8.7544e-08, + -5.1921e-08, 1.1409e-08], + [-2.2817e-07, -1.1083e-07, 5.5414e-08, ..., -5.6112e-08, + 6.7987e-08, 2.1886e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0306, -0.0289, -0.0157, 0.0014, 0.0008, 0.0036, 0.0066, -0.0145, + 0.0184, -0.0371], device='cuda:0'), grad: tensor([-1.2061e-07, 6.4969e-06, 1.3551e-07, 6.4773e-07, 1.4277e-06, + 2.6985e-07, -1.2331e-06, -1.1295e-05, 3.0058e-07, 3.3658e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 217.21, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5227 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.09 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0383, 0.0009, 0.1081, ..., 0.0083, -0.0689, -0.0555], + [-0.1279, -0.1252, -0.1526, ..., 0.0560, 0.0475, 0.3161], + [-0.1057, -0.1192, -0.1521, ..., -0.1564, 0.0247, -0.1618], + ..., + [-0.1566, -0.1690, 0.0203, ..., -0.1792, -0.0823, -0.2872], + [ 0.0998, 0.0371, -0.1074, ..., -0.1662, 0.0174, -0.1654], + [ 0.0840, 0.0728, -0.1037, ..., 0.0732, -0.1677, -0.1257]], + device='cuda:0'), grad: tensor([[-4.5146e-07, -3.5856e-06, -3.1609e-06, ..., -3.9227e-06, + 8.3353e-08, 1.3970e-09], + [ 8.6799e-07, 2.9569e-08, 5.9325e-07, ..., 0.0000e+00, + 1.1604e-06, -9.4762e-08], + [ 1.5600e-08, 2.6310e-08, 5.4017e-08, ..., 3.3295e-08, + 3.7951e-08, 7.6834e-09], + ..., + [ 2.2352e-08, 1.8394e-08, 5.1456e-08, ..., 3.7951e-08, + 5.7975e-08, 4.7497e-08], + [-1.5311e-06, 3.1432e-08, -7.7114e-07, ..., 7.4273e-08, + -2.0191e-06, 2.7940e-09], + [ 4.8894e-07, 2.7195e-06, 3.6154e-06, ..., 2.9169e-06, + 1.0980e-06, 1.7462e-08]], device='cuda:0') +Epoch 190, bias, value: tensor([-3.0595e-02, -2.8477e-02, -1.6047e-02, 1.1960e-03, -4.2502e-05, + 4.5360e-03, 6.4396e-03, -1.4688e-02, 1.8372e-02, -3.7117e-02], + device='cuda:0'), grad: tensor([-8.1658e-06, 4.5151e-06, 2.9686e-07, 6.5705e-07, -5.2564e-06, + 1.1008e-06, 3.3211e-06, -1.6647e-07, -7.0632e-06, 1.0736e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 217.22, cls_loss 0.0011 cls_loss_mapping 0.0026 cls_loss_causal 0.4893 re_mapping 0.0049 re_causal 0.0136 /// teacc 99.05 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0388, 0.0010, 0.1081, ..., 0.0083, -0.0691, -0.0562], + [-0.1287, -0.1258, -0.1534, ..., 0.0562, 0.0478, 0.3168], + [-0.1063, -0.1195, -0.1525, ..., -0.1572, 0.0245, -0.1628], + ..., + [-0.1569, -0.1697, 0.0209, ..., -0.1794, -0.0829, -0.2880], + [ 0.1001, 0.0372, -0.1083, ..., -0.1671, 0.0173, -0.1657], + [ 0.0849, 0.0730, -0.1037, ..., 0.0733, -0.1685, -0.1276]], + device='cuda:0'), grad: tensor([[ 2.4657e-07, 2.4843e-07, -3.0501e-08, ..., 1.4971e-07, + 4.4238e-08, 0.0000e+00], + [ 2.1933e-07, 1.9744e-06, 2.0186e-07, ..., 1.0128e-07, + 4.4378e-07, -2.5146e-08], + [ 2.5611e-07, 2.9313e-07, 2.0256e-08, ..., 1.9907e-07, + 6.6822e-08, 9.3132e-10], + ..., + [ 1.2363e-07, -2.9281e-06, 2.6776e-08, ..., 9.0105e-08, + 1.1688e-07, 8.6147e-09], + [ 2.3679e-07, 2.8801e-07, 4.8894e-08, ..., 1.8161e-07, + 5.4948e-08, 6.9849e-10], + [ 1.0971e-06, 2.3227e-06, 2.5518e-07, ..., 9.5554e-07, + 5.4995e-07, 9.3132e-09]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0307, -0.0282, -0.0164, 0.0014, -0.0003, 0.0043, 0.0065, -0.0147, + 0.0182, -0.0371], device='cuda:0'), grad: tensor([ 9.4064e-07, 9.2089e-05, 1.1958e-06, 2.5481e-06, -1.5786e-06, + 1.2266e-06, -2.0210e-07, -1.3959e-04, 2.5146e-06, 4.1008e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 217.03, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.5609 re_mapping 0.0049 re_causal 0.0151 /// teacc 99.11 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0390, 0.0010, 0.1082, ..., 0.0083, -0.0692, -0.0564], + [-0.1293, -0.1262, -0.1538, ..., 0.0562, 0.0478, 0.3169], + [-0.1074, -0.1206, -0.1527, ..., -0.1575, 0.0243, -0.1626], + ..., + [-0.1576, -0.1710, 0.0209, ..., -0.1795, -0.0834, -0.2885], + [ 0.1014, 0.0376, -0.1087, ..., -0.1672, 0.0179, -0.1658], + [ 0.0848, 0.0729, -0.1037, ..., 0.0733, -0.1689, -0.1285]], + device='cuda:0'), grad: tensor([[ 2.0955e-08, -1.6647e-07, 7.8138e-07, ..., 4.5449e-07, + 1.6997e-08, 3.0268e-09], + [ 1.3970e-08, 4.4238e-08, 4.5402e-08, ..., -4.1910e-09, + 1.6764e-08, -8.2888e-08], + [ 2.0023e-08, -2.1886e-08, 5.0990e-08, ..., 4.4703e-08, + 1.5134e-08, 9.0804e-09], + ..., + [ 1.4226e-07, 1.1805e-07, 1.6298e-08, ..., 1.5111e-07, + 2.0489e-08, 1.1176e-08], + [-1.5367e-07, -9.7323e-08, 6.8685e-08, ..., 5.9372e-08, + -1.5087e-07, 8.1491e-09], + [-1.8277e-07, 1.3807e-07, 4.9174e-07, ..., 6.2864e-09, + 3.3760e-07, 1.4668e-08]], device='cuda:0') +Epoch 192, bias, value: tensor([-3.0652e-02, -2.8257e-02, -1.6616e-02, 1.2821e-03, -7.6303e-05, + 4.5300e-03, 6.5501e-03, -1.4596e-02, 1.8562e-02, -3.7201e-02], + device='cuda:0'), grad: tensor([ 2.1644e-06, 2.9453e-07, -2.7800e-07, 4.7730e-07, -1.0673e-06, + 4.7428e-07, -3.3602e-06, 1.0268e-07, -7.1479e-08, 1.2983e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 217.03, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4977 re_mapping 0.0047 re_causal 0.0143 /// teacc 99.13 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0397, 0.0010, 0.1082, ..., 0.0083, -0.0690, -0.0562], + [-0.1316, -0.1270, -0.1543, ..., 0.0578, 0.0479, 0.3172], + [-0.1082, -0.1212, -0.1530, ..., -0.1583, 0.0242, -0.1632], + ..., + [-0.1582, -0.1714, 0.0208, ..., -0.1816, -0.0836, -0.2888], + [ 0.1020, 0.0381, -0.1087, ..., -0.1674, 0.0183, -0.1659], + [ 0.0861, 0.0731, -0.1037, ..., 0.0734, -0.1696, -0.1290]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, -6.0536e-09, -4.6566e-10, ..., 7.9162e-09, + 4.8894e-09, 1.1642e-09], + [ 7.7561e-06, 5.5879e-09, 4.4238e-09, ..., -1.0943e-08, + 1.3970e-08, -5.2387e-08], + [ 1.8533e-07, 1.3271e-08, 5.8208e-09, ..., 7.2177e-09, + -2.9337e-08, 1.3504e-08], + ..., + [-1.3746e-05, 3.3760e-08, 1.0245e-08, ..., 7.7067e-08, + 2.2585e-08, 1.4435e-08], + [-1.1409e-08, -3.9581e-08, 9.7789e-09, ..., 2.4913e-08, + -3.7486e-08, 6.2864e-09], + [ 5.5581e-06, -8.6613e-08, 9.3132e-08, ..., -2.0722e-07, + 1.2154e-07, 6.2864e-09]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0306, -0.0268, -0.0170, 0.0012, 0.0001, 0.0041, 0.0065, -0.0161, + 0.0190, -0.0371], device='cuda:0'), grad: tensor([ 2.6892e-07, 1.8907e-04, 3.7495e-06, 9.7416e-07, 1.5255e-06, + 1.2456e-07, -1.5995e-07, -3.3379e-04, 4.1025e-07, 1.3781e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 217.27, cls_loss 0.0007 cls_loss_mapping 0.0029 cls_loss_causal 0.4883 re_mapping 0.0043 re_causal 0.0136 /// teacc 99.06 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0403, 0.0009, 0.1082, ..., 0.0083, -0.0692, -0.0565], + [-0.1326, -0.1274, -0.1544, ..., 0.0578, 0.0479, 0.3173], + [-0.1087, -0.1217, -0.1531, ..., -0.1588, 0.0242, -0.1633], + ..., + [-0.1588, -0.1720, 0.0206, ..., -0.1820, -0.0837, -0.2891], + [ 0.1020, 0.0382, -0.1089, ..., -0.1678, 0.0184, -0.1660], + [ 0.0875, 0.0734, -0.1037, ..., 0.0736, -0.1698, -0.1295]], + device='cuda:0'), grad: tensor([[ 6.1560e-07, 1.2219e-06, -1.3434e-07, ..., -2.7008e-08, + 5.6857e-07, 7.6834e-09], + [ 8.0094e-08, 1.1292e-07, 6.3330e-08, ..., -7.7067e-08, + 1.2503e-07, -1.7602e-07], + [ 2.8871e-08, -4.1304e-07, 1.4505e-07, ..., 1.3737e-08, + -2.8522e-07, 1.5367e-08], + ..., + [ 9.6159e-08, 5.2387e-08, 1.0547e-07, ..., 2.9104e-08, + 3.2154e-07, 3.7951e-08], + [-7.0455e-07, -7.9675e-07, 5.3551e-08, ..., -7.2643e-08, + -1.4435e-07, 4.6799e-08], + [ 2.5276e-06, 3.8464e-07, 2.0210e-06, ..., 3.7719e-08, + 7.0781e-06, 2.2119e-08]], device='cuda:0') +Epoch 194, bias, value: tensor([-3.0727e-02, -2.7567e-02, -1.6955e-02, 6.0893e-04, -7.2062e-05, + 4.2131e-03, 6.6394e-03, -1.5283e-02, 1.8897e-02, -3.6955e-02], + device='cuda:0'), grad: tensor([ 3.1050e-06, 7.3854e-07, -3.7905e-06, -1.1944e-07, -2.4825e-05, + -4.6194e-07, 6.7800e-07, 1.1958e-06, 2.4750e-07, 2.3216e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 217.25, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.4944 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.10 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0429, 0.0006, 0.1082, ..., 0.0080, -0.0693, -0.0576], + [-0.1332, -0.1283, -0.1549, ..., 0.0578, 0.0481, 0.3176], + [-0.1092, -0.1221, -0.1534, ..., -0.1596, 0.0240, -0.1635], + ..., + [-0.1591, -0.1724, 0.0206, ..., -0.1821, -0.0839, -0.2896], + [ 0.1020, 0.0375, -0.1093, ..., -0.1688, 0.0183, -0.1664], + [ 0.0887, 0.0738, -0.1037, ..., 0.0739, -0.1704, -0.1310]], + device='cuda:0'), grad: tensor([[ 4.1444e-08, -2.1653e-08, -9.2899e-08, ..., -2.3283e-10, + 2.0955e-08, 6.9849e-10], + [ 7.4506e-08, 4.2608e-08, 8.1491e-09, ..., 6.9384e-08, + 2.1653e-08, -1.5367e-08], + [ 1.6531e-08, 1.5600e-08, 1.4668e-08, ..., 2.0256e-08, + -1.3178e-07, 3.0268e-09], + ..., + [ 1.8557e-07, 8.2888e-08, 1.3039e-08, ..., 1.6810e-07, + 6.2864e-09, 1.0245e-08], + [ 6.9849e-10, 2.7940e-09, 1.3504e-08, ..., 7.1246e-08, + -4.6566e-09, 1.6298e-09], + [-2.4084e-06, -1.0971e-06, 1.7229e-08, ..., -2.2314e-06, + 2.3749e-08, 1.6298e-09]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0311, -0.0276, -0.0172, 0.0005, -0.0005, 0.0042, 0.0072, -0.0150, + 0.0185, -0.0367], device='cuda:0'), grad: tensor([ 1.5832e-08, 3.8929e-07, -4.6170e-07, 1.8859e-07, 6.9998e-06, + 1.1083e-07, 1.3458e-07, 6.0443e-07, 2.4633e-07, -8.2180e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 217.03, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.5290 re_mapping 0.0047 re_causal 0.0146 /// teacc 99.10 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0427, 0.0008, 0.1083, ..., 0.0081, -0.0693, -0.0590], + [-0.1348, -0.1289, -0.1552, ..., 0.0580, 0.0481, 0.3181], + [-0.1100, -0.1229, -0.1540, ..., -0.1612, 0.0241, -0.1635], + ..., + [-0.1603, -0.1730, 0.0206, ..., -0.1824, -0.0840, -0.2904], + [ 0.1022, 0.0375, -0.1097, ..., -0.1692, 0.0188, -0.1666], + [ 0.0889, 0.0738, -0.1038, ..., 0.0740, -0.1724, -0.1317]], + device='cuda:0'), grad: tensor([[ 1.4831e-07, 4.6566e-10, -4.8429e-08, ..., 1.0547e-07, + 1.7928e-08, 2.2817e-08], + [ 9.2853e-07, 2.8126e-07, 1.8626e-08, ..., 1.6089e-07, + -5.3039e-07, -1.3523e-06], + [ 2.1071e-07, 1.5879e-07, 1.9092e-08, ..., 2.5821e-07, + 1.3295e-07, 2.1001e-07], + ..., + [ 6.5193e-07, 2.1630e-07, 3.2596e-09, ..., 8.7963e-07, + 3.9558e-07, 8.6520e-07], + [ 1.3481e-07, 5.7044e-08, 2.4401e-07, ..., 5.7463e-07, + -4.4238e-08, 1.7928e-08], + [-3.1404e-06, -6.6496e-07, 1.9791e-08, ..., -2.2613e-06, + 5.9139e-08, 9.5461e-08]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0310, -0.0277, -0.0172, 0.0002, 0.0001, 0.0049, 0.0068, -0.0151, + 0.0186, -0.0368], device='cuda:0'), grad: tensor([ 4.7660e-07, -2.0955e-08, 1.2936e-06, -2.9374e-06, 2.0694e-06, + 5.5134e-06, -4.5262e-06, 2.1197e-06, 2.0228e-06, -6.0312e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 216.87, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.5193 re_mapping 0.0044 re_causal 0.0138 /// teacc 99.13 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0428, 0.0008, 0.1084, ..., 0.0081, -0.0695, -0.0598], + [-0.1363, -0.1309, -0.1553, ..., 0.0582, 0.0482, 0.3187], + [-0.1104, -0.1234, -0.1543, ..., -0.1619, 0.0241, -0.1641], + ..., + [-0.1601, -0.1707, 0.0206, ..., -0.1825, -0.0837, -0.2913], + [ 0.1024, 0.0375, -0.1100, ..., -0.1697, 0.0193, -0.1668], + [ 0.0912, 0.0741, -0.1039, ..., 0.0741, -0.1727, -0.1324]], + device='cuda:0'), grad: tensor([[ 3.6787e-08, -4.3539e-07, -2.4005e-07, ..., -2.5635e-07, + 1.3504e-08, -7.8231e-08], + [ 5.4482e-08, 7.7765e-08, 8.2189e-08, ..., 1.9791e-08, + -3.0501e-08, -1.6694e-07], + [ 1.1339e-07, 1.1781e-07, 1.0990e-07, ..., 1.8370e-07, + 6.7055e-08, 4.1910e-08], + ..., + [ 9.4529e-08, 9.0571e-08, 3.9814e-08, ..., 1.0873e-07, + 3.8184e-08, 5.7742e-08], + [ 1.2340e-07, 1.1013e-07, 5.9139e-08, ..., 1.2945e-07, + 3.6322e-08, 2.5379e-08], + [-2.9895e-07, -3.7253e-09, 4.5891e-07, ..., 3.0524e-07, + 2.8871e-08, 6.9849e-08]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0310, -0.0279, -0.0171, -0.0009, -0.0009, 0.0049, 0.0066, -0.0145, + 0.0186, -0.0366], device='cuda:0'), grad: tensor([ 1.4007e-06, 9.2853e-07, -8.4266e-06, 2.8359e-07, 9.1735e-07, + 2.1942e-06, -3.2149e-06, 2.7008e-06, 2.3879e-06, 8.2189e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 217.06, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.4994 re_mapping 0.0041 re_causal 0.0131 /// teacc 99.05 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0430, 0.0009, 0.1084, ..., 0.0082, -0.0697, -0.0604], + [-0.1371, -0.1322, -0.1555, ..., 0.0585, 0.0482, 0.3194], + [-0.1113, -0.1241, -0.1545, ..., -0.1626, 0.0240, -0.1645], + ..., + [-0.1610, -0.1713, 0.0204, ..., -0.1827, -0.0837, -0.2921], + [ 0.1023, 0.0373, -0.1104, ..., -0.1706, 0.0194, -0.1670], + [ 0.0923, 0.0743, -0.1040, ..., 0.0742, -0.1737, -0.1336]], + device='cuda:0'), grad: tensor([[ 2.1188e-08, 1.2806e-08, -7.5288e-06, ..., -7.6666e-06, + -3.1944e-06, 9.5461e-09], + [ 1.0617e-07, 1.9721e-07, 3.8147e-06, ..., 3.6694e-06, + 1.6103e-06, -5.9977e-07], + [ 1.0710e-07, 2.0955e-07, 2.1122e-06, ..., 2.2240e-06, + 1.0487e-06, 6.8685e-08], + ..., + [ 3.6554e-08, 8.3586e-08, 1.1362e-07, ..., 2.4308e-07, + 1.4482e-07, 3.0431e-07], + [ 3.1665e-08, 1.5227e-07, 1.4063e-07, ..., 1.9907e-07, + 2.2585e-08, 3.1432e-08], + [-1.5367e-07, 9.0571e-08, 2.6287e-07, ..., 7.4506e-09, + 1.1711e-07, 1.0547e-07]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0310, -0.0281, -0.0171, 0.0021, -0.0005, 0.0020, 0.0065, -0.0142, + 0.0185, -0.0367], device='cuda:0'), grad: tensor([-2.3246e-05, 1.1466e-05, 9.2946e-07, -6.3032e-06, -1.0408e-07, + 5.1185e-06, 3.1851e-06, 6.5528e-06, 1.3579e-06, 1.0449e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 217.02, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4955 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0434, 0.0008, 0.1084, ..., 0.0082, -0.0697, -0.0611], + [-0.1378, -0.1331, -0.1559, ..., 0.0589, 0.0483, 0.3198], + [-0.1119, -0.1245, -0.1549, ..., -0.1632, 0.0240, -0.1647], + ..., + [-0.1595, -0.1724, 0.0202, ..., -0.1825, -0.0843, -0.2936], + [ 0.1024, 0.0370, -0.1108, ..., -0.1712, 0.0195, -0.1674], + [ 0.0911, 0.0744, -0.1042, ..., 0.0741, -0.1748, -0.1345]], + device='cuda:0'), grad: tensor([[ 7.6601e-08, 1.0733e-07, 4.5402e-08, ..., 1.5786e-07, + 7.5903e-08, 5.5879e-09], + [ 1.0082e-07, 1.3737e-07, 2.5844e-08, ..., -1.3737e-07, + -1.8161e-08, -5.2247e-07], + [ 4.0978e-08, 1.0221e-07, 7.1246e-08, ..., 1.3085e-07, + 5.9372e-08, 5.1223e-08], + ..., + [ 1.7956e-06, 3.0622e-06, 5.9605e-08, ..., 5.5786e-07, + 1.6578e-06, 4.1025e-07], + [-2.4661e-06, -4.4592e-06, 3.9116e-08, ..., 7.4739e-08, + -2.4792e-06, 6.7521e-09], + [ 5.2620e-08, 1.8720e-07, -2.6450e-06, ..., -9.7379e-06, + 7.8417e-07, 2.0722e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0310, -0.0280, -0.0167, 0.0022, 0.0003, 0.0020, 0.0063, -0.0139, + 0.0183, -0.0374], device='cuda:0'), grad: tensor([ 1.4249e-06, 4.2329e-07, 1.0673e-06, 1.1856e-06, 2.6971e-05, + -4.7199e-06, 6.3367e-06, 2.5198e-05, -3.7372e-05, -2.0504e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 217.05, cls_loss 0.0011 cls_loss_mapping 0.0029 cls_loss_causal 0.4551 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.04 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0445, 0.0006, 0.1085, ..., 0.0081, -0.0698, -0.0627], + [-0.1385, -0.1339, -0.1562, ..., 0.0591, 0.0486, 0.3199], + [-0.1120, -0.1252, -0.1555, ..., -0.1633, 0.0256, -0.1629], + ..., + [-0.1601, -0.1751, 0.0201, ..., -0.1830, -0.0867, -0.2954], + [ 0.1025, 0.0372, -0.1113, ..., -0.1725, 0.0179, -0.1679], + [ 0.0923, 0.0752, -0.1042, ..., 0.0743, -0.1754, -0.1360]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, -4.2608e-08, -3.1898e-08, ..., -2.9104e-08, + 1.0943e-08, 1.9558e-08], + [ 1.2969e-07, 6.0536e-09, 2.3283e-09, ..., -2.8056e-07, + -2.2142e-07, -7.2224e-07], + [ 2.4680e-08, 2.3050e-08, 1.1642e-08, ..., 1.3434e-07, + 1.0012e-07, 1.1991e-07], + ..., + [-3.1595e-07, 6.0536e-09, 4.6566e-10, ..., 1.8999e-07, + 1.4319e-07, 4.5751e-07], + [-6.7987e-08, -8.8941e-08, 1.5832e-08, ..., 1.3597e-07, + 5.7509e-08, 1.9558e-08], + [ 5.0757e-08, 4.8894e-09, 1.1642e-08, ..., 8.1491e-09, + 2.1188e-08, 6.1234e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0311, -0.0284, -0.0139, 0.0021, 0.0001, 0.0020, 0.0061, -0.0148, + 0.0175, -0.0371], device='cuda:0'), grad: tensor([ 6.6822e-08, 3.3621e-06, 1.1222e-06, 4.2282e-06, 2.6473e-07, + 6.5193e-07, -1.8217e-06, -1.1712e-05, 7.2038e-07, 3.1255e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 217.05, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.4963 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.06 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0448, 0.0007, 0.1085, ..., 0.0082, -0.0700, -0.0635], + [-0.1392, -0.1345, -0.1564, ..., 0.0594, 0.0490, 0.3205], + [-0.1142, -0.1270, -0.1558, ..., -0.1648, 0.0253, -0.1634], + ..., + [-0.1606, -0.1759, 0.0199, ..., -0.1834, -0.0868, -0.2964], + [ 0.1021, 0.0370, -0.1116, ..., -0.1738, 0.0177, -0.1684], + [ 0.0928, 0.0755, -0.1042, ..., 0.0744, -0.1760, -0.1371]], + device='cuda:0'), grad: tensor([[ 1.8161e-08, -6.1467e-08, 4.2655e-07, ..., 1.2293e-06, + 1.4901e-08, 3.7253e-09], + [ 1.4855e-07, 3.6322e-08, 1.9558e-08, ..., -9.4995e-08, + 3.7253e-09, -1.7835e-07], + [ 7.9162e-08, 4.4238e-08, 1.1176e-08, ..., 6.3330e-08, + 1.2573e-08, 2.1886e-08], + ..., + [-2.7940e-08, 3.5390e-08, 7.4506e-09, ..., 9.6392e-08, + 1.0710e-07, 8.0094e-08], + [ 6.2864e-08, 1.8626e-08, 1.6764e-08, ..., 7.7300e-08, + 1.6298e-08, 2.7940e-09], + [-2.1374e-07, -1.0245e-07, 3.8650e-08, ..., -2.9802e-08, + 5.7276e-08, 4.7032e-08]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0310, -0.0283, -0.0146, 0.0023, -0.0002, 0.0020, 0.0061, -0.0145, + 0.0172, -0.0370], device='cuda:0'), grad: tensor([ 3.1441e-06, 8.1182e-05, 4.2934e-07, 2.2585e-07, 3.7160e-07, + 3.3667e-07, -3.5353e-06, -8.9526e-05, 2.5481e-06, 4.7982e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 217.01, cls_loss 0.0010 cls_loss_mapping 0.0029 cls_loss_causal 0.5225 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.08 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0450, 0.0008, 0.1087, ..., 0.0083, -0.0701, -0.0636], + [-0.1420, -0.1355, -0.1576, ..., 0.0593, 0.0485, 0.3208], + [-0.1149, -0.1276, -0.1561, ..., -0.1654, 0.0252, -0.1637], + ..., + [-0.1612, -0.1768, 0.0195, ..., -0.1835, -0.0870, -0.2973], + [ 0.1014, 0.0367, -0.1118, ..., -0.1748, 0.0177, -0.1685], + [ 0.0940, 0.0758, -0.1043, ..., 0.0746, -0.1768, -0.1374]], + device='cuda:0'), grad: tensor([[ 3.1991e-07, 2.6543e-07, -2.7474e-08, ..., 2.3935e-07, + 4.5868e-07, 5.4948e-08], + [ 1.2573e-07, 1.5041e-07, 3.8650e-07, ..., 7.6042e-07, + 7.1106e-07, 4.8894e-08], + [-2.7614e-07, -3.4459e-08, 5.7276e-08, ..., 3.0315e-07, + -3.6974e-07, 2.9290e-07], + ..., + [ 1.0803e-07, 1.1316e-07, 8.3819e-09, ..., 1.0803e-07, + 1.3225e-07, 1.9092e-08], + [ 5.5041e-07, 5.9558e-07, 4.0978e-07, ..., 1.3737e-06, + 1.4491e-06, 4.2282e-07], + [ 3.6545e-06, 4.4033e-06, 3.2131e-08, ..., 3.1907e-06, + 2.6524e-06, 7.4506e-09]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0309, -0.0288, -0.0147, 0.0022, 0.0006, 0.0020, 0.0057, -0.0144, + 0.0168, -0.0367], device='cuda:0'), grad: tensor([ 1.8384e-06, 5.4911e-06, -1.3988e-06, -1.5117e-05, 2.9653e-06, + 9.6187e-06, -2.0742e-05, -5.3551e-07, 7.6592e-06, 1.0170e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 217.01, cls_loss 0.0011 cls_loss_mapping 0.0035 cls_loss_causal 0.5005 re_mapping 0.0046 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0450, 0.0009, 0.1087, ..., 0.0083, -0.0705, -0.0639], + [-0.1427, -0.1366, -0.1579, ..., 0.0596, 0.0487, 0.3213], + [-0.1154, -0.1281, -0.1565, ..., -0.1664, 0.0253, -0.1640], + ..., + [-0.1619, -0.1771, 0.0204, ..., -0.1838, -0.0874, -0.2985], + [ 0.1017, 0.0368, -0.1120, ..., -0.1754, 0.0184, -0.1687], + [ 0.0935, 0.0754, -0.1046, ..., 0.0746, -0.1796, -0.1391]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, -1.2107e-07, -3.7206e-07, ..., -3.2969e-07, + 4.0047e-08, 4.6566e-10], + [ 1.1036e-07, 8.1491e-08, 1.7229e-08, ..., 4.0978e-08, + 6.7521e-08, -4.0513e-08], + [ 1.9791e-07, 1.4948e-07, 1.9558e-08, ..., 6.6590e-08, + -1.5507e-07, 7.4506e-09], + ..., + [ 5.5134e-07, 2.1327e-07, 1.1176e-08, ..., 3.1525e-07, + 1.2945e-07, 8.8476e-09], + [-9.2667e-08, -6.9384e-08, 4.1910e-09, ..., 5.8208e-08, + -1.2200e-07, 3.2596e-09], + [-6.6217e-07, -1.4808e-07, 4.7963e-08, ..., -3.1479e-07, + 6.9384e-08, 2.7940e-09]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0309, -0.0289, -0.0147, 0.0021, 0.0012, 0.0021, 0.0061, -0.0138, + 0.0167, -0.0379], device='cuda:0'), grad: tensor([-4.1630e-07, 7.4925e-07, -8.6147e-08, -1.1893e-06, 5.6019e-07, + 1.0561e-06, 5.1269e-07, 4.1910e-09, 1.3085e-07, -1.2927e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 217.21, cls_loss 0.0009 cls_loss_mapping 0.0034 cls_loss_causal 0.4953 re_mapping 0.0044 re_causal 0.0129 /// teacc 99.09 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0452, 0.0009, 0.1089, ..., 0.0084, -0.0706, -0.0641], + [-0.1434, -0.1376, -0.1588, ..., 0.0596, 0.0483, 0.3217], + [-0.1159, -0.1288, -0.1571, ..., -0.1672, 0.0253, -0.1641], + ..., + [-0.1624, -0.1778, 0.0202, ..., -0.1842, -0.0870, -0.3001], + [ 0.1020, 0.0362, -0.1131, ..., -0.1765, 0.0192, -0.1700], + [ 0.0937, 0.0753, -0.1047, ..., 0.0747, -0.1804, -0.1404]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, 6.0536e-08, 6.0536e-09, ..., 1.1642e-08, + 1.1642e-08, 6.5193e-09], + [-4.6100e-08, 5.5879e-08, 8.3819e-09, ..., -3.0734e-07, + -3.5996e-07, -5.0850e-07], + [ 5.4017e-08, 5.5414e-08, 1.9092e-08, ..., 8.2422e-08, + 5.9605e-08, -1.2107e-08], + ..., + [ 6.2399e-08, 9.7789e-08, 1.8626e-09, ..., 1.8440e-07, + 2.5425e-07, 3.2876e-07], + [ 4.6566e-10, 2.0955e-07, 6.9849e-09, ..., 2.5611e-08, + -9.3132e-10, 2.7940e-09], + [ 7.5437e-08, 2.0443e-07, 9.3132e-10, ..., 9.5461e-08, + 1.0198e-07, 1.1036e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0308, -0.0294, -0.0149, 0.0020, 0.0013, 0.0021, 0.0067, -0.0133, + 0.0164, -0.0382], device='cuda:0'), grad: tensor([ 2.9290e-07, -9.7416e-07, -5.3225e-07, 4.9174e-06, 1.7788e-07, + -7.3351e-06, 4.8848e-07, 1.1623e-06, 6.1095e-07, 1.1846e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 216.99, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5110 re_mapping 0.0045 re_causal 0.0131 /// teacc 98.96 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0457, 0.0010, 0.1091, ..., 0.0085, -0.0709, -0.0646], + [-0.1446, -0.1392, -0.1595, ..., 0.0599, 0.0486, 0.3222], + [-0.1168, -0.1299, -0.1576, ..., -0.1679, 0.0252, -0.1644], + ..., + [-0.1627, -0.1782, 0.0201, ..., -0.1845, -0.0876, -0.3014], + [ 0.1036, 0.0372, -0.1135, ..., -0.1773, 0.0201, -0.1706], + [ 0.0960, 0.0760, -0.1045, ..., 0.0750, -0.1795, -0.1417]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.5856e-08, 1.8580e-07, ..., 1.5553e-07, + 2.3283e-08, 1.8626e-09], + [ 4.1910e-09, 1.1642e-08, 5.4948e-08, ..., -7.9162e-09, + -2.6543e-08, -1.3458e-07], + [ 3.7253e-09, 8.8476e-09, 2.1234e-07, ..., 1.7229e-07, + 3.9581e-08, 2.0023e-08], + ..., + [ 4.6566e-09, 1.7695e-08, 1.2573e-08, ..., 2.3283e-08, + 3.4925e-08, 5.5879e-08], + [-1.8626e-08, 1.7136e-07, 3.2131e-08, ..., 4.2841e-08, + -7.9162e-09, 1.3970e-09], + [-4.1910e-09, 5.6345e-08, 9.9652e-08, ..., 1.4901e-08, + 1.1548e-07, 1.0710e-08]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0307, -0.0293, -0.0149, 0.0023, 0.0002, 0.0018, 0.0065, -0.0133, + 0.0171, -0.0377], device='cuda:0'), grad: tensor([ 7.6974e-07, 5.4482e-08, 5.4343e-07, 1.4585e-06, 4.3865e-07, + -1.8207e-06, -2.7455e-06, 4.4517e-07, 3.9348e-07, 4.7917e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 217.21, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.5015 re_mapping 0.0042 re_causal 0.0131 /// teacc 99.01 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0458, 0.0012, 0.1092, ..., 0.0086, -0.0712, -0.0648], + [-0.1439, -0.1401, -0.1598, ..., 0.0613, 0.0496, 0.3228], + [-0.1177, -0.1317, -0.1583, ..., -0.1685, 0.0251, -0.1646], + ..., + [-0.1628, -0.1784, 0.0196, ..., -0.1859, -0.0892, -0.3031], + [ 0.1044, 0.0373, -0.1151, ..., -0.1780, 0.0202, -0.1707], + [ 0.0962, 0.0759, -0.1046, ..., 0.0750, -0.1798, -0.1428]], + device='cuda:0'), grad: tensor([[ 2.6124e-07, 4.9360e-08, -2.4168e-07, ..., -3.3528e-08, + 3.5856e-07, 1.0384e-07], + [ 2.1234e-07, 1.3085e-07, 3.4459e-08, ..., -3.1572e-06, + -2.9001e-06, -2.0117e-06], + [ 9.0804e-08, 8.8476e-08, 4.9826e-08, ..., 1.3951e-06, + 1.2657e-06, 7.9628e-07], + ..., + [ 2.2585e-07, 1.1222e-07, 9.7789e-09, ..., 7.1758e-07, + 4.1956e-07, 2.9616e-07], + [-1.2573e-06, -9.9652e-07, -5.7276e-08, ..., 2.5844e-07, + -8.4378e-07, 1.2666e-07], + [-9.5740e-07, -1.9185e-07, 1.6578e-07, ..., -1.0878e-06, + 2.1001e-07, 6.3796e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-3.0532e-02, -2.8676e-02, -1.5140e-02, 2.1309e-03, 2.4091e-06, + 1.9951e-03, 6.6160e-03, -1.3774e-02, 1.6945e-02, -3.7748e-02], + device='cuda:0'), grad: tensor([ 6.1514e-07, -9.3952e-06, 4.4629e-06, 2.2650e-06, 4.1574e-06, + -2.9709e-07, 1.1045e-06, 2.1514e-06, -2.8554e-06, -2.2296e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 217.20, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5146 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.05 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0461, 0.0015, 0.1093, ..., 0.0087, -0.0716, -0.0650], + [-0.1455, -0.1417, -0.1606, ..., 0.0615, 0.0498, 0.3231], + [-0.1187, -0.1289, -0.1590, ..., -0.1692, 0.0294, -0.1648], + ..., + [-0.1638, -0.1791, 0.0184, ..., -0.1865, -0.0905, -0.3044], + [ 0.1054, 0.0338, -0.1162, ..., -0.1790, 0.0157, -0.1711], + [ 0.0967, 0.0760, -0.1047, ..., 0.0751, -0.1808, -0.1437]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, -4.9826e-08, -4.6566e-07, ..., -1.4668e-07, + 7.9162e-09, 7.9162e-09], + [ 1.3504e-08, 1.3039e-08, 1.4435e-08, ..., -1.7788e-07, + -2.2212e-07, -6.1467e-07], + [ 2.2817e-08, 2.7474e-08, 4.9360e-08, ..., 1.1176e-07, + 1.2061e-07, 2.6496e-07], + ..., + [ 4.3306e-08, 4.9826e-08, 2.8266e-07, ..., 1.6950e-07, + 5.8208e-08, 1.3923e-07], + [-4.5635e-08, -5.7742e-08, 2.0023e-08, ..., 1.9092e-08, + -3.2596e-08, 6.5193e-09], + [-4.6566e-10, -9.7789e-09, 2.7753e-07, ..., -2.8871e-08, + 2.6356e-07, 2.7474e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0305, -0.0288, -0.0109, 0.0020, 0.0008, 0.0022, 0.0064, -0.0140, + 0.0122, -0.0380], device='cuda:0'), grad: tensor([-1.3430e-06, -9.2527e-07, 6.3982e-07, 4.4703e-08, -4.9733e-07, + 1.1176e-07, -1.1409e-07, 1.0952e-06, -1.3364e-07, 1.1269e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 217.43, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.5199 re_mapping 0.0043 re_causal 0.0130 /// teacc 99.09 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0457, 0.0016, 0.1095, ..., 0.0088, -0.0718, -0.0654], + [-0.1463, -0.1429, -0.1609, ..., 0.0619, 0.0500, 0.3237], + [-0.1218, -0.1291, -0.1597, ..., -0.1704, 0.0295, -0.1652], + ..., + [-0.1642, -0.1798, 0.0182, ..., -0.1870, -0.0911, -0.3058], + [ 0.1074, 0.0341, -0.1166, ..., -0.1797, 0.0156, -0.1712], + [ 0.0971, 0.0761, -0.1048, ..., 0.0752, -0.1815, -0.1451]], + device='cuda:0'), grad: tensor([[ 4.1910e-08, 1.5181e-07, 6.5938e-07, ..., 3.9255e-07, + 3.4925e-08, 3.2596e-08], + [-9.2667e-08, -3.7765e-07, 1.3504e-08, ..., -9.9838e-07, + -3.9674e-07, -2.4121e-06], + [ 4.1537e-07, 3.5809e-07, 6.0536e-09, ..., 1.7835e-07, + 3.5996e-07, 8.0559e-08], + ..., + [ 1.2992e-07, 2.6915e-07, 8.3819e-09, ..., 4.9034e-07, + 2.3469e-07, 1.0524e-06], + [ 7.7020e-07, 1.2368e-06, 5.8673e-07, ..., 4.4843e-07, + 4.8755e-07, 9.7789e-08], + [ 2.8079e-07, 9.7603e-07, 6.4587e-07, ..., 5.3318e-07, + 1.8394e-07, 7.4878e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0303, -0.0288, -0.0109, 0.0020, 0.0007, 0.0022, 0.0061, -0.0140, + 0.0122, -0.0380], device='cuda:0'), grad: tensor([ 1.4715e-06, -1.1578e-05, 1.2759e-06, -2.1979e-06, 1.7462e-07, + -2.5164e-06, -1.0170e-06, 5.6475e-06, 3.2336e-06, 5.4911e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 217.34, cls_loss 0.0010 cls_loss_mapping 0.0030 cls_loss_causal 0.5371 re_mapping 0.0041 re_causal 0.0133 /// teacc 98.97 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.0460, 0.0016, 0.1095, ..., 0.0088, -0.0721, -0.0656], + [-0.1463, -0.1437, -0.1610, ..., 0.0622, 0.0502, 0.3243], + [-0.1223, -0.1292, -0.1599, ..., -0.1708, 0.0295, -0.1654], + ..., + [-0.1648, -0.1805, 0.0179, ..., -0.1872, -0.0912, -0.3070], + [ 0.1077, 0.0343, -0.1170, ..., -0.1813, 0.0156, -0.1714], + [ 0.0985, 0.0766, -0.1048, ..., 0.0754, -0.1816, -0.1491]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -3.5390e-08, -2.7940e-08, ..., -9.3132e-09, + 1.0710e-08, 1.3970e-09], + [ 5.0291e-08, 1.9558e-08, 1.9092e-08, ..., 8.3819e-09, + 6.0536e-09, -1.2433e-07], + [ 1.8626e-08, 1.5832e-08, 1.8626e-08, ..., 2.1886e-08, + 3.9116e-08, 4.1910e-08], + ..., + [ 1.6987e-06, 3.8836e-07, 2.0023e-08, ..., 1.5507e-06, + 4.1444e-08, 6.0070e-08], + [-1.8161e-08, -1.8161e-08, 5.5879e-09, ..., 5.5879e-09, + -2.5611e-08, -7.4506e-09], + [-1.8133e-06, -3.8277e-07, 8.6613e-08, ..., -1.6298e-06, + 1.2247e-07, 8.8476e-09]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0303, -0.0297, -0.0109, 0.0013, 0.0003, 0.0028, 0.0060, -0.0130, + 0.0123, -0.0378], device='cuda:0'), grad: tensor([ 6.8964e-07, 7.7039e-06, 5.8673e-06, 9.2536e-06, -1.4389e-07, + 1.3877e-07, 7.9162e-08, -2.1711e-05, -5.8208e-08, -1.8254e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 217.13, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.5015 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.05 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.0463, 0.0016, 0.1097, ..., 0.0088, -0.0717, -0.0652], + [-0.1466, -0.1448, -0.1614, ..., 0.0621, 0.0502, 0.3247], + [-0.1233, -0.1293, -0.1611, ..., -0.1724, 0.0295, -0.1659], + ..., + [-0.1650, -0.1809, 0.0175, ..., -0.1875, -0.0915, -0.3078], + [ 0.1090, 0.0347, -0.1180, ..., -0.1819, 0.0157, -0.1715], + [ 0.0991, 0.0769, -0.1049, ..., 0.0755, -0.1819, -0.1499]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -4.2422e-07, -5.5507e-07, ..., -1.7043e-07, + 1.3039e-08, 4.1910e-09], + [ 9.7789e-09, 7.6834e-08, 1.9511e-07, ..., -3.6322e-08, + 6.6590e-08, -2.4308e-07], + [ 1.5367e-08, 5.7742e-08, 1.1362e-07, ..., 3.6787e-08, + 8.1025e-08, 3.5390e-08], + ..., + [ 5.5879e-09, 4.0978e-08, 2.0955e-08, ..., 6.1933e-08, + 6.6124e-08, 1.4715e-07], + [-4.4703e-08, 1.1409e-07, 4.2189e-07, ..., 2.2631e-07, + -5.1223e-08, 6.5193e-09], + [ 8.3819e-09, 9.6858e-08, 1.5181e-07, ..., 5.3085e-08, + 3.7253e-08, 1.6298e-08]], device='cuda:0') +Epoch 210, bias, value: tensor([-3.0341e-02, -3.0034e-02, -1.0942e-02, 4.1137e-04, 7.7389e-05, + 3.6310e-03, 6.2156e-03, -1.2507e-02, 1.2375e-02, -3.7897e-02], + device='cuda:0'), grad: tensor([-1.0757e-06, 3.6368e-07, 2.9849e-07, 1.1036e-07, -5.1176e-07, + 2.1281e-07, -5.2666e-07, -3.9116e-07, 7.5856e-07, 7.8138e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 217.37, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5222 re_mapping 0.0043 re_causal 0.0133 /// teacc 99.10 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0467, 0.0016, 0.1097, ..., 0.0088, -0.0721, -0.0655], + [-0.1472, -0.1456, -0.1625, ..., 0.0637, 0.0505, 0.3258], + [-0.1239, -0.1294, -0.1615, ..., -0.1731, 0.0295, -0.1663], + ..., + [-0.1655, -0.1817, 0.0185, ..., -0.1893, -0.0917, -0.3106], + [ 0.1095, 0.0348, -0.1192, ..., -0.1831, 0.0157, -0.1716], + [ 0.0997, 0.0772, -0.1049, ..., 0.0756, -0.1823, -0.1526]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -7.8231e-08, -7.4040e-08, ..., -9.7789e-08, + 1.3970e-09, 0.0000e+00], + [ 2.7940e-09, 5.1223e-09, 3.7253e-09, ..., 5.5879e-09, + 2.7940e-09, 0.0000e+00], + [ 6.5193e-09, 8.3819e-09, 3.2596e-09, ..., 4.6566e-09, + -1.3970e-09, 0.0000e+00], + ..., + [ 4.6566e-09, 1.3504e-08, 2.7940e-09, ..., 6.5193e-09, + 4.1910e-09, 0.0000e+00], + [-5.8673e-08, -5.0757e-08, 6.5193e-09, ..., 1.1642e-08, + -2.5611e-08, 0.0000e+00], + [-2.7940e-09, 6.2864e-08, 5.8208e-08, ..., 6.2399e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0304, -0.0292, -0.0110, 0.0003, -0.0004, 0.0037, 0.0071, -0.0132, + 0.0124, -0.0378], device='cuda:0'), grad: tensor([-2.1048e-07, 4.2515e-07, -6.3749e-07, 1.1176e-07, 2.4680e-08, + 2.7418e-06, -2.7083e-06, 1.2340e-07, -1.2759e-07, 2.6403e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 217.04, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.5063 re_mapping 0.0042 re_causal 0.0130 /// teacc 99.12 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0469, 0.0018, 0.1098, ..., 0.0088, -0.0724, -0.0655], + [-0.1480, -0.1463, -0.1628, ..., 0.0639, 0.0505, 0.3264], + [-0.1247, -0.1294, -0.1619, ..., -0.1735, 0.0295, -0.1665], + ..., + [-0.1655, -0.1821, 0.0184, ..., -0.1896, -0.0918, -0.3130], + [ 0.1100, 0.0348, -0.1196, ..., -0.1839, 0.0157, -0.1718], + [ 0.1002, 0.0773, -0.1050, ..., 0.0757, -0.1826, -0.1533]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -2.0023e-08, -2.7940e-08, ..., -1.6298e-08, + 6.5193e-09, 2.7940e-09], + [ 2.2817e-08, 5.5879e-09, 4.1910e-09, ..., -5.7742e-08, + -6.2399e-08, -2.6636e-07], + [ 9.3132e-09, 9.7789e-09, 3.2596e-09, ..., 4.1444e-08, + 2.5146e-08, 1.5879e-07], + ..., + [-1.3039e-08, 1.0245e-08, 4.1910e-09, ..., 3.4925e-08, + 3.8650e-08, 6.7521e-08], + [-1.2573e-08, -1.7695e-08, 3.3062e-08, ..., 3.7253e-08, + -6.9849e-09, 2.7940e-09], + [-7.4506e-09, 1.0710e-08, 7.4040e-08, ..., -6.9849e-09, + 6.7521e-08, 9.3132e-09]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0304, -0.0291, -0.0112, 0.0002, -0.0005, 0.0037, 0.0071, -0.0127, + 0.0124, -0.0379], device='cuda:0'), grad: tensor([-2.7008e-08, -1.7229e-07, 6.6124e-08, -2.5146e-08, -1.0524e-07, + 4.4238e-08, -1.2107e-07, 2.7940e-08, 7.1712e-08, 2.5565e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 217.39, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4816 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.10 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0464, 0.0023, 0.1100, ..., 0.0089, -0.0731, -0.0657], + [-0.1494, -0.1473, -0.1632, ..., 0.0638, 0.0504, 0.3269], + [-0.1252, -0.1295, -0.1630, ..., -0.1743, 0.0295, -0.1670], + ..., + [-0.1658, -0.1826, 0.0182, ..., -0.1897, -0.0920, -0.3135], + [ 0.1106, 0.0350, -0.1185, ..., -0.1839, 0.0158, -0.1719], + [ 0.1006, 0.0773, -0.1051, ..., 0.0759, -0.1830, -0.1538]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -7.0315e-08, -9.2667e-08, ..., -1.0617e-07, + 9.3132e-09, 3.7253e-09], + [ 6.5193e-09, 2.5146e-08, 3.3528e-08, ..., -1.3039e-08, + -1.0990e-07, -3.2922e-07], + [ 9.7789e-09, 1.0245e-08, 7.4506e-09, ..., 2.1886e-08, + 2.5146e-08, 4.4238e-08], + ..., + [ 1.6764e-08, 1.3504e-08, 8.8476e-09, ..., 3.5390e-08, + 5.1223e-08, 9.9186e-08], + [ 2.3283e-09, 5.5879e-09, 7.9162e-09, ..., 1.4435e-08, + 1.3970e-08, 2.6077e-08], + [-3.3993e-08, -6.5193e-09, 6.3330e-08, ..., -1.3039e-08, + 7.6834e-08, 1.4435e-08]], device='cuda:0') +Epoch 213, bias, value: tensor([-3.0219e-02, -2.9289e-02, -1.1152e-02, 8.2337e-05, -6.4182e-04, + 3.6797e-03, 6.8922e-03, -1.2694e-02, 1.2478e-02, -3.7782e-02], + device='cuda:0'), grad: tensor([-2.8685e-07, 2.4401e-06, 1.9744e-07, 4.7963e-08, 1.1232e-06, + 4.2841e-08, 1.5507e-07, -4.3996e-06, 1.0151e-07, 5.7928e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 217.27, cls_loss 0.0009 cls_loss_mapping 0.0036 cls_loss_causal 0.4988 re_mapping 0.0046 re_causal 0.0134 /// teacc 99.11 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0466, 0.0024, 0.1099, ..., 0.0088, -0.0739, -0.0655], + [-0.1518, -0.1496, -0.1637, ..., 0.0635, 0.0504, 0.3277], + [-0.1260, -0.1296, -0.1634, ..., -0.1749, 0.0295, -0.1675], + ..., + [-0.1659, -0.1831, 0.0179, ..., -0.1898, -0.0924, -0.3149], + [ 0.1113, 0.0352, -0.1186, ..., -0.1843, 0.0158, -0.1720], + [ 0.1020, 0.0782, -0.1050, ..., 0.0763, -0.1838, -0.1560]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -3.4925e-08, -6.0536e-08, ..., -5.8673e-08, + 2.7474e-08, 8.8476e-09], + [ 6.5193e-09, 2.5146e-08, 1.4901e-08, ..., -4.8196e-07, + 8.8476e-09, -5.4808e-07], + [ 1.1176e-08, -4.1956e-07, 5.3085e-08, ..., 3.0734e-08, + -8.4518e-07, 2.0955e-08], + ..., + [ 3.1199e-08, 1.5181e-07, 3.0966e-07, ..., 2.9523e-07, + 4.4517e-07, 2.9616e-07], + [-1.3504e-08, 1.3970e-08, 2.4214e-08, ..., 6.9849e-09, + 5.8673e-08, 4.1910e-09], + [-5.4017e-08, 1.3504e-08, 2.6356e-07, ..., 1.4575e-07, + 1.6438e-07, 1.8114e-07]], device='cuda:0') +Epoch 214, bias, value: tensor([-3.0328e-02, -2.9536e-02, -1.1143e-02, 1.9435e-05, -1.6688e-03, + 3.6965e-03, 7.3846e-03, -1.2631e-02, 1.2550e-02, -3.7208e-02], + device='cuda:0'), grad: tensor([ 1.8626e-09, -1.2247e-06, -4.2841e-06, 9.4483e-07, -1.5860e-06, + 1.3895e-06, 3.4878e-07, 3.0212e-06, 3.0082e-07, 1.0841e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 217.35, cls_loss 0.0009 cls_loss_mapping 0.0031 cls_loss_causal 0.5230 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.04 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0467, 0.0025, 0.1106, ..., 0.0092, -0.0716, -0.0624], + [-0.1528, -0.1511, -0.1673, ..., 0.0627, 0.0498, 0.3275], + [-0.1277, -0.1298, -0.1643, ..., -0.1765, 0.0295, -0.1684], + ..., + [-0.1663, -0.1835, 0.0177, ..., -0.1900, -0.0927, -0.3164], + [ 0.1134, 0.0356, -0.1191, ..., -0.1850, 0.0159, -0.1723], + [ 0.1022, 0.0773, -0.1051, ..., 0.0764, -0.1843, -0.1574]], + device='cuda:0'), grad: tensor([[ 3.0268e-08, -6.2864e-08, 2.7940e-09, ..., 4.6566e-10, + 1.9092e-08, 4.1910e-08], + [ 1.4622e-07, 4.3772e-08, 4.8429e-08, ..., -3.8603e-07, + 4.9826e-08, -9.2154e-07], + [ 1.1530e-06, 7.2317e-07, 5.9139e-08, ..., 1.7881e-07, + 7.1339e-07, 1.2107e-07], + ..., + [ 1.7649e-07, 1.9139e-07, -4.3167e-07, ..., 2.6403e-07, + 9.9186e-08, 5.0943e-07], + [-1.1176e-08, -2.7008e-08, 2.5146e-08, ..., 2.7940e-08, + -1.9092e-08, 2.6543e-08], + [-9.4529e-08, -4.0047e-08, 3.1618e-07, ..., -4.8429e-08, + 5.6811e-08, 9.6858e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0298, -0.0302, -0.0112, 0.0001, -0.0018, 0.0036, 0.0075, -0.0125, + 0.0127, -0.0375], device='cuda:0'), grad: tensor([ 6.6869e-07, -7.0222e-07, 4.7348e-06, -6.3777e-06, 6.0070e-07, + 7.4739e-07, 1.2107e-08, -2.0023e-06, 1.5227e-07, 2.1495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 217.23, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4873 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.11 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.0468, 0.0026, 0.1108, ..., 0.0093, -0.0715, -0.0623], + [-0.1534, -0.1519, -0.1680, ..., 0.0626, 0.0497, 0.3279], + [-0.1290, -0.1299, -0.1648, ..., -0.1779, 0.0295, -0.1690], + ..., + [-0.1667, -0.1840, 0.0177, ..., -0.1903, -0.0928, -0.3173], + [ 0.1142, 0.0355, -0.1223, ..., -0.1880, 0.0159, -0.1724], + [ 0.1025, 0.0773, -0.1053, ..., 0.0764, -0.1846, -0.1588]], + device='cuda:0'), grad: tensor([[ 1.3923e-07, 1.7462e-07, -3.4571e-06, ..., -6.7195e-07, + 4.1910e-09, 0.0000e+00], + [ 3.3528e-08, 6.2864e-08, 4.1910e-08, ..., 2.1886e-08, + 3.7253e-09, 0.0000e+00], + [ 1.9837e-07, 1.9418e-07, 4.2282e-07, ..., 9.6858e-08, + 1.0338e-07, 0.0000e+00], + ..., + [ 1.2247e-07, 1.7276e-07, 4.0513e-08, ..., 6.8452e-08, + 1.7695e-08, 0.0000e+00], + [-2.1700e-07, -8.8010e-08, 7.3249e-07, ..., 2.7241e-07, + -1.8394e-07, 0.0000e+00], + [-9.1596e-07, -9.7789e-07, 4.1956e-07, ..., -5.8115e-07, + -1.2107e-08, 0.0000e+00]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0297, -0.0305, -0.0111, 0.0002, -0.0019, 0.0034, 0.0091, -0.0122, + 0.0126, -0.0377], device='cuda:0'), grad: tensor([-4.9472e-06, 2.1141e-07, 1.3877e-06, -4.0710e-05, 7.8324e-07, + 4.1783e-05, 2.2128e-06, 4.9826e-07, 2.4633e-07, -1.5981e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 217.22, cls_loss 0.0009 cls_loss_mapping 0.0030 cls_loss_causal 0.4842 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.09 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0470, 0.0026, 0.1108, ..., 0.0093, -0.0717, -0.0624], + [-0.1542, -0.1527, -0.1682, ..., 0.0628, 0.0500, 0.3294], + [-0.1281, -0.1287, -0.1617, ..., -0.1796, 0.0302, -0.1713], + ..., + [-0.1674, -0.1856, 0.0179, ..., -0.1905, -0.0930, -0.3184], + [ 0.1145, 0.0345, -0.1254, ..., -0.1883, 0.0153, -0.1725], + [ 0.1027, 0.0771, -0.1056, ..., 0.0766, -0.1853, -0.1600]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, -1.8626e-08, 8.2888e-07, ..., 2.2445e-07, + 2.7940e-09, 1.1176e-08], + [ 2.4214e-08, 1.4435e-08, 2.5705e-07, ..., -4.6659e-07, + -1.4156e-07, -1.0803e-06], + [ 7.1712e-08, 7.7765e-08, 1.1045e-06, ..., 4.1304e-07, + 4.7963e-08, 1.7695e-07], + ..., + [ 1.5786e-07, 7.6368e-08, 1.1642e-07, ..., 4.8941e-07, + 8.1956e-08, 6.4261e-07], + [-7.8697e-08, -4.1910e-08, 5.4482e-07, ..., 1.7881e-07, + -3.0268e-08, 3.1199e-08], + [-4.4284e-07, -1.3830e-07, -2.7940e-08, ..., -2.6729e-07, + 1.9092e-08, 1.4761e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0297, -0.0305, -0.0103, 0.0002, -0.0017, 0.0035, 0.0094, -0.0120, + 0.0116, -0.0381], device='cuda:0'), grad: tensor([ 2.4103e-06, -1.6503e-06, 3.9525e-06, 2.9430e-07, 2.9355e-06, + 9.6112e-06, -1.9938e-05, 1.8626e-06, 1.2927e-06, -7.4692e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 217.25, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.5116 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.11 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0471, 0.0032, 0.1111, ..., 0.0094, -0.0718, -0.0624], + [-0.1543, -0.1538, -0.1684, ..., 0.0631, 0.0501, 0.3301], + [-0.1291, -0.1288, -0.1619, ..., -0.1811, 0.0301, -0.1722], + ..., + [-0.1677, -0.1861, 0.0171, ..., -0.1908, -0.0932, -0.3191], + [ 0.1145, 0.0343, -0.1258, ..., -0.1892, 0.0153, -0.1728], + [ 0.1031, 0.0770, -0.1059, ..., 0.0766, -0.1858, -0.1606]], + device='cuda:0'), grad: tensor([[ 1.1409e-07, 8.6473e-07, 4.8755e-07, ..., 1.9092e-08, + 5.1223e-08, 7.4506e-09], + [ 2.1374e-07, 3.0622e-06, 3.7719e-08, ..., 1.3039e-08, + 6.7987e-08, -9.9186e-08], + [ 4.1677e-07, 1.7872e-06, 7.8697e-07, ..., 1.0943e-07, + 3.1479e-07, 1.9558e-08], + ..., + [ 5.5879e-08, -5.0925e-06, 7.9628e-08, ..., 9.4995e-08, + 1.1316e-07, 1.1642e-08], + [-1.7975e-07, 6.4541e-07, 2.5844e-07, ..., 7.3109e-08, + 4.7032e-08, 2.6543e-08], + [ 4.6566e-08, 4.2329e-07, 4.8429e-08, ..., -5.4482e-08, + 4.5169e-08, 1.8626e-09]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0294, -0.0304, -0.0103, 0.0002, -0.0013, 0.0035, 0.0097, -0.0121, + 0.0116, -0.0384], device='cuda:0'), grad: tensor([ 2.0415e-06, 2.2233e-05, 4.9248e-06, 1.6510e-05, 2.4727e-07, + -1.1429e-05, 4.5123e-07, -4.0293e-05, 2.8182e-06, 2.5183e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 217.34, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4770 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.14 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0474, 0.0032, 0.1112, ..., 0.0094, -0.0721, -0.0624], + [-0.1551, -0.1548, -0.1687, ..., 0.0636, 0.0503, 0.3306], + [-0.1306, -0.1290, -0.1620, ..., -0.1821, 0.0301, -0.1726], + ..., + [-0.1685, -0.1867, 0.0170, ..., -0.1916, -0.0938, -0.3197], + [ 0.1179, 0.0347, -0.1258, ..., -0.1895, 0.0153, -0.1730], + [ 0.1027, 0.0767, -0.1066, ..., 0.0765, -0.1871, -0.1616]], + device='cuda:0'), grad: tensor([[ 7.7765e-08, -4.9826e-07, -1.6876e-06, ..., -1.0552e-06, + 3.1665e-08, -3.7765e-07], + [ 2.2352e-08, 3.5856e-08, 6.2864e-08, ..., 4.8429e-08, + 9.3132e-09, 7.9162e-09], + [ 7.9162e-08, 5.3085e-08, 9.6392e-08, ..., 4.6100e-08, + 6.9384e-08, 1.3504e-08], + ..., + [ 1.2945e-07, 6.1933e-08, 1.0710e-08, ..., 9.5461e-08, + 5.5879e-09, 2.3283e-09], + [ 1.8999e-07, 1.4389e-07, 3.0734e-08, ..., 1.3551e-07, + -5.1688e-08, 4.6566e-09], + [-6.9616e-07, -2.2817e-08, 2.2072e-07, ..., -2.4308e-07, + 5.0291e-08, 3.6787e-08]], device='cuda:0') +Epoch 219, bias, value: tensor([-2.9466e-02, -3.0425e-02, -1.0354e-02, 7.6214e-05, -5.0525e-04, + 3.5466e-03, 9.8483e-03, -1.2111e-02, 1.1776e-02, -3.9061e-02], + device='cuda:0'), grad: tensor([-3.2913e-06, 3.1758e-07, 8.1863e-07, 3.3481e-07, -1.2107e-08, + -1.5628e-06, 3.5763e-06, -2.0349e-07, 7.6508e-07, -7.5903e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 217.20, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5125 re_mapping 0.0040 re_causal 0.0129 /// teacc 99.03 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.0479, 0.0034, 0.1114, ..., 0.0095, -0.0722, -0.0622], + [-0.1561, -0.1558, -0.1688, ..., 0.0637, 0.0503, 0.3310], + [-0.1322, -0.1291, -0.1622, ..., -0.1829, 0.0301, -0.1730], + ..., + [-0.1692, -0.1869, 0.0167, ..., -0.1918, -0.0939, -0.3204], + [ 0.1191, 0.0357, -0.1257, ..., -0.1879, 0.0154, -0.1731], + [ 0.1036, 0.0769, -0.1066, ..., 0.0766, -0.1873, -0.1625]], + device='cuda:0'), grad: tensor([[ 2.3330e-07, 1.0803e-07, 3.8650e-08, ..., 2.8871e-08, + 4.5635e-08, 9.3132e-10], + [ 5.1688e-08, 3.3993e-08, 2.1420e-08, ..., 1.1642e-08, + 2.1886e-08, -3.0268e-08], + [ 1.4203e-07, 8.9407e-08, 4.4703e-08, ..., 2.8871e-08, + 5.9139e-08, 4.1910e-09], + ..., + [ 2.6077e-08, 4.1444e-08, 1.8626e-08, ..., 4.3306e-08, + 2.6077e-08, 1.3504e-08], + [-5.6252e-07, -3.3481e-07, -1.4482e-07, ..., -3.3993e-08, + -2.0256e-07, 1.8626e-09], + [ 3.8650e-08, 2.6543e-08, 4.2729e-06, ..., -7.1712e-08, + 4.0978e-08, 2.7940e-09]], device='cuda:0') +Epoch 220, bias, value: tensor([-2.9374e-02, -3.0543e-02, -1.0421e-02, 4.3541e-05, -5.3197e-04, + 3.4600e-03, 8.1794e-03, -1.2119e-02, 1.2232e-02, -3.8959e-02], + device='cuda:0'), grad: tensor([ 6.6543e-07, 6.2911e-07, -1.0207e-06, -2.1467e-07, -1.1444e-05, + 8.5682e-07, 1.8300e-07, 9.2853e-07, -2.4736e-06, 1.1913e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 217.28, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4766 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.0485, 0.0032, 0.1114, ..., 0.0094, -0.0724, -0.0622], + [-0.1570, -0.1574, -0.1689, ..., 0.0641, 0.0502, 0.3315], + [-0.1333, -0.1292, -0.1622, ..., -0.1837, 0.0300, -0.1733], + ..., + [-0.1698, -0.1874, 0.0164, ..., -0.1925, -0.0939, -0.3215], + [ 0.1211, 0.0358, -0.1258, ..., -0.1874, 0.0155, -0.1726], + [ 0.1041, 0.0773, -0.1066, ..., 0.0769, -0.1876, -0.1640]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, -3.5856e-08, -2.6852e-05, ..., -2.7269e-05, + 1.5134e-08, 9.7789e-09], + [ 3.9581e-09, 6.9849e-09, 9.8953e-08, ..., -1.6298e-09, + -6.1002e-08, -2.3982e-07], + [ 6.7521e-09, 7.9162e-09, 5.4762e-07, ..., 5.6624e-07, + 1.7695e-08, 4.0280e-08], + ..., + [ 1.1874e-08, 9.7789e-09, 1.3039e-08, ..., 5.1688e-08, + 1.6997e-08, 6.4261e-08], + [ 3.4925e-09, 7.5670e-08, 3.2131e-08, ..., 5.3784e-08, + 1.0245e-08, 3.8650e-08], + [-7.4739e-08, -5.3551e-09, 1.4948e-07, ..., 9.3132e-08, + 1.2107e-08, 3.4925e-08]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0295, -0.0303, -0.0105, -0.0001, -0.0009, 0.0036, 0.0084, -0.0123, + 0.0124, -0.0388], device='cuda:0'), grad: tensor([-6.8367e-05, -9.2667e-08, 1.6931e-06, 1.4203e-08, 2.6147e-07, + -3.7230e-07, 6.6519e-05, -1.6973e-07, 3.5111e-07, 2.6007e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 217.29, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.5069 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.08 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0482, 0.0035, 0.1118, ..., 0.0096, -0.0725, -0.0623], + [-0.1584, -0.1580, -0.1690, ..., 0.0643, 0.0505, 0.3328], + [-0.1342, -0.1293, -0.1623, ..., -0.1849, 0.0300, -0.1742], + ..., + [-0.1705, -0.1880, 0.0161, ..., -0.1928, -0.0941, -0.3226], + [ 0.1214, 0.0358, -0.1258, ..., -0.1882, 0.0155, -0.1740], + [ 0.1050, 0.0775, -0.1067, ..., 0.0771, -0.1883, -0.1657]], + device='cuda:0'), grad: tensor([[ 1.6913e-06, 6.1393e-06, 3.9972e-06, ..., 1.0356e-05, + 7.9162e-09, 7.4506e-09], + [ 3.2131e-08, 5.4250e-08, 5.0059e-08, ..., -1.1874e-08, + -6.4261e-08, -8.8243e-08], + [ 5.3318e-08, 8.5216e-08, 1.1688e-07, ..., 1.5320e-07, + 1.8859e-08, 1.7928e-08], + ..., + [ 5.0059e-08, 7.7300e-08, 7.5903e-08, ..., 1.4389e-07, + 2.4214e-08, 3.3062e-08], + [ 2.7544e-07, 1.8440e-07, 2.6077e-08, ..., 3.0105e-07, + 3.2596e-09, 6.0536e-09], + [-4.1388e-06, -8.2180e-06, -4.4517e-06, ..., -1.2964e-05, + 1.6345e-07, 7.4506e-09]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0293, -0.0304, -0.0105, 0.0003, -0.0008, 0.0033, 0.0077, -0.0121, + 0.0123, -0.0388], device='cuda:0'), grad: tensor([ 2.1785e-05, -1.1665e-07, 4.6683e-07, 4.1686e-06, 1.6252e-07, + 9.1502e-08, 7.1619e-07, 4.6846e-07, 5.9605e-07, -2.8297e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 217.16, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5118 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0473, 0.0042, 0.1123, ..., 0.0102, -0.0731, -0.0629], + [-0.1594, -0.1588, -0.1693, ..., 0.0651, 0.0484, 0.3353], + [-0.1322, -0.1294, -0.1624, ..., -0.1868, 0.0298, -0.1757], + ..., + [-0.1715, -0.1889, 0.0194, ..., -0.1932, -0.0909, -0.3240], + [ 0.1215, 0.0357, -0.1260, ..., -0.1889, 0.0154, -0.1753], + [ 0.1059, 0.0772, -0.1071, ..., 0.0771, -0.1887, -0.1673]], + device='cuda:0'), grad: tensor([[ 1.4296e-06, 1.7062e-06, 7.5996e-07, ..., 4.2431e-06, + 9.4296e-07, 3.3975e-06], + [ 1.4668e-08, 1.5134e-08, 1.4901e-08, ..., -1.3784e-05, + -4.1239e-06, -1.5378e-05], + [ 1.6764e-08, 2.1420e-08, 1.2107e-08, ..., 2.8964e-07, + -1.6298e-07, 3.1409e-07], + ..., + [ 5.9837e-08, 3.8184e-08, 5.2387e-08, ..., 2.0000e-07, + 1.7788e-07, 1.5274e-07], + [ 2.3022e-06, 3.0622e-06, 1.3858e-06, ..., 5.5060e-06, + 1.0692e-06, 3.7700e-06], + [-3.8296e-06, -4.7013e-06, -1.8254e-06, ..., -3.2783e-06, + 2.9989e-07, 1.5274e-07]], device='cuda:0') +Epoch 223, bias, value: tensor([-2.8865e-02, -3.1918e-02, -1.0795e-02, 5.1149e-05, -1.3629e-03, + 3.3954e-03, 6.9416e-03, -9.8029e-03, 1.2214e-02, -3.8911e-02], + device='cuda:0'), grad: tensor([ 1.3687e-05, -3.3975e-05, -1.5855e-05, 2.5034e-06, -1.2759e-06, + 1.1018e-06, 1.6838e-05, 6.3181e-06, 1.9774e-05, -9.0972e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 217.30, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.4891 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.07 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0472, 0.0045, 0.1125, ..., 0.0103, -0.0733, -0.0632], + [-0.1600, -0.1601, -0.1693, ..., 0.0656, 0.0486, 0.3365], + [-0.1327, -0.1296, -0.1625, ..., -0.1880, 0.0297, -0.1761], + ..., + [-0.1720, -0.1893, 0.0186, ..., -0.1936, -0.0910, -0.3254], + [ 0.1220, 0.0357, -0.1261, ..., -0.1896, 0.0155, -0.1761], + [ 0.1065, 0.0770, -0.1073, ..., 0.0771, -0.1891, -0.1693]], + device='cuda:0'), grad: tensor([[ 3.3993e-08, -2.1886e-08, -3.3528e-08, ..., -3.3760e-08, + 1.3737e-08, 3.4925e-09], + [ 3.9814e-08, 5.5414e-08, 4.8894e-09, ..., -5.3551e-09, + 1.1409e-08, -5.1688e-08], + [ 2.1677e-07, 1.3411e-07, 2.7940e-09, ..., 2.5146e-08, + 8.0792e-08, 2.1420e-08], + ..., + [ 1.0291e-07, 3.2363e-08, 4.1910e-09, ..., 5.9139e-08, + 4.8894e-08, 1.8161e-08], + [-1.2182e-06, -6.2957e-07, 2.0955e-09, ..., 7.4506e-09, + -3.8696e-07, 1.6298e-09], + [ 1.7975e-07, 1.9302e-07, 2.3050e-08, ..., -6.1467e-08, + 5.6345e-08, 3.7253e-09]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0288, -0.0319, -0.0109, 0.0004, -0.0013, 0.0038, 0.0048, -0.0096, + 0.0122, -0.0390], device='cuda:0'), grad: tensor([ 2.7940e-09, 2.7590e-07, 2.1793e-07, 1.2424e-06, -3.2363e-08, + 1.9139e-07, 1.2736e-07, 1.5507e-07, -2.9039e-06, 7.4599e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 217.26, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.4814 re_mapping 0.0043 re_causal 0.0123 /// teacc 98.95 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.0468, 0.0047, 0.1126, ..., 0.0104, -0.0735, -0.0632], + [-0.1618, -0.1612, -0.1694, ..., 0.0662, 0.0488, 0.3378], + [-0.1331, -0.1297, -0.1626, ..., -0.1889, 0.0298, -0.1764], + ..., + [-0.1730, -0.1905, 0.0181, ..., -0.1948, -0.0912, -0.3289], + [ 0.1229, 0.0359, -0.1262, ..., -0.1900, 0.0155, -0.1763], + [ 0.1087, 0.0779, -0.1073, ..., 0.0776, -0.1894, -0.1722]], + device='cuda:0'), grad: tensor([[ 5.2853e-08, 7.1013e-08, 2.3283e-10, ..., 2.0256e-08, + 1.7695e-08, 1.8394e-08], + [ 1.3574e-07, 1.7765e-07, 4.6566e-10, ..., -7.4506e-09, + 5.0757e-08, -1.0128e-07], + [-8.6706e-07, -1.4491e-06, 0.0000e+00, ..., 1.1176e-08, + 5.2154e-08, 1.4435e-08], + ..., + [ 3.4203e-07, 2.9500e-07, 5.5879e-09, ..., 2.7940e-07, + -1.6461e-07, 3.8883e-08], + [-2.1141e-07, -2.3888e-07, 0.0000e+00, ..., 1.0012e-08, + -1.8929e-07, 7.4506e-09], + [ 4.4797e-07, 1.0375e-06, 1.3970e-09, ..., -3.0594e-07, + 1.5134e-07, 6.5193e-09]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0287, -0.0320, -0.0108, 0.0003, -0.0022, 0.0037, 0.0044, -0.0099, + 0.0124, -0.0383], device='cuda:0'), grad: tensor([ 3.9209e-07, 1.2610e-06, -7.2382e-06, 5.2340e-07, 3.8324e-07, + 5.4343e-07, 1.1595e-07, -3.7090e-07, -7.7579e-07, 5.1819e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 217.50, cls_loss 0.0010 cls_loss_mapping 0.0029 cls_loss_causal 0.4846 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.06 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.0474, 0.0046, 0.1126, ..., 0.0103, -0.0739, -0.0635], + [-0.1627, -0.1625, -0.1696, ..., 0.0666, 0.0502, 0.3389], + [-0.1335, -0.1298, -0.1627, ..., -0.1910, 0.0295, -0.1773], + ..., + [-0.1734, -0.1909, 0.0176, ..., -0.1952, -0.0917, -0.3295], + [ 0.1242, 0.0358, -0.1262, ..., -0.1903, 0.0155, -0.1772], + [ 0.1104, 0.0787, -0.1073, ..., 0.0781, -0.1897, -0.1734]], + device='cuda:0'), grad: tensor([[ 2.4913e-08, 2.8405e-08, 4.6799e-08, ..., 5.0757e-08, + 3.2829e-08, 1.2340e-08], + [ 1.8114e-07, 2.2631e-07, 6.0536e-09, ..., -6.2864e-09, + -6.4727e-08, -1.0971e-06], + [ 1.7229e-07, 1.9069e-07, 2.0955e-09, ..., 4.1910e-09, + 3.2643e-07, 4.4820e-07], + ..., + [ 1.2270e-07, 1.3108e-07, 5.3784e-08, ..., 1.0710e-08, + 1.5809e-07, 3.7556e-07], + [-1.0729e-06, -1.2787e-06, 1.3039e-08, ..., 1.5134e-08, + -1.5059e-06, 7.9162e-09], + [ 6.7288e-08, 1.0151e-07, 1.5018e-07, ..., -2.1653e-08, + 5.1456e-08, 6.9849e-09]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0289, -0.0312, -0.0105, -0.0012, -0.0030, 0.0051, 0.0032, -0.0104, + 0.0124, -0.0379], device='cuda:0'), grad: tensor([ 4.4052e-07, 3.5577e-06, -1.9632e-06, 8.2701e-07, 1.3225e-07, + 4.1425e-06, 2.4540e-07, 1.3048e-06, -1.0394e-05, 1.6857e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 217.05, cls_loss 0.0010 cls_loss_mapping 0.0032 cls_loss_causal 0.5393 re_mapping 0.0041 re_causal 0.0125 /// teacc 98.99 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0480, 0.0048, 0.1128, ..., 0.0103, -0.0736, -0.0637], + [-0.1637, -0.1639, -0.1700, ..., 0.0670, 0.0504, 0.3397], + [-0.1339, -0.1299, -0.1630, ..., -0.1925, 0.0296, -0.1778], + ..., + [-0.1746, -0.1913, 0.0172, ..., -0.1960, -0.0920, -0.3303], + [ 0.1246, 0.0356, -0.1264, ..., -0.1911, 0.0155, -0.1776], + [ 0.1121, 0.0794, -0.1073, ..., 0.0783, -0.1901, -0.1743]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.6531e-08, 1.5926e-07, ..., 2.3120e-07, + 2.3516e-08, 8.7544e-08], + [ 4.8894e-09, 3.7253e-09, 4.8196e-08, ..., -8.9407e-07, + -2.8149e-07, -1.3793e-06], + [ 2.0489e-08, 6.0536e-09, 1.8626e-09, ..., 1.3644e-07, + 4.0978e-08, 1.9884e-07], + ..., + [ 1.9092e-08, 9.5461e-09, 1.1642e-08, ..., 2.9872e-07, + 1.4459e-07, 4.4145e-07], + [-3.9348e-08, 9.0804e-09, 4.7265e-08, ..., 1.2829e-07, + -4.7032e-08, 1.0012e-07], + [-6.6822e-08, -3.2363e-08, 7.1479e-08, ..., 1.8766e-07, + 1.5087e-07, 3.5227e-07]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0288, -0.0313, -0.0119, -0.0016, -0.0032, 0.0053, 0.0032, -0.0085, + 0.0123, -0.0374], device='cuda:0'), grad: tensor([ 6.2771e-07, -2.8890e-06, 3.2689e-07, 1.3295e-07, -3.0594e-07, + 2.6380e-07, -5.9605e-07, 1.1986e-06, 1.8347e-07, 1.0757e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 217.22, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4660 re_mapping 0.0043 re_causal 0.0122 /// teacc 99.03 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0486, 0.0048, 0.1128, ..., 0.0103, -0.0739, -0.0639], + [-0.1647, -0.1652, -0.1702, ..., 0.0680, 0.0506, 0.3415], + [-0.1352, -0.1302, -0.1631, ..., -0.1942, 0.0295, -0.1792], + ..., + [-0.1746, -0.1904, 0.0166, ..., -0.1964, -0.0921, -0.3329], + [ 0.1261, 0.0358, -0.1265, ..., -0.1922, 0.0156, -0.1773], + [ 0.1123, 0.0793, -0.1074, ..., 0.0783, -0.1909, -0.1776]], + device='cuda:0'), grad: tensor([[ 2.2096e-07, 1.5483e-07, -3.1665e-08, ..., 2.2002e-07, + 2.3283e-09, 4.1910e-09], + [ 4.7265e-08, 3.7951e-08, 4.1910e-09, ..., 1.8626e-09, + -3.6089e-08, -1.1455e-07], + [ 1.2573e-08, 7.4506e-09, 9.3132e-10, ..., 1.6298e-08, + 1.5367e-08, 2.5146e-08], + ..., + [ 2.4308e-07, 1.5483e-07, 1.3853e-07, ..., 2.7800e-07, + 2.2119e-08, 5.1456e-08], + [ 4.6589e-07, 4.1397e-07, 2.7940e-09, ..., 5.1828e-07, + 6.9849e-10, 6.5193e-09], + [-1.1427e-06, -8.9640e-07, 1.9185e-07, ..., -1.2154e-06, + 7.2177e-09, 1.0477e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0289, -0.0311, -0.0124, -0.0015, -0.0032, 0.0052, 0.0034, -0.0079, + 0.0125, -0.0377], device='cuda:0'), grad: tensor([ 5.3551e-07, 4.5495e-07, -8.4052e-08, 3.6787e-07, -6.9477e-07, + 7.1712e-08, 4.8894e-08, 4.4028e-07, 1.2862e-06, -2.4214e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 217.19, cls_loss 0.0009 cls_loss_mapping 0.0030 cls_loss_causal 0.4975 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.17 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0485, 0.0051, 0.1129, ..., 0.0103, -0.0741, -0.0640], + [-0.1654, -0.1656, -0.1706, ..., 0.0685, 0.0506, 0.3428], + [-0.1360, -0.1304, -0.1632, ..., -0.1955, 0.0295, -0.1799], + ..., + [-0.1759, -0.1918, 0.0154, ..., -0.1980, -0.0923, -0.3348], + [ 0.1273, 0.0359, -0.1267, ..., -0.1931, 0.0157, -0.1780], + [ 0.1135, 0.0795, -0.1084, ..., 0.0782, -0.1918, -0.1795]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -3.6322e-08, -7.3109e-08, ..., -6.1467e-08, + 1.1642e-08, 2.3283e-09], + [ 2.9802e-08, 1.6298e-08, 3.7253e-09, ..., -5.3085e-08, + 2.8405e-08, -1.2061e-07], + [ 1.2200e-07, 5.6345e-08, 5.5879e-09, ..., 2.5146e-08, + 1.0990e-07, 8.8476e-09], + ..., + [ 3.0268e-08, 2.1420e-08, 3.2596e-09, ..., 4.0978e-08, + 1.1642e-08, 3.4925e-08], + [-2.0070e-07, -5.3085e-08, 4.6566e-09, ..., 1.4901e-08, + -2.4447e-07, 3.7253e-09], + [-3.9581e-08, 1.1642e-08, 4.1910e-08, ..., 2.6077e-08, + 6.9849e-09, 4.6566e-08]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0288, -0.0310, -0.0124, -0.0015, -0.0018, 0.0053, 0.0032, -0.0083, + 0.0126, -0.0383], device='cuda:0'), grad: tensor([-1.3551e-07, -4.6100e-08, 2.9476e-07, -9.2667e-08, 1.4529e-07, + 1.0291e-07, 1.1642e-07, 1.6298e-07, -6.5472e-07, 1.2107e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 217.44, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.5426 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.19 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0484, 0.0054, 0.1131, ..., 0.0105, -0.0742, -0.0641], + [-0.1659, -0.1661, -0.1707, ..., 0.0696, 0.0508, 0.3438], + [-0.1362, -0.1305, -0.1633, ..., -0.1966, 0.0294, -0.1807], + ..., + [-0.1766, -0.1929, 0.0158, ..., -0.1992, -0.0924, -0.3364], + [ 0.1281, 0.0360, -0.1269, ..., -0.1936, 0.0158, -0.1781], + [ 0.1136, 0.0789, -0.1094, ..., 0.0780, -0.1925, -0.1814]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, -4.4238e-08, -7.7765e-08, ..., -4.5169e-08, + 9.3132e-09, -5.1223e-09], + [ 9.0804e-08, 5.4017e-08, 6.5193e-09, ..., 4.8894e-08, + 5.9605e-08, -2.0489e-08], + [ 1.1129e-07, 6.9849e-08, 5.1223e-09, ..., 3.7719e-08, + 1.2340e-07, 1.4435e-08], + ..., + [ 5.2154e-08, 3.0268e-08, 9.3132e-10, ..., 3.3993e-08, + 3.4459e-08, 9.7789e-09], + [-3.5390e-08, -2.2817e-08, 3.8650e-08, ..., 6.6590e-08, + -8.9407e-08, -8.8476e-09], + [-6.5658e-08, -2.5611e-08, 1.7695e-08, ..., -4.7032e-08, + 1.3970e-08, 3.2596e-09]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0286, -0.0306, -0.0125, -0.0014, -0.0008, 0.0052, 0.0031, -0.0085, + 0.0127, -0.0393], device='cuda:0'), grad: tensor([-1.1781e-07, 2.7753e-07, 4.5169e-07, -6.9942e-07, 1.3784e-07, + 3.0687e-07, -1.9651e-07, 5.0757e-08, -1.3085e-07, -7.9628e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 217.36, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.5020 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.12 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0488, 0.0053, 0.1132, ..., 0.0105, -0.0748, -0.0640], + [-0.1667, -0.1669, -0.1709, ..., 0.0698, 0.0508, 0.3444], + [-0.1368, -0.1306, -0.1633, ..., -0.1975, 0.0294, -0.1810], + ..., + [-0.1774, -0.1941, 0.0151, ..., -0.1995, -0.0925, -0.3372], + [ 0.1285, 0.0360, -0.1271, ..., -0.1947, 0.0158, -0.1783], + [ 0.1138, 0.0789, -0.1096, ..., 0.0781, -0.1931, -0.1827]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, -3.6787e-08, -1.2666e-07, ..., -6.9849e-08, + 9.7789e-09, -1.4435e-08], + [ 1.0431e-07, 7.8231e-08, 1.3970e-09, ..., 1.2154e-07, + -6.7521e-08, -1.4994e-07], + [ 1.3039e-08, 1.2107e-08, 6.5193e-09, ..., 3.6322e-08, + 4.1444e-08, 4.0978e-08], + ..., + [ 3.9581e-08, 2.0489e-08, 9.3132e-10, ..., 5.4482e-08, + 9.7789e-09, 1.8626e-08], + [ 2.4214e-08, 1.2573e-08, 1.3970e-09, ..., 5.3085e-08, + -6.7987e-08, 4.3772e-08], + [-1.9139e-07, -8.1025e-08, 1.1129e-07, ..., -2.6869e-07, + 2.7940e-09, 3.5390e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0287, -0.0307, -0.0125, -0.0010, -0.0006, 0.0051, 0.0035, -0.0086, + 0.0127, -0.0395], device='cuda:0'), grad: tensor([-2.6356e-07, 1.1409e-07, 1.9139e-07, -1.1455e-06, 2.3609e-07, + 1.2126e-06, 5.8673e-08, -5.6205e-07, -2.4680e-08, 2.0023e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 217.32, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.4902 re_mapping 0.0043 re_causal 0.0122 /// teacc 99.10 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.0491, 0.0056, 0.1140, ..., 0.0107, -0.0751, -0.0639], + [-0.1694, -0.1690, -0.1713, ..., 0.0692, 0.0505, 0.3446], + [-0.1373, -0.1307, -0.1635, ..., -0.1987, 0.0294, -0.1812], + ..., + [-0.1785, -0.1961, 0.0125, ..., -0.2004, -0.0926, -0.3376], + [ 0.1295, 0.0361, -0.1272, ..., -0.1948, 0.0159, -0.1785], + [ 0.1154, 0.0800, -0.1094, ..., 0.0787, -0.1934, -0.1835]], + device='cuda:0'), grad: tensor([[ 5.2853e-07, 3.7579e-07, -2.1001e-07, ..., -1.4575e-07, + 3.3667e-07, -5.5879e-09], + [ 3.7765e-07, 3.5157e-07, 1.0245e-08, ..., 8.8476e-09, + 3.8045e-07, -4.6100e-08], + [ 2.6729e-06, 1.7667e-06, 2.1420e-08, ..., 2.2352e-08, + 1.5935e-06, 1.2107e-08], + ..., + [ 5.6392e-07, 3.8324e-07, 7.4506e-09, ..., 1.3504e-08, + 4.8801e-07, 1.2107e-08], + [ 1.1018e-06, 1.1828e-07, 2.3283e-09, ..., 1.0245e-08, + -9.2341e-07, 1.8626e-09], + [ 6.1188e-07, 7.0175e-07, 1.2526e-07, ..., 4.9360e-08, + 4.2701e-07, 1.3504e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0282, -0.0314, -0.0125, 0.0023, -0.0014, 0.0017, 0.0037, -0.0086, + 0.0128, -0.0385], device='cuda:0'), grad: tensor([ 1.5739e-06, 2.1085e-06, 8.7023e-06, -2.0832e-05, 2.3469e-07, + -6.4494e-07, 2.5667e-06, 2.5406e-06, 8.8569e-07, 2.8480e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 217.31, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4848 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.11 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.0496, 0.0057, 0.1149, ..., 0.0114, -0.0725, -0.0639], + [-0.1699, -0.1701, -0.1737, ..., 0.0685, 0.0501, 0.3464], + [-0.1381, -0.1309, -0.1645, ..., -0.2024, 0.0292, -0.1829], + ..., + [-0.1790, -0.1967, 0.0124, ..., -0.2009, -0.0927, -0.3391], + [ 0.1310, 0.0362, -0.1272, ..., -0.1951, 0.0159, -0.1790], + [ 0.1160, 0.0801, -0.1095, ..., 0.0788, -0.1938, -0.1846]], + device='cuda:0'), grad: tensor([[ 1.8161e-08, 1.1176e-08, -9.3132e-09, ..., 8.8476e-09, + 2.0023e-08, 1.0710e-08], + [ 4.2375e-08, 3.2596e-08, 9.3132e-09, ..., -7.2177e-08, + -4.6100e-08, -2.9663e-07], + [ 1.4342e-07, 1.1083e-07, 4.6566e-10, ..., 1.1129e-07, + 1.4948e-07, 1.2759e-07], + ..., + [ 3.9581e-08, 3.0268e-08, 5.1223e-09, ..., 5.6345e-08, + 6.4261e-08, 1.0198e-07], + [ 3.7253e-09, 5.1223e-09, 4.6566e-10, ..., 9.3132e-09, + 4.6566e-09, 5.5879e-09], + [-1.1176e-08, -7.4506e-09, 2.5146e-08, ..., -1.2573e-08, + 4.4238e-08, 1.3970e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0269, -0.0315, -0.0128, 0.0019, -0.0015, 0.0021, 0.0035, -0.0087, + 0.0129, -0.0385], device='cuda:0'), grad: tensor([ 9.4995e-08, -1.4529e-07, 1.0710e-06, -9.4110e-07, -1.2899e-07, + 1.3923e-07, 4.9826e-08, -2.9802e-07, 4.1910e-08, 1.3271e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 217.41, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4882 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.07 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.0498, 0.0057, 0.1150, ..., 0.0113, -0.0726, -0.0640], + [-0.1701, -0.1715, -0.1738, ..., 0.0687, 0.0502, 0.3472], + [-0.1384, -0.1309, -0.1645, ..., -0.2029, 0.0291, -0.1835], + ..., + [-0.1792, -0.1962, 0.0120, ..., -0.2012, -0.0928, -0.3405], + [ 0.1310, 0.0362, -0.1274, ..., -0.1957, 0.0161, -0.1791], + [ 0.1163, 0.0802, -0.1098, ..., 0.0789, -0.1945, -0.1851]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.3318e-07, -5.1456e-07, ..., -1.9046e-07, + 7.9162e-09, 2.1886e-08], + [ 9.3132e-09, 1.4435e-08, 2.8871e-08, ..., -2.4168e-07, + -1.4855e-07, -6.4261e-07], + [ 7.9162e-09, 1.4901e-08, 3.2596e-08, ..., 1.5227e-07, + 7.5903e-08, 3.2783e-07], + ..., + [ 3.1665e-08, 5.6345e-08, 1.2293e-07, ..., 1.4016e-07, + 4.0047e-08, 1.5693e-07], + [ 1.7695e-08, 2.2352e-08, 1.4901e-08, ..., 4.3306e-08, + 1.2107e-08, 4.5635e-08], + [-1.0384e-07, -4.1910e-08, 1.4575e-07, ..., -4.1444e-08, + 9.7789e-09, 2.4680e-08]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0270, -0.0315, -0.0128, 0.0019, -0.0012, 0.0021, 0.0036, -0.0087, + 0.0130, -0.0387], device='cuda:0'), grad: tensor([-1.2033e-06, -1.1874e-06, 6.5099e-07, 8.8941e-08, 1.9930e-07, + 6.1002e-08, 3.1060e-07, 6.7847e-07, 2.8079e-07, 1.2806e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 217.36, cls_loss 0.0007 cls_loss_mapping 0.0026 cls_loss_causal 0.5102 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.14 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.0500, 0.0061, 0.1152, ..., 0.0114, -0.0728, -0.0640], + [-0.1705, -0.1733, -0.1741, ..., 0.0691, 0.0505, 0.3475], + [-0.1387, -0.1310, -0.1647, ..., -0.2052, 0.0290, -0.1838], + ..., + [-0.1795, -0.1968, 0.0118, ..., -0.2015, -0.0930, -0.3412], + [ 0.1312, 0.0361, -0.1276, ..., -0.1974, 0.0161, -0.1794], + [ 0.1172, 0.0806, -0.1102, ..., 0.0791, -0.1947, -0.1857]], + device='cuda:0'), grad: tensor([[-1.8720e-07, -2.7008e-07, -1.7416e-07, ..., 6.7055e-08, + 9.3132e-09, 3.0734e-08], + [ 1.7229e-08, 3.8650e-08, 3.6787e-08, ..., -1.6345e-07, + 1.8626e-08, -3.5437e-07], + [ 4.6566e-08, 9.6858e-08, 7.6368e-08, ..., 5.0291e-08, + 1.5507e-07, 6.3796e-08], + ..., + [ 6.1002e-08, 4.9360e-08, 7.0781e-08, ..., 1.0012e-07, + 6.4261e-08, 1.2014e-07], + [ 2.0023e-08, 4.8429e-08, 7.4971e-08, ..., 9.5926e-08, + -2.0256e-07, 2.3749e-08], + [-6.6124e-08, -1.3970e-09, 5.7742e-08, ..., -3.0734e-08, + 8.3819e-09, 5.5414e-08]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0268, -0.0313, -0.0129, 0.0018, -0.0014, 0.0021, 0.0043, -0.0088, + 0.0129, -0.0386], device='cuda:0'), grad: tensor([-5.8394e-07, -5.0385e-07, 1.0040e-06, 5.9558e-07, -1.4156e-07, + -9.9186e-08, -4.3167e-07, 3.2037e-07, -2.5798e-07, 1.0524e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 217.42, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4865 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.10 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.0484, 0.0066, 0.1156, ..., 0.0117, -0.0729, -0.0641], + [-0.1712, -0.1753, -0.1743, ..., 0.0692, 0.0508, 0.3482], + [-0.1392, -0.1311, -0.1648, ..., -0.2057, 0.0289, -0.1844], + ..., + [-0.1798, -0.1967, 0.0112, ..., -0.2018, -0.0933, -0.3423], + [ 0.1326, 0.0367, -0.1277, ..., -0.1974, 0.0162, -0.1795], + [ 0.1171, 0.0804, -0.1106, ..., 0.0791, -0.1949, -0.1866]], + device='cuda:0'), grad: tensor([[ 1.0710e-08, 5.1223e-09, -8.8476e-09, ..., 1.8626e-09, + 2.3283e-09, 3.2596e-09], + [ 1.4901e-08, 1.2573e-08, -1.4761e-07, ..., -6.0676e-07, + -2.2491e-07, -4.4983e-07], + [ 4.1910e-09, 4.1910e-09, 5.1223e-09, ..., 1.2107e-08, + 6.5193e-09, 7.4506e-09], + ..., + [ 1.1129e-07, 8.5216e-08, 7.4971e-08, ..., 4.5262e-07, + 1.4901e-07, 2.9011e-07], + [ 1.7276e-07, 1.2945e-07, 8.3819e-09, ..., 7.2643e-08, + -4.6566e-10, 1.3970e-09], + [-4.7591e-07, -3.4086e-07, 2.1886e-08, ..., -1.2899e-07, + 2.6077e-08, 5.5414e-08]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0265, -0.0311, -0.0129, 0.0018, -0.0012, 0.0021, 0.0041, -0.0090, + 0.0131, -0.0389], device='cuda:0'), grad: tensor([ 2.8405e-08, -2.1979e-06, 6.6124e-08, 3.1199e-08, 7.6555e-07, + 4.7497e-08, 5.3085e-08, 1.6391e-06, 5.0897e-07, -9.2434e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 217.13, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.4931 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.14 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.0486, 0.0067, 0.1158, ..., 0.0118, -0.0728, -0.0640], + [-0.1722, -0.1783, -0.1749, ..., 0.0690, 0.0506, 0.3488], + [-0.1409, -0.1317, -0.1649, ..., -0.2062, 0.0290, -0.1847], + ..., + [-0.1824, -0.1973, 0.0085, ..., -0.2027, -0.0935, -0.3431], + [ 0.1336, 0.0372, -0.1277, ..., -0.1973, 0.0163, -0.1796], + [ 0.1190, 0.0804, -0.1102, ..., 0.0794, -0.1951, -0.1878]], + device='cuda:0'), grad: tensor([[-1.8626e-09, -2.2352e-08, -9.7789e-09, ..., 1.6298e-08, + 6.0536e-09, 2.0489e-08], + [ 1.3970e-09, 1.8626e-09, 3.7253e-09, ..., -5.5414e-08, + -3.3341e-07, -9.1409e-07], + [ 3.2596e-09, 3.7253e-09, 8.8476e-09, ..., 3.8184e-08, + 8.6613e-08, 3.0687e-07], + ..., + [ 5.1223e-09, 3.2596e-09, 4.6566e-10, ..., 1.1642e-08, + 8.4750e-08, 1.8161e-07], + [-4.1910e-09, -3.7253e-09, 4.7497e-08, ..., 7.5437e-08, + 6.0536e-09, 3.3993e-08], + [-5.1223e-09, 1.0710e-08, 2.7940e-08, ..., 2.3749e-08, + 1.6298e-08, 2.5611e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0264, -0.0317, -0.0128, 0.0018, -0.0016, 0.0021, 0.0047, -0.0097, + 0.0133, -0.0378], device='cuda:0'), grad: tensor([ 1.1222e-07, -1.3867e-06, 5.4017e-07, 4.4238e-08, 5.8906e-07, + 8.2888e-08, -4.4471e-07, -1.8673e-06, 3.0082e-07, 2.0377e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 217.39, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4775 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.18 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.0470, 0.0068, 0.1162, ..., 0.0118, -0.0729, -0.0641], + [-0.1725, -0.1788, -0.1751, ..., 0.0691, 0.0506, 0.3495], + [-0.1410, -0.1318, -0.1650, ..., -0.2065, 0.0290, -0.1852], + ..., + [-0.1835, -0.1981, 0.0086, ..., -0.2040, -0.0936, -0.3439], + [ 0.1333, 0.0370, -0.1281, ..., -0.1981, 0.0163, -0.1798], + [ 0.1196, 0.0802, -0.1105, ..., 0.0798, -0.1955, -0.1883]], + device='cuda:0'), grad: tensor([[ 1.8161e-08, 1.9092e-08, -1.0245e-08, ..., 4.1910e-09, + 3.7253e-09, 2.7940e-09], + [ 9.6858e-08, 1.0058e-07, 4.6566e-10, ..., -3.7253e-09, + 2.4214e-08, -5.6811e-08], + [ 8.8941e-08, 8.5216e-08, 0.0000e+00, ..., 1.8626e-08, + -1.2945e-07, 1.1176e-08], + ..., + [ 1.9837e-07, 1.9884e-07, 4.6566e-10, ..., 4.2375e-08, + 2.5565e-07, 3.1665e-08], + [ 3.4459e-08, 4.5635e-08, 0.0000e+00, ..., 1.2107e-08, + 1.3039e-08, 3.2596e-09], + [ 4.7032e-08, 7.4506e-08, 8.8476e-09, ..., 3.4459e-08, + 1.3039e-08, 9.3132e-09]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0262, -0.0318, -0.0128, 0.0017, -0.0017, 0.0022, 0.0050, -0.0100, + 0.0132, -0.0375], device='cuda:0'), grad: tensor([ 2.6962e-07, 2.3283e-06, -6.2026e-07, -1.7136e-06, 2.7940e-08, + 3.7765e-07, 3.6322e-08, -1.4585e-06, 1.5739e-07, 6.0117e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 217.38, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.4560 re_mapping 0.0041 re_causal 0.0116 /// teacc 99.01 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.0476, 0.0068, 0.1164, ..., 0.0119, -0.0730, -0.0641], + [-0.1749, -0.1814, -0.1754, ..., 0.0718, 0.0524, 0.3501], + [-0.1414, -0.1320, -0.1651, ..., -0.2070, 0.0283, -0.1852], + ..., + [-0.1834, -0.1966, 0.0087, ..., -0.2068, -0.0952, -0.3450], + [ 0.1308, 0.0344, -0.1304, ..., -0.2016, 0.0164, -0.1803], + [ 0.1200, 0.0801, -0.1107, ..., 0.0799, -0.1959, -0.1897]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -2.9337e-08, -1.8161e-08, ..., -1.5367e-08, + 1.8394e-07, 6.5193e-09], + [ 1.8300e-07, 4.7032e-08, 1.4203e-07, ..., 7.4040e-08, + 1.2014e-07, -3.3807e-07], + [ 1.0710e-07, 8.2888e-08, 2.4773e-07, ..., 2.6543e-08, + 2.2957e-07, 5.7276e-08], + ..., + [ 1.6345e-07, 1.0058e-07, 2.8731e-07, ..., 8.5216e-08, + 4.2655e-07, 1.5507e-07], + [ 8.3121e-07, 4.4890e-07, 7.0781e-08, ..., 2.8312e-07, + -3.7253e-09, 1.1176e-08], + [-1.3635e-06, -5.5414e-07, 6.1002e-08, ..., -4.6520e-07, + 3.6322e-08, 5.7276e-08]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0263, -0.0323, -0.0133, 0.0019, -0.0018, 0.0025, 0.0023, -0.0090, + 0.0125, -0.0377], device='cuda:0'), grad: tensor([ 4.1863e-07, 7.8371e-07, 5.8161e-07, -3.9861e-07, -3.8669e-06, + 2.6589e-07, 9.3086e-07, 2.1495e-06, 2.4065e-06, -3.2801e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 217.14, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4838 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.01 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.0480, 0.0067, 0.1159, ..., 0.0115, -0.0730, -0.0642], + [-0.1755, -0.1833, -0.1758, ..., 0.0717, 0.0531, 0.3507], + [-0.1415, -0.1321, -0.1652, ..., -0.2073, 0.0283, -0.1855], + ..., + [-0.1836, -0.1970, 0.0084, ..., -0.2070, -0.0960, -0.3460], + [ 0.1303, 0.0337, -0.1310, ..., -0.2020, 0.0165, -0.1806], + [ 0.1215, 0.0806, -0.1109, ..., 0.0800, -0.1964, -0.1911]], + device='cuda:0'), grad: tensor([[ 4.8289e-07, 1.4808e-07, -3.6787e-08, ..., 1.8626e-07, + 8.6147e-08, 5.1223e-09], + [ 1.9651e-07, 7.3574e-08, 2.3283e-09, ..., 4.0978e-08, + 3.7253e-09, -1.1967e-07], + [ 4.5542e-07, 1.9232e-07, 3.8650e-08, ..., 1.8161e-07, + 1.2526e-07, 5.1688e-08], + ..., + [ 2.2165e-07, 9.1735e-08, 9.3132e-10, ..., 1.0012e-07, + 4.4238e-08, 2.8871e-08], + [ 1.8626e-09, -3.2131e-08, 4.6566e-10, ..., 1.1921e-07, + -1.1735e-07, 6.0536e-09], + [ 2.7474e-08, -2.6543e-08, 1.0710e-08, ..., 2.3283e-09, + 6.4261e-08, 1.0710e-08]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0270, -0.0311, -0.0131, 0.0019, -0.0017, 0.0026, 0.0030, -0.0101, + 0.0122, -0.0378], device='cuda:0'), grad: tensor([ 1.4231e-06, 3.2550e-07, 8.8103e-07, -3.6396e-06, 4.4284e-07, + 2.1420e-08, -3.1665e-08, 4.9779e-07, -2.0955e-08, 9.1270e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 217.76, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.5082 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.10 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.0491, 0.0065, 0.1161, ..., 0.0116, -0.0730, -0.0642], + [-0.1764, -0.1841, -0.1762, ..., 0.0719, 0.0532, 0.3525], + [-0.1422, -0.1322, -0.1655, ..., -0.2090, 0.0282, -0.1862], + ..., + [-0.1849, -0.1975, 0.0084, ..., -0.2074, -0.0961, -0.3476], + [ 0.1322, 0.0351, -0.1310, ..., -0.2018, 0.0168, -0.1813], + [ 0.1207, 0.0793, -0.1111, ..., 0.0801, -0.1985, -0.1944]], + device='cuda:0'), grad: tensor([[ 5.5600e-07, 6.7521e-08, 1.8300e-07, ..., -7.9162e-08, + 3.1199e-08, 0.0000e+00], + [ 9.4064e-08, 3.7253e-08, 1.9511e-07, ..., 6.3330e-08, + 4.5029e-07, -3.2596e-09], + [ 6.2957e-07, 2.6543e-07, 4.3446e-07, ..., 8.3819e-09, + 3.9116e-08, 4.6566e-10], + ..., + [ 4.9360e-08, 4.3306e-08, 1.7323e-06, ..., 2.4680e-08, + 7.8231e-07, 9.3132e-10], + [-1.3541e-06, -5.4110e-07, -7.0175e-07, ..., 9.3132e-09, + 8.5216e-08, 0.0000e+00], + [ 8.1025e-08, 1.4389e-07, 1.3523e-05, ..., -2.2352e-08, + 8.7172e-06, 9.3132e-10]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0271, -0.0312, -0.0131, 0.0016, -0.0014, 0.0028, 0.0034, -0.0102, + 0.0129, -0.0383], device='cuda:0'), grad: tensor([ 2.7735e-06, 8.8364e-06, 3.7625e-06, 1.9511e-07, -9.1970e-05, + 1.0012e-07, 5.6298e-07, 2.4326e-06, -6.7577e-06, 8.0049e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 217.43, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.5048 re_mapping 0.0037 re_causal 0.0120 /// teacc 99.16 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.0496, 0.0066, 0.1163, ..., 0.0116, -0.0732, -0.0643], + [-0.1772, -0.1849, -0.1765, ..., 0.0720, 0.0532, 0.3534], + [-0.1428, -0.1324, -0.1656, ..., -0.2095, 0.0282, -0.1867], + ..., + [-0.1849, -0.1977, 0.0082, ..., -0.2075, -0.0962, -0.3483], + [ 0.1321, 0.0350, -0.1312, ..., -0.2023, 0.0168, -0.1815], + [ 0.1207, 0.0791, -0.1114, ..., 0.0802, -0.1996, -0.1977]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.2596e-09, -5.5879e-09, ..., -1.3970e-09, + 2.3283e-09, 2.7940e-09], + [ 1.3970e-09, 1.3970e-09, 4.6566e-10, ..., -5.9139e-08, + -3.8184e-08, -2.0675e-07], + [ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 3.2596e-09, + -1.7788e-07, 2.1886e-08], + ..., + [ 3.7253e-09, 2.3283e-09, 4.6566e-10, ..., 2.3283e-08, + 3.1199e-08, 6.9384e-08], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09], + [-9.3132e-09, -1.8626e-09, 4.1910e-09, ..., 2.4214e-08, + 1.8626e-08, 9.0804e-08]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0271, -0.0314, -0.0130, 0.0016, -0.0010, 0.0028, 0.0041, -0.0101, + 0.0129, -0.0389], device='cuda:0'), grad: tensor([ 9.3132e-09, -2.7800e-07, -4.3400e-07, 1.3039e-08, 6.9104e-07, + 1.2107e-08, 6.5193e-09, -1.7835e-07, 3.3062e-08, 1.5320e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 217.44, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4800 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.14 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.0495, 0.0069, 0.1166, ..., 0.0118, -0.0731, -0.0642], + [-0.1778, -0.1854, -0.1777, ..., 0.0717, 0.0531, 0.3537], + [-0.1434, -0.1325, -0.1657, ..., -0.2099, 0.0282, -0.1872], + ..., + [-0.1851, -0.1981, 0.0079, ..., -0.2075, -0.0963, -0.3489], + [ 0.1321, 0.0349, -0.1315, ..., -0.2025, 0.0169, -0.1817], + [ 0.1211, 0.0792, -0.1116, ..., 0.0802, -0.1999, -0.1990]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -1.3970e-09, 4.6566e-10, ..., 9.3132e-10, + 1.0710e-08, 2.7940e-09], + [ 6.9849e-09, 2.7940e-09, 1.2107e-08, ..., -6.5193e-09, + 7.2643e-08, -1.4435e-08], + [ 6.9849e-09, 4.1910e-09, 2.3283e-09, ..., 4.6566e-09, + -1.3364e-07, 3.2596e-09], + ..., + [ 1.3970e-08, 6.5193e-09, 1.1642e-08, ..., 6.5193e-09, + 5.1223e-08, 1.8626e-09], + [ 6.5193e-09, 2.7940e-09, 1.8626e-09, ..., 5.1223e-09, + 2.1886e-08, 2.3283e-09], + [ 6.0536e-09, 1.8626e-09, 6.1002e-08, ..., 9.3132e-10, + 3.5856e-08, 4.1910e-09]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0269, -0.0318, -0.0131, 0.0016, -0.0011, 0.0027, 0.0049, -0.0097, + 0.0128, -0.0390], device='cuda:0'), grad: tensor([ 5.7742e-08, 4.8056e-07, -1.0114e-06, -9.0804e-08, -2.3050e-07, + 3.9116e-08, 6.7055e-08, 2.5425e-07, 1.6158e-07, 2.6356e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 217.58, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.5022 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.0494, 0.0079, 0.1173, ..., 0.0123, -0.0733, -0.0643], + [-0.1781, -0.1863, -0.1781, ..., 0.0719, 0.0533, 0.3552], + [-0.1443, -0.1327, -0.1659, ..., -0.2107, 0.0282, -0.1887], + ..., + [-0.1856, -0.1982, 0.0056, ..., -0.2078, -0.0965, -0.3504], + [ 0.1321, 0.0348, -0.1317, ..., -0.2029, 0.0169, -0.1826], + [ 0.1204, 0.0788, -0.1141, ..., 0.0799, -0.2015, -0.2004]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, -1.1921e-07, -2.0163e-07, ..., -1.1409e-07, + 3.5390e-08, 0.0000e+00], + [ 6.9849e-09, 1.3970e-08, 2.3283e-08, ..., 1.6298e-08, + 2.4820e-07, -5.1223e-09], + [ 1.4435e-08, 9.7789e-09, 1.6252e-07, ..., 1.0757e-07, + -4.1444e-07, 1.3970e-09], + ..., + [ 3.9116e-08, 2.6543e-08, 1.0245e-08, ..., 3.1199e-08, + 1.5832e-08, 2.3283e-09], + [-4.5169e-08, 5.8673e-08, 5.5879e-08, ..., 4.0047e-08, + -5.1223e-09, 1.3970e-09], + [-7.0315e-08, 2.0023e-08, 1.2387e-07, ..., 1.0710e-08, + 7.9162e-09, 1.3970e-09]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0262, -0.0317, -0.0131, 0.0016, 0.0017, 0.0027, 0.0049, -0.0099, + 0.0127, -0.0414], device='cuda:0'), grad: tensor([-2.8871e-07, 1.1530e-06, -1.5199e-06, 1.2452e-06, 4.8103e-07, + -1.1763e-06, -1.5227e-07, 1.9139e-07, 5.3085e-08, 2.5611e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 217.69, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.4963 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.06 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.0493, 0.0082, 0.1176, ..., 0.0124, -0.0734, -0.0650], + [-0.1784, -0.1869, -0.1783, ..., 0.0724, 0.0533, 0.3570], + [-0.1446, -0.1327, -0.1660, ..., -0.2111, 0.0283, -0.1891], + ..., + [-0.1859, -0.1995, 0.0056, ..., -0.2081, -0.0967, -0.3525], + [ 0.1322, 0.0351, -0.1318, ..., -0.2032, 0.0169, -0.1835], + [ 0.1208, 0.0789, -0.1145, ..., 0.0799, -0.2024, -0.2028]], + device='cuda:0'), grad: tensor([[ 2.3749e-08, 2.8405e-08, 4.6566e-10, ..., 5.5879e-09, + 6.0536e-09, 2.3283e-09], + [ 2.9523e-07, 3.5251e-07, 6.9849e-09, ..., 8.3819e-09, + 3.3993e-08, -4.5635e-08], + [ 1.2070e-06, 1.4380e-06, 4.6566e-10, ..., 1.5367e-08, + -1.7695e-08, 9.7789e-09], + ..., + [ 9.7230e-07, 1.0785e-06, 4.6566e-10, ..., 1.4575e-07, + 5.2620e-08, 2.4214e-08], + [ 3.7253e-08, 1.0431e-07, 4.6566e-10, ..., 2.0955e-08, + -7.3109e-08, 3.7253e-09], + [-3.9116e-08, -9.3132e-10, 1.6764e-08, ..., -8.5216e-08, + 1.8161e-08, 3.2596e-09]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0262, -0.0316, -0.0127, 0.0016, 0.0019, 0.0027, 0.0051, -0.0100, + 0.0127, -0.0418], device='cuda:0'), grad: tensor([ 7.6927e-07, 1.1101e-05, -5.6773e-06, -1.5706e-05, 2.1746e-07, + 2.3544e-06, 8.2422e-08, 6.2548e-06, 3.7625e-07, 2.1467e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 217.52, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.5015 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.13 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.0483, 0.0090, 0.1181, ..., 0.0129, -0.0735, -0.0652], + [-0.1789, -0.1875, -0.1789, ..., 0.0723, 0.0534, 0.3576], + [-0.1450, -0.1328, -0.1661, ..., -0.2118, 0.0284, -0.1893], + ..., + [-0.1861, -0.1999, 0.0062, ..., -0.2083, -0.0968, -0.3533], + [ 0.1322, 0.0351, -0.1321, ..., -0.2036, 0.0169, -0.1836], + [ 0.1210, 0.0786, -0.1150, ..., 0.0797, -0.2032, -0.2035]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 7.4506e-09, -2.7940e-09, ..., -9.3132e-10, + 7.4506e-09, 0.0000e+00], + [ 1.2107e-08, 1.5367e-08, 1.1642e-08, ..., 6.5193e-09, + 2.5146e-08, 0.0000e+00], + [ 3.5856e-08, 2.7008e-08, 2.7940e-09, ..., 1.7695e-08, + 2.7940e-08, 4.6566e-10], + ..., + [ 9.7789e-09, 8.8476e-09, 1.8626e-09, ..., 5.1223e-09, + 6.9849e-09, 0.0000e+00], + [ 1.5832e-08, 1.5320e-07, 4.6566e-10, ..., 1.8626e-09, + -1.3039e-08, 0.0000e+00], + [ 3.5856e-08, 1.8859e-07, 1.9092e-08, ..., -9.3132e-10, + 2.5611e-08, 0.0000e+00]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0257, -0.0319, -0.0126, 0.0015, 0.0018, 0.0027, 0.0059, -0.0099, + 0.0127, -0.0423], device='cuda:0'), grad: tensor([ 7.4040e-08, 2.4028e-07, -3.3807e-07, 7.2271e-07, -1.3504e-07, + -1.3616e-06, 1.3178e-07, 2.4680e-08, 2.2585e-07, 4.4564e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 217.61, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4930 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.15 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.0491, 0.0092, 0.1183, ..., 0.0130, -0.0737, -0.0654], + [-0.1793, -0.1880, -0.1790, ..., 0.0724, 0.0534, 0.3585], + [-0.1456, -0.1330, -0.1662, ..., -0.2124, 0.0283, -0.1898], + ..., + [-0.1866, -0.2005, 0.0062, ..., -0.2085, -0.0969, -0.3544], + [ 0.1321, 0.0350, -0.1322, ..., -0.2040, 0.0169, -0.1844], + [ 0.1216, 0.0786, -0.1152, ..., 0.0798, -0.2036, -0.2044]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 7.9162e-09, 3.2596e-09, ..., 3.5390e-08, + 1.3970e-08, 3.2131e-08], + [ 6.9849e-09, -7.3109e-08, 3.3993e-08, ..., -2.8918e-07, + -1.1548e-07, -3.7765e-07], + [ 1.5832e-08, 2.1886e-08, 3.2596e-09, ..., 3.1199e-08, + 1.8626e-08, 2.6543e-08], + ..., + [ 8.3819e-09, 2.2817e-08, 1.3970e-09, ..., 5.4482e-08, + 2.6077e-08, 5.9605e-08], + [-1.3504e-08, -1.8161e-08, 2.0303e-07, ..., 2.2864e-07, + -2.1886e-08, 1.0710e-08], + [-2.1420e-08, 1.1642e-08, 2.5146e-08, ..., 7.4506e-08, + 6.3796e-08, 1.0384e-07]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0256, -0.0319, -0.0126, 0.0016, 0.0017, 0.0026, 0.0060, -0.0100, + 0.0127, -0.0423], device='cuda:0'), grad: tensor([ 1.4482e-07, -1.2014e-07, 2.7753e-07, -2.0489e-08, -4.6566e-08, + 6.5677e-06, -7.1004e-06, -7.4925e-07, 6.4680e-07, 4.1211e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 217.73, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4996 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.08 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.0495, 0.0092, 0.1183, ..., 0.0130, -0.0739, -0.0656], + [-0.1798, -0.1886, -0.1793, ..., 0.0725, 0.0536, 0.3605], + [-0.1469, -0.1332, -0.1665, ..., -0.2132, 0.0283, -0.1915], + ..., + [-0.1868, -0.2008, 0.0063, ..., -0.2087, -0.0970, -0.3558], + [ 0.1322, 0.0350, -0.1320, ..., -0.2044, 0.0170, -0.1852], + [ 0.1223, 0.0787, -0.1153, ..., 0.0800, -0.2042, -0.2057]], + device='cuda:0'), grad: tensor([[ 1.0477e-07, 4.1863e-07, 7.8697e-08, ..., 1.1586e-06, + 1.0710e-08, 6.9849e-08], + [ 5.8673e-08, 1.4901e-08, -1.3970e-08, ..., -1.3988e-06, + -3.3062e-07, -2.6189e-06], + [ 8.0559e-08, 6.5193e-09, 2.7940e-09, ..., 2.1886e-08, + 6.9849e-09, 2.7008e-08], + ..., + [-1.4417e-06, 1.2573e-08, 8.8476e-09, ..., 6.3330e-08, + 1.2573e-08, 5.7276e-08], + [ 2.3283e-08, 1.2107e-08, 1.2107e-08, ..., 3.5390e-07, + 8.5216e-08, 6.4587e-07], + [ 1.1511e-06, -4.7404e-07, -5.8208e-08, ..., -1.2582e-06, + 1.2107e-08, 2.0023e-08]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0257, -0.0320, -0.0128, 0.0016, 0.0013, 0.0026, 0.0058, -0.0096, + 0.0130, -0.0423], device='cuda:0'), grad: tensor([ 2.2501e-06, -3.5111e-06, 1.0021e-06, 2.9756e-07, 1.5274e-07, + -5.3179e-07, 3.1348e-06, -1.7226e-05, 1.4119e-06, 1.3031e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 217.49, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.4840 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.17 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.0509, 0.0093, 0.1184, ..., 0.0130, -0.0741, -0.0657], + [-0.1811, -0.1899, -0.1795, ..., 0.0727, 0.0535, 0.3616], + [-0.1477, -0.1334, -0.1666, ..., -0.2139, 0.0283, -0.1919], + ..., + [-0.1876, -0.2016, 0.0060, ..., -0.2089, -0.0971, -0.3571], + [ 0.1325, 0.0353, -0.1319, ..., -0.2049, 0.0170, -0.1864], + [ 0.1229, 0.0788, -0.1155, ..., 0.0801, -0.2048, -0.2064]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -2.3749e-08, -2.5379e-08, ..., -3.8650e-08, + 2.2352e-08, 9.3132e-10], + [ 5.1223e-09, 9.0804e-09, 1.6298e-09, ..., -9.5461e-09, + 1.8626e-07, -2.4913e-08], + [ 1.8626e-09, 3.0268e-09, 6.9849e-10, ..., 2.0955e-09, + -3.4925e-07, 2.7940e-09], + ..., + [ 1.8626e-09, 2.0955e-09, 4.6566e-10, ..., 9.3132e-09, + 7.4506e-09, 1.8394e-08], + [-2.8173e-07, -5.2387e-07, 5.5879e-09, ..., 4.4238e-09, + -6.3563e-07, 1.8626e-09], + [ 1.6298e-09, 2.7008e-08, 5.1456e-08, ..., 4.3306e-08, + 4.6100e-08, 3.0268e-09]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0259, -0.0324, -0.0125, 0.0017, 0.0012, 0.0026, 0.0060, -0.0098, + 0.0132, -0.0425], device='cuda:0'), grad: tensor([ 7.7533e-08, 1.0217e-06, -2.1048e-06, 3.9348e-08, -5.7509e-08, + 1.5181e-07, 3.4571e-06, 1.5576e-07, -2.9728e-06, 2.4587e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 217.50, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4822 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.13 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.0510, 0.0097, 0.1203, ..., 0.0140, -0.0742, -0.0658], + [-0.1818, -0.1905, -0.1798, ..., 0.0729, 0.0534, 0.3630], + [-0.1488, -0.1338, -0.1670, ..., -0.2147, 0.0282, -0.1922], + ..., + [-0.1880, -0.2022, 0.0061, ..., -0.2091, -0.0972, -0.3579], + [ 0.1331, 0.0356, -0.1319, ..., -0.2051, 0.0172, -0.1875], + [ 0.1229, 0.0784, -0.1160, ..., 0.0800, -0.2064, -0.2068]], + device='cuda:0'), grad: tensor([[ 4.7963e-07, 8.2562e-07, -1.1064e-06, ..., -2.3283e-08, + 2.5891e-07, 0.0000e+00], + [ 6.2399e-08, 1.0524e-07, 3.2131e-08, ..., 3.4925e-08, + 5.8208e-08, -9.3132e-09], + [ 3.0780e-07, 5.0338e-07, 4.1071e-07, ..., 9.3132e-09, + 2.9057e-07, 0.0000e+00], + ..., + [ 1.1921e-07, 1.4761e-07, 3.2596e-08, ..., 3.9116e-08, + 4.5635e-08, 2.3283e-09], + [-1.8794e-06, -3.1423e-06, -9.4576e-07, ..., -3.0361e-07, + -1.6615e-06, 4.6566e-10], + [-3.7253e-08, 5.4017e-08, 5.8208e-08, ..., -8.9873e-08, + 4.9826e-08, 4.6566e-10]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0241, -0.0325, -0.0126, 0.0016, 0.0016, 0.0026, 0.0044, -0.0098, + 0.0137, -0.0431], device='cuda:0'), grad: tensor([-2.1746e-07, 9.9279e-07, 1.9427e-06, 5.6159e-07, 1.1194e-06, + 3.6620e-06, 7.4506e-06, 7.1386e-07, -1.6466e-05, 2.3283e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 217.36, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4682 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.19 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.0525, 0.0093, 0.1204, ..., 0.0136, -0.0744, -0.0662], + [-0.1815, -0.1905, -0.1801, ..., 0.0734, 0.0534, 0.3646], + [-0.1496, -0.1340, -0.1672, ..., -0.2157, 0.0282, -0.1926], + ..., + [-0.1863, -0.2028, 0.0065, ..., -0.2094, -0.0973, -0.3595], + [ 0.1333, 0.0358, -0.1319, ..., -0.2056, 0.0174, -0.1897], + [ 0.1238, 0.0804, -0.1160, ..., 0.0812, -0.2064, -0.2072]], + device='cuda:0'), grad: tensor([[-4.6566e-10, -4.3306e-08, -5.3551e-08, ..., -6.2864e-08, + 2.3283e-09, -6.9849e-09], + [ 7.9162e-09, 5.5879e-09, 6.5193e-09, ..., -1.2573e-08, + -1.0710e-08, -6.4261e-08], + [-1.3504e-08, 1.8626e-09, 3.7253e-09, ..., 6.5193e-09, + -1.9092e-08, 6.9849e-09], + ..., + [ 1.1642e-07, 5.4482e-08, 1.0710e-08, ..., 8.3819e-08, + 1.8626e-08, 4.6100e-08], + [-5.0291e-08, -2.4680e-08, 3.2596e-09, ..., 1.9558e-08, + 5.5879e-09, 5.1223e-09], + [-9.8720e-08, -1.8626e-08, 3.3528e-08, ..., -5.0757e-08, + 4.6566e-10, 6.9849e-09]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0245, -0.0324, -0.0125, 0.0016, 0.0002, 0.0026, 0.0042, -0.0091, + 0.0138, -0.0430], device='cuda:0'), grad: tensor([ 3.4496e-06, 6.0536e-08, -7.9796e-06, 4.2748e-07, 2.1514e-07, + 4.4703e-08, 1.2713e-07, 5.0711e-07, 3.0249e-06, 1.2433e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 217.61, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4596 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.18 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.0533, 0.0092, 0.1222, ..., 0.0146, -0.0745, -0.0663], + [-0.1819, -0.1909, -0.1804, ..., 0.0735, 0.0535, 0.3657], + [-0.1510, -0.1342, -0.1674, ..., -0.2169, 0.0281, -0.1931], + ..., + [-0.1866, -0.2032, 0.0061, ..., -0.2096, -0.0974, -0.3612], + [ 0.1333, 0.0357, -0.1320, ..., -0.2060, 0.0174, -0.1903], + [ 0.1250, 0.0813, -0.1160, ..., 0.0817, -0.2066, -0.2083]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, -5.0757e-08, -8.8941e-08, ..., -5.4948e-08, + 4.6566e-10, 4.6566e-10], + [ 8.8941e-08, 7.0315e-08, 1.1176e-08, ..., 1.7229e-08, + 6.9849e-09, -1.1642e-08], + [ 5.7369e-07, 4.0280e-07, 2.8405e-08, ..., 2.0955e-08, + 6.9849e-09, 4.1910e-09], + ..., + [ 3.3528e-08, 2.6543e-08, 1.3970e-08, ..., 1.9558e-08, + 7.4506e-09, 6.9849e-09], + [-1.1157e-06, -7.8185e-07, 4.6566e-10, ..., -4.7497e-08, + -2.8871e-08, 1.8626e-09], + [-2.1420e-08, 1.9558e-08, 2.9337e-08, ..., -3.4459e-08, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0230, -0.0323, -0.0125, 0.0016, -0.0002, 0.0026, 0.0027, -0.0092, + 0.0138, -0.0426], device='cuda:0'), grad: tensor([-1.8906e-07, 3.9535e-07, 1.9819e-06, 8.6147e-07, 4.5169e-08, + 3.9814e-07, 3.3341e-07, 1.2666e-07, -4.0531e-06, 9.3598e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 217.39, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4805 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.12 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.0538, 0.0092, 0.1222, ..., 0.0145, -0.0747, -0.0664], + [-0.1811, -0.1929, -0.1805, ..., 0.0743, 0.0522, 0.3689], + [-0.1521, -0.1345, -0.1677, ..., -0.2161, 0.0286, -0.1947], + ..., + [-0.1880, -0.2023, 0.0060, ..., -0.2110, -0.0975, -0.3645], + [ 0.1333, 0.0357, -0.1321, ..., -0.2064, 0.0174, -0.1906], + [ 0.1256, 0.0817, -0.1162, ..., 0.0820, -0.2073, -0.2088]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 4.1910e-09, 0.0000e+00, ..., 4.6566e-09, + 3.2596e-09, 9.7789e-09], + [ 1.3970e-08, 1.2573e-08, 0.0000e+00, ..., -6.7987e-08, + -9.5461e-08, -2.9802e-07], + [ 2.0023e-08, 1.8626e-08, 0.0000e+00, ..., 1.0245e-08, + 1.3970e-08, 2.6077e-08], + ..., + [ 2.4214e-08, 2.1886e-08, 0.0000e+00, ..., 3.2596e-08, + 4.4238e-08, 1.1502e-07], + [ 2.4214e-08, 1.9558e-08, 4.6566e-10, ..., 1.3504e-08, + 4.1910e-09, 9.7789e-09], + [-3.3993e-08, -2.1886e-08, 1.8626e-09, ..., -9.3132e-09, + 6.0536e-09, 1.5832e-08]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0231, -0.0321, -0.0122, 0.0016, -0.0001, 0.0026, 0.0030, -0.0099, + 0.0137, -0.0426], device='cuda:0'), grad: tensor([ 1.1688e-07, -2.4494e-07, -4.2096e-06, -1.9651e-07, 2.5425e-07, + 1.3737e-07, 1.3830e-07, 2.6170e-07, 3.6899e-06, 4.5635e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 217.34, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.5002 re_mapping 0.0040 re_causal 0.0113 /// teacc 99.17 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.0559, 0.0076, 0.1215, ..., 0.0135, -0.0746, -0.0665], + [-0.1812, -0.1936, -0.1812, ..., 0.0743, 0.0521, 0.3695], + [-0.1531, -0.1347, -0.1680, ..., -0.2167, 0.0286, -0.1957], + ..., + [-0.1886, -0.2032, 0.0059, ..., -0.2114, -0.0975, -0.3648], + [ 0.1330, 0.0351, -0.1326, ..., -0.2075, 0.0174, -0.1913], + [ 0.1270, 0.0830, -0.1165, ..., 0.0826, -0.2080, -0.2092]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 4.6566e-09, -2.7940e-09, ..., 2.3283e-09, + 1.3970e-09, 2.3283e-09], + [ 5.1223e-09, 1.1642e-08, 4.6566e-10, ..., -1.3504e-08, + -3.2596e-09, -4.7497e-08], + [ 1.4901e-08, 1.5367e-08, 4.6566e-10, ..., 7.9162e-09, + 1.2107e-08, 9.3132e-09], + ..., + [ 9.3132e-09, 8.8476e-09, 0.0000e+00, ..., 7.9162e-09, + 2.3283e-09, 6.0536e-09], + [ 4.5169e-08, 1.6624e-07, 0.0000e+00, ..., 3.9581e-08, + -1.0710e-08, 5.5879e-09], + [-7.7300e-08, -5.3085e-08, 3.2596e-09, ..., -5.8208e-08, + 1.3970e-09, 2.7940e-09]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0243, -0.0326, -0.0118, 0.0017, 0.0002, 0.0025, 0.0042, -0.0099, + 0.0134, -0.0424], device='cuda:0'), grad: tensor([ 5.9605e-08, 1.7043e-06, 9.2620e-07, 8.8941e-08, 7.7765e-08, + -1.0557e-05, 9.9689e-06, -3.4794e-06, 1.0561e-06, 1.3644e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 217.05, cls_loss 0.0006 cls_loss_mapping 0.0023 cls_loss_causal 0.4799 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.11 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.0562, 0.0078, 0.1215, ..., 0.0135, -0.0747, -0.0668], + [-0.1813, -0.1943, -0.1815, ..., 0.0744, 0.0521, 0.3704], + [-0.1541, -0.1351, -0.1682, ..., -0.2174, 0.0286, -0.1968], + ..., + [-0.1892, -0.2041, 0.0073, ..., -0.2117, -0.0976, -0.3651], + [ 0.1331, 0.0352, -0.1326, ..., -0.2077, 0.0175, -0.1919], + [ 0.1278, 0.0832, -0.1169, ..., 0.0828, -0.2082, -0.2115]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, -4.6566e-09, -2.8405e-08, ..., -1.5832e-08, + 7.4506e-09, 9.3132e-10], + [ 1.4901e-08, 1.3504e-08, 6.8452e-08, ..., 1.7695e-08, + 4.2375e-08, -4.3306e-08], + [ 1.8626e-08, 2.5611e-08, 4.6566e-09, ..., 6.5193e-09, + 6.0536e-09, 5.5879e-09], + ..., + [ 3.3993e-08, 3.1665e-08, 3.9581e-08, ..., 3.9581e-08, + 2.8871e-08, 1.2573e-08], + [ 4.0047e-08, 2.3283e-07, 4.1910e-09, ..., 1.3039e-08, + -1.4901e-08, 1.4435e-08], + [-3.6787e-08, -1.2107e-08, 8.2422e-08, ..., -1.3970e-09, + 5.4482e-08, 3.7253e-09]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0244, -0.0326, -0.0118, 0.0016, 0.0001, 0.0026, 0.0041, -0.0096, + 0.0135, -0.0428], device='cuda:0'), grad: tensor([-7.4506e-09, 2.6822e-07, 5.7276e-08, -4.1462e-06, -8.9407e-07, + 3.6247e-06, 1.7136e-07, 2.7195e-07, 4.0838e-07, 2.5099e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 217.29, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4808 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.14 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.0565, 0.0078, 0.1216, ..., 0.0136, -0.0748, -0.0669], + [-0.1793, -0.1927, -0.1817, ..., 0.0770, 0.0524, 0.3725], + [-0.1546, -0.1353, -0.1685, ..., -0.2178, 0.0285, -0.1980], + ..., + [-0.1911, -0.2064, 0.0072, ..., -0.2144, -0.0976, -0.3669], + [ 0.1333, 0.0353, -0.1333, ..., -0.2083, 0.0174, -0.1942], + [ 0.1278, 0.0829, -0.1170, ..., 0.0829, -0.2086, -0.2122]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, -2.7940e-09, -1.5832e-08, ..., -1.4435e-08, + 2.7940e-09, 1.4901e-08], + [-1.5348e-06, -1.6317e-06, 1.8626e-09, ..., -5.6438e-07, + -1.0245e-08, -4.4741e-06], + [ 3.1665e-08, 4.0978e-08, 2.7940e-09, ..., 1.8626e-08, + 1.2107e-08, 6.4727e-08], + ..., + [ 1.3541e-06, 1.4678e-06, 0.0000e+00, ..., 4.5402e-07, + 8.8476e-09, 3.8706e-06], + [ 1.2387e-07, 1.5367e-07, 4.6566e-10, ..., 5.4482e-08, + 8.3819e-09, 3.0966e-07], + [ 2.1886e-08, 3.4459e-08, 1.1176e-08, ..., 4.8429e-08, + 3.2596e-09, 9.0338e-08]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0244, -0.0306, -0.0117, 0.0016, 0.0002, 0.0026, 0.0044, -0.0115, + 0.0133, -0.0430], device='cuda:0'), grad: tensor([ 2.3749e-08, -2.1890e-05, 3.3528e-07, -2.5611e-07, 1.7695e-08, + -6.8918e-08, 6.9290e-07, 1.9163e-05, 1.6652e-06, 2.7474e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 217.38, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.5013 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.17 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.0570, 0.0079, 0.1216, ..., 0.0136, -0.0750, -0.0670], + [-0.1794, -0.1930, -0.1819, ..., 0.0770, 0.0524, 0.3733], + [-0.1576, -0.1363, -0.1686, ..., -0.2185, 0.0284, -0.1999], + ..., + [-0.1917, -0.2072, 0.0071, ..., -0.2146, -0.0977, -0.3672], + [ 0.1333, 0.0352, -0.1333, ..., -0.2081, 0.0175, -0.1948], + [ 0.1287, 0.0831, -0.1171, ..., 0.0832, -0.2089, -0.2127]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 5.1223e-09, 4.6566e-10, ..., 2.3283e-08, + 6.0536e-09, 2.1886e-08], + [ 2.7940e-09, 1.3970e-09, -3.2596e-09, ..., -6.2305e-07, + -2.2585e-07, -9.1316e-07], + [-1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 1.0710e-08, + -5.5414e-08, 1.6764e-08], + ..., + [ 2.2817e-08, 8.8476e-09, 4.6566e-10, ..., 1.7229e-08, + 6.0536e-09, 6.9849e-09], + [ 1.9092e-08, 2.1420e-08, 0.0000e+00, ..., 4.3772e-08, + 1.3970e-08, 5.3551e-08], + [-1.9744e-07, -6.7987e-08, 1.8626e-09, ..., -1.7043e-07, + 1.8626e-09, 2.3283e-09]], device='cuda:0') +Epoch 257, bias, value: tensor([-2.4453e-02, -3.0728e-02, -1.1687e-02, 1.7160e-03, -6.7318e-05, + 2.6795e-03, 3.6791e-03, -1.1682e-02, 1.3500e-02, -4.2743e-02], + device='cuda:0'), grad: tensor([ 9.0804e-08, -1.7602e-06, -5.1744e-06, 7.2876e-07, 4.2375e-07, + -4.2701e-07, 1.5367e-06, 4.8913e-06, 2.0629e-07, -5.0198e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 217.55, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4712 re_mapping 0.0039 re_causal 0.0114 /// teacc 99.14 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.0589, 0.0075, 0.1220, ..., 0.0137, -0.0750, -0.0654], + [-0.1797, -0.1933, -0.1840, ..., 0.0766, 0.0523, 0.3737], + [-0.1591, -0.1366, -0.1692, ..., -0.2197, 0.0283, -0.2010], + ..., + [-0.1921, -0.2078, 0.0071, ..., -0.2148, -0.0979, -0.3673], + [ 0.1339, 0.0352, -0.1334, ..., -0.2082, 0.0177, -0.1953], + [ 0.1298, 0.0835, -0.1176, ..., 0.0840, -0.2123, -0.2130]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, -1.5832e-08, ..., -3.7253e-09, + 5.5879e-09, 5.5879e-09], + [ 9.1270e-08, 4.7497e-08, 9.3132e-10, ..., -3.6322e-08, + -1.3970e-08, -4.5635e-08], + [ 5.4948e-08, 3.3528e-08, 5.5879e-09, ..., 9.3132e-09, + 2.3283e-08, 6.5193e-09], + ..., + [-1.8580e-06, -9.4902e-07, 9.3132e-10, ..., 3.7253e-09, + 2.7940e-09, 4.6566e-09], + [ 9.0338e-08, 4.1910e-08, -1.8626e-09, ..., 5.5879e-09, + -3.8184e-08, 1.8626e-09], + [ 1.4333e-06, 7.3947e-07, 1.4529e-07, ..., 8.3819e-09, + 1.4715e-07, 2.7940e-09]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0243, -0.0311, -0.0118, 0.0016, 0.0007, 0.0028, 0.0034, -0.0117, + 0.0138, -0.0430], device='cuda:0'), grad: tensor([ 7.2643e-08, 9.4064e-07, 3.6135e-07, 9.9652e-07, -1.4901e-07, + 2.9337e-07, 4.0978e-08, -1.9684e-05, 1.5227e-06, 1.5587e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 217.64, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4732 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.20 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.0596, 0.0073, 0.1220, ..., 0.0135, -0.0752, -0.0660], + [-0.1804, -0.1944, -0.1841, ..., 0.0766, 0.0521, 0.3745], + [-0.1612, -0.1374, -0.1694, ..., -0.2206, 0.0282, -0.2025], + ..., + [-0.1922, -0.2080, 0.0070, ..., -0.2149, -0.0981, -0.3675], + [ 0.1345, 0.0354, -0.1333, ..., -0.2077, 0.0180, -0.1968], + [ 0.1311, 0.0840, -0.1179, ..., 0.0845, -0.2130, -0.2138]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 1.0245e-08, 0.0000e+00, ..., 4.8429e-08, + 2.2352e-08, 2.8871e-08], + [ 1.3970e-08, 8.3819e-09, 9.3132e-10, ..., -3.1106e-07, + -3.3621e-07, -6.4634e-07], + [ 1.1269e-07, 2.9802e-08, 0.0000e+00, ..., 4.3772e-08, + 1.4994e-07, 6.8918e-08], + ..., + [ 1.3225e-07, 2.2352e-08, -9.3132e-09, ..., 1.5646e-07, + -9.9652e-08, 4.4703e-08], + [-1.0245e-07, -2.4214e-08, 0.0000e+00, ..., 1.7136e-07, + 1.0617e-07, 3.3248e-07], + [-2.8405e-07, -5.1223e-08, 0.0000e+00, ..., -2.9244e-07, + 1.3039e-08, 5.5879e-09]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0246, -0.0314, -0.0120, 0.0016, 0.0004, 0.0028, 0.0036, -0.0116, + 0.0144, -0.0429], device='cuda:0'), grad: tensor([ 2.3469e-07, -5.6624e-07, 5.0850e-07, 2.1420e-08, 6.5379e-07, + 1.4622e-07, 2.0396e-07, -1.8720e-07, 9.6858e-08, -1.0952e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 217.66, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4787 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.15 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.0607, 0.0067, 0.1219, ..., 0.0133, -0.0758, -0.0662], + [-0.1805, -0.1950, -0.1843, ..., 0.0767, 0.0517, 0.3750], + [-0.1626, -0.1377, -0.1695, ..., -0.2209, 0.0285, -0.2039], + ..., + [-0.1925, -0.2083, 0.0068, ..., -0.2150, -0.0983, -0.3675], + [ 0.1348, 0.0354, -0.1334, ..., -0.2080, 0.0179, -0.1972], + [ 0.1316, 0.0842, -0.1180, ..., 0.0847, -0.2132, -0.2141]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 5.5879e-09, + 5.5879e-09, 8.3819e-09], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., -1.6112e-07, + -1.6764e-08, -3.7067e-07], + [ 6.5193e-09, 5.5879e-09, 9.3132e-10, ..., 2.2352e-08, + -1.6484e-07, 3.4459e-08], + ..., + [ 6.5193e-09, 5.5879e-09, 0.0000e+00, ..., 8.4750e-08, + 7.8231e-08, 1.6391e-07], + [ 2.5146e-08, 4.1910e-08, 9.3132e-10, ..., 5.0291e-08, + 7.7300e-08, 1.0617e-07], + [-1.4901e-08, -2.7940e-09, 0.0000e+00, ..., 8.3819e-09, + 5.5879e-09, 2.5146e-08]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0249, -0.0322, -0.0114, 0.0016, 0.0003, 0.0028, 0.0038, -0.0115, + 0.0145, -0.0428], device='cuda:0'), grad: tensor([ 7.1712e-08, 9.8720e-08, -2.3283e-06, 3.6322e-08, 9.1270e-08, + -4.6566e-09, 8.9407e-08, 7.4785e-07, 1.0794e-06, 9.1270e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 217.59, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4576 re_mapping 0.0038 re_causal 0.0111 /// teacc 99.11 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.0611, 0.0066, 0.1218, ..., 0.0131, -0.0759, -0.0670], + [-0.1795, -0.1954, -0.1843, ..., 0.0770, 0.0517, 0.3777], + [-0.1633, -0.1379, -0.1696, ..., -0.2215, 0.0284, -0.2055], + ..., + [-0.1936, -0.2087, 0.0067, ..., -0.2152, -0.0983, -0.3700], + [ 0.1350, 0.0354, -0.1335, ..., -0.2082, 0.0181, -0.1981], + [ 0.1321, 0.0844, -0.1180, ..., 0.0850, -0.2133, -0.2147]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 3.6322e-08, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 1.8626e-09], + [ 3.2596e-08, 3.8184e-08, 0.0000e+00, ..., -4.4703e-08, + -2.7940e-09, -1.1362e-07], + [ 1.0524e-07, 1.0990e-07, 0.0000e+00, ..., 1.6764e-08, + 4.6566e-09, 1.6764e-08], + ..., + [ 2.5146e-08, 3.0734e-08, 0.0000e+00, ..., 4.3772e-08, + 3.2596e-08, 6.8918e-08], + [-1.2480e-07, -7.4506e-08, 9.3132e-10, ..., 8.3819e-09, + 9.3132e-10, 6.5193e-09], + [ 3.7253e-09, 1.7695e-08, 1.8626e-09, ..., 7.4506e-09, + 3.7253e-09, 1.0245e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0251, -0.0313, -0.0119, 0.0016, 0.0001, 0.0027, 0.0044, -0.0124, + 0.0155, -0.0426], device='cuda:0'), grad: tensor([ 1.5181e-07, 6.5193e-08, -1.2834e-06, 1.3262e-06, 3.8184e-08, + -1.7136e-06, 1.6484e-07, 1.5758e-06, -4.8522e-07, 1.7975e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 217.33, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4870 re_mapping 0.0036 re_causal 0.0112 /// teacc 99.19 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.0616, 0.0065, 0.1217, ..., 0.0130, -0.0760, -0.0673], + [-0.1795, -0.1968, -0.1853, ..., 0.0772, 0.0518, 0.3781], + [-0.1640, -0.1381, -0.1700, ..., -0.2221, 0.0283, -0.2064], + ..., + [-0.1938, -0.2088, 0.0070, ..., -0.2153, -0.0984, -0.3702], + [ 0.1351, 0.0356, -0.1337, ..., -0.2084, 0.0182, -0.2001], + [ 0.1324, 0.0845, -0.1185, ..., 0.0850, -0.2149, -0.2154]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 1.8626e-08, -1.3970e-08, ..., 4.6566e-09, + 1.2107e-08, 2.7940e-09], + [ 2.5146e-08, 2.7008e-08, 4.6566e-09, ..., -3.8184e-08, + 0.0000e+00, -1.2107e-07], + [ 1.6205e-07, 1.3784e-07, 3.7253e-09, ..., 3.4459e-08, + -6.5193e-09, 2.6077e-08], + ..., + [ 9.3132e-09, 1.0245e-08, 7.4506e-09, ..., 2.7008e-08, + 3.0734e-08, 4.0047e-08], + [-2.9523e-07, -7.9162e-08, 3.7253e-09, ..., -7.4506e-09, + -6.9849e-08, 1.1176e-08], + [ 5.4017e-08, 6.9849e-08, 8.5682e-08, ..., 2.6077e-08, + 9.6858e-08, 1.1176e-08]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0253, -0.0312, -0.0124, 0.0016, 0.0006, 0.0032, 0.0024, -0.0124, + 0.0162, -0.0430], device='cuda:0'), grad: tensor([ 1.0617e-07, -6.5193e-09, 4.5635e-08, 1.8906e-06, -2.3190e-07, + -2.1532e-06, 1.9744e-07, 5.6811e-08, -4.5076e-07, 5.8021e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 217.52, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4505 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.18 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.0619, 0.0068, 0.1218, ..., 0.0130, -0.0760, -0.0673], + [-0.1797, -0.1972, -0.1854, ..., 0.0772, 0.0517, 0.3786], + [-0.1650, -0.1384, -0.1702, ..., -0.2225, 0.0284, -0.2073], + ..., + [-0.1941, -0.2093, 0.0068, ..., -0.2154, -0.0986, -0.3704], + [ 0.1353, 0.0356, -0.1340, ..., -0.2088, 0.0183, -0.2004], + [ 0.1333, 0.0847, -0.1189, ..., 0.0853, -0.2159, -0.2158]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, 8.3819e-09, ..., 8.3819e-09, + 4.6566e-09, 3.7253e-09], + [ 4.6566e-09, 9.3132e-10, -3.2596e-08, ..., -1.1083e-07, + -4.6566e-08, -1.4249e-07], + [ 3.7253e-09, 2.7940e-09, 3.7253e-09, ..., 5.5879e-09, + 8.3819e-09, 4.6566e-09], + ..., + [ 1.0245e-08, 5.5879e-09, 2.8871e-08, ..., 1.8626e-08, + 5.5879e-08, 1.3039e-08], + [-2.7940e-09, 9.3132e-10, 3.7253e-09, ..., 1.0245e-08, + -9.3132e-10, 6.5193e-09], + [-1.9558e-08, -5.5879e-09, 7.5437e-08, ..., 1.2107e-08, + 1.4715e-07, 3.2596e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0252, -0.0314, -0.0122, 0.0015, 0.0006, 0.0032, 0.0026, -0.0124, + 0.0164, -0.0431], device='cuda:0'), grad: tensor([ 5.4017e-08, -2.0582e-07, -2.0023e-07, -2.6077e-08, 1.3504e-07, + -2.2352e-08, -2.5798e-07, 1.8161e-07, 5.6811e-08, 2.8778e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 217.47, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4566 re_mapping 0.0035 re_causal 0.0108 /// teacc 99.19 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.0626, 0.0068, 0.1219, ..., 0.0130, -0.0762, -0.0674], + [-0.1801, -0.1980, -0.1855, ..., 0.0773, 0.0515, 0.3791], + [-0.1653, -0.1386, -0.1704, ..., -0.2228, 0.0284, -0.2075], + ..., + [-0.1940, -0.2094, 0.0070, ..., -0.2154, -0.0983, -0.3706], + [ 0.1353, 0.0353, -0.1342, ..., -0.2095, 0.0183, -0.2021], + [ 0.1337, 0.0850, -0.1189, ..., 0.0855, -0.2161, -0.2165]], + device='cuda:0'), grad: tensor([[-2.2352e-08, -5.7742e-08, -3.6322e-08, ..., -1.9558e-08, + 0.0000e+00, 9.3132e-10], + [ 1.8626e-09, 1.5832e-08, 6.7987e-08, ..., -2.7940e-09, + -1.8626e-09, -1.5832e-08], + [ 5.5879e-09, 5.5879e-09, 8.3819e-09, ..., 2.7940e-09, + 6.5193e-09, 3.7253e-09], + ..., + [ 4.6566e-09, -1.9558e-08, -1.4435e-07, ..., 5.5879e-09, + 0.0000e+00, 2.7940e-09], + [ 6.5193e-09, 2.7940e-08, 2.5146e-08, ..., 1.3970e-08, + -4.6566e-09, 2.7940e-09], + [-2.7940e-09, 2.7008e-08, 4.9360e-08, ..., 9.3132e-10, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0254, -0.0322, -0.0120, 0.0017, 0.0003, 0.0030, 0.0028, -0.0118, + 0.0161, -0.0430], device='cuda:0'), grad: tensor([ 6.7987e-08, 5.7928e-07, 3.6322e-08, 7.4506e-09, 1.1548e-07, + 3.6322e-08, -4.5635e-08, -1.0170e-06, 8.1956e-08, 1.3970e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 217.44, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4895 re_mapping 0.0036 re_causal 0.0112 /// teacc 99.10 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.0626, 0.0069, 0.1216, ..., 0.0128, -0.0762, -0.0674], + [-0.1802, -0.1986, -0.1856, ..., 0.0773, 0.0516, 0.3797], + [-0.1679, -0.1410, -0.1707, ..., -0.2234, 0.0283, -0.2093], + ..., + [-0.1946, -0.2106, 0.0069, ..., -0.2156, -0.0983, -0.3707], + [ 0.1352, 0.0351, -0.1344, ..., -0.2099, 0.0183, -0.2025], + [ 0.1344, 0.0856, -0.1191, ..., 0.0858, -0.2164, -0.2173]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.7253e-09, -2.7940e-09, ..., 1.8626e-09, + 4.6566e-09, 1.8626e-09], + [ 1.0245e-08, 1.4901e-08, 9.3132e-10, ..., -1.3970e-08, + 4.6566e-09, -3.9116e-08], + [ 1.2107e-08, 1.7695e-08, 1.8626e-09, ..., 1.2107e-08, + 1.8626e-08, 9.3132e-09], + ..., + [ 4.6566e-09, 7.4506e-09, 4.8429e-08, ..., 5.5879e-09, + 1.1642e-07, 3.7253e-09], + [ 5.5879e-09, 4.7591e-07, 0.0000e+00, ..., 1.8626e-08, + -1.8626e-09, 1.8626e-08], + [-1.2107e-08, -1.8626e-09, 8.7544e-08, ..., -1.0245e-08, + 1.9837e-07, 1.8626e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0257, -0.0322, -0.0124, 0.0018, 0.0003, 0.0031, 0.0031, -0.0117, + 0.0160, -0.0430], device='cuda:0'), grad: tensor([ 1.5832e-08, 2.5146e-08, 7.6368e-08, -2.2817e-07, -5.4110e-07, + -9.4436e-07, 1.0710e-07, 9.1270e-08, 1.0123e-06, 3.8557e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 217.77, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4882 re_mapping 0.0039 re_causal 0.0111 /// teacc 99.17 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.0629, 0.0066, 0.1213, ..., 0.0125, -0.0764, -0.0677], + [-0.1805, -0.2001, -0.1861, ..., 0.0773, 0.0517, 0.3807], + [-0.1685, -0.1413, -0.1712, ..., -0.2239, 0.0282, -0.2122], + ..., + [-0.1950, -0.2106, 0.0075, ..., -0.2156, -0.0982, -0.3709], + [ 0.1352, 0.0349, -0.1347, ..., -0.2102, 0.0183, -0.2034], + [ 0.1351, 0.0861, -0.1194, ..., 0.0860, -0.2169, -0.2180]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-09, 1.8626e-09, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00], + [ 8.3819e-09, 1.4901e-08, 0.0000e+00, ..., 9.3132e-10, + 5.5879e-09, -5.5879e-09], + [ 4.6566e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -3.3528e-08, 1.8626e-09], + ..., + [ 4.1723e-07, 3.7998e-07, 0.0000e+00, ..., 1.6857e-07, + 9.3132e-09, 1.8626e-09], + [-3.1665e-08, 1.4715e-07, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [-2.5146e-08, 5.3085e-08, 9.3132e-10, ..., -2.9802e-08, + 7.4506e-09, 9.3132e-10]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0261, -0.0324, -0.0131, 0.0019, -0.0002, 0.0031, 0.0035, -0.0109, + 0.0158, -0.0430], device='cuda:0'), grad: tensor([ 4.2841e-08, 1.2666e-07, -4.4517e-07, -2.8312e-07, 6.6124e-08, + -1.6224e-06, 3.5204e-07, 1.2200e-06, 4.3865e-07, 9.2201e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 217.66, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4526 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.14 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.0627, 0.0061, 0.1212, ..., 0.0121, -0.0765, -0.0678], + [-0.1807, -0.2007, -0.1864, ..., 0.0775, 0.0518, 0.3814], + [-0.1691, -0.1418, -0.1715, ..., -0.2245, 0.0282, -0.2125], + ..., + [-0.1954, -0.2114, 0.0080, ..., -0.2159, -0.0985, -0.3712], + [ 0.1353, 0.0348, -0.1349, ..., -0.2106, 0.0185, -0.2037], + [ 0.1362, 0.0866, -0.1199, ..., 0.0870, -0.2181, -0.2197]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -5.2247e-07, -8.8476e-07, ..., -2.4308e-07, + 0.0000e+00, -5.5879e-09], + [ 1.8626e-09, 6.0536e-08, 9.9652e-08, ..., 2.5146e-08, + 0.0000e+00, -6.5193e-09], + [ 2.7940e-09, 3.7253e-08, 5.7742e-08, ..., 1.6764e-08, + 1.8626e-09, 9.3132e-10], + ..., + [ 9.3132e-10, 9.3132e-09, 1.3970e-08, ..., 7.4506e-09, + 9.3132e-10, 3.7253e-09], + [-7.4506e-09, 3.3248e-07, 5.6531e-07, ..., 1.3970e-07, + -6.5193e-09, 5.5879e-09], + [-2.7940e-09, 6.9849e-08, 9.7789e-08, ..., 3.5390e-08, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0264, -0.0324, -0.0131, 0.0017, -0.0006, 0.0033, 0.0037, -0.0109, + 0.0159, -0.0430], device='cuda:0'), grad: tensor([-1.9055e-06, 2.7195e-07, -2.9430e-07, 1.0990e-07, 3.8184e-08, + -2.7008e-08, 5.9605e-08, -3.4459e-08, 1.3094e-06, 4.7125e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 217.51, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.5045 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.16 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.0631, 0.0060, 0.1212, ..., 0.0120, -0.0767, -0.0677], + [-0.1808, -0.2012, -0.1866, ..., 0.0777, 0.0522, 0.3822], + [-0.1693, -0.1419, -0.1718, ..., -0.2251, 0.0281, -0.2140], + ..., + [-0.1957, -0.2121, 0.0079, ..., -0.2161, -0.0988, -0.3716], + [ 0.1354, 0.0346, -0.1353, ..., -0.2110, 0.0185, -0.2040], + [ 0.1366, 0.0867, -0.1200, ..., 0.0872, -0.2185, -0.2209]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 9.3132e-09, 0.0000e+00, ..., 1.8626e-09, + 7.4506e-09, 9.3132e-10], + [ 9.0338e-08, 6.1467e-08, 6.5193e-09, ..., 9.3132e-08, + 8.0094e-08, -1.8626e-09], + [ 4.1910e-08, 5.0291e-08, 9.3132e-09, ..., 1.8626e-08, + 3.5390e-08, 9.3132e-10], + ..., + [ 3.1665e-08, 2.3283e-08, 2.1420e-08, ..., 2.7940e-08, + 4.3772e-08, 9.3132e-10], + [-2.7940e-08, -4.9360e-08, 6.5193e-09, ..., 1.4901e-08, + -2.5146e-08, 9.3132e-10], + [-8.3819e-09, 9.3132e-09, 4.6566e-09, ..., -8.3819e-09, + 6.5193e-09, 9.3132e-10]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0265, -0.0323, -0.0128, 0.0017, -0.0007, 0.0034, 0.0034, -0.0111, + 0.0158, -0.0431], device='cuda:0'), grad: tensor([ 1.0338e-07, 1.0878e-06, 4.0513e-07, -5.0291e-08, -1.1269e-07, + -2.7195e-07, -8.5682e-08, -1.0300e-06, -2.2445e-07, 1.9930e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 217.80, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4675 re_mapping 0.0039 re_causal 0.0111 /// teacc 99.12 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.0634, 0.0063, 0.1213, ..., 0.0121, -0.0768, -0.0676], + [-0.1823, -0.2017, -0.1868, ..., 0.0779, 0.0522, 0.3828], + [-0.1696, -0.1420, -0.1722, ..., -0.2255, 0.0282, -0.2151], + ..., + [-0.1926, -0.2121, 0.0080, ..., -0.2152, -0.0990, -0.3717], + [ 0.1354, 0.0345, -0.1357, ..., -0.2119, 0.0186, -0.2046], + [ 0.1338, 0.0876, -0.1200, ..., 0.0865, -0.2187, -0.2214]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 7.4506e-09, 1.8626e-09, ..., 6.6124e-08, + 2.8871e-08, 8.9407e-08], + [-4.3772e-08, 3.7253e-09, 9.3132e-10, ..., -2.3935e-07, + -1.0990e-07, -3.6974e-07], + [ 3.2596e-08, 2.3283e-08, 1.0245e-08, ..., 3.6322e-08, + 2.2352e-08, 5.4017e-08], + ..., + [ 1.7695e-08, 1.0245e-08, 0.0000e+00, ..., 1.7695e-08, + 4.6566e-09, 1.6764e-08], + [-6.7987e-08, -1.5926e-07, -9.0338e-08, ..., 9.1270e-08, + 5.5879e-09, 1.3318e-07], + [-2.4214e-08, 3.3528e-08, 2.7940e-09, ..., -1.8626e-08, + 4.6566e-09, 1.3039e-08]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0265, -0.0333, -0.0127, 0.0018, -0.0015, 0.0033, 0.0037, -0.0079, + 0.0157, -0.0460], device='cuda:0'), grad: tensor([ 2.4680e-07, -5.8673e-07, 2.6356e-07, 0.0000e+00, 6.3330e-08, + 2.1420e-07, 4.0419e-07, -4.0419e-07, -4.0419e-07, 2.0303e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 217.36, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4776 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.16 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.0647, 0.0047, 0.1212, ..., 0.0119, -0.0770, -0.0679], + [-0.1823, -0.2019, -0.1869, ..., 0.0781, 0.0525, 0.3838], + [-0.1719, -0.1424, -0.1728, ..., -0.2263, 0.0280, -0.2161], + ..., + [-0.1927, -0.2127, 0.0080, ..., -0.2153, -0.0991, -0.3722], + [ 0.1357, 0.0341, -0.1354, ..., -0.2123, 0.0188, -0.2050], + [ 0.1339, 0.0877, -0.1202, ..., 0.0868, -0.2190, -0.2236]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -4.6566e-09, -9.3132e-09, ..., -1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., -1.8626e-09, + 9.3132e-09, -4.6566e-09], + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 9.3132e-10, + -8.8476e-08, 9.3132e-10], + ..., + [ 2.7940e-09, 1.8626e-09, 9.3132e-10, ..., 2.7940e-09, + 5.2154e-08, 2.7940e-09], + [-1.8626e-09, 4.3772e-08, 4.6566e-09, ..., 7.4506e-09, + 1.6764e-08, 0.0000e+00], + [-4.2841e-08, -1.3970e-08, 5.5879e-09, ..., -2.3283e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0269, -0.0332, -0.0132, 0.0018, -0.0017, 0.0035, 0.0035, -0.0079, + 0.0161, -0.0460], device='cuda:0'), grad: tensor([-5.5879e-09, 1.0617e-07, -8.6706e-07, 6.2399e-08, 1.0431e-07, + -1.5926e-07, 7.9162e-08, 5.0664e-07, 2.5518e-07, -8.4750e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 217.45, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4844 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.04 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.0652, 0.0039, 0.1215, ..., 0.0111, -0.0772, -0.0681], + [-0.1827, -0.2027, -0.1871, ..., 0.0781, 0.0525, 0.3842], + [-0.1726, -0.1427, -0.1733, ..., -0.2267, 0.0280, -0.2166], + ..., + [-0.1927, -0.2138, 0.0077, ..., -0.2155, -0.0994, -0.3723], + [ 0.1354, 0.0337, -0.1357, ..., -0.2132, 0.0191, -0.2051], + [ 0.1340, 0.0888, -0.1207, ..., 0.0882, -0.2195, -0.2241]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 8.3819e-09, 0.0000e+00, ..., 1.8626e-09, + 5.5879e-09, 9.3132e-10], + [ 1.3690e-07, 1.8440e-07, -9.3132e-10, ..., 3.4459e-08, + 1.1083e-07, -2.4214e-08], + [ 2.4214e-08, 3.1665e-08, 9.3132e-10, ..., 7.4506e-09, + 2.3283e-08, 9.3132e-10], + ..., + [ 1.5181e-07, 1.5553e-07, 5.5879e-09, ..., 8.6613e-08, + 8.4750e-08, 1.5832e-08], + [ 7.4506e-08, 2.2072e-07, 0.0000e+00, ..., 3.9116e-08, + 9.4995e-08, 1.8626e-09], + [ 2.5146e-08, 2.3283e-08, 9.3132e-09, ..., 1.7695e-08, + 1.6764e-08, 1.8626e-09]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0274, -0.0333, -0.0128, 0.0018, -0.0021, 0.0035, 0.0032, -0.0079, + 0.0162, -0.0459], device='cuda:0'), grad: tensor([ 2.3283e-08, 4.2282e-07, 6.0536e-08, -1.3243e-06, -9.3132e-10, + -6.1095e-07, 4.0606e-07, 4.6473e-07, 4.5635e-07, 9.6858e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 217.52, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4905 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.10 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.0656, 0.0041, 0.1216, ..., 0.0111, -0.0775, -0.0683], + [-0.1830, -0.2040, -0.1878, ..., 0.0780, 0.0523, 0.3846], + [-0.1728, -0.1429, -0.1736, ..., -0.2271, 0.0280, -0.2170], + ..., + [-0.1928, -0.2146, 0.0070, ..., -0.2157, -0.0997, -0.3725], + [ 0.1353, 0.0332, -0.1359, ..., -0.2137, 0.0191, -0.2054], + [ 0.1341, 0.0882, -0.1214, ..., 0.0883, -0.2220, -0.2244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2387e-07, -1.6484e-07, ..., -1.2480e-07, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 6.5193e-09, ..., -2.7940e-09, + 4.6566e-09, -1.0245e-08], + [ 0.0000e+00, 1.0431e-07, 1.9930e-07, ..., 1.0431e-07, + 6.0536e-08, 2.7940e-09], + ..., + [ 5.5879e-09, 5.5879e-09, 1.5832e-08, ..., 1.0245e-08, + 1.6764e-08, 9.3132e-09], + [-9.3132e-09, -6.5193e-09, 1.8626e-09, ..., 1.8626e-09, + -5.5879e-09, 1.8626e-09], + [-5.5879e-09, 1.0245e-08, 1.1371e-06, ..., 7.4506e-09, + 1.2098e-06, 0.0000e+00]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0274, -0.0335, -0.0125, 0.0018, -0.0011, 0.0037, 0.0030, -0.0080, + 0.0160, -0.0460], device='cuda:0'), grad: tensor([-8.2795e-07, 6.6124e-08, 1.4594e-06, 4.0978e-08, -1.4991e-05, + 2.3283e-08, 2.4214e-08, 2.0117e-07, -1.7695e-08, 1.4015e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 217.93, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4798 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.11 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.0663, 0.0041, 0.1216, ..., 0.0110, -0.0777, -0.0685], + [-0.1831, -0.2042, -0.1880, ..., 0.0780, 0.0524, 0.3852], + [-0.1728, -0.1430, -0.1737, ..., -0.2276, 0.0280, -0.2175], + ..., + [-0.1928, -0.2152, 0.0069, ..., -0.2159, -0.0997, -0.3728], + [ 0.1352, 0.0330, -0.1360, ..., -0.2142, 0.0191, -0.2058], + [ 0.1343, 0.0888, -0.1217, ..., 0.0886, -0.2228, -0.2255]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.4901e-08, -2.1420e-08, ..., -1.5832e-08, + 2.7940e-09, 2.7940e-09], + [ 2.7940e-09, 3.7253e-09, 0.0000e+00, ..., -2.2352e-08, + -4.6566e-09, -6.5193e-08], + [ 1.4901e-08, 8.3819e-09, 9.3132e-10, ..., 1.3039e-08, + 8.3819e-09, 2.5146e-08], + ..., + [ 3.7253e-09, -3.7253e-09, 9.3132e-10, ..., 1.1176e-08, + 4.6566e-09, 2.7940e-08], + [-1.2852e-07, -7.5437e-08, 9.3132e-10, ..., 6.5193e-09, + -4.0047e-08, 4.6566e-09], + [ 3.8184e-08, 4.0978e-08, 1.3970e-08, ..., 1.3970e-08, + 1.2107e-08, 9.3132e-10]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0276, -0.0336, -0.0121, 0.0017, -0.0011, 0.0037, 0.0032, -0.0081, + 0.0158, -0.0460], device='cuda:0'), grad: tensor([-3.3528e-08, 4.3772e-08, 2.8685e-07, 3.2317e-07, 1.5832e-08, + -2.7940e-09, -1.6764e-08, -6.4634e-07, -3.0547e-07, 3.3900e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 272---------------------------------------------------- +epoch 272, time 218.24, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4843 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.23 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.0671, 0.0043, 0.1217, ..., 0.0111, -0.0778, -0.0687], + [-0.1833, -0.2046, -0.1882, ..., 0.0782, 0.0525, 0.3856], + [-0.1737, -0.1433, -0.1740, ..., -0.2283, 0.0278, -0.2185], + ..., + [-0.1929, -0.2157, 0.0069, ..., -0.2164, -0.0999, -0.3729], + [ 0.1356, 0.0331, -0.1362, ..., -0.2145, 0.0193, -0.2058], + [ 0.1345, 0.0889, -0.1219, ..., 0.0893, -0.2243, -0.2258]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2107e-08, -1.1176e-08, ..., -1.3970e-08, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 2.7940e-09, 1.8626e-09, ..., 2.7940e-09, + -1.8626e-09, -4.6566e-09], + [ 0.0000e+00, 1.8626e-09, 4.6566e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 1.3970e-08, 4.6566e-09, 5.1223e-08, ..., 1.8626e-08, + 4.0047e-08, 3.7253e-09], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [-1.6764e-08, -1.8626e-09, 1.0245e-08, ..., -1.8626e-08, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0277, -0.0337, -0.0124, 0.0015, -0.0012, 0.0039, 0.0034, -0.0079, + 0.0160, -0.0460], device='cuda:0'), grad: tensor([-5.4017e-08, 1.1176e-08, -6.2399e-07, 5.5879e-09, -1.5739e-07, + 1.9558e-08, -1.4901e-08, 7.7114e-07, 1.6764e-08, 2.1420e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 217.23, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.4962 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.19 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.0679, 0.0044, 0.1222, ..., 0.0112, -0.0779, -0.0690], + [-0.1833, -0.2049, -0.1891, ..., 0.0785, 0.0528, 0.3872], + [-0.1747, -0.1440, -0.1745, ..., -0.2294, 0.0277, -0.2204], + ..., + [-0.1930, -0.2181, 0.0058, ..., -0.2171, -0.1005, -0.3738], + [ 0.1360, 0.0334, -0.1364, ..., -0.2149, 0.0195, -0.2067], + [ 0.1348, 0.0897, -0.1224, ..., 0.0900, -0.2247, -0.2277]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.8626e-09, -6.5193e-09, ..., -3.7253e-09, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., -3.7253e-09, + -2.1420e-08, -4.5635e-08], + [ 2.7940e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + ..., + [ 7.4506e-09, 1.0245e-08, 0.0000e+00, ..., 4.6566e-09, + 2.0489e-08, 3.8184e-08], + [ 1.1176e-08, -1.1176e-08, 0.0000e+00, ..., 1.6764e-08, + -1.3039e-08, 0.0000e+00], + [-9.6858e-08, -6.4261e-08, 8.3819e-09, ..., -5.3085e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0274, -0.0334, -0.0124, 0.0013, -0.0011, 0.0039, 0.0034, -0.0083, + 0.0165, -0.0460], device='cuda:0'), grad: tensor([ 1.0245e-07, 3.7253e-09, -2.1551e-06, 2.6450e-07, 3.1199e-07, + 1.2293e-07, 6.7055e-07, 6.9290e-07, 7.3574e-08, -8.1956e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 217.46, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4772 re_mapping 0.0036 re_causal 0.0109 /// teacc 99.14 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.0685, 0.0048, 0.1228, ..., 0.0115, -0.0778, -0.0692], + [-0.1833, -0.2051, -0.1897, ..., 0.0790, 0.0530, 0.3888], + [-0.1755, -0.1444, -0.1748, ..., -0.2299, 0.0277, -0.2213], + ..., + [-0.1933, -0.2205, 0.0054, ..., -0.2176, -0.1008, -0.3746], + [ 0.1359, 0.0333, -0.1366, ..., -0.2154, 0.0196, -0.2076], + [ 0.1346, 0.0886, -0.1228, ..., 0.0896, -0.2250, -0.2334]], + device='cuda:0'), grad: tensor([[ 8.1025e-08, 8.8476e-08, -1.7043e-07, ..., -9.5926e-08, + 4.6566e-09, -2.7940e-09], + [ 5.0291e-08, 5.6811e-08, 5.0291e-08, ..., 2.7008e-08, + 7.4506e-09, 0.0000e+00], + [ 7.4506e-08, 8.4750e-08, 7.4506e-09, ..., 4.6566e-09, + 1.1176e-08, 9.3132e-10], + ..., + [ 4.7497e-08, 5.0291e-08, -1.1176e-08, ..., 9.3132e-10, + 1.4901e-08, 9.3132e-10], + [ 1.2703e-06, 1.5041e-06, 4.5635e-08, ..., 2.7940e-08, + 2.7008e-08, 0.0000e+00], + [ 1.1548e-07, 1.3877e-07, 3.6322e-08, ..., 2.7940e-09, + 8.3819e-09, 1.8626e-09]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0270, -0.0332, -0.0121, -0.0009, -0.0013, 0.0064, 0.0035, -0.0086, + 0.0165, -0.0463], device='cuda:0'), grad: tensor([-2.6356e-07, 1.0012e-06, 4.1351e-07, -7.9647e-06, 1.0245e-07, + 3.4794e-06, 1.4622e-07, -1.2685e-06, 3.5875e-06, 7.4785e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 217.44, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4947 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.13 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.0703, 0.0053, 0.1235, ..., 0.0118, -0.0778, -0.0692], + [-0.1835, -0.2058, -0.1909, ..., 0.0791, 0.0530, 0.3895], + [-0.1765, -0.1449, -0.1754, ..., -0.2308, 0.0276, -0.2218], + ..., + [-0.1935, -0.2226, 0.0051, ..., -0.2182, -0.1009, -0.3750], + [ 0.1364, 0.0334, -0.1372, ..., -0.2162, 0.0198, -0.2082], + [ 0.1345, 0.0878, -0.1235, ..., 0.0899, -0.2266, -0.2347]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -4.6566e-09, -6.1467e-08, ..., -2.5146e-08, + 2.7940e-09, 0.0000e+00], + [ 5.1223e-08, 4.2841e-08, 2.7940e-09, ..., 3.2596e-08, + 6.7055e-08, -2.7940e-09], + [ 1.5087e-07, 1.6112e-07, 5.5879e-09, ..., 9.3132e-10, + -1.0524e-07, 0.0000e+00], + ..., + [ 1.0338e-07, 9.1270e-08, -3.7253e-09, ..., 4.6566e-08, + 2.4214e-08, 2.7940e-09], + [-3.8184e-07, -4.0233e-07, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [-1.7323e-07, -1.1362e-07, 8.3819e-09, ..., -2.0117e-07, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0264, -0.0333, -0.0121, -0.0008, -0.0005, 0.0062, 0.0038, -0.0088, + 0.0166, -0.0467], device='cuda:0'), grad: tensor([-2.2352e-08, 1.1772e-06, -8.9034e-07, 1.6019e-07, 2.7381e-07, + 9.5274e-07, 4.3400e-07, 1.7453e-06, -3.5390e-06, -3.0454e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 217.60, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4690 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.13 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.0714, 0.0060, 0.1242, ..., 0.0120, -0.0781, -0.0693], + [-0.1836, -0.2073, -0.1913, ..., 0.0794, 0.0534, 0.3905], + [-0.1778, -0.1454, -0.1759, ..., -0.2319, 0.0273, -0.2248], + ..., + [-0.1936, -0.2237, 0.0049, ..., -0.2186, -0.1012, -0.3755], + [ 0.1392, 0.0363, -0.1378, ..., -0.2150, 0.0203, -0.2089], + [ 0.1346, 0.0876, -0.1242, ..., 0.0901, -0.2274, -0.2356]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.5146e-08, -5.1223e-09, ..., 5.5879e-09, + 1.3970e-09, 4.6566e-10], + [ 1.3877e-07, 1.0571e-07, 3.2596e-09, ..., 7.2177e-08, + 1.2387e-07, -4.6566e-09], + [ 2.3097e-07, 1.7742e-07, 1.8626e-09, ..., 1.2154e-07, + 2.1048e-07, 9.3132e-10], + ..., + [ 3.0734e-08, 2.0023e-08, 0.0000e+00, ..., 2.0955e-08, + 1.5832e-08, 2.7940e-09], + [ 1.1176e-07, 8.5216e-08, 2.9802e-08, ..., 9.7789e-08, + 1.0151e-07, 2.3283e-09], + [-7.1246e-08, -1.3970e-08, 2.3749e-08, ..., -4.0047e-08, + 4.6566e-09, 4.6566e-10]], device='cuda:0') +Epoch 278, bias, value: tensor([-2.6068e-02, -3.3088e-02, -1.2604e-02, -7.4336e-04, 3.1777e-05, + 5.7711e-03, 1.9292e-03, -8.8935e-03, 1.9330e-02, -4.6858e-02], + device='cuda:0'), grad: tensor([ 4.6566e-08, 5.8766e-07, 9.8348e-07, -2.2240e-06, 1.1688e-07, + 1.8068e-07, -3.1898e-07, 9.2667e-08, 6.3097e-07, -8.4285e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 217.41, cls_loss 0.0006 cls_loss_mapping 0.0025 cls_loss_causal 0.4789 re_mapping 0.0039 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.0704, 0.0070, 0.1246, ..., 0.0123, -0.0785, -0.0703], + [-0.1839, -0.2080, -0.1912, ..., 0.0800, 0.0538, 0.3915], + [-0.1787, -0.1458, -0.1763, ..., -0.2329, 0.0272, -0.2264], + ..., + [-0.1937, -0.2245, 0.0046, ..., -0.2192, -0.1017, -0.3759], + [ 0.1396, 0.0365, -0.1380, ..., -0.2150, 0.0205, -0.2087], + [ 0.1339, 0.0855, -0.1252, ..., 0.0892, -0.2321, -0.2374]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -2.9802e-08, -3.9581e-08, ..., -3.1665e-08, + 1.3970e-09, 9.7789e-09], + [-2.7940e-09, 7.4506e-09, 1.8626e-09, ..., -4.3260e-07, + -1.5367e-08, -4.6566e-07], + [ 1.1921e-07, 3.0734e-08, 9.3132e-10, ..., 1.2107e-08, + 3.3062e-08, 8.8476e-09], + ..., + [ 1.3970e-08, 4.1910e-09, 9.3132e-10, ..., 2.4354e-07, + 1.4435e-08, 2.5099e-07], + [-3.2596e-09, -2.3283e-09, 4.6566e-10, ..., 1.2573e-08, + -1.3970e-09, 1.1176e-08], + [ 5.3085e-08, 5.6811e-08, 3.1199e-08, ..., 2.4913e-07, + 3.3993e-08, 1.6531e-07]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0258, -0.0329, -0.0127, -0.0005, 0.0039, 0.0055, 0.0019, -0.0091, + 0.0195, -0.0486], device='cuda:0'), grad: tensor([-8.8476e-08, -1.3057e-06, -1.5507e-07, -6.4541e-07, -2.7940e-09, + 1.6345e-07, 2.2352e-08, 1.2107e-06, 8.3353e-08, 7.1386e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 217.46, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.5053 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.15 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.0706, 0.0077, 0.1249, ..., 0.0124, -0.0790, -0.0715], + [-0.1843, -0.2089, -0.1913, ..., 0.0805, 0.0539, 0.3932], + [-0.1793, -0.1463, -0.1766, ..., -0.2340, 0.0271, -0.2269], + ..., + [-0.1937, -0.2249, 0.0045, ..., -0.2195, -0.1019, -0.3771], + [ 0.1396, 0.0366, -0.1384, ..., -0.2153, 0.0210, -0.2104], + [ 0.1340, 0.0854, -0.1255, ..., 0.0893, -0.2322, -0.2382]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -9.3132e-08, -1.7975e-07, ..., -9.3598e-08, + 4.6566e-10, -3.5390e-08], + [ 3.7253e-09, 6.9849e-09, 4.1910e-09, ..., -2.3283e-09, + 4.6566e-10, -1.7229e-08], + [ 1.7229e-08, 2.3749e-08, 3.2596e-09, ..., 2.7940e-09, + 2.9337e-08, 2.3283e-09], + ..., + [ 1.2107e-08, 7.9162e-09, 4.6566e-10, ..., 9.7789e-09, + 1.7229e-08, 1.2573e-08], + [-1.9185e-07, -2.3469e-07, 4.6566e-09, ..., 3.2596e-09, + -3.1246e-07, 1.8626e-09], + [-1.8626e-08, 1.8626e-09, 1.7229e-08, ..., -9.3132e-09, + 9.3132e-10, 4.1910e-09]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0257, -0.0324, -0.0128, -0.0004, 0.0038, 0.0055, 0.0017, -0.0092, + 0.0197, -0.0487], device='cuda:0'), grad: tensor([-3.5716e-07, 1.9325e-07, 1.6997e-07, 2.3469e-07, 4.8429e-08, + 6.7893e-07, 2.9756e-07, -2.9290e-07, -9.5554e-07, 2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 217.58, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4628 re_mapping 0.0036 re_causal 0.0109 /// teacc 99.03 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.0706, 0.0083, 0.1251, ..., 0.0126, -0.0792, -0.0720], + [-0.1845, -0.2092, -0.1913, ..., 0.0807, 0.0540, 0.3940], + [-0.1800, -0.1466, -0.1767, ..., -0.2347, 0.0269, -0.2271], + ..., + [-0.1939, -0.2259, 0.0063, ..., -0.2202, -0.1020, -0.3774], + [ 0.1398, 0.0366, -0.1387, ..., -0.2158, 0.0212, -0.2115], + [ 0.1343, 0.0855, -0.1257, ..., 0.0898, -0.2323, -0.2388]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 9.3132e-10, 1.9837e-07, ..., 4.9220e-07, + 9.3132e-10, 1.0477e-07], + [ 6.0536e-09, 3.2596e-09, 1.1409e-07, ..., 2.6403e-07, + -9.7789e-09, -5.1223e-09], + [ 2.3283e-09, 2.3283e-09, 1.6764e-08, ..., 4.3306e-08, + 2.7940e-09, 1.4435e-08], + ..., + [ 2.4214e-08, 8.8476e-09, 9.3132e-10, ..., 3.3993e-08, + 8.8476e-09, 3.3993e-08], + [ 2.7940e-09, 4.1910e-09, 2.3283e-09, ..., 1.0710e-08, + 3.7253e-09, 1.0710e-08], + [-3.6787e-08, -1.1176e-08, 1.3970e-09, ..., -3.0268e-08, + 9.3132e-10, 3.2596e-09]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0255, -0.0331, -0.0125, -0.0004, 0.0034, 0.0054, 0.0016, -0.0089, + 0.0198, -0.0486], device='cuda:0'), grad: tensor([ 1.0002e-06, 5.1269e-07, 9.7789e-08, -1.8626e-08, 3.3062e-08, + 3.1665e-08, -1.6820e-06, 5.6811e-08, 3.7719e-08, -6.9384e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 217.43, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4604 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.16 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.0752, 0.0049, 0.1232, ..., 0.0098, -0.0820, -0.0729], + [-0.1849, -0.2102, -0.1918, ..., 0.0828, 0.0558, 0.3973], + [-0.1816, -0.1471, -0.1770, ..., -0.2363, 0.0267, -0.2281], + ..., + [-0.1940, -0.2288, 0.0066, ..., -0.2213, -0.1032, -0.3781], + [ 0.1392, 0.0362, -0.1391, ..., -0.2193, 0.0213, -0.2152], + [ 0.1352, 0.0873, -0.1256, ..., 0.0938, -0.2320, -0.2395]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 1.2340e-07, ..., 7.0315e-08, + 3.5856e-08, 2.6543e-08], + [ 3.2596e-09, 4.6566e-09, 6.2399e-08, ..., 3.3528e-08, + 2.0489e-08, 6.9849e-09], + [ 1.8626e-09, 4.6566e-09, 3.0734e-08, ..., 1.9092e-08, + 1.2107e-08, 7.4506e-09], + ..., + [ 1.0245e-08, 5.5879e-09, 2.3283e-09, ..., 1.1176e-08, + 2.3283e-09, 3.7253e-09], + [ 2.7940e-09, 3.7253e-09, 1.7928e-07, ..., 1.0151e-07, + 5.3551e-08, 3.8650e-08], + [-4.3772e-08, -1.5832e-08, 1.3970e-09, ..., -3.8184e-08, + 1.3970e-09, 3.2596e-09]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0284, -0.0326, -0.0127, -0.0004, 0.0020, 0.0055, 0.0023, -0.0088, + 0.0193, -0.0475], device='cuda:0'), grad: tensor([ 4.2841e-07, 3.0641e-07, 1.4529e-07, -8.7544e-08, 1.9651e-07, + 4.8010e-07, -1.8552e-06, -4.7171e-07, 6.0955e-07, 2.6124e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 217.27, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4535 re_mapping 0.0037 re_causal 0.0107 /// teacc 98.98 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.0757, 0.0047, 0.1231, ..., 0.0096, -0.0825, -0.0730], + [-0.1853, -0.2092, -0.1924, ..., 0.0838, 0.0560, 0.3987], + [-0.1824, -0.1475, -0.1772, ..., -0.2371, 0.0265, -0.2285], + ..., + [-0.1941, -0.2308, 0.0064, ..., -0.2225, -0.1038, -0.3803], + [ 0.1389, 0.0356, -0.1395, ..., -0.2201, 0.0211, -0.2158], + [ 0.1353, 0.0872, -0.1261, ..., 0.0944, -0.2331, -0.2400]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 4.6566e-09, 5.4017e-08, ..., 7.7765e-08, + 5.8673e-08, 1.1735e-07], + [ 3.7253e-09, 2.7940e-09, 6.0396e-07, ..., 5.0990e-07, + 7.6089e-07, -4.0233e-07], + [ 2.3283e-09, 1.8626e-09, 4.3772e-08, ..., 1.9092e-08, + 4.6566e-08, 2.9802e-08], + ..., + [ 6.5193e-09, 5.1223e-09, 1.9278e-07, ..., 2.4680e-08, + 2.0768e-07, 5.4482e-08], + [ 8.3819e-09, 1.3970e-08, 2.1560e-07, ..., 7.1153e-07, + 1.9977e-07, 1.5078e-06], + [-2.8871e-08, -1.9092e-08, 6.1141e-07, ..., -1.5832e-08, + 6.6170e-07, 1.8626e-08]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0287, -0.0333, -0.0127, -0.0004, 0.0024, 0.0056, 0.0024, -0.0083, + 0.0190, -0.0477], device='cuda:0'), grad: tensor([ 4.6054e-07, 1.1660e-06, 4.9360e-08, 9.2667e-08, -5.2936e-06, + 5.7183e-06, -1.0416e-05, 1.2293e-06, 3.6061e-06, 3.3956e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 217.54, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4915 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.15 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.0762, 0.0054, 0.1241, ..., 0.0099, -0.0826, -0.0727], + [-0.1858, -0.2092, -0.1934, ..., 0.0841, 0.0563, 0.4003], + [-0.1834, -0.1479, -0.1784, ..., -0.2387, 0.0264, -0.2299], + ..., + [-0.1942, -0.2318, 0.0069, ..., -0.2231, -0.1044, -0.3817], + [ 0.1383, 0.0352, -0.1399, ..., -0.2211, 0.0209, -0.2177], + [ 0.1355, 0.0874, -0.1266, ..., 0.0950, -0.2336, -0.2408]], + device='cuda:0'), grad: tensor([[-1.9558e-08, -9.7789e-08, -1.4575e-07, ..., -1.1316e-07, + 1.5367e-08, 4.6566e-10], + [ 1.2573e-08, 8.8476e-09, 1.7695e-08, ..., 9.3132e-10, + 5.0291e-08, -5.1223e-09], + [ 3.0734e-08, 1.6298e-08, 1.6764e-08, ..., 4.1910e-09, + 7.6368e-08, 1.8626e-09], + ..., + [ 5.1223e-09, 3.7253e-09, 6.5193e-09, ..., 1.8626e-09, + 1.7695e-08, 2.7940e-09], + [-8.1491e-08, -2.7940e-08, 6.9849e-09, ..., 6.0536e-09, + -1.4529e-07, 2.3283e-09], + [-9.6392e-08, -1.0617e-07, 6.9849e-08, ..., -2.3749e-08, + 1.0058e-07, 0.0000e+00]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0283, -0.0333, -0.0124, -0.0003, 0.0028, 0.0055, 0.0021, -0.0083, + 0.0185, -0.0477], device='cuda:0'), grad: tensor([-3.2643e-07, 1.8766e-07, 1.4249e-07, 6.9849e-08, -3.8138e-07, + 4.2655e-07, 2.6682e-07, 8.5682e-08, -4.3446e-07, -2.3749e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 217.63, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4724 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.0770, 0.0053, 0.1242, ..., 0.0098, -0.0829, -0.0728], + [-0.1867, -0.2099, -0.1944, ..., 0.0839, 0.0559, 0.4006], + [-0.1847, -0.1486, -0.1791, ..., -0.2408, 0.0262, -0.2310], + ..., + [-0.1946, -0.2329, 0.0069, ..., -0.2238, -0.1049, -0.3821], + [ 0.1384, 0.0352, -0.1407, ..., -0.2216, 0.0209, -0.2181], + [ 0.1361, 0.0877, -0.1273, ..., 0.0958, -0.2344, -0.2413]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 1.3970e-09, ..., 9.7789e-09, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 2.3283e-09, 4.6566e-10, ..., 8.3819e-09, + 2.7940e-09, 0.0000e+00], + [ 1.3970e-09, 1.3970e-09, 4.1910e-09, ..., -5.5414e-08, + 8.3819e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 4.1444e-08, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 1.2107e-08, 4.6566e-09, ..., 1.4435e-08, + -2.7940e-09, 0.0000e+00], + [-1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 285, bias, value: tensor([-2.8351e-02, -3.3530e-02, -1.2366e-02, -2.3023e-05, 3.2461e-03, + 5.2888e-03, 2.1953e-03, -8.5822e-03, 1.8369e-02, -4.7609e-02], + device='cuda:0'), grad: tensor([ 7.4506e-08, 9.3132e-08, -9.3644e-07, 7.4878e-07, 2.7940e-08, + -6.3283e-07, -5.4482e-08, 5.4715e-07, 1.2061e-07, 1.3970e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 217.54, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4920 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.15 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.0774, 0.0054, 0.1246, ..., 0.0101, -0.0825, -0.0724], + [-0.1882, -0.2113, -0.1956, ..., 0.0833, 0.0553, 0.4011], + [-0.1870, -0.1494, -0.1804, ..., -0.2429, 0.0264, -0.2315], + ..., + [-0.1947, -0.2335, 0.0070, ..., -0.2242, -0.1053, -0.3823], + [ 0.1392, 0.0353, -0.1410, ..., -0.2220, 0.0212, -0.2184], + [ 0.1363, 0.0877, -0.1277, ..., 0.0967, -0.2347, -0.2427]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -2.5332e-07, -5.8394e-07, ..., -3.4319e-07, + 6.9849e-09, 1.6158e-07], + [ 2.4214e-08, 5.2154e-08, -8.7684e-07, ..., -4.9733e-07, + 3.0268e-08, -3.1665e-06], + [ 9.7789e-08, 1.4668e-07, 2.9895e-07, ..., 1.8254e-07, + 1.4994e-07, 8.2003e-07], + ..., + [ 3.4925e-08, 4.7963e-08, 3.9162e-07, ..., 2.2771e-07, + 5.2620e-08, 1.3132e-06], + [ 8.3819e-09, 9.5461e-08, 1.5320e-07, ..., 1.1036e-07, + 3.3993e-08, 6.9849e-08], + [ 2.3283e-09, 1.4901e-08, 8.7544e-08, ..., 4.1444e-08, + 1.1176e-08, 1.8207e-07]], device='cuda:0') +Epoch 286, bias, value: tensor([-2.8035e-02, -3.4053e-02, -1.2450e-02, -6.3765e-05, 3.1482e-03, + 5.3886e-03, 2.2422e-03, -8.6133e-03, 1.8569e-02, -4.7515e-02], + device='cuda:0'), grad: tensor([-1.0692e-06, -6.2995e-06, 2.3134e-06, -1.0170e-06, 2.4959e-07, + 1.0692e-06, 7.2643e-07, 2.8573e-06, 6.8080e-07, 4.7637e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 217.64, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4607 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.13 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.0778, 0.0055, 0.1248, ..., 0.0101, -0.0827, -0.0727], + [-0.1899, -0.2124, -0.1959, ..., 0.0833, 0.0549, 0.4013], + [-0.1881, -0.1499, -0.1811, ..., -0.2438, 0.0262, -0.2332], + ..., + [-0.1947, -0.2347, 0.0068, ..., -0.2243, -0.1055, -0.3826], + [ 0.1401, 0.0356, -0.1409, ..., -0.2220, 0.0219, -0.2158], + [ 0.1365, 0.0877, -0.1278, ..., 0.0969, -0.2349, -0.2433]], + device='cuda:0'), grad: tensor([[ 1.9092e-08, 1.6764e-08, 1.2573e-08, ..., 1.3970e-08, + 6.0536e-09, -4.6566e-10], + [ 1.1176e-08, 1.3504e-08, 1.4901e-08, ..., 9.3132e-09, + 5.5879e-09, 0.0000e+00], + [ 4.6566e-10, 3.7253e-09, 8.3819e-09, ..., 9.3132e-10, + 3.7253e-09, 9.3132e-10], + ..., + [ 3.4645e-07, 2.4214e-07, 3.9581e-08, ..., 2.8452e-07, + 1.3039e-08, 3.2596e-09], + [ 1.3970e-09, 1.9558e-08, 7.3574e-08, ..., 3.4925e-08, + 1.3970e-08, 4.6566e-10], + [-4.0885e-07, 6.7428e-07, 2.6058e-06, ..., -3.8836e-07, + 1.0943e-06, 4.6566e-10]], device='cuda:0') +Epoch 287, bias, value: tensor([-2.8044e-02, -3.4831e-02, -1.2592e-02, -3.8633e-05, 2.7945e-03, + 5.3607e-03, 2.3705e-03, -8.3422e-03, 1.9084e-02, -4.7445e-02], + device='cuda:0'), grad: tensor([ 8.2888e-08, 7.4971e-08, 3.7719e-08, 1.2899e-07, -8.5309e-06, + -3.2131e-08, -9.4529e-08, 8.5216e-07, 2.2771e-07, 7.2420e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 217.60, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4886 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.09 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.0794, 0.0052, 0.1246, ..., 0.0099, -0.0830, -0.0733], + [-0.1906, -0.2159, -0.1983, ..., 0.0829, 0.0549, 0.3999], + [-0.1891, -0.1506, -0.1825, ..., -0.2452, 0.0260, -0.2337], + ..., + [-0.1948, -0.2359, 0.0066, ..., -0.2247, -0.1059, -0.3829], + [ 0.1402, 0.0355, -0.1424, ..., -0.2236, 0.0221, -0.2176], + [ 0.1367, 0.0879, -0.1286, ..., 0.0975, -0.2354, -0.2439]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, -1.5274e-07, -2.8824e-07, ..., -1.0803e-07, + 5.1223e-09, -1.0803e-07], + [ 6.7055e-08, 9.0338e-08, 5.1223e-09, ..., 1.8626e-08, + 9.4995e-08, -4.3306e-08], + [ 1.6997e-07, 2.6776e-07, 8.5682e-08, ..., 1.0012e-07, + 2.6915e-07, 5.7276e-08], + ..., + [ 4.4238e-07, 6.3796e-07, 3.2596e-09, ..., 2.3702e-07, + 5.4808e-07, 1.6764e-08], + [-3.8650e-08, 7.3109e-08, 6.1933e-08, ..., 3.7253e-08, + -2.0210e-07, 2.9802e-08], + [-5.3551e-08, 6.1467e-08, 7.1712e-08, ..., -4.9826e-08, + 5.7276e-08, 3.3528e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([-2.8504e-02, -3.5946e-02, -1.2423e-02, -4.7324e-05, 2.9058e-03, + 5.7517e-03, 2.5184e-03, -8.2940e-03, 1.8934e-02, -4.7448e-02], + device='cuda:0'), grad: tensor([-6.7474e-07, 2.8871e-07, 1.0962e-06, -2.8796e-06, 9.3132e-08, + -1.9372e-07, 7.3388e-07, 1.8040e-06, -3.6275e-07, 7.0315e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 217.62, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4600 re_mapping 0.0036 re_causal 0.0105 /// teacc 99.17 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.0800, 0.0046, 0.1240, ..., 0.0092, -0.0834, -0.0743], + [-0.1911, -0.2180, -0.2007, ..., 0.0825, 0.0553, 0.3994], + [-0.1913, -0.1515, -0.1833, ..., -0.2465, 0.0258, -0.2349], + ..., + [-0.1951, -0.2378, 0.0062, ..., -0.2252, -0.1065, -0.3838], + [ 0.1405, 0.0350, -0.1443, ..., -0.2249, 0.0223, -0.2207], + [ 0.1369, 0.0882, -0.1287, ..., 0.0981, -0.2355, -0.2443]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -1.5367e-08, 2.1886e-08, ..., 6.0536e-09, + 2.3749e-08, 3.2596e-08], + [ 3.4925e-08, 3.4925e-08, -1.1176e-08, ..., -9.8255e-08, + 2.7008e-08, -1.8859e-07], + [ 4.3306e-08, 7.8697e-08, 4.6566e-10, ..., 9.7789e-09, + 2.0955e-08, 3.2596e-09], + ..., + [ 3.5856e-07, 4.3120e-07, 1.0710e-08, ..., 2.0023e-08, + 5.7742e-08, 1.8626e-08], + [ 5.6811e-08, 4.1164e-06, 2.9989e-07, ..., 1.3318e-07, + 1.0151e-07, 9.6392e-08], + [ 4.6566e-10, 1.0850e-07, 2.9802e-08, ..., 1.6764e-08, + 5.1223e-09, 7.4506e-09]], device='cuda:0') +Epoch 289, bias, value: tensor([-2.9473e-02, -3.6608e-02, -1.2539e-02, -1.0262e-05, 2.7585e-03, + 7.8030e-03, -2.3370e-04, -8.5232e-03, 1.8508e-02, -4.7310e-02], + device='cuda:0'), grad: tensor([ 2.7986e-07, -4.9826e-08, 1.9185e-07, 3.8054e-06, 1.2014e-07, + -1.3880e-05, -2.1271e-06, 1.6699e-06, 9.7379e-06, 2.9849e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 217.61, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4741 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.07 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.0804, 0.0047, 0.1241, ..., 0.0092, -0.0837, -0.0751], + [-0.1904, -0.2178, -0.2009, ..., 0.0837, 0.0559, 0.4005], + [-0.1920, -0.1521, -0.1845, ..., -0.2478, 0.0256, -0.2367], + ..., + [-0.1953, -0.2400, 0.0045, ..., -0.2263, -0.1079, -0.3846], + [ 0.1408, 0.0351, -0.1439, ..., -0.2258, 0.0226, -0.2218], + [ 0.1370, 0.0880, -0.1307, ..., 0.0984, -0.2360, -0.2451]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, -1.7695e-08, -4.4238e-08, ..., -1.6298e-08, + 7.9162e-09, 0.0000e+00], + [ 3.8184e-08, 7.8231e-08, 2.1420e-08, ..., 1.8626e-09, + 1.1642e-08, -4.6566e-09], + [ 1.6764e-08, 4.3306e-08, 6.0536e-08, ..., 1.5832e-08, + 8.3819e-09, 4.6566e-10], + ..., + [ 3.5856e-08, 2.9802e-08, -3.5856e-08, ..., 1.5367e-08, + 5.5879e-09, 4.6566e-10], + [ 3.6089e-07, 2.6310e-07, 1.0524e-07, ..., 1.3551e-07, + 3.7253e-08, 9.3132e-10], + [-4.5542e-07, -2.6729e-07, 3.3062e-08, ..., -1.9139e-07, + 3.2596e-09, 4.6566e-10]], device='cuda:0') +Epoch 290, bias, value: tensor([-2.9584e-02, -3.5955e-02, -1.2713e-02, -8.5790e-05, 3.8995e-03, + 7.6141e-03, 1.6536e-04, -8.9637e-03, 1.8612e-02, -4.7790e-02], + device='cuda:0'), grad: tensor([-2.0955e-08, 2.9476e-07, 1.1222e-07, -1.1206e-05, 4.9081e-07, + 1.0952e-05, -7.9721e-07, -2.1234e-07, 1.5013e-06, -1.1353e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 217.56, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4770 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.18 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.0797, 0.0053, 0.1251, ..., 0.0095, -0.0841, -0.0754], + [-0.1918, -0.2181, -0.2011, ..., 0.0836, 0.0555, 0.4008], + [-0.1923, -0.1524, -0.1853, ..., -0.2489, 0.0258, -0.2369], + ..., + [-0.1953, -0.2414, 0.0043, ..., -0.2268, -0.1074, -0.3847], + [ 0.1405, 0.0349, -0.1440, ..., -0.2262, 0.0228, -0.2221], + [ 0.1373, 0.0883, -0.1317, ..., 0.0993, -0.2362, -0.2453]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 1.0431e-07, 1.8626e-08, ..., 2.3283e-09, + 1.8626e-09, 9.3132e-10], + [ 2.9337e-08, 2.0955e-08, 0.0000e+00, ..., 1.0710e-08, + 5.5879e-09, -4.3772e-08], + [-8.8476e-09, 6.9849e-09, 0.0000e+00, ..., 8.8476e-09, + 6.0536e-09, 1.3970e-09], + ..., + [ 1.9558e-08, 1.0245e-08, 4.6566e-10, ..., 1.1642e-08, + 4.6566e-09, 2.7940e-09], + [ 2.4680e-08, 3.9581e-08, 9.3132e-10, ..., 2.6543e-08, + 1.2573e-08, 3.3062e-08], + [ 2.3749e-08, 1.8626e-08, 2.3283e-09, ..., 2.2817e-08, + 1.4901e-08, 3.7253e-09]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0291, -0.0370, -0.0124, -0.0001, 0.0039, 0.0076, 0.0001, -0.0084, + 0.0185, -0.0476], device='cuda:0'), grad: tensor([ 2.3935e-07, 6.4261e-08, -1.1008e-06, 2.1234e-07, 2.0023e-08, + -5.4855e-07, 1.6764e-08, 4.9267e-07, 5.2899e-07, 7.8231e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 217.62, cls_loss 0.0006 cls_loss_mapping 0.0022 cls_loss_causal 0.4612 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.0799, 0.0053, 0.1251, ..., 0.0093, -0.0845, -0.0758], + [-0.1920, -0.2182, -0.2012, ..., 0.0840, 0.0562, 0.4015], + [-0.1927, -0.1526, -0.1866, ..., -0.2501, 0.0256, -0.2381], + ..., + [-0.1954, -0.2421, 0.0039, ..., -0.2272, -0.1077, -0.3852], + [ 0.1404, 0.0347, -0.1444, ..., -0.2271, 0.0227, -0.2227], + [ 0.1363, 0.0873, -0.1338, ..., 0.0969, -0.2385, -0.2456]], + device='cuda:0'), grad: tensor([[ 1.4296e-07, 1.1269e-07, 2.0489e-08, ..., -1.8626e-09, + 1.3970e-09, 0.0000e+00], + [ 8.0094e-08, 6.5193e-08, 1.4901e-08, ..., -4.6566e-10, + -2.7940e-09, -1.0710e-08], + [ 2.4680e-08, 3.1199e-08, 2.7940e-09, ..., 1.8626e-09, + 1.5367e-08, 2.7940e-09], + ..., + [ 2.1281e-07, 1.7695e-07, 3.4925e-08, ..., 2.4680e-08, + 9.3132e-09, 3.7253e-09], + [ 4.9360e-08, 2.0023e-08, 1.6764e-08, ..., 1.3970e-09, + -3.2596e-08, 9.3132e-10], + [-5.5408e-04, -4.2677e-04, -1.0669e-04, ..., -5.1456e-07, + 4.6566e-09, 4.6566e-10]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0293, -0.0371, -0.0123, -0.0001, 0.0052, 0.0076, 0.0003, -0.0084, + 0.0182, -0.0492], device='cuda:0'), grad: tensor([ 5.0711e-07, 3.1339e-07, 1.0943e-07, 1.0720e-06, 1.9217e-03, + -8.8941e-07, 8.6147e-08, 2.8824e-07, 1.3551e-07, -1.9255e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 217.54, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4649 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.21 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.0801, 0.0053, 0.1254, ..., 0.0090, -0.0847, -0.0764], + [-0.1924, -0.2185, -0.2015, ..., 0.0841, 0.0565, 0.4020], + [-0.1932, -0.1530, -0.1874, ..., -0.2514, 0.0255, -0.2389], + ..., + [-0.1954, -0.2425, 0.0019, ..., -0.2275, -0.1080, -0.3855], + [ 0.1404, 0.0347, -0.1445, ..., -0.2275, 0.0227, -0.2229], + [ 0.1390, 0.0900, -0.1321, ..., 0.0970, -0.2389, -0.2460]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 1.0245e-08, 1.1642e-08, ..., 2.7474e-08, + 2.0023e-08, 1.2107e-08], + [ 1.4435e-08, 1.1176e-08, 7.4506e-09, ..., -8.9128e-07, + -6.3889e-07, -1.2247e-06], + [ 6.9849e-09, 1.3039e-08, 1.9092e-08, ..., 5.6205e-07, + 4.3865e-07, 7.1246e-07], + ..., + [ 1.9558e-08, 1.5367e-08, -4.1910e-09, ..., 3.1758e-07, + 2.3656e-07, 4.0885e-07], + [ 4.6566e-09, -1.3039e-08, 1.0710e-08, ..., 2.2817e-08, + -7.4971e-08, 2.1420e-08], + [-4.6566e-07, -2.6822e-07, 1.7229e-08, ..., -4.2235e-07, + 1.8626e-08, 1.0710e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0294, -0.0372, -0.0124, -0.0001, 0.0028, 0.0076, 0.0004, -0.0085, + 0.0181, -0.0468], device='cuda:0'), grad: tensor([ 1.8114e-07, -4.1462e-06, 3.0585e-06, 3.4180e-07, 7.4180e-07, + 2.3935e-07, -6.4727e-08, 6.8219e-07, -2.4447e-07, -7.9302e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 217.98, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4380 re_mapping 0.0037 re_causal 0.0103 /// teacc 99.11 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.0808, 0.0054, 0.1257, ..., 0.0090, -0.0852, -0.0774], + [-0.1928, -0.2188, -0.2016, ..., 0.0845, 0.0570, 0.4033], + [-0.1936, -0.1535, -0.1881, ..., -0.2533, 0.0253, -0.2413], + ..., + [-0.1955, -0.2420, 0.0018, ..., -0.2282, -0.1085, -0.3865], + [ 0.1405, 0.0349, -0.1448, ..., -0.2281, 0.0229, -0.2234], + [ 0.1392, 0.0901, -0.1322, ..., 0.0974, -0.2390, -0.2474]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.8626e-09, -3.2596e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 0.0000e+00], + ..., + [ 3.2596e-09, 1.8626e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, -1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + -1.4435e-08, 0.0000e+00], + [-1.4435e-08, -4.6566e-09, 2.7940e-09, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0294, -0.0363, -0.0127, -0.0003, 0.0028, 0.0075, 0.0005, -0.0088, + 0.0182, -0.0467], device='cuda:0'), grad: tensor([ 5.1223e-08, 5.3272e-07, -3.3202e-07, 3.2596e-08, 3.8231e-07, + 3.2596e-08, 8.8476e-09, -1.0170e-06, 3.7719e-08, 2.7288e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 217.88, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4713 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.06 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.0815, 0.0053, 0.1255, ..., 0.0087, -0.0856, -0.0774], + [-0.1930, -0.2190, -0.2016, ..., 0.0851, 0.0575, 0.4040], + [-0.1945, -0.1541, -0.1889, ..., -0.2546, 0.0252, -0.2428], + ..., + [-0.1958, -0.2434, 0.0026, ..., -0.2293, -0.1088, -0.3867], + [ 0.1406, 0.0350, -0.1451, ..., -0.2285, 0.0229, -0.2236], + [ 0.1393, 0.0902, -0.1323, ..., 0.0978, -0.2391, -0.2478]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 4.6566e-09, -1.9092e-08, ..., 2.7008e-08, + 1.3970e-09, 5.5879e-09], + [ 2.3283e-09, 1.1176e-08, 1.3970e-09, ..., -7.8697e-08, + -7.4971e-08, -3.2317e-07], + [ 4.6566e-10, 4.1910e-09, 3.7253e-09, ..., 3.3528e-08, + 4.1444e-08, 2.0117e-07], + ..., + [ 7.9162e-09, 1.2107e-08, 4.6566e-10, ..., 6.3330e-08, + 2.7474e-08, 9.5926e-08], + [ 1.0710e-08, 5.0291e-08, 1.8626e-09, ..., 9.3132e-09, + 4.6566e-10, 4.6566e-09], + [-1.0757e-07, -8.2422e-08, 7.9162e-09, ..., -3.2177e-07, + 1.3970e-09, 4.1910e-09]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0298, -0.0361, -0.0130, -0.0003, 0.0027, 0.0075, 0.0006, -0.0088, + 0.0183, -0.0467], device='cuda:0'), grad: tensor([ 7.7300e-08, -5.8347e-07, 2.7893e-07, 1.9651e-07, 6.8638e-07, + -4.8941e-07, 7.2643e-08, 3.3155e-07, 2.5239e-07, -8.0653e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 217.85, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4575 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.11 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.0818, 0.0055, 0.1257, ..., 0.0087, -0.0857, -0.0779], + [-0.1935, -0.2191, -0.2018, ..., 0.0855, 0.0578, 0.4047], + [-0.1954, -0.1545, -0.1895, ..., -0.2561, 0.0254, -0.2436], + ..., + [-0.1959, -0.2437, 0.0026, ..., -0.2296, -0.1091, -0.3870], + [ 0.1407, 0.0350, -0.1453, ..., -0.2291, 0.0228, -0.2241], + [ 0.1394, 0.0903, -0.1324, ..., 0.0980, -0.2393, -0.2491]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.3970e-09, -2.7940e-09, ..., 5.1223e-09, + 4.6566e-10, 0.0000e+00], + [ 1.3970e-09, 4.6566e-10, 3.7253e-09, ..., 5.7742e-08, + -9.3132e-10, -6.5193e-09], + [ 1.8626e-09, 1.8626e-09, 4.6566e-10, ..., 5.1223e-09, + 1.8626e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, -1.1642e-08, ..., -2.0489e-07, + 2.3283e-09, 5.5879e-09], + [-1.3970e-08, -1.3039e-08, 4.6566e-10, ..., -4.6566e-10, + -1.4901e-08, 0.0000e+00], + [-4.1910e-09, -1.8626e-09, 9.3132e-09, ..., 1.0990e-07, + 1.8626e-09, 1.3970e-09]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0297, -0.0361, -0.0126, -0.0003, 0.0027, 0.0075, 0.0006, -0.0087, + 0.0181, -0.0467], device='cuda:0'), grad: tensor([ 7.5437e-08, 6.7567e-07, -4.5169e-07, 1.2573e-07, 2.4773e-07, + 5.8208e-08, -1.4901e-08, -1.8645e-06, 4.2375e-08, 1.1250e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 217.90, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4638 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.19 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.0823, 0.0050, 0.1259, ..., 0.0074, -0.0861, -0.0780], + [-0.1944, -0.2194, -0.2020, ..., 0.0850, 0.0579, 0.4050], + [-0.1959, -0.1548, -0.1899, ..., -0.2573, 0.0253, -0.2443], + ..., + [-0.1960, -0.2441, 0.0025, ..., -0.2299, -0.1095, -0.3873], + [ 0.1410, 0.0352, -0.1456, ..., -0.2295, 0.0232, -0.2241], + [ 0.1395, 0.0904, -0.1324, ..., 0.0991, -0.2395, -0.2495]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 6.0536e-09, 9.3132e-10, ..., 1.0245e-08, + 9.3132e-10, 9.3132e-10], + [ 1.5832e-08, 1.5367e-08, 1.3970e-09, ..., 1.4435e-08, + -4.1910e-09, -2.7008e-08], + [ 2.0955e-08, 1.4901e-08, 6.5193e-09, ..., 7.9162e-09, + 4.1910e-09, 3.7253e-09], + ..., + [ 3.8650e-08, 3.5856e-08, 4.6566e-10, ..., 4.8894e-08, + 1.0710e-08, 1.8161e-08], + [-3.2131e-08, -7.4506e-09, 9.3132e-10, ..., 6.0536e-09, + -5.5879e-09, 1.3970e-09], + [-4.6147e-07, -4.1258e-07, 3.6787e-08, ..., -5.7463e-07, + 2.2352e-08, 9.3132e-10]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0306, -0.0365, -0.0125, -0.0003, 0.0025, 0.0075, 0.0007, -0.0087, + 0.0183, -0.0465], device='cuda:0'), grad: tensor([ 4.3772e-08, 4.7963e-08, -5.9046e-07, 4.4750e-07, 1.0524e-06, + -2.4680e-08, 1.5832e-08, 2.3097e-07, 1.0431e-07, -1.3225e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 217.82, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4680 re_mapping 0.0035 re_causal 0.0102 /// teacc 99.08 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.0830, 0.0052, 0.1263, ..., 0.0076, -0.0864, -0.0785], + [-0.1949, -0.2197, -0.2022, ..., 0.0838, 0.0584, 0.4056], + [-0.1965, -0.1551, -0.1905, ..., -0.2592, 0.0256, -0.2455], + ..., + [-0.1962, -0.2450, 0.0023, ..., -0.2314, -0.1110, -0.3880], + [ 0.1413, 0.0352, -0.1461, ..., -0.2299, 0.0232, -0.2242], + [ 0.1396, 0.0904, -0.1327, ..., 0.1002, -0.2404, -0.2502]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, 1.3039e-08, 3.1199e-08, ..., 4.4238e-09, + 3.2363e-08, 9.3132e-10], + [ 8.8476e-09, 1.1642e-06, 1.9558e-08, ..., 3.0035e-08, + 2.7008e-08, -2.5146e-08], + [ 5.8208e-09, 2.7241e-08, 3.1432e-08, ..., 7.6834e-09, + 6.0769e-08, 9.5461e-09], + ..., + [-1.4668e-08, -2.0880e-06, 5.8208e-08, ..., -6.7055e-08, + 9.0105e-08, 5.5879e-09], + [-9.7789e-09, 6.5658e-08, 2.5611e-09, ..., 4.6566e-09, + -1.7928e-08, 2.3283e-09], + [ 5.1223e-09, 1.6615e-06, 3.0901e-06, ..., 2.6310e-08, + 2.8647e-06, 3.4925e-09]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0307, -0.0370, -0.0117, -0.0004, 0.0026, 0.0076, 0.0009, -0.0092, + 0.0181, -0.0464], device='cuda:0'), grad: tensor([ 1.7020e-07, 7.4171e-06, 2.7544e-07, 4.4610e-07, -1.4648e-05, + 1.3504e-07, 1.6415e-07, -1.2808e-05, 4.2794e-07, 1.8418e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 217.81, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4812 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.17 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.0839, 0.0055, 0.1265, ..., 0.0075, -0.0870, -0.0788], + [-0.1956, -0.2201, -0.2024, ..., 0.0845, 0.0577, 0.4062], + [-0.1978, -0.1559, -0.1919, ..., -0.2629, 0.0254, -0.2463], + ..., + [-0.1964, -0.2462, 0.0026, ..., -0.2337, -0.1109, -0.3887], + [ 0.1412, 0.0349, -0.1465, ..., -0.2309, 0.0231, -0.2243], + [ 0.1396, 0.0904, -0.1338, ..., 0.1004, -0.2429, -0.2509]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -2.0489e-08, -3.3993e-08, ..., -1.9325e-08, + 9.3132e-10, 4.6566e-10], + [ 3.0268e-09, 4.6566e-09, 2.3283e-09, ..., -3.0268e-09, + 1.8626e-09, -9.7789e-09], + [ 2.3283e-09, 4.1910e-09, 2.5611e-09, ..., 1.8626e-09, + 2.5844e-08, 9.3132e-10], + ..., + [ 8.1956e-08, 2.3050e-08, 4.6566e-09, ..., 6.2864e-09, + 9.7789e-09, 4.4238e-09], + [-4.6566e-09, -6.0536e-09, 1.3970e-09, ..., 1.6298e-09, + -6.3563e-08, 1.6298e-09], + [-1.6531e-08, -6.5193e-09, 2.1653e-08, ..., -7.6834e-09, + 2.0955e-09, 1.8626e-09]], device='cuda:0') +Epoch 299, bias, value: tensor([-3.0974e-02, -3.7755e-02, -1.2167e-02, -8.0047e-05, 2.8039e-03, + 7.7094e-03, 7.2783e-04, -8.9296e-03, 1.7812e-02, -4.6612e-02], + device='cuda:0'), grad: tensor([-5.6112e-08, 1.1642e-07, 2.4633e-07, 8.1724e-08, 6.4028e-08, + -6.9849e-09, 6.6357e-08, -3.8790e-07, -1.5995e-07, 5.3784e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 217.80, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4463 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.19 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.0845, 0.0068, 0.1276, ..., 0.0078, -0.0879, -0.0811], + [-0.1963, -0.2206, -0.2026, ..., 0.0849, 0.0578, 0.4069], + [-0.1985, -0.1565, -0.1924, ..., -0.2637, 0.0252, -0.2466], + ..., + [-0.1966, -0.2465, 0.0024, ..., -0.2346, -0.1113, -0.3890], + [ 0.1415, 0.0349, -0.1466, ..., -0.2316, 0.0240, -0.2245], + [ 0.1397, 0.0904, -0.1340, ..., 0.1010, -0.2431, -0.2524]], + device='cuda:0'), grad: tensor([[ 3.4226e-08, 1.0943e-08, -2.8638e-08, ..., -1.0477e-08, + 2.3283e-08, 1.8626e-09], + [ 3.2363e-08, 1.6997e-08, -1.1642e-09, ..., -1.5134e-08, + 5.8208e-09, -3.2829e-08], + [ 3.7253e-08, 2.1420e-08, 4.1910e-09, ..., 9.3132e-09, + 2.2352e-08, 4.4238e-09], + ..., + [ 3.8184e-08, 5.8673e-08, 9.3132e-10, ..., 1.8394e-08, + 5.2154e-08, 4.6566e-09], + [-9.4995e-08, -5.7044e-08, 7.9162e-09, ..., 2.7940e-09, + -6.4960e-08, 3.9581e-09], + [ 2.4447e-08, 1.5134e-08, 1.1874e-08, ..., 4.4238e-09, + 1.5832e-08, 1.8626e-09]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0304, -0.0393, -0.0116, -0.0008, 0.0028, 0.0082, 0.0006, -0.0081, + 0.0181, -0.0466], device='cuda:0'), grad: tensor([ 6.7754e-08, 4.2794e-07, 4.1700e-07, -1.9092e-07, 1.1967e-07, + 2.2259e-07, -1.1176e-08, -1.2545e-06, -2.0768e-07, 4.0908e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 217.59, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4622 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.05 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.0852, 0.0068, 0.1280, ..., 0.0078, -0.0883, -0.0816], + [-0.1966, -0.2210, -0.2027, ..., 0.0856, 0.0582, 0.4082], + [-0.2001, -0.1573, -0.1933, ..., -0.2672, 0.0248, -0.2489], + ..., + [-0.1967, -0.2472, 0.0021, ..., -0.2352, -0.1116, -0.3898], + [ 0.1416, 0.0346, -0.1494, ..., -0.2343, 0.0247, -0.2250], + [ 0.1398, 0.0905, -0.1341, ..., 0.1014, -0.2433, -0.2539]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 9.3132e-10, -2.7940e-09, ..., -2.3283e-10, + 1.1642e-09, 0.0000e+00], + [ 6.5193e-09, 9.0804e-09, 4.6566e-10, ..., 4.8894e-09, + 4.1910e-09, -2.3283e-09], + [ 2.1886e-08, 3.0734e-08, 4.6566e-10, ..., 2.0023e-08, + 1.7695e-08, 4.6566e-10], + ..., + [ 1.8161e-08, 2.6776e-08, 4.6566e-10, ..., 1.6997e-08, + 1.3271e-08, 2.7940e-09], + [ 3.7253e-09, 4.6566e-09, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 3.2596e-09, 7.9162e-09, 1.6298e-09, ..., 1.8626e-09, + 1.0477e-08, 2.3283e-10]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0304, -0.0392, -0.0119, -0.0005, 0.0027, 0.0081, 0.0008, -0.0082, + 0.0177, -0.0465], device='cuda:0'), grad: tensor([ 3.4925e-09, 3.7253e-08, 1.4412e-07, -4.7591e-07, 8.3819e-09, + 1.5879e-07, 3.2596e-09, 9.8953e-08, 1.6997e-08, 2.4447e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 217.87, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4713 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.14 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.0857, 0.0072, 0.1287, ..., 0.0081, -0.0886, -0.0827], + [-0.1968, -0.2212, -0.2029, ..., 0.0876, 0.0607, 0.4111], + [-0.2008, -0.1579, -0.1941, ..., -0.2714, 0.0233, -0.2530], + ..., + [-0.1968, -0.2480, 0.0020, ..., -0.2357, -0.1123, -0.3910], + [ 0.1416, 0.0344, -0.1495, ..., -0.2350, 0.0250, -0.2260], + [ 0.1398, 0.0904, -0.1343, ..., 0.1014, -0.2438, -0.2545]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.7928e-08, -2.0023e-08, ..., -2.4214e-08, + 1.1642e-09, 2.3283e-09], + [ 6.5193e-09, 1.8626e-09, 9.3132e-10, ..., -2.4191e-07, + -3.8184e-08, -1.8533e-07], + [ 3.7253e-08, -2.7940e-09, 3.0268e-09, ..., 3.4226e-08, + -1.1642e-09, 2.1886e-08], + ..., + [ 6.9849e-10, 6.9849e-10, 0.0000e+00, ..., 1.9162e-07, + 3.3062e-08, 1.4598e-07], + [-4.8196e-08, 2.0955e-09, -2.5611e-09, ..., 1.3970e-09, + 3.9581e-09, 2.3283e-10], + [-6.9849e-10, 1.1642e-08, 1.3504e-08, ..., 2.3283e-08, + 1.3970e-09, 3.4925e-09]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0301, -0.0380, -0.0142, -0.0001, 0.0027, 0.0080, 0.0007, -0.0084, + 0.0177, -0.0466], device='cuda:0'), grad: tensor([-4.9360e-08, -6.2026e-07, 2.0373e-07, 2.7707e-08, 1.7229e-08, + 4.8662e-08, 5.7742e-08, 6.0443e-07, -3.5297e-07, 6.5425e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 217.55, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4742 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.11 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.0859, 0.0076, 0.1288, ..., 0.0081, -0.0890, -0.0835], + [-0.1969, -0.2213, -0.2029, ..., 0.0861, 0.0613, 0.4118], + [-0.2010, -0.1581, -0.1945, ..., -0.2715, 0.0232, -0.2534], + ..., + [-0.1968, -0.2482, 0.0019, ..., -0.2341, -0.1128, -0.3919], + [ 0.1416, 0.0344, -0.1520, ..., -0.2378, 0.0250, -0.2261], + [ 0.1398, 0.0904, -0.1343, ..., 0.1017, -0.2439, -0.2553]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.3283e-08, -3.7486e-08, ..., -1.9325e-08, + 2.5611e-09, 9.3132e-10], + [ 7.2177e-09, 3.9581e-09, 9.3132e-10, ..., -2.7940e-09, + 4.6799e-08, -3.0734e-08], + [ 2.1397e-07, 1.9395e-07, 7.2177e-09, ..., -6.9849e-10, + 5.9512e-07, 6.0536e-09], + ..., + [ 7.4739e-08, 3.2596e-09, 1.3970e-09, ..., 1.2899e-07, + -6.1747e-07, 1.2340e-08], + [-2.7218e-07, -2.7800e-07, -6.1234e-08, ..., 4.4238e-09, + -2.7381e-07, 2.0955e-09], + [-5.1223e-08, 4.0280e-08, 2.2119e-08, ..., -1.3621e-07, + 4.4005e-08, 2.3283e-09]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0304, -0.0401, -0.0111, -0.0002, 0.0026, 0.0078, 0.0022, -0.0084, + 0.0159, -0.0466], device='cuda:0'), grad: tensor([ 3.2154e-07, 1.7416e-06, 1.1399e-05, 5.9232e-07, 2.5462e-06, + 2.0093e-07, 3.1083e-07, -1.6466e-05, -1.4119e-06, 7.1479e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 217.44, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4865 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.10 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.0868, 0.0085, 0.1298, ..., 0.0087, -0.0904, -0.0859], + [-0.1972, -0.2210, -0.2027, ..., 0.0887, 0.0622, 0.4135], + [-0.2025, -0.1587, -0.1952, ..., -0.2719, 0.0230, -0.2538], + ..., + [-0.1969, -0.2493, 0.0015, ..., -0.2367, -0.1137, -0.3945], + [ 0.1421, 0.0345, -0.1529, ..., -0.2389, 0.0257, -0.2264], + [ 0.1399, 0.0905, -0.1345, ..., 0.1020, -0.2442, -0.2576]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 2.0955e-09, 4.4238e-09, ..., 2.7940e-09, + 4.6566e-09, 6.9849e-10], + [ 5.5879e-09, 6.0536e-09, 3.0268e-09, ..., -4.1910e-09, + -2.0955e-09, -2.4680e-08], + [ 3.7253e-09, 3.9581e-09, 2.0955e-09, ..., 3.2596e-09, + 7.2177e-09, 1.1642e-09], + ..., + [ 1.3271e-08, 1.3039e-08, 1.8626e-09, ..., 2.3283e-08, + 3.2596e-09, 5.8208e-09], + [ 9.3132e-10, 1.1176e-08, 1.3970e-09, ..., 2.3283e-09, + 1.8626e-09, 1.8626e-09], + [-2.6170e-07, -2.3702e-07, 2.0955e-09, ..., -3.3644e-07, + -3.3528e-08, 9.3132e-10]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0302, -0.0392, -0.0112, -0.0003, 0.0026, 0.0078, 0.0026, -0.0089, + 0.0152, -0.0466], device='cuda:0'), grad: tensor([ 2.1188e-08, -2.5611e-09, 2.4680e-08, -6.0769e-08, 7.3295e-07, + 2.9802e-08, 3.0268e-08, 2.3050e-08, 3.6089e-08, -8.1398e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 217.59, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4464 re_mapping 0.0035 re_causal 0.0101 /// teacc 99.22 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.0872, 0.0087, 0.1303, ..., 0.0088, -0.0910, -0.0872], + [-0.1975, -0.2208, -0.2028, ..., 0.0895, 0.0623, 0.4142], + [-0.2032, -0.1594, -0.1959, ..., -0.2718, 0.0229, -0.2539], + ..., + [-0.1970, -0.2503, 0.0014, ..., -0.2374, -0.1133, -0.3950], + [ 0.1422, 0.0342, -0.1529, ..., -0.2392, 0.0259, -0.2276], + [ 0.1402, 0.0907, -0.1344, ..., 0.1028, -0.2442, -0.2584]], + device='cuda:0'), grad: tensor([[ 2.2817e-08, 3.5390e-08, 0.0000e+00, ..., 7.6834e-09, + 4.1910e-09, 0.0000e+00], + [ 5.0757e-08, 8.5915e-08, 0.0000e+00, ..., 2.5611e-09, + 1.3970e-08, 0.0000e+00], + [ 1.3201e-07, 2.2748e-07, 0.0000e+00, ..., 9.3132e-10, + 1.5600e-08, 2.3283e-10], + ..., + [ 1.1246e-07, 1.9232e-07, 2.3283e-09, ..., 2.0955e-09, + 2.9104e-08, 1.6298e-09], + [ 9.8720e-08, 1.6764e-07, 0.0000e+00, ..., 1.0245e-08, + 2.4447e-08, 0.0000e+00], + [-2.1420e-08, 1.2806e-07, 6.2864e-09, ..., -3.0035e-08, + 2.7940e-09, 2.3283e-10]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0301, -0.0394, -0.0112, -0.0004, 0.0024, 0.0080, 0.0025, -0.0087, + 0.0150, -0.0464], device='cuda:0'), grad: tensor([ 1.1385e-07, 3.0012e-07, -2.5611e-07, -2.5943e-05, 4.2608e-08, + 2.3916e-05, 4.5169e-08, 9.9931e-07, 6.3516e-07, 2.1490e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 217.74, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4851 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.21 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.0880, 0.0083, 0.1302, ..., 0.0081, -0.0915, -0.0878], + [-0.1981, -0.2210, -0.2035, ..., 0.0896, 0.0623, 0.4150], + [-0.2043, -0.1603, -0.1969, ..., -0.2723, 0.0227, -0.2539], + ..., + [-0.1972, -0.2510, 0.0016, ..., -0.2377, -0.1138, -0.3959], + [ 0.1420, 0.0337, -0.1535, ..., -0.2397, 0.0261, -0.2282], + [ 0.1404, 0.0910, -0.1345, ..., 0.1037, -0.2442, -0.2625]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 2.5611e-09, 2.0955e-09, ..., 8.3819e-09, + 6.7521e-09, 0.0000e+00], + [ 4.5169e-08, 1.5832e-08, 1.3970e-09, ..., 5.2154e-08, + 5.5879e-08, -7.4506e-09], + [ 9.9884e-08, 5.4250e-08, 2.0955e-09, ..., 4.8894e-09, + 6.3097e-08, 2.3283e-10], + ..., + [ 1.6997e-08, 1.0943e-08, 1.3970e-09, ..., 2.1653e-08, + 6.3330e-08, 7.6834e-09], + [-8.9640e-08, -4.8662e-08, 1.0710e-08, ..., 1.8626e-08, + -1.1479e-07, 4.6566e-10], + [-3.0501e-08, -1.6298e-08, 3.9581e-09, ..., -1.0291e-07, + 1.9325e-08, 4.6566e-10]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0308, -0.0395, -0.0113, -0.0003, 0.0021, 0.0077, 0.0031, -0.0086, + 0.0146, -0.0461], device='cuda:0'), grad: tensor([ 3.5390e-08, 2.8056e-07, -1.9628e-07, -3.7858e-07, 2.2585e-07, + -2.0955e-08, -1.5367e-08, 4.6706e-07, -2.3097e-07, -1.5390e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 217.69, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4676 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.20 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.0881, 0.0087, 0.1305, ..., 0.0083, -0.0923, -0.0900], + [-0.1982, -0.2211, -0.2039, ..., 0.0898, 0.0624, 0.4154], + [-0.2053, -0.1608, -0.1975, ..., -0.2725, 0.0231, -0.2541], + ..., + [-0.1972, -0.2514, 0.0015, ..., -0.2380, -0.1144, -0.3966], + [ 0.1419, 0.0335, -0.1537, ..., -0.2400, 0.0260, -0.2284], + [ 0.1405, 0.0911, -0.1346, ..., 0.1040, -0.2441, -0.2648]], + device='cuda:0'), grad: tensor([[ 2.8871e-08, -2.0023e-08, -3.6322e-08, ..., 2.8871e-08, + 6.0536e-09, 0.0000e+00], + [ 7.4506e-09, 7.4506e-09, 8.3819e-09, ..., 9.7789e-09, + 1.3504e-08, -2.7940e-09], + [ 3.6322e-08, 3.8650e-08, 2.2352e-08, ..., 1.6764e-08, + 5.2620e-08, 4.6566e-10], + ..., + [ 3.2131e-08, 1.5832e-08, 7.9162e-09, ..., 3.7719e-08, + 3.3528e-08, 3.2596e-09], + [-6.8918e-08, -5.1688e-08, -1.3970e-09, ..., 7.9162e-09, + -1.1222e-07, 0.0000e+00], + [-1.0151e-07, -8.8476e-09, 8.8476e-09, ..., -2.0070e-07, + 1.6298e-08, 4.6566e-10]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0308, -0.0394, -0.0110, -0.0002, 0.0021, 0.0077, 0.0031, -0.0087, + 0.0144, -0.0461], device='cuda:0'), grad: tensor([-6.4261e-08, 1.1874e-07, 2.9430e-07, 1.6764e-08, 1.0803e-07, + 1.4203e-07, 1.3132e-07, -2.3004e-07, -4.3958e-07, -6.6590e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 217.53, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4739 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.09 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.0885, 0.0090, 0.1305, ..., 0.0085, -0.0927, -0.0910], + [-0.1985, -0.2213, -0.2041, ..., 0.0899, 0.0625, 0.4158], + [-0.2065, -0.1616, -0.1973, ..., -0.2728, 0.0232, -0.2541], + ..., + [-0.1976, -0.2525, 0.0011, ..., -0.2387, -0.1154, -0.3971], + [ 0.1418, 0.0333, -0.1538, ..., -0.2401, 0.0260, -0.2286], + [ 0.1406, 0.0911, -0.1347, ..., 0.1043, -0.2445, -0.2660]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.2596e-09, -2.3283e-09, ..., -1.3970e-09, + 9.3132e-10, 4.6566e-10], + [ 3.7253e-09, 1.9092e-08, 5.1223e-09, ..., 9.1735e-08, + -4.6566e-10, -6.9849e-09], + [ 8.8476e-09, 4.6566e-09, 1.3970e-09, ..., 2.3283e-09, + 5.5879e-09, 4.6566e-10], + ..., + [ 6.5193e-09, 7.9162e-09, 4.6566e-10, ..., -8.6613e-08, + 2.3283e-09, 2.7940e-09], + [-6.5658e-08, -2.4214e-08, 1.8626e-09, ..., 3.7253e-09, + -4.2841e-08, 0.0000e+00], + [ 1.1176e-08, 6.9849e-09, 7.9162e-09, ..., -2.7940e-09, + 2.1420e-08, 1.8626e-09]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0309, -0.0396, -0.0108, 0.0005, 0.0021, 0.0073, 0.0029, -0.0087, + 0.0143, -0.0461], device='cuda:0'), grad: tensor([ 1.0245e-08, 2.1011e-06, 7.6368e-08, 1.4752e-06, 6.5193e-08, + -1.2666e-06, 4.5169e-08, -2.5481e-06, -2.0862e-07, 2.4261e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 217.66, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4585 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.13 lr 0.00010000 +Epoch 309, weight, value: tensor([[-8.9001e-02, 9.4323e-03, 1.3089e-01, ..., 8.7768e-03, + -9.3003e-02, -9.0977e-02], + [-1.9885e-01, -2.2149e-01, -2.0413e-01, ..., 9.0122e-02, + 6.2587e-02, 4.1637e-01], + [-2.0785e-01, -1.6240e-01, -1.9838e-01, ..., -2.7313e-01, + 2.2853e-02, -2.5422e-01], + ..., + [-1.9766e-01, -2.5319e-01, 3.8456e-04, ..., -2.3916e-01, + -1.1619e-01, -3.9836e-01], + [ 1.4271e-01, 3.3523e-02, -1.5381e-01, ..., -2.4004e-01, + 2.7375e-02, -2.2862e-01], + [ 1.4148e-01, 9.1950e-02, -1.3411e-01, ..., 1.0723e-01, + -2.4191e-01, -2.6633e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 2.7940e-09, ..., 5.5879e-09, + 3.7253e-09, 2.7940e-09], + [ 4.6566e-10, 4.6566e-10, 9.3132e-09, ..., -2.5146e-08, + -5.5879e-09, -5.1688e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.3283e-09, + 3.2596e-09, 4.6566e-10], + ..., + [ 1.3970e-09, 4.6566e-10, 4.1910e-09, ..., 1.6298e-08, + 2.2352e-08, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + 1.8626e-09, 3.2596e-09], + [-9.3132e-09, -3.7253e-09, 1.8114e-07, ..., 1.0012e-07, + 2.0023e-07, 1.3970e-08]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0305, -0.0397, -0.0111, 0.0004, 0.0005, 0.0075, 0.0030, -0.0087, + 0.0146, -0.0446], device='cuda:0'), grad: tensor([ 2.7940e-08, -2.7474e-08, 3.3528e-08, 1.0710e-08, -6.9989e-07, + 9.7789e-09, -1.1967e-07, 2.5146e-08, 1.1176e-08, 7.5297e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 217.79, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4786 re_mapping 0.0035 re_causal 0.0098 /// teacc 99.17 lr 0.00010000 +Epoch 310, weight, value: tensor([[-8.9404e-02, 1.0055e-02, 1.3138e-01, ..., 9.4183e-03, + -9.3965e-02, -9.2785e-02], + [-2.0034e-01, -2.2238e-01, -2.0504e-01, ..., 9.0401e-02, + 6.5659e-02, 4.1941e-01], + [-2.0947e-01, -1.6313e-01, -1.9903e-01, ..., -2.7276e-01, + 1.9661e-02, -2.5751e-01], + ..., + [-1.9803e-01, -2.5427e-01, 1.3077e-04, ..., -2.4013e-01, + -1.1794e-01, -4.0113e-01], + [ 1.4260e-01, 3.3274e-02, -1.5405e-01, ..., -2.4123e-01, + 2.7202e-02, -2.3247e-01], + [ 1.4161e-01, 9.1923e-02, -1.3486e-01, ..., 1.0737e-01, + -2.4238e-01, -2.6780e-01]], device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.3504e-08, -1.7695e-08, ..., -1.0710e-08, + 3.7253e-09, -2.3283e-09], + [ 1.8626e-08, 1.4901e-08, 1.8626e-09, ..., 1.0710e-08, + 8.8476e-09, -7.4506e-09], + [ 4.0978e-08, 2.9802e-08, 1.3970e-09, ..., 2.4680e-08, + 2.3283e-08, 1.3970e-09], + ..., + [ 3.9861e-07, 2.7800e-07, 0.0000e+00, ..., 2.2212e-07, + 2.2259e-07, 6.9849e-09], + [ 2.0489e-08, 1.6298e-08, 1.8626e-09, ..., 1.3504e-08, + 1.1642e-08, 9.3132e-10], + [-1.0245e-08, -3.7253e-09, 3.7253e-09, ..., -1.3970e-08, + 5.1223e-09, 1.3970e-09]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0303, -0.0367, -0.0145, 0.0006, 0.0006, 0.0077, 0.0030, -0.0089, + 0.0141, -0.0447], device='cuda:0'), grad: tensor([-4.4238e-08, 1.7229e-07, 1.8207e-07, -2.2613e-06, 2.7474e-08, + 1.9372e-07, 3.7719e-08, 1.5255e-06, 1.0943e-07, 6.9849e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 217.69, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4750 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.15 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.0893, 0.0110, 0.1320, ..., 0.0101, -0.0948, -0.0932], + [-0.2006, -0.2226, -0.2054, ..., 0.0911, 0.0657, 0.4195], + [-0.2104, -0.1637, -0.1998, ..., -0.2729, 0.0197, -0.2575], + ..., + [-0.1982, -0.2550, -0.0019, ..., -0.2411, -0.1188, -0.4025], + [ 0.1426, 0.0332, -0.1543, ..., -0.2416, 0.0273, -0.2326], + [ 0.1417, 0.0919, -0.1360, ..., 0.1075, -0.2429, -0.2692]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-10, -4.6566e-10, ..., 4.6566e-10, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 4.6566e-09, ..., -4.6566e-09, + -4.6566e-10, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 9.3132e-10, + 8.8476e-09, 1.3970e-09], + ..., + [ 4.1910e-09, 3.2596e-09, 6.6124e-08, ..., 6.5193e-09, + 5.8208e-08, 3.7253e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [-1.6298e-08, -8.8476e-09, 3.2596e-09, ..., -1.7695e-08, + 9.3132e-10, 1.3970e-09]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0303, -0.0367, -0.0145, 0.0008, 0.0008, 0.0076, 0.0032, -0.0092, + 0.0138, -0.0449], device='cuda:0'), grad: tensor([ 6.0536e-09, 2.3749e-08, 3.0734e-08, 1.7229e-08, -2.1560e-07, + -1.3504e-08, 9.7789e-09, 1.4296e-07, 6.9849e-09, -8.3819e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 217.48, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4900 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.22 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.0899, 0.0113, 0.1323, ..., 0.0105, -0.0954, -0.0946], + [-0.2012, -0.2228, -0.2053, ..., 0.0914, 0.0657, 0.4195], + [-0.2111, -0.1642, -0.2005, ..., -0.2730, 0.0197, -0.2575], + ..., + [-0.1988, -0.2574, -0.0023, ..., -0.2423, -0.1195, -0.4034], + [ 0.1427, 0.0333, -0.1544, ..., -0.2419, 0.0275, -0.2328], + [ 0.1418, 0.0919, -0.1364, ..., 0.1076, -0.2431, -0.2711]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -7.4506e-09, -1.3970e-08, ..., -1.1642e-08, + 4.6566e-10, 1.8626e-09], + [ 1.3970e-09, 1.0245e-08, 1.8626e-09, ..., -2.4214e-08, + -1.4901e-08, -5.3085e-08], + [ 0.0000e+00, -5.6811e-08, -1.0245e-08, ..., 6.0536e-09, + -6.0536e-09, 1.1642e-08], + ..., + [ 2.7940e-09, 2.1420e-08, 5.1223e-09, ..., 6.0536e-09, + 6.9849e-09, 1.1642e-08], + [-5.1223e-09, 1.6298e-08, 6.0536e-09, ..., 1.8626e-09, + -1.8626e-09, 3.7253e-09], + [-9.7789e-09, 9.3132e-10, 6.9849e-09, ..., -2.7940e-09, + 2.7940e-09, 1.3970e-09]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0306, -0.0367, -0.0144, 0.0011, 0.0009, 0.0075, 0.0032, -0.0096, + 0.0138, -0.0449], device='cuda:0'), grad: tensor([-1.5367e-08, -2.0023e-08, -3.4086e-07, 8.8941e-08, 1.4435e-08, + -6.7987e-08, 1.0896e-07, -3.4506e-07, 1.1455e-07, 4.5542e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 311---------------------------------------------------- +epoch 311, time 218.60, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4692 re_mapping 0.0034 re_causal 0.0104 /// teacc 99.24 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.0906, 0.0116, 0.1327, ..., 0.0109, -0.0957, -0.0948], + [-0.2014, -0.2228, -0.2050, ..., 0.0918, 0.0657, 0.4195], + [-0.2117, -0.1646, -0.2011, ..., -0.2734, 0.0197, -0.2575], + ..., + [-0.1989, -0.2579, -0.0028, ..., -0.2428, -0.1201, -0.4044], + [ 0.1428, 0.0333, -0.1546, ..., -0.2420, 0.0278, -0.2330], + [ 0.1419, 0.0919, -0.1365, ..., 0.1077, -0.2433, -0.2722]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.8161e-08, 2.3283e-08, ..., 2.3283e-09, + 1.8626e-09, 2.3283e-09], + [ 4.6566e-10, 5.8208e-08, 7.4040e-08, ..., -6.6590e-08, + -3.4459e-08, -1.0477e-07], + [ 0.0000e+00, 2.3283e-09, 2.7940e-09, ..., 1.8626e-08, + 9.7789e-09, 2.9337e-08], + ..., + [ 0.0000e+00, -8.8941e-08, -1.1502e-07, ..., 9.3132e-10, + 4.6566e-10, 1.3970e-09], + [ 2.3283e-09, 5.5879e-09, 0.0000e+00, ..., 3.7253e-09, + 1.8626e-09, 3.2596e-09], + [-9.3132e-10, 6.5193e-09, 1.2107e-08, ..., 9.3132e-10, + 3.7253e-09, 1.8626e-09]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0303, -0.0367, -0.0144, 0.0013, 0.0009, 0.0074, 0.0031, -0.0096, + 0.0138, -0.0449], device='cuda:0'), grad: tensor([ 1.6391e-07, 3.5716e-07, 6.4261e-08, -1.7229e-08, 2.8405e-08, + 2.1886e-08, 1.2107e-07, -8.3726e-07, 3.2596e-08, 7.2643e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 217.86, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4657 re_mapping 0.0034 re_causal 0.0098 /// teacc 99.15 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.0914, 0.0118, 0.1326, ..., 0.0107, -0.0963, -0.0948], + [-0.2018, -0.2232, -0.2054, ..., 0.0924, 0.0657, 0.4196], + [-0.2128, -0.1653, -0.2028, ..., -0.2741, 0.0196, -0.2575], + ..., + [-0.1990, -0.2583, -0.0026, ..., -0.2436, -0.1212, -0.4058], + [ 0.1430, 0.0330, -0.1554, ..., -0.2425, 0.0281, -0.2331], + [ 0.1420, 0.0918, -0.1364, ..., 0.1080, -0.2432, -0.2727]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.3970e-09, 7.4506e-09, ..., 1.3970e-08, + 4.6566e-10, 4.6566e-10], + [ 6.5193e-09, 7.4506e-09, 1.6484e-07, ..., 4.3306e-08, + -6.9849e-09, -5.5414e-08], + [ 4.3772e-08, 5.9605e-08, 2.7940e-09, ..., 6.0536e-09, + 1.3970e-09, 1.3970e-09], + ..., + [ 2.4214e-08, 2.9337e-08, -3.6089e-07, ..., -1.0012e-07, + 1.2573e-08, 3.3528e-08], + [-7.0315e-08, -9.2667e-08, 1.3970e-09, ..., 2.7940e-09, + 9.3132e-10, 4.6566e-10], + [-4.5169e-08, -2.1886e-08, 1.7369e-07, ..., 4.6566e-09, + 6.0536e-09, 1.7229e-08]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0308, -0.0366, -0.0145, 0.0012, 0.0007, 0.0077, 0.0037, -0.0097, + 0.0133, -0.0449], device='cuda:0'), grad: tensor([ 4.9826e-08, 1.1800e-06, 2.9709e-07, 1.3458e-07, 2.8266e-07, + -5.1223e-08, -1.0850e-07, -2.5183e-06, -4.3027e-07, 1.1707e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 217.80, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4195 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.22 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.0916, 0.0119, 0.1327, ..., 0.0107, -0.0966, -0.0949], + [-0.2023, -0.2255, -0.2057, ..., 0.0921, 0.0656, 0.4196], + [-0.2131, -0.1627, -0.2037, ..., -0.2743, 0.0196, -0.2575], + ..., + [-0.1991, -0.2588, -0.0028, ..., -0.2435, -0.1197, -0.4061], + [ 0.1433, 0.0331, -0.1555, ..., -0.2427, 0.0279, -0.2331], + [ 0.1421, 0.0919, -0.1366, ..., 0.1081, -0.2434, -0.2732]], + device='cuda:0'), grad: tensor([[ 1.1036e-07, 3.8091e-07, 1.1316e-07, ..., 9.2387e-07, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 3.7253e-09, 4.6566e-10, ..., -6.0536e-09, + -2.3283e-09, -2.4680e-08], + [ 1.3970e-09, 1.8626e-09, 0.0000e+00, ..., 2.3283e-09, + 1.3970e-09, 4.6566e-10], + ..., + [ 1.0710e-08, 1.0245e-08, 4.6566e-10, ..., 2.9802e-08, + 7.9162e-09, 2.4680e-08], + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., 3.2596e-09, + 9.3132e-10, 4.6566e-10], + [-1.2992e-07, -4.0932e-07, -1.1828e-07, ..., -9.9000e-07, + 1.3970e-09, 9.3132e-10]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0309, -0.0373, -0.0139, 0.0006, 0.0007, 0.0076, 0.0038, -0.0086, + 0.0131, -0.0449], device='cuda:0'), grad: tensor([ 1.7509e-06, -2.3749e-08, -1.2666e-07, -2.0955e-08, 7.3109e-08, + 1.4808e-07, -1.4342e-07, 2.0675e-07, 2.0955e-08, -1.8757e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 217.62, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4450 re_mapping 0.0034 re_causal 0.0100 /// teacc 99.21 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.0920, 0.0123, 0.1332, ..., 0.0109, -0.0969, -0.0949], + [-0.2027, -0.2257, -0.2059, ..., 0.0921, 0.0657, 0.4196], + [-0.2134, -0.1627, -0.2040, ..., -0.2744, 0.0196, -0.2575], + ..., + [-0.1992, -0.2591, -0.0029, ..., -0.2436, -0.1201, -0.4064], + [ 0.1445, 0.0340, -0.1555, ..., -0.2426, 0.0285, -0.2332], + [ 0.1425, 0.0920, -0.1367, ..., 0.1087, -0.2431, -0.2734]], + device='cuda:0'), grad: tensor([[ 2.5332e-07, 6.8964e-07, 3.1898e-07, ..., 5.3039e-07, + 5.7695e-07, 0.0000e+00], + [ 1.0710e-08, 1.3504e-08, 2.7940e-09, ..., -5.4017e-08, + -5.8673e-08, -7.4506e-08], + [ 2.4214e-08, 4.9360e-08, 2.0489e-08, ..., 3.6322e-08, + 4.7497e-08, 4.6566e-09], + ..., + [ 2.4214e-08, 2.4214e-08, 9.3132e-10, ..., 6.1467e-08, + 8.4750e-08, 6.8918e-08], + [-5.4017e-08, -5.1688e-08, 8.8476e-09, ..., 2.4214e-08, + -6.9384e-08, 0.0000e+00], + [-2.1979e-07, -1.3132e-07, 1.8626e-09, ..., -1.6298e-07, + 1.0245e-08, 9.3132e-10]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0307, -0.0373, -0.0139, 0.0006, 0.0003, 0.0075, 0.0038, -0.0085, + 0.0135, -0.0446], device='cuda:0'), grad: tensor([ 2.6654e-06, -2.3656e-07, 2.3609e-07, -3.1684e-06, 5.8347e-07, + 2.5565e-07, 3.5902e-07, 3.9162e-07, -3.1758e-07, -7.5577e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 217.33, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4690 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.22 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.0928, 0.0129, 0.1353, ..., 0.0122, -0.0968, -0.0946], + [-0.2031, -0.2260, -0.2072, ..., 0.0918, 0.0656, 0.4196], + [-0.2139, -0.1628, -0.2046, ..., -0.2748, 0.0196, -0.2575], + ..., + [-0.1995, -0.2600, -0.0028, ..., -0.2443, -0.1206, -0.4070], + [ 0.1447, 0.0340, -0.1557, ..., -0.2427, 0.0286, -0.2334], + [ 0.1428, 0.0922, -0.1368, ..., 0.1092, -0.2430, -0.2738]], + device='cuda:0'), grad: tensor([[-1.5460e-07, -1.7555e-07, -4.2561e-07, ..., -2.8824e-07, + -2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 2.3283e-09, 6.0536e-09, ..., 2.7940e-09, + 4.1910e-09, -2.3283e-09], + [ 3.2596e-09, 3.7253e-09, 1.3970e-08, ..., 5.5879e-09, + -5.4948e-08, 9.3132e-10], + ..., + [ 4.6566e-10, 9.3132e-10, 3.7253e-09, ..., 2.3283e-09, + 4.2841e-08, 2.7940e-09], + [ 3.0268e-08, 2.8405e-08, 7.4506e-08, ..., 5.0757e-08, + 1.0710e-08, 0.0000e+00], + [ 8.8010e-08, 8.2888e-08, 2.1560e-07, ..., 1.4342e-07, + 1.8161e-08, 4.6566e-10]], device='cuda:0') +Epoch 317, bias, value: tensor([-2.9444e-02, -3.7395e-02, -1.3889e-02, 7.2543e-04, 4.5095e-05, + 7.5157e-03, 3.6122e-03, -8.2224e-03, 1.3433e-02, -4.4423e-02], + device='cuda:0'), grad: tensor([-1.0040e-06, 7.5437e-08, 1.3970e-08, 5.3365e-07, -9.9186e-08, + 4.9360e-08, 1.8347e-07, -5.7742e-07, 2.9057e-07, 5.3877e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 316---------------------------------------------------- +epoch 316, time 218.51, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4852 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.28 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.0931, 0.0130, 0.1357, ..., 0.0122, -0.0973, -0.0952], + [-0.2036, -0.2262, -0.2076, ..., 0.0924, 0.0657, 0.4197], + [-0.2149, -0.1630, -0.2053, ..., -0.2766, 0.0196, -0.2576], + ..., + [-0.2001, -0.2620, -0.0029, ..., -0.2456, -0.1223, -0.4079], + [ 0.1453, 0.0339, -0.1556, ..., -0.2429, 0.0294, -0.2335], + [ 0.1428, 0.0922, -0.1370, ..., 0.1093, -0.2432, -0.2743]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 1.3970e-09, ..., 8.6613e-08, + 0.0000e+00, 1.4016e-07], + [ 1.3970e-09, 9.3132e-10, 4.6566e-10, ..., -2.0070e-07, + 0.0000e+00, -3.5251e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.0710e-08, + 0.0000e+00, 1.5367e-08], + ..., + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 2.8405e-08, + 0.0000e+00, 4.0513e-08], + [ 4.6566e-10, 0.0000e+00, 3.2596e-09, ..., 5.0291e-08, + 0.0000e+00, 7.3109e-08], + [-5.3085e-08, -3.6322e-08, 1.3970e-09, ..., -8.5682e-08, + 4.6566e-10, 2.2817e-08]], device='cuda:0') +Epoch 318, bias, value: tensor([-2.9446e-02, -3.7196e-02, -1.3949e-02, 1.0028e-03, 1.7857e-05, + 9.6802e-03, 9.4074e-04, -9.2173e-03, 1.3513e-02, -4.4438e-02], + device='cuda:0'), grad: tensor([ 3.5902e-07, -8.4611e-07, 3.3528e-08, 1.7695e-08, 1.9511e-07, + 1.2247e-07, -1.9092e-08, -6.0536e-09, 2.0349e-07, -4.4238e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 217.72, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4534 re_mapping 0.0038 re_causal 0.0103 /// teacc 99.16 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.0935, 0.0125, 0.1357, ..., 0.0119, -0.0989, -0.0955], + [-0.2038, -0.2266, -0.2085, ..., 0.0930, 0.0657, 0.4197], + [-0.2154, -0.1632, -0.2064, ..., -0.2769, 0.0196, -0.2576], + ..., + [-0.2003, -0.2623, -0.0031, ..., -0.2465, -0.1236, -0.4095], + [ 0.1449, 0.0338, -0.1555, ..., -0.2432, 0.0298, -0.2337], + [ 0.1435, 0.0925, -0.1373, ..., 0.1103, -0.2436, -0.2764]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -1.3970e-09, -1.8626e-09, ..., 2.3283e-09, + 2.3283e-09, 4.6566e-10], + [ 3.2596e-09, 3.2596e-09, 3.2596e-08, ..., 9.3132e-10, + 6.5658e-08, -1.5367e-08], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 6.9849e-09, + 3.2596e-09, 4.6566e-10], + ..., + [ 1.2107e-08, -1.3970e-09, 1.8626e-09, ..., 1.7695e-08, + 6.5193e-09, 7.9162e-09], + [ 2.7707e-07, 1.8766e-07, 9.3132e-10, ..., 2.3283e-09, + 1.2852e-07, 1.3970e-09], + [-7.0781e-08, -3.4459e-08, 6.0536e-09, ..., -9.4064e-08, + 6.9849e-09, 2.7940e-09]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0300, -0.0372, -0.0139, 0.0010, -0.0003, 0.0096, 0.0009, -0.0095, + 0.0137, -0.0440], device='cuda:0'), grad: tensor([ 1.1642e-08, 2.0256e-07, 3.9116e-08, -6.2538e-07, 6.0536e-09, + 3.0268e-08, -5.7276e-08, -7.4506e-08, 6.6170e-07, -2.0023e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 217.73, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4469 re_mapping 0.0034 re_causal 0.0097 /// teacc 99.19 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.0937, 0.0126, 0.1351, ..., 0.0114, -0.0995, -0.0956], + [-0.2040, -0.2267, -0.2086, ..., 0.0941, 0.0658, 0.4198], + [-0.2159, -0.1637, -0.2070, ..., -0.2776, 0.0196, -0.2576], + ..., + [-0.2005, -0.2631, -0.0029, ..., -0.2472, -0.1248, -0.4103], + [ 0.1448, 0.0337, -0.1556, ..., -0.2442, 0.0290, -0.2350], + [ 0.1437, 0.0926, -0.1375, ..., 0.1106, -0.2439, -0.2772]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 9.3132e-09, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 3.2596e-09, 8.8476e-09, 0.0000e+00, ..., 1.3970e-09, + 3.7253e-09, 0.0000e+00], + [ 5.5879e-09, 6.9849e-09, 0.0000e+00, ..., 1.8626e-09, + 2.5146e-08, 0.0000e+00], + ..., + [ 2.2817e-08, 3.0734e-08, 0.0000e+00, ..., 6.9849e-09, + 3.7253e-09, 0.0000e+00], + [-7.4040e-08, 1.1642e-08, 0.0000e+00, ..., 4.6566e-10, + -4.5635e-08, 0.0000e+00], + [ 9.3132e-09, 5.1688e-08, -4.6566e-10, ..., -3.7253e-08, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0308, -0.0371, -0.0139, 0.0014, -0.0003, 0.0095, 0.0010, -0.0098, + 0.0131, -0.0439], device='cuda:0'), grad: tensor([ 3.8184e-08, 3.8184e-08, 7.4971e-08, -3.8184e-08, 8.1025e-08, + -1.6093e-06, 1.2470e-06, 9.9652e-08, 7.4506e-09, 7.3574e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 217.55, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4944 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.13 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.0943, 0.0126, 0.1356, ..., 0.0115, -0.0990, -0.0954], + [-0.2042, -0.2269, -0.2099, ..., 0.0949, 0.0658, 0.4198], + [-0.2168, -0.1639, -0.2081, ..., -0.2778, 0.0196, -0.2576], + ..., + [-0.2007, -0.2634, -0.0020, ..., -0.2487, -0.1252, -0.4107], + [ 0.1452, 0.0340, -0.1556, ..., -0.2444, 0.0298, -0.2350], + [ 0.1439, 0.0928, -0.1376, ..., 0.1109, -0.2439, -0.2776]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-10, 1.3970e-08, ..., 2.3283e-08, + 9.3132e-10, 9.3132e-10], + [ 1.8626e-09, 3.7253e-09, 4.6566e-09, ..., -1.7229e-08, + -3.5390e-08, -6.4261e-08], + [ 1.3970e-09, 3.2596e-09, 7.9162e-09, ..., 1.4901e-08, + 9.3132e-09, 1.1176e-08], + ..., + [ 3.7253e-09, 7.9162e-09, -8.1491e-08, ..., 2.0955e-08, + 3.3528e-08, 4.3306e-08], + [-6.9849e-09, -7.9162e-09, 2.8405e-08, ..., 4.0513e-08, + -4.6566e-10, 1.3970e-09], + [-1.5832e-08, -7.9162e-09, 5.1223e-09, ..., -1.2107e-08, + 2.7940e-09, 2.7940e-09]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0306, -0.0371, -0.0139, 0.0014, -0.0005, 0.0094, 0.0010, -0.0101, + 0.0133, -0.0437], device='cuda:0'), grad: tensor([ 6.5658e-08, -1.1222e-07, 6.1467e-08, -1.1828e-07, 3.7719e-07, + 2.6124e-07, -3.9348e-07, -1.8952e-07, 8.1956e-08, -2.4680e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 217.78, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4609 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.15 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.0945, 0.0130, 0.1359, ..., 0.0118, -0.0993, -0.0960], + [-0.2045, -0.2270, -0.2104, ..., 0.0964, 0.0660, 0.4199], + [-0.2178, -0.1640, -0.2098, ..., -0.2805, 0.0194, -0.2577], + ..., + [-0.2008, -0.2638, -0.0013, ..., -0.2493, -0.1263, -0.4115], + [ 0.1453, 0.0338, -0.1560, ..., -0.2446, 0.0300, -0.2355], + [ 0.1440, 0.0928, -0.1380, ..., 0.1110, -0.2442, -0.2787]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 7.4506e-09, 4.6566e-10, ..., 9.7789e-09, + 3.2596e-09, 0.0000e+00], + [ 1.0710e-08, 1.7695e-08, 1.8626e-09, ..., 6.9849e-09, + 1.3504e-08, -1.1176e-08], + [ 1.1642e-08, 1.7695e-08, 0.0000e+00, ..., 1.0710e-08, + 1.4901e-08, 4.6566e-10], + ..., + [ 2.2817e-08, 2.7940e-08, 1.8626e-09, ..., 2.7474e-08, + 2.3749e-08, 7.4506e-09], + [ 2.3283e-09, 3.2596e-09, 4.6566e-10, ..., 9.7789e-09, + 1.3970e-09, 0.0000e+00], + [-6.2399e-08, -1.8161e-08, 1.8161e-08, ..., -3.4459e-08, + 3.3993e-08, 2.3283e-09]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0306, -0.0369, -0.0141, 0.0015, -0.0005, 0.0094, 0.0010, -0.0099, + 0.0132, -0.0438], device='cuda:0'), grad: tensor([ 4.0513e-08, 6.8452e-08, 1.4156e-07, -8.3540e-07, 2.6077e-08, + 5.5740e-07, 7.4506e-09, 1.7229e-08, 1.6764e-08, -4.6100e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 217.67, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4563 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.16 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.0949, 0.0129, 0.1362, ..., 0.0117, -0.1005, -0.0961], + [-0.2048, -0.2272, -0.2105, ..., 0.0972, 0.0661, 0.4200], + [-0.2187, -0.1644, -0.2107, ..., -0.2809, 0.0193, -0.2577], + ..., + [-0.2009, -0.2639, -0.0026, ..., -0.2502, -0.1268, -0.4121], + [ 0.1454, 0.0338, -0.1561, ..., -0.2448, 0.0303, -0.2355], + [ 0.1442, 0.0928, -0.1384, ..., 0.1111, -0.2444, -0.2793]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3970e-09, 1.2061e-07, ..., 9.0338e-08, + 9.3132e-10, 1.3970e-09], + [ 4.1910e-09, 1.3970e-09, 9.3132e-10, ..., -4.1444e-08, + -2.2352e-08, -4.9826e-08], + [ 7.9162e-09, 9.3132e-09, 4.6566e-10, ..., 2.3283e-09, + 6.0536e-09, 1.3970e-09], + ..., + [ 4.7032e-08, 1.0710e-08, 2.7940e-09, ..., -5.4529e-07, + 1.6298e-08, 1.9558e-08], + [-3.2596e-09, -4.1910e-09, 4.6566e-10, ..., 4.6566e-10, + -3.7253e-09, 0.0000e+00], + [ 9.3132e-10, 1.3970e-09, 3.2596e-09, ..., 5.6811e-07, + 1.0245e-08, 1.8626e-08]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0307, -0.0369, -0.0141, 0.0016, -0.0005, 0.0094, 0.0011, -0.0100, + 0.0133, -0.0439], device='cuda:0'), grad: tensor([ 3.0827e-07, -6.0536e-09, -2.6496e-07, -8.6473e-07, -9.3132e-10, + 9.4157e-07, -2.5285e-07, -1.7181e-05, 3.4459e-08, 1.7270e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 217.93, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4716 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.18 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.0954, 0.0131, 0.1366, ..., 0.0118, -0.1011, -0.0962], + [-0.2050, -0.2274, -0.2109, ..., 0.0975, 0.0661, 0.4200], + [-0.2192, -0.1645, -0.2114, ..., -0.2809, 0.0193, -0.2577], + ..., + [-0.2011, -0.2641, -0.0029, ..., -0.2505, -0.1277, -0.4128], + [ 0.1457, 0.0342, -0.1558, ..., -0.2449, 0.0313, -0.2356], + [ 0.1443, 0.0930, -0.1392, ..., 0.1113, -0.2454, -0.2799]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -9.3132e-10, -2.7940e-09, ..., -1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 3.2596e-09, 3.7253e-09, 1.3970e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 4.6566e-10], + [-1.2573e-08, -1.2107e-08, 2.3283e-09, ..., 1.8626e-09, + -9.3132e-09, 0.0000e+00], + [-6.9849e-09, -4.1910e-09, 9.3132e-10, ..., -7.9162e-09, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0307, -0.0369, -0.0141, 0.0013, -0.0004, 0.0095, 0.0010, -0.0099, + 0.0137, -0.0441], device='cuda:0'), grad: tensor([ 4.6566e-09, 5.7742e-08, -1.0235e-06, 1.5926e-07, 2.0023e-08, + 3.8184e-08, -1.6764e-08, 5.5740e-07, 2.2119e-07, -1.2107e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 217.91, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4577 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.23 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.0959, 0.0131, 0.1355, ..., 0.0108, -0.1019, -0.0970], + [-0.2055, -0.2276, -0.2110, ..., 0.0989, 0.0661, 0.4201], + [-0.2196, -0.1647, -0.2117, ..., -0.2816, 0.0193, -0.2577], + ..., + [-0.2012, -0.2622, -0.0031, ..., -0.2514, -0.1256, -0.4145], + [ 0.1459, 0.0341, -0.1560, ..., -0.2457, 0.0302, -0.2372], + [ 0.1446, 0.0932, -0.1392, ..., 0.1116, -0.2456, -0.2805]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 7.4506e-09, 3.7253e-09, ..., -1.8626e-09, + 0.0000e+00, -8.3819e-09], + [ 1.1176e-08, 1.6764e-08, 0.0000e+00, ..., 8.3819e-09, + 1.5832e-08, 2.7940e-09], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, + 5.5879e-09, 5.5879e-09], + [-1.8626e-08, -1.1409e-06, -2.2352e-06, ..., 9.3132e-10, + -1.0710e-06, -5.2154e-08], + [-6.5193e-09, -2.0489e-08, 1.7975e-07, ..., -2.1420e-08, + 2.3842e-07, 0.0000e+00]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0316, -0.0370, -0.0141, 0.0007, -0.0007, 0.0094, 0.0012, -0.0081, + 0.0130, -0.0439], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.7008e-08, 2.6077e-08, 3.0734e-08, -5.3644e-07, + 1.7881e-07, 5.0813e-06, 3.7253e-09, -5.2974e-06, 4.9081e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 217.96, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4763 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.22 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.0969, 0.0130, 0.1351, ..., 0.0105, -0.1026, -0.0973], + [-0.2082, -0.2282, -0.2114, ..., 0.0984, 0.0660, 0.4202], + [-0.2202, -0.1648, -0.2124, ..., -0.2818, 0.0193, -0.2577], + ..., + [-0.2014, -0.2624, -0.0033, ..., -0.2520, -0.1260, -0.4157], + [ 0.1458, 0.0345, -0.1558, ..., -0.2462, 0.0304, -0.2374], + [ 0.1451, 0.0933, -0.1394, ..., 0.1123, -0.2452, -0.2811]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -7.7300e-08, -1.3877e-07, ..., -7.7300e-08, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-08, 3.9116e-08, 5.7742e-08, ..., 2.5146e-08, + 4.1910e-08, -1.2107e-08], + [ 1.3970e-08, -1.0245e-08, 3.7253e-09, ..., 4.6566e-09, + 1.7695e-08, 7.4506e-09], + ..., + [ 3.7253e-09, 3.7253e-09, 4.6566e-09, ..., 3.7253e-09, + 1.0245e-08, 1.8626e-09], + [-1.0151e-07, 1.6764e-08, 6.4261e-08, ..., 3.6322e-08, + -2.4959e-07, 2.7940e-09], + [ 0.0000e+00, 5.5879e-09, 1.5832e-08, ..., 3.7253e-09, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0323, -0.0371, -0.0142, 0.0007, -0.0008, 0.0092, 0.0015, -0.0081, + 0.0130, -0.0436], device='cuda:0'), grad: tensor([-4.7032e-07, 3.3528e-07, -3.2596e-08, 6.4261e-08, -2.3283e-08, + 1.5460e-07, 3.0175e-07, 5.7742e-08, -4.5169e-07, 6.5193e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 217.89, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4702 re_mapping 0.0032 re_causal 0.0089 /// teacc 99.25 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.0987, 0.0132, 0.1362, ..., 0.0109, -0.1044, -0.0974], + [-0.2095, -0.2286, -0.2129, ..., 0.0977, 0.0659, 0.4202], + [-0.2234, -0.1660, -0.2157, ..., -0.2829, 0.0191, -0.2578], + ..., + [-0.2020, -0.2636, -0.0033, ..., -0.2531, -0.1268, -0.4166], + [ 0.1444, 0.0332, -0.1558, ..., -0.2470, 0.0316, -0.2362], + [ 0.1454, 0.0933, -0.1398, ..., 0.1130, -0.2455, -0.2814]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 3.7253e-09, ..., -9.3132e-10, + 5.7742e-08, -4.6566e-09], + [ 8.3819e-09, 9.3132e-09, 0.0000e+00, ..., 4.6566e-09, + -5.4948e-08, 9.3132e-10], + ..., + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 9.3132e-10], + [ 1.3039e-08, 1.5832e-08, 1.8626e-09, ..., 8.3819e-09, + 1.3039e-08, 9.3132e-10], + [-9.3132e-09, -4.6566e-09, 4.6566e-09, ..., -9.3132e-09, + 3.7253e-09, 9.3132e-10]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0321, -0.0372, -0.0143, 0.0012, -0.0008, 0.0094, 0.0016, -0.0083, + 0.0133, -0.0433], device='cuda:0'), grad: tensor([ 2.1420e-08, 5.2713e-07, -6.0257e-07, -8.0094e-08, 2.7940e-09, + -4.8354e-06, 4.8801e-06, 2.0489e-08, 6.7055e-08, -1.0245e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 217.54, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4801 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.21 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.0991, 0.0133, 0.1364, ..., 0.0101, -0.1059, -0.0975], + [-0.2102, -0.2290, -0.2142, ..., 0.0980, 0.0659, 0.4203], + [-0.2255, -0.1665, -0.2176, ..., -0.2834, 0.0191, -0.2578], + ..., + [-0.2022, -0.2637, -0.0033, ..., -0.2540, -0.1269, -0.4176], + [ 0.1454, 0.0334, -0.1560, ..., -0.2471, 0.0333, -0.2361], + [ 0.1456, 0.0934, -0.1401, ..., 0.1133, -0.2457, -0.2820]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -6.5193e-09, -1.8626e-08, ..., -1.1176e-08, + 9.3132e-10, 9.3132e-10], + [ 4.6566e-09, 9.3132e-10, 1.8626e-09, ..., -4.6566e-09, + 8.6613e-08, -4.6566e-09], + [ 6.5193e-09, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 2.5146e-08, 8.3819e-09], + ..., + [ 2.7940e-09, 1.8626e-09, 2.7940e-09, ..., 6.5193e-09, + 8.3819e-09, 1.2107e-08], + [-4.5635e-08, -1.0245e-08, 9.3132e-10, ..., 1.8626e-09, + -1.6112e-07, -1.3970e-08], + [-6.5193e-09, -1.8626e-09, 1.6764e-08, ..., -5.5879e-09, + 1.0245e-08, 9.3132e-10]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0326, -0.0377, -0.0143, 0.0008, -0.0011, 0.0099, 0.0014, -0.0066, + 0.0140, -0.0433], device='cuda:0'), grad: tensor([-4.3772e-08, 2.2817e-07, 7.8231e-08, 3.4459e-08, -5.1223e-08, + 1.6764e-08, 1.2387e-07, 4.7497e-08, -4.5914e-07, 2.0489e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 217.90, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4661 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.18 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.0994, 0.0136, 0.1376, ..., 0.0108, -0.1069, -0.0978], + [-0.2105, -0.2292, -0.2151, ..., 0.0986, 0.0658, 0.4203], + [-0.2266, -0.1667, -0.2176, ..., -0.2846, 0.0190, -0.2578], + ..., + [-0.2025, -0.2643, -0.0040, ..., -0.2545, -0.1272, -0.4182], + [ 0.1466, 0.0335, -0.1562, ..., -0.2473, 0.0353, -0.2348], + [ 0.1457, 0.0934, -0.1406, ..., 0.1134, -0.2459, -0.2829]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -4.6566e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 2.0955e-07, ..., 1.8626e-09, + 3.5390e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + -9.3132e-10, 0.0000e+00], + [-6.5193e-09, -1.8626e-09, 3.7253e-09, ..., -4.6566e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0317, -0.0378, -0.0143, 0.0010, -0.0011, 0.0098, 0.0012, -0.0067, + 0.0150, -0.0433], device='cuda:0'), grad: tensor([-9.3132e-09, 4.6566e-09, 2.7940e-09, 1.3039e-08, -5.3365e-07, + -1.3039e-08, -9.3132e-09, 5.3737e-07, 8.3819e-09, -2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 217.76, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4743 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.22 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.0999, 0.0141, 0.1384, ..., 0.0112, -0.1073, -0.0979], + [-0.2108, -0.2294, -0.2166, ..., 0.0986, 0.0658, 0.4203], + [-0.2283, -0.1673, -0.2181, ..., -0.2847, 0.0190, -0.2578], + ..., + [-0.2026, -0.2645, -0.0043, ..., -0.2548, -0.1275, -0.4192], + [ 0.1467, 0.0333, -0.1566, ..., -0.2479, 0.0350, -0.2357], + [ 0.1457, 0.0934, -0.1408, ..., 0.1134, -0.2460, -0.2833]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, 0.0000e+00, ..., -4.6566e-09, + -1.4901e-08, -1.1176e-08], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 6.5193e-09, -9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 4.6566e-09], + [ 0.0000e+00, 9.6858e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-9.3132e-10, 1.8626e-09, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0313, -0.0378, -0.0143, 0.0013, -0.0011, 0.0095, 0.0018, -0.0069, + 0.0145, -0.0433], device='cuda:0'), grad: tensor([ 1.6764e-08, 7.4506e-09, 1.0245e-08, 3.3341e-06, 2.8871e-08, + -4.4554e-06, 7.1712e-07, 1.5832e-08, 3.0268e-07, 1.3039e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 329---------------------------------------------------- +epoch 329, time 218.76, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4728 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.29 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.1004, 0.0147, 0.1385, ..., 0.0114, -0.1075, -0.0980], + [-0.2109, -0.2295, -0.2171, ..., 0.1010, 0.0658, 0.4204], + [-0.2284, -0.1674, -0.2182, ..., -0.2849, 0.0190, -0.2579], + ..., + [-0.2027, -0.2647, -0.0044, ..., -0.2574, -0.1277, -0.4204], + [ 0.1471, 0.0332, -0.1567, ..., -0.2479, 0.0353, -0.2358], + [ 0.1459, 0.0936, -0.1410, ..., 0.1136, -0.2462, -0.2847]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-10, -9.3132e-10, ..., 2.7940e-09, + 1.8626e-09, 1.8626e-09], + [ 4.6566e-09, 5.5879e-09, 1.9558e-08, ..., 2.0489e-08, + 2.7008e-08, 1.4901e-08], + [ 1.8626e-09, 1.8626e-09, 2.7940e-09, ..., 9.3132e-10, + -4.8429e-08, 9.3132e-10], + ..., + [ 1.2107e-08, 1.1176e-08, 0.0000e+00, ..., 9.3132e-10, + 1.0245e-08, 1.8626e-09], + [-8.1956e-08, -7.8231e-08, 1.0245e-08, ..., 1.5832e-08, + 8.3819e-09, 8.3819e-09], + [ 1.6764e-08, 1.8626e-08, 2.7940e-09, ..., -9.3132e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0313, -0.0375, -0.0142, 0.0008, -0.0012, 0.0096, 0.0018, -0.0079, + 0.0145, -0.0433], device='cuda:0'), grad: tensor([ 1.3970e-08, 2.5611e-07, -2.8033e-07, 5.7742e-08, -1.5832e-08, + 2.5500e-06, -2.4587e-06, 1.1828e-07, -3.4645e-07, 1.2573e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 217.79, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4532 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.17 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.1008, 0.0146, 0.1386, ..., 0.0110, -0.1077, -0.0986], + [-0.2111, -0.2296, -0.2172, ..., 0.1013, 0.0658, 0.4205], + [-0.2295, -0.1678, -0.2196, ..., -0.2852, 0.0190, -0.2579], + ..., + [-0.2028, -0.2652, -0.0059, ..., -0.2576, -0.1279, -0.4211], + [ 0.1473, 0.0325, -0.1580, ..., -0.2482, 0.0354, -0.2363], + [ 0.1460, 0.0935, -0.1420, ..., 0.1137, -0.2469, -0.2865]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, -4.6566e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0319, -0.0374, -0.0142, 0.0004, -0.0008, 0.0119, -0.0004, -0.0081, + 0.0140, -0.0436], device='cuda:0'), grad: tensor([ 3.7253e-09, 2.3823e-06, 2.3749e-07, -1.4808e-07, 5.5879e-09, + 1.8533e-07, 8.3819e-09, -2.9206e-06, 2.3004e-07, 2.0489e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 217.52, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4603 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.25 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.1012, 0.0148, 0.1389, ..., 0.0111, -0.1079, -0.0989], + [-0.2114, -0.2299, -0.2171, ..., 0.1018, 0.0659, 0.4205], + [-0.2304, -0.1681, -0.2204, ..., -0.2853, 0.0190, -0.2579], + ..., + [-0.2028, -0.2654, -0.0046, ..., -0.2576, -0.1280, -0.4218], + [ 0.1469, 0.0321, -0.1581, ..., -0.2492, 0.0355, -0.2373], + [ 0.1462, 0.0936, -0.1433, ..., 0.1139, -0.2474, -0.2870]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 5.5879e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + -9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-08, ..., 9.3132e-09, + 0.0000e+00, 0.0000e+00], + [-3.1665e-08, -1.5832e-08, -1.8626e-09, ..., -3.3528e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 333, bias, value: tensor([-3.1827e-02, -3.7383e-02, -1.4210e-02, 5.7662e-05, -6.2588e-04, + 1.2480e-02, -8.9661e-04, -7.5800e-03, 1.3585e-02, -4.4142e-02], + device='cuda:0'), grad: tensor([ 1.9558e-08, 4.2841e-08, 3.7253e-09, -9.3132e-10, 8.8476e-08, + 6.6776e-07, -8.8196e-07, -2.7940e-08, 1.9185e-07, -9.8720e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 217.25, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.5091 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.15 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.1020, 0.0148, 0.1391, ..., 0.0104, -0.1084, -0.0989], + [-0.2119, -0.2302, -0.2172, ..., 0.1018, 0.0659, 0.4205], + [-0.2315, -0.1685, -0.2210, ..., -0.2855, 0.0191, -0.2579], + ..., + [-0.2030, -0.2657, -0.0043, ..., -0.2577, -0.1282, -0.4223], + [ 0.1471, 0.0319, -0.1586, ..., -0.2496, 0.0355, -0.2373], + [ 0.1462, 0.0935, -0.1437, ..., 0.1140, -0.2478, -0.2874]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 4.6566e-09, 9.3132e-10, 1.8626e-09, ..., 9.3132e-10, + -2.3935e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -3.7253e-08, ..., -3.7253e-09, + 2.7940e-09, 0.0000e+00], + [-4.6566e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-07, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 4.0978e-08, ..., 1.8626e-09, + 1.6764e-08, 0.0000e+00]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0324, -0.0374, -0.0140, 0.0006, -0.0007, 0.0121, -0.0007, -0.0076, + 0.0129, -0.0442], device='cuda:0'), grad: tensor([ 3.2596e-08, 9.2201e-08, -1.1334e-06, 1.4901e-08, -5.7742e-08, + 1.4901e-08, 6.5193e-09, -3.0082e-07, 1.1139e-06, 2.1979e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 217.40, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4884 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.19 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.1027, 0.0139, 0.1390, ..., 0.0089, -0.1094, -0.1014], + [-0.2124, -0.2305, -0.2161, ..., 0.1023, 0.0660, 0.4207], + [-0.2327, -0.1687, -0.2218, ..., -0.2863, 0.0191, -0.2579], + ..., + [-0.2031, -0.2661, -0.0050, ..., -0.2579, -0.1287, -0.4238], + [ 0.1479, 0.0322, -0.1591, ..., -0.2495, 0.0357, -0.2379], + [ 0.1463, 0.0936, -0.1441, ..., 0.1143, -0.2483, -0.2883]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., -9.3132e-10, + 9.3132e-09, 1.8626e-09], + [ 2.7940e-09, 8.3819e-09, 2.7940e-08, ..., -2.7940e-09, + 4.1910e-08, -2.6077e-08], + [ 2.0489e-08, 1.3970e-08, 3.7253e-09, ..., 9.3132e-10, + 2.2352e-08, 3.7253e-09], + ..., + [ 2.7940e-09, 2.7940e-09, 4.6566e-09, ..., 2.7940e-09, + 1.3039e-08, 1.2107e-08], + [-6.6124e-08, -4.0978e-08, 5.5879e-09, ..., 2.7940e-09, + -5.3085e-08, 0.0000e+00], + [ 3.7253e-09, 9.3132e-09, 7.0781e-08, ..., 1.8626e-09, + 9.8720e-08, 3.7253e-09]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0337, -0.0373, -0.0140, 0.0003, -0.0006, 0.0121, -0.0005, -0.0077, + 0.0127, -0.0443], device='cuda:0'), grad: tensor([ 2.1420e-08, 9.7789e-08, -1.2014e-07, -6.5193e-09, -4.9733e-07, + 1.7323e-07, 5.3085e-08, 2.7008e-08, -4.4703e-08, 2.9337e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 217.50, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4370 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.20 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.1043, 0.0135, 0.1392, ..., 0.0083, -0.1101, -0.1025], + [-0.2129, -0.2308, -0.2163, ..., 0.1026, 0.0660, 0.4208], + [-0.2334, -0.1690, -0.2227, ..., -0.2867, 0.0191, -0.2579], + ..., + [-0.2032, -0.2664, -0.0050, ..., -0.2580, -0.1290, -0.4244], + [ 0.1484, 0.0320, -0.1608, ..., -0.2506, 0.0367, -0.2384], + [ 0.1465, 0.0939, -0.1442, ..., 0.1146, -0.2484, -0.2889]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 1.8626e-08, ..., 0.0000e+00, + 3.3528e-08, 9.3132e-10], + [ 9.3132e-10, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + -1.8626e-08, 0.0000e+00], + [ 2.5146e-08, 2.5146e-08, 9.3132e-10, ..., 1.3970e-08, + -8.3819e-09, 0.0000e+00], + [-3.9116e-08, -3.6322e-08, 1.8626e-09, ..., -1.8626e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0344, -0.0373, -0.0140, -0.0001, -0.0006, 0.0129, -0.0012, -0.0078, + 0.0122, -0.0441], device='cuda:0'), grad: tensor([ 1.0245e-08, 4.8243e-07, 2.2538e-07, 1.1548e-07, -7.1712e-08, + 3.6322e-08, 1.3970e-08, -8.5589e-07, 6.7987e-08, -2.3283e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 217.61, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4793 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.22 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.1050, 0.0135, 0.1393, ..., 0.0082, -0.1106, -0.1026], + [-0.2139, -0.2355, -0.2208, ..., 0.1015, 0.0650, 0.4198], + [-0.2359, -0.1698, -0.2237, ..., -0.2870, 0.0190, -0.2579], + ..., + [-0.2035, -0.2669, -0.0051, ..., -0.2582, -0.1294, -0.4246], + [ 0.1493, 0.0345, -0.1580, ..., -0.2504, 0.0402, -0.2358], + [ 0.1468, 0.0942, -0.1443, ..., 0.1148, -0.2485, -0.2890]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 7.4506e-09, 5.5879e-09, 0.0000e+00, ..., 4.6566e-09, + 5.5879e-09, 9.3132e-10], + ..., + [ 2.6077e-08, 1.1176e-08, 0.0000e+00, ..., 1.1176e-08, + 5.5879e-09, 9.3132e-10], + [ 4.6566e-09, 3.7253e-09, 9.3132e-10, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [-5.5879e-08, -3.4459e-08, 0.0000e+00, ..., -3.5390e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 337, bias, value: tensor([-3.4663e-02, -3.9074e-02, -1.3987e-02, -4.9964e-06, -6.9308e-04, + 1.3304e-02, -1.5729e-03, -7.8389e-03, 1.5363e-02, -4.3956e-02], + device='cuda:0'), grad: tensor([ 6.5193e-09, 1.0245e-08, 2.6077e-08, -9.3132e-08, 3.3528e-08, + 7.5437e-08, 0.0000e+00, 9.3132e-08, 2.1420e-08, -1.6298e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 217.59, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4655 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.10 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.1054, 0.0144, 0.1392, ..., 0.0082, -0.1116, -0.1033], + [-0.2146, -0.2356, -0.2209, ..., 0.1017, 0.0650, 0.4199], + [-0.2370, -0.1702, -0.2259, ..., -0.2873, 0.0190, -0.2580], + ..., + [-0.2036, -0.2673, -0.0053, ..., -0.2585, -0.1297, -0.4255], + [ 0.1480, 0.0339, -0.1587, ..., -0.2518, 0.0400, -0.2363], + [ 0.1470, 0.0943, -0.1444, ..., 0.1150, -0.2486, -0.2894]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -2.7008e-08, -3.8184e-08, ..., -5.1223e-08, + 4.6566e-09, 0.0000e+00], + [ 3.9116e-08, 2.7008e-08, 9.3132e-10, ..., 9.3132e-09, + 7.4506e-09, -4.9360e-08], + [ 3.5390e-08, 2.2352e-08, 9.3132e-10, ..., 2.4214e-08, + 1.4901e-08, 0.0000e+00], + ..., + [ 1.7695e-08, 1.1176e-08, 2.7940e-09, ..., 2.8871e-08, + 1.9558e-08, 3.7253e-08], + [ 5.5879e-08, 3.7253e-08, 1.8626e-09, ..., 3.9116e-08, + 3.0734e-08, 9.3132e-10], + [ 2.4214e-08, 3.1665e-08, 1.9558e-08, ..., 4.6566e-08, + 1.6764e-08, 7.4506e-09]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0349, -0.0391, -0.0140, 0.0002, -0.0008, 0.0132, -0.0010, -0.0080, + 0.0151, -0.0439], device='cuda:0'), grad: tensor([-8.1956e-08, 4.9360e-08, -9.3132e-10, -7.3854e-07, 2.1420e-08, + 2.1793e-07, 1.3039e-08, 1.3970e-07, 2.0862e-07, 1.6857e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 217.85, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4769 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.24 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.1095, 0.0122, 0.1374, ..., 0.0065, -0.1127, -0.1049], + [-0.2150, -0.2357, -0.2210, ..., 0.1019, 0.0651, 0.4200], + [-0.2381, -0.1705, -0.2278, ..., -0.2876, 0.0190, -0.2580], + ..., + [-0.2039, -0.2680, -0.0067, ..., -0.2588, -0.1302, -0.4267], + [ 0.1475, 0.0334, -0.1594, ..., -0.2527, 0.0398, -0.2368], + [ 0.1474, 0.0947, -0.1448, ..., 0.1154, -0.2489, -0.2908]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -2.7940e-09, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -1.1176e-08, + 3.7253e-09, -3.3528e-08], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 6.6496e-07, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -1.0841e-06, 6.5193e-09], + [-9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 2.7940e-09, 2.7940e-09]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0383, -0.0390, -0.0140, 0.0002, -0.0005, 0.0130, -0.0002, -0.0082, + 0.0148, -0.0437], device='cuda:0'), grad: tensor([ 2.7940e-09, -2.2352e-08, 3.0436e-06, -6.3330e-08, 1.7630e-06, + 1.4156e-07, -3.7253e-09, -4.8988e-06, 1.3039e-08, 1.7695e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 217.78, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4830 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.14 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.1109, 0.0117, 0.1382, ..., 0.0046, -0.1125, -0.1051], + [-0.2157, -0.2360, -0.2210, ..., 0.1022, 0.0651, 0.4200], + [-0.2411, -0.1714, -0.2286, ..., -0.2881, 0.0189, -0.2580], + ..., + [-0.2050, -0.2704, -0.0069, ..., -0.2601, -0.1303, -0.4284], + [ 0.1479, 0.0334, -0.1596, ..., -0.2532, 0.0399, -0.2368], + [ 0.1479, 0.0955, -0.1448, ..., 0.1167, -0.2488, -0.2916]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -1.3039e-08, ..., -1.2107e-08, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., -2.7940e-09, + -2.7940e-09, -1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + 1.8626e-09, 7.4506e-09], + ..., + [ 9.3132e-10, 9.3132e-10, -2.4214e-08, ..., 3.7253e-09, + 1.8626e-09, 5.5879e-09], + [ 8.3819e-09, 7.4506e-09, 9.3132e-10, ..., 4.6566e-09, + 2.7940e-09, 9.3132e-10], + [-2.7940e-09, 9.3132e-10, 4.6566e-09, ..., 1.8626e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0393, -0.0391, -0.0140, 0.0006, -0.0010, 0.0129, -0.0001, -0.0086, + 0.0148, -0.0430], device='cuda:0'), grad: tensor([-5.1223e-08, 3.7253e-09, -3.7253e-09, -2.5146e-08, 1.4249e-07, + 1.1176e-08, 2.7940e-09, -1.1083e-07, 1.9558e-08, 1.3970e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 217.54, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4581 re_mapping 0.0032 re_causal 0.0089 /// teacc 99.20 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.1110, 0.0122, 0.1400, ..., 0.0057, -0.1129, -0.1033], + [-0.2166, -0.2362, -0.2217, ..., 0.1017, 0.0650, 0.4200], + [-0.2426, -0.1718, -0.2305, ..., -0.2887, 0.0189, -0.2580], + ..., + [-0.2055, -0.2716, -0.0069, ..., -0.2605, -0.1308, -0.4297], + [ 0.1484, 0.0322, -0.1599, ..., -0.2532, 0.0401, -0.2369], + [ 0.1480, 0.0955, -0.1462, ..., 0.1168, -0.2497, -0.2924]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 6.5193e-09, ..., 9.3132e-10, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 2.7940e-09, 1.8626e-09, ..., -4.6566e-09, + -2.7940e-09, -1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 3.7253e-09], + ..., + [ 0.0000e+00, 2.7940e-09, 7.4506e-09, ..., 2.7940e-09, + 5.5879e-09, 7.4506e-09], + [ 9.3132e-10, 1.3970e-08, 1.1176e-08, ..., 5.5879e-09, + 2.7940e-09, 1.8626e-09], + [-7.4506e-09, 5.3085e-08, 2.2165e-07, ..., -8.3819e-09, + 1.0058e-07, 9.3132e-10]], device='cuda:0') +Epoch 341, bias, value: tensor([-3.8235e-02, -3.9247e-02, -1.3799e-02, -3.3703e-04, -7.1460e-04, + 1.3432e-02, 1.6923e-05, -8.9485e-03, 1.4767e-02, -4.3258e-02], + device='cuda:0'), grad: tensor([ 2.7940e-08, -1.1176e-08, -1.3970e-08, 3.3528e-08, -9.9652e-07, + -6.5193e-08, 3.6322e-08, 5.2154e-08, 8.3819e-08, 8.5961e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 217.58, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4530 re_mapping 0.0032 re_causal 0.0092 /// teacc 99.25 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.1131, 0.0102, 0.1402, ..., 0.0046, -0.1133, -0.1036], + [-0.2177, -0.2364, -0.2217, ..., 0.1023, 0.0650, 0.4203], + [-0.2441, -0.1727, -0.2312, ..., -0.2910, 0.0189, -0.2581], + ..., + [-0.2057, -0.2724, -0.0068, ..., -0.2609, -0.1315, -0.4309], + [ 0.1480, 0.0313, -0.1606, ..., -0.2542, 0.0403, -0.2374], + [ 0.1493, 0.0976, -0.1469, ..., 0.1180, -0.2505, -0.2941]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -2.0489e-08, + -2.1420e-08, -9.3132e-08], + [ 2.1420e-08, 1.7695e-08, 0.0000e+00, ..., 5.5879e-09, + 2.3283e-08, 3.5390e-08], + ..., + [ 3.7253e-09, 4.6566e-09, 1.8626e-09, ..., 7.4506e-09, + 1.2107e-08, 2.9802e-08], + [-3.2596e-08, -3.3528e-08, 6.5193e-09, ..., 6.5193e-09, + -3.2596e-08, 1.2107e-08], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 3.7253e-09]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0400, -0.0392, -0.0138, 0.0011, -0.0008, 0.0120, 0.0002, -0.0092, + 0.0145, -0.0418], device='cuda:0'), grad: tensor([ 8.1956e-08, -1.6484e-07, -9.3132e-08, 1.2759e-07, 9.3132e-09, + -4.4983e-07, 1.8720e-07, 7.0781e-08, 2.0117e-07, 3.6322e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 217.63, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4509 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.12 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.1141, 0.0093, 0.1398, ..., 0.0036, -0.1139, -0.1041], + [-0.2182, -0.2366, -0.2219, ..., 0.1028, 0.0650, 0.4205], + [-0.2452, -0.1732, -0.2319, ..., -0.2923, 0.0188, -0.2582], + ..., + [-0.2059, -0.2728, -0.0074, ..., -0.2613, -0.1323, -0.4333], + [ 0.1485, 0.0313, -0.1609, ..., -0.2548, 0.0406, -0.2377], + [ 0.1499, 0.0978, -0.1533, ..., 0.1177, -0.2583, -0.2961]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.3132e-10, 5.5879e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 9.3132e-10, 4.6566e-09, ..., 4.6566e-09, + 1.8626e-09, 0.0000e+00], + [ 2.7940e-09, 9.3132e-10, 1.8626e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 9.3132e-10, -4.6566e-09, ..., 6.5193e-09, + 9.3132e-10, 0.0000e+00], + [-1.3039e-08, -5.5879e-09, 2.7940e-09, ..., 3.7253e-09, + -1.5832e-08, 0.0000e+00], + [-1.6764e-08, -5.5879e-09, 2.7940e-09, ..., -2.8033e-07, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0411, -0.0394, -0.0139, 0.0014, 0.0043, 0.0110, 0.0013, -0.0081, + 0.0145, -0.0465], device='cuda:0'), grad: tensor([ 2.3283e-08, 5.8673e-08, 4.6566e-08, -1.3970e-08, 9.7696e-07, + 1.1083e-07, -2.3283e-08, -4.1258e-07, -5.0291e-08, -7.0687e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 217.77, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4616 re_mapping 0.0033 re_causal 0.0092 /// teacc 99.23 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.1146, 0.0096, 0.1409, ..., 0.0038, -0.1143, -0.1046], + [-0.2186, -0.2367, -0.2221, ..., 0.1031, 0.0649, 0.4206], + [-0.2459, -0.1734, -0.2329, ..., -0.2928, 0.0188, -0.2582], + ..., + [-0.2065, -0.2732, -0.0078, ..., -0.2622, -0.1331, -0.4347], + [ 0.1487, 0.0312, -0.1613, ..., -0.2556, 0.0408, -0.2384], + [ 0.1503, 0.0980, -0.1528, ..., 0.1202, -0.2590, -0.2969]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 8.3819e-09, 9.3132e-10, 2.7008e-08, ..., 1.0245e-08, + 1.4901e-08, 0.0000e+00], + [ 9.3132e-10, 5.5879e-09, 1.2107e-08, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 5.5879e-09, 7.1712e-08, -5.9605e-08, ..., 2.7940e-09, + -3.1665e-08, 0.0000e+00], + [-1.8626e-09, -7.2643e-08, 1.0245e-08, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00], + [-1.0151e-07, 9.3132e-10, 3.7253e-09, ..., -1.1269e-07, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0408, -0.0394, -0.0136, 0.0011, 0.0037, 0.0110, 0.0012, -0.0089, + 0.0143, -0.0456], device='cuda:0'), grad: tensor([ 8.3819e-09, 2.5425e-07, 1.0710e-07, -9.1270e-07, 4.2841e-07, + 9.1363e-07, 5.5879e-09, -7.8231e-08, -3.1013e-07, -4.2561e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 217.74, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4818 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.24 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.1146, 0.0096, 0.1410, ..., 0.0037, -0.1145, -0.1046], + [-0.2190, -0.2368, -0.2222, ..., 0.1058, 0.0649, 0.4207], + [-0.2462, -0.1735, -0.2333, ..., -0.2931, 0.0188, -0.2582], + ..., + [-0.2066, -0.2734, -0.0069, ..., -0.2652, -0.1334, -0.4358], + [ 0.1489, 0.0314, -0.1615, ..., -0.2560, 0.0409, -0.2386], + [ 0.1505, 0.0975, -0.1527, ..., 0.1213, -0.2608, -0.2976]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-10, 9.3132e-10], + [ 8.8476e-09, 2.3283e-09, 4.6566e-10, ..., 4.6566e-10, + -7.9162e-09, -3.9116e-08], + [ 6.5193e-09, 1.3970e-09, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 8.3819e-09], + ..., + [ 1.0664e-07, 9.3132e-09, 0.0000e+00, ..., 1.5367e-08, + 4.1910e-09, 2.1886e-08], + [ 3.2131e-08, 6.0536e-09, 0.0000e+00, ..., 6.5193e-09, + -1.8626e-09, 4.6566e-09], + [-5.2107e-07, -2.8871e-08, 9.7789e-09, ..., -4.3306e-08, + 7.9162e-09, 9.3132e-10]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0410, -0.0416, -0.0136, 0.0012, 0.0040, 0.0109, 0.0014, -0.0077, + 0.0143, -0.0461], device='cuda:0'), grad: tensor([ 6.5658e-08, -2.0023e-08, -8.2422e-08, 3.1758e-07, 1.1753e-06, + 1.0524e-07, 3.3993e-08, 6.4727e-07, 1.6810e-07, -2.4028e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 217.80, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4662 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.24 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.1149, 0.0096, 0.1409, ..., 0.0034, -0.1152, -0.1049], + [-0.2195, -0.2369, -0.2223, ..., 0.1064, 0.0650, 0.4209], + [-0.2472, -0.1739, -0.2336, ..., -0.2934, 0.0188, -0.2583], + ..., + [-0.2070, -0.2742, -0.0070, ..., -0.2657, -0.1341, -0.4408], + [ 0.1500, 0.0318, -0.1617, ..., -0.2565, 0.0410, -0.2388], + [ 0.1506, 0.0976, -0.1528, ..., 0.1214, -0.2608, -0.3007]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, -1.3970e-09, ..., -9.3132e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., -2.3283e-09, + 2.3283e-09, -1.1176e-08], + [ 0.0000e+00, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + 7.1386e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 1.5832e-08, 6.9849e-09], + [ 4.6566e-10, 3.2596e-09, 0.0000e+00, ..., 4.6566e-10, + -7.4878e-07, 1.8626e-09], + [ 0.0000e+00, 2.3283e-09, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0414, -0.0417, -0.0136, 0.0009, 0.0040, 0.0109, 0.0015, -0.0075, + 0.0144, -0.0461], device='cuda:0'), grad: tensor([ 1.3970e-09, -4.6566e-10, 1.2806e-06, 1.8114e-06, 7.9162e-09, + -1.8626e-06, 6.1467e-08, 5.8208e-08, -1.3635e-06, 1.2573e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 217.78, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4860 re_mapping 0.0030 re_causal 0.0092 /// teacc 99.18 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.1153, 0.0104, 0.1432, ..., 0.0055, -0.1155, -0.1050], + [-0.2198, -0.2370, -0.2224, ..., 0.1077, 0.0651, 0.4211], + [-0.2479, -0.1742, -0.2340, ..., -0.2940, 0.0188, -0.2583], + ..., + [-0.2072, -0.2742, -0.0071, ..., -0.2670, -0.1346, -0.4429], + [ 0.1505, 0.0318, -0.1618, ..., -0.2569, 0.0412, -0.2389], + [ 0.1509, 0.0976, -0.1533, ..., 0.1212, -0.2609, -0.3023]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.3970e-09, 1.0850e-07, ..., -5.5879e-09, + 1.3970e-09, 0.0000e+00], + [ 7.4506e-09, 1.6764e-08, 6.0536e-09, ..., -2.3283e-09, + 6.5193e-09, -6.9849e-09], + [ 2.5611e-08, 2.0489e-08, -6.6962e-07, ..., 4.6566e-10, + 3.1199e-08, 0.0000e+00], + ..., + [ 2.3283e-09, 1.8626e-09, 5.0757e-07, ..., 1.3970e-09, + 2.7940e-09, 2.7940e-09], + [-6.8918e-08, -4.6566e-08, 6.0536e-09, ..., 4.6566e-10, + -8.6613e-08, 4.6566e-10], + [-1.0710e-08, -2.3283e-09, 1.2573e-08, ..., -9.3132e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0403, -0.0417, -0.0135, 0.0005, 0.0039, 0.0110, 0.0017, -0.0076, + 0.0144, -0.0463], device='cuda:0'), grad: tensor([ 3.8836e-07, 6.8452e-08, -2.1514e-06, 3.0175e-07, 6.1467e-08, + -1.7835e-07, 4.4703e-08, 1.6149e-06, -2.3842e-07, 8.6613e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 217.53, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4501 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.16 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.1155, 0.0103, 0.1430, ..., 0.0053, -0.1157, -0.1053], + [-0.2200, -0.2371, -0.2231, ..., 0.1089, 0.0652, 0.4212], + [-0.2483, -0.1749, -0.2337, ..., -0.2949, 0.0187, -0.2583], + ..., + [-0.2072, -0.2739, -0.0054, ..., -0.2679, -0.1354, -0.4450], + [ 0.1511, 0.0322, -0.1620, ..., -0.2574, 0.0415, -0.2391], + [ 0.1516, 0.0977, -0.1530, ..., 0.1217, -0.2610, -0.3044]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 7.9162e-09, 9.3132e-10, ..., 6.9849e-09, + 9.7789e-09, 0.0000e+00], + [ 1.1176e-08, 9.7789e-09, 4.6566e-10, ..., 1.9092e-08, + 9.7789e-09, -5.1223e-09], + [ 1.5832e-08, 1.3970e-08, 4.6566e-10, ..., 1.1642e-08, + 1.6298e-08, 1.8626e-09], + ..., + [ 1.1176e-08, 9.7789e-09, 0.0000e+00, ..., -5.1223e-09, + 1.2107e-08, 3.2596e-09], + [ 1.2387e-07, 1.0710e-07, 1.3970e-09, ..., 8.2888e-08, + 1.2340e-07, 4.6566e-10], + [ 1.7229e-08, 1.4901e-08, 0.0000e+00, ..., 1.1176e-08, + 1.6764e-08, 0.0000e+00]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0407, -0.0417, -0.0136, 0.0002, 0.0036, 0.0111, 0.0017, -0.0075, + 0.0146, -0.0461], device='cuda:0'), grad: tensor([ 3.6787e-08, 2.3283e-07, 7.1246e-08, -1.5423e-06, 2.0023e-08, + 8.5821e-07, -1.5367e-08, -1.9977e-07, 4.4890e-07, 8.8010e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 217.44, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4232 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.19 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.1157, 0.0102, 0.1430, ..., 0.0052, -0.1161, -0.1054], + [-0.2201, -0.2372, -0.2232, ..., 0.1089, 0.0653, 0.4214], + [-0.2489, -0.1752, -0.2349, ..., -0.2955, 0.0186, -0.2584], + ..., + [-0.2074, -0.2733, -0.0052, ..., -0.2679, -0.1361, -0.4467], + [ 0.1507, 0.0322, -0.1629, ..., -0.2588, 0.0414, -0.2391], + [ 0.1519, 0.0979, -0.1534, ..., 0.1218, -0.2611, -0.3047]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.3970e-09, 3.7253e-08, ..., 2.0955e-08, + 8.3819e-09, 0.0000e+00], + [ 4.6566e-10, 1.3970e-09, 9.7789e-09, ..., 2.7940e-09, + 1.3970e-09, -7.9162e-09], + [ 3.7253e-09, 7.4506e-09, 9.3132e-09, ..., 5.1223e-09, + 1.0245e-08, 5.1223e-09], + ..., + [ 9.3132e-10, 2.3283e-09, 1.0710e-08, ..., 9.3132e-10, + 6.0536e-09, 1.8626e-09], + [-8.3819e-09, -1.6298e-08, 1.0803e-07, ..., 1.7649e-07, + -4.1910e-09, 0.0000e+00], + [ 1.1642e-08, 6.1002e-08, 1.7136e-07, ..., 9.3132e-10, + 1.3504e-08, 4.6566e-10]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0409, -0.0417, -0.0138, 0.0002, 0.0038, 0.0108, 0.0022, -0.0075, + 0.0145, -0.0464], device='cuda:0'), grad: tensor([ 1.2387e-07, 3.0734e-08, 8.8476e-08, 1.5367e-08, -8.4657e-07, + 5.7137e-07, -1.1874e-06, -7.4971e-08, 5.5926e-07, 7.4506e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 217.73, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4787 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.24 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.1156, 0.0118, 0.1446, ..., 0.0062, -0.1164, -0.1055], + [-0.2212, -0.2373, -0.2232, ..., 0.1097, 0.0654, 0.4217], + [-0.2497, -0.1754, -0.2357, ..., -0.2962, 0.0185, -0.2585], + ..., + [-0.2078, -0.2739, -0.0053, ..., -0.2689, -0.1372, -0.4501], + [ 0.1506, 0.0319, -0.1633, ..., -0.2596, 0.0414, -0.2392], + [ 0.1521, 0.0977, -0.1541, ..., 0.1218, -0.2612, -0.3053]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, -2.3283e-09], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10], + ..., + [ 3.7253e-09, -1.3970e-09, 0.0000e+00, ..., 2.3283e-09, + 9.3132e-10, 2.3283e-09], + [ 1.3970e-09, 9.3132e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-2.5146e-08, -1.9092e-08, -3.7253e-09, ..., -2.4680e-08, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0394, -0.0417, -0.0139, 0.0002, 0.0039, 0.0107, 0.0027, -0.0075, + 0.0143, -0.0466], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.3039e-08, 1.4435e-08, 6.7055e-08, 9.7789e-08, + -2.6217e-07, 1.5646e-07, -3.4459e-08, 3.0734e-08, -7.8697e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 217.69, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4587 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.15 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.1157, 0.0120, 0.1449, ..., 0.0062, -0.1170, -0.1060], + [-0.2217, -0.2376, -0.2233, ..., 0.1106, 0.0657, 0.4219], + [-0.2505, -0.1759, -0.2372, ..., -0.2993, 0.0183, -0.2585], + ..., + [-0.2074, -0.2719, -0.0054, ..., -0.2695, -0.1387, -0.4525], + [ 0.1504, 0.0306, -0.1636, ..., -0.2602, 0.0414, -0.2396], + [ 0.1522, 0.0977, -0.1542, ..., 0.1221, -0.2614, -0.3059]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 7.9162e-09, -6.9849e-09, ..., 4.1910e-09, + 8.3819e-09, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 1.3970e-09, ..., 4.6566e-10, + 4.6566e-10, -5.1223e-09], + [ 9.3132e-10, 1.3970e-09, 9.3132e-10, ..., 1.3970e-09, + 9.3132e-10, 9.3132e-10], + ..., + [ 2.3283e-09, 2.7940e-09, -4.6566e-10, ..., 4.1910e-09, + 1.3970e-09, 4.1910e-09], + [ 4.1910e-09, 6.5193e-09, 0.0000e+00, ..., 6.0536e-09, + 5.1223e-09, 4.6566e-10], + [ 4.6566e-10, 9.3132e-10, 3.2596e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0394, -0.0416, -0.0141, -0.0010, 0.0039, 0.0119, 0.0020, -0.0075, + 0.0138, -0.0468], device='cuda:0'), grad: tensor([ 1.7695e-08, 1.6298e-08, -1.3970e-08, -1.0943e-07, 1.8626e-09, + 2.8871e-08, 1.0710e-08, 1.2573e-08, 3.1199e-08, 1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 217.69, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4323 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.09 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.1158, 0.0120, 0.1450, ..., 0.0062, -0.1175, -0.1067], + [-0.2234, -0.2378, -0.2233, ..., 0.1121, 0.0661, 0.4224], + [-0.2513, -0.1763, -0.2377, ..., -0.3016, 0.0182, -0.2587], + ..., + [-0.2106, -0.2745, -0.0055, ..., -0.2723, -0.1406, -0.4579], + [ 0.1509, 0.0304, -0.1637, ..., -0.2606, 0.0415, -0.2400], + [ 0.1536, 0.0982, -0.1543, ..., 0.1229, -0.2614, -0.3062]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -3.2596e-09, ..., -4.6566e-09, + -1.8626e-09, -1.6298e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 3.7253e-09, + 1.3970e-09, 1.3039e-08], + [-3.2596e-09, -4.1910e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 4.6566e-10], + [ 1.3970e-09, 2.3283e-09, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 1.3970e-09]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0396, -0.0418, -0.0143, -0.0011, 0.0040, 0.0121, 0.0018, -0.0073, + 0.0136, -0.0464], device='cuda:0'), grad: tensor([ 3.2596e-09, -2.8405e-08, -2.7008e-08, 6.5193e-09, 1.8626e-09, + -3.7253e-09, 8.8476e-09, 2.7474e-08, 9.7789e-09, 7.9162e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 217.56, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4686 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.14 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.1160, 0.0122, 0.1455, ..., 0.0063, -0.1180, -0.1069], + [-0.2257, -0.2380, -0.2231, ..., 0.1125, 0.0662, 0.4226], + [-0.2528, -0.1767, -0.2385, ..., -0.3019, 0.0182, -0.2587], + ..., + [-0.2115, -0.2750, -0.0059, ..., -0.2735, -0.1416, -0.4592], + [ 0.1507, 0.0301, -0.1638, ..., -0.2613, 0.0415, -0.2402], + [ 0.1544, 0.0985, -0.1544, ..., 0.1235, -0.2615, -0.3070]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., -5.2154e-08, + -9.7789e-09, -1.9092e-08], + [ 9.3132e-10, 9.3132e-10, 4.6566e-10, ..., 9.3132e-10, + -4.1444e-08, 1.8626e-09], + ..., + [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 5.3551e-08, + 5.4482e-08, 1.7229e-08], + [-6.0536e-09, -4.1910e-09, 0.0000e+00, ..., 0.0000e+00, + -1.7695e-08, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 2.3283e-09, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0395, -0.0420, -0.0141, -0.0005, 0.0039, 0.0120, 0.0016, -0.0072, + 0.0134, -0.0461], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.0151e-07, -2.0862e-07, 7.4506e-09, -1.0245e-08, + 2.6543e-08, 1.0710e-08, 3.0734e-07, -4.2375e-08, 1.4901e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 217.35, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4500 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.20 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.1159, 0.0132, 0.1472, ..., 0.0071, -0.1184, -0.1072], + [-0.2259, -0.2395, -0.2231, ..., 0.1120, 0.0662, 0.4227], + [-0.2541, -0.1774, -0.2389, ..., -0.3024, 0.0182, -0.2587], + ..., + [-0.2116, -0.2737, -0.0060, ..., -0.2731, -0.1422, -0.4605], + [ 0.1509, 0.0303, -0.1639, ..., -0.2615, 0.0417, -0.2402], + [ 0.1545, 0.0985, -0.1546, ..., 0.1235, -0.2617, -0.3077]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -5.5879e-09, -4.6566e-09, ..., -2.7940e-09, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 3.7253e-09, 2.7940e-09, ..., 2.7940e-09, + 1.8626e-09, -1.8626e-09], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 1.8626e-09, + 2.7940e-09, 9.3132e-10], + ..., + [ 5.5879e-09, 1.1176e-08, 0.0000e+00, ..., 8.3819e-09, + 4.6566e-09, 2.7940e-09], + [ 3.7253e-09, 9.3132e-09, 1.8626e-08, ..., 1.6764e-08, + 1.5832e-08, 0.0000e+00], + [-7.0781e-08, -1.4342e-07, 2.7940e-09, ..., -8.8476e-08, + -5.1223e-08, 0.0000e+00]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0384, -0.0423, -0.0140, -0.0006, 0.0039, 0.0120, 0.0013, -0.0068, + 0.0134, -0.0462], device='cuda:0'), grad: tensor([-7.4506e-09, 1.7695e-08, 1.9558e-08, 4.3772e-08, 2.5053e-07, + 6.9849e-08, -1.5274e-07, -1.2480e-07, 9.7789e-08, -2.1234e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 217.92, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4550 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.19 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.1161, 0.0132, 0.1466, ..., 0.0060, -0.1202, -0.1082], + [-0.2263, -0.2398, -0.2236, ..., 0.1122, 0.0664, 0.4230], + [-0.2551, -0.1783, -0.2410, ..., -0.3035, 0.0180, -0.2588], + ..., + [-0.2118, -0.2740, -0.0060, ..., -0.2736, -0.1430, -0.4622], + [ 0.1512, 0.0307, -0.1639, ..., -0.2624, 0.0421, -0.2405], + [ 0.1550, 0.0989, -0.1549, ..., 0.1237, -0.2618, -0.3091]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 9.3132e-09, 9.3132e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -2.1607e-07, + -2.1979e-07, -2.1420e-07], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 8.6613e-08, + 8.8476e-08, 8.5682e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0524e-07, + 1.0803e-07, 1.0617e-07], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0394, -0.0424, -0.0141, -0.0008, 0.0038, 0.0118, 0.0020, -0.0068, + 0.0136, -0.0462], device='cuda:0'), grad: tensor([ 6.7055e-08, -6.6590e-07, -3.5763e-06, 3.7253e-08, 9.3132e-09, + 2.8964e-07, 1.9558e-08, 3.7104e-06, 8.0094e-08, 2.2352e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 217.67, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4712 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.13 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.1161, 0.0140, 0.1472, ..., 0.0063, -0.1207, -0.1089], + [-0.2273, -0.2401, -0.2237, ..., 0.1126, 0.0665, 0.4233], + [-0.2558, -0.1789, -0.2418, ..., -0.3051, 0.0179, -0.2589], + ..., + [-0.2119, -0.2742, -0.0059, ..., -0.2738, -0.1439, -0.4642], + [ 0.1521, 0.0316, -0.1639, ..., -0.2626, 0.0425, -0.2406], + [ 0.1553, 0.0989, -0.1551, ..., 0.1240, -0.2618, -0.3100]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.9558e-08, -2.4214e-08, 0.0000e+00, ..., -3.2596e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0392, -0.0423, -0.0142, -0.0005, 0.0037, 0.0114, 0.0022, -0.0068, + 0.0138, -0.0462], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.2200e-07, -2.0489e-07, -2.7940e-09, 2.3283e-08, + 5.9605e-08, 2.7940e-09, 3.8184e-08, 1.7695e-08, -5.4948e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 217.68, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4321 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.17 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.1164, 0.0143, 0.1474, ..., 0.0063, -0.1211, -0.1091], + [-0.2276, -0.2403, -0.2237, ..., 0.1138, 0.0670, 0.4238], + [-0.2569, -0.1794, -0.2420, ..., -0.3056, 0.0179, -0.2589], + ..., + [-0.2119, -0.2751, -0.0059, ..., -0.2749, -0.1474, -0.4680], + [ 0.1526, 0.0322, -0.1639, ..., -0.2631, 0.0428, -0.2411], + [ 0.1558, 0.0996, -0.1552, ..., 0.1244, -0.2618, -0.3118]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -5.5879e-09, + -5.5879e-09, -8.3819e-09], + [ 2.7940e-09, 9.3132e-10, -1.3690e-07, ..., 2.7940e-09, + 9.3132e-09, 9.3132e-10], + ..., + [ 9.3132e-10, 1.8626e-09, 9.1270e-08, ..., 3.7253e-09, + 2.7940e-09, 6.5193e-09], + [-9.3132e-10, 9.3132e-10, 7.4506e-09, ..., 9.3132e-10, + -4.6566e-09, 0.0000e+00], + [-7.3574e-08, -1.7043e-07, 9.3132e-10, ..., -1.1176e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0398, -0.0425, -0.0140, -0.0004, 0.0035, 0.0112, 0.0021, -0.0066, + 0.0139, -0.0462], device='cuda:0'), grad: tensor([ 1.6578e-07, -6.5193e-09, -7.0315e-07, -9.3132e-10, 6.7987e-08, + 4.6566e-07, -8.3819e-09, 4.9174e-07, 2.2352e-08, -5.0571e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 217.80, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4465 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.19 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.1173, 0.0131, 0.1470, ..., 0.0055, -0.1237, -0.1119], + [-0.2279, -0.2404, -0.2236, ..., 0.1145, 0.0672, 0.4240], + [-0.2581, -0.1800, -0.2419, ..., -0.3059, 0.0178, -0.2589], + ..., + [-0.2121, -0.2755, -0.0080, ..., -0.2754, -0.1488, -0.4698], + [ 0.1527, 0.0321, -0.1641, ..., -0.2639, 0.0430, -0.2413], + [ 0.1564, 0.1001, -0.1556, ..., 0.1247, -0.2620, -0.3132]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, -1.8626e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 3.7253e-09], + [-1.1455e-07, -1.4342e-07, 0.0000e+00, ..., 0.0000e+00, + -2.4214e-08, 9.3132e-10], + [ 1.0245e-07, 1.2852e-07, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-08, 2.7940e-09]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0409, -0.0424, -0.0140, -0.0004, 0.0036, 0.0113, 0.0019, -0.0067, + 0.0140, -0.0462], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.5832e-08, 2.3283e-08, 1.3039e-08, 3.9116e-08, + 2.7008e-08, 6.5193e-09, -8.8476e-08, -5.3085e-07, 4.9919e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 217.53, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4483 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.14 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.1175, 0.0130, 0.1473, ..., 0.0050, -0.1242, -0.1120], + [-0.2283, -0.2405, -0.2237, ..., 0.1147, 0.0673, 0.4242], + [-0.2586, -0.1792, -0.2429, ..., -0.3061, 0.0176, -0.2590], + ..., + [-0.2126, -0.2768, -0.0099, ..., -0.2759, -0.1497, -0.4707], + [ 0.1532, 0.0324, -0.1642, ..., -0.2642, 0.0437, -0.2414], + [ 0.1566, 0.1002, -0.1557, ..., 0.1249, -0.2622, -0.3140]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 1.8626e-09, ..., -8.3819e-09, + -4.6566e-09, -6.5193e-09], + [ 4.6566e-09, 1.4901e-08, 0.0000e+00, ..., 9.3132e-10, + 5.5879e-09, 1.8626e-09], + ..., + [ 1.3970e-08, 4.2841e-08, 1.8626e-09, ..., 4.6566e-09, + 2.0489e-08, 4.6566e-09], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.7300e-08, ..., 0.0000e+00, + 3.7253e-08, 9.3132e-10]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0413, -0.0427, -0.0142, -0.0003, 0.0038, 0.0112, 0.0019, -0.0064, + 0.0142, -0.0463], device='cuda:0'), grad: tensor([ 5.5879e-09, -1.3970e-08, 3.6322e-08, -1.7229e-07, -5.1130e-07, + 3.9116e-08, 1.0245e-08, 1.2852e-07, 4.6566e-09, 4.7963e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 218.09, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4631 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.07 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.1175, 0.0134, 0.1476, ..., 0.0050, -0.1245, -0.1125], + [-0.2284, -0.2406, -0.2238, ..., 0.1185, 0.0675, 0.4245], + [-0.2587, -0.1786, -0.2433, ..., -0.3067, 0.0177, -0.2590], + ..., + [-0.2127, -0.2770, -0.0105, ..., -0.2794, -0.1506, -0.4726], + [ 0.1534, 0.0325, -0.1642, ..., -0.2650, 0.0437, -0.2420], + [ 0.1566, 0.1001, -0.1561, ..., 0.1247, -0.2623, -0.3155]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + ..., + [ 1.0245e-08, 7.4506e-09, 0.0000e+00, ..., 6.5193e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [-2.6077e-08, -1.8626e-08, 0.0000e+00, ..., -2.1420e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0415, -0.0417, -0.0136, -0.0010, 0.0041, 0.0111, 0.0020, -0.0074, + 0.0140, -0.0467], device='cuda:0'), grad: tensor([ 1.8626e-09, 8.3819e-08, 1.3970e-08, 1.3039e-08, 5.4948e-08, + 2.7940e-09, -1.8626e-09, -4.0978e-08, -4.0047e-08, -8.1025e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 217.99, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4506 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.14 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.1177, 0.0136, 0.1477, ..., 0.0050, -0.1246, -0.1125], + [-0.2287, -0.2407, -0.2239, ..., 0.1189, 0.0676, 0.4247], + [-0.2594, -0.1784, -0.2438, ..., -0.3071, 0.0178, -0.2591], + ..., + [-0.2131, -0.2776, -0.0107, ..., -0.2799, -0.1514, -0.4731], + [ 0.1534, 0.0318, -0.1647, ..., -0.2668, 0.0437, -0.2421], + [ 0.1570, 0.1004, -0.1562, ..., 0.1251, -0.2624, -0.3157]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.7695e-08, -2.7940e-08, ..., -1.9558e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -3.3528e-08, 2.7940e-09, ..., 1.8626e-09, + -4.6566e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 4.6566e-09, 2.7940e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 1.1176e-08, 6.5193e-09, ..., 6.5193e-09, + 0.0000e+00, 0.0000e+00], + [-3.1665e-08, -1.3970e-08, 6.5193e-09, ..., -1.3970e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0416, -0.0416, -0.0126, -0.0020, 0.0041, 0.0113, 0.0027, -0.0077, + 0.0136, -0.0465], device='cuda:0'), grad: tensor([-8.2888e-08, 9.3132e-09, -1.5181e-07, 1.4342e-07, 6.7987e-08, + 1.0245e-08, 2.0489e-08, 2.3283e-08, 5.3085e-08, -7.3574e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 217.60, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4335 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.11 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.1178, 0.0138, 0.1476, ..., 0.0049, -0.1248, -0.1126], + [-0.2291, -0.2409, -0.2240, ..., 0.1189, 0.0676, 0.4248], + [-0.2601, -0.1789, -0.2442, ..., -0.3073, 0.0178, -0.2591], + ..., + [-0.2132, -0.2777, -0.0108, ..., -0.2800, -0.1514, -0.4736], + [ 0.1539, 0.0320, -0.1650, ..., -0.2676, 0.0438, -0.2422], + [ 0.1571, 0.1004, -0.1563, ..., 0.1252, -0.2624, -0.3159]], + device='cuda:0'), grad: tensor([[1.8626e-09, 2.7940e-08, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, + 0.0000e+00], + [6.5193e-09, 1.0245e-08, 0.0000e+00, ..., 4.6566e-09, 1.0245e-08, + 0.0000e+00], + [9.3132e-09, 1.4901e-08, 0.0000e+00, ..., 1.8626e-09, 1.0245e-08, + 0.0000e+00], + ..., + [5.5879e-09, 1.5832e-08, 0.0000e+00, ..., 2.7940e-09, 4.6566e-09, + 0.0000e+00], + [2.7940e-09, 4.7497e-08, 0.0000e+00, ..., 1.8626e-09, 3.7253e-09, + 0.0000e+00], + [8.3819e-09, 2.1420e-08, 3.2596e-08, ..., 9.3132e-10, 6.0536e-08, + 0.0000e+00]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0419, -0.0416, -0.0128, -0.0022, 0.0040, 0.0111, 0.0033, -0.0077, + 0.0136, -0.0466], device='cuda:0'), grad: tensor([ 1.3132e-07, 5.7742e-08, 2.7008e-08, -7.4506e-09, -1.8626e-07, + -8.9593e-07, 3.5204e-07, 7.4506e-08, 1.9465e-07, 2.5611e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 217.67, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4662 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.13 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.1180, 0.0137, 0.1479, ..., 0.0048, -0.1251, -0.1122], + [-0.2296, -0.2411, -0.2242, ..., 0.1190, 0.0676, 0.4249], + [-0.2615, -0.1797, -0.2443, ..., -0.3076, 0.0177, -0.2591], + ..., + [-0.2135, -0.2784, -0.0109, ..., -0.2801, -0.1518, -0.4745], + [ 0.1541, 0.0320, -0.1656, ..., -0.2691, 0.0440, -0.2424], + [ 0.1573, 0.0995, -0.1571, ..., 0.1253, -0.2626, -0.3162]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + -6.5193e-09, -1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 3.7253e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 3.7253e-09, 5.5879e-09], + [-4.6566e-09, -3.7253e-09, 9.3132e-10, ..., -6.5193e-09, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0422, -0.0416, -0.0127, -0.0035, 0.0044, 0.0118, 0.0039, -0.0077, + 0.0134, -0.0470], device='cuda:0'), grad: tensor([ 9.3132e-10, -1.8626e-09, -2.7940e-09, 4.6566e-09, 2.3283e-08, + -3.7253e-09, 2.7940e-09, -1.7695e-08, 1.8626e-08, -1.3039e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 217.84, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4789 re_mapping 0.0028 re_causal 0.0085 /// teacc 99.15 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.1178, 0.0151, 0.1494, ..., 0.0063, -0.1255, -0.1125], + [-0.2307, -0.2414, -0.2243, ..., 0.1190, 0.0676, 0.4251], + [-0.2635, -0.1806, -0.2450, ..., -0.3082, 0.0177, -0.2592], + ..., + [-0.2138, -0.2793, -0.0108, ..., -0.2801, -0.1524, -0.4751], + [ 0.1572, 0.0349, -0.1662, ..., -0.2730, 0.0457, -0.2429], + [ 0.1574, 0.0993, -0.1573, ..., 0.1253, -0.2628, -0.3167]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.5832e-08, -3.3528e-08, ..., -1.9558e-08, + 1.8626e-09, 0.0000e+00], + [ 6.5193e-09, 1.5832e-08, 4.6566e-09, ..., 2.7940e-09, + 2.7940e-09, -4.6566e-09], + [ 1.0245e-08, 2.4214e-08, 4.6566e-09, ..., 5.5879e-09, + 7.4506e-09, 9.3132e-10], + ..., + [ 3.7253e-09, 7.4506e-09, 9.3132e-10, ..., 2.7940e-09, + 2.7940e-09, 1.8626e-09], + [ 2.7940e-09, 7.4506e-09, 1.8626e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 2.0489e-08, 4.5635e-08, 7.4506e-09, ..., 6.5193e-09, + 1.5832e-08, 9.3132e-10]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0412, -0.0416, -0.0127, -0.0027, 0.0045, 0.0106, 0.0046, -0.0077, + 0.0135, -0.0472], device='cuda:0'), grad: tensor([-5.1223e-08, 3.3341e-07, -4.6566e-09, -3.1702e-06, 3.1665e-08, + 2.9318e-06, 2.8871e-08, -3.0361e-07, 4.4703e-08, 1.6484e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 217.34, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4275 re_mapping 0.0030 re_causal 0.0083 /// teacc 99.17 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.1178, 0.0162, 0.1505, ..., 0.0068, -0.1261, -0.1127], + [-0.2336, -0.2438, -0.2244, ..., 0.1189, 0.0674, 0.4255], + [-0.2659, -0.1818, -0.2454, ..., -0.3091, 0.0178, -0.2593], + ..., + [-0.2143, -0.2803, -0.0109, ..., -0.2803, -0.1543, -0.4767], + [ 0.1577, 0.0353, -0.1665, ..., -0.2730, 0.0463, -0.2431], + [ 0.1574, 0.0990, -0.1576, ..., 0.1252, -0.2632, -0.3173]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -1.8626e-09, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0403, -0.0417, -0.0127, 0.0013, 0.0045, 0.0091, 0.0042, -0.0077, + 0.0139, -0.0475], device='cuda:0'), grad: tensor([ 2.7940e-09, -4.6566e-09, 4.6566e-09, 5.5879e-09, -9.3132e-09, + -6.5193e-09, 0.0000e+00, 7.4506e-09, 2.7940e-09, 8.3819e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 217.58, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4443 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.18 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.1180, 0.0157, 0.1505, ..., 0.0061, -0.1270, -0.1129], + [-0.2343, -0.2442, -0.2245, ..., 0.1189, 0.0675, 0.4258], + [-0.2666, -0.1822, -0.2461, ..., -0.3102, 0.0177, -0.2594], + ..., + [-0.2146, -0.2807, -0.0110, ..., -0.2803, -0.1549, -0.4775], + [ 0.1579, 0.0354, -0.1665, ..., -0.2731, 0.0463, -0.2441], + [ 0.1565, 0.0979, -0.1595, ..., 0.1249, -0.2649, -0.3178]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., -9.3132e-10, + -3.7253e-09, -1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09], + ..., + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 1.8626e-09], + [ 4.6566e-09, 5.5879e-09, 0.0000e+00, ..., 5.5879e-09, + 9.3132e-10, 5.5879e-09], + [-1.3411e-07, -1.6112e-07, 0.0000e+00, ..., -1.3784e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0409, -0.0417, -0.0128, 0.0015, 0.0053, 0.0094, 0.0041, -0.0077, + 0.0140, -0.0483], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.4901e-08, 4.6566e-09, 2.6543e-07, 1.0245e-08, + 1.1921e-07, 4.6566e-09, 1.1176e-08, 2.2352e-08, -4.2003e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 217.53, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4895 re_mapping 0.0027 re_causal 0.0085 /// teacc 99.07 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.1198, 0.0167, 0.1511, ..., 0.0059, -0.1281, -0.1135], + [-0.2348, -0.2445, -0.2247, ..., 0.1192, 0.0679, 0.4266], + [-0.2703, -0.1842, -0.2474, ..., -0.3116, 0.0171, -0.2596], + ..., + [-0.2153, -0.2812, -0.0112, ..., -0.2807, -0.1563, -0.4811], + [ 0.1582, 0.0354, -0.1666, ..., -0.2733, 0.0470, -0.2452], + [ 0.1578, 0.0986, -0.1600, ..., 0.1259, -0.2654, -0.3186]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 9.3132e-10, -3.7253e-09, ..., 4.6566e-09, + 1.8626e-09, 0.0000e+00], + [ 5.1223e-08, 1.8626e-09, 6.5193e-09, ..., 2.7940e-09, + 7.5437e-08, 0.0000e+00], + [ 5.5879e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [-1.8440e-07, -3.0734e-08, -1.9558e-08, ..., 9.3132e-10, + -2.4308e-07, 9.3132e-10], + [-5.5879e-09, -4.6566e-09, 9.3132e-10, ..., -2.1420e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0409, -0.0414, -0.0132, 0.0003, 0.0055, 0.0099, 0.0040, -0.0080, + 0.0141, -0.0480], device='cuda:0'), grad: tensor([ 1.2107e-08, 3.1106e-07, 3.6322e-08, 2.8871e-08, 2.9802e-08, + 2.4121e-07, 4.0792e-07, 1.1176e-08, -1.0403e-06, -3.2596e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 217.68, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4444 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.1207, 0.0167, 0.1516, ..., 0.0063, -0.1297, -0.1136], + [-0.2363, -0.2450, -0.2248, ..., 0.1193, 0.0680, 0.4269], + [-0.2728, -0.1869, -0.2487, ..., -0.3129, 0.0168, -0.2597], + ..., + [-0.2165, -0.2822, -0.0113, ..., -0.2808, -0.1562, -0.4815], + [ 0.1585, 0.0356, -0.1667, ..., -0.2736, 0.0476, -0.2457], + [ 0.1603, 0.1003, -0.1601, ..., 0.1274, -0.2655, -0.3189]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.7940e-09, 3.7253e-09, ..., 9.3132e-10, + 5.5879e-09, 0.0000e+00], + [ 2.7940e-09, 3.7253e-09, 4.6566e-09, ..., 1.8626e-09, + 6.5193e-09, 0.0000e+00], + [ 3.7253e-09, 5.5879e-09, 5.5879e-09, ..., 2.7940e-09, + 8.3819e-09, 9.3132e-10], + ..., + [ 2.3283e-08, 3.2596e-08, 3.5390e-08, ..., 1.9558e-08, + 5.9605e-08, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [-1.0245e-08, 3.1665e-08, 7.5437e-08, ..., -1.4901e-08, + 1.2387e-07, 0.0000e+00]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0412, -0.0414, -0.0133, -0.0002, 0.0049, 0.0101, 0.0038, -0.0080, + 0.0142, -0.0469], device='cuda:0'), grad: tensor([ 1.5832e-08, 2.2352e-08, 2.5146e-08, -2.7940e-09, -4.5449e-07, + -1.1176e-08, 1.2107e-08, 1.8440e-07, 3.7253e-09, 2.1141e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 217.51, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4369 re_mapping 0.0032 re_causal 0.0087 /// teacc 99.28 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.1209, 0.0168, 0.1517, ..., 0.0062, -0.1301, -0.1139], + [-0.2365, -0.2453, -0.2248, ..., 0.1194, 0.0682, 0.4273], + [-0.2731, -0.1871, -0.2490, ..., -0.3138, 0.0166, -0.2599], + ..., + [-0.2165, -0.2822, -0.0113, ..., -0.2809, -0.1562, -0.4819], + [ 0.1585, 0.0355, -0.1668, ..., -0.2738, 0.0476, -0.2475], + [ 0.1637, 0.1032, -0.1583, ..., 0.1306, -0.2650, -0.3191]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.2107e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0424, -0.0414, -0.0131, -0.0004, 0.0020, 0.0104, 0.0038, -0.0080, + 0.0141, -0.0441], device='cuda:0'), grad: tensor([ 1.1176e-08, 6.7055e-08, 4.6566e-09, 5.6997e-07, 4.0047e-07, + -1.4529e-06, 2.9802e-08, 2.7753e-07, 9.6858e-08, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 217.67, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4639 re_mapping 0.0033 re_causal 0.0094 /// teacc 99.21 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.1211, 0.0171, 0.1519, ..., 0.0062, -0.1309, -0.1143], + [-0.2367, -0.2454, -0.2250, ..., 0.1194, 0.0677, 0.4275], + [-0.2735, -0.1872, -0.2495, ..., -0.3141, 0.0168, -0.2600], + ..., + [-0.2161, -0.2817, -0.0114, ..., -0.2809, -0.1553, -0.4824], + [ 0.1586, 0.0355, -0.1670, ..., -0.2739, 0.0478, -0.2478], + [ 0.1637, 0.1032, -0.1585, ..., 0.1307, -0.2652, -0.3196]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-9.3132e-10, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0432, -0.0415, -0.0125, -0.0018, 0.0020, 0.0111, 0.0040, -0.0079, + 0.0141, -0.0441], device='cuda:0'), grad: tensor([ 0.0000e+00, 2.7940e-08, -2.3283e-08, 7.4506e-09, -2.7940e-09, + -1.8626e-08, 9.3132e-09, 5.5879e-09, -5.5879e-09, 2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 217.60, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4626 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.15 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.1212, 0.0177, 0.1525, ..., 0.0065, -0.1311, -0.1142], + [-0.2369, -0.2455, -0.2254, ..., 0.1195, 0.0674, 0.4278], + [-0.2739, -0.1875, -0.2527, ..., -0.3148, 0.0165, -0.2600], + ..., + [-0.2164, -0.2821, -0.0114, ..., -0.2810, -0.1544, -0.4839], + [ 0.1585, 0.0355, -0.1676, ..., -0.2741, 0.0477, -0.2483], + [ 0.1639, 0.1033, -0.1585, ..., 0.1307, -0.2652, -0.3204]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 9.3132e-10], + [ 1.6764e-08, 5.5879e-09, 0.0000e+00, ..., 6.5193e-09, + -1.1176e-08, -4.9360e-08], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-09, + 4.6566e-09, 1.8626e-09], + ..., + [ 1.8626e-08, 8.3819e-09, 0.0000e+00, ..., 2.5146e-08, + 8.3819e-09, 1.8626e-08], + [-7.4506e-09, -1.0245e-08, 0.0000e+00, ..., 1.1176e-08, + -7.4506e-09, 0.0000e+00], + [-1.0524e-07, -2.4214e-08, 0.0000e+00, ..., -1.1735e-07, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0428, -0.0416, -0.0125, -0.0019, 0.0019, 0.0110, 0.0043, -0.0079, + 0.0140, -0.0441], device='cuda:0'), grad: tensor([ 2.4214e-08, 1.3039e-08, 4.0978e-08, -5.6811e-08, 3.1292e-07, + 7.5065e-07, -8.1304e-07, 1.2200e-07, 2.7940e-08, -4.2655e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 217.79, cls_loss 0.0004 cls_loss_mapping 0.0008 cls_loss_causal 0.4562 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.12 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.1230, 0.0174, 0.1528, ..., 0.0057, -0.1312, -0.1143], + [-0.2371, -0.2456, -0.2254, ..., 0.1195, 0.0675, 0.4279], + [-0.2741, -0.1878, -0.2531, ..., -0.3151, 0.0164, -0.2600], + ..., + [-0.2165, -0.2824, -0.0115, ..., -0.2810, -0.1549, -0.4844], + [ 0.1587, 0.0355, -0.1676, ..., -0.2741, 0.0482, -0.2483], + [ 0.1640, 0.1035, -0.1585, ..., 0.1309, -0.2653, -0.3205]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, + 9.3132e-10], + [0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 3.7253e-09, + 0.0000e+00]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0433, -0.0415, -0.0125, -0.0020, 0.0019, 0.0109, 0.0042, -0.0079, + 0.0141, -0.0440], device='cuda:0'), grad: tensor([ 2.7940e-09, 1.6764e-08, -6.6124e-08, 7.8231e-08, -1.2107e-08, + -7.4506e-08, 5.5879e-09, 1.5832e-08, 2.6077e-08, 1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 217.75, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4756 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.10 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.1231, 0.0178, 0.1526, ..., 0.0062, -0.1311, -0.1142], + [-0.2373, -0.2458, -0.2257, ..., 0.1195, 0.0675, 0.4282], + [-0.2744, -0.1881, -0.2535, ..., -0.3157, 0.0163, -0.2601], + ..., + [-0.2167, -0.2826, -0.0117, ..., -0.2810, -0.1548, -0.4850], + [ 0.1584, 0.0354, -0.1677, ..., -0.2743, 0.0483, -0.2486], + [ 0.1643, 0.1036, -0.1587, ..., 0.1310, -0.2654, -0.3208]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, -2.7940e-09, ..., -2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., -8.3819e-09, + 0.0000e+00, -1.7695e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 7.4506e-09], + [-6.5193e-09, -6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 4.6566e-09]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0432, -0.0412, -0.0125, -0.0025, 0.0020, 0.0113, 0.0042, -0.0083, + 0.0140, -0.0440], device='cuda:0'), grad: tensor([-6.5193e-09, 2.1048e-07, -1.8161e-07, 9.3132e-09, 3.7253e-09, + 1.3970e-08, 1.6764e-08, -4.0792e-07, -2.7008e-08, 3.6508e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 217.55, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4610 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.20 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.1233, 0.0180, 0.1530, ..., 0.0064, -0.1318, -0.1144], + [-0.2383, -0.2462, -0.2260, ..., 0.1195, 0.0685, 0.4289], + [-0.2754, -0.1888, -0.2540, ..., -0.3158, 0.0162, -0.2602], + ..., + [-0.2161, -0.2826, -0.0119, ..., -0.2810, -0.1574, -0.4874], + [ 0.1621, 0.0371, -0.1678, ..., -0.2731, 0.0491, -0.2487], + [ 0.1630, 0.1025, -0.1589, ..., 0.1307, -0.2661, -0.3241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -4.6566e-09, ..., -4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 4.6566e-10, 9.3132e-10], + [-2.7940e-09, -9.3132e-10, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0432, -0.0411, -0.0123, -0.0033, 0.0019, 0.0113, 0.0041, -0.0084, + 0.0156, -0.0449], device='cuda:0'), grad: tensor([-1.3970e-08, 7.4506e-09, 2.7940e-09, 1.8626e-09, 1.4435e-08, + 4.1910e-09, -6.0536e-09, -2.3283e-08, 5.1223e-09, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 217.68, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4149 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.1235, 0.0184, 0.1536, ..., 0.0066, -0.1322, -0.1146], + [-0.2385, -0.2463, -0.2260, ..., 0.1200, 0.0686, 0.4293], + [-0.2757, -0.1890, -0.2534, ..., -0.3163, 0.0170, -0.2603], + ..., + [-0.2162, -0.2827, -0.0117, ..., -0.2813, -0.1590, -0.4886], + [ 0.1623, 0.0371, -0.1683, ..., -0.2735, 0.0491, -0.2493], + [ 0.1631, 0.1025, -0.1592, ..., 0.1307, -0.2662, -0.3251]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.5367e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 4.6566e-10], + [-6.5193e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., -2.7940e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0428, -0.0410, -0.0115, -0.0034, 0.0019, 0.0111, 0.0042, -0.0086, + 0.0156, -0.0450], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.9116e-08, -8.1491e-08, 3.6787e-08, 3.2596e-09, + -3.5856e-08, 7.9162e-09, 1.6764e-08, 6.9849e-09, 4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 217.88, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4528 re_mapping 0.0029 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.1235, 0.0186, 0.1535, ..., 0.0064, -0.1331, -0.1147], + [-0.2388, -0.2464, -0.2264, ..., 0.1201, 0.0686, 0.4296], + [-0.2759, -0.1892, -0.2535, ..., -0.3180, 0.0169, -0.2603], + ..., + [-0.2164, -0.2831, -0.0120, ..., -0.2816, -0.1592, -0.4901], + [ 0.1625, 0.0373, -0.1685, ..., -0.2736, 0.0493, -0.2497], + [ 0.1631, 0.1025, -0.1594, ..., 0.1308, -0.2666, -0.3260]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-09, + 3.7253e-09, 0.0000e+00], + [ 9.3132e-10, 2.3283e-09, 1.8626e-09, ..., 5.1223e-09, + 5.1223e-09, 4.6566e-10], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00], + [-4.1910e-09, 5.1223e-09, 3.2596e-09, ..., 2.3283e-09, + -1.8626e-09, -2.7940e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0429, -0.0409, -0.0113, -0.0036, 0.0019, 0.0110, 0.0046, -0.0086, + 0.0156, -0.0451], device='cuda:0'), grad: tensor([ 2.0023e-08, 3.1665e-08, 4.6566e-10, 6.9849e-09, 5.1223e-09, + -3.3993e-08, -2.1886e-08, 6.5193e-09, -6.0536e-09, 4.1910e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 217.41, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4625 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.11 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.1238, 0.0186, 0.1540, ..., 0.0066, -0.1336, -0.1148], + [-0.2399, -0.2467, -0.2268, ..., 0.1201, 0.0685, 0.4297], + [-0.2768, -0.1898, -0.2542, ..., -0.3187, 0.0171, -0.2604], + ..., + [-0.2164, -0.2838, -0.0120, ..., -0.2816, -0.1594, -0.4904], + [ 0.1627, 0.0372, -0.1683, ..., -0.2738, 0.0496, -0.2499], + [ 0.1634, 0.1026, -0.1596, ..., 0.1310, -0.2668, -0.3261]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 3.7253e-09, 2.7940e-09, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-5.1223e-09, -2.3283e-09, 0.0000e+00, ..., -5.1223e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0430, -0.0411, -0.0112, -0.0037, 0.0019, 0.0114, 0.0044, -0.0085, + 0.0157, -0.0450], device='cuda:0'), grad: tensor([ 4.6566e-10, 4.6566e-09, 6.5193e-09, -9.7789e-09, 6.0536e-09, + 1.3039e-08, -8.8476e-09, 9.7789e-09, 3.2596e-09, -1.3039e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 217.64, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4313 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.17 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.1240, 0.0187, 0.1539, ..., 0.0065, -0.1342, -0.1147], + [-0.2407, -0.2470, -0.2270, ..., 0.1199, 0.0684, 0.4297], + [-0.2775, -0.1903, -0.2546, ..., -0.3189, 0.0171, -0.2604], + ..., + [-0.2176, -0.2843, -0.0118, ..., -0.2816, -0.1594, -0.4905], + [ 0.1624, 0.0373, -0.1684, ..., -0.2741, 0.0501, -0.2501], + [ 0.1641, 0.1029, -0.1598, ..., 0.1314, -0.2670, -0.3262]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -5.5879e-09, -5.1223e-09, ..., -5.1223e-09, + -1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 5.1223e-09, ..., 0.0000e+00, + 4.6566e-10, -1.8626e-09], + [ 0.0000e+00, 9.3132e-10, 1.1642e-08, ..., 1.3970e-09, + 3.2596e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, -4.6566e-09, ..., 1.3970e-09, + 1.3970e-09, 2.3283e-09], + [ 4.1910e-09, 4.6566e-09, 1.3970e-09, ..., 4.6566e-09, + 1.8626e-09, 4.6566e-10], + [ 8.3819e-09, 8.8476e-09, 7.4506e-09, ..., 6.5193e-09, + 6.0536e-09, 0.0000e+00]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0434, -0.0416, -0.0113, -0.0035, 0.0018, 0.0108, 0.0045, -0.0080, + 0.0156, -0.0447], device='cuda:0'), grad: tensor([-2.7940e-08, 4.7963e-08, 5.6345e-08, -3.0268e-08, -3.8138e-07, + 2.1420e-08, 2.9849e-07, -4.8429e-08, 2.0489e-08, 5.6811e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 217.95, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4601 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.24 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.1241, 0.0191, 0.1544, ..., 0.0069, -0.1346, -0.1147], + [-0.2413, -0.2473, -0.2271, ..., 0.1199, 0.0689, 0.4301], + [-0.2780, -0.1906, -0.2551, ..., -0.3194, 0.0172, -0.2604], + ..., + [-0.2179, -0.2847, -0.0119, ..., -0.2817, -0.1605, -0.4920], + [ 0.1624, 0.0373, -0.1687, ..., -0.2744, 0.0502, -0.2503], + [ 0.1643, 0.1029, -0.1599, ..., 0.1315, -0.2671, -0.3265]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 3.2596e-09, 0.0000e+00, ..., 3.7253e-09, + 8.8476e-09, 0.0000e+00], + [ 1.2107e-08, 4.1910e-09, 4.6566e-10, ..., 2.7940e-09, + 7.4506e-09, 0.0000e+00], + ..., + [ 7.9162e-09, 1.8626e-09, 0.0000e+00, ..., 7.9162e-09, + 0.0000e+00, 4.6566e-10], + [-1.4435e-08, -1.2107e-08, 0.0000e+00, ..., 4.6566e-10, + -2.7940e-08, 0.0000e+00], + [-6.2399e-08, -1.3970e-09, 0.0000e+00, ..., -2.7008e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0430, -0.0416, -0.0113, -0.0036, 0.0018, 0.0106, 0.0047, -0.0081, + 0.0156, -0.0447], device='cuda:0'), grad: tensor([ 2.1886e-08, 3.9116e-08, 6.1467e-08, 2.2817e-08, 1.5367e-07, + 2.4680e-08, 1.3504e-08, 1.8626e-08, -1.1595e-07, -2.2817e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 217.83, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4357 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.19 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.1243, 0.0195, 0.1550, ..., 0.0072, -0.1350, -0.1150], + [-0.2425, -0.2480, -0.2272, ..., 0.1199, 0.0689, 0.4306], + [-0.2798, -0.1903, -0.2556, ..., -0.3203, 0.0169, -0.2607], + ..., + [-0.2179, -0.2849, -0.0120, ..., -0.2817, -0.1607, -0.4928], + [ 0.1629, 0.0375, -0.1688, ..., -0.2744, 0.0509, -0.2501], + [ 0.1646, 0.1032, -0.1601, ..., 0.1318, -0.2673, -0.3285]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -4.6566e-10, -2.7940e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-1.3970e-09, -4.6566e-10, 3.7253e-09, ..., -9.3132e-10, + 5.1223e-09, 4.6566e-10]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0429, -0.0416, -0.0113, -0.0037, 0.0015, 0.0101, 0.0049, -0.0080, + 0.0159, -0.0446], device='cuda:0'), grad: tensor([ 0.0000e+00, 3.7253e-09, 3.2596e-09, 9.7789e-09, -1.7695e-08, + -6.9849e-09, 9.3132e-10, -3.2596e-09, 4.6566e-10, 2.0955e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 217.71, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4912 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.21 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.1243, 0.0197, 0.1559, ..., 0.0080, -0.1346, -0.1138], + [-0.2433, -0.2481, -0.2279, ..., 0.1199, 0.0689, 0.4308], + [-0.2800, -0.1906, -0.2558, ..., -0.3209, 0.0169, -0.2609], + ..., + [-0.2179, -0.2854, -0.0121, ..., -0.2817, -0.1607, -0.4931], + [ 0.1629, 0.0375, -0.1690, ..., -0.2746, 0.0511, -0.2503], + [ 0.1648, 0.1033, -0.1602, ..., 0.1319, -0.2675, -0.3294]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.7789e-09, -1.4435e-08, ..., -1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 1.3970e-09, 1.3504e-08, ..., -1.3504e-08, + 1.6298e-08, -3.2596e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.0710e-08, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.6566e-09, 2.3283e-09, 1.3970e-09, ..., 2.5611e-08, + 9.7789e-09, 3.1199e-08], + [-9.3132e-10, -9.3132e-10, 1.4435e-08, ..., 5.6811e-08, + 1.3970e-09, 1.8626e-09], + [-1.9092e-08, -1.3039e-08, 2.7940e-09, ..., -1.7229e-08, + 3.7253e-09, 2.3283e-09]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0421, -0.0418, -0.0114, -0.0041, 0.0015, 0.0099, 0.0052, -0.0078, + 0.0159, -0.0446], device='cuda:0'), grad: tensor([-4.2375e-08, 3.4552e-07, 4.6100e-08, 1.5832e-08, -5.3085e-08, + 4.8429e-08, -3.4319e-07, -2.3982e-07, 2.6124e-07, -3.3528e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 217.54, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4739 re_mapping 0.0030 re_causal 0.0086 /// teacc 99.18 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.1248, 0.0202, 0.1568, ..., 0.0090, -0.1347, -0.1139], + [-0.2446, -0.2486, -0.2280, ..., 0.1199, 0.0690, 0.4310], + [-0.2807, -0.1911, -0.2565, ..., -0.3212, 0.0168, -0.2609], + ..., + [-0.2182, -0.2857, -0.0116, ..., -0.2818, -0.1607, -0.4934], + [ 0.1629, 0.0375, -0.1692, ..., -0.2752, 0.0512, -0.2511], + [ 0.1658, 0.1044, -0.1605, ..., 0.1325, -0.2677, -0.3301]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, -2.3283e-09, ..., -2.3283e-09, + 9.3132e-10, 0.0000e+00], + [ 6.0536e-09, 7.9162e-09, 3.2596e-09, ..., 2.7940e-09, + 1.3039e-08, -1.5367e-08], + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 3.2596e-09, + -1.3504e-08, 9.3132e-10], + ..., + [-8.3819e-09, 0.0000e+00, -1.9092e-08, ..., 9.3132e-10, + 1.4901e-08, 1.7695e-08], + [-1.5832e-08, -2.7940e-08, 1.8161e-08, ..., 5.5879e-09, + -5.0757e-08, -6.9849e-09], + [ 9.3132e-10, 9.3132e-10, 6.9849e-09, ..., 3.2596e-09, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0416, -0.0418, -0.0102, -0.0056, 0.0008, 0.0096, 0.0053, -0.0078, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([-8.8476e-09, 5.0757e-08, -3.7719e-08, 1.4901e-08, -3.7253e-09, + 1.1362e-07, -4.5635e-08, 1.3970e-09, -1.1595e-07, 3.2596e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 217.46, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4089 re_mapping 0.0031 re_causal 0.0083 /// teacc 99.14 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.1250, 0.0205, 0.1578, ..., 0.0095, -0.1343, -0.1133], + [-0.2455, -0.2491, -0.2288, ..., 0.1199, 0.0688, 0.4312], + [-0.2810, -0.1914, -0.2580, ..., -0.3216, 0.0168, -0.2609], + ..., + [-0.2183, -0.2859, -0.0118, ..., -0.2819, -0.1606, -0.4941], + [ 0.1631, 0.0375, -0.1697, ..., -0.2754, 0.0512, -0.2510], + [ 0.1660, 0.1044, -0.1608, ..., 0.1327, -0.2679, -0.3309]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 3.2596e-09, 4.6566e-10, ..., -9.3132e-10, + -9.3132e-10, -4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 9.3132e-10, 1.8626e-09, 4.6566e-10, ..., 9.3132e-10, + 9.3132e-10, 2.7940e-09], + [ 1.3970e-09, 2.7940e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + [ 2.3283e-09, 8.8476e-09, 1.8626e-09, ..., -1.3970e-09, + 4.6566e-10, 9.3132e-10]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0410, -0.0420, -0.0096, -0.0064, 0.0009, 0.0099, 0.0052, -0.0077, + 0.0156, -0.0440], device='cuda:0'), grad: tensor([ 4.6566e-10, 2.3283e-09, 4.6566e-10, 1.4156e-07, -3.7253e-09, + -1.8021e-07, 3.2596e-09, 9.7789e-09, 1.0710e-08, 2.1886e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 217.26, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4478 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.26 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.1252, 0.0211, 0.1590, ..., 0.0103, -0.1338, -0.1125], + [-0.2498, -0.2503, -0.2296, ..., 0.1196, 0.0688, 0.4315], + [-0.2823, -0.1923, -0.2589, ..., -0.3222, 0.0166, -0.2611], + ..., + [-0.2205, -0.2897, -0.0147, ..., -0.2822, -0.1609, -0.4953], + [ 0.1633, 0.0377, -0.1700, ..., -0.2756, 0.0517, -0.2519], + [ 0.1673, 0.1046, -0.1619, ..., 0.1340, -0.2681, -0.3311]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -5.1223e-09, -1.1642e-08, ..., -4.1910e-09, + 1.3970e-09, 0.0000e+00], + [ 2.8405e-08, 3.0268e-08, 1.3970e-09, ..., 3.2596e-09, + 3.1199e-08, 1.0710e-08], + [ 3.2596e-09, 3.7253e-09, 9.3132e-10, ..., 4.6566e-10, + 1.8626e-09, 0.0000e+00], + ..., + [ 4.1910e-09, 4.1910e-09, 0.0000e+00, ..., 2.3283e-09, + 1.8626e-09, 4.6566e-10], + [-1.7136e-07, -1.8394e-07, 0.0000e+00, ..., 3.7253e-09, + -2.0023e-07, -7.3574e-08], + [-1.4901e-08, -3.7253e-09, 0.0000e+00, ..., -1.5832e-08, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0402, -0.0421, -0.0099, -0.0067, 0.0018, 0.0096, 0.0053, -0.0080, + 0.0159, -0.0436], device='cuda:0'), grad: tensor([-9.7789e-09, 2.3935e-07, 4.9826e-08, 4.2841e-08, 8.7079e-08, + 6.1048e-07, 4.6566e-10, -5.9186e-07, -8.1072e-07, 3.8184e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 217.54, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4675 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.11 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.1260, 0.0212, 0.1593, ..., 0.0102, -0.1343, -0.1125], + [-0.2504, -0.2507, -0.2297, ..., 0.1196, 0.0687, 0.4317], + [-0.2837, -0.1931, -0.2596, ..., -0.3226, 0.0166, -0.2611], + ..., + [-0.2208, -0.2899, -0.0146, ..., -0.2822, -0.1610, -0.4957], + [ 0.1640, 0.0381, -0.1701, ..., -0.2756, 0.0528, -0.2517], + [ 0.1676, 0.1048, -0.1622, ..., 0.1343, -0.2684, -0.3316]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-10, -1.3970e-09, ..., 3.7253e-09, + 4.6566e-10, 6.0536e-09], + [-3.2596e-09, -1.1176e-08, 0.0000e+00, ..., -7.1712e-08, + -5.5879e-09, -1.0245e-07], + [ 4.6566e-10, 1.3970e-09, 4.6566e-10, ..., 3.2596e-09, + 1.8626e-09, 4.6566e-09], + ..., + [ 4.6566e-09, 6.9849e-09, 0.0000e+00, ..., 3.5390e-08, + -9.3132e-10, 4.3772e-08], + [ 1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 6.9849e-09, + 3.2596e-09, 1.2573e-08], + [-7.9162e-09, -4.6566e-09, 9.3132e-10, ..., 1.0245e-08, + 9.3132e-10, 2.2352e-08]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0404, -0.0421, -0.0099, -0.0073, 0.0018, 0.0098, 0.0052, -0.0080, + 0.0164, -0.0437], device='cuda:0'), grad: tensor([ 1.2573e-08, -2.3609e-07, 2.2817e-08, -2.1420e-08, 3.0268e-08, + 4.2375e-08, 1.8161e-08, 8.5216e-08, 2.8871e-08, 3.1199e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 217.73, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4630 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.1259, 0.0215, 0.1600, ..., 0.0109, -0.1339, -0.1116], + [-0.2497, -0.2504, -0.2301, ..., 0.1202, 0.0689, 0.4343], + [-0.2841, -0.1934, -0.2595, ..., -0.3230, 0.0167, -0.2612], + ..., + [-0.2198, -0.2877, -0.0147, ..., -0.2826, -0.1615, -0.5011], + [ 0.1646, 0.0386, -0.1701, ..., -0.2757, 0.0530, -0.2518], + [ 0.1676, 0.1046, -0.1629, ..., 0.1342, -0.2688, -0.3332]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -9.3132e-10, -1.2107e-08, ..., 0.0000e+00, + 7.9162e-09, -9.3132e-10], + [ 9.7789e-09, -7.0781e-08, -1.9511e-07, ..., -8.0746e-07, + -3.6927e-07, -1.7006e-06], + [ 4.1444e-08, 3.7719e-08, 1.2573e-08, ..., 2.5146e-08, + 3.2131e-08, 4.1910e-09], + ..., + [ 3.2596e-09, 8.1491e-08, 1.9651e-07, ..., 8.0187e-07, + 3.7253e-07, 1.6745e-06], + [-1.9604e-07, -1.7276e-07, -4.8429e-08, ..., -1.0524e-07, + -1.4622e-07, 2.7940e-09], + [ 3.4459e-08, 3.3993e-08, 1.3504e-08, ..., 2.5611e-08, + 2.8405e-08, 1.2107e-08]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0400, -0.0416, -0.0098, -0.0098, 0.0022, 0.0093, 0.0048, -0.0081, + 0.0166, -0.0442], device='cuda:0'), grad: tensor([ 1.3039e-08, -4.5262e-06, 2.4540e-07, 3.9581e-08, 1.5367e-08, + 1.1316e-07, 4.0792e-07, 4.5262e-06, -1.0766e-06, 2.3609e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 217.68, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4531 re_mapping 0.0032 re_causal 0.0091 /// teacc 99.05 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.1259, 0.0220, 0.1609, ..., 0.0116, -0.1341, -0.1115], + [-0.2497, -0.2505, -0.2299, ..., 0.1205, 0.0695, 0.4351], + [-0.2848, -0.1938, -0.2596, ..., -0.3246, 0.0168, -0.2615], + ..., + [-0.2200, -0.2878, -0.0149, ..., -0.2829, -0.1625, -0.5019], + [ 0.1652, 0.0389, -0.1703, ..., -0.2757, 0.0531, -0.2520], + [ 0.1677, 0.1047, -0.1634, ..., 0.1343, -0.2692, -0.3340]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 7.9162e-09, 2.3283e-09, ..., 4.6566e-10, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + ..., + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 4.6566e-10, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 7.4971e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [-1.3970e-09, 0.0000e+00, 2.0489e-08, ..., -1.3970e-09, + 2.5611e-08, 0.0000e+00]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0394, -0.0415, -0.0098, -0.0097, 0.0022, 0.0089, 0.0047, -0.0080, + 0.0168, -0.0450], device='cuda:0'), grad: tensor([ 4.1910e-09, 8.6613e-08, 1.0198e-07, 4.1910e-09, -1.2992e-07, + -4.6827e-06, 4.0904e-06, -1.1967e-07, 5.5088e-07, 9.4064e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 218.03, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4337 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.08 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.1264, 0.0207, 0.1605, ..., 0.0099, -0.1345, -0.1117], + [-0.2487, -0.2518, -0.2304, ..., 0.1209, 0.0696, 0.4354], + [-0.2853, -0.1941, -0.2602, ..., -0.3252, 0.0179, -0.2617], + ..., + [-0.2214, -0.2880, -0.0148, ..., -0.2833, -0.1626, -0.5023], + [ 0.1652, 0.0388, -0.1708, ..., -0.2759, 0.0515, -0.2529], + [ 0.1679, 0.1049, -0.1636, ..., 0.1346, -0.2696, -0.3360]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 1.3970e-09, 7.4506e-09, 0.0000e+00, ..., -8.8476e-09, + -6.5193e-09, -1.6298e-08], + [ 9.3132e-10, 6.0536e-09, 0.0000e+00, ..., 6.0536e-09, + 5.1223e-09, 8.8476e-09], + ..., + [ 3.7253e-09, 6.0536e-09, 0.0000e+00, ..., 6.0536e-09, + 1.3970e-09, 3.2596e-09], + [ 4.6566e-10, 3.7253e-09, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 1.8626e-09], + [-1.0710e-08, 6.0536e-09, 0.0000e+00, ..., -1.2107e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0408, -0.0414, -0.0089, -0.0102, 0.0023, 0.0079, 0.0065, -0.0080, + 0.0156, -0.0450], device='cuda:0'), grad: tensor([ 1.3970e-08, -7.4506e-09, 5.7742e-08, 5.7183e-07, 4.5635e-08, + -9.0990e-07, 1.6857e-07, 2.2817e-08, 2.3283e-08, 3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 217.74, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4519 re_mapping 0.0028 re_causal 0.0086 /// teacc 99.16 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.1272, 0.0204, 0.1617, ..., 0.0102, -0.1349, -0.1118], + [-0.2492, -0.2523, -0.2305, ..., 0.1211, 0.0700, 0.4360], + [-0.2864, -0.1958, -0.2609, ..., -0.3268, 0.0180, -0.2619], + ..., + [-0.2217, -0.2881, -0.0148, ..., -0.2834, -0.1629, -0.5027], + [ 0.1654, 0.0388, -0.1711, ..., -0.2760, 0.0512, -0.2538], + [ 0.1680, 0.1058, -0.1644, ..., 0.1354, -0.2700, -0.3374]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 1.3970e-09, ..., 1.1176e-08, + 4.7497e-08, -5.3085e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 3.2596e-09], + ..., + [ 0.0000e+00, -4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 1.4901e-08, 5.0757e-08], + [ 4.1910e-09, 4.1910e-09, 4.6566e-10, ..., 4.1910e-09, + 0.0000e+00, 4.6566e-10], + [-6.9849e-09, -6.9849e-09, 2.6962e-07, ..., -5.5879e-09, + 3.3015e-07, -9.3132e-10]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0408, -0.0413, -0.0089, -0.0089, 0.0026, 0.0071, 0.0067, -0.0080, + 0.0155, -0.0450], device='cuda:0'), grad: tensor([ 2.3283e-09, 8.5682e-08, 9.3132e-10, 4.6566e-09, -2.0936e-06, + 4.6566e-09, 9.3132e-09, 7.4971e-08, 1.8161e-08, 1.8906e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 217.51, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4855 re_mapping 0.0028 re_causal 0.0091 /// teacc 99.19 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.1276, 0.0204, 0.1621, ..., 0.0103, -0.1352, -0.1119], + [-0.2498, -0.2527, -0.2307, ..., 0.1211, 0.0701, 0.4363], + [-0.2869, -0.1961, -0.2610, ..., -0.3271, 0.0182, -0.2620], + ..., + [-0.2222, -0.2883, -0.0149, ..., -0.2836, -0.1634, -0.5030], + [ 0.1654, 0.0385, -0.1718, ..., -0.2762, 0.0513, -0.2539], + [ 0.1683, 0.1060, -0.1666, ..., 0.1357, -0.2708, -0.3380]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 4.6566e-10, 1.8626e-09, ..., 1.8626e-09, + 1.3970e-09, 0.0000e+00], + [ 1.3970e-09, 7.9628e-08, 4.6566e-10, ..., 9.3132e-10, + 1.8626e-09, 5.0291e-08], + [ 6.5193e-09, -9.3132e-10, 3.7253e-09, ..., 1.3970e-09, + 7.4506e-09, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 1.3970e-09], + [-1.5367e-08, -9.3132e-10, -1.1176e-08, ..., -1.2573e-08, + -9.3132e-09, 0.0000e+00], + [ 9.3132e-10, 5.1223e-09, 4.6566e-10, ..., 3.7253e-09, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0409, -0.0414, -0.0085, -0.0086, 0.0040, 0.0067, 0.0071, -0.0081, + 0.0152, -0.0464], device='cuda:0'), grad: tensor([ 2.2817e-08, 2.1001e-07, -5.9605e-08, -5.5414e-08, 9.3132e-09, + -1.9325e-07, 8.1491e-08, 8.8476e-09, -9.0338e-08, 7.9162e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 217.63, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4484 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.15 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.1281, 0.0200, 0.1621, ..., 0.0101, -0.1356, -0.1119], + [-0.2502, -0.2531, -0.2310, ..., 0.1211, 0.0701, 0.4365], + [-0.2875, -0.1966, -0.2611, ..., -0.3274, 0.0183, -0.2620], + ..., + [-0.2226, -0.2885, -0.0149, ..., -0.2837, -0.1636, -0.5032], + [ 0.1655, 0.0385, -0.1720, ..., -0.2763, 0.0512, -0.2539], + [ 0.1686, 0.1062, -0.1667, ..., 0.1360, -0.2710, -0.3381]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-09, 5.5879e-09, ..., 1.1642e-08, + 9.3132e-10, 5.1223e-09], + [ 4.6566e-10, 4.6566e-10, 1.2945e-07, ..., 2.7381e-07, + 0.0000e+00, 1.1735e-07], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-6.9849e-09, -4.6566e-09, 9.3132e-10, ..., 2.7940e-09, + -3.7253e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0414, -0.0427, -0.0080, -0.0082, 0.0037, 0.0071, 0.0066, -0.0069, + 0.0151, -0.0462], device='cuda:0'), grad: tensor([ 2.7940e-08, 5.5134e-07, 2.7940e-09, 9.3132e-09, 1.8626e-09, + 3.2596e-09, -5.7509e-07, -2.7940e-09, -1.5832e-08, 2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 218.05, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4553 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.14 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.1290, 0.0211, 0.1643, ..., 0.0107, -0.1361, -0.1119], + [-0.2503, -0.2534, -0.2313, ..., 0.1211, 0.0701, 0.4367], + [-0.2882, -0.1972, -0.2615, ..., -0.3276, 0.0183, -0.2620], + ..., + [-0.2229, -0.2888, -0.0150, ..., -0.2837, -0.1637, -0.5034], + [ 0.1656, 0.0385, -0.1725, ..., -0.2765, 0.0513, -0.2540], + [ 0.1687, 0.1062, -0.1669, ..., 0.1360, -0.2717, -0.3391]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 1.3970e-08, ..., 3.7253e-09, + 1.8626e-08, -8.3819e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 9.3132e-10], + ..., + [ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 9.3132e-10, + -5.5879e-08, 3.7253e-09], + [ 3.7253e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [-5.5879e-09, 3.7253e-09, -1.8626e-09, ..., -1.5832e-08, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0403, -0.0428, -0.0075, -0.0089, 0.0038, 0.0073, 0.0067, -0.0069, + 0.0151, -0.0464], device='cuda:0'), grad: tensor([ 1.0245e-08, 7.0781e-08, 4.7497e-08, 7.2643e-08, 1.3039e-07, + -2.4680e-07, 6.7055e-08, -1.8254e-07, 2.4214e-08, -2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 217.97, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4628 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.1294, 0.0209, 0.1643, ..., 0.0103, -0.1366, -0.1121], + [-0.2506, -0.2535, -0.2314, ..., 0.1212, 0.0705, 0.4373], + [-0.2892, -0.1977, -0.2622, ..., -0.3287, 0.0180, -0.2626], + ..., + [-0.2233, -0.2889, -0.0149, ..., -0.2838, -0.1638, -0.5036], + [ 0.1657, 0.0385, -0.1726, ..., -0.2765, 0.0516, -0.2541], + [ 0.1690, 0.1064, -0.1670, ..., 0.1363, -0.2719, -0.3394]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [-1.7695e-08, -6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.1176e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0409, -0.0427, -0.0082, -0.0094, 0.0037, 0.0075, 0.0067, -0.0068, + 0.0151, -0.0463], device='cuda:0'), grad: tensor([ 0.0000e+00, 3.7253e-09, 7.4506e-09, 2.3469e-07, 1.0245e-08, + -2.4028e-07, 2.7940e-09, -9.4064e-08, 1.7695e-08, 6.5193e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 217.61, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4822 re_mapping 0.0028 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.1297, 0.0213, 0.1658, ..., 0.0102, -0.1368, -0.1118], + [-0.2509, -0.2537, -0.2322, ..., 0.1212, 0.0705, 0.4378], + [-0.2900, -0.1982, -0.2627, ..., -0.3292, 0.0179, -0.2627], + ..., + [-0.2235, -0.2891, -0.0147, ..., -0.2838, -0.1641, -0.5044], + [ 0.1659, 0.0384, -0.1731, ..., -0.2767, 0.0520, -0.2543], + [ 0.1691, 0.1063, -0.1675, ..., 0.1364, -0.2724, -0.3402]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.1467e-08, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-6.4261e-08, -1.8626e-08, 0.0000e+00, ..., 0.0000e+00, + -2.5146e-08, -9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 4.4703e-08, ..., 9.3132e-10, + 4.0978e-08, 0.0000e+00]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0401, -0.0427, -0.0078, -0.0095, 0.0037, 0.0076, 0.0067, -0.0068, + 0.0152, -0.0467], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.7253e-09, 3.8836e-07, 9.3132e-10, -2.4308e-07, + -1.8626e-09, 0.0000e+00, -2.7940e-09, -4.0419e-07, 2.5984e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 217.90, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4262 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.24 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.1298, 0.0220, 0.1661, ..., 0.0103, -0.1374, -0.1118], + [-0.2511, -0.2539, -0.2325, ..., 0.1214, 0.0706, 0.4392], + [-0.2907, -0.1988, -0.2629, ..., -0.3294, 0.0178, -0.2628], + ..., + [-0.2243, -0.2895, -0.0146, ..., -0.2842, -0.1643, -0.5064], + [ 0.1663, 0.0388, -0.1732, ..., -0.2768, 0.0523, -0.2544], + [ 0.1693, 0.1064, -0.1676, ..., 0.1367, -0.2725, -0.3405]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -2.7940e-08, ..., -8.3819e-09, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -9.3132e-09, + -2.1420e-08, -7.9162e-08], + [ 0.0000e+00, 9.3132e-10, 4.6566e-09, ..., 4.6566e-09, + 4.6566e-09, 1.7695e-08], + ..., + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 2.7940e-09, + 1.8626e-09, 1.0245e-08], + [ 9.3132e-10, 2.7940e-09, 1.1176e-08, ..., 8.3819e-09, + -9.3132e-10, 1.3039e-08], + [-1.2107e-08, -9.3132e-09, 2.7940e-09, ..., -3.7253e-09, + 2.7940e-09, 6.5193e-09]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0401, -0.0422, -0.0079, -0.0094, 0.0035, 0.0074, 0.0067, -0.0073, + 0.0155, -0.0465], device='cuda:0'), grad: tensor([-4.4703e-08, -5.9605e-08, 5.9605e-08, 3.7253e-09, 3.7253e-08, + 3.5390e-08, 4.7497e-08, -2.2911e-07, 5.2154e-08, 1.0245e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 217.48, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4750 re_mapping 0.0028 re_causal 0.0084 /// teacc 99.20 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.1302, 0.0220, 0.1672, ..., 0.0107, -0.1385, -0.1120], + [-0.2515, -0.2542, -0.2329, ..., 0.1215, 0.0707, 0.4397], + [-0.2924, -0.2007, -0.2642, ..., -0.3316, 0.0175, -0.2628], + ..., + [-0.2246, -0.2897, -0.0145, ..., -0.2843, -0.1645, -0.5067], + [ 0.1664, 0.0387, -0.1737, ..., -0.2770, 0.0527, -0.2542], + [ 0.1696, 0.1067, -0.1677, ..., 0.1370, -0.2726, -0.3408]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 5.5879e-09, 3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.2107e-08, -1.2107e-08, 0.0000e+00, ..., 1.8626e-09, + -1.9558e-08, 0.0000e+00], + [-8.3819e-09, -1.8626e-09, 0.0000e+00, ..., -1.3970e-08, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0399, -0.0422, -0.0081, -0.0085, 0.0033, 0.0074, 0.0067, -0.0073, + 0.0155, -0.0464], device='cuda:0'), grad: tensor([ 1.5832e-08, 1.3225e-07, -7.4506e-08, 3.1665e-08, 2.1420e-08, + 1.6764e-08, 1.0245e-08, -7.2643e-08, -1.0617e-07, 3.2596e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 217.81, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4754 re_mapping 0.0027 re_causal 0.0085 /// teacc 99.15 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.1315, 0.0217, 0.1677, ..., 0.0104, -0.1388, -0.1120], + [-0.2518, -0.2545, -0.2328, ..., 0.1216, 0.0708, 0.4401], + [-0.2940, -0.2021, -0.2643, ..., -0.3328, 0.0177, -0.2629], + ..., + [-0.2250, -0.2900, -0.0145, ..., -0.2843, -0.1649, -0.5069], + [ 0.1667, 0.0389, -0.1739, ..., -0.2770, 0.0527, -0.2542], + [ 0.1703, 0.1071, -0.1681, ..., 0.1371, -0.2739, -0.3410]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 4.6566e-09, ..., 0.0000e+00, + 1.3039e-08, -9.8720e-08], + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 3.7253e-09, + 2.7940e-09, 2.5146e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 9.3132e-10, 7.4506e-08], + [-5.5879e-09, -5.5879e-09, 5.5879e-08, ..., 1.9930e-07, + 7.1712e-08, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 1.1176e-08, 9.3132e-10]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0403, -0.0423, -0.0077, -0.0080, 0.0031, 0.0072, 0.0069, -0.0073, + 0.0156, -0.0463], device='cuda:0'), grad: tensor([ 2.8871e-08, -1.3225e-07, 8.3819e-08, 1.0617e-07, -6.7055e-08, + 2.1420e-07, -1.8580e-06, 1.6484e-07, 1.4035e-06, 5.4017e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 217.74, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4165 re_mapping 0.0031 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.1320, 0.0215, 0.1679, ..., 0.0103, -0.1390, -0.1121], + [-0.2522, -0.2546, -0.2330, ..., 0.1216, 0.0708, 0.4404], + [-0.2943, -0.2021, -0.2645, ..., -0.3330, 0.0178, -0.2630], + ..., + [-0.2259, -0.2903, -0.0144, ..., -0.2844, -0.1651, -0.5072], + [ 0.1668, 0.0390, -0.1741, ..., -0.2773, 0.0528, -0.2544], + [ 0.1706, 0.1072, -0.1682, ..., 0.1373, -0.2741, -0.3411]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, -9.3132e-10, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 9.3132e-10, 5.5879e-09, ..., -9.3132e-10, + 1.5832e-08, -5.5879e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 3.7253e-09], + ..., + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-2.0489e-08, 5.5879e-09, -2.1420e-08, ..., 9.3132e-10, + -7.3574e-08, 9.3132e-10], + [-9.3132e-10, 1.5832e-08, 0.0000e+00, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0408, -0.0423, -0.0072, -0.0081, 0.0031, 0.0067, 0.0073, -0.0074, + 0.0156, -0.0463], device='cuda:0'), grad: tensor([ 2.7940e-09, 9.6858e-08, 2.1420e-08, -3.1665e-08, 4.6566e-09, + 4.1910e-08, 1.3411e-07, -1.6764e-07, -2.7753e-07, 1.7602e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 217.65, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4966 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.1322, 0.0216, 0.1679, ..., 0.0102, -0.1396, -0.1124], + [-0.2527, -0.2549, -0.2331, ..., 0.1223, 0.0719, 0.4409], + [-0.2957, -0.2023, -0.2649, ..., -0.3333, 0.0176, -0.2631], + ..., + [-0.2266, -0.2906, -0.0144, ..., -0.2851, -0.1669, -0.5077], + [ 0.1671, 0.0392, -0.1741, ..., -0.2774, 0.0536, -0.2544], + [ 0.1707, 0.1071, -0.1684, ..., 0.1374, -0.2748, -0.3413]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, -2.7940e-09], + [ 6.5193e-09, 3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 1.1176e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [-7.4506e-09, -4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + -1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0413, -0.0422, -0.0078, -0.0070, 0.0031, 0.0063, 0.0074, -0.0074, + 0.0160, -0.0466], device='cuda:0'), grad: tensor([ 3.4459e-08, 5.0291e-08, 1.1548e-07, 9.3132e-10, 1.1735e-07, + 6.5193e-09, -3.2317e-07, 6.5193e-09, -1.3970e-08, 8.3819e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 217.68, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4535 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.1323, 0.0198, 0.1677, ..., 0.0078, -0.1401, -0.1130], + [-0.2536, -0.2550, -0.2337, ..., 0.1223, 0.0720, 0.4414], + [-0.2965, -0.2024, -0.2668, ..., -0.3337, 0.0165, -0.2635], + ..., + [-0.2270, -0.2908, -0.0143, ..., -0.2852, -0.1670, -0.5079], + [ 0.1672, 0.0392, -0.1739, ..., -0.2776, 0.0550, -0.2534], + [ 0.1735, 0.1086, -0.1686, ..., 0.1382, -0.2754, -0.3418]], + device='cuda:0'), grad: tensor([[ 5.4948e-08, 4.5635e-08, 1.6764e-08, ..., 1.0803e-07, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.6566e-09, -2.7940e-09, 9.3132e-10, ..., 3.7253e-09, + -4.6566e-09, 0.0000e+00], + [-5.7742e-08, -4.7497e-08, -1.6764e-08, ..., -1.1269e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0438, -0.0423, -0.0079, -0.0070, 0.0007, 0.0064, 0.0074, -0.0073, + 0.0164, -0.0441], device='cuda:0'), grad: tensor([ 2.5518e-07, 8.3819e-09, 4.0047e-08, 1.8626e-09, 3.4459e-08, + 1.8626e-08, -4.2841e-08, 7.4506e-09, -5.3085e-08, -2.6263e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 217.79, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4494 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.15 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.1325, 0.0202, 0.1689, ..., 0.0081, -0.1405, -0.1130], + [-0.2540, -0.2553, -0.2343, ..., 0.1224, 0.0718, 0.4416], + [-0.2968, -0.2024, -0.2671, ..., -0.3337, 0.0166, -0.2636], + ..., + [-0.2273, -0.2909, -0.0144, ..., -0.2854, -0.1671, -0.5083], + [ 0.1658, 0.0384, -0.1744, ..., -0.2778, 0.0544, -0.2534], + [ 0.1738, 0.1088, -0.1687, ..., 0.1383, -0.2756, -0.3420]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.7416e-07, ..., 0.0000e+00, + 8.4750e-08, 6.5193e-09], + [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., -2.7940e-09, + 9.3132e-10, -1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 7.8231e-08, ..., 1.8626e-09, + 4.0047e-08, 1.3039e-08], + ..., + [ 2.7940e-09, 1.8626e-09, 4.6566e-09, ..., 9.3132e-10, + 3.7253e-09, 3.7253e-09], + [-6.5193e-09, -4.6566e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 1.8626e-09, 2.7940e-09, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0431, -0.0423, -0.0075, -0.0061, 0.0005, 0.0063, 0.0074, -0.0074, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([ 4.5728e-07, 1.8626e-09, 2.1514e-07, 8.3819e-09, -8.5682e-07, + 5.5879e-09, 1.4808e-07, 2.7940e-08, -1.3039e-08, 1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 217.83, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4416 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.16 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.1324, 0.0203, 0.1690, ..., 0.0081, -0.1406, -0.1130], + [-0.2540, -0.2553, -0.2344, ..., 0.1224, 0.0718, 0.4417], + [-0.2969, -0.2024, -0.2673, ..., -0.3337, 0.0166, -0.2636], + ..., + [-0.2273, -0.2909, -0.0144, ..., -0.2855, -0.1672, -0.5083], + [ 0.1658, 0.0385, -0.1744, ..., -0.2778, 0.0544, -0.2534], + [ 0.1738, 0.1088, -0.1687, ..., 0.1384, -0.2756, -0.3420]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 9.3132e-10], + [ 9.3132e-09, 4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + -4.6566e-09, -3.0734e-08], + [ 4.6566e-09, 1.8626e-09, 0.0000e+00, ..., 5.5879e-09, + 2.7940e-09, 5.5879e-09], + ..., + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, + 2.7940e-09, 8.3819e-09], + [-7.1712e-08, -7.0781e-08, -2.7940e-09, ..., -7.4506e-09, + -4.0978e-08, 9.3132e-10], + [ 5.9605e-08, 5.7742e-08, 3.7253e-09, ..., 1.2107e-08, + 3.4459e-08, 9.3132e-10]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0430, -0.0423, -0.0075, -0.0060, 0.0005, 0.0063, 0.0074, -0.0074, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([ 1.3039e-08, 3.0734e-08, -7.6368e-08, -8.6613e-08, 1.3970e-08, + 4.0978e-08, 2.5146e-08, 5.6811e-08, -2.2165e-07, 2.0955e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 217.77, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4499 re_mapping 0.0025 re_causal 0.0080 /// teacc 99.14 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.1325, 0.0202, 0.1691, ..., 0.0081, -0.1406, -0.1130], + [-0.2540, -0.2553, -0.2344, ..., 0.1224, 0.0718, 0.4417], + [-0.2969, -0.2024, -0.2673, ..., -0.3337, 0.0166, -0.2636], + ..., + [-0.2273, -0.2909, -0.0144, ..., -0.2855, -0.1672, -0.5084], + [ 0.1658, 0.0385, -0.1744, ..., -0.2778, 0.0545, -0.2534], + [ 0.1738, 0.1088, -0.1687, ..., 0.1384, -0.2757, -0.3420]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.9652e-08, + -5.4948e-08, -2.0489e-08], + [ 1.8626e-09, 9.3132e-10, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.5926e-08, + 5.3085e-08, 2.0489e-08], + [-2.7940e-09, 9.3132e-10, 1.8626e-09, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., 5.5879e-09, + 3.7253e-09, 9.3132e-10]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0430, -0.0425, -0.0075, -0.0060, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([ 8.3819e-09, -2.4214e-07, -3.7253e-09, -1.2107e-08, -7.4506e-09, + 1.2107e-08, -1.7695e-08, 2.0210e-07, 4.0047e-08, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 217.56, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4139 re_mapping 0.0025 re_causal 0.0078 /// teacc 99.11 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.1326, 0.0202, 0.1692, ..., 0.0081, -0.1406, -0.1130], + [-0.2542, -0.2554, -0.2345, ..., 0.1224, 0.0718, 0.4418], + [-0.2969, -0.2024, -0.2674, ..., -0.3337, 0.0166, -0.2637], + ..., + [-0.2274, -0.2910, -0.0144, ..., -0.2855, -0.1672, -0.5084], + [ 0.1659, 0.0385, -0.1745, ..., -0.2778, 0.0545, -0.2535], + [ 0.1738, 0.1088, -0.1688, ..., 0.1384, -0.2758, -0.3421]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, 5.5879e-09, ..., 9.3132e-10, + 1.3970e-08, 2.7940e-09], + [ 1.4901e-08, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 5.6811e-08, 3.3528e-08], + ..., + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 1.8626e-09], + [-1.8626e-08, 9.3132e-10, -1.0245e-08, ..., 9.3132e-10, + -6.7055e-08, -3.8184e-08], + [-1.6764e-08, -1.0245e-08, 1.0245e-08, ..., -1.7695e-08, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0430, -0.0425, -0.0074, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([-9.3132e-10, 4.9360e-08, 1.7416e-07, 6.4261e-08, -3.0734e-08, + -7.8231e-08, 1.1176e-08, 9.3132e-09, -2.0023e-07, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 217.85, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4298 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.14 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.1327, 0.0201, 0.1692, ..., 0.0081, -0.1406, -0.1130], + [-0.2542, -0.2554, -0.2345, ..., 0.1225, 0.0718, 0.4418], + [-0.2970, -0.2024, -0.2674, ..., -0.3337, 0.0166, -0.2637], + ..., + [-0.2274, -0.2910, -0.0144, ..., -0.2855, -0.1672, -0.5085], + [ 0.1659, 0.0385, -0.1745, ..., -0.2778, 0.0545, -0.2535], + [ 0.1738, 0.1088, -0.1688, ..., 0.1385, -0.2759, -0.3421]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + -9.3132e-10, -1.0245e-08], + [ 2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 2.7940e-09, 9.3132e-10], + ..., + [ 1.3039e-08, 6.5193e-09, 0.0000e+00, ..., 1.3039e-08, + 8.3819e-09, 5.5879e-09], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 4.6566e-09, + 9.3132e-10, 9.3132e-10], + [ 2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0430, -0.0425, -0.0074, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([ 4.6566e-09, 0.0000e+00, 1.2107e-08, -1.2759e-07, 5.5879e-09, + 5.5879e-09, -2.5146e-08, 7.6368e-08, 2.2352e-08, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 217.70, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4610 re_mapping 0.0024 re_causal 0.0078 /// teacc 99.16 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.1329, 0.0200, 0.1692, ..., 0.0079, -0.1406, -0.1130], + [-0.2543, -0.2554, -0.2345, ..., 0.1225, 0.0718, 0.4419], + [-0.2970, -0.2025, -0.2674, ..., -0.3338, 0.0166, -0.2637], + ..., + [-0.2275, -0.2910, -0.0144, ..., -0.2855, -0.1672, -0.5085], + [ 0.1659, 0.0385, -0.1745, ..., -0.2778, 0.0546, -0.2535], + [ 0.1738, 0.1088, -0.1688, ..., 0.1385, -0.2759, -0.3421]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -6.5193e-09, + -5.5879e-09, -2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 7.4506e-09], + ..., + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, + 1.8626e-09, 7.4506e-09], + [-9.3132e-10, -9.3132e-10, 1.9558e-08, ..., 1.9558e-08, + -1.8626e-09, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0431, -0.0425, -0.0073, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0160, -0.0439], device='cuda:0'), grad: tensor([ 1.2107e-08, -2.6077e-08, -1.8626e-08, 1.4901e-07, 5.5879e-09, + -1.5460e-07, -1.2573e-07, 3.4459e-08, 1.2014e-07, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 217.78, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4287 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.1329, 0.0200, 0.1692, ..., 0.0079, -0.1407, -0.1130], + [-0.2543, -0.2554, -0.2345, ..., 0.1225, 0.0718, 0.4419], + [-0.2970, -0.2025, -0.2675, ..., -0.3338, 0.0166, -0.2637], + ..., + [-0.2275, -0.2910, -0.0144, ..., -0.2855, -0.1673, -0.5085], + [ 0.1659, 0.0385, -0.1745, ..., -0.2779, 0.0546, -0.2535], + [ 0.1739, 0.1088, -0.1688, ..., 0.1385, -0.2760, -0.3422]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [-1.3039e-08, -7.4506e-09, 0.0000e+00, ..., -1.6764e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0432, -0.0425, -0.0073, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0160, -0.0439], device='cuda:0'), grad: tensor([ 2.7940e-09, 2.1420e-08, -1.0710e-07, 2.5239e-07, 1.0245e-08, + -2.2631e-07, -9.3132e-10, 7.1712e-08, 2.0489e-08, -3.8184e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 217.58, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4136 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.1330, 0.0200, 0.1692, ..., 0.0079, -0.1407, -0.1130], + [-0.2543, -0.2554, -0.2346, ..., 0.1225, 0.0718, 0.4420], + [-0.2970, -0.2025, -0.2675, ..., -0.3338, 0.0166, -0.2637], + ..., + [-0.2275, -0.2910, -0.0144, ..., -0.2856, -0.1673, -0.5086], + [ 0.1659, 0.0385, -0.1746, ..., -0.2779, 0.0546, -0.2536], + [ 0.1739, 0.1088, -0.1689, ..., 0.1386, -0.2761, -0.3422]], + device='cuda:0'), grad: tensor([[-1.8626e-09, -3.7253e-09, 6.8918e-08, ..., 9.0338e-08, + 2.6077e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 1.7695e-08, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, + 3.7253e-09, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 3.6322e-08, ..., 9.3132e-10, + 3.2596e-08, 0.0000e+00]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0432, -0.0425, -0.0073, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0160, -0.0439], device='cuda:0'), grad: tensor([ 3.1386e-07, 6.7055e-08, -1.0245e-08, -3.7253e-09, 4.1910e-08, + 1.0151e-07, -6.5565e-07, 0.0000e+00, 2.7940e-08, 1.1828e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 217.88, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4217 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.1330, 0.0199, 0.1692, ..., 0.0078, -0.1408, -0.1131], + [-0.2544, -0.2555, -0.2346, ..., 0.1225, 0.0719, 0.4421], + [-0.2971, -0.2025, -0.2675, ..., -0.3338, 0.0166, -0.2638], + ..., + [-0.2276, -0.2911, -0.0144, ..., -0.2856, -0.1673, -0.5087], + [ 0.1660, 0.0386, -0.1746, ..., -0.2779, 0.0546, -0.2536], + [ 0.1739, 0.1089, -0.1689, ..., 0.1386, -0.2761, -0.3423]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, -2.7940e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 1.8626e-09], + [-2.7940e-09, -9.3132e-10, 1.8626e-09, ..., 2.7940e-09, + -2.7940e-09, 9.3132e-10], + [-1.8626e-09, -9.3132e-10, 0.0000e+00, ..., -2.7940e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0433, -0.0425, -0.0073, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0160, -0.0439], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.3970e-08, -1.0990e-07, -1.2573e-07, 2.7940e-09, + 1.1362e-07, -7.4506e-09, 1.1642e-07, -3.7253e-09, -4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 217.78, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4255 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.24 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.1330, 0.0200, 0.1692, ..., 0.0078, -0.1408, -0.1131], + [-0.2544, -0.2555, -0.2346, ..., 0.1225, 0.0719, 0.4421], + [-0.2971, -0.2025, -0.2676, ..., -0.3338, 0.0166, -0.2638], + ..., + [-0.2276, -0.2911, -0.0144, ..., -0.2856, -0.1673, -0.5088], + [ 0.1660, 0.0386, -0.1746, ..., -0.2779, 0.0547, -0.2536], + [ 0.1739, 0.1089, -0.1689, ..., 0.1386, -0.2762, -0.3423]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 5.5879e-09, 9.3132e-10, ..., 6.5193e-09, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 2.7940e-09], + ..., + [ 1.0245e-08, 7.4506e-09, -9.3132e-10, ..., 9.3132e-09, + 1.8626e-09, 1.8626e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, -9.3132e-10, 0.0000e+00, ..., -1.4901e-08, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0433, -0.0425, -0.0073, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0160, -0.0439], device='cuda:0'), grad: tensor([ 8.3819e-09, 5.3085e-08, -3.4459e-08, -1.3970e-07, 5.2154e-08, + 2.2352e-08, 7.4506e-09, 4.0978e-08, 1.9558e-08, -1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 217.47, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4259 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.24 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.1330, 0.0200, 0.1694, ..., 0.0079, -0.1408, -0.1131], + [-0.2545, -0.2555, -0.2346, ..., 0.1225, 0.0719, 0.4422], + [-0.2971, -0.2025, -0.2676, ..., -0.3339, 0.0166, -0.2638], + ..., + [-0.2277, -0.2911, -0.0144, ..., -0.2856, -0.1673, -0.5088], + [ 0.1660, 0.0386, -0.1746, ..., -0.2779, 0.0547, -0.2537], + [ 0.1739, 0.1089, -0.1689, ..., 0.1387, -0.2762, -0.3424]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, 9.3132e-10, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, 7.4506e-09, + 0.0000e+00]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0432, -0.0425, -0.0073, -0.0059, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0160, -0.0439], device='cuda:0'), grad: tensor([ 2.7940e-09, 1.0245e-08, 1.8626e-09, 1.9558e-08, -3.4459e-08, + -6.3330e-08, 2.0489e-08, -6.5193e-09, 1.3039e-08, 3.7253e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 217.34, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4379 re_mapping 0.0023 re_causal 0.0075 /// teacc 99.22 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.1330, 0.0200, 0.1694, ..., 0.0079, -0.1409, -0.1131], + [-0.2545, -0.2555, -0.2347, ..., 0.1226, 0.0719, 0.4423], + [-0.2972, -0.2025, -0.2677, ..., -0.3339, 0.0166, -0.2638], + ..., + [-0.2278, -0.2912, -0.0144, ..., -0.2856, -0.1673, -0.5089], + [ 0.1658, 0.0384, -0.1746, ..., -0.2780, 0.0546, -0.2537], + [ 0.1739, 0.1089, -0.1689, ..., 0.1387, -0.2762, -0.3424]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -1.8626e-09, -5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 1.8626e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 1.8626e-09], + [-9.3132e-10, -9.3132e-10, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0432, -0.0425, -0.0073, -0.0058, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([ 0.0000e+00, -4.6566e-09, 4.6566e-10, -2.7940e-09, 1.8626e-09, + 2.7940e-09, 9.3132e-10, 9.3132e-10, 3.2596e-09, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 217.49, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4403 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.25 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.1330, 0.0200, 0.1694, ..., 0.0079, -0.1409, -0.1131], + [-0.2546, -0.2556, -0.2347, ..., 0.1226, 0.0719, 0.4423], + [-0.2972, -0.2025, -0.2677, ..., -0.3340, 0.0166, -0.2638], + ..., + [-0.2279, -0.2912, -0.0144, ..., -0.2857, -0.1674, -0.5089], + [ 0.1658, 0.0384, -0.1747, ..., -0.2780, 0.0546, -0.2538], + [ 0.1739, 0.1089, -0.1690, ..., 0.1387, -0.2763, -0.3424]], + device='cuda:0'), grad: tensor([[1.8626e-09, 0.0000e+00, 1.8161e-08, ..., 2.7008e-08, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 4.1910e-09, 0.0000e+00, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 2.7940e-09, ..., 3.7253e-09, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0432, -0.0425, -0.0072, -0.0058, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0159, -0.0439], device='cuda:0'), grad: tensor([ 6.5193e-08, 4.1910e-09, 1.8626e-09, -2.2352e-08, 4.6566e-10, + 2.9802e-08, -9.2667e-08, -3.2596e-09, 1.0245e-08, 1.0710e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 217.82, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4235 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.24 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.1331, 0.0200, 0.1694, ..., 0.0079, -0.1410, -0.1132], + [-0.2546, -0.2556, -0.2347, ..., 0.1226, 0.0719, 0.4424], + [-0.2973, -0.2025, -0.2677, ..., -0.3340, 0.0165, -0.2639], + ..., + [-0.2279, -0.2912, -0.0144, ..., -0.2857, -0.1674, -0.5090], + [ 0.1658, 0.0384, -0.1747, ..., -0.2780, 0.0546, -0.2538], + [ 0.1739, 0.1089, -0.1690, ..., 0.1387, -0.2764, -0.3425]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 1.3970e-09, ..., 4.6566e-10, + 1.3970e-09, 4.6566e-10], + [ 1.3970e-09, 2.3283e-09, 1.3970e-09, ..., -1.8626e-09, + 4.6566e-10, -4.1910e-09], + [ 6.5193e-09, 7.4506e-09, 1.0710e-08, ..., 4.6566e-10, + 1.4901e-08, 4.6566e-10], + ..., + [ 4.6566e-10, 9.3132e-10, 9.3132e-10, ..., 4.6566e-10, + 1.3970e-09, 4.6566e-10], + [-1.6764e-08, -2.0023e-08, -2.8405e-08, ..., 2.3283e-09, + -3.8184e-08, 2.7940e-09], + [ 9.3132e-10, 1.3970e-09, 2.7940e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0432, -0.0425, -0.0073, -0.0057, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0158, -0.0439], device='cuda:0'), grad: tensor([ 1.1642e-08, 2.0489e-08, 1.1036e-07, -3.3062e-08, 4.1910e-09, + 6.4727e-08, 7.1246e-08, -1.8626e-08, -2.5518e-07, 2.2817e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 217.55, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4146 re_mapping 0.0023 re_causal 0.0075 /// teacc 99.26 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.1331, 0.0200, 0.1694, ..., 0.0078, -0.1410, -0.1132], + [-0.2547, -0.2556, -0.2347, ..., 0.1226, 0.0719, 0.4425], + [-0.2974, -0.2026, -0.2678, ..., -0.3341, 0.0165, -0.2639], + ..., + [-0.2280, -0.2913, -0.0144, ..., -0.2857, -0.1674, -0.5091], + [ 0.1657, 0.0383, -0.1747, ..., -0.2781, 0.0546, -0.2538], + [ 0.1739, 0.1089, -0.1690, ..., 0.1388, -0.2764, -0.3425]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., -5.5879e-09, + -9.3132e-10, -8.8476e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 2.3283e-09], + ..., + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, + 4.6566e-10, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [-9.3132e-10, -4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0432, -0.0426, -0.0073, -0.0056, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0157, -0.0439], device='cuda:0'), grad: tensor([ 3.2596e-09, -1.3039e-08, -7.9162e-09, -2.3283e-09, 4.6566e-10, + 3.7253e-09, 4.6566e-10, 1.6764e-08, 4.6566e-10, 4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 217.58, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4011 re_mapping 0.0023 re_causal 0.0073 /// teacc 99.25 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.1331, 0.0200, 0.1695, ..., 0.0079, -0.1410, -0.1132], + [-0.2547, -0.2556, -0.2348, ..., 0.1226, 0.0719, 0.4425], + [-0.2974, -0.2026, -0.2678, ..., -0.3341, 0.0165, -0.2639], + ..., + [-0.2280, -0.2913, -0.0144, ..., -0.2857, -0.1674, -0.5091], + [ 0.1657, 0.0383, -0.1748, ..., -0.2781, 0.0547, -0.2539], + [ 0.1740, 0.1089, -0.1691, ..., 0.1388, -0.2765, -0.3425]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.0536e-09, + 2.3283e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.0978e-08, + -1.8161e-08, -2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.3970e-09, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 3.7253e-09, + 1.8626e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4680e-08, + 1.0710e-08, 9.7789e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 4.6566e-10]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0432, -0.0426, -0.0073, -0.0056, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0158, -0.0439], device='cuda:0'), grad: tensor([ 1.7229e-08, -9.0338e-08, 3.6787e-08, 2.7474e-08, 1.3970e-09, + -1.5367e-08, 1.3970e-09, -4.9360e-08, 7.1246e-08, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 217.58, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4230 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.25 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.1331, 0.0201, 0.1696, ..., 0.0079, -0.1410, -0.1132], + [-0.2548, -0.2556, -0.2348, ..., 0.1226, 0.0720, 0.4426], + [-0.2974, -0.2026, -0.2679, ..., -0.3341, 0.0165, -0.2640], + ..., + [-0.2281, -0.2913, -0.0144, ..., -0.2857, -0.1675, -0.5091], + [ 0.1657, 0.0383, -0.1748, ..., -0.2781, 0.0547, -0.2539], + [ 0.1740, 0.1089, -0.1691, ..., 0.1388, -0.2766, -0.3425]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 3.2596e-09, 2.3283e-09, ..., -4.6566e-10, + 1.3039e-08, -1.8626e-09], + [ 4.6566e-10, 4.1910e-09, 3.2596e-09, ..., 4.6566e-10, + 1.8626e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 9.3132e-10, 7.4506e-09, ..., 0.0000e+00, + 6.0536e-09, 4.6566e-10], + [ 3.2596e-09, -5.1223e-09, -6.9849e-09, ..., 4.1910e-09, + -4.0047e-08, 0.0000e+00], + [-1.2107e-08, -8.8476e-09, 0.0000e+00, ..., -1.7695e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0432, -0.0426, -0.0073, -0.0056, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0158, -0.0439], device='cuda:0'), grad: tensor([ 2.3283e-09, 5.7276e-08, 7.0315e-08, 5.1223e-09, -1.4435e-08, + 2.2817e-08, 6.5193e-09, 2.5146e-08, -1.5087e-07, -2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 217.69, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4210 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.27 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.1331, 0.0201, 0.1696, ..., 0.0079, -0.1411, -0.1132], + [-0.2548, -0.2557, -0.2348, ..., 0.1226, 0.0720, 0.4426], + [-0.2975, -0.2026, -0.2680, ..., -0.3342, 0.0165, -0.2640], + ..., + [-0.2281, -0.2913, -0.0144, ..., -0.2857, -0.1675, -0.5092], + [ 0.1658, 0.0383, -0.1748, ..., -0.2781, 0.0547, -0.2540], + [ 0.1740, 0.1089, -0.1691, ..., 0.1388, -0.2767, -0.3425]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-09, + -9.3132e-10, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [-9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + -4.6566e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0432, -0.0426, -0.0073, -0.0056, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0158, -0.0440], device='cuda:0'), grad: tensor([ 1.8626e-09, -8.3819e-09, 7.9162e-09, 4.6566e-10, -1.8626e-09, + 2.1886e-08, -1.3970e-08, 1.8626e-09, -1.3970e-09, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 217.52, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4305 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.24 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.1331, 0.0201, 0.1698, ..., 0.0079, -0.1411, -0.1132], + [-0.2549, -0.2557, -0.2349, ..., 0.1227, 0.0720, 0.4427], + [-0.2975, -0.2026, -0.2681, ..., -0.3342, 0.0164, -0.2640], + ..., + [-0.2281, -0.2913, -0.0144, ..., -0.2858, -0.1675, -0.5093], + [ 0.1657, 0.0383, -0.1749, ..., -0.2782, 0.0547, -0.2540], + [ 0.1740, 0.1089, -0.1692, ..., 0.1389, -0.2767, -0.3426]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + -9.3132e-10, -4.6566e-09], + [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 9.3132e-10, 9.3132e-10], + ..., + [ 3.7253e-09, 1.3970e-09, 0.0000e+00, ..., 3.2596e-09, + 2.3283e-09, 5.5879e-09], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 4.1910e-09, + -5.1223e-09, 0.0000e+00], + [-1.5832e-08, -6.0536e-09, 0.0000e+00, ..., -1.3039e-08, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0431, -0.0426, -0.0073, -0.0056, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 2.3283e-09, 8.3819e-09, 5.1223e-09, 1.8161e-08, 1.1176e-08, + -2.5146e-08, 1.9558e-08, 1.6764e-08, -6.0536e-09, -4.6100e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 217.44, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0022 re_causal 0.0074 /// teacc 99.27 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.1331, 0.0202, 0.1698, ..., 0.0079, -0.1411, -0.1132], + [-0.2549, -0.2557, -0.2349, ..., 0.1227, 0.0720, 0.4428], + [-0.2976, -0.2026, -0.2682, ..., -0.3343, 0.0164, -0.2640], + ..., + [-0.2282, -0.2914, -0.0144, ..., -0.2858, -0.1675, -0.5093], + [ 0.1657, 0.0383, -0.1749, ..., -0.2782, 0.0548, -0.2540], + [ 0.1740, 0.1089, -0.1692, ..., 0.1389, -0.2768, -0.3426]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -4.6566e-10, 0.0000e+00, ..., -3.2596e-09, + -1.8626e-09, -1.1642e-08], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 1.3970e-09], + ..., + [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 2.7940e-09], + [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 2.3283e-09], + [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0431, -0.0425, -0.0074, -0.0056, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 1.8626e-09, -1.5367e-08, -3.3993e-08, -6.5193e-09, 6.0536e-09, + -1.8626e-09, 1.1176e-08, 4.5169e-08, 1.0710e-08, -5.5879e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 217.63, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4295 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.27 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.1331, 0.0202, 0.1699, ..., 0.0079, -0.1411, -0.1132], + [-0.2550, -0.2558, -0.2349, ..., 0.1227, 0.0721, 0.4429], + [-0.2976, -0.2026, -0.2682, ..., -0.3344, 0.0164, -0.2641], + ..., + [-0.2283, -0.2914, -0.0144, ..., -0.2858, -0.1676, -0.5094], + [ 0.1657, 0.0383, -0.1749, ..., -0.2782, 0.0548, -0.2541], + [ 0.1740, 0.1089, -0.1692, ..., 0.1389, -0.2768, -0.3426]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -9.3132e-10, -2.3283e-09, ..., 0.0000e+00, + 1.3970e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.2352e-08, + -1.4435e-08, -4.0513e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.6764e-08, + 1.1642e-08, 2.9337e-08], + ..., + [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 2.3283e-09, + 1.8626e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 4.6566e-10, 1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0430, -0.0425, -0.0074, -0.0055, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([-4.6566e-10, -7.4506e-08, 8.7544e-08, -1.2573e-08, 3.7253e-09, + 6.0536e-09, 8.3819e-09, -2.7940e-08, 7.9162e-09, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 217.50, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4132 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.27 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.1331, 0.0202, 0.1700, ..., 0.0079, -0.1411, -0.1132], + [-0.2550, -0.2558, -0.2349, ..., 0.1228, 0.0721, 0.4430], + [-0.2977, -0.2026, -0.2682, ..., -0.3344, 0.0164, -0.2641], + ..., + [-0.2283, -0.2914, -0.0144, ..., -0.2859, -0.1676, -0.5095], + [ 0.1657, 0.0383, -0.1749, ..., -0.2783, 0.0548, -0.2541], + [ 0.1740, 0.1090, -0.1693, ..., 0.1390, -0.2768, -0.3427]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., -4.6566e-10, + 4.6566e-10, -5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 9.3132e-10], + ..., + [ 1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 2.7940e-09], + [ 4.1910e-09, 9.3132e-10, 0.0000e+00, ..., 3.2596e-09, + -3.7253e-09, 0.0000e+00], + [-5.1223e-09, -2.7940e-09, 0.0000e+00, ..., -4.1910e-09, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0430, -0.0425, -0.0074, -0.0055, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0157, -0.0439], device='cuda:0'), grad: tensor([ 9.3132e-10, 7.9162e-09, -6.2864e-08, -2.2817e-08, 6.5193e-09, + 3.2596e-08, 3.2596e-09, 4.5169e-08, 7.9162e-09, -9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 217.56, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4291 re_mapping 0.0022 re_causal 0.0074 /// teacc 99.26 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.1331, 0.0202, 0.1700, ..., 0.0079, -0.1412, -0.1133], + [-0.2551, -0.2558, -0.2350, ..., 0.1228, 0.0721, 0.4431], + [-0.2977, -0.2027, -0.2683, ..., -0.3344, 0.0163, -0.2641], + ..., + [-0.2284, -0.2915, -0.0144, ..., -0.2859, -0.1676, -0.5096], + [ 0.1657, 0.0383, -0.1749, ..., -0.2783, 0.0548, -0.2542], + [ 0.1740, 0.1090, -0.1693, ..., 0.1390, -0.2769, -0.3427]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -4.6566e-10, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [-1.3970e-09, -4.6566e-10, 9.3132e-10, ..., -2.7940e-09, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0430, -0.0425, -0.0074, -0.0055, 0.0005, 0.0063, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 9.3132e-10, -3.7253e-09, 3.7253e-09, -1.1176e-08, 5.1223e-09, + 1.0710e-08, 1.3970e-09, 4.1910e-09, 1.8626e-09, -1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 217.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3983 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.25 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.1332, 0.0203, 0.1701, ..., 0.0080, -0.1412, -0.1133], + [-0.2551, -0.2559, -0.2350, ..., 0.1228, 0.0721, 0.4433], + [-0.2977, -0.2027, -0.2683, ..., -0.3345, 0.0163, -0.2642], + ..., + [-0.2284, -0.2915, -0.0144, ..., -0.2859, -0.1677, -0.5098], + [ 0.1657, 0.0383, -0.1750, ..., -0.2784, 0.0549, -0.2542], + [ 0.1741, 0.1090, -0.1693, ..., 0.1391, -0.2770, -0.3427]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 1.3970e-09, ..., 1.8626e-09, + 4.6566e-10, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 4.6566e-10, ..., 0.0000e+00, + 1.3970e-09, -3.7253e-09], + [ 2.3283e-09, 9.3132e-10, 4.6566e-10, ..., 1.8626e-09, + 1.8626e-09, 1.3970e-09], + ..., + [ 1.8626e-09, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 1.8626e-09, 9.3132e-10], + [-1.8626e-08, -1.4435e-08, 4.6566e-10, ..., -1.0245e-08, + -1.7229e-08, 2.7940e-09], + [ 1.3970e-09, 1.3970e-09, 4.6566e-10, ..., 1.8626e-09, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0429, -0.0425, -0.0075, -0.0055, 0.0005, 0.0063, 0.0074, -0.0073, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 1.2107e-08, 1.5367e-08, 3.0268e-08, 4.7032e-08, 3.7253e-09, + 5.2620e-08, -5.7742e-08, -1.0710e-08, -1.0105e-07, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 217.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3898 re_mapping 0.0022 re_causal 0.0073 /// teacc 99.25 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.1332, 0.0203, 0.1701, ..., 0.0080, -0.1412, -0.1133], + [-0.2552, -0.2559, -0.2350, ..., 0.1228, 0.0721, 0.4433], + [-0.2978, -0.2027, -0.2684, ..., -0.3346, 0.0163, -0.2642], + ..., + [-0.2285, -0.2915, -0.0144, ..., -0.2859, -0.1677, -0.5098], + [ 0.1657, 0.0383, -0.1749, ..., -0.2784, 0.0549, -0.2543], + [ 0.1741, 0.1090, -0.1693, ..., 0.1391, -0.2770, -0.3428]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.5146e-08, + -1.7229e-08, -5.3085e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.4214e-08, + 1.6764e-08, 5.2620e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-9.3132e-10, -4.6566e-10, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0429, -0.0426, -0.0075, -0.0054, 0.0005, 0.0062, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 0.0000e+00, -1.1548e-07, 2.3283e-09, -2.7940e-08, 3.7253e-09, + 2.9337e-08, 9.3132e-10, 1.1595e-07, 1.8626e-09, -2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 217.79, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4251 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.24 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.1332, 0.0203, 0.1701, ..., 0.0080, -0.1412, -0.1133], + [-0.2552, -0.2559, -0.2351, ..., 0.1229, 0.0722, 0.4435], + [-0.2978, -0.2027, -0.2684, ..., -0.3347, 0.0163, -0.2643], + ..., + [-0.2285, -0.2916, -0.0144, ..., -0.2860, -0.1678, -0.5100], + [ 0.1657, 0.0382, -0.1749, ..., -0.2784, 0.0549, -0.2543], + [ 0.1741, 0.1090, -0.1694, ..., 0.1391, -0.2770, -0.3428]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.6566e-10, -2.7940e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 2.3283e-09, 1.3970e-09, ..., -2.7940e-09, + -2.3283e-09, -1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 5.1223e-09], + ..., + [ 7.9162e-09, 4.6566e-09, 0.0000e+00, ..., 9.3132e-09, + 1.3970e-09, 9.3132e-09], + [ 4.6566e-09, 3.2596e-09, 4.6566e-10, ..., 6.0536e-09, + 0.0000e+00, 9.3132e-10], + [-3.6322e-08, -2.0023e-08, 8.3819e-09, ..., -3.6322e-08, + 6.5193e-09, 9.3132e-10]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0429, -0.0426, -0.0075, -0.0053, 0.0005, 0.0062, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([-8.3819e-09, -9.7789e-09, 1.3970e-08, 4.1910e-09, 3.4459e-08, + 6.9849e-09, 2.3283e-09, 2.4214e-08, 2.1420e-08, -8.3353e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 217.63, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4032 re_mapping 0.0022 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.1332, 0.0203, 0.1702, ..., 0.0079, -0.1412, -0.1133], + [-0.2553, -0.2559, -0.2351, ..., 0.1229, 0.0722, 0.4436], + [-0.2979, -0.2027, -0.2684, ..., -0.3347, 0.0162, -0.2643], + ..., + [-0.2286, -0.2916, -0.0144, ..., -0.2860, -0.1678, -0.5101], + [ 0.1657, 0.0382, -0.1750, ..., -0.2785, 0.0550, -0.2544], + [ 0.1741, 0.1090, -0.1694, ..., 0.1391, -0.2770, -0.3428]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-2.7940e-09, -3.7253e-09, 4.6566e-10, ..., 4.6566e-10, + -1.8626e-09, 0.0000e+00], + [-1.3970e-09, 0.0000e+00, 9.3132e-10, ..., -3.2596e-09, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0429, -0.0426, -0.0075, -0.0053, 0.0005, 0.0062, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.1910e-09, 3.2596e-09, 1.8626e-09, 4.1910e-09, + 1.8626e-08, -2.0023e-08, 1.8626e-09, -7.9162e-09, -3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 217.72, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4205 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.1332, 0.0203, 0.1702, ..., 0.0079, -0.1413, -0.1134], + [-0.2553, -0.2560, -0.2351, ..., 0.1229, 0.0722, 0.4437], + [-0.2979, -0.2027, -0.2685, ..., -0.3348, 0.0162, -0.2644], + ..., + [-0.2286, -0.2916, -0.0144, ..., -0.2860, -0.1678, -0.5102], + [ 0.1657, 0.0382, -0.1750, ..., -0.2785, 0.0550, -0.2544], + [ 0.1741, 0.1090, -0.1694, ..., 0.1392, -0.2771, -0.3428]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 1.3970e-09, 0.0000e+00, ..., -4.0047e-08, + -9.3132e-10, -4.7497e-08], + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 1.3970e-09, + 4.6566e-10, 1.8626e-09], + ..., + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 4.0978e-08, + 0.0000e+00, 4.5635e-08], + [ 9.3132e-10, 2.3283e-09, 4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [-7.9162e-09, -1.8626e-09, 4.6566e-10, ..., -1.2107e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0430, -0.0426, -0.0075, -0.0053, 0.0005, 0.0062, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 1.8626e-09, -1.8720e-07, 6.9849e-09, 4.9826e-08, 6.5193e-09, + -7.4971e-08, 1.5832e-08, 1.9278e-07, 1.0245e-08, -1.2573e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 217.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4151 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.22 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.1332, 0.0204, 0.1703, ..., 0.0080, -0.1413, -0.1134], + [-0.2554, -0.2560, -0.2352, ..., 0.1229, 0.0723, 0.4438], + [-0.2980, -0.2028, -0.2685, ..., -0.3348, 0.0162, -0.2644], + ..., + [-0.2286, -0.2916, -0.0144, ..., -0.2861, -0.1679, -0.5103], + [ 0.1657, 0.0382, -0.1750, ..., -0.2786, 0.0550, -0.2544], + [ 0.1741, 0.1090, -0.1694, ..., 0.1392, -0.2772, -0.3428]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., -4.0047e-08, + -5.1223e-09, -4.5169e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 1.3970e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 4.6566e-10, 4.1910e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.0734e-08, + 4.6566e-09, 3.3993e-08]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0429, -0.0426, -0.0075, -0.0053, 0.0005, 0.0062, 0.0074, -0.0072, + 0.0157, -0.0440], device='cuda:0'), grad: tensor([ 3.2596e-09, -1.2061e-07, 6.0536e-09, -7.9162e-09, -4.1910e-09, + 3.7253e-09, 1.7229e-08, 5.1223e-09, 3.7253e-09, 1.0571e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 217.55, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4060 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.1332, 0.0204, 0.1703, ..., 0.0080, -0.1413, -0.1134], + [-0.2554, -0.2561, -0.2352, ..., 0.1230, 0.0723, 0.4439], + [-0.2980, -0.2028, -0.2686, ..., -0.3349, 0.0162, -0.2644], + ..., + [-0.2287, -0.2917, -0.0144, ..., -0.2861, -0.1679, -0.5103], + [ 0.1657, 0.0382, -0.1751, ..., -0.2786, 0.0550, -0.2545], + [ 0.1741, 0.1090, -0.1695, ..., 0.1393, -0.2773, -0.3429]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -9.3132e-10, -2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 1.8626e-09], + [-2.3283e-09, 9.3132e-10, -4.6566e-10, ..., 0.0000e+00, + -9.3132e-10, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0429, -0.0426, -0.0075, -0.0052, 0.0005, 0.0062, 0.0074, -0.0071, + 0.0156, -0.0440], device='cuda:0'), grad: tensor([ 0.0000e+00, 5.1223e-09, -7.4506e-09, 9.3132e-10, 4.6566e-10, + -6.5193e-09, 7.4506e-09, 6.0536e-09, -3.7253e-09, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 217.38, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.1332, 0.0204, 0.1704, ..., 0.0080, -0.1414, -0.1134], + [-0.2555, -0.2561, -0.2352, ..., 0.1230, 0.0723, 0.4440], + [-0.2981, -0.2028, -0.2686, ..., -0.3350, 0.0162, -0.2645], + ..., + [-0.2287, -0.2917, -0.0144, ..., -0.2861, -0.1679, -0.5104], + [ 0.1657, 0.0382, -0.1751, ..., -0.2787, 0.0551, -0.2546], + [ 0.1742, 0.1090, -0.1695, ..., 0.1393, -0.2773, -0.3429]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, + -6.5193e-09, -2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, + 4.6566e-09, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0429, -0.0426, -0.0075, -0.0052, 0.0005, 0.0062, 0.0074, -0.0071, + 0.0156, -0.0440], device='cuda:0'), grad: tensor([ 9.3132e-10, -3.5856e-08, 3.7253e-09, 1.3970e-09, 3.7253e-09, + 4.6566e-09, -2.7940e-09, 2.4680e-08, 1.3970e-09, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 217.58, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4387 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.1333, 0.0205, 0.1705, ..., 0.0080, -0.1414, -0.1134], + [-0.2556, -0.2562, -0.2353, ..., 0.1230, 0.0723, 0.4441], + [-0.2981, -0.2028, -0.2687, ..., -0.3350, 0.0162, -0.2645], + ..., + [-0.2288, -0.2917, -0.0145, ..., -0.2861, -0.1680, -0.5104], + [ 0.1657, 0.0382, -0.1752, ..., -0.2787, 0.0551, -0.2547], + [ 0.1742, 0.1091, -0.1696, ..., 0.1394, -0.2774, -0.3429]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3970e-09, -1.6298e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + -4.6566e-10, -2.5611e-09], + [ 0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 1.8626e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 1.1642e-09, 1.3970e-09, ..., 1.6298e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0428, -0.0426, -0.0075, -0.0052, 0.0005, 0.0062, 0.0074, -0.0071, + 0.0156, -0.0440], device='cuda:0'), grad: tensor([-6.5193e-09, 2.3283e-10, 1.3271e-08, 3.9581e-09, 4.6566e-10, + -1.6298e-09, -6.9849e-10, -1.6531e-08, 3.4925e-09, 6.7521e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 217.61, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4204 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.1333, 0.0205, 0.1705, ..., 0.0080, -0.1414, -0.1134], + [-0.2556, -0.2562, -0.2353, ..., 0.1230, 0.0724, 0.4442], + [-0.2981, -0.2028, -0.2687, ..., -0.3351, 0.0161, -0.2646], + ..., + [-0.2288, -0.2918, -0.0145, ..., -0.2862, -0.1680, -0.5105], + [ 0.1657, 0.0382, -0.1752, ..., -0.2788, 0.0551, -0.2547], + [ 0.1742, 0.1091, -0.1696, ..., 0.1394, -0.2775, -0.3429]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -2.0955e-09, -6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 6.9849e-10, 2.5611e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 2.0955e-09], + [-1.1642e-09, -4.6566e-10, 0.0000e+00, ..., -1.1642e-09, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0428, -0.0426, -0.0075, -0.0052, 0.0006, 0.0062, 0.0073, -0.0071, + 0.0156, -0.0440], device='cuda:0'), grad: tensor([ 1.1642e-09, 1.0710e-08, 8.2422e-07, 2.0955e-09, 4.1910e-09, + 3.2596e-09, 2.3283e-09, -8.4750e-07, 7.6834e-09, -2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 217.42, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4055 re_mapping 0.0021 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.1332, 0.0206, 0.1708, ..., 0.0082, -0.1414, -0.1134], + [-0.2557, -0.2563, -0.2354, ..., 0.1231, 0.0724, 0.4444], + [-0.2982, -0.2028, -0.2688, ..., -0.3351, 0.0161, -0.2646], + ..., + [-0.2289, -0.2918, -0.0145, ..., -0.2862, -0.1681, -0.5108], + [ 0.1657, 0.0382, -0.1753, ..., -0.2788, 0.0552, -0.2548], + [ 0.1742, 0.1091, -0.1697, ..., 0.1394, -0.2776, -0.3430]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 1.8626e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -6.5193e-09, + -4.8894e-09, -8.6147e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 1.1642e-09, 2.5611e-09], + ..., + [ 2.3283e-09, 2.3283e-10, 0.0000e+00, ..., 3.9581e-09, + 6.9849e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 1.8626e-09], + [-1.6298e-09, 0.0000e+00, 4.6566e-10, ..., -1.8626e-09, + 6.9849e-10, 6.9849e-10]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0425, -0.0426, -0.0075, -0.0052, 0.0006, 0.0062, 0.0073, -0.0072, + 0.0156, -0.0440], device='cuda:0'), grad: tensor([ 5.1223e-09, -1.0710e-08, 1.3970e-08, -2.5611e-09, 3.7253e-09, + 9.5461e-09, 2.7940e-09, -3.8184e-08, 5.8208e-09, 2.5611e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 217.61, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4363 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.1332, 0.0207, 0.1708, ..., 0.0082, -0.1415, -0.1135], + [-0.2557, -0.2563, -0.2354, ..., 0.1231, 0.0725, 0.4445], + [-0.2982, -0.2029, -0.2688, ..., -0.3352, 0.0161, -0.2647], + ..., + [-0.2289, -0.2918, -0.0145, ..., -0.2863, -0.1682, -0.5109], + [ 0.1657, 0.0382, -0.1753, ..., -0.2789, 0.0552, -0.2548], + [ 0.1742, 0.1091, -0.1698, ..., 0.1394, -0.2777, -0.3430]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.4901e-08, -1.7229e-08, ..., -1.4203e-08, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 9.3132e-10, 1.3970e-09, ..., -2.5611e-09, + -1.8626e-09, -1.3737e-08], + [ 0.0000e+00, 6.9849e-10, 6.9849e-10, ..., 1.1642e-09, + 4.6566e-10, 2.3283e-09], + ..., + [ 0.0000e+00, 4.6566e-10, -2.3283e-10, ..., 1.1642e-09, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 6.9849e-10, 9.3132e-10, ..., 1.6298e-09, + 4.6566e-10, 3.7253e-09], + [ 0.0000e+00, 4.6566e-09, 6.9849e-09, ..., 4.6566e-09, + 9.3132e-10, 2.3283e-10]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0425, -0.0426, -0.0075, -0.0052, 0.0006, 0.0062, 0.0073, -0.0072, + 0.0156, -0.0441], device='cuda:0'), grad: tensor([-6.5891e-08, -1.7695e-08, 2.2119e-08, 1.4203e-08, 1.3970e-09, + 2.3050e-08, 9.7789e-09, -1.5832e-08, 1.1409e-08, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 217.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4438 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.1332, 0.0207, 0.1708, ..., 0.0082, -0.1415, -0.1135], + [-0.2558, -0.2563, -0.2355, ..., 0.1232, 0.0725, 0.4448], + [-0.2983, -0.2029, -0.2688, ..., -0.3353, 0.0160, -0.2647], + ..., + [-0.2290, -0.2919, -0.0145, ..., -0.2863, -0.1682, -0.5110], + [ 0.1657, 0.0382, -0.1753, ..., -0.2789, 0.0552, -0.2549], + [ 0.1742, 0.1091, -0.1699, ..., 0.1395, -0.2778, -0.3431]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-10, -1.1642e-09, ..., -1.3970e-09, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3970e-09, + -1.3970e-09, -1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 6.9849e-10, 4.8894e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, + 4.6566e-10, 1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 2.0955e-09, ..., 6.9849e-10, + 1.3970e-09, 2.3283e-10]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0425, -0.0426, -0.0075, -0.0052, 0.0006, 0.0062, 0.0073, -0.0072, + 0.0156, -0.0441], device='cuda:0'), grad: tensor([-3.2596e-09, -1.1176e-08, 4.4238e-09, 1.1642e-09, -1.0710e-08, + 1.3970e-09, 2.0955e-09, 3.4925e-09, 3.7253e-09, 9.7789e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 217.33, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4348 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.24 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.1332, 0.0207, 0.1709, ..., 0.0082, -0.1415, -0.1135], + [-0.2558, -0.2564, -0.2355, ..., 0.1233, 0.0725, 0.4449], + [-0.2983, -0.2029, -0.2689, ..., -0.3353, 0.0160, -0.2648], + ..., + [-0.2290, -0.2919, -0.0145, ..., -0.2864, -0.1683, -0.5111], + [ 0.1657, 0.0382, -0.1754, ..., -0.2790, 0.0553, -0.2549], + [ 0.1742, 0.1090, -0.1700, ..., 0.1395, -0.2780, -0.3431]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.1642e-09, -2.3283e-09, ..., -1.3970e-09, + 0.0000e+00, 2.3283e-10], + [ 2.7940e-09, 1.6298e-09, 2.3283e-10, ..., 3.0268e-09, + -1.3970e-09, -4.1910e-09], + [ 4.6566e-10, 4.6566e-10, 2.3283e-10, ..., 4.6566e-10, + 2.3283e-10, 4.6566e-10], + ..., + [ 6.5193e-09, 3.0268e-09, 0.0000e+00, ..., 1.0012e-08, + 4.6566e-10, 1.1642e-09], + [ 1.0710e-08, 7.4506e-09, 2.3283e-10, ..., 4.6566e-10, + 3.0268e-09, 2.3283e-10], + [-1.5367e-08, -5.8208e-09, 2.3283e-10, ..., -2.3283e-08, + 4.6566e-10, 2.3283e-10]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0426, -0.0427, -0.0076, -0.0051, 0.0007, 0.0062, 0.0073, -0.0070, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([-5.5879e-09, 1.0012e-08, 3.4925e-09, -2.0023e-08, 3.2131e-08, + -1.5832e-08, 1.2573e-08, 1.6764e-08, 2.8638e-08, -5.3085e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 217.94, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3991 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.23 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.1332, 0.0207, 0.1709, ..., 0.0082, -0.1415, -0.1135], + [-0.2559, -0.2564, -0.2356, ..., 0.1233, 0.0726, 0.4450], + [-0.2984, -0.2029, -0.2689, ..., -0.3354, 0.0160, -0.2648], + ..., + [-0.2291, -0.2919, -0.0145, ..., -0.2864, -0.1683, -0.5113], + [ 0.1657, 0.0382, -0.1754, ..., -0.2790, 0.0553, -0.2550], + [ 0.1743, 0.1091, -0.1701, ..., 0.1395, -0.2781, -0.3432]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, -1.6298e-09, ..., -2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., -6.9849e-10, + 3.7253e-09, -2.3283e-09], + [ 0.0000e+00, 2.3283e-10, 4.6566e-10, ..., 4.6566e-10, + 2.3283e-10, 1.3970e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 4.6566e-10, 1.6298e-09], + [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.7521e-09, 2.3283e-10], + [-1.8626e-09, -2.3283e-10, 0.0000e+00, ..., -2.5611e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0425, -0.0427, -0.0076, -0.0051, 0.0007, 0.0061, 0.0073, -0.0070, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([ 1.1642e-09, 2.1188e-08, -2.1420e-07, 3.1665e-08, 4.8894e-09, + 3.4925e-09, 9.0804e-09, 1.5693e-07, -2.3283e-09, -2.5611e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 217.59, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4324 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.22 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.1333, 0.0208, 0.1709, ..., 0.0082, -0.1416, -0.1135], + [-0.2559, -0.2564, -0.2356, ..., 0.1233, 0.0726, 0.4451], + [-0.2984, -0.2029, -0.2689, ..., -0.3354, 0.0160, -0.2649], + ..., + [-0.2291, -0.2920, -0.0145, ..., -0.2865, -0.1684, -0.5114], + [ 0.1657, 0.0382, -0.1755, ..., -0.2791, 0.0553, -0.2551], + [ 0.1743, 0.1091, -0.1701, ..., 0.1396, -0.2781, -0.3432]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 2.7940e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, -1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-1.1176e-08, -4.6566e-09, 0.0000e+00, ..., -1.5367e-08, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0425, -0.0428, -0.0076, -0.0050, 0.0007, 0.0060, 0.0074, -0.0069, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([ 0.0000e+00, -6.9849e-09, 9.3132e-10, -9.3132e-10, 5.1223e-09, + 1.6764e-08, 1.7229e-08, 7.9162e-09, 3.7253e-09, -3.8650e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 217.66, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4146 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.1333, 0.0208, 0.1709, ..., 0.0082, -0.1416, -0.1135], + [-0.2559, -0.2565, -0.2357, ..., 0.1234, 0.0727, 0.4453], + [-0.2984, -0.2030, -0.2690, ..., -0.3355, 0.0159, -0.2649], + ..., + [-0.2292, -0.2920, -0.0145, ..., -0.2866, -0.1685, -0.5115], + [ 0.1657, 0.0382, -0.1755, ..., -0.2791, 0.0554, -0.2551], + [ 0.1743, 0.1091, -0.1701, ..., 0.1396, -0.2782, -0.3432]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -2.3283e-09, + -1.3970e-09, -3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.3970e-09, + 1.3970e-09, 2.3283e-09], + [ 9.3132e-10, 9.3132e-10, 1.3970e-09, ..., 2.7940e-09, + 4.6566e-10, 4.6566e-10], + [-1.9558e-08, -1.9558e-08, 1.3039e-08, ..., -9.3132e-09, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0426, -0.0428, -0.0076, -0.0050, 0.0006, 0.0060, 0.0074, -0.0070, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([ 2.3283e-09, -1.8626e-09, 9.7789e-09, 4.6566e-10, -1.2573e-08, + 1.2619e-07, -1.2992e-07, 4.6566e-09, 1.3504e-08, -4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 217.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4032 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.22 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.1333, 0.0208, 0.1710, ..., 0.0082, -0.1416, -0.1135], + [-0.2560, -0.2565, -0.2358, ..., 0.1235, 0.0727, 0.4455], + [-0.2985, -0.2030, -0.2690, ..., -0.3355, 0.0159, -0.2649], + ..., + [-0.2292, -0.2920, -0.0145, ..., -0.2866, -0.1685, -0.5117], + [ 0.1657, 0.0382, -0.1756, ..., -0.2792, 0.0554, -0.2552], + [ 0.1744, 0.1092, -0.1702, ..., 0.1397, -0.2782, -0.3433]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [-1.6764e-08, -5.5879e-09, 0.0000e+00, ..., -1.4435e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0426, -0.0428, -0.0076, -0.0050, 0.0006, 0.0060, 0.0074, -0.0070, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.4435e-08, 6.0536e-09, 2.7940e-09, 3.8184e-08, + -2.7940e-09, 1.3970e-09, -5.2154e-08, 1.0245e-08, -1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 217.83, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4177 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.21 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.1333, 0.0209, 0.1710, ..., 0.0082, -0.1416, -0.1135], + [-0.2561, -0.2566, -0.2358, ..., 0.1235, 0.0727, 0.4456], + [-0.2985, -0.2030, -0.2691, ..., -0.3356, 0.0159, -0.2650], + ..., + [-0.2293, -0.2921, -0.0145, ..., -0.2867, -0.1686, -0.5118], + [ 0.1658, 0.0382, -0.1756, ..., -0.2792, 0.0555, -0.2552], + [ 0.1744, 0.1092, -0.1702, ..., 0.1398, -0.2783, -0.3433]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 2.7940e-09, 4.6566e-10, ..., -2.3283e-09, + -1.8626e-09, -6.5193e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 1.3970e-09, 3.7253e-09], + ..., + [ 4.6566e-10, 3.2596e-09, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 2.3283e-09], + [ 0.0000e+00, 4.6566e-09, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [-1.3970e-09, 4.1910e-09, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0425, -0.0428, -0.0075, -0.0050, 0.0006, 0.0060, 0.0074, -0.0070, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([ 4.6566e-09, 6.5193e-09, 6.0536e-09, 6.6590e-08, 8.8476e-09, + -1.4948e-07, -4.6566e-10, 2.1420e-08, 2.3749e-08, 1.8161e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 217.77, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4642 re_mapping 0.0020 re_causal 0.0076 /// teacc 99.22 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.1333, 0.0209, 0.1711, ..., 0.0082, -0.1417, -0.1136], + [-0.2561, -0.2566, -0.2359, ..., 0.1236, 0.0728, 0.4458], + [-0.2985, -0.2030, -0.2691, ..., -0.3356, 0.0159, -0.2651], + ..., + [-0.2294, -0.2921, -0.0145, ..., -0.2868, -0.1687, -0.5120], + [ 0.1658, 0.0382, -0.1757, ..., -0.2793, 0.0555, -0.2553], + [ 0.1744, 0.1093, -0.1703, ..., 0.1399, -0.2784, -0.3433]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10], + [-4.6566e-10, -4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0425, -0.0428, -0.0074, -0.0050, 0.0006, 0.0060, 0.0074, -0.0070, + 0.0155, -0.0441], device='cuda:0'), grad: tensor([ 4.6566e-10, 0.0000e+00, -9.3132e-09, 0.0000e+00, -3.2596e-09, + 1.3970e-09, 9.3132e-10, 5.1223e-09, 3.7253e-09, 5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 217.77, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4212 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.25 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.1333, 0.0210, 0.1712, ..., 0.0082, -0.1417, -0.1136], + [-0.2562, -0.2567, -0.2359, ..., 0.1237, 0.0729, 0.4460], + [-0.2986, -0.2030, -0.2691, ..., -0.3357, 0.0158, -0.2651], + ..., + [-0.2295, -0.2922, -0.0146, ..., -0.2869, -0.1687, -0.5122], + [ 0.1658, 0.0383, -0.1758, ..., -0.2794, 0.0555, -0.2553], + [ 0.1745, 0.1093, -0.1704, ..., 0.1400, -0.2785, -0.3433]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0850e-07, + -4.6566e-10, -2.9057e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0617e-07, + 0.0000e+00, 2.8452e-07], + [-4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0425, -0.0428, -0.0074, -0.0050, 0.0006, 0.0060, 0.0074, -0.0070, + 0.0154, -0.0441], device='cuda:0'), grad: tensor([ 0.0000e+00, -1.0245e-06, 1.3970e-08, 4.1910e-09, 4.6566e-10, + 3.2596e-09, 1.3970e-09, 1.0068e-06, -6.0536e-09, 4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 217.59, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4202 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.23 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.1333, 0.0211, 0.1714, ..., 0.0084, -0.1417, -0.1136], + [-0.2562, -0.2567, -0.2360, ..., 0.1237, 0.0730, 0.4462], + [-0.2986, -0.2030, -0.2692, ..., -0.3358, 0.0158, -0.2652], + ..., + [-0.2295, -0.2922, -0.0146, ..., -0.2870, -0.1688, -0.5124], + [ 0.1658, 0.0382, -0.1759, ..., -0.2795, 0.0556, -0.2554], + [ 0.1745, 0.1094, -0.1705, ..., 0.1400, -0.2785, -0.3434]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., -1.3970e-09, + 4.6566e-10, -1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, 4.6566e-10], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0423, -0.0427, -0.0073, -0.0050, 0.0006, 0.0060, 0.0074, -0.0070, + 0.0154, -0.0441], device='cuda:0'), grad: tensor([ 4.1910e-09, 1.6298e-08, -5.3551e-08, 0.0000e+00, -1.9092e-08, + 2.7940e-09, 4.1910e-09, 1.3970e-08, 1.5367e-08, 1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 217.64, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4117 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.1333, 0.0211, 0.1714, ..., 0.0084, -0.1417, -0.1136], + [-0.2563, -0.2567, -0.2361, ..., 0.1238, 0.0730, 0.4464], + [-0.2986, -0.2031, -0.2692, ..., -0.3359, 0.0158, -0.2652], + ..., + [-0.2296, -0.2923, -0.0146, ..., -0.2871, -0.1689, -0.5125], + [ 0.1658, 0.0383, -0.1760, ..., -0.2796, 0.0556, -0.2555], + [ 0.1746, 0.1094, -0.1705, ..., 0.1401, -0.2786, -0.3434]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 4.6566e-10, ..., -9.3132e-10, + 4.6566e-10, -1.8626e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 2.7940e-09], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 9.3132e-10], + [-1.8626e-09, -4.6566e-10, 4.6566e-10, ..., 9.3132e-10, + -1.3970e-09, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0423, -0.0427, -0.0072, -0.0050, 0.0006, 0.0059, 0.0074, -0.0071, + 0.0153, -0.0441], device='cuda:0'), grad: tensor([ 6.0536e-09, 4.6566e-09, -2.6077e-08, 4.6566e-09, 3.2596e-09, + -2.1886e-08, 8.8476e-09, 2.7474e-08, -4.6566e-10, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 217.85, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4120 re_mapping 0.0020 re_causal 0.0069 /// teacc 99.22 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.1333, 0.0211, 0.1715, ..., 0.0084, -0.1418, -0.1136], + [-0.2563, -0.2568, -0.2361, ..., 0.1239, 0.0730, 0.4466], + [-0.2987, -0.2031, -0.2692, ..., -0.3359, 0.0158, -0.2653], + ..., + [-0.2297, -0.2924, -0.0146, ..., -0.2872, -0.1689, -0.5127], + [ 0.1659, 0.0383, -0.1760, ..., -0.2796, 0.0557, -0.2555], + [ 0.1746, 0.1095, -0.1706, ..., 0.1402, -0.2786, -0.3434]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 9.3132e-10, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, 6.0536e-09, + 0.0000e+00], + [4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, + 0.0000e+00], + ..., + [9.3132e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, 9.3132e-10, + 0.0000e+00], + [9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, + 0.0000e+00], + [4.6566e-10, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, 1.5367e-08, + 0.0000e+00]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0423, -0.0428, -0.0073, -0.0050, 0.0006, 0.0059, 0.0074, -0.0070, + 0.0153, -0.0441], device='cuda:0'), grad: tensor([ 3.7253e-09, 2.5611e-08, 5.1223e-09, -1.4901e-08, -1.1502e-07, + 7.4506e-09, 1.1642e-08, 7.9162e-09, 4.6566e-09, 7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 217.86, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4174 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.24 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.1333, 0.0212, 0.1715, ..., 0.0084, -0.1418, -0.1136], + [-0.2564, -0.2568, -0.2362, ..., 0.1240, 0.0731, 0.4468], + [-0.2987, -0.2031, -0.2693, ..., -0.3360, 0.0157, -0.2654], + ..., + [-0.2298, -0.2924, -0.0146, ..., -0.2872, -0.1690, -0.5128], + [ 0.1659, 0.0383, -0.1760, ..., -0.2797, 0.0557, -0.2556], + [ 0.1747, 0.1096, -0.1706, ..., 0.1404, -0.2787, -0.3434]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., -4.6566e-10, + -1.3970e-09, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 3.2596e-09], + [ 0.0000e+00, 1.3970e-09, 1.1176e-08, ..., 9.3132e-09, + 4.6566e-10, 1.3970e-09], + [-2.7940e-09, -9.3132e-10, 1.8626e-09, ..., -3.7253e-09, + 1.3970e-09, 4.6566e-10]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0423, -0.0428, -0.0072, -0.0050, 0.0006, 0.0059, 0.0074, -0.0070, + 0.0153, -0.0441], device='cuda:0'), grad: tensor([ 4.6566e-09, -1.8626e-09, 4.6566e-09, 2.3283e-09, 6.5193e-09, + 1.6298e-08, -7.9162e-08, 9.3132e-09, 4.7497e-08, -4.1910e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 217.91, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4213 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.25 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.1333, 0.0212, 0.1715, ..., 0.0084, -0.1418, -0.1137], + [-0.2565, -0.2569, -0.2362, ..., 0.1240, 0.0731, 0.4469], + [-0.2988, -0.2031, -0.2693, ..., -0.3360, 0.0157, -0.2654], + ..., + [-0.2299, -0.2925, -0.0146, ..., -0.2873, -0.1690, -0.5129], + [ 0.1659, 0.0382, -0.1762, ..., -0.2798, 0.0557, -0.2557], + [ 0.1748, 0.1096, -0.1706, ..., 0.1405, -0.2787, -0.3435]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 4.6566e-10, ..., -1.3970e-09, + -4.6566e-10, -4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + ..., + [ 0.0000e+00, 2.7940e-09, 6.5193e-09, ..., 9.3132e-10, + 4.6566e-10, 2.3283e-09], + [-6.0536e-09, -6.5193e-09, 0.0000e+00, ..., -1.8626e-09, + -4.1910e-09, 4.6566e-10], + [ 3.2596e-09, 8.8476e-09, 1.3970e-08, ..., 9.3132e-10, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0423, -0.0428, -0.0072, -0.0049, 0.0005, 0.0058, 0.0075, -0.0071, + 0.0152, -0.0441], device='cuda:0'), grad: tensor([ 1.3970e-09, -2.3283e-09, 3.2596e-09, 3.7253e-09, -1.0198e-07, + 7.4506e-09, 7.4506e-09, 3.5390e-08, -2.9337e-08, 8.1956e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 217.64, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4214 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.23 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.1333, 0.0213, 0.1716, ..., 0.0085, -0.1418, -0.1137], + [-0.2565, -0.2569, -0.2363, ..., 0.1241, 0.0732, 0.4472], + [-0.2988, -0.2031, -0.2693, ..., -0.3361, 0.0157, -0.2655], + ..., + [-0.2300, -0.2926, -0.0146, ..., -0.2874, -0.1691, -0.5131], + [ 0.1659, 0.0383, -0.1762, ..., -0.2799, 0.0558, -0.2557], + [ 0.1748, 0.1097, -0.1707, ..., 0.1405, -0.2788, -0.3435]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-1.1176e-08, -1.8626e-09, 0.0000e+00, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0422, -0.0428, -0.0071, -0.0050, 0.0005, 0.0058, 0.0075, -0.0070, + 0.0152, -0.0441], device='cuda:0'), grad: tensor([ 1.3970e-09, 2.7940e-09, 2.7940e-09, -1.3970e-08, 2.3283e-08, + 1.6298e-08, 1.3970e-09, -4.6566e-10, 4.6566e-09, -2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 217.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4425 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.24 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.1333, 0.0214, 0.1717, ..., 0.0085, -0.1419, -0.1137], + [-0.2565, -0.2570, -0.2364, ..., 0.1242, 0.0733, 0.4474], + [-0.2988, -0.2032, -0.2694, ..., -0.3361, 0.0157, -0.2655], + ..., + [-0.2301, -0.2926, -0.0146, ..., -0.2875, -0.1692, -0.5133], + [ 0.1659, 0.0383, -0.1763, ..., -0.2800, 0.0558, -0.2558], + [ 0.1748, 0.1097, -0.1708, ..., 0.1406, -0.2789, -0.3435]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -4.6566e-10, -1.8626e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 1.8626e-09], + ..., + [ 4.6566e-10, 4.6566e-10, -4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 1.3970e-09], + [-2.3283e-09, -1.8626e-09, 9.3132e-10, ..., 1.3970e-09, + -2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0421, -0.0428, -0.0070, -0.0050, 0.0005, 0.0057, 0.0076, -0.0071, + 0.0152, -0.0441], device='cuda:0'), grad: tensor([-3.7253e-09, -4.6566e-10, 2.3283e-09, 5.5879e-09, 0.0000e+00, + 3.7253e-09, 1.8626e-09, 2.7940e-09, -1.3970e-08, 5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 217.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4027 re_mapping 0.0020 re_causal 0.0070 /// teacc 99.22 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.1333, 0.0214, 0.1717, ..., 0.0085, -0.1419, -0.1137], + [-0.2566, -0.2570, -0.2364, ..., 0.1244, 0.0734, 0.4478], + [-0.2988, -0.2032, -0.2694, ..., -0.3362, 0.0157, -0.2656], + ..., + [-0.2302, -0.2927, -0.0146, ..., -0.2877, -0.1694, -0.5137], + [ 0.1659, 0.0383, -0.1763, ..., -0.2801, 0.0559, -0.2558], + [ 0.1748, 0.1097, -0.1708, ..., 0.1406, -0.2789, -0.3436]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 2.3283e-09, 0.0000e+00, ..., 1.8626e-09, + 4.1910e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 9.3132e-10], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + [-6.0536e-09, -4.6566e-09, 0.0000e+00, ..., -1.8626e-09, + -9.3132e-09, 4.6566e-10], + [-6.0536e-09, -3.7253e-09, 0.0000e+00, ..., -6.0536e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0421, -0.0427, -0.0070, -0.0050, 0.0005, 0.0058, 0.0076, -0.0071, + 0.0151, -0.0441], device='cuda:0'), grad: tensor([ 9.3132e-10, 2.1420e-08, 2.7940e-09, 4.6566e-10, 1.3504e-08, + 1.3970e-08, 4.6566e-09, 3.2596e-09, -3.8650e-08, -2.0023e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 217.43, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4351 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.1333, 0.0215, 0.1717, ..., 0.0085, -0.1419, -0.1137], + [-0.2566, -0.2570, -0.2365, ..., 0.1244, 0.0735, 0.4479], + [-0.2989, -0.2032, -0.2694, ..., -0.3363, 0.0156, -0.2657], + ..., + [-0.2303, -0.2927, -0.0146, ..., -0.2878, -0.1695, -0.5138], + [ 0.1660, 0.0383, -0.1764, ..., -0.2801, 0.0559, -0.2559], + [ 0.1749, 0.1097, -0.1709, ..., 0.1407, -0.2790, -0.3437]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.5193e-09, -1.5367e-08, ..., -8.8476e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., -4.6566e-10, + -4.6566e-10, -1.8626e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 2.7940e-09, 3.2596e-09, 1.3970e-09, ..., 4.6566e-10, + 4.6566e-10, 1.3970e-09], + [-2.3283e-09, -1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0421, -0.0427, -0.0069, -0.0050, 0.0005, 0.0058, 0.0077, -0.0072, + 0.0151, -0.0441], device='cuda:0'), grad: tensor([-3.9581e-08, 0.0000e+00, 3.7253e-09, 2.2585e-07, 9.3132e-10, + -2.3050e-07, 3.1199e-08, 1.3970e-08, -7.4506e-09, 7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 217.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4159 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.1333, 0.0215, 0.1718, ..., 0.0085, -0.1419, -0.1137], + [-0.2567, -0.2571, -0.2366, ..., 0.1245, 0.0735, 0.4481], + [-0.2989, -0.2032, -0.2695, ..., -0.3363, 0.0156, -0.2657], + ..., + [-0.2303, -0.2928, -0.0146, ..., -0.2878, -0.1696, -0.5140], + [ 0.1660, 0.0382, -0.1764, ..., -0.2802, 0.0560, -0.2560], + [ 0.1749, 0.1097, -0.1709, ..., 0.1407, -0.2791, -0.3437]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-2.3283e-09, -1.8626e-09, 0.0000e+00, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0421, -0.0427, -0.0068, -0.0051, 0.0005, 0.0058, 0.0077, -0.0072, + 0.0151, -0.0441], device='cuda:0'), grad: tensor([ 4.6566e-10, 6.5193e-09, 5.1223e-09, 1.0710e-08, 4.1910e-09, + -2.7940e-09, 1.3970e-09, -1.4901e-08, 1.8626e-09, -4.1910e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 217.49, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4161 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.1333, 0.0216, 0.1718, ..., 0.0086, -0.1419, -0.1137], + [-0.2567, -0.2571, -0.2366, ..., 0.1246, 0.0736, 0.4483], + [-0.2989, -0.2032, -0.2695, ..., -0.3364, 0.0156, -0.2658], + ..., + [-0.2304, -0.2928, -0.0146, ..., -0.2880, -0.1696, -0.5142], + [ 0.1660, 0.0382, -0.1765, ..., -0.2803, 0.0561, -0.2560], + [ 0.1750, 0.1098, -0.1710, ..., 0.1409, -0.2792, -0.3438]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + -4.6566e-10, -2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0421, -0.0427, -0.0068, -0.0051, 0.0005, 0.0059, 0.0077, -0.0072, + 0.0150, -0.0441], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.8626e-09, -6.0536e-09, 9.3132e-10, 0.0000e+00, + 4.6566e-10, 1.3970e-09, 4.1910e-09, 9.3132e-10, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 217.59, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4016 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.23 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.1333, 0.0216, 0.1718, ..., 0.0085, -0.1419, -0.1137], + [-0.2568, -0.2572, -0.2367, ..., 0.1247, 0.0736, 0.4485], + [-0.2990, -0.2033, -0.2696, ..., -0.3364, 0.0156, -0.2658], + ..., + [-0.2305, -0.2929, -0.0146, ..., -0.2881, -0.1697, -0.5143], + [ 0.1660, 0.0382, -0.1765, ..., -0.2803, 0.0561, -0.2560], + [ 0.1750, 0.1098, -0.1711, ..., 0.1410, -0.2793, -0.3438]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 4.6566e-10], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 9.3132e-10], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 4.6566e-10, + 4.6566e-10], + [0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, + 0.0000e+00]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0421, -0.0427, -0.0068, -0.0052, 0.0005, 0.0059, 0.0077, -0.0073, + 0.0150, -0.0441], device='cuda:0'), grad: tensor([ 1.3970e-08, 1.0710e-08, -4.2375e-08, 9.3132e-10, -2.7940e-09, + -4.6566e-10, 4.6566e-09, 1.9092e-08, 4.6566e-10, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 217.57, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4250 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.25 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.1334, 0.0216, 0.1718, ..., 0.0085, -0.1420, -0.1137], + [-0.2569, -0.2572, -0.2367, ..., 0.1248, 0.0737, 0.4486], + [-0.2990, -0.2033, -0.2696, ..., -0.3365, 0.0155, -0.2659], + ..., + [-0.2306, -0.2930, -0.0146, ..., -0.2882, -0.1698, -0.5144], + [ 0.1660, 0.0382, -0.1766, ..., -0.2804, 0.0562, -0.2561], + [ 0.1751, 0.1099, -0.1712, ..., 0.1411, -0.2794, -0.3439]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -3.2596e-09, + -2.7940e-09, -8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.3283e-09, + 1.8626e-09, 6.9849e-09], + ..., + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 2.3283e-09], + [-4.6566e-10, -4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + -9.3132e-10, 4.6566e-10], + [-1.8626e-09, -2.3283e-09, 0.0000e+00, ..., -2.3283e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0421, -0.0426, -0.0068, -0.0052, 0.0005, 0.0059, 0.0077, -0.0073, + 0.0150, -0.0441], device='cuda:0'), grad: tensor([ 1.8626e-09, -3.2596e-09, -3.1199e-08, 3.7253e-09, 3.7253e-09, + -4.6566e-10, 4.6566e-09, -3.2596e-09, 8.3819e-09, 1.9092e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 217.61, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4142 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.24 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.1334, 0.0216, 0.1719, ..., 0.0085, -0.1420, -0.1138], + [-0.2569, -0.2572, -0.2368, ..., 0.1249, 0.0737, 0.4489], + [-0.2990, -0.2033, -0.2696, ..., -0.3366, 0.0155, -0.2660], + ..., + [-0.2307, -0.2930, -0.0146, ..., -0.2883, -0.1699, -0.5147], + [ 0.1660, 0.0382, -0.1766, ..., -0.2805, 0.0562, -0.2562], + [ 0.1751, 0.1100, -0.1712, ..., 0.1412, -0.2795, -0.3439]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.1223e-09, -9.7789e-09, ..., -9.7789e-09, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, + -4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 3.2596e-09, 6.0536e-09, ..., 6.0536e-09, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0421, -0.0426, -0.0068, -0.0052, 0.0005, 0.0059, 0.0077, -0.0073, + 0.0149, -0.0441], device='cuda:0'), grad: tensor([-3.6787e-08, 4.6566e-09, -4.6566e-09, 2.3283e-09, 6.0536e-09, + 4.6566e-10, 4.1910e-09, 3.2596e-09, 2.6077e-08, 2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 217.58, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4003 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.23 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.1334, 0.0217, 0.1719, ..., 0.0086, -0.1420, -0.1138], + [-0.2570, -0.2573, -0.2368, ..., 0.1250, 0.0738, 0.4490], + [-0.2991, -0.2033, -0.2696, ..., -0.3366, 0.0155, -0.2661], + ..., + [-0.2308, -0.2931, -0.0146, ..., -0.2884, -0.1700, -0.5148], + [ 0.1660, 0.0382, -0.1767, ..., -0.2806, 0.0563, -0.2562], + [ 0.1752, 0.1100, -0.1713, ..., 0.1413, -0.2796, -0.3439]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 2.3283e-10, 6.2864e-09], + [ 0.0000e+00, 2.3283e-10, -1.1642e-09, ..., -4.7032e-08, + -2.0955e-08, -1.3295e-07], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.5134e-08, + 7.2177e-09, 3.6322e-08], + ..., + [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 6.9849e-09, + 9.3132e-10, 3.3295e-08], + [ 0.0000e+00, 6.9849e-10, 1.8626e-09, ..., 2.2352e-08, + 1.1874e-08, 4.2142e-08], + [-1.1642e-09, -6.9849e-10, 4.6566e-10, ..., -9.3132e-10, + 4.6566e-10, 1.8626e-09]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0421, -0.0426, -0.0068, -0.0052, 0.0005, 0.0059, 0.0077, -0.0073, + 0.0149, -0.0441], device='cuda:0'), grad: tensor([ 1.0245e-08, -2.4727e-07, 7.3342e-08, 9.0804e-09, 6.0536e-09, + 9.3132e-09, -3.0268e-09, 5.1456e-08, 1.0175e-07, 2.0955e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 217.87, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4076 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.1334, 0.0217, 0.1720, ..., 0.0086, -0.1420, -0.1138], + [-0.2570, -0.2574, -0.2369, ..., 0.1250, 0.0738, 0.4492], + [-0.2991, -0.2034, -0.2697, ..., -0.3367, 0.0154, -0.2661], + ..., + [-0.2309, -0.2932, -0.0146, ..., -0.2885, -0.1700, -0.5149], + [ 0.1661, 0.0382, -0.1768, ..., -0.2807, 0.0564, -0.2563], + [ 0.1752, 0.1100, -0.1714, ..., 0.1414, -0.2797, -0.3439]], + device='cuda:0'), grad: tensor([[-2.3283e-10, -9.3132e-10, 1.3970e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., -2.5611e-09, + -4.6566e-10, -3.4925e-09], + [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 3.0268e-09, + 1.3970e-09, 2.0955e-09], + ..., + [ 2.3283e-10, 0.0000e+00, -4.6566e-10, ..., 1.8626e-09, + 4.6566e-10, 1.6298e-09], + [-6.9849e-10, -4.6566e-10, 3.2596e-09, ..., 3.4925e-09, + -6.9849e-10, 0.0000e+00], + [-4.6566e-10, -2.3283e-10, 6.2864e-09, ..., 4.6566e-10, + 6.0536e-09, 0.0000e+00]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0421, -0.0426, -0.0068, -0.0052, 0.0005, 0.0059, 0.0078, -0.0073, + 0.0149, -0.0442], device='cuda:0'), grad: tensor([ 6.5193e-09, -1.6298e-09, 1.0012e-08, 1.6298e-09, -1.6065e-08, + 1.8859e-08, -4.7265e-08, 2.0955e-09, 9.3132e-09, 2.0256e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 217.59, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4076 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.1334, 0.0217, 0.1720, ..., 0.0086, -0.1420, -0.1138], + [-0.2571, -0.2574, -0.2369, ..., 0.1252, 0.0739, 0.4495], + [-0.2991, -0.2034, -0.2697, ..., -0.3368, 0.0154, -0.2662], + ..., + [-0.2309, -0.2932, -0.0146, ..., -0.2886, -0.1701, -0.5151], + [ 0.1661, 0.0382, -0.1768, ..., -0.2808, 0.0564, -0.2563], + [ 0.1753, 0.1101, -0.1715, ..., 0.1415, -0.2798, -0.3440]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 4.6566e-10], + [ 1.1642e-09, 9.3132e-10, -4.6566e-10, ..., -6.9849e-10, + -1.3970e-09, -7.2177e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 9.3132e-10, + 1.1642e-09, 3.9581e-09], + ..., + [ 1.3970e-09, 1.1642e-09, 0.0000e+00, ..., 2.0955e-09, + 4.6566e-10, 1.6298e-09], + [-4.8894e-09, -5.3551e-09, 2.3283e-10, ..., -1.3970e-09, + -3.0268e-09, 6.9849e-10], + [-4.4238e-09, -2.5611e-09, 0.0000e+00, ..., -8.8476e-09, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0421, -0.0426, -0.0069, -0.0052, 0.0005, 0.0060, 0.0078, -0.0074, + 0.0148, -0.0442], device='cuda:0'), grad: tensor([ 2.7940e-09, -5.8208e-09, 6.9849e-09, 4.1910e-09, 1.5134e-08, + 5.3551e-09, 2.0955e-09, 8.8476e-09, -1.8394e-08, -1.7229e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 217.58, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4106 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.1334, 0.0218, 0.1720, ..., 0.0086, -0.1421, -0.1138], + [-0.2571, -0.2574, -0.2370, ..., 0.1253, 0.0740, 0.4496], + [-0.2992, -0.2034, -0.2697, ..., -0.3369, 0.0153, -0.2663], + ..., + [-0.2310, -0.2932, -0.0146, ..., -0.2887, -0.1702, -0.5152], + [ 0.1661, 0.0382, -0.1768, ..., -0.2809, 0.0566, -0.2564], + [ 0.1753, 0.1101, -0.1716, ..., 0.1415, -0.2799, -0.3440]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -6.9849e-10, + -1.3970e-09, -5.1223e-09], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 9.3132e-10], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.1642e-09], + [ 2.3283e-10, -2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + -4.6566e-10, 0.0000e+00], + [-1.8626e-08, -1.3039e-08, -9.3132e-10, ..., -2.2119e-08, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0421, -0.0425, -0.0069, -0.0053, 0.0005, 0.0060, 0.0078, -0.0074, + 0.0149, -0.0442], device='cuda:0'), grad: tensor([ 3.2596e-09, 2.4214e-08, 2.5611e-09, 9.5461e-09, 5.8440e-08, + 3.2596e-09, 1.1642e-09, -3.7020e-08, 0.0000e+00, -5.8440e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 217.53, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4341 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.25 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.1334, 0.0218, 0.1721, ..., 0.0086, -0.1421, -0.1138], + [-0.2572, -0.2575, -0.2370, ..., 0.1253, 0.0740, 0.4497], + [-0.2992, -0.2035, -0.2698, ..., -0.3370, 0.0153, -0.2664], + ..., + [-0.2310, -0.2933, -0.0146, ..., -0.2888, -0.1703, -0.5153], + [ 0.1662, 0.0382, -0.1770, ..., -0.2810, 0.0566, -0.2564], + [ 0.1754, 0.1101, -0.1717, ..., 0.1417, -0.2800, -0.3440]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., -6.9849e-10, + -4.6566e-10, -2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 2.3283e-10, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0420, -0.0425, -0.0069, -0.0053, 0.0005, 0.0060, 0.0078, -0.0074, + 0.0149, -0.0442], device='cuda:0'), grad: tensor([ 1.6298e-09, -1.8626e-09, 2.2352e-08, 1.8626e-09, 1.3970e-08, + 6.0536e-09, -1.8859e-08, -2.6543e-08, 4.6566e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 217.45, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3808 re_mapping 0.0020 re_causal 0.0068 /// teacc 99.22 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.1335, 0.0218, 0.1721, ..., 0.0086, -0.1421, -0.1138], + [-0.2573, -0.2575, -0.2371, ..., 0.1254, 0.0740, 0.4498], + [-0.2993, -0.2035, -0.2698, ..., -0.3370, 0.0153, -0.2665], + ..., + [-0.2311, -0.2934, -0.0147, ..., -0.2889, -0.1704, -0.5154], + [ 0.1662, 0.0383, -0.1770, ..., -0.2811, 0.0568, -0.2564], + [ 0.1755, 0.1101, -0.1718, ..., 0.1418, -0.2802, -0.3440]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 2.3283e-10, ..., -4.6566e-10, + -4.6566e-10, -2.3283e-09], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 1.1642e-09], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 1.1642e-09], + [-1.6298e-09, -1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 2.3283e-10], + [ 0.0000e+00, 4.6566e-10, 8.1491e-09, ..., 0.0000e+00, + 2.5611e-09, 2.3283e-10]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0421, -0.0425, -0.0069, -0.0054, 0.0005, 0.0060, 0.0078, -0.0074, + 0.0149, -0.0443], device='cuda:0'), grad: tensor([ 4.6566e-10, -1.8626e-09, 3.9581e-09, 6.9849e-10, -1.9092e-08, + 5.8208e-09, 3.7253e-09, 5.1223e-09, -1.0943e-08, 1.9092e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 217.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4227 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.24 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.1335, 0.0219, 0.1721, ..., 0.0087, -0.1421, -0.1138], + [-0.2573, -0.2576, -0.2372, ..., 0.1254, 0.0741, 0.4499], + [-0.2993, -0.2035, -0.2699, ..., -0.3371, 0.0152, -0.2666], + ..., + [-0.2312, -0.2934, -0.0147, ..., -0.2889, -0.1704, -0.5154], + [ 0.1662, 0.0383, -0.1771, ..., -0.2812, 0.0569, -0.2564], + [ 0.1755, 0.1102, -0.1719, ..., 0.1419, -0.2802, -0.3440]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.1642e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-6.9849e-10, -4.6566e-10, 6.9849e-10, ..., 6.9849e-10, + -2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 6.9849e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0421, -0.0426, -0.0070, -0.0053, 0.0005, 0.0058, 0.0079, -0.0073, + 0.0149, -0.0443], device='cuda:0'), grad: tensor([ 2.5611e-09, 8.8476e-09, -3.7253e-09, 5.1223e-09, 5.8208e-09, + 9.3132e-09, -7.4506e-09, -4.6799e-08, 7.9162e-09, 2.1188e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 217.72, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4228 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.24 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.1335, 0.0219, 0.1722, ..., 0.0087, -0.1421, -0.1139], + [-0.2573, -0.2576, -0.2373, ..., 0.1254, 0.0741, 0.4500], + [-0.2994, -0.2036, -0.2699, ..., -0.3371, 0.0152, -0.2666], + ..., + [-0.2312, -0.2935, -0.0146, ..., -0.2890, -0.1705, -0.5156], + [ 0.1662, 0.0383, -0.1771, ..., -0.2812, 0.0569, -0.2565], + [ 0.1756, 0.1102, -0.1720, ..., 0.1420, -0.2804, -0.3441]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.9849e-10, -3.7253e-09, ..., -2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 2.3283e-10, -2.5611e-09, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 2.3283e-09, ..., -2.3283e-10, + 4.6566e-10, 9.3132e-10], + [-2.5611e-09, -1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + -1.1642e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0421, -0.0426, -0.0071, -0.0053, 0.0006, 0.0058, 0.0080, -0.0073, + 0.0149, -0.0444], device='cuda:0'), grad: tensor([-4.4238e-09, 2.6310e-08, -1.3364e-07, 7.2177e-09, 1.1642e-09, + -2.0955e-09, 9.3132e-09, 9.4995e-08, -6.0536e-09, 1.9791e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.51, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3785 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.22 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.1335, 0.0219, 0.1722, ..., 0.0086, -0.1421, -0.1139], + [-0.2574, -0.2576, -0.2373, ..., 0.1255, 0.0741, 0.4501], + [-0.2994, -0.2036, -0.2700, ..., -0.3371, 0.0151, -0.2667], + ..., + [-0.2313, -0.2935, -0.0146, ..., -0.2890, -0.1705, -0.5156], + [ 0.1663, 0.0383, -0.1772, ..., -0.2813, 0.0570, -0.2565], + [ 0.1757, 0.1103, -0.1721, ..., 0.1422, -0.2805, -0.3441]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 4.6566e-10, 6.9849e-10, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, + 2.3283e-10], + ..., + [2.3283e-10, 3.9581e-09, 7.6834e-09, ..., 2.3283e-10, 0.0000e+00, + 2.3283e-10], + [0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 1.5134e-08, 2.8871e-08, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0421, -0.0427, -0.0071, -0.0054, 0.0006, 0.0058, 0.0080, -0.0072, + 0.0148, -0.0444], device='cuda:0'), grad: tensor([ 6.9849e-10, 5.8208e-09, 2.3283e-09, 1.1642e-09, -1.8836e-07, + 5.8208e-09, 3.4925e-09, 3.8417e-08, 4.6566e-10, 1.4226e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 217.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4161 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.25 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.1335, 0.0220, 0.1724, ..., 0.0088, -0.1422, -0.1139], + [-0.2574, -0.2577, -0.2374, ..., 0.1255, 0.0742, 0.4503], + [-0.2994, -0.2036, -0.2700, ..., -0.3372, 0.0151, -0.2668], + ..., + [-0.2314, -0.2936, -0.0147, ..., -0.2891, -0.1706, -0.5157], + [ 0.1663, 0.0383, -0.1772, ..., -0.2814, 0.0570, -0.2566], + [ 0.1757, 0.1103, -0.1723, ..., 0.1423, -0.2806, -0.3441]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 9.3132e-10, 6.9849e-09, ..., 5.1223e-09, + 1.3970e-09, 0.0000e+00], + [ 2.5611e-09, 1.6298e-09, 2.3283e-09, ..., 1.8626e-09, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 2.7940e-09, ..., 2.3283e-09, + 6.9849e-10, 2.3283e-10], + ..., + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 4.6566e-10], + [-1.5600e-08, -1.0710e-08, 2.0722e-08, ..., 1.5600e-08, + -1.6531e-08, 0.0000e+00], + [ 2.0955e-09, 1.6298e-09, 1.8626e-09, ..., 2.3283e-10, + 4.1910e-09, 0.0000e+00]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0419, -0.0427, -0.0070, -0.0054, 0.0006, 0.0059, 0.0079, -0.0073, + 0.0148, -0.0445], device='cuda:0'), grad: tensor([ 4.1444e-08, 2.8405e-08, 1.2107e-08, 1.9092e-08, 8.8476e-09, + 6.6590e-08, -2.1909e-07, -4.1910e-09, 2.7474e-08, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 217.13, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4220 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.26 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.1335, 0.0221, 0.1724, ..., 0.0088, -0.1422, -0.1139], + [-0.2575, -0.2577, -0.2374, ..., 0.1256, 0.0742, 0.4504], + [-0.2995, -0.2036, -0.2700, ..., -0.3373, 0.0151, -0.2668], + ..., + [-0.2314, -0.2936, -0.0147, ..., -0.2892, -0.1706, -0.5158], + [ 0.1663, 0.0383, -0.1773, ..., -0.2815, 0.0572, -0.2566], + [ 0.1758, 0.1104, -0.1724, ..., 0.1424, -0.2807, -0.3441]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, -4.1910e-09, -1.1176e-08, ..., -4.4238e-09, + 0.0000e+00, 0.0000e+00], + [ 1.6298e-09, 1.3970e-09, 2.3283e-10, ..., 2.3283e-10, + -4.6566e-10, -1.0477e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 1.1642e-09], + ..., + [ 3.0268e-09, 2.3283e-09, 2.3283e-10, ..., 5.1223e-09, + 6.9849e-10, 5.8208e-09], + [ 9.3132e-10, 6.9849e-10, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 6.9849e-10], + [-1.4668e-08, -1.0477e-08, 7.9162e-09, ..., -1.3970e-08, + 6.2864e-09, 1.1642e-09]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0419, -0.0427, -0.0070, -0.0055, 0.0006, 0.0058, 0.0079, -0.0072, + 0.0148, -0.0445], device='cuda:0'), grad: tensor([-2.9104e-08, -1.0477e-08, -9.0804e-09, -3.7253e-09, 7.9162e-09, + 1.4435e-08, 2.5379e-08, 2.5611e-08, 6.0536e-09, -2.0722e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 217.33, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4433 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.25 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.1335, 0.0220, 0.1724, ..., 0.0088, -0.1422, -0.1139], + [-0.2575, -0.2578, -0.2374, ..., 0.1257, 0.0742, 0.4505], + [-0.2995, -0.2037, -0.2701, ..., -0.3373, 0.0150, -0.2669], + ..., + [-0.2315, -0.2937, -0.0147, ..., -0.2893, -0.1707, -0.5159], + [ 0.1663, 0.0383, -0.1773, ..., -0.2816, 0.0573, -0.2567], + [ 0.1759, 0.1105, -0.1725, ..., 0.1426, -0.2808, -0.3442]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + -1.6298e-09, -3.9581e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 1.1642e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-09, 2.3283e-09], + [-9.3132e-09, -8.3819e-09, 0.0000e+00, ..., -8.1491e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0420, -0.0428, -0.0072, -0.0055, 0.0005, 0.0058, 0.0079, -0.0071, + 0.0149, -0.0445], device='cuda:0'), grad: tensor([ 4.6566e-10, -3.7253e-09, 2.7940e-09, 4.6566e-10, 2.7940e-08, + 1.6298e-09, 1.3970e-09, 2.0955e-09, 4.1910e-09, -2.4214e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 217.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3989 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.25 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.1335, 0.0221, 0.1724, ..., 0.0087, -0.1422, -0.1139], + [-0.2576, -0.2578, -0.2375, ..., 0.1259, 0.0744, 0.4508], + [-0.2995, -0.2037, -0.2701, ..., -0.3375, 0.0149, -0.2671], + ..., + [-0.2316, -0.2937, -0.0146, ..., -0.2894, -0.1708, -0.5161], + [ 0.1664, 0.0383, -0.1773, ..., -0.2817, 0.0573, -0.2567], + [ 0.1759, 0.1105, -0.1726, ..., 0.1427, -0.2809, -0.3442]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -2.3283e-10, -9.3132e-10, ..., -6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-09, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 1.6298e-09, 9.3132e-10, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 4.6566e-10], + ..., + [ 2.5611e-09, 1.6298e-09, 0.0000e+00, ..., 6.9849e-10, + 9.3132e-10, 9.3132e-10], + [ 1.6997e-08, 1.0245e-08, 0.0000e+00, ..., 2.3283e-09, + 5.3551e-09, 4.6566e-10], + [ 2.3283e-10, 2.3283e-10, 6.2864e-09, ..., 2.3283e-10, + 4.1910e-09, 0.0000e+00]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0420, -0.0427, -0.0073, -0.0055, 0.0005, 0.0058, 0.0079, -0.0071, + 0.0148, -0.0446], device='cuda:0'), grad: tensor([-2.7940e-09, 5.5879e-09, 6.0536e-09, -1.3784e-07, -1.6997e-08, + 6.2399e-08, 7.2177e-09, 1.0245e-08, 5.5414e-08, 1.9325e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 217.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4026 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.24 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.1335, 0.0222, 0.1725, ..., 0.0088, -0.1422, -0.1139], + [-0.2576, -0.2579, -0.2375, ..., 0.1260, 0.0744, 0.4510], + [-0.2996, -0.2038, -0.2701, ..., -0.3376, 0.0149, -0.2672], + ..., + [-0.2316, -0.2938, -0.0146, ..., -0.2895, -0.1708, -0.5162], + [ 0.1664, 0.0382, -0.1774, ..., -0.2817, 0.0574, -0.2568], + [ 0.1760, 0.1106, -0.1727, ..., 0.1428, -0.2810, -0.3443]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + -6.9849e-09, -5.2387e-08], + [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., 0.0000e+00, + 1.3970e-09, 2.5611e-09], + ..., + [ 1.3970e-09, 1.1642e-09, 0.0000e+00, ..., 1.6298e-09, + 8.6147e-09, 5.0059e-08], + [-2.5611e-09, 4.6566e-10, 2.3283e-10, ..., 1.8626e-09, + -1.0710e-08, -6.9849e-10], + [-7.2177e-09, -6.5193e-09, 0.0000e+00, ..., -9.0804e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0419, -0.0427, -0.0074, -0.0056, 0.0005, 0.0058, 0.0079, -0.0071, + 0.0148, -0.0446], device='cuda:0'), grad: tensor([ 1.8626e-09, -9.5693e-08, 8.6147e-09, 5.3551e-09, 2.1188e-08, + 8.1491e-09, 1.4668e-08, 1.0617e-07, -2.6077e-08, -3.0966e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 217.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4217 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.1335, 0.0222, 0.1726, ..., 0.0089, -0.1422, -0.1139], + [-0.2576, -0.2579, -0.2376, ..., 0.1261, 0.0745, 0.4512], + [-0.2996, -0.2038, -0.2702, ..., -0.3376, 0.0149, -0.2673], + ..., + [-0.2317, -0.2938, -0.0146, ..., -0.2897, -0.1709, -0.5163], + [ 0.1664, 0.0382, -0.1774, ..., -0.2818, 0.0574, -0.2568], + [ 0.1761, 0.1106, -0.1728, ..., 0.1428, -0.2811, -0.3443]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0012e-08, ..., 8.1491e-09, + 1.1642e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.3737e-08, ..., 2.3283e-10, + 2.3283e-10, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 6.2864e-09, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0418, -0.0429, -0.0074, -0.0057, 0.0005, 0.0058, 0.0079, -0.0068, + 0.0147, -0.0447], device='cuda:0'), grad: tensor([ 7.7533e-08, 4.8894e-08, -3.6415e-07, 5.6811e-08, 1.6065e-08, + 6.9849e-09, 2.5611e-09, -3.7486e-08, 1.7183e-07, 4.2375e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 217.42, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4185 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.24 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.1335, 0.0223, 0.1727, ..., 0.0089, -0.1422, -0.1139], + [-0.2577, -0.2580, -0.2376, ..., 0.1262, 0.0746, 0.4514], + [-0.2997, -0.2038, -0.2702, ..., -0.3377, 0.0148, -0.2674], + ..., + [-0.2318, -0.2939, -0.0147, ..., -0.2898, -0.1710, -0.5165], + [ 0.1665, 0.0382, -0.1775, ..., -0.2819, 0.0575, -0.2569], + [ 0.1762, 0.1107, -0.1729, ..., 0.1430, -0.2812, -0.3444]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 4.4238e-09, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 6.9849e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + ..., + [ 4.6566e-10, 6.9849e-10, -1.0245e-08, ..., 4.6566e-10, + 4.6566e-10, 6.9849e-10], + [-1.3970e-09, -6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + -1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0418, -0.0429, -0.0074, -0.0058, 0.0004, 0.0059, 0.0078, -0.0069, + 0.0147, -0.0447], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.0966e-08, 1.7928e-08, -3.2596e-09, 1.5832e-08, + 5.1223e-09, 2.7940e-09, -8.0559e-08, -5.3551e-09, 2.1420e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 217.52, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4465 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.22 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.1335, 0.0224, 0.1728, ..., 0.0089, -0.1423, -0.1139], + [-0.2578, -0.2580, -0.2377, ..., 0.1264, 0.0747, 0.4517], + [-0.2997, -0.2039, -0.2703, ..., -0.3378, 0.0147, -0.2675], + ..., + [-0.2318, -0.2940, -0.0147, ..., -0.2899, -0.1711, -0.5168], + [ 0.1666, 0.0383, -0.1775, ..., -0.2820, 0.0576, -0.2569], + [ 0.1763, 0.1107, -0.1730, ..., 0.1431, -0.2813, -0.3444]], + device='cuda:0'), grad: tensor([[0.0000e+00, 4.6566e-10, 9.3132e-10, ..., 6.9849e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 2.3283e-10], + ..., + [2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, + 4.6566e-10], + [0.0000e+00, 9.3132e-10, 6.9849e-10, ..., 6.9849e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 0.0000e+00, 6.9849e-10, + 0.0000e+00]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0417, -0.0428, -0.0073, -0.0059, 0.0004, 0.0060, 0.0078, -0.0070, + 0.0147, -0.0447], device='cuda:0'), grad: tensor([ 1.0477e-08, 1.0245e-08, 2.5611e-09, 2.0256e-08, -1.8626e-09, + -8.1491e-08, 3.6554e-08, -7.4506e-09, 1.4668e-08, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 217.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4307 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.24 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.1335, 0.0226, 0.1731, ..., 0.0092, -0.1423, -0.1139], + [-0.2578, -0.2581, -0.2378, ..., 0.1265, 0.0747, 0.4519], + [-0.2997, -0.2039, -0.2704, ..., -0.3379, 0.0147, -0.2675], + ..., + [-0.2319, -0.2940, -0.0146, ..., -0.2901, -0.1711, -0.5170], + [ 0.1665, 0.0382, -0.1776, ..., -0.2821, 0.0577, -0.2569], + [ 0.1764, 0.1107, -0.1733, ..., 0.1431, -0.2814, -0.3444]], + device='cuda:0'), grad: tensor([[-2.5844e-08, -1.0501e-07, -5.7369e-07, ..., -4.6263e-07, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 3.9581e-09, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 1.6298e-09, 6.9849e-09, 3.8417e-08, ..., 3.0966e-08, + 0.0000e+00, 2.3283e-10], + ..., + [ 2.5611e-09, 9.3132e-10, 2.3283e-10, ..., 2.3283e-09, + 0.0000e+00, 2.3283e-10], + [ 1.3970e-09, 6.9849e-09, 2.6310e-08, ..., 2.1420e-08, + 0.0000e+00, 0.0000e+00], + [ 8.6147e-09, 4.3306e-08, 2.3982e-07, ..., 1.9139e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0413, -0.0428, -0.0074, -0.0061, 0.0004, 0.0062, 0.0076, -0.0070, + 0.0146, -0.0449], device='cuda:0'), grad: tensor([-1.6261e-06, 1.5600e-08, 9.8255e-08, 1.1246e-07, 3.3528e-08, + 2.0955e-08, 5.8394e-07, 1.2340e-08, 8.2189e-08, 6.7521e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 217.45, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4513 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.26 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.1335, 0.0227, 0.1733, ..., 0.0093, -0.1423, -0.1139], + [-0.2578, -0.2582, -0.2379, ..., 0.1266, 0.0748, 0.4522], + [-0.2998, -0.2040, -0.2705, ..., -0.3380, 0.0147, -0.2677], + ..., + [-0.2320, -0.2941, -0.0146, ..., -0.2903, -0.1712, -0.5173], + [ 0.1666, 0.0382, -0.1776, ..., -0.2822, 0.0578, -0.2570], + [ 0.1764, 0.1107, -0.1735, ..., 0.1431, -0.2815, -0.3444]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.0268e-09, -6.5193e-09, ..., -2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 6.9849e-10, ..., -5.5879e-09, + -3.0268e-09, -8.8476e-09], + [ 0.0000e+00, 9.3132e-10, 2.5611e-09, ..., 6.9849e-09, + 3.0268e-09, 8.6147e-09], + ..., + [ 0.0000e+00, 6.9849e-10, 1.3970e-09, ..., 9.3132e-10, + 2.3283e-10, 6.9849e-10], + [ 0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0412, -0.0427, -0.0074, -0.0061, 0.0004, 0.0063, 0.0074, -0.0070, + 0.0146, -0.0450], device='cuda:0'), grad: tensor([-2.1886e-08, -1.7928e-08, 3.1199e-08, 1.6298e-09, 1.6298e-09, + 9.3132e-10, 4.6566e-10, 6.2864e-09, 1.6298e-09, 7.2177e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 217.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3972 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.24 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.1334, 0.0228, 0.1734, ..., 0.0095, -0.1423, -0.1139], + [-0.2579, -0.2582, -0.2379, ..., 0.1267, 0.0748, 0.4524], + [-0.2998, -0.2040, -0.2705, ..., -0.3381, 0.0146, -0.2677], + ..., + [-0.2320, -0.2941, -0.0146, ..., -0.2904, -0.1712, -0.5174], + [ 0.1666, 0.0382, -0.1777, ..., -0.2823, 0.0578, -0.2571], + [ 0.1765, 0.1107, -0.1736, ..., 0.1431, -0.2816, -0.3445]], + device='cuda:0'), grad: tensor([[ 3.4925e-09, 0.0000e+00, 2.3283e-10, ..., 2.0955e-09, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00], + [ 2.3283e-09, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 6.9849e-10, 0.0000e+00], + ..., + [ 6.9849e-10, 2.3283e-10, 3.7253e-09, ..., 2.3283e-10, + 3.4925e-09, 0.0000e+00], + [ 0.0000e+00, -1.1642e-09, 4.6566e-10, ..., 1.1642e-09, + -9.3132e-10, 0.0000e+00], + [-2.2817e-08, -2.5611e-09, 3.0268e-09, ..., -1.7695e-08, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0410, -0.0428, -0.0075, -0.0061, 0.0004, 0.0062, 0.0074, -0.0070, + 0.0145, -0.0451], device='cuda:0'), grad: tensor([ 1.9791e-08, 5.8208e-09, 1.0245e-08, -2.0955e-09, 2.1420e-08, + 1.0477e-08, 6.2864e-09, 2.3749e-08, 6.2864e-09, -9.8720e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 217.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4062 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.26 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.1334, 0.0229, 0.1735, ..., 0.0096, -0.1423, -0.1139], + [-0.2579, -0.2583, -0.2380, ..., 0.1268, 0.0748, 0.4526], + [-0.2999, -0.2041, -0.2706, ..., -0.3381, 0.0146, -0.2678], + ..., + [-0.2321, -0.2941, -0.0146, ..., -0.2905, -0.1713, -0.5176], + [ 0.1666, 0.0382, -0.1777, ..., -0.2824, 0.0579, -0.2571], + [ 0.1766, 0.1107, -0.1738, ..., 0.1432, -0.2817, -0.3445]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1642e-10, -2.7940e-09, ..., -1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-10, + 1.1642e-10, 0.0000e+00], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 0.0000e+00], + ..., + [ 5.8208e-10, 5.8208e-10, 2.3283e-10, ..., 6.9849e-10, + 1.1642e-10, 1.1642e-10], + [-2.3283e-10, -3.4925e-10, 0.0000e+00, ..., 1.1642e-10, + -3.4925e-10, 0.0000e+00], + [-1.2806e-09, -4.6566e-10, 4.6566e-10, ..., -1.1642e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0409, -0.0427, -0.0075, -0.0062, 0.0004, 0.0062, 0.0073, -0.0070, + 0.0145, -0.0452], device='cuda:0'), grad: tensor([-6.8685e-09, 4.5402e-09, -1.2806e-09, 1.2107e-08, 2.6776e-09, + -5.3551e-09, 1.2573e-08, -2.2817e-08, -9.3132e-10, 5.2387e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 217.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3960 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.24 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.1334, 0.0230, 0.1736, ..., 0.0096, -0.1423, -0.1139], + [-0.2580, -0.2583, -0.2380, ..., 0.1269, 0.0749, 0.4527], + [-0.2999, -0.2041, -0.2706, ..., -0.3382, 0.0145, -0.2679], + ..., + [-0.2321, -0.2942, -0.0146, ..., -0.2906, -0.1713, -0.5176], + [ 0.1666, 0.0382, -0.1778, ..., -0.2825, 0.0580, -0.2572], + [ 0.1767, 0.1108, -0.1739, ..., 0.1433, -0.2818, -0.3445]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.5134e-09, 1.1642e-09, ..., 6.9849e-10, + 2.2119e-09, 6.9849e-10], + [ 1.3970e-09, 2.5611e-09, 2.2119e-09, ..., -8.4983e-09, + -3.3760e-09, -1.7113e-08], + [ 4.8894e-09, 7.2177e-09, 9.3132e-09, ..., 1.1642e-09, + 1.2922e-08, 2.3283e-09], + ..., + [ 1.6298e-09, 1.3970e-09, 1.5134e-09, ..., 1.8626e-09, + 2.5611e-09, 2.5611e-09], + [-1.2922e-08, -1.1758e-08, -2.8405e-08, ..., 3.3760e-09, + -3.4459e-08, 6.1700e-09], + [-2.3283e-10, 1.3970e-09, 3.4808e-08, ..., -9.3132e-10, + 6.2049e-08, 1.1642e-10]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0409, -0.0428, -0.0076, -0.0062, 0.0003, 0.0062, 0.0073, -0.0068, + 0.0144, -0.0452], device='cuda:0'), grad: tensor([ 1.4435e-08, -1.1409e-08, 8.8359e-08, 1.2247e-07, -2.0198e-07, + -1.1805e-07, 9.5926e-08, 2.0838e-08, -2.1479e-07, 2.1572e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 217.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4395 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.25 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.1334, 0.0230, 0.1736, ..., 0.0096, -0.1423, -0.1139], + [-0.2580, -0.2584, -0.2381, ..., 0.1270, 0.0750, 0.4529], + [-0.3000, -0.2041, -0.2707, ..., -0.3382, 0.0145, -0.2680], + ..., + [-0.2322, -0.2942, -0.0145, ..., -0.2907, -0.1714, -0.5178], + [ 0.1667, 0.0382, -0.1778, ..., -0.2826, 0.0581, -0.2573], + [ 0.1768, 0.1108, -0.1740, ..., 0.1434, -0.2819, -0.3445]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.1642e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 8.1491e-10, 5.8208e-10, 5.8208e-10, ..., -1.5134e-09, + -1.1642e-09, -1.2224e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 1.1642e-10, 6.9849e-10], + ..., + [ 1.2806e-09, 8.1491e-10, 0.0000e+00, ..., 4.7730e-09, + 1.2806e-09, 9.3132e-09], + [ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 3.4925e-10, + 1.1642e-10, 5.8208e-10], + [-4.8894e-09, -3.3760e-09, 0.0000e+00, ..., -7.5670e-09, + 1.1642e-10, 3.4925e-10]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0409, -0.0428, -0.0077, -0.0063, 0.0003, 0.0062, 0.0073, -0.0068, + 0.0145, -0.0453], device='cuda:0'), grad: tensor([ 1.3970e-09, -9.4296e-09, -5.5879e-09, 9.3132e-09, 1.5949e-08, + -8.7311e-09, -3.3760e-09, 2.2352e-08, 1.0012e-08, -2.0722e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 217.42, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4514 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.24 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.1334, 0.0230, 0.1736, ..., 0.0096, -0.1423, -0.1139], + [-0.2581, -0.2584, -0.2382, ..., 0.1271, 0.0750, 0.4531], + [-0.3000, -0.2042, -0.2707, ..., -0.3383, 0.0144, -0.2680], + ..., + [-0.2323, -0.2943, -0.0146, ..., -0.2908, -0.1715, -0.5180], + [ 0.1667, 0.0382, -0.1779, ..., -0.2827, 0.0582, -0.2573], + [ 0.1768, 0.1108, -0.1742, ..., 0.1435, -0.2822, -0.3445]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.6089e-09, ..., 1.1642e-09, + 1.5134e-09, 1.3970e-09], + [ 1.1642e-09, 3.4925e-10, -3.4925e-10, ..., -1.4435e-08, + -6.0536e-09, -2.1071e-08], + [ 0.0000e+00, 0.0000e+00, 1.0477e-09, ..., 1.7462e-09, + 1.3970e-09, 4.3074e-09], + ..., + [ 1.1642e-10, 0.0000e+00, 2.3283e-10, ..., 5.2387e-09, + 1.8626e-09, 5.8208e-09], + [ 2.3283e-10, 2.3283e-10, 2.2119e-09, ..., 6.6357e-09, + 1.7462e-09, 5.8208e-09], + [-1.1642e-09, -1.1642e-10, 2.5611e-09, ..., 0.0000e+00, + 1.6298e-09, 1.9791e-09]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0409, -0.0428, -0.0075, -0.0062, 0.0005, 0.0062, 0.0074, -0.0069, + 0.0143, -0.0455], device='cuda:0'), grad: tensor([ 1.3504e-08, -4.4703e-08, 9.0804e-09, -2.5611e-09, -1.8277e-08, + 4.4238e-09, -1.0477e-09, 8.4983e-09, 2.6659e-08, 1.1176e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 217.42, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3907 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.24 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.1334, 0.0231, 0.1738, ..., 0.0098, -0.1424, -0.1140], + [-0.2581, -0.2585, -0.2382, ..., 0.1272, 0.0751, 0.4533], + [-0.3000, -0.2042, -0.2708, ..., -0.3385, 0.0144, -0.2682], + ..., + [-0.2324, -0.2944, -0.0146, ..., -0.2909, -0.1716, -0.5181], + [ 0.1667, 0.0381, -0.1780, ..., -0.2828, 0.0582, -0.2574], + [ 0.1770, 0.1109, -0.1744, ..., 0.1435, -0.2823, -0.3445]], + device='cuda:0'), grad: tensor([[ 1.0477e-09, 2.3283e-10, 0.0000e+00, ..., 1.0477e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-08, 2.9104e-09, 0.0000e+00, ..., 1.0943e-08, + 1.1642e-10, 0.0000e+00], + [ 3.4925e-10, 0.0000e+00, 1.1642e-10, ..., 2.3283e-10, + -9.3132e-10, 1.1642e-10], + ..., + [ 4.0745e-09, 1.1642e-09, 0.0000e+00, ..., 3.9581e-09, + 1.1642e-10, 2.3283e-10], + [ 1.9791e-09, 2.3283e-09, 1.1642e-10, ..., 1.7462e-09, + 6.9849e-10, 1.1642e-10], + [-6.8918e-08, -1.7229e-08, 0.0000e+00, ..., -6.4494e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0407, -0.0427, -0.0077, -0.0062, 0.0005, 0.0062, 0.0073, -0.0069, + 0.0143, -0.0456], device='cuda:0'), grad: tensor([ 4.7730e-09, 4.9942e-08, -1.2224e-08, 1.9791e-09, 1.9651e-07, + -9.3132e-10, 7.9162e-09, 1.8044e-08, 2.6543e-08, -2.8242e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 217.82, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4318 re_mapping 0.0018 re_causal 0.0069 /// teacc 99.27 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.1334, 0.0232, 0.1738, ..., 0.0098, -0.1424, -0.1140], + [-0.2582, -0.2585, -0.2383, ..., 0.1272, 0.0751, 0.4534], + [-0.3001, -0.2042, -0.2708, ..., -0.3385, 0.0144, -0.2683], + ..., + [-0.2324, -0.2944, -0.0145, ..., -0.2910, -0.1716, -0.5181], + [ 0.1667, 0.0381, -0.1780, ..., -0.2829, 0.0582, -0.2574], + [ 0.1770, 0.1109, -0.1744, ..., 0.1436, -0.2824, -0.3446]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-10, -2.2119e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 3.4925e-10], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-10, 8.1491e-10], + [-2.6776e-09, -1.1642e-10, -3.4925e-10, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0407, -0.0428, -0.0077, -0.0062, 0.0005, 0.0063, 0.0072, -0.0069, + 0.0142, -0.0456], device='cuda:0'), grad: tensor([ 3.4925e-10, -5.8208e-10, 8.1491e-10, 3.6089e-09, 1.1176e-08, + -5.5879e-09, 3.8417e-09, 2.6776e-09, 3.7253e-09, -1.0594e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 217.66, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4224 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.24 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.1334, 0.0232, 0.1738, ..., 0.0098, -0.1424, -0.1140], + [-0.2582, -0.2586, -0.2383, ..., 0.1273, 0.0752, 0.4536], + [-0.3001, -0.2043, -0.2709, ..., -0.3386, 0.0143, -0.2684], + ..., + [-0.2325, -0.2945, -0.0145, ..., -0.2910, -0.1717, -0.5182], + [ 0.1667, 0.0381, -0.1781, ..., -0.2830, 0.0582, -0.2576], + [ 0.1771, 0.1110, -0.1746, ..., 0.1438, -0.2825, -0.3446]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9791e-09, ..., 1.6298e-09, + 3.4925e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.2713e-08, ..., 2.5844e-08, + 8.6147e-09, 1.6298e-08], + [ 1.1642e-10, 0.0000e+00, 1.1642e-10, ..., 2.3283e-10, + 8.1491e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 1.1642e-10, + 2.3283e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 2.0955e-09, + -6.1700e-09, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 3.4925e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0407, -0.0427, -0.0079, -0.0062, 0.0005, 0.0064, 0.0072, -0.0068, + 0.0141, -0.0457], device='cuda:0'), grad: tensor([ 8.0327e-09, 1.3213e-07, -9.6625e-09, 3.4925e-10, 1.7579e-08, + 1.5926e-07, -3.0384e-07, 6.1700e-09, -1.0477e-08, 3.0268e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 217.62, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4193 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.26 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.1334, 0.0232, 0.1738, ..., 0.0098, -0.1424, -0.1140], + [-0.2583, -0.2586, -0.2384, ..., 0.1274, 0.0752, 0.4539], + [-0.3002, -0.2043, -0.2709, ..., -0.3386, 0.0143, -0.2685], + ..., + [-0.2326, -0.2945, -0.0145, ..., -0.2912, -0.1718, -0.5185], + [ 0.1668, 0.0381, -0.1781, ..., -0.2831, 0.0583, -0.2576], + [ 0.1774, 0.1112, -0.1746, ..., 0.1440, -0.2827, -0.3446]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, 0.0000e+00, + 0.0000e+00], + [1.1642e-10, 1.1642e-10, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, + 0.0000e+00], + [1.1642e-10, 0.0000e+00, 3.4925e-10, ..., 4.6566e-10, 0.0000e+00, + 2.3283e-10], + ..., + [1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 3.4925e-10, 2.3283e-10, + 4.6566e-10], + [1.1642e-10, 2.3283e-10, 2.6776e-09, ..., 2.4447e-09, 1.1642e-10, + 0.0000e+00], + [3.4925e-10, 4.6566e-10, 0.0000e+00, ..., 3.4925e-10, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0408, -0.0427, -0.0079, -0.0062, 0.0003, 0.0063, 0.0072, -0.0069, + 0.0141, -0.0456], device='cuda:0'), grad: tensor([ 1.0477e-09, 3.7253e-09, 1.1642e-08, -3.2596e-09, 4.6566e-10, + 6.6357e-09, -1.4552e-08, -1.3271e-08, 1.1292e-08, 1.7462e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 217.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4013 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.26 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.1335, 0.0232, 0.1739, ..., 0.0098, -0.1424, -0.1140], + [-0.2584, -0.2587, -0.2385, ..., 0.1276, 0.0754, 0.4541], + [-0.3002, -0.2043, -0.2709, ..., -0.3387, 0.0142, -0.2686], + ..., + [-0.2327, -0.2946, -0.0146, ..., -0.2914, -0.1719, -0.5186], + [ 0.1668, 0.0381, -0.1782, ..., -0.2833, 0.0584, -0.2577], + [ 0.1776, 0.1113, -0.1747, ..., 0.1441, -0.2827, -0.3446]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 5.8208e-10, + 2.3283e-10, 6.9849e-10], + [ 1.1642e-10, 5.8208e-10, 1.1642e-10, ..., -3.7951e-08, + -1.7579e-08, -5.9023e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4447e-09, + 1.2806e-09, 4.4238e-09], + ..., + [ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 2.6426e-08, + 1.2340e-08, 4.2142e-08], + [ 1.1642e-10, 3.4925e-10, 1.0477e-09, ..., 4.0745e-09, + 1.5134e-09, 5.7044e-09], + [-1.3970e-09, -1.0477e-09, 0.0000e+00, ..., 2.5611e-09, + 1.9791e-09, 6.7521e-09]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0408, -0.0427, -0.0079, -0.0062, 0.0002, 0.0063, 0.0073, -0.0069, + 0.0139, -0.0456], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.4238e-07, 7.5670e-09, 5.9372e-09, 9.4296e-09, + -5.5879e-09, 5.1223e-09, 1.0733e-07, 1.9209e-08, 9.5461e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 217.73, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4272 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.25 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.1335, 0.0232, 0.1739, ..., 0.0098, -0.1425, -0.1141], + [-0.2584, -0.2587, -0.2385, ..., 0.1278, 0.0756, 0.4545], + [-0.3002, -0.2044, -0.2709, ..., -0.3388, 0.0142, -0.2688], + ..., + [-0.2328, -0.2947, -0.0146, ..., -0.2916, -0.1721, -0.5189], + [ 0.1668, 0.0380, -0.1783, ..., -0.2834, 0.0583, -0.2579], + [ 0.1777, 0.1114, -0.1748, ..., 0.1443, -0.2828, -0.3447]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1642e-10, 6.9849e-10, ..., 6.9849e-10, + 3.4925e-10, 4.6566e-10], + [ 6.9849e-10, 4.6566e-10, 3.4925e-10, ..., -1.5832e-08, + -3.8417e-09, -2.5728e-08], + [ 8.1491e-10, 9.3132e-10, 4.6566e-10, ..., 1.3970e-09, + 1.7462e-09, 2.4447e-09], + ..., + [ 1.2806e-09, 9.3132e-10, -1.1642e-10, ..., 9.3132e-09, + 1.5134e-09, 1.1292e-08], + [-8.1491e-10, -1.5134e-09, 3.2596e-09, ..., 4.6566e-09, + 1.0477e-09, 7.4506e-09], + [ 3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 3.7253e-09, + 2.3283e-10, 3.8417e-09]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0408, -0.0425, -0.0079, -0.0062, 0.0001, 0.0063, 0.0074, -0.0070, + 0.0136, -0.0456], device='cuda:0'), grad: tensor([ 3.6089e-09, -6.2631e-08, 1.2573e-08, -1.0128e-08, 6.7521e-09, + 1.1292e-08, -1.4668e-08, 3.6089e-08, 1.5832e-08, 1.4319e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 217.54, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4215 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.26 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.1335, 0.0233, 0.1739, ..., 0.0098, -0.1425, -0.1141], + [-0.2585, -0.2588, -0.2386, ..., 0.1279, 0.0757, 0.4547], + [-0.3003, -0.2044, -0.2709, ..., -0.3388, 0.0141, -0.2688], + ..., + [-0.2328, -0.2947, -0.0146, ..., -0.2917, -0.1722, -0.5191], + [ 0.1668, 0.0380, -0.1784, ..., -0.2836, 0.0584, -0.2580], + [ 0.1779, 0.1115, -0.1749, ..., 0.1444, -0.2829, -0.3447]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 3.4925e-10, + 0.0000e+00, 0.0000e+00], + [ 8.1491e-10, 1.0477e-09, 1.1642e-10, ..., 1.3970e-09, + 3.4925e-10, 0.0000e+00], + [ 1.1642e-10, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, + 1.1642e-10, 1.1642e-10], + ..., + [ 4.6566e-10, 6.9849e-10, -3.4925e-10, ..., 9.3132e-10, + 2.3283e-10, 1.1642e-10], + [ 0.0000e+00, 3.4925e-10, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [-8.6147e-09, -6.2864e-09, 3.4925e-10, ..., -1.3853e-08, + 1.1642e-10, 0.0000e+00]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0408, -0.0424, -0.0079, -0.0062, 0.0001, 0.0063, 0.0074, -0.0071, + 0.0134, -0.0456], device='cuda:0'), grad: tensor([ 8.1491e-10, 1.0012e-08, 5.7044e-09, -6.2864e-09, 3.7020e-08, + 9.3132e-10, 2.0955e-09, -3.2596e-09, 1.8626e-09, -4.1327e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 217.67, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4100 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.23 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.1335, 0.0233, 0.1739, ..., 0.0098, -0.1425, -0.1141], + [-0.2585, -0.2588, -0.2386, ..., 0.1281, 0.0758, 0.4548], + [-0.3003, -0.2044, -0.2709, ..., -0.3389, 0.0141, -0.2689], + ..., + [-0.2329, -0.2948, -0.0146, ..., -0.2919, -0.1722, -0.5191], + [ 0.1667, 0.0379, -0.1785, ..., -0.2838, 0.0584, -0.2580], + [ 0.1780, 0.1116, -0.1750, ..., 0.1446, -0.2830, -0.3448]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-10, 4.6566e-10], + [ 5.8208e-10, 2.3283e-10, -5.8208e-10, ..., -3.1432e-09, + -5.9372e-09, -1.8277e-08], + [ 1.1642e-10, 1.1642e-10, 1.1642e-10, ..., 9.3132e-10, + 1.6298e-09, 4.5402e-09], + ..., + [ 5.8208e-10, 2.3283e-10, 1.1642e-10, ..., 1.5134e-09, + 1.1642e-09, 3.4925e-09], + [ 1.1642e-10, 2.3283e-10, 0.0000e+00, ..., 5.8208e-10, + 4.6566e-10, 1.2806e-09], + [-3.4925e-09, -1.5134e-09, 1.1642e-10, ..., -4.7730e-09, + 2.3283e-10, 4.6566e-10]], device='cuda:0') +Epoch 490, bias, value: tensor([-4.0773e-02, -4.2563e-02, -7.7614e-03, -6.2187e-03, 7.5489e-05, + 6.3689e-03, 7.3327e-03, -6.9412e-03, 1.3092e-02, -4.5542e-02], + device='cuda:0'), grad: tensor([ 1.1642e-09, -2.8755e-08, 8.7311e-09, 1.2806e-09, 1.1176e-08, + 2.2119e-09, 4.5402e-09, 8.9640e-09, 3.0268e-09, -1.3388e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 217.91, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4028 re_mapping 0.0018 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.1335, 0.0233, 0.1739, ..., 0.0098, -0.1425, -0.1141], + [-0.2586, -0.2589, -0.2386, ..., 0.1282, 0.0759, 0.4551], + [-0.3003, -0.2044, -0.2710, ..., -0.3390, 0.0141, -0.2690], + ..., + [-0.2330, -0.2949, -0.0146, ..., -0.2921, -0.1724, -0.5193], + [ 0.1667, 0.0379, -0.1786, ..., -0.2839, 0.0584, -0.2581], + [ 0.1782, 0.1118, -0.1751, ..., 0.1448, -0.2831, -0.3448]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 4.6566e-10, ..., 5.8208e-10, + 2.3283e-10, 8.1491e-10], + [ 3.4925e-10, 0.0000e+00, 6.9849e-10, ..., -4.5402e-09, + -3.3760e-09, -2.0955e-08], + [ 5.8208e-10, 2.3283e-10, 2.3283e-10, ..., 8.1491e-10, + 9.3132e-10, 3.3760e-09], + ..., + [ 2.2119e-09, 0.0000e+00, 1.2806e-09, ..., 3.3760e-09, + 2.9104e-09, 1.1409e-08], + [ 9.3132e-10, 2.3283e-10, 2.3283e-10, ..., 8.1491e-10, + 4.6566e-10, 1.6298e-09], + [-5.5879e-09, 0.0000e+00, 1.1292e-08, ..., -1.2806e-09, + 7.4506e-09, 2.3283e-09]], device='cuda:0') +Epoch 491, bias, value: tensor([-4.0819e-02, -4.2616e-02, -7.7872e-03, -6.2157e-03, 4.3897e-05, + 6.3607e-03, 7.2601e-03, -6.8821e-03, 1.3030e-02, -4.5478e-02], + device='cuda:0'), grad: tensor([ 8.6147e-09, -3.1199e-08, -1.4086e-08, 3.0268e-09, -5.2736e-08, + 8.7311e-09, 1.8626e-09, 4.4121e-08, 1.7462e-08, 3.0501e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 217.80, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3950 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.1335, 0.0233, 0.1739, ..., 0.0098, -0.1425, -0.1141], + [-0.2586, -0.2589, -0.2387, ..., 0.1283, 0.0760, 0.4552], + [-0.3004, -0.2045, -0.2710, ..., -0.3390, 0.0140, -0.2691], + ..., + [-0.2332, -0.2949, -0.0146, ..., -0.2922, -0.1725, -0.5194], + [ 0.1667, 0.0378, -0.1786, ..., -0.2840, 0.0585, -0.2581], + [ 0.1784, 0.1118, -0.1752, ..., 0.1450, -0.2833, -0.3448]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 1.0477e-09], + [ 3.4925e-10, 2.3283e-10, 0.0000e+00, ..., -6.6357e-09, + -4.7730e-09, -2.4447e-08], + [ 3.4925e-10, 2.3283e-10, 0.0000e+00, ..., 1.5134e-09, + 9.3132e-10, 4.1910e-09], + ..., + [ 9.3132e-10, 5.8208e-10, 0.0000e+00, ..., 3.6089e-09, + 2.2119e-09, 1.1176e-08], + [ 2.3283e-10, 3.4925e-10, 0.0000e+00, ..., 1.1642e-09, + 4.6566e-10, 2.3283e-09], + [-5.5879e-09, -3.0268e-09, -2.3283e-10, ..., -2.9104e-09, + 5.8208e-10, 2.4447e-09]], device='cuda:0') +Epoch 492, bias, value: tensor([-4.0819e-02, -4.2615e-02, -7.8146e-03, -6.1641e-03, 5.2491e-05, + 6.4032e-03, 7.1880e-03, -6.8993e-03, 1.2928e-02, -4.5462e-02], + device='cuda:0'), grad: tensor([ 2.6776e-09, -4.2142e-08, 9.4296e-09, -6.0536e-09, 2.2817e-08, + -3.9209e-07, 3.9442e-07, 2.3516e-08, 6.0536e-09, -1.4203e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 217.68, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4414 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.23 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.1336, 0.0233, 0.1740, ..., 0.0098, -0.1425, -0.1141], + [-0.2587, -0.2589, -0.2387, ..., 0.1285, 0.0760, 0.4554], + [-0.3004, -0.2045, -0.2710, ..., -0.3391, 0.0140, -0.2691], + ..., + [-0.2332, -0.2950, -0.0146, ..., -0.2924, -0.1725, -0.5196], + [ 0.1667, 0.0378, -0.1787, ..., -0.2841, 0.0585, -0.2582], + [ 0.1785, 0.1119, -0.1753, ..., 0.1451, -0.2834, -0.3448]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 1.1642e-10, 1.1642e-10, 9.3132e-10, ..., -1.0477e-09, + 5.8208e-10, -4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 3.1432e-09, ..., 2.3283e-10, + 3.0268e-09, 1.5134e-09], + ..., + [ 3.4925e-10, 3.4925e-10, -5.2387e-09, ..., 9.3132e-10, + -4.1910e-09, 1.1642e-09], + [ 0.0000e+00, 3.4925e-10, 3.4925e-10, ..., 0.0000e+00, + 3.4925e-10, 3.4925e-10], + [-6.9849e-10, -4.6566e-10, 5.0059e-09, ..., -1.5134e-09, + 3.4925e-09, 1.1642e-10]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0408, -0.0426, -0.0078, -0.0062, 0.0001, 0.0065, 0.0071, -0.0069, + 0.0128, -0.0456], device='cuda:0'), grad: tensor([ 4.6566e-10, -5.8208e-10, 2.6193e-08, 1.7462e-09, -8.1491e-09, + 4.6566e-09, 2.3283e-09, -3.3528e-08, 3.3760e-09, 1.2107e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 217.69, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4182 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.1336, 0.0234, 0.1740, ..., 0.0098, -0.1426, -0.1141], + [-0.2587, -0.2590, -0.2388, ..., 0.1285, 0.0761, 0.4555], + [-0.3004, -0.2045, -0.2710, ..., -0.3391, 0.0140, -0.2692], + ..., + [-0.2333, -0.2951, -0.0146, ..., -0.2925, -0.1726, -0.5197], + [ 0.1667, 0.0378, -0.1787, ..., -0.2842, 0.0586, -0.2582], + [ 0.1787, 0.1120, -0.1754, ..., 0.1453, -0.2835, -0.3449]], + device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00], + [2.3283e-10, 1.1642e-10, 0.0000e+00, ..., 5.8208e-10, 1.1642e-10, + 1.1642e-10], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 2.3283e-10], + [0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1642e-09, 0.0000e+00, + 0.0000e+00], + [1.1642e-10, 1.1642e-10, 2.3283e-10, ..., 1.1642e-10, 2.3283e-10, + 0.0000e+00]], device='cuda:0') +Epoch 494, bias, value: tensor([-4.0828e-02, -4.2549e-02, -7.7843e-03, -6.1556e-03, 2.5391e-05, + 6.4926e-03, 7.1603e-03, -6.9491e-03, 1.2596e-02, -4.5518e-02], + device='cuda:0'), grad: tensor([ 2.2119e-09, 7.3342e-09, -1.5949e-08, 3.0268e-09, 8.1491e-10, + 4.1910e-09, -1.4435e-08, 7.6834e-09, 9.3132e-09, 2.6776e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 217.73, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4007 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.23 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.1336, 0.0234, 0.1740, ..., 0.0098, -0.1426, -0.1141], + [-0.2588, -0.2590, -0.2388, ..., 0.1286, 0.0761, 0.4557], + [-0.3005, -0.2046, -0.2711, ..., -0.3392, 0.0140, -0.2693], + ..., + [-0.2334, -0.2951, -0.0146, ..., -0.2926, -0.1726, -0.5197], + [ 0.1668, 0.0378, -0.1788, ..., -0.2843, 0.0586, -0.2583], + [ 0.1788, 0.1121, -0.1755, ..., 0.1454, -0.2836, -0.3449]], + device='cuda:0'), grad: tensor([[ 2.4447e-09, 1.9791e-09, 5.8208e-10, ..., 3.8417e-09, + 0.0000e+00, 0.0000e+00], + [ 1.0477e-09, 1.0477e-09, 6.9849e-10, ..., 1.0477e-09, + -3.4925e-10, -1.6298e-09], + [ 1.1642e-10, 2.3283e-10, 6.9849e-10, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + ..., + [ 4.6566e-10, -5.8208e-10, -2.5611e-09, ..., 8.1491e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 4.6566e-10, 3.4925e-10, ..., 1.3970e-09, + 2.3283e-10, 0.0000e+00], + [-1.3388e-08, -1.0012e-08, 2.3283e-10, ..., -2.2119e-08, + 1.1642e-10, 2.3283e-10]], device='cuda:0') +Epoch 495, bias, value: tensor([-4.0821e-02, -4.2541e-02, -7.8652e-03, -6.1364e-03, -4.2546e-05, + 6.4397e-03, 7.1976e-03, -6.9497e-03, 1.2545e-02, -4.5493e-02], + device='cuda:0'), grad: tensor([ 1.4668e-08, 6.7521e-09, 4.8894e-09, 1.3970e-08, 2.1770e-08, + 1.2806e-08, -3.0268e-09, -1.6647e-08, 5.3551e-09, -5.7160e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 217.65, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4383 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.24 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.1336, 0.0235, 0.1741, ..., 0.0099, -0.1426, -0.1142], + [-0.2588, -0.2591, -0.2389, ..., 0.1287, 0.0762, 0.4558], + [-0.3006, -0.2046, -0.2711, ..., -0.3392, 0.0139, -0.2693], + ..., + [-0.2335, -0.2952, -0.0146, ..., -0.2927, -0.1727, -0.5198], + [ 0.1668, 0.0377, -0.1789, ..., -0.2844, 0.0586, -0.2583], + [ 0.1789, 0.1122, -0.1756, ..., 0.1456, -0.2837, -0.3449]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-10, + 3.4925e-10, 0.0000e+00], + [ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + ..., + [ 1.2806e-09, 6.9849e-10, 0.0000e+00, ..., 9.3132e-10, + 5.8208e-10, 5.8208e-10], + [-4.0745e-09, -1.1642e-09, 0.0000e+00, ..., 5.8208e-10, + -5.0059e-09, 2.3283e-10], + [-2.3283e-09, -1.8626e-09, 0.0000e+00, ..., -1.8626e-09, + 1.1642e-10, 0.0000e+00]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0408, -0.0425, -0.0079, -0.0060, -0.0001, 0.0064, 0.0073, -0.0070, + 0.0124, -0.0455], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.6065e-08, 4.0745e-09, 1.3271e-08, 1.3970e-09, + 4.3074e-09, 2.9104e-09, -1.3039e-08, -2.4564e-08, -2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 217.65, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3996 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.24 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.1336, 0.0235, 0.1741, ..., 0.0099, -0.1426, -0.1142], + [-0.2589, -0.2592, -0.2389, ..., 0.1288, 0.0762, 0.4559], + [-0.3006, -0.2047, -0.2712, ..., -0.3393, 0.0139, -0.2694], + ..., + [-0.2336, -0.2953, -0.0145, ..., -0.2928, -0.1727, -0.5198], + [ 0.1669, 0.0377, -0.1790, ..., -0.2845, 0.0587, -0.2584], + [ 0.1791, 0.1122, -0.1757, ..., 0.1458, -0.2838, -0.3449]], + device='cuda:0'), grad: tensor([[ 1.5134e-09, 1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 2.3283e-10, 0.0000e+00], + [ 1.0477e-09, 9.3132e-10, 0.0000e+00, ..., 8.1491e-10, + 4.6566e-10, 0.0000e+00], + [ 5.8208e-10, 5.8208e-10, 0.0000e+00, ..., 3.4925e-10, + 2.3283e-10, 1.1642e-10], + ..., + [ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 1.6298e-09, + 5.8208e-10, 0.0000e+00], + [ 2.4447e-09, 2.2119e-09, 0.0000e+00, ..., 1.2806e-09, + 1.3970e-09, 1.1642e-10], + [ 1.2806e-09, 3.6089e-09, 1.2806e-09, ..., -5.8208e-10, + 1.5134e-09, 0.0000e+00]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0408, -0.0426, -0.0081, -0.0059, -0.0001, 0.0063, 0.0073, -0.0070, + 0.0123, -0.0455], device='cuda:0'), grad: tensor([ 5.7044e-09, 2.7241e-08, -3.2596e-09, -2.4796e-08, 2.7940e-09, + 1.5134e-08, 6.9849e-10, -1.7218e-07, 1.5832e-08, 1.4796e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 217.73, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3912 re_mapping 0.0018 re_causal 0.0068 /// teacc 99.23 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.1336, 0.0236, 0.1742, ..., 0.0099, -0.1426, -0.1142], + [-0.2589, -0.2592, -0.2389, ..., 0.1288, 0.0762, 0.4560], + [-0.3007, -0.2047, -0.2712, ..., -0.3394, 0.0138, -0.2695], + ..., + [-0.2337, -0.2954, -0.0146, ..., -0.2928, -0.1728, -0.5199], + [ 0.1669, 0.0377, -0.1790, ..., -0.2846, 0.0587, -0.2584], + [ 0.1791, 0.1122, -0.1759, ..., 0.1458, -0.2838, -0.3450]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.1642e-10, ..., 0.0000e+00, + -4.6566e-10, 1.1642e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 3.4925e-10], + [ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 1.1642e-10, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.1642e-10, + 2.3283e-10, 1.1642e-10]], device='cuda:0') +Epoch 498, bias, value: tensor([-4.0681e-02, -4.2569e-02, -8.1047e-03, -5.8771e-03, -8.1411e-05, + 6.2939e-03, 7.3545e-03, -6.9353e-03, 1.2329e-02, -4.5701e-02], + device='cuda:0'), grad: tensor([ 1.9791e-09, 2.3283e-09, -2.1770e-08, 1.1642e-09, -3.4925e-09, + 4.6566e-10, 1.1642e-09, 8.6147e-09, 9.7789e-09, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 217.72, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4033 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.23 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.1336, 0.0236, 0.1743, ..., 0.0100, -0.1426, -0.1142], + [-0.2590, -0.2593, -0.2390, ..., 0.1289, 0.0763, 0.4560], + [-0.3007, -0.2048, -0.2713, ..., -0.3394, 0.0138, -0.2695], + ..., + [-0.2338, -0.2954, -0.0146, ..., -0.2930, -0.1729, -0.5199], + [ 0.1670, 0.0377, -0.1791, ..., -0.2846, 0.0588, -0.2584], + [ 0.1792, 0.1123, -0.1760, ..., 0.1459, -0.2839, -0.3450]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 5.8208e-11, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 5.8208e-11, 1.1642e-10, ..., 0.0000e+00, + 5.8208e-11, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1642e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.8208e-11, 5.8208e-11, -2.3283e-10, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 7.5670e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [-9.8371e-09, -8.8476e-09, 8.7311e-10, ..., -4.8894e-09, + 6.4028e-10, 0.0000e+00]], device='cuda:0') +Epoch 499, bias, value: tensor([-4.0610e-02, -4.2564e-02, -8.1439e-03, -5.8141e-03, -8.4912e-05, + 6.2664e-03, 7.2158e-03, -6.9572e-03, 1.2412e-02, -4.5755e-02], + device='cuda:0'), grad: tensor([ 2.9104e-10, 1.1642e-09, 5.8208e-10, 6.9849e-10, 3.1258e-08, + 3.4925e-10, 9.3132e-10, -2.4447e-09, 2.6776e-09, -2.5670e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 217.70, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4026 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.25 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.1337, 0.0236, 0.1743, ..., 0.0100, -0.1426, -0.1142], + [-0.2591, -0.2593, -0.2390, ..., 0.1291, 0.0763, 0.4563], + [-0.3007, -0.2048, -0.2713, ..., -0.3395, 0.0138, -0.2696], + ..., + [-0.2338, -0.2955, -0.0146, ..., -0.2932, -0.1729, -0.5202], + [ 0.1671, 0.0378, -0.1791, ..., -0.2847, 0.0589, -0.2584], + [ 0.1793, 0.1123, -0.1760, ..., 0.1460, -0.2840, -0.3450]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9791e-09, ..., -1.1059e-09, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 5.8208e-11, 1.7462e-10, ..., -1.6298e-09, + -1.4552e-09, -9.2550e-09], + [ 5.8208e-11, 1.7462e-10, 2.9104e-10, ..., 8.1491e-10, + 3.4925e-10, 2.7358e-09], + ..., + [ 5.8208e-11, 5.8208e-11, 0.0000e+00, ..., 1.5716e-09, + 1.3388e-09, 7.7998e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 1.1642e-10, 3.4925e-10], + [ 5.8208e-11, 5.8208e-11, 1.7462e-10, ..., 2.3283e-10, + 1.7462e-10, 5.2387e-10]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0406, -0.0424, -0.0082, -0.0059, -0.0001, 0.0062, 0.0072, -0.0071, + 0.0125, -0.0458], device='cuda:0'), grad: tensor([-3.4925e-09, -1.4203e-08, 5.5297e-09, 4.0745e-10, 1.3388e-09, + 1.3388e-09, 7.5670e-09, 1.2747e-08, 8.1491e-10, 1.7462e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 217.48, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4309 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.27 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps4', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps4/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.089996 99.209999 ... 86.945686 69.764622 +ShearY 98.669998 98.909996 ... 86.945686 67.276324 +AutoContrast 99.089996 99.129997 ... 86.945686 61.274684 +Invert 98.720001 93.839996 ... 86.945686 50.101730 +Equalize 98.220001 98.409996 ... 86.945686 73.226051 +Solarize 98.059998 97.689995 ... 86.945686 63.925620 +SolarizeAdd 98.250000 97.829994 ... 86.945686 70.808240 +Posterize 99.010002 99.159996 ... 86.945686 73.015031 +Contrast 98.989998 99.260002 ... 86.945686 67.801760 +Color 99.049995 99.269997 ... 86.945686 59.026607 +Brightness 98.839996 99.279999 ... 86.945686 66.790533 +Sharpness 99.070000 99.209999 ... 86.945686 69.948109 +NoiseSalt 99.159996 99.220001 ... 86.945686 56.808372 +NoiseGaussian 99.070000 99.279999 ... 86.945686 58.316548 +w/o do (original x) 99.270000 0.000000 ... 0.000000 73.227632 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.21 68.461893 78.657927 76.499529 86.098655 77.429501 diff --git a/Meta-causal/code-withStyleAttack/66567.error b/Meta-causal/code-withStyleAttack/66567.error new file mode 100644 index 0000000000000000000000000000000000000000..5b09d3d759b1a068653d824f401cf6b8e10ef88f --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66567.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 37: eduler: command not found diff --git a/Meta-causal/code-withStyleAttack/66567.log b/Meta-causal/code-withStyleAttack/66567.log new file mode 100644 index 0000000000000000000000000000000000000000..70dd75ed682b118b030b317fe053d42981749b47 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66567.log @@ -0,0 +1,14349 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps4', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0271, 0.0207, 0.0215, ..., -0.0289, 0.0029, -0.0146], + [-0.0148, 0.0150, -0.0160, ..., -0.0022, 0.0053, 0.0148], + [-0.0263, -0.0228, -0.0024, ..., 0.0142, -0.0181, 0.0305], + ..., + [ 0.0153, 0.0045, 0.0014, ..., 0.0168, 0.0223, 0.0115], + [ 0.0200, 0.0085, 0.0071, ..., 0.0254, 0.0155, 0.0160], + [-0.0279, 0.0077, -0.0190, ..., 0.0294, -0.0245, -0.0030]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0260, -0.0171, 0.0127, -0.0213, 0.0291, 0.0182, 0.0038, -0.0096, + -0.0190, -0.0048], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 222.47, cls_loss 1.1909 cls_loss_mapping 1.7588 cls_loss_causal 2.2050 re_mapping 0.1691 re_causal 0.1843 /// teacc 89.91 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0326, 0.0279, 0.0215, ..., -0.0361, 0.0006, -0.0122], + [-0.0113, 0.0079, -0.0160, ..., 0.0018, 0.0068, 0.0058], + [-0.0321, -0.0242, -0.0024, ..., 0.0165, -0.0124, 0.0228], + ..., + [ 0.0152, 0.0090, 0.0014, ..., 0.0114, 0.0157, 0.0141], + [ 0.0207, 0.0022, 0.0071, ..., 0.0225, 0.0163, 0.0124], + [-0.0240, 0.0069, -0.0190, ..., 0.0340, -0.0324, 0.0001]], + device='cuda:0'), grad: tensor([[ 0.0077, -0.0236, 0.0000, ..., 0.0048, -0.0097, -0.0010], + [ 0.0134, 0.0004, 0.0000, ..., 0.0023, 0.0079, 0.0174], + [ 0.0138, 0.0122, 0.0000, ..., 0.0147, 0.0112, 0.0086], + ..., + [-0.0009, -0.0062, 0.0000, ..., -0.0022, 0.0072, -0.0067], + [ 0.0082, 0.0028, 0.0000, ..., 0.0154, 0.0054, 0.0232], + [-0.0198, 0.0028, 0.0000, ..., -0.0201, -0.0074, -0.0303]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0228, -0.0146, 0.0122, -0.0209, 0.0288, 0.0188, 0.0038, -0.0104, + -0.0199, -0.0033], device='cuda:0'), grad: tensor([ 0.0010, 0.0184, 0.0276, -0.0201, -0.0305, -0.0240, 0.0409, -0.0018, + 0.0214, -0.0328], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 222.00, cls_loss 0.3407 cls_loss_mapping 0.7050 cls_loss_causal 1.9079 re_mapping 0.2080 re_causal 0.2844 /// teacc 92.91 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0358, 0.0322, 0.0215, ..., -0.0381, -0.0007, -0.0125], + [-0.0099, 0.0067, -0.0160, ..., 0.0039, 0.0045, 0.0019], + [-0.0357, -0.0267, -0.0024, ..., 0.0153, -0.0102, 0.0202], + ..., + [ 0.0149, 0.0123, 0.0014, ..., 0.0080, 0.0149, 0.0163], + [ 0.0217, -0.0006, 0.0071, ..., 0.0202, 0.0183, 0.0095], + [-0.0267, 0.0057, -0.0190, ..., 0.0360, -0.0359, 0.0032]], + device='cuda:0'), grad: tensor([[ 2.3308e-03, -7.9575e-03, 0.0000e+00, ..., 1.4915e-03, + -4.8661e-04, 1.0872e-03], + [ 4.4365e-03, 2.1458e-04, 0.0000e+00, ..., 2.0695e-04, + 2.3708e-03, 2.4624e-03], + [-1.2329e-02, 2.8343e-03, 0.0000e+00, ..., -1.1711e-03, + -1.9417e-03, 4.4594e-03], + ..., + [-6.1035e-03, -6.1684e-03, 0.0000e+00, ..., 3.7308e-03, + 4.6539e-03, 1.9140e-03], + [ 2.2400e-02, 9.2936e-04, 0.0000e+00, ..., 3.5534e-03, + 2.3961e-04, -5.1193e-03], + [-4.1008e-05, 3.6716e-03, 0.0000e+00, ..., -1.6174e-02, + 3.4447e-03, -1.3779e-02]], device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0225, -0.0144, 0.0119, -0.0210, 0.0290, 0.0191, 0.0033, -0.0110, + -0.0196, -0.0029], device='cuda:0'), grad: tensor([-0.0027, 0.0048, -0.0056, -0.0110, 0.0160, -0.0029, -0.0032, 0.0054, + 0.0090, -0.0099], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 222.00, cls_loss 0.2320 cls_loss_mapping 0.4427 cls_loss_causal 1.6614 re_mapping 0.1442 re_causal 0.2336 /// teacc 95.20 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0380, 0.0355, 0.0215, ..., -0.0396, -0.0006, -0.0129], + [-0.0085, 0.0056, -0.0160, ..., 0.0053, 0.0039, -0.0004], + [-0.0376, -0.0298, -0.0024, ..., 0.0145, -0.0091, 0.0186], + ..., + [ 0.0152, 0.0148, 0.0014, ..., 0.0064, 0.0162, 0.0176], + [ 0.0216, -0.0013, 0.0071, ..., 0.0192, 0.0179, 0.0079], + [-0.0292, 0.0044, -0.0190, ..., 0.0368, -0.0395, 0.0051]], + device='cuda:0'), grad: tensor([[ 0.0021, -0.0089, 0.0000, ..., 0.0013, -0.0018, 0.0009], + [-0.0088, 0.0003, 0.0000, ..., -0.0040, -0.0026, 0.0010], + [ 0.0073, 0.0024, 0.0000, ..., 0.0046, -0.0022, 0.0019], + ..., + [ 0.0068, 0.0019, 0.0000, ..., 0.0227, 0.0046, 0.0399], + [ 0.0047, 0.0007, 0.0000, ..., 0.0109, 0.0025, 0.0096], + [-0.0223, -0.0034, 0.0000, ..., -0.0535, -0.0038, -0.0618]], + device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0225, -0.0145, 0.0122, -0.0210, 0.0290, 0.0192, 0.0031, -0.0112, + -0.0196, -0.0027], device='cuda:0'), grad: tensor([-0.0027, -0.0086, 0.0042, -0.0007, 0.0146, 0.0085, -0.0024, 0.0344, + 0.0138, -0.0609], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 222.12, cls_loss 0.1640 cls_loss_mapping 0.3067 cls_loss_causal 1.5048 re_mapping 0.1139 re_causal 0.2005 /// teacc 95.96 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0401, 0.0390, 0.0215, ..., -0.0411, -0.0011, -0.0130], + [-0.0080, 0.0043, -0.0160, ..., 0.0066, 0.0038, -0.0017], + [-0.0384, -0.0314, -0.0024, ..., 0.0140, -0.0083, 0.0172], + ..., + [ 0.0151, 0.0168, 0.0014, ..., 0.0043, 0.0181, 0.0185], + [ 0.0217, -0.0033, 0.0071, ..., 0.0180, 0.0178, 0.0068], + [-0.0305, 0.0029, -0.0190, ..., 0.0383, -0.0430, 0.0064]], + device='cuda:0'), grad: tensor([[ 0.0030, 0.0018, 0.0000, ..., 0.0017, 0.0019, 0.0055], + [-0.0059, 0.0001, 0.0000, ..., -0.0061, -0.0265, 0.0037], + [ 0.0032, 0.0016, 0.0000, ..., 0.0042, 0.0309, 0.0083], + ..., + [ 0.0005, 0.0014, 0.0000, ..., 0.0061, 0.0051, 0.0076], + [-0.0012, -0.0092, 0.0000, ..., -0.0075, -0.0070, -0.0121], + [ 0.0052, 0.0022, 0.0000, ..., 0.0061, 0.0097, 0.0072]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0228, -0.0143, 0.0121, -0.0211, 0.0292, 0.0190, 0.0028, -0.0114, + -0.0195, -0.0025], device='cuda:0'), grad: tensor([ 0.0086, -0.0283, 0.0347, -0.0283, -0.0015, 0.0189, -0.0077, 0.0100, + -0.0221, 0.0156], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 222.00, cls_loss 0.1243 cls_loss_mapping 0.2199 cls_loss_causal 1.3502 re_mapping 0.0925 re_causal 0.1784 /// teacc 96.79 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0407, 0.0416, 0.0215, ..., -0.0420, -0.0016, -0.0131], + [-0.0067, 0.0029, -0.0160, ..., 0.0073, 0.0043, -0.0032], + [-0.0398, -0.0331, -0.0024, ..., 0.0132, -0.0078, 0.0158], + ..., + [ 0.0160, 0.0179, 0.0014, ..., 0.0028, 0.0188, 0.0187], + [ 0.0222, -0.0044, 0.0071, ..., 0.0173, 0.0172, 0.0059], + [-0.0323, 0.0021, -0.0190, ..., 0.0391, -0.0461, 0.0078]], + device='cuda:0'), grad: tensor([[-4.9706e-03, -8.4686e-03, 0.0000e+00, ..., -2.0862e-05, + 5.5885e-04, -7.4577e-03], + [ 9.9564e-04, 1.8239e-04, 0.0000e+00, ..., -2.9945e-04, + 2.3899e-03, 2.3460e-03], + [ 1.3552e-03, 1.9455e-03, 0.0000e+00, ..., 2.3518e-03, + 2.0008e-03, 4.8714e-03], + ..., + [-5.6314e-04, -9.5510e-04, 0.0000e+00, ..., 1.1177e-03, + -2.4147e-03, -3.9864e-03], + [ 6.9332e-04, 1.4420e-03, 0.0000e+00, ..., 2.3537e-03, + -6.9275e-03, -7.4043e-03], + [ 6.1722e-03, 2.6741e-03, 0.0000e+00, ..., -6.6681e-03, + 4.5166e-03, 4.8904e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0231, -0.0141, 0.0120, -0.0207, 0.0293, 0.0185, 0.0024, -0.0117, + -0.0193, -0.0025], device='cuda:0'), grad: tensor([-0.0271, 0.0043, 0.0108, 0.0197, 0.0048, -0.0041, 0.0053, -0.0024, + -0.0189, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 222.09, cls_loss 0.1065 cls_loss_mapping 0.1843 cls_loss_causal 1.2670 re_mapping 0.0778 re_causal 0.1577 /// teacc 97.39 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0426, 0.0440, 0.0215, ..., -0.0440, -0.0018, -0.0133], + [-0.0069, 0.0020, -0.0160, ..., 0.0081, 0.0038, -0.0046], + [-0.0406, -0.0340, -0.0024, ..., 0.0130, -0.0071, 0.0148], + ..., + [ 0.0165, 0.0187, 0.0014, ..., 0.0015, 0.0192, 0.0196], + [ 0.0227, -0.0054, 0.0071, ..., 0.0166, 0.0170, 0.0050], + [-0.0335, 0.0012, -0.0190, ..., 0.0403, -0.0486, 0.0086]], + device='cuda:0'), grad: tensor([[ 0.0007, -0.0082, 0.0000, ..., -0.0015, -0.0017, -0.0008], + [-0.0002, 0.0002, 0.0000, ..., -0.0008, 0.0006, 0.0001], + [ 0.0011, 0.0014, 0.0000, ..., 0.0012, -0.0013, 0.0003], + ..., + [ 0.0005, 0.0010, 0.0000, ..., 0.0010, 0.0014, 0.0004], + [-0.0048, 0.0004, 0.0000, ..., 0.0007, -0.0009, 0.0001], + [ 0.0003, 0.0010, 0.0000, ..., -0.0010, 0.0003, -0.0006]], + device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0229, -0.0146, 0.0124, -0.0205, 0.0291, 0.0183, 0.0024, -0.0118, + -0.0191, -0.0024], device='cuda:0'), grad: tensor([-0.0069, 0.0006, -0.0007, -0.0041, 0.0023, 0.0112, -0.0027, 0.0025, + -0.0024, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 222.37, cls_loss 0.0928 cls_loss_mapping 0.1614 cls_loss_causal 1.2338 re_mapping 0.0651 re_causal 0.1447 /// teacc 97.51 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0441, 0.0462, 0.0215, ..., -0.0454, -0.0017, -0.0132], + [-0.0065, 0.0005, -0.0160, ..., 0.0085, 0.0040, -0.0059], + [-0.0414, -0.0361, -0.0024, ..., 0.0124, -0.0070, 0.0137], + ..., + [ 0.0173, 0.0192, 0.0014, ..., 0.0007, 0.0207, 0.0205], + [ 0.0229, -0.0051, 0.0071, ..., 0.0162, 0.0169, 0.0043], + [-0.0345, 0.0012, -0.0190, ..., 0.0413, -0.0513, 0.0095]], + device='cuda:0'), grad: tensor([[ 3.0446e-04, -1.6165e-03, 0.0000e+00, ..., 3.4046e-04, + -1.3649e-04, -1.8895e-04], + [-7.6771e-04, 5.5313e-05, 0.0000e+00, ..., -8.4114e-04, + 3.6383e-04, 2.1732e-04], + [-2.6398e-03, 2.6870e-04, 0.0000e+00, ..., 3.9434e-04, + -6.9351e-03, 5.4455e-04], + ..., + [ 1.1574e-02, 4.8971e-04, 0.0000e+00, ..., 1.8814e-02, + 1.2941e-03, 2.9160e-02], + [ 3.5267e-03, 2.2662e-04, 0.0000e+00, ..., 1.9932e-03, + 5.0011e-03, 1.0605e-03], + [-1.0948e-02, -2.4959e-05, 0.0000e+00, ..., -2.2629e-02, + -1.6987e-04, -3.2471e-02]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0230, -0.0145, 0.0122, -0.0205, 0.0289, 0.0178, 0.0023, -0.0113, + -0.0188, -0.0024], device='cuda:0'), grad: tensor([-0.0002, -0.0002, -0.0075, 0.0015, 0.0019, -0.0008, -0.0003, 0.0285, + 0.0059, -0.0288], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 222.24, cls_loss 0.0855 cls_loss_mapping 0.1470 cls_loss_causal 1.1677 re_mapping 0.0560 re_causal 0.1304 /// teacc 97.79 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0450, 0.0481, 0.0215, ..., -0.0463, -0.0028, -0.0126], + [-0.0063, -0.0007, -0.0160, ..., 0.0092, 0.0037, -0.0069], + [-0.0412, -0.0376, -0.0024, ..., 0.0121, -0.0061, 0.0128], + ..., + [ 0.0174, 0.0196, 0.0014, ..., -0.0004, 0.0206, 0.0207], + [ 0.0230, -0.0060, 0.0071, ..., 0.0155, 0.0167, 0.0035], + [-0.0357, 0.0005, -0.0190, ..., 0.0418, -0.0533, 0.0103]], + device='cuda:0'), grad: tensor([[ 1.1559e-03, -4.4346e-04, 0.0000e+00, ..., 9.0504e-04, + 8.9550e-04, 1.1873e-03], + [-1.2362e-04, 9.0301e-05, 0.0000e+00, ..., 7.8559e-05, + 1.0328e-03, 1.0757e-03], + [ 4.4203e-04, 2.8157e-04, 0.0000e+00, ..., -2.6894e-04, + 5.2605e-03, 2.0447e-03], + ..., + [-1.2188e-03, -1.3704e-03, 0.0000e+00, ..., 2.8839e-03, + -1.3542e-03, -1.2016e-03], + [ 1.1787e-03, 4.7177e-05, 0.0000e+00, ..., 2.5387e-03, + 1.3018e-03, 1.4200e-03], + [ 1.0376e-03, 5.7268e-04, 0.0000e+00, ..., -3.2825e-03, + 1.0462e-03, -3.1834e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0231, -0.0146, 0.0128, -0.0204, 0.0290, 0.0177, 0.0019, -0.0118, + -0.0184, -0.0025], device='cuda:0'), grad: tensor([ 0.0026, 0.0014, 0.0020, -0.0085, -0.0035, 0.0015, 0.0012, 0.0005, + 0.0035, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 222.05, cls_loss 0.0748 cls_loss_mapping 0.1289 cls_loss_causal 1.1284 re_mapping 0.0504 re_causal 0.1194 /// teacc 97.84 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0464, 0.0494, 0.0215, ..., -0.0470, -0.0029, -0.0125], + [-0.0058, -0.0019, -0.0160, ..., 0.0099, 0.0036, -0.0075], + [-0.0423, -0.0398, -0.0024, ..., 0.0115, -0.0058, 0.0116], + ..., + [ 0.0181, 0.0202, 0.0014, ..., -0.0012, 0.0211, 0.0212], + [ 0.0232, -0.0065, 0.0071, ..., 0.0150, 0.0165, 0.0024], + [-0.0369, 0.0009, -0.0190, ..., 0.0428, -0.0554, 0.0110]], + device='cuda:0'), grad: tensor([[-1.7166e-03, -7.5150e-03, 0.0000e+00, ..., 2.3043e-04, + -2.4414e-03, -3.3408e-05], + [-5.3406e-04, 3.9124e-04, 0.0000e+00, ..., -1.1396e-03, + 6.8712e-04, 5.7364e-04], + [ 2.0676e-03, 4.6806e-03, 0.0000e+00, ..., 8.5783e-04, + 1.7509e-03, 5.2261e-04], + ..., + [ 5.0640e-04, 2.5964e-04, 0.0000e+00, ..., 1.4687e-03, + 7.4768e-04, 1.5736e-03], + [ 7.0715e-04, 7.4053e-04, 0.0000e+00, ..., 2.1992e-03, + 1.9531e-03, 2.8839e-03], + [ 6.6102e-05, 1.9240e-04, 0.0000e+00, ..., -5.3558e-03, + 2.6345e-04, -3.4599e-03]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0230, -0.0147, 0.0129, -0.0204, 0.0288, 0.0175, 0.0022, -0.0118, + -0.0183, -0.0023], device='cuda:0'), grad: tensor([-0.0118, 0.0004, 0.0085, -0.0050, -0.0004, 0.0026, 0.0023, 0.0028, + 0.0051, -0.0045], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 222.12, cls_loss 0.0710 cls_loss_mapping 0.1232 cls_loss_causal 1.0820 re_mapping 0.0462 re_causal 0.1106 /// teacc 97.88 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0472, 0.0510, 0.0215, ..., -0.0474, -0.0029, -0.0123], + [-0.0059, -0.0028, -0.0160, ..., 0.0103, 0.0033, -0.0087], + [-0.0430, -0.0412, -0.0024, ..., 0.0111, -0.0059, 0.0106], + ..., + [ 0.0185, 0.0208, 0.0014, ..., -0.0021, 0.0222, 0.0220], + [ 0.0236, -0.0071, 0.0071, ..., 0.0143, 0.0161, 0.0017], + [-0.0375, 0.0004, -0.0190, ..., 0.0436, -0.0577, 0.0117]], + device='cuda:0'), grad: tensor([[ 2.8419e-04, -1.2035e-03, 0.0000e+00, ..., 2.3735e-04, + 3.0175e-05, 8.2076e-05], + [ 2.9812e-03, 2.7552e-05, 0.0000e+00, ..., 1.2207e-04, + 1.2703e-03, 1.1188e-04], + [ 5.9891e-04, 3.1257e-04, 0.0000e+00, ..., 1.5879e-04, + 4.6194e-05, 1.2898e-04], + ..., + [ 6.1131e-04, 1.3542e-04, 0.0000e+00, ..., 1.8034e-03, + 2.9278e-04, 1.5326e-03], + [-9.4299e-03, 1.1575e-04, 0.0000e+00, ..., 3.1018e-04, + -4.1161e-03, -3.9399e-05], + [ 7.4100e-04, 1.1611e-04, 0.0000e+00, ..., -2.9678e-03, + 2.0480e-04, -3.5019e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0233, -0.0151, 0.0128, -0.0199, 0.0287, 0.0173, 0.0018, -0.0117, + -0.0183, -0.0024], device='cuda:0'), grad: tensor([ 6.6876e-05, 4.1542e-03, 8.0156e-04, -4.5052e-03, 3.0577e-05, + 1.4893e-02, -2.6608e-03, 2.4548e-03, -1.3496e-02, -1.7366e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 222.41, cls_loss 0.0545 cls_loss_mapping 0.0974 cls_loss_causal 1.0696 re_mapping 0.0414 re_causal 0.1082 /// teacc 98.07 lr 0.00010000 +Epoch 12, weight, value: tensor([[-4.8286e-02, 5.2557e-02, 2.1474e-02, ..., -4.7808e-02, + -3.0306e-03, -1.2115e-02], + [-5.4672e-03, -3.8802e-03, -1.6035e-02, ..., 1.0723e-02, + 3.2854e-03, -9.2491e-03], + [-4.3902e-02, -4.2428e-02, -2.3694e-03, ..., 1.0608e-02, + -6.0441e-03, 9.4452e-03], + ..., + [ 1.8583e-02, 2.1708e-02, 1.4217e-03, ..., -3.4607e-03, + 2.2753e-02, 2.2221e-02], + [ 2.4385e-02, -7.3145e-03, 7.1494e-03, ..., 1.3876e-02, + 1.6189e-02, 1.0914e-03], + [-3.8162e-02, -6.2173e-05, -1.9007e-02, ..., 4.4448e-02, + -5.9792e-02, 1.2404e-02]], device='cuda:0'), grad: tensor([[-7.8630e-04, -2.9335e-03, 0.0000e+00, ..., 5.9891e-04, + -1.6081e-04, -3.9520e-03], + [-1.4770e-04, 5.3793e-05, 0.0000e+00, ..., -6.9761e-04, + 6.5517e-04, 1.2672e-04], + [ 1.0176e-03, 3.5095e-04, 0.0000e+00, ..., 9.7847e-04, + -2.0294e-03, 7.2956e-04], + ..., + [-4.4464e-02, -1.1883e-03, 0.0000e+00, ..., -2.3727e-02, + -1.9264e-03, -2.0065e-03], + [-1.4400e-03, 6.2799e-04, 0.0000e+00, ..., 1.2726e-02, + 4.1795e-04, 1.1139e-02], + [ 1.5688e-03, 1.2665e-03, 0.0000e+00, ..., -1.5625e-02, + 8.6975e-04, -1.2680e-02]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0233, -0.0150, 0.0128, -0.0200, 0.0286, 0.0174, 0.0017, -0.0120, + -0.0179, -0.0023], device='cuda:0'), grad: tensor([-0.0041, 0.0007, -0.0025, 0.0027, 0.0276, 0.0106, 0.0019, -0.0316, + 0.0140, -0.0193], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 222.50, cls_loss 0.0579 cls_loss_mapping 0.1039 cls_loss_causal 1.0564 re_mapping 0.0377 re_causal 0.1009 /// teacc 98.28 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0492, 0.0537, 0.0215, ..., -0.0487, -0.0036, -0.0120], + [-0.0054, -0.0046, -0.0160, ..., 0.0110, 0.0028, -0.0103], + [-0.0445, -0.0434, -0.0024, ..., 0.0101, -0.0055, 0.0088], + ..., + [ 0.0194, 0.0224, 0.0014, ..., -0.0040, 0.0231, 0.0227], + [ 0.0244, -0.0078, 0.0071, ..., 0.0133, 0.0160, 0.0001], + [-0.0391, -0.0004, -0.0190, ..., 0.0454, -0.0618, 0.0131]], + device='cuda:0'), grad: tensor([[ 5.8603e-04, -4.7159e-04, 0.0000e+00, ..., 1.7023e-04, + 1.1712e-04, 5.8472e-05], + [ 1.5383e-03, 2.5958e-05, 0.0000e+00, ..., 1.2159e-04, + 7.6389e-04, 1.5078e-03], + [ 1.0037e-04, 9.7871e-05, 0.0000e+00, ..., 1.5783e-04, + -9.9373e-04, 3.2759e-04], + ..., + [-6.1188e-03, -1.3602e-04, 0.0000e+00, ..., -1.4839e-03, + -1.4687e-03, -3.5534e-03], + [-1.4982e-03, 8.7202e-05, 0.0000e+00, ..., 1.2045e-03, + 8.2874e-04, 5.2834e-04], + [ 3.0785e-03, 1.3697e-04, 0.0000e+00, ..., 1.4362e-03, + 5.7268e-04, 1.6379e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0230, -0.0153, 0.0131, -0.0199, 0.0285, 0.0174, 0.0014, -0.0118, + -0.0179, -0.0022], device='cuda:0'), grad: tensor([ 5.5790e-04, 2.6817e-03, -2.3556e-03, 1.1420e-04, -1.3399e-04, + 2.9736e-03, 7.7486e-05, -6.6147e-03, -2.1672e-04, 2.9163e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 221.59, cls_loss 0.0477 cls_loss_mapping 0.0857 cls_loss_causal 1.0102 re_mapping 0.0346 re_causal 0.0943 /// teacc 98.28 lr 0.00010000 +Epoch 14, weight, value: tensor([[-4.9821e-02, 5.4745e-02, 2.1474e-02, ..., -4.8854e-02, + -4.3319e-03, -1.2319e-02], + [-5.0357e-03, -5.4368e-03, -1.6035e-02, ..., 1.1131e-02, + 2.3343e-03, -1.0792e-02], + [-4.4913e-02, -4.4125e-02, -2.3694e-03, ..., 9.4166e-03, + -5.3669e-03, 7.5606e-03], + ..., + [ 2.0087e-02, 2.2409e-02, 1.4217e-03, ..., -4.8106e-03, + 2.4123e-02, 2.3376e-02], + [ 2.4623e-02, -7.8029e-03, 7.1494e-03, ..., 1.3405e-02, + 1.5700e-02, 1.1892e-05], + [-3.9960e-02, -8.8761e-04, -1.9007e-02, ..., 4.6218e-02, + -6.3923e-02, 1.3724e-02]], device='cuda:0'), grad: tensor([[ 3.7622e-04, -1.0414e-03, 0.0000e+00, ..., 2.5344e-04, + -4.8709e-04, -2.1970e-04], + [ 3.5381e-04, 1.8328e-05, 0.0000e+00, ..., -1.6522e-04, + 9.1851e-05, 1.8048e-04], + [ 2.0180e-03, 6.8855e-04, 0.0000e+00, ..., 1.0653e-03, + 9.4032e-04, 8.0681e-04], + ..., + [ 6.7472e-04, 3.7163e-05, 0.0000e+00, ..., 7.4720e-04, + -4.4894e-04, -8.6975e-04], + [-2.4338e-03, 2.7761e-05, 0.0000e+00, ..., -8.1420e-05, + -3.1509e-03, -3.7842e-03], + [ 4.0054e-03, 6.0081e-05, 0.0000e+00, ..., 3.3398e-03, + 2.3174e-03, 3.6068e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0228, -0.0154, 0.0130, -0.0199, 0.0285, 0.0174, 0.0014, -0.0115, + -0.0178, -0.0022], device='cuda:0'), grad: tensor([-0.0003, 0.0005, 0.0033, 0.0018, -0.0029, 0.0081, -0.0123, 0.0002, + -0.0069, 0.0084], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 222.17, cls_loss 0.0438 cls_loss_mapping 0.0797 cls_loss_causal 0.9708 re_mapping 0.0332 re_causal 0.0900 /// teacc 98.30 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0506, 0.0561, 0.0215, ..., -0.0489, -0.0047, -0.0122], + [-0.0048, -0.0062, -0.0160, ..., 0.0115, 0.0021, -0.0112], + [-0.0458, -0.0449, -0.0024, ..., 0.0089, -0.0052, 0.0069], + ..., + [ 0.0208, 0.0225, 0.0014, ..., -0.0055, 0.0247, 0.0235], + [ 0.0251, -0.0082, 0.0071, ..., 0.0127, 0.0157, -0.0005], + [-0.0406, -0.0011, -0.0190, ..., 0.0465, -0.0657, 0.0144]], + device='cuda:0'), grad: tensor([[ 1.8919e-04, -6.3705e-04, 0.0000e+00, ..., 5.1141e-05, + 2.7791e-05, -3.0637e-05], + [-7.7438e-04, 4.4890e-06, 0.0000e+00, ..., -8.8406e-04, + 6.4850e-04, 6.0463e-04], + [ 7.5054e-04, 7.9751e-05, 0.0000e+00, ..., 1.1158e-04, + 2.0142e-02, 4.3144e-03], + ..., + [-3.8300e-03, 2.0653e-05, 0.0000e+00, ..., 2.5392e-04, + -2.7054e-02, -1.1108e-02], + [ 2.2526e-03, 1.1462e-04, 0.0000e+00, ..., 1.1015e-03, + 2.4681e-03, 3.1166e-03], + [ 7.1526e-04, 1.2141e-04, 0.0000e+00, ..., -1.1072e-03, + 1.6623e-03, 8.7786e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0228, -0.0156, 0.0129, -0.0198, 0.0284, 0.0171, 0.0016, -0.0115, + -0.0176, -0.0021], device='cuda:0'), grad: tensor([-1.4365e-04, 4.1783e-05, 3.2684e-02, 3.5286e-03, 5.2834e-04, + 2.2793e-03, -2.4261e-03, -4.4861e-02, 6.3324e-03, 2.0504e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 222.14, cls_loss 0.0416 cls_loss_mapping 0.0777 cls_loss_causal 0.9779 re_mapping 0.0303 re_causal 0.0858 /// teacc 98.45 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0518, 0.0571, 0.0215, ..., -0.0493, -0.0051, -0.0120], + [-0.0044, -0.0068, -0.0160, ..., 0.0120, 0.0021, -0.0113], + [-0.0459, -0.0458, -0.0024, ..., 0.0086, -0.0052, 0.0059], + ..., + [ 0.0212, 0.0229, 0.0014, ..., -0.0065, 0.0252, 0.0237], + [ 0.0253, -0.0090, 0.0071, ..., 0.0122, 0.0154, -0.0012], + [-0.0413, -0.0016, -0.0190, ..., 0.0469, -0.0675, 0.0150]], + device='cuda:0'), grad: tensor([[-5.5981e-04, -6.2103e-03, 0.0000e+00, ..., -6.6185e-04, + -3.6621e-04, -2.8563e-04], + [ 1.6737e-04, 6.2227e-05, 0.0000e+00, ..., -6.5708e-04, + -3.1686e-04, 6.6102e-05], + [ 4.8876e-04, 6.6948e-04, 0.0000e+00, ..., 3.3236e-04, + 2.5439e-04, 1.2720e-04], + ..., + [-5.6362e-04, 4.4823e-05, 0.0000e+00, ..., 1.5497e-04, + -1.8382e-04, -5.2166e-04], + [-2.0618e-03, 9.7871e-05, 0.0000e+00, ..., -5.9605e-07, + 4.9889e-05, 2.5773e-04], + [ 8.7309e-04, 1.4079e-04, 0.0000e+00, ..., -7.6294e-04, + 1.8573e-04, -1.0133e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0228, -0.0155, 0.0130, -0.0197, 0.0287, 0.0171, 0.0017, -0.0114, + -0.0177, -0.0026], device='cuda:0'), grad: tensor([-0.0067, -0.0007, 0.0017, 0.0028, 0.0017, -0.0011, 0.0047, -0.0005, + -0.0022, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 222.53, cls_loss 0.0405 cls_loss_mapping 0.0761 cls_loss_causal 0.9295 re_mapping 0.0297 re_causal 0.0825 /// teacc 98.59 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0523, 0.0580, 0.0215, ..., -0.0499, -0.0047, -0.0117], + [-0.0043, -0.0073, -0.0160, ..., 0.0121, 0.0019, -0.0120], + [-0.0470, -0.0469, -0.0024, ..., 0.0080, -0.0056, 0.0049], + ..., + [ 0.0226, 0.0229, 0.0014, ..., -0.0071, 0.0263, 0.0239], + [ 0.0255, -0.0100, 0.0071, ..., 0.0118, 0.0153, -0.0015], + [-0.0422, -0.0010, -0.0190, ..., 0.0475, -0.0695, 0.0156]], + device='cuda:0'), grad: tensor([[ 8.8155e-05, -1.9526e-04, 0.0000e+00, ..., 8.3685e-05, + -2.6062e-05, -1.9297e-06], + [-1.2903e-03, 2.3749e-06, 0.0000e+00, ..., -1.6613e-03, + -1.2321e-03, -5.2071e-04], + [ 4.6945e-04, 2.9624e-05, 0.0000e+00, ..., 5.7316e-04, + 6.2990e-04, 3.0184e-04], + ..., + [ 3.5381e-04, -9.4771e-06, 0.0000e+00, ..., 5.7173e-04, + 3.7408e-04, 2.5773e-04], + [ 5.7077e-04, 5.0180e-06, 0.0000e+00, ..., 7.0620e-04, + -6.0558e-05, 2.9135e-04], + [ 2.7990e-04, 3.0577e-05, 0.0000e+00, ..., -7.7629e-04, + 2.3812e-05, -8.9455e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0229, -0.0156, 0.0126, -0.0197, 0.0284, 0.0168, 0.0020, -0.0111, + -0.0174, -0.0026], device='cuda:0'), grad: tensor([ 6.4373e-05, -4.6043e-03, 1.9817e-03, 4.8041e-04, 3.5477e-04, + -1.0061e-03, 1.5345e-03, 1.4563e-03, 3.6716e-04, -6.2704e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 221.28, cls_loss 0.0343 cls_loss_mapping 0.0651 cls_loss_causal 0.8965 re_mapping 0.0283 re_causal 0.0798 /// teacc 98.34 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0527, 0.0588, 0.0215, ..., -0.0506, -0.0049, -0.0118], + [-0.0037, -0.0080, -0.0160, ..., 0.0126, 0.0018, -0.0129], + [-0.0481, -0.0478, -0.0024, ..., 0.0075, -0.0058, 0.0043], + ..., + [ 0.0230, 0.0230, 0.0014, ..., -0.0075, 0.0267, 0.0242], + [ 0.0263, -0.0096, 0.0071, ..., 0.0112, 0.0155, -0.0017], + [-0.0429, -0.0007, -0.0190, ..., 0.0481, -0.0711, 0.0163]], + device='cuda:0'), grad: tensor([[ 2.7680e-04, -9.0241e-05, 0.0000e+00, ..., 1.3649e-04, + 1.2839e-04, 1.7035e-04], + [-1.6193e-03, 1.3016e-05, 0.0000e+00, ..., -1.9627e-03, + -1.2894e-03, -9.7466e-04], + [ 5.1737e-05, 2.6286e-05, 0.0000e+00, ..., 3.1304e-04, + 3.4630e-05, 4.5109e-04], + ..., + [ 2.1183e-04, -2.8920e-04, 0.0000e+00, ..., 8.9025e-04, + 6.5231e-04, 4.7350e-04], + [ 4.9162e-04, 2.8074e-05, 0.0000e+00, ..., 3.6812e-04, + 1.0860e-04, 1.1134e-04], + [ 2.0301e-04, 8.2552e-05, 0.0000e+00, ..., -7.5150e-04, + 6.5267e-05, -8.1539e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0227, -0.0154, 0.0123, -0.0197, 0.0284, 0.0170, 0.0017, -0.0111, + -0.0171, -0.0025], device='cuda:0'), grad: tensor([ 0.0005, -0.0044, -0.0004, 0.0010, 0.0007, 0.0195, -0.0193, 0.0019, + 0.0008, -0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 221.34, cls_loss 0.0330 cls_loss_mapping 0.0648 cls_loss_causal 0.9057 re_mapping 0.0267 re_causal 0.0775 /// teacc 98.37 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0533, 0.0599, 0.0215, ..., -0.0509, -0.0055, -0.0112], + [-0.0035, -0.0088, -0.0160, ..., 0.0126, 0.0022, -0.0127], + [-0.0483, -0.0486, -0.0024, ..., 0.0073, -0.0057, 0.0035], + ..., + [ 0.0229, 0.0232, 0.0014, ..., -0.0084, 0.0272, 0.0245], + [ 0.0266, -0.0100, 0.0071, ..., 0.0115, 0.0152, -0.0020], + [-0.0433, -0.0012, -0.0190, ..., 0.0485, -0.0726, 0.0166]], + device='cuda:0'), grad: tensor([[ 8.3208e-05, -2.7847e-04, 0.0000e+00, ..., 1.0031e-04, + -2.1145e-05, -3.6180e-05], + [-1.4601e-03, 8.8587e-06, 0.0000e+00, ..., -1.4009e-03, + 1.0538e-04, 5.8293e-05], + [ 5.8442e-05, 4.1664e-05, 0.0000e+00, ..., 1.3959e-04, + -4.9829e-04, 7.5877e-05], + ..., + [ 8.0228e-05, -3.6061e-05, 0.0000e+00, ..., 2.8801e-04, + -1.4579e-04, -1.0443e-04], + [ 9.4032e-04, 1.3448e-05, 0.0000e+00, ..., 1.4725e-03, + 3.3402e-04, 9.0075e-04], + [-5.1260e-04, 4.8816e-05, 0.0000e+00, ..., -1.8234e-03, + 1.4591e-04, -1.4648e-03]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0228, -0.0152, 0.0124, -0.0197, 0.0284, 0.0169, 0.0015, -0.0114, + -0.0169, -0.0026], device='cuda:0'), grad: tensor([-8.8960e-06, -2.0504e-03, -5.3930e-04, 1.5509e-04, 7.0858e-04, + 5.4073e-04, 5.7173e-04, 2.1386e-04, 2.5806e-03, -2.1744e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 221.23, cls_loss 0.0323 cls_loss_mapping 0.0640 cls_loss_causal 0.9150 re_mapping 0.0256 re_causal 0.0750 /// teacc 98.37 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0540, 0.0614, 0.0215, ..., -0.0511, -0.0053, -0.0108], + [-0.0037, -0.0097, -0.0160, ..., 0.0129, 0.0019, -0.0131], + [-0.0488, -0.0494, -0.0024, ..., 0.0071, -0.0054, 0.0028], + ..., + [ 0.0235, 0.0234, 0.0014, ..., -0.0092, 0.0278, 0.0250], + [ 0.0269, -0.0105, 0.0071, ..., 0.0111, 0.0148, -0.0027], + [-0.0439, -0.0014, -0.0190, ..., 0.0490, -0.0745, 0.0170]], + device='cuda:0'), grad: tensor([[ 1.9431e-04, -7.9632e-05, 0.0000e+00, ..., 1.0687e-04, + 9.5189e-05, 1.3041e-04], + [ 7.4615e-03, 6.5416e-06, 0.0000e+00, ..., 3.6836e-04, + 2.8362e-03, 2.3890e-04], + [ 5.5361e-04, 5.0366e-05, 0.0000e+00, ..., 3.7813e-04, + 1.7774e-04, 1.5891e-04], + ..., + [-8.9798e-03, -1.4806e-04, 0.0000e+00, ..., 1.3256e-03, + -5.0247e-05, 1.5821e-03], + [-8.9798e-03, 6.1765e-06, 0.0000e+00, ..., 7.0286e-04, + -3.2272e-03, 7.1526e-04], + [-1.5283e-04, 1.2457e-04, 0.0000e+00, ..., -6.4850e-03, + -5.3940e-03, -8.5220e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0232, -0.0155, 0.0127, -0.0198, 0.0283, 0.0169, 0.0013, -0.0112, + -0.0168, -0.0027], device='cuda:0'), grad: tensor([ 0.0004, 0.0104, 0.0010, 0.0084, 0.0078, 0.0022, 0.0010, -0.0053, + -0.0119, -0.0140], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 221.32, cls_loss 0.0314 cls_loss_mapping 0.0634 cls_loss_causal 0.8812 re_mapping 0.0250 re_causal 0.0728 /// teacc 98.28 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0548, 0.0624, 0.0215, ..., -0.0514, -0.0053, -0.0108], + [-0.0037, -0.0102, -0.0160, ..., 0.0130, 0.0019, -0.0137], + [-0.0496, -0.0507, -0.0024, ..., 0.0067, -0.0054, 0.0024], + ..., + [ 0.0241, 0.0236, 0.0014, ..., -0.0098, 0.0283, 0.0254], + [ 0.0273, -0.0107, 0.0071, ..., 0.0107, 0.0146, -0.0031], + [-0.0447, -0.0017, -0.0190, ..., 0.0496, -0.0756, 0.0176]], + device='cuda:0'), grad: tensor([[ 8.5950e-05, -8.2970e-04, 0.0000e+00, ..., 1.7285e-05, + -2.2995e-04, -1.2755e-04], + [-5.0497e-04, 5.1588e-05, 0.0000e+00, ..., -2.7871e-04, + 4.2200e-04, 1.3983e-04], + [ 2.2340e-04, 2.1231e-04, 0.0000e+00, ..., 7.3791e-05, + -4.5466e-04, 6.6817e-05], + ..., + [ 8.1968e-04, 2.7871e-04, 0.0000e+00, ..., 5.4789e-04, + 8.2493e-04, 1.0128e-03], + [ 3.2616e-04, 6.5327e-05, 0.0000e+00, ..., 1.8668e-04, + 5.3453e-04, 8.6737e-04], + [ 1.0222e-04, 6.4373e-05, 0.0000e+00, ..., -4.4322e-04, + 2.6169e-03, 2.4757e-03]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0229, -0.0156, 0.0125, -0.0203, 0.0284, 0.0171, 0.0013, -0.0108, + -0.0165, -0.0028], device='cuda:0'), grad: tensor([-5.7459e-04, -2.8920e-04, -4.3058e-04, -5.3139e-03, -5.7459e-05, + -3.5620e-04, 7.7963e-04, 2.1324e-03, 1.2169e-03, 2.8973e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 221.31, cls_loss 0.0286 cls_loss_mapping 0.0520 cls_loss_causal 0.9051 re_mapping 0.0227 re_causal 0.0712 /// teacc 98.38 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0553, 0.0633, 0.0215, ..., -0.0508, -0.0056, -0.0106], + [-0.0040, -0.0107, -0.0160, ..., 0.0130, 0.0018, -0.0142], + [-0.0504, -0.0512, -0.0024, ..., 0.0063, -0.0053, 0.0018], + ..., + [ 0.0240, 0.0238, 0.0014, ..., -0.0104, 0.0283, 0.0255], + [ 0.0276, -0.0101, 0.0071, ..., 0.0100, 0.0147, -0.0033], + [-0.0450, -0.0020, -0.0190, ..., 0.0501, -0.0772, 0.0183]], + device='cuda:0'), grad: tensor([[ 1.0121e-04, -3.4761e-04, 0.0000e+00, ..., 2.0516e-04, + -1.6773e-04, -2.4414e-04], + [-8.0526e-05, 1.8343e-05, 0.0000e+00, ..., -1.2636e-04, + 3.5524e-05, 2.8357e-05], + [ 5.1856e-05, 7.6771e-05, 0.0000e+00, ..., 1.0157e-04, + 7.2837e-05, 5.8979e-05], + ..., + [ 1.4350e-05, 2.2039e-05, 0.0000e+00, ..., 3.7402e-05, + 5.2154e-05, 7.1347e-05], + [ 2.4724e-04, 8.2850e-05, 0.0000e+00, ..., 1.8859e-04, + 6.2227e-05, 9.6202e-05], + [ 4.4048e-05, 8.2433e-05, 0.0000e+00, ..., -1.4269e-04, + 8.5771e-05, -1.0371e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0230, -0.0159, 0.0123, -0.0196, 0.0284, 0.0166, 0.0012, -0.0113, + -0.0164, -0.0023], device='cuda:0'), grad: tensor([-1.2124e-04, -1.5187e-04, 2.4748e-04, 2.1422e-04, 3.8338e-04, + 3.0369e-05, -1.3084e-03, 1.1969e-04, 5.8079e-04, 6.0089e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 222.31, cls_loss 0.0292 cls_loss_mapping 0.0582 cls_loss_causal 0.8706 re_mapping 0.0225 re_causal 0.0687 /// teacc 98.73 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0559, 0.0643, 0.0215, ..., -0.0511, -0.0058, -0.0103], + [-0.0042, -0.0112, -0.0160, ..., 0.0131, 0.0011, -0.0151], + [-0.0512, -0.0517, -0.0024, ..., 0.0060, -0.0053, 0.0011], + ..., + [ 0.0246, 0.0241, 0.0014, ..., -0.0111, 0.0289, 0.0259], + [ 0.0277, -0.0105, 0.0071, ..., 0.0095, 0.0144, -0.0039], + [-0.0454, -0.0025, -0.0190, ..., 0.0504, -0.0785, 0.0187]], + device='cuda:0'), grad: tensor([[ 3.8266e-05, -2.5773e-04, 0.0000e+00, ..., 3.2276e-05, + -3.5346e-05, -1.7092e-05], + [-2.1088e-04, 7.9274e-06, 0.0000e+00, ..., -9.8228e-05, + 3.1769e-05, 5.6475e-05], + [ 5.9277e-05, 5.8591e-05, 0.0000e+00, ..., 3.5971e-05, + -3.8433e-04, 4.5985e-05], + ..., + [-1.9813e-04, -4.6194e-05, 0.0000e+00, ..., 1.9395e-04, + 6.4850e-05, -1.6677e-04], + [ 9.6202e-05, 3.5912e-05, 0.0000e+00, ..., 8.6367e-05, + 8.2254e-05, 9.3400e-05], + [ 1.9336e-04, 5.8353e-05, 0.0000e+00, ..., -6.4659e-03, + -6.5422e-04, -7.2327e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0230, -0.0164, 0.0121, -0.0198, 0.0288, 0.0169, 0.0010, -0.0108, + -0.0163, -0.0025], device='cuda:0'), grad: tensor([-6.5863e-05, -1.8454e-04, -5.2881e-04, 1.1116e-02, 4.5443e-04, + 6.0976e-05, 9.3699e-05, 1.5509e-04, 2.7657e-04, -1.1375e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 221.47, cls_loss 0.0250 cls_loss_mapping 0.0526 cls_loss_causal 0.8822 re_mapping 0.0222 re_causal 0.0668 /// teacc 98.56 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0565, 0.0651, 0.0215, ..., -0.0512, -0.0053, -0.0103], + [-0.0039, -0.0118, -0.0160, ..., 0.0138, 0.0013, -0.0145], + [-0.0519, -0.0524, -0.0024, ..., 0.0053, -0.0054, 0.0005], + ..., + [ 0.0252, 0.0244, 0.0014, ..., -0.0118, 0.0289, 0.0261], + [ 0.0278, -0.0107, 0.0071, ..., 0.0089, 0.0143, -0.0046], + [-0.0458, -0.0027, -0.0190, ..., 0.0512, -0.0801, 0.0190]], + device='cuda:0'), grad: tensor([[ 2.4188e-04, 1.4031e-04, 0.0000e+00, ..., 1.1396e-04, + 3.1447e-04, 3.2377e-04], + [ 2.2447e-04, 3.3766e-05, 0.0000e+00, ..., 3.9935e-04, + 1.1253e-04, 2.0671e-04], + [ 9.2566e-05, 2.4021e-05, 0.0000e+00, ..., 1.1420e-04, + -7.1573e-04, 5.4866e-05], + ..., + [-6.8331e-04, -6.4993e-04, 0.0000e+00, ..., 6.1035e-04, + -1.4997e-04, -3.6240e-04], + [ 4.4942e-04, 7.1108e-05, 0.0000e+00, ..., 5.8222e-04, + 2.8491e-04, 2.4152e-04], + [-6.7997e-04, 2.6536e-04, 0.0000e+00, ..., -1.0887e-02, + 1.5414e-04, -3.8185e-03]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0229, -0.0159, 0.0119, -0.0198, 0.0286, 0.0167, 0.0014, -0.0109, + -0.0164, -0.0024], device='cuda:0'), grad: tensor([ 0.0010, 0.0007, -0.0013, 0.0015, 0.0087, -0.0009, -0.0012, -0.0007, + 0.0014, -0.0092], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 221.35, cls_loss 0.0205 cls_loss_mapping 0.0444 cls_loss_causal 0.8495 re_mapping 0.0209 re_causal 0.0642 /// teacc 98.60 lr 0.00010000 +Epoch 25, weight, value: tensor([[-5.7242e-02, 6.6043e-02, 2.1474e-02, ..., -5.1369e-02, + -4.6786e-03, -1.0276e-02], + [-3.5799e-03, -1.2495e-02, -1.6035e-02, ..., 1.4097e-02, + 8.7243e-04, -1.4583e-02], + [-5.2669e-02, -5.2875e-02, -2.3694e-03, ..., 4.7535e-03, + -5.3805e-03, -7.5696e-05], + ..., + [ 2.5545e-02, 2.4855e-02, 1.4217e-03, ..., -1.2643e-02, + 2.9314e-02, 2.6116e-02], + [ 2.8137e-02, -1.1221e-02, 7.1494e-03, ..., 8.7158e-03, + 1.4116e-02, -4.9226e-03], + [-4.6163e-02, -3.1379e-03, -1.9007e-02, ..., 5.1910e-02, + -8.1943e-02, 1.9632e-02]], device='cuda:0'), grad: tensor([[ 1.1671e-04, -1.8978e-04, 0.0000e+00, ..., 6.8784e-05, + 4.9782e-04, 2.1386e-04], + [-4.7874e-03, 8.8662e-06, 0.0000e+00, ..., -5.4626e-03, + -3.5114e-03, -6.2103e-03], + [ 3.0971e-04, 3.8087e-05, 0.0000e+00, ..., 2.0587e-04, + -3.4943e-03, -9.0647e-04], + ..., + [-5.5170e-04, -4.3058e-04, 0.0000e+00, ..., 9.1791e-04, + 7.2050e-04, -8.6641e-04], + [ 5.7411e-04, 3.7611e-05, 0.0000e+00, ..., 1.0414e-03, + 7.8440e-04, 1.0872e-03], + [ 3.0155e-03, 2.4509e-04, 0.0000e+00, ..., 1.5745e-03, + 2.8877e-03, 4.2038e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0231, -0.0160, 0.0117, -0.0200, 0.0285, 0.0169, 0.0011, -0.0110, + -0.0162, -0.0022], device='cuda:0'), grad: tensor([ 0.0007, -0.0137, -0.0039, 0.0028, 0.0008, 0.0015, 0.0008, 0.0012, + 0.0026, 0.0073], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 221.03, cls_loss 0.0180 cls_loss_mapping 0.0395 cls_loss_causal 0.8501 re_mapping 0.0209 re_causal 0.0649 /// teacc 98.70 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0577, 0.0669, 0.0215, ..., -0.0514, -0.0049, -0.0102], + [-0.0034, -0.0130, -0.0160, ..., 0.0141, 0.0008, -0.0150], + [-0.0533, -0.0535, -0.0024, ..., 0.0044, -0.0053, -0.0004], + ..., + [ 0.0260, 0.0253, 0.0014, ..., -0.0131, 0.0297, 0.0265], + [ 0.0282, -0.0117, 0.0071, ..., 0.0082, 0.0138, -0.0054], + [-0.0467, -0.0035, -0.0190, ..., 0.0524, -0.0839, 0.0200]], + device='cuda:0'), grad: tensor([[ 1.3578e-04, -2.2221e-04, 0.0000e+00, ..., 1.4400e-04, + -2.6718e-05, -6.1989e-06], + [ 1.3089e-04, 7.8827e-06, 0.0000e+00, ..., -4.1872e-05, + 9.7871e-05, 4.9978e-05], + [ 5.6863e-05, 5.2184e-05, 0.0000e+00, ..., 2.8104e-05, + 2.0355e-05, 3.9518e-05], + ..., + [-4.1306e-05, -9.7990e-05, 0.0000e+00, ..., 9.0539e-05, + -3.5733e-05, -7.7367e-05], + [ 3.3617e-04, 2.4974e-05, 0.0000e+00, ..., -3.2926e-04, + -2.4259e-04, -1.9014e-04], + [ 1.1688e-04, 1.3721e-04, 0.0000e+00, ..., -3.4571e-05, + 1.1373e-04, -1.3912e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0230, -0.0159, 0.0118, -0.0199, 0.0283, 0.0170, 0.0011, -0.0108, + -0.0164, -0.0023], device='cuda:0'), grad: tensor([ 2.2304e-04, 4.6277e-04, 9.2208e-05, 1.1063e-03, 9.4891e-05, + 1.7796e-03, -3.3169e-03, 2.6196e-05, -7.0858e-04, 2.3973e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 221.56, cls_loss 0.0217 cls_loss_mapping 0.0452 cls_loss_causal 0.8418 re_mapping 0.0202 re_causal 0.0643 /// teacc 98.52 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0585, 0.0674, 0.0215, ..., -0.0518, -0.0051, -0.0104], + [-0.0030, -0.0136, -0.0160, ..., 0.0144, 0.0006, -0.0144], + [-0.0536, -0.0538, -0.0024, ..., 0.0041, -0.0049, -0.0009], + ..., + [ 0.0266, 0.0255, 0.0014, ..., -0.0135, 0.0297, 0.0267], + [ 0.0286, -0.0122, 0.0071, ..., 0.0080, 0.0136, -0.0056], + [-0.0473, -0.0038, -0.0190, ..., 0.0526, -0.0850, 0.0201]], + device='cuda:0'), grad: tensor([[ 6.4194e-05, -4.2009e-04, 0.0000e+00, ..., 1.8048e-04, + -2.3329e-04, -8.5175e-05], + [ 1.7846e-04, 2.4647e-05, 0.0000e+00, ..., 4.6164e-05, + 2.7990e-04, 2.5916e-04], + [ 1.2529e-04, 4.5151e-05, 0.0000e+00, ..., 1.1861e-04, + 2.2173e-04, 2.5463e-04], + ..., + [ 5.2786e-04, -1.8156e-04, 0.0000e+00, ..., 4.7135e-04, + -8.3208e-04, -9.5701e-04], + [ 3.8326e-05, 3.0458e-05, 0.0000e+00, ..., -8.8644e-04, + -2.9635e-04, -1.5230e-03], + [ 7.3004e-04, 1.1194e-04, 0.0000e+00, ..., 6.2084e-04, + 8.0633e-04, 1.1702e-03]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0225, -0.0159, 0.0120, -0.0196, 0.0284, 0.0167, 0.0014, -0.0108, + -0.0162, -0.0026], device='cuda:0'), grad: tensor([ 0.0002, 0.0006, 0.0007, 0.0013, -0.0016, 0.0007, -0.0002, -0.0006, + -0.0049, 0.0038], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 221.92, cls_loss 0.0208 cls_loss_mapping 0.0435 cls_loss_causal 0.8071 re_mapping 0.0190 re_causal 0.0591 /// teacc 98.87 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0590, 0.0680, 0.0215, ..., -0.0517, -0.0048, -0.0105], + [-0.0032, -0.0143, -0.0160, ..., 0.0144, 0.0007, -0.0149], + [-0.0542, -0.0547, -0.0024, ..., 0.0038, -0.0050, -0.0015], + ..., + [ 0.0272, 0.0257, 0.0014, ..., -0.0140, 0.0303, 0.0269], + [ 0.0289, -0.0128, 0.0071, ..., 0.0076, 0.0131, -0.0054], + [-0.0480, -0.0032, -0.0190, ..., 0.0529, -0.0864, 0.0204]], + device='cuda:0'), grad: tensor([[ 5.8979e-05, -4.3273e-04, 0.0000e+00, ..., 6.4909e-05, + -4.8250e-05, -1.5783e-04], + [ 4.9233e-05, 6.8545e-06, 0.0000e+00, ..., 3.1561e-05, + 1.4156e-05, 6.2808e-06], + [ 4.5985e-05, 4.2170e-05, 0.0000e+00, ..., 2.4512e-05, + -8.3208e-05, 1.9938e-05], + ..., + [ 1.1450e-04, 3.7923e-06, 0.0000e+00, ..., 1.7035e-04, + 6.5491e-06, 3.2693e-05], + [-5.0187e-05, 4.0591e-05, 0.0000e+00, ..., 8.8453e-05, + -1.2945e-06, 4.2737e-05], + [ 1.2994e-04, 1.0192e-04, 0.0000e+00, ..., -1.9372e-07, + 3.6240e-05, -7.0930e-05]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0224, -0.0161, 0.0120, -0.0201, 0.0286, 0.0168, 0.0014, -0.0106, + -0.0159, -0.0027], device='cuda:0'), grad: tensor([-3.1757e-04, 9.8228e-05, 8.3447e-07, 1.5223e-04, -7.7710e-06, + 1.2517e-04, -3.6216e-04, 2.3365e-04, -1.1969e-04, 1.9693e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 221.23, cls_loss 0.0164 cls_loss_mapping 0.0368 cls_loss_causal 0.8083 re_mapping 0.0191 re_causal 0.0601 /// teacc 98.76 lr 0.00010000 +Epoch 29, weight, value: tensor([[-5.9608e-02, 6.8900e-02, 2.1474e-02, ..., -5.1820e-02, + -4.8102e-03, -1.0217e-02], + [-3.1428e-03, -1.4941e-02, -1.6035e-02, ..., 1.4709e-02, + 7.7120e-05, -1.5244e-02], + [-5.4266e-02, -5.5666e-02, -2.3694e-03, ..., 3.2383e-03, + -4.2649e-03, -2.1302e-03], + ..., + [ 2.7268e-02, 2.6016e-02, 1.4217e-03, ..., -1.4595e-02, + 3.0371e-02, 2.6945e-02], + [ 2.9114e-02, -1.3270e-02, 7.1494e-03, ..., 7.2195e-03, + 1.2816e-02, -5.6487e-03], + [-4.8255e-02, -3.1679e-03, -1.9007e-02, ..., 5.3250e-02, + -8.7291e-02, 2.0705e-02]], device='cuda:0'), grad: tensor([[ 2.3246e-04, -4.6539e-04, 0.0000e+00, ..., 5.6887e-04, + 4.8041e-05, -4.9353e-05], + [-8.5115e-05, 2.4229e-05, 0.0000e+00, ..., -1.5438e-04, + 2.0519e-05, 2.3112e-05], + [ 1.0687e-04, 1.5545e-04, 0.0000e+00, ..., 2.0909e-04, + 1.1969e-04, 9.0420e-05], + ..., + [-6.4149e-06, 2.8670e-05, 0.0000e+00, ..., 1.0484e-04, + 8.8736e-06, -1.9506e-05], + [-1.6257e-05, 5.1439e-05, 0.0000e+00, ..., 1.4842e-04, + 2.3723e-05, 7.0870e-05], + [ 2.3758e-04, 1.5342e-04, 0.0000e+00, ..., 3.1233e-04, + 4.6802e-04, 2.2733e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0225, -0.0165, 0.0125, -0.0201, 0.0289, 0.0166, 0.0014, -0.0110, + -0.0158, -0.0027], device='cuda:0'), grad: tensor([ 5.9223e-04, -1.9848e-04, 6.3086e-04, -2.8419e-03, 2.4815e-03, + 1.4210e-03, -3.6526e-03, 1.3745e-04, 7.5161e-05, 1.3533e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 221.58, cls_loss 0.0155 cls_loss_mapping 0.0351 cls_loss_causal 0.7662 re_mapping 0.0183 re_causal 0.0564 /// teacc 98.67 lr 0.00010000 +Epoch 30, weight, value: tensor([[-6.0235e-02, 6.9886e-02, 2.1474e-02, ..., -5.1922e-02, + -4.8931e-03, -9.7667e-03], + [-2.9859e-03, -1.5341e-02, -1.6035e-02, ..., 1.5390e-02, + -8.1908e-05, -1.5872e-02], + [-5.4863e-02, -5.6175e-02, -2.3694e-03, ..., 2.9079e-03, + -4.2371e-03, -2.5689e-03], + ..., + [ 2.7616e-02, 2.6457e-02, 1.4217e-03, ..., -1.5213e-02, + 3.0804e-02, 2.7380e-02], + [ 2.9388e-02, -1.3704e-02, 7.1494e-03, ..., 6.9659e-03, + 1.2675e-02, -6.0729e-03], + [-4.8932e-02, -3.5080e-03, -1.9007e-02, ..., 5.3486e-02, + -8.8450e-02, 2.0943e-02]], device='cuda:0'), grad: tensor([[ 3.9887e-04, -7.4148e-04, 0.0000e+00, ..., 2.8044e-05, + -1.1069e-04, -2.0432e-04], + [-8.3017e-04, 4.1157e-05, 0.0000e+00, ..., -7.5936e-05, + -1.3876e-03, 3.6031e-05], + [ 7.8440e-04, 2.6870e-04, 0.0000e+00, ..., 1.0753e-04, + 1.1578e-03, 1.2201e-04], + ..., + [-6.0701e-04, -4.2796e-04, 0.0000e+00, ..., 8.0705e-05, + -1.6212e-04, -3.7599e-04], + [ 1.6749e-04, 1.6379e-04, 0.0000e+00, ..., 3.4750e-05, + 1.3840e-04, 1.2153e-04], + [ 2.9182e-04, 2.7704e-04, 0.0000e+00, ..., -1.2207e-04, + 1.7154e-04, 6.2585e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0225, -0.0163, 0.0123, -0.0204, 0.0291, 0.0167, 0.0014, -0.0107, + -0.0158, -0.0030], device='cuda:0'), grad: tensor([-0.0006, -0.0033, 0.0030, 0.0007, 0.0002, -0.0009, 0.0006, -0.0007, + 0.0004, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 220.98, cls_loss 0.0178 cls_loss_mapping 0.0387 cls_loss_causal 0.7877 re_mapping 0.0180 re_causal 0.0550 /// teacc 98.85 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0610, 0.0706, 0.0215, ..., -0.0522, -0.0045, -0.0094], + [-0.0032, -0.0159, -0.0160, ..., 0.0153, -0.0006, -0.0166], + [-0.0552, -0.0571, -0.0024, ..., 0.0027, -0.0039, -0.0031], + ..., + [ 0.0289, 0.0268, 0.0014, ..., -0.0157, 0.0320, 0.0281], + [ 0.0297, -0.0143, 0.0071, ..., 0.0068, 0.0118, -0.0066], + [-0.0494, -0.0033, -0.0190, ..., 0.0543, -0.0904, 0.0214]], + device='cuda:0'), grad: tensor([[ 1.6963e-04, -1.1539e-04, 0.0000e+00, ..., 2.1601e-04, + -4.8459e-05, -5.9366e-05], + [-8.8394e-05, 2.3633e-05, 0.0000e+00, ..., 2.4572e-05, + 3.9712e-06, 2.9296e-05], + [ 9.7752e-05, 4.3780e-05, 0.0000e+00, ..., 8.6486e-05, + 2.1905e-05, 3.8594e-05], + ..., + [-2.5535e-04, -1.7655e-04, 0.0000e+00, ..., 3.4899e-05, + -1.5092e-04, -2.3866e-04], + [ 2.4509e-04, 1.2469e-04, 0.0000e+00, ..., 2.9945e-04, + 9.7096e-05, 9.0778e-05], + [ 1.7107e-04, 1.0687e-04, 0.0000e+00, ..., 1.2338e-04, + 1.3065e-04, 1.4031e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0224, -0.0165, 0.0127, -0.0206, 0.0286, 0.0168, 0.0008, -0.0100, + -0.0158, -0.0028], device='cuda:0'), grad: tensor([ 2.0432e-04, -6.3002e-05, 1.8549e-04, -2.8682e-04, 2.0275e-03, + 1.8191e-04, -3.1414e-03, -3.5357e-04, 7.3671e-04, 5.0974e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 221.50, cls_loss 0.0147 cls_loss_mapping 0.0345 cls_loss_causal 0.7589 re_mapping 0.0180 re_causal 0.0547 /// teacc 98.85 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0614, 0.0713, 0.0215, ..., -0.0522, -0.0038, -0.0094], + [-0.0033, -0.0169, -0.0160, ..., 0.0154, -0.0009, -0.0170], + [-0.0558, -0.0579, -0.0024, ..., 0.0024, -0.0038, -0.0035], + ..., + [ 0.0291, 0.0269, 0.0014, ..., -0.0163, 0.0320, 0.0278], + [ 0.0299, -0.0147, 0.0071, ..., 0.0064, 0.0113, -0.0073], + [-0.0497, -0.0030, -0.0190, ..., 0.0545, -0.0914, 0.0221]], + device='cuda:0'), grad: tensor([[ 1.0026e-04, -1.8883e-04, 0.0000e+00, ..., 3.3319e-05, + -1.3396e-05, -9.7394e-05], + [ 1.2314e-04, 3.6899e-06, 0.0000e+00, ..., 6.1393e-05, + 6.1333e-05, 5.4926e-05], + [ 4.9591e-05, 2.2665e-05, 0.0000e+00, ..., 1.9148e-05, + -1.0049e-04, 6.5446e-05], + ..., + [ 3.6001e-05, -2.6356e-06, 0.0000e+00, ..., 4.7415e-05, + 1.1814e-04, 1.8287e-04], + [-1.9491e-04, 7.1824e-06, 0.0000e+00, ..., 8.1301e-05, + -3.3259e-05, -3.6210e-05], + [ 1.1700e-04, 2.5004e-05, 0.0000e+00, ..., 1.2779e-04, + 8.8358e-04, 1.5059e-03]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0225, -0.0166, 0.0126, -0.0204, 0.0284, 0.0167, 0.0012, -0.0105, + -0.0160, -0.0022], device='cuda:0'), grad: tensor([ 0.0003, 0.0005, -0.0001, -0.0015, -0.0002, 0.0005, -0.0003, 0.0003, + -0.0013, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 221.61, cls_loss 0.0155 cls_loss_mapping 0.0341 cls_loss_causal 0.7861 re_mapping 0.0176 re_causal 0.0551 /// teacc 98.78 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0621, 0.0718, 0.0215, ..., -0.0525, -0.0040, -0.0093], + [-0.0035, -0.0172, -0.0160, ..., 0.0158, -0.0014, -0.0168], + [-0.0561, -0.0583, -0.0024, ..., 0.0022, -0.0034, -0.0039], + ..., + [ 0.0295, 0.0274, 0.0014, ..., -0.0169, 0.0325, 0.0281], + [ 0.0301, -0.0150, 0.0071, ..., 0.0061, 0.0112, -0.0075], + [-0.0503, -0.0034, -0.0190, ..., 0.0547, -0.0929, 0.0223]], + device='cuda:0'), grad: tensor([[ 3.9250e-05, -1.2481e-04, 0.0000e+00, ..., 6.3181e-05, + 5.8040e-06, 6.1244e-06], + [ 5.3406e-04, 3.0771e-06, 0.0000e+00, ..., 1.0407e-04, + 4.6492e-04, 5.8889e-04], + [ 3.3164e-04, 3.0041e-05, 0.0000e+00, ..., 3.6418e-05, + 1.0204e-03, 9.9468e-04], + ..., + [-9.1791e-04, -6.1803e-06, 0.0000e+00, ..., 4.8846e-05, + -2.8458e-03, -3.1013e-03], + [ 1.0580e-04, 2.1234e-05, 0.0000e+00, ..., 1.7202e-04, + 7.0751e-05, 2.4199e-04], + [ 5.4550e-04, 3.0786e-05, 0.0000e+00, ..., -2.7013e-04, + 2.9659e-04, 2.7990e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0221, -0.0167, 0.0129, -0.0206, 0.0285, 0.0169, 0.0013, -0.0105, + -0.0156, -0.0026], device='cuda:0'), grad: tensor([ 5.6744e-05, 1.6975e-03, 1.7691e-03, 5.1689e-03, 8.7559e-05, + -5.4474e-03, 1.5163e-04, -5.0850e-03, 7.0858e-04, 8.8549e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 221.43, cls_loss 0.0150 cls_loss_mapping 0.0368 cls_loss_causal 0.7783 re_mapping 0.0168 re_causal 0.0546 /// teacc 98.70 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0625, 0.0721, 0.0107, ..., -0.0528, -0.0042, -0.0097], + [-0.0029, -0.0178, -0.0210, ..., 0.0165, -0.0015, -0.0167], + [-0.0565, -0.0588, -0.0007, ..., 0.0020, -0.0034, -0.0047], + ..., + [ 0.0295, 0.0277, -0.0062, ..., -0.0177, 0.0331, 0.0285], + [ 0.0301, -0.0154, -0.0022, ..., 0.0057, 0.0106, -0.0082], + [-0.0507, -0.0030, -0.0300, ..., 0.0553, -0.0937, 0.0227]], + device='cuda:0'), grad: tensor([[ 3.6925e-05, -6.3705e-04, 0.0000e+00, ..., -1.7524e-04, + -3.8028e-04, -7.1287e-04], + [ 2.9278e-03, 5.6699e-06, 0.0000e+00, ..., 2.4452e-03, + 1.4746e-04, 2.8953e-05], + [ 2.3150e-04, 4.2021e-05, 0.0000e+00, ..., 2.4128e-04, + 1.2791e-04, 1.1277e-04], + ..., + [ 3.7289e-04, -1.1005e-05, 0.0000e+00, ..., 3.2687e-04, + 1.6749e-05, -2.6360e-05], + [-1.2360e-03, 9.8869e-06, 0.0000e+00, ..., -9.1553e-04, + -2.4068e-04, -7.1824e-05], + [ 4.8804e-04, 4.9782e-04, 0.0000e+00, ..., 7.6294e-04, + 4.8614e-04, 6.5231e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0216, -0.0166, 0.0129, -0.0203, 0.0285, 0.0170, 0.0010, -0.0102, + -0.0160, -0.0024], device='cuda:0'), grad: tensor([-0.0011, 0.0034, 0.0007, -0.0002, -0.0038, 0.0005, 0.0012, 0.0005, + -0.0034, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 221.26, cls_loss 0.0153 cls_loss_mapping 0.0356 cls_loss_causal 0.7715 re_mapping 0.0164 re_causal 0.0539 /// teacc 98.84 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0628, 0.0729, 0.0107, ..., -0.0531, -0.0039, -0.0094], + [-0.0028, -0.0184, -0.0210, ..., 0.0166, -0.0012, -0.0163], + [-0.0569, -0.0594, -0.0007, ..., 0.0017, -0.0033, -0.0056], + ..., + [ 0.0296, 0.0280, -0.0062, ..., -0.0181, 0.0330, 0.0288], + [ 0.0301, -0.0158, -0.0022, ..., 0.0054, 0.0100, -0.0085], + [-0.0513, -0.0034, -0.0300, ..., 0.0557, -0.0953, 0.0231]], + device='cuda:0'), grad: tensor([[ 2.4468e-05, -5.9336e-05, 0.0000e+00, ..., 1.9073e-05, + -5.8115e-06, -1.3769e-05], + [ 2.2784e-05, 3.0678e-06, 0.0000e+00, ..., 9.1851e-05, + 2.5034e-05, 2.3574e-05], + [ 3.1531e-05, 8.4862e-06, 0.0000e+00, ..., 2.8551e-05, + 1.1750e-05, 2.7061e-05], + ..., + [ 3.9309e-05, -3.0264e-05, 0.0000e+00, ..., 7.3433e-05, + 3.4899e-05, 8.2105e-06], + [ 1.2589e-04, 5.5470e-06, 0.0000e+00, ..., 3.5495e-05, + 7.1764e-05, 3.0413e-05], + [ 2.7847e-04, 2.5854e-05, 0.0000e+00, ..., 5.8603e-04, + 1.1796e-04, 1.0818e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0217, -0.0166, 0.0131, -0.0199, 0.0286, 0.0165, 0.0012, -0.0102, + -0.0160, -0.0028], device='cuda:0'), grad: tensor([ 3.4451e-05, 1.5855e-04, 3.0875e-05, -3.2949e-04, -9.9277e-04, + 5.1975e-05, -5.1689e-04, 1.7905e-04, 3.6645e-04, 1.0195e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 221.34, cls_loss 0.0107 cls_loss_mapping 0.0303 cls_loss_causal 0.7347 re_mapping 0.0162 re_causal 0.0508 /// teacc 98.79 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0632, 0.0736, 0.0107, ..., -0.0532, -0.0034, -0.0092], + [-0.0027, -0.0190, -0.0209, ..., 0.0165, -0.0014, -0.0170], + [-0.0572, -0.0599, -0.0008, ..., 0.0016, -0.0030, -0.0062], + ..., + [ 0.0297, 0.0282, -0.0063, ..., -0.0186, 0.0332, 0.0289], + [ 0.0303, -0.0163, -0.0022, ..., 0.0051, 0.0096, -0.0086], + [-0.0517, -0.0037, -0.0300, ..., 0.0563, -0.0966, 0.0235]], + device='cuda:0'), grad: tensor([[ 2.2471e-04, -8.0466e-06, 0.0000e+00, ..., 3.4952e-04, + -1.4281e-04, -1.4901e-07], + [ 4.0442e-05, 1.7807e-05, 0.0000e+00, ..., 4.7058e-05, + 1.4007e-05, 1.1086e-05], + [ 9.2328e-05, 9.1553e-05, 0.0000e+00, ..., 7.8142e-05, + 4.0054e-05, 5.6684e-05], + ..., + [-4.5568e-05, 4.8243e-06, 0.0000e+00, ..., 4.1157e-05, + -9.6023e-05, -1.5664e-04], + [ 8.5831e-05, 4.3750e-05, 0.0000e+00, ..., 2.7120e-05, + 6.0767e-05, 5.6088e-05], + [ 1.1677e-04, 5.6565e-05, 0.0000e+00, ..., -1.1927e-04, + 5.3734e-05, -1.9419e-04]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0218, -0.0167, 0.0131, -0.0198, 0.0285, 0.0163, 0.0013, -0.0103, + -0.0161, -0.0025], device='cuda:0'), grad: tensor([ 1.2569e-03, 1.5962e-04, 3.6454e-04, 7.3481e-04, 4.9204e-05, + 3.3259e-04, -3.0460e-03, -5.2929e-05, 2.3460e-04, -3.3736e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 221.94, cls_loss 0.0124 cls_loss_mapping 0.0302 cls_loss_causal 0.6983 re_mapping 0.0160 re_causal 0.0488 /// teacc 98.88 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0639, 0.0739, 0.0104, ..., -0.0533, -0.0029, -0.0090], + [-0.0028, -0.0194, -0.0205, ..., 0.0164, -0.0018, -0.0171], + [-0.0579, -0.0605, -0.0010, ..., 0.0009, -0.0029, -0.0071], + ..., + [ 0.0302, 0.0289, -0.0064, ..., -0.0191, 0.0332, 0.0290], + [ 0.0306, -0.0167, -0.0024, ..., 0.0049, 0.0094, -0.0090], + [-0.0523, -0.0036, -0.0302, ..., 0.0564, -0.0978, 0.0237]], + device='cuda:0'), grad: tensor([[ 3.0845e-05, -2.4065e-05, 0.0000e+00, ..., 5.1931e-06, + 5.4501e-06, 1.3851e-05], + [ 9.2089e-05, 3.2559e-06, 0.0000e+00, ..., -1.8433e-05, + 4.9770e-05, 4.8399e-05], + [ 6.2168e-05, 1.2785e-05, 0.0000e+00, ..., 1.8716e-05, + 4.9323e-05, 1.7440e-04], + ..., + [-3.7098e-04, -2.1026e-05, 0.0000e+00, ..., 3.2842e-05, + -3.0398e-05, -2.2614e-04], + [-1.3053e-05, 3.0231e-06, 0.0000e+00, ..., 1.2562e-05, + 2.5600e-05, 3.2723e-05], + [ 1.1754e-04, 8.7321e-06, 0.0000e+00, ..., -8.5473e-05, + 4.4107e-05, -8.3923e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0216, -0.0171, 0.0129, -0.0200, 0.0293, 0.0164, 0.0017, -0.0104, + -0.0160, -0.0027], device='cuda:0'), grad: tensor([ 6.4075e-05, 4.0984e-04, -8.9109e-05, -8.5354e-05, 1.9360e-04, + -3.5167e-05, 1.8644e-04, -7.0620e-04, -4.7863e-05, 1.0961e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 221.25, cls_loss 0.0119 cls_loss_mapping 0.0281 cls_loss_causal 0.7292 re_mapping 0.0154 re_causal 0.0481 /// teacc 98.67 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0644, 0.0748, 0.0102, ..., -0.0533, -0.0026, -0.0089], + [-0.0027, -0.0198, -0.0205, ..., 0.0165, -0.0019, -0.0176], + [-0.0586, -0.0613, -0.0010, ..., 0.0004, -0.0032, -0.0077], + ..., + [ 0.0308, 0.0293, -0.0064, ..., -0.0196, 0.0338, 0.0295], + [ 0.0306, -0.0168, -0.0024, ..., 0.0049, 0.0090, -0.0092], + [-0.0527, -0.0040, -0.0303, ..., 0.0566, -0.0989, 0.0244]], + device='cuda:0'), grad: tensor([[ 1.3614e-04, -3.4404e-04, 0.0000e+00, ..., 1.2130e-04, + 5.5075e-05, 2.6017e-05], + [ 8.2397e-03, 6.2324e-06, 0.0000e+00, ..., 7.2556e-03, + 8.7917e-05, 2.1565e-04], + [ 7.4863e-05, 1.0657e-04, 0.0000e+00, ..., 6.5207e-05, + -3.7098e-04, 9.3699e-05], + ..., + [ 3.0403e-03, -7.8529e-06, 0.0000e+00, ..., 3.3150e-03, + 6.8235e-04, 2.2640e-03], + [-6.4201e-03, 2.3812e-05, 0.0000e+00, ..., -6.6185e-03, + 3.0565e-04, -4.0436e-03], + [ 2.7771e-03, 1.5128e-04, 0.0000e+00, ..., 2.7370e-03, + 4.5300e-04, 2.1076e-03]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0217, -0.0171, 0.0126, -0.0201, 0.0296, 0.0168, 0.0011, -0.0101, + -0.0161, -0.0028], device='cuda:0'), grad: tensor([ 8.1718e-05, 1.1063e-02, -4.7278e-04, -5.7745e-04, -1.0658e-02, + -6.1464e-04, 5.2595e-04, 7.0610e-03, -1.2627e-02, 6.2256e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 221.99, cls_loss 0.0090 cls_loss_mapping 0.0248 cls_loss_causal 0.7195 re_mapping 0.0150 re_causal 0.0487 /// teacc 98.93 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0648, 0.0753, 0.0080, ..., -0.0536, -0.0024, -0.0090], + [-0.0026, -0.0205, -0.0196, ..., 0.0168, -0.0020, -0.0179], + [-0.0591, -0.0619, -0.0030, ..., 0.0001, -0.0031, -0.0082], + ..., + [ 0.0313, 0.0300, -0.0069, ..., -0.0205, 0.0345, 0.0298], + [ 0.0308, -0.0171, -0.0036, ..., 0.0048, 0.0087, -0.0096], + [-0.0533, -0.0040, -0.0312, ..., 0.0571, -0.1002, 0.0250]], + device='cuda:0'), grad: tensor([[ 9.9421e-05, -1.6522e-04, 0.0000e+00, ..., 6.0529e-05, + -4.6372e-05, 3.4273e-05], + [-2.7210e-05, 2.2009e-05, 0.0000e+00, ..., -4.8369e-05, + 2.2411e-05, 2.1577e-05], + [ 1.2159e-04, 1.2374e-04, 0.0000e+00, ..., 8.6427e-05, + 7.4863e-05, 4.8995e-05], + ..., + [-1.8954e-04, -2.0850e-04, 0.0000e+00, ..., 3.9637e-05, + -5.0694e-05, -1.6320e-04], + [-3.6526e-04, 8.8513e-05, 0.0000e+00, ..., -1.5044e-04, + -4.5091e-05, 8.9049e-05], + [ 5.6362e-04, 8.9049e-05, 0.0000e+00, ..., 3.1638e-04, + 3.7026e-04, 7.2765e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0214, -0.0168, 0.0125, -0.0206, 0.0297, 0.0171, 0.0010, -0.0101, + -0.0162, -0.0026], device='cuda:0'), grad: tensor([ 5.5999e-05, -2.5570e-05, 4.4775e-04, -3.3259e-04, 3.3021e-04, + -1.0128e-03, -8.6427e-05, -2.8348e-04, -9.1600e-04, 1.8225e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 221.10, cls_loss 0.0082 cls_loss_mapping 0.0224 cls_loss_causal 0.7292 re_mapping 0.0149 re_causal 0.0487 /// teacc 98.91 lr 0.00010000 +Epoch 40, weight, value: tensor([[-6.5262e-02, 7.6111e-02, 7.1190e-03, ..., -5.3943e-02, + -2.1221e-03, -8.7471e-03], + [-2.4465e-03, -2.1303e-02, -1.9655e-02, ..., 1.7238e-02, + -2.1946e-03, -1.7789e-02], + [-5.9595e-02, -6.2375e-02, -2.9460e-03, ..., -1.2674e-05, + -3.2499e-03, -8.6540e-03], + ..., + [ 3.1424e-02, 3.0306e-02, -6.9804e-03, ..., -2.1184e-02, + 3.4483e-02, 2.9966e-02], + [ 3.0894e-02, -1.7569e-02, -3.7024e-03, ..., 4.3204e-03, + 8.3091e-03, -1.0087e-02], + [-5.3773e-02, -4.3447e-03, -3.1355e-02, ..., 5.7592e-02, + -1.0164e-01, 2.5250e-02]], device='cuda:0'), grad: tensor([[ 7.1824e-05, -5.3495e-05, 0.0000e+00, ..., 5.9932e-05, + 2.8908e-05, -7.7188e-06], + [ 1.7548e-04, 2.4606e-06, 0.0000e+00, ..., 7.3612e-05, + 1.2922e-04, 5.6982e-05], + [ 5.5122e-04, 4.3884e-06, 0.0000e+00, ..., -2.4319e-04, + -1.7810e-04, 4.2248e-04], + ..., + [-1.4153e-03, 7.4878e-06, 0.0000e+00, ..., -2.5868e-04, + -1.0548e-03, -7.8535e-04], + [ 1.3840e-04, 1.0513e-05, 0.0000e+00, ..., 7.9036e-05, + 1.3161e-04, 7.1049e-05], + [ 1.2255e-04, 1.0066e-05, 0.0000e+00, ..., -6.8009e-05, + 1.3304e-04, -3.9786e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0215, -0.0166, 0.0126, -0.0200, 0.0293, 0.0169, 0.0014, -0.0105, + -0.0164, -0.0027], device='cuda:0'), grad: tensor([ 0.0003, 0.0007, -0.0005, 0.0006, 0.0014, 0.0004, -0.0014, -0.0021, + 0.0004, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 221.36, cls_loss 0.0079 cls_loss_mapping 0.0229 cls_loss_causal 0.7067 re_mapping 0.0142 re_causal 0.0460 /// teacc 98.82 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0658, 0.0769, 0.0050, ..., -0.0541, -0.0018, -0.0085], + [-0.0020, -0.0221, -0.0197, ..., 0.0176, -0.0023, -0.0182], + [-0.0601, -0.0632, -0.0033, ..., -0.0003, -0.0033, -0.0094], + ..., + [ 0.0317, 0.0309, -0.0072, ..., -0.0217, 0.0347, 0.0304], + [ 0.0313, -0.0180, -0.0038, ..., 0.0040, 0.0081, -0.0104], + [-0.0546, -0.0046, -0.0325, ..., 0.0576, -0.1027, 0.0255]], + device='cuda:0'), grad: tensor([[ 2.3633e-05, -1.3538e-05, 2.3283e-10, ..., 2.6494e-05, + 2.4229e-05, 4.5240e-05], + [ 4.1783e-05, 3.9637e-06, 2.3283e-10, ..., 3.2783e-05, + 1.2219e-04, 1.2410e-04], + [ 3.1620e-05, 4.2021e-06, -2.0955e-09, ..., 9.4771e-06, + 8.1539e-05, 7.7844e-05], + ..., + [ 6.6817e-05, -8.7097e-06, 6.9849e-10, ..., 2.1875e-04, + 1.7035e-04, 4.3774e-04], + [ 1.8167e-04, 4.6827e-06, 4.6566e-10, ..., 2.4527e-05, + 2.3866e-04, 1.9300e-04], + [-8.6546e-05, 7.4133e-06, 0.0000e+00, ..., -1.6727e-03, + 5.7757e-05, -9.3842e-04]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0214, -0.0163, 0.0124, -0.0200, 0.0299, 0.0169, 0.0009, -0.0104, + -0.0161, -0.0031], device='cuda:0'), grad: tensor([ 1.0985e-04, 2.8968e-04, 1.2481e-04, -1.2474e-03, 1.9913e-03, + 1.4174e-04, 2.9337e-06, 7.9155e-04, 4.9496e-04, -2.6989e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 221.43, cls_loss 0.0120 cls_loss_mapping 0.0299 cls_loss_causal 0.7175 re_mapping 0.0140 re_causal 0.0453 /// teacc 98.59 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0665, 0.0773, 0.0044, ..., -0.0553, -0.0014, -0.0090], + [-0.0022, -0.0230, -0.0232, ..., 0.0176, -0.0031, -0.0180], + [-0.0600, -0.0639, 0.0005, ..., -0.0006, -0.0027, -0.0099], + ..., + [ 0.0318, 0.0311, -0.0131, ..., -0.0223, 0.0351, 0.0305], + [ 0.0317, -0.0183, -0.0067, ..., 0.0039, 0.0071, -0.0103], + [-0.0553, -0.0043, -0.0373, ..., 0.0583, -0.1035, 0.0261]], + device='cuda:0'), grad: tensor([[ 5.8144e-05, -2.2137e-04, 5.0291e-08, ..., -8.5905e-06, + -1.8686e-05, -4.2468e-05], + [-1.3037e-03, 3.3647e-05, 3.0734e-08, ..., -7.8297e-04, + 3.3438e-05, 5.7429e-05], + [ 5.6744e-05, 1.8373e-05, -4.2911e-07, ..., 2.1189e-05, + -2.7686e-05, 1.5944e-05], + ..., + [-1.2919e-05, -1.9717e-04, 7.6136e-08, ..., 1.2410e-04, + -1.2362e-04, -2.4581e-04], + [-4.7898e-04, 9.1493e-05, 1.0547e-07, ..., -3.5286e-04, + 2.8268e-05, 1.6183e-05], + [ 6.8092e-04, 1.2875e-04, 4.8894e-09, ..., 9.6798e-04, + 5.1916e-05, 1.7762e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0206, -0.0171, 0.0132, -0.0200, 0.0300, 0.0165, 0.0010, -0.0103, + -0.0157, -0.0029], device='cuda:0'), grad: tensor([-7.7844e-05, -1.4973e-03, -6.0111e-05, 2.5964e-04, 6.8045e-04, + -1.9989e-03, 2.3613e-03, -1.8346e-04, -9.9087e-04, 1.5030e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 221.15, cls_loss 0.0095 cls_loss_mapping 0.0236 cls_loss_causal 0.6774 re_mapping 0.0151 re_causal 0.0451 /// teacc 98.82 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0669, 0.0780, 0.0011, ..., -0.0554, -0.0012, -0.0086], + [-0.0022, -0.0238, -0.0233, ..., 0.0180, -0.0035, -0.0191], + [-0.0606, -0.0645, 0.0013, ..., -0.0011, -0.0031, -0.0111], + ..., + [ 0.0325, 0.0313, -0.0159, ..., -0.0228, 0.0357, 0.0312], + [ 0.0316, -0.0186, -0.0064, ..., 0.0038, 0.0066, -0.0106], + [-0.0556, -0.0047, -0.0400, ..., 0.0589, -0.1046, 0.0265]], + device='cuda:0'), grad: tensor([[ 1.5295e-04, -2.9624e-05, 9.7811e-05, ..., 9.3758e-05, + -5.1856e-05, 1.7047e-05], + [ 5.4896e-05, 2.1100e-05, 5.9679e-06, ..., 6.5528e-06, + 4.0114e-05, 1.3277e-05], + [-1.2341e-03, 8.4281e-05, -2.5302e-05, ..., 2.2799e-06, + -1.0586e-03, 2.9534e-05], + ..., + [ 1.0786e-03, -1.7393e-04, 8.5160e-06, ..., 1.4082e-05, + 8.3542e-04, -1.0645e-04], + [ 2.2650e-05, 2.1547e-05, 5.9679e-06, ..., 4.8317e-06, + 7.1943e-05, 1.2137e-05], + [ 4.6968e-05, 6.0737e-05, 2.4997e-06, ..., 6.0908e-06, + 8.8394e-05, 2.9817e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0206, -0.0170, 0.0127, -0.0200, 0.0294, 0.0166, 0.0008, -0.0098, + -0.0156, -0.0028], device='cuda:0'), grad: tensor([ 3.6597e-04, 1.8382e-04, -4.3907e-03, 1.4746e-04, 1.5885e-05, + 7.9453e-05, -2.9588e-04, 3.1757e-03, 2.9588e-04, 4.2176e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 221.06, cls_loss 0.0073 cls_loss_mapping 0.0221 cls_loss_causal 0.6960 re_mapping 0.0141 re_causal 0.0444 /// teacc 98.83 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0672, 0.0787, 0.0002, ..., -0.0554, -0.0010, -0.0083], + [-0.0024, -0.0252, -0.0231, ..., 0.0179, -0.0038, -0.0194], + [-0.0607, -0.0650, 0.0015, ..., -0.0013, -0.0027, -0.0115], + ..., + [ 0.0329, 0.0316, -0.0174, ..., -0.0231, 0.0361, 0.0313], + [ 0.0318, -0.0191, -0.0067, ..., 0.0035, 0.0060, -0.0109], + [-0.0561, -0.0050, -0.0408, ..., 0.0593, -0.1056, 0.0268]], + device='cuda:0'), grad: tensor([[ 3.3289e-05, -4.0293e-05, 0.0000e+00, ..., 5.3525e-05, + 2.8491e-05, 6.9290e-06], + [ 1.9133e-05, 7.6368e-06, 0.0000e+00, ..., -1.0870e-05, + 5.3763e-05, 4.3422e-05], + [ 1.0049e-04, 1.4484e-05, 0.0000e+00, ..., 4.3362e-05, + -1.1522e-04, 4.5061e-05], + ..., + [-9.6798e-05, -7.8857e-05, 0.0000e+00, ..., 2.3827e-05, + -1.4651e-04, -2.2733e-04], + [-3.7718e-04, 3.5316e-05, 0.0000e+00, ..., -1.4156e-05, + -7.0214e-05, 1.8299e-05], + [ 1.3053e-04, 5.0426e-05, 0.0000e+00, ..., 3.3289e-05, + 7.6115e-05, 7.1228e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0207, -0.0172, 0.0131, -0.0203, 0.0292, 0.0169, 0.0008, -0.0100, + -0.0154, -0.0027], device='cuda:0'), grad: tensor([ 0.0001, 0.0001, 0.0001, 0.0003, 0.0002, 0.0002, -0.0003, -0.0004, + -0.0008, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 221.18, cls_loss 0.0081 cls_loss_mapping 0.0217 cls_loss_causal 0.6877 re_mapping 0.0135 re_causal 0.0420 /// teacc 98.84 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0678, 0.0793, -0.0035, ..., -0.0559, -0.0005, -0.0083], + [-0.0024, -0.0263, -0.0231, ..., 0.0180, -0.0040, -0.0196], + [-0.0611, -0.0660, 0.0011, ..., -0.0017, -0.0025, -0.0122], + ..., + [ 0.0333, 0.0319, -0.0176, ..., -0.0235, 0.0364, 0.0317], + [ 0.0320, -0.0191, -0.0071, ..., 0.0033, 0.0055, -0.0113], + [-0.0570, -0.0048, -0.0421, ..., 0.0593, -0.1066, 0.0270]], + device='cuda:0'), grad: tensor([[ 4.8459e-05, -1.3679e-05, 0.0000e+00, ..., 1.3314e-05, + 3.4541e-05, 2.1592e-05], + [ 1.2589e-04, 1.4208e-05, 0.0000e+00, ..., 2.8983e-05, + 1.4138e-04, 1.6212e-04], + [ 1.2064e-04, 2.8104e-05, 0.0000e+00, ..., 1.4096e-05, + 4.1127e-04, 2.1243e-04], + ..., + [-2.4056e-04, -3.2157e-05, 0.0000e+00, ..., 1.5900e-05, + -5.2261e-04, -5.9032e-04], + [-3.2353e-04, -4.0114e-05, 0.0000e+00, ..., 6.4135e-05, + -8.5458e-06, 7.6652e-05], + [ 6.7428e-06, 2.3525e-06, 0.0000e+00, ..., -5.1975e-04, + 1.0383e-04, -2.3365e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0204, -0.0171, 0.0130, -0.0206, 0.0300, 0.0170, 0.0008, -0.0098, + -0.0152, -0.0033], device='cuda:0'), grad: tensor([ 1.5557e-04, 3.2496e-04, 9.1457e-04, -8.7559e-05, 3.8362e-04, + 5.9456e-05, 3.0541e-04, -9.6798e-04, -7.1430e-04, -3.7408e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 220.93, cls_loss 0.0059 cls_loss_mapping 0.0184 cls_loss_causal 0.6791 re_mapping 0.0135 re_causal 0.0429 /// teacc 98.82 lr 0.00010000 +Epoch 46, weight, value: tensor([[-6.8233e-02, 8.0117e-02, -4.3298e-03, ..., -5.6040e-02, + 7.6915e-05, -8.0567e-03], + [-2.3160e-03, -2.7044e-02, -2.3154e-02, ..., 1.8236e-02, + -4.3729e-03, -2.0043e-02], + [-6.1246e-02, -6.6555e-02, 1.0193e-03, ..., -2.1805e-03, + -2.4769e-03, -1.2742e-02], + ..., + [ 3.3516e-02, 3.1953e-02, -1.7660e-02, ..., -2.3926e-02, + 3.7032e-02, 3.2286e-02], + [ 3.2228e-02, -1.9606e-02, -7.1846e-03, ..., 2.9861e-03, + 5.0148e-03, -1.1506e-02], + [-5.7243e-02, -5.0434e-03, -4.2215e-02, ..., 5.9838e-02, + -1.0774e-01, 2.7391e-02]], device='cuda:0'), grad: tensor([[ 1.6958e-05, -4.3726e-04, 0.0000e+00, ..., -1.3626e-04, + -7.5638e-05, -2.2799e-05], + [ 6.5422e-04, 6.0685e-06, 0.0000e+00, ..., 2.0042e-06, + 8.8453e-05, 1.3375e-04], + [ 3.9250e-05, 3.7849e-05, 0.0000e+00, ..., 2.8893e-05, + -1.8537e-05, 2.1026e-05], + ..., + [-2.5153e-04, 1.0766e-05, 0.0000e+00, ..., -7.7188e-06, + -2.0730e-04, -2.9063e-04], + [-6.8140e-04, 3.4124e-05, 0.0000e+00, ..., 3.9965e-05, + 3.7998e-05, 4.3660e-05], + [ 9.1910e-05, 2.2888e-05, 0.0000e+00, ..., -2.1267e-04, + 5.3406e-05, -1.6272e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0205, -0.0173, 0.0128, -0.0209, 0.0296, 0.0173, 0.0008, -0.0096, + -0.0149, -0.0033], device='cuda:0'), grad: tensor([-3.9816e-04, 1.0509e-03, 2.8998e-05, 3.3116e-04, 4.6223e-05, + 1.8454e-04, 3.6597e-04, -3.8981e-04, -9.0742e-04, -3.1257e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 221.26, cls_loss 0.0097 cls_loss_mapping 0.0268 cls_loss_causal 0.7394 re_mapping 0.0129 re_causal 0.0422 /// teacc 98.77 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0687, 0.0808, -0.0044, ..., -0.0551, 0.0008, -0.0081], + [-0.0020, -0.0300, -0.0232, ..., 0.0182, -0.0048, -0.0206], + [-0.0617, -0.0679, 0.0010, ..., -0.0025, -0.0021, -0.0135], + ..., + [ 0.0338, 0.0320, -0.0177, ..., -0.0246, 0.0374, 0.0327], + [ 0.0324, -0.0204, -0.0072, ..., 0.0024, 0.0046, -0.0120], + [-0.0580, -0.0048, -0.0424, ..., 0.0601, -0.1083, 0.0282]], + device='cuda:0'), grad: tensor([[ 1.6958e-05, -2.4825e-05, 0.0000e+00, ..., 2.4453e-05, + 9.4771e-06, 6.7651e-06], + [-1.6287e-05, 2.2314e-06, 0.0000e+00, ..., 4.4666e-06, + -7.1712e-06, 2.6569e-05], + [ 3.2723e-05, 7.9274e-06, 0.0000e+00, ..., 1.1005e-05, + 1.4186e-04, 4.4674e-05], + ..., + [ 1.4633e-05, -3.0026e-06, 0.0000e+00, ..., 7.2300e-05, + 7.6115e-05, 7.8380e-05], + [ 5.5969e-05, 1.3247e-05, 0.0000e+00, ..., 8.3208e-05, + 1.1128e-04, 6.1154e-05], + [ 1.5587e-05, 7.6070e-06, 0.0000e+00, ..., -2.2900e-04, + 4.1664e-05, -1.3959e-04]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0210, -0.0177, 0.0131, -0.0208, 0.0299, 0.0164, 0.0012, -0.0095, + -0.0152, -0.0031], device='cuda:0'), grad: tensor([ 5.0694e-05, -7.2956e-05, 2.4247e-04, -5.1165e-04, 2.3246e-04, + -3.1257e-04, 8.4877e-05, 2.3329e-04, 3.0780e-04, -2.5344e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 46---------------------------------------------------- +epoch 46, time 221.37, cls_loss 0.0047 cls_loss_mapping 0.0153 cls_loss_causal 0.6930 re_mapping 0.0133 re_causal 0.0425 /// teacc 98.95 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0690, 0.0815, -0.0046, ..., -0.0552, 0.0009, -0.0080], + [-0.0019, -0.0305, -0.0232, ..., 0.0187, -0.0049, -0.0208], + [-0.0621, -0.0685, 0.0010, ..., -0.0027, -0.0018, -0.0141], + ..., + [ 0.0342, 0.0325, -0.0178, ..., -0.0252, 0.0375, 0.0329], + [ 0.0325, -0.0209, -0.0073, ..., 0.0022, 0.0041, -0.0123], + [-0.0584, -0.0051, -0.0424, ..., 0.0604, -0.1096, 0.0286]], + device='cuda:0'), grad: tensor([[ 3.2842e-05, -2.7016e-05, 3.0710e-07, ..., 2.6435e-05, + -8.2701e-06, 2.3860e-06], + [ 6.3002e-05, 1.2498e-06, 4.6683e-08, ..., 7.1116e-06, + 6.1870e-05, 4.0978e-05], + [ 1.2565e-04, 6.4671e-06, 2.1537e-08, ..., 7.3127e-06, + 1.0663e-04, 4.3839e-05], + ..., + [-2.8417e-05, 3.4692e-07, 3.7253e-09, ..., 3.5882e-05, + 7.3463e-06, 3.1114e-05], + [-6.3992e-04, 9.8124e-06, 1.0198e-07, ..., -5.3227e-05, + -2.8396e-04, -3.7360e-04], + [ 2.4235e-04, 8.9929e-06, 1.0012e-08, ..., -5.2166e-04, + 9.4712e-05, -9.3162e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0210, -0.0177, 0.0132, -0.0207, 0.0296, 0.0166, 0.0012, -0.0095, + -0.0153, -0.0033], device='cuda:0'), grad: tensor([ 0.0002, 0.0002, 0.0003, 0.0002, 0.0008, 0.0003, -0.0003, 0.0001, + -0.0020, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 221.09, cls_loss 0.0086 cls_loss_mapping 0.0256 cls_loss_causal 0.6915 re_mapping 0.0135 re_causal 0.0431 /// teacc 98.74 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0695, 0.0822, -0.0046, ..., -0.0555, 0.0011, -0.0080], + [-0.0012, -0.0310, -0.0232, ..., 0.0185, -0.0043, -0.0203], + [-0.0622, -0.0691, 0.0010, ..., -0.0029, -0.0023, -0.0149], + ..., + [ 0.0338, 0.0333, -0.0178, ..., -0.0262, 0.0376, 0.0323], + [ 0.0325, -0.0214, -0.0073, ..., 0.0019, 0.0038, -0.0126], + [-0.0584, -0.0055, -0.0424, ..., 0.0612, -0.1107, 0.0297]], + device='cuda:0'), grad: tensor([[ 6.7770e-05, -8.2329e-06, 0.0000e+00, ..., 2.9340e-05, + 7.7963e-05, 5.8383e-05], + [-2.3413e-04, 1.1884e-05, 0.0000e+00, ..., -1.1128e-04, + -3.4541e-05, -4.0859e-05], + [ 1.0854e-04, 8.7395e-06, 0.0000e+00, ..., 1.8135e-05, + 1.7852e-05, 1.0215e-05], + ..., + [ 1.6844e-04, -4.0457e-06, 0.0000e+00, ..., 1.0455e-04, + 2.3514e-05, 3.1352e-05], + [-1.4341e-04, 1.3649e-05, 0.0000e+00, ..., 1.7360e-05, + 1.1707e-06, 1.5602e-05], + [ 1.8090e-05, -1.3992e-05, 0.0000e+00, ..., -5.1260e-04, + 5.6893e-05, -1.1998e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0210, -0.0170, 0.0129, -0.0208, 0.0298, 0.0169, 0.0009, -0.0103, + -0.0155, -0.0027], device='cuda:0'), grad: tensor([ 2.2578e-04, -6.2704e-04, 2.5249e-04, 2.1458e-04, 5.4646e-04, + -4.9019e-04, 2.8927e-06, 4.9591e-04, -2.7919e-04, -3.4332e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 221.13, cls_loss 0.0077 cls_loss_mapping 0.0199 cls_loss_causal 0.6518 re_mapping 0.0132 re_causal 0.0399 /// teacc 98.91 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0698, 0.0832, -0.0045, ..., -0.0552, 0.0014, -0.0077], + [-0.0021, -0.0315, -0.0232, ..., 0.0179, -0.0052, -0.0212], + [-0.0630, -0.0697, 0.0010, ..., -0.0032, -0.0025, -0.0155], + ..., + [ 0.0343, 0.0343, -0.0179, ..., -0.0265, 0.0384, 0.0329], + [ 0.0327, -0.0218, -0.0073, ..., 0.0017, 0.0032, -0.0126], + [-0.0583, -0.0060, -0.0425, ..., 0.0614, -0.1114, 0.0297]], + device='cuda:0'), grad: tensor([[-5.8532e-05, -3.0708e-04, 0.0000e+00, ..., 6.2995e-06, + -4.1664e-05, -4.0978e-05], + [-1.3471e-05, 9.3086e-07, 0.0000e+00, ..., -2.4319e-05, + 8.2031e-06, 1.3702e-05], + [ 1.0401e-05, 6.8396e-06, 0.0000e+00, ..., 3.2820e-06, + 6.9737e-06, 1.1593e-05], + ..., + [-5.7369e-05, 2.4717e-06, 0.0000e+00, ..., 2.3067e-05, + -5.9456e-05, 6.3360e-05], + [-6.5845e-07, 6.7428e-06, 0.0000e+00, ..., 8.8140e-06, + 5.9754e-06, 9.6709e-06], + [ 3.2812e-05, 1.4916e-05, 0.0000e+00, ..., -3.3587e-05, + 3.4183e-05, -1.5759e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0215, -0.0182, 0.0126, -0.0206, 0.0295, 0.0169, 0.0015, -0.0097, + -0.0153, -0.0029], device='cuda:0'), grad: tensor([-3.3402e-04, -1.2130e-05, 1.1683e-05, 5.1111e-05, 7.8559e-05, + 9.4026e-06, 2.8801e-04, 1.0437e-04, 3.7607e-06, -2.0182e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 220.58, cls_loss 0.0056 cls_loss_mapping 0.0188 cls_loss_causal 0.6506 re_mapping 0.0124 re_causal 0.0404 /// teacc 98.87 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0702, 0.0840, -0.0052, ..., -0.0551, 0.0018, -0.0076], + [-0.0012, -0.0333, -0.0237, ..., 0.0184, -0.0054, -0.0213], + [-0.0631, -0.0703, 0.0026, ..., -0.0033, -0.0020, -0.0161], + ..., + [ 0.0343, 0.0346, -0.0183, ..., -0.0271, 0.0383, 0.0331], + [ 0.0329, -0.0222, -0.0073, ..., 0.0014, 0.0028, -0.0129], + [-0.0589, -0.0062, -0.0432, ..., 0.0616, -0.1124, 0.0299]], + device='cuda:0'), grad: tensor([[ 1.2241e-05, -2.0817e-05, 0.0000e+00, ..., 1.6049e-05, + 2.3358e-06, -2.1011e-06], + [-4.7423e-06, 2.8443e-06, 0.0000e+00, ..., 1.7658e-05, + 7.7933e-06, 7.5512e-06], + [-1.6568e-06, 5.7705e-06, 0.0000e+00, ..., 1.9684e-05, + -6.3002e-05, 8.4713e-06], + ..., + [ 8.2701e-06, -1.2808e-05, 0.0000e+00, ..., 4.2886e-05, + 1.2152e-05, -3.6597e-05], + [ 9.8348e-05, 2.6766e-06, 0.0000e+00, ..., 1.8740e-04, + 4.2200e-05, 6.6042e-05], + [-2.3060e-06, 8.3447e-06, 0.0000e+00, ..., -1.9765e-04, + 2.2113e-05, -9.1612e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0218, -0.0181, 0.0133, -0.0202, 0.0293, 0.0164, 0.0015, -0.0099, + -0.0153, -0.0032], device='cuda:0'), grad: tensor([ 4.6819e-05, 5.2214e-05, -1.6236e-04, 3.4642e-04, 7.4387e-05, + -2.0576e-04, -4.8065e-04, 9.2268e-05, 5.3883e-04, -3.0208e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 221.17, cls_loss 0.0074 cls_loss_mapping 0.0228 cls_loss_causal 0.6920 re_mapping 0.0124 re_causal 0.0391 /// teacc 98.74 lr 0.00010000 +Epoch 52, weight, value: tensor([[-7.0674e-02, 8.5008e-02, -5.8993e-03, ..., -5.4769e-02, + 2.0804e-03, -7.5159e-03], + [-3.3988e-06, -3.4676e-02, -2.3881e-02, ..., 1.9237e-02, + -5.1566e-03, -2.1151e-02], + [-6.3981e-02, -7.1344e-02, 2.9943e-03, ..., -3.8594e-03, + -2.9203e-03, -1.6955e-02], + ..., + [ 3.3825e-02, 3.4966e-02, -1.8558e-02, ..., -2.8273e-02, + 3.8029e-02, 3.2884e-02], + [ 3.3201e-02, -2.2031e-02, -7.4828e-03, ..., 9.6565e-04, + 2.4571e-03, -1.2991e-02], + [-5.9303e-02, -6.4522e-03, -4.3781e-02, ..., 6.2167e-02, + -1.1311e-01, 3.0492e-02]], device='cuda:0'), grad: tensor([[ 5.6624e-06, -2.1343e-03, 0.0000e+00, ..., -7.8869e-04, + 9.6634e-06, 9.0003e-06], + [ 3.4750e-05, 5.5456e-04, 0.0000e+00, ..., 2.3067e-04, + 1.6421e-05, 1.3486e-05], + [ 1.4924e-05, 6.5446e-05, 0.0000e+00, ..., 2.6152e-05, + 2.2098e-05, 1.8716e-05], + ..., + [ 7.3435e-07, -9.4473e-06, 0.0000e+00, ..., 2.3663e-05, + -3.9525e-06, -1.0379e-05], + [-4.1515e-05, 4.7922e-05, 0.0000e+00, ..., 2.6748e-05, + 4.1723e-05, 3.4958e-05], + [ 2.3127e-05, 2.0340e-05, 0.0000e+00, ..., -4.5374e-06, + 6.3419e-05, 4.2975e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0221, -0.0170, 0.0123, -0.0197, 0.0295, 0.0163, 0.0011, -0.0110, + -0.0150, -0.0029], device='cuda:0'), grad: tensor([-3.7441e-03, 1.0796e-03, 1.7357e-04, -1.1711e-03, 2.1830e-05, + 1.0281e-03, 2.3479e-03, 2.2814e-05, 6.7174e-05, 1.7095e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 221.29, cls_loss 0.0056 cls_loss_mapping 0.0191 cls_loss_causal 0.6691 re_mapping 0.0123 re_causal 0.0380 /// teacc 98.94 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0711, 0.0860, -0.0053, ..., -0.0553, 0.0023, -0.0075], + [-0.0002, -0.0364, -0.0239, ..., 0.0195, -0.0059, -0.0215], + [-0.0646, -0.0721, 0.0030, ..., -0.0041, -0.0031, -0.0176], + ..., + [ 0.0345, 0.0359, -0.0187, ..., -0.0287, 0.0389, 0.0333], + [ 0.0336, -0.0233, -0.0075, ..., 0.0006, 0.0018, -0.0134], + [-0.0603, -0.0067, -0.0440, ..., 0.0624, -0.1140, 0.0308]], + device='cuda:0'), grad: tensor([[ 2.0355e-05, -2.8431e-05, 0.0000e+00, ..., 2.3112e-05, + -4.2543e-06, 1.9267e-05], + [ 3.3341e-06, 1.0565e-05, 0.0000e+00, ..., 6.4746e-06, + 5.2117e-06, 1.3337e-05], + [ 1.6749e-05, 1.2681e-05, 0.0000e+00, ..., 7.5512e-06, + -1.1420e-04, 1.4141e-05], + ..., + [-5.0098e-05, -9.9957e-05, 0.0000e+00, ..., 1.7142e-04, + 1.1086e-04, 9.0659e-05], + [ 1.3597e-06, 1.3635e-05, 0.0000e+00, ..., 8.4996e-05, + 1.0788e-05, 5.2184e-05], + [ 3.3796e-05, 5.3376e-05, 0.0000e+00, ..., -5.1451e-04, + 5.6550e-06, -3.9101e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0221, -0.0174, 0.0122, -0.0201, 0.0298, 0.0164, 0.0012, -0.0105, + -0.0148, -0.0031], device='cuda:0'), grad: tensor([ 4.4376e-05, 3.2902e-05, -3.5357e-04, 8.3029e-05, 2.7680e-04, + 1.0800e-04, -8.1956e-05, 5.4598e-04, 2.7791e-05, -6.8331e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 220.99, cls_loss 0.0055 cls_loss_mapping 0.0188 cls_loss_causal 0.6613 re_mapping 0.0128 re_causal 0.0396 /// teacc 98.91 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0716, 0.0861, -0.0052, ..., -0.0543, 0.0026, -0.0078], + [-0.0002, -0.0365, -0.0239, ..., 0.0194, -0.0060, -0.0219], + [-0.0648, -0.0730, 0.0030, ..., -0.0043, -0.0029, -0.0180], + ..., + [ 0.0350, 0.0362, -0.0188, ..., -0.0288, 0.0388, 0.0336], + [ 0.0340, -0.0239, -0.0075, ..., 0.0002, 0.0016, -0.0136], + [-0.0613, -0.0059, -0.0441, ..., 0.0623, -0.1150, 0.0311]], + device='cuda:0'), grad: tensor([[ 3.2544e-05, -1.2204e-05, 0.0000e+00, ..., 3.4302e-05, + 4.1397e-07, 2.4214e-06], + [ 4.6206e-04, -3.7968e-05, 0.0000e+00, ..., -1.3685e-04, + 1.9526e-04, 2.9850e-04], + [-1.5140e-04, 5.1931e-06, 0.0000e+00, ..., 1.7494e-05, + -2.6774e-04, 2.5555e-05], + ..., + [-1.4515e-03, -5.6140e-06, 0.0000e+00, ..., -1.2338e-04, + -3.8838e-04, -7.9298e-04], + [ 5.7507e-04, 1.0535e-05, 0.0000e+00, ..., 1.0288e-04, + 2.3496e-04, 2.3663e-04], + [ 2.4629e-04, 6.7204e-06, 0.0000e+00, ..., 3.0264e-05, + 8.2374e-05, 1.2106e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0219, -0.0175, 0.0126, -0.0203, 0.0297, 0.0169, 0.0010, -0.0106, + -0.0147, -0.0034], device='cuda:0'), grad: tensor([ 8.5950e-05, 7.4053e-04, -5.3024e-04, 1.4269e-04, 2.6822e-04, + 2.1017e-04, -2.9802e-05, -2.4853e-03, 1.1349e-03, 4.6301e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 221.39, cls_loss 0.0046 cls_loss_mapping 0.0149 cls_loss_causal 0.7017 re_mapping 0.0120 re_causal 0.0390 /// teacc 98.93 lr 0.00010000 +Epoch 55, weight, value: tensor([[-7.2176e-02, 8.6397e-02, -5.4029e-03, ..., -5.4611e-02, + 2.0133e-03, -7.9563e-03], + [-1.0996e-04, -3.7073e-02, -2.3919e-02, ..., 1.9474e-02, + -5.8294e-03, -2.1906e-02], + [-6.5552e-02, -7.4036e-02, 2.9647e-03, ..., -4.6805e-03, + -3.1414e-03, -1.8535e-02], + ..., + [ 3.5616e-02, 3.7037e-02, -1.8806e-02, ..., -2.9336e-02, + 3.9233e-02, 3.3840e-02], + [ 3.4053e-02, -2.4045e-02, -7.5287e-03, ..., -3.1543e-05, + 1.1349e-03, -1.4024e-02], + [-6.1832e-02, -6.1282e-03, -4.4135e-02, ..., 6.2954e-02, + -1.1622e-01, 3.1481e-02]], device='cuda:0'), grad: tensor([[ 1.4044e-05, -7.2196e-06, 0.0000e+00, ..., 7.2531e-06, + 7.6145e-06, 7.1265e-06], + [ 9.9897e-05, 3.6508e-06, 0.0000e+00, ..., 4.1015e-06, + 9.5308e-05, 6.0320e-05], + [ 3.9041e-05, 7.7635e-06, 0.0000e+00, ..., 5.7630e-06, + 1.0327e-05, 2.4930e-05], + ..., + [-2.5940e-04, -4.4197e-05, 0.0000e+00, ..., 2.6286e-05, + -2.3961e-04, -1.6153e-04], + [-5.3570e-06, 3.9302e-06, 0.0000e+00, ..., 3.3826e-05, + 2.5675e-05, 2.9653e-05], + [ 5.7101e-05, 1.1347e-05, 0.0000e+00, ..., -5.7489e-05, + 3.1501e-05, -4.7207e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0214, -0.0174, 0.0122, -0.0203, 0.0296, 0.0172, 0.0014, -0.0102, + -0.0149, -0.0036], device='cuda:0'), grad: tensor([ 4.2766e-05, 2.6202e-04, -1.3327e-06, 1.5759e-04, -2.7597e-05, + 1.9407e-04, -1.0073e-04, -5.6505e-04, 4.6164e-05, -8.5905e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 220.91, cls_loss 0.0046 cls_loss_mapping 0.0161 cls_loss_causal 0.6731 re_mapping 0.0115 re_causal 0.0376 /// teacc 98.87 lr 0.00010000 +Epoch 56, weight, value: tensor([[-7.2568e-02, 8.6877e-02, -5.8136e-03, ..., -5.4798e-02, + 2.1685e-03, -7.8439e-03], + [ 5.4685e-05, -3.7505e-02, -2.3951e-02, ..., 1.9525e-02, + -6.0424e-03, -2.2274e-02], + [-6.6257e-02, -7.4615e-02, 2.8815e-03, ..., -4.8897e-03, + -3.1396e-03, -1.9128e-02], + ..., + [ 3.6129e-02, 3.7681e-02, -1.9107e-02, ..., -2.9752e-02, + 3.9771e-02, 3.4251e-02], + [ 3.4081e-02, -2.4505e-02, -7.7546e-03, ..., -2.8677e-04, + 5.6306e-04, -1.4653e-02], + [-6.2432e-02, -6.3567e-03, -4.4707e-02, ..., 6.3389e-02, + -1.1683e-01, 3.2167e-02]], device='cuda:0'), grad: tensor([[ 1.1027e-05, -2.1949e-05, 0.0000e+00, ..., -1.6112e-06, + 4.4145e-06, -7.8231e-06], + [ 1.4153e-03, 2.5798e-07, 0.0000e+00, ..., 3.1495e-04, + 5.6252e-06, 5.2340e-06], + [-4.1187e-05, 2.2277e-06, 0.0000e+00, ..., 1.1101e-05, + -1.5938e-04, -5.9940e-06], + ..., + [ 5.8919e-05, 4.7637e-07, 0.0000e+00, ..., 2.3007e-05, + 1.6582e-04, 1.0276e-04], + [-1.5736e-03, 3.4133e-07, 0.0000e+00, ..., -3.5214e-04, + 1.2076e-04, 8.1718e-05], + [ 8.0943e-05, 1.1519e-05, 0.0000e+00, ..., -9.3699e-05, + 1.5259e-05, -5.7757e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0212, -0.0176, 0.0120, -0.0204, 0.0296, 0.0173, 0.0018, -0.0098, + -0.0150, -0.0036], device='cuda:0'), grad: tensor([ 1.2912e-05, 2.8934e-03, -4.9067e-04, -9.1076e-05, 2.1863e-04, + -2.5839e-05, 4.1932e-05, 4.3964e-04, -3.0365e-03, 3.6120e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 221.28, cls_loss 0.0058 cls_loss_mapping 0.0182 cls_loss_causal 0.6790 re_mapping 0.0119 re_causal 0.0370 /// teacc 98.81 lr 0.00010000 +Epoch 57, weight, value: tensor([[-7.2897e-02, 8.7638e-02, -6.2224e-03, ..., -5.5206e-02, + 1.6315e-03, -7.6453e-03], + [ 4.2282e-04, -3.7886e-02, -2.3964e-02, ..., 2.0465e-02, + -6.3355e-03, -2.2256e-02], + [-6.6745e-02, -7.5230e-02, 2.8744e-03, ..., -5.2249e-03, + -2.2338e-03, -1.9102e-02], + ..., + [ 3.6460e-02, 3.7862e-02, -1.9204e-02, ..., -3.0786e-02, + 3.9187e-02, 3.4333e-02], + [ 3.4271e-02, -2.5016e-02, -7.9140e-03, ..., -6.8168e-04, + 4.1204e-05, -1.5379e-02], + [-6.3125e-02, -6.6369e-03, -4.4907e-02, ..., 6.3702e-02, + -1.1782e-01, 3.2361e-02]], device='cuda:0'), grad: tensor([[ 1.3046e-05, -1.4372e-05, 3.5344e-07, ..., 2.8741e-06, + 1.1593e-05, 2.2119e-07], + [-1.0896e-04, 6.0862e-07, 5.6997e-07, ..., -5.3614e-05, + -2.4773e-06, -1.1548e-05], + [ 4.4584e-05, 2.4512e-06, 1.8184e-07, ..., 9.7081e-06, + -6.1452e-05, 4.1187e-05], + ..., + [-1.4555e-04, 1.6438e-07, 3.3178e-08, ..., 3.8445e-05, + -7.8976e-06, -1.1301e-04], + [ 4.8310e-05, 8.4797e-07, 1.5479e-06, ..., 1.7211e-05, + 5.0515e-05, 4.5598e-05], + [ 4.0561e-05, 4.3139e-06, 2.1223e-07, ..., -3.1024e-05, + 1.9997e-05, -1.2487e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0211, -0.0169, 0.0123, -0.0204, 0.0297, 0.0172, 0.0022, -0.0104, + -0.0149, -0.0039], device='cuda:0'), grad: tensor([ 5.3793e-05, -1.4246e-04, -2.2018e-04, -4.1686e-06, 7.2360e-05, + 2.7925e-05, 4.7386e-05, -5.2720e-05, 1.9383e-04, 2.4691e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 220.61, cls_loss 0.0051 cls_loss_mapping 0.0148 cls_loss_causal 0.6433 re_mapping 0.0113 re_causal 0.0349 /// teacc 98.78 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0732, 0.0885, -0.0064, ..., -0.0555, 0.0020, -0.0074], + [ 0.0009, -0.0383, -0.0240, ..., 0.0206, -0.0066, -0.0227], + [-0.0673, -0.0757, 0.0029, ..., -0.0054, -0.0020, -0.0195], + ..., + [ 0.0363, 0.0380, -0.0198, ..., -0.0314, 0.0389, 0.0344], + [ 0.0343, -0.0253, -0.0081, ..., -0.0010, -0.0004, -0.0158], + [-0.0638, -0.0070, -0.0453, ..., 0.0636, -0.1187, 0.0327]], + device='cuda:0'), grad: tensor([[ 2.8163e-05, 1.5646e-06, 4.4703e-08, ..., 2.9460e-05, + -1.8906e-06, 2.2590e-05], + [ 1.6436e-05, 6.8024e-06, 1.3039e-08, ..., 2.4229e-05, + 7.0445e-06, 1.1824e-05], + [ 1.9684e-05, 1.7174e-06, 1.0361e-08, ..., 2.3037e-05, + -5.7295e-06, 9.3877e-06], + ..., + [-5.2750e-05, -5.0426e-05, 1.6764e-08, ..., 2.7761e-05, + -7.5884e-06, -3.7134e-05], + [ 1.1995e-05, 2.1346e-06, 3.2363e-07, ..., 8.7619e-05, + 1.0133e-05, 5.1439e-05], + [ 2.3156e-05, 1.2085e-05, 5.5181e-08, ..., -2.1422e-04, + 1.0118e-05, -1.2791e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0211, -0.0168, 0.0128, -0.0202, 0.0304, 0.0172, 0.0021, -0.0108, + -0.0152, -0.0044], device='cuda:0'), grad: tensor([ 9.2745e-05, 7.1764e-05, 2.8452e-07, 1.9595e-05, 1.3220e-04, + -6.2346e-05, -8.4758e-05, -5.2392e-05, 1.7893e-04, -2.9612e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 221.21, cls_loss 0.0050 cls_loss_mapping 0.0151 cls_loss_causal 0.6316 re_mapping 0.0110 re_causal 0.0343 /// teacc 98.91 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0736, 0.0899, -0.0064, ..., -0.0555, 0.0028, -0.0073], + [ 0.0006, -0.0401, -0.0240, ..., 0.0204, -0.0070, -0.0231], + [-0.0682, -0.0769, 0.0029, ..., -0.0056, -0.0024, -0.0209], + ..., + [ 0.0369, 0.0383, -0.0200, ..., -0.0320, 0.0398, 0.0351], + [ 0.0345, -0.0258, -0.0082, ..., -0.0015, -0.0008, -0.0163], + [-0.0641, -0.0072, -0.0457, ..., 0.0643, -0.1194, 0.0334]], + device='cuda:0'), grad: tensor([[ 3.1330e-06, -9.5293e-06, 0.0000e+00, ..., 3.6396e-06, + 1.9139e-07, -3.5297e-06], + [-1.2958e-04, 4.8662e-07, 0.0000e+00, ..., -2.1189e-05, + -6.0439e-05, -3.4630e-05], + [ 1.3322e-05, 2.0713e-06, 0.0000e+00, ..., 4.0308e-06, + 3.3947e-07, 7.4357e-06], + ..., + [ 8.9049e-05, -2.4810e-06, 0.0000e+00, ..., 8.8364e-06, + 4.5598e-05, 1.9237e-05], + [ 5.5246e-06, 9.6671e-07, 0.0000e+00, ..., 6.9961e-06, + 7.7635e-06, 2.9653e-06], + [ 2.7895e-05, 6.9998e-06, 0.0000e+00, ..., 3.2037e-05, + 6.5975e-06, 2.9467e-06]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0216, -0.0175, 0.0121, -0.0208, 0.0300, 0.0175, 0.0023, -0.0102, + -0.0153, -0.0038], device='cuda:0'), grad: tensor([ 6.5975e-06, -2.9588e-04, 1.8969e-05, 2.1726e-05, -4.9144e-05, + -9.5814e-06, 8.1509e-06, 2.1935e-04, 1.4439e-05, 6.5029e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 220.80, cls_loss 0.0041 cls_loss_mapping 0.0144 cls_loss_causal 0.6365 re_mapping 0.0111 re_causal 0.0340 /// teacc 98.95 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0739, 0.0906, -0.0065, ..., -0.0557, 0.0034, -0.0070], + [ 0.0003, -0.0400, -0.0240, ..., 0.0204, -0.0073, -0.0235], + [-0.0686, -0.0778, 0.0029, ..., -0.0059, -0.0024, -0.0219], + ..., + [ 0.0373, 0.0383, -0.0201, ..., -0.0325, 0.0402, 0.0356], + [ 0.0346, -0.0264, -0.0079, ..., -0.0018, -0.0015, -0.0166], + [-0.0646, -0.0073, -0.0462, ..., 0.0647, -0.1200, 0.0339]], + device='cuda:0'), grad: tensor([[ 2.4904e-06, -3.8773e-05, -2.2370e-06, ..., 5.8338e-06, + -1.7155e-06, -7.2718e-06], + [-1.0175e-04, 1.6382e-06, 2.3574e-07, ..., -8.0824e-05, + 2.1264e-05, 2.0549e-05], + [ 9.7454e-06, 4.2096e-06, 6.7893e-07, ..., 8.9854e-06, + -3.7432e-05, 2.2367e-05], + ..., + [ 3.4332e-05, -4.8336e-07, 1.0803e-07, ..., 3.5465e-05, + 2.0817e-05, 8.0541e-06], + [-4.3780e-05, 1.8179e-06, 9.0851e-07, ..., -9.4250e-06, + 1.3471e-05, 1.0982e-05], + [ 2.0847e-05, 1.4119e-05, 5.4855e-07, ..., -6.0558e-05, + 1.8150e-05, -4.7199e-06]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0217, -0.0177, 0.0123, -0.0212, 0.0301, 0.0177, 0.0024, -0.0100, + -0.0156, -0.0038], device='cuda:0'), grad: tensor([-1.5736e-05, -1.1361e-04, -6.2168e-05, -1.2708e-04, 1.4091e-04, + 9.4891e-05, 1.6496e-05, 1.2386e-04, -5.8353e-05, 5.8208e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 220.94, cls_loss 0.0043 cls_loss_mapping 0.0153 cls_loss_causal 0.6697 re_mapping 0.0116 re_causal 0.0368 /// teacc 98.93 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0745, 0.0911, -0.0066, ..., -0.0560, 0.0034, -0.0072], + [ 0.0004, -0.0406, -0.0247, ..., 0.0206, -0.0075, -0.0238], + [-0.0692, -0.0783, 0.0028, ..., -0.0061, -0.0023, -0.0223], + ..., + [ 0.0374, 0.0388, -0.0205, ..., -0.0334, 0.0404, 0.0355], + [ 0.0347, -0.0269, -0.0079, ..., -0.0023, -0.0020, -0.0171], + [-0.0652, -0.0073, -0.0470, ..., 0.0650, -0.1208, 0.0347]], + device='cuda:0'), grad: tensor([[ 8.3268e-05, 5.1320e-05, 3.7812e-07, ..., 4.9889e-05, + 8.3447e-06, 5.0403e-06], + [ 1.6820e-04, 5.3555e-05, 2.1455e-07, ..., -5.4926e-05, + 7.0669e-06, -3.1549e-07], + [ 3.5137e-05, 6.4932e-06, -3.0501e-08, ..., 8.1956e-06, + 6.5006e-06, 6.2585e-06], + ..., + [ 4.9353e-05, 1.8603e-07, 3.9069e-07, ..., 1.8820e-05, + -2.4065e-06, -3.2008e-05], + [-6.3276e-04, -1.4412e-04, 7.4506e-07, ..., 9.4101e-06, + 1.7598e-05, -1.2815e-05], + [ 7.9274e-05, 1.0036e-05, 2.6496e-07, ..., 3.5409e-06, + 1.8567e-05, 1.6272e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0214, -0.0176, 0.0121, -0.0213, 0.0304, 0.0178, 0.0031, -0.0104, + -0.0159, -0.0035], device='cuda:0'), grad: tensor([ 3.4070e-04, 6.8331e-04, 1.0777e-04, 2.7180e-04, 9.3579e-05, + 2.0409e-04, 2.1279e-04, 1.6057e-04, -2.2717e-03, 1.9467e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 220.74, cls_loss 0.0059 cls_loss_mapping 0.0188 cls_loss_causal 0.6592 re_mapping 0.0106 re_causal 0.0334 /// teacc 98.86 lr 0.00010000 +Epoch 62, weight, value: tensor([[-7.4889e-02, 9.2146e-02, -6.7970e-03, ..., -5.6346e-02, + 3.8993e-03, -6.9587e-03], + [ 7.0992e-05, -4.1111e-02, -2.5225e-02, ..., 2.0566e-02, + -7.6503e-03, -2.4275e-02], + [-6.9238e-02, -7.9071e-02, 2.8112e-03, ..., -6.5107e-03, + -2.0477e-03, -2.3439e-02], + ..., + [ 3.7938e-02, 3.9203e-02, -2.1067e-02, ..., -3.4214e-02, + 4.0797e-02, 3.5874e-02], + [ 3.4964e-02, -2.7258e-02, -1.0568e-02, ..., -2.7603e-03, + -2.5984e-03, -1.7847e-02], + [-6.5653e-02, -7.6623e-03, -4.8064e-02, ..., 6.5061e-02, + -1.2154e-01, 3.5581e-02]], device='cuda:0'), grad: tensor([[ 8.1286e-06, -5.2862e-06, 0.0000e+00, ..., 7.8529e-06, + 4.7982e-06, 1.9185e-06], + [-4.2915e-06, 7.4832e-07, 0.0000e+00, ..., -5.0552e-06, + 5.5209e-06, 3.7849e-06], + [ 7.1883e-05, 2.9206e-06, 0.0000e+00, ..., 1.2666e-05, + 4.9233e-05, 1.8939e-05], + ..., + [-1.8701e-05, -3.0622e-06, 0.0000e+00, ..., 1.0401e-05, + -1.3053e-05, -1.0870e-05], + [-1.0830e-04, 1.0468e-05, 0.0000e+00, ..., 1.1705e-05, + -4.5121e-05, 2.2158e-05], + [ 2.3142e-05, 4.3139e-06, 0.0000e+00, ..., -9.5785e-05, + 1.5013e-05, -6.6042e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0217, -0.0180, 0.0123, -0.0215, 0.0311, 0.0181, 0.0023, -0.0107, + -0.0158, -0.0034], device='cuda:0'), grad: tensor([ 3.7909e-05, 5.1744e-06, 2.4652e-04, -1.0192e-05, 1.3208e-04, + -1.6600e-05, -8.4043e-06, -3.1859e-05, -2.8920e-04, -6.5267e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 220.84, cls_loss 0.0033 cls_loss_mapping 0.0120 cls_loss_causal 0.6874 re_mapping 0.0102 re_causal 0.0346 /// teacc 98.74 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0752, 0.0928, -0.0069, ..., -0.0563, 0.0040, -0.0067], + [ 0.0004, -0.0416, -0.0253, ..., 0.0207, -0.0077, -0.0245], + [-0.0698, -0.0797, 0.0030, ..., -0.0067, -0.0017, -0.0236], + ..., + [ 0.0382, 0.0398, -0.0218, ..., -0.0345, 0.0410, 0.0362], + [ 0.0352, -0.0276, -0.0107, ..., -0.0031, -0.0032, -0.0182], + [-0.0660, -0.0081, -0.0492, ..., 0.0656, -0.1225, 0.0359]], + device='cuda:0'), grad: tensor([[ 7.4208e-06, 3.5986e-06, 0.0000e+00, ..., 1.4871e-05, + 3.5707e-06, 3.0417e-06], + [-1.1057e-04, 4.8503e-06, 0.0000e+00, ..., 5.2035e-05, + -6.1512e-05, 1.3816e-04], + [ 1.3757e-04, 2.1458e-06, 0.0000e+00, ..., 7.3798e-06, + 6.8367e-05, 7.6294e-06], + ..., + [-2.8276e-04, -1.1764e-05, 0.0000e+00, ..., -9.9719e-05, + -1.1265e-04, -1.8811e-04], + [ 2.4021e-05, 1.2200e-06, 0.0000e+00, ..., 1.7852e-05, + 1.2130e-05, 6.2734e-06], + [ 7.2062e-05, 4.1574e-06, 0.0000e+00, ..., 6.8426e-05, + 1.7971e-05, -1.0645e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0216, -0.0178, 0.0128, -0.0216, 0.0306, 0.0177, 0.0026, -0.0106, + -0.0158, -0.0034], device='cuda:0'), grad: tensor([ 4.0442e-05, -2.9349e-04, 3.1853e-04, 6.9141e-05, 1.4591e-04, + -1.4603e-05, 5.3614e-05, -5.7983e-04, 6.8307e-05, 1.9193e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 221.01, cls_loss 0.0035 cls_loss_mapping 0.0134 cls_loss_causal 0.6306 re_mapping 0.0104 re_causal 0.0331 /// teacc 98.87 lr 0.00010000 +Epoch 64, weight, value: tensor([[-7.5496e-02, 9.3501e-02, -6.9794e-03, ..., -5.6648e-02, + 4.1431e-03, -6.7076e-03], + [ 5.6886e-05, -4.2313e-02, -2.5277e-02, ..., 2.0711e-02, + -7.7130e-03, -2.4739e-02], + [-6.9950e-02, -8.0322e-02, 2.9624e-03, ..., -6.9391e-03, + -1.8660e-03, -2.4268e-02], + ..., + [ 3.8577e-02, 4.0461e-02, -2.1767e-02, ..., -3.5000e-02, + 4.1174e-02, 3.6529e-02], + [ 3.5262e-02, -2.7934e-02, -1.0756e-02, ..., -3.5484e-03, + -3.6464e-03, -1.8947e-02], + [-6.6322e-02, -8.4466e-03, -4.9238e-02, ..., 6.6246e-02, + -1.2344e-01, 3.6481e-02]], device='cuda:0'), grad: tensor([[ 2.3581e-06, -4.2152e-04, 9.9302e-08, ..., 4.2208e-06, + -2.0874e-04, -1.0949e-04], + [-4.1366e-05, 1.8012e-06, 1.3900e-07, ..., -7.4625e-05, + 2.2985e-06, 2.0675e-06], + [ 1.2830e-05, 3.6979e-04, 3.0710e-07, ..., 1.3500e-05, + 1.8859e-04, 1.0049e-04], + ..., + [-4.2208e-06, 1.6600e-05, 2.1770e-08, ..., 7.3649e-06, + -3.1777e-06, -5.1185e-06], + [-2.5779e-05, 3.5409e-06, -1.4538e-06, ..., 8.0541e-06, + 4.0121e-06, 7.2345e-06], + [ 1.5825e-05, 9.6560e-06, 5.7509e-08, ..., -2.4348e-05, + 9.9912e-06, -2.6464e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0216, -0.0185, 0.0133, -0.0218, 0.0304, 0.0180, 0.0024, -0.0105, + -0.0159, -0.0031], device='cuda:0'), grad: tensor([-4.5371e-04, -1.4985e-04, 4.6754e-04, 1.8418e-05, 9.7573e-05, + 7.0453e-05, 5.1051e-05, 1.5497e-05, -1.1784e-04, 1.2089e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 220.98, cls_loss 0.0045 cls_loss_mapping 0.0146 cls_loss_causal 0.6192 re_mapping 0.0106 re_causal 0.0330 /// teacc 98.94 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0761, 0.0944, -0.0076, ..., -0.0567, 0.0043, -0.0068], + [-0.0002, -0.0426, -0.0271, ..., 0.0206, -0.0077, -0.0246], + [-0.0702, -0.0811, 0.0029, ..., -0.0063, -0.0017, -0.0248], + ..., + [ 0.0391, 0.0407, -0.0228, ..., -0.0356, 0.0418, 0.0372], + [ 0.0355, -0.0284, -0.0113, ..., -0.0037, -0.0043, -0.0195], + [-0.0671, -0.0087, -0.0506, ..., 0.0666, -0.1246, 0.0367]], + device='cuda:0'), grad: tensor([[ 2.4848e-06, -6.9011e-07, 1.3749e-07, ..., 2.1085e-06, + 4.8801e-06, 1.8459e-06], + [ 5.5283e-06, 4.7637e-07, 3.3644e-08, ..., -9.1502e-07, + 1.5125e-05, 1.8016e-05], + [ 6.6869e-06, 6.3283e-07, 5.9605e-08, ..., 4.2915e-06, + -3.2961e-05, 5.1856e-06], + ..., + [-3.9041e-05, -4.2990e-06, 6.6473e-08, ..., 4.0382e-06, + -6.1572e-05, -1.0842e-04], + [ 4.5747e-06, 3.8650e-07, 6.1141e-07, ..., 4.4137e-05, + 9.1419e-06, 3.6985e-05], + [ 3.8981e-05, 1.9409e-06, 3.4552e-07, ..., 1.2964e-05, + 1.9297e-05, -1.6183e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0217, -0.0186, 0.0139, -0.0219, 0.0296, 0.0175, 0.0035, -0.0103, + -0.0159, -0.0034], device='cuda:0'), grad: tensor([ 2.0221e-05, 3.5256e-05, -1.3494e-04, 1.1396e-04, -1.0800e-04, + 6.4485e-06, 1.5944e-05, -1.0711e-04, 9.5785e-05, 6.3062e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 64---------------------------------------------------- +epoch 64, time 221.38, cls_loss 0.0041 cls_loss_mapping 0.0146 cls_loss_causal 0.6342 re_mapping 0.0102 re_causal 0.0320 /// teacc 99.00 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0765, 0.0949, -0.0101, ..., -0.0574, 0.0047, -0.0071], + [-0.0002, -0.0431, -0.0272, ..., 0.0205, -0.0079, -0.0247], + [-0.0706, -0.0821, 0.0028, ..., -0.0067, -0.0018, -0.0256], + ..., + [ 0.0395, 0.0412, -0.0247, ..., -0.0361, 0.0420, 0.0378], + [ 0.0359, -0.0284, -0.0117, ..., -0.0036, -0.0049, -0.0190], + [-0.0678, -0.0086, -0.0530, ..., 0.0670, -0.1263, 0.0366]], + device='cuda:0'), grad: tensor([[ 2.0579e-05, -4.3586e-06, 4.8708e-07, ..., 2.1189e-05, + 9.6932e-06, 2.6882e-05], + [ 7.1228e-05, 1.5751e-07, 1.4249e-07, ..., 5.3287e-05, + 4.2766e-05, 1.1384e-04], + [ 6.0499e-06, 1.1567e-06, 1.2713e-07, ..., 4.6313e-05, + 9.1642e-06, 1.1683e-05], + ..., + [-1.0166e-03, -7.5437e-07, 6.7870e-08, ..., -7.2050e-04, + -4.8709e-04, -1.4629e-03], + [ 1.0580e-05, 4.9267e-07, 1.4603e-06, ..., 2.2948e-05, + 1.4894e-05, 1.6481e-05], + [ 7.5960e-04, 1.8822e-06, 1.7381e-07, ..., 5.3167e-04, + 3.6454e-04, 1.0710e-03]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0212, -0.0187, 0.0138, -0.0221, 0.0297, 0.0177, 0.0035, -0.0098, + -0.0153, -0.0039], device='cuda:0'), grad: tensor([ 7.9572e-05, 2.1839e-04, 2.0039e-04, 6.9857e-05, 2.5654e-04, + 2.4045e-04, -4.3392e-04, -2.8152e-03, 7.6354e-05, 2.1057e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 220.87, cls_loss 0.0050 cls_loss_mapping 0.0155 cls_loss_causal 0.6177 re_mapping 0.0103 re_causal 0.0309 /// teacc 98.94 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0777, 0.0952, -0.0120, ..., -0.0580, 0.0048, -0.0075], + [-0.0002, -0.0440, -0.0263, ..., 0.0199, -0.0077, -0.0256], + [-0.0709, -0.0829, 0.0024, ..., -0.0070, -0.0020, -0.0262], + ..., + [ 0.0397, 0.0420, -0.0310, ..., -0.0366, 0.0420, 0.0380], + [ 0.0357, -0.0291, -0.0104, ..., -0.0042, -0.0057, -0.0194], + [-0.0677, -0.0087, -0.0616, ..., 0.0678, -0.1272, 0.0373]], + device='cuda:0'), grad: tensor([[ 4.1872e-06, -8.3387e-05, 1.5097e-06, ..., 4.8243e-06, + -1.5438e-05, -3.2902e-05], + [-1.4342e-05, -9.6858e-06, -4.3660e-06, ..., -2.9579e-05, + 1.3128e-05, 1.7986e-05], + [ 1.2688e-05, 2.0131e-05, 7.5735e-06, ..., 2.4915e-05, + 5.8934e-06, 1.9968e-05], + ..., + [-5.6177e-05, -1.2763e-05, 4.5192e-07, ..., 5.5321e-07, + -2.8521e-05, -4.4733e-05], + [ 6.8136e-06, 7.0743e-06, 3.0175e-06, ..., 1.5348e-05, + 1.3784e-05, 1.2159e-05], + [ 1.6183e-05, 5.2392e-05, 2.9849e-07, ..., -3.6478e-05, + 3.2395e-05, 5.1484e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0204, -0.0182, 0.0130, -0.0220, 0.0297, 0.0186, 0.0027, -0.0100, + -0.0154, -0.0033], device='cuda:0'), grad: tensor([-6.3300e-05, -4.0203e-05, 6.3181e-05, -4.7773e-05, 8.3029e-05, + 4.6760e-05, -8.4102e-05, -8.3625e-05, 6.8545e-05, 5.7369e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 220.97, cls_loss 0.0031 cls_loss_mapping 0.0138 cls_loss_causal 0.5931 re_mapping 0.0103 re_causal 0.0315 /// teacc 98.79 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0780, 0.0960, -0.0106, ..., -0.0582, 0.0054, -0.0072], + [ 0.0004, -0.0443, -0.0264, ..., 0.0207, -0.0082, -0.0252], + [-0.0714, -0.0837, 0.0022, ..., -0.0072, -0.0020, -0.0269], + ..., + [ 0.0389, 0.0423, -0.0335, ..., -0.0380, 0.0420, 0.0382], + [ 0.0359, -0.0294, -0.0106, ..., -0.0044, -0.0061, -0.0197], + [-0.0690, -0.0090, -0.0651, ..., 0.0674, -0.1285, 0.0371]], + device='cuda:0'), grad: tensor([[ 5.7705e-06, 8.7544e-07, 1.7462e-09, ..., 1.1593e-05, + 2.7008e-06, 2.5351e-06], + [ 1.6809e-05, 2.8983e-06, 3.4925e-09, ..., 2.4110e-05, + 2.5295e-06, 2.8852e-06], + [ 4.0680e-05, 3.2037e-06, -2.5146e-08, ..., 7.4506e-05, + -1.1973e-05, 3.0864e-06], + ..., + [-5.1670e-06, -1.2308e-05, 3.8417e-09, ..., 2.2516e-05, + -6.7577e-06, -1.6913e-05], + [ 5.8450e-06, 1.4156e-06, 8.6147e-09, ..., 3.8862e-05, + 3.5539e-06, 8.7619e-06], + [ 3.9995e-05, 5.2750e-06, 1.1642e-10, ..., 3.8117e-05, + 4.0270e-06, -1.4350e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0205, -0.0181, 0.0130, -0.0216, 0.0309, 0.0185, 0.0026, -0.0106, + -0.0153, -0.0041], device='cuda:0'), grad: tensor([ 2.7537e-05, 5.2989e-05, 8.7023e-05, 2.8566e-05, -4.8447e-04, + 2.9832e-05, 1.0926e-04, 4.7013e-06, 5.2571e-05, 9.1791e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 221.12, cls_loss 0.0030 cls_loss_mapping 0.0112 cls_loss_causal 0.6020 re_mapping 0.0103 re_causal 0.0317 /// teacc 98.92 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0783, 0.0966, -0.0138, ..., -0.0584, 0.0056, -0.0068], + [ 0.0004, -0.0444, -0.0267, ..., 0.0207, -0.0084, -0.0253], + [-0.0723, -0.0841, 0.0013, ..., -0.0076, -0.0019, -0.0273], + ..., + [ 0.0393, 0.0425, -0.0374, ..., -0.0383, 0.0421, 0.0382], + [ 0.0364, -0.0301, -0.0085, ..., -0.0046, -0.0068, -0.0201], + [-0.0695, -0.0092, -0.0687, ..., 0.0675, -0.1293, 0.0374]], + device='cuda:0'), grad: tensor([[ 1.4482e-06, -1.7807e-05, 9.2317e-08, ..., 1.1725e-06, + 3.5074e-06, -2.6566e-07], + [-2.6390e-05, 3.1805e-07, -3.5483e-06, ..., -2.8759e-05, + 9.1642e-06, 9.8124e-06], + [ 2.0619e-06, 1.0803e-06, 3.3760e-08, ..., 1.3774e-06, + -2.3276e-05, -1.9461e-05], + ..., + [-7.2531e-06, -2.1439e-06, 2.0128e-07, ..., 2.8625e-05, + 1.9252e-05, 2.4408e-05], + [ 1.8328e-06, 4.9137e-06, 3.8813e-07, ..., 1.0617e-05, + 1.8701e-05, 2.6971e-05], + [-2.6301e-06, 5.3085e-06, 5.6345e-08, ..., -3.6836e-05, + 8.7142e-05, 7.1228e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0205, -0.0181, 0.0128, -0.0218, 0.0315, 0.0182, 0.0031, -0.0107, + -0.0150, -0.0043], device='cuda:0'), grad: tensor([-5.6103e-06, -2.4483e-05, -1.4770e-04, -2.7823e-04, 4.1127e-05, + 4.0233e-05, 2.6047e-05, 1.6057e-04, 5.3406e-05, 1.3471e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 68---------------------------------------------------- +epoch 68, time 221.57, cls_loss 0.0045 cls_loss_mapping 0.0136 cls_loss_causal 0.6285 re_mapping 0.0097 re_causal 0.0292 /// teacc 99.03 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0786, 0.0982, -0.0137, ..., -0.0583, 0.0062, -0.0065], + [ 0.0002, -0.0453, -0.0298, ..., 0.0206, -0.0082, -0.0254], + [-0.0721, -0.0854, 0.0048, ..., -0.0081, -0.0018, -0.0279], + ..., + [ 0.0390, 0.0428, -0.0404, ..., -0.0390, 0.0417, 0.0383], + [ 0.0366, -0.0313, -0.0081, ..., -0.0052, -0.0074, -0.0207], + [-0.0697, -0.0097, -0.0712, ..., 0.0688, -0.1306, 0.0383]], + device='cuda:0'), grad: tensor([[ 7.0184e-06, -1.0207e-05, 1.3330e-07, ..., 7.4133e-06, + -1.6093e-06, 3.9674e-06], + [-5.3167e-05, -7.5027e-06, -9.4101e-06, ..., -3.6925e-05, + -3.1739e-06, 4.1090e-06], + [ 9.1046e-06, 3.7402e-06, 1.2722e-06, ..., 7.1414e-06, + 5.4836e-06, 4.2245e-06], + ..., + [-2.1353e-05, -1.6272e-05, 2.4214e-06, ..., 2.5004e-05, + -7.7635e-06, -1.1422e-05], + [ 1.4760e-05, 8.0690e-06, 5.9418e-07, ..., 3.1918e-05, + 8.5980e-06, 3.0369e-05], + [ 1.5959e-05, 1.1623e-05, 3.2922e-07, ..., -1.1110e-04, + 7.5661e-06, -7.9393e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0214, -0.0184, 0.0133, -0.0218, 0.0310, 0.0180, 0.0027, -0.0114, + -0.0153, -0.0033], device='cuda:0'), grad: tensor([ 1.4052e-05, -1.3661e-04, 3.2991e-05, 2.2814e-05, 1.0157e-04, + 4.3541e-05, 2.3797e-05, 2.1294e-05, 1.0139e-04, -2.2483e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 69---------------------------------------------------- +epoch 69, time 221.71, cls_loss 0.0032 cls_loss_mapping 0.0108 cls_loss_causal 0.6136 re_mapping 0.0093 re_causal 0.0298 /// teacc 99.09 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0794, 0.0984, -0.0135, ..., -0.0583, 0.0063, -0.0064], + [ 0.0003, -0.0462, -0.0298, ..., 0.0208, -0.0084, -0.0258], + [-0.0729, -0.0861, 0.0048, ..., -0.0084, -0.0019, -0.0284], + ..., + [ 0.0399, 0.0437, -0.0399, ..., -0.0392, 0.0422, 0.0390], + [ 0.0370, -0.0320, -0.0081, ..., -0.0054, -0.0079, -0.0211], + [-0.0709, -0.0100, -0.0734, ..., 0.0685, -0.1317, 0.0383]], + device='cuda:0'), grad: tensor([[ 2.5108e-06, -6.2823e-05, 0.0000e+00, ..., 6.0312e-06, + -1.6749e-05, -2.3335e-05], + [-2.9758e-05, 2.0079e-06, 0.0000e+00, ..., -2.6584e-05, + 2.9081e-07, 3.5837e-06], + [ 1.8924e-05, 8.6650e-06, 0.0000e+00, ..., 1.2606e-05, + -1.8537e-05, 7.6033e-06], + ..., + [ 9.9167e-06, -7.4226e-07, 0.0000e+00, ..., 4.6343e-05, + -7.1973e-06, 1.8269e-05], + [ 3.7123e-06, 2.1327e-06, 0.0000e+00, ..., 2.2307e-05, + 4.7907e-06, 1.8626e-05], + [-1.0073e-04, 2.5794e-05, 0.0000e+00, ..., -3.4547e-04, + 1.2748e-05, -2.8205e-04]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0208, -0.0183, 0.0130, -0.0218, 0.0320, 0.0179, 0.0028, -0.0108, + -0.0151, -0.0043], device='cuda:0'), grad: tensor([-6.0290e-05, -1.0377e-04, -3.3498e-05, 7.2658e-05, 1.7941e-04, + 2.9898e-04, 8.0466e-06, 7.9513e-05, 7.1824e-05, -5.1308e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 221.09, cls_loss 0.0032 cls_loss_mapping 0.0122 cls_loss_causal 0.6522 re_mapping 0.0096 re_causal 0.0311 /// teacc 98.99 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0798, 0.0994, -0.0139, ..., -0.0586, 0.0063, -0.0060], + [ 0.0003, -0.0465, -0.0298, ..., 0.0210, -0.0088, -0.0261], + [-0.0735, -0.0871, 0.0041, ..., -0.0087, -0.0024, -0.0288], + ..., + [ 0.0405, 0.0440, -0.0413, ..., -0.0397, 0.0427, 0.0392], + [ 0.0373, -0.0323, -0.0081, ..., -0.0057, -0.0082, -0.0215], + [-0.0713, -0.0102, -0.0745, ..., 0.0690, -0.1325, 0.0390]], + device='cuda:0'), grad: tensor([[ 5.4725e-06, -1.5408e-05, 0.0000e+00, ..., 4.4219e-06, + -8.4266e-06, -1.9036e-06], + [-1.8910e-05, 5.3905e-06, 0.0000e+00, ..., -2.2620e-05, + 1.7732e-05, 4.1611e-06], + [ 1.3031e-05, 5.5395e-06, 0.0000e+00, ..., 5.3495e-06, + 1.6659e-05, 1.6719e-05], + ..., + [-1.1325e-04, -4.6641e-05, 0.0000e+00, ..., 1.8284e-05, + -1.6719e-05, -8.0407e-05], + [ 2.3648e-05, 1.4767e-05, 0.0000e+00, ..., 1.6838e-05, + 6.4261e-06, 2.5630e-05], + [ 6.5923e-05, 1.2808e-05, 0.0000e+00, ..., -2.9743e-05, + 1.4663e-05, 1.4350e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0208, -0.0184, 0.0125, -0.0226, 0.0315, 0.0186, 0.0033, -0.0105, + -0.0149, -0.0041], device='cuda:0'), grad: tensor([-2.8405e-06, -6.0856e-05, 6.9976e-05, -1.8314e-05, 5.0157e-05, + 1.0759e-05, -5.3532e-06, -2.6917e-04, 9.4652e-05, 1.3089e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 221.18, cls_loss 0.0039 cls_loss_mapping 0.0143 cls_loss_causal 0.6273 re_mapping 0.0098 re_causal 0.0306 /// teacc 98.99 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0803, 0.0998, -0.0144, ..., -0.0591, 0.0063, -0.0063], + [ 0.0001, -0.0466, -0.0303, ..., 0.0214, -0.0093, -0.0266], + [-0.0737, -0.0876, 0.0042, ..., -0.0089, -0.0023, -0.0291], + ..., + [ 0.0409, 0.0448, -0.0403, ..., -0.0406, 0.0425, 0.0391], + [ 0.0376, -0.0329, -0.0083, ..., -0.0062, -0.0086, -0.0221], + [-0.0718, -0.0103, -0.0758, ..., 0.0697, -0.1332, 0.0399]], + device='cuda:0'), grad: tensor([[ 5.2154e-06, -1.4164e-05, 0.0000e+00, ..., 1.1750e-05, + 1.4171e-05, -3.6694e-06], + [-5.5581e-06, 2.2389e-06, 0.0000e+00, ..., -8.7246e-06, + 5.2527e-06, 2.3358e-06], + [-5.9716e-06, -1.1533e-05, 0.0000e+00, ..., 3.5986e-06, + -1.0389e-04, 1.2644e-05], + ..., + [-3.1646e-06, 7.3984e-06, 0.0000e+00, ..., 1.1370e-05, + 2.3276e-05, -3.3230e-05], + [ 2.2966e-06, 3.3304e-06, 0.0000e+00, ..., 1.4685e-05, + 1.1548e-05, 2.8126e-06], + [ 3.6955e-05, 9.0897e-06, 0.0000e+00, ..., 1.8135e-05, + 9.5665e-06, 1.0757e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0202, -0.0185, 0.0127, -0.0223, 0.0313, 0.0186, 0.0031, -0.0109, + -0.0147, -0.0037], device='cuda:0'), grad: tensor([ 9.3818e-05, 6.7847e-07, -3.7456e-04, 9.6679e-05, -4.2856e-05, + 4.8369e-05, -4.4405e-05, 1.4055e-04, -1.7462e-07, 8.1658e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 221.14, cls_loss 0.0034 cls_loss_mapping 0.0126 cls_loss_causal 0.6044 re_mapping 0.0099 re_causal 0.0310 /// teacc 98.85 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0806, 0.0995, -0.0145, ..., -0.0593, 0.0064, -0.0062], + [ 0.0013, -0.0473, -0.0304, ..., 0.0219, -0.0086, -0.0256], + [-0.0749, -0.0884, 0.0042, ..., -0.0091, -0.0029, -0.0300], + ..., + [ 0.0417, 0.0457, -0.0405, ..., -0.0413, 0.0431, 0.0390], + [ 0.0377, -0.0334, -0.0083, ..., -0.0066, -0.0092, -0.0226], + [-0.0719, -0.0106, -0.0767, ..., 0.0706, -0.1342, 0.0407]], + device='cuda:0'), grad: tensor([[ 8.1956e-06, -1.2565e-04, -1.2694e-06, ..., 5.4948e-06, + -4.9770e-05, -3.0175e-06], + [-5.9605e-05, -6.2808e-06, 7.3342e-09, ..., 6.1691e-05, + -1.2979e-05, 7.1451e-06], + [ 1.3068e-05, 1.4231e-05, 1.0442e-07, ..., 5.2862e-06, + 6.1616e-06, 9.9535e-08], + ..., + [ 1.2569e-05, -3.7998e-06, 1.4785e-08, ..., 2.9325e-05, + 9.2667e-07, 1.2673e-05], + [ 5.3912e-05, 1.3366e-05, 8.5216e-08, ..., 1.4581e-05, + 1.0788e-05, 1.3307e-05], + [ 2.2694e-05, 1.6570e-05, 1.4051e-07, ..., -3.4904e-04, + 6.0052e-06, -3.7956e-04]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0196, -0.0174, 0.0119, -0.0222, 0.0303, 0.0182, 0.0035, -0.0105, + -0.0150, -0.0031], device='cuda:0'), grad: tensor([-1.2827e-04, -3.5334e-04, 9.7081e-06, 7.0477e-04, -7.0238e-04, + 1.1158e-04, 8.4209e-04, 8.8274e-05, 2.6727e-04, -8.3828e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 221.10, cls_loss 0.0030 cls_loss_mapping 0.0110 cls_loss_causal 0.6055 re_mapping 0.0093 re_causal 0.0296 /// teacc 98.98 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0805, 0.1012, -0.0139, ..., -0.0596, 0.0072, -0.0053], + [ 0.0015, -0.0478, -0.0304, ..., 0.0221, -0.0087, -0.0259], + [-0.0752, -0.0891, 0.0041, ..., -0.0092, -0.0027, -0.0303], + ..., + [ 0.0419, 0.0461, -0.0410, ..., -0.0423, 0.0433, 0.0390], + [ 0.0376, -0.0339, -0.0084, ..., -0.0073, -0.0097, -0.0233], + [-0.0725, -0.0109, -0.0777, ..., 0.0710, -0.1352, 0.0414]], + device='cuda:0'), grad: tensor([[ 2.9542e-06, 1.1586e-06, 0.0000e+00, ..., 6.5193e-06, + 1.9260e-06, -1.0536e-07], + [-1.5438e-05, 2.0545e-06, 0.0000e+00, ..., -1.4536e-05, + 2.6412e-06, 3.0585e-06], + [ 5.8208e-08, 5.8226e-06, 0.0000e+00, ..., -2.0508e-06, + -8.7172e-06, 2.5611e-06], + ..., + [-4.7684e-07, -9.9838e-07, 0.0000e+00, ..., 3.8482e-06, + -9.1642e-06, -1.2293e-05], + [-3.8967e-06, 4.7199e-06, 0.0000e+00, ..., 7.4506e-06, + 6.2063e-06, 2.7027e-06], + [ 1.1101e-05, 3.2075e-06, 0.0000e+00, ..., 4.1164e-06, + 6.0834e-06, -3.9302e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0205, -0.0172, 0.0121, -0.0220, 0.0301, 0.0183, 0.0032, -0.0108, + -0.0156, -0.0030], device='cuda:0'), grad: tensor([ 5.4061e-05, -1.8433e-05, -1.3545e-05, 1.3077e-04, 1.6344e-04, + -8.6308e-05, -2.9421e-04, -1.0520e-05, 4.4286e-05, 3.0458e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 221.08, cls_loss 0.0028 cls_loss_mapping 0.0105 cls_loss_causal 0.6049 re_mapping 0.0092 re_causal 0.0292 /// teacc 98.89 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0810, 0.1020, -0.0137, ..., -0.0598, 0.0075, -0.0053], + [ 0.0012, -0.0483, -0.0304, ..., 0.0220, -0.0092, -0.0261], + [-0.0751, -0.0896, 0.0042, ..., -0.0095, -0.0025, -0.0308], + ..., + [ 0.0419, 0.0466, -0.0417, ..., -0.0428, 0.0433, 0.0392], + [ 0.0386, -0.0345, -0.0085, ..., -0.0074, -0.0095, -0.0235], + [-0.0732, -0.0110, -0.0780, ..., 0.0709, -0.1359, 0.0416]], + device='cuda:0'), grad: tensor([[ 4.4964e-06, -4.4852e-06, 0.0000e+00, ..., 4.1388e-06, + 1.2532e-05, 1.3322e-05], + [-1.8144e-04, 3.3015e-07, 0.0000e+00, ..., -1.5235e-04, + -2.8715e-05, -3.7849e-05], + [ 2.1368e-05, 1.2927e-06, 0.0000e+00, ..., 1.6183e-05, + -1.7928e-06, 8.3223e-06], + ..., + [ 5.6118e-05, -1.8170e-06, 0.0000e+00, ..., 4.4733e-05, + 1.2822e-05, 1.2167e-05], + [ 1.7345e-05, 4.6054e-07, 0.0000e+00, ..., 2.1115e-05, + 8.5086e-06, 9.4697e-06], + [ 9.6679e-05, 2.4680e-06, 0.0000e+00, ..., 1.5998e-04, + 4.0770e-05, 4.9680e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0205, -0.0177, 0.0124, -0.0217, 0.0306, 0.0177, 0.0031, -0.0110, + -0.0145, -0.0034], device='cuda:0'), grad: tensor([ 4.0859e-05, -5.4502e-04, 4.8280e-05, -1.6928e-04, -1.2839e-04, + 1.3113e-04, 1.6883e-05, 1.7226e-04, 5.4300e-05, 3.7932e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 220.97, cls_loss 0.0036 cls_loss_mapping 0.0123 cls_loss_causal 0.6271 re_mapping 0.0088 re_causal 0.0287 /// teacc 98.82 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0812, 0.1025, -0.0138, ..., -0.0602, 0.0074, -0.0053], + [ 0.0010, -0.0488, -0.0305, ..., 0.0227, -0.0096, -0.0263], + [-0.0748, -0.0903, 0.0043, ..., -0.0096, -0.0022, -0.0314], + ..., + [ 0.0421, 0.0474, -0.0421, ..., -0.0437, 0.0436, 0.0392], + [ 0.0388, -0.0348, -0.0086, ..., -0.0077, -0.0100, -0.0233], + [-0.0739, -0.0114, -0.0782, ..., 0.0713, -0.1371, 0.0422]], + device='cuda:0'), grad: tensor([[ 1.5423e-05, -2.4922e-06, 0.0000e+00, ..., 5.9903e-06, + 2.1998e-06, 6.5640e-06], + [-2.5973e-05, 2.4540e-07, 0.0000e+00, ..., -6.0171e-05, + 1.2398e-05, -3.5584e-05], + [ 1.2763e-05, 4.6100e-07, 0.0000e+00, ..., 5.7109e-06, + -8.5905e-06, 4.1388e-06], + ..., + [ 8.4192e-06, -3.1246e-07, 0.0000e+00, ..., 1.0245e-05, + 3.6836e-05, 4.2588e-05], + [-3.1561e-05, 1.9069e-07, 0.0000e+00, ..., 4.0494e-06, + 2.9765e-06, 5.8860e-06], + [ 5.3674e-05, 7.7812e-07, 0.0000e+00, ..., 1.8752e-04, + 1.7703e-05, 3.8862e-05]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0203, -0.0176, 0.0132, -0.0221, 0.0304, 0.0178, 0.0032, -0.0115, + -0.0143, -0.0033], device='cuda:0'), grad: tensor([ 9.3102e-05, -1.0329e-04, -1.4119e-05, -4.5091e-05, -2.0003e-04, + -3.9315e-04, 2.3687e-04, 1.1122e-04, -1.0794e-04, 4.2319e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 220.78, cls_loss 0.0035 cls_loss_mapping 0.0107 cls_loss_causal 0.6077 re_mapping 0.0088 re_causal 0.0276 /// teacc 98.92 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0816, 0.1028, -0.0135, ..., -0.0598, 0.0082, -0.0063], + [ 0.0003, -0.0513, -0.0305, ..., 0.0223, -0.0106, -0.0283], + [-0.0751, -0.0916, 0.0042, ..., -0.0099, -0.0021, -0.0318], + ..., + [ 0.0431, 0.0482, -0.0431, ..., -0.0441, 0.0444, 0.0402], + [ 0.0391, -0.0353, -0.0088, ..., -0.0081, -0.0105, -0.0238], + [-0.0746, -0.0102, -0.0772, ..., 0.0717, -0.1384, 0.0435]], + device='cuda:0'), grad: tensor([[ 1.1928e-05, -1.8477e-04, 1.9791e-09, ..., 1.7196e-05, + -3.5137e-05, -3.2187e-05], + [-1.0066e-05, 6.2697e-06, 5.8208e-10, ..., -1.2875e-05, + 1.0915e-05, 1.8194e-05], + [ 2.0862e-05, 3.2604e-05, -8.9640e-09, ..., 9.0301e-06, + 3.3639e-06, 1.4767e-05], + ..., + [-1.9133e-05, 7.9349e-06, 1.1642e-09, ..., 3.2246e-05, + -1.8463e-05, -3.4839e-05], + [ 1.2338e-05, 1.6794e-05, 2.0955e-09, ..., 1.8880e-05, + 9.3579e-06, 9.7081e-06], + [ 5.9068e-05, 8.1003e-05, 1.1642e-10, ..., -2.3574e-05, + 8.6963e-05, 5.4538e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0201, -0.0188, 0.0134, -0.0223, 0.0304, 0.0177, 0.0028, -0.0108, + -0.0139, -0.0026], device='cuda:0'), grad: tensor([-2.7418e-04, -4.1574e-05, 7.7069e-05, 7.5400e-05, 1.4257e-04, + -3.4046e-04, -5.3704e-05, 1.1211e-07, 1.0121e-04, 3.1304e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 220.94, cls_loss 0.0027 cls_loss_mapping 0.0097 cls_loss_causal 0.5898 re_mapping 0.0096 re_causal 0.0283 /// teacc 98.94 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0819, 0.1037, -0.0144, ..., -0.0603, 0.0096, -0.0061], + [ 0.0004, -0.0521, -0.0305, ..., 0.0225, -0.0111, -0.0285], + [-0.0753, -0.0927, 0.0044, ..., -0.0101, -0.0024, -0.0323], + ..., + [ 0.0432, 0.0485, -0.0443, ..., -0.0449, 0.0448, 0.0406], + [ 0.0389, -0.0357, -0.0089, ..., -0.0088, -0.0112, -0.0243], + [-0.0754, -0.0104, -0.0752, ..., 0.0720, -0.1400, 0.0439]], + device='cuda:0'), grad: tensor([[ 4.6231e-06, -1.1533e-05, 1.4820e-07, ..., 9.6299e-07, + -2.4941e-06, -2.8815e-06], + [-3.7146e-04, -3.2187e-05, 3.8836e-07, ..., -1.9476e-05, + -2.4691e-05, -1.2803e-04], + [ 3.3617e-05, 2.6152e-06, 3.0510e-06, ..., 5.3085e-06, + -2.4766e-05, 2.9415e-05], + ..., + [ 1.6868e-04, 1.9968e-05, -9.3728e-06, ..., 9.2760e-06, + 8.0690e-06, 2.6345e-05], + [ 8.3387e-05, 7.6890e-06, 1.3541e-06, ..., 8.6427e-06, + 3.7283e-05, 3.6329e-05], + [ 1.8120e-05, 5.7444e-06, -2.0742e-05, ..., -8.4341e-05, + 1.4775e-05, -2.2769e-05]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0201, -0.0191, 0.0136, -0.0223, 0.0306, 0.0178, 0.0037, -0.0109, + -0.0146, -0.0028], device='cuda:0'), grad: tensor([ 7.4245e-06, -1.2426e-03, -1.0118e-05, 1.3566e-04, 1.1504e-04, + -1.5303e-05, 1.8597e-05, 6.5136e-04, 3.6597e-04, -2.6196e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 220.84, cls_loss 0.0031 cls_loss_mapping 0.0090 cls_loss_causal 0.6186 re_mapping 0.0092 re_causal 0.0294 /// teacc 99.07 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0822, 0.1048, -0.0150, ..., -0.0586, 0.0104, -0.0047], + [ 0.0006, -0.0523, -0.0306, ..., 0.0226, -0.0120, -0.0289], + [-0.0756, -0.0935, 0.0048, ..., -0.0106, -0.0027, -0.0336], + ..., + [ 0.0433, 0.0486, -0.0489, ..., -0.0454, 0.0455, 0.0411], + [ 0.0390, -0.0362, -0.0091, ..., -0.0090, -0.0119, -0.0247], + [-0.0760, -0.0112, -0.0699, ..., 0.0717, -0.1408, 0.0437]], + device='cuda:0'), grad: tensor([[ 4.7013e-06, -2.0933e-04, 2.4261e-07, ..., 2.0817e-05, + -7.5817e-05, -6.6936e-05], + [-6.6087e-06, 4.7795e-06, 7.2410e-08, ..., 2.9542e-06, + 5.6177e-06, 8.4043e-06], + [ 4.0084e-06, 5.4985e-06, -1.7253e-07, ..., 9.9912e-06, + 1.3418e-05, 1.0714e-05], + ..., + [ 4.9546e-07, -1.9139e-07, 7.5623e-07, ..., 5.7757e-05, + -9.5740e-06, 3.9414e-06], + [ 3.3170e-05, 9.5367e-06, 1.6834e-07, ..., 1.7130e-04, + 9.6112e-06, 6.1035e-05], + [-2.2364e-04, 8.1897e-05, -2.1029e-06, ..., -1.0204e-03, + 2.7835e-05, -3.0994e-04]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0213, -0.0196, 0.0135, -0.0223, 0.0308, 0.0179, 0.0032, -0.0103, + -0.0147, -0.0034], device='cuda:0'), grad: tensor([-2.1386e-04, 1.4819e-05, 4.4137e-05, 4.6670e-05, 1.8435e-03, + 1.4043e-04, 1.4913e-04, 1.1510e-04, 4.2272e-04, -2.5635e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 220.84, cls_loss 0.0030 cls_loss_mapping 0.0120 cls_loss_causal 0.6440 re_mapping 0.0092 re_causal 0.0294 /// teacc 98.93 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0824, 0.1050, -0.0155, ..., -0.0590, 0.0107, -0.0050], + [ 0.0012, -0.0526, -0.0310, ..., 0.0231, -0.0123, -0.0279], + [-0.0757, -0.0941, 0.0054, ..., -0.0108, -0.0024, -0.0343], + ..., + [ 0.0431, 0.0489, -0.0488, ..., -0.0462, 0.0451, 0.0406], + [ 0.0392, -0.0365, -0.0092, ..., -0.0092, -0.0121, -0.0250], + [-0.0764, -0.0107, -0.0700, ..., 0.0721, -0.1413, 0.0444]], + device='cuda:0'), grad: tensor([[ 5.8524e-06, -6.2466e-05, -4.0978e-08, ..., 1.7673e-05, + -2.6003e-05, -1.9997e-05], + [-9.0376e-06, 1.1595e-06, 8.1491e-10, ..., -1.5900e-05, + 3.3192e-06, 3.4496e-06], + [ 5.2676e-06, 8.2627e-06, 7.7998e-09, ..., 1.0505e-05, + 1.5572e-05, 1.2837e-05], + ..., + [ 3.4664e-06, 3.7551e-06, 5.9372e-09, ..., 5.2005e-06, + 1.1839e-05, 8.9556e-06], + [-4.5560e-06, 3.0417e-06, 3.8417e-09, ..., 3.9749e-06, + 3.2429e-06, 3.9153e-06], + [ 1.0602e-05, 2.2665e-05, 2.3283e-09, ..., 3.1050e-06, + 1.0535e-05, 3.1274e-06]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0209, -0.0199, 0.0148, -0.0225, 0.0307, 0.0180, 0.0029, -0.0113, + -0.0145, -0.0030], device='cuda:0'), grad: tensor([-2.7061e-05, -5.5999e-05, 8.3506e-05, -5.0694e-05, 2.0564e-05, + 1.0335e-04, -1.5008e-04, 4.1217e-05, -1.0765e-04, 1.4305e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 80---------------------------------------------------- +epoch 80, time 221.48, cls_loss 0.0035 cls_loss_mapping 0.0098 cls_loss_causal 0.5995 re_mapping 0.0085 re_causal 0.0261 /// teacc 99.14 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0827, 0.1055, -0.0154, ..., -0.0592, 0.0108, -0.0047], + [ 0.0003, -0.0530, -0.0308, ..., 0.0222, -0.0120, -0.0281], + [-0.0757, -0.0948, 0.0055, ..., -0.0112, -0.0020, -0.0350], + ..., + [ 0.0439, 0.0494, -0.0499, ..., -0.0467, 0.0447, 0.0411], + [ 0.0391, -0.0369, -0.0094, ..., -0.0096, -0.0127, -0.0259], + [-0.0766, -0.0109, -0.0708, ..., 0.0727, -0.1425, 0.0445]], + device='cuda:0'), grad: tensor([[ 3.3118e-06, 5.0813e-06, 4.0629e-08, ..., 7.7784e-06, + 2.4531e-06, -3.8557e-06], + [-6.1244e-06, 5.6578e-07, 1.7229e-08, ..., -6.9439e-06, + -2.2507e-04, -8.6248e-05], + [ 8.8066e-06, 9.4436e-07, 1.1758e-08, ..., 3.1125e-06, + 1.1975e-04, 5.0455e-05], + ..., + [-5.7444e-06, -2.1560e-07, 4.9011e-08, ..., 1.8198e-06, + 8.7321e-05, 3.1114e-05], + [ 1.8459e-06, 2.4009e-06, 6.8569e-08, ..., 9.7677e-06, + 1.3202e-05, 3.6731e-06], + [ 7.3835e-06, 4.6231e-06, 1.0885e-07, ..., 3.7719e-06, + 2.0862e-05, 1.0550e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0207, -0.0206, 0.0149, -0.0230, 0.0307, 0.0182, 0.0033, -0.0105, + -0.0150, -0.0028], device='cuda:0'), grad: tensor([ 5.3972e-05, -9.6273e-04, 4.8804e-04, 9.4652e-05, 4.2260e-05, + -8.2254e-05, -1.1313e-04, 3.6550e-04, 5.5373e-05, 5.8323e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 220.72, cls_loss 0.0035 cls_loss_mapping 0.0122 cls_loss_causal 0.6171 re_mapping 0.0082 re_causal 0.0261 /// teacc 98.95 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0832, 0.1068, -0.0149, ..., -0.0595, 0.0118, -0.0043], + [ 0.0005, -0.0528, -0.0312, ..., 0.0224, -0.0122, -0.0286], + [-0.0765, -0.0960, 0.0058, ..., -0.0116, -0.0020, -0.0361], + ..., + [ 0.0442, 0.0501, -0.0513, ..., -0.0472, 0.0452, 0.0416], + [ 0.0385, -0.0381, -0.0098, ..., -0.0100, -0.0143, -0.0266], + [-0.0771, -0.0116, -0.0711, ..., 0.0730, -0.1442, 0.0449]], + device='cuda:0'), grad: tensor([[ 3.3360e-06, -3.8669e-06, -3.0873e-07, ..., 1.8952e-06, + -3.5600e-07, -6.5472e-07], + [-3.2544e-05, -3.4012e-06, -1.1793e-07, ..., -2.2411e-05, + 4.1872e-06, 6.9179e-06], + [ 8.4490e-06, 9.6858e-07, 3.1665e-08, ..., 2.7008e-06, + 4.9174e-07, 1.9148e-06], + ..., + [-2.2069e-05, 4.7358e-07, 4.5751e-08, ..., 6.6124e-06, + -2.1800e-05, -3.1829e-05], + [ 1.6659e-05, 1.3122e-06, 1.6519e-07, ..., 4.6752e-06, + 7.6108e-06, 2.7642e-06], + [ 9.1866e-06, 1.3346e-06, 4.5868e-08, ..., 1.1418e-06, + 3.9712e-06, -2.0873e-07]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0209, -0.0203, 0.0147, -0.0220, 0.0307, 0.0176, 0.0037, -0.0102, + -0.0161, -0.0031], device='cuda:0'), grad: tensor([ 5.9940e-06, -9.1136e-05, 2.2009e-05, 6.2048e-05, 1.5140e-05, + -1.6943e-05, 5.2154e-06, -3.7909e-05, 1.5795e-05, 1.9729e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 220.61, cls_loss 0.0029 cls_loss_mapping 0.0092 cls_loss_causal 0.5959 re_mapping 0.0082 re_causal 0.0254 /// teacc 98.94 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0839, 0.1077, -0.0151, ..., -0.0598, 0.0124, -0.0040], + [ 0.0006, -0.0526, -0.0317, ..., 0.0224, -0.0123, -0.0288], + [-0.0767, -0.0970, 0.0064, ..., -0.0120, -0.0019, -0.0371], + ..., + [ 0.0446, 0.0502, -0.0508, ..., -0.0477, 0.0452, 0.0418], + [ 0.0381, -0.0385, -0.0105, ..., -0.0102, -0.0155, -0.0264], + [-0.0778, -0.0119, -0.0714, ..., 0.0734, -0.1455, 0.0455]], + device='cuda:0'), grad: tensor([[ 6.6683e-06, -5.0180e-06, 9.5367e-07, ..., 1.3642e-05, + -1.0617e-06, -1.5309e-07], + [-5.6505e-05, 1.7812e-07, 7.1758e-07, ..., -6.7234e-05, + -1.5488e-06, 2.0508e-06], + [ 3.1769e-05, 1.7406e-06, 5.2564e-06, ..., 6.1750e-05, + 3.2596e-06, 3.2224e-06], + ..., + [ 6.3777e-06, -6.4494e-07, 2.1365e-06, ..., 3.0294e-05, + -7.2867e-06, -9.2164e-06], + [ 2.4617e-05, 1.3607e-06, 3.7346e-06, ..., 5.1498e-05, + 2.0973e-06, 5.8338e-06], + [ 3.9279e-05, 6.0257e-07, 4.6901e-06, ..., 3.6657e-05, + 1.3165e-05, -8.2031e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0209, -0.0202, 0.0149, -0.0219, 0.0306, 0.0177, 0.0042, -0.0103, + -0.0165, -0.0033], device='cuda:0'), grad: tensor([ 2.2069e-05, -1.7333e-04, 1.2720e-04, 5.0329e-06, -5.1546e-04, + -7.9023e-07, 2.8563e-04, 3.7819e-05, 1.0449e-04, 1.0735e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 220.55, cls_loss 0.0029 cls_loss_mapping 0.0107 cls_loss_causal 0.6312 re_mapping 0.0082 re_causal 0.0265 /// teacc 98.95 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0846, 0.1085, -0.0148, ..., -0.0600, 0.0129, -0.0039], + [ 0.0009, -0.0536, -0.0314, ..., 0.0228, -0.0126, -0.0289], + [-0.0775, -0.0980, 0.0066, ..., -0.0123, -0.0021, -0.0383], + ..., + [ 0.0449, 0.0511, -0.0508, ..., -0.0488, 0.0457, 0.0421], + [ 0.0383, -0.0392, -0.0107, ..., -0.0107, -0.0161, -0.0268], + [-0.0788, -0.0122, -0.0719, ..., 0.0736, -0.1467, 0.0464]], + device='cuda:0'), grad: tensor([[ 3.6526e-06, -4.1816e-07, 4.1211e-07, ..., 1.2122e-05, + 1.9237e-05, -6.6170e-07], + [-2.6131e-04, 1.7472e-06, -3.4094e-05, ..., 3.5107e-05, + 6.3814e-06, 3.1274e-06], + [ 9.0659e-05, 3.8370e-06, 1.0677e-05, ..., 4.6119e-06, + 1.9848e-05, 9.7901e-06], + ..., + [ 8.1733e-06, 2.3663e-05, 9.7789e-07, ..., 3.1292e-05, + 9.8586e-05, 2.2626e-04], + [-1.1615e-05, 2.4997e-06, 7.7859e-07, ..., 1.7853e-06, + -3.3099e-06, 3.6042e-06], + [ 2.2352e-05, -2.4602e-05, 1.1956e-07, ..., 3.7402e-05, + 1.7613e-05, -1.1164e-04]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0207, -0.0200, 0.0145, -0.0220, 0.0312, 0.0174, 0.0043, -0.0104, + -0.0164, -0.0034], device='cuda:0'), grad: tensor([ 1.3459e-04, -6.0511e-04, 2.9016e-04, -1.1247e-04, 8.0168e-06, + 1.4842e-04, -7.1228e-05, 3.2115e-04, -5.7191e-05, -5.7012e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 220.61, cls_loss 0.0028 cls_loss_mapping 0.0083 cls_loss_causal 0.5845 re_mapping 0.0084 re_causal 0.0262 /// teacc 98.99 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0852, 0.1097, -0.0152, ..., -0.0604, 0.0135, -0.0034], + [ 0.0010, -0.0540, -0.0316, ..., 0.0228, -0.0128, -0.0291], + [-0.0779, -0.0986, 0.0066, ..., -0.0126, -0.0019, -0.0388], + ..., + [ 0.0453, 0.0521, -0.0508, ..., -0.0492, 0.0458, 0.0425], + [ 0.0385, -0.0398, -0.0109, ..., -0.0107, -0.0166, -0.0273], + [-0.0794, -0.0126, -0.0721, ..., 0.0737, -0.1478, 0.0466]], + device='cuda:0'), grad: tensor([[ 1.5780e-05, -6.0834e-06, 1.6857e-07, ..., 2.3857e-05, + 6.5193e-08, 1.3530e-05], + [-1.3602e-04, 7.4431e-06, -1.0297e-05, ..., -1.2290e-04, + -2.0489e-06, 2.4199e-05], + [ 4.2260e-05, 1.2010e-05, 3.9637e-06, ..., 2.4334e-05, + 9.8050e-06, 3.2336e-05], + ..., + [-3.4285e-04, -4.5347e-04, 2.1956e-07, ..., 3.2157e-05, + -2.4164e-04, -1.5593e-03], + [ 6.1512e-05, 1.8522e-05, 5.1260e-06, ..., 6.9141e-05, + 1.1601e-05, 6.3360e-05], + [ 3.2496e-04, 3.9601e-04, 4.5169e-08, ..., 1.7643e-04, + 2.0230e-04, 1.3752e-03]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0208, -0.0201, 0.0144, -0.0222, 0.0314, 0.0191, 0.0024, -0.0102, + -0.0163, -0.0037], device='cuda:0'), grad: tensor([ 1.0574e-04, -4.8232e-04, 2.1863e-04, 7.0810e-05, -3.1304e-04, + 1.9658e-04, -1.6403e-04, -4.7569e-03, 4.0030e-04, 4.7226e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 220.41, cls_loss 0.0024 cls_loss_mapping 0.0098 cls_loss_causal 0.5999 re_mapping 0.0081 re_causal 0.0251 /// teacc 99.01 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0857, 0.1108, -0.0154, ..., -0.0606, 0.0151, -0.0030], + [ 0.0015, -0.0543, -0.0311, ..., 0.0233, -0.0130, -0.0292], + [-0.0784, -0.1001, 0.0065, ..., -0.0131, -0.0020, -0.0397], + ..., + [ 0.0455, 0.0539, -0.0513, ..., -0.0501, 0.0463, 0.0434], + [ 0.0388, -0.0404, -0.0110, ..., -0.0110, -0.0174, -0.0278], + [-0.0804, -0.0134, -0.0724, ..., 0.0738, -0.1496, 0.0464]], + device='cuda:0'), grad: tensor([[ 5.4948e-07, 1.0394e-05, 0.0000e+00, ..., 4.7684e-06, + 1.7146e-06, 1.0375e-06], + [-3.7067e-06, 9.5868e-08, 0.0000e+00, ..., -2.6878e-06, + 7.2597e-07, 8.3260e-07], + [ 2.1700e-06, 1.6857e-07, 0.0000e+00, ..., 1.1427e-06, + 2.1923e-06, 2.5369e-06], + ..., + [ 3.6992e-06, 8.1898e-08, 0.0000e+00, ..., 4.2655e-06, + 9.5135e-07, 1.4650e-06], + [-4.6901e-06, 3.6461e-07, 0.0000e+00, ..., 6.6608e-06, + 2.3674e-06, 5.7854e-06], + [ 4.4964e-06, 1.5588e-07, 0.0000e+00, ..., -2.6748e-05, + 2.2594e-06, -2.2247e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0210, -0.0197, 0.0140, -0.0221, 0.0318, 0.0188, 0.0024, -0.0096, + -0.0160, -0.0047], device='cuda:0'), grad: tensor([ 3.5167e-05, -7.8902e-06, 9.6411e-06, 1.1943e-05, 4.8988e-06, + 1.7881e-05, -4.5985e-05, 1.3970e-05, 6.7381e-07, -4.0352e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 220.80, cls_loss 0.0025 cls_loss_mapping 0.0089 cls_loss_causal 0.6006 re_mapping 0.0085 re_causal 0.0264 /// teacc 99.06 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0864, 0.1112, -0.0154, ..., -0.0608, 0.0155, -0.0030], + [ 0.0018, -0.0533, -0.0310, ..., 0.0236, -0.0131, -0.0292], + [-0.0789, -0.1007, 0.0067, ..., -0.0135, -0.0019, -0.0401], + ..., + [ 0.0457, 0.0541, -0.0517, ..., -0.0509, 0.0464, 0.0435], + [ 0.0390, -0.0410, -0.0110, ..., -0.0115, -0.0178, -0.0284], + [-0.0815, -0.0134, -0.0723, ..., 0.0744, -0.1502, 0.0478]], + device='cuda:0'), grad: tensor([[ 5.8580e-07, -5.5879e-06, 0.0000e+00, ..., 2.6785e-06, + 1.6123e-05, -7.0874e-07], + [-9.2983e-06, 2.1094e-07, 0.0000e+00, ..., -5.3272e-06, + 2.4781e-05, 3.4273e-06], + [ 4.0308e-06, 9.1270e-07, 0.0000e+00, ..., 1.4575e-06, + -3.8052e-04, 2.5164e-06], + ..., + [-9.0599e-05, 1.1089e-07, 0.0000e+00, ..., 2.0973e-06, + -7.9095e-05, -1.8275e-04], + [-2.5649e-06, 6.4634e-07, 0.0000e+00, ..., 1.0677e-05, + 5.5544e-06, 6.1654e-06], + [ 1.5527e-05, 1.7434e-06, 0.0000e+00, ..., -7.5102e-06, + 2.0012e-05, 2.3156e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0207, -0.0194, 0.0140, -0.0220, 0.0317, 0.0183, 0.0029, -0.0099, + -0.0161, -0.0043], device='cuda:0'), grad: tensor([ 4.3392e-05, 3.4630e-05, -7.8011e-04, 8.5831e-04, 2.6077e-05, + 2.3201e-05, -2.7984e-05, -2.3687e-04, 1.7926e-05, 4.0859e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 220.28, cls_loss 0.0024 cls_loss_mapping 0.0084 cls_loss_causal 0.6018 re_mapping 0.0084 re_causal 0.0258 /// teacc 98.99 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0867, 0.1120, -0.0155, ..., -0.0611, 0.0160, -0.0027], + [ 0.0018, -0.0535, -0.0310, ..., 0.0236, -0.0136, -0.0297], + [-0.0798, -0.1017, 0.0068, ..., -0.0138, -0.0021, -0.0411], + ..., + [ 0.0459, 0.0543, -0.0520, ..., -0.0515, 0.0461, 0.0436], + [ 0.0391, -0.0414, -0.0111, ..., -0.0118, -0.0183, -0.0287], + [-0.0817, -0.0135, -0.0721, ..., 0.0750, -0.1509, 0.0484]], + device='cuda:0'), grad: tensor([[ 4.6119e-06, -6.1393e-05, 1.0952e-06, ..., 1.6484e-06, + -3.5819e-06, -3.6824e-06], + [ 3.5297e-06, 1.6823e-05, 1.5879e-07, ..., 4.1798e-06, + 4.1686e-06, 2.0638e-05], + [ 1.3746e-05, 1.4290e-05, -1.2107e-05, ..., 3.9721e-07, + -1.2713e-06, 1.2524e-05], + ..., + [ 1.0226e-06, -1.0908e-05, 4.7730e-07, ..., 5.2899e-05, + -9.7556e-08, 8.4877e-05], + [ 5.5611e-05, 2.2784e-05, 2.3693e-06, ..., 3.2540e-06, + 2.8268e-05, 2.4393e-05], + [ 1.6138e-05, 5.1707e-06, 4.2543e-06, ..., -7.5161e-05, + 1.5765e-05, -1.1098e-04]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0209, -0.0197, 0.0138, -0.0214, 0.0315, 0.0177, 0.0037, -0.0102, + -0.0162, -0.0038], device='cuda:0'), grad: tensor([-5.6893e-05, 5.6744e-05, -1.8477e-05, 3.4857e-04, 3.8624e-05, + -7.1573e-04, 4.0323e-05, 1.5891e-04, 2.4629e-04, -9.8705e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 220.67, cls_loss 0.0030 cls_loss_mapping 0.0094 cls_loss_causal 0.5944 re_mapping 0.0083 re_causal 0.0250 /// teacc 99.00 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0872, 0.1128, -0.0157, ..., -0.0611, 0.0165, -0.0030], + [ 0.0020, -0.0526, -0.0310, ..., 0.0238, -0.0137, -0.0296], + [-0.0804, -0.1028, 0.0069, ..., -0.0143, -0.0025, -0.0419], + ..., + [ 0.0469, 0.0547, -0.0524, ..., -0.0519, 0.0466, 0.0443], + [ 0.0392, -0.0413, -0.0113, ..., -0.0120, -0.0188, -0.0289], + [-0.0831, -0.0136, -0.0717, ..., 0.0744, -0.1517, 0.0482]], + device='cuda:0'), grad: tensor([[ 2.5611e-06, -1.7345e-05, 1.4610e-07, ..., 1.7080e-06, + 1.9185e-06, -3.2429e-06], + [-1.2910e-04, 2.4978e-06, -1.8612e-05, ..., 3.9548e-05, + -3.6418e-05, -7.8231e-06], + [ 9.3281e-05, 1.5581e-06, 9.5591e-06, ..., 4.4346e-05, + 2.5660e-05, 1.1943e-05], + ..., + [ 7.1190e-06, -5.6140e-06, 5.3644e-07, ..., 1.2569e-05, + 1.6615e-05, -1.1347e-05], + [ 6.7353e-06, 1.2293e-06, 3.9418e-07, ..., 7.8753e-06, + 1.4767e-05, 5.3942e-06], + [ 5.9977e-06, 1.3195e-05, 6.6299e-08, ..., -2.6915e-07, + 2.2054e-05, 5.8040e-06]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0209, -0.0193, 0.0131, -0.0211, 0.0322, 0.0173, 0.0036, -0.0094, + -0.0160, -0.0050], device='cuda:0'), grad: tensor([ 6.0201e-06, -6.3801e-04, 4.2272e-04, -1.5128e-04, -1.6344e-04, + 1.4043e-04, 1.2130e-04, 9.9897e-05, 7.9870e-05, 8.2314e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 220.83, cls_loss 0.0027 cls_loss_mapping 0.0091 cls_loss_causal 0.5901 re_mapping 0.0080 re_causal 0.0234 /// teacc 99.14 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0877, 0.1142, -0.0157, ..., -0.0611, 0.0173, -0.0031], + [ 0.0018, -0.0549, -0.0318, ..., 0.0238, -0.0141, -0.0302], + [-0.0806, -0.1058, 0.0079, ..., -0.0149, -0.0026, -0.0427], + ..., + [ 0.0475, 0.0560, -0.0527, ..., -0.0523, 0.0467, 0.0451], + [ 0.0395, -0.0419, -0.0113, ..., -0.0122, -0.0196, -0.0292], + [-0.0837, -0.0137, -0.0721, ..., 0.0746, -0.1529, 0.0482]], + device='cuda:0'), grad: tensor([[ 6.2352e-07, -3.8370e-06, 0.0000e+00, ..., 1.1278e-06, + 2.1025e-07, -2.9104e-07], + [ 1.0012e-06, 2.6566e-07, 0.0000e+00, ..., 1.0021e-06, + 3.1292e-06, 1.9521e-06], + [ 3.1572e-06, 4.9733e-07, 0.0000e+00, ..., 4.7423e-06, + 6.2026e-06, 3.7998e-06], + ..., + [-5.1931e-06, -1.7211e-06, 0.0000e+00, ..., 2.8033e-06, + -6.6683e-06, -1.1064e-05], + [ 1.9372e-06, 5.1828e-07, 0.0000e+00, ..., 3.2280e-06, + 7.5251e-06, 3.9898e-06], + [ 5.5283e-06, 1.7071e-06, 0.0000e+00, ..., -2.5183e-06, + 7.0781e-06, 2.9244e-06]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0215, -0.0199, 0.0130, -0.0215, 0.0317, 0.0178, 0.0037, -0.0086, + -0.0157, -0.0056], device='cuda:0'), grad: tensor([ 2.6394e-06, 9.6560e-06, 2.0698e-05, 1.8626e-05, -2.0877e-05, + -4.8876e-05, -7.8231e-06, -1.8731e-05, 2.4065e-05, 2.0519e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 220.35, cls_loss 0.0021 cls_loss_mapping 0.0078 cls_loss_causal 0.6056 re_mapping 0.0078 re_causal 0.0258 /// teacc 98.98 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0880, 0.1146, -0.0155, ..., -0.0614, 0.0175, -0.0031], + [ 0.0019, -0.0552, -0.0321, ..., 0.0240, -0.0144, -0.0302], + [-0.0809, -0.1064, 0.0080, ..., -0.0151, -0.0023, -0.0433], + ..., + [ 0.0477, 0.0564, -0.0529, ..., -0.0531, 0.0468, 0.0452], + [ 0.0393, -0.0425, -0.0116, ..., -0.0126, -0.0204, -0.0297], + [-0.0839, -0.0138, -0.0723, ..., 0.0752, -0.1539, 0.0486]], + device='cuda:0'), grad: tensor([[ 1.2079e-06, -2.7250e-06, 4.5309e-07, ..., 7.4953e-06, + 4.6217e-07, 2.9150e-07], + [ 2.2743e-06, 9.6217e-08, 9.6508e-08, ..., 6.2920e-06, + 1.4352e-06, 1.2657e-06], + [ 3.1497e-06, 1.9663e-07, 1.2014e-07, ..., 4.3288e-06, + -1.1794e-05, 1.0040e-06], + ..., + [ 3.6489e-06, 2.2957e-07, 1.5774e-08, ..., 1.0714e-05, + 4.6915e-07, 3.4529e-07], + [ 1.0274e-05, 2.5961e-07, 4.2655e-06, ..., 5.1856e-05, + 2.4270e-06, 4.1761e-06], + [ 1.2946e-04, 1.4864e-06, 9.5693e-08, ..., 2.1827e-04, + 1.9073e-06, -2.0877e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0213, -0.0200, 0.0136, -0.0218, 0.0311, 0.0185, 0.0039, -0.0088, + -0.0162, -0.0055], device='cuda:0'), grad: tensor([ 2.0370e-05, 1.6853e-05, -2.8849e-05, 1.2624e-04, -4.5538e-04, + -7.3075e-05, -1.2517e-04, 2.5958e-05, 1.3566e-04, 3.5787e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 220.55, cls_loss 0.0022 cls_loss_mapping 0.0067 cls_loss_causal 0.6204 re_mapping 0.0079 re_causal 0.0248 /// teacc 98.93 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0883, 0.1150, -0.0157, ..., -0.0620, 0.0176, -0.0029], + [ 0.0019, -0.0556, -0.0317, ..., 0.0240, -0.0142, -0.0305], + [-0.0808, -0.1069, 0.0075, ..., -0.0154, -0.0018, -0.0439], + ..., + [ 0.0476, 0.0567, -0.0531, ..., -0.0537, 0.0463, 0.0453], + [ 0.0395, -0.0429, -0.0119, ..., -0.0131, -0.0211, -0.0302], + [-0.0845, -0.0143, -0.0722, ..., 0.0753, -0.1548, 0.0489]], + device='cuda:0'), grad: tensor([[ 1.1008e-06, 1.8179e-05, 1.0186e-08, ..., 6.6578e-05, + 6.1281e-07, 5.0277e-05], + [ 2.0489e-07, 3.6065e-07, 2.2701e-09, ..., -4.4284e-07, + 1.3094e-06, 1.5665e-06], + [ 3.2298e-06, 2.4540e-07, -3.2596e-08, ..., 3.3975e-06, + -4.0941e-06, 2.1104e-06], + ..., + [ 1.6481e-05, 6.4634e-07, 1.4843e-08, ..., 1.9133e-05, + 6.2399e-08, 1.8356e-06], + [-4.0919e-05, 2.7642e-05, 2.3108e-08, ..., 8.6427e-05, + -8.3148e-06, 6.7413e-05], + [ 9.5814e-06, -5.2065e-05, 1.7288e-08, ..., -1.6785e-04, + 2.0545e-06, -1.3065e-04]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0208, -0.0195, 0.0136, -0.0212, 0.0316, 0.0184, 0.0039, -0.0093, + -0.0163, -0.0058], device='cuda:0'), grad: tensor([ 1.1057e-04, -3.7998e-06, -3.6545e-06, 5.6714e-05, -2.7448e-05, + 2.0042e-05, 5.1446e-06, 4.5508e-05, 5.0277e-05, -2.5296e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 220.72, cls_loss 0.0021 cls_loss_mapping 0.0086 cls_loss_causal 0.5993 re_mapping 0.0077 re_causal 0.0246 /// teacc 99.00 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0887, 0.1156, -0.0160, ..., -0.0626, 0.0177, -0.0031], + [ 0.0018, -0.0562, -0.0328, ..., 0.0239, -0.0144, -0.0306], + [-0.0805, -0.1076, 0.0094, ..., -0.0157, -0.0014, -0.0443], + ..., + [ 0.0476, 0.0569, -0.0541, ..., -0.0544, 0.0460, 0.0453], + [ 0.0400, -0.0438, -0.0121, ..., -0.0136, -0.0221, -0.0309], + [-0.0851, -0.0140, -0.0723, ..., 0.0759, -0.1560, 0.0494]], + device='cuda:0'), grad: tensor([[ 1.2433e-06, -1.6820e-06, -2.1688e-07, ..., 8.9966e-07, + 2.7148e-07, 8.4867e-08], + [ 7.8455e-06, 1.8964e-07, 1.5716e-08, ..., 7.2606e-06, + 6.7335e-07, 1.2703e-06], + [ 9.9093e-07, 3.6438e-07, -5.4133e-08, ..., 1.2890e-06, + 8.9081e-07, 9.2341e-07], + ..., + [ 5.8860e-06, -2.7451e-07, 3.0734e-08, ..., 9.2164e-06, + 3.6508e-07, 2.7772e-06], + [ 9.4995e-06, 2.5029e-07, 1.8044e-08, ..., 1.6570e-05, + 2.5518e-06, 6.1803e-06], + [ 5.0031e-06, 7.5996e-07, 6.9733e-08, ..., -7.2122e-05, + 1.6829e-06, -3.8505e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0205, -0.0199, 0.0144, -0.0216, 0.0316, 0.0189, 0.0041, -0.0099, + -0.0163, -0.0056], device='cuda:0'), grad: tensor([ 4.8615e-06, 1.5989e-05, 6.0983e-06, 2.5496e-05, 3.4779e-05, + -3.8773e-05, 1.2510e-05, 1.9342e-05, 3.8058e-05, -1.1832e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 220.65, cls_loss 0.0021 cls_loss_mapping 0.0072 cls_loss_causal 0.5920 re_mapping 0.0077 re_causal 0.0240 /// teacc 99.06 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0890, 0.1167, -0.0161, ..., -0.0628, 0.0194, -0.0027], + [ 0.0019, -0.0572, -0.0334, ..., 0.0238, -0.0144, -0.0306], + [-0.0806, -0.1087, 0.0109, ..., -0.0159, -0.0012, -0.0448], + ..., + [ 0.0479, 0.0572, -0.0549, ..., -0.0547, 0.0461, 0.0455], + [ 0.0399, -0.0444, -0.0124, ..., -0.0142, -0.0230, -0.0314], + [-0.0855, -0.0143, -0.0724, ..., 0.0763, -0.1567, 0.0497]], + device='cuda:0'), grad: tensor([[ 4.9127e-07, -4.0084e-06, -3.2154e-07, ..., 2.1942e-06, + -1.0151e-07, 8.1584e-07], + [-1.5600e-08, 5.5460e-07, 2.7474e-08, ..., 3.4384e-06, + 6.5239e-07, 2.3302e-06], + [ 1.5004e-06, 2.0247e-06, -8.7311e-07, ..., 1.1902e-06, + -1.4380e-06, 1.0952e-06], + ..., + [-2.0470e-06, -3.4645e-06, 8.9174e-08, ..., 1.3851e-05, + -2.6710e-06, 1.4724e-06], + [ 2.5686e-06, 1.2163e-06, 2.5332e-07, ..., 5.2303e-06, + 6.4867e-07, 1.0477e-06], + [ 1.1072e-05, 1.4305e-06, 4.9826e-08, ..., 3.0380e-06, + 9.5367e-07, -2.1368e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0210, -0.0197, 0.0149, -0.0222, 0.0314, 0.0188, 0.0044, -0.0101, + -0.0167, -0.0056], device='cuda:0'), grad: tensor([ 1.1278e-06, 1.1332e-05, -6.3144e-06, 1.1779e-05, -2.3901e-05, + 1.4439e-05, -1.7276e-06, 2.3663e-05, 9.8124e-06, -4.0233e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 220.65, cls_loss 0.0023 cls_loss_mapping 0.0081 cls_loss_causal 0.6184 re_mapping 0.0077 re_causal 0.0237 /// teacc 99.07 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0893, 0.1170, -0.0161, ..., -0.0633, 0.0193, -0.0027], + [ 0.0020, -0.0582, -0.0340, ..., 0.0241, -0.0151, -0.0309], + [-0.0807, -0.1094, 0.0114, ..., -0.0162, -0.0012, -0.0454], + ..., + [ 0.0481, 0.0590, -0.0555, ..., -0.0556, 0.0464, 0.0457], + [ 0.0398, -0.0449, -0.0126, ..., -0.0147, -0.0235, -0.0318], + [-0.0863, -0.0150, -0.0725, ..., 0.0768, -0.1582, 0.0501]], + device='cuda:0'), grad: tensor([[ 3.7393e-07, -1.9923e-05, -5.1409e-07, ..., -5.3085e-06, + -1.9337e-07, -6.5416e-06], + [ 2.0582e-06, 1.8310e-06, 1.1059e-08, ..., -8.2981e-07, + 7.6517e-06, 6.7800e-06], + [ 9.9763e-06, 1.7891e-06, 6.1234e-08, ..., 2.2314e-06, + 1.0259e-05, 7.2382e-06], + ..., + [-1.4603e-05, -6.6496e-06, 7.2061e-08, ..., 3.9712e-06, + -3.2961e-05, -2.3082e-05], + [-1.3341e-07, 1.4491e-06, 3.2922e-07, ..., 1.5451e-06, + 3.4235e-06, 2.5611e-06], + [ 2.6114e-06, 1.6093e-05, 8.0094e-08, ..., -2.8685e-06, + 9.6634e-06, 3.5912e-06]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0205, -0.0198, 0.0151, -0.0220, 0.0313, 0.0188, 0.0051, -0.0100, + -0.0172, -0.0058], device='cuda:0'), grad: tensor([-2.5481e-05, 1.4789e-05, 3.0756e-05, 2.0906e-05, -3.0696e-05, + -1.1779e-05, 3.6269e-05, -6.6817e-05, 5.2936e-06, 2.6807e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 220.68, cls_loss 0.0024 cls_loss_mapping 0.0078 cls_loss_causal 0.5855 re_mapping 0.0078 re_causal 0.0231 /// teacc 98.96 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0896, 0.1182, -0.0159, ..., -0.0632, 0.0203, -0.0025], + [ 0.0020, -0.0614, -0.0340, ..., 0.0244, -0.0164, -0.0319], + [-0.0811, -0.1107, 0.0117, ..., -0.0167, -0.0005, -0.0464], + ..., + [ 0.0489, 0.0608, -0.0559, ..., -0.0564, 0.0473, 0.0471], + [ 0.0397, -0.0464, -0.0127, ..., -0.0152, -0.0241, -0.0324], + [-0.0870, -0.0153, -0.0728, ..., 0.0773, -0.1595, 0.0504]], + device='cuda:0'), grad: tensor([[ 1.6391e-07, -1.4333e-06, 6.4028e-09, ..., 1.7984e-06, + -1.1362e-07, 1.1409e-06], + [ 7.7672e-07, 8.8941e-08, 1.8626e-09, ..., 1.7822e-05, + 7.3249e-07, 1.4894e-05], + [ 2.2743e-06, 3.4063e-07, -3.6787e-08, ..., 1.9092e-06, + 3.5167e-06, 3.4235e-06], + ..., + [-7.4226e-07, 2.0023e-07, 8.0327e-09, ..., 4.6007e-06, + -2.0191e-06, 2.3358e-06], + [ 5.5879e-08, 2.9826e-07, 2.2119e-09, ..., 3.3285e-06, + 1.1856e-06, 3.9451e-06], + [ 1.0729e-06, 2.0501e-07, 6.9849e-10, ..., -4.3005e-05, + 3.1712e-07, -3.7253e-05]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0212, -0.0202, 0.0159, -0.0221, 0.0313, 0.0184, 0.0047, -0.0098, + -0.0174, -0.0059], device='cuda:0'), grad: tensor([ 3.3602e-06, 3.8564e-05, 1.0140e-05, -1.3644e-06, 1.0498e-05, + 4.6864e-06, 5.9046e-06, 8.4564e-06, 9.1866e-06, -8.9467e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 220.56, cls_loss 0.0022 cls_loss_mapping 0.0078 cls_loss_causal 0.6119 re_mapping 0.0075 re_causal 0.0242 /// teacc 99.05 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0902, 0.1193, -0.0155, ..., -0.0631, 0.0210, -0.0022], + [ 0.0021, -0.0627, -0.0340, ..., 0.0247, -0.0168, -0.0322], + [-0.0819, -0.1121, 0.0118, ..., -0.0171, -0.0009, -0.0476], + ..., + [ 0.0495, 0.0621, -0.0558, ..., -0.0573, 0.0481, 0.0474], + [ 0.0397, -0.0472, -0.0128, ..., -0.0157, -0.0244, -0.0327], + [-0.0881, -0.0156, -0.0730, ..., 0.0774, -0.1604, 0.0513]], + device='cuda:0'), grad: tensor([[ 5.9418e-07, -1.2481e-04, 1.7695e-08, ..., 7.7859e-06, + -3.4928e-05, -3.8594e-05], + [-5.2862e-06, 3.2913e-06, 3.4925e-09, ..., -2.6710e-06, + 1.7853e-06, 2.3898e-06], + [ 2.2259e-06, 1.4611e-05, -5.0873e-08, ..., 5.6587e-06, + 4.6454e-06, 5.9418e-06], + ..., + [-5.8264e-06, 2.6301e-06, 4.1910e-09, ..., 2.5313e-06, + -1.7583e-06, -5.2601e-06], + [ 1.2163e-06, 1.0811e-05, 8.4983e-09, ..., 2.8908e-06, + 4.0382e-06, 5.0180e-06], + [ 3.8408e-06, 3.3021e-05, 8.1491e-10, ..., -7.0594e-07, + 9.8497e-06, 7.4059e-06]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0217, -0.0202, 0.0157, -0.0227, 0.0320, 0.0190, 0.0040, -0.0097, + -0.0178, -0.0059], device='cuda:0'), grad: tensor([-1.4091e-04, -9.4473e-06, 3.4243e-05, 3.8981e-05, 8.2552e-05, + -4.7743e-05, -4.0382e-05, 2.0256e-07, 2.9132e-05, 5.3138e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 220.44, cls_loss 0.0019 cls_loss_mapping 0.0085 cls_loss_causal 0.5781 re_mapping 0.0080 re_causal 0.0238 /// teacc 99.01 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0906, 0.1179, -0.0185, ..., -0.0650, 0.0207, -0.0019], + [ 0.0023, -0.0630, -0.0341, ..., 0.0248, -0.0172, -0.0325], + [-0.0829, -0.1136, 0.0117, ..., -0.0174, -0.0012, -0.0484], + ..., + [ 0.0500, 0.0626, -0.0553, ..., -0.0578, 0.0484, 0.0476], + [ 0.0400, -0.0479, -0.0128, ..., -0.0160, -0.0249, -0.0332], + [-0.0889, -0.0159, -0.0731, ..., 0.0776, -0.1614, 0.0517]], + device='cuda:0'), grad: tensor([[ 1.1325e-06, -1.0058e-06, 2.3888e-07, ..., 1.7546e-06, + -2.4331e-08, 2.1327e-06], + [ 1.8324e-07, 8.8941e-07, 1.1297e-06, ..., -3.1316e-08, + 2.6040e-06, 1.8664e-06], + [-3.2447e-06, 6.5658e-07, -4.4554e-06, ..., 2.1923e-06, + -5.7481e-06, 2.4140e-06], + ..., + [-5.7667e-06, -1.5527e-05, 9.2946e-07, ..., 1.0364e-05, + 2.1625e-06, -5.6028e-06], + [ 1.9129e-06, 6.2119e-07, 4.4378e-07, ..., 4.6156e-06, + 1.6894e-06, 3.6452e-06], + [ 7.4022e-06, 8.6427e-06, 1.5309e-07, ..., -3.4124e-05, + 2.6505e-06, -1.8775e-05]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0199, -0.0200, 0.0147, -0.0229, 0.0321, 0.0193, 0.0055, -0.0095, + -0.0173, -0.0060], device='cuda:0'), grad: tensor([ 7.2792e-06, 1.0490e-05, -3.5733e-05, -3.6024e-06, 1.1921e-05, + 3.4302e-05, 4.4480e-06, 1.6596e-06, 2.2605e-05, -5.3436e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 220.34, cls_loss 0.0024 cls_loss_mapping 0.0080 cls_loss_causal 0.5691 re_mapping 0.0075 re_causal 0.0221 /// teacc 98.99 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0909, 0.1193, -0.0185, ..., -0.0639, 0.0220, -0.0011], + [ 0.0022, -0.0633, -0.0356, ..., 0.0247, -0.0179, -0.0331], + [-0.0836, -0.1146, 0.0135, ..., -0.0176, -0.0009, -0.0494], + ..., + [ 0.0509, 0.0628, -0.0559, ..., -0.0590, 0.0490, 0.0479], + [ 0.0403, -0.0486, -0.0123, ..., -0.0163, -0.0254, -0.0338], + [-0.0889, -0.0165, -0.0725, ..., 0.0792, -0.1626, 0.0532]], + device='cuda:0'), grad: tensor([[ 1.3625e-06, -5.8115e-07, 0.0000e+00, ..., 1.0096e-06, + 1.4564e-07, 6.3702e-07], + [-6.2995e-06, 6.8336e-08, 0.0000e+00, ..., -2.6301e-06, + 6.9477e-07, -2.8424e-06], + [ 1.0096e-06, 1.0943e-07, 0.0000e+00, ..., 7.8930e-07, + 6.4410e-06, 3.3174e-06], + ..., + [ 2.7865e-06, -3.2946e-08, 0.0000e+00, ..., 2.2799e-06, + -1.6054e-07, 5.9791e-07], + [-4.0948e-05, -2.9523e-07, 0.0000e+00, ..., 2.1793e-06, + 1.7174e-06, 2.4885e-06], + [ 1.3402e-06, 1.7288e-07, 0.0000e+00, ..., -1.7315e-05, + 9.7882e-07, -1.2226e-05]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0209, -0.0207, 0.0155, -0.0228, 0.0313, 0.0187, 0.0047, -0.0097, + -0.0170, -0.0049], device='cuda:0'), grad: tensor([ 6.1058e-06, -2.3618e-05, 1.5303e-05, 1.7703e-05, 2.0087e-05, + 2.8670e-05, 2.5675e-05, 1.2547e-05, -7.0453e-05, -3.2008e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 220.44, cls_loss 0.0023 cls_loss_mapping 0.0080 cls_loss_causal 0.5889 re_mapping 0.0076 re_causal 0.0225 /// teacc 98.98 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0914, 0.1202, -0.0186, ..., -0.0640, 0.0221, -0.0009], + [ 0.0021, -0.0641, -0.0356, ..., 0.0247, -0.0181, -0.0335], + [-0.0839, -0.1143, 0.0135, ..., -0.0181, -0.0003, -0.0499], + ..., + [ 0.0515, 0.0636, -0.0563, ..., -0.0597, 0.0487, 0.0480], + [ 0.0411, -0.0490, -0.0124, ..., -0.0164, -0.0257, -0.0341], + [-0.0900, -0.0167, -0.0726, ..., 0.0791, -0.1634, 0.0536]], + device='cuda:0'), grad: tensor([[ 8.2608e-07, -1.2912e-05, 2.7940e-09, ..., 2.2016e-06, + -4.6417e-06, -1.7211e-06], + [ 3.6117e-06, 4.2841e-07, 6.8569e-08, ..., 3.3490e-06, + 1.5497e-06, 2.0005e-06], + [ 1.6680e-06, 7.8510e-07, -9.0338e-08, ..., 2.8629e-06, + -1.4761e-06, 1.9502e-06], + ..., + [ 8.5449e-07, -2.3283e-08, 3.0268e-09, ..., 1.2890e-05, + -5.6485e-07, 5.3309e-06], + [-4.9055e-05, 3.3062e-07, 4.7730e-09, ..., -8.1807e-06, + -2.8824e-07, 5.2415e-06], + [ 1.7853e-06, 6.9141e-06, 1.2806e-09, ..., -1.8668e-04, + 3.7290e-06, -9.5785e-05]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0208, -0.0208, 0.0165, -0.0229, 0.0317, 0.0186, 0.0041, -0.0104, + -0.0160, -0.0052], device='cuda:0'), grad: tensor([-1.8803e-06, 2.1055e-05, -6.4559e-06, 5.2333e-05, 2.9039e-04, + 5.0277e-05, 3.6538e-05, 3.1054e-05, -1.1826e-04, -3.5477e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 220.77, cls_loss 0.0019 cls_loss_mapping 0.0066 cls_loss_causal 0.5607 re_mapping 0.0075 re_causal 0.0230 /// teacc 99.05 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0916, 0.1208, -0.0187, ..., -0.0639, 0.0230, 0.0003], + [ 0.0022, -0.0643, -0.0356, ..., 0.0247, -0.0183, -0.0337], + [-0.0845, -0.1146, 0.0135, ..., -0.0184, -0.0007, -0.0508], + ..., + [ 0.0519, 0.0640, -0.0566, ..., -0.0603, 0.0492, 0.0486], + [ 0.0415, -0.0494, -0.0146, ..., -0.0170, -0.0263, -0.0348], + [-0.0905, -0.0177, -0.0728, ..., 0.0793, -0.1649, 0.0534]], + device='cuda:0'), grad: tensor([[ 3.5297e-06, -2.5749e-05, 1.7462e-09, ..., 4.3167e-07, + -1.0200e-05, -9.6485e-06], + [-1.5395e-06, 2.1532e-06, 8.1491e-10, ..., -4.8615e-06, + 1.2703e-06, 1.4883e-06], + [ 1.1757e-05, 9.7677e-06, -8.8476e-09, ..., 9.0711e-07, + 5.0254e-06, 6.0201e-06], + ..., + [-5.3942e-05, -4.4294e-06, 8.1491e-10, ..., 3.7719e-06, + -8.9109e-06, -3.3025e-06], + [ 2.8238e-06, 2.4978e-06, 2.0955e-09, ..., 2.2538e-06, + 1.7295e-06, 2.3674e-06], + [ 1.8412e-06, 6.5304e-06, 0.0000e+00, ..., -5.6624e-06, + 5.0440e-06, 1.7695e-07]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0208, -0.0206, 0.0157, -0.0232, 0.0317, 0.0186, 0.0033, -0.0097, + -0.0146, -0.0054], device='cuda:0'), grad: tensor([-2.4751e-05, -1.1288e-05, 3.7938e-05, -3.0667e-05, 5.5969e-05, + 2.9206e-05, 1.8943e-06, -7.7844e-05, 1.4976e-05, 4.5411e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 220.43, cls_loss 0.0022 cls_loss_mapping 0.0074 cls_loss_causal 0.5463 re_mapping 0.0071 re_causal 0.0213 /// teacc 98.91 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0920, 0.1219, -0.0186, ..., -0.0641, 0.0240, 0.0003], + [ 0.0025, -0.0647, -0.0364, ..., 0.0251, -0.0188, -0.0346], + [-0.0849, -0.1153, 0.0147, ..., -0.0188, -0.0009, -0.0514], + ..., + [ 0.0523, 0.0642, -0.0572, ..., -0.0611, 0.0497, 0.0490], + [ 0.0414, -0.0499, -0.0149, ..., -0.0177, -0.0269, -0.0355], + [-0.0916, -0.0179, -0.0736, ..., 0.0797, -0.1660, 0.0539]], + device='cuda:0'), grad: tensor([[ 5.0012e-07, -4.1910e-06, 0.0000e+00, ..., 4.7730e-07, + 5.6326e-06, 1.5472e-07], + [ 1.2070e-06, 4.9779e-07, 0.0000e+00, ..., -1.3120e-07, + 4.9695e-06, 3.0361e-06], + [ 8.1211e-06, 5.0198e-07, 0.0000e+00, ..., 1.5181e-06, + -1.4909e-05, 1.2200e-06], + ..., + [-2.1625e-06, 1.9674e-07, 0.0000e+00, ..., 2.7902e-06, + 1.9162e-07, -5.2620e-07], + [-1.9431e-05, 4.9593e-07, 0.0000e+00, ..., 6.0303e-07, + 5.8413e-06, 1.8645e-06], + [ 1.2033e-06, -2.2119e-09, 0.0000e+00, ..., -4.2133e-06, + 1.9874e-06, -2.7344e-06]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0214, -0.0205, 0.0156, -0.0233, 0.0313, 0.0189, 0.0029, -0.0094, + -0.0149, -0.0054], device='cuda:0'), grad: tensor([ 5.4926e-05, 3.2812e-05, -5.6863e-05, 4.0859e-05, 5.4538e-06, + 2.5719e-05, 1.9193e-05, 1.3851e-05, -1.3936e-04, 3.4086e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 220.91, cls_loss 0.0025 cls_loss_mapping 0.0097 cls_loss_causal 0.6108 re_mapping 0.0074 re_causal 0.0238 /// teacc 98.99 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0925, 0.1222, -0.0187, ..., -0.0647, 0.0243, 0.0004], + [ 0.0022, -0.0649, -0.0368, ..., 0.0252, -0.0197, -0.0350], + [-0.0855, -0.1161, 0.0148, ..., -0.0193, -0.0007, -0.0520], + ..., + [ 0.0530, 0.0644, -0.0575, ..., -0.0610, 0.0499, 0.0505], + [ 0.0417, -0.0503, -0.0152, ..., -0.0184, -0.0275, -0.0360], + [-0.0927, -0.0181, -0.0740, ..., 0.0797, -0.1667, 0.0535]], + device='cuda:0'), grad: tensor([[ 3.3248e-07, -6.2678e-07, 1.3970e-08, ..., 4.4005e-07, + 9.2164e-06, -1.5600e-08], + [-3.6228e-06, 1.6938e-07, 2.1188e-08, ..., -4.4368e-06, + 6.5506e-05, 1.0468e-06], + [ 7.9535e-07, 2.0198e-07, -1.5635e-07, ..., 6.0350e-07, + 1.2267e-04, 6.5146e-07], + ..., + [-3.1348e-06, -1.6652e-06, 3.5740e-08, ..., 1.7928e-06, + 2.1160e-06, -7.2941e-06], + [ 3.7309e-06, 1.5111e-07, 1.5367e-08, ..., 1.1861e-05, + 1.6510e-05, 5.6364e-06], + [ 4.4852e-06, 1.4659e-06, 3.7253e-09, ..., -1.1154e-05, + 1.5542e-05, 2.6729e-07]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0209, -0.0210, 0.0156, -0.0227, 0.0317, 0.0181, 0.0041, -0.0085, + -0.0150, -0.0063], device='cuda:0'), grad: tensor([ 3.4571e-05, 2.2388e-04, 4.3797e-04, -9.1743e-04, 1.1019e-05, + 7.3075e-05, 5.9456e-06, 5.1642e-07, 8.3148e-05, 4.6015e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 220.67, cls_loss 0.0020 cls_loss_mapping 0.0070 cls_loss_causal 0.5876 re_mapping 0.0071 re_causal 0.0218 /// teacc 99.00 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0929, 0.1222, -0.0187, ..., -0.0656, 0.0247, 0.0004], + [ 0.0025, -0.0653, -0.0371, ..., 0.0255, -0.0204, -0.0354], + [-0.0862, -0.1172, 0.0151, ..., -0.0197, -0.0011, -0.0529], + ..., + [ 0.0535, 0.0652, -0.0579, ..., -0.0622, 0.0505, 0.0508], + [ 0.0418, -0.0510, -0.0155, ..., -0.0197, -0.0278, -0.0366], + [-0.0936, -0.0182, -0.0742, ..., 0.0802, -0.1677, 0.0543]], + device='cuda:0'), grad: tensor([[ 1.1211e-07, -9.1419e-06, -5.0850e-07, ..., -4.1304e-07, + -1.7658e-06, -1.8906e-06], + [-1.5041e-06, 1.3062e-07, 6.5193e-09, ..., -1.0990e-06, + 1.1101e-06, 1.3812e-06], + [ 3.1432e-07, 9.1596e-07, 3.7020e-08, ..., 4.6054e-07, + -6.3982e-07, 1.9409e-06], + ..., + [ 1.7777e-07, -9.2434e-08, 4.5402e-09, ..., 3.2075e-06, + 1.6959e-06, 4.3884e-06], + [-3.1758e-07, 5.6019e-07, 2.0140e-08, ..., 9.4529e-07, + 6.7055e-06, 4.8950e-06], + [ 4.3726e-07, 5.7444e-06, 3.5227e-07, ..., -5.4352e-06, + 2.1551e-06, -5.5209e-06]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0203, -0.0212, 0.0151, -0.0235, 0.0318, 0.0184, 0.0054, -0.0082, + -0.0152, -0.0063], device='cuda:0'), grad: tensor([-9.5442e-06, -1.0012e-08, -4.0196e-06, -2.1607e-05, 4.5933e-06, + 5.6848e-06, 1.5670e-07, 1.2346e-05, 1.5467e-05, -3.1684e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 220.97, cls_loss 0.0015 cls_loss_mapping 0.0059 cls_loss_causal 0.5443 re_mapping 0.0072 re_causal 0.0217 /// teacc 99.03 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0933, 0.1227, -0.0187, ..., -0.0655, 0.0240, 0.0004], + [ 0.0029, -0.0656, -0.0372, ..., 0.0257, -0.0205, -0.0353], + [-0.0866, -0.1179, 0.0152, ..., -0.0201, -0.0010, -0.0533], + ..., + [ 0.0536, 0.0657, -0.0583, ..., -0.0628, 0.0504, 0.0506], + [ 0.0416, -0.0519, -0.0158, ..., -0.0202, -0.0284, -0.0370], + [-0.0941, -0.0185, -0.0746, ..., 0.0804, -0.1688, 0.0547]], + device='cuda:0'), grad: tensor([[ 8.0327e-07, -9.0152e-06, 7.4506e-08, ..., 3.8147e-06, + -9.9745e-07, 3.7439e-07], + [-4.7758e-06, 2.7493e-06, 2.3749e-08, ..., -8.3074e-06, + 1.2964e-06, 5.9721e-08], + [ 1.5870e-06, 2.9635e-06, -1.0547e-07, ..., 7.4878e-06, + 2.5029e-08, 6.1048e-07], + ..., + [ 1.3085e-06, 3.5809e-07, 8.0792e-08, ..., 1.7891e-06, + -2.9197e-07, -7.0874e-07], + [ 8.2515e-07, 1.4147e-06, 6.4168e-07, ..., 1.2517e-06, + 2.6245e-06, 1.2200e-06], + [ 3.8296e-06, 1.0282e-06, 3.2596e-08, ..., 2.9299e-06, + 1.3942e-06, -1.3388e-08]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0202, -0.0209, 0.0151, -0.0237, 0.0318, 0.0190, 0.0056, -0.0086, + -0.0156, -0.0065], device='cuda:0'), grad: tensor([-2.2911e-07, -1.6704e-05, 2.2665e-05, 6.8210e-06, 9.4548e-06, + -2.8238e-05, -1.9595e-05, 6.4857e-06, 3.6899e-06, 1.5616e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 220.79, cls_loss 0.0015 cls_loss_mapping 0.0061 cls_loss_causal 0.5838 re_mapping 0.0074 re_causal 0.0226 /// teacc 98.97 lr 0.00010000 +Epoch 107, weight, value: tensor([[-9.3715e-02, 1.2358e-01, -1.8597e-02, ..., -6.5391e-02, + 2.4342e-02, 7.5395e-04], + [ 2.5178e-03, -6.6111e-02, -3.7216e-02, ..., 2.5703e-02, + -2.0907e-02, -3.5493e-02], + [-8.7000e-02, -1.1843e-01, 1.6030e-02, ..., -2.0451e-02, + -2.9550e-05, -5.3902e-02], + ..., + [ 5.3872e-02, 6.6164e-02, -5.8684e-02, ..., -6.3178e-02, + 5.0525e-02, 5.0905e-02], + [ 4.1980e-02, -5.3026e-02, -1.6616e-02, ..., -2.1096e-02, + -2.8867e-02, -3.7698e-02], + [-9.4526e-02, -1.8773e-02, -7.5252e-02, ..., 8.1105e-02, + -1.6960e-01, 5.5257e-02]], device='cuda:0'), grad: tensor([[ 4.9546e-07, -1.8105e-05, -6.6357e-08, ..., 1.5134e-06, + -3.0510e-06, 1.2191e-06], + [ 2.4363e-06, 2.0303e-06, 2.0536e-07, ..., 4.3698e-06, + 1.8701e-06, 1.6261e-06], + [ 1.1809e-06, 5.3607e-06, 1.9791e-07, ..., 2.4699e-06, + 2.2218e-05, 1.0081e-05], + ..., + [ 9.5740e-07, 6.1374e-07, 9.0338e-08, ..., 4.8168e-06, + 3.2838e-06, 3.7029e-06], + [ 5.5786e-07, 1.2619e-06, 1.6973e-07, ..., 4.2431e-06, + 1.6820e-06, 4.2431e-06], + [-7.4692e-06, 2.8755e-07, 3.7090e-07, ..., -4.3416e-04, + 4.0233e-06, -2.4164e-04]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0206, -0.0213, 0.0158, -0.0243, 0.0311, 0.0193, 0.0056, -0.0087, + -0.0158, -0.0061], device='cuda:0'), grad: tensor([-1.7479e-05, 1.4067e-05, 4.2319e-05, -3.4362e-05, 6.7759e-04, + -6.0126e-06, 2.1324e-05, 1.7464e-05, 3.5968e-06, -7.1907e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 220.80, cls_loss 0.0022 cls_loss_mapping 0.0064 cls_loss_causal 0.5457 re_mapping 0.0071 re_causal 0.0202 /// teacc 99.04 lr 0.00010000 +Epoch 108, weight, value: tensor([[-9.4231e-02, 1.2426e-01, -1.8423e-02, ..., -6.5438e-02, + 2.4460e-02, 7.9014e-04], + [ 3.1821e-03, -6.5055e-02, -3.8028e-02, ..., 2.5876e-02, + -2.0493e-02, -3.5400e-02], + [-8.7990e-02, -1.1898e-01, 1.6350e-02, ..., -2.1196e-02, + -9.4269e-05, -5.4634e-02], + ..., + [ 5.3773e-02, 6.6105e-02, -5.9064e-02, ..., -6.3832e-02, + 5.0405e-02, 5.0946e-02], + [ 4.2024e-02, -5.3797e-02, -1.6914e-02, ..., -2.1493e-02, + -3.0128e-02, -3.8407e-02], + [-9.5116e-02, -1.9066e-02, -7.5213e-02, ..., 8.1397e-02, + -1.7119e-01, 5.5579e-02]], device='cuda:0'), grad: tensor([[ 4.0559e-07, -3.3509e-06, 3.0501e-08, ..., 8.6892e-07, + -1.0123e-06, -6.4867e-07], + [-2.7474e-06, 6.5193e-08, 1.6531e-08, ..., -4.9882e-06, + -3.3993e-06, 2.9150e-07], + [ 1.3700e-06, 4.8848e-07, 4.0047e-08, ..., 1.2824e-06, + 9.9000e-07, 3.6089e-07], + ..., + [ 3.2764e-06, 1.0571e-07, 9.3132e-10, ..., 2.2352e-06, + 9.2760e-07, 7.4133e-07], + [-1.0477e-07, 3.0058e-07, 2.9337e-08, ..., 3.4310e-06, + 6.4960e-07, 1.6456e-06], + [ 1.6987e-06, 1.5218e-06, 1.3970e-09, ..., -4.1276e-06, + 1.5693e-06, -3.1795e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0206, -0.0203, 0.0151, -0.0230, 0.0311, 0.0180, 0.0065, -0.0093, + -0.0158, -0.0063], device='cuda:0'), grad: tensor([ 1.2387e-06, -4.2349e-05, 9.2685e-06, 1.4216e-05, -5.9307e-06, + 5.2340e-06, 8.0559e-07, 1.1861e-05, 5.5805e-06, 1.9092e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 220.45, cls_loss 0.0016 cls_loss_mapping 0.0060 cls_loss_causal 0.5883 re_mapping 0.0069 re_causal 0.0221 /// teacc 98.90 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0948, 0.1245, -0.0184, ..., -0.0659, 0.0248, 0.0006], + [ 0.0038, -0.0653, -0.0381, ..., 0.0263, -0.0215, -0.0354], + [-0.0885, -0.1196, 0.0164, ..., -0.0216, 0.0005, -0.0551], + ..., + [ 0.0538, 0.0666, -0.0592, ..., -0.0645, 0.0504, 0.0511], + [ 0.0419, -0.0543, -0.0170, ..., -0.0221, -0.0313, -0.0390], + [-0.0960, -0.0191, -0.0748, ..., 0.0817, -0.1724, 0.0558]], + device='cuda:0'), grad: tensor([[ 1.0934e-06, -3.2242e-06, 3.6554e-08, ..., 1.7113e-07, + -1.1828e-07, -4.8801e-07], + [-2.5742e-06, 2.8149e-07, 4.0978e-08, ..., -2.5202e-06, + 2.5867e-07, 3.8580e-07], + [ 4.1276e-06, 1.0217e-06, 6.9849e-08, ..., 3.4925e-06, + 5.1921e-07, 7.8324e-07], + ..., + [ 2.2426e-05, -2.0750e-06, 9.0338e-08, ..., 2.2426e-05, + -1.1232e-06, -1.4366e-07], + [-2.0579e-05, 1.7369e-07, 6.8825e-07, ..., 5.1316e-07, + 2.3693e-06, -3.1814e-06], + [ 2.4781e-05, 1.8775e-06, 8.3819e-08, ..., 1.1519e-05, + 1.3029e-06, 5.5647e-07]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0203, -0.0212, 0.0166, -0.0231, 0.0314, 0.0187, 0.0061, -0.0095, + -0.0164, -0.0065], device='cuda:0'), grad: tensor([ 2.4680e-08, -9.5069e-06, 1.3702e-05, 8.9854e-06, -7.7486e-05, + -9.5248e-05, 9.0480e-05, 6.6102e-05, -6.3837e-05, 6.6817e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 220.47, cls_loss 0.0024 cls_loss_mapping 0.0076 cls_loss_causal 0.5528 re_mapping 0.0071 re_causal 0.0211 /// teacc 98.87 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0957, 0.1252, -0.0185, ..., -0.0658, 0.0253, 0.0005], + [ 0.0040, -0.0671, -0.0385, ..., 0.0256, -0.0212, -0.0354], + [-0.0894, -0.1205, 0.0168, ..., -0.0224, 0.0003, -0.0556], + ..., + [ 0.0534, 0.0677, -0.0595, ..., -0.0663, 0.0504, 0.0513], + [ 0.0424, -0.0550, -0.0155, ..., -0.0229, -0.0318, -0.0395], + [-0.0951, -0.0196, -0.0727, ..., 0.0846, -0.1738, 0.0577]], + device='cuda:0'), grad: tensor([[ 4.4890e-06, -6.1691e-06, 3.5134e-07, ..., 3.7830e-06, + -5.6205e-07, 1.6559e-06], + [ 3.0637e-05, 1.0086e-06, 5.4128e-06, ..., 3.4064e-05, + 2.2594e-06, 3.7868e-06], + [ 4.1127e-06, 2.3469e-06, -3.2899e-07, ..., 3.4031e-06, + 1.1530e-06, 1.9111e-06], + ..., + [-1.3858e-06, -6.7614e-06, 3.8277e-07, ..., 5.9754e-06, + -7.7039e-06, -5.9046e-06], + [ 2.4244e-05, 1.4566e-06, 3.2745e-06, ..., 2.0936e-05, + 5.0515e-06, 7.2382e-06], + [ 2.1577e-05, 1.2405e-06, 9.3365e-08, ..., -7.0455e-07, + 7.8976e-06, -3.2806e-07]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0206, -0.0214, 0.0162, -0.0235, 0.0298, 0.0193, 0.0050, -0.0098, + -0.0156, -0.0042], device='cuda:0'), grad: tensor([ 1.6302e-05, 1.6415e-04, 1.6421e-05, 8.6486e-05, -3.5733e-05, + -1.7357e-04, -2.7776e-04, -5.2787e-06, 1.3328e-04, 7.6056e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 220.94, cls_loss 0.0019 cls_loss_mapping 0.0072 cls_loss_causal 0.5952 re_mapping 0.0068 re_causal 0.0215 /// teacc 99.10 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0962, 0.1264, -0.0186, ..., -0.0657, 0.0262, 0.0006], + [ 0.0045, -0.0679, -0.0387, ..., 0.0255, -0.0209, -0.0350], + [-0.0899, -0.1216, 0.0173, ..., -0.0231, 0.0002, -0.0563], + ..., + [ 0.0532, 0.0683, -0.0597, ..., -0.0672, 0.0506, 0.0512], + [ 0.0422, -0.0560, -0.0148, ..., -0.0243, -0.0325, -0.0402], + [-0.0971, -0.0201, -0.0722, ..., 0.0839, -0.1750, 0.0583]], + device='cuda:0'), grad: tensor([[ 2.6030e-07, 6.0759e-06, 1.5181e-07, ..., 1.1744e-06, + 3.9846e-05, 8.0541e-06], + [-2.3376e-06, 9.2201e-07, 3.2829e-08, ..., -2.4531e-06, + 1.0710e-06, 1.7378e-06], + [ 1.0198e-06, 2.6990e-06, 2.9569e-08, ..., 6.8219e-07, + 4.8503e-06, 2.6580e-06], + ..., + [-2.1476e-06, -2.1324e-05, 1.1642e-09, ..., 1.6876e-06, + -1.0379e-05, -2.6196e-05], + [-3.7067e-07, 3.8021e-07, 1.6321e-07, ..., 1.7332e-06, + 8.0233e-07, 6.7987e-07], + [ 1.9781e-06, 9.5814e-06, 3.7253e-09, ..., -1.0110e-05, + 6.2957e-06, 2.2426e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0212, -0.0211, 0.0159, -0.0238, 0.0311, 0.0192, 0.0050, -0.0102, + -0.0159, -0.0052], device='cuda:0'), grad: tensor([ 8.1658e-05, -8.2105e-06, 1.6108e-05, 2.0730e-04, 1.6823e-05, + -2.7752e-04, 4.0494e-06, -5.2750e-05, 2.3544e-06, 1.0550e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 221.14, cls_loss 0.0016 cls_loss_mapping 0.0061 cls_loss_causal 0.5565 re_mapping 0.0070 re_causal 0.0212 /// teacc 99.06 lr 0.00010000 +Epoch 112, weight, value: tensor([[-9.6783e-02, 1.2696e-01, -1.8516e-02, ..., -6.5924e-02, + 2.6452e-02, 8.2896e-04], + [ 4.4556e-03, -6.8293e-02, -3.9653e-02, ..., 2.5114e-02, + -2.1135e-02, -3.5397e-02], + [-9.0321e-02, -1.2218e-01, 1.7537e-02, ..., -2.3388e-02, + -1.6888e-06, -5.7135e-02], + ..., + [ 5.3522e-02, 6.9029e-02, -6.1251e-02, ..., -6.7670e-02, + 5.1099e-02, 5.1791e-02], + [ 4.2486e-02, -5.6541e-02, -1.4784e-02, ..., -2.4909e-02, + -3.2881e-02, -4.0576e-02], + [-9.6299e-02, -2.0254e-02, -7.0767e-02, ..., 8.5230e-02, + -1.7611e-01, 5.9111e-02]], device='cuda:0'), grad: tensor([[ 1.2945e-07, -9.3058e-06, 0.0000e+00, ..., 1.3746e-06, + -9.3691e-07, -6.6124e-07], + [-7.2923e-07, 7.8557e-07, 0.0000e+00, ..., -1.9604e-07, + 1.7881e-06, 8.9779e-07], + [ 5.2620e-07, 6.7893e-07, 0.0000e+00, ..., 4.3144e-07, + -3.1292e-05, -2.9877e-06], + ..., + [-1.4743e-06, -6.4261e-06, 0.0000e+00, ..., 4.4075e-07, + 9.0674e-06, -4.3474e-06], + [-3.1595e-07, 4.1025e-07, 0.0000e+00, ..., 1.1045e-06, + 1.3620e-05, 1.8068e-06], + [ 1.3243e-06, 5.3756e-06, 0.0000e+00, ..., -1.4035e-06, + 1.0319e-06, 2.7809e-06]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0213, -0.0212, 0.0156, -0.0238, 0.0299, 0.0193, 0.0047, -0.0098, + -0.0159, -0.0040], device='cuda:0'), grad: tensor([-4.2319e-06, 4.1425e-06, -9.4116e-05, 1.4193e-05, 4.3325e-06, + 4.9472e-06, 1.1250e-06, 1.9297e-05, 4.1366e-05, 8.9481e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 220.75, cls_loss 0.0014 cls_loss_mapping 0.0063 cls_loss_causal 0.5776 re_mapping 0.0071 re_causal 0.0213 /// teacc 98.98 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0973, 0.1276, -0.0187, ..., -0.0662, 0.0271, 0.0010], + [ 0.0043, -0.0691, -0.0399, ..., 0.0248, -0.0214, -0.0357], + [-0.0907, -0.1233, 0.0176, ..., -0.0240, -0.0002, -0.0576], + ..., + [ 0.0537, 0.0698, -0.0618, ..., -0.0683, 0.0513, 0.0520], + [ 0.0423, -0.0578, -0.0150, ..., -0.0256, -0.0338, -0.0412], + [-0.0967, -0.0206, -0.0701, ..., 0.0853, -0.1772, 0.0592]], + device='cuda:0'), grad: tensor([[ 4.7078e-07, -2.1197e-06, -3.0501e-08, ..., 9.0618e-07, + -5.0012e-07, 2.2352e-08], + [-1.7649e-07, 1.3504e-07, 2.7870e-07, ..., 1.5749e-06, + 4.7009e-07, 4.3362e-06], + [-2.6971e-06, 4.0070e-07, -1.0617e-05, ..., 8.9779e-07, + -7.4953e-06, 3.8254e-07], + ..., + [ 3.3025e-06, -2.7171e-07, 9.7007e-06, ..., 4.2729e-06, + 6.9998e-06, 5.5172e-06], + [-1.6559e-06, -3.4459e-08, 1.7369e-07, ..., 1.8720e-06, + -6.1467e-07, 1.7937e-06], + [ 2.8238e-06, 1.3355e-06, 1.7020e-07, ..., -5.9158e-06, + 1.1809e-06, -1.6510e-05]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0214, -0.0215, 0.0154, -0.0245, 0.0301, 0.0202, 0.0051, -0.0097, + -0.0165, -0.0042], device='cuda:0'), grad: tensor([ 2.0918e-06, 6.8210e-06, -2.3305e-05, 1.3495e-06, -3.1106e-06, + 9.4399e-06, -5.2266e-06, 3.8803e-05, -5.3868e-06, -2.1547e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 220.65, cls_loss 0.0019 cls_loss_mapping 0.0071 cls_loss_causal 0.6030 re_mapping 0.0068 re_causal 0.0208 /// teacc 99.05 lr 0.00010000 +Epoch 114, weight, value: tensor([[-9.8338e-02, 1.2833e-01, -1.8725e-02, ..., -6.6373e-02, + 2.7028e-02, 1.0439e-03], + [ 4.9620e-03, -6.8916e-02, -4.0215e-02, ..., 2.5429e-02, + -2.0912e-02, -3.5232e-02], + [-9.1127e-02, -1.2470e-01, 1.8010e-02, ..., -2.4485e-02, + -5.5518e-05, -5.8218e-02], + ..., + [ 5.2495e-02, 7.1102e-02, -6.3368e-02, ..., -6.9819e-02, + 5.0847e-02, 5.2038e-02], + [ 4.2890e-02, -5.8494e-02, -1.5217e-02, ..., -2.4575e-02, + -3.4673e-02, -4.1255e-02], + [-9.7580e-02, -2.1138e-02, -7.1239e-02, ..., 8.4884e-02, + -1.7924e-01, 5.9122e-02]], device='cuda:0'), grad: tensor([[ 1.1064e-06, -1.1465e-06, 1.2596e-07, ..., 5.2936e-06, + 3.0617e-07, 1.4091e-06], + [ 1.6727e-06, 6.9477e-07, 1.8626e-08, ..., 2.7684e-07, + 9.3598e-07, 2.4084e-06], + [ 7.3835e-06, 2.6431e-06, 1.7695e-08, ..., 1.4305e-05, + 3.2615e-06, 7.3686e-06], + ..., + [-3.4302e-05, -8.7619e-06, 2.6543e-08, ..., 9.3970e-07, + -9.3281e-06, -2.7984e-05], + [ 1.9111e-06, 4.6799e-07, 6.7288e-08, ..., 9.1270e-06, + 1.5665e-06, 2.2445e-06], + [ 2.0847e-05, 5.0776e-06, -3.7625e-07, ..., -4.3958e-07, + 6.1207e-06, 1.4998e-05]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0211, -0.0210, 0.0153, -0.0248, 0.0308, 0.0203, 0.0052, -0.0104, + -0.0154, -0.0050], device='cuda:0'), grad: tensor([ 1.9759e-05, 1.0438e-05, 7.7963e-05, -8.1658e-06, 2.0611e-04, + 6.0081e-05, -3.4618e-04, -1.4353e-04, 3.6061e-05, 8.6963e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 220.49, cls_loss 0.0015 cls_loss_mapping 0.0056 cls_loss_causal 0.5785 re_mapping 0.0069 re_causal 0.0208 /// teacc 99.05 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0991, 0.1290, -0.0187, ..., -0.0670, 0.0277, 0.0014], + [ 0.0052, -0.0691, -0.0403, ..., 0.0258, -0.0216, -0.0358], + [-0.0915, -0.1252, 0.0180, ..., -0.0250, 0.0005, -0.0587], + ..., + [ 0.0527, 0.0718, -0.0652, ..., -0.0701, 0.0507, 0.0523], + [ 0.0429, -0.0591, -0.0154, ..., -0.0249, -0.0355, -0.0419], + [-0.0980, -0.0214, -0.0719, ..., 0.0850, -0.1804, 0.0593]], + device='cuda:0'), grad: tensor([[ 3.4906e-06, -2.9877e-05, 1.1409e-08, ..., 2.3656e-06, + -7.7859e-06, -1.2763e-05], + [-1.8880e-05, -2.4084e-06, 1.0245e-08, ..., -1.7956e-05, + 4.2543e-06, 3.7737e-06], + [ 2.9728e-06, 1.2591e-06, 2.7474e-08, ..., 2.2668e-06, + 1.7546e-06, 1.5851e-06], + ..., + [ 1.7881e-07, 1.0775e-06, 1.6298e-09, ..., 5.2378e-06, + -3.6862e-06, -2.8778e-06], + [ 1.3951e-06, 9.0748e-06, 2.7940e-08, ..., 3.3025e-06, + 5.5097e-06, 6.3516e-06], + [ 3.4850e-06, 5.2676e-06, 5.5879e-09, ..., 3.3733e-06, + 2.0154e-06, -6.6776e-07]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0210, -0.0209, 0.0156, -0.0247, 0.0306, 0.0203, 0.0058, -0.0106, + -0.0156, -0.0051], device='cuda:0'), grad: tensor([-3.3438e-05, -8.6725e-05, 1.7524e-05, -1.1228e-05, -1.4812e-05, + 1.7688e-05, 4.9651e-05, 1.4618e-05, 2.6360e-05, 2.0251e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 220.60, cls_loss 0.0018 cls_loss_mapping 0.0061 cls_loss_causal 0.5819 re_mapping 0.0066 re_causal 0.0203 /// teacc 98.99 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0997, 0.1302, -0.0186, ..., -0.0676, 0.0285, 0.0013], + [ 0.0055, -0.0694, -0.0403, ..., 0.0256, -0.0220, -0.0360], + [-0.0918, -0.1266, 0.0184, ..., -0.0255, 0.0010, -0.0594], + ..., + [ 0.0534, 0.0720, -0.0649, ..., -0.0706, 0.0509, 0.0528], + [ 0.0430, -0.0602, -0.0155, ..., -0.0256, -0.0362, -0.0427], + [-0.0986, -0.0212, -0.0716, ..., 0.0852, -0.1815, 0.0598]], + device='cuda:0'), grad: tensor([[ 2.8349e-06, -8.2850e-06, 2.6776e-08, ..., 8.8438e-06, + -2.5984e-06, 2.6859e-06], + [ 2.0266e-06, 1.0328e-06, 1.6298e-08, ..., 1.2182e-05, + 2.1365e-06, 5.4240e-06], + [ 2.8387e-06, 2.2557e-06, 4.3074e-08, ..., 6.5789e-06, + -3.5819e-06, 3.8967e-06], + ..., + [ 1.7956e-05, -6.0014e-06, 1.1642e-09, ..., 1.1516e-04, + -1.2191e-06, 3.9965e-05], + [-1.3687e-05, 3.9348e-07, 3.1898e-08, ..., 5.0664e-06, + 2.0582e-06, 3.4273e-06], + [-1.6224e-04, 3.4329e-06, 6.7521e-09, ..., -1.0757e-03, + 5.0962e-06, -4.0984e-04]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0211, -0.0210, 0.0162, -0.0244, 0.0305, 0.0194, 0.0062, -0.0107, + -0.0158, -0.0050], device='cuda:0'), grad: tensor([ 3.6240e-05, 4.0531e-05, -3.4389e-07, 5.9038e-05, 1.9760e-03, + 5.3704e-05, 2.9892e-05, 2.5558e-04, -2.2602e-04, -2.2240e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 220.37, cls_loss 0.0022 cls_loss_mapping 0.0080 cls_loss_causal 0.5687 re_mapping 0.0063 re_causal 0.0194 /// teacc 99.06 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.1004, 0.1311, -0.0187, ..., -0.0676, 0.0290, 0.0016], + [ 0.0058, -0.0698, -0.0402, ..., 0.0257, -0.0222, -0.0361], + [-0.0927, -0.1292, 0.0181, ..., -0.0264, 0.0008, -0.0609], + ..., + [ 0.0538, 0.0741, -0.0652, ..., -0.0717, 0.0518, 0.0535], + [ 0.0431, -0.0607, -0.0156, ..., -0.0260, -0.0368, -0.0432], + [-0.0989, -0.0219, -0.0707, ..., 0.0850, -0.1829, 0.0603]], + device='cuda:0'), grad: tensor([[ 6.6832e-06, -3.1642e-07, 0.0000e+00, ..., 5.1744e-06, + 2.5565e-07, 7.3481e-07], + [-7.4923e-05, 8.0792e-08, 0.0000e+00, ..., -5.9247e-05, + -9.9242e-06, -1.3433e-05], + [ 6.5379e-06, 1.3784e-07, 0.0000e+00, ..., 3.6750e-06, + 2.0005e-06, 3.3136e-06], + ..., + [ 6.5863e-06, -3.8394e-07, 0.0000e+00, ..., 4.6715e-06, + 2.6040e-06, 2.3376e-06], + [ 2.9579e-05, 4.6333e-08, 0.0000e+00, ..., 2.7880e-05, + 1.3486e-06, 1.7621e-06], + [ 9.8273e-06, 1.8184e-07, 0.0000e+00, ..., 3.1628e-06, + 3.8296e-06, 2.9579e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0213, -0.0208, 0.0158, -0.0249, 0.0315, 0.0198, 0.0056, -0.0103, + -0.0159, -0.0059], device='cuda:0'), grad: tensor([ 3.9577e-05, -4.4847e-04, 3.3587e-05, 3.9190e-05, 3.8952e-05, + -1.1347e-05, 3.9816e-05, 4.6223e-05, 1.7035e-04, 5.2392e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 220.71, cls_loss 0.0014 cls_loss_mapping 0.0051 cls_loss_causal 0.5392 re_mapping 0.0066 re_causal 0.0198 /// teacc 99.05 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.1011, 0.1315, -0.0187, ..., -0.0681, 0.0290, 0.0015], + [ 0.0057, -0.0702, -0.0403, ..., 0.0257, -0.0224, -0.0365], + [-0.0930, -0.1296, 0.0181, ..., -0.0269, 0.0008, -0.0618], + ..., + [ 0.0542, 0.0753, -0.0648, ..., -0.0725, 0.0520, 0.0540], + [ 0.0428, -0.0616, -0.0156, ..., -0.0265, -0.0376, -0.0442], + [-0.0993, -0.0221, -0.0708, ..., 0.0851, -0.1840, 0.0607]], + device='cuda:0'), grad: tensor([[ 5.7044e-08, -2.7916e-07, 4.3306e-08, ..., 1.9781e-06, + 2.6077e-08, -1.0128e-07], + [-4.0513e-07, 6.5193e-08, 1.3039e-08, ..., -2.0489e-08, + -9.5461e-09, 8.7311e-08], + [ 2.8103e-07, 9.4064e-08, 1.8161e-08, ..., 2.5961e-07, + 5.3924e-07, 3.1851e-07], + ..., + [-7.7998e-08, -3.8580e-07, 3.0268e-09, ..., 1.6927e-07, + -6.9151e-08, -2.2002e-07], + [-3.6089e-08, 8.4285e-08, 6.2166e-08, ..., 5.1036e-07, + 3.2433e-07, 2.5146e-07], + [ 9.6392e-08, 3.0315e-07, 1.9791e-08, ..., -1.1530e-06, + 1.1595e-07, -5.4995e-07]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0210, -0.0208, 0.0160, -0.0249, 0.0317, 0.0203, 0.0055, -0.0102, + -0.0163, -0.0061], device='cuda:0'), grad: tensor([ 3.9153e-06, -3.1013e-06, 3.2075e-06, -1.5739e-06, 8.8010e-07, + 4.3064e-06, -8.5756e-06, 4.6869e-07, 1.6512e-06, -1.2117e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 220.63, cls_loss 0.0016 cls_loss_mapping 0.0059 cls_loss_causal 0.5700 re_mapping 0.0064 re_causal 0.0199 /// teacc 98.98 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.1016, 0.1319, -0.0187, ..., -0.0689, 0.0291, 0.0012], + [ 0.0061, -0.0710, -0.0403, ..., 0.0261, -0.0225, -0.0363], + [-0.0932, -0.1304, 0.0185, ..., -0.0273, 0.0010, -0.0625], + ..., + [ 0.0544, 0.0782, -0.0649, ..., -0.0735, 0.0522, 0.0549], + [ 0.0432, -0.0629, -0.0156, ..., -0.0269, -0.0379, -0.0443], + [-0.0998, -0.0228, -0.0706, ..., 0.0853, -0.1848, 0.0607]], + device='cuda:0'), grad: tensor([[ 2.7963e-07, -1.6913e-06, 7.4506e-09, ..., 8.8941e-07, + -3.7043e-07, -2.1467e-07], + [-4.1336e-05, 6.4308e-07, 1.2643e-07, ..., -1.5050e-05, + 5.0571e-07, 8.3912e-07], + [ 6.9570e-07, 3.2759e-07, 1.3737e-08, ..., 1.3998e-06, + 2.6566e-07, 2.1514e-07], + ..., + [ 1.8608e-06, -8.5682e-07, 2.3283e-10, ..., 1.5181e-06, + -5.0617e-07, -9.8720e-07], + [ 1.3202e-05, 2.1979e-07, 2.7241e-08, ..., 1.1057e-05, + 1.8976e-07, 1.8207e-07], + [ 1.1362e-06, 2.8871e-07, 1.3970e-09, ..., 3.0361e-07, + 2.1607e-07, -3.4296e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0203, -0.0206, 0.0160, -0.0249, 0.0317, 0.0195, 0.0061, -0.0099, + -0.0160, -0.0061], device='cuda:0'), grad: tensor([ 1.6093e-06, -3.4451e-05, 7.0371e-06, -6.5193e-08, 4.5687e-05, + 9.0450e-06, -6.9082e-05, 2.3525e-06, 3.4899e-05, 3.0436e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 220.76, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5576 re_mapping 0.0062 re_causal 0.0191 /// teacc 98.99 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.1015, 0.1330, -0.0187, ..., -0.0693, 0.0292, 0.0027], + [ 0.0057, -0.0717, -0.0404, ..., 0.0260, -0.0229, -0.0372], + [-0.0935, -0.1311, 0.0190, ..., -0.0278, 0.0009, -0.0633], + ..., + [ 0.0550, 0.0795, -0.0651, ..., -0.0738, 0.0523, 0.0556], + [ 0.0435, -0.0635, -0.0157, ..., -0.0274, -0.0390, -0.0449], + [-0.1004, -0.0231, -0.0695, ..., 0.0853, -0.1855, 0.0611]], + device='cuda:0'), grad: tensor([[ 6.0489e-07, -1.2550e-07, 5.4855e-07, ..., 5.4389e-07, + 7.0315e-07, 2.2189e-07], + [-1.4156e-06, -4.0559e-07, 7.7765e-08, ..., -6.7335e-07, + 1.8775e-06, 1.8096e-06], + [ 5.3877e-07, 2.3516e-08, 8.8941e-08, ..., 4.6706e-07, + 2.1122e-06, 3.3081e-06], + ..., + [ 5.0385e-07, -1.0245e-07, 3.3760e-08, ..., 9.4809e-07, + 4.7423e-06, 4.3362e-06], + [ 2.0601e-06, 1.8068e-07, 2.7120e-06, ..., 2.1830e-06, + 8.1360e-06, 8.3670e-06], + [ 8.6203e-06, 2.1583e-07, 1.1316e-07, ..., 1.7807e-05, + 5.0897e-07, 6.1607e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0215, -0.0214, 0.0160, -0.0250, 0.0319, 0.0185, 0.0061, -0.0095, + -0.0153, -0.0064], device='cuda:0'), grad: tensor([ 6.5789e-06, 9.5833e-07, 1.4111e-05, -9.1732e-05, -3.2306e-05, + 4.2230e-05, -5.3227e-05, 2.6494e-05, 5.2750e-05, 3.4124e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 220.75, cls_loss 0.0019 cls_loss_mapping 0.0062 cls_loss_causal 0.5633 re_mapping 0.0063 re_causal 0.0195 /// teacc 99.02 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.1020, 0.1339, -0.0188, ..., -0.0699, 0.0291, 0.0032], + [ 0.0051, -0.0721, -0.0404, ..., 0.0251, -0.0233, -0.0381], + [-0.0939, -0.1317, 0.0190, ..., -0.0282, 0.0014, -0.0640], + ..., + [ 0.0549, 0.0775, -0.0653, ..., -0.0766, 0.0525, 0.0545], + [ 0.0435, -0.0642, -0.0159, ..., -0.0279, -0.0399, -0.0459], + [-0.1006, -0.0216, -0.0678, ..., 0.0858, -0.1862, 0.0631]], + device='cuda:0'), grad: tensor([[ 3.5879e-07, -7.2569e-06, 3.2596e-08, ..., 2.2221e-06, + -1.2363e-07, -3.6862e-06], + [ 3.4133e-07, 5.4482e-07, 1.2084e-07, ..., 1.5544e-06, + 9.3551e-07, 1.3020e-06], + [ 8.6240e-07, 5.4063e-07, 5.3318e-08, ..., 1.3560e-06, + 1.5395e-06, 1.7937e-06], + ..., + [ 1.2107e-06, -4.1910e-09, 2.0396e-07, ..., 4.0792e-06, + 4.8801e-07, 1.0394e-06], + [-1.0524e-07, 7.0967e-07, 5.4948e-08, ..., 3.6154e-06, + 3.6880e-06, 4.0941e-06], + [ 7.5661e-06, 4.1090e-06, 8.4238e-07, ..., -1.9640e-05, + 8.5775e-07, -8.0317e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0217, -0.0217, 0.0159, -0.0272, 0.0318, 0.0204, 0.0070, -0.0107, + -0.0158, -0.0057], device='cuda:0'), grad: tensor([-5.0897e-07, 5.8226e-06, 1.0043e-05, -1.9729e-05, -1.9506e-05, + 9.3132e-06, 2.2724e-06, 1.1832e-05, 1.1221e-05, -1.0602e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 220.74, cls_loss 0.0017 cls_loss_mapping 0.0057 cls_loss_causal 0.5452 re_mapping 0.0063 re_causal 0.0192 /// teacc 99.00 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.1022, 0.1336, -0.0187, ..., -0.0715, 0.0296, 0.0014], + [ 0.0043, -0.0725, -0.0405, ..., 0.0233, -0.0236, -0.0383], + [-0.0943, -0.1336, 0.0191, ..., -0.0286, 0.0014, -0.0651], + ..., + [ 0.0550, 0.0782, -0.0654, ..., -0.0772, 0.0528, 0.0552], + [ 0.0432, -0.0650, -0.0160, ..., -0.0288, -0.0404, -0.0468], + [-0.1015, -0.0201, -0.0677, ..., 0.0866, -0.1874, 0.0642]], + device='cuda:0'), grad: tensor([[ 6.2725e-07, -7.4565e-05, 2.3283e-09, ..., 5.2787e-06, + -3.3438e-05, -4.0270e-06], + [ 1.3877e-07, 4.4852e-06, 2.3982e-08, ..., 1.4771e-06, + 2.0470e-06, 2.3060e-06], + [ 2.2333e-06, 2.8521e-05, 7.8930e-08, ..., 2.6710e-06, + 1.3687e-05, 7.2867e-06], + ..., + [-1.9874e-06, -3.4086e-06, 2.0955e-09, ..., 8.1118e-07, + -4.7148e-07, -8.0392e-06], + [ 2.2352e-06, -1.8049e-06, 2.3283e-10, ..., -7.5549e-06, + 5.0887e-06, -5.4874e-06], + [ 4.4778e-06, 1.4864e-05, 1.5600e-08, ..., 8.4937e-06, + 9.8124e-06, 8.2701e-06]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0205, -0.0225, 0.0160, -0.0274, 0.0318, 0.0205, 0.0073, -0.0104, + -0.0165, -0.0049], device='cuda:0'), grad: tensor([-9.0778e-05, 4.9733e-06, 6.0558e-05, 2.9849e-07, -2.3857e-05, + 1.2338e-05, 1.1615e-05, -1.1772e-05, -2.2516e-05, 5.9158e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 220.46, cls_loss 0.0022 cls_loss_mapping 0.0072 cls_loss_causal 0.5797 re_mapping 0.0064 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.1026, 0.1347, -0.0185, ..., -0.0710, 0.0301, 0.0022], + [ 0.0045, -0.0731, -0.0405, ..., 0.0229, -0.0238, -0.0385], + [-0.0954, -0.1344, 0.0192, ..., -0.0293, 0.0015, -0.0676], + ..., + [ 0.0553, 0.0794, -0.0655, ..., -0.0777, 0.0535, 0.0565], + [ 0.0447, -0.0645, -0.0161, ..., -0.0272, -0.0409, -0.0457], + [-0.1026, -0.0207, -0.0661, ..., 0.0864, -0.1886, 0.0640]], + device='cuda:0'), grad: tensor([[ 9.5833e-07, 1.2596e-07, 0.0000e+00, ..., 5.4855e-07, + 3.0268e-07, 5.4948e-07], + [-2.7418e-06, 4.7521e-07, 0.0000e+00, ..., 1.2806e-07, + 4.5192e-07, -1.9632e-06], + [ 1.6289e-06, 7.4413e-07, 0.0000e+00, ..., 5.2247e-07, + 5.9698e-07, 1.0263e-06], + ..., + [-2.6263e-06, -4.2841e-06, 0.0000e+00, ..., 2.0489e-07, + -2.2370e-06, -3.2652e-06], + [-1.4238e-05, 1.3364e-07, 0.0000e+00, ..., -1.2122e-05, + 9.7230e-07, 9.7230e-07], + [ 4.8801e-06, 1.2387e-06, 0.0000e+00, ..., 1.0263e-06, + 4.7982e-06, 8.0988e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0210, -0.0228, 0.0157, -0.0272, 0.0330, 0.0195, 0.0051, -0.0099, + -0.0138, -0.0057], device='cuda:0'), grad: tensor([ 7.2606e-06, -1.4968e-05, -1.0058e-06, 2.2829e-05, 8.6576e-06, + -3.3498e-05, 3.6806e-05, 2.8103e-07, -7.7903e-05, 5.1320e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 220.35, cls_loss 0.0016 cls_loss_mapping 0.0044 cls_loss_causal 0.5348 re_mapping 0.0064 re_causal 0.0194 /// teacc 99.01 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.1033, 0.1355, -0.0185, ..., -0.0707, 0.0298, 0.0025], + [ 0.0022, -0.0735, -0.0432, ..., 0.0211, -0.0241, -0.0389], + [-0.0927, -0.1341, 0.0216, ..., -0.0279, 0.0015, -0.0684], + ..., + [ 0.0555, 0.0802, -0.0655, ..., -0.0781, 0.0535, 0.0571], + [ 0.0444, -0.0657, -0.0163, ..., -0.0276, -0.0415, -0.0461], + [-0.1034, -0.0210, -0.0665, ..., 0.0865, -0.1895, 0.0641]], + device='cuda:0'), grad: tensor([[ 3.7486e-07, -1.0520e-05, 4.6333e-08, ..., 8.5915e-08, + 1.2964e-06, -8.7731e-07], + [ 2.3586e-07, 1.3914e-06, 3.1223e-07, ..., -8.6054e-07, + 1.3048e-06, 8.6008e-07], + [ 1.2051e-06, 3.8520e-06, 3.8417e-08, ..., 4.5658e-07, + 2.3648e-05, 3.8333e-06], + ..., + [-3.1712e-07, -9.8422e-06, 1.4738e-07, ..., 8.7777e-08, + -8.9332e-06, -1.6183e-05], + [-4.9062e-06, 1.7043e-06, -1.5143e-06, ..., 1.8487e-07, + 2.6878e-06, 1.2275e-06], + [ 9.6112e-07, 6.5044e-06, 9.0804e-09, ..., -3.1432e-08, + 3.3285e-06, 3.5930e-06]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0210, -0.0253, 0.0187, -0.0270, 0.0332, 0.0196, 0.0049, -0.0097, + -0.0142, -0.0061], device='cuda:0'), grad: tensor([-7.5065e-06, 5.0403e-06, 5.9634e-05, -5.5641e-05, 7.3686e-06, + 2.1420e-06, 2.9057e-06, -3.0488e-05, 4.0070e-07, 1.6108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 220.91, cls_loss 0.0019 cls_loss_mapping 0.0063 cls_loss_causal 0.5523 re_mapping 0.0062 re_causal 0.0181 /// teacc 99.01 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.1042, 0.1365, -0.0199, ..., -0.0707, 0.0300, 0.0028], + [ 0.0028, -0.0746, -0.0428, ..., 0.0217, -0.0233, -0.0381], + [-0.0929, -0.1364, 0.0216, ..., -0.0283, 0.0017, -0.0720], + ..., + [ 0.0557, 0.0816, -0.0659, ..., -0.0788, 0.0536, 0.0586], + [ 0.0444, -0.0651, -0.0168, ..., -0.0289, -0.0423, -0.0475], + [-0.1043, -0.0214, -0.0668, ..., 0.0866, -0.1907, 0.0639]], + device='cuda:0'), grad: tensor([[ 1.0105e-07, -6.4727e-07, 1.9325e-08, ..., 3.6787e-06, + 7.9162e-08, 1.2037e-07], + [ 3.0664e-07, 3.5111e-07, 3.4925e-08, ..., -1.6531e-08, + 8.5402e-07, 7.3668e-07], + [ 3.4906e-06, 1.2852e-07, -5.1688e-08, ..., 6.5565e-07, + 6.5975e-06, 4.4629e-06], + ..., + [ 9.7044e-07, -7.3668e-07, 4.2422e-07, ..., 1.2182e-06, + 4.7218e-07, 4.0070e-07], + [-4.0345e-06, 6.7754e-08, 3.4226e-08, ..., 3.4589e-06, + 1.0449e-06, 1.5059e-06], + [ 4.2445e-07, -3.0501e-08, -2.5146e-07, ..., -5.1521e-06, + 4.6310e-07, -2.8200e-06]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0211, -0.0252, 0.0183, -0.0272, 0.0332, 0.0191, 0.0056, -0.0084, + -0.0143, -0.0064], device='cuda:0'), grad: tensor([ 1.7002e-05, 3.2727e-06, 3.2812e-05, -2.9713e-05, 1.5693e-06, + 6.6869e-06, -2.9117e-05, 6.7540e-06, 2.1029e-06, -1.1399e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 220.45, cls_loss 0.0021 cls_loss_mapping 0.0078 cls_loss_causal 0.5791 re_mapping 0.0062 re_causal 0.0188 /// teacc 99.02 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.1049, 0.1372, -0.0198, ..., -0.0712, 0.0302, 0.0027], + [ 0.0027, -0.0751, -0.0428, ..., 0.0217, -0.0236, -0.0388], + [-0.0932, -0.1375, 0.0216, ..., -0.0287, 0.0012, -0.0731], + ..., + [ 0.0571, 0.0823, -0.0662, ..., -0.0794, 0.0541, 0.0595], + [ 0.0442, -0.0664, -0.0171, ..., -0.0299, -0.0441, -0.0486], + [-0.1053, -0.0215, -0.0669, ..., 0.0870, -0.1920, 0.0646]], + device='cuda:0'), grad: tensor([[ 4.9826e-07, -9.1270e-07, 1.3062e-07, ..., 4.2468e-07, + -2.9942e-07, 2.6589e-07], + [ 3.0398e-06, 1.1292e-07, 9.1270e-08, ..., -2.1923e-06, + 5.2080e-06, 7.7859e-06], + [ 1.5972e-06, 1.3900e-07, 2.7288e-07, ..., 1.3867e-06, + 1.2284e-06, 1.6307e-06], + ..., + [-1.0118e-05, -9.2015e-07, 4.6799e-08, ..., 6.7288e-07, + -1.1183e-05, -1.8284e-05], + [-1.6103e-06, 2.8405e-08, -1.3243e-06, ..., 2.6636e-07, + 1.6233e-06, 1.2014e-06], + [ 3.0957e-06, 9.7789e-07, 2.7078e-07, ..., 1.4873e-06, + 1.9046e-06, 3.1646e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0211, -0.0255, 0.0179, -0.0251, 0.0329, 0.0174, 0.0060, -0.0070, + -0.0154, -0.0064], device='cuda:0'), grad: tensor([ 3.8184e-06, 2.0519e-05, 1.4059e-05, 1.0505e-05, 4.4033e-06, + 7.9349e-06, -6.4410e-06, -5.4121e-05, -2.5198e-05, 2.4512e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 220.33, cls_loss 0.0016 cls_loss_mapping 0.0058 cls_loss_causal 0.5259 re_mapping 0.0060 re_causal 0.0184 /// teacc 99.02 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.1056, 0.1381, -0.0199, ..., -0.0713, 0.0308, 0.0031], + [ 0.0029, -0.0755, -0.0428, ..., 0.0231, -0.0238, -0.0391], + [-0.0935, -0.1392, 0.0215, ..., -0.0292, 0.0009, -0.0740], + ..., + [ 0.0576, 0.0824, -0.0662, ..., -0.0803, 0.0543, 0.0600], + [ 0.0445, -0.0673, -0.0173, ..., -0.0302, -0.0445, -0.0491], + [-0.1059, -0.0218, -0.0663, ..., 0.0869, -0.1927, 0.0646]], + device='cuda:0'), grad: tensor([[ 1.5390e-07, -1.9714e-05, 1.4203e-08, ..., 3.0035e-07, + -3.6340e-06, -1.0446e-05], + [-2.7530e-06, 3.7067e-07, 1.7462e-09, ..., -5.3458e-06, + 7.4622e-08, -7.3621e-07], + [ 4.5146e-07, 1.2275e-06, -2.2119e-09, ..., 8.1351e-07, + 1.9232e-07, 8.7544e-07], + ..., + [-3.6834e-07, -1.1027e-06, 8.1491e-10, ..., 8.2143e-07, + -1.5064e-07, -7.3295e-07], + [ 2.6915e-07, 3.7951e-07, 1.4785e-08, ..., 6.4913e-07, + 1.0571e-07, 2.4773e-07], + [ 9.5367e-07, 1.5706e-05, 1.9791e-09, ..., 4.6976e-06, + 2.6021e-06, 6.6608e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0215, -0.0252, 0.0176, -0.0251, 0.0327, 0.0179, 0.0059, -0.0068, + -0.0154, -0.0068], device='cuda:0'), grad: tensor([-2.0415e-05, -2.3499e-05, 4.0345e-06, 8.2552e-06, 9.5293e-06, + 5.2005e-06, -2.1964e-05, -6.9197e-07, 3.4291e-06, 3.6091e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 220.49, cls_loss 0.0012 cls_loss_mapping 0.0046 cls_loss_causal 0.5561 re_mapping 0.0059 re_causal 0.0190 /// teacc 99.06 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.1064, 0.1384, -0.0201, ..., -0.0716, 0.0305, 0.0028], + [ 0.0029, -0.0758, -0.0428, ..., 0.0232, -0.0239, -0.0393], + [-0.0939, -0.1404, 0.0215, ..., -0.0297, 0.0008, -0.0745], + ..., + [ 0.0580, 0.0830, -0.0662, ..., -0.0808, 0.0545, 0.0604], + [ 0.0452, -0.0677, -0.0158, ..., -0.0304, -0.0449, -0.0495], + [-0.1063, -0.0218, -0.0662, ..., 0.0870, -0.1933, 0.0648]], + device='cuda:0'), grad: tensor([[ 6.7009e-07, -1.4587e-07, 5.8208e-10, ..., 5.2620e-07, + 7.1013e-09, 4.2957e-08], + [-5.3160e-06, 8.6147e-09, 4.6566e-10, ..., -3.1237e-06, + 1.3097e-07, 1.2130e-07], + [ 3.1143e-06, 2.0489e-08, -1.0827e-08, ..., 7.2457e-07, + -1.6263e-07, 1.5832e-07], + ..., + [ 2.1551e-06, -4.1793e-08, 2.5611e-09, ..., 1.3243e-06, + 2.2282e-07, 4.1630e-07], + [-8.6948e-06, 1.5483e-08, 4.6566e-10, ..., -9.5833e-07, + -1.5879e-07, 1.3504e-07], + [ 1.2182e-06, 4.3306e-08, 2.3283e-10, ..., -8.2515e-07, + 1.8440e-07, -8.5775e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0212, -0.0251, 0.0172, -0.0252, 0.0327, 0.0182, 0.0050, -0.0066, + -0.0141, -0.0069], device='cuda:0'), grad: tensor([ 4.3623e-06, -1.3947e-05, 8.5905e-06, 9.1344e-06, 1.8016e-05, + 1.5914e-05, -1.2577e-05, 1.1899e-05, -4.4793e-05, 3.4273e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 220.35, cls_loss 0.0012 cls_loss_mapping 0.0047 cls_loss_causal 0.5526 re_mapping 0.0060 re_causal 0.0181 /// teacc 98.95 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.1070, 0.1398, -0.0201, ..., -0.0719, 0.0309, 0.0033], + [ 0.0032, -0.0760, -0.0428, ..., 0.0235, -0.0240, -0.0393], + [-0.0939, -0.1413, 0.0215, ..., -0.0301, 0.0008, -0.0749], + ..., + [ 0.0579, 0.0829, -0.0663, ..., -0.0815, 0.0544, 0.0603], + [ 0.0454, -0.0687, -0.0148, ..., -0.0314, -0.0457, -0.0502], + [-0.1067, -0.0225, -0.0663, ..., 0.0873, -0.1939, 0.0653]], + device='cuda:0'), grad: tensor([[ 5.3179e-07, 1.8033e-07, 0.0000e+00, ..., 2.6450e-06, + -3.9698e-08, -7.7998e-08], + [ 4.1783e-05, 1.0128e-07, 0.0000e+00, ..., 1.9523e-07, + 2.6217e-07, 1.9942e-07], + [ 2.4457e-06, 2.4377e-07, 0.0000e+00, ..., 9.8720e-07, + 2.4750e-07, 2.3155e-07], + ..., + [ 2.5071e-06, -2.5425e-07, 0.0000e+00, ..., 9.9465e-07, + 6.3097e-07, -4.0862e-08], + [-4.8757e-05, 2.8615e-07, 0.0000e+00, ..., 3.6834e-07, + 2.1746e-07, 1.3725e-07], + [ 2.3413e-06, 5.8999e-07, 0.0000e+00, ..., 2.5295e-06, + 6.5239e-07, 5.6392e-07]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0217, -0.0249, 0.0171, -0.0256, 0.0328, 0.0182, 0.0044, -0.0070, + -0.0135, -0.0067], device='cuda:0'), grad: tensor([ 7.8902e-06, 9.9421e-05, 6.4522e-06, -3.4422e-06, -7.0445e-06, + 1.8356e-06, -8.0839e-06, 8.0243e-06, -1.1456e-04, 9.3356e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 220.36, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5311 re_mapping 0.0059 re_causal 0.0178 /// teacc 99.11 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.1074, 0.1406, -0.0202, ..., -0.0721, 0.0293, 0.0034], + [ 0.0036, -0.0763, -0.0428, ..., 0.0237, -0.0242, -0.0394], + [-0.0942, -0.1417, 0.0215, ..., -0.0303, 0.0009, -0.0753], + ..., + [ 0.0581, 0.0834, -0.0663, ..., -0.0820, 0.0546, 0.0607], + [ 0.0451, -0.0679, -0.0150, ..., -0.0317, -0.0466, -0.0503], + [-0.1071, -0.0232, -0.0658, ..., 0.0874, -0.1955, 0.0652]], + device='cuda:0'), grad: tensor([[ 5.9791e-07, -2.2620e-05, 0.0000e+00, ..., 1.0319e-06, + -8.5980e-06, -4.7944e-06], + [ 2.7986e-07, 1.0757e-06, 0.0000e+00, ..., 2.6217e-07, + 3.9791e-07, 1.0151e-06], + [ 2.9244e-07, 6.4634e-07, 0.0000e+00, ..., 3.9721e-07, + -3.1013e-07, 5.6531e-07], + ..., + [-1.0163e-05, -2.2173e-05, 0.0000e+00, ..., 2.2613e-06, + 8.6334e-07, -1.8463e-05], + [ 1.0002e-06, 2.1365e-06, 0.0000e+00, ..., 1.6950e-06, + 6.5099e-07, 2.5555e-06], + [ 7.2047e-06, 1.6049e-05, 0.0000e+00, ..., -5.7593e-06, + 6.2631e-07, 1.0341e-05]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0214, -0.0246, 0.0172, -0.0261, 0.0327, 0.0189, 0.0048, -0.0070, + -0.0137, -0.0071], device='cuda:0'), grad: tensor([-3.1799e-05, 4.2096e-06, -3.8296e-06, 0.0000e+00, 1.4707e-05, + 2.6494e-05, -2.0023e-06, -4.7714e-05, 1.1139e-05, 2.8744e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 220.67, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5101 re_mapping 0.0061 re_causal 0.0184 /// teacc 99.01 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.1076, 0.1411, -0.0202, ..., -0.0722, 0.0294, 0.0034], + [ 0.0034, -0.0767, -0.0428, ..., 0.0235, -0.0244, -0.0399], + [-0.0943, -0.1422, 0.0215, ..., -0.0306, 0.0007, -0.0757], + ..., + [ 0.0585, 0.0838, -0.0664, ..., -0.0824, 0.0547, 0.0609], + [ 0.0448, -0.0682, -0.0150, ..., -0.0319, -0.0475, -0.0510], + [-0.1075, -0.0234, -0.0652, ..., 0.0875, -0.1965, 0.0659]], + device='cuda:0'), grad: tensor([[ 4.3749e-07, 0.0000e+00, 2.3283e-10, ..., 8.0466e-06, + 1.9133e-05, 7.4767e-06], + [-1.5035e-05, 0.0000e+00, 1.1642e-10, ..., -4.9099e-06, + 5.5134e-06, 1.7416e-06], + [ 4.6045e-06, 0.0000e+00, -1.6298e-09, ..., 1.4147e-06, + -2.7362e-06, 1.7304e-06], + ..., + [ 4.4890e-06, 0.0000e+00, 3.4925e-10, ..., 3.5688e-06, + 6.0275e-06, 6.0396e-07], + [ 1.5078e-06, 0.0000e+00, 3.4925e-10, ..., 6.7912e-06, + 2.3752e-05, 7.5959e-06], + [ 3.2671e-06, 0.0000e+00, 0.0000e+00, ..., 2.4997e-06, + 4.0941e-06, 5.5786e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0215, -0.0249, 0.0170, -0.0255, 0.0329, 0.0192, 0.0040, -0.0070, + -0.0137, -0.0070], device='cuda:0'), grad: tensor([ 8.8036e-05, -2.3797e-05, -8.5458e-06, -1.9464e-03, 7.5661e-06, + 1.7090e-03, 8.0988e-06, 4.1395e-05, 1.0002e-04, 2.5496e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 220.22, cls_loss 0.0014 cls_loss_mapping 0.0057 cls_loss_causal 0.5552 re_mapping 0.0060 re_causal 0.0179 /// teacc 98.93 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.1093, 0.1414, -0.0202, ..., -0.0734, 0.0291, 0.0031], + [ 0.0045, -0.0769, -0.0427, ..., 0.0235, -0.0246, -0.0397], + [-0.0944, -0.1432, 0.0216, ..., -0.0313, 0.0008, -0.0762], + ..., + [ 0.0575, 0.0837, -0.0665, ..., -0.0830, 0.0548, 0.0607], + [ 0.0447, -0.0688, -0.0150, ..., -0.0327, -0.0481, -0.0517], + [-0.1082, -0.0228, -0.0652, ..., 0.0873, -0.1966, 0.0669]], + device='cuda:0'), grad: tensor([[ 1.0594e-07, -3.9116e-07, 1.9325e-08, ..., 2.8778e-07, + 2.0722e-08, 1.7113e-08], + [-1.0617e-06, 7.5670e-08, 1.3970e-08, ..., -7.5484e-07, + 2.7241e-07, 5.0897e-07], + [ 2.9430e-07, 6.1700e-08, -2.5425e-07, ..., 4.8475e-07, + -2.3399e-08, 2.2142e-07], + ..., + [ 4.6683e-08, -1.4144e-07, 1.3737e-08, ..., 2.0005e-06, + -3.4343e-08, 2.1718e-06], + [ 2.7753e-06, 4.4005e-08, 1.4482e-07, ..., 1.1519e-05, + 2.7940e-06, 6.7465e-06], + [-1.1716e-06, 1.0384e-07, 5.2387e-09, ..., -1.5676e-05, + 4.4005e-07, -9.6783e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0203, -0.0242, 0.0170, -0.0255, 0.0334, 0.0194, 0.0043, -0.0080, + -0.0142, -0.0070], device='cuda:0'), grad: tensor([ 1.0068e-06, -2.2128e-06, -6.6832e-06, -4.6611e-05, 4.2506e-06, + 3.8683e-05, 9.6485e-07, 6.1318e-06, 4.0710e-05, -3.6329e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 220.37, cls_loss 0.0015 cls_loss_mapping 0.0057 cls_loss_causal 0.5402 re_mapping 0.0060 re_causal 0.0181 /// teacc 99.07 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.1099, 0.1421, -0.0203, ..., -0.0736, 0.0291, 0.0031], + [ 0.0042, -0.0777, -0.0429, ..., 0.0237, -0.0248, -0.0409], + [-0.0949, -0.1446, 0.0216, ..., -0.0319, 0.0012, -0.0767], + ..., + [ 0.0583, 0.0841, -0.0665, ..., -0.0842, 0.0551, 0.0612], + [ 0.0456, -0.0698, -0.0154, ..., -0.0333, -0.0487, -0.0527], + [-0.1088, -0.0229, -0.0653, ..., 0.0875, -0.1973, 0.0679]], + device='cuda:0'), grad: tensor([[ 6.8825e-07, -5.5544e-06, 2.3283e-10, ..., 1.5348e-05, + -6.7404e-08, -3.5111e-07], + [ 5.9605e-06, 4.5635e-06, 1.1642e-10, ..., 7.0110e-06, + 4.3809e-06, 7.6182e-06], + [ 4.1611e-06, 1.3728e-06, 1.1642e-10, ..., 3.5893e-06, + 3.5092e-06, 3.3062e-06], + ..., + [-1.2487e-05, -1.1757e-05, 0.0000e+00, ..., 1.6754e-06, + -2.0400e-05, -2.8446e-05], + [ 6.6496e-07, 9.3970e-07, 2.3283e-10, ..., 2.9579e-06, + 1.2387e-06, 8.8336e-07], + [ 1.8686e-05, 7.4804e-06, 0.0000e+00, ..., 2.5451e-05, + 5.4166e-06, 8.5309e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0204, -0.0243, 0.0164, -0.0257, 0.0335, 0.0191, 0.0042, -0.0075, + -0.0136, -0.0068], device='cuda:0'), grad: tensor([ 3.9876e-05, 3.9756e-05, 2.0266e-05, 2.7955e-05, -5.4568e-05, + 1.4462e-05, -7.5936e-05, -9.9480e-05, 9.5442e-06, 7.8022e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 220.26, cls_loss 0.0015 cls_loss_mapping 0.0051 cls_loss_causal 0.5369 re_mapping 0.0061 re_causal 0.0181 /// teacc 98.98 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.1104, 0.1427, -0.0203, ..., -0.0739, 0.0292, 0.0032], + [ 0.0045, -0.0779, -0.0429, ..., 0.0237, -0.0251, -0.0411], + [-0.0950, -0.1452, 0.0216, ..., -0.0324, 0.0012, -0.0771], + ..., + [ 0.0567, 0.0846, -0.0665, ..., -0.0849, 0.0537, 0.0595], + [ 0.0483, -0.0702, -0.0154, ..., -0.0338, -0.0466, -0.0501], + [-0.1090, -0.0231, -0.0651, ..., 0.0863, -0.1980, 0.0682]], + device='cuda:0'), grad: tensor([[ 1.0128e-07, -3.9265e-06, 0.0000e+00, ..., 5.1130e-07, + -1.3802e-06, -1.8859e-08], + [-2.5798e-07, 1.6368e-07, 0.0000e+00, ..., 3.7765e-07, + 1.3942e-06, 1.1101e-06], + [ 2.7148e-07, 1.5218e-06, 0.0000e+00, ..., 2.4983e-07, + 1.5809e-07, 6.8732e-07], + ..., + [-1.9907e-07, -2.6939e-07, 0.0000e+00, ..., 2.0843e-06, + 1.9260e-06, 5.0627e-06], + [-1.7951e-07, 4.8708e-07, 0.0000e+00, ..., 1.4305e-06, + 1.7490e-06, 2.3432e-06], + [ 8.7917e-07, 7.2690e-07, 0.0000e+00, ..., -6.3069e-06, + 3.5716e-07, -9.1866e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0204, -0.0239, 0.0160, -0.0257, 0.0349, 0.0190, 0.0042, -0.0096, + -0.0112, -0.0082], device='cuda:0'), grad: tensor([-4.8913e-06, 2.7075e-05, -2.8551e-05, -2.5585e-05, 1.8319e-06, + 1.1683e-05, 3.1423e-06, 2.3380e-05, 4.2766e-06, -1.2480e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 220.62, cls_loss 0.0017 cls_loss_mapping 0.0065 cls_loss_causal 0.5752 re_mapping 0.0058 re_causal 0.0180 /// teacc 99.03 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.1109, 0.1432, -0.0204, ..., -0.0742, 0.0295, 0.0031], + [ 0.0051, -0.0782, -0.0429, ..., 0.0236, -0.0249, -0.0411], + [-0.0953, -0.1463, 0.0216, ..., -0.0328, 0.0009, -0.0780], + ..., + [ 0.0565, 0.0857, -0.0665, ..., -0.0869, 0.0547, 0.0597], + [ 0.0478, -0.0713, -0.0155, ..., -0.0348, -0.0475, -0.0508], + [-0.1091, -0.0231, -0.0651, ..., 0.0871, -0.2004, 0.0695]], + device='cuda:0'), grad: tensor([[ 1.4976e-06, -1.6779e-05, 1.2806e-08, ..., 9.2387e-07, + 4.1188e-07, 3.4506e-07], + [-2.5526e-05, 6.7800e-07, 1.4203e-08, ..., -9.1419e-06, + 5.0990e-07, -9.0804e-08], + [ 7.0967e-06, -9.0944e-07, -1.9022e-07, ..., 3.5670e-06, + 1.8012e-06, 1.1055e-06], + ..., + [ 7.8380e-06, -3.6159e-07, 2.0955e-09, ..., 3.1181e-06, + 4.4750e-07, 4.7497e-08], + [ 1.7714e-06, 8.1956e-07, 3.4692e-08, ..., 5.5656e-06, + 5.1269e-07, 4.2957e-07], + [ 2.6729e-06, 8.9733e-07, 1.8626e-09, ..., 1.0813e-06, + 5.8301e-07, 5.3365e-07]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0203, -0.0230, 0.0159, -0.0262, 0.0346, 0.0197, 0.0037, -0.0100, + -0.0120, -0.0075], device='cuda:0'), grad: tensor([-7.7933e-06, -7.5340e-05, -2.5928e-05, -1.6868e-05, 1.6391e-04, + 6.6608e-06, -1.3125e-04, 3.9816e-05, 3.2127e-05, 1.4357e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 220.52, cls_loss 0.0014 cls_loss_mapping 0.0043 cls_loss_causal 0.5540 re_mapping 0.0058 re_causal 0.0174 /// teacc 99.10 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.1113, 0.1436, -0.0207, ..., -0.0751, 0.0295, 0.0029], + [ 0.0053, -0.0785, -0.0429, ..., 0.0237, -0.0250, -0.0419], + [-0.0955, -0.1471, 0.0216, ..., -0.0331, 0.0006, -0.0788], + ..., + [ 0.0566, 0.0862, -0.0665, ..., -0.0878, 0.0547, 0.0600], + [ 0.0476, -0.0720, -0.0155, ..., -0.0355, -0.0482, -0.0510], + [-0.1095, -0.0232, -0.0638, ..., 0.0876, -0.2016, 0.0703]], + device='cuda:0'), grad: tensor([[ 5.4250e-07, -1.0252e-05, 0.0000e+00, ..., 5.3179e-07, + -1.2759e-07, 1.9302e-07], + [ 5.9307e-06, 4.0489e-07, 0.0000e+00, ..., 3.1851e-07, + 4.2003e-07, 4.2655e-07], + [ 1.9348e-04, 2.8908e-06, 0.0000e+00, ..., 2.9127e-07, + 1.3206e-06, 5.5972e-07], + ..., + [-1.1153e-07, 9.9884e-08, 0.0000e+00, ..., 8.5123e-07, + -1.8831e-06, -1.8217e-06], + [-2.2161e-04, 8.3027e-07, 0.0000e+00, ..., 1.7099e-06, + 1.9614e-06, 1.3700e-06], + [ 3.7672e-07, 2.1607e-06, 0.0000e+00, ..., -1.0014e-05, + 2.9639e-07, -5.4277e-06]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0199, -0.0230, 0.0158, -0.0248, 0.0343, 0.0188, 0.0043, -0.0101, + -0.0126, -0.0072], device='cuda:0'), grad: tensor([-1.5944e-05, 1.2882e-05, 3.0875e-04, -2.3134e-06, 2.0415e-05, + 2.3961e-05, 8.8215e-06, 3.1176e-07, -3.3903e-04, -1.7405e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 220.53, cls_loss 0.0013 cls_loss_mapping 0.0048 cls_loss_causal 0.5511 re_mapping 0.0055 re_causal 0.0175 /// teacc 99.06 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.1117, 0.1444, -0.0208, ..., -0.0757, 0.0300, 0.0029], + [ 0.0065, -0.0788, -0.0429, ..., 0.0251, -0.0252, -0.0408], + [-0.0960, -0.1482, 0.0216, ..., -0.0336, 0.0003, -0.0796], + ..., + [ 0.0558, 0.0865, -0.0665, ..., -0.0896, 0.0549, 0.0595], + [ 0.0478, -0.0725, -0.0156, ..., -0.0362, -0.0484, -0.0512], + [-0.1104, -0.0234, -0.0628, ..., 0.0877, -0.2022, 0.0709]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -8.3353e-08, 0.0000e+00, ..., 5.2340e-07, + 2.0489e-08, 1.5413e-07], + [ 7.8464e-08, 6.0536e-09, 0.0000e+00, ..., 3.4156e-07, + 2.2841e-07, 3.7858e-07], + [ 4.4471e-07, 2.5844e-08, 0.0000e+00, ..., 1.2219e-06, + 4.7241e-07, 1.3039e-06], + ..., + [-5.5181e-07, -3.6322e-08, 0.0000e+00, ..., 1.1921e-07, + -1.2694e-06, -2.0601e-06], + [ 5.7509e-08, 2.1886e-08, 0.0000e+00, ..., 1.6382e-06, + 3.4855e-07, 4.7963e-07], + [ 1.7835e-07, 7.1945e-08, 0.0000e+00, ..., -5.8673e-06, + 9.8255e-08, -2.9393e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0200, -0.0221, 0.0153, -0.0247, 0.0343, 0.0184, 0.0051, -0.0106, + -0.0127, -0.0072], device='cuda:0'), grad: tensor([ 1.5534e-06, 2.5611e-06, 4.3847e-06, 6.6161e-06, 2.5099e-07, + 4.1053e-06, -1.2636e-05, -4.2841e-06, 1.0535e-05, -1.3113e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 220.36, cls_loss 0.0017 cls_loss_mapping 0.0049 cls_loss_causal 0.5559 re_mapping 0.0062 re_causal 0.0184 /// teacc 98.87 lr 0.00010000 +Epoch 138, weight, value: tensor([[-1.1200e-01, 1.4502e-01, -2.0914e-02, ..., -7.6035e-02, + 3.0057e-02, 2.8392e-03], + [ 6.4162e-03, -7.8966e-02, -4.3028e-02, ..., 2.5370e-02, + -2.5310e-02, -4.0940e-02], + [-9.5724e-02, -1.4927e-01, 2.1733e-02, ..., -3.3942e-02, + 7.8999e-05, -8.0287e-02], + ..., + [ 5.5779e-02, 8.6508e-02, -6.6598e-02, ..., -9.0418e-02, + 5.4424e-02, 5.9070e-02], + [ 4.7582e-02, -7.3131e-02, -1.5641e-02, ..., -3.7032e-02, + -4.8818e-02, -5.1496e-02], + [-1.1080e-01, -2.3532e-02, -6.2216e-02, ..., 8.8183e-02, + -2.0258e-01, 7.1984e-02]], device='cuda:0'), grad: tensor([[ 8.7824e-07, -2.7753e-06, 8.9407e-08, ..., -1.9348e-07, + 8.6147e-09, -1.0827e-07], + [ 1.6438e-06, 1.2713e-07, 4.6799e-07, ..., -1.3458e-07, + 6.2678e-07, 1.2591e-06], + [ 1.1194e-06, 7.4785e-07, -1.6317e-06, ..., 4.7032e-08, + 3.9185e-07, 1.0375e-06], + ..., + [-4.4331e-07, -1.2759e-07, 6.2864e-09, ..., 1.2363e-07, + -9.0944e-07, -2.8014e-06], + [ 2.6803e-06, 2.8685e-07, 2.2585e-08, ..., 7.4506e-08, + 9.7556e-08, 1.4203e-07], + [ 7.7672e-07, 7.5996e-07, 1.0943e-08, ..., -9.2201e-08, + 6.5425e-08, 7.1246e-08]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0201, -0.0236, 0.0174, -0.0256, 0.0341, 0.0189, 0.0054, -0.0114, + -0.0128, -0.0068], device='cuda:0'), grad: tensor([ 1.5777e-06, 1.3173e-05, 9.5274e-07, 4.4028e-07, 1.3113e-06, + -8.4877e-05, 4.5747e-05, -5.1335e-06, 2.1026e-05, 5.6922e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 220.26, cls_loss 0.0019 cls_loss_mapping 0.0075 cls_loss_causal 0.5772 re_mapping 0.0058 re_causal 0.0177 /// teacc 99.05 lr 0.00010000 +Epoch 139, weight, value: tensor([[-1.1231e-01, 1.4571e-01, -2.1060e-02, ..., -7.6400e-02, + 3.0040e-02, 2.8314e-03], + [ 6.4603e-03, -7.9176e-02, -4.3148e-02, ..., 2.5405e-02, + -2.5346e-02, -4.1241e-02], + [-9.5789e-02, -1.5016e-01, 2.1835e-02, ..., -3.5122e-02, + -9.1360e-05, -8.1008e-02], + ..., + [ 5.6407e-02, 8.7159e-02, -6.6612e-02, ..., -9.1657e-02, + 5.4724e-02, 5.9709e-02], + [ 4.6026e-02, -7.3698e-02, -1.5670e-02, ..., -3.8075e-02, + -4.8984e-02, -5.1849e-02], + [-1.1158e-01, -2.4067e-02, -6.2085e-02, ..., 8.8306e-02, + -2.0378e-01, 7.2323e-02]], device='cuda:0'), grad: tensor([[ 1.0268e-07, -2.5798e-07, 0.0000e+00, ..., 4.0885e-07, + 1.3039e-08, 1.0966e-07], + [-4.9919e-07, 1.9954e-07, 0.0000e+00, ..., 1.5972e-07, + 5.0291e-08, -6.9849e-09], + [ 6.9886e-06, 2.2817e-08, 0.0000e+00, ..., 1.1967e-07, + -2.5309e-07, 4.6864e-06], + ..., + [-8.2105e-06, -7.8557e-07, 0.0000e+00, ..., 2.0005e-06, + 6.4494e-08, -4.7721e-06], + [ 5.2433e-07, 3.4459e-08, 0.0000e+00, ..., 4.8392e-06, + 2.8405e-08, 4.4424e-07], + [ 5.0943e-07, 5.5972e-07, 0.0000e+00, ..., -5.1484e-06, + 1.8394e-08, -2.8126e-06]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0201, -0.0239, 0.0176, -0.0264, 0.0341, 0.0210, 0.0049, -0.0110, + -0.0140, -0.0070], device='cuda:0'), grad: tensor([ 1.7919e-06, -7.2345e-06, 2.4512e-05, 2.2501e-06, 8.4639e-06, + 2.3935e-06, -3.2902e-05, -1.8179e-05, 3.2544e-05, -1.3635e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 220.60, cls_loss 0.0012 cls_loss_mapping 0.0039 cls_loss_causal 0.5055 re_mapping 0.0059 re_causal 0.0169 /// teacc 99.06 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1127, 0.1466, -0.0211, ..., -0.0769, 0.0300, 0.0028], + [ 0.0066, -0.0795, -0.0432, ..., 0.0253, -0.0258, -0.0412], + [-0.0962, -0.1517, 0.0218, ..., -0.0360, -0.0003, -0.0815], + ..., + [ 0.0566, 0.0876, -0.0666, ..., -0.0922, 0.0551, 0.0599], + [ 0.0459, -0.0748, -0.0157, ..., -0.0391, -0.0492, -0.0521], + [-0.1121, -0.0245, -0.0615, ..., 0.0885, -0.2044, 0.0727]], + device='cuda:0'), grad: tensor([[ 4.9593e-08, -6.5658e-08, 2.8405e-08, ..., 2.3912e-07, + 1.4435e-08, 1.7695e-08], + [-1.8650e-07, 6.3796e-08, 3.0268e-08, ..., -1.2852e-07, + 1.6997e-08, 6.9384e-08], + [ 1.9418e-07, 5.7509e-08, 9.5228e-08, ..., 5.9931e-07, + 2.3050e-08, 6.2631e-08], + ..., + [-1.0896e-07, -3.2014e-07, 9.3132e-10, ..., 3.5577e-07, + -3.2363e-08, -1.5320e-07], + [-8.2888e-08, 1.6531e-08, 5.5879e-09, ..., 1.0803e-07, + 4.4005e-08, 2.9802e-08], + [ 3.8790e-07, 1.4063e-07, 1.1642e-09, ..., -3.5297e-07, + 5.0990e-08, -4.5192e-07]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0204, -0.0237, 0.0174, -0.0263, 0.0340, 0.0203, 0.0060, -0.0110, + -0.0145, -0.0069], device='cuda:0'), grad: tensor([ 1.0449e-06, 4.6799e-08, 4.1514e-07, 1.5087e-06, 1.0747e-06, + -5.6252e-07, -4.6752e-06, 6.4075e-07, 3.8301e-07, 9.7090e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 139---------------------------------------------------- +epoch 139, time 221.49, cls_loss 0.0012 cls_loss_mapping 0.0042 cls_loss_causal 0.5312 re_mapping 0.0058 re_causal 0.0172 /// teacc 99.18 lr 0.00010000 +Epoch 141, weight, value: tensor([[-1.1309e-01, 1.4734e-01, -2.1516e-02, ..., -7.7231e-02, + 2.9374e-02, 2.9644e-03], + [ 6.7303e-03, -8.0517e-02, -4.3395e-02, ..., 2.5133e-02, + -2.6092e-02, -4.1443e-02], + [-9.6459e-02, -1.5269e-01, 2.2012e-02, ..., -3.6246e-02, + -2.0173e-04, -8.1947e-02], + ..., + [ 5.6776e-02, 8.8541e-02, -6.6595e-02, ..., -9.2721e-02, + 5.5296e-02, 6.0273e-02], + [ 4.5529e-02, -7.5276e-02, -1.5705e-02, ..., -3.9834e-02, + -4.9406e-02, -5.2312e-02], + [-1.1276e-01, -2.4922e-02, -6.1620e-02, ..., 8.8474e-02, + -2.0507e-01, 7.2641e-02]], device='cuda:0'), grad: tensor([[ 1.6578e-07, -6.7661e-07, 5.5879e-09, ..., 3.1013e-07, + -8.3353e-08, 2.6077e-08], + [-2.3544e-06, 1.7835e-07, 1.9558e-08, ..., -2.4047e-06, + 5.5879e-08, 1.8161e-07], + [ 1.4361e-06, 1.7788e-07, 2.8871e-08, ..., 1.4286e-06, + 4.6566e-08, 8.5216e-08], + ..., + [ 1.3085e-07, -3.7672e-07, 1.3970e-09, ..., 7.0361e-07, + -1.3132e-07, -2.2585e-07], + [-1.8524e-06, 6.4261e-08, 4.6566e-10, ..., 3.1274e-06, + 3.5856e-08, 1.0598e-06], + [ 5.9605e-07, 1.3970e-07, 8.5216e-08, ..., -6.5602e-06, + 4.1444e-08, -3.1386e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0202, -0.0239, 0.0173, -0.0263, 0.0343, 0.0201, 0.0067, -0.0108, + -0.0146, -0.0072], device='cuda:0'), grad: tensor([ 4.4378e-07, -1.1660e-05, 8.8513e-06, 1.9342e-05, 1.6876e-06, + 1.3851e-05, 3.2857e-06, 2.2873e-06, -2.6345e-05, -1.1727e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 220.65, cls_loss 0.0013 cls_loss_mapping 0.0042 cls_loss_causal 0.5280 re_mapping 0.0056 re_causal 0.0169 /// teacc 99.06 lr 0.00010000 +Epoch 142, weight, value: tensor([[-1.1344e-01, 1.4859e-01, -2.1881e-02, ..., -7.7339e-02, + 2.9855e-02, 3.1426e-03], + [ 6.7061e-03, -8.2356e-02, -4.3544e-02, ..., 2.4862e-02, + -2.6215e-02, -4.2262e-02], + [-9.6682e-02, -1.5177e-01, 2.2230e-02, ..., -3.6772e-02, + -1.6289e-04, -8.2255e-02], + ..., + [ 5.7100e-02, 8.9886e-02, -6.6671e-02, ..., -9.3984e-02, + 5.5337e-02, 6.0171e-02], + [ 4.5753e-02, -7.6188e-02, -1.5780e-02, ..., -4.0380e-02, + -4.9739e-02, -5.2543e-02], + [-1.1344e-01, -2.5733e-02, -6.1893e-02, ..., 8.8781e-02, + -2.0564e-01, 7.3483e-02]], device='cuda:0'), grad: tensor([[ 4.0978e-07, -4.6134e-05, 6.5193e-09, ..., -3.9667e-05, + 1.1642e-08, 1.2852e-07], + [-8.6874e-06, 1.0300e-06, 6.0536e-09, ..., -5.5954e-06, + 8.8941e-08, 7.3621e-07], + [ 3.6508e-06, 4.0755e-06, 1.8626e-09, ..., 5.9009e-06, + 6.8452e-08, 3.3760e-07], + ..., + [ 2.2817e-08, -8.9733e-07, 4.6566e-10, ..., 2.8796e-06, + 3.0268e-08, 7.5344e-07], + [ 1.4426e-06, 2.2687e-06, 2.7940e-09, ..., 5.8934e-06, + 5.9139e-08, 1.8422e-06], + [ 1.2759e-06, 2.0117e-05, 9.3132e-10, ..., 9.4026e-06, + 5.8673e-08, -7.9572e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0205, -0.0241, 0.0177, -0.0263, 0.0344, 0.0199, 0.0065, -0.0109, + -0.0147, -0.0071], device='cuda:0'), grad: tensor([-1.8561e-04, -3.1054e-05, 2.7820e-05, 2.8446e-05, 1.8224e-05, + 9.7156e-06, 4.2349e-05, 9.4175e-06, 1.8373e-05, 6.2108e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 221.05, cls_loss 0.0016 cls_loss_mapping 0.0056 cls_loss_causal 0.5557 re_mapping 0.0057 re_causal 0.0168 /// teacc 98.92 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1139, 0.1495, -0.0219, ..., -0.0783, 0.0304, 0.0029], + [ 0.0057, -0.0843, -0.0439, ..., 0.0249, -0.0266, -0.0444], + [-0.0968, -0.1529, 0.0227, ..., -0.0372, -0.0003, -0.0831], + ..., + [ 0.0583, 0.0913, -0.0665, ..., -0.0952, 0.0559, 0.0609], + [ 0.0460, -0.0769, -0.0159, ..., -0.0407, -0.0499, -0.0526], + [-0.1146, -0.0257, -0.0621, ..., 0.0893, -0.2076, 0.0750]], + device='cuda:0'), grad: tensor([[ 5.9372e-07, -2.7288e-06, -4.4238e-08, ..., 5.3504e-07, + -1.4110e-07, 6.1747e-07], + [-1.0673e-06, 1.6764e-07, 4.6566e-10, ..., 6.6217e-07, + 2.0675e-07, 1.8939e-05], + [ 4.7823e-07, 8.2422e-07, 3.2596e-09, ..., 6.4913e-07, + 2.2911e-07, 1.4240e-06], + ..., + [-9.1735e-08, -5.6392e-07, 0.0000e+00, ..., 1.1712e-05, + -2.1793e-07, 6.7353e-05], + [-2.9858e-06, 1.6112e-07, 2.7940e-09, ..., -8.1817e-07, + 1.4435e-07, 2.2873e-06], + [ 8.7023e-06, 1.2703e-06, 2.1420e-08, ..., -1.7196e-05, + 4.6678e-06, -9.7513e-05]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0203, -0.0255, 0.0181, -0.0265, 0.0343, 0.0195, 0.0069, -0.0101, + -0.0145, -0.0067], device='cuda:0'), grad: tensor([ 4.5821e-06, 4.0054e-05, 1.0058e-05, 1.5736e-05, 3.0428e-05, + -6.1452e-05, 4.9800e-05, 1.7452e-04, -5.7518e-05, -2.0587e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 220.44, cls_loss 0.0012 cls_loss_mapping 0.0043 cls_loss_causal 0.5439 re_mapping 0.0055 re_causal 0.0168 /// teacc 99.01 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1147, 0.1502, -0.0220, ..., -0.0786, 0.0303, 0.0032], + [ 0.0060, -0.0844, -0.0440, ..., 0.0260, -0.0270, -0.0448], + [-0.0973, -0.1536, 0.0228, ..., -0.0378, -0.0003, -0.0838], + ..., + [ 0.0587, 0.0915, -0.0666, ..., -0.0959, 0.0562, 0.0612], + [ 0.0460, -0.0768, -0.0160, ..., -0.0412, -0.0502, -0.0528], + [-0.1156, -0.0261, -0.0616, ..., 0.0893, -0.2085, 0.0754]], + device='cuda:0'), grad: tensor([[ 4.1444e-08, -3.4180e-06, 0.0000e+00, ..., 4.2329e-07, + -6.9430e-07, -7.8557e-07], + [ 2.2305e-07, 8.0094e-08, 0.0000e+00, ..., 6.3563e-07, + 3.9116e-08, 2.4168e-07], + [ 2.3749e-07, 2.7753e-07, 0.0000e+00, ..., 2.2259e-07, + 6.5193e-08, 1.5693e-07], + ..., + [ 1.2666e-07, 7.0781e-08, 0.0000e+00, ..., 8.2655e-07, + 3.9581e-08, 5.9837e-07], + [-5.2946e-07, 1.1874e-07, 0.0000e+00, ..., 4.8243e-07, + 7.2643e-08, 3.2363e-07], + [-7.4394e-06, 2.3004e-06, 0.0000e+00, ..., -1.5783e-04, + 4.6426e-07, -7.9811e-05]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0200, -0.0253, 0.0181, -0.0258, 0.0343, 0.0191, 0.0068, -0.0100, + -0.0145, -0.0069], device='cuda:0'), grad: tensor([-3.9674e-06, 1.8114e-06, 1.2973e-06, 1.0654e-06, 3.7813e-04, + 1.4063e-06, -2.2445e-07, 2.3618e-06, -1.1288e-06, -3.8052e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 220.65, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.5251 re_mapping 0.0057 re_causal 0.0165 /// teacc 98.96 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1151, 0.1510, -0.0220, ..., -0.0787, 0.0302, 0.0033], + [ 0.0057, -0.0856, -0.0443, ..., 0.0258, -0.0273, -0.0457], + [-0.0976, -0.1545, 0.0230, ..., -0.0381, -0.0003, -0.0844], + ..., + [ 0.0594, 0.0930, -0.0667, ..., -0.0963, 0.0563, 0.0619], + [ 0.0457, -0.0773, -0.0161, ..., -0.0415, -0.0504, -0.0530], + [-0.1162, -0.0267, -0.0613, ..., 0.0898, -0.2094, 0.0758]], + device='cuda:0'), grad: tensor([[ 5.4948e-08, 0.0000e+00, 0.0000e+00, ..., 6.5193e-08, + 9.5926e-08, 8.3353e-08], + [ 7.7300e-08, 0.0000e+00, 0.0000e+00, ..., -6.8918e-08, + 3.3621e-06, 2.2966e-06], + [ 1.9372e-07, 0.0000e+00, 0.0000e+00, ..., 2.0489e-07, + 1.5991e-06, 1.0692e-06], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 3.3714e-07, + 1.2014e-06, 5.4669e-07], + [ 1.7602e-07, 0.0000e+00, 0.0000e+00, ..., 2.0396e-07, + 5.5209e-06, 3.5483e-06], + [ 1.0468e-06, 0.0000e+00, 0.0000e+00, ..., 1.3113e-06, + 2.8312e-07, 1.6065e-07]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0201, -0.0256, 0.0180, -0.0258, 0.0339, 0.0192, 0.0068, -0.0094, + -0.0148, -0.0066], device='cuda:0'), grad: tensor([ 5.9046e-07, 8.7768e-06, 4.4964e-06, -3.4362e-05, -6.8732e-06, + -1.1176e-08, 2.2575e-06, 4.1798e-06, 1.6347e-05, 4.5709e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 220.16, cls_loss 0.0012 cls_loss_mapping 0.0039 cls_loss_causal 0.5303 re_mapping 0.0056 re_causal 0.0166 /// teacc 99.01 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1156, 0.1517, -0.0221, ..., -0.0789, 0.0303, 0.0037], + [ 0.0057, -0.0857, -0.0444, ..., 0.0259, -0.0276, -0.0459], + [-0.0978, -0.1551, 0.0233, ..., -0.0384, -0.0005, -0.0850], + ..., + [ 0.0595, 0.0932, -0.0668, ..., -0.0970, 0.0563, 0.0620], + [ 0.0456, -0.0777, -0.0161, ..., -0.0421, -0.0513, -0.0533], + [-0.1169, -0.0271, -0.0599, ..., 0.0899, -0.2106, 0.0761]], + device='cuda:0'), grad: tensor([[ 1.4761e-07, -5.1130e-07, -5.1223e-08, ..., 2.3711e-06, + 1.4342e-07, 8.3819e-08], + [-8.5682e-08, 1.2107e-08, -4.1910e-09, ..., -2.8592e-07, + 7.3202e-07, 9.0851e-07], + [-4.0755e-06, 3.1199e-08, 1.0245e-08, ..., 1.3830e-07, + 2.3283e-08, 5.4576e-07], + ..., + [-1.1642e-08, -9.3132e-08, 3.7253e-09, ..., 6.4401e-07, + -2.8014e-06, -2.5034e-06], + [ 6.2026e-07, 1.3504e-08, 4.6566e-09, ..., 2.1234e-07, + 4.2748e-07, 4.1816e-07], + [ 3.5902e-07, 1.3178e-07, 1.2573e-08, ..., -1.3560e-06, + 5.3225e-07, -2.0787e-06]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0202, -0.0253, 0.0180, -0.0247, 0.0344, 0.0187, 0.0063, -0.0096, + -0.0152, -0.0068], device='cuda:0'), grad: tensor([ 8.5384e-06, 1.5125e-06, -1.6838e-05, 6.5677e-06, 7.4431e-06, + 2.9858e-06, -1.1146e-05, 5.1223e-09, 3.5204e-06, -2.5798e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 220.70, cls_loss 0.0013 cls_loss_mapping 0.0037 cls_loss_causal 0.5041 re_mapping 0.0053 re_causal 0.0160 /// teacc 99.01 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.1159, 0.1520, -0.0222, ..., -0.0799, 0.0309, 0.0043], + [ 0.0061, -0.0860, -0.0444, ..., 0.0270, -0.0278, -0.0453], + [-0.0981, -0.1560, 0.0233, ..., -0.0389, -0.0008, -0.0857], + ..., + [ 0.0594, 0.0933, -0.0669, ..., -0.0982, 0.0561, 0.0615], + [ 0.0467, -0.0781, -0.0161, ..., -0.0407, -0.0514, -0.0534], + [-0.1180, -0.0276, -0.0573, ..., 0.0897, -0.2123, 0.0765]], + device='cuda:0'), grad: tensor([[ 8.2422e-08, 1.6391e-07, 5.1223e-09, ..., 4.9919e-07, + 1.1409e-07, 1.4622e-07], + [ 5.2294e-07, 1.4761e-07, 1.3970e-09, ..., -3.0268e-08, + 1.8254e-07, 3.7346e-07], + [ 7.5623e-07, 1.4529e-07, 9.3132e-10, ..., 4.9919e-07, + 4.6566e-07, 4.3865e-07], + ..., + [-8.4285e-08, -9.9465e-07, 0.0000e+00, ..., 5.3085e-07, + 4.0047e-07, -5.3458e-07], + [-2.1551e-06, 1.0943e-07, 3.2596e-09, ..., 4.1025e-07, + 3.1618e-07, 2.3982e-07], + [ 7.8976e-07, 7.2923e-07, 4.6566e-10, ..., -1.7565e-06, + 1.8859e-07, -7.6089e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0199, -0.0250, 0.0177, -0.0251, 0.0348, 0.0190, 0.0062, -0.0102, + -0.0140, -0.0074], device='cuda:0'), grad: tensor([ 2.4624e-06, 1.6401e-06, 4.0978e-06, -5.0068e-06, 2.6897e-06, + 5.0711e-07, -4.7088e-06, 2.0023e-08, -8.2143e-07, -8.6846e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 220.47, cls_loss 0.0013 cls_loss_mapping 0.0040 cls_loss_causal 0.5412 re_mapping 0.0057 re_causal 0.0169 /// teacc 99.07 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.1164, 0.1532, -0.0228, ..., -0.0805, 0.0316, 0.0041], + [ 0.0063, -0.0863, -0.0444, ..., 0.0270, -0.0281, -0.0454], + [-0.0983, -0.1582, 0.0236, ..., -0.0394, -0.0004, -0.0863], + ..., + [ 0.0595, 0.0935, -0.0673, ..., -0.0992, 0.0563, 0.0615], + [ 0.0467, -0.0791, -0.0162, ..., -0.0412, -0.0524, -0.0536], + [-0.1185, -0.0278, -0.0576, ..., 0.0903, -0.2134, 0.0773]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, -8.0047e-07, 0.0000e+00, ..., 3.4738e-07, + -1.1316e-07, 2.7940e-09], + [ 7.7300e-08, 2.2817e-08, 0.0000e+00, ..., 1.1502e-07, + 9.5926e-07, 1.0561e-06], + [ 1.0571e-06, 3.3528e-08, 0.0000e+00, ..., 5.7695e-07, + 3.3434e-07, 2.9150e-07], + ..., + [-9.9652e-08, -7.5437e-08, 0.0000e+00, ..., 1.8114e-07, + 1.2919e-05, 1.4223e-05], + [-2.7334e-07, 2.4214e-08, 0.0000e+00, ..., 4.3213e-07, + 1.7788e-07, 2.6310e-07], + [ 1.5600e-07, 2.3050e-07, 0.0000e+00, ..., -2.5202e-06, + 2.6869e-07, -1.0096e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0198, -0.0251, 0.0182, -0.0253, 0.0345, 0.0184, 0.0074, -0.0103, + -0.0147, -0.0070], device='cuda:0'), grad: tensor([-4.0047e-08, 2.9467e-06, 3.5781e-06, -3.6150e-05, 6.1607e-07, + 2.6189e-06, -3.1330e-06, 3.4750e-05, -6.8452e-08, -5.1931e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 220.22, cls_loss 0.0015 cls_loss_mapping 0.0060 cls_loss_causal 0.5296 re_mapping 0.0059 re_causal 0.0164 /// teacc 98.96 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1172, 0.1555, -0.0238, ..., -0.0787, 0.0316, 0.0038], + [ 0.0064, -0.0891, -0.0451, ..., 0.0259, -0.0284, -0.0454], + [-0.0986, -0.1608, 0.0240, ..., -0.0408, -0.0005, -0.0869], + ..., + [ 0.0597, 0.0943, -0.0679, ..., -0.0997, 0.0564, 0.0616], + [ 0.0468, -0.0801, -0.0164, ..., -0.0421, -0.0529, -0.0539], + [-0.1188, -0.0278, -0.0563, ..., 0.0919, -0.2143, 0.0790]], + device='cuda:0'), grad: tensor([[ 6.7521e-08, -7.9647e-06, 2.5611e-08, ..., 1.2014e-07, + -3.5316e-06, -2.1718e-06], + [ 2.3842e-07, 5.5414e-08, 1.3039e-08, ..., -4.7125e-07, + 2.9523e-07, 1.6112e-07], + [ 6.8871e-07, 1.9092e-08, -8.9873e-08, ..., 1.6158e-07, + 8.0094e-08, 1.4016e-07], + ..., + [-1.6559e-06, 2.6543e-08, 2.0489e-08, ..., 1.0291e-07, + -5.9698e-07, -3.6275e-07], + [ 4.1444e-08, 9.6392e-08, 4.8429e-08, ..., 3.1665e-08, + 2.2771e-07, 9.3132e-08], + [ 3.4459e-07, 5.0664e-07, 9.7789e-09, ..., 4.2375e-07, + 2.8359e-07, 1.5413e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0219, -0.0257, 0.0179, -0.0252, 0.0329, 0.0182, 0.0075, -0.0105, + -0.0150, -0.0055], device='cuda:0'), grad: tensor([-1.1526e-05, 6.2166e-07, 5.8394e-07, 1.3188e-06, 2.7800e-07, + 8.2552e-06, 2.2817e-06, -4.3102e-06, 1.3318e-07, 2.3730e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 220.28, cls_loss 0.0011 cls_loss_mapping 0.0048 cls_loss_causal 0.5443 re_mapping 0.0053 re_causal 0.0163 /// teacc 99.04 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1179, 0.1579, -0.0235, ..., -0.0788, 0.0330, 0.0053], + [ 0.0064, -0.0896, -0.0450, ..., 0.0259, -0.0287, -0.0458], + [-0.0988, -0.1618, 0.0241, ..., -0.0411, -0.0007, -0.0874], + ..., + [ 0.0598, 0.0946, -0.0681, ..., -0.1000, 0.0552, 0.0614], + [ 0.0467, -0.0823, -0.0164, ..., -0.0428, -0.0534, -0.0542], + [-0.1200, -0.0290, -0.0564, ..., 0.0926, -0.2162, 0.0796]], + device='cuda:0'), grad: tensor([[ 2.6543e-08, -1.2554e-06, 0.0000e+00, ..., 1.3113e-06, + 7.9162e-09, -7.5437e-08], + [ 3.4459e-08, 8.3353e-08, 0.0000e+00, ..., 1.8012e-06, + 4.6706e-07, 2.5239e-07], + [ 1.5413e-07, 9.4064e-08, 0.0000e+00, ..., 3.5942e-05, + 6.9896e-07, 3.7067e-07], + ..., + [-4.7032e-07, -2.5798e-07, 0.0000e+00, ..., 5.7742e-08, + -7.0315e-08, -4.0792e-07], + [ 7.7765e-08, 2.0023e-08, 0.0000e+00, ..., 2.5937e-07, + 7.6555e-07, 2.6124e-07], + [ 8.0559e-08, 7.2643e-08, 0.0000e+00, ..., 1.0710e-08, + 1.3085e-07, -2.4680e-08]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0227, -0.0256, 0.0178, -0.0247, 0.0323, 0.0187, 0.0076, -0.0111, + -0.0153, -0.0054], device='cuda:0'), grad: tensor([ 4.1276e-06, 7.5959e-06, 1.3018e-04, -1.2672e-04, 5.2676e-06, + 1.2118e-04, -1.4389e-04, -8.2608e-07, 2.6505e-06, 4.9779e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 220.73, cls_loss 0.0010 cls_loss_mapping 0.0033 cls_loss_causal 0.5163 re_mapping 0.0054 re_causal 0.0162 /// teacc 99.13 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1183, 0.1587, -0.0236, ..., -0.0785, 0.0331, 0.0055], + [ 0.0062, -0.0899, -0.0448, ..., 0.0257, -0.0289, -0.0463], + [-0.0991, -0.1624, 0.0238, ..., -0.0420, -0.0012, -0.0881], + ..., + [ 0.0603, 0.0953, -0.0683, ..., -0.1005, 0.0554, 0.0616], + [ 0.0469, -0.0832, -0.0165, ..., -0.0432, -0.0536, -0.0544], + [-0.1206, -0.0293, -0.0564, ..., 0.0926, -0.2168, 0.0802]], + device='cuda:0'), grad: tensor([[ 1.2713e-07, -3.3993e-08, 0.0000e+00, ..., 1.1036e-07, + 8.8476e-08, 2.0489e-08], + [-2.4606e-06, -1.3690e-07, 0.0000e+00, ..., -1.5032e-06, + 1.6578e-07, -2.9383e-07], + [ 2.2259e-07, 2.7940e-08, 0.0000e+00, ..., 1.9325e-07, + -3.1618e-07, 4.8429e-08], + ..., + [ 2.5798e-07, -1.1548e-07, 0.0000e+00, ..., 3.4133e-07, + -8.5682e-08, -1.4203e-07], + [ 8.0513e-07, 5.9605e-08, 0.0000e+00, ..., 5.9698e-07, + 5.5414e-08, 1.7742e-07], + [ 3.2857e-06, 1.5041e-07, 0.0000e+00, ..., 3.6638e-06, + 6.1607e-07, -4.3074e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0232, -0.0254, 0.0167, -0.0247, 0.0325, 0.0186, 0.0076, -0.0107, + -0.0153, -0.0056], device='cuda:0'), grad: tensor([ 1.3076e-06, -1.3068e-05, -1.2126e-06, 1.9670e-06, -7.3910e-06, + -3.9041e-06, 5.6811e-07, 1.4193e-06, 5.0031e-06, 1.5274e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 220.90, cls_loss 0.0010 cls_loss_mapping 0.0036 cls_loss_causal 0.5156 re_mapping 0.0051 re_causal 0.0164 /// teacc 99.10 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1185, 0.1595, -0.0236, ..., -0.0784, 0.0332, 0.0055], + [ 0.0061, -0.0903, -0.0449, ..., 0.0256, -0.0293, -0.0469], + [-0.0993, -0.1631, 0.0240, ..., -0.0423, -0.0011, -0.0888], + ..., + [ 0.0607, 0.0961, -0.0682, ..., -0.1023, 0.0560, 0.0606], + [ 0.0470, -0.0837, -0.0165, ..., -0.0440, -0.0538, -0.0546], + [-0.1211, -0.0296, -0.0564, ..., 0.0929, -0.2174, 0.0820]], + device='cuda:0'), grad: tensor([[ 1.9697e-07, -3.7393e-07, 0.0000e+00, ..., 3.6135e-07, + 5.8208e-08, 1.2107e-08], + [ 5.5879e-09, 4.0978e-08, -9.3132e-10, ..., -5.0245e-07, + 3.3481e-07, 2.7521e-07], + [ 2.6356e-07, 9.0804e-08, 0.0000e+00, ..., 4.9081e-07, + 9.5740e-07, 4.8801e-07], + ..., + [ 1.7136e-07, 1.0710e-08, 4.6566e-10, ..., 3.4180e-07, + -1.4585e-06, -4.6268e-06], + [ 3.6322e-07, 2.5611e-08, 0.0000e+00, ..., 2.0070e-07, + 7.8604e-07, 3.5902e-07], + [ 2.5947e-06, 2.3795e-07, 0.0000e+00, ..., 5.1744e-06, + 9.5833e-07, 2.5649e-06]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0237, -0.0257, 0.0168, -0.0249, 0.0325, 0.0183, 0.0076, -0.0113, + -0.0155, -0.0050], device='cuda:0'), grad: tensor([ 6.8918e-07, -7.4599e-07, 4.0829e-06, -5.3234e-06, -1.6347e-05, + -1.7313e-06, 5.7854e-06, -6.1952e-06, 3.7663e-06, 1.5974e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 220.32, cls_loss 0.0012 cls_loss_mapping 0.0045 cls_loss_causal 0.5172 re_mapping 0.0051 re_causal 0.0156 /// teacc 99.10 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1192, 0.1602, -0.0235, ..., -0.0785, 0.0330, 0.0057], + [ 0.0055, -0.0903, -0.0456, ..., 0.0247, -0.0296, -0.0471], + [-0.0996, -0.1635, 0.0248, ..., -0.0430, -0.0016, -0.0900], + ..., + [ 0.0609, 0.0964, -0.0683, ..., -0.1029, 0.0561, 0.0606], + [ 0.0468, -0.0841, -0.0163, ..., -0.0450, -0.0543, -0.0548], + [-0.1241, -0.0299, -0.0565, ..., 0.0920, -0.2187, 0.0817]], + device='cuda:0'), grad: tensor([[ 1.6624e-07, -2.7940e-09, 0.0000e+00, ..., 1.7043e-07, + 1.1455e-07, 3.8184e-08], + [-6.6031e-07, 1.3970e-09, 0.0000e+00, ..., -3.4133e-07, + 6.1467e-08, 8.9407e-08], + [ 7.0920e-07, 1.5693e-07, 0.0000e+00, ..., 5.6997e-07, + -1.7229e-08, 1.2899e-07], + ..., + [ 2.1560e-07, -4.5262e-07, 0.0000e+00, ..., 4.5775e-07, + -5.7276e-08, -4.8662e-07], + [-2.2585e-07, 1.2107e-08, 0.0000e+00, ..., 2.1048e-07, + 4.5635e-08, 8.0094e-08], + [ 2.9933e-06, 1.6857e-07, 0.0000e+00, ..., 4.8839e-06, + 7.0315e-08, 5.2154e-08]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0237, -0.0264, 0.0168, -0.0245, 0.0339, 0.0181, 0.0080, -0.0114, + -0.0154, -0.0062], device='cuda:0'), grad: tensor([ 1.3318e-06, -2.2687e-06, 9.5740e-07, 1.7453e-06, -1.7151e-05, + 1.1940e-06, 1.7006e-06, 2.1085e-06, -4.8243e-06, 1.5192e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 220.05, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.5115 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.05 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.1197, 0.1606, -0.0236, ..., -0.0786, 0.0324, 0.0057], + [ 0.0053, -0.0906, -0.0458, ..., 0.0245, -0.0301, -0.0474], + [-0.0998, -0.1641, 0.0249, ..., -0.0436, -0.0020, -0.0905], + ..., + [ 0.0613, 0.0969, -0.0683, ..., -0.1033, 0.0563, 0.0608], + [ 0.0470, -0.0846, -0.0164, ..., -0.0453, -0.0550, -0.0551], + [-0.1248, -0.0302, -0.0565, ..., 0.0920, -0.2198, 0.0817]], + device='cuda:0'), grad: tensor([[ 3.6648e-07, -4.4778e-06, -1.4417e-06, ..., 2.1793e-07, + -7.0641e-07, -2.4103e-06], + [-9.7752e-06, -1.5367e-08, 1.5367e-08, ..., -9.2164e-06, + 7.6834e-08, 2.0489e-07], + [ 4.9248e-06, 4.0932e-07, 5.3551e-08, ..., 1.7444e-06, + 3.7067e-07, 1.4659e-06], + ..., + [-1.6898e-05, 2.4214e-07, 6.5658e-08, ..., -2.0992e-06, + -9.5740e-07, -4.8429e-06], + [ 9.6671e-07, 2.3609e-07, 9.7323e-08, ..., 1.5078e-06, + 7.6834e-08, 6.1654e-07], + [ 6.8769e-06, 7.6788e-07, 1.3737e-07, ..., 1.3439e-06, + 4.8988e-07, 1.4659e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0236, -0.0267, 0.0169, -0.0239, 0.0340, 0.0180, 0.0081, -0.0113, + -0.0152, -0.0064], device='cuda:0'), grad: tensor([-6.6534e-06, -1.7154e-04, 1.3983e-04, 5.8934e-06, 1.9386e-05, + 1.0461e-05, 1.0416e-05, -2.9951e-05, 4.1276e-06, 1.7956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 220.55, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5583 re_mapping 0.0057 re_causal 0.0171 /// teacc 99.15 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1205, 0.1613, -0.0234, ..., -0.0788, 0.0325, 0.0060], + [ 0.0038, -0.0931, -0.0455, ..., 0.0238, -0.0304, -0.0492], + [-0.1005, -0.1648, 0.0244, ..., -0.0449, -0.0021, -0.0909], + ..., + [ 0.0631, 0.0994, -0.0689, ..., -0.1034, 0.0564, 0.0615], + [ 0.0474, -0.0871, -0.0165, ..., -0.0480, -0.0555, -0.0558], + [-0.1284, -0.0302, -0.0549, ..., 0.0895, -0.2206, 0.0795]], + device='cuda:0'), grad: tensor([[ 2.0023e-07, 3.0268e-08, 9.3132e-10, ..., 5.0571e-07, + 2.6915e-07, 1.2433e-07], + [-2.3037e-05, -5.1223e-08, 0.0000e+00, ..., -3.6299e-05, + 2.6776e-07, 1.1642e-07], + [ 1.6531e-07, 7.1712e-08, 0.0000e+00, ..., 1.7229e-07, + -1.0006e-05, 1.3364e-07], + ..., + [ 4.0745e-07, -4.6380e-07, 0.0000e+00, ..., 1.3215e-06, + 1.8300e-07, 1.7695e-07], + [ 4.3400e-07, 5.4948e-08, 4.6566e-10, ..., 7.2923e-07, + 7.1665e-07, 3.2084e-07], + [ 8.8057e-07, 6.1002e-08, 0.0000e+00, ..., -2.2072e-06, + 5.1688e-08, -2.7064e-06]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0237, -0.0280, 0.0161, -0.0239, 0.0368, 0.0174, 0.0084, -0.0101, + -0.0150, -0.0089], device='cuda:0'), grad: tensor([ 2.4568e-06, -8.0407e-05, -3.6836e-05, 3.2753e-05, 7.4148e-05, + 4.9695e-06, 8.1165e-07, 3.9041e-06, 3.0082e-06, -4.7386e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 220.24, cls_loss 0.0012 cls_loss_mapping 0.0039 cls_loss_causal 0.5270 re_mapping 0.0051 re_causal 0.0153 /// teacc 99.08 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1213, 0.1615, -0.0249, ..., -0.0798, 0.0323, 0.0054], + [ 0.0042, -0.0930, -0.0457, ..., 0.0242, -0.0306, -0.0495], + [-0.1008, -0.1656, 0.0247, ..., -0.0455, -0.0017, -0.0915], + ..., + [ 0.0631, 0.0997, -0.0691, ..., -0.1038, 0.0566, 0.0618], + [ 0.0475, -0.0880, -0.0166, ..., -0.0484, -0.0561, -0.0561], + [-0.1286, -0.0301, -0.0547, ..., 0.0896, -0.2216, 0.0798]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, -8.3121e-07, 9.3132e-10, ..., 3.5251e-07, + -5.3551e-08, 6.2864e-08], + [-1.2014e-06, 9.3132e-08, 1.3970e-09, ..., -2.0340e-06, + 6.9384e-08, 2.0908e-07], + [ 6.4913e-07, 4.3726e-07, 9.3132e-10, ..., 9.6485e-07, + -2.1048e-07, 3.1898e-07], + ..., + [-4.7358e-07, -5.9698e-07, -2.2352e-08, ..., 5.8394e-07, + -2.7334e-07, -4.3679e-07], + [ 2.3283e-08, 6.4727e-08, 9.3132e-10, ..., 3.0268e-07, + 2.5146e-08, 1.2992e-07], + [ 1.6065e-07, 3.2177e-07, 1.8626e-09, ..., -6.0238e-06, + 6.4261e-08, -4.7013e-06]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0227, -0.0271, 0.0156, -0.0239, 0.0368, 0.0170, 0.0091, -0.0102, + -0.0151, -0.0088], device='cuda:0'), grad: tensor([ 1.9372e-07, -8.6203e-06, 3.7253e-06, 1.3728e-06, 1.2703e-05, + 1.0692e-06, 2.1420e-06, -3.0827e-07, 8.2701e-07, -1.3128e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 220.21, cls_loss 0.0012 cls_loss_mapping 0.0034 cls_loss_causal 0.5310 re_mapping 0.0053 re_causal 0.0156 /// teacc 99.09 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1224, 0.1621, -0.0259, ..., -0.0805, 0.0323, 0.0056], + [ 0.0048, -0.0922, -0.0464, ..., 0.0245, -0.0306, -0.0491], + [-0.1010, -0.1662, 0.0254, ..., -0.0458, -0.0024, -0.0922], + ..., + [ 0.0629, 0.0993, -0.0691, ..., -0.1043, 0.0572, 0.0617], + [ 0.0477, -0.0885, -0.0167, ..., -0.0484, -0.0568, -0.0563], + [-0.1289, -0.0306, -0.0546, ..., 0.0896, -0.2228, 0.0799]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, -2.3648e-05, 9.3132e-10, ..., -1.1604e-06, + -4.5113e-06, -5.4725e-06], + [ 1.2573e-07, 6.3749e-07, 4.6566e-10, ..., -1.7695e-08, + 8.6427e-07, 1.1045e-06], + [ 5.0897e-07, 1.2387e-06, 4.6566e-10, ..., 3.1758e-07, + 8.2655e-07, 1.0952e-06], + ..., + [-2.3581e-06, -3.6396e-06, 0.0000e+00, ..., 1.5367e-07, + -2.2039e-05, -3.0220e-05], + [ 8.3307e-07, 1.3404e-05, 2.3283e-09, ..., 7.1246e-07, + 1.0960e-05, 1.4670e-05], + [ 2.8731e-07, 2.2147e-06, 0.0000e+00, ..., -1.1269e-07, + 2.7716e-06, 3.4329e-06]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0222, -0.0266, 0.0155, -0.0242, 0.0370, 0.0174, 0.0089, -0.0104, + -0.0148, -0.0090], device='cuda:0'), grad: tensor([-5.5403e-05, 3.8967e-06, 6.1952e-06, 4.2766e-05, -1.1316e-06, + 4.4610e-07, 1.1466e-05, -8.1062e-05, 6.0260e-05, 1.2510e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 220.28, cls_loss 0.0012 cls_loss_mapping 0.0041 cls_loss_causal 0.5048 re_mapping 0.0056 re_causal 0.0152 /// teacc 98.98 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.1229, 0.1627, -0.0259, ..., -0.0809, 0.0312, 0.0060], + [ 0.0045, -0.0923, -0.0464, ..., 0.0242, -0.0310, -0.0498], + [-0.1014, -0.1668, 0.0255, ..., -0.0468, -0.0032, -0.0943], + ..., + [ 0.0638, 0.0998, -0.0691, ..., -0.1047, 0.0590, 0.0633], + [ 0.0474, -0.0893, -0.0167, ..., -0.0491, -0.0580, -0.0570], + [-0.1295, -0.0311, -0.0546, ..., 0.0894, -0.2244, 0.0797]], + device='cuda:0'), grad: tensor([[ 7.3109e-08, -1.0051e-05, 9.3132e-10, ..., 9.7789e-08, + 1.6978e-06, -2.9616e-07], + [ 2.1281e-07, 5.0850e-07, -8.3819e-09, ..., -3.2131e-08, + 3.6322e-07, 3.2736e-07], + [-2.3097e-07, 3.2820e-06, -5.8208e-08, ..., 7.2643e-08, + 5.4436e-07, 3.7765e-07], + ..., + [-1.3653e-06, -7.0920e-07, 6.1467e-08, ..., 1.6950e-07, + -9.4157e-07, -1.4016e-06], + [ 6.9570e-07, 1.4622e-06, 6.0536e-09, ..., 8.4750e-08, + 9.0711e-07, 8.4331e-07], + [ 2.9579e-06, 6.2119e-07, 3.2596e-09, ..., 1.2726e-05, + 4.9593e-07, 2.4447e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0220, -0.0266, 0.0144, -0.0248, 0.0375, 0.0178, 0.0088, -0.0091, + -0.0153, -0.0094], device='cuda:0'), grad: tensor([-1.2033e-05, 4.4890e-06, -1.4715e-06, -1.4409e-05, -2.0832e-05, + 7.4320e-06, 5.4017e-06, -1.4063e-06, 7.6368e-06, 2.5168e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 220.47, cls_loss 0.0012 cls_loss_mapping 0.0033 cls_loss_causal 0.5353 re_mapping 0.0053 re_causal 0.0157 /// teacc 99.00 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1238, 0.1635, -0.0260, ..., -0.0808, 0.0312, 0.0061], + [ 0.0048, -0.0927, -0.0465, ..., 0.0244, -0.0315, -0.0500], + [-0.1020, -0.1677, 0.0256, ..., -0.0472, -0.0038, -0.0952], + ..., + [ 0.0642, 0.1002, -0.0691, ..., -0.1053, 0.0594, 0.0635], + [ 0.0472, -0.0898, -0.0167, ..., -0.0499, -0.0586, -0.0573], + [-0.1297, -0.0314, -0.0541, ..., 0.0899, -0.2244, 0.0802]], + device='cuda:0'), grad: tensor([[ 2.3516e-07, 2.1281e-07, 2.0023e-08, ..., 4.5635e-08, + 2.3283e-09, 2.1234e-07], + [ 3.0641e-07, 1.6857e-07, 8.3819e-09, ..., 1.7835e-07, + 2.7008e-08, 1.9046e-07], + [ 1.0524e-07, 2.6543e-08, 4.6566e-09, ..., 2.4214e-08, + 1.9558e-08, 6.1002e-08], + ..., + [-6.4773e-07, -1.1437e-06, 4.6566e-10, ..., 4.1304e-07, + -1.1828e-07, -9.3505e-07], + [ 5.5619e-06, 5.5879e-08, 6.9570e-07, ..., 3.0501e-07, + 8.4750e-08, 3.1013e-07], + [ 4.9919e-07, 4.8662e-07, 7.9162e-09, ..., -1.0487e-06, + 2.0023e-08, -5.9931e-07]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0222, -0.0287, 0.0166, -0.0246, 0.0371, 0.0180, 0.0085, -0.0090, + -0.0158, -0.0090], device='cuda:0'), grad: tensor([ 7.3807e-07, 9.3691e-07, 2.8685e-07, 6.0210e-07, 5.6252e-07, + -9.6440e-05, 8.3566e-05, -1.5311e-06, 1.2711e-05, -1.5134e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 220.57, cls_loss 0.0013 cls_loss_mapping 0.0041 cls_loss_causal 0.5594 re_mapping 0.0052 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.1248, 0.1643, -0.0260, ..., -0.0808, 0.0317, 0.0062], + [ 0.0063, -0.0929, -0.0465, ..., 0.0254, -0.0304, -0.0487], + [-0.1023, -0.1693, 0.0257, ..., -0.0475, -0.0038, -0.0968], + ..., + [ 0.0630, 0.1005, -0.0696, ..., -0.1067, 0.0588, 0.0627], + [ 0.0464, -0.0901, -0.0170, ..., -0.0506, -0.0597, -0.0577], + [-0.1300, -0.0309, -0.0541, ..., 0.0898, -0.2253, 0.0805]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 8.5216e-08, 0.0000e+00, ..., 5.6811e-08, + 2.3423e-07, 1.8068e-07], + [ 1.7975e-07, 1.4110e-07, 0.0000e+00, ..., -7.9162e-09, + 2.1001e-07, 3.8138e-07], + [ 9.6951e-07, 6.2864e-07, 0.0000e+00, ..., 3.2131e-08, + 8.0653e-07, 1.3374e-06], + ..., + [-3.4124e-06, -2.1849e-06, 0.0000e+00, ..., 2.9290e-07, + -2.4103e-06, -4.9137e-06], + [ 3.1386e-07, 3.5390e-08, 0.0000e+00, ..., 1.2107e-07, + 2.9802e-07, 5.1735e-07], + [ 1.8338e-06, 8.0513e-07, 0.0000e+00, ..., 1.5236e-06, + 5.0152e-07, 1.5125e-06]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0222, -0.0276, 0.0164, -0.0236, 0.0372, 0.0169, 0.0089, -0.0100, + -0.0168, -0.0090], device='cuda:0'), grad: tensor([ 7.8836e-07, 1.1064e-06, 4.7982e-06, 3.3509e-06, -4.9286e-06, + 8.5635e-07, 1.9521e-06, -1.4968e-05, 9.0757e-07, 6.0946e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 220.41, cls_loss 0.0011 cls_loss_mapping 0.0040 cls_loss_causal 0.5331 re_mapping 0.0049 re_causal 0.0156 /// teacc 99.11 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1255, 0.1648, -0.0260, ..., -0.0810, 0.0318, 0.0049], + [ 0.0069, -0.0930, -0.0466, ..., 0.0258, -0.0282, -0.0488], + [-0.1025, -0.1699, 0.0257, ..., -0.0477, -0.0059, -0.0975], + ..., + [ 0.0627, 0.1010, -0.0697, ..., -0.1071, 0.0592, 0.0629], + [ 0.0467, -0.0906, -0.0163, ..., -0.0511, -0.0602, -0.0581], + [-0.1305, -0.0306, -0.0543, ..., 0.0897, -0.2263, 0.0808]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, -1.1906e-05, -5.9232e-06, ..., 1.4901e-08, + -4.3563e-07, -6.9216e-06], + [-1.4086e-07, 1.6624e-07, 4.7963e-08, ..., -1.4435e-07, + 2.0722e-08, 7.3574e-08], + [ 4.4238e-08, 7.8604e-07, 5.5879e-08, ..., 3.6322e-08, + 1.4366e-07, 1.8044e-07], + ..., + [ 6.7521e-09, 3.4482e-07, 1.7066e-07, ..., 2.5146e-08, + 1.1176e-08, 1.8114e-07], + [ 1.0477e-07, 4.0308e-06, 2.3376e-06, ..., 3.5623e-08, + 6.8685e-08, 2.5798e-06], + [ 2.5379e-08, 3.2689e-06, 1.5730e-06, ..., -1.2410e-07, + 1.7858e-07, 1.9073e-06]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0218, -0.0266, 0.0158, -0.0239, 0.0374, 0.0172, 0.0076, -0.0102, + -0.0164, -0.0090], device='cuda:0'), grad: tensor([-3.1799e-05, -6.9616e-08, 1.3439e-06, 9.6112e-07, 4.2212e-07, + 5.2042e-06, 2.4531e-06, 9.6578e-07, 1.1832e-05, 8.6650e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 220.16, cls_loss 0.0010 cls_loss_mapping 0.0038 cls_loss_causal 0.5554 re_mapping 0.0049 re_causal 0.0163 /// teacc 99.05 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1261, 0.1654, -0.0259, ..., -0.0814, 0.0315, 0.0052], + [ 0.0058, -0.0933, -0.0478, ..., 0.0244, -0.0284, -0.0490], + [-0.1028, -0.1710, 0.0259, ..., -0.0480, -0.0061, -0.0989], + ..., + [ 0.0635, 0.1023, -0.0705, ..., -0.1076, 0.0600, 0.0634], + [ 0.0461, -0.0914, -0.0161, ..., -0.0518, -0.0607, -0.0584], + [-0.1314, -0.0315, -0.0543, ..., 0.0896, -0.2285, 0.0808]], + device='cuda:0'), grad: tensor([[ 4.1793e-07, 2.1141e-07, 0.0000e+00, ..., 7.7207e-07, + 7.0501e-07, 6.4494e-08], + [-4.4852e-06, 3.8650e-08, 0.0000e+00, ..., 3.2363e-08, + -1.5125e-05, 2.6659e-07], + [ 4.8988e-06, 2.3097e-07, 0.0000e+00, ..., 2.8452e-07, + -1.4566e-06, 9.0571e-08], + ..., + [ 1.0366e-06, -1.2410e-07, 0.0000e+00, ..., 2.1663e-06, + 2.8759e-06, 5.3495e-06], + [-1.6123e-05, -1.5330e-06, 0.0000e+00, ..., -1.8878e-06, + 9.2480e-07, 2.5169e-07], + [ 5.5879e-08, 4.7009e-07, 0.0000e+00, ..., -3.0380e-06, + 1.9441e-07, -7.0892e-06]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0215, -0.0270, 0.0157, -0.0240, 0.0378, 0.0174, 0.0085, -0.0099, + -0.0169, -0.0094], device='cuda:0'), grad: tensor([ 9.6932e-06, -8.0884e-05, 1.0699e-05, 6.0230e-05, 4.5747e-06, + 2.2516e-05, 1.5073e-05, 2.7731e-05, -5.8621e-05, -1.1057e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 220.56, cls_loss 0.0011 cls_loss_mapping 0.0035 cls_loss_causal 0.5347 re_mapping 0.0051 re_causal 0.0153 /// teacc 99.07 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1270, 0.1662, -0.0260, ..., -0.0817, 0.0321, 0.0057], + [ 0.0046, -0.0947, -0.0487, ..., 0.0239, -0.0292, -0.0501], + [-0.1007, -0.1715, 0.0263, ..., -0.0483, -0.0060, -0.0994], + ..., + [ 0.0644, 0.1037, -0.0706, ..., -0.1083, 0.0608, 0.0638], + [ 0.0442, -0.0918, -0.0162, ..., -0.0522, -0.0610, -0.0586], + [-0.1307, -0.0318, -0.0537, ..., 0.0902, -0.2278, 0.0818]], + device='cuda:0'), grad: tensor([[ 4.4657e-07, -1.6042e-07, -6.9849e-10, ..., 1.7229e-08, + 1.3760e-07, 6.7707e-07], + [-1.1805e-07, 4.4238e-09, 7.6834e-09, ..., -1.9977e-07, + 2.6752e-07, 3.0966e-07], + [ 1.8114e-07, 6.5193e-09, -1.8626e-08, ..., 6.5193e-08, + 1.8487e-07, 3.5740e-07], + ..., + [-9.1493e-06, -9.7789e-09, 2.0955e-09, ..., 8.8010e-07, + 2.4820e-07, -1.3813e-05], + [ 5.2713e-07, 3.7253e-09, 5.1223e-09, ..., 2.4913e-08, + 4.4145e-07, 3.0315e-07], + [ 4.7311e-06, 1.0151e-07, 1.1642e-09, ..., -1.6317e-06, + 8.7265e-07, 5.9530e-06]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0216, -0.0276, 0.0173, -0.0242, 0.0372, 0.0174, 0.0084, -0.0096, + -0.0190, -0.0087], device='cuda:0'), grad: tensor([ 1.9781e-06, 4.8522e-07, 3.8301e-07, -5.9716e-06, 1.5497e-06, + 1.4283e-05, 1.0412e-06, -3.3081e-05, 2.2985e-06, 1.6958e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 220.41, cls_loss 0.0009 cls_loss_mapping 0.0032 cls_loss_causal 0.5287 re_mapping 0.0050 re_causal 0.0153 /// teacc 99.07 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1281, 0.1666, -0.0258, ..., -0.0819, 0.0320, 0.0057], + [ 0.0043, -0.0947, -0.0488, ..., 0.0231, -0.0291, -0.0508], + [-0.1009, -0.1718, 0.0264, ..., -0.0486, -0.0062, -0.0997], + ..., + [ 0.0640, 0.1039, -0.0735, ..., -0.1088, 0.0604, 0.0640], + [ 0.0445, -0.0922, -0.0164, ..., -0.0527, -0.0615, -0.0589], + [-0.1313, -0.0321, -0.0541, ..., 0.0901, -0.2284, 0.0820]], + device='cuda:0'), grad: tensor([[ 7.3574e-08, -4.7288e-07, 0.0000e+00, ..., 1.4016e-07, + -7.4506e-09, 9.8487e-08], + [-3.3528e-06, 6.2166e-08, -2.0955e-09, ..., -1.8431e-06, + -6.1234e-07, -6.8685e-07], + [ 2.6682e-07, 8.6613e-08, 0.0000e+00, ..., 1.5879e-07, + 1.5227e-07, 2.3656e-07], + ..., + [ 1.5572e-06, -2.3656e-07, 6.9849e-10, ..., 1.1772e-06, + 3.7532e-07, 3.1688e-07], + [-4.2515e-07, 3.2363e-08, 0.0000e+00, ..., 1.9115e-07, + 2.7940e-08, 1.7742e-07], + [ 8.3447e-07, 2.8941e-07, 0.0000e+00, ..., -7.9162e-08, + 1.5786e-07, -6.2678e-07]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0214, -0.0278, 0.0174, -0.0244, 0.0374, 0.0180, 0.0083, -0.0101, + -0.0188, -0.0087], device='cuda:0'), grad: tensor([ 1.4622e-07, -7.9796e-06, 1.4268e-06, -8.0373e-07, 1.3644e-06, + 1.4603e-06, 6.6590e-07, 4.9509e-06, -1.0207e-06, -2.4284e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 220.31, cls_loss 0.0010 cls_loss_mapping 0.0053 cls_loss_causal 0.5174 re_mapping 0.0052 re_causal 0.0151 /// teacc 99.05 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1299, 0.1670, -0.0256, ..., -0.0823, 0.0323, 0.0060], + [ 0.0056, -0.0948, -0.0486, ..., 0.0233, -0.0292, -0.0508], + [-0.1021, -0.1741, 0.0263, ..., -0.0496, -0.0062, -0.1005], + ..., + [ 0.0643, 0.1048, -0.0736, ..., -0.1101, 0.0605, 0.0634], + [ 0.0442, -0.0914, -0.0165, ..., -0.0531, -0.0620, -0.0590], + [-0.1313, -0.0324, -0.0542, ..., 0.0904, -0.2294, 0.0824]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, -4.0159e-06, -9.7789e-09, ..., 1.3039e-07, + -1.0268e-07, -5.5134e-07], + [-8.5449e-08, 6.9942e-07, 1.1176e-08, ..., -1.0268e-07, + 2.6776e-08, 1.1991e-07], + [ 1.1362e-07, 4.0024e-07, 7.2177e-09, ..., 5.1223e-08, + 6.7987e-08, 1.0827e-07], + ..., + [ 9.7789e-09, 4.7963e-08, 2.7940e-09, ..., 8.1491e-08, + -1.1059e-07, -9.7090e-08], + [ 2.1397e-07, 2.2608e-07, 2.7358e-07, ..., 8.3819e-08, + 1.3527e-07, 5.3784e-08], + [ 8.0792e-08, 6.9989e-07, 2.2585e-08, ..., -4.1886e-07, + 7.0082e-08, -3.0990e-07]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0210, -0.0270, 0.0167, -0.0243, 0.0373, 0.0180, 0.0083, -0.0104, + -0.0190, -0.0085], device='cuda:0'), grad: tensor([-6.1244e-06, 8.8941e-07, 2.5947e-06, 1.7714e-06, 1.1344e-06, + -7.1973e-06, 8.6576e-06, 4.6659e-07, -2.3656e-06, 1.7718e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 220.39, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5404 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.00 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1307, 0.1692, -0.0263, ..., -0.0825, 0.0290, 0.0059], + [ 0.0054, -0.0951, -0.0490, ..., 0.0235, -0.0300, -0.0514], + [-0.1023, -0.1750, 0.0283, ..., -0.0499, -0.0054, -0.1012], + ..., + [ 0.0647, 0.1051, -0.0747, ..., -0.1117, 0.0613, 0.0630], + [ 0.0447, -0.0916, -0.0171, ..., -0.0535, -0.0625, -0.0594], + [-0.1316, -0.0324, -0.0543, ..., 0.0906, -0.2303, 0.0832]], + device='cuda:0'), grad: tensor([[ 6.0536e-08, -6.1318e-06, -5.1688e-08, ..., -1.7229e-08, + -5.4389e-07, -2.0908e-07], + [ 2.7986e-07, 4.2608e-07, -1.1874e-08, ..., 1.8789e-07, + 6.3051e-07, 7.3668e-07], + [ 5.2107e-07, 3.2899e-07, 8.3819e-09, ..., 1.8673e-07, + 9.0571e-07, 1.0831e-06], + ..., + [-2.8666e-06, 3.2852e-07, 4.1910e-09, ..., 4.0210e-07, + -5.5730e-06, -6.8992e-06], + [ 1.5320e-07, 4.8522e-07, 1.1176e-08, ..., 6.9477e-07, + 3.6578e-07, 7.1200e-07], + [ 1.8375e-06, 6.6031e-07, 6.9849e-09, ..., -8.8755e-07, + 3.0808e-06, 2.8145e-06]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0214, -0.0269, 0.0170, -0.0241, 0.0372, 0.0177, 0.0071, -0.0112, + -0.0185, -0.0082], device='cuda:0'), grad: tensor([-1.2085e-05, 7.0222e-06, -4.0941e-06, 7.7337e-06, -1.4426e-06, + 4.2021e-06, 5.1484e-06, -3.1769e-05, 6.1914e-06, 1.9133e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 220.54, cls_loss 0.0009 cls_loss_mapping 0.0034 cls_loss_causal 0.5118 re_mapping 0.0052 re_causal 0.0149 /// teacc 99.01 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1312, 0.1700, -0.0262, ..., -0.0826, 0.0279, 0.0059], + [ 0.0053, -0.0954, -0.0494, ..., 0.0233, -0.0304, -0.0518], + [-0.1025, -0.1753, 0.0289, ..., -0.0504, -0.0055, -0.1025], + ..., + [ 0.0653, 0.1051, -0.0744, ..., -0.1122, 0.0618, 0.0634], + [ 0.0445, -0.0917, -0.0168, ..., -0.0545, -0.0630, -0.0596], + [-0.1319, -0.0325, -0.0546, ..., 0.0907, -0.2311, 0.0835]], + device='cuda:0'), grad: tensor([[ 2.1327e-06, -2.4331e-07, 1.1008e-06, ..., 1.7025e-06, + 2.0433e-06, -1.6531e-08], + [ 1.3066e-06, 6.0536e-09, 6.7940e-07, ..., 2.4983e-07, + 1.5236e-06, 2.3260e-07], + [ 1.8757e-06, 7.6834e-09, 9.1130e-07, ..., 7.1200e-07, + 1.8803e-06, 8.6846e-08], + ..., + [ 2.4557e-05, 5.3551e-09, 1.3046e-05, ..., 1.2387e-07, + 2.4542e-05, 1.5600e-07], + [ 2.2314e-06, 2.3516e-08, 1.1371e-06, ..., 8.0913e-06, + 2.5500e-06, 5.4203e-07], + [ 3.3956e-06, 1.0198e-07, 1.7295e-06, ..., -1.0338e-07, + 6.2659e-06, 2.3600e-06]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0213, -0.0270, 0.0169, -0.0247, 0.0372, 0.0183, 0.0070, -0.0109, + -0.0186, -0.0081], device='cuda:0'), grad: tensor([ 1.7688e-05, 9.1866e-06, 1.2510e-05, 2.8804e-05, 1.3329e-05, + -2.4819e-04, -4.2945e-05, 1.3626e-04, 4.3333e-05, 3.0085e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 220.19, cls_loss 0.0011 cls_loss_mapping 0.0044 cls_loss_causal 0.5427 re_mapping 0.0051 re_causal 0.0150 /// teacc 99.11 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1318, 0.1708, -0.0261, ..., -0.0827, 0.0281, 0.0054], + [ 0.0054, -0.0958, -0.0495, ..., 0.0231, -0.0312, -0.0522], + [-0.1027, -0.1759, 0.0290, ..., -0.0508, -0.0055, -0.1032], + ..., + [ 0.0655, 0.1067, -0.0748, ..., -0.1129, 0.0635, 0.0642], + [ 0.0445, -0.0923, -0.0169, ..., -0.0552, -0.0634, -0.0600], + [-0.1321, -0.0329, -0.0546, ..., 0.0907, -0.2325, 0.0837]], + device='cuda:0'), grad: tensor([[ 3.5856e-08, 1.5483e-07, 1.2117e-06, ..., 5.3868e-06, + -3.4925e-09, 3.2261e-06], + [ 5.7090e-07, 2.1304e-07, -3.7253e-09, ..., -7.6834e-08, + 4.2375e-08, 1.0990e-07], + [-3.1968e-07, 1.1455e-07, 6.2864e-09, ..., 6.4261e-08, + 1.0547e-07, 1.2130e-07], + ..., + [ 2.8173e-07, -2.4331e-07, 6.7521e-09, ..., 8.7544e-08, + -2.8079e-07, -4.1607e-07], + [-1.2554e-06, 3.5390e-08, 1.0710e-08, ..., 1.8161e-07, + 5.5879e-09, 1.1781e-07], + [ 8.0559e-08, -5.6252e-07, -1.2694e-06, ..., -6.2101e-06, + 7.9395e-08, -3.7160e-06]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0213, -0.0272, 0.0171, -0.0247, 0.0372, 0.0179, 0.0071, -0.0104, + -0.0188, -0.0082], device='cuda:0'), grad: tensor([ 1.0215e-05, 3.4645e-06, -8.7023e-06, 8.3214e-07, 1.3523e-06, + 1.9707e-06, 6.8732e-07, 7.8157e-06, -5.3942e-06, -1.2197e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 220.21, cls_loss 0.0010 cls_loss_mapping 0.0036 cls_loss_causal 0.5315 re_mapping 0.0047 re_causal 0.0147 /// teacc 99.11 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1325, 0.1713, -0.0261, ..., -0.0830, 0.0283, 0.0054], + [ 0.0054, -0.0960, -0.0498, ..., 0.0229, -0.0315, -0.0525], + [-0.1027, -0.1762, 0.0292, ..., -0.0488, -0.0047, -0.1037], + ..., + [ 0.0661, 0.1071, -0.0751, ..., -0.1131, 0.0634, 0.0646], + [ 0.0446, -0.0926, -0.0171, ..., -0.0562, -0.0640, -0.0603], + [-0.1325, -0.0333, -0.0546, ..., 0.0907, -0.2338, 0.0838]], + device='cuda:0'), grad: tensor([[ 1.7020e-07, -2.0955e-08, 1.0263e-06, ..., 8.8476e-07, + 1.5111e-07, 2.8661e-07], + [ 1.6741e-07, 2.0955e-08, 1.3057e-06, ..., 2.4051e-07, + 1.2005e-06, 5.9698e-07], + [ 9.3877e-07, 3.4459e-08, -7.0296e-06, ..., 3.1479e-07, + -1.7853e-06, 1.2293e-07], + ..., + [-3.4925e-09, -1.5832e-07, 7.3295e-07, ..., 2.4885e-06, + 1.6135e-07, 3.1721e-06], + [-1.8254e-06, 7.6834e-09, 1.9204e-06, ..., 1.5097e-06, + 4.8848e-07, 1.9884e-07], + [ 1.8859e-07, 6.1002e-08, 2.0000e-07, ..., -5.5656e-06, + 1.8603e-07, -6.3144e-06]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0212, -0.0274, 0.0184, -0.0265, 0.0371, 0.0193, 0.0061, -0.0105, + -0.0187, -0.0083], device='cuda:0'), grad: tensor([ 8.8364e-06, 1.9848e-05, -5.9754e-05, 2.4661e-05, 1.2942e-05, + -1.1533e-05, -2.2817e-06, 1.6108e-05, 8.4862e-06, -1.7419e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 220.14, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.5086 re_mapping 0.0047 re_causal 0.0143 /// teacc 99.03 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1333, 0.1716, -0.0263, ..., -0.0832, 0.0269, 0.0054], + [ 0.0054, -0.0961, -0.0498, ..., 0.0227, -0.0311, -0.0526], + [-0.1030, -0.1763, 0.0300, ..., -0.0491, -0.0072, -0.1072], + ..., + [ 0.0663, 0.1074, -0.0762, ..., -0.1135, 0.0649, 0.0660], + [ 0.0449, -0.0929, -0.0174, ..., -0.0568, -0.0645, -0.0605], + [-0.1327, -0.0336, -0.0545, ..., 0.0908, -0.2346, 0.0839]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, -6.2399e-07, 4.6566e-10, ..., 6.8685e-08, + 3.9767e-07, -7.9162e-08], + [ 3.0710e-07, 1.5832e-08, 5.2387e-08, ..., 2.9011e-07, + 3.8650e-08, 2.5146e-08], + [ 2.1886e-08, 2.3516e-08, -4.4121e-07, ..., 3.3062e-08, + -1.5097e-06, 8.1491e-09], + ..., + [ 3.3993e-08, 2.1420e-08, 2.5122e-07, ..., 2.3819e-07, + 1.5320e-07, 2.2934e-07], + [-6.0536e-09, 7.7067e-08, 5.6345e-08, ..., 2.1188e-07, + 7.1712e-08, 9.3132e-08], + [ 1.3388e-07, 1.5274e-07, 5.6811e-08, ..., -4.6240e-07, + 4.8894e-09, -6.5332e-07]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0208, -0.0272, 0.0165, -0.0263, 0.0372, 0.0191, 0.0063, -0.0088, + -0.0182, -0.0083], device='cuda:0'), grad: tensor([ 6.7614e-07, 1.4044e-06, -7.2159e-06, 2.0806e-06, -6.0210e-07, + 5.2154e-07, 9.8022e-08, 2.6636e-06, 1.1148e-06, -7.2317e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 220.65, cls_loss 0.0014 cls_loss_mapping 0.0041 cls_loss_causal 0.5418 re_mapping 0.0049 re_causal 0.0145 /// teacc 99.01 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1358, 0.1727, -0.0272, ..., -0.0828, 0.0257, 0.0060], + [ 0.0058, -0.0959, -0.0504, ..., 0.0242, -0.0318, -0.0533], + [-0.1028, -0.1771, 0.0325, ..., -0.0497, -0.0065, -0.1080], + ..., + [ 0.0663, 0.1077, -0.0793, ..., -0.1143, 0.0646, 0.0665], + [ 0.0455, -0.0933, -0.0178, ..., -0.0568, -0.0651, -0.0602], + [-0.1330, -0.0340, -0.0548, ..., 0.0909, -0.2353, 0.0842]], + device='cuda:0'), grad: tensor([[ 3.4226e-08, -5.4017e-06, 2.3283e-10, ..., 1.6794e-05, + -6.6962e-07, -5.1921e-08], + [ 6.9849e-08, 2.3935e-07, 0.0000e+00, ..., 2.4233e-06, + 8.9640e-08, 1.5600e-08], + [ 3.0571e-07, 1.1940e-06, -2.3283e-10, ..., 1.2945e-06, + 2.6985e-07, 1.4203e-08], + ..., + [ 1.9372e-07, 1.4342e-07, 2.3283e-10, ..., 7.9535e-07, + 4.2841e-08, 1.1292e-07], + [-1.0841e-06, 2.1374e-07, 2.3283e-10, ..., 3.3583e-06, + 4.5635e-08, 5.9372e-08], + [ 1.3690e-07, 3.7975e-07, 0.0000e+00, ..., 1.1541e-05, + 3.7719e-08, -5.7742e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0208, -0.0266, 0.0164, -0.0264, 0.0367, 0.0201, 0.0052, -0.0089, + -0.0175, -0.0083], device='cuda:0'), grad: tensor([ 4.7147e-05, 9.0674e-06, 7.3574e-06, 5.0664e-06, 1.5748e-04, + 6.0126e-06, -2.8348e-04, 4.2841e-06, 5.7407e-06, 4.1336e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 220.64, cls_loss 0.0009 cls_loss_mapping 0.0028 cls_loss_causal 0.5286 re_mapping 0.0052 re_causal 0.0155 /// teacc 98.93 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1367, 0.1740, -0.0277, ..., -0.0828, 0.0260, 0.0058], + [ 0.0057, -0.0975, -0.0504, ..., 0.0235, -0.0320, -0.0534], + [-0.1031, -0.1790, 0.0327, ..., -0.0504, -0.0064, -0.1082], + ..., + [ 0.0662, 0.1081, -0.0799, ..., -0.1151, 0.0643, 0.0665], + [ 0.0458, -0.0936, -0.0181, ..., -0.0577, -0.0655, -0.0606], + [-0.1332, -0.0341, -0.0542, ..., 0.0911, -0.2358, 0.0846]], + device='cuda:0'), grad: tensor([[ 2.0452e-06, -1.8328e-05, 8.3819e-07, ..., 5.0105e-07, + -1.4357e-05, -2.4792e-06], + [ 1.1278e-06, 1.4924e-07, 5.0478e-07, ..., 8.8476e-09, + 8.8336e-07, 1.5926e-07], + [ 2.1309e-06, 1.1310e-05, -5.0291e-08, ..., 1.0710e-07, + 1.0036e-05, 1.6922e-06], + ..., + [ 1.2234e-05, 3.1851e-07, 4.5151e-06, ..., 3.4925e-08, + 6.8098e-06, 7.1898e-07], + [ 3.7942e-06, 5.0478e-07, 1.4966e-06, ..., 6.6822e-08, + 2.6412e-06, 3.8766e-07], + [ 2.7269e-06, 4.7684e-07, 9.8720e-07, ..., -7.6788e-07, + 2.1346e-06, 9.7323e-08]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0212, -0.0275, 0.0168, -0.0266, 0.0366, 0.0201, 0.0057, -0.0093, + -0.0173, -0.0081], device='cuda:0'), grad: tensor([-2.3171e-05, 6.1691e-06, 1.8820e-05, 2.5317e-05, 2.3529e-05, + -1.3208e-04, 3.7625e-06, 4.9233e-05, 1.7196e-05, 1.1094e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 220.37, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.5132 re_mapping 0.0050 re_causal 0.0145 /// teacc 98.95 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1373, 0.1733, -0.0277, ..., -0.0838, 0.0250, 0.0030], + [ 0.0054, -0.0979, -0.0504, ..., 0.0233, -0.0326, -0.0538], + [-0.1033, -0.1799, 0.0330, ..., -0.0506, -0.0061, -0.1083], + ..., + [ 0.0662, 0.1082, -0.0803, ..., -0.1156, 0.0640, 0.0667], + [ 0.0460, -0.0928, -0.0184, ..., -0.0584, -0.0658, -0.0608], + [-0.1337, -0.0317, -0.0541, ..., 0.0913, -0.2343, 0.0857]], + device='cuda:0'), grad: tensor([[ 4.8196e-08, -1.1083e-06, 0.0000e+00, ..., 5.2853e-08, + -3.4785e-07, 1.7346e-07], + [ 1.8859e-08, 4.4843e-07, -4.6566e-10, ..., 2.2585e-08, + 2.6287e-07, 4.5518e-07], + [ 9.0338e-08, 6.5099e-07, 2.3283e-10, ..., 1.9791e-08, + 3.9814e-07, 2.9220e-07], + ..., + [-5.3272e-07, -1.9278e-06, 0.0000e+00, ..., 3.0035e-08, + -6.6403e-07, -2.0359e-06], + [ 5.1456e-08, 1.5809e-07, 0.0000e+00, ..., 4.0047e-08, + 8.3121e-08, 9.9652e-08], + [ 2.8405e-07, 9.2806e-07, 0.0000e+00, ..., -1.7486e-07, + 2.5821e-07, 3.3784e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0194, -0.0278, 0.0169, -0.0267, 0.0368, 0.0201, 0.0055, -0.0095, + -0.0169, -0.0075], device='cuda:0'), grad: tensor([-1.4286e-06, 1.5004e-06, 1.8086e-06, 3.0007e-06, 5.2992e-07, + -1.6447e-06, -5.1782e-07, -5.8748e-06, 5.9046e-07, 2.0340e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 220.19, cls_loss 0.0009 cls_loss_mapping 0.0030 cls_loss_causal 0.5329 re_mapping 0.0047 re_causal 0.0148 /// teacc 98.95 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1378, 0.1735, -0.0278, ..., -0.0840, 0.0250, 0.0029], + [ 0.0051, -0.0982, -0.0504, ..., 0.0232, -0.0331, -0.0546], + [-0.1034, -0.1797, 0.0332, ..., -0.0508, -0.0060, -0.1085], + ..., + [ 0.0672, 0.1098, -0.0806, ..., -0.1157, 0.0645, 0.0681], + [ 0.0459, -0.0934, -0.0185, ..., -0.0589, -0.0663, -0.0609], + [-0.1346, -0.0327, -0.0543, ..., 0.0906, -0.2362, 0.0849]], + device='cuda:0'), grad: tensor([[ 1.5600e-08, -1.0841e-06, 1.1642e-09, ..., 1.8510e-07, + -1.6368e-07, -3.9744e-07], + [ 1.0803e-07, 5.5647e-08, 2.3283e-10, ..., 1.2619e-07, + 1.6345e-07, 1.3597e-07], + [ 3.2596e-08, 1.7276e-07, -1.3970e-09, ..., 9.4296e-08, + 5.7928e-07, 2.9453e-07], + ..., + [-5.8208e-08, -1.8021e-07, 2.3283e-10, ..., 1.3923e-07, + 1.6578e-07, 3.2363e-08], + [ 2.1420e-08, 8.6380e-08, 3.2596e-09, ..., 1.4640e-06, + 1.2172e-06, 1.2107e-06], + [ 2.2072e-07, 4.1584e-07, 0.0000e+00, ..., -4.9826e-07, + 1.4342e-07, -4.9267e-07]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0192, -0.0280, 0.0170, -0.0271, 0.0375, 0.0205, 0.0056, -0.0090, + -0.0171, -0.0085], device='cuda:0'), grad: tensor([-7.7579e-07, 7.2736e-07, 1.8422e-06, -6.4783e-06, -2.3395e-06, + 1.8403e-06, -3.7011e-06, 5.4296e-07, 9.2238e-06, -8.9221e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 220.71, cls_loss 0.0012 cls_loss_mapping 0.0040 cls_loss_causal 0.5264 re_mapping 0.0049 re_causal 0.0146 /// teacc 99.08 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1389, 0.1744, -0.0280, ..., -0.0839, 0.0250, 0.0033], + [ 0.0047, -0.0988, -0.0513, ..., 0.0232, -0.0336, -0.0553], + [-0.1039, -0.1808, 0.0326, ..., -0.0511, -0.0057, -0.1088], + ..., + [ 0.0681, 0.1112, -0.0807, ..., -0.1173, 0.0648, 0.0685], + [ 0.0458, -0.0950, -0.0187, ..., -0.0601, -0.0671, -0.0617], + [-0.1350, -0.0334, -0.0550, ..., 0.0911, -0.2372, 0.0859]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -2.7451e-07, 0.0000e+00, ..., 8.8708e-08, + 5.3551e-09, -2.3749e-08], + [ 1.5832e-08, 1.4435e-08, 0.0000e+00, ..., 7.6368e-08, + 7.2876e-08, 6.0769e-08], + [ 2.7707e-08, 2.6077e-08, -2.3283e-10, ..., 1.7649e-07, + 8.5915e-08, 6.2166e-08], + ..., + [-3.2363e-08, -6.7055e-08, 0.0000e+00, ..., 4.0513e-08, + 1.3737e-08, -5.3318e-08], + [-9.0804e-09, 6.5193e-09, 0.0000e+00, ..., 5.0757e-08, + 2.1188e-08, 2.5611e-08], + [ 3.4412e-07, 2.2841e-07, 0.0000e+00, ..., 8.8941e-07, + 3.5157e-08, 1.6391e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0194, -0.0282, 0.0169, -0.0278, 0.0372, 0.0205, 0.0059, -0.0088, + -0.0173, -0.0080], device='cuda:0'), grad: tensor([-1.6764e-08, 4.4773e-07, 8.5402e-07, -1.0952e-06, -2.2482e-06, + 7.5996e-07, -1.3085e-06, 1.9558e-08, 1.5600e-07, 2.4289e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 220.74, cls_loss 0.0008 cls_loss_mapping 0.0029 cls_loss_causal 0.5213 re_mapping 0.0049 re_causal 0.0146 /// teacc 99.03 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1392, 0.1749, -0.0279, ..., -0.0841, 0.0250, 0.0034], + [ 0.0046, -0.0993, -0.0512, ..., 0.0232, -0.0340, -0.0559], + [-0.1042, -0.1821, 0.0323, ..., -0.0515, -0.0059, -0.1091], + ..., + [ 0.0686, 0.1125, -0.0805, ..., -0.1177, 0.0651, 0.0691], + [ 0.0459, -0.0953, -0.0189, ..., -0.0610, -0.0675, -0.0623], + [-0.1354, -0.0340, -0.0549, ..., 0.0917, -0.2381, 0.0864]], + device='cuda:0'), grad: tensor([[ 8.4285e-08, -2.8638e-08, 0.0000e+00, ..., 1.1176e-07, + 3.4925e-09, 8.2655e-08], + [ 3.1106e-07, 9.3132e-09, 0.0000e+00, ..., 2.9593e-07, + 2.7567e-07, 7.9442e-07], + [ 2.8405e-07, 6.2864e-09, 0.0000e+00, ..., 1.6787e-07, + 4.9127e-08, 7.0082e-08], + ..., + [ 5.1316e-07, -1.4179e-07, 0.0000e+00, ..., 9.9558e-07, + -4.9081e-07, -7.1945e-07], + [ 5.1688e-08, 5.5879e-09, 0.0000e+00, ..., -4.8615e-07, + 3.0734e-08, -8.6799e-07], + [ 5.8766e-07, 1.1083e-07, 0.0000e+00, ..., 1.0827e-07, + 8.3819e-08, 8.6613e-08]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0194, -0.0283, 0.0167, -0.0277, 0.0370, 0.0205, 0.0052, -0.0085, + -0.0172, -0.0076], device='cuda:0'), grad: tensor([ 5.2107e-07, 2.2557e-06, 9.4529e-07, 7.4226e-07, -8.9109e-06, + 1.1921e-06, 1.6913e-06, 2.2110e-06, -9.0078e-06, 8.3148e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 220.81, cls_loss 0.0013 cls_loss_mapping 0.0053 cls_loss_causal 0.5393 re_mapping 0.0050 re_causal 0.0148 /// teacc 99.04 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1397, 0.1754, -0.0281, ..., -0.0843, 0.0250, 0.0034], + [ 0.0043, -0.0998, -0.0513, ..., 0.0231, -0.0342, -0.0564], + [-0.1043, -0.1824, 0.0326, ..., -0.0518, -0.0059, -0.1092], + ..., + [ 0.0689, 0.1113, -0.0806, ..., -0.1215, 0.0623, 0.0666], + [ 0.0460, -0.0957, -0.0189, ..., -0.0635, -0.0684, -0.0640], + [-0.1361, -0.0325, -0.0551, ..., 0.0941, -0.2358, 0.0894]], + device='cuda:0'), grad: tensor([[ 4.8662e-08, -2.2608e-07, -1.1409e-08, ..., 3.1223e-07, + 4.9081e-07, 9.1642e-07], + [ 3.3993e-08, 8.3586e-08, 2.3283e-10, ..., 3.5460e-07, + 1.1064e-06, 2.2072e-06], + [ 8.3586e-08, 1.0058e-07, 3.7253e-09, ..., 2.1677e-07, + 4.8894e-07, 9.4436e-07], + ..., + [ 1.9427e-06, 4.8429e-07, -1.1642e-09, ..., 2.6003e-06, + 2.1216e-06, 4.8801e-06], + [-9.2760e-07, -2.4540e-07, 4.6566e-10, ..., 6.8452e-07, + 1.1176e-07, -1.9325e-07], + [ 7.9647e-06, 1.5292e-06, 2.3283e-09, ..., 1.6049e-05, + 1.2308e-05, 2.1473e-05]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0194, -0.0286, 0.0168, -0.0276, 0.0371, 0.0210, 0.0049, -0.0112, + -0.0178, -0.0050], device='cuda:0'), grad: tensor([ 3.2950e-06, 8.4490e-06, 4.1127e-06, -1.2118e-04, -3.6120e-05, + 1.1817e-05, 5.7090e-07, 2.3127e-05, -4.2357e-06, 1.1027e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 220.33, cls_loss 0.0010 cls_loss_mapping 0.0049 cls_loss_causal 0.5488 re_mapping 0.0048 re_causal 0.0147 /// teacc 99.06 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1404, 0.1784, -0.0281, ..., -0.0838, 0.0259, 0.0047], + [ 0.0030, -0.1000, -0.0514, ..., 0.0231, -0.0360, -0.0576], + [-0.1045, -0.1845, 0.0330, ..., -0.0524, -0.0060, -0.1094], + ..., + [ 0.0706, 0.1113, -0.0806, ..., -0.1217, 0.0624, 0.0665], + [ 0.0476, -0.0977, -0.0191, ..., -0.0641, -0.0690, -0.0641], + [-0.1369, -0.0331, -0.0553, ..., 0.0941, -0.2357, 0.0895]], + device='cuda:0'), grad: tensor([[ 6.9616e-08, -7.8380e-06, 6.2864e-09, ..., -4.5113e-06, + -4.9360e-08, -1.6848e-06], + [-3.3993e-08, 2.4447e-08, 5.8208e-09, ..., -9.3365e-08, + 3.6322e-08, 2.6543e-08], + [ 5.8906e-08, 1.0128e-07, -1.3364e-07, ..., 6.8918e-08, + -3.9581e-09, 4.6799e-08], + ..., + [-1.1199e-07, -8.2655e-08, 3.0035e-08, ..., 5.2620e-08, + -1.6298e-08, -4.5751e-07], + [ 3.1199e-07, 1.3760e-07, 7.2410e-08, ..., 3.8324e-07, + 6.4075e-07, 4.0676e-07], + [ 3.0245e-07, 6.9328e-06, 1.3970e-09, ..., 3.4012e-06, + 2.6333e-07, 1.5656e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0206, -0.0293, 0.0167, -0.0276, 0.0372, 0.0204, 0.0047, -0.0112, + -0.0164, -0.0050], device='cuda:0'), grad: tensor([-1.9506e-05, 1.1642e-07, -2.4848e-06, 3.0883e-06, 1.4128e-06, + 5.5581e-06, -1.0729e-05, -4.2608e-08, 4.7833e-06, 1.7792e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 220.09, cls_loss 0.0008 cls_loss_mapping 0.0031 cls_loss_causal 0.4763 re_mapping 0.0051 re_causal 0.0143 /// teacc 98.99 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1414, 0.1788, -0.0282, ..., -0.0839, 0.0260, 0.0051], + [ 0.0026, -0.0998, -0.0516, ..., 0.0227, -0.0361, -0.0576], + [-0.1046, -0.1849, 0.0331, ..., -0.0529, -0.0059, -0.1096], + ..., + [ 0.0707, 0.1114, -0.0805, ..., -0.1217, 0.0624, 0.0665], + [ 0.0480, -0.0958, -0.0192, ..., -0.0645, -0.0695, -0.0642], + [-0.1377, -0.0334, -0.0553, ..., 0.0940, -0.2358, 0.0895]], + device='cuda:0'), grad: tensor([[ 1.9721e-07, 7.0548e-08, 2.0489e-08, ..., 1.6298e-08, + 8.2655e-08, 1.4203e-07], + [ 9.7230e-07, 5.1828e-07, 1.2713e-07, ..., 1.2922e-07, + 3.2829e-08, 5.1782e-07], + [ 5.1456e-07, 1.8254e-07, -2.7474e-07, ..., 2.6543e-08, + 3.5344e-07, 3.4785e-07], + ..., + [-4.9293e-05, -1.8403e-06, 2.2352e-08, ..., 4.1211e-08, + -1.7315e-05, -2.4244e-05], + [ 6.4261e-08, 3.3760e-08, 2.1653e-08, ..., 2.1653e-08, + 7.3574e-08, 7.9395e-08], + [ 1.6484e-06, 8.2329e-07, 4.4238e-09, ..., -1.3947e-07, + 4.8056e-07, 1.1465e-06]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0202, -0.0293, 0.0167, -0.0277, 0.0376, 0.0203, 0.0049, -0.0112, + -0.0157, -0.0052], device='cuda:0'), grad: tensor([ 7.1432e-07, 7.9907e-07, 1.6093e-06, 1.3337e-06, 7.7307e-05, + 1.0394e-05, 4.2096e-07, -9.8228e-05, 5.2154e-07, 5.1372e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 220.05, cls_loss 0.0010 cls_loss_mapping 0.0030 cls_loss_causal 0.4839 re_mapping 0.0049 re_causal 0.0137 /// teacc 99.09 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.1422, 0.1798, -0.0283, ..., -0.0838, 0.0264, 0.0055], + [ 0.0033, -0.0993, -0.0532, ..., 0.0237, -0.0362, -0.0577], + [-0.1050, -0.1855, 0.0348, ..., -0.0531, -0.0059, -0.1098], + ..., + [ 0.0713, 0.1114, -0.0804, ..., -0.1217, 0.0626, 0.0666], + [ 0.0480, -0.0964, -0.0195, ..., -0.0650, -0.0702, -0.0646], + [-0.1382, -0.0337, -0.0555, ..., 0.0939, -0.2359, 0.0894]], + device='cuda:0'), grad: tensor([[ 8.6147e-09, -5.2512e-05, 2.3283e-09, ..., -4.1761e-06, + -2.2933e-05, -2.5585e-05], + [-1.3201e-07, 7.9162e-07, 2.3283e-10, ..., -3.0035e-08, + 3.8277e-07, 4.9639e-07], + [ 3.5390e-08, 2.7046e-06, 4.8894e-09, ..., 5.4762e-07, + 1.4789e-06, 1.5497e-06], + ..., + [ 2.6543e-08, 2.5593e-06, 4.6566e-10, ..., 4.2398e-07, + 1.0999e-06, 1.4901e-06], + [ 1.8394e-08, 1.5721e-06, 2.3283e-10, ..., 2.5821e-07, + 8.8662e-07, 7.9256e-07], + [ 3.8650e-08, 1.6496e-05, -2.7940e-08, ..., -1.0552e-06, + 5.9642e-06, 5.6624e-06]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0205, -0.0294, 0.0171, -0.0308, 0.0376, 0.0230, 0.0046, -0.0111, + -0.0157, -0.0053], device='cuda:0'), grad: tensor([-1.1241e-04, 1.4706e-06, 7.4171e-06, 1.3366e-05, 5.2080e-06, + 3.7491e-05, 1.1563e-05, 6.1467e-06, 4.0494e-06, 2.5615e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 220.20, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.5208 re_mapping 0.0047 re_causal 0.0137 /// teacc 99.03 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1429, 0.1809, -0.0275, ..., -0.0839, 0.0271, 0.0059], + [ 0.0033, -0.0994, -0.0532, ..., 0.0242, -0.0364, -0.0578], + [-0.1057, -0.1860, 0.0350, ..., -0.0536, -0.0062, -0.1102], + ..., + [ 0.0720, 0.1115, -0.0804, ..., -0.1218, 0.0627, 0.0667], + [ 0.0477, -0.0968, -0.0201, ..., -0.0652, -0.0707, -0.0650], + [-0.1387, -0.0339, -0.0560, ..., 0.0939, -0.2360, 0.0894]], + device='cuda:0'), grad: tensor([[ 9.1968e-08, 5.5879e-09, 7.5903e-08, ..., 2.3190e-07, + 1.2561e-07, 2.3050e-08], + [ 2.7684e-07, 1.5367e-08, 2.6776e-08, ..., 1.5385e-06, + 7.0781e-08, 4.4471e-08], + [-3.1665e-07, -1.7113e-07, -4.6194e-07, ..., 1.9688e-06, + -2.5928e-06, 2.2468e-08], + ..., + [ 9.9186e-08, -1.0000e-07, 1.4552e-08, ..., 1.7730e-07, + 3.0780e-07, -1.9604e-07], + [ 4.1118e-07, 1.4307e-07, 9.4296e-09, ..., 4.7078e-07, + 1.6633e-06, 1.0815e-07], + [ 1.6880e-07, 7.1712e-08, 5.8208e-10, ..., -2.7791e-06, + 7.9162e-08, -1.3933e-06]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0208, -0.0292, 0.0168, -0.0308, 0.0375, 0.0233, 0.0040, -0.0111, + -0.0160, -0.0054], device='cuda:0'), grad: tensor([ 2.4531e-06, 7.4022e-06, -1.2301e-05, 5.6699e-06, 8.4192e-06, + 3.7607e-06, -2.2978e-05, 3.0212e-06, 8.2925e-06, -3.7607e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 220.36, cls_loss 0.0009 cls_loss_mapping 0.0039 cls_loss_causal 0.5250 re_mapping 0.0049 re_causal 0.0144 /// teacc 98.91 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.1437, 0.1814, -0.0281, ..., -0.0841, 0.0271, 0.0061], + [ 0.0047, -0.0993, -0.0519, ..., 0.0255, -0.0362, -0.0572], + [-0.1061, -0.1863, 0.0353, ..., -0.0542, -0.0062, -0.1103], + ..., + [ 0.0715, 0.1115, -0.0818, ..., -0.1219, 0.0627, 0.0667], + [ 0.0477, -0.0971, -0.0203, ..., -0.0655, -0.0713, -0.0654], + [-0.1395, -0.0340, -0.0573, ..., 0.0940, -0.2361, 0.0894]], + device='cuda:0'), grad: tensor([[ 7.1945e-08, -2.3679e-07, 0.0000e+00, ..., 6.8918e-08, + 2.0955e-09, -6.5193e-09], + [-2.3283e-09, 4.0047e-08, 0.0000e+00, ..., -1.2061e-07, + 3.1246e-07, 5.0943e-07], + [ 4.7730e-08, 2.4680e-08, 0.0000e+00, ..., 5.5181e-08, + 9.7230e-07, 2.2701e-07], + ..., + [ 3.1828e-07, -1.8463e-07, 0.0000e+00, ..., 6.6357e-08, + 1.4389e-07, 9.7556e-08], + [ 2.0186e-07, 3.2596e-08, 0.0000e+00, ..., 2.3283e-08, + 1.4040e-07, 1.8417e-07], + [ 3.2852e-07, 2.4098e-07, 0.0000e+00, ..., -2.9872e-07, + 2.1094e-07, 1.3364e-07]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0206, -0.0288, 0.0171, -0.0305, 0.0374, 0.0223, 0.0057, -0.0111, + -0.0160, -0.0055], device='cuda:0'), grad: tensor([ 7.8510e-07, 2.0172e-06, -7.5102e-05, 5.8208e-08, 2.7586e-06, + -3.7532e-06, 4.1351e-07, 6.3360e-05, 6.8992e-06, 2.7027e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 220.11, cls_loss 0.0010 cls_loss_mapping 0.0047 cls_loss_causal 0.5385 re_mapping 0.0047 re_causal 0.0144 /// teacc 99.05 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1442, 0.1847, -0.0282, ..., -0.0842, 0.0268, 0.0090], + [ 0.0051, -0.0996, -0.0520, ..., 0.0261, -0.0360, -0.0571], + [-0.1064, -0.1868, 0.0354, ..., -0.0544, -0.0062, -0.1106], + ..., + [ 0.0719, 0.1116, -0.0818, ..., -0.1220, 0.0628, 0.0668], + [ 0.0479, -0.0982, -0.0201, ..., -0.0664, -0.0719, -0.0666], + [-0.1402, -0.0370, -0.0574, ..., 0.0938, -0.2363, 0.0888]], + device='cuda:0'), grad: tensor([[ 8.7544e-08, -7.5204e-07, 0.0000e+00, ..., 2.8126e-07, + -1.3504e-07, -5.1688e-08], + [ 1.0617e-06, 1.2023e-06, 0.0000e+00, ..., -1.1758e-07, + 2.3050e-07, 1.2089e-06], + [ 2.1188e-07, 4.3190e-07, 0.0000e+00, ..., 2.9290e-07, + 1.6135e-07, 4.9453e-07], + ..., + [-2.4159e-06, -2.5369e-06, 0.0000e+00, ..., 5.8906e-08, + -4.1095e-07, -2.8815e-06], + [ 2.3516e-08, 6.0769e-08, 0.0000e+00, ..., 7.5903e-07, + 2.7358e-07, 1.7323e-07], + [ 4.0629e-07, 6.4820e-07, 0.0000e+00, ..., 4.1910e-08, + 1.6368e-07, 4.3935e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0234, -0.0285, 0.0172, -0.0306, 0.0377, 0.0220, 0.0062, -0.0111, + -0.0158, -0.0067], device='cuda:0'), grad: tensor([-1.6857e-07, 4.2468e-06, 2.7940e-06, -7.2002e-05, 4.0159e-06, + 7.3135e-05, -7.4990e-06, -1.0230e-05, 2.9020e-06, 2.6487e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 220.57, cls_loss 0.0010 cls_loss_mapping 0.0038 cls_loss_causal 0.5246 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.04 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1452, 0.1845, -0.0284, ..., -0.0858, 0.0266, 0.0085], + [ 0.0054, -0.1002, -0.0520, ..., 0.0268, -0.0363, -0.0573], + [-0.1069, -0.1874, 0.0353, ..., -0.0552, -0.0064, -0.1110], + ..., + [ 0.0727, 0.1120, -0.0818, ..., -0.1220, 0.0632, 0.0669], + [ 0.0482, -0.0982, -0.0199, ..., -0.0664, -0.0725, -0.0667], + [-0.1408, -0.0367, -0.0568, ..., 0.0939, -0.2364, 0.0889]], + device='cuda:0'), grad: tensor([[ 3.9581e-09, -4.8429e-08, 0.0000e+00, ..., 1.0943e-08, + 7.2177e-09, 2.0722e-08], + [-6.1700e-08, 3.0268e-09, 0.0000e+00, ..., -5.1688e-08, + 6.7521e-08, 1.0268e-07], + [ 6.1467e-08, 7.6834e-09, 0.0000e+00, ..., 2.7940e-08, + 8.6147e-08, 1.2643e-07], + ..., + [-1.0128e-07, -1.0943e-08, 0.0000e+00, ..., 4.4936e-08, + 2.7008e-08, -7.6601e-08], + [ 1.2573e-08, 4.6566e-09, 0.0000e+00, ..., 1.5832e-08, + 6.9384e-08, 9.5693e-08], + [ 5.3318e-08, 2.2352e-08, 0.0000e+00, ..., -8.1491e-08, + 7.4040e-08, -3.0268e-09]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0229, -0.0284, 0.0170, -0.0306, 0.0376, 0.0215, 0.0062, -0.0109, + -0.0140, -0.0066], device='cuda:0'), grad: tensor([ 8.1025e-08, -4.3074e-08, 2.8219e-07, -1.3538e-05, 1.2037e-07, + 1.2681e-05, -7.5670e-08, -2.0023e-08, 3.8301e-07, 1.5437e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 220.36, cls_loss 0.0011 cls_loss_mapping 0.0030 cls_loss_causal 0.5038 re_mapping 0.0047 re_causal 0.0137 /// teacc 99.09 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1464, 0.1848, -0.0285, ..., -0.0867, 0.0285, 0.0086], + [ 0.0052, -0.1010, -0.0521, ..., 0.0268, -0.0369, -0.0586], + [-0.1076, -0.1889, 0.0356, ..., -0.0557, -0.0069, -0.1122], + ..., + [ 0.0721, 0.1121, -0.0820, ..., -0.1222, 0.0631, 0.0670], + [ 0.0474, -0.0988, -0.0199, ..., -0.0677, -0.0738, -0.0677], + [-0.1412, -0.0366, -0.0569, ..., 0.0941, -0.2364, 0.0890]], + device='cuda:0'), grad: tensor([[-2.3749e-08, -3.5297e-07, 0.0000e+00, ..., 2.6077e-08, + 6.3097e-08, -6.6590e-08], + [-1.2338e-04, -8.2701e-06, 0.0000e+00, ..., -2.8871e-08, + 2.0792e-07, -8.6427e-05], + [ 1.4226e-07, 1.5367e-08, 0.0000e+00, ..., 1.6764e-08, + 9.1642e-07, 7.9349e-07], + ..., + [ 1.1915e-04, 8.0094e-06, 0.0000e+00, ..., 1.6065e-08, + 5.0291e-07, 8.4043e-05], + [ 2.4494e-06, 2.5122e-07, 0.0000e+00, ..., -1.5856e-07, + 1.0687e-07, 2.0936e-06], + [ 4.6776e-07, 5.3551e-08, 0.0000e+00, ..., 3.7951e-08, + 3.3365e-07, 6.0070e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0230, -0.0291, 0.0170, -0.0302, 0.0377, 0.0207, 0.0075, -0.0109, + -0.0150, -0.0065], device='cuda:0'), grad: tensor([-2.9989e-07, -3.1137e-04, -3.5320e-07, -6.7428e-06, 6.0815e-07, + 3.3900e-06, -8.6147e-09, 3.0541e-04, 6.8098e-06, 2.2799e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 220.03, cls_loss 0.0011 cls_loss_mapping 0.0029 cls_loss_causal 0.5304 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.11 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1481, 0.1849, -0.0285, ..., -0.0868, 0.0284, 0.0086], + [ 0.0018, -0.1010, -0.0536, ..., 0.0265, -0.0388, -0.0623], + [-0.1075, -0.1893, 0.0370, ..., -0.0556, -0.0072, -0.1129], + ..., + [ 0.0756, 0.1121, -0.0821, ..., -0.1222, 0.0634, 0.0677], + [ 0.0470, -0.0992, -0.0200, ..., -0.0683, -0.0753, -0.0685], + [-0.1417, -0.0367, -0.0569, ..., 0.0942, -0.2366, 0.0890]], + device='cuda:0'), grad: tensor([[ 1.0943e-07, -6.7661e-07, 0.0000e+00, ..., 1.5320e-07, + 3.2829e-08, 6.8918e-08], + [-8.5123e-07, 3.9814e-08, 0.0000e+00, ..., -6.2026e-07, + 2.4983e-07, 2.5448e-07], + [ 3.2596e-09, 2.2701e-07, 0.0000e+00, ..., 3.7509e-07, + 3.8464e-07, 4.0489e-07], + ..., + [ 6.1374e-07, 3.7253e-09, 0.0000e+00, ..., 4.5006e-07, + 4.5821e-07, 4.3726e-07], + [ 2.2817e-07, 4.1211e-08, 0.0000e+00, ..., 1.1176e-07, + 5.1921e-07, 4.1933e-07], + [ 4.2492e-07, 1.7369e-07, 0.0000e+00, ..., 5.4203e-07, + 2.2585e-07, 5.4715e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0229, -0.0330, 0.0174, -0.0296, 0.0375, 0.0205, 0.0079, -0.0095, + -0.0155, -0.0066], device='cuda:0'), grad: tensor([ 5.5367e-07, 1.5777e-06, -1.0125e-05, -6.2734e-06, -1.8189e-06, + 4.8336e-07, 4.3539e-07, 7.4543e-06, 2.7195e-06, 4.9509e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 220.12, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5157 re_mapping 0.0047 re_causal 0.0138 /// teacc 99.10 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1491, 0.1851, -0.0285, ..., -0.0864, 0.0284, 0.0087], + [ 0.0020, -0.1011, -0.0536, ..., 0.0266, -0.0391, -0.0624], + [-0.1078, -0.1899, 0.0371, ..., -0.0558, -0.0073, -0.1131], + ..., + [ 0.0765, 0.1122, -0.0824, ..., -0.1222, 0.0638, 0.0678], + [ 0.0469, -0.0997, -0.0200, ..., -0.0688, -0.0764, -0.0692], + [-0.1416, -0.0368, -0.0570, ..., 0.0942, -0.2368, 0.0892]], + device='cuda:0'), grad: tensor([[ 1.7462e-08, -2.6659e-07, 0.0000e+00, ..., 7.8231e-08, + 1.2107e-08, 1.3039e-08], + [ 1.1595e-07, 7.9162e-09, 0.0000e+00, ..., 4.9127e-08, + 6.4727e-08, 6.2631e-08], + [ 4.1211e-08, 2.2585e-08, 0.0000e+00, ..., 2.9802e-08, + 3.1199e-08, 2.2352e-08], + ..., + [ 3.5623e-08, 7.4506e-09, 0.0000e+00, ..., 1.1432e-07, + 2.3562e-07, 3.2433e-07], + [-2.3167e-07, 5.0291e-08, 0.0000e+00, ..., -9.5461e-09, + 3.4692e-08, 5.5647e-08], + [ 1.0175e-07, 9.0338e-08, 0.0000e+00, ..., -3.7509e-07, + 2.0186e-07, -1.4668e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0231, -0.0330, 0.0175, -0.0297, 0.0367, 0.0207, 0.0078, -0.0093, + -0.0156, -0.0065], device='cuda:0'), grad: tensor([-2.2282e-07, 5.2806e-07, 2.5705e-07, -4.7944e-06, 3.7486e-07, + 2.0564e-06, 1.2200e-06, 8.6613e-07, -4.0373e-07, 8.5449e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 220.50, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.4927 re_mapping 0.0045 re_causal 0.0136 /// teacc 99.05 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1499, 0.1852, -0.0285, ..., -0.0865, 0.0285, 0.0088], + [ 0.0034, -0.1013, -0.0521, ..., 0.0282, -0.0379, -0.0610], + [-0.1080, -0.1903, 0.0373, ..., -0.0559, -0.0073, -0.1133], + ..., + [ 0.0755, 0.1123, -0.0840, ..., -0.1234, 0.0632, 0.0671], + [ 0.0471, -0.1000, -0.0200, ..., -0.0688, -0.0770, -0.0694], + [-0.1425, -0.0368, -0.0584, ..., 0.0944, -0.2372, 0.0890]], + device='cuda:0'), grad: tensor([[ 8.1258e-07, 1.0058e-07, 3.9116e-08, ..., 1.4598e-07, + 5.7183e-07, 7.7626e-07], + [ 2.8405e-07, 7.6136e-08, 2.7940e-09, ..., -1.2852e-07, + 2.5565e-07, 3.2689e-07], + [ 2.0675e-07, 5.0291e-08, 4.4238e-09, ..., 5.6578e-08, + 1.3318e-07, 1.8231e-07], + ..., + [-2.7586e-06, -6.5612e-07, 2.3283e-10, ..., 2.7474e-08, + -2.0452e-06, -2.8685e-06], + [-1.1851e-07, 1.7229e-08, 4.4936e-08, ..., 1.1921e-07, + 8.7079e-08, 7.1479e-08], + [ 5.6252e-07, 1.4808e-07, 3.9581e-09, ..., 6.2864e-09, + 3.9465e-07, 5.0198e-07]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0231, -0.0316, 0.0177, -0.0282, 0.0364, 0.0193, 0.0078, -0.0103, + -0.0154, -0.0067], device='cuda:0'), grad: tensor([ 3.9004e-06, 1.2163e-06, 7.8976e-07, 8.9779e-07, 4.0196e-06, + -6.0257e-07, -1.6508e-07, -1.2554e-05, -1.0966e-07, 2.5667e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 220.85, cls_loss 0.0013 cls_loss_mapping 0.0044 cls_loss_causal 0.5083 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.00 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1524, 0.1852, -0.0287, ..., -0.0876, 0.0284, 0.0085], + [ 0.0049, -0.1017, -0.0504, ..., 0.0297, -0.0363, -0.0594], + [-0.1089, -0.1906, 0.0373, ..., -0.0599, -0.0075, -0.1137], + ..., + [ 0.0740, 0.1126, -0.0857, ..., -0.1247, 0.0620, 0.0662], + [ 0.0475, -0.1003, -0.0191, ..., -0.0700, -0.0780, -0.0702], + [-0.1442, -0.0368, -0.0596, ..., 0.0945, -0.2374, 0.0891]], + device='cuda:0'), grad: tensor([[ 2.0466e-07, -2.4680e-06, 1.6298e-09, ..., 4.8429e-08, + 1.6764e-08, -8.2236e-07], + [ 1.4687e-04, 1.6810e-07, 2.3283e-10, ..., 1.8597e-05, + 1.0571e-07, 1.9255e-07], + [ 4.2911e-07, 1.4366e-07, -2.1886e-08, ..., 7.0315e-08, + 1.4389e-07, 2.6869e-07], + ..., + [-1.1572e-07, 4.4936e-08, 2.5611e-09, ..., 7.1479e-08, + 9.9652e-08, 7.7998e-08], + [ 7.1488e-06, 9.5926e-08, 2.3283e-09, ..., 9.7416e-07, + 4.6799e-08, 1.1199e-07], + [ 2.1793e-06, 1.2238e-06, 2.3283e-10, ..., -1.2899e-07, + 5.5647e-08, 2.5518e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0228, -0.0302, 0.0171, -0.0279, 0.0368, 0.0193, 0.0070, -0.0115, + -0.0148, -0.0066], device='cuda:0'), grad: tensor([-3.3267e-06, 3.4380e-04, 1.6931e-06, 4.9502e-05, 8.9221e-07, + -4.2343e-04, 7.0184e-06, 4.2957e-07, 1.6943e-05, 6.3814e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 220.09, cls_loss 0.0008 cls_loss_mapping 0.0037 cls_loss_causal 0.5026 re_mapping 0.0049 re_causal 0.0146 /// teacc 99.03 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1535, 0.1850, -0.0288, ..., -0.0884, 0.0284, 0.0081], + [ 0.0048, -0.1018, -0.0504, ..., 0.0298, -0.0364, -0.0595], + [-0.1097, -0.1920, 0.0373, ..., -0.0616, -0.0078, -0.1139], + ..., + [ 0.0740, 0.1140, -0.0857, ..., -0.1247, 0.0621, 0.0667], + [ 0.0475, -0.1007, -0.0189, ..., -0.0708, -0.0785, -0.0708], + [-0.1447, -0.0370, -0.0595, ..., 0.0946, -0.2374, 0.0887]], + device='cuda:0'), grad: tensor([[ 1.0012e-08, -1.9209e-07, 4.6566e-10, ..., 2.0023e-08, + 5.7509e-08, -7.7300e-08], + [ 4.4238e-09, 3.7253e-09, 2.3283e-10, ..., 5.5879e-09, + 2.8173e-08, 1.6065e-08], + [ 6.0536e-09, 3.0268e-09, -1.4435e-08, ..., 2.2352e-08, + -8.7637e-07, 2.3749e-08], + ..., + [-6.0769e-08, -2.0955e-09, 9.3132e-10, ..., 6.7521e-09, + 3.9162e-07, -2.3120e-07], + [ 4.2841e-08, 3.2596e-09, 5.5879e-09, ..., 6.1234e-08, + 2.7171e-07, 5.0291e-08], + [ 9.0804e-08, 1.6950e-07, 2.3283e-10, ..., -2.4214e-07, + 1.1991e-07, 1.5600e-08]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0224, -0.0302, 0.0166, -0.0280, 0.0369, 0.0200, 0.0062, -0.0112, + -0.0149, -0.0070], device='cuda:0'), grad: tensor([ 1.3015e-07, 2.3330e-07, -5.4538e-06, 1.2405e-06, 5.5972e-07, + -1.3756e-06, 9.4529e-08, 2.5444e-06, 1.8245e-06, 2.0256e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 219.78, cls_loss 0.0010 cls_loss_mapping 0.0032 cls_loss_causal 0.5114 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.13 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1543, 0.1852, -0.0289, ..., -0.0886, 0.0282, 0.0082], + [ 0.0049, -0.1022, -0.0504, ..., 0.0298, -0.0364, -0.0594], + [-0.1100, -0.1924, 0.0374, ..., -0.0621, -0.0077, -0.1142], + ..., + [ 0.0740, 0.1141, -0.0857, ..., -0.1248, 0.0621, 0.0667], + [ 0.0472, -0.1008, -0.0191, ..., -0.0713, -0.0792, -0.0713], + [-0.1452, -0.0371, -0.0595, ..., 0.0953, -0.2375, 0.0889]], + device='cuda:0'), grad: tensor([[ 7.4040e-08, -1.4203e-07, 4.7032e-08, ..., 2.2212e-07, + -1.1642e-08, -3.6787e-08], + [ 7.1479e-07, 7.9162e-09, -8.4750e-08, ..., 4.6752e-07, + 5.3085e-08, 6.3330e-08], + [ 2.6124e-07, 1.1176e-08, 4.3772e-08, ..., 1.5134e-07, + 1.8300e-07, 2.9523e-07], + ..., + [ 1.6615e-06, 1.1176e-08, 7.8231e-08, ..., 8.7637e-07, + 2.2165e-07, -2.4866e-07], + [ 2.2110e-06, 1.4435e-08, 4.6752e-07, ..., 1.4296e-06, + 8.5123e-07, -3.0734e-08], + [ 5.0152e-07, 5.4482e-08, 1.0710e-08, ..., 5.1782e-07, + 3.3062e-08, 1.9837e-07]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0224, -0.0301, 0.0172, -0.0281, 0.0360, 0.0203, 0.0062, -0.0114, + -0.0151, -0.0067], device='cuda:0'), grad: tensor([ 6.7241e-07, -4.9174e-07, 2.6487e-06, 8.7544e-08, -8.4043e-06, + -1.5706e-05, 7.7412e-06, 4.1351e-06, 5.6215e-06, 3.6992e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 220.54, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.5011 re_mapping 0.0046 re_causal 0.0139 /// teacc 99.04 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1553, 0.1853, -0.0290, ..., -0.0888, 0.0281, 0.0082], + [ 0.0050, -0.1022, -0.0504, ..., 0.0299, -0.0364, -0.0594], + [-0.1106, -0.1930, 0.0376, ..., -0.0624, -0.0079, -0.1146], + ..., + [ 0.0739, 0.1142, -0.0857, ..., -0.1249, 0.0621, 0.0667], + [ 0.0469, -0.1013, -0.0194, ..., -0.0724, -0.0799, -0.0720], + [-0.1439, -0.0371, -0.0595, ..., 0.0966, -0.2376, 0.0893]], + device='cuda:0'), grad: tensor([[ 1.3039e-07, -2.4680e-08, 9.3132e-10, ..., 8.6147e-08, + 2.5146e-08, 1.3970e-09], + [ 1.4901e-07, 7.9162e-09, 5.1223e-09, ..., 2.2817e-08, + 9.7323e-08, 2.1886e-07], + [ 2.0862e-07, 8.3819e-09, 7.4506e-09, ..., 1.4203e-07, + 4.8429e-08, 5.5414e-08], + ..., + [ 4.8894e-08, -2.4214e-08, -6.8918e-08, ..., 6.8452e-08, + -9.3132e-10, -5.2247e-07], + [-9.7789e-07, 1.8626e-09, 1.3970e-09, ..., 8.9873e-08, + 1.7183e-07, 2.3283e-08], + [ 5.2247e-07, 2.1886e-08, 1.6298e-08, ..., -2.9337e-08, + 8.0094e-08, 7.1246e-08]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0224, -0.0300, 0.0170, -0.0280, 0.0344, 0.0204, 0.0060, -0.0114, + -0.0154, -0.0058], device='cuda:0'), grad: tensor([ 1.2908e-06, 1.2806e-06, 1.8291e-06, 5.5619e-06, 4.2617e-06, + -4.0680e-05, 3.2097e-05, 9.9931e-07, -1.0923e-05, 4.2841e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 220.31, cls_loss 0.0009 cls_loss_mapping 0.0030 cls_loss_causal 0.5024 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.02 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.1565, 0.1853, -0.0291, ..., -0.0897, 0.0270, 0.0079], + [ 0.0049, -0.1026, -0.0504, ..., 0.0299, -0.0366, -0.0595], + [-0.1102, -0.1934, 0.0381, ..., -0.0631, -0.0080, -0.1149], + ..., + [ 0.0740, 0.1144, -0.0857, ..., -0.1249, 0.0623, 0.0667], + [ 0.0468, -0.1019, -0.0196, ..., -0.0726, -0.0811, -0.0725], + [-0.1447, -0.0371, -0.0597, ..., 0.0966, -0.2377, 0.0893]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -2.5611e-08, 0.0000e+00, ..., 3.6787e-08, + 0.0000e+00, -5.1223e-09], + [ 2.7940e-09, 7.4506e-09, -4.6566e-10, ..., -2.0023e-08, + 2.5611e-08, 3.0268e-08], + [ 7.1712e-08, 4.6566e-09, -4.6566e-10, ..., 9.7789e-09, + 6.1933e-08, 4.1910e-08], + ..., + [-1.4901e-08, -3.1665e-08, 0.0000e+00, ..., 1.3039e-08, + -5.7276e-08, -1.2200e-07], + [-4.3400e-07, 5.5879e-09, 0.0000e+00, ..., 1.5832e-08, + 6.9849e-09, 5.1223e-09], + [ 1.4156e-07, 4.7032e-08, 0.0000e+00, ..., 3.6322e-08, + 6.4727e-08, 1.1409e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0220, -0.0300, 0.0171, -0.0279, 0.0346, 0.0207, 0.0054, -0.0114, + -0.0154, -0.0059], device='cuda:0'), grad: tensor([ 1.6624e-07, 1.8766e-07, 6.2073e-07, 2.2212e-07, 1.5367e-08, + 9.4017e-07, 2.0023e-08, -4.2375e-08, -2.9076e-06, 7.9395e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 220.56, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4897 re_mapping 0.0046 re_causal 0.0137 /// teacc 99.05 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1571, 0.1854, -0.0292, ..., -0.0898, 0.0267, 0.0079], + [ 0.0049, -0.1029, -0.0504, ..., 0.0299, -0.0366, -0.0595], + [-0.1105, -0.1939, 0.0381, ..., -0.0633, -0.0081, -0.1151], + ..., + [ 0.0741, 0.1145, -0.0857, ..., -0.1250, 0.0624, 0.0668], + [ 0.0469, -0.1023, -0.0199, ..., -0.0729, -0.0818, -0.0730], + [-0.1449, -0.0372, -0.0597, ..., 0.0967, -0.2378, 0.0894]], + device='cuda:0'), grad: tensor([[ 4.4843e-07, 1.8161e-08, 1.9558e-08, ..., 1.8515e-06, + 3.0734e-08, 2.3702e-07], + [ 2.9104e-07, 4.5169e-08, 2.7940e-09, ..., 1.0990e-06, + 2.6636e-07, 2.8545e-07], + [ 1.4575e-07, 1.8161e-08, -2.6077e-08, ..., 6.1095e-07, + -8.2050e-07, 8.0559e-08], + ..., + [ 9.9465e-07, -2.3842e-07, 2.3283e-09, ..., 5.7407e-06, + 1.7975e-07, 3.9022e-07], + [ 4.5635e-06, 3.6787e-08, 5.5879e-09, ..., 1.7166e-05, + 9.6858e-08, 1.2796e-06], + [-1.3542e-04, -2.2771e-07, 1.3970e-09, ..., -5.2309e-04, + 5.5414e-08, -3.7730e-05]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0220, -0.0300, 0.0170, -0.0279, 0.0345, 0.0206, 0.0055, -0.0113, + -0.0154, -0.0059], device='cuda:0'), grad: tensor([ 4.4927e-06, 3.7178e-06, -1.6680e-06, -9.8906e-07, 1.1301e-03, + 7.2233e-06, 8.0690e-06, 1.3113e-05, 4.0084e-05, -1.2054e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 220.19, cls_loss 0.0012 cls_loss_mapping 0.0043 cls_loss_causal 0.5225 re_mapping 0.0044 re_causal 0.0133 /// teacc 99.06 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1583, 0.1859, -0.0312, ..., -0.0894, 0.0268, 0.0078], + [ 0.0049, -0.1047, -0.0505, ..., 0.0297, -0.0367, -0.0596], + [-0.1108, -0.1964, 0.0390, ..., -0.0645, -0.0083, -0.1156], + ..., + [ 0.0741, 0.1146, -0.0857, ..., -0.1250, 0.0624, 0.0669], + [ 0.0470, -0.1035, -0.0193, ..., -0.0734, -0.0817, -0.0732], + [-0.1460, -0.0373, -0.0589, ..., 0.0973, -0.2380, 0.0894]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 3.9954e-07, + 3.2596e-09, 6.7521e-08], + [-4.7963e-08, 1.3970e-09, 0.0000e+00, ..., -4.5169e-08, + 8.8476e-09, 5.0757e-08], + [ 3.1199e-08, 9.3132e-10, 0.0000e+00, ..., 8.4285e-08, + 1.3504e-08, 2.2817e-08], + ..., + [-3.8184e-08, -8.8476e-09, 0.0000e+00, ..., 2.6729e-07, + -3.3528e-08, 1.5041e-07], + [-1.3970e-08, 9.3132e-10, 0.0000e+00, ..., 1.5181e-07, + 4.6566e-09, 5.4948e-08], + [ 3.1199e-08, 3.7253e-09, 0.0000e+00, ..., -3.1665e-06, + 2.7008e-08, -1.4938e-06]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0223, -0.0302, 0.0170, -0.0270, 0.0345, 0.0201, 0.0046, -0.0113, + -0.0148, -0.0058], device='cuda:0'), grad: tensor([ 1.2219e-06, -7.8231e-08, 4.6985e-07, 3.2829e-07, 5.2005e-06, + 3.6042e-07, -1.1846e-06, 5.2899e-07, -1.6373e-06, -5.2229e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 220.58, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4951 re_mapping 0.0046 re_causal 0.0137 /// teacc 99.03 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1592, 0.1861, -0.0312, ..., -0.0895, 0.0268, 0.0078], + [ 0.0049, -0.1048, -0.0505, ..., 0.0297, -0.0367, -0.0595], + [-0.1121, -0.1974, 0.0386, ..., -0.0660, -0.0083, -0.1162], + ..., + [ 0.0741, 0.1147, -0.0857, ..., -0.1251, 0.0625, 0.0669], + [ 0.0479, -0.1041, -0.0197, ..., -0.0734, -0.0819, -0.0735], + [-0.1465, -0.0374, -0.0589, ..., 0.0973, -0.2381, 0.0894]], + device='cuda:0'), grad: tensor([[ 9.7789e-08, -6.6170e-07, 8.3819e-09, ..., 1.0245e-07, + 9.3132e-10, -8.6147e-08], + [-2.2296e-06, 2.0489e-08, 4.7963e-08, ..., -2.2966e-06, + 2.7940e-09, 1.1642e-08], + [ 2.1234e-07, 5.0757e-08, -1.0347e-06, ..., 2.1467e-07, + 3.2596e-09, 2.2352e-08], + ..., + [ 9.6858e-08, 3.4459e-08, 8.4890e-07, ..., 1.5972e-07, + -1.8626e-09, -9.2667e-08], + [ 3.5251e-07, 3.1665e-08, 6.4261e-08, ..., 2.7614e-07, + 2.2817e-08, 9.7789e-09], + [ 1.0123e-06, 4.2655e-07, 1.3970e-09, ..., 1.5069e-06, + 1.2107e-08, -6.3377e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0224, -0.0301, 0.0165, -0.0271, 0.0346, 0.0200, 0.0047, -0.0113, + -0.0140, -0.0059], device='cuda:0'), grad: tensor([-6.5798e-07, -9.4250e-06, -6.7092e-06, 4.0699e-07, -3.7011e-06, + 7.0035e-07, 6.4820e-06, 7.0371e-06, 1.6764e-06, 4.1835e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 220.40, cls_loss 0.0009 cls_loss_mapping 0.0034 cls_loss_causal 0.4958 re_mapping 0.0044 re_causal 0.0129 /// teacc 98.99 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1601, 0.1861, -0.0313, ..., -0.0898, 0.0267, 0.0078], + [ 0.0050, -0.1055, -0.0505, ..., 0.0298, -0.0368, -0.0596], + [-0.1131, -0.1985, 0.0377, ..., -0.0666, -0.0085, -0.1168], + ..., + [ 0.0742, 0.1151, -0.0856, ..., -0.1251, 0.0626, 0.0670], + [ 0.0475, -0.1044, -0.0200, ..., -0.0745, -0.0826, -0.0742], + [-0.1469, -0.0374, -0.0590, ..., 0.0974, -0.2382, 0.0895]], + device='cuda:0'), grad: tensor([[ 6.7521e-08, -6.1933e-08, 2.1420e-08, ..., 8.6613e-08, + 7.9162e-09, -6.5193e-09], + [ 3.8370e-06, 1.0245e-08, 1.8161e-08, ..., 3.5632e-06, + 1.1642e-08, 3.0901e-06], + [ 8.5682e-08, 1.7695e-08, -3.9395e-07, ..., 5.9139e-08, + -2.2631e-07, 8.5682e-08], + ..., + [-1.2470e-06, -3.8184e-08, 1.0710e-08, ..., 2.3749e-07, + -1.3970e-09, -4.3288e-06], + [ 2.3749e-07, 1.9092e-08, 1.0245e-08, ..., 1.7788e-07, + 1.8161e-08, 3.1292e-07], + [ 7.8464e-07, 7.9162e-08, 1.8626e-09, ..., 9.6671e-07, + 3.1665e-08, 1.8533e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0222, -0.0300, 0.0154, -0.0279, 0.0347, 0.0206, 0.0041, -0.0111, + -0.0134, -0.0060], device='cuda:0'), grad: tensor([ 2.9569e-07, 1.6898e-05, -2.3488e-06, 3.4496e-06, -1.5289e-05, + 1.2061e-07, 1.1362e-06, -8.5533e-06, 1.1157e-06, 3.1646e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 220.41, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.5293 re_mapping 0.0044 re_causal 0.0134 /// teacc 99.10 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1615, 0.1862, -0.0314, ..., -0.0906, 0.0266, 0.0078], + [ 0.0050, -0.1058, -0.0505, ..., 0.0298, -0.0369, -0.0596], + [-0.1135, -0.1990, 0.0382, ..., -0.0668, -0.0086, -0.1170], + ..., + [ 0.0742, 0.1153, -0.0856, ..., -0.1251, 0.0626, 0.0671], + [ 0.0472, -0.1050, -0.0204, ..., -0.0757, -0.0836, -0.0746], + [-0.1475, -0.0375, -0.0590, ..., 0.0973, -0.2383, 0.0895]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -7.2177e-08, 0.0000e+00, ..., 2.4866e-07, + 3.1199e-08, 2.4680e-07], + [-4.8708e-07, 7.4506e-09, -3.1665e-08, ..., -3.2969e-07, + 2.5658e-07, 9.2341e-07], + [ 3.1665e-08, 1.6764e-08, 9.3132e-10, ..., 4.3772e-08, + 3.4831e-07, 5.7463e-07], + ..., + [-2.4168e-07, -1.6764e-08, 3.7253e-09, ..., 1.4789e-06, + -7.3481e-07, -4.4629e-06], + [ 2.1840e-07, 3.2596e-09, 2.0489e-08, ..., 1.4342e-06, + 8.8476e-08, 1.2405e-06], + [-2.9942e-07, 4.4238e-08, 0.0000e+00, ..., -8.0764e-06, + 1.1604e-06, -2.3153e-06]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0220, -0.0300, 0.0156, -0.0285, 0.0349, 0.0211, 0.0047, -0.0111, + -0.0138, -0.0061], device='cuda:0'), grad: tensor([ 8.7218e-07, 7.5530e-07, 1.4305e-06, -2.3581e-06, 1.5706e-05, + 1.7583e-06, 1.8161e-07, -8.1286e-06, 4.7572e-06, -1.5028e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 220.37, cls_loss 0.0009 cls_loss_mapping 0.0034 cls_loss_causal 0.4928 re_mapping 0.0046 re_causal 0.0130 /// teacc 99.08 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1624, 0.1865, -0.0318, ..., -0.0908, 0.0267, 0.0079], + [ 0.0050, -0.1062, -0.0505, ..., 0.0298, -0.0369, -0.0597], + [-0.1138, -0.1998, 0.0400, ..., -0.0672, -0.0088, -0.1173], + ..., + [ 0.0742, 0.1155, -0.0856, ..., -0.1252, 0.0626, 0.0671], + [ 0.0468, -0.1059, -0.0208, ..., -0.0765, -0.0860, -0.0757], + [-0.1485, -0.0377, -0.0590, ..., 0.0968, -0.2384, 0.0895]], + device='cuda:0'), grad: tensor([[ 4.9826e-08, -1.9558e-08, 0.0000e+00, ..., 9.2667e-08, + 1.4435e-08, 2.4214e-08], + [ 4.3772e-08, 1.3970e-09, 0.0000e+00, ..., 1.4761e-07, + 1.4435e-08, 4.5635e-08], + [ 1.8161e-08, 3.2596e-09, 0.0000e+00, ..., 4.7032e-08, + 3.7253e-09, 2.6077e-08], + ..., + [ 6.7521e-08, -1.8626e-09, 0.0000e+00, ..., 1.4063e-07, + 1.8161e-08, 1.4575e-07], + [ 1.3737e-07, 3.7253e-09, 0.0000e+00, ..., 1.2433e-07, + 7.2177e-08, 9.3132e-08], + [ 7.6834e-08, 3.7253e-09, 0.0000e+00, ..., -6.6916e-07, + 3.8650e-08, -6.8778e-07]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0221, -0.0301, 0.0158, -0.0254, 0.0357, 0.0182, 0.0051, -0.0112, + -0.0146, -0.0065], device='cuda:0'), grad: tensor([ 4.3306e-07, 5.3225e-07, 2.1746e-07, 8.8289e-06, 9.8906e-07, + -1.1623e-05, 4.1584e-07, 7.4925e-07, 9.6299e-07, -1.5013e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 220.46, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4819 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.09 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1634, 0.1864, -0.0323, ..., -0.0919, 0.0240, 0.0076], + [ 0.0051, -0.1070, -0.0505, ..., 0.0298, -0.0370, -0.0597], + [-0.1140, -0.2008, 0.0414, ..., -0.0670, -0.0088, -0.1175], + ..., + [ 0.0742, 0.1158, -0.0858, ..., -0.1254, 0.0626, 0.0670], + [ 0.0451, -0.1069, -0.0210, ..., -0.0777, -0.0870, -0.0769], + [-0.1489, -0.0376, -0.0591, ..., 0.0969, -0.2385, 0.0897]], + device='cuda:0'), grad: tensor([[ 7.7765e-08, -1.3830e-07, 0.0000e+00, ..., 2.0582e-07, + -2.6077e-08, -5.0757e-08], + [-4.8336e-07, 5.2154e-08, -1.5832e-08, ..., -5.2992e-07, + 4.2375e-08, 9.7789e-09], + [ 1.4855e-07, 9.1270e-08, 4.6566e-10, ..., 3.1712e-07, + 6.6124e-08, 2.5146e-08], + ..., + [ 2.3330e-07, -2.7940e-09, 9.3132e-10, ..., 2.5332e-07, + 4.2375e-08, -6.0536e-09], + [ 2.7800e-07, 1.3504e-08, 9.7789e-09, ..., 1.7695e-07, + 9.4995e-08, 1.2573e-08], + [ 3.7393e-07, 8.3819e-08, 4.6566e-10, ..., 5.7789e-07, + 4.4703e-08, 2.4214e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0210, -0.0301, 0.0162, -0.0256, 0.0358, 0.0189, 0.0055, -0.0112, + -0.0158, -0.0065], device='cuda:0'), grad: tensor([ 1.0487e-06, -1.7527e-06, -5.5209e-06, 1.9461e-05, -4.9174e-06, + -2.2039e-05, 5.6773e-06, 4.5896e-06, 1.5991e-06, 1.8729e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 220.53, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4864 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.10 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1645, 0.1868, -0.0326, ..., -0.0919, 0.0244, 0.0077], + [ 0.0052, -0.1074, -0.0505, ..., 0.0299, -0.0370, -0.0597], + [-0.1143, -0.2033, 0.0420, ..., -0.0681, -0.0090, -0.1178], + ..., + [ 0.0741, 0.1158, -0.0858, ..., -0.1255, 0.0627, 0.0670], + [ 0.0448, -0.1075, -0.0232, ..., -0.0791, -0.0884, -0.0778], + [-0.1496, -0.0376, -0.0591, ..., 0.0969, -0.2386, 0.0898]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 1.7229e-08, 1.8626e-09, ..., 4.0187e-07, + 8.3819e-09, 5.1223e-09], + [-5.3607e-06, 1.8626e-08, 4.6566e-10, ..., -6.5565e-06, + 8.8476e-09, 1.4435e-08], + [ 4.4238e-08, 8.2422e-08, -1.3970e-09, ..., 2.3330e-07, + -3.4925e-08, 1.6298e-08], + ..., + [ 1.7509e-07, -9.3132e-10, 4.6566e-10, ..., 2.8219e-07, + -4.6566e-09, 1.0896e-07], + [-2.5705e-07, 5.4017e-08, 1.8626e-09, ..., 1.6019e-07, + 4.1910e-09, -3.3528e-08], + [ 1.0263e-06, 2.2352e-08, 0.0000e+00, ..., 1.4901e-06, + 5.1223e-09, -1.6904e-07]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0213, -0.0300, 0.0163, -0.0258, 0.0359, 0.0191, 0.0057, -0.0114, + -0.0165, -0.0064], device='cuda:0'), grad: tensor([ 2.1905e-06, -1.4000e-05, 1.0226e-06, 5.2480e-07, 1.2331e-05, + 9.3225e-07, -6.4336e-06, 9.2667e-07, -2.2948e-06, 4.8019e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 220.23, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4946 re_mapping 0.0042 re_causal 0.0134 /// teacc 99.13 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1655, 0.1864, -0.0327, ..., -0.0931, 0.0243, 0.0066], + [ 0.0052, -0.1080, -0.0505, ..., 0.0299, -0.0371, -0.0598], + [-0.1150, -0.2038, 0.0418, ..., -0.0690, -0.0086, -0.1182], + ..., + [ 0.0742, 0.1159, -0.0858, ..., -0.1257, 0.0628, 0.0667], + [ 0.0445, -0.1078, -0.0233, ..., -0.0802, -0.0895, -0.0783], + [-0.1500, -0.0372, -0.0592, ..., 0.0973, -0.2387, 0.0906]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, -1.2899e-07, 7.4506e-09, ..., 1.5600e-07, + 1.1176e-08, 8.8476e-09], + [-3.0734e-08, 4.1910e-09, 4.6100e-08, ..., -1.6298e-08, + 8.2422e-08, 1.8626e-08], + [ 4.6566e-08, 2.2817e-08, -2.1188e-07, ..., 2.9802e-08, + -3.3807e-07, 3.3528e-08], + ..., + [ 3.3528e-08, -2.7474e-08, 5.2154e-08, ..., 7.6834e-08, + 6.4261e-08, 9.3132e-10], + [-3.0175e-07, 1.8626e-09, 1.2107e-08, ..., 1.8813e-07, + 2.2817e-08, 1.0850e-07], + [ 3.7253e-08, 1.2107e-08, 4.6566e-10, ..., -5.1968e-07, + 6.5193e-09, -4.3027e-07]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0203, -0.0299, 0.0164, -0.0260, 0.0358, 0.0190, 0.0066, -0.0117, + -0.0172, -0.0056], device='cuda:0'), grad: tensor([ 6.5751e-07, 9.1596e-07, -3.5353e-06, 2.8182e-06, 1.1362e-06, + 6.1002e-08, -1.3104e-06, 1.2424e-06, -7.3761e-07, -1.2433e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 220.04, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4997 re_mapping 0.0041 re_causal 0.0129 /// teacc 99.08 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1663, 0.1866, -0.0332, ..., -0.0932, 0.0242, 0.0066], + [ 0.0052, -0.1083, -0.0505, ..., 0.0299, -0.0372, -0.0598], + [-0.1157, -0.2042, 0.0420, ..., -0.0697, -0.0086, -0.1185], + ..., + [ 0.0743, 0.1160, -0.0858, ..., -0.1258, 0.0628, 0.0668], + [ 0.0444, -0.1081, -0.0235, ..., -0.0808, -0.0903, -0.0788], + [-0.1508, -0.0373, -0.0592, ..., 0.0975, -0.2388, 0.0906]], + device='cuda:0'), grad: tensor([[ 4.1444e-08, -1.5600e-07, 9.3132e-10, ..., 5.9605e-08, + -1.5367e-08, -5.5879e-08], + [-4.1910e-09, 4.1910e-09, 9.3132e-10, ..., 6.9197e-07, + 4.1910e-08, 8.3260e-07], + [ 2.4680e-08, 1.2107e-08, 1.3970e-09, ..., 5.6345e-08, + -2.7940e-09, 3.5856e-08], + ..., + [ 2.1886e-08, -1.8626e-09, 0.0000e+00, ..., 6.9011e-07, + 1.1548e-07, 7.9907e-07], + [ 2.8871e-08, 5.5879e-09, 4.6566e-10, ..., 8.1956e-08, + 5.0291e-08, 1.3132e-07], + [ 4.1910e-09, 1.0151e-07, 0.0000e+00, ..., -2.9393e-06, + 3.1060e-07, -2.3376e-06]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0203, -0.0299, 0.0164, -0.0260, 0.0357, 0.0191, 0.0063, -0.0116, + -0.0173, -0.0057], device='cuda:0'), grad: tensor([ 4.6566e-10, 2.4289e-06, 1.5041e-07, -2.4140e-06, 3.8333e-06, + -7.2457e-07, 1.2992e-07, 2.6543e-06, 5.4389e-07, -6.5900e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 220.61, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.5045 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1670, 0.1870, -0.0337, ..., -0.0932, 0.0240, 0.0066], + [ 0.0052, -0.1097, -0.0505, ..., 0.0299, -0.0373, -0.0599], + [-0.1161, -0.2062, 0.0425, ..., -0.0700, -0.0086, -0.1189], + ..., + [ 0.0744, 0.1162, -0.0858, ..., -0.1259, 0.0626, 0.0668], + [ 0.0443, -0.1098, -0.0235, ..., -0.0822, -0.0909, -0.0803], + [-0.1512, -0.0373, -0.0592, ..., 0.0977, -0.2389, 0.0908]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, -1.1828e-07, 4.6566e-09, ..., 1.5367e-08, + -9.3132e-10, -3.3993e-08], + [-8.3819e-08, 3.7253e-09, 6.5193e-09, ..., -1.5646e-07, + 4.3772e-08, 3.1665e-08], + [ 2.2817e-08, 1.0710e-08, -1.3970e-07, ..., 3.5390e-08, + -1.1921e-07, 1.1222e-07], + ..., + [ 1.2573e-08, 6.9849e-09, 1.6298e-08, ..., 3.3993e-08, + 2.6915e-07, 1.7695e-07], + [ 1.4435e-08, 6.9849e-09, 1.3039e-08, ..., 3.3528e-08, + 1.9232e-07, 1.4948e-07], + [ 1.0710e-08, 5.5414e-08, 1.8626e-09, ..., 1.2107e-08, + 5.1223e-08, 5.9139e-08]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0203, -0.0300, 0.0166, -0.0258, 0.0355, 0.0190, 0.0061, -0.0117, + -0.0175, -0.0055], device='cuda:0'), grad: tensor([ 5.2154e-08, -6.0676e-07, -2.0936e-06, -4.3726e-07, 1.4715e-07, + 3.6554e-07, 1.0431e-07, 1.3150e-06, 8.6986e-07, 3.0082e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 220.11, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.5175 re_mapping 0.0044 re_causal 0.0127 /// teacc 99.17 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1696, 0.1875, -0.0338, ..., -0.0935, 0.0240, 0.0069], + [ 0.0054, -0.1100, -0.0505, ..., 0.0300, -0.0374, -0.0598], + [-0.1171, -0.2078, 0.0425, ..., -0.0710, -0.0086, -0.1197], + ..., + [ 0.0745, 0.1164, -0.0858, ..., -0.1259, 0.0626, 0.0670], + [ 0.0414, -0.1108, -0.0237, ..., -0.0856, -0.0918, -0.0819], + [-0.1520, -0.0375, -0.0592, ..., 0.0976, -0.2390, 0.0906]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -5.0850e-07, 0.0000e+00, ..., 3.6461e-07, + -2.7008e-08, -6.0536e-08], + [ 3.7253e-09, 9.1270e-08, 0.0000e+00, ..., 2.2817e-08, + 5.3551e-08, 8.2422e-08], + [ 1.7229e-08, 9.3598e-08, -4.6566e-10, ..., 2.1420e-08, + 6.4727e-08, 6.8918e-08], + ..., + [-3.2596e-08, -1.3923e-07, 0.0000e+00, ..., 1.8626e-08, + -1.3970e-07, -2.5611e-07], + [ 2.7008e-08, 2.3935e-07, 0.0000e+00, ..., 5.9139e-08, + 3.3528e-08, 2.3283e-08], + [ 7.1060e-07, 1.6252e-07, 0.0000e+00, ..., 1.1828e-06, + 8.4285e-08, 1.1269e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0205, -0.0298, 0.0164, -0.0257, 0.0357, 0.0189, 0.0083, -0.0115, + -0.0202, -0.0060], device='cuda:0'), grad: tensor([ 1.3504e-08, 3.3621e-07, 3.7393e-07, -8.8010e-08, -2.7381e-06, + 2.0955e-07, -1.1409e-06, -6.1234e-07, 6.3470e-07, 2.9970e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 220.32, cls_loss 0.0006 cls_loss_mapping 0.0024 cls_loss_causal 0.5079 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.11 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1702, 0.1876, -0.0338, ..., -0.0935, 0.0241, 0.0070], + [ 0.0055, -0.1103, -0.0505, ..., 0.0300, -0.0375, -0.0599], + [-0.1173, -0.2084, 0.0435, ..., -0.0713, -0.0086, -0.1214], + ..., + [ 0.0745, 0.1165, -0.0859, ..., -0.1260, 0.0626, 0.0671], + [ 0.0415, -0.1108, -0.0237, ..., -0.0859, -0.0926, -0.0824], + [-0.1524, -0.0376, -0.0592, ..., 0.0976, -0.2392, 0.0906]], + device='cuda:0'), grad: tensor([[ 1.9092e-08, -7.6415e-07, 1.3970e-09, ..., 1.5926e-07, + -2.2165e-07, -2.1514e-07], + [ 1.6717e-07, 6.4727e-08, 1.7695e-08, ..., 1.4435e-08, + 5.1688e-08, 1.0571e-07], + [ 1.4249e-07, 3.8184e-07, 2.4214e-08, ..., 3.4459e-08, + 2.4633e-07, 2.1560e-07], + ..., + [-2.9895e-07, -1.5041e-07, -5.9605e-08, ..., 6.9849e-09, + -1.3504e-07, -3.1013e-07], + [-5.9465e-07, 2.8405e-08, 6.9849e-09, ..., 3.6927e-07, + 2.1886e-08, 3.0268e-08], + [ 4.5169e-07, 3.9395e-07, 3.2596e-09, ..., 1.3411e-07, + 3.1199e-08, 1.1781e-07]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0205, -0.0298, 0.0165, -0.0256, 0.0358, 0.0189, 0.0084, -0.0116, + -0.0202, -0.0061], device='cuda:0'), grad: tensor([-9.6019e-07, 1.1642e-06, 1.8394e-06, 4.2608e-07, 3.0268e-08, + 1.9092e-06, -3.4180e-06, -1.1828e-06, -4.4070e-06, 4.6007e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 220.04, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.5436 re_mapping 0.0044 re_causal 0.0137 /// teacc 99.08 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1710, 0.1878, -0.0338, ..., -0.0939, 0.0239, 0.0070], + [ 0.0054, -0.1118, -0.0505, ..., 0.0299, -0.0376, -0.0600], + [-0.1192, -0.2106, 0.0439, ..., -0.0717, -0.0093, -0.1239], + ..., + [ 0.0747, 0.1175, -0.0859, ..., -0.1260, 0.0631, 0.0674], + [ 0.0413, -0.1115, -0.0238, ..., -0.0862, -0.0935, -0.0836], + [-0.1542, -0.0378, -0.0592, ..., 0.0958, -0.2395, 0.0903]], + device='cuda:0'), grad: tensor([[ 5.0757e-08, -2.1886e-08, 0.0000e+00, ..., 3.4459e-08, + 9.3132e-10, 3.8650e-08], + [ 1.5320e-07, 7.0315e-08, 0.0000e+00, ..., 1.2340e-07, + 2.3283e-09, 2.1420e-07], + [ 1.9092e-08, 1.2107e-08, 0.0000e+00, ..., 1.3970e-08, + 1.0710e-08, 3.6787e-08], + ..., + [ 1.4529e-07, -5.7509e-07, 0.0000e+00, ..., 1.6950e-07, + -1.8626e-09, -1.2647e-06], + [-4.8149e-07, 8.8476e-09, 0.0000e+00, ..., 3.0734e-08, + 2.7940e-09, -1.8673e-07], + [ 3.3900e-07, 4.5169e-07, 0.0000e+00, ..., -2.5500e-06, + 2.7940e-09, -5.5460e-07]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0205, -0.0298, 0.0157, -0.0256, 0.0377, 0.0188, 0.0084, -0.0112, + -0.0204, -0.0075], device='cuda:0'), grad: tensor([ 2.8405e-07, 1.4715e-06, 2.0303e-07, 3.4738e-07, 5.4538e-06, + -8.4788e-06, 8.5309e-06, -1.0058e-06, -5.0366e-06, -1.7863e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 220.42, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4760 re_mapping 0.0045 re_causal 0.0130 /// teacc 99.09 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1721, 0.1886, -0.0323, ..., -0.0943, 0.0235, 0.0071], + [ 0.0053, -0.1144, -0.0505, ..., 0.0299, -0.0377, -0.0604], + [-0.1198, -0.2127, 0.0440, ..., -0.0729, -0.0104, -0.1253], + ..., + [ 0.0749, 0.1186, -0.0859, ..., -0.1261, 0.0629, 0.0678], + [ 0.0417, -0.1119, -0.0239, ..., -0.0864, -0.0943, -0.0842], + [-0.1544, -0.0380, -0.0593, ..., 0.0959, -0.2396, 0.0904]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, -1.6997e-07, 1.3970e-09, ..., 7.0315e-08, + -1.3504e-08, 5.3085e-08], + [ 1.7695e-08, 5.5879e-09, -1.3970e-09, ..., -9.7789e-09, + 2.6077e-08, 5.1223e-08], + [ 1.7695e-07, 2.9337e-08, 3.9581e-08, ..., 1.2107e-08, + 4.9826e-07, 7.7020e-07], + ..., + [-3.0966e-07, -4.9826e-08, -6.4727e-08, ..., 1.7509e-07, + -8.3027e-07, -1.0151e-06], + [ 2.4214e-08, 1.1176e-08, 3.2596e-09, ..., 1.7695e-07, + 3.3993e-08, 1.7323e-07], + [ 4.5635e-08, 1.8161e-08, 7.4506e-09, ..., -7.5297e-07, + 1.0291e-07, -8.0094e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0210, -0.0301, 0.0151, -0.0255, 0.0377, 0.0190, 0.0068, -0.0109, + -0.0198, -0.0075], device='cuda:0'), grad: tensor([-6.6590e-08, 9.8720e-08, 3.2075e-06, 7.9488e-07, 9.2061e-07, + 7.9954e-07, -2.3330e-07, -4.6045e-06, 6.5379e-07, -1.5665e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 220.23, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.5242 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.13 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1737, 0.1887, -0.0327, ..., -0.0953, 0.0229, 0.0070], + [ 0.0056, -0.1153, -0.0501, ..., 0.0302, -0.0374, -0.0598], + [-0.1207, -0.2135, 0.0435, ..., -0.0738, -0.0118, -0.1268], + ..., + [ 0.0747, 0.1192, -0.0862, ..., -0.1265, 0.0635, 0.0676], + [ 0.0419, -0.1118, -0.0246, ..., -0.0891, -0.0954, -0.0861], + [-0.1549, -0.0381, -0.0597, ..., 0.0973, -0.2399, 0.0910]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -4.6892e-07, 0.0000e+00, ..., -1.2107e-08, + 3.7253e-09, -2.7474e-08], + [-9.3132e-09, 1.0245e-08, 0.0000e+00, ..., -1.0710e-08, + 7.9162e-09, 1.0841e-06], + [ 3.7253e-09, 1.2107e-07, 0.0000e+00, ..., 3.2596e-09, + 2.7940e-08, 3.3900e-07], + ..., + [ 4.6566e-10, 6.5193e-09, 0.0000e+00, ..., 1.8626e-09, + 4.6147e-07, -1.4845e-06], + [ 1.7695e-08, 2.8405e-08, 0.0000e+00, ..., 6.0536e-09, + 8.3819e-09, 2.0256e-07], + [ 6.5193e-09, 1.1502e-07, 0.0000e+00, ..., 1.0245e-08, + 3.3062e-08, 1.4063e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0204, -0.0293, 0.0148, -0.0256, 0.0363, 0.0181, 0.0080, -0.0113, + -0.0192, -0.0064], device='cuda:0'), grad: tensor([-6.8452e-07, 2.3469e-06, 8.5542e-07, -8.8802e-07, 1.1176e-08, + 1.9092e-08, 3.7951e-07, -3.0920e-06, 5.8115e-07, 4.8522e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 220.35, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4771 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.07 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1750, 0.1891, -0.0327, ..., -0.0953, 0.0229, 0.0071], + [ 0.0056, -0.1156, -0.0500, ..., 0.0300, -0.0375, -0.0598], + [-0.1211, -0.2144, 0.0441, ..., -0.0741, -0.0119, -0.1272], + ..., + [ 0.0748, 0.1196, -0.0864, ..., -0.1267, 0.0637, 0.0678], + [ 0.0416, -0.1121, -0.0250, ..., -0.0895, -0.0963, -0.0866], + [-0.1556, -0.0382, -0.0598, ..., 0.0982, -0.2401, 0.0909]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, -4.1910e-09, 2.3283e-08, ..., 1.5832e-08, + 2.3283e-09, 1.8626e-09], + [ 1.0990e-07, 1.6298e-08, 8.3819e-09, ..., 2.5611e-08, + 6.8452e-08, 1.2293e-07], + [ 1.3039e-08, 1.8626e-09, -6.4727e-08, ..., 1.3039e-08, + 1.3039e-08, 1.5832e-08], + ..., + [-1.5507e-07, -2.8405e-08, 4.6566e-10, ..., 2.7474e-08, + -1.1502e-07, -2.2585e-07], + [ 1.3039e-08, 1.3970e-09, 2.7940e-09, ..., 7.9162e-09, + 9.7789e-09, 8.3819e-09], + [ 6.6077e-07, 9.3132e-09, 9.3132e-10, ..., 1.4361e-06, + 3.2596e-08, 4.9826e-08]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0206, -0.0293, 0.0148, -0.0256, 0.0356, 0.0180, 0.0075, -0.0112, + -0.0189, -0.0060], device='cuda:0'), grad: tensor([ 1.7323e-07, 4.7358e-07, -2.3702e-07, 1.2759e-07, -3.7160e-06, + -7.3574e-08, 3.2596e-09, -5.6066e-07, 8.7079e-08, 3.7309e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 220.67, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.5284 re_mapping 0.0042 re_causal 0.0131 /// teacc 99.13 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1771, 0.1883, -0.0327, ..., -0.0983, 0.0228, 0.0071], + [ 0.0057, -0.1160, -0.0500, ..., 0.0299, -0.0376, -0.0598], + [-0.1216, -0.2154, 0.0442, ..., -0.0743, -0.0123, -0.1275], + ..., + [ 0.0747, 0.1198, -0.0864, ..., -0.1268, 0.0636, 0.0677], + [ 0.0414, -0.1125, -0.0251, ..., -0.0898, -0.0977, -0.0868], + [-0.1560, -0.0383, -0.0598, ..., 0.0981, -0.2402, 0.0910]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -6.8452e-07, 2.3283e-10, ..., -3.5856e-08, + 3.5786e-07, 1.0058e-07], + [-2.6007e-07, 9.3132e-09, -2.5379e-08, ..., -9.1037e-08, + 1.1502e-07, 1.4086e-07], + [ 8.1258e-08, 3.9581e-09, 8.1491e-09, ..., 1.2340e-08, + 1.5297e-07, 1.6484e-07], + ..., + [-1.1642e-08, -1.5832e-08, 2.3283e-10, ..., 4.3074e-08, + 5.4576e-07, 6.2492e-07], + [ 3.9581e-09, 2.4913e-08, 4.6566e-10, ..., 2.1188e-08, + 1.6089e-07, 1.9255e-07], + [ 2.1188e-08, 2.0210e-07, 0.0000e+00, ..., -6.9616e-08, + 2.8964e-06, 3.1721e-06]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0184, -0.0294, 0.0151, -0.0256, 0.0358, 0.0181, 0.0099, -0.0114, + -0.0190, -0.0060], device='cuda:0'), grad: tensor([ 4.0862e-07, -4.0489e-07, 1.2256e-06, -2.1368e-05, 4.4657e-07, + 5.0180e-06, 8.9873e-07, 2.2724e-06, -1.0710e-08, 1.1489e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 220.45, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.5074 re_mapping 0.0041 re_causal 0.0128 /// teacc 99.03 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1783, 0.1885, -0.0330, ..., -0.0986, 0.0226, 0.0070], + [ 0.0057, -0.1163, -0.0500, ..., 0.0299, -0.0377, -0.0598], + [-0.1219, -0.2163, 0.0445, ..., -0.0745, -0.0123, -0.1278], + ..., + [ 0.0747, 0.1200, -0.0864, ..., -0.1269, 0.0635, 0.0677], + [ 0.0426, -0.1130, -0.0252, ..., -0.0887, -0.0984, -0.0846], + [-0.1570, -0.0383, -0.0599, ..., 0.0980, -0.2403, 0.0910]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -3.5297e-07, 0.0000e+00, ..., 7.8930e-08, + 6.9849e-09, -4.9826e-08], + [ 3.3062e-08, 1.5832e-08, 0.0000e+00, ..., 5.4017e-08, + 4.8429e-08, 1.2340e-07], + [ 2.1886e-08, 1.2107e-08, -4.6566e-10, ..., 3.1432e-08, + 2.1653e-08, 4.0513e-08], + ..., + [ 1.1642e-09, 3.9581e-09, 0.0000e+00, ..., 1.3527e-07, + -1.2456e-07, -3.4319e-07], + [ 2.3493e-07, 1.7229e-08, 0.0000e+00, ..., 1.5809e-07, + 7.4971e-08, 5.9372e-08], + [ 1.0408e-07, 4.8894e-08, 0.0000e+00, ..., -6.7502e-06, + 9.6858e-08, -3.4031e-06]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0183, -0.0294, 0.0152, -0.0255, 0.0361, 0.0182, 0.0099, -0.0114, + -0.0180, -0.0063], device='cuda:0'), grad: tensor([-4.1677e-08, 5.0850e-07, 2.8755e-07, 6.1207e-06, 1.4991e-05, + -1.0207e-05, 9.3179e-07, -2.7171e-07, 1.4119e-06, -1.3724e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 220.55, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.4735 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.12 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1797, 0.1905, -0.0331, ..., -0.0986, 0.0224, 0.0074], + [ 0.0057, -0.1169, -0.0501, ..., 0.0299, -0.0375, -0.0598], + [-0.1221, -0.2174, 0.0454, ..., -0.0745, -0.0120, -0.1280], + ..., + [ 0.0748, 0.1205, -0.0864, ..., -0.1270, 0.0636, 0.0679], + [ 0.0427, -0.1147, -0.0254, ..., -0.0890, -0.0991, -0.0849], + [-0.1574, -0.0386, -0.0598, ..., 0.0979, -0.2409, 0.0908]], + device='cuda:0'), grad: tensor([[ 1.9092e-08, -2.6380e-07, 0.0000e+00, ..., 2.5611e-09, + 1.0245e-08, -2.6776e-08], + [ 2.3283e-09, 3.0268e-09, -6.9849e-10, ..., -2.9104e-08, + 1.7928e-08, 6.7521e-09], + [ 1.3737e-07, 5.8906e-08, 0.0000e+00, ..., 1.3271e-08, + 7.8930e-08, 1.0827e-07], + ..., + [ 7.7067e-08, 2.7940e-09, 2.3283e-10, ..., 2.1886e-08, + 5.2387e-08, -8.8476e-08], + [ 2.8638e-07, 1.8859e-08, 0.0000e+00, ..., 1.3271e-08, + 1.7229e-07, 1.1874e-08], + [ 5.6112e-08, 8.3353e-08, 0.0000e+00, ..., 5.5647e-08, + 1.7462e-08, 1.6531e-08]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0197, -0.0294, 0.0164, -0.0256, 0.0363, 0.0183, 0.0080, -0.0116, + -0.0179, -0.0067], device='cuda:0'), grad: tensor([-3.7905e-07, 9.0804e-09, 8.0280e-07, 6.6170e-07, -1.3015e-07, + -2.8163e-06, 2.2841e-07, 2.1467e-07, 9.5461e-07, 4.4797e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 220.16, cls_loss 0.0009 cls_loss_mapping 0.0028 cls_loss_causal 0.5074 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.12 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1803, 0.1910, -0.0332, ..., -0.0988, 0.0228, 0.0075], + [ 0.0056, -0.1175, -0.0501, ..., 0.0298, -0.0377, -0.0599], + [-0.1222, -0.2191, 0.0458, ..., -0.0752, -0.0122, -0.1287], + ..., + [ 0.0748, 0.1207, -0.0864, ..., -0.1270, 0.0637, 0.0680], + [ 0.0433, -0.1170, -0.0257, ..., -0.0900, -0.0998, -0.0856], + [-0.1581, -0.0387, -0.0576, ..., 0.0980, -0.2411, 0.0911]], + device='cuda:0'), grad: tensor([[ 8.6147e-09, -3.7998e-07, 6.0536e-09, ..., 6.6357e-08, + 1.2852e-07, -6.9849e-10], + [-2.7637e-07, 1.6997e-08, -5.9558e-07, ..., -8.6334e-07, + 1.4482e-07, 1.1292e-07], + [ 1.8394e-08, 1.3970e-08, 2.0489e-08, ..., 1.0571e-07, + 6.5053e-07, 4.8336e-07], + ..., + [ 7.6834e-09, 7.4506e-09, 1.4435e-08, ..., 6.5658e-08, + 5.5460e-07, 4.8056e-07], + [ 1.5972e-07, 4.9826e-08, 3.2200e-07, ..., 5.3504e-07, + 1.9884e-07, 1.4901e-07], + [ 1.5832e-08, 8.6613e-08, 2.5611e-09, ..., 2.2817e-07, + 2.1863e-07, -3.7719e-08]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0199, -0.0296, 0.0165, -0.0256, 0.0367, 0.0183, 0.0079, -0.0115, + -0.0178, -0.0069], device='cuda:0'), grad: tensor([-1.3784e-07, -2.9225e-06, 2.3525e-06, -2.0787e-05, 1.6280e-06, + 1.4670e-05, -9.0338e-07, 1.9111e-06, 2.8666e-06, 1.2890e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 220.92, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.5057 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.08 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1810, 0.1912, -0.0334, ..., -0.0990, 0.0229, 0.0076], + [ 0.0056, -0.1177, -0.0501, ..., 0.0298, -0.0377, -0.0599], + [-0.1228, -0.2201, 0.0459, ..., -0.0760, -0.0125, -0.1298], + ..., + [ 0.0749, 0.1211, -0.0864, ..., -0.1271, 0.0638, 0.0682], + [ 0.0433, -0.1174, -0.0258, ..., -0.0901, -0.1005, -0.0860], + [-0.1584, -0.0388, -0.0576, ..., 0.0982, -0.2413, 0.0912]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, -2.5984e-07, 4.6566e-10, ..., 1.9558e-08, + -3.4925e-08, -7.1479e-08], + [ 2.7986e-07, 5.9837e-08, 3.9581e-09, ..., 1.9791e-08, + 3.4971e-07, 9.1409e-07], + [ 1.4435e-08, 6.6357e-08, -1.1874e-08, ..., 3.9581e-09, + 2.7940e-08, 4.6799e-08], + ..., + [-3.8021e-07, -6.9849e-08, 4.8894e-09, ..., 5.1223e-08, + -4.0443e-07, -1.0394e-06], + [ 3.6089e-08, 1.2107e-08, 1.3970e-09, ..., 1.7229e-08, + 1.8999e-07, 1.9744e-07], + [ 3.3062e-08, 1.2992e-07, 0.0000e+00, ..., -1.4044e-06, + 5.3551e-08, -4.3656e-07]], device='cuda:0') +Epoch 215, bias, value: tensor([ 0.0199, -0.0296, 0.0163, -0.0257, 0.0365, 0.0184, 0.0079, -0.0114, + -0.0178, -0.0068], device='cuda:0'), grad: tensor([-2.8056e-07, 2.3190e-06, -5.6298e-07, -6.0955e-07, 3.0901e-06, + 3.4692e-08, 1.7672e-07, -2.0619e-06, 8.8289e-07, -2.9821e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 220.50, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5061 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.05 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1824, 0.1913, -0.0338, ..., -0.0994, 0.0228, 0.0076], + [ 0.0054, -0.1183, -0.0503, ..., 0.0297, -0.0378, -0.0601], + [-0.1235, -0.2205, 0.0458, ..., -0.0769, -0.0129, -0.1304], + ..., + [ 0.0749, 0.1225, -0.0864, ..., -0.1273, 0.0644, 0.0686], + [ 0.0427, -0.1177, -0.0281, ..., -0.0905, -0.1014, -0.0867], + [-0.1585, -0.0393, -0.0576, ..., 0.0986, -0.2419, 0.0911]], + device='cuda:0'), grad: tensor([[ 2.0955e-08, -6.7521e-08, 6.7521e-09, ..., 8.8476e-09, + 2.3749e-08, 2.7707e-08], + [ 1.9465e-07, 3.9581e-09, 7.2177e-09, ..., -1.8626e-09, + 3.4785e-07, 4.4424e-07], + [ 8.7777e-08, 1.4668e-08, 2.0955e-09, ..., 7.9162e-09, + 1.4226e-07, 2.2096e-07], + ..., + [-1.0338e-06, 1.1642e-09, 6.9849e-10, ..., 3.3062e-08, + -1.7108e-06, -2.6282e-06], + [ 7.7998e-08, 2.7940e-09, 3.5390e-08, ..., 1.1874e-08, + 8.7079e-08, 1.4226e-07], + [ 4.6613e-07, 2.2585e-08, 1.3970e-09, ..., -8.6147e-09, + 6.9290e-07, 1.0850e-06]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0198, -0.0299, 0.0161, -0.0261, 0.0361, 0.0180, 0.0099, -0.0109, + -0.0185, -0.0069], device='cuda:0'), grad: tensor([ 7.7533e-08, 1.5534e-06, 7.3016e-07, 1.7723e-06, 5.1921e-08, + 5.0962e-06, -4.6119e-06, -9.7305e-06, 6.7614e-07, 4.3586e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 220.69, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.5043 re_mapping 0.0043 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1831, 0.1917, -0.0338, ..., -0.0995, 0.0226, 0.0076], + [ 0.0055, -0.1186, -0.0503, ..., 0.0297, -0.0379, -0.0601], + [-0.1240, -0.2215, 0.0457, ..., -0.0777, -0.0131, -0.1307], + ..., + [ 0.0750, 0.1227, -0.0864, ..., -0.1274, 0.0646, 0.0688], + [ 0.0428, -0.1182, -0.0283, ..., -0.0910, -0.1023, -0.0871], + [-0.1590, -0.0394, -0.0576, ..., 0.0988, -0.2421, 0.0911]], + device='cuda:0'), grad: tensor([[ 1.4203e-08, -1.5767e-06, 0.0000e+00, ..., -3.8417e-08, + -1.8184e-07, -6.0163e-07], + [ 5.4482e-08, 4.7032e-08, 0.0000e+00, ..., 1.2084e-07, + 1.5367e-08, 4.0047e-08], + [ 6.0536e-09, 5.9139e-08, -6.9849e-10, ..., 2.2585e-08, + 2.3516e-08, 3.7020e-08], + ..., + [ 3.3528e-08, 2.2817e-08, 2.3283e-10, ..., 1.1269e-07, + 1.6298e-09, -3.2596e-09], + [ 4.2841e-08, 8.0094e-08, 2.3283e-10, ..., 1.5832e-08, + 2.2352e-08, 2.7241e-08], + [ 6.4261e-08, 4.0606e-07, 0.0000e+00, ..., -1.6228e-07, + 2.2119e-08, -2.5844e-08]], device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0198, -0.0299, 0.0160, -0.0264, 0.0360, 0.0186, 0.0095, -0.0108, + -0.0185, -0.0070], device='cuda:0'), grad: tensor([-3.5502e-06, 7.6275e-07, 1.4203e-08, 5.3551e-07, -8.1165e-07, + -1.2368e-06, 2.3469e-06, 5.7789e-07, 2.5053e-07, 1.1204e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 220.64, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4941 re_mapping 0.0044 re_causal 0.0123 /// teacc 99.07 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1842, 0.1920, -0.0343, ..., -0.0996, 0.0224, 0.0076], + [ 0.0054, -0.1194, -0.0503, ..., 0.0297, -0.0381, -0.0602], + [-0.1247, -0.2225, 0.0459, ..., -0.0783, -0.0148, -0.1312], + ..., + [ 0.0751, 0.1229, -0.0864, ..., -0.1274, 0.0646, 0.0688], + [ 0.0426, -0.1188, -0.0285, ..., -0.0914, -0.1043, -0.0876], + [-0.1593, -0.0396, -0.0577, ..., 0.0991, -0.2423, 0.0912]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -1.9856e-06, 2.7707e-08, ..., 4.8103e-07, + -9.4064e-08, 6.9849e-09], + [-2.1653e-08, 1.6484e-07, 4.6566e-10, ..., 2.8405e-08, + 8.8476e-09, 2.1653e-08], + [ 1.8161e-08, 1.1735e-07, 9.3132e-10, ..., 8.7079e-08, + 1.2340e-08, 3.6787e-08], + ..., + [ 2.9104e-07, 1.1688e-07, 0.0000e+00, ..., 1.1995e-06, + 6.2864e-09, 1.9488e-07], + [ 2.3446e-07, 5.7276e-08, 2.5146e-08, ..., 1.2573e-06, + 6.7521e-09, 1.8929e-07], + [ 8.8215e-06, 1.3341e-07, 4.8894e-09, ..., 3.3170e-05, + 9.3132e-09, 2.4624e-06]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0196, -0.0300, 0.0153, -0.0244, 0.0358, 0.0171, 0.0097, -0.0107, + -0.0188, -0.0069], device='cuda:0'), grad: tensor([-4.7944e-06, 6.1467e-07, 5.9186e-07, 4.3772e-07, -9.8705e-05, + 1.1325e-06, 2.9877e-06, 4.0531e-06, 3.9414e-06, 8.9705e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 220.24, cls_loss 0.0010 cls_loss_mapping 0.0033 cls_loss_causal 0.5378 re_mapping 0.0043 re_causal 0.0127 /// teacc 99.05 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1864, 0.1926, -0.0350, ..., -0.0998, 0.0222, 0.0078], + [ 0.0051, -0.1216, -0.0504, ..., 0.0293, -0.0382, -0.0609], + [-0.1247, -0.2238, 0.0470, ..., -0.0786, -0.0153, -0.1320], + ..., + [ 0.0756, 0.1238, -0.0864, ..., -0.1277, 0.0650, 0.0693], + [ 0.0422, -0.1198, -0.0302, ..., -0.0921, -0.1055, -0.0884], + [-0.1598, -0.0399, -0.0552, ..., 0.1008, -0.2427, 0.0920]], + device='cuda:0'), grad: tensor([[ 1.8394e-07, -2.0862e-06, 1.0710e-08, ..., 1.1479e-07, + 2.7940e-09, -2.8638e-08], + [-2.2519e-06, 3.2759e-07, -7.1619e-07, ..., -1.8133e-06, + 1.6298e-09, 1.8394e-08], + [ 6.8918e-08, 7.5670e-08, 4.0769e-07, ..., 2.9290e-07, + 2.0955e-09, 7.2177e-09], + ..., + [ 6.7241e-07, 1.3388e-07, 1.5600e-08, ..., 4.0885e-07, + -1.7229e-08, -2.9849e-07], + [ 5.1968e-07, 1.4692e-07, 6.5425e-08, ..., 2.4191e-07, + 4.6566e-10, 1.8626e-09], + [ 1.6438e-07, 3.0501e-07, 1.8324e-07, ..., 2.1933e-07, + 3.2829e-08, 2.5798e-07]], device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0198, -0.0310, 0.0152, -0.0240, 0.0341, 0.0166, 0.0103, -0.0099, + -0.0195, -0.0057], device='cuda:0'), grad: tensor([-5.4464e-06, -2.2233e-05, 7.4357e-06, 2.0117e-06, 2.0191e-06, + -6.9663e-07, 4.5002e-06, 3.6657e-06, 3.6713e-06, 5.0664e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 220.86, cls_loss 0.0011 cls_loss_mapping 0.0035 cls_loss_causal 0.5303 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.05 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1900, 0.1933, -0.0353, ..., -0.1007, 0.0223, 0.0084], + [ 0.0078, -0.1190, -0.0475, ..., 0.0310, -0.0383, -0.0590], + [-0.1256, -0.2253, 0.0469, ..., -0.0798, -0.0157, -0.1325], + ..., + [ 0.0731, 0.1231, -0.0893, ..., -0.1293, 0.0653, 0.0679], + [ 0.0417, -0.1211, -0.0307, ..., -0.0938, -0.1063, -0.0887], + [-0.1610, -0.0407, -0.0558, ..., 0.0996, -0.2429, 0.0922]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, -6.2864e-09, 0.0000e+00, ..., 7.7998e-08, + 7.6834e-09, 5.3551e-09], + [ 1.0477e-08, 1.3271e-08, 6.9849e-10, ..., 1.2573e-08, + 1.7695e-08, 3.3295e-08], + [ 2.3516e-08, 1.5832e-08, -3.4925e-09, ..., 1.1642e-08, + -9.0804e-09, 4.5169e-08], + ..., + [-3.5134e-07, -4.1677e-08, 2.3283e-09, ..., 1.2340e-08, + -4.1910e-08, -9.3365e-08], + [ 1.6298e-09, 2.5611e-09, 0.0000e+00, ..., 6.9849e-09, + 1.3271e-08, 1.0012e-08], + [ 3.3295e-08, 1.2806e-08, 0.0000e+00, ..., -5.5647e-08, + 7.9628e-08, 5.5181e-08]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0201, -0.0279, 0.0148, -0.0243, 0.0353, 0.0168, 0.0105, -0.0126, + -0.0207, -0.0068], device='cuda:0'), grad: tensor([ 3.8045e-07, 2.8708e-07, -1.6196e-06, -6.8825e-07, 6.7428e-07, + 5.2433e-07, -2.0140e-07, 1.4366e-07, 3.0478e-07, 2.1281e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 220.51, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.4928 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.10 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1914, 0.1953, -0.0355, ..., -0.1008, 0.0222, 0.0101], + [ 0.0079, -0.1194, -0.0475, ..., 0.0309, -0.0383, -0.0590], + [-0.1261, -0.2264, 0.0467, ..., -0.0801, -0.0159, -0.1329], + ..., + [ 0.0731, 0.1233, -0.0893, ..., -0.1293, 0.0654, 0.0679], + [ 0.0418, -0.1214, -0.0308, ..., -0.0940, -0.1066, -0.0889], + [-0.1616, -0.0425, -0.0559, ..., 0.0998, -0.2430, 0.0919]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, -6.5379e-06, 0.0000e+00, ..., -1.3402e-06, + 2.1188e-08, -2.3283e-06], + [-4.8662e-08, -2.3283e-09, -1.6298e-09, ..., 1.2363e-07, + 3.7951e-08, 2.9337e-08], + [ 3.2363e-08, 9.0804e-09, -2.3283e-10, ..., 6.4494e-08, + 5.4948e-08, 2.8638e-08], + ..., + [-4.9360e-08, -4.4238e-09, 2.3283e-10, ..., 4.7265e-08, + -3.5390e-08, -6.7288e-08], + [ 8.1491e-09, 8.8476e-09, 2.3283e-10, ..., 9.5461e-09, + 4.6566e-08, 2.0955e-08], + [ 4.3958e-07, 1.3201e-07, 2.3283e-10, ..., 8.8941e-07, + 3.8417e-08, 5.4482e-08]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0219, -0.0278, 0.0144, -0.0237, 0.0351, 0.0162, 0.0103, -0.0126, + -0.0206, -0.0074], device='cuda:0'), grad: tensor([-1.1005e-05, 1.0128e-07, 5.3132e-07, -1.1930e-06, -2.8387e-06, + 6.9197e-07, 1.0885e-05, -2.4727e-07, 2.5984e-07, 2.8089e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 220.51, cls_loss 0.0008 cls_loss_mapping 0.0030 cls_loss_causal 0.5245 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.05 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1921, 0.1960, -0.0356, ..., -0.1010, 0.0221, 0.0106], + [ 0.0079, -0.1199, -0.0475, ..., 0.0309, -0.0384, -0.0590], + [-0.1265, -0.2280, 0.0467, ..., -0.0804, -0.0163, -0.1338], + ..., + [ 0.0732, 0.1238, -0.0893, ..., -0.1294, 0.0655, 0.0680], + [ 0.0417, -0.1219, -0.0309, ..., -0.0946, -0.1077, -0.0893], + [-0.1622, -0.0431, -0.0560, ..., 0.0998, -0.2432, 0.0919]], + device='cuda:0'), grad: tensor([[ 5.3551e-09, -6.1234e-08, 1.1642e-09, ..., 8.8476e-09, + 1.3970e-09, -3.9581e-09], + [-5.0059e-08, 7.9162e-09, 2.1653e-08, ..., -7.3574e-08, + 3.4925e-09, 6.9849e-09], + [ 1.3271e-08, 1.3271e-08, -5.2527e-07, ..., 1.5600e-08, + 1.6065e-08, 9.3132e-09], + ..., + [-2.3516e-08, -6.3796e-08, 4.9779e-07, ..., 2.3516e-08, + -1.7928e-08, -8.3121e-08], + [ 3.2596e-09, 6.7521e-09, 2.3283e-10, ..., 6.9849e-09, + 2.3283e-09, 1.8626e-09], + [ 1.0105e-07, 6.4261e-08, 2.3283e-10, ..., 1.9558e-07, + 1.8626e-08, 6.3097e-08]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0225, -0.0278, 0.0142, -0.0266, 0.0353, 0.0189, 0.0104, -0.0125, + -0.0209, -0.0077], device='cuda:0'), grad: tensor([-9.0105e-08, -1.6578e-07, -2.2445e-06, -7.9861e-08, -3.4040e-07, + 9.2899e-08, 1.4296e-07, 2.1495e-06, -7.1712e-08, 6.2399e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 220.28, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.4942 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.08 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1934, 0.1965, -0.0358, ..., -0.1012, 0.0220, 0.0108], + [ 0.0078, -0.1199, -0.0474, ..., 0.0309, -0.0385, -0.0591], + [-0.1267, -0.2285, 0.0464, ..., -0.0811, -0.0165, -0.1342], + ..., + [ 0.0732, 0.1240, -0.0893, ..., -0.1295, 0.0655, 0.0681], + [ 0.0416, -0.1224, -0.0313, ..., -0.0951, -0.1062, -0.0889], + [-0.1631, -0.0432, -0.0560, ..., 0.0999, -0.2434, 0.0920]], + device='cuda:0'), grad: tensor([[ 8.6147e-09, -3.3528e-08, 0.0000e+00, ..., 1.3970e-08, + 6.7521e-09, 2.3283e-09], + [-2.9104e-08, 1.1642e-09, 0.0000e+00, ..., -3.9348e-08, + 4.2142e-08, 7.2876e-08], + [ 1.2107e-08, 1.3970e-09, 0.0000e+00, ..., 1.0012e-08, + 1.9092e-08, 2.5611e-08], + ..., + [-9.4902e-07, 1.8626e-09, 0.0000e+00, ..., -6.2259e-07, + -8.0839e-07, -2.3376e-06], + [ 6.0536e-09, 6.9849e-10, 0.0000e+00, ..., 1.3970e-08, + 3.0268e-09, 2.5611e-09], + [ 9.0897e-07, 1.0245e-08, 0.0000e+00, ..., 6.6590e-07, + 7.6089e-07, 2.1402e-06]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0229, -0.0277, 0.0137, -0.0238, 0.0355, 0.0162, 0.0099, -0.0126, + -0.0207, -0.0077], device='cuda:0'), grad: tensor([ 1.8394e-08, 7.4506e-08, 1.0571e-07, -2.2538e-07, 3.2573e-07, + 1.7998e-07, 4.6566e-10, -8.5160e-06, -1.1176e-08, 8.0839e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 220.23, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4894 re_mapping 0.0040 re_causal 0.0122 /// teacc 99.04 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1944, 0.1968, -0.0331, ..., -0.1014, 0.0219, 0.0109], + [ 0.0077, -0.1213, -0.0474, ..., 0.0309, -0.0386, -0.0592], + [-0.1271, -0.2290, 0.0470, ..., -0.0813, -0.0167, -0.1347], + ..., + [ 0.0733, 0.1252, -0.0893, ..., -0.1296, 0.0656, 0.0683], + [ 0.0414, -0.1247, -0.0316, ..., -0.0953, -0.1065, -0.0896], + [-0.1639, -0.0433, -0.0561, ..., 0.1000, -0.2436, 0.0919]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -5.5879e-08, 0.0000e+00, ..., 7.4506e-09, + 1.2107e-08, -4.1910e-09], + [ 1.1176e-08, 3.7253e-09, 0.0000e+00, ..., -1.3970e-08, + 2.0023e-08, 9.6392e-08], + [ 1.6298e-08, 5.1223e-09, 0.0000e+00, ..., 8.3819e-09, + 5.9605e-08, 1.0477e-07], + ..., + [-4.7497e-08, -4.6566e-10, 0.0000e+00, ..., 7.4506e-09, + -2.3749e-08, -2.4680e-07], + [-7.9162e-09, 6.9849e-09, 0.0000e+00, ..., 8.7079e-08, + 2.0722e-07, 1.4016e-07], + [ 1.9558e-08, 1.9558e-08, 0.0000e+00, ..., -5.7323e-07, + 1.3970e-08, -4.9826e-08]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0235, -0.0278, 0.0137, -0.0238, 0.0355, 0.0161, 0.0100, -0.0124, + -0.0209, -0.0079], device='cuda:0'), grad: tensor([ 2.2817e-08, 1.7509e-07, 6.0955e-07, -2.2203e-06, 6.9058e-07, + 3.7719e-07, 5.6345e-08, -5.6904e-07, 1.4650e-06, -6.1793e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 220.40, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4934 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.08 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1959, 0.1970, -0.0335, ..., -0.1017, 0.0218, 0.0109], + [ 0.0077, -0.1216, -0.0474, ..., 0.0308, -0.0387, -0.0593], + [-0.1274, -0.2295, 0.0471, ..., -0.0814, -0.0170, -0.1351], + ..., + [ 0.0733, 0.1253, -0.0894, ..., -0.1296, 0.0657, 0.0684], + [ 0.0402, -0.1253, -0.0317, ..., -0.0956, -0.1081, -0.0901], + [-0.1651, -0.0433, -0.0559, ..., 0.0996, -0.2439, 0.0919]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, -5.8264e-06, -6.2399e-07, ..., 1.3551e-07, + 2.0955e-08, -2.5798e-07], + [ 6.9384e-08, 1.2200e-07, 5.1223e-09, ..., 6.2864e-08, + 8.9873e-08, 1.4761e-07], + [ 7.1246e-08, 3.0035e-07, 2.8405e-08, ..., 6.4261e-08, + 6.9849e-08, 1.2293e-07], + ..., + [-5.4855e-07, -5.6578e-07, 3.7253e-09, ..., 6.3330e-08, + -6.8871e-07, -8.6334e-07], + [-1.8161e-08, 1.1967e-07, 3.7253e-09, ..., 1.7378e-06, + 6.2399e-08, 1.1129e-06], + [ 3.7113e-07, 6.4727e-07, 2.0489e-08, ..., -7.0482e-06, + 3.5157e-07, -3.5185e-06]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0235, -0.0279, 0.0140, -0.0238, 0.0359, 0.0162, 0.0101, -0.0124, + -0.0217, -0.0083], device='cuda:0'), grad: tensor([-1.1809e-05, 9.9186e-07, 1.8273e-06, 1.0105e-06, 7.2382e-06, + 9.6392e-07, 1.3568e-05, -3.7197e-06, 2.3060e-06, -1.2428e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 220.46, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5052 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.04 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1977, 0.1997, -0.0335, ..., -0.1019, 0.0212, 0.0122], + [ 0.0077, -0.1221, -0.0474, ..., 0.0304, -0.0388, -0.0594], + [-0.1286, -0.2314, 0.0471, ..., -0.0817, -0.0178, -0.1362], + ..., + [ 0.0734, 0.1258, -0.0894, ..., -0.1297, 0.0661, 0.0686], + [ 0.0400, -0.1261, -0.0315, ..., -0.0966, -0.1089, -0.0912], + [-0.1651, -0.0436, -0.0559, ..., 0.0998, -0.2442, 0.0919]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -2.6450e-07, 2.7940e-08, ..., 8.9407e-08, + 2.3283e-09, -1.3504e-08], + [-2.8871e-08, 3.8650e-07, 4.6566e-09, ..., 4.6566e-08, + 9.3132e-09, 5.7276e-07], + [ 6.0536e-09, 2.8685e-07, -9.3598e-08, ..., 1.9558e-08, + 3.7253e-09, 3.4925e-07], + ..., + [-1.8626e-08, -1.2275e-06, 9.3132e-09, ..., 4.3772e-08, + -3.0268e-08, -1.8282e-06], + [ 2.6077e-08, 1.3504e-08, 1.9092e-08, ..., 6.5193e-08, + 2.3283e-09, 9.7789e-09], + [ 2.0023e-08, 6.6962e-07, 6.5193e-09, ..., 3.6210e-06, + 9.7789e-09, 8.2050e-07]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0250, -0.0281, 0.0139, -0.0238, 0.0359, 0.0162, 0.0093, -0.0122, + -0.0219, -0.0085], device='cuda:0'), grad: tensor([ 3.3062e-08, 2.0433e-06, 8.7172e-07, 2.3562e-07, -9.1270e-06, + -1.0896e-07, 9.8348e-07, -6.7577e-06, 5.1875e-07, 1.1310e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 219.97, cls_loss 0.0008 cls_loss_mapping 0.0027 cls_loss_causal 0.5222 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.07 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1989, 0.2003, -0.0340, ..., -0.1015, 0.0209, 0.0124], + [ 0.0077, -0.1227, -0.0474, ..., 0.0304, -0.0389, -0.0595], + [-0.1303, -0.2325, 0.0474, ..., -0.0826, -0.0190, -0.1375], + ..., + [ 0.0735, 0.1265, -0.0894, ..., -0.1298, 0.0667, 0.0689], + [ 0.0414, -0.1263, -0.0316, ..., -0.0950, -0.1093, -0.0903], + [-0.1663, -0.0437, -0.0559, ..., 0.0997, -0.2447, 0.0918]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -4.1910e-09, 7.4506e-09, ..., 4.8988e-07, + 5.9605e-08, 9.0338e-08], + [ 2.7940e-09, 1.3970e-09, 9.3132e-10, ..., 1.6857e-07, + 5.7742e-08, 8.5682e-08], + [ 1.8626e-09, 9.3132e-10, 4.6566e-10, ..., 3.7719e-08, + 4.9826e-08, 7.1712e-08], + ..., + [-4.9826e-08, -5.5879e-09, 0.0000e+00, ..., 2.2352e-08, + 8.2888e-08, 4.1910e-08], + [ 4.1910e-09, 0.0000e+00, 1.6764e-08, ..., 6.6496e-07, + 1.6484e-07, 2.4168e-07], + [ 8.2422e-08, 3.7253e-09, 9.3132e-10, ..., 1.4668e-07, + 3.2131e-07, 4.8662e-07]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0253, -0.0282, 0.0133, -0.0238, 0.0362, 0.0162, 0.0087, -0.0119, + -0.0200, -0.0091], device='cuda:0'), grad: tensor([ 2.0526e-06, 8.0653e-07, 2.2724e-07, -2.0933e-04, 3.1432e-07, + 2.0921e-04, -9.2983e-06, 4.7078e-07, 3.0361e-06, 2.5518e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 220.29, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.5315 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.04 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.2004, 0.2003, -0.0345, ..., -0.1016, 0.0208, 0.0124], + [ 0.0076, -0.1236, -0.0477, ..., 0.0304, -0.0390, -0.0599], + [-0.1309, -0.2332, 0.0494, ..., -0.0816, -0.0194, -0.1380], + ..., + [ 0.0736, 0.1271, -0.0894, ..., -0.1299, 0.0668, 0.0693], + [ 0.0411, -0.1268, -0.0321, ..., -0.0960, -0.1098, -0.0906], + [-0.1666, -0.0438, -0.0559, ..., 0.0999, -0.2448, 0.0920]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.3283e-09, 0.0000e+00, ..., 5.0291e-08, + 4.6566e-10, 2.6543e-08], + [ 1.1967e-07, 5.5879e-09, 0.0000e+00, ..., 7.9162e-09, + 3.2596e-09, 3.7346e-07], + [ 3.3993e-08, 2.3283e-09, 0.0000e+00, ..., 1.5832e-08, + 1.3970e-09, 8.7079e-08], + ..., + [-2.1420e-07, -3.7253e-09, 0.0000e+00, ..., 4.1910e-08, + -1.4901e-08, -6.5984e-07], + [-4.7963e-08, -3.5390e-08, 0.0000e+00, ..., -1.6950e-07, + 9.3132e-10, 1.4901e-08], + [ 1.0664e-07, 1.1176e-08, 0.0000e+00, ..., -7.7765e-08, + 1.1176e-08, -3.9581e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0253, -0.0296, 0.0156, -0.0238, 0.0361, 0.0162, 0.0087, -0.0115, + -0.0205, -0.0090], device='cuda:0'), grad: tensor([ 7.0129e-07, 1.6969e-06, 6.3051e-07, 4.4797e-07, 4.7404e-07, + 5.5507e-07, 1.2480e-06, -1.5236e-06, -5.6252e-06, 1.4128e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 220.32, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4969 re_mapping 0.0040 re_causal 0.0122 /// teacc 99.08 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.2019, 0.2004, -0.0347, ..., -0.1018, 0.0207, 0.0122], + [ 0.0076, -0.1237, -0.0477, ..., 0.0305, -0.0392, -0.0600], + [-0.1313, -0.2335, 0.0495, ..., -0.0817, -0.0196, -0.1382], + ..., + [ 0.0737, 0.1273, -0.0894, ..., -0.1301, 0.0671, 0.0694], + [ 0.0410, -0.1271, -0.0323, ..., -0.0965, -0.1103, -0.0911], + [-0.1670, -0.0438, -0.0559, ..., 0.1000, -0.2450, 0.0921]], + device='cuda:0'), grad: tensor([[ 6.8452e-08, -1.4855e-07, 5.3085e-08, ..., 2.1607e-07, + 3.9116e-08, 4.1910e-08], + [ 8.2254e-05, 8.3819e-09, 6.9384e-08, ..., 2.6776e-07, + 6.5804e-05, 1.1146e-04], + [ 2.3702e-07, 6.5193e-09, 7.9162e-09, ..., 3.5856e-08, + 1.9977e-07, 3.4319e-07], + ..., + [-8.3566e-05, -2.3749e-08, 1.3970e-09, ..., -3.1199e-08, + -6.7890e-05, -1.1504e-04], + [-2.2314e-06, 1.2107e-08, 6.5193e-09, ..., 3.0268e-08, + 1.5087e-07, -2.7567e-07], + [ 2.3395e-06, 2.3749e-08, 2.3283e-09, ..., -1.1269e-07, + 1.0142e-06, 2.0675e-06]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0252, -0.0296, 0.0156, -0.0238, 0.0362, 0.0161, 0.0090, -0.0115, + -0.0208, -0.0090], device='cuda:0'), grad: tensor([ 8.7637e-07, 4.7708e-04, 4.3726e-07, 4.7609e-06, 1.4640e-06, + 4.0643e-06, -4.9099e-06, -4.7326e-04, -4.4137e-05, 3.4124e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 220.35, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4945 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.2041, 0.2004, -0.0349, ..., -0.1020, 0.0205, 0.0121], + [ 0.0074, -0.1243, -0.0477, ..., 0.0305, -0.0406, -0.0616], + [-0.1319, -0.2350, 0.0495, ..., -0.0820, -0.0200, -0.1396], + ..., + [ 0.0740, 0.1285, -0.0894, ..., -0.1301, 0.0684, 0.0709], + [ 0.0410, -0.1283, -0.0323, ..., -0.0973, -0.1110, -0.0934], + [-0.1676, -0.0441, -0.0559, ..., 0.1001, -0.2454, 0.0921]], + device='cuda:0'), grad: tensor([[ 1.4435e-08, -1.8207e-07, 0.0000e+00, ..., 3.0315e-07, + -1.6298e-08, -5.7276e-08], + [-2.7940e-08, 1.7695e-08, 0.0000e+00, ..., -1.7229e-08, + 3.2596e-09, 9.3132e-09], + [ 4.1910e-09, 1.3039e-08, 0.0000e+00, ..., 3.1665e-08, + 1.8626e-09, 5.1223e-09], + ..., + [ 4.6566e-10, 3.2131e-08, 0.0000e+00, ..., 1.1176e-08, + -1.2107e-08, -2.0023e-08], + [-1.8626e-08, 1.0245e-08, 0.0000e+00, ..., -2.6077e-08, + 9.3132e-10, 2.7940e-09], + [ 1.4435e-08, 4.2375e-08, 0.0000e+00, ..., 8.8476e-09, + 1.8161e-08, 3.3993e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0251, -0.0308, 0.0155, -0.0238, 0.0361, 0.0160, 0.0096, -0.0099, + -0.0217, -0.0091], device='cuda:0'), grad: tensor([ 5.3924e-07, -1.2107e-07, 1.2480e-07, 1.3690e-07, 6.3628e-06, + 2.2724e-07, -7.1973e-06, 4.4703e-08, -3.1339e-07, 2.1141e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 220.47, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.5087 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.05 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.2048, 0.2005, -0.0355, ..., -0.1022, 0.0204, 0.0120], + [ 0.0074, -0.1244, -0.0477, ..., 0.0306, -0.0407, -0.0639], + [-0.1324, -0.2359, 0.0495, ..., -0.0821, -0.0202, -0.1418], + ..., + [ 0.0741, 0.1290, -0.0894, ..., -0.1302, 0.0691, 0.0731], + [ 0.0415, -0.1287, -0.0314, ..., -0.0973, -0.1116, -0.0936], + [-0.1682, -0.0443, -0.0559, ..., 0.1001, -0.2459, 0.0920]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -5.5879e-08, 0.0000e+00, ..., 4.4238e-08, + 0.0000e+00, -2.4214e-08], + [ 4.6566e-09, 9.3132e-10, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-10, 3.2596e-09], + [ 1.8161e-08, 4.6566e-10, 0.0000e+00, ..., 3.0268e-08, + 3.7253e-09, 9.3132e-09], + ..., + [ 3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 4.0513e-08, + -4.6566e-09, -1.2573e-08], + [ 1.2107e-08, 1.3970e-09, 0.0000e+00, ..., 3.1199e-08, + 3.7253e-09, 8.8476e-09], + [ 4.1863e-07, 3.3528e-08, 0.0000e+00, ..., 9.5041e-07, + 1.8626e-09, -3.4925e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0250, -0.0321, 0.0153, -0.0239, 0.0361, 0.0160, 0.0096, -0.0082, + -0.0207, -0.0094], device='cuda:0'), grad: tensor([ 8.8476e-09, -7.8231e-08, 1.2852e-07, 7.1246e-08, -2.6673e-06, + 2.5611e-08, 4.0652e-07, 6.7521e-08, 8.8476e-08, 1.9409e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 220.23, cls_loss 0.0009 cls_loss_mapping 0.0035 cls_loss_causal 0.4867 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.03 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.2052, 0.2009, -0.0356, ..., -0.1021, 0.0203, 0.0124], + [ 0.0074, -0.1246, -0.0477, ..., 0.0307, -0.0408, -0.0640], + [-0.1327, -0.2363, 0.0496, ..., -0.0822, -0.0203, -0.1423], + ..., + [ 0.0741, 0.1295, -0.0894, ..., -0.1300, 0.0681, 0.0732], + [ 0.0414, -0.1295, -0.0313, ..., -0.0956, -0.1097, -0.0913], + [-0.1695, -0.0445, -0.0559, ..., 0.1003, -0.2474, 0.0914]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -3.3528e-08, 9.3132e-10, ..., 4.3772e-08, + 2.3283e-09, -5.1223e-09], + [ 1.5832e-08, 4.1910e-09, 1.3970e-08, ..., -2.0303e-07, + -4.1910e-09, -1.9558e-08], + [-2.1886e-07, -4.1910e-09, -1.3923e-07, ..., 1.1176e-08, + -2.7474e-08, 3.7253e-09], + ..., + [ 1.7229e-08, 1.8626e-09, 1.2573e-08, ..., 7.8231e-08, + 9.7789e-09, 3.7719e-08], + [ 1.7323e-07, 7.4506e-09, 1.0664e-07, ..., 2.5146e-08, + 7.9162e-09, 9.7789e-09], + [ 2.3283e-09, 1.1642e-08, 0.0000e+00, ..., -4.7963e-08, + 7.9162e-09, -9.9652e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0253, -0.0321, 0.0154, -0.0236, 0.0356, 0.0159, 0.0091, -0.0083, + -0.0173, -0.0104], device='cuda:0'), grad: tensor([ 2.4540e-07, -5.5647e-07, -2.5518e-06, 3.2363e-07, 8.3353e-08, + 6.4727e-08, 1.1642e-08, 8.8476e-07, 1.4286e-06, 6.5658e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 220.52, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4994 re_mapping 0.0039 re_causal 0.0124 /// teacc 99.01 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.2056, 0.2009, -0.0358, ..., -0.1023, 0.0203, 0.0121], + [ 0.0075, -0.1247, -0.0477, ..., 0.0308, -0.0409, -0.0639], + [-0.1332, -0.2367, 0.0496, ..., -0.0823, -0.0207, -0.1425], + ..., + [ 0.0741, 0.1296, -0.0894, ..., -0.1303, 0.0678, 0.0731], + [ 0.0413, -0.1303, -0.0314, ..., -0.0964, -0.1100, -0.0915], + [-0.1702, -0.0445, -0.0560, ..., 0.1011, -0.2476, 0.0922]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, -2.8405e-08, -1.3970e-09, ..., 4.0978e-08, + 3.7253e-09, -1.0245e-08], + [ 3.8184e-08, 9.3132e-10, 0.0000e+00, ..., 4.1258e-07, + 4.1910e-09, 3.7253e-09], + [ 9.7789e-09, 1.3970e-09, 0.0000e+00, ..., 1.5367e-08, + 4.0047e-08, 4.7497e-08], + ..., + [ 2.5146e-08, 1.8626e-09, 0.0000e+00, ..., 3.3528e-08, + 3.4459e-08, 4.2375e-08], + [ 8.6147e-08, 3.2596e-09, 0.0000e+00, ..., 9.3132e-08, + 1.3551e-07, 1.4296e-07], + [ 1.1595e-07, 1.7695e-08, 9.3132e-10, ..., 1.2526e-07, + 1.9092e-08, -3.7253e-09]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0252, -0.0320, 0.0155, -0.0237, 0.0357, 0.0159, 0.0093, -0.0085, + -0.0174, -0.0099], device='cuda:0'), grad: tensor([ 1.1781e-07, 1.0952e-06, 1.8999e-07, 9.3924e-07, -9.2760e-07, + -2.8107e-06, -3.7206e-07, 2.4214e-07, 8.9686e-07, 6.3516e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 220.51, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.5300 re_mapping 0.0040 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.2063, 0.2010, -0.0359, ..., -0.1026, 0.0202, 0.0121], + [ 0.0075, -0.1249, -0.0477, ..., 0.0309, -0.0410, -0.0640], + [-0.1337, -0.2377, 0.0517, ..., -0.0824, -0.0214, -0.1400], + ..., + [ 0.0741, 0.1298, -0.0900, ..., -0.1305, 0.0680, 0.0730], + [ 0.0412, -0.1311, -0.0315, ..., -0.0972, -0.1107, -0.0918], + [-0.1707, -0.0445, -0.0560, ..., 0.1013, -0.2478, 0.0924]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -2.1681e-06, 0.0000e+00, ..., -2.9011e-07, + 4.0978e-08, 1.8487e-07], + [-1.2107e-08, 5.5879e-09, 0.0000e+00, ..., 4.0280e-07, + 6.8452e-08, 3.0780e-07], + [ 1.8626e-09, 6.1002e-08, 0.0000e+00, ..., 1.0617e-07, + 2.8405e-08, 8.1025e-08], + ..., + [ 5.1223e-09, -8.3819e-09, 0.0000e+00, ..., 1.0148e-05, + 1.1362e-06, 6.5155e-06], + [ 9.7789e-09, 2.8219e-07, 0.0000e+00, ..., 6.8126e-07, + 8.4285e-08, 4.2887e-07], + [ 3.7253e-09, 1.5553e-07, 0.0000e+00, ..., -5.3495e-05, + -7.3500e-06, -3.7402e-05]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0250, -0.0320, 0.0184, -0.0238, 0.0356, 0.0159, 0.0100, -0.0105, + -0.0181, -0.0099], device='cuda:0'), grad: tensor([-5.2005e-06, 1.2740e-06, 4.0373e-07, 2.5351e-06, 1.1933e-04, + -1.8040e-06, 7.2047e-06, 2.8446e-05, 2.0508e-06, -1.5414e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 220.36, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.4884 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.11 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.2068, 0.2012, -0.0359, ..., -0.1026, 0.0201, 0.0122], + [ 0.0075, -0.1251, -0.0477, ..., 0.0309, -0.0411, -0.0640], + [-0.1340, -0.2384, 0.0517, ..., -0.0824, -0.0217, -0.1404], + ..., + [ 0.0741, 0.1298, -0.0900, ..., -0.1306, 0.0681, 0.0730], + [ 0.0411, -0.1311, -0.0316, ..., -0.0977, -0.1111, -0.0921], + [-0.1709, -0.0447, -0.0559, ..., 0.1017, -0.2477, 0.0927]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, -1.9092e-08, 0.0000e+00, ..., 2.2352e-08, + 2.3283e-08, 4.4238e-08], + [-3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 2.0023e-08, + 2.2352e-08, 4.8894e-08], + [ 5.1223e-09, 1.8626e-09, 0.0000e+00, ..., 6.9849e-09, + 1.1176e-08, 2.3283e-08], + ..., + [ 1.8626e-09, -2.3283e-09, 0.0000e+00, ..., 7.7300e-08, + 2.0489e-08, 4.8894e-08], + [-5.1223e-09, 9.3132e-10, 0.0000e+00, ..., 7.4506e-08, + 1.2107e-08, 5.7276e-08], + [ 2.2352e-08, 8.3819e-09, 0.0000e+00, ..., -1.4156e-06, + 2.3749e-08, -7.1712e-07]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0250, -0.0320, 0.0188, -0.0238, 0.0353, 0.0160, 0.0099, -0.0107, + -0.0184, -0.0097], device='cuda:0'), grad: tensor([ 1.7416e-07, 1.6298e-07, 3.4925e-08, -2.4270e-06, 2.6077e-06, + 1.8869e-06, 9.7323e-08, 3.1572e-07, 1.9604e-07, -3.0175e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 220.41, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4965 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.10 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.2077, 0.2012, -0.0360, ..., -0.1026, 0.0200, 0.0122], + [ 0.0079, -0.1251, -0.0475, ..., 0.0320, -0.0409, -0.0638], + [-0.1342, -0.2387, 0.0517, ..., -0.0825, -0.0220, -0.1407], + ..., + [ 0.0738, 0.1299, -0.0901, ..., -0.1317, 0.0677, 0.0729], + [ 0.0409, -0.1315, -0.0316, ..., -0.0982, -0.1117, -0.0925], + [-0.1716, -0.0447, -0.0562, ..., 0.1017, -0.2480, 0.0927]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 2.3283e-09, 0.0000e+00, ..., 2.0489e-08, + 2.3283e-09, 1.4435e-08], + [ 2.3283e-09, 3.3062e-08, 0.0000e+00, ..., -1.2573e-08, + 2.0023e-08, 6.1002e-08], + [ 1.5367e-08, 4.1910e-09, 0.0000e+00, ..., 1.7695e-08, + 6.0536e-09, 1.2107e-08], + ..., + [-8.0094e-08, -1.5227e-07, 0.0000e+00, ..., 4.6566e-08, + -8.7079e-08, -2.0815e-07], + [ 8.8476e-09, 1.0245e-08, 0.0000e+00, ..., 1.0710e-08, + 9.3132e-09, 2.6077e-08], + [ 4.7032e-08, 4.4703e-08, 0.0000e+00, ..., -2.8983e-06, + 2.5146e-08, -2.1160e-06]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0250, -0.0317, 0.0188, -0.0240, 0.0355, 0.0163, 0.0099, -0.0110, + -0.0188, -0.0099], device='cuda:0'), grad: tensor([ 7.7300e-08, 2.6077e-08, 3.8184e-07, -1.4761e-07, 5.5209e-06, + 6.6264e-07, 4.1910e-08, -3.5251e-07, -1.6615e-06, -4.5188e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 220.65, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4997 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.04 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.2089, 0.2013, -0.0361, ..., -0.1028, 0.0199, 0.0121], + [ 0.0079, -0.1252, -0.0475, ..., 0.0320, -0.0411, -0.0639], + [-0.1351, -0.2392, 0.0518, ..., -0.0826, -0.0225, -0.1413], + ..., + [ 0.0738, 0.1301, -0.0901, ..., -0.1318, 0.0680, 0.0730], + [ 0.0394, -0.1320, -0.0317, ..., -0.0990, -0.1124, -0.0929], + [-0.1720, -0.0448, -0.0562, ..., 0.1020, -0.2483, 0.0929]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.1297e-06, 0.0000e+00, ..., 5.1223e-09, + -1.2573e-08, -1.8580e-07], + [ 8.3819e-08, 1.1781e-07, 0.0000e+00, ..., -5.5879e-09, + 1.1548e-07, 1.7928e-07], + [ 7.4506e-09, 1.7229e-08, 0.0000e+00, ..., 2.3283e-09, + 1.0245e-08, 1.7695e-08], + ..., + [-1.0245e-07, 1.3970e-08, 0.0000e+00, ..., 1.1176e-08, + -1.1222e-07, -1.6904e-07], + [-1.7695e-08, 4.2375e-08, 0.0000e+00, ..., 1.8626e-09, + 2.0955e-08, 3.3062e-08], + [ 7.4506e-09, 6.1467e-08, 0.0000e+00, ..., -1.3970e-08, + 1.3504e-08, 2.2352e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([ 0.0249, -0.0317, 0.0188, -0.0241, 0.0352, 0.0165, 0.0094, -0.0110, + -0.0198, -0.0098], device='cuda:0'), grad: tensor([-2.8666e-06, 6.9942e-07, 9.3598e-08, 9.3132e-09, 6.0070e-08, + 1.0058e-06, 1.1791e-06, -3.6415e-07, 6.0536e-09, 1.8347e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 220.38, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.5048 re_mapping 0.0038 re_causal 0.0116 /// teacc 99.09 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.2104, 0.2016, -0.0363, ..., -0.1028, 0.0200, 0.0126], + [ 0.0079, -0.1255, -0.0474, ..., 0.0317, -0.0412, -0.0644], + [-0.1357, -0.2411, 0.0525, ..., -0.0826, -0.0227, -0.1419], + ..., + [ 0.0738, 0.1304, -0.0904, ..., -0.1323, 0.0677, 0.0733], + [ 0.0394, -0.1328, -0.0318, ..., -0.1005, -0.1113, -0.0928], + [-0.1722, -0.0453, -0.0565, ..., 0.1022, -0.2486, 0.0931]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, -6.9384e-08, 0.0000e+00, ..., 1.9139e-07, + -7.4506e-09, 9.5461e-08], + [-3.5856e-08, 7.4506e-09, 0.0000e+00, ..., 2.1886e-08, + 3.7253e-09, 2.3283e-08], + [ 6.2399e-08, 2.6077e-08, -9.7789e-09, ..., 7.7765e-08, + 1.0710e-08, 4.8894e-08], + ..., + [ 1.4808e-07, -3.8184e-08, 9.3132e-10, ..., 2.8452e-07, + -2.4680e-08, -4.8894e-08], + [ 3.4925e-08, 4.6566e-09, 3.2596e-09, ..., 1.9046e-07, + 2.7940e-09, 1.1874e-07], + [ 4.7311e-07, 2.0955e-08, 0.0000e+00, ..., -4.2981e-07, + 6.0536e-09, -8.8988e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0250, -0.0320, 0.0189, -0.0241, 0.0353, 0.0166, 0.0097, -0.0110, + -0.0198, -0.0097], device='cuda:0'), grad: tensor([ 4.1723e-07, -5.1223e-08, 2.8545e-07, 1.0366e-06, -2.8536e-06, + 1.0878e-06, -1.0757e-07, 5.7183e-07, 5.8720e-07, -9.8441e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 220.33, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4870 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.04 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.2118, 0.2018, -0.0366, ..., -0.1029, 0.0200, 0.0127], + [ 0.0079, -0.1257, -0.0474, ..., 0.0316, -0.0413, -0.0645], + [-0.1366, -0.2420, 0.0525, ..., -0.0827, -0.0230, -0.1423], + ..., + [ 0.0738, 0.1305, -0.0904, ..., -0.1326, 0.0675, 0.0732], + [ 0.0393, -0.1336, -0.0321, ..., -0.1023, -0.1116, -0.0936], + [-0.1734, -0.0454, -0.0565, ..., 0.1024, -0.2488, 0.0935]], + device='cuda:0'), grad: tensor([[ 3.7253e-08, -1.5832e-07, 4.6566e-10, ..., 8.3819e-08, + -6.0536e-09, 5.5879e-09], + [ 2.7660e-07, 4.6566e-09, 0.0000e+00, ..., 3.3760e-07, + 3.1665e-08, 6.3330e-08], + [ 2.3283e-08, 2.1886e-08, -1.8626e-09, ..., 4.2375e-08, + 7.3109e-08, 4.7497e-07], + ..., + [ 1.0850e-07, -9.3132e-10, 9.3132e-10, ..., 3.2969e-07, + 7.9162e-09, -2.7614e-07], + [ 2.8871e-08, 1.8626e-08, 4.6566e-10, ..., 3.5390e-08, + 3.6322e-08, 3.4459e-08], + [ 3.0547e-07, 7.8697e-08, 0.0000e+00, ..., -4.0280e-07, + 8.8476e-09, -8.2888e-07]], device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0250, -0.0321, 0.0190, -0.0242, 0.0357, 0.0166, 0.0097, -0.0112, + -0.0206, -0.0096], device='cuda:0'), grad: tensor([-1.2107e-08, 9.8720e-07, 1.2256e-06, 9.3132e-10, -1.3933e-06, + -1.6484e-07, 3.8650e-07, -1.3690e-07, 3.7625e-07, -1.2787e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 220.18, cls_loss 0.0007 cls_loss_mapping 0.0030 cls_loss_causal 0.5050 re_mapping 0.0039 re_causal 0.0114 /// teacc 99.09 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.2136, 0.2022, -0.0368, ..., -0.1032, 0.0212, 0.0131], + [ 0.0080, -0.1261, -0.0474, ..., 0.0316, -0.0414, -0.0646], + [-0.1374, -0.2435, 0.0525, ..., -0.0829, -0.0233, -0.1441], + ..., + [ 0.0738, 0.1304, -0.0904, ..., -0.1329, 0.0672, 0.0733], + [ 0.0392, -0.1351, -0.0322, ..., -0.1028, -0.1120, -0.0940], + [-0.1756, -0.0456, -0.0561, ..., 0.1023, -0.2491, 0.0937]], + device='cuda:0'), grad: tensor([[ 1.2871e-06, 1.2219e-06, 0.0000e+00, ..., 1.6447e-06, + 4.6566e-10, 9.7789e-08], + [-1.7975e-06, -1.5916e-06, 0.0000e+00, ..., -1.8682e-06, + 1.8626e-09, 1.7835e-07], + [ 7.3574e-08, 2.5611e-08, 0.0000e+00, ..., 7.0315e-08, + 2.3283e-09, 1.8161e-08], + ..., + [ 1.0375e-06, 2.9802e-08, 0.0000e+00, ..., 8.2254e-06, + 3.2596e-09, 6.6459e-06], + [ 6.0536e-09, 3.1199e-08, 0.0000e+00, ..., 8.4750e-08, + 9.3132e-10, 3.1199e-08], + [-2.2799e-06, 1.0245e-08, 0.0000e+00, ..., -2.0370e-05, + 1.3970e-09, -1.6674e-05]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0251, -0.0321, 0.0190, -0.0241, 0.0362, 0.0165, 0.0096, -0.0112, + -0.0207, -0.0099], device='cuda:0'), grad: tensor([ 7.6443e-06, -9.3952e-06, 3.9395e-07, 1.6391e-07, 3.1292e-05, + 2.1886e-08, 1.4426e-06, 2.1800e-05, -9.9652e-08, -5.3227e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 220.24, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4875 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.18 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.2169, 0.2023, -0.0370, ..., -0.1033, 0.0211, 0.0131], + [ 0.0082, -0.1264, -0.0474, ..., 0.0318, -0.0416, -0.0646], + [-0.1400, -0.2463, 0.0524, ..., -0.0831, -0.0246, -0.1448], + ..., + [ 0.0739, 0.1312, -0.0903, ..., -0.1348, 0.0676, 0.0726], + [ 0.0391, -0.1356, -0.0324, ..., -0.1033, -0.1118, -0.0940], + [-0.1766, -0.0458, -0.0560, ..., 0.1034, -0.2494, 0.0951]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -4.9220e-07, 0.0000e+00, ..., 1.3895e-06, + -5.0757e-08, -2.2771e-07], + [-2.0489e-08, 1.0710e-08, 0.0000e+00, ..., -5.5879e-09, + 5.1223e-09, 9.7789e-09], + [ 2.3283e-09, 1.3551e-07, 0.0000e+00, ..., 5.5879e-09, + 2.1420e-08, 7.1246e-08], + ..., + [ 5.1223e-09, 8.8476e-09, 0.0000e+00, ..., 1.1642e-08, + 5.1223e-09, 4.6566e-09], + [ 5.5879e-09, 3.4459e-08, 0.0000e+00, ..., 4.3772e-08, + 1.3970e-08, 2.7474e-08], + [ 4.8894e-08, 1.9465e-07, 0.0000e+00, ..., 1.0431e-07, + 2.8405e-08, 8.7079e-08]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0251, -0.0320, 0.0188, -0.0241, 0.0361, 0.0165, 0.0096, -0.0117, + -0.0206, -0.0086], device='cuda:0'), grad: tensor([ 2.4661e-06, 2.1886e-08, 3.1525e-07, 1.9884e-07, -1.1828e-07, + -9.4529e-08, -3.7402e-06, 6.6124e-08, 2.2305e-07, 6.7474e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 220.69, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4950 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.06 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.2186, 0.2024, -0.0374, ..., -0.1036, 0.0208, 0.0132], + [ 0.0081, -0.1267, -0.0478, ..., 0.0311, -0.0417, -0.0647], + [-0.1434, -0.2490, 0.0538, ..., -0.0821, -0.0271, -0.1470], + ..., + [ 0.0743, 0.1319, -0.0903, ..., -0.1350, 0.0682, 0.0729], + [ 0.0400, -0.1366, -0.0326, ..., -0.1036, -0.1120, -0.0942], + [-0.1779, -0.0460, -0.0558, ..., 0.1037, -0.2497, 0.0954]], + device='cuda:0'), grad: tensor([[ 2.7474e-08, -5.9605e-08, 4.6566e-09, ..., 1.0245e-07, + 4.6566e-10, -7.9162e-09], + [-4.7963e-08, -1.0151e-07, 1.8626e-09, ..., -2.1653e-07, + 1.8626e-09, 5.1223e-09], + [ 8.8476e-09, 1.6764e-08, -4.6566e-10, ..., 3.2131e-08, + 4.6566e-10, 9.3132e-10], + ..., + [ 2.0023e-08, 5.1223e-09, 9.3132e-10, ..., 2.4680e-08, + 5.1223e-09, 1.4435e-08], + [-5.0291e-08, 2.1420e-08, 1.3970e-09, ..., 1.9372e-07, + 3.2596e-09, 9.7789e-08], + [ 9.0338e-08, 4.7497e-08, 0.0000e+00, ..., -6.1933e-08, + 9.3132e-10, -1.2619e-07]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0249, -0.0324, 0.0190, -0.0240, 0.0358, 0.0165, 0.0097, -0.0115, + -0.0201, -0.0086], device='cuda:0'), grad: tensor([ 2.2957e-07, -1.0151e-06, 3.3015e-07, -9.8199e-06, -4.6659e-07, + 1.0043e-05, 5.1409e-07, 1.7881e-07, 7.4971e-08, -4.7963e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 220.80, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.5090 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.09 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.2200, 0.2024, -0.0377, ..., -0.1037, 0.0207, 0.0131], + [ 0.0082, -0.1271, -0.0481, ..., 0.0310, -0.0420, -0.0648], + [-0.1457, -0.2507, 0.0549, ..., -0.0815, -0.0290, -0.1489], + ..., + [ 0.0745, 0.1330, -0.0904, ..., -0.1352, 0.0691, 0.0732], + [ 0.0398, -0.1372, -0.0326, ..., -0.1030, -0.1137, -0.0952], + [-0.1789, -0.0461, -0.0557, ..., 0.1041, -0.2499, 0.0957]], + device='cuda:0'), grad: tensor([[ 7.6834e-08, 1.8161e-07, 3.4925e-08, ..., 3.5251e-07, + 4.6566e-10, 2.6636e-07], + [ 1.9791e-07, 4.8149e-07, 6.9849e-09, ..., 8.2608e-07, + -1.7695e-08, 7.1153e-07], + [ 4.1444e-08, 2.3749e-08, -2.5146e-08, ..., 1.6810e-07, + 4.6566e-09, 4.4238e-08], + ..., + [-1.0785e-06, -5.8636e-06, 0.0000e+00, ..., 1.2573e-07, + 2.3283e-09, -8.6650e-06], + [ 4.3306e-08, 1.6252e-07, 2.0815e-07, ..., 6.3470e-07, + 4.6566e-09, 2.4680e-07], + [ 1.0561e-06, 4.8131e-06, 0.0000e+00, ..., 3.3993e-07, + 3.2596e-09, 7.0967e-06]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0248, -0.0328, 0.0194, -0.0243, 0.0355, 0.0167, 0.0090, -0.0112, + -0.0190, -0.0085], device='cuda:0'), grad: tensor([ 2.3432e-06, 3.9898e-06, 2.2724e-07, 8.5216e-08, -3.2224e-06, + 7.7114e-07, -4.1835e-06, -3.2395e-05, 4.5896e-06, 2.7761e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 220.45, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4912 re_mapping 0.0037 re_causal 0.0116 /// teacc 99.13 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.2211, 0.2027, -0.0381, ..., -0.1039, 0.0205, 0.0132], + [ 0.0082, -0.1274, -0.0481, ..., 0.0311, -0.0421, -0.0649], + [-0.1469, -0.2513, 0.0549, ..., -0.0816, -0.0294, -0.1494], + ..., + [ 0.0746, 0.1335, -0.0904, ..., -0.1355, 0.0691, 0.0733], + [ 0.0397, -0.1380, -0.0329, ..., -0.1034, -0.1149, -0.0959], + [-0.1792, -0.0463, -0.0556, ..., 0.1045, -0.2501, 0.0959]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 0.0000e+00, 9.3132e-10, ..., 2.1420e-08, + 1.3970e-09, 2.7940e-09], + [ 1.6913e-05, 1.6298e-09, 5.7276e-08, ..., 2.7850e-05, + 1.2573e-08, 2.6310e-08], + [ 8.8476e-09, 2.3283e-10, 4.6566e-10, ..., 1.4901e-08, + 2.0955e-09, 2.0955e-09], + ..., + [ 6.5193e-09, -4.8894e-09, 4.6566e-10, ..., 4.4238e-08, + -1.4203e-08, -6.4494e-08], + [ 5.3551e-09, 4.6566e-10, 2.3283e-10, ..., 1.3970e-09, + 7.6834e-09, 9.3132e-09], + [ 1.7998e-07, 1.8626e-09, 2.0955e-09, ..., 2.6892e-07, + 2.3516e-08, 8.8476e-09]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0249, -0.0328, 0.0193, -0.0242, 0.0352, 0.0167, 0.0090, -0.0112, + -0.0192, -0.0083], device='cuda:0'), grad: tensor([ 6.7055e-08, 5.5432e-05, -4.2375e-08, 4.0412e-05, -5.6326e-05, + -4.0323e-05, 2.4401e-07, 1.0012e-08, 3.7253e-08, 6.6124e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 220.45, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5173 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.15 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.2236, 0.2029, -0.0389, ..., -0.1041, 0.0203, 0.0133], + [ 0.0080, -0.1278, -0.0483, ..., 0.0305, -0.0422, -0.0649], + [-0.1477, -0.2521, 0.0549, ..., -0.0817, -0.0298, -0.1498], + ..., + [ 0.0747, 0.1337, -0.0904, ..., -0.1356, 0.0692, 0.0735], + [ 0.0393, -0.1380, -0.0323, ..., -0.1034, -0.1158, -0.0965], + [-0.1810, -0.0464, -0.0558, ..., 0.1042, -0.2506, 0.0958]], + device='cuda:0'), grad: tensor([[ 1.4668e-08, 1.2759e-07, 0.0000e+00, ..., 1.2927e-06, + 1.1409e-08, 1.4459e-07], + [-1.0477e-08, 9.0804e-09, 0.0000e+00, ..., 8.5915e-08, + 5.8208e-09, 6.4028e-08], + [ 3.7253e-09, 1.3970e-09, 0.0000e+00, ..., 1.8859e-08, + 6.5193e-09, 1.2107e-08], + ..., + [ 1.7462e-08, 3.9581e-09, 0.0000e+00, ..., 1.0803e-07, + 1.8394e-08, 9.2667e-08], + [ 5.8208e-09, 7.6834e-09, 0.0000e+00, ..., 1.5600e-07, + 4.8894e-09, 1.5832e-07], + [ 1.3341e-07, -7.8697e-08, 0.0000e+00, ..., -8.0094e-07, + 2.9802e-08, -1.1288e-06]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0248, -0.0330, 0.0194, -0.0242, 0.0358, 0.0168, 0.0089, -0.0112, + -0.0185, -0.0088], device='cuda:0'), grad: tensor([ 5.2787e-06, -1.5376e-06, 1.7649e-06, 1.3616e-06, -2.7730e-07, + -4.0536e-07, -4.9137e-06, 4.8289e-07, 4.3935e-07, -2.1458e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 220.37, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4972 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.14 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.2248, 0.2028, -0.0392, ..., -0.1041, 0.0202, 0.0113], + [ 0.0080, -0.1281, -0.0484, ..., 0.0304, -0.0424, -0.0655], + [-0.1485, -0.2524, 0.0549, ..., -0.0818, -0.0302, -0.1502], + ..., + [ 0.0749, 0.1341, -0.0904, ..., -0.1357, 0.0702, 0.0741], + [ 0.0391, -0.1385, -0.0324, ..., -0.1036, -0.1166, -0.0970], + [-0.1816, -0.0457, -0.0558, ..., 0.1045, -0.2512, 0.0966]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.3609e-07, 0.0000e+00, ..., 4.1910e-09, + 2.9104e-08, 4.6566e-10], + [ 6.7521e-09, 3.7719e-08, 0.0000e+00, ..., 7.2177e-09, + 4.6566e-08, 5.1688e-08], + [ 9.0804e-09, 3.8883e-08, 0.0000e+00, ..., 1.5367e-08, + 3.7486e-08, 4.3539e-08], + ..., + [ 1.1642e-09, -1.7299e-07, 0.0000e+00, ..., 2.0722e-08, + -6.2166e-08, -1.0757e-07], + [-3.0268e-09, 2.5611e-09, 0.0000e+00, ..., 7.4506e-09, + 6.5193e-09, 9.5461e-09], + [ 2.8638e-08, 1.5949e-07, 0.0000e+00, ..., 1.4203e-08, + 7.2876e-08, 9.3132e-08]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0239, -0.0333, 0.0197, -0.0243, 0.0355, 0.0168, 0.0091, -0.0109, + -0.0197, -0.0080], device='cuda:0'), grad: tensor([-3.2224e-07, 2.4168e-07, 3.1409e-07, -8.0559e-07, -3.1432e-07, + 5.3085e-07, 2.6636e-07, -2.5937e-07, -1.9814e-07, 5.5507e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 220.81, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4981 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.11 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.2259, 0.2034, -0.0364, ..., -0.1041, 0.0225, 0.0124], + [ 0.0080, -0.1289, -0.0484, ..., 0.0304, -0.0428, -0.0657], + [-0.1494, -0.2536, 0.0550, ..., -0.0817, -0.0315, -0.1510], + ..., + [ 0.0751, 0.1351, -0.0904, ..., -0.1358, 0.0709, 0.0743], + [ 0.0391, -0.1400, -0.0329, ..., -0.1039, -0.1175, -0.0973], + [-0.1820, -0.0463, -0.0559, ..., 0.1047, -0.2521, 0.0965]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, -2.6310e-08, 2.7940e-09, ..., 6.4727e-08, + 1.6298e-09, 6.9849e-10], + [-2.5611e-07, 4.8894e-09, -1.4435e-08, ..., -1.5087e-07, + 7.6834e-09, 7.4506e-09], + [ 5.1921e-08, 6.0536e-09, -3.0501e-08, ..., 8.6147e-09, + 2.5611e-08, 1.3504e-08], + ..., + [ 9.3132e-10, -2.0955e-08, 6.9849e-09, ..., 4.8196e-08, + 6.9849e-10, -1.3271e-08], + [-2.3283e-10, 1.6298e-09, 9.3132e-10, ..., 1.6787e-07, + 5.3318e-08, 3.3062e-08], + [ 8.1491e-09, 3.6322e-08, 0.0000e+00, ..., -7.8138e-07, + 2.0023e-08, -2.6310e-08]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0246, -0.0333, 0.0196, -0.0242, 0.0353, 0.0166, 0.0092, -0.0108, + -0.0195, -0.0081], device='cuda:0'), grad: tensor([ 1.6694e-07, -5.4296e-07, 5.5181e-08, -1.1353e-06, 1.0096e-06, + 1.6838e-06, 8.8243e-08, 1.1269e-07, 5.9884e-07, -2.0154e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 220.78, cls_loss 0.0006 cls_loss_mapping 0.0024 cls_loss_causal 0.4820 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.07 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.2268, 0.2035, -0.0364, ..., -0.1043, 0.0221, 0.0124], + [ 0.0082, -0.1281, -0.0484, ..., 0.0306, -0.0429, -0.0656], + [-0.1503, -0.2544, 0.0551, ..., -0.0821, -0.0339, -0.1526], + ..., + [ 0.0751, 0.1351, -0.0904, ..., -0.1361, 0.0710, 0.0743], + [ 0.0386, -0.1416, -0.0332, ..., -0.1049, -0.1180, -0.0975], + [-0.1830, -0.0466, -0.0559, ..., 0.1039, -0.2527, 0.0964]], + device='cuda:0'), grad: tensor([[ 5.3551e-09, -5.8562e-06, 2.4680e-08, ..., 2.8871e-08, + 1.3970e-09, -2.3339e-06], + [-2.5611e-08, 7.4040e-08, 4.9826e-08, ..., -4.0047e-08, + 2.5379e-08, 1.1083e-07], + [ 3.8650e-08, 1.1642e-08, -5.8115e-06, ..., 1.9325e-08, + 1.2573e-08, 3.6787e-08], + ..., + [-7.4040e-08, 2.7264e-07, 5.0617e-07, ..., 4.0280e-08, + -6.0769e-08, -4.5868e-08], + [-2.7474e-08, 2.8173e-08, 3.0035e-08, ..., 3.0734e-08, + 2.3283e-09, 2.2119e-08], + [ 3.4226e-08, 3.6843e-06, 3.3760e-08, ..., -5.6345e-08, + 1.6298e-08, 1.4696e-06]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0246, -0.0332, 0.0193, -0.0241, 0.0367, 0.0166, 0.0093, -0.0108, + -0.0201, -0.0091], device='cuda:0'), grad: tensor([-9.0078e-06, 6.6031e-07, -2.7716e-05, 2.4274e-05, 2.5937e-07, + 1.3988e-06, 1.1362e-06, 2.4717e-06, 2.9220e-07, 6.1691e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 220.55, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4831 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.16 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.2286, 0.2038, -0.0364, ..., -0.1045, 0.0220, 0.0124], + [ 0.0083, -0.1279, -0.0484, ..., 0.0306, -0.0430, -0.0657], + [-0.1532, -0.2570, 0.0553, ..., -0.0824, -0.0343, -0.1544], + ..., + [ 0.0751, 0.1356, -0.0904, ..., -0.1367, 0.0704, 0.0743], + [ 0.0390, -0.1436, -0.0336, ..., -0.1057, -0.1185, -0.0978], + [-0.1841, -0.0468, -0.0557, ..., 0.1045, -0.2526, 0.0971]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -5.7044e-08, 0.0000e+00, ..., 1.6298e-08, + 5.3551e-09, -4.6566e-09], + [-3.9581e-09, 5.1223e-09, 0.0000e+00, ..., -1.3271e-08, + 2.0489e-08, 2.8173e-08], + [ 8.1491e-09, 2.7940e-09, 0.0000e+00, ..., 1.3504e-08, + 3.7253e-09, 5.3551e-09], + ..., + [-4.7730e-08, -2.3050e-08, 0.0000e+00, ..., 6.6590e-08, + -8.9174e-08, -5.3085e-08], + [-5.8208e-09, 3.0268e-09, 0.0000e+00, ..., 7.2177e-09, + 4.1910e-09, 8.1491e-09], + [ 7.9162e-08, 3.6787e-08, 0.0000e+00, ..., 4.2492e-07, + 3.0734e-08, -5.3318e-08]], device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0246, -0.0332, 0.0193, -0.0241, 0.0363, 0.0166, 0.0096, -0.0110, + -0.0203, -0.0085], device='cuda:0'), grad: tensor([ 1.0245e-08, 9.0105e-08, -7.4180e-07, 5.2527e-07, -1.5274e-06, + 4.5169e-08, 7.0082e-08, -4.6100e-08, 2.0489e-07, 1.3784e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 220.85, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.4940 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.2298, 0.2041, -0.0364, ..., -0.1050, 0.0220, 0.0125], + [ 0.0087, -0.1244, -0.0484, ..., 0.0304, -0.0425, -0.0652], + [-0.1545, -0.2576, 0.0554, ..., -0.0827, -0.0347, -0.1548], + ..., + [ 0.0745, 0.1338, -0.0904, ..., -0.1380, 0.0707, 0.0736], + [ 0.0409, -0.1462, -0.0339, ..., -0.1062, -0.1168, -0.0961], + [-0.1854, -0.0471, -0.0554, ..., 0.1054, -0.2533, 0.0979]], + device='cuda:0'), grad: tensor([[ 1.4924e-07, 1.1642e-09, 2.3283e-10, ..., 3.1106e-07, + 9.3132e-10, 3.7486e-08], + [ 1.4054e-06, -3.1432e-08, 2.3283e-10, ..., 2.7521e-07, + 1.1176e-08, 1.1083e-06], + [ 4.0676e-07, 4.4238e-09, -4.6566e-10, ..., 1.2778e-06, + 3.0501e-08, 3.2061e-07], + ..., + [-1.9334e-06, 1.0012e-08, 0.0000e+00, ..., 2.8429e-07, + -1.0408e-07, -2.0359e-06], + [ 1.7905e-07, 6.5193e-09, 0.0000e+00, ..., 2.4540e-07, + 1.8626e-09, -2.1071e-07], + [ 5.1856e-06, 2.5611e-09, 0.0000e+00, ..., 1.2323e-05, + 6.6590e-08, 6.6822e-07]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0246, -0.0327, 0.0195, -0.0244, 0.0362, 0.0168, 0.0090, -0.0121, + -0.0190, -0.0080], device='cuda:0'), grad: tensor([ 1.4137e-06, 8.5682e-06, 6.7316e-06, 4.5751e-07, -5.2869e-05, + 8.4285e-07, -3.5912e-06, -1.1742e-05, -1.5413e-07, 5.0396e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 220.27, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4958 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.16 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.2318, 0.2042, -0.0365, ..., -0.1053, 0.0220, 0.0119], + [ 0.0094, -0.1246, -0.0483, ..., 0.0321, -0.0423, -0.0650], + [-0.1557, -0.2582, 0.0554, ..., -0.0830, -0.0350, -0.1550], + ..., + [ 0.0737, 0.1343, -0.0906, ..., -0.1396, 0.0700, 0.0731], + [ 0.0416, -0.1459, -0.0337, ..., -0.1080, -0.1163, -0.0957], + [-0.1855, -0.0472, -0.0556, ..., 0.1061, -0.2537, 0.0987]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -6.5193e-08, 4.6566e-10, ..., 1.7812e-07, + 1.1642e-09, -1.3039e-08], + [-1.5553e-07, 4.6566e-09, 9.3132e-10, ..., -2.4866e-07, + -1.6764e-08, 4.2608e-08], + [ 9.5461e-09, 5.8208e-09, -1.1176e-08, ..., 6.8219e-08, + 9.3132e-10, 7.2177e-09], + ..., + [ 5.4017e-08, -9.7789e-09, 1.8626e-09, ..., 1.3341e-07, + 3.9581e-09, -5.2853e-08], + [ 1.0012e-08, 2.0955e-09, 9.3132e-10, ..., -1.6000e-06, + 1.3504e-08, 1.5832e-08], + [ 1.0338e-07, 4.1444e-08, 4.6566e-10, ..., 6.0070e-07, + 8.3121e-08, 7.1246e-08]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0242, -0.0322, 0.0195, -0.0244, 0.0363, 0.0167, 0.0090, -0.0129, + -0.0185, -0.0072], device='cuda:0'), grad: tensor([ 1.8533e-06, -7.0501e-07, 5.4762e-07, 4.0978e-06, 1.1446e-06, + 1.4920e-06, 1.5292e-06, 4.5961e-07, -1.7926e-05, 7.5214e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 220.43, cls_loss 0.0009 cls_loss_mapping 0.0031 cls_loss_causal 0.4964 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.12 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.2329, 0.2047, -0.0365, ..., -0.1057, 0.0219, 0.0122], + [ 0.0095, -0.1247, -0.0483, ..., 0.0323, -0.0425, -0.0651], + [-0.1561, -0.2589, 0.0554, ..., -0.0831, -0.0351, -0.1556], + ..., + [ 0.0737, 0.1345, -0.0906, ..., -0.1401, 0.0694, 0.0732], + [ 0.0415, -0.1463, -0.0338, ..., -0.1084, -0.1164, -0.0959], + [-0.1860, -0.0475, -0.0556, ..., 0.1074, -0.2546, 0.0988]], + device='cuda:0'), grad: tensor([[ 4.8894e-09, -1.5134e-07, 1.4366e-07, ..., 2.9523e-07, + 0.0000e+00, -9.3132e-10], + [ 1.6135e-07, 3.4925e-09, 1.0245e-08, ..., -3.9302e-07, + 4.6566e-10, 2.0955e-09], + [ 6.0536e-09, 4.4238e-09, 1.8626e-09, ..., 1.3201e-07, + 4.6566e-10, 1.1642e-09], + ..., + [ 1.2340e-08, -2.3283e-10, 1.3970e-09, ..., 5.7044e-08, + -2.5611e-09, -2.3283e-09], + [ 1.4668e-08, 1.6298e-08, 9.5461e-09, ..., 3.4925e-08, + 0.0000e+00, 4.6566e-10], + [ 3.6554e-08, 4.7963e-08, 7.2177e-09, ..., 5.5181e-08, + 9.3132e-10, -1.4435e-08]], device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0243, -0.0322, 0.0195, -0.0242, 0.0349, 0.0166, 0.0088, -0.0130, + -0.0183, -0.0064], device='cuda:0'), grad: tensor([ 6.9011e-07, 7.4785e-07, -7.6368e-07, 6.0443e-07, 3.4319e-07, + -1.0952e-06, -1.3439e-06, 1.6601e-07, 3.2387e-07, 3.2084e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 220.77, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4781 re_mapping 0.0040 re_causal 0.0113 /// teacc 99.09 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.2337, 0.2050, -0.0365, ..., -0.1059, 0.0220, 0.0121], + [ 0.0095, -0.1250, -0.0483, ..., 0.0323, -0.0427, -0.0651], + [-0.1565, -0.2597, 0.0554, ..., -0.0832, -0.0353, -0.1562], + ..., + [ 0.0741, 0.1363, -0.0906, ..., -0.1403, 0.0695, 0.0740], + [ 0.0415, -0.1489, -0.0343, ..., -0.1088, -0.1169, -0.0962], + [-0.1878, -0.0485, -0.0556, ..., 0.1080, -0.2550, 0.0981]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 1.6298e-09, ..., 1.3271e-08, + 4.4238e-08, 1.0035e-07], + [ 2.2119e-08, 1.6298e-09, 6.9849e-10, ..., 2.1653e-08, + 1.5320e-07, 3.6461e-07], + [ 1.3737e-08, 2.3283e-10, -7.9162e-09, ..., 1.0710e-08, + 4.5053e-07, 1.3467e-06], + ..., + [-4.9174e-07, -1.5134e-08, -2.7032e-07, ..., 1.9441e-07, + -5.7230e-07, -3.2876e-06], + [-1.2573e-08, 0.0000e+00, 7.9162e-09, ..., 5.1223e-09, + 2.5774e-07, 8.4750e-07], + [ 5.3365e-07, 1.2107e-08, 2.2817e-08, ..., 7.3435e-07, + 1.6140e-06, 3.7551e-06]], device='cuda:0') +Epoch 253, bias, value: tensor([ 0.0243, -0.0322, 0.0196, -0.0239, 0.0344, 0.0164, 0.0088, -0.0124, + -0.0187, -0.0068], device='cuda:0'), grad: tensor([ 4.6077e-07, 1.7779e-06, 7.6108e-06, -4.5985e-05, -1.1390e-06, + 2.9206e-05, 8.0327e-08, -8.9258e-06, 4.6492e-06, 1.2219e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 220.29, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4788 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.10 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.2357, 0.2053, -0.0363, ..., -0.1063, 0.0220, 0.0122], + [ 0.0094, -0.1254, -0.0482, ..., 0.0322, -0.0429, -0.0653], + [-0.1572, -0.2613, 0.0554, ..., -0.0834, -0.0357, -0.1568], + ..., + [ 0.0745, 0.1385, -0.0906, ..., -0.1406, 0.0698, 0.0750], + [ 0.0412, -0.1503, -0.0354, ..., -0.1092, -0.1170, -0.0964], + [-0.1908, -0.0498, -0.0557, ..., 0.1055, -0.2561, 0.0976]], + device='cuda:0'), grad: tensor([[ 1.8859e-08, -1.2238e-06, 0.0000e+00, ..., 4.4238e-09, + -3.0687e-07, -5.3365e-07], + [-3.3760e-08, 1.3039e-08, 0.0000e+00, ..., -4.6566e-08, + 6.6124e-08, 8.3353e-08], + [ 5.3551e-09, 2.0256e-08, 0.0000e+00, ..., 3.9581e-09, + 3.2829e-08, 4.0978e-08], + ..., + [ 2.4214e-08, 3.2131e-08, 0.0000e+00, ..., 2.5146e-08, + 3.3528e-08, 3.6089e-08], + [ 5.3225e-07, 4.6566e-08, 0.0000e+00, ..., 5.5879e-09, + 1.3039e-08, 1.3597e-07], + [ 1.3597e-07, 2.5937e-07, 0.0000e+00, ..., 6.7055e-08, + 7.0548e-08, 1.4482e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0244, -0.0323, 0.0195, -0.0239, 0.0377, 0.0166, 0.0088, -0.0117, + -0.0190, -0.0105], device='cuda:0'), grad: tensor([-3.2745e-06, 3.5157e-07, -3.0338e-07, 6.6590e-07, -4.0536e-07, + -2.2370e-06, 8.8941e-07, 4.2794e-07, 2.3823e-06, 1.4985e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 220.20, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.5041 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.11 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.2371, 0.2057, -0.0357, ..., -0.1060, 0.0220, 0.0124], + [ 0.0088, -0.1255, -0.0483, ..., 0.0322, -0.0434, -0.0655], + [-0.1576, -0.2615, 0.0555, ..., -0.0834, -0.0360, -0.1572], + ..., + [ 0.0747, 0.1386, -0.0907, ..., -0.1412, 0.0699, 0.0748], + [ 0.0414, -0.1505, -0.0331, ..., -0.1099, -0.1171, -0.0965], + [-0.1911, -0.0500, -0.0560, ..., 0.1060, -0.2563, 0.0985]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.9581e-09, 0.0000e+00, ..., 1.3737e-08, + 0.0000e+00, 2.3283e-10], + [-1.0477e-08, 1.8626e-09, 0.0000e+00, ..., 9.7789e-09, + 6.0536e-09, 1.4901e-08], + [ 4.3306e-08, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + 2.4913e-08, 5.6578e-08], + ..., + [-3.3760e-08, -2.0955e-09, 0.0000e+00, ..., 2.6310e-08, + -6.3796e-08, -1.1153e-07], + [ 5.5879e-09, 4.6566e-10, 0.0000e+00, ..., 3.0268e-09, + 6.0536e-09, 1.2806e-08], + [ 3.7253e-08, 4.4238e-09, 0.0000e+00, ..., -6.5193e-09, + 6.9849e-09, -4.6100e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0246, -0.0326, 0.0194, -0.0238, 0.0374, 0.0167, 0.0078, -0.0119, + -0.0189, -0.0100], device='cuda:0'), grad: tensor([ 4.5868e-08, -8.6147e-08, 2.8173e-07, 7.4273e-08, -5.4250e-08, + 1.3993e-07, -1.3993e-07, -2.5495e-07, -2.3283e-09, 1.7695e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 220.62, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4719 re_mapping 0.0039 re_causal 0.0112 /// teacc 99.12 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.2382, 0.2058, -0.0361, ..., -0.1063, 0.0220, 0.0124], + [ 0.0083, -0.1283, -0.0483, ..., 0.0322, -0.0436, -0.0680], + [-0.1579, -0.2618, 0.0555, ..., -0.0836, -0.0357, -0.1580], + ..., + [ 0.0752, 0.1410, -0.0907, ..., -0.1415, 0.0700, 0.0769], + [ 0.0413, -0.1507, -0.0347, ..., -0.1104, -0.1171, -0.0966], + [-0.1941, -0.0501, -0.0562, ..., 0.1052, -0.2565, 0.0984]], + device='cuda:0'), grad: tensor([[ 3.8184e-08, 0.0000e+00, 2.3283e-10, ..., 3.8673e-07, + 1.1642e-09, 1.1642e-09], + [ 3.9581e-09, 6.9849e-10, 4.6566e-10, ..., 7.4506e-09, + -3.3993e-08, -2.0415e-06], + [ 1.9558e-08, 2.0955e-09, -1.6298e-09, ..., 2.1188e-08, + 2.5611e-09, 1.8161e-08], + ..., + [ 3.0268e-09, -5.8208e-09, 2.3283e-10, ..., 1.2340e-08, + 3.0268e-08, 1.8077e-06], + [-6.6217e-07, 0.0000e+00, 2.3283e-10, ..., 1.6298e-09, + 2.3283e-10, -3.9581e-09], + [ 1.5716e-07, 2.3283e-09, 0.0000e+00, ..., 1.1101e-06, + 4.6566e-09, 4.6566e-09]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0245, -0.0344, 0.0194, -0.0239, 0.0384, 0.0168, 0.0082, -0.0104, + -0.0189, -0.0110], device='cuda:0'), grad: tensor([ 1.8720e-06, -5.1856e-06, 3.4226e-07, 6.6962e-07, -3.9898e-06, + 1.1750e-05, 1.0841e-06, 4.9062e-06, -1.6943e-05, 5.4613e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 220.66, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.5158 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.07 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.2406, 0.2059, -0.0360, ..., -0.1065, 0.0220, 0.0120], + [ 0.0090, -0.1291, -0.0483, ..., 0.0348, -0.0419, -0.0678], + [-0.1622, -0.2649, 0.0554, ..., -0.0837, -0.0388, -0.1592], + ..., + [ 0.0754, 0.1420, -0.0907, ..., -0.1433, 0.0707, 0.0772], + [ 0.0411, -0.1515, -0.0352, ..., -0.1113, -0.1173, -0.0967], + [-0.1951, -0.0502, -0.0537, ..., 0.1050, -0.2589, 0.0981]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, -2.3283e-09, 4.6566e-10, ..., 4.7032e-08, + 2.3283e-10, 4.2841e-08], + [ 4.8429e-08, 2.9337e-08, 1.6298e-09, ..., 1.4971e-07, + 1.3737e-08, 2.2282e-07], + [ 2.6776e-08, 1.8626e-09, 2.3283e-09, ..., 1.3737e-08, + 1.1642e-09, 2.7474e-08], + ..., + [-1.8626e-07, -8.7311e-08, -4.8894e-09, ..., 8.3586e-08, + -4.1444e-08, -2.1327e-07], + [ 5.3551e-08, 9.3132e-10, 3.4925e-09, ..., 5.7742e-08, + 2.3283e-10, 9.4064e-08], + [ 9.5926e-08, 4.9826e-08, 2.5844e-08, ..., -3.3528e-07, + 2.3283e-08, -3.3295e-07]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0242, -0.0332, 0.0191, -0.0239, 0.0383, 0.0171, 0.0074, -0.0106, + -0.0193, -0.0116], device='cuda:0'), grad: tensor([ 1.9348e-07, 1.1781e-06, 3.4738e-07, 1.1479e-07, 9.3132e-10, + 1.3644e-07, -5.5879e-09, -2.0824e-06, 9.0804e-07, -7.9768e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 220.57, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4828 re_mapping 0.0039 re_causal 0.0112 /// teacc 99.16 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.2429, 0.2061, -0.0360, ..., -0.1066, 0.0218, 0.0120], + [ 0.0119, -0.1291, -0.0453, ..., 0.0379, -0.0428, -0.0654], + [-0.1628, -0.2653, 0.0554, ..., -0.0838, -0.0385, -0.1659], + ..., + [ 0.0725, 0.1423, -0.0937, ..., -0.1467, 0.0712, 0.0757], + [ 0.0406, -0.1517, -0.0358, ..., -0.1124, -0.1177, -0.0971], + [-0.1955, -0.0504, -0.0541, ..., 0.1050, -0.2599, 0.0980]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, -9.3132e-10, 0.0000e+00, ..., 3.3528e-08, + 6.5193e-09, 1.3039e-08], + [-3.7253e-08, 1.0524e-07, -1.3970e-09, ..., -1.2573e-08, + 3.0734e-08, 8.9407e-08], + [ 5.5879e-09, 1.4435e-08, 0.0000e+00, ..., 4.1910e-09, + 4.6566e-09, 1.3504e-08], + ..., + [-1.3504e-08, -2.8778e-07, 0.0000e+00, ..., 3.3993e-08, + -8.6613e-08, -2.0908e-07], + [-2.1420e-08, 5.5879e-09, 0.0000e+00, ..., 1.9139e-07, + 1.3970e-09, 1.5460e-07], + [ 2.7474e-08, 1.3970e-07, 0.0000e+00, ..., -2.9383e-07, + 3.5856e-08, -1.3551e-07]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0242, -0.0303, 0.0178, -0.0239, 0.0383, 0.0173, 0.0074, -0.0132, + -0.0200, -0.0117], device='cuda:0'), grad: tensor([ 1.6717e-07, 4.9174e-07, -4.3539e-07, 8.1025e-08, 2.4494e-07, + 2.4587e-07, -3.1153e-07, -4.8289e-07, 3.6322e-07, -3.6880e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 220.87, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4996 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.17 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.2446, 0.2063, -0.0361, ..., -0.1070, 0.0219, 0.0120], + [ 0.0119, -0.1293, -0.0458, ..., 0.0378, -0.0429, -0.0653], + [-0.1624, -0.2671, 0.0583, ..., -0.0813, -0.0396, -0.1664], + ..., + [ 0.0724, 0.1426, -0.0938, ..., -0.1467, 0.0719, 0.0757], + [ 0.0404, -0.1524, -0.0358, ..., -0.1129, -0.1178, -0.0972], + [-0.1959, -0.0508, -0.0542, ..., 0.1052, -0.2605, 0.0982]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, 0.0000e+00, ..., 5.1223e-09, + 9.3132e-10, 2.7940e-09], + [-9.7789e-09, 2.3283e-09, 0.0000e+00, ..., 6.5193e-09, + 2.3283e-09, 2.0023e-08], + [ 5.5879e-09, 1.8626e-09, -3.2596e-09, ..., 6.9849e-09, + 1.8626e-09, 9.3132e-10], + ..., + [ 1.3970e-09, -4.1910e-09, 9.3132e-10, ..., 5.3085e-08, + 1.8626e-09, 4.0047e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3970e-09, + 1.3970e-09, 2.7940e-09], + [ 1.3970e-09, 3.7253e-09, 0.0000e+00, ..., -2.9802e-07, + 2.7940e-09, -2.3935e-07]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0241, -0.0308, 0.0200, -0.0239, 0.0383, 0.0173, 0.0078, -0.0132, + -0.0203, -0.0117], device='cuda:0'), grad: tensor([ 1.2107e-08, 3.1665e-08, -9.9186e-08, -1.8021e-07, 5.9605e-07, + 1.2806e-07, 2.4680e-08, 2.2538e-07, 2.8871e-08, -7.5670e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 258---------------------------------------------------- +epoch 258, time 221.17, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4954 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.22 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.2454, 0.2068, -0.0361, ..., -0.1069, 0.0218, 0.0123], + [ 0.0119, -0.1294, -0.0458, ..., 0.0378, -0.0431, -0.0653], + [-0.1632, -0.2680, 0.0583, ..., -0.0815, -0.0399, -0.1666], + ..., + [ 0.0724, 0.1428, -0.0938, ..., -0.1467, 0.0721, 0.0757], + [ 0.0403, -0.1528, -0.0358, ..., -0.1127, -0.1178, -0.0973], + [-0.1960, -0.0512, -0.0542, ..., 0.1053, -0.2608, 0.0985]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, -1.0477e-07, 3.7253e-09, ..., 2.2817e-08, + -6.0536e-09, 3.8650e-08], + [ 2.3982e-07, -5.7509e-07, 0.0000e+00, ..., 2.1514e-07, + 2.7940e-09, 2.8405e-08], + [ 2.0489e-08, 3.6461e-07, 0.0000e+00, ..., 1.8626e-08, + 6.0536e-09, 1.3039e-07], + ..., + [ 7.4506e-08, -2.5891e-07, -1.7229e-08, ..., 1.0896e-07, + 1.3039e-08, -4.0745e-07], + [ 6.7987e-08, 3.3528e-08, 4.6566e-10, ..., 1.0710e-08, + 6.5193e-09, 1.5367e-08], + [ 4.2282e-07, 3.9302e-07, 9.7789e-09, ..., 7.7067e-07, + 7.9162e-09, 1.5646e-07]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0243, -0.0308, 0.0200, -0.0239, 0.0382, 0.0173, 0.0076, -0.0132, + -0.0203, -0.0117], device='cuda:0'), grad: tensor([ 1.0384e-07, -9.7789e-07, 1.0524e-06, 4.8280e-06, -2.2929e-06, + -6.5006e-06, 3.7346e-07, 3.4738e-07, 3.9069e-07, 2.6859e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 220.58, cls_loss 0.0006 cls_loss_mapping 0.0023 cls_loss_causal 0.5173 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.20 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.2463, 0.2072, -0.0361, ..., -0.1069, 0.0211, 0.0123], + [ 0.0119, -0.1297, -0.0458, ..., 0.0377, -0.0435, -0.0654], + [-0.1635, -0.2690, 0.0583, ..., -0.0817, -0.0403, -0.1668], + ..., + [ 0.0726, 0.1433, -0.0938, ..., -0.1467, 0.0724, 0.0760], + [ 0.0402, -0.1533, -0.0359, ..., -0.1129, -0.1179, -0.0974], + [-0.1964, -0.0515, -0.0542, ..., 0.1054, -0.2610, 0.0990]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -9.4995e-08, 0.0000e+00, ..., -7.4506e-09, + 0.0000e+00, -4.1910e-08], + [-6.0536e-09, 9.7789e-09, 0.0000e+00, ..., -2.7940e-08, + 1.8161e-08, 5.7742e-08], + [ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 4.1910e-09, + 1.3970e-09, 2.7940e-09], + ..., + [-2.5146e-08, -1.8626e-08, 4.6566e-10, ..., 9.3132e-09, + -3.5856e-08, -1.1222e-07], + [-9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, -1.1642e-08], + [ 5.0291e-08, 7.9162e-08, 9.3132e-10, ..., 2.4168e-07, + 1.9558e-08, 9.6392e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([ 0.0244, -0.0308, 0.0200, -0.0239, 0.0377, 0.0174, 0.0073, -0.0131, + -0.0204, -0.0116], device='cuda:0'), grad: tensor([-1.5879e-07, 1.1036e-07, 0.0000e+00, 1.7229e-08, -3.4925e-07, + 2.7940e-08, 4.3772e-08, -2.9989e-07, -1.8626e-07, 8.1072e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 220.74, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4822 re_mapping 0.0038 re_causal 0.0111 /// teacc 99.14 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.2472, 0.2077, -0.0363, ..., -0.1070, 0.0208, 0.0124], + [ 0.0119, -0.1300, -0.0458, ..., 0.0377, -0.0437, -0.0655], + [-0.1637, -0.2701, 0.0585, ..., -0.0819, -0.0403, -0.1669], + ..., + [ 0.0726, 0.1435, -0.0938, ..., -0.1467, 0.0725, 0.0761], + [ 0.0402, -0.1538, -0.0368, ..., -0.1132, -0.1180, -0.0975], + [-0.1966, -0.0518, -0.0543, ..., 0.1054, -0.2611, 0.0991]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.7940e-09, 1.4901e-08], + [ 1.6671e-07, 9.5926e-08, 1.4435e-08, ..., 6.0536e-09, + 2.8871e-08, 1.7509e-07], + [ 2.7008e-08, 8.8476e-09, -4.6566e-10, ..., 4.6566e-10, + 4.1910e-09, 3.3528e-08], + ..., + [-7.9069e-07, -5.4389e-07, 2.3283e-09, ..., 8.3819e-09, + -1.4156e-07, -1.1502e-06], + [ 4.9826e-08, 7.1246e-08, -1.9558e-08, ..., 1.8626e-09, + 1.8626e-08, 1.6578e-07], + [ 5.5740e-07, 2.7753e-07, 0.0000e+00, ..., 6.9384e-08, + 1.2526e-07, 7.4459e-07]], device='cuda:0') +Epoch 262, bias, value: tensor([ 0.0245, -0.0308, 0.0201, -0.0239, 0.0378, 0.0175, 0.0072, -0.0130, + -0.0206, -0.0117], device='cuda:0'), grad: tensor([ 5.2620e-08, 8.2469e-07, -2.7008e-07, 1.9409e-06, 3.3947e-07, + -2.6003e-06, 2.4680e-07, -2.7902e-06, 1.0105e-07, 2.1495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 220.51, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5206 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.16 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.2488, 0.2080, -0.0363, ..., -0.1071, 0.0209, 0.0123], + [ 0.0119, -0.1303, -0.0459, ..., 0.0377, -0.0438, -0.0656], + [-0.1652, -0.2716, 0.0592, ..., -0.0806, -0.0406, -0.1672], + ..., + [ 0.0726, 0.1439, -0.0938, ..., -0.1468, 0.0728, 0.0762], + [ 0.0393, -0.1553, -0.0368, ..., -0.1148, -0.1181, -0.0977], + [-0.1970, -0.0519, -0.0542, ..., 0.1053, -0.2617, 0.0992]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, -4.6566e-09, 0.0000e+00, ..., 2.7940e-09, + 2.3283e-09, 9.3132e-09], + [ 7.9162e-08, 1.1874e-07, 0.0000e+00, ..., -1.3504e-08, + 4.0978e-08, 1.8673e-07], + [ 7.8604e-07, 1.7919e-06, 0.0000e+00, ..., 6.9849e-09, + 3.0268e-08, 1.7583e-06], + ..., + [-1.0123e-06, -2.1067e-06, 0.0000e+00, ..., 4.6566e-09, + -8.4285e-08, -2.1681e-06], + [ 3.3062e-08, 4.2841e-08, 0.0000e+00, ..., -2.3283e-09, + 1.2573e-08, 6.3330e-08], + [ 6.0070e-08, 5.6345e-08, 0.0000e+00, ..., -2.3283e-08, + 3.0734e-08, 9.7323e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0246, -0.0310, 0.0209, -0.0237, 0.0380, 0.0173, 0.0066, -0.0130, + -0.0212, -0.0119], device='cuda:0'), grad: tensor([ 6.9849e-08, 9.9093e-07, 1.1370e-05, 1.9325e-07, 1.4203e-07, + -2.7902e-06, 2.9318e-06, -1.3851e-05, 3.3481e-07, 5.9092e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 220.60, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4900 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.12 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.2528, 0.2081, -0.0363, ..., -0.1073, 0.0203, 0.0122], + [ 0.0119, -0.1307, -0.0459, ..., 0.0377, -0.0442, -0.0656], + [-0.1663, -0.2727, 0.0592, ..., -0.0808, -0.0408, -0.1674], + ..., + [ 0.0726, 0.1449, -0.0938, ..., -0.1468, 0.0734, 0.0764], + [ 0.0393, -0.1563, -0.0368, ..., -0.1145, -0.1182, -0.0978], + [-0.1972, -0.0523, -0.0542, ..., 0.1073, -0.2626, 0.1011]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 5.1223e-09, 0.0000e+00, ..., 1.3504e-08, + 5.1223e-09, 4.6566e-09], + [-8.8476e-09, 9.3132e-10, 0.0000e+00, ..., -3.5390e-08, + 1.8626e-09, 3.2596e-09], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 1.6298e-08, + -1.8626e-09, 9.3132e-10], + ..., + [ 3.7253e-09, -4.6566e-10, 0.0000e+00, ..., 1.3504e-08, + 2.3283e-09, -1.3970e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 9.3132e-10, 0.0000e+00, ..., -4.6566e-09, + 1.8626e-09, -1.3970e-09]], device='cuda:0') +Epoch 264, bias, value: tensor([ 0.0244, -0.0310, 0.0213, -0.0236, 0.0363, 0.0172, 0.0064, -0.0130, + -0.0209, -0.0104], device='cuda:0'), grad: tensor([ 1.2200e-07, -5.2620e-08, -5.1223e-07, 5.6345e-08, 1.7229e-08, + -1.8720e-07, 1.6298e-08, 5.2294e-07, 1.1176e-08, 2.1886e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 221.39, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4908 re_mapping 0.0039 re_causal 0.0110 /// teacc 99.13 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.2535, 0.2083, -0.0364, ..., -0.1077, 0.0202, 0.0124], + [ 0.0119, -0.1309, -0.0459, ..., 0.0377, -0.0444, -0.0657], + [-0.1670, -0.2731, 0.0592, ..., -0.0809, -0.0410, -0.1676], + ..., + [ 0.0727, 0.1453, -0.0937, ..., -0.1468, 0.0738, 0.0767], + [ 0.0393, -0.1586, -0.0370, ..., -0.1146, -0.1182, -0.0980], + [-0.1973, -0.0527, -0.0544, ..., 0.1073, -0.2636, 0.1010]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.6822e-07, -7.4506e-08, ..., 1.8626e-09, + 2.0955e-08, -2.2119e-07], + [-1.8626e-09, 5.5879e-09, 9.3132e-10, ..., -4.1910e-09, + 3.7253e-09, 2.2352e-08], + [ 2.3283e-09, 1.3970e-09, -3.7253e-09, ..., 3.7253e-09, + 1.8626e-09, 5.5879e-09], + ..., + [ 0.0000e+00, 9.3132e-10, 4.6566e-10, ..., 7.4506e-09, + 2.3283e-09, -8.4750e-08], + [ 4.6566e-10, 2.1420e-08, 3.7253e-09, ..., 4.6566e-10, + 3.3528e-08, 3.0268e-08], + [ 6.9849e-09, 8.1491e-08, 4.6566e-09, ..., 1.1176e-08, + 9.7789e-09, 1.0198e-07]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0243, -0.0310, 0.0212, -0.0236, 0.0362, 0.0172, 0.0063, -0.0129, + -0.0210, -0.0105], device='cuda:0'), grad: tensor([-1.2154e-06, 1.9511e-07, -4.4098e-07, 1.0729e-06, -4.6566e-10, + -1.1045e-06, 6.0257e-07, 3.2224e-07, 1.8999e-07, 3.7532e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 220.60, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4762 re_mapping 0.0038 re_causal 0.0107 /// teacc 99.16 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.2542, 0.2085, -0.0366, ..., -0.1079, 0.0203, 0.0125], + [ 0.0119, -0.1312, -0.0458, ..., 0.0378, -0.0445, -0.0659], + [-0.1684, -0.2743, 0.0590, ..., -0.0812, -0.0415, -0.1689], + ..., + [ 0.0728, 0.1457, -0.0936, ..., -0.1468, 0.0742, 0.0774], + [ 0.0394, -0.1590, -0.0372, ..., -0.1145, -0.1182, -0.0980], + [-0.1978, -0.0529, -0.0551, ..., 0.1072, -0.2643, 0.1007]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, -2.0955e-08, 0.0000e+00, ..., 5.0757e-08, + 1.1176e-08, 3.6787e-08], + [ 2.3283e-09, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + 2.8871e-08, 8.1491e-08], + [ 1.8161e-08, 2.7940e-09, 0.0000e+00, ..., 6.5193e-09, + 2.3982e-07, 2.7986e-07], + ..., + [-2.1420e-08, -1.8626e-09, 0.0000e+00, ..., 1.3970e-08, + 1.2619e-07, 1.0757e-07], + [ 1.2433e-07, 1.8626e-09, 0.0000e+00, ..., 3.2596e-09, + 2.3283e-08, 2.7940e-08], + [ 7.9162e-09, 5.5879e-09, 0.0000e+00, ..., -1.9651e-07, + 1.1642e-08, -1.1316e-07]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0243, -0.0310, 0.0208, -0.0241, 0.0358, 0.0178, 0.0060, -0.0126, + -0.0205, -0.0109], device='cuda:0'), grad: tensor([ 1.6857e-07, 2.3562e-07, 6.8359e-07, -1.7276e-07, 9.5926e-08, + -2.7679e-06, 4.6520e-07, 3.4366e-07, 9.9652e-07, -5.5879e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 220.63, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4817 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.2549, 0.2102, -0.0355, ..., -0.1080, 0.0203, 0.0149], + [ 0.0119, -0.1314, -0.0458, ..., 0.0378, -0.0447, -0.0660], + [-0.1705, -0.2784, 0.0589, ..., -0.0818, -0.0421, -0.1714], + ..., + [ 0.0729, 0.1461, -0.0935, ..., -0.1468, 0.0748, 0.0779], + [ 0.0394, -0.1603, -0.0373, ..., -0.1142, -0.1186, -0.0983], + [-0.1988, -0.0533, -0.0553, ..., 0.1068, -0.2652, 0.1001]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, -5.2154e-07, 0.0000e+00, ..., 3.9581e-08, + 0.0000e+00, -1.4808e-07], + [-1.3923e-07, 2.8405e-08, 0.0000e+00, ..., -1.8766e-07, + 9.3132e-10, -9.6392e-08], + [ 1.5367e-08, 2.2352e-08, -4.6566e-10, ..., 2.3283e-08, + 1.3970e-09, 1.2573e-08], + ..., + [ 5.0291e-08, 5.1223e-09, 0.0000e+00, ..., 2.4214e-07, + 1.3970e-09, 1.9930e-07], + [-1.0338e-07, 2.1420e-08, 0.0000e+00, ..., -7.5158e-07, + 0.0000e+00, 2.1886e-08], + [ 5.1223e-08, 1.0198e-07, 0.0000e+00, ..., -5.0291e-08, + 0.0000e+00, -9.3132e-08]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0259, -0.0309, 0.0198, -0.0243, 0.0363, 0.0180, 0.0055, -0.0125, + -0.0206, -0.0116], device='cuda:0'), grad: tensor([-8.2608e-07, 2.8033e-07, -6.5193e-07, 3.0035e-07, 1.6904e-07, + 3.0966e-07, 2.0862e-06, 8.2096e-07, -2.5444e-06, 6.1002e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 220.64, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4725 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.2565, 0.2106, -0.0354, ..., -0.1086, 0.0205, 0.0151], + [ 0.0119, -0.1317, -0.0458, ..., 0.0377, -0.0445, -0.0660], + [-0.1711, -0.2789, 0.0590, ..., -0.0820, -0.0425, -0.1716], + ..., + [ 0.0729, 0.1462, -0.0935, ..., -0.1469, 0.0744, 0.0778], + [ 0.0394, -0.1609, -0.0368, ..., -0.1142, -0.1186, -0.0984], + [-0.1990, -0.0535, -0.0553, ..., 0.1069, -0.2657, 0.1004]], + device='cuda:0'), grad: tensor([[ 7.9162e-09, -1.6764e-08, 0.0000e+00, ..., 1.3970e-08, + 4.1910e-09, 5.1223e-09], + [ 3.8417e-07, 1.5832e-08, 0.0000e+00, ..., 9.7789e-09, + 1.8952e-07, 4.4797e-07], + [ 4.9826e-08, 3.7253e-09, 0.0000e+00, ..., 7.4506e-09, + 2.1886e-08, 5.2154e-08], + ..., + [-5.5833e-07, -2.5146e-08, 0.0000e+00, ..., 1.2107e-08, + -2.7893e-07, -6.5938e-07], + [-9.3132e-09, 2.3283e-09, 0.0000e+00, ..., 8.8476e-09, + 5.5879e-09, 1.2107e-08], + [ 7.2643e-08, 1.1176e-08, 0.0000e+00, ..., 2.3283e-09, + 3.2596e-08, 6.8452e-08]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0258, -0.0309, 0.0199, -0.0243, 0.0362, 0.0183, 0.0050, -0.0126, + -0.0203, -0.0116], device='cuda:0'), grad: tensor([ 7.0781e-08, 1.7136e-06, 1.8952e-07, 9.8720e-08, 2.4214e-08, + 2.8387e-06, -2.8145e-06, -2.4289e-06, 0.0000e+00, 3.0687e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 220.58, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4723 re_mapping 0.0037 re_causal 0.0109 /// teacc 99.14 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.2573, 0.2108, -0.0355, ..., -0.1088, 0.0205, 0.0151], + [ 0.0119, -0.1319, -0.0458, ..., 0.0377, -0.0448, -0.0660], + [-0.1725, -0.2791, 0.0590, ..., -0.0821, -0.0428, -0.1720], + ..., + [ 0.0729, 0.1465, -0.0935, ..., -0.1469, 0.0745, 0.0779], + [ 0.0395, -0.1614, -0.0366, ..., -0.1143, -0.1187, -0.0985], + [-0.1993, -0.0537, -0.0554, ..., 0.1070, -0.2660, 0.1004]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.8626e-09, 0.0000e+00, ..., 1.4901e-08, + 0.0000e+00, 6.0536e-09], + [ 9.3132e-10, 1.4435e-08, 0.0000e+00, ..., -6.9849e-09, + 1.7229e-08, 3.1665e-08], + [ 4.1910e-09, 3.2596e-09, 0.0000e+00, ..., 3.7253e-09, + 2.7940e-09, 3.7253e-09], + ..., + [-7.4506e-09, -3.0734e-08, 0.0000e+00, ..., 6.1467e-08, + -2.7940e-08, -1.1642e-08], + [-9.3132e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-09, + 4.6566e-10, 7.4506e-09], + [ 4.6566e-09, 6.9849e-09, 0.0000e+00, ..., -1.6112e-07, + 6.5193e-09, -1.1176e-07]], device='cuda:0') +Epoch 269, bias, value: tensor([ 0.0258, -0.0310, 0.0199, -0.0244, 0.0362, 0.0184, 0.0051, -0.0125, + -0.0202, -0.0116], device='cuda:0'), grad: tensor([ 4.5635e-08, 3.8184e-08, 3.3528e-08, 9.7323e-08, 1.9185e-07, + -2.3283e-08, -6.8452e-08, 6.1002e-08, -3.6787e-08, -3.2783e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 220.75, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4758 re_mapping 0.0040 re_causal 0.0111 /// teacc 99.04 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.2577, 0.2109, -0.0363, ..., -0.1092, 0.0202, 0.0149], + [ 0.0120, -0.1322, -0.0458, ..., 0.0378, -0.0451, -0.0662], + [-0.1729, -0.2792, 0.0589, ..., -0.0822, -0.0431, -0.1721], + ..., + [ 0.0729, 0.1470, -0.0936, ..., -0.1470, 0.0748, 0.0782], + [ 0.0396, -0.1616, -0.0375, ..., -0.1139, -0.1184, -0.0985], + [-0.1996, -0.0540, -0.0555, ..., 0.1071, -0.2666, 0.1009]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -2.3283e-09, 5.1223e-09, ..., 9.3132e-09, + 4.6566e-10, 3.2596e-09], + [ 2.7940e-08, 4.6566e-10, 9.9186e-08, ..., 5.9605e-08, + 2.7940e-09, 7.4506e-09], + [ 8.8476e-09, 4.6566e-10, -3.0873e-07, ..., 8.3819e-09, + 2.3283e-09, 1.1642e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 2.0862e-07, ..., 3.6322e-08, + 8.3819e-09, 1.3970e-08], + [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 2.7940e-09, + 1.3970e-09, 3.2596e-09], + [ 8.6613e-08, 2.7940e-09, 0.0000e+00, ..., 1.0198e-07, + 1.3970e-09, -7.9162e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([ 0.0255, -0.0310, 0.0199, -0.0248, 0.0362, 0.0191, 0.0036, -0.0125, + -0.0199, -0.0114], device='cuda:0'), grad: tensor([ 3.6322e-08, 6.2212e-07, -1.3839e-06, 1.4110e-07, -6.1514e-07, + -1.7835e-07, 6.6590e-08, 1.0543e-06, 1.6764e-08, 2.4168e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 220.55, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4951 re_mapping 0.0039 re_causal 0.0111 /// teacc 99.17 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.2587, 0.2107, -0.0364, ..., -0.1099, 0.0198, 0.0124], + [ 0.0120, -0.1326, -0.0458, ..., 0.0378, -0.0454, -0.0664], + [-0.1732, -0.2792, 0.0594, ..., -0.0823, -0.0431, -0.1728], + ..., + [ 0.0729, 0.1474, -0.0936, ..., -0.1470, 0.0751, 0.0787], + [ 0.0396, -0.1626, -0.0375, ..., -0.1139, -0.1179, -0.0986], + [-0.1999, -0.0533, -0.0549, ..., 0.1072, -0.2669, 0.1018]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -7.9162e-09, 0.0000e+00, ..., 2.0955e-08, + 4.6566e-10, 2.0489e-08], + [ 6.9849e-09, 8.3819e-09, -4.6566e-10, ..., -6.0536e-09, + 2.3283e-09, 3.5390e-08], + [ 3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 2.3283e-09, + 9.3132e-10, 1.3970e-08], + ..., + [-9.5926e-08, -7.6368e-08, 0.0000e+00, ..., 1.7695e-08, + 3.7253e-09, -3.3341e-07], + [ 4.6566e-09, 4.1910e-09, 0.0000e+00, ..., -2.3283e-08, + 4.6566e-10, 1.6764e-08], + [ 6.3330e-08, 5.5879e-08, 0.0000e+00, ..., 4.7032e-08, + 1.8626e-09, 2.2352e-07]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0239, -0.0310, 0.0203, -0.0251, 0.0362, 0.0193, 0.0034, -0.0125, + -0.0192, -0.0111], device='cuda:0'), grad: tensor([ 1.1548e-07, 1.0245e-07, -2.0117e-06, -2.9337e-08, -1.6531e-07, + 1.2154e-07, 2.2957e-07, 7.5484e-07, -1.3551e-07, 1.0375e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 220.34, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4638 re_mapping 0.0037 re_causal 0.0104 /// teacc 99.17 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.2592, 0.2108, -0.0373, ..., -0.1103, 0.0198, 0.0120], + [ 0.0120, -0.1326, -0.0458, ..., 0.0378, -0.0455, -0.0666], + [-0.1736, -0.2791, 0.0594, ..., -0.0824, -0.0431, -0.1739], + ..., + [ 0.0729, 0.1474, -0.0936, ..., -0.1471, 0.0749, 0.0789], + [ 0.0397, -0.1628, -0.0383, ..., -0.1141, -0.1179, -0.0987], + [-0.2004, -0.0533, -0.0553, ..., 0.1071, -0.2672, 0.1020]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, -3.2596e-09, 0.0000e+00, ..., 3.7253e-09, + 4.6566e-10, -4.6566e-10], + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 4.6566e-10, 5.5879e-09], + [ 5.9139e-08, 4.6566e-10, 0.0000e+00, ..., 1.8161e-08, + 0.0000e+00, 5.4482e-08], + ..., + [ 9.2667e-08, 4.6566e-10, 0.0000e+00, ..., 2.0396e-07, + 1.3970e-09, -5.2620e-08], + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-10, 1.3970e-09], + [ 1.0664e-07, 4.1910e-09, 0.0000e+00, ..., 8.5216e-08, + 1.3970e-09, -4.5635e-08]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0235, -0.0311, 0.0202, -0.0251, 0.0364, 0.0192, 0.0033, -0.0124, + -0.0185, -0.0112], device='cuda:0'), grad: tensor([ 6.4727e-08, 1.6764e-08, -1.0105e-07, 3.7253e-07, -1.0412e-06, + -4.7870e-07, 1.3039e-07, 4.2561e-07, 2.7800e-07, 3.3434e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 220.69, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.5039 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.22 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.2604, 0.2109, -0.0378, ..., -0.1109, 0.0196, 0.0119], + [ 0.0120, -0.1327, -0.0458, ..., 0.0378, -0.0456, -0.0667], + [-0.1746, -0.2792, 0.0592, ..., -0.0828, -0.0433, -0.1741], + ..., + [ 0.0729, 0.1474, -0.0936, ..., -0.1471, 0.0747, 0.0790], + [ 0.0396, -0.1630, -0.0385, ..., -0.1151, -0.1181, -0.0989], + [-0.2010, -0.0534, -0.0553, ..., 0.1070, -0.2674, 0.1023]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -2.5611e-08, 0.0000e+00, ..., 1.5832e-08, + -4.6566e-10, -4.6566e-09], + [ 1.5832e-08, 4.1910e-09, 1.8626e-09, ..., 5.4948e-08, + 1.8626e-09, 2.2352e-08], + [ 1.3504e-08, 6.9849e-09, 6.5193e-09, ..., 3.0734e-08, + 2.7940e-09, 2.4214e-08], + ..., + [ 5.0291e-08, -6.5193e-09, -1.0710e-08, ..., 2.8452e-07, + -9.3132e-10, 1.3970e-08], + [ 1.8626e-09, 4.6566e-10, 4.6566e-10, ..., -1.4901e-08, + 1.8626e-09, 6.9849e-09], + [ 1.2731e-06, 9.3132e-09, 4.6566e-10, ..., 4.7758e-06, + -1.8626e-09, 6.1980e-07]], device='cuda:0') +Epoch 273, bias, value: tensor([ 0.0230, -0.0310, 0.0202, -0.0256, 0.0367, 0.0200, 0.0027, -0.0125, + -0.0187, -0.0113], device='cuda:0'), grad: tensor([ 1.0384e-07, 1.9511e-07, 3.9954e-07, 2.9756e-07, -1.3769e-05, + 7.7998e-07, 2.5425e-07, 6.0769e-07, -1.6727e-06, 1.2808e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 272---------------------------------------------------- +epoch 272, time 221.36, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4852 re_mapping 0.0036 re_causal 0.0105 /// teacc 99.23 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.2608, 0.2111, -0.0383, ..., -0.1115, 0.0199, 0.0120], + [ 0.0119, -0.1333, -0.0457, ..., 0.0378, -0.0458, -0.0672], + [-0.1764, -0.2793, 0.0589, ..., -0.0833, -0.0436, -0.1745], + ..., + [ 0.0730, 0.1480, -0.0936, ..., -0.1472, 0.0749, 0.0796], + [ 0.0396, -0.1634, -0.0386, ..., -0.1155, -0.1182, -0.0991], + [-0.2013, -0.0537, -0.0554, ..., 0.1071, -0.2680, 0.1021]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 4.1444e-08, + 4.6566e-10, 3.3993e-08], + [-1.2107e-08, -3.2596e-09, 0.0000e+00, ..., 2.8871e-08, + -9.7789e-09, 7.0781e-08], + [ 5.1223e-09, 4.6566e-10, -4.6566e-10, ..., 1.5832e-08, + 3.6322e-08, 3.6787e-08], + ..., + [ 9.3132e-09, 1.3970e-09, 0.0000e+00, ..., 1.2293e-07, + 8.8476e-09, 1.1129e-07], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.1886e-08, + 8.9873e-08, 1.2899e-07], + [-5.8627e-07, 4.6566e-10, 0.0000e+00, ..., -1.6928e-05, + 3.2596e-09, -1.7211e-05]], device='cuda:0') +Epoch 274, bias, value: tensor([ 0.0227, -0.0312, 0.0200, -0.0253, 0.0368, 0.0196, 0.0031, -0.0122, + -0.0186, -0.0115], device='cuda:0'), grad: tensor([ 1.9697e-07, -1.0161e-06, 7.0874e-07, -7.1619e-07, 6.0409e-05, + 5.7975e-07, 3.2596e-09, 8.3167e-07, 7.0268e-07, -6.1750e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 220.83, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4712 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.16 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.2612, 0.2112, -0.0383, ..., -0.1118, 0.0199, 0.0118], + [ 0.0119, -0.1334, -0.0457, ..., 0.0379, -0.0460, -0.0676], + [-0.1770, -0.2795, 0.0590, ..., -0.0835, -0.0438, -0.1748], + ..., + [ 0.0731, 0.1483, -0.0936, ..., -0.1473, 0.0753, 0.0804], + [ 0.0388, -0.1637, -0.0389, ..., -0.1166, -0.1183, -0.0995], + [-0.2019, -0.0538, -0.0555, ..., 0.1070, -0.2686, 0.1018]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, -4.6566e-09, 0.0000e+00, ..., 4.1910e-09, + -4.6566e-10, 0.0000e+00], + [-1.7835e-07, 1.8626e-09, 0.0000e+00, ..., -2.7334e-07, + 4.1910e-09, -5.1688e-08], + [ 1.4435e-08, 9.3132e-10, 0.0000e+00, ..., 3.0268e-08, + 2.0955e-08, 5.5414e-08], + ..., + [ 6.9384e-08, -1.8626e-09, 0.0000e+00, ..., 1.2107e-07, + -1.0245e-08, -2.9337e-08], + [ 5.5879e-09, 4.6566e-10, 0.0000e+00, ..., 1.0245e-08, + 4.6566e-10, 3.7253e-09], + [ 6.4261e-08, 2.3283e-09, 0.0000e+00, ..., 6.0070e-08, + 5.5879e-09, 4.5169e-08]], device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0225, -0.0312, 0.0200, -0.0256, 0.0369, 0.0200, 0.0030, -0.0121, + -0.0191, -0.0118], device='cuda:0'), grad: tensor([ 2.6543e-08, -1.0561e-06, -7.1013e-07, -1.5367e-08, 8.8010e-08, + 3.9116e-08, 4.0513e-08, 1.1232e-06, 1.2573e-07, 3.4366e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 220.66, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4857 re_mapping 0.0036 re_causal 0.0103 /// teacc 99.00 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.2617, 0.2116, -0.0385, ..., -0.1118, 0.0191, 0.0121], + [ 0.0120, -0.1336, -0.0457, ..., 0.0379, -0.0463, -0.0675], + [-0.1775, -0.2797, 0.0590, ..., -0.0838, -0.0446, -0.1750], + ..., + [ 0.0730, 0.1486, -0.0936, ..., -0.1473, 0.0755, 0.0804], + [ 0.0387, -0.1643, -0.0395, ..., -0.1175, -0.1185, -0.0998], + [-0.2024, -0.0546, -0.0556, ..., 0.1069, -0.2691, 0.1018]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.9092e-08, 2.7940e-09, ..., 5.5414e-08, + 2.7940e-09, 2.5611e-08], + [-1.3970e-09, 4.6566e-10, 4.6566e-10, ..., 2.6776e-07, + 7.9162e-09, 6.2771e-07], + [ 0.0000e+00, 4.6566e-10, -6.5193e-09, ..., 1.3504e-08, + 5.1223e-09, 4.2375e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 4.6566e-10, ..., 6.6590e-08, + -9.3132e-10, -1.4110e-07], + [ 1.3970e-09, 9.3132e-10, 1.3970e-09, ..., 6.5658e-08, + 2.6077e-08, -9.3132e-10], + [ 9.3132e-10, 3.7253e-09, 4.6566e-10, ..., -4.0000e-07, + 2.3283e-08, -6.8033e-07]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0226, -0.0311, 0.0201, -0.0260, 0.0370, 0.0203, 0.0033, -0.0122, + -0.0191, -0.0121], device='cuda:0'), grad: tensor([ 2.4075e-07, 1.4696e-06, 4.3306e-08, 1.7062e-06, 2.2585e-07, + -1.7732e-06, -6.7987e-07, -2.3702e-07, 2.7660e-07, -1.2796e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 220.81, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4917 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.22 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.2623, 0.2120, -0.0385, ..., -0.1122, 0.0189, 0.0121], + [ 0.0120, -0.1339, -0.0458, ..., 0.0378, -0.0466, -0.0677], + [-0.1780, -0.2799, 0.0590, ..., -0.0841, -0.0451, -0.1754], + ..., + [ 0.0730, 0.1490, -0.0936, ..., -0.1474, 0.0761, 0.0806], + [ 0.0386, -0.1648, -0.0433, ..., -0.1190, -0.1187, -0.1002], + [-0.2025, -0.0550, -0.0557, ..., 0.1073, -0.2698, 0.1026]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 1.3039e-08, + 0.0000e+00, 4.6566e-09], + [ 2.7940e-09, 2.7940e-09, 4.6566e-10, ..., 2.7940e-09, + 1.3970e-09, 7.4506e-09], + [ 1.8626e-09, 1.3970e-09, 4.6566e-10, ..., 6.9849e-09, + 9.3132e-10, 1.3970e-09], + ..., + [ 4.6566e-10, -5.1223e-09, 0.0000e+00, ..., 2.7474e-08, + -9.3132e-10, 8.8476e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 1.3970e-09], + [ 8.8476e-09, 1.8626e-09, 0.0000e+00, ..., -3.2503e-07, + 9.3132e-10, -1.8487e-07]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0223, -0.0312, 0.0199, -0.0269, 0.0368, 0.0198, 0.0068, -0.0121, + -0.0199, -0.0118], device='cuda:0'), grad: tensor([ 6.1467e-08, -3.2131e-08, 5.9605e-08, -1.4901e-08, 5.2899e-07, + 1.6904e-07, 1.0291e-07, 8.4750e-08, 2.5146e-08, -9.7044e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 220.88, cls_loss 0.0006 cls_loss_mapping 0.0024 cls_loss_causal 0.5029 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.17 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.2627, 0.2130, -0.0383, ..., -0.1127, 0.0187, 0.0116], + [ 0.0120, -0.1344, -0.0458, ..., 0.0380, -0.0470, -0.0677], + [-0.1788, -0.2801, 0.0591, ..., -0.0843, -0.0456, -0.1758], + ..., + [ 0.0730, 0.1499, -0.0936, ..., -0.1475, 0.0764, 0.0807], + [ 0.0374, -0.1655, -0.0434, ..., -0.1223, -0.1189, -0.1005], + [-0.2027, -0.0569, -0.0558, ..., 0.1074, -0.2704, 0.1029]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, -8.7544e-08, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 2.3283e-10], + [ 1.1642e-09, 2.6310e-08, 0.0000e+00, ..., -1.8626e-09, + 1.0710e-08, 2.6077e-08], + [ 2.0955e-09, 1.3970e-08, 0.0000e+00, ..., 9.3132e-10, + 3.4925e-09, 1.4668e-08], + ..., + [-2.0955e-08, -1.2806e-08, 0.0000e+00, ..., 1.3039e-08, + -3.7253e-08, -6.1700e-08], + [ 4.6566e-10, 4.4238e-09, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 1.1642e-09], + [ 1.3970e-09, 1.6065e-08, 0.0000e+00, ..., -1.8859e-08, + 1.8626e-09, -1.2107e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0222, -0.0310, 0.0198, -0.0266, 0.0368, 0.0193, 0.0073, -0.0121, + -0.0216, -0.0119], device='cuda:0'), grad: tensor([-9.1968e-08, 1.7202e-06, -5.6773e-06, 1.3504e-08, 2.5183e-06, + 2.4447e-08, 1.5441e-06, -1.1059e-07, 6.4727e-08, 0.0000e+00], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 220.45, cls_loss 0.0006 cls_loss_mapping 0.0022 cls_loss_causal 0.4973 re_mapping 0.0036 re_causal 0.0109 /// teacc 99.17 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.2633, 0.2134, -0.0383, ..., -0.1129, 0.0186, 0.0116], + [ 0.0120, -0.1348, -0.0457, ..., 0.0380, -0.0472, -0.0680], + [-0.1792, -0.2806, 0.0592, ..., -0.0844, -0.0457, -0.1761], + ..., + [ 0.0731, 0.1511, -0.0937, ..., -0.1475, 0.0767, 0.0814], + [ 0.0375, -0.1667, -0.0435, ..., -0.1224, -0.1190, -0.1008], + [-0.2033, -0.0575, -0.0564, ..., 0.1073, -0.2708, 0.1023]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.6298e-09, 6.9849e-10, ..., 2.2585e-08, + 6.9849e-10, 2.7940e-09], + [-1.7462e-08, 7.7300e-08, 1.3970e-09, ..., -4.4471e-08, + 2.3050e-08, 4.9826e-08], + [ 1.1874e-08, 1.4668e-08, 6.7521e-09, ..., 1.6531e-08, + 5.5879e-09, 1.5600e-08], + ..., + [-2.0955e-09, -1.3527e-07, 1.1176e-08, ..., 2.1188e-08, + -3.9116e-08, -8.7079e-08], + [ 1.3970e-09, 4.6566e-10, -9.4296e-08, ..., 3.2596e-09, + 2.3283e-09, 1.3970e-09], + [ 3.4925e-09, 4.1211e-08, 4.4238e-09, ..., -1.2573e-08, + 1.1874e-08, 1.2107e-08]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0222, -0.0311, 0.0199, -0.0245, 0.0369, 0.0173, 0.0074, -0.0118, + -0.0215, -0.0127], device='cuda:0'), grad: tensor([ 1.1455e-07, 1.2130e-07, -1.1325e-06, 9.8906e-07, 5.7509e-08, + -6.3982e-07, -5.2154e-08, 6.2725e-07, -1.7718e-07, 1.1153e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 220.40, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4912 re_mapping 0.0036 re_causal 0.0106 /// teacc 99.13 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.2637, 0.2139, -0.0379, ..., -0.1130, 0.0186, 0.0117], + [ 0.0120, -0.1349, -0.0457, ..., 0.0381, -0.0473, -0.0682], + [-0.1795, -0.2807, 0.0593, ..., -0.0844, -0.0459, -0.1777], + ..., + [ 0.0731, 0.1513, -0.0937, ..., -0.1476, 0.0765, 0.0818], + [ 0.0376, -0.1674, -0.0435, ..., -0.1227, -0.1192, -0.1010], + [-0.2034, -0.0577, -0.0565, ..., 0.1073, -0.2710, 0.1023]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -3.0268e-09, 0.0000e+00, ..., 6.0536e-09, + 6.9849e-10, 1.0943e-08], + [ 6.9849e-10, 7.2177e-09, 0.0000e+00, ..., -8.1491e-09, + 5.1223e-09, 9.7789e-09], + [ 3.9581e-09, 8.6147e-09, 0.0000e+00, ..., 4.6566e-09, + 3.0291e-07, 3.6787e-07], + ..., + [ 9.5461e-09, -2.7474e-08, 2.3283e-10, ..., 8.3121e-08, + -1.3737e-08, 1.1781e-07], + [ 1.3271e-08, 2.3283e-10, 0.0000e+00, ..., 1.8626e-09, + 9.3132e-10, 2.3283e-09], + [ 7.6834e-09, 2.3283e-09, 0.0000e+00, ..., -9.4296e-08, + 9.3132e-10, -1.7532e-07]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0225, -0.0311, 0.0195, -0.0246, 0.0369, 0.0172, 0.0076, -0.0117, + -0.0216, -0.0128], device='cuda:0'), grad: tensor([ 1.8859e-08, 2.5844e-08, 2.0899e-06, -2.0824e-06, -2.4214e-08, + 2.1420e-08, 5.4250e-08, 2.8871e-07, -4.6799e-08, -3.2829e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 220.32, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.5112 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.16 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.2646, 0.2142, -0.0382, ..., -0.1134, 0.0180, 0.0115], + [ 0.0120, -0.1352, -0.0458, ..., 0.0380, -0.0476, -0.0682], + [-0.1806, -0.2810, 0.0595, ..., -0.0846, -0.0466, -0.1782], + ..., + [ 0.0731, 0.1520, -0.0937, ..., -0.1477, 0.0767, 0.0819], + [ 0.0384, -0.1677, -0.0433, ..., -0.1226, -0.1193, -0.1012], + [-0.2038, -0.0578, -0.0567, ..., 0.1075, -0.2717, 0.1028]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, -3.6997e-07, 0.0000e+00, ..., 1.0477e-08, + 0.0000e+00, 1.3970e-09], + [-8.6147e-08, 7.3574e-08, 6.9849e-10, ..., -1.9209e-07, + 4.6566e-10, 4.4238e-09], + [ 4.9593e-08, 3.3528e-08, -5.3551e-09, ..., 9.5461e-08, + 2.3283e-10, 2.3283e-09], + ..., + [ 1.6531e-08, 1.6764e-08, 4.6566e-09, ..., 5.7509e-08, + -2.0955e-09, -1.7695e-08], + [ 2.0955e-09, 1.6484e-07, 0.0000e+00, ..., 8.1491e-09, + 0.0000e+00, 9.3132e-10], + [ 3.4226e-08, 2.0489e-08, 0.0000e+00, ..., 1.1153e-07, + 6.9849e-10, 1.2107e-08]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0224, -0.0311, 0.0193, -0.0246, 0.0369, 0.0172, 0.0074, -0.0117, + -0.0201, -0.0126], device='cuda:0'), grad: tensor([-6.6031e-07, -1.1222e-06, 6.7847e-07, 4.1444e-08, -2.1071e-07, + 7.9861e-08, 6.5425e-08, 4.0513e-07, 3.6275e-07, 3.6135e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 220.60, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.5230 re_mapping 0.0036 re_causal 0.0104 /// teacc 99.19 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.2653, 0.2158, -0.0384, ..., -0.1139, 0.0177, 0.0116], + [ 0.0120, -0.1355, -0.0458, ..., 0.0380, -0.0480, -0.0683], + [-0.1817, -0.2814, 0.0593, ..., -0.0850, -0.0471, -0.1783], + ..., + [ 0.0731, 0.1522, -0.0937, ..., -0.1477, 0.0774, 0.0820], + [ 0.0386, -0.1685, -0.0431, ..., -0.1223, -0.1196, -0.1013], + [-0.2040, -0.0585, -0.0570, ..., 0.1076, -0.2725, 0.1030]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -6.1467e-08, 0.0000e+00, ..., 2.0955e-09, + 1.8626e-09, -1.3970e-09], + [ 6.9849e-10, 1.6298e-09, 0.0000e+00, ..., 9.3132e-10, + 6.7521e-09, 2.7940e-09], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., 2.3283e-10, + 4.4238e-09, 1.8626e-09], + ..., + [ 4.1910e-09, 1.1642e-09, 0.0000e+00, ..., 5.1223e-09, + 3.2829e-08, 1.3039e-08], + [ 2.4680e-08, 1.3970e-09, 0.0000e+00, ..., -2.3283e-10, + 6.8266e-07, 2.6217e-07], + [ 1.1642e-08, 1.3970e-08, 0.0000e+00, ..., -2.8638e-08, + 4.1211e-08, -1.6065e-08]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0228, -0.0311, 0.0193, -0.0248, 0.0368, 0.0172, 0.0078, -0.0118, + -0.0196, -0.0127], device='cuda:0'), grad: tensor([-1.1176e-07, 5.7742e-08, 4.6799e-08, 3.4153e-05, 7.6601e-08, + -3.9011e-05, 6.9523e-07, 2.3888e-07, 3.5316e-06, 2.7567e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 220.17, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4793 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.17 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.2666, 0.2164, -0.0384, ..., -0.1140, 0.0174, 0.0117], + [ 0.0122, -0.1358, -0.0452, ..., 0.0381, -0.0482, -0.0683], + [-0.1866, -0.2817, 0.0568, ..., -0.0869, -0.0473, -0.1787], + ..., + [ 0.0731, 0.1528, -0.0938, ..., -0.1478, 0.0776, 0.0821], + [ 0.0387, -0.1690, -0.0433, ..., -0.1227, -0.1198, -0.1015], + [-0.2043, -0.0592, -0.0577, ..., 0.1077, -0.2731, 0.1031]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.3970e-09, 0.0000e+00, ..., 6.9849e-10, + 8.3819e-09, 8.8476e-09], + [ 6.0536e-09, 2.4517e-07, -4.6566e-10, ..., -3.0268e-09, + 1.5204e-07, 3.7765e-07], + [ 3.7253e-09, 8.8476e-09, 0.0000e+00, ..., 3.7253e-09, + 4.1211e-08, 4.3306e-08], + ..., + [-4.6566e-10, -3.1246e-07, 0.0000e+00, ..., 9.7789e-09, + -9.1968e-08, -3.8720e-07], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 4.8894e-09, 4.4238e-09], + [ 8.3819e-09, 5.4482e-08, 0.0000e+00, ..., 3.4925e-09, + 3.4925e-08, 7.6601e-08]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0229, -0.0308, 0.0179, -0.0250, 0.0368, 0.0169, 0.0091, -0.0118, + -0.0189, -0.0128], device='cuda:0'), grad: tensor([ 3.3528e-08, 1.1884e-06, 1.8603e-07, -5.3318e-07, -1.2689e-07, + 1.8929e-07, 2.7241e-08, -1.0617e-06, -1.8207e-07, 3.0082e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 220.32, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4377 re_mapping 0.0035 re_causal 0.0099 /// teacc 99.13 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.2695, 0.2174, -0.0385, ..., -0.1129, 0.0171, 0.0117], + [ 0.0123, -0.1370, -0.0450, ..., 0.0384, -0.0487, -0.0684], + [-0.1870, -0.2820, 0.0569, ..., -0.0872, -0.0474, -0.1788], + ..., + [ 0.0730, 0.1539, -0.0939, ..., -0.1481, 0.0779, 0.0822], + [ 0.0388, -0.1706, -0.0435, ..., -0.1230, -0.1199, -0.1016], + [-0.2049, -0.0596, -0.0583, ..., 0.1074, -0.2735, 0.1033]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-10, 4.6566e-10], + [ 1.1642e-09, 2.0955e-09, 0.0000e+00, ..., 6.9849e-10, + 2.0955e-09, 2.7940e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 1.6298e-09, + 6.9849e-10, 9.3132e-10], + ..., + [ 4.6566e-10, -3.2596e-09, 0.0000e+00, ..., 2.3283e-09, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 4.6566e-10], + [ 5.8208e-09, 9.3132e-10, 0.0000e+00, ..., -9.7789e-09, + 3.0268e-09, -4.4238e-09]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0238, -0.0306, 0.0181, -0.0251, 0.0372, 0.0169, 0.0090, -0.0120, + -0.0187, -0.0133], device='cuda:0'), grad: tensor([ 3.2596e-08, 2.1188e-08, -1.1176e-08, 7.8930e-08, 1.4435e-08, + -2.8638e-07, 1.3271e-08, 2.8405e-08, -5.5879e-09, 1.2433e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 220.54, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4711 re_mapping 0.0034 re_causal 0.0104 /// teacc 99.21 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.2699, 0.2179, -0.0382, ..., -0.1130, 0.0170, 0.0118], + [ 0.0123, -0.1373, -0.0450, ..., 0.0384, -0.0489, -0.0685], + [-0.1871, -0.2822, 0.0574, ..., -0.0872, -0.0470, -0.1789], + ..., + [ 0.0730, 0.1541, -0.0940, ..., -0.1482, 0.0779, 0.0823], + [ 0.0388, -0.1721, -0.0435, ..., -0.1230, -0.1200, -0.1017], + [-0.2051, -0.0599, -0.0584, ..., 0.1074, -0.2738, 0.1034]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -7.7765e-08, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, -1.8626e-09], + [-1.6298e-09, 3.4925e-09, 0.0000e+00, ..., 9.3132e-10, + 6.9849e-10, -8.3819e-08], + [ 1.8626e-09, 7.9162e-09, 0.0000e+00, ..., 3.2596e-09, + 2.3283e-10, 1.3039e-08], + ..., + [ 2.7940e-09, 1.1642e-09, 0.0000e+00, ..., 8.1491e-09, + -9.3132e-10, 7.0082e-08], + [ 4.6566e-10, 6.5193e-09, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 6.9849e-10], + [ 7.6834e-09, 3.4226e-08, 0.0000e+00, ..., 1.0477e-08, + 4.6566e-10, -4.8894e-09]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0241, -0.0306, 0.0184, -0.0252, 0.0373, 0.0170, 0.0086, -0.0121, + -0.0182, -0.0133], device='cuda:0'), grad: tensor([-1.6438e-07, -6.7148e-07, 1.3015e-07, 2.3283e-08, -8.2422e-08, + 1.1642e-08, 6.8685e-08, 5.7463e-07, 2.0489e-08, 1.1688e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 220.96, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4878 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.2704, 0.2181, -0.0383, ..., -0.1132, 0.0163, 0.0117], + [ 0.0123, -0.1372, -0.0451, ..., 0.0383, -0.0491, -0.0685], + [-0.1874, -0.2823, 0.0575, ..., -0.0873, -0.0472, -0.1790], + ..., + [ 0.0730, 0.1539, -0.0940, ..., -0.1484, 0.0777, 0.0816], + [ 0.0387, -0.1730, -0.0435, ..., -0.1233, -0.1199, -0.1018], + [-0.2052, -0.0600, -0.0585, ..., 0.1076, -0.2747, 0.1048]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -2.0955e-09, 0.0000e+00, ..., 2.3283e-09, + 2.3283e-10, 1.1642e-09], + [-2.3749e-08, 2.3283e-09, 0.0000e+00, ..., -2.5611e-08, + 4.1910e-09, 5.5879e-09], + [ 1.4435e-08, 1.1642e-09, 0.0000e+00, ..., 2.0256e-08, + 1.6298e-09, 2.0955e-09], + ..., + [ 3.2596e-08, 1.5134e-08, 0.0000e+00, ..., 1.4668e-08, + -2.0955e-09, 2.3283e-10], + [-2.6077e-08, -1.9092e-08, 0.0000e+00, ..., -9.3132e-10, + -1.8626e-09, 6.9849e-10], + [ 2.0955e-09, 3.4925e-09, 0.0000e+00, ..., -4.3772e-08, + 4.4238e-09, -3.6787e-08]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0240, -0.0306, 0.0184, -0.0252, 0.0373, 0.0171, 0.0084, -0.0123, + -0.0181, -0.0127], device='cuda:0'), grad: tensor([ 9.7789e-09, -9.8255e-08, 9.6625e-08, 2.9593e-07, 5.8906e-08, + -4.4284e-07, 1.4063e-07, 1.8161e-07, -1.6973e-07, -6.0536e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 220.48, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4930 re_mapping 0.0036 re_causal 0.0101 /// teacc 99.07 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.2712, 0.2186, -0.0383, ..., -0.1135, 0.0158, 0.0115], + [ 0.0123, -0.1377, -0.0452, ..., 0.0382, -0.0496, -0.0689], + [-0.1876, -0.2826, 0.0583, ..., -0.0876, -0.0476, -0.1792], + ..., + [ 0.0730, 0.1542, -0.0941, ..., -0.1486, 0.0766, 0.0812], + [ 0.0386, -0.1741, -0.0435, ..., -0.1239, -0.1204, -0.1024], + [-0.2054, -0.0602, -0.0561, ..., 0.1095, -0.2758, 0.1076]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.0513e-08, 1.6298e-09, ..., 2.3283e-08, + 9.3132e-10, 1.6065e-08], + [-5.5879e-09, 2.3283e-09, 5.3551e-09, ..., 1.4203e-08, + 3.9581e-09, 2.0256e-08], + [ 5.3551e-09, 3.9581e-09, 7.9162e-09, ..., 1.1409e-08, + 3.7253e-09, 1.1176e-08], + ..., + [-1.9558e-08, -2.7940e-09, -1.7532e-07, ..., 3.5367e-07, + -6.0303e-08, 9.3132e-08], + [ 2.3283e-10, 9.3132e-10, 2.3283e-10, ..., 5.5879e-09, + 2.8405e-08, 3.4925e-09], + [ 2.0489e-08, 1.5134e-08, 1.0245e-08, ..., -2.0955e-06, + 1.3271e-08, -1.4855e-06]], device='cuda:0') +Epoch 287, bias, value: tensor([ 0.0239, -0.0308, 0.0187, -0.0251, 0.0358, 0.0171, 0.0085, -0.0125, + -0.0187, -0.0106], device='cuda:0'), grad: tensor([ 2.7940e-09, 3.6554e-08, 5.5181e-08, 7.2718e-06, 4.0382e-06, + -6.9775e-06, 6.3796e-08, 2.6426e-07, 1.7579e-07, -4.9099e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 220.46, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4887 re_mapping 0.0038 re_causal 0.0106 /// teacc 99.12 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.2721, 0.2187, -0.0384, ..., -0.1139, 0.0148, 0.0114], + [ 0.0124, -0.1382, -0.0441, ..., 0.0387, -0.0511, -0.0692], + [-0.1882, -0.2828, 0.0583, ..., -0.0878, -0.0483, -0.1796], + ..., + [ 0.0730, 0.1549, -0.0950, ..., -0.1491, 0.0782, 0.0816], + [ 0.0384, -0.1746, -0.0431, ..., -0.1242, -0.1208, -0.1027], + [-0.2059, -0.0605, -0.0584, ..., 0.1093, -0.2771, 0.1077]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, -2.1281e-07, -3.4925e-09, ..., 9.3132e-10, + -4.5635e-08, -4.8429e-08], + [-9.6858e-08, 9.0804e-09, 4.6566e-10, ..., -8.8476e-08, + 3.7253e-09, 4.1910e-09], + [ 3.2131e-08, 7.4506e-09, -4.6566e-10, ..., 2.7241e-08, + 3.2596e-09, 3.0268e-09], + ..., + [ 2.5146e-08, 8.8476e-09, 2.3283e-10, ..., 2.0489e-08, + -9.3132e-10, -4.4238e-09], + [-9.0804e-09, 1.1874e-08, 4.6566e-10, ..., 4.4238e-09, + 2.7940e-09, 3.9581e-09], + [ 1.6065e-08, 5.0990e-08, 6.9849e-10, ..., 1.3970e-08, + 8.8476e-09, 1.0477e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0235, -0.0305, 0.0188, -0.0256, 0.0362, 0.0175, 0.0084, -0.0127, + -0.0187, -0.0110], device='cuda:0'), grad: tensor([-5.4389e-07, 3.2410e-07, -1.0617e-06, 1.2387e-07, 9.3132e-08, + 1.6461e-07, 2.6845e-07, 3.7928e-07, 9.0804e-09, 2.4703e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 220.29, cls_loss 0.0008 cls_loss_mapping 0.0027 cls_loss_causal 0.4796 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.13 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.2728, 0.2189, -0.0389, ..., -0.1143, 0.0147, 0.0114], + [ 0.0124, -0.1383, -0.0442, ..., 0.0385, -0.0515, -0.0698], + [-0.1886, -0.2830, 0.0584, ..., -0.0881, -0.0488, -0.1802], + ..., + [ 0.0730, 0.1554, -0.0950, ..., -0.1495, 0.0784, 0.0814], + [ 0.0382, -0.1757, -0.0431, ..., -0.1245, -0.1213, -0.1031], + [-0.2063, -0.0608, -0.0585, ..., 0.1062, -0.2777, 0.1050]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, -1.9325e-08, 0.0000e+00, ..., 6.9849e-10, + 1.6298e-09, 0.0000e+00], + [ 6.2864e-09, 1.8626e-09, 1.3970e-09, ..., 4.6566e-10, + 1.4668e-08, 1.2806e-08], + [ 2.3283e-10, 1.6298e-09, 0.0000e+00, ..., 4.6566e-10, + 3.9581e-09, 2.7940e-09], + ..., + [-5.8208e-09, -2.0955e-09, 4.6566e-10, ..., 1.1642e-09, + -2.7241e-08, -5.1456e-08], + [ 7.6834e-09, 2.3283e-09, -4.1910e-09, ..., 9.3132e-10, + 3.0268e-08, 9.3132e-10], + [ 7.2177e-09, 2.7940e-09, 0.0000e+00, ..., -2.0955e-09, + 2.2352e-08, 3.2131e-08]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0233, -0.0308, 0.0189, -0.0264, 0.0395, 0.0184, 0.0082, -0.0129, + -0.0190, -0.0142], device='cuda:0'), grad: tensor([-2.1420e-08, 9.4995e-08, 1.6531e-08, 7.9582e-07, -6.0536e-09, + -1.0338e-06, 1.8207e-07, -1.3970e-07, 1.6298e-09, 1.2456e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 220.85, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.5005 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.10 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.2735, 0.2192, -0.0391, ..., -0.1149, 0.0148, 0.0113], + [ 0.0124, -0.1388, -0.0442, ..., 0.0385, -0.0522, -0.0700], + [-0.1889, -0.2832, 0.0584, ..., -0.0884, -0.0490, -0.1808], + ..., + [ 0.0730, 0.1572, -0.0950, ..., -0.1496, 0.0804, 0.0825], + [ 0.0381, -0.1782, -0.0432, ..., -0.1251, -0.1219, -0.1037], + [-0.2068, -0.0620, -0.0587, ..., 0.1062, -0.2816, 0.1048]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.2131e-08, -9.3132e-09, ..., 2.5146e-08, + 6.2864e-09, 7.4506e-09], + [ 1.0710e-08, 8.1491e-09, 6.5193e-09, ..., 6.5193e-09, + 4.6333e-08, 9.3831e-08], + [ 1.1642e-09, 2.3283e-09, 6.9849e-10, ..., 1.6298e-09, + 4.1910e-09, 6.2864e-09], + ..., + [-1.3039e-08, -8.3819e-09, -8.1491e-09, ..., 1.2340e-08, + -3.9116e-08, -1.1688e-07], + [ 4.1910e-09, 1.3970e-09, 2.3283e-10, ..., 4.4238e-09, + 3.1199e-08, 5.8208e-09], + [ 6.0536e-09, 2.0955e-08, 6.9849e-09, ..., -4.1653e-07, + 1.8859e-08, -2.2701e-07]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0230, -0.0308, 0.0187, -0.0265, 0.0395, 0.0185, 0.0085, -0.0125, + -0.0195, -0.0143], device='cuda:0'), grad: tensor([ 1.5134e-08, 7.1712e-07, -3.5134e-07, 4.4480e-06, 2.1397e-07, + -4.2841e-06, 1.9162e-07, -2.2980e-07, 4.2142e-08, -7.4599e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 221.23, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4884 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.16 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.2743, 0.2193, -0.0396, ..., -0.1157, 0.0153, 0.0115], + [ 0.0123, -0.1391, -0.0443, ..., 0.0385, -0.0532, -0.0709], + [-0.1902, -0.2838, 0.0585, ..., -0.0887, -0.0501, -0.1814], + ..., + [ 0.0732, 0.1580, -0.0949, ..., -0.1497, 0.0815, 0.0835], + [ 0.0381, -0.1794, -0.0435, ..., -0.1253, -0.1220, -0.1040], + [-0.2075, -0.0623, -0.0591, ..., 0.1061, -0.2827, 0.1047]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.2573e-08, 1.6298e-09, ..., 1.9558e-08, + 3.2596e-09, 1.9558e-08], + [ 4.9546e-07, 1.3271e-08, 4.5635e-08, ..., 5.4529e-07, + 8.8243e-08, 3.0175e-07], + [ 2.5146e-08, 1.3970e-08, 7.6834e-09, ..., 8.8476e-09, + 6.4727e-08, 7.4739e-08], + ..., + [-1.9558e-08, -2.3982e-08, -3.0734e-08, ..., 8.8708e-08, + 1.0524e-07, 6.7987e-08], + [ 9.7789e-09, 1.6298e-09, 9.3132e-10, ..., 1.5600e-08, + 1.3039e-08, 2.2817e-08], + [ 1.6275e-07, 4.1910e-09, -3.9814e-08, ..., -2.4401e-07, + 1.0943e-08, -3.5460e-07]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0218, -0.0311, 0.0187, -0.0265, 0.0396, 0.0184, 0.0091, -0.0121, + -0.0195, -0.0143], device='cuda:0'), grad: tensor([ 8.0327e-08, 3.3509e-06, 1.8184e-07, -1.7509e-06, -3.0771e-06, + 9.8534e-07, 3.6089e-08, 5.4296e-07, 1.3225e-07, -4.6473e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 220.86, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4890 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.19 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.2750, 0.2231, -0.0396, ..., -0.1128, 0.0158, 0.0147], + [ 0.0122, -0.1410, -0.0442, ..., 0.0385, -0.0549, -0.0720], + [-0.1916, -0.2840, 0.0582, ..., -0.0889, -0.0503, -0.1818], + ..., + [ 0.0734, 0.1597, -0.0947, ..., -0.1497, 0.0823, 0.0846], + [ 0.0367, -0.1818, -0.0457, ..., -0.1254, -0.1223, -0.1043], + [-0.2078, -0.0629, -0.0592, ..., 0.1061, -0.2837, 0.1047]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.1958e-06, 0.0000e+00, ..., 6.4541e-07, + 0.0000e+00, 1.0012e-08], + [ 1.2107e-08, 6.3563e-08, 0.0000e+00, ..., 1.9558e-08, + 3.2596e-09, 7.6834e-08], + [ 3.7253e-09, 9.9419e-08, 0.0000e+00, ..., 5.1921e-08, + 9.3132e-10, 1.7928e-08], + ..., + [-5.3318e-08, -7.9861e-08, 0.0000e+00, ..., 8.1491e-09, + -7.9162e-09, -3.0827e-07], + [ 1.6298e-09, 2.8405e-08, 0.0000e+00, ..., 1.4668e-08, + 0.0000e+00, 5.5879e-09], + [ 3.6089e-08, 9.4296e-08, 0.0000e+00, ..., 1.0012e-08, + 2.5611e-09, 1.8044e-07]], device='cuda:0') +Epoch 292, bias, value: tensor([ 0.0253, -0.0314, 0.0187, -0.0265, 0.0396, 0.0184, 0.0059, -0.0118, + -0.0205, -0.0143], device='cuda:0'), grad: tensor([ 4.3511e-06, 4.6333e-07, 3.1595e-07, 1.7579e-07, 6.5751e-07, + 1.3234e-06, -7.0930e-06, -1.2051e-06, 1.0431e-07, 9.2667e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 220.88, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4911 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.16 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.2762, 0.2232, -0.0402, ..., -0.1129, 0.0157, 0.0148], + [ 0.0123, -0.1413, -0.0432, ..., 0.0388, -0.0555, -0.0722], + [-0.1920, -0.2841, 0.0583, ..., -0.0891, -0.0508, -0.1821], + ..., + [ 0.0733, 0.1601, -0.0958, ..., -0.1501, 0.0825, 0.0848], + [ 0.0366, -0.1826, -0.0459, ..., -0.1257, -0.1227, -0.1047], + [-0.2081, -0.0631, -0.0595, ..., 0.1061, -0.2847, 0.1047]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 5.1456e-08, ..., 2.0489e-07, + 2.3283e-10, 4.4238e-09], + [ 7.4506e-09, 0.0000e+00, 3.2596e-09, ..., 1.6531e-08, + 1.3504e-08, 5.4715e-08], + [ 6.2864e-09, 0.0000e+00, 6.7521e-09, ..., 2.7940e-08, + 6.2864e-09, 1.1642e-08], + ..., + [-1.3504e-08, -2.3283e-10, 0.0000e+00, ..., 3.0268e-09, + -3.0734e-08, -1.4063e-07], + [-6.7521e-09, 0.0000e+00, 3.0268e-09, ..., 1.3504e-08, + 4.6566e-10, 2.7940e-09], + [ 5.5879e-09, 0.0000e+00, 4.6566e-10, ..., -1.0873e-07, + 2.0955e-08, 3.0268e-09]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0253, -0.0312, 0.0190, -0.0265, 0.0396, 0.0185, 0.0058, -0.0120, + -0.0206, -0.0144], device='cuda:0'), grad: tensor([ 1.0226e-06, 2.4540e-07, 1.1036e-07, 6.8219e-08, 1.9185e-07, + -1.4831e-07, -1.2852e-06, -2.7358e-07, 1.1362e-07, -3.6089e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 221.52, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4814 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.16 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.2772, 0.2233, -0.0390, ..., -0.1129, 0.0189, 0.0139], + [ 0.0124, -0.1425, -0.0432, ..., 0.0389, -0.0560, -0.0722], + [-0.1923, -0.2849, 0.0581, ..., -0.0897, -0.0516, -0.1825], + ..., + [ 0.0733, 0.1602, -0.0956, ..., -0.1503, 0.0836, 0.0859], + [ 0.0365, -0.1845, -0.0461, ..., -0.1268, -0.1236, -0.1053], + [-0.2084, -0.0626, -0.0596, ..., 0.1061, -0.2871, 0.1044]], + device='cuda:0'), grad: tensor([[ 2.7264e-07, 5.2387e-07, 0.0000e+00, ..., 2.3330e-07, + -1.3504e-08, -5.3551e-09], + [-3.3458e-07, -7.0268e-07, 2.3283e-10, ..., -2.9290e-07, + 1.2340e-08, 2.3516e-08], + [ 1.5367e-08, 5.1456e-08, -1.3970e-09, ..., 1.7462e-08, + 3.8650e-08, 4.2841e-08], + ..., + [ 1.8859e-08, 3.1432e-08, 1.1642e-09, ..., 1.0035e-07, + 2.6077e-08, 1.6857e-07], + [ 1.3039e-08, 2.8405e-08, 0.0000e+00, ..., 1.0245e-08, + 2.3283e-09, 2.3283e-09], + [ 2.0955e-09, 1.0012e-08, 0.0000e+00, ..., -9.1968e-08, + 5.1223e-09, -1.6904e-07]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0252, -0.0311, 0.0188, -0.0263, 0.0396, 0.0184, 0.0057, -0.0117, + -0.0213, -0.0145], device='cuda:0'), grad: tensor([ 2.5705e-06, -3.2019e-06, 2.9686e-07, -1.1409e-07, 5.6578e-08, + -2.0117e-07, 1.9185e-07, 6.1747e-07, 1.5344e-07, -3.6275e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 220.77, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4736 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.10 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.2783, 0.2234, -0.0395, ..., -0.1129, 0.0191, 0.0140], + [ 0.0124, -0.1426, -0.0432, ..., 0.0389, -0.0562, -0.0723], + [-0.1930, -0.2851, 0.0580, ..., -0.0900, -0.0520, -0.1827], + ..., + [ 0.0733, 0.1604, -0.0956, ..., -0.1504, 0.0836, 0.0860], + [ 0.0363, -0.1838, -0.0461, ..., -0.1270, -0.1238, -0.1052], + [-0.2088, -0.0629, -0.0596, ..., 0.1061, -0.2874, 0.1044]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.6531e-08, 0.0000e+00, ..., 1.0710e-08, + 0.0000e+00, 4.1910e-09], + [-5.1688e-08, 9.3132e-10, 0.0000e+00, ..., -1.0431e-07, + 9.3132e-10, 4.4238e-09], + [ 5.3085e-08, 2.3283e-10, 0.0000e+00, ..., 1.2014e-07, + 2.3283e-10, 4.6566e-10], + ..., + [ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 5.5181e-08, + 4.6566e-09, 4.5169e-08], + [ 1.1642e-09, 4.6566e-10, 0.0000e+00, ..., 5.1223e-09, + 4.6566e-10, 1.3970e-09], + [ 2.4214e-08, 3.2596e-09, 0.0000e+00, ..., -2.4447e-07, + 4.6566e-10, -2.8056e-07]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0252, -0.0311, 0.0187, -0.0264, 0.0396, 0.0185, 0.0057, -0.0117, + -0.0211, -0.0145], device='cuda:0'), grad: tensor([ 3.2596e-09, -8.5775e-07, 8.5263e-07, 9.3132e-09, 4.4703e-07, + 4.9360e-08, -1.8626e-08, 1.6275e-07, -7.6834e-09, -6.3330e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 220.86, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4724 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.11 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.2793, 0.2234, -0.0395, ..., -0.1134, 0.0184, 0.0136], + [ 0.0124, -0.1428, -0.0432, ..., 0.0389, -0.0566, -0.0724], + [-0.1934, -0.2855, 0.0581, ..., -0.0902, -0.0548, -0.1842], + ..., + [ 0.0733, 0.1609, -0.0956, ..., -0.1505, 0.0833, 0.0859], + [ 0.0356, -0.1877, -0.0463, ..., -0.1298, -0.1246, -0.1081], + [-0.2094, -0.0620, -0.0597, ..., 0.1062, -0.2881, 0.1049]], + device='cuda:0'), grad: tensor([[ 4.4238e-09, -5.6811e-08, 0.0000e+00, ..., 3.0035e-08, + 9.3132e-10, -2.0955e-09], + [ 2.1234e-07, 1.6298e-09, 0.0000e+00, ..., 2.5542e-07, + 1.6298e-09, 2.3283e-09], + [ 5.3551e-09, 1.3970e-09, 0.0000e+00, ..., 7.9162e-09, + 4.1910e-09, 4.6566e-09], + ..., + [ 1.9325e-08, 2.7940e-09, 0.0000e+00, ..., 2.4447e-08, + 2.0023e-08, 2.3749e-08], + [ 3.3760e-08, 2.7940e-09, 0.0000e+00, ..., 4.8662e-08, + 6.7521e-09, 7.2177e-09], + [ 3.7742e-07, 1.7462e-08, 0.0000e+00, ..., 4.2492e-07, + 3.7253e-08, 3.2131e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0250, -0.0311, 0.0169, -0.0257, 0.0396, 0.0187, 0.0058, -0.0118, + -0.0230, -0.0143], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.3858e-06, 3.1898e-08, -2.2096e-07, -3.9414e-06, + -9.4064e-08, -1.9092e-07, 2.0326e-07, 2.9732e-07, 2.5313e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 220.43, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4324 re_mapping 0.0035 re_causal 0.0102 /// teacc 99.17 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.2799, 0.2234, -0.0396, ..., -0.1134, 0.0180, 0.0136], + [ 0.0124, -0.1432, -0.0423, ..., 0.0391, -0.0559, -0.0724], + [-0.1942, -0.2856, 0.0581, ..., -0.0903, -0.0550, -0.1846], + ..., + [ 0.0733, 0.1615, -0.0965, ..., -0.1511, 0.0830, 0.0857], + [ 0.0356, -0.1878, -0.0463, ..., -0.1299, -0.1250, -0.1083], + [-0.2102, -0.0623, -0.0603, ..., 0.1062, -0.2890, 0.1050]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -2.0256e-08, 0.0000e+00, ..., 1.1479e-07, + -4.6566e-10, -4.6566e-10], + [-1.6298e-09, 4.6566e-10, 0.0000e+00, ..., 8.1491e-09, + 4.6566e-10, 1.1642e-09], + [ 1.6298e-09, 5.1223e-09, 0.0000e+00, ..., 5.1223e-09, + 2.0955e-09, 1.6298e-09], + ..., + [ 6.0536e-09, 4.6566e-10, 2.3283e-10, ..., 1.7229e-08, + 1.1642e-09, 2.0955e-09], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 5.4017e-08, + 4.6566e-10, 6.9849e-10], + [ 1.3271e-08, 7.4506e-09, 6.9849e-10, ..., 7.9162e-08, + 4.6566e-10, -5.1223e-09]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0250, -0.0308, 0.0168, -0.0259, 0.0396, 0.0190, 0.0058, -0.0121, + -0.0231, -0.0143], device='cuda:0'), grad: tensor([ 5.8254e-07, 5.0990e-08, 3.7253e-09, -2.2352e-08, -1.8650e-07, + 1.6555e-05, -1.7598e-05, 8.6613e-08, 2.9476e-07, 2.8475e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 221.22, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4990 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.16 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.2825, 0.2235, -0.0397, ..., -0.1136, 0.0200, 0.0135], + [ 0.0125, -0.1464, -0.0412, ..., 0.0399, -0.0568, -0.0746], + [-0.1956, -0.2863, 0.0580, ..., -0.0907, -0.0553, -0.1857], + ..., + [ 0.0737, 0.1647, -0.0969, ..., -0.1515, 0.0853, 0.0880], + [ 0.0347, -0.1885, -0.0464, ..., -0.1322, -0.1261, -0.1108], + [-0.2117, -0.0626, -0.0630, ..., 0.1063, -0.2908, 0.1053]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 0.0000e+00, 6.9849e-10, ..., 1.9791e-09, + 1.1642e-10, 1.6298e-09], + [ 4.1910e-09, 1.8626e-09, -5.8208e-10, ..., 5.9372e-09, + 2.4447e-09, 6.4028e-09], + [ 1.8626e-09, 5.8208e-10, -1.7462e-09, ..., -5.0059e-09, + 5.8208e-10, 1.7462e-09], + ..., + [ 2.7125e-08, -3.8417e-09, 6.9849e-10, ..., 1.1642e-08, + 6.2864e-09, -8.2655e-09], + [ 1.1525e-08, 2.3283e-10, 2.3283e-10, ..., 1.2806e-09, + 4.6566e-09, 1.3970e-09], + [ 1.0245e-08, 1.9791e-09, 2.3283e-10, ..., -1.2107e-08, + 1.3970e-09, -1.6065e-08]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0250, -0.0309, 0.0165, -0.0261, 0.0395, 0.0191, 0.0058, -0.0110, + -0.0258, -0.0143], device='cuda:0'), grad: tensor([ 1.0023e-07, 7.5321e-08, -4.1071e-07, 1.9628e-07, -5.2969e-08, + -2.8918e-07, 1.2550e-07, 1.6694e-07, 6.5193e-08, 3.4226e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 297---------------------------------------------------- +epoch 297, time 221.42, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4844 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.24 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.2836, 0.2235, -0.0397, ..., -0.1138, 0.0206, 0.0135], + [ 0.0125, -0.1464, -0.0412, ..., 0.0399, -0.0569, -0.0746], + [-0.1960, -0.2865, 0.0581, ..., -0.0909, -0.0554, -0.1858], + ..., + [ 0.0737, 0.1648, -0.0969, ..., -0.1516, 0.0854, 0.0881], + [ 0.0344, -0.1886, -0.0463, ..., -0.1341, -0.1263, -0.1117], + [-0.2121, -0.0628, -0.0630, ..., 0.1063, -0.2912, 0.1054]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -5.3551e-09, 0.0000e+00, ..., 1.4203e-08, + -1.0477e-09, 2.0489e-08], + [ 3.8417e-09, 4.6566e-10, 0.0000e+00, ..., 3.1432e-09, + 1.0477e-09, 7.7998e-09], + [ 5.8208e-10, 3.4925e-10, 0.0000e+00, ..., 6.9849e-10, + 5.8208e-10, 1.0477e-09], + ..., + [ 3.6089e-09, 2.3283e-10, 0.0000e+00, ..., 1.5949e-08, + 3.8417e-09, 2.2235e-08], + [-1.9791e-08, 0.0000e+00, 0.0000e+00, ..., 1.8976e-08, + 9.3132e-10, 2.1770e-08], + [ 3.8417e-09, 4.8894e-09, 0.0000e+00, ..., -1.2259e-07, + 1.3970e-09, -1.5122e-07]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0250, -0.0308, 0.0166, -0.0263, 0.0395, 0.0193, 0.0057, -0.0110, + -0.0269, -0.0143], device='cuda:0'), grad: tensor([ 6.7637e-08, -5.7416e-07, 4.0862e-08, 1.1723e-07, 1.8708e-07, + 1.4447e-07, 2.0023e-08, 6.3609e-07, -2.4564e-07, -3.7975e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 220.84, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4572 re_mapping 0.0035 re_causal 0.0102 /// teacc 99.12 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.2840, 0.2237, -0.0397, ..., -0.1141, 0.0166, 0.0134], + [ 0.0125, -0.1466, -0.0412, ..., 0.0397, -0.0573, -0.0747], + [-0.1966, -0.2869, 0.0582, ..., -0.0910, -0.0551, -0.1874], + ..., + [ 0.0737, 0.1649, -0.0969, ..., -0.1518, 0.0846, 0.0879], + [ 0.0343, -0.1894, -0.0464, ..., -0.1356, -0.1267, -0.1113], + [-0.2123, -0.0630, -0.0630, ..., 0.1064, -0.2915, 0.1057]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, -4.8894e-09, 0.0000e+00, ..., 3.4925e-09, + 4.4238e-09, 2.3283e-09], + [ 3.0501e-08, 3.2596e-09, 0.0000e+00, ..., 6.9849e-09, + 8.4052e-08, 5.0524e-08], + [ 4.6217e-07, 2.5611e-09, 0.0000e+00, ..., 1.8626e-09, + 2.0228e-06, 1.1381e-06], + ..., + [-7.5484e-07, -6.7521e-09, 0.0000e+00, ..., 6.7521e-09, + -2.5667e-06, -1.4426e-06], + [ 1.8161e-08, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, + 7.5437e-08, 4.3074e-08], + [ 3.0268e-09, 3.9581e-09, 0.0000e+00, ..., 4.1444e-08, + 6.0536e-09, -1.6298e-09]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0249, -0.0311, 0.0173, -0.0262, 0.0394, 0.0194, 0.0060, -0.0113, + -0.0270, -0.0142], device='cuda:0'), grad: tensor([ 2.3050e-08, 4.0373e-07, 7.8827e-06, 1.0589e-06, 7.5856e-07, + -9.5228e-08, 1.2107e-07, -1.0580e-05, 3.1176e-07, 1.5600e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 220.75, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4847 re_mapping 0.0034 re_causal 0.0098 /// teacc 99.22 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.2851, 0.2240, -0.0397, ..., -0.1145, 0.0163, 0.0137], + [ 0.0125, -0.1469, -0.0410, ..., 0.0399, -0.0596, -0.0750], + [-0.1977, -0.2871, 0.0581, ..., -0.0918, -0.0554, -0.1883], + ..., + [ 0.0738, 0.1662, -0.0970, ..., -0.1518, 0.0880, 0.0909], + [ 0.0333, -0.1894, -0.0465, ..., -0.1358, -0.1297, -0.1143], + [-0.2125, -0.0645, -0.0635, ..., 0.1064, -0.2943, 0.1056]], + device='cuda:0'), grad: tensor([[ 9.0804e-09, 1.3271e-08, 0.0000e+00, ..., 2.8173e-08, + 9.3132e-10, 6.9849e-10], + [-1.7695e-08, -2.5146e-08, 0.0000e+00, ..., -5.8906e-08, + 6.7521e-09, 8.1491e-09], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 1.6298e-09, + 4.6566e-09, 4.1910e-09], + ..., + [ 4.6566e-10, 3.0268e-09, 0.0000e+00, ..., 4.4238e-09, + 1.7462e-08, 1.5832e-08], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 4.6566e-10], + [ 4.6566e-10, 2.5611e-09, 0.0000e+00, ..., 2.0955e-09, + 1.4668e-08, 1.3737e-08]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0249, -0.0313, 0.0179, -0.0268, 0.0394, 0.0192, 0.0048, -0.0084, + -0.0291, -0.0143], device='cuda:0'), grad: tensor([ 8.7079e-08, -7.9162e-08, -7.1945e-08, -1.0035e-07, 3.9814e-08, + -7.1712e-08, 6.6357e-08, 3.4925e-09, 2.5844e-08, 1.1595e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 221.09, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4467 re_mapping 0.0038 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.2855, 0.2240, -0.0398, ..., -0.1146, 0.0164, 0.0137], + [ 0.0125, -0.1468, -0.0409, ..., 0.0401, -0.0579, -0.0752], + [-0.1980, -0.2869, 0.0581, ..., -0.0916, -0.0552, -0.1885], + ..., + [ 0.0738, 0.1664, -0.0970, ..., -0.1519, 0.0880, 0.0911], + [ 0.0334, -0.1895, -0.0464, ..., -0.1357, -0.1297, -0.1143], + [-0.2126, -0.0653, -0.0636, ..., 0.1063, -0.2961, 0.1055]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 6.9849e-10, ..., 2.7008e-08, + 4.6566e-10, 2.2119e-08], + [ 4.2841e-08, 0.0000e+00, 2.0489e-08, ..., 4.0513e-08, + 6.0536e-09, 1.4435e-08], + [ 4.8662e-08, 0.0000e+00, 2.1188e-08, ..., 3.5390e-08, + 6.7521e-09, 7.2177e-09], + ..., + [ 2.4773e-07, 0.0000e+00, 1.1781e-07, ..., 3.0175e-07, + 7.4506e-09, 1.0384e-07], + [ 4.8894e-09, 0.0000e+00, 3.2596e-09, ..., 2.7474e-08, + 8.3819e-09, 1.9791e-08], + [ 2.9104e-08, 6.9849e-10, 1.1409e-08, ..., -1.3495e-06, + 3.7253e-09, -1.1818e-06]], device='cuda:0') +Epoch 302, bias, value: tensor([ 0.0249, -0.0313, 0.0183, -0.0267, 0.0394, 0.0193, 0.0048, -0.0085, + -0.0291, -0.0144], device='cuda:0'), grad: tensor([ 9.1037e-08, 2.8731e-07, 6.7987e-08, -5.7975e-08, 1.1064e-06, + 8.3214e-07, -2.1583e-07, 1.5181e-06, 1.8394e-07, -3.8072e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 220.40, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4838 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.12 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.2859, 0.2243, -0.0397, ..., -0.1148, 0.0171, 0.0138], + [ 0.0125, -0.1468, -0.0410, ..., 0.0401, -0.0581, -0.0753], + [-0.1983, -0.2873, 0.0581, ..., -0.0921, -0.0553, -0.1890], + ..., + [ 0.0738, 0.1664, -0.0970, ..., -0.1527, 0.0880, 0.0909], + [ 0.0333, -0.1903, -0.0465, ..., -0.1359, -0.1297, -0.1144], + [-0.2134, -0.0662, -0.0636, ..., 0.1064, -0.2962, 0.1058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.0745e-08, 0.0000e+00, ..., 9.7789e-09, + -4.8894e-09, -5.5879e-09], + [ 3.2596e-09, 3.2596e-09, 0.0000e+00, ..., -2.6124e-07, + -9.2667e-08, -3.6624e-07], + [ 3.4925e-09, 9.0804e-09, 0.0000e+00, ..., 5.8208e-09, + 1.8626e-09, 8.6147e-09], + ..., + [-1.7928e-08, -6.5193e-09, 0.0000e+00, ..., 2.5495e-07, + 8.6380e-08, 3.1316e-07], + [ 2.3283e-10, 9.3132e-10, 2.3283e-10, ..., 6.7521e-09, + 2.3283e-10, 4.6566e-10], + [ 3.0268e-09, 2.5379e-08, 0.0000e+00, ..., 1.8370e-07, + 4.6566e-09, 2.5611e-08]], device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0251, -0.0313, 0.0180, -0.0271, 0.0395, 0.0196, 0.0047, -0.0087, + -0.0291, -0.0144], device='cuda:0'), grad: tensor([-3.2131e-08, -2.7064e-06, 7.0082e-08, 2.2515e-07, -7.3109e-07, + 1.7267e-06, -1.7108e-06, 2.5053e-06, -2.1001e-07, 8.6986e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 220.67, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4640 re_mapping 0.0035 re_causal 0.0101 /// teacc 99.17 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.2892, 0.2243, -0.0394, ..., -0.1150, 0.0171, 0.0138], + [ 0.0126, -0.1466, -0.0409, ..., 0.0404, -0.0584, -0.0753], + [-0.1989, -0.2874, 0.0581, ..., -0.0927, -0.0553, -0.1893], + ..., + [ 0.0737, 0.1662, -0.0971, ..., -0.1530, 0.0880, 0.0909], + [ 0.0333, -0.1903, -0.0465, ..., -0.1360, -0.1297, -0.1144], + [-0.2140, -0.0664, -0.0636, ..., 0.1065, -0.2966, 0.1064]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -5.1223e-09, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 2.3283e-10], + [ 5.3551e-09, 4.6566e-10, 0.0000e+00, ..., 1.0245e-08, + 2.5611e-09, 6.0536e-09], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 1.6298e-09, + 3.2596e-09, 2.7940e-09], + ..., + [ 1.3201e-07, 6.9849e-10, 0.0000e+00, ..., 2.0489e-07, + 1.3970e-09, 1.4692e-07], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 4.6566e-10], + [ 2.8266e-07, 6.9849e-10, 0.0000e+00, ..., 4.6194e-07, + 0.0000e+00, 2.9430e-07]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0251, -0.0310, 0.0181, -0.0273, 0.0394, 0.0196, 0.0046, -0.0087, + -0.0291, -0.0143], device='cuda:0'), grad: tensor([ 5.3085e-08, 6.6357e-08, -2.7218e-07, 1.7951e-07, -2.2575e-06, + -1.6810e-07, -6.5193e-09, 7.0501e-07, 1.1106e-07, 1.5991e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 220.59, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4552 re_mapping 0.0033 re_causal 0.0099 /// teacc 99.19 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.2899, 0.2244, -0.0391, ..., -0.1153, 0.0171, 0.0137], + [ 0.0126, -0.1467, -0.0409, ..., 0.0404, -0.0586, -0.0755], + [-0.1992, -0.2877, 0.0581, ..., -0.0931, -0.0553, -0.1895], + ..., + [ 0.0738, 0.1664, -0.0971, ..., -0.1530, 0.0881, 0.0911], + [ 0.0332, -0.1904, -0.0465, ..., -0.1360, -0.1297, -0.1144], + [-0.2144, -0.0665, -0.0637, ..., 0.1065, -0.2969, 0.1064]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.0804e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -1.8626e-09], + [-6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 7.9162e-09, + 2.3283e-10, 4.6566e-09], + [ 1.1642e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 3.9581e-09, + 0.0000e+00, 1.1642e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10], + [ 2.5611e-09, 4.6566e-09, 0.0000e+00, ..., -2.0489e-08, + 0.0000e+00, -8.6147e-09]], device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0251, -0.0310, 0.0183, -0.0270, 0.0394, 0.0192, 0.0048, -0.0087, + -0.0291, -0.0143], device='cuda:0'), grad: tensor([-1.3737e-08, 1.1642e-08, 9.0804e-09, 1.9092e-08, 1.6531e-08, + -1.8626e-08, 7.4506e-09, 1.1642e-08, -1.8626e-09, -1.8626e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 220.68, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4593 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.11 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.2902, 0.2244, -0.0392, ..., -0.1154, 0.0171, 0.0137], + [ 0.0127, -0.1467, -0.0409, ..., 0.0403, -0.0588, -0.0755], + [-0.2013, -0.2879, 0.0581, ..., -0.0932, -0.0554, -0.1897], + ..., + [ 0.0738, 0.1665, -0.0972, ..., -0.1531, 0.0880, 0.0910], + [ 0.0331, -0.1904, -0.0465, ..., -0.1365, -0.1297, -0.1145], + [-0.2155, -0.0667, -0.0637, ..., 0.1066, -0.2971, 0.1066]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3504e-08, 0.0000e+00, ..., 1.6298e-09, + 0.0000e+00, -2.0955e-09], + [ 9.3132e-10, 4.4238e-09, 0.0000e+00, ..., -6.9849e-10, + 9.3132e-10, 6.7521e-09], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 1.6298e-09], + ..., + [-9.5461e-09, -2.6310e-08, 0.0000e+00, ..., 3.0268e-09, + -9.3132e-10, -3.7253e-08], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 9.5461e-09, + 0.0000e+00, 7.2177e-09], + [ 9.0804e-09, 2.5379e-08, 0.0000e+00, ..., -1.3830e-07, + 2.3283e-10, -6.6590e-08]], device='cuda:0') +Epoch 306, bias, value: tensor([ 0.0250, -0.0309, 0.0178, -0.0268, 0.0394, 0.0190, 0.0047, -0.0087, + -0.0291, -0.0142], device='cuda:0'), grad: tensor([-2.0489e-08, 1.5832e-08, -2.8871e-08, 4.7497e-08, 3.7532e-07, + 1.0012e-08, 8.1491e-09, -8.9174e-08, 1.3737e-08, -3.1851e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 221.01, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4676 re_mapping 0.0035 re_causal 0.0097 /// teacc 99.15 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.2912, 0.2245, -0.0393, ..., -0.1160, 0.0172, 0.0136], + [ 0.0126, -0.1470, -0.0409, ..., 0.0402, -0.0595, -0.0760], + [-0.2018, -0.2884, 0.0581, ..., -0.0934, -0.0555, -0.1900], + ..., + [ 0.0740, 0.1670, -0.0972, ..., -0.1539, 0.0881, 0.0909], + [ 0.0329, -0.1914, -0.0466, ..., -0.1367, -0.1298, -0.1146], + [-0.2172, -0.0670, -0.0636, ..., 0.1069, -0.2977, 0.1075]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 9.5461e-09, + 2.3283e-10, 1.1642e-08], + [ 7.4506e-09, 3.7253e-09, 0.0000e+00, ..., 3.0268e-09, + 7.9162e-09, 1.3737e-08], + [ 2.7940e-09, 4.4238e-09, 0.0000e+00, ..., 2.3283e-10, + 5.1223e-09, 6.0536e-09], + ..., + [-2.2119e-08, -1.8626e-08, 0.0000e+00, ..., 1.3039e-08, + -1.9092e-08, -2.3982e-08], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 8.3819e-09, + 4.6566e-10, 2.0955e-09], + [ 4.3539e-08, 7.9162e-09, 0.0000e+00, ..., 4.7032e-08, + 6.7521e-09, -5.7044e-08]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0249, -0.0312, 0.0179, -0.0267, 0.0391, 0.0190, 0.0051, -0.0088, + -0.0292, -0.0139], device='cuda:0'), grad: tensor([ 4.2841e-08, 9.6159e-08, -3.5414e-07, 4.0513e-08, -2.0931e-07, + 1.5786e-07, -8.5682e-08, 1.4785e-07, 5.1456e-08, 1.3853e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 220.85, cls_loss 0.0006 cls_loss_mapping 0.0024 cls_loss_causal 0.4671 re_mapping 0.0033 re_causal 0.0096 /// teacc 99.16 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.2914, 0.2246, -0.0393, ..., -0.1161, 0.0172, 0.0136], + [ 0.0126, -0.1471, -0.0409, ..., 0.0402, -0.0597, -0.0762], + [-0.2021, -0.2886, 0.0582, ..., -0.0936, -0.0556, -0.1903], + ..., + [ 0.0739, 0.1671, -0.0972, ..., -0.1542, 0.0882, 0.0909], + [ 0.0331, -0.1915, -0.0467, ..., -0.1368, -0.1298, -0.1146], + [-0.2185, -0.0674, -0.0636, ..., 0.1068, -0.2981, 0.1076]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 1.6298e-09, 1.8626e-09], + [ 2.3283e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 1.6298e-08, 1.9325e-08], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 3.2596e-09, + 3.2363e-08, 4.3074e-08], + ..., + [-1.6298e-09, -4.6566e-10, 0.0000e+00, ..., 4.1910e-09, + 2.0955e-09, -4.1910e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 3.0268e-09, + 6.0536e-09, 1.0012e-08], + [ 2.7940e-09, 6.9849e-10, 0.0000e+00, ..., -1.5236e-06, + 5.3551e-09, -1.8883e-07]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0250, -0.0310, 0.0178, -0.0269, 0.0392, 0.0190, 0.0051, -0.0089, + -0.0292, -0.0140], device='cuda:0'), grad: tensor([ 1.3504e-08, 1.1385e-07, 1.7206e-07, -3.3202e-07, 3.1162e-06, + -6.9849e-10, 5.2154e-08, 2.4913e-08, -7.6368e-08, -3.0585e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 221.09, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4554 re_mapping 0.0035 re_causal 0.0100 /// teacc 99.15 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.2919, 0.2247, -0.0393, ..., -0.1163, 0.0172, 0.0135], + [ 0.0126, -0.1473, -0.0409, ..., 0.0402, -0.0601, -0.0763], + [-0.2030, -0.2892, 0.0582, ..., -0.0938, -0.0556, -0.1906], + ..., + [ 0.0740, 0.1676, -0.0972, ..., -0.1542, 0.0882, 0.0910], + [ 0.0332, -0.1916, -0.0467, ..., -0.1369, -0.1298, -0.1146], + [-0.2190, -0.0677, -0.0636, ..., 0.1068, -0.2984, 0.1076]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, -1.1874e-08, 0.0000e+00, ..., 1.7928e-08, + 6.9849e-10, 3.9581e-09], + [ 4.7032e-08, 9.3132e-10, 4.6566e-10, ..., 9.0105e-08, + 7.9209e-07, 1.0924e-06], + [ 1.1642e-08, 4.6566e-10, 4.6566e-10, ..., 2.0256e-08, + 1.6531e-08, 2.1420e-08], + ..., + [ 1.0803e-07, -1.6298e-09, 7.4506e-09, ..., 1.9767e-07, + -1.0803e-06, -1.4957e-06], + [ 1.9092e-08, 4.6566e-10, 2.3283e-10, ..., 5.5647e-08, + 6.9151e-08, 1.1432e-07], + [ 3.5483e-07, 3.4925e-09, 2.3283e-10, ..., 5.5647e-07, + 1.9092e-08, -3.3062e-08]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0251, -0.0312, 0.0179, -0.0267, 0.0392, 0.0190, 0.0050, -0.0088, + -0.0292, -0.0140], device='cuda:0'), grad: tensor([ 5.1688e-08, 4.0457e-06, 1.6950e-07, 8.4285e-07, -4.5411e-06, + 1.4435e-07, 1.1642e-07, -4.0531e-06, 5.7463e-07, 2.6431e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 220.84, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4953 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.12 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.2923, 0.2248, -0.0395, ..., -0.1163, 0.0170, 0.0127], + [ 0.0127, -0.1475, -0.0408, ..., 0.0402, -0.0609, -0.0768], + [-0.2043, -0.2896, 0.0577, ..., -0.0941, -0.0557, -0.1908], + ..., + [ 0.0739, 0.1680, -0.0976, ..., -0.1549, 0.0882, 0.0908], + [ 0.0331, -0.1916, -0.0472, ..., -0.1370, -0.1298, -0.1146], + [-0.2197, -0.0682, -0.0638, ..., 0.1069, -0.2986, 0.1083]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0943e-08, 0.0000e+00, ..., 1.7229e-08, + 2.3283e-10, -9.3132e-10], + [ 3.2596e-09, 2.5611e-09, 0.0000e+00, ..., 6.5193e-09, + 1.3039e-08, 3.5856e-08], + [ 9.3132e-10, -1.1642e-09, 0.0000e+00, ..., 4.8894e-09, + 8.1491e-09, 4.0280e-08], + ..., + [-8.6147e-09, -3.0268e-09, 0.0000e+00, ..., 2.6310e-08, + -3.8883e-08, -1.1898e-07], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 2.7241e-08, + 3.9581e-09, 5.8208e-09], + [ 3.4925e-09, 9.7789e-09, 0.0000e+00, ..., 8.3353e-08, + 1.2340e-08, 1.8161e-08]], device='cuda:0') +Epoch 310, bias, value: tensor([ 0.0248, -0.0313, 0.0178, -0.0269, 0.0392, 0.0191, 0.0050, -0.0090, + -0.0292, -0.0138], device='cuda:0'), grad: tensor([ 2.1211e-07, 1.8906e-06, -2.2519e-06, 1.2363e-07, -3.7136e-07, + 1.1874e-07, -1.5739e-07, -9.5926e-08, 1.0710e-07, 4.2818e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 220.92, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4910 re_mapping 0.0034 re_causal 0.0100 /// teacc 99.19 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.2927, 0.2249, -0.0395, ..., -0.1165, 0.0170, 0.0127], + [ 0.0128, -0.1476, -0.0407, ..., 0.0404, -0.0614, -0.0771], + [-0.2048, -0.2899, 0.0577, ..., -0.0947, -0.0558, -0.1913], + ..., + [ 0.0740, 0.1684, -0.0976, ..., -0.1551, 0.0883, 0.0910], + [ 0.0329, -0.1917, -0.0473, ..., -0.1370, -0.1298, -0.1147], + [-0.2201, -0.0688, -0.0639, ..., 0.1069, -0.2991, 0.1085]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.1910e-09, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, -2.3283e-10], + [-8.1491e-09, 6.9849e-10, 0.0000e+00, ..., -1.0501e-07, + 6.9849e-10, 1.6298e-09], + [ 3.2596e-09, 1.3970e-09, 0.0000e+00, ..., 9.2434e-08, + 0.0000e+00, 4.6566e-10], + ..., + [ 5.1223e-09, -4.6566e-10, 0.0000e+00, ..., 1.0710e-08, + -9.3132e-10, -1.8626e-09], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3271e-08, 1.8626e-09, 0.0000e+00, ..., 3.5856e-08, + 0.0000e+00, -1.8626e-09]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0248, -0.0313, 0.0177, -0.0270, 0.0392, 0.0191, 0.0051, -0.0089, + -0.0292, -0.0138], device='cuda:0'), grad: tensor([ 1.1642e-09, -9.6671e-07, 8.6240e-07, 7.2177e-09, -7.5437e-08, + 3.6554e-08, -2.9569e-08, 4.1910e-08, 9.0804e-09, 1.1618e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 220.56, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4830 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.20 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.2930, 0.2251, -0.0395, ..., -0.1166, 0.0171, 0.0128], + [ 0.0128, -0.1476, -0.0408, ..., 0.0404, -0.0618, -0.0774], + [-0.2053, -0.2904, 0.0576, ..., -0.0953, -0.0558, -0.1915], + ..., + [ 0.0740, 0.1684, -0.0975, ..., -0.1554, 0.0883, 0.0910], + [ 0.0327, -0.1921, -0.0472, ..., -0.1371, -0.1298, -0.1147], + [-0.2216, -0.0697, -0.0640, ..., 0.1066, -0.2992, 0.1087]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.3551e-09, 0.0000e+00, ..., 9.3132e-10, + 2.3283e-10, -2.3283e-10], + [ 1.6298e-09, 0.0000e+00, 0.0000e+00, ..., -6.7521e-09, + 4.2142e-08, 5.1223e-08], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 2.7474e-08, 3.0734e-08], + ..., + [-1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 6.7521e-09, + -9.0105e-08, -1.0850e-07], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.5611e-08, 3.1199e-08], + [ 1.6298e-09, 0.0000e+00, 0.0000e+00, ..., -4.8894e-09, + 3.0268e-09, -1.3970e-09]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0250, -0.0313, 0.0176, -0.0271, 0.0395, 0.0189, 0.0051, -0.0090, + -0.0292, -0.0141], device='cuda:0'), grad: tensor([-5.8208e-09, 1.6321e-07, 1.2014e-07, -3.3062e-08, 1.3504e-08, + 8.8476e-09, 6.7521e-09, -3.7369e-07, 1.1316e-07, 3.9581e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 220.77, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4880 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.17 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.2933, 0.2253, -0.0395, ..., -0.1167, 0.0172, 0.0129], + [ 0.0130, -0.1478, -0.0407, ..., 0.0404, -0.0635, -0.0777], + [-0.2059, -0.2909, 0.0576, ..., -0.0958, -0.0567, -0.1936], + ..., + [ 0.0739, 0.1685, -0.0976, ..., -0.1557, 0.0885, 0.0912], + [ 0.0326, -0.1922, -0.0473, ..., -0.1372, -0.1298, -0.1147], + [-0.2227, -0.0702, -0.0640, ..., 0.1066, -0.2994, 0.1089]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.0035e-08, ..., 8.8476e-09, + 2.3283e-10, 1.1642e-09], + [-7.4506e-09, 0.0000e+00, 3.9861e-07, ..., 8.7544e-08, + 2.5611e-09, 1.8161e-08], + [ 2.7940e-09, 0.0000e+00, 3.4226e-08, ..., 1.1409e-08, + 1.8626e-09, 1.6578e-07], + ..., + [ 5.8208e-09, 0.0000e+00, 1.3970e-09, ..., 1.0291e-07, + 3.2596e-09, -2.1770e-07], + [-6.7521e-09, 0.0000e+00, 1.5972e-07, ..., 4.2608e-08, + 7.6834e-09, 8.6147e-09], + [ 1.1642e-09, 4.6566e-10, 5.1223e-09, ..., -1.2317e-07, + 6.9849e-10, -1.5250e-07]], device='cuda:0') +Epoch 313, bias, value: tensor([ 0.0251, -0.0311, 0.0165, -0.0272, 0.0395, 0.0191, 0.0049, -0.0089, + -0.0292, -0.0141], device='cuda:0'), grad: tensor([ 1.6461e-07, 2.1234e-06, 8.8150e-07, -2.3376e-07, 7.9628e-07, + 1.5289e-05, -1.7539e-05, -8.6101e-07, -1.0873e-07, -5.0152e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 220.53, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5109 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.16 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.2943, 0.2254, -0.0396, ..., -0.1172, 0.0175, 0.0129], + [ 0.0130, -0.1480, -0.0407, ..., 0.0404, -0.0641, -0.0781], + [-0.2066, -0.2913, 0.0578, ..., -0.0963, -0.0568, -0.1940], + ..., + [ 0.0740, 0.1690, -0.0976, ..., -0.1565, 0.0886, 0.0911], + [ 0.0324, -0.1926, -0.0475, ..., -0.1376, -0.1298, -0.1147], + [-0.2250, -0.0713, -0.0641, ..., 0.1067, -0.3004, 0.1093]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, -1.2834e-06, 4.6566e-10, ..., 1.1642e-09, + 0.0000e+00, -4.9034e-07], + [-1.6065e-08, 1.1409e-08, -1.3271e-08, ..., -1.0012e-08, + 8.6147e-09, 1.8626e-08], + [ 3.9581e-09, 1.3970e-09, 2.3283e-09, ..., 3.2596e-09, + 4.6566e-10, 9.3132e-10], + ..., + [ 4.6566e-10, 2.7940e-09, 1.3970e-09, ..., 3.2596e-09, + -8.6147e-09, -1.3504e-08], + [ 9.7789e-09, 3.4925e-09, 6.9849e-09, ..., 3.4925e-09, + 1.1642e-09, 2.0955e-09], + [ 3.7253e-09, 5.5879e-08, 2.3283e-10, ..., 7.9162e-09, + 1.3970e-09, 2.1188e-08]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0251, -0.0312, 0.0167, -0.0275, 0.0396, 0.0185, 0.0056, -0.0091, + -0.0292, -0.0140], device='cuda:0'), grad: tensor([-2.7381e-06, -1.8859e-08, 9.6159e-08, 7.3574e-08, 2.8405e-08, + 1.8366e-06, 1.2694e-06, 3.5390e-08, -1.8943e-06, 1.3309e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 220.56, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4956 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.08 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.2945, 0.2255, -0.0396, ..., -0.1173, 0.0173, 0.0129], + [ 0.0128, -0.1482, -0.0407, ..., 0.0402, -0.0645, -0.0783], + [-0.2069, -0.2916, 0.0578, ..., -0.0966, -0.0568, -0.1941], + ..., + [ 0.0740, 0.1695, -0.0976, ..., -0.1567, 0.0887, 0.0912], + [ 0.0323, -0.1927, -0.0475, ..., -0.1377, -0.1298, -0.1147], + [-0.2261, -0.0722, -0.0641, ..., 0.1067, -0.3010, 0.1093]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 1.1642e-09], + [-4.8894e-09, 1.0477e-08, 0.0000e+00, ..., -2.1188e-08, + 3.4925e-09, 8.3819e-09], + [ 4.6566e-10, 2.0955e-09, 0.0000e+00, ..., 1.1642e-09, + 9.3132e-10, 1.6298e-09], + ..., + [ 4.6566e-10, -1.8161e-08, 0.0000e+00, ..., 1.7229e-08, + -6.9849e-09, -8.8476e-09], + [-2.3283e-10, -2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, -3.2131e-08], + [ 1.6298e-09, 6.7521e-09, 0.0000e+00, ..., -6.9849e-09, + 2.5611e-09, 1.9325e-08]], device='cuda:0') +Epoch 315, bias, value: tensor([ 0.0252, -0.0315, 0.0170, -0.0276, 0.0397, 0.0186, 0.0056, -0.0091, + -0.0291, -0.0141], device='cuda:0'), grad: tensor([ 6.0536e-09, -3.7486e-08, -3.8650e-08, 1.1362e-07, 3.2596e-09, + 1.0477e-08, 1.9791e-08, 4.9127e-08, -3.8324e-07, 2.6589e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 220.65, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4564 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.17 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.2950, 0.2256, -0.0399, ..., -0.1175, 0.0172, 0.0129], + [ 0.0128, -0.1487, -0.0407, ..., 0.0402, -0.0652, -0.0785], + [-0.2072, -0.2918, 0.0577, ..., -0.0968, -0.0568, -0.1943], + ..., + [ 0.0740, 0.1702, -0.0976, ..., -0.1569, 0.0881, 0.0908], + [ 0.0322, -0.1928, -0.0476, ..., -0.1378, -0.1299, -0.1148], + [-0.2273, -0.0726, -0.0642, ..., 0.1067, -0.3017, 0.1092]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.5134e-08, 0.0000e+00, ..., 1.6298e-09, + 6.9849e-10, 2.3283e-10], + [-9.3132e-09, 2.5611e-09, 0.0000e+00, ..., -1.2340e-08, + 5.3551e-09, 5.8208e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 1.8626e-09, 1.1642e-09], + ..., + [ 1.2806e-08, -2.5611e-09, 0.0000e+00, ..., 2.8638e-08, + 1.1409e-08, 8.8476e-09], + [ 2.3283e-10, 6.9849e-10, 0.0000e+00, ..., 6.9849e-10, + 2.1420e-08, 1.1874e-08], + [ 3.7253e-09, 4.6566e-09, 0.0000e+00, ..., -1.0501e-07, + 3.4925e-09, -5.0990e-08]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0251, -0.0316, 0.0175, -0.0264, 0.0397, 0.0186, 0.0056, -0.0095, + -0.0291, -0.0142], device='cuda:0'), grad: tensor([-1.6298e-08, -1.0408e-07, -1.8091e-07, -2.8703e-06, 1.4273e-07, + 2.7493e-06, 2.8405e-08, 3.6880e-07, 6.7288e-08, -1.5739e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 221.19, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4648 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.13 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.2960, 0.2257, -0.0399, ..., -0.1180, 0.0171, 0.0127], + [ 0.0116, -0.1516, -0.0411, ..., 0.0401, -0.0676, -0.0808], + [-0.2081, -0.2921, 0.0577, ..., -0.0976, -0.0568, -0.1956], + ..., + [ 0.0753, 0.1733, -0.0970, ..., -0.1572, 0.0883, 0.0924], + [ 0.0321, -0.1929, -0.0477, ..., -0.1384, -0.1299, -0.1148], + [-0.2290, -0.0736, -0.0641, ..., 0.1067, -0.3024, 0.1092]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -6.9477e-07, 0.0000e+00, ..., -1.0873e-07, + 0.0000e+00, -1.1642e-09], + [ 2.1164e-07, 3.3528e-08, 0.0000e+00, ..., 7.6834e-09, + 0.0000e+00, 6.9849e-10], + [ 1.8626e-09, 1.4715e-07, 0.0000e+00, ..., 3.2363e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.2806e-08, 2.8638e-08, 0.0000e+00, ..., 6.9849e-09, + 4.6566e-10, 1.8626e-09], + [ 1.8626e-09, 9.0571e-08, 0.0000e+00, ..., 1.8161e-08, + 0.0000e+00, 2.3283e-10], + [ 2.0955e-09, 2.3027e-07, 0.0000e+00, ..., 3.2363e-08, + 0.0000e+00, -5.8208e-09]], device='cuda:0') +Epoch 317, bias, value: tensor([ 0.0250, -0.0337, 0.0201, -0.0266, 0.0397, 0.0192, 0.0052, -0.0085, + -0.0293, -0.0142], device='cuda:0'), grad: tensor([-2.0452e-06, 6.3935e-07, 3.7020e-07, 3.4203e-07, 7.0315e-08, + -4.8243e-07, -5.7975e-08, 1.5576e-07, 2.9057e-07, 7.2736e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 220.97, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4756 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.18 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.2968, 0.2257, -0.0402, ..., -0.1185, 0.0169, 0.0127], + [ 0.0114, -0.1522, -0.0411, ..., 0.0400, -0.0679, -0.0813], + [-0.2085, -0.2922, 0.0577, ..., -0.0982, -0.0569, -0.1958], + ..., + [ 0.0755, 0.1739, -0.0971, ..., -0.1573, 0.0883, 0.0927], + [ 0.0320, -0.1931, -0.0478, ..., -0.1385, -0.1299, -0.1148], + [-0.2302, -0.0741, -0.0641, ..., 0.1068, -0.3027, 0.1093]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 3.9581e-09, + 2.3283e-10, 1.6298e-09], + [-8.6147e-09, 6.9849e-10, 0.0000e+00, ..., -7.6834e-09, + 2.3283e-09, 4.1910e-09], + [ 2.5611e-09, 6.9849e-10, 0.0000e+00, ..., 4.4238e-09, + 2.3283e-09, 3.2596e-09], + ..., + [ 6.2864e-09, -1.1642e-09, 0.0000e+00, ..., 4.8662e-08, + 5.9837e-08, 9.5461e-08], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 9.0804e-09, + 9.3132e-10, 5.1223e-09], + [ 2.9104e-08, 9.3132e-10, 0.0000e+00, ..., -9.0338e-08, + 1.1642e-09, -9.5228e-08]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0248, -0.0342, 0.0206, -0.0262, 0.0397, 0.0187, 0.0057, -0.0083, + -0.0293, -0.0142], device='cuda:0'), grad: tensor([ 1.9558e-08, 7.9162e-09, -3.7113e-07, -2.1327e-07, -5.4063e-07, + 2.7171e-07, 1.2107e-08, 3.8813e-07, 2.5146e-08, 4.0862e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 220.93, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4983 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.21 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.2974, 0.2259, -0.0402, ..., -0.1189, 0.0169, 0.0128], + [ 0.0115, -0.1522, -0.0411, ..., 0.0403, -0.0682, -0.0817], + [-0.2098, -0.2928, 0.0577, ..., -0.1001, -0.0569, -0.1984], + ..., + [ 0.0755, 0.1741, -0.0971, ..., -0.1575, 0.0883, 0.0932], + [ 0.0317, -0.1934, -0.0477, ..., -0.1387, -0.1299, -0.1149], + [-0.2315, -0.0746, -0.0642, ..., 0.1067, -0.3037, 0.1094]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.0477e-08, 0.0000e+00, ..., 2.0955e-09, + -2.3283e-10, -2.0955e-09], + [ 1.9092e-08, 3.0966e-08, 0.0000e+00, ..., 2.0489e-08, + 1.9092e-08, 2.9104e-08], + [ 6.9849e-09, 6.5193e-09, 0.0000e+00, ..., 6.2864e-09, + 7.6834e-09, 1.2806e-08], + ..., + [ 1.3039e-07, -4.1211e-08, 0.0000e+00, ..., 1.1059e-07, + -2.3982e-08, -5.1921e-08], + [ 1.6298e-09, 4.6566e-10, 0.0000e+00, ..., 1.6298e-09, + 2.5146e-08, 1.5134e-08], + [ 9.2201e-08, 6.0536e-09, 0.0000e+00, ..., 1.5041e-07, + 3.9581e-09, 1.2806e-08]], device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0248, -0.0341, 0.0192, -0.0242, 0.0397, 0.0169, 0.0056, -0.0081, + -0.0293, -0.0142], device='cuda:0'), grad: tensor([-7.9162e-09, 1.9162e-07, 8.4518e-08, -1.3667e-07, -1.2275e-06, + 4.6333e-08, 4.9826e-08, 3.7998e-07, 1.1409e-07, 5.1502e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 220.83, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4771 re_mapping 0.0033 re_causal 0.0094 /// teacc 99.13 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.2976, 0.2260, -0.0402, ..., -0.1192, 0.0174, 0.0128], + [ 0.0116, -0.1522, -0.0411, ..., 0.0403, -0.0684, -0.0818], + [-0.2101, -0.2936, 0.0577, ..., -0.1004, -0.0569, -0.1985], + ..., + [ 0.0754, 0.1740, -0.0971, ..., -0.1577, 0.0883, 0.0932], + [ 0.0318, -0.1936, -0.0476, ..., -0.1401, -0.1299, -0.1149], + [-0.2321, -0.0751, -0.0642, ..., 0.1068, -0.3040, 0.1097]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, -3.6322e-08, 0.0000e+00, ..., 8.1491e-10, + -1.1642e-10, -4.6566e-10], + [-5.3551e-09, 2.4447e-09, 0.0000e+00, ..., -5.9372e-09, + 1.3970e-09, 2.7940e-09], + [ 3.4925e-10, 2.5611e-09, 0.0000e+00, ..., 5.8208e-10, + -3.7253e-09, 4.6566e-10], + ..., + [ 3.2596e-09, -2.0955e-09, 0.0000e+00, ..., 5.7044e-09, + 2.3283e-09, -1.6298e-09], + [ 1.5134e-09, 9.3132e-10, 0.0000e+00, ..., 1.2806e-09, + 0.0000e+00, 8.1491e-10], + [ 1.7462e-09, 1.4319e-08, 0.0000e+00, ..., -9.1968e-09, + 8.1491e-10, -6.1700e-09]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0249, -0.0340, 0.0193, -0.0240, 0.0397, 0.0162, 0.0061, -0.0082, + -0.0294, -0.0142], device='cuda:0'), grad: tensor([-6.6007e-08, -1.0361e-08, -1.3656e-07, 1.9325e-08, 1.6880e-08, + -2.9104e-09, 6.7404e-08, 1.5285e-07, -2.9104e-08, 1.1176e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 220.69, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4987 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.21 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.2984, 0.2262, -0.0406, ..., -0.1196, 0.0173, 0.0123], + [ 0.0118, -0.1527, -0.0409, ..., 0.0407, -0.0670, -0.0819], + [-0.2111, -0.2944, 0.0577, ..., -0.1008, -0.0572, -0.1988], + ..., + [ 0.0753, 0.1746, -0.0972, ..., -0.1583, 0.0882, 0.0931], + [ 0.0318, -0.1938, -0.0476, ..., -0.1407, -0.1300, -0.1149], + [-0.2342, -0.0751, -0.0637, ..., 0.1068, -0.3044, 0.1101]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.1491e-10, + 1.1642e-10, 1.1642e-10], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 8.1491e-10, + 1.3970e-09, 2.7940e-09], + [ 3.7253e-09, 1.1642e-10, 0.0000e+00, ..., 5.8208e-10, + 2.5146e-08, 4.5402e-09], + ..., + [-1.1059e-08, -4.4238e-09, 0.0000e+00, ..., 1.7462e-09, + -5.8208e-09, -3.5157e-08], + [ 3.2596e-09, 2.3283e-10, 2.3283e-10, ..., 1.6298e-09, + 2.6776e-09, 5.2387e-09], + [ 4.7730e-09, 4.6566e-09, 0.0000e+00, ..., -4.3423e-08, + 5.9372e-09, 2.3283e-10]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0248, -0.0338, 0.0194, -0.0239, 0.0398, 0.0161, 0.0060, -0.0085, + -0.0293, -0.0142], device='cuda:0'), grad: tensor([ 4.5402e-09, 2.0955e-08, 1.8545e-07, -1.9313e-07, 1.0279e-07, + 1.2456e-08, -2.5611e-09, -1.1444e-07, 4.5053e-08, -4.0978e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 220.54, cls_loss 0.0004 cls_loss_mapping 0.0022 cls_loss_causal 0.4672 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.17 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.2986, 0.2262, -0.0406, ..., -0.1212, 0.0175, 0.0122], + [ 0.0118, -0.1527, -0.0409, ..., 0.0408, -0.0673, -0.0821], + [-0.2114, -0.2947, 0.0579, ..., -0.1011, -0.0568, -0.1989], + ..., + [ 0.0753, 0.1747, -0.0972, ..., -0.1587, 0.0882, 0.0931], + [ 0.0323, -0.1938, -0.0477, ..., -0.1408, -0.1299, -0.1149], + [-0.2347, -0.0755, -0.0637, ..., 0.1068, -0.3048, 0.1103]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.1642e-09, 1.1642e-10, ..., 1.4552e-08, + 1.1642e-10, 6.2864e-09], + [ 4.9011e-08, 0.0000e+00, 5.3551e-09, ..., 5.1339e-08, + 2.3283e-09, 3.2596e-09], + [ 4.3074e-09, 1.1642e-10, 2.3283e-10, ..., 8.4983e-09, + -1.0128e-08, 5.8208e-10], + ..., + [ 4.5635e-08, 0.0000e+00, 4.0745e-09, ..., 8.0210e-08, + 2.9104e-09, 1.2340e-08], + [ 1.0477e-09, 0.0000e+00, 1.1642e-10, ..., 3.0734e-08, + 3.9581e-09, 1.6531e-08], + [ 1.6415e-08, 0.0000e+00, 1.5134e-09, ..., -5.0431e-07, + 3.4925e-10, -2.9942e-07]], device='cuda:0') +Epoch 322, bias, value: tensor([ 0.0245, -0.0336, 0.0198, -0.0240, 0.0398, 0.0160, 0.0063, -0.0087, + -0.0292, -0.0141], device='cuda:0'), grad: tensor([ 4.8312e-08, 1.5332e-07, -1.5169e-07, 5.0152e-07, 8.6147e-09, + 5.2620e-07, 7.6834e-08, 2.7567e-07, 3.1781e-08, -1.4529e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 220.94, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4760 re_mapping 0.0033 re_causal 0.0092 /// teacc 99.19 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.2991, 0.2264, -0.0408, ..., -0.1215, 0.0181, 0.0123], + [ 0.0118, -0.1528, -0.0409, ..., 0.0408, -0.0677, -0.0822], + [-0.2125, -0.2956, 0.0580, ..., -0.1007, -0.0566, -0.1990], + ..., + [ 0.0753, 0.1748, -0.0972, ..., -0.1589, 0.0882, 0.0932], + [ 0.0326, -0.1940, -0.0479, ..., -0.1409, -0.1299, -0.1149], + [-0.2360, -0.0761, -0.0634, ..., 0.1068, -0.3051, 0.1104]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 1.6298e-09, + 5.8208e-10, 4.6566e-10], + [ 3.6089e-09, 5.1223e-09, 0.0000e+00, ..., 3.2596e-09, + 2.2119e-09, 1.9791e-09], + [ 8.1491e-10, 1.0477e-09, 0.0000e+00, ..., 2.7940e-09, + 3.4925e-10, 3.4925e-10], + ..., + [-2.2119e-09, -9.6625e-09, 0.0000e+00, ..., 1.1642e-08, + -1.7462e-09, 5.7044e-09], + [ 5.3551e-09, 1.1642e-10, 0.0000e+00, ..., 1.3970e-09, + 1.2456e-08, 9.3132e-10], + [ 3.6205e-08, 1.8626e-09, 0.0000e+00, ..., 5.6112e-08, + 1.6182e-08, -1.2573e-08]], device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0247, -0.0338, 0.0208, -0.0242, 0.0398, 0.0159, 0.0061, -0.0088, + -0.0289, -0.0141], device='cuda:0'), grad: tensor([ 1.1874e-08, 2.6426e-08, 2.2002e-08, 1.9316e-06, -1.2666e-07, + -2.0899e-06, -6.4028e-08, 2.1537e-08, 6.5309e-08, 2.2596e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 221.16, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4550 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.22 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.2993, 0.2265, -0.0408, ..., -0.1219, 0.0182, 0.0123], + [ 0.0118, -0.1529, -0.0412, ..., 0.0408, -0.0685, -0.0823], + [-0.2132, -0.2959, 0.0580, ..., -0.1009, -0.0566, -0.1991], + ..., + [ 0.0754, 0.1750, -0.0972, ..., -0.1591, 0.0882, 0.0932], + [ 0.0325, -0.1941, -0.0480, ..., -0.1412, -0.1299, -0.1150], + [-0.2368, -0.0765, -0.0634, ..., 0.1069, -0.3056, 0.1106]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, -2.0373e-08, 0.0000e+00, ..., 6.9849e-10, + -3.7253e-09, -3.4925e-10], + [-2.3283e-09, 6.9849e-10, 0.0000e+00, ..., -5.7044e-09, + 4.6566e-10, 8.1491e-10], + [ 4.6566e-10, 1.6298e-09, 0.0000e+00, ..., 1.5134e-09, + 5.8208e-10, 1.1642e-10], + ..., + [ 1.7462e-09, 5.8208e-10, 1.1642e-10, ..., 9.1968e-09, + 3.4925e-10, 3.3760e-09], + [-1.2806e-09, 5.8208e-10, 0.0000e+00, ..., 1.8277e-08, + 1.1642e-10, 1.0245e-08], + [ 3.4925e-10, 1.6298e-09, 0.0000e+00, ..., -3.0035e-08, + 3.4925e-10, -1.9907e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0246, -0.0339, 0.0210, -0.0243, 0.0397, 0.0158, 0.0064, -0.0089, + -0.0289, -0.0141], device='cuda:0'), grad: tensor([-4.4121e-08, 2.7940e-08, -8.6147e-09, 2.7823e-08, 3.1199e-08, + 2.8173e-08, 1.9791e-08, 3.9698e-08, -4.6217e-08, -6.0536e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 221.08, cls_loss 0.0003 cls_loss_mapping 0.0007 cls_loss_causal 0.4725 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.18 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.2996, 0.2265, -0.0408, ..., -0.1220, 0.0183, 0.0123], + [ 0.0118, -0.1529, -0.0412, ..., 0.0410, -0.0688, -0.0823], + [-0.2137, -0.2962, 0.0582, ..., -0.1022, -0.0560, -0.1992], + ..., + [ 0.0754, 0.1751, -0.0972, ..., -0.1593, 0.0882, 0.0933], + [ 0.0324, -0.1943, -0.0480, ..., -0.1414, -0.1299, -0.1150], + [-0.2374, -0.0770, -0.0635, ..., 0.1069, -0.3059, 0.1107]], + device='cuda:0'), grad: tensor([[ 5.8208e-10, -3.1316e-08, 0.0000e+00, ..., 1.7462e-09, + 0.0000e+00, -4.8894e-09], + [-9.7789e-09, 1.0477e-09, 0.0000e+00, ..., -4.4936e-08, + 1.9791e-09, 7.3342e-09], + [ 1.6298e-09, 6.9849e-10, 0.0000e+00, ..., 5.5879e-09, + 1.1642e-10, 3.4925e-10], + ..., + [-2.7707e-08, 1.0477e-09, 0.0000e+00, ..., 6.5193e-09, + -1.3039e-08, -5.8906e-08], + [ 1.6298e-09, 1.7462e-09, 0.0000e+00, ..., -1.2573e-08, + -3.4925e-09, 1.7462e-09], + [ 2.0722e-08, 1.1525e-08, 0.0000e+00, ..., 8.1491e-10, + 8.7311e-09, 3.9814e-08]], device='cuda:0') +Epoch 325, bias, value: tensor([ 0.0246, -0.0338, 0.0212, -0.0245, 0.0398, 0.0159, 0.0060, -0.0090, + -0.0288, -0.0141], device='cuda:0'), grad: tensor([-5.7276e-08, 2.6589e-07, -1.4696e-06, 5.4389e-07, 3.8301e-08, + 5.5879e-09, 2.4587e-07, 4.5821e-07, -1.6287e-07, 1.5832e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 221.24, cls_loss 0.0006 cls_loss_mapping 0.0025 cls_loss_causal 0.4854 re_mapping 0.0034 re_causal 0.0096 /// teacc 99.02 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.3002, 0.2268, -0.0407, ..., -0.1223, 0.0183, 0.0123], + [ 0.0124, -0.1528, -0.0411, ..., 0.0432, -0.0658, -0.0810], + [-0.2138, -0.2969, 0.0583, ..., -0.1037, -0.0554, -0.1994], + ..., + [ 0.0750, 0.1752, -0.0971, ..., -0.1614, 0.0870, 0.0928], + [ 0.0305, -0.1948, -0.0483, ..., -0.1419, -0.1301, -0.1154], + [-0.2398, -0.0779, -0.0636, ..., 0.1069, -0.3089, 0.1106]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, -1.1676e-07, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, -3.6089e-09], + [-2.6193e-08, -1.6298e-09, 0.0000e+00, ..., -2.0606e-08, + 3.4925e-10, 6.9849e-10], + [ 1.1642e-09, 2.0955e-09, 0.0000e+00, ..., 2.7940e-09, + 4.6566e-10, 4.6566e-10], + ..., + [ 1.7812e-08, 2.3283e-10, 0.0000e+00, ..., 1.0594e-08, + 2.5611e-09, 5.8208e-10], + [-5.8208e-10, 1.5134e-09, 0.0000e+00, ..., 1.9791e-09, + 0.0000e+00, 2.3283e-10], + [ 1.8626e-09, 9.8953e-09, 0.0000e+00, ..., 1.0477e-09, + 2.3283e-10, 6.9849e-10]], device='cuda:0') +Epoch 326, bias, value: tensor([ 0.0249, -0.0318, 0.0215, -0.0267, 0.0399, 0.0180, 0.0057, -0.0101, + -0.0291, -0.0142], device='cuda:0'), grad: tensor([-2.5611e-07, -1.4482e-07, 2.0140e-08, -1.5134e-09, 1.9907e-08, + 4.2142e-08, 2.3737e-07, 9.0804e-08, -2.7358e-08, 2.9104e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 220.92, cls_loss 0.0004 cls_loss_mapping 0.0020 cls_loss_causal 0.4696 re_mapping 0.0034 re_causal 0.0098 /// teacc 99.18 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.3006, 0.2270, -0.0408, ..., -0.1224, 0.0183, 0.0122], + [ 0.0125, -0.1529, -0.0411, ..., 0.0434, -0.0659, -0.0811], + [-0.2147, -0.2974, 0.0586, ..., -0.1046, -0.0557, -0.1996], + ..., + [ 0.0749, 0.1753, -0.0971, ..., -0.1616, 0.0870, 0.0928], + [ 0.0305, -0.1953, -0.0484, ..., -0.1420, -0.1301, -0.1154], + [-0.2420, -0.0786, -0.0636, ..., 0.1068, -0.3093, 0.1106]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.9791e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, -1.1642e-10], + [-4.7730e-09, 2.3283e-10, 0.0000e+00, ..., -3.7253e-09, + 3.4925e-10, 1.9791e-09], + [ 1.8626e-09, 1.1642e-10, 0.0000e+00, ..., 2.6776e-09, + 6.9849e-10, 4.6566e-10], + ..., + [ 1.0710e-07, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, + 4.3074e-09, 6.8685e-09], + [ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 1.7462e-09, + 3.4925e-10, 8.1491e-10], + [ 9.1968e-09, 2.6776e-09, 0.0000e+00, ..., -1.8743e-08, + 2.3283e-10, -2.4564e-08]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0251, -0.0319, 0.0219, -0.0268, 0.0401, 0.0182, 0.0053, -0.0102, + -0.0291, -0.0144], device='cuda:0'), grad: tensor([ 1.7462e-09, -3.2014e-08, -1.0245e-08, -2.9802e-08, -1.8755e-07, + 2.5029e-08, -1.3970e-08, 2.7055e-07, 1.8976e-08, -4.8196e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 220.77, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4988 re_mapping 0.0033 re_causal 0.0099 /// teacc 99.08 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.3010, 0.2271, -0.0416, ..., -0.1228, 0.0183, 0.0122], + [ 0.0125, -0.1532, -0.0407, ..., 0.0434, -0.0660, -0.0813], + [-0.2174, -0.2977, 0.0575, ..., -0.1052, -0.0560, -0.1998], + ..., + [ 0.0750, 0.1756, -0.0971, ..., -0.1618, 0.0870, 0.0929], + [ 0.0310, -0.1956, -0.0499, ..., -0.1421, -0.1301, -0.1154], + [-0.2436, -0.0791, -0.0636, ..., 0.1046, -0.3095, 0.1114]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.6531e-08, 0.0000e+00, ..., 3.0268e-09, + -2.3283e-10, 2.3283e-10], + [ 1.1874e-08, 2.0023e-08, 0.0000e+00, ..., 7.2177e-09, + 4.1910e-09, 7.2177e-09], + [ 1.1642e-08, 1.0710e-08, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 6.0536e-09], + ..., + [-2.0955e-08, -2.9569e-08, 0.0000e+00, ..., 2.0489e-08, + -2.0955e-09, 3.0268e-09], + [ 4.6566e-10, 1.1642e-09, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, -3.2596e-08], + [ 6.5193e-09, 5.3551e-09, 0.0000e+00, ..., -4.2608e-08, + 6.9849e-10, -1.8394e-08]], device='cuda:0') +Epoch 328, bias, value: tensor([ 0.0251, -0.0318, 0.0213, -0.0267, 0.0423, 0.0180, 0.0055, -0.0102, + -0.0290, -0.0166], device='cuda:0'), grad: tensor([ 1.4924e-07, 1.4179e-07, 1.6834e-07, 8.1211e-07, 5.5181e-08, + 9.8906e-07, 4.7777e-07, 1.5181e-07, -3.6433e-06, 7.2364e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 221.09, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4507 re_mapping 0.0033 re_causal 0.0090 /// teacc 99.12 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.3012, 0.2281, -0.0416, ..., -0.1233, 0.0183, 0.0143], + [ 0.0124, -0.1545, -0.0407, ..., 0.0433, -0.0663, -0.0820], + [-0.2178, -0.2983, 0.0575, ..., -0.1057, -0.0562, -0.2001], + ..., + [ 0.0751, 0.1770, -0.0971, ..., -0.1620, 0.0873, 0.0934], + [ 0.0308, -0.1964, -0.0499, ..., -0.1430, -0.1302, -0.1155], + [-0.2445, -0.0823, -0.0636, ..., 0.1042, -0.3103, 0.1111]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -2.7940e-08, 0.0000e+00, ..., 2.7940e-09, + -9.3132e-10, -9.3132e-10], + [-2.7241e-08, 7.2177e-09, -1.0245e-08, ..., -5.2620e-08, + -4.1910e-09, 1.6298e-09], + [ 1.6298e-09, 1.6298e-09, 6.9849e-10, ..., 3.7253e-09, + 4.6566e-10, 2.3283e-10], + ..., + [ 2.6776e-08, 2.3283e-10, 8.8476e-09, ..., 5.1688e-08, + 4.8894e-09, 2.0955e-09], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 4.6566e-10, + -1.3970e-09, 2.3283e-10], + [ 9.3132e-10, 3.0268e-09, 0.0000e+00, ..., -4.2142e-08, + 4.6566e-10, -4.4936e-08]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0263, -0.0324, 0.0214, -0.0270, 0.0427, 0.0181, 0.0056, -0.0098, + -0.0291, -0.0171], device='cuda:0'), grad: tensor([ 2.7474e-08, -4.0652e-07, 4.0047e-08, 7.0781e-08, 5.6112e-08, + 1.2666e-07, 1.1548e-07, 1.4417e-06, -1.5395e-06, 7.2410e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 220.66, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4724 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.12 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.3016, 0.2282, -0.0416, ..., -0.1238, 0.0183, 0.0143], + [ 0.0131, -0.1535, -0.0407, ..., 0.0435, -0.0662, -0.0820], + [-0.2183, -0.2988, 0.0575, ..., -0.1059, -0.0565, -0.2006], + ..., + [ 0.0746, 0.1763, -0.0971, ..., -0.1627, 0.0874, 0.0933], + [ 0.0305, -0.1966, -0.0499, ..., -0.1438, -0.1302, -0.1155], + [-0.2455, -0.0826, -0.0636, ..., 0.1042, -0.3116, 0.1118]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 0.0000e+00, 0.0000e+00], + [-2.2585e-08, 2.3283e-10, -9.3132e-09, ..., -5.5414e-08, + 0.0000e+00, 0.0000e+00], + [ 6.2864e-09, 2.3283e-10, 2.3283e-09, ..., 1.6997e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3504e-08, 0.0000e+00, 5.8208e-09, ..., 3.2596e-08, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 0.0000e+00, 2.3283e-10, ..., 3.0268e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, -9.3132e-10]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0262, -0.0312, 0.0213, -0.0272, 0.0427, 0.0181, 0.0057, -0.0107, + -0.0290, -0.0170], device='cuda:0'), grad: tensor([ 2.4913e-08, -6.0536e-07, 1.0733e-07, 6.0536e-09, 3.5507e-07, + 2.5844e-08, -3.4273e-07, 4.1677e-07, 1.7229e-08, 5.3551e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 221.24, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4627 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.16 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.3019, 0.2285, -0.0416, ..., -0.1242, 0.0183, 0.0143], + [ 0.0130, -0.1548, -0.0407, ..., 0.0436, -0.0669, -0.0828], + [-0.2202, -0.3026, 0.0574, ..., -0.1066, -0.0561, -0.2025], + ..., + [ 0.0749, 0.1805, -0.0971, ..., -0.1630, 0.0876, 0.0945], + [ 0.0301, -0.1975, -0.0500, ..., -0.1460, -0.1302, -0.1159], + [-0.2464, -0.0840, -0.0636, ..., 0.1043, -0.3150, 0.1118]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.3283e-09, 0.0000e+00, ..., 6.9849e-09, + 6.9849e-10, 5.1223e-09], + [ 4.6566e-10, 4.1910e-09, 0.0000e+00, ..., 1.6298e-09, + 2.5611e-09, 6.7521e-09], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 4.6566e-10, + 2.5611e-09, 1.8626e-09], + ..., + [ 9.3132e-10, -5.1223e-09, 0.0000e+00, ..., 2.9337e-08, + 6.9849e-10, 1.5134e-08], + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 3.2596e-09, + 2.3283e-10, 2.3283e-09], + [ 1.6298e-09, 2.7940e-09, 0.0000e+00, ..., -2.2678e-07, + -2.4680e-08, -1.7742e-07]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0263, -0.0318, 0.0197, -0.0270, 0.0427, 0.0181, 0.0058, -0.0100, + -0.0291, -0.0170], device='cuda:0'), grad: tensor([ 1.3039e-08, 1.1874e-08, 1.0710e-08, -1.1642e-09, 3.7369e-07, + -2.1653e-08, 2.4913e-08, 5.2154e-08, 3.9581e-09, -4.4773e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 220.85, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4529 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.13 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.3022, 0.2287, -0.0416, ..., -0.1247, 0.0182, 0.0143], + [ 0.0130, -0.1557, -0.0406, ..., 0.0435, -0.0674, -0.0833], + [-0.2205, -0.3027, 0.0574, ..., -0.1068, -0.0560, -0.2030], + ..., + [ 0.0750, 0.1813, -0.0971, ..., -0.1636, 0.0877, 0.0947], + [ 0.0299, -0.1982, -0.0500, ..., -0.1463, -0.1302, -0.1160], + [-0.2473, -0.0845, -0.0637, ..., 0.1043, -0.3155, 0.1121]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, -9.8720e-08, 0.0000e+00, ..., 7.5670e-08, + 0.0000e+00, -6.7521e-09], + [ 2.0489e-08, 1.2107e-08, 0.0000e+00, ..., 2.8638e-08, + 0.0000e+00, 1.2340e-08], + [ 3.9581e-09, 3.1432e-08, 0.0000e+00, ..., 5.3551e-09, + 0.0000e+00, 3.0268e-09], + ..., + [ 4.3306e-08, -9.7090e-08, 0.0000e+00, ..., 1.5832e-08, + 0.0000e+00, -9.3132e-08], + [-9.8720e-08, 7.9162e-09, 0.0000e+00, ..., 8.1491e-09, + 0.0000e+00, 4.6566e-10], + [ 3.5623e-08, 1.1269e-07, 0.0000e+00, ..., 9.9186e-07, + 0.0000e+00, 7.1945e-08]], device='cuda:0') +Epoch 332, bias, value: tensor([ 0.0264, -0.0321, 0.0203, -0.0269, 0.0427, 0.0179, 0.0057, -0.0099, + -0.0292, -0.0170], device='cuda:0'), grad: tensor([-9.1502e-08, 1.9441e-07, 1.5390e-07, 3.9116e-08, -1.9129e-06, + 1.6671e-07, -3.7486e-08, 7.4273e-08, -7.4785e-07, 2.1681e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 221.01, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4881 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.18 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.3029, 0.2290, -0.0417, ..., -0.1249, 0.0188, 0.0144], + [ 0.0129, -0.1558, -0.0404, ..., 0.0437, -0.0675, -0.0834], + [-0.2209, -0.3031, 0.0567, ..., -0.1090, -0.0561, -0.2032], + ..., + [ 0.0750, 0.1814, -0.0971, ..., -0.1637, 0.0878, 0.0950], + [ 0.0296, -0.1984, -0.0500, ..., -0.1446, -0.1302, -0.1160], + [-0.2484, -0.0847, -0.0637, ..., 0.1043, -0.3160, 0.1120]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.9104e-08, + 0.0000e+00, 0.0000e+00], + [-1.1642e-09, 1.1176e-08, 0.0000e+00, ..., -1.2573e-08, + 2.3283e-10, 1.4668e-08], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 7.2177e-09, + 2.3283e-10, 4.6566e-10], + ..., + [-1.7695e-08, -2.2352e-08, 0.0000e+00, ..., -6.9849e-10, + 2.3283e-10, -3.1432e-08], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 1.0943e-08, + 0.0000e+00, 4.6566e-10], + [ 2.2817e-08, 1.0710e-08, 0.0000e+00, ..., 5.3318e-08, + 2.3283e-10, 1.4203e-08]], device='cuda:0') +Epoch 333, bias, value: tensor([ 0.0266, -0.0320, 0.0204, -0.0268, 0.0427, 0.0178, 0.0048, -0.0100, + -0.0287, -0.0170], device='cuda:0'), grad: tensor([ 1.1479e-07, 3.6042e-07, -2.5984e-07, 1.2806e-08, 9.9419e-08, + 9.7090e-08, -7.0967e-07, -7.4506e-08, 1.9744e-07, 1.6694e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 221.08, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4751 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.21 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.3034, 0.2290, -0.0417, ..., -0.1253, 0.0188, 0.0143], + [ 0.0129, -0.1567, -0.0404, ..., 0.0437, -0.0676, -0.0837], + [-0.2213, -0.3031, 0.0567, ..., -0.1092, -0.0561, -0.2034], + ..., + [ 0.0751, 0.1821, -0.0971, ..., -0.1638, 0.0879, 0.0953], + [ 0.0289, -0.1986, -0.0501, ..., -0.1448, -0.1303, -0.1161], + [-0.2493, -0.0848, -0.0637, ..., 0.1043, -0.3163, 0.1121]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -6.2864e-09, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 6.9849e-10], + [ 3.8184e-08, 1.9791e-08, -1.1642e-09, ..., -2.5611e-09, + 2.0722e-08, 2.5146e-08], + [ 4.6566e-09, 1.3970e-09, 2.3283e-10, ..., 9.3132e-10, + 1.1642e-09, 1.6298e-09], + ..., + [-5.3085e-08, -2.4680e-08, 4.6566e-10, ..., 4.6566e-09, + -2.3283e-08, -2.8638e-08], + [ 3.0268e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + -1.6298e-09, 1.1642e-09], + [ 1.6298e-09, 1.1642e-09, 0.0000e+00, ..., -4.0047e-08, + 6.9849e-10, -1.8626e-08]], device='cuda:0') +Epoch 334, bias, value: tensor([ 0.0265, -0.0325, 0.0205, -0.0269, 0.0427, 0.0178, 0.0049, -0.0096, + -0.0288, -0.0170], device='cuda:0'), grad: tensor([-2.5611e-09, 1.3271e-07, -1.0710e-08, 2.0489e-08, 8.0094e-08, + -3.7253e-08, 4.0513e-08, -1.3853e-07, -2.7940e-09, -6.9616e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 221.08, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4580 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.18 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.3044, 0.2291, -0.0419, ..., -0.1259, 0.0189, 0.0143], + [ 0.0131, -0.1568, -0.0404, ..., 0.0440, -0.0676, -0.0836], + [-0.2220, -0.3032, 0.0567, ..., -0.1098, -0.0556, -0.2037], + ..., + [ 0.0749, 0.1822, -0.0971, ..., -0.1643, 0.0878, 0.0953], + [ 0.0290, -0.1988, -0.0503, ..., -0.1449, -0.1303, -0.1161], + [-0.2537, -0.0850, -0.0637, ..., 0.1042, -0.3164, 0.1122]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.0268e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, -2.3283e-10], + [-1.5600e-08, 2.3283e-10, 2.3283e-10, ..., -3.2596e-08, + 0.0000e+00, 2.3283e-10], + [ 3.2596e-09, 0.0000e+00, -6.9849e-10, ..., 6.7521e-09, + 6.9849e-10, 2.3283e-10], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 2.3283e-10, 4.6566e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 9.3132e-10, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, -1.3970e-09]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0265, -0.0321, 0.0202, -0.0265, 0.0428, 0.0174, 0.0048, -0.0098, + -0.0288, -0.0171], device='cuda:0'), grad: tensor([ 2.3283e-10, -1.0803e-07, -2.1933e-07, 9.0804e-09, 1.0012e-08, + 2.3283e-09, 8.2888e-08, 1.4133e-07, 1.0128e-07, -2.0955e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 221.22, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4589 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.14 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.3049, 0.2292, -0.0422, ..., -0.1265, 0.0187, 0.0143], + [ 0.0131, -0.1570, -0.0404, ..., 0.0439, -0.0677, -0.0838], + [-0.2224, -0.3032, 0.0569, ..., -0.1100, -0.0554, -0.2041], + ..., + [ 0.0749, 0.1824, -0.0972, ..., -0.1675, 0.0877, 0.0923], + [ 0.0289, -0.1990, -0.0505, ..., -0.1451, -0.1303, -0.1161], + [-0.2544, -0.0851, -0.0637, ..., 0.1052, -0.3166, 0.1167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.5078e-06, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -7.0594e-07], + [ 4.6566e-10, 4.6566e-09, 0.0000e+00, ..., -1.3970e-09, + 1.8626e-09, 5.1223e-09], + [ 6.9849e-10, 2.0955e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-09, 2.7940e-09], + ..., + [-1.8626e-09, -1.1176e-08, 0.0000e+00, ..., 9.3132e-10, + -3.2596e-09, -1.5832e-08], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 2.7940e-09], + [ 1.3970e-09, 1.4827e-06, 0.0000e+00, ..., 1.3970e-09, + 9.5461e-09, 7.0967e-07]], device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0264, -0.0291, 0.0177, -0.0264, 0.0425, 0.0174, 0.0049, -0.0126, + -0.0288, -0.0160], device='cuda:0'), grad: tensor([-3.8780e-06, 1.4668e-08, 1.3271e-08, -2.6580e-06, 1.5600e-08, + 2.6207e-06, 3.3062e-08, -3.9116e-08, 1.6764e-08, 3.8594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 220.61, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4905 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.16 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.3063, 0.2294, -0.0425, ..., -0.1269, 0.0182, 0.0143], + [ 0.0130, -0.1576, -0.0405, ..., 0.0437, -0.0678, -0.0852], + [-0.2232, -0.3033, 0.0569, ..., -0.1102, -0.0557, -0.2045], + ..., + [ 0.0750, 0.1832, -0.0972, ..., -0.1676, 0.0878, 0.0927], + [ 0.0292, -0.1993, -0.0507, ..., -0.1453, -0.1303, -0.1161], + [-0.2573, -0.0857, -0.0638, ..., 0.1053, -0.3168, 0.1168]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.6368e-08, 0.0000e+00, ..., 2.8871e-08, + 4.6566e-10, 7.2177e-09], + [ 3.0268e-09, 5.3551e-09, 0.0000e+00, ..., 4.2841e-08, + 2.5611e-09, 4.4238e-09], + [ 4.6566e-10, 1.2340e-08, 0.0000e+00, ..., 1.0245e-08, + -9.3132e-10, 9.3132e-10], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 3.9581e-09, + 9.3132e-10, 4.6566e-10], + [ 2.3283e-10, 3.2596e-09, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 6.9849e-10], + [ 5.8208e-09, 1.8394e-08, 0.0000e+00, ..., -1.2782e-07, + 2.3283e-10, -5.8906e-08]], device='cuda:0') +Epoch 337, bias, value: tensor([ 0.0265, -0.0292, 0.0177, -0.0261, 0.0424, 0.0175, 0.0043, -0.0123, + -0.0287, -0.0160], device='cuda:0'), grad: tensor([-1.0035e-07, 2.1770e-07, 4.4936e-08, 8.4518e-08, 1.5134e-07, + 1.9465e-07, -4.2212e-07, 1.7928e-08, 1.2806e-08, -1.8883e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 221.00, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4661 re_mapping 0.0033 re_causal 0.0092 /// teacc 99.18 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.3085, 0.2297, -0.0426, ..., -0.1274, 0.0180, 0.0143], + [ 0.0109, -0.1605, -0.0405, ..., 0.0432, -0.0690, -0.0866], + [-0.2250, -0.3043, 0.0569, ..., -0.1109, -0.0586, -0.2066], + ..., + [ 0.0769, 0.1877, -0.0971, ..., -0.1676, 0.0897, 0.0933], + [ 0.0328, -0.2016, -0.0507, ..., -0.1455, -0.1304, -0.1162], + [-0.2582, -0.0861, -0.0638, ..., 0.1054, -0.3168, 0.1172]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 3.1898e-08, + 2.3283e-10, 1.5367e-08], + [ 1.3970e-09, 2.3283e-09, 0.0000e+00, ..., 3.7090e-07, + 3.4925e-09, 1.6182e-07], + [ 2.8173e-08, 2.2817e-08, 0.0000e+00, ..., 7.9162e-09, + 2.8173e-08, 3.2363e-08], + ..., + [-1.7928e-08, -2.5844e-08, 0.0000e+00, ..., 3.3993e-08, + -2.0955e-08, -2.7940e-08], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-09, 1.1642e-09], + [ 4.5169e-08, -1.3970e-09, 0.0000e+00, ..., -1.3541e-06, + 2.3283e-10, -6.5286e-07]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0267, -0.0295, 0.0176, -0.0275, 0.0424, 0.0180, 0.0039, -0.0109, + -0.0278, -0.0159], device='cuda:0'), grad: tensor([ 1.0687e-07, 1.1316e-06, -4.4773e-07, -4.7497e-08, 1.7751e-06, + 9.7789e-08, -2.8638e-08, 5.6345e-08, 3.3295e-08, -2.6636e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 220.92, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4732 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.04 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.3101, 0.2298, -0.0430, ..., -0.1283, 0.0179, 0.0143], + [ 0.0106, -0.1614, -0.0405, ..., 0.0436, -0.0692, -0.0872], + [-0.2276, -0.3048, 0.0569, ..., -0.1118, -0.0592, -0.2068], + ..., + [ 0.0777, 0.1887, -0.0971, ..., -0.1675, 0.0899, 0.0935], + [ 0.0341, -0.2018, -0.0514, ..., -0.1449, -0.1304, -0.1160], + [-0.2597, -0.0863, -0.0639, ..., 0.1054, -0.3170, 0.1172]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + 0.0000e+00, 1.3970e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 5.1223e-09, + -2.3283e-10, 2.0955e-09], + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 2.0955e-09], + [ 1.6298e-09, 2.3283e-10, 0.0000e+00, ..., -3.5157e-08, + 0.0000e+00, -2.3749e-08]], device='cuda:0') +Epoch 339, bias, value: tensor([ 0.0265, -0.0296, 0.0176, -0.0277, 0.0423, 0.0179, 0.0043, -0.0105, + -0.0275, -0.0159], device='cuda:0'), grad: tensor([ 2.9756e-07, 9.3132e-09, 9.3132e-10, 1.7858e-07, 1.4435e-08, + -7.4226e-07, 2.8778e-07, 8.3819e-09, 1.8626e-08, -7.1013e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 220.71, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4921 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.19 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.3115, 0.2298, -0.0431, ..., -0.1298, 0.0170, 0.0138], + [ 0.0107, -0.1614, -0.0404, ..., 0.0438, -0.0692, -0.0873], + [-0.2290, -0.3048, 0.0570, ..., -0.1139, -0.0594, -0.2073], + ..., + [ 0.0777, 0.1888, -0.0972, ..., -0.1676, 0.0899, 0.0936], + [ 0.0342, -0.2019, -0.0516, ..., -0.1454, -0.1304, -0.1161], + [-0.2606, -0.0863, -0.0639, ..., 0.1054, -0.3172, 0.1174]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 2.0955e-09, + 4.6566e-10, 4.1910e-09], + [ 2.3283e-09, 1.4901e-08, 0.0000e+00, ..., 2.3283e-09, + 2.3283e-10, 1.6065e-08], + [ 2.3283e-10, 9.3132e-10, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 1.1642e-09], + ..., + [-1.5600e-08, -1.2014e-07, 0.0000e+00, ..., -1.3271e-08, + 4.6566e-10, -1.2596e-07], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.6298e-09, + 2.3283e-10, 9.3132e-10], + [ 1.8859e-08, 8.5682e-08, 0.0000e+00, ..., 2.6543e-08, + 1.3970e-09, 8.9407e-08]], device='cuda:0') +Epoch 340, bias, value: tensor([ 0.0259, -0.0295, 0.0176, -0.0279, 0.0423, 0.0183, 0.0043, -0.0105, + -0.0275, -0.0159], device='cuda:0'), grad: tensor([ 1.5600e-08, 4.7497e-08, 7.4506e-09, 3.7486e-07, -2.6077e-08, + 2.5867e-07, -6.3330e-07, -3.4040e-07, 1.1409e-08, 2.9104e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 221.03, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4896 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.07 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.3122, 0.2298, -0.0433, ..., -0.1302, 0.0164, 0.0137], + [ 0.0132, -0.1615, -0.0403, ..., 0.0451, -0.0694, -0.0876], + [-0.2312, -0.3049, 0.0566, ..., -0.1146, -0.0596, -0.2077], + ..., + [ 0.0776, 0.1890, -0.0972, ..., -0.1677, 0.0900, 0.0936], + [ 0.0311, -0.2021, -0.0520, ..., -0.1481, -0.1304, -0.1162], + [-0.2614, -0.0866, -0.0643, ..., 0.1054, -0.3175, 0.1175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8161e-08, + 9.3132e-10, 2.3283e-10], + [ 2.3283e-09, 0.0000e+00, 2.3283e-10, ..., 5.1223e-09, + 4.1910e-09, 1.3970e-09], + [ 4.6566e-10, 0.0000e+00, -8.1491e-09, ..., 1.7462e-08, + 4.1211e-08, 1.9558e-08], + ..., + [ 3.4925e-09, 0.0000e+00, 2.3283e-10, ..., 3.2596e-09, + 6.2864e-09, 3.7253e-09], + [-7.6834e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-09, + 3.0268e-09, 1.1642e-09], + [ 6.9849e-10, 2.3283e-10, 2.3283e-10, ..., -1.1642e-09, + 2.5611e-09, -9.3132e-10]], device='cuda:0') +Epoch 341, bias, value: tensor([ 0.0259, -0.0289, 0.0176, -0.0281, 0.0423, 0.0182, 0.0046, -0.0105, + -0.0303, -0.0159], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.0687e-07, 1.5437e-07, 1.0014e-05, 6.0769e-08, + -9.1195e-06, -1.2550e-07, 2.8079e-07, -1.5553e-06, 1.1153e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 220.30, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4834 re_mapping 0.0033 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.3125, 0.2299, -0.0453, ..., -0.1315, 0.0146, 0.0137], + [ 0.0132, -0.1616, -0.0403, ..., 0.0441, -0.0695, -0.0887], + [-0.2317, -0.3049, 0.0567, ..., -0.1151, -0.0596, -0.2083], + ..., + [ 0.0776, 0.1894, -0.0972, ..., -0.1672, 0.0901, 0.0947], + [ 0.0311, -0.2023, -0.0521, ..., -0.1482, -0.1304, -0.1161], + [-0.2620, -0.0878, -0.0645, ..., 0.1054, -0.3188, 0.1169]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -9.3132e-10, 4.6566e-10, ..., 7.6834e-09, + 2.3283e-10, 6.9849e-10], + [ 1.4179e-07, 4.2375e-08, -6.9849e-10, ..., 2.1816e-07, + 4.0047e-08, 2.9197e-07], + [ 3.0268e-09, -4.6566e-09, 0.0000e+00, ..., 5.1223e-09, + 5.5879e-09, 1.1409e-08], + ..., + [-1.8440e-07, -5.8906e-08, 2.3283e-10, ..., -2.6496e-07, + -3.7486e-08, -3.7323e-07], + [ 2.3283e-10, 1.8626e-09, 4.6566e-10, ..., 4.8894e-09, + 9.3132e-09, 1.0245e-08], + [ 4.7497e-08, 1.8161e-08, 0.0000e+00, ..., 2.6007e-07, + 1.0245e-08, 6.3330e-08]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0256, -0.0290, 0.0176, -0.0284, 0.0423, 0.0185, 0.0047, -0.0103, + -0.0300, -0.0159], device='cuda:0'), grad: tensor([ 7.6834e-08, 1.0896e-06, -3.4808e-07, -1.6135e-07, -5.1223e-07, + 8.8476e-08, -9.3132e-09, -1.2647e-06, 1.3993e-07, 9.1083e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 221.15, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4628 re_mapping 0.0030 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.3127, 0.2302, -0.0455, ..., -0.1318, 0.0144, 0.0136], + [ 0.0132, -0.1617, -0.0403, ..., 0.0441, -0.0696, -0.0889], + [-0.2330, -0.3051, 0.0567, ..., -0.1156, -0.0600, -0.2097], + ..., + [ 0.0777, 0.1896, -0.0969, ..., -0.1670, 0.0900, 0.0951], + [ 0.0311, -0.2027, -0.0522, ..., -0.1484, -0.1305, -0.1162], + [-0.2634, -0.0885, -0.0645, ..., 0.1053, -0.3193, 0.1166]], + device='cuda:0'), grad: tensor([[ 3.0268e-09, -2.1677e-07, -1.3970e-09, ..., -3.6554e-08, + 2.3283e-10, -6.5193e-08], + [ 3.7253e-09, 1.1642e-08, 0.0000e+00, ..., 1.3434e-07, + 0.0000e+00, 7.2177e-09], + [ 2.3283e-09, 3.2596e-09, 2.3283e-10, ..., 2.1653e-08, + 0.0000e+00, 6.9849e-10], + ..., + [ 6.9849e-10, 1.8161e-08, 0.0000e+00, ..., 1.6531e-08, + 0.0000e+00, 1.0943e-08], + [-3.4925e-09, 7.9162e-09, 0.0000e+00, ..., 1.0710e-08, + 0.0000e+00, 3.7253e-09], + [ 3.7253e-09, 7.5903e-08, 6.9849e-10, ..., -7.6136e-08, + 6.9849e-10, -5.0990e-08]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0258, -0.0290, 0.0176, -0.0269, 0.0423, 0.0172, 0.0047, -0.0101, + -0.0300, -0.0160], device='cuda:0'), grad: tensor([-5.6159e-07, 7.4971e-07, 1.3225e-07, 2.2678e-07, 1.7602e-07, + 2.7940e-09, -7.8790e-07, 8.7311e-08, 2.6077e-08, -3.0268e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 220.73, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4502 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.21 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.3133, 0.2305, -0.0455, ..., -0.1324, 0.0139, 0.0136], + [ 0.0132, -0.1616, -0.0403, ..., 0.0443, -0.0697, -0.0891], + [-0.2337, -0.3051, 0.0565, ..., -0.1160, -0.0601, -0.2110], + ..., + [ 0.0778, 0.1896, -0.0968, ..., -0.1672, 0.0899, 0.0952], + [ 0.0311, -0.2039, -0.0526, ..., -0.1485, -0.1305, -0.1163], + [-0.2638, -0.0889, -0.0646, ..., 0.1054, -0.3196, 0.1168]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.7521e-09, 0.0000e+00, ..., 2.5844e-08, + 6.9849e-10, 1.0477e-08], + [ 2.3283e-10, 9.3132e-10, 0.0000e+00, ..., 2.2817e-08, + 3.0268e-09, 1.1874e-08], + [ 4.6566e-10, 1.1642e-09, 0.0000e+00, ..., 1.1642e-09, + 1.3970e-08, 1.1642e-08], + ..., + [-4.6566e-10, -1.6298e-09, 0.0000e+00, ..., 1.1525e-07, + 4.1910e-09, 4.9593e-08], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.1653e-08, + 1.3970e-09, 9.7789e-09], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., -8.1072e-07, + 1.6298e-09, -3.1688e-07]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0258, -0.0290, 0.0176, -0.0273, 0.0423, 0.0176, 0.0046, -0.0102, + -0.0300, -0.0160], device='cuda:0'), grad: tensor([ 4.7032e-08, 6.2399e-08, 5.1223e-08, -1.6158e-07, 1.2964e-06, + 1.1595e-07, 1.8626e-08, 2.5122e-07, 6.1467e-08, -1.7360e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 220.92, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4701 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.19 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.3143, 0.2309, -0.0455, ..., -0.1323, 0.0140, 0.0136], + [ 0.0123, -0.1641, -0.0403, ..., 0.0441, -0.0701, -0.0928], + [-0.2367, -0.3054, 0.0565, ..., -0.1169, -0.0611, -0.2135], + ..., + [ 0.0793, 0.1918, -0.0967, ..., -0.1672, 0.0904, 0.0966], + [ 0.0311, -0.2042, -0.0529, ..., -0.1486, -0.1305, -0.1164], + [-0.2645, -0.0892, -0.0646, ..., 0.1054, -0.3199, 0.1169]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, -6.5425e-08, 0.0000e+00, ..., 2.7940e-09, + 2.3283e-10, -1.9092e-08], + [-4.9127e-08, 2.3283e-10, 0.0000e+00, ..., -9.8487e-08, + 1.1642e-09, 2.0955e-09], + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 2.4680e-08, 0.0000e+00], + ..., + [ 2.0256e-08, 4.6566e-10, 0.0000e+00, ..., 4.5868e-08, + 6.9849e-10, 2.5611e-09], + [ 1.7928e-08, 2.3283e-10, 0.0000e+00, ..., 3.6322e-08, + 3.0268e-09, 0.0000e+00], + [ 3.6089e-08, 1.3970e-09, 0.0000e+00, ..., 5.9837e-08, + 9.3132e-10, -1.1874e-08]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0260, -0.0297, 0.0176, -0.0276, 0.0422, 0.0176, 0.0045, -0.0080, + -0.0300, -0.0160], device='cuda:0'), grad: tensor([-1.0361e-07, -6.6590e-07, 2.3213e-07, -3.2596e-07, -1.5181e-07, + 1.1665e-07, 1.2014e-07, 3.1549e-07, 2.5914e-07, 2.0722e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 220.56, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4643 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.14 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.3145, 0.2315, -0.0451, ..., -0.1326, 0.0141, 0.0137], + [ 0.0126, -0.1629, -0.0388, ..., 0.0453, -0.0702, -0.0919], + [-0.2382, -0.3055, 0.0568, ..., -0.1172, -0.0612, -0.2140], + ..., + [ 0.0790, 0.1907, -0.0982, ..., -0.1675, 0.0903, 0.0962], + [ 0.0310, -0.2050, -0.0528, ..., -0.1488, -0.1305, -0.1164], + [-0.2650, -0.0896, -0.0656, ..., 0.1054, -0.3201, 0.1169]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -8.6147e-09, -6.9849e-10, ..., 0.0000e+00, + 0.0000e+00, -6.9849e-10], + [-3.7253e-09, 4.6566e-10, 0.0000e+00, ..., -7.2177e-09, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 2.3283e-10], + ..., + [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10], + [ 1.1642e-09, 4.6566e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 4.1910e-09, 3.2596e-09, 2.3283e-10, ..., 2.7940e-09, + 4.6566e-10, 9.3132e-10]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0265, -0.0294, 0.0176, -0.0277, 0.0422, 0.0177, 0.0041, -0.0088, + -0.0300, -0.0160], device='cuda:0'), grad: tensor([-1.1874e-08, -2.9569e-08, -7.2317e-07, 1.9325e-08, 6.0536e-09, + -3.8650e-08, 2.7707e-08, 1.9558e-08, 7.1060e-07, 2.7474e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 221.13, cls_loss 0.0004 cls_loss_mapping 0.0014 cls_loss_causal 0.4742 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.11 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.3147, 0.2317, -0.0449, ..., -0.1334, 0.0140, 0.0137], + [ 0.0126, -0.1629, -0.0388, ..., 0.0453, -0.0702, -0.0916], + [-0.2388, -0.3057, 0.0574, ..., -0.1175, -0.0613, -0.2148], + ..., + [ 0.0790, 0.1908, -0.0982, ..., -0.1675, 0.0898, 0.0960], + [ 0.0310, -0.2052, -0.0541, ..., -0.1478, -0.1305, -0.1165], + [-0.2659, -0.0900, -0.0659, ..., 0.1054, -0.3206, 0.1169]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.5367e-08, 0.0000e+00, ..., 2.3283e-10, + 2.7940e-09, -2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.0955e-09, + 5.3551e-09, 7.5437e-08], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.6065e-08, 1.3970e-08], + ..., + [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 5.1223e-09, -7.1013e-08], + [ 2.5379e-08, 2.3283e-10, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 2.0955e-09], + [ 5.8208e-09, 7.2177e-09, 0.0000e+00, ..., -1.6298e-08, + 1.5832e-08, 8.3819e-09]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0265, -0.0294, 0.0176, -0.0275, 0.0422, 0.0179, 0.0030, -0.0089, + -0.0297, -0.0160], device='cuda:0'), grad: tensor([-1.2340e-08, 3.6089e-07, 4.8662e-08, -4.7963e-07, 2.5379e-08, + 3.4925e-08, 1.3853e-07, -3.2806e-07, 1.5274e-07, 7.5903e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 220.23, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4526 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.19 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.3154, 0.2320, -0.0450, ..., -0.1340, 0.0137, 0.0139], + [ 0.0127, -0.1629, -0.0386, ..., 0.0454, -0.0702, -0.0917], + [-0.2398, -0.3058, 0.0569, ..., -0.1178, -0.0615, -0.2159], + ..., + [ 0.0790, 0.1908, -0.0982, ..., -0.1675, 0.0898, 0.0961], + [ 0.0310, -0.2055, -0.0544, ..., -0.1474, -0.1306, -0.1164], + [-0.2665, -0.0908, -0.0660, ..., 0.1054, -0.3208, 0.1169]], + device='cuda:0'), grad: tensor([[ 8.1491e-10, -3.2596e-09, 2.3283e-10, ..., 5.8208e-10, + 0.0000e+00, 1.1642e-10], + [-1.1642e-08, -1.1642e-09, -2.3283e-09, ..., -8.1491e-09, + 3.4925e-10, 6.9849e-10], + [ 1.5134e-09, 1.1642e-09, 1.1642e-10, ..., 1.3970e-09, + 8.1491e-10, 1.0477e-09], + ..., + [ 6.0536e-09, 1.2806e-09, 5.8208e-10, ..., 7.3342e-09, + 2.3283e-10, 1.9791e-09], + [ 2.5611e-09, 2.3283e-10, 6.9849e-10, ..., 1.3970e-09, + 8.1491e-10, 1.0477e-09], + [ 6.9849e-10, 2.5611e-09, 0.0000e+00, ..., -5.8208e-09, + 4.6566e-10, -4.0745e-09]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0264, -0.0294, 0.0176, -0.0274, 0.0422, 0.0180, 0.0027, -0.0089, + -0.0295, -0.0160], device='cuda:0'), grad: tensor([ 2.1537e-08, -1.9092e-08, -1.2061e-07, 7.7067e-08, 2.3283e-08, + -1.0396e-07, 3.6438e-08, 5.4948e-08, 5.3551e-08, -3.4925e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 220.26, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4629 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.13 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.3162, 0.2323, -0.0449, ..., -0.1358, 0.0135, 0.0132], + [ 0.0129, -0.1630, -0.0368, ..., 0.0469, -0.0703, -0.0913], + [-0.2401, -0.3060, 0.0572, ..., -0.1180, -0.0615, -0.2166], + ..., + [ 0.0788, 0.1910, -0.1000, ..., -0.1680, 0.0898, 0.0961], + [ 0.0309, -0.2057, -0.0558, ..., -0.1476, -0.1306, -0.1165], + [-0.2675, -0.0910, -0.0676, ..., 0.1054, -0.3211, 0.1170]], + device='cuda:0'), grad: tensor([[ 1.9791e-09, 2.8335e-07, 0.0000e+00, ..., 3.1921e-07, + -3.4925e-10, 4.6566e-10], + [ 2.0978e-07, 1.1059e-08, 4.6566e-10, ..., 7.1712e-08, + 1.1642e-10, 6.7940e-07], + [ 4.4238e-09, 4.2259e-08, 1.0477e-09, ..., 4.7148e-08, + 5.8208e-10, 4.5402e-09], + ..., + [-3.5297e-07, 2.5611e-09, -6.6357e-09, ..., -8.9989e-08, + -1.5134e-09, -1.1194e-06], + [ 1.2224e-08, 2.6426e-08, -1.1642e-09, ..., 3.7602e-08, + 6.9849e-10, 8.4401e-08], + [ 1.0466e-07, 8.1491e-09, 5.8208e-10, ..., -4.7265e-08, + 2.3283e-10, 2.8056e-07]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0257, -0.0293, 0.0176, -0.0278, 0.0422, 0.0184, 0.0028, -0.0091, + -0.0295, -0.0161], device='cuda:0'), grad: tensor([ 1.5199e-06, 2.0806e-06, 2.8475e-07, 1.7579e-08, 4.4191e-07, + 5.6392e-07, -2.4941e-06, -3.3136e-06, 8.2073e-08, 8.1118e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 220.42, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4813 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.12 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.3164, 0.2328, -0.0451, ..., -0.1370, 0.0133, 0.0128], + [ 0.0131, -0.1631, -0.0351, ..., 0.0478, -0.0704, -0.0910], + [-0.2416, -0.3063, 0.0576, ..., -0.1186, -0.0617, -0.2173], + ..., + [ 0.0787, 0.1911, -0.1012, ..., -0.1684, 0.0897, 0.0959], + [ 0.0309, -0.2065, -0.0574, ..., -0.1482, -0.1306, -0.1169], + [-0.2685, -0.0910, -0.0687, ..., 0.1055, -0.3212, 0.1172]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.3632e-08, -1.7695e-08, ..., -1.0477e-09, + 0.0000e+00, -1.8626e-09], + [ 2.3283e-10, 9.3132e-10, 2.3283e-10, ..., 3.8417e-09, + 0.0000e+00, 4.1910e-09], + [ 1.1642e-10, 4.6566e-10, -1.1642e-10, ..., 5.8208e-10, + 0.0000e+00, 4.6566e-10], + ..., + [-3.4925e-10, -6.9849e-10, 2.3283e-10, ..., 3.1432e-09, + 0.0000e+00, -8.4983e-09], + [ 0.0000e+00, 5.8208e-10, 4.6566e-10, ..., 5.8208e-10, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 2.3283e-09, 1.1642e-09, ..., -8.2771e-08, + 0.0000e+00, -5.6811e-08]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0256, -0.0290, 0.0175, -0.0274, 0.0422, 0.0179, 0.0029, -0.0095, + -0.0296, -0.0160], device='cuda:0'), grad: tensor([-5.2038e-08, 5.3085e-08, 8.1491e-10, 8.8476e-09, 2.9220e-07, + 2.3749e-08, 3.7835e-08, -1.4715e-07, -4.3074e-09, -2.1083e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 220.95, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4613 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.13 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.3166, 0.2330, -0.0445, ..., -0.1377, 0.0137, 0.0127], + [ 0.0133, -0.1637, -0.0343, ..., 0.0478, -0.0709, -0.0912], + [-0.2435, -0.3064, 0.0556, ..., -0.1189, -0.0619, -0.2179], + ..., + [ 0.0787, 0.1917, -0.1012, ..., -0.1684, 0.0900, 0.0960], + [ 0.0309, -0.2067, -0.0575, ..., -0.1484, -0.1306, -0.1169], + [-0.2686, -0.0912, -0.0687, ..., 0.1057, -0.3213, 0.1177]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 0.0000e+00, 4.6566e-09], + [ 2.3283e-10, 9.3132e-10, 2.3283e-10, ..., 4.4238e-09, + 2.3283e-10, 3.7253e-09], + [ 0.0000e+00, 3.0268e-09, -4.8894e-09, ..., 2.0955e-09, + 6.9849e-10, 3.9581e-09], + ..., + [ 0.0000e+00, -3.7253e-09, 1.8626e-09, ..., 1.7462e-08, + -2.3283e-10, 1.6764e-08], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.3283e-09, + 2.3283e-10, 2.7940e-09], + [ 3.4925e-09, 4.6566e-10, 4.6566e-10, ..., 3.6322e-08, + 6.9849e-10, -6.7521e-08]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0255, -0.0288, 0.0172, -0.0276, 0.0420, 0.0178, 0.0029, -0.0094, + -0.0295, -0.0158], device='cuda:0'), grad: tensor([ 2.1886e-08, 2.5146e-08, -4.5402e-08, 3.8184e-08, -1.7066e-07, + -2.6077e-08, 7.9162e-09, 7.0315e-08, 1.2806e-08, 9.2201e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 220.67, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4661 re_mapping 0.0031 re_causal 0.0091 /// teacc 98.95 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.3175, 0.2331, -0.0443, ..., -0.1383, 0.0136, 0.0126], + [ 0.0133, -0.1637, -0.0342, ..., 0.0480, -0.0710, -0.0914], + [-0.2437, -0.3065, 0.0557, ..., -0.1194, -0.0617, -0.2185], + ..., + [ 0.0787, 0.1920, -0.1013, ..., -0.1685, 0.0896, 0.0959], + [ 0.0309, -0.2074, -0.0575, ..., -0.1486, -0.1307, -0.1169], + [-0.2695, -0.0924, -0.0688, ..., 0.1057, -0.3224, 0.1178]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, -2.5611e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 4.6566e-10], + [-4.0280e-08, -7.9162e-09, 0.0000e+00, ..., -3.0268e-08, + 2.3283e-10, -2.7940e-09], + [ 1.6298e-09, 2.0955e-09, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 2.5611e-09], + ..., + [ 3.3993e-08, 6.7521e-09, 0.0000e+00, ..., 2.7474e-08, + -2.3283e-10, 6.9849e-10], + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., -3.0268e-09, + 2.3283e-10, -3.9581e-09]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0252, -0.0288, 0.0172, -0.0275, 0.0419, 0.0188, 0.0023, -0.0096, + -0.0295, -0.0158], device='cuda:0'), grad: tensor([ 4.4238e-09, -2.0163e-07, -4.1910e-09, 4.6566e-09, 1.1409e-08, + -8.6147e-09, 1.4435e-08, 1.8720e-07, 5.8208e-09, -9.3132e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 220.47, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4551 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.08 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.3180, 0.2331, -0.0442, ..., -0.1394, 0.0116, 0.0126], + [ 0.0133, -0.1637, -0.0341, ..., 0.0480, -0.0711, -0.0915], + [-0.2439, -0.3067, 0.0557, ..., -0.1198, -0.0611, -0.2190], + ..., + [ 0.0786, 0.1921, -0.1014, ..., -0.1686, 0.0888, 0.0954], + [ 0.0308, -0.2082, -0.0575, ..., -0.1489, -0.1307, -0.1171], + [-0.2702, -0.0926, -0.0688, ..., 0.1057, -0.3231, 0.1179]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [-3.0268e-09, 3.9581e-09, 0.0000e+00, ..., -3.0268e-09, + 4.6566e-10, 5.1223e-09], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 3.0268e-09, -1.3271e-08, 0.0000e+00, ..., 6.2864e-09, + -1.1642e-09, -9.5461e-09], + [ 1.1642e-09, 6.9849e-10, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 6.9849e-10], + [ 1.9325e-08, 8.3819e-09, 0.0000e+00, ..., 5.2620e-08, + 4.6566e-10, 2.3982e-08]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0246, -0.0288, 0.0173, -0.0274, 0.0420, 0.0198, 0.0021, -0.0099, + -0.0296, -0.0158], device='cuda:0'), grad: tensor([ 1.6298e-09, -6.2864e-09, 3.2596e-09, 2.0955e-09, -2.0280e-07, + 4.1910e-09, 7.2177e-09, -7.4506e-09, 9.7789e-09, 2.0349e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 220.73, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4768 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.15 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.3182, 0.2334, -0.0441, ..., -0.1396, 0.0117, 0.0125], + [ 0.0133, -0.1638, -0.0342, ..., 0.0480, -0.0712, -0.0921], + [-0.2445, -0.3073, 0.0558, ..., -0.1198, -0.0616, -0.2207], + ..., + [ 0.0787, 0.1925, -0.1014, ..., -0.1686, 0.0890, 0.0958], + [ 0.0308, -0.2088, -0.0574, ..., -0.1492, -0.1308, -0.1173], + [-0.2711, -0.0933, -0.0688, ..., 0.1058, -0.3234, 0.1179]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 6.9849e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + 5.3551e-09, 1.1874e-08], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 2.0023e-08, 2.3050e-08], + ..., + [-1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + -1.8394e-08, -1.2573e-08], + [ 4.6566e-10, 6.9849e-10, 0.0000e+00, ..., 1.5367e-08, + 1.0943e-08, 2.1420e-08], + [ 0.0000e+00, 2.5611e-09, 0.0000e+00, ..., -1.2107e-08, + 6.2864e-09, -1.8626e-09]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0247, -0.0288, 0.0172, -0.0287, 0.0420, 0.0210, 0.0019, -0.0096, + -0.0296, -0.0158], device='cuda:0'), grad: tensor([-4.1910e-09, 5.2154e-08, 1.1316e-07, -2.5169e-07, 2.1653e-08, + 6.9151e-08, -2.7800e-07, -5.4017e-08, 3.2457e-07, 1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 220.77, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4616 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.24 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.3184, 0.2337, -0.0441, ..., -0.1399, 0.0118, 0.0123], + [ 0.0133, -0.1640, -0.0342, ..., 0.0478, -0.0713, -0.0925], + [-0.2465, -0.3084, 0.0557, ..., -0.1197, -0.0631, -0.2230], + ..., + [ 0.0790, 0.1956, -0.1012, ..., -0.1687, 0.0892, 0.0989], + [ 0.0308, -0.2119, -0.0578, ..., -0.1495, -0.1308, -0.1178], + [-0.2738, -0.0967, -0.0688, ..., 0.1058, -0.3240, 0.1151]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 3.2596e-09, 4.8894e-09, 4.6566e-10, ..., -2.7940e-09, + 1.3970e-09, 8.6147e-09], + [ 2.3283e-10, 6.9849e-10, -4.6566e-10, ..., 2.3283e-10, + 2.3283e-10, 9.3132e-10], + ..., + [-3.7253e-09, -5.1223e-09, 0.0000e+00, ..., 2.3283e-09, + -1.6298e-09, -9.0804e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0248, -0.0289, 0.0172, -0.0261, 0.0420, 0.0183, 0.0020, -0.0066, + -0.0297, -0.0185], device='cuda:0'), grad: tensor([ 9.0804e-09, 5.8906e-08, -4.7032e-08, 6.7521e-09, 3.2596e-09, + 2.2817e-08, -4.6566e-08, -1.5134e-08, 1.6065e-08, 4.4238e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 221.00, cls_loss 0.0004 cls_loss_mapping 0.0015 cls_loss_causal 0.4671 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.18 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.3188, 0.2337, -0.0440, ..., -0.1414, 0.0117, 0.0123], + [ 0.0134, -0.1642, -0.0342, ..., 0.0478, -0.0714, -0.0917], + [-0.2472, -0.3086, 0.0556, ..., -0.1200, -0.0634, -0.2236], + ..., + [ 0.0790, 0.1959, -0.1011, ..., -0.1688, 0.0893, 0.0990], + [ 0.0308, -0.2118, -0.0579, ..., -0.1491, -0.1308, -0.1179], + [-0.2754, -0.0969, -0.0688, ..., 0.1058, -0.3243, 0.1149]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.2864e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, -2.3283e-10], + [ 4.4238e-09, 1.6065e-08, 6.5193e-09, ..., 2.0722e-08, + 7.4506e-09, 2.3283e-10], + [ 0.0000e+00, -2.0489e-08, -8.6147e-09, ..., 2.3283e-10, + 3.2131e-08, 0.0000e+00], + ..., + [ 4.6566e-10, 4.4238e-09, 1.8626e-09, ..., 3.9581e-09, + 3.7253e-09, 1.6298e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 2.5844e-08, 0.0000e+00], + [ 1.3621e-07, 2.3283e-09, 0.0000e+00, ..., 6.3889e-07, + 4.6566e-10, -1.8626e-09]], device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0243, -0.0288, 0.0172, -0.0256, 0.0420, 0.0178, 0.0021, -0.0065, + -0.0296, -0.0187], device='cuda:0'), grad: tensor([-8.1491e-09, 6.4727e-07, -6.1654e-07, -2.8964e-07, -1.5479e-06, + 2.0256e-08, 5.5879e-09, 1.8510e-07, 1.0990e-07, 1.4957e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 220.55, cls_loss 0.0004 cls_loss_mapping 0.0015 cls_loss_causal 0.4714 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.09 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.3188, 0.2342, -0.0441, ..., -0.1415, 0.0117, 0.0123], + [ 0.0134, -0.1642, -0.0342, ..., 0.0477, -0.0715, -0.0917], + [-0.2474, -0.3089, 0.0558, ..., -0.1201, -0.0635, -0.2242], + ..., + [ 0.0791, 0.1959, -0.1011, ..., -0.1692, 0.0894, 0.0990], + [ 0.0308, -0.2126, -0.0589, ..., -0.1488, -0.1308, -0.1180], + [-0.2767, -0.0970, -0.0689, ..., 0.1059, -0.3253, 0.1150]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.4925e-09, 0.0000e+00, ..., 1.3039e-08, + 0.0000e+00, 6.7521e-09], + [ 2.3283e-10, 2.0955e-09, 0.0000e+00, ..., 6.8918e-08, + 6.9849e-10, 2.0489e-08], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 3.9581e-09, + 1.1642e-09, 2.0955e-09], + ..., + [ 0.0000e+00, 3.9581e-09, 0.0000e+00, ..., 9.7789e-08, + 4.1910e-09, 4.1444e-08], + [-2.3283e-10, 1.6298e-09, 0.0000e+00, ..., 1.1991e-07, + 0.0000e+00, 3.4925e-08], + [ 6.9849e-10, -2.0023e-08, 0.0000e+00, ..., -3.3677e-06, + 4.6566e-10, -9.4529e-07]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0246, -0.0288, 0.0172, -0.0255, 0.0420, 0.0177, 0.0017, -0.0065, + -0.0295, -0.0186], device='cuda:0'), grad: tensor([ 4.1677e-08, 2.1909e-07, 5.1456e-08, 6.2864e-09, 1.1362e-05, + 1.1409e-07, 4.0978e-08, 3.9884e-07, 3.7649e-07, -1.2614e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 220.82, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4475 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.23 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.3190, 0.2346, -0.0440, ..., -0.1420, 0.0123, 0.0116], + [ 0.0133, -0.1644, -0.0343, ..., 0.0475, -0.0716, -0.0919], + [-0.2475, -0.3091, 0.0559, ..., -0.1208, -0.0637, -0.2244], + ..., + [ 0.0791, 0.1960, -0.1012, ..., -0.1694, 0.0894, 0.0990], + [ 0.0308, -0.2129, -0.0589, ..., -0.1495, -0.1309, -0.1182], + [-0.2778, -0.0970, -0.0689, ..., 0.1072, -0.3255, 0.1152]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.5611e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3970e-09, 2.3283e-10, 0.0000e+00, ..., -5.3551e-09, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.0955e-09, + 9.3132e-10, 6.9849e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 0.0000e+00], + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0244, -0.0288, 0.0173, -0.0255, 0.0411, 0.0176, 0.0014, -0.0065, + -0.0295, -0.0182], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.3411e-06, -1.6302e-05, 2.8173e-08, 6.4494e-08, + -6.2864e-09, 3.2829e-08, 1.4767e-05, 2.7940e-08, 2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 220.92, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4647 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.18 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.3193, 0.2350, -0.0433, ..., -0.1424, 0.0124, 0.0114], + [ 0.0134, -0.1645, -0.0343, ..., 0.0475, -0.0717, -0.0920], + [-0.2476, -0.3093, 0.0560, ..., -0.1211, -0.0639, -0.2246], + ..., + [ 0.0791, 0.1960, -0.1012, ..., -0.1694, 0.0894, 0.0990], + [ 0.0308, -0.2132, -0.0589, ..., -0.1497, -0.1310, -0.1182], + [-0.2785, -0.0970, -0.0689, ..., 0.1072, -0.3256, 0.1152]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -2.1397e-07, 1.3970e-09, ..., 4.8894e-09, + -2.9337e-08, 0.0000e+00], + [ 9.3132e-10, 1.0943e-08, 4.6566e-10, ..., 9.3132e-10, + 3.9581e-09, 2.0955e-09], + [ 6.9849e-10, 3.0268e-08, 9.3132e-10, ..., 3.0268e-09, + 5.1223e-09, 2.3283e-10], + ..., + [ 4.6566e-10, 3.0268e-09, 0.0000e+00, ..., 1.0245e-08, + 4.6566e-10, 7.6834e-09], + [-3.7253e-09, 3.7486e-08, 9.3132e-10, ..., 2.7940e-09, + 5.5879e-09, 4.6566e-10], + [ 0.0000e+00, 1.7928e-08, 2.3283e-10, ..., -1.8859e-08, + 2.5611e-09, -1.8161e-08]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0246, -0.0289, 0.0174, -0.0254, 0.0411, 0.0177, 0.0013, -0.0066, + -0.0296, -0.0182], device='cuda:0'), grad: tensor([-4.0932e-07, 6.5193e-08, 1.1129e-07, 2.0140e-07, 3.2131e-08, + 9.2201e-08, 1.2224e-07, 1.2293e-07, -3.1549e-07, -1.7229e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 220.83, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4916 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.17 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.3195, 0.2353, -0.0426, ..., -0.1429, 0.0126, 0.0113], + [ 0.0133, -0.1647, -0.0342, ..., 0.0475, -0.0719, -0.0920], + [-0.2483, -0.3095, 0.0560, ..., -0.1220, -0.0642, -0.2249], + ..., + [ 0.0791, 0.1961, -0.1014, ..., -0.1696, 0.0895, 0.0990], + [ 0.0309, -0.2140, -0.0582, ..., -0.1500, -0.1310, -0.1182], + [-0.2803, -0.0970, -0.0689, ..., 0.1072, -0.3260, 0.1152]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [ 1.1176e-08, 1.8626e-09, 0.0000e+00, ..., 1.6298e-08, + 2.7940e-09, 2.3283e-09], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 1.8626e-09], + ..., + [-4.6566e-10, -4.6566e-09, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, -1.3970e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.3504e-08, 2.3283e-09, 0.0000e+00, ..., 1.7229e-08, + 4.6566e-10, -4.6566e-10]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0248, -0.0289, 0.0174, -0.0254, 0.0411, 0.0174, 0.0014, -0.0067, + -0.0288, -0.0182], device='cuda:0'), grad: tensor([-1.5832e-08, 7.4040e-08, 1.7695e-08, -5.4948e-07, -1.1642e-07, + 5.2666e-07, 4.6566e-09, -5.1223e-09, 5.1223e-09, 6.5193e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 220.71, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4796 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.13 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.3212, 0.2354, -0.0425, ..., -0.1438, 0.0121, 0.0102], + [ 0.0126, -0.1670, -0.0342, ..., 0.0475, -0.0735, -0.0935], + [-0.2486, -0.3099, 0.0564, ..., -0.1224, -0.0645, -0.2255], + ..., + [ 0.0803, 0.1966, -0.1014, ..., -0.1697, 0.0904, 0.0991], + [ 0.0309, -0.2148, -0.0584, ..., -0.1502, -0.1311, -0.1183], + [-0.2814, -0.0971, -0.0690, ..., 0.1072, -0.3266, 0.1153]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-09, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 4.1910e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 2.3283e-09], + ..., + [-4.6566e-10, -4.6566e-09, 0.0000e+00, ..., 4.6566e-10, + 1.7695e-08, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 1.3970e-09, + 1.3970e-09, 4.6566e-09]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0239, -0.0291, 0.0174, -0.0256, 0.0411, 0.0176, 0.0014, -0.0065, + -0.0288, -0.0182], device='cuda:0'), grad: tensor([ 0.0000e+00, 5.7276e-08, 3.1665e-08, -8.8476e-08, 2.3283e-09, + 4.4703e-08, 3.6787e-08, 4.5169e-08, -1.7649e-07, 5.6811e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 220.94, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4628 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.09 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.3213, 0.2358, -0.0426, ..., -0.1438, 0.0122, 0.0101], + [ 0.0129, -0.1654, -0.0343, ..., 0.0495, -0.0735, -0.0937], + [-0.2488, -0.3101, 0.0566, ..., -0.1229, -0.0647, -0.2259], + ..., + [ 0.0803, 0.1967, -0.1015, ..., -0.1698, 0.0904, 0.0991], + [ 0.0308, -0.2155, -0.0586, ..., -0.1506, -0.1311, -0.1185], + [-0.2829, -0.0971, -0.0690, ..., 0.1072, -0.3268, 0.1153]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.8626e-09, 1.8626e-09, ..., 5.5879e-09, + 3.2596e-09, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 4.6566e-09, ..., 1.3970e-09, + 2.7940e-09, 2.7940e-09], + [ 1.2154e-07, 9.1735e-08, -1.0710e-08, ..., 0.0000e+00, + 2.1234e-07, 2.7986e-07], + ..., + [-1.2340e-07, -9.6858e-08, 0.0000e+00, ..., 1.3970e-09, + -2.1607e-07, -2.8312e-07], + [ 4.6566e-10, 9.3132e-10, 4.6566e-10, ..., 9.3132e-10, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -4.6566e-09, + 0.0000e+00, -4.6566e-09]], device='cuda:0') +Epoch 362, bias, value: tensor([ 2.4050e-02, -2.8578e-02, 1.7374e-02, -2.5808e-02, 4.1172e-02, + 1.6985e-02, 9.0099e-05, -6.5194e-03, -2.8890e-02, -1.8177e-02], + device='cuda:0'), grad: tensor([ 6.0070e-08, 7.0315e-08, 9.8906e-07, 7.4506e-09, 3.5390e-08, + 5.5879e-09, -7.4506e-08, -1.0971e-06, 4.6566e-09, -3.2596e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 221.20, cls_loss 0.0004 cls_loss_mapping 0.0017 cls_loss_causal 0.4525 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.05 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.3214, 0.2362, -0.0427, ..., -0.1441, 0.0122, 0.0101], + [ 0.0129, -0.1654, -0.0344, ..., 0.0497, -0.0737, -0.0939], + [-0.2491, -0.3100, 0.0568, ..., -0.1233, -0.0650, -0.2264], + ..., + [ 0.0804, 0.1967, -0.1015, ..., -0.1700, 0.0905, 0.0991], + [ 0.0308, -0.2161, -0.0587, ..., -0.1509, -0.1312, -0.1186], + [-0.2835, -0.0971, -0.0690, ..., 0.1073, -0.3269, 0.1153]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + -4.6566e-10, 0.0000e+00], + [-8.8476e-09, -7.2177e-08, 0.0000e+00, ..., -2.9802e-08, + 1.3970e-09, 5.1223e-09], + [ 5.1223e-09, 4.1910e-08, 0.0000e+00, ..., 1.3970e-08, + 9.3132e-10, 4.6566e-10], + ..., + [ 1.8626e-09, 1.2573e-08, 0.0000e+00, ..., 1.0710e-08, + -6.0536e-09, -2.4214e-08], + [ 4.6566e-10, 3.2596e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 9.7789e-09, 0.0000e+00, ..., 2.3283e-09, + 5.5879e-09, 2.0023e-08]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0241, -0.0287, 0.0176, -0.0258, 0.0412, 0.0169, -0.0001, -0.0065, + -0.0289, -0.0181], device='cuda:0'), grad: tensor([ 2.2352e-08, -3.5670e-07, 1.8161e-07, 6.5193e-09, 6.5193e-09, + 1.0710e-08, -6.9849e-09, 5.9139e-08, 1.4435e-08, 7.5437e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 220.84, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4429 re_mapping 0.0030 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.3215, 0.2363, -0.0422, ..., -0.1457, 0.0123, 0.0095], + [ 0.0129, -0.1662, -0.0346, ..., 0.0497, -0.0743, -0.0948], + [-0.2494, -0.3103, 0.0568, ..., -0.1239, -0.0667, -0.2282], + ..., + [ 0.0805, 0.1970, -0.1012, ..., -0.1699, 0.0910, 0.0992], + [ 0.0308, -0.2164, -0.0587, ..., -0.1511, -0.1312, -0.1188], + [-0.2862, -0.0972, -0.0695, ..., 0.1074, -0.3289, 0.1153]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -4.6566e-10, + 3.2131e-08, 8.8476e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 1.4435e-08, 1.0710e-08], + ..., + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 3.9581e-08, 3.4925e-08], + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., -1.5367e-08, + 2.0489e-08, 1.3970e-09], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 5.1223e-09, + 5.1223e-09, 5.5879e-09]], device='cuda:0') +Epoch 364, bias, value: tensor([ 2.3500e-02, -2.8931e-02, 1.7540e-02, -2.5866e-02, 4.1079e-02, + 1.6972e-02, 4.8077e-05, -6.3687e-03, -2.8810e-02, -1.8146e-02], + device='cuda:0'), grad: tensor([ 4.0047e-08, 2.1886e-07, 1.0990e-07, -6.4261e-07, -3.2596e-09, + 1.9511e-07, 2.9802e-08, 1.1781e-07, -1.4808e-07, 9.4529e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 220.24, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4607 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.13 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.3216, 0.2371, -0.0415, ..., -0.1461, 0.0124, 0.0094], + [ 0.0132, -0.1663, -0.0322, ..., 0.0516, -0.0731, -0.0954], + [-0.2495, -0.3106, 0.0575, ..., -0.1244, -0.0668, -0.2288], + ..., + [ 0.0800, 0.1971, -0.1041, ..., -0.1713, 0.0902, 0.0992], + [ 0.0307, -0.2168, -0.0597, ..., -0.1514, -0.1313, -0.1189], + [-0.2856, -0.0972, -0.0699, ..., 0.1077, -0.3294, 0.1154]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -4.6566e-10, ..., 3.4459e-08, + 4.6566e-10, 1.0245e-08], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 6.1467e-08, + 2.3283e-09, 1.8161e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 5.1223e-09, 5.5879e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.4808e-07, + 1.0245e-08, 4.6100e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.5600e-07, + 1.8626e-09, 4.1444e-08], + [ 6.0536e-09, 9.3132e-10, 0.0000e+00, ..., -2.0996e-05, + 1.3970e-09, -5.4277e-06]], device='cuda:0') +Epoch 365, bias, value: tensor([ 2.3922e-02, -2.8393e-02, 1.7571e-02, -2.5619e-02, 4.0832e-02, + 1.6721e-02, 5.6092e-05, -6.7365e-03, -2.8788e-02, -1.8035e-02], + device='cuda:0'), grad: tensor([ 1.1595e-07, 2.1653e-07, 4.7963e-08, -2.0023e-08, 6.8605e-05, + 1.0608e-06, 8.6613e-08, 5.3225e-07, 5.3039e-07, -7.1228e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 220.91, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4667 re_mapping 0.0031 re_causal 0.0090 /// teacc 99.18 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.3217, 0.2382, -0.0413, ..., -0.1464, 0.0126, 0.0094], + [ 0.0131, -0.1664, -0.0321, ..., 0.0517, -0.0727, -0.0952], + [-0.2496, -0.3106, 0.0577, ..., -0.1246, -0.0670, -0.2293], + ..., + [ 0.0800, 0.1971, -0.1042, ..., -0.1714, 0.0900, 0.0992], + [ 0.0307, -0.2176, -0.0601, ..., -0.1516, -0.1313, -0.1190], + [-0.2859, -0.0973, -0.0699, ..., 0.1080, -0.3296, 0.1155]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0245e-08, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 5.1223e-09, 6.9849e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 1.3970e-09], + ..., + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 4.1910e-09, 3.7253e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + [ 4.6566e-10, 2.7940e-09, 0.0000e+00, ..., -3.2596e-09, + 4.6566e-10, -4.6566e-10]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0246, -0.0283, 0.0176, -0.0255, 0.0405, 0.0162, 0.0001, -0.0068, + -0.0287, -0.0179], device='cuda:0'), grad: tensor([-1.6298e-08, 2.8405e-08, 0.0000e+00, -9.3132e-10, 0.0000e+00, + -4.3306e-08, 1.3970e-08, 2.0023e-08, 3.7253e-09, 4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 220.75, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4377 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.09 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.3217, 0.2394, -0.0422, ..., -0.1467, 0.0122, 0.0089], + [ 0.0131, -0.1664, -0.0321, ..., 0.0516, -0.0728, -0.0955], + [-0.2497, -0.3109, 0.0578, ..., -0.1249, -0.0672, -0.2312], + ..., + [ 0.0800, 0.1971, -0.1043, ..., -0.1715, 0.0899, 0.0992], + [ 0.0307, -0.2181, -0.0606, ..., -0.1521, -0.1313, -0.1192], + [-0.2862, -0.0973, -0.0698, ..., 0.1081, -0.3298, 0.1155]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 3.7253e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 3.2596e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + ..., + [-6.9849e-09, -9.7789e-09, 0.0000e+00, ..., 3.2596e-09, + 0.0000e+00, -5.5879e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 4.6566e-09, 0.0000e+00, ..., -1.2573e-08, + 0.0000e+00, -5.1223e-09]], device='cuda:0') +Epoch 367, bias, value: tensor([ 2.5568e-02, -2.8364e-02, 1.7587e-02, -2.5473e-02, 4.0444e-02, + 1.6046e-02, 9.3103e-05, -6.8096e-03, -2.8749e-02, -1.7818e-02], + device='cuda:0'), grad: tensor([-2.7940e-09, 1.8626e-08, 9.3132e-10, 5.5879e-09, 2.6077e-08, + -1.4435e-08, 1.3039e-08, -2.5611e-08, 5.5879e-09, -1.3970e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 220.53, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4667 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.14 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.3218, 0.2397, -0.0422, ..., -0.1484, 0.0124, 0.0083], + [ 0.0130, -0.1663, -0.0322, ..., 0.0517, -0.0729, -0.0928], + [-0.2489, -0.3111, 0.0581, ..., -0.1252, -0.0674, -0.2318], + ..., + [ 0.0800, 0.1971, -0.1044, ..., -0.1716, 0.0900, 0.0983], + [ 0.0307, -0.2186, -0.0612, ..., -0.1524, -0.1314, -0.1193], + [-0.2867, -0.0973, -0.0698, ..., 0.1082, -0.3299, 0.1155]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 2.3283e-09, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 4.1910e-09], + [ 4.6566e-10, 1.8626e-09, -3.2596e-09, ..., 4.6566e-10, + 9.3132e-10, 2.3283e-09], + ..., + [-4.6566e-10, -3.7253e-09, 0.0000e+00, ..., 3.7253e-09, + 4.6566e-10, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + [ 6.5193e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-09, + -4.6566e-10, -4.1910e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0253, -0.0268, 0.0175, -0.0258, 0.0403, 0.0162, 0.0002, -0.0079, + -0.0288, -0.0177], device='cuda:0'), grad: tensor([ 6.0536e-09, 1.8161e-08, -3.4459e-08, -5.1223e-09, -2.3749e-08, + 2.7940e-09, 9.3132e-10, 1.0245e-08, 4.1910e-09, 2.3283e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 220.53, cls_loss 0.0003 cls_loss_mapping 0.0014 cls_loss_causal 0.4476 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.17 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.3218, 0.2398, -0.0422, ..., -0.1486, 0.0126, 0.0083], + [ 0.0130, -0.1663, -0.0323, ..., 0.0516, -0.0729, -0.0925], + [-0.2487, -0.3113, 0.0592, ..., -0.1248, -0.0675, -0.2321], + ..., + [ 0.0800, 0.1971, -0.1045, ..., -0.1717, 0.0900, 0.0982], + [ 0.0307, -0.2188, -0.0619, ..., -0.1526, -0.1314, -0.1194], + [-0.2879, -0.0973, -0.0698, ..., 0.1085, -0.3300, 0.1156]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -1.8626e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 3.9581e-08, 2.2817e-08, 0.0000e+00, ..., 3.0734e-08, + 1.3970e-08, 1.1688e-07], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.8626e-09], + ..., + [-5.6811e-08, -3.3062e-08, 0.0000e+00, ..., -3.7253e-08, + -2.0489e-08, -1.6298e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 1.4435e-08, 9.3132e-09, 0.0000e+00, ..., 4.6566e-10, + 5.1223e-09, 3.3528e-08]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0253, -0.0267, 0.0177, -0.0260, 0.0402, 0.0162, 0.0002, -0.0080, + -0.0287, -0.0176], device='cuda:0'), grad: tensor([ 2.3283e-09, 4.1816e-07, 8.8476e-09, 2.3283e-09, 3.1665e-08, + 8.8476e-09, 4.1910e-09, -5.7928e-07, -8.3819e-09, 1.2526e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 220.77, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4899 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.14 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.3224, 0.2400, -0.0422, ..., -0.1493, 0.0126, 0.0082], + [ 0.0130, -0.1665, -0.0323, ..., 0.0516, -0.0731, -0.0925], + [-0.2489, -0.3112, 0.0594, ..., -0.1253, -0.0677, -0.2335], + ..., + [ 0.0801, 0.1971, -0.1045, ..., -0.1718, 0.0901, 0.0982], + [ 0.0307, -0.2191, -0.0621, ..., -0.1528, -0.1314, -0.1196], + [-0.2892, -0.0973, -0.0699, ..., 0.1084, -0.3303, 0.1156]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3749e-07, 1.8626e-08, 5.5414e-08, ..., 2.1514e-07, + 1.1176e-08, 4.8429e-08], + [ 3.7253e-09, 1.0245e-08, 4.6566e-10, ..., 1.3970e-09, + 5.5879e-09, 2.2817e-08], + ..., + [ 1.0803e-07, -3.6322e-08, 2.7474e-08, ..., 1.1455e-07, + -1.6764e-08, -6.9384e-08], + [ 1.0338e-07, 2.7940e-09, 2.4214e-08, ..., 9.2667e-08, + 9.3132e-10, 2.3283e-09], + [ 7.8697e-08, 1.8626e-09, 1.8626e-08, ..., 5.1688e-08, + 4.6566e-10, -8.8476e-09]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0251, -0.0268, 0.0177, -0.0263, 0.0404, 0.0162, 0.0003, -0.0080, + -0.0287, -0.0177], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.2079e-06, 1.0058e-07, -8.8476e-09, -2.4289e-06, + 1.5367e-08, 9.1735e-08, 2.4308e-07, 4.7125e-07, 3.0920e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 220.48, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4605 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.10 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.3246, 0.2404, -0.0422, ..., -0.1496, 0.0128, 0.0076], + [ 0.0109, -0.1696, -0.0323, ..., 0.0514, -0.0732, -0.0931], + [-0.2494, -0.3116, 0.0594, ..., -0.1256, -0.0679, -0.2343], + ..., + [ 0.0825, 0.1989, -0.1046, ..., -0.1719, 0.0901, 0.0985], + [ 0.0307, -0.2214, -0.0625, ..., -0.1532, -0.1314, -0.1197], + [-0.2921, -0.0974, -0.0699, ..., 0.1084, -0.3307, 0.1156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 2.2817e-08, 2.3283e-09, 8.8476e-09, ..., 3.4459e-08, + 3.7253e-09, 6.5193e-09], + [ 2.5611e-08, 1.8626e-09, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-10, 4.1910e-09], + ..., + [ 5.5879e-09, 1.9558e-08, 4.6566e-10, ..., 6.9849e-09, + 8.5216e-08, 7.0315e-08], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 6.0536e-09, + 0.0000e+00, 9.3132e-10], + [ 6.9849e-09, 3.2596e-09, 9.3132e-10, ..., -5.5879e-09, + 1.8626e-09, -4.6566e-10]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0250, -0.0276, 0.0177, -0.0264, 0.0404, 0.0162, 0.0003, -0.0073, + -0.0286, -0.0179], device='cuda:0'), grad: tensor([ 1.9558e-08, -2.2817e-08, -6.0536e-09, -1.9884e-07, -1.7323e-07, + 9.9652e-08, -1.5693e-07, 2.8405e-07, 1.4435e-07, 1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 220.80, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4765 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.07 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.3247, 0.2409, -0.0422, ..., -0.1501, 0.0126, 0.0075], + [ 0.0107, -0.1697, -0.0320, ..., 0.0517, -0.0734, -0.0932], + [-0.2490, -0.3100, 0.0597, ..., -0.1260, -0.0681, -0.2331], + ..., + [ 0.0828, 0.1989, -0.1051, ..., -0.1721, 0.0900, 0.0985], + [ 0.0306, -0.2220, -0.0627, ..., -0.1534, -0.1315, -0.1198], + [-0.2932, -0.0975, -0.0702, ..., 0.1085, -0.3309, 0.1156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 2.3283e-09], + [ 3.7253e-09, 4.6566e-10, 4.6566e-10, ..., 2.7940e-09, + 1.1176e-08, 1.2573e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.7497e-08, 4.1910e-09], + ..., + [ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., 3.7253e-08, + 6.2399e-08, 5.5414e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.7940e-09, 4.1910e-09], + [ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., -7.9628e-08, + 7.9162e-09, -9.3132e-08]], device='cuda:0') +Epoch 372, bias, value: tensor([ 0.0249, -0.0278, 0.0182, -0.0263, 0.0403, 0.0173, -0.0003, -0.0073, + -0.0288, -0.0180], device='cuda:0'), grad: tensor([ 1.4435e-08, 1.2945e-07, -1.3877e-06, -1.4715e-06, -4.8429e-08, + 1.4901e-06, -9.3598e-08, 1.5255e-06, 2.9802e-08, -1.9046e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 220.66, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4477 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.10 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.3249, 0.2411, -0.0424, ..., -0.1512, 0.0120, 0.0074], + [ 0.0105, -0.1698, -0.0321, ..., 0.0517, -0.0736, -0.0932], + [-0.2497, -0.3102, 0.0596, ..., -0.1271, -0.0693, -0.2363], + ..., + [ 0.0827, 0.1989, -0.1053, ..., -0.1723, 0.0901, 0.0985], + [ 0.0304, -0.2194, -0.0623, ..., -0.1531, -0.1316, -0.1201], + [-0.2987, -0.0975, -0.0702, ..., 0.1074, -0.3319, 0.1156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 1.8626e-09, 1.3970e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 2.3283e-09], + ..., + [-9.3132e-10, -2.3283e-09, 0.0000e+00, ..., 2.3283e-09, + 6.0536e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, -1.8626e-09]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0242, -0.0279, 0.0181, -0.0244, 0.0421, 0.0185, -0.0023, -0.0073, + -0.0286, -0.0188], device='cuda:0'), grad: tensor([ 4.1910e-09, 2.7474e-08, 2.2817e-08, 2.1420e-08, -1.6298e-08, + 6.8452e-08, 1.3970e-08, 2.2352e-08, -2.0768e-07, 5.7742e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 220.74, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4625 re_mapping 0.0028 re_causal 0.0086 /// teacc 99.11 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.3246, 0.2412, -0.0421, ..., -0.1532, 0.0119, 0.0063], + [ 0.0105, -0.1698, -0.0321, ..., 0.0516, -0.0737, -0.0932], + [-0.2499, -0.3101, 0.0596, ..., -0.1275, -0.0697, -0.2367], + ..., + [ 0.0827, 0.1990, -0.1052, ..., -0.1728, 0.0897, 0.0985], + [ 0.0304, -0.2195, -0.0621, ..., -0.1531, -0.1316, -0.1201], + [-0.2990, -0.0975, -0.0703, ..., 0.1075, -0.3323, 0.1157]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 2.2352e-08, + 0.0000e+00, 4.6566e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 9.3132e-10], + [-4.0513e-08, 0.0000e+00, 0.0000e+00, ..., 8.0559e-08, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-08, 9.3132e-10, 0.0000e+00, ..., 1.0710e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0231, -0.0279, 0.0182, -0.0243, 0.0420, 0.0179, -0.0017, -0.0074, + -0.0281, -0.0187], device='cuda:0'), grad: tensor([ 1.7695e-08, 5.8208e-08, 1.2107e-08, 2.1420e-08, -4.4657e-07, + 3.2410e-07, -3.1199e-07, 2.1420e-08, -6.1002e-08, 3.8370e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 220.23, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4738 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.10 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.3248, 0.2419, -0.0418, ..., -0.1536, 0.0118, 0.0064], + [ 0.0102, -0.1701, -0.0318, ..., 0.0516, -0.0739, -0.0933], + [-0.2505, -0.3102, 0.0598, ..., -0.1290, -0.0701, -0.2370], + ..., + [ 0.0831, 0.1991, -0.1055, ..., -0.1731, 0.0897, 0.0985], + [ 0.0304, -0.2197, -0.0616, ..., -0.1537, -0.1317, -0.1203], + [-0.2995, -0.0976, -0.0703, ..., 0.1075, -0.3324, 0.1157]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 2.7940e-09], + [-4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 2.3283e-09, + 9.3132e-10, 5.1223e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 9.3132e-10], + ..., + [-4.6566e-10, -1.3970e-09, 0.0000e+00, ..., 3.0268e-08, + -4.6566e-10, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.8476e-09, + 0.0000e+00, 9.7789e-09], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., -8.2888e-08, + 0.0000e+00, -7.9162e-08]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0232, -0.0280, 0.0183, -0.0246, 0.0421, 0.0177, -0.0016, -0.0074, + -0.0281, -0.0188], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.1176e-08, 4.6566e-09, 4.6566e-09, 8.6613e-08, + 1.5832e-08, 1.8626e-09, 9.4064e-08, 2.9802e-08, -2.5705e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 220.72, cls_loss 0.0004 cls_loss_mapping 0.0008 cls_loss_causal 0.4278 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.11 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.3246, 0.2426, -0.0417, ..., -0.1541, 0.0117, 0.0065], + [ 0.0102, -0.1701, -0.0318, ..., 0.0517, -0.0739, -0.0934], + [-0.2528, -0.3104, 0.0597, ..., -0.1305, -0.0715, -0.2390], + ..., + [ 0.0834, 0.1992, -0.1055, ..., -0.1733, 0.0900, 0.0986], + [ 0.0304, -0.2202, -0.0618, ..., -0.1545, -0.1317, -0.1209], + [-0.2996, -0.0976, -0.0704, ..., 0.1076, -0.3326, 0.1158]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 7.4506e-09], + [ 1.3970e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 3.2596e-09], + ..., + [-6.9849e-09, -1.7229e-08, 0.0000e+00, ..., 1.8626e-09, + -2.7940e-09, -3.2131e-08], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 3.2596e-09], + [ 7.9162e-09, 1.0245e-08, 4.6566e-10, ..., 6.0536e-09, + 1.8626e-09, 1.8626e-08]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0235, -0.0280, 0.0182, -0.0247, 0.0421, 0.0177, -0.0016, -0.0074, + -0.0282, -0.0187], device='cuda:0'), grad: tensor([ 4.6566e-10, 2.1420e-08, 1.0245e-08, 7.9628e-08, -3.8184e-08, + -8.1491e-08, 1.5367e-08, -7.0781e-08, 1.0710e-08, 6.7521e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 220.80, cls_loss 0.0005 cls_loss_mapping 0.0022 cls_loss_causal 0.4578 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.16 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.3247, 0.2435, -0.0418, ..., -0.1551, 0.0118, 0.0067], + [ 0.0101, -0.1704, -0.0318, ..., 0.0517, -0.0743, -0.0935], + [-0.2530, -0.3106, 0.0598, ..., -0.1313, -0.0718, -0.2395], + ..., + [ 0.0834, 0.1992, -0.1055, ..., -0.1734, 0.0891, 0.0986], + [ 0.0305, -0.2179, -0.0618, ..., -0.1547, -0.1302, -0.1201], + [-0.3009, -0.0977, -0.0704, ..., 0.1071, -0.3337, 0.1158]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 2.3283e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, -6.5193e-09], + ..., + [-4.6566e-10, -1.8626e-09, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-09, + 0.0000e+00, 1.3970e-09], + [ 6.9849e-09, 1.3970e-09, 0.0000e+00, ..., 1.3039e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0240, -0.0282, 0.0182, -0.0251, 0.0427, 0.0182, -0.0020, -0.0076, + -0.0258, -0.0190], device='cuda:0'), grad: tensor([ 1.3504e-08, 1.2573e-08, -7.6834e-08, 3.2131e-08, -5.6811e-08, + 1.4342e-07, -1.9325e-07, 5.2620e-08, 4.3306e-08, 5.2620e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 220.92, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4426 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.14 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.3250, 0.2447, -0.0405, ..., -0.1555, 0.0114, 0.0066], + [ 0.0101, -0.1707, -0.0319, ..., 0.0523, -0.0749, -0.0935], + [-0.2535, -0.3116, 0.0597, ..., -0.1327, -0.0725, -0.2415], + ..., + [ 0.0834, 0.1996, -0.1055, ..., -0.1740, 0.0897, 0.0987], + [ 0.0306, -0.2197, -0.0613, ..., -0.1572, -0.1302, -0.1225], + [-0.3016, -0.0977, -0.0704, ..., 0.1071, -0.3353, 0.1159]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -4.6566e-10], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., -1.3970e-09, + 0.0000e+00, -9.3132e-10]], device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0238, -0.0281, 0.0181, -0.0255, 0.0429, 0.0151, 0.0011, -0.0076, + -0.0267, -0.0190], device='cuda:0'), grad: tensor([-1.3970e-09, 1.1642e-08, 2.7940e-09, 2.5053e-07, 5.5879e-09, + -4.5076e-07, 1.2340e-07, 4.1910e-09, 4.0978e-08, 1.7695e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 220.82, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4601 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.23 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.3249, 0.2454, -0.0404, ..., -0.1562, 0.0113, 0.0066], + [ 0.0101, -0.1708, -0.0319, ..., 0.0524, -0.0752, -0.0936], + [-0.2557, -0.3120, 0.0599, ..., -0.1333, -0.0750, -0.2442], + ..., + [ 0.0837, 0.1997, -0.1055, ..., -0.1742, 0.0905, 0.0987], + [ 0.0305, -0.2206, -0.0615, ..., -0.1582, -0.1304, -0.1235], + [-0.3020, -0.0976, -0.0704, ..., 0.1069, -0.3358, 0.1160]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.0431e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.7789e-08], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 4.6566e-10, + 1.3970e-09, 4.6566e-09], + [ 0.0000e+00, 9.3132e-10, -4.6566e-10, ..., 0.0000e+00, + 2.3283e-09, 4.1910e-09], + ..., + [ 0.0000e+00, 1.9092e-08, 4.6566e-10, ..., 4.6566e-10, + 9.3132e-10, 3.7253e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 2.7940e-09, 1.2945e-07, 0.0000e+00, ..., 8.8476e-09, + 0.0000e+00, 2.6077e-08]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0236, -0.0281, 0.0181, -0.0259, 0.0432, 0.0152, 0.0011, -0.0075, + -0.0275, -0.0190], device='cuda:0'), grad: tensor([-8.0466e-07, 2.3749e-08, 5.5879e-09, -2.9802e-08, -2.0955e-08, + 4.2282e-07, 1.3690e-07, 4.5635e-08, 1.8626e-09, 2.2864e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 221.03, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4883 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.17 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.3250, 0.2460, -0.0404, ..., -0.1572, 0.0119, 0.0066], + [ 0.0100, -0.1709, -0.0319, ..., 0.0523, -0.0754, -0.0937], + [-0.2562, -0.3123, 0.0600, ..., -0.1336, -0.0756, -0.2449], + ..., + [ 0.0838, 0.1999, -0.1055, ..., -0.1744, 0.0905, 0.0988], + [ 0.0305, -0.2206, -0.0610, ..., -0.1584, -0.1304, -0.1236], + [-0.3023, -0.0978, -0.0705, ..., 0.1069, -0.3367, 0.1160]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, -3.0268e-08, 0.0000e+00, ..., 2.3283e-09, + 2.7940e-09, 5.3551e-09], + [ 1.1642e-09, 1.8626e-09, 0.0000e+00, ..., 1.3970e-09, + 2.9802e-08, 2.9569e-08], + [ 0.0000e+00, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + 4.9593e-08, 4.3306e-08], + ..., + [ 9.3132e-10, 1.6065e-08, 0.0000e+00, ..., 8.6147e-08, + 3.2596e-08, 1.3504e-07], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 4.1910e-09, 4.1910e-09], + [ 5.3551e-09, -1.6531e-08, 0.0000e+00, ..., -1.1967e-07, + 3.4925e-09, -1.6578e-07]], device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0237, -0.0282, 0.0181, -0.0259, 0.0432, 0.0153, 0.0010, -0.0076, + -0.0269, -0.0191], device='cuda:0'), grad: tensor([-4.0513e-08, 2.3632e-07, 3.6438e-07, -9.8068e-07, 9.4762e-08, + 8.0792e-08, 6.1467e-08, 4.8243e-07, 3.4692e-08, -3.2340e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 220.70, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4918 re_mapping 0.0028 re_causal 0.0086 /// teacc 99.15 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.3250, 0.2466, -0.0405, ..., -0.1578, 0.0120, 0.0065], + [ 0.0099, -0.1710, -0.0319, ..., 0.0522, -0.0754, -0.0937], + [-0.2567, -0.3123, 0.0600, ..., -0.1345, -0.0757, -0.2454], + ..., + [ 0.0837, 0.1999, -0.1057, ..., -0.1748, 0.0905, 0.0988], + [ 0.0307, -0.2207, -0.0609, ..., -0.1585, -0.1304, -0.1237], + [-0.3031, -0.0979, -0.0705, ..., 0.1069, -0.3371, 0.1160]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 6.9849e-09, + 0.0000e+00, 0.0000e+00], + [ 3.9581e-09, 6.5193e-09, 2.3283e-10, ..., -3.2596e-09, + 0.0000e+00, 4.1910e-09], + [ 0.0000e+00, 2.3283e-10, -3.4925e-09, ..., 1.1642e-09, + 0.0000e+00, 0.0000e+00], + ..., + [-3.7253e-09, -6.0536e-09, 3.4925e-09, ..., 2.5611e-09, + 2.3283e-10, -3.4925e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 0.0000e+00], + [ 1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, -1.3970e-09]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0238, -0.0282, 0.0181, -0.0260, 0.0433, 0.0153, 0.0010, -0.0077, + -0.0263, -0.0191], device='cuda:0'), grad: tensor([ 3.0035e-08, 1.1874e-08, -8.3353e-08, 9.1223e-07, 5.1223e-09, + -9.2806e-07, -2.3982e-08, 8.1025e-08, -8.6147e-09, 7.2177e-09], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 380---------------------------------------------------- +epoch 380, time 221.27, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4541 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.27 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.3250, 0.2477, -0.0405, ..., -0.1582, 0.0121, 0.0068], + [ 0.0099, -0.1710, -0.0320, ..., 0.0523, -0.0755, -0.0938], + [-0.2569, -0.3125, 0.0601, ..., -0.1354, -0.0756, -0.2450], + ..., + [ 0.0837, 0.1999, -0.1057, ..., -0.1751, 0.0904, 0.0988], + [ 0.0308, -0.2207, -0.0611, ..., -0.1587, -0.1305, -0.1237], + [-0.3065, -0.0979, -0.0705, ..., 0.1050, -0.3374, 0.1156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 1.8626e-09], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 2.0955e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, -6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 1.1642e-09], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., -6.3330e-08, + 0.0000e+00, -5.1223e-08]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0243, -0.0283, 0.0202, -0.0265, 0.0450, 0.0153, 0.0010, -0.0089, + -0.0260, -0.0202], device='cuda:0'), grad: tensor([ 5.8208e-09, 9.0804e-09, -5.1223e-09, -1.6065e-08, 1.1688e-07, + 2.4913e-08, 0.0000e+00, 2.3283e-10, 4.4238e-09, -1.3853e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 221.06, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4423 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.25 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.3251, 0.2478, -0.0405, ..., -0.1612, 0.0122, 0.0066], + [ 0.0097, -0.1712, -0.0320, ..., 0.0519, -0.0757, -0.0939], + [-0.2587, -0.3134, 0.0591, ..., -0.1362, -0.0769, -0.2477], + ..., + [ 0.0845, 0.2001, -0.1052, ..., -0.1753, 0.0906, 0.0989], + [ 0.0307, -0.2208, -0.0611, ..., -0.1589, -0.1307, -0.1239], + [-0.3066, -0.0980, -0.0706, ..., 0.1052, -0.3378, 0.1157]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1874e-08, 0.0000e+00, ..., 3.2596e-09, + -1.3970e-09, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 3.0268e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 6.9849e-10, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + 0.0000e+00, 1.4901e-08], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 3.8417e-08, + 2.3283e-10, 4.1211e-08], + [ 6.9849e-10, 1.1642e-09, 0.0000e+00, ..., -1.2317e-07, + 2.3283e-10, -1.3830e-07]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0226, -0.0284, 0.0200, -0.0261, 0.0450, 0.0153, 0.0011, -0.0088, + -0.0260, -0.0201], device='cuda:0'), grad: tensor([-1.2340e-08, -4.6566e-10, -1.1176e-08, 1.9092e-08, 2.8405e-08, + 6.2399e-07, -2.8755e-07, 6.3330e-08, 2.1816e-07, -6.2957e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 220.83, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4531 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.3253, 0.2481, -0.0405, ..., -0.1616, 0.0123, 0.0063], + [ 0.0097, -0.1713, -0.0318, ..., 0.0520, -0.0752, -0.0940], + [-0.2590, -0.3134, 0.0590, ..., -0.1365, -0.0770, -0.2480], + ..., + [ 0.0846, 0.2003, -0.1054, ..., -0.1756, 0.0905, 0.0989], + [ 0.0307, -0.2208, -0.0613, ..., -0.1592, -0.1307, -0.1240], + [-0.3062, -0.0981, -0.0707, ..., 0.1081, -0.3381, 0.1177]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.1432e-08, 0.0000e+00, ..., 3.9581e-09, + 0.0000e+00, 1.1642e-09], + [ 6.2864e-09, 1.2340e-08, 2.3283e-10, ..., 2.0256e-08, + 2.3283e-10, 5.8208e-09], + [ 1.1642e-09, -1.2154e-07, 0.0000e+00, ..., 3.4925e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 3.4925e-08, 1.7462e-08, -9.3132e-10, ..., 1.1316e-07, + 8.3819e-09, 2.0722e-08], + [ 9.3132e-10, 3.4925e-09, 0.0000e+00, ..., 2.7940e-09, + 2.3283e-10, 9.3132e-10], + [ 3.7681e-06, 8.1491e-09, 2.3283e-10, ..., 1.0505e-05, + 6.9849e-10, 4.5053e-07]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0225, -0.0284, 0.0200, -0.0264, 0.0418, 0.0153, 0.0011, -0.0087, + -0.0261, -0.0175], device='cuda:0'), grad: tensor([ 1.5972e-07, 1.3364e-07, -5.4808e-07, 2.1607e-07, -4.1634e-05, + -1.0710e-08, 2.5611e-08, 5.2433e-07, 2.8405e-08, 4.1127e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 220.54, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4703 re_mapping 0.0028 re_causal 0.0085 /// teacc 99.14 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.3254, 0.2487, -0.0406, ..., -0.1619, 0.0122, 0.0061], + [ 0.0096, -0.1714, -0.0319, ..., 0.0518, -0.0754, -0.0940], + [-0.2595, -0.3136, 0.0590, ..., -0.1381, -0.0776, -0.2487], + ..., + [ 0.0847, 0.2003, -0.1053, ..., -0.1761, 0.0907, 0.0989], + [ 0.0307, -0.2209, -0.0614, ..., -0.1594, -0.1308, -0.1241], + [-0.3070, -0.0981, -0.0707, ..., 0.1079, -0.3408, 0.1178]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 6.9849e-10], + [ 3.2596e-09, 0.0000e+00, 4.6566e-10, ..., 5.1223e-09, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, + 6.9849e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., -2.0955e-09, + 0.0000e+00, -2.7940e-09]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0227, -0.0285, 0.0200, -0.0272, 0.0421, 0.0155, 0.0010, -0.0088, + -0.0262, -0.0177], device='cuda:0'), grad: tensor([ 1.6298e-09, 1.0245e-08, 4.6566e-09, 1.8626e-09, -1.2806e-08, + 1.1642e-09, 1.1642e-09, 1.0012e-08, 1.1642e-09, -4.1910e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 220.85, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4517 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.15 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.3259, 0.2493, -0.0396, ..., -0.1638, 0.0122, 0.0036], + [ 0.0090, -0.1720, -0.0318, ..., 0.0517, -0.0759, -0.0943], + [-0.2597, -0.3139, 0.0589, ..., -0.1397, -0.0780, -0.2489], + ..., + [ 0.0854, 0.2007, -0.1053, ..., -0.1770, 0.0910, 0.0989], + [ 0.0313, -0.2210, -0.0621, ..., -0.1584, -0.1309, -0.1236], + [-0.3074, -0.0980, -0.0710, ..., 0.1081, -0.3413, 0.1180]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 9.3132e-10], + [-2.1420e-08, 9.3132e-10, 0.0000e+00, ..., -1.1548e-07, + 7.9162e-09, 3.0268e-08], + [ 2.5611e-09, 4.6566e-10, 0.0000e+00, ..., 1.4901e-08, + 0.0000e+00, 6.9849e-10], + ..., + [ 1.2107e-08, -1.1642e-09, 0.0000e+00, ..., 8.8243e-08, + -1.1176e-08, -6.2864e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 6.9849e-10], + [ 4.6566e-09, 6.9849e-10, 0.0000e+00, ..., -1.0477e-08, + 3.0268e-09, -4.2142e-08]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0217, -0.0287, 0.0200, -0.0276, 0.0421, 0.0156, 0.0010, -0.0088, + -0.0249, -0.0177], device='cuda:0'), grad: tensor([ 5.1223e-09, -3.8301e-07, 5.4948e-08, 1.1642e-08, 6.9384e-08, + 5.3551e-09, 4.8894e-09, 2.7381e-07, 3.2596e-09, -3.6089e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 220.68, cls_loss 0.0004 cls_loss_mapping 0.0008 cls_loss_causal 0.4416 re_mapping 0.0030 re_causal 0.0085 /// teacc 99.20 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.3260, 0.2495, -0.0397, ..., -0.1645, 0.0121, 0.0035], + [ 0.0089, -0.1722, -0.0318, ..., 0.0516, -0.0761, -0.0944], + [-0.2598, -0.3140, 0.0588, ..., -0.1410, -0.0780, -0.2491], + ..., + [ 0.0856, 0.2008, -0.1053, ..., -0.1773, 0.0909, 0.0989], + [ 0.0313, -0.2210, -0.0622, ..., -0.1586, -0.1309, -0.1236], + [-0.3075, -0.0980, -0.0710, ..., 0.1082, -0.3415, 0.1181]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 4.6566e-10], + [ 2.3562e-07, 1.0245e-07, 0.0000e+00, ..., 1.1176e-08, + 8.7311e-08, 7.9861e-08], + [ 7.4506e-09, 3.2596e-09, 0.0000e+00, ..., 6.9849e-10, + 3.2596e-09, 2.7940e-09], + ..., + [-2.7195e-07, -1.1805e-07, 0.0000e+00, ..., -1.3737e-08, + -9.9884e-08, -9.2201e-08], + [ 6.5193e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-10, + 1.3970e-09, 1.3970e-09], + [ 5.3551e-09, 2.3283e-09, 0.0000e+00, ..., -3.0268e-09, + 1.8626e-09, -4.6566e-10]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0208, -0.0286, 0.0200, -0.0242, 0.0419, 0.0135, 0.0012, -0.0089, + -0.0250, -0.0176], device='cuda:0'), grad: tensor([ 1.4435e-08, 8.8755e-07, 3.2596e-08, 4.5635e-08, 8.6147e-08, + -6.4168e-07, 5.3644e-07, -1.0114e-06, 4.1910e-08, 1.4901e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 221.31, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4387 re_mapping 0.0030 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.3261, 0.2486, -0.0373, ..., -0.1668, 0.0120, 0.0003], + [ 0.0087, -0.1723, -0.0318, ..., 0.0515, -0.0765, -0.0945], + [-0.2600, -0.3151, 0.0584, ..., -0.1431, -0.0782, -0.2492], + ..., + [ 0.0859, 0.2010, -0.1052, ..., -0.1783, 0.0910, 0.0989], + [ 0.0313, -0.2210, -0.0623, ..., -0.1593, -0.1311, -0.1238], + [-0.3088, -0.0974, -0.0712, ..., 0.1093, -0.3419, 0.1191]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0955e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.0955e-09, 0.0000e+00, ..., -3.0268e-09, + 0.0000e+00, -1.3970e-09]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0182, -0.0287, 0.0199, -0.0241, 0.0408, 0.0134, 0.0012, -0.0090, + -0.0252, -0.0163], device='cuda:0'), grad: tensor([-3.2596e-09, 1.3970e-09, 4.6566e-10, 1.8626e-09, 7.2177e-09, + 6.5193e-09, 1.3970e-09, 4.1910e-09, -6.9849e-09, -1.1642e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 220.42, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4361 re_mapping 0.0031 re_causal 0.0087 /// teacc 99.20 lr 0.00010000 +Epoch 389, weight, value: tensor([[-3.2621e-01, 2.4921e-01, -3.7195e-02, ..., -1.6714e-01, + 1.1867e-02, 3.4114e-04], + [ 8.1562e-03, -1.7257e-01, -3.1990e-02, ..., 5.1401e-02, + -7.6636e-02, -9.4635e-02], + [-2.6025e-01, -3.1541e-01, 5.8098e-02, ..., -1.4545e-01, + -7.8299e-02, -2.4936e-01], + ..., + [ 8.6180e-02, 2.0165e-01, -1.0525e-01, ..., -1.7855e-01, + 9.1116e-02, 9.9101e-02], + [ 3.1259e-02, -2.2110e-01, -6.2353e-02, ..., -1.6102e-01, + -1.3121e-01, -1.2470e-01], + [-3.0886e-01, -9.8109e-02, -7.1258e-02, ..., 1.0959e-01, + -3.4212e-01, 1.1916e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.9581e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 6.9849e-10], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -2.3283e-10, + 2.7940e-09, 8.1491e-09], + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 4.6566e-09, 1.1642e-08], + ..., + [-5.1688e-08, 0.0000e+00, 0.0000e+00, ..., -1.6764e-08, + -1.1874e-08, -2.0233e-07], + [ 3.4925e-09, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 2.2585e-08, 5.3551e-08], + [ 4.1444e-08, 6.9849e-10, 0.0000e+00, ..., 1.4901e-08, + 1.6531e-08, 1.8417e-07]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0183, -0.0289, 0.0199, -0.0241, 0.0406, 0.0135, 0.0012, -0.0087, + -0.0256, -0.0162], device='cuda:0'), grad: tensor([-3.0268e-09, 5.5181e-08, 1.0105e-07, -7.2271e-07, 3.4459e-08, + 1.2945e-07, 1.0943e-08, -7.6974e-07, 3.9348e-07, 7.7346e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 220.87, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4666 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.23 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.3263, 0.2497, -0.0373, ..., -0.1668, 0.0119, 0.0004], + [ 0.0075, -0.1755, -0.0321, ..., 0.0515, -0.0770, -0.0969], + [-0.2606, -0.3152, 0.0579, ..., -0.1460, -0.0784, -0.2496], + ..., + [ 0.0864, 0.2044, -0.1052, ..., -0.1793, 0.0913, 0.1001], + [ 0.0313, -0.2212, -0.0631, ..., -0.1612, -0.1313, -0.1248], + [-0.3090, -0.0983, -0.0714, ..., 0.1097, -0.3423, 0.1193]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.8894e-09, 1.0943e-08, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-09, 2.0955e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + ..., + [-4.6566e-09, -1.1176e-08, 0.0000e+00, ..., 2.3283e-10, + -7.2177e-09, -2.1420e-08], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 2.0955e-09]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0184, -0.0320, 0.0199, -0.0242, 0.0406, 0.0136, 0.0011, -0.0061, + -0.0254, -0.0161], device='cuda:0'), grad: tensor([ 2.5611e-09, 7.6136e-08, -1.7695e-08, 3.4925e-08, 1.6298e-09, + -3.6019e-07, 3.1688e-07, -6.2631e-08, 9.5461e-09, 8.3819e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 221.10, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4689 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.18 lr 0.00010000 +Epoch 391, weight, value: tensor([[-3.2637e-01, 2.5159e-01, -3.4897e-02, ..., -1.6601e-01, + 1.2238e-02, 2.1787e-04], + [ 7.4879e-03, -1.7553e-01, -3.2067e-02, ..., 5.1556e-02, + -7.7308e-02, -9.7041e-02], + [-2.6083e-01, -3.1624e-01, 5.7948e-02, ..., -1.4720e-01, + -7.8488e-02, -2.5014e-01], + ..., + [ 8.6529e-02, 2.0447e-01, -1.0520e-01, ..., -1.7965e-01, + 9.1042e-02, 1.0012e-01], + [ 3.1211e-02, -2.2155e-01, -6.4297e-02, ..., -1.6158e-01, + -1.3147e-01, -1.2494e-01], + [-3.0920e-01, -9.8442e-02, -7.1565e-02, ..., 1.0980e-01, + -3.4285e-01, 1.1935e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0245e-08, 0.0000e+00, ..., 1.3970e-09, + -2.0955e-09, 1.8626e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 1.6298e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 9.5461e-09, 0.0000e+00, ..., 2.0955e-09, + 2.0955e-09, 3.0268e-09], + ..., + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.5146e-08, + 0.0000e+00, 3.1199e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-08, + 0.0000e+00, 1.8394e-08], + [ 4.8894e-09, 6.9849e-10, 2.3283e-10, ..., -4.5868e-08, + 0.0000e+00, -6.6357e-08]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0193, -0.0320, 0.0199, -0.0242, 0.0405, 0.0137, 0.0009, -0.0060, + -0.0255, -0.0161], device='cuda:0'), grad: tensor([-1.2107e-08, -7.4506e-09, 5.0291e-08, 7.6834e-09, 1.5367e-08, + 1.6764e-08, 5.3551e-09, 1.5041e-07, 9.5461e-08, -3.0850e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 221.13, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4606 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.18 lr 0.00010000 +Epoch 392, weight, value: tensor([[-3.2658e-01, 2.5226e-01, -3.4654e-02, ..., -1.6637e-01, + 1.2370e-02, 1.9492e-04], + [ 7.8969e-03, -1.7555e-01, -3.1046e-02, ..., 5.2072e-02, + -7.5672e-02, -9.7093e-02], + [-2.6111e-01, -3.1737e-01, 5.8194e-02, ..., -1.4776e-01, + -7.8626e-02, -2.5043e-01], + ..., + [ 8.6084e-02, 2.0451e-01, -1.0630e-01, ..., -1.8019e-01, + 8.8644e-02, 1.0010e-01], + [ 3.1564e-02, -2.2188e-01, -6.5169e-02, ..., -1.6216e-01, + -1.3172e-01, -1.2527e-01], + [-3.0938e-01, -9.8540e-02, -7.1752e-02, ..., 1.0982e-01, + -3.4387e-01, 1.1939e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 3.9581e-09, + 2.3283e-10, 3.4925e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.9581e-09, + 1.3970e-09, 7.6834e-09], + [ 2.3283e-10, -2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 9.3132e-10, 1.1642e-09], + ..., + [ 2.0955e-09, 2.3283e-10, 0.0000e+00, ..., 6.6822e-08, + 7.4506e-09, 6.5425e-08], + [ 9.0804e-09, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 4.6566e-10, 8.1491e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -4.4308e-07, + 9.3132e-10, -4.0559e-07]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0192, -0.0320, 0.0199, -0.0241, 0.0405, 0.0135, 0.0013, -0.0061, + -0.0255, -0.0161], device='cuda:0'), grad: tensor([ 2.9569e-08, 3.0268e-08, -1.9372e-07, 2.1188e-08, 1.0924e-06, + -3.4948e-07, 2.3074e-07, 2.6426e-07, 8.1491e-08, -1.1986e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 221.22, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4850 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.13 lr 0.00010000 +Epoch 393, weight, value: tensor([[-3.2665e-01, 2.5262e-01, -3.4589e-02, ..., -1.6692e-01, + 1.2508e-02, 1.5477e-04], + [ 8.0086e-03, -1.7555e-01, -3.1074e-02, ..., 5.2078e-02, + -7.5691e-02, -9.7151e-02], + [-2.6195e-01, -3.1889e-01, 5.8233e-02, ..., -1.4833e-01, + -7.8858e-02, -2.5079e-01], + ..., + [ 8.6125e-02, 2.0461e-01, -1.0637e-01, ..., -1.8055e-01, + 8.8746e-02, 1.0015e-01], + [ 3.1463e-02, -2.2174e-01, -6.7624e-02, ..., -1.6271e-01, + -1.3189e-01, -1.2552e-01], + [-3.0949e-01, -9.8734e-02, -7.2050e-02, ..., 1.1005e-01, + -3.4494e-01, 1.1953e-01]], device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.9325e-08, 0.0000e+00, ..., 6.9849e-10, + -1.6298e-09, 0.0000e+00], + [ 1.3271e-08, 4.6566e-10, 0.0000e+00, ..., -4.6566e-09, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 1.1642e-08, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-09, 2.3283e-10], + ..., + [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 4.8894e-09, + 2.3283e-10, 2.5611e-09], + [ 3.4925e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, + 2.3283e-10, 2.3283e-10], + [ 4.1910e-09, 4.4238e-09, 0.0000e+00, ..., -9.3132e-10, + 4.6566e-10, -3.0268e-09]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0190, -0.0320, 0.0198, -0.0242, 0.0403, 0.0160, -0.0012, -0.0062, + -0.0259, -0.0159], device='cuda:0'), grad: tensor([-3.6322e-08, 7.9162e-08, 5.7509e-08, 1.8720e-07, 4.6566e-10, + -3.3411e-07, 5.5879e-09, 2.8871e-08, -6.9849e-09, 3.4692e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 220.99, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4527 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +Epoch 394, weight, value: tensor([[-3.2672e-01, 2.5290e-01, -3.4613e-02, ..., -1.6707e-01, + 1.2451e-02, 1.2952e-04], + [ 8.8908e-03, -1.7557e-01, -2.9968e-02, ..., 5.3933e-02, + -7.4679e-02, -9.7210e-02], + [-2.6216e-01, -3.1926e-01, 5.8244e-02, ..., -1.5008e-01, + -7.9291e-02, -2.5129e-01], + ..., + [ 8.5155e-02, 2.0466e-01, -1.0755e-01, ..., -1.8169e-01, + 8.8198e-02, 1.0020e-01], + [ 3.1464e-02, -2.2169e-01, -6.7881e-02, ..., -1.6294e-01, + -1.3251e-01, -1.2573e-01], + [-3.0957e-01, -9.8817e-02, -7.2304e-02, ..., 1.1005e-01, + -3.4542e-01, 1.1954e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.6298e-09, 4.6566e-10, 0.0000e+00, ..., -6.9849e-10, + 6.9849e-10, 3.9581e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 6.9849e-10], + ..., + [-2.5611e-09, -2.3283e-09, 0.0000e+00, ..., -2.3283e-10, + -1.6298e-09, -8.8476e-09], + [-8.0559e-08, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 6.9849e-10], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., -1.1642e-09, + 6.9849e-10, 3.2596e-09]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0188, -0.0318, 0.0198, -0.0240, 0.0403, 0.0160, -0.0013, -0.0063, + -0.0258, -0.0159], device='cuda:0'), grad: tensor([ 3.0268e-09, 2.8056e-07, -3.1223e-07, 5.3551e-09, 6.5193e-09, + 1.1334e-06, 2.0140e-07, -1.1642e-08, -1.3206e-06, 2.5379e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 221.04, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4665 re_mapping 0.0028 re_causal 0.0083 /// teacc 99.10 lr 0.00010000 +Epoch 395, weight, value: tensor([[-3.2727e-01, 2.5338e-01, -3.4351e-02, ..., -1.6747e-01, + 1.2515e-02, 1.1239e-05], + [ 6.9107e-03, -1.7591e-01, -2.9999e-02, ..., 5.2757e-02, + -7.4827e-02, -9.7793e-02], + [-2.6253e-01, -3.1979e-01, 5.8320e-02, ..., -1.5070e-01, + -7.9300e-02, -2.5170e-01], + ..., + [ 8.7294e-02, 2.0504e-01, -1.0758e-01, ..., -1.8307e-01, + 8.8177e-02, 1.0033e-01], + [ 3.1384e-02, -2.2202e-01, -6.7927e-02, ..., -1.6324e-01, + -1.3265e-01, -1.2597e-01], + [-3.0968e-01, -9.8879e-02, -7.2511e-02, ..., 1.1083e-01, + -3.4556e-01, 1.2015e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 6.9849e-10], + [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 4.4238e-09, + 6.9849e-10, 3.0268e-09], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, + 4.6566e-10, 6.9849e-10], + ..., + [ 7.2177e-09, 0.0000e+00, -6.9849e-10, ..., 2.3749e-08, + 0.0000e+00, 7.9162e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-09, + 2.3283e-10, 2.7940e-09], + [ 1.5111e-07, 4.6566e-10, 2.3283e-10, ..., 2.9756e-07, + 0.0000e+00, -5.5181e-08]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0187, -0.0320, 0.0198, -0.0244, 0.0398, 0.0161, -0.0012, -0.0062, + -0.0260, -0.0153], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.3737e-08, 1.2340e-08, 2.1886e-08, -1.0366e-06, + -2.3283e-08, 2.0955e-08, 8.7079e-08, 1.6065e-08, 9.1037e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 220.67, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4553 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +Epoch 396, weight, value: tensor([[-3.2747e-01, 2.5416e-01, -3.4166e-02, ..., -1.6773e-01, + 1.2524e-02, 1.5293e-05], + [ 6.9355e-03, -1.7592e-01, -3.0016e-02, ..., 5.2739e-02, + -7.4982e-02, -9.7864e-02], + [-2.6326e-01, -3.1998e-01, 5.8275e-02, ..., -1.5125e-01, + -7.9484e-02, -2.5192e-01], + ..., + [ 8.7255e-02, 2.0503e-01, -1.0756e-01, ..., -1.8367e-01, + 8.8169e-02, 1.0031e-01], + [ 3.1302e-02, -2.2210e-01, -6.8126e-02, ..., -1.6343e-01, + -1.3274e-01, -1.2609e-01], + [-3.1039e-01, -9.8977e-02, -7.2615e-02, ..., 1.1079e-01, + -3.4566e-01, 1.2027e-01]], device='cuda:0'), grad: tensor([[ 1.6298e-09, -2.1188e-08, 0.0000e+00, ..., 2.0489e-08, + -2.3283e-09, -2.0955e-09], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 2.5611e-09, 0.0000e+00, ..., 2.3283e-09, + 2.3283e-10, 9.3132e-10], + [ 9.3132e-10, 1.1642e-09, 0.0000e+00, ..., 1.2107e-08, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 9.0804e-09, 0.0000e+00, ..., -9.0804e-09, + 1.1642e-09, -1.3970e-09]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0188, -0.0319, 0.0196, -0.0244, 0.0399, 0.0161, -0.0012, -0.0063, + -0.0260, -0.0153], device='cuda:0'), grad: tensor([ 9.5228e-08, 1.6531e-08, 4.8894e-09, 3.8417e-08, 2.3050e-08, + 1.2852e-07, -3.8953e-07, 1.3504e-08, 7.8464e-08, -8.1491e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 220.67, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4661 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.16 lr 0.00010000 +Epoch 397, weight, value: tensor([[-3.2765e-01, 2.5432e-01, -3.4236e-02, ..., -1.6823e-01, + 1.2528e-02, -6.4569e-06], + [ 6.7951e-03, -1.7599e-01, -3.0007e-02, ..., 5.2858e-02, + -7.5879e-02, -9.7979e-02], + [-2.6346e-01, -3.2011e-01, 5.8062e-02, ..., -1.5233e-01, + -7.9610e-02, -2.5215e-01], + ..., + [ 8.7428e-02, 2.0512e-01, -1.0752e-01, ..., -1.8393e-01, + 8.8783e-02, 1.0038e-01], + [ 3.0438e-02, -2.2095e-01, -6.8374e-02, ..., -1.6373e-01, + -1.3284e-01, -1.2685e-01], + [-3.1059e-01, -9.9053e-02, -7.2768e-02, ..., 1.1076e-01, + -3.4605e-01, 1.2029e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 2.5611e-09, 0.0000e+00, ..., -1.1642e-09, + 2.7940e-09, 6.0536e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-09, 2.3283e-09], + ..., + [ 0.0000e+00, -7.4506e-09, 0.0000e+00, ..., 1.1642e-09, + -4.6566e-09, -1.2573e-08], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., -4.6566e-10, + 1.3970e-09, 3.2596e-09], + [ 6.9849e-10, 2.3283e-09, 0.0000e+00, ..., 1.8626e-09, + 3.0268e-09, 5.8208e-09]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0185, -0.0320, 0.0202, -0.0258, 0.0399, 0.0167, -0.0012, -0.0064, + -0.0266, -0.0154], device='cuda:0'), grad: tensor([ 1.6298e-09, 5.5879e-09, 1.0710e-08, -5.4715e-08, -2.5611e-09, + 5.1456e-08, 6.5193e-09, -3.8184e-08, 1.1642e-09, 3.1199e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 220.29, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4541 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.07 lr 0.00010000 +Epoch 398, weight, value: tensor([[-3.2785e-01, 2.5585e-01, -3.4245e-02, ..., -1.6812e-01, + 1.2542e-02, -3.8356e-05], + [ 6.9530e-03, -1.7600e-01, -2.9907e-02, ..., 5.2741e-02, + -7.5931e-02, -9.7979e-02], + [-2.6364e-01, -3.2022e-01, 5.8044e-02, ..., -1.5312e-01, + -7.9469e-02, -2.5240e-01], + ..., + [ 8.7477e-02, 2.0516e-01, -1.0764e-01, ..., -1.8424e-01, + 8.8920e-02, 1.0040e-01], + [ 3.0134e-02, -2.2122e-01, -6.8428e-02, ..., -1.6392e-01, + -1.3305e-01, -1.2698e-01], + [-3.1067e-01, -9.9328e-02, -7.2846e-02, ..., 1.1080e-01, + -3.4698e-01, 1.2031e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + [ 2.0722e-08, 1.1176e-08, 9.3132e-10, ..., 0.0000e+00, + 2.7474e-08, 9.1968e-08], + [ 1.3970e-09, 9.3132e-10, -4.4238e-09, ..., 0.0000e+00, + 1.0245e-08, 3.3295e-08], + ..., + [-2.7940e-08, -1.2107e-08, -1.1176e-08, ..., 0.0000e+00, + -6.1467e-08, -2.1374e-07], + [ 1.1642e-09, -1.3970e-09, 4.6566e-10, ..., -1.8626e-09, + 3.2596e-09, 1.0477e-08], + [ 1.3970e-09, 1.3970e-09, 9.3132e-10, ..., 9.3132e-10, + 1.5134e-08, 4.8662e-08]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0190, -0.0319, 0.0202, -0.0263, 0.0399, 0.0168, -0.0010, -0.0064, + -0.0270, -0.0154], device='cuda:0'), grad: tensor([ 2.3283e-09, 5.1688e-07, 1.0501e-07, 8.9873e-08, 4.9826e-08, + 1.8626e-08, 2.3749e-08, -1.0561e-06, 1.3271e-08, 2.4866e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 220.40, cls_loss 0.0004 cls_loss_mapping 0.0015 cls_loss_causal 0.4688 re_mapping 0.0028 re_causal 0.0086 /// teacc 99.14 lr 0.00010000 +Epoch 399, weight, value: tensor([[-3.2787e-01, 2.5719e-01, -3.2904e-02, ..., -1.6826e-01, + 1.2585e-02, -5.0085e-05], + [ 6.9720e-03, -1.7602e-01, -2.9912e-02, ..., 5.2741e-02, + -7.6043e-02, -9.8027e-02], + [-2.6415e-01, -3.2061e-01, 5.8173e-02, ..., -1.5346e-01, + -7.9035e-02, -2.5245e-01], + ..., + [ 8.7615e-02, 2.0520e-01, -1.0764e-01, ..., -1.8430e-01, + 8.8983e-02, 1.0044e-01], + [ 2.9960e-02, -2.2007e-01, -6.8444e-02, ..., -1.6411e-01, + -1.3317e-01, -1.2712e-01], + [-3.1099e-01, -9.9464e-02, -7.2955e-02, ..., 1.1071e-01, + -3.4719e-01, 1.2032e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.6298e-09, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 2.3283e-10, 1.8626e-09, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, -6.9849e-10, -4.2375e-08, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3283e-10, 1.8859e-08, ..., 3.4925e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.5611e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 6.9849e-10, 2.3283e-09, ..., -3.7020e-08, + 0.0000e+00, -1.8161e-08]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0188, -0.0319, 0.0203, -0.0264, 0.0400, 0.0169, -0.0012, -0.0064, + -0.0264, -0.0155], device='cuda:0'), grad: tensor([ 9.0804e-09, 1.2340e-08, -1.9255e-07, 7.5903e-08, 6.4960e-08, + 5.8208e-09, 5.5879e-09, 9.2434e-08, 8.6147e-09, -7.3807e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 220.74, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4836 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.07 lr 0.00010000 +Epoch 400, weight, value: tensor([[-3.2812e-01, 2.5816e-01, -3.2531e-02, ..., -1.6906e-01, + 1.2653e-02, -1.1100e-04], + [ 6.6018e-03, -1.7606e-01, -2.9961e-02, ..., 5.2229e-02, + -7.6214e-02, -9.8147e-02], + [-2.6521e-01, -3.2090e-01, 5.8321e-02, ..., -1.5393e-01, + -7.9760e-02, -2.5293e-01], + ..., + [ 8.7919e-02, 2.0525e-01, -1.0761e-01, ..., -1.8447e-01, + 8.9039e-02, 1.0050e-01], + [ 2.9332e-02, -2.2028e-01, -6.8678e-02, ..., -1.6473e-01, + -1.3331e-01, -1.2737e-01], + [-3.1173e-01, -9.9545e-02, -7.3022e-02, ..., 1.1058e-01, + -3.4750e-01, 1.2040e-01]], device='cuda:0'), grad: tensor([[ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 2.3283e-10], + [ 4.1910e-09, -4.6566e-10, 0.0000e+00, ..., -1.2340e-08, + 1.3970e-09, 6.9849e-10], + [ 3.7253e-09, 2.3283e-10, 0.0000e+00, ..., 1.6298e-09, + 1.3970e-09, 6.9849e-10], + ..., + [ 1.5367e-08, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 4.8894e-09, 1.3970e-09], + [-5.0059e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.3039e-08, 2.3283e-10], + [ 9.5461e-09, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 2.7940e-09, -1.1642e-09]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0189, -0.0320, 0.0204, -0.0271, 0.0402, 0.0168, -0.0012, -0.0064, + -0.0256, -0.0157], device='cuda:0'), grad: tensor([ 3.4226e-08, 1.8161e-08, -6.0536e-09, 1.5972e-07, 1.9791e-08, + 7.4506e-08, 7.1013e-08, 2.8964e-07, -8.2189e-07, 1.6950e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 220.31, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4533 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 401, weight, value: tensor([[-3.2819e-01, 2.5927e-01, -3.2768e-02, ..., -1.7018e-01, + 1.2589e-02, -1.5251e-04], + [ 6.4919e-03, -1.7609e-01, -3.0117e-02, ..., 5.2070e-02, + -7.6484e-02, -9.8227e-02], + [-2.6576e-01, -3.2157e-01, 5.8268e-02, ..., -1.5466e-01, + -8.0278e-02, -2.5327e-01], + ..., + [ 8.8023e-02, 2.0528e-01, -1.0760e-01, ..., -1.8475e-01, + 8.9326e-02, 1.0055e-01], + [ 2.9324e-02, -2.2076e-01, -6.9715e-02, ..., -1.6505e-01, + -1.3344e-01, -1.2746e-01], + [-3.1210e-01, -9.9774e-02, -7.3121e-02, ..., 1.1058e-01, + -3.4882e-01, 1.2042e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 1.3039e-08, 2.5611e-09, 0.0000e+00, ..., 1.5832e-08, + 8.8476e-09, 4.6566e-08], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 1.6298e-09], + ..., + [-2.6310e-08, -5.3551e-09, 0.0000e+00, ..., -2.9337e-08, + -1.7695e-08, -9.3132e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 1.6298e-09], + [ 1.1642e-08, 2.7940e-09, 0.0000e+00, ..., 1.3039e-08, + 7.6834e-09, 4.0978e-08]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0191, -0.0320, 0.0204, -0.0283, 0.0402, 0.0166, -0.0011, -0.0064, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([ 2.0955e-09, 1.2824e-06, -5.3793e-06, 4.4238e-09, 1.0710e-08, + 1.7928e-08, -6.5193e-09, 3.8929e-06, 1.1642e-08, 1.5297e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 220.58, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4333 re_mapping 0.0028 re_causal 0.0085 /// teacc 99.11 lr 0.00001000 +Epoch 402, weight, value: tensor([[-3.2820e-01, 2.5931e-01, -3.2770e-02, ..., -1.7020e-01, + 1.2589e-02, -1.5411e-04], + [ 6.4736e-03, -1.7609e-01, -3.0125e-02, ..., 5.2032e-02, + -7.6497e-02, -9.8234e-02], + [-2.6578e-01, -3.2159e-01, 5.8268e-02, ..., -1.5471e-01, + -8.0288e-02, -2.5329e-01], + ..., + [ 8.8031e-02, 2.0529e-01, -1.0760e-01, ..., -1.8479e-01, + 8.9329e-02, 1.0055e-01], + [ 2.9331e-02, -2.2075e-01, -6.9733e-02, ..., -1.6507e-01, + -1.3345e-01, -1.2747e-01], + [-3.1212e-01, -9.9779e-02, -7.3123e-02, ..., 1.1058e-01, + -3.4884e-01, 1.2043e-01]], device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.6298e-09, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 3.7951e-08, 6.2864e-09, 0.0000e+00, ..., 5.9837e-08, + 0.0000e+00, 7.4506e-09], + [ 4.6566e-10, 1.6298e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + ..., + [ 3.9581e-09, -1.2806e-08, 0.0000e+00, ..., 1.1409e-08, + 4.6566e-10, -7.6834e-09], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 2.3283e-10], + [ 7.7765e-08, 6.0536e-09, 0.0000e+00, ..., 1.0058e-07, + -9.3132e-10, -1.3504e-08]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0191, -0.0322, 0.0208, -0.0283, 0.0402, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([ 4.6566e-10, 1.7858e-07, 7.2177e-09, 3.7253e-09, -4.4587e-07, + -1.7229e-08, 3.9581e-09, -8.6147e-09, 6.2864e-09, 2.8033e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 221.26, cls_loss 0.0004 cls_loss_mapping 0.0006 cls_loss_causal 0.4398 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.11 lr 0.00001000 +Epoch 403, weight, value: tensor([[-3.2820e-01, 2.5936e-01, -3.2768e-02, ..., -1.7024e-01, + 1.2586e-02, -1.6123e-04], + [ 6.4410e-03, -1.7609e-01, -3.0138e-02, ..., 5.1989e-02, + -7.6528e-02, -9.8243e-02], + [-2.6580e-01, -3.2162e-01, 5.8306e-02, ..., -1.5477e-01, + -8.0299e-02, -2.5330e-01], + ..., + [ 8.8040e-02, 2.0529e-01, -1.0761e-01, ..., -1.8481e-01, + 8.9344e-02, 1.0055e-01], + [ 2.9337e-02, -2.2078e-01, -6.9735e-02, ..., -1.6509e-01, + -1.3346e-01, -1.2748e-01], + [-3.1214e-01, -9.9785e-02, -7.3128e-02, ..., 1.1058e-01, + -3.4887e-01, 1.2044e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0955e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 2.5611e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.0955e-09, 0.0000e+00, 0.0000e+00, ..., 3.9581e-09, + 2.3283e-10, 4.8894e-09], + [-1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., -8.1491e-09, + 0.0000e+00, -9.3132e-09]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0191, -0.0322, 0.0209, -0.0283, 0.0402, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([-3.4925e-09, 8.6147e-09, 2.0955e-09, -3.0268e-09, 7.2177e-09, + 4.4238e-09, 4.1910e-09, 3.1432e-08, -1.2107e-08, -2.7241e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 220.76, cls_loss 0.0003 cls_loss_mapping 0.0005 cls_loss_causal 0.4358 re_mapping 0.0026 re_causal 0.0081 /// teacc 99.12 lr 0.00001000 +Epoch 404, weight, value: tensor([[-3.2820e-01, 2.5942e-01, -3.2770e-02, ..., -1.7027e-01, + 1.2586e-02, -1.6134e-04], + [ 6.4243e-03, -1.7609e-01, -3.0179e-02, ..., 5.1971e-02, + -7.6539e-02, -9.8253e-02], + [-2.6583e-01, -3.2164e-01, 5.8311e-02, ..., -1.5483e-01, + -8.0306e-02, -2.5331e-01], + ..., + [ 8.8044e-02, 2.0529e-01, -1.0762e-01, ..., -1.8483e-01, + 8.9344e-02, 1.0055e-01], + [ 2.9338e-02, -2.2080e-01, -6.9790e-02, ..., -1.6511e-01, + -1.3346e-01, -1.2749e-01], + [-3.1216e-01, -9.9791e-02, -7.3139e-02, ..., 1.1058e-01, + -3.4887e-01, 1.2044e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-6.9849e-10, 0.0000e+00, -2.3283e-10, ..., -2.7940e-09, + 2.3283e-10, 2.3283e-10], + [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 6.9849e-10, 6.9849e-10], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([ 4.6566e-10, 2.2352e-08, -3.2363e-08, -1.0245e-08, -9.7789e-09, + 4.6566e-09, 2.5146e-08, 1.0477e-08, -1.2806e-08, 9.5461e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 220.91, cls_loss 0.0003 cls_loss_mapping 0.0006 cls_loss_causal 0.4339 re_mapping 0.0025 re_causal 0.0079 /// teacc 99.13 lr 0.00001000 +Epoch 405, weight, value: tensor([[-3.2821e-01, 2.5948e-01, -3.2771e-02, ..., -1.7029e-01, + 1.2600e-02, -1.6519e-04], + [ 6.4107e-03, -1.7609e-01, -3.0185e-02, ..., 5.1964e-02, + -7.6552e-02, -9.8258e-02], + [-2.6588e-01, -3.2169e-01, 5.8311e-02, ..., -1.5501e-01, + -8.0317e-02, -2.5332e-01], + ..., + [ 8.8050e-02, 2.0529e-01, -1.0762e-01, ..., -1.8488e-01, + 8.9349e-02, 1.0055e-01], + [ 2.9339e-02, -2.2085e-01, -6.9791e-02, ..., -1.6513e-01, + -1.3347e-01, -1.2750e-01], + [-3.1219e-01, -9.9798e-02, -7.3142e-02, ..., 1.1058e-01, + -3.4890e-01, 1.2045e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.1642e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([ 3.2596e-09, 2.5611e-08, -1.5367e-07, 2.3283e-09, 7.9162e-09, + 2.0955e-09, 4.8894e-09, 5.6345e-08, 5.8673e-08, 8.1491e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 220.38, cls_loss 0.0003 cls_loss_mapping 0.0005 cls_loss_causal 0.4456 re_mapping 0.0026 re_causal 0.0082 /// teacc 99.16 lr 0.00001000 +Epoch 406, weight, value: tensor([[-3.2822e-01, 2.5954e-01, -3.2769e-02, ..., -1.7033e-01, + 1.2601e-02, -1.6588e-04], + [ 6.4169e-03, -1.7609e-01, -3.0191e-02, ..., 5.1974e-02, + -7.6558e-02, -9.8261e-02], + [-2.6592e-01, -3.2170e-01, 5.8311e-02, ..., -1.5509e-01, + -8.0332e-02, -2.5333e-01], + ..., + [ 8.8050e-02, 2.0529e-01, -1.0762e-01, ..., -1.8491e-01, + 8.9340e-02, 1.0055e-01], + [ 2.9338e-02, -2.2087e-01, -6.9792e-02, ..., -1.6514e-01, + -1.3348e-01, -1.2750e-01], + [-3.1221e-01, -9.9803e-02, -7.3143e-02, ..., 1.1058e-01, + -3.4891e-01, 1.2045e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.6298e-09, 2.3283e-10, 1.6298e-09, ..., 4.6566e-10, + 2.3283e-10, 6.9849e-10], + [ 2.3283e-10, 2.3283e-10, 2.3283e-10, ..., 0.0000e+00, + 1.1642e-09, 1.3970e-09], + ..., + [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-09, 2.5611e-09], + [-7.6834e-09, 0.0000e+00, 4.4238e-09, ..., 1.3970e-09, + 2.3283e-10, 4.6566e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., -3.2596e-09, + 0.0000e+00, -3.9581e-09]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([ 1.1642e-09, 1.9558e-08, -5.8208e-09, 1.3271e-08, 1.2806e-08, + 2.0992e-06, -2.1346e-06, 2.1653e-08, -1.0477e-08, -8.8476e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 220.48, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4045 re_mapping 0.0025 re_causal 0.0078 /// teacc 99.13 lr 0.00001000 +Epoch 407, weight, value: tensor([[-3.2822e-01, 2.5958e-01, -3.2768e-02, ..., -1.7038e-01, + 1.2597e-02, -1.7076e-04], + [ 6.4174e-03, -1.7610e-01, -3.0192e-02, ..., 5.1975e-02, + -7.6569e-02, -9.8268e-02], + [-2.6595e-01, -3.2171e-01, 5.8310e-02, ..., -1.5511e-01, + -8.0343e-02, -2.5335e-01], + ..., + [ 8.8050e-02, 2.0529e-01, -1.0762e-01, ..., -1.8493e-01, + 8.9333e-02, 1.0055e-01], + [ 2.9339e-02, -2.2090e-01, -6.9825e-02, ..., -1.6516e-01, + -1.3348e-01, -1.2751e-01], + [-3.1223e-01, -9.9807e-02, -7.3147e-02, ..., 1.1059e-01, + -3.4890e-01, 1.2046e-01]], device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4925e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([ 1.6065e-08, 2.7241e-08, -7.6368e-08, 0.0000e+00, 2.0955e-09, + 8.1491e-09, -5.7044e-08, 6.7288e-08, 1.1409e-08, 4.8894e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 221.04, cls_loss 0.0003 cls_loss_mapping 0.0005 cls_loss_causal 0.4504 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.12 lr 0.00001000 +Epoch 408, weight, value: tensor([[-3.2823e-01, 2.5961e-01, -3.2769e-02, ..., -1.7042e-01, + 1.2594e-02, -1.7172e-04], + [ 6.4106e-03, -1.7610e-01, -3.0199e-02, ..., 5.1964e-02, + -7.6578e-02, -9.8278e-02], + [-2.6599e-01, -3.2173e-01, 5.8304e-02, ..., -1.5514e-01, + -8.0353e-02, -2.5337e-01], + ..., + [ 8.8049e-02, 2.0530e-01, -1.0762e-01, ..., -1.8496e-01, + 8.9333e-02, 1.0055e-01], + [ 2.9334e-02, -2.2090e-01, -6.9859e-02, ..., -1.6518e-01, + -1.3349e-01, -1.2751e-01], + [-3.1227e-01, -9.9812e-02, -7.3151e-02, ..., 1.1058e-01, + -3.4892e-01, 1.2046e-01]], device='cuda:0'), grad: tensor([[ 1.1642e-09, -7.2177e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 9.3132e-10], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 9.3132e-10, 6.9849e-10], + [ 0.0000e+00, 3.0268e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 2.3283e-10], + ..., + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-09, 1.1642e-09], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-10], + [ 1.1176e-08, 2.5611e-09, 0.0000e+00, ..., 2.3283e-09, + 1.7928e-08, 6.7521e-09]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([-6.9849e-09, 2.0256e-08, -4.1910e-09, 7.1479e-08, -5.5879e-09, + -2.3004e-07, 3.9581e-09, 1.7695e-08, 4.6566e-09, 1.3132e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 220.37, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4224 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.14 lr 0.00001000 +Epoch 409, weight, value: tensor([[-3.2823e-01, 2.5964e-01, -3.2767e-02, ..., -1.7046e-01, + 1.2595e-02, -1.7732e-04], + [ 6.4101e-03, -1.7610e-01, -3.0201e-02, ..., 5.1954e-02, + -7.6590e-02, -9.8284e-02], + [-2.6605e-01, -3.2176e-01, 5.8298e-02, ..., -1.5518e-01, + -8.0365e-02, -2.5339e-01], + ..., + [ 8.8052e-02, 2.0530e-01, -1.0762e-01, ..., -1.8500e-01, + 8.9329e-02, 1.0055e-01], + [ 2.9338e-02, -2.2092e-01, -6.9888e-02, ..., -1.6519e-01, + -1.3350e-01, -1.2752e-01], + [-3.1230e-01, -9.9814e-02, -7.3155e-02, ..., 1.1058e-01, + -3.4892e-01, 1.2047e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2619e-07, 0.0000e+00, ..., 2.3283e-10, + -9.3132e-10, 2.3283e-10], + [-1.0477e-08, -1.2107e-08, 0.0000e+00, ..., -1.9092e-08, + 0.0000e+00, -8.8476e-09], + [ 4.6566e-10, 7.2177e-09, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 9.0804e-09, 1.8626e-08, 0.0000e+00, ..., 4.4238e-08, + 2.0955e-09, 3.5157e-08], + [ 0.0000e+00, 8.8476e-09, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 6.9849e-10], + [ 4.6566e-10, 6.9849e-09, 0.0000e+00, ..., -4.6100e-08, + -3.4925e-09, -4.8894e-08]], device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0158], device='cuda:0'), grad: tensor([-3.6578e-07, -2.0280e-07, 1.9791e-08, 1.3504e-08, 5.6811e-08, + 7.6834e-09, 2.7474e-07, 2.8685e-07, 2.8173e-08, -1.0803e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 220.59, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4337 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.14 lr 0.00001000 +Epoch 410, weight, value: tensor([[-3.2824e-01, 2.5969e-01, -3.2764e-02, ..., -1.7050e-01, + 1.2592e-02, -1.8863e-04], + [ 6.4112e-03, -1.7610e-01, -3.0209e-02, ..., 5.1951e-02, + -7.6601e-02, -9.8294e-02], + [-2.6607e-01, -3.2179e-01, 5.8298e-02, ..., -1.5523e-01, + -8.0383e-02, -2.5341e-01], + ..., + [ 8.8058e-02, 2.0527e-01, -1.0762e-01, ..., -1.8529e-01, + 8.9323e-02, 1.0047e-01], + [ 2.9336e-02, -2.2095e-01, -6.9895e-02, ..., -1.6521e-01, + -1.3351e-01, -1.2753e-01], + [-3.1232e-01, -9.9752e-02, -7.3157e-02, ..., 1.1064e-01, + -3.4894e-01, 1.2055e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, -6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 4.6566e-10, 0.0000e+00, ..., 3.0268e-09, + 0.0000e+00, 2.3283e-09], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 4.6566e-10, 6.9849e-10], + [ 4.4238e-09, 6.9849e-10, 0.0000e+00, ..., -2.7940e-09, + 4.6566e-10, -6.7521e-09]], device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 1.6298e-09, 7.2177e-09, -9.3132e-09, 7.5204e-08, -1.8626e-09, + -8.5915e-08, 1.1642e-09, 1.9325e-08, 3.2596e-09, -6.9849e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 220.94, cls_loss 0.0003 cls_loss_mapping 0.0005 cls_loss_causal 0.4416 re_mapping 0.0023 re_causal 0.0078 /// teacc 99.15 lr 0.00001000 +Epoch 411, weight, value: tensor([[-3.2825e-01, 2.5975e-01, -3.2763e-02, ..., -1.7056e-01, + 1.2595e-02, -1.9223e-04], + [ 6.4140e-03, -1.7610e-01, -3.0209e-02, ..., 5.1957e-02, + -7.6616e-02, -9.8299e-02], + [-2.6611e-01, -3.2182e-01, 5.8297e-02, ..., -1.5527e-01, + -8.0399e-02, -2.5344e-01], + ..., + [ 8.8065e-02, 2.0527e-01, -1.0762e-01, ..., -1.8540e-01, + 8.9335e-02, 1.0045e-01], + [ 2.9339e-02, -2.2096e-01, -6.9898e-02, ..., -1.6523e-01, + -1.3352e-01, -1.2754e-01], + [-3.1233e-01, -9.9746e-02, -7.3159e-02, ..., 1.1067e-01, + -3.4896e-01, 1.2059e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-09, 2.7940e-09, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 9.5461e-09], + [ 6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 9.3132e-10], + ..., + [-9.7789e-09, -4.6566e-09, 0.0000e+00, ..., -1.6298e-09, + -4.6566e-10, -1.4668e-08], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 1.6298e-09, 1.3970e-09, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 3.0268e-09]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 2.5611e-09, 4.6333e-08, 7.4506e-09, 4.6566e-10, 4.6566e-09, + 4.8894e-09, -1.0245e-08, -6.6357e-08, 3.7253e-09, 1.4435e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 220.72, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4490 re_mapping 0.0023 re_causal 0.0078 /// teacc 99.15 lr 0.00001000 +Epoch 412, weight, value: tensor([[-3.2825e-01, 2.5979e-01, -3.2761e-02, ..., -1.7060e-01, + 1.2598e-02, -2.0017e-04], + [ 6.4158e-03, -1.7611e-01, -3.0208e-02, ..., 5.1977e-02, + -7.6633e-02, -9.8304e-02], + [-2.6616e-01, -3.2177e-01, 5.8292e-02, ..., -1.5541e-01, + -8.0416e-02, -2.5346e-01], + ..., + [ 8.8088e-02, 2.0527e-01, -1.0762e-01, ..., -1.8541e-01, + 8.9346e-02, 1.0045e-01], + [ 2.9334e-02, -2.2097e-01, -6.9929e-02, ..., -1.6526e-01, + -1.3354e-01, -1.2755e-01], + [-3.1237e-01, -9.9754e-02, -7.3163e-02, ..., 1.1067e-01, + -3.4899e-01, 1.2059e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.3970e-09, 2.5611e-09, -4.6566e-10, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 6.9849e-10, 4.6566e-10, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 3.0268e-09, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 3.2596e-09, 0.0000e+00, ..., 5.3551e-09, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 2.3283e-09, 0.0000e+00, ..., -6.9849e-09, + 0.0000e+00, -7.9162e-09]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([-1.8161e-08, 1.1642e-09, 6.9849e-09, 6.9849e-10, 9.5461e-09, + 2.0722e-08, -3.1898e-08, 1.1409e-08, 2.1420e-08, -1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 220.40, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4301 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.15 lr 0.00001000 +Epoch 413, weight, value: tensor([[-3.2826e-01, 2.5985e-01, -3.2762e-02, ..., -1.7064e-01, + 1.2599e-02, -2.0492e-04], + [ 6.4118e-03, -1.7611e-01, -3.0210e-02, ..., 5.1991e-02, + -7.6651e-02, -9.8318e-02], + [-2.6620e-01, -3.2180e-01, 5.8289e-02, ..., -1.5555e-01, + -8.0426e-02, -2.5350e-01], + ..., + [ 8.8103e-02, 2.0528e-01, -1.0762e-01, ..., -1.8542e-01, + 8.9364e-02, 1.0046e-01], + [ 2.9342e-02, -2.2099e-01, -6.9978e-02, ..., -1.6528e-01, + -1.3355e-01, -1.2756e-01], + [-3.1239e-01, -9.9767e-02, -7.3170e-02, ..., 1.1067e-01, + -3.4904e-01, 1.2059e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.6298e-09, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 3.0268e-09], + [ 2.3283e-10, 2.3283e-10, -4.6566e-10, ..., 2.3283e-10, + 2.5611e-09, 2.3283e-09], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 1.3970e-09, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 6.9849e-10], + [ 4.6566e-10, 1.1642e-09, 0.0000e+00, ..., 2.3283e-10, + 6.9849e-10, 4.6566e-10]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.7229e-08, -1.5600e-08, 7.9628e-08, 4.1910e-09, + -1.0571e-07, -4.8894e-09, 1.9092e-08, 1.3970e-08, 9.7789e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 220.86, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4314 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.13 lr 0.00001000 +Epoch 414, weight, value: tensor([[-3.2826e-01, 2.5988e-01, -3.2764e-02, ..., -1.7067e-01, + 1.2600e-02, -2.0372e-04], + [ 6.4017e-03, -1.7611e-01, -3.0201e-02, ..., 5.2003e-02, + -7.6669e-02, -9.8334e-02], + [-2.6623e-01, -3.2183e-01, 5.8289e-02, ..., -1.5560e-01, + -8.0447e-02, -2.5352e-01], + ..., + [ 8.8123e-02, 2.0528e-01, -1.0762e-01, ..., -1.8543e-01, + 8.9363e-02, 1.0047e-01], + [ 2.9341e-02, -2.2101e-01, -7.0022e-02, ..., -1.6531e-01, + -1.3356e-01, -1.2757e-01], + [-3.1241e-01, -9.9774e-02, -7.3189e-02, ..., 1.1067e-01, + -3.4906e-01, 1.2060e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 4.6566e-10], + [ 4.6566e-10, -2.3283e-10, 0.0000e+00, ..., -3.2596e-09, + 4.6566e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 6.9849e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-09, + 6.9849e-10, 5.5879e-09], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.6298e-09, 3.0268e-09], + [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., -2.2585e-08, + 3.2596e-09, -2.5379e-08]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0065, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 3.0268e-09, -1.6764e-08, 4.4238e-09, -7.6834e-09, 3.5390e-08, + -1.7928e-08, 1.7928e-08, 2.9104e-08, 1.9558e-08, -6.2166e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 220.98, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4260 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.16 lr 0.00001000 +Epoch 415, weight, value: tensor([[-3.2827e-01, 2.5993e-01, -3.2764e-02, ..., -1.7070e-01, + 1.2602e-02, -2.0768e-04], + [ 6.4050e-03, -1.7612e-01, -3.0202e-02, ..., 5.2018e-02, + -7.6693e-02, -9.8343e-02], + [-2.6628e-01, -3.2186e-01, 5.8291e-02, ..., -1.5567e-01, + -8.0461e-02, -2.5356e-01], + ..., + [ 8.8135e-02, 2.0529e-01, -1.0762e-01, ..., -1.8555e-01, + 8.9392e-02, 1.0044e-01], + [ 2.9348e-02, -2.2102e-01, -7.0025e-02, ..., -1.6533e-01, + -1.3357e-01, -1.2758e-01], + [-3.1243e-01, -9.9774e-02, -7.3190e-02, ..., 1.1070e-01, + -3.4912e-01, 1.2063e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.0268e-09, 8.6147e-09, -1.3970e-09, ..., -2.3283e-09, + 1.9092e-08, 2.9569e-08], + [ 1.7928e-08, -6.9849e-10, 1.1642e-09, ..., 4.6566e-10, + 2.3283e-10, 4.6566e-10], + ..., + [-1.6764e-08, -1.0245e-08, 0.0000e+00, ..., 1.1642e-09, + -1.8626e-08, -3.3993e-08], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 4.4238e-09, 4.6566e-09], + [ 1.8626e-09, 3.0268e-09, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-09, 6.2864e-09]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 6.9849e-10, -5.9372e-08, 1.3621e-07, -1.7835e-07, 2.5611e-09, + 1.4785e-07, 1.6298e-09, -9.3132e-08, 3.1665e-08, 2.1886e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 220.87, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4164 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.13 lr 0.00001000 +Epoch 416, weight, value: tensor([[-3.2828e-01, 2.5999e-01, -3.2764e-02, ..., -1.7077e-01, + 1.2601e-02, -2.0926e-04], + [ 6.4081e-03, -1.7612e-01, -3.0202e-02, ..., 5.2023e-02, + -7.6722e-02, -9.8350e-02], + [-2.6636e-01, -3.2185e-01, 5.8289e-02, ..., -1.5573e-01, + -8.0483e-02, -2.5360e-01], + ..., + [ 8.8146e-02, 2.0529e-01, -1.0763e-01, ..., -1.8557e-01, + 8.9415e-02, 1.0044e-01], + [ 2.9335e-02, -2.2105e-01, -7.0029e-02, ..., -1.6537e-01, + -1.3358e-01, -1.2760e-01], + [-3.1248e-01, -9.9784e-02, -7.3197e-02, ..., 1.1070e-01, + -3.4916e-01, 1.2064e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 1.1642e-09, -2.3283e-10, 0.0000e+00, ..., 2.5611e-09, + -2.3283e-10, -9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 8.3819e-09, 2.3283e-10, 0.0000e+00, ..., 1.0710e-08, + 2.3283e-10, 1.3970e-09]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 0.0000e+00, 6.0536e-09, -4.4238e-09, 4.6566e-10, -3.9348e-08, + 1.1642e-09, 1.1642e-09, 1.2340e-08, 3.9581e-09, 3.0501e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 220.66, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4350 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.13 lr 0.00001000 +Epoch 417, weight, value: tensor([[-3.2829e-01, 2.6004e-01, -3.2760e-02, ..., -1.7086e-01, + 1.2606e-02, -2.1300e-04], + [ 6.4199e-03, -1.7612e-01, -3.0202e-02, ..., 5.2038e-02, + -7.6740e-02, -9.8356e-02], + [-2.6643e-01, -3.2189e-01, 5.8282e-02, ..., -1.5579e-01, + -8.0513e-02, -2.5363e-01], + ..., + [ 8.8164e-02, 2.0530e-01, -1.0763e-01, ..., -1.8558e-01, + 8.9436e-02, 1.0044e-01], + [ 2.9339e-02, -2.2107e-01, -7.0038e-02, ..., -1.6541e-01, + -1.3359e-01, -1.2761e-01], + [-3.1255e-01, -9.9794e-02, -7.3204e-02, ..., 1.1069e-01, + -3.4923e-01, 1.2064e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.4925e-09, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., -1.8626e-09, + 2.3283e-10, 1.1642e-09], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 1.1642e-09], + ..., + [-6.9849e-10, -2.5611e-09, 0.0000e+00, ..., 6.7521e-09, + 0.0000e+00, 2.5611e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.6298e-09, + 4.6566e-10, 9.3132e-10], + [ 3.0268e-09, 2.7940e-09, 0.0000e+00, ..., -3.7253e-09, + 6.9849e-10, -6.0536e-09]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0236, -0.0157], device='cuda:0'), grad: tensor([ 3.0734e-08, 1.8626e-08, -1.0617e-07, -9.5461e-09, 1.2107e-08, + 3.9581e-09, -1.6065e-08, 4.4703e-08, 3.5390e-08, -1.0477e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 220.44, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4191 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.13 lr 0.00001000 +Epoch 418, weight, value: tensor([[-3.2829e-01, 2.6009e-01, -3.2762e-02, ..., -1.7095e-01, + 1.2601e-02, -2.2686e-04], + [ 6.4310e-03, -1.7613e-01, -3.0203e-02, ..., 5.2045e-02, + -7.6756e-02, -9.8362e-02], + [-2.6652e-01, -3.2193e-01, 5.8307e-02, ..., -1.5585e-01, + -8.0544e-02, -2.5367e-01], + ..., + [ 8.8172e-02, 2.0529e-01, -1.0763e-01, ..., -1.8588e-01, + 8.9442e-02, 1.0034e-01], + [ 2.9331e-02, -2.2109e-01, -7.0070e-02, ..., -1.6544e-01, + -1.3361e-01, -1.2763e-01], + [-3.1259e-01, -9.9760e-02, -7.3213e-02, ..., 1.1077e-01, + -3.4925e-01, 1.2075e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., -1.1642e-09, + 4.1910e-09, 2.7940e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [-6.7521e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -4.8894e-09, -3.2131e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 2.3283e-10], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., -1.3970e-09, + 9.3132e-10, 3.4925e-09]], device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0192, -0.0323, 0.0210, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0236, -0.0156], device='cuda:0'), grad: tensor([ 6.9849e-10, 5.0757e-08, 3.9581e-09, 3.2596e-09, 4.8894e-09, + -4.8894e-09, 2.3283e-09, -7.2876e-08, 4.6566e-09, 1.2340e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 220.33, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4481 re_mapping 0.0022 re_causal 0.0077 /// teacc 99.13 lr 0.00001000 +Epoch 419, weight, value: tensor([[-3.2830e-01, 2.6012e-01, -3.2762e-02, ..., -1.7105e-01, + 1.2597e-02, -2.4714e-04], + [ 6.4554e-03, -1.7613e-01, -3.0188e-02, ..., 5.2074e-02, + -7.6774e-02, -9.8367e-02], + [-2.6660e-01, -3.2194e-01, 5.8307e-02, ..., -1.5591e-01, + -8.0566e-02, -2.5371e-01], + ..., + [ 8.8174e-02, 2.0529e-01, -1.0765e-01, ..., -1.8589e-01, + 8.9449e-02, 1.0034e-01], + [ 2.9331e-02, -2.2109e-01, -7.0085e-02, ..., -1.6548e-01, + -1.3362e-01, -1.2765e-01], + [-3.1261e-01, -9.9766e-02, -7.3229e-02, ..., 1.1078e-01, + -3.4928e-01, 1.2076e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-08, 4.8894e-09, 0.0000e+00, ..., -6.9849e-10, + 1.2806e-08, 1.8161e-08], + [ 6.9849e-10, 6.9849e-10, 0.0000e+00, ..., 1.1642e-09, + 1.3970e-09, 2.3283e-09], + ..., + [-1.2107e-08, -6.0536e-09, 0.0000e+00, ..., 3.0268e-09, + -7.9162e-09, -1.2340e-08], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 1.6298e-09, 2.5611e-09], + [ 1.6298e-09, 4.6566e-10, 0.0000e+00, ..., -3.7253e-09, + 6.9849e-10, -4.6566e-09]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0191, -0.0323, 0.0210, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0236, -0.0156], device='cuda:0'), grad: tensor([ 4.6566e-10, 3.1432e-08, 2.0256e-08, -4.2142e-08, -4.6566e-09, + 6.0536e-09, 1.3970e-09, -2.4214e-08, 1.8161e-08, -1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 220.52, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4017 re_mapping 0.0022 re_causal 0.0074 /// teacc 99.13 lr 0.00001000 +Epoch 420, weight, value: tensor([[-3.2831e-01, 2.6016e-01, -3.2761e-02, ..., -1.7110e-01, + 1.2594e-02, -2.5414e-04], + [ 6.4477e-03, -1.7613e-01, -3.0195e-02, ..., 5.2076e-02, + -7.6802e-02, -9.8377e-02], + [-2.6664e-01, -3.2200e-01, 5.8316e-02, ..., -1.5597e-01, + -8.0620e-02, -2.5376e-01], + ..., + [ 8.8183e-02, 2.0529e-01, -1.0764e-01, ..., -1.8591e-01, + 8.9466e-02, 1.0034e-01], + [ 2.9332e-02, -2.2110e-01, -7.0131e-02, ..., -1.6552e-01, + -1.3364e-01, -1.2766e-01], + [-3.1263e-01, -9.9772e-02, -7.3238e-02, ..., 1.1079e-01, + -3.4932e-01, 1.2077e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.2468e-08, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 1.1642e-10], + [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-10, 5.3551e-09], + [ 1.1642e-10, 9.3132e-10, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, 1.3970e-09], + ..., + [-9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-09, + 3.4925e-10, -7.6834e-09], + [ 0.0000e+00, 1.7462e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 2.6776e-09], + [ 2.3283e-10, 4.0745e-09, 0.0000e+00, ..., -1.0943e-08, + 2.3283e-10, -7.6834e-09]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0191, -0.0323, 0.0210, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0237, -0.0156], device='cuda:0'), grad: tensor([-5.2969e-08, 3.3155e-07, -3.1199e-07, 2.6776e-09, 2.3050e-08, + 7.9162e-09, 2.7241e-08, -2.8289e-08, 1.6298e-08, -8.8476e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 220.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4387 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.16 lr 0.00001000 +Epoch 421, weight, value: tensor([[-3.2831e-01, 2.6019e-01, -3.2761e-02, ..., -1.7119e-01, + 1.2591e-02, -2.7214e-04], + [ 6.4426e-03, -1.7613e-01, -3.0197e-02, ..., 5.2082e-02, + -7.6825e-02, -9.8383e-02], + [-2.6669e-01, -3.2205e-01, 5.8312e-02, ..., -1.5604e-01, + -8.0661e-02, -2.5382e-01], + ..., + [ 8.8198e-02, 2.0530e-01, -1.0764e-01, ..., -1.8592e-01, + 8.9474e-02, 1.0034e-01], + [ 2.9330e-02, -2.2110e-01, -7.0171e-02, ..., -1.6556e-01, + -1.3366e-01, -1.2768e-01], + [-3.1265e-01, -9.9784e-02, -7.3244e-02, ..., 1.1080e-01, + -3.4935e-01, 1.2078e-01]], device='cuda:0'), grad: tensor([[ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 1.1642e-10], + [ 6.9849e-10, 1.1642e-10, 0.0000e+00, ..., -2.3283e-10, + 0.0000e+00, 1.5134e-09], + [ 9.3132e-10, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-10, 2.0955e-09], + ..., + [ 1.9791e-09, 1.1642e-10, 0.0000e+00, ..., 1.1059e-08, + 0.0000e+00, 5.4715e-09], + [ 2.4447e-09, 1.1642e-10, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 2.3283e-09], + [-2.6543e-08, 6.9849e-10, 1.1642e-10, ..., -6.8452e-08, + 0.0000e+00, -7.0082e-08]], device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0191, -0.0323, 0.0210, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0237, -0.0156], device='cuda:0'), grad: tensor([ 1.3853e-08, 5.9372e-09, 1.2689e-08, 8.2655e-09, 2.1420e-07, + -7.9861e-08, 3.6322e-08, 2.2352e-08, 2.8056e-08, -2.4517e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 220.77, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4212 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.14 lr 0.00001000 +Epoch 422, weight, value: tensor([[-3.2832e-01, 2.6024e-01, -3.2761e-02, ..., -1.7124e-01, + 1.2584e-02, -2.8064e-04], + [ 6.4425e-03, -1.7613e-01, -3.0199e-02, ..., 5.2150e-02, + -7.6845e-02, -9.8390e-02], + [-2.6672e-01, -3.2210e-01, 5.8309e-02, ..., -1.5624e-01, + -8.0713e-02, -2.5387e-01], + ..., + [ 8.8205e-02, 2.0530e-01, -1.0764e-01, ..., -1.8594e-01, + 8.9488e-02, 1.0034e-01], + [ 2.9330e-02, -2.2112e-01, -7.0182e-02, ..., -1.6559e-01, + -1.3367e-01, -1.2770e-01], + [-3.1269e-01, -9.9796e-02, -7.3250e-02, ..., 1.1080e-01, + -3.4942e-01, 1.2079e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.9791e-09, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, + 6.9849e-10, 2.5611e-09], + [ 8.1491e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 3.4925e-10, -4.6566e-10], + [ 3.4925e-10, 5.8208e-10, 0.0000e+00, ..., -1.4552e-08, + 2.3283e-10, -1.1874e-08]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0237, -0.0156], device='cuda:0'), grad: tensor([ 3.1432e-09, 1.3621e-08, 3.9581e-09, 1.1525e-08, 3.9348e-08, + -1.8161e-08, -1.6647e-08, 1.5134e-08, -2.9569e-08, -1.0594e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 220.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4061 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.13 lr 0.00001000 +Epoch 423, weight, value: tensor([[-3.2833e-01, 2.6031e-01, -3.2761e-02, ..., -1.7128e-01, + 1.2578e-02, -2.8421e-04], + [ 6.4573e-03, -1.7614e-01, -3.0194e-02, ..., 5.2205e-02, + -7.6871e-02, -9.8400e-02], + [-2.6680e-01, -3.2218e-01, 5.8329e-02, ..., -1.5630e-01, + -8.0748e-02, -2.5394e-01], + ..., + [ 8.8230e-02, 2.0531e-01, -1.0765e-01, ..., -1.8595e-01, + 8.9511e-02, 1.0035e-01], + [ 2.9326e-02, -2.2115e-01, -7.0186e-02, ..., -1.6562e-01, + -1.3370e-01, -1.2772e-01], + [-3.1272e-01, -9.9809e-02, -7.3260e-02, ..., 1.1081e-01, + -3.4948e-01, 1.2080e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -8.1491e-10, + 4.6566e-10, 1.0477e-09], + [ 3.4925e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 3.4925e-10, 8.1491e-10], + ..., + [-7.6834e-09, -5.7044e-09, 0.0000e+00, ..., 8.1491e-10, + -8.0327e-09, -1.8510e-08], + [ 5.5879e-09, 4.0745e-09, 0.0000e+00, ..., 1.1642e-10, + 5.8208e-09, 1.3271e-08], + [ 1.0477e-09, 5.8208e-10, 0.0000e+00, ..., 2.3283e-10, + 8.1491e-10, 1.8626e-09]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0237, -0.0156], device='cuda:0'), grad: tensor([ 2.3283e-10, 2.5611e-09, 6.4028e-09, 7.4506e-09, 9.3132e-10, + 3.6089e-09, 9.3132e-10, -1.1048e-07, 8.2771e-08, 1.3737e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 221.01, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4389 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.14 lr 0.00001000 +Epoch 424, weight, value: tensor([[-3.2833e-01, 2.6037e-01, -3.2762e-02, ..., -1.7130e-01, + 1.2573e-02, -2.8513e-04], + [ 6.4439e-03, -1.7614e-01, -3.0200e-02, ..., 5.2213e-02, + -7.6902e-02, -9.8412e-02], + [-2.6685e-01, -3.2228e-01, 5.8327e-02, ..., -1.5633e-01, + -8.0784e-02, -2.5407e-01], + ..., + [ 8.8255e-02, 2.0532e-01, -1.0764e-01, ..., -1.8597e-01, + 8.9522e-02, 1.0036e-01], + [ 2.9347e-02, -2.2116e-01, -7.0188e-02, ..., -1.6564e-01, + -1.3371e-01, -1.2773e-01], + [-3.1275e-01, -9.9819e-02, -7.3263e-02, ..., 1.1081e-01, + -3.4952e-01, 1.2080e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 2.3283e-10], + ..., + [-2.3283e-10, -2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, -8.1491e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 2.3283e-10], + [ 2.3283e-10, 3.4925e-10, 0.0000e+00, ..., -1.0477e-09, + 1.1642e-10, 3.4925e-10]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0237, -0.0156], device='cuda:0'), grad: tensor([ 1.1642e-10, -6.1700e-09, 1.9791e-09, 1.0710e-08, 5.0059e-09, + 1.8626e-09, 1.7229e-08, 1.6298e-09, -2.5844e-08, 1.0477e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 220.59, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4276 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.15 lr 0.00001000 +Epoch 425, weight, value: tensor([[-3.2834e-01, 2.6043e-01, -3.2760e-02, ..., -1.7134e-01, + 1.2570e-02, -2.9990e-04], + [ 6.4550e-03, -1.7614e-01, -3.0199e-02, ..., 5.2227e-02, + -7.6923e-02, -9.8432e-02], + [-2.6691e-01, -3.2233e-01, 5.8330e-02, ..., -1.5636e-01, + -8.0821e-02, -2.5413e-01], + ..., + [ 8.8273e-02, 2.0533e-01, -1.0764e-01, ..., -1.8598e-01, + 8.9542e-02, 1.0037e-01], + [ 2.9360e-02, -2.2117e-01, -7.0196e-02, ..., -1.6567e-01, + -1.3373e-01, -1.2775e-01], + [-3.1278e-01, -9.9828e-02, -7.3266e-02, ..., 1.1081e-01, + -3.4956e-01, 1.2081e-01]], device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 8.1491e-10, + 1.1642e-10, 0.0000e+00], + [-1.1642e-10, 1.1642e-10, 0.0000e+00, ..., -1.0477e-09, + 1.1642e-10, 2.3283e-10], + [ 3.4925e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 8.1491e-10, 8.1491e-10], + ..., + [ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 3.4925e-10, 2.3283e-10], + [-5.8208e-09, 0.0000e+00, 0.0000e+00, ..., -1.0477e-09, + 2.3283e-10, 2.3283e-10], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-10, 2.3283e-10]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0191, -0.0323, 0.0210, -0.0283, 0.0403, 0.0166, -0.0011, -0.0066, + -0.0237, -0.0156], device='cuda:0'), grad: tensor([ 4.6566e-09, -2.3283e-10, -6.9849e-10, -1.5134e-09, 3.8417e-09, + 1.4552e-08, 3.3528e-08, 1.1292e-08, -5.7742e-08, 4.5402e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 220.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4311 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 426, weight, value: tensor([[-3.2835e-01, 2.6048e-01, -3.2754e-02, ..., -1.7144e-01, + 1.2563e-02, -3.0826e-04], + [ 6.4755e-03, -1.7615e-01, -3.0191e-02, ..., 5.2305e-02, + -7.6949e-02, -9.8452e-02], + [-2.6700e-01, -3.2242e-01, 5.8361e-02, ..., -1.5649e-01, + -8.0861e-02, -2.5425e-01], + ..., + [ 8.8301e-02, 2.0532e-01, -1.0766e-01, ..., -1.8620e-01, + 8.9557e-02, 1.0030e-01], + [ 2.9349e-02, -2.2119e-01, -7.0199e-02, ..., -1.6572e-01, + -1.3376e-01, -1.2778e-01], + [-3.1282e-01, -9.9776e-02, -7.3277e-02, ..., 1.1089e-01, + -3.4964e-01, 1.2090e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.2806e-09, 9.3132e-10, 5.8208e-10, ..., 1.2806e-09, + 1.1642e-09, 1.9791e-09], + [ 4.4238e-09, 1.2573e-08, 8.1491e-10, ..., 1.0477e-09, + 1.0012e-08, 1.5716e-08], + ..., + [-3.6089e-09, -1.4435e-08, 3.4925e-10, ..., 1.3970e-09, + -4.0745e-09, -6.6357e-09], + [ 2.3283e-10, 1.1642e-10, 1.1642e-10, ..., 6.9849e-10, + 5.8208e-10, 1.0477e-09], + [ 2.5611e-09, 1.2806e-09, 2.3283e-10, ..., -8.1491e-10, + 3.1432e-09, 2.9104e-09]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 2.3283e-10, 1.2224e-08, 7.7416e-08, -3.3877e-08, -1.5250e-08, + 7.2177e-09, -2.0955e-09, -4.9942e-08, 4.8894e-09, 1.5134e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 220.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4098 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.16 lr 0.00001000 +Epoch 427, weight, value: tensor([[-3.2836e-01, 2.6054e-01, -3.2757e-02, ..., -1.7149e-01, + 1.2562e-02, -3.2324e-04], + [ 6.4697e-03, -1.7616e-01, -3.0190e-02, ..., 5.2305e-02, + -7.6985e-02, -9.8466e-02], + [-2.6708e-01, -3.2257e-01, 5.8360e-02, ..., -1.5653e-01, + -8.0917e-02, -2.5440e-01], + ..., + [ 8.8343e-02, 2.0533e-01, -1.0765e-01, ..., -1.8631e-01, + 8.9602e-02, 1.0027e-01], + [ 2.9353e-02, -2.2122e-01, -7.0208e-02, ..., -1.6576e-01, + -1.3378e-01, -1.2780e-01], + [-3.1285e-01, -9.9759e-02, -7.3283e-02, ..., 1.1093e-01, + -3.4968e-01, 1.2094e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2806e-09, 0.0000e+00, ..., 1.6298e-09, + 0.0000e+00, 0.0000e+00], + [-1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 2.3283e-09], + [ 6.9849e-10, 1.1642e-10, 0.0000e+00, ..., 4.6566e-10, + 6.7521e-09, 8.0327e-09], + ..., + [-5.8208e-10, -2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + -9.7789e-09, -1.0594e-08], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 3.4925e-10, 1.0477e-09], + [ 2.3283e-10, 1.6298e-09, 0.0000e+00, ..., -3.1432e-09, + 4.6566e-10, -1.9791e-09]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0402, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 4.0745e-09, 1.1642e-08, 5.1106e-08, -4.6566e-10, 9.8953e-09, + 1.5367e-08, -2.9104e-08, -6.5425e-08, 1.5134e-08, -1.9791e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 220.79, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4205 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.16 lr 0.00001000 +Epoch 428, weight, value: tensor([[-3.2836e-01, 2.6068e-01, -3.2758e-02, ..., -1.7152e-01, + 1.2563e-02, -3.2990e-04], + [ 6.4618e-03, -1.7616e-01, -3.0190e-02, ..., 5.2309e-02, + -7.7003e-02, -9.8477e-02], + [-2.6715e-01, -3.2262e-01, 5.8349e-02, ..., -1.5656e-01, + -8.0953e-02, -2.5446e-01], + ..., + [ 8.8371e-02, 2.0534e-01, -1.0766e-01, ..., -1.8632e-01, + 8.9621e-02, 1.0028e-01], + [ 2.9363e-02, -2.2126e-01, -7.0215e-02, ..., -1.6580e-01, + -1.3379e-01, -1.2782e-01], + [-3.1289e-01, -9.9777e-02, -7.3292e-02, ..., 1.1093e-01, + -3.4973e-01, 1.2095e-01]], device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 1.1642e-10, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 9.3132e-10, 1.9791e-09, 1.1642e-10, ..., -4.6566e-10, + 1.3970e-09, 5.5879e-09], + [ 1.6298e-09, 5.8208e-10, 0.0000e+00, ..., 1.1642e-10, + 2.2119e-09, 4.5402e-09], + ..., + [-8.8476e-09, -8.9640e-09, 0.0000e+00, ..., 5.8208e-10, + -4.5402e-09, -2.3283e-08], + [-2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 3.4925e-10, 9.3132e-10], + [ 5.1223e-09, 5.2387e-09, 0.0000e+00, ..., 3.4925e-10, + 2.5611e-09, 1.2224e-08]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0402, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 2.3283e-09, 4.5402e-09, 1.8626e-08, -1.5134e-09, 5.8208e-09, + 6.8685e-09, 0.0000e+00, -6.7637e-08, 2.3283e-10, 4.2375e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 220.66, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4039 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.17 lr 0.00001000 +Epoch 429, weight, value: tensor([[-3.2837e-01, 2.6078e-01, -3.2753e-02, ..., -1.7156e-01, + 1.2557e-02, -3.3071e-04], + [ 6.4621e-03, -1.7617e-01, -3.0183e-02, ..., 5.2368e-02, + -7.7030e-02, -9.8487e-02], + [-2.6720e-01, -3.2265e-01, 5.8346e-02, ..., -1.5674e-01, + -8.0994e-02, -2.5450e-01], + ..., + [ 8.8382e-02, 2.0533e-01, -1.0766e-01, ..., -1.8641e-01, + 8.9645e-02, 1.0025e-01], + [ 2.9357e-02, -2.2128e-01, -7.0218e-02, ..., -1.6585e-01, + -1.3380e-01, -1.2784e-01], + [-3.1297e-01, -9.9775e-02, -7.3298e-02, ..., 1.1096e-01, + -3.4980e-01, 1.2099e-01]], device='cuda:0'), grad: tensor([[ 3.4925e-10, -1.7462e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 2.6776e-09, 2.3283e-10, 2.3283e-09, ..., -2.9104e-09, + 1.6298e-09, 5.8208e-09], + [ 8.1491e-10, 3.9581e-09, 0.0000e+00, ..., 1.0477e-09, + 8.1491e-10, 2.4447e-09], + ..., + [ 2.4447e-09, -3.9581e-09, 0.0000e+00, ..., 1.3970e-09, + 1.1642e-09, 1.8626e-09], + [ 6.9849e-10, 3.4925e-10, 0.0000e+00, ..., 5.8208e-10, + 4.6566e-10, 1.1642e-09], + [ 6.9849e-10, 1.5134e-09, 1.1642e-10, ..., 5.8208e-10, + 9.3132e-10, 3.3760e-09]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0283, 0.0402, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([-2.3283e-10, -3.0501e-08, 1.6764e-08, -5.1805e-08, -2.8522e-08, + 1.3504e-08, 4.5984e-08, 2.2352e-08, 9.6625e-09, 1.2922e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 220.62, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4203 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.19 lr 0.00001000 +Epoch 430, weight, value: tensor([[-3.2837e-01, 2.6083e-01, -3.2754e-02, ..., -1.7168e-01, + 1.2554e-02, -3.3528e-04], + [ 6.4754e-03, -1.7617e-01, -3.0187e-02, ..., 5.2406e-02, + -7.7057e-02, -9.8513e-02], + [-2.6730e-01, -3.2277e-01, 5.8345e-02, ..., -1.5682e-01, + -8.1047e-02, -2.5463e-01], + ..., + [ 8.8400e-02, 2.0535e-01, -1.0766e-01, ..., -1.8643e-01, + 8.9654e-02, 1.0026e-01], + [ 2.9362e-02, -2.2130e-01, -7.0221e-02, ..., -1.6589e-01, + -1.3382e-01, -1.2786e-01], + [-3.1304e-01, -9.9790e-02, -7.3301e-02, ..., 1.1096e-01, + -3.4983e-01, 1.2100e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.2596e-09, 0.0000e+00, ..., 2.5611e-09, + -2.3283e-10, 2.7940e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 6.9849e-09, 4.9826e-08], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 3.9581e-09, 1.7695e-08], + ..., + [ 1.1642e-09, 1.1642e-09, 0.0000e+00, ..., 3.0268e-09, + -1.3271e-08, -9.3365e-08], + [-6.9849e-10, -9.3132e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 4.6566e-10], + [ 6.9151e-08, 1.6298e-09, 0.0000e+00, ..., 8.2655e-08, + 4.1910e-09, 1.2806e-08]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 6.2864e-09, 1.4924e-07, 4.9127e-08, 5.8208e-09, -3.0547e-07, + -8.6147e-09, -1.1642e-09, -2.5542e-07, -6.9849e-10, 3.7509e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 220.39, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4495 re_mapping 0.0021 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +Epoch 431, weight, value: tensor([[-3.2838e-01, 2.6090e-01, -3.2752e-02, ..., -1.7175e-01, + 1.2556e-02, -3.4230e-04], + [ 6.4761e-03, -1.7617e-01, -3.0188e-02, ..., 5.2414e-02, + -7.7087e-02, -9.8528e-02], + [-2.6736e-01, -3.2286e-01, 5.8346e-02, ..., -1.5684e-01, + -8.1092e-02, -2.5475e-01], + ..., + [ 8.8434e-02, 2.0536e-01, -1.0766e-01, ..., -1.8644e-01, + 8.9658e-02, 1.0027e-01], + [ 2.9376e-02, -2.2132e-01, -7.0240e-02, ..., -1.6592e-01, + -1.3384e-01, -1.2788e-01], + [-3.1309e-01, -9.9804e-02, -7.3308e-02, ..., 1.1097e-01, + -3.4988e-01, 1.2101e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 4.6566e-10], + [ 0.0000e+00, -6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 6.9849e-10], + ..., + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.0955e-09, + 2.3283e-10, 1.8626e-09], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 6.9849e-10], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -1.3039e-08, + 0.0000e+00, -1.0477e-08]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.6764e-08, -6.9151e-08, 1.6298e-09, 2.5844e-08, + 3.7253e-09, 1.8626e-09, 3.9116e-08, 8.6147e-09, -2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 220.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4163 re_mapping 0.0021 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 +Epoch 432, weight, value: tensor([[-3.2838e-01, 2.6108e-01, -3.2753e-02, ..., -1.7178e-01, + 1.2547e-02, -3.4407e-04], + [ 6.4643e-03, -1.7618e-01, -3.0193e-02, ..., 5.2398e-02, + -7.7124e-02, -9.8549e-02], + [-2.6744e-01, -3.2294e-01, 5.8338e-02, ..., -1.5690e-01, + -8.1166e-02, -2.5485e-01], + ..., + [ 8.8472e-02, 2.0537e-01, -1.0766e-01, ..., -1.8647e-01, + 8.9667e-02, 1.0028e-01], + [ 2.9374e-02, -2.2138e-01, -7.0244e-02, ..., -1.6598e-01, + -1.3387e-01, -1.2792e-01], + [-3.1317e-01, -9.9828e-02, -7.3312e-02, ..., 1.1098e-01, + -3.4993e-01, 1.2102e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 2.3283e-10], + [ 6.9849e-10, 9.3132e-10, -6.9849e-10, ..., -4.6566e-10, + 4.6566e-10, 3.7253e-09], + [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.0955e-09, + 9.3132e-10, 1.1642e-09], + ..., + [ 4.6566e-10, -9.3132e-10, 4.6566e-10, ..., 1.9791e-08, + -2.3283e-10, 1.7229e-08], + [-6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, -2.3283e-09], + [ 1.1642e-09, 2.3283e-10, 6.9849e-10, ..., -2.1653e-08, + 9.3132e-10, -2.0256e-08]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 8.6147e-09, -8.3586e-08, 6.5425e-08, 7.8231e-08, 0.0000e+00, + -8.1956e-08, 0.0000e+00, 5.8673e-08, -2.0489e-08, -2.3516e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 220.84, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4346 re_mapping 0.0021 re_causal 0.0075 /// teacc 99.19 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.3284, 0.2612, -0.0328, ..., -0.1719, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0772, -0.0986], + [-0.2675, -0.3230, 0.0583, ..., -0.1569, -0.0812, -0.2550], + ..., + [ 0.0885, 0.2054, -0.1077, ..., -0.1865, 0.0897, 0.1003], + [ 0.0294, -0.2214, -0.0703, ..., -0.1660, -0.1339, -0.1279], + [-0.3132, -0.0998, -0.0733, ..., 0.1110, -0.3500, 0.1210]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.2177e-09, + 0.0000e+00, 2.3283e-10], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 8.3819e-09, + 1.3970e-09, 6.0536e-09], + [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 6.2864e-09, + 4.6566e-10, 1.3970e-09], + ..., + [-3.3993e-08, -5.9372e-08, 0.0000e+00, ..., -1.9558e-08, + -3.8417e-08, -1.2643e-07], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.5367e-08, + 2.3283e-10, 9.3132e-10], + [ 3.3295e-08, 5.4948e-08, 0.0000e+00, ..., 1.9558e-08, + 3.5856e-08, 1.1642e-07]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0283, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 2.9569e-08, 5.6112e-08, 1.2340e-08, 9.3132e-09, 5.7369e-07, + 4.5868e-08, -7.6182e-07, -3.9232e-07, 6.5658e-08, 3.7206e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 220.22, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4268 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.3284, 0.2612, -0.0328, ..., -0.1720, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0772, -0.0986], + [-0.2676, -0.3231, 0.0583, ..., -0.1569, -0.0812, -0.2551], + ..., + [ 0.0886, 0.2054, -0.1077, ..., -0.1865, 0.0897, 0.1003], + [ 0.0293, -0.2214, -0.0703, ..., -0.1661, -0.1339, -0.1280], + [-0.3133, -0.0999, -0.0733, ..., 0.1110, -0.3501, 0.1210]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.3970e-09, 0.0000e+00, ..., 1.6298e-09, + 2.3283e-10, 2.7940e-09], + [ 6.2166e-08, 4.6566e-10, 0.0000e+00, ..., 5.2853e-08, + 2.3283e-10, 1.3970e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 9.3132e-10], + ..., + [ 2.2585e-08, 9.3132e-10, 0.0000e+00, ..., 2.1188e-08, + 2.3283e-10, 3.0268e-09], + [ 6.4494e-08, 4.6566e-10, 0.0000e+00, ..., 5.5647e-08, + 2.3283e-10, 1.3970e-09], + [ 2.7940e-08, -6.5193e-09, 0.0000e+00, ..., 8.6147e-09, + 6.9849e-10, -2.1420e-08]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 6.2864e-09, 2.4796e-07, -2.2119e-08, 1.2107e-08, -6.5565e-07, + -1.7928e-08, 3.6787e-08, 1.1851e-07, 2.6100e-07, 2.4447e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 220.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4101 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.3284, 0.2614, -0.0328, ..., -0.1721, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0772, -0.0986], + [-0.2677, -0.3231, 0.0583, ..., -0.1570, -0.0813, -0.2551], + ..., + [ 0.0886, 0.2054, -0.1077, ..., -0.1865, 0.0898, 0.1003], + [ 0.0293, -0.2215, -0.0704, ..., -0.1662, -0.1339, -0.1280], + [-0.3135, -0.0999, -0.0733, ..., 0.1110, -0.3501, 0.1210]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 1.6298e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 4.6566e-10, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 1.6298e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 3.7253e-09, + 2.3283e-10, 4.8894e-09], + [ 4.1910e-09, -1.6298e-09, 0.0000e+00, ..., -3.9581e-09, + 2.3283e-10, -1.2806e-08]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0403, 0.0166, -0.0011, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 6.0536e-09, 7.4506e-09, -5.3551e-09, 2.0955e-09, -6.2864e-09, + 9.7789e-09, -6.9849e-09, 1.3039e-08, 2.1188e-08, -2.6776e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 220.15, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4231 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.3284, 0.2614, -0.0328, ..., -0.1722, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0773, -0.0986], + [-0.2678, -0.3232, 0.0583, ..., -0.1570, -0.0814, -0.2552], + ..., + [ 0.0886, 0.2054, -0.1077, ..., -0.1865, 0.0898, 0.1003], + [ 0.0293, -0.2215, -0.0704, ..., -0.1662, -0.1340, -0.1281], + [-0.3136, -0.0999, -0.0733, ..., 0.1109, -0.3502, 0.1211]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 7.6834e-09, + 1.1642e-09, 1.1409e-08], + [ 1.4668e-08, 2.3283e-10, 0.0000e+00, ..., 9.7789e-09, + 2.3283e-10, 2.7940e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 3.0268e-09, + 4.6566e-10, 3.9581e-09], + ..., + [-2.0955e-09, 2.3283e-10, 0.0000e+00, ..., 4.4471e-08, + 2.3283e-10, 3.3062e-08], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 6.0536e-09, + 9.3132e-10, 8.6147e-09], + [ 3.7020e-08, -4.1910e-09, 0.0000e+00, ..., -5.1688e-08, + -3.7253e-09, -1.0221e-07]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0403, 0.0166, -0.0010, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 4.3074e-08, 2.0349e-07, 1.9325e-08, 6.5193e-09, -5.5879e-08, + -2.3283e-09, 1.3737e-08, -3.0966e-08, 1.4435e-08, -2.0210e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 220.67, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4164 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.18 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.3284, 0.2615, -0.0328, ..., -0.1723, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0523, -0.0773, -0.0986], + [-0.2678, -0.3232, 0.0583, ..., -0.1571, -0.0814, -0.2553], + ..., + [ 0.0886, 0.2054, -0.1077, ..., -0.1867, 0.0898, 0.1003], + [ 0.0293, -0.2216, -0.0704, ..., -0.1663, -0.1340, -0.1281], + [-0.3138, -0.0999, -0.0734, ..., 0.1110, -0.3502, 0.1211]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., 4.8894e-09, + 0.0000e+00, 6.9849e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 4.6566e-10], + ..., + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 2.5611e-09, + 0.0000e+00, 2.3283e-10], + [-1.4901e-08, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 9.5461e-09, 2.3283e-10, 0.0000e+00, ..., 3.9581e-09, + 2.3283e-10, 4.6566e-10]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0403, 0.0166, -0.0010, -0.0067, + -0.0237, -0.0155], device='cuda:0'), grad: tensor([ 1.1409e-08, 5.4715e-08, -4.5868e-08, -1.6531e-08, -5.8208e-08, + 9.3365e-08, 1.8626e-08, 2.6077e-08, -2.0792e-07, 1.3178e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 220.44, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4340 re_mapping 0.0021 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.3284, 0.2616, -0.0328, ..., -0.1724, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0523, -0.0773, -0.0987], + [-0.2679, -0.3233, 0.0583, ..., -0.1571, -0.0815, -0.2553], + ..., + [ 0.0887, 0.2054, -0.1077, ..., -0.1867, 0.0898, 0.1003], + [ 0.0293, -0.2216, -0.0704, ..., -0.1664, -0.1340, -0.1281], + [-0.3139, -0.0999, -0.0734, ..., 0.1110, -0.3503, 0.1211]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [-8.6147e-09, -9.3132e-10, 1.3970e-09, ..., -6.5193e-09, + 4.6566e-10, -1.2387e-07], + [ 2.3283e-09, 2.3283e-10, 2.3283e-10, ..., 1.6298e-09, + -2.3283e-10, 1.2806e-08], + ..., + [ 1.3970e-09, 4.6566e-10, -1.8626e-09, ..., 3.9581e-09, + 1.5600e-08, 1.2224e-07], + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-10], + [ 2.5611e-09, 2.3283e-10, 0.0000e+00, ..., -4.4238e-09, + 0.0000e+00, -5.1223e-09]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0067, + -0.0238, -0.0155], device='cuda:0'), grad: tensor([ 1.2340e-08, -1.2014e-06, 8.0094e-08, -2.8173e-08, 1.5367e-08, + 2.3283e-09, -1.0245e-08, 1.1278e-06, 1.0477e-08, 2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 220.38, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4170 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.15 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.3284, 0.2617, -0.0328, ..., -0.1725, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0523, -0.0774, -0.0987], + [-0.2679, -0.3233, 0.0583, ..., -0.1571, -0.0815, -0.2554], + ..., + [ 0.0887, 0.2054, -0.1077, ..., -0.1867, 0.0898, 0.1003], + [ 0.0292, -0.2216, -0.0704, ..., -0.1664, -0.1340, -0.1281], + [-0.3140, -0.0999, -0.0734, ..., 0.1110, -0.3504, 0.1211]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-10], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 1.6298e-09], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 4.6566e-10], + ..., + [-4.1910e-09, -6.7521e-09, 0.0000e+00, ..., 9.3132e-10, + -1.6298e-09, -4.4238e-09], + [-4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 1.3970e-09], + [ 1.1642e-09, 1.6298e-09, 0.0000e+00, ..., -9.3132e-09, + 4.6566e-10, -1.0477e-08]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0067, + -0.0238, -0.0155], device='cuda:0'), grad: tensor([ 2.5611e-09, 8.8476e-09, 2.7940e-09, 3.9581e-09, 4.3074e-08, + 2.8173e-08, 6.9849e-09, -2.0955e-08, -2.6310e-08, -3.6089e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 220.59, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4508 re_mapping 0.0021 re_causal 0.0076 /// teacc 99.16 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.3284, 0.2618, -0.0328, ..., -0.1725, 0.0125, -0.0004], + [ 0.0064, -0.1762, -0.0302, ..., 0.0523, -0.0774, -0.0987], + [-0.2681, -0.3234, 0.0583, ..., -0.1571, -0.0816, -0.2555], + ..., + [ 0.0887, 0.2055, -0.1077, ..., -0.1867, 0.0899, 0.1003], + [ 0.0292, -0.2217, -0.0704, ..., -0.1665, -0.1341, -0.1282], + [-0.3141, -0.1000, -0.0734, ..., 0.1110, -0.3504, 0.1211]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 9.3132e-10], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 6.9849e-10, + 1.1642e-09, 2.0955e-09], + [ 2.3283e-10, 1.3970e-09, 0.0000e+00, ..., 2.0955e-09, + 2.7940e-09, 3.0268e-09], + ..., + [-1.1642e-09, -4.6566e-09, 0.0000e+00, ..., 3.0268e-09, + -9.3132e-10, -2.3283e-09], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.5611e-09, + 6.9849e-10, 2.3283e-09], + [ 2.3283e-10, -2.3283e-10, 0.0000e+00, ..., -1.5832e-08, + 2.3283e-10, -1.4901e-08]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0067, + -0.0238, -0.0155], device='cuda:0'), grad: tensor([ 1.2573e-08, -6.5891e-08, -9.7090e-08, -1.6764e-08, 3.1199e-08, + 1.9791e-08, -4.5635e-08, 3.2829e-08, 1.7136e-07, -4.3306e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 220.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4289 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.15 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.3285, 0.2619, -0.0328, ..., -0.1726, 0.0125, -0.0005], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0774, -0.0987], + [-0.2681, -0.3235, 0.0584, ..., -0.1572, -0.0816, -0.2556], + ..., + [ 0.0887, 0.2055, -0.1077, ..., -0.1868, 0.0899, 0.1003], + [ 0.0292, -0.2217, -0.0704, ..., -0.1666, -0.1341, -0.1282], + [-0.3142, -0.1000, -0.0734, ..., 0.1110, -0.3505, 0.1212]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., -2.3283e-10, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + ..., + [ 2.5611e-09, -4.6566e-10, 0.0000e+00, ..., 3.7253e-09, + 0.0000e+00, 3.7253e-09], + [-1.7928e-08, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 2.3283e-10, 0.0000e+00, ..., -5.6112e-08, + 0.0000e+00, -6.9151e-08]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0067, + -0.0238, -0.0155], device='cuda:0'), grad: tensor([ 1.6298e-09, 7.9162e-09, 9.0804e-09, 4.6566e-09, 1.7392e-07, + 5.5879e-08, 1.4901e-08, 2.9104e-08, -1.1967e-07, -1.6438e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 220.31, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4409 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.15 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.3285, 0.2621, -0.0327, ..., -0.1727, 0.0125, -0.0005], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0775, -0.0987], + [-0.2682, -0.3235, 0.0584, ..., -0.1572, -0.0817, -0.2556], + ..., + [ 0.0887, 0.2054, -0.1077, ..., -0.1870, 0.0899, 0.1002], + [ 0.0292, -0.2218, -0.0704, ..., -0.1666, -0.1341, -0.1282], + [-0.3143, -0.0999, -0.0734, ..., 0.1110, -0.3505, 0.1213]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, 0.0000e+00, ..., 2.3283e-10, + 1.3970e-09, 9.3132e-10], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 2.0955e-09, 1.3970e-09], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 6.9849e-10], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 4.6566e-10], + [ 9.3132e-10, 1.3970e-09, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 4.6566e-10]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0193, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0068, + -0.0238, -0.0155], device='cuda:0'), grad: tensor([ 3.0268e-09, 2.1886e-08, -7.6834e-09, -2.2585e-08, 2.5611e-09, + -2.1886e-08, 5.1223e-09, 6.9849e-09, 4.8894e-09, 8.1491e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 220.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4118 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.3285, 0.2623, -0.0327, ..., -0.1728, 0.0125, -0.0005], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0775, -0.0988], + [-0.2682, -0.3236, 0.0584, ..., -0.1573, -0.0817, -0.2557], + ..., + [ 0.0887, 0.2055, -0.1077, ..., -0.1870, 0.0899, 0.1002], + [ 0.0292, -0.2218, -0.0704, ..., -0.1667, -0.1341, -0.1283], + [-0.3144, -0.1000, -0.0734, ..., 0.1110, -0.3506, 0.1213]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 7.7300e-08, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 1.3970e-09, 0.0000e+00, ..., 6.5193e-09, + 2.3283e-10, 9.3132e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 3.0268e-09, + 2.3283e-10, 0.0000e+00], + ..., + [-6.9849e-10, -2.5611e-09, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, -4.6566e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 1.8626e-09, + 0.0000e+00, 2.3283e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0194, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0068, + -0.0238, -0.0155], device='cuda:0'), grad: tensor([ 3.0710e-07, 2.4913e-08, 6.7521e-09, 1.6531e-08, 1.0896e-07, + 4.6566e-09, -4.7125e-07, 6.9849e-09, 3.4925e-09, 5.3551e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 220.53, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4161 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.15 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.3285, 0.2624, -0.0327, ..., -0.1729, 0.0125, -0.0005], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0775, -0.0988], + [-0.2683, -0.3236, 0.0584, ..., -0.1573, -0.0818, -0.2558], + ..., + [ 0.0888, 0.2055, -0.1077, ..., -0.1871, 0.0899, 0.1002], + [ 0.0292, -0.2218, -0.0705, ..., -0.1668, -0.1342, -0.1283], + [-0.3145, -0.0999, -0.0734, ..., 0.1111, -0.3506, 0.1214]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.4715e-08, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 4.8894e-09], + [-9.3132e-10, 6.9849e-10, 0.0000e+00, ..., -6.9849e-10, + 0.0000e+00, 6.9849e-10], + [ 1.1642e-09, 3.9581e-09, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 1.3970e-09], + ..., + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, + 0.0000e+00, 1.1409e-08], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 7.9162e-09, + 0.0000e+00, 5.3551e-09], + [ 8.8941e-08, 3.2596e-09, 0.0000e+00, ..., 9.3132e-08, + 0.0000e+00, -3.4692e-08]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0194, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0068, + -0.0238, -0.0154], device='cuda:0'), grad: tensor([-7.2876e-08, -3.7253e-09, 1.4435e-08, 5.8208e-09, -3.2713e-07, + -1.3970e-09, 4.7497e-08, 4.5635e-08, 4.4703e-08, 2.6915e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 220.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4235 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.18 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.3285, 0.2626, -0.0327, ..., -0.1730, 0.0125, -0.0005], + [ 0.0064, -0.1762, -0.0302, ..., 0.0524, -0.0775, -0.0988], + [-0.2684, -0.3237, 0.0584, ..., -0.1573, -0.0818, -0.2559], + ..., + [ 0.0888, 0.2055, -0.1077, ..., -0.1871, 0.0899, 0.1002], + [ 0.0292, -0.2219, -0.0705, ..., -0.1668, -0.1342, -0.1284], + [-0.3146, -0.1000, -0.0734, ..., 0.1111, -0.3506, 0.1214]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.8208e-09, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 0.0000e+00], + [ 4.8894e-09, 1.0245e-08, 0.0000e+00, ..., -1.1874e-08, + 6.9849e-09, 1.7462e-08], + [ 1.3970e-09, 1.6298e-09, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 2.0955e-09], + ..., + [-1.5134e-08, -1.8626e-08, 2.3283e-10, ..., 8.3819e-09, + -1.2573e-08, -3.0734e-08], + [ 1.8626e-09, 3.0268e-09, 0.0000e+00, ..., 6.9849e-10, + 9.3132e-10, 2.0955e-09], + [ 5.3551e-09, 8.3819e-09, 0.0000e+00, ..., 9.3132e-10, + 3.2596e-09, 8.1491e-09]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0194, -0.0323, 0.0209, -0.0284, 0.0404, 0.0166, -0.0010, -0.0068, + -0.0238, -0.0154], device='cuda:0'), grad: tensor([-1.3737e-08, -3.2829e-08, 1.6298e-08, 2.5611e-09, 1.5600e-08, + 5.1223e-09, 3.7253e-09, -6.3097e-08, 1.9325e-08, 5.1223e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 220.62, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4130 re_mapping 0.0020 re_causal 0.0070 /// teacc 99.17 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.3285, 0.2627, -0.0327, ..., -0.1730, 0.0125, -0.0005], + [ 0.0064, -0.1762, -0.0302, ..., 0.0525, -0.0776, -0.0988], + [-0.2685, -0.3237, 0.0584, ..., -0.1574, -0.0819, -0.2559], + ..., + [ 0.0888, 0.2055, -0.1077, ..., -0.1872, 0.0899, 0.1002], + [ 0.0292, -0.2220, -0.0705, ..., -0.1669, -0.1342, -0.1284], + [-0.3147, -0.1000, -0.0734, ..., 0.1111, -0.3507, 0.1214]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2806e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 4.6566e-10], + [-4.6566e-10, 0.0000e+00, 0.0000e+00, ..., -6.9849e-10, + 4.6566e-10, 6.9849e-10], + [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 1.1642e-09], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 6.9849e-10], + [-2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 6.9849e-10], + [ 6.9849e-10, 1.3504e-08, 0.0000e+00, ..., 2.3283e-10, + 4.6566e-10, 1.3970e-09]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0195, -0.0323, 0.0209, -0.0284, 0.0404, 0.0167, -0.0010, -0.0069, + -0.0239, -0.0155], device='cuda:0'), grad: tensor([-3.1432e-08, 1.3970e-09, 6.9849e-09, -4.7730e-08, 1.6298e-09, + 3.0501e-08, 2.3283e-09, 8.8476e-09, 6.9849e-10, 4.0280e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 220.65, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4210 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.3285, 0.2629, -0.0327, ..., -0.1731, 0.0125, -0.0005], + [ 0.0065, -0.1762, -0.0302, ..., 0.0525, -0.0776, -0.0989], + [-0.2686, -0.3238, 0.0584, ..., -0.1574, -0.0819, -0.2560], + ..., + [ 0.0888, 0.2055, -0.1077, ..., -0.1872, 0.0899, 0.1002], + [ 0.0292, -0.2220, -0.0705, ..., -0.1670, -0.1342, -0.1284], + [-0.3148, -0.1000, -0.0734, ..., 0.1111, -0.3507, 0.1214]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [-1.8626e-09, -9.3132e-10, 0.0000e+00, ..., 2.5611e-09, + -6.9849e-10, 9.3132e-10], + [ 1.3970e-09, 6.9849e-10, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 3.7253e-09], + [ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., -5.3551e-09, + 0.0000e+00, -1.0477e-08]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0195, -0.0323, 0.0209, -0.0284, 0.0405, 0.0167, -0.0010, -0.0069, + -0.0239, -0.0155], device='cuda:0'), grad: tensor([ 1.8626e-09, 4.9826e-08, -5.2387e-08, 2.3283e-10, 8.3819e-09, + 1.1642e-09, 5.3551e-09, 9.3132e-10, 1.5367e-08, -2.2119e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 220.86, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4588 re_mapping 0.0020 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.3285, 0.2630, -0.0327, ..., -0.1732, 0.0125, -0.0005], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0776, -0.0989], + [-0.2687, -0.3238, 0.0584, ..., -0.1575, -0.0820, -0.2560], + ..., + [ 0.0889, 0.2055, -0.1077, ..., -0.1872, 0.0899, 0.1002], + [ 0.0291, -0.2220, -0.0705, ..., -0.1671, -0.1343, -0.1285], + [-0.3149, -0.1001, -0.0734, ..., 0.1111, -0.3507, 0.1214]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.8894e-09, 0.0000e+00, ..., 1.0664e-07, + 0.0000e+00, 0.0000e+00], + [ 2.5611e-09, 4.1910e-09, 0.0000e+00, ..., 1.0943e-08, + 4.6566e-10, 3.2596e-09], + [ 2.3283e-10, 6.9849e-10, 0.0000e+00, ..., 2.0955e-09, + 4.6566e-09, 3.4925e-09], + ..., + [-3.7253e-09, -4.8894e-09, 0.0000e+00, ..., 4.6566e-10, + -2.3283e-10, -4.4238e-09], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., 1.8626e-09, + 2.3283e-10, 0.0000e+00], + [ 1.1642e-09, 2.3283e-09, 0.0000e+00, ..., 1.3970e-09, + 2.3283e-10, 1.1642e-09]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0195, -0.0323, 0.0209, -0.0285, 0.0405, 0.0167, -0.0010, -0.0069, + -0.0239, -0.0155], device='cuda:0'), grad: tensor([ 4.0536e-07, 5.6578e-08, 2.8638e-08, -9.3132e-09, 1.1572e-07, + 1.6065e-08, -6.0629e-07, -1.5600e-08, 9.7789e-09, 9.5461e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 220.63, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4389 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.3285, 0.2631, -0.0327, ..., -0.1733, 0.0125, -0.0005], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0777, -0.0989], + [-0.2688, -0.3239, 0.0584, ..., -0.1575, -0.0820, -0.2561], + ..., + [ 0.0889, 0.2055, -0.1077, ..., -0.1872, 0.0900, 0.1002], + [ 0.0291, -0.2221, -0.0705, ..., -0.1672, -0.1343, -0.1285], + [-0.3151, -0.1001, -0.0734, ..., 0.1111, -0.3508, 0.1214]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 9.3132e-10, ..., 3.2596e-09, + 2.3283e-10, 1.1642e-09], + [-2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 1.1642e-09, + 1.8626e-09, 5.5879e-09], + [ 0.0000e+00, 2.3283e-10, 4.6566e-10, ..., 1.8626e-09, + 2.3283e-10, 9.3132e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.0256e-08, + 7.9162e-09, 3.3062e-08], + [-1.3970e-09, 0.0000e+00, 2.3283e-10, ..., 2.3283e-09, + 4.6566e-10, 9.3132e-10], + [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., -3.6089e-08, + -1.0477e-08, -4.7730e-08]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0195, -0.0323, 0.0209, -0.0285, 0.0405, 0.0167, -0.0010, -0.0069, + -0.0239, -0.0155], device='cuda:0'), grad: tensor([ 1.3271e-08, 1.8859e-08, -3.4925e-09, -4.0978e-08, 7.3574e-08, + 1.3039e-08, -4.3074e-08, 9.8487e-08, -2.4680e-08, -9.0105e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 220.42, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4188 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.16 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.3285, 0.2633, -0.0327, ..., -0.1734, 0.0125, -0.0005], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0777, -0.0989], + [-0.2689, -0.3239, 0.0584, ..., -0.1575, -0.0821, -0.2562], + ..., + [ 0.0889, 0.2055, -0.1077, ..., -0.1872, 0.0900, 0.1002], + [ 0.0291, -0.2221, -0.0705, ..., -0.1673, -0.1343, -0.1286], + [-0.3152, -0.1001, -0.0734, ..., 0.1111, -0.3508, 0.1215]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 8.1491e-10, 1.1642e-10, ..., 1.0477e-09, + 0.0000e+00, 1.0477e-09], + [ 4.2026e-08, 4.8429e-08, 4.6566e-10, ..., 1.2806e-09, + 0.0000e+00, 4.6799e-08], + [ 1.2806e-09, 9.3132e-10, 0.0000e+00, ..., 1.0477e-09, + 0.0000e+00, 9.3132e-10], + ..., + [-4.8196e-08, -5.5414e-08, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, -4.9360e-08], + [ 1.0477e-09, 9.3132e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, -3.4925e-10], + [ 3.0501e-08, 1.8626e-09, 0.0000e+00, ..., 2.3399e-08, + 0.0000e+00, -1.3737e-08]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0195, -0.0323, 0.0209, -0.0285, 0.0405, 0.0166, -0.0010, -0.0069, + -0.0239, -0.0155], device='cuda:0'), grad: tensor([ 6.7521e-09, 1.8976e-07, 8.1491e-09, 2.6776e-09, -1.4086e-07, + 1.4901e-08, 1.6298e-09, -1.7835e-07, -2.8638e-08, 1.3306e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 220.39, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4366 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.16 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.3285, 0.2634, -0.0327, ..., -0.1735, 0.0125, -0.0006], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0777, -0.0989], + [-0.2690, -0.3239, 0.0584, ..., -0.1576, -0.0821, -0.2563], + ..., + [ 0.0889, 0.2055, -0.1077, ..., -0.1873, 0.0900, 0.1002], + [ 0.0291, -0.2221, -0.0706, ..., -0.1674, -0.1343, -0.1286], + [-0.3154, -0.1002, -0.0734, ..., 0.1110, -0.3508, 0.1215]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10], + [ 1.5134e-09, 5.8208e-10, 0.0000e+00, ..., 3.4925e-10, + 1.1642e-09, 4.0745e-09], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 8.1491e-10], + ..., + [-4.8894e-09, -1.2806e-09, 0.0000e+00, ..., 1.5134e-09, + -1.7462e-09, -1.0245e-08], + [ 1.8626e-09, 1.1642e-10, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 1.7462e-09], + [ 4.5984e-08, 3.4925e-10, 0.0000e+00, ..., 2.9104e-08, + 1.0477e-09, 8.4983e-09]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0195, -0.0323, 0.0209, -0.0285, 0.0406, 0.0166, -0.0010, -0.0069, + -0.0239, -0.0156], device='cuda:0'), grad: tensor([ 1.2806e-09, 1.5483e-08, 5.1223e-09, 6.3796e-08, -1.0617e-07, + -1.4459e-07, 6.8685e-09, -2.9569e-08, 8.6147e-09, 1.7800e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 220.27, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4307 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.14 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.3285, 0.2635, -0.0327, ..., -0.1736, 0.0125, -0.0006], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0778, -0.0989], + [-0.2690, -0.3240, 0.0584, ..., -0.1577, -0.0822, -0.2564], + ..., + [ 0.0890, 0.2056, -0.1077, ..., -0.1873, 0.0901, 0.1002], + [ 0.0291, -0.2222, -0.0706, ..., -0.1674, -0.1344, -0.1286], + [-0.3155, -0.1002, -0.0735, ..., 0.1110, -0.3509, 0.1215]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3039e-08, 0.0000e+00, ..., 3.1432e-09, + 0.0000e+00, 0.0000e+00], + [ 1.2456e-08, 1.1642e-10, 3.4925e-10, ..., 1.0477e-09, + 3.4925e-10, 2.2119e-09], + [ 4.1910e-09, 2.3283e-10, 5.8208e-10, ..., 2.0955e-09, + 9.3132e-10, 4.0745e-09], + ..., + [-4.5402e-09, 0.0000e+00, -1.1642e-09, ..., 0.0000e+00, + 1.2806e-09, -4.7730e-09], + [-3.2131e-08, 4.6566e-10, 1.1642e-10, ..., 6.9849e-10, + 3.4925e-10, 1.0477e-09], + [ 1.2806e-09, 2.3283e-10, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 3.4925e-10]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0194, -0.0323, 0.0209, -0.0284, 0.0406, 0.0166, -0.0010, -0.0069, + -0.0239, -0.0156], device='cuda:0'), grad: tensor([-6.7521e-09, 2.0918e-06, -2.0005e-06, 6.0536e-09, 9.3132e-09, + 1.2678e-07, 2.1071e-08, -1.0128e-08, -2.3679e-07, 1.0245e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 220.34, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4189 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.14 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.3285, 0.2636, -0.0327, ..., -0.1738, 0.0125, -0.0006], + [ 0.0065, -0.1763, -0.0302, ..., 0.0526, -0.0778, -0.0990], + [-0.2691, -0.3240, 0.0584, ..., -0.1577, -0.0823, -0.2564], + ..., + [ 0.0890, 0.2056, -0.1077, ..., -0.1874, 0.0900, 0.1002], + [ 0.0291, -0.2222, -0.0706, ..., -0.1675, -0.1344, -0.1287], + [-0.3156, -0.1002, -0.0735, ..., 0.1111, -0.3509, 0.1215]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + [-1.2573e-08, 9.3132e-10, 0.0000e+00, ..., 6.9849e-10, + 9.3132e-10, 2.4447e-09], + [ 1.2922e-08, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, 4.6566e-10], + ..., + [-2.4913e-08, -3.3760e-08, 0.0000e+00, ..., -8.4983e-09, + -2.7241e-08, -8.3004e-08], + [ 3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 3.4925e-10, 8.1491e-10], + [ 2.6659e-08, 3.1898e-08, 0.0000e+00, ..., 1.1758e-08, + 2.5728e-08, 7.7882e-08]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0193, -0.0323, 0.0209, -0.0284, 0.0406, 0.0166, -0.0009, -0.0069, + -0.0239, -0.0156], device='cuda:0'), grad: tensor([ 1.8626e-09, -1.9511e-07, 1.8370e-07, -8.1491e-10, -9.1968e-09, + -4.6799e-08, 5.4017e-08, -2.5122e-07, 8.1491e-10, 2.7055e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 220.92, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4507 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.3285, 0.2637, -0.0327, ..., -0.1740, 0.0125, -0.0007], + [ 0.0066, -0.1763, -0.0302, ..., 0.0525, -0.0779, -0.0990], + [-0.2692, -0.3241, 0.0584, ..., -0.1578, -0.0824, -0.2565], + ..., + [ 0.0891, 0.2056, -0.1077, ..., -0.1874, 0.0901, 0.1002], + [ 0.0291, -0.2222, -0.0706, ..., -0.1676, -0.1345, -0.1288], + [-0.3157, -0.1002, -0.0735, ..., 0.1111, -0.3511, 0.1216]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-10, 1.1642e-10], + [ 8.1491e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.1642e-10, 5.8208e-10], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, 2.3283e-10], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.0955e-09, + 1.1642e-10, 1.7462e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 1.6298e-09, 1.6298e-09], + [ 1.2340e-08, 3.4925e-10, 0.0000e+00, ..., 1.3970e-08, + 0.0000e+00, -6.1700e-09]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0192, -0.0323, 0.0209, -0.0284, 0.0406, 0.0166, -0.0009, -0.0069, + -0.0239, -0.0156], device='cuda:0'), grad: tensor([ 1.1642e-09, 7.3342e-09, -1.8626e-09, -1.1409e-08, -5.6461e-08, + 3.3760e-09, 3.0268e-09, 7.9162e-09, 1.3737e-08, 4.1095e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 220.52, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3999 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.14 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.3286, 0.2638, -0.0327, ..., -0.1741, 0.0125, -0.0007], + [ 0.0066, -0.1763, -0.0302, ..., 0.0525, -0.0780, -0.0991], + [-0.2693, -0.3241, 0.0585, ..., -0.1578, -0.0824, -0.2566], + ..., + [ 0.0891, 0.2056, -0.1078, ..., -0.1874, 0.0902, 0.1003], + [ 0.0291, -0.2223, -0.0706, ..., -0.1677, -0.1345, -0.1288], + [-0.3159, -0.1003, -0.0735, ..., 0.1111, -0.3511, 0.1216]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.1642e-09, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, 1.3970e-09], + [ 3.6205e-08, 4.5402e-08, 2.3283e-10, ..., 5.8208e-10, + 1.7462e-09, 4.3889e-08], + [ 6.9849e-10, 9.3132e-10, -1.1642e-10, ..., 2.3283e-10, + 1.8626e-09, 2.4447e-09], + ..., + [-4.4936e-08, -5.6112e-08, -1.0477e-09, ..., 2.9104e-09, + 1.4203e-08, -4.3889e-08], + [ 8.1491e-10, 1.0477e-09, 0.0000e+00, ..., 4.6566e-10, + 2.6193e-08, 2.7474e-08], + [ 4.1910e-09, 4.0745e-09, 5.8208e-10, ..., -4.4238e-09, + 3.1432e-09, 2.2119e-09]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0191, -0.0323, 0.0209, -0.0284, 0.0407, 0.0166, -0.0009, -0.0069, + -0.0239, -0.0156], device='cuda:0'), grad: tensor([ 7.5670e-09, 2.4191e-07, -7.1246e-08, -1.9674e-07, 1.2107e-08, + 3.0734e-08, 4.6566e-10, -1.4354e-07, 1.2666e-07, 8.0327e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 220.70, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3946 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.11 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.3286, 0.2639, -0.0327, ..., -0.1742, 0.0125, -0.0007], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0780, -0.0992], + [-0.2694, -0.3242, 0.0585, ..., -0.1578, -0.0825, -0.2567], + ..., + [ 0.0892, 0.2056, -0.1078, ..., -0.1876, 0.0901, 0.1002], + [ 0.0291, -0.2223, -0.0706, ..., -0.1678, -0.1346, -0.1289], + [-0.3160, -0.1003, -0.0735, ..., 0.1111, -0.3512, 0.1217]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 3.4925e-10, 8.1491e-10], + [-4.8894e-09, -2.3283e-10, 0.0000e+00, ..., -6.9849e-10, + 4.6566e-10, 1.5134e-09], + [ 2.3283e-10, 1.1642e-10, 0.0000e+00, ..., 2.3283e-10, + 8.1491e-10, 1.3970e-09], + ..., + [ 3.0268e-09, 2.3283e-10, 0.0000e+00, ..., 5.4715e-09, + 2.3283e-10, 7.7998e-09], + [ 1.6298e-09, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 3.4925e-10, 1.1642e-09], + [ 2.5611e-09, 2.3283e-10, 0.0000e+00, ..., -1.8626e-08, + 4.6566e-10, -2.1304e-08]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0191, -0.0324, 0.0209, -0.0283, 0.0407, 0.0165, -0.0009, -0.0069, + -0.0240, -0.0155], device='cuda:0'), grad: tensor([ 6.4028e-09, -1.6182e-08, 8.9640e-09, -6.2282e-08, 3.3062e-08, + 4.2841e-08, 3.0268e-09, 3.6554e-08, 9.5461e-09, -4.8778e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 220.67, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4308 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.3286, 0.2640, -0.0327, ..., -0.1743, 0.0125, -0.0007], + [ 0.0065, -0.1763, -0.0302, ..., 0.0525, -0.0780, -0.0992], + [-0.2695, -0.3243, 0.0585, ..., -0.1579, -0.0826, -0.2569], + ..., + [ 0.0893, 0.2057, -0.1078, ..., -0.1877, 0.0902, 0.1002], + [ 0.0291, -0.2223, -0.0706, ..., -0.1679, -0.1346, -0.1289], + [-0.3161, -0.1003, -0.0735, ..., 0.1112, -0.3512, 0.1217]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-5.8208e-10, -6.9849e-10, 0.0000e+00, ..., 3.4925e-10, + -2.3283e-10, -9.3132e-10], + [ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 5.9372e-09, 4.6566e-10, 0.0000e+00, ..., 7.5670e-09, + 1.1642e-10, 5.8208e-10]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0191, -0.0324, 0.0209, -0.0284, 0.0407, 0.0166, -0.0009, -0.0070, + -0.0240, -0.0155], device='cuda:0'), grad: tensor([ 8.1491e-10, 1.7462e-09, 1.0477e-09, 2.7940e-09, -2.1770e-08, + -3.0966e-08, 2.6077e-08, -1.9791e-09, 2.2119e-09, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 220.35, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4332 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.14 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.3286, 0.2642, -0.0327, ..., -0.1744, 0.0125, -0.0008], + [ 0.0065, -0.1764, -0.0302, ..., 0.0526, -0.0781, -0.0993], + [-0.2695, -0.3244, 0.0585, ..., -0.1579, -0.0827, -0.2571], + ..., + [ 0.0894, 0.2057, -0.1078, ..., -0.1877, 0.0902, 0.1003], + [ 0.0291, -0.2224, -0.0706, ..., -0.1680, -0.1347, -0.1290], + [-0.3163, -0.1003, -0.0735, ..., 0.1112, -0.3513, 0.1218]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 2.3283e-10], + [ 1.2806e-09, 2.3283e-09, 0.0000e+00, ..., 3.4925e-10, + 8.1491e-10, 2.7940e-09], + [ 8.1491e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 2.4447e-09], + ..., + [-4.0745e-09, -9.8953e-09, 0.0000e+00, ..., 2.3283e-10, + -2.2119e-09, -1.0827e-08], + [ 9.3132e-10, 2.2119e-09, 0.0000e+00, ..., 0.0000e+00, + 1.2806e-09, 3.2596e-09], + [ 1.3970e-09, 2.6776e-09, 0.0000e+00, ..., 5.8208e-10, + 1.1642e-09, 3.4925e-09]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0191, -0.0324, 0.0209, -0.0283, 0.0407, 0.0166, -0.0009, -0.0069, + -0.0240, -0.0155], device='cuda:0'), grad: tensor([ 6.9849e-10, 1.2689e-08, 8.9640e-09, 8.2888e-08, 1.5134e-09, + -9.5111e-08, 8.3819e-09, -3.7951e-08, 1.4901e-08, 1.3388e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 220.27, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4232 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.13 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.3286, 0.2643, -0.0327, ..., -0.1745, 0.0125, -0.0008], + [ 0.0065, -0.1764, -0.0302, ..., 0.0526, -0.0781, -0.0993], + [-0.2696, -0.3245, 0.0585, ..., -0.1579, -0.0828, -0.2571], + ..., + [ 0.0894, 0.2057, -0.1078, ..., -0.1878, 0.0902, 0.1002], + [ 0.0290, -0.2225, -0.0706, ..., -0.1681, -0.1347, -0.1291], + [-0.3164, -0.1003, -0.0735, ..., 0.1112, -0.3513, 0.1218]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.9791e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 6.9849e-10, + 1.1642e-10, 2.3283e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, 3.4925e-10], + ..., + [ 6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 1.8626e-09, + 1.1642e-10, 1.0477e-09], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 3.4925e-10, + 1.1642e-10, 2.3283e-10], + [ 5.3551e-09, 1.2806e-09, 0.0000e+00, ..., 8.1491e-09, + 1.1642e-10, -8.1491e-10]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0190, -0.0324, 0.0209, -0.0283, 0.0407, 0.0165, -0.0008, -0.0070, + -0.0240, -0.0155], device='cuda:0'), grad: tensor([-2.4447e-09, 1.5600e-08, -1.2689e-08, 1.6531e-08, -3.0617e-08, + 0.0000e+00, 6.1700e-09, 9.4296e-09, -1.7695e-08, 2.7125e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 220.81, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4068 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.13 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.3286, 0.2644, -0.0327, ..., -0.1746, 0.0125, -0.0008], + [ 0.0066, -0.1764, -0.0302, ..., 0.0526, -0.0782, -0.0994], + [-0.2697, -0.3246, 0.0585, ..., -0.1580, -0.0829, -0.2573], + ..., + [ 0.0895, 0.2058, -0.1078, ..., -0.1879, 0.0903, 0.1003], + [ 0.0290, -0.2225, -0.0706, ..., -0.1682, -0.1347, -0.1292], + [-0.3165, -0.1004, -0.0735, ..., 0.1113, -0.3514, 0.1219]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 8.1491e-10], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.6776e-09, + 1.1642e-10, 2.3283e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 2.3283e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.6357e-09, + 0.0000e+00, 5.8208e-09], + [-3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 8.8476e-09, + 1.1642e-10, 7.6834e-09], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., -5.5996e-08, + 0.0000e+00, -4.9826e-08]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0190, -0.0324, 0.0209, -0.0283, 0.0407, 0.0165, -0.0008, -0.0070, + -0.0240, -0.0155], device='cuda:0'), grad: tensor([ 1.7113e-08, 7.3458e-08, -1.1560e-07, 4.5868e-08, 1.8510e-07, + 6.1700e-09, 2.2119e-09, 3.4575e-08, 4.0163e-08, -2.7241e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 220.46, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3943 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.15 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.3286, 0.2645, -0.0327, ..., -0.1747, 0.0125, -0.0008], + [ 0.0066, -0.1764, -0.0302, ..., 0.0527, -0.0783, -0.0995], + [-0.2698, -0.3247, 0.0585, ..., -0.1580, -0.0829, -0.2574], + ..., + [ 0.0896, 0.2058, -0.1078, ..., -0.1879, 0.0903, 0.1003], + [ 0.0290, -0.2225, -0.0707, ..., -0.1683, -0.1348, -0.1292], + [-0.3166, -0.1004, -0.0735, ..., 0.1113, -0.3514, 0.1219]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 5.8208e-10, ..., 1.6298e-09, + 0.0000e+00, 1.1642e-10], + [ 2.3283e-09, 4.6566e-10, 1.0477e-09, ..., 3.9581e-09, + 1.1642e-10, 6.9849e-10], + [ 1.1642e-10, 1.1642e-10, 2.3283e-10, ..., 6.9849e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.4925e-10, -6.9849e-10, 0.0000e+00, ..., 1.6298e-09, + -1.1642e-10, -4.6566e-10], + [ 0.0000e+00, 1.1642e-10, 1.1642e-10, ..., 5.8208e-10, + 1.1642e-10, 1.1642e-10], + [ 4.6566e-10, 4.6566e-10, 1.1642e-10, ..., -6.9849e-10, + 1.1642e-10, -5.8208e-10]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0190, -0.0325, 0.0210, -0.0283, 0.0407, 0.0165, -0.0008, -0.0070, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([ 6.0536e-09, 9.4296e-09, 3.2596e-09, 1.1642e-10, 1.0361e-08, + 1.9791e-09, -2.8405e-08, 4.4238e-09, 2.7940e-09, -4.6566e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 220.38, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4331 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.18 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.3286, 0.2646, -0.0327, ..., -0.1748, 0.0125, -0.0009], + [ 0.0066, -0.1764, -0.0302, ..., 0.0526, -0.0783, -0.0995], + [-0.2699, -0.3248, 0.0585, ..., -0.1580, -0.0830, -0.2575], + ..., + [ 0.0897, 0.2059, -0.1078, ..., -0.1879, 0.0904, 0.1003], + [ 0.0290, -0.2226, -0.0707, ..., -0.1683, -0.1348, -0.1293], + [-0.3167, -0.1005, -0.0735, ..., 0.1113, -0.3515, 0.1219]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., -8.1491e-09, + 0.0000e+00, 2.0955e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 1.1642e-10], + ..., + [-8.2655e-09, -2.5611e-09, 0.0000e+00, ..., -3.7253e-09, + 0.0000e+00, -2.3516e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 1.1642e-10], + [ 2.7940e-09, 1.1642e-09, 0.0000e+00, ..., -4.3074e-09, + 0.0000e+00, 2.3283e-09]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0189, -0.0325, 0.0210, -0.0283, 0.0407, 0.0165, -0.0008, -0.0069, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([ 1.6298e-09, -2.8405e-08, 8.2655e-09, 2.1420e-08, 5.0757e-08, + 4.2957e-08, 4.6683e-08, -4.8894e-08, -1.1176e-07, 3.0384e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 220.37, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4393 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.3286, 0.2648, -0.0327, ..., -0.1749, 0.0125, -0.0009], + [ 0.0066, -0.1765, -0.0302, ..., 0.0527, -0.0785, -0.0996], + [-0.2700, -0.3248, 0.0585, ..., -0.1581, -0.0831, -0.2576], + ..., + [ 0.0899, 0.2059, -0.1078, ..., -0.1880, 0.0906, 0.1003], + [ 0.0289, -0.2226, -0.0707, ..., -0.1684, -0.1348, -0.1293], + [-0.3169, -0.1006, -0.0735, ..., 0.1113, -0.3516, 0.1220]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.7789e-09, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-10, 8.1491e-10, 2.3283e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 2.3283e-10, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 1.1642e-10, 3.4925e-10], + ..., + [-5.8208e-10, -1.8626e-09, 5.8208e-10, ..., 2.7940e-09, + -3.4925e-10, -9.3132e-10], + [ 1.1642e-10, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 3.4925e-10], + [ 1.8626e-09, 8.6147e-09, 2.3283e-10, ..., -2.5611e-09, + 2.3283e-10, -8.1491e-10]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0189, -0.0325, 0.0210, -0.0283, 0.0407, 0.0165, -0.0008, -0.0069, + -0.0242, -0.0155], device='cuda:0'), grad: tensor([-2.7241e-08, 8.3819e-09, -2.3283e-10, 6.9849e-10, 1.5134e-09, + 3.6089e-09, 8.1491e-10, 1.1642e-09, 2.6776e-09, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 220.41, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4358 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.16 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.3286, 0.2649, -0.0327, ..., -0.1749, 0.0125, -0.0009], + [ 0.0066, -0.1765, -0.0302, ..., 0.0527, -0.0785, -0.0996], + [-0.2701, -0.3249, 0.0585, ..., -0.1581, -0.0831, -0.2577], + ..., + [ 0.0900, 0.2060, -0.1078, ..., -0.1880, 0.0906, 0.1004], + [ 0.0289, -0.2227, -0.0707, ..., -0.1685, -0.1349, -0.1294], + [-0.3169, -0.1007, -0.0736, ..., 0.1113, -0.3516, 0.1220]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.0047e-08, 0.0000e+00, ..., -2.9104e-09, + -8.1491e-10, -2.3283e-10], + [ 2.3283e-10, 3.3760e-09, 0.0000e+00, ..., 3.4925e-10, + 2.3283e-10, 4.6566e-10], + [ 1.1642e-10, 3.3760e-09, 0.0000e+00, ..., 3.4925e-10, + 3.4925e-10, 4.6566e-10], + ..., + [ 1.1642e-10, 1.1642e-09, 0.0000e+00, ..., 5.8208e-10, + 1.1642e-10, 8.1491e-10], + [-4.6566e-10, 1.0710e-08, 0.0000e+00, ..., 1.0477e-09, + 2.3283e-10, 2.3283e-10], + [ 4.6566e-10, 1.2456e-08, 0.0000e+00, ..., -1.9791e-09, + 4.6566e-10, -3.2596e-09]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0190, -0.0325, 0.0210, -0.0283, 0.0407, 0.0165, -0.0008, -0.0069, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([-9.2434e-08, 1.5134e-08, 7.1013e-09, 6.6357e-09, 3.6089e-09, + 3.4925e-09, 1.4319e-08, 8.6147e-09, 1.5250e-08, 2.4564e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 220.47, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4224 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.15 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.3286, 0.2651, -0.0327, ..., -0.1750, 0.0125, -0.0009], + [ 0.0065, -0.1765, -0.0302, ..., 0.0527, -0.0785, -0.0997], + [-0.2701, -0.3250, 0.0585, ..., -0.1581, -0.0831, -0.2578], + ..., + [ 0.0901, 0.2061, -0.1078, ..., -0.1880, 0.0907, 0.1004], + [ 0.0290, -0.2227, -0.0707, ..., -0.1685, -0.1349, -0.1294], + [-0.3170, -0.1008, -0.0736, ..., 0.1113, -0.3517, 0.1220]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-10, 5.8208e-10], + [ 3.4925e-10, 5.8208e-10, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 6.9849e-10], + ..., + [-4.6566e-10, -6.9849e-10, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, -5.8208e-10], + [ 1.1642e-10, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 1.1642e-10, 6.9849e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0190, -0.0325, 0.0209, -0.0283, 0.0407, 0.0165, -0.0008, -0.0069, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([ 2.2119e-09, 5.8208e-09, 6.8685e-09, 7.7998e-09, 3.6089e-09, + 6.1700e-09, -5.9372e-09, -4.6566e-09, -1.2573e-08, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 220.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4197 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.16 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.3286, 0.2654, -0.0327, ..., -0.1750, 0.0125, -0.0009], + [ 0.0065, -0.1766, -0.0302, ..., 0.0527, -0.0786, -0.0998], + [-0.2702, -0.3251, 0.0585, ..., -0.1581, -0.0832, -0.2579], + ..., + [ 0.0901, 0.2061, -0.1078, ..., -0.1882, 0.0907, 0.1004], + [ 0.0290, -0.2228, -0.0707, ..., -0.1686, -0.1349, -0.1295], + [-0.3171, -0.1008, -0.0736, ..., 0.1114, -0.3518, 0.1221]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.9849e-10, 0.0000e+00, ..., -5.8208e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + ..., + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 0.0000e+00], + [ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 1.1642e-10], + [ 1.6298e-09, 6.9849e-10, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0191, -0.0325, 0.0209, -0.0283, 0.0406, 0.0165, -0.0008, -0.0069, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([ 1.2806e-09, 8.2306e-08, -8.2538e-08, 1.2224e-08, -1.6298e-09, + -4.6799e-08, 2.4098e-08, 4.4238e-09, 6.4028e-09, 1.2806e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 220.66, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4251 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.14 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.3286, 0.2656, -0.0327, ..., -0.1751, 0.0125, -0.0009], + [ 0.0065, -0.1766, -0.0302, ..., 0.0527, -0.0786, -0.0998], + [-0.2703, -0.3251, 0.0585, ..., -0.1582, -0.0833, -0.2580], + ..., + [ 0.0902, 0.2061, -0.1078, ..., -0.1882, 0.0908, 0.1004], + [ 0.0290, -0.2228, -0.0707, ..., -0.1686, -0.1349, -0.1295], + [-0.3171, -0.1009, -0.0736, ..., 0.1114, -0.3518, 0.1221]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-10, 1.1642e-10], + [ 1.1642e-10, 8.1491e-10, 3.4925e-10, ..., -3.4925e-10, + 3.0268e-09, 2.9104e-09], + [ 4.6566e-10, 9.3132e-10, -3.3760e-09, ..., 4.6566e-10, + 2.9104e-09, 3.6089e-09], + ..., + [-1.6298e-09, -4.0745e-09, 3.0268e-09, ..., 5.8208e-10, + 2.4447e-09, -3.4925e-10], + [ 2.3283e-10, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 5.8208e-10, 5.8208e-10], + [ 1.0477e-09, 2.2119e-09, 0.0000e+00, ..., 1.1642e-10, + 1.0477e-09, 1.8626e-09]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0192, -0.0325, 0.0210, -0.0283, 0.0406, 0.0165, -0.0008, -0.0069, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([ 1.2806e-09, 2.7008e-08, -5.8208e-08, -6.1584e-08, 2.2119e-09, + 1.3853e-08, 8.4983e-09, 6.2981e-08, 6.5193e-09, 8.9640e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 220.45, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4385 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.3286, 0.2658, -0.0327, ..., -0.1752, 0.0125, -0.0009], + [ 0.0065, -0.1766, -0.0302, ..., 0.0527, -0.0787, -0.0998], + [-0.2703, -0.3252, 0.0586, ..., -0.1582, -0.0833, -0.2581], + ..., + [ 0.0902, 0.2061, -0.1078, ..., -0.1882, 0.0908, 0.1004], + [ 0.0289, -0.2228, -0.0707, ..., -0.1687, -0.1349, -0.1296], + [-0.3172, -0.1010, -0.0736, ..., 0.1114, -0.3518, 0.1221]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 3.4925e-10], + [ 1.9325e-08, 2.5262e-08, 2.3283e-10, ..., 3.3760e-09, + 0.0000e+00, 4.5286e-08], + [ 1.0477e-09, 1.2806e-09, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 2.0955e-09], + ..., + [-2.0023e-08, -2.8987e-08, 0.0000e+00, ..., 1.0594e-08, + 0.0000e+00, -3.7951e-08], + [ 3.4925e-10, 5.8208e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 8.1491e-10], + [ 3.0268e-09, 2.4447e-09, 0.0000e+00, ..., -1.0827e-08, + 0.0000e+00, -1.4435e-08]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0192, -0.0325, 0.0210, -0.0284, 0.0406, 0.0165, -0.0007, -0.0070, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([-2.5611e-09, 1.7288e-07, 1.0943e-08, 5.4715e-09, -1.1758e-08, + 4.6566e-10, 5.8208e-09, -1.5192e-07, 4.3074e-09, -2.3632e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 220.91, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3981 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.16 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.3286, 0.2659, -0.0327, ..., -0.1753, 0.0125, -0.0009], + [ 0.0065, -0.1766, -0.0302, ..., 0.0528, -0.0787, -0.0998], + [-0.2704, -0.3253, 0.0586, ..., -0.1583, -0.0834, -0.2581], + ..., + [ 0.0902, 0.2062, -0.1078, ..., -0.1883, 0.0908, 0.1004], + [ 0.0289, -0.2229, -0.0707, ..., -0.1688, -0.1350, -0.1296], + [-0.3174, -0.1010, -0.0736, ..., 0.1114, -0.3518, 0.1222]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.2387e-09, + 0.0000e+00, 0.0000e+00], + [ 4.3074e-09, 2.6776e-09, 0.0000e+00, ..., 1.1642e-10, + 1.1642e-10, 2.0955e-09], + [ 2.3283e-10, 1.1642e-10, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + ..., + [-3.3760e-09, -2.2119e-09, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, -1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 1.1642e-10], + [ 4.6566e-10, 3.4925e-10, 0.0000e+00, ..., -3.4925e-09, + 2.3283e-10, -2.3283e-09]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0192, -0.0325, 0.0210, -0.0284, 0.0406, 0.0165, -0.0007, -0.0070, + -0.0241, -0.0155], device='cuda:0'), grad: tensor([ 2.9104e-08, 1.9209e-08, 3.2596e-09, 6.9849e-10, 1.3388e-08, + 1.5949e-08, -4.8778e-08, -8.7311e-09, 3.9581e-09, -6.2864e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 221.11, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4220 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.19 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.3287, 0.2661, -0.0327, ..., -0.1753, 0.0125, -0.0009], + [ 0.0065, -0.1766, -0.0302, ..., 0.0529, -0.0787, -0.0999], + [-0.2705, -0.3254, 0.0586, ..., -0.1584, -0.0834, -0.2582], + ..., + [ 0.0903, 0.2062, -0.1078, ..., -0.1883, 0.0909, 0.1004], + [ 0.0290, -0.2229, -0.0707, ..., -0.1688, -0.1350, -0.1297], + [-0.3175, -0.1011, -0.0736, ..., 0.1114, -0.3519, 0.1222]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7462e-10, + 5.8208e-11, 2.3283e-10], + [ 1.1642e-10, 4.0745e-10, 5.8208e-11, ..., 1.7462e-10, + 5.8208e-10, 2.2119e-09], + [ 4.6566e-10, 6.9849e-10, -2.4447e-09, ..., 2.3283e-10, + 2.2701e-09, 2.8522e-09], + ..., + [-4.1910e-09, -8.7311e-10, 5.8208e-11, ..., 1.9209e-09, + -2.2701e-09, -8.2073e-09], + [ 1.1642e-10, 1.1642e-10, 1.8626e-09, ..., 2.9104e-10, + 3.4925e-10, 5.8208e-10], + [ 3.6089e-09, 4.6566e-10, 1.7462e-10, ..., -1.3097e-08, + 3.1432e-09, -9.0804e-09]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0193, -0.0325, 0.0209, -0.0284, 0.0407, 0.0165, -0.0007, -0.0070, + -0.0241, -0.0156], device='cuda:0'), grad: tensor([ 2.6193e-09, 2.7416e-08, -4.5868e-08, -1.5367e-08, 1.1409e-08, + 4.2783e-08, -5.5879e-09, -7.5088e-09, 2.7183e-08, -3.1316e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 220.73, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4155 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.18 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.3287, 0.2664, -0.0327, ..., -0.1754, 0.0125, -0.0009], + [ 0.0066, -0.1766, -0.0302, ..., 0.0529, -0.0788, -0.0999], + [-0.2705, -0.3254, 0.0586, ..., -0.1584, -0.0834, -0.2583], + ..., + [ 0.0903, 0.2063, -0.1078, ..., -0.1884, 0.0909, 0.1005], + [ 0.0291, -0.2230, -0.0707, ..., -0.1689, -0.1350, -0.1297], + [-0.3177, -0.1012, -0.0736, ..., 0.1115, -0.3520, 0.1222]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, -2.5029e-09, 0.0000e+00, ..., 2.3283e-10, + -5.8208e-11, 1.7462e-10], + [ 8.1491e-10, 5.8208e-10, 0.0000e+00, ..., 5.8208e-10, + 3.4925e-10, 1.1059e-09], + [ 5.2387e-10, 5.8208e-10, 0.0000e+00, ..., 1.7462e-10, + 0.0000e+00, 5.8208e-10], + ..., + [-1.3388e-09, -9.8953e-10, 0.0000e+00, ..., 4.5402e-09, + -1.0477e-09, -3.4925e-10], + [-8.1491e-10, 5.8208e-10, 0.0000e+00, ..., 4.6566e-10, + 1.7462e-10, 9.3132e-10], + [ 1.2224e-09, 7.5670e-10, 0.0000e+00, ..., -5.4715e-09, + 5.8208e-10, -3.8417e-09]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0195, -0.0325, 0.0210, -0.0284, 0.0407, 0.0164, -0.0007, -0.0070, + -0.0240, -0.0156], device='cuda:0'), grad: tensor([-6.0536e-09, 8.2073e-09, 2.1537e-09, -8.1491e-10, 3.5507e-09, + 1.1001e-08, 4.1910e-09, 3.3760e-09, -9.6625e-09, 1.1642e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 220.55, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4205 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.3287, 0.2666, -0.0327, ..., -0.1754, 0.0125, -0.0009], + [ 0.0066, -0.1767, -0.0302, ..., 0.0530, -0.0789, -0.1000], + [-0.2706, -0.3255, 0.0586, ..., -0.1585, -0.0835, -0.2584], + ..., + [ 0.0904, 0.2063, -0.1078, ..., -0.1884, 0.0910, 0.1005], + [ 0.0291, -0.2230, -0.0707, ..., -0.1690, -0.1351, -0.1298], + [-0.3179, -0.1013, -0.0736, ..., 0.1115, -0.3520, 0.1223]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.4377e-08, 5.8208e-11, ..., 7.5670e-10, + -1.6298e-09, 6.9849e-10], + [-2.7940e-09, 9.8953e-10, 2.3283e-10, ..., -2.6776e-09, + 1.7462e-10, 4.6566e-10], + [ 1.1642e-10, 1.8044e-09, 0.0000e+00, ..., 3.4925e-10, + 2.3283e-10, 1.1642e-10], + ..., + [ 2.7358e-09, 1.1642e-09, 0.0000e+00, ..., 1.1350e-08, + 1.9791e-09, 1.0710e-08], + [ 1.7462e-10, 4.8894e-09, 0.0000e+00, ..., 5.2387e-10, + 6.9849e-10, 5.8208e-10], + [ 1.4552e-09, 4.0163e-09, 0.0000e+00, ..., -1.4668e-08, + -3.4925e-09, -2.2643e-08]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0196, -0.0325, 0.0210, -0.0284, 0.0407, 0.0164, -0.0008, -0.0070, + -0.0240, -0.0156], device='cuda:0'), grad: tensor([-2.9861e-08, -1.8335e-08, -4.6566e-09, 5.9372e-09, 2.5961e-08, + 8.4983e-09, -1.2806e-09, 5.8732e-08, 7.2760e-09, -3.8766e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 221.06, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.3986 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.3287, 0.2667, -0.0327, ..., -0.1755, 0.0125, -0.0010], + [ 0.0065, -0.1767, -0.0302, ..., 0.0530, -0.0789, -0.1000], + [-0.2707, -0.3256, 0.0586, ..., -0.1586, -0.0836, -0.2585], + ..., + [ 0.0905, 0.2064, -0.1078, ..., -0.1885, 0.0910, 0.1005], + [ 0.0291, -0.2231, -0.0707, ..., -0.1691, -0.1351, -0.1298], + [-0.3181, -0.1013, -0.0736, ..., 0.1114, -0.3521, 0.1223]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3970e-08, 0.0000e+00, ..., 4.6566e-10, + 0.0000e+00, 2.3283e-10], + [ 0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 3.4925e-10, + 1.0477e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 3.6089e-09, 2.4447e-09], + ..., + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 5.2387e-09, + 1.7462e-09, 7.9162e-09], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, + 9.3132e-10, 6.9849e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., -7.5670e-09, + 3.4925e-10, -9.8953e-09]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0196, -0.0325, 0.0210, -0.0284, 0.0408, 0.0164, -0.0008, -0.0070, + -0.0240, -0.0157], device='cuda:0'), grad: tensor([-2.5961e-08, 6.2864e-09, 1.2689e-08, -3.6671e-08, 1.1991e-08, + 2.3632e-08, 1.8626e-08, 2.6776e-08, 5.0059e-09, -2.7474e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 220.69, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4101 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.3287, 0.2669, -0.0327, ..., -0.1755, 0.0125, -0.0010], + [ 0.0065, -0.1767, -0.0302, ..., 0.0530, -0.0790, -0.1001], + [-0.2708, -0.3256, 0.0586, ..., -0.1587, -0.0837, -0.2586], + ..., + [ 0.0905, 0.2065, -0.1078, ..., -0.1885, 0.0911, 0.1005], + [ 0.0291, -0.2231, -0.0708, ..., -0.1692, -0.1351, -0.1299], + [-0.3182, -0.1015, -0.0736, ..., 0.1115, -0.3521, 0.1224]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 8.1491e-10], + [ 6.1700e-09, 4.8894e-09, 2.3283e-10, ..., 1.1642e-09, + 2.4447e-09, 9.1968e-09], + [ 3.9581e-09, 3.0268e-09, 2.3283e-10, ..., 1.1642e-10, + 1.5134e-09, 5.4715e-09], + ..., + [-3.5390e-08, -9.8953e-09, -1.4435e-08, ..., 7.2177e-09, + -4.7730e-09, -3.4343e-08], + [ 1.5134e-09, 1.0477e-09, 0.0000e+00, ..., 1.3970e-09, + 8.1491e-10, 3.7253e-09], + [ 2.2119e-09, 8.1491e-10, 6.9849e-10, ..., -3.6322e-08, + 4.6566e-10, -3.4575e-08]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0196, -0.0325, 0.0209, -0.0284, 0.0407, 0.0164, -0.0008, -0.0070, + -0.0240, -0.0156], device='cuda:0'), grad: tensor([ 3.9581e-09, 5.5064e-08, 2.0955e-08, -2.5495e-08, 1.3539e-07, + 1.0605e-07, -3.3760e-09, -2.0862e-07, 1.8976e-08, -9.5344e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 220.77, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4297 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.15 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.3287, 0.2670, -0.0327, ..., -0.1756, 0.0125, -0.0011], + [ 0.0065, -0.1768, -0.0302, ..., 0.0530, -0.0791, -0.1001], + [-0.2709, -0.3257, 0.0586, ..., -0.1587, -0.0837, -0.2587], + ..., + [ 0.0907, 0.2066, -0.1078, ..., -0.1885, 0.0912, 0.1006], + [ 0.0291, -0.2232, -0.0708, ..., -0.1693, -0.1352, -0.1300], + [-0.3183, -0.1015, -0.0736, ..., 0.1115, -0.3522, 0.1224]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.8208e-10, ..., 6.7521e-09, + 1.1642e-10, 1.1642e-10], + [ 1.1642e-10, 1.1642e-10, 2.3283e-10, ..., 2.2119e-09, + 5.8208e-10, 8.1491e-10], + [ 1.1642e-10, 0.0000e+00, 1.1642e-10, ..., 1.1642e-09, + 4.6566e-10, 3.4925e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2806e-09, + 1.0477e-09, 1.9791e-09], + [ 0.0000e+00, 2.3283e-10, 2.3283e-10, ..., 1.8626e-09, + 3.4925e-10, 2.3283e-10], + [ 4.6566e-10, 3.4925e-10, 0.0000e+00, ..., -2.3283e-09, + 2.3283e-10, -2.7940e-09]], device='cuda:0') +Epoch 475, bias, value: tensor([ 0.0195, -0.0326, 0.0209, -0.0284, 0.0408, 0.0165, -0.0007, -0.0070, + -0.0240, -0.0157], device='cuda:0'), grad: tensor([ 2.9686e-08, 1.3504e-08, 7.7998e-09, -1.0710e-08, 1.4552e-08, + 8.9640e-09, -6.6007e-08, 1.1292e-08, 1.0012e-08, -6.6357e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 220.86, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4258 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.17 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.3287, 0.2671, -0.0327, ..., -0.1757, 0.0125, -0.0011], + [ 0.0064, -0.1768, -0.0302, ..., 0.0530, -0.0791, -0.1002], + [-0.2709, -0.3257, 0.0586, ..., -0.1587, -0.0838, -0.2587], + ..., + [ 0.0907, 0.2066, -0.1078, ..., -0.1886, 0.0912, 0.1007], + [ 0.0291, -0.2232, -0.0708, ..., -0.1693, -0.1352, -0.1301], + [-0.3185, -0.1016, -0.0736, ..., 0.1115, -0.3522, 0.1224]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 1.1642e-10], + [ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., -1.0477e-09, + 0.0000e+00, -1.0477e-09]], device='cuda:0') +Epoch 476, bias, value: tensor([ 0.0195, -0.0326, 0.0210, -0.0285, 0.0408, 0.0165, -0.0007, -0.0070, + -0.0240, -0.0157], device='cuda:0'), grad: tensor([ 3.7253e-09, 2.5611e-09, 1.2806e-09, 1.0594e-08, 4.1910e-09, + 5.2038e-08, -6.5775e-08, 5.8208e-10, 2.5611e-09, -1.5134e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 220.58, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4280 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.15 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.3287, 0.2673, -0.0327, ..., -0.1758, 0.0125, -0.0011], + [ 0.0064, -0.1769, -0.0302, ..., 0.0531, -0.0792, -0.1003], + [-0.2711, -0.3258, 0.0585, ..., -0.1588, -0.0839, -0.2589], + ..., + [ 0.0909, 0.2067, -0.1078, ..., -0.1887, 0.0913, 0.1007], + [ 0.0291, -0.2232, -0.0708, ..., -0.1694, -0.1352, -0.1301], + [-0.3187, -0.1017, -0.0736, ..., 0.1115, -0.3523, 0.1225]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + [ 2.2119e-09, 3.4925e-10, 0.0000e+00, ..., 4.1910e-09, + 4.6566e-10, 8.1491e-10], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -1.9791e-09, 0.0000e+00], + ..., + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.6776e-09, + 0.0000e+00, 1.5134e-09], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 2.5029e-08, 1.1642e-10, 0.0000e+00, ..., 4.0280e-08, + 0.0000e+00, -3.4925e-09]], device='cuda:0') +Epoch 477, bias, value: tensor([ 0.0196, -0.0326, 0.0209, -0.0285, 0.0408, 0.0166, -0.0008, -0.0070, + -0.0240, -0.0158], device='cuda:0'), grad: tensor([ 3.2596e-09, 2.2352e-08, -2.8755e-08, 2.7474e-08, -1.8394e-07, + 1.1642e-10, 2.5029e-08, 9.0804e-09, 2.2119e-09, 1.4051e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 220.82, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4310 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.15 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.3287, 0.2675, -0.0327, ..., -0.1758, 0.0125, -0.0011], + [ 0.0064, -0.1769, -0.0302, ..., 0.0530, -0.0792, -0.1003], + [-0.2712, -0.3260, 0.0585, ..., -0.1588, -0.0839, -0.2590], + ..., + [ 0.0910, 0.2068, -0.1078, ..., -0.1887, 0.0913, 0.1007], + [ 0.0292, -0.2233, -0.0708, ..., -0.1695, -0.1353, -0.1302], + [-0.3189, -0.1018, -0.0736, ..., 0.1115, -0.3523, 0.1225]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 5.8208e-10, 1.6298e-09, 0.0000e+00, ..., -3.9581e-09, + 4.6566e-10, 2.5611e-09], + [ 3.4925e-10, 9.3132e-10, 0.0000e+00, ..., 1.5134e-09, + 3.4925e-10, 1.2806e-09], + ..., + [-2.0955e-09, -5.9372e-09, 0.0000e+00, ..., 6.1700e-09, + -1.9791e-09, -3.2596e-09], + [ 2.3283e-10, 8.1491e-10, 0.0000e+00, ..., 6.9849e-10, + 2.3283e-10, 1.2806e-09], + [ 1.0477e-09, 2.5611e-09, 0.0000e+00, ..., -7.4506e-09, + 6.9849e-10, -6.5193e-09]], device='cuda:0') +Epoch 478, bias, value: tensor([ 0.0197, -0.0326, 0.0209, -0.0286, 0.0409, 0.0166, -0.0008, -0.0070, + -0.0239, -0.0158], device='cuda:0'), grad: tensor([ 1.1642e-09, -1.7229e-08, 7.5670e-09, 3.9581e-09, 1.5949e-08, + 1.0477e-09, 1.1642e-09, 2.3283e-10, 6.5193e-09, -6.7521e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 220.75, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4375 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.16 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.3287, 0.2677, -0.0327, ..., -0.1759, 0.0125, -0.0012], + [ 0.0063, -0.1769, -0.0302, ..., 0.0530, -0.0793, -0.1004], + [-0.2713, -0.3260, 0.0585, ..., -0.1589, -0.0840, -0.2591], + ..., + [ 0.0911, 0.2069, -0.1078, ..., -0.1888, 0.0914, 0.1007], + [ 0.0292, -0.2233, -0.0708, ..., -0.1695, -0.1353, -0.1302], + [-0.3191, -0.1019, -0.0737, ..., 0.1116, -0.3524, 0.1226]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.9791e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 2.3283e-10, 4.6566e-10, 1.1642e-10, ..., 3.4925e-10, + 0.0000e+00, 3.4925e-10], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + ..., + [ 3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 7.3342e-09, + 0.0000e+00, 9.7789e-09], + [ 9.3132e-10, 1.0477e-09, 0.0000e+00, ..., 8.1491e-10, + 1.1642e-10, 8.1491e-10], + [ 2.3283e-10, 1.9791e-09, 0.0000e+00, ..., -8.2655e-09, + 0.0000e+00, -1.1758e-08]], device='cuda:0') +Epoch 479, bias, value: tensor([ 0.0197, -0.0327, 0.0209, -0.0287, 0.0409, 0.0167, -0.0009, -0.0070, + -0.0239, -0.0158], device='cuda:0'), grad: tensor([-1.2806e-09, 1.5716e-08, 3.2596e-09, 1.1339e-07, 3.0152e-08, + -3.3132e-07, 2.1874e-07, 3.9348e-08, -2.7241e-08, -3.1665e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 220.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4396 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.3287, 0.2680, -0.0327, ..., -0.1759, 0.0125, -0.0012], + [ 0.0063, -0.1770, -0.0302, ..., 0.0531, -0.0793, -0.1005], + [-0.2714, -0.3262, 0.0585, ..., -0.1589, -0.0841, -0.2593], + ..., + [ 0.0913, 0.2070, -0.1078, ..., -0.1889, 0.0915, 0.1008], + [ 0.0292, -0.2234, -0.0708, ..., -0.1696, -0.1354, -0.1303], + [-0.3193, -0.1020, -0.0737, ..., 0.1116, -0.3525, 0.1227]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 8.1491e-10], + [ 2.1071e-08, 3.1898e-08, 0.0000e+00, ..., 1.6298e-09, + 0.0000e+00, 5.6229e-08], + [ 9.3132e-10, 5.8208e-10, 0.0000e+00, ..., 5.8208e-10, + -2.3283e-10, 1.8626e-09], + ..., + [-2.4098e-08, -3.5041e-08, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, -5.6694e-08], + [ 1.1642e-10, 4.6566e-10, 0.0000e+00, ..., 1.2806e-09, + 0.0000e+00, 2.0955e-09], + [ 2.5611e-09, 2.2119e-09, 0.0000e+00, ..., -3.1549e-08, + 1.1642e-10, -2.5844e-08]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0199, -0.0327, 0.0209, -0.0287, 0.0409, 0.0167, -0.0009, -0.0070, + -0.0239, -0.0158], device='cuda:0'), grad: tensor([ 2.7940e-09, 2.1094e-07, -3.0268e-09, 3.3760e-09, 7.5088e-08, + -2.0955e-09, 3.6089e-09, -2.0058e-07, 4.4238e-09, -8.1607e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 220.76, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3925 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.11 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.3288, 0.2682, -0.0327, ..., -0.1760, 0.0125, -0.0012], + [ 0.0063, -0.1770, -0.0302, ..., 0.0531, -0.0794, -0.1006], + [-0.2715, -0.3263, 0.0586, ..., -0.1590, -0.0842, -0.2594], + ..., + [ 0.0914, 0.2071, -0.1078, ..., -0.1889, 0.0916, 0.1008], + [ 0.0292, -0.2234, -0.0708, ..., -0.1697, -0.1354, -0.1304], + [-0.3195, -0.1022, -0.0737, ..., 0.1116, -0.3525, 0.1227]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-09, 0.0000e+00, ..., 9.3132e-10, + -6.9849e-10, 1.5134e-09], + [ 2.3283e-10, 2.3283e-10, 1.1642e-10, ..., 2.5611e-09, + 0.0000e+00, 4.5402e-09], + [ 8.4983e-09, 3.4925e-10, 5.4715e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10], + ..., + [-8.9640e-09, 1.1642e-10, -5.7044e-09, ..., 1.1642e-09, + 1.1642e-10, 1.7462e-09], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 1.1642e-10], + [ 1.2806e-09, 2.7940e-09, 0.0000e+00, ..., -1.0477e-08, + 3.4925e-10, -1.8394e-08]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0201, -0.0327, 0.0209, -0.0287, 0.0410, 0.0167, -0.0009, -0.0070, + -0.0239, -0.0159], device='cuda:0'), grad: tensor([-6.6357e-09, 1.6764e-08, 5.6112e-08, -1.1642e-10, 1.8859e-08, + -1.0477e-09, 2.9104e-09, -5.4715e-08, 1.5134e-09, -4.0396e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 220.47, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4335 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.14 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.3288, 0.2684, -0.0327, ..., -0.1761, 0.0125, -0.0012], + [ 0.0063, -0.1771, -0.0302, ..., 0.0532, -0.0794, -0.1006], + [-0.2716, -0.3264, 0.0585, ..., -0.1590, -0.0843, -0.2595], + ..., + [ 0.0914, 0.2072, -0.1078, ..., -0.1890, 0.0916, 0.1009], + [ 0.0292, -0.2235, -0.0708, ..., -0.1699, -0.1354, -0.1304], + [-0.3196, -0.1022, -0.0737, ..., 0.1116, -0.3526, 0.1228]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0955e-09, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 5.8208e-10, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, + 0.0000e+00, 0.0000e+00], + [ 3.4925e-10, 2.3283e-10, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 1.1642e-10]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0201, -0.0327, 0.0209, -0.0288, 0.0410, 0.0168, -0.0009, -0.0070, + -0.0239, -0.0159], device='cuda:0'), grad: tensor([ 9.5461e-09, 7.7998e-09, -1.5134e-09, 3.1432e-09, 1.5134e-09, + 2.6776e-09, -3.1432e-08, 3.2596e-09, 2.1071e-08, 3.0268e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 220.52, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4126 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.17 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.3288, 0.2686, -0.0326, ..., -0.1762, 0.0125, -0.0012], + [ 0.0064, -0.1771, -0.0302, ..., 0.0532, -0.0795, -0.1007], + [-0.2717, -0.3265, 0.0585, ..., -0.1591, -0.0843, -0.2597], + ..., + [ 0.0915, 0.2072, -0.1078, ..., -0.1892, 0.0916, 0.1009], + [ 0.0291, -0.2235, -0.0708, ..., -0.1700, -0.1355, -0.1305], + [-0.3198, -0.1023, -0.0737, ..., 0.1117, -0.3526, 0.1229]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0827e-08, 0.0000e+00, ..., 2.3283e-10, + -1.2806e-09, -4.6566e-10], + [-5.8208e-10, 1.5134e-09, 0.0000e+00, ..., 8.1491e-10, + 8.1491e-10, 1.5134e-09], + [ 2.3283e-10, 1.1642e-09, 0.0000e+00, ..., 1.1642e-10, + 1.1642e-10, 2.3283e-10], + ..., + [ 8.1491e-10, 1.7462e-09, 0.0000e+00, ..., 1.6298e-09, + 2.5611e-09, 4.6566e-09], + [ 4.6566e-10, 2.3283e-09, 0.0000e+00, ..., 2.3283e-10, + 1.3970e-09, 1.0477e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., -5.5879e-09, + 6.4028e-09, 5.2387e-09]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0200, -0.0327, 0.0209, -0.0288, 0.0410, 0.0169, -0.0010, -0.0071, + -0.0240, -0.0159], device='cuda:0'), grad: tensor([-2.0606e-08, 5.5879e-09, 4.3074e-09, -7.4506e-08, 8.4983e-09, + 2.6077e-08, 8.7311e-09, 2.3865e-08, 1.3388e-08, 2.2585e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 220.50, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4212 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.3288, 0.2688, -0.0326, ..., -0.1764, 0.0125, -0.0013], + [ 0.0064, -0.1771, -0.0301, ..., 0.0534, -0.0795, -0.1008], + [-0.2718, -0.3266, 0.0585, ..., -0.1591, -0.0844, -0.2598], + ..., + [ 0.0915, 0.2073, -0.1078, ..., -0.1893, 0.0917, 0.1009], + [ 0.0291, -0.2236, -0.0708, ..., -0.1701, -0.1355, -0.1306], + [-0.3200, -0.1024, -0.0737, ..., 0.1118, -0.3526, 0.1230]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 2.3283e-10], + [ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 5.8208e-10, + 1.1642e-10, 5.8208e-10], + [ 1.2806e-09, 1.6298e-09, 0.0000e+00, ..., 2.3283e-10, + 5.8208e-10, 1.2806e-09], + ..., + [ 6.9849e-10, -1.1642e-09, 2.3283e-10, ..., 2.9104e-09, + 2.3283e-10, 1.2806e-09], + [-2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 1.0477e-09], + [ 2.2119e-09, 3.4925e-10, 1.1642e-10, ..., -4.1910e-09, + 0.0000e+00, -6.7521e-09]], device='cuda:0') +Epoch 484, bias, value: tensor([ 0.0200, -0.0327, 0.0209, -0.0288, 0.0410, 0.0169, -0.0009, -0.0072, + -0.0240, -0.0159], device='cuda:0'), grad: tensor([ 1.1642e-09, 4.0745e-09, 1.3970e-08, 1.1642e-10, -5.7044e-09, + 3.7253e-09, 3.4925e-09, 9.5461e-09, -2.2119e-09, -1.1409e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 220.54, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4116 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.19 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.3288, 0.2688, -0.0326, ..., -0.1764, 0.0125, -0.0013], + [ 0.0065, -0.1771, -0.0301, ..., 0.0534, -0.0796, -0.1009], + [-0.2719, -0.3266, 0.0585, ..., -0.1592, -0.0845, -0.2599], + ..., + [ 0.0915, 0.2073, -0.1079, ..., -0.1894, 0.0917, 0.1009], + [ 0.0291, -0.2236, -0.0709, ..., -0.1702, -0.1355, -0.1306], + [-0.3203, -0.1025, -0.0737, ..., 0.1117, -0.3527, 0.1231]], + device='cuda:0'), grad: tensor([[5.8208e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 3.4925e-10, 0.0000e+00, ..., 0.0000e+00, 1.1642e-10, + 0.0000e+00], + ..., + [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 1.1642e-10, + 1.1642e-10], + [3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.1642e-10, + 1.1642e-10], + [1.5134e-09, 2.3283e-10, 0.0000e+00, ..., 1.2806e-09, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 485, bias, value: tensor([ 0.0200, -0.0327, 0.0209, -0.0289, 0.0411, 0.0170, -0.0010, -0.0072, + -0.0240, -0.0160], device='cuda:0'), grad: tensor([ 9.4296e-09, 1.6298e-09, 1.0477e-09, 7.2177e-09, -1.6298e-09, + -2.5961e-08, 1.4901e-08, 1.6298e-09, -1.1874e-08, 8.0327e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 220.81, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4330 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.18 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.3288, 0.2691, -0.0327, ..., -0.1765, 0.0125, -0.0013], + [ 0.0065, -0.1772, -0.0301, ..., 0.0535, -0.0796, -0.1010], + [-0.2720, -0.3267, 0.0585, ..., -0.1592, -0.0845, -0.2601], + ..., + [ 0.0916, 0.2075, -0.1079, ..., -0.1894, 0.0918, 0.1010], + [ 0.0291, -0.2237, -0.0709, ..., -0.1703, -0.1356, -0.1307], + [-0.3205, -0.1026, -0.0737, ..., 0.1117, -0.3527, 0.1231]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 0.0000e+00], + [-1.9791e-09, 1.1642e-10, 0.0000e+00, ..., -3.6089e-09, + 0.0000e+00, 2.3283e-10], + [ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 3.4925e-10], + ..., + [ 1.2806e-09, 0.0000e+00, 0.0000e+00, ..., 2.2119e-09, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., -1.1642e-09, + 0.0000e+00, -1.2806e-09]], device='cuda:0') +Epoch 486, bias, value: tensor([ 0.0202, -0.0327, 0.0208, -0.0289, 0.0412, 0.0170, -0.0010, -0.0071, + -0.0240, -0.0161], device='cuda:0'), grad: tensor([ 4.6566e-10, -1.7462e-08, 2.9104e-09, 3.6089e-09, 5.1223e-09, + -6.0536e-09, 4.6566e-09, 1.1642e-08, 7.7998e-09, -1.6298e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 220.75, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4086 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.15 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.3288, 0.2692, -0.0327, ..., -0.1765, 0.0125, -0.0014], + [ 0.0065, -0.1773, -0.0300, ..., 0.0535, -0.0797, -0.1011], + [-0.2721, -0.3268, 0.0585, ..., -0.1593, -0.0846, -0.2602], + ..., + [ 0.0917, 0.2076, -0.1079, ..., -0.1895, 0.0918, 0.1011], + [ 0.0290, -0.2237, -0.0709, ..., -0.1704, -0.1356, -0.1308], + [-0.3207, -0.1027, -0.0737, ..., 0.1117, -0.3528, 0.1232]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2806e-09, + 1.1642e-10, 2.3283e-10], + [ 4.6566e-10, 8.1491e-10, 0.0000e+00, ..., 4.5402e-09, + 1.6298e-09, 4.4238e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 5.0059e-09, 3.6089e-09], + ..., + [-4.6566e-10, -9.3132e-10, 0.0000e+00, ..., 9.3132e-10, + 1.2806e-09, 5.8208e-10], + [ 0.0000e+00, 1.1642e-10, 0.0000e+00, ..., 6.9849e-10, + 1.1642e-09, 9.3132e-10], + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., -8.4983e-09, + 4.6566e-10, -4.1910e-09]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0202, -0.0326, 0.0207, -0.0290, 0.0413, 0.0171, -0.0010, -0.0071, + -0.0241, -0.0162], device='cuda:0'), grad: tensor([ 1.0012e-08, 2.3283e-08, 2.6776e-08, -5.9372e-08, 1.4203e-08, + 1.2410e-07, -1.3353e-07, 5.3551e-09, 1.2806e-08, -1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 220.70, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4121 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.3288, 0.2693, -0.0327, ..., -0.1767, 0.0125, -0.0014], + [ 0.0065, -0.1774, -0.0300, ..., 0.0536, -0.0797, -0.1011], + [-0.2722, -0.3269, 0.0585, ..., -0.1593, -0.0846, -0.2603], + ..., + [ 0.0918, 0.2077, -0.1079, ..., -0.1896, 0.0919, 0.1011], + [ 0.0290, -0.2238, -0.0709, ..., -0.1705, -0.1356, -0.1309], + [-0.3208, -0.1028, -0.0737, ..., 0.1117, -0.3528, 0.1232]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, -1.2689e-08, 0.0000e+00, ..., 6.9849e-10, + 0.0000e+00, 1.1642e-10], + [ 2.3283e-09, 4.0745e-09, 0.0000e+00, ..., 1.5134e-09, + 1.0477e-09, 3.1432e-09], + [ 1.1642e-09, 2.3283e-09, 0.0000e+00, ..., 1.0477e-09, + 6.9849e-10, 1.9791e-09], + ..., + [-5.8208e-10, -8.6147e-09, 0.0000e+00, ..., 4.8894e-09, + -2.7940e-09, -8.0327e-09], + [ 7.3342e-09, 1.1642e-09, 0.0000e+00, ..., 9.6625e-09, + 1.1642e-10, 5.8208e-10], + [ 1.5320e-07, 3.0268e-09, 0.0000e+00, ..., 1.9209e-07, + 5.8208e-10, 1.7462e-09]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0201, -0.0326, 0.0207, -0.0291, 0.0413, 0.0172, -0.0010, -0.0072, + -0.0242, -0.0162], device='cuda:0'), grad: tensor([-3.4459e-08, 1.8626e-08, 1.1758e-08, 2.3283e-09, -6.3423e-07, + 6.7521e-09, 2.1770e-08, -1.3155e-08, 3.4110e-08, 6.0257e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 220.22, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4037 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.3288, 0.2695, -0.0327, ..., -0.1767, 0.0125, -0.0014], + [ 0.0065, -0.1774, -0.0300, ..., 0.0535, -0.0797, -0.1012], + [-0.2723, -0.3269, 0.0585, ..., -0.1594, -0.0847, -0.2603], + ..., + [ 0.0918, 0.2078, -0.1079, ..., -0.1898, 0.0919, 0.1011], + [ 0.0289, -0.2238, -0.0709, ..., -0.1707, -0.1357, -0.1310], + [-0.3211, -0.1029, -0.0737, ..., 0.1117, -0.3529, 0.1234]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.2806e-09, 0.0000e+00, ..., -2.3283e-10, + 9.3132e-10, 2.2119e-09], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + -1.3970e-09, 0.0000e+00], + ..., + [-9.3132e-10, -2.2119e-09, 0.0000e+00, ..., 3.4925e-10, + -1.1642e-10, -3.7253e-09], + [ 5.8208e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 6.9849e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 1.2806e-09]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0202, -0.0326, 0.0207, -0.0292, 0.0414, 0.0172, -0.0009, -0.0072, + -0.0243, -0.0162], device='cuda:0'), grad: tensor([ 1.0477e-09, 6.7521e-09, -3.8650e-08, 7.3342e-09, 2.2119e-09, + -2.0606e-08, 6.1700e-09, 2.9686e-08, 1.2573e-08, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 220.71, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4232 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.3288, 0.2697, -0.0327, ..., -0.1769, 0.0125, -0.0014], + [ 0.0066, -0.1775, -0.0300, ..., 0.0537, -0.0798, -0.1013], + [-0.2724, -0.3270, 0.0585, ..., -0.1594, -0.0847, -0.2604], + ..., + [ 0.0918, 0.2079, -0.1080, ..., -0.1899, 0.0919, 0.1011], + [ 0.0289, -0.2239, -0.0709, ..., -0.1709, -0.1357, -0.1311], + [-0.3214, -0.1030, -0.0737, ..., 0.1118, -0.3529, 0.1235]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 0.0000e+00, 1.1642e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-7.0315e-08, 2.3283e-10, 1.1642e-10, ..., -1.2701e-07, + 0.0000e+00, 2.3283e-10], + [ 5.3551e-09, 0.0000e+00, 1.1642e-10, ..., 9.4296e-09, + 2.3283e-10, 1.1642e-10], + ..., + [ 2.1886e-08, 0.0000e+00, 1.1642e-10, ..., 3.9698e-08, + 1.1642e-10, 1.1642e-10], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 1.1642e-10, 0.0000e+00], + [ 3.1549e-08, 4.6566e-10, 2.3283e-10, ..., 5.5297e-08, + 1.1642e-10, 1.1642e-10]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0201, -0.0326, 0.0207, -0.0292, 0.0415, 0.0171, -0.0008, -0.0073, + -0.0244, -0.0162], device='cuda:0'), grad: tensor([ 3.6089e-09, -4.6287e-07, 3.6787e-08, -6.2631e-08, 6.9384e-08, + 6.4843e-08, 3.2596e-09, 1.4808e-07, 3.1432e-09, 2.0384e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 220.37, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4043 re_mapping 0.0018 re_causal 0.0068 /// teacc 99.17 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.3288, 0.2698, -0.0327, ..., -0.1769, 0.0125, -0.0015], + [ 0.0067, -0.1776, -0.0299, ..., 0.0538, -0.0798, -0.1014], + [-0.2725, -0.3271, 0.0585, ..., -0.1595, -0.0847, -0.2606], + ..., + [ 0.0919, 0.2081, -0.1080, ..., -0.1900, 0.0920, 0.1012], + [ 0.0289, -0.2239, -0.0709, ..., -0.1710, -0.1357, -0.1313], + [-0.3216, -0.1032, -0.0738, ..., 0.1118, -0.3530, 0.1236]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 1.1642e-10], + [ 1.2806e-09, 2.0955e-09, 0.0000e+00, ..., 3.4925e-10, + 5.8208e-10, 1.8626e-09], + [ 1.5134e-09, 2.5611e-09, 0.0000e+00, ..., 1.1642e-10, + 6.9849e-10, 1.7462e-09], + ..., + [-4.1910e-09, -7.1013e-09, 0.0000e+00, ..., 1.8626e-09, + -1.8626e-09, -3.4925e-09], + [ 4.6566e-10, 8.1491e-10, 0.0000e+00, ..., 3.4925e-10, + 2.3283e-10, 6.9849e-10], + [ 3.4925e-10, 8.1491e-10, 0.0000e+00, ..., -2.7707e-08, + 1.1642e-10, -2.7125e-08]], device='cuda:0') +Epoch 491, bias, value: tensor([ 0.0202, -0.0325, 0.0206, -0.0293, 0.0415, 0.0171, -0.0008, -0.0073, + -0.0245, -0.0162], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.4086e-08, -3.3993e-08, 8.1491e-09, 8.1491e-08, + 3.9581e-09, 2.6776e-09, -1.2922e-08, 8.8476e-09, -6.4843e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 220.72, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4290 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.3288, 0.2701, -0.0327, ..., -0.1769, 0.0125, -0.0015], + [ 0.0067, -0.1777, -0.0299, ..., 0.0539, -0.0799, -0.1016], + [-0.2726, -0.3272, 0.0585, ..., -0.1597, -0.0848, -0.2607], + ..., + [ 0.0920, 0.2083, -0.1081, ..., -0.1902, 0.0921, 0.1013], + [ 0.0289, -0.2240, -0.0709, ..., -0.1712, -0.1358, -0.1314], + [-0.3218, -0.1033, -0.0738, ..., 0.1119, -0.3530, 0.1237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, + 0.0000e+00, 1.0477e-09], + [-3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 4.1910e-09, + 1.1642e-10, 3.0268e-09], + [ 4.6566e-10, 1.1642e-10, 0.0000e+00, ..., 1.8626e-09, + 4.6566e-10, 3.4925e-10], + ..., + [ 3.4925e-10, 0.0000e+00, 0.0000e+00, ..., 4.4005e-08, + 3.4925e-10, 5.7742e-08], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 1.5134e-09, + 0.0000e+00, 5.8208e-10], + [ 2.3283e-10, 3.4925e-10, 0.0000e+00, ..., -7.7300e-08, + 0.0000e+00, -8.4634e-08]], device='cuda:0') +Epoch 492, bias, value: tensor([ 0.0204, -0.0325, 0.0205, -0.0293, 0.0416, 0.0171, -0.0008, -0.0073, + -0.0245, -0.0163], device='cuda:0'), grad: tensor([ 4.6566e-09, 2.4331e-08, 1.3504e-08, 1.7462e-09, 8.2422e-08, + 1.5018e-08, -4.0396e-08, 1.6205e-07, 8.0327e-09, -2.5751e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 220.40, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4189 re_mapping 0.0019 re_causal 0.0068 /// teacc 99.17 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.3289, 0.2702, -0.0327, ..., -0.1770, 0.0125, -0.0015], + [ 0.0068, -0.1778, -0.0299, ..., 0.0540, -0.0800, -0.1019], + [-0.2727, -0.3273, 0.0585, ..., -0.1597, -0.0849, -0.2608], + ..., + [ 0.0921, 0.2084, -0.1081, ..., -0.1903, 0.0921, 0.1014], + [ 0.0289, -0.2240, -0.0709, ..., -0.1712, -0.1358, -0.1314], + [-0.3219, -0.1034, -0.0738, ..., 0.1120, -0.3531, 0.1238]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 7.5670e-10, + 0.0000e+00, 1.7462e-10], + [ 4.0745e-10, 1.8626e-09, 0.0000e+00, ..., -6.9849e-10, + 1.8044e-09, 5.1223e-09], + [ 1.4552e-09, 2.8522e-09, 0.0000e+00, ..., 6.4028e-10, + 2.9104e-09, 5.9954e-09], + ..., + [-2.7940e-09, -5.5879e-09, 0.0000e+00, ..., 2.2701e-09, + -5.7626e-09, -1.3155e-08], + [ 6.9849e-10, 5.8208e-10, 0.0000e+00, ..., 8.7311e-10, + 4.6566e-10, 1.3970e-09], + [ 8.1491e-10, 1.2224e-09, 0.0000e+00, ..., -3.2596e-09, + 1.1059e-09, -9.8953e-10]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0204, -0.0326, 0.0206, -0.0294, 0.0415, 0.0171, -0.0008, -0.0073, + -0.0245, -0.0163], device='cuda:0'), grad: tensor([ 5.2387e-09, 3.7893e-08, -4.7905e-08, 2.9104e-09, 1.8626e-08, + 1.9791e-09, 9.0222e-09, -4.1269e-08, 2.6484e-08, 2.6776e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 220.44, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3947 re_mapping 0.0019 re_causal 0.0069 /// teacc 99.16 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.3289, 0.2703, -0.0327, ..., -0.1770, 0.0125, -0.0015], + [ 0.0069, -0.1778, -0.0298, ..., 0.0541, -0.0800, -0.1019], + [-0.2728, -0.3273, 0.0585, ..., -0.1598, -0.0849, -0.2609], + ..., + [ 0.0921, 0.2085, -0.1081, ..., -0.1903, 0.0922, 0.1015], + [ 0.0289, -0.2241, -0.0710, ..., -0.1713, -0.1358, -0.1315], + [-0.3220, -0.1036, -0.0738, ..., 0.1119, -0.3531, 0.1239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.8208e-11, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.7462e-10, 2.9104e-10], + [ 1.7462e-10, 0.0000e+00, 0.0000e+00, ..., 1.7462e-10, + 0.0000e+00, 6.9849e-10], + ..., + [-2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, -2.6193e-09], + [ 4.0745e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 2.3283e-10, 1.7462e-10], + [ 1.1642e-10, 5.8208e-11, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.9104e-10]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0205, -0.0326, 0.0205, -0.0294, 0.0416, 0.0171, -0.0008, -0.0073, + -0.0246, -0.0163], device='cuda:0'), grad: tensor([ 4.6566e-10, 2.6193e-09, 6.9849e-10, 3.4343e-09, 4.6566e-09, + -1.7462e-09, 1.3970e-09, -7.5670e-09, 4.1327e-09, 1.9209e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 220.89, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4345 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.17 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.3289, 0.2704, -0.0327, ..., -0.1771, 0.0125, -0.0016], + [ 0.0068, -0.1779, -0.0298, ..., 0.0541, -0.0800, -0.1021], + [-0.2729, -0.3274, 0.0585, ..., -0.1599, -0.0850, -0.2610], + ..., + [ 0.0923, 0.2086, -0.1081, ..., -0.1904, 0.0922, 0.1016], + [ 0.0288, -0.2241, -0.0710, ..., -0.1714, -0.1359, -0.1315], + [-0.3221, -0.1036, -0.0738, ..., 0.1121, -0.3531, 0.1240]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 5.8208e-11, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 2.3283e-10], + [ 2.9104e-09, 5.2387e-10, 0.0000e+00, ..., 2.7358e-09, + 2.9104e-10, 8.1491e-10], + [ 1.6880e-09, 1.2806e-09, 0.0000e+00, ..., 7.5670e-10, + 1.1642e-10, 1.8044e-09], + ..., + [-1.3970e-09, -1.9791e-09, 0.0000e+00, ..., 5.2387e-10, + 1.7462e-10, -2.5029e-09], + [ 6.9849e-10, 1.7462e-10, 0.0000e+00, ..., 5.8208e-11, + 1.1642e-10, 2.3283e-10], + [ 2.5611e-09, 1.3970e-09, 0.0000e+00, ..., 1.6880e-09, + 2.9104e-10, 1.4552e-09]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0205, -0.0327, 0.0206, -0.0295, 0.0415, 0.0171, -0.0008, -0.0072, + -0.0246, -0.0163], device='cuda:0'), grad: tensor([ 2.9104e-09, 1.4727e-08, 1.2689e-08, -7.0431e-09, -1.9267e-08, + -9.4005e-08, 8.6497e-08, -1.0536e-08, 3.6089e-09, 1.6589e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 220.84, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4296 re_mapping 0.0018 re_causal 0.0069 /// teacc 99.18 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.3289, 0.2705, -0.0327, ..., -0.1771, 0.0125, -0.0016], + [ 0.0068, -0.1779, -0.0298, ..., 0.0542, -0.0801, -0.1023], + [-0.2730, -0.3275, 0.0585, ..., -0.1599, -0.0851, -0.2611], + ..., + [ 0.0924, 0.2087, -0.1081, ..., -0.1907, 0.0922, 0.1016], + [ 0.0289, -0.2241, -0.0710, ..., -0.1715, -0.1359, -0.1316], + [-0.3222, -0.1037, -0.0738, ..., 0.1123, -0.3532, 0.1242]], + device='cuda:0'), grad: tensor([[ 5.8208e-10, -6.3446e-09, 5.8208e-11, ..., 5.8208e-11, + 0.0000e+00, 1.1642e-10], + [ 2.0955e-09, 2.6193e-09, 1.1642e-10, ..., 2.9104e-10, + 5.8208e-11, 1.2224e-09], + [ 5.8790e-09, -4.0745e-10, 2.9104e-10, ..., 1.7462e-10, + 5.8208e-11, 2.9104e-10], + ..., + [ 3.4925e-10, -2.4447e-09, 5.8208e-11, ..., 3.9581e-09, + 5.8208e-11, 3.2014e-09], + [-1.5018e-08, 4.0745e-10, -7.5670e-10, ..., 2.3283e-10, + 5.8208e-11, 2.9104e-10], + [ 1.6880e-09, 1.5716e-09, 5.8208e-11, ..., -4.7148e-09, + 0.0000e+00, -5.9954e-09]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0205, -0.0328, 0.0206, -0.0295, 0.0414, 0.0171, -0.0008, -0.0074, + -0.0245, -0.0161], device='cuda:0'), grad: tensor([-4.4820e-09, 2.3574e-08, 2.0664e-08, -5.8208e-11, 3.6671e-09, + 2.3108e-08, 2.1129e-08, 1.9500e-08, -8.6613e-08, -8.7311e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 220.72, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4355 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.19 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.3289, 0.2706, -0.0327, ..., -0.1772, 0.0125, -0.0016], + [ 0.0068, -0.1780, -0.0298, ..., 0.0542, -0.0801, -0.1024], + [-0.2731, -0.3275, 0.0584, ..., -0.1600, -0.0851, -0.2611], + ..., + [ 0.0924, 0.2088, -0.1081, ..., -0.1908, 0.0922, 0.1016], + [ 0.0289, -0.2242, -0.0710, ..., -0.1715, -0.1359, -0.1317], + [-0.3224, -0.1039, -0.0738, ..., 0.1123, -0.3532, 0.1244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.4028e-10, ..., 8.7311e-10, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 5.8208e-11, 5.8208e-11, ..., -5.8208e-11, + 0.0000e+00, 2.3283e-10], + [ 2.9104e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 5.8208e-11, 1.7462e-10], + ..., + [ 8.1491e-10, -2.3283e-10, 1.7462e-10, ..., 8.7311e-10, + 5.8208e-11, -1.0477e-09], + [ 3.4925e-10, 5.8208e-11, 1.7462e-10, ..., 3.4925e-10, + 1.7462e-10, 3.4925e-10], + [ 6.9849e-10, 5.8208e-11, 1.1642e-10, ..., 4.6566e-10, + 0.0000e+00, 4.0745e-10]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0205, -0.0328, 0.0207, -0.0296, 0.0414, 0.0171, -0.0008, -0.0074, + -0.0245, -0.0161], device='cuda:0'), grad: tensor([ 3.6671e-09, 2.3283e-10, 4.3656e-09, -2.8522e-09, -9.3714e-09, + 1.1118e-08, -1.0710e-08, 4.0163e-09, 4.2492e-09, 5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 220.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4392 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.22 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.3289, 0.2708, -0.0327, ..., -0.1772, 0.0125, -0.0016], + [ 0.0069, -0.1780, -0.0298, ..., 0.0542, -0.0802, -0.1024], + [-0.2732, -0.3275, 0.0584, ..., -0.1600, -0.0851, -0.2612], + ..., + [ 0.0925, 0.2089, -0.1081, ..., -0.1909, 0.0922, 0.1016], + [ 0.0289, -0.2242, -0.0710, ..., -0.1716, -0.1359, -0.1317], + [-0.3226, -0.1039, -0.0738, ..., 0.1124, -0.3532, 0.1245]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.7462e-10], + [ 1.6298e-09, 1.9034e-08, 0.0000e+00, ..., 0.0000e+00, + 4.2492e-09, 1.1642e-08], + [ 3.0268e-09, 5.6461e-09, 0.0000e+00, ..., 6.9849e-10, + 2.6193e-09, 5.1223e-09], + ..., + [-3.7253e-09, -4.1269e-08, 0.0000e+00, ..., 5.8208e-11, + -8.9640e-09, -2.7299e-08], + [-4.0745e-10, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 9.3132e-10], + [ 6.9849e-10, 1.5134e-08, 0.0000e+00, ..., 0.0000e+00, + 2.2119e-09, 1.0070e-08]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0207, -0.0328, 0.0207, -0.0296, 0.0415, 0.0170, -0.0008, -0.0075, + -0.0245, -0.0162], device='cuda:0'), grad: tensor([-2.9686e-09, 5.3027e-08, 2.9802e-08, 1.3970e-09, -1.3388e-09, + 3.8999e-09, 1.3970e-09, -1.1595e-07, -1.8626e-09, 4.1095e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 220.25, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4200 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.3289, 0.2710, -0.0327, ..., -0.1773, 0.0125, -0.0017], + [ 0.0068, -0.1781, -0.0298, ..., 0.0543, -0.0802, -0.1025], + [-0.2732, -0.3276, 0.0585, ..., -0.1601, -0.0852, -0.2613], + ..., + [ 0.0925, 0.2090, -0.1081, ..., -0.1910, 0.0923, 0.1015], + [ 0.0289, -0.2242, -0.0710, ..., -0.1717, -0.1360, -0.1318], + [-0.3227, -0.1040, -0.0738, ..., 0.1125, -0.3533, 0.1246]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.8208e-11, + 0.0000e+00, 0.0000e+00], + [-1.2224e-09, 1.7462e-10, -9.8953e-10, ..., -1.1642e-09, + 2.3283e-10, 1.3388e-09], + [ 2.3283e-10, 1.7462e-10, 3.4925e-10, ..., 4.0745e-10, + -3.4925e-10, 4.0745e-10], + ..., + [ 5.8208e-10, 0.0000e+00, -6.4028e-10, ..., 5.8208e-10, + 3.4925e-10, -3.4925e-09], + [ 9.8953e-10, 4.0745e-10, 8.1491e-10, ..., 1.1642e-10, + 5.8208e-11, 2.0955e-09], + [ 5.8208e-10, 7.5670e-10, 3.4925e-10, ..., 1.3388e-09, + 8.1491e-10, 1.3388e-09]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0208, -0.0328, 0.0208, -0.0295, 0.0414, 0.0170, -0.0008, -0.0076, + -0.0244, -0.0161], device='cuda:0'), grad: tensor([ 1.2806e-09, -3.0268e-08, 2.7940e-09, 1.4552e-09, -4.0745e-10, + -5.8208e-10, 5.4133e-09, 8.3819e-09, 1.0361e-08, 1.4843e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 219.91, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4384 re_mapping 0.0019 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.3289, 0.2712, -0.0327, ..., -0.1774, 0.0125, -0.0017], + [ 0.0068, -0.1782, -0.0298, ..., 0.0543, -0.0803, -0.1026], + [-0.2733, -0.3276, 0.0585, ..., -0.1601, -0.0852, -0.2614], + ..., + [ 0.0926, 0.2091, -0.1081, ..., -0.1912, 0.0922, 0.1015], + [ 0.0290, -0.2243, -0.0710, ..., -0.1718, -0.1360, -0.1319], + [-0.3228, -0.1041, -0.0738, ..., 0.1127, -0.3533, 0.1249]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 5.8208e-11, 0.0000e+00, ..., 4.0745e-10, + 0.0000e+00, 5.8208e-10], + [ 1.3388e-09, 1.1059e-09, 0.0000e+00, ..., 1.2224e-09, + 5.8208e-10, 2.5611e-09], + [ 1.1642e-10, 5.8208e-11, 0.0000e+00, ..., 8.1491e-10, + 5.8208e-11, 1.1642e-09], + ..., + [-2.8871e-08, -2.8638e-08, 0.0000e+00, ..., -1.4785e-08, + -1.6822e-08, -6.6822e-08], + [ 2.9104e-10, 2.3283e-10, 0.0000e+00, ..., 6.2864e-09, + 1.1642e-10, 8.6147e-09], + [ 3.1840e-08, 2.7474e-08, 0.0000e+00, ..., 1.2165e-08, + 1.6065e-08, 4.9826e-08]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0208, -0.0330, 0.0209, -0.0295, 0.0413, 0.0169, -0.0007, -0.0077, + -0.0244, -0.0159], device='cuda:0'), grad: tensor([ 3.6671e-09, 3.1665e-08, -3.6089e-08, 8.8476e-09, -3.8999e-09, + -6.9849e-10, 3.6671e-09, -2.0768e-07, 3.5798e-08, 1.7590e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 220.18, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4016 re_mapping 0.0019 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps4', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps4/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.970001 99.040001 ... 86.945686 69.764622 +ShearY 98.790001 98.940002 ... 86.945686 67.276324 +AutoContrast 99.089996 99.129997 ... 86.945686 61.274684 +Invert 98.720001 93.839996 ... 86.945686 50.101730 +Equalize 98.220001 98.409996 ... 86.945686 73.226051 +Solarize 98.059998 97.689995 ... 86.945686 63.925620 +SolarizeAdd 98.250000 97.829994 ... 86.945686 70.808240 +Posterize 99.010002 99.159996 ... 86.945686 73.015031 +Contrast 98.989998 99.260002 ... 86.945686 67.801760 +Color 99.049995 99.269997 ... 86.945686 59.026607 +Brightness 98.839996 99.279999 ... 86.945686 66.790533 +Sharpness 99.070000 99.209999 ... 86.945686 69.948109 +NoiseSalt 99.180000 99.220001 ... 86.945686 56.808372 +NoiseGaussian 99.070000 99.269997 ... 86.945686 58.316548 +w/o do (original x) 99.270000 0.000000 ... 0.000000 73.227632 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.2 68.411955 78.757916 76.478593 86.098655 77.43678 diff --git a/Meta-causal/code-withStyleAttack/66570.error b/Meta-causal/code-withStyleAttack/66570.error new file mode 100644 index 0000000000000000000000000000000000000000..73d7ba71d3cfe95f3c5799618de65f9325cb0186 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66570.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 38: clsadapt: command not found diff --git a/Meta-causal/code-withStyleAttack/66570.log b/Meta-causal/code-withStyleAttack/66570.log new file mode 100644 index 0000000000000000000000000000000000000000..22da83561d860d7326cc2e39aac95d998e2b4f70 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66570.log @@ -0,0 +1,14081 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-2.0338e-02, 2.8873e-02, -4.7888e-03, ..., -1.2311e-02, + 2.2054e-02, 1.7408e-02], + [-2.0231e-02, 6.1802e-03, 1.2685e-02, ..., -1.8912e-02, + 2.6775e-02, -1.1620e-02], + [-1.9306e-02, 1.4900e-02, -2.6086e-02, ..., -5.8901e-03, + 9.1994e-03, 2.9826e-02], + ..., + [ 8.8666e-03, 2.0376e-02, 5.6834e-03, ..., 6.7483e-03, + 1.1188e-02, 3.1058e-02], + [ 1.8620e-02, 2.2866e-02, -1.3200e-02, ..., -3.5290e-05, + -9.7229e-03, 5.5188e-03], + [ 8.4471e-04, -2.3735e-02, -4.7123e-03, ..., 1.0032e-02, + -2.1812e-02, 2.6384e-02]], device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0307, -0.0057, -0.0157, 0.0173, -0.0013, -0.0175, -0.0030, 0.0068, + -0.0251, 0.0207], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 221.11, cls_loss 1.2116 cls_loss_mapping 1.7897 cls_loss_causal 2.1944 re_mapping 0.1720 re_causal 0.1828 /// teacc 87.95 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0172, 0.0213, -0.0022, ..., -0.0190, 0.0143, 0.0114], + [-0.0274, 0.0060, 0.0047, ..., -0.0263, 0.0295, -0.0188], + [-0.0218, 0.0105, -0.0284, ..., -0.0093, 0.0098, 0.0346], + ..., + [ 0.0168, 0.0218, 0.0048, ..., 0.0135, 0.0113, 0.0245], + [ 0.0192, 0.0225, -0.0104, ..., -0.0043, -0.0142, 0.0052], + [ 0.0037, -0.0215, -0.0055, ..., 0.0140, -0.0198, 0.0262]], + device='cuda:0'), grad: tensor([[-0.0015, 0.0030, -0.0203, ..., 0.0029, 0.0006, 0.0042], + [ 0.0110, 0.0158, 0.0072, ..., 0.0115, 0.0029, 0.0109], + [ 0.0067, 0.0088, 0.0028, ..., -0.0056, -0.0185, -0.0535], + ..., + [ 0.0002, -0.0096, 0.0317, ..., 0.0119, 0.0051, 0.0142], + [-0.0444, -0.0373, -0.0276, ..., -0.0711, -0.0139, -0.0612], + [-0.0226, -0.0166, -0.0166, ..., -0.0589, -0.0157, -0.0084]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0287, -0.0047, -0.0155, 0.0172, -0.0021, -0.0166, -0.0034, 0.0083, + -0.0266, 0.0197], device='cuda:0'), grad: tensor([-0.0113, 0.0240, 0.0001, 0.0371, 0.0612, -0.0600, 0.0221, 0.0098, + -0.0606, -0.0224], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 220.87, cls_loss 0.3204 cls_loss_mapping 0.7082 cls_loss_causal 1.8723 re_mapping 0.2134 re_causal 0.2817 /// teacc 93.84 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0156, 0.0175, -0.0010, ..., -0.0212, 0.0130, 0.0094], + [-0.0310, 0.0075, 0.0021, ..., -0.0284, 0.0314, -0.0208], + [-0.0205, 0.0081, -0.0295, ..., -0.0110, 0.0084, 0.0366], + ..., + [ 0.0191, 0.0215, 0.0018, ..., 0.0152, 0.0103, 0.0226], + [ 0.0219, 0.0225, -0.0078, ..., -0.0078, -0.0176, 0.0068], + [ 0.0078, -0.0195, -0.0035, ..., 0.0163, -0.0210, 0.0271]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0025, -0.0090, ..., 0.0018, 0.0022, 0.0017], + [ 0.0027, 0.0062, 0.0061, ..., 0.0037, 0.0040, 0.0066], + [ 0.0005, -0.0014, -0.0147, ..., 0.0031, 0.0059, -0.0241], + ..., + [-0.0231, -0.0037, 0.0009, ..., -0.0158, 0.0015, -0.0042], + [-0.0075, -0.0100, -0.0054, ..., 0.0017, 0.0063, 0.0075], + [ 0.0070, -0.0002, -0.0046, ..., -0.0047, -0.0077, 0.0029]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0289, -0.0046, -0.0156, 0.0169, -0.0029, -0.0162, -0.0039, 0.0080, + -0.0260, 0.0204], device='cuda:0'), grad: tensor([-0.0036, 0.0114, -0.0141, -0.0050, -0.0393, 0.0229, 0.0439, -0.0081, + -0.0078, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 220.60, cls_loss 0.2093 cls_loss_mapping 0.4317 cls_loss_causal 1.6596 re_mapping 0.1496 re_causal 0.2387 /// teacc 95.57 lr 0.00010000 +Epoch 4, weight, value: tensor([[-1.3835e-02, 1.5192e-02, -2.6956e-04, ..., -2.3209e-02, + 1.1168e-02, 8.0131e-03], + [-3.2520e-02, 8.4930e-03, 3.2596e-04, ..., -2.9654e-02, + 3.3284e-02, -2.2537e-02], + [-2.1327e-02, 6.1256e-03, -3.0191e-02, ..., -1.2489e-02, + 6.8937e-03, 3.7925e-02], + ..., + [ 2.1169e-02, 2.1431e-02, -3.8800e-05, ..., 1.6649e-02, + 9.3628e-03, 2.1579e-02], + [ 2.3683e-02, 2.2640e-02, -6.6909e-03, ..., -9.4372e-03, + -1.9386e-02, 7.6527e-03], + [ 8.5494e-03, -1.8640e-02, -1.7218e-03, ..., 1.7103e-02, + -2.2801e-02, 2.7186e-02]], device='cuda:0'), grad: tensor([[-0.0004, 0.0011, -0.0002, ..., 0.0006, 0.0006, 0.0014], + [ 0.0039, 0.0107, 0.0012, ..., 0.0031, 0.0011, 0.0038], + [ 0.0051, 0.0072, -0.0010, ..., -0.0397, -0.0367, -0.0419], + ..., + [-0.0083, -0.0075, 0.0015, ..., -0.0127, -0.0001, -0.0126], + [-0.0126, -0.0221, 0.0021, ..., 0.0030, 0.0034, -0.0017], + [ 0.0057, 0.0154, 0.0035, ..., 0.0089, 0.0046, 0.0073]], + device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0293, -0.0049, -0.0155, 0.0169, -0.0033, -0.0162, -0.0040, 0.0080, + -0.0259, 0.0207], device='cuda:0'), grad: tensor([ 0.0002, 0.0117, -0.0146, 0.0112, 0.0144, 0.0131, -0.0163, -0.0142, + -0.0214, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 220.55, cls_loss 0.1576 cls_loss_mapping 0.3001 cls_loss_causal 1.4660 re_mapping 0.1147 re_causal 0.1962 /// teacc 96.10 lr 0.00010000 +Epoch 5, weight, value: tensor([[-1.2819e-02, 1.2800e-02, -5.7126e-06, ..., -2.4761e-02, + 9.9418e-03, 6.7867e-03], + [-3.4546e-02, 9.3752e-03, -1.0822e-03, ..., -3.0659e-02, + 3.4044e-02, -2.4124e-02], + [-2.1273e-02, 3.6643e-03, -3.0604e-02, ..., -1.3637e-02, + 6.5717e-03, 3.9395e-02], + ..., + [ 2.2334e-02, 2.0352e-02, -1.5579e-03, ..., 1.7287e-02, + 7.9728e-03, 2.0681e-02], + [ 2.5590e-02, 2.3290e-02, -5.9635e-03, ..., -1.0940e-02, + -2.1025e-02, 8.5825e-03], + [ 9.4028e-03, -1.7710e-02, -2.6248e-04, ..., 1.8092e-02, + -2.3775e-02, 2.7445e-02]], device='cuda:0'), grad: tensor([[-1.8511e-03, 6.1989e-04, -1.6422e-03, ..., 7.1001e-04, + 3.2187e-04, -4.7535e-05], + [ 2.5845e-03, -8.0795e-03, 7.4577e-04, ..., -3.3817e-03, + -3.7384e-03, 3.8834e-03], + [-5.7449e-03, -3.8414e-03, 4.2419e-03, ..., -3.0918e-03, + 3.5620e-04, -1.2184e-02], + ..., + [-7.0906e-04, 1.9684e-03, 2.1915e-03, ..., -3.9053e-04, + 8.5354e-04, 3.8071e-03], + [-3.1872e-03, 2.3270e-03, 5.9052e-03, ..., 2.2125e-03, + 9.6893e-04, 4.7340e-03], + [ 5.1956e-03, 6.8245e-03, 2.6875e-03, ..., 2.5349e-03, + 1.2388e-03, 2.7390e-03]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0292, -0.0051, -0.0150, 0.0170, -0.0035, -0.0162, -0.0042, 0.0078, + -0.0259, 0.0209], device='cuda:0'), grad: tensor([-0.0015, -0.0012, -0.0146, -0.0032, 0.0029, 0.0174, -0.0219, 0.0033, + 0.0100, 0.0088], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 220.68, cls_loss 0.1190 cls_loss_mapping 0.2190 cls_loss_causal 1.3125 re_mapping 0.0955 re_causal 0.1723 /// teacc 97.06 lr 0.00010000 +Epoch 6, weight, value: tensor([[-1.2686e-02, 1.0770e-02, 4.0780e-05, ..., -2.5924e-02, + 9.5085e-03, 6.6466e-03], + [-3.5054e-02, 1.0110e-02, -2.1269e-03, ..., -3.0457e-02, + 3.4784e-02, -2.5700e-02], + [-2.1650e-02, 1.7044e-03, -3.1771e-02, ..., -1.4879e-02, + 5.9343e-03, 4.0409e-02], + ..., + [ 2.3341e-02, 1.9823e-02, -2.6772e-03, ..., 1.8167e-02, + 6.6918e-03, 1.9521e-02], + [ 2.6894e-02, 2.3559e-02, -5.1741e-03, ..., -1.2062e-02, + -2.2126e-02, 1.0126e-02], + [ 9.9560e-03, -1.7401e-02, 6.5516e-04, ..., 1.8316e-02, + -2.5046e-02, 2.7003e-02]], device='cuda:0'), grad: tensor([[-1.2338e-04, 6.4707e-04, -3.5739e-04, ..., 5.6791e-04, + 4.9496e-04, 6.9666e-04], + [ 3.7537e-03, 7.9117e-03, 2.5845e-03, ..., 2.9640e-03, + 1.6117e-03, 4.8943e-03], + [ 7.0238e-04, 8.0526e-05, 2.3949e-04, ..., 5.0688e-04, + 8.5258e-04, -3.8624e-03], + ..., + [-2.1881e-02, -1.9775e-02, -1.9226e-03, ..., -2.9373e-02, + 3.0351e-04, 4.8685e-04], + [-4.5204e-03, -9.5444e-03, -4.0092e-03, ..., 1.3142e-03, + -3.1242e-03, -8.0643e-03], + [ 1.4099e-02, 2.0828e-02, 4.9171e-03, ..., 2.0355e-02, + -6.3419e-04, 2.7370e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0293, -0.0055, -0.0149, 0.0173, -0.0035, -0.0162, -0.0042, 0.0077, + -0.0256, 0.0205], device='cuda:0'), grad: tensor([-0.0004, 0.0097, -0.0009, 0.0059, 0.0047, -0.0128, 0.0037, -0.0201, + -0.0127, 0.0229], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 220.63, cls_loss 0.1108 cls_loss_mapping 0.1942 cls_loss_causal 1.2632 re_mapping 0.0785 re_causal 0.1482 /// teacc 97.20 lr 0.00010000 +Epoch 7, weight, value: tensor([[-1.2187e-02, 9.0787e-03, 1.4907e-04, ..., -2.7287e-02, + 8.5178e-03, 6.1161e-03], + [-3.5633e-02, 1.0341e-02, -2.8623e-03, ..., -3.1916e-02, + 3.4223e-02, -2.6920e-02], + [-2.2411e-02, 8.2805e-05, -3.2625e-02, ..., -1.5888e-02, + 5.5865e-03, 4.1615e-02], + ..., + [ 2.4301e-02, 1.9375e-02, -3.7758e-03, ..., 1.8854e-02, + 5.7177e-03, 1.8441e-02], + [ 2.7874e-02, 2.3951e-02, -4.6641e-03, ..., -1.2754e-02, + -2.2706e-02, 1.1112e-02], + [ 1.0750e-02, -1.6933e-02, 1.9538e-03, ..., 1.8652e-02, + -2.6157e-02, 2.7072e-02]], device='cuda:0'), grad: tensor([[-1.2660e-04, 7.4148e-04, 4.1962e-04, ..., 6.2704e-05, + 3.9840e-04, 6.2752e-04], + [ 4.2987e-04, -1.4214e-02, -1.3900e-04, ..., -5.6535e-05, + -7.0000e-04, 2.3770e-04], + [-1.4439e-03, 3.6755e-03, 8.3065e-04, ..., 2.7156e-04, + 3.7384e-04, -3.0308e-03], + ..., + [ 4.1628e-04, 3.5439e-03, 3.5667e-04, ..., 3.4690e-04, + 9.7215e-05, 1.0881e-03], + [-3.7766e-04, 2.5101e-03, 1.1343e-04, ..., 2.1625e-04, + 4.1342e-04, 1.2350e-04], + [ 3.9792e-04, 1.9054e-03, 1.0176e-03, ..., 6.3276e-04, + 1.7607e-04, 5.1785e-04]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0295, -0.0059, -0.0148, 0.0179, -0.0034, -0.0163, -0.0046, 0.0074, + -0.0254, 0.0204], device='cuda:0'), grad: tensor([ 0.0016, -0.0124, 0.0002, 0.0030, 0.0023, -0.0065, 0.0003, 0.0044, + 0.0035, 0.0036], device='cuda:0') +100 +0.0001 +changing lr +epoch 6, time 220.44, cls_loss 0.1000 cls_loss_mapping 0.1771 cls_loss_causal 1.1792 re_mapping 0.0676 re_causal 0.1338 /// teacc 96.81 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0113, 0.0074, 0.0003, ..., -0.0289, 0.0066, 0.0053], + [-0.0360, 0.0113, -0.0034, ..., -0.0315, 0.0342, -0.0288], + [-0.0226, -0.0018, -0.0335, ..., -0.0170, 0.0050, 0.0428], + ..., + [ 0.0251, 0.0190, -0.0046, ..., 0.0196, 0.0051, 0.0175], + [ 0.0290, 0.0244, -0.0041, ..., -0.0129, -0.0232, 0.0123], + [ 0.0113, -0.0168, 0.0030, ..., 0.0189, -0.0270, 0.0271]], + device='cuda:0'), grad: tensor([[ 2.3991e-05, 3.6120e-04, 7.6199e-04, ..., 7.2145e-04, + 5.5981e-04, 6.3801e-04], + [ 2.5201e-04, -4.5586e-04, 6.8140e-04, ..., 3.7694e-04, + -2.1601e-04, 4.4370e-04], + [ 7.6342e-04, 7.3481e-04, 1.1253e-03, ..., 9.2220e-04, + 5.5456e-04, 3.3426e-04], + ..., + [ 5.8889e-04, 8.7595e-04, 6.3896e-04, ..., 9.3937e-04, + 3.8052e-04, 8.6498e-04], + [ 8.7619e-05, 4.7150e-03, 1.6775e-03, ..., 5.3101e-03, + 3.4161e-03, 3.5439e-03], + [-4.7989e-03, -2.1439e-03, -3.5477e-03, ..., -7.0724e-03, + -1.6165e-03, -2.4433e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0294, -0.0059, -0.0144, 0.0178, -0.0036, -0.0163, -0.0047, 0.0074, + -0.0250, 0.0201], device='cuda:0'), grad: tensor([ 0.0007, -0.0002, 0.0008, 0.0017, -0.0015, -0.0055, 0.0018, 0.0015, + 0.0063, -0.0057], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 220.74, cls_loss 0.1022 cls_loss_mapping 0.1662 cls_loss_causal 1.1195 re_mapping 0.0585 re_causal 0.1186 /// teacc 97.69 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0097, 0.0065, 0.0007, ..., -0.0302, 0.0052, 0.0047], + [-0.0369, 0.0114, -0.0041, ..., -0.0323, 0.0331, -0.0308], + [-0.0239, -0.0028, -0.0348, ..., -0.0183, 0.0046, 0.0436], + ..., + [ 0.0256, 0.0189, -0.0056, ..., 0.0201, 0.0044, 0.0169], + [ 0.0308, 0.0245, -0.0034, ..., -0.0138, -0.0237, 0.0137], + [ 0.0120, -0.0164, 0.0037, ..., 0.0193, -0.0277, 0.0270]], + device='cuda:0'), grad: tensor([[-3.1948e-05, 4.7636e-04, 4.9889e-05, ..., 2.8205e-04, + 1.8549e-04, 2.9922e-04], + [-1.1384e-04, 4.6182e-04, 2.1243e-04, ..., 1.2553e-04, + 6.1035e-04, 2.0370e-03], + [ 5.0354e-04, 1.5488e-03, 8.1730e-04, ..., 8.7500e-04, + 1.0605e-03, -1.1301e-03], + ..., + [-1.5545e-03, 3.8266e-04, 7.3624e-04, ..., -1.3103e-03, + 3.2806e-04, 5.8746e-04], + [ 1.8635e-03, 3.0346e-03, 1.8282e-03, ..., 3.0422e-03, + 5.4789e-04, 1.0853e-03], + [-3.0346e-03, -3.1776e-03, -3.9101e-03, ..., -5.1079e-03, + -6.7520e-04, -2.0828e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0294, -0.0064, -0.0143, 0.0182, -0.0036, -0.0161, -0.0050, 0.0071, + -0.0248, 0.0201], device='cuda:0'), grad: tensor([ 0.0007, 0.0027, 0.0008, -0.0098, -0.0097, 0.0053, 0.0094, 0.0001, + 0.0040, -0.0036], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 220.71, cls_loss 0.0765 cls_loss_mapping 0.1339 cls_loss_causal 1.1052 re_mapping 0.0525 re_causal 0.1136 /// teacc 98.08 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0094, 0.0058, 0.0011, ..., -0.0315, 0.0041, 0.0041], + [-0.0369, 0.0120, -0.0050, ..., -0.0316, 0.0326, -0.0318], + [-0.0242, -0.0040, -0.0352, ..., -0.0194, 0.0038, 0.0443], + ..., + [ 0.0261, 0.0180, -0.0067, ..., 0.0205, 0.0039, 0.0160], + [ 0.0315, 0.0251, -0.0030, ..., -0.0147, -0.0244, 0.0142], + [ 0.0129, -0.0159, 0.0047, ..., 0.0199, -0.0281, 0.0270]], + device='cuda:0'), grad: tensor([[ 0.0063, 0.0098, 0.0076, ..., 0.0066, 0.0007, 0.0086], + [ 0.0011, 0.0006, 0.0006, ..., 0.0020, 0.0004, 0.0004], + [-0.0047, -0.0082, -0.0105, ..., 0.0007, 0.0002, -0.0109], + ..., + [-0.0030, -0.0003, 0.0026, ..., -0.0083, -0.0015, 0.0017], + [ 0.0003, 0.0031, 0.0028, ..., 0.0007, 0.0003, 0.0020], + [-0.0059, -0.0044, -0.0062, ..., -0.0094, -0.0020, -0.0034]], + device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0299, -0.0064, -0.0141, 0.0181, -0.0036, -0.0161, -0.0052, 0.0068, + -0.0249, 0.0203], device='cuda:0'), grad: tensor([ 0.0137, 0.0016, -0.0142, -0.0006, 0.0046, -0.0011, 0.0035, -0.0031, + 0.0042, -0.0088], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 220.88, cls_loss 0.0707 cls_loss_mapping 0.1242 cls_loss_causal 1.0703 re_mapping 0.0474 re_causal 0.1065 /// teacc 98.28 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0092, 0.0041, 0.0012, ..., -0.0323, 0.0029, 0.0035], + [-0.0377, 0.0126, -0.0054, ..., -0.0325, 0.0318, -0.0327], + [-0.0246, -0.0055, -0.0355, ..., -0.0207, 0.0031, 0.0449], + ..., + [ 0.0267, 0.0180, -0.0074, ..., 0.0213, 0.0039, 0.0156], + [ 0.0319, 0.0256, -0.0026, ..., -0.0154, -0.0245, 0.0149], + [ 0.0137, -0.0155, 0.0051, ..., 0.0204, -0.0290, 0.0267]], + device='cuda:0'), grad: tensor([[ 2.1243e-04, 4.7565e-04, 3.1090e-04, ..., 7.8201e-05, + 2.0361e-04, 9.8896e-04], + [ 3.1686e-04, 6.3324e-04, 2.8586e-04, ..., 1.6916e-04, + 3.9101e-05, 3.2115e-04], + [ 1.4296e-03, 4.8676e-03, 1.6441e-03, ..., 1.4937e-04, + 4.6492e-05, -3.8624e-04], + ..., + [-5.9080e-04, 6.4969e-05, 4.0293e-04, ..., -1.2207e-03, + 8.4639e-05, 4.5323e-04], + [-4.5776e-03, -1.5373e-02, -5.0278e-03, ..., 1.3292e-04, + -6.4671e-06, -4.0665e-03], + [ 2.4104e-04, -1.6522e-04, -5.2691e-05, ..., -1.0376e-03, + -2.8372e-04, -2.5344e-04]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0297, -0.0064, -0.0143, 0.0183, -0.0039, -0.0163, -0.0051, 0.0073, + -0.0248, 0.0202], device='cuda:0'), grad: tensor([ 0.0057, 0.0016, 0.0059, -0.0050, 0.0020, 0.0022, 0.0047, 0.0008, + -0.0203, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 220.19, cls_loss 0.0481 cls_loss_mapping 0.0926 cls_loss_causal 1.0223 re_mapping 0.0458 re_causal 0.1079 /// teacc 98.10 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0084, 0.0035, 0.0015, ..., -0.0333, 0.0018, 0.0032], + [-0.0384, 0.0131, -0.0057, ..., -0.0334, 0.0307, -0.0338], + [-0.0249, -0.0068, -0.0360, ..., -0.0216, 0.0028, 0.0459], + ..., + [ 0.0273, 0.0181, -0.0080, ..., 0.0220, 0.0037, 0.0150], + [ 0.0329, 0.0260, -0.0022, ..., -0.0158, -0.0247, 0.0157], + [ 0.0141, -0.0152, 0.0055, ..., 0.0205, -0.0297, 0.0264]], + device='cuda:0'), grad: tensor([[ 4.8981e-03, 1.7815e-03, 1.0330e-02, ..., 8.5533e-05, + 5.4264e-04, 6.7253e-03], + [ 1.1474e-04, -1.1003e-04, 2.3818e-04, ..., 1.4424e-04, + 2.1175e-05, 5.9223e-04], + [ 1.7614e-03, -1.0824e-03, -1.3447e-03, ..., 7.5245e-04, + 1.6415e-04, -4.5586e-03], + ..., + [-2.7294e-03, -1.3580e-03, 4.4346e-04, ..., -1.5545e-03, + 9.1612e-05, 2.7990e-04], + [ 2.7013e-04, 1.0222e-04, 4.0627e-04, ..., 2.4724e-04, + 5.6028e-05, 1.7381e-04], + [-7.9956e-03, -8.0299e-04, -1.3649e-02, ..., -1.2331e-03, + -8.5878e-04, -5.0850e-03]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0298, -0.0066, -0.0140, 0.0182, -0.0039, -0.0160, -0.0055, 0.0074, + -0.0247, 0.0199], device='cuda:0'), grad: tensor([ 0.0214, 0.0008, -0.0067, 0.0036, 0.0011, 0.0024, 0.0008, -0.0022, + 0.0013, -0.0226], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 221.00, cls_loss 0.0529 cls_loss_mapping 0.0968 cls_loss_causal 1.0256 re_mapping 0.0404 re_causal 0.0976 /// teacc 98.34 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0078, 0.0025, 0.0016, ..., -0.0339, 0.0014, 0.0027], + [-0.0389, 0.0136, -0.0061, ..., -0.0334, 0.0302, -0.0348], + [-0.0255, -0.0080, -0.0369, ..., -0.0227, 0.0022, 0.0464], + ..., + [ 0.0280, 0.0182, -0.0085, ..., 0.0227, 0.0034, 0.0146], + [ 0.0338, 0.0261, -0.0019, ..., -0.0166, -0.0253, 0.0162], + [ 0.0148, -0.0153, 0.0062, ..., 0.0206, -0.0305, 0.0264]], + device='cuda:0'), grad: tensor([[-3.6764e-04, 2.1267e-04, -1.1263e-03, ..., 2.3827e-05, + 5.4073e-04, 7.1049e-04], + [ 4.7803e-05, 6.7616e-04, 2.9278e-04, ..., -1.7926e-05, + 1.9288e-04, 4.3416e-04], + [-6.1810e-05, 1.9178e-03, 7.0906e-04, ..., 1.5640e-04, + 2.5773e-04, -1.1559e-03], + ..., + [ 1.9321e-03, 9.1400e-03, 4.4036e-04, ..., 1.2741e-03, + 9.7632e-05, 1.3971e-03], + [-3.9244e-04, -6.2981e-03, -1.6317e-03, ..., 1.4734e-04, + 2.5272e-04, -4.4174e-03], + [-5.2404e-04, 1.2231e-04, 1.1188e-04, ..., -1.0548e-03, + 5.3674e-05, 4.4435e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0299, -0.0069, -0.0139, 0.0181, -0.0041, -0.0160, -0.0054, 0.0078, + -0.0246, 0.0197], device='cuda:0'), grad: tensor([-0.0019, 0.0013, 0.0003, -0.0124, 0.0021, 0.0238, 0.0006, 0.0110, + -0.0254, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 220.71, cls_loss 0.0441 cls_loss_mapping 0.0870 cls_loss_causal 0.9800 re_mapping 0.0385 re_causal 0.0958 /// teacc 98.40 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0076, 0.0013, 0.0016, ..., -0.0350, 0.0004, 0.0023], + [-0.0392, 0.0141, -0.0065, ..., -0.0336, 0.0295, -0.0358], + [-0.0264, -0.0091, -0.0376, ..., -0.0237, 0.0017, 0.0467], + ..., + [ 0.0286, 0.0178, -0.0092, ..., 0.0233, 0.0030, 0.0143], + [ 0.0349, 0.0263, -0.0013, ..., -0.0167, -0.0259, 0.0170], + [ 0.0154, -0.0148, 0.0069, ..., 0.0208, -0.0310, 0.0261]], + device='cuda:0'), grad: tensor([[ 3.7402e-05, 2.3782e-04, -1.7893e-04, ..., 6.4015e-05, + 2.0817e-05, 7.0632e-05], + [-5.4312e-04, -1.9684e-03, -1.5700e-04, ..., 1.3947e-04, + 6.9618e-05, -1.8716e-05], + [ 1.9872e-04, 3.1257e-04, 1.9777e-04, ..., 1.1832e-04, + 1.4231e-05, -1.3030e-04], + ..., + [-1.2016e-04, 2.9016e-04, 1.0663e-04, ..., -3.1757e-04, + 2.0102e-05, 1.2374e-04], + [-5.7678e-03, -8.2932e-03, -5.4741e-03, ..., -3.6240e-03, + 8.0049e-05, -3.6583e-03], + [ 4.9706e-03, 8.1024e-03, 4.7455e-03, ..., 3.2749e-03, + 5.7459e-05, 3.1967e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0298, -0.0071, -0.0142, 0.0182, -0.0040, -0.0161, -0.0055, 0.0080, + -0.0242, 0.0197], device='cuda:0'), grad: tensor([-8.4415e-06, -1.7462e-03, 2.0683e-04, 9.8801e-04, 1.3018e-04, + 1.0157e-04, 2.1386e-04, 7.7009e-05, -8.3847e-03, 8.4152e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 220.14, cls_loss 0.0456 cls_loss_mapping 0.0848 cls_loss_causal 0.9933 re_mapping 0.0364 re_causal 0.0931 /// teacc 98.22 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0074, 0.0001, 0.0018, ..., -0.0351, 0.0005, 0.0027], + [-0.0399, 0.0144, -0.0068, ..., -0.0344, 0.0287, -0.0368], + [-0.0265, -0.0101, -0.0382, ..., -0.0247, 0.0009, 0.0474], + ..., + [ 0.0291, 0.0175, -0.0098, ..., 0.0237, 0.0027, 0.0134], + [ 0.0354, 0.0265, -0.0010, ..., -0.0174, -0.0266, 0.0173], + [ 0.0162, -0.0144, 0.0074, ..., 0.0214, -0.0311, 0.0260]], + device='cuda:0'), grad: tensor([[-6.8283e-04, 6.8545e-05, 1.0574e-04, ..., 2.1148e-04, + 1.8120e-04, -6.2108e-05], + [ 4.0245e-04, 6.4611e-04, 1.5855e-04, ..., 8.6689e-04, + 5.1689e-04, 1.5793e-03], + [ 6.5660e-04, 1.6010e-04, 4.0460e-04, ..., 2.5678e-04, + -3.5667e-04, -2.5291e-03], + ..., + [-1.5526e-03, -3.2020e-04, 2.0170e-04, ..., -1.7481e-03, + 7.9334e-05, 2.8706e-04], + [ 2.0921e-04, 7.6246e-04, 5.3930e-04, ..., 8.7690e-04, + 1.7715e-04, 2.3460e-04], + [ 2.4014e-03, 4.6539e-03, 5.3215e-04, ..., 6.1417e-03, + 2.3899e-03, 2.3975e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0300, -0.0073, -0.0139, 0.0181, -0.0040, -0.0162, -0.0057, 0.0078, + -0.0241, 0.0198], device='cuda:0'), grad: tensor([-0.0014, 0.0018, -0.0010, 0.0001, -0.0045, 0.0073, -0.0086, -0.0009, + 0.0013, 0.0058], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 220.92, cls_loss 0.0411 cls_loss_mapping 0.0762 cls_loss_causal 0.9661 re_mapping 0.0341 re_causal 0.0868 /// teacc 98.42 lr 0.00010000 +Epoch 16, weight, value: tensor([[-6.9643e-03, -7.7086e-07, 2.3059e-03, ..., -3.5503e-02, + 5.3579e-05, 2.8913e-03], + [-3.9664e-02, 1.4836e-02, -7.0415e-03, ..., -3.4469e-02, + 2.8013e-02, -3.7907e-02], + [-2.7280e-02, -1.1383e-02, -3.8997e-02, ..., -2.5754e-02, + 3.3229e-04, 4.7770e-02], + ..., + [ 2.9317e-02, 1.7459e-02, -1.0336e-02, ..., 2.4176e-02, + 2.3521e-03, 1.2958e-02], + [ 3.6289e-02, 2.6503e-02, -7.2695e-04, ..., -1.7311e-02, + -2.7153e-02, 1.7820e-02], + [ 1.6474e-02, -1.4518e-02, 7.6950e-03, ..., 2.1405e-02, + -3.1781e-02, 2.5660e-02]], device='cuda:0'), grad: tensor([[ 4.5389e-05, 3.2991e-05, 7.9215e-05, ..., 1.2207e-04, + 1.1593e-04, 1.4544e-04], + [ 1.1420e-04, 1.6880e-04, 2.3603e-05, ..., 5.9366e-04, + 2.3484e-04, 2.5272e-04], + [ 5.9366e-05, 3.6865e-05, 2.1830e-05, ..., 9.4414e-05, + 2.5272e-05, -2.1815e-04], + ..., + [-4.3631e-04, -3.2187e-05, 2.4343e-04, ..., -3.3092e-04, + 1.7452e-04, 2.5964e-04], + [ 2.0695e-04, 6.2525e-05, 1.4675e-04, ..., 4.3082e-04, + 9.5129e-05, 1.6046e-04], + [ 2.7132e-04, 5.2032e-03, -4.0197e-04, ..., 1.3824e-02, + 5.9242e-03, 4.4479e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0305, -0.0069, -0.0143, 0.0182, -0.0042, -0.0160, -0.0058, 0.0078, + -0.0242, 0.0195], device='cuda:0'), grad: tensor([ 2.2948e-04, 5.0783e-04, -2.5105e-04, 5.7411e-04, -9.0790e-03, + 1.9860e-04, -1.2100e-05, -1.4901e-04, 4.2486e-04, 7.5569e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 220.17, cls_loss 0.0305 cls_loss_mapping 0.0608 cls_loss_causal 0.9110 re_mapping 0.0319 re_causal 0.0833 /// teacc 98.27 lr 0.00010000 +Epoch 17, weight, value: tensor([[-6.4511e-03, -2.7394e-04, 2.3617e-03, ..., -3.6016e-02, + -8.8198e-04, 2.1699e-03], + [-4.0170e-02, 1.5129e-02, -7.5262e-03, ..., -3.4977e-02, + 2.7475e-02, -3.8656e-02], + [-2.7520e-02, -1.2377e-02, -3.9481e-02, ..., -2.6549e-02, + 8.1455e-05, 4.8429e-02], + ..., + [ 2.9889e-02, 1.7052e-02, -1.0838e-02, ..., 2.4603e-02, + 2.0595e-03, 1.2469e-02], + [ 3.6728e-02, 2.6888e-02, -3.5531e-04, ..., -1.7857e-02, + -2.7695e-02, 1.8166e-02], + [ 1.6710e-02, -1.4388e-02, 8.1810e-03, ..., 2.1483e-02, + -3.2257e-02, 2.5818e-02]], device='cuda:0'), grad: tensor([[ 7.2420e-05, 1.0949e-04, 4.1097e-05, ..., 1.8954e-04, + 1.9610e-04, 1.6642e-04], + [ 9.6989e-04, 8.4400e-04, 6.7663e-04, ..., 9.8801e-04, + 3.1066e-04, 5.4121e-04], + [ 6.2525e-05, 1.8513e-04, 1.1438e-04, ..., 9.1553e-05, + 9.5904e-05, -1.0890e-04], + ..., + [-8.3780e-04, -1.0955e-04, 2.5660e-05, ..., -1.2770e-03, + 6.2883e-05, 7.8201e-05], + [-1.5078e-03, -7.4816e-04, -1.4296e-03, ..., 2.6059e-04, + 1.5509e-04, -7.9918e-04], + [-6.3896e-03, -1.5808e-02, -4.8981e-03, ..., -1.8631e-02, + -8.3160e-03, -7.3853e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0302, -0.0069, -0.0143, 0.0183, -0.0044, -0.0157, -0.0058, 0.0076, + -0.0242, 0.0195], device='cuda:0'), grad: tensor([ 0.0003, 0.0019, 0.0002, 0.0004, 0.0173, 0.0006, 0.0003, -0.0012, + -0.0027, -0.0172], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 220.80, cls_loss 0.0347 cls_loss_mapping 0.0705 cls_loss_causal 0.9086 re_mapping 0.0307 re_causal 0.0815 /// teacc 98.47 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0061, -0.0006, 0.0024, ..., -0.0367, -0.0017, 0.0014], + [-0.0405, 0.0152, -0.0080, ..., -0.0353, 0.0267, -0.0399], + [-0.0278, -0.0129, -0.0398, ..., -0.0271, -0.0005, 0.0491], + ..., + [ 0.0301, 0.0166, -0.0114, ..., 0.0249, 0.0020, 0.0125], + [ 0.0377, 0.0271, 0.0004, ..., -0.0182, -0.0276, 0.0192], + [ 0.0168, -0.0144, 0.0086, ..., 0.0216, -0.0327, 0.0256]], + device='cuda:0'), grad: tensor([[-9.6130e-04, 1.3268e-04, -4.1342e-04, ..., 4.4942e-04, + 1.9283e-03, 6.6900e-04], + [ 4.5717e-05, -5.6171e-04, 2.5392e-04, ..., -8.6427e-05, + 5.9795e-04, 5.3978e-04], + [ 1.3485e-03, 2.8634e-04, 1.5011e-03, ..., 1.6379e-04, + 2.1493e-04, 1.4381e-03], + ..., + [ 6.7854e-04, 1.0195e-03, 4.1533e-04, ..., 1.2007e-03, + 2.8419e-04, 2.2292e-04], + [ 2.0721e-02, 4.5853e-03, 2.2705e-02, ..., 3.4618e-03, + 1.2674e-03, 2.0172e-02], + [-2.3819e-02, -8.6670e-03, -2.6398e-02, ..., -9.3842e-03, + -1.8082e-03, -2.3453e-02]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0300, -0.0069, -0.0139, 0.0186, -0.0044, -0.0157, -0.0059, 0.0073, + -0.0237, 0.0190], device='cuda:0'), grad: tensor([-5.1230e-05, 2.3496e-04, 2.7752e-03, 1.8625e-03, 8.4000e-03, + -1.1578e-03, -7.8125e-03, 1.5297e-03, 2.8259e-02, -3.4027e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 220.73, cls_loss 0.0317 cls_loss_mapping 0.0626 cls_loss_causal 0.8684 re_mapping 0.0276 re_causal 0.0756 /// teacc 98.61 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0058, -0.0018, 0.0024, ..., -0.0373, -0.0025, 0.0009], + [-0.0412, 0.0151, -0.0082, ..., -0.0358, 0.0261, -0.0408], + [-0.0283, -0.0135, -0.0406, ..., -0.0281, -0.0010, 0.0495], + ..., + [ 0.0306, 0.0167, -0.0120, ..., 0.0254, 0.0017, 0.0122], + [ 0.0379, 0.0272, 0.0006, ..., -0.0188, -0.0281, 0.0193], + [ 0.0175, -0.0143, 0.0092, ..., 0.0218, -0.0331, 0.0257]], + device='cuda:0'), grad: tensor([[-4.3106e-04, 1.2130e-04, -1.2960e-03, ..., 1.1367e-04, + -9.8586e-05, -1.0834e-03], + [ 3.2401e-04, 3.1519e-04, 2.5082e-04, ..., 3.9268e-04, + 8.7440e-05, 2.0885e-04], + [ 2.0111e-04, 1.2815e-04, 2.0385e-04, ..., 2.2876e-04, + 3.1799e-05, 1.5593e-04], + ..., + [-1.4000e-03, 8.8751e-05, 1.2165e-04, ..., -2.2907e-03, + 7.7665e-05, 8.6188e-05], + [-2.9445e-04, -2.2621e-03, -1.9474e-03, ..., 1.9264e-04, + 8.1956e-05, -2.0542e-03], + [ 4.5753e-04, -7.5758e-05, 1.7512e-04, ..., 3.5143e-04, + 5.5134e-06, 1.2779e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0300, -0.0071, -0.0139, 0.0189, -0.0042, -0.0159, -0.0057, 0.0073, + -0.0240, 0.0189], device='cuda:0'), grad: tensor([-0.0021, 0.0009, 0.0008, 0.0010, 0.0010, -0.0003, 0.0029, -0.0020, + -0.0037, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 220.42, cls_loss 0.0293 cls_loss_mapping 0.0613 cls_loss_causal 0.8654 re_mapping 0.0267 re_causal 0.0730 /// teacc 98.64 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0056, -0.0025, 0.0028, ..., -0.0379, -0.0033, 0.0007], + [-0.0413, 0.0154, -0.0085, ..., -0.0360, 0.0257, -0.0417], + [-0.0282, -0.0145, -0.0413, ..., -0.0283, -0.0014, 0.0504], + ..., + [ 0.0309, 0.0163, -0.0125, ..., 0.0258, 0.0013, 0.0114], + [ 0.0386, 0.0275, 0.0010, ..., -0.0191, -0.0287, 0.0196], + [ 0.0180, -0.0140, 0.0098, ..., 0.0221, -0.0333, 0.0259]], + device='cuda:0'), grad: tensor([[-9.8199e-06, 5.3972e-05, 2.4289e-05, ..., 8.5175e-05, + 3.9309e-05, 1.5771e-04], + [ 4.9740e-05, 1.7822e-04, 2.3276e-05, ..., 8.7678e-05, + 1.9953e-05, 2.0075e-04], + [-1.2279e-04, -2.3210e-04, 9.8825e-05, ..., -1.2493e-04, + 6.3777e-05, -8.6308e-04], + ..., + [ 4.9543e-04, 5.1451e-04, 3.7956e-04, ..., 8.8930e-04, + 2.4402e-04, 7.8869e-04], + [-3.3164e-04, -1.5950e-04, -4.1938e-04, ..., 2.6798e-04, + 8.4341e-05, -2.9588e-04], + [-6.3610e-04, -3.7456e-04, -4.1676e-04, ..., -1.5450e-03, + -4.0698e-04, -3.0327e-04]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0300, -0.0074, -0.0134, 0.0188, -0.0046, -0.0159, -0.0057, 0.0074, + -0.0239, 0.0190], device='cuda:0'), grad: tensor([ 0.0002, 0.0015, -0.0043, 0.0004, 0.0004, 0.0014, -0.0011, 0.0026, + 0.0001, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 219.62, cls_loss 0.0298 cls_loss_mapping 0.0588 cls_loss_causal 0.8830 re_mapping 0.0262 re_causal 0.0732 /// teacc 98.44 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0053, -0.0030, 0.0032, ..., -0.0384, -0.0037, 0.0003], + [-0.0414, 0.0156, -0.0089, ..., -0.0360, 0.0253, -0.0425], + [-0.0289, -0.0153, -0.0418, ..., -0.0294, -0.0018, 0.0510], + ..., + [ 0.0313, 0.0162, -0.0132, ..., 0.0262, 0.0008, 0.0111], + [ 0.0395, 0.0278, 0.0015, ..., -0.0195, -0.0290, 0.0197], + [ 0.0181, -0.0141, 0.0099, ..., 0.0219, -0.0339, 0.0255]], + device='cuda:0'), grad: tensor([[ 1.0529e-02, 6.3479e-05, 3.7823e-03, ..., 9.8705e-04, + 9.4548e-06, 3.1281e-03], + [ 6.2323e-04, 1.4675e-04, 1.0991e-04, ..., 6.1846e-04, + 9.6262e-05, 1.7703e-04], + [-1.2428e-02, 1.3781e-04, -4.4899e-03, ..., -9.8038e-04, + 8.4117e-06, -3.8776e-03], + ..., + [ 2.2292e-04, -1.4699e-04, 4.7731e-04, ..., -2.6774e-04, + 8.4043e-05, 4.5156e-04], + [ 4.8637e-04, 1.6558e-04, 8.0764e-05, ..., 3.2520e-04, + 7.8321e-05, 6.9499e-05], + [-4.9067e-04, -5.8126e-04, -6.5088e-04, ..., -1.8444e-03, + -1.3912e-04, -5.7316e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0300, -0.0074, -0.0133, 0.0186, -0.0041, -0.0159, -0.0058, 0.0075, + -0.0238, 0.0186], device='cuda:0'), grad: tensor([ 0.0130, 0.0008, -0.0156, -0.0002, 0.0009, -0.0002, 0.0011, 0.0004, + 0.0010, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 220.77, cls_loss 0.0314 cls_loss_mapping 0.0616 cls_loss_causal 0.8644 re_mapping 0.0251 re_causal 0.0681 /// teacc 98.76 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0057, -0.0040, 0.0032, ..., -0.0392, -0.0043, -0.0002], + [-0.0421, 0.0152, -0.0095, ..., -0.0366, 0.0244, -0.0431], + [-0.0293, -0.0161, -0.0422, ..., -0.0302, -0.0020, 0.0515], + ..., + [ 0.0319, 0.0165, -0.0135, ..., 0.0266, 0.0001, 0.0110], + [ 0.0401, 0.0279, 0.0019, ..., -0.0201, -0.0296, 0.0201], + [ 0.0190, -0.0138, 0.0105, ..., 0.0226, -0.0343, 0.0255]], + device='cuda:0'), grad: tensor([[ 1.7333e-04, 2.9683e-04, 2.6011e-04, ..., 1.4210e-04, + 1.0687e-04, 4.4227e-04], + [-8.3065e-04, -2.0657e-03, 1.5986e-04, ..., -1.7328e-03, + 4.5300e-05, 2.3341e-04], + [ 1.9884e-04, 3.5191e-04, 3.4261e-04, ..., 1.8108e-04, + 1.1212e-04, 3.2520e-04], + ..., + [ 2.0862e-05, 7.9727e-04, 8.2612e-05, ..., 4.6587e-04, + 2.1666e-05, 2.1791e-04], + [-1.0052e-03, -4.2987e-04, -2.0180e-03, ..., 1.0008e-04, + -5.1832e-04, -4.1351e-03], + [ 7.2777e-05, -5.5170e-04, -1.7595e-03, ..., -3.8815e-04, + 4.0680e-05, -1.4460e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0296, -0.0080, -0.0131, 0.0186, -0.0044, -0.0158, -0.0059, 0.0081, + -0.0241, 0.0190], device='cuda:0'), grad: tensor([ 1.2836e-03, -4.4708e-03, 1.7662e-03, 3.3264e-03, 8.6117e-04, + 2.0027e-03, 2.1572e-03, 1.8644e-03, -8.8577e-03, 6.9082e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 220.78, cls_loss 0.0238 cls_loss_mapping 0.0507 cls_loss_causal 0.8334 re_mapping 0.0237 re_causal 0.0674 /// teacc 98.87 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0057, -0.0042, 0.0033, ..., -0.0401, -0.0053, -0.0009], + [-0.0424, 0.0158, -0.0098, ..., -0.0369, 0.0235, -0.0439], + [-0.0295, -0.0174, -0.0428, ..., -0.0310, -0.0023, 0.0519], + ..., + [ 0.0322, 0.0161, -0.0138, ..., 0.0269, -0.0006, 0.0105], + [ 0.0407, 0.0281, 0.0022, ..., -0.0204, -0.0299, 0.0205], + [ 0.0195, -0.0135, 0.0109, ..., 0.0229, -0.0345, 0.0255]], + device='cuda:0'), grad: tensor([[ 5.6773e-05, 7.4267e-05, 3.9548e-05, ..., 1.2374e-04, + 8.8155e-05, 1.2255e-04], + [-3.8719e-03, -4.4861e-03, 4.5091e-05, ..., -7.3013e-03, + -6.0081e-04, -2.2912e-04], + [ 7.0870e-05, 9.3222e-05, 3.0816e-05, ..., 9.9659e-05, + 5.6177e-05, -1.6153e-04], + ..., + [ 9.4891e-04, 2.3632e-03, 5.3972e-05, ..., 2.8896e-03, + -2.4724e-04, 4.6158e-04], + [-8.2016e-05, -1.0672e-03, -5.1451e-04, ..., 5.0879e-04, + 8.5413e-05, -5.2547e-04], + [ 2.6379e-03, 2.6512e-03, 5.2750e-05, ..., 7.4654e-03, + 2.2869e-03, 1.1072e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0295, -0.0075, -0.0136, 0.0186, -0.0043, -0.0158, -0.0059, 0.0082, + -0.0242, 0.0191], device='cuda:0'), grad: tensor([ 2.7657e-04, -8.8272e-03, -4.3392e-05, 5.2452e-04, -1.4420e-03, + 9.0027e-04, -2.2995e-04, 4.7417e-03, -1.0929e-03, 5.1956e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 219.69, cls_loss 0.0215 cls_loss_mapping 0.0495 cls_loss_causal 0.8088 re_mapping 0.0242 re_causal 0.0680 /// teacc 98.74 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0045, -0.0046, 0.0037, ..., -0.0399, -0.0062, -0.0013], + [-0.0428, 0.0162, -0.0101, ..., -0.0373, 0.0229, -0.0450], + [-0.0296, -0.0177, -0.0435, ..., -0.0315, -0.0026, 0.0526], + ..., + [ 0.0326, 0.0157, -0.0144, ..., 0.0274, -0.0009, 0.0100], + [ 0.0412, 0.0284, 0.0024, ..., -0.0209, -0.0300, 0.0209], + [ 0.0195, -0.0132, 0.0112, ..., 0.0231, -0.0345, 0.0256]], + device='cuda:0'), grad: tensor([[-2.3818e-04, 1.8492e-05, -5.9605e-05, ..., 1.2070e-05, + 1.2887e-04, 1.0175e-04], + [ 3.5226e-05, -2.4259e-04, -7.6711e-05, ..., 3.6120e-05, + -1.2207e-04, -5.5408e-04], + [ 9.5963e-05, 2.5201e-04, 1.9205e-04, ..., 4.6790e-05, + 2.4939e-04, 5.9557e-04], + ..., + [ 3.2812e-05, 5.5760e-05, 2.8729e-05, ..., 1.5259e-04, + 1.0878e-04, 4.8578e-05], + [-1.5354e-04, -1.1462e-04, 2.2113e-04, ..., -2.2352e-05, + 6.8665e-04, 6.4278e-04], + [ 1.0192e-04, 2.0003e-04, 1.1021e-04, ..., 3.1233e-04, + 2.0933e-04, 1.3423e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0298, -0.0078, -0.0129, 0.0185, -0.0047, -0.0160, -0.0059, 0.0084, + -0.0242, 0.0189], device='cuda:0'), grad: tensor([-2.4045e-04, -9.8515e-04, 1.1511e-03, 7.1943e-05, -2.9206e-04, + 4.1275e-03, -5.3329e-03, 2.4164e-04, 8.2111e-04, 4.4084e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 219.46, cls_loss 0.0218 cls_loss_mapping 0.0521 cls_loss_causal 0.8210 re_mapping 0.0230 re_causal 0.0646 /// teacc 98.68 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0046, -0.0053, 0.0039, ..., -0.0408, -0.0069, -0.0016], + [-0.0430, 0.0165, -0.0105, ..., -0.0375, 0.0221, -0.0461], + [-0.0300, -0.0183, -0.0441, ..., -0.0324, -0.0035, 0.0529], + ..., + [ 0.0329, 0.0152, -0.0149, ..., 0.0275, -0.0015, 0.0095], + [ 0.0418, 0.0287, 0.0031, ..., -0.0211, -0.0308, 0.0214], + [ 0.0200, -0.0130, 0.0119, ..., 0.0237, -0.0342, 0.0261]], + device='cuda:0'), grad: tensor([[-4.4629e-06, 1.0151e-04, 5.6326e-06, ..., 4.2289e-05, + 5.5164e-05, 1.1420e-04], + [ 1.8907e-04, -2.0294e-03, 2.2519e-04, ..., 2.8044e-05, + 4.0293e-05, -9.2363e-04], + [ 4.9353e-04, 2.1267e-03, 4.4331e-06, ..., 3.6788e-04, + -1.3389e-05, -1.5297e-03], + ..., + [-1.4067e-03, -5.7650e-04, 4.2081e-05, ..., -8.9931e-04, + 7.4804e-06, 2.0397e-04], + [ 5.9605e-05, -8.3804e-05, -6.5327e-04, ..., 1.7369e-04, + -1.5152e-04, 1.2627e-03], + [-1.0446e-05, -1.3542e-03, -4.8804e-04, ..., -3.3212e-04, + 4.1813e-05, -1.0884e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0296, -0.0080, -0.0131, 0.0185, -0.0050, -0.0162, -0.0058, 0.0083, + -0.0238, 0.0195], device='cuda:0'), grad: tensor([ 0.0002, -0.0034, -0.0008, 0.0006, 0.0014, -0.0073, 0.0080, -0.0010, + 0.0038, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 219.76, cls_loss 0.0158 cls_loss_mapping 0.0397 cls_loss_causal 0.8049 re_mapping 0.0221 re_causal 0.0627 /// teacc 98.80 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0044, -0.0058, 0.0042, ..., -0.0411, -0.0077, -0.0019], + [-0.0433, 0.0168, -0.0108, ..., -0.0381, 0.0212, -0.0468], + [-0.0303, -0.0191, -0.0446, ..., -0.0330, -0.0038, 0.0534], + ..., + [ 0.0333, 0.0149, -0.0153, ..., 0.0277, -0.0022, 0.0091], + [ 0.0422, 0.0289, 0.0033, ..., -0.0215, -0.0312, 0.0217], + [ 0.0203, -0.0129, 0.0119, ..., 0.0238, -0.0346, 0.0260]], + device='cuda:0'), grad: tensor([[ 4.2844e-04, 4.6992e-04, 7.2241e-04, ..., 4.5242e-03, + 1.4429e-03, 3.8395e-03], + [-1.8663e-03, -6.1607e-03, -3.5429e-04, ..., 2.1052e-04, + 1.0586e-03, -9.1457e-04], + [ 6.6221e-05, 2.7180e-04, 2.5940e-04, ..., 1.1292e-03, + 3.5882e-04, 4.7874e-04], + ..., + [ 2.5725e-04, 4.7970e-04, 2.4486e-04, ..., 1.1969e-03, + 4.0698e-04, 9.3699e-04], + [ 5.4312e-04, 2.0561e-03, -2.3785e-03, ..., 3.1877e-04, + 2.6798e-04, -1.1625e-03], + [-1.0319e-03, 2.8491e-04, -2.1133e-03, ..., -1.6678e-02, + -4.9858e-03, -1.2413e-02]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0296, -0.0080, -0.0132, 0.0186, -0.0048, -0.0162, -0.0057, 0.0085, + -0.0240, 0.0192], device='cuda:0'), grad: tensor([ 7.4463e-03, -2.6913e-03, 1.0185e-03, -5.9986e-04, 1.3901e-02, + 6.7253e-03, -4.2534e-03, 2.0046e-03, -6.6221e-05, -2.3483e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 219.76, cls_loss 0.0180 cls_loss_mapping 0.0447 cls_loss_causal 0.8292 re_mapping 0.0210 re_causal 0.0615 /// teacc 98.80 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0039, -0.0063, 0.0043, ..., -0.0417, -0.0083, -0.0022], + [-0.0440, 0.0168, -0.0111, ..., -0.0385, 0.0201, -0.0478], + [-0.0305, -0.0194, -0.0451, ..., -0.0335, -0.0041, 0.0540], + ..., + [ 0.0336, 0.0148, -0.0158, ..., 0.0280, -0.0025, 0.0086], + [ 0.0427, 0.0293, 0.0037, ..., -0.0216, -0.0318, 0.0222], + [ 0.0206, -0.0129, 0.0123, ..., 0.0241, -0.0350, 0.0259]], + device='cuda:0'), grad: tensor([[ 5.5134e-05, 7.2360e-05, 1.6415e-04, ..., 9.6619e-05, + 3.5316e-05, 1.9276e-04], + [-7.1451e-06, -7.1168e-05, 2.4348e-05, ..., 3.5055e-06, + 1.2487e-05, 5.4240e-05], + [-3.5644e-05, -3.0136e-04, -1.2302e-03, ..., 1.0145e-04, + 3.5856e-06, -1.8854e-03], + ..., + [-2.6658e-05, 1.8537e-04, 1.6594e-04, ..., 1.1817e-05, + 6.3181e-06, 1.2052e-04], + [ 3.2812e-05, 5.2601e-05, 1.6463e-04, ..., 5.6237e-05, + 2.9430e-05, 1.1629e-04], + [-1.0639e-04, 4.1842e-05, 9.5892e-04, ..., -8.1360e-05, + 5.0873e-05, 1.0185e-03]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0295, -0.0087, -0.0128, 0.0184, -0.0047, -0.0159, -0.0056, 0.0086, + -0.0240, 0.0192], device='cuda:0'), grad: tensor([ 3.5524e-04, -5.1081e-05, -2.3899e-03, 3.0136e-04, 3.9244e-04, + -1.0614e-03, -6.9380e-05, 1.8752e-04, 3.6740e-04, 1.9684e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 219.78, cls_loss 0.0171 cls_loss_mapping 0.0406 cls_loss_causal 0.7746 re_mapping 0.0211 re_causal 0.0594 /// teacc 98.81 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0038, -0.0067, 0.0045, ..., -0.0422, -0.0088, -0.0026], + [-0.0448, 0.0168, -0.0115, ..., -0.0391, 0.0194, -0.0485], + [-0.0306, -0.0201, -0.0454, ..., -0.0339, -0.0042, 0.0546], + ..., + [ 0.0343, 0.0148, -0.0163, ..., 0.0284, -0.0030, 0.0082], + [ 0.0432, 0.0299, 0.0043, ..., -0.0220, -0.0320, 0.0226], + [ 0.0208, -0.0128, 0.0125, ..., 0.0243, -0.0352, 0.0258]], + device='cuda:0'), grad: tensor([[ 7.5579e-05, 2.6703e-05, 1.4126e-04, ..., 2.6321e-04, + 8.2552e-05, 2.1625e-04], + [ 1.1867e-04, 3.3617e-05, 8.3566e-05, ..., 1.1116e-04, + 1.0788e-05, 1.2314e-04], + [ 4.0269e-04, 3.6359e-04, 5.5695e-04, ..., 1.5497e-04, + 3.7163e-05, 8.1444e-04], + ..., + [-1.1586e-05, 8.1062e-05, 1.9538e-04, ..., 3.0756e-05, + 6.3300e-05, 2.2435e-04], + [-3.1400e-04, -5.1355e-04, -5.4693e-04, ..., 2.5797e-04, + 5.8204e-05, -9.7561e-04], + [-7.8440e-04, -1.7917e-04, -9.3412e-04, ..., -1.4362e-03, + -3.2806e-04, -1.0519e-03]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0296, -0.0090, -0.0128, 0.0186, -0.0048, -0.0161, -0.0054, 0.0089, + -0.0238, 0.0188], device='cuda:0'), grad: tensor([ 0.0003, 0.0003, 0.0018, 0.0002, 0.0006, 0.0001, 0.0003, 0.0002, + -0.0019, -0.0020], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 219.44, cls_loss 0.0135 cls_loss_mapping 0.0352 cls_loss_causal 0.8006 re_mapping 0.0197 re_causal 0.0575 /// teacc 98.84 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0035, -0.0071, 0.0050, ..., -0.0426, -0.0093, -0.0027], + [-0.0451, 0.0167, -0.0118, ..., -0.0393, 0.0187, -0.0490], + [-0.0309, -0.0210, -0.0460, ..., -0.0345, -0.0046, 0.0550], + ..., + [ 0.0346, 0.0150, -0.0168, ..., 0.0287, -0.0036, 0.0075], + [ 0.0436, 0.0301, 0.0045, ..., -0.0224, -0.0325, 0.0228], + [ 0.0212, -0.0128, 0.0130, ..., 0.0245, -0.0354, 0.0260]], + device='cuda:0'), grad: tensor([[-1.9565e-05, 5.7697e-05, -3.7456e-04, ..., 3.4511e-05, + -5.8591e-05, -1.0020e-04], + [-1.8883e-04, -3.9887e-04, 2.3708e-05, ..., -8.2922e-04, + -3.1650e-05, 4.0710e-05], + [ 2.5243e-05, 6.2168e-05, 4.9919e-05, ..., 6.0111e-05, + 2.0623e-05, -1.8394e-04], + ..., + [ 1.7393e-04, 3.7909e-04, 1.1390e-04, ..., 7.9107e-04, + 8.2076e-05, 1.2779e-04], + [ 2.2203e-05, 1.3947e-04, 6.7830e-05, ..., 1.9956e-04, + 4.3511e-05, 5.0962e-05], + [-1.8406e-04, 3.2020e-04, 4.2886e-05, ..., 1.2617e-03, + 5.7602e-04, 8.3876e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0299, -0.0091, -0.0129, 0.0186, -0.0048, -0.0159, -0.0058, 0.0091, + -0.0239, 0.0187], device='cuda:0'), grad: tensor([-0.0004, -0.0014, -0.0002, -0.0009, -0.0012, 0.0008, 0.0004, 0.0012, + 0.0004, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 219.78, cls_loss 0.0166 cls_loss_mapping 0.0378 cls_loss_causal 0.7609 re_mapping 0.0194 re_causal 0.0564 /// teacc 98.78 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0033, -0.0076, 0.0053, ..., -0.0432, -0.0100, -0.0030], + [-0.0452, 0.0172, -0.0123, ..., -0.0395, 0.0180, -0.0497], + [-0.0311, -0.0216, -0.0460, ..., -0.0352, -0.0050, 0.0556], + ..., + [ 0.0346, 0.0143, -0.0177, ..., 0.0290, -0.0038, 0.0069], + [ 0.0443, 0.0304, 0.0047, ..., -0.0228, -0.0329, 0.0230], + [ 0.0216, -0.0126, 0.0136, ..., 0.0246, -0.0359, 0.0260]], + device='cuda:0'), grad: tensor([[ 2.7455e-06, 3.7819e-05, -6.4433e-05, ..., 3.1799e-05, + 5.0105e-06, -6.9402e-06], + [ 4.2439e-05, 3.7372e-05, 3.1084e-05, ..., 7.0572e-05, + 2.3127e-05, 1.2136e-04], + [-3.9011e-05, -6.4313e-05, 5.2214e-05, ..., 3.1978e-05, + 8.5756e-06, -3.8791e-04], + ..., + [-3.3617e-05, -4.0680e-06, 4.9114e-05, ..., -9.1910e-05, + -5.4464e-06, 7.8321e-05], + [ 2.5129e-04, 4.1485e-04, 3.8958e-04, ..., 3.2258e-04, + 3.9786e-05, 1.7715e-04], + [-9.3842e-04, -1.2741e-03, -1.2732e-03, ..., -7.9393e-04, + -6.0529e-05, -5.4151e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0300, -0.0091, -0.0129, 0.0185, -0.0046, -0.0157, -0.0057, 0.0086, + -0.0237, 0.0187], device='cuda:0'), grad: tensor([-8.1480e-05, 2.6059e-04, -6.6710e-04, 1.0233e-03, 1.1653e-04, + 2.3389e-04, 1.0002e-04, 6.0946e-05, 9.4461e-04, -1.9932e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 220.44, cls_loss 0.0146 cls_loss_mapping 0.0387 cls_loss_causal 0.7654 re_mapping 0.0192 re_causal 0.0555 /// teacc 98.88 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0028, -0.0086, 0.0054, ..., -0.0434, -0.0105, -0.0034], + [-0.0454, 0.0175, -0.0126, ..., -0.0397, 0.0170, -0.0506], + [-0.0316, -0.0223, -0.0466, ..., -0.0359, -0.0053, 0.0561], + ..., + [ 0.0354, 0.0146, -0.0181, ..., 0.0295, -0.0040, 0.0066], + [ 0.0446, 0.0307, 0.0051, ..., -0.0233, -0.0337, 0.0235], + [ 0.0216, -0.0127, 0.0139, ..., 0.0247, -0.0362, 0.0261]], + device='cuda:0'), grad: tensor([[-5.3078e-05, 2.6241e-05, 3.0056e-05, ..., 1.2003e-05, + 2.1315e-04, 2.0075e-04], + [ 1.2323e-05, -2.7561e-04, 3.5018e-05, ..., -2.4363e-05, + 7.8082e-06, 3.4869e-05], + [ 3.9548e-05, 2.5892e-04, 8.7261e-05, ..., 2.5317e-05, + 2.8223e-05, 1.2696e-05], + ..., + [-6.1132e-06, 1.0186e-04, 3.4064e-05, ..., 4.2009e-04, + 9.6440e-05, 1.2815e-04], + [-1.2708e-04, -2.9236e-05, -1.1712e-04, ..., 3.3468e-05, + -3.3481e-07, -2.3282e-04], + [-3.0160e-05, 3.5405e-05, -3.1292e-06, ..., 2.6137e-05, + 2.1011e-05, 9.9093e-06]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0300, -0.0092, -0.0129, 0.0183, -0.0047, -0.0156, -0.0055, 0.0090, + -0.0238, 0.0185], device='cuda:0'), grad: tensor([ 1.2338e-04, -2.3899e-03, 1.8148e-03, -5.0068e-05, -1.2970e-04, + 1.5318e-04, -2.8896e-04, 7.5626e-04, -1.1635e-04, 1.2732e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 219.48, cls_loss 0.0152 cls_loss_mapping 0.0399 cls_loss_causal 0.7750 re_mapping 0.0181 re_causal 0.0532 /// teacc 98.82 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0027, -0.0090, 0.0057, ..., -0.0442, -0.0115, -0.0035], + [-0.0453, 0.0179, -0.0130, ..., -0.0396, 0.0162, -0.0514], + [-0.0318, -0.0228, -0.0470, ..., -0.0365, -0.0054, 0.0567], + ..., + [ 0.0356, 0.0140, -0.0186, ..., 0.0298, -0.0044, 0.0061], + [ 0.0452, 0.0310, 0.0052, ..., -0.0235, -0.0343, 0.0237], + [ 0.0217, -0.0126, 0.0141, ..., 0.0248, -0.0367, 0.0258]], + device='cuda:0'), grad: tensor([[ 4.6104e-05, 1.1820e-04, 3.6061e-05, ..., 1.2016e-04, + 7.9453e-05, 2.4915e-04], + [-5.5790e-04, -2.7084e-03, 2.0728e-05, ..., -1.2767e-04, + 2.0280e-05, 8.3387e-05], + [ 1.9085e-04, 9.9945e-04, -2.5153e-05, ..., 1.7393e-04, + 6.0916e-05, -2.4796e-04], + ..., + [-2.7132e-04, 8.2397e-04, 2.9638e-05, ..., -1.8330e-03, + 2.9564e-05, 9.6202e-05], + [ 2.0698e-05, 1.3137e-04, 5.0694e-05, ..., 5.3167e-05, + 5.5164e-05, 1.9610e-04], + [ 2.5058e-04, 2.0838e-04, -4.3422e-05, ..., 1.0881e-03, + -7.5847e-06, 2.3365e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0297, -0.0093, -0.0127, 0.0185, -0.0047, -0.0156, -0.0052, 0.0089, + -0.0237, 0.0183], device='cuda:0'), grad: tensor([ 7.6866e-04, -5.3749e-03, 1.6775e-03, 3.9972e-06, 6.4182e-04, + 1.5774e-03, -1.5984e-03, 2.8253e-04, 5.8222e-04, 1.4381e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 219.77, cls_loss 0.0126 cls_loss_mapping 0.0326 cls_loss_causal 0.7683 re_mapping 0.0177 re_causal 0.0524 /// teacc 98.86 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0027, -0.0093, 0.0061, ..., -0.0449, -0.0120, -0.0037], + [-0.0453, 0.0186, -0.0133, ..., -0.0397, 0.0156, -0.0522], + [-0.0318, -0.0240, -0.0476, ..., -0.0369, -0.0058, 0.0571], + ..., + [ 0.0356, 0.0138, -0.0190, ..., 0.0299, -0.0048, 0.0055], + [ 0.0457, 0.0313, 0.0055, ..., -0.0238, -0.0348, 0.0243], + [ 0.0224, -0.0124, 0.0144, ..., 0.0253, -0.0367, 0.0259]], + device='cuda:0'), grad: tensor([[ 8.0526e-05, 4.1515e-05, 2.2739e-05, ..., 2.5392e-05, + 1.8179e-05, 7.8857e-05], + [ 9.6679e-05, -1.1706e-04, 2.0280e-05, ..., 2.2396e-05, + 3.5632e-06, 7.1585e-05], + [-5.0926e-04, -1.1533e-04, 4.2021e-05, ..., 1.0908e-04, + 1.1519e-05, -4.8423e-04], + ..., + [-4.0308e-06, 1.0949e-04, 3.0190e-05, ..., -2.3305e-04, + 1.2010e-05, 1.6046e-04], + [-4.0960e-04, -3.7646e-04, -4.4155e-04, ..., -2.6003e-05, + -1.9252e-05, -2.9683e-04], + [ 3.2425e-05, -1.3471e-05, 6.8247e-06, ..., -8.4519e-05, + 1.7241e-05, 6.0856e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0298, -0.0089, -0.0129, 0.0184, -0.0050, -0.0158, -0.0056, 0.0088, + -0.0234, 0.0186], device='cuda:0'), grad: tensor([ 2.0480e-04, 9.1344e-06, -1.0777e-03, 8.4496e-04, 7.7963e-05, + 1.9073e-04, 5.2363e-05, 4.9561e-05, -4.7445e-04, 1.2082e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 220.60, cls_loss 0.0110 cls_loss_mapping 0.0334 cls_loss_causal 0.7547 re_mapping 0.0169 re_causal 0.0511 /// teacc 98.94 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0022, -0.0101, 0.0063, ..., -0.0452, -0.0129, -0.0041], + [-0.0460, 0.0192, -0.0138, ..., -0.0400, 0.0153, -0.0529], + [-0.0322, -0.0248, -0.0479, ..., -0.0375, -0.0061, 0.0576], + ..., + [ 0.0361, 0.0137, -0.0196, ..., 0.0304, -0.0049, 0.0052], + [ 0.0467, 0.0316, 0.0061, ..., -0.0241, -0.0353, 0.0248], + [ 0.0223, -0.0124, 0.0145, ..., 0.0252, -0.0374, 0.0258]], + device='cuda:0'), grad: tensor([[-2.1309e-05, 1.0081e-05, -1.4625e-05, ..., -9.4576e-07, + 2.2039e-05, 4.5449e-06], + [-8.0187e-07, -2.6894e-04, 8.9556e-06, ..., -4.3064e-05, + -1.4566e-06, -1.9833e-05], + [ 1.5765e-05, 7.8857e-05, 2.0951e-05, ..., 2.1398e-05, + 1.1891e-05, 1.0937e-05], + ..., + [-2.4319e-05, 4.4972e-05, 2.4483e-05, ..., 1.4521e-05, + 8.2925e-06, 1.9193e-05], + [ 9.6262e-06, 1.3316e-04, 2.4378e-05, ..., 3.5465e-05, + 2.0042e-05, 2.5779e-05], + [-4.5091e-05, -4.6343e-05, -3.6687e-05, ..., -3.3051e-05, + 4.3452e-05, 1.3895e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0299, -0.0089, -0.0130, 0.0186, -0.0048, -0.0160, -0.0056, 0.0091, + -0.0232, 0.0181], device='cuda:0'), grad: tensor([-8.1211e-06, -4.7326e-04, 1.5128e-04, -5.2154e-06, 1.4938e-05, + 6.1572e-05, -5.8711e-05, 4.8935e-05, 2.5702e-04, 1.1764e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 220.63, cls_loss 0.0093 cls_loss_mapping 0.0283 cls_loss_causal 0.7678 re_mapping 0.0166 re_causal 0.0507 /// teacc 98.96 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0022, -0.0105, 0.0064, ..., -0.0457, -0.0136, -0.0046], + [-0.0466, 0.0192, -0.0142, ..., -0.0405, 0.0146, -0.0536], + [-0.0321, -0.0254, -0.0482, ..., -0.0379, -0.0065, 0.0582], + ..., + [ 0.0364, 0.0136, -0.0201, ..., 0.0307, -0.0053, 0.0046], + [ 0.0472, 0.0317, 0.0063, ..., -0.0244, -0.0356, 0.0250], + [ 0.0226, -0.0121, 0.0149, ..., 0.0255, -0.0377, 0.0259]], + device='cuda:0'), grad: tensor([[ 3.8326e-05, 3.5614e-05, 4.1306e-05, ..., 3.7670e-05, + 3.8534e-05, 4.6104e-05], + [ 4.6283e-05, -1.8671e-05, 4.1604e-05, ..., 2.5868e-05, + 5.3078e-05, 8.1718e-05], + [ 1.3375e-04, 9.9778e-05, 8.1301e-05, ..., 1.4353e-04, + 5.6595e-05, -3.2485e-05], + ..., + [-6.0034e-04, -3.1948e-05, -5.0068e-05, ..., -5.8985e-04, + 3.4310e-06, 3.2723e-05], + [ 2.3454e-05, -9.9063e-05, -1.1635e-04, ..., 1.2946e-04, + 4.3899e-05, -1.4627e-04], + [ 6.8426e-05, 2.8819e-05, 2.2024e-05, ..., 4.3213e-07, + 4.8690e-06, -1.9774e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0297, -0.0091, -0.0128, 0.0186, -0.0049, -0.0160, -0.0054, 0.0091, + -0.0234, 0.0182], device='cuda:0'), grad: tensor([ 0.0002, 0.0002, 0.0003, 0.0001, 0.0003, -0.0002, -0.0005, -0.0009, + 0.0002, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 219.72, cls_loss 0.0112 cls_loss_mapping 0.0331 cls_loss_causal 0.7896 re_mapping 0.0162 re_causal 0.0502 /// teacc 98.81 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0021, -0.0112, 0.0065, ..., -0.0461, -0.0141, -0.0049], + [-0.0468, 0.0194, -0.0146, ..., -0.0410, 0.0137, -0.0544], + [-0.0321, -0.0261, -0.0484, ..., -0.0388, -0.0073, 0.0586], + ..., + [ 0.0366, 0.0133, -0.0208, ..., 0.0312, -0.0058, 0.0041], + [ 0.0475, 0.0319, 0.0067, ..., -0.0249, -0.0360, 0.0255], + [ 0.0231, -0.0119, 0.0152, ..., 0.0257, -0.0381, 0.0257]], + device='cuda:0'), grad: tensor([[-1.1104e-04, 3.2395e-05, -3.3155e-07, ..., 6.8545e-06, + 1.3538e-05, 4.2617e-05], + [ 6.9678e-05, 6.6936e-05, 9.5248e-05, ..., -9.2834e-06, + -1.6645e-05, 7.3671e-05], + [ 2.3589e-05, 1.4913e-04, 9.0599e-05, ..., -5.8830e-05, + 1.0945e-05, -1.0556e-04], + ..., + [ 4.8503e-06, 9.2864e-05, 5.3585e-05, ..., -2.9370e-05, + 5.7630e-06, 6.1870e-05], + [-2.3103e-04, -1.7154e-04, -1.7953e-04, ..., 2.1070e-05, + -2.9244e-06, -3.4642e-04], + [ 1.1325e-05, 9.2566e-05, 7.3671e-05, ..., -6.8545e-05, + -1.0328e-06, 1.6332e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0296, -0.0091, -0.0129, 0.0189, -0.0048, -0.0160, -0.0054, 0.0092, + -0.0235, 0.0180], device='cuda:0'), grad: tensor([-1.1861e-04, 9.3699e-05, -1.7512e-04, -1.2398e-03, 2.3079e-04, + 4.9496e-04, 5.2118e-04, 2.3210e-04, -2.9230e-04, 2.5392e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 220.20, cls_loss 0.0107 cls_loss_mapping 0.0293 cls_loss_causal 0.7512 re_mapping 0.0156 re_causal 0.0481 /// teacc 98.89 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0016, -0.0118, 0.0065, ..., -0.0463, -0.0151, -0.0054], + [-0.0473, 0.0194, -0.0149, ..., -0.0415, 0.0132, -0.0552], + [-0.0323, -0.0268, -0.0490, ..., -0.0393, -0.0076, 0.0590], + ..., + [ 0.0367, 0.0131, -0.0215, ..., 0.0315, -0.0062, 0.0035], + [ 0.0480, 0.0321, 0.0070, ..., -0.0252, -0.0366, 0.0259], + [ 0.0233, -0.0118, 0.0153, ..., 0.0258, -0.0385, 0.0256]], + device='cuda:0'), grad: tensor([[ 8.4102e-05, 3.8862e-05, 3.0264e-05, ..., 3.1531e-05, + 3.7640e-05, 7.6890e-05], + [ 1.2767e-04, -3.9637e-05, 2.0951e-05, ..., 5.8711e-05, + 1.3404e-05, 2.7657e-05], + [ 6.2180e-04, 3.5810e-04, 2.5272e-05, ..., 2.9945e-04, + 9.6112e-06, -8.1420e-05], + ..., + [-1.1683e-03, -5.4026e-04, 2.4393e-05, ..., -7.5006e-04, + 8.5086e-06, -1.1347e-05], + [-4.0579e-04, 2.1780e-04, 3.9697e-05, ..., 4.1634e-05, + 1.9264e-04, -1.3804e-04], + [ 2.1899e-04, 3.0696e-05, 5.2869e-05, ..., 6.3896e-05, + 1.7419e-05, 4.3422e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0296, -0.0093, -0.0129, 0.0192, -0.0045, -0.0162, -0.0055, 0.0091, + -0.0233, 0.0179], device='cuda:0'), grad: tensor([ 2.3746e-04, 5.4479e-05, 1.1311e-03, 2.2984e-03, 2.0754e-04, + -3.0231e-03, 6.1703e-04, -2.1935e-03, 2.9802e-04, 3.7193e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 219.82, cls_loss 0.0098 cls_loss_mapping 0.0266 cls_loss_causal 0.6884 re_mapping 0.0161 re_causal 0.0467 /// teacc 98.82 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0015, -0.0122, 0.0068, ..., -0.0467, -0.0156, -0.0054], + [-0.0470, 0.0196, -0.0153, ..., -0.0416, 0.0128, -0.0556], + [-0.0325, -0.0270, -0.0493, ..., -0.0400, -0.0078, 0.0593], + ..., + [ 0.0372, 0.0130, -0.0219, ..., 0.0319, -0.0064, 0.0032], + [ 0.0483, 0.0326, 0.0072, ..., -0.0254, -0.0369, 0.0262], + [ 0.0235, -0.0116, 0.0156, ..., 0.0263, -0.0384, 0.0256]], + device='cuda:0'), grad: tensor([[-7.1168e-05, 2.9698e-05, -5.7131e-05, ..., -4.5717e-05, + 8.4043e-05, 9.9182e-05], + [ 2.3067e-05, -5.0187e-05, 2.3246e-05, ..., 2.6658e-05, + 2.6196e-05, 4.7266e-05], + [ 3.9339e-05, 8.3745e-05, 4.3422e-05, ..., 3.4213e-05, + 3.2127e-05, -1.7852e-05], + ..., + [-3.6538e-05, 2.4557e-05, 3.2544e-05, ..., -2.4050e-05, + 1.0461e-05, 4.5717e-05], + [-2.1911e-04, -2.7084e-04, -1.2314e-04, ..., -1.1808e-04, + 7.2122e-05, -1.4794e-04], + [ 6.6102e-05, 6.6817e-05, -1.0662e-05, ..., 7.5579e-05, + 3.1561e-05, 4.7326e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0297, -0.0095, -0.0126, 0.0186, -0.0049, -0.0155, -0.0058, 0.0093, + -0.0234, 0.0181], device='cuda:0'), grad: tensor([-4.4793e-05, -2.6420e-05, 6.5565e-05, 1.7548e-04, 1.5974e-04, + 2.4214e-05, -4.5371e-04, 2.2501e-05, -1.3387e-04, 2.1124e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 220.62, cls_loss 0.0101 cls_loss_mapping 0.0285 cls_loss_causal 0.7411 re_mapping 0.0150 re_causal 0.0471 /// teacc 98.99 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0019, -0.0125, 0.0067, ..., -0.0477, -0.0163, -0.0064], + [-0.0477, 0.0196, -0.0157, ..., -0.0421, 0.0122, -0.0564], + [-0.0325, -0.0276, -0.0497, ..., -0.0407, -0.0082, 0.0597], + ..., + [ 0.0377, 0.0130, -0.0223, ..., 0.0328, -0.0064, 0.0026], + [ 0.0488, 0.0329, 0.0074, ..., -0.0258, -0.0373, 0.0268], + [ 0.0238, -0.0116, 0.0159, ..., 0.0263, -0.0388, 0.0256]], + device='cuda:0'), grad: tensor([[-4.7028e-05, 2.0802e-05, -9.9838e-05, ..., 9.3505e-06, + 3.2932e-06, -7.4923e-05], + [-1.0198e-06, -2.2221e-03, 1.4871e-05, ..., -7.3433e-05, + -6.1616e-06, 1.2539e-05], + [ 5.2243e-05, 1.5221e-03, 3.0637e-05, ..., 1.4298e-05, + 2.6394e-06, 1.8239e-05], + ..., + [-1.8656e-05, 4.5156e-04, 3.0309e-05, ..., 1.7136e-05, + 2.4028e-06, 2.4468e-05], + [-4.5300e-05, 6.1750e-05, 1.8045e-05, ..., 3.1143e-05, + 4.0308e-06, -4.3392e-05], + [-2.9713e-05, -1.6302e-05, -5.0247e-05, ..., -1.0663e-04, + -2.2817e-06, -2.0280e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0293, -0.0098, -0.0126, 0.0189, -0.0050, -0.0156, -0.0056, 0.0098, + -0.0234, 0.0178], device='cuda:0'), grad: tensor([-2.0540e-04, -1.1330e-02, 7.8278e-03, -1.4257e-03, 2.3174e-04, + 2.4414e-03, 1.8752e-04, 2.0466e-03, 1.8728e-04, 3.6448e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 220.05, cls_loss 0.0088 cls_loss_mapping 0.0261 cls_loss_causal 0.6739 re_mapping 0.0151 re_causal 0.0438 /// teacc 98.95 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0016, -0.0130, 0.0069, ..., -0.0479, -0.0170, -0.0069], + [-0.0482, 0.0200, -0.0160, ..., -0.0426, 0.0117, -0.0571], + [-0.0328, -0.0285, -0.0501, ..., -0.0412, -0.0087, 0.0600], + ..., + [ 0.0378, 0.0125, -0.0228, ..., 0.0331, -0.0068, 0.0020], + [ 0.0490, 0.0330, 0.0076, ..., -0.0264, -0.0379, 0.0271], + [ 0.0246, -0.0113, 0.0163, ..., 0.0267, -0.0392, 0.0257]], + device='cuda:0'), grad: tensor([[ 2.0131e-05, 3.4332e-05, 1.6570e-05, ..., 3.7313e-05, + 3.3230e-05, 4.6849e-05], + [ 4.3422e-05, 3.6836e-05, 3.6389e-05, ..., 5.1528e-05, + 3.5614e-05, 4.4763e-05], + [ 1.5408e-05, 3.9697e-05, 1.9759e-05, ..., 1.9863e-05, + 1.1414e-05, -3.7432e-05], + ..., + [ 8.9169e-05, 1.9884e-04, 9.7513e-05, ..., 1.6332e-04, + 8.4817e-05, 9.4235e-05], + [-2.1899e-04, -3.3712e-04, -2.8658e-04, ..., -1.3649e-04, + -5.7518e-05, -2.3580e-04], + [-6.9857e-04, -7.9441e-04, -4.8137e-04, ..., -1.4553e-03, + -6.1321e-04, -4.8304e-04]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0294, -0.0095, -0.0129, 0.0190, -0.0052, -0.0159, -0.0052, 0.0095, + -0.0235, 0.0181], device='cuda:0'), grad: tensor([ 8.5950e-05, 9.8109e-05, 1.0552e-06, 1.5152e-04, 1.4906e-03, + 5.8204e-05, 1.4447e-05, 3.1137e-04, -6.6185e-04, -1.5488e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 219.74, cls_loss 0.0099 cls_loss_mapping 0.0271 cls_loss_causal 0.6988 re_mapping 0.0143 re_causal 0.0441 /// teacc 98.93 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0015, -0.0138, 0.0070, ..., -0.0487, -0.0177, -0.0074], + [-0.0483, 0.0201, -0.0163, ..., -0.0424, 0.0110, -0.0582], + [-0.0332, -0.0292, -0.0504, ..., -0.0419, -0.0090, 0.0604], + ..., + [ 0.0382, 0.0127, -0.0233, ..., 0.0331, -0.0074, 0.0016], + [ 0.0492, 0.0335, 0.0079, ..., -0.0269, -0.0382, 0.0274], + [ 0.0249, -0.0115, 0.0164, ..., 0.0266, -0.0398, 0.0253]], + device='cuda:0'), grad: tensor([[ 2.1592e-05, 2.6882e-05, -1.0915e-05, ..., 2.0638e-05, + 4.5300e-05, 3.9309e-05], + [ 1.6088e-03, 8.3208e-04, 7.9349e-06, ..., 1.5612e-03, + 1.7256e-05, 1.8582e-05], + [ 5.3120e-04, 2.6560e-04, 1.3657e-05, ..., 7.6115e-05, + 1.2219e-05, -4.7266e-05], + ..., + [-2.9068e-03, -1.4648e-03, 2.0489e-05, ..., -1.9779e-03, + 7.7784e-06, -5.6535e-05], + [ 3.8356e-05, 1.0550e-05, 7.2904e-06, ..., 2.6226e-05, + 2.8163e-05, 1.4305e-05], + [ 2.4652e-04, 1.2314e-04, 2.9087e-05, ..., 1.9908e-04, + 2.8405e-06, 1.4201e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0292, -0.0096, -0.0129, 0.0187, -0.0043, -0.0156, -0.0053, 0.0098, + -0.0237, 0.0177], device='cuda:0'), grad: tensor([ 0.0004, 0.0034, 0.0015, 0.0015, 0.0002, -0.0009, -0.0004, -0.0069, + 0.0002, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 219.77, cls_loss 0.0084 cls_loss_mapping 0.0260 cls_loss_causal 0.7027 re_mapping 0.0147 re_causal 0.0448 /// teacc 98.92 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0013, -0.0132, 0.0071, ..., -0.0490, -0.0184, -0.0077], + [-0.0488, 0.0202, -0.0167, ..., -0.0430, 0.0098, -0.0584], + [-0.0335, -0.0301, -0.0508, ..., -0.0422, -0.0094, 0.0607], + ..., + [ 0.0389, 0.0127, -0.0237, ..., 0.0333, -0.0080, 0.0010], + [ 0.0498, 0.0338, 0.0084, ..., -0.0270, -0.0386, 0.0279], + [ 0.0249, -0.0115, 0.0167, ..., 0.0269, -0.0397, 0.0253]], + device='cuda:0'), grad: tensor([[-1.8358e-04, 1.4178e-05, 5.4479e-05, ..., 7.2494e-06, + 1.3721e-04, 4.7773e-05], + [ 7.2241e-05, 3.0488e-05, 4.5806e-05, ..., 2.8387e-05, + 5.8785e-06, 5.4359e-05], + [ 9.9182e-05, 3.0145e-05, 3.7253e-05, ..., 1.2286e-05, + 2.4922e-06, 1.2837e-05], + ..., + [-5.4091e-06, 2.1592e-05, 2.8759e-05, ..., -2.3752e-05, + 5.0887e-06, 2.0519e-05], + [-1.7321e-04, -3.1137e-04, -1.9956e-04, ..., 1.8487e-06, + 5.1260e-06, -2.8181e-04], + [ 1.7628e-05, 1.1772e-05, -3.4034e-05, ..., -2.1607e-05, + 2.9653e-05, 2.3961e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0296, -0.0098, -0.0133, 0.0186, -0.0044, -0.0157, -0.0053, 0.0102, + -0.0234, 0.0175], device='cuda:0'), grad: tensor([-1.6391e-04, 1.3983e-04, 2.3198e-04, 1.9562e-04, 1.2435e-05, + 2.7323e-04, -2.6226e-04, 9.2685e-05, -5.8270e-04, 6.3598e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 219.71, cls_loss 0.0088 cls_loss_mapping 0.0236 cls_loss_causal 0.6887 re_mapping 0.0141 re_causal 0.0426 /// teacc 98.86 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0009, -0.0138, 0.0072, ..., -0.0492, -0.0193, -0.0082], + [-0.0490, 0.0204, -0.0171, ..., -0.0430, 0.0094, -0.0589], + [-0.0339, -0.0306, -0.0511, ..., -0.0428, -0.0099, 0.0608], + ..., + [ 0.0390, 0.0121, -0.0248, ..., 0.0336, -0.0085, 0.0007], + [ 0.0505, 0.0340, 0.0086, ..., -0.0271, -0.0390, 0.0282], + [ 0.0251, -0.0114, 0.0174, ..., 0.0270, -0.0395, 0.0258]], + device='cuda:0'), grad: tensor([[-4.2766e-05, 9.2983e-05, -3.9726e-05, ..., 7.7784e-06, + 4.5747e-06, 1.9863e-05], + [-5.2065e-05, -5.4550e-04, -2.0945e-04, ..., 1.5533e-04, + 7.0408e-06, 6.8069e-05], + [ 6.2704e-05, 4.0084e-05, 4.3035e-05, ..., 2.4959e-05, + 3.1471e-05, -1.6797e-04], + ..., + [-4.4799e-04, -3.0684e-04, -1.0170e-05, ..., -4.4227e-04, + 4.2953e-06, 4.8488e-05], + [ 1.1480e-04, 3.5667e-04, 1.1975e-04, ..., 1.6809e-05, + 4.8466e-06, 4.2245e-06], + [ 2.2626e-04, 2.5535e-04, 9.1672e-05, ..., 1.6463e-04, + 6.5416e-06, 8.4117e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0296, -0.0098, -0.0135, 0.0193, -0.0047, -0.0159, -0.0053, 0.0101, + -0.0236, 0.0178], device='cuda:0'), grad: tensor([ 6.2048e-05, -4.1151e-04, -2.9850e-04, 1.0233e-03, 7.9513e-05, + -9.8038e-04, 2.3022e-05, -6.0463e-04, 5.4264e-04, 5.6505e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 219.86, cls_loss 0.0086 cls_loss_mapping 0.0265 cls_loss_causal 0.6864 re_mapping 0.0139 re_causal 0.0413 /// teacc 98.93 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0008, -0.0144, 0.0076, ..., -0.0496, -0.0199, -0.0085], + [-0.0493, 0.0206, -0.0174, ..., -0.0433, 0.0086, -0.0597], + [-0.0340, -0.0310, -0.0514, ..., -0.0433, -0.0103, 0.0614], + ..., + [ 0.0393, 0.0119, -0.0253, ..., 0.0338, -0.0088, 0.0002], + [ 0.0509, 0.0341, 0.0088, ..., -0.0275, -0.0392, 0.0284], + [ 0.0254, -0.0111, 0.0178, ..., 0.0274, -0.0397, 0.0256]], + device='cuda:0'), grad: tensor([[-2.0772e-05, 1.2472e-05, -1.7250e-04, ..., 1.3158e-05, + 1.4782e-05, -7.5817e-05], + [ 3.5077e-05, -4.3698e-06, 1.8030e-05, ..., 7.7188e-05, + 2.4244e-05, 4.0650e-05], + [ 1.4402e-05, 3.2961e-05, 4.8459e-05, ..., 2.4945e-05, + 1.4327e-05, 5.5522e-05], + ..., + [-2.4095e-05, 8.4281e-05, 3.4392e-05, ..., -2.6971e-05, + 2.9519e-05, 3.4660e-05], + [-4.1425e-05, -1.5771e-04, -6.3241e-05, ..., 8.0764e-05, + 1.4126e-05, -1.5974e-04], + [-1.0008e-04, -1.6809e-04, -7.9632e-05, ..., -1.6844e-04, + 4.2319e-05, 6.0827e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0297, -0.0101, -0.0131, 0.0187, -0.0043, -0.0155, -0.0052, 0.0100, + -0.0237, 0.0175], device='cuda:0'), grad: tensor([-3.8266e-04, 6.5506e-05, 1.9407e-04, 3.4022e-04, -5.5820e-05, + -1.1700e-04, 1.1200e-04, 7.5817e-05, -1.6928e-04, -6.2704e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 219.50, cls_loss 0.0084 cls_loss_mapping 0.0241 cls_loss_causal 0.7132 re_mapping 0.0137 re_causal 0.0417 /// teacc 98.91 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0005, -0.0152, 0.0078, ..., -0.0499, -0.0207, -0.0088], + [-0.0492, 0.0209, -0.0175, ..., -0.0434, 0.0076, -0.0602], + [-0.0343, -0.0308, -0.0519, ..., -0.0438, -0.0105, 0.0618], + ..., + [ 0.0394, 0.0112, -0.0257, ..., 0.0339, -0.0092, -0.0005], + [ 0.0513, 0.0343, 0.0090, ..., -0.0279, -0.0395, 0.0288], + [ 0.0257, -0.0108, 0.0180, ..., 0.0277, -0.0399, 0.0256]], + device='cuda:0'), grad: tensor([[-4.6864e-06, 8.0988e-06, 1.9610e-05, ..., 3.4012e-06, + 3.5375e-05, 3.0413e-05], + [ 7.7635e-06, -5.5023e-06, 9.3877e-06, ..., 1.4216e-05, + 1.6856e-04, 8.9347e-05], + [ 8.6501e-06, 8.5771e-05, 3.2127e-05, ..., 6.4559e-06, + 1.6212e-05, -1.3322e-05], + ..., + [-3.3975e-05, 6.0312e-06, 3.3788e-06, ..., -2.9370e-05, + 6.7912e-06, 9.0897e-06], + [-9.5591e-06, -1.9707e-06, 2.1886e-07, ..., 4.1761e-06, + 2.2754e-05, 3.5875e-06], + [ 9.1493e-06, 6.2250e-06, 1.3039e-05, ..., 6.7711e-05, + 4.6790e-05, 2.1070e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0298, -0.0101, -0.0127, 0.0187, -0.0046, -0.0154, -0.0049, 0.0096, + -0.0239, 0.0178], device='cuda:0'), grad: tensor([ 5.1647e-05, 2.3389e-04, 1.2803e-04, -1.3947e-04, -2.6003e-05, + 7.0393e-05, -4.2415e-04, -2.0117e-05, 3.8475e-05, 8.7440e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 219.73, cls_loss 0.0103 cls_loss_mapping 0.0277 cls_loss_causal 0.6814 re_mapping 0.0137 re_causal 0.0402 /// teacc 98.82 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0003, -0.0160, 0.0081, ..., -0.0501, -0.0216, -0.0093], + [-0.0497, 0.0211, -0.0180, ..., -0.0438, 0.0068, -0.0613], + [-0.0346, -0.0316, -0.0525, ..., -0.0444, -0.0107, 0.0625], + ..., + [ 0.0400, 0.0111, -0.0262, ..., 0.0349, -0.0089, -0.0008], + [ 0.0522, 0.0349, 0.0097, ..., -0.0282, -0.0400, 0.0291], + [ 0.0259, -0.0109, 0.0187, ..., 0.0276, -0.0400, 0.0258]], + device='cuda:0'), grad: tensor([[ 2.0564e-05, 6.7651e-05, 2.5582e-04, ..., 4.4815e-06, + 3.3736e-04, 3.3474e-04], + [ 1.4693e-05, 4.0650e-05, 1.0091e-04, ..., 1.1362e-05, + 1.2779e-04, 1.1939e-04], + [ 2.8789e-05, 6.8069e-05, 8.6844e-05, ..., 3.3945e-05, + 7.1108e-05, 6.6459e-05], + ..., + [ 3.0667e-05, 7.9274e-05, 7.8261e-05, ..., 5.2750e-05, + 1.7136e-05, 4.4793e-05], + [-1.6236e-04, -6.3360e-05, 5.5104e-05, ..., 1.1280e-05, + 2.5725e-04, -4.4703e-05], + [ 4.9859e-05, 7.1764e-04, 6.6948e-04, ..., 1.3340e-04, + 9.3937e-05, 1.9205e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0296, -0.0104, -0.0128, 0.0190, -0.0050, -0.0158, -0.0050, 0.0103, + -0.0238, 0.0178], device='cuda:0'), grad: tensor([ 7.8344e-04, 3.1495e-04, 2.4402e-04, 6.7024e-03, 1.9558e-06, + -9.7961e-03, -1.5116e-03, 2.9778e-04, 6.9094e-04, 2.2659e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 219.85, cls_loss 0.0081 cls_loss_mapping 0.0218 cls_loss_causal 0.6664 re_mapping 0.0130 re_causal 0.0392 /// teacc 98.90 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0002, -0.0157, 0.0085, ..., -0.0505, -0.0229, -0.0100], + [-0.0503, 0.0207, -0.0182, ..., -0.0441, 0.0063, -0.0620], + [-0.0347, -0.0321, -0.0529, ..., -0.0448, -0.0109, 0.0631], + ..., + [ 0.0404, 0.0114, -0.0268, ..., 0.0351, -0.0093, -0.0011], + [ 0.0525, 0.0349, 0.0098, ..., -0.0285, -0.0406, 0.0294], + [ 0.0261, -0.0110, 0.0185, ..., 0.0276, -0.0404, 0.0256]], + device='cuda:0'), grad: tensor([[-1.1707e-06, 1.6570e-05, 1.1444e-05, ..., 7.1824e-06, + 2.4408e-05, 2.8461e-05], + [ 1.2644e-05, -4.3750e-05, 1.2495e-05, ..., -9.5963e-06, + 2.9448e-06, 1.8880e-05], + [ 2.6822e-05, 4.4167e-05, 2.5541e-05, ..., 1.1370e-05, + 1.2867e-05, 1.3523e-05], + ..., + [ 8.0094e-06, 4.8608e-05, 1.7166e-05, ..., 3.5375e-05, + 1.9655e-05, 2.4602e-05], + [-1.3292e-04, 3.1590e-05, -5.0664e-05, ..., 4.8652e-06, + 1.5043e-05, -7.7784e-05], + [-3.6001e-05, 3.9153e-06, -2.2382e-05, ..., -5.4419e-05, + 7.9498e-06, 4.0717e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0296, -0.0109, -0.0123, 0.0192, -0.0040, -0.0161, -0.0052, 0.0106, + -0.0240, 0.0174], device='cuda:0'), grad: tensor([ 3.5226e-05, -7.6830e-05, 7.2181e-05, -3.7622e-04, 2.5511e-04, + 2.1410e-04, -2.0552e-04, 1.0437e-04, -4.0054e-05, 1.8343e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 219.72, cls_loss 0.0087 cls_loss_mapping 0.0216 cls_loss_causal 0.6556 re_mapping 0.0124 re_causal 0.0376 /// teacc 98.95 lr 0.00010000 +Epoch 48, weight, value: tensor([[ 1.0740e-05, -1.5618e-02, 8.7789e-03, ..., -5.1208e-02, + -2.3957e-02, -1.0562e-02], + [-4.9773e-02, 2.1763e-02, -1.8757e-02, ..., -4.4016e-02, + 5.6473e-03, -6.3223e-02], + [-3.5299e-02, -3.2546e-02, -5.3253e-02, ..., -4.5823e-02, + -1.1438e-02, 6.3599e-02], + ..., + [ 4.0679e-02, 1.0473e-02, -2.7122e-02, ..., 3.5520e-02, + -9.7560e-03, -1.2694e-03], + [ 5.3007e-02, 3.5353e-02, 1.0067e-02, ..., -2.8683e-02, + -4.1542e-02, 2.9722e-02], + [ 2.6278e-02, -1.0946e-02, 1.8780e-02, ..., 2.7671e-02, + -4.0165e-02, 2.5677e-02]], device='cuda:0'), grad: tensor([[ 1.5557e-05, 9.3102e-05, 8.9884e-05, ..., 6.3241e-05, + 1.2362e-04, 1.1134e-04], + [ 4.3124e-05, -2.2297e-03, 8.0585e-05, ..., -2.3079e-03, + -1.3733e-03, 5.2333e-05], + [ 5.8323e-05, 3.1757e-04, 1.5640e-04, ..., 1.7798e-04, + 1.8978e-04, 1.1986e-04], + ..., + [-3.0845e-05, 5.5504e-04, 1.6809e-04, ..., 4.1032e-04, + 4.0340e-04, 1.5354e-04], + [ 3.2544e-04, 1.0357e-03, 8.4448e-04, ..., 9.3222e-04, + 1.4277e-03, 1.1358e-03], + [-1.0386e-03, -1.5230e-03, -1.4191e-03, ..., -2.1362e-03, + -1.5125e-03, -1.6260e-03]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0295, -0.0103, -0.0123, 0.0191, -0.0039, -0.0161, -0.0054, 0.0103, + -0.0242, 0.0173], device='cuda:0'), grad: tensor([ 0.0003, -0.0060, 0.0008, 0.0001, 0.0060, 0.0006, -0.0021, 0.0013, + 0.0034, -0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 219.92, cls_loss 0.0064 cls_loss_mapping 0.0203 cls_loss_causal 0.7045 re_mapping 0.0126 re_causal 0.0393 /// teacc 98.92 lr 0.00010000 +Epoch 49, weight, value: tensor([[ 0.0005, -0.0162, 0.0091, ..., -0.0516, -0.0248, -0.0108], + [-0.0501, 0.0221, -0.0190, ..., -0.0442, 0.0054, -0.0638], + [-0.0353, -0.0334, -0.0539, ..., -0.0463, -0.0119, 0.0636], + ..., + [ 0.0410, 0.0104, -0.0277, ..., 0.0360, -0.0103, -0.0018], + [ 0.0534, 0.0356, 0.0102, ..., -0.0291, -0.0419, 0.0303], + [ 0.0264, -0.0107, 0.0192, ..., 0.0277, -0.0402, 0.0258]], + device='cuda:0'), grad: tensor([[-4.5538e-05, -1.2052e-04, -9.1672e-05, ..., 9.2804e-05, + 1.8978e-04, 6.3516e-06], + [ 1.4171e-05, 1.9111e-06, 3.9637e-05, ..., 6.0439e-05, + 1.5175e-04, 1.1909e-04], + [ 1.9714e-05, 7.4625e-05, 1.0484e-04, ..., 3.1853e-04, + 3.7122e-04, 4.0555e-04], + ..., + [-2.0787e-05, 3.6180e-05, 2.9087e-05, ..., -2.2873e-06, + 3.6180e-05, 6.8069e-05], + [ 3.0309e-05, 3.1531e-05, 3.2544e-05, ..., 6.0439e-05, + 6.2168e-05, 6.4909e-05], + [-7.6354e-05, -9.7334e-05, -4.6849e-05, ..., -1.3626e-04, + -3.2634e-06, -6.7711e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0297, -0.0102, -0.0124, 0.0193, -0.0044, -0.0159, -0.0052, 0.0102, + -0.0242, 0.0171], device='cuda:0'), grad: tensor([-5.4646e-04, 3.1424e-04, 9.8991e-04, 2.7180e-04, 1.2655e-03, + 9.7379e-06, -2.4548e-03, 4.2886e-05, 1.9884e-04, -9.1851e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 48---------------------------------------------------- +epoch 48, time 220.62, cls_loss 0.0069 cls_loss_mapping 0.0192 cls_loss_causal 0.7000 re_mapping 0.0120 re_causal 0.0373 /// teacc 99.05 lr 0.00010000 +Epoch 50, weight, value: tensor([[ 0.0006, -0.0169, 0.0091, ..., -0.0518, -0.0259, -0.0114], + [-0.0506, 0.0225, -0.0194, ..., -0.0445, 0.0046, -0.0646], + [-0.0357, -0.0341, -0.0542, ..., -0.0470, -0.0123, 0.0643], + ..., + [ 0.0416, 0.0102, -0.0281, ..., 0.0363, -0.0110, -0.0022], + [ 0.0539, 0.0360, 0.0106, ..., -0.0295, -0.0423, 0.0307], + [ 0.0266, -0.0106, 0.0196, ..., 0.0277, -0.0404, 0.0257]], + device='cuda:0'), grad: tensor([[ 5.0627e-06, 1.4953e-05, 2.9653e-05, ..., 2.9919e-07, + 3.9607e-05, 2.8566e-05], + [ 1.7852e-05, 2.0593e-05, 4.1217e-05, ..., 7.5363e-06, + 8.4460e-05, 5.3406e-05], + [ 1.3985e-05, 1.7062e-05, 2.0429e-05, ..., -1.3836e-05, + 3.0145e-05, -9.2924e-05], + ..., + [-4.3474e-06, 3.2395e-05, 2.6897e-05, ..., -6.5975e-06, + 6.7279e-06, 3.2455e-05], + [-2.8908e-06, 9.8348e-05, 1.1182e-04, ..., -6.4336e-06, + 1.3089e-04, -3.7283e-05], + [ 1.8299e-04, 1.4555e-04, 3.8433e-04, ..., 1.5974e-05, + 2.0698e-05, 4.8071e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0294, -0.0103, -0.0121, 0.0191, -0.0046, -0.0160, -0.0045, 0.0104, + -0.0242, 0.0169], device='cuda:0'), grad: tensor([ 0.0001, 0.0002, -0.0001, -0.0004, 0.0002, -0.0011, -0.0005, 0.0001, + 0.0003, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 219.86, cls_loss 0.0061 cls_loss_mapping 0.0200 cls_loss_causal 0.6596 re_mapping 0.0119 re_causal 0.0366 /// teacc 98.85 lr 0.00010000 +Epoch 51, weight, value: tensor([[ 0.0008, -0.0171, 0.0093, ..., -0.0523, -0.0267, -0.0117], + [-0.0505, 0.0228, -0.0197, ..., -0.0447, 0.0029, -0.0655], + [-0.0361, -0.0347, -0.0545, ..., -0.0476, -0.0124, 0.0650], + ..., + [ 0.0419, 0.0099, -0.0287, ..., 0.0364, -0.0115, -0.0027], + [ 0.0543, 0.0361, 0.0108, ..., -0.0298, -0.0430, 0.0309], + [ 0.0267, -0.0104, 0.0199, ..., 0.0278, -0.0408, 0.0254]], + device='cuda:0'), grad: tensor([[-1.5044e-04, -1.6642e-04, -2.2399e-04, ..., -2.0236e-05, + 2.7925e-05, -3.9816e-05], + [ 1.3220e-04, 1.3137e-04, 1.1511e-05, ..., 5.2786e-04, + 2.0301e-04, 1.2279e-04], + [ 5.7995e-05, 4.5329e-05, 3.4690e-05, ..., 8.2791e-05, + 3.3468e-05, 4.9382e-05], + ..., + [-4.6301e-04, -3.2330e-04, 1.9863e-05, ..., -1.5802e-03, + -3.8815e-04, -1.7786e-04], + [-1.0267e-05, 2.0012e-05, 1.2383e-05, ..., 7.7069e-05, + 4.9412e-05, -1.9781e-06], + [ 5.1856e-05, 5.5075e-05, 5.2303e-05, ..., 1.3256e-04, + 6.4611e-05, 5.4985e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0294, -0.0106, -0.0119, 0.0193, -0.0044, -0.0159, -0.0042, 0.0103, + -0.0246, 0.0168], device='cuda:0'), grad: tensor([-7.6628e-04, 5.4979e-04, 1.9503e-04, 1.9264e-04, 6.0797e-04, + 1.6654e-04, -1.9148e-05, -1.4944e-03, 1.0794e-04, 4.5967e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 219.71, cls_loss 0.0053 cls_loss_mapping 0.0168 cls_loss_causal 0.6159 re_mapping 0.0117 re_causal 0.0342 /// teacc 98.86 lr 0.00010000 +Epoch 52, weight, value: tensor([[ 0.0010, -0.0174, 0.0096, ..., -0.0526, -0.0272, -0.0121], + [-0.0512, 0.0227, -0.0202, ..., -0.0450, 0.0023, -0.0664], + [-0.0365, -0.0354, -0.0549, ..., -0.0480, -0.0128, 0.0653], + ..., + [ 0.0422, 0.0097, -0.0292, ..., 0.0368, -0.0118, -0.0031], + [ 0.0551, 0.0367, 0.0111, ..., -0.0301, -0.0435, 0.0316], + [ 0.0269, -0.0103, 0.0202, ..., 0.0280, -0.0409, 0.0253]], + device='cuda:0'), grad: tensor([[ 1.6373e-06, 2.9728e-05, 1.1630e-05, ..., 2.5228e-05, + 4.7833e-05, 5.7280e-05], + [-8.6725e-06, -8.0645e-05, 4.2677e-05, ..., 3.0287e-06, + 3.7998e-05, 6.9499e-05], + [ 4.2129e-04, 6.1750e-05, 3.1590e-05, ..., 4.3178e-04, + 1.4281e-04, 1.4496e-04], + ..., + [-4.7588e-04, 6.0827e-05, 2.5272e-05, ..., -4.9257e-04, + 9.9167e-06, -2.2173e-05], + [-7.3254e-05, -8.3828e-04, -3.2783e-04, ..., 1.7273e-04, + -1.5306e-04, -4.7636e-04], + [-2.6774e-04, -2.2268e-04, -2.7394e-04, ..., -3.5977e-04, + -1.5795e-04, -2.4652e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0294, -0.0111, -0.0120, 0.0192, -0.0046, -0.0153, -0.0045, 0.0103, + -0.0240, 0.0167], device='cuda:0'), grad: tensor([ 1.0967e-04, -7.4685e-05, 1.2617e-03, 8.1110e-04, 5.9462e-04, + 6.9201e-05, 1.7178e-04, -1.1387e-03, -1.1196e-03, -6.8474e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 220.16, cls_loss 0.0053 cls_loss_mapping 0.0174 cls_loss_causal 0.6285 re_mapping 0.0111 re_causal 0.0348 /// teacc 98.97 lr 0.00010000 +Epoch 53, weight, value: tensor([[ 0.0006, -0.0184, 0.0094, ..., -0.0542, -0.0283, -0.0125], + [-0.0513, 0.0230, -0.0205, ..., -0.0454, 0.0018, -0.0671], + [-0.0371, -0.0360, -0.0552, ..., -0.0490, -0.0135, 0.0655], + ..., + [ 0.0428, 0.0095, -0.0296, ..., 0.0375, -0.0117, -0.0030], + [ 0.0552, 0.0368, 0.0111, ..., -0.0306, -0.0442, 0.0316], + [ 0.0276, -0.0100, 0.0210, ..., 0.0282, -0.0411, 0.0254]], + device='cuda:0'), grad: tensor([[ 6.2399e-06, 1.9699e-05, -3.5763e-06, ..., 2.6152e-05, + 4.6119e-06, 1.7837e-05], + [ 2.1681e-05, 5.7481e-06, 1.8567e-05, ..., 7.4916e-06, + 1.5255e-06, 2.4572e-05], + [ 5.1081e-05, 5.1886e-05, 3.5048e-05, ..., 6.2227e-05, + 5.8748e-06, 1.4700e-05], + ..., + [-4.1574e-05, 4.0472e-05, 2.3380e-05, ..., 9.5367e-05, + 1.0639e-05, 2.5943e-05], + [-1.6487e-04, -1.6677e-04, -1.7083e-04, ..., 5.4777e-05, + 7.9796e-06, -2.2948e-04], + [-6.4254e-05, -5.3942e-05, -4.1157e-05, ..., -1.5700e-04, + -4.4703e-05, -3.5495e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0289, -0.0109, -0.0123, 0.0192, -0.0046, -0.0155, -0.0041, 0.0108, + -0.0243, 0.0169], device='cuda:0'), grad: tensor([ 1.3888e-04, 6.0320e-05, 4.4227e-04, 5.8365e-04, 2.9755e-04, + -2.0103e-03, 1.3888e-04, 5.6219e-04, -1.9538e-04, -1.8075e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 220.28, cls_loss 0.0054 cls_loss_mapping 0.0162 cls_loss_causal 0.6783 re_mapping 0.0114 re_causal 0.0350 /// teacc 99.03 lr 0.00010000 +Epoch 54, weight, value: tensor([[ 0.0010, -0.0189, 0.0098, ..., -0.0542, -0.0282, -0.0127], + [-0.0515, 0.0235, -0.0208, ..., -0.0455, 0.0012, -0.0676], + [-0.0373, -0.0366, -0.0554, ..., -0.0498, -0.0140, 0.0661], + ..., + [ 0.0429, 0.0092, -0.0302, ..., 0.0377, -0.0121, -0.0035], + [ 0.0558, 0.0372, 0.0115, ..., -0.0310, -0.0444, 0.0321], + [ 0.0280, -0.0099, 0.0212, ..., 0.0286, -0.0412, 0.0254]], + device='cuda:0'), grad: tensor([[ 6.7770e-05, 2.8655e-05, 1.0973e-04, ..., 7.7844e-05, + 1.9550e-04, 2.0671e-04], + [ 3.0443e-05, 1.3761e-05, 1.1697e-05, ..., 2.3827e-05, + 2.9311e-05, 1.1432e-04], + [-5.9187e-05, -1.3554e-04, 2.2426e-05, ..., 2.8253e-05, + 2.0638e-05, -3.1400e-04], + ..., + [ 1.9789e-04, 2.8992e-04, 1.9872e-04, ..., 2.6703e-04, + 7.6175e-05, 2.7132e-04], + [ 1.2386e-04, 1.3542e-04, 1.4257e-04, ..., 2.1839e-04, + 3.8534e-05, 1.1802e-04], + [-5.7268e-04, -5.3740e-04, -5.9843e-04, ..., -7.8058e-04, + -1.2589e-04, -3.6049e-04]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0293, -0.0107, -0.0122, 0.0191, -0.0049, -0.0155, -0.0045, 0.0109, + -0.0243, 0.0168], device='cuda:0'), grad: tensor([ 0.0004, 0.0003, -0.0013, 0.0003, 0.0003, 0.0001, -0.0004, 0.0011, + 0.0004, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 219.98, cls_loss 0.0054 cls_loss_mapping 0.0194 cls_loss_causal 0.6798 re_mapping 0.0110 re_causal 0.0342 /// teacc 98.89 lr 0.00010000 +Epoch 55, weight, value: tensor([[ 0.0011, -0.0194, 0.0098, ..., -0.0546, -0.0290, -0.0133], + [-0.0517, 0.0239, -0.0212, ..., -0.0457, 0.0006, -0.0683], + [-0.0377, -0.0372, -0.0558, ..., -0.0505, -0.0144, 0.0664], + ..., + [ 0.0434, 0.0089, -0.0306, ..., 0.0381, -0.0126, -0.0041], + [ 0.0563, 0.0376, 0.0119, ..., -0.0313, -0.0447, 0.0327], + [ 0.0281, -0.0100, 0.0216, ..., 0.0286, -0.0414, 0.0256]], + device='cuda:0'), grad: tensor([[ 8.0287e-05, 1.7440e-04, 1.0550e-04, ..., 2.3283e-06, + 7.9095e-05, 2.6870e-04], + [ 4.5806e-05, 3.5822e-05, 1.6063e-05, ..., 6.7353e-05, + 6.9571e-04, 5.0354e-04], + [ 5.3823e-05, 6.8784e-05, 5.6863e-05, ..., 2.0340e-05, + 9.4223e-04, 7.1621e-04], + ..., + [-2.3782e-05, 7.9721e-06, 3.2634e-05, ..., -2.5347e-05, + 6.1631e-05, 7.4565e-05], + [-2.6584e-04, -4.3297e-04, -2.2924e-04, ..., 2.3529e-05, + 1.7196e-05, -5.6553e-04], + [ 1.1399e-05, 3.4004e-05, 1.3635e-05, ..., -7.5877e-05, + -1.1891e-05, 5.5909e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0290, -0.0106, -0.0124, 0.0194, -0.0051, -0.0156, -0.0043, 0.0110, + -0.0242, 0.0167], device='cuda:0'), grad: tensor([ 0.0006, 0.0025, 0.0033, -0.0002, 0.0004, 0.0005, -0.0067, 0.0001, + -0.0006, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 219.85, cls_loss 0.0043 cls_loss_mapping 0.0141 cls_loss_causal 0.6528 re_mapping 0.0110 re_causal 0.0348 /// teacc 98.86 lr 0.00010000 +Epoch 56, weight, value: tensor([[ 0.0014, -0.0198, 0.0099, ..., -0.0547, -0.0295, -0.0137], + [-0.0523, 0.0240, -0.0216, ..., -0.0464, -0.0003, -0.0690], + [-0.0379, -0.0379, -0.0562, ..., -0.0510, -0.0150, 0.0667], + ..., + [ 0.0440, 0.0088, -0.0310, ..., 0.0386, -0.0128, -0.0045], + [ 0.0567, 0.0380, 0.0121, ..., -0.0316, -0.0451, 0.0330], + [ 0.0282, -0.0099, 0.0219, ..., 0.0286, -0.0415, 0.0257]], + device='cuda:0'), grad: tensor([[-9.0674e-06, 1.1064e-05, -1.5140e-05, ..., -2.3469e-06, + 7.2271e-06, 1.2837e-05], + [ 1.8030e-06, 7.8678e-06, 1.0870e-05, ..., 1.6466e-05, + 2.9445e-05, 3.2336e-05], + [ 9.1791e-06, 2.0742e-05, 1.3895e-05, ..., 3.2410e-06, + 1.4976e-05, -7.2837e-05], + ..., + [-8.9630e-06, 2.3097e-05, 1.4283e-05, ..., 9.5591e-06, + 9.7081e-06, 2.7448e-05], + [ 4.7594e-05, -7.3719e-04, -4.1103e-04, ..., 2.8133e-05, + -7.5960e-04, -9.5177e-04], + [-8.6784e-05, -4.0382e-05, -8.4460e-05, ..., -8.0228e-05, + -1.3776e-05, -5.8115e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0291, -0.0110, -0.0127, 0.0196, -0.0049, -0.0158, -0.0040, 0.0113, + -0.0242, 0.0166], device='cuda:0'), grad: tensor([-2.3335e-05, 4.5031e-05, -5.4806e-05, -2.0877e-05, 4.4614e-05, + -4.2021e-05, 1.3485e-03, 4.1425e-05, -1.2245e-03, -1.1462e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 219.59, cls_loss 0.0062 cls_loss_mapping 0.0182 cls_loss_causal 0.6388 re_mapping 0.0113 re_causal 0.0338 /// teacc 98.91 lr 0.00010000 +Epoch 57, weight, value: tensor([[ 0.0017, -0.0199, 0.0102, ..., -0.0549, -0.0303, -0.0142], + [-0.0522, 0.0248, -0.0217, ..., -0.0462, -0.0009, -0.0696], + [-0.0382, -0.0381, -0.0565, ..., -0.0513, -0.0152, 0.0674], + ..., + [ 0.0440, 0.0082, -0.0316, ..., 0.0384, -0.0132, -0.0049], + [ 0.0570, 0.0379, 0.0121, ..., -0.0321, -0.0455, 0.0335], + [ 0.0286, -0.0097, 0.0223, ..., 0.0291, -0.0416, 0.0256]], + device='cuda:0'), grad: tensor([[-3.1106e-06, 1.0476e-05, 7.5661e-06, ..., 4.7982e-06, + 8.1584e-06, 1.7524e-05], + [ 1.1556e-05, -3.7178e-06, 9.9838e-06, ..., 2.1800e-05, + 7.9200e-06, 2.6867e-05], + [ 7.1041e-06, 1.8597e-05, -7.3947e-07, ..., 9.7454e-06, + 4.8168e-06, -4.1872e-05], + ..., + [-3.5614e-05, 1.0484e-04, 4.6074e-05, ..., -1.9729e-05, + 6.1728e-06, 2.1830e-05], + [-1.5087e-05, -1.5691e-05, -3.9309e-05, ..., 8.3223e-06, + 5.3719e-06, -4.6819e-05], + [-3.2056e-06, 3.6042e-06, -1.5900e-05, ..., 1.4074e-05, + 3.2425e-05, 2.1532e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0293, -0.0104, -0.0125, 0.0196, -0.0050, -0.0155, -0.0046, 0.0108, + -0.0246, 0.0168], device='cuda:0'), grad: tensor([ 4.1604e-05, 7.6234e-05, -1.1170e-04, -2.1112e-04, -6.6638e-05, + 6.4492e-05, 3.1688e-07, 1.6856e-04, -1.4104e-05, 5.2869e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 219.69, cls_loss 0.0055 cls_loss_mapping 0.0156 cls_loss_causal 0.6628 re_mapping 0.0105 re_causal 0.0320 /// teacc 98.88 lr 0.00010000 +Epoch 58, weight, value: tensor([[ 0.0015, -0.0202, 0.0102, ..., -0.0554, -0.0310, -0.0146], + [-0.0527, 0.0251, -0.0221, ..., -0.0470, -0.0015, -0.0703], + [-0.0384, -0.0387, -0.0568, ..., -0.0519, -0.0157, 0.0678], + ..., + [ 0.0445, 0.0080, -0.0320, ..., 0.0388, -0.0136, -0.0054], + [ 0.0574, 0.0380, 0.0127, ..., -0.0325, -0.0462, 0.0338], + [ 0.0290, -0.0095, 0.0228, ..., 0.0295, -0.0416, 0.0257]], + device='cuda:0'), grad: tensor([[ 5.2959e-05, 4.6730e-05, 1.3255e-05, ..., 1.6704e-05, + 7.6115e-05, 9.9719e-05], + [-6.1989e-05, -2.6321e-04, -4.6253e-05, ..., 1.1399e-05, + 3.0518e-05, 3.1859e-05], + [ 1.2743e-04, 7.4923e-05, 6.8069e-05, ..., 4.3392e-05, + 2.8268e-05, 8.5890e-05], + ..., + [-9.3877e-05, 4.8161e-05, 3.7581e-05, ..., -5.4717e-05, + 1.2510e-05, 3.3110e-05], + [-3.4118e-04, 4.2677e-05, -4.5443e-04, ..., 7.6234e-05, + -3.9428e-05, -5.3692e-04], + [ 9.0003e-05, -9.0420e-05, 1.4439e-05, ..., -1.0705e-04, + 4.1038e-05, 1.4961e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0292, -0.0104, -0.0123, 0.0193, -0.0051, -0.0156, -0.0052, 0.0109, + -0.0241, 0.0171], device='cuda:0'), grad: tensor([ 1.3256e-04, -5.4836e-04, 2.7800e-04, 5.3596e-04, 3.8910e-04, + 2.6441e-04, -4.6039e-04, 4.2051e-05, -9.3079e-04, 2.9778e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 219.61, cls_loss 0.0048 cls_loss_mapping 0.0151 cls_loss_causal 0.6410 re_mapping 0.0107 re_causal 0.0324 /// teacc 99.05 lr 0.00010000 +Epoch 59, weight, value: tensor([[ 0.0021, -0.0205, 0.0105, ..., -0.0556, -0.0320, -0.0152], + [-0.0528, 0.0256, -0.0223, ..., -0.0470, -0.0020, -0.0710], + [-0.0384, -0.0393, -0.0573, ..., -0.0526, -0.0161, 0.0683], + ..., + [ 0.0448, 0.0077, -0.0324, ..., 0.0392, -0.0139, -0.0059], + [ 0.0578, 0.0383, 0.0129, ..., -0.0330, -0.0467, 0.0341], + [ 0.0291, -0.0094, 0.0230, ..., 0.0297, -0.0417, 0.0256]], + device='cuda:0'), grad: tensor([[-6.0350e-06, 5.4128e-06, -2.6170e-06, ..., 5.1223e-06, + 9.4771e-06, 1.1355e-05], + [-3.2857e-06, -3.5167e-05, 5.6773e-06, ..., 2.4453e-05, + 2.0966e-05, 1.9118e-05], + [ 1.5991e-06, 1.4924e-05, 6.4224e-06, ..., 1.0766e-05, + 4.9323e-05, 3.3170e-05], + ..., + [ 2.5760e-06, 2.4930e-05, 6.6794e-06, ..., 2.6375e-05, + 1.8567e-05, 1.7717e-05], + [-5.2154e-08, 7.4148e-05, 2.2501e-05, ..., 2.8476e-05, + 1.3322e-05, 6.8843e-06], + [-1.9684e-05, 8.4490e-06, -1.6093e-05, ..., 1.9029e-05, + 3.8028e-05, 1.7241e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0292, -0.0103, -0.0123, 0.0190, -0.0053, -0.0157, -0.0044, 0.0112, + -0.0242, 0.0168], device='cuda:0'), grad: tensor([ 1.2912e-05, -4.9204e-05, 7.1585e-05, -1.0598e-04, -1.2493e-04, + -1.4082e-05, -6.4135e-05, 6.4850e-05, 1.5628e-04, 5.2661e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 219.65, cls_loss 0.0050 cls_loss_mapping 0.0126 cls_loss_causal 0.6334 re_mapping 0.0105 re_causal 0.0328 /// teacc 98.81 lr 0.00010000 +Epoch 60, weight, value: tensor([[ 0.0022, -0.0210, 0.0106, ..., -0.0559, -0.0326, -0.0155], + [-0.0531, 0.0259, -0.0226, ..., -0.0477, -0.0026, -0.0717], + [-0.0386, -0.0399, -0.0577, ..., -0.0533, -0.0167, 0.0691], + ..., + [ 0.0449, 0.0074, -0.0328, ..., 0.0395, -0.0143, -0.0070], + [ 0.0583, 0.0384, 0.0132, ..., -0.0333, -0.0472, 0.0344], + [ 0.0297, -0.0090, 0.0232, ..., 0.0301, -0.0417, 0.0257]], + device='cuda:0'), grad: tensor([[ 3.1650e-05, 3.8981e-05, 4.7624e-05, ..., 9.7901e-06, + 3.6936e-06, 1.7986e-05], + [ 2.2948e-04, 2.1422e-04, 1.6415e-04, ..., 3.8445e-05, + 3.8743e-06, 8.5235e-05], + [ 8.2433e-05, 5.9128e-05, 3.5942e-05, ..., 7.3254e-05, + 4.0494e-06, 1.2629e-05], + ..., + [-1.3709e-04, 1.5097e-06, 2.7642e-05, ..., -2.3341e-04, + 2.1812e-06, -4.6402e-05], + [-3.8123e-04, -6.6471e-04, -2.2352e-04, ..., 2.2963e-05, + 4.5002e-06, -2.4986e-04], + [-1.2159e-04, -1.6165e-04, 3.4064e-05, ..., -3.2544e-04, + -7.6592e-05, -9.8586e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0292, -0.0103, -0.0121, 0.0191, -0.0054, -0.0158, -0.0044, 0.0107, + -0.0243, 0.0172], device='cuda:0'), grad: tensor([ 0.0002, 0.0006, 0.0003, 0.0010, 0.0005, -0.0035, 0.0013, -0.0003, + -0.0005, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 219.65, cls_loss 0.0053 cls_loss_mapping 0.0177 cls_loss_causal 0.6333 re_mapping 0.0102 re_causal 0.0308 /// teacc 98.89 lr 0.00010000 +Epoch 61, weight, value: tensor([[ 0.0020, -0.0216, 0.0106, ..., -0.0566, -0.0332, -0.0160], + [-0.0534, 0.0265, -0.0229, ..., -0.0476, -0.0031, -0.0722], + [-0.0391, -0.0408, -0.0578, ..., -0.0545, -0.0172, 0.0696], + ..., + [ 0.0456, 0.0072, -0.0331, ..., 0.0406, -0.0141, -0.0073], + [ 0.0588, 0.0388, 0.0136, ..., -0.0336, -0.0476, 0.0348], + [ 0.0298, -0.0091, 0.0234, ..., 0.0299, -0.0425, 0.0251]], + device='cuda:0'), grad: tensor([[-3.0294e-05, 6.5528e-06, -2.3037e-05, ..., 2.0340e-06, + 1.9103e-05, 1.8999e-05], + [ 1.4499e-05, 2.3842e-05, 5.5991e-06, ..., 1.3447e-04, + 8.7500e-05, 3.8058e-05], + [ 1.3888e-05, 7.2643e-06, 5.3011e-06, ..., 2.4110e-05, + 1.8001e-05, -4.4182e-06], + ..., + [-1.7095e-04, -1.0157e-04, -1.4054e-06, ..., 6.0908e-06, + 1.3089e-04, 4.9502e-05], + [-3.8296e-05, -5.1081e-05, -3.3617e-05, ..., 8.7693e-06, + 7.2122e-06, -4.1217e-05], + [ 7.0751e-05, 5.4687e-05, 3.2723e-05, ..., 6.4671e-05, + 1.5900e-05, 2.5570e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0289, -0.0099, -0.0124, 0.0193, -0.0053, -0.0157, -0.0045, 0.0111, + -0.0243, 0.0166], device='cuda:0'), grad: tensor([-4.9204e-05, 2.0981e-04, 4.2111e-05, 2.1708e-04, -5.2881e-04, + 3.7998e-05, 2.7940e-05, -7.7903e-05, -7.4923e-05, 1.9574e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 219.32, cls_loss 0.0052 cls_loss_mapping 0.0136 cls_loss_causal 0.6298 re_mapping 0.0100 re_causal 0.0303 /// teacc 98.98 lr 0.00010000 +Epoch 62, weight, value: tensor([[ 0.0025, -0.0213, 0.0110, ..., -0.0572, -0.0335, -0.0163], + [-0.0536, 0.0265, -0.0232, ..., -0.0480, -0.0038, -0.0738], + [-0.0396, -0.0409, -0.0582, ..., -0.0551, -0.0179, 0.0701], + ..., + [ 0.0458, 0.0068, -0.0338, ..., 0.0408, -0.0148, -0.0078], + [ 0.0592, 0.0391, 0.0139, ..., -0.0341, -0.0484, 0.0352], + [ 0.0302, -0.0092, 0.0239, ..., 0.0300, -0.0430, 0.0250]], + device='cuda:0'), grad: tensor([[ 2.7567e-07, 1.7747e-05, 1.9595e-05, ..., 6.0350e-06, + 2.7549e-06, 2.4587e-05], + [ 9.4473e-05, -9.6619e-05, 6.6519e-05, ..., -1.0610e-04, + -4.0203e-05, 8.8453e-05], + [-3.9220e-05, 7.5459e-05, -1.0318e-04, ..., 1.1139e-05, + 3.5409e-06, -2.2483e-04], + ..., + [-1.6391e-05, 7.4029e-05, 4.4703e-05, ..., -8.1733e-06, + 6.2250e-06, 4.1008e-05], + [ 3.7581e-05, 1.1635e-04, 4.4763e-05, ..., 3.8534e-05, + -9.2201e-07, 8.6069e-05], + [-2.2209e-04, -4.1080e-04, -1.2755e-04, ..., -8.6665e-05, + 3.3043e-06, -1.9073e-04]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0294, -0.0103, -0.0126, 0.0194, -0.0051, -0.0157, -0.0046, 0.0111, + -0.0242, 0.0165], device='cuda:0'), grad: tensor([ 1.4198e-04, -1.3149e-04, -8.7500e-04, -9.5904e-05, 4.8208e-04, + 5.1451e-04, 4.1485e-05, 3.0208e-04, 4.8828e-04, -8.6689e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 219.81, cls_loss 0.0047 cls_loss_mapping 0.0155 cls_loss_causal 0.6351 re_mapping 0.0100 re_causal 0.0304 /// teacc 98.76 lr 0.00010000 +Epoch 63, weight, value: tensor([[ 0.0027, -0.0218, 0.0112, ..., -0.0575, -0.0346, -0.0172], + [-0.0543, 0.0266, -0.0235, ..., -0.0488, -0.0045, -0.0747], + [-0.0398, -0.0417, -0.0585, ..., -0.0556, -0.0184, 0.0707], + ..., + [ 0.0462, 0.0069, -0.0343, ..., 0.0409, -0.0150, -0.0082], + [ 0.0595, 0.0393, 0.0140, ..., -0.0346, -0.0491, 0.0353], + [ 0.0307, -0.0089, 0.0241, ..., 0.0306, -0.0432, 0.0248]], + device='cuda:0'), grad: tensor([[ 7.8008e-06, 3.3915e-05, 1.8030e-05, ..., 1.5676e-05, + 1.1578e-05, 1.6108e-05], + [-1.3304e-04, -8.5783e-04, -3.5548e-04, ..., 1.5020e-04, + -3.3379e-04, -2.7847e-04], + [ 8.7321e-06, 2.7373e-05, 4.2111e-05, ..., 9.9242e-05, + 4.5151e-05, 1.5700e-04], + ..., + [-8.5384e-06, 3.4660e-05, 2.6554e-05, ..., 9.3126e-04, + 3.0875e-04, 6.1798e-04], + [ 6.3896e-05, 3.2973e-04, 1.0228e-04, ..., 6.2644e-05, + 1.0538e-04, -5.8085e-05], + [-9.3102e-05, -7.4506e-05, -6.1333e-05, ..., -7.8499e-05, + -1.5631e-05, -3.8475e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0293, -0.0109, -0.0123, 0.0192, -0.0049, -0.0154, -0.0041, 0.0111, + -0.0246, 0.0166], device='cuda:0'), grad: tensor([ 1.2898e-04, -1.8263e-03, 3.7408e-04, -2.8744e-03, -2.0046e-03, + 2.6054e-03, 1.4534e-03, 1.5097e-03, 6.6519e-04, -2.8476e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 219.41, cls_loss 0.0070 cls_loss_mapping 0.0184 cls_loss_causal 0.6339 re_mapping 0.0098 re_causal 0.0292 /// teacc 99.00 lr 0.00010000 +Epoch 64, weight, value: tensor([[ 0.0023, -0.0233, 0.0108, ..., -0.0583, -0.0354, -0.0179], + [-0.0540, 0.0275, -0.0238, ..., -0.0485, -0.0050, -0.0756], + [-0.0405, -0.0427, -0.0591, ..., -0.0566, -0.0187, 0.0712], + ..., + [ 0.0465, 0.0062, -0.0349, ..., 0.0407, -0.0156, -0.0082], + [ 0.0600, 0.0399, 0.0146, ..., -0.0351, -0.0494, 0.0357], + [ 0.0314, -0.0092, 0.0243, ..., 0.0303, -0.0441, 0.0242]], + device='cuda:0'), grad: tensor([[-2.3887e-05, 3.7625e-06, -1.0774e-05, ..., 2.9802e-06, + 6.7465e-06, 2.1890e-05], + [ 9.5293e-06, -1.6749e-05, 6.1020e-06, ..., 1.4063e-06, + 6.6869e-06, 3.2872e-05], + [ 9.7230e-06, 4.9919e-06, 3.5726e-06, ..., 9.7528e-06, + 5.6811e-06, -2.1291e-04], + ..., + [-8.7172e-06, 8.4713e-06, 4.3213e-06, ..., -4.5300e-06, + 3.3025e-06, 8.2493e-05], + [-1.3627e-05, -1.6570e-05, -1.2755e-05, ..., 7.9796e-06, + 9.2611e-06, 1.8448e-05], + [-6.7875e-06, -7.5549e-06, -8.1658e-06, ..., -5.7630e-06, + 1.0043e-05, -6.7754e-07]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0288, -0.0102, -0.0117, 0.0195, -0.0038, -0.0159, -0.0040, 0.0103, + -0.0246, 0.0160], device='cuda:0'), grad: tensor([ 1.6671e-06, 5.2482e-05, -5.2357e-04, 1.0431e-04, 5.1945e-05, + -5.5730e-05, 3.9876e-05, 2.2304e-04, 7.3910e-05, 3.1978e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 219.59, cls_loss 0.0055 cls_loss_mapping 0.0165 cls_loss_causal 0.6271 re_mapping 0.0099 re_causal 0.0301 /// teacc 98.94 lr 0.00010000 +Epoch 65, weight, value: tensor([[ 0.0024, -0.0238, 0.0109, ..., -0.0587, -0.0360, -0.0184], + [-0.0538, 0.0281, -0.0244, ..., -0.0480, -0.0054, -0.0765], + [-0.0406, -0.0434, -0.0595, ..., -0.0572, -0.0191, 0.0717], + ..., + [ 0.0466, 0.0056, -0.0357, ..., 0.0412, -0.0160, -0.0086], + [ 0.0608, 0.0403, 0.0150, ..., -0.0363, -0.0501, 0.0360], + [ 0.0317, -0.0085, 0.0251, ..., 0.0308, -0.0440, 0.0245]], + device='cuda:0'), grad: tensor([[ 5.2786e-04, 1.5259e-04, 4.0817e-04, ..., 1.5805e-06, + 2.9802e-05, 6.6376e-04], + [ 1.2249e-05, -8.8215e-06, 1.2770e-05, ..., 1.0133e-05, + 1.7196e-05, 3.0667e-05], + [ 1.2398e-04, 3.5316e-05, 1.1307e-04, ..., 3.3956e-06, + 9.7990e-05, 2.9159e-04], + ..., + [-2.8953e-05, 8.1062e-06, 1.0274e-05, ..., -1.4877e-04, + -4.2245e-06, 1.7568e-05], + [-7.8678e-04, -2.1291e-04, -5.9700e-04, ..., 6.5602e-06, + 3.7670e-05, -8.7786e-04], + [ 3.6240e-05, -5.8487e-06, 2.1458e-05, ..., -3.0687e-07, + -1.4054e-06, 4.1991e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0287, -0.0103, -0.0115, 0.0186, -0.0045, -0.0151, -0.0041, 0.0107, + -0.0248, 0.0162], device='cuda:0'), grad: tensor([ 1.4935e-03, 4.0919e-05, 5.3787e-04, 8.6010e-05, 2.4307e-04, + 4.4703e-05, -3.8290e-04, -8.3864e-05, -2.0943e-03, 1.1146e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 219.60, cls_loss 0.0044 cls_loss_mapping 0.0140 cls_loss_causal 0.6236 re_mapping 0.0097 re_causal 0.0296 /// teacc 98.83 lr 0.00010000 +Epoch 66, weight, value: tensor([[ 0.0025, -0.0245, 0.0110, ..., -0.0591, -0.0368, -0.0190], + [-0.0547, 0.0277, -0.0255, ..., -0.0485, -0.0062, -0.0779], + [-0.0410, -0.0440, -0.0600, ..., -0.0578, -0.0195, 0.0721], + ..., + [ 0.0469, 0.0053, -0.0361, ..., 0.0413, -0.0167, -0.0092], + [ 0.0616, 0.0409, 0.0155, ..., -0.0369, -0.0507, 0.0367], + [ 0.0323, -0.0083, 0.0257, ..., 0.0310, -0.0444, 0.0246]], + device='cuda:0'), grad: tensor([[-2.7150e-05, 2.3842e-06, -1.1273e-05, ..., 2.2173e-05, + 1.7047e-05, 1.4231e-05], + [ 2.4885e-06, 1.0934e-06, 4.2059e-06, ..., 1.5363e-05, + 1.2845e-05, 1.1168e-05], + [ 8.8587e-06, 9.2313e-06, 6.7055e-06, ..., 1.2577e-05, + 1.3210e-05, 1.0200e-05], + ..., + [ 5.5134e-07, 1.4201e-05, 9.4026e-06, ..., 6.9439e-06, + 5.0254e-06, 4.9546e-06], + [ 1.6466e-05, 2.1905e-05, 1.5944e-05, ..., 1.9699e-05, + 5.8934e-06, 5.4762e-06], + [-3.2276e-05, -3.2693e-05, -2.1607e-05, ..., 1.5989e-05, + 3.8266e-05, 2.4468e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0287, -0.0111, -0.0116, 0.0186, -0.0041, -0.0149, -0.0041, 0.0107, + -0.0245, 0.0163], device='cuda:0'), grad: tensor([-1.8239e-05, 2.5839e-05, 5.0336e-05, -3.9816e-05, -9.8407e-05, + 2.8610e-05, -4.5240e-05, 2.7344e-05, 6.1393e-05, 7.7710e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 219.32, cls_loss 0.0047 cls_loss_mapping 0.0162 cls_loss_causal 0.6285 re_mapping 0.0098 re_causal 0.0291 /// teacc 99.02 lr 0.00010000 +Epoch 67, weight, value: tensor([[ 0.0028, -0.0239, 0.0113, ..., -0.0593, -0.0372, -0.0193], + [-0.0550, 0.0278, -0.0260, ..., -0.0492, -0.0068, -0.0786], + [-0.0414, -0.0447, -0.0603, ..., -0.0585, -0.0198, 0.0725], + ..., + [ 0.0477, 0.0052, -0.0366, ..., 0.0419, -0.0170, -0.0095], + [ 0.0621, 0.0408, 0.0156, ..., -0.0372, -0.0522, 0.0369], + [ 0.0321, -0.0081, 0.0259, ..., 0.0310, -0.0449, 0.0243]], + device='cuda:0'), grad: tensor([[ 6.4135e-05, 8.0317e-06, 3.7588e-06, ..., 1.3798e-05, + 1.1399e-05, 3.3200e-05], + [ 1.6689e-04, -2.2769e-05, 6.6198e-06, ..., 1.1072e-05, + 6.1765e-06, 4.3511e-05], + [-6.4278e-04, 2.5362e-05, 2.1100e-05, ..., -3.8475e-05, + 1.1414e-05, -1.6510e-04], + ..., + [ 3.2926e-04, 2.5034e-05, 1.3463e-05, ..., 2.1443e-05, + 1.2271e-05, 1.0073e-04], + [ 1.9535e-05, 1.1250e-06, -1.8731e-05, ..., 3.0607e-05, + 1.7747e-05, -3.1501e-05], + [-2.3395e-05, -3.1590e-05, -2.9549e-05, ..., -6.9380e-05, + -1.2822e-05, -2.3857e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0293, -0.0113, -0.0118, 0.0185, -0.0040, -0.0149, -0.0037, 0.0111, + -0.0251, 0.0161], device='cuda:0'), grad: tensor([ 1.7786e-04, 4.5705e-04, -1.8425e-03, -2.1413e-05, 6.7139e-04, + -1.1730e-03, 5.6744e-04, 1.1110e-03, 5.5283e-05, -5.0366e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 219.60, cls_loss 0.0044 cls_loss_mapping 0.0139 cls_loss_causal 0.6213 re_mapping 0.0092 re_causal 0.0285 /// teacc 98.93 lr 0.00010000 +Epoch 68, weight, value: tensor([[ 0.0029, -0.0244, 0.0112, ..., -0.0597, -0.0379, -0.0201], + [-0.0558, 0.0276, -0.0264, ..., -0.0496, -0.0074, -0.0790], + [-0.0417, -0.0455, -0.0610, ..., -0.0590, -0.0202, 0.0731], + ..., + [ 0.0486, 0.0055, -0.0371, ..., 0.0421, -0.0175, -0.0101], + [ 0.0624, 0.0410, 0.0159, ..., -0.0377, -0.0525, 0.0373], + [ 0.0322, -0.0081, 0.0262, ..., 0.0310, -0.0454, 0.0240]], + device='cuda:0'), grad: tensor([[-1.0738e-06, 1.3504e-06, -5.0012e-07, ..., 1.0924e-06, + 2.6543e-06, 3.8296e-06], + [ 4.1798e-06, 8.7917e-06, 3.9637e-06, ..., 2.5053e-06, + 3.3267e-06, 1.4007e-05], + [-3.9511e-07, 5.8822e-06, 4.4480e-06, ..., 1.2130e-05, + 3.8669e-06, -4.3422e-05], + ..., + [-1.3620e-05, 1.8343e-05, 5.1931e-06, ..., -1.4015e-05, + 1.9576e-06, -5.0925e-06], + [ 5.8822e-06, 5.7220e-05, 1.6913e-05, ..., 5.4128e-06, + 1.1072e-05, 3.3170e-05], + [-1.1101e-05, 6.3665e-06, -6.5714e-06, ..., -7.1637e-06, + 4.7535e-06, 1.8803e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0288, -0.0117, -0.0121, 0.0184, -0.0038, -0.0141, -0.0035, 0.0115, + -0.0255, 0.0160], device='cuda:0'), grad: tensor([ 3.3136e-06, 3.0488e-05, -1.8820e-05, -7.0000e-04, 4.0568e-06, + 5.6982e-04, -1.6719e-05, -5.8152e-06, 1.2064e-04, 1.1764e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 219.57, cls_loss 0.0038 cls_loss_mapping 0.0133 cls_loss_causal 0.6270 re_mapping 0.0095 re_causal 0.0294 /// teacc 99.05 lr 0.00010000 +Epoch 69, weight, value: tensor([[ 0.0032, -0.0252, 0.0114, ..., -0.0598, -0.0388, -0.0210], + [-0.0566, 0.0277, -0.0267, ..., -0.0502, -0.0078, -0.0795], + [-0.0418, -0.0461, -0.0610, ..., -0.0598, -0.0205, 0.0741], + ..., + [ 0.0482, 0.0051, -0.0379, ..., 0.0422, -0.0178, -0.0112], + [ 0.0642, 0.0417, 0.0164, ..., -0.0370, -0.0529, 0.0381], + [ 0.0324, -0.0078, 0.0264, ..., 0.0314, -0.0454, 0.0240]], + device='cuda:0'), grad: tensor([[ 1.4463e-06, 3.5036e-06, 2.5099e-07, ..., 2.0433e-06, + 3.1628e-06, 1.1444e-05], + [ 4.6156e-06, -1.6708e-06, 3.3434e-06, ..., 1.2368e-06, + 2.5500e-06, 6.4149e-06], + [ 8.1910e-07, 9.7081e-06, 6.1952e-06, ..., 5.0440e-06, + 9.4026e-06, -2.1264e-05], + ..., + [ 2.0154e-06, 6.6198e-06, 6.0648e-06, ..., -2.4512e-06, + 1.5842e-06, 6.2846e-06], + [ 9.5814e-06, 5.4650e-06, 8.6948e-06, ..., 1.3597e-05, + 1.9614e-06, -7.5661e-06], + [-5.5075e-05, -3.5763e-05, -4.3184e-05, ..., -2.4632e-05, + 2.0750e-06, -3.9116e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0287, -0.0119, -0.0117, 0.0184, -0.0041, -0.0143, -0.0035, 0.0114, + -0.0248, 0.0160], device='cuda:0'), grad: tensor([ 1.7449e-05, 9.7454e-06, -2.7820e-05, 1.3910e-05, 1.2875e-05, + 2.9787e-05, -1.2539e-05, 1.2495e-05, 2.1741e-05, -7.7665e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 219.68, cls_loss 0.0044 cls_loss_mapping 0.0140 cls_loss_causal 0.6168 re_mapping 0.0086 re_causal 0.0267 /// teacc 98.94 lr 0.00010000 +Epoch 70, weight, value: tensor([[ 0.0038, -0.0257, 0.0117, ..., -0.0599, -0.0398, -0.0215], + [-0.0568, 0.0280, -0.0270, ..., -0.0505, -0.0084, -0.0800], + [-0.0419, -0.0465, -0.0612, ..., -0.0603, -0.0212, 0.0748], + ..., + [ 0.0489, 0.0050, -0.0382, ..., 0.0427, -0.0182, -0.0117], + [ 0.0645, 0.0419, 0.0165, ..., -0.0377, -0.0531, 0.0382], + [ 0.0326, -0.0074, 0.0271, ..., 0.0317, -0.0453, 0.0241]], + device='cuda:0'), grad: tensor([[-1.2638e-06, 3.4813e-06, -4.6343e-06, ..., 3.8594e-06, + 7.5549e-06, 1.6078e-05], + [ 1.9759e-05, -4.4182e-06, 2.1104e-06, ..., 8.4490e-06, + 9.2760e-06, 3.5226e-05], + [ 4.3064e-05, 6.2324e-06, 1.1027e-06, ..., 7.1973e-06, + 2.2128e-05, 6.1870e-05], + ..., + [ 2.3067e-05, 2.8014e-05, 1.1139e-05, ..., 4.7922e-05, + 1.6525e-05, 3.3885e-05], + [-2.3198e-04, 5.8673e-07, -1.7211e-06, ..., 1.7643e-05, + 1.5132e-05, -3.2377e-04], + [-1.4812e-05, -6.3442e-06, -2.1413e-05, ..., 1.5199e-06, + 1.8418e-05, 2.2173e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0289, -0.0116, -0.0114, 0.0183, -0.0045, -0.0147, -0.0036, 0.0117, + -0.0252, 0.0160], device='cuda:0'), grad: tensor([ 1.4409e-05, 6.3717e-05, 1.3959e-04, 2.9013e-05, -6.5804e-05, + 3.8004e-04, 4.9442e-05, 1.2088e-04, -7.5674e-04, 2.5705e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 219.44, cls_loss 0.0047 cls_loss_mapping 0.0168 cls_loss_causal 0.6399 re_mapping 0.0087 re_causal 0.0274 /// teacc 98.87 lr 0.00010000 +Epoch 71, weight, value: tensor([[ 0.0039, -0.0262, 0.0118, ..., -0.0608, -0.0404, -0.0218], + [-0.0572, 0.0282, -0.0274, ..., -0.0508, -0.0092, -0.0808], + [-0.0421, -0.0471, -0.0614, ..., -0.0613, -0.0217, 0.0753], + ..., + [ 0.0494, 0.0048, -0.0383, ..., 0.0431, -0.0188, -0.0122], + [ 0.0647, 0.0421, 0.0165, ..., -0.0384, -0.0535, 0.0385], + [ 0.0327, -0.0080, 0.0269, ..., 0.0308, -0.0462, 0.0239]], + device='cuda:0'), grad: tensor([[ 7.7020e-07, 2.6882e-05, 1.0863e-05, ..., 2.6985e-07, + 3.6005e-06, 7.0222e-06], + [ 4.7795e-06, 1.1690e-05, 6.2659e-06, ..., 1.2871e-06, + 1.6866e-06, 1.4238e-05], + [ 5.2042e-06, 2.4945e-05, 1.1504e-05, ..., 8.2003e-07, + 1.5385e-06, 3.9376e-06], + ..., + [ 2.9616e-06, 1.9684e-05, 7.7039e-06, ..., 2.0973e-06, + 1.8692e-06, 1.1027e-05], + [-7.9870e-05, -9.3937e-05, -7.4744e-05, ..., -1.1809e-06, + 2.3632e-07, -2.9516e-04], + [ 6.0610e-06, 2.2769e-05, 9.8497e-06, ..., 1.8319e-06, + 2.6170e-06, 1.1258e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0289, -0.0117, -0.0112, 0.0181, -0.0030, -0.0143, -0.0039, 0.0118, + -0.0256, 0.0150], device='cuda:0'), grad: tensor([ 4.8578e-05, 3.8743e-05, 3.1203e-05, -1.4269e-04, 8.7768e-06, + 3.2473e-04, 1.8024e-04, 4.7177e-05, -5.8222e-04, 4.4882e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 70---------------------------------------------------- +epoch 70, time 220.59, cls_loss 0.0036 cls_loss_mapping 0.0107 cls_loss_causal 0.6075 re_mapping 0.0085 re_causal 0.0264 /// teacc 99.10 lr 0.00010000 +Epoch 72, weight, value: tensor([[ 0.0040, -0.0266, 0.0117, ..., -0.0618, -0.0410, -0.0224], + [-0.0575, 0.0282, -0.0277, ..., -0.0511, -0.0099, -0.0818], + [-0.0425, -0.0483, -0.0617, ..., -0.0620, -0.0219, 0.0760], + ..., + [ 0.0499, 0.0046, -0.0386, ..., 0.0434, -0.0193, -0.0130], + [ 0.0651, 0.0427, 0.0170, ..., -0.0389, -0.0537, 0.0392], + [ 0.0331, -0.0079, 0.0275, ..., 0.0311, -0.0466, 0.0240]], + device='cuda:0'), grad: tensor([[ 1.6764e-06, 4.3698e-06, 1.3877e-07, ..., 3.9674e-06, + 1.7779e-06, 4.6156e-06], + [ 4.4912e-05, 2.1115e-05, 1.3135e-05, ..., 8.4639e-06, + 3.2261e-06, 5.3942e-05], + [ 4.1664e-05, 4.5657e-05, 1.1906e-05, ..., 2.2858e-05, + 4.7646e-06, 5.4508e-05], + ..., + [ 6.8486e-05, 3.5435e-05, 2.4781e-05, ..., 2.1234e-05, + 3.0063e-06, 1.1519e-05], + [-1.1486e-04, -9.6977e-05, -3.5763e-05, ..., 5.2489e-06, + 2.7642e-06, -1.4842e-04], + [-8.7678e-05, -1.1936e-05, -2.1040e-05, ..., -9.9242e-05, + 4.4219e-06, 1.6779e-05]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0284, -0.0120, -0.0113, 0.0182, -0.0029, -0.0146, -0.0040, 0.0118, + -0.0251, 0.0153], device='cuda:0'), grad: tensor([ 1.1958e-05, 9.9778e-05, 1.7738e-04, 6.0610e-06, 1.7330e-05, + 5.3979e-06, 1.8179e-05, 5.6326e-05, -3.0637e-04, -8.6546e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 220.05, cls_loss 0.0037 cls_loss_mapping 0.0120 cls_loss_causal 0.6089 re_mapping 0.0089 re_causal 0.0270 /// teacc 99.01 lr 0.00010000 +Epoch 73, weight, value: tensor([[ 0.0038, -0.0271, 0.0117, ..., -0.0623, -0.0417, -0.0233], + [-0.0578, 0.0284, -0.0280, ..., -0.0513, -0.0106, -0.0830], + [-0.0429, -0.0486, -0.0620, ..., -0.0625, -0.0222, 0.0770], + ..., + [ 0.0502, 0.0044, -0.0389, ..., 0.0437, -0.0199, -0.0139], + [ 0.0660, 0.0432, 0.0174, ..., -0.0393, -0.0538, 0.0401], + [ 0.0335, -0.0077, 0.0278, ..., 0.0314, -0.0467, 0.0238]], + device='cuda:0'), grad: tensor([[ 2.4531e-06, 4.8280e-05, 1.9103e-05, ..., 2.0847e-05, + 9.3654e-06, 1.3836e-05], + [ 6.6087e-06, -4.2990e-06, 3.2317e-06, ..., 1.3269e-05, + 1.2554e-06, 3.8445e-06], + [ 3.0339e-05, 2.4945e-05, 1.0476e-05, ..., 8.1241e-05, + 3.8259e-06, 2.9832e-05], + ..., + [-1.7083e-04, -3.5822e-05, 1.0490e-05, ..., -3.2735e-04, + -4.2804e-06, -4.5955e-05], + [ 1.9968e-06, 4.6074e-05, 1.6078e-05, ..., 3.4362e-05, + 2.2203e-06, 4.9174e-07], + [ 2.6911e-05, 8.4162e-05, 1.8805e-05, ..., 7.3493e-05, + -1.8976e-08, 4.5039e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0280, -0.0123, -0.0108, 0.0177, -0.0031, -0.0145, -0.0039, 0.0119, + -0.0246, 0.0153], device='cuda:0'), grad: tensor([ 1.0461e-04, 8.6874e-06, 1.5450e-04, -2.9540e-04, 7.0393e-05, + 1.4484e-04, -3.2604e-05, -4.8876e-04, 1.0550e-04, 2.2829e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 219.55, cls_loss 0.0054 cls_loss_mapping 0.0147 cls_loss_causal 0.6212 re_mapping 0.0088 re_causal 0.0268 /// teacc 98.95 lr 0.00010000 +Epoch 74, weight, value: tensor([[ 0.0035, -0.0279, 0.0117, ..., -0.0641, -0.0423, -0.0240], + [-0.0580, 0.0289, -0.0282, ..., -0.0516, -0.0113, -0.0836], + [-0.0434, -0.0499, -0.0623, ..., -0.0635, -0.0232, 0.0774], + ..., + [ 0.0505, 0.0042, -0.0392, ..., 0.0438, -0.0206, -0.0151], + [ 0.0669, 0.0433, 0.0178, ..., -0.0391, -0.0540, 0.0408], + [ 0.0350, -0.0068, 0.0288, ..., 0.0331, -0.0460, 0.0250]], + device='cuda:0'), grad: tensor([[ 7.6294e-06, 1.2554e-05, 1.0945e-05, ..., 2.0102e-05, + 4.9621e-06, 1.2189e-05], + [ 4.6380e-06, -7.4446e-05, 4.1351e-06, ..., 3.3882e-06, + 1.2800e-05, 5.6028e-06], + [-1.2493e-04, 5.4002e-05, 3.9265e-06, ..., 3.8370e-06, + 2.9989e-06, -1.4687e-04], + ..., + [ 1.3161e-04, 1.0975e-05, 3.9861e-06, ..., 6.7651e-06, + 7.4366e-07, 1.5724e-04], + [-1.3337e-05, -1.9014e-05, -1.5557e-05, ..., 8.9779e-06, + 3.5875e-06, -1.2919e-05], + [-3.7283e-05, -3.2574e-05, -3.5435e-05, ..., -6.9857e-05, + -1.7062e-06, -2.7984e-05]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0277, -0.0123, -0.0110, 0.0185, -0.0046, -0.0149, -0.0036, 0.0117, + -0.0245, 0.0166], device='cuda:0'), grad: tensor([ 1.0002e-04, -2.1625e-04, -6.9380e-04, 1.8084e-04, 5.8860e-05, + 1.0419e-04, -3.0696e-05, 5.7888e-04, 2.6137e-05, -1.0949e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 219.16, cls_loss 0.0044 cls_loss_mapping 0.0128 cls_loss_causal 0.5954 re_mapping 0.0087 re_causal 0.0268 /// teacc 99.02 lr 0.00010000 +Epoch 75, weight, value: tensor([[ 0.0049, -0.0278, 0.0123, ..., -0.0643, -0.0429, -0.0245], + [-0.0587, 0.0297, -0.0284, ..., -0.0520, -0.0124, -0.0847], + [-0.0439, -0.0508, -0.0628, ..., -0.0639, -0.0238, 0.0777], + ..., + [ 0.0507, 0.0032, -0.0394, ..., 0.0435, -0.0215, -0.0159], + [ 0.0678, 0.0439, 0.0182, ..., -0.0392, -0.0543, 0.0421], + [ 0.0350, -0.0073, 0.0288, ..., 0.0325, -0.0472, 0.0242]], + device='cuda:0'), grad: tensor([[-2.4185e-05, 1.8328e-05, -7.2084e-06, ..., 4.5039e-06, + 2.0284e-06, 4.8019e-06], + [ 2.4009e-06, -4.0717e-06, 8.1658e-06, ..., 1.5134e-06, + 3.7588e-06, 4.5225e-06], + [ 1.1943e-05, 5.8889e-05, 2.3469e-05, ..., 3.8236e-05, + 1.8075e-05, 4.0643e-06], + ..., + [-9.9167e-06, 8.3745e-05, 3.6061e-05, ..., -7.0445e-06, + 3.8967e-06, 7.8306e-06], + [ 1.0170e-05, 9.9301e-05, 4.4405e-05, ..., 1.1988e-05, + 2.1793e-06, 8.9854e-06], + [-5.4128e-06, 3.0011e-05, 9.1940e-06, ..., -4.2617e-06, + 4.0382e-06, 3.8706e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0284, -0.0121, -0.0113, 0.0185, -0.0036, -0.0155, -0.0036, 0.0112, + -0.0237, 0.0157], device='cuda:0'), grad: tensor([-7.0691e-05, -1.8373e-05, 1.3590e-04, -4.7493e-04, -3.9399e-05, + 2.9847e-05, 9.2030e-05, 1.0538e-04, 1.7905e-04, 6.0827e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 218.76, cls_loss 0.0036 cls_loss_mapping 0.0111 cls_loss_causal 0.6157 re_mapping 0.0084 re_causal 0.0260 /// teacc 98.89 lr 0.00010000 +Epoch 76, weight, value: tensor([[ 0.0051, -0.0283, 0.0124, ..., -0.0645, -0.0436, -0.0250], + [-0.0591, 0.0299, -0.0288, ..., -0.0524, -0.0131, -0.0852], + [-0.0447, -0.0514, -0.0633, ..., -0.0646, -0.0244, 0.0782], + ..., + [ 0.0512, 0.0031, -0.0397, ..., 0.0439, -0.0218, -0.0164], + [ 0.0682, 0.0443, 0.0186, ..., -0.0398, -0.0547, 0.0424], + [ 0.0356, -0.0071, 0.0289, ..., 0.0330, -0.0472, 0.0241]], + device='cuda:0'), grad: tensor([[-1.4575e-06, 1.6093e-05, 4.1798e-06, ..., 8.5086e-06, + 4.3698e-06, 8.5831e-06], + [ 1.1779e-05, 1.9655e-05, 1.2666e-05, ..., 3.0756e-05, + 1.9267e-05, 2.7061e-05], + [ 2.2560e-05, 5.3048e-05, 2.3544e-05, ..., 5.4866e-05, + 2.7633e-04, 2.9850e-04], + ..., + [-2.3231e-05, 1.8124e-06, 1.2964e-05, ..., -7.6592e-05, + -6.3404e-06, -5.6699e-06], + [ 3.4273e-05, 8.8573e-05, 5.4061e-05, ..., 4.9978e-05, + 1.1802e-05, 4.4376e-05], + [-1.5855e-04, -3.9554e-04, -2.5845e-04, ..., -1.3244e-04, + 3.3319e-05, -7.2896e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0284, -0.0119, -0.0113, 0.0183, -0.0038, -0.0156, -0.0038, 0.0115, + -0.0239, 0.0159], device='cuda:0'), grad: tensor([ 1.7241e-05, 8.3506e-05, 6.2561e-04, 3.4666e-04, 1.3065e-04, + 5.5075e-05, -6.2561e-04, -9.9480e-05, 2.2149e-04, -7.5483e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 75---------------------------------------------------- +epoch 75, time 219.24, cls_loss 0.0029 cls_loss_mapping 0.0098 cls_loss_causal 0.5937 re_mapping 0.0088 re_causal 0.0263 /// teacc 99.11 lr 0.00010000 +Epoch 77, weight, value: tensor([[ 0.0053, -0.0289, 0.0125, ..., -0.0647, -0.0438, -0.0254], + [-0.0592, 0.0302, -0.0292, ..., -0.0526, -0.0140, -0.0859], + [-0.0450, -0.0519, -0.0636, ..., -0.0654, -0.0249, 0.0786], + ..., + [ 0.0512, 0.0028, -0.0401, ..., 0.0441, -0.0224, -0.0168], + [ 0.0689, 0.0450, 0.0191, ..., -0.0402, -0.0551, 0.0428], + [ 0.0360, -0.0069, 0.0292, ..., 0.0330, -0.0476, 0.0239]], + device='cuda:0'), grad: tensor([[-6.9559e-05, 2.9299e-06, -1.1884e-05, ..., 2.0396e-06, + 4.7758e-06, 3.4496e-06], + [ 5.8636e-06, -5.1633e-06, 9.8161e-07, ..., -1.4121e-07, + 2.4885e-06, 1.6168e-06], + [ 6.1393e-06, 5.3868e-06, 7.5018e-07, ..., 5.9567e-06, + 1.6754e-06, -2.3581e-06], + ..., + [-3.2663e-05, -2.8592e-06, 2.8983e-06, ..., -2.4214e-05, + 3.3360e-06, 1.3839e-06], + [ 1.2405e-05, 7.3984e-06, 6.4913e-07, ..., 1.2815e-05, + 5.1260e-06, 2.3860e-06], + [-6.8583e-06, 7.0333e-06, -5.9530e-06, ..., -3.8631e-06, + 3.6173e-06, -8.2934e-07]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0285, -0.0119, -0.0114, 0.0178, -0.0038, -0.0152, -0.0037, 0.0113, + -0.0236, 0.0159], device='cuda:0'), grad: tensor([-4.5657e-04, 1.2452e-06, 1.5736e-05, -2.6241e-05, 1.5177e-05, + 2.8062e-04, 1.4973e-04, -5.3078e-05, 4.6164e-05, 2.6599e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 218.90, cls_loss 0.0036 cls_loss_mapping 0.0101 cls_loss_causal 0.5948 re_mapping 0.0079 re_causal 0.0242 /// teacc 98.99 lr 0.00010000 +Epoch 78, weight, value: tensor([[ 0.0059, -0.0294, 0.0126, ..., -0.0649, -0.0446, -0.0259], + [-0.0594, 0.0304, -0.0295, ..., -0.0529, -0.0150, -0.0869], + [-0.0452, -0.0524, -0.0638, ..., -0.0660, -0.0255, 0.0792], + ..., + [ 0.0513, 0.0023, -0.0407, ..., 0.0445, -0.0225, -0.0175], + [ 0.0692, 0.0448, 0.0191, ..., -0.0407, -0.0558, 0.0429], + [ 0.0363, -0.0068, 0.0295, ..., 0.0332, -0.0479, 0.0238]], + device='cuda:0'), grad: tensor([[-8.6566e-07, 5.5954e-06, 3.2842e-05, ..., 2.6859e-06, + 1.9610e-05, 3.3289e-05], + [ 2.4140e-06, -9.9480e-05, 7.3053e-06, ..., 2.3529e-05, + 1.9118e-05, 1.5035e-05], + [ 2.8964e-06, 9.2983e-05, 7.2494e-06, ..., 7.4022e-06, + 5.7854e-06, 2.9206e-06], + ..., + [-7.0557e-06, 3.0547e-05, 4.5747e-06, ..., 6.1281e-07, + 4.7311e-06, 4.7795e-06], + [ 3.7272e-06, 2.9221e-05, 1.1571e-05, ..., 2.7522e-05, + 1.5825e-05, 1.5765e-05], + [ 1.1642e-06, 3.8445e-05, 2.1845e-05, ..., 7.0870e-05, + 4.7743e-05, 4.3452e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0287, -0.0122, -0.0107, 0.0181, -0.0043, -0.0153, -0.0029, 0.0113, + -0.0242, 0.0158], device='cuda:0'), grad: tensor([ 7.3493e-05, -1.4889e-04, 1.7178e-04, -1.4031e-04, -2.6345e-04, + -3.7611e-05, 4.1842e-05, 4.8548e-05, 8.1599e-05, 1.7297e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 218.49, cls_loss 0.0028 cls_loss_mapping 0.0102 cls_loss_causal 0.5778 re_mapping 0.0081 re_causal 0.0251 /// teacc 98.95 lr 0.00010000 +Epoch 79, weight, value: tensor([[ 0.0064, -0.0297, 0.0132, ..., -0.0650, -0.0450, -0.0259], + [-0.0597, 0.0305, -0.0298, ..., -0.0532, -0.0163, -0.0878], + [-0.0458, -0.0532, -0.0643, ..., -0.0668, -0.0253, 0.0797], + ..., + [ 0.0519, 0.0021, -0.0409, ..., 0.0450, -0.0228, -0.0177], + [ 0.0693, 0.0447, 0.0191, ..., -0.0412, -0.0562, 0.0431], + [ 0.0365, -0.0066, 0.0297, ..., 0.0334, -0.0479, 0.0238]], + device='cuda:0'), grad: tensor([[ 1.2228e-06, 2.6375e-06, 2.1625e-06, ..., 3.0361e-06, + 5.7742e-07, 2.0415e-06], + [ 1.9264e-04, 7.2241e-05, 1.4743e-06, ..., 8.0168e-05, + 1.2945e-06, 1.7965e-06], + [ 2.6911e-05, 2.8849e-05, 1.7798e-06, ..., 1.3985e-05, + 9.6299e-07, -3.9712e-06], + ..., + [-2.2411e-04, -9.4116e-05, 3.0417e-06, ..., -9.1076e-05, + 1.3225e-06, 2.6785e-06], + [ 2.8219e-07, 2.1085e-06, -1.3914e-06, ..., 3.2485e-06, + 4.2142e-07, -2.3153e-06], + [-6.9886e-06, 2.4540e-07, -6.1989e-06, ..., -5.4613e-06, + 1.2703e-06, 8.8650e-08]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0293, -0.0126, -0.0109, 0.0186, -0.0045, -0.0152, -0.0031, 0.0118, + -0.0247, 0.0158], device='cuda:0'), grad: tensor([ 1.1131e-05, 1.2436e-03, 2.3925e-04, -1.0371e-05, -1.0384e-06, + 1.5512e-05, 8.5086e-06, -1.5259e-03, 1.7971e-05, 4.1514e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 218.70, cls_loss 0.0028 cls_loss_mapping 0.0098 cls_loss_causal 0.5869 re_mapping 0.0084 re_causal 0.0251 /// teacc 98.88 lr 0.00010000 +Epoch 80, weight, value: tensor([[ 0.0068, -0.0302, 0.0134, ..., -0.0653, -0.0457, -0.0264], + [-0.0602, 0.0305, -0.0301, ..., -0.0535, -0.0168, -0.0888], + [-0.0462, -0.0536, -0.0646, ..., -0.0676, -0.0259, 0.0801], + ..., + [ 0.0526, 0.0019, -0.0412, ..., 0.0453, -0.0233, -0.0182], + [ 0.0693, 0.0447, 0.0191, ..., -0.0421, -0.0561, 0.0433], + [ 0.0368, -0.0064, 0.0303, ..., 0.0333, -0.0486, 0.0236]], + device='cuda:0'), grad: tensor([[-1.2958e-04, 4.8466e-06, 8.5728e-07, ..., 7.1600e-06, + 2.6003e-06, 4.0680e-06], + [ 9.6038e-06, -1.3299e-06, 4.0345e-06, ..., 1.3828e-05, + 6.3106e-06, 1.0774e-05], + [ 1.6177e-06, 8.1286e-06, 4.5411e-06, ..., 9.8944e-06, + 4.8280e-06, -2.0519e-05], + ..., + [ 1.3041e-04, 5.3704e-05, 4.8488e-05, ..., 1.4937e-04, + 1.4201e-05, 3.0011e-05], + [-1.6645e-05, -1.3739e-05, -9.6560e-06, ..., 9.6783e-06, + 4.2468e-06, -1.9729e-05], + [-1.8096e-04, -1.8585e-04, -1.4675e-04, ..., -3.6764e-04, + -4.8220e-05, -5.4836e-05]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0292, -0.0129, -0.0110, 0.0181, -0.0040, -0.0148, -0.0034, 0.0122, + -0.0250, 0.0157], device='cuda:0'), grad: tensor([-2.1136e-04, 2.7612e-05, -6.6102e-05, 6.2108e-05, 2.4867e-04, + 7.9155e-05, 1.2660e-04, 3.1948e-04, -5.9716e-06, -5.8031e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 218.62, cls_loss 0.0038 cls_loss_mapping 0.0121 cls_loss_causal 0.6005 re_mapping 0.0082 re_causal 0.0250 /// teacc 98.90 lr 0.00010000 +Epoch 81, weight, value: tensor([[ 0.0073, -0.0303, 0.0137, ..., -0.0658, -0.0461, -0.0270], + [-0.0608, 0.0309, -0.0308, ..., -0.0536, -0.0175, -0.0895], + [-0.0465, -0.0541, -0.0648, ..., -0.0680, -0.0265, 0.0807], + ..., + [ 0.0525, 0.0013, -0.0420, ..., 0.0452, -0.0236, -0.0189], + [ 0.0702, 0.0454, 0.0197, ..., -0.0421, -0.0564, 0.0438], + [ 0.0372, -0.0062, 0.0307, ..., 0.0335, -0.0488, 0.0234]], + device='cuda:0'), grad: tensor([[-5.5432e-06, -3.3565e-06, -1.6382e-06, ..., 6.6916e-07, + 1.7792e-05, 1.5691e-05], + [ 1.4883e-06, -6.6236e-06, 2.1197e-06, ..., 3.8333e-06, + 6.2659e-06, 9.9391e-06], + [ 4.7334e-07, 1.7397e-06, 1.6512e-06, ..., 1.2480e-06, + 5.0440e-06, -3.3796e-05], + ..., + [-9.7975e-07, 3.9712e-06, 1.9372e-06, ..., 1.1018e-06, + 2.4065e-06, 3.5707e-06], + [ 2.1644e-06, 6.1169e-06, 8.8140e-06, ..., 2.8890e-06, + 1.9133e-05, 1.8626e-05], + [-4.3660e-06, -2.7902e-06, -2.8536e-06, ..., -6.2697e-06, + 4.6566e-06, 2.1514e-06]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0299, -0.0133, -0.0110, 0.0192, -0.0041, -0.0142, -0.0044, 0.0118, + -0.0251, 0.0157], device='cuda:0'), grad: tensor([ 1.8492e-05, 1.9699e-05, -1.3912e-04, 5.2810e-05, 2.1026e-05, + 2.3946e-05, -7.2122e-05, 1.7032e-05, 5.4777e-05, 3.4515e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 218.67, cls_loss 0.0030 cls_loss_mapping 0.0088 cls_loss_causal 0.5878 re_mapping 0.0082 re_causal 0.0248 /// teacc 99.05 lr 0.00010000 +Epoch 82, weight, value: tensor([[ 0.0073, -0.0307, 0.0138, ..., -0.0660, -0.0467, -0.0276], + [-0.0611, 0.0313, -0.0311, ..., -0.0539, -0.0182, -0.0901], + [-0.0465, -0.0551, -0.0652, ..., -0.0686, -0.0270, 0.0818], + ..., + [ 0.0528, 0.0011, -0.0423, ..., 0.0454, -0.0242, -0.0197], + [ 0.0706, 0.0457, 0.0200, ..., -0.0425, -0.0567, 0.0440], + [ 0.0376, -0.0060, 0.0312, ..., 0.0337, -0.0489, 0.0233]], + device='cuda:0'), grad: tensor([[ 1.1370e-05, 1.4035e-06, -3.1702e-06, ..., 3.8743e-06, + 3.0193e-06, 9.6709e-06], + [ 2.5984e-06, -5.1502e-07, 7.4599e-07, ..., 1.8878e-06, + 3.6266e-06, 3.1944e-06], + [-6.5029e-05, 2.2277e-06, 6.3516e-07, ..., -1.3843e-05, + 2.1365e-06, -3.3289e-05], + ..., + [ 3.7700e-05, 3.3174e-06, 1.0198e-06, ..., 9.0897e-06, + 9.8441e-07, 2.0161e-05], + [ 3.1162e-06, 1.3612e-05, 1.6494e-06, ..., 1.6280e-06, + 1.9781e-06, 1.7704e-06], + [-2.0899e-06, 4.2282e-06, -2.9895e-06, ..., 4.6134e-05, + 3.6836e-05, 2.2799e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0295, -0.0130, -0.0111, 0.0192, -0.0040, -0.0139, -0.0046, 0.0118, + -0.0255, 0.0157], device='cuda:0'), grad: tensor([ 2.2635e-05, 9.4473e-06, -9.5904e-05, -8.1718e-05, -3.6836e-05, + 4.6879e-05, -4.5478e-05, 6.7353e-05, 5.1290e-05, 6.2287e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 218.91, cls_loss 0.0034 cls_loss_mapping 0.0124 cls_loss_causal 0.5640 re_mapping 0.0079 re_causal 0.0235 /// teacc 99.02 lr 0.00010000 +Epoch 83, weight, value: tensor([[ 0.0073, -0.0309, 0.0137, ..., -0.0665, -0.0483, -0.0287], + [-0.0617, 0.0313, -0.0315, ..., -0.0543, -0.0189, -0.0906], + [-0.0469, -0.0556, -0.0655, ..., -0.0691, -0.0278, 0.0820], + ..., + [ 0.0531, 0.0007, -0.0426, ..., 0.0457, -0.0245, -0.0202], + [ 0.0716, 0.0465, 0.0207, ..., -0.0430, -0.0572, 0.0451], + [ 0.0382, -0.0058, 0.0314, ..., 0.0340, -0.0491, 0.0232]], + device='cuda:0'), grad: tensor([[-6.7472e-05, -4.8578e-06, -2.4498e-05, ..., -2.4885e-05, + 2.2200e-07, 1.1837e-06], + [ 2.8666e-06, 1.3839e-06, 2.1160e-06, ..., 1.5683e-06, + 3.7788e-07, 1.3625e-06], + [-4.3400e-06, 3.4291e-06, 2.1830e-06, ..., 2.1625e-06, + 4.0093e-07, -3.9458e-05], + ..., + [ 1.3616e-06, 6.3069e-06, 2.7772e-06, ..., 1.3562e-07, + 3.5716e-07, 1.6969e-06], + [ 8.5682e-06, 4.4964e-06, 1.4594e-06, ..., 1.6810e-06, + 3.4925e-07, 2.9579e-05], + [ 3.5226e-05, 2.0824e-06, 1.1168e-05, ..., 9.3728e-06, + -1.5832e-06, -1.8638e-07]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0289, -0.0132, -0.0116, 0.0192, -0.0041, -0.0145, -0.0040, 0.0120, + -0.0248, 0.0160], device='cuda:0'), grad: tensor([-1.6654e-04, 9.1046e-06, -3.6567e-05, -1.4797e-05, 1.2040e-05, + -8.3596e-06, 3.3677e-05, 1.6302e-05, 5.7191e-05, 9.7692e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 218.97, cls_loss 0.0029 cls_loss_mapping 0.0111 cls_loss_causal 0.5985 re_mapping 0.0078 re_causal 0.0242 /// teacc 98.92 lr 0.00010000 +Epoch 84, weight, value: tensor([[ 0.0077, -0.0314, 0.0141, ..., -0.0667, -0.0490, -0.0292], + [-0.0619, 0.0317, -0.0318, ..., -0.0545, -0.0196, -0.0909], + [-0.0474, -0.0560, -0.0657, ..., -0.0701, -0.0287, 0.0825], + ..., + [ 0.0535, 0.0004, -0.0430, ..., 0.0460, -0.0247, -0.0205], + [ 0.0720, 0.0461, 0.0209, ..., -0.0435, -0.0583, 0.0451], + [ 0.0383, -0.0061, 0.0313, ..., 0.0340, -0.0495, 0.0229]], + device='cuda:0'), grad: tensor([[-3.3975e-05, 6.9616e-07, -2.1577e-05, ..., 7.1526e-07, + 3.9227e-06, -1.0051e-05], + [ 3.4943e-06, 1.1735e-06, 1.8328e-06, ..., 4.9509e-06, + 4.6343e-06, 9.9093e-06], + [ 7.9423e-06, 3.0212e-06, 4.6380e-06, ..., 7.1824e-06, + 5.6922e-06, -1.3202e-05], + ..., + [-7.0930e-06, 2.2314e-06, 2.4084e-06, ..., -1.3178e-06, + 4.3996e-06, 1.3024e-05], + [ 4.3921e-06, -1.4916e-05, 3.2187e-06, ..., 2.4885e-06, + -4.1374e-07, -6.9179e-06], + [ 5.8189e-06, -1.4044e-05, -9.6783e-06, ..., -1.1958e-05, + -1.2303e-06, 1.8897e-06]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0291, -0.0129, -0.0115, 0.0194, -0.0038, -0.0142, -0.0041, 0.0121, + -0.0256, 0.0155], device='cuda:0'), grad: tensor([-8.9467e-05, 1.7822e-05, -1.1787e-05, 3.3110e-05, 8.2999e-06, + 1.3433e-05, -2.0526e-06, 8.8736e-06, 2.4468e-05, -2.8070e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 218.49, cls_loss 0.0027 cls_loss_mapping 0.0085 cls_loss_causal 0.5411 re_mapping 0.0073 re_causal 0.0221 /// teacc 99.06 lr 0.00010000 +Epoch 85, weight, value: tensor([[ 0.0079, -0.0318, 0.0142, ..., -0.0669, -0.0497, -0.0299], + [-0.0622, 0.0319, -0.0320, ..., -0.0548, -0.0200, -0.0917], + [-0.0478, -0.0568, -0.0660, ..., -0.0711, -0.0292, 0.0831], + ..., + [ 0.0538, 0.0002, -0.0434, ..., 0.0464, -0.0252, -0.0208], + [ 0.0731, 0.0467, 0.0215, ..., -0.0443, -0.0587, 0.0457], + [ 0.0383, -0.0059, 0.0315, ..., 0.0343, -0.0494, 0.0229]], + device='cuda:0'), grad: tensor([[-3.3411e-07, 5.6438e-06, 2.5257e-06, ..., 1.9372e-06, + 1.7494e-05, 2.0698e-05], + [ 2.2445e-06, -2.1353e-05, -3.3528e-07, ..., 5.8077e-06, + 1.5259e-05, 1.6928e-05], + [ 1.5255e-06, 4.2729e-06, 3.2932e-06, ..., 3.5334e-06, + 1.4842e-05, -1.0528e-05], + ..., + [-1.1615e-05, 3.5726e-06, 8.6240e-07, ..., -8.7023e-06, + 2.1625e-06, 3.4664e-06], + [ 3.3416e-06, 5.6654e-05, 7.2241e-05, ..., 6.8583e-06, + 4.1747e-04, 4.8423e-04], + [-1.6559e-06, 4.8168e-06, -3.0454e-06, ..., 3.2127e-05, + 3.2455e-05, 9.7081e-06]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0291, -0.0131, -0.0113, 0.0187, -0.0039, -0.0137, -0.0043, 0.0124, + -0.0251, 0.0154], device='cuda:0'), grad: tensor([ 3.6687e-05, -8.2776e-06, -1.5073e-05, 2.4512e-05, -3.5256e-05, + 2.1607e-06, -7.4148e-04, -1.2472e-05, 6.9952e-04, 4.9770e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 218.88, cls_loss 0.0035 cls_loss_mapping 0.0115 cls_loss_causal 0.6015 re_mapping 0.0074 re_causal 0.0230 /// teacc 99.03 lr 0.00010000 +Epoch 86, weight, value: tensor([[ 0.0088, -0.0314, 0.0146, ..., -0.0669, -0.0506, -0.0304], + [-0.0613, 0.0323, -0.0330, ..., -0.0541, -0.0209, -0.0928], + [-0.0483, -0.0577, -0.0662, ..., -0.0720, -0.0299, 0.0839], + ..., + [ 0.0536, -0.0005, -0.0438, ..., 0.0465, -0.0259, -0.0223], + [ 0.0732, 0.0476, 0.0221, ..., -0.0455, -0.0594, 0.0460], + [ 0.0385, -0.0057, 0.0317, ..., 0.0342, -0.0497, 0.0228]], + device='cuda:0'), grad: tensor([[-9.9838e-05, 1.0505e-06, -8.7917e-06, ..., 1.0412e-06, + -7.8185e-07, -9.2313e-06], + [ 1.8971e-06, -5.1856e-06, 8.6753e-07, ..., -2.2426e-06, + 4.2352e-07, 1.0030e-06], + [ 5.7966e-06, 1.7788e-06, 2.3879e-06, ..., 1.1008e-06, + 7.6089e-07, 3.3304e-06], + ..., + [ 3.1739e-06, 4.0270e-06, 1.4221e-06, ..., 3.1535e-06, + 9.0618e-07, 9.2294e-07], + [ 5.4501e-06, 2.1812e-06, 2.2203e-06, ..., 1.4119e-06, + 6.6869e-07, 2.4214e-06], + [-3.6135e-06, -1.5246e-06, -3.4235e-06, ..., -3.4124e-06, + 1.4631e-06, 2.1365e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0298, -0.0130, -0.0113, 0.0188, -0.0038, -0.0142, -0.0044, 0.0117, + -0.0245, 0.0153], device='cuda:0'), grad: tensor([-2.1040e-04, -1.4536e-05, 2.0459e-05, -2.0176e-05, -1.0189e-06, + 5.4628e-05, 1.3149e-04, 1.6421e-05, 2.0251e-05, 2.7046e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 218.59, cls_loss 0.0024 cls_loss_mapping 0.0083 cls_loss_causal 0.5561 re_mapping 0.0075 re_causal 0.0229 /// teacc 98.91 lr 0.00010000 +Epoch 87, weight, value: tensor([[ 0.0090, -0.0320, 0.0148, ..., -0.0671, -0.0513, -0.0308], + [-0.0617, 0.0324, -0.0334, ..., -0.0545, -0.0222, -0.0936], + [-0.0479, -0.0580, -0.0665, ..., -0.0718, -0.0302, 0.0851], + ..., + [ 0.0538, -0.0006, -0.0441, ..., 0.0468, -0.0262, -0.0238], + [ 0.0738, 0.0483, 0.0225, ..., -0.0459, -0.0598, 0.0466], + [ 0.0387, -0.0057, 0.0320, ..., 0.0342, -0.0502, 0.0225]], + device='cuda:0'), grad: tensor([[-1.6138e-05, 4.8950e-06, -3.5614e-06, ..., 3.6368e-07, + 1.9614e-06, 1.7941e-05], + [ 3.6154e-06, -1.1781e-07, 1.9092e-06, ..., 3.3043e-06, + 1.0906e-06, 7.5847e-06], + [ 3.0268e-06, 1.5255e-06, 1.5143e-06, ..., 3.1162e-06, + 1.3672e-06, -1.0175e-04], + ..., + [ 1.3754e-05, 1.3217e-05, 1.1876e-05, ..., 1.3970e-05, + 3.0454e-06, 1.0289e-05], + [ 1.1569e-04, 8.9943e-05, 8.0287e-05, ..., 9.9659e-05, + 1.3269e-05, 9.3460e-05], + [-1.5450e-04, -1.2720e-04, -1.1146e-04, ..., -1.3685e-04, + -1.6272e-05, -4.5508e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0300, -0.0132, -0.0107, 0.0182, -0.0036, -0.0142, -0.0040, 0.0116, + -0.0244, 0.0150], device='cuda:0'), grad: tensor([-3.9451e-06, 1.4149e-05, -1.6546e-04, 2.5228e-05, 3.0845e-05, + 1.4022e-05, 1.0878e-05, 3.5971e-05, 3.1185e-04, -2.7323e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 218.54, cls_loss 0.0029 cls_loss_mapping 0.0108 cls_loss_causal 0.5818 re_mapping 0.0074 re_causal 0.0226 /// teacc 98.84 lr 0.00010000 +Epoch 88, weight, value: tensor([[ 0.0092, -0.0324, 0.0152, ..., -0.0675, -0.0523, -0.0314], + [-0.0621, 0.0325, -0.0336, ..., -0.0549, -0.0235, -0.0947], + [-0.0484, -0.0580, -0.0669, ..., -0.0727, -0.0307, 0.0855], + ..., + [ 0.0542, -0.0009, -0.0446, ..., 0.0472, -0.0265, -0.0243], + [ 0.0746, 0.0486, 0.0231, ..., -0.0459, -0.0601, 0.0472], + [ 0.0389, -0.0055, 0.0321, ..., 0.0343, -0.0505, 0.0222]], + device='cuda:0'), grad: tensor([[ 8.0233e-07, 2.1458e-06, -3.6182e-07, ..., 2.5723e-06, + 7.8144e-09, 1.8897e-06], + [ 3.2187e-06, 4.3493e-07, 1.6140e-06, ..., 2.3656e-06, + 4.8982e-08, 4.2608e-07], + [ 5.7593e-06, 3.8855e-06, 1.3988e-06, ..., 3.8967e-06, + 3.8766e-08, -6.7279e-06], + ..., + [ 7.8559e-05, 4.2140e-05, 2.6494e-05, ..., 5.8383e-05, + 1.6845e-07, 9.5461e-07], + [ 4.3921e-06, 3.3490e-06, 1.5730e-06, ..., 4.0196e-06, + 1.1554e-08, -4.5728e-07], + [-1.1200e-04, -4.2319e-05, -3.0875e-05, ..., -8.6427e-05, + 1.7369e-07, 3.3434e-07]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0300, -0.0135, -0.0102, 0.0172, -0.0037, -0.0136, -0.0036, 0.0118, + -0.0243, 0.0149], device='cuda:0'), grad: tensor([ 8.9705e-06, 2.9150e-06, 1.0533e-06, -2.5958e-05, 1.4745e-05, + 1.4417e-05, 6.3851e-06, 1.6940e-04, 1.3612e-05, -2.0516e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 218.56, cls_loss 0.0034 cls_loss_mapping 0.0126 cls_loss_causal 0.5941 re_mapping 0.0074 re_causal 0.0226 /// teacc 99.06 lr 0.00010000 +Epoch 89, weight, value: tensor([[ 0.0080, -0.0332, 0.0147, ..., -0.0692, -0.0534, -0.0325], + [-0.0624, 0.0328, -0.0340, ..., -0.0552, -0.0241, -0.0956], + [-0.0488, -0.0610, -0.0671, ..., -0.0734, -0.0310, 0.0863], + ..., + [ 0.0546, -0.0013, -0.0452, ..., 0.0475, -0.0272, -0.0251], + [ 0.0745, 0.0488, 0.0232, ..., -0.0472, -0.0606, 0.0477], + [ 0.0402, -0.0050, 0.0336, ..., 0.0350, -0.0507, 0.0223]], + device='cuda:0'), grad: tensor([[-1.5032e-06, 1.9725e-06, 6.5984e-07, ..., -5.6624e-07, + 1.1744e-06, 2.8443e-06], + [-5.5134e-06, -4.6492e-05, 8.6054e-07, ..., -6.8136e-06, + 5.2992e-07, 1.2830e-05], + [ 3.9265e-06, 5.2378e-06, 1.0226e-06, ..., 3.3714e-06, + 4.5635e-07, -2.3261e-05], + ..., + [-5.3644e-06, 1.7285e-05, 8.0699e-07, ..., -8.2999e-06, + 1.1902e-06, 5.1223e-06], + [-9.5963e-06, 1.0133e-05, -9.2611e-06, ..., 5.2415e-06, + 6.1002e-07, -7.8678e-06], + [-2.6356e-06, 9.7975e-06, -4.1327e-07, ..., -2.9102e-05, + -1.4096e-05, -8.4862e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0288, -0.0134, -0.0118, 0.0186, -0.0037, -0.0138, -0.0034, 0.0118, + -0.0244, 0.0154], device='cuda:0'), grad: tensor([ 2.6971e-06, -4.6998e-05, -6.5744e-05, 1.4966e-06, 4.6164e-05, + 1.6287e-05, 6.2361e-06, 3.2783e-05, 1.4402e-05, -7.4729e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 218.68, cls_loss 0.0025 cls_loss_mapping 0.0085 cls_loss_causal 0.6037 re_mapping 0.0075 re_causal 0.0230 /// teacc 98.99 lr 0.00010000 +Epoch 90, weight, value: tensor([[ 0.0083, -0.0336, 0.0148, ..., -0.0694, -0.0550, -0.0333], + [-0.0630, 0.0332, -0.0343, ..., -0.0559, -0.0252, -0.0970], + [-0.0491, -0.0616, -0.0673, ..., -0.0741, -0.0312, 0.0871], + ..., + [ 0.0552, -0.0014, -0.0455, ..., 0.0482, -0.0276, -0.0256], + [ 0.0750, 0.0490, 0.0236, ..., -0.0476, -0.0611, 0.0480], + [ 0.0404, -0.0048, 0.0341, ..., 0.0351, -0.0508, 0.0223]], + device='cuda:0'), grad: tensor([[ 1.3644e-07, 1.9055e-06, 9.4250e-07, ..., 6.4448e-07, + 3.8408e-06, 5.4128e-06], + [ 7.8650e-07, 1.1940e-06, 1.1306e-06, ..., 4.6715e-06, + 3.7830e-06, 5.4501e-06], + [-2.2538e-06, -2.7660e-06, -2.7996e-06, ..., 6.6124e-06, + 6.1803e-06, -4.2915e-05], + ..., + [ 1.1729e-07, 3.6303e-06, 1.4435e-06, ..., 3.5837e-06, + 3.1795e-06, 4.0494e-06], + [-6.0946e-06, -1.4849e-05, -9.4622e-06, ..., 5.6857e-07, + 5.1269e-07, -1.6481e-05], + [ 2.4401e-06, 8.4862e-06, 4.6752e-06, ..., 1.9059e-05, + 1.4775e-05, 4.7535e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0286, -0.0136, -0.0117, 0.0185, -0.0040, -0.0137, -0.0032, 0.0122, + -0.0244, 0.0154], device='cuda:0'), grad: tensor([ 1.1690e-05, 1.1787e-05, -8.0407e-05, 6.3963e-06, -2.9340e-05, + -4.0047e-06, -5.1744e-06, 1.4104e-05, -3.5822e-05, 1.1063e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 218.95, cls_loss 0.0028 cls_loss_mapping 0.0092 cls_loss_causal 0.6066 re_mapping 0.0071 re_causal 0.0220 /// teacc 99.05 lr 0.00010000 +Epoch 91, weight, value: tensor([[ 0.0087, -0.0339, 0.0149, ..., -0.0697, -0.0557, -0.0339], + [-0.0634, 0.0332, -0.0347, ..., -0.0566, -0.0260, -0.0977], + [-0.0500, -0.0617, -0.0676, ..., -0.0754, -0.0318, 0.0876], + ..., + [ 0.0556, -0.0019, -0.0462, ..., 0.0487, -0.0283, -0.0259], + [ 0.0755, 0.0497, 0.0240, ..., -0.0479, -0.0617, 0.0487], + [ 0.0406, -0.0047, 0.0345, ..., 0.0353, -0.0512, 0.0220]], + device='cuda:0'), grad: tensor([[-1.0341e-05, 5.3272e-06, -1.0490e-05, ..., 4.2506e-06, + 5.3197e-06, 7.7784e-06], + [ 1.8450e-06, -4.5747e-05, 4.3027e-06, ..., -3.2224e-06, + 4.7795e-06, 1.7151e-05], + [ 5.6177e-06, 2.9787e-05, 5.6960e-06, ..., 1.2316e-05, + 8.7470e-06, -4.1604e-05], + ..., + [-9.7007e-06, 6.1870e-05, 8.0094e-06, ..., 2.2739e-05, + 2.3142e-05, 3.4362e-05], + [-1.1146e-05, 7.4267e-05, 7.2382e-06, ..., 1.0371e-05, + 9.3281e-06, 4.4256e-06], + [ 5.8971e-06, 6.6310e-06, -3.3956e-06, ..., -8.3447e-06, + -2.3711e-06, 1.1869e-05]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0288, -0.0138, -0.0118, 0.0184, -0.0039, -0.0137, -0.0031, 0.0124, + -0.0242, 0.0152], device='cuda:0'), grad: tensor([ 1.9874e-06, -2.2638e-04, 3.1501e-05, -2.0468e-04, -1.1331e-04, + 7.2896e-05, 3.6925e-05, 1.8859e-04, 1.4806e-04, 6.4731e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 218.27, cls_loss 0.0028 cls_loss_mapping 0.0099 cls_loss_causal 0.5550 re_mapping 0.0073 re_causal 0.0220 /// teacc 98.95 lr 0.00010000 +Epoch 92, weight, value: tensor([[ 0.0088, -0.0344, 0.0151, ..., -0.0699, -0.0566, -0.0345], + [-0.0638, 0.0334, -0.0352, ..., -0.0568, -0.0268, -0.0985], + [-0.0507, -0.0622, -0.0679, ..., -0.0764, -0.0321, 0.0882], + ..., + [ 0.0559, -0.0024, -0.0468, ..., 0.0499, -0.0285, -0.0261], + [ 0.0761, 0.0502, 0.0243, ..., -0.0485, -0.0621, 0.0491], + [ 0.0410, -0.0042, 0.0353, ..., 0.0356, -0.0515, 0.0219]], + device='cuda:0'), grad: tensor([[-1.1213e-06, 9.0711e-07, -6.8126e-07, ..., 8.8429e-07, + 5.5321e-07, 2.5481e-06], + [ 1.4445e-06, -4.9695e-06, 9.2061e-07, ..., 1.8775e-06, + 1.0775e-06, 3.3472e-06], + [ 7.4022e-06, 4.8988e-06, 4.2692e-06, ..., 1.2238e-06, + 5.4203e-07, -1.3590e-05], + ..., + [ 3.5390e-08, 1.6112e-06, 5.4436e-07, ..., 1.7602e-07, + 8.4471e-07, 3.2075e-06], + [-2.3112e-05, -1.0453e-05, -1.2532e-05, ..., 5.9092e-07, + 3.8650e-07, -6.9663e-06], + [ 5.9642e-06, 7.7933e-06, 3.0231e-06, ..., 4.3750e-05, + 2.6032e-05, 2.9862e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0288, -0.0141, -0.0116, 0.0185, -0.0048, -0.0139, -0.0027, 0.0128, + -0.0242, 0.0155], device='cuda:0'), grad: tensor([ 3.9265e-06, -8.3968e-06, -5.5395e-06, 2.9013e-05, -5.8621e-05, + -4.4584e-05, 2.0429e-05, 1.0498e-05, -2.3350e-05, 7.6532e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 218.28, cls_loss 0.0024 cls_loss_mapping 0.0069 cls_loss_causal 0.5765 re_mapping 0.0071 re_causal 0.0213 /// teacc 99.03 lr 0.00010000 +Epoch 93, weight, value: tensor([[ 0.0090, -0.0346, 0.0153, ..., -0.0703, -0.0578, -0.0353], + [-0.0645, 0.0334, -0.0355, ..., -0.0572, -0.0276, -0.0992], + [-0.0514, -0.0625, -0.0679, ..., -0.0785, -0.0326, 0.0887], + ..., + [ 0.0565, -0.0024, -0.0472, ..., 0.0502, -0.0300, -0.0265], + [ 0.0766, 0.0504, 0.0246, ..., -0.0489, -0.0625, 0.0494], + [ 0.0412, -0.0042, 0.0357, ..., 0.0358, -0.0518, 0.0218]], + device='cuda:0'), grad: tensor([[-1.1832e-05, 1.1154e-05, -6.3069e-06, ..., 2.6673e-06, + 3.1060e-07, 2.5332e-06], + [ 4.8608e-05, 1.1049e-05, 3.3788e-06, ..., 6.6236e-06, + 1.2284e-06, 1.6272e-05], + [ 2.1413e-05, 2.6375e-05, 5.5693e-06, ..., 2.3529e-05, + 5.5917e-06, -1.1101e-05], + ..., + [ 6.0685e-06, 2.0850e-04, 1.8001e-05, ..., -8.7678e-05, + -3.1412e-05, 1.1884e-05], + [-1.7834e-04, 2.0728e-05, -1.2241e-05, ..., 6.0312e-06, + 5.1223e-07, -5.4032e-05], + [ 2.3276e-05, 1.2189e-05, 7.2047e-06, ..., 3.6538e-05, + 7.6815e-06, 4.9174e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0287, -0.0146, -0.0118, 0.0184, -0.0045, -0.0137, -0.0026, 0.0134, + -0.0243, 0.0154], device='cuda:0'), grad: tensor([ 1.2815e-06, 1.4806e-04, 1.0300e-04, -9.2936e-04, 1.0800e-04, + 1.6201e-04, 4.2766e-05, 5.8460e-04, -3.2425e-04, 1.0329e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 218.16, cls_loss 0.0029 cls_loss_mapping 0.0081 cls_loss_causal 0.6082 re_mapping 0.0067 re_causal 0.0213 /// teacc 99.00 lr 0.00010000 +Epoch 94, weight, value: tensor([[ 0.0091, -0.0352, 0.0154, ..., -0.0706, -0.0584, -0.0360], + [-0.0648, 0.0335, -0.0358, ..., -0.0574, -0.0283, -0.0998], + [-0.0517, -0.0629, -0.0685, ..., -0.0792, -0.0330, 0.0892], + ..., + [ 0.0564, -0.0030, -0.0481, ..., 0.0502, -0.0309, -0.0272], + [ 0.0777, 0.0512, 0.0255, ..., -0.0491, -0.0627, 0.0504], + [ 0.0414, -0.0043, 0.0360, ..., 0.0358, -0.0524, 0.0214]], + device='cuda:0'), grad: tensor([[ 1.4184e-06, 7.7933e-06, 1.1601e-05, ..., 2.9057e-06, + 3.4481e-05, 3.1263e-05], + [ 7.5698e-05, 1.7047e-05, 7.7859e-06, ..., 5.2541e-05, + 1.7956e-05, 1.8418e-05], + [ 9.9465e-06, 6.5714e-06, 5.5172e-06, ..., 8.5086e-06, + 1.3083e-05, 3.5693e-07], + ..., + [-1.8299e-04, -5.8055e-05, 1.8589e-06, ..., -1.5378e-04, + 2.7865e-06, 5.3495e-06], + [ 3.7402e-06, 7.0512e-05, 1.1963e-04, ..., 7.8231e-06, + 2.8253e-04, 2.5892e-04], + [ 6.5386e-05, 2.1726e-05, -7.8604e-06, ..., 4.6343e-05, + -6.1141e-07, 6.4773e-07]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0287, -0.0146, -0.0115, 0.0181, -0.0042, -0.0131, -0.0031, 0.0126, + -0.0234, 0.0149], device='cuda:0'), grad: tensor([ 1.1081e-04, 2.1207e-04, 3.4094e-05, 1.2034e-04, 1.3030e-04, + 3.3617e-04, -1.5860e-03, -4.2558e-04, 9.2793e-04, 1.3971e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 218.18, cls_loss 0.0023 cls_loss_mapping 0.0069 cls_loss_causal 0.5834 re_mapping 0.0070 re_causal 0.0216 /// teacc 99.03 lr 0.00010000 +Epoch 95, weight, value: tensor([[ 0.0096, -0.0356, 0.0156, ..., -0.0708, -0.0589, -0.0366], + [-0.0649, 0.0341, -0.0360, ..., -0.0576, -0.0286, -0.1002], + [-0.0519, -0.0632, -0.0686, ..., -0.0801, -0.0337, 0.0899], + ..., + [ 0.0566, -0.0033, -0.0490, ..., 0.0504, -0.0314, -0.0276], + [ 0.0779, 0.0513, 0.0256, ..., -0.0496, -0.0632, 0.0506], + [ 0.0418, -0.0041, 0.0364, ..., 0.0359, -0.0528, 0.0212]], + device='cuda:0'), grad: tensor([[ 2.3022e-05, 1.1111e-06, -3.5332e-08, ..., 2.1249e-05, + 6.1886e-07, 7.0035e-07], + [ 1.1325e-05, -1.3132e-06, 1.0263e-06, ..., 1.2197e-05, + 2.1569e-06, 2.3395e-06], + [ 3.4451e-05, 3.7756e-06, 6.2957e-07, ..., 4.0352e-05, + 5.1269e-07, 7.0408e-07], + ..., + [-1.8764e-04, -7.7710e-06, 8.4192e-07, ..., -1.8108e-04, + 1.4054e-06, -2.1197e-06], + [-1.1204e-06, -9.7454e-06, -8.9258e-06, ..., 1.0282e-05, + 1.1548e-06, -5.7034e-06], + [ 5.0664e-05, 9.2089e-06, 6.6217e-07, ..., 5.8025e-05, + 1.0103e-05, 1.1146e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0288, -0.0144, -0.0110, 0.0181, -0.0042, -0.0134, -0.0030, 0.0126, + -0.0237, 0.0149], device='cuda:0'), grad: tensor([ 1.2577e-04, 4.1395e-05, 1.3471e-04, 1.9157e-04, 5.4426e-06, + -1.9693e-04, 3.8773e-05, -5.1308e-04, 1.7062e-05, 1.5533e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 218.43, cls_loss 0.0028 cls_loss_mapping 0.0080 cls_loss_causal 0.5711 re_mapping 0.0070 re_causal 0.0206 /// teacc 98.97 lr 0.00010000 +Epoch 96, weight, value: tensor([[ 0.0100, -0.0360, 0.0157, ..., -0.0711, -0.0604, -0.0376], + [-0.0651, 0.0345, -0.0364, ..., -0.0579, -0.0292, -0.1007], + [-0.0520, -0.0636, -0.0689, ..., -0.0806, -0.0343, 0.0907], + ..., + [ 0.0568, -0.0040, -0.0498, ..., 0.0507, -0.0320, -0.0284], + [ 0.0790, 0.0524, 0.0267, ..., -0.0508, -0.0642, 0.0510], + [ 0.0417, -0.0043, 0.0360, ..., 0.0366, -0.0528, 0.0210]], + device='cuda:0'), grad: tensor([[ 3.9395e-07, 6.6916e-07, 1.0254e-06, ..., 7.5810e-07, + 4.3586e-06, 6.6012e-06], + [ 2.4419e-06, 5.3551e-07, 4.5821e-07, ..., 5.1819e-06, + 2.2501e-06, 8.7097e-06], + [-2.0210e-06, 1.4110e-06, -1.2042e-06, ..., 1.8291e-06, + 1.2834e-06, -2.1994e-05], + ..., + [-2.3589e-05, -9.0227e-06, 5.7369e-07, ..., -3.4362e-05, + 6.5984e-07, 2.6356e-06], + [ 3.0883e-06, 3.0734e-06, 2.2966e-06, ..., 2.6003e-06, + 1.4137e-06, 8.4341e-06], + [ 1.5073e-05, 6.7316e-06, -1.4734e-06, ..., 2.1428e-05, + 3.6974e-07, 5.9791e-07]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0290, -0.0144, -0.0107, 0.0179, -0.0046, -0.0134, -0.0021, 0.0122, + -0.0232, 0.0144], device='cuda:0'), grad: tensor([ 1.8343e-05, 2.6390e-05, -5.5552e-05, 8.9174e-08, -7.5949e-07, + -1.7956e-05, -2.8145e-06, -4.1485e-05, 3.5286e-05, 3.8385e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 218.23, cls_loss 0.0027 cls_loss_mapping 0.0090 cls_loss_causal 0.5902 re_mapping 0.0070 re_causal 0.0208 /// teacc 99.11 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 0.0102, -0.0366, 0.0158, ..., -0.0715, -0.0623, -0.0386], + [-0.0653, 0.0350, -0.0365, ..., -0.0583, -0.0303, -0.1013], + [-0.0524, -0.0642, -0.0694, ..., -0.0811, -0.0347, 0.0913], + ..., + [ 0.0569, -0.0044, -0.0505, ..., 0.0508, -0.0329, -0.0293], + [ 0.0795, 0.0528, 0.0271, ..., -0.0515, -0.0653, 0.0516], + [ 0.0423, -0.0042, 0.0365, ..., 0.0367, -0.0530, 0.0209]], + device='cuda:0'), grad: tensor([[-5.2571e-05, -5.7742e-06, -2.1264e-05, ..., 1.1437e-06, + 3.6275e-07, 3.6508e-07], + [ 4.6417e-06, -7.8306e-06, 1.9558e-06, ..., 4.2133e-06, + 6.9570e-07, 1.2135e-06], + [ 3.3170e-05, 1.1191e-05, 8.7544e-06, ..., 2.3007e-05, + 6.6590e-07, 5.5227e-07], + ..., + [-3.1590e-05, 5.4017e-06, 3.2708e-06, ..., -5.1945e-05, + 6.6590e-07, 1.0617e-06], + [ 1.6382e-06, 1.2573e-06, 4.5693e-08, ..., 2.9616e-06, + 4.4634e-07, -1.5255e-06], + [ 1.2003e-05, 3.4273e-06, 3.1777e-06, ..., 2.0579e-05, + 6.8918e-06, 7.2867e-06]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0284, -0.0143, -0.0106, 0.0176, -0.0045, -0.0131, -0.0014, 0.0118, + -0.0232, 0.0144], device='cuda:0'), grad: tensor([-1.9145e-04, -1.1551e-04, 1.7726e-04, 1.5177e-05, 9.7007e-06, + 5.8949e-05, 1.8075e-05, -5.1320e-05, 1.3627e-05, 6.5088e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 218.08, cls_loss 0.0022 cls_loss_mapping 0.0065 cls_loss_causal 0.5788 re_mapping 0.0070 re_causal 0.0211 /// teacc 99.05 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0109, -0.0367, 0.0160, ..., -0.0717, -0.0627, -0.0391], + [-0.0658, 0.0350, -0.0372, ..., -0.0588, -0.0310, -0.1021], + [-0.0527, -0.0646, -0.0697, ..., -0.0819, -0.0351, 0.0918], + ..., + [ 0.0573, -0.0046, -0.0508, ..., 0.0512, -0.0335, -0.0298], + [ 0.0796, 0.0529, 0.0274, ..., -0.0522, -0.0657, 0.0520], + [ 0.0425, -0.0041, 0.0368, ..., 0.0367, -0.0535, 0.0207]], + device='cuda:0'), grad: tensor([[ 1.2768e-06, 4.5709e-06, -1.1399e-05, ..., 1.7341e-06, + 4.5891e-07, -3.9302e-06], + [-1.4365e-05, -3.5822e-05, 6.6450e-07, ..., 8.4564e-06, + 1.1809e-06, 7.3574e-06], + [ 2.2203e-06, 4.9360e-06, 3.6974e-06, ..., 1.4855e-06, + 4.5588e-07, -1.1446e-06], + ..., + [ 4.1500e-06, 8.3223e-06, 1.6224e-06, ..., 5.9307e-06, + 2.1625e-06, 3.3546e-06], + [ 2.0370e-05, 4.7594e-05, 3.1274e-06, ..., 6.3032e-06, + 7.2364e-07, 1.1558e-06], + [-4.0025e-05, -8.1658e-05, -7.0706e-06, ..., -3.2157e-05, + 2.1569e-06, -1.4842e-05]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0290, -0.0145, -0.0107, 0.0176, -0.0042, -0.0131, -0.0015, 0.0121, + -0.0233, 0.0140], device='cuda:0'), grad: tensor([-1.1146e-05, -1.9863e-05, 1.1571e-05, 3.1084e-05, 2.2560e-05, + 7.3195e-05, 7.6219e-06, 2.6420e-05, 8.4817e-05, -2.2638e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 97---------------------------------------------------- +epoch 97, time 218.64, cls_loss 0.0024 cls_loss_mapping 0.0077 cls_loss_causal 0.5212 re_mapping 0.0068 re_causal 0.0197 /// teacc 99.14 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0085, -0.0372, 0.0143, ..., -0.0744, -0.0633, -0.0396], + [-0.0660, 0.0350, -0.0375, ..., -0.0595, -0.0327, -0.1034], + [-0.0534, -0.0650, -0.0700, ..., -0.0818, -0.0352, 0.0930], + ..., + [ 0.0578, -0.0048, -0.0513, ..., 0.0514, -0.0340, -0.0310], + [ 0.0797, 0.0529, 0.0275, ..., -0.0527, -0.0660, 0.0521], + [ 0.0449, -0.0039, 0.0386, ..., 0.0379, -0.0541, 0.0205]], + device='cuda:0'), grad: tensor([[-2.3469e-07, 4.9919e-07, -1.0780e-07, ..., 1.0841e-06, + 3.6024e-06, 4.0904e-06], + [-1.7826e-06, -6.3777e-06, 5.3551e-07, ..., 4.9546e-06, + 4.8541e-06, 4.6119e-06], + [ 1.5972e-06, 1.9148e-06, 6.5705e-07, ..., 4.9807e-06, + 5.2452e-05, 5.9277e-05], + ..., + [-2.9318e-06, 4.6119e-06, 8.7917e-07, ..., 9.5069e-06, + 9.2983e-06, 7.8678e-06], + [-2.6077e-07, 3.2950e-06, -5.2946e-07, ..., 5.3346e-06, + 4.3437e-06, 2.7008e-06], + [ 1.1474e-06, 2.5228e-05, 1.2860e-05, ..., 7.9572e-05, + 5.0873e-05, 4.3005e-05]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0269, -0.0150, -0.0102, 0.0177, -0.0037, -0.0133, -0.0016, 0.0120, + -0.0236, 0.0154], device='cuda:0'), grad: tensor([ 7.2010e-06, -1.5035e-05, 9.5665e-05, 2.6394e-06, -1.4496e-04, + 1.0580e-05, -9.6679e-05, 1.2018e-05, 1.5885e-05, 1.1283e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 218.32, cls_loss 0.0024 cls_loss_mapping 0.0072 cls_loss_causal 0.5939 re_mapping 0.0066 re_causal 0.0210 /// teacc 99.03 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0088, -0.0377, 0.0144, ..., -0.0744, -0.0643, -0.0404], + [-0.0663, 0.0352, -0.0379, ..., -0.0603, -0.0339, -0.1042], + [-0.0540, -0.0656, -0.0707, ..., -0.0824, -0.0357, 0.0935], + ..., + [ 0.0582, -0.0053, -0.0521, ..., 0.0520, -0.0344, -0.0313], + [ 0.0802, 0.0531, 0.0278, ..., -0.0535, -0.0667, 0.0523], + [ 0.0449, -0.0039, 0.0389, ..., 0.0378, -0.0544, 0.0204]], + device='cuda:0'), grad: tensor([[-3.3863e-06, 7.4096e-06, 1.8384e-06, ..., 1.3374e-06, + 2.2482e-06, 1.0274e-05], + [ 1.0706e-05, 2.0891e-05, 1.6451e-05, ..., 3.3341e-06, + 7.5623e-06, 2.3186e-05], + [ 2.2680e-05, 2.6554e-05, 2.8089e-05, ..., 1.6257e-05, + 2.2110e-06, 6.6817e-05], + ..., + [-1.3463e-05, -2.7455e-06, 2.3711e-06, ..., -3.8892e-05, + -6.4960e-07, -3.2894e-06], + [-2.3055e-04, -5.2309e-04, -3.7313e-04, ..., 4.6715e-06, + -1.3340e-04, -5.2977e-04], + [ 9.2611e-06, 1.8388e-05, 1.0885e-05, ..., 3.3021e-05, + 1.7479e-05, 3.4273e-05]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0271, -0.0153, -0.0104, 0.0184, -0.0037, -0.0135, -0.0012, 0.0121, + -0.0239, 0.0151], device='cuda:0'), grad: tensor([ 1.1608e-05, 6.1005e-05, 1.8823e-04, 2.7254e-05, -3.2336e-05, + 4.3392e-05, 1.1110e-03, -4.0323e-05, -1.4734e-03, 1.0437e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 218.28, cls_loss 0.0022 cls_loss_mapping 0.0059 cls_loss_causal 0.5575 re_mapping 0.0068 re_causal 0.0203 /// teacc 98.98 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0089, -0.0380, 0.0145, ..., -0.0744, -0.0654, -0.0410], + [-0.0667, 0.0353, -0.0383, ..., -0.0607, -0.0347, -0.1049], + [-0.0541, -0.0661, -0.0709, ..., -0.0828, -0.0364, 0.0945], + ..., + [ 0.0586, -0.0054, -0.0524, ..., 0.0526, -0.0356, -0.0320], + [ 0.0808, 0.0538, 0.0284, ..., -0.0539, -0.0673, 0.0528], + [ 0.0449, -0.0038, 0.0392, ..., 0.0377, -0.0549, 0.0200]], + device='cuda:0'), grad: tensor([[-7.3910e-06, 2.6952e-06, -1.9781e-06, ..., 3.1758e-07, + 9.3738e-07, 3.9674e-06], + [ 1.2629e-05, 1.3076e-05, 1.0125e-05, ..., 7.4320e-07, + 2.8731e-07, 2.3082e-05], + [-7.3621e-07, 1.6000e-06, 1.5832e-06, ..., -8.2701e-07, + 1.0403e-06, -2.9594e-05], + ..., + [ 1.3988e-06, 1.4501e-06, 1.3774e-06, ..., 1.1856e-06, + 3.2131e-07, 2.3283e-06], + [-6.9380e-05, -8.4996e-05, -5.9485e-05, ..., 1.6047e-06, + 5.8487e-07, -8.1420e-05], + [-8.8708e-07, -1.6736e-06, -1.9008e-06, ..., -7.5698e-06, + -4.5681e-07, 4.2804e-06]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0272, -0.0152, -0.0101, 0.0176, -0.0036, -0.0133, -0.0010, 0.0123, + -0.0238, 0.0147], device='cuda:0'), grad: tensor([-2.8446e-05, 6.3419e-05, -5.9724e-05, 2.4930e-05, 1.3888e-05, + 2.5243e-05, 1.7667e-04, 7.9349e-06, -2.2995e-04, 5.5768e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 218.76, cls_loss 0.0022 cls_loss_mapping 0.0073 cls_loss_causal 0.5466 re_mapping 0.0062 re_causal 0.0194 /// teacc 99.03 lr 0.00010000 +Epoch 102, weight, value: tensor([[ 0.0090, -0.0383, 0.0146, ..., -0.0744, -0.0659, -0.0415], + [-0.0669, 0.0359, -0.0388, ..., -0.0610, -0.0356, -0.1043], + [-0.0548, -0.0668, -0.0712, ..., -0.0840, -0.0373, 0.0946], + ..., + [ 0.0591, -0.0057, -0.0528, ..., 0.0533, -0.0361, -0.0322], + [ 0.0812, 0.0539, 0.0288, ..., -0.0547, -0.0675, 0.0530], + [ 0.0451, -0.0035, 0.0396, ..., 0.0380, -0.0549, 0.0202]], + device='cuda:0'), grad: tensor([[ 1.6473e-07, 3.7672e-07, 1.0245e-08, ..., 2.9244e-07, + 5.3830e-07, 8.2934e-07], + [ 4.2305e-07, -7.3649e-06, 6.4494e-08, ..., 7.4040e-07, + 5.9046e-07, -1.5311e-06], + [-8.4424e-07, 3.5390e-06, 5.3318e-08, ..., 1.4529e-06, + 5.5227e-07, -3.1274e-06], + ..., + [-3.4645e-06, 6.9104e-07, 8.5915e-08, ..., -4.7348e-06, + 5.4110e-07, 1.2051e-06], + [ 9.7416e-07, 6.3516e-07, -1.4901e-07, ..., 2.0918e-06, + 1.1018e-06, 2.7139e-06], + [-3.3120e-08, -1.8813e-07, -6.0257e-07, ..., 1.5404e-06, + 1.5320e-06, 1.0096e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0273, -0.0144, -0.0108, 0.0174, -0.0039, -0.0126, -0.0014, 0.0128, + -0.0241, 0.0147], device='cuda:0'), grad: tensor([ 2.4084e-06, -1.9535e-05, 1.0356e-06, 4.6343e-06, -1.6494e-06, + 1.0528e-05, -7.5698e-06, -6.3404e-06, 1.2085e-05, 4.3400e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 218.48, cls_loss 0.0026 cls_loss_mapping 0.0092 cls_loss_causal 0.5611 re_mapping 0.0065 re_causal 0.0203 /// teacc 99.08 lr 0.00010000 +Epoch 103, weight, value: tensor([[ 0.0091, -0.0387, 0.0147, ..., -0.0745, -0.0665, -0.0420], + [-0.0681, 0.0359, -0.0391, ..., -0.0619, -0.0369, -0.1053], + [-0.0549, -0.0674, -0.0714, ..., -0.0844, -0.0377, 0.0953], + ..., + [ 0.0580, -0.0061, -0.0550, ..., 0.0529, -0.0370, -0.0327], + [ 0.0820, 0.0540, 0.0290, ..., -0.0553, -0.0679, 0.0531], + [ 0.0463, -0.0035, 0.0406, ..., 0.0384, -0.0553, 0.0200]], + device='cuda:0'), grad: tensor([[-1.7285e-05, 1.7732e-06, -3.0603e-06, ..., -3.7206e-07, + 4.9360e-06, -1.4680e-07], + [ 1.2837e-05, -4.6100e-07, 1.3607e-06, ..., 1.7315e-05, + 1.1716e-06, 2.2948e-06], + [ 1.3009e-05, 5.5842e-06, 5.9558e-07, ..., 1.7196e-05, + 2.8498e-06, 7.7812e-07], + ..., + [-5.8293e-05, -1.2562e-05, 6.9663e-07, ..., -8.2493e-05, + -1.5251e-05, -1.5453e-05], + [ 1.3143e-05, 1.0282e-05, 3.3882e-06, ..., 1.8165e-05, + 3.9712e-06, 6.0312e-06], + [-2.9500e-07, -1.7196e-05, -9.9614e-06, ..., -8.0541e-06, + 2.3711e-06, -8.0913e-06]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0275, -0.0154, -0.0103, 0.0164, -0.0031, -0.0117, -0.0016, 0.0117, + -0.0241, 0.0153], device='cuda:0'), grad: tensor([-2.6941e-05, 2.3350e-05, 4.1515e-05, -1.3724e-05, 6.6280e-05, + 5.3607e-06, 1.2308e-05, -1.5414e-04, 4.4197e-05, 1.8375e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 218.59, cls_loss 0.0020 cls_loss_mapping 0.0066 cls_loss_causal 0.5596 re_mapping 0.0066 re_causal 0.0196 /// teacc 99.07 lr 0.00010000 +Epoch 104, weight, value: tensor([[ 0.0094, -0.0391, 0.0148, ..., -0.0745, -0.0670, -0.0424], + [-0.0681, 0.0364, -0.0394, ..., -0.0622, -0.0378, -0.1058], + [-0.0550, -0.0679, -0.0716, ..., -0.0847, -0.0380, 0.0957], + ..., + [ 0.0581, -0.0068, -0.0554, ..., 0.0528, -0.0376, -0.0330], + [ 0.0825, 0.0544, 0.0293, ..., -0.0557, -0.0681, 0.0535], + [ 0.0464, -0.0030, 0.0409, ..., 0.0388, -0.0555, 0.0198]], + device='cuda:0'), grad: tensor([[ 4.8382e-07, 7.3761e-07, 2.5495e-07, ..., 3.9339e-06, + 9.8348e-06, 7.7412e-06], + [ 3.6880e-07, -1.6661e-06, 4.4960e-07, ..., 6.7614e-06, + 1.4231e-05, 1.2808e-05], + [-1.3842e-07, 1.1893e-06, 2.7660e-07, ..., 1.4700e-05, + 4.1187e-05, 2.3708e-05], + ..., + [ 4.2804e-06, 3.7346e-06, 1.0571e-06, ..., 9.7603e-06, + 1.1489e-05, 9.0972e-06], + [ 1.6853e-05, 1.2547e-05, 1.3262e-05, ..., 2.3872e-05, + 3.6687e-05, 3.8087e-05], + [-2.5988e-05, -1.5914e-05, -1.5855e-05, ..., -1.3769e-05, + 1.4186e-05, -1.9986e-06]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0278, -0.0152, -0.0103, 0.0165, -0.0032, -0.0119, -0.0018, 0.0114, + -0.0239, 0.0154], device='cuda:0'), grad: tensor([ 2.5377e-05, 3.1769e-05, 7.9215e-05, 1.4856e-05, -3.8767e-04, + 1.4529e-05, 7.6413e-05, 4.0621e-05, 1.1683e-04, -1.2308e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 218.39, cls_loss 0.0022 cls_loss_mapping 0.0077 cls_loss_causal 0.5337 re_mapping 0.0065 re_causal 0.0193 /// teacc 98.91 lr 0.00010000 +Epoch 105, weight, value: tensor([[ 0.0095, -0.0397, 0.0148, ..., -0.0746, -0.0675, -0.0428], + [-0.0687, 0.0360, -0.0398, ..., -0.0629, -0.0388, -0.1068], + [-0.0551, -0.0684, -0.0717, ..., -0.0851, -0.0386, 0.0963], + ..., + [ 0.0583, -0.0068, -0.0556, ..., 0.0533, -0.0380, -0.0335], + [ 0.0832, 0.0548, 0.0294, ..., -0.0563, -0.0686, 0.0539], + [ 0.0466, -0.0024, 0.0412, ..., 0.0383, -0.0569, 0.0191]], + device='cuda:0'), grad: tensor([[ 2.6170e-06, 1.8058e-06, -6.3842e-07, ..., 5.9716e-06, + 1.3364e-06, 3.5502e-06], + [ 4.9360e-07, 2.0082e-07, 3.1944e-07, ..., 5.3970e-07, + 4.6543e-07, 7.6368e-07], + [ 4.2515e-07, 1.3048e-06, 5.2666e-07, ..., 1.4277e-06, + 6.4075e-07, -1.7323e-07], + ..., + [ 1.4417e-06, 1.1101e-06, 5.1828e-07, ..., 2.1681e-06, + 5.6531e-07, 1.6717e-06], + [-1.2629e-06, -1.1949e-06, -1.1390e-06, ..., 7.1945e-07, + 2.0266e-06, 6.6636e-07], + [-1.0751e-05, -4.4480e-06, -2.0619e-06, ..., -1.7628e-05, + -1.8720e-06, -8.7693e-06]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0279, -0.0160, -0.0100, 0.0165, -0.0019, -0.0123, -0.0017, 0.0117, + -0.0239, 0.0149], device='cuda:0'), grad: tensor([ 5.2676e-06, 1.3905e-06, 1.0356e-06, -9.7230e-07, 1.0923e-05, + 1.4780e-06, -7.4366e-07, 5.8785e-06, 2.3469e-06, -2.6554e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 218.24, cls_loss 0.0022 cls_loss_mapping 0.0086 cls_loss_causal 0.5645 re_mapping 0.0064 re_causal 0.0198 /// teacc 99.02 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 0.0095, -0.0408, 0.0149, ..., -0.0747, -0.0684, -0.0433], + [-0.0693, 0.0360, -0.0401, ..., -0.0632, -0.0397, -0.1078], + [-0.0561, -0.0690, -0.0721, ..., -0.0861, -0.0404, 0.0963], + ..., + [ 0.0584, -0.0078, -0.0569, ..., 0.0533, -0.0381, -0.0336], + [ 0.0838, 0.0554, 0.0297, ..., -0.0566, -0.0696, 0.0540], + [ 0.0471, -0.0017, 0.0418, ..., 0.0388, -0.0571, 0.0189]], + device='cuda:0'), grad: tensor([[-1.8299e-04, -8.5458e-06, -4.6581e-05, ..., 4.6566e-06, + -2.6417e-04, -2.2233e-05], + [ 1.3441e-05, 4.3362e-05, 4.6901e-06, ..., 3.9995e-05, + 1.9461e-05, 6.2324e-06], + [ 9.9316e-06, 1.1802e-05, 3.5726e-06, ..., 9.0003e-06, + 8.2925e-06, -4.3124e-05], + ..., + [-6.6496e-06, 1.5989e-05, 3.4217e-06, ..., -1.7986e-05, + 1.2830e-05, 5.4277e-06], + [-2.3395e-05, -1.8850e-05, -9.8348e-06, ..., 6.2361e-06, + 1.7315e-05, -1.7628e-05], + [ 4.2737e-05, 4.6045e-05, 9.4920e-06, ..., 9.5129e-05, + 1.3411e-04, 9.5487e-05]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0278, -0.0166, -0.0105, 0.0165, -0.0023, -0.0121, -0.0007, 0.0117, + -0.0240, 0.0151], device='cuda:0'), grad: tensor([-7.6008e-04, 1.4579e-04, -9.2030e-05, 3.0056e-05, -3.8052e-04, + 1.8430e-04, 4.7469e-04, 3.6955e-05, -2.5332e-06, 3.6216e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 218.51, cls_loss 0.0020 cls_loss_mapping 0.0082 cls_loss_causal 0.5396 re_mapping 0.0065 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0095, -0.0413, 0.0150, ..., -0.0749, -0.0685, -0.0440], + [-0.0697, 0.0362, -0.0404, ..., -0.0639, -0.0409, -0.1084], + [-0.0562, -0.0696, -0.0722, ..., -0.0868, -0.0407, 0.0973], + ..., + [ 0.0591, -0.0081, -0.0573, ..., 0.0540, -0.0383, -0.0341], + [ 0.0839, 0.0554, 0.0298, ..., -0.0573, -0.0703, 0.0538], + [ 0.0471, -0.0013, 0.0423, ..., 0.0391, -0.0574, 0.0189]], + device='cuda:0'), grad: tensor([[-6.3926e-06, 1.0608e-06, 1.4040e-07, ..., 1.6643e-06, + 9.5135e-07, -1.3475e-08], + [ 6.8871e-07, 6.4075e-06, 4.2189e-07, ..., 1.4015e-05, + 9.8348e-06, 6.0350e-06], + [ 9.4064e-07, 1.0794e-06, 5.5833e-07, ..., 1.6764e-06, + -5.2229e-06, -1.4268e-05], + ..., + [-1.5339e-06, 5.2713e-06, 3.0827e-07, ..., 8.2701e-06, + 7.4282e-06, 3.5055e-06], + [-2.8033e-06, -5.0757e-07, -3.7160e-06, ..., 6.7353e-06, + 3.9376e-06, -2.3432e-06], + [ 3.8254e-07, 8.7395e-06, -3.1851e-07, ..., 1.7986e-05, + 1.0490e-05, 2.1625e-06]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0277, -0.0163, -0.0106, 0.0165, -0.0025, -0.0121, -0.0007, 0.0123, + -0.0245, 0.0151], device='cuda:0'), grad: tensor([-1.0841e-05, 3.3617e-05, -3.6955e-05, 8.1435e-06, -8.7261e-05, + 4.7311e-06, 3.5584e-05, 2.0429e-05, 1.6438e-06, 3.0845e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 218.32, cls_loss 0.0026 cls_loss_mapping 0.0075 cls_loss_causal 0.5765 re_mapping 0.0065 re_causal 0.0194 /// teacc 99.12 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0096, -0.0423, 0.0150, ..., -0.0749, -0.0701, -0.0449], + [-0.0703, 0.0363, -0.0410, ..., -0.0644, -0.0419, -0.1093], + [-0.0566, -0.0707, -0.0725, ..., -0.0874, -0.0413, 0.0980], + ..., + [ 0.0596, -0.0085, -0.0575, ..., 0.0544, -0.0389, -0.0347], + [ 0.0843, 0.0561, 0.0302, ..., -0.0582, -0.0709, 0.0543], + [ 0.0475, -0.0004, 0.0430, ..., 0.0398, -0.0571, 0.0192]], + device='cuda:0'), grad: tensor([[ 5.0478e-06, 1.2450e-05, 3.4943e-06, ..., 9.0525e-06, + 8.5589e-07, 6.5193e-07], + [ 2.6464e-05, 1.1170e-04, 5.3607e-06, ..., 1.2350e-04, + 1.8612e-05, 7.2382e-06], + [-2.1029e-06, 2.4721e-05, 7.6666e-06, ..., 1.0222e-05, + -2.4699e-06, -3.8147e-05], + ..., + [-2.0415e-05, 7.7605e-05, 1.8865e-05, ..., 2.7448e-05, + 1.4894e-05, 2.1067e-06], + [ 4.0382e-06, 3.5524e-05, 1.1124e-05, ..., 9.0078e-06, + 1.7155e-06, 7.5698e-06], + [-8.9407e-06, -3.1646e-06, 2.2024e-05, ..., -3.0100e-05, + 7.1414e-06, 8.9733e-07]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0277, -0.0166, -0.0107, 0.0179, -0.0031, -0.0136, -0.0009, 0.0124, + -0.0245, 0.0157], device='cuda:0'), grad: tensor([ 4.3839e-05, 3.7932e-04, -1.2323e-05, -5.4169e-04, -2.0719e-04, + 5.0634e-05, 1.0654e-05, 1.7798e-04, 1.1355e-04, -1.5453e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 218.23, cls_loss 0.0020 cls_loss_mapping 0.0081 cls_loss_causal 0.5245 re_mapping 0.0066 re_causal 0.0187 /// teacc 98.98 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0096, -0.0425, 0.0150, ..., -0.0755, -0.0704, -0.0454], + [-0.0706, 0.0370, -0.0413, ..., -0.0648, -0.0416, -0.1089], + [-0.0572, -0.0719, -0.0727, ..., -0.0880, -0.0420, 0.0980], + ..., + [ 0.0605, -0.0084, -0.0577, ..., 0.0553, -0.0401, -0.0351], + [ 0.0844, 0.0561, 0.0305, ..., -0.0591, -0.0714, 0.0546], + [ 0.0475, -0.0003, 0.0433, ..., 0.0397, -0.0577, 0.0189]], + device='cuda:0'), grad: tensor([[-4.9412e-05, 5.4277e-06, 7.1898e-07, ..., 1.2275e-06, + 2.8357e-05, 1.5974e-05], + [ 9.2853e-07, -4.7721e-06, 5.0664e-07, ..., -1.3420e-06, + 3.1460e-06, 3.9674e-06], + [ 2.2575e-06, 1.6382e-06, 4.3563e-07, ..., 1.0133e-06, + 1.9046e-06, -6.9797e-05], + ..., + [ 5.0142e-06, 2.7530e-06, 9.9000e-07, ..., 6.3283e-07, + 6.8080e-07, 3.2540e-06], + [ 2.4036e-05, 2.2575e-05, 1.4909e-05, ..., 2.3633e-05, + 1.1757e-05, 6.4135e-05], + [-3.2097e-05, -6.7055e-05, -4.1842e-05, ..., -6.4969e-05, + -3.2902e-05, -2.9951e-05]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0277, -0.0156, -0.0114, 0.0177, -0.0028, -0.0134, -0.0010, 0.0130, + -0.0249, 0.0153], device='cuda:0'), grad: tensor([-1.7300e-05, -3.2838e-06, -1.0836e-04, 8.5831e-05, 1.7595e-04, + 3.9279e-05, -1.6749e-04, 1.4558e-05, 1.6153e-04, -1.8084e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 108---------------------------------------------------- +epoch 108, time 219.30, cls_loss 0.0023 cls_loss_mapping 0.0071 cls_loss_causal 0.5543 re_mapping 0.0064 re_causal 0.0190 /// teacc 99.19 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 0.0098, -0.0428, 0.0151, ..., -0.0755, -0.0712, -0.0458], + [-0.0708, 0.0378, -0.0417, ..., -0.0650, -0.0426, -0.1097], + [-0.0577, -0.0730, -0.0730, ..., -0.0885, -0.0422, 0.0985], + ..., + [ 0.0608, -0.0089, -0.0579, ..., 0.0562, -0.0403, -0.0356], + [ 0.0854, 0.0567, 0.0310, ..., -0.0599, -0.0704, 0.0565], + [ 0.0475, -0.0002, 0.0436, ..., 0.0395, -0.0582, 0.0185]], + device='cuda:0'), grad: tensor([[-3.4575e-07, 4.7265e-07, 2.2966e-06, ..., 6.3516e-07, + 2.1886e-06, 1.0394e-05], + [ 4.3795e-07, -3.4366e-06, -2.0862e-07, ..., 1.0237e-05, + 1.0610e-05, 6.6422e-06], + [ 2.0713e-06, 5.2666e-07, -1.0483e-05, ..., 2.3954e-06, + 1.5628e-06, -3.6865e-05], + ..., + [-5.9642e-06, 1.3355e-06, 3.1898e-07, ..., -4.2990e-06, + 1.3094e-06, 1.8086e-06], + [ 2.6682e-07, 1.2003e-05, 1.0446e-05, ..., 8.5682e-06, + 6.4112e-06, 2.2992e-05], + [ 1.6736e-06, -7.6219e-06, -6.5938e-06, ..., -6.0163e-06, + -1.7202e-06, -4.1425e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0280, -0.0150, -0.0117, 0.0175, -0.0031, -0.0133, -0.0021, 0.0134, + -0.0240, 0.0149], device='cuda:0'), grad: tensor([ 2.5257e-05, 6.0499e-06, -9.5904e-05, 2.3335e-05, -1.2659e-05, + 8.5980e-06, -1.2577e-05, -9.7156e-06, 7.0512e-05, -2.7921e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 218.76, cls_loss 0.0021 cls_loss_mapping 0.0066 cls_loss_causal 0.5626 re_mapping 0.0063 re_causal 0.0188 /// teacc 99.01 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0098, -0.0432, 0.0151, ..., -0.0757, -0.0721, -0.0465], + [-0.0705, 0.0391, -0.0419, ..., -0.0647, -0.0440, -0.1113], + [-0.0582, -0.0736, -0.0736, ..., -0.0892, -0.0422, 0.0999], + ..., + [ 0.0610, -0.0102, -0.0584, ..., 0.0564, -0.0407, -0.0363], + [ 0.0859, 0.0568, 0.0312, ..., -0.0605, -0.0711, 0.0565], + [ 0.0478, 0.0001, 0.0440, ..., 0.0399, -0.0579, 0.0187]], + device='cuda:0'), grad: tensor([[ 5.0180e-06, 5.2750e-06, 4.9509e-06, ..., 4.3362e-06, + 1.2955e-06, 1.4277e-06], + [ 2.9318e-06, 1.7583e-06, 2.6394e-06, ..., 3.3136e-06, + 1.1735e-06, 1.1958e-06], + [ 2.5686e-06, 2.8219e-06, 1.7053e-06, ..., 2.5090e-06, + 9.9000e-07, 1.2657e-06], + ..., + [ 1.3970e-05, 1.3769e-05, 1.2740e-05, ..., 1.1973e-05, + 2.8759e-06, 3.8408e-06], + [ 5.0038e-05, 4.0174e-05, 4.2945e-05, ..., 3.4213e-05, + 8.2031e-06, 8.4415e-06], + [-1.0490e-04, -8.1420e-05, -8.8453e-05, ..., -6.6400e-05, + -1.1697e-05, -1.7762e-05]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0278, -0.0151, -0.0106, 0.0179, -0.0036, -0.0138, -0.0016, 0.0131, + -0.0242, 0.0149], device='cuda:0'), grad: tensor([ 1.3001e-05, 7.6592e-06, 1.0066e-05, 2.0474e-05, 9.1419e-06, + 1.3247e-05, 2.4308e-06, 4.0233e-05, 1.1808e-04, -2.3448e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 218.91, cls_loss 0.0020 cls_loss_mapping 0.0054 cls_loss_causal 0.5576 re_mapping 0.0059 re_causal 0.0177 /// teacc 98.98 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0100, -0.0436, 0.0153, ..., -0.0758, -0.0728, -0.0469], + [-0.0710, 0.0393, -0.0424, ..., -0.0654, -0.0458, -0.1123], + [-0.0587, -0.0742, -0.0741, ..., -0.0901, -0.0432, 0.1004], + ..., + [ 0.0614, -0.0103, -0.0587, ..., 0.0569, -0.0410, -0.0369], + [ 0.0861, 0.0569, 0.0314, ..., -0.0611, -0.0715, 0.0568], + [ 0.0479, 0.0002, 0.0443, ..., 0.0399, -0.0583, 0.0185]], + device='cuda:0'), grad: tensor([[-1.9162e-07, 5.6066e-07, 4.9477e-10, ..., 1.6578e-07, + 4.8941e-07, 4.6939e-07], + [ 2.1271e-06, 1.5683e-06, 3.3621e-07, ..., 1.2275e-06, + 3.4482e-07, 7.8185e-07], + [ 7.3109e-08, 2.6524e-06, 3.0221e-07, ..., 3.4506e-07, + 4.6380e-07, -3.6322e-07], + ..., + [-6.2659e-06, 3.0268e-06, 2.1723e-07, ..., -8.4117e-06, + 1.7916e-07, 6.6776e-07], + [-2.0899e-06, -3.3863e-06, -1.3541e-06, ..., 4.1258e-07, + -6.8173e-07, -4.5672e-06], + [ 4.5300e-06, 3.0305e-06, 3.0571e-07, ..., 5.7444e-06, + 6.8964e-07, 1.4165e-06]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0281, -0.0155, -0.0108, 0.0179, -0.0037, -0.0137, -0.0011, 0.0135, + -0.0245, 0.0147], device='cuda:0'), grad: tensor([-1.6186e-06, 7.9349e-06, 3.2625e-08, -2.3261e-05, 4.8950e-06, + 6.0350e-06, -5.2014e-07, -3.6359e-06, -4.5039e-06, 1.4581e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 218.62, cls_loss 0.0021 cls_loss_mapping 0.0058 cls_loss_causal 0.5645 re_mapping 0.0057 re_causal 0.0178 /// teacc 98.97 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 0.0100, -0.0442, 0.0155, ..., -0.0759, -0.0737, -0.0479], + [-0.0716, 0.0388, -0.0430, ..., -0.0659, -0.0466, -0.1134], + [-0.0587, -0.0751, -0.0750, ..., -0.0906, -0.0439, 0.1011], + ..., + [ 0.0615, -0.0108, -0.0595, ..., 0.0573, -0.0412, -0.0377], + [ 0.0867, 0.0577, 0.0316, ..., -0.0615, -0.0719, 0.0573], + [ 0.0481, 0.0004, 0.0446, ..., 0.0400, -0.0585, 0.0184]], + device='cuda:0'), grad: tensor([[ 5.6403e-08, 5.0366e-06, 1.2433e-07, ..., 6.7567e-07, + 4.2142e-07, 2.0601e-06], + [ 5.1921e-07, -7.5877e-05, 2.7008e-07, ..., -2.6934e-06, + 2.3330e-07, -1.7926e-05], + [ 3.7961e-06, 1.2301e-05, 2.2654e-07, ..., 2.0206e-05, + 1.3914e-06, -1.5898e-06], + ..., + [-1.2152e-05, -7.3463e-06, 7.1130e-08, ..., -6.1095e-05, + -4.1351e-06, 2.7791e-06], + [-1.0710e-06, 2.8059e-05, -2.9728e-06, ..., 2.1867e-06, + 8.8126e-08, 5.0813e-06], + [ 7.4646e-07, 1.4395e-05, 1.2468e-07, ..., 4.0978e-06, + 7.0455e-07, 3.9712e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0282, -0.0163, -0.0112, 0.0186, -0.0038, -0.0134, -0.0010, 0.0134, + -0.0242, 0.0147], device='cuda:0'), grad: tensor([ 1.5087e-05, -1.4532e-04, 3.7730e-05, 6.0171e-05, 3.3647e-05, + -2.1189e-05, 1.4804e-05, -8.8274e-05, 5.3108e-05, 4.0054e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 218.45, cls_loss 0.0019 cls_loss_mapping 0.0057 cls_loss_causal 0.5633 re_mapping 0.0061 re_causal 0.0187 /// teacc 99.07 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0102, -0.0447, 0.0158, ..., -0.0760, -0.0747, -0.0486], + [-0.0717, 0.0394, -0.0433, ..., -0.0664, -0.0475, -0.1140], + [-0.0591, -0.0755, -0.0751, ..., -0.0912, -0.0444, 0.1018], + ..., + [ 0.0620, -0.0111, -0.0597, ..., 0.0581, -0.0413, -0.0381], + [ 0.0871, 0.0578, 0.0319, ..., -0.0622, -0.0724, 0.0578], + [ 0.0480, 0.0003, 0.0446, ..., 0.0399, -0.0588, 0.0182]], + device='cuda:0'), grad: tensor([[ 6.1467e-08, 1.3290e-06, 2.0105e-07, ..., 3.5088e-07, + 6.4308e-07, 1.4696e-06], + [ 2.9169e-06, 2.4419e-06, 7.2084e-07, ..., 4.5337e-06, + 1.7683e-07, 1.3914e-06], + [ 3.8184e-06, 4.4331e-06, 8.5961e-07, ..., 6.0424e-06, + 3.1479e-07, -2.7984e-05], + ..., + [-1.3232e-05, 8.6948e-06, 1.0980e-06, ..., -2.0683e-05, + 2.0850e-07, 6.3097e-07], + [-9.3644e-07, 3.4440e-06, -2.6147e-07, ..., 2.2482e-06, + -1.5320e-06, 1.3731e-05], + [ 3.3621e-06, 4.4405e-06, 1.2722e-06, ..., 2.8349e-06, + 1.7872e-06, 3.5707e-06]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0288, -0.0161, -0.0111, 0.0184, -0.0035, -0.0138, -0.0007, 0.0138, + -0.0240, 0.0138], device='cuda:0'), grad: tensor([ 4.6678e-06, 1.7092e-05, -2.0862e-05, -5.3614e-05, 1.1168e-05, + 6.2101e-06, 3.4329e-06, -2.4632e-05, 4.1068e-05, 1.5348e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 218.53, cls_loss 0.0020 cls_loss_mapping 0.0055 cls_loss_causal 0.5529 re_mapping 0.0061 re_causal 0.0178 /// teacc 99.09 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 0.0105, -0.0454, 0.0159, ..., -0.0761, -0.0749, -0.0485], + [-0.0728, 0.0393, -0.0436, ..., -0.0673, -0.0484, -0.1148], + [-0.0594, -0.0759, -0.0754, ..., -0.0919, -0.0447, 0.1025], + ..., + [ 0.0621, -0.0113, -0.0600, ..., 0.0585, -0.0417, -0.0388], + [ 0.0876, 0.0577, 0.0319, ..., -0.0628, -0.0729, 0.0578], + [ 0.0483, 0.0005, 0.0448, ..., 0.0401, -0.0592, 0.0181]], + device='cuda:0'), grad: tensor([[ 1.6689e-06, 6.3470e-07, 3.2107e-07, ..., 2.0303e-06, + 1.5404e-06, 1.1744e-06], + [ 4.1872e-06, 5.3085e-07, 4.4703e-07, ..., 3.6377e-06, + 6.6357e-07, 5.8720e-07], + [ 8.8066e-06, 7.7998e-07, 4.0699e-07, ..., 7.4394e-06, + 8.5542e-07, 3.7393e-07], + ..., + [-4.8578e-05, 1.6242e-06, -2.3786e-06, ..., -4.7088e-05, + 1.0501e-07, 4.7334e-07], + [ 7.0371e-06, 3.5036e-06, 1.6158e-06, ..., 7.7784e-06, + 4.8466e-06, 3.1590e-06], + [ 8.1509e-06, -7.5810e-06, -4.7125e-06, ..., 8.8736e-06, + 9.3179e-07, -9.0944e-07]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0291, -0.0167, -0.0106, 0.0178, -0.0035, -0.0136, -0.0002, 0.0138, + -0.0243, 0.0139], device='cuda:0'), grad: tensor([ 8.0839e-06, 1.3754e-05, 2.6375e-05, 1.0200e-05, 7.1824e-06, + 4.4048e-05, -2.4199e-05, -1.3435e-04, 3.2753e-05, 1.6049e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 218.47, cls_loss 0.0019 cls_loss_mapping 0.0065 cls_loss_causal 0.5491 re_mapping 0.0063 re_causal 0.0182 /// teacc 99.16 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0113, -0.0455, 0.0167, ..., -0.0763, -0.0754, -0.0483], + [-0.0735, 0.0395, -0.0440, ..., -0.0679, -0.0491, -0.1152], + [-0.0601, -0.0765, -0.0756, ..., -0.0927, -0.0451, 0.1027], + ..., + [ 0.0624, -0.0114, -0.0605, ..., 0.0590, -0.0419, -0.0393], + [ 0.0879, 0.0582, 0.0321, ..., -0.0633, -0.0732, 0.0582], + [ 0.0485, 0.0007, 0.0449, ..., 0.0403, -0.0595, 0.0178]], + device='cuda:0'), grad: tensor([[ 2.1588e-06, 2.1663e-06, 4.3609e-07, ..., 2.7288e-07, + 1.7742e-07, 1.7937e-06], + [-1.3912e-07, 2.7139e-06, 1.1763e-06, ..., 4.8988e-07, + 4.1956e-07, 1.0836e-04], + [-8.4657e-07, 4.6343e-06, 1.0505e-06, ..., -3.1386e-07, + 2.7055e-07, -1.3113e-04], + ..., + [ 2.9756e-07, 5.4650e-06, 1.0375e-06, ..., 2.8987e-07, + 2.2631e-07, 2.6841e-06], + [ 2.0787e-06, 3.4235e-06, 1.4156e-06, ..., 3.7514e-06, + 1.4054e-06, 5.2564e-06], + [-2.8368e-06, 5.3784e-07, -2.3618e-06, ..., -3.2317e-06, + -8.3994e-08, -9.6112e-07]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0309, -0.0170, -0.0112, 0.0174, -0.0036, -0.0135, -0.0007, 0.0143, + -0.0244, 0.0135], device='cuda:0'), grad: tensor([ 2.3499e-05, 3.1662e-04, -3.6526e-04, -5.1200e-05, 3.8557e-06, + -3.6925e-05, 5.6148e-05, 1.9804e-05, 2.7999e-05, 5.0403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 218.73, cls_loss 0.0020 cls_loss_mapping 0.0064 cls_loss_causal 0.5571 re_mapping 0.0061 re_causal 0.0175 /// teacc 99.09 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 0.0117, -0.0463, 0.0168, ..., -0.0764, -0.0762, -0.0487], + [-0.0737, 0.0398, -0.0444, ..., -0.0683, -0.0501, -0.1162], + [-0.0595, -0.0770, -0.0759, ..., -0.0934, -0.0457, 0.1037], + ..., + [ 0.0623, -0.0118, -0.0609, ..., 0.0593, -0.0428, -0.0407], + [ 0.0884, 0.0589, 0.0324, ..., -0.0641, -0.0735, 0.0588], + [ 0.0486, 0.0008, 0.0454, ..., 0.0403, -0.0599, 0.0173]], + device='cuda:0'), grad: tensor([[ 8.5235e-06, 9.4436e-07, -2.7823e-07, ..., 1.9874e-06, + 5.5181e-07, 6.9514e-06], + [ 8.2180e-06, -6.2864e-07, 6.2678e-07, ..., 4.1164e-06, + 4.4284e-07, 6.7763e-06], + [-7.7128e-05, 1.5385e-06, 1.2228e-06, ..., -4.1991e-05, + 6.0769e-07, -6.5386e-05], + ..., + [ 5.0992e-05, 1.8282e-06, 7.1712e-07, ..., 2.5317e-05, + 5.1782e-07, 3.8564e-05], + [-4.9248e-06, -9.7379e-06, -7.1600e-06, ..., 3.4850e-06, + 6.0303e-07, -7.9796e-06], + [ 8.2776e-06, 2.9430e-06, 6.7893e-07, ..., 7.3500e-06, + 3.6173e-06, 6.8843e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0308, -0.0171, -0.0105, 0.0171, -0.0025, -0.0135, -0.0011, 0.0138, + -0.0243, 0.0132], device='cuda:0'), grad: tensor([ 5.6565e-05, 3.4004e-05, -1.9240e-04, 9.3758e-05, 4.3400e-06, + -3.3188e-04, 3.8654e-05, 2.1291e-04, 3.5554e-05, 4.7535e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 218.74, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5619 re_mapping 0.0058 re_causal 0.0179 /// teacc 99.08 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 0.0113, -0.0471, 0.0166, ..., -0.0773, -0.0774, -0.0496], + [-0.0743, 0.0399, -0.0448, ..., -0.0688, -0.0510, -0.1168], + [-0.0600, -0.0776, -0.0764, ..., -0.0938, -0.0462, 0.1038], + ..., + [ 0.0626, -0.0122, -0.0614, ..., 0.0595, -0.0433, -0.0410], + [ 0.0891, 0.0595, 0.0330, ..., -0.0648, -0.0737, 0.0594], + [ 0.0491, 0.0009, 0.0457, ..., 0.0408, -0.0600, 0.0172]], + device='cuda:0'), grad: tensor([[ 7.9393e-05, 6.1467e-07, 9.2834e-06, ..., 9.4831e-05, + 1.8382e-07, 8.7777e-07], + [ 2.5537e-06, -4.1281e-07, 5.8115e-07, ..., 3.5744e-06, + 5.1688e-08, 7.0361e-07], + [-3.5372e-06, 1.4147e-06, 6.4634e-07, ..., -1.1120e-06, + 5.9837e-08, -1.1317e-05], + ..., + [ 4.5337e-06, 1.2331e-06, 1.4799e-06, ..., 3.2019e-06, + 2.3516e-08, 7.2010e-06], + [ 4.4890e-06, 1.3392e-06, 2.5239e-06, ..., 4.4629e-06, + 7.1246e-08, 1.1921e-06], + [-1.0204e-04, -3.2056e-06, -1.6361e-05, ..., -1.1939e-04, + 5.5035e-08, -1.4883e-06]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0302, -0.0174, -0.0108, 0.0172, -0.0021, -0.0130, -0.0018, 0.0138, + -0.0241, 0.0134], device='cuda:0'), grad: tensor([ 1.5593e-04, 6.0871e-06, -2.6271e-05, -4.0140e-07, 2.2218e-05, + 4.5598e-06, 1.7630e-06, 1.8701e-05, 1.5438e-05, -1.9801e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 218.87, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5358 re_mapping 0.0064 re_causal 0.0187 /// teacc 99.02 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0114, -0.0475, 0.0166, ..., -0.0774, -0.0793, -0.0508], + [-0.0750, 0.0401, -0.0451, ..., -0.0693, -0.0529, -0.1181], + [-0.0611, -0.0791, -0.0766, ..., -0.0949, -0.0463, 0.1043], + ..., + [ 0.0629, -0.0128, -0.0617, ..., 0.0600, -0.0447, -0.0422], + [ 0.0894, 0.0598, 0.0333, ..., -0.0657, -0.0742, 0.0597], + [ 0.0495, 0.0012, 0.0464, ..., 0.0408, -0.0605, 0.0170]], + device='cuda:0'), grad: tensor([[-3.8296e-06, 9.9360e-08, -1.7174e-06, ..., 1.4668e-07, + 4.3102e-06, 4.5635e-06], + [ 6.6822e-07, 7.2760e-11, 3.8836e-07, ..., 1.6752e-07, + 3.3062e-07, 5.6904e-07], + [ 1.5153e-06, 1.7183e-07, 8.0373e-07, ..., 4.1071e-07, + 1.1129e-06, -3.7532e-07], + ..., + [-1.0710e-06, 1.6880e-07, 1.4948e-07, ..., -1.1260e-06, + 1.2992e-07, 3.0780e-07], + [ 1.7276e-06, 4.7934e-08, 1.0878e-06, ..., 1.7276e-07, + 2.3544e-06, 2.8796e-06], + [ 1.1232e-06, 4.6653e-08, 3.8208e-07, ..., 5.8953e-07, + 6.2818e-07, 8.0513e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0300, -0.0179, -0.0118, 0.0192, -0.0024, -0.0132, -0.0014, 0.0133, + -0.0243, 0.0134], device='cuda:0'), grad: tensor([-5.5917e-06, 3.1572e-06, 1.5590e-06, 1.1668e-05, -1.6391e-06, + -8.4266e-06, -1.5453e-05, -1.4994e-06, 1.0833e-05, 5.3719e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 218.89, cls_loss 0.0016 cls_loss_mapping 0.0068 cls_loss_causal 0.5725 re_mapping 0.0060 re_causal 0.0187 /// teacc 99.07 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0115, -0.0480, 0.0165, ..., -0.0776, -0.0803, -0.0517], + [-0.0752, 0.0412, -0.0447, ..., -0.0696, -0.0535, -0.1181], + [-0.0612, -0.0796, -0.0771, ..., -0.0958, -0.0468, 0.1048], + ..., + [ 0.0632, -0.0133, -0.0621, ..., 0.0605, -0.0451, -0.0433], + [ 0.0897, 0.0595, 0.0333, ..., -0.0662, -0.0748, 0.0600], + [ 0.0498, 0.0015, 0.0468, ..., 0.0409, -0.0607, 0.0169]], + device='cuda:0'), grad: tensor([[-1.2843e-06, 1.1381e-06, 2.0803e-07, ..., 1.4417e-06, + -1.7672e-07, -5.9744e-07], + [ 1.2545e-06, 1.3821e-05, 7.8790e-07, ..., 2.1324e-05, + 2.0102e-05, 1.3635e-05], + [ 1.6717e-06, 8.1584e-07, 4.7404e-07, ..., 3.5129e-06, + 1.8701e-06, -4.9584e-06], + ..., + [-6.4448e-06, 2.2631e-06, 6.2445e-07, ..., -7.4431e-06, + 2.5984e-06, 2.4047e-06], + [-5.4128e-06, -2.5779e-06, -1.9968e-06, ..., 1.7844e-06, + 1.6373e-06, -1.9558e-06], + [ 2.2836e-06, 8.2403e-06, 5.2573e-07, ..., 4.9859e-05, + 4.3631e-05, 3.4332e-05]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0299, -0.0172, -0.0123, 0.0189, -0.0025, -0.0134, -0.0012, 0.0139, + -0.0246, 0.0134], device='cuda:0'), grad: tensor([-9.7007e-06, 6.0707e-05, -9.0199e-07, 1.2629e-05, -1.8787e-04, + -6.0303e-07, 3.3826e-05, -1.1936e-05, -3.0305e-06, 1.0681e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 218.50, cls_loss 0.0017 cls_loss_mapping 0.0061 cls_loss_causal 0.5789 re_mapping 0.0058 re_causal 0.0186 /// teacc 99.07 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0117, -0.0483, 0.0165, ..., -0.0776, -0.0813, -0.0522], + [-0.0760, 0.0415, -0.0449, ..., -0.0703, -0.0546, -0.1188], + [-0.0620, -0.0800, -0.0773, ..., -0.0964, -0.0469, 0.1055], + ..., + [ 0.0642, -0.0136, -0.0625, ..., 0.0611, -0.0458, -0.0438], + [ 0.0902, 0.0596, 0.0335, ..., -0.0667, -0.0753, 0.0602], + [ 0.0495, 0.0018, 0.0473, ..., 0.0407, -0.0612, 0.0167]], + device='cuda:0'), grad: tensor([[ 3.6135e-07, 4.8010e-07, 2.7404e-07, ..., 8.0606e-07, + 1.6093e-05, 1.0826e-05], + [ 1.1893e-06, 8.7684e-07, 5.9558e-07, ..., 9.8627e-07, + 3.0976e-06, 3.4403e-06], + [ 3.4403e-06, 2.7250e-06, 1.0412e-06, ..., 6.2725e-07, + 1.9465e-06, 5.6438e-07], + ..., + [ 7.4366e-07, 9.4343e-07, 4.1118e-07, ..., 5.0180e-06, + 4.0680e-06, 3.1162e-06], + [-1.3895e-05, -1.1869e-05, -5.5730e-06, ..., 2.6380e-07, + -1.5888e-06, -1.2204e-05], + [ 1.4743e-06, 1.8394e-06, 3.6997e-07, ..., 6.9402e-06, + 6.6161e-06, 4.5337e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0302, -0.0176, -0.0123, 0.0185, -0.0022, -0.0138, 0.0003, 0.0145, + -0.0248, 0.0129], device='cuda:0'), grad: tensor([ 2.9087e-05, 1.0699e-05, -4.7944e-06, 3.9302e-06, 1.4886e-05, + 1.2241e-05, -6.0976e-05, 1.2904e-05, -3.5495e-05, 1.7628e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 218.55, cls_loss 0.0019 cls_loss_mapping 0.0051 cls_loss_causal 0.5686 re_mapping 0.0057 re_causal 0.0181 /// teacc 99.06 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0119, -0.0488, 0.0166, ..., -0.0778, -0.0818, -0.0526], + [-0.0766, 0.0423, -0.0455, ..., -0.0707, -0.0551, -0.1182], + [-0.0620, -0.0814, -0.0777, ..., -0.0967, -0.0477, 0.1061], + ..., + [ 0.0647, -0.0140, -0.0628, ..., 0.0614, -0.0466, -0.0444], + [ 0.0909, 0.0586, 0.0328, ..., -0.0670, -0.0755, 0.0597], + [ 0.0496, 0.0021, 0.0482, ..., 0.0416, -0.0596, 0.0173]], + device='cuda:0'), grad: tensor([[ 1.9011e-07, 3.5740e-07, -9.5218e-06, ..., 4.5123e-07, + 3.4482e-07, -1.6153e-05], + [ 1.2182e-06, -1.0934e-06, 1.1576e-06, ..., 1.0459e-06, + 3.0850e-07, 2.0172e-06], + [ 6.7940e-07, 3.6997e-07, 4.1798e-06, ..., 3.9185e-07, + 3.0198e-07, 5.1074e-06], + ..., + [-1.7043e-06, 6.7195e-07, 7.3342e-07, ..., -2.0228e-06, + 3.9279e-07, 9.2573e-07], + [-3.1143e-06, -2.0005e-06, -2.1630e-07, ..., -2.5076e-07, + -8.7777e-07, 2.3062e-07], + [-1.9260e-06, -1.2564e-06, -2.1514e-06, ..., -2.0247e-06, + 3.1572e-07, 7.4320e-07]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0303, -0.0173, -0.0120, 0.0184, -0.0030, -0.0127, -0.0004, 0.0146, + -0.0258, 0.0133], device='cuda:0'), grad: tensor([-1.4961e-04, 1.1340e-05, 5.8919e-05, 1.4961e-05, 9.3058e-06, + 5.6066e-06, 3.6150e-05, 9.0164e-08, 1.5274e-05, -2.0955e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 218.71, cls_loss 0.0024 cls_loss_mapping 0.0063 cls_loss_causal 0.5215 re_mapping 0.0057 re_causal 0.0171 /// teacc 99.00 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0109, -0.0506, 0.0156, ..., -0.0788, -0.0825, -0.0524], + [-0.0777, 0.0423, -0.0462, ..., -0.0715, -0.0564, -0.1193], + [-0.0624, -0.0820, -0.0782, ..., -0.0977, -0.0485, 0.1065], + ..., + [ 0.0651, -0.0143, -0.0637, ..., 0.0619, -0.0469, -0.0449], + [ 0.0917, 0.0590, 0.0333, ..., -0.0678, -0.0760, 0.0604], + [ 0.0508, 0.0034, 0.0500, ..., 0.0427, -0.0595, 0.0173]], + device='cuda:0'), grad: tensor([[-1.9721e-07, 1.5891e-07, 7.9512e-08, ..., 1.6997e-07, + 9.6206e-07, 1.0207e-06], + [ 3.2806e-07, -1.2089e-06, 9.8837e-08, ..., 1.0459e-06, + 5.9931e-07, 4.3819e-07], + [ 3.3434e-06, 4.9546e-07, 8.9756e-08, ..., 6.0759e-06, + 2.5379e-07, 6.4168e-07], + ..., + [-5.3532e-06, -4.5123e-07, 5.9197e-08, ..., -1.0550e-05, + 4.2259e-07, -7.4413e-07], + [-5.9186e-07, 4.4741e-06, 3.7365e-06, ..., 3.2363e-07, + 6.0862e-07, 9.9838e-07], + [ 5.2527e-07, 6.6124e-07, 2.3248e-07, ..., 1.8766e-06, + 1.2796e-06, 8.3679e-07]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0293, -0.0182, -0.0120, 0.0179, -0.0036, -0.0120, -0.0003, 0.0149, + -0.0260, 0.0146], device='cuda:0'), grad: tensor([ 1.7891e-06, -9.8348e-07, 1.0096e-05, 2.7731e-05, -3.8091e-06, + -4.8578e-05, 1.2003e-05, -1.6659e-05, 1.3806e-05, 4.6045e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 218.52, cls_loss 0.0024 cls_loss_mapping 0.0074 cls_loss_causal 0.5726 re_mapping 0.0055 re_causal 0.0170 /// teacc 98.98 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0111, -0.0507, 0.0157, ..., -0.0786, -0.0860, -0.0555], + [-0.0788, 0.0425, -0.0466, ..., -0.0724, -0.0576, -0.1207], + [-0.0627, -0.0826, -0.0788, ..., -0.0983, -0.0488, 0.1077], + ..., + [ 0.0655, -0.0144, -0.0641, ..., 0.0622, -0.0478, -0.0453], + [ 0.0929, 0.0595, 0.0338, ..., -0.0681, -0.0765, 0.0610], + [ 0.0508, 0.0038, 0.0504, ..., 0.0429, -0.0589, 0.0179]], + device='cuda:0'), grad: tensor([[ 1.5378e-05, 2.8568e-07, 1.7568e-05, ..., 1.2115e-05, + 4.6939e-06, 1.6361e-05], + [ 4.5518e-07, 3.4762e-07, 1.9022e-07, ..., 9.7901e-06, + 1.1936e-05, 4.0606e-06], + [ 1.9930e-06, 5.3504e-07, 7.2503e-07, ..., 3.4608e-06, + 2.0918e-06, -1.5181e-06], + ..., + [-4.3064e-06, 1.2657e-06, 6.3423e-07, ..., 2.1756e-06, + 7.4618e-06, 2.9560e-06], + [ 1.5292e-06, 4.1677e-07, 1.2089e-06, ..., 2.6301e-06, + 2.7083e-06, 3.5260e-06], + [-2.1636e-05, 6.8806e-06, -2.6107e-05, ..., 4.5896e-05, + 7.3731e-05, 2.3898e-06]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0291, -0.0188, -0.0118, 0.0197, -0.0046, -0.0141, 0.0007, 0.0154, + -0.0257, 0.0147], device='cuda:0'), grad: tensor([ 4.7982e-05, 2.1055e-05, 4.0308e-06, 3.6471e-06, -2.4629e-04, + 3.7462e-05, 3.6806e-05, 4.8131e-06, 1.2621e-05, 7.7963e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 218.59, cls_loss 0.0018 cls_loss_mapping 0.0053 cls_loss_causal 0.5244 re_mapping 0.0058 re_causal 0.0171 /// teacc 99.05 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0111, -0.0510, 0.0157, ..., -0.0786, -0.0863, -0.0561], + [-0.0789, 0.0429, -0.0469, ..., -0.0730, -0.0591, -0.1214], + [-0.0631, -0.0830, -0.0792, ..., -0.0990, -0.0492, 0.1080], + ..., + [ 0.0659, -0.0153, -0.0648, ..., 0.0622, -0.0488, -0.0458], + [ 0.0935, 0.0608, 0.0350, ..., -0.0685, -0.0768, 0.0619], + [ 0.0509, 0.0037, 0.0508, ..., 0.0428, -0.0598, 0.0170]], + device='cuda:0'), grad: tensor([[ 3.1153e-07, 1.3653e-06, 6.9477e-07, ..., 3.4715e-07, + 5.7369e-07, 4.2398e-07], + [ 4.7591e-07, 2.7921e-06, 1.2787e-06, ..., 4.1211e-07, + 4.0489e-07, 3.8627e-07], + [ 6.8266e-07, 5.6922e-06, 2.3711e-06, ..., 3.8976e-07, + 3.3621e-07, -2.6333e-07], + ..., + [-1.7295e-06, 5.6289e-06, 2.6226e-06, ..., -1.7975e-06, + 9.1677e-08, 1.6415e-07], + [ 1.6382e-06, 1.3560e-05, 5.8524e-06, ..., 1.0282e-06, + 2.6990e-06, 1.8897e-06], + [-2.0079e-06, -1.3057e-06, -2.4084e-06, ..., -2.6617e-06, + 5.3784e-07, -2.5216e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0291, -0.0187, -0.0118, 0.0190, -0.0038, -0.0138, 0.0010, 0.0154, + -0.0252, 0.0145], device='cuda:0'), grad: tensor([ 4.3772e-06, 7.8380e-06, 1.3329e-05, -7.2420e-05, 3.6675e-06, + 1.0185e-05, -8.2254e-06, 6.7279e-06, 3.6567e-05, -2.0564e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 218.44, cls_loss 0.0018 cls_loss_mapping 0.0053 cls_loss_causal 0.5485 re_mapping 0.0056 re_causal 0.0172 /// teacc 99.08 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0111, -0.0514, 0.0157, ..., -0.0786, -0.0870, -0.0567], + [-0.0798, 0.0424, -0.0483, ..., -0.0742, -0.0614, -0.1234], + [-0.0625, -0.0831, -0.0795, ..., -0.0995, -0.0501, 0.1091], + ..., + [ 0.0659, -0.0157, -0.0652, ..., 0.0643, -0.0498, -0.0466], + [ 0.0940, 0.0612, 0.0353, ..., -0.0693, -0.0775, 0.0621], + [ 0.0510, 0.0046, 0.0511, ..., 0.0430, -0.0600, 0.0169]], + device='cuda:0'), grad: tensor([[-2.4773e-07, 9.9558e-07, 6.5425e-07, ..., 3.3039e-07, + 1.2405e-06, 2.4885e-06], + [ 4.6170e-07, 5.7556e-07, 1.2573e-06, ..., 3.5344e-07, + 3.6717e-07, 8.2701e-07], + [ 2.2016e-06, 3.0342e-06, 3.2205e-06, ..., 4.0117e-07, + 1.3718e-06, -5.0738e-06], + ..., + [ 1.2424e-06, 3.7868e-06, 3.8706e-06, ..., -7.0408e-07, + 4.6729e-07, 1.5246e-06], + [-6.7204e-06, -6.6403e-07, 2.1514e-06, ..., 5.2992e-07, + 6.7241e-07, -1.7835e-06], + [ 1.5330e-06, 3.8326e-05, 7.1712e-06, ..., 1.2004e-04, + 8.2791e-05, 8.2791e-05]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0291, -0.0201, -0.0109, 0.0189, -0.0048, -0.0140, 0.0020, 0.0162, + -0.0254, 0.0145], device='cuda:0'), grad: tensor([ 8.4937e-06, 4.9546e-06, -1.4715e-06, -6.0946e-05, -2.1970e-04, + 1.6913e-05, -1.0334e-05, 1.9357e-05, 1.5602e-05, 2.2686e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 218.56, cls_loss 0.0015 cls_loss_mapping 0.0046 cls_loss_causal 0.5174 re_mapping 0.0055 re_causal 0.0165 /// teacc 99.02 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0112, -0.0519, 0.0157, ..., -0.0787, -0.0873, -0.0570], + [-0.0800, 0.0427, -0.0486, ..., -0.0744, -0.0621, -0.1241], + [-0.0628, -0.0834, -0.0798, ..., -0.1005, -0.0501, 0.1103], + ..., + [ 0.0662, -0.0165, -0.0661, ..., 0.0643, -0.0520, -0.0475], + [ 0.0943, 0.0613, 0.0356, ..., -0.0702, -0.0775, 0.0622], + [ 0.0511, 0.0048, 0.0513, ..., 0.0430, -0.0605, 0.0164]], + device='cuda:0'), grad: tensor([[ 6.0303e-07, 4.2259e-07, 2.9104e-07, ..., 9.6485e-07, + 1.2130e-07, 1.3178e-07], + [ 6.9104e-07, -1.7718e-07, 2.8615e-07, ..., 1.1865e-06, + 1.6252e-07, 1.2503e-07], + [ 9.2899e-07, 2.6822e-07, 1.3574e-07, ..., 1.0198e-06, + 1.1642e-07, -2.1234e-07], + ..., + [-6.4112e-06, 2.9467e-06, 1.9539e-06, ..., -4.8988e-06, + 5.1083e-07, 2.0349e-07], + [-1.4603e-06, -1.2219e-06, -1.6168e-06, ..., 4.3004e-07, + 4.1467e-07, -9.2853e-07], + [-4.3549e-06, -1.2584e-05, -8.0988e-06, ..., -1.7047e-05, + -1.7621e-06, 9.0292e-07]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0292, -0.0202, -0.0105, 0.0191, -0.0042, -0.0140, 0.0016, 0.0160, + -0.0257, 0.0144], device='cuda:0'), grad: tensor([ 1.7723e-06, 7.0501e-07, 8.8895e-07, 1.2353e-05, 2.0400e-05, + 8.8569e-07, -2.7649e-09, -7.8455e-06, -1.4883e-06, -2.7657e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 218.73, cls_loss 0.0018 cls_loss_mapping 0.0056 cls_loss_causal 0.5440 re_mapping 0.0054 re_causal 0.0164 /// teacc 98.93 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0112, -0.0526, 0.0156, ..., -0.0787, -0.0883, -0.0577], + [-0.0804, 0.0428, -0.0490, ..., -0.0750, -0.0635, -0.1248], + [-0.0636, -0.0839, -0.0800, ..., -0.1006, -0.0505, 0.1114], + ..., + [ 0.0663, -0.0170, -0.0669, ..., 0.0643, -0.0527, -0.0484], + [ 0.0955, 0.0619, 0.0364, ..., -0.0708, -0.0780, 0.0626], + [ 0.0511, 0.0051, 0.0515, ..., 0.0432, -0.0605, 0.0164]], + device='cuda:0'), grad: tensor([[-4.0350e-07, 5.6298e-07, 1.8772e-08, ..., 2.0105e-07, + 1.7285e-06, 1.4184e-06], + [-1.1944e-07, -1.1874e-08, 4.0792e-07, ..., 2.6682e-07, + 7.8045e-07, 4.8010e-07], + [ 3.6834e-07, 2.7064e-06, 9.4669e-07, ..., 8.8336e-07, + 8.0792e-07, 5.0478e-07], + ..., + [-7.7905e-07, 2.2724e-06, 5.2340e-07, ..., 1.0403e-06, + 1.8356e-06, 8.6892e-07], + [ 1.0151e-07, 5.4687e-06, 2.1327e-06, ..., 3.6671e-07, + 1.9968e-06, 1.6941e-06], + [ 2.2189e-07, 1.9008e-06, 2.0058e-07, ..., 2.7083e-06, + 2.6692e-06, 1.3625e-06]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0291, -0.0204, -0.0096, 0.0188, -0.0042, -0.0140, 0.0017, 0.0156, + -0.0257, 0.0145], device='cuda:0'), grad: tensor([ 1.4119e-06, -5.3551e-07, 7.8306e-06, -2.4527e-05, -8.6725e-06, + 2.8759e-06, -5.1856e-06, 4.7386e-06, 1.3582e-05, 8.4490e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 218.67, cls_loss 0.0015 cls_loss_mapping 0.0042 cls_loss_causal 0.5171 re_mapping 0.0056 re_causal 0.0163 /// teacc 98.95 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0113, -0.0530, 0.0156, ..., -0.0787, -0.0895, -0.0588], + [-0.0797, 0.0441, -0.0492, ..., -0.0749, -0.0637, -0.1251], + [-0.0640, -0.0846, -0.0804, ..., -0.1013, -0.0510, 0.1120], + ..., + [ 0.0667, -0.0178, -0.0673, ..., 0.0646, -0.0530, -0.0489], + [ 0.0958, 0.0619, 0.0366, ..., -0.0713, -0.0786, 0.0626], + [ 0.0511, 0.0050, 0.0516, ..., 0.0431, -0.0608, 0.0159]], + device='cuda:0'), grad: tensor([[-4.6492e-06, 1.3690e-07, -1.5162e-06, ..., 1.9744e-07, + 1.6904e-07, 9.7207e-08], + [ 1.4016e-07, -3.6485e-07, 1.3539e-07, ..., 2.2817e-07, + 1.7788e-07, 2.1781e-07], + [ 3.2503e-07, 2.5984e-07, 1.4738e-07, ..., 2.0256e-07, + 1.8498e-07, 1.5006e-07], + ..., + [ 7.5670e-07, 3.9465e-07, 3.8603e-07, ..., 2.3004e-07, + 2.0652e-07, 2.5518e-07], + [-9.9465e-07, -1.0878e-06, -4.1421e-07, ..., 3.3481e-07, + 1.9139e-07, -1.6131e-06], + [ 1.3150e-06, 3.5949e-06, 1.7928e-07, ..., 9.3281e-06, + 6.7465e-06, 6.7204e-06]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0291, -0.0194, -0.0097, 0.0187, -0.0040, -0.0138, 0.0017, 0.0155, + -0.0260, 0.0143], device='cuda:0'), grad: tensor([-7.4357e-06, -9.6625e-08, 1.0934e-06, 1.4249e-06, -1.9535e-05, + 2.2575e-06, 1.2033e-06, 2.1551e-06, -6.3516e-07, 1.9535e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 218.53, cls_loss 0.0025 cls_loss_mapping 0.0068 cls_loss_causal 0.5335 re_mapping 0.0053 re_causal 0.0157 /// teacc 98.99 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0113, -0.0535, 0.0155, ..., -0.0788, -0.0897, -0.0593], + [-0.0790, 0.0445, -0.0498, ..., -0.0752, -0.0643, -0.1259], + [-0.0632, -0.0883, -0.0796, ..., -0.1022, -0.0518, 0.1129], + ..., + [ 0.0661, -0.0187, -0.0681, ..., 0.0643, -0.0540, -0.0506], + [ 0.0958, 0.0648, 0.0361, ..., -0.0726, -0.0789, 0.0628], + [ 0.0513, 0.0061, 0.0523, ..., 0.0434, -0.0611, 0.0157]], + device='cuda:0'), grad: tensor([[-4.2804e-06, 3.8708e-08, -7.4357e-06, ..., -2.3487e-08, + 2.9540e-08, -2.9709e-06], + [ 3.5390e-07, -1.5204e-07, 3.6717e-07, ..., 1.4785e-07, + 2.8609e-08, 9.1619e-08], + [ 5.8580e-07, 8.2771e-08, 6.6776e-07, ..., 9.4180e-08, + 2.6979e-08, -1.7136e-07], + ..., + [ 3.7346e-07, -1.1059e-08, 1.1716e-06, ..., -1.6075e-06, + -1.7171e-08, 6.1817e-08], + [ 7.8557e-07, 1.4401e-07, 1.0738e-06, ..., 3.1141e-08, + 2.0082e-08, 2.5332e-07], + [ 3.3557e-05, 1.1234e-07, 3.3528e-05, ..., 6.0024e-07, + 1.3923e-07, 4.9081e-07]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0291, -0.0186, -0.0088, 0.0182, -0.0036, -0.0139, 0.0012, 0.0146, + -0.0264, 0.0145], device='cuda:0'), grad: tensor([-2.4945e-05, 1.8477e-06, 2.6338e-06, 1.7971e-05, 4.1462e-06, + -2.2948e-04, 2.0072e-05, 4.4778e-06, 5.5917e-06, 1.9765e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 218.42, cls_loss 0.0014 cls_loss_mapping 0.0055 cls_loss_causal 0.5372 re_mapping 0.0055 re_causal 0.0168 /// teacc 99.02 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0113, -0.0538, 0.0155, ..., -0.0788, -0.0900, -0.0592], + [-0.0794, 0.0446, -0.0503, ..., -0.0759, -0.0647, -0.1264], + [-0.0632, -0.0884, -0.0798, ..., -0.1034, -0.0522, 0.1132], + ..., + [ 0.0664, -0.0187, -0.0683, ..., 0.0650, -0.0545, -0.0513], + [ 0.0960, 0.0650, 0.0364, ..., -0.0731, -0.0786, 0.0630], + [ 0.0513, 0.0062, 0.0524, ..., 0.0433, -0.0614, 0.0154]], + device='cuda:0'), grad: tensor([[-4.3102e-06, 1.7695e-08, -2.2445e-06, ..., 8.3307e-07, + 1.0934e-06, 1.3718e-06], + [ 3.5716e-07, 7.4971e-07, 2.4494e-07, ..., 1.9949e-06, + 1.8366e-06, 1.7099e-06], + [ 4.3004e-07, 3.1781e-07, 2.6380e-07, ..., 7.9023e-07, + 1.3523e-06, 1.2293e-06], + ..., + [ 7.1852e-07, 7.3900e-07, 6.4261e-07, ..., 2.7269e-06, + 1.2871e-06, 1.0720e-06], + [ 4.6147e-07, -2.9569e-07, 1.8615e-07, ..., 6.1840e-07, + 9.7044e-07, 8.1863e-07], + [-5.4203e-07, 1.0081e-05, -1.5469e-06, ..., 1.2800e-05, + 1.2465e-05, 1.0267e-05]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0293, -0.0189, -0.0087, 0.0179, -0.0034, -0.0138, 0.0004, 0.0153, + -0.0263, 0.0144], device='cuda:0'), grad: tensor([-8.2925e-06, 4.8317e-06, 4.6305e-06, 2.9691e-06, -3.7640e-05, + 1.3672e-05, -2.1487e-05, 5.3085e-06, 5.5693e-06, 3.0413e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 218.53, cls_loss 0.0020 cls_loss_mapping 0.0063 cls_loss_causal 0.5500 re_mapping 0.0052 re_causal 0.0162 /// teacc 98.99 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0115, -0.0543, 0.0157, ..., -0.0788, -0.0904, -0.0603], + [-0.0798, 0.0446, -0.0510, ..., -0.0765, -0.0655, -0.1269], + [-0.0633, -0.0888, -0.0800, ..., -0.1045, -0.0528, 0.1134], + ..., + [ 0.0668, -0.0188, -0.0693, ..., 0.0653, -0.0553, -0.0516], + [ 0.0962, 0.0649, 0.0364, ..., -0.0740, -0.0792, 0.0633], + [ 0.0515, 0.0063, 0.0530, ..., 0.0433, -0.0619, 0.0156]], + device='cuda:0'), grad: tensor([[-1.7826e-06, 2.4792e-06, 1.6177e-06, ..., 8.1840e-08, + 1.3824e-08, 3.1088e-06], + [ 6.6636e-07, 2.6170e-07, 4.5821e-07, ..., 2.8382e-07, + 1.7532e-07, 7.1013e-07], + [ 2.0638e-06, 1.2582e-06, 8.8941e-07, ..., 1.1944e-07, + 3.4604e-08, 1.2601e-06], + ..., + [ 6.2678e-07, 8.2422e-07, 8.5682e-07, ..., 3.1944e-07, + 7.9861e-08, 5.0385e-07], + [-9.7379e-06, -1.0498e-05, -7.7933e-06, ..., 3.4575e-07, + 2.0082e-08, -1.2338e-05], + [ 4.5784e-06, 4.3362e-06, 2.5313e-06, ..., -4.4308e-07, + 8.9128e-07, 5.7667e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0293, -0.0193, -0.0088, 0.0177, -0.0031, -0.0135, 0.0005, 0.0157, + -0.0264, 0.0143], device='cuda:0'), grad: tensor([-7.9721e-06, 9.6392e-07, 6.5677e-06, 2.2035e-06, -9.3132e-08, + -3.0473e-06, 5.2042e-06, 2.9430e-06, -1.9819e-05, 1.3039e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 218.24, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5369 re_mapping 0.0056 re_causal 0.0166 /// teacc 99.03 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0117, -0.0546, 0.0157, ..., -0.0788, -0.0907, -0.0604], + [-0.0799, 0.0452, -0.0514, ..., -0.0770, -0.0665, -0.1277], + [-0.0636, -0.0891, -0.0800, ..., -0.1052, -0.0533, 0.1136], + ..., + [ 0.0665, -0.0196, -0.0705, ..., 0.0648, -0.0561, -0.0521], + [ 0.0962, 0.0649, 0.0363, ..., -0.0746, -0.0798, 0.0632], + [ 0.0516, 0.0066, 0.0533, ..., 0.0435, -0.0622, 0.0153]], + device='cuda:0'), grad: tensor([[-1.6543e-07, -2.0750e-06, -1.5780e-05, ..., 1.9930e-07, + 1.8636e-06, -3.6284e-06], + [ 1.2596e-07, 2.9174e-07, 1.7812e-07, ..., 2.1867e-06, + 2.3916e-06, 1.8468e-06], + [ 1.1723e-07, 1.7518e-06, 9.3430e-06, ..., 3.1712e-07, + -4.2981e-07, 2.9546e-07], + ..., + [-1.1432e-07, 2.7404e-07, 1.2037e-07, ..., 1.2871e-06, + 1.2126e-06, 8.5728e-07], + [ 2.6636e-07, 3.5623e-07, 1.5749e-06, ..., 8.4238e-07, + 9.7509e-07, 1.5302e-06], + [-2.6673e-06, -1.1921e-06, 9.6951e-07, ..., -3.0883e-06, + 4.1258e-07, -1.9707e-06]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0297, -0.0192, -0.0089, 0.0177, -0.0030, -0.0133, 0.0005, 0.0153, + -0.0266, 0.0143], device='cuda:0'), grad: tensor([-2.1085e-05, 7.1451e-06, 1.0125e-05, -1.4913e-07, -5.2992e-07, + 3.0138e-06, -3.2447e-06, 3.6564e-06, 5.7258e-06, -4.7274e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 218.62, cls_loss 0.0015 cls_loss_mapping 0.0040 cls_loss_causal 0.5271 re_mapping 0.0051 re_causal 0.0162 /// teacc 99.06 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0117, -0.0550, 0.0157, ..., -0.0789, -0.0920, -0.0615], + [-0.0805, 0.0444, -0.0522, ..., -0.0786, -0.0689, -0.1288], + [-0.0636, -0.0892, -0.0801, ..., -0.1059, -0.0536, 0.1137], + ..., + [ 0.0669, -0.0188, -0.0709, ..., 0.0654, -0.0566, -0.0525], + [ 0.0962, 0.0652, 0.0365, ..., -0.0753, -0.0801, 0.0632], + [ 0.0516, 0.0067, 0.0537, ..., 0.0435, -0.0625, 0.0152]], + device='cuda:0'), grad: tensor([[-2.2689e-07, 8.0909e-08, -1.0943e-07, ..., 7.3225e-08, + 2.2672e-08, 5.6252e-07], + [ 3.0221e-07, -7.2177e-07, 5.0786e-08, ..., 3.2294e-07, + 3.3935e-08, 5.5917e-06], + [ 1.4594e-06, 1.6438e-06, 1.0002e-06, ..., 2.9407e-07, + 3.6234e-08, -9.3505e-06], + ..., + [-3.8520e-06, 5.6112e-07, 4.1531e-08, ..., -4.9248e-06, + 3.6031e-08, 9.1270e-07], + [-1.7080e-06, -2.7493e-06, -1.3905e-06, ..., 1.0675e-07, + -5.5006e-09, -2.6803e-06], + [ 2.8815e-06, 1.0233e-07, 1.0827e-08, ..., 3.5353e-06, + 4.9215e-08, 2.6356e-07]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0295, -0.0209, -0.0089, 0.0172, -0.0026, -0.0126, 0.0006, 0.0164, + -0.0267, 0.0142], device='cuda:0'), grad: tensor([ 1.2033e-06, 1.5482e-05, -2.6301e-05, 1.0222e-05, 2.6990e-06, + -1.1958e-06, 4.0643e-06, -3.8706e-06, -1.0021e-05, 7.6182e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 218.42, cls_loss 0.0016 cls_loss_mapping 0.0047 cls_loss_causal 0.5309 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.13 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0119, -0.0551, 0.0158, ..., -0.0788, -0.0921, -0.0619], + [-0.0807, 0.0455, -0.0527, ..., -0.0787, -0.0694, -0.1292], + [-0.0637, -0.0895, -0.0801, ..., -0.1067, -0.0547, 0.1138], + ..., + [ 0.0674, -0.0194, -0.0711, ..., 0.0658, -0.0568, -0.0529], + [ 0.0963, 0.0653, 0.0366, ..., -0.0761, -0.0803, 0.0633], + [ 0.0516, 0.0067, 0.0538, ..., 0.0433, -0.0629, 0.0148]], + device='cuda:0'), grad: tensor([[-1.0408e-05, 5.0813e-06, -1.1292e-07, ..., 5.5786e-07, + -2.4345e-06, -4.2319e-06], + [ 1.5495e-07, -7.4841e-06, 5.8149e-08, ..., 6.7987e-07, + 4.3958e-07, 3.0431e-07], + [ 2.8964e-07, 7.8697e-07, 1.6484e-07, ..., 7.1805e-07, + 4.1188e-07, 2.5448e-07], + ..., + [-4.7591e-07, 8.2050e-07, 7.9046e-08, ..., 1.7136e-06, + 1.7537e-06, 1.1995e-06], + [-3.2876e-07, 5.2096e-08, -1.5926e-07, ..., 7.8790e-07, + 4.8336e-07, 1.1455e-07], + [ 3.7905e-07, 1.6307e-06, 2.5902e-09, ..., 1.7717e-05, + 1.1124e-05, 7.8604e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0300, -0.0201, -0.0090, 0.0178, -0.0030, -0.0135, 0.0015, 0.0163, + -0.0267, 0.0139], device='cuda:0'), grad: tensor([-5.9754e-06, -1.6570e-05, 2.9150e-06, -2.3423e-07, -3.4958e-05, + 2.0768e-06, 1.9148e-05, 3.9972e-06, 9.7323e-07, 2.8595e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 218.56, cls_loss 0.0016 cls_loss_mapping 0.0046 cls_loss_causal 0.5296 re_mapping 0.0053 re_causal 0.0161 /// teacc 98.99 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0120, -0.0556, 0.0159, ..., -0.0788, -0.0931, -0.0626], + [-0.0810, 0.0462, -0.0531, ..., -0.0790, -0.0703, -0.1297], + [-0.0638, -0.0898, -0.0804, ..., -0.1065, -0.0552, 0.1140], + ..., + [ 0.0672, -0.0203, -0.0719, ..., 0.0659, -0.0578, -0.0545], + [ 0.0967, 0.0655, 0.0368, ..., -0.0769, -0.0790, 0.0635], + [ 0.0515, 0.0068, 0.0539, ..., 0.0432, -0.0636, 0.0144]], + device='cuda:0'), grad: tensor([[ 7.7952e-07, 4.1618e-08, 4.0920e-08, ..., 1.7080e-06, + 3.5740e-07, 5.0850e-07], + [ 4.8336e-07, -1.0654e-06, 7.5670e-08, ..., 4.0792e-07, + 5.0524e-07, 5.6345e-07], + [ 9.4809e-07, 1.8987e-07, 1.5099e-07, ..., 9.0711e-07, + 2.7637e-07, 1.4110e-07], + ..., + [-1.1530e-06, 1.8696e-07, 4.9639e-07, ..., -3.1409e-07, + 3.7282e-08, 3.1502e-07], + [ 5.4855e-07, 3.2014e-07, -1.2876e-07, ..., 2.8498e-07, + 3.4682e-06, 3.1460e-06], + [-3.9637e-06, -1.3015e-07, -1.9129e-06, ..., -7.7263e-06, + 1.4494e-07, -5.2899e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0299, -0.0197, -0.0089, 0.0179, -0.0027, -0.0132, 0.0014, 0.0156, + -0.0265, 0.0135], device='cuda:0'), grad: tensor([ 1.7732e-06, -2.1793e-06, 2.9206e-06, 1.9800e-06, 6.3218e-06, + 5.2676e-06, -1.4782e-05, -2.1067e-06, 9.7156e-06, -8.9332e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 218.57, cls_loss 0.0014 cls_loss_mapping 0.0055 cls_loss_causal 0.5238 re_mapping 0.0054 re_causal 0.0164 /// teacc 99.00 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0121, -0.0558, 0.0160, ..., -0.0788, -0.0933, -0.0627], + [-0.0812, 0.0465, -0.0534, ..., -0.0797, -0.0710, -0.1305], + [-0.0638, -0.0900, -0.0805, ..., -0.1073, -0.0559, 0.1143], + ..., + [ 0.0676, -0.0204, -0.0721, ..., 0.0663, -0.0584, -0.0552], + [ 0.0968, 0.0658, 0.0370, ..., -0.0777, -0.0792, 0.0636], + [ 0.0515, 0.0065, 0.0540, ..., 0.0431, -0.0641, 0.0139]], + device='cuda:0'), grad: tensor([[-5.0813e-06, 1.4680e-07, -1.4585e-06, ..., 2.1607e-07, + 7.1828e-08, 5.0408e-08], + [ 3.8445e-06, -9.4529e-07, 1.3562e-07, ..., 1.7602e-06, + 2.1211e-07, 9.3458e-07], + [ 6.1020e-06, 2.6054e-07, 3.4645e-07, ..., 3.8072e-06, + 9.5135e-07, -2.6226e-06], + ..., + [-8.6650e-06, 3.7556e-07, 4.4773e-07, ..., 2.7921e-06, + 3.3975e-06, 3.8669e-06], + [-4.2637e-08, -5.3225e-07, -4.9081e-07, ..., 1.6100e-07, + 3.2946e-08, -5.7567e-08], + [ 4.0908e-07, -1.8207e-07, -9.4587e-08, ..., 3.0757e-07, + 6.4820e-07, 5.7463e-07]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0302, -0.0198, -0.0088, 0.0176, -0.0025, -0.0133, 0.0014, 0.0157, + -0.0266, 0.0133], device='cuda:0'), grad: tensor([-1.2189e-05, 1.3925e-05, 1.5900e-05, 6.2399e-06, -1.2435e-05, + 2.4810e-06, 3.3081e-06, -2.2247e-05, 2.1495e-06, 2.7865e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 218.59, cls_loss 0.0023 cls_loss_mapping 0.0058 cls_loss_causal 0.5235 re_mapping 0.0053 re_causal 0.0161 /// teacc 99.01 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0122, -0.0564, 0.0161, ..., -0.0789, -0.0938, -0.0632], + [-0.0830, 0.0469, -0.0540, ..., -0.0822, -0.0717, -0.1311], + [-0.0639, -0.0902, -0.0807, ..., -0.1086, -0.0569, 0.1146], + ..., + [ 0.0686, -0.0208, -0.0726, ..., 0.0671, -0.0608, -0.0569], + [ 0.0969, 0.0660, 0.0372, ..., -0.0784, -0.0797, 0.0637], + [ 0.0510, 0.0091, 0.0549, ..., 0.0426, -0.0642, 0.0114]], + device='cuda:0'), grad: tensor([[-3.0510e-06, 1.1260e-06, -3.4273e-07, ..., 3.4197e-08, + -3.0571e-07, 5.1642e-07], + [-8.6846e-07, 4.8354e-06, 1.6519e-07, ..., -1.5963e-06, + 3.8475e-08, 5.8822e-06], + [ 4.9593e-07, 5.8599e-06, 2.5425e-07, ..., 1.0254e-06, + 1.9418e-07, -2.8938e-05], + ..., + [ 2.0617e-07, 5.3905e-06, 1.7590e-07, ..., 1.5786e-07, + 1.6676e-08, 4.7381e-07], + [ 2.4098e-07, 2.4159e-06, 1.3772e-07, ..., 4.0716e-08, + 1.1700e-07, 1.4470e-07], + [ 1.1204e-06, 1.2331e-06, 2.2957e-07, ..., -1.5064e-07, + 1.5716e-07, 1.6775e-07]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0303, -0.0209, -0.0088, 0.0177, -0.0029, -0.0140, 0.0025, 0.0164, + -0.0266, 0.0129], device='cuda:0'), grad: tensor([-4.3325e-06, 2.4527e-05, -5.8800e-05, -3.6061e-06, 6.3591e-06, + -6.9082e-05, 7.0333e-05, 1.3806e-05, 1.1779e-05, 9.1121e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 218.94, cls_loss 0.0022 cls_loss_mapping 0.0063 cls_loss_causal 0.5364 re_mapping 0.0052 re_causal 0.0161 /// teacc 99.13 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 0.0123, -0.0570, 0.0162, ..., -0.0789, -0.0948, -0.0638], + [-0.0835, 0.0470, -0.0544, ..., -0.0827, -0.0726, -0.1318], + [-0.0640, -0.0904, -0.0809, ..., -0.1095, -0.0587, 0.1145], + ..., + [ 0.0659, -0.0212, -0.0756, ..., 0.0643, -0.0617, -0.0571], + [ 0.0971, 0.0662, 0.0386, ..., -0.0792, -0.0807, 0.0642], + [ 0.0529, 0.0085, 0.0562, ..., 0.0426, -0.0656, 0.0098]], + device='cuda:0'), grad: tensor([[ 2.2142e-07, 1.3039e-06, 6.3702e-07, ..., 3.3923e-07, + 6.9523e-07, 1.4845e-06], + [ 1.3784e-06, 1.2312e-06, 4.9733e-07, ..., 5.3318e-07, + 4.6077e-07, 1.9390e-06], + [ 4.9919e-07, 1.7248e-06, 8.5495e-07, ..., 8.3726e-07, + 6.5705e-07, -2.0996e-05], + ..., + [-3.2354e-06, 3.9162e-07, 3.4319e-07, ..., -6.5081e-06, + -4.3958e-07, 1.0673e-06], + [-7.8678e-06, -5.3085e-06, 2.3562e-06, ..., 3.0780e-07, + -1.2172e-06, 6.1207e-06], + [ 2.3805e-06, 4.8466e-06, 2.9989e-06, ..., 3.5185e-06, + 1.2927e-06, 1.9688e-06]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0303, -0.0211, -0.0089, 0.0178, -0.0014, -0.0130, 0.0009, 0.0137, + -0.0268, 0.0130], device='cuda:0'), grad: tensor([ 5.3346e-06, 7.0035e-06, -4.3660e-05, -2.4840e-05, 6.4075e-06, + 3.1888e-05, 2.0191e-06, -8.0913e-06, 5.6401e-06, 1.8254e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 218.73, cls_loss 0.0016 cls_loss_mapping 0.0061 cls_loss_causal 0.5288 re_mapping 0.0054 re_causal 0.0157 /// teacc 99.07 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 0.0124, -0.0575, 0.0164, ..., -0.0790, -0.0953, -0.0638], + [-0.0840, 0.0473, -0.0547, ..., -0.0831, -0.0732, -0.1324], + [-0.0641, -0.0907, -0.0811, ..., -0.1100, -0.0591, 0.1146], + ..., + [ 0.0660, -0.0217, -0.0758, ..., 0.0646, -0.0619, -0.0576], + [ 0.0973, 0.0665, 0.0385, ..., -0.0800, -0.0809, 0.0643], + [ 0.0529, 0.0087, 0.0565, ..., 0.0426, -0.0656, 0.0098]], + device='cuda:0'), grad: tensor([[-4.9360e-07, 3.3365e-07, 3.4372e-08, ..., 1.2480e-07, + 3.7195e-08, 8.0734e-08], + [ 8.0909e-08, -7.1190e-06, 4.9506e-08, ..., 2.0768e-07, + 1.0122e-07, 2.3388e-07], + [ 1.7253e-07, 9.6764e-07, 2.7183e-08, ..., 1.1036e-07, + -1.3737e-07, -1.7909e-06], + ..., + [-1.0887e-06, 1.4557e-06, 2.6519e-07, ..., -6.4559e-06, + -4.3632e-07, 1.2689e-07], + [ 3.4546e-08, 7.1572e-07, 1.3650e-08, ..., 1.3923e-07, + 1.8656e-08, 3.1694e-08], + [-1.8452e-07, -1.2564e-06, -6.1048e-07, ..., -1.3458e-07, + 2.9104e-10, 4.1880e-08]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0305, -0.0211, -0.0090, 0.0175, -0.0016, -0.0129, 0.0009, 0.0139, + -0.0268, 0.0130], device='cuda:0'), grad: tensor([-7.4692e-07, -1.0669e-05, -1.1399e-06, -5.0254e-06, 1.1481e-05, + 5.6736e-06, 3.0305e-06, -3.7607e-06, 1.9688e-06, -8.4005e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 218.55, cls_loss 0.0014 cls_loss_mapping 0.0044 cls_loss_causal 0.5020 re_mapping 0.0051 re_causal 0.0154 /// teacc 99.09 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 0.0125, -0.0583, 0.0164, ..., -0.0790, -0.0956, -0.0641], + [-0.0845, 0.0473, -0.0552, ..., -0.0834, -0.0738, -0.1331], + [-0.0642, -0.0909, -0.0814, ..., -0.1104, -0.0599, 0.1147], + ..., + [ 0.0660, -0.0231, -0.0763, ..., 0.0645, -0.0622, -0.0583], + [ 0.0973, 0.0666, 0.0385, ..., -0.0814, -0.0810, 0.0643], + [ 0.0531, 0.0094, 0.0571, ..., 0.0428, -0.0657, 0.0098]], + device='cuda:0'), grad: tensor([[-1.6494e-06, 2.9826e-07, 2.9657e-08, ..., 7.0315e-08, + 3.2736e-07, 4.1686e-06], + [ 4.6636e-07, 1.8964e-07, 3.6228e-07, ..., 5.2806e-07, + 5.1297e-06, 1.0616e-04], + [-1.2983e-06, 6.2119e-07, 1.9744e-07, ..., 2.7311e-07, + -7.2606e-06, -1.6212e-04], + ..., + [ 4.1421e-07, 8.4564e-07, 3.3877e-07, ..., -1.4948e-07, + 3.0361e-07, 7.1675e-06], + [ 3.6554e-07, 1.0513e-05, 2.9635e-06, ..., 8.0699e-07, + 6.2445e-07, 7.0669e-06], + [-2.4915e-05, -5.0157e-05, -4.3243e-05, ..., -1.1224e-04, + -4.6045e-05, -5.6773e-05]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0306, -0.0216, -0.0091, 0.0176, -0.0016, -0.0130, 0.0009, 0.0138, + -0.0267, 0.0132], device='cuda:0'), grad: tensor([ 1.0334e-05, 3.8958e-04, -5.9175e-04, 2.0117e-05, 1.9133e-04, + 5.5470e-06, 7.0333e-05, 2.2799e-05, 4.3929e-05, -1.6141e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 218.28, cls_loss 0.0016 cls_loss_mapping 0.0046 cls_loss_causal 0.5238 re_mapping 0.0055 re_causal 0.0163 /// teacc 99.05 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 0.0125, -0.0588, 0.0164, ..., -0.0790, -0.0976, -0.0651], + [-0.0851, 0.0478, -0.0567, ..., -0.0829, -0.0729, -0.1344], + [-0.0643, -0.0913, -0.0815, ..., -0.1109, -0.0602, 0.1150], + ..., + [ 0.0662, -0.0238, -0.0766, ..., 0.0646, -0.0627, -0.0588], + [ 0.0973, 0.0664, 0.0377, ..., -0.0829, -0.0815, 0.0641], + [ 0.0533, 0.0103, 0.0582, ..., 0.0428, -0.0657, 0.0100]], + device='cuda:0'), grad: tensor([[-1.1232e-06, 3.0198e-07, -1.5413e-07, ..., 4.3097e-07, + 4.7870e-07, 5.8394e-07], + [ 8.0047e-07, -2.0408e-07, 9.6858e-08, ..., 6.0163e-07, + 4.2631e-07, 9.2154e-07], + [-6.5342e-06, 3.8673e-07, 1.1519e-07, ..., 4.5123e-07, + 2.9290e-07, -5.0664e-06], + ..., + [ 2.0284e-06, 4.0955e-07, 6.3330e-08, ..., -1.2135e-06, + 1.8836e-07, 2.1309e-06], + [ 2.4438e-06, -5.5617e-08, 4.2928e-08, ..., 3.1106e-07, + 1.4342e-06, 3.0026e-06], + [ 3.4366e-07, 1.2629e-06, -2.6356e-07, ..., 2.7977e-06, + 2.7027e-06, 1.2005e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0307, -0.0204, -0.0091, 0.0174, -0.0018, -0.0131, 0.0015, 0.0138, + -0.0272, 0.0133], device='cuda:0'), grad: tensor([-3.5781e-06, 2.6971e-06, -1.7941e-05, 1.0908e-05, -1.0923e-05, + -9.4414e-05, 8.6844e-05, 4.3288e-06, 1.3135e-05, 9.0823e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 218.31, cls_loss 0.0015 cls_loss_mapping 0.0039 cls_loss_causal 0.5645 re_mapping 0.0052 re_causal 0.0165 /// teacc 99.15 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 0.0126, -0.0596, 0.0165, ..., -0.0791, -0.0980, -0.0655], + [-0.0857, 0.0481, -0.0571, ..., -0.0834, -0.0746, -0.1353], + [-0.0644, -0.0917, -0.0817, ..., -0.1117, -0.0608, 0.1150], + ..., + [ 0.0664, -0.0242, -0.0768, ..., 0.0647, -0.0630, -0.0593], + [ 0.0975, 0.0668, 0.0380, ..., -0.0836, -0.0819, 0.0642], + [ 0.0532, 0.0099, 0.0585, ..., 0.0430, -0.0656, 0.0104]], + device='cuda:0'), grad: tensor([[-4.1761e-06, 1.4657e-07, -4.2878e-06, ..., 3.2992e-07, + 1.0856e-07, 1.5984e-07], + [ 6.0722e-07, -1.9069e-07, 2.0489e-07, ..., 1.0384e-06, + 1.1607e-07, 2.3888e-07], + [ 1.8636e-06, 8.8057e-07, 6.2864e-07, ..., 1.6596e-06, + 7.1363e-08, 6.0303e-07], + ..., + [-2.6599e-06, 2.7637e-07, 1.2049e-07, ..., -5.7034e-06, + 2.2352e-08, 1.0029e-07], + [-1.8468e-06, -2.4736e-06, -1.1623e-06, ..., 4.1351e-07, + 1.1281e-07, -1.8170e-06], + [ 2.3074e-07, -1.0217e-06, -1.2503e-07, ..., -3.9139e-07, + 3.7299e-07, -3.1525e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0308, -0.0209, -0.0094, 0.0175, -0.0020, -0.0129, 0.0019, 0.0140, + -0.0272, 0.0133], device='cuda:0'), grad: tensor([-3.3617e-05, 1.4231e-06, 8.0764e-06, 1.9982e-05, 3.0529e-06, + 4.4592e-06, 9.1344e-06, -1.0997e-05, -5.5917e-06, 4.0382e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 218.44, cls_loss 0.0017 cls_loss_mapping 0.0044 cls_loss_causal 0.5147 re_mapping 0.0052 re_causal 0.0154 /// teacc 98.93 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0127, -0.0599, 0.0166, ..., -0.0792, -0.0986, -0.0661], + [-0.0863, 0.0482, -0.0578, ..., -0.0839, -0.0753, -0.1365], + [-0.0646, -0.0918, -0.0819, ..., -0.1134, -0.0640, 0.1143], + ..., + [ 0.0666, -0.0247, -0.0769, ..., 0.0649, -0.0635, -0.0598], + [ 0.0977, 0.0668, 0.0381, ..., -0.0853, -0.0821, 0.0643], + [ 0.0534, 0.0104, 0.0591, ..., 0.0433, -0.0654, 0.0109]], + device='cuda:0'), grad: tensor([[-6.2287e-06, 7.7533e-08, -3.7346e-06, ..., 2.6822e-07, + 5.5041e-07, 5.9837e-07], + [ 4.5798e-07, -5.1968e-07, 2.0023e-07, ..., 8.3726e-07, + 2.1420e-07, 3.8045e-07], + [ 2.7008e-06, 5.0291e-07, 3.5018e-07, ..., 1.4476e-05, + 1.3516e-07, 2.2016e-06], + ..., + [-2.8554e-06, 2.6240e-07, 5.1409e-07, ..., -2.4527e-05, + 2.3327e-08, -4.4703e-06], + [-9.6741e-08, -3.8045e-07, -2.0629e-07, ..., 6.6031e-07, + 6.6776e-07, 1.8300e-07], + [ 2.4177e-06, -1.9977e-07, 8.9314e-07, ..., 4.5262e-06, + 9.4122e-08, 1.0207e-06]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0310, -0.0215, -0.0098, 0.0175, -0.0022, -0.0138, 0.0023, 0.0143, + -0.0271, 0.0136], device='cuda:0'), grad: tensor([-1.4998e-05, 1.5022e-06, 2.1711e-05, 9.3579e-06, 7.2643e-06, + 1.7226e-05, -1.5117e-05, -4.5717e-05, 2.6766e-06, 1.6034e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 218.92, cls_loss 0.0017 cls_loss_mapping 0.0052 cls_loss_causal 0.5376 re_mapping 0.0048 re_causal 0.0151 /// teacc 98.91 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0128, -0.0603, 0.0167, ..., -0.0792, -0.0989, -0.0668], + [-0.0867, 0.0486, -0.0585, ..., -0.0841, -0.0762, -0.1382], + [-0.0647, -0.0921, -0.0821, ..., -0.1144, -0.0641, 0.1148], + ..., + [ 0.0667, -0.0256, -0.0772, ..., 0.0649, -0.0639, -0.0606], + [ 0.0979, 0.0670, 0.0383, ..., -0.0856, -0.0822, 0.0645], + [ 0.0535, 0.0101, 0.0596, ..., 0.0430, -0.0659, 0.0100]], + device='cuda:0'), grad: tensor([[ 1.8952e-07, 3.7788e-07, -1.6444e-08, ..., 5.3970e-07, + 2.2680e-05, 1.4506e-05], + [ 5.4622e-07, -2.5947e-06, 2.0990e-07, ..., 1.2629e-06, + 9.0944e-07, 9.9186e-07], + [-5.9465e-07, 4.1611e-06, 7.1758e-07, ..., 2.0489e-06, + 6.3237e-07, -8.8289e-06], + ..., + [-4.3656e-09, 2.0247e-06, 6.2864e-07, ..., -1.3337e-06, + 7.8371e-07, 3.3658e-06], + [-1.9539e-06, -2.1793e-06, -1.1735e-06, ..., 1.8757e-06, + 1.6717e-06, 1.1632e-06], + [-2.4643e-06, -3.5614e-06, -2.0750e-06, ..., -4.3884e-06, + 1.4119e-06, 1.4352e-06]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0312, -0.0221, -0.0096, 0.0176, -0.0017, -0.0140, 0.0019, 0.0144, + -0.0270, 0.0132], device='cuda:0'), grad: tensor([ 4.3780e-05, -2.7772e-06, -1.0557e-05, 2.7344e-05, 2.7623e-06, + -4.4852e-05, -3.4541e-05, 4.4182e-06, 1.9893e-05, -5.5432e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 217.57, cls_loss 0.0016 cls_loss_mapping 0.0052 cls_loss_causal 0.5126 re_mapping 0.0052 re_causal 0.0150 /// teacc 99.05 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0128, -0.0609, 0.0167, ..., -0.0793, -0.0996, -0.0681], + [-0.0871, 0.0487, -0.0591, ..., -0.0839, -0.0756, -0.1375], + [-0.0648, -0.0919, -0.0823, ..., -0.1156, -0.0642, 0.1156], + ..., + [ 0.0665, -0.0270, -0.0785, ..., 0.0650, -0.0645, -0.0628], + [ 0.0984, 0.0674, 0.0393, ..., -0.0848, -0.0822, 0.0647], + [ 0.0536, 0.0103, 0.0599, ..., 0.0431, -0.0660, 0.0099]], + device='cuda:0'), grad: tensor([[-4.7660e-07, 9.0688e-08, -1.0524e-07, ..., 6.5018e-08, + 1.0233e-07, 1.2096e-07], + [ 3.3935e-08, -5.9605e-07, 3.2305e-08, ..., 1.2328e-07, + 4.6799e-08, 2.4121e-07], + [ 1.0768e-07, 1.7346e-07, 6.3097e-08, ..., 1.0082e-07, + 1.0850e-07, -7.1116e-06], + ..., + [ 2.7823e-08, 1.4249e-07, 3.6205e-08, ..., 7.4040e-08, + 6.6299e-08, 5.1921e-07], + [-4.6915e-08, 4.2288e-08, -6.1584e-08, ..., 3.3004e-08, + 1.8231e-07, 5.0813e-06], + [ 1.3830e-07, 5.8115e-07, 4.0192e-08, ..., 1.2843e-06, + 1.0859e-06, 8.0839e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0311, -0.0217, -0.0090, 0.0173, -0.0019, -0.0139, 0.0017, 0.0142, + -0.0269, 0.0132], device='cuda:0'), grad: tensor([-8.8383e-07, -9.1363e-07, -1.7285e-05, 2.8554e-06, -2.9430e-06, + -2.3027e-07, 9.0105e-07, 1.7732e-06, 1.3366e-05, 3.3211e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 217.15, cls_loss 0.0012 cls_loss_mapping 0.0036 cls_loss_causal 0.5287 re_mapping 0.0052 re_causal 0.0154 /// teacc 99.00 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 0.0129, -0.0613, 0.0169, ..., -0.0793, -0.0998, -0.0680], + [-0.0878, 0.0488, -0.0596, ..., -0.0842, -0.0760, -0.1379], + [-0.0649, -0.0921, -0.0825, ..., -0.1170, -0.0645, 0.1157], + ..., + [ 0.0666, -0.0271, -0.0787, ..., 0.0652, -0.0649, -0.0631], + [ 0.0986, 0.0677, 0.0397, ..., -0.0853, -0.0823, 0.0648], + [ 0.0537, 0.0115, 0.0601, ..., 0.0434, -0.0658, 0.0102]], + device='cuda:0'), grad: tensor([[-2.3411e-07, 1.6892e-07, 1.1758e-07, ..., 5.7230e-07, + 4.5961e-07, 9.7696e-07], + [ 7.8056e-08, -2.2980e-07, 5.7044e-08, ..., 4.7544e-07, + 3.2969e-07, 1.0217e-06], + [ 1.4994e-07, 7.6718e-08, 2.8958e-08, ..., -2.1067e-06, + -2.4140e-06, -1.0543e-05], + ..., + [-1.7090e-07, 3.7020e-07, 1.7800e-07, ..., -1.4622e-07, + 3.3318e-07, 7.8883e-07], + [ 1.6042e-07, 8.2073e-09, 1.8859e-08, ..., 7.5158e-07, + 5.5972e-07, 1.6354e-06], + [-2.0508e-06, -2.0470e-06, -1.7071e-06, ..., -4.1164e-06, + -4.4075e-07, -1.2503e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0315, -0.0219, -0.0091, 0.0171, -0.0024, -0.0140, 0.0017, 0.0144, + -0.0268, 0.0135], device='cuda:0'), grad: tensor([ 1.8370e-07, 2.3376e-06, -2.4125e-05, 5.3570e-06, 1.4156e-05, + -5.6624e-06, 6.5900e-06, 3.0879e-08, 6.4336e-06, -5.3532e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 217.45, cls_loss 0.0017 cls_loss_mapping 0.0048 cls_loss_causal 0.5581 re_mapping 0.0048 re_causal 0.0145 /// teacc 99.06 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0129, -0.0631, 0.0167, ..., -0.0795, -0.0986, -0.0691], + [-0.0883, 0.0491, -0.0603, ..., -0.0846, -0.0766, -0.1392], + [-0.0651, -0.0926, -0.0834, ..., -0.1199, -0.0648, 0.1155], + ..., + [ 0.0668, -0.0281, -0.0791, ..., 0.0656, -0.0651, -0.0627], + [ 0.0990, 0.0691, 0.0411, ..., -0.0863, -0.0825, 0.0656], + [ 0.0537, 0.0115, 0.0604, ..., 0.0435, -0.0659, 0.0101]], + device='cuda:0'), grad: tensor([[ 2.6682e-07, -4.7055e-07, -3.7928e-07, ..., 3.0128e-07, + 2.1532e-06, 1.9912e-06], + [ 3.7090e-07, 3.6694e-07, 4.6333e-07, ..., 1.8300e-07, + 1.4156e-07, 4.8941e-07], + [-2.6673e-06, 4.7870e-07, 2.6822e-07, ..., 1.8510e-07, + 2.1944e-07, -4.7088e-06], + ..., + [ 8.8708e-07, 6.1188e-07, 2.7916e-07, ..., 2.4703e-07, + 1.3737e-07, 1.4054e-06], + [-2.2189e-07, -1.7583e-06, -1.4855e-06, ..., 2.7451e-07, + 3.6042e-07, 7.6089e-07], + [-9.5554e-07, -9.5647e-07, -1.6140e-06, ..., -1.6922e-06, + 3.3923e-07, -7.4692e-07]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0317, -0.0222, -0.0096, 0.0168, -0.0025, -0.0141, 0.0011, 0.0146, + -0.0261, 0.0135], device='cuda:0'), grad: tensor([ 2.0303e-06, 2.1830e-06, -1.5348e-05, 8.9034e-07, 2.3413e-06, + 1.2420e-05, -1.4670e-05, 5.6587e-06, 4.2282e-06, 1.8161e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 217.46, cls_loss 0.0017 cls_loss_mapping 0.0052 cls_loss_causal 0.5391 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.14 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0130, -0.0638, 0.0169, ..., -0.0795, -0.0981, -0.0698], + [-0.0894, 0.0493, -0.0613, ..., -0.0856, -0.0772, -0.1394], + [-0.0655, -0.0928, -0.0840, ..., -0.1229, -0.0656, 0.1154], + ..., + [ 0.0672, -0.0274, -0.0794, ..., 0.0660, -0.0658, -0.0629], + [ 0.0993, 0.0693, 0.0417, ..., -0.0876, -0.0828, 0.0657], + [ 0.0538, 0.0116, 0.0608, ..., 0.0434, -0.0661, 0.0101]], + device='cuda:0'), grad: tensor([[ 4.4657e-07, 1.6042e-07, 1.7136e-07, ..., 5.0478e-07, + 1.3653e-06, 1.1483e-06], + [ 3.6485e-07, -1.3609e-07, 1.5018e-07, ..., 1.0687e-07, + 4.4378e-07, 5.2247e-07], + [ 5.8301e-07, 3.3039e-07, 2.5635e-07, ..., 6.2515e-08, + 2.0617e-07, -3.5949e-07], + ..., + [-1.6866e-06, 2.1257e-07, 1.9604e-07, ..., -1.4715e-06, + 7.8813e-08, 3.2247e-07], + [-2.5313e-06, -1.1874e-06, -1.0850e-06, ..., 3.8533e-07, + -4.4308e-07, -1.2880e-06], + [-6.6543e-07, -2.6589e-07, -7.0920e-07, ..., -1.2731e-06, + 2.1909e-07, -2.5006e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0321, -0.0222, -0.0100, 0.0168, -0.0023, -0.0147, 0.0005, 0.0153, + -0.0262, 0.0136], device='cuda:0'), grad: tensor([ 3.0324e-06, 6.4028e-07, -3.4319e-07, 3.3993e-06, 4.9144e-05, + 4.9127e-07, -4.6909e-05, -5.0068e-06, -4.1090e-06, -3.4762e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 217.19, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.5361 re_mapping 0.0048 re_causal 0.0151 /// teacc 99.08 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0132, -0.0647, 0.0169, ..., -0.0796, -0.0981, -0.0698], + [-0.0895, 0.0495, -0.0632, ..., -0.0864, -0.0777, -0.1406], + [-0.0656, -0.0931, -0.0842, ..., -0.1233, -0.0657, 0.1158], + ..., + [ 0.0672, -0.0279, -0.0798, ..., 0.0662, -0.0662, -0.0635], + [ 0.0994, 0.0694, 0.0419, ..., -0.0882, -0.0828, 0.0657], + [ 0.0539, 0.0120, 0.0612, ..., 0.0435, -0.0661, 0.0101]], + device='cuda:0'), grad: tensor([[ 7.5437e-07, 1.3644e-07, 2.5053e-07, ..., 4.3097e-07, + 2.6338e-06, 1.7332e-06], + [ 9.1852e-08, -7.5474e-06, 1.4290e-08, ..., 1.4016e-07, + 1.1316e-07, -8.1435e-06], + [ 3.1502e-07, 3.9861e-06, 6.2340e-08, ..., 2.8498e-07, + 1.3085e-07, 4.3958e-06], + ..., + [-2.5742e-06, -4.5821e-07, 1.3015e-07, ..., -2.5425e-06, + 7.2876e-08, -1.0151e-07], + [ 9.3412e-07, 2.1923e-06, 6.2515e-08, ..., 5.8487e-07, + 2.8852e-06, 4.0233e-06], + [ 6.5891e-07, 4.7777e-07, -5.5926e-07, ..., 8.0559e-07, + 2.0512e-07, 3.8464e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([ 3.2202e-02, -2.2596e-02, -9.8498e-03, 1.7229e-02, -2.2204e-03, + -1.4884e-02, 5.5253e-05, 1.5279e-02, -2.6350e-02, 1.3674e-02], + device='cuda:0'), grad: tensor([ 5.5954e-06, -3.2216e-05, 1.7807e-05, 1.0185e-05, 4.6641e-06, + -1.7509e-05, -1.8757e-06, -4.8354e-06, 1.5363e-05, 2.7511e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 216.87, cls_loss 0.0016 cls_loss_mapping 0.0042 cls_loss_causal 0.5403 re_mapping 0.0048 re_causal 0.0149 /// teacc 99.00 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0130, -0.0658, 0.0167, ..., -0.0796, -0.1005, -0.0726], + [-0.0899, 0.0500, -0.0638, ..., -0.0869, -0.0784, -0.1406], + [-0.0656, -0.0933, -0.0844, ..., -0.1243, -0.0662, 0.1159], + ..., + [ 0.0672, -0.0286, -0.0802, ..., 0.0663, -0.0665, -0.0641], + [ 0.0998, 0.0697, 0.0423, ..., -0.0892, -0.0829, 0.0661], + [ 0.0540, 0.0122, 0.0616, ..., 0.0436, -0.0662, 0.0100]], + device='cuda:0'), grad: tensor([[ 3.6019e-07, 3.9535e-07, 1.6054e-07, ..., 5.7323e-07, + 2.6566e-07, 5.4063e-07], + [ 6.0955e-07, -2.5053e-07, 1.5181e-07, ..., 1.0598e-06, + 2.7753e-07, 2.3679e-07], + [ 2.1351e-07, 5.1176e-07, 7.0664e-08, ..., 3.1968e-07, + 3.4372e-08, -2.5854e-06], + ..., + [ 1.7649e-06, 2.5518e-06, 1.2061e-06, ..., 2.9448e-06, + 1.8217e-06, 2.4196e-06], + [ 1.0999e-06, 1.2927e-06, 5.2666e-07, ..., 1.7267e-06, + 7.4785e-07, 1.1148e-06], + [-2.4617e-05, -2.3559e-05, -1.0960e-05, ..., -3.4899e-05, + -1.4372e-05, -2.0668e-05]], device='cuda:0') +Epoch 151, bias, value: tensor([ 3.1667e-02, -2.2049e-02, -9.8336e-03, 1.6618e-02, -2.2240e-03, + -1.4322e-02, -2.9679e-05, 1.5170e-02, -2.6093e-02, 1.3706e-02], + device='cuda:0'), grad: tensor([ 1.6559e-06, -6.7800e-07, -3.0156e-06, -6.9849e-07, 5.0515e-05, + 9.3039e-07, 3.9786e-06, 5.7705e-06, 4.1276e-06, -6.2644e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 217.43, cls_loss 0.0014 cls_loss_mapping 0.0045 cls_loss_causal 0.5438 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.08 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0132, -0.0664, 0.0168, ..., -0.0797, -0.1008, -0.0727], + [-0.0903, 0.0501, -0.0641, ..., -0.0877, -0.0793, -0.1410], + [-0.0657, -0.0936, -0.0845, ..., -0.1250, -0.0668, 0.1162], + ..., + [ 0.0672, -0.0292, -0.0806, ..., 0.0663, -0.0674, -0.0653], + [ 0.0998, 0.0698, 0.0423, ..., -0.0900, -0.0831, 0.0659], + [ 0.0541, 0.0123, 0.0618, ..., 0.0435, -0.0664, 0.0099]], + device='cuda:0'), grad: tensor([[-6.8471e-06, 6.6997e-08, -1.9167e-06, ..., -1.2573e-06, + -1.5199e-06, -1.4193e-06], + [ 5.0711e-07, -5.2620e-07, 5.2853e-07, ..., 8.0909e-08, + 4.1281e-07, 4.5286e-07], + [ 5.5647e-07, -2.1886e-08, 1.8103e-07, ..., 1.9348e-07, + 1.5099e-07, -7.8185e-07], + ..., + [ 1.6261e-06, 1.0002e-06, 9.9745e-07, ..., 1.2973e-06, + 2.1886e-08, 3.2736e-07], + [ 5.3225e-07, 2.7986e-07, 1.6252e-07, ..., 1.6484e-07, + 1.5320e-07, 7.0781e-07], + [-3.2189e-08, -1.9204e-06, -1.7639e-06, ..., -4.0904e-06, + 5.2067e-08, -5.8999e-07]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0319, -0.0220, -0.0097, 0.0170, -0.0020, -0.0139, -0.0006, 0.0150, + -0.0264, 0.0135], device='cuda:0'), grad: tensor([-3.4124e-05, 4.6864e-06, 1.6931e-06, 2.4997e-06, 2.0787e-06, + 7.3574e-06, 9.9242e-06, 4.6417e-06, 3.7644e-06, -2.5462e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 217.56, cls_loss 0.0013 cls_loss_mapping 0.0040 cls_loss_causal 0.5319 re_mapping 0.0049 re_causal 0.0154 /// teacc 99.09 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 0.0131, -0.0670, 0.0168, ..., -0.0801, -0.1011, -0.0730], + [-0.0906, 0.0507, -0.0647, ..., -0.0881, -0.0798, -0.1411], + [-0.0658, -0.0940, -0.0848, ..., -0.1256, -0.0674, 0.1163], + ..., + [ 0.0673, -0.0297, -0.0809, ..., 0.0664, -0.0684, -0.0660], + [ 0.1004, 0.0708, 0.0437, ..., -0.0906, -0.0828, 0.0665], + [ 0.0543, 0.0124, 0.0622, ..., 0.0437, -0.0664, 0.0099]], + device='cuda:0'), grad: tensor([[-1.7822e-05, 5.3580e-08, -5.2229e-06, ..., 3.2852e-07, + -5.1335e-06, -4.0643e-06], + [ 4.1490e-07, -1.1252e-07, 1.0524e-07, ..., 1.4598e-07, + 1.6834e-07, 2.5798e-07], + [ 8.0140e-07, 1.9907e-07, 2.0128e-07, ..., 1.5879e-07, + 8.4657e-07, -5.9605e-07], + ..., + [-2.1921e-07, 9.5577e-08, 8.5682e-08, ..., -1.1083e-06, + 1.0408e-07, 3.3202e-07], + [ 4.4256e-06, 1.3774e-06, 2.6543e-06, ..., 1.6363e-06, + 3.1553e-06, 4.1835e-06], + [ 7.0035e-07, -3.2373e-06, -3.1814e-06, ..., -2.5984e-06, + 4.7358e-07, -1.6047e-06]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0318, -0.0215, -0.0098, 0.0168, -0.0021, -0.0145, -0.0008, 0.0149, + -0.0258, 0.0136], device='cuda:0'), grad: tensor([-6.2644e-05, 1.4743e-06, 3.8892e-06, 2.8670e-05, 7.1526e-06, + -5.3763e-05, 4.3392e-05, -1.6676e-08, 3.0145e-05, 1.7183e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 217.17, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.5401 re_mapping 0.0048 re_causal 0.0150 /// teacc 99.00 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0131, -0.0677, 0.0167, ..., -0.0803, -0.1014, -0.0737], + [-0.0905, 0.0513, -0.0655, ..., -0.0887, -0.0801, -0.1416], + [-0.0659, -0.0943, -0.0851, ..., -0.1272, -0.0679, 0.1163], + ..., + [ 0.0674, -0.0302, -0.0812, ..., 0.0666, -0.0693, -0.0666], + [ 0.1005, 0.0711, 0.0442, ..., -0.0915, -0.0828, 0.0667], + [ 0.0544, 0.0125, 0.0624, ..., 0.0437, -0.0665, 0.0098]], + device='cuda:0'), grad: tensor([[ 1.7341e-06, 2.4196e-06, 1.7714e-06, ..., 1.4831e-07, + 2.2864e-07, 1.1716e-06], + [ 3.0617e-07, 3.0338e-07, 2.2247e-07, ..., 3.2573e-07, + 1.8696e-07, 2.3958e-07], + [ 2.7986e-07, 3.0361e-07, 1.9441e-07, ..., 1.3201e-07, + 9.1677e-08, 1.5041e-07], + ..., + [ 4.1793e-08, 2.4447e-07, 1.4727e-07, ..., -1.5320e-07, + 1.3772e-07, 1.6426e-07], + [-5.5321e-06, -5.2452e-06, -4.3437e-06, ..., 1.2293e-07, + 7.9686e-08, -2.5518e-06], + [ 9.4529e-07, 6.6170e-07, 5.8021e-07, ..., 9.6567e-08, + 2.2806e-07, 6.0117e-07]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0319, -0.0211, -0.0099, 0.0162, -0.0021, -0.0137, -0.0010, 0.0149, + -0.0258, 0.0136], device='cuda:0'), grad: tensor([ 4.3884e-06, 1.3728e-06, 1.1837e-06, 8.3819e-07, -1.3430e-06, + 2.3134e-06, 6.6869e-07, 3.0315e-07, -1.2808e-05, 3.0603e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 217.34, cls_loss 0.0014 cls_loss_mapping 0.0040 cls_loss_causal 0.5278 re_mapping 0.0047 re_causal 0.0147 /// teacc 99.06 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 0.0131, -0.0690, 0.0166, ..., -0.0804, -0.1019, -0.0747], + [-0.0915, 0.0514, -0.0665, ..., -0.0896, -0.0832, -0.1431], + [-0.0659, -0.0946, -0.0854, ..., -0.1303, -0.0700, 0.1161], + ..., + [ 0.0674, -0.0311, -0.0815, ..., 0.0671, -0.0684, -0.0654], + [ 0.1010, 0.0716, 0.0450, ..., -0.0925, -0.0829, 0.0669], + [ 0.0545, 0.0126, 0.0626, ..., 0.0438, -0.0666, 0.0095]], + device='cuda:0'), grad: tensor([[-5.6103e-06, 1.7358e-07, 2.5786e-08, ..., -1.6484e-07, + 4.5868e-08, 8.2655e-08], + [ 1.4086e-07, -1.7136e-06, -5.9837e-08, ..., 3.0559e-09, + -7.4622e-08, 1.3714e-07], + [ 7.6182e-07, 4.1723e-07, 3.1432e-07, ..., 2.0867e-08, + 6.2399e-08, -3.4762e-07], + ..., + [ 3.8277e-07, 1.4750e-07, 6.6531e-08, ..., 2.7969e-08, + 8.7020e-09, 1.1467e-07], + [-6.5099e-07, 4.6776e-07, -4.7358e-07, ..., 1.4435e-08, + 9.1386e-08, -2.1863e-07], + [ 4.4368e-06, 1.2410e-07, -1.8044e-09, ..., 8.6147e-08, + 2.9715e-08, 5.1834e-08]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0317, -0.0232, -0.0103, 0.0162, -0.0020, -0.0138, 0.0002, 0.0154, + -0.0256, 0.0135], device='cuda:0'), grad: tensor([-1.8671e-05, -3.0864e-06, 1.2880e-06, 1.3197e-06, 6.2678e-07, + -3.8221e-06, 2.1122e-06, 1.7304e-06, 1.8151e-06, 1.6704e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 217.23, cls_loss 0.0014 cls_loss_mapping 0.0046 cls_loss_causal 0.4882 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.08 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 0.0142, -0.0662, 0.0177, ..., -0.0804, -0.1024, -0.0742], + [-0.0919, 0.0514, -0.0670, ..., -0.0900, -0.0831, -0.1437], + [-0.0661, -0.0949, -0.0855, ..., -0.1304, -0.0702, 0.1169], + ..., + [ 0.0676, -0.0320, -0.0817, ..., 0.0673, -0.0692, -0.0672], + [ 0.1011, 0.0717, 0.0450, ..., -0.0936, -0.0830, 0.0670], + [ 0.0539, 0.0119, 0.0622, ..., 0.0440, -0.0666, 0.0095]], + device='cuda:0'), grad: tensor([[ 3.0827e-07, 2.8522e-07, -6.6357e-09, ..., 4.9267e-07, + 1.8044e-07, 6.4634e-07], + [ 2.8824e-07, -1.3076e-06, 5.5472e-08, ..., -2.7212e-08, + -6.3388e-08, 3.2852e-07], + [-4.5933e-06, 6.5751e-07, 4.9709e-08, ..., 1.6333e-07, + 1.0640e-07, -7.9349e-06], + ..., + [ 3.0976e-06, 1.3150e-06, 5.4669e-07, ..., 7.4646e-07, + 1.2910e-07, 4.2915e-06], + [ 1.1530e-06, 1.6531e-07, -2.4564e-08, ..., 1.2992e-07, + 9.3307e-08, 1.8394e-06], + [-2.3022e-06, -2.6207e-06, -1.2526e-06, ..., -2.6450e-06, + 2.1304e-08, -7.0722e-09]], device='cuda:0') +Epoch 156, bias, value: tensor([ 3.3375e-02, -2.3580e-02, -9.5301e-03, 1.5859e-02, -2.1910e-03, + -1.3171e-02, -5.8726e-05, 1.4723e-02, -2.5616e-02, 1.2923e-02], + device='cuda:0'), grad: tensor([ 9.4902e-07, -2.3842e-06, -1.9282e-05, 2.3618e-06, 3.2652e-06, + 7.4646e-07, 7.8930e-08, 1.3679e-05, 5.6848e-06, -5.1409e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 217.17, cls_loss 0.0015 cls_loss_mapping 0.0042 cls_loss_causal 0.5543 re_mapping 0.0048 re_causal 0.0152 /// teacc 99.10 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 0.0140, -0.0665, 0.0176, ..., -0.0807, -0.1027, -0.0747], + [-0.0932, 0.0516, -0.0674, ..., -0.0905, -0.0833, -0.1453], + [-0.0657, -0.0952, -0.0855, ..., -0.1307, -0.0709, 0.1181], + ..., + [ 0.0676, -0.0342, -0.0831, ..., 0.0672, -0.0710, -0.0693], + [ 0.1012, 0.0719, 0.0448, ..., -0.0947, -0.0830, 0.0668], + [ 0.0542, 0.0120, 0.0626, ..., 0.0440, -0.0668, 0.0093]], + device='cuda:0'), grad: tensor([[ 5.4686e-08, 1.4610e-07, 1.0157e-07, ..., 1.2992e-07, + 3.0012e-07, 3.0268e-07], + [ 6.0012e-08, -3.9442e-07, 5.5355e-08, ..., 8.0443e-08, + 1.3341e-07, 1.7218e-07], + [-5.1310e-08, 2.5611e-07, 8.3179e-08, ..., 1.1933e-07, + 2.3062e-07, -5.7183e-07], + ..., + [-1.4424e-07, 3.4808e-07, 1.5914e-07, ..., -1.7672e-07, + 1.5961e-07, 2.3190e-07], + [ 1.0885e-07, 2.6636e-07, 1.0645e-06, ..., 1.3632e-07, + 3.6210e-06, 3.1460e-06], + [-6.4587e-07, -6.9663e-07, -7.4739e-07, ..., -3.2340e-07, + 4.2870e-08, -6.6939e-08]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0331, -0.0244, -0.0084, 0.0172, -0.0020, -0.0138, -0.0002, 0.0137, + -0.0258, 0.0130], device='cuda:0'), grad: tensor([ 7.9209e-07, -6.5845e-07, -6.0070e-07, 2.7870e-07, 2.2259e-06, + -4.0117e-07, -7.8306e-06, 2.3411e-07, 6.8173e-06, -8.8383e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 216.73, cls_loss 0.0013 cls_loss_mapping 0.0039 cls_loss_causal 0.5491 re_mapping 0.0050 re_causal 0.0154 /// teacc 99.04 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 0.0140, -0.0667, 0.0176, ..., -0.0808, -0.1035, -0.0755], + [-0.0940, 0.0528, -0.0674, ..., -0.0911, -0.0836, -0.1464], + [-0.0657, -0.0954, -0.0858, ..., -0.1311, -0.0715, 0.1184], + ..., + [ 0.0676, -0.0349, -0.0835, ..., 0.0672, -0.0724, -0.0700], + [ 0.1015, 0.0723, 0.0451, ..., -0.0955, -0.0831, 0.0671], + [ 0.0542, 0.0122, 0.0630, ..., 0.0440, -0.0669, 0.0092]], + device='cuda:0'), grad: tensor([[ 3.9907e-07, 4.5681e-07, 6.2166e-07, ..., 1.1688e-06, + 4.5425e-07, 7.3016e-07], + [ 8.8802e-07, 6.9663e-07, 1.8848e-07, ..., 5.8785e-06, + 2.1067e-06, 1.8049e-06], + [ 3.7299e-07, 6.8638e-07, 2.0629e-07, ..., 1.0533e-06, + 3.6112e-07, -3.0790e-06], + ..., + [-1.2182e-06, -4.2608e-07, 2.9267e-07, ..., -1.3830e-06, + 5.9046e-07, -2.0314e-07], + [ 1.5879e-06, 5.9139e-07, 2.7250e-06, ..., 4.1015e-06, + 1.6512e-06, 2.0061e-06], + [-6.2175e-06, -5.8152e-06, -1.2450e-05, ..., -1.2740e-05, + -4.7870e-06, -1.0207e-05]], device='cuda:0') +Epoch 158, bias, value: tensor([ 3.3105e-02, -2.4437e-02, -8.0865e-03, 1.9677e-02, -1.8779e-03, + -1.6710e-02, 4.4499e-05, 1.3388e-02, -2.5679e-02, 1.2878e-02], + device='cuda:0'), grad: tensor([ 3.5446e-06, 1.3918e-05, -5.9158e-06, 1.0765e-04, 8.8140e-06, + -1.0520e-04, 3.6303e-06, -4.8913e-06, 9.9540e-06, -3.1680e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 217.17, cls_loss 0.0016 cls_loss_mapping 0.0033 cls_loss_causal 0.5289 re_mapping 0.0047 re_causal 0.0142 /// teacc 98.92 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 0.0137, -0.0674, 0.0172, ..., -0.0813, -0.1043, -0.0767], + [-0.0947, 0.0530, -0.0690, ..., -0.0919, -0.0839, -0.1476], + [-0.0658, -0.0957, -0.0862, ..., -0.1315, -0.0721, 0.1189], + ..., + [ 0.0677, -0.0355, -0.0839, ..., 0.0673, -0.0731, -0.0709], + [ 0.1018, 0.0728, 0.0452, ..., -0.0965, -0.0837, 0.0672], + [ 0.0547, 0.0129, 0.0642, ..., 0.0442, -0.0670, 0.0093]], + device='cuda:0'), grad: tensor([[ 1.2317e-07, 6.3423e-07, 4.2915e-06, ..., 5.6531e-07, + 6.3155e-09, 1.8452e-07], + [-1.0495e-07, -2.2398e-07, 1.0664e-06, ..., 3.1199e-07, + 1.1059e-08, 1.6950e-07], + [ 3.5670e-07, 1.2005e-06, 5.0157e-05, ..., 1.1502e-06, + 8.4692e-09, 3.0664e-07], + ..., + [-1.0978e-07, 7.8138e-07, 2.4177e-06, ..., -1.3411e-07, + 9.6043e-09, 1.6077e-07], + [ 3.5809e-07, 2.5071e-06, 2.0891e-05, ..., 2.4959e-06, + 6.2864e-09, 6.5798e-07], + [-6.6832e-06, -3.2783e-05, -3.3051e-05, ..., -2.9907e-05, + 1.3981e-07, -1.0327e-05]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0325, -0.0249, -0.0079, 0.0194, -0.0018, -0.0167, 0.0009, 0.0132, + -0.0259, 0.0132], device='cuda:0'), grad: tensor([ 6.9380e-05, 8.5756e-06, 8.9550e-04, 6.9094e-04, 1.8969e-05, + -2.0180e-03, 1.8850e-05, 3.4928e-05, 3.4022e-04, -5.9366e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 217.27, cls_loss 0.0013 cls_loss_mapping 0.0041 cls_loss_causal 0.5297 re_mapping 0.0048 re_causal 0.0145 /// teacc 99.11 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0132, -0.0679, 0.0163, ..., -0.0823, -0.1048, -0.0780], + [-0.0955, 0.0530, -0.0707, ..., -0.0927, -0.0847, -0.1487], + [-0.0658, -0.0957, -0.0869, ..., -0.1319, -0.0728, 0.1191], + ..., + [ 0.0676, -0.0363, -0.0846, ..., 0.0672, -0.0738, -0.0720], + [ 0.1018, 0.0735, 0.0458, ..., -0.0972, -0.0842, 0.0671], + [ 0.0553, 0.0133, 0.0653, ..., 0.0445, -0.0673, 0.0091]], + device='cuda:0'), grad: tensor([[ 1.0629e-07, 1.3527e-07, 1.2980e-08, ..., 9.0688e-08, + 1.2910e-07, 4.9034e-07], + [-5.7090e-07, -2.9840e-06, 3.6962e-08, ..., -1.4540e-07, + 4.8865e-08, 8.8301e-08], + [ 6.3889e-07, 1.2340e-06, 3.5681e-08, ..., 4.3958e-07, + 1.0442e-07, -7.4971e-07], + ..., + [-1.1707e-06, 7.2736e-07, 3.0239e-08, ..., -9.9931e-07, + 3.5565e-08, 7.6951e-08], + [-1.7695e-07, -1.8871e-07, -1.3551e-07, ..., 7.3458e-08, + 1.8731e-07, 1.4564e-07], + [ 3.7416e-07, -2.5175e-08, -8.0327e-08, ..., 1.8440e-07, + 2.0431e-08, 4.4674e-08]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0317, -0.0256, -0.0081, 0.0189, -0.0015, -0.0160, 0.0011, 0.0129, + -0.0260, 0.0136], device='cuda:0'), grad: tensor([ 2.7064e-06, -9.0674e-06, 3.8277e-07, 3.9265e-06, 1.3690e-06, + 6.0862e-07, -7.8557e-07, -1.2647e-06, 3.7206e-07, 1.7770e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 217.59, cls_loss 0.0011 cls_loss_mapping 0.0039 cls_loss_causal 0.5030 re_mapping 0.0048 re_causal 0.0143 /// teacc 99.03 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0133, -0.0681, 0.0164, ..., -0.0824, -0.1052, -0.0785], + [-0.0961, 0.0538, -0.0713, ..., -0.0930, -0.0848, -0.1489], + [-0.0660, -0.0965, -0.0871, ..., -0.1324, -0.0730, 0.1191], + ..., + [ 0.0681, -0.0358, -0.0843, ..., 0.0676, -0.0742, -0.0722], + [ 0.1020, 0.0739, 0.0461, ..., -0.0978, -0.0843, 0.0673], + [ 0.0553, 0.0132, 0.0655, ..., 0.0444, -0.0674, 0.0089]], + device='cuda:0'), grad: tensor([[ 7.2352e-08, 8.0152e-08, 2.6659e-08, ..., 7.0082e-08, + 5.4017e-08, 6.7404e-08], + [ 3.5553e-07, 1.3469e-07, 7.9744e-08, ..., 9.0746e-08, + 6.5309e-08, 1.4657e-07], + [ 1.1008e-06, 3.4319e-07, 4.5664e-08, ..., 2.6613e-07, + 8.8185e-08, -2.5542e-07], + ..., + [-3.8482e-06, -7.7300e-07, 5.1281e-08, ..., 2.9360e-07, + 6.8126e-07, 7.1479e-07], + [-1.7472e-06, -1.3895e-06, -4.6100e-07, ..., 9.2201e-08, + 1.5914e-07, -4.7823e-07], + [ 7.5321e-08, -7.1130e-08, -1.5087e-07, ..., 3.0478e-07, + 2.9593e-07, 1.5704e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0317, -0.0259, -0.0081, 0.0186, -0.0015, -0.0160, 0.0012, 0.0137, + -0.0259, 0.0134], device='cuda:0'), grad: tensor([ 3.5600e-07, 9.2946e-07, 2.5965e-06, 5.5060e-06, -1.9539e-06, + 4.3921e-06, 1.2247e-07, -7.6741e-06, -5.1372e-06, 8.5589e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 217.33, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.4999 re_mapping 0.0047 re_causal 0.0140 /// teacc 98.97 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0133, -0.0684, 0.0163, ..., -0.0825, -0.1057, -0.0794], + [-0.0977, 0.0536, -0.0718, ..., -0.0943, -0.0851, -0.1493], + [-0.0662, -0.0972, -0.0875, ..., -0.1327, -0.0730, 0.1191], + ..., + [ 0.0688, -0.0347, -0.0839, ..., 0.0679, -0.0745, -0.0723], + [ 0.1026, 0.0746, 0.0471, ..., -0.0983, -0.0844, 0.0678], + [ 0.0552, 0.0130, 0.0653, ..., 0.0441, -0.0679, 0.0080]], + device='cuda:0'), grad: tensor([[ 1.8289e-07, 1.3085e-07, 2.7940e-07, ..., 4.2841e-07, + 7.2352e-08, 4.3586e-07], + [-1.0617e-07, -9.2573e-07, 4.3539e-08, ..., 9.4587e-09, + 1.5600e-08, 1.0978e-07], + [ 4.4447e-07, 1.6030e-07, 1.9209e-07, ..., 7.2177e-07, + 7.5437e-08, -2.0433e-06], + ..., + [ 7.8056e-08, 5.2201e-07, 3.6717e-07, ..., -3.8091e-07, + -1.3679e-09, 9.0944e-07], + [-2.3667e-07, -4.0774e-08, -1.9348e-07, ..., 2.7916e-07, + 7.6718e-08, 4.4354e-08], + [-1.3150e-06, -3.0530e-08, -1.0468e-06, ..., -2.0303e-06, + 2.2992e-09, -4.7823e-07]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0316, -0.0266, -0.0083, 0.0183, -0.0009, -0.0157, 0.0011, 0.0145, + -0.0256, 0.0129], device='cuda:0'), grad: tensor([ 1.6866e-06, -3.3602e-06, -1.0714e-05, 4.1090e-06, 2.8946e-06, + 1.4314e-06, 7.4273e-08, 6.4895e-06, 1.3057e-06, -3.9339e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 217.44, cls_loss 0.0015 cls_loss_mapping 0.0034 cls_loss_causal 0.5231 re_mapping 0.0045 re_causal 0.0140 /// teacc 98.96 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0133, -0.0686, 0.0165, ..., -0.0826, -0.1066, -0.0804], + [-0.0992, 0.0533, -0.0728, ..., -0.0956, -0.0852, -0.1504], + [-0.0662, -0.0978, -0.0883, ..., -0.1329, -0.0731, 0.1196], + ..., + [ 0.0691, -0.0344, -0.0844, ..., 0.0680, -0.0751, -0.0734], + [ 0.1029, 0.0751, 0.0473, ..., -0.0998, -0.0848, 0.0680], + [ 0.0556, 0.0136, 0.0664, ..., 0.0443, -0.0681, 0.0085]], + device='cuda:0'), grad: tensor([[-6.3442e-06, 1.1525e-07, -4.9913e-08, ..., -1.4319e-08, + 1.7462e-10, 5.2067e-08], + [ 2.3178e-07, -2.4393e-05, 1.9412e-08, ..., 3.4343e-09, + 1.2224e-09, 1.6851e-08], + [-5.3383e-06, 3.2578e-06, 1.0181e-07, ..., 4.4529e-09, + 1.7462e-10, -5.5917e-06], + ..., + [ 6.6534e-06, 1.9908e-05, 3.8097e-08, ..., 9.4296e-09, + 2.3283e-10, 5.5395e-06], + [-1.4785e-07, -7.4646e-07, -4.1164e-07, ..., 9.8080e-09, + 1.1642e-10, -5.4156e-07], + [ 1.5981e-06, 8.0839e-07, 3.2457e-07, ..., -1.5600e-08, + 4.0745e-10, 3.7532e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0314, -0.0267, -0.0079, 0.0175, -0.0013, -0.0150, 0.0009, 0.0145, + -0.0257, 0.0132], device='cuda:0'), grad: tensor([-2.1473e-05, -3.8838e-04, 3.3021e-05, 1.9610e-05, 1.1019e-05, + -9.9689e-06, 9.7826e-06, 3.3689e-04, 6.5472e-07, 8.8066e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 217.24, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5165 re_mapping 0.0047 re_causal 0.0139 /// teacc 99.10 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0133, -0.0689, 0.0165, ..., -0.0826, -0.1071, -0.0814], + [-0.0979, 0.0548, -0.0742, ..., -0.0972, -0.0854, -0.1511], + [-0.0659, -0.0984, -0.0887, ..., -0.1332, -0.0731, 0.1200], + ..., + [ 0.0686, -0.0364, -0.0843, ..., 0.0686, -0.0750, -0.0733], + [ 0.1031, 0.0754, 0.0479, ..., -0.1002, -0.0850, 0.0683], + [ 0.0555, 0.0143, 0.0665, ..., 0.0444, -0.0681, 0.0085]], + device='cuda:0'), grad: tensor([[ 1.9092e-07, 4.0466e-07, 2.9220e-07, ..., 2.5425e-07, + -6.8685e-09, 6.8266e-07], + [ 1.1484e-07, -9.1549e-07, 2.8114e-08, ..., -4.0187e-07, + 2.9162e-08, 5.1106e-08], + [ 8.0187e-07, 4.0233e-07, 7.1363e-08, ..., 9.4995e-07, + 1.5367e-08, -3.6496e-08], + ..., + [-3.4701e-06, -8.2143e-07, 1.6997e-07, ..., -4.1723e-06, + 4.9185e-08, 1.0623e-07], + [-2.5630e-06, -1.2890e-06, -1.5730e-06, ..., 2.1560e-07, + 8.5449e-08, -5.4156e-07], + [ 3.4552e-06, 3.3341e-06, 4.3004e-07, ..., 1.2748e-05, + 5.7295e-06, 4.6790e-06]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0313, -0.0260, -0.0079, 0.0175, -0.0015, -0.0150, 0.0011, 0.0141, + -0.0260, 0.0134], device='cuda:0'), grad: tensor([-7.8529e-06, -2.5835e-06, 1.7202e-06, 8.5589e-07, -1.3806e-05, + 5.1409e-07, 8.0466e-06, -5.4799e-06, -5.1036e-06, 2.3693e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 217.35, cls_loss 0.0019 cls_loss_mapping 0.0058 cls_loss_causal 0.4869 re_mapping 0.0049 re_causal 0.0137 /// teacc 98.98 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0141, -0.0686, 0.0174, ..., -0.0828, -0.1076, -0.0805], + [-0.0958, 0.0579, -0.0757, ..., -0.0994, -0.0858, -0.1519], + [-0.0662, -0.0993, -0.0891, ..., -0.1360, -0.0735, 0.1202], + ..., + [ 0.0670, -0.0393, -0.0848, ..., 0.0700, -0.0757, -0.0740], + [ 0.1027, 0.0758, 0.0476, ..., -0.1026, -0.0854, 0.0683], + [ 0.0560, 0.0155, 0.0690, ..., 0.0450, -0.0676, 0.0099]], + device='cuda:0'), grad: tensor([[ 2.3225e-08, 7.3621e-07, 5.5588e-08, ..., 5.5588e-08, + 7.3342e-09, 3.1316e-08], + [ 1.3737e-08, 3.9004e-06, 3.9465e-08, ..., 5.4948e-08, + 2.0431e-08, 3.5565e-08], + [ 3.1083e-08, 1.1539e-06, 1.2212e-07, ..., 4.3074e-08, + 1.5250e-08, -1.8149e-07], + ..., + [-1.3795e-07, 1.1213e-06, 1.2305e-07, ..., -3.0326e-08, + 8.8592e-08, 9.3132e-08], + [-1.1642e-09, 2.4820e-07, 5.6403e-08, ..., 7.8406e-08, + 1.0768e-08, -5.8382e-08], + [-5.2445e-08, 4.8243e-07, -5.0350e-08, ..., 1.7253e-07, + 1.7392e-07, 9.5344e-08]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0323, -0.0242, -0.0091, 0.0170, -0.0026, -0.0153, 0.0008, 0.0126, + -0.0263, 0.0146], device='cuda:0'), grad: tensor([ 2.7660e-06, 9.2387e-06, 1.9241e-06, -1.9088e-05, 1.3597e-07, + -1.4342e-05, 8.3223e-06, 2.6673e-06, 6.0350e-06, 2.3507e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 217.28, cls_loss 0.0014 cls_loss_mapping 0.0034 cls_loss_causal 0.5084 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.14 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0139, -0.0693, 0.0171, ..., -0.0833, -0.1080, -0.0815], + [-0.0959, 0.0578, -0.0772, ..., -0.1001, -0.0861, -0.1528], + [-0.0664, -0.0998, -0.0895, ..., -0.1368, -0.0738, 0.1206], + ..., + [ 0.0670, -0.0393, -0.0841, ..., 0.0709, -0.0764, -0.0741], + [ 0.1037, 0.0768, 0.0491, ..., -0.1033, -0.0852, 0.0688], + [ 0.0561, 0.0153, 0.0690, ..., 0.0447, -0.0680, 0.0091]], + device='cuda:0'), grad: tensor([[-1.6941e-06, 2.7847e-07, -1.0775e-06, ..., 1.0282e-06, + 2.3935e-07, 7.3761e-07], + [ 1.1490e-07, 1.7381e-07, 1.4296e-07, ..., 4.6496e-07, + 8.1141e-08, 2.6613e-07], + [ 1.9395e-07, 2.0361e-07, 2.1281e-07, ..., 4.2375e-07, + 2.0117e-07, 4.4005e-08], + ..., + [ 4.3167e-07, 6.0955e-07, 5.2201e-07, ..., 1.3420e-06, + 1.0221e-07, 6.0303e-07], + [ 2.2771e-07, 2.7195e-07, 2.6124e-07, ..., 1.0980e-06, + 3.7393e-07, 9.0618e-07], + [-3.8780e-06, -6.3442e-06, -6.5230e-06, ..., -4.5955e-05, + -1.1392e-05, -2.8029e-05]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0319, -0.0243, -0.0089, 0.0166, -0.0021, -0.0152, 0.0001, 0.0128, + -0.0256, 0.0141], device='cuda:0'), grad: tensor([-8.4713e-06, 1.2331e-06, 4.1025e-07, 1.1967e-06, 4.5955e-05, + 5.4725e-06, 9.7975e-06, 3.4012e-06, 3.2298e-06, -6.2168e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 217.34, cls_loss 0.0015 cls_loss_mapping 0.0051 cls_loss_causal 0.5004 re_mapping 0.0047 re_causal 0.0139 /// teacc 99.06 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0140, -0.0705, 0.0168, ..., -0.0836, -0.1090, -0.0825], + [-0.0959, 0.0577, -0.0787, ..., -0.1013, -0.0866, -0.1560], + [-0.0665, -0.0987, -0.0901, ..., -0.1363, -0.0736, 0.1234], + ..., + [ 0.0670, -0.0395, -0.0848, ..., 0.0710, -0.0774, -0.0754], + [ 0.1036, 0.0770, 0.0494, ..., -0.1042, -0.0858, 0.0687], + [ 0.0563, 0.0156, 0.0695, ..., 0.0449, -0.0682, 0.0089]], + device='cuda:0'), grad: tensor([[-1.4435e-08, 4.2492e-08, 1.5832e-08, ..., 4.8371e-08, + 6.3097e-08, 9.4296e-08], + [ 3.8708e-08, -2.0408e-07, 3.2131e-08, ..., 8.9640e-09, + 2.0838e-08, 4.1095e-08], + [ 1.4540e-07, 1.1129e-07, 6.4261e-08, ..., 1.6869e-07, + 7.4971e-08, -1.9500e-08], + ..., + [-5.4017e-07, 4.5460e-08, 4.4005e-08, ..., -5.9186e-07, + 1.2689e-08, 4.0745e-08], + [-5.1572e-08, -6.9907e-08, -1.0792e-07, ..., 9.4296e-08, + 5.7218e-08, -2.2061e-08], + [ 1.7462e-07, 1.1176e-08, -1.2247e-07, ..., 1.3877e-07, + 5.3435e-08, -9.8546e-08]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0317, -0.0246, -0.0066, 0.0163, -0.0021, -0.0154, 0.0008, 0.0127, + -0.0262, 0.0142], device='cuda:0'), grad: tensor([ 1.1368e-07, -5.2247e-07, 4.5775e-07, -5.1502e-07, 4.2655e-07, + 5.6997e-07, -3.4645e-07, -1.1735e-06, -1.6007e-08, 9.8720e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 217.35, cls_loss 0.0011 cls_loss_mapping 0.0030 cls_loss_causal 0.5017 re_mapping 0.0046 re_causal 0.0139 /// teacc 99.07 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0141, -0.0706, 0.0168, ..., -0.0837, -0.1093, -0.0829], + [-0.0960, 0.0577, -0.0789, ..., -0.1017, -0.0868, -0.1562], + [-0.0668, -0.0991, -0.0903, ..., -0.1370, -0.0742, 0.1236], + ..., + [ 0.0677, -0.0395, -0.0852, ..., 0.0720, -0.0778, -0.0765], + [ 0.1040, 0.0777, 0.0497, ..., -0.1035, -0.0859, 0.0691], + [ 0.0555, 0.0155, 0.0696, ..., 0.0445, -0.0686, 0.0083]], + device='cuda:0'), grad: tensor([[ 7.8580e-08, 7.7474e-08, 7.4680e-08, ..., 1.4622e-07, + 1.8929e-07, 1.9115e-07], + [ 4.3889e-08, -1.6822e-07, 1.3039e-08, ..., 3.5157e-08, + 2.0373e-08, 5.1456e-08], + [-2.6240e-07, 3.4110e-08, 1.1292e-08, ..., 3.2131e-08, + 5.3144e-08, -8.5682e-07], + ..., + [ 9.6741e-08, 1.0821e-07, 6.7579e-08, ..., 8.4983e-09, + 4.0163e-09, 7.1013e-07], + [ 1.1327e-07, 7.0489e-08, 3.5041e-08, ..., 1.1770e-07, + 1.0076e-07, 1.6997e-07], + [-6.1048e-07, -8.3679e-07, -9.0012e-07, ..., -1.5777e-06, + 5.5064e-08, -2.7567e-07]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0319, -0.0246, -0.0067, 0.0161, -0.0017, -0.0155, 0.0008, 0.0130, + -0.0260, 0.0133], device='cuda:0'), grad: tensor([ 6.3516e-07, -1.9534e-07, -2.8126e-06, 4.8336e-07, 1.8319e-06, + -5.3085e-07, -4.1001e-07, 2.2165e-06, 1.0822e-06, -2.3097e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 217.24, cls_loss 0.0015 cls_loss_mapping 0.0035 cls_loss_causal 0.5156 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.13 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0143, -0.0707, 0.0167, ..., -0.0838, -0.1101, -0.0837], + [-0.0961, 0.0577, -0.0793, ..., -0.1024, -0.0872, -0.1564], + [-0.0672, -0.0999, -0.0906, ..., -0.1377, -0.0749, 0.1235], + ..., + [ 0.0681, -0.0396, -0.0851, ..., 0.0723, -0.0796, -0.0771], + [ 0.1035, 0.0785, 0.0502, ..., -0.1044, -0.0873, 0.0680], + [ 0.0554, 0.0156, 0.0698, ..., 0.0442, -0.0689, 0.0079]], + device='cuda:0'), grad: tensor([[ 9.6159e-08, 6.4401e-07, 1.4633e-07, ..., 2.8475e-07, + 3.6845e-08, 1.6810e-07], + [ 9.9279e-07, 6.0024e-07, 2.6706e-07, ..., 1.1716e-06, + 5.2969e-09, 6.0303e-07], + [ 4.1537e-06, 1.6987e-06, 5.0850e-07, ..., 5.1968e-06, + 5.6461e-09, 2.2408e-06], + ..., + [-1.4871e-05, 2.2259e-06, 2.3167e-07, ..., -1.8537e-05, + 1.2806e-09, -9.0823e-06], + [ 2.8289e-07, 2.8536e-06, 6.7102e-07, ..., 5.4203e-07, + 3.7835e-09, 1.4843e-08], + [ 1.8748e-06, 5.5619e-06, 1.4408e-06, ..., 2.2277e-06, + 8.7894e-09, 1.1660e-06]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0323, -0.0247, -0.0072, 0.0160, -0.0012, -0.0157, 0.0023, 0.0131, + -0.0270, 0.0128], device='cuda:0'), grad: tensor([ 3.1274e-06, 5.9046e-06, 1.8612e-05, -8.4698e-05, 1.6287e-05, + 2.2531e-05, 5.4948e-07, -2.6584e-05, 1.4722e-05, 2.9504e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 216.92, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.5030 re_mapping 0.0046 re_causal 0.0139 /// teacc 99.12 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0144, -0.0708, 0.0167, ..., -0.0838, -0.1105, -0.0842], + [-0.0960, 0.0585, -0.0800, ..., -0.1001, -0.0874, -0.1568], + [-0.0674, -0.1004, -0.0909, ..., -0.1380, -0.0746, 0.1240], + ..., + [ 0.0682, -0.0401, -0.0853, ..., 0.0718, -0.0799, -0.0774], + [ 0.1038, 0.0790, 0.0506, ..., -0.1048, -0.0875, 0.0682], + [ 0.0554, 0.0154, 0.0700, ..., 0.0440, -0.0692, 0.0077]], + device='cuda:0'), grad: tensor([[ 1.6193e-07, 2.6473e-07, 1.0093e-07, ..., 1.6065e-07, + 9.0280e-08, 3.7835e-07], + [ 4.5836e-05, 1.9634e-04, 7.5638e-05, ..., 1.1581e-04, + 7.2585e-08, 2.6584e-05], + [ 6.6881e-08, 1.5786e-07, 5.3609e-08, ..., 1.4133e-07, + -1.7891e-06, -6.2138e-06], + ..., + [ 3.7660e-08, 2.8545e-07, 1.2526e-07, ..., -4.9651e-08, + 1.4901e-08, 1.5181e-07], + [ 9.9931e-07, 1.2629e-06, 6.7241e-07, ..., 7.7207e-07, + 2.4750e-07, 1.0664e-06], + [-4.7028e-05, -2.0206e-04, -7.7903e-05, ..., -1.1927e-04, + 1.1176e-08, -2.7001e-05]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0323, -0.0243, -0.0070, 0.0162, -0.0011, -0.0157, 0.0020, 0.0129, + -0.0269, 0.0123], device='cuda:0'), grad: tensor([ 1.7304e-06, 4.0030e-04, -1.2092e-05, 9.6951e-07, 9.1270e-06, + -1.0915e-05, 1.3061e-05, 7.9814e-07, 7.4133e-06, -4.1056e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 217.17, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5187 re_mapping 0.0044 re_causal 0.0141 /// teacc 99.02 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0145, -0.0710, 0.0172, ..., -0.0839, -0.1110, -0.0849], + [-0.0963, 0.0583, -0.0821, ..., -0.1011, -0.0878, -0.1570], + [-0.0675, -0.1010, -0.0918, ..., -0.1383, -0.0742, 0.1244], + ..., + [ 0.0684, -0.0402, -0.0857, ..., 0.0720, -0.0808, -0.0778], + [ 0.1040, 0.0792, 0.0511, ..., -0.1054, -0.0875, 0.0685], + [ 0.0555, 0.0158, 0.0703, ..., 0.0436, -0.0698, 0.0071]], + device='cuda:0'), grad: tensor([[ 1.9209e-08, 6.4378e-08, 3.1432e-08, ..., 1.4692e-07, + 7.6485e-08, 7.8813e-08], + [ 1.5949e-08, 5.3435e-08, 3.0501e-08, ..., 1.7439e-07, + 8.0210e-08, 7.5786e-08], + [ 4.5868e-08, 9.0338e-08, 3.7253e-08, ..., 1.5658e-07, + 7.4739e-08, 7.1712e-08], + ..., + [-4.7847e-08, 2.6287e-07, 1.1444e-07, ..., 6.1747e-07, + 2.9476e-07, 2.9802e-07], + [ 8.6147e-09, 1.1257e-07, 4.4238e-08, ..., 2.5798e-07, + 1.5087e-07, 1.3353e-07], + [-1.3295e-07, 2.0291e-07, 6.5193e-09, ..., 1.1660e-06, + 6.2631e-07, 5.0897e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0325, -0.0243, -0.0071, 0.0164, -0.0005, -0.0155, 0.0013, 0.0129, + -0.0269, 0.0119], device='cuda:0'), grad: tensor([ 4.5076e-07, 4.1211e-07, 7.5018e-07, 3.0883e-06, -4.8168e-06, + -5.4725e-06, 1.1269e-06, 1.2489e-06, 9.8255e-07, 2.2445e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 217.14, cls_loss 0.0010 cls_loss_mapping 0.0030 cls_loss_causal 0.4895 re_mapping 0.0045 re_causal 0.0139 /// teacc 99.15 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0143, -0.0711, 0.0172, ..., -0.0839, -0.1147, -0.0877], + [-0.0963, 0.0583, -0.0828, ..., -0.1013, -0.0881, -0.1573], + [-0.0677, -0.1013, -0.0921, ..., -0.1387, -0.0750, 0.1244], + ..., + [ 0.0684, -0.0403, -0.0862, ..., 0.0720, -0.0808, -0.0779], + [ 0.1040, 0.0792, 0.0512, ..., -0.1059, -0.0878, 0.0685], + [ 0.0557, 0.0162, 0.0707, ..., 0.0437, -0.0699, 0.0071]], + device='cuda:0'), grad: tensor([[-7.9721e-07, -4.0140e-07, 2.5961e-08, ..., 2.0023e-08, + 6.0303e-07, 7.2783e-07], + [ 2.3632e-08, -2.8312e-07, 3.8068e-08, ..., 1.2340e-07, + 5.6997e-07, 9.5461e-07], + [ 5.0175e-08, 5.5879e-07, 1.8685e-07, ..., 1.4692e-07, + 2.2771e-07, -2.0023e-06], + ..., + [ 2.9919e-08, 2.6426e-07, 4.2142e-08, ..., 5.5530e-08, + 7.8930e-08, 1.0547e-07], + [ 5.0873e-08, 3.7951e-07, 7.8231e-08, ..., 6.1002e-08, + 2.0396e-06, 2.6543e-06], + [ 1.7229e-07, 2.2689e-07, 1.9558e-08, ..., 3.2876e-07, + 4.2259e-07, 4.2375e-07]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0316, -0.0244, -0.0071, 0.0167, -0.0006, -0.0153, 0.0024, 0.0128, + -0.0271, 0.0120], device='cuda:0'), grad: tensor([-1.8198e-06, 9.0990e-07, -2.5295e-06, -2.1607e-06, -5.2294e-07, + 4.0047e-06, -7.0632e-06, 1.2033e-06, 5.8077e-06, 2.1309e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 217.34, cls_loss 0.0013 cls_loss_mapping 0.0034 cls_loss_causal 0.5198 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0168, -0.0681, 0.0172, ..., -0.0840, -0.1150, -0.0855], + [-0.0965, 0.0584, -0.0839, ..., -0.1017, -0.0885, -0.1576], + [-0.0680, -0.1017, -0.0923, ..., -0.1396, -0.0762, 0.1245], + ..., + [ 0.0686, -0.0404, -0.0862, ..., 0.0726, -0.0807, -0.0783], + [ 0.1038, 0.0791, 0.0513, ..., -0.1067, -0.0884, 0.0682], + [ 0.0557, 0.0165, 0.0712, ..., 0.0437, -0.0700, 0.0071]], + device='cuda:0'), grad: tensor([[ 5.9372e-09, 1.2515e-07, 1.3388e-08, ..., 1.2340e-07, + 4.0536e-07, 3.9674e-07], + [ 1.3853e-08, -4.2305e-07, 5.3435e-08, ..., 2.7916e-07, + 1.9022e-07, 2.1921e-07], + [ 2.9569e-08, 6.4168e-07, 6.6240e-08, ..., 1.3772e-07, + 1.1630e-07, 1.0722e-07], + ..., + [-1.7462e-08, 5.8953e-07, 2.1537e-08, ..., 3.4459e-07, + 1.7637e-07, 2.3388e-07], + [-1.2142e-07, -1.0571e-07, -2.2689e-07, ..., 1.7788e-07, + 4.4727e-07, 2.0454e-07], + [ 7.6834e-08, 5.7369e-07, 1.2456e-07, ..., 9.0105e-07, + 4.5542e-07, 6.0257e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0345, -0.0245, -0.0072, 0.0163, -0.0006, -0.0171, 0.0020, 0.0130, + -0.0279, 0.0119], device='cuda:0'), grad: tensor([ 1.1744e-06, -1.1958e-06, 1.8450e-06, -3.1814e-06, -4.4256e-06, + 5.9046e-06, -5.9530e-06, 2.1979e-06, 8.0978e-07, 2.7921e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 216.91, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.5045 re_mapping 0.0040 re_causal 0.0127 /// teacc 99.12 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0168, -0.0682, 0.0164, ..., -0.0841, -0.1159, -0.0867], + [-0.0965, 0.0584, -0.0845, ..., -0.1020, -0.0889, -0.1579], + [-0.0681, -0.1020, -0.0926, ..., -0.1399, -0.0763, 0.1247], + ..., + [ 0.0687, -0.0405, -0.0866, ..., 0.0724, -0.0825, -0.0795], + [ 0.1040, 0.0797, 0.0516, ..., -0.1071, -0.0887, 0.0684], + [ 0.0557, 0.0166, 0.0713, ..., 0.0438, -0.0701, 0.0069]], + device='cuda:0'), grad: tensor([[ 6.4028e-09, 1.2461e-06, 2.7008e-08, ..., 1.2026e-07, + 6.6124e-08, 6.7055e-08], + [ 6.2818e-07, -4.8846e-05, 2.0955e-08, ..., 1.6056e-06, + 3.7253e-09, 1.8626e-08], + [ 1.5448e-07, 3.4690e-05, 3.0501e-08, ..., 2.9895e-07, + 6.1700e-09, -2.0023e-08], + ..., + [-9.3272e-07, 7.4431e-06, 1.4878e-07, ..., -2.6990e-06, + 8.1491e-10, 1.9558e-08], + [-7.0431e-08, 4.7334e-07, -1.1770e-07, ..., 1.2014e-07, + -7.4506e-09, -2.0838e-07], + [-3.2317e-07, -1.7916e-07, -2.9919e-07, ..., -4.2608e-07, + 1.2456e-08, 2.3632e-08]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0341, -0.0245, -0.0072, 0.0161, -0.0004, -0.0169, 0.0030, 0.0130, + -0.0280, 0.0118], device='cuda:0'), grad: tensor([ 4.6864e-06, -1.8525e-04, 1.3578e-04, 1.2279e-05, 5.6513e-06, + 2.0377e-06, 1.3215e-06, 2.0415e-05, 2.0713e-06, 1.0757e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 217.06, cls_loss 0.0010 cls_loss_mapping 0.0029 cls_loss_causal 0.5226 re_mapping 0.0042 re_causal 0.0132 /// teacc 98.96 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0168, -0.0682, 0.0164, ..., -0.0842, -0.1159, -0.0870], + [-0.0966, 0.0586, -0.0849, ..., -0.1022, -0.0891, -0.1577], + [-0.0683, -0.1036, -0.0929, ..., -0.1404, -0.0767, 0.1246], + ..., + [ 0.0689, -0.0406, -0.0870, ..., 0.0719, -0.0853, -0.0808], + [ 0.1043, 0.0809, 0.0523, ..., -0.1074, -0.0887, 0.0687], + [ 0.0558, 0.0168, 0.0714, ..., 0.0439, -0.0702, 0.0069]], + device='cuda:0'), grad: tensor([[ 2.5751e-07, 4.2468e-07, 2.7637e-07, ..., 1.0279e-07, + 1.2340e-08, 1.0375e-06], + [ 1.4575e-06, 8.8150e-07, 6.9570e-07, ..., 1.1045e-06, + 1.0012e-08, 1.7196e-05], + [ 8.2096e-07, 1.3029e-06, 7.8045e-07, ..., 2.4633e-07, + 1.6647e-08, -2.0191e-05], + ..., + [-1.6224e-06, -4.7684e-07, -4.9500e-07, ..., -1.5507e-06, + -1.1525e-08, 1.7285e-06], + [-9.1568e-06, -9.4250e-06, -4.3586e-06, ..., -6.3423e-07, + -1.3374e-06, -1.3679e-05], + [ 1.0068e-06, 1.7118e-06, 8.7125e-07, ..., -6.6007e-08, + 4.7730e-09, 1.9725e-06]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0341, -0.0243, -0.0080, 0.0161, -0.0003, -0.0169, 0.0027, 0.0129, + -0.0277, 0.0118], device='cuda:0'), grad: tensor([ 4.1611e-06, 7.9334e-05, -8.6010e-05, 6.0014e-06, 1.7211e-06, + 4.4890e-06, 2.4647e-05, -2.3786e-06, -3.8147e-05, 6.1803e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 217.10, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.4919 re_mapping 0.0041 re_causal 0.0125 /// teacc 98.95 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0169, -0.0684, 0.0163, ..., -0.0844, -0.1164, -0.0876], + [-0.0967, 0.0586, -0.0860, ..., -0.1023, -0.0893, -0.1583], + [-0.0688, -0.1043, -0.0934, ..., -0.1409, -0.0779, 0.1245], + ..., + [ 0.0687, -0.0407, -0.0882, ..., 0.0716, -0.0856, -0.0813], + [ 0.1048, 0.0818, 0.0530, ..., -0.1084, -0.0892, 0.0691], + [ 0.0563, 0.0176, 0.0721, ..., 0.0443, -0.0703, 0.0069]], + device='cuda:0'), grad: tensor([[-5.0105e-07, 6.7172e-08, 3.0035e-08, ..., 7.8231e-08, + 3.8417e-08, 3.1106e-07], + [ 1.7660e-07, -1.2014e-07, 3.5274e-08, ..., 8.8010e-08, + 1.7812e-08, 4.9500e-07], + [-1.0012e-06, 2.3295e-07, 1.3504e-08, ..., 6.1421e-07, + 1.4273e-07, -4.4331e-06], + ..., + [ 7.1665e-07, 1.5774e-07, 6.4960e-08, ..., -1.2247e-06, + -2.3586e-07, 2.4047e-06], + [ 1.0640e-07, 5.1223e-08, 9.6043e-08, ..., 1.0815e-07, + 3.8184e-08, 3.0454e-07], + [-1.0885e-07, -4.4098e-07, -3.3667e-07, ..., -3.1851e-07, + 4.4703e-08, 1.1118e-07]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0341, -0.0244, -0.0085, 0.0162, -0.0004, -0.0171, 0.0032, 0.0128, + -0.0271, 0.0122], device='cuda:0'), grad: tensor([-1.5600e-07, 1.6643e-06, -1.6868e-05, 6.3814e-06, 1.6596e-06, + -9.2164e-06, 1.2368e-06, 8.7544e-06, 1.9632e-06, 4.5635e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 217.48, cls_loss 0.0013 cls_loss_mapping 0.0044 cls_loss_causal 0.5329 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.05 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0169, -0.0685, 0.0162, ..., -0.0846, -0.1193, -0.0905], + [-0.0968, 0.0586, -0.0866, ..., -0.1027, -0.0895, -0.1587], + [-0.0686, -0.1044, -0.0935, ..., -0.1414, -0.0788, 0.1252], + ..., + [ 0.0687, -0.0408, -0.0888, ..., 0.0719, -0.0858, -0.0819], + [ 0.1054, 0.0826, 0.0536, ..., -0.1089, -0.0895, 0.0693], + [ 0.0564, 0.0186, 0.0728, ..., 0.0449, -0.0702, 0.0074]], + device='cuda:0'), grad: tensor([[ 5.8790e-08, 7.5670e-08, 5.1805e-08, ..., 1.8359e-07, + 9.0804e-08, 1.1199e-07], + [ 8.6939e-07, -9.6159e-08, 2.1036e-07, ..., 1.8589e-06, + 2.2468e-08, 5.1921e-08], + [-1.1092e-06, 1.6124e-07, 6.5076e-08, ..., -5.1735e-07, + -5.6904e-07, -1.9372e-06], + ..., + [-9.8199e-06, 7.2177e-07, -1.9930e-06, ..., -2.5287e-05, + 9.1270e-08, 1.2852e-07], + [ 5.6066e-07, 2.8475e-07, 1.5588e-07, ..., 4.3353e-07, + 4.5355e-07, 9.6299e-07], + [ 8.3447e-06, 2.9639e-07, 2.3898e-06, ..., 2.2411e-05, + 4.2235e-07, 4.6962e-07]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0327, -0.0245, -0.0080, 0.0158, -0.0009, -0.0170, 0.0052, 0.0128, + -0.0270, 0.0126], device='cuda:0'), grad: tensor([ 5.1688e-07, 3.1814e-06, -5.7518e-06, -1.1269e-06, 5.0059e-07, + 8.4937e-07, 6.9733e-08, -4.3005e-05, 3.8296e-06, 4.0889e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 217.02, cls_loss 0.0016 cls_loss_mapping 0.0046 cls_loss_causal 0.5052 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.15 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0168, -0.0686, 0.0162, ..., -0.0850, -0.1195, -0.0905], + [-0.0969, 0.0587, -0.0878, ..., -0.1030, -0.0902, -0.1590], + [-0.0687, -0.1057, -0.0940, ..., -0.1418, -0.0796, 0.1257], + ..., + [ 0.0691, -0.0409, -0.0891, ..., 0.0730, -0.0860, -0.0835], + [ 0.1055, 0.0855, 0.0557, ..., -0.1106, -0.0905, 0.0691], + [ 0.0564, 0.0192, 0.0737, ..., 0.0450, -0.0703, 0.0075]], + device='cuda:0'), grad: tensor([[-4.0838e-07, 4.2189e-07, 7.9977e-08, ..., 2.2317e-07, + 2.7940e-08, 1.4505e-07], + [ 8.5565e-08, -2.6748e-06, 4.1677e-08, ..., 1.2666e-07, + 1.0710e-08, -1.9721e-07], + [ 9.6951e-07, 1.0785e-06, 4.8429e-08, ..., 1.2768e-06, + 7.6019e-08, 1.8894e-07], + ..., + [-1.4137e-06, 6.1840e-07, 7.3691e-08, ..., -1.8803e-06, + 9.0804e-09, -5.7509e-08], + [ 5.5181e-08, 5.1782e-07, 6.8569e-08, ..., 2.4354e-07, + 4.2957e-08, 1.3434e-07], + [ 4.8545e-08, -1.0636e-06, -1.3262e-06, ..., -2.9188e-06, + -8.3121e-08, -1.4976e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0330, -0.0246, -0.0071, 0.0149, -0.0010, -0.0151, 0.0027, 0.0128, + -0.0274, 0.0125], device='cuda:0'), grad: tensor([-1.1653e-07, -8.3670e-06, 5.8562e-06, -8.8476e-07, 5.2452e-06, + 5.1409e-07, 4.9872e-07, -1.6829e-06, 2.0918e-06, -3.1553e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 217.29, cls_loss 0.0013 cls_loss_mapping 0.0028 cls_loss_causal 0.5346 re_mapping 0.0046 re_causal 0.0140 /// teacc 99.15 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0169, -0.0686, 0.0162, ..., -0.0852, -0.1197, -0.0905], + [-0.0970, 0.0595, -0.0883, ..., -0.1032, -0.0908, -0.1593], + [-0.0690, -0.1067, -0.0944, ..., -0.1427, -0.0805, 0.1257], + ..., + [ 0.0693, -0.0414, -0.0898, ..., 0.0734, -0.0863, -0.0843], + [ 0.1057, 0.0857, 0.0560, ..., -0.1119, -0.0914, 0.0686], + [ 0.0564, 0.0191, 0.0742, ..., 0.0450, -0.0705, 0.0075]], + device='cuda:0'), grad: tensor([[-6.6473e-08, 1.1048e-07, 5.7044e-08, ..., 6.8103e-08, + 2.0023e-08, 1.0105e-07], + [ 9.9069e-08, 1.1048e-07, 1.9011e-07, ..., 2.6729e-07, + 7.8464e-08, 4.0326e-07], + [ 7.1130e-08, 2.2817e-07, 5.6461e-08, ..., 8.2701e-07, + 1.8917e-07, 5.0291e-07], + ..., + [ 9.4529e-08, 2.3062e-07, 6.0536e-08, ..., -3.5041e-07, + 9.7207e-08, 2.8918e-07], + [-4.9546e-07, -1.8664e-06, -1.3188e-06, ..., 1.9639e-07, + 1.4668e-08, -1.6782e-06], + [-3.3900e-07, -2.7893e-07, -1.9185e-07, ..., -1.7136e-07, + 1.9791e-07, 2.2980e-07]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0331, -0.0242, -0.0074, 0.0155, -0.0009, -0.0147, 0.0020, 0.0125, + -0.0282, 0.0122], device='cuda:0'), grad: tensor([-3.8301e-07, 7.2503e-07, 2.1886e-06, 8.5076e-07, -1.9073e-06, + 1.6224e-06, 2.8014e-06, -1.0282e-06, -5.1484e-06, 2.9616e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.36, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5266 re_mapping 0.0043 re_causal 0.0131 /// teacc 99.14 lr 0.00010000 +Epoch 180, weight, value: tensor([[ 0.0170, -0.0696, 0.0154, ..., -0.0853, -0.1211, -0.0916], + [-0.0971, 0.0594, -0.0895, ..., -0.1035, -0.0911, -0.1604], + [-0.0690, -0.1070, -0.0949, ..., -0.1432, -0.0809, 0.1265], + ..., + [ 0.0695, -0.0415, -0.0902, ..., 0.0738, -0.0867, -0.0852], + [ 0.1059, 0.0871, 0.0570, ..., -0.1133, -0.0913, 0.0690], + [ 0.0563, 0.0194, 0.0747, ..., 0.0451, -0.0709, 0.0071]], + device='cuda:0'), grad: tensor([[-7.9256e-07, 8.2189e-07, -3.0594e-07, ..., 2.6380e-07, + -5.5600e-07, -2.6636e-07], + [ 2.4261e-07, 2.3155e-07, 1.8894e-07, ..., 1.6566e-07, + 9.7207e-08, 3.2363e-07], + [ 5.3085e-07, 7.7533e-07, 4.3493e-07, ..., 1.1048e-07, + 9.6508e-08, 5.7369e-07], + ..., + [ 2.4866e-07, 3.5181e-07, 1.9406e-07, ..., 7.0082e-07, + 3.7439e-07, 5.7975e-07], + [-6.2808e-06, -8.3148e-06, -4.1798e-06, ..., 1.2852e-07, + 2.3399e-08, -7.1153e-06], + [-1.5518e-07, 2.1781e-07, -3.3760e-09, ..., 2.3358e-06, + 2.1942e-06, 2.1439e-06]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0322, -0.0244, -0.0066, 0.0156, -0.0009, -0.0148, 0.0029, 0.0125, + -0.0276, 0.0121], device='cuda:0'), grad: tensor([-2.3376e-06, 1.1502e-06, 2.9709e-06, 5.2638e-06, -6.4895e-06, + 9.0078e-06, 1.0662e-05, 2.2575e-06, -2.8908e-05, 6.4038e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 217.04, cls_loss 0.0011 cls_loss_mapping 0.0034 cls_loss_causal 0.4821 re_mapping 0.0045 re_causal 0.0135 /// teacc 99.13 lr 0.00010000 +Epoch 181, weight, value: tensor([[ 0.0172, -0.0698, 0.0152, ..., -0.0854, -0.1212, -0.0917], + [-0.0970, 0.0597, -0.0901, ..., -0.1033, -0.0914, -0.1606], + [-0.0694, -0.1077, -0.0954, ..., -0.1440, -0.0812, 0.1264], + ..., + [ 0.0693, -0.0419, -0.0916, ..., 0.0740, -0.0871, -0.0857], + [ 0.1080, 0.0897, 0.0595, ..., -0.1151, -0.0913, 0.0701], + [ 0.0557, 0.0177, 0.0735, ..., 0.0453, -0.0710, 0.0061]], + device='cuda:0'), grad: tensor([[-4.7195e-07, 3.1199e-08, -7.5321e-08, ..., 6.7404e-08, + 2.3865e-08, 4.5751e-08], + [ 9.2248e-07, 1.2794e-07, 1.4319e-08, ..., 2.2650e-06, + -1.3621e-08, 3.0966e-07], + [ 2.2119e-07, 3.1898e-08, 2.0373e-08, ..., 7.8464e-08, + 8.3819e-09, 6.1700e-09], + ..., + [-2.0266e-06, -1.0421e-06, -2.2526e-07, ..., -7.4953e-06, + 8.2655e-09, -5.0571e-07], + [ 4.5169e-08, 5.7858e-08, 3.6089e-08, ..., 1.4610e-07, + 1.3737e-08, 2.8638e-08], + [ 2.5611e-07, -2.1013e-07, -3.2363e-07, ..., 3.4180e-07, + 1.1874e-08, -1.8382e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0322, -0.0242, -0.0068, 0.0157, -0.0012, -0.0150, 0.0028, 0.0124, + -0.0258, 0.0109], device='cuda:0'), grad: tensor([-2.2948e-06, 2.3600e-06, 1.0915e-06, 7.7337e-06, 2.2687e-06, + 1.5856e-07, 1.8289e-07, -1.2912e-05, 3.4133e-07, 1.1204e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 217.32, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.4922 re_mapping 0.0042 re_causal 0.0131 /// teacc 99.11 lr 0.00010000 +Epoch 182, weight, value: tensor([[ 0.0170, -0.0700, 0.0148, ..., -0.0859, -0.1212, -0.0919], + [-0.0971, 0.0598, -0.0907, ..., -0.1040, -0.0921, -0.1612], + [-0.0694, -0.1080, -0.0956, ..., -0.1446, -0.0826, 0.1264], + ..., + [ 0.0692, -0.0421, -0.0926, ..., 0.0739, -0.0879, -0.0871], + [ 0.1083, 0.0898, 0.0597, ..., -0.1164, -0.0914, 0.0702], + [ 0.0562, 0.0185, 0.0738, ..., 0.0458, -0.0716, 0.0063]], + device='cuda:0'), grad: tensor([[-7.8115e-08, 1.2899e-07, 7.6485e-08, ..., 1.2014e-07, + 3.7369e-08, 5.5297e-08], + [ 3.4692e-08, -1.3865e-07, 9.8953e-09, ..., 1.5984e-07, + 7.8930e-08, 9.2667e-08], + [ 3.7556e-07, 3.7393e-07, 2.1886e-08, ..., 1.3597e-06, + 5.8208e-08, 7.5297e-07], + ..., + [-4.1537e-07, 1.9895e-07, 4.3539e-08, ..., -1.5767e-06, + 2.9686e-08, -7.9675e-07], + [ 1.5495e-07, 2.5565e-07, 1.1176e-07, ..., 1.8626e-07, + 1.1350e-07, 1.6496e-07], + [-2.1805e-07, -3.0105e-07, -3.2014e-07, ..., -1.6589e-07, + 1.6624e-07, 8.0676e-08]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0320, -0.0243, -0.0069, 0.0154, -0.0012, -0.0147, 0.0028, 0.0123, + -0.0258, 0.0111], device='cuda:0'), grad: tensor([-1.2212e-07, -2.1502e-07, 4.2655e-06, -1.4892e-06, -3.5437e-07, + 4.0070e-07, -2.6682e-07, -3.3639e-06, 1.3523e-06, -2.1444e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 217.28, cls_loss 0.0011 cls_loss_mapping 0.0027 cls_loss_causal 0.4907 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.07 lr 0.00010000 +Epoch 183, weight, value: tensor([[ 0.0172, -0.0700, 0.0151, ..., -0.0861, -0.1208, -0.0913], + [-0.0973, 0.0597, -0.0911, ..., -0.1044, -0.0924, -0.1615], + [-0.0693, -0.1083, -0.0957, ..., -0.1455, -0.0830, 0.1267], + ..., + [ 0.0694, -0.0422, -0.0929, ..., 0.0745, -0.0879, -0.0876], + [ 0.1083, 0.0900, 0.0597, ..., -0.1175, -0.0916, 0.0701], + [ 0.0563, 0.0192, 0.0744, ..., 0.0460, -0.0714, 0.0068]], + device='cuda:0'), grad: tensor([[-2.3737e-07, 6.4611e-08, 7.6834e-09, ..., 3.3481e-07, + 5.0059e-09, 1.9791e-08], + [ 1.3004e-07, -5.3644e-07, 1.8510e-08, ..., 1.8894e-07, + 4.6566e-09, 4.1910e-08], + [ 1.2165e-07, 1.6601e-07, 3.7020e-08, ..., 5.8440e-08, + 1.5949e-08, -1.0431e-07], + ..., + [-4.2049e-07, 1.9674e-07, 1.3388e-08, ..., -1.0068e-06, + -3.9581e-09, 3.9581e-08], + [-1.5032e-06, -1.8291e-06, -4.0117e-07, ..., 3.9698e-08, + 2.3632e-08, -2.8918e-07], + [ 5.1456e-07, 4.8289e-07, 4.4121e-08, ..., -1.9826e-07, + -4.2259e-08, -1.5215e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0326, -0.0245, -0.0060, 0.0149, -0.0020, -0.0147, 0.0026, 0.0122, + -0.0259, 0.0116], device='cuda:0'), grad: tensor([-9.2853e-07, -6.9104e-07, 1.5739e-07, 1.2424e-06, 4.9360e-07, + 4.1574e-06, 2.4075e-07, -2.1141e-06, -4.4182e-06, 1.8422e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 217.01, cls_loss 0.0011 cls_loss_mapping 0.0034 cls_loss_causal 0.5070 re_mapping 0.0045 re_causal 0.0137 /// teacc 99.14 lr 0.00010000 +Epoch 184, weight, value: tensor([[ 0.0175, -0.0701, 0.0150, ..., -0.0863, -0.1201, -0.0905], + [-0.0973, 0.0598, -0.0916, ..., -0.1047, -0.0928, -0.1620], + [-0.0698, -0.1098, -0.0961, ..., -0.1471, -0.0836, 0.1268], + ..., + [ 0.0697, -0.0423, -0.0934, ..., 0.0753, -0.0879, -0.0879], + [ 0.1088, 0.0905, 0.0599, ..., -0.1180, -0.0915, 0.0706], + [ 0.0562, 0.0191, 0.0747, ..., 0.0455, -0.0719, 0.0060]], + device='cuda:0'), grad: tensor([[ 7.2177e-09, 3.6089e-07, 2.0070e-07, ..., 3.3644e-08, + 3.7136e-08, 3.6019e-07], + [ 3.5390e-07, -2.8294e-06, 3.9255e-07, ..., -1.3865e-07, + 7.9512e-08, 6.2026e-07], + [ 1.1548e-06, 4.4480e-06, 1.4380e-06, ..., 1.4156e-07, + 1.8580e-07, 2.1178e-06], + ..., + [-1.0943e-07, 7.3947e-07, 1.6857e-07, ..., -3.4831e-07, + 7.8580e-08, 3.1665e-07], + [-3.5260e-06, -4.9695e-06, -3.8370e-06, ..., 5.4948e-08, + -4.8429e-07, -6.9141e-06], + [ 2.7148e-07, 5.9837e-07, 1.8440e-07, ..., 9.1828e-07, + 4.3400e-07, 6.8359e-07]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0331, -0.0245, -0.0068, 0.0148, -0.0016, -0.0147, 0.0019, 0.0124, + -0.0254, 0.0112], device='cuda:0'), grad: tensor([ 8.2143e-07, -2.7493e-05, 3.6567e-05, -1.0040e-06, 1.2498e-06, + 4.5113e-06, 5.8264e-06, 2.7660e-06, -2.7195e-05, 3.9861e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 217.26, cls_loss 0.0011 cls_loss_mapping 0.0030 cls_loss_causal 0.4804 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.07 lr 0.00010000 +Epoch 185, weight, value: tensor([[ 0.0175, -0.0702, 0.0150, ..., -0.0867, -0.1201, -0.0906], + [-0.0975, 0.0600, -0.0922, ..., -0.1053, -0.0931, -0.1622], + [-0.0701, -0.1109, -0.0966, ..., -0.1478, -0.0841, 0.1268], + ..., + [ 0.0700, -0.0424, -0.0937, ..., 0.0760, -0.0879, -0.0886], + [ 0.1090, 0.0907, 0.0600, ..., -0.1189, -0.0917, 0.0709], + [ 0.0563, 0.0191, 0.0749, ..., 0.0455, -0.0721, 0.0059]], + device='cuda:0'), grad: tensor([[ 1.3376e-07, 1.2806e-07, 9.9419e-08, ..., 2.3784e-07, + 5.5530e-08, 9.5344e-08], + [ 5.3784e-07, 7.9861e-08, 6.4494e-08, ..., 7.1572e-07, + 8.8476e-09, 6.4727e-08], + [ 1.0217e-06, 3.4575e-08, 1.3853e-08, ..., 1.1995e-06, + 1.4668e-08, 1.2654e-07], + ..., + [-3.1944e-06, 8.0443e-08, 5.9721e-08, ..., -3.9078e-06, + 2.3283e-10, -2.2550e-07], + [ 7.2550e-07, 3.5902e-07, 2.3888e-07, ..., 9.2294e-07, + 4.0862e-08, 1.9628e-07], + [ 2.4633e-07, -7.6834e-07, -6.7800e-07, ..., -8.8010e-08, + 1.3970e-09, -2.9267e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0330, -0.0245, -0.0070, 0.0152, -0.0014, -0.0148, 0.0019, 0.0125, + -0.0253, 0.0110], device='cuda:0'), grad: tensor([ 9.2201e-07, 3.3956e-06, 6.6906e-06, 2.3022e-06, 1.4687e-06, + -3.7458e-06, 5.3644e-07, -1.7941e-05, 6.6385e-06, -3.0361e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 184---------------------------------------------------- +epoch 184, time 217.82, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.5228 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.23 lr 0.00010000 +Epoch 186, weight, value: tensor([[ 0.0176, -0.0703, 0.0149, ..., -0.0868, -0.1202, -0.0907], + [-0.0977, 0.0602, -0.0927, ..., -0.1060, -0.0936, -0.1626], + [-0.0705, -0.1116, -0.0971, ..., -0.1487, -0.0864, 0.1263], + ..., + [ 0.0704, -0.0426, -0.0942, ..., 0.0765, -0.0880, -0.0889], + [ 0.1091, 0.0908, 0.0601, ..., -0.1201, -0.0918, 0.0711], + [ 0.0562, 0.0192, 0.0752, ..., 0.0456, -0.0722, 0.0060]], + device='cuda:0'), grad: tensor([[-2.3609e-07, 7.6368e-08, 1.8044e-08, ..., 3.7253e-08, + 3.3528e-08, 1.1327e-07], + [ 1.1851e-07, 1.3970e-08, 6.3563e-08, ..., 2.7008e-07, + 1.6880e-08, 2.4447e-08], + [ 2.3225e-07, 1.0338e-07, 5.5064e-08, ..., 6.7102e-07, + 2.0838e-08, -3.3155e-07], + ..., + [-9.0664e-07, 2.0757e-07, 6.1817e-08, ..., -2.3525e-06, + 4.0745e-09, -6.9733e-08], + [ 5.1456e-08, 3.2084e-07, 6.7637e-08, ..., 9.1735e-08, + 5.3784e-08, 2.3609e-07], + [ 5.0617e-07, 2.0303e-07, 4.2492e-08, ..., 8.0839e-07, + 1.2107e-08, 4.3656e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0332, -0.0246, -0.0073, 0.0152, -0.0014, -0.0147, 0.0019, 0.0126, + -0.0253, 0.0110], device='cuda:0'), grad: tensor([-4.4331e-07, 5.3737e-07, 1.2796e-06, -2.9355e-06, 6.8266e-07, + 5.8627e-07, 2.8196e-07, -4.1053e-06, 1.4501e-06, 2.6543e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 217.03, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.5052 re_mapping 0.0042 re_causal 0.0131 /// teacc 99.13 lr 0.00010000 +Epoch 187, weight, value: tensor([[ 0.0177, -0.0700, 0.0146, ..., -0.0870, -0.1206, -0.0904], + [-0.0978, 0.0602, -0.0931, ..., -0.1068, -0.0943, -0.1631], + [-0.0708, -0.1120, -0.0972, ..., -0.1495, -0.0868, 0.1264], + ..., + [ 0.0706, -0.0427, -0.0947, ..., 0.0766, -0.0888, -0.0897], + [ 0.1092, 0.0910, 0.0601, ..., -0.1216, -0.0919, 0.0713], + [ 0.0563, 0.0189, 0.0758, ..., 0.0479, -0.0707, 0.0080]], + device='cuda:0'), grad: tensor([[ 2.4820e-07, 9.8255e-08, 5.6345e-08, ..., 4.5192e-07, + 2.1444e-07, 2.1921e-07], + [ 4.3679e-07, 1.3213e-07, 6.1584e-08, ..., 7.7859e-07, + 3.5949e-07, 3.2503e-07], + [ 2.1060e-07, 5.8440e-08, 2.6659e-08, ..., 4.9965e-07, + 2.6822e-07, 1.8708e-07], + ..., + [ 2.6554e-05, 5.3570e-06, 2.5406e-06, ..., 1.8269e-05, + 2.5518e-07, 2.7046e-06], + [ 8.0746e-07, 1.9453e-07, 1.5413e-07, ..., 1.0049e-06, + 1.9569e-07, 2.9802e-07], + [-2.9176e-05, -6.2510e-06, -3.0808e-06, ..., -1.3262e-05, + 5.3123e-06, 6.3330e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0334, -0.0247, -0.0073, 0.0151, -0.0036, -0.0152, 0.0027, 0.0125, + -0.0253, 0.0131], device='cuda:0'), grad: tensor([ 2.1216e-06, 2.9672e-06, 1.7239e-06, 1.6829e-06, -1.6257e-05, + -9.1642e-06, 3.0492e-06, 6.6102e-05, 3.7216e-06, -5.6148e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 216.92, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4983 re_mapping 0.0042 re_causal 0.0130 /// teacc 99.14 lr 0.00010000 +Epoch 188, weight, value: tensor([[ 0.0177, -0.0701, 0.0145, ..., -0.0871, -0.1206, -0.0905], + [-0.0980, 0.0604, -0.0936, ..., -0.1074, -0.0947, -0.1631], + [-0.0709, -0.1131, -0.0976, ..., -0.1500, -0.0864, 0.1269], + ..., + [ 0.0711, -0.0427, -0.0949, ..., 0.0776, -0.0889, -0.0899], + [ 0.1093, 0.0911, 0.0602, ..., -0.1223, -0.0919, 0.0714], + [ 0.0561, 0.0188, 0.0758, ..., 0.0478, -0.0710, 0.0077]], + device='cuda:0'), grad: tensor([[-1.2328e-07, -3.4925e-09, -4.7381e-08, ..., 1.0896e-07, + 9.6159e-08, 1.0629e-07], + [ 5.2340e-07, 2.1304e-08, 1.5250e-08, ..., 6.8452e-07, + 6.9384e-08, 1.0710e-07], + [ 8.5123e-07, 4.7614e-08, 1.0361e-08, ..., 1.4603e-06, + 4.2701e-07, 5.7369e-07], + ..., + [-3.2205e-06, 4.3074e-08, 1.1525e-08, ..., -4.7646e-06, + -6.0210e-07, -8.0699e-07], + [ 1.1874e-07, -2.6915e-07, 4.4936e-08, ..., 2.2305e-07, + -6.0536e-09, -3.9185e-07], + [ 1.2852e-06, 7.8580e-08, 4.2841e-08, ..., 1.5451e-06, + 6.7404e-08, 1.4319e-07]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0336, -0.0247, -0.0079, 0.0154, -0.0035, -0.0153, 0.0027, 0.0128, + -0.0253, 0.0128], device='cuda:0'), grad: tensor([-3.6974e-07, 2.3991e-06, 4.4070e-06, -2.3574e-07, 1.1642e-06, + -1.8915e-06, 1.3486e-06, -1.3188e-05, 1.4557e-06, 4.8764e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 217.09, cls_loss 0.0011 cls_loss_mapping 0.0036 cls_loss_causal 0.4733 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.03 lr 0.00010000 +Epoch 189, weight, value: tensor([[ 0.0178, -0.0701, 0.0145, ..., -0.0872, -0.1208, -0.0907], + [-0.0981, 0.0592, -0.0938, ..., -0.1077, -0.0950, -0.1652], + [-0.0712, -0.1116, -0.0975, ..., -0.1507, -0.0873, 0.1284], + ..., + [ 0.0715, -0.0430, -0.0952, ..., 0.0787, -0.0894, -0.0906], + [ 0.1094, 0.0912, 0.0602, ..., -0.1236, -0.0920, 0.0714], + [ 0.0561, 0.0190, 0.0761, ..., 0.0478, -0.0711, 0.0078]], + device='cuda:0'), grad: tensor([[-1.5553e-06, -2.4009e-06, -1.4447e-07, ..., 3.1781e-08, + 2.0606e-08, -1.8859e-06], + [ 6.3330e-08, 1.8964e-07, 3.6205e-08, ..., 7.7882e-08, + 2.6310e-08, 9.3365e-08], + [ 1.0454e-07, 2.1071e-07, 2.2701e-08, ..., 4.2259e-08, + 1.9209e-08, 1.4040e-07], + ..., + [ 3.9465e-08, 2.5728e-07, 4.5169e-08, ..., 2.9802e-08, + 1.1758e-08, 8.3237e-08], + [ 1.2142e-07, 2.4727e-07, 3.5157e-08, ..., 5.3667e-08, + 1.6065e-08, 1.5646e-07], + [ 5.3691e-07, 1.0449e-06, 3.5507e-08, ..., 1.1874e-07, + 2.1223e-07, 8.5402e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0336, -0.0260, -0.0058, 0.0150, -0.0037, -0.0152, 0.0028, 0.0130, + -0.0254, 0.0128], device='cuda:0'), grad: tensor([-1.4000e-05, 8.1677e-07, 1.0896e-06, -7.3342e-07, -4.2794e-07, + 1.3839e-06, 3.6955e-06, 8.7405e-07, 1.2275e-06, 6.0722e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 217.36, cls_loss 0.0010 cls_loss_mapping 0.0032 cls_loss_causal 0.5232 re_mapping 0.0043 re_causal 0.0131 /// teacc 99.07 lr 0.00010000 +Epoch 190, weight, value: tensor([[ 0.0178, -0.0702, 0.0145, ..., -0.0874, -0.1209, -0.0909], + [-0.0986, 0.0596, -0.0941, ..., -0.1093, -0.0960, -0.1652], + [-0.0710, -0.1118, -0.0978, ..., -0.1514, -0.0899, 0.1281], + ..., + [ 0.0722, -0.0433, -0.0954, ..., 0.0801, -0.0903, -0.0912], + [ 0.1094, 0.0913, 0.0603, ..., -0.1245, -0.0922, 0.0714], + [ 0.0558, 0.0190, 0.0762, ..., 0.0477, -0.0713, 0.0076]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, -1.0282e-06, 1.8510e-08, ..., 2.7055e-07, + 8.2422e-08, -9.6019e-07], + [ 5.3551e-08, -1.3970e-09, 1.6298e-08, ..., 2.0314e-07, + 9.5111e-08, 2.8522e-07], + [-1.4110e-07, 2.0838e-07, 1.0943e-08, ..., 1.1094e-07, + 8.1374e-08, -3.2922e-07], + ..., + [ 1.8044e-07, 2.5146e-07, 8.8126e-08, ..., 1.0226e-06, + 3.2713e-07, 7.1339e-07], + [ 5.2620e-08, 5.0408e-08, -2.7940e-09, ..., 2.2445e-07, + 1.3981e-07, 1.9895e-07], + [-2.7404e-07, -9.8348e-07, -3.5018e-07, ..., -3.0585e-06, + 1.8009e-07, -1.4836e-06]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0337, -0.0261, -0.0061, 0.0147, -0.0036, -0.0151, 0.0031, 0.0134, + -0.0255, 0.0126], device='cuda:0'), grad: tensor([-1.2435e-05, 1.2834e-06, -2.2131e-07, 5.4529e-07, 2.0899e-06, + 4.4098e-07, 6.0312e-06, 3.6992e-06, 1.6196e-06, -3.0529e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 217.21, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.4996 re_mapping 0.0039 re_causal 0.0123 /// teacc 99.02 lr 0.00010000 +Epoch 191, weight, value: tensor([[ 0.0175, -0.0711, 0.0131, ..., -0.0874, -0.1210, -0.0928], + [-0.0987, 0.0597, -0.0950, ..., -0.1097, -0.0966, -0.1656], + [-0.0713, -0.1121, -0.0985, ..., -0.1522, -0.0902, 0.1285], + ..., + [ 0.0725, -0.0434, -0.0958, ..., 0.0804, -0.0910, -0.0923], + [ 0.1104, 0.0922, 0.0615, ..., -0.1254, -0.0925, 0.0726], + [ 0.0558, 0.0192, 0.0766, ..., 0.0477, -0.0714, 0.0076]], + device='cuda:0'), grad: tensor([[-5.5227e-07, 1.1141e-07, -4.2003e-07, ..., 1.8545e-07, + 1.9092e-08, 5.2038e-08], + [ 7.5786e-08, 9.1852e-08, 3.8999e-08, ..., 3.0082e-07, + 6.7579e-08, 1.1263e-07], + [ 4.3423e-08, 7.7591e-08, 3.4517e-08, ..., 6.6415e-08, + 3.8708e-08, -5.5472e-08], + ..., + [ 6.9709e-07, 7.0874e-07, 5.0897e-07, ..., 1.2517e-06, + 7.0257e-08, 2.3236e-07], + [ 3.8708e-08, 1.4086e-07, 4.0978e-08, ..., 1.2992e-07, + 2.2235e-08, 8.0501e-08], + [-7.9051e-06, -2.0698e-05, -2.6133e-06, ..., -3.0845e-05, + -1.5665e-06, -1.2077e-05]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0333, -0.0261, -0.0060, 0.0142, -0.0035, -0.0140, 0.0017, 0.0134, + -0.0254, 0.0125], device='cuda:0'), grad: tensor([-7.0892e-06, 6.7428e-07, 7.6275e-07, 1.4780e-06, 7.4744e-05, + 3.0305e-06, 2.6152e-06, 3.1423e-06, 1.0459e-06, -8.0407e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 217.05, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.5032 re_mapping 0.0038 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 192, weight, value: tensor([[ 0.0176, -0.0712, 0.0131, ..., -0.0876, -0.1210, -0.0930], + [-0.0988, 0.0598, -0.0954, ..., -0.1101, -0.0973, -0.1659], + [-0.0717, -0.1123, -0.0990, ..., -0.1531, -0.0905, 0.1285], + ..., + [ 0.0729, -0.0435, -0.0963, ..., 0.0810, -0.0915, -0.0923], + [ 0.1105, 0.0922, 0.0615, ..., -0.1270, -0.0928, 0.0725], + [ 0.0559, 0.0201, 0.0770, ..., 0.0477, -0.0719, 0.0076]], + device='cuda:0'), grad: tensor([[ 1.4668e-08, 5.2096e-08, 8.3179e-08, ..., 2.5379e-08, + 2.0361e-07, 1.7160e-07], + [ 1.8510e-08, -1.0099e-07, 4.6624e-08, ..., -5.8208e-11, + 8.6846e-08, 9.0105e-08], + [ 8.9232e-08, 1.5786e-07, 8.5391e-08, ..., 1.0949e-07, + 7.9803e-08, 1.1834e-07], + ..., + [-1.1036e-07, 9.1677e-08, 2.5204e-08, ..., -4.4610e-07, + -6.1293e-08, -6.3505e-08], + [-3.0710e-07, -9.3319e-07, -3.5460e-07, ..., 6.6124e-08, + 5.5996e-08, -4.2515e-07], + [-1.9150e-08, -2.1653e-08, -1.8161e-07, ..., -5.5705e-08, + 3.9756e-08, -2.2526e-08]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0334, -0.0262, -0.0062, 0.0144, -0.0036, -0.0140, 0.0016, 0.0136, + -0.0256, 0.0127], device='cuda:0'), grad: tensor([ 7.2503e-07, 1.1991e-07, 8.4192e-07, 1.8226e-06, 1.2079e-06, + -1.5534e-06, -2.6878e-06, 6.3516e-07, -1.3635e-06, 2.3760e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 217.30, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.4876 re_mapping 0.0041 re_causal 0.0129 /// teacc 99.05 lr 0.00010000 +Epoch 193, weight, value: tensor([[ 0.0177, -0.0712, 0.0131, ..., -0.0877, -0.1208, -0.0929], + [-0.0997, 0.0594, -0.0968, ..., -0.1120, -0.0982, -0.1656], + [-0.0718, -0.1125, -0.0994, ..., -0.1534, -0.0908, 0.1287], + ..., + [ 0.0738, -0.0437, -0.0966, ..., 0.0833, -0.0918, -0.0934], + [ 0.1106, 0.0920, 0.0614, ..., -0.1285, -0.0930, 0.0719], + [ 0.0560, 0.0215, 0.0784, ..., 0.0478, -0.0720, 0.0079]], + device='cuda:0'), grad: tensor([[-2.5448e-07, 1.2817e-07, 6.7288e-08, ..., -6.0187e-08, + 7.4040e-08, 1.7288e-07], + [ 6.4634e-07, 5.5600e-07, 3.0268e-07, ..., -1.3621e-08, + 3.3807e-07, 6.2212e-07], + [ 5.1549e-07, 2.3411e-07, 2.6869e-07, ..., 7.3342e-09, + 2.7893e-07, -4.3097e-07], + ..., + [ 7.0333e-06, 7.6443e-06, 3.2280e-06, ..., 6.1700e-08, + 3.9004e-06, 6.3144e-06], + [-1.0841e-05, -1.1824e-05, -5.1856e-06, ..., 1.8976e-08, + -5.9977e-06, -9.5963e-06], + [ 1.7509e-07, 1.5448e-07, -6.4028e-09, ..., -9.8953e-08, + 5.3551e-08, 2.6962e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0336, -0.0262, -0.0065, 0.0143, -0.0036, -0.0139, 0.0012, 0.0140, + -0.0262, 0.0128], device='cuda:0'), grad: tensor([-6.7707e-07, 2.6673e-06, -4.7684e-07, 1.0654e-05, 8.9221e-07, + 9.8906e-07, 6.3563e-07, 3.1680e-05, -4.7863e-05, 1.4454e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 217.30, cls_loss 0.0009 cls_loss_mapping 0.0035 cls_loss_causal 0.5346 re_mapping 0.0039 re_causal 0.0130 /// teacc 99.02 lr 0.00010000 +Epoch 194, weight, value: tensor([[ 0.0177, -0.0712, 0.0132, ..., -0.0878, -0.1210, -0.0933], + [-0.0997, 0.0596, -0.0971, ..., -0.1116, -0.0993, -0.1659], + [-0.0718, -0.1126, -0.0997, ..., -0.1538, -0.0913, 0.1291], + ..., + [ 0.0738, -0.0441, -0.0973, ..., 0.0833, -0.0923, -0.0940], + [ 0.1108, 0.0923, 0.0617, ..., -0.1292, -0.0932, 0.0719], + [ 0.0560, 0.0216, 0.0788, ..., 0.0477, -0.0722, 0.0078]], + device='cuda:0'), grad: tensor([[-1.7171e-07, 2.7590e-08, 1.4901e-08, ..., 2.7823e-08, + 8.6427e-07, 4.2981e-07], + [ 9.4878e-08, -2.0140e-08, 2.2352e-08, ..., 6.7521e-08, + 5.1036e-07, 2.7195e-07], + [ 9.2550e-08, 1.3621e-08, 5.0059e-09, ..., 1.1153e-07, + 2.1898e-07, 1.1118e-07], + ..., + [-1.7695e-07, 3.2014e-08, 7.1013e-09, ..., -2.7474e-07, + 2.9453e-08, 2.4680e-08], + [-4.4564e-07, -4.4797e-07, -2.3155e-07, ..., 8.8243e-08, + 2.4447e-07, -1.6193e-07], + [ 3.8673e-07, 2.6543e-07, 9.1270e-08, ..., 2.2503e-07, + 2.8755e-07, 2.8452e-07]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0336, -0.0262, -0.0063, 0.0147, -0.0035, -0.0140, 0.0014, 0.0138, + -0.0262, 0.0127], device='cuda:0'), grad: tensor([-2.2128e-06, 2.0489e-06, 1.3076e-06, 2.0713e-06, 3.9898e-06, + -1.5218e-06, -8.2403e-06, -4.6194e-07, 1.5402e-07, 2.8610e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 217.10, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.4666 re_mapping 0.0040 re_causal 0.0120 /// teacc 98.88 lr 0.00010000 +Epoch 195, weight, value: tensor([[ 0.0181, -0.0710, 0.0132, ..., -0.0879, -0.1209, -0.0927], + [-0.0998, 0.0598, -0.0973, ..., -0.1118, -0.1006, -0.1663], + [-0.0717, -0.1128, -0.0996, ..., -0.1542, -0.0919, 0.1297], + ..., + [ 0.0736, -0.0444, -0.0981, ..., 0.0831, -0.0935, -0.0954], + [ 0.1108, 0.0924, 0.0618, ..., -0.1317, -0.0935, 0.0717], + [ 0.0564, 0.0225, 0.0792, ..., 0.0485, -0.0716, 0.0084]], + device='cuda:0'), grad: tensor([[ 3.4925e-08, 1.8708e-07, 8.4634e-08, ..., 7.0897e-08, + 3.1292e-07, 1.0189e-06], + [ 6.8452e-08, 4.1444e-07, 1.8510e-07, ..., 2.0803e-07, + 1.3388e-08, 1.6857e-07], + [ 4.4587e-08, 2.6496e-07, 8.6031e-08, ..., 9.3598e-08, + 3.9814e-08, -7.3127e-06], + ..., + [ 3.5274e-08, 1.4901e-06, 4.8103e-07, ..., 4.7125e-07, + 4.6566e-09, 3.8883e-07], + [-1.6838e-06, -8.5263e-07, -8.2562e-07, ..., 1.8103e-07, + -2.3399e-08, 1.5963e-06], + [ 1.3248e-07, 2.0210e-06, 6.7987e-07, ..., 1.8801e-07, + 1.4319e-08, 3.6787e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0340, -0.0261, -0.0060, 0.0144, -0.0042, -0.0141, 0.0015, 0.0136, + -0.0264, 0.0133], device='cuda:0'), grad: tensor([ 4.1053e-06, 1.7062e-06, -2.6360e-05, -1.0192e-05, 2.2314e-06, + 5.4538e-06, 2.0880e-06, 6.0461e-06, 8.0243e-06, 6.9141e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 217.06, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.5112 re_mapping 0.0037 re_causal 0.0120 /// teacc 98.94 lr 0.00010000 +Epoch 196, weight, value: tensor([[ 0.0183, -0.0711, 0.0133, ..., -0.0880, -0.1209, -0.0928], + [-0.1000, 0.0599, -0.0974, ..., -0.1122, -0.1011, -0.1665], + [-0.0717, -0.1130, -0.1000, ..., -0.1547, -0.0924, 0.1302], + ..., + [ 0.0736, -0.0446, -0.0985, ..., 0.0831, -0.0940, -0.0965], + [ 0.1111, 0.0927, 0.0620, ..., -0.1326, -0.0937, 0.0718], + [ 0.0565, 0.0234, 0.0793, ..., 0.0487, -0.0722, 0.0084]], + device='cuda:0'), grad: tensor([[ 3.1898e-08, 1.3993e-07, -9.5228e-08, ..., 8.1491e-08, + 1.4796e-07, 1.4389e-07], + [ 7.5321e-08, -3.0035e-07, -2.2119e-08, ..., 8.6729e-08, + 3.8417e-08, 3.1502e-07], + [-2.9616e-06, 3.4459e-08, 3.7369e-08, ..., -1.5832e-06, + 4.2608e-08, -1.9968e-05], + ..., + [ 1.2152e-05, 9.1456e-07, 1.0207e-06, ..., 1.6674e-05, + 1.9209e-08, 1.1787e-05], + [ 7.1805e-07, -3.8301e-08, -6.4727e-08, ..., 3.4459e-08, + 1.4447e-07, 4.3362e-06], + [-1.0669e-05, -8.7777e-07, -1.1642e-06, ..., -1.5706e-05, + -8.3935e-08, -1.6717e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0343, -0.0261, -0.0060, 0.0150, -0.0041, -0.0144, 0.0014, 0.0134, + -0.0263, 0.0132], device='cuda:0'), grad: tensor([-2.4345e-06, 2.0186e-07, -4.7773e-05, 7.2196e-06, 1.1418e-06, + -1.1246e-07, 2.3711e-06, 5.3048e-05, 1.0744e-05, -2.4423e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 216.87, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.5204 re_mapping 0.0038 re_causal 0.0127 /// teacc 98.96 lr 0.00010000 +Epoch 197, weight, value: tensor([[ 0.0183, -0.0712, 0.0134, ..., -0.0881, -0.1211, -0.0931], + [-0.1002, 0.0599, -0.0977, ..., -0.1130, -0.1018, -0.1668], + [-0.0719, -0.1131, -0.1002, ..., -0.1558, -0.0926, 0.1313], + ..., + [ 0.0736, -0.0448, -0.0999, ..., 0.0829, -0.0942, -0.0981], + [ 0.1111, 0.0926, 0.0619, ..., -0.1338, -0.0940, 0.0714], + [ 0.0570, 0.0235, 0.0797, ..., 0.0488, -0.0725, 0.0081]], + device='cuda:0'), grad: tensor([[-2.7940e-09, 4.4121e-08, 3.5274e-08, ..., 1.2538e-07, + 6.6357e-09, 4.3772e-08], + [ 2.5728e-08, 9.6275e-08, 5.6927e-08, ..., 2.4564e-07, + 7.9395e-08, 1.2130e-07], + [ 1.1059e-08, 4.3306e-08, 1.6182e-08, ..., 1.6531e-08, + 7.2177e-09, -3.6205e-08], + ..., + [ 7.0315e-07, 1.3933e-06, 1.1604e-06, ..., 5.1074e-06, + 5.7044e-09, 1.4808e-06], + [-2.5611e-09, 5.5647e-08, 2.5495e-08, ..., 1.1001e-07, + 1.2806e-09, -2.6776e-09], + [-9.8441e-07, -1.8040e-06, -1.5898e-06, ..., -6.9477e-06, + 1.1467e-07, -1.9222e-06]], device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0345, -0.0264, -0.0057, 0.0149, -0.0039, -0.0144, 0.0015, 0.0132, + -0.0268, 0.0134], device='cuda:0'), grad: tensor([ 1.3318e-07, 7.2364e-07, 4.3889e-08, -7.1106e-07, 3.0044e-06, + 6.8732e-07, 2.7986e-07, 1.3471e-05, 3.6089e-07, -1.7956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 216.83, cls_loss 0.0009 cls_loss_mapping 0.0035 cls_loss_causal 0.5165 re_mapping 0.0040 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +Epoch 198, weight, value: tensor([[ 0.0183, -0.0713, 0.0134, ..., -0.0884, -0.1213, -0.0933], + [-0.1003, 0.0598, -0.0980, ..., -0.1133, -0.1028, -0.1670], + [-0.0721, -0.1133, -0.1005, ..., -0.1564, -0.0932, 0.1314], + ..., + [ 0.0736, -0.0450, -0.1011, ..., 0.0828, -0.0946, -0.0986], + [ 0.1115, 0.0929, 0.0623, ..., -0.1348, -0.0944, 0.0715], + [ 0.0573, 0.0234, 0.0800, ..., 0.0487, -0.0728, 0.0080]], + device='cuda:0'), grad: tensor([[ 2.0862e-07, 4.0885e-07, 5.2620e-08, ..., 4.9081e-07, + 4.3074e-09, 3.9022e-07], + [ 9.6275e-08, -7.6275e-07, 3.5157e-08, ..., 1.5623e-07, + 1.0477e-09, -1.1479e-07], + [ 4.7311e-07, 6.1467e-07, 1.5064e-07, ..., 5.2154e-08, + 3.6089e-09, 3.3621e-07], + ..., + [ 1.8543e-06, 1.3318e-07, 1.3625e-06, ..., 2.8908e-06, + 2.3283e-10, 3.7230e-07], + [-7.4320e-07, -8.2841e-07, -2.4540e-07, ..., 1.9034e-07, + 2.9104e-09, -1.0207e-06], + [-2.9262e-06, -2.6077e-07, -1.5441e-06, ..., -6.0685e-06, + 1.6298e-09, -1.5451e-06]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0344, -0.0265, -0.0059, 0.0157, -0.0037, -0.0145, 0.0016, 0.0131, + -0.0267, 0.0133], device='cuda:0'), grad: tensor([ 2.5965e-06, -2.6803e-06, 1.6093e-06, 1.1967e-06, 4.7423e-06, + 5.0059e-08, 1.1753e-06, 7.3761e-06, -2.5239e-06, -1.3538e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 217.03, cls_loss 0.0012 cls_loss_mapping 0.0040 cls_loss_causal 0.5341 re_mapping 0.0038 re_causal 0.0124 /// teacc 98.92 lr 0.00010000 +Epoch 199, weight, value: tensor([[ 0.0179, -0.0714, 0.0132, ..., -0.0897, -0.1215, -0.0936], + [-0.1005, 0.0598, -0.0984, ..., -0.1140, -0.1034, -0.1678], + [-0.0727, -0.1136, -0.1009, ..., -0.1576, -0.0934, 0.1320], + ..., + [ 0.0748, -0.0449, -0.1010, ..., 0.0851, -0.0951, -0.0986], + [ 0.1118, 0.0935, 0.0626, ..., -0.1359, -0.0945, 0.0720], + [ 0.0572, 0.0233, 0.0802, ..., 0.0485, -0.0731, 0.0078]], + device='cuda:0'), grad: tensor([[ 1.5716e-08, 3.0734e-08, 7.3342e-09, ..., 2.8522e-08, + 1.4866e-07, 2.9197e-07], + [ 2.2235e-08, -5.5879e-08, 1.3504e-08, ..., 5.5181e-08, + 5.5763e-08, 2.0792e-07], + [-1.8312e-07, 3.6205e-08, 5.3551e-09, ..., 4.4238e-09, + 8.1258e-08, -1.6680e-06], + ..., + [ 5.6461e-08, 9.6275e-08, 3.5740e-08, ..., 1.9069e-07, + 3.9814e-08, 1.3679e-07], + [-6.6822e-08, -5.2969e-08, 9.6625e-09, ..., 1.6182e-08, + 2.4145e-07, 5.8068e-07], + [-9.4064e-08, -1.0477e-08, -7.5437e-08, ..., -5.8440e-08, + 2.6403e-07, 2.3586e-07]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0341, -0.0268, -0.0058, 0.0155, -0.0036, -0.0144, 0.0011, 0.0140, + -0.0265, 0.0130], device='cuda:0'), grad: tensor([ 1.0068e-06, 4.5262e-07, -5.2936e-06, 9.4995e-07, 5.7044e-08, + 5.6718e-07, -9.6112e-07, 8.9081e-07, 1.9046e-06, 4.1816e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 216.89, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5250 re_mapping 0.0036 re_causal 0.0117 /// teacc 99.06 lr 0.00010000 +Epoch 200, weight, value: tensor([[ 0.0180, -0.0715, 0.0131, ..., -0.0900, -0.1220, -0.0941], + [-0.1008, 0.0599, -0.0987, ..., -0.1145, -0.1058, -0.1688], + [-0.0733, -0.1137, -0.1012, ..., -0.1587, -0.0952, 0.1324], + ..., + [ 0.0752, -0.0451, -0.1014, ..., 0.0856, -0.0955, -0.0992], + [ 0.1122, 0.0935, 0.0628, ..., -0.1373, -0.0953, 0.0720], + [ 0.0573, 0.0234, 0.0803, ..., 0.0484, -0.0734, 0.0077]], + device='cuda:0'), grad: tensor([[-5.9558e-07, 1.0990e-06, 5.7183e-07, ..., -6.7172e-08, + 7.2177e-09, 6.5006e-07], + [ 5.7393e-08, 1.0812e-04, 6.0111e-05, ..., 1.8161e-08, + 6.4028e-09, 6.7115e-05], + [ 1.9872e-07, 1.4253e-05, 3.8520e-06, ..., 2.9453e-08, + 8.0327e-09, 4.2580e-06], + ..., + [ 9.6625e-08, 6.0070e-07, 2.0501e-07, ..., 1.8510e-08, + 8.4983e-09, 2.3725e-07], + [-2.2782e-07, -1.2791e-04, -7.1347e-05, ..., 1.8510e-08, + -9.6625e-09, -7.9811e-05], + [ 1.9441e-07, 3.3760e-07, 1.0477e-07, ..., -1.7579e-08, + 2.0722e-08, 1.7544e-07]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0340, -0.0271, -0.0058, 0.0153, -0.0033, -0.0145, 0.0018, 0.0142, + -0.0268, 0.0128], device='cuda:0'), grad: tensor([ 9.8534e-07, 4.4227e-04, 6.9857e-05, -4.2021e-05, 6.3283e-07, + 2.0459e-05, 2.2560e-05, 3.3062e-06, -5.2118e-04, 2.8498e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 216.93, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.4578 re_mapping 0.0037 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 201, weight, value: tensor([[ 0.0185, -0.0718, 0.0132, ..., -0.0900, -0.1212, -0.0930], + [-0.1010, 0.0599, -0.1000, ..., -0.1150, -0.1070, -0.1696], + [-0.0736, -0.1141, -0.1017, ..., -0.1593, -0.0959, 0.1323], + ..., + [ 0.0752, -0.0453, -0.1019, ..., 0.0859, -0.0957, -0.0995], + [ 0.1128, 0.0940, 0.0635, ..., -0.1381, -0.0954, 0.0724], + [ 0.0571, 0.0231, 0.0803, ..., 0.0483, -0.0737, 0.0075]], + device='cuda:0'), grad: tensor([[-4.2631e-07, -9.3482e-08, 5.0059e-09, ..., -1.4284e-07, + 1.4203e-08, 2.7590e-08], + [ 2.0606e-08, 1.7777e-07, 1.1991e-08, ..., 1.2107e-08, + 5.4715e-09, 2.2561e-07], + [ 1.0128e-07, 7.4320e-07, 3.1898e-08, ..., 9.3132e-09, + 1.1525e-08, -3.7998e-07], + ..., + [ 3.3062e-08, 3.0873e-07, 1.6764e-08, ..., 2.7823e-08, + 0.0000e+00, 3.6904e-08], + [ 4.2375e-08, 1.2584e-07, -1.2689e-08, ..., 4.1095e-08, + 2.8638e-08, 5.7975e-08], + [ 1.2608e-07, 1.3504e-07, -7.3342e-09, ..., -3.4925e-10, + 5.8208e-10, 2.3865e-08]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0346, -0.0272, -0.0061, 0.0172, -0.0030, -0.0148, 0.0012, 0.0141, + -0.0264, 0.0124], device='cuda:0'), grad: tensor([-2.8480e-06, 1.3132e-06, 1.6289e-06, -4.9099e-06, 2.0897e-07, + 8.2422e-07, 2.6752e-07, 1.1688e-06, 1.0245e-06, 1.3188e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 217.16, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5259 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.01 lr 0.00010000 +Epoch 202, weight, value: tensor([[ 0.0163, -0.0722, 0.0134, ..., -0.0930, -0.1214, -0.0921], + [-0.1012, 0.0606, -0.1002, ..., -0.1153, -0.1077, -0.1697], + [-0.0734, -0.1144, -0.1020, ..., -0.1599, -0.0965, 0.1324], + ..., + [ 0.0753, -0.0455, -0.1025, ..., 0.0860, -0.0959, -0.1000], + [ 0.1117, 0.0937, 0.0636, ..., -0.1399, -0.0958, 0.0717], + [ 0.0598, 0.0230, 0.0806, ..., 0.0490, -0.0739, 0.0074]], + device='cuda:0'), grad: tensor([[ 4.3423e-08, 3.5507e-08, 1.2456e-08, ..., 1.6077e-07, + 6.0117e-07, 3.7439e-07], + [ 1.3842e-07, -5.2713e-07, 7.7998e-09, ..., 6.4820e-07, + 5.9092e-07, 2.8894e-07], + [ 1.3027e-07, 4.9127e-08, 5.2387e-09, ..., 1.9791e-07, + 2.3760e-07, 7.2992e-08], + ..., + [-5.2433e-07, 6.3330e-07, 6.4843e-08, ..., 7.3854e-07, + 7.3016e-07, 3.3434e-07], + [ 7.1013e-08, -2.6776e-08, -2.2585e-08, ..., 1.4610e-07, + 2.4214e-07, 1.2910e-07], + [-1.5355e-07, 2.8615e-07, -1.0105e-07, ..., 4.7833e-06, + 3.9451e-06, 1.6596e-06]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0330, -0.0271, -0.0060, 0.0166, -0.0029, -0.0144, 0.0012, 0.0140, + -0.0276, 0.0135], device='cuda:0'), grad: tensor([ 1.8738e-06, -1.1362e-06, 1.0589e-06, 1.1772e-06, -1.8373e-05, + -1.4372e-05, 5.7854e-06, 3.0976e-06, 1.0684e-05, 1.0200e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 217.34, cls_loss 0.0008 cls_loss_mapping 0.0026 cls_loss_causal 0.4935 re_mapping 0.0038 re_causal 0.0123 /// teacc 98.98 lr 0.00010000 +Epoch 203, weight, value: tensor([[ 0.0165, -0.0722, 0.0135, ..., -0.0931, -0.1219, -0.0923], + [-0.1010, 0.0611, -0.1003, ..., -0.1155, -0.1077, -0.1696], + [-0.0737, -0.1148, -0.1022, ..., -0.1603, -0.0974, 0.1326], + ..., + [ 0.0752, -0.0459, -0.1029, ..., 0.0860, -0.0967, -0.1007], + [ 0.1118, 0.0936, 0.0637, ..., -0.1405, -0.0960, 0.0716], + [ 0.0599, 0.0230, 0.0807, ..., 0.0491, -0.0740, 0.0073]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, 2.4331e-07, 4.8894e-08, ..., 7.6601e-08, + 1.3271e-08, 2.9453e-08], + [ 3.2363e-08, -1.4284e-07, 6.5193e-09, ..., 7.1479e-08, + 3.2596e-09, 2.4913e-08], + [ 2.2119e-08, 7.3947e-07, 1.3935e-07, ..., 2.9104e-08, + 2.6776e-09, -4.3539e-08], + ..., + [-7.0594e-07, 3.5437e-07, 8.1956e-08, ..., -8.7661e-08, + 1.6298e-09, 1.5250e-08], + [-1.8915e-06, -1.3523e-06, -1.4408e-06, ..., -5.8580e-07, + -1.9278e-07, -2.6077e-06], + [ 2.2873e-06, 1.2852e-06, 1.2452e-06, ..., 9.5461e-07, + 1.9418e-07, 2.3376e-06]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0332, -0.0268, -0.0061, 0.0161, -0.0029, -0.0141, 0.0013, 0.0137, + -0.0278, 0.0134], device='cuda:0'), grad: tensor([ 9.8255e-07, 6.8336e-08, 2.6952e-06, -6.5863e-06, 5.6624e-07, + -1.3206e-06, 5.1875e-07, 2.0973e-06, -5.5619e-06, 6.5081e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 217.04, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.4810 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.09 lr 0.00010000 +Epoch 204, weight, value: tensor([[ 0.0166, -0.0722, 0.0135, ..., -0.0932, -0.1223, -0.0925], + [-0.1011, 0.0614, -0.1005, ..., -0.1158, -0.1083, -0.1698], + [-0.0739, -0.1150, -0.1025, ..., -0.1607, -0.0978, 0.1329], + ..., + [ 0.0748, -0.0464, -0.1055, ..., 0.0858, -0.0973, -0.1017], + [ 0.1124, 0.0939, 0.0638, ..., -0.1413, -0.0964, 0.0716], + [ 0.0602, 0.0233, 0.0818, ..., 0.0491, -0.0743, 0.0072]], + device='cuda:0'), grad: tensor([[-3.3411e-08, 3.3062e-08, 3.1432e-09, ..., 3.0524e-07, + 1.4203e-08, 2.6310e-08], + [ 2.0035e-07, 3.4692e-08, 5.1223e-09, ..., 4.6659e-07, + 1.8859e-08, 2.8126e-07], + [ 2.9081e-07, 1.4610e-07, 7.9162e-09, ..., 5.3085e-07, + 1.7113e-08, -6.3423e-07], + ..., + [-1.0496e-06, 1.5693e-07, 7.1013e-09, ..., -2.3674e-06, + 3.0384e-08, 1.5704e-07], + [ 2.7940e-09, 2.4447e-07, -9.3132e-10, ..., 3.5157e-08, + 5.0059e-09, 2.7707e-08], + [ 4.3726e-07, 5.7160e-08, -1.7928e-08, ..., 9.6671e-07, + 1.5728e-07, 7.8231e-08]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0332, -0.0268, -0.0061, 0.0159, -0.0028, -0.0143, 0.0017, 0.0134, + -0.0278, 0.0135], device='cuda:0'), grad: tensor([ 6.2026e-07, 2.4475e-06, -8.0140e-07, 1.3746e-06, 6.2864e-09, + -5.4576e-06, 1.2983e-06, -4.6417e-06, 2.1085e-06, 3.0082e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 217.12, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.4835 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.11 lr 0.00010000 +Epoch 205, weight, value: tensor([[ 0.0168, -0.0722, 0.0137, ..., -0.0933, -0.1224, -0.0926], + [-0.1019, 0.0616, -0.1006, ..., -0.1170, -0.1089, -0.1701], + [-0.0748, -0.1153, -0.1034, ..., -0.1620, -0.0983, 0.1328], + ..., + [ 0.0755, -0.0466, -0.1064, ..., 0.0863, -0.0976, -0.1018], + [ 0.1128, 0.0941, 0.0642, ..., -0.1415, -0.0968, 0.0718], + [ 0.0603, 0.0234, 0.0819, ..., 0.0487, -0.0752, 0.0065]], + device='cuda:0'), grad: tensor([[ 6.6007e-08, 1.0652e-07, 6.9384e-08, ..., 1.7928e-08, + 4.6100e-08, 1.7183e-07], + [ 2.0058e-07, 2.6776e-08, 5.3667e-08, ..., 9.0688e-08, + 1.8626e-08, 1.3900e-07], + [ 1.5162e-06, 1.3066e-06, 7.2271e-07, ..., 5.7393e-08, + 1.7812e-08, 1.6326e-06], + ..., + [-1.0361e-07, 1.0803e-07, 7.3807e-08, ..., -7.1013e-08, + 2.4214e-08, 1.4994e-07], + [-2.6077e-06, -2.3451e-06, -1.2852e-06, ..., 7.7998e-09, + 4.0047e-08, -2.9895e-06], + [ 2.4098e-07, 2.4121e-07, 3.8883e-08, ..., -5.9488e-08, + 1.4924e-07, 3.5949e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0335, -0.0269, -0.0064, 0.0160, -0.0024, -0.0144, 0.0020, 0.0136, + -0.0279, 0.0131], device='cuda:0'), grad: tensor([ 2.6543e-07, 6.5798e-07, 5.8413e-06, 1.2619e-06, 3.2131e-08, + 5.8999e-07, 3.5297e-07, -2.3702e-07, -1.0125e-05, 1.3327e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 217.26, cls_loss 0.0012 cls_loss_mapping 0.0033 cls_loss_causal 0.4851 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.10 lr 0.00010000 +Epoch 206, weight, value: tensor([[ 0.0168, -0.0724, 0.0134, ..., -0.0934, -0.1232, -0.0933], + [-0.1020, 0.0620, -0.1009, ..., -0.1173, -0.1095, -0.1704], + [-0.0752, -0.1156, -0.1040, ..., -0.1614, -0.0985, 0.1338], + ..., + [ 0.0756, -0.0469, -0.1075, ..., 0.0863, -0.0978, -0.1037], + [ 0.1143, 0.0947, 0.0649, ..., -0.1422, -0.0966, 0.0726], + [ 0.0604, 0.0236, 0.0827, ..., 0.0485, -0.0759, 0.0061]], + device='cuda:0'), grad: tensor([[ 1.9930e-07, 1.1567e-06, 7.3854e-07, ..., 2.5146e-07, + 3.0152e-08, 1.0140e-07], + [ 3.8906e-07, 2.6170e-07, 1.1409e-07, ..., 4.6147e-07, + 1.7590e-07, 2.9034e-07], + [ 1.5339e-06, 3.5726e-06, 2.2501e-06, ..., 2.7288e-06, + 3.1898e-08, 1.0831e-06], + ..., + [ 5.0217e-06, 3.4533e-06, 1.4482e-06, ..., 3.3118e-06, + 5.6578e-08, 1.5814e-06], + [ 4.1677e-08, 4.4587e-08, -3.3760e-09, ..., 2.7334e-07, + 1.5146e-07, -1.4319e-07], + [-8.7619e-06, -4.2506e-06, -2.2240e-06, ..., -8.3372e-06, + 4.8708e-07, -2.5555e-06]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0333, -0.0266, -0.0038, 0.0133, -0.0023, -0.0145, 0.0019, 0.0134, + -0.0273, 0.0128], device='cuda:0'), grad: tensor([ 2.8405e-06, 1.7015e-06, 1.6212e-05, -1.0356e-05, 2.3725e-07, + 3.3714e-06, 4.3726e-07, 1.5348e-05, 4.0047e-07, -3.0130e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 217.25, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.5131 re_mapping 0.0037 re_causal 0.0122 /// teacc 99.05 lr 0.00010000 +Epoch 207, weight, value: tensor([[ 0.0168, -0.0726, 0.0131, ..., -0.0935, -0.1237, -0.0937], + [-0.1022, 0.0621, -0.1020, ..., -0.1176, -0.1107, -0.1711], + [-0.0757, -0.1159, -0.1048, ..., -0.1622, -0.0990, 0.1339], + ..., + [ 0.0758, -0.0471, -0.1078, ..., 0.0870, -0.0985, -0.1044], + [ 0.1149, 0.0951, 0.0656, ..., -0.1430, -0.0969, 0.0729], + [ 0.0605, 0.0236, 0.0831, ..., 0.0484, -0.0766, 0.0058]], + device='cuda:0'), grad: tensor([[-1.5646e-06, 1.3469e-07, 8.6729e-08, ..., -5.6927e-08, + -1.0738e-06, -4.4680e-07], + [ 4.3423e-08, 1.2806e-09, 5.9721e-08, ..., 2.3399e-08, + 3.3877e-08, 3.8766e-08], + [ 9.4296e-08, 7.6275e-07, 1.9802e-07, ..., 1.8044e-08, + 7.8464e-08, 5.9721e-08], + ..., + [ 2.2061e-07, 4.5123e-07, 2.3283e-07, ..., 3.0012e-07, + 3.0850e-08, 1.9092e-08], + [-2.5844e-07, 1.7625e-07, 1.0477e-07, ..., 1.7008e-07, + -4.8289e-07, -6.1980e-07], + [-4.7591e-07, -4.7730e-07, -7.7114e-07, ..., -7.0175e-07, + 2.5076e-07, -3.0734e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0331, -0.0268, -0.0030, 0.0120, -0.0021, -0.0141, 0.0024, 0.0136, + -0.0270, 0.0126], device='cuda:0'), grad: tensor([-5.2489e-06, 2.2212e-07, 2.3618e-06, -6.2250e-06, 7.9861e-07, + 6.2445e-07, 6.5416e-06, 1.7369e-06, 6.7893e-07, -1.5097e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 217.28, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.4877 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 208, weight, value: tensor([[ 0.0169, -0.0728, 0.0131, ..., -0.0935, -0.1241, -0.0940], + [-0.1027, 0.0621, -0.1022, ..., -0.1182, -0.1127, -0.1721], + [-0.0765, -0.1163, -0.1057, ..., -0.1635, -0.0994, 0.1347], + ..., + [ 0.0747, -0.0470, -0.1087, ..., 0.0876, -0.0991, -0.1048], + [ 0.1173, 0.0950, 0.0658, ..., -0.1442, -0.0971, 0.0729], + [ 0.0606, 0.0236, 0.0833, ..., 0.0484, -0.0769, 0.0058]], + device='cuda:0'), grad: tensor([[ 9.7230e-07, 2.2119e-08, 5.5996e-08, ..., 2.3842e-07, + 2.3283e-10, 1.2340e-08], + [ 3.4105e-06, 1.0442e-07, 5.9884e-07, ..., 2.5462e-06, + 1.3888e-07, 1.3830e-07], + [ 5.5768e-06, 5.4482e-07, 8.6031e-08, ..., 1.6287e-07, + 4.1677e-08, 3.6671e-07], + ..., + [-2.9111e-04, 4.5961e-07, -5.5060e-06, ..., -2.4348e-05, + 2.6426e-08, 1.8976e-08], + [ 2.5845e-04, -9.4809e-07, -1.3970e-07, ..., 1.9372e-07, + -3.1199e-08, -7.4366e-07], + [ 1.9580e-05, 3.0128e-07, 4.8503e-06, ..., 2.0891e-05, + 4.1723e-07, 1.6927e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0330, -0.0273, -0.0029, 0.0122, -0.0020, -0.0139, 0.0018, 0.0136, + -0.0264, 0.0125], device='cuda:0'), grad: tensor([ 1.8049e-06, 9.3281e-06, 1.1303e-05, -9.9465e-07, -2.8056e-07, + 5.1446e-06, 8.5495e-07, -5.2404e-04, 4.3654e-04, 6.0052e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 217.35, cls_loss 0.0010 cls_loss_mapping 0.0030 cls_loss_causal 0.5091 re_mapping 0.0036 re_causal 0.0119 /// teacc 99.12 lr 0.00010000 +Epoch 209, weight, value: tensor([[ 0.0168, -0.0731, 0.0128, ..., -0.0936, -0.1249, -0.0945], + [-0.1031, 0.0622, -0.1028, ..., -0.1188, -0.1132, -0.1725], + [-0.0780, -0.1166, -0.1064, ..., -0.1648, -0.0997, 0.1352], + ..., + [ 0.0774, -0.0471, -0.1090, ..., 0.0897, -0.0994, -0.1044], + [ 0.1156, 0.0951, 0.0661, ..., -0.1458, -0.0976, 0.0730], + [ 0.0600, 0.0233, 0.0831, ..., 0.0484, -0.0768, 0.0059]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 6.9151e-08, 1.1059e-08, ..., 3.2247e-08, + 5.9139e-08, 1.1362e-07], + [ 1.0128e-07, 7.2177e-09, 1.2107e-08, ..., 1.4552e-07, + 2.7823e-08, 5.2853e-08], + [-4.2608e-08, 4.0978e-08, 1.0594e-08, ..., 2.7288e-07, + 3.2480e-08, -3.5577e-07], + ..., + [-6.0350e-07, 4.4471e-08, 1.6880e-08, ..., -1.4026e-06, + -1.0070e-07, 6.2864e-08], + [-8.5635e-07, -1.1642e-06, -1.6775e-07, ..., -6.2864e-09, + -5.6904e-07, -1.2182e-06], + [ 5.1409e-07, 4.8755e-07, 7.3807e-08, ..., 3.1386e-07, + 3.0594e-07, 5.7416e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0327, -0.0274, -0.0030, 0.0124, -0.0023, -0.0139, 0.0024, 0.0149, + -0.0273, 0.0121], device='cuda:0'), grad: tensor([ 5.8790e-08, 4.5029e-07, -1.3551e-06, 5.5274e-07, 1.2368e-06, + 5.5786e-07, 7.0874e-07, -2.4531e-06, -1.9968e-06, 2.2035e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 217.41, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4819 re_mapping 0.0037 re_causal 0.0119 /// teacc 99.09 lr 0.00010000 +Epoch 210, weight, value: tensor([[ 0.0167, -0.0733, 0.0123, ..., -0.0938, -0.1252, -0.0948], + [-0.1034, 0.0626, -0.1029, ..., -0.1193, -0.1137, -0.1724], + [-0.0781, -0.1170, -0.1066, ..., -0.1654, -0.1001, 0.1355], + ..., + [ 0.0777, -0.0480, -0.1095, ..., 0.0896, -0.0998, -0.1050], + [ 0.1157, 0.0947, 0.0662, ..., -0.1466, -0.0983, 0.0727], + [ 0.0603, 0.0240, 0.0846, ..., 0.0486, -0.0768, 0.0059]], + device='cuda:0'), grad: tensor([[ 6.7288e-08, 1.4552e-08, 1.2852e-07, ..., 3.7719e-08, + 2.2687e-06, 1.2275e-06], + [ 2.2305e-07, 9.0804e-09, 1.0827e-08, ..., 9.7440e-08, + 3.0734e-08, 2.9453e-08], + [ 2.3454e-05, 4.7032e-08, 2.5262e-08, ..., 1.0341e-05, + 6.1118e-08, 3.7951e-08], + ..., + [-2.3961e-05, 3.2946e-08, 1.3621e-08, ..., -1.0580e-05, + 7.9162e-09, 1.6415e-08], + [-2.1001e-07, -1.8557e-07, -1.5565e-07, ..., 3.6671e-08, + 1.4331e-07, -1.1385e-07], + [ 9.8022e-08, 3.7020e-08, 2.2817e-08, ..., 2.6776e-09, + 1.3621e-08, 6.7987e-08]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0325, -0.0273, -0.0031, 0.0128, -0.0025, -0.0138, 0.0027, 0.0144, + -0.0278, 0.0123], device='cuda:0'), grad: tensor([ 7.4878e-06, 9.0431e-07, 7.9393e-05, -1.2957e-07, 7.6881e-07, + 1.5190e-06, -9.3207e-06, -8.1062e-05, -5.6927e-08, 3.5926e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 217.22, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4818 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.07 lr 0.00010000 +Epoch 211, weight, value: tensor([[ 0.0169, -0.0731, 0.0125, ..., -0.0939, -0.1252, -0.0946], + [-0.1036, 0.0629, -0.1030, ..., -0.1197, -0.1153, -0.1726], + [-0.0792, -0.1178, -0.1070, ..., -0.1667, -0.1010, 0.1354], + ..., + [ 0.0781, -0.0482, -0.1099, ..., 0.0900, -0.1007, -0.1053], + [ 0.1157, 0.0946, 0.0663, ..., -0.1473, -0.0994, 0.0725], + [ 0.0602, 0.0238, 0.0847, ..., 0.0485, -0.0774, 0.0056]], + device='cuda:0'), grad: tensor([[ 2.1933e-07, 1.0058e-07, 1.1001e-07, ..., 1.3143e-07, + 1.6182e-08, 8.7661e-08], + [ 8.1607e-08, -6.0908e-07, 2.5611e-08, ..., 4.4820e-08, + -4.0955e-07, 3.0035e-08], + [ 1.6729e-07, 8.9407e-08, 7.1595e-08, ..., 5.4599e-08, + 1.2224e-08, 8.3935e-08], + ..., + [-3.8417e-07, 4.2655e-07, 7.6601e-08, ..., -2.5518e-07, + 2.5681e-07, 4.4936e-08], + [-6.0303e-07, -3.1758e-07, -4.6100e-07, ..., 4.5518e-08, + 4.5286e-08, -6.7288e-07], + [-1.4296e-07, -9.1619e-08, -1.4913e-07, ..., -2.3888e-07, + 1.1176e-08, 5.8208e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0329, -0.0272, -0.0034, 0.0131, -0.0021, -0.0144, 0.0030, 0.0147, + -0.0283, 0.0119], device='cuda:0'), grad: tensor([ 9.9279e-07, -1.8254e-05, 1.9614e-06, 8.6892e-07, 9.0972e-06, + 1.9725e-06, 3.7923e-06, 1.2927e-06, -1.6093e-06, -1.0361e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 217.24, cls_loss 0.0010 cls_loss_mapping 0.0028 cls_loss_causal 0.5111 re_mapping 0.0034 re_causal 0.0115 /// teacc 99.06 lr 0.00010000 +Epoch 212, weight, value: tensor([[ 0.0169, -0.0736, 0.0119, ..., -0.0940, -0.1257, -0.0949], + [-0.1038, 0.0630, -0.1032, ..., -0.1202, -0.1169, -0.1728], + [-0.0771, -0.1181, -0.1074, ..., -0.1648, -0.1021, 0.1368], + ..., + [ 0.0781, -0.0486, -0.1099, ..., 0.0906, -0.1011, -0.1081], + [ 0.1159, 0.0949, 0.0667, ..., -0.1480, -0.1005, 0.0722], + [ 0.0597, 0.0237, 0.0847, ..., 0.0481, -0.0777, 0.0055]], + device='cuda:0'), grad: tensor([[-4.1793e-08, 6.6240e-08, 6.0885e-08, ..., 4.7847e-08, + 2.1071e-08, 9.1270e-08], + [ 8.7894e-08, -1.6252e-07, 2.4098e-08, ..., 1.6950e-07, + 1.5134e-08, 1.2922e-08], + [ 6.2515e-08, 1.6449e-07, 2.7707e-08, ..., 1.0268e-07, + 3.3528e-08, -8.4983e-08], + ..., + [-1.9232e-07, 3.2107e-07, 7.7300e-08, ..., -2.5844e-07, + 3.1432e-09, 1.5134e-08], + [ 9.3132e-10, 3.0943e-07, 1.6764e-08, ..., 7.3924e-08, + 3.6554e-08, -6.7521e-09], + [-9.6974e-08, -1.5448e-07, -1.4435e-07, ..., -3.0501e-07, + -1.3737e-08, -3.0850e-08]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0328, -0.0273, -0.0029, 0.0133, -0.0018, -0.0147, 0.0036, 0.0143, + -0.0283, 0.0115], device='cuda:0'), grad: tensor([ 1.8040e-06, -1.6019e-07, -8.2538e-08, -1.8515e-06, 5.2154e-07, + -1.5199e-06, 3.2224e-07, 1.7870e-07, 1.0487e-06, -2.6822e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 217.21, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4766 re_mapping 0.0036 re_causal 0.0116 /// teacc 99.07 lr 0.00010000 +Epoch 213, weight, value: tensor([[ 0.0169, -0.0738, 0.0117, ..., -0.0940, -0.1260, -0.0952], + [-0.1040, 0.0638, -0.1034, ..., -0.1206, -0.1173, -0.1724], + [-0.0770, -0.1194, -0.1081, ..., -0.1651, -0.1028, 0.1365], + ..., + [ 0.0779, -0.0489, -0.1112, ..., 0.0906, -0.1016, -0.1086], + [ 0.1163, 0.0952, 0.0674, ..., -0.1486, -0.1004, 0.0728], + [ 0.0599, 0.0239, 0.0852, ..., 0.0483, -0.0778, 0.0054]], + device='cuda:0'), grad: tensor([[ 3.4575e-08, 6.7696e-08, 6.0594e-08, ..., 1.9511e-07, + 1.0635e-07, 5.9546e-08], + [ 2.0431e-08, -5.9488e-08, 4.7381e-08, ..., 4.1840e-07, + 2.8312e-07, 1.3947e-07], + [ 3.7835e-08, 7.2585e-08, 3.1956e-08, ..., 2.9127e-07, + 2.0489e-07, 1.2643e-07], + ..., + [ 4.6508e-08, 2.3725e-07, 1.3411e-07, ..., 9.4762e-07, + 6.3982e-07, 3.1758e-07], + [-2.0023e-08, 1.8394e-08, 5.2562e-08, ..., 2.3236e-07, + 1.2701e-07, -1.7462e-09], + [-3.3900e-07, 1.3644e-07, 1.3423e-07, ..., 7.5884e-06, + 5.9381e-06, 2.7716e-06]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0327, -0.0268, -0.0033, 0.0131, -0.0019, -0.0146, 0.0040, 0.0142, + -0.0281, 0.0115], device='cuda:0'), grad: tensor([ 4.5029e-07, 5.1782e-07, 7.4646e-07, 3.9767e-07, -2.0742e-05, + -1.7346e-08, 6.7055e-07, 2.3339e-06, 3.9116e-07, 1.5274e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 217.42, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.5000 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 +Epoch 214, weight, value: tensor([[ 0.0167, -0.0739, 0.0109, ..., -0.0943, -0.1262, -0.0956], + [-0.1043, 0.0646, -0.1036, ..., -0.1209, -0.1178, -0.1727], + [-0.0771, -0.1200, -0.1089, ..., -0.1655, -0.1031, 0.1369], + ..., + [ 0.0779, -0.0499, -0.1123, ..., 0.0906, -0.1022, -0.1093], + [ 0.1165, 0.0952, 0.0676, ..., -0.1498, -0.1004, 0.0730], + [ 0.0606, 0.0245, 0.0865, ..., 0.0489, -0.0775, 0.0060]], + device='cuda:0'), grad: tensor([[ 7.3342e-09, 2.8056e-08, 8.6147e-09, ..., 1.4296e-07, + 6.4727e-07, 5.0757e-07], + [ 4.3423e-07, 2.1770e-08, 1.4610e-08, ..., 5.9791e-07, + 2.8983e-06, 1.3318e-06], + [ 1.0487e-06, 5.7183e-07, 1.7474e-07, ..., 1.1269e-06, + 1.1317e-05, 5.4687e-06], + ..., + [-1.3299e-06, 3.8708e-08, 8.5565e-09, ..., -1.2545e-06, + 6.3237e-07, 3.8301e-07], + [-7.7439e-07, -7.2410e-07, -2.2980e-07, ..., 1.3551e-07, + 9.5135e-07, -2.9081e-07], + [ 3.4645e-07, 3.2771e-08, 4.8312e-09, ..., 5.2564e-06, + 4.8466e-06, 2.7362e-06]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0324, -0.0264, -0.0033, 0.0131, -0.0025, -0.0146, 0.0039, 0.0137, + -0.0282, 0.0123], device='cuda:0'), grad: tensor([ 1.8403e-06, 9.9167e-06, 3.6657e-05, 7.2829e-07, -1.1355e-05, + 2.9489e-05, -7.7724e-05, -2.5947e-06, -1.6321e-07, 1.3277e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 217.11, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4882 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.08 lr 0.00010000 +Epoch 215, weight, value: tensor([[ 0.0166, -0.0742, 0.0108, ..., -0.0946, -0.1266, -0.0959], + [-0.1049, 0.0654, -0.1038, ..., -0.1219, -0.1197, -0.1729], + [-0.0779, -0.1207, -0.1095, ..., -0.1670, -0.1052, 0.1367], + ..., + [ 0.0786, -0.0509, -0.1129, ..., 0.0917, -0.1006, -0.1093], + [ 0.1169, 0.0956, 0.0679, ..., -0.1507, -0.1008, 0.0734], + [ 0.0606, 0.0245, 0.0861, ..., 0.0474, -0.0800, 0.0044]], + device='cuda:0'), grad: tensor([[ 1.6822e-08, 3.5623e-08, 2.3923e-08, ..., 3.4634e-08, + 3.0617e-08, 4.7847e-08], + [ 3.3597e-07, 1.0908e-07, 1.2969e-07, ..., 2.5891e-07, + 3.7835e-09, 1.6706e-07], + [ 2.0559e-07, 9.4238e-08, 4.0804e-08, ..., 1.9395e-07, + 3.0268e-09, 6.1060e-08], + ..., + [-1.2415e-06, 1.0320e-07, 1.7579e-08, ..., -1.4091e-06, + 2.2119e-09, 9.8953e-09], + [-1.7916e-07, -1.0207e-06, -3.8976e-07, ..., 3.0175e-07, + 3.8417e-09, -6.5798e-07], + [ 2.8638e-07, 1.0466e-07, 1.0006e-07, ..., 3.0641e-07, + 2.3865e-08, 4.0338e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([ 0.0324, -0.0262, -0.0035, 0.0132, -0.0008, -0.0149, 0.0042, 0.0139, + -0.0280, 0.0107], device='cuda:0'), grad: tensor([ 5.7649e-07, 1.6382e-06, 1.0245e-06, 1.8060e-05, 8.3528e-08, + -2.4617e-05, 8.9966e-07, -4.3884e-06, 2.1067e-06, 4.6119e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 217.20, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4730 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.06 lr 0.00010000 +Epoch 216, weight, value: tensor([[ 0.0171, -0.0743, 0.0112, ..., -0.0947, -0.1270, -0.0958], + [-0.1052, 0.0663, -0.1038, ..., -0.1225, -0.1203, -0.1732], + [-0.0781, -0.1210, -0.1097, ..., -0.1676, -0.1056, 0.1375], + ..., + [ 0.0784, -0.0522, -0.1136, ..., 0.0916, -0.1010, -0.1097], + [ 0.1169, 0.0945, 0.0678, ..., -0.1519, -0.1011, 0.0730], + [ 0.0610, 0.0252, 0.0876, ..., 0.0477, -0.0799, 0.0045]], + device='cuda:0'), grad: tensor([[ 3.0675e-08, 3.3935e-08, 1.1118e-08, ..., 9.1968e-09, + 5.5239e-08, 8.0036e-08], + [ 4.9418e-08, -9.6043e-09, 1.2806e-08, ..., 4.7323e-08, + 2.1362e-08, 3.4051e-08], + [ 9.5519e-08, 8.6438e-08, 2.6252e-08, ..., 6.3737e-08, + 2.5262e-08, 2.7532e-08], + ..., + [-1.8091e-07, 3.5390e-08, 5.5297e-09, ..., -2.5285e-07, + -3.6322e-08, -1.8044e-09], + [-2.8987e-07, -2.3912e-07, -1.3784e-07, ..., 1.8917e-08, + 3.7311e-08, -4.4471e-08], + [ 1.3248e-07, 8.5274e-08, 2.0955e-08, ..., 3.1060e-07, + 1.9162e-07, 1.0373e-07]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0331, -0.0260, -0.0034, 0.0133, -0.0010, -0.0146, 0.0033, 0.0135, + -0.0290, 0.0110], device='cuda:0'), grad: tensor([ 3.2736e-07, 9.9128e-08, 3.3737e-07, 8.8150e-07, -3.5530e-07, + -2.8275e-06, 7.4133e-07, -4.0955e-07, 1.9034e-07, 1.0319e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 217.28, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4699 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.18 lr 0.00010000 +Epoch 217, weight, value: tensor([[ 0.0172, -0.0745, 0.0113, ..., -0.0948, -0.1274, -0.0962], + [-0.1055, 0.0666, -0.1040, ..., -0.1230, -0.1211, -0.1733], + [-0.0791, -0.1218, -0.1102, ..., -0.1688, -0.1061, 0.1374], + ..., + [ 0.0788, -0.0529, -0.1141, ..., 0.0923, -0.1012, -0.1098], + [ 0.1178, 0.0956, 0.0683, ..., -0.1525, -0.1010, 0.0739], + [ 0.0609, 0.0253, 0.0878, ..., 0.0477, -0.0800, 0.0045]], + device='cuda:0'), grad: tensor([[-7.9744e-09, 6.1118e-09, 2.1537e-09, ..., 3.0443e-08, + 1.7870e-08, 3.0617e-08], + [ 4.2724e-08, -1.2573e-08, 2.7358e-09, ..., 8.0559e-08, + 2.2352e-08, 3.8475e-08], + [ 4.2084e-08, 9.7207e-09, 3.7835e-09, ..., 7.8406e-08, + 2.4796e-08, -1.0803e-07], + ..., + [-2.7171e-07, 2.2119e-08, 7.3342e-09, ..., -2.7940e-07, + 3.1781e-08, 2.5844e-08], + [ 8.2073e-09, 5.5996e-08, 1.1176e-08, ..., 3.3004e-08, + 1.0652e-08, 5.5647e-08], + [ 1.2619e-07, -2.8580e-08, -1.6764e-08, ..., 6.9616e-07, + 4.2818e-07, 2.6310e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0331, -0.0260, -0.0037, 0.0133, -0.0011, -0.0152, 0.0034, 0.0137, + -0.0277, 0.0109], device='cuda:0'), grad: tensor([ 1.0349e-07, 2.3108e-07, -2.7940e-07, 2.9663e-07, -1.4277e-06, + -9.8720e-07, 5.3877e-07, -5.6997e-07, 5.5693e-07, 1.5292e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 217.50, cls_loss 0.0012 cls_loss_mapping 0.0029 cls_loss_causal 0.4605 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.17 lr 0.00010000 +Epoch 218, weight, value: tensor([[ 0.0171, -0.0748, 0.0112, ..., -0.0951, -0.1277, -0.0964], + [-0.1062, 0.0664, -0.1042, ..., -0.1237, -0.1216, -0.1735], + [-0.0793, -0.1221, -0.1106, ..., -0.1695, -0.1068, 0.1375], + ..., + [ 0.0792, -0.0527, -0.1146, ..., 0.0924, -0.1026, -0.1103], + [ 0.1155, 0.0935, 0.0662, ..., -0.1534, -0.1011, 0.0730], + [ 0.0611, 0.0253, 0.0881, ..., 0.0477, -0.0803, 0.0043]], + device='cuda:0'), grad: tensor([[ 2.2235e-07, 2.0186e-07, 1.3190e-07, ..., 2.1537e-07, + 1.2945e-07, 2.1746e-07], + [ 8.9814e-08, 2.1420e-08, 3.0443e-08, ..., 1.7369e-07, + 4.9709e-08, 4.9884e-08], + [ 6.8313e-07, 3.8464e-07, 1.7160e-07, ..., 8.5915e-07, + 3.2689e-07, 5.5227e-07], + ..., + [-1.0733e-07, 7.7998e-08, 3.2480e-08, ..., -3.5320e-07, + 9.9535e-09, 4.3248e-08], + [-6.0583e-07, -1.1725e-06, -1.0207e-06, ..., 2.1770e-08, + 4.9360e-08, -3.7160e-07], + [-1.0366e-06, -3.6904e-07, -7.4098e-08, ..., -1.3364e-06, + -4.5751e-07, -8.4518e-07]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0331, -0.0263, -0.0038, 0.0144, -0.0008, -0.0145, 0.0038, 0.0140, + -0.0301, 0.0108], device='cuda:0'), grad: tensor([ 1.1055e-06, 4.4936e-07, 3.3397e-06, 1.1856e-06, 1.1623e-06, + 6.7381e-07, -3.5623e-07, -4.6892e-07, -2.6133e-06, -4.4927e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 217.37, cls_loss 0.0011 cls_loss_mapping 0.0029 cls_loss_causal 0.4768 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.09 lr 0.00010000 +Epoch 219, weight, value: tensor([[ 0.0174, -0.0752, 0.0109, ..., -0.0952, -0.1284, -0.0969], + [-0.1064, 0.0655, -0.1044, ..., -0.1243, -0.1230, -0.1720], + [-0.0799, -0.1237, -0.1113, ..., -0.1714, -0.1088, 0.1375], + ..., + [ 0.0795, -0.0534, -0.1151, ..., 0.0929, -0.1025, -0.1107], + [ 0.1155, 0.0944, 0.0665, ..., -0.1542, -0.1011, 0.0724], + [ 0.0612, 0.0256, 0.0884, ..., 0.0479, -0.0804, 0.0043]], + device='cuda:0'), grad: tensor([[ 5.9954e-09, 1.4144e-08, 1.1118e-08, ..., 1.1717e-07, + 6.8033e-07, 5.5972e-07], + [ 5.5297e-09, 6.9849e-10, 1.1642e-08, ..., 1.1991e-08, + 2.5029e-08, 2.2177e-08], + [ 6.6939e-09, 1.8219e-08, 7.3924e-09, ..., 4.7730e-09, + 3.1083e-08, 9.3132e-09], + ..., + [ 4.8312e-09, 1.9150e-08, 1.1234e-08, ..., 8.1491e-10, + 1.2224e-09, 5.6461e-09], + [-2.5029e-09, 2.7881e-08, 3.6729e-08, ..., 4.9360e-08, + 1.0472e-07, 6.3330e-08], + [-6.0420e-08, -2.1164e-07, -1.8068e-07, ..., -1.4261e-07, + 1.8917e-08, 1.2922e-08]], device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0331, -0.0285, -0.0043, 0.0144, -0.0008, -0.0144, 0.0037, 0.0139, + -0.0287, 0.0109], device='cuda:0'), grad: tensor([ 2.0117e-06, 7.2701e-08, 9.5810e-08, 3.4506e-07, 1.1834e-07, + 3.0827e-07, -3.2373e-06, 6.7987e-08, 4.1933e-07, -1.8475e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 217.38, cls_loss 0.0010 cls_loss_mapping 0.0033 cls_loss_causal 0.4984 re_mapping 0.0036 re_causal 0.0112 /// teacc 99.03 lr 0.00010000 +Epoch 220, weight, value: tensor([[ 0.0174, -0.0756, 0.0108, ..., -0.0954, -0.1282, -0.0974], + [-0.1065, 0.0648, -0.1074, ..., -0.1244, -0.1260, -0.1746], + [-0.0796, -0.1254, -0.1125, ..., -0.1706, -0.1096, 0.1391], + ..., + [ 0.0794, -0.0545, -0.1155, ..., 0.0927, -0.1044, -0.1133], + [ 0.1155, 0.0955, 0.0682, ..., -0.1553, -0.0991, 0.0744], + [ 0.0612, 0.0256, 0.0887, ..., 0.0480, -0.0804, 0.0044]], + device='cuda:0'), grad: tensor([[-2.8347e-08, 2.8522e-08, 2.2410e-08, ..., 2.3225e-08, + 1.1339e-07, 7.8231e-08], + [ 2.4855e-08, 1.9791e-09, 1.4668e-08, ..., 6.9558e-08, + 2.2585e-08, 1.6589e-08], + [ 4.2550e-08, 5.5821e-08, 3.2946e-08, ..., 5.5821e-08, + 2.2934e-08, 3.5740e-08], + ..., + [-1.1630e-07, 1.9441e-08, 7.5670e-09, ..., -3.9651e-07, + 2.1188e-08, 6.2864e-09], + [-4.8429e-08, -1.7998e-07, -1.5937e-07, ..., 1.2631e-08, + 1.1350e-08, -1.2619e-07], + [ 2.0373e-08, -8.8301e-08, -6.6590e-08, ..., 1.4994e-07, + 7.1246e-08, 8.0327e-09]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0337, -0.0289, -0.0043, 0.0144, -0.0008, -0.0145, 0.0017, 0.0130, + -0.0283, 0.0109], device='cuda:0'), grad: tensor([ 2.5472e-07, 3.8324e-07, 4.2515e-07, 8.5449e-07, 4.7917e-07, + -6.9067e-06, 4.6156e-06, -5.8860e-07, -9.8837e-08, 5.6345e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 217.04, cls_loss 0.0013 cls_loss_mapping 0.0037 cls_loss_causal 0.5118 re_mapping 0.0037 re_causal 0.0116 /// teacc 99.05 lr 0.00010000 +Epoch 221, weight, value: tensor([[ 0.0152, -0.0774, 0.0105, ..., -0.0978, -0.1307, -0.1006], + [-0.1075, 0.0646, -0.1075, ..., -0.1254, -0.1264, -0.1752], + [-0.0816, -0.1270, -0.1131, ..., -0.1720, -0.1108, 0.1387], + ..., + [ 0.0812, -0.0526, -0.1159, ..., 0.0944, -0.1034, -0.1109], + [ 0.1155, 0.0955, 0.0684, ..., -0.1570, -0.0990, 0.0744], + [ 0.0619, 0.0268, 0.0893, ..., 0.0485, -0.0805, 0.0048]], + device='cuda:0'), grad: tensor([[-7.9744e-08, 3.7998e-07, 2.1502e-07, ..., 2.0175e-07, + 5.4366e-08, 6.1002e-08], + [ 1.6938e-08, 4.2003e-07, 2.5472e-07, ..., 2.2841e-07, + 8.3237e-09, 4.2142e-08], + [ 4.2783e-08, 5.8720e-07, 3.2666e-07, ..., 2.9057e-07, + 2.4156e-08, 2.5728e-08], + ..., + [ 3.3178e-09, 2.1455e-07, 1.2608e-07, ..., 8.7486e-08, + 4.6566e-10, 2.2817e-08], + [-3.9116e-08, 4.3698e-06, 2.6561e-06, ..., 2.3060e-06, + 4.2899e-08, 3.3015e-07], + [-6.1642e-08, -4.0472e-05, -2.4259e-05, ..., -2.1085e-05, + 3.1432e-09, -2.9355e-06]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0317, -0.0291, -0.0046, 0.0140, -0.0010, -0.0131, 0.0019, 0.0150, + -0.0284, 0.0093], device='cuda:0'), grad: tensor([ 8.0559e-07, 1.2405e-06, 1.4249e-06, 9.0659e-05, 3.4496e-06, + 8.1374e-08, -7.7125e-08, 5.9977e-07, 1.2197e-05, -1.1051e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 217.39, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.4887 re_mapping 0.0038 re_causal 0.0114 /// teacc 98.96 lr 0.00010000 +Epoch 222, weight, value: tensor([[ 0.0154, -0.0776, 0.0105, ..., -0.0978, -0.1309, -0.1007], + [-0.1077, 0.0650, -0.1076, ..., -0.1259, -0.1267, -0.1752], + [-0.0822, -0.1294, -0.1151, ..., -0.1724, -0.1115, 0.1377], + ..., + [ 0.0827, -0.0516, -0.1173, ..., 0.0957, -0.1038, -0.1111], + [ 0.1156, 0.0957, 0.0686, ..., -0.1580, -0.0988, 0.0752], + [ 0.0609, 0.0267, 0.0912, ..., 0.0477, -0.0807, 0.0047]], + device='cuda:0'), grad: tensor([[ 4.1036e-08, 1.1828e-07, 7.7474e-08, ..., 7.1595e-08, + 1.4086e-08, 6.4902e-08], + [ 2.3050e-08, -1.0070e-08, 1.4377e-08, ..., 2.5379e-08, + 9.3714e-09, 1.3330e-08], + [ 1.4552e-08, 2.5786e-08, 4.8894e-09, ..., 1.0710e-08, + 1.1758e-08, 1.4144e-08], + ..., + [ 3.8231e-07, 4.1979e-07, 2.8731e-07, ..., 2.6496e-07, + 1.0245e-08, 1.6647e-08], + [-1.9034e-08, 1.9022e-07, -2.0955e-09, ..., 3.9581e-08, + -2.2759e-08, -7.8348e-08], + [-5.8115e-07, -6.1560e-07, -4.5728e-07, ..., -2.2724e-07, + 1.2619e-07, 1.5122e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0319, -0.0290, -0.0052, 0.0136, -0.0003, -0.0131, 0.0009, 0.0164, + -0.0283, 0.0083], device='cuda:0'), grad: tensor([ 9.3132e-08, 4.5809e-08, 7.8464e-08, 3.9190e-06, -3.3155e-07, + -5.3383e-06, 3.8277e-07, 1.2442e-06, 1.0431e-06, -1.1465e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 217.17, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.4840 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 223, weight, value: tensor([[ 0.0153, -0.0780, 0.0102, ..., -0.0983, -0.1311, -0.1011], + [-0.1086, 0.0650, -0.1079, ..., -0.1301, -0.1271, -0.1768], + [-0.0824, -0.1288, -0.1156, ..., -0.1752, -0.1145, 0.1387], + ..., + [ 0.0841, -0.0532, -0.1147, ..., 0.0966, -0.1049, -0.1121], + [ 0.1156, 0.0956, 0.0686, ..., -0.1618, -0.0993, 0.0750], + [ 0.0596, 0.0282, 0.0882, ..., 0.0470, -0.0814, 0.0044]], + device='cuda:0'), grad: tensor([[-1.6345e-07, 1.2224e-08, -1.0477e-09, ..., 1.6356e-08, + 5.4799e-06, 4.1127e-06], + [ 6.9907e-08, 4.6683e-08, 1.0477e-09, ..., 2.0396e-07, + 1.2293e-07, 1.1077e-07], + [ 4.4541e-07, 6.6473e-08, 2.3865e-09, ..., 3.3900e-06, + 8.5309e-07, 3.8510e-07], + ..., + [-5.7276e-07, 5.6461e-09, 1.2806e-09, ..., -4.1462e-06, + -8.5356e-07, -3.3597e-07], + [-2.6915e-07, -2.5588e-07, 1.7928e-08, ..., 1.4319e-08, + 5.4855e-07, 2.3632e-07], + [ 1.4575e-07, 2.6799e-07, 1.8044e-09, ..., 6.9663e-07, + 1.6717e-07, 1.9139e-07]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0314, -0.0295, -0.0048, 0.0131, -0.0003, -0.0133, 0.0010, 0.0176, + -0.0284, 0.0074], device='cuda:0'), grad: tensor([ 1.3813e-05, 9.6299e-07, 6.0126e-06, 7.1386e-07, 1.8138e-07, + 7.0296e-06, -2.3946e-05, -7.3798e-06, 6.1188e-07, 2.0247e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 217.10, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.4718 re_mapping 0.0036 re_causal 0.0114 /// teacc 99.10 lr 0.00010000 +Epoch 224, weight, value: tensor([[ 0.0153, -0.0782, 0.0097, ..., -0.0986, -0.1317, -0.1017], + [-0.1090, 0.0650, -0.1081, ..., -0.1309, -0.1274, -0.1769], + [-0.0827, -0.1289, -0.1163, ..., -0.1766, -0.1161, 0.1385], + ..., + [ 0.0841, -0.0532, -0.1147, ..., 0.0966, -0.1048, -0.1124], + [ 0.1156, 0.0956, 0.0687, ..., -0.1637, -0.0997, 0.0750], + [ 0.0594, 0.0285, 0.0890, ..., 0.0464, -0.0835, 0.0027]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.7986e-08, 2.3283e-09, ..., 3.7951e-08, + 1.3295e-07, 1.2328e-07], + [ 2.2352e-08, 2.4331e-08, 2.0955e-09, ..., 7.9686e-08, + 5.7393e-08, 1.6042e-07], + [ 1.5926e-07, 1.3853e-08, 1.2806e-09, ..., 3.0105e-07, + 9.3190e-08, -2.4354e-07], + ..., + [-2.1001e-07, 2.0489e-08, 4.5984e-09, ..., -3.6694e-07, + 6.6357e-09, 1.2433e-07], + [ 3.0268e-09, 1.1700e-08, -7.5670e-10, ..., 3.3935e-08, + 1.2061e-07, 1.3644e-07], + [-3.3295e-08, 2.9989e-07, -1.7055e-08, ..., 8.2375e-07, + 9.8720e-07, 8.9500e-07]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0312, -0.0296, -0.0049, 0.0124, 0.0013, -0.0141, 0.0015, 0.0176, + -0.0284, 0.0074], device='cuda:0'), grad: tensor([ 3.5902e-07, 7.4646e-07, -4.0024e-07, 2.1164e-07, -3.5781e-06, + 1.1791e-06, -1.6494e-06, -4.6939e-07, 4.7428e-07, 3.1255e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 217.24, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.5310 re_mapping 0.0034 re_causal 0.0113 /// teacc 98.95 lr 0.00010000 +Epoch 225, weight, value: tensor([[ 0.0154, -0.0784, 0.0095, ..., -0.0987, -0.1322, -0.1021], + [-0.1093, 0.0652, -0.1082, ..., -0.1315, -0.1280, -0.1771], + [-0.0830, -0.1295, -0.1174, ..., -0.1770, -0.1165, 0.1385], + ..., + [ 0.0841, -0.0532, -0.1147, ..., 0.0967, -0.1038, -0.1125], + [ 0.1157, 0.0957, 0.0691, ..., -0.1645, -0.0999, 0.0757], + [ 0.0593, 0.0284, 0.0891, ..., 0.0459, -0.0846, 0.0016]], + device='cuda:0'), grad: tensor([[ 3.7660e-08, 3.6613e-08, 1.7695e-08, ..., 1.5425e-08, + 7.1188e-08, 9.2143e-08], + [ 1.2969e-07, -3.0850e-09, 4.7323e-08, ..., 5.3085e-08, + 8.5565e-09, 1.5995e-07], + [ 1.5944e-06, 9.2993e-07, 6.6496e-07, ..., 3.9639e-08, + 2.4913e-08, 1.9837e-06], + ..., + [ 1.1286e-07, 1.5821e-07, 7.4739e-08, ..., -1.1135e-07, + 2.9104e-10, 2.3935e-07], + [-2.2277e-06, -1.3858e-06, -9.4622e-07, ..., 1.0594e-08, + 2.4331e-08, -2.8759e-06], + [-2.6368e-08, -4.0978e-08, -1.3621e-08, ..., -9.5111e-08, + 1.7462e-09, 4.2550e-08]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0312, -0.0296, -0.0052, 0.0124, 0.0022, -0.0142, 0.0023, 0.0176, + -0.0284, 0.0072], device='cuda:0'), grad: tensor([ 3.3528e-07, 4.4843e-07, 6.2585e-06, 8.1304e-07, 3.0501e-07, + 4.6310e-07, -6.8918e-08, 6.1328e-07, -9.0823e-06, -7.0839e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 217.26, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4558 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.03 lr 0.00010000 +Epoch 226, weight, value: tensor([[ 0.0155, -0.0782, 0.0089, ..., -0.0988, -0.1323, -0.1023], + [-0.1097, 0.0652, -0.1083, ..., -0.1317, -0.1281, -0.1772], + [-0.0836, -0.1299, -0.1183, ..., -0.1772, -0.1166, 0.1386], + ..., + [ 0.0840, -0.0533, -0.1148, ..., 0.0966, -0.1039, -0.1126], + [ 0.1159, 0.0959, 0.0695, ..., -0.1653, -0.1003, 0.0760], + [ 0.0593, 0.0284, 0.0892, ..., 0.0460, -0.0847, 0.0016]], + device='cuda:0'), grad: tensor([[-5.6112e-08, 2.1595e-08, 3.7253e-09, ..., 3.8417e-08, + 2.9104e-09, 1.0594e-08], + [ 1.3120e-07, -1.0460e-07, 9.1386e-09, ..., 2.2189e-07, + 2.1537e-09, 2.7241e-08], + [ 4.2724e-08, 3.9930e-08, 2.3283e-09, ..., 6.2748e-08, + 1.5716e-09, -5.6170e-08], + ..., + [-2.4773e-07, 4.0396e-08, 2.2701e-09, ..., -4.8429e-07, + 3.2014e-09, 1.6473e-08], + [ 2.5728e-08, 1.9965e-08, -4.6566e-10, ..., 4.2666e-08, + 4.8894e-09, 1.5367e-08], + [-3.5344e-07, -4.7521e-07, -9.9419e-08, ..., -2.0000e-07, + 8.5449e-08, -2.8696e-08]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0316, -0.0296, -0.0052, 0.0120, 0.0022, -0.0142, 0.0020, 0.0176, + -0.0283, 0.0072], device='cuda:0'), grad: tensor([-5.4296e-07, 3.2061e-07, 5.7102e-08, 3.1618e-07, 1.4435e-06, + 1.8650e-07, 4.2748e-07, -1.3029e-06, 2.8592e-07, -1.1744e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 217.20, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4662 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.13 lr 0.00010000 +Epoch 227, weight, value: tensor([[ 0.0154, -0.0784, 0.0092, ..., -0.0992, -0.1325, -0.1025], + [-0.1100, 0.0653, -0.1084, ..., -0.1321, -0.1285, -0.1775], + [-0.0839, -0.1309, -0.1188, ..., -0.1774, -0.1166, 0.1389], + ..., + [ 0.0840, -0.0533, -0.1148, ..., 0.0967, -0.1037, -0.1128], + [ 0.1159, 0.0959, 0.0695, ..., -0.1659, -0.1009, 0.0761], + [ 0.0594, 0.0284, 0.0892, ..., 0.0460, -0.0847, 0.0016]], + device='cuda:0'), grad: tensor([[-5.1083e-07, 2.4796e-08, 8.4401e-09, ..., 1.3853e-08, + 5.1805e-09, 3.3237e-08], + [ 1.2631e-08, -4.7206e-08, 6.6939e-09, ..., 1.1583e-08, + 1.0477e-09, 1.1525e-08], + [ 1.5949e-08, 4.8371e-08, 1.3504e-08, ..., 8.6147e-09, + 3.2014e-09, -1.6124e-08], + ..., + [-2.3283e-09, 2.0897e-08, 4.0163e-09, ..., -7.3924e-09, + 2.6193e-09, 7.6252e-09], + [ 6.4145e-08, 3.5390e-08, 2.1537e-09, ..., 3.8068e-08, + 6.9267e-09, 7.5786e-08], + [-2.1618e-07, -9.1328e-08, -1.7602e-07, ..., -1.2154e-07, + 1.4051e-07, -1.3458e-07]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0314, -0.0295, -0.0064, 0.0133, 0.0021, -0.0142, 0.0028, 0.0176, + -0.0283, 0.0072], device='cuda:0'), grad: tensor([-2.6189e-06, -2.4401e-07, 2.2759e-08, -1.0978e-07, 1.8370e-07, + 1.1232e-06, 1.6652e-06, 7.9046e-08, 2.8545e-07, -3.7532e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 217.32, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4771 re_mapping 0.0034 re_causal 0.0113 /// teacc 99.09 lr 0.00010000 +Epoch 228, weight, value: tensor([[ 0.0160, -0.0785, 0.0092, ..., -0.0992, -0.1322, -0.1023], + [-0.1079, 0.0657, -0.1085, ..., -0.1309, -0.1288, -0.1776], + [-0.0806, -0.1308, -0.1191, ..., -0.1772, -0.1169, 0.1416], + ..., + [ 0.0837, -0.0534, -0.1148, ..., 0.0966, -0.1043, -0.1159], + [ 0.1154, 0.0959, 0.0696, ..., -0.1666, -0.1010, 0.0758], + [ 0.0594, 0.0284, 0.0893, ..., 0.0460, -0.0848, 0.0016]], + device='cuda:0'), grad: tensor([[-4.2957e-08, 3.8883e-08, 3.4343e-09, ..., 6.6939e-09, + 3.7835e-09, 3.7428e-08], + [ 1.6915e-07, -1.1118e-07, 2.3865e-09, ..., 1.7078e-07, + 1.0477e-09, 1.2107e-08], + [ 1.3551e-07, 6.9034e-08, 1.4494e-08, ..., 5.5123e-08, + 2.0373e-09, 7.2701e-08], + ..., + [-2.4773e-07, 5.8790e-08, 5.4715e-09, ..., -3.0780e-07, + 1.3970e-09, 1.3621e-08], + [-3.0082e-07, -2.2852e-07, -2.6426e-08, ..., 1.7753e-08, + 2.7358e-09, -3.4226e-07], + [ 5.5297e-08, -2.2119e-09, -1.6240e-08, ..., 2.3341e-08, + 7.9162e-09, 1.9383e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0320, -0.0286, -0.0043, 0.0132, 0.0021, -0.0142, 0.0027, 0.0172, + -0.0285, 0.0072], device='cuda:0'), grad: tensor([-1.5367e-07, 3.2876e-07, 4.6240e-07, 1.5600e-07, 1.0332e-07, + 3.0571e-07, 2.6869e-07, -9.3831e-07, -7.7160e-07, 2.3143e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 217.25, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4776 re_mapping 0.0034 re_causal 0.0112 /// teacc 99.11 lr 0.00010000 +Epoch 229, weight, value: tensor([[ 0.0161, -0.0786, 0.0090, ..., -0.0992, -0.1321, -0.1021], + [-0.1085, 0.0657, -0.1086, ..., -0.1313, -0.1290, -0.1777], + [-0.0809, -0.1323, -0.1225, ..., -0.1777, -0.1171, 0.1412], + ..., + [ 0.0838, -0.0534, -0.1149, ..., 0.0967, -0.1043, -0.1161], + [ 0.1154, 0.0960, 0.0699, ..., -0.1675, -0.1011, 0.0761], + [ 0.0594, 0.0285, 0.0894, ..., 0.0460, -0.0848, 0.0016]], + device='cuda:0'), grad: tensor([[ 5.9255e-08, 7.5670e-09, 7.5088e-09, ..., 5.0641e-09, + 6.4611e-08, 1.7253e-07], + [ 9.4704e-08, -2.3236e-07, 2.2701e-09, ..., 3.1258e-08, + 6.5775e-09, 1.4633e-07], + [-5.9605e-07, 6.9966e-08, 2.8522e-09, ..., 6.7404e-08, + 6.2864e-09, -1.3588e-06], + ..., + [ 1.5926e-07, 9.9884e-08, 1.9791e-09, ..., -1.2678e-07, + 1.1059e-08, 6.1607e-07], + [ 6.8976e-08, -8.7894e-09, -7.9744e-09, ..., 1.4959e-08, + 5.5472e-08, 1.9884e-07], + [ 3.8592e-08, 9.7265e-08, -2.3283e-10, ..., 2.1176e-07, + 1.6997e-07, 1.9546e-07]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0323, -0.0288, -0.0050, 0.0136, 0.0020, -0.0142, 0.0027, 0.0172, + -0.0285, 0.0072], device='cuda:0'), grad: tensor([ 6.6822e-07, -2.0629e-07, -5.3570e-06, 7.2038e-07, -2.4377e-07, + 4.7148e-08, -3.3877e-08, 2.5667e-06, 8.7637e-07, 9.5926e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 217.26, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.4938 re_mapping 0.0035 re_causal 0.0110 /// teacc 99.07 lr 0.00010000 +Epoch 230, weight, value: tensor([[ 0.0161, -0.0788, 0.0089, ..., -0.0993, -0.1338, -0.1032], + [-0.1087, 0.0658, -0.1087, ..., -0.1317, -0.1293, -0.1779], + [-0.0810, -0.1328, -0.1233, ..., -0.1779, -0.1175, 0.1413], + ..., + [ 0.0838, -0.0535, -0.1149, ..., 0.0967, -0.1051, -0.1164], + [ 0.1155, 0.0961, 0.0701, ..., -0.1691, -0.1014, 0.0763], + [ 0.0594, 0.0285, 0.0895, ..., 0.0460, -0.0849, 0.0015]], + device='cuda:0'), grad: tensor([[ 2.9278e-08, 3.8475e-08, 1.3097e-08, ..., 3.5390e-08, + 6.8860e-08, 7.5030e-08], + [ 1.2689e-08, -2.4564e-08, 3.6671e-09, ..., 1.5716e-08, + 1.7462e-08, 2.2759e-08], + [ 1.3330e-08, 1.5600e-08, 3.7835e-09, ..., 4.9477e-09, + 2.1886e-08, -1.5076e-08], + ..., + [ 9.5519e-08, 9.5461e-08, 2.2352e-08, ..., 1.3458e-07, + 2.4971e-08, 8.6438e-08], + [ 1.3201e-07, 1.5367e-07, 1.1967e-07, ..., 1.5204e-07, + 7.2992e-08, 8.4809e-08], + [-4.0047e-07, -3.7509e-07, -1.9767e-07, ..., -3.9302e-07, + 1.4261e-08, -1.5553e-07]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0310, -0.0288, -0.0050, 0.0135, 0.0020, -0.0142, 0.0039, 0.0172, + -0.0285, 0.0072], device='cuda:0'), grad: tensor([ 2.4564e-07, -7.1013e-09, -1.8603e-07, 1.5914e-07, 1.8172e-07, + 2.3562e-07, -6.7614e-07, 5.2294e-07, 6.4634e-07, -1.1120e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 217.47, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4597 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.03 lr 0.00010000 +Epoch 231, weight, value: tensor([[ 0.0161, -0.0790, 0.0087, ..., -0.0995, -0.1340, -0.1033], + [-0.1090, 0.0658, -0.1089, ..., -0.1326, -0.1297, -0.1781], + [-0.0809, -0.1330, -0.1237, ..., -0.1781, -0.1177, 0.1419], + ..., + [ 0.0837, -0.0535, -0.1149, ..., 0.0966, -0.1058, -0.1169], + [ 0.1155, 0.0961, 0.0701, ..., -0.1708, -0.1014, 0.0764], + [ 0.0595, 0.0286, 0.0896, ..., 0.0461, -0.0849, 0.0017]], + device='cuda:0'), grad: tensor([[-2.2957e-07, 4.9360e-08, 5.0757e-08, ..., -1.8626e-08, + 6.1118e-09, 7.1595e-09], + [ 1.0425e-07, -3.6578e-07, 4.6741e-08, ..., 1.5437e-07, + 4.1910e-09, 7.1595e-09], + [ 1.6531e-07, 4.6450e-08, 1.6706e-08, ..., 1.4366e-07, + 1.4552e-09, -3.7428e-08], + ..., + [-1.1496e-07, 3.4855e-07, 8.9989e-08, ..., -2.1013e-07, + 1.4552e-09, 2.9104e-09], + [ 3.7311e-08, 5.9779e-08, 2.8114e-08, ..., 5.0699e-08, + 2.3865e-09, 1.9209e-08], + [-2.2328e-07, -2.7381e-07, -4.4727e-07, ..., -6.1560e-07, + 8.6729e-09, 3.9581e-09]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0310, -0.0288, -0.0047, 0.0134, 0.0019, -0.0142, 0.0038, 0.0172, + -0.0285, 0.0073], device='cuda:0'), grad: tensor([ 1.0796e-05, -8.5775e-07, 1.2457e-05, 5.8524e-06, 1.4361e-06, + -5.8591e-05, 4.1798e-06, 3.8706e-06, 1.5348e-05, 5.5619e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 217.43, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4966 re_mapping 0.0034 re_causal 0.0110 /// teacc 98.91 lr 0.00010000 +Epoch 232, weight, value: tensor([[ 0.0160, -0.0794, 0.0080, ..., -0.0999, -0.1344, -0.1038], + [-0.1092, 0.0658, -0.1091, ..., -0.1332, -0.1301, -0.1784], + [-0.0814, -0.1368, -0.1248, ..., -0.1794, -0.1180, 0.1422], + ..., + [ 0.0838, -0.0536, -0.1150, ..., 0.0967, -0.1061, -0.1171], + [ 0.1156, 0.0963, 0.0701, ..., -0.1727, -0.1016, 0.0766], + [ 0.0595, 0.0286, 0.0898, ..., 0.0462, -0.0850, 0.0017]], + device='cuda:0'), grad: tensor([[-9.4878e-09, 1.9791e-08, 1.2631e-08, ..., 2.1886e-08, + 6.4028e-09, 4.4820e-09], + [ 2.4855e-08, -1.4994e-07, 1.4086e-08, ..., 2.9395e-08, + 1.5949e-08, 1.8626e-08], + [ 5.9954e-09, 2.0489e-07, 9.9535e-09, ..., 4.8603e-08, + 2.2817e-08, 8.6147e-08], + ..., + [-4.8720e-08, 6.6531e-08, 2.1246e-08, ..., -4.3889e-08, + 1.1059e-08, 8.5565e-09], + [ 3.1199e-08, -5.1165e-08, 1.4028e-08, ..., 2.8929e-08, + 1.8277e-08, -1.5541e-07], + [-6.5484e-08, -4.8720e-08, -5.6811e-08, ..., -5.5472e-08, + 2.0606e-08, 9.5461e-09]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0308, -0.0289, -0.0081, 0.0159, 0.0019, -0.0140, 0.0036, 0.0172, + -0.0284, 0.0073], device='cuda:0'), grad: tensor([-1.9209e-09, -1.8300e-07, -8.1724e-07, 5.5041e-07, -1.0245e-07, + 2.1339e-07, 1.1391e-07, 1.8044e-07, 1.5728e-07, -9.8662e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 217.27, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4965 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.08 lr 0.00010000 +Epoch 233, weight, value: tensor([[ 0.0160, -0.0794, 0.0070, ..., -0.1002, -0.1350, -0.1045], + [-0.1095, 0.0658, -0.1093, ..., -0.1344, -0.1308, -0.1788], + [-0.0815, -0.1370, -0.1252, ..., -0.1807, -0.1189, 0.1424], + ..., + [ 0.0838, -0.0536, -0.1150, ..., 0.0968, -0.1060, -0.1172], + [ 0.1155, 0.0964, 0.0689, ..., -0.1742, -0.1040, 0.0756], + [ 0.0595, 0.0287, 0.0900, ..., 0.0462, -0.0851, 0.0016]], + device='cuda:0'), grad: tensor([[-1.1234e-07, 1.0169e-07, 1.5821e-07, ..., 3.9674e-07, + 2.5349e-08, 2.5053e-07], + [ 3.3341e-07, -1.4727e-08, 1.4028e-08, ..., 5.4296e-07, + 3.9698e-07, 1.6135e-07], + [ 9.6043e-08, 1.7404e-08, 5.3842e-09, ..., 8.3819e-08, + 3.8272e-08, 2.9919e-07], + ..., + [-2.7847e-07, 8.1956e-08, 1.1118e-07, ..., -4.1298e-08, + 1.2224e-08, 4.9389e-08], + [-1.2253e-08, -5.0233e-08, -6.3737e-09, ..., 4.2521e-08, + 1.5280e-08, 2.7963e-07], + [-4.3283e-07, -2.4261e-07, -4.1071e-07, ..., -8.6753e-07, + 2.0768e-07, 2.3528e-07]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0305, -0.0291, -0.0082, 0.0158, 0.0019, -0.0140, 0.0051, 0.0173, + -0.0285, 0.0073], device='cuda:0'), grad: tensor([ 3.3956e-06, 1.7937e-06, 3.4478e-06, 1.8310e-06, -7.1749e-06, + -1.6779e-05, 9.6336e-06, -3.9511e-07, 3.8594e-06, 3.5693e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 217.36, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4818 re_mapping 0.0034 re_causal 0.0113 /// teacc 99.09 lr 0.00010000 +Epoch 234, weight, value: tensor([[ 0.0162, -0.0795, 0.0074, ..., -0.1003, -0.1350, -0.1046], + [-0.1100, 0.0658, -0.1094, ..., -0.1347, -0.1309, -0.1789], + [-0.0816, -0.1371, -0.1256, ..., -0.1811, -0.1191, 0.1425], + ..., + [ 0.0839, -0.0536, -0.1150, ..., 0.0968, -0.1058, -0.1175], + [ 0.1156, 0.0964, 0.0690, ..., -0.1756, -0.1042, 0.0757], + [ 0.0595, 0.0287, 0.0900, ..., 0.0462, -0.0851, 0.0018]], + device='cuda:0'), grad: tensor([[-7.4564e-08, 1.1583e-08, 1.8673e-07, ..., 1.5018e-08, + 7.9302e-07, 7.1060e-07], + [ 2.7940e-09, -9.0338e-08, 3.0268e-09, ..., 5.2387e-09, + 5.5821e-08, 4.9302e-08], + [ 1.9209e-09, 1.7753e-08, 2.5611e-09, ..., 6.4028e-10, + 2.3108e-08, -3.3178e-09], + ..., + [ 1.5192e-08, 6.4145e-08, 9.8953e-10, ..., 2.3516e-08, + 5.1223e-09, 1.7579e-08], + [ 1.0303e-08, 1.9500e-08, 2.5844e-08, ..., 8.9058e-09, + 3.5716e-07, 3.0990e-07], + [-2.7823e-08, -3.1898e-08, 1.6880e-08, ..., -5.9372e-08, + 5.4599e-08, 4.8429e-08]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0308, -0.0292, -0.0082, 0.0158, 0.0018, -0.0140, 0.0049, 0.0173, + -0.0286, 0.0073], device='cuda:0'), grad: tensor([ 2.2314e-06, -1.0285e-07, 6.3330e-08, -1.4808e-07, 7.9069e-07, + 1.6727e-06, -6.3106e-06, 2.8987e-07, 1.2917e-06, 2.3691e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 217.14, cls_loss 0.0009 cls_loss_mapping 0.0031 cls_loss_causal 0.4968 re_mapping 0.0032 re_causal 0.0106 /// teacc 99.07 lr 0.00010000 +Epoch 235, weight, value: tensor([[ 0.0163, -0.0798, 0.0072, ..., -0.1006, -0.1351, -0.1048], + [-0.1101, 0.0659, -0.1098, ..., -0.1350, -0.1310, -0.1793], + [-0.0816, -0.1373, -0.1260, ..., -0.1814, -0.1195, 0.1431], + ..., + [ 0.0838, -0.0537, -0.1151, ..., 0.0968, -0.1072, -0.1182], + [ 0.1157, 0.0969, 0.0696, ..., -0.1776, -0.1044, 0.0763], + [ 0.0596, 0.0288, 0.0902, ..., 0.0464, -0.0850, 0.0019]], + device='cuda:0'), grad: tensor([[ 5.8091e-08, 5.7626e-08, 3.2305e-08, ..., 7.5845e-08, + 1.5483e-08, 6.6531e-08], + [ 3.4866e-08, -2.5681e-07, 2.3574e-08, ..., 7.1072e-08, + 4.8778e-08, 1.4820e-07], + [ 6.1793e-07, 5.5763e-08, 2.7707e-08, ..., 5.2294e-07, + 5.4424e-08, -6.8778e-07], + ..., + [-4.9127e-07, 2.0675e-07, 5.1339e-08, ..., -4.3493e-07, + 4.4762e-08, 3.5553e-07], + [-2.1188e-07, -2.7940e-09, -2.2189e-07, ..., 1.5507e-07, + 8.7777e-08, 4.3947e-08], + [-1.8440e-07, -1.4249e-07, -1.9383e-08, ..., -1.8487e-07, + 2.2026e-07, 3.0780e-07]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0311, -0.0291, -0.0082, 0.0159, 0.0016, -0.0141, 0.0051, 0.0172, + -0.0284, 0.0074], device='cuda:0'), grad: tensor([ 4.4890e-07, -4.0745e-10, -1.9167e-06, -5.5600e-07, -1.3160e-06, + 1.5171e-06, 4.5728e-07, 5.4203e-07, 3.0594e-07, 5.2666e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 217.30, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.4989 re_mapping 0.0033 re_causal 0.0110 /// teacc 98.99 lr 0.00010000 +Epoch 236, weight, value: tensor([[ 0.0161, -0.0804, 0.0056, ..., -0.1012, -0.1353, -0.1055], + [-0.1095, 0.0679, -0.1099, ..., -0.1337, -0.1312, -0.1795], + [-0.0816, -0.1374, -0.1265, ..., -0.1816, -0.1210, 0.1438], + ..., + [ 0.0838, -0.0547, -0.1151, ..., 0.0962, -0.1082, -0.1190], + [ 0.1158, 0.0969, 0.0699, ..., -0.1794, -0.1047, 0.0764], + [ 0.0597, 0.0289, 0.0904, ..., 0.0464, -0.0851, 0.0020]], + device='cuda:0'), grad: tensor([[ 4.4284e-07, 6.1188e-07, 6.1700e-09, ..., 8.0140e-07, + 5.7626e-09, 5.6531e-07], + [ 2.0792e-07, 1.0068e-06, 3.7206e-07, ..., 1.1651e-06, + -1.7753e-08, 1.8161e-07], + [ 5.0757e-07, 8.1200e-08, 3.2014e-09, ..., 1.6857e-06, + 2.6193e-09, 5.9546e-08], + ..., + [-3.9954e-07, 2.6124e-07, 5.4715e-09, ..., -1.5162e-06, + 9.3132e-10, 2.3586e-07], + [ 1.8172e-07, 2.4936e-07, -5.4133e-09, ..., 3.4645e-07, + 7.2760e-09, 2.2352e-07], + [-2.2389e-06, -4.0159e-06, -4.3469e-07, ..., -4.9248e-06, + 1.8044e-09, -2.8480e-06]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0307, -0.0275, -0.0082, 0.0157, 0.0015, -0.0141, 0.0052, 0.0164, + -0.0285, 0.0075], device='cuda:0'), grad: tensor([ 2.3805e-06, 2.8908e-06, 3.3379e-06, 2.5239e-07, 6.0722e-06, + -2.7916e-07, 8.4285e-07, -2.5779e-06, 1.0552e-06, -1.3977e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 217.41, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4981 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.03 lr 0.00010000 +Epoch 237, weight, value: tensor([[ 0.0159, -0.0810, 0.0053, ..., -0.1019, -0.1354, -0.1058], + [-0.1092, 0.0690, -0.1101, ..., -0.1326, -0.1319, -0.1797], + [-0.0818, -0.1375, -0.1270, ..., -0.1825, -0.1214, 0.1448], + ..., + [ 0.0837, -0.0554, -0.1152, ..., 0.0957, -0.1089, -0.1201], + [ 0.1159, 0.0970, 0.0701, ..., -0.1817, -0.1046, 0.0767], + [ 0.0598, 0.0290, 0.0906, ..., 0.0465, -0.0851, 0.0020]], + device='cuda:0'), grad: tensor([[ 4.7944e-06, 4.1351e-06, 2.2873e-06, ..., 5.5097e-06, + 2.8056e-08, 2.1961e-06], + [ 4.3423e-08, -2.0047e-07, 1.9209e-08, ..., 7.0722e-08, + 1.2224e-08, 3.3062e-08], + [ 1.7812e-07, 1.5995e-07, 5.7335e-08, ..., 2.6962e-07, + 4.4121e-08, 4.0105e-08], + ..., + [-9.5053e-08, 2.6030e-07, 8.0501e-08, ..., -1.2689e-07, + 5.0815e-08, 1.1956e-07], + [ 2.7698e-06, 5.0217e-06, 2.3320e-06, ..., 3.0734e-06, + 6.1700e-08, 2.5425e-06], + [-8.0690e-06, -9.6709e-06, -4.9621e-06, ..., -8.1882e-06, + 8.2143e-07, -4.4852e-06]], device='cuda:0') +Epoch 237, bias, value: tensor([ 0.0304, -0.0265, -0.0082, 0.0157, 0.0016, -0.0142, 0.0051, 0.0158, + -0.0284, 0.0076], device='cuda:0'), grad: tensor([ 1.9789e-05, -4.0745e-07, 1.4482e-07, 7.5065e-07, -1.8384e-06, + 1.7555e-07, 2.8545e-07, 2.4121e-07, 1.0543e-05, -2.9698e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 217.34, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.4924 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.06 lr 0.00010000 +Epoch 238, weight, value: tensor([[ 0.0159, -0.0816, 0.0049, ..., -0.1021, -0.1355, -0.1061], + [-0.1093, 0.0692, -0.1103, ..., -0.1325, -0.1315, -0.1800], + [-0.0821, -0.1377, -0.1288, ..., -0.1826, -0.1216, 0.1451], + ..., + [ 0.0837, -0.0554, -0.1153, ..., 0.0958, -0.1095, -0.1206], + [ 0.1182, 0.0993, 0.0732, ..., -0.1826, -0.1047, 0.0797], + [ 0.0592, 0.0281, 0.0891, ..., 0.0465, -0.0851, 0.0008]], + device='cuda:0'), grad: tensor([[ 1.0675e-07, -5.0873e-08, 1.3795e-08, ..., 9.4704e-08, + 1.1001e-08, 4.6566e-08], + [ 7.0501e-07, 6.7172e-08, 2.4564e-08, ..., 4.5775e-07, + 2.2526e-08, 8.3877e-08], + [ 5.4250e-08, 1.8056e-07, 6.1060e-08, ..., -2.1607e-07, + 1.1234e-08, -6.1060e-08], + ..., + [-8.3633e-07, 7.2352e-08, 2.8696e-08, ..., -3.8510e-07, + 1.4785e-08, 1.8510e-07], + [-2.2806e-07, -5.1688e-07, -1.9628e-07, ..., 3.7369e-08, + -9.3132e-10, -5.3737e-07], + [ 3.7893e-08, 1.1822e-07, -1.8626e-09, ..., 3.9488e-07, + 2.8731e-07, 4.0932e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0303, -0.0264, -0.0080, 0.0156, 0.0013, -0.0142, 0.0052, 0.0158, + -0.0272, 0.0070], device='cuda:0'), grad: tensor([-4.1304e-07, 2.5649e-06, -1.0496e-06, 3.6135e-07, -9.6858e-07, + 2.0885e-07, 4.2003e-07, -1.2759e-06, -1.4510e-06, 1.6056e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 217.33, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4629 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.11 lr 0.00010000 +Epoch 239, weight, value: tensor([[ 0.0157, -0.0818, 0.0034, ..., -0.1031, -0.1357, -0.1063], + [-0.1096, 0.0692, -0.1105, ..., -0.1326, -0.1318, -0.1811], + [-0.0829, -0.1378, -0.1295, ..., -0.1859, -0.1220, 0.1459], + ..., + [ 0.0838, -0.0554, -0.1153, ..., 0.0959, -0.1096, -0.1204], + [ 0.1182, 0.0994, 0.0734, ..., -0.1842, -0.1048, 0.0798], + [ 0.0592, 0.0281, 0.0893, ..., 0.0466, -0.0852, 0.0008]], + device='cuda:0'), grad: tensor([[-1.9500e-08, 1.5658e-08, 2.8813e-08, ..., 7.1304e-08, + 8.7079e-07, 7.2923e-07], + [ 7.5670e-09, -1.6275e-07, 5.4715e-09, ..., 5.3318e-08, + 5.3202e-08, 5.2329e-08], + [ 5.2387e-09, 3.9523e-08, 2.4447e-09, ..., -8.7311e-09, + 2.8929e-08, -1.4226e-07], + ..., + [-9.3714e-09, 5.0641e-08, 9.4296e-09, ..., -4.5984e-09, + 7.2177e-09, 1.8161e-08], + [-9.8953e-10, 6.0536e-08, 7.4506e-09, ..., 2.3865e-08, + 4.7870e-07, 4.0815e-07], + [-2.0314e-08, -8.5449e-08, -7.6718e-08, ..., 8.1165e-07, + 6.6916e-07, 2.8685e-07]], device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0298, -0.0266, -0.0080, 0.0156, 0.0012, -0.0143, 0.0060, 0.0159, + -0.0272, 0.0070], device='cuda:0'), grad: tensor([ 2.0023e-06, -2.6682e-07, -4.9407e-07, 1.0151e-06, -1.0515e-06, + -1.0971e-06, -3.3155e-06, 2.5099e-07, 1.4845e-06, 1.4817e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 217.37, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4684 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.11 lr 0.00010000 +Epoch 240, weight, value: tensor([[ 0.0158, -0.0822, 0.0027, ..., -0.1035, -0.1358, -0.1066], + [-0.1098, 0.0675, -0.1108, ..., -0.1352, -0.1350, -0.1843], + [-0.0832, -0.1380, -0.1298, ..., -0.1869, -0.1233, 0.1462], + ..., + [ 0.0836, -0.0555, -0.1154, ..., 0.0957, -0.1096, -0.1208], + [ 0.1183, 0.0995, 0.0736, ..., -0.1859, -0.1049, 0.0800], + [ 0.0594, 0.0281, 0.0895, ..., 0.0467, -0.0853, 0.0006]], + device='cuda:0'), grad: tensor([[-3.5077e-05, -3.0234e-05, -3.6329e-05, ..., 1.5134e-09, + 1.5018e-08, -8.2627e-06], + [ 2.1013e-08, -2.5122e-07, 2.0780e-08, ..., 4.2492e-09, + 3.3760e-09, 3.4284e-08], + [ 3.0268e-08, 5.8208e-08, 2.9278e-08, ..., 4.1910e-09, + 1.6880e-08, -3.9069e-07], + ..., + [ 8.3412e-08, 1.2899e-07, 9.0164e-08, ..., -3.4343e-09, + 5.2387e-10, 2.9663e-07], + [ 1.1567e-06, 1.1260e-06, 1.1967e-06, ..., 1.3970e-09, + 1.9965e-08, 3.7719e-07], + [ 3.3349e-05, 2.8759e-05, 3.4541e-05, ..., 1.2689e-08, + 1.9849e-08, 7.8976e-06]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0300, -0.0285, -0.0080, 0.0154, 0.0038, -0.0143, 0.0060, 0.0158, + -0.0272, 0.0071], device='cuda:0'), grad: tensor([-1.5140e-04, -7.6694e-07, -1.9763e-06, 1.5879e-06, 3.5600e-07, + 3.4645e-07, -1.7835e-07, 2.1365e-06, 5.9679e-06, 1.4424e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 217.61, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.5028 re_mapping 0.0032 re_causal 0.0106 /// teacc 99.08 lr 0.00010000 +Epoch 241, weight, value: tensor([[ 0.0163, -0.0814, 0.0046, ..., -0.1037, -0.1360, -0.1068], + [-0.1100, 0.0675, -0.1109, ..., -0.1354, -0.1350, -0.1844], + [-0.0832, -0.1380, -0.1302, ..., -0.1880, -0.1243, 0.1472], + ..., + [ 0.0837, -0.0555, -0.1155, ..., 0.0959, -0.1095, -0.1211], + [ 0.1183, 0.0995, 0.0736, ..., -0.1876, -0.1049, 0.0800], + [ 0.0594, 0.0281, 0.0894, ..., 0.0467, -0.0853, 0.0005]], + device='cuda:0'), grad: tensor([[-2.4633e-07, 2.0838e-07, -2.5332e-06, ..., 1.7346e-07, + 7.5088e-09, 7.1479e-08], + [ 9.5635e-08, 1.3625e-06, 4.8568e-07, ..., 7.1386e-07, + 4.2492e-09, 6.3749e-07], + [ 6.4669e-08, 1.8943e-06, 8.1444e-07, ..., 1.5344e-07, + 2.5029e-09, 9.7603e-07], + ..., + [-5.7800e-08, 1.0915e-06, 1.4971e-07, ..., 1.2759e-06, + 2.6776e-09, 8.2597e-08], + [-2.2049e-07, -4.5262e-06, -1.9046e-06, ..., 7.8115e-08, + 4.3074e-09, -2.7549e-06], + [-1.1653e-07, -4.9807e-06, 1.3448e-06, ..., -8.7917e-06, + 4.4238e-09, 2.0897e-08]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0301, -0.0286, -0.0078, 0.0152, 0.0037, -0.0143, 0.0061, 0.0158, + -0.0272, 0.0070], device='cuda:0'), grad: tensor([-1.4313e-05, 5.9754e-06, 3.3509e-06, 2.5723e-06, 1.2986e-05, + 3.1441e-06, 3.2037e-06, 4.3660e-06, -1.3508e-05, -7.7784e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 217.54, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4829 re_mapping 0.0033 re_causal 0.0105 /// teacc 99.06 lr 0.00010000 +Epoch 242, weight, value: tensor([[ 0.0164, -0.0818, 0.0048, ..., -0.1039, -0.1361, -0.1070], + [-0.1100, 0.0677, -0.1112, ..., -0.1355, -0.1351, -0.1845], + [-0.0826, -0.1382, -0.1309, ..., -0.1879, -0.1250, 0.1478], + ..., + [ 0.0836, -0.0555, -0.1156, ..., 0.0959, -0.1105, -0.1217], + [ 0.1184, 0.0995, 0.0737, ..., -0.1904, -0.1051, 0.0801], + [ 0.0594, 0.0286, 0.0910, ..., 0.0475, -0.0845, 0.0015]], + device='cuda:0'), grad: tensor([[-1.4086e-08, 1.2876e-07, 9.9419e-08, ..., 2.2491e-07, + 2.0326e-07, 1.4785e-07], + [ 2.7032e-07, 7.7765e-08, 7.3458e-08, ..., 5.2247e-07, + 3.7777e-08, 2.8696e-08], + [ 2.2759e-08, 2.5029e-08, 1.1409e-08, ..., 5.6229e-08, + 2.2061e-08, 2.0664e-08], + ..., + [-9.3074e-08, 5.5321e-07, 2.6333e-07, ..., 3.4622e-07, + 1.3388e-09, 2.7358e-09], + [ 2.8987e-08, 2.3120e-07, 1.5565e-07, ..., 3.8161e-07, + 1.1560e-07, 6.9966e-08], + [-3.9185e-07, -5.6364e-06, -4.9323e-06, ..., -9.8795e-06, + 1.7055e-08, 1.3621e-08]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0303, -0.0286, -0.0071, 0.0151, 0.0029, -0.0142, 0.0057, 0.0157, + -0.0273, 0.0076], device='cuda:0'), grad: tensor([ 7.9209e-07, 1.6727e-06, 3.0268e-07, 6.3097e-07, 2.9784e-06, + 2.8476e-05, -4.4852e-06, 8.9314e-07, 1.5907e-06, -3.2872e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 217.60, cls_loss 0.0009 cls_loss_mapping 0.0037 cls_loss_causal 0.5023 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.09 lr 0.00010000 +Epoch 243, weight, value: tensor([[ 0.0167, -0.0787, 0.0083, ..., -0.1042, -0.1365, -0.1076], + [-0.1107, 0.0656, -0.1143, ..., -0.1356, -0.1352, -0.1845], + [-0.0827, -0.1383, -0.1317, ..., -0.1886, -0.1252, 0.1481], + ..., + [ 0.0837, -0.0556, -0.1156, ..., 0.0960, -0.1109, -0.1222], + [ 0.1184, 0.0994, 0.0736, ..., -0.1937, -0.1058, 0.0802], + [ 0.0594, 0.0287, 0.0911, ..., 0.0475, -0.0845, 0.0015]], + device='cuda:0'), grad: tensor([[ 3.5914e-08, 3.5507e-08, 1.2165e-08, ..., 5.4657e-08, + 8.3644e-08, 9.7556e-08], + [ 3.8592e-08, -3.4168e-08, 1.4959e-08, ..., 9.5984e-08, + 8.6729e-09, 1.4552e-08], + [ 2.1828e-08, 2.6310e-08, 2.9104e-09, ..., 2.7148e-07, + 1.0652e-08, 1.5541e-08], + ..., + [ 2.9430e-07, 4.4494e-07, 1.5274e-07, ..., 4.5751e-07, + 3.2480e-08, 1.3155e-08], + [-9.7265e-08, -4.0163e-09, -3.1898e-08, ..., 2.3108e-08, + 8.6497e-08, -2.0955e-09], + [-4.9220e-07, -7.6834e-07, -2.8382e-07, ..., -1.3020e-06, + -5.1223e-09, -2.9046e-08]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0336, -0.0306, -0.0071, 0.0152, 0.0029, -0.0143, 0.0062, 0.0158, + -0.0274, 0.0075], device='cuda:0'), grad: tensor([ 3.8510e-07, 1.3271e-08, 6.1374e-07, 9.7847e-08, 9.7137e-07, + 5.5181e-07, -7.8185e-07, 1.6810e-06, 2.3982e-08, -3.5409e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 218.13, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4996 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.06 lr 0.00010000 +Epoch 244, weight, value: tensor([[ 0.0168, -0.0787, 0.0084, ..., -0.1043, -0.1363, -0.1077], + [-0.1109, 0.0657, -0.1144, ..., -0.1357, -0.1353, -0.1847], + [-0.0842, -0.1387, -0.1330, ..., -0.1910, -0.1260, 0.1484], + ..., + [ 0.0838, -0.0557, -0.1158, ..., 0.0962, -0.1117, -0.1220], + [ 0.1185, 0.0995, 0.0738, ..., -0.1953, -0.1054, 0.0805], + [ 0.0595, 0.0288, 0.0913, ..., 0.0475, -0.0846, 0.0014]], + device='cuda:0'), grad: tensor([[ 3.3062e-08, 3.0326e-08, 1.6065e-08, ..., 6.3446e-09, + 4.3656e-09, 5.8557e-08], + [ 3.3469e-08, -1.8370e-07, 2.3807e-08, ..., 1.8277e-08, + -4.2492e-09, 4.1910e-08], + [ 2.4308e-07, 1.8068e-07, 1.5274e-07, ..., 2.9802e-08, + 3.9581e-09, 2.9337e-07], + ..., + [ 2.7532e-08, 1.0838e-07, 2.9569e-08, ..., -1.2375e-07, + 2.9104e-10, 3.3528e-08], + [-6.6916e-07, -2.2526e-07, -4.1397e-07, ..., 3.0850e-09, + 3.4925e-09, -8.4657e-07], + [ 4.9477e-09, -7.7998e-09, 7.5088e-09, ..., -1.8277e-08, + 3.4925e-10, 3.6496e-08]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0336, -0.0306, -0.0075, 0.0150, 0.0028, -0.0143, 0.0058, 0.0158, + -0.0274, 0.0075], device='cuda:0'), grad: tensor([ 1.2596e-07, -5.9372e-07, 1.1763e-06, -2.0128e-07, 2.6845e-07, + 8.0978e-07, 2.5379e-07, 1.2922e-07, -2.1774e-06, 2.0349e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 217.50, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4752 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.14 lr 0.00010000 +Epoch 245, weight, value: tensor([[ 0.0165, -0.0787, 0.0082, ..., -0.1046, -0.1371, -0.1090], + [-0.1111, 0.0657, -0.1145, ..., -0.1357, -0.1354, -0.1848], + [-0.0842, -0.1387, -0.1341, ..., -0.1913, -0.1267, 0.1498], + ..., + [ 0.0838, -0.0557, -0.1158, ..., 0.0962, -0.1122, -0.1226], + [ 0.1186, 0.0997, 0.0741, ..., -0.1963, -0.1056, 0.0806], + [ 0.0595, 0.0288, 0.0914, ..., 0.0475, -0.0847, 0.0013]], + device='cuda:0'), grad: tensor([[ 5.0990e-07, 3.2666e-07, 2.9476e-07, ..., 2.8918e-07, + 1.9209e-08, 4.5286e-07], + [ 3.0943e-07, 1.8871e-07, 1.1519e-07, ..., 2.2876e-08, + 2.6193e-09, 3.2829e-07], + [ 9.3016e-08, 7.4622e-08, 3.2014e-08, ..., 2.7067e-08, + 3.9581e-09, 6.7230e-08], + ..., + [ 2.8731e-07, 1.4738e-07, 2.4051e-07, ..., 3.2620e-07, + 5.8208e-10, 1.1903e-07], + [-3.1311e-06, -2.2873e-06, -1.1390e-06, ..., 1.6007e-08, + 1.4086e-08, -3.5837e-06], + [-5.3085e-07, -1.3877e-07, -5.0664e-07, ..., -9.8627e-07, + 4.4820e-09, 1.5891e-08]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0335, -0.0306, -0.0073, 0.0148, 0.0028, -0.0140, 0.0061, 0.0158, + -0.0274, 0.0075], device='cuda:0'), grad: tensor([ 2.5276e-06, 1.5181e-06, 5.2806e-07, -1.1763e-06, 8.5589e-07, + 1.1578e-05, 1.8254e-06, 1.4622e-06, -1.6704e-05, -2.4009e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 217.42, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.5248 re_mapping 0.0033 re_causal 0.0111 /// teacc 99.09 lr 0.00010000 +Epoch 246, weight, value: tensor([[ 0.0164, -0.0788, 0.0081, ..., -0.1049, -0.1373, -0.1094], + [-0.1113, 0.0658, -0.1145, ..., -0.1358, -0.1355, -0.1849], + [-0.0848, -0.1392, -0.1348, ..., -0.1932, -0.1279, 0.1493], + ..., + [ 0.0838, -0.0558, -0.1159, ..., 0.0963, -0.1105, -0.1223], + [ 0.1188, 0.0998, 0.0742, ..., -0.1977, -0.1060, 0.0808], + [ 0.0596, 0.0292, 0.0920, ..., 0.0477, -0.0845, 0.0020]], + device='cuda:0'), grad: tensor([[-6.9849e-10, 1.5716e-09, 1.6298e-09, ..., 1.1642e-09, + 1.1292e-08, 9.3714e-09], + [ 3.1956e-08, -1.7462e-10, 1.0477e-09, ..., 3.4110e-08, + 5.6461e-09, 5.0641e-09], + [ 7.6834e-09, 6.5193e-09, 1.5134e-09, ..., 1.2398e-08, + 1.5716e-09, 1.2224e-09], + ..., + [-6.7521e-08, 1.2224e-09, 4.4820e-09, ..., -7.4680e-08, + 5.8208e-10, -2.3283e-10], + [ 1.1059e-09, 9.5286e-08, 1.2037e-07, ..., 1.3970e-09, + 3.0559e-08, 8.8359e-08], + [ 1.7637e-08, 4.4820e-09, 6.9849e-10, ..., 3.6263e-08, + 1.5134e-08, 2.2701e-08]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0335, -0.0305, -0.0080, 0.0152, 0.0026, -0.0140, 0.0050, 0.0159, + -0.0274, 0.0077], device='cuda:0'), grad: tensor([ 2.3574e-08, 1.0553e-07, 5.1747e-08, 1.2689e-08, -2.5262e-08, + -1.2564e-06, -7.7009e-08, -1.8335e-07, 1.1660e-06, 1.9209e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 217.34, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4955 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.13 lr 0.00010000 +Epoch 247, weight, value: tensor([[ 0.0164, -0.0788, 0.0080, ..., -0.1051, -0.1374, -0.1096], + [-0.1122, 0.0658, -0.1145, ..., -0.1359, -0.1356, -0.1850], + [-0.0849, -0.1393, -0.1356, ..., -0.1937, -0.1283, 0.1497], + ..., + [ 0.0839, -0.0559, -0.1160, ..., 0.0963, -0.1116, -0.1232], + [ 0.1209, 0.0997, 0.0759, ..., -0.1987, -0.1034, 0.0838], + [ 0.0596, 0.0293, 0.0922, ..., 0.0477, -0.0846, 0.0019]], + device='cuda:0'), grad: tensor([[ 2.2701e-08, 6.5775e-08, 1.6356e-08, ..., 3.9989e-08, + 9.0222e-09, 9.6578e-07], + [ 1.9674e-07, -5.8208e-11, 3.8941e-08, ..., 3.2876e-07, + 5.8208e-09, 1.3039e-08], + [ 1.6019e-07, 1.8906e-07, 4.7497e-08, ..., 2.5542e-07, + 9.1968e-09, -1.1325e-06], + ..., + [-1.8273e-06, 2.9337e-07, 7.2527e-08, ..., -2.1905e-06, + 1.3562e-08, 2.0082e-08], + [ 1.0571e-07, 1.0468e-06, 3.0221e-07, ..., 1.7742e-07, + 7.6136e-08, 1.1729e-07], + [ 1.1399e-06, 1.8049e-06, 5.3365e-07, ..., 1.3476e-06, + 1.1991e-07, 3.5740e-08]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0335, -0.0307, -0.0080, 0.0152, 0.0028, -0.0157, 0.0020, 0.0160, + -0.0257, 0.0077], device='cuda:0'), grad: tensor([ 2.7046e-06, 8.1072e-07, -1.6708e-06, -1.2144e-05, 3.4226e-07, + 4.0047e-06, 5.4541e-08, -6.4708e-06, 3.3304e-06, 9.0376e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 217.61, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.5013 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 248, weight, value: tensor([[ 0.0163, -0.0788, 0.0079, ..., -0.1053, -0.1379, -0.1104], + [-0.1126, 0.0662, -0.1146, ..., -0.1358, -0.1358, -0.1852], + [-0.0851, -0.1396, -0.1370, ..., -0.1939, -0.1305, 0.1497], + ..., + [ 0.0836, -0.0564, -0.1160, ..., 0.0964, -0.1121, -0.1264], + [ 0.1220, 0.1001, 0.0759, ..., -0.2012, -0.1037, 0.0843], + [ 0.0596, 0.0294, 0.0926, ..., 0.0477, -0.0846, 0.0021]], + device='cuda:0'), grad: tensor([[ 9.1782e-07, 1.0878e-06, 1.0347e-06, ..., 1.2876e-07, + 1.9791e-08, 1.1530e-06], + [ 1.7020e-07, 1.5402e-07, 1.6869e-07, ..., 4.3830e-08, + 2.6776e-09, 4.4890e-07], + [ 1.9022e-07, 2.1595e-07, 1.9954e-07, ..., 3.0966e-08, + 6.2282e-09, -1.0487e-06], + ..., + [ 1.3020e-06, 1.4063e-06, 1.8617e-06, ..., 1.4566e-06, + 8.7311e-10, 9.7230e-07], + [-5.4426e-06, -6.3553e-06, -5.8897e-06, ..., -3.5181e-07, + 3.8533e-08, -6.6385e-06], + [ 1.7891e-06, 2.2408e-06, 1.2461e-06, ..., -2.0787e-06, + -1.0489e-07, 4.0717e-06]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0334, -0.0304, -0.0080, 0.0149, 0.0027, -0.0155, 0.0025, 0.0155, + -0.0253, 0.0077], device='cuda:0'), grad: tensor([ 3.4571e-06, 1.6373e-06, -4.2841e-06, 1.9167e-06, 6.5472e-07, + 1.6922e-06, 3.8277e-07, 7.8827e-06, -1.9640e-05, 6.2846e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 217.53, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4869 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.08 lr 0.00010000 +Epoch 249, weight, value: tensor([[ 0.0169, -0.0788, 0.0079, ..., -0.1054, -0.1385, -0.1112], + [-0.1129, 0.0662, -0.1147, ..., -0.1358, -0.1359, -0.1853], + [-0.0851, -0.1397, -0.1378, ..., -0.1939, -0.1312, 0.1504], + ..., + [ 0.0857, -0.0565, -0.1149, ..., 0.0981, -0.1128, -0.1238], + [ 0.1191, 0.1002, 0.0739, ..., -0.2051, -0.1039, 0.0823], + [ 0.0596, 0.0294, 0.0927, ..., 0.0477, -0.0847, 0.0021]], + device='cuda:0'), grad: tensor([[-4.5169e-08, 1.1560e-07, 5.1281e-08, ..., 4.5984e-09, + 1.0879e-07, 8.9407e-08], + [ 1.5658e-07, 4.4936e-08, 6.4611e-09, ..., 2.8452e-07, + 4.0163e-09, 1.1350e-07], + [ 5.0641e-09, 1.8452e-08, 4.7730e-09, ..., 4.2375e-08, + 6.1700e-09, -1.2224e-07], + ..., + [-3.6135e-07, -1.0111e-07, 1.2922e-08, ..., -7.5577e-07, + 2.1537e-09, 1.2631e-08], + [ 9.0804e-09, 4.1968e-08, 1.3621e-08, ..., 1.8219e-08, + 2.7998e-08, 2.2061e-08], + [ 1.2480e-07, 7.7358e-08, 9.1968e-09, ..., 2.2340e-07, + 4.7730e-09, 1.6298e-09]], device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0334, -0.0304, -0.0079, 0.0148, 0.0027, -0.0154, 0.0029, 0.0175, + -0.0281, 0.0077], device='cuda:0'), grad: tensor([ 4.3469e-07, 1.2210e-06, -3.4412e-07, -1.6112e-06, 1.2340e-07, + 1.3877e-06, -4.0955e-07, -1.8673e-06, 2.6170e-07, 8.1118e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 217.57, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4965 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 +Epoch 250, weight, value: tensor([[ 0.0164, -0.0789, 0.0075, ..., -0.1063, -0.1389, -0.1119], + [-0.1133, 0.0662, -0.1148, ..., -0.1359, -0.1360, -0.1855], + [-0.0852, -0.1399, -0.1389, ..., -0.1944, -0.1315, 0.1509], + ..., + [ 0.0858, -0.0567, -0.1149, ..., 0.0981, -0.1130, -0.1237], + [ 0.1190, 0.1003, 0.0739, ..., -0.2052, -0.1040, 0.0824], + [ 0.0597, 0.0295, 0.0932, ..., 0.0478, -0.0847, 0.0022]], + device='cuda:0'), grad: tensor([[-6.4028e-09, 1.5134e-09, -6.4028e-09, ..., 5.0117e-08, + 2.1886e-08, 1.9267e-08], + [ 3.0792e-08, -1.9791e-08, 9.8953e-10, ..., 1.6647e-07, + 4.7265e-08, 4.3015e-08], + [ 4.7381e-08, 9.3132e-09, 3.1432e-09, ..., 2.4284e-07, + 6.8045e-08, 6.3505e-08], + ..., + [-2.1444e-07, 8.9058e-09, 8.1491e-10, ..., -2.7823e-07, + 1.1316e-07, 1.0192e-07], + [-3.8999e-09, -1.0827e-08, -1.5134e-08, ..., 8.8476e-09, + -2.0373e-09, -1.6880e-08], + [ 9.9186e-08, 9.7207e-09, 9.7207e-09, ..., 3.6117e-06, + 1.6959e-06, 1.5264e-06]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0334, -0.0304, -0.0079, 0.0155, 0.0027, -0.0155, 0.0029, 0.0175, + -0.0282, 0.0078], device='cuda:0'), grad: tensor([ 5.8673e-08, 2.4633e-07, 5.5274e-07, 1.1770e-07, -7.0594e-06, + 2.8638e-08, 2.4354e-07, -8.3726e-07, 6.3446e-09, 6.6273e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 217.45, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4488 re_mapping 0.0032 re_causal 0.0103 /// teacc 98.99 lr 0.00010000 +Epoch 251, weight, value: tensor([[ 0.0168, -0.0789, 0.0075, ..., -0.1065, -0.1388, -0.1119], + [-0.1135, 0.0663, -0.1148, ..., -0.1360, -0.1361, -0.1855], + [-0.0853, -0.1402, -0.1395, ..., -0.1946, -0.1318, 0.1510], + ..., + [ 0.0858, -0.0567, -0.1150, ..., 0.0981, -0.1134, -0.1238], + [ 0.1190, 0.1003, 0.0740, ..., -0.2052, -0.1046, 0.0823], + [ 0.0597, 0.0296, 0.0935, ..., 0.0479, -0.0848, 0.0022]], + device='cuda:0'), grad: tensor([[-9.8953e-09, 7.0431e-09, 4.2492e-09, ..., 2.2876e-08, + 5.8627e-07, 3.4645e-07], + [ 1.1409e-08, 2.9337e-08, 1.0594e-08, ..., 3.8359e-08, + 2.3865e-07, 1.7229e-07], + [ 3.5216e-08, 1.2224e-07, 6.2282e-09, ..., 3.7777e-08, + 1.7462e-07, 2.3888e-07], + ..., + [ 1.4086e-07, 3.9930e-08, 4.4645e-08, ..., 2.7590e-07, + 5.7044e-08, 1.4482e-07], + [-6.7463e-08, -2.5844e-07, -1.1298e-07, ..., 4.3132e-08, + 4.4191e-07, -3.6438e-08], + [-2.6752e-07, -2.8056e-08, -8.7661e-08, ..., -2.2398e-07, + 1.9884e-07, 2.6368e-08]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0334, -0.0304, -0.0079, 0.0155, 0.0027, -0.0154, 0.0024, 0.0175, + -0.0282, 0.0079], device='cuda:0'), grad: tensor([ 1.1232e-06, 6.7148e-07, 5.4156e-07, 2.0526e-06, 3.7393e-07, + -8.4983e-07, -4.9323e-06, 7.9861e-07, 4.4354e-07, -2.3399e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 217.41, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4514 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.09 lr 0.00010000 +Epoch 252, weight, value: tensor([[ 0.0169, -0.0789, 0.0074, ..., -0.1070, -0.1390, -0.1124], + [-0.1137, 0.0664, -0.1149, ..., -0.1361, -0.1362, -0.1856], + [-0.0852, -0.1404, -0.1405, ..., -0.1949, -0.1323, 0.1516], + ..., + [ 0.0858, -0.0568, -0.1152, ..., 0.0981, -0.1141, -0.1239], + [ 0.1188, 0.0999, 0.0735, ..., -0.2054, -0.1048, 0.0821], + [ 0.0601, 0.0299, 0.0943, ..., 0.0482, -0.0848, 0.0031]], + device='cuda:0'), grad: tensor([[ 5.2387e-09, 1.3271e-08, 8.7311e-09, ..., 2.5437e-08, + 7.6834e-09, 1.5891e-08], + [ 5.3260e-08, -2.8696e-08, 8.1491e-09, ..., 1.1979e-07, + 2.3691e-08, 3.9639e-08], + [ 8.0676e-08, 9.4529e-08, 2.6543e-08, ..., 8.9756e-08, + 1.5541e-08, 8.5798e-08], + ..., + [-1.9127e-07, 5.5996e-08, 2.1560e-07, ..., -3.5297e-07, + 4.0862e-08, 1.9604e-07], + [-1.1700e-07, -1.7462e-07, -6.1817e-08, ..., 1.1642e-08, + 4.0745e-09, -1.7462e-07], + [-6.9907e-08, -1.1560e-07, -3.9442e-07, ..., -3.5623e-07, + -1.5728e-07, -4.5518e-07]], device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0334, -0.0304, -0.0079, 0.0155, 0.0026, -0.0153, 0.0020, 0.0174, + -0.0284, 0.0082], device='cuda:0'), grad: tensor([ 7.0606e-08, -3.1292e-07, 5.5460e-07, 2.7299e-08, 1.0589e-06, + 2.1793e-07, 5.4482e-08, -3.7556e-07, -5.3318e-07, -7.5344e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 217.43, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4807 re_mapping 0.0030 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 +Epoch 253, weight, value: tensor([[ 0.0146, -0.0789, 0.0061, ..., -0.1094, -0.1392, -0.1150], + [-0.1140, 0.0665, -0.1150, ..., -0.1361, -0.1362, -0.1858], + [-0.0850, -0.1405, -0.1424, ..., -0.1951, -0.1328, 0.1532], + ..., + [ 0.0857, -0.0569, -0.1153, ..., 0.0981, -0.1143, -0.1241], + [ 0.1190, 0.1003, 0.0743, ..., -0.2054, -0.1050, 0.0824], + [ 0.0605, 0.0298, 0.0950, ..., 0.0485, -0.0846, 0.0035]], + device='cuda:0'), grad: tensor([[ 3.6089e-08, 4.7497e-08, 1.6938e-08, ..., 4.2899e-08, + 6.7113e-08, 6.1118e-08], + [ 1.5751e-07, -7.7940e-08, 3.6613e-08, ..., 7.6485e-08, + 1.0571e-07, 1.2037e-07], + [ 9.9186e-07, 1.1306e-06, 3.0710e-07, ..., 1.3644e-07, + 4.6846e-07, 7.5903e-07], + ..., + [-1.9814e-07, 5.1223e-07, 1.4738e-07, ..., -6.8266e-07, + 1.9046e-07, 2.8545e-07], + [-1.9837e-06, -2.4568e-06, -6.7614e-07, ..., 4.0804e-08, + -7.0501e-07, -1.5423e-06], + [ 2.3888e-07, 1.3027e-07, -7.4680e-08, ..., 1.7858e-07, + 2.8894e-07, 1.2247e-07]], device='cuda:0') +Epoch 253, bias, value: tensor([ 0.0329, -0.0303, -0.0075, 0.0154, 0.0023, -0.0153, 0.0020, 0.0174, + -0.0284, 0.0086], device='cuda:0'), grad: tensor([ 5.0059e-07, 3.3434e-07, 4.9472e-06, 3.3695e-06, 7.8836e-07, + -1.5616e-05, 1.0967e-05, -6.9430e-07, -7.1451e-06, 2.5108e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 217.10, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4830 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.08 lr 0.00010000 +Epoch 254, weight, value: tensor([[ 0.0140, -0.0790, 0.0057, ..., -0.1101, -0.1394, -0.1157], + [-0.1142, 0.0665, -0.1150, ..., -0.1361, -0.1363, -0.1859], + [-0.0852, -0.1408, -0.1437, ..., -0.1957, -0.1333, 0.1530], + ..., + [ 0.0858, -0.0569, -0.1153, ..., 0.0981, -0.1146, -0.1241], + [ 0.1190, 0.1002, 0.0742, ..., -0.2054, -0.1050, 0.0823], + [ 0.0606, 0.0299, 0.0956, ..., 0.0486, -0.0846, 0.0038]], + device='cuda:0'), grad: tensor([[-2.4389e-08, 2.0897e-08, 1.8044e-09, ..., 3.4925e-10, + 2.7940e-09, 4.4820e-09], + [ 9.7207e-09, 3.7602e-08, 1.4727e-08, ..., 1.2224e-09, + 6.9849e-10, 1.1642e-08], + [ 6.6939e-09, 1.0384e-06, 3.8883e-08, ..., -1.1642e-09, + 2.3283e-10, 7.5554e-08], + ..., + [ 1.9674e-08, 2.8452e-07, 4.0338e-08, ..., 4.0745e-09, + 1.7462e-10, 1.3330e-08], + [-1.1176e-08, 2.4820e-07, 1.0885e-08, ..., 2.8522e-09, + 7.8580e-09, 3.6671e-09], + [ 1.6473e-08, 5.7218e-08, 1.1176e-08, ..., 1.1118e-08, + 2.9104e-09, 4.3656e-09]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0328, -0.0303, -0.0077, 0.0156, 0.0022, -0.0152, 0.0018, 0.0174, + -0.0284, 0.0087], device='cuda:0'), grad: tensor([ 3.2014e-09, 4.8138e-08, 3.7514e-06, -6.3367e-06, 3.9465e-08, + 4.1258e-07, 1.4552e-08, 9.4436e-07, 8.6846e-07, 2.5332e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 217.05, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.5038 re_mapping 0.0029 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 +Epoch 255, weight, value: tensor([[ 0.0141, -0.0790, 0.0057, ..., -0.1102, -0.1396, -0.1158], + [-0.1143, 0.0666, -0.1151, ..., -0.1361, -0.1364, -0.1868], + [-0.0855, -0.1415, -0.1449, ..., -0.1960, -0.1338, 0.1546], + ..., + [ 0.0857, -0.0570, -0.1155, ..., 0.0980, -0.1161, -0.1243], + [ 0.1190, 0.1003, 0.0743, ..., -0.2054, -0.1054, 0.0825], + [ 0.0607, 0.0299, 0.0962, ..., 0.0488, -0.0847, 0.0040]], + device='cuda:0'), grad: tensor([[ 3.0093e-08, 3.9465e-08, 1.7462e-09, ..., 5.5297e-09, + 8.5216e-08, 8.5100e-08], + [ 5.4715e-08, -3.8594e-06, 8.1491e-10, ..., 1.7753e-08, + 2.5844e-08, 6.1758e-08], + [ 2.6589e-07, 2.5332e-07, 2.3283e-10, ..., -1.6182e-08, + 1.9383e-08, 7.6310e-08], + ..., + [ 5.6403e-08, 3.6452e-06, 8.7311e-09, ..., 2.0023e-08, + 6.4611e-09, 1.1572e-07], + [-4.8662e-07, -2.6263e-07, 6.4028e-10, ..., 1.0652e-08, + 9.3190e-08, -2.7148e-07], + [ 1.2980e-08, 1.5867e-07, -2.0547e-08, ..., 1.6706e-08, + 5.3027e-08, 3.6554e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0328, -0.0304, -0.0068, 0.0156, 0.0022, -0.0152, 0.0018, 0.0171, + -0.0284, 0.0089], device='cuda:0'), grad: tensor([ 1.4994e-06, -2.8580e-05, 8.5775e-07, 5.0329e-06, 3.6391e-07, + -3.5858e-04, -3.8790e-07, 3.9309e-05, 3.3951e-04, 1.3765e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 217.32, cls_loss 0.0016 cls_loss_mapping 0.0039 cls_loss_causal 0.4978 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.16 lr 0.00010000 +Epoch 256, weight, value: tensor([[ 0.0142, -0.0790, 0.0056, ..., -0.1103, -0.1400, -0.1161], + [-0.1163, 0.0666, -0.1152, ..., -0.1363, -0.1366, -0.1870], + [-0.0882, -0.1423, -0.1467, ..., -0.2023, -0.1351, 0.1563], + ..., + [ 0.0828, -0.0569, -0.1155, ..., 0.0954, -0.1148, -0.1261], + [ 0.1221, 0.0997, 0.0741, ..., -0.2024, -0.1068, 0.0833], + [ 0.0606, 0.0301, 0.0964, ..., 0.0487, -0.0847, 0.0041]], + device='cuda:0'), grad: tensor([[-2.0373e-09, 8.7894e-09, 2.3283e-10, ..., 2.2701e-09, + 2.0470e-06, 7.4692e-07], + [-2.9220e-08, -2.6892e-07, 1.8626e-09, ..., 7.7998e-09, + 3.3877e-08, 1.9837e-07], + [ 6.7521e-09, 5.5647e-08, 5.8208e-10, ..., 1.8976e-08, + 3.5623e-08, -3.3225e-07], + ..., + [ 1.3853e-08, 3.8138e-07, 2.3865e-08, ..., -2.9686e-08, + 7.5670e-09, 9.7789e-08], + [ 9.3132e-10, 5.4832e-08, 1.1642e-09, ..., 4.6566e-10, + 1.3399e-07, 6.5775e-08], + [ 5.4715e-09, 3.9523e-08, 1.2806e-09, ..., 1.2049e-08, + 3.3586e-08, 2.6776e-08]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0327, -0.0306, -0.0079, 0.0154, 0.0019, -0.0149, 0.0025, 0.0145, + -0.0257, 0.0088], device='cuda:0'), grad: tensor([ 4.6529e-06, -4.7311e-07, -6.3609e-07, -1.2396e-06, 6.2166e-07, + 6.4028e-07, -5.9605e-06, 1.5879e-06, 5.3924e-07, 2.7427e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 217.53, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4695 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.12 lr 0.00010000 +Epoch 257, weight, value: tensor([[ 0.0143, -0.0790, 0.0061, ..., -0.1105, -0.1407, -0.1167], + [-0.1191, 0.0667, -0.1152, ..., -0.1369, -0.1367, -0.1871], + [-0.0885, -0.1425, -0.1513, ..., -0.2027, -0.1360, 0.1563], + ..., + [ 0.0828, -0.0569, -0.1157, ..., 0.0956, -0.1149, -0.1261], + [ 0.1221, 0.0996, 0.0741, ..., -0.2024, -0.1075, 0.0832], + [ 0.0607, 0.0302, 0.0966, ..., 0.0487, -0.0848, 0.0041]], + device='cuda:0'), grad: tensor([[ 3.6671e-09, 2.6019e-08, 7.4506e-09, ..., 9.8953e-10, + 1.6880e-09, 1.3446e-08], + [ 7.6834e-09, -4.2142e-07, 1.1118e-08, ..., 1.0477e-09, + 1.0477e-09, 2.3458e-08], + [ 2.0314e-08, 7.1304e-08, 2.9511e-08, ..., 1.1642e-10, + 5.8208e-10, -3.4925e-09], + ..., + [ 1.3097e-08, 3.4866e-08, 1.5076e-08, ..., 1.3271e-08, + 5.2387e-10, 2.9453e-08], + [-1.5274e-07, -2.7241e-07, -2.1583e-07, ..., 2.9104e-10, + 2.4913e-08, -2.9383e-07], + [ 6.0885e-08, 1.7369e-07, 9.6275e-08, ..., -2.0431e-08, + 5.1223e-09, 1.6019e-07]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0329, -0.0308, -0.0081, 0.0154, 0.0017, -0.0147, 0.0032, 0.0145, + -0.0258, 0.0088], device='cuda:0'), grad: tensor([ 2.4447e-07, -4.2655e-06, -5.0152e-07, 4.1304e-07, 2.7940e-06, + 2.2445e-07, 3.4692e-07, 6.1048e-07, -6.9989e-07, 8.1956e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 217.45, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.5103 re_mapping 0.0030 re_causal 0.0099 /// teacc 99.07 lr 0.00010000 +Epoch 258, weight, value: tensor([[ 0.0143, -0.0791, 0.0060, ..., -0.1107, -0.1411, -0.1172], + [-0.1193, 0.0668, -0.1154, ..., -0.1369, -0.1367, -0.1877], + [-0.0887, -0.1423, -0.1525, ..., -0.2031, -0.1367, 0.1570], + ..., + [ 0.0828, -0.0570, -0.1158, ..., 0.0956, -0.1160, -0.1262], + [ 0.1221, 0.0997, 0.0745, ..., -0.2024, -0.1077, 0.0834], + [ 0.0608, 0.0305, 0.0976, ..., 0.0491, -0.0845, 0.0042]], + device='cuda:0'), grad: tensor([[-1.4119e-06, 1.0873e-07, 1.1933e-08, ..., 1.3737e-08, + -1.3476e-06, -1.2228e-06], + [ 2.8231e-08, -1.1092e-06, 2.1188e-08, ..., 3.5274e-08, + -3.3109e-07, -4.1677e-07], + [ 5.6811e-08, 6.5716e-08, 8.6147e-09, ..., 1.0477e-08, + 2.9046e-08, 8.1083e-08], + ..., + [ 5.7044e-09, 5.3085e-08, 2.3923e-08, ..., 2.0606e-08, + 2.1479e-08, 2.4214e-08], + [-2.3283e-08, 8.5449e-07, 6.1118e-09, ..., 1.5949e-08, + 3.1362e-07, 2.9476e-07], + [-1.9034e-08, -1.8347e-07, -2.7730e-07, ..., 5.1456e-07, + 5.5553e-07, 7.8348e-08]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0329, -0.0309, -0.0074, 0.0151, 0.0011, -0.0148, 0.0025, 0.0145, + -0.0258, 0.0092], device='cuda:0'), grad: tensor([-8.5309e-06, -4.9472e-06, 4.8522e-07, -2.7381e-07, -6.3889e-07, + 6.2678e-07, 7.9647e-06, 2.6077e-07, 3.9935e-06, 1.0813e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 217.46, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4612 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.02 lr 0.00010000 +Epoch 259, weight, value: tensor([[ 0.0147, -0.0791, 0.0059, ..., -0.1108, -0.1402, -0.1166], + [-0.1196, 0.0669, -0.1155, ..., -0.1370, -0.1368, -0.1878], + [-0.0889, -0.1426, -0.1533, ..., -0.2034, -0.1374, 0.1570], + ..., + [ 0.0828, -0.0571, -0.1159, ..., 0.0956, -0.1168, -0.1262], + [ 0.1221, 0.0998, 0.0747, ..., -0.2024, -0.1084, 0.0834], + [ 0.0609, 0.0306, 0.0979, ..., 0.0493, -0.0844, 0.0043]], + device='cuda:0'), grad: tensor([[-4.5402e-09, 1.6298e-08, -6.4028e-10, ..., 4.3656e-09, + -1.1642e-10, 3.8417e-09], + [ 1.5157e-07, -1.1595e-07, 2.9104e-10, ..., 4.9884e-08, + 4.0745e-10, 1.3306e-07], + [ 7.8557e-07, 4.9477e-09, 4.0745e-10, ..., 6.0070e-08, + 1.1642e-10, -8.5402e-07], + ..., + [-1.7174e-06, 3.2305e-08, 2.3283e-10, ..., -1.7055e-07, + 3.4925e-10, 6.7288e-07], + [ 5.2992e-07, 5.1572e-08, -1.6880e-09, ..., 1.5250e-08, + 5.8208e-11, -2.2119e-09], + [ 3.5623e-08, 3.4925e-09, 3.4925e-10, ..., 1.7229e-08, + 1.0477e-09, 1.2864e-08]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0331, -0.0308, -0.0076, 0.0152, 0.0011, -0.0171, 0.0045, 0.0145, + -0.0258, 0.0093], device='cuda:0'), grad: tensor([ 1.6869e-07, 1.2219e-06, 4.3064e-06, 1.4845e-06, 1.8743e-07, + 8.2655e-09, 1.2666e-07, -1.3299e-05, 5.4725e-06, 3.1595e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 217.49, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.5077 re_mapping 0.0031 re_causal 0.0104 /// teacc 99.11 lr 0.00010000 +Epoch 260, weight, value: tensor([[ 0.0150, -0.0791, 0.0059, ..., -0.1110, -0.1402, -0.1165], + [-0.1198, 0.0671, -0.1156, ..., -0.1371, -0.1369, -0.1880], + [-0.0890, -0.1421, -0.1538, ..., -0.2036, -0.1373, 0.1570], + ..., + [ 0.0829, -0.0572, -0.1160, ..., 0.0956, -0.1193, -0.1263], + [ 0.1221, 0.0994, 0.0747, ..., -0.2024, -0.1088, 0.0835], + [ 0.0610, 0.0307, 0.0984, ..., 0.0494, -0.0843, 0.0047]], + device='cuda:0'), grad: tensor([[-3.0245e-07, 7.4040e-08, 3.1607e-08, ..., 1.2980e-08, + -5.6170e-08, 1.1583e-07], + [ 1.7392e-07, 9.8487e-08, 5.1281e-08, ..., 8.1374e-08, + 2.2294e-08, 1.3364e-07], + [ 1.8673e-07, 1.1129e-07, 5.4250e-08, ..., 1.1589e-07, + 1.8394e-08, 9.4296e-08], + ..., + [ 2.4145e-07, 2.3155e-07, 1.1834e-07, ..., -1.2829e-07, + -7.1595e-09, 1.5064e-07], + [-1.3337e-06, -1.5190e-06, -6.4308e-07, ..., 1.0186e-08, + -6.0396e-07, -2.0638e-06], + [ 1.2061e-07, -8.6729e-09, 2.2992e-08, ..., -2.0489e-07, + 6.0943e-08, 1.5018e-07]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0332, -0.0308, -0.0062, 0.0141, 0.0011, -0.0179, 0.0053, 0.0145, + -0.0258, 0.0094], device='cuda:0'), grad: tensor([-2.8927e-06, 7.8930e-07, 5.7602e-07, 1.0896e-06, 2.5588e-07, + 8.8988e-07, 3.1777e-06, 5.9884e-07, -5.4613e-06, 9.5181e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 217.71, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4767 re_mapping 0.0030 re_causal 0.0099 /// teacc 99.11 lr 0.00010000 +Epoch 261, weight, value: tensor([[ 0.0153, -0.0791, 0.0060, ..., -0.1112, -0.1410, -0.1172], + [-0.1200, 0.0672, -0.1157, ..., -0.1371, -0.1372, -0.1882], + [-0.0892, -0.1426, -0.1543, ..., -0.2038, -0.1379, 0.1577], + ..., + [ 0.0829, -0.0572, -0.1161, ..., 0.0956, -0.1213, -0.1265], + [ 0.1221, 0.0996, 0.0751, ..., -0.2024, -0.1090, 0.0836], + [ 0.0610, 0.0307, 0.0985, ..., 0.0492, -0.0848, 0.0041]], + device='cuda:0'), grad: tensor([[ 1.0477e-09, 2.1537e-09, 7.5670e-10, ..., 6.4611e-09, + 6.0536e-09, 9.0804e-09], + [ 3.2946e-08, -5.7975e-08, 8.7311e-10, ..., 9.7789e-09, + 2.7940e-09, 3.4168e-08], + [-2.1642e-07, 6.6939e-09, 7.5670e-10, ..., 2.0373e-09, + 8.1491e-10, -2.3167e-07], + ..., + [ 1.9616e-07, 7.2119e-08, 1.2515e-08, ..., 7.5088e-08, + 1.9209e-09, 1.7486e-07], + [ 4.1910e-09, 5.8208e-10, -1.5716e-09, ..., 3.3178e-09, + 8.1491e-09, 8.3819e-09], + [-4.5053e-08, -3.7777e-08, -2.1071e-08, ..., -5.9837e-08, + 1.8685e-08, 2.2235e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([ 0.0332, -0.0308, -0.0056, 0.0139, 0.0013, -0.0179, 0.0053, 0.0145, + -0.0258, 0.0092], device='cuda:0'), grad: tensor([ 6.6066e-08, 3.9465e-08, -1.7695e-06, 6.1118e-08, 2.4855e-08, + -1.8300e-07, -2.6193e-09, 1.6550e-06, 9.4820e-08, 2.4680e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 217.58, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.4826 re_mapping 0.0029 re_causal 0.0098 /// teacc 99.11 lr 0.00010000 +Epoch 262, weight, value: tensor([[ 0.0160, -0.0791, 0.0061, ..., -0.1114, -0.1408, -0.1170], + [-0.1221, 0.0671, -0.1158, ..., -0.1373, -0.1377, -0.1887], + [-0.0895, -0.1427, -0.1550, ..., -0.2041, -0.1390, 0.1583], + ..., + [ 0.0829, -0.0572, -0.1164, ..., 0.0954, -0.1244, -0.1272], + [ 0.1221, 0.0997, 0.0751, ..., -0.2024, -0.1100, 0.0835], + [ 0.0609, 0.0309, 0.0991, ..., 0.0493, -0.0859, 0.0029]], + device='cuda:0'), grad: tensor([[ 1.3213e-08, 9.0222e-09, 3.4343e-09, ..., 4.8894e-09, + 1.4133e-07, 1.4016e-07], + [ 1.4843e-08, -6.5833e-08, 6.5775e-09, ..., 1.5774e-08, + 4.7148e-09, 2.0314e-08], + [ 5.1805e-08, 1.1874e-07, 5.2212e-08, ..., 1.7462e-08, + 4.9477e-09, 1.1857e-07], + ..., + [-4.8138e-08, 2.5670e-08, 1.7462e-09, ..., -9.0455e-08, + -4.8312e-09, -6.1118e-09], + [-7.6718e-08, -1.8138e-07, -1.0111e-07, ..., 1.0827e-08, + 5.6403e-08, -1.9232e-07], + [ 2.1595e-08, 5.2853e-08, 1.3970e-08, ..., -7.5670e-10, + 1.0477e-09, 6.5716e-08]], device='cuda:0') +Epoch 262, bias, value: tensor([ 0.0334, -0.0314, -0.0054, 0.0116, 0.0019, -0.0177, 0.0053, 0.0145, + -0.0258, 0.0090], device='cuda:0'), grad: tensor([ 3.2946e-07, -4.1118e-07, 4.4657e-07, 4.4645e-08, 1.3877e-07, + 6.6415e-08, -4.4028e-07, 1.5891e-08, -4.1025e-07, 2.2887e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 217.53, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4927 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.03 lr 0.00010000 +Epoch 263, weight, value: tensor([[ 0.0157, -0.0793, 0.0055, ..., -0.1118, -0.1419, -0.1179], + [-0.1222, 0.0671, -0.1159, ..., -0.1373, -0.1379, -0.1886], + [-0.0895, -0.1433, -0.1562, ..., -0.2045, -0.1398, 0.1584], + ..., + [ 0.0829, -0.0573, -0.1166, ..., 0.0954, -0.1253, -0.1274], + [ 0.1221, 0.0998, 0.0753, ..., -0.2025, -0.1105, 0.0835], + [ 0.0611, 0.0310, 0.0999, ..., 0.0492, -0.0862, 0.0025]], + device='cuda:0'), grad: tensor([[ 2.9826e-07, 4.7451e-07, 5.2294e-07, ..., 5.1409e-07, + 1.5949e-08, 2.4331e-08], + [ 5.4482e-08, 1.8801e-08, 5.3493e-08, ..., 7.9395e-08, + 3.3178e-09, 1.7346e-08], + [-7.5495e-08, 2.2759e-08, 1.7171e-08, ..., 2.3167e-08, + 3.7253e-09, -2.6263e-07], + ..., + [ 1.3988e-06, 2.2706e-06, 2.4885e-06, ..., 2.3860e-06, + 5.8208e-10, 6.6764e-08], + [ 6.4378e-08, 7.1421e-08, 7.6368e-08, ..., 9.2667e-08, + 5.6403e-08, 1.1391e-07], + [-4.0084e-06, -6.4336e-06, -7.1079e-06, ..., -6.9700e-06, + 3.3760e-09, 4.3306e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0332, -0.0313, -0.0055, 0.0116, 0.0021, -0.0177, 0.0053, 0.0145, + -0.0258, 0.0090], device='cuda:0'), grad: tensor([ 6.1579e-06, 3.5623e-07, -1.0087e-07, 1.5631e-05, 3.0873e-07, + -5.8711e-05, 3.4412e-07, 1.5222e-05, 4.4346e-05, -2.3484e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 217.40, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.5273 re_mapping 0.0031 re_causal 0.0105 /// teacc 99.00 lr 0.00010000 +Epoch 264, weight, value: tensor([[ 0.0156, -0.0794, 0.0052, ..., -0.1121, -0.1427, -0.1186], + [-0.1224, 0.0672, -0.1160, ..., -0.1374, -0.1381, -0.1887], + [-0.0890, -0.1439, -0.1568, ..., -0.2050, -0.1408, 0.1585], + ..., + [ 0.0829, -0.0573, -0.1168, ..., 0.0955, -0.1258, -0.1275], + [ 0.1221, 0.0999, 0.0756, ..., -0.2025, -0.1106, 0.0833], + [ 0.0611, 0.0311, 0.1004, ..., 0.0491, -0.0867, 0.0020]], + device='cuda:0'), grad: tensor([[ 3.1898e-08, 2.5844e-08, 1.0536e-08, ..., 2.0955e-08, + 3.1199e-08, 4.2317e-08], + [ 1.2564e-06, 1.3330e-08, 2.1770e-08, ..., 1.3476e-06, + 3.8010e-08, 9.5519e-08], + [ 1.8091e-07, 1.1956e-07, 7.1246e-08, ..., 1.3551e-07, + 4.8429e-08, 1.6787e-07], + ..., + [-1.8589e-06, 4.3074e-08, 2.3807e-08, ..., -2.3581e-06, + 3.7020e-08, 4.3714e-08], + [-5.7882e-07, -4.2072e-07, -2.5169e-07, ..., 1.8219e-08, + -1.2224e-09, -6.3656e-07], + [ 6.4308e-07, 2.0955e-08, -8.2073e-09, ..., 8.8941e-07, + 1.0402e-07, 8.2247e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([ 0.0331, -0.0313, -0.0054, 0.0117, 0.0022, -0.0173, 0.0050, 0.0145, + -0.0259, 0.0090], device='cuda:0'), grad: tensor([ 2.6589e-07, 4.1053e-06, 1.0123e-06, 4.9546e-07, -2.0314e-07, + -2.4345e-06, 2.0694e-06, -6.2659e-06, -1.8319e-06, 2.7902e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 217.57, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4541 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.06 lr 0.00010000 +Epoch 265, weight, value: tensor([[ 0.0155, -0.0794, 0.0050, ..., -0.1123, -0.1435, -0.1194], + [-0.1226, 0.0673, -0.1161, ..., -0.1375, -0.1382, -0.1888], + [-0.0893, -0.1444, -0.1577, ..., -0.2055, -0.1411, 0.1585], + ..., + [ 0.0829, -0.0574, -0.1169, ..., 0.0956, -0.1261, -0.1275], + [ 0.1221, 0.1001, 0.0758, ..., -0.2025, -0.1110, 0.0834], + [ 0.0612, 0.0311, 0.1007, ..., 0.0489, -0.0877, 0.0013]], + device='cuda:0'), grad: tensor([[-1.8277e-08, 1.4319e-08, 4.3074e-09, ..., 4.3074e-09, + 9.0804e-09, 7.5088e-09], + [ 4.9477e-09, 2.9744e-08, 8.4983e-09, ..., 1.3039e-08, + 6.5076e-08, 4.9942e-08], + [ 2.4098e-08, 1.3912e-08, 2.7940e-09, ..., 1.0361e-08, + 2.1537e-09, 2.2817e-08], + ..., + [-3.0850e-09, 7.0664e-08, 1.4435e-08, ..., -1.1583e-08, + 1.1642e-10, 2.3283e-09], + [-2.1013e-08, 5.6461e-08, 1.0303e-08, ..., 9.6043e-09, + 4.0745e-09, -3.4925e-08], + [-3.4343e-09, 1.0937e-07, -9.8953e-09, ..., -1.1409e-08, + 4.6566e-10, 1.9209e-09]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0330, -0.0313, -0.0056, 0.0117, 0.0024, -0.0173, 0.0050, 0.0145, + -0.0259, 0.0089], device='cuda:0'), grad: tensor([-4.9942e-08, 2.1036e-07, 8.7370e-08, -1.2353e-05, 1.0344e-07, + 1.1437e-05, -1.2678e-07, 2.0687e-07, 1.2293e-07, 3.5437e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 217.56, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4823 re_mapping 0.0030 re_causal 0.0099 /// teacc 99.05 lr 0.00010000 +Epoch 266, weight, value: tensor([[ 0.0159, -0.0794, 0.0050, ..., -0.1124, -0.1431, -0.1192], + [-0.1227, 0.0677, -0.1162, ..., -0.1375, -0.1384, -0.1890], + [-0.0896, -0.1450, -0.1587, ..., -0.2058, -0.1415, 0.1585], + ..., + [ 0.0829, -0.0578, -0.1170, ..., 0.0956, -0.1262, -0.1276], + [ 0.1221, 0.1004, 0.0761, ..., -0.2025, -0.1111, 0.0836], + [ 0.0612, 0.0312, 0.1010, ..., 0.0490, -0.0878, 0.0013]], + device='cuda:0'), grad: tensor([[-9.4995e-08, 9.3714e-09, -9.1386e-09, ..., 1.1176e-08, + 1.4377e-08, 2.7532e-08], + [ 6.7288e-08, 1.1933e-07, 7.2119e-08, ..., 1.2584e-07, + 8.6147e-09, 3.5157e-08], + [ 9.5519e-08, 1.7066e-07, 4.2492e-09, ..., 7.1130e-08, + 2.1537e-09, 3.5483e-07], + ..., + [-2.4168e-07, 6.3505e-08, 2.4214e-08, ..., -5.1595e-07, + 2.7940e-09, 3.6904e-08], + [-6.2631e-08, -2.7427e-07, 8.0327e-09, ..., 1.5541e-08, + 2.0256e-08, -6.2212e-07], + [ 1.4494e-07, 7.5670e-10, -1.2224e-08, ..., 2.4750e-07, + 1.9732e-08, 3.1840e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0332, -0.0310, -0.0059, 0.0120, 0.0023, -0.0173, 0.0050, 0.0144, + -0.0259, 0.0089], device='cuda:0'), grad: tensor([-4.2748e-07, 7.5344e-07, 2.5760e-06, 1.2806e-07, 7.8988e-08, + 1.2154e-06, -5.9046e-07, -6.3563e-07, -3.9600e-06, 8.6986e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 217.74, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4855 re_mapping 0.0029 re_causal 0.0100 /// teacc 99.05 lr 0.00010000 +Epoch 267, weight, value: tensor([[ 0.0160, -0.0795, 0.0049, ..., -0.1126, -0.1440, -0.1197], + [-0.1231, 0.0678, -0.1163, ..., -0.1376, -0.1385, -0.1892], + [-0.0900, -0.1453, -0.1595, ..., -0.2070, -0.1417, 0.1587], + ..., + [ 0.0829, -0.0580, -0.1171, ..., 0.0956, -0.1264, -0.1276], + [ 0.1221, 0.0997, 0.0754, ..., -0.2025, -0.1113, 0.0836], + [ 0.0613, 0.0314, 0.1012, ..., 0.0491, -0.0878, 0.0013]], + device='cuda:0'), grad: tensor([[-1.0594e-07, 6.4611e-09, -2.3749e-08, ..., 3.3760e-09, + 3.5507e-09, 1.0477e-08], + [ 7.6252e-09, 3.1549e-08, 4.9302e-08, ..., 5.3551e-09, + 2.9686e-09, 4.0513e-08], + [ 3.1723e-08, 5.1048e-08, 3.0443e-08, ..., 3.2596e-09, + 2.2701e-09, 1.1933e-08], + ..., + [ 3.0268e-09, 1.8568e-08, 6.9267e-09, ..., 1.2864e-08, + 1.2573e-08, 2.0606e-08], + [-9.6101e-08, -2.9989e-07, -2.3772e-07, ..., 2.7940e-09, + -1.9209e-09, -2.8824e-07], + [ 5.8499e-08, 1.8510e-08, 2.5379e-08, ..., 2.0722e-08, + 2.1071e-08, 2.9802e-08]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0331, -0.0310, -0.0059, 0.0125, 0.0021, -0.0172, 0.0050, 0.0144, + -0.0259, 0.0090], device='cuda:0'), grad: tensor([-3.4901e-07, 3.7020e-07, 6.7172e-08, -3.9290e-08, -4.9302e-08, + 3.6042e-07, 2.8219e-07, 1.2084e-07, -1.1269e-06, 3.6927e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 217.84, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.5195 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +Epoch 268, weight, value: tensor([[ 0.0162, -0.0795, 0.0051, ..., -0.1129, -0.1447, -0.1202], + [-0.1236, 0.0680, -0.1164, ..., -0.1377, -0.1388, -0.1893], + [-0.0902, -0.1459, -0.1600, ..., -0.2079, -0.1421, 0.1588], + ..., + [ 0.0829, -0.0581, -0.1172, ..., 0.0957, -0.1266, -0.1276], + [ 0.1221, 0.0998, 0.0755, ..., -0.2025, -0.1120, 0.0837], + [ 0.0613, 0.0315, 0.1014, ..., 0.0491, -0.0880, 0.0013]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 5.1688e-08, 5.6229e-08, ..., 1.0710e-08, + 2.6193e-09, 8.9116e-08], + [ 8.9058e-09, -1.5716e-07, 2.2992e-08, ..., 1.0652e-08, + 6.9849e-10, 3.7253e-09], + [ 1.3504e-08, 1.3039e-07, 9.4413e-08, ..., 8.2655e-09, + 1.7462e-10, 1.0943e-08], + ..., + [ 6.8685e-09, 2.8266e-07, 1.4366e-07, ..., 2.6193e-08, + 2.9104e-10, 1.0768e-08], + [-9.0688e-08, 3.4575e-08, 1.5716e-09, ..., 8.9058e-09, + 8.0909e-09, -1.7416e-07], + [-3.1549e-08, -2.2585e-08, -1.7462e-10, ..., -1.1240e-07, + 5.5297e-09, 3.9348e-08]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0332, -0.0310, -0.0062, 0.0124, 0.0019, -0.0172, 0.0050, 0.0145, + -0.0259, 0.0090], device='cuda:0'), grad: tensor([ 1.7439e-07, -3.4412e-07, 5.4762e-07, -2.2370e-06, 1.5134e-07, + 8.3447e-07, -7.5786e-08, 9.5554e-07, -2.4156e-08, 3.5274e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 217.47, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4897 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.15 lr 0.00010000 +Epoch 269, weight, value: tensor([[ 0.0164, -0.0795, 0.0050, ..., -0.1132, -0.1451, -0.1209], + [-0.1241, 0.0681, -0.1166, ..., -0.1378, -0.1389, -0.1871], + [-0.0906, -0.1463, -0.1614, ..., -0.2085, -0.1425, 0.1591], + ..., + [ 0.0830, -0.0582, -0.1174, ..., 0.0958, -0.1268, -0.1276], + [ 0.1221, 0.1000, 0.0760, ..., -0.2025, -0.1120, 0.0831], + [ 0.0613, 0.0315, 0.1014, ..., 0.0492, -0.0882, 0.0011]], + device='cuda:0'), grad: tensor([[-4.4121e-08, 7.0431e-09, 2.9686e-09, ..., 1.4959e-08, + 3.0675e-08, 2.9395e-08], + [ 3.3760e-08, -3.6671e-09, 6.7521e-09, ..., 6.1700e-08, + 1.8743e-08, 2.2934e-08], + [ 3.1607e-08, 1.3213e-08, 6.2282e-09, ..., 3.9930e-08, + 1.3039e-08, 1.2806e-08], + ..., + [-7.3342e-08, 3.2887e-08, 1.4086e-08, ..., -8.9407e-08, + 2.2061e-08, 3.1432e-08], + [-2.3923e-08, -6.7870e-08, -3.8766e-08, ..., 2.1770e-08, + 3.8068e-08, -2.2235e-08], + [ 1.0361e-08, 1.0233e-07, 1.5658e-08, ..., 1.2526e-06, + 7.0827e-07, 7.6788e-07]], device='cuda:0') +Epoch 269, bias, value: tensor([ 0.0331, -0.0304, -0.0062, 0.0128, 0.0018, -0.0172, 0.0049, 0.0145, + -0.0261, 0.0089], device='cuda:0'), grad: tensor([-1.1566e-07, 2.0314e-07, 1.9395e-07, 1.4692e-07, -2.8200e-06, + 2.4401e-07, -2.1292e-07, -3.5809e-07, -2.3516e-08, 2.7623e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 217.66, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4881 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.15 lr 0.00010000 +Epoch 270, weight, value: tensor([[ 0.0165, -0.0796, 0.0049, ..., -0.1133, -0.1482, -0.1237], + [-0.1244, 0.0687, -0.1165, ..., -0.1379, -0.1391, -0.1872], + [-0.0908, -0.1467, -0.1620, ..., -0.2091, -0.1435, 0.1597], + ..., + [ 0.0830, -0.0586, -0.1175, ..., 0.0958, -0.1274, -0.1277], + [ 0.1221, 0.1002, 0.0763, ..., -0.2025, -0.1122, 0.0831], + [ 0.0613, 0.0315, 0.1016, ..., 0.0491, -0.0886, 0.0008]], + device='cuda:0'), grad: tensor([[ 5.8790e-09, 3.4343e-08, 2.5611e-08, ..., 3.4459e-08, + 3.6089e-09, 4.9418e-08], + [ 5.1630e-08, 6.7346e-08, 4.2608e-08, ..., 3.8475e-08, + 1.4959e-08, 1.4389e-07], + [ 5.4017e-08, 8.3004e-08, 4.3015e-08, ..., 1.6415e-08, + 4.8894e-09, -5.7789e-07], + ..., + [ 6.6939e-08, 8.2247e-08, 6.6357e-08, ..., 7.1246e-08, + 2.4156e-08, 1.0827e-07], + [-3.3854e-07, -6.6590e-07, -3.5227e-07, ..., 1.5949e-08, + 3.7253e-09, -2.7940e-07], + [-7.0315e-08, 2.0547e-08, -3.4808e-08, ..., 3.5670e-07, + 3.1502e-07, 3.5809e-07]], device='cuda:0') +Epoch 270, bias, value: tensor([ 0.0326, -0.0301, -0.0060, 0.0126, 0.0019, -0.0172, 0.0050, 0.0145, + -0.0261, 0.0087], device='cuda:0'), grad: tensor([ 2.9104e-07, 8.9407e-07, -4.3511e-06, 1.5115e-06, -9.7789e-07, + 8.3353e-07, 4.8475e-07, 7.9442e-07, -7.5344e-07, 1.2740e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 217.60, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4729 re_mapping 0.0029 re_causal 0.0094 /// teacc 99.05 lr 0.00010000 +Epoch 271, weight, value: tensor([[ 0.0166, -0.0797, 0.0047, ..., -0.1137, -0.1484, -0.1244], + [-0.1251, 0.0690, -0.1167, ..., -0.1381, -0.1395, -0.1874], + [-0.0912, -0.1475, -0.1641, ..., -0.2097, -0.1440, 0.1601], + ..., + [ 0.0830, -0.0587, -0.1176, ..., 0.0959, -0.1276, -0.1278], + [ 0.1221, 0.0973, 0.0737, ..., -0.2025, -0.1151, 0.0802], + [ 0.0612, 0.0316, 0.1020, ..., 0.0489, -0.0887, 0.0007]], + device='cuda:0'), grad: tensor([[ 2.6543e-08, 2.0431e-08, 1.5483e-08, ..., 1.5425e-08, + 3.8999e-09, 7.7533e-08], + [ 6.7870e-08, 3.5798e-08, 3.0443e-08, ..., 3.3469e-08, + 1.4552e-09, 4.2084e-08], + [ 7.8988e-08, 2.5146e-08, 1.0070e-08, ..., 5.9546e-08, + 5.8208e-10, -5.6403e-08], + ..., + [-1.9581e-07, 3.7719e-08, 1.7812e-08, ..., -1.8557e-07, + 0.0000e+00, 1.4319e-08], + [-6.6182e-08, -2.7358e-07, -7.3516e-08, ..., 8.5158e-08, + 3.1432e-09, -1.5809e-07], + [-1.1438e-07, -8.3935e-08, -1.2026e-07, ..., -7.0722e-08, + 9.3132e-10, -7.0722e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0325, -0.0299, -0.0062, 0.0122, 0.0019, -0.0171, 0.0065, 0.0145, + -0.0274, 0.0086], device='cuda:0'), grad: tensor([ 5.3085e-07, 2.0838e-07, -2.0955e-07, 1.4307e-07, 1.5763e-07, + 5.8999e-07, 2.7125e-08, -5.1036e-07, -7.5065e-07, -1.6834e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 217.72, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4748 re_mapping 0.0032 re_causal 0.0101 /// teacc 99.14 lr 0.00010000 +Epoch 272, weight, value: tensor([[ 0.0179, -0.0797, 0.0046, ..., -0.1139, -0.1486, -0.1249], + [-0.1255, 0.0691, -0.1168, ..., -0.1382, -0.1414, -0.1877], + [-0.0920, -0.1478, -0.1655, ..., -0.2106, -0.1449, 0.1601], + ..., + [ 0.0830, -0.0588, -0.1177, ..., 0.0960, -0.1281, -0.1278], + [ 0.1221, 0.0973, 0.0738, ..., -0.2025, -0.1152, 0.0802], + [ 0.0611, 0.0316, 0.1022, ..., 0.0487, -0.0891, 0.0003]], + device='cuda:0'), grad: tensor([[ 4.8720e-08, 9.8196e-08, 3.9756e-08, ..., 5.9314e-08, + 4.6566e-09, 3.3586e-08], + [ 5.6461e-08, 1.2212e-07, 2.9220e-08, ..., 8.0618e-08, + 1.1176e-08, 1.7637e-08], + [ 2.6077e-08, 7.1828e-08, 1.9441e-08, ..., 1.7695e-08, + 3.6089e-09, 7.9744e-09], + ..., + [-2.6077e-08, 1.1967e-07, 4.4063e-08, ..., -4.6624e-08, + 3.1723e-08, 3.0559e-08], + [ 6.0129e-08, 2.0652e-07, 8.6613e-08, ..., 1.2037e-07, + 3.1432e-09, 4.2026e-08], + [-5.3179e-07, -4.7963e-07, -3.7020e-07, ..., -5.2853e-07, + 3.0850e-08, -3.5577e-07]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0327, -0.0300, -0.0065, 0.0117, 0.0022, -0.0171, 0.0065, 0.0145, + -0.0275, 0.0083], device='cuda:0'), grad: tensor([ 4.2794e-07, 5.6066e-07, 3.0175e-07, -3.0193e-06, 5.6112e-07, + 1.9111e-06, 2.0780e-07, 3.0291e-07, 9.3831e-07, -2.1923e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 217.48, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4561 re_mapping 0.0028 re_causal 0.0091 /// teacc 99.14 lr 0.00010000 +Epoch 273, weight, value: tensor([[ 0.0179, -0.0798, 0.0043, ..., -0.1142, -0.1488, -0.1255], + [-0.1260, 0.0692, -0.1169, ..., -0.1384, -0.1420, -0.1879], + [-0.0939, -0.1483, -0.1661, ..., -0.2135, -0.1457, 0.1601], + ..., + [ 0.0831, -0.0589, -0.1179, ..., 0.0961, -0.1283, -0.1279], + [ 0.1221, 0.0974, 0.0739, ..., -0.2025, -0.1152, 0.0802], + [ 0.0613, 0.0317, 0.1028, ..., 0.0488, -0.0895, 0.0004]], + device='cuda:0'), grad: tensor([[-8.4983e-09, 1.5425e-08, 1.0419e-08, ..., 2.1246e-08, + 4.4238e-09, 9.1968e-09], + [ 1.6880e-08, -5.0582e-08, 4.7730e-09, ..., 2.6193e-08, + -5.2387e-10, 4.8894e-09], + [ 2.8987e-08, 5.4715e-09, 4.6566e-10, ..., 4.4587e-08, + 4.6566e-10, -4.0745e-10], + ..., + [-2.3865e-09, 5.6927e-08, 3.4284e-08, ..., -9.6043e-09, + 6.9849e-10, 3.0384e-08], + [ 6.2864e-09, 7.7416e-09, 3.7253e-09, ..., 1.0128e-08, + 7.2760e-09, 6.1118e-09], + [-1.7090e-07, -1.7835e-07, -1.6298e-07, ..., -4.6683e-07, + 1.1059e-09, -9.3540e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([ 0.0326, -0.0300, -0.0084, 0.0118, 0.0023, -0.0170, 0.0065, 0.0147, + -0.0275, 0.0084], device='cuda:0'), grad: tensor([-1.3737e-08, -1.3993e-07, 1.4668e-07, 5.1514e-08, 4.6450e-07, + 8.7963e-07, -9.0688e-08, 9.0105e-08, 6.9558e-08, -1.4640e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 217.56, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4633 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.16 lr 0.00010000 +Epoch 274, weight, value: tensor([[ 0.0159, -0.0799, 0.0031, ..., -0.1169, -0.1490, -0.1262], + [-0.1267, 0.0694, -0.1173, ..., -0.1385, -0.1422, -0.1880], + [-0.0945, -0.1489, -0.1675, ..., -0.2140, -0.1464, 0.1602], + ..., + [ 0.0831, -0.0592, -0.1189, ..., 0.0961, -0.1284, -0.1280], + [ 0.1221, 0.0974, 0.0740, ..., -0.2026, -0.1152, 0.0802], + [ 0.0623, 0.0325, 0.1065, ..., 0.0507, -0.0879, 0.0027]], + device='cuda:0'), grad: tensor([[ 2.7649e-09, 1.9267e-08, 5.1514e-09, ..., 1.4406e-08, + 1.4156e-07, 1.4075e-07], + [ 2.1362e-08, 6.6881e-08, 2.4738e-09, ..., 3.3266e-08, + 4.5693e-09, 4.8312e-08], + [ 1.1671e-08, 5.9430e-08, 6.1118e-10, ..., 3.6904e-08, + 1.1083e-07, 1.3935e-07], + ..., + [-1.0116e-07, 2.6147e-07, 1.9762e-08, ..., -1.4005e-07, + 5.1805e-09, 4.0542e-08], + [ 2.8958e-08, -7.7020e-07, 6.3737e-09, ..., 3.7107e-08, + 2.8260e-08, -2.1625e-06], + [-2.3516e-08, -4.4180e-08, -6.3097e-08, ..., 2.0722e-08, + 2.6193e-08, 5.0379e-08]], device='cuda:0') +Epoch 274, bias, value: tensor([ 0.0321, -0.0299, -0.0086, 0.0126, 0.0004, -0.0171, 0.0065, 0.0147, + -0.0275, 0.0103], device='cuda:0'), grad: tensor([ 4.0466e-07, 4.7497e-07, 5.9512e-07, -2.0415e-06, 3.9255e-07, + 5.2340e-06, -5.3411e-07, 7.8091e-07, -5.8301e-06, 5.0152e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 217.61, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4743 re_mapping 0.0029 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +Epoch 275, weight, value: tensor([[ 0.0160, -0.0800, 0.0029, ..., -0.1171, -0.1491, -0.1266], + [-0.1270, 0.0695, -0.1176, ..., -0.1386, -0.1424, -0.1881], + [-0.0950, -0.1494, -0.1688, ..., -0.2151, -0.1470, 0.1603], + ..., + [ 0.0831, -0.0593, -0.1191, ..., 0.0961, -0.1290, -0.1281], + [ 0.1222, 0.0974, 0.0741, ..., -0.2026, -0.1152, 0.0802], + [ 0.0624, 0.0300, 0.1041, ..., 0.0489, -0.0903, 0.0013]], + device='cuda:0'), grad: tensor([[ 7.6252e-09, 3.0152e-08, 1.4843e-08, ..., 9.2259e-09, + 4.1327e-09, 3.6671e-08], + [ 1.0402e-07, 4.9220e-07, 3.6176e-08, ..., 8.4843e-07, + 3.1269e-07, 4.1770e-07], + [ 1.4273e-07, 2.3481e-07, 1.4168e-07, ..., 1.4174e-08, + 3.3178e-09, 9.1968e-09], + ..., + [-1.6368e-07, 3.3848e-08, 1.1700e-08, ..., -1.9977e-07, + 7.4797e-09, 9.0571e-08], + [-3.8627e-07, -6.3609e-07, -4.1653e-07, ..., 3.2538e-08, + 9.6043e-10, -6.6357e-07], + [ 1.7590e-07, 3.1851e-06, 1.1653e-07, ..., 4.6901e-06, + 1.9260e-06, 2.2221e-06]], device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0321, -0.0299, -0.0090, 0.0127, 0.0022, -0.0172, 0.0065, 0.0147, + -0.0275, 0.0084], device='cuda:0'), grad: tensor([ 1.4878e-07, 1.9241e-06, -1.5250e-08, 9.3423e-09, -1.0632e-05, + 2.3621e-07, 3.2643e-07, -1.8021e-07, -2.0973e-06, 1.0282e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 217.56, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4717 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.16 lr 0.00010000 +Epoch 276, weight, value: tensor([[ 0.0160, -0.0800, 0.0027, ..., -0.1175, -0.1492, -0.1269], + [-0.1279, 0.0694, -0.1180, ..., -0.1388, -0.1430, -0.1885], + [-0.0951, -0.1498, -0.1697, ..., -0.2153, -0.1475, 0.1607], + ..., + [ 0.0831, -0.0594, -0.1194, ..., 0.0961, -0.1306, -0.1283], + [ 0.1222, 0.0975, 0.0742, ..., -0.2026, -0.1152, 0.0802], + [ 0.0625, 0.0302, 0.1041, ..., 0.0491, -0.0904, 0.0017]], + device='cuda:0'), grad: tensor([[-1.5704e-07, 1.3708e-08, 3.5128e-08, ..., 5.7189e-08, + 1.1933e-09, 4.5315e-08], + [ 4.1531e-08, 3.3877e-08, 2.0169e-08, ..., 1.0565e-08, + 9.3132e-10, 7.3342e-08], + [ 5.9430e-08, 8.6729e-08, 3.3615e-08, ..., 4.9185e-09, + 2.9104e-10, 1.3481e-07], + ..., + [ 3.8766e-08, 2.2206e-08, 2.2643e-08, ..., 2.6135e-08, + 3.2014e-10, 2.3778e-08], + [-1.5053e-07, -3.0035e-07, -1.1089e-07, ..., 1.0419e-08, + 6.6939e-10, -4.9453e-07], + [-4.8254e-08, -8.4110e-09, -7.1479e-08, ..., -1.1129e-07, + 1.4348e-08, -4.3277e-08]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0321, -0.0301, -0.0088, 0.0126, 0.0021, -0.0172, 0.0065, 0.0147, + -0.0275, 0.0086], device='cuda:0'), grad: tensor([-7.0594e-07, 1.8044e-07, 3.8673e-07, -9.2259e-09, 2.4884e-08, + 2.8452e-07, 7.2736e-07, 1.7462e-07, -1.1120e-06, 5.2067e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 217.66, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4955 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.20 lr 0.00010000 +Epoch 277, weight, value: tensor([[ 0.0161, -0.0802, 0.0023, ..., -0.1178, -0.1495, -0.1272], + [-0.1282, 0.0695, -0.1183, ..., -0.1389, -0.1431, -0.1886], + [-0.0950, -0.1502, -0.1706, ..., -0.2155, -0.1479, 0.1609], + ..., + [ 0.0831, -0.0595, -0.1196, ..., 0.0961, -0.1308, -0.1283], + [ 0.1222, 0.0974, 0.0741, ..., -0.2026, -0.1152, 0.0803], + [ 0.0627, 0.0304, 0.1042, ..., 0.0491, -0.0904, 0.0017]], + device='cuda:0'), grad: tensor([[-2.8958e-08, 4.4238e-09, 1.7462e-09, ..., 2.7544e-07, + 2.5448e-07, 1.3260e-07], + [ 5.1805e-09, -4.0251e-08, 2.9395e-09, ..., 1.1420e-07, + 6.5425e-08, 6.1409e-08], + [ 3.9698e-08, 7.5379e-09, 1.5134e-09, ..., 2.2177e-07, + 3.3731e-08, -1.7439e-07], + ..., + [-3.8213e-08, 1.7637e-08, 3.9872e-09, ..., 6.0594e-08, + 1.1560e-07, 3.4401e-08], + [-1.7521e-08, -3.9872e-09, -6.1118e-10, ..., 4.4063e-08, + 2.8289e-08, 1.5658e-07], + [ 8.5856e-09, -1.9209e-09, -2.0314e-08, ..., 4.7162e-06, + 2.5537e-06, 1.5004e-06]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0320, -0.0301, -0.0088, 0.0128, 0.0021, -0.0171, 0.0064, 0.0147, + -0.0275, 0.0086], device='cuda:0'), grad: tensor([ 6.3004e-07, 1.5250e-07, -3.0687e-07, 1.5262e-07, -2.0027e-05, + -2.2829e-07, 1.0505e-05, 1.9860e-07, 7.3295e-07, 8.1807e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 217.47, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4865 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.12 lr 0.00010000 +Epoch 278, weight, value: tensor([[ 0.0165, -0.0803, 0.0021, ..., -0.1180, -0.1495, -0.1274], + [-0.1293, 0.0695, -0.1188, ..., -0.1391, -0.1433, -0.1890], + [-0.0961, -0.1517, -0.1720, ..., -0.2160, -0.1484, 0.1610], + ..., + [ 0.0831, -0.0596, -0.1197, ..., 0.0962, -0.1310, -0.1284], + [ 0.1222, 0.0976, 0.0743, ..., -0.2026, -0.1153, 0.0803], + [ 0.0627, 0.0304, 0.1043, ..., 0.0491, -0.0904, 0.0017]], + device='cuda:0'), grad: tensor([[-1.2829e-07, -1.6589e-09, -2.1042e-08, ..., 8.7311e-09, + 5.6345e-08, 4.3772e-08], + [ 2.3836e-08, 6.4203e-08, 2.0518e-08, ..., 7.1828e-08, + 8.8534e-08, 5.6985e-08], + [ 1.3009e-08, 9.6508e-08, 3.5885e-08, ..., 2.3865e-08, + 1.3184e-08, -1.6444e-08], + ..., + [-9.4296e-09, 7.2294e-08, 2.1333e-08, ..., 1.8976e-08, + 7.4564e-08, 5.0903e-08], + [ 3.5798e-08, 5.5443e-08, 2.5524e-08, ..., 8.0036e-09, + 3.3266e-08, 2.1479e-08], + [ 2.2672e-08, 5.2183e-08, 1.9209e-09, ..., 5.9401e-08, + 9.4646e-08, 5.5152e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0322, -0.0302, -0.0090, 0.0124, 0.0021, -0.0171, 0.0064, 0.0147, + -0.0275, 0.0086], device='cuda:0'), grad: tensor([-7.2829e-07, 5.1083e-07, 3.3411e-07, -1.5218e-06, -6.6496e-07, + 2.0105e-07, 3.6578e-07, 3.0361e-07, 6.2678e-07, 5.7882e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 217.45, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5230 re_mapping 0.0030 re_causal 0.0100 /// teacc 99.11 lr 0.00010000 +Epoch 279, weight, value: tensor([[ 0.0169, -0.0803, 0.0021, ..., -0.1181, -0.1498, -0.1277], + [-0.1298, 0.0698, -0.1191, ..., -0.1394, -0.1445, -0.1893], + [-0.0964, -0.1523, -0.1698, ..., -0.2163, -0.1522, 0.1630], + ..., + [ 0.0832, -0.0598, -0.1199, ..., 0.0963, -0.1316, -0.1285], + [ 0.1222, 0.0976, 0.0745, ..., -0.2026, -0.1153, 0.0804], + [ 0.0627, 0.0304, 0.1042, ..., 0.0491, -0.0905, 0.0016]], + device='cuda:0'), grad: tensor([[ 2.5029e-09, 8.0327e-08, 6.3796e-08, ..., 5.1659e-08, + 1.1874e-07, 1.2538e-07], + [ 2.2701e-08, 1.4348e-08, 4.6333e-08, ..., 6.8627e-08, + 4.1066e-08, 4.8836e-08], + [ 3.0617e-08, 5.6694e-08, 3.7253e-08, ..., 9.6392e-08, + 1.9529e-08, 2.5611e-09], + ..., + [-5.7975e-08, 3.1316e-08, 1.7521e-08, ..., -2.5937e-07, + 8.5274e-09, 3.2305e-09], + [-4.7177e-08, -1.9674e-08, -1.2224e-08, ..., 6.6240e-08, + 1.1188e-07, 4.0920e-08], + [ 1.5221e-08, 6.0536e-08, 4.8662e-08, ..., 3.3382e-08, + 3.1025e-08, 2.6979e-08]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0323, -0.0303, -0.0073, 0.0124, 0.0021, -0.0171, 0.0064, 0.0147, + -0.0275, 0.0085], device='cuda:0'), grad: tensor([ 9.0059e-07, 5.3272e-07, 5.3970e-07, -3.3379e-06, 2.2561e-07, + 1.8384e-06, -1.0869e-06, -6.7847e-07, 5.2992e-07, 5.5134e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 217.39, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4738 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.16 lr 0.00010000 +Epoch 280, weight, value: tensor([[ 0.0171, -0.0804, 0.0020, ..., -0.1183, -0.1499, -0.1278], + [-0.1301, 0.0700, -0.1193, ..., -0.1395, -0.1448, -0.1895], + [-0.0969, -0.1529, -0.1700, ..., -0.2171, -0.1525, 0.1630], + ..., + [ 0.0832, -0.0600, -0.1200, ..., 0.0963, -0.1319, -0.1285], + [ 0.1222, 0.0976, 0.0746, ..., -0.2027, -0.1153, 0.0804], + [ 0.0628, 0.0306, 0.1043, ..., 0.0493, -0.0904, 0.0017]], + device='cuda:0'), grad: tensor([[ 1.1700e-08, 1.4348e-08, 1.6880e-09, ..., 1.3097e-09, + 2.3923e-08, 2.3138e-08], + [ 1.1642e-08, -2.3702e-07, 5.5006e-09, ..., 9.4587e-09, + 5.6170e-09, 1.0565e-08], + [ 1.3766e-08, 1.1252e-07, 1.7084e-08, ..., 1.4319e-08, + 1.0594e-08, 1.4930e-08], + ..., + [-2.8522e-08, 1.0850e-07, 4.7148e-09, ..., -3.6525e-08, + 4.3656e-09, 7.3633e-09], + [ 9.1968e-09, 1.1059e-08, -2.2323e-08, ..., 7.3633e-09, + 6.7521e-08, 2.8289e-08], + [ 1.3941e-08, 1.7782e-08, 8.7311e-11, ..., 1.8335e-08, + 6.7230e-09, 8.0909e-09]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0323, -0.0302, -0.0076, 0.0135, 0.0020, -0.0172, 0.0064, 0.0148, + -0.0275, 0.0086], device='cuda:0'), grad: tensor([ 1.0739e-07, -8.1863e-07, 3.5716e-07, -2.0256e-07, 4.4063e-08, + 4.2608e-08, -1.6764e-07, 2.9244e-07, 2.4121e-07, 1.1653e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 217.22, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4435 re_mapping 0.0033 re_causal 0.0099 /// teacc 99.14 lr 0.00010000 +Epoch 281, weight, value: tensor([[ 0.0174, -0.0804, 0.0020, ..., -0.1184, -0.1500, -0.1280], + [-0.1301, 0.0705, -0.1193, ..., -0.1396, -0.1449, -0.1895], + [-0.0973, -0.1535, -0.1706, ..., -0.2174, -0.1530, 0.1630], + ..., + [ 0.0832, -0.0603, -0.1201, ..., 0.0964, -0.1322, -0.1285], + [ 0.1222, 0.0976, 0.0746, ..., -0.2027, -0.1153, 0.0804], + [ 0.0627, 0.0307, 0.1043, ..., 0.0493, -0.0904, 0.0017]], + device='cuda:0'), grad: tensor([[ 1.2026e-07, 3.9057e-08, 2.7387e-08, ..., 1.8044e-07, + 2.6223e-08, 8.5565e-08], + [ 9.0804e-09, 3.7428e-08, 1.2165e-08, ..., 1.3999e-08, + 2.1828e-09, 1.5367e-08], + [-5.9663e-09, 1.1444e-07, 4.2492e-09, ..., -7.3342e-09, + -2.5611e-09, -4.8167e-08], + ..., + [ 7.4971e-08, 8.1491e-08, 2.8260e-08, ..., 1.1473e-07, + 1.9005e-08, 7.1130e-08], + [-2.6484e-09, 6.5367e-08, 1.0885e-08, ..., 3.6089e-09, + 2.7940e-09, 7.8580e-10], + [-2.8219e-07, -4.2317e-08, -5.9168e-08, ..., -4.2026e-07, + -5.1165e-08, -1.8626e-07]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0324, -0.0299, -0.0075, 0.0136, 0.0020, -0.0172, 0.0064, 0.0148, + -0.0276, 0.0086], device='cuda:0'), grad: tensor([ 3.3691e-07, 7.8115e-08, 2.0233e-07, -2.3879e-06, 2.6193e-07, + 1.3802e-06, 1.2340e-07, 4.2166e-07, 2.3178e-07, -6.2631e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 217.63, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4826 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.16 lr 0.00010000 +Epoch 282, weight, value: tensor([[ 0.0176, -0.0805, 0.0019, ..., -0.1186, -0.1502, -0.1286], + [-0.1318, 0.0705, -0.1196, ..., -0.1398, -0.1461, -0.1898], + [-0.0975, -0.1539, -0.1710, ..., -0.2178, -0.1534, 0.1635], + ..., + [ 0.0833, -0.0604, -0.1203, ..., 0.0965, -0.1325, -0.1287], + [ 0.1222, 0.0978, 0.0750, ..., -0.2027, -0.1153, 0.0804], + [ 0.0626, 0.0306, 0.1043, ..., 0.0493, -0.0904, 0.0017]], + device='cuda:0'), grad: tensor([[ 7.0198e-08, 1.8277e-07, 1.2736e-07, ..., 8.2422e-08, + 9.0804e-09, 8.3877e-08], + [ 2.6822e-07, 3.4808e-07, 1.8789e-07, ..., 3.5414e-07, + 2.7358e-09, 1.2503e-07], + [ 4.5472e-07, 1.6834e-07, 1.1991e-07, ..., 5.8301e-07, + 1.8044e-09, 8.3761e-08], + ..., + [-1.1260e-06, 6.4634e-07, 4.7917e-07, ..., -7.2233e-06, + 1.1642e-10, 3.0617e-07], + [-2.8498e-06, -8.4341e-06, -6.2324e-06, ..., 9.2143e-08, + 1.1874e-08, -4.0233e-06], + [ 1.0822e-06, 2.0508e-06, 1.6000e-06, ..., 3.1628e-06, + 1.5716e-09, 1.0533e-06]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0323, -0.0302, -0.0072, 0.0135, 0.0033, -0.0171, 0.0046, 0.0148, + -0.0276, 0.0086], device='cuda:0'), grad: tensor([ 7.7253e-07, 2.3581e-06, 2.3730e-06, 1.3620e-05, 9.1828e-07, + 6.9402e-06, 1.7160e-07, -1.6406e-05, -2.4080e-05, 1.3344e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 217.45, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4569 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.16 lr 0.00010000 +Epoch 283, weight, value: tensor([[ 0.0178, -0.0806, 0.0016, ..., -0.1188, -0.1507, -0.1293], + [-0.1330, 0.0705, -0.1202, ..., -0.1402, -0.1465, -0.1901], + [-0.0980, -0.1549, -0.1717, ..., -0.2185, -0.1557, 0.1634], + ..., + [ 0.0833, -0.0608, -0.1208, ..., 0.0967, -0.1328, -0.1288], + [ 0.1222, 0.0979, 0.0755, ..., -0.2027, -0.1153, 0.0804], + [ 0.0628, 0.0308, 0.1044, ..., 0.0493, -0.0905, 0.0017]], + device='cuda:0'), grad: tensor([[-3.0093e-08, 6.1118e-09, 2.1537e-09, ..., 9.0222e-09, + 2.6193e-09, 1.5891e-08], + [ 2.1420e-08, -2.3399e-08, 2.5029e-09, ..., 4.5868e-08, + 6.9849e-10, 1.3504e-08], + [ 7.7416e-09, 1.2689e-08, 1.1642e-09, ..., 9.2550e-09, + 4.0745e-10, -1.5565e-07], + ..., + [-6.1234e-08, 3.0035e-08, 1.0477e-08, ..., -1.2410e-07, + 2.5029e-09, 3.6031e-08], + [-9.2550e-09, -7.1595e-09, -1.0245e-08, ..., 1.2049e-08, + 1.9209e-09, 3.6962e-08], + [ 2.8173e-08, -5.9546e-08, -4.0221e-08, ..., -7.5030e-08, + 5.4715e-09, -2.8696e-08]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0327, -0.0307, -0.0069, 0.0137, 0.0033, -0.0172, 0.0045, 0.0149, + -0.0275, 0.0086], device='cuda:0'), grad: tensor([-6.0478e-08, 7.4506e-09, -3.4110e-07, 6.8336e-08, 2.2608e-07, + 2.3283e-08, 4.8545e-08, -9.7905e-08, 1.7020e-07, -3.8592e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 217.67, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4685 re_mapping 0.0029 re_causal 0.0097 /// teacc 99.07 lr 0.00010000 +Epoch 284, weight, value: tensor([[ 0.0178, -0.0807, 0.0014, ..., -0.1190, -0.1511, -0.1300], + [-0.1355, 0.0707, -0.1204, ..., -0.1403, -0.1468, -0.1903], + [-0.0991, -0.1557, -0.1728, ..., -0.2191, -0.1559, 0.1636], + ..., + [ 0.0834, -0.0610, -0.1211, ..., 0.0968, -0.1333, -0.1288], + [ 0.1223, 0.0980, 0.0757, ..., -0.2027, -0.1153, 0.0805], + [ 0.0628, 0.0308, 0.1044, ..., 0.0492, -0.0905, 0.0017]], + device='cuda:0'), grad: tensor([[-4.8894e-09, 8.4401e-09, 1.1059e-09, ..., 1.2456e-08, + 5.0641e-09, 6.5775e-09], + [ 4.8312e-09, -5.0757e-08, 1.1059e-09, ..., 2.4855e-08, + 1.3155e-08, 1.2282e-08], + [ 1.1642e-08, 1.2340e-08, 4.7148e-09, ..., 1.9267e-08, + 7.9162e-09, 1.1118e-08], + ..., + [ 2.2643e-08, 6.2282e-08, 1.8335e-08, ..., 5.8964e-08, + 2.6193e-08, 3.6904e-08], + [-3.4692e-08, 1.1415e-07, -1.6706e-08, ..., 1.0978e-07, + 4.8487e-08, 4.7556e-08], + [-1.5891e-08, -1.4552e-09, -1.5774e-08, ..., 8.0140e-07, + 3.8883e-07, 4.4657e-07]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0325, -0.0311, -0.0078, 0.0141, 0.0033, -0.0173, 0.0045, 0.0152, + -0.0275, 0.0085], device='cuda:0'), grad: tensor([ 6.0129e-08, -1.8801e-07, 1.1700e-07, 4.4852e-06, -2.8498e-06, + -6.0163e-06, 5.0198e-07, 4.3632e-07, 1.1409e-06, 2.3376e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 217.54, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4820 re_mapping 0.0030 re_causal 0.0102 /// teacc 99.03 lr 0.00010000 +Epoch 285, weight, value: tensor([[ 0.0180, -0.0808, 0.0012, ..., -0.1192, -0.1510, -0.1294], + [-0.1324, 0.0738, -0.1208, ..., -0.1407, -0.1472, -0.1906], + [-0.0998, -0.1561, -0.1730, ..., -0.2199, -0.1559, 0.1640], + ..., + [ 0.0819, -0.0641, -0.1215, ..., 0.0972, -0.1340, -0.1287], + [ 0.1222, 0.0980, 0.0757, ..., -0.2028, -0.1154, 0.0804], + [ 0.0623, 0.0309, 0.1045, ..., 0.0492, -0.0905, 0.0017]], + device='cuda:0'), grad: tensor([[ 3.0268e-08, 8.8650e-08, 2.2410e-08, ..., 4.6508e-08, + 1.6298e-07, 1.4633e-07], + [ 3.9407e-08, 3.9884e-07, 1.4377e-08, ..., 5.1165e-08, + 6.5134e-08, 5.3027e-08], + [ 2.1886e-08, 1.0338e-06, 5.8208e-09, ..., 2.5611e-08, + 3.1781e-08, 3.2480e-08], + ..., + [-1.5949e-08, 2.7358e-07, 3.7253e-08, ..., -1.7346e-08, + 1.2747e-08, 1.4494e-08], + [ 2.1327e-07, 7.3621e-07, 7.7940e-08, ..., 3.0617e-07, + 2.3544e-06, 2.7642e-06], + [-4.2492e-07, -3.9022e-07, -3.2224e-07, ..., -5.2992e-07, + -2.2154e-07, -2.3795e-07]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0329, -0.0280, -0.0078, 0.0156, 0.0033, -0.0174, 0.0044, 0.0128, + -0.0276, 0.0085], device='cuda:0'), grad: tensor([ 6.9616e-07, 1.8785e-06, 4.5188e-06, -1.9833e-05, 1.7965e-06, + 1.1139e-05, -9.0450e-06, 9.1782e-07, 9.5442e-06, -1.6410e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 217.65, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4756 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.02 lr 0.00010000 +Epoch 286, weight, value: tensor([[ 0.0178, -0.0809, 0.0010, ..., -0.1197, -0.1520, -0.1306], + [-0.1324, 0.0738, -0.1210, ..., -0.1415, -0.1479, -0.1911], + [-0.1012, -0.1574, -0.1736, ..., -0.2212, -0.1572, 0.1645], + ..., + [ 0.0820, -0.0642, -0.1219, ..., 0.0973, -0.1346, -0.1288], + [ 0.1222, 0.0980, 0.0759, ..., -0.2029, -0.1154, 0.0804], + [ 0.0626, 0.0311, 0.1046, ..., 0.0493, -0.0905, 0.0017]], + device='cuda:0'), grad: tensor([[-2.4447e-09, 1.6298e-09, 7.5670e-10, ..., 5.9954e-09, + 1.5716e-08, 1.2922e-08], + [ 1.1409e-08, 1.0477e-08, 1.1642e-09, ..., 5.0291e-08, + 3.9232e-08, 1.8685e-08], + [ 5.9372e-09, 6.4028e-09, 1.3388e-09, ..., 7.9162e-09, + 4.3074e-09, 1.2224e-09], + ..., + [-2.4447e-08, 1.1467e-08, 3.9581e-09, ..., 1.6880e-08, + 2.5611e-08, 1.3039e-08], + [-2.7940e-09, -2.7358e-09, -1.4552e-09, ..., 4.0163e-09, + 4.5053e-08, 3.9698e-08], + [ 9.3132e-09, 6.0536e-09, -5.0059e-09, ..., 1.2135e-06, + 5.9418e-07, 2.5192e-07]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0325, -0.0281, -0.0079, 0.0191, 0.0033, -0.0182, 0.0045, 0.0128, + -0.0277, 0.0085], device='cuda:0'), grad: tensor([ 4.2433e-08, 1.1845e-07, 3.4401e-08, 8.2830e-08, -1.7993e-06, + -3.2946e-07, 9.3132e-10, 3.9407e-08, 1.6170e-07, 1.6633e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 217.48, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.4935 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.05 lr 0.00010000 +Epoch 287, weight, value: tensor([[ 0.0180, -0.0810, 0.0007, ..., -0.1199, -0.1523, -0.1310], + [-0.1325, 0.0738, -0.1220, ..., -0.1417, -0.1484, -0.1916], + [-0.1021, -0.1586, -0.1748, ..., -0.2216, -0.1577, 0.1647], + ..., + [ 0.0819, -0.0642, -0.1234, ..., 0.0971, -0.1350, -0.1289], + [ 0.1223, 0.0983, 0.0762, ..., -0.2029, -0.1154, 0.0805], + [ 0.0634, 0.0312, 0.1048, ..., 0.0494, -0.0905, 0.0017]], + device='cuda:0'), grad: tensor([[ 1.2282e-08, 6.7055e-08, 1.8917e-08, ..., 1.5949e-07, + 9.5810e-08, 7.1770e-08], + [ 9.2899e-08, 3.4855e-07, 2.2235e-08, ..., 1.0366e-06, + 5.8347e-07, 3.7719e-07], + [ 1.5239e-07, 1.7835e-07, 1.2084e-07, ..., 4.1095e-08, + 1.0652e-08, 1.2200e-07], + ..., + [-5.2806e-07, 8.0734e-08, 3.5740e-08, ..., -1.1094e-07, + 9.6334e-08, 6.5600e-08], + [-8.8185e-08, -3.0175e-07, -2.2550e-07, ..., 1.0792e-07, + 2.6822e-07, -8.6613e-08], + [ 2.0384e-07, 5.2061e-07, -5.1805e-09, ..., 1.5702e-06, + 8.8196e-07, 5.7090e-07]], device='cuda:0') +Epoch 287, bias, value: tensor([ 0.0325, -0.0281, -0.0081, 0.0187, 0.0033, -0.0180, 0.0044, 0.0128, + -0.0277, 0.0086], device='cuda:0'), grad: tensor([ 4.1560e-07, 2.8070e-06, 8.7079e-07, 5.7137e-07, -7.3463e-06, + -1.3700e-06, 2.5472e-07, -1.0738e-06, 2.4028e-07, 4.6268e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 217.50, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4983 re_mapping 0.0028 re_causal 0.0097 /// teacc 99.03 lr 0.00010000 +Epoch 288, weight, value: tensor([[ 0.0185, -0.0811, 0.0006, ..., -0.1201, -0.1525, -0.1314], + [-0.1326, 0.0738, -0.1234, ..., -0.1423, -0.1492, -0.1922], + [-0.0997, -0.1593, -0.1756, ..., -0.2211, -0.1583, 0.1659], + ..., + [ 0.0819, -0.0643, -0.1241, ..., 0.0972, -0.1354, -0.1294], + [ 0.1223, 0.0984, 0.0766, ..., -0.2029, -0.1154, 0.0805], + [ 0.0636, 0.0313, 0.1049, ..., 0.0494, -0.0906, 0.0017]], + device='cuda:0'), grad: tensor([[-3.1490e-08, 1.7462e-09, 9.8953e-10, ..., 1.7462e-09, + 4.6566e-10, 4.0745e-10], + [ 4.8312e-09, -2.5728e-08, 1.2224e-09, ..., 5.4133e-09, + 1.4552e-09, 3.2596e-09], + [ 2.5029e-09, 5.2387e-09, 5.2387e-10, ..., 2.3283e-09, + 8.1491e-10, -4.0745e-10], + ..., + [-1.9209e-09, 6.8103e-09, 2.6776e-09, ..., -1.8626e-09, + 5.2387e-10, 1.9791e-09], + [ 3.2596e-09, -9.5461e-09, -8.0327e-09, ..., 2.5611e-09, + -8.3237e-09, -2.0780e-08], + [-5.2387e-10, -1.9791e-09, -8.4401e-09, ..., -1.1118e-08, + 3.8417e-09, 2.7358e-09]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0326, -0.0282, -0.0059, 0.0186, 0.0033, -0.0179, 0.0044, 0.0126, + -0.0277, 0.0086], device='cuda:0'), grad: tensor([-7.3051e-08, -6.0129e-08, 1.6415e-08, 1.6333e-07, 1.0070e-08, + -2.3108e-07, 1.2573e-07, 2.2526e-08, 2.0955e-08, 1.8626e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 217.63, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4650 re_mapping 0.0028 re_causal 0.0093 /// teacc 98.92 lr 0.00010000 +Epoch 289, weight, value: tensor([[ 1.8731e-02, -8.1264e-02, 1.7369e-04, ..., -1.2037e-01, + -1.5261e-01, -1.3174e-01], + [-1.3277e-01, 7.3782e-02, -1.2377e-01, ..., -1.4282e-01, + -1.5047e-01, -1.9334e-01], + [-1.0084e-01, -1.6071e-01, -1.7620e-01, ..., -2.1931e-01, + -1.5565e-01, 1.6927e-01], + ..., + [ 8.1991e-02, -6.4285e-02, -1.2440e-01, ..., 9.7352e-02, + -1.3588e-01, -1.2903e-01], + [ 1.2228e-01, 9.8514e-02, 7.6588e-02, ..., -2.0296e-01, + -1.1549e-01, 8.0574e-02], + [ 6.3974e-02, 3.1591e-02, 1.0515e-01, ..., 4.9492e-02, + -9.0535e-02, 1.6995e-03]], device='cuda:0'), grad: tensor([[ 1.2135e-06, 6.9104e-07, 3.9395e-07, ..., 8.7917e-07, + 2.6403e-07, 5.0245e-07], + [ 1.3364e-07, 3.9057e-08, 3.0443e-08, ..., 8.0967e-08, + 1.5483e-08, 5.6636e-08], + [ 2.4820e-07, 9.7207e-08, 4.5227e-08, ..., 1.5053e-07, + 2.2235e-08, 3.7835e-08], + ..., + [-8.1304e-07, 1.9325e-07, 1.0862e-07, ..., -4.8429e-07, + 6.4611e-08, 1.4575e-07], + [ 4.3190e-07, 7.3749e-08, 7.8289e-08, ..., 3.8021e-07, + 8.2015e-08, 1.7346e-08], + [-3.4962e-06, -2.3376e-06, -1.3793e-06, ..., -2.6450e-06, + -9.0990e-07, -1.6754e-06]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0326, -0.0283, -0.0031, 0.0184, 0.0014, -0.0179, 0.0044, 0.0127, + -0.0277, 0.0088], device='cuda:0'), grad: tensor([ 3.9674e-06, 3.8114e-07, 6.1933e-07, 1.4611e-05, 4.7870e-06, + -1.2420e-05, 1.0041e-07, -1.5888e-06, 1.2526e-06, -1.1727e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 217.81, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4734 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 +Epoch 290, weight, value: tensor([[ 1.8785e-02, -8.1407e-02, -9.9547e-05, ..., -1.2069e-01, + -1.5302e-01, -1.3246e-01], + [-1.3283e-01, 7.3821e-02, -1.2401e-01, ..., -1.4317e-01, + -1.5095e-01, -1.9369e-01], + [-1.0163e-01, -1.6189e-01, -1.7674e-01, ..., -2.1929e-01, + -1.5563e-01, 1.6933e-01], + ..., + [ 8.2075e-02, -6.4324e-02, -1.2463e-01, ..., 9.7497e-02, + -1.3692e-01, -1.2908e-01], + [ 1.2228e-01, 9.8551e-02, 7.6665e-02, ..., -2.0299e-01, + -1.1556e-01, 8.0554e-02], + [ 6.3830e-02, 3.1641e-02, 1.0520e-01, ..., 4.9464e-02, + -9.0581e-02, 1.6804e-03]], device='cuda:0'), grad: tensor([[ 1.3330e-08, 4.6333e-08, 2.5029e-08, ..., 8.2073e-09, + 4.0745e-10, 3.2713e-08], + [ 8.9582e-08, 9.9419e-08, 6.2631e-08, ..., 3.2422e-08, + 1.1642e-09, 8.9349e-08], + [ 5.0291e-08, 7.0489e-08, 3.5099e-08, ..., 8.7894e-09, + 7.5670e-10, 4.1444e-08], + ..., + [ 3.8301e-08, 7.4680e-08, 4.4645e-08, ..., 9.3132e-09, + 1.1642e-09, 4.8371e-08], + [-5.8161e-07, -9.5833e-07, -5.0059e-07, ..., 1.0012e-08, + 3.4925e-10, -7.2410e-07], + [ 3.3411e-08, 1.6124e-07, 6.1234e-08, ..., -1.6741e-07, + 7.2177e-09, 1.5751e-07]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0325, -0.0283, -0.0031, 0.0182, 0.0013, -0.0178, 0.0044, 0.0128, + -0.0277, 0.0087], device='cuda:0'), grad: tensor([ 6.1817e-08, 3.9791e-07, 2.3562e-07, 4.3237e-07, 2.6450e-07, + 6.5193e-07, 2.7264e-07, 2.0547e-07, -2.9802e-06, 4.5309e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 217.30, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4684 re_mapping 0.0028 re_causal 0.0092 /// teacc 98.96 lr 0.00010000 +Epoch 291, weight, value: tensor([[ 0.0206, -0.0816, -0.0006, ..., -0.1211, -0.1508, -0.1318], + [-0.1329, 0.0738, -0.1245, ..., -0.1436, -0.1516, -0.1940], + [-0.1026, -0.1627, -0.1774, ..., -0.2193, -0.1557, 0.1693], + ..., + [ 0.0821, -0.0644, -0.1252, ..., 0.0976, -0.1378, -0.1293], + [ 0.1223, 0.0988, 0.0772, ..., -0.2030, -0.1156, 0.0807], + [ 0.0641, 0.0317, 0.1053, ..., 0.0495, -0.0906, 0.0017]], + device='cuda:0'), grad: tensor([[ 1.7288e-08, 3.8184e-08, 3.5507e-08, ..., 6.5425e-08, + 1.4028e-08, 1.6764e-08], + [ 1.1944e-07, 4.4762e-08, 5.3202e-08, ..., 1.9511e-07, + 3.9756e-08, 4.0163e-08], + [ 4.3423e-08, 2.3865e-08, 1.0419e-08, ..., 5.2969e-08, + 1.8277e-08, 7.8580e-09], + ..., + [ 2.4866e-07, 2.7614e-07, 3.0873e-07, ..., 5.0617e-07, + 5.2620e-08, 7.1421e-08], + [-9.7149e-08, -8.1724e-08, -3.9814e-08, ..., 9.8546e-08, + 2.0897e-08, -6.6007e-08], + [-5.6857e-07, -1.9814e-07, -5.2387e-07, ..., -4.0070e-07, + 4.8988e-07, 3.6042e-07]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0340, -0.0284, -0.0031, 0.0180, 0.0013, -0.0178, 0.0043, 0.0128, + -0.0277, 0.0087], device='cuda:0'), grad: tensor([ 1.2456e-07, 4.6985e-07, 1.6566e-07, 2.0443e-07, -1.6466e-06, + 2.3353e-07, 4.4936e-08, 1.2172e-06, -1.2515e-07, -6.7428e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 217.56, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4692 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.08 lr 0.00010000 +Epoch 292, weight, value: tensor([[ 0.0204, -0.0818, -0.0009, ..., -0.1215, -0.1532, -0.1344], + [-0.1329, 0.0740, -0.1248, ..., -0.1439, -0.1525, -0.1931], + [-0.1017, -0.1655, -0.1781, ..., -0.2194, -0.1557, 0.1693], + ..., + [ 0.0821, -0.0644, -0.1256, ..., 0.0976, -0.1391, -0.1297], + [ 0.1223, 0.0988, 0.0774, ..., -0.2030, -0.1156, 0.0806], + [ 0.0642, 0.0316, 0.1052, ..., 0.0494, -0.0909, 0.0016]], + device='cuda:0'), grad: tensor([[-2.0571e-07, -4.5053e-08, -3.5157e-08, ..., -8.7311e-09, + -8.0909e-09, -2.2526e-08], + [ 1.6182e-08, 1.3947e-07, 1.6298e-09, ..., 1.2049e-08, + 3.3341e-07, 3.6648e-07], + [ 2.0023e-08, 1.6822e-08, 1.9209e-09, ..., 8.2655e-09, + 6.6357e-09, -6.5193e-09], + ..., + [-1.4494e-08, 2.2876e-08, 5.7626e-09, ..., -3.2596e-08, + 2.2119e-09, 1.0594e-08], + [-3.5681e-08, 3.6582e-06, -2.3283e-10, ..., 1.1467e-08, + 7.5772e-06, 8.0168e-06], + [ 1.8033e-07, 6.7404e-08, 1.6589e-08, ..., 1.1234e-08, + 4.3074e-08, 6.6473e-08]], device='cuda:0') +Epoch 292, bias, value: tensor([ 0.0326, -0.0283, -0.0031, 0.0179, 0.0015, -0.0177, 0.0045, 0.0128, + -0.0278, 0.0086], device='cuda:0'), grad: tensor([-4.9639e-07, 1.0626e-06, 6.8685e-09, 4.5518e-08, 7.6718e-08, + 2.1905e-06, -2.7880e-05, -1.3388e-08, 2.4408e-05, 5.9139e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 217.46, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4701 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.14 lr 0.00010000 +Epoch 293, weight, value: tensor([[ 0.0205, -0.0817, -0.0010, ..., -0.1215, -0.1533, -0.1346], + [-0.1329, 0.0740, -0.1253, ..., -0.1442, -0.1533, -0.1934], + [-0.1021, -0.1660, -0.1785, ..., -0.2194, -0.1557, 0.1693], + ..., + [ 0.0822, -0.0644, -0.1257, ..., 0.0977, -0.1392, -0.1297], + [ 0.1223, 0.0990, 0.0777, ..., -0.2031, -0.1156, 0.0807], + [ 0.0641, 0.0316, 0.1052, ..., 0.0493, -0.0909, 0.0015]], + device='cuda:0'), grad: tensor([[-3.2305e-08, -5.5879e-09, -1.7462e-08, ..., 1.0303e-08, + 4.8312e-09, 3.7835e-09], + [ 5.1688e-08, -5.5297e-08, 5.8790e-09, ..., 5.7393e-08, + 4.3656e-09, 7.2760e-09], + [ 5.8324e-08, 8.4401e-09, 2.5029e-09, ..., 1.0611e-07, + 3.8999e-09, 9.5461e-09], + ..., + [-5.5693e-07, 1.7288e-08, 3.0268e-09, ..., -6.7102e-07, + 4.1910e-09, -2.7998e-08], + [ 6.6240e-08, 1.2398e-08, 9.8953e-10, ..., 8.0036e-08, + 5.7626e-09, 2.6776e-09], + [ 3.4412e-07, 2.1828e-08, -5.9954e-09, ..., 4.0233e-07, + 5.3085e-08, 2.7940e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0328, -0.0283, -0.0032, 0.0179, 0.0015, -0.0177, 0.0045, 0.0128, + -0.0278, 0.0086], device='cuda:0'), grad: tensor([-1.6124e-07, 5.1281e-08, 2.4983e-07, 1.5949e-07, -3.1490e-08, + -5.4669e-07, 1.7299e-07, -1.5497e-06, 5.5600e-07, 1.0850e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 218.42, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4579 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.08 lr 0.00010000 +Epoch 294, weight, value: tensor([[ 0.0200, -0.0823, -0.0016, ..., -0.1220, -0.1535, -0.1353], + [-0.1307, 0.0763, -0.1266, ..., -0.1453, -0.1543, -0.1946], + [-0.1037, -0.1688, -0.1799, ..., -0.2194, -0.1558, 0.1693], + ..., + [ 0.0821, -0.0647, -0.1264, ..., 0.0980, -0.1413, -0.1299], + [ 0.1207, 0.0965, 0.0783, ..., -0.2031, -0.1156, 0.0811], + [ 0.0645, 0.0318, 0.1054, ..., 0.0494, -0.0909, 0.0016]], + device='cuda:0'), grad: tensor([[ 4.8894e-09, 3.8359e-08, 1.4319e-08, ..., 8.5565e-09, + 7.6892e-08, 5.3609e-08], + [ 3.7777e-08, 8.6089e-08, 3.1258e-08, ..., 5.5879e-08, + 2.7358e-09, 4.7730e-09], + [ 3.1432e-08, 3.8603e-07, 1.2037e-07, ..., 4.5111e-08, + 1.3388e-09, -6.4028e-09], + ..., + [-1.0070e-07, 1.3900e-07, 5.6403e-08, ..., -1.4831e-07, + 1.7462e-10, 2.1537e-09], + [ 6.2282e-09, 2.8173e-07, 8.7079e-08, ..., 1.7229e-08, + 3.8359e-08, 2.3632e-08], + [-2.9686e-08, 1.2864e-08, -4.4121e-08, ..., -6.5367e-08, + 2.6776e-09, 3.4925e-10]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0324, -0.0257, -0.0033, 0.0176, 0.0014, -0.0181, 0.0044, 0.0127, + -0.0303, 0.0086], device='cuda:0'), grad: tensor([ 2.7660e-07, 4.1421e-07, 1.3188e-06, -3.7830e-06, 2.1537e-07, + 6.4215e-07, -2.0373e-07, 2.9919e-08, 1.0142e-06, 8.6613e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 217.74, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4932 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.10 lr 0.00010000 +Epoch 295, weight, value: tensor([[ 0.0194, -0.0832, -0.0034, ..., -0.1223, -0.1535, -0.1363], + [-0.1307, 0.0763, -0.1276, ..., -0.1458, -0.1555, -0.1970], + [-0.1046, -0.1693, -0.1807, ..., -0.2196, -0.1558, 0.1694], + ..., + [ 0.0822, -0.0647, -0.1271, ..., 0.0982, -0.1425, -0.1301], + [ 0.1208, 0.0969, 0.0806, ..., -0.2032, -0.1156, 0.0815], + [ 0.0647, 0.0319, 0.1055, ..., 0.0495, -0.0908, 0.0016]], + device='cuda:0'), grad: tensor([[-2.0838e-08, 9.3132e-09, 7.5670e-10, ..., 4.6566e-10, + 4.5984e-09, 1.1118e-08], + [ 4.0978e-08, -9.9614e-06, 1.6298e-09, ..., 9.5984e-08, + 1.5134e-09, 4.0745e-08], + [ 7.2177e-09, 5.3924e-07, 2.5611e-09, ..., 7.9744e-09, + 2.5029e-09, -4.7288e-07], + ..., + [-9.1910e-08, 8.8736e-06, 2.0955e-09, ..., -2.3376e-07, + 5.8208e-10, 1.2398e-08], + [-1.0012e-08, -2.3865e-08, -1.9383e-08, ..., 1.0477e-09, + -2.0373e-09, 3.5507e-07], + [ 6.2457e-08, 1.8044e-07, -1.5716e-09, ..., 1.2550e-07, + 1.3388e-09, 1.4261e-08]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0321, -0.0264, -0.0022, 0.0170, 0.0014, -0.0180, 0.0043, 0.0129, + -0.0302, 0.0087], device='cuda:0'), grad: tensor([ 7.7067e-08, -1.5342e-04, 6.7316e-06, 3.0827e-06, 1.8524e-06, + -2.0431e-08, 1.8126e-07, 1.3661e-04, 1.5581e-06, 3.2503e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 218.03, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4690 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.14 lr 0.00010000 +Epoch 296, weight, value: tensor([[ 0.0196, -0.0833, -0.0037, ..., -0.1227, -0.1537, -0.1374], + [-0.1307, 0.0763, -0.1283, ..., -0.1461, -0.1560, -0.1974], + [-0.1043, -0.1700, -0.1815, ..., -0.2196, -0.1558, 0.1694], + ..., + [ 0.0822, -0.0649, -0.1275, ..., 0.0983, -0.1434, -0.1305], + [ 0.1209, 0.0970, 0.0811, ..., -0.2032, -0.1156, 0.0816], + [ 0.0648, 0.0320, 0.1056, ..., 0.0495, -0.0909, 0.0016]], + device='cuda:0'), grad: tensor([[-2.3644e-07, 6.1700e-09, -6.9616e-08, ..., 8.7311e-09, + -2.7649e-08, 3.4925e-10], + [ 5.4657e-08, -1.7462e-10, 2.5262e-08, ..., 3.5856e-08, + 9.3132e-09, 4.3947e-08], + [ 5.9663e-08, 7.8639e-08, 4.0454e-08, ..., 1.0710e-08, + 7.2177e-09, 1.0565e-07], + ..., + [-1.7462e-08, 2.6135e-08, 7.1013e-09, ..., -3.9639e-08, + 3.0850e-09, 1.0594e-08], + [-8.6799e-07, -1.4380e-06, -5.7835e-07, ..., 7.9162e-09, + 2.7358e-09, -2.1160e-06], + [ 5.3784e-07, 8.3353e-07, 3.0594e-07, ..., 5.3900e-08, + 1.2200e-07, 1.3392e-06]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0321, -0.0264, -0.0022, 0.0171, 0.0014, -0.0180, 0.0043, 0.0128, + -0.0302, 0.0087], device='cuda:0'), grad: tensor([-1.5162e-06, 2.5355e-07, 3.4692e-07, 8.6380e-08, -6.5251e-08, + 8.8103e-07, 1.4231e-06, -7.7591e-08, -4.2692e-06, 2.9542e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 218.02, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4790 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.05 lr 0.00010000 +Epoch 297, weight, value: tensor([[ 0.0196, -0.0834, -0.0038, ..., -0.1230, -0.1537, -0.1377], + [-0.1307, 0.0763, -0.1286, ..., -0.1463, -0.1563, -0.1976], + [-0.1047, -0.1708, -0.1820, ..., -0.2196, -0.1559, 0.1694], + ..., + [ 0.0823, -0.0649, -0.1278, ..., 0.0985, -0.1444, -0.1305], + [ 0.1209, 0.0970, 0.0813, ..., -0.2032, -0.1156, 0.0817], + [ 0.0647, 0.0322, 0.1057, ..., 0.0495, -0.0909, 0.0016]], + device='cuda:0'), grad: tensor([[-2.7008e-08, 2.2119e-09, 4.0745e-10, ..., 1.5716e-09, + 4.2492e-09, 4.4238e-09], + [ 2.9104e-09, -4.8953e-08, 4.0745e-10, ..., 5.1223e-09, + 2.4447e-09, 2.6193e-09], + [ 2.2701e-09, 1.0477e-08, 4.6566e-10, ..., 1.3970e-09, + 1.6880e-09, -8.4401e-09], + ..., + [ 6.4028e-10, 2.0140e-08, 2.2119e-09, ..., 1.5134e-08, + 1.0652e-08, 4.8894e-09], + [ 3.8417e-09, 2.0314e-08, 8.7311e-10, ..., 5.5879e-09, + 1.8161e-08, 2.4214e-08], + [ 5.0641e-09, -1.7870e-08, -1.2282e-08, ..., 6.2922e-08, + 8.9814e-08, 2.9861e-08]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0321, -0.0264, -0.0022, 0.0173, 0.0014, -0.0181, 0.0043, 0.0129, + -0.0301, 0.0087], device='cuda:0'), grad: tensor([-7.3051e-08, -1.3853e-07, 2.2643e-08, 1.8976e-08, -1.3411e-07, + -1.4401e-07, 6.4436e-08, 7.5845e-08, 1.6205e-07, 1.7323e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 217.61, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4632 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.11 lr 0.00010000 +Epoch 298, weight, value: tensor([[ 0.0199, -0.0834, -0.0037, ..., -0.1234, -0.1538, -0.1378], + [-0.1308, 0.0764, -0.1288, ..., -0.1467, -0.1562, -0.1980], + [-0.1053, -0.1723, -0.1847, ..., -0.2197, -0.1559, 0.1694], + ..., + [ 0.0824, -0.0650, -0.1279, ..., 0.0990, -0.1447, -0.1300], + [ 0.1209, 0.0971, 0.0814, ..., -0.2033, -0.1156, 0.0818], + [ 0.0643, 0.0322, 0.1058, ..., 0.0495, -0.0909, 0.0015]], + device='cuda:0'), grad: tensor([[ 3.1432e-09, 2.0955e-09, 7.5670e-10, ..., 6.1700e-09, + 2.9104e-10, 3.4925e-10], + [ 1.3947e-07, -1.5658e-08, 1.3388e-09, ..., 3.4529e-07, + 7.6252e-09, 4.9477e-09], + [ 3.7486e-08, 4.2492e-09, 5.2387e-10, ..., 7.6019e-08, + 1.7462e-10, -4.7730e-09], + ..., + [-3.8254e-07, 3.1083e-08, 1.4144e-08, ..., -7.7067e-07, + 4.2492e-09, 3.3760e-09], + [ 1.6589e-08, 7.7998e-09, 2.3865e-09, ..., 3.2538e-08, + 3.4925e-10, -1.1642e-10], + [ 1.4924e-07, -3.8766e-08, -3.2072e-08, ..., 2.8731e-07, + 1.9674e-08, 7.7416e-09]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0321, -0.0264, -0.0023, 0.0171, 0.0014, -0.0178, 0.0042, 0.0130, + -0.0301, 0.0086], device='cuda:0'), grad: tensor([ 1.8161e-08, 6.6403e-07, 1.5367e-07, 1.4377e-08, 2.5204e-08, + 2.1129e-08, 1.0186e-08, -1.6047e-06, 8.9873e-08, 6.0024e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 217.82, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4733 re_mapping 0.0027 re_causal 0.0092 /// teacc 99.04 lr 0.00010000 +Epoch 299, weight, value: tensor([[ 0.0201, -0.0835, -0.0038, ..., -0.1236, -0.1539, -0.1381], + [-0.1308, 0.0763, -0.1293, ..., -0.1473, -0.1567, -0.1983], + [-0.1056, -0.1725, -0.1850, ..., -0.2197, -0.1559, 0.1695], + ..., + [ 0.0825, -0.0650, -0.1283, ..., 0.0995, -0.1454, -0.1312], + [ 0.1210, 0.0971, 0.0816, ..., -0.2033, -0.1156, 0.0819], + [ 0.0643, 0.0322, 0.1059, ..., 0.0494, -0.0910, 0.0015]], + device='cuda:0'), grad: tensor([[ 1.4086e-08, 5.0641e-09, 1.7637e-08, ..., 4.8429e-08, + 4.1095e-08, 4.8487e-08], + [ 4.6333e-08, -5.2387e-09, 1.1001e-08, ..., 4.8149e-07, + 3.5565e-08, 2.6543e-08], + [ 2.0373e-08, 1.3039e-08, 4.8894e-09, ..., 1.2396e-06, + 9.8953e-09, 1.0536e-08], + ..., + [-1.2456e-07, 1.9209e-08, 1.5483e-08, ..., -2.2743e-06, + 2.3341e-08, 1.0594e-08], + [ 2.6077e-08, -2.1828e-08, 6.4785e-08, ..., 1.5344e-07, + 2.2980e-07, 2.4145e-07], + [ 4.8894e-09, -5.9372e-08, -6.7928e-08, ..., 1.9569e-07, + 9.3656e-08, 4.5169e-08]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0321, -0.0264, -0.0020, 0.0170, 0.0013, -0.0177, 0.0041, 0.0129, + -0.0301, 0.0085], device='cuda:0'), grad: tensor([ 2.5285e-07, 1.0692e-06, 2.7120e-06, 1.2957e-07, -8.4168e-08, + 3.8976e-07, -9.6392e-07, -4.8615e-06, 1.0235e-06, 3.3271e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 217.71, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4681 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.06 lr 0.00010000 +Epoch 300, weight, value: tensor([[ 0.0202, -0.0835, -0.0037, ..., -0.1239, -0.1541, -0.1384], + [-0.1308, 0.0763, -0.1298, ..., -0.1477, -0.1591, -0.1988], + [-0.1049, -0.1726, -0.1852, ..., -0.2197, -0.1559, 0.1696], + ..., + [ 0.0825, -0.0650, -0.1288, ..., 0.0996, -0.1464, -0.1316], + [ 0.1210, 0.0972, 0.0819, ..., -0.2034, -0.1156, 0.0819], + [ 0.0644, 0.0328, 0.1061, ..., 0.0496, -0.0910, 0.0016]], + device='cuda:0'), grad: tensor([[-4.4878e-08, -7.5670e-10, 2.0256e-08, ..., 4.5227e-08, + 1.8044e-09, 1.1816e-08], + [ 8.6147e-09, -1.9150e-08, 1.0768e-08, ..., 4.2666e-08, + 9.2550e-09, 1.3853e-08], + [ 5.2969e-09, 9.4878e-09, 3.8417e-09, ..., 8.4983e-09, + 3.4925e-10, -6.5193e-09], + ..., + [ 5.1223e-09, 9.7847e-08, 9.5752e-08, ..., 2.0815e-07, + 1.7462e-09, 4.9477e-08], + [ 5.4482e-08, 2.9569e-08, 3.8242e-08, ..., 8.6555e-08, + 2.5029e-09, 1.8801e-08], + [-2.0920e-07, -5.8161e-07, -7.5204e-07, ..., -1.4827e-06, + 9.5810e-08, -3.0454e-07]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0320, -0.0265, -0.0020, 0.0170, 0.0012, -0.0177, 0.0042, 0.0129, + -0.0301, 0.0086], device='cuda:0'), grad: tensor([-2.7521e-07, 4.3481e-08, 3.8999e-09, 8.5565e-08, 2.1681e-06, + 7.9221e-08, 2.3050e-08, 4.6915e-07, 4.1258e-07, -3.0063e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 217.63, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4546 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.08 lr 0.00010000 +Epoch 301, weight, value: tensor([[ 0.0194, -0.0839, -0.0046, ..., -0.1262, -0.1544, -0.1397], + [-0.1310, 0.0763, -0.1308, ..., -0.1505, -0.1593, -0.1992], + [-0.1060, -0.1733, -0.1859, ..., -0.2198, -0.1559, 0.1696], + ..., + [ 0.0827, -0.0651, -0.1295, ..., 0.1004, -0.1467, -0.1319], + [ 0.1211, 0.0977, 0.0841, ..., -0.2034, -0.1156, 0.0823], + [ 0.0640, 0.0324, 0.1059, ..., 0.0496, -0.0911, 0.0014]], + device='cuda:0'), grad: tensor([[ 7.1013e-09, 6.9849e-09, 1.1991e-08, ..., 1.3271e-08, + 9.0629e-08, 6.8685e-09], + [ 4.1735e-08, -3.0617e-08, 2.6193e-09, ..., 6.8918e-08, + 1.9791e-09, 5.2969e-09], + [-4.8836e-08, 1.8568e-08, 2.0373e-09, ..., 1.3446e-08, + 5.8208e-10, -2.1071e-07], + ..., + [-1.5693e-07, 5.6694e-08, 2.8405e-08, ..., -2.6659e-07, + 5.8208e-11, 1.1409e-08], + [ 7.0664e-08, 8.6729e-09, 3.0850e-09, ..., 3.0443e-08, + 1.3330e-08, 1.9651e-07], + [ 3.4925e-08, -5.7975e-08, -5.1339e-08, ..., -2.9744e-08, + 1.3388e-09, 5.2387e-10]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0314, -0.0266, -0.0020, 0.0167, 0.0012, -0.0178, 0.0042, 0.0132, + -0.0299, 0.0085], device='cuda:0'), grad: tensor([ 3.7206e-07, 1.1257e-07, -1.7602e-06, 1.2270e-07, 1.6624e-07, + 5.8860e-07, -9.3365e-07, -5.4389e-07, 1.8487e-06, 2.4447e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 217.62, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4709 re_mapping 0.0028 re_causal 0.0095 /// teacc 99.00 lr 0.00010000 +Epoch 302, weight, value: tensor([[ 0.0193, -0.0841, -0.0051, ..., -0.1266, -0.1546, -0.1404], + [-0.1311, 0.0761, -0.1322, ..., -0.1515, -0.1611, -0.1998], + [-0.1064, -0.1740, -0.1863, ..., -0.2198, -0.1560, 0.1696], + ..., + [ 0.0825, -0.0653, -0.1318, ..., 0.0998, -0.1501, -0.1328], + [ 0.1212, 0.0978, 0.0846, ..., -0.2035, -0.1158, 0.0823], + [ 0.0654, 0.0336, 0.1062, ..., 0.0499, -0.0910, 0.0015]], + device='cuda:0'), grad: tensor([[ 1.3097e-08, 1.7491e-08, 7.6543e-09, ..., 2.0460e-08, + 4.1036e-09, 9.5461e-09], + [ 1.5309e-08, -8.1817e-07, 2.5902e-09, ..., 2.6834e-08, + 4.3656e-10, -6.8452e-08], + [ 1.1525e-08, 1.9593e-07, 1.2224e-09, ..., 7.6252e-09, + 2.9104e-10, 6.8103e-09], + ..., + [ 1.2689e-07, 2.8242e-07, 9.5810e-08, ..., 1.6880e-07, + 2.9104e-11, 5.5006e-08], + [ 7.5379e-09, 2.0757e-07, 2.4738e-09, ..., 1.1030e-08, + 3.4343e-09, 1.8714e-08], + [-2.1292e-07, 9.6043e-10, -1.2759e-07, ..., -2.9197e-07, + 1.7462e-10, -3.6642e-08]], device='cuda:0') +Epoch 302, bias, value: tensor([ 0.0313, -0.0267, -0.0020, 0.0165, 0.0012, -0.0179, 0.0044, 0.0130, + -0.0299, 0.0089], device='cuda:0'), grad: tensor([ 2.9430e-07, -2.4755e-06, 8.2934e-07, 1.8058e-06, 5.7509e-07, + -9.9167e-06, 2.1216e-06, 1.4901e-06, 2.1029e-06, 3.1684e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 217.65, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4740 re_mapping 0.0029 re_causal 0.0091 /// teacc 98.97 lr 0.00010000 +Epoch 303, weight, value: tensor([[ 0.0193, -0.0843, -0.0052, ..., -0.1269, -0.1551, -0.1411], + [-0.1311, 0.0761, -0.1329, ..., -0.1520, -0.1620, -0.2005], + [-0.1074, -0.1756, -0.1876, ..., -0.2199, -0.1560, 0.1696], + ..., + [ 0.0825, -0.0654, -0.1329, ..., 0.0998, -0.1526, -0.1332], + [ 0.1213, 0.0980, 0.0849, ..., -0.2035, -0.1158, 0.0824], + [ 0.0659, 0.0337, 0.1064, ..., 0.0500, -0.0911, 0.0015]], + device='cuda:0'), grad: tensor([[ 4.0163e-09, 5.7044e-09, 4.3656e-09, ..., 6.4028e-09, + 6.1700e-09, 7.3924e-09], + [ 1.7288e-08, -1.7462e-09, 1.6298e-09, ..., 2.7067e-08, + 4.8312e-09, 5.0641e-09], + [ 1.3388e-08, 2.9104e-09, 8.7311e-10, ..., 1.9558e-08, + 1.9791e-09, -2.0023e-08], + ..., + [-1.0012e-07, 6.4611e-09, 4.0163e-09, ..., -1.3760e-07, + 4.2492e-09, 7.0431e-09], + [ 4.7730e-08, 6.2631e-08, 5.1805e-08, ..., 5.1339e-08, + 3.6671e-09, 4.6275e-08], + [-1.2398e-08, -9.9244e-08, -8.7661e-08, ..., 4.7148e-08, + 3.7951e-08, -4.8720e-08]], device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0310, -0.0268, -0.0021, 0.0168, 0.0011, -0.0174, 0.0040, 0.0129, + -0.0299, 0.0090], device='cuda:0'), grad: tensor([ 4.0513e-08, 9.0280e-08, -1.1583e-08, 7.8056e-08, -7.1898e-07, + -2.6659e-07, 9.0292e-07, -3.3434e-07, 3.3295e-07, -8.9465e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 217.30, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4804 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.10 lr 0.00010000 +Epoch 304, weight, value: tensor([[ 0.0191, -0.0844, -0.0054, ..., -0.1275, -0.1554, -0.1416], + [-0.1313, 0.0760, -0.1339, ..., -0.1526, -0.1623, -0.2008], + [-0.1083, -0.1758, -0.1878, ..., -0.2199, -0.1560, 0.1697], + ..., + [ 0.0827, -0.0653, -0.1332, ..., 0.1000, -0.1527, -0.1333], + [ 0.1212, 0.0979, 0.0850, ..., -0.2035, -0.1158, 0.0824], + [ 0.0662, 0.0340, 0.1066, ..., 0.0501, -0.0911, 0.0015]], + device='cuda:0'), grad: tensor([[ 2.8568e-07, 1.3388e-09, 8.8010e-08, ..., 3.4925e-10, + 3.0920e-07, 4.8708e-07], + [ 9.1270e-08, 1.9616e-08, 2.7474e-08, ..., 7.3342e-09, + 5.7393e-08, 1.3737e-07], + [-8.4459e-08, 2.0547e-08, 1.0477e-08, ..., 5.2387e-09, + 9.8371e-09, -1.6252e-07], + ..., + [ 2.1362e-08, 2.0955e-08, 8.5565e-09, ..., -1.9034e-08, + 9.3132e-10, 5.7509e-08], + [ 9.6951e-07, -1.8370e-07, 2.1351e-07, ..., 2.1537e-09, + 1.0598e-06, 1.6708e-06], + [ 3.0093e-08, 1.7521e-08, 1.1874e-08, ..., 2.0955e-09, + 2.1129e-08, 4.1327e-08]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0308, -0.0269, -0.0021, 0.0160, 0.0011, -0.0170, 0.0040, 0.0133, + -0.0300, 0.0090], device='cuda:0'), grad: tensor([ 2.2799e-06, 7.4971e-07, -6.6217e-07, 1.7406e-06, 2.1001e-07, + 1.2876e-07, -1.3642e-05, 3.6042e-07, 8.4788e-06, 3.5320e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 217.76, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4816 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.19 lr 0.00010000 +Epoch 305, weight, value: tensor([[ 0.0204, -0.0843, -0.0057, ..., -0.1277, -0.1584, -0.1448], + [-0.1313, 0.0761, -0.1350, ..., -0.1531, -0.1631, -0.2012], + [-0.1088, -0.1773, -0.1884, ..., -0.2199, -0.1560, 0.1697], + ..., + [ 0.0828, -0.0654, -0.1339, ..., 0.1000, -0.1544, -0.1336], + [ 0.1212, 0.0979, 0.0853, ..., -0.2036, -0.1159, 0.0823], + [ 0.0687, 0.0357, 0.1090, ..., 0.0519, -0.0891, 0.0031]], + device='cuda:0'), grad: tensor([[ 1.2980e-08, 1.4901e-08, 5.2387e-09, ..., 2.1770e-08, + 3.3760e-09, 1.2689e-08], + [ 1.5693e-07, 6.1933e-08, 2.8114e-08, ..., 2.1048e-07, + 4.5402e-09, 1.4319e-07], + [ 2.0082e-07, -4.0350e-07, 1.1642e-08, ..., 2.4145e-07, + 2.1537e-09, -5.9837e-07], + ..., + [-2.0750e-06, 5.2678e-08, 3.3062e-08, ..., -3.3267e-06, + 5.2969e-09, 2.3341e-08], + [-2.4936e-07, -1.8440e-07, -4.2934e-07, ..., 8.9000e-08, + -6.9849e-10, 3.6613e-08], + [ 1.5469e-06, -3.7428e-08, -1.4016e-07, ..., 2.5891e-06, + 3.9290e-08, 5.7451e-08]], device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0291, -0.0269, -0.0021, 0.0152, -0.0006, -0.0167, 0.0043, 0.0133, + -0.0300, 0.0110], device='cuda:0'), grad: tensor([ 1.2759e-07, 1.1660e-06, -1.6866e-06, 1.0543e-06, 3.5856e-07, + 4.9360e-07, 5.5367e-07, -9.5814e-06, 2.8696e-08, 7.4729e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 217.68, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4642 re_mapping 0.0026 re_causal 0.0087 /// teacc 99.04 lr 0.00010000 +Epoch 306, weight, value: tensor([[ 0.0215, -0.0845, -0.0059, ..., -0.1280, -0.1583, -0.1449], + [-0.1314, 0.0762, -0.1356, ..., -0.1536, -0.1635, -0.2016], + [-0.1091, -0.1779, -0.1891, ..., -0.2200, -0.1561, 0.1698], + ..., + [ 0.0829, -0.0655, -0.1348, ..., 0.1001, -0.1555, -0.1342], + [ 0.1214, 0.0982, 0.0866, ..., -0.2036, -0.1156, 0.0829], + [ 0.0691, 0.0362, 0.1095, ..., 0.0522, -0.0888, 0.0034]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 4.3656e-09, 1.3388e-09, ..., 4.0163e-09, + 1.3388e-09, 1.8044e-09], + [ 3.3760e-09, -8.2597e-08, 1.9791e-09, ..., 5.1805e-09, + 5.8208e-10, 1.1583e-08], + [ 2.7940e-09, 1.0128e-08, 2.2119e-09, ..., 1.1642e-09, + 4.0745e-10, -2.8173e-08], + ..., + [ 7.9453e-08, 7.3283e-08, 3.7719e-08, ..., 1.4005e-07, + 7.5670e-10, 1.5309e-08], + [-1.0885e-08, -5.0641e-09, -1.0594e-08, ..., 4.2492e-09, + 1.1642e-09, -5.7626e-09], + [-1.3690e-07, -7.3633e-08, -6.4087e-08, ..., -2.3190e-07, + 4.5402e-09, -2.0955e-09]], device='cuda:0') +Epoch 306, bias, value: tensor([ 0.0295, -0.0269, -0.0020, 0.0152, -0.0010, -0.0168, 0.0040, 0.0133, + -0.0299, 0.0114], device='cuda:0'), grad: tensor([ 2.5320e-08, -2.8173e-07, -4.2084e-08, 3.8010e-08, 2.3935e-07, + -9.7381e-08, 4.1095e-08, 4.0187e-07, 5.6520e-08, -3.7323e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 217.85, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4443 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.04 lr 0.00010000 +Epoch 307, weight, value: tensor([[ 0.0200, -0.0868, -0.0095, ..., -0.1286, -0.1590, -0.1481], + [-0.1315, 0.0762, -0.1362, ..., -0.1541, -0.1639, -0.2019], + [-0.1098, -0.1788, -0.1898, ..., -0.2200, -0.1561, 0.1698], + ..., + [ 0.0828, -0.0658, -0.1378, ..., 0.0997, -0.1566, -0.1349], + [ 0.1215, 0.0987, 0.0885, ..., -0.2036, -0.1153, 0.0834], + [ 0.0696, 0.0366, 0.1096, ..., 0.0524, -0.0888, 0.0034]], + device='cuda:0'), grad: tensor([[-4.5111e-08, 8.1491e-10, 5.8208e-11, ..., -3.3760e-09, + 5.1875e-07, 5.2433e-07], + [ 1.5832e-08, -3.1432e-09, 3.2014e-09, ..., 1.3271e-08, + 9.1386e-09, 2.4447e-07], + [ 6.6939e-09, 3.3178e-09, 8.1491e-10, ..., 3.5507e-09, + 2.7940e-09, -3.1595e-07], + ..., + [-2.2119e-09, 1.4552e-08, 1.1350e-08, ..., 4.1327e-09, + 4.6566e-10, 3.1432e-08], + [-5.0059e-09, -9.8953e-09, -5.7044e-09, ..., 4.4238e-09, + 1.4133e-07, 1.4226e-07], + [ 1.1001e-08, -1.6473e-08, -1.7870e-08, ..., -4.0687e-08, + 2.6776e-09, 5.4715e-09]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0284, -0.0269, -0.0021, 0.0152, -0.0010, -0.0169, 0.0040, 0.0131, + -0.0297, 0.0115], device='cuda:0'), grad: tensor([ 1.4110e-06, 6.5612e-07, -7.7300e-07, 1.8103e-08, 9.8662e-08, + 2.0524e-07, -2.1998e-06, 9.8196e-08, 4.5868e-07, 3.1025e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 217.60, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4590 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.14 lr 0.00010000 +Epoch 308, weight, value: tensor([[ 0.0207, -0.0875, -0.0105, ..., -0.1286, -0.1589, -0.1486], + [-0.1315, 0.0765, -0.1366, ..., -0.1543, -0.1643, -0.2026], + [-0.1098, -0.1793, -0.1901, ..., -0.2200, -0.1561, 0.1699], + ..., + [ 0.0828, -0.0661, -0.1386, ..., 0.0998, -0.1574, -0.1357], + [ 0.1216, 0.0991, 0.0895, ..., -0.2037, -0.1151, 0.0835], + [ 0.0696, 0.0363, 0.1097, ..., 0.0524, -0.0889, 0.0033]], + device='cuda:0'), grad: tensor([[-4.6566e-09, 7.7998e-09, 3.4925e-10, ..., 6.4028e-10, + 5.8790e-09, 4.5402e-09], + [ 2.3923e-08, -1.9267e-08, 3.4925e-10, ..., 3.5274e-08, + 5.8208e-10, 7.5670e-10], + [ 5.1805e-09, 3.8417e-09, 8.7311e-10, ..., 5.9372e-09, + 6.4028e-10, 1.6880e-09], + ..., + [-4.7323e-08, 8.3121e-08, 2.3923e-08, ..., -6.4785e-08, + 5.8790e-09, 4.9477e-09], + [ 1.3504e-08, 2.4622e-08, 1.5716e-09, ..., 2.1013e-08, + 4.2492e-09, 3.2014e-09], + [ 3.7835e-09, -9.3132e-10, -7.6834e-09, ..., 1.1933e-08, + 8.6729e-09, 4.0163e-09]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0289, -0.0268, -0.0020, 0.0147, -0.0010, -0.0161, 0.0035, 0.0130, + -0.0296, 0.0114], device='cuda:0'), grad: tensor([ 4.1502e-08, 2.6834e-08, 3.2363e-08, 8.2003e-07, -1.2224e-09, + -1.7965e-06, 5.1688e-07, 5.1339e-08, 2.5611e-07, 7.3051e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 217.67, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4749 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.16 lr 0.00010000 +Epoch 309, weight, value: tensor([[ 0.0210, -0.0881, -0.0112, ..., -0.1286, -0.1585, -0.1487], + [-0.1316, 0.0766, -0.1371, ..., -0.1550, -0.1646, -0.2028], + [-0.1106, -0.1799, -0.1909, ..., -0.2200, -0.1560, 0.1700], + ..., + [ 0.0829, -0.0661, -0.1390, ..., 0.0999, -0.1582, -0.1360], + [ 0.1216, 0.0991, 0.0891, ..., -0.2037, -0.1153, 0.0835], + [ 0.0696, 0.0363, 0.1097, ..., 0.0523, -0.0889, 0.0033]], + device='cuda:0'), grad: tensor([[-1.6356e-08, 9.8953e-10, 1.3388e-09, ..., -6.4611e-09, + 1.1059e-09, 8.7894e-09], + [ 1.6461e-07, 3.3178e-09, 8.7311e-10, ..., 2.0431e-08, + 8.7311e-10, 2.4983e-07], + [-2.2934e-07, -3.4750e-08, 1.7462e-10, ..., 1.0303e-08, + 4.6566e-10, -6.8638e-07], + ..., + [ 4.0804e-08, 5.2387e-09, 5.4133e-09, ..., -4.0163e-08, + 1.4552e-09, 1.4342e-07], + [ 2.8813e-08, 2.6717e-08, 3.1432e-08, ..., 1.1059e-08, + 2.4447e-09, 1.8976e-07], + [ 1.2224e-09, -1.1059e-08, -2.6252e-08, ..., 4.8894e-09, + 2.1362e-08, 2.7067e-08]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0291, -0.0268, -0.0019, 0.0153, -0.0010, -0.0160, 0.0034, 0.0130, + -0.0297, 0.0113], device='cuda:0'), grad: tensor([-1.4412e-07, 9.8161e-07, -2.2873e-06, 1.3644e-07, 1.5460e-07, + -1.7346e-07, 4.2957e-08, 4.3400e-07, 7.7905e-07, 9.9011e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 217.71, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4569 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 +Epoch 310, weight, value: tensor([[ 0.0211, -0.0883, -0.0114, ..., -0.1288, -0.1587, -0.1492], + [-0.1315, 0.0771, -0.1377, ..., -0.1557, -0.1650, -0.2033], + [-0.1112, -0.1804, -0.1918, ..., -0.2199, -0.1561, 0.1701], + ..., + [ 0.0830, -0.0664, -0.1396, ..., 0.1000, -0.1587, -0.1373], + [ 0.1216, 0.0991, 0.0894, ..., -0.2038, -0.1153, 0.0835], + [ 0.0694, 0.0357, 0.1098, ..., 0.0524, -0.0890, 0.0033]], + device='cuda:0'), grad: tensor([[-1.9441e-08, 5.9954e-09, 1.6880e-09, ..., 1.1642e-10, + -1.2747e-08, 6.0536e-09], + [ 1.6706e-08, 2.3225e-08, 1.1350e-08, ..., 9.8953e-10, + 5.1805e-09, 2.9919e-08], + [ 6.5193e-09, 1.7462e-08, 4.7148e-09, ..., 1.7462e-10, + 2.9686e-09, -1.4552e-09], + ..., + [ 1.0827e-08, 2.4738e-08, 7.5670e-09, ..., 1.9209e-09, + 3.0268e-09, 2.3865e-08], + [-1.1083e-07, -2.6380e-07, -8.5973e-08, ..., 1.1642e-10, + -1.3155e-08, -2.0827e-07], + [ 3.0966e-08, 6.3912e-08, 2.2468e-08, ..., 2.7940e-09, + 1.3213e-08, 5.7451e-08]], device='cuda:0') +Epoch 310, bias, value: tensor([ 0.0290, -0.0266, -0.0018, 0.0152, -0.0010, -0.0161, 0.0034, 0.0129, + -0.0297, 0.0112], device='cuda:0'), grad: tensor([-1.0466e-07, 8.3179e-08, 9.6625e-09, 1.4598e-07, 5.8208e-09, + 7.6951e-08, 1.3970e-07, 1.0116e-07, -6.8173e-07, 2.4075e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 217.56, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4428 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.07 lr 0.00010000 +Epoch 311, weight, value: tensor([[ 0.0213, -0.0883, -0.0115, ..., -0.1289, -0.1595, -0.1505], + [-0.1317, 0.0769, -0.1393, ..., -0.1569, -0.1668, -0.2068], + [-0.1120, -0.1805, -0.1922, ..., -0.2200, -0.1561, 0.1703], + ..., + [ 0.0831, -0.0664, -0.1400, ..., 0.1003, -0.1591, -0.1379], + [ 0.1216, 0.0993, 0.0898, ..., -0.2038, -0.1155, 0.0835], + [ 0.0695, 0.0360, 0.1098, ..., 0.0525, -0.0889, 0.0033]], + device='cuda:0'), grad: tensor([[-8.7311e-10, 7.2876e-08, 3.9581e-09, ..., 2.0373e-09, + 6.7870e-08, 3.8417e-09], + [ 8.9640e-09, -5.3570e-06, 1.2515e-08, ..., 1.2864e-08, + -5.2825e-06, -2.1106e-07], + [ 3.6671e-09, 5.0291e-06, 4.4820e-09, ..., 2.0955e-09, + 4.9509e-06, 2.0349e-07], + ..., + [ 3.2596e-08, 1.8126e-07, 7.5204e-08, ..., 6.5716e-08, + 8.1491e-08, 1.2806e-08], + [-6.1467e-08, -1.3015e-07, -5.7800e-08, ..., 8.5565e-09, + 5.2270e-08, -9.9652e-08], + [-2.2992e-08, -4.7730e-09, -2.4971e-08, ..., -9.4413e-08, + 3.0675e-08, 4.1444e-08]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0287, -0.0268, -0.0017, 0.0147, -0.0011, -0.0160, 0.0036, 0.0131, + -0.0297, 0.0112], device='cuda:0'), grad: tensor([ 7.5763e-07, -5.8055e-05, 5.4449e-05, 2.1104e-06, 3.9674e-07, + -2.5164e-06, 3.9814e-07, 1.1669e-06, 1.1623e-06, 2.5914e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 217.80, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4566 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.07 lr 0.00010000 +Epoch 312, weight, value: tensor([[ 0.0217, -0.0886, -0.0119, ..., -0.1293, -0.1596, -0.1508], + [-0.1318, 0.0800, -0.1379, ..., -0.1580, -0.1660, -0.2034], + [-0.1123, -0.1856, -0.1937, ..., -0.2201, -0.1565, 0.1687], + ..., + [ 0.0832, -0.0666, -0.1410, ..., 0.1004, -0.1602, -0.1384], + [ 0.1217, 0.0974, 0.0872, ..., -0.2038, -0.1156, 0.0836], + [ 0.0696, 0.0361, 0.1099, ..., 0.0525, -0.0890, 0.0033]], + device='cuda:0'), grad: tensor([[ 5.1397e-08, 1.4598e-07, 7.5146e-08, ..., 2.7707e-08, + 2.2002e-08, 9.5577e-08], + [ 7.1304e-08, 1.8103e-07, 9.8255e-08, ..., 4.6683e-08, + 3.6904e-08, 1.5402e-07], + [ 7.8348e-08, 2.4145e-07, 9.7090e-08, ..., 1.5774e-08, + 4.2899e-08, 1.3236e-07], + ..., + [ 6.8627e-08, 1.8941e-07, 1.6019e-07, ..., 1.1653e-07, + 3.3586e-08, 1.2899e-07], + [-3.3970e-07, -1.2517e-06, -4.8848e-07, ..., 4.6100e-08, + -1.9337e-07, -8.7358e-07], + [-8.2131e-08, 8.7079e-08, -1.8138e-07, ..., 1.7866e-05, + 7.4692e-06, 6.9849e-06]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0288, -0.0245, -0.0042, 0.0142, -0.0011, -0.0159, 0.0036, 0.0131, + -0.0306, 0.0113], device='cuda:0'), grad: tensor([ 5.6438e-07, 8.1025e-07, 8.5402e-07, 5.2573e-07, -3.0413e-05, + 6.6299e-08, 1.4831e-07, 1.0533e-06, -4.1574e-06, 3.0577e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 217.59, cls_loss 0.0008 cls_loss_mapping 0.0031 cls_loss_causal 0.4520 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.11 lr 0.00010000 +Epoch 313, weight, value: tensor([[ 0.0218, -0.0889, -0.0123, ..., -0.1294, -0.1597, -0.1512], + [-0.1319, 0.0795, -0.1401, ..., -0.1589, -0.1684, -0.2039], + [-0.1124, -0.1872, -0.1968, ..., -0.2201, -0.1566, 0.1680], + ..., + [ 0.0833, -0.0667, -0.1417, ..., 0.1007, -0.1608, -0.1396], + [ 0.1243, 0.1019, 0.0927, ..., -0.2039, -0.1153, 0.0878], + [ 0.0692, 0.0346, 0.1093, ..., 0.0524, -0.0891, 0.0027]], + device='cuda:0'), grad: tensor([[-9.0455e-08, 3.1432e-09, 1.8626e-09, ..., 7.5670e-10, + 2.9104e-09, 2.8522e-09], + [ 9.8662e-08, -3.0850e-09, 3.8417e-09, ..., 2.5379e-07, + 1.3970e-08, 1.0716e-07], + [ 1.1059e-07, 5.4133e-09, 1.9791e-09, ..., 2.5239e-07, + 1.4144e-08, 9.2608e-08], + ..., + [-1.8300e-07, 1.7928e-08, 9.4296e-09, ..., -5.1316e-07, + -2.7881e-08, -1.9907e-07], + [ 1.0012e-08, 4.5309e-07, 2.7032e-07, ..., 4.0745e-09, + 5.5297e-09, -8.7311e-10], + [ 1.9791e-08, 2.6193e-08, 8.6729e-09, ..., -1.4843e-08, + 9.3132e-10, 2.4447e-09]], device='cuda:0') +Epoch 313, bias, value: tensor([ 0.0288, -0.0239, -0.0060, 0.0125, -0.0010, -0.0185, 0.0035, 0.0131, + -0.0267, 0.0108], device='cuda:0'), grad: tensor([-3.1618e-07, 7.6229e-07, 8.0978e-07, 1.2219e-05, 5.2212e-08, + -1.4201e-05, 4.9768e-08, -1.4277e-06, 1.8720e-06, 1.8789e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 217.72, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4778 re_mapping 0.0026 re_causal 0.0091 /// teacc 99.08 lr 0.00010000 +Epoch 314, weight, value: tensor([[ 0.0219, -0.0890, -0.0124, ..., -0.1296, -0.1600, -0.1517], + [-0.1321, 0.0795, -0.1401, ..., -0.1600, -0.1691, -0.2039], + [-0.1140, -0.1874, -0.1972, ..., -0.2203, -0.1567, 0.1680], + ..., + [ 0.0835, -0.0668, -0.1422, ..., 0.1011, -0.1614, -0.1397], + [ 0.1247, 0.1024, 0.0931, ..., -0.2039, -0.1154, 0.0881], + [ 0.0691, 0.0345, 0.1093, ..., 0.0524, -0.0892, 0.0026]], + device='cuda:0'), grad: tensor([[-1.8277e-08, 6.4028e-10, 0.0000e+00, ..., 9.1386e-09, + -4.3074e-09, 1.1642e-10], + [ 1.0186e-08, 6.9849e-09, 4.6566e-10, ..., 1.2643e-07, + 2.3283e-10, 1.5134e-09], + [ 2.3865e-09, 4.4820e-09, 2.9104e-10, ..., 6.1002e-08, + 1.1642e-10, -2.4447e-09], + ..., + [-9.0047e-08, -1.4773e-07, 3.4925e-10, ..., -2.2743e-06, + 2.3283e-10, 6.9849e-10], + [ 8.3819e-09, -2.3283e-10, -8.1491e-10, ..., 4.5926e-08, + 0.0000e+00, 0.0000e+00], + [ 5.6345e-08, 1.0786e-07, 1.1642e-10, ..., 1.5935e-06, + 2.0955e-09, 9.3132e-10]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0286, -0.0240, -0.0061, 0.0114, -0.0010, -0.0188, 0.0036, 0.0132, + -0.0263, 0.0107], device='cuda:0'), grad: tensor([-7.0955e-08, 2.4866e-07, 1.1554e-07, 4.3120e-07, 3.9861e-07, + 7.4040e-08, 4.2317e-08, -4.4368e-06, 9.2492e-08, 3.1367e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 217.55, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4750 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.04 lr 0.00010000 +Epoch 315, weight, value: tensor([[ 0.0220, -0.0888, -0.0124, ..., -0.1300, -0.1598, -0.1520], + [-0.1322, 0.0794, -0.1402, ..., -0.1611, -0.1707, -0.2040], + [-0.1135, -0.1874, -0.1977, ..., -0.2203, -0.1568, 0.1680], + ..., + [ 0.0836, -0.0669, -0.1427, ..., 0.1016, -0.1610, -0.1403], + [ 0.1247, 0.1025, 0.0932, ..., -0.2040, -0.1154, 0.0884], + [ 0.0692, 0.0347, 0.1094, ..., 0.0524, -0.0893, 0.0026]], + device='cuda:0'), grad: tensor([[ 4.8894e-09, 1.2980e-08, 7.0431e-09, ..., 1.3039e-08, + 7.5670e-08, 4.9593e-08], + [ 1.3364e-07, -1.7812e-08, 1.0536e-08, ..., 1.2154e-07, + 2.4040e-08, 1.7753e-08], + [ 8.6147e-09, 3.1432e-08, 2.7358e-09, ..., 8.5565e-09, + 2.6368e-08, -7.1479e-08], + ..., + [-1.4924e-07, 3.1258e-08, 1.9209e-08, ..., -1.0064e-07, + 2.4447e-08, 7.9744e-09], + [-3.6845e-08, -3.7253e-08, -3.9057e-08, ..., 2.2352e-08, + 8.1840e-08, 2.4622e-08], + [-1.1869e-07, -1.6857e-07, -1.5728e-07, ..., -4.8382e-07, + -9.8778e-08, 1.8044e-09]], device='cuda:0') +Epoch 315, bias, value: tensor([ 0.0291, -0.0241, -0.0061, 0.0102, -0.0010, -0.0186, 0.0036, 0.0133, + -0.0263, 0.0107], device='cuda:0'), grad: tensor([ 2.8266e-07, 4.5891e-07, -3.1502e-07, 1.0245e-06, 3.0752e-06, + -5.8189e-06, 2.2855e-06, -2.8522e-07, 2.1630e-07, -9.2294e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 217.32, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4652 re_mapping 0.0026 re_causal 0.0092 /// teacc 99.04 lr 0.00010000 +Epoch 316, weight, value: tensor([[ 0.0221, -0.0889, -0.0126, ..., -0.1308, -0.1600, -0.1523], + [-0.1323, 0.0794, -0.1402, ..., -0.1615, -0.1711, -0.2041], + [-0.1141, -0.1875, -0.1980, ..., -0.2204, -0.1568, 0.1680], + ..., + [ 0.0836, -0.0669, -0.1431, ..., 0.1018, -0.1617, -0.1405], + [ 0.1247, 0.1025, 0.0932, ..., -0.2040, -0.1155, 0.0885], + [ 0.0693, 0.0347, 0.1094, ..., 0.0523, -0.0895, 0.0025]], + device='cuda:0'), grad: tensor([[ 1.9209e-09, 4.2492e-09, 6.4028e-10, ..., 5.9663e-08, + 5.6520e-08, 8.3819e-09], + [ 1.1292e-07, 7.0431e-09, 1.0477e-09, ..., 1.6124e-07, + 2.5495e-08, 3.8883e-08], + [ 6.6683e-07, 1.6252e-07, 5.8208e-11, ..., 9.6578e-07, + 8.2655e-09, 5.0431e-07], + ..., + [-1.0449e-06, 1.3097e-08, 2.0373e-09, ..., -1.4678e-06, + 1.4377e-08, -6.9151e-07], + [ 1.2014e-07, 1.7753e-08, 1.7462e-09, ..., 1.7765e-07, + 8.0327e-09, 8.7777e-08], + [ 3.4343e-08, 4.5984e-09, -8.9058e-09, ..., 9.2294e-07, + 8.5775e-07, 2.4273e-08]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0291, -0.0241, -0.0061, 0.0103, -0.0009, -0.0186, 0.0035, 0.0133, + -0.0263, 0.0106], device='cuda:0'), grad: tensor([ 9.3482e-08, 5.0012e-07, 3.4384e-06, -1.0390e-07, -1.4827e-06, + 1.7637e-08, 7.7067e-08, -4.6417e-06, 6.2119e-07, 1.4929e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 217.57, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4801 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.14 lr 0.00010000 +Epoch 317, weight, value: tensor([[ 0.0224, -0.0891, -0.0128, ..., -0.1313, -0.1602, -0.1527], + [-0.1323, 0.0799, -0.1402, ..., -0.1619, -0.1706, -0.2041], + [-0.1148, -0.1876, -0.1983, ..., -0.2206, -0.1576, 0.1680], + ..., + [ 0.0839, -0.0670, -0.1438, ..., 0.1024, -0.1641, -0.1413], + [ 0.1247, 0.1025, 0.0932, ..., -0.2041, -0.1156, 0.0884], + [ 0.0690, 0.0340, 0.1096, ..., 0.0522, -0.0896, 0.0025]], + device='cuda:0'), grad: tensor([[-2.1397e-07, 8.6147e-07, 3.3225e-07, ..., 1.1092e-06, + 4.2492e-09, 3.2037e-07], + [ 6.2864e-07, 4.0010e-06, 1.5460e-06, ..., 5.2862e-06, + 9.6625e-09, 1.4910e-06], + [ 1.6880e-07, 4.4331e-07, 1.7066e-07, ..., 6.7288e-07, + 2.0955e-09, 1.4331e-07], + ..., + [-8.1724e-07, 2.9188e-06, 1.1260e-06, ..., 2.3842e-06, + 2.2352e-08, 1.0924e-06], + [ 4.0117e-07, 1.5255e-06, 5.8720e-07, ..., 2.1365e-06, + 6.4611e-09, 5.6811e-07], + [-2.5500e-06, -2.7806e-05, -1.1504e-05, ..., -5.4717e-05, + -1.2621e-05, -1.8716e-05]], device='cuda:0') +Epoch 317, bias, value: tensor([ 0.0290, -0.0240, -0.0061, 0.0103, -0.0009, -0.0186, 0.0035, 0.0135, + -0.0264, 0.0104], device='cuda:0'), grad: tensor([ 2.3954e-06, 1.7777e-05, 2.2408e-06, 1.9725e-06, 1.1349e-04, + 1.0896e-06, 5.8580e-07, 8.8289e-06, 7.6964e-06, -1.5604e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 217.83, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4413 re_mapping 0.0026 re_causal 0.0083 /// teacc 99.04 lr 0.00010000 +Epoch 318, weight, value: tensor([[ 0.0229, -0.0892, -0.0130, ..., -0.1318, -0.1605, -0.1532], + [-0.1324, 0.0799, -0.1403, ..., -0.1633, -0.1717, -0.2042], + [-0.1147, -0.1876, -0.1986, ..., -0.2206, -0.1576, 0.1680], + ..., + [ 0.0839, -0.0673, -0.1451, ..., 0.1024, -0.1660, -0.1419], + [ 0.1247, 0.1025, 0.0932, ..., -0.2042, -0.1158, 0.0884], + [ 0.0687, 0.0348, 0.1098, ..., 0.0518, -0.0905, 0.0019]], + device='cuda:0'), grad: tensor([[ 2.2375e-07, 8.2375e-07, 2.3912e-07, ..., 7.5437e-07, + 8.1607e-08, 1.5507e-07], + [ 7.5437e-07, 2.7269e-06, 7.9023e-07, ..., 2.4959e-06, + 9.5752e-08, 4.8708e-07], + [ 1.2550e-07, 4.5542e-07, 1.2410e-07, ..., 4.3027e-07, + 5.1106e-08, 2.6776e-08], + ..., + [ 9.7230e-07, 3.9265e-06, 1.1744e-06, ..., 3.2950e-06, + 6.0070e-08, 6.1933e-07], + [ 6.5472e-07, 2.9225e-06, 7.3155e-07, ..., 2.8163e-06, + 7.4925e-07, 6.9011e-07], + [-3.4142e-06, -1.3471e-05, -3.7234e-06, ..., -1.2472e-05, + -1.6633e-06, -2.5611e-06]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0290, -0.0241, -0.0061, 0.0103, -0.0005, -0.0186, 0.0036, 0.0134, + -0.0264, 0.0100], device='cuda:0'), grad: tensor([ 3.2410e-06, 1.0759e-05, 1.6280e-06, 2.3358e-06, 6.3367e-06, + 2.4419e-06, -5.4191e-08, 1.3985e-05, 1.1757e-05, -5.2422e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 217.61, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4624 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.02 lr 0.00010000 +Epoch 319, weight, value: tensor([[ 0.0239, -0.0894, -0.0132, ..., -0.1321, -0.1607, -0.1533], + [-0.1326, 0.0799, -0.1403, ..., -0.1641, -0.1729, -0.2042], + [-0.1162, -0.1876, -0.1990, ..., -0.2207, -0.1578, 0.1680], + ..., + [ 0.0840, -0.0675, -0.1463, ..., 0.1025, -0.1686, -0.1425], + [ 0.1248, 0.1025, 0.0932, ..., -0.2043, -0.1159, 0.0885], + [ 0.0689, 0.0353, 0.1099, ..., 0.0518, -0.0905, 0.0019]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.8510e-08, 1.7812e-08, ..., 1.7229e-08, + 7.2760e-09, 5.0641e-09], + [ 1.1525e-08, 8.5565e-09, 1.0594e-08, ..., 1.0012e-08, + 1.9209e-09, 2.7940e-09], + [ 1.7462e-09, 4.5344e-08, 7.0431e-09, ..., 1.4552e-09, + 8.1491e-10, 6.9849e-10], + ..., + [ 1.4901e-08, 1.8917e-08, 1.7521e-08, ..., 1.1118e-08, + 1.6880e-09, 1.6880e-09], + [ 2.7940e-07, 2.3108e-07, 2.7427e-07, ..., 2.1490e-07, + 1.3970e-08, -4.0745e-10], + [-3.7532e-07, -3.5856e-07, -3.8138e-07, ..., -3.4133e-07, + -4.1153e-08, -2.5670e-08]], device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0292, -0.0242, -0.0061, 0.0100, -0.0004, -0.0186, 0.0036, 0.0135, + -0.0264, 0.0101], device='cuda:0'), grad: tensor([ 4.9709e-08, 3.6962e-08, 1.2130e-07, -1.4249e-07, 2.0489e-07, + -4.0699e-07, 5.0233e-08, 6.4319e-08, 1.3430e-06, -1.3048e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 217.64, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4642 re_mapping 0.0026 re_causal 0.0091 /// teacc 99.09 lr 0.00010000 +Epoch 320, weight, value: tensor([[ 0.0235, -0.0897, -0.0134, ..., -0.1328, -0.1614, -0.1543], + [-0.1329, 0.0798, -0.1403, ..., -0.1653, -0.1738, -0.2043], + [-0.1179, -0.1877, -0.1995, ..., -0.2208, -0.1579, 0.1681], + ..., + [ 0.0841, -0.0678, -0.1488, ..., 0.1030, -0.1668, -0.1424], + [ 0.1248, 0.1025, 0.0932, ..., -0.2046, -0.1160, 0.0885], + [ 0.0706, 0.0381, 0.1111, ..., 0.0526, -0.0897, 0.0027]], + device='cuda:0'), grad: tensor([[ 7.5612e-08, 1.0227e-07, 6.9616e-08, ..., 6.9267e-08, + 8.1491e-10, 6.4028e-10], + [ 4.7032e-08, 4.7730e-09, 1.1874e-08, ..., 6.9325e-08, + 1.3388e-09, 1.8044e-08], + [ 2.4564e-08, 6.9849e-09, 1.3388e-09, ..., 5.8382e-08, + 1.2224e-09, 1.6589e-08], + ..., + [ 1.6182e-07, 3.3318e-07, 2.2701e-07, ..., 7.3342e-09, + -9.8953e-10, -6.3621e-08], + [ 1.1321e-07, 1.6321e-07, 1.0472e-07, ..., 1.2154e-07, + 1.5134e-09, -2.0955e-09], + [-5.0385e-07, -6.5519e-07, -4.6147e-07, ..., -4.0838e-07, + 1.0361e-08, 1.4028e-08]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0287, -0.0243, -0.0061, 0.0098, -0.0013, -0.0187, 0.0037, 0.0137, + -0.0264, 0.0112], device='cuda:0'), grad: tensor([ 2.6845e-07, 1.8720e-07, 1.6275e-07, -4.8021e-08, 1.8475e-07, + 8.7777e-08, 8.9640e-09, 3.2061e-07, 4.8801e-07, -1.6605e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 217.55, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4428 re_mapping 0.0026 re_causal 0.0090 /// teacc 99.11 lr 0.00010000 +Epoch 321, weight, value: tensor([[ 0.0229, -0.0897, -0.0136, ..., -0.1333, -0.1625, -0.1560], + [-0.1330, 0.0798, -0.1404, ..., -0.1660, -0.1743, -0.2044], + [-0.1193, -0.1878, -0.1998, ..., -0.2209, -0.1579, 0.1681], + ..., + [ 0.0841, -0.0681, -0.1504, ..., 0.1032, -0.1689, -0.1428], + [ 0.1248, 0.1025, 0.0931, ..., -0.2048, -0.1162, 0.0885], + [ 0.0723, 0.0404, 0.1117, ..., 0.0533, -0.0888, 0.0034]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.9104e-09, 3.8417e-09, ..., 4.8894e-09, + 8.0327e-09, 6.6357e-09], + [ 1.4901e-08, 4.6566e-10, 2.0955e-09, ..., 1.3155e-08, + 7.3342e-09, 8.1491e-09], + [ 1.0012e-08, 1.3970e-09, 4.6566e-10, ..., 8.3819e-09, + 1.7462e-09, -1.7812e-08], + ..., + [-2.9104e-09, 2.0489e-08, 2.6426e-08, ..., 6.1700e-09, + 1.1642e-10, 7.6834e-09], + [ 8.3819e-09, 3.4925e-09, 3.9581e-09, ..., 1.0361e-08, + 1.6065e-08, 1.5367e-08], + [-4.8894e-08, -3.2829e-08, -5.1106e-08, ..., -6.0070e-08, + 2.9104e-09, 1.8626e-09]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0279, -0.0243, -0.0061, 0.0097, -0.0018, -0.0187, 0.0038, 0.0138, + -0.0264, 0.0119], device='cuda:0'), grad: tensor([ 3.8650e-08, 7.3691e-08, -1.3970e-08, 1.3271e-08, 6.8569e-08, + 3.7136e-08, -1.0431e-07, 2.8987e-08, 9.4180e-08, -2.1525e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 217.74, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4806 re_mapping 0.0027 re_causal 0.0092 /// teacc 99.14 lr 0.00010000 +Epoch 322, weight, value: tensor([[ 0.0231, -0.0901, -0.0140, ..., -0.1343, -0.1627, -0.1563], + [-0.1342, 0.0794, -0.1404, ..., -0.1672, -0.1751, -0.2045], + [-0.1215, -0.1878, -0.2003, ..., -0.2210, -0.1580, 0.1682], + ..., + [ 0.0842, -0.0687, -0.1526, ..., 0.1026, -0.1702, -0.1446], + [ 0.1247, 0.1024, 0.0931, ..., -0.2050, -0.1163, 0.0885], + [ 0.0748, 0.0419, 0.1120, ..., 0.0531, -0.0893, 0.0030]], + device='cuda:0'), grad: tensor([[ 1.1292e-07, 2.4866e-07, 1.5018e-08, ..., 4.0373e-07, + 1.1572e-07, 2.8359e-07], + [ 1.3341e-07, 1.5716e-07, 1.1222e-07, ..., 7.2061e-08, + 1.4319e-08, 2.0990e-07], + [ 4.7032e-08, 6.4960e-08, 3.2247e-08, ..., 3.9698e-08, + 3.5157e-08, -5.2503e-08], + ..., + [ 1.2224e-08, 9.9768e-08, 3.0734e-08, ..., 5.9954e-08, + 3.2480e-08, 1.8696e-07], + [-1.9744e-06, -2.9262e-06, -2.0210e-06, ..., 3.3993e-08, + 2.5611e-08, -2.7884e-06], + [ 1.3048e-06, 1.8207e-06, 1.5292e-06, ..., -1.2573e-07, + 4.3283e-07, 2.0452e-06]], device='cuda:0') +Epoch 322, bias, value: tensor([ 0.0278, -0.0251, -0.0061, 0.0094, -0.0015, -0.0187, 0.0038, 0.0142, + -0.0264, 0.0122], device='cuda:0'), grad: tensor([ 1.7509e-06, 9.5181e-07, -2.6217e-07, 1.3396e-05, -5.9232e-07, + -1.6570e-05, 3.6117e-06, 9.3225e-07, -9.6262e-06, 6.4224e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 217.89, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4671 re_mapping 0.0027 re_causal 0.0089 /// teacc 99.00 lr 0.00010000 +Epoch 323, weight, value: tensor([[ 0.0235, -0.0902, -0.0141, ..., -0.1346, -0.1623, -0.1559], + [-0.1344, 0.0796, -0.1404, ..., -0.1680, -0.1764, -0.2046], + [-0.1223, -0.1879, -0.2005, ..., -0.2210, -0.1580, 0.1682], + ..., + [ 0.0844, -0.0691, -0.1529, ..., 0.1026, -0.1708, -0.1449], + [ 0.1248, 0.1025, 0.0931, ..., -0.2052, -0.1164, 0.0887], + [ 0.0751, 0.0425, 0.1121, ..., 0.0534, -0.0892, 0.0032]], + device='cuda:0'), grad: tensor([[-1.1967e-07, -2.0023e-08, 0.0000e+00, ..., 6.0536e-09, + 4.6566e-09, -1.0943e-08], + [ 4.4354e-08, -2.3283e-10, 1.1642e-10, ..., 1.8126e-07, + 1.5786e-07, 1.4529e-07], + [ 8.7311e-09, 6.8685e-09, 1.1642e-10, ..., 3.0315e-07, + 4.9127e-07, 4.0652e-07], + ..., + [-1.6042e-07, 3.0734e-08, 9.3132e-10, ..., -2.2072e-07, + 1.0012e-07, 7.0082e-08], + [ 1.9674e-08, 3.0780e-07, 2.3283e-10, ..., 2.3283e-08, + 1.1199e-07, 1.5949e-08], + [ 1.1199e-07, 3.1991e-07, -1.2806e-09, ..., 1.1250e-06, + 8.3866e-07, 6.0024e-07]], device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0293, -0.0251, -0.0061, 0.0098, -0.0016, -0.0185, 0.0029, 0.0142, + -0.0265, 0.0124], device='cuda:0'), grad: tensor([-9.5833e-07, 6.0629e-07, 1.2293e-06, 1.1949e-06, -4.0866e-06, + -2.3656e-06, 1.3290e-06, -7.3900e-07, 9.5554e-07, 2.8368e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 217.63, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4547 re_mapping 0.0026 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +Epoch 324, weight, value: tensor([[ 0.0236, -0.0902, -0.0141, ..., -0.1350, -0.1625, -0.1562], + [-0.1345, 0.0795, -0.1405, ..., -0.1689, -0.1783, -0.2047], + [-0.1226, -0.1879, -0.2007, ..., -0.2211, -0.1581, 0.1689], + ..., + [ 0.0845, -0.0692, -0.1533, ..., 0.1028, -0.1717, -0.1481], + [ 0.1248, 0.1026, 0.0932, ..., -0.2052, -0.1167, 0.0887], + [ 0.0749, 0.0424, 0.1121, ..., 0.0533, -0.0895, 0.0030]], + device='cuda:0'), grad: tensor([[-2.3283e-10, 5.8208e-10, 1.1642e-10, ..., 5.8208e-10, + 2.6776e-09, 1.5832e-08], + [ 6.2981e-08, -5.8208e-10, 4.6566e-10, ..., 6.3912e-08, + 1.6182e-08, 9.0571e-08], + [ 4.1211e-08, 2.2119e-09, 3.4925e-10, ..., 4.0513e-08, + 4.1910e-09, -1.1422e-05], + ..., + [-1.4435e-07, 6.1700e-09, 1.5134e-09, ..., -1.3679e-07, + 2.5844e-08, 1.0997e-05], + [ 1.8626e-08, 4.1910e-09, 8.1491e-10, ..., 2.4098e-08, + 1.4692e-07, 1.2293e-07], + [ 1.6415e-08, 9.3132e-10, -3.4925e-10, ..., 1.5215e-07, + 8.6729e-08, 7.1595e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0294, -0.0252, -0.0055, 0.0093, -0.0015, -0.0186, 0.0030, 0.0137, + -0.0264, 0.0123], device='cuda:0'), grad: tensor([ 9.1153e-08, 6.9523e-07, -6.0946e-05, 9.1875e-07, 1.8366e-06, + -8.1137e-06, 5.4650e-06, 5.8532e-05, 1.2051e-06, 3.8557e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 217.96, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4519 re_mapping 0.0025 re_causal 0.0087 /// teacc 99.10 lr 0.00010000 +Epoch 325, weight, value: tensor([[ 0.0236, -0.0904, -0.0143, ..., -0.1353, -0.1627, -0.1565], + [-0.1346, 0.0795, -0.1405, ..., -0.1694, -0.1788, -0.2047], + [-0.1232, -0.1881, -0.2009, ..., -0.2212, -0.1581, 0.1690], + ..., + [ 0.0846, -0.0693, -0.1536, ..., 0.1031, -0.1725, -0.1486], + [ 0.1248, 0.1025, 0.0931, ..., -0.2056, -0.1179, 0.0885], + [ 0.0750, 0.0433, 0.1123, ..., 0.0533, -0.0892, 0.0031]], + device='cuda:0'), grad: tensor([[-2.3283e-09, 5.6112e-07, 4.0838e-07, ..., 4.4238e-09, + 4.6217e-08, 1.9092e-08], + [ 1.0827e-08, 1.2026e-07, 5.8091e-08, ..., 1.0268e-07, + 6.3563e-08, 1.4086e-08], + [ 2.5611e-09, 3.0268e-08, 1.7579e-08, ..., 7.7998e-09, + 7.4506e-09, 3.6089e-09], + ..., + [-1.3737e-08, 9.6625e-08, 4.0513e-08, ..., 4.1910e-08, + 4.5868e-08, 7.3342e-09], + [ 3.9581e-09, 7.6601e-08, 4.8429e-08, ..., 1.6880e-08, + 4.7614e-08, 2.7358e-08], + [-2.1420e-08, 4.6147e-07, -3.0850e-08, ..., 8.5309e-07, + 6.0629e-07, 8.6147e-08]], device='cuda:0') +Epoch 325, bias, value: tensor([ 0.0293, -0.0252, -0.0054, 0.0089, -0.0015, -0.0185, 0.0029, 0.0137, + -0.0266, 0.0125], device='cuda:0'), grad: tensor([ 3.5372e-06, 7.1991e-07, 1.9313e-07, -7.2233e-06, -1.0682e-06, + 2.1514e-06, -6.8871e-07, 3.3830e-07, 5.6205e-07, 1.4594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 217.85, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4735 re_mapping 0.0027 re_causal 0.0092 /// teacc 99.10 lr 0.00010000 +Epoch 326, weight, value: tensor([[ 0.0241, -0.0903, -0.0145, ..., -0.1355, -0.1656, -0.1598], + [-0.1348, 0.0795, -0.1405, ..., -0.1710, -0.1806, -0.2049], + [-0.1242, -0.1882, -0.2012, ..., -0.2212, -0.1582, 0.1690], + ..., + [ 0.0851, -0.0691, -0.1539, ..., 0.1042, -0.1721, -0.1485], + [ 0.1249, 0.1025, 0.0931, ..., -0.2057, -0.1180, 0.0885], + [ 0.0744, 0.0436, 0.1125, ..., 0.0533, -0.0893, 0.0033]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.6298e-09, 3.4925e-10, ..., 2.4447e-09, + 3.8417e-09, 4.3074e-09], + [ 1.1008e-06, 2.4796e-08, 9.8953e-09, ..., 1.8347e-06, + 9.2480e-07, 1.0906e-06], + [ 5.1921e-08, 1.2107e-08, 9.3132e-10, ..., 8.1258e-08, + 4.3772e-08, 5.2038e-08], + ..., + [-1.2377e-06, 1.0012e-08, 6.9849e-10, ..., -2.2110e-06, + -1.0747e-06, -1.2442e-06], + [-6.9500e-08, -8.9174e-08, -1.7579e-08, ..., 1.4051e-07, + 1.0547e-07, 4.7265e-08], + [ 4.9244e-08, 1.5134e-09, -3.9581e-09, ..., 2.3434e-07, + 2.5425e-07, 2.1816e-07]], device='cuda:0') +Epoch 326, bias, value: tensor([ 0.0267, -0.0253, -0.0055, 0.0087, -0.0016, -0.0183, 0.0038, 0.0143, + -0.0266, 0.0124], device='cuda:0'), grad: tensor([-1.2107e-08, 6.3218e-06, 3.2759e-07, 1.2224e-08, -4.3027e-07, + 1.9709e-07, -1.1327e-07, -7.2531e-06, 1.8743e-08, 9.2201e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 218.12, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4455 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.17 lr 0.00010000 +Epoch 327, weight, value: tensor([[ 0.0250, -0.0906, -0.0141, ..., -0.1357, -0.1656, -0.1598], + [-0.1353, 0.0792, -0.1405, ..., -0.1720, -0.1824, -0.2050], + [-0.1268, -0.1884, -0.2015, ..., -0.2218, -0.1582, 0.1687], + ..., + [ 0.0852, -0.0693, -0.1543, ..., 0.1048, -0.1738, -0.1469], + [ 0.1249, 0.1026, 0.0932, ..., -0.2058, -0.1181, 0.0886], + [ 0.0751, 0.0442, 0.1126, ..., 0.0531, -0.0898, 0.0029]], + device='cuda:0'), grad: tensor([[-5.8208e-10, 2.9104e-09, 3.4925e-10, ..., 3.1432e-09, + 4.5402e-09, 3.6089e-09], + [ 1.3039e-08, 1.4086e-08, 1.3970e-09, ..., 2.8522e-08, + 1.5716e-08, 1.4785e-08], + [ 5.4715e-09, 1.0477e-08, 2.3283e-10, ..., 5.8208e-09, + 2.6776e-09, 8.3819e-09], + ..., + [-1.3388e-08, 1.6880e-08, 6.9849e-10, ..., -2.5611e-08, + 9.5461e-09, 1.1758e-08], + [-2.1770e-08, -5.0291e-08, -5.7044e-09, ..., 7.5670e-09, + 3.0035e-08, -3.1898e-08], + [ 8.7311e-09, 1.5972e-07, 1.0477e-09, ..., 3.8650e-07, + 3.4296e-07, 1.7730e-07]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0270, -0.0260, -0.0066, 0.0088, -0.0014, -0.0183, 0.0036, 0.0156, + -0.0266, 0.0130], device='cuda:0'), grad: tensor([ 9.6625e-09, 1.0093e-07, 4.2957e-08, 2.7474e-08, -1.0086e-06, + -7.8138e-07, 6.4820e-07, -1.8394e-08, -1.8044e-08, 1.0058e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 217.53, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4757 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.05 lr 0.00010000 +Epoch 328, weight, value: tensor([[ 0.0250, -0.0908, -0.0141, ..., -0.1362, -0.1655, -0.1599], + [-0.1355, 0.0792, -0.1406, ..., -0.1728, -0.1832, -0.2050], + [-0.1267, -0.1884, -0.2017, ..., -0.2219, -0.1583, 0.1688], + ..., + [ 0.0853, -0.0694, -0.1547, ..., 0.1050, -0.1752, -0.1473], + [ 0.1249, 0.1026, 0.0932, ..., -0.2059, -0.1182, 0.0886], + [ 0.0751, 0.0442, 0.1126, ..., 0.0529, -0.0901, 0.0026]], + device='cuda:0'), grad: tensor([[ 8.4983e-09, -2.8522e-08, 6.9849e-10, ..., 1.8743e-08, + 2.0955e-09, 8.1491e-10], + [ 2.0140e-08, 3.5809e-07, 3.2596e-09, ..., 3.7532e-07, + -8.8476e-09, 4.1910e-09], + [ 7.5670e-09, 8.2655e-09, 1.2806e-09, ..., 1.3970e-08, + 9.3132e-10, 2.3283e-10], + ..., + [ 2.3632e-08, 2.8461e-06, 2.5611e-09, ..., 2.2557e-06, + 4.1910e-09, 2.4447e-09], + [-1.0361e-08, 3.2596e-09, -1.2224e-08, ..., 2.7707e-08, + 1.7695e-08, -1.5832e-08], + [-1.0722e-07, -3.7588e-06, -1.0477e-09, ..., -3.1423e-06, + 4.6566e-09, 2.6776e-09]], device='cuda:0') +Epoch 328, bias, value: tensor([ 0.0271, -0.0261, -0.0065, 0.0090, -0.0013, -0.0183, 0.0035, 0.0156, + -0.0266, 0.0128], device='cuda:0'), grad: tensor([ 8.2655e-09, 9.7416e-07, 7.7998e-08, 2.4196e-06, 7.2736e-07, + -3.8855e-06, 1.8887e-06, 6.8545e-06, 3.1898e-07, -9.3803e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 217.82, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4495 re_mapping 0.0026 re_causal 0.0088 /// teacc 99.13 lr 0.00010000 +Epoch 329, weight, value: tensor([[ 0.0272, -0.0910, -0.0142, ..., -0.1367, -0.1656, -0.1599], + [-0.1357, 0.0793, -0.1406, ..., -0.1740, -0.1843, -0.2051], + [-0.1275, -0.1885, -0.2020, ..., -0.2219, -0.1584, 0.1689], + ..., + [ 0.0858, -0.0692, -0.1551, ..., 0.1055, -0.1746, -0.1476], + [ 0.1249, 0.1026, 0.0932, ..., -0.2061, -0.1183, 0.0887], + [ 0.0749, 0.0444, 0.1127, ..., 0.0528, -0.0903, 0.0025]], + device='cuda:0'), grad: tensor([[ 7.6834e-09, 8.1491e-10, 1.1642e-10, ..., 1.0827e-08, + 4.6566e-10, 1.0477e-09], + [ 2.4284e-07, -2.2736e-07, 5.8208e-10, ..., 1.9255e-07, + 2.9104e-09, -7.3807e-08], + [ 8.6846e-08, 2.0617e-07, 0.0000e+00, ..., 7.1712e-08, + 2.3283e-09, 3.7369e-08], + ..., + [-9.4902e-07, 9.6625e-09, 3.4925e-10, ..., -1.0151e-06, + 1.8626e-09, 1.6997e-08], + [ 5.9488e-08, 2.2817e-08, -3.4925e-10, ..., 6.9849e-08, + 2.0955e-09, 1.7579e-08], + [ 3.5344e-07, -2.7940e-09, -2.6776e-09, ..., 4.5775e-07, + 2.0838e-08, 1.6531e-08]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0275, -0.0261, -0.0065, 0.0081, -0.0012, -0.0183, 0.0035, 0.0161, + -0.0267, 0.0127], device='cuda:0'), grad: tensor([ 5.2969e-08, 1.0356e-06, 1.1921e-06, 1.0347e-06, -8.1491e-10, + 1.4657e-07, 1.9558e-08, -6.3442e-06, 5.2387e-07, 2.3469e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 217.94, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4471 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.18 lr 0.00010000 +Epoch 330, weight, value: tensor([[ 0.0274, -0.0911, -0.0142, ..., -0.1368, -0.1656, -0.1600], + [-0.1359, 0.0795, -0.1406, ..., -0.1756, -0.1851, -0.2051], + [-0.1278, -0.1886, -0.2023, ..., -0.2220, -0.1583, 0.1690], + ..., + [ 0.0860, -0.0694, -0.1554, ..., 0.1060, -0.1745, -0.1477], + [ 0.1250, 0.1025, 0.0932, ..., -0.2061, -0.1185, 0.0887], + [ 0.0748, 0.0442, 0.1127, ..., 0.0527, -0.0904, 0.0024]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.9244e-08, 0.0000e+00, ..., 5.8208e-10, + 1.0245e-08, 8.4983e-09], + [ 5.0059e-09, 1.1525e-08, 2.3283e-10, ..., 2.5728e-08, + 1.3970e-09, 5.5879e-09], + [ 9.3132e-10, 7.5670e-09, 2.3283e-10, ..., 6.2864e-09, + 1.3970e-09, -1.7229e-08], + ..., + [-1.2806e-08, -1.1642e-10, 4.6566e-10, ..., -2.2934e-07, + 1.2806e-09, 3.9581e-09], + [ 1.6298e-09, 1.1502e-07, -2.0955e-09, ..., 7.5670e-09, + 2.7940e-09, 1.1642e-10], + [ 2.7940e-09, 1.3737e-07, 3.4925e-10, ..., 2.7567e-07, + 9.7207e-08, 4.6217e-08]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0276, -0.0261, -0.0065, 0.0080, -0.0011, -0.0183, 0.0035, 0.0162, + -0.0267, 0.0126], device='cuda:0'), grad: tensor([ 2.7055e-07, 1.2817e-07, -2.3283e-10, 9.7789e-07, -2.1548e-07, + -2.9262e-06, 2.9569e-07, -4.0466e-07, 6.7800e-07, 1.1958e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 217.84, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4396 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.06 lr 0.00010000 +Epoch 331, weight, value: tensor([[ 0.0277, -0.0915, -0.0140, ..., -0.1374, -0.1658, -0.1601], + [-0.1360, 0.0796, -0.1406, ..., -0.1766, -0.1860, -0.2053], + [-0.1272, -0.1882, -0.2025, ..., -0.2220, -0.1579, 0.1695], + ..., + [ 0.0861, -0.0698, -0.1557, ..., 0.1063, -0.1744, -0.1482], + [ 0.1249, 0.1026, 0.0932, ..., -0.2063, -0.1188, 0.0886], + [ 0.0748, 0.0445, 0.1128, ..., 0.0527, -0.0907, 0.0022]], + device='cuda:0'), grad: tensor([[-2.1886e-08, 4.6566e-09, 8.1491e-10, ..., -2.3283e-10, + 9.5461e-09, 1.8626e-08], + [ 6.0536e-09, 4.1910e-09, 1.8626e-09, ..., 1.5716e-08, + 4.4238e-09, 1.1059e-08], + [ 1.3504e-08, 9.7789e-09, 1.3970e-09, ..., 7.2177e-09, + 5.5879e-09, 2.1653e-08], + ..., + [ 5.7044e-09, 1.1758e-08, 2.0955e-09, ..., 1.5250e-08, + 4.5402e-09, 1.8394e-08], + [ 4.0745e-09, -1.5134e-09, -6.9849e-10, ..., 2.9919e-08, + 3.7253e-09, 1.8510e-08], + [-8.9291e-08, -9.6741e-08, -2.3050e-08, ..., -2.8755e-08, + 1.3306e-07, -1.0524e-07]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0276, -0.0260, -0.0062, 0.0081, -0.0011, -0.0184, 0.0037, 0.0159, + -0.0268, 0.0124], device='cuda:0'), grad: tensor([-7.7998e-09, 4.5169e-08, 8.0676e-08, 9.1968e-08, 1.3958e-07, + -3.2876e-07, 1.1653e-07, 7.2643e-08, 1.9069e-07, -3.7299e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 217.70, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4544 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.03 lr 0.00010000 +Epoch 332, weight, value: tensor([[ 0.0277, -0.0922, -0.0141, ..., -0.1381, -0.1660, -0.1603], + [-0.1361, 0.0797, -0.1406, ..., -0.1774, -0.1909, -0.2057], + [-0.1294, -0.1882, -0.2030, ..., -0.2221, -0.1582, 0.1696], + ..., + [ 0.0863, -0.0707, -0.1578, ..., 0.1062, -0.1755, -0.1482], + [ 0.1249, 0.1027, 0.0932, ..., -0.2069, -0.1197, 0.0885], + [ 0.0754, 0.0451, 0.1131, ..., 0.0528, -0.0910, 0.0020]], + device='cuda:0'), grad: tensor([[ 2.2119e-09, 2.0838e-08, 3.1432e-09, ..., 1.5600e-08, + 2.4331e-08, 1.8161e-08], + [ 1.3621e-08, -1.0794e-06, 5.9372e-09, ..., -8.8126e-08, + 3.4925e-09, 1.5018e-08], + [ 9.5461e-09, 2.6426e-08, 6.9849e-10, ..., 2.4680e-08, + 2.4447e-09, 5.4715e-09], + ..., + [ 8.3819e-09, 3.4086e-07, 5.4599e-08, ..., 2.5472e-07, + 5.5879e-09, -4.5402e-09], + [-9.5461e-09, 5.0873e-08, 7.1013e-09, ..., 7.6834e-08, + 2.2119e-08, -8.9640e-09], + [-8.1840e-08, -2.8731e-07, -1.4633e-07, ..., -6.8732e-07, + 4.4703e-08, 2.2934e-08]], device='cuda:0') +Epoch 332, bias, value: tensor([ 0.0275, -0.0259, -0.0063, 0.0092, -0.0011, -0.0154, 0.0007, 0.0154, + -0.0270, 0.0126], device='cuda:0'), grad: tensor([ 2.7963e-07, -7.0482e-06, 2.1502e-07, -1.1385e-07, 5.4613e-06, + -8.1122e-05, 8.0824e-05, 1.7146e-06, 6.7521e-07, -7.6089e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 217.46, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4992 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.07 lr 0.00010000 +Epoch 333, weight, value: tensor([[ 0.0285, -0.0925, -0.0142, ..., -0.1386, -0.1660, -0.1603], + [-0.1362, 0.0797, -0.1407, ..., -0.1786, -0.1925, -0.2059], + [-0.1290, -0.1877, -0.2032, ..., -0.2217, -0.1584, 0.1702], + ..., + [ 0.0861, -0.0714, -0.1583, ..., 0.1075, -0.1737, -0.1493], + [ 0.1249, 0.1027, 0.0932, ..., -0.2075, -0.1200, 0.0886], + [ 0.0754, 0.0450, 0.1133, ..., 0.0524, -0.0920, 0.0015]], + device='cuda:0'), grad: tensor([[ 1.0594e-08, 1.2806e-08, 2.3283e-10, ..., 2.3283e-10, + 1.2806e-09, 9.1968e-09], + [ 4.8429e-08, 3.7719e-08, 3.1432e-09, ..., 2.5611e-09, + 8.1491e-10, 1.3283e-07], + [ 1.4435e-08, 3.0152e-08, 2.5611e-09, ..., 2.3283e-10, + 9.3132e-10, -3.2736e-07], + ..., + [ 1.6252e-07, 1.4273e-07, 2.7940e-09, ..., 9.1968e-09, + 4.8894e-09, 1.1700e-07], + [-4.9453e-07, -2.5285e-07, -2.2235e-08, ..., 5.8208e-10, + 9.0804e-09, -1.6089e-07], + [ 3.0384e-08, 1.8743e-08, 4.6566e-10, ..., 1.4901e-08, + 1.8859e-08, 2.9919e-08]], device='cuda:0') +Epoch 333, bias, value: tensor([ 0.0277, -0.0260, -0.0057, 0.0081, -0.0011, -0.0153, 0.0007, 0.0155, + -0.0271, 0.0121], device='cuda:0'), grad: tensor([ 7.3528e-07, 3.0845e-05, -1.0288e-04, 4.0799e-05, 7.5670e-07, + -1.2049e-07, 1.5432e-06, 2.5898e-05, -9.2294e-07, 3.3639e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 217.92, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4629 re_mapping 0.0027 re_causal 0.0085 /// teacc 99.07 lr 0.00010000 +Epoch 334, weight, value: tensor([[ 0.0287, -0.0929, -0.0144, ..., -0.1387, -0.1662, -0.1606], + [-0.1363, 0.0797, -0.1407, ..., -0.1795, -0.1932, -0.2062], + [-0.1263, -0.1860, -0.2040, ..., -0.2203, -0.1586, 0.1719], + ..., + [ 0.0855, -0.0730, -0.1587, ..., 0.1064, -0.1737, -0.1522], + [ 0.1251, 0.1032, 0.0933, ..., -0.2076, -0.1200, 0.0893], + [ 0.0746, 0.0438, 0.1131, ..., 0.0521, -0.0926, 0.0008]], + device='cuda:0'), grad: tensor([[-9.5926e-08, -3.2596e-08, 1.8626e-09, ..., 5.8208e-10, + 2.3283e-10, -2.2701e-08], + [ 1.9209e-08, -6.6706e-08, 8.1491e-09, ..., 2.5262e-08, + 2.5611e-09, 1.2224e-08], + [ 1.3737e-08, 4.7614e-08, 6.8685e-09, ..., 1.9558e-08, + 1.5134e-09, 5.9372e-09], + ..., + [-2.9104e-08, 6.4727e-08, 7.5670e-09, ..., -7.4855e-08, + -4.7730e-09, -6.0536e-09], + [ 1.1642e-09, 3.2131e-08, -7.2177e-09, ..., 1.2224e-08, + 1.9791e-09, -2.1537e-08], + [ 3.7835e-08, 8.4983e-08, 4.0745e-09, ..., 3.3877e-08, + 1.5018e-08, 2.2002e-08]], device='cuda:0') +Epoch 334, bias, value: tensor([ 0.0277, -0.0261, -0.0041, 0.0064, -0.0009, -0.0152, 0.0007, 0.0138, + -0.0270, 0.0114], device='cuda:0'), grad: tensor([-7.3947e-07, -1.2678e-07, 2.5425e-07, -1.0803e-06, -3.9581e-09, + 5.5274e-07, 3.3714e-07, -5.4832e-08, 3.0827e-07, 5.6392e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 217.63, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4612 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.14 lr 0.00010000 +Epoch 335, weight, value: tensor([[ 0.0295, -0.0929, -0.0145, ..., -0.1391, -0.1663, -0.1607], + [-0.1364, 0.0798, -0.1408, ..., -0.1802, -0.1936, -0.2063], + [-0.1264, -0.1862, -0.2045, ..., -0.2205, -0.1588, 0.1719], + ..., + [ 0.0856, -0.0732, -0.1594, ..., 0.1067, -0.1742, -0.1522], + [ 0.1251, 0.1032, 0.0933, ..., -0.2078, -0.1204, 0.0891], + [ 0.0747, 0.0440, 0.1132, ..., 0.0518, -0.0932, 0.0005]], + device='cuda:0'), grad: tensor([[-5.8208e-10, 8.1491e-09, 2.9220e-08, ..., 1.9791e-09, + 9.5554e-07, 1.0384e-06], + [ 9.0804e-09, -8.9407e-08, 5.3551e-09, ..., 1.4319e-08, + 9.4529e-08, 1.0547e-07], + [ 4.3074e-09, 1.5832e-08, 1.6298e-09, ..., 6.8685e-09, + 2.1304e-08, 2.4447e-08], + ..., + [-2.1886e-08, 5.2154e-08, 8.2655e-09, ..., -2.6426e-08, + 3.4925e-09, -1.3970e-09], + [-2.2119e-09, 2.8405e-08, 7.2177e-08, ..., 1.1176e-08, + 1.6633e-06, 1.7844e-06], + [ 4.7730e-09, 6.5193e-09, 2.4680e-08, ..., 4.0280e-08, + 8.1537e-07, 8.8941e-07]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0278, -0.0261, -0.0041, 0.0065, -0.0008, -0.0152, 0.0007, 0.0138, + -0.0271, 0.0113], device='cuda:0'), grad: tensor([ 3.2168e-06, -2.2724e-07, 1.9616e-07, 1.8021e-07, 2.9672e-06, + 1.5376e-06, -1.6361e-05, 1.5402e-07, 5.2005e-06, 3.1255e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 217.54, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4667 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.11 lr 0.00010000 +Epoch 336, weight, value: tensor([[ 2.9844e-02, -9.3301e-02, -1.4643e-02, ..., -1.3953e-01, + -1.6674e-01, -1.6103e-01], + [-1.3646e-01, 8.0034e-02, -1.4083e-01, ..., -1.8030e-01, + -1.9461e-01, -2.0643e-01], + [-1.2645e-01, -1.8633e-01, -2.0526e-01, ..., -2.2059e-01, + -1.5930e-01, 1.7196e-01], + ..., + [ 8.5616e-02, -7.3495e-02, -1.5986e-01, ..., 1.0659e-01, + -1.7479e-01, -1.5225e-01], + [ 1.2523e-01, 1.0329e-01, 9.3445e-02, ..., -2.0785e-01, + -1.2073e-01, 8.9188e-02], + [ 7.4545e-02, 4.3490e-02, 1.1299e-01, ..., 5.1031e-02, + -9.4713e-02, -1.9185e-04]], device='cuda:0'), grad: tensor([[ 1.1642e-10, 7.7998e-09, 0.0000e+00, ..., 2.0303e-07, + 2.2375e-07, 1.6752e-07], + [ 1.6298e-09, -9.5554e-07, 4.6566e-10, ..., 3.3225e-07, + 3.6368e-07, 2.7474e-07], + [ 1.2806e-09, 3.8301e-08, 1.7462e-09, ..., 1.0263e-06, + 1.1371e-06, 8.4750e-07], + ..., + [-5.8208e-10, 9.0525e-07, 1.7462e-09, ..., 1.7521e-07, + 1.8848e-07, 1.4144e-07], + [-5.7044e-09, 6.4913e-07, 2.7940e-09, ..., 1.5844e-07, + 9.9302e-08, 7.1363e-08], + [-2.5611e-09, 3.5646e-07, -9.3132e-09, ..., 2.0713e-06, + 2.2911e-06, 1.7155e-06]], device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0276, -0.0260, -0.0041, 0.0072, -0.0007, -0.0153, 0.0008, 0.0136, + -0.0271, 0.0106], device='cuda:0'), grad: tensor([ 9.6392e-07, -1.5795e-06, 4.8093e-06, 2.2221e-06, -2.0608e-05, + -9.0301e-06, 3.4459e-06, 3.7905e-06, 4.4070e-06, 1.1586e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 217.48, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4596 re_mapping 0.0028 re_causal 0.0088 /// teacc 99.16 lr 0.00010000 +Epoch 337, weight, value: tensor([[ 0.0316, -0.0926, -0.0146, ..., -0.1401, -0.1667, -0.1608], + [-0.1370, 0.0801, -0.1409, ..., -0.1839, -0.1966, -0.2069], + [-0.1268, -0.1865, -0.2058, ..., -0.2208, -0.1595, 0.1720], + ..., + [ 0.0862, -0.0735, -0.1602, ..., 0.1080, -0.1723, -0.1520], + [ 0.1252, 0.1034, 0.0935, ..., -0.2080, -0.1209, 0.0893], + [ 0.0746, 0.0442, 0.1135, ..., 0.0518, -0.0944, -0.0004]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.0179e-06, 3.8417e-09, ..., 6.8685e-09, + -3.2573e-07, 8.8476e-09], + [ 2.9686e-08, 1.3167e-07, 3.4925e-09, ..., 4.9127e-08, + 9.8837e-08, 8.6613e-08], + [ 8.0210e-08, 3.6205e-08, 9.3132e-10, ..., 1.2654e-07, + 2.2119e-09, -2.2736e-07], + ..., + [-1.5681e-07, 4.3772e-08, 1.0245e-08, ..., -2.5239e-07, + 4.0745e-09, 2.9453e-08], + [ 3.2946e-08, 1.7218e-07, 1.5367e-08, ..., 9.4296e-08, + 3.0966e-08, 9.1735e-08], + [-2.8522e-08, -4.9127e-08, -6.8918e-08, ..., -1.1385e-07, + 5.0059e-09, -7.9162e-09]], device='cuda:0') +Epoch 337, bias, value: tensor([ 0.0287, -0.0263, -0.0042, 0.0072, -0.0015, -0.0153, 0.0008, 0.0143, + -0.0271, 0.0110], device='cuda:0'), grad: tensor([-1.0453e-05, 3.2336e-06, -3.0408e-07, 1.3446e-07, 2.9942e-07, + 2.5844e-07, 5.8152e-06, -6.7567e-07, 1.6959e-06, -3.8417e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 217.74, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4590 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 338, weight, value: tensor([[ 0.0321, -0.0927, -0.0148, ..., -0.1411, -0.1668, -0.1608], + [-0.1371, 0.0801, -0.1410, ..., -0.1847, -0.1969, -0.2071], + [-0.1269, -0.1866, -0.2066, ..., -0.2210, -0.1597, 0.1722], + ..., + [ 0.0863, -0.0738, -0.1616, ..., 0.1101, -0.1705, -0.1520], + [ 0.1253, 0.1035, 0.0936, ..., -0.2083, -0.1210, 0.0900], + [ 0.0751, 0.0463, 0.1138, ..., 0.0536, -0.0926, 0.0004]], + device='cuda:0'), grad: tensor([[ 3.1432e-09, 1.1642e-09, 3.4925e-10, ..., 4.5402e-09, + 5.5879e-09, 4.8894e-09], + [ 5.1688e-08, 1.0477e-09, 2.0955e-09, ..., 3.1083e-07, + 1.2224e-08, 8.4983e-09], + [ 5.4599e-08, 5.7044e-09, 1.2806e-09, ..., 1.0617e-07, + 9.8953e-09, 5.7044e-09], + ..., + [-2.9989e-07, -1.7579e-08, 8.1491e-10, ..., -7.3016e-07, + -1.2922e-08, 1.2806e-09], + [-1.6298e-09, -3.4226e-08, -2.4680e-08, ..., 2.5146e-08, + 8.8476e-09, -3.1316e-08], + [ 1.6706e-07, 2.1886e-08, 7.7998e-09, ..., 2.4866e-07, + 1.8626e-09, 1.1642e-08]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0290, -0.0264, -0.0042, 0.0072, -0.0038, -0.0153, 0.0008, 0.0149, + -0.0270, 0.0129], device='cuda:0'), grad: tensor([ 3.2247e-08, 1.3653e-06, 4.5612e-07, 9.9535e-08, 4.1211e-08, + 1.3609e-07, -1.1572e-07, -3.0082e-06, 1.3271e-08, 9.9093e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 217.82, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4631 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.20 lr 0.00010000 +Epoch 339, weight, value: tensor([[ 0.0313, -0.0929, -0.0149, ..., -0.1440, -0.1670, -0.1611], + [-0.1382, 0.0796, -0.1410, ..., -0.1875, -0.1974, -0.2073], + [-0.1273, -0.1867, -0.2072, ..., -0.2212, -0.1598, 0.1722], + ..., + [ 0.0871, -0.0730, -0.1620, ..., 0.1103, -0.1723, -0.1521], + [ 0.1254, 0.1039, 0.0937, ..., -0.2084, -0.1214, 0.0902], + [ 0.0754, 0.0466, 0.1137, ..., 0.0545, -0.0917, 0.0004]], + device='cuda:0'), grad: tensor([[ 1.7346e-08, 4.8894e-09, 8.1491e-09, ..., 2.4564e-08, + 3.4925e-10, 9.0804e-09], + [ 1.5134e-08, 2.3632e-08, 4.4238e-09, ..., 1.5716e-08, + 1.2806e-09, 6.2864e-09], + [ 5.3551e-09, 1.4203e-08, 2.0955e-09, ..., 4.1910e-09, + 2.3283e-10, -1.5367e-08], + ..., + [-1.1316e-07, 1.0768e-07, 2.3749e-08, ..., -7.1013e-09, + 3.4925e-09, 3.2596e-09], + [ 4.3423e-08, -5.0524e-08, 3.6089e-09, ..., 3.7486e-08, + -2.5262e-08, -5.7509e-08], + [-1.1525e-08, -2.0140e-08, -3.2014e-08, ..., -2.0675e-07, + -1.0710e-08, -5.3551e-09]], device='cuda:0') +Epoch 339, bias, value: tensor([ 0.0287, -0.0273, -0.0043, 0.0068, -0.0045, -0.0153, 0.0008, 0.0161, + -0.0269, 0.0135], device='cuda:0'), grad: tensor([ 7.2876e-08, 1.2899e-07, -2.9569e-08, -6.0350e-07, 3.5018e-07, + -2.6822e-07, 1.8126e-07, -7.0781e-08, 3.6042e-07, -9.7672e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 217.46, cls_loss 0.0004 cls_loss_mapping 0.0014 cls_loss_causal 0.4446 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.17 lr 0.00010000 +Epoch 340, weight, value: tensor([[ 0.0315, -0.0930, -0.0150, ..., -0.1442, -0.1671, -0.1612], + [-0.1383, 0.0793, -0.1410, ..., -0.1879, -0.1977, -0.2074], + [-0.1273, -0.1867, -0.2075, ..., -0.2213, -0.1598, 0.1723], + ..., + [ 0.0872, -0.0732, -0.1627, ..., 0.1101, -0.1737, -0.1523], + [ 0.1254, 0.1040, 0.0937, ..., -0.2088, -0.1216, 0.0902], + [ 0.0757, 0.0469, 0.1138, ..., 0.0546, -0.0918, 0.0004]], + device='cuda:0'), grad: tensor([[ 7.7998e-09, 5.6112e-08, 1.6764e-08, ..., 1.4319e-08, + 9.3132e-09, 6.5193e-09], + [ 6.8336e-08, 2.4098e-08, 1.3853e-08, ..., 1.8755e-07, + 1.8626e-09, 1.0710e-08], + [ 3.1432e-08, 3.1549e-08, 5.5879e-09, ..., 7.6019e-08, + 6.5193e-09, -5.7044e-09], + ..., + [-5.8440e-08, 1.4366e-07, 6.3679e-08, ..., -2.3097e-07, + 6.0536e-09, 7.6834e-09], + [ 4.0629e-08, 2.4633e-07, 7.9628e-08, ..., 1.0047e-07, + 3.6554e-08, 1.1874e-08], + [-1.3656e-07, -2.6356e-07, -1.3353e-07, ..., -2.1688e-07, + 2.4447e-08, 1.5250e-08]], device='cuda:0') +Epoch 340, bias, value: tensor([ 0.0287, -0.0275, -0.0042, 0.0063, -0.0045, -0.0153, 0.0008, 0.0160, + -0.0269, 0.0136], device='cuda:0'), grad: tensor([ 2.0221e-07, 8.2329e-07, 3.0827e-07, -2.1160e-06, 9.4878e-08, + 1.2703e-06, -1.3970e-09, -9.0152e-07, 9.5647e-07, -6.3656e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 217.64, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4868 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.15 lr 0.00010000 +Epoch 341, weight, value: tensor([[ 0.0320, -0.0931, -0.0151, ..., -0.1445, -0.1671, -0.1611], + [-0.1384, 0.0795, -0.1410, ..., -0.1884, -0.1980, -0.2075], + [-0.1274, -0.1870, -0.2077, ..., -0.2216, -0.1600, 0.1724], + ..., + [ 0.0873, -0.0735, -0.1632, ..., 0.1103, -0.1741, -0.1525], + [ 0.1254, 0.1041, 0.0937, ..., -0.2090, -0.1216, 0.0904], + [ 0.0757, 0.0473, 0.1139, ..., 0.0534, -0.0930, -0.0015]], + device='cuda:0'), grad: tensor([[-4.3074e-09, 2.3225e-08, 0.0000e+00, ..., 1.6473e-08, + 1.6810e-07, 1.1059e-07], + [ 5.3551e-09, 1.2666e-07, 1.1642e-10, ..., 6.3388e-08, + 5.5938e-08, 4.7963e-08], + [-6.9849e-10, 6.1654e-07, 5.8208e-11, ..., 1.7055e-08, + 1.9034e-08, -3.0384e-08], + ..., + [-1.2282e-08, 1.2864e-07, 5.8208e-11, ..., 6.2864e-09, + 2.6950e-08, 2.2410e-08], + [ 1.8044e-09, 2.5029e-08, -3.4925e-10, ..., 3.1781e-08, + 3.1956e-08, 2.9162e-08], + [ 7.8580e-09, 3.9814e-08, 5.8208e-11, ..., 2.6543e-07, + 2.3632e-07, 1.5483e-07]], device='cuda:0') +Epoch 341, bias, value: tensor([ 0.0290, -0.0274, -0.0044, 0.0067, -0.0034, -0.0153, 0.0008, 0.0160, + -0.0270, 0.0125], device='cuda:0'), grad: tensor([ 9.0431e-07, 7.2177e-07, 2.3041e-06, -4.5523e-06, -1.3439e-06, + 9.8161e-07, -6.5751e-07, 5.4855e-07, 2.6473e-07, 8.3400e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 217.79, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4759 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.20 lr 0.00010000 +Epoch 342, weight, value: tensor([[ 0.0323, -0.0936, -0.0153, ..., -0.1447, -0.1674, -0.1613], + [-0.1386, 0.0797, -0.1411, ..., -0.1897, -0.1990, -0.2077], + [-0.1283, -0.1874, -0.2084, ..., -0.2227, -0.1603, 0.1722], + ..., + [ 0.0878, -0.0742, -0.1636, ..., 0.1109, -0.1741, -0.1522], + [ 0.1257, 0.1041, 0.0941, ..., -0.2097, -0.1212, 0.0914], + [ 0.0758, 0.0481, 0.1141, ..., 0.0536, -0.0931, -0.0015]], + device='cuda:0'), grad: tensor([[ 6.4028e-10, 1.5716e-09, 1.1642e-10, ..., 1.7462e-10, + 6.4028e-10, 1.2224e-09], + [ 2.0198e-08, 1.3737e-08, 1.5716e-09, ..., 1.1059e-08, + 5.8208e-11, 1.4494e-08], + [ 2.1944e-08, 1.6356e-08, 1.0477e-09, ..., 1.9558e-08, + 5.8208e-11, 9.0804e-09], + ..., + [-2.2235e-08, 1.6007e-08, 1.2224e-09, ..., -4.9244e-08, + 0.0000e+00, 7.6252e-09], + [-5.4890e-08, -7.6427e-08, -7.4506e-09, ..., 1.1001e-08, + 6.4028e-10, -6.3679e-08], + [ 5.7044e-09, 4.5402e-09, -5.8208e-11, ..., 5.5879e-09, + 9.3132e-10, 4.3074e-09]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0290, -0.0276, -0.0048, 0.0077, -0.0034, -0.0153, 0.0008, 0.0164, + -0.0268, 0.0127], device='cuda:0'), grad: tensor([ 6.8685e-09, 5.2562e-08, 1.0786e-07, -6.4669e-08, 5.2387e-09, + 1.3760e-07, 1.4203e-08, -6.8801e-08, -2.1304e-07, 3.0850e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 217.63, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4793 re_mapping 0.0026 re_causal 0.0088 /// teacc 99.21 lr 0.00010000 +Epoch 343, weight, value: tensor([[ 0.0327, -0.0939, -0.0153, ..., -0.1447, -0.1675, -0.1613], + [-0.1387, 0.0798, -0.1411, ..., -0.1900, -0.1993, -0.2078], + [-0.1284, -0.1875, -0.2087, ..., -0.2228, -0.1606, 0.1723], + ..., + [ 0.0879, -0.0751, -0.1661, ..., 0.1111, -0.1745, -0.1524], + [ 0.1257, 0.1041, 0.0942, ..., -0.2098, -0.1215, 0.0915], + [ 0.0756, 0.0480, 0.1141, ..., 0.0535, -0.0932, -0.0017]], + device='cuda:0'), grad: tensor([[-2.2352e-08, -1.4959e-08, 1.1642e-10, ..., 2.2119e-09, + 4.6566e-10, 8.7311e-10], + [ 3.1851e-07, 2.1828e-08, 2.9104e-10, ..., 4.2981e-07, + 1.0768e-08, 1.8743e-08], + [ 7.6834e-09, 1.1642e-09, 1.1642e-10, ..., 1.3679e-08, + 3.4343e-09, 4.0745e-10], + ..., + [-4.5542e-07, -1.7462e-08, 2.3865e-09, ..., -6.4401e-07, + -1.9965e-08, -1.8219e-08], + [ 3.2305e-08, 4.0745e-09, -4.6566e-10, ..., 5.9605e-08, + 5.2387e-09, -1.7462e-10], + [ 8.2655e-08, -1.1525e-08, -3.6089e-09, ..., 1.9290e-07, + 5.6694e-08, 2.9511e-08]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0294, -0.0276, -0.0048, 0.0089, -0.0033, -0.0153, 0.0008, 0.0162, + -0.0269, 0.0126], device='cuda:0'), grad: tensor([-2.1677e-07, 1.1120e-06, 3.2014e-08, 1.5716e-08, -8.7777e-08, + 8.5449e-08, 9.6625e-08, -1.5656e-06, 1.6170e-07, 4.0419e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 217.80, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.4762 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.13 lr 0.00010000 +Epoch 344, weight, value: tensor([[ 0.0329, -0.0945, -0.0155, ..., -0.1447, -0.1676, -0.1614], + [-0.1412, 0.0773, -0.1412, ..., -0.1902, -0.1994, -0.2080], + [-0.1285, -0.1876, -0.2093, ..., -0.2230, -0.1607, 0.1726], + ..., + [ 0.0900, -0.0725, -0.1665, ..., 0.1115, -0.1748, -0.1531], + [ 0.1259, 0.1043, 0.0942, ..., -0.2100, -0.1219, 0.0917], + [ 0.0754, 0.0478, 0.1141, ..., 0.0534, -0.0933, -0.0018]], + device='cuda:0'), grad: tensor([[ 2.4447e-09, 6.4611e-09, 2.3865e-09, ..., 3.9581e-09, + 4.5402e-09, 6.2282e-09], + [ 2.6252e-08, -5.0990e-08, 3.4925e-09, ..., 5.0059e-08, + 2.8522e-09, 7.6252e-09], + [ 6.5833e-08, 5.6461e-09, 1.1642e-09, ..., 2.7218e-07, + 1.9791e-09, -6.1700e-09], + ..., + [-2.1083e-07, 2.6659e-08, 4.3074e-09, ..., -4.7684e-07, + 1.6880e-09, 6.0536e-09], + [ 5.5414e-08, 1.6880e-07, 7.9744e-08, ..., 1.5297e-07, + 1.6356e-08, 7.2876e-08], + [ 3.2014e-08, -2.1956e-07, -1.1880e-07, ..., 9.7207e-09, + 4.1444e-08, -7.7242e-08]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0294, -0.0300, -0.0046, 0.0106, -0.0033, -0.0154, 0.0008, 0.0187, + -0.0269, 0.0124], device='cuda:0'), grad: tensor([ 3.6554e-08, -2.2061e-08, 8.8802e-07, 3.4168e-08, -4.0745e-09, + 3.7160e-07, -2.9663e-07, -1.6484e-06, 7.1898e-07, -6.1467e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 217.74, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4470 re_mapping 0.0024 re_causal 0.0085 /// teacc 99.20 lr 0.00010000 +Epoch 345, weight, value: tensor([[ 0.0330, -0.0946, -0.0155, ..., -0.1448, -0.1678, -0.1616], + [-0.1412, 0.0775, -0.1412, ..., -0.1905, -0.1997, -0.2081], + [-0.1289, -0.1877, -0.2095, ..., -0.2231, -0.1610, 0.1726], + ..., + [ 0.0901, -0.0727, -0.1672, ..., 0.1118, -0.1750, -0.1531], + [ 0.1260, 0.1043, 0.0942, ..., -0.2102, -0.1227, 0.0916], + [ 0.0753, 0.0479, 0.1142, ..., 0.0532, -0.0935, -0.0021]], + device='cuda:0'), grad: tensor([[ 2.8114e-08, 2.8347e-08, 1.5541e-08, ..., 1.9209e-09, + 1.5774e-08, 5.2969e-08], + [ 1.2759e-07, 1.0373e-07, 6.9092e-08, ..., 3.3760e-09, + 4.4587e-08, 2.2422e-07], + [ 9.1968e-09, 8.9640e-09, 4.7148e-09, ..., 1.0477e-09, + 5.0059e-09, -2.0314e-07], + ..., + [ 7.1595e-08, 4.6159e-08, 6.8103e-09, ..., 1.7649e-07, + 2.8522e-09, 1.8883e-07], + [-8.4378e-07, -7.8278e-07, -4.6007e-07, ..., 2.9802e-08, + -2.3656e-07, -1.3234e-06], + [-6.2864e-08, -2.2002e-08, 3.7253e-09, ..., -1.9791e-07, + 1.3853e-08, 2.8929e-08]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0293, -0.0300, -0.0047, 0.0102, -0.0031, -0.0154, 0.0009, 0.0187, + -0.0269, 0.0122], device='cuda:0'), grad: tensor([ 1.6484e-07, 6.6310e-07, -5.5507e-07, 6.1933e-08, 1.7043e-07, + -2.1351e-07, 3.2224e-06, 8.8243e-07, -4.0606e-06, -3.1339e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 217.51, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4693 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.16 lr 0.00010000 +Epoch 346, weight, value: tensor([[ 0.0330, -0.0949, -0.0156, ..., -0.1450, -0.1691, -0.1624], + [-0.1413, 0.0778, -0.1412, ..., -0.1895, -0.1982, -0.2081], + [-0.1292, -0.1878, -0.2101, ..., -0.2233, -0.1613, 0.1727], + ..., + [ 0.0903, -0.0729, -0.1674, ..., 0.1121, -0.1752, -0.1532], + [ 0.1261, 0.1044, 0.0943, ..., -0.2104, -0.1225, 0.0920], + [ 0.0749, 0.0477, 0.1141, ..., 0.0530, -0.0937, -0.0027]], + device='cuda:0'), grad: tensor([[ 2.9104e-10, 2.3283e-10, 0.0000e+00, ..., 4.0745e-10, + 1.4552e-09, 8.7311e-10], + [ 7.9162e-09, 3.1432e-09, 2.3283e-10, ..., 6.6939e-09, + 1.6298e-09, 7.9744e-09], + [ 6.9267e-09, 9.3132e-10, 5.8208e-11, ..., 8.8476e-09, + 8.7311e-10, 9.3132e-10], + ..., + [-3.9581e-08, 7.2177e-09, 2.3283e-10, ..., -5.5355e-08, + 8.7311e-10, 4.0745e-10], + [-2.0373e-09, -1.1234e-08, -5.8208e-10, ..., 3.0268e-09, + 3.5507e-09, -1.0303e-08], + [ 2.3283e-08, -1.3970e-09, -6.4028e-10, ..., 3.9989e-08, + 4.3656e-09, 3.3178e-09]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0287, -0.0299, -0.0047, 0.0101, -0.0031, -0.0154, 0.0009, 0.0187, + -0.0268, 0.0119], device='cuda:0'), grad: tensor([ 1.1001e-08, 4.7730e-08, 3.7020e-08, 2.3399e-08, 1.2806e-09, + -7.1619e-07, 5.2061e-07, -1.4168e-07, 9.7731e-08, 1.4040e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 217.84, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4584 re_mapping 0.0027 re_causal 0.0086 /// teacc 99.06 lr 0.00010000 +Epoch 347, weight, value: tensor([[ 0.0314, -0.0952, -0.0157, ..., -0.1470, -0.1694, -0.1626], + [-0.1413, 0.0780, -0.1412, ..., -0.1903, -0.1994, -0.2082], + [-0.1298, -0.1883, -0.2104, ..., -0.2238, -0.1619, 0.1727], + ..., + [ 0.0902, -0.0733, -0.1684, ..., 0.1121, -0.1753, -0.1532], + [ 0.1261, 0.1044, 0.0943, ..., -0.2106, -0.1228, 0.0919], + [ 0.0771, 0.0477, 0.1141, ..., 0.0529, -0.0943, -0.0035]], + device='cuda:0'), grad: tensor([[ 6.4028e-10, 2.0955e-09, 1.7462e-09, ..., 3.2596e-09, + 5.2387e-10, 8.7311e-10], + [ 4.3074e-09, -9.4296e-09, 3.6671e-09, ..., 8.7311e-09, + 1.3970e-09, 6.6357e-09], + [ 3.6671e-09, 5.1223e-09, 9.8953e-10, ..., 2.2701e-09, + -5.5879e-09, -3.8126e-08], + ..., + [ 2.9104e-10, 1.0303e-08, 7.3924e-09, ..., 1.2689e-08, + 2.0373e-09, 3.4925e-09], + [-2.0373e-08, -1.8044e-09, 5.2387e-10, ..., 9.3714e-09, + 2.0373e-09, -1.5600e-08], + [-9.8953e-09, -4.6392e-08, -5.1979e-08, ..., -6.1933e-08, + 2.6426e-08, 8.2655e-09]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0271, -0.0298, -0.0049, 0.0102, -0.0026, -0.0154, 0.0009, 0.0186, + -0.0269, 0.0118], device='cuda:0'), grad: tensor([ 1.7055e-08, 4.4296e-08, -3.2224e-07, 1.3714e-07, 1.4273e-07, + 9.5053e-08, -8.1491e-09, 7.5321e-08, 3.8417e-09, -1.6787e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 217.63, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4612 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.21 lr 0.00010000 +Epoch 348, weight, value: tensor([[ 0.0302, -0.0954, -0.0158, ..., -0.1482, -0.1695, -0.1628], + [-0.1414, 0.0784, -0.1413, ..., -0.1910, -0.2002, -0.2083], + [-0.1301, -0.1899, -0.2109, ..., -0.2240, -0.1621, 0.1728], + ..., + [ 0.0904, -0.0734, -0.1688, ..., 0.1126, -0.1751, -0.1533], + [ 0.1262, 0.1046, 0.0944, ..., -0.2109, -0.1231, 0.0920], + [ 0.0780, 0.0478, 0.1142, ..., 0.0531, -0.0944, -0.0037]], + device='cuda:0'), grad: tensor([[-2.7358e-09, 5.1223e-09, 2.3283e-10, ..., 4.0745e-10, + 1.7346e-08, 1.7113e-08], + [ 5.9954e-09, -9.1316e-07, 1.9209e-09, ..., 5.9954e-09, + 5.4715e-09, 8.2073e-09], + [ 2.7358e-09, 1.2340e-08, 5.8208e-10, ..., 2.6193e-09, + 4.7148e-09, 5.4715e-09], + ..., + [-1.3388e-09, 4.3015e-08, 1.3970e-09, ..., -8.4983e-09, + 6.9849e-10, 2.0373e-09], + [-2.3341e-08, 7.3062e-07, -1.3679e-08, ..., 2.5029e-09, + 2.2235e-08, 5.1805e-09], + [-4.5984e-09, 1.7812e-08, -2.1537e-09, ..., -2.1246e-08, + 2.1537e-09, 3.8999e-09]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0262, -0.0296, -0.0053, 0.0098, -0.0025, -0.0154, 0.0009, 0.0187, + -0.0269, 0.0120], device='cuda:0'), grad: tensor([-3.8650e-08, -2.4196e-06, 5.9954e-08, 1.0245e-08, 7.2701e-08, + -1.7090e-07, 2.3865e-07, 1.1042e-07, 2.1178e-06, 4.7323e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 217.64, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4472 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.18 lr 0.00010000 +Epoch 349, weight, value: tensor([[ 0.0302, -0.0957, -0.0159, ..., -0.1483, -0.1701, -0.1636], + [-0.1416, 0.0784, -0.1413, ..., -0.1919, -0.2015, -0.2088], + [-0.1307, -0.1900, -0.2123, ..., -0.2250, -0.1626, 0.1731], + ..., + [ 0.0908, -0.0735, -0.1690, ..., 0.1140, -0.1751, -0.1531], + [ 0.1263, 0.1048, 0.0944, ..., -0.2113, -0.1233, 0.0922], + [ 0.0779, 0.0479, 0.1143, ..., 0.0530, -0.0946, -0.0040]], + device='cuda:0'), grad: tensor([[-1.6356e-08, 1.0652e-08, 5.8208e-11, ..., 1.7462e-09, + 1.8801e-08, 1.2224e-08], + [ 6.5775e-09, -1.4435e-07, 1.7462e-10, ..., 9.3714e-09, + 6.0536e-09, -1.4435e-08], + [ 1.8685e-08, 1.1991e-08, 5.8208e-11, ..., 8.8476e-09, + 4.4820e-09, 1.1642e-08], + ..., + [-1.1059e-09, 2.8871e-08, 7.5670e-10, ..., -4.0978e-08, + 3.2596e-09, 5.0059e-09], + [-3.5856e-08, 4.7323e-08, 0.0000e+00, ..., 4.7148e-09, + 4.3656e-09, 1.0070e-08], + [-6.2282e-09, -6.5193e-09, -6.2282e-09, ..., 1.5495e-07, + 1.1758e-07, 6.9966e-08]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0257, -0.0298, -0.0062, 0.0098, -0.0027, -0.0154, 0.0010, 0.0197, + -0.0269, 0.0118], device='cuda:0'), grad: tensor([-5.2387e-08, -6.7707e-07, 1.4284e-07, 8.5391e-08, -1.1281e-07, + -1.0559e-07, -1.1659e-07, 5.0990e-08, 4.2375e-07, 3.6182e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 217.64, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4630 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.17 lr 0.00010000 +Epoch 350, weight, value: tensor([[ 0.0298, -0.0960, -0.0160, ..., -0.1489, -0.1703, -0.1641], + [-0.1417, 0.0785, -0.1414, ..., -0.1925, -0.2020, -0.2089], + [-0.1311, -0.1901, -0.2129, ..., -0.2244, -0.1630, 0.1737], + ..., + [ 0.0909, -0.0738, -0.1696, ..., 0.1134, -0.1752, -0.1537], + [ 0.1264, 0.1047, 0.0944, ..., -0.2119, -0.1238, 0.0921], + [ 0.0789, 0.0489, 0.1148, ..., 0.0534, -0.0945, -0.0037]], + device='cuda:0'), grad: tensor([[ 1.2955e-06, 1.6880e-09, 1.7462e-10, ..., 2.3320e-06, + 1.5483e-08, 4.1816e-07], + [ 9.5693e-08, 1.2806e-09, 5.2387e-10, ..., 1.7486e-07, + 2.2701e-09, 4.2492e-08], + [ 2.6135e-08, 1.3271e-08, 6.9849e-10, ..., -5.6054e-08, + 2.4447e-09, -1.7742e-07], + ..., + [ 3.3365e-07, 1.8335e-08, 2.3865e-09, ..., 6.3563e-07, + 6.9849e-10, 2.1874e-07], + [-1.5716e-08, -3.4459e-08, -1.8626e-09, ..., 2.0373e-08, + 6.9267e-09, -7.3924e-09], + [-1.9334e-06, -2.1188e-08, -5.5297e-09, ..., -3.4552e-06, + 1.2340e-08, -5.4622e-07]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0252, -0.0298, -0.0057, 0.0097, -0.0028, -0.0154, 0.0010, 0.0193, + -0.0271, 0.0121], device='cuda:0'), grad: tensor([ 4.4331e-06, 3.6322e-07, -3.2410e-07, 2.8696e-08, 6.5193e-07, + 1.2964e-06, -1.3476e-06, 1.3448e-06, -1.5192e-08, -6.4112e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 217.87, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4404 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.08 lr 0.00010000 +Epoch 351, weight, value: tensor([[ 0.0298, -0.0966, -0.0159, ..., -0.1490, -0.1706, -0.1644], + [-0.1419, 0.0788, -0.1414, ..., -0.1932, -0.2050, -0.2095], + [-0.1314, -0.1904, -0.2141, ..., -0.2244, -0.1631, 0.1738], + ..., + [ 0.0909, -0.0743, -0.1700, ..., 0.1134, -0.1755, -0.1538], + [ 0.1265, 0.1048, 0.0945, ..., -0.2120, -0.1242, 0.0923], + [ 0.0792, 0.0493, 0.1149, ..., 0.0542, -0.0941, -0.0028]], + device='cuda:0'), grad: tensor([[ 1.3795e-08, 2.5088e-08, 3.3178e-09, ..., 4.6508e-08, + 2.2643e-08, 2.8696e-08], + [ 3.9057e-08, 1.5541e-08, 6.9849e-09, ..., 4.9267e-07, + 7.4739e-08, 1.4016e-07], + [ 1.2224e-08, 1.1642e-06, 2.0373e-09, ..., 7.1861e-06, + 7.2364e-07, 2.0564e-06], + ..., + [ 2.4913e-08, 7.9686e-08, 1.3504e-08, ..., -9.1419e-06, + 6.5542e-08, -4.6357e-07], + [-1.1979e-07, -1.3746e-06, 6.4028e-10, ..., 3.9651e-07, + -6.9151e-07, -1.8468e-06], + [-8.6788e-08, -1.6019e-07, -7.9162e-08, ..., 1.7453e-06, + 9.7044e-07, 8.1677e-07]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0251, -0.0298, -0.0057, 0.0099, -0.0032, -0.0155, 0.0011, 0.0191, + -0.0271, 0.0127], device='cuda:0'), grad: tensor([ 2.3190e-07, 1.3541e-06, 2.2873e-05, 1.1083e-06, -4.5821e-06, + 5.7789e-07, 4.9779e-07, -2.0638e-05, -6.6087e-06, 5.1595e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 217.83, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4711 re_mapping 0.0026 re_causal 0.0089 /// teacc 99.15 lr 0.00010000 +Epoch 352, weight, value: tensor([[ 0.0299, -0.0970, -0.0162, ..., -0.1491, -0.1715, -0.1649], + [-0.1420, 0.0794, -0.1415, ..., -0.1936, -0.2053, -0.2096], + [-0.1319, -0.1907, -0.2150, ..., -0.2245, -0.1633, 0.1738], + ..., + [ 0.0910, -0.0751, -0.1708, ..., 0.1134, -0.1755, -0.1538], + [ 0.1266, 0.1050, 0.0945, ..., -0.2122, -0.1250, 0.0925], + [ 0.0794, 0.0494, 0.1150, ..., 0.0541, -0.0943, -0.0032]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 5.8208e-11, 3.4925e-10, ..., 1.2806e-09, + 4.5402e-09, 4.1910e-09], + [ 4.5402e-09, 1.7462e-09, 1.6880e-09, ..., 4.8312e-09, + 2.3865e-09, 3.8999e-09], + [ 3.5390e-08, 1.9209e-09, 7.5670e-10, ..., 3.1490e-08, + 1.7462e-09, -2.2119e-09], + ..., + [-4.0105e-08, 4.8312e-09, 9.3132e-10, ..., -2.8289e-08, + 4.4820e-09, 4.4238e-09], + [-1.1409e-08, -2.2701e-08, -1.6124e-08, ..., 4.8894e-09, + 5.7044e-09, -5.8790e-09], + [ 3.3178e-09, 5.1863e-08, 3.7253e-09, ..., 3.3644e-07, + 2.1281e-07, 1.4482e-07]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0248, -0.0296, -0.0058, 0.0101, -0.0031, -0.0155, 0.0012, 0.0190, + -0.0270, 0.0125], device='cuda:0'), grad: tensor([-1.0419e-08, 2.4156e-08, 1.4657e-07, -5.0059e-09, -6.8080e-07, + 1.9209e-09, -7.2760e-09, -1.4051e-07, 1.4727e-08, 6.8545e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 217.74, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4433 re_mapping 0.0025 re_causal 0.0082 /// teacc 99.23 lr 0.00010000 +Epoch 353, weight, value: tensor([[ 0.0298, -0.0966, -0.0163, ..., -0.1492, -0.1720, -0.1653], + [-0.1421, 0.0798, -0.1415, ..., -0.1939, -0.2055, -0.2097], + [-0.1326, -0.1912, -0.2157, ..., -0.2245, -0.1636, 0.1738], + ..., + [ 0.0905, -0.0759, -0.1717, ..., 0.1133, -0.1756, -0.1538], + [ 0.1267, 0.1051, 0.0946, ..., -0.2126, -0.1253, 0.0927], + [ 0.0804, 0.0499, 0.1150, ..., 0.0543, -0.0945, -0.0034]], + device='cuda:0'), grad: tensor([[-9.1386e-09, 2.6193e-09, 1.5716e-09, ..., 2.9104e-09, + 1.1059e-09, 1.6298e-09], + [ 3.7835e-09, -1.0477e-09, 1.9209e-09, ..., 4.4238e-09, + 7.5670e-10, 1.2165e-08], + [ 3.1432e-09, -1.2747e-08, 6.9849e-10, ..., 3.4925e-09, + 4.6566e-10, -6.1234e-08], + ..., + [ 1.9034e-08, 3.1258e-08, 1.7521e-08, ..., 3.1549e-08, + 3.4925e-10, 7.8580e-09], + [ 1.3795e-08, 1.9441e-08, 1.1001e-08, ..., 2.4796e-08, + 5.0641e-09, 4.8312e-09], + [-5.3726e-08, -5.4250e-08, -4.5460e-08, ..., -8.8417e-08, + 1.9791e-09, -6.1118e-09]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0249, -0.0293, -0.0058, 0.0100, -0.0030, -0.0161, 0.0018, 0.0188, + -0.0270, 0.0127], device='cuda:0'), grad: tensor([-7.9221e-08, 1.2130e-07, -7.0035e-07, 2.5169e-07, 5.8790e-08, + 2.0966e-07, -1.0710e-08, 1.6158e-07, 1.8638e-07, -1.8929e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 217.33, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4810 re_mapping 0.0025 re_causal 0.0087 /// teacc 99.11 lr 0.00010000 +Epoch 354, weight, value: tensor([[ 0.0286, -0.0974, -0.0169, ..., -0.1508, -0.1728, -0.1662], + [-0.1421, 0.0801, -0.1416, ..., -0.1946, -0.2058, -0.2099], + [-0.1329, -0.1914, -0.2161, ..., -0.2246, -0.1641, 0.1740], + ..., + [ 0.0909, -0.0762, -0.1726, ..., 0.1134, -0.1761, -0.1539], + [ 0.1267, 0.1052, 0.0947, ..., -0.2131, -0.1262, 0.0923], + [ 0.0817, 0.0503, 0.1152, ..., 0.0548, -0.0946, -0.0035]], + device='cuda:0'), grad: tensor([[-6.7521e-09, 1.4959e-08, 0.0000e+00, ..., 1.2224e-09, + 1.0477e-09, 1.8626e-09], + [ 2.5029e-09, 2.8405e-08, 5.8208e-11, ..., 5.0641e-09, + 1.5716e-09, 1.9441e-08], + [ 1.2806e-08, 1.9674e-08, 5.8208e-11, ..., 2.1188e-08, + 2.6193e-09, -1.3399e-07], + ..., + [-2.0606e-08, 3.4517e-08, 2.9104e-10, ..., -2.9278e-08, + 6.4028e-10, 8.5856e-08], + [ 5.1223e-09, 1.1880e-07, 5.2387e-10, ..., 1.2561e-07, + 1.0908e-07, 1.1327e-07], + [ 1.3388e-09, 1.0844e-07, -1.1642e-09, ..., 2.7264e-07, + 2.4866e-07, 2.1537e-07]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0238, -0.0293, -0.0058, 0.0084, -0.0029, -0.0165, 0.0022, 0.0190, + -0.0272, 0.0129], device='cuda:0'), grad: tensor([ 2.2119e-08, 1.3737e-07, -2.9709e-07, -1.5553e-06, -1.0338e-06, + 9.3039e-07, 1.6321e-07, 2.6310e-07, 6.5891e-07, 7.4832e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 217.52, cls_loss 0.0004 cls_loss_mapping 0.0014 cls_loss_causal 0.4371 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.12 lr 0.00010000 +Epoch 355, weight, value: tensor([[ 0.0287, -0.0973, -0.0165, ..., -0.1508, -0.1731, -0.1664], + [-0.1426, 0.0798, -0.1418, ..., -0.1972, -0.2061, -0.2101], + [-0.1329, -0.1914, -0.2163, ..., -0.2246, -0.1642, 0.1740], + ..., + [ 0.0912, -0.0762, -0.1732, ..., 0.1135, -0.1762, -0.1539], + [ 0.1267, 0.1052, 0.0947, ..., -0.2133, -0.1267, 0.0920], + [ 0.0818, 0.0510, 0.1154, ..., 0.0549, -0.0947, -0.0036]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 6.4028e-09, 0.0000e+00, ..., 6.2864e-09, + 2.9313e-07, 1.8475e-07], + [ 7.5670e-10, 1.9209e-09, 1.1642e-10, ..., 4.1910e-09, + 8.3237e-09, 7.2177e-09], + [ 3.4925e-10, 2.9686e-09, 0.0000e+00, ..., 5.6461e-09, + 2.2235e-08, 1.2631e-08], + ..., + [ 2.4447e-09, 7.7416e-09, 9.3132e-10, ..., 1.5891e-08, + 1.1059e-08, 6.4028e-09], + [-4.4820e-09, -2.6601e-08, -1.7462e-10, ..., 5.2969e-09, + 3.9407e-08, 1.2340e-08], + [-3.6089e-09, 1.6368e-07, -1.5716e-09, ..., 1.3998e-06, + 1.3774e-06, 6.8266e-07]], device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0238, -0.0297, -0.0058, 0.0080, -0.0030, -0.0164, 0.0022, 0.0192, + -0.0273, 0.0130], device='cuda:0'), grad: tensor([ 7.8790e-07, 3.1025e-08, 6.3446e-08, 4.1968e-08, -3.6079e-06, + 1.0635e-07, -1.1865e-06, 5.5181e-08, 2.2294e-08, 3.6974e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 217.46, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4855 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.09 lr 0.00010000 +Epoch 356, weight, value: tensor([[ 0.0294, -0.0958, -0.0149, ..., -0.1508, -0.1736, -0.1664], + [-0.1427, 0.0800, -0.1419, ..., -0.1980, -0.2065, -0.2105], + [-0.1331, -0.1916, -0.2170, ..., -0.2246, -0.1646, 0.1743], + ..., + [ 0.0905, -0.0773, -0.1766, ..., 0.1130, -0.1799, -0.1546], + [ 0.1270, 0.1053, 0.0947, ..., -0.2137, -0.1275, 0.0923], + [ 0.0820, 0.0519, 0.1159, ..., 0.0542, -0.0954, -0.0051]], + device='cuda:0'), grad: tensor([[-2.9104e-09, 2.0373e-09, 2.9104e-10, ..., 1.6298e-09, + 3.0850e-09, 4.0163e-09], + [ 8.3237e-09, 8.9058e-09, 1.5716e-09, ..., 2.2876e-08, + 1.6880e-09, 1.2107e-08], + [ 2.3283e-09, 1.2398e-08, 6.4028e-10, ..., -3.0268e-08, + 2.6484e-08, -5.5588e-08], + ..., + [-3.2654e-08, 2.4447e-08, 5.0641e-09, ..., -6.5193e-08, + 8.1491e-10, 1.9732e-08], + [-5.2969e-09, 1.0751e-07, 3.3760e-09, ..., 2.3283e-08, + 6.1700e-09, 2.9278e-08], + [-9.0222e-09, -1.4179e-07, -4.0920e-08, ..., -1.7276e-07, + 2.8871e-08, -5.4308e-08]], device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0244, -0.0297, -0.0057, 0.0061, -0.0020, -0.0163, 0.0022, 0.0187, + -0.0273, 0.0126], device='cuda:0'), grad: tensor([-1.5134e-09, 8.0443e-08, -1.9604e-07, -3.7579e-07, 5.3272e-07, + 5.5938e-08, -7.1421e-08, -3.9232e-08, 4.2235e-07, -3.9255e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 217.32, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4389 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.15 lr 0.00010000 +Epoch 357, weight, value: tensor([[ 0.0296, -0.0959, -0.0141, ..., -0.1508, -0.1736, -0.1665], + [-0.1429, 0.0798, -0.1419, ..., -0.1995, -0.2068, -0.2109], + [-0.1335, -0.1918, -0.2174, ..., -0.2246, -0.1646, 0.1746], + ..., + [ 0.0908, -0.0773, -0.1769, ..., 0.1130, -0.1800, -0.1549], + [ 0.1271, 0.1055, 0.0947, ..., -0.2139, -0.1276, 0.0926], + [ 0.0821, 0.0521, 0.1161, ..., 0.0543, -0.0954, -0.0052]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 3.7020e-08, 0.0000e+00, ..., 2.3283e-10, + 8.1491e-10, 1.3970e-09], + [ 1.7812e-07, 2.1346e-06, 1.1642e-10, ..., 1.0012e-08, + 2.3283e-10, 4.6217e-08], + [ 5.0641e-09, -8.7894e-09, 5.8208e-11, ..., 7.2177e-09, + 2.3283e-10, -1.1473e-07], + ..., + [-7.7416e-09, 4.1677e-08, 2.3283e-10, ..., -2.1537e-08, + 3.4925e-10, 3.0734e-08], + [-2.9313e-07, -3.5726e-06, -4.0745e-10, ..., 2.9686e-09, + 1.5134e-09, -2.8929e-08], + [ 2.0955e-09, 4.5984e-09, -4.6566e-10, ..., 3.6089e-09, + 2.3283e-09, 8.9640e-09]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0246, -0.0300, -0.0056, 0.0067, -0.0020, -0.0163, 0.0022, 0.0187, + -0.0273, 0.0126], device='cuda:0'), grad: tensor([ 1.1263e-07, 8.2776e-06, -3.3784e-07, 4.0606e-07, 1.1350e-08, + 4.1500e-06, 6.2864e-07, 1.6461e-07, -1.3508e-05, 5.7975e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 217.10, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4764 re_mapping 0.0024 re_causal 0.0080 /// teacc 99.21 lr 0.00010000 +Epoch 358, weight, value: tensor([[ 0.0298, -0.0960, -0.0140, ..., -0.1508, -0.1738, -0.1669], + [-0.1431, 0.0802, -0.1420, ..., -0.2000, -0.2071, -0.2111], + [-0.1340, -0.1922, -0.2180, ..., -0.2246, -0.1647, 0.1747], + ..., + [ 0.0910, -0.0778, -0.1774, ..., 0.1132, -0.1801, -0.1550], + [ 0.1271, 0.1057, 0.0947, ..., -0.2157, -0.1282, 0.0931], + [ 0.0823, 0.0525, 0.1165, ..., 0.0544, -0.0955, -0.0051]], + device='cuda:0'), grad: tensor([[-1.2503e-07, -3.2654e-08, 5.8208e-11, ..., -1.2224e-09, + 3.2014e-09, -3.2189e-08], + [ 4.1677e-08, -1.4273e-07, 4.6566e-10, ..., 1.2049e-08, + 7.4506e-09, -3.5216e-08], + [ 2.6193e-08, 8.0676e-08, 1.6880e-09, ..., 2.8114e-08, + 2.0547e-08, 4.9477e-09], + ..., + [ 1.1583e-08, 2.9220e-08, 1.6880e-09, ..., -6.3446e-09, + 3.7253e-09, 3.1665e-08], + [ 2.1944e-08, 4.8196e-08, -2.2119e-09, ..., 1.5134e-09, + 4.4820e-09, 4.2666e-08], + [ 8.9058e-09, 7.2177e-09, -2.5029e-09, ..., 4.0571e-08, + 3.3644e-08, 3.5914e-08]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0248, -0.0300, -0.0057, 0.0066, -0.0021, -0.0163, 0.0022, 0.0189, + -0.0277, 0.0127], device='cuda:0'), grad: tensor([-1.5656e-06, -5.5239e-08, 5.3830e-07, 6.1700e-07, -3.7951e-07, + -1.6550e-06, 9.9931e-07, 2.8079e-07, 8.7591e-07, 3.5809e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 217.18, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4834 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.11 lr 0.00010000 +Epoch 359, weight, value: tensor([[ 0.0300, -0.0962, -0.0141, ..., -0.1510, -0.1747, -0.1674], + [-0.1459, 0.0775, -0.1420, ..., -0.2013, -0.2075, -0.2131], + [-0.1354, -0.1927, -0.2190, ..., -0.2247, -0.1650, 0.1746], + ..., + [ 0.0882, -0.0809, -0.1783, ..., 0.1102, -0.1802, -0.1545], + [ 0.1271, 0.1058, 0.0947, ..., -0.2163, -0.1282, 0.0936], + [ 0.0820, 0.0524, 0.1166, ..., 0.0538, -0.0958, -0.0057]], + device='cuda:0'), grad: tensor([[ 7.1013e-09, 5.8208e-10, 2.9104e-10, ..., 1.6880e-09, + 7.3342e-09, 3.7544e-08], + [ 9.8196e-08, -1.1059e-09, 1.4552e-09, ..., 2.2468e-08, + 8.1491e-10, 5.7276e-07], + [-7.0548e-07, 5.8208e-10, 5.8208e-11, ..., -1.0047e-07, + 4.0745e-10, -2.8070e-06], + ..., + [ 3.3854e-07, -2.3807e-08, 4.7730e-09, ..., -1.0768e-07, + 5.2387e-10, 1.4640e-06], + [ 7.4506e-09, 1.8044e-09, 6.9849e-10, ..., 6.0536e-09, + 1.7870e-08, 5.0233e-08], + [ 1.8370e-07, -1.5483e-08, -1.0186e-08, ..., 1.9558e-08, + 7.3924e-09, 6.3051e-07]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0248, -0.0327, -0.0058, 0.0101, -0.0018, -0.0163, 0.0023, 0.0161, + -0.0277, 0.0121], device='cuda:0'), grad: tensor([ 1.3341e-07, 1.7323e-06, -8.7097e-06, 8.2608e-07, 3.5099e-08, + -3.5111e-07, 1.3364e-07, 3.9525e-06, 3.6531e-07, 1.8673e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 217.31, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4657 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.19 lr 0.00010000 +Epoch 360, weight, value: tensor([[ 0.0301, -0.0967, -0.0143, ..., -0.1511, -0.1752, -0.1678], + [-0.1460, 0.0782, -0.1420, ..., -0.2023, -0.2080, -0.2133], + [-0.1356, -0.1931, -0.2194, ..., -0.2247, -0.1654, 0.1749], + ..., + [ 0.0882, -0.0809, -0.1786, ..., 0.1103, -0.1803, -0.1547], + [ 0.1271, 0.1055, 0.0947, ..., -0.2171, -0.1290, 0.0934], + [ 0.0823, 0.0530, 0.1170, ..., 0.0537, -0.0961, -0.0059]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 3.2596e-09, 1.1642e-10, ..., 6.5775e-09, + 1.0827e-08, 1.3446e-08], + [ 1.1234e-08, -4.9418e-08, 1.0477e-09, ..., 5.9314e-08, + 2.5437e-08, 5.9139e-08], + [ 6.9267e-09, 5.7626e-09, 6.4028e-10, ..., 1.5018e-08, + 3.0268e-09, -1.8021e-07], + ..., + [ 3.0547e-06, 1.5264e-06, 5.8208e-10, ..., 9.3281e-06, + 5.6461e-09, 2.0606e-08], + [ 5.8208e-11, 4.0920e-08, -2.2701e-09, ..., 2.0082e-08, + 8.9826e-07, 8.2934e-07], + [-3.8110e-06, -1.8533e-06, 5.2387e-10, ..., -1.1548e-05, + 4.5809e-08, 3.5274e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0247, -0.0323, -0.0058, 0.0100, -0.0017, -0.0163, 0.0023, 0.0161, + -0.0282, 0.0120], device='cuda:0'), grad: tensor([ 6.9966e-08, 5.7102e-08, -6.1234e-07, 1.7602e-07, 3.7681e-06, + -2.4051e-07, -2.6263e-06, 1.6361e-05, 3.0752e-06, -1.9982e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 217.32, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4794 re_mapping 0.0024 re_causal 0.0078 /// teacc 99.21 lr 0.00010000 +Epoch 361, weight, value: tensor([[ 0.0303, -0.0972, -0.0146, ..., -0.1511, -0.1756, -0.1681], + [-0.1462, 0.0785, -0.1422, ..., -0.2032, -0.2085, -0.2136], + [-0.1349, -0.1932, -0.2204, ..., -0.2247, -0.1656, 0.1751], + ..., + [ 0.0882, -0.0809, -0.1789, ..., 0.1103, -0.1803, -0.1548], + [ 0.1272, 0.1051, 0.0949, ..., -0.2175, -0.1289, 0.0938], + [ 0.0825, 0.0538, 0.1174, ..., 0.0539, -0.0962, -0.0061]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 8.3528e-09, 4.5984e-09, ..., 5.9954e-09, + 4.1327e-09, 3.7835e-09], + [ 3.7544e-09, -4.7637e-07, 3.4343e-09, ..., 6.7230e-09, + 1.2224e-09, 1.0943e-08], + [ 2.4447e-09, 7.8289e-09, 1.7462e-09, ..., 2.4738e-09, + 7.8580e-10, -1.8568e-08], + ..., + [ 2.9977e-09, 1.2631e-08, 4.5111e-09, ..., 9.8953e-10, + 4.9477e-10, 7.7998e-09], + [ 2.3225e-08, 4.5612e-07, 1.5978e-08, ..., 3.7486e-08, + 1.0594e-08, -9.8953e-10], + [-9.0920e-08, -9.5344e-08, -9.5461e-08, ..., -1.2794e-07, + 1.9791e-09, -3.2072e-08]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0249, -0.0323, -0.0056, 0.0100, -0.0018, -0.0163, 0.0023, 0.0161, + -0.0286, 0.0121], device='cuda:0'), grad: tensor([ 2.4622e-08, -1.2806e-06, -1.9470e-08, 3.1461e-08, 2.8918e-07, + 1.0122e-07, -4.5344e-08, 3.9901e-08, 1.3635e-06, -4.8289e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 217.15, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4503 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.09 lr 0.00010000 +Epoch 362, weight, value: tensor([[ 0.0304, -0.0975, -0.0148, ..., -0.1512, -0.1758, -0.1684], + [-0.1463, 0.0785, -0.1422, ..., -0.2051, -0.2089, -0.2138], + [-0.1355, -0.1933, -0.2211, ..., -0.2248, -0.1659, 0.1752], + ..., + [ 0.0883, -0.0809, -0.1792, ..., 0.1104, -0.1804, -0.1549], + [ 0.1272, 0.1052, 0.0949, ..., -0.2178, -0.1291, 0.0940], + [ 0.0822, 0.0539, 0.1177, ..., 0.0536, -0.0965, -0.0064]], + device='cuda:0'), grad: tensor([[ 1.4261e-09, 2.4738e-09, 2.3283e-10, ..., 4.1910e-09, + 1.5367e-07, 8.0676e-08], + [ 4.3627e-08, -4.3516e-07, 2.9104e-10, ..., 3.5245e-08, + 2.6892e-08, -1.7055e-08], + [ 2.4535e-08, 5.4977e-08, 1.1642e-10, ..., 1.8539e-08, + 6.6066e-09, 6.8103e-09], + ..., + [-8.5507e-08, 3.0617e-07, 1.5425e-09, ..., -4.5286e-08, + 3.8417e-09, 2.8347e-08], + [ 9.1677e-09, 1.3591e-08, 2.6193e-10, ..., 1.1671e-08, + 2.0931e-07, 1.1036e-07], + [-2.5262e-08, -4.5315e-08, -1.6793e-08, ..., -2.7809e-06, + -1.0058e-06, -7.9302e-07]], device='cuda:0') +Epoch 362, bias, value: tensor([ 0.0248, -0.0324, -0.0058, 0.0100, -0.0017, -0.0162, 0.0023, 0.0161, + -0.0286, 0.0118], device='cuda:0'), grad: tensor([ 3.0780e-07, -9.0338e-07, 2.5798e-07, 2.5233e-08, 4.6343e-06, + 1.1576e-06, -2.1476e-06, 3.5996e-07, 4.9593e-07, -4.1761e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 217.32, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4373 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.15 lr 0.00010000 +Epoch 363, weight, value: tensor([[ 0.0304, -0.0952, -0.0149, ..., -0.1512, -0.1760, -0.1684], + [-0.1465, 0.0787, -0.1422, ..., -0.2059, -0.2098, -0.2143], + [-0.1364, -0.1935, -0.2218, ..., -0.2249, -0.1664, 0.1752], + ..., + [ 0.0883, -0.0809, -0.1795, ..., 0.1104, -0.1805, -0.1548], + [ 0.1274, 0.1055, 0.0950, ..., -0.2180, -0.1295, 0.0945], + [ 0.0822, 0.0539, 0.1178, ..., 0.0534, -0.0967, -0.0070]], + device='cuda:0'), grad: tensor([[ 1.0419e-08, 1.2195e-08, 2.0082e-09, ..., 2.6484e-09, + 8.7311e-11, 1.4086e-08], + [ 2.5320e-09, 2.6776e-09, 1.4552e-09, ..., 2.2992e-09, + 2.3283e-10, 3.2596e-09], + [ 9.8953e-10, 7.8580e-10, 1.1642e-10, ..., 8.7311e-10, + 4.6566e-10, -1.1642e-10], + ..., + [ 7.0431e-09, 6.0245e-09, 8.4401e-10, ..., 1.6880e-09, + 4.3656e-10, 6.9849e-09], + [-1.3679e-08, 6.1642e-08, 4.3336e-08, ..., 5.5588e-08, + 1.4552e-10, 1.0943e-08], + [-3.1403e-08, -1.1723e-07, -6.1118e-08, ..., -6.3621e-08, + 1.0565e-08, -5.0961e-08]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0259, -0.0324, -0.0058, 0.0100, -0.0014, -0.0161, 0.0022, 0.0161, + -0.0286, 0.0116], device='cuda:0'), grad: tensor([ 2.8958e-08, 1.0594e-08, 1.6007e-09, 2.6193e-09, 2.0460e-08, + 3.1781e-08, 2.1741e-08, 2.9395e-08, 7.1595e-08, -2.1723e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 217.26, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4609 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.08 lr 0.00010000 +Epoch 364, weight, value: tensor([[ 0.0304, -0.0956, -0.0153, ..., -0.1513, -0.1762, -0.1686], + [-0.1466, 0.0782, -0.1423, ..., -0.2080, -0.2101, -0.2155], + [-0.1366, -0.1938, -0.2230, ..., -0.2249, -0.1667, 0.1753], + ..., + [ 0.0874, -0.0809, -0.1821, ..., 0.1094, -0.1807, -0.1549], + [ 0.1272, 0.1041, 0.0946, ..., -0.2212, -0.1321, 0.0938], + [ 0.0867, 0.0583, 0.1195, ..., 0.0573, -0.0962, -0.0055]], + device='cuda:0'), grad: tensor([[ 1.8878e-06, 1.4703e-07, 1.0768e-09, ..., 3.3379e-06, + 2.9663e-07, 5.8813e-07], + [ 1.5483e-08, 2.6193e-09, 3.7544e-09, ..., 1.6880e-08, + 4.8894e-09, 1.5047e-08], + [ 1.1001e-08, 1.1874e-08, 3.3178e-09, ..., 1.2631e-08, + 4.0163e-09, 6.7230e-09], + ..., + [ 9.3132e-09, 1.0768e-08, 2.4447e-09, ..., 1.0536e-08, + 4.8894e-09, 1.2486e-08], + [-5.1339e-08, -9.3365e-08, -3.9028e-08, ..., 3.7835e-10, + -3.7689e-08, -1.0786e-07], + [-1.9502e-06, -1.3248e-07, 7.2469e-09, ..., -3.4627e-06, + -2.9779e-07, -5.8627e-07]], device='cuda:0') +Epoch 364, bias, value: tensor([ 0.0258, -0.0327, -0.0058, 0.0100, -0.0016, -0.0162, 0.0022, 0.0157, + -0.0293, 0.0156], device='cuda:0'), grad: tensor([ 6.2697e-06, 4.0483e-08, 5.2212e-08, 6.6939e-09, 1.7253e-07, + 1.1211e-07, 5.5850e-08, 4.9680e-08, -3.1595e-07, -6.4224e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 217.07, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4560 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.10 lr 0.00010000 +Epoch 365, weight, value: tensor([[ 0.0303, -0.0962, -0.0154, ..., -0.1515, -0.1801, -0.1724], + [-0.1467, 0.0783, -0.1424, ..., -0.2091, -0.2106, -0.2157], + [-0.1371, -0.1944, -0.2248, ..., -0.2249, -0.1675, 0.1752], + ..., + [ 0.0873, -0.0809, -0.1823, ..., 0.1092, -0.1807, -0.1549], + [ 0.1274, 0.1044, 0.0947, ..., -0.2213, -0.1322, 0.0944], + [ 0.0872, 0.0580, 0.1195, ..., 0.0577, -0.0964, -0.0058]], + device='cuda:0'), grad: tensor([[-2.3819e-07, -6.0827e-09, -1.4273e-07, ..., 1.8917e-09, + -1.1350e-09, 7.2177e-09], + [ 2.8755e-07, 9.0571e-08, 7.4215e-09, ..., 2.3935e-07, + 6.4028e-09, 9.5228e-08], + [ 1.4319e-07, 1.5250e-08, 7.8580e-09, ..., 2.0280e-07, + 1.6880e-09, 1.2107e-08], + ..., + [-3.2922e-07, -5.7626e-09, 1.8626e-09, ..., -5.7323e-07, + 2.7067e-09, 1.2980e-08], + [-3.7905e-07, -2.3295e-07, -1.3417e-08, ..., 1.9470e-08, + 1.7462e-10, -2.7427e-07], + [ 1.8440e-07, 7.6252e-09, 8.7661e-08, ..., 9.0338e-08, + 2.8056e-08, 2.2497e-08]], device='cuda:0') +Epoch 365, bias, value: tensor([ 0.0223, -0.0327, -0.0059, 0.0100, -0.0015, -0.0161, 0.0029, 0.0156, + -0.0293, 0.0160], device='cuda:0'), grad: tensor([-3.7458e-06, 1.7323e-06, 1.1828e-06, 3.2550e-07, -4.5809e-08, + 8.2562e-07, 1.1334e-06, -2.3823e-06, -1.5059e-06, 2.4978e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 217.54, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4364 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.11 lr 0.00010000 +Epoch 366, weight, value: tensor([[ 0.0303, -0.0964, -0.0154, ..., -0.1515, -0.1801, -0.1725], + [-0.1468, 0.0789, -0.1424, ..., -0.2091, -0.2109, -0.2160], + [-0.1383, -0.1953, -0.2259, ..., -0.2250, -0.1678, 0.1752], + ..., + [ 0.0873, -0.0810, -0.1825, ..., 0.1093, -0.1808, -0.1550], + [ 0.1275, 0.1045, 0.0947, ..., -0.2213, -0.1323, 0.0949], + [ 0.0871, 0.0578, 0.1195, ..., 0.0576, -0.0967, -0.0061]], + device='cuda:0'), grad: tensor([[ 4.3365e-08, 3.8708e-09, 5.2387e-10, ..., 6.4436e-08, + 1.8859e-08, 2.2788e-08], + [ 2.0955e-09, -1.6589e-08, 5.5297e-10, ..., 4.6857e-09, + 1.0477e-09, 5.4715e-09], + [ 5.5879e-09, 8.3528e-09, 4.0745e-10, ..., 9.8080e-09, + 1.7462e-09, -4.4238e-09], + ..., + [-1.0390e-08, 9.6043e-09, 5.8208e-10, ..., -2.4680e-08, + 1.8917e-09, 3.2596e-09], + [-8.7311e-11, 5.1892e-08, 1.1205e-08, ..., 1.9500e-08, + 1.5367e-08, 2.3021e-08], + [-6.9034e-08, -7.2690e-07, -2.6333e-07, ..., -4.7102e-07, + -2.7451e-07, -5.0012e-07]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0223, -0.0324, -0.0060, 0.0100, -0.0014, -0.0161, 0.0029, 0.0156, + -0.0293, 0.0158], device='cuda:0'), grad: tensor([ 1.6915e-07, -3.5798e-08, 2.0169e-08, -9.9943e-08, 2.1439e-06, + 4.1153e-08, 2.5728e-08, -4.4005e-08, 1.7439e-07, -2.3693e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 217.41, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4502 re_mapping 0.0025 re_causal 0.0086 /// teacc 99.08 lr 0.00010000 +Epoch 367, weight, value: tensor([[ 0.0303, -0.0994, -0.0155, ..., -0.1515, -0.1806, -0.1731], + [-0.1469, 0.0790, -0.1424, ..., -0.2096, -0.2117, -0.2165], + [-0.1390, -0.1958, -0.2269, ..., -0.2250, -0.1685, 0.1754], + ..., + [ 0.0874, -0.0810, -0.1827, ..., 0.1093, -0.1809, -0.1550], + [ 0.1276, 0.1017, 0.0948, ..., -0.2214, -0.1355, 0.0919], + [ 0.0870, 0.0576, 0.1196, ..., 0.0575, -0.0968, -0.0062]], + device='cuda:0'), grad: tensor([[-1.0477e-09, 2.3865e-09, 2.9104e-10, ..., 0.0000e+00, + 1.7462e-08, 1.3155e-08], + [ 1.3388e-09, -4.9185e-08, 7.5670e-10, ..., 5.8208e-11, + 2.0955e-09, -3.4925e-09], + [ 9.3132e-10, 8.0909e-09, 1.1059e-09, ..., 5.8208e-11, + 2.0373e-09, 2.6193e-09], + ..., + [ 5.8208e-10, 1.7928e-08, 9.8953e-10, ..., 3.4925e-10, + 2.9104e-10, 2.6776e-09], + [-6.4611e-09, 2.0373e-09, -4.0745e-10, ..., 0.0000e+00, + 6.0885e-08, 3.9814e-08], + [ 1.2806e-09, 1.4610e-08, 3.4925e-10, ..., 1.5716e-09, + 1.3970e-09, 3.0850e-09]], device='cuda:0') +Epoch 367, bias, value: tensor([ 0.0216, -0.0325, -0.0060, 0.0109, -0.0013, -0.0160, 0.0024, 0.0156, + -0.0321, 0.0156], device='cuda:0'), grad: tensor([ 4.3015e-08, -1.4901e-07, 2.8173e-08, -2.7416e-08, 5.0990e-08, + 5.3609e-08, -2.6636e-07, 5.7044e-08, 1.6438e-07, 6.2864e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 217.53, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4693 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.16 lr 0.00010000 +Epoch 368, weight, value: tensor([[ 0.0303, -0.0996, -0.0157, ..., -0.1516, -0.1807, -0.1732], + [-0.1470, 0.0793, -0.1425, ..., -0.2102, -0.2121, -0.2166], + [-0.1397, -0.1966, -0.2295, ..., -0.2251, -0.1690, 0.1755], + ..., + [ 0.0874, -0.0810, -0.1831, ..., 0.1094, -0.1811, -0.1551], + [ 0.1279, 0.1018, 0.0950, ..., -0.2215, -0.1356, 0.0920], + [ 0.0869, 0.0575, 0.1199, ..., 0.0573, -0.0976, -0.0068]], + device='cuda:0'), grad: tensor([[ 1.2806e-09, 1.8626e-09, 1.8044e-09, ..., 4.3074e-09, + 4.5402e-09, 2.9686e-09], + [ 1.6589e-08, -2.6543e-08, 1.5716e-09, ..., 3.7893e-08, + 1.1642e-09, 1.6298e-09], + [ 2.7358e-09, 5.8790e-09, 1.2806e-09, ..., 6.8685e-09, + 2.9104e-10, -3.0850e-09], + ..., + [-2.3749e-08, 1.5309e-08, 6.5775e-09, ..., -4.8196e-08, + 1.8044e-09, 2.3283e-09], + [ 1.8044e-09, 6.9849e-09, 1.5716e-09, ..., 6.2864e-09, + 1.1001e-08, 8.0327e-09], + [-1.8976e-08, -2.1188e-08, -3.5740e-08, ..., -7.2643e-08, + -1.4319e-08, -7.7998e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0215, -0.0323, -0.0060, 0.0109, -0.0011, -0.0160, 0.0024, 0.0156, + -0.0321, 0.0154], device='cuda:0'), grad: tensor([ 2.3982e-08, 2.7823e-08, -4.4238e-09, 7.7998e-09, 2.1618e-07, + -1.4727e-08, -6.9966e-08, -6.7754e-08, 8.7835e-08, -1.8859e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 218.02, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4492 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.22 lr 0.00010000 +Epoch 369, weight, value: tensor([[ 0.0304, -0.0997, -0.0163, ..., -0.1518, -0.1807, -0.1733], + [-0.1473, 0.0794, -0.1428, ..., -0.2111, -0.2127, -0.2171], + [-0.1405, -0.1969, -0.2303, ..., -0.2252, -0.1693, 0.1760], + ..., + [ 0.0875, -0.0810, -0.1836, ..., 0.1094, -0.1811, -0.1552], + [ 0.1282, 0.1019, 0.0958, ..., -0.2217, -0.1356, 0.0921], + [ 0.0869, 0.0575, 0.1198, ..., 0.0573, -0.0977, -0.0069]], + device='cuda:0'), grad: tensor([[-1.2806e-09, 3.8999e-09, 0.0000e+00, ..., 1.1642e-10, + 1.1118e-08, 9.7789e-09], + [ 1.8626e-09, -4.2026e-08, 2.3283e-10, ..., 1.0477e-09, + 6.8103e-09, 2.3341e-08], + [ 1.0477e-09, 1.7521e-08, 5.8208e-11, ..., 6.4028e-10, + 2.3865e-09, -2.0606e-08], + ..., + [ 1.9791e-09, 2.3923e-08, 1.7462e-10, ..., 7.5670e-10, + 5.8208e-10, 5.0059e-09], + [-7.8580e-09, -1.9791e-09, -9.8953e-10, ..., 2.9104e-10, + 1.5541e-08, 3.9581e-09], + [ 8.1491e-10, 1.0827e-08, 1.7462e-10, ..., 2.1595e-08, + 2.0256e-08, 1.2806e-08]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0215, -0.0325, -0.0059, 0.0109, -0.0014, -0.0161, 0.0025, 0.0157, + -0.0321, 0.0153], device='cuda:0'), grad: tensor([ 2.5379e-08, -5.6869e-08, -1.9500e-08, -6.2981e-08, -3.5041e-08, + 7.0955e-08, -1.0955e-07, 8.2597e-08, 3.3411e-08, 8.1083e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 217.72, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4782 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.21 lr 0.00010000 +Epoch 370, weight, value: tensor([[ 0.0276, -0.1006, -0.0169, ..., -0.1547, -0.1815, -0.1734], + [-0.1473, 0.0796, -0.1430, ..., -0.2117, -0.2138, -0.2175], + [-0.1407, -0.1970, -0.2312, ..., -0.2252, -0.1696, 0.1762], + ..., + [ 0.0875, -0.0810, -0.1847, ..., 0.1094, -0.1813, -0.1553], + [ 0.1284, 0.1020, 0.0961, ..., -0.2218, -0.1356, 0.0922], + [ 0.0880, 0.0576, 0.1201, ..., 0.0581, -0.0970, -0.0074]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 3.2538e-08, 1.8044e-08, ..., 1.8103e-08, + -8.7311e-10, 0.0000e+00], + [ 1.9674e-08, -6.9849e-10, 3.6671e-09, ..., 6.0711e-08, + 5.2387e-10, 2.3283e-10], + [ 1.0303e-08, 2.1537e-09, 8.7311e-10, ..., 2.7998e-08, + 5.2387e-10, 2.3283e-10], + ..., + [-3.9465e-08, 5.8033e-08, 2.3458e-08, ..., -1.5018e-07, + 4.0745e-10, 2.3283e-10], + [ 2.9337e-08, 6.4960e-08, 3.5740e-08, ..., 4.2666e-08, + 2.3283e-10, 1.1642e-10], + [-1.2130e-07, -3.4668e-07, -1.9651e-07, ..., -1.4470e-07, + 1.2573e-08, 5.5879e-09]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0196, -0.0324, -0.0058, 0.0109, -0.0011, -0.0162, 0.0026, 0.0157, + -0.0320, 0.0161], device='cuda:0'), grad: tensor([ 5.1805e-08, 1.3574e-07, 7.6659e-08, 3.9348e-07, 6.4203e-08, + 9.5402e-08, 2.2526e-08, -3.0035e-07, 1.8417e-07, -7.1293e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 217.69, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4357 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.11 lr 0.00010000 +Epoch 371, weight, value: tensor([[ 0.0272, -0.1007, -0.0195, ..., -0.1550, -0.1815, -0.1735], + [-0.1470, 0.0807, -0.1432, ..., -0.2120, -0.2145, -0.2178], + [-0.1407, -0.1972, -0.2326, ..., -0.2252, -0.1705, 0.1764], + ..., + [ 0.0875, -0.0811, -0.1856, ..., 0.1094, -0.1815, -0.1555], + [ 0.1290, 0.1022, 0.0967, ..., -0.2219, -0.1356, 0.0925], + [ 0.0882, 0.0571, 0.1203, ..., 0.0580, -0.0976, -0.0079]], + device='cuda:0'), grad: tensor([[-1.8757e-06, -8.0327e-09, 2.7940e-09, ..., -8.6287e-07, + -3.4794e-06, -1.3784e-06], + [ 1.0099e-07, 8.7894e-09, 3.6089e-09, ..., 7.7067e-08, + 2.3784e-07, 9.7381e-08], + [ 1.0518e-07, 2.0955e-09, 9.3132e-10, ..., 5.6811e-08, + 2.0023e-07, 8.1025e-08], + ..., + [ 1.1554e-07, 1.2747e-08, 6.7521e-09, ..., 5.3202e-08, + 2.0559e-07, 9.1153e-08], + [ 6.1002e-08, -3.6089e-08, -2.3225e-08, ..., 4.4878e-08, + 1.6065e-07, 2.2817e-08], + [ 7.5297e-07, -2.2817e-08, -1.4552e-08, ..., 3.8836e-07, + 1.5916e-06, 6.6310e-07]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0194, -0.0316, -0.0056, 0.0109, -0.0005, -0.0163, 0.0026, 0.0156, + -0.0318, 0.0160], device='cuda:0'), grad: tensor([-1.7866e-05, 1.0710e-06, 1.0114e-06, 2.3190e-06, -1.0841e-06, + 7.3900e-07, 4.4331e-06, 1.0887e-06, 6.1328e-07, 7.6517e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 217.79, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4556 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.11 lr 0.00010000 +Epoch 372, weight, value: tensor([[ 0.0282, -0.0992, -0.0202, ..., -0.1550, -0.1813, -0.1735], + [-0.1471, 0.0809, -0.1433, ..., -0.2126, -0.2150, -0.2182], + [-0.1410, -0.1968, -0.2332, ..., -0.2252, -0.1708, 0.1768], + ..., + [ 0.0875, -0.0812, -0.1861, ..., 0.1095, -0.1819, -0.1557], + [ 0.1291, 0.1022, 0.0968, ..., -0.2222, -0.1356, 0.0926], + [ 0.0878, 0.0564, 0.1207, ..., 0.0579, -0.0983, -0.0082]], + device='cuda:0'), grad: tensor([[-1.6182e-08, 3.4925e-10, 0.0000e+00, ..., 1.8626e-09, + -2.9104e-09, 2.3283e-09], + [ 9.0222e-09, -4.3656e-09, 5.8208e-11, ..., 1.6298e-08, + 4.7730e-09, 5.1223e-09], + [ 9.8953e-09, -4.0745e-10, 0.0000e+00, ..., 2.3167e-08, + 9.3132e-09, -1.5716e-08], + ..., + [-3.7660e-08, 3.4343e-09, 5.8208e-11, ..., 1.3504e-08, + 4.1560e-08, 2.4796e-08], + [ 7.3924e-09, 1.3388e-09, 0.0000e+00, ..., 7.1013e-09, + 1.2049e-08, 3.9581e-09], + [ 1.6647e-08, 9.3132e-10, 5.8208e-11, ..., 6.5367e-08, + 3.5565e-08, 1.5309e-08]], device='cuda:0') +Epoch 372, bias, value: tensor([ 2.0082e-02, -3.1623e-02, -5.3906e-03, 1.0850e-02, -2.4483e-05, + -1.6266e-02, 2.5046e-03, 1.5540e-02, -3.1824e-02, 1.5554e-02], + device='cuda:0'), grad: tensor([-5.4366e-08, 5.3202e-08, -4.9535e-08, 3.5274e-08, -3.2876e-07, + -4.1164e-07, 4.2887e-07, 6.9442e-08, 8.6322e-08, 1.9046e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 217.61, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4426 re_mapping 0.0024 re_causal 0.0078 /// teacc 99.09 lr 0.00010000 +Epoch 373, weight, value: tensor([[ 0.0286, -0.0990, -0.0207, ..., -0.1550, -0.1812, -0.1736], + [-0.1474, 0.0809, -0.1435, ..., -0.2136, -0.2161, -0.2189], + [-0.1419, -0.1968, -0.2343, ..., -0.2253, -0.1714, 0.1770], + ..., + [ 0.0875, -0.0812, -0.1869, ..., 0.1095, -0.1832, -0.1560], + [ 0.1295, 0.1023, 0.0973, ..., -0.2226, -0.1357, 0.0928], + [ 0.0877, 0.0566, 0.1211, ..., 0.0578, -0.0991, -0.0092]], + device='cuda:0'), grad: tensor([[-6.6007e-08, -4.8894e-09, 5.8208e-11, ..., 2.3283e-10, + 9.3132e-10, 7.5670e-10], + [ 1.5774e-08, -1.7276e-07, 1.2224e-09, ..., 1.1350e-08, + 1.1642e-10, 4.4238e-09], + [ 1.6298e-08, 8.2830e-08, 1.1642e-09, ..., 9.7207e-09, + 1.7462e-10, 1.5134e-09], + ..., + [-8.6380e-08, 9.1153e-08, 9.3132e-10, ..., -1.1508e-07, + 0.0000e+00, 4.3074e-09], + [-7.0431e-09, -1.2631e-08, -8.4401e-09, ..., 3.3760e-09, + 2.6776e-09, -2.5262e-08], + [ 1.0949e-07, 1.0710e-08, 1.9791e-09, ..., 8.9989e-08, + 1.1642e-10, 5.1223e-09]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0203, -0.0318, -0.0054, 0.0108, 0.0009, -0.0162, 0.0025, 0.0155, + -0.0318, 0.0152], device='cuda:0'), grad: tensor([-2.9453e-07, -7.6042e-07, 4.3237e-07, -8.6322e-08, 2.4040e-08, + 1.1042e-07, 1.5891e-08, 1.4168e-07, 1.6764e-08, 4.1374e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 217.57, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4403 re_mapping 0.0024 re_causal 0.0080 /// teacc 99.06 lr 0.00010000 +Epoch 374, weight, value: tensor([[ 0.0287, -0.0990, -0.0209, ..., -0.1551, -0.1813, -0.1737], + [-0.1475, 0.0812, -0.1437, ..., -0.2135, -0.2164, -0.2193], + [-0.1420, -0.1968, -0.2351, ..., -0.2253, -0.1718, 0.1773], + ..., + [ 0.0875, -0.0812, -0.1882, ..., 0.1095, -0.1832, -0.1561], + [ 0.1300, 0.1024, 0.0980, ..., -0.2227, -0.1358, 0.0929], + [ 0.0876, 0.0562, 0.1210, ..., 0.0577, -0.0993, -0.0097]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 8.4983e-09, 5.8208e-11, ..., 5.2387e-10, + 1.4552e-09, 3.4343e-09], + [ 2.4447e-09, -6.5798e-07, 4.6566e-10, ..., 6.4028e-10, + 1.5134e-09, -1.9395e-07], + [ 7.4506e-09, 2.4098e-08, 1.7462e-10, ..., 2.6543e-08, + 2.0955e-09, -1.3737e-08], + ..., + [-1.3388e-08, 1.1834e-07, 2.9104e-10, ..., -2.9919e-08, + 6.5775e-09, 3.9232e-08], + [-4.0163e-09, 6.0245e-08, -4.4820e-09, ..., 5.5879e-09, + -1.7462e-10, 1.9674e-08], + [ 1.8626e-09, 3.3760e-08, -5.8208e-10, ..., 1.3865e-07, + 1.0082e-07, 5.7684e-08]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0204, -0.0317, -0.0053, 0.0108, 0.0010, -0.0164, 0.0026, 0.0155, + -0.0318, 0.0150], device='cuda:0'), grad: tensor([ 4.9826e-08, -3.4831e-06, 1.1671e-07, 4.8662e-08, 1.7118e-06, + -6.0536e-09, 2.1723e-07, 5.4063e-07, 4.0419e-07, 4.0838e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 217.70, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4324 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.11 lr 0.00010000 +Epoch 375, weight, value: tensor([[ 0.0287, -0.0991, -0.0212, ..., -0.1551, -0.1813, -0.1737], + [-0.1476, 0.0816, -0.1438, ..., -0.2138, -0.2160, -0.2191], + [-0.1411, -0.1970, -0.2356, ..., -0.2253, -0.1720, 0.1776], + ..., + [ 0.0875, -0.0812, -0.1910, ..., 0.1095, -0.1832, -0.1563], + [ 0.1304, 0.1024, 0.0983, ..., -0.2230, -0.1359, 0.0928], + [ 0.0875, 0.0562, 0.1214, ..., 0.0576, -0.0995, -0.0099]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 2.0373e-09, 1.1642e-10, ..., 1.1642e-10, + 3.4925e-10, 5.2387e-10], + [ 1.6298e-09, -2.9337e-08, 5.2387e-10, ..., 1.5716e-09, + -1.1642e-10, -1.5716e-09], + [ 2.8522e-09, 5.4715e-09, 4.6566e-10, ..., 4.2492e-09, + 1.1642e-10, 5.8208e-11], + ..., + [-4.8894e-09, 8.3237e-09, 4.6566e-10, ..., -1.0477e-08, + -5.8208e-11, -1.7462e-10], + [-6.0536e-09, 9.3132e-10, -4.3074e-09, ..., 2.0955e-09, + 3.6089e-09, -5.2969e-09], + [ 1.5134e-09, 3.4925e-09, -2.3283e-10, ..., 2.9104e-09, + 1.6298e-09, 2.3865e-09]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0203, -0.0315, -0.0051, 0.0108, 0.0007, -0.0164, 0.0027, 0.0155, + -0.0318, 0.0149], device='cuda:0'), grad: tensor([ 5.6461e-09, -1.0774e-07, 2.1711e-08, 1.3108e-07, 5.2794e-08, + -1.9895e-07, 4.0745e-10, 1.5541e-08, 4.5460e-08, 3.0850e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 217.74, cls_loss 0.0004 cls_loss_mapping 0.0008 cls_loss_causal 0.4285 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.12 lr 0.00010000 +Epoch 376, weight, value: tensor([[ 0.0287, -0.0992, -0.0214, ..., -0.1551, -0.1813, -0.1738], + [-0.1477, 0.0819, -0.1438, ..., -0.2143, -0.2154, -0.2194], + [-0.1414, -0.1973, -0.2363, ..., -0.2254, -0.1722, 0.1779], + ..., + [ 0.0875, -0.0812, -0.1916, ..., 0.1096, -0.1833, -0.1564], + [ 0.1305, 0.1025, 0.0985, ..., -0.2233, -0.1359, 0.0928], + [ 0.0875, 0.0564, 0.1216, ..., 0.0576, -0.0997, -0.0100]], + device='cuda:0'), grad: tensor([[-6.9849e-10, 1.1059e-09, -4.0745e-10, ..., 4.2492e-09, + 4.5402e-09, 3.3760e-09], + [ 3.2014e-09, -1.3446e-08, 5.8208e-10, ..., 1.0186e-08, + 4.0163e-09, 3.7253e-09], + [ 1.7462e-09, 1.9209e-09, 1.7462e-10, ..., 1.9674e-08, + 1.3737e-08, 7.3342e-09], + ..., + [-4.8487e-08, 1.2049e-08, 2.3283e-09, ..., -1.5949e-07, + 5.8790e-09, 4.8312e-09], + [-3.8999e-09, -5.2387e-10, -1.0477e-09, ..., 8.1491e-09, + 9.4296e-09, 2.7940e-09], + [ 3.5390e-08, -1.4668e-08, -9.8953e-09, ..., 4.7288e-07, + 3.0501e-07, 1.9732e-07]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0203, -0.0314, -0.0049, 0.0108, 0.0007, -0.0164, 0.0027, 0.0155, + -0.0318, 0.0148], device='cuda:0'), grad: tensor([ 1.2806e-08, -3.1490e-08, 4.4296e-08, -4.7730e-09, -8.9267e-07, + 3.4343e-08, 2.7241e-08, -2.1362e-07, 3.1141e-08, 1.0142e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 217.54, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4376 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.04 lr 0.00010000 +Epoch 377, weight, value: tensor([[ 0.0287, -0.0993, -0.0215, ..., -0.1551, -0.1815, -0.1740], + [-0.1481, 0.0819, -0.1440, ..., -0.2164, -0.2165, -0.2200], + [-0.1419, -0.1980, -0.2388, ..., -0.2254, -0.1725, 0.1779], + ..., + [ 0.0875, -0.0813, -0.1933, ..., 0.1095, -0.1838, -0.1565], + [ 0.1308, 0.1025, 0.0986, ..., -0.2237, -0.1360, 0.0927], + [ 0.0877, 0.0577, 0.1224, ..., 0.0579, -0.0999, -0.0104]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 7.6834e-09, 1.7462e-10, ..., 8.7311e-10, + 1.6298e-09, 1.5716e-09], + [ 1.8335e-08, -5.0524e-08, 2.0955e-09, ..., 2.5611e-08, + 5.5879e-09, 8.0327e-09], + [ 8.1491e-09, 6.3446e-09, 1.8044e-09, ..., 1.4319e-08, + 8.4401e-09, 5.4133e-09], + ..., + [-5.7626e-08, 3.0035e-08, 2.3865e-09, ..., -1.0303e-07, + 3.6671e-09, 3.9581e-09], + [-1.8976e-08, 4.4005e-08, -1.0419e-08, ..., 2.2992e-08, + 4.1095e-08, -1.3737e-08], + [ 3.1490e-08, 7.7416e-09, -2.6776e-09, ..., 5.9197e-08, + 1.4959e-08, 7.9162e-09]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0202, -0.0316, -0.0050, 0.0108, 0.0012, -0.0165, 0.0028, 0.0155, + -0.0319, 0.0151], device='cuda:0'), grad: tensor([ 5.0350e-08, -1.6659e-07, 4.4936e-08, -3.6322e-08, 4.4238e-09, + -1.6205e-06, 5.8673e-07, -5.2678e-08, 9.0944e-07, 2.9174e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 217.83, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4724 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.14 lr 0.00010000 +Epoch 378, weight, value: tensor([[ 0.0289, -0.0994, -0.0217, ..., -0.1551, -0.1816, -0.1741], + [-0.1482, 0.0819, -0.1441, ..., -0.2177, -0.2178, -0.2206], + [-0.1427, -0.1983, -0.2395, ..., -0.2255, -0.1735, 0.1782], + ..., + [ 0.0876, -0.0814, -0.1966, ..., 0.1095, -0.1842, -0.1566], + [ 0.1312, 0.1026, 0.0989, ..., -0.2240, -0.1360, 0.0928], + [ 0.0876, 0.0577, 0.1225, ..., 0.0576, -0.1006, -0.0117]], + device='cuda:0'), grad: tensor([[-6.4028e-10, 6.9849e-10, 3.4925e-10, ..., 6.4028e-10, + 2.9104e-10, 4.0745e-10], + [ 5.9954e-09, 7.5670e-10, 1.6880e-09, ..., 1.7462e-09, + 6.4028e-10, 4.7730e-09], + [ 9.8953e-10, 1.9209e-09, 2.3283e-10, ..., 4.0745e-10, + 5.2387e-10, 1.0477e-09], + ..., + [ 5.0059e-09, 2.5495e-08, 3.5507e-09, ..., 6.9849e-09, + 0.0000e+00, 2.7940e-09], + [-6.5134e-08, -7.7300e-08, -1.2806e-08, ..., 4.8894e-09, + 1.3970e-09, -5.4366e-08], + [-2.0373e-09, 8.0909e-09, -8.9640e-09, ..., -2.6543e-08, + 1.3388e-09, 1.4843e-08]], device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0202, -0.0317, -0.0049, 0.0110, 0.0020, -0.0165, 0.0028, 0.0154, + -0.0319, 0.0147], device='cuda:0'), grad: tensor([-6.2282e-09, 1.5716e-09, 1.0943e-08, -2.8638e-08, 6.0943e-08, + 1.4366e-07, -1.7404e-08, 8.5915e-08, -2.0536e-07, -2.8173e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 217.81, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4796 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.12 lr 0.00010000 +Epoch 379, weight, value: tensor([[ 0.0291, -0.0994, -0.0221, ..., -0.1551, -0.1817, -0.1742], + [-0.1485, 0.0820, -0.1443, ..., -0.2189, -0.2186, -0.2212], + [-0.1438, -0.1984, -0.2401, ..., -0.2256, -0.1741, 0.1795], + ..., + [ 0.0876, -0.0814, -0.1970, ..., 0.1096, -0.1842, -0.1580], + [ 0.1314, 0.1027, 0.0992, ..., -0.2243, -0.1361, 0.0929], + [ 0.0876, 0.0577, 0.1227, ..., 0.0575, -0.1010, -0.0121]], + device='cuda:0'), grad: tensor([[ 2.0198e-08, 1.2224e-09, 4.0745e-10, ..., 1.1642e-09, + 3.3062e-07, 3.5577e-07], + [ 2.7940e-09, -5.1048e-08, 2.9104e-10, ..., 1.5716e-09, + 3.3353e-08, 2.0780e-08], + [ 2.0373e-09, 3.5507e-09, 5.8208e-11, ..., 2.5611e-09, + 1.9732e-08, 2.2002e-08], + ..., + [ 5.8208e-10, 7.8580e-09, 1.0477e-09, ..., -6.9849e-10, + 6.9849e-10, 2.4447e-09], + [ 3.1432e-09, 4.1910e-09, 4.0745e-10, ..., 1.9209e-09, + 5.5588e-08, 5.9896e-08], + [-8.0327e-09, -2.9104e-09, -6.8103e-09, ..., -1.3737e-08, + 1.2456e-08, 1.3621e-08]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0203, -0.0318, -0.0043, 0.0109, 0.0023, -0.0164, 0.0028, 0.0153, + -0.0320, 0.0146], device='cuda:0'), grad: tensor([ 8.7544e-07, -1.4086e-07, 7.4797e-08, 1.0536e-08, 2.9686e-07, + -2.0675e-07, -1.4063e-06, 2.9337e-08, 4.6426e-07, 1.6822e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 217.91, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4613 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.06 lr 0.00010000 +Epoch 380, weight, value: tensor([[ 0.0291, -0.0995, -0.0224, ..., -0.1552, -0.1817, -0.1743], + [-0.1489, 0.0821, -0.1444, ..., -0.2195, -0.2191, -0.2213], + [-0.1472, -0.1987, -0.2404, ..., -0.2259, -0.1750, 0.1792], + ..., + [ 0.0878, -0.0814, -0.1972, ..., 0.1096, -0.1843, -0.1577], + [ 0.1316, 0.1027, 0.0993, ..., -0.2245, -0.1361, 0.0929], + [ 0.0876, 0.0576, 0.1229, ..., 0.0573, -0.1015, -0.0130]], + device='cuda:0'), grad: tensor([[ 1.1059e-09, 5.0291e-08, 0.0000e+00, ..., 1.4552e-09, + 1.1642e-10, 1.1642e-10], + [ 4.8371e-08, -4.7521e-07, 2.9104e-10, ..., 6.4611e-08, + 5.8208e-11, 1.2806e-09], + [ 4.1851e-08, 2.4447e-08, 1.1642e-10, ..., 5.5705e-08, + 1.1642e-10, -8.1491e-10], + ..., + [-1.9395e-07, 1.9593e-07, 2.3283e-10, ..., -2.5937e-07, + 0.0000e+00, 6.9849e-10], + [ 2.6484e-08, 1.7323e-07, -8.1491e-10, ..., 3.5507e-08, + 1.1642e-10, -6.4028e-10], + [ 6.2922e-08, 4.5868e-08, 5.8208e-11, ..., 8.7311e-08, + 1.3970e-09, 1.3388e-09]], device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0203, -0.0319, -0.0046, 0.0109, 0.0025, -0.0161, 0.0027, 0.0154, + -0.0321, 0.0144], device='cuda:0'), grad: tensor([ 1.5681e-07, -1.1716e-06, 2.2398e-07, 1.1977e-06, 3.8533e-08, + -2.2110e-06, 1.9150e-07, -1.1321e-07, 1.3057e-06, 3.8836e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 217.52, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4437 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.01 lr 0.00010000 +Epoch 381, weight, value: tensor([[ 0.0292, -0.0998, -0.0231, ..., -0.1552, -0.1818, -0.1743], + [-0.1492, 0.0824, -0.1446, ..., -0.2201, -0.2199, -0.2216], + [-0.1498, -0.2036, -0.2427, ..., -0.2264, -0.1811, 0.1780], + ..., + [ 0.0879, -0.0814, -0.1976, ..., 0.1097, -0.1847, -0.1576], + [ 0.1324, 0.1029, 0.0997, ..., -0.2247, -0.1360, 0.0933], + [ 0.0876, 0.0577, 0.1232, ..., 0.0573, -0.1019, -0.0135]], + device='cuda:0'), grad: tensor([[ 1.5134e-09, 2.7358e-09, 1.0477e-09, ..., 2.3283e-10, + 1.7462e-10, 2.9686e-09], + [ 1.1583e-08, -1.9209e-09, 4.4238e-09, ..., 8.6729e-09, + 7.5670e-10, 2.0722e-08], + [ 9.5461e-09, 5.8208e-09, 2.2119e-09, ..., 1.1176e-08, + 6.4028e-10, -2.2608e-07], + ..., + [-1.9674e-08, 1.7113e-08, 2.7940e-09, ..., -3.9581e-08, + 4.6566e-10, 2.5379e-08], + [-6.2340e-08, -1.0338e-07, -4.3015e-08, ..., 1.3970e-09, + -3.3760e-09, -8.2888e-08], + [ 1.5483e-08, 1.0652e-08, 3.7835e-09, ..., 2.6368e-08, + 1.2107e-08, 2.0780e-08]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0203, -0.0319, -0.0064, 0.0111, 0.0031, -0.0158, 0.0025, 0.0154, + -0.0321, 0.0143], device='cuda:0'), grad: tensor([ 1.2864e-08, 5.4191e-08, -1.6689e-06, 1.3253e-06, -2.9104e-09, + 1.4948e-07, 9.0571e-08, 1.3446e-07, -2.0466e-07, 1.1409e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 217.32, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4673 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.13 lr 0.00010000 +Epoch 382, weight, value: tensor([[ 0.0293, -0.1002, -0.0244, ..., -0.1552, -0.1819, -0.1746], + [-0.1493, 0.0830, -0.1447, ..., -0.2205, -0.2210, -0.2234], + [-0.1491, -0.2038, -0.2431, ..., -0.2265, -0.1816, 0.1791], + ..., + [ 0.0878, -0.0815, -0.1989, ..., 0.1096, -0.1848, -0.1581], + [ 0.1326, 0.1027, 0.0998, ..., -0.2250, -0.1361, 0.0932], + [ 0.0878, 0.0586, 0.1239, ..., 0.0574, -0.1020, -0.0137]], + device='cuda:0'), grad: tensor([[ 1.4727e-08, 5.5297e-09, 0.0000e+00, ..., 2.9104e-08, + 9.0571e-08, 5.2562e-08], + [ 3.6263e-08, 9.8953e-10, 0.0000e+00, ..., 6.3388e-08, + 4.7730e-09, 5.7044e-09], + [ 1.2107e-08, 2.3283e-10, 0.0000e+00, ..., 2.1537e-08, + 1.6298e-09, -3.8999e-09], + ..., + [-1.1025e-07, 1.1642e-09, 0.0000e+00, ..., -1.9511e-07, + 1.7462e-10, 1.7462e-09], + [ 5.9372e-09, 5.7044e-09, 1.7462e-10, ..., 9.8953e-09, + 3.5274e-08, 2.3982e-08], + [ 3.0617e-08, 3.4925e-10, 0.0000e+00, ..., 5.1688e-08, + 9.3132e-10, 8.7311e-10]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0203, -0.0318, -0.0058, 0.0111, 0.0030, -0.0158, 0.0026, 0.0153, + -0.0323, 0.0145], device='cuda:0'), grad: tensor([ 3.0803e-07, 1.9115e-07, 5.0059e-08, 1.2107e-08, 1.5914e-07, + 3.0617e-07, -7.9256e-07, -5.0245e-07, 1.5087e-07, 1.4319e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 217.22, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4488 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.10 lr 0.00010000 +Epoch 383, weight, value: tensor([[ 0.0296, -0.1002, -0.0244, ..., -0.1552, -0.1819, -0.1747], + [-0.1495, 0.0831, -0.1448, ..., -0.2208, -0.2212, -0.2240], + [-0.1463, -0.2040, -0.2433, ..., -0.2262, -0.1817, 0.1799], + ..., + [ 0.0877, -0.0815, -0.1990, ..., 0.1096, -0.1848, -0.1586], + [ 0.1326, 0.1024, 0.0999, ..., -0.2251, -0.1361, 0.0931], + [ 0.0874, 0.0586, 0.1238, ..., 0.0574, -0.1022, -0.0144]], + device='cuda:0'), grad: tensor([[-3.6904e-08, 3.4925e-09, 0.0000e+00, ..., -3.7835e-09, + 5.9197e-08, 5.3551e-08], + [ 6.2282e-09, -6.3097e-08, 2.3283e-10, ..., 2.3865e-09, + 5.2387e-09, 5.7626e-09], + [ 4.3074e-09, 5.8790e-09, 5.8208e-11, ..., 2.7358e-09, + 1.9209e-09, 8.1491e-10], + ..., + [ 5.8208e-10, 2.3458e-08, 5.8208e-11, ..., -5.7044e-09, + 2.3283e-10, 6.9849e-10], + [ 4.0745e-10, 1.1642e-08, -9.3132e-10, ..., 9.3132e-10, + 5.2678e-08, 4.8254e-08], + [ 1.5891e-08, 8.0909e-09, 1.1642e-10, ..., 8.5565e-09, + 8.3237e-09, 4.1910e-09]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0205, -0.0319, -0.0051, 0.0111, 0.0029, -0.0158, 0.0026, 0.0152, + -0.0325, 0.0142], device='cuda:0'), grad: tensor([ 6.9907e-08, -4.3889e-07, 9.1095e-08, 1.6415e-08, 1.9965e-07, + 5.5414e-07, -1.0272e-06, 1.5949e-07, 2.5937e-07, 1.3527e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 217.72, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4571 re_mapping 0.0023 re_causal 0.0081 /// teacc 99.10 lr 0.00010000 +Epoch 384, weight, value: tensor([[ 0.0300, -0.1021, -0.0246, ..., -0.1552, -0.1823, -0.1747], + [-0.1496, 0.0864, -0.1450, ..., -0.2212, -0.2180, -0.2209], + [-0.1465, -0.2072, -0.2437, ..., -0.2262, -0.1849, 0.1770], + ..., + [ 0.0877, -0.0816, -0.1997, ..., 0.1097, -0.1854, -0.1587], + [ 0.1340, 0.1026, 0.1002, ..., -0.2254, -0.1361, 0.0934], + [ 0.0873, 0.0590, 0.1252, ..., 0.0574, -0.1022, -0.0144]], + device='cuda:0'), grad: tensor([[ 1.3923e-07, 1.3062e-07, 1.7055e-07, ..., 2.1153e-07, + 1.7521e-07, 6.4727e-08], + [ 1.0745e-07, 1.1036e-07, 9.6858e-08, ..., 4.5868e-08, + 1.5949e-08, 1.4761e-07], + [ 7.9162e-09, -6.9966e-08, 6.5193e-09, ..., 2.7940e-09, + 3.0268e-09, -4.0885e-07], + ..., + [ 5.1805e-08, 5.7509e-08, 4.7148e-08, ..., 2.7590e-08, + 5.4715e-09, 8.8708e-08], + [-1.7956e-06, -5.6438e-07, -1.4296e-06, ..., 2.4587e-07, + 3.8417e-09, -1.6792e-06], + [-3.2550e-07, -1.1800e-06, -6.0908e-07, ..., -9.7323e-07, + -2.0023e-07, 1.3155e-08]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0203, -0.0287, -0.0081, 0.0112, 0.0023, -0.0162, 0.0026, 0.0152, + -0.0323, 0.0143], device='cuda:0'), grad: tensor([ 7.7020e-07, 1.0272e-06, -1.3364e-06, 5.7649e-07, 5.3272e-07, + 1.3180e-05, 9.8906e-07, 5.4017e-07, -1.0088e-05, -6.1542e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 217.28, cls_loss 0.0005 cls_loss_mapping 0.0021 cls_loss_causal 0.4493 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.12 lr 0.00010000 +Epoch 385, weight, value: tensor([[ 0.0302, -0.1023, -0.0252, ..., -0.1552, -0.1822, -0.1748], + [-0.1500, 0.0864, -0.1453, ..., -0.2230, -0.2180, -0.2210], + [-0.1465, -0.2072, -0.2435, ..., -0.2263, -0.1849, 0.1771], + ..., + [ 0.0875, -0.0816, -0.2030, ..., 0.1098, -0.1854, -0.1599], + [ 0.1363, 0.1027, 0.1010, ..., -0.2260, -0.1362, 0.0940], + [ 0.0872, 0.0600, 0.1271, ..., 0.0575, -0.1022, -0.0143]], + device='cuda:0'), grad: tensor([[-3.1851e-07, 3.1432e-08, 1.3504e-08, ..., -8.0909e-08, + 2.3982e-08, 5.7626e-08], + [ 1.9209e-08, 1.9092e-08, 7.1013e-09, ..., 1.4785e-08, + 1.0594e-08, 3.0152e-08], + [ 2.6659e-08, 1.5891e-07, 1.0943e-08, ..., 1.1758e-08, + 1.8394e-08, 2.2934e-08], + ..., + [-1.9441e-08, 2.3516e-08, 4.6566e-09, ..., -3.4110e-08, + 5.0059e-09, 1.7579e-08], + [ 6.4727e-08, -5.9139e-08, -2.9919e-08, ..., 6.6590e-08, + 5.4017e-08, 5.3435e-08], + [-1.6298e-09, -5.9139e-08, -3.3178e-08, ..., -2.6892e-08, + 1.3737e-08, 2.7358e-08]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0203, -0.0288, -0.0080, 0.0112, 0.0018, -0.0163, 0.0026, 0.0152, + -0.0321, 0.0145], device='cuda:0'), grad: tensor([-7.6927e-07, 1.5995e-07, 7.3621e-07, -1.4398e-06, -1.3271e-08, + 1.6605e-06, -9.5135e-07, -5.4482e-08, 5.9837e-07, 6.7404e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 217.71, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4545 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.12 lr 0.00010000 +Epoch 386, weight, value: tensor([[ 0.0302, -0.1025, -0.0259, ..., -0.1552, -0.1823, -0.1749], + [-0.1506, 0.0864, -0.1455, ..., -0.2244, -0.2180, -0.2211], + [-0.1471, -0.2072, -0.2411, ..., -0.2264, -0.1849, 0.1773], + ..., + [ 0.0877, -0.0816, -0.2034, ..., 0.1099, -0.1855, -0.1598], + [ 0.1367, 0.1028, 0.1017, ..., -0.2264, -0.1363, 0.0940], + [ 0.0870, 0.0599, 0.1267, ..., 0.0575, -0.1024, -0.0152]], + device='cuda:0'), grad: tensor([[ 4.5402e-09, 6.7521e-09, 2.2119e-09, ..., 8.9640e-09, + 8.1491e-10, 1.1642e-09], + [ 2.0256e-08, 2.3283e-09, 5.7044e-09, ..., 4.9477e-08, + 1.7229e-08, 1.7695e-08], + [ 2.3167e-08, 4.1910e-09, 5.8208e-10, ..., 2.8871e-08, + 6.5193e-09, -6.2864e-09], + ..., + [ 2.9919e-08, 6.8685e-08, 2.6310e-08, ..., 3.8650e-08, + -7.5670e-09, -1.4552e-08], + [ 1.5949e-08, 1.1409e-08, 3.8417e-09, ..., 2.2352e-08, + 2.7940e-09, 3.0268e-09], + [-1.5844e-07, 2.0559e-07, -2.7823e-08, ..., 3.0454e-07, + 4.3167e-07, 1.8091e-07]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0202, -0.0289, -0.0080, 0.0112, 0.0017, -0.0164, 0.0027, 0.0153, + -0.0321, 0.0143], device='cuda:0'), grad: tensor([ 4.1793e-08, 7.6834e-08, 4.8312e-08, 3.2946e-08, -8.5728e-07, + 6.2049e-08, 9.3132e-09, 3.2829e-07, 1.0955e-07, 1.5239e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 217.88, cls_loss 0.0005 cls_loss_mapping 0.0020 cls_loss_causal 0.4523 re_mapping 0.0025 re_causal 0.0082 /// teacc 99.19 lr 0.00010000 +Epoch 387, weight, value: tensor([[ 0.0302, -0.1028, -0.0272, ..., -0.1553, -0.1826, -0.1752], + [-0.1510, 0.0866, -0.1457, ..., -0.2256, -0.2181, -0.2211], + [-0.1474, -0.2072, -0.2412, ..., -0.2265, -0.1849, 0.1774], + ..., + [ 0.0882, -0.0817, -0.2036, ..., 0.1101, -0.1857, -0.1599], + [ 0.1352, 0.1029, 0.1019, ..., -0.2290, -0.1363, 0.0941], + [ 0.0870, 0.0599, 0.1267, ..., 0.0572, -0.1035, -0.0164]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.1642e-10, 3.4925e-10, ..., 3.4925e-10, + 1.0361e-08, 7.2177e-09], + [ 6.9849e-10, -4.6566e-09, 2.3283e-10, ..., 1.0477e-09, + 8.1491e-10, 1.5134e-09], + [ 4.6566e-10, 5.8208e-10, 0.0000e+00, ..., 8.1491e-10, + 1.9791e-09, -1.0477e-09], + ..., + [-1.0477e-09, 2.5611e-09, 1.1642e-10, ..., -1.3970e-09, + 2.3283e-10, 6.9849e-10], + [-1.2922e-08, -1.4785e-08, -1.1409e-08, ..., 1.7462e-09, + 7.4506e-09, -2.5611e-09], + [-9.0804e-09, 1.5134e-09, -1.8626e-08, ..., -1.8277e-08, + 5.1223e-09, 3.9581e-09]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0201, -0.0288, -0.0080, 0.0111, 0.0024, -0.0164, 0.0028, 0.0155, + -0.0326, 0.0139], device='cuda:0'), grad: tensor([ 2.1886e-08, -1.9092e-08, 6.9849e-10, 1.3039e-08, 6.9733e-08, + -3.8533e-08, -4.1910e-08, 1.3388e-08, 1.5600e-08, -1.9209e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 217.96, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4547 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.15 lr 0.00010000 +Epoch 388, weight, value: tensor([[ 0.0302, -0.1028, -0.0276, ..., -0.1553, -0.1826, -0.1753], + [-0.1513, 0.0872, -0.1458, ..., -0.2264, -0.2181, -0.2212], + [-0.1472, -0.2072, -0.2413, ..., -0.2265, -0.1848, 0.1776], + ..., + [ 0.0884, -0.0818, -0.2033, ..., 0.1103, -0.1859, -0.1602], + [ 0.1353, 0.1029, 0.1021, ..., -0.2292, -0.1364, 0.0939], + [ 0.0866, 0.0600, 0.1264, ..., 0.0567, -0.1038, -0.0167]], + device='cuda:0'), grad: tensor([[ 4.0745e-09, 6.5193e-09, 3.3760e-09, ..., 9.3132e-10, + -1.1642e-09, -3.3760e-09], + [ 9.1968e-09, 6.9849e-09, 4.5402e-09, ..., 1.5134e-09, + 2.2119e-09, 8.1491e-09], + [ 2.5611e-09, 6.1700e-09, 1.7462e-09, ..., 4.6566e-10, + 1.8626e-09, 3.2596e-09], + ..., + [ 3.6787e-08, 1.8044e-08, 2.1304e-08, ..., 1.5134e-09, + 1.4203e-08, 3.0850e-08], + [-1.0780e-07, -4.0862e-08, -5.2736e-08, ..., 1.6298e-09, + -3.2131e-08, -9.1037e-08], + [ 6.5193e-09, -4.4238e-09, -9.1968e-09, ..., -2.8289e-08, + 2.9104e-09, 1.5367e-08]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0200, -0.0286, -0.0077, 0.0110, 0.0029, -0.0164, 0.0028, 0.0154, + -0.0328, 0.0135], device='cuda:0'), grad: tensor([-1.0349e-07, 5.7393e-08, 4.3539e-08, -4.2608e-08, 9.5693e-08, + 1.4121e-07, 4.1910e-08, 1.2049e-07, -3.3760e-07, 2.2119e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 217.69, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.5078 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.11 lr 0.00010000 +Epoch 389, weight, value: tensor([[ 0.0303, -0.1030, -0.0285, ..., -0.1553, -0.1828, -0.1755], + [-0.1520, 0.0872, -0.1461, ..., -0.2280, -0.2182, -0.2212], + [-0.1486, -0.2072, -0.2426, ..., -0.2269, -0.1849, 0.1781], + ..., + [ 0.0885, -0.0819, -0.2038, ..., 0.1105, -0.1863, -0.1617], + [ 0.1361, 0.1032, 0.1030, ..., -0.2294, -0.1384, 0.0933], + [ 0.0865, 0.0600, 0.1265, ..., 0.0566, -0.1042, -0.0172]], + device='cuda:0'), grad: tensor([[-3.4925e-09, 1.8626e-09, 2.3283e-10, ..., 1.1642e-10, + 6.5193e-09, 7.7998e-09], + [ 1.3155e-08, -9.5461e-09, 1.0710e-08, ..., 3.8417e-09, + 5.0059e-09, 3.4808e-08], + [ 4.7730e-09, 1.9441e-08, 7.7998e-09, ..., 3.3760e-09, + 5.5879e-09, 6.4028e-09], + ..., + [-1.0477e-09, 3.9698e-08, 6.2864e-09, ..., -9.3132e-09, + 2.3283e-10, 1.4435e-08], + [-3.6554e-08, -9.9069e-08, -4.2608e-08, ..., 3.7253e-09, + 1.5250e-08, -8.2189e-08], + [-1.1642e-10, 4.0745e-09, -1.5134e-09, ..., -4.7730e-09, + 2.4447e-09, 8.1491e-09]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0199, -0.0287, -0.0076, 0.0110, 0.0030, -0.0137, 0.0026, 0.0154, + -0.0357, 0.0132], device='cuda:0'), grad: tensor([-7.2177e-09, -3.8417e-08, 2.0373e-08, 1.1642e-09, 7.2992e-08, + 1.6147e-07, -1.5215e-07, 1.6065e-07, -2.1735e-07, 1.8394e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 217.31, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4598 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.22 lr 0.00010000 +Epoch 390, weight, value: tensor([[ 0.0305, -0.1033, -0.0295, ..., -0.1554, -0.1829, -0.1757], + [-0.1522, 0.0873, -0.1463, ..., -0.2289, -0.2182, -0.2213], + [-0.1493, -0.2073, -0.2430, ..., -0.2271, -0.1849, 0.1781], + ..., + [ 0.0886, -0.0819, -0.2045, ..., 0.1106, -0.1872, -0.1619], + [ 0.1365, 0.1033, 0.1034, ..., -0.2298, -0.1385, 0.0934], + [ 0.0867, 0.0610, 0.1277, ..., 0.0568, -0.1044, -0.0169]], + device='cuda:0'), grad: tensor([[ 5.0059e-09, 3.4925e-09, 9.3132e-10, ..., 1.1642e-10, + 7.5670e-09, 8.1491e-09], + [ 4.7032e-08, 3.1316e-08, 8.4983e-09, ..., 5.0059e-09, + 3.4925e-09, 1.9092e-08], + [ 4.0745e-09, 2.0955e-09, 5.8208e-10, ..., 2.6776e-09, + 3.1432e-09, -4.6566e-10], + ..., + [ 1.1642e-07, 8.4285e-08, 2.2701e-08, ..., -8.4983e-09, + 4.6566e-10, 4.2142e-08], + [-2.8405e-07, -1.9534e-07, -5.2969e-08, ..., 3.4925e-10, + 8.6147e-09, -8.8359e-08], + [ 3.7835e-08, 2.4447e-08, 6.9849e-09, ..., 1.1525e-08, + 6.0536e-09, 1.4319e-08]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0199, -0.0287, -0.0076, 0.0110, 0.0027, -0.0137, 0.0027, 0.0154, + -0.0357, 0.0135], device='cuda:0'), grad: tensor([ 3.4226e-08, 1.7369e-07, 9.8953e-09, 4.7963e-08, 2.2119e-09, + 2.4354e-07, -1.0408e-07, 3.8138e-07, -9.1549e-07, 1.4226e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 217.63, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4706 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.11 lr 0.00010000 +Epoch 391, weight, value: tensor([[ 0.0305, -0.1034, -0.0299, ..., -0.1554, -0.1829, -0.1758], + [-0.1527, 0.0873, -0.1465, ..., -0.2305, -0.2183, -0.2213], + [-0.1495, -0.2074, -0.2432, ..., -0.2273, -0.1849, 0.1782], + ..., + [ 0.0884, -0.0840, -0.2051, ..., 0.1106, -0.1874, -0.1620], + [ 0.1368, 0.1033, 0.1043, ..., -0.2300, -0.1385, 0.0934], + [ 0.0867, 0.0609, 0.1282, ..., 0.0568, -0.1045, -0.0172]], + device='cuda:0'), grad: tensor([[-2.9686e-08, 2.3283e-10, 8.1491e-10, ..., -2.1420e-08, + 8.1491e-10, 1.3970e-09], + [ 1.6135e-07, 6.2864e-09, 9.3132e-10, ..., 1.6287e-07, + 4.7381e-08, 1.1444e-07], + [ 1.5367e-08, -5.9372e-09, 1.1642e-10, ..., 1.2573e-08, + 3.4925e-10, -1.3993e-07], + ..., + [-5.7230e-07, 3.2596e-09, 2.3283e-09, ..., -4.8429e-07, + 4.6566e-10, 1.3039e-08], + [-2.9104e-09, -2.7008e-08, -1.0012e-08, ..., 1.1292e-08, + -2.3283e-10, -1.8743e-08], + [ 6.5891e-08, -1.1805e-07, -4.0187e-07, ..., -2.0943e-07, + -2.9569e-07, -8.9407e-08]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0198, -0.0288, -0.0077, 0.0136, 0.0026, -0.0137, 0.0028, 0.0128, + -0.0357, 0.0134], device='cuda:0'), grad: tensor([-1.2352e-07, 1.1884e-06, -7.5391e-07, 3.1129e-07, 1.2200e-06, + -1.8510e-08, 9.9069e-08, -2.2557e-06, 3.8883e-08, 3.2876e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 217.65, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4427 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.03 lr 0.00010000 +Epoch 392, weight, value: tensor([[ 0.0305, -0.1034, -0.0304, ..., -0.1554, -0.1830, -0.1760], + [-0.1532, 0.0873, -0.1468, ..., -0.2320, -0.2183, -0.2214], + [-0.1500, -0.2074, -0.2434, ..., -0.2274, -0.1850, 0.1783], + ..., + [ 0.0886, -0.0840, -0.2054, ..., 0.1108, -0.1875, -0.1620], + [ 0.1379, 0.1036, 0.1060, ..., -0.2302, -0.1385, 0.0939], + [ 0.0863, 0.0604, 0.1277, ..., 0.0566, -0.1048, -0.0179]], + device='cuda:0'), grad: tensor([[-1.6415e-08, 1.0477e-09, 8.1491e-10, ..., 6.9849e-10, + 1.1642e-09, 1.5134e-09], + [ 1.4435e-08, 7.5670e-09, 5.7044e-09, ..., 8.9640e-09, + 8.1491e-09, 2.0256e-08], + [ 3.2596e-09, 4.5402e-09, 2.0955e-09, ..., 3.3760e-09, + -3.3760e-09, -1.4203e-08], + ..., + [-7.3342e-09, 4.7730e-09, 3.7253e-09, ..., -1.7579e-08, + 3.4925e-09, 8.8476e-09], + [-6.4145e-08, -6.6473e-08, -4.0280e-08, ..., 1.1642e-09, + -3.0035e-08, -1.1176e-07], + [ 2.6892e-08, 1.3271e-08, 7.3342e-09, ..., 1.1513e-07, + 1.0547e-07, 7.1828e-08]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0198, -0.0289, -0.0077, 0.0135, 0.0026, -0.0137, 0.0030, 0.0128, + -0.0356, 0.0128], device='cuda:0'), grad: tensor([-7.6136e-08, 8.4634e-08, -1.3970e-08, 1.5018e-08, -2.9011e-07, + 6.4145e-08, 1.5472e-07, -1.6997e-08, -3.1246e-07, 4.0838e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 217.48, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4731 re_mapping 0.0024 re_causal 0.0085 /// teacc 99.01 lr 0.00010000 +Epoch 393, weight, value: tensor([[ 0.0306, -0.1034, -0.0309, ..., -0.1554, -0.1830, -0.1761], + [-0.1536, 0.0874, -0.1469, ..., -0.2334, -0.2185, -0.2216], + [-0.1501, -0.2074, -0.2436, ..., -0.2276, -0.1850, 0.1786], + ..., + [ 0.0888, -0.0840, -0.2057, ..., 0.1111, -0.1877, -0.1621], + [ 0.1380, 0.1037, 0.1063, ..., -0.2304, -0.1386, 0.0936], + [ 0.0860, 0.0610, 0.1289, ..., 0.0565, -0.1049, -0.0185]], + device='cuda:0'), grad: tensor([[ 1.5134e-09, 6.9849e-10, 6.9849e-10, ..., 1.3970e-09, + 2.3283e-10, 3.8417e-09], + [ 8.1491e-10, -2.8056e-08, 1.1642e-10, ..., 8.1491e-10, + 0.0000e+00, 1.1292e-08], + [ 5.8208e-10, 3.3760e-09, 0.0000e+00, ..., 4.6566e-10, + 1.1642e-10, -3.7835e-08], + ..., + [ 1.1642e-10, 2.7940e-09, 0.0000e+00, ..., 2.3283e-10, + 0.0000e+00, 8.9640e-09], + [ 1.5134e-09, 1.8859e-08, 6.9849e-10, ..., 1.6298e-09, + 3.4925e-10, 1.0477e-08], + [-1.5018e-08, 2.3283e-10, -6.8685e-09, ..., -1.3155e-08, + 1.0477e-09, -2.6776e-08]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0200, -0.0289, -0.0077, 0.0136, 0.0020, -0.0137, 0.0030, 0.0129, + -0.0356, 0.0122], device='cuda:0'), grad: tensor([ 1.6764e-08, -5.9488e-08, -1.3318e-07, 2.9686e-08, 7.6718e-08, + 2.2119e-09, 1.8976e-08, 4.5053e-08, 1.1246e-07, -9.1968e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 217.61, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4562 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.05 lr 0.00010000 +Epoch 394, weight, value: tensor([[ 0.0314, -0.1033, -0.0294, ..., -0.1554, -0.1827, -0.1753], + [-0.1543, 0.0876, -0.1472, ..., -0.2359, -0.2186, -0.2217], + [-0.1511, -0.2076, -0.2440, ..., -0.2280, -0.1850, 0.1789], + ..., + [ 0.0889, -0.0841, -0.2071, ..., 0.1113, -0.1882, -0.1626], + [ 0.1389, 0.1038, 0.1078, ..., -0.2309, -0.1387, 0.0932], + [ 0.0861, 0.0615, 0.1288, ..., 0.0566, -0.1052, -0.0196]], + device='cuda:0'), grad: tensor([[-1.1642e-10, 2.3283e-10, 1.1642e-10, ..., 3.4925e-10, + 4.6566e-10, 6.9849e-10], + [ 1.4785e-08, 3.3760e-09, 1.8626e-09, ..., 2.4331e-08, + 1.3970e-09, 1.7579e-08], + [ 2.0256e-08, -4.6566e-09, 6.9849e-10, ..., 3.7835e-08, + 4.6566e-10, -1.8626e-08], + ..., + [-4.3074e-08, 7.2177e-09, 1.2806e-09, ..., -8.2422e-08, + 3.4925e-10, 7.3342e-09], + [-8.6147e-09, -2.2817e-08, -1.3155e-08, ..., 8.1491e-09, + -2.2119e-09, -2.3632e-08], + [ 3.2596e-09, 8.1491e-10, 2.3283e-10, ..., 3.1432e-09, + 1.3970e-09, 2.7940e-09]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0207, -0.0285, -0.0078, 0.0136, 0.0018, -0.0137, 0.0023, 0.0129, + -0.0358, 0.0120], device='cuda:0'), grad: tensor([-1.1642e-10, 8.6264e-08, 7.5204e-08, 2.7707e-08, 2.1420e-08, + 6.4960e-08, -1.8626e-08, -2.1479e-07, -4.6799e-08, 2.3516e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 217.83, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4489 re_mapping 0.0024 re_causal 0.0080 /// teacc 99.13 lr 0.00010000 +Epoch 395, weight, value: tensor([[ 0.0315, -0.1035, -0.0301, ..., -0.1555, -0.1828, -0.1756], + [-0.1548, 0.0876, -0.1477, ..., -0.2377, -0.2188, -0.2218], + [-0.1522, -0.2076, -0.2444, ..., -0.2284, -0.1850, 0.1790], + ..., + [ 0.0890, -0.0841, -0.2082, ..., 0.1114, -0.1883, -0.1624], + [ 0.1388, 0.1039, 0.1082, ..., -0.2322, -0.1388, 0.0933], + [ 0.0865, 0.0625, 0.1299, ..., 0.0569, -0.1053, -0.0197]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.0477e-09, 0.0000e+00, ..., 2.3283e-10, + 1.9325e-08, 1.7113e-08], + [ 2.3283e-09, 1.0128e-08, 0.0000e+00, ..., 1.3970e-09, + 4.0745e-09, 5.8208e-09], + [ 1.8626e-09, 5.2387e-09, 0.0000e+00, ..., 2.2119e-09, + 6.9849e-10, -6.0536e-09], + ..., + [-4.7730e-09, 1.3853e-08, 1.1642e-10, ..., -4.5402e-09, + 1.1642e-10, 6.0536e-09], + [-3.4925e-09, -1.5949e-08, 0.0000e+00, ..., 4.6566e-10, + 3.2829e-08, 2.3749e-08], + [-4.8894e-09, 1.1642e-10, -1.5134e-09, ..., -8.4983e-09, + 0.0000e+00, 8.1491e-10]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0206, -0.0286, -0.0078, 0.0136, 0.0014, -0.0139, 0.0027, 0.0130, + -0.0360, 0.0124], device='cuda:0'), grad: tensor([ 4.1793e-08, 4.0513e-08, -4.5402e-09, -1.1479e-07, 1.9232e-07, + 1.3225e-07, -3.6648e-07, 4.1793e-08, 5.1223e-08, -1.0361e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 217.62, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4593 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.09 lr 0.00010000 +Epoch 396, weight, value: tensor([[ 0.0318, -0.1035, -0.0302, ..., -0.1555, -0.1828, -0.1756], + [-0.1559, 0.0877, -0.1479, ..., -0.2387, -0.2188, -0.2219], + [-0.1521, -0.2076, -0.2448, ..., -0.2284, -0.1850, 0.1793], + ..., + [ 0.0892, -0.0841, -0.2084, ..., 0.1114, -0.1883, -0.1625], + [ 0.1389, 0.1036, 0.1085, ..., -0.2326, -0.1389, 0.0932], + [ 0.0865, 0.0624, 0.1301, ..., 0.0570, -0.1053, -0.0201]], + device='cuda:0'), grad: tensor([[ 7.3342e-09, 1.0594e-08, 2.9104e-09, ..., 2.2934e-08, + 5.3551e-09, 1.1758e-08], + [ 1.8510e-08, 2.2119e-09, 1.1176e-08, ..., 2.2585e-08, + 5.0059e-09, 2.1188e-08], + [ 4.8894e-08, 4.2375e-08, 2.9104e-09, ..., 1.3784e-07, + 2.8289e-08, 3.2713e-08], + ..., + [ 4.0745e-09, 1.8510e-08, 3.7253e-09, ..., 2.9220e-08, + 9.6625e-09, 2.0722e-08], + [-3.5507e-08, -6.0652e-08, -7.5903e-08, ..., 7.2177e-09, + -1.0594e-08, -7.6019e-08], + [-9.4995e-08, -7.7882e-08, 8.0327e-09, ..., -2.9220e-07, + -5.4366e-08, -8.9058e-08]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0207, -0.0285, -0.0077, 0.0142, 0.0011, -0.0152, 0.0027, 0.0130, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 7.0897e-08, 9.0338e-08, 3.1339e-07, 2.8522e-08, 1.6694e-07, + 9.3714e-08, 1.1758e-07, 7.8813e-08, -2.9616e-07, -6.5472e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 217.47, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4416 re_mapping 0.0021 re_causal 0.0075 /// teacc 99.07 lr 0.00010000 +Epoch 397, weight, value: tensor([[ 0.0320, -0.1033, -0.0302, ..., -0.1555, -0.1827, -0.1756], + [-0.1564, 0.0878, -0.1481, ..., -0.2396, -0.2188, -0.2220], + [-0.1521, -0.2076, -0.2451, ..., -0.2284, -0.1850, 0.1796], + ..., + [ 0.0892, -0.0841, -0.2087, ..., 0.1115, -0.1885, -0.1627], + [ 0.1390, 0.1037, 0.1088, ..., -0.2329, -0.1390, 0.0932], + [ 0.0865, 0.0626, 0.1303, ..., 0.0569, -0.1059, -0.0206]], + device='cuda:0'), grad: tensor([[-5.5879e-09, 5.8208e-10, 1.1642e-10, ..., 7.9162e-09, + 7.6834e-09, 6.0536e-09], + [ 5.7044e-09, -1.0477e-08, 2.5611e-09, ..., 3.3411e-08, + 3.6089e-08, 3.3411e-08], + [ 4.0745e-09, 2.2119e-09, 3.4925e-10, ..., 1.4668e-08, + 1.0012e-08, -7.3691e-08], + ..., + [-6.0536e-09, 1.1176e-08, 1.0477e-09, ..., 2.9104e-09, + 1.2689e-08, 1.8510e-08], + [-9.5461e-09, -1.2340e-08, -8.2655e-09, ..., 2.5611e-09, + 1.6298e-09, 2.9220e-08], + [ 1.2806e-09, -9.3132e-10, -2.3283e-09, ..., 7.6601e-07, + 7.2177e-07, 4.7428e-07]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0210, -0.0285, -0.0075, 0.0141, 0.0013, -0.0151, 0.0026, 0.0130, + -0.0364, 0.0121], device='cuda:0'), grad: tensor([ 8.1491e-09, 5.6927e-08, -2.3225e-07, 3.6205e-08, -2.5611e-06, + 2.5728e-08, 1.6659e-07, 1.4226e-07, 1.2899e-07, 2.2203e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 217.41, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4503 re_mapping 0.0022 re_causal 0.0078 /// teacc 99.16 lr 0.00010000 +Epoch 398, weight, value: tensor([[ 0.0321, -0.1037, -0.0305, ..., -0.1555, -0.1835, -0.1763], + [-0.1582, 0.0879, -0.1484, ..., -0.2414, -0.2189, -0.2221], + [-0.1533, -0.2077, -0.2457, ..., -0.2287, -0.1851, 0.1796], + ..., + [ 0.0897, -0.0842, -0.2089, ..., 0.1116, -0.1888, -0.1627], + [ 0.1392, 0.1037, 0.1091, ..., -0.2334, -0.1393, 0.0928], + [ 0.0863, 0.0658, 0.1334, ..., 0.0595, -0.1033, -0.0181]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 3.4925e-10, 0.0000e+00, ..., 1.7462e-09, + 7.6136e-08, 6.1467e-08], + [ 2.6892e-08, 9.6625e-09, 1.2806e-09, ..., 3.4459e-08, + 1.7928e-08, 1.5018e-08], + [ 1.6065e-08, 3.4925e-10, 0.0000e+00, ..., 1.8859e-08, + 4.8894e-09, 4.0745e-09], + ..., + [-1.3097e-07, 7.6834e-09, 9.3132e-10, ..., -1.5367e-07, + 5.8208e-10, 5.8208e-10], + [ 1.8976e-08, 1.1642e-09, 2.3283e-10, ..., 2.3516e-08, + 1.6671e-07, 1.3399e-07], + [ 4.9593e-08, -1.6997e-08, -2.4447e-09, ..., 5.3085e-08, + 6.1700e-09, 5.0059e-09]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0203, -0.0287, -0.0076, 0.0141, -0.0017, -0.0152, 0.0034, 0.0132, + -0.0366, 0.0145], device='cuda:0'), grad: tensor([ 2.1281e-07, 1.4796e-07, 6.3796e-08, 2.6776e-08, 1.3364e-07, + 3.4068e-06, -4.2655e-06, -3.8603e-07, 5.1782e-07, 1.4377e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 217.69, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4169 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.06 lr 0.00010000 +Epoch 399, weight, value: tensor([[ 0.0322, -0.1037, -0.0306, ..., -0.1555, -0.1838, -0.1766], + [-0.1590, 0.0879, -0.1486, ..., -0.2420, -0.2189, -0.2221], + [-0.1536, -0.2077, -0.2461, ..., -0.2288, -0.1851, 0.1796], + ..., + [ 0.0898, -0.0842, -0.2092, ..., 0.1117, -0.1899, -0.1628], + [ 0.1406, 0.1043, 0.1105, ..., -0.2336, -0.1394, 0.0937], + [ 0.0859, 0.0654, 0.1329, ..., 0.0591, -0.1038, -0.0192]], + device='cuda:0'), grad: tensor([[-1.0477e-09, 1.0710e-08, 2.3283e-10, ..., 2.3283e-10, + 4.0745e-09, 3.7253e-09], + [ 7.9162e-09, -2.3842e-07, 2.0955e-09, ..., 1.1642e-09, + 3.6089e-09, 1.0128e-08], + [ 1.1059e-08, 2.2002e-08, 5.0059e-09, ..., 4.6566e-10, + 9.4296e-09, 1.9674e-08], + ..., + [ 1.2806e-08, 1.8883e-07, 6.1700e-09, ..., 1.6182e-08, + 1.1642e-10, 5.7044e-09], + [-7.0315e-08, -7.6252e-08, -2.6077e-08, ..., 2.2119e-09, + -5.3085e-08, -1.1956e-07], + [-1.0012e-08, 1.2806e-09, -5.1223e-09, ..., -2.2119e-08, + 3.9581e-09, 6.1700e-09]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0201, -0.0288, -0.0077, 0.0141, -0.0013, -0.0153, 0.0036, 0.0132, + -0.0363, 0.0139], device='cuda:0'), grad: tensor([ 4.7497e-08, -1.0319e-06, 1.0047e-07, 1.8743e-08, 1.1933e-07, + 1.6263e-07, 8.8708e-08, 8.3540e-07, -3.2340e-07, -4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 217.55, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4556 re_mapping 0.0022 re_causal 0.0077 /// teacc 99.10 lr 0.00010000 +Epoch 400, weight, value: tensor([[ 0.0324, -0.1039, -0.0303, ..., -0.1556, -0.1840, -0.1768], + [-0.1600, 0.0879, -0.1490, ..., -0.2462, -0.2189, -0.2222], + [-0.1546, -0.2078, -0.2465, ..., -0.2292, -0.1851, 0.1797], + ..., + [ 0.0915, -0.0817, -0.2097, ..., 0.1135, -0.1902, -0.1628], + [ 0.1408, 0.1044, 0.1105, ..., -0.2344, -0.1395, 0.0936], + [ 0.0853, 0.0643, 0.1335, ..., 0.0577, -0.1039, -0.0193]], + device='cuda:0'), grad: tensor([[ 2.5029e-08, 2.6776e-09, 6.9849e-10, ..., 3.3760e-09, + 2.5611e-09, 7.9279e-08], + [ 2.0722e-08, -8.7661e-08, 2.0955e-09, ..., 1.2573e-08, + 1.2806e-09, 1.2852e-07], + [ 9.5577e-08, 2.8173e-08, 8.1491e-10, ..., 7.9721e-07, + 9.3132e-10, -6.3144e-07], + ..., + [-1.2701e-07, 7.4739e-08, 3.3178e-08, ..., -6.7474e-07, + 0.0000e+00, -1.0675e-07], + [ 4.8312e-08, -2.0722e-08, -1.1991e-08, ..., 6.4028e-09, + 8.2655e-09, 2.1944e-07], + [-9.7440e-08, -5.6112e-08, -3.5740e-08, ..., -1.6449e-07, + 1.1642e-10, 2.3167e-08]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0200, -0.0290, -0.0078, 0.0141, -0.0015, -0.0155, 0.0043, 0.0157, + -0.0364, 0.0124], device='cuda:0'), grad: tensor([ 2.9034e-07, 7.4692e-07, -3.5297e-06, 2.2110e-06, 1.1374e-07, + 3.4226e-08, 8.6962e-08, -4.2119e-07, 6.8732e-07, -2.1618e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 217.71, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4567 re_mapping 0.0022 re_causal 0.0073 /// teacc 99.17 lr 0.00001000 +Epoch 401, weight, value: tensor([[ 0.0329, -0.1040, -0.0304, ..., -0.1556, -0.1865, -0.1782], + [-0.1608, 0.0880, -0.1501, ..., -0.2481, -0.2190, -0.2223], + [-0.1547, -0.2079, -0.2493, ..., -0.2294, -0.1852, 0.1800], + ..., + [ 0.0916, -0.0816, -0.2105, ..., 0.1137, -0.1906, -0.1634], + [ 0.1412, 0.1047, 0.1113, ..., -0.2347, -0.1395, 0.0939], + [ 0.0852, 0.0642, 0.1339, ..., 0.0576, -0.1039, -0.0193]], + device='cuda:0'), grad: tensor([[-3.8417e-09, 4.5402e-08, 1.5483e-08, ..., 4.3889e-08, + 0.0000e+00, 1.1292e-08], + [ 1.3853e-08, -3.0384e-07, 6.2166e-08, ..., 1.7777e-07, + 1.1642e-10, 5.1106e-08], + [ 7.4506e-09, 3.4575e-08, 2.2119e-09, ..., 9.0804e-09, + 1.1642e-10, -2.1537e-08], + ..., + [ 3.8417e-09, 3.6578e-07, 9.8487e-08, ..., 2.7288e-07, + 0.0000e+00, 7.0897e-08], + [ 1.6764e-08, 1.7253e-07, 6.0070e-08, ..., 1.7311e-07, + 1.1642e-10, 4.2492e-08], + [-1.5122e-07, -2.0582e-06, -8.3726e-07, ..., -2.3711e-06, + 2.3283e-10, -5.6485e-07]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0179, -0.0291, -0.0076, 0.0140, -0.0016, -0.0155, 0.0053, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 1.6182e-07, -3.5707e-06, 2.3143e-07, 6.9849e-08, 7.2345e-06, + 1.1583e-07, 4.5751e-08, 1.8552e-06, 7.4226e-07, -6.8732e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 217.84, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4231 re_mapping 0.0021 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 402, weight, value: tensor([[ 0.0329, -0.1040, -0.0304, ..., -0.1556, -0.1865, -0.1782], + [-0.1607, 0.0880, -0.1502, ..., -0.2483, -0.2190, -0.2223], + [-0.1546, -0.2079, -0.2494, ..., -0.2294, -0.1852, 0.1801], + ..., + [ 0.0916, -0.0816, -0.2106, ..., 0.1137, -0.1906, -0.1634], + [ 0.1412, 0.1048, 0.1114, ..., -0.2348, -0.1395, 0.0942], + [ 0.0852, 0.0642, 0.1341, ..., 0.0577, -0.1039, -0.0193]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 5.0059e-09, 1.9791e-09, ..., 3.4925e-09, + 2.2119e-09, 4.5402e-09], + [ 7.3342e-09, 3.8417e-09, 3.3760e-09, ..., 8.6147e-09, + 1.1642e-09, 8.3819e-09], + [ 6.4028e-09, 5.1223e-09, 1.8626e-09, ..., 6.7521e-09, + 9.3132e-10, -2.0722e-08], + ..., + [ 4.6566e-09, 3.7369e-08, 1.3039e-08, ..., 5.8208e-09, + 4.0745e-09, 1.1525e-08], + [ 2.0140e-08, 4.3772e-08, 1.6415e-08, ..., 3.1898e-08, + 5.5879e-09, 1.0361e-08], + [-7.1712e-08, -1.3830e-07, -5.7975e-08, ..., -9.1619e-08, + -9.8953e-09, -2.4680e-08]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0179, -0.0290, -0.0076, 0.0140, -0.0016, -0.0154, 0.0051, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 1.8859e-08, 3.8766e-08, -5.2154e-08, -5.2853e-07, 1.3958e-07, + 6.0257e-07, -4.8894e-09, 1.2841e-07, 1.8999e-07, -5.2759e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 217.68, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4119 re_mapping 0.0021 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 403, weight, value: tensor([[ 0.0329, -0.1040, -0.0304, ..., -0.1556, -0.1865, -0.1782], + [-0.1607, 0.0880, -0.1503, ..., -0.2484, -0.2190, -0.2223], + [-0.1545, -0.2079, -0.2494, ..., -0.2294, -0.1852, 0.1802], + ..., + [ 0.0916, -0.0816, -0.2107, ..., 0.1137, -0.1907, -0.1635], + [ 0.1412, 0.1048, 0.1115, ..., -0.2348, -0.1395, 0.0942], + [ 0.0852, 0.0642, 0.1341, ..., 0.0577, -0.1039, -0.0192]], + device='cuda:0'), grad: tensor([[-4.0745e-09, 1.1525e-08, 4.8894e-09, ..., 5.1223e-09, + -5.8208e-10, 1.7695e-08], + [ 7.2294e-08, 4.1560e-08, 2.1537e-08, ..., 3.6089e-08, + 1.1642e-10, 6.9034e-08], + [-5.6066e-07, 3.4925e-10, 8.1491e-09, ..., -2.3132e-07, + 0.0000e+00, -4.2608e-07], + ..., + [ 1.2340e-08, 3.6205e-08, 8.0327e-09, ..., -1.4086e-07, + 0.0000e+00, 2.2317e-07], + [ 1.2876e-07, -2.8964e-07, -1.3190e-07, ..., 1.8347e-07, + -1.1642e-10, -2.1618e-07], + [ 2.5611e-07, 1.0827e-07, 4.5635e-08, ..., 1.0384e-07, + 1.1642e-10, 2.3656e-07]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0179, -0.0290, -0.0076, 0.0140, -0.0016, -0.0154, 0.0050, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 2.4564e-08, 3.6787e-07, -3.1292e-06, 1.5390e-07, 7.1246e-08, + 2.1770e-07, 1.0000e-07, 9.7509e-07, -2.6729e-07, 1.4985e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 217.77, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4537 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 404, weight, value: tensor([[ 0.0329, -0.1041, -0.0304, ..., -0.1557, -0.1865, -0.1782], + [-0.1606, 0.0881, -0.1504, ..., -0.2484, -0.2190, -0.2223], + [-0.1545, -0.2079, -0.2495, ..., -0.2295, -0.1852, 0.1802], + ..., + [ 0.0916, -0.0816, -0.2107, ..., 0.1137, -0.1907, -0.1636], + [ 0.1413, 0.1048, 0.1116, ..., -0.2349, -0.1395, 0.0942], + [ 0.0852, 0.0642, 0.1341, ..., 0.0577, -0.1039, -0.0193]], + device='cuda:0'), grad: tensor([[ 8.1491e-10, 8.1491e-10, 8.7311e-10, ..., 1.6298e-09, + 4.6566e-10, 6.9849e-10], + [ 9.8953e-09, 4.5984e-09, 1.5134e-09, ..., 2.5029e-09, + 5.8208e-10, 1.1991e-08], + [ 5.4715e-09, 3.7253e-09, 4.6566e-10, ..., 1.7462e-09, + 3.4925e-10, -1.0419e-08], + ..., + [ 8.7311e-10, 7.6252e-09, 4.4238e-09, ..., 1.5716e-09, + 1.1642e-09, 5.5879e-09], + [-2.4447e-08, -1.9907e-08, -1.3388e-09, ..., 3.4343e-09, + 6.4028e-10, -2.1537e-08], + [-1.6124e-08, -1.6997e-08, -2.5902e-08, ..., -3.9698e-08, + -4.0745e-10, 4.7730e-09]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0179, -0.0290, -0.0076, 0.0140, -0.0016, -0.0153, 0.0049, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 7.9744e-09, 5.0990e-08, -4.2142e-08, 2.5320e-08, 8.1083e-08, + -1.2200e-07, 6.0885e-08, 3.6205e-08, 2.4913e-08, -1.0815e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 217.59, cls_loss 0.0004 cls_loss_mapping 0.0006 cls_loss_causal 0.4064 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 405, weight, value: tensor([[ 0.0329, -0.1041, -0.0305, ..., -0.1557, -0.1865, -0.1782], + [-0.1607, 0.0881, -0.1504, ..., -0.2485, -0.2190, -0.2223], + [-0.1545, -0.2079, -0.2495, ..., -0.2294, -0.1852, 0.1803], + ..., + [ 0.0916, -0.0816, -0.2108, ..., 0.1137, -0.1908, -0.1637], + [ 0.1413, 0.1048, 0.1116, ..., -0.2349, -0.1395, 0.0943], + [ 0.0852, 0.0642, 0.1341, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-1.7637e-08, 8.3237e-09, 3.5507e-09, ..., 1.1642e-09, + 4.8894e-09, 6.7521e-09], + [ 1.8917e-08, 9.2143e-08, 9.4878e-09, ..., 9.4355e-08, + 1.2165e-08, 1.9616e-08], + [ 1.7521e-08, 1.6473e-08, 7.8580e-09, ..., 3.3760e-09, + 9.4878e-09, 1.5309e-08], + ..., + [ 5.1805e-09, 2.0314e-08, 9.1968e-09, ..., -8.3819e-09, + 1.1933e-08, 1.7870e-08], + [-2.0862e-07, -2.5448e-07, -1.4110e-07, ..., 3.7835e-09, + -1.8324e-07, -2.6799e-07], + [ 2.4971e-08, -2.4028e-07, 6.1118e-09, ..., -3.0594e-07, + 1.2398e-08, 1.5600e-08]], device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0179, -0.0290, -0.0075, 0.0140, -0.0016, -0.0153, 0.0049, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([-3.7835e-08, 4.6170e-07, 7.7300e-08, 2.2771e-07, 8.4890e-07, + 4.7288e-07, 1.8999e-07, 6.3970e-08, -1.1604e-06, -1.1167e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 217.81, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4210 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 +Epoch 406, weight, value: tensor([[ 0.0329, -0.1041, -0.0305, ..., -0.1557, -0.1865, -0.1782], + [-0.1607, 0.0881, -0.1504, ..., -0.2486, -0.2190, -0.2224], + [-0.1545, -0.2079, -0.2496, ..., -0.2294, -0.1852, 0.1803], + ..., + [ 0.0916, -0.0816, -0.2109, ..., 0.1137, -0.1908, -0.1637], + [ 0.1414, 0.1048, 0.1116, ..., -0.2350, -0.1395, 0.0943], + [ 0.0852, 0.0642, 0.1342, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-3.1432e-09, 1.1642e-10, 5.8208e-11, ..., 5.8208e-11, + 2.5437e-08, 1.3504e-08], + [ 1.7462e-09, -2.2701e-09, 2.9104e-10, ..., 1.4552e-09, + 7.5670e-10, 1.7462e-09], + [ 1.5134e-09, 3.4925e-10, 0.0000e+00, ..., 1.9209e-09, + 8.1491e-10, -4.8312e-09], + ..., + [-5.4715e-09, 3.8417e-09, 6.9849e-10, ..., -6.2282e-09, + 1.1642e-10, 6.9849e-10], + [ 1.1059e-09, -7.5670e-10, -5.8208e-10, ..., 3.0850e-09, + 1.2806e-09, 2.2701e-09], + [ 1.3388e-09, -1.9791e-09, -1.3970e-09, ..., 9.8953e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0180, -0.0290, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 4.8894e-08, -3.8999e-09, -2.5029e-09, 1.1467e-08, 1.9791e-09, + 2.0955e-09, -6.2108e-08, 3.9581e-09, 1.7346e-08, 5.7044e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 217.65, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4353 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 +Epoch 407, weight, value: tensor([[ 0.0329, -0.1041, -0.0305, ..., -0.1557, -0.1865, -0.1782], + [-0.1607, 0.0881, -0.1505, ..., -0.2486, -0.2190, -0.2224], + [-0.1545, -0.2079, -0.2496, ..., -0.2295, -0.1852, 0.1803], + ..., + [ 0.0916, -0.0816, -0.2110, ..., 0.1137, -0.1909, -0.1638], + [ 0.1414, 0.1049, 0.1117, ..., -0.2351, -0.1395, 0.0944], + [ 0.0852, 0.0642, 0.1343, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-2.2119e-09, 1.1642e-10, 0.0000e+00, ..., 5.8208e-11, + 6.9849e-10, 5.8208e-10], + [ 2.0373e-09, 9.3132e-10, 2.3283e-10, ..., 2.0955e-09, + 9.8953e-10, 1.3970e-09], + [ 2.2119e-09, 5.2387e-10, 1.1642e-10, ..., 3.0850e-09, + 7.5670e-10, 8.7311e-10], + ..., + [-7.5088e-09, 1.1642e-09, 4.0745e-10, ..., -1.3621e-08, + 5.8208e-11, 5.8208e-10], + [-7.5670e-10, 7.2177e-09, -7.5670e-10, ..., 1.9791e-09, + 1.8044e-09, -6.9849e-10], + [ 2.3865e-09, 5.2387e-10, 5.8208e-11, ..., 1.5076e-08, + 1.0943e-08, 7.9744e-09]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0181, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([-8.6147e-09, 1.2398e-08, 1.1874e-08, 3.4401e-08, -1.5134e-08, + -2.1711e-07, 4.7963e-08, -2.9569e-08, 1.3830e-07, 3.7020e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 217.67, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4091 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.22 lr 0.00001000 +Epoch 408, weight, value: tensor([[ 0.0329, -0.1041, -0.0305, ..., -0.1557, -0.1865, -0.1782], + [-0.1607, 0.0881, -0.1505, ..., -0.2487, -0.2190, -0.2224], + [-0.1544, -0.2080, -0.2496, ..., -0.2294, -0.1852, 0.1803], + ..., + [ 0.0916, -0.0816, -0.2110, ..., 0.1137, -0.1910, -0.1638], + [ 0.1415, 0.1049, 0.1117, ..., -0.2351, -0.1395, 0.0944], + [ 0.0852, 0.0642, 0.1343, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-2.9104e-10, 1.3388e-09, 5.8208e-10, ..., 2.0373e-09, + 5.8208e-11, 2.3283e-10], + [ 2.7416e-08, 4.7148e-09, 2.5611e-09, ..., 3.2596e-08, + 2.9104e-10, 2.6776e-09], + [ 6.1118e-09, 5.8208e-10, 3.4925e-10, ..., 9.1968e-09, + 1.1642e-10, 4.6566e-10], + ..., + [-1.0710e-07, 1.7986e-08, 7.6252e-09, ..., -1.3283e-07, + 6.4028e-10, 1.5134e-09], + [-7.6252e-09, -8.2073e-09, -7.2760e-09, ..., 7.0431e-09, + -8.7311e-10, -1.2631e-08], + [ 5.9430e-08, -4.5518e-08, -1.8568e-08, ..., 5.2503e-08, + 9.3132e-10, 1.5716e-09]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0181, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([-5.8208e-11, 1.1059e-07, 2.7649e-08, 2.1479e-08, 6.4785e-08, + 3.2014e-08, 1.6356e-08, -3.8231e-07, -2.3283e-08, 1.4249e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 217.53, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4268 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.21 lr 0.00001000 +Epoch 409, weight, value: tensor([[ 0.0330, -0.1041, -0.0305, ..., -0.1557, -0.1865, -0.1782], + [-0.1608, 0.0881, -0.1505, ..., -0.2488, -0.2190, -0.2224], + [-0.1544, -0.2080, -0.2496, ..., -0.2295, -0.1852, 0.1804], + ..., + [ 0.0916, -0.0816, -0.2111, ..., 0.1137, -0.1911, -0.1639], + [ 0.1415, 0.1049, 0.1117, ..., -0.2352, -0.1395, 0.0944], + [ 0.0852, 0.0642, 0.1343, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-1.0547e-07, -1.1642e-09, 0.0000e+00, ..., -3.4925e-09, + 4.0745e-10, -3.4925e-09], + [ 1.8335e-08, 1.7462e-10, 5.8208e-11, ..., 7.7416e-09, + 1.0477e-09, 7.5670e-10], + [ 9.1968e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 4.0745e-10, 3.4925e-10], + ..., + [-4.8894e-09, 5.2387e-10, 2.9104e-10, ..., -1.4785e-08, + 8.7311e-10, 5.8208e-10], + [ 2.4505e-08, 5.8208e-10, 5.8208e-11, ..., 2.3865e-09, + 2.8522e-09, 2.3865e-09], + [ 2.7008e-08, 5.2387e-10, 0.0000e+00, ..., 2.2002e-08, + 1.7171e-08, 8.9640e-09]], device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0181, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([-4.5006e-07, 9.7381e-08, 5.0408e-08, 2.1327e-07, 1.1700e-08, + -2.4550e-06, 8.1584e-07, 5.3551e-09, 1.8021e-07, 1.5432e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 217.33, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4458 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 +Epoch 410, weight, value: tensor([[ 0.0330, -0.1041, -0.0305, ..., -0.1557, -0.1865, -0.1782], + [-0.1608, 0.0881, -0.1506, ..., -0.2488, -0.2190, -0.2224], + [-0.1544, -0.2080, -0.2497, ..., -0.2295, -0.1852, 0.1804], + ..., + [ 0.0916, -0.0816, -0.2111, ..., 0.1137, -0.1911, -0.1639], + [ 0.1415, 0.1049, 0.1118, ..., -0.2352, -0.1395, 0.0944], + [ 0.0852, 0.0642, 0.1344, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-7.5670e-10, 6.9849e-10, 4.0745e-10, ..., 2.9104e-10, + 5.8208e-10, 2.9686e-08], + [ 2.6193e-09, -8.0327e-09, 4.0745e-10, ..., 2.6193e-09, + 2.3283e-10, 1.4971e-07], + [ 1.4552e-09, 4.0745e-10, 5.8208e-11, ..., 2.9686e-09, + 1.1642e-10, -9.6951e-07], + ..., + [-4.1327e-09, 9.5461e-09, 2.2701e-09, ..., -6.3446e-09, + 0.0000e+00, 1.3597e-07], + [ 1.1059e-09, 4.0745e-10, 1.7462e-10, ..., 3.0268e-09, + 2.3283e-09, 2.2410e-08], + [-5.1223e-09, -6.4611e-09, -6.2864e-09, ..., -7.3342e-09, + 5.8208e-11, 2.0606e-08]], device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0181, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 2.4447e-07, 1.1995e-06, -8.1360e-06, 4.9509e-06, 9.2434e-08, + 1.6042e-07, 9.0804e-09, 1.1595e-06, 1.7462e-07, 1.5297e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 217.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4281 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.14 lr 0.00001000 +Epoch 411, weight, value: tensor([[ 0.0330, -0.1041, -0.0306, ..., -0.1557, -0.1865, -0.1782], + [-0.1608, 0.0881, -0.1506, ..., -0.2489, -0.2190, -0.2224], + [-0.1543, -0.2080, -0.2497, ..., -0.2295, -0.1852, 0.1804], + ..., + [ 0.0916, -0.0816, -0.2112, ..., 0.1137, -0.1911, -0.1639], + [ 0.1415, 0.1049, 0.1118, ..., -0.2353, -0.1395, 0.0944], + [ 0.0852, 0.0642, 0.1344, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-5.4133e-09, 1.1642e-10, 0.0000e+00, ..., 1.8044e-09, + 2.1537e-09, 2.6776e-09], + [ 3.0850e-09, -1.2689e-08, 0.0000e+00, ..., 4.2492e-09, + 1.4552e-09, 7.1013e-09], + [-1.1059e-09, 7.5670e-09, 0.0000e+00, ..., 1.7462e-10, + 3.2596e-09, -1.6938e-08], + ..., + [-1.9791e-09, 5.0059e-09, 5.8208e-11, ..., 1.2224e-09, + 2.9686e-09, 9.3714e-09], + [ 1.2806e-09, 5.8208e-11, 0.0000e+00, ..., 1.0768e-08, + 5.9954e-09, 1.0070e-08], + [ 3.2014e-09, -9.3132e-10, -7.5670e-10, ..., 1.9395e-07, + 1.6717e-07, 1.5495e-07]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0182, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([-1.5774e-08, -9.0920e-08, 5.9954e-09, 5.8440e-08, -5.6392e-07, + -2.6869e-07, -8.2655e-09, 1.0175e-07, 1.7369e-07, 6.2911e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 217.89, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4324 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.16 lr 0.00001000 +Epoch 412, weight, value: tensor([[ 0.0330, -0.1041, -0.0306, ..., -0.1557, -0.1865, -0.1782], + [-0.1609, 0.0881, -0.1507, ..., -0.2490, -0.2190, -0.2224], + [-0.1543, -0.2080, -0.2497, ..., -0.2295, -0.1852, 0.1805], + ..., + [ 0.0916, -0.0816, -0.2112, ..., 0.1137, -0.1912, -0.1640], + [ 0.1416, 0.1050, 0.1118, ..., -0.2354, -0.1395, 0.0945], + [ 0.0852, 0.0642, 0.1345, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 3.4925e-10, 1.1642e-10, ..., 6.9849e-10, + 3.4925e-10, 3.4925e-10], + [ 3.8417e-09, -5.4133e-09, 2.3283e-10, ..., 7.4506e-09, + 1.0477e-09, 4.7148e-09], + [ 1.5937e-07, 7.5670e-10, 5.8208e-11, ..., 3.3015e-07, + 4.6566e-10, 1.6857e-07], + ..., + [-1.7323e-07, 5.0059e-09, 7.5670e-10, ..., -3.5227e-07, + 1.2224e-09, -1.8417e-07], + [ 3.4925e-09, 4.6566e-10, 4.6566e-10, ..., 6.0536e-09, + -2.3283e-10, 3.2596e-09], + [ 4.6566e-09, -1.5716e-09, -1.6298e-09, ..., 4.9942e-08, + 3.0617e-08, 1.8277e-08]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0182, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 2.9104e-09, -6.8103e-09, 8.1584e-07, 6.8685e-09, -7.9337e-08, + 7.3342e-09, 5.9954e-09, -8.6194e-07, 2.1188e-08, 1.0605e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 217.32, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4255 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.15 lr 0.00001000 +Epoch 413, weight, value: tensor([[ 0.0330, -0.1041, -0.0306, ..., -0.1557, -0.1865, -0.1782], + [-0.1609, 0.0881, -0.1507, ..., -0.2491, -0.2190, -0.2224], + [-0.1543, -0.2080, -0.2497, ..., -0.2295, -0.1852, 0.1805], + ..., + [ 0.0916, -0.0816, -0.2113, ..., 0.1137, -0.1912, -0.1640], + [ 0.1416, 0.1050, 0.1119, ..., -0.2354, -0.1395, 0.0945], + [ 0.0852, 0.0642, 0.1345, ..., 0.0577, -0.1040, -0.0193]], + device='cuda:0'), grad: tensor([[-3.0210e-08, 6.9849e-09, 4.6566e-09, ..., 1.7462e-10, + 5.8208e-10, 6.1118e-09], + [ 5.5879e-08, 1.1234e-07, 7.6892e-08, ..., 2.0373e-09, + 9.3132e-10, 9.1153e-08], + [ 2.2934e-08, 4.6275e-08, 3.0850e-08, ..., -4.6566e-10, + 1.7462e-10, 1.5018e-08], + ..., + [ 2.0897e-08, 4.8312e-08, 3.1025e-08, ..., -1.6298e-09, + 3.4925e-10, 3.7893e-08], + [-2.6426e-07, -5.5507e-07, -3.7486e-07, ..., 1.2806e-09, + 1.1059e-09, -4.1048e-07], + [ 6.8976e-08, 9.8196e-08, 6.5076e-08, ..., 2.9686e-09, + 3.4925e-09, 7.5321e-08]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0182, -0.0291, -0.0075, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([-1.6415e-07, 4.1816e-07, 1.0425e-07, 2.2445e-07, 1.8859e-08, + 5.1269e-07, 1.6578e-07, 1.8650e-07, -1.9353e-06, 4.9034e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 217.61, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4372 re_mapping 0.0018 re_causal 0.0073 /// teacc 99.16 lr 0.00001000 +Epoch 414, weight, value: tensor([[ 0.0331, -0.1041, -0.0306, ..., -0.1557, -0.1865, -0.1782], + [-0.1609, 0.0881, -0.1508, ..., -0.2492, -0.2190, -0.2225], + [-0.1543, -0.2080, -0.2498, ..., -0.2295, -0.1852, 0.1805], + ..., + [ 0.0916, -0.0816, -0.2113, ..., 0.1137, -0.1912, -0.1640], + [ 0.1416, 0.1050, 0.1119, ..., -0.2356, -0.1395, 0.0945], + [ 0.0852, 0.0642, 0.1346, ..., 0.0577, -0.1040, -0.0194]], + device='cuda:0'), grad: tensor([[-2.7940e-09, 1.5716e-09, 9.8953e-10, ..., 3.1432e-09, + 6.9849e-10, 1.5134e-09], + [ 1.2980e-08, -2.4447e-09, 2.5029e-09, ..., 1.5716e-08, + 2.7358e-09, 6.8685e-09], + [ 1.3853e-08, 3.8999e-09, 8.7311e-10, ..., 2.2177e-08, + 2.6193e-09, -2.3283e-09], + ..., + [-2.9337e-08, 3.4051e-08, 2.2061e-08, ..., -2.9395e-08, + 1.6356e-08, 3.3178e-08], + [-4.1095e-08, -2.9395e-08, -2.8929e-08, ..., 1.1642e-08, + 1.2806e-09, -5.5006e-08], + [ 1.0943e-08, -3.1141e-08, -1.5076e-08, ..., 2.7660e-07, + 1.7998e-07, 5.4250e-08]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0182, -0.0291, -0.0074, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([-1.1933e-08, 3.0617e-08, 6.8860e-08, -1.5483e-08, -7.6042e-07, + 6.4203e-08, 4.2899e-08, -2.2352e-08, -1.0082e-07, 7.2084e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 217.27, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4214 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.13 lr 0.00001000 +Epoch 415, weight, value: tensor([[ 0.0330, -0.1041, -0.0307, ..., -0.1559, -0.1865, -0.1782], + [-0.1610, 0.0881, -0.1509, ..., -0.2493, -0.2190, -0.2225], + [-0.1543, -0.2080, -0.2498, ..., -0.2295, -0.1852, 0.1805], + ..., + [ 0.0916, -0.0816, -0.2115, ..., 0.1137, -0.1913, -0.1641], + [ 0.1416, 0.1050, 0.1120, ..., -0.2356, -0.1395, 0.0945], + [ 0.0852, 0.0642, 0.1346, ..., 0.0577, -0.1040, -0.0194]], + device='cuda:0'), grad: tensor([[-2.5611e-09, 5.8208e-11, 0.0000e+00, ..., 5.8208e-11, + 5.8208e-11, 1.7462e-10], + [ 1.6298e-09, -1.1642e-10, 2.3283e-10, ..., 2.0955e-09, + 1.7462e-10, 4.8312e-09], + [ 2.0373e-09, 2.9104e-10, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, -1.1700e-08], + ..., + [-8.1491e-10, 5.9372e-09, 4.0163e-09, ..., 5.8208e-09, + 5.8208e-11, 4.3074e-09], + [-5.8208e-11, -6.4028e-10, -7.5670e-10, ..., 1.0477e-09, + 1.1642e-10, 4.6566e-10], + [-2.3865e-09, -5.8790e-09, -5.2387e-09, ..., -1.3795e-08, + 4.0745e-10, 5.2387e-10]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0183, -0.0292, -0.0074, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([-1.0128e-08, 1.7171e-08, -2.6543e-08, 9.4878e-09, 3.5507e-09, + -1.3271e-08, 1.1409e-08, 2.4738e-08, 1.3271e-08, -1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 217.54, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4527 re_mapping 0.0017 re_causal 0.0073 /// teacc 99.10 lr 0.00001000 +Epoch 416, weight, value: tensor([[ 0.0330, -0.1041, -0.0307, ..., -0.1559, -0.1865, -0.1783], + [-0.1608, 0.0882, -0.1509, ..., -0.2493, -0.2191, -0.2225], + [-0.1544, -0.2080, -0.2499, ..., -0.2295, -0.1852, 0.1805], + ..., + [ 0.0916, -0.0816, -0.2116, ..., 0.1137, -0.1913, -0.1641], + [ 0.1417, 0.1050, 0.1120, ..., -0.2357, -0.1395, 0.0945], + [ 0.0852, 0.0642, 0.1347, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[ 2.9104e-10, 5.8208e-10, 2.9104e-10, ..., 6.4028e-10, + 9.8953e-10, 1.2806e-09], + [ 4.4238e-09, -2.5029e-09, 1.2224e-09, ..., 4.1327e-09, + 4.6566e-10, 3.3760e-09], + [ 2.5611e-09, 1.2224e-09, 1.7462e-10, ..., 3.1432e-09, + 1.3970e-09, -3.2596e-09], + ..., + [-8.2073e-09, 5.5879e-09, 2.0955e-09, ..., -7.1595e-09, + 1.7462e-10, 1.1642e-09], + [-3.6089e-09, -5.5297e-09, -4.6566e-09, ..., 4.3656e-09, + 1.7462e-09, -4.4820e-09], + [-1.0652e-08, -2.2934e-08, -1.2922e-08, ..., -3.0035e-08, + -6.9849e-10, 1.7462e-09]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 5.8790e-09, -4.6566e-09, 1.9209e-09, 1.3271e-08, 7.9861e-08, + 2.1479e-08, -2.5146e-08, -6.4028e-10, 6.4028e-10, -7.9221e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 217.69, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4335 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.17 lr 0.00001000 +Epoch 417, weight, value: tensor([[ 0.0329, -0.1041, -0.0307, ..., -0.1561, -0.1866, -0.1783], + [-0.1608, 0.0882, -0.1510, ..., -0.2494, -0.2191, -0.2225], + [-0.1544, -0.2080, -0.2500, ..., -0.2295, -0.1852, 0.1806], + ..., + [ 0.0916, -0.0816, -0.2117, ..., 0.1137, -0.1914, -0.1642], + [ 0.1417, 0.1051, 0.1120, ..., -0.2358, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1348, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[-2.9104e-10, 1.9791e-09, 1.1642e-09, ..., 2.1537e-09, + 1.5192e-08, 1.6880e-08], + [ 3.0850e-09, -3.7486e-07, 8.1491e-10, ..., 2.1537e-09, + 1.3388e-08, 1.6997e-08], + [ 1.6298e-09, 1.4203e-08, 1.1642e-10, ..., 1.7462e-10, + 1.4785e-08, 1.7288e-08], + ..., + [ 1.2631e-08, 3.8557e-07, 1.2165e-08, ..., 2.6426e-08, + 1.3970e-09, 1.8626e-09], + [-4.6566e-10, 1.3446e-08, 5.0641e-09, ..., 1.0419e-08, + 2.1502e-07, 2.4168e-07], + [-4.1735e-08, -5.1863e-08, -3.9581e-08, ..., -1.6356e-08, + 5.5355e-08, 8.1491e-09]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 4.7497e-08, -1.7090e-06, 9.7032e-08, -1.9325e-08, 1.3481e-07, + 6.3935e-07, -1.5637e-06, 1.7909e-06, 6.7567e-07, -9.4122e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 217.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4638 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 +Epoch 418, weight, value: tensor([[ 0.0329, -0.1041, -0.0307, ..., -0.1561, -0.1866, -0.1783], + [-0.1609, 0.0882, -0.1510, ..., -0.2494, -0.2191, -0.2225], + [-0.1544, -0.2080, -0.2500, ..., -0.2295, -0.1852, 0.1806], + ..., + [ 0.0916, -0.0816, -0.2118, ..., 0.1137, -0.1914, -0.1642], + [ 0.1417, 0.1051, 0.1121, ..., -0.2358, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1349, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 5.8208e-11, 5.8208e-11, ..., 1.7462e-10, + 4.6566e-10, 3.6089e-09], + [ 1.3388e-09, -3.4925e-10, 5.8208e-11, ..., 1.1642e-09, + 1.2224e-09, 2.3108e-08], + [-2.8522e-09, 5.8208e-11, 0.0000e+00, ..., 3.4925e-10, + 8.1491e-10, -1.6298e-07], + ..., + [ 1.5716e-09, 1.6298e-09, 4.0745e-10, ..., 2.0373e-09, + 9.8953e-10, 2.1653e-08], + [ 5.2387e-10, 3.4925e-10, 1.1642e-10, ..., 3.4925e-10, + 9.3132e-10, 9.6974e-08], + [-1.0477e-09, -1.8044e-09, -1.2224e-09, ..., 1.0012e-08, + 1.3213e-08, 6.8685e-09]], device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 1.6880e-08, 9.2841e-08, -7.1852e-07, 6.9733e-08, -2.2759e-08, + 8.9640e-09, 1.3970e-09, 1.0210e-07, 4.3423e-07, 3.0326e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 217.91, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4302 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.16 lr 0.00001000 +Epoch 419, weight, value: tensor([[ 0.0329, -0.1041, -0.0307, ..., -0.1561, -0.1866, -0.1783], + [-0.1608, 0.0882, -0.1510, ..., -0.2495, -0.2191, -0.2225], + [-0.1544, -0.2080, -0.2500, ..., -0.2295, -0.1852, 0.1806], + ..., + [ 0.0916, -0.0816, -0.2118, ..., 0.1137, -0.1914, -0.1642], + [ 0.1417, 0.1051, 0.1121, ..., -0.2359, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1349, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[-5.9954e-09, 4.6566e-10, 3.4925e-10, ..., 6.9849e-10, + 1.5134e-09, 4.3074e-09], + [ 4.9477e-09, 1.9791e-09, 1.4552e-09, ..., 8.5565e-09, + 1.7462e-09, 7.9162e-09], + [ 5.7626e-09, 1.7462e-10, 5.8208e-11, ..., 8.4983e-09, + 9.8953e-10, -3.2480e-08], + ..., + [-1.7695e-08, 2.9686e-09, 1.9791e-09, ..., -2.6252e-08, + 5.2387e-10, 5.5879e-09], + [ 5.8208e-09, 2.4447e-08, 1.7404e-08, ..., 1.3853e-08, + 1.0128e-08, 1.6240e-08], + [-4.2492e-09, -7.5437e-08, -5.4075e-08, ..., -2.8929e-08, + 5.8208e-10, 1.9209e-09]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0153, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([-1.3039e-08, 5.2503e-08, -8.6846e-08, 9.3540e-08, 2.8813e-08, + 7.3109e-08, -7.7474e-08, -4.7265e-08, 9.9477e-08, -1.0699e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 217.83, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4217 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 +Epoch 420, weight, value: tensor([[ 0.0328, -0.1041, -0.0307, ..., -0.1562, -0.1866, -0.1783], + [-0.1608, 0.0882, -0.1511, ..., -0.2496, -0.2191, -0.2226], + [-0.1544, -0.2080, -0.2500, ..., -0.2296, -0.1852, 0.1807], + ..., + [ 0.0916, -0.0816, -0.2119, ..., 0.1137, -0.1914, -0.1642], + [ 0.1417, 0.1051, 0.1121, ..., -0.2359, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1350, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[-2.0373e-09, 1.1642e-10, 0.0000e+00, ..., 6.4028e-10, + 1.6880e-09, 1.3970e-09], + [ 2.6776e-09, -2.0373e-09, 2.3283e-10, ..., 5.0641e-09, + 5.2387e-10, 5.9954e-09], + [ 3.8417e-09, 2.9104e-10, 5.8208e-11, ..., 8.2655e-09, + 6.4028e-10, -7.9744e-09], + ..., + [-1.4203e-08, 1.6880e-09, 1.7462e-10, ..., -2.8289e-08, + 1.7462e-10, -6.9849e-10], + [ 1.7462e-10, -5.8208e-10, -5.2387e-10, ..., 1.6473e-08, + 2.5611e-09, 2.6776e-09], + [ 5.2387e-09, 6.4028e-10, 5.8208e-11, ..., 2.7590e-08, + 1.4086e-08, 1.0943e-08]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0154, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 4.0163e-09, 3.3993e-08, 5.9954e-09, 7.8755e-08, -1.6880e-08, + -4.2096e-07, 1.4459e-07, -5.8382e-08, 1.4924e-07, 8.9232e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 217.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3967 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.16 lr 0.00001000 +Epoch 421, weight, value: tensor([[ 0.0328, -0.1041, -0.0308, ..., -0.1562, -0.1866, -0.1783], + [-0.1609, 0.0882, -0.1511, ..., -0.2498, -0.2191, -0.2226], + [-0.1544, -0.2080, -0.2501, ..., -0.2296, -0.1852, 0.1807], + ..., + [ 0.0916, -0.0816, -0.2119, ..., 0.1137, -0.1915, -0.1643], + [ 0.1417, 0.1051, 0.1122, ..., -0.2360, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1350, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[ 1.7462e-10, 4.0745e-10, 1.1642e-10, ..., 1.7462e-10, + 5.8208e-11, 2.3283e-10], + [ 3.0268e-09, 3.1432e-09, 2.2701e-09, ..., 1.6298e-09, + 1.1642e-10, 4.1910e-09], + [ 3.4925e-10, 2.7940e-09, 8.1491e-10, ..., 2.0373e-09, + 1.1642e-10, -3.2014e-09], + ..., + [-2.3283e-09, 3.7253e-09, 1.5134e-09, ..., -5.2387e-09, + 5.8208e-11, 3.3178e-09], + [-1.3039e-08, -1.4959e-08, -1.6182e-08, ..., 9.2550e-09, + 5.8208e-11, -2.1304e-08], + [ 3.8417e-09, 2.2701e-09, 1.8044e-09, ..., 8.6147e-09, + 1.3970e-09, 3.0850e-09]], device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0154, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 2.9686e-09, 2.2934e-08, -1.1642e-09, -7.1595e-09, 9.8953e-10, + -1.8557e-07, 3.5274e-08, 2.0780e-08, 6.1933e-08, 5.9255e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 217.55, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4386 re_mapping 0.0017 re_causal 0.0073 /// teacc 99.19 lr 0.00001000 +Epoch 422, weight, value: tensor([[ 0.0329, -0.1042, -0.0308, ..., -0.1562, -0.1866, -0.1783], + [-0.1609, 0.0882, -0.1511, ..., -0.2499, -0.2191, -0.2226], + [-0.1544, -0.2080, -0.2501, ..., -0.2296, -0.1852, 0.1807], + ..., + [ 0.0916, -0.0816, -0.2120, ..., 0.1137, -0.1915, -0.1643], + [ 0.1418, 0.1051, 0.1122, ..., -0.2361, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1350, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[ 7.5670e-10, 9.8953e-10, 5.2387e-10, ..., 1.6298e-09, + 1.3388e-09, 1.2806e-09], + [ 1.5134e-09, -5.8790e-09, 6.9849e-10, ..., 2.9104e-09, + 1.0477e-09, 2.2701e-09], + [ 3.4925e-10, 2.0373e-09, 0.0000e+00, ..., -1.1642e-10, + 4.0745e-10, -3.3760e-09], + ..., + [ 6.4028e-09, 1.3330e-08, 4.6566e-09, ..., 1.0943e-08, + 9.3132e-10, 1.1642e-09], + [ 1.2224e-09, 2.6193e-09, 6.4028e-10, ..., 2.4447e-09, + 5.1805e-09, 5.4133e-09], + [-6.3737e-08, -3.1316e-08, -4.1793e-08, ..., -1.1519e-07, + -9.4878e-09, -6.1700e-09]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0183, -0.0291, -0.0074, 0.0140, -0.0016, -0.0154, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 8.4983e-09, -1.6124e-08, -1.0477e-09, -1.2398e-08, 2.5867e-07, + 9.3190e-08, -8.9058e-08, 5.6229e-08, 2.4098e-08, -2.9500e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 216.94, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4332 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 423, weight, value: tensor([[ 0.0329, -0.1041, -0.0308, ..., -0.1562, -0.1866, -0.1783], + [-0.1610, 0.0882, -0.1512, ..., -0.2500, -0.2191, -0.2226], + [-0.1544, -0.2080, -0.2501, ..., -0.2296, -0.1852, 0.1808], + ..., + [ 0.0916, -0.0816, -0.2120, ..., 0.1137, -0.1915, -0.1644], + [ 0.1418, 0.1051, 0.1122, ..., -0.2362, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1351, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 2.4447e-09, 4.0745e-10, ..., 1.9791e-09, + 1.4831e-07, 1.1007e-07], + [ 3.6671e-09, 4.7730e-09, 1.4552e-09, ..., 5.1223e-09, + 5.7276e-08, 5.2678e-08], + [-4.8312e-09, 1.3388e-09, 1.7462e-10, ..., 8.7311e-10, + 7.5088e-08, 2.6019e-08], + ..., + [ 2.1537e-09, 1.0128e-08, 9.3132e-10, ..., 1.5716e-09, + 3.7835e-09, 1.3388e-08], + [-1.7462e-09, 6.8103e-09, -8.1491e-10, ..., 9.3132e-10, + 8.9058e-08, 6.5658e-08], + [-2.5320e-08, -1.3737e-07, -4.3539e-08, ..., -1.9977e-07, + -5.2387e-09, 7.6834e-09]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0184, -0.0292, -0.0074, 0.0140, -0.0017, -0.0154, 0.0048, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 8.4378e-07, 3.7998e-07, 3.1572e-07, 4.5809e-08, 1.0235e-06, + 7.2978e-06, -9.9912e-06, 8.9523e-08, 4.7218e-07, -4.5216e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 217.58, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4455 re_mapping 0.0017 re_causal 0.0073 /// teacc 99.23 lr 0.00001000 +Epoch 424, weight, value: tensor([[ 0.0329, -0.1041, -0.0308, ..., -0.1562, -0.1866, -0.1783], + [-0.1610, 0.0882, -0.1513, ..., -0.2501, -0.2191, -0.2226], + [-0.1545, -0.2080, -0.2502, ..., -0.2296, -0.1852, 0.1808], + ..., + [ 0.0916, -0.0816, -0.2121, ..., 0.1137, -0.1916, -0.1644], + [ 0.1419, 0.1052, 0.1123, ..., -0.2362, -0.1395, 0.0946], + [ 0.0852, 0.0642, 0.1351, ..., 0.0577, -0.1041, -0.0194]], + device='cuda:0'), grad: tensor([[ 8.7311e-11, 3.4925e-10, 2.9104e-11, ..., 8.7311e-11, + 4.3656e-10, 1.2515e-09], + [ 3.4051e-09, -5.7335e-09, 1.0768e-09, ..., 2.8813e-09, + 6.4028e-10, 5.6782e-08], + [ 4.0454e-09, 3.7835e-09, 1.7462e-09, ..., 3.0559e-09, + 6.4028e-10, -8.2934e-07], + ..., + [-1.1147e-08, 7.4797e-09, 2.0664e-09, ..., -1.9005e-08, + 5.5297e-10, 7.1572e-07], + [-6.9849e-09, -1.1263e-08, -8.0909e-09, ..., 3.5216e-09, + 6.1118e-10, 2.9657e-08], + [ 4.6566e-09, 1.0477e-09, 2.3283e-10, ..., 7.3051e-09, + 1.0768e-09, 2.4738e-09]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0184, -0.0292, -0.0074, 0.0140, -0.0016, -0.0154, 0.0049, 0.0158, + -0.0364, 0.0123], device='cuda:0'), grad: tensor([ 5.6752e-09, 9.9593e-08, -2.2259e-06, 5.7393e-08, 1.2747e-08, + 1.8161e-08, -8.7311e-09, 1.9297e-06, 9.5228e-08, 2.9540e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 217.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4330 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 425, weight, value: tensor([[ 0.0329, -0.1041, -0.0308, ..., -0.1562, -0.1866, -0.1783], + [-0.1611, 0.0882, -0.1513, ..., -0.2502, -0.2191, -0.2226], + [-0.1545, -0.2080, -0.2503, ..., -0.2296, -0.1852, 0.1808], + ..., + [ 0.0916, -0.0816, -0.2121, ..., 0.1137, -0.1917, -0.1645], + [ 0.1420, 0.1052, 0.1124, ..., -0.2363, -0.1395, 0.0947], + [ 0.0852, 0.0642, 0.1351, ..., 0.0577, -0.1042, -0.0195]], + device='cuda:0'), grad: tensor([[-2.4156e-09, 3.1432e-09, 0.0000e+00, ..., 0.0000e+00, + 5.8208e-10, 1.7171e-09], + [ 3.2654e-08, 4.7730e-09, 2.9104e-10, ..., 1.5338e-08, + 3.5216e-09, 3.0064e-08], + [ 1.4232e-08, 4.6857e-09, 8.7311e-11, ..., 2.0867e-08, + 2.7940e-09, -2.2497e-08], + ..., + [-8.5856e-09, 5.8790e-09, 1.4552e-10, ..., -2.0571e-07, + 2.5029e-09, 1.2456e-08], + [-6.9267e-08, -2.8085e-08, -5.8208e-10, ..., 3.6642e-08, + 9.8953e-10, -4.0891e-08], + [ 8.4692e-09, 1.3679e-09, 2.9104e-11, ..., 4.7556e-08, + 3.1810e-08, 2.3254e-08]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0184, -0.0292, -0.0074, 0.0140, -0.0016, -0.0154, 0.0049, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 5.9081e-09, 1.5018e-07, 8.4692e-09, 4.6042e-08, 7.7649e-08, + 1.6636e-07, 3.5507e-08, -5.0850e-07, -1.2759e-07, 1.5553e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 217.89, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3938 re_mapping 0.0017 re_causal 0.0068 /// teacc 99.22 lr 0.00001000 +Epoch 426, weight, value: tensor([[ 0.0330, -0.1040, -0.0308, ..., -0.1562, -0.1866, -0.1784], + [-0.1611, 0.0882, -0.1514, ..., -0.2503, -0.2191, -0.2227], + [-0.1546, -0.2080, -0.2503, ..., -0.2297, -0.1852, 0.1809], + ..., + [ 0.0916, -0.0816, -0.2123, ..., 0.1137, -0.1918, -0.1646], + [ 0.1420, 0.1053, 0.1125, ..., -0.2363, -0.1395, 0.0947], + [ 0.0852, 0.0642, 0.1351, ..., 0.0577, -0.1042, -0.0195]], + device='cuda:0'), grad: tensor([[-6.3446e-09, 1.1059e-09, 6.1118e-10, ..., 8.4401e-10, + 6.8685e-09, 6.1409e-09], + [ 1.1496e-08, -1.6298e-09, 1.7462e-09, ..., 1.2689e-08, + 2.2119e-09, 3.9581e-09], + [ 1.9354e-08, 3.9581e-09, 2.0955e-09, ..., 2.2788e-08, + 1.5134e-09, 3.7835e-09], + ..., + [-5.4308e-08, 5.2387e-09, 1.5134e-09, ..., -7.0198e-08, + 2.7940e-09, 3.3178e-09], + [-3.5798e-08, -5.7335e-08, -4.0367e-08, ..., 7.2760e-09, + 5.4133e-09, -5.0932e-08], + [ 3.7311e-08, 2.1246e-08, 1.4581e-08, ..., 8.5682e-08, + 4.2754e-08, 4.0542e-08]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0185, -0.0292, -0.0073, 0.0140, -0.0016, -0.0154, 0.0049, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 1.0186e-09, 3.2771e-08, 8.3004e-08, 2.0780e-08, -1.1193e-07, + 1.0635e-07, -5.5763e-08, -1.8917e-07, -1.4016e-07, 2.6962e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 217.73, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4116 re_mapping 0.0017 re_causal 0.0069 /// teacc 99.22 lr 0.00001000 +Epoch 427, weight, value: tensor([[ 0.0330, -0.1041, -0.0309, ..., -0.1562, -0.1866, -0.1784], + [-0.1611, 0.0883, -0.1515, ..., -0.2504, -0.2191, -0.2227], + [-0.1545, -0.2080, -0.2504, ..., -0.2297, -0.1852, 0.1809], + ..., + [ 0.0916, -0.0816, -0.2124, ..., 0.1137, -0.1918, -0.1646], + [ 0.1421, 0.1053, 0.1126, ..., -0.2364, -0.1396, 0.0947], + [ 0.0852, 0.0642, 0.1352, ..., 0.0577, -0.1042, -0.0195]], + device='cuda:0'), grad: tensor([[-2.9104e-09, 6.4028e-10, 1.1642e-10, ..., 2.9104e-10, + -6.9849e-10, 3.4925e-10], + [ 8.7311e-10, -1.1642e-10, 5.8208e-10, ..., 3.7835e-09, + 4.1910e-09, 2.7358e-09], + [ 4.0745e-10, 4.0163e-09, 6.9849e-10, ..., 1.2806e-09, + 9.3132e-10, 5.8208e-10], + ..., + [-8.7311e-10, 6.1118e-09, 9.8953e-10, ..., -1.0477e-09, + 1.1642e-09, 1.4552e-09], + [-1.6473e-08, 4.7730e-09, -7.1013e-09, ..., 8.7311e-10, + 6.9849e-10, -1.4261e-08], + [-8.7311e-10, -8.1491e-10, -9.3132e-10, ..., 1.2747e-08, + 1.2806e-08, 8.2655e-09]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0185, -0.0292, -0.0073, 0.0140, -0.0016, -0.0154, 0.0049, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([-1.1700e-08, 4.5984e-09, 1.6007e-08, -6.1525e-08, -5.0466e-08, + 6.1409e-08, 2.2468e-08, 2.4331e-08, -1.5891e-08, 3.1898e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 217.81, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3918 re_mapping 0.0017 re_causal 0.0069 /// teacc 99.22 lr 0.00001000 +Epoch 428, weight, value: tensor([[ 0.0330, -0.1040, -0.0309, ..., -0.1562, -0.1866, -0.1784], + [-0.1611, 0.0883, -0.1516, ..., -0.2505, -0.2191, -0.2227], + [-0.1545, -0.2080, -0.2504, ..., -0.2297, -0.1852, 0.1810], + ..., + [ 0.0916, -0.0816, -0.2124, ..., 0.1137, -0.1919, -0.1648], + [ 0.1422, 0.1053, 0.1127, ..., -0.2364, -0.1396, 0.0948], + [ 0.0852, 0.0642, 0.1352, ..., 0.0577, -0.1042, -0.0196]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.4028e-10, 8.1491e-10], + [ 1.3388e-09, -1.0128e-08, 2.3283e-10, ..., 2.5029e-09, + 2.3283e-10, 3.4925e-09], + [ 9.8953e-09, 1.3970e-09, 5.8208e-11, ..., 2.8929e-08, + 5.8208e-11, 1.9209e-09], + ..., + [-1.1933e-08, 1.0128e-08, 1.1642e-10, ..., -3.4692e-08, + 0.0000e+00, -8.1491e-09], + [-5.8208e-10, -1.7462e-09, -9.3132e-10, ..., 1.9209e-09, + 1.3970e-09, 1.1059e-09], + [ 7.5670e-10, 1.7462e-10, 0.0000e+00, ..., 1.9791e-09, + 6.9849e-10, 6.4028e-10]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0185, -0.0292, -0.0073, 0.0140, -0.0016, -0.0154, 0.0050, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 2.3865e-09, -1.9674e-08, 5.8557e-08, 4.0862e-08, 6.6939e-09, + -4.6042e-08, 1.8044e-09, -4.9011e-08, 2.0023e-08, 7.7998e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 217.53, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4085 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.23 lr 0.00001000 +Epoch 429, weight, value: tensor([[ 0.0330, -0.1041, -0.0309, ..., -0.1562, -0.1866, -0.1785], + [-0.1611, 0.0883, -0.1517, ..., -0.2506, -0.2191, -0.2228], + [-0.1546, -0.2080, -0.2505, ..., -0.2297, -0.1852, 0.1811], + ..., + [ 0.0916, -0.0817, -0.2125, ..., 0.1137, -0.1919, -0.1648], + [ 0.1422, 0.1054, 0.1128, ..., -0.2365, -0.1396, 0.0948], + [ 0.0852, 0.0642, 0.1353, ..., 0.0577, -0.1042, -0.0196]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.3970e-09, 5.8208e-10, ..., 5.8208e-10, + 6.9849e-10, 1.2806e-09], + [ 1.0652e-08, 4.3074e-09, 2.9686e-09, ..., 5.7044e-09, + 6.4028e-10, 6.3446e-09], + [ 7.5088e-09, 4.8894e-09, 1.2224e-09, ..., 6.4611e-09, + 5.2387e-10, 1.2224e-09], + ..., + [-1.5425e-08, 1.2864e-08, 2.9686e-09, ..., -2.3108e-08, + 5.8208e-10, 3.3178e-09], + [-4.7556e-08, -2.7241e-08, -2.3632e-08, ..., 8.0909e-09, + 2.0955e-09, -4.6392e-08], + [ 1.0594e-08, 2.7940e-09, 1.6298e-09, ..., -2.4447e-09, + 1.2806e-09, 1.0361e-08]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0185, -0.0292, -0.0073, 0.0140, -0.0016, -0.0155, 0.0050, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.2957e-08, 3.9581e-08, -3.4762e-07, 1.9441e-08, + 3.7579e-07, 2.7358e-08, -4.2899e-08, -1.5600e-07, 4.9011e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 217.35, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4382 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 430, weight, value: tensor([[ 0.0330, -0.1041, -0.0310, ..., -0.1562, -0.1867, -0.1785], + [-0.1611, 0.0883, -0.1518, ..., -0.2507, -0.2192, -0.2228], + [-0.1547, -0.2080, -0.2505, ..., -0.2297, -0.1852, 0.1811], + ..., + [ 0.0916, -0.0817, -0.2126, ..., 0.1137, -0.1919, -0.1649], + [ 0.1423, 0.1054, 0.1129, ..., -0.2366, -0.1396, 0.0949], + [ 0.0852, 0.0642, 0.1353, ..., 0.0577, -0.1043, -0.0196]], + device='cuda:0'), grad: tensor([[-8.1491e-10, 1.0477e-09, 4.0745e-10, ..., 3.4925e-10, + 1.3970e-09, 1.8044e-09], + [ 1.0536e-08, -1.2806e-09, 4.4238e-09, ..., 2.3632e-08, + 2.2526e-08, 2.2643e-08], + [ 4.3656e-09, 4.7148e-09, 1.9791e-09, ..., 1.1642e-09, + 1.5716e-09, -1.6880e-09], + ..., + [ 3.7253e-09, 1.4086e-08, 1.8626e-09, ..., 3.1432e-09, + 2.5611e-09, 5.2969e-09], + [-3.7078e-08, -2.3050e-08, -1.5250e-08, ..., 1.7462e-10, + 3.6089e-09, -2.3050e-08], + [ 2.1537e-09, 2.0373e-09, 8.1491e-10, ..., 1.6415e-08, + 1.4028e-08, 8.1491e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0185, -0.0292, -0.0073, 0.0140, -0.0017, -0.0155, 0.0050, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 0.0000e+00, 3.7719e-08, 9.0222e-09, -2.6252e-08, -1.0757e-07, + -5.4482e-08, 1.1816e-07, 8.6264e-08, -9.1095e-08, 4.9185e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 217.30, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4442 re_mapping 0.0016 re_causal 0.0072 /// teacc 99.23 lr 0.00001000 +Epoch 431, weight, value: tensor([[ 0.0331, -0.1041, -0.0310, ..., -0.1562, -0.1867, -0.1785], + [-0.1612, 0.0883, -0.1519, ..., -0.2509, -0.2192, -0.2228], + [-0.1547, -0.2080, -0.2506, ..., -0.2298, -0.1852, 0.1812], + ..., + [ 0.0916, -0.0817, -0.2128, ..., 0.1137, -0.1919, -0.1650], + [ 0.1424, 0.1055, 0.1130, ..., -0.2367, -0.1396, 0.0949], + [ 0.0852, 0.0642, 0.1354, ..., 0.0577, -0.1043, -0.0196]], + device='cuda:0'), grad: tensor([[-1.6880e-09, 1.2224e-09, 2.3283e-10, ..., 4.6566e-10, + 9.8953e-10, 1.5134e-09], + [ 1.6298e-09, -9.4296e-09, 1.7462e-10, ..., 1.9209e-09, + 6.4028e-10, 9.1968e-09], + [ 1.1642e-10, 1.3970e-09, 0.0000e+00, ..., 2.5029e-09, + 2.3283e-10, -5.3027e-08], + ..., + [-3.8999e-09, 3.6671e-09, 1.0477e-09, ..., -7.9162e-09, + 2.3283e-10, 1.1642e-08], + [ 1.2224e-09, 5.9372e-09, 4.0745e-10, ..., 1.4552e-09, + 4.7148e-09, 2.9337e-08], + [-5.0059e-09, -9.3714e-09, -8.9640e-09, ..., -1.1234e-08, + -2.4447e-09, -5.8208e-10]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0185, -0.0292, -0.0072, 0.0140, -0.0017, -0.0155, 0.0050, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 2.8522e-09, 5.8208e-11, -1.7078e-07, 1.4063e-07, 4.4995e-08, + -1.2852e-07, -1.6007e-08, 4.0920e-08, 1.2456e-07, -2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 217.53, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3994 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 432, weight, value: tensor([[ 0.0331, -0.1041, -0.0311, ..., -0.1562, -0.1867, -0.1785], + [-0.1613, 0.0883, -0.1519, ..., -0.2510, -0.2192, -0.2228], + [-0.1548, -0.2080, -0.2506, ..., -0.2298, -0.1852, 0.1812], + ..., + [ 0.0916, -0.0817, -0.2129, ..., 0.1137, -0.1919, -0.1651], + [ 0.1425, 0.1055, 0.1131, ..., -0.2369, -0.1396, 0.0949], + [ 0.0852, 0.0642, 0.1355, ..., 0.0577, -0.1043, -0.0196]], + device='cuda:0'), grad: tensor([[ 2.9104e-10, 6.4028e-10, 3.4925e-10, ..., 5.8208e-10, + 7.5670e-10, 1.0477e-09], + [ 1.1758e-08, 8.7311e-10, 2.5611e-09, ..., 1.0303e-08, + 5.8208e-10, 3.7835e-09], + [ 8.6147e-09, 1.1642e-09, 6.9849e-10, ..., 9.9535e-09, + 5.8208e-10, 1.1059e-09], + ..., + [-2.6892e-08, 6.2282e-09, 3.5507e-09, ..., -3.8533e-08, + 2.1537e-09, 5.4715e-09], + [-3.5565e-08, -3.2946e-08, -2.6484e-08, ..., 3.8417e-09, + -1.3970e-09, -3.3819e-08], + [ 1.4727e-08, -1.8626e-09, -5.8208e-10, ..., 1.8103e-08, + 7.2760e-09, 2.0373e-09]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0185, -0.0292, -0.0072, 0.0140, -0.0017, -0.0155, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 3.1781e-08, 5.1048e-08, 3.9174e-08, 1.1120e-06, 7.1595e-09, + -1.5860e-06, 3.5274e-07, -5.7044e-08, -6.8103e-09, 7.5263e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 217.44, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4280 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 +Epoch 433, weight, value: tensor([[ 0.0329, -0.1042, -0.0311, ..., -0.1565, -0.1867, -0.1786], + [-0.1613, 0.0883, -0.1520, ..., -0.2511, -0.2192, -0.2229], + [-0.1548, -0.2080, -0.2507, ..., -0.2298, -0.1853, 0.1812], + ..., + [ 0.0917, -0.0817, -0.2129, ..., 0.1137, -0.1920, -0.1651], + [ 0.1425, 0.1055, 0.1132, ..., -0.2370, -0.1396, 0.0949], + [ 0.0852, 0.0642, 0.1356, ..., 0.0577, -0.1043, -0.0197]], + device='cuda:0'), grad: tensor([[ 7.5670e-10, 3.4925e-10, 4.0745e-10, ..., 1.8626e-09, + 1.2224e-09, 1.3388e-09], + [ 8.0327e-09, 1.1642e-10, 4.6566e-10, ..., 2.1188e-08, + 1.4552e-09, 1.6298e-09], + [ 1.0303e-08, 5.2387e-10, 2.3283e-10, ..., 2.8289e-08, + 2.5029e-09, 1.8626e-09], + ..., + [-1.5832e-08, 1.9791e-09, 1.2224e-09, ..., -6.5076e-08, + -3.7835e-09, -2.4447e-09], + [ 1.2806e-09, -1.6298e-09, -1.6880e-09, ..., 4.7730e-09, + 1.6880e-09, -2.9104e-10], + [-3.1432e-09, -3.7835e-09, -3.8999e-09, ..., 1.0012e-08, + 1.3097e-08, 4.0163e-09]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0184, -0.0292, -0.0072, 0.0140, -0.0017, -0.0155, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 8.8476e-09, 8.7719e-08, 8.6613e-08, 4.9011e-08, -3.6089e-09, + -2.1830e-06, 8.0327e-09, 1.5479e-06, 3.9628e-07, 2.1770e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 217.83, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4209 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.13 lr 0.00001000 +Epoch 434, weight, value: tensor([[ 0.0329, -0.1042, -0.0311, ..., -0.1565, -0.1867, -0.1786], + [-0.1614, 0.0884, -0.1521, ..., -0.2513, -0.2192, -0.2229], + [-0.1549, -0.2080, -0.2507, ..., -0.2299, -0.1853, 0.1813], + ..., + [ 0.0917, -0.0817, -0.2130, ..., 0.1137, -0.1921, -0.1652], + [ 0.1426, 0.1056, 0.1133, ..., -0.2371, -0.1396, 0.0950], + [ 0.0852, 0.0642, 0.1356, ..., 0.0577, -0.1043, -0.0197]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 2.5611e-09, 9.3132e-10, ..., 5.8208e-10, + 4.0745e-09, 5.8208e-09], + [ 4.6566e-09, 8.4401e-09, 2.9104e-09, ..., 9.3132e-09, + 1.4028e-08, 1.8917e-08], + [ 5.7626e-09, 1.1816e-08, 2.7940e-09, ..., 4.8312e-09, + 7.3924e-09, 6.2282e-09], + ..., + [-1.6880e-09, 5.8208e-09, 1.7462e-09, ..., 2.3865e-09, + 5.5297e-09, 6.0536e-09], + [-2.8929e-08, -5.6403e-08, -2.3167e-08, ..., 2.5029e-09, + -2.2002e-08, -8.1316e-08], + [-9.3132e-10, 2.1537e-09, -6.9849e-10, ..., 5.6461e-09, + 1.9732e-08, 1.8917e-08]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0184, -0.0292, -0.0072, 0.0140, -0.0017, -0.0155, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 1.7637e-08, 7.0431e-08, 3.8650e-08, -2.2410e-08, -4.4354e-08, + 6.0943e-08, 3.8825e-08, 2.2119e-08, -2.2794e-07, 5.4366e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 217.51, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4226 re_mapping 0.0016 re_causal 0.0070 /// teacc 99.13 lr 0.00001000 +Epoch 435, weight, value: tensor([[ 0.0329, -0.1041, -0.0311, ..., -0.1565, -0.1868, -0.1787], + [-0.1614, 0.0883, -0.1522, ..., -0.2514, -0.2192, -0.2229], + [-0.1550, -0.2080, -0.2508, ..., -0.2299, -0.1853, 0.1813], + ..., + [ 0.0917, -0.0817, -0.2131, ..., 0.1137, -0.1921, -0.1652], + [ 0.1426, 0.1056, 0.1134, ..., -0.2372, -0.1396, 0.0950], + [ 0.0852, 0.0643, 0.1357, ..., 0.0577, -0.1044, -0.0197]], + device='cuda:0'), grad: tensor([[-3.7835e-09, 6.9849e-10, -5.8208e-11, ..., -2.3283e-10, + 9.0804e-09, 9.4296e-09], + [ 4.5402e-09, -1.1758e-08, 2.9104e-09, ..., 3.4925e-10, + 9.3132e-09, 2.2643e-08], + [ 1.0361e-08, 2.6543e-08, 7.9162e-09, ..., 2.3283e-10, + 2.7940e-09, 4.0338e-08], + ..., + [ 4.1910e-09, 1.3388e-08, 2.5029e-09, ..., 1.1642e-09, + 1.1642e-10, 1.0419e-08], + [-2.0722e-08, -5.5530e-08, -2.0198e-08, ..., 6.4028e-10, + 2.1560e-07, 1.3155e-07], + [ 3.3760e-09, 1.9209e-09, 5.8208e-10, ..., 1.9209e-09, + 2.0955e-09, 3.7253e-09]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0185, -0.0293, -0.0072, 0.0140, -0.0017, -0.0155, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 3.6671e-09, -2.0547e-08, 1.7928e-07, 1.3201e-07, 7.4564e-08, + 1.6624e-07, -1.0030e-06, 9.4355e-08, 3.2037e-07, 6.6473e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 217.51, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4061 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.19 lr 0.00001000 +Epoch 436, weight, value: tensor([[ 0.0328, -0.1042, -0.0311, ..., -0.1567, -0.1868, -0.1787], + [-0.1615, 0.0883, -0.1522, ..., -0.2516, -0.2192, -0.2230], + [-0.1551, -0.2081, -0.2508, ..., -0.2300, -0.1853, 0.1813], + ..., + [ 0.0917, -0.0817, -0.2132, ..., 0.1137, -0.1921, -0.1652], + [ 0.1427, 0.1057, 0.1134, ..., -0.2374, -0.1396, 0.0951], + [ 0.0852, 0.0643, 0.1358, ..., 0.0577, -0.1044, -0.0197]], + device='cuda:0'), grad: tensor([[ 1.7462e-10, 1.1642e-10, 0.0000e+00, ..., 1.7462e-10, + 1.3970e-09, 1.2806e-09], + [ 4.3074e-09, -5.8208e-10, 4.0745e-10, ..., 3.8999e-09, + 6.9849e-10, 3.3178e-09], + [ 4.4238e-09, 8.7311e-10, 1.1642e-10, ..., 5.1805e-09, + 5.2387e-10, -4.4820e-09], + ..., + [-1.5192e-08, 2.7358e-09, 2.3283e-10, ..., -1.5192e-08, + 1.0477e-09, 2.2119e-09], + [-2.7358e-09, -5.7626e-09, -2.2701e-09, ..., 2.0373e-09, + 2.0373e-09, -1.5134e-09], + [ 3.7253e-09, 2.9104e-10, 5.8208e-11, ..., 6.8103e-09, + 2.3283e-09, 1.3388e-09]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0184, -0.0293, -0.0072, 0.0140, -0.0017, -0.0155, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 4.7148e-09, 1.8743e-08, 5.0059e-09, 2.4331e-08, -5.5879e-09, + 1.6531e-08, -2.0547e-08, -4.9127e-08, 0.0000e+00, 2.2817e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 217.61, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3804 re_mapping 0.0016 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 437, weight, value: tensor([[ 0.0328, -0.1042, -0.0312, ..., -0.1567, -0.1868, -0.1787], + [-0.1616, 0.0883, -0.1523, ..., -0.2517, -0.2192, -0.2230], + [-0.1550, -0.2081, -0.2509, ..., -0.2300, -0.1853, 0.1814], + ..., + [ 0.0917, -0.0817, -0.2133, ..., 0.1137, -0.1922, -0.1653], + [ 0.1427, 0.1057, 0.1134, ..., -0.2375, -0.1396, 0.0951], + [ 0.0852, 0.0643, 0.1359, ..., 0.0577, -0.1044, -0.0197]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 4.6566e-10, ..., 1.8626e-09, + 3.4925e-10, 6.4028e-10], + [ 3.5507e-09, 6.5775e-09, 1.5134e-09, ..., 3.5507e-09, + 8.1491e-10, 1.6182e-08], + [-3.0850e-09, 2.4389e-08, 4.8894e-09, ..., 9.3132e-10, + 5.2387e-10, 4.7323e-08], + ..., + [ 7.3924e-09, 1.2340e-08, 4.1910e-09, ..., 1.5716e-08, + 9.3132e-10, 1.0303e-08], + [ 4.4820e-09, -4.1386e-08, -6.2282e-09, ..., 7.2760e-09, + 4.5984e-09, -9.7381e-08], + [-2.9628e-08, -3.2305e-08, -1.5891e-08, ..., -5.3551e-08, + 6.2864e-09, -8.7311e-10]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0184, -0.0293, -0.0072, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 6.2282e-09, 5.0466e-08, 1.0349e-07, 1.6415e-08, 7.3458e-08, + -9.2899e-08, 1.0029e-07, 6.8510e-08, -1.6741e-07, -1.4447e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 217.86, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4107 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.14 lr 0.00001000 +Epoch 438, weight, value: tensor([[ 0.0328, -0.1042, -0.0312, ..., -0.1568, -0.1868, -0.1787], + [-0.1616, 0.0884, -0.1523, ..., -0.2518, -0.2192, -0.2230], + [-0.1551, -0.2081, -0.2509, ..., -0.2300, -0.1853, 0.1814], + ..., + [ 0.0917, -0.0817, -0.2134, ..., 0.1137, -0.1922, -0.1654], + [ 0.1427, 0.1057, 0.1135, ..., -0.2376, -0.1396, 0.0951], + [ 0.0852, 0.0643, 0.1360, ..., 0.0577, -0.1044, -0.0198]], + device='cuda:0'), grad: tensor([[ 4.0745e-10, 2.0955e-09, 2.3283e-10, ..., 3.4925e-10, + 6.9849e-10, 9.8953e-10], + [ 5.1223e-09, -8.9873e-08, 9.8953e-10, ..., -1.0477e-09, + 1.1642e-10, -5.1805e-09], + [ 1.0594e-08, 1.1001e-08, 1.0477e-09, ..., 1.2165e-08, + 5.8208e-11, 1.3970e-09], + ..., + [-3.8417e-09, 8.9582e-08, 3.8999e-09, ..., -3.1432e-09, + 0.0000e+00, 6.3446e-09], + [-1.5367e-08, -1.4086e-08, -1.2456e-08, ..., 3.3760e-09, + 7.5670e-10, -1.9209e-08], + [-3.7253e-09, 1.6531e-08, -1.1059e-09, ..., -1.2806e-08, + 0.0000e+00, 5.9954e-09]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0184, -0.0293, -0.0072, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 9.1386e-09, -2.7264e-07, 7.2760e-08, -4.6275e-08, 1.6124e-08, + 1.1234e-08, 4.1910e-09, 1.9558e-07, -3.0734e-08, 5.1339e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 217.68, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4306 re_mapping 0.0016 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 +Epoch 439, weight, value: tensor([[ 0.0326, -0.1044, -0.0312, ..., -0.1570, -0.1868, -0.1788], + [-0.1618, 0.0884, -0.1524, ..., -0.2521, -0.2193, -0.2230], + [-0.1551, -0.2081, -0.2510, ..., -0.2300, -0.1853, 0.1815], + ..., + [ 0.0917, -0.0817, -0.2135, ..., 0.1137, -0.1922, -0.1654], + [ 0.1428, 0.1057, 0.1135, ..., -0.2377, -0.1396, 0.0952], + [ 0.0853, 0.0643, 0.1361, ..., 0.0577, -0.1044, -0.0198]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 2.5611e-09, 5.2387e-10, ..., 8.7311e-10, + 3.2713e-08, 1.6589e-08], + [ 3.4925e-09, -6.9267e-08, 1.4552e-09, ..., 1.5716e-09, + 2.2119e-09, -5.9546e-08], + [ 8.1491e-10, 8.2655e-09, 2.9104e-10, ..., 2.3283e-10, + 6.9849e-10, 7.1595e-09], + ..., + [ 3.5507e-09, 6.9267e-09, 3.3178e-09, ..., 5.8208e-09, + 5.2387e-10, 1.8044e-09], + [-9.3132e-09, 1.0012e-08, -1.6880e-09, ..., 2.1537e-09, + 3.6089e-09, 9.0222e-09], + [-1.0419e-08, -1.7812e-08, -1.3970e-08, ..., -1.9150e-08, + 4.6566e-09, 2.5611e-09]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0183, -0.0294, -0.0072, 0.0140, -0.0017, -0.0156, 0.0052, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 8.5798e-08, -2.3074e-07, 3.0675e-08, -3.4343e-09, 7.0140e-08, + 3.1607e-08, 2.8871e-08, 3.0093e-08, 4.4471e-08, -6.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 217.64, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4185 re_mapping 0.0016 re_causal 0.0070 /// teacc 99.23 lr 0.00001000 +Epoch 440, weight, value: tensor([[ 0.0327, -0.1044, -0.0313, ..., -0.1570, -0.1868, -0.1788], + [-0.1618, 0.0884, -0.1525, ..., -0.2523, -0.2193, -0.2231], + [-0.1552, -0.2081, -0.2510, ..., -0.2301, -0.1853, 0.1815], + ..., + [ 0.0917, -0.0817, -0.2136, ..., 0.1137, -0.1922, -0.1654], + [ 0.1428, 0.1057, 0.1135, ..., -0.2379, -0.1397, 0.0952], + [ 0.0853, 0.0643, 0.1362, ..., 0.0577, -0.1044, -0.0198]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 2.3283e-10, 5.8208e-11, ..., 1.7462e-10, + 1.5134e-09, 1.3388e-09], + [ 5.0059e-09, -3.3760e-09, 5.8208e-11, ..., 3.9581e-09, + 8.1491e-10, 3.6089e-09], + [ 5.1805e-09, 7.5670e-10, 0.0000e+00, ..., 6.3446e-09, + 4.6566e-10, -5.2387e-10], + ..., + [-9.8953e-09, 3.7253e-09, 4.0745e-10, ..., -1.4668e-08, + 0.0000e+00, 1.2806e-09], + [-1.0768e-08, -4.5402e-09, 1.7462e-10, ..., 2.0955e-09, + 8.0327e-09, 8.1491e-10], + [ 2.3283e-09, -4.0745e-10, -6.4028e-10, ..., 2.7940e-09, + 1.7462e-10, 2.9104e-10]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0183, -0.0294, -0.0072, 0.0140, -0.0017, -0.0156, 0.0052, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 4.4820e-09, 7.5670e-09, 1.8510e-08, 4.4820e-09, 1.5076e-08, + 6.1351e-08, -5.7567e-08, -2.4214e-08, -1.7055e-08, 9.4878e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 217.68, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4060 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.22 lr 0.00001000 +Epoch 441, weight, value: tensor([[ 0.0327, -0.1044, -0.0313, ..., -0.1570, -0.1869, -0.1788], + [-0.1619, 0.0884, -0.1526, ..., -0.2524, -0.2193, -0.2231], + [-0.1553, -0.2081, -0.2511, ..., -0.2301, -0.1853, 0.1816], + ..., + [ 0.0917, -0.0817, -0.2137, ..., 0.1137, -0.1923, -0.1655], + [ 0.1430, 0.1058, 0.1136, ..., -0.2380, -0.1397, 0.0952], + [ 0.0853, 0.0643, 0.1363, ..., 0.0577, -0.1044, -0.0198]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.2387e-10, 8.7311e-10, ..., 5.8208e-11, + 4.8312e-09, 4.7148e-09], + [ 2.7940e-09, -2.2119e-09, 5.2387e-10, ..., 3.3178e-09, + 2.5611e-09, 2.3865e-09], + [ 3.4925e-09, 3.6089e-09, 1.7462e-10, ..., 4.7148e-09, + 4.6566e-10, 7.5670e-10], + ..., + [-7.1595e-09, 6.2864e-09, 5.8208e-11, ..., -1.0361e-08, + 1.1642e-10, 8.7311e-10], + [-2.0373e-09, -1.2806e-09, 2.0955e-09, ..., 6.9849e-10, + 1.4785e-08, 1.1467e-08], + [ 1.7462e-09, 6.9849e-10, 0.0000e+00, ..., 2.3283e-09, + 4.0745e-10, 3.4925e-10]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0183, -0.0294, -0.0072, 0.0140, -0.0017, -0.0156, 0.0052, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([ 2.0023e-08, -3.7835e-09, 2.7183e-08, -2.4738e-08, -1.3039e-08, + -9.3947e-08, -1.5658e-08, 4.5984e-09, 1.0186e-07, 1.1991e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 217.50, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4146 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.23 lr 0.00001000 +Epoch 442, weight, value: tensor([[ 0.0327, -0.1045, -0.0314, ..., -0.1570, -0.1869, -0.1789], + [-0.1620, 0.0884, -0.1527, ..., -0.2524, -0.2193, -0.2231], + [-0.1554, -0.2081, -0.2511, ..., -0.2302, -0.1853, 0.1816], + ..., + [ 0.0917, -0.0817, -0.2138, ..., 0.1137, -0.1924, -0.1655], + [ 0.1431, 0.1058, 0.1137, ..., -0.2381, -0.1397, 0.0953], + [ 0.0853, 0.0643, 0.1364, ..., 0.0577, -0.1045, -0.0198]], + device='cuda:0'), grad: tensor([[-9.3714e-09, 7.5670e-10, 4.6566e-10, ..., 1.1059e-09, + 7.0431e-09, 5.7626e-09], + [ 4.1910e-09, 6.2864e-09, 2.9104e-10, ..., 2.1362e-08, + 1.8685e-08, 4.9477e-08], + [ 1.3388e-09, 2.3283e-09, 1.1642e-10, ..., 2.6776e-09, + 5.8208e-10, -8.3644e-08], + ..., + [-7.7416e-09, 5.0641e-09, 2.4447e-09, ..., -8.8476e-09, + 4.6566e-10, 1.4086e-08], + [ 5.1805e-09, 6.4028e-10, -5.2387e-10, ..., 6.6357e-09, + 2.5029e-09, 9.5461e-09], + [-1.0419e-08, -4.7730e-09, -7.7416e-09, ..., -4.0163e-09, + 1.2456e-08, 6.5775e-09]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0184, -0.0294, -0.0072, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0363, 0.0123], device='cuda:0'), grad: tensor([-3.2014e-09, 1.9616e-07, -2.7963e-07, 3.4343e-09, -3.1956e-08, + 7.3342e-08, -2.5844e-08, 3.0734e-08, 6.5542e-08, -8.7311e-09], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 441---------------------------------------------------- +epoch 441, time 218.38, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4096 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.24 lr 0.00001000 +Epoch 443, weight, value: tensor([[ 0.0328, -0.1045, -0.0314, ..., -0.1570, -0.1869, -0.1789], + [-0.1620, 0.0884, -0.1528, ..., -0.2525, -0.2193, -0.2232], + [-0.1555, -0.2081, -0.2512, ..., -0.2302, -0.1853, 0.1816], + ..., + [ 0.0917, -0.0817, -0.2139, ..., 0.1137, -0.1924, -0.1656], + [ 0.1432, 0.1059, 0.1139, ..., -0.2382, -0.1396, 0.0954], + [ 0.0853, 0.0643, 0.1365, ..., 0.0577, -0.1045, -0.0198]], + device='cuda:0'), grad: tensor([[ 1.7462e-09, 2.4447e-09, 1.1059e-09, ..., 3.8417e-09, + 4.6566e-10, 2.3283e-10], + [ 1.4144e-08, -1.3912e-08, 7.9162e-09, ..., 3.9057e-08, + 8.7311e-10, -4.1910e-09], + [ 1.8044e-09, 2.0955e-08, 4.6566e-10, ..., 2.6193e-09, + 5.8208e-10, 3.3760e-09], + ..., + [ 8.2480e-08, 8.9232e-08, 6.2806e-08, ..., 1.2876e-07, + 1.9034e-08, 3.0268e-09], + [ 1.2806e-09, 1.3970e-09, 1.1059e-09, ..., 7.4506e-09, + 1.2806e-09, -2.9686e-09], + [-1.3201e-07, -1.4307e-07, -9.2783e-08, ..., -2.5402e-07, + -1.6240e-08, 3.5507e-09]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0184, -0.0294, -0.0072, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 1.3679e-08, -2.9511e-08, 9.4820e-08, 6.5193e-09, 1.9907e-07, + 1.1642e-08, 1.0303e-08, 4.5914e-07, 1.9383e-08, -7.6788e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 217.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4081 re_mapping 0.0016 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +Epoch 444, weight, value: tensor([[ 0.0327, -0.1046, -0.0315, ..., -0.1570, -0.1869, -0.1789], + [-0.1622, 0.0884, -0.1529, ..., -0.2528, -0.2193, -0.2232], + [-0.1555, -0.2081, -0.2512, ..., -0.2302, -0.1853, 0.1817], + ..., + [ 0.0917, -0.0817, -0.2141, ..., 0.1137, -0.1925, -0.1657], + [ 0.1434, 0.1060, 0.1141, ..., -0.2383, -0.1396, 0.0955], + [ 0.0853, 0.0643, 0.1366, ..., 0.0577, -0.1045, -0.0199]], + device='cuda:0'), grad: tensor([[ 4.0745e-10, 1.1642e-09, 1.7462e-10, ..., 6.4028e-10, + 3.7253e-09, 2.6776e-09], + [ 2.9686e-09, -1.8626e-08, 6.4028e-10, ..., 2.2701e-09, + 1.5134e-09, 2.5611e-09], + [ 2.1537e-09, 9.4180e-08, 4.0745e-10, ..., 1.3388e-09, + 2.3865e-09, 2.7358e-09], + ..., + [ 2.0780e-08, 1.4284e-07, 5.6461e-09, ..., 3.9581e-08, + 5.8208e-11, 4.0745e-10], + [-3.2014e-09, 1.2922e-08, -1.4552e-09, ..., 1.6880e-09, + 6.9267e-09, 9.3132e-10], + [-2.4971e-08, -1.8394e-08, -6.6939e-09, ..., -4.8254e-08, + 2.3283e-10, 1.7462e-10]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0183, -0.0294, -0.0072, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 1.4377e-08, -8.9174e-08, 3.3807e-07, -7.6368e-07, 5.4541e-08, + 1.4727e-08, -9.2841e-08, 5.2666e-07, 1.0460e-07, -8.7894e-08], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 443---------------------------------------------------- +epoch 443, time 217.94, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4626 re_mapping 0.0016 re_causal 0.0074 /// teacc 99.25 lr 0.00001000 +Epoch 445, weight, value: tensor([[ 0.0328, -0.1046, -0.0316, ..., -0.1570, -0.1869, -0.1789], + [-0.1624, 0.0884, -0.1530, ..., -0.2530, -0.2194, -0.2232], + [-0.1556, -0.2081, -0.2513, ..., -0.2303, -0.1853, 0.1817], + ..., + [ 0.0917, -0.0817, -0.2142, ..., 0.1138, -0.1925, -0.1657], + [ 0.1435, 0.1061, 0.1141, ..., -0.2385, -0.1396, 0.0956], + [ 0.0853, 0.0643, 0.1367, ..., 0.0577, -0.1046, -0.0199]], + device='cuda:0'), grad: tensor([[-6.2864e-09, 1.1059e-09, 1.1642e-10, ..., 7.5670e-10, + 5.8208e-10, 1.9791e-09], + [ 3.0268e-09, -1.0477e-08, 5.2387e-10, ..., 2.9104e-09, + 2.2701e-09, 1.4552e-08], + [-1.5134e-09, -2.2119e-09, 2.3283e-10, ..., 1.8626e-09, + 9.3132e-10, -4.9826e-08], + ..., + [ 6.0536e-09, 1.2631e-08, 4.0745e-09, ..., 6.6357e-09, + 2.7358e-09, 2.5320e-08], + [-8.0327e-09, -8.8476e-09, -1.8044e-09, ..., 2.6776e-09, + 1.3388e-09, -4.4238e-09], + [-3.6962e-08, -1.3853e-08, -3.0908e-08, ..., -1.4610e-08, + 2.6776e-08, -6.0536e-09]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0184, -0.0295, -0.0072, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([-1.4144e-08, 1.7462e-10, -1.8138e-07, 4.7788e-08, 1.1874e-08, + -2.3458e-08, 8.1956e-08, 1.3667e-07, 4.5984e-09, -3.8999e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 217.65, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4056 re_mapping 0.0016 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +Epoch 446, weight, value: tensor([[ 0.0328, -0.1046, -0.0316, ..., -0.1570, -0.1869, -0.1790], + [-0.1624, 0.0885, -0.1530, ..., -0.2530, -0.2194, -0.2233], + [-0.1556, -0.2081, -0.2513, ..., -0.2303, -0.1853, 0.1818], + ..., + [ 0.0917, -0.0817, -0.2143, ..., 0.1138, -0.1926, -0.1658], + [ 0.1435, 0.1061, 0.1142, ..., -0.2386, -0.1396, 0.0957], + [ 0.0853, 0.0643, 0.1367, ..., 0.0577, -0.1046, -0.0200]], + device='cuda:0'), grad: tensor([[-1.0477e-09, 3.0850e-09, 0.0000e+00, ..., 2.9104e-10, + 1.7462e-10, 1.1642e-10], + [ 3.4925e-10, -8.2655e-08, 1.1642e-10, ..., 5.2387e-10, + 3.4925e-10, 6.4028e-10], + [ 2.3283e-10, 2.9104e-09, 0.0000e+00, ..., 2.9104e-10, + 2.9104e-10, 1.1642e-10], + ..., + [ 8.1491e-10, 5.4482e-08, 5.8208e-10, ..., 3.1432e-09, + 1.1642e-09, 6.9849e-10], + [ 3.4925e-10, 1.3271e-08, 1.7462e-10, ..., 5.2387e-10, + 1.7462e-10, 2.3283e-10], + [-1.5134e-09, 4.1327e-09, -1.2806e-09, ..., 1.0827e-08, + 1.2398e-08, 5.3551e-09]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0184, -0.0295, -0.0071, 0.0140, -0.0017, -0.0156, 0.0051, 0.0158, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 5.7044e-09, -3.4622e-07, 1.4028e-08, 7.6252e-09, -2.3923e-08, + 1.2340e-08, 7.7416e-09, 2.3213e-07, 5.6345e-08, 5.1688e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 217.68, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4211 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.20 lr 0.00001000 +Epoch 447, weight, value: tensor([[ 0.0328, -0.1047, -0.0317, ..., -0.1572, -0.1869, -0.1790], + [-0.1625, 0.0885, -0.1531, ..., -0.2531, -0.2194, -0.2233], + [-0.1556, -0.2082, -0.2514, ..., -0.2303, -0.1853, 0.1819], + ..., + [ 0.0917, -0.0817, -0.2145, ..., 0.1138, -0.1927, -0.1659], + [ 0.1437, 0.1063, 0.1145, ..., -0.2388, -0.1396, 0.0958], + [ 0.0853, 0.0643, 0.1368, ..., 0.0577, -0.1047, -0.0200]], + device='cuda:0'), grad: tensor([[-8.4168e-08, 3.5798e-09, 5.8208e-11, ..., -4.0163e-09, + 1.4843e-09, -2.9278e-08], + [ 6.6357e-09, -1.6851e-08, 1.7171e-09, ..., 1.3795e-08, + 2.1013e-08, 2.6484e-08], + [ 1.3504e-08, 1.2515e-09, 1.1642e-10, ..., 3.1141e-09, + 3.6671e-09, 2.8231e-09], + ..., + [ 5.6752e-09, 1.6909e-08, 1.2515e-09, ..., 4.5402e-09, + 2.3283e-10, 3.2014e-09], + [-1.0128e-08, -4.5926e-08, -1.2311e-08, ..., 5.8208e-10, + -9.0222e-10, -6.5193e-08], + [ 9.6625e-09, -1.3970e-09, -8.7311e-10, ..., -1.5716e-09, + 3.4634e-09, 6.9558e-09]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0183, -0.0295, -0.0071, 0.0140, -0.0017, -0.0157, 0.0051, 0.0158, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([-6.3656e-07, 6.4028e-09, 1.0582e-07, 2.7008e-08, -1.2405e-06, + 1.5309e-07, 1.5628e-06, 8.9407e-08, -1.3481e-07, 9.5461e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 217.63, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4311 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.22 lr 0.00001000 +Epoch 448, weight, value: tensor([[ 0.0328, -0.1047, -0.0317, ..., -0.1572, -0.1869, -0.1790], + [-0.1626, 0.0885, -0.1532, ..., -0.2532, -0.2194, -0.2234], + [-0.1557, -0.2082, -0.2514, ..., -0.2304, -0.1853, 0.1820], + ..., + [ 0.0917, -0.0817, -0.2146, ..., 0.1138, -0.1928, -0.1660], + [ 0.1438, 0.1064, 0.1146, ..., -0.2389, -0.1396, 0.0959], + [ 0.0853, 0.0643, 0.1369, ..., 0.0577, -0.1047, -0.0201]], + device='cuda:0'), grad: tensor([[-2.7067e-09, 9.6043e-10, 1.5425e-09, ..., 8.2946e-09, + 2.5902e-09, 2.5029e-09], + [ 4.2492e-09, 4.3656e-10, 3.2014e-10, ..., 5.2678e-09, + 7.2760e-10, 6.3737e-09], + [ 4.4820e-09, 1.7462e-10, 1.7462e-10, ..., 3.6380e-09, + 2.6193e-10, -2.1071e-08], + ..., + [-6.7812e-09, 1.8335e-09, 2.7067e-09, ..., -3.0850e-09, + 1.3679e-09, 8.2655e-09], + [-2.0373e-10, -1.8626e-09, -6.6939e-10, ..., 2.9104e-09, + 2.3574e-09, 3.6962e-09], + [-5.6461e-09, -3.4051e-09, -5.9081e-09, ..., -2.4884e-08, + -2.0082e-09, -1.8044e-09]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0184, -0.0295, -0.0071, 0.0140, -0.0017, -0.0157, 0.0051, 0.0158, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([-9.4005e-09, 4.7818e-08, -7.1595e-08, 2.4622e-08, 2.4767e-08, + 3.3353e-08, -2.4796e-08, -1.8917e-08, 2.6746e-08, -1.9703e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 217.53, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4037 re_mapping 0.0016 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 449, weight, value: tensor([[ 0.0329, -0.1047, -0.0318, ..., -0.1572, -0.1869, -0.1791], + [-0.1628, 0.0885, -0.1533, ..., -0.2534, -0.2195, -0.2234], + [-0.1558, -0.2082, -0.2515, ..., -0.2304, -0.1853, 0.1820], + ..., + [ 0.0917, -0.0817, -0.2147, ..., 0.1138, -0.1929, -0.1661], + [ 0.1439, 0.1064, 0.1147, ..., -0.2392, -0.1396, 0.0960], + [ 0.0853, 0.0643, 0.1371, ..., 0.0577, -0.1048, -0.0201]], + device='cuda:0'), grad: tensor([[ 4.3656e-10, 4.9477e-10, 1.4552e-10, ..., 4.0745e-09, + 1.0448e-08, 4.0745e-09], + [ 4.5984e-09, 1.5134e-09, 7.5670e-10, ..., 2.4855e-08, + 9.3132e-10, 2.7067e-09], + [ 1.4872e-08, 2.3574e-09, 2.3283e-10, ..., 1.9604e-07, + 3.7835e-10, -6.4611e-09], + ..., + [-1.5978e-08, 3.2160e-08, 4.7439e-09, ..., -2.8824e-07, + 1.4552e-10, 1.5425e-09], + [ 4.1036e-09, 1.5076e-08, 2.0955e-09, ..., 6.1118e-09, + 4.1036e-09, 5.3260e-09], + [-1.5716e-09, -7.0140e-09, -5.2387e-09, ..., 2.5757e-08, + -1.1642e-10, -7.8580e-10]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0184, -0.0296, -0.0071, 0.0140, -0.0017, -0.0157, 0.0051, 0.0158, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 3.1141e-08, 5.7800e-08, 4.1560e-07, -1.4214e-07, 6.7055e-08, + 8.5798e-08, -8.2538e-08, -5.2666e-07, 7.1072e-08, 5.4104e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 217.48, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4297 re_mapping 0.0016 re_causal 0.0070 /// teacc 99.18 lr 0.00001000 +Epoch 450, weight, value: tensor([[ 0.0329, -0.1047, -0.0318, ..., -0.1572, -0.1870, -0.1792], + [-0.1629, 0.0885, -0.1534, ..., -0.2536, -0.2195, -0.2235], + [-0.1559, -0.2082, -0.2516, ..., -0.2305, -0.1854, 0.1820], + ..., + [ 0.0917, -0.0817, -0.2148, ..., 0.1138, -0.1929, -0.1662], + [ 0.1440, 0.1065, 0.1148, ..., -0.2393, -0.1396, 0.0960], + [ 0.0853, 0.0643, 0.1372, ..., 0.0577, -0.1048, -0.0202]], + device='cuda:0'), grad: tensor([[ 1.0477e-09, 1.1642e-09, 6.9849e-10, ..., 1.2224e-09, + 1.4843e-09, 1.9791e-09], + [ 4.3656e-09, 2.3283e-09, 3.6089e-09, ..., 2.7649e-09, + 4.2783e-09, 7.9744e-09], + [ 5.8208e-10, 1.0477e-09, 4.3656e-10, ..., 7.5670e-10, + 1.0477e-09, -1.4552e-09], + ..., + [ 6.3737e-09, 7.5961e-09, 1.9209e-09, ..., 1.3009e-08, + 1.4843e-09, 1.9209e-09], + [-2.8347e-08, -2.8289e-08, -3.2654e-08, ..., 3.6671e-09, + -1.0186e-08, -2.0693e-08], + [-1.4319e-08, -1.1787e-08, 6.6939e-10, ..., -2.6048e-08, + 1.6444e-08, 1.3504e-08]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0184, -0.0296, -0.0071, 0.0140, -0.0017, -0.0157, 0.0052, 0.0158, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 2.0373e-08, 4.3568e-08, 7.8580e-10, 5.0728e-08, 2.7736e-08, + -2.7451e-07, 9.0047e-08, 7.6601e-08, 2.0955e-09, -2.8522e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 217.42, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4043 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 451, weight, value: tensor([[ 0.0330, -0.1047, -0.0318, ..., -0.1572, -0.1870, -0.1792], + [-0.1630, 0.0885, -0.1535, ..., -0.2536, -0.2195, -0.2235], + [-0.1560, -0.2082, -0.2517, ..., -0.2305, -0.1854, 0.1821], + ..., + [ 0.0917, -0.0817, -0.2150, ..., 0.1138, -0.1930, -0.1662], + [ 0.1441, 0.1065, 0.1149, ..., -0.2395, -0.1396, 0.0961], + [ 0.0853, 0.0643, 0.1373, ..., 0.0577, -0.1048, -0.0202]], + device='cuda:0'), grad: tensor([[ 8.4401e-10, 3.7835e-10, 1.4552e-10, ..., 2.9104e-11, + 1.2486e-08, 1.3039e-08], + [ 7.6834e-09, 2.9977e-09, 1.3970e-09, ..., 1.6298e-09, + 3.1752e-08, 6.1817e-08], + [ 5.5879e-09, 1.2806e-09, 6.6939e-10, ..., 4.8021e-09, + 7.1595e-09, -1.9209e-07], + ..., + [ 1.7957e-08, 8.4983e-09, 6.9849e-10, ..., -9.4587e-09, + 4.0745e-10, 1.4703e-07], + [-4.5373e-08, -1.9470e-08, -8.8185e-09, ..., 1.2224e-09, + 2.2555e-08, -9.1095e-09], + [ 1.2515e-09, 3.7835e-10, 1.1642e-10, ..., 1.1933e-09, + 1.0477e-09, 3.7544e-09]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0184, -0.0296, -0.0071, 0.0140, -0.0017, -0.0158, 0.0052, 0.0158, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 5.2940e-08, 2.2480e-07, -4.8149e-07, 3.0355e-08, 1.4843e-07, + 1.1092e-06, -1.4361e-06, 3.9628e-07, -5.5152e-08, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 217.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4347 re_mapping 0.0016 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 +Epoch 452, weight, value: tensor([[ 0.0330, -0.1048, -0.0319, ..., -0.1572, -0.1870, -0.1793], + [-0.1631, 0.0886, -0.1536, ..., -0.2537, -0.2195, -0.2236], + [-0.1562, -0.2083, -0.2517, ..., -0.2306, -0.1854, 0.1821], + ..., + [ 0.0917, -0.0817, -0.2152, ..., 0.1138, -0.1931, -0.1663], + [ 0.1443, 0.1066, 0.1150, ..., -0.2397, -0.1397, 0.0962], + [ 0.0853, 0.0643, 0.1375, ..., 0.0577, -0.1049, -0.0202]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, -1.1642e-10, 3.2014e-10, ..., 4.0745e-10, + 1.4552e-10, 5.2387e-10], + [ 5.2678e-09, 2.9977e-09, 1.5425e-09, ..., 1.4843e-09, + 3.7835e-10, 5.0059e-09], + [ 1.2951e-08, 8.1200e-09, 2.9395e-09, ..., 1.0768e-09, + 1.4552e-10, 1.2922e-08], + ..., + [ 3.7544e-09, 4.0745e-09, 2.5902e-09, ..., -3.4925e-10, + 4.3656e-10, 5.0641e-09], + [-2.6717e-08, -1.5105e-08, -6.6066e-09, ..., 2.5029e-09, + 2.0373e-10, -2.8842e-08], + [-4.5984e-09, -5.0932e-09, -5.0059e-09, ..., -6.8103e-09, + 1.0186e-09, -2.3283e-10]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0184, -0.0296, -0.0071, 0.0140, -0.0017, -0.0159, 0.0053, 0.0158, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([-4.2492e-09, 2.6077e-08, 5.8266e-08, 1.0390e-08, 1.0186e-08, + 4.2783e-09, 1.5192e-08, 2.0809e-08, -1.1188e-07, -1.6880e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 217.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3942 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 453, weight, value: tensor([[ 0.0331, -0.1047, -0.0319, ..., -0.1572, -0.1870, -0.1793], + [-0.1631, 0.0886, -0.1537, ..., -0.2537, -0.2196, -0.2236], + [-0.1562, -0.2083, -0.2518, ..., -0.2306, -0.1854, 0.1822], + ..., + [ 0.0917, -0.0817, -0.2154, ..., 0.1138, -0.1932, -0.1664], + [ 0.1445, 0.1067, 0.1152, ..., -0.2398, -0.1397, 0.0963], + [ 0.0853, 0.0643, 0.1376, ..., 0.0577, -0.1049, -0.0203]], + device='cuda:0'), grad: tensor([[ 2.9104e-10, 2.3283e-10, 1.4552e-10, ..., 4.0745e-10, + 3.7835e-10, 2.6193e-10], + [ 1.2398e-08, 2.2119e-09, 6.1118e-10, ..., 1.2427e-08, + 1.2806e-09, 2.5320e-09], + [ 5.6170e-09, 6.6939e-10, 2.9104e-10, ..., 5.9663e-09, + 5.8208e-10, -1.2515e-09], + ..., + [-2.4011e-08, 3.2887e-09, 1.2806e-09, ..., -2.5757e-08, + -1.2224e-09, -1.0477e-09], + [-1.6880e-09, -1.5425e-09, -2.2701e-09, ..., 3.3178e-09, + 7.5670e-10, -2.8231e-09], + [ 3.2305e-09, -1.4843e-09, -1.7753e-09, ..., 3.3178e-09, + 2.0082e-09, 1.1350e-09]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0185, -0.0297, -0.0071, 0.0140, -0.0017, -0.0159, 0.0054, 0.0158, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 2.9977e-09, 8.2888e-08, 2.0926e-08, 1.1903e-08, 8.9349e-09, + -7.2760e-09, 8.1491e-09, -1.2887e-07, 6.5193e-09, 1.7841e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 217.60, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3889 re_mapping 0.0016 re_causal 0.0066 /// teacc 99.18 lr 0.00001000 +Epoch 454, weight, value: tensor([[ 0.0331, -0.1048, -0.0320, ..., -0.1572, -0.1871, -0.1794], + [-0.1632, 0.0886, -0.1539, ..., -0.2538, -0.2196, -0.2237], + [-0.1564, -0.2083, -0.2519, ..., -0.2307, -0.1854, 0.1823], + ..., + [ 0.0917, -0.0817, -0.2155, ..., 0.1138, -0.1932, -0.1665], + [ 0.1445, 0.1068, 0.1153, ..., -0.2401, -0.1397, 0.0964], + [ 0.0853, 0.0643, 0.1377, ..., 0.0577, -0.1050, -0.0203]], + device='cuda:0'), grad: tensor([[ 1.1467e-08, 1.1380e-08, 7.8580e-09, ..., 7.5670e-10, + 4.1327e-09, 2.2759e-08], + [ 1.2718e-08, -1.1583e-08, 7.9162e-09, ..., 3.2887e-09, + 3.4634e-09, 1.6822e-08], + [-1.9936e-08, 1.6356e-08, 2.5029e-09, ..., 2.7358e-09, + 1.3097e-09, -9.3598e-08], + ..., + [ 7.3342e-09, 1.2893e-08, 6.5775e-09, ..., -3.8417e-09, + 2.3574e-09, 1.7666e-08], + [-7.3924e-08, -1.0582e-07, -7.9861e-08, ..., 1.4552e-09, + -3.6758e-08, -4.0280e-08], + [ 1.8452e-08, 1.9878e-08, 1.4756e-08, ..., -1.2427e-08, + 1.0768e-08, 2.7649e-08]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0185, -0.0297, -0.0071, 0.0140, -0.0017, -0.0159, 0.0054, 0.0158, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 7.6019e-08, -8.8708e-08, -1.5763e-07, 2.4535e-08, 5.7829e-08, + 1.5774e-08, 1.1898e-07, 8.2131e-08, -1.7299e-07, 5.6723e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 217.51, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4360 re_mapping 0.0015 re_causal 0.0069 /// teacc 99.19 lr 0.00001000 +Epoch 455, weight, value: tensor([[ 0.0331, -0.1048, -0.0320, ..., -0.1572, -0.1871, -0.1795], + [-0.1634, 0.0886, -0.1540, ..., -0.2540, -0.2196, -0.2237], + [-0.1565, -0.2083, -0.2520, ..., -0.2308, -0.1854, 0.1824], + ..., + [ 0.0918, -0.0817, -0.2157, ..., 0.1138, -0.1932, -0.1666], + [ 0.1447, 0.1069, 0.1155, ..., -0.2402, -0.1397, 0.0965], + [ 0.0853, 0.0643, 0.1378, ..., 0.0577, -0.1051, -0.0204]], + device='cuda:0'), grad: tensor([[-2.6601e-08, 1.8335e-09, 2.3283e-10, ..., -6.7521e-09, + -9.3132e-10, 6.9849e-10], + [ 1.7637e-08, 1.2136e-08, 1.2806e-09, ..., 8.4401e-10, + 1.1642e-10, 7.4797e-09], + [-5.4133e-09, 1.0768e-09, 0.0000e+00, ..., 9.8953e-10, + 8.7311e-11, -1.0012e-08], + ..., + [ 1.4232e-08, 8.4401e-09, 1.2224e-09, ..., 2.6484e-09, + 8.7311e-11, 1.0768e-08], + [-7.1479e-08, -6.8976e-08, -4.2492e-09, ..., 1.2806e-09, + 3.7835e-10, -3.0792e-08], + [ 4.5402e-09, -2.0664e-09, -1.9791e-09, ..., -2.6193e-09, + 3.2014e-10, 1.4552e-10]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0186, -0.0298, -0.0071, 0.0140, -0.0017, -0.0159, 0.0054, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([-2.2200e-07, 6.8278e-08, -3.1432e-08, 2.2107e-07, 1.9936e-08, + -2.3318e-07, 1.9453e-07, 1.0780e-07, -1.7253e-07, 5.9168e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 217.63, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3924 re_mapping 0.0016 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 456, weight, value: tensor([[ 0.0332, -0.1048, -0.0321, ..., -0.1572, -0.1871, -0.1796], + [-0.1634, 0.0886, -0.1541, ..., -0.2540, -0.2197, -0.2238], + [-0.1566, -0.2083, -0.2521, ..., -0.2308, -0.1854, 0.1824], + ..., + [ 0.0918, -0.0817, -0.2158, ..., 0.1138, -0.1932, -0.1666], + [ 0.1449, 0.1070, 0.1156, ..., -0.2404, -0.1397, 0.0966], + [ 0.0853, 0.0643, 0.1380, ..., 0.0577, -0.1051, -0.0205]], + device='cuda:0'), grad: tensor([[-1.5134e-09, 4.3656e-10, 2.9104e-11, ..., 5.8208e-11, + 8.1491e-10, 7.2760e-10], + [ 4.0745e-10, 8.4401e-10, 8.7311e-11, ..., 4.9477e-10, + 4.0745e-10, 3.4925e-10], + [ 4.6566e-10, 1.1642e-09, 0.0000e+00, ..., 4.9477e-10, + 2.6193e-10, 2.3283e-10], + ..., + [ 2.0373e-10, 3.5216e-09, 2.3283e-10, ..., 3.7835e-10, + 0.0000e+00, 0.0000e+00], + [ 2.6193e-10, 3.6380e-09, 3.7835e-10, ..., 1.7462e-10, + 2.1537e-09, 1.8044e-09], + [-2.6193e-10, -2.6193e-10, -6.4028e-10, ..., -9.3132e-10, + 2.0373e-10, 1.1642e-10]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0186, -0.0298, -0.0071, 0.0139, -0.0017, -0.0160, 0.0054, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([-2.9395e-09, 6.1409e-09, 6.4902e-09, 2.2672e-08, 7.0431e-09, + -2.1548e-07, 6.9442e-08, 1.5541e-08, 9.9652e-08, 3.4634e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 217.49, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4252 re_mapping 0.0015 re_causal 0.0068 /// teacc 99.20 lr 0.00001000 +Epoch 457, weight, value: tensor([[ 0.0332, -0.1048, -0.0321, ..., -0.1572, -0.1871, -0.1796], + [-0.1635, 0.0886, -0.1542, ..., -0.2541, -0.2197, -0.2238], + [-0.1568, -0.2083, -0.2521, ..., -0.2309, -0.1854, 0.1825], + ..., + [ 0.0918, -0.0818, -0.2160, ..., 0.1138, -0.1933, -0.1667], + [ 0.1449, 0.1070, 0.1156, ..., -0.2405, -0.1397, 0.0967], + [ 0.0853, 0.0643, 0.1382, ..., 0.0577, -0.1051, -0.0205]], + device='cuda:0'), grad: tensor([[-4.8312e-09, 1.8335e-09, 8.7311e-11, ..., 6.9849e-10, + -1.4552e-09, 1.4552e-09], + [ 6.8976e-09, 1.2718e-08, 4.0745e-10, ..., 9.7789e-09, + 3.3178e-09, 2.1333e-08], + [ 1.3679e-09, 2.0955e-08, 9.6043e-10, ..., 1.6007e-09, + 4.3656e-10, -2.9104e-08], + ..., + [-1.6298e-09, 1.7928e-08, 1.0768e-09, ..., -1.5716e-09, + 3.7835e-10, 9.0513e-09], + [-1.6880e-09, 1.4174e-08, -2.6193e-10, ..., 1.2806e-09, + 5.1223e-09, 3.8417e-09], + [-1.8656e-08, -3.7020e-08, -1.7753e-09, ..., -2.0780e-08, + 9.6043e-09, -2.1653e-08]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0186, -0.0298, -0.0071, 0.0139, -0.0017, -0.0160, 0.0055, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([-9.9826e-09, 8.6206e-08, -2.8987e-08, -1.9278e-07, 5.7451e-08, + 2.7427e-07, -2.3912e-07, 7.6136e-08, 7.9337e-08, -1.0448e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 217.70, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4105 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.19 lr 0.00001000 +Epoch 458, weight, value: tensor([[ 0.0333, -0.1048, -0.0322, ..., -0.1572, -0.1872, -0.1797], + [-0.1636, 0.0887, -0.1543, ..., -0.2542, -0.2197, -0.2239], + [-0.1568, -0.2084, -0.2522, ..., -0.2310, -0.1854, 0.1826], + ..., + [ 0.0918, -0.0818, -0.2162, ..., 0.1138, -0.1934, -0.1669], + [ 0.1450, 0.1070, 0.1157, ..., -0.2408, -0.1397, 0.0967], + [ 0.0853, 0.0643, 0.1384, ..., 0.0577, -0.1052, -0.0206]], + device='cuda:0'), grad: tensor([[-6.6939e-10, 9.6043e-10, 8.7311e-11, ..., 1.1642e-10, + 4.6566e-10, 1.1059e-09], + [ 5.5879e-09, 7.8289e-09, 6.4028e-10, ..., 2.5320e-09, + 4.6566e-10, 1.0303e-08], + [ 3.6671e-09, 1.1729e-08, 9.6043e-10, ..., 3.3469e-09, + 5.2387e-10, -1.3184e-08], + ..., + [-3.6671e-09, 1.0070e-08, 1.0186e-09, ..., -3.7544e-09, + 4.6566e-10, 6.4611e-09], + [-1.0623e-08, 3.5798e-09, -1.0768e-09, ..., 9.6043e-10, + 7.2760e-10, -3.0268e-09], + [ 2.0955e-09, 3.1141e-09, 2.0373e-10, ..., 4.2783e-09, + 2.7940e-09, 2.7649e-09]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0187, -0.0298, -0.0070, 0.0139, -0.0018, -0.0160, 0.0055, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([ 2.3283e-09, 5.3988e-08, -1.6647e-08, -1.2061e-07, -6.6939e-10, + -1.2561e-07, 1.3690e-07, 3.5332e-08, 2.9366e-08, 2.5379e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 217.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4177 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +Epoch 459, weight, value: tensor([[ 0.0333, -0.1049, -0.0323, ..., -0.1572, -0.1872, -0.1798], + [-0.1637, 0.0887, -0.1544, ..., -0.2543, -0.2198, -0.2240], + [-0.1569, -0.2084, -0.2522, ..., -0.2311, -0.1855, 0.1827], + ..., + [ 0.0918, -0.0818, -0.2165, ..., 0.1138, -0.1933, -0.1670], + [ 0.1451, 0.1072, 0.1160, ..., -0.2410, -0.1397, 0.0968], + [ 0.0853, 0.0643, 0.1385, ..., 0.0577, -0.1053, -0.0207]], + device='cuda:0'), grad: tensor([[ 8.7311e-11, 1.7753e-09, 2.9104e-11, ..., 6.9849e-10, + 5.8790e-09, 3.8417e-09], + [ 2.5611e-09, -1.4086e-08, 6.1118e-10, ..., 7.2760e-09, + 4.2783e-09, 2.5932e-08], + [ 1.9500e-09, 2.0955e-09, 2.0373e-10, ..., -3.2887e-09, + 1.5134e-09, -7.6951e-08], + ..., + [-5.1514e-09, 9.0222e-09, 6.1118e-10, ..., -2.2992e-09, + 2.2410e-09, 1.9325e-08], + [-1.5716e-09, -1.5134e-09, -1.8044e-09, ..., 2.8522e-09, + 1.4115e-08, 1.6007e-08], + [ 1.7462e-09, -2.7940e-09, -1.8335e-09, ..., 9.9477e-08, + 9.2667e-08, 1.4494e-08]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0187, -0.0298, -0.0070, 0.0139, -0.0017, -0.0161, 0.0056, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([ 1.8190e-08, 4.9185e-08, -2.3842e-07, 6.2108e-08, -2.5379e-07, + 1.2270e-07, -1.3702e-07, 7.4273e-08, 6.8743e-08, 2.6776e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 217.64, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4126 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.16 lr 0.00001000 +Epoch 460, weight, value: tensor([[ 0.0333, -0.1049, -0.0323, ..., -0.1573, -0.1872, -0.1798], + [-0.1638, 0.0888, -0.1545, ..., -0.2544, -0.2198, -0.2240], + [-0.1570, -0.2084, -0.2523, ..., -0.2312, -0.1855, 0.1828], + ..., + [ 0.0918, -0.0818, -0.2167, ..., 0.1138, -0.1934, -0.1671], + [ 0.1453, 0.1073, 0.1161, ..., -0.2412, -0.1397, 0.0969], + [ 0.0853, 0.0644, 0.1387, ..., 0.0577, -0.1054, -0.0208]], + device='cuda:0'), grad: tensor([[-6.4611e-09, 1.7171e-09, 8.7311e-10, ..., 1.3970e-09, + 3.2305e-09, 2.9977e-09], + [ 3.5798e-09, 2.6484e-09, 1.6880e-09, ..., 4.9768e-09, + 3.6089e-09, 1.0594e-08], + [ 1.4843e-09, 1.9500e-09, 3.2014e-10, ..., -2.4447e-09, + 1.8626e-09, -3.5099e-08], + ..., + [ 1.4843e-08, 4.4616e-08, 2.3923e-08, ..., 3.6089e-08, + 1.4785e-08, 1.4086e-08], + [-3.2887e-09, 1.8452e-08, 9.3132e-10, ..., 7.8871e-09, + 1.2893e-08, 2.7183e-08], + [-2.5029e-08, -5.0408e-08, -3.9290e-08, ..., 5.9663e-08, + 8.3179e-08, 4.6246e-08]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0186, -0.0298, -0.0070, 0.0139, -0.0017, -0.0161, 0.0056, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([-6.3155e-09, 5.3173e-08, -9.8255e-08, 9.5286e-08, -2.3027e-07, + -3.7136e-07, -4.9069e-08, 2.4377e-07, 3.0082e-07, 9.0338e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 217.61, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4248 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.20 lr 0.00001000 +Epoch 461, weight, value: tensor([[ 0.0333, -0.1050, -0.0324, ..., -0.1573, -0.1873, -0.1800], + [-0.1639, 0.0888, -0.1546, ..., -0.2545, -0.2198, -0.2241], + [-0.1571, -0.2084, -0.2524, ..., -0.2312, -0.1855, 0.1829], + ..., + [ 0.0918, -0.0818, -0.2169, ..., 0.1139, -0.1935, -0.1672], + [ 0.1454, 0.1074, 0.1163, ..., -0.2414, -0.1397, 0.0970], + [ 0.0853, 0.0644, 0.1388, ..., 0.0577, -0.1056, -0.0209]], + device='cuda:0'), grad: tensor([[ 5.5297e-10, 1.7462e-10, 1.1642e-10, ..., 8.7311e-11, + 1.7462e-10, 2.3283e-10], + [ 2.9890e-08, -3.3760e-09, 4.9477e-10, ..., 3.3557e-08, + 1.7462e-10, 1.1438e-08], + [ 6.7812e-09, 1.8626e-09, 2.9104e-10, ..., 7.3924e-09, + 3.4925e-10, 2.7940e-09], + ..., + [-3.7777e-08, 3.4634e-09, 5.2387e-10, ..., -4.5140e-08, + 2.0373e-10, -1.3330e-08], + [ 7.3051e-09, -1.6880e-09, -1.5134e-09, ..., 2.6193e-09, + 1.1642e-10, -2.3283e-09], + [ 5.6461e-09, 1.4261e-09, 1.4261e-09, ..., 9.4296e-09, + 3.9290e-09, 4.6857e-09]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0186, -0.0298, -0.0070, 0.0139, -0.0017, -0.0161, 0.0057, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([ 9.5752e-09, 1.0373e-07, 3.2043e-08, 1.6793e-08, -5.2678e-09, + -4.1677e-07, 1.6100e-07, -1.2584e-07, 2.1013e-07, 3.8883e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 217.67, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4183 re_mapping 0.0015 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 +Epoch 462, weight, value: tensor([[ 0.0334, -0.1050, -0.0324, ..., -0.1573, -0.1874, -0.1801], + [-0.1640, 0.0888, -0.1547, ..., -0.2546, -0.2199, -0.2242], + [-0.1571, -0.2085, -0.2524, ..., -0.2313, -0.1855, 0.1830], + ..., + [ 0.0918, -0.0818, -0.2170, ..., 0.1139, -0.1935, -0.1673], + [ 0.1455, 0.1074, 0.1163, ..., -0.2416, -0.1398, 0.0970], + [ 0.0853, 0.0644, 0.1390, ..., 0.0577, -0.1057, -0.0210]], + device='cuda:0'), grad: tensor([[ 5.5297e-10, 4.6566e-10, 1.1642e-10, ..., 2.9104e-10, + 4.0745e-09, 5.9663e-09], + [ 5.2678e-09, -1.3126e-08, 3.2014e-10, ..., 6.0536e-09, + 1.1176e-08, 1.9412e-08], + [ 5.4424e-09, 3.6962e-09, 1.4552e-10, ..., 4.0163e-09, + 4.4238e-09, -7.1595e-09], + ..., + [-2.3749e-08, 7.1886e-09, 1.4552e-10, ..., -2.8755e-08, + 4.3656e-10, 4.5984e-09], + [-2.7649e-09, -9.3132e-10, -1.2224e-09, ..., 3.4925e-09, + 3.1223e-07, 4.1560e-07], + [ 1.0739e-08, 8.7311e-10, 5.8208e-11, ..., 2.4505e-08, + 1.0768e-08, 8.0327e-09]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0186, -0.0299, -0.0070, 0.0139, -0.0016, -0.0162, 0.0058, 0.0158, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([ 1.8714e-08, 1.3388e-08, 5.3551e-09, 3.3528e-08, 1.4552e-10, + 1.6997e-08, -1.3001e-06, -4.9069e-08, 1.1995e-06, 8.0909e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 217.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4149 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.25 lr 0.00001000 +Epoch 463, weight, value: tensor([[ 0.0334, -0.1051, -0.0325, ..., -0.1573, -0.1874, -0.1802], + [-0.1642, 0.0889, -0.1548, ..., -0.2548, -0.2199, -0.2243], + [-0.1572, -0.2085, -0.2525, ..., -0.2314, -0.1855, 0.1831], + ..., + [ 0.0918, -0.0818, -0.2172, ..., 0.1139, -0.1935, -0.1673], + [ 0.1455, 0.1075, 0.1165, ..., -0.2419, -0.1398, 0.0970], + [ 0.0853, 0.0644, 0.1391, ..., 0.0577, -0.1058, -0.0212]], + device='cuda:0'), grad: tensor([[ 1.6880e-09, 1.5716e-09, 7.8580e-10, ..., 2.2701e-09, + 3.3178e-09, 2.6776e-09], + [ 1.4028e-08, -1.2427e-08, 3.1723e-09, ..., -2.3283e-10, + 1.4552e-09, 2.6514e-08], + [-1.4377e-08, 6.2282e-09, 5.5297e-10, ..., -3.1723e-09, + 1.0477e-09, -1.0006e-07], + ..., + [ 1.5600e-08, 2.2206e-08, 6.5484e-09, ..., 1.7521e-08, + 2.3283e-09, 1.7142e-08], + [-1.0594e-08, -1.2631e-08, -6.5484e-09, ..., 5.7626e-09, + 2.9686e-09, 4.0483e-08], + [-2.5349e-08, -2.2847e-08, -1.2253e-08, ..., -7.3633e-09, + 3.4604e-08, 1.6706e-08]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0186, -0.0299, -0.0070, 0.0139, -0.0016, -0.0162, 0.0058, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 1.7229e-08, 4.2288e-08, -3.1455e-07, 4.4267e-08, -2.4243e-08, + 5.2591e-08, -3.6409e-08, 1.4855e-07, 1.1118e-07, -3.5623e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 217.64, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4209 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +Epoch 464, weight, value: tensor([[ 0.0335, -0.1051, -0.0326, ..., -0.1574, -0.1874, -0.1803], + [-0.1643, 0.0890, -0.1550, ..., -0.2549, -0.2200, -0.2244], + [-0.1574, -0.2085, -0.2526, ..., -0.2316, -0.1855, 0.1832], + ..., + [ 0.0918, -0.0818, -0.2173, ..., 0.1139, -0.1936, -0.1674], + [ 0.1457, 0.1077, 0.1168, ..., -0.2421, -0.1398, 0.0971], + [ 0.0853, 0.0644, 0.1393, ..., 0.0577, -0.1060, -0.0213]], + device='cuda:0'), grad: tensor([[-1.7753e-09, 3.7544e-09, 8.7311e-11, ..., 8.7311e-11, + 4.8312e-09, 4.8312e-09], + [ 2.9686e-09, -2.0547e-08, 5.8208e-10, ..., 1.6007e-09, + 4.6566e-10, 2.5611e-09], + [ 9.3132e-10, 2.3283e-09, 5.8208e-11, ..., 1.3679e-09, + 3.7835e-10, -5.9954e-09], + ..., + [-1.1642e-10, 1.0215e-08, 4.9477e-10, ..., -2.3865e-09, + 0.0000e+00, 1.1933e-09], + [-5.1514e-09, -4.0454e-09, -3.0559e-09, ..., 7.8580e-10, + 2.7358e-09, 1.7462e-10], + [ 5.8208e-10, 1.3970e-09, -6.1118e-10, ..., 3.2014e-10, + 5.8208e-11, 5.2387e-10]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0186, -0.0299, -0.0070, 0.0139, -0.0016, -0.0162, 0.0059, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 3.0210e-08, -5.2707e-08, -6.8103e-09, 3.3388e-07, 3.2567e-08, + -5.2061e-07, 7.5961e-08, 4.6770e-08, 6.4785e-08, 1.4028e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 217.79, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3971 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.22 lr 0.00001000 +Epoch 465, weight, value: tensor([[ 0.0335, -0.1052, -0.0326, ..., -0.1574, -0.1875, -0.1804], + [-0.1646, 0.0890, -0.1552, ..., -0.2551, -0.2201, -0.2245], + [-0.1575, -0.2085, -0.2526, ..., -0.2317, -0.1856, 0.1833], + ..., + [ 0.0919, -0.0818, -0.2175, ..., 0.1139, -0.1935, -0.1674], + [ 0.1460, 0.1079, 0.1171, ..., -0.2423, -0.1398, 0.0973], + [ 0.0853, 0.0644, 0.1394, ..., 0.0577, -0.1061, -0.0215]], + device='cuda:0'), grad: tensor([[ 3.1432e-09, 8.0618e-09, 5.2969e-09, ..., 7.2760e-10, + 5.1805e-09, 8.1200e-09], + [ 1.3300e-08, 9.9826e-09, 5.4424e-09, ..., 5.3260e-09, + 3.4343e-09, 1.0768e-08], + [ 1.2486e-08, 4.3947e-09, 1.4552e-09, ..., 1.2689e-08, + 6.3155e-09, 7.8580e-09], + ..., + [-1.5338e-08, 4.6275e-09, 2.1828e-09, ..., -1.7666e-08, + 1.9791e-09, 2.4447e-09], + [-6.8394e-08, -1.0029e-07, -6.1176e-08, ..., 3.1141e-09, + -2.5058e-08, -7.7533e-08], + [ 1.5280e-08, 1.7404e-08, 8.7020e-09, ..., 1.7753e-09, + 2.0402e-08, 2.3370e-08]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0186, -0.0300, -0.0070, 0.0139, -0.0016, -0.0164, 0.0061, 0.0159, + -0.0360, 0.0123], device='cuda:0'), grad: tensor([ 2.4476e-08, 5.5035e-08, 6.6590e-08, 1.6036e-08, -4.1997e-08, + 4.0076e-08, 1.0082e-07, -5.1805e-08, -2.8475e-07, 9.2492e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 217.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4149 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.22 lr 0.00001000 +Epoch 466, weight, value: tensor([[ 0.0336, -0.1052, -0.0327, ..., -0.1574, -0.1876, -0.1805], + [-0.1647, 0.0891, -0.1553, ..., -0.2553, -0.2201, -0.2246], + [-0.1577, -0.2086, -0.2527, ..., -0.2318, -0.1856, 0.1834], + ..., + [ 0.0919, -0.0818, -0.2177, ..., 0.1139, -0.1935, -0.1675], + [ 0.1460, 0.1079, 0.1172, ..., -0.2426, -0.1398, 0.0973], + [ 0.0853, 0.0644, 0.1396, ..., 0.0577, -0.1062, -0.0215]], + device='cuda:0'), grad: tensor([[ 4.6275e-09, 4.7439e-09, 2.0082e-09, ..., 7.3342e-09, + 1.5425e-09, 2.0082e-09], + [ 1.4639e-08, 5.5879e-09, 3.5216e-09, ..., 2.0373e-08, + 3.8708e-09, 6.8685e-09], + [ 8.4983e-09, 2.7067e-09, 1.0186e-09, ..., 1.3912e-08, + 2.6193e-09, 3.1432e-09], + ..., + [-2.5902e-08, 3.1578e-08, 1.1729e-08, ..., -5.5821e-08, + -1.1350e-08, -4.3947e-09], + [-7.6252e-09, -1.6036e-08, -1.0477e-08, ..., 1.9325e-08, + 1.5425e-09, -2.1246e-08], + [-1.2422e-07, -1.2130e-07, -7.0489e-08, ..., -1.9837e-07, + -3.2451e-08, -3.5099e-08]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0186, -0.0300, -0.0070, 0.0139, -0.0016, -0.0165, 0.0063, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 2.7096e-08, 6.8801e-08, 4.9127e-08, 3.5361e-08, 5.7137e-07, + 7.2876e-08, 1.3300e-08, -1.1758e-07, -3.8825e-08, -6.6962e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.35, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4078 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.20 lr 0.00001000 +Epoch 467, weight, value: tensor([[ 0.0336, -0.1053, -0.0328, ..., -0.1574, -0.1876, -0.1806], + [-0.1649, 0.0891, -0.1554, ..., -0.2554, -0.2202, -0.2247], + [-0.1579, -0.2086, -0.2527, ..., -0.2320, -0.1856, 0.1835], + ..., + [ 0.0919, -0.0818, -0.2178, ..., 0.1139, -0.1934, -0.1675], + [ 0.1462, 0.1081, 0.1173, ..., -0.2428, -0.1399, 0.0974], + [ 0.0853, 0.0644, 0.1398, ..., 0.0577, -0.1063, -0.0216]], + device='cuda:0'), grad: tensor([[-6.6648e-09, 6.6939e-10, 4.0745e-10, ..., 9.6043e-10, + 1.6298e-09, 1.3824e-09], + [ 1.2631e-08, -3.3178e-09, 7.4215e-10, ..., 1.0288e-08, + 1.7171e-09, 6.9413e-09], + [ 1.5716e-08, 1.3970e-09, 2.4738e-10, ..., 1.3504e-08, + 1.9500e-09, 6.7230e-09], + ..., + [-3.8592e-08, 7.6398e-09, 3.3760e-09, ..., -2.9613e-08, + 2.2701e-09, -1.5076e-08], + [ 2.3720e-09, -2.5611e-09, 1.8917e-10, ..., 1.1772e-08, + 2.4593e-09, -2.1100e-09], + [-8.8767e-10, -9.9535e-09, -9.2259e-09, ..., 4.2259e-08, + 3.9174e-08, 1.8394e-08]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0186, -0.0300, -0.0070, 0.0139, -0.0016, -0.0166, 0.0063, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([-3.0268e-08, 3.2305e-08, 7.0257e-08, 2.0678e-08, -9.8953e-08, + 3.0268e-08, -1.7462e-09, -1.2747e-07, 4.7963e-08, 8.3121e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 217.26, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4044 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 468, weight, value: tensor([[ 0.0335, -0.1055, -0.0328, ..., -0.1576, -0.1876, -0.1807], + [-0.1650, 0.0891, -0.1555, ..., -0.2556, -0.2202, -0.2248], + [-0.1580, -0.2086, -0.2528, ..., -0.2321, -0.1856, 0.1837], + ..., + [ 0.0919, -0.0819, -0.2179, ..., 0.1140, -0.1935, -0.1676], + [ 0.1462, 0.1082, 0.1174, ..., -0.2431, -0.1399, 0.0974], + [ 0.0853, 0.0644, 0.1400, ..., 0.0577, -0.1064, -0.0217]], + device='cuda:0'), grad: tensor([[ 2.0373e-10, 2.0664e-09, 7.8580e-10, ..., 1.2951e-09, + 1.2806e-09, 4.3947e-09], + [ 4.0105e-08, 7.8435e-09, 2.7503e-09, ..., 1.7099e-08, + 2.8522e-09, 2.7765e-08], + [ 4.7003e-09, 8.3965e-09, 3.0850e-09, ..., -8.1491e-10, + 3.2014e-09, -1.1467e-08], + ..., + [-3.6700e-08, 2.8813e-09, 8.4401e-10, ..., -1.3839e-08, + 8.2946e-10, -4.1327e-09], + [-2.2483e-08, -5.5705e-08, -2.0867e-08, ..., 1.0768e-09, + -1.5789e-08, -6.2049e-08], + [ 2.4884e-09, 5.6898e-09, 1.9354e-09, ..., -3.5361e-09, + 4.9768e-09, 9.9681e-09]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0185, -0.0301, -0.0069, 0.0139, -0.0016, -0.0166, 0.0064, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 1.3461e-08, 2.1688e-07, -4.9535e-08, 3.3295e-08, 1.0521e-08, + -3.4808e-08, 9.2317e-08, -1.1845e-07, -1.7066e-07, 3.3964e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 217.26, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4216 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.17 lr 0.00001000 +Epoch 469, weight, value: tensor([[ 0.0336, -0.1055, -0.0329, ..., -0.1576, -0.1877, -0.1808], + [-0.1653, 0.0892, -0.1557, ..., -0.2558, -0.2203, -0.2249], + [-0.1581, -0.2086, -0.2529, ..., -0.2322, -0.1857, 0.1839], + ..., + [ 0.0919, -0.0819, -0.2182, ..., 0.1140, -0.1935, -0.1677], + [ 0.1464, 0.1084, 0.1177, ..., -0.2435, -0.1398, 0.0976], + [ 0.0853, 0.0644, 0.1403, ..., 0.0577, -0.1065, -0.0218]], + device='cuda:0'), grad: tensor([[ 6.6939e-10, 1.1205e-09, 5.9663e-10, ..., 1.4115e-09, + 9.4587e-10, 1.0186e-09], + [ 5.0495e-09, 4.8458e-09, 2.2847e-09, ..., 7.3051e-09, + 5.3114e-09, 5.2096e-09], + [ 3.8126e-09, 1.1059e-09, 4.2201e-10, ..., 6.5920e-09, + 1.5134e-09, 1.5571e-09], + ..., + [-6.3592e-09, 8.4110e-09, 5.3406e-09, ..., -5.5879e-09, + 1.5716e-09, 2.2119e-09], + [-5.9488e-08, -1.0780e-07, -5.0291e-08, ..., 2.3429e-09, + -4.8691e-08, -7.6077e-08], + [-7.1304e-10, -7.1450e-09, -6.6066e-09, ..., 4.3656e-08, + 5.6752e-08, 4.8953e-08]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0186, -0.0302, -0.0069, 0.0139, -0.0016, -0.0166, 0.0064, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 6.0099e-09, 3.2567e-08, 1.9369e-08, 2.5451e-08, -1.7486e-07, + 1.0442e-07, 1.5029e-07, -2.0082e-09, -2.8592e-07, 1.4016e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 217.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4197 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.19 lr 0.00001000 +Epoch 470, weight, value: tensor([[ 0.0337, -0.1056, -0.0329, ..., -0.1576, -0.1877, -0.1809], + [-0.1654, 0.0892, -0.1558, ..., -0.2559, -0.2204, -0.2250], + [-0.1582, -0.2087, -0.2529, ..., -0.2323, -0.1857, 0.1840], + ..., + [ 0.0919, -0.0819, -0.2184, ..., 0.1140, -0.1935, -0.1678], + [ 0.1465, 0.1085, 0.1178, ..., -0.2437, -0.1398, 0.0977], + [ 0.0853, 0.0644, 0.1406, ..., 0.0577, -0.1066, -0.0219]], + device='cuda:0'), grad: tensor([[ 3.6380e-10, 8.1491e-10, 3.7835e-10, ..., 8.4401e-10, + 1.4668e-08, 9.6625e-09], + [ 2.3152e-08, -1.9645e-09, 1.2515e-09, ..., 5.6229e-08, + 2.6193e-09, 2.2264e-09], + [ 3.3615e-09, 7.5670e-10, 1.1642e-10, ..., 6.7375e-09, + 1.1642e-09, -2.1828e-10], + ..., + [-2.2410e-08, 3.2160e-08, 1.5818e-08, ..., -5.1397e-08, + 1.3068e-08, 8.4983e-09], + [ 6.8394e-10, 3.5943e-09, 1.3824e-09, ..., 3.7835e-09, + 1.8510e-08, 1.1409e-08], + [-1.7753e-08, -5.7131e-08, -3.1549e-08, ..., -4.4995e-08, + -2.4942e-08, -1.5134e-08]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0186, -0.0302, -0.0068, 0.0139, -0.0016, -0.0165, 0.0063, 0.0159, + -0.0361, 0.0123], device='cuda:0'), grad: tensor([ 4.1851e-08, 1.6554e-07, 2.4258e-08, 4.6450e-08, 1.0064e-07, + 4.9535e-08, -2.0396e-07, -1.1490e-07, 7.2818e-08, -1.8382e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 217.72, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4177 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.19 lr 0.00001000 +Epoch 471, weight, value: tensor([[ 0.0336, -0.1057, -0.0330, ..., -0.1578, -0.1878, -0.1810], + [-0.1659, 0.0892, -0.1559, ..., -0.2564, -0.2204, -0.2252], + [-0.1585, -0.2087, -0.2530, ..., -0.2325, -0.1857, 0.1841], + ..., + [ 0.0920, -0.0819, -0.2187, ..., 0.1140, -0.1934, -0.1679], + [ 0.1465, 0.1087, 0.1180, ..., -0.2442, -0.1399, 0.0978], + [ 0.0853, 0.0644, 0.1409, ..., 0.0577, -0.1067, -0.0220]], + device='cuda:0'), grad: tensor([[ 4.9477e-10, 3.0559e-10, 4.3656e-11, ..., 1.1642e-09, + 1.5658e-08, 1.5076e-08], + [ 1.9791e-09, 8.7311e-10, 1.7462e-10, ..., 6.5338e-09, + 7.7998e-09, 1.2282e-08], + [-1.0768e-08, -3.5070e-09, 1.3097e-10, ..., 1.0376e-08, + 1.4261e-08, -1.1752e-07], + ..., + [-3.5216e-09, 2.7212e-09, 3.2014e-10, ..., -1.3679e-09, + 4.3219e-09, 7.4651e-09], + [ 8.1782e-09, 2.7940e-09, -9.1677e-10, ..., 1.7462e-09, + 1.0885e-08, 8.8126e-08], + [ 2.0955e-09, 1.0041e-09, 3.9290e-10, ..., 1.8641e-08, + 1.7026e-08, 1.7535e-08]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0185, -0.0304, -0.0068, 0.0138, -0.0016, -0.0166, 0.0063, 0.0160, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 8.4634e-08, 6.2399e-08, -3.4622e-07, 1.6822e-08, -1.0896e-07, + 6.6264e-07, -7.9395e-07, 2.6892e-08, 3.2387e-07, 9.0804e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 217.52, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4429 re_mapping 0.0015 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 472, weight, value: tensor([[ 0.0337, -0.1058, -0.0330, ..., -0.1578, -0.1878, -0.1811], + [-0.1660, 0.0892, -0.1560, ..., -0.2565, -0.2205, -0.2252], + [-0.1586, -0.2087, -0.2530, ..., -0.2327, -0.1857, 0.1843], + ..., + [ 0.0920, -0.0819, -0.2188, ..., 0.1141, -0.1934, -0.1680], + [ 0.1465, 0.1088, 0.1181, ..., -0.2444, -0.1399, 0.0979], + [ 0.0853, 0.0645, 0.1410, ..., 0.0577, -0.1068, -0.0221]], + device='cuda:0'), grad: tensor([[-3.7835e-10, 7.2760e-10, 4.6566e-10, ..., 1.7608e-09, + 6.9849e-09, 1.4057e-08], + [ 5.7335e-09, 3.9290e-10, 4.5111e-10, ..., 8.2073e-09, + 3.3469e-09, 1.7099e-08], + [ 1.4843e-09, 2.6193e-10, 7.2760e-11, ..., 1.8335e-09, + 7.6107e-09, -4.5984e-08], + ..., + [-1.8190e-09, 1.2122e-08, 6.6648e-09, ..., 3.6234e-09, + 8.2946e-10, 1.9572e-08], + [ 3.1869e-09, 1.7317e-09, 1.3097e-09, ..., 4.8458e-09, + 2.0678e-08, 4.6624e-08], + [-2.7227e-08, -2.7896e-08, -1.8408e-08, ..., -4.9768e-08, + 2.0227e-09, 8.2946e-10]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0185, -0.0305, -0.0068, 0.0138, -0.0016, -0.0166, 0.0064, 0.0160, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 3.1287e-08, 8.5507e-08, -2.4564e-07, 6.1118e-08, 1.1199e-07, + 1.4406e-08, -1.5809e-07, 1.0798e-07, 1.2992e-07, -1.3167e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 217.40, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3934 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.18 lr 0.00001000 +Epoch 473, weight, value: tensor([[ 0.0337, -0.1058, -0.0331, ..., -0.1578, -0.1878, -0.1812], + [-0.1661, 0.0893, -0.1562, ..., -0.2566, -0.2205, -0.2254], + [-0.1588, -0.2088, -0.2531, ..., -0.2329, -0.1858, 0.1844], + ..., + [ 0.0921, -0.0819, -0.2189, ..., 0.1141, -0.1934, -0.1681], + [ 0.1465, 0.1090, 0.1184, ..., -0.2447, -0.1399, 0.0980], + [ 0.0853, 0.0645, 0.1411, ..., 0.0577, -0.1070, -0.0224]], + device='cuda:0'), grad: tensor([[ 4.3074e-09, 1.0274e-08, 6.0390e-09, ..., 8.0036e-09, + 4.0600e-09, 1.1933e-09], + [ 5.5443e-09, -5.9110e-08, 1.3533e-09, ..., 6.2282e-09, + 1.4261e-09, 1.2951e-09], + [ 4.4092e-09, 2.1391e-09, 2.6193e-10, ..., 5.3551e-09, + 3.3469e-10, 2.9104e-10], + ..., + [-1.6982e-08, 4.1327e-08, 7.2469e-09, ..., -1.5469e-08, + 1.0477e-09, 6.5484e-10], + [-1.5134e-09, 1.2267e-08, 2.5466e-09, ..., 6.8831e-09, + 6.7375e-09, -1.9791e-09], + [-9.3423e-09, -2.4622e-08, -2.7838e-08, ..., -2.5408e-08, + -2.8085e-09, 1.0768e-09]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0186, -0.0306, -0.0068, 0.0138, -0.0016, -0.0167, 0.0065, 0.0160, + -0.0362, 0.0123], device='cuda:0'), grad: tensor([ 4.8574e-08, -2.3423e-07, 2.5844e-08, 2.5102e-08, 3.4575e-08, + 9.7905e-08, -9.2608e-08, 9.2143e-08, 6.0303e-08, -5.1572e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 217.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4264 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.17 lr 0.00001000 +Epoch 474, weight, value: tensor([[ 0.0338, -0.1059, -0.0331, ..., -0.1578, -0.1879, -0.1813], + [-0.1663, 0.0893, -0.1563, ..., -0.2567, -0.2206, -0.2255], + [-0.1592, -0.2088, -0.2532, ..., -0.2331, -0.1858, 0.1845], + ..., + [ 0.0921, -0.0819, -0.2191, ..., 0.1141, -0.1934, -0.1680], + [ 0.1466, 0.1091, 0.1185, ..., -0.2449, -0.1399, 0.0981], + [ 0.0853, 0.0645, 0.1413, ..., 0.0577, -0.1071, -0.0226]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 5.0495e-09, 5.8208e-11, ..., 2.1828e-10, + 1.0186e-09, 9.8953e-10], + [ 3.8126e-09, -9.4529e-08, 1.1642e-10, ..., 4.6857e-09, + 3.4925e-10, 3.4197e-09], + [ 1.0041e-09, 7.0431e-09, 1.4552e-11, ..., 1.6735e-09, + 9.1677e-10, -4.0745e-09], + ..., + [-2.0242e-08, 6.0478e-08, 3.4925e-10, ..., -2.5320e-08, + 5.8208e-11, 1.9209e-09], + [ 6.2573e-10, 1.8277e-08, 1.0186e-10, ..., 8.4401e-10, + 9.6043e-10, 1.2369e-09], + [ 1.1918e-08, 3.7835e-09, -1.2224e-09, ..., 1.4785e-08, + 9.1677e-10, 6.4028e-10]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0186, -0.0306, -0.0068, 0.0138, -0.0016, -0.0167, 0.0065, 0.0160, + -0.0362, 0.0122], device='cuda:0'), grad: tensor([ 4.4878e-08, -7.4459e-07, 1.2908e-08, 2.6819e-08, 1.3461e-08, + -7.0722e-09, -2.4593e-09, 4.6706e-07, 1.6321e-07, 6.4843e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 217.57, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3949 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.18 lr 0.00001000 +Epoch 475, weight, value: tensor([[ 0.0339, -0.1059, -0.0331, ..., -0.1578, -0.1879, -0.1814], + [-0.1668, 0.0894, -0.1565, ..., -0.2572, -0.2206, -0.2256], + [-0.1594, -0.2088, -0.2532, ..., -0.2333, -0.1858, 0.1846], + ..., + [ 0.0922, -0.0819, -0.2193, ..., 0.1142, -0.1933, -0.1681], + [ 0.1468, 0.1093, 0.1188, ..., -0.2450, -0.1399, 0.0983], + [ 0.0853, 0.0645, 0.1415, ..., 0.0577, -0.1073, -0.0227]], + device='cuda:0'), grad: tensor([[-4.6566e-10, 2.6193e-10, 1.6007e-10, ..., 6.9849e-10, + -2.1828e-10, 2.7649e-10], + [ 1.8335e-08, -6.7812e-09, 1.3097e-10, ..., 2.5218e-08, + 1.4697e-09, 1.1205e-09], + [ 1.6153e-08, 3.0122e-09, 1.4552e-11, ..., 1.7666e-08, + 2.4884e-09, 1.5425e-09], + ..., + [-6.3330e-08, 2.7503e-09, 3.9290e-10, ..., -7.3866e-08, + -1.5862e-09, -1.5134e-09], + [ 5.1368e-09, 2.1973e-09, 1.2515e-09, ..., 5.7189e-09, + 9.8953e-10, 5.9663e-10], + [ 9.7207e-09, -2.3429e-09, -2.4011e-09, ..., 3.8010e-08, + 2.1901e-08, 1.1496e-08]], device='cuda:0') +Epoch 475, bias, value: tensor([ 0.0187, -0.0309, -0.0068, 0.0138, -0.0016, -0.0167, 0.0065, 0.0161, + -0.0362, 0.0122], device='cuda:0'), grad: tensor([-1.1656e-08, 7.2003e-08, 9.9884e-08, 4.8371e-08, -3.5885e-08, + 7.0286e-09, 7.8580e-09, -3.1851e-07, 3.4517e-08, 1.0815e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 217.49, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4422 re_mapping 0.0015 re_causal 0.0068 /// teacc 99.17 lr 0.00001000 +Epoch 476, weight, value: tensor([[ 0.0340, -0.1060, -0.0332, ..., -0.1578, -0.1879, -0.1814], + [-0.1668, 0.0895, -0.1566, ..., -0.2572, -0.2207, -0.2257], + [-0.1596, -0.2089, -0.2533, ..., -0.2335, -0.1859, 0.1847], + ..., + [ 0.0922, -0.0819, -0.2194, ..., 0.1142, -0.1933, -0.1681], + [ 0.1469, 0.1093, 0.1189, ..., -0.2452, -0.1399, 0.0983], + [ 0.0853, 0.0645, 0.1416, ..., 0.0577, -0.1074, -0.0228]], + device='cuda:0'), grad: tensor([[ 1.6007e-10, 1.4552e-10, 2.9104e-11, ..., 3.3469e-10, + 7.4215e-10, 1.1933e-09], + [ 3.1287e-09, 3.6380e-10, 5.5297e-10, ..., 3.0559e-09, + 1.2660e-09, 8.0618e-09], + [ 2.3574e-09, 8.0036e-10, 1.4552e-10, ..., 2.5902e-09, + 7.7125e-10, -3.5478e-08], + ..., + [-6.1700e-09, 2.1682e-09, 4.5111e-10, ..., -5.3551e-09, + 1.0623e-09, 1.9107e-08], + [-3.9145e-09, -4.7294e-09, -2.5029e-09, ..., 8.1491e-10, + 3.4051e-09, 2.3429e-09], + [ 1.9209e-09, 2.7649e-10, 5.8208e-11, ..., 2.9191e-08, + 2.7547e-08, 1.4072e-08]], device='cuda:0') +Epoch 476, bias, value: tensor([ 0.0187, -0.0309, -0.0069, 0.0138, -0.0016, -0.0167, 0.0065, 0.0161, + -0.0362, 0.0122], device='cuda:0'), grad: tensor([ 5.5152e-09, 4.6857e-08, -1.7288e-07, 3.3120e-08, -8.1433e-08, + 1.3752e-08, -3.9872e-09, 7.3691e-08, 9.9826e-09, 8.7661e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 217.48, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4179 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.19 lr 0.00001000 +Epoch 477, weight, value: tensor([[ 0.0340, -0.1061, -0.0333, ..., -0.1578, -0.1879, -0.1815], + [-0.1669, 0.0895, -0.1567, ..., -0.2573, -0.2207, -0.2258], + [-0.1597, -0.2089, -0.2533, ..., -0.2337, -0.1859, 0.1849], + ..., + [ 0.0922, -0.0819, -0.2196, ..., 0.1143, -0.1933, -0.1682], + [ 0.1470, 0.1095, 0.1191, ..., -0.2452, -0.1399, 0.0984], + [ 0.0853, 0.0645, 0.1418, ..., 0.0577, -0.1075, -0.0229]], + device='cuda:0'), grad: tensor([[ 1.3679e-09, 1.9354e-09, 5.2387e-10, ..., 1.0186e-10, + 1.4406e-09, 2.4011e-09], + [ 7.2469e-09, -2.1522e-08, 2.1828e-09, ..., 4.8021e-10, + 5.0932e-10, 7.6252e-09], + [ 6.0681e-09, 7.3924e-09, 1.5716e-09, ..., -4.3656e-11, + 3.6380e-10, 2.1246e-09], + ..., + [ 1.1059e-09, 2.5349e-08, 1.1642e-09, ..., -2.4447e-09, + 1.3097e-10, 4.1327e-09], + [-3.3877e-08, -2.4360e-08, -1.1089e-08, ..., 1.3242e-09, + -4.3656e-11, -3.1607e-08], + [ 4.5693e-09, 5.8935e-09, 1.4697e-09, ..., 7.2760e-10, + 1.8917e-10, 3.9290e-09]], device='cuda:0') +Epoch 477, bias, value: tensor([ 0.0187, -0.0310, -0.0069, 0.0138, -0.0016, -0.0166, 0.0065, 0.0161, + -0.0363, 0.0122], device='cuda:0'), grad: tensor([ 1.3606e-08, -8.1083e-08, 2.6310e-08, 4.1444e-08, 5.8062e-09, + -1.0361e-07, 5.1863e-08, 1.0582e-07, -6.4785e-08, 3.3644e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 217.29, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4273 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.19 lr 0.00001000 +Epoch 478, weight, value: tensor([[ 0.0341, -0.1062, -0.0334, ..., -0.1578, -0.1880, -0.1816], + [-0.1670, 0.0896, -0.1568, ..., -0.2573, -0.2208, -0.2259], + [-0.1599, -0.2090, -0.2534, ..., -0.2338, -0.1859, 0.1849], + ..., + [ 0.0923, -0.0820, -0.2198, ..., 0.1143, -0.1933, -0.1682], + [ 0.1472, 0.1097, 0.1193, ..., -0.2454, -0.1399, 0.0987], + [ 0.0853, 0.0645, 0.1421, ..., 0.0577, -0.1076, -0.0230]], + device='cuda:0'), grad: tensor([[ 3.6380e-10, 1.6007e-10, 2.9104e-11, ..., 4.0745e-10, + 1.5425e-09, 1.1350e-09], + [ 3.8068e-08, -1.2224e-09, 8.7311e-11, ..., 3.8999e-08, + 3.9290e-10, 2.8667e-09], + [ 1.0623e-08, 7.2760e-10, 0.0000e+00, ..., 1.5352e-08, + 2.6193e-10, 3.6525e-09], + ..., + [-1.3970e-07, -2.1246e-09, 3.0559e-10, ..., -1.5053e-07, + 2.9104e-10, -1.1089e-08], + [ 7.3633e-09, 1.4552e-10, -2.4447e-09, ..., 1.6065e-08, + 1.7317e-09, -4.3219e-09], + [ 7.1246e-08, 2.7212e-09, 5.8208e-11, ..., 7.4389e-08, + 1.2515e-09, 4.3656e-09]], device='cuda:0') +Epoch 478, bias, value: tensor([ 0.0187, -0.0310, -0.0069, 0.0137, -0.0017, -0.0165, 0.0064, 0.0161, + -0.0362, 0.0122], device='cuda:0'), grad: tensor([ 7.0286e-09, 1.2980e-07, 5.0233e-08, 5.1572e-08, 1.4203e-08, + -1.4808e-07, 4.1182e-08, -5.0711e-07, 9.3831e-08, 2.7427e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 217.58, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4174 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.21 lr 0.00001000 +Epoch 479, weight, value: tensor([[ 0.0341, -0.1062, -0.0335, ..., -0.1579, -0.1880, -0.1818], + [-0.1670, 0.0897, -0.1570, ..., -0.2573, -0.2209, -0.2260], + [-0.1601, -0.2090, -0.2535, ..., -0.2340, -0.1860, 0.1851], + ..., + [ 0.0923, -0.0820, -0.2200, ..., 0.1143, -0.1933, -0.1683], + [ 0.1474, 0.1099, 0.1195, ..., -0.2455, -0.1399, 0.0988], + [ 0.0853, 0.0645, 0.1423, ..., 0.0577, -0.1076, -0.0231]], + device='cuda:0'), grad: tensor([[ 4.7148e-09, 7.4215e-10, 6.2573e-10, ..., 8.7311e-10, + 1.0332e-09, 1.4639e-08], + [ 1.2020e-08, 2.4302e-09, 1.6153e-09, ..., 4.5984e-09, + 5.0932e-10, 1.7259e-08], + [-1.1886e-07, 1.4843e-09, 6.1118e-10, ..., 4.8312e-09, + -6.2573e-10, -2.4075e-07], + ..., + [-3.6380e-09, 4.2492e-09, 3.0559e-09, ..., -1.6997e-08, + 3.2014e-10, 1.9048e-08], + [ 8.8243e-08, -1.2515e-08, -4.1036e-09, ..., 5.3842e-09, + 1.4115e-09, 1.7066e-07], + [-1.8044e-09, -1.5905e-08, -1.8554e-08, ..., -2.5102e-08, + -5.2823e-09, -2.3720e-09]], device='cuda:0') +Epoch 479, bias, value: tensor([ 0.0187, -0.0310, -0.0069, 0.0137, -0.0018, -0.0166, 0.0065, 0.0161, + -0.0362, 0.0122], device='cuda:0'), grad: tensor([ 5.3638e-08, 8.5332e-08, -1.0822e-06, 2.1464e-08, 8.8243e-08, + 1.7462e-08, 2.2061e-08, 3.5623e-08, 8.2282e-07, -5.7102e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 217.82, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4225 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.21 lr 0.00001000 +Epoch 480, weight, value: tensor([[ 0.0341, -0.1062, -0.0335, ..., -0.1579, -0.1881, -0.1819], + [-0.1671, 0.0897, -0.1571, ..., -0.2574, -0.2209, -0.2262], + [-0.1602, -0.2091, -0.2536, ..., -0.2343, -0.1860, 0.1853], + ..., + [ 0.0923, -0.0820, -0.2203, ..., 0.1143, -0.1933, -0.1683], + [ 0.1476, 0.1102, 0.1198, ..., -0.2457, -0.1398, 0.0990], + [ 0.0853, 0.0645, 0.1425, ..., 0.0577, -0.1077, -0.0232]], + device='cuda:0'), grad: tensor([[ 3.8708e-09, 3.2160e-09, 1.2369e-09, ..., 4.0891e-09, + 1.3533e-09, 6.2282e-09], + [ 2.1697e-08, 5.1165e-08, 6.8394e-09, ..., 2.9861e-08, + 5.8353e-09, 2.5248e-08], + [-6.3010e-09, 1.4959e-08, 1.9936e-09, ..., -2.9977e-08, + 1.8772e-09, -6.3679e-08], + ..., + [-1.3883e-08, 5.7713e-08, 1.1132e-08, ..., -3.7544e-08, + -9.4587e-09, 3.1083e-08], + [-3.3935e-08, -2.3152e-08, -1.8568e-08, ..., 1.1220e-08, + 4.8603e-09, -5.7800e-08], + [-1.8961e-08, -2.5728e-08, -2.8114e-08, ..., -1.0623e-08, + 3.4604e-08, 1.8685e-08]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0187, -0.0310, -0.0069, 0.0137, -0.0018, -0.0165, 0.0064, 0.0162, + -0.0361, 0.0122], device='cuda:0'), grad: tensor([ 4.4121e-08, 2.8335e-07, -2.5262e-07, -4.0466e-07, 1.0745e-07, + 2.1362e-07, 6.7288e-08, 8.0501e-08, -1.1607e-07, 3.4051e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 217.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4037 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.20 lr 0.00001000 +Epoch 481, weight, value: tensor([[ 0.0340, -0.1064, -0.0336, ..., -0.1581, -0.1882, -0.1821], + [-0.1672, 0.0899, -0.1573, ..., -0.2575, -0.2210, -0.2263], + [-0.1603, -0.2091, -0.2537, ..., -0.2344, -0.1860, 0.1855], + ..., + [ 0.0923, -0.0820, -0.2205, ..., 0.1143, -0.1934, -0.1686], + [ 0.1478, 0.1104, 0.1200, ..., -0.2458, -0.1398, 0.0992], + [ 0.0853, 0.0645, 0.1428, ..., 0.0577, -0.1079, -0.0234]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 5.9081e-09, 1.4552e-11, ..., 1.6007e-10, + 1.3824e-09, 9.6043e-10], + [ 6.1118e-10, -1.1537e-07, 1.4552e-11, ..., 1.3533e-09, + 4.9477e-10, 3.2014e-10], + [ 3.3469e-10, 1.4741e-08, 0.0000e+00, ..., 7.8580e-10, + 3.6380e-10, 2.1828e-10], + ..., + [-4.6566e-10, 3.3120e-08, 1.3097e-10, ..., -6.9849e-10, + 3.0559e-10, 1.6007e-10], + [ 1.6007e-10, 2.8813e-09, 4.3656e-11, ..., 1.8917e-10, + 1.1059e-09, 8.5856e-10], + [ 3.6380e-10, 1.8044e-08, 0.0000e+00, ..., 5.6607e-09, + 4.9185e-09, 2.6193e-09]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0186, -0.0311, -0.0068, 0.0137, -0.0019, -0.0164, 0.0063, 0.0162, + -0.0361, 0.0122], device='cuda:0'), grad: tensor([ 3.7893e-08, -6.5984e-07, 8.6962e-08, 1.8394e-07, 1.2777e-08, + 1.4843e-08, 1.7695e-08, 1.8731e-07, 2.2002e-08, 1.1828e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 217.44, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3759 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.19 lr 0.00001000 +Epoch 482, weight, value: tensor([[ 0.0340, -0.1064, -0.0337, ..., -0.1581, -0.1883, -0.1822], + [-0.1673, 0.0899, -0.1574, ..., -0.2575, -0.2210, -0.2265], + [-0.1605, -0.2092, -0.2537, ..., -0.2345, -0.1860, 0.1856], + ..., + [ 0.0923, -0.0821, -0.2208, ..., 0.1144, -0.1934, -0.1687], + [ 0.1481, 0.1107, 0.1203, ..., -0.2459, -0.1397, 0.0995], + [ 0.0853, 0.0645, 0.1429, ..., 0.0577, -0.1080, -0.0235]], + device='cuda:0'), grad: tensor([[-1.6444e-09, 1.5862e-09, -9.8953e-10, ..., 2.9104e-10, + 3.4488e-09, 2.2847e-09], + [ 4.5839e-09, -8.3674e-09, 4.9477e-10, ..., 4.2201e-09, + 1.4552e-09, 5.4570e-09], + [ 2.2992e-09, 8.7311e-10, 1.3097e-10, ..., 2.5320e-09, + 2.1537e-09, -8.1636e-09], + ..., + [-1.1147e-08, 6.7375e-09, 1.4843e-09, ..., -1.1700e-08, + 2.1828e-10, 1.5862e-09], + [-1.1496e-09, -3.9145e-09, -9.1677e-10, ..., 3.7544e-09, + 5.5006e-09, -7.2760e-11], + [ 3.3324e-09, -2.6193e-09, -3.0559e-09, ..., -1.3679e-09, + 4.5111e-10, 8.7311e-10]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0187, -0.0311, -0.0068, 0.0137, -0.0019, -0.0165, 0.0063, 0.0162, + -0.0360, 0.0122], device='cuda:0'), grad: tensor([ 5.7480e-09, 1.5905e-08, -4.8400e-08, 1.1106e-07, 2.5902e-08, + -2.6473e-07, 3.8475e-08, -2.1100e-09, 1.1793e-07, 1.4625e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 217.47, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4220 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +Epoch 483, weight, value: tensor([[ 0.0341, -0.1065, -0.0337, ..., -0.1581, -0.1884, -0.1823], + [-0.1673, 0.0901, -0.1576, ..., -0.2576, -0.2211, -0.2266], + [-0.1606, -0.2092, -0.2538, ..., -0.2346, -0.1861, 0.1858], + ..., + [ 0.0923, -0.0821, -0.2209, ..., 0.1144, -0.1934, -0.1689], + [ 0.1483, 0.1109, 0.1205, ..., -0.2461, -0.1397, 0.0996], + [ 0.0853, 0.0645, 0.1430, ..., 0.0577, -0.1083, -0.0238]], + device='cuda:0'), grad: tensor([[ 1.3097e-10, 6.6939e-10, 0.0000e+00, ..., 5.3842e-10, + 2.6048e-09, 2.4738e-09], + [ 2.9831e-09, -4.9185e-08, 1.6007e-10, ..., 5.5879e-09, + 3.3178e-09, -2.7925e-08], + [ 2.0809e-09, 7.5088e-09, 1.4552e-11, ..., 3.7835e-09, + 7.1304e-10, -7.5495e-08], + ..., + [-2.9802e-08, 4.9185e-09, 2.3283e-10, ..., -4.1240e-08, + 4.6566e-10, 7.2061e-08], + [ 2.2410e-09, 7.5670e-09, 1.4552e-11, ..., 3.0414e-09, + 5.8935e-09, 1.2966e-08], + [ 2.0809e-08, 2.9162e-08, 1.4552e-11, ..., 1.5949e-07, + 1.2980e-07, 5.4366e-08]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0187, -0.0310, -0.0067, 0.0136, -0.0018, -0.0163, 0.0062, 0.0162, + -0.0360, 0.0122], device='cuda:0'), grad: tensor([ 1.1103e-08, -1.4226e-07, -2.0594e-07, 2.5349e-08, -2.0931e-07, + 8.7020e-09, 5.6665e-08, 9.7963e-08, 6.6531e-08, 3.0594e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 217.80, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4367 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +Epoch 484, weight, value: tensor([[ 0.0341, -0.1065, -0.0338, ..., -0.1581, -0.1884, -0.1825], + [-0.1674, 0.0903, -0.1577, ..., -0.2576, -0.2212, -0.2267], + [-0.1607, -0.2093, -0.2539, ..., -0.2347, -0.1861, 0.1860], + ..., + [ 0.0924, -0.0821, -0.2210, ..., 0.1144, -0.1934, -0.1690], + [ 0.1484, 0.1110, 0.1207, ..., -0.2462, -0.1397, 0.0997], + [ 0.0853, 0.0645, 0.1431, ..., 0.0577, -0.1086, -0.0241]], + device='cuda:0'), grad: tensor([[ 7.8580e-10, 1.2806e-09, 8.5856e-10, ..., 1.6298e-09, + 8.6729e-09, 4.0745e-09], + [ 7.6107e-09, 5.7480e-09, 4.1327e-09, ..., 4.7294e-09, + 1.9500e-09, 7.9308e-09], + [ 6.3446e-09, 6.7521e-09, 2.9249e-09, ..., 5.2823e-09, + 1.8190e-09, 8.7166e-09], + ..., + [-5.2969e-09, 9.5170e-09, 5.0059e-09, ..., -7.7271e-09, + 1.3679e-09, 6.6211e-09], + [-3.0501e-08, -3.1752e-08, -1.8146e-08, ..., 8.7748e-09, + 1.3883e-08, -2.5917e-08], + [ 3.0559e-10, -1.2311e-08, -1.0143e-08, ..., 8.7311e-10, + 6.6357e-09, 6.1846e-09]], device='cuda:0') +Epoch 484, bias, value: tensor([ 0.0187, -0.0310, -0.0067, 0.0136, -0.0017, -0.0163, 0.0062, 0.0162, + -0.0360, 0.0121], device='cuda:0'), grad: tensor([ 2.6150e-08, 3.7253e-08, 4.6653e-08, 1.4494e-08, -2.3632e-08, + 9.1386e-08, -5.5443e-08, 8.6147e-09, -1.1863e-07, -4.1764e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 217.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4136 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.16 lr 0.00001000 +Epoch 485, weight, value: tensor([[ 0.0341, -0.1066, -0.0338, ..., -0.1582, -0.1885, -0.1826], + [-0.1676, 0.0904, -0.1578, ..., -0.2578, -0.2212, -0.2269], + [-0.1609, -0.2094, -0.2539, ..., -0.2349, -0.1861, 0.1861], + ..., + [ 0.0924, -0.0821, -0.2211, ..., 0.1145, -0.1934, -0.1690], + [ 0.1485, 0.1112, 0.1209, ..., -0.2464, -0.1397, 0.0998], + [ 0.0853, 0.0646, 0.1432, ..., 0.0577, -0.1088, -0.0242]], + device='cuda:0'), grad: tensor([[-1.9951e-08, 1.6589e-09, 2.7649e-10, ..., -2.3574e-09, + 3.8854e-09, 5.3842e-09], + [ 4.5111e-09, -2.7794e-09, 8.2946e-10, ..., 3.0122e-09, + 5.6752e-09, 7.0868e-09], + [ 3.5798e-09, 1.9645e-09, 1.3097e-10, ..., 2.5175e-09, + 1.7899e-09, -1.9005e-08], + ..., + [-2.4011e-09, 3.2305e-09, 6.1118e-10, ..., -5.3260e-09, + 1.7317e-09, 7.8435e-09], + [ 0.0000e+00, -1.6735e-09, -2.8085e-09, ..., 2.9686e-09, + 3.8708e-08, 1.8437e-08], + [ 5.9663e-09, -6.7812e-09, -3.9290e-09, ..., -1.8190e-09, + 1.4261e-09, -3.0414e-09]], device='cuda:0') +Epoch 485, bias, value: tensor([ 0.0186, -0.0311, -0.0068, 0.0136, -0.0016, -0.0161, 0.0060, 0.0162, + -0.0360, 0.0121], device='cuda:0'), grad: tensor([-6.7987e-08, 5.2969e-08, -6.3330e-08, 3.5740e-08, 4.1502e-08, + -1.3150e-06, 1.0319e-06, 4.1968e-08, 2.2317e-07, 2.4767e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 217.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4207 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.16 lr 0.00001000 +Epoch 486, weight, value: tensor([[ 0.0341, -0.1067, -0.0338, ..., -0.1583, -0.1886, -0.1827], + [-0.1678, 0.0905, -0.1579, ..., -0.2580, -0.2213, -0.2270], + [-0.1611, -0.2095, -0.2540, ..., -0.2351, -0.1862, 0.1862], + ..., + [ 0.0925, -0.0822, -0.2212, ..., 0.1145, -0.1933, -0.1690], + [ 0.1485, 0.1112, 0.1210, ..., -0.2467, -0.1398, 0.0999], + [ 0.0853, 0.0646, 0.1433, ..., 0.0577, -0.1090, -0.0245]], + device='cuda:0'), grad: tensor([[ 2.0373e-10, 4.3656e-10, 1.4552e-10, ..., 2.6193e-10, + 7.3342e-09, 6.2719e-09], + [ 4.0745e-09, 1.1496e-09, 1.3533e-09, ..., 4.1182e-09, + 4.8167e-09, 8.9494e-09], + [ 6.0681e-09, 8.0327e-09, 2.1973e-09, ..., 6.4611e-09, + 1.8044e-09, 1.1001e-08], + ..., + [-7.9890e-09, 7.6252e-09, 1.9936e-09, ..., -1.5410e-08, + 3.4925e-10, 4.1618e-09], + [-9.5606e-09, -2.2090e-08, -7.8144e-09, ..., 1.3242e-09, + 2.8231e-08, -9.2841e-09], + [ 6.1118e-10, -1.7317e-09, -1.6007e-09, ..., 4.9768e-09, + 4.2346e-09, 4.3074e-09]], device='cuda:0') +Epoch 486, bias, value: tensor([ 0.0187, -0.0311, -0.0068, 0.0136, -0.0015, -0.0161, 0.0059, 0.0162, + -0.0361, 0.0121], device='cuda:0'), grad: tensor([ 2.3632e-08, 3.1519e-08, 6.5833e-08, 8.6147e-08, 1.2224e-08, + 5.2387e-10, -2.2398e-07, -1.8772e-09, 2.4884e-09, 2.5044e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 217.87, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4156 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +Epoch 487, weight, value: tensor([[ 0.0342, -0.1067, -0.0338, ..., -0.1583, -0.1886, -0.1828], + [-0.1678, 0.0906, -0.1580, ..., -0.2580, -0.2213, -0.2271], + [-0.1613, -0.2095, -0.2540, ..., -0.2354, -0.1862, 0.1863], + ..., + [ 0.0925, -0.0822, -0.2214, ..., 0.1146, -0.1933, -0.1691], + [ 0.1485, 0.1113, 0.1211, ..., -0.2468, -0.1398, 0.1000], + [ 0.0853, 0.0646, 0.1434, ..., 0.0576, -0.1092, -0.0247]], + device='cuda:0'), grad: tensor([[ 4.5111e-10, 2.0009e-09, 0.0000e+00, ..., 6.9849e-10, + 2.1828e-10, 1.5280e-10], + [ 1.5047e-08, -3.5710e-08, 2.9104e-11, ..., 8.4692e-09, + 6.6939e-10, -1.0914e-09], + [ 2.9191e-08, 2.6994e-09, 1.4552e-11, ..., 2.9890e-08, + 5.6752e-10, 4.6566e-10], + ..., + [-1.0675e-07, 1.4530e-08, 2.9104e-11, ..., -9.8487e-08, + 6.9122e-10, 1.3824e-09], + [ 4.2492e-09, 1.5352e-09, 7.2760e-12, ..., 4.1473e-09, + 1.0914e-10, 4.6566e-10], + [ 5.2678e-08, 1.7390e-09, 7.2760e-12, ..., 6.7696e-08, + 1.4377e-08, 6.8030e-09]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0187, -0.0311, -0.0069, 0.0135, -0.0015, -0.0158, 0.0059, 0.0162, + -0.0362, 0.0121], device='cuda:0'), grad: tensor([ 1.0041e-08, -6.1002e-08, 1.1775e-07, 1.5862e-08, -9.6697e-09, + 5.5516e-09, 1.0506e-08, -3.1455e-07, 2.3370e-08, 2.1572e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 217.63, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4295 re_mapping 0.0015 re_causal 0.0068 /// teacc 99.16 lr 0.00001000 +Epoch 488, weight, value: tensor([[ 0.0342, -0.1068, -0.0339, ..., -0.1583, -0.1887, -0.1829], + [-0.1679, 0.0908, -0.1580, ..., -0.2581, -0.2214, -0.2272], + [-0.1615, -0.2096, -0.2541, ..., -0.2355, -0.1863, 0.1864], + ..., + [ 0.0926, -0.0822, -0.2215, ..., 0.1146, -0.1933, -0.1691], + [ 0.1487, 0.1114, 0.1212, ..., -0.2470, -0.1399, 0.1001], + [ 0.0853, 0.0646, 0.1436, ..., 0.0576, -0.1094, -0.0249]], + device='cuda:0'), grad: tensor([[-1.0965e-08, 1.2951e-09, 2.1828e-10, ..., 1.0914e-10, + -3.6438e-08, 5.6025e-10], + [ 6.3010e-09, -1.1256e-08, 9.7498e-10, ..., 1.0390e-08, + 2.2119e-09, 3.8272e-09], + [ 8.9785e-09, 2.7067e-09, 2.3283e-10, ..., 1.7753e-08, + 2.8813e-09, 4.1036e-09], + ..., + [-2.1886e-08, 8.6366e-09, 9.6043e-10, ..., -4.5169e-08, + -6.7739e-09, -8.8039e-09], + [-4.4311e-09, 2.1915e-08, 2.7940e-09, ..., 2.2119e-09, + 1.5789e-09, -4.7148e-09], + [ 4.9986e-09, 4.7366e-09, 2.3283e-09, ..., 5.2096e-09, + 2.2119e-09, 4.7876e-09]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0187, -0.0311, -0.0069, 0.0135, -0.0015, -0.0158, 0.0059, 0.0162, + -0.0363, 0.0121], device='cuda:0'), grad: tensor([-5.5647e-07, -9.2623e-09, 6.6299e-08, -4.3074e-08, 1.2384e-08, + 3.1985e-08, 5.2201e-07, -1.0652e-07, 4.5169e-08, 4.4820e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 217.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4116 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.23 lr 0.00001000 +Epoch 489, weight, value: tensor([[ 0.0344, -0.1068, -0.0339, ..., -0.1583, -0.1887, -0.1830], + [-0.1680, 0.0910, -0.1582, ..., -0.2581, -0.2215, -0.2274], + [-0.1617, -0.2097, -0.2541, ..., -0.2357, -0.1863, 0.1866], + ..., + [ 0.0926, -0.0823, -0.2216, ..., 0.1146, -0.1932, -0.1691], + [ 0.1488, 0.1115, 0.1214, ..., -0.2471, -0.1400, 0.1001], + [ 0.0852, 0.0646, 0.1437, ..., 0.0576, -0.1096, -0.0251]], + device='cuda:0'), grad: tensor([[-1.7186e-08, 4.5111e-10, 4.3656e-11, ..., 1.8917e-10, + 8.2364e-09, 3.2160e-09], + [ 5.7626e-09, -2.9744e-08, 8.7311e-11, ..., 5.1077e-09, + 2.9686e-09, -1.2107e-08], + [ 5.3697e-09, 4.9622e-09, 1.4552e-11, ..., 4.7876e-09, + 9.4587e-10, 3.8854e-09], + ..., + [-1.4203e-08, 1.8044e-08, 9.0222e-10, ..., -1.6778e-08, + 2.1828e-10, 6.9413e-09], + [ 3.6525e-09, 1.4552e-09, 2.4738e-10, ..., 1.8335e-09, + 1.6560e-08, 1.0274e-08], + [ 4.9331e-09, -1.5716e-09, -2.1391e-09, ..., 3.4488e-09, + 2.2847e-09, 1.6298e-09]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0188, -0.0311, -0.0070, 0.0135, -0.0015, -0.0156, 0.0060, 0.0162, + -0.0364, 0.0120], device='cuda:0'), grad: tensor([-6.3912e-08, -9.2783e-08, 4.0105e-08, 2.5801e-08, 1.2398e-08, + 6.4261e-08, -1.1723e-07, 2.5320e-08, 6.7928e-08, 3.2713e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 217.72, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3826 re_mapping 0.0014 re_causal 0.0062 /// teacc 99.22 lr 0.00001000 +Epoch 490, weight, value: tensor([[ 0.0345, -0.1069, -0.0339, ..., -0.1583, -0.1888, -0.1832], + [-0.1681, 0.0912, -0.1583, ..., -0.2583, -0.2215, -0.2275], + [-0.1619, -0.2098, -0.2542, ..., -0.2359, -0.1864, 0.1867], + ..., + [ 0.0926, -0.0823, -0.2218, ..., 0.1147, -0.1932, -0.1692], + [ 0.1490, 0.1117, 0.1216, ..., -0.2472, -0.1400, 0.1003], + [ 0.0852, 0.0646, 0.1439, ..., 0.0576, -0.1098, -0.0253]], + device='cuda:0'), grad: tensor([[-5.5792e-08, 7.1450e-09, 4.6275e-09, ..., -1.2733e-08, + 1.3242e-09, 1.6589e-09], + [ 4.0629e-08, 1.9791e-09, 2.6193e-09, ..., 4.2055e-08, + 1.6444e-08, 4.2783e-09], + [ 4.7323e-08, 2.6193e-09, 1.2951e-09, ..., 3.7777e-08, + 8.3819e-09, 1.8772e-09], + ..., + [-1.7823e-07, 1.4872e-08, 8.8330e-09, ..., -1.8405e-07, + -8.6380e-08, 8.9931e-09], + [-3.0122e-09, -1.5265e-08, -1.6909e-08, ..., 1.9805e-08, + 1.3504e-08, -2.1493e-08], + [ 1.2486e-08, -3.0268e-08, -2.5553e-08, ..., -2.6776e-08, + 3.7107e-08, 2.0344e-08]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0189, -0.0310, -0.0070, 0.0135, -0.0015, -0.0156, 0.0062, 0.0162, + -0.0364, 0.0120], device='cuda:0'), grad: tensor([-2.7381e-07, 1.5541e-07, 2.1548e-07, 2.2794e-07, 1.7881e-07, + 3.7486e-08, -5.9517e-09, -6.1048e-07, 6.1293e-08, 4.3656e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 217.72, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4051 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.22 lr 0.00001000 +Epoch 491, weight, value: tensor([[ 0.0345, -0.1070, -0.0340, ..., -0.1584, -0.1889, -0.1833], + [-0.1682, 0.0913, -0.1584, ..., -0.2583, -0.2216, -0.2276], + [-0.1621, -0.2099, -0.2542, ..., -0.2361, -0.1864, 0.1868], + ..., + [ 0.0927, -0.0823, -0.2220, ..., 0.1147, -0.1932, -0.1693], + [ 0.1491, 0.1118, 0.1218, ..., -0.2473, -0.1401, 0.1003], + [ 0.0852, 0.0646, 0.1441, ..., 0.0576, -0.1100, -0.0255]], + device='cuda:0'), grad: tensor([[ 4.3656e-10, 3.7835e-10, 1.0186e-10, ..., 4.3656e-10, + 2.7649e-10, 4.0745e-10], + [ 5.0495e-09, -5.2241e-09, 7.5670e-10, ..., 4.5984e-09, + 2.9104e-10, 5.6752e-09], + [ 9.2841e-09, 1.4697e-09, 7.2760e-11, ..., 7.6543e-09, + 2.1828e-10, -8.9203e-09], + ..., + [-5.8790e-08, 5.3551e-09, 1.0186e-09, ..., -4.4995e-08, + 1.4552e-10, 2.0518e-09], + [ 5.2096e-09, -5.4133e-09, -3.7107e-09, ..., 8.7603e-09, + 5.2387e-10, -4.5839e-09], + [ 2.4491e-08, -1.6444e-09, -1.8481e-09, ..., 1.8306e-08, + 1.4406e-09, 1.5862e-09]], device='cuda:0') +Epoch 491, bias, value: tensor([ 0.0189, -0.0310, -0.0070, 0.0134, -0.0015, -0.0156, 0.0062, 0.0162, + -0.0364, 0.0120], device='cuda:0'), grad: tensor([ 3.8563e-09, 1.5032e-08, -9.7498e-10, 5.7800e-08, 1.0259e-08, + -7.0868e-09, 3.2305e-09, -1.4878e-07, 1.5876e-08, 7.5786e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 218.02, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4051 re_mapping 0.0015 re_causal 0.0063 /// teacc 99.23 lr 0.00001000 +Epoch 492, weight, value: tensor([[ 0.0346, -0.1071, -0.0340, ..., -0.1584, -0.1890, -0.1834], + [-0.1683, 0.0915, -0.1585, ..., -0.2584, -0.2217, -0.2278], + [-0.1622, -0.2100, -0.2543, ..., -0.2363, -0.1865, 0.1870], + ..., + [ 0.0927, -0.0823, -0.2222, ..., 0.1147, -0.1932, -0.1694], + [ 0.1493, 0.1121, 0.1221, ..., -0.2474, -0.1401, 0.1005], + [ 0.0852, 0.0646, 0.1442, ..., 0.0576, -0.1103, -0.0258]], + device='cuda:0'), grad: tensor([[-3.3033e-09, 7.1304e-10, 2.9104e-11, ..., 0.0000e+00, + 1.2267e-08, 1.1802e-08], + [ 9.8953e-10, -7.0140e-09, 1.7462e-10, ..., 7.5670e-10, + 7.1159e-09, 8.7603e-09], + [ 9.7498e-10, 4.6857e-09, 1.1642e-10, ..., -2.6193e-10, + 4.5111e-09, -2.5175e-09], + ..., + [ 3.4925e-10, 4.5839e-09, 4.3656e-10, ..., 5.8208e-10, + 1.1642e-10, 2.0809e-09], + [ 1.1642e-10, 2.2701e-09, -4.8021e-10, ..., 1.0332e-09, + 3.0384e-08, 2.8886e-08], + [ 1.2078e-09, 1.2224e-09, 1.6007e-10, ..., 6.6939e-10, + 7.8580e-10, 9.0222e-10]], device='cuda:0') +Epoch 492, bias, value: tensor([ 0.0190, -0.0310, -0.0070, 0.0134, -0.0014, -0.0157, 0.0063, 0.0162, + -0.0364, 0.0120], device='cuda:0'), grad: tensor([ 2.1028e-08, -3.1287e-09, 2.1973e-09, -9.8953e-10, 6.7870e-08, + 1.6298e-07, -3.7812e-07, 2.7663e-08, 1.0798e-07, 1.6182e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 217.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4177 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.22 lr 0.00001000 +Epoch 493, weight, value: tensor([[ 0.0348, -0.1071, -0.0340, ..., -0.1584, -0.1891, -0.1836], + [-0.1684, 0.0917, -0.1586, ..., -0.2585, -0.2218, -0.2279], + [-0.1624, -0.2101, -0.2543, ..., -0.2364, -0.1866, 0.1872], + ..., + [ 0.0927, -0.0824, -0.2225, ..., 0.1148, -0.1932, -0.1696], + [ 0.1497, 0.1123, 0.1224, ..., -0.2475, -0.1402, 0.1007], + [ 0.0852, 0.0646, 0.1443, ..., 0.0576, -0.1104, -0.0259]], + device='cuda:0'), grad: tensor([[ 5.3842e-10, 2.0518e-09, 6.5484e-10, ..., 1.1787e-09, + 2.1828e-10, 4.8021e-10], + [ 3.4051e-09, -8.6380e-07, 1.1642e-09, ..., -3.5565e-08, + -1.1089e-08, 6.2573e-10], + [ 1.2369e-09, 3.6671e-08, 6.9849e-10, ..., 2.9104e-09, + 7.1304e-10, 8.4401e-10], + ..., + [ 2.4156e-09, 7.9442e-07, 2.4011e-09, ..., 2.5873e-08, + 6.6793e-09, 4.7439e-09], + [-1.4683e-08, -3.7398e-09, -1.2515e-09, ..., 1.0463e-08, + 8.4401e-10, -1.3402e-08], + [-1.2064e-08, -1.1394e-08, -2.1057e-08, ..., -3.4488e-08, + 3.0850e-09, -5.0495e-09]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0191, -0.0310, -0.0070, 0.0134, -0.0015, -0.0157, 0.0065, 0.0162, + -0.0363, 0.0120], device='cuda:0'), grad: tensor([ 1.0754e-08, -3.9376e-06, 1.7008e-07, 2.4884e-08, 8.7079e-08, + 4.5635e-08, 3.3877e-08, 3.6694e-06, 4.3947e-09, -8.6904e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 217.83, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4056 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.19 lr 0.00001000 +Epoch 494, weight, value: tensor([[ 0.0349, -0.1071, -0.0341, ..., -0.1584, -0.1892, -0.1838], + [-0.1685, 0.0919, -0.1587, ..., -0.2586, -0.2219, -0.2281], + [-0.1625, -0.2102, -0.2544, ..., -0.2365, -0.1866, 0.1875], + ..., + [ 0.0928, -0.0825, -0.2226, ..., 0.1148, -0.1932, -0.1699], + [ 0.1499, 0.1125, 0.1225, ..., -0.2477, -0.1402, 0.1009], + [ 0.0852, 0.0647, 0.1445, ..., 0.0576, -0.1105, -0.0261]], + device='cuda:0'), grad: tensor([[ 2.1828e-10, 1.3679e-09, 2.9104e-11, ..., 3.9290e-10, + 1.7069e-08, 1.5789e-08], + [ 5.6752e-09, -4.3801e-09, 1.4552e-11, ..., 1.1409e-08, + 2.2264e-09, 1.9354e-09], + [ 1.4188e-08, 7.1450e-09, 0.0000e+00, ..., 3.2451e-08, + 6.9849e-10, 5.5297e-10], + ..., + [-4.3074e-08, 1.9587e-08, 5.6752e-10, ..., -8.1607e-08, + 2.9104e-10, 2.9104e-11], + [ 2.4738e-09, 4.3510e-09, 1.4552e-10, ..., 3.9727e-09, + 1.2049e-08, 1.0987e-08], + [ 1.6415e-08, 2.9686e-09, 0.0000e+00, ..., 2.3691e-08, + 8.4401e-10, 7.2760e-10]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0191, -0.0309, -0.0069, 0.0133, -0.0016, -0.0157, 0.0066, 0.0162, + -0.0363, 0.0120], device='cuda:0'), grad: tensor([ 5.8673e-08, 1.4654e-08, 1.1711e-07, -1.7241e-07, 1.0442e-07, + 3.0221e-07, -4.1025e-07, -1.5646e-07, 8.0501e-08, 7.9803e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 217.43, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4260 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.19 lr 0.00001000 +Epoch 495, weight, value: tensor([[ 0.0349, -0.1072, -0.0341, ..., -0.1584, -0.1892, -0.1839], + [-0.1686, 0.0921, -0.1588, ..., -0.2586, -0.2219, -0.2283], + [-0.1627, -0.2102, -0.2545, ..., -0.2367, -0.1866, 0.1877], + ..., + [ 0.0928, -0.0825, -0.2228, ..., 0.1148, -0.1934, -0.1701], + [ 0.1500, 0.1126, 0.1226, ..., -0.2479, -0.1402, 0.1010], + [ 0.0853, 0.0647, 0.1449, ..., 0.0576, -0.1106, -0.0261]], + device='cuda:0'), grad: tensor([[-8.0036e-10, 1.8917e-10, 1.3097e-10, ..., 1.7462e-10, + 2.5757e-09, 3.1141e-09], + [ 7.1595e-09, -2.9104e-10, 3.3469e-10, ..., 6.9122e-09, + 8.2946e-10, 2.5175e-09], + [ 4.2637e-09, 6.7812e-09, 3.3469e-10, ..., 5.5006e-09, + 8.2946e-10, -2.4593e-09], + ..., + [-2.0518e-08, 7.2614e-09, 2.7649e-10, ..., -2.4433e-08, + 1.4552e-11, 1.8481e-09], + [ 1.6880e-09, 4.2055e-09, 1.6007e-10, ..., 2.8813e-09, + 1.7419e-08, 2.4869e-08], + [ 5.4133e-09, 4.2201e-10, 2.9104e-10, ..., 5.4279e-09, + 8.7311e-11, 4.2201e-10]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0192, -0.0309, -0.0069, 0.0133, -0.0017, -0.0156, 0.0065, 0.0162, + -0.0363, 0.0120], device='cuda:0'), grad: tensor([ 5.7044e-09, 2.7663e-08, 2.8114e-08, -4.7905e-08, 1.7157e-08, + 1.3225e-07, -2.1025e-07, -4.7323e-08, 9.2376e-08, 2.0576e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 217.95, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4013 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.19 lr 0.00001000 +Epoch 496, weight, value: tensor([[ 0.0349, -0.1073, -0.0341, ..., -0.1585, -0.1893, -0.1840], + [-0.1688, 0.0922, -0.1589, ..., -0.2588, -0.2220, -0.2285], + [-0.1629, -0.2103, -0.2545, ..., -0.2370, -0.1867, 0.1879], + ..., + [ 0.0928, -0.0825, -0.2230, ..., 0.1149, -0.1937, -0.1703], + [ 0.1500, 0.1127, 0.1227, ..., -0.2481, -0.1403, 0.1011], + [ 0.0853, 0.0647, 0.1452, ..., 0.0576, -0.1108, -0.0262]], + device='cuda:0'), grad: tensor([[ 1.8772e-09, 3.0559e-09, 1.5425e-09, ..., 5.9663e-10, + 1.7171e-09, 2.4593e-09], + [ 1.0201e-08, 1.3533e-08, 6.4319e-09, ..., 2.3574e-09, + 4.4529e-09, 1.1525e-08], + [ 4.4674e-09, 5.0495e-09, 1.8772e-09, ..., 2.6193e-09, + 1.0041e-09, 4.3510e-09], + ..., + [ 1.7899e-09, 9.4151e-09, 3.9290e-09, ..., -4.2492e-09, + 2.5757e-09, 5.1368e-09], + [-1.2026e-07, -1.6601e-07, -7.8813e-08, ..., 9.5606e-09, + -1.0419e-07, -1.5646e-07], + [ 8.5856e-10, -1.5352e-08, -1.2675e-08, ..., -2.4549e-08, + 6.9558e-09, 7.5961e-09]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0191, -0.0310, -0.0069, 0.0133, -0.0015, -0.0156, 0.0066, 0.0162, + -0.0364, 0.0120], device='cuda:0'), grad: tensor([ 1.2136e-08, 5.7335e-08, 2.7634e-08, 4.0978e-08, 1.2034e-08, + 3.6298e-07, 1.4645e-07, 1.3417e-08, -5.7230e-07, -6.9442e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 217.59, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4021 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.20 lr 0.00001000 +Epoch 497, weight, value: tensor([[ 0.0350, -0.1074, -0.0342, ..., -0.1585, -0.1894, -0.1841], + [-0.1689, 0.0923, -0.1590, ..., -0.2589, -0.2221, -0.2287], + [-0.1631, -0.2104, -0.2546, ..., -0.2371, -0.1868, 0.1880], + ..., + [ 0.0929, -0.0826, -0.2232, ..., 0.1149, -0.1937, -0.1704], + [ 0.1502, 0.1128, 0.1228, ..., -0.2483, -0.1404, 0.1012], + [ 0.0853, 0.0648, 0.1456, ..., 0.0576, -0.1109, -0.0262]], + device='cuda:0'), grad: tensor([[-1.2966e-08, -6.8540e-09, 3.9290e-10, ..., 2.6048e-09, + -4.5693e-09, 5.8208e-11], + [ 1.0041e-08, 3.6816e-09, 1.6735e-09, ..., 1.1467e-08, + 2.0664e-09, 4.5984e-09], + [ 4.8021e-09, 6.8248e-09, 9.6043e-10, ..., 7.1886e-09, + 3.6671e-09, 3.7689e-09], + ..., + [-2.8987e-08, 6.0536e-09, 1.2806e-09, ..., -5.2154e-08, + 1.9063e-08, 1.4421e-08], + [-1.1772e-08, -5.6752e-10, -4.7439e-09, ..., 2.7212e-09, + 1.3533e-09, -1.1103e-08], + [ 2.4374e-08, 5.8208e-09, -1.1933e-09, ..., 9.1153e-08, + 2.5961e-08, 1.7142e-08]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0192, -0.0310, -0.0070, 0.0132, -0.0016, -0.0156, 0.0065, 0.0162, + -0.0365, 0.0120], device='cuda:0'), grad: tensor([-1.3702e-07, 5.8324e-08, 4.3947e-08, -2.6339e-09, -1.4086e-07, + -1.6636e-07, 1.4715e-07, -6.6124e-08, 6.3970e-08, 2.2911e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 217.53, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3993 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.22 lr 0.00001000 +Epoch 498, weight, value: tensor([[ 0.0352, -0.1074, -0.0342, ..., -0.1585, -0.1894, -0.1841], + [-0.1690, 0.0924, -0.1591, ..., -0.2589, -0.2222, -0.2288], + [-0.1633, -0.2105, -0.2546, ..., -0.2373, -0.1869, 0.1881], + ..., + [ 0.0929, -0.0826, -0.2234, ..., 0.1149, -0.1937, -0.1705], + [ 0.1503, 0.1130, 0.1230, ..., -0.2484, -0.1404, 0.1014], + [ 0.0853, 0.0648, 0.1459, ..., 0.0576, -0.1110, -0.0263]], + device='cuda:0'), grad: tensor([[-1.1714e-08, 1.8917e-10, -1.5862e-09, ..., -1.1642e-10, + 8.5856e-10, 8.4401e-10], + [ 2.8958e-09, 1.3388e-09, 6.8394e-10, ..., 5.8208e-09, + 2.6193e-10, 1.3533e-09], + [ 4.3947e-09, 8.5856e-10, 3.3469e-10, ..., 2.8522e-09, + 5.2387e-10, 8.7311e-10], + ..., + [-1.5425e-08, 2.2847e-09, 6.1118e-10, ..., -3.7835e-10, + 8.7311e-11, 7.7125e-10], + [ 6.5484e-10, -2.7212e-09, -1.7026e-09, ..., 5.2823e-09, + 1.0332e-09, -2.7794e-09], + [ 1.2937e-08, 6.6939e-10, 5.0932e-10, ..., 1.6706e-08, + 9.4587e-10, 9.3132e-10]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0193, -0.0310, -0.0070, 0.0132, -0.0017, -0.0156, 0.0065, 0.0162, + -0.0364, 0.0120], device='cuda:0'), grad: tensor([-4.5868e-08, 4.4995e-08, 2.5786e-08, 5.5297e-08, 1.8365e-08, + -3.6508e-07, 6.1933e-08, 1.0309e-07, 3.7922e-08, 8.0676e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 217.48, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4104 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.21 lr 0.00001000 +Epoch 499, weight, value: tensor([[ 0.0354, -0.1074, -0.0342, ..., -0.1586, -0.1895, -0.1842], + [-0.1690, 0.0926, -0.1592, ..., -0.2590, -0.2223, -0.2289], + [-0.1635, -0.2106, -0.2546, ..., -0.2375, -0.1869, 0.1882], + ..., + [ 0.0930, -0.0827, -0.2235, ..., 0.1150, -0.1937, -0.1705], + [ 0.1505, 0.1131, 0.1231, ..., -0.2486, -0.1405, 0.1016], + [ 0.0853, 0.0648, 0.1461, ..., 0.0576, -0.1112, -0.0264]], + device='cuda:0'), grad: tensor([[ 9.0222e-10, 2.9249e-09, 1.5425e-09, ..., 3.4925e-09, + 6.1118e-10, 6.6939e-10], + [ 1.4406e-09, -7.5204e-08, 7.8580e-10, ..., 3.3760e-09, + 1.4697e-09, -6.9413e-09], + [ 1.1059e-09, 8.1636e-09, 2.0373e-10, ..., 2.0955e-09, + 1.3533e-09, 1.8772e-09], + ..., + [ 9.3860e-09, 1.4043e-08, 7.2032e-09, ..., 1.7477e-08, + 4.4674e-09, 3.2596e-09], + [ 3.5216e-09, 5.0029e-08, 2.2264e-09, ..., 5.5588e-09, + 8.4401e-10, 6.3446e-09], + [-3.8825e-08, -3.2625e-08, -3.1840e-08, ..., 8.3237e-09, + 6.9558e-08, 3.0268e-08]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0195, -0.0310, -0.0071, 0.0132, -0.0017, -0.0156, 0.0065, 0.0162, + -0.0364, 0.0119], device='cuda:0'), grad: tensor([ 1.1860e-08, -3.0920e-07, 4.2433e-08, 7.8580e-09, -5.4919e-08, + -3.0210e-08, 8.0559e-08, 7.2294e-08, 2.3982e-07, -3.3819e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 217.80, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4113 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.20 lr 0.00001000 +Epoch 500, weight, value: tensor([[ 0.0355, -0.1075, -0.0342, ..., -0.1586, -0.1895, -0.1843], + [-0.1691, 0.0928, -0.1592, ..., -0.2591, -0.2224, -0.2291], + [-0.1638, -0.2107, -0.2547, ..., -0.2378, -0.1870, 0.1883], + ..., + [ 0.0930, -0.0827, -0.2237, ..., 0.1151, -0.1937, -0.1706], + [ 0.1505, 0.1133, 0.1233, ..., -0.2488, -0.1405, 0.1017], + [ 0.0853, 0.0648, 0.1463, ..., 0.0576, -0.1114, -0.0266]], + device='cuda:0'), grad: tensor([[-2.2847e-09, 8.1491e-10, 4.2201e-10, ..., 1.1642e-09, + 7.5670e-10, 1.6880e-09], + [ 5.4279e-09, 9.4587e-10, 9.0222e-10, ..., 6.6502e-09, + 1.3679e-09, 1.8699e-08], + [ 9.6043e-10, 5.5588e-09, 2.8376e-09, ..., -3.9290e-09, + 4.2346e-09, -7.3807e-08], + ..., + [-5.3697e-09, 7.5379e-09, 3.8999e-09, ..., -7.7271e-09, + 1.5716e-09, 3.7311e-08], + [-2.5029e-09, -7.9890e-09, -4.3365e-09, ..., 9.2841e-09, + -1.3388e-09, -1.9354e-09], + [-1.0885e-08, -1.4596e-08, -8.4547e-09, ..., -4.7585e-09, + 9.0513e-09, 8.1200e-09]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0196, -0.0309, -0.0072, 0.0131, -0.0017, -0.0155, 0.0065, 0.0162, + -0.0365, 0.0119], device='cuda:0'), grad: tensor([-2.5029e-09, 6.9849e-08, -2.2969e-07, 4.3539e-08, -8.3528e-09, + 2.1959e-08, 2.6630e-09, 9.7847e-08, 2.3618e-08, -1.2442e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 217.83, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4199 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.18 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.010002 99.180000 ... 90.134529 74.229075 +ShearY 98.889999 98.979996 ... 90.134529 69.590206 +AutoContrast 99.159996 99.220001 ... 90.134529 64.794886 +Invert 98.909996 97.829994 ... 90.134529 61.473919 +Equalize 98.540001 98.699997 ... 90.134529 70.410120 +Solarize 98.320000 98.409996 ... 90.134529 60.968220 +SolarizeAdd 98.629997 98.529999 ... 90.134529 68.521472 +Posterize 99.019997 99.080002 ... 90.134529 73.605137 +Contrast 99.089996 99.269997 ... 90.134529 70.217152 +Color 99.070000 99.250000 ... 90.134529 61.023748 +Brightness 99.040001 99.220001 ... 90.134529 71.119184 +Sharpness 99.139999 99.260002 ... 90.134529 74.840681 +NoiseSalt 99.159996 99.199997 ... 90.134529 65.811692 +NoiseGaussian 99.110001 99.239998 ... 90.134529 63.527483 +w/o do (original x) 99.250000 0.000000 ... 0.000000 76.780433 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.24 68.742317 78.524608 78.237203 89.586447 78.772644 diff --git a/Meta-causal/code-withStyleAttack/66576.error b/Meta-causal/code-withStyleAttack/66576.error new file mode 100644 index 0000000000000000000000000000000000000000..1bcb69be82d680b0fab7b86e577d164aa0960707 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66576.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 39: um: command not found diff --git a/Meta-causal/code-withStyleAttack/66576.log b/Meta-causal/code-withStyleAttack/66576.log new file mode 100644 index 0000000000000000000000000000000000000000..44eee368d2c3c11babff61d5e9261aa2075739e5 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66576.log @@ -0,0 +1,14609 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0225, 0.0167, 0.0002, ..., 0.0177, 0.0233, 0.0266], + [ 0.0205, 0.0032, -0.0094, ..., -0.0179, 0.0109, -0.0055], + [ 0.0307, -0.0056, 0.0230, ..., -0.0012, -0.0074, -0.0211], + ..., + [ 0.0139, -0.0308, 0.0127, ..., -0.0037, -0.0164, -0.0115], + [-0.0048, -0.0046, -0.0162, ..., -0.0063, -0.0257, -0.0111], + [ 0.0046, 0.0199, 0.0034, ..., 0.0211, -0.0023, -0.0060]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0309, 0.0096, 0.0175, 0.0242, 0.0277, -0.0111, 0.0008, -0.0064, + 0.0129, -0.0281], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 220.36, cls_loss 1.5267 cls_loss_mapping 1.9075 cls_loss_causal 2.2132 re_mapping 0.1092 re_causal 0.1090 /// teacc 76.68 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0201, 0.0167, -0.0037, ..., 0.0152, 0.0256, 0.0267], + [ 0.0177, 0.0032, -0.0068, ..., -0.0277, 0.0075, -0.0060], + [ 0.0289, -0.0056, 0.0189, ..., -0.0083, -0.0071, -0.0237], + ..., + [ 0.0049, -0.0308, 0.0190, ..., -0.0020, -0.0202, -0.0192], + [-0.0054, -0.0046, -0.0150, ..., -0.0035, -0.0280, -0.0132], + [-0.0039, 0.0199, 0.0064, ..., 0.0266, -0.0072, -0.0137]], + device='cuda:0'), grad: tensor([[ 0.0073, 0.0000, 0.0025, ..., 0.0043, -0.0046, 0.0015], + [-0.0229, 0.0000, -0.0339, ..., -0.0133, 0.0003, -0.0093], + [ 0.0087, 0.0000, 0.0047, ..., 0.0064, 0.0014, 0.0115], + ..., + [ 0.0060, 0.0000, -0.0197, ..., -0.0346, 0.0003, 0.0047], + [ 0.0094, 0.0000, -0.0047, ..., -0.0160, 0.0012, 0.0116], + [ 0.0079, 0.0000, 0.0296, ..., 0.0594, 0.0003, 0.0078]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0286, 0.0107, 0.0166, 0.0242, 0.0282, -0.0103, 0.0009, -0.0055, + 0.0122, -0.0278], device='cuda:0'), grad: tensor([ 0.0085, -0.0685, 0.0128, 0.0494, 0.0425, -0.0736, 0.0026, -0.0192, + -0.0084, 0.0539], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 219.53, cls_loss 0.5938 cls_loss_mapping 0.8408 cls_loss_causal 1.9057 re_mapping 0.2141 re_causal 0.2563 /// teacc 90.99 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0245, 0.0261, -0.0054, ..., 0.0157, 0.0307, 0.0270], + [ 0.0198, -0.0102, -0.0059, ..., -0.0310, 0.0052, -0.0060], + [ 0.0230, -0.0053, 0.0215, ..., -0.0099, -0.0045, -0.0271], + ..., + [-0.0012, -0.0333, 0.0206, ..., -0.0030, -0.0232, -0.0226], + [-0.0053, -0.0102, -0.0147, ..., -0.0009, -0.0321, -0.0142], + [-0.0063, 0.0145, 0.0101, ..., 0.0293, -0.0117, -0.0174]], + device='cuda:0'), grad: tensor([[ 0.0179, 0.0370, 0.0008, ..., 0.0012, 0.0406, 0.0329], + [ 0.0097, 0.0006, 0.0047, ..., 0.0054, 0.0051, 0.0108], + [-0.0152, -0.0220, 0.0103, ..., 0.0091, -0.0278, -0.0152], + ..., + [ 0.0012, -0.0083, -0.0141, ..., -0.0144, 0.0011, 0.0011], + [-0.0235, -0.0016, -0.0084, ..., -0.0048, -0.0111, -0.0231], + [ 0.0034, 0.0105, 0.0216, ..., 0.0271, 0.0012, 0.0041]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0286, 0.0112, 0.0167, 0.0236, 0.0278, -0.0088, 0.0002, -0.0061, + 0.0122, -0.0276], device='cuda:0'), grad: tensor([ 0.0415, 0.0196, -0.0148, -0.0119, 0.0145, -0.0011, -0.0177, -0.0120, + -0.0407, 0.0226], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 219.37, cls_loss 0.3528 cls_loss_mapping 0.5048 cls_loss_causal 1.6672 re_mapping 0.1607 re_causal 0.2469 /// teacc 94.06 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0282, 0.0310, -0.0071, ..., 0.0152, 0.0327, 0.0265], + [ 0.0220, -0.0141, -0.0044, ..., -0.0327, 0.0038, -0.0054], + [ 0.0187, -0.0068, 0.0225, ..., -0.0110, -0.0029, -0.0305], + ..., + [-0.0082, -0.0330, 0.0225, ..., -0.0034, -0.0251, -0.0254], + [-0.0052, -0.0127, -0.0148, ..., -0.0003, -0.0336, -0.0147], + [-0.0057, 0.0149, 0.0114, ..., 0.0309, -0.0133, -0.0183]], + device='cuda:0'), grad: tensor([[-5.7335e-03, -3.5362e-03, 2.6655e-04, ..., 1.9569e-03, + -5.1361e-02, -3.3386e-02], + [-5.5351e-03, 5.1171e-05, -2.9163e-03, ..., 1.6575e-03, + -6.1131e-04, -2.6169e-03], + [-2.0523e-03, 7.4863e-04, -3.0365e-02, ..., -1.9188e-03, + -9.5673e-03, 1.0605e-03], + ..., + [ 4.6616e-03, 8.5783e-04, 1.5022e-02, ..., 1.2039e-02, + 2.7351e-03, 1.4219e-03], + [-4.7607e-02, 2.5058e-04, 8.6212e-03, ..., -3.5095e-02, + -6.2218e-03, -4.8737e-02], + [ 2.0504e-03, 1.8203e-04, -1.2543e-02, ..., -2.1454e-02, + 2.8725e-03, 2.3060e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0282, 0.0117, 0.0164, 0.0236, 0.0277, -0.0085, 0.0001, -0.0060, + 0.0122, -0.0277], device='cuda:0'), grad: tensor([-0.0397, -0.0046, -0.0274, 0.0409, 0.0020, 0.0201, 0.0434, 0.0181, + -0.0376, -0.0151], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 219.32, cls_loss 0.2471 cls_loss_mapping 0.3487 cls_loss_causal 1.5058 re_mapping 0.1274 re_causal 0.2318 /// teacc 94.89 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0305, 0.0348, -0.0081, ..., 0.0146, 0.0350, 0.0269], + [ 0.0237, -0.0174, -0.0036, ..., -0.0346, 0.0042, -0.0050], + [ 0.0150, -0.0087, 0.0233, ..., -0.0131, -0.0026, -0.0332], + ..., + [-0.0123, -0.0327, 0.0232, ..., -0.0039, -0.0274, -0.0271], + [-0.0064, -0.0147, -0.0144, ..., 0.0005, -0.0354, -0.0153], + [-0.0041, 0.0154, 0.0124, ..., 0.0326, -0.0140, -0.0189]], + device='cuda:0'), grad: tensor([[ 0.0059, -0.0004, 0.0047, ..., 0.0031, 0.0001, 0.0063], + [-0.0031, 0.0005, -0.0039, ..., 0.0015, -0.0007, -0.0007], + [ 0.0047, -0.0002, -0.0274, ..., -0.0009, 0.0035, 0.0104], + ..., + [ 0.0120, 0.0036, 0.0207, ..., 0.0140, 0.0009, 0.0075], + [-0.0013, 0.0037, 0.0043, ..., -0.0010, -0.0032, -0.0052], + [ 0.0015, 0.0003, -0.0022, ..., -0.0103, 0.0013, 0.0033]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0289, 0.0120, 0.0162, 0.0235, 0.0277, -0.0086, -0.0001, -0.0063, + 0.0120, -0.0274], device='cuda:0'), grad: tensor([ 0.0010, -0.0056, 0.0031, 0.0058, -0.0155, -0.0104, 0.0064, 0.0269, + -0.0172, 0.0054], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 219.06, cls_loss 0.2034 cls_loss_mapping 0.2719 cls_loss_causal 1.3978 re_mapping 0.1024 re_causal 0.2110 /// teacc 95.77 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0336, 0.0380, -0.0092, ..., 0.0140, 0.0356, 0.0264], + [ 0.0246, -0.0194, -0.0027, ..., -0.0355, 0.0041, -0.0055], + [ 0.0132, -0.0101, 0.0234, ..., -0.0151, -0.0022, -0.0348], + ..., + [-0.0149, -0.0333, 0.0243, ..., -0.0042, -0.0294, -0.0289], + [-0.0068, -0.0169, -0.0139, ..., 0.0011, -0.0368, -0.0154], + [-0.0043, 0.0150, 0.0127, ..., 0.0337, -0.0149, -0.0198]], + device='cuda:0'), grad: tensor([[-0.0215, -0.0197, -0.0004, ..., 0.0006, -0.0313, -0.0342], + [-0.0025, 0.0009, -0.0030, ..., 0.0004, 0.0004, 0.0003], + [ 0.0017, 0.0007, 0.0006, ..., 0.0008, 0.0004, 0.0021], + ..., + [ 0.0009, 0.0003, 0.0013, ..., 0.0076, 0.0008, 0.0017], + [ 0.0024, 0.0010, 0.0002, ..., -0.0002, 0.0025, 0.0024], + [ 0.0015, 0.0006, -0.0052, ..., -0.0186, 0.0010, -0.0017]], + device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0289, 0.0119, 0.0163, 0.0234, 0.0277, -0.0084, -0.0004, -0.0061, + 0.0122, -0.0277], device='cuda:0'), grad: tensor([-0.0379, -0.0033, -0.0021, 0.0110, 0.0090, 0.0096, 0.0133, 0.0057, + 0.0056, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 219.04, cls_loss 0.1859 cls_loss_mapping 0.2417 cls_loss_causal 1.3359 re_mapping 0.0876 re_causal 0.1969 /// teacc 96.10 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0356, 0.0414, -0.0098, ..., 0.0135, 0.0370, 0.0266], + [ 0.0256, -0.0229, -0.0024, ..., -0.0365, 0.0039, -0.0058], + [ 0.0115, -0.0113, 0.0240, ..., -0.0155, -0.0014, -0.0362], + ..., + [-0.0170, -0.0329, 0.0251, ..., -0.0049, -0.0309, -0.0308], + [-0.0071, -0.0179, -0.0137, ..., 0.0010, -0.0385, -0.0153], + [-0.0042, 0.0147, 0.0135, ..., 0.0347, -0.0164, -0.0202]], + device='cuda:0'), grad: tensor([[ 0.0012, -0.0014, 0.0003, ..., 0.0004, 0.0009, 0.0018], + [-0.0003, 0.0002, -0.0026, ..., 0.0002, -0.0009, 0.0009], + [ 0.0018, 0.0010, -0.0075, ..., 0.0006, -0.0043, 0.0026], + ..., + [ 0.0007, 0.0002, 0.0012, ..., 0.0028, 0.0010, 0.0004], + [ 0.0069, 0.0012, 0.0025, ..., -0.0003, 0.0045, 0.0062], + [ 0.0040, 0.0010, -0.0029, ..., -0.0068, 0.0010, 0.0036]], + device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0288, 0.0118, 0.0165, 0.0234, 0.0279, -0.0086, -0.0008, -0.0061, + 0.0123, -0.0276], device='cuda:0'), grad: tensor([ 0.0027, -0.0022, -0.0161, 0.0092, 0.0109, 0.0067, -0.0277, 0.0042, + 0.0111, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 219.04, cls_loss 0.1534 cls_loss_mapping 0.1970 cls_loss_causal 1.2614 re_mapping 0.0771 re_causal 0.1781 /// teacc 96.11 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0374, 0.0436, -0.0102, ..., 0.0129, 0.0382, 0.0266], + [ 0.0263, -0.0280, -0.0018, ..., -0.0369, 0.0033, -0.0068], + [ 0.0105, -0.0131, 0.0244, ..., -0.0165, 0.0005, -0.0374], + ..., + [-0.0190, -0.0330, 0.0256, ..., -0.0057, -0.0341, -0.0313], + [-0.0077, -0.0182, -0.0137, ..., 0.0014, -0.0391, -0.0153], + [-0.0046, 0.0158, 0.0143, ..., 0.0359, -0.0170, -0.0214]], + device='cuda:0'), grad: tensor([[-2.6155e-04, -3.1494e-02, -6.0501e-03, ..., -1.1474e-05, + -1.3229e-02, -1.6586e-02], + [-2.8248e-03, 8.6117e-04, -1.1597e-03, ..., 1.0719e-03, + 5.3120e-04, -1.8024e-04], + [ 2.1133e-03, 5.4398e-03, 7.6532e-04, ..., 3.3903e-04, + 2.3289e-03, 4.4250e-03], + ..., + [ 6.7616e-04, 7.1096e-04, -4.5824e-04, ..., 5.4502e-04, + 4.7159e-04, 1.0986e-03], + [ 3.6831e-03, 1.4236e-02, 3.1090e-03, ..., 9.8801e-04, + 6.5308e-03, 1.3779e-02], + [ 2.2774e-03, 2.2945e-03, 1.0262e-03, ..., 3.2921e-03, + 1.2302e-03, 4.7417e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0290, 0.0119, 0.0168, 0.0232, 0.0277, -0.0090, -0.0007, -0.0062, + 0.0125, -0.0274], device='cuda:0'), grad: tensor([-0.0267, -0.0005, 0.0052, -0.0034, -0.0115, 0.0005, 0.0030, 0.0023, + 0.0206, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 219.06, cls_loss 0.1330 cls_loss_mapping 0.1680 cls_loss_causal 1.2426 re_mapping 0.0670 re_causal 0.1663 /// teacc 96.35 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0384, 0.0466, -0.0114, ..., 0.0127, 0.0388, 0.0269], + [ 0.0282, -0.0300, -0.0014, ..., -0.0375, 0.0033, -0.0076], + [ 0.0084, -0.0148, 0.0257, ..., -0.0170, 0.0016, -0.0392], + ..., + [-0.0206, -0.0332, 0.0261, ..., -0.0062, -0.0364, -0.0321], + [-0.0084, -0.0180, -0.0137, ..., 0.0016, -0.0407, -0.0149], + [-0.0041, 0.0156, 0.0147, ..., 0.0369, -0.0173, -0.0220]], + device='cuda:0'), grad: tensor([[ 1.2808e-03, -5.5170e-04, 7.5817e-04, ..., 1.9479e-04, + 6.9571e-04, 1.1396e-03], + [ 6.9678e-05, 1.2994e-05, 1.2722e-03, ..., 4.8876e-04, + -3.2687e-04, 8.2970e-04], + [ 1.3905e-03, 1.5545e-04, -6.9580e-03, ..., 2.0676e-03, + -4.1313e-03, 2.5387e-03], + ..., + [ 1.9360e-04, 1.6674e-05, 9.2773e-03, ..., 2.6779e-03, + 3.0923e-04, 1.9054e-03], + [ 5.1880e-03, -2.1815e-04, 9.9716e-03, ..., -2.6855e-03, + 2.5368e-03, 1.3151e-03], + [ 2.0752e-03, 2.2650e-04, 3.8109e-03, ..., 1.0757e-03, + 4.2224e-04, 2.5024e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0292, 0.0122, 0.0171, 0.0229, 0.0275, -0.0090, -0.0012, -0.0061, + 0.0125, -0.0273], device='cuda:0'), grad: tensor([ 0.0025, 0.0020, -0.0109, -0.0014, 0.0043, -0.0030, -0.0199, 0.0101, + 0.0113, 0.0050], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 219.05, cls_loss 0.1259 cls_loss_mapping 0.1528 cls_loss_causal 1.1532 re_mapping 0.0622 re_causal 0.1539 /// teacc 96.76 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0397, 0.0494, -0.0122, ..., 0.0121, 0.0395, 0.0270], + [ 0.0286, -0.0336, -0.0016, ..., -0.0384, 0.0021, -0.0086], + [ 0.0074, -0.0158, 0.0263, ..., -0.0178, 0.0034, -0.0404], + ..., + [-0.0222, -0.0344, 0.0266, ..., -0.0064, -0.0380, -0.0328], + [-0.0089, -0.0185, -0.0133, ..., 0.0020, -0.0423, -0.0142], + [-0.0031, 0.0155, 0.0158, ..., 0.0379, -0.0177, -0.0223]], + device='cuda:0'), grad: tensor([[ 0.0145, 0.0186, 0.0106, ..., 0.0003, 0.0076, 0.0002], + [-0.0010, 0.0003, -0.0007, ..., 0.0003, -0.0003, 0.0003], + [ 0.0024, 0.0031, 0.0032, ..., 0.0003, 0.0022, 0.0003], + ..., + [ 0.0015, 0.0009, 0.0100, ..., 0.0069, 0.0008, 0.0001], + [ 0.0019, 0.0014, -0.0022, ..., -0.0004, 0.0005, -0.0003], + [-0.0215, -0.0328, -0.0349, ..., -0.0106, -0.0123, 0.0008]], + device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0295, 0.0115, 0.0174, 0.0232, 0.0277, -0.0093, -0.0013, -0.0063, + 0.0125, -0.0272], device='cuda:0'), grad: tensor([ 0.0189, -0.0006, 0.0060, 0.0105, 0.0013, 0.0077, -0.0054, 0.0057, + -0.0009, -0.0433], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 219.37, cls_loss 0.1196 cls_loss_mapping 0.1447 cls_loss_causal 1.1710 re_mapping 0.0586 re_causal 0.1453 /// teacc 96.94 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0412, 0.0514, -0.0133, ..., 0.0117, 0.0398, 0.0272], + [ 0.0299, -0.0366, -0.0011, ..., -0.0388, 0.0017, -0.0094], + [ 0.0054, -0.0170, 0.0261, ..., -0.0183, 0.0042, -0.0417], + ..., + [-0.0239, -0.0328, 0.0272, ..., -0.0067, -0.0389, -0.0333], + [-0.0094, -0.0193, -0.0125, ..., 0.0021, -0.0432, -0.0137], + [-0.0031, 0.0165, 0.0161, ..., 0.0390, -0.0178, -0.0230]], + device='cuda:0'), grad: tensor([[ 1.7858e-04, 3.5000e-04, 5.4979e-04, ..., 1.7166e-04, + 4.4632e-03, 9.9659e-05], + [ 2.1315e-04, 4.6229e-04, 1.3285e-03, ..., 8.7881e-04, + 5.0354e-04, 9.9564e-04], + [ 3.3712e-04, -1.5230e-03, -2.3499e-03, ..., 5.8413e-04, + -4.7836e-03, 3.2473e-04], + ..., + [-6.2108e-05, -1.7262e-03, -3.7262e-02, ..., -2.4017e-02, + 1.9193e-04, -1.2197e-03], + [-6.1493e-03, -4.5776e-03, -1.4305e-03, ..., -3.0422e-03, + 2.0294e-03, -7.7324e-03], + [ 5.4932e-03, 1.8673e-03, 1.6541e-02, ..., 1.2100e-02, + 6.5613e-04, 5.7602e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0295, 0.0116, 0.0171, 0.0232, 0.0275, -0.0095, -0.0014, -0.0061, + 0.0128, -0.0271], device='cuda:0'), grad: tensor([ 0.0056, 0.0024, -0.0086, 0.0046, 0.0048, 0.0123, -0.0035, -0.0315, + -0.0051, 0.0189], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 219.19, cls_loss 0.1014 cls_loss_mapping 0.1269 cls_loss_causal 1.0954 re_mapping 0.0542 re_causal 0.1359 /// teacc 97.24 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0417, 0.0532, -0.0136, ..., 0.0113, 0.0404, 0.0272], + [ 0.0313, -0.0384, -0.0005, ..., -0.0390, 0.0016, -0.0101], + [ 0.0035, -0.0182, 0.0260, ..., -0.0193, 0.0052, -0.0429], + ..., + [-0.0252, -0.0327, 0.0278, ..., -0.0069, -0.0403, -0.0338], + [-0.0099, -0.0188, -0.0121, ..., 0.0025, -0.0440, -0.0135], + [-0.0031, 0.0169, 0.0166, ..., 0.0397, -0.0179, -0.0240]], + device='cuda:0'), grad: tensor([[-0.0103, -0.0351, -0.0093, ..., -0.0177, -0.0022, -0.0048], + [-0.0035, 0.0003, -0.0008, ..., -0.0002, 0.0015, -0.0002], + [ 0.0009, 0.0017, -0.0043, ..., 0.0010, -0.0035, 0.0004], + ..., + [-0.0001, -0.0006, -0.0025, ..., 0.0004, 0.0008, 0.0001], + [ 0.0026, 0.0034, 0.0014, ..., -0.0024, 0.0006, 0.0010], + [ 0.0078, 0.0181, 0.0078, ..., 0.0111, 0.0003, 0.0032]], + device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0296, 0.0119, 0.0169, 0.0233, 0.0276, -0.0097, -0.0017, -0.0061, + 0.0130, -0.0271], device='cuda:0'), grad: tensor([-0.0312, 0.0011, 0.0004, 0.0174, 0.0125, -0.0067, 0.0052, -0.0013, + -0.0167, 0.0193], device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 218.47, cls_loss 0.0930 cls_loss_mapping 0.1163 cls_loss_causal 1.0807 re_mapping 0.0520 re_causal 0.1254 /// teacc 97.11 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0425, 0.0557, -0.0136, ..., 0.0114, 0.0411, 0.0274], + [ 0.0319, -0.0413, -0.0009, ..., -0.0393, 0.0009, -0.0111], + [ 0.0036, -0.0194, 0.0262, ..., -0.0201, 0.0070, -0.0436], + ..., + [-0.0263, -0.0328, 0.0282, ..., -0.0074, -0.0409, -0.0346], + [-0.0108, -0.0199, -0.0117, ..., 0.0025, -0.0455, -0.0134], + [-0.0040, 0.0164, 0.0172, ..., 0.0405, -0.0183, -0.0258]], + device='cuda:0'), grad: tensor([[ 4.4942e-04, 3.6812e-04, 2.0618e-03, ..., 1.3943e-03, + 1.0147e-03, 4.7421e-04], + [-9.3746e-04, 2.1744e-04, -4.7505e-05, ..., 4.6539e-04, + -2.5177e-04, 9.9087e-04], + [ 1.0538e-03, 3.8195e-04, 7.8812e-03, ..., 1.8988e-03, + 2.1057e-03, 3.9816e-04], + ..., + [ 9.9182e-04, 1.9372e-04, -4.3449e-03, ..., 1.0023e-03, + -4.7445e-04, 7.9918e-04], + [-4.6234e-03, -1.8740e-03, -8.3618e-03, ..., -6.6452e-03, + -2.7714e-03, -5.6305e-03], + [ 2.2864e-04, 6.7532e-05, -4.8828e-03, ..., -1.5907e-03, + 7.4387e-04, 2.5272e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0301, 0.0111, 0.0174, 0.0235, 0.0275, -0.0097, -0.0021, -0.0059, + 0.0131, -0.0275], device='cuda:0'), grad: tensor([ 0.0035, 0.0002, 0.0072, 0.0028, -0.0035, 0.0038, 0.0012, 0.0024, + -0.0130, -0.0046], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 218.94, cls_loss 0.0899 cls_loss_mapping 0.1074 cls_loss_causal 1.0240 re_mapping 0.0483 re_causal 0.1159 /// teacc 97.20 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0435, 0.0570, -0.0147, ..., 0.0111, 0.0411, 0.0273], + [ 0.0329, -0.0411, -0.0005, ..., -0.0395, 0.0009, -0.0113], + [ 0.0027, -0.0205, 0.0258, ..., -0.0209, 0.0071, -0.0448], + ..., + [-0.0271, -0.0328, 0.0290, ..., -0.0077, -0.0411, -0.0355], + [-0.0116, -0.0209, -0.0111, ..., 0.0024, -0.0459, -0.0131], + [-0.0042, 0.0166, 0.0177, ..., 0.0412, -0.0191, -0.0264]], + device='cuda:0'), grad: tensor([[ 1.5235e-04, -4.3488e-04, -8.3780e-04, ..., 5.6952e-05, + 5.0545e-04, 4.3945e-03], + [-9.1648e-04, 3.5429e-04, -7.9060e-04, ..., 2.2805e-04, + 1.5831e-04, 2.9349e-04], + [ 1.0395e-04, 1.1435e-03, 7.4673e-04, ..., 2.3437e-04, + 4.1294e-04, 7.1955e-04], + ..., + [ 3.1352e-04, 2.3861e-03, 2.6608e-03, ..., 1.4410e-03, + 1.0700e-03, 3.0422e-03], + [ 6.9237e-04, 5.9605e-04, 4.4203e-04, ..., -3.6985e-05, + 5.5027e-04, 1.2007e-03], + [ 2.9492e-04, 3.6335e-03, -6.4011e-03, ..., -4.5433e-03, + 1.6642e-03, 6.7711e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0296, 0.0110, 0.0171, 0.0232, 0.0275, -0.0098, -0.0016, -0.0057, + 0.0133, -0.0272], device='cuda:0'), grad: tensor([ 0.0051, -0.0008, 0.0018, -0.0272, 0.0041, 0.0036, 0.0012, 0.0079, + 0.0026, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 219.30, cls_loss 0.0991 cls_loss_mapping 0.1125 cls_loss_causal 1.0146 re_mapping 0.0465 re_causal 0.1097 /// teacc 97.36 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0440, 0.0589, -0.0150, ..., 0.0109, 0.0419, 0.0276], + [ 0.0331, -0.0424, -0.0003, ..., -0.0401, 0.0006, -0.0119], + [ 0.0018, -0.0220, 0.0264, ..., -0.0211, 0.0083, -0.0458], + ..., + [-0.0277, -0.0327, 0.0290, ..., -0.0083, -0.0428, -0.0359], + [-0.0117, -0.0204, -0.0110, ..., 0.0028, -0.0470, -0.0128], + [-0.0046, 0.0161, 0.0181, ..., 0.0415, -0.0192, -0.0276]], + device='cuda:0'), grad: tensor([[ 2.2459e-04, -1.2484e-03, -2.8515e-04, ..., 2.8148e-05, + 4.1199e-04, -8.4519e-05], + [-1.4143e-03, 5.9843e-05, -4.0703e-03, ..., 2.7046e-05, + -1.3390e-03, -1.1673e-03], + [ 8.2207e-04, 2.3913e-04, 7.6246e-04, ..., 5.9098e-05, + -7.7128e-05, 5.8842e-04], + ..., + [ 2.8944e-04, 6.9678e-05, -1.2169e-03, ..., 5.9426e-05, + 3.3617e-04, 2.2340e-04], + [ 2.3727e-03, 9.3341e-05, 2.7103e-03, ..., 1.5330e-04, + 1.5392e-03, 1.5879e-03], + [ 1.6582e-04, 3.5191e-04, -2.9445e-04, ..., -9.2697e-04, + -5.5552e-05, 1.5032e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0299, 0.0109, 0.0174, 0.0233, 0.0275, -0.0099, -0.0019, -0.0059, + 0.0134, -0.0272], device='cuda:0'), grad: tensor([-0.0001, -0.0042, 0.0020, -0.0100, 0.0010, 0.0076, -0.0018, -0.0009, + 0.0062, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 219.45, cls_loss 0.0832 cls_loss_mapping 0.0993 cls_loss_causal 1.0426 re_mapping 0.0441 re_causal 0.1102 /// teacc 97.42 lr 0.00010000 +Epoch 16, weight, value: tensor([[-4.4734e-02, 6.0386e-02, -1.4948e-02, ..., 1.0627e-02, + 4.2105e-02, 2.7441e-02], + [ 3.3678e-02, -4.3849e-02, -4.8130e-05, ..., -4.0221e-02, + -2.9464e-04, -1.2517e-02], + [ 1.6860e-03, -2.3486e-02, 2.6096e-02, ..., -2.1325e-02, + 9.3325e-03, -4.7144e-02], + ..., + [-2.8553e-02, -3.2688e-02, 2.9642e-02, ..., -8.4565e-03, + -4.2681e-02, -3.6918e-02], + [-1.1891e-02, -1.9113e-02, -1.0603e-02, ..., 3.0223e-03, + -4.7771e-02, -1.2133e-02], + [-4.7448e-03, 1.5314e-02, 1.8377e-02, ..., 4.2159e-02, + -1.9740e-02, -2.8544e-02]], device='cuda:0'), grad: tensor([[ 2.7466e-04, -8.4305e-04, 1.1749e-03, ..., 4.6283e-05, + -6.9737e-06, 7.8321e-05], + [-5.9271e-04, 2.8872e-04, -3.7646e-04, ..., 9.6142e-05, + 1.9670e-04, -7.0274e-05], + [ 3.5214e-04, 9.7811e-05, -9.7322e-04, ..., 5.4598e-05, + -8.8739e-04, 1.6153e-04], + ..., + [ 3.3760e-04, -7.0870e-05, -8.9502e-04, ..., -1.3518e-04, + 2.8563e-04, -2.8563e-04], + [-3.2120e-03, -3.2425e-03, -4.4365e-03, ..., -1.3316e-04, + 5.9271e-04, -5.0449e-04], + [ 4.4489e-04, 1.8036e-04, -2.5043e-03, ..., -1.3161e-03, + -1.0262e-03, 9.0075e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0298, 0.0107, 0.0172, 0.0236, 0.0274, -0.0100, -0.0023, -0.0053, + 0.0137, -0.0276], device='cuda:0'), grad: tensor([ 6.9332e-04, -9.3281e-05, -1.4238e-03, -2.1347e-02, -6.9313e-03, + 2.7252e-02, 9.0256e-03, -4.0507e-04, -4.5471e-03, -2.2278e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 218.93, cls_loss 0.0793 cls_loss_mapping 0.0891 cls_loss_causal 0.9795 re_mapping 0.0410 re_causal 0.1002 /// teacc 97.57 lr 0.00010000 +Epoch 17, weight, value: tensor([[-4.5602e-02, 6.1512e-02, -1.5708e-02, ..., 9.9933e-03, + 4.2156e-02, 2.7156e-02], + [ 3.4906e-02, -4.4998e-02, -6.8769e-05, ..., -4.0637e-02, + -1.3987e-03, -1.2916e-02], + [ 4.3029e-04, -2.4813e-02, 2.6310e-02, ..., -2.2162e-02, + 1.0940e-02, -4.8528e-02], + ..., + [-2.9407e-02, -3.2944e-02, 3.0067e-02, ..., -8.4086e-03, + -4.3730e-02, -3.8085e-02], + [-1.2785e-02, -1.9664e-02, -1.0396e-02, ..., 3.1274e-03, + -4.8599e-02, -1.2237e-02], + [-4.7859e-03, 1.6342e-02, 1.9213e-02, ..., 4.3115e-02, + -1.9459e-02, -2.8609e-02]], device='cuda:0'), grad: tensor([[ 6.4135e-05, -2.6274e-04, 5.1528e-05, ..., 3.5197e-05, + 1.7536e-04, 3.6418e-05], + [ 9.4461e-04, 6.1691e-06, 4.5700e-03, ..., 2.1114e-03, + 2.8372e-04, 1.0610e-04], + [ 1.1009e-04, 4.6492e-05, -2.5501e-03, ..., 1.1235e-04, + -1.6165e-03, 1.4365e-04], + ..., + [ 3.8433e-03, 1.5035e-05, 1.5335e-02, ..., 8.5983e-03, + 1.7452e-04, 4.4322e-04], + [ 1.2245e-03, 2.0832e-05, 2.7485e-03, ..., 1.1234e-03, + 3.4666e-04, 1.0872e-03], + [-1.1734e-02, 7.2002e-05, -2.9129e-02, ..., -1.6876e-02, + 6.0976e-05, -6.0463e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0294, 0.0106, 0.0171, 0.0236, 0.0274, -0.0104, -0.0020, -0.0054, + 0.0139, -0.0272], device='cuda:0'), grad: tensor([ 0.0004, 0.0037, -0.0045, 0.0015, 0.0005, 0.0057, 0.0006, 0.0119, + 0.0035, -0.0234], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 219.59, cls_loss 0.0766 cls_loss_mapping 0.0861 cls_loss_causal 0.9726 re_mapping 0.0376 re_causal 0.0937 /// teacc 97.69 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0466, 0.0634, -0.0154, ..., 0.0099, 0.0429, 0.0274], + [ 0.0357, -0.0459, 0.0003, ..., -0.0410, -0.0018, -0.0134], + [-0.0005, -0.0255, 0.0266, ..., -0.0217, 0.0112, -0.0493], + ..., + [-0.0307, -0.0334, 0.0304, ..., -0.0089, -0.0440, -0.0396], + [-0.0132, -0.0196, -0.0104, ..., 0.0035, -0.0493, -0.0121], + [-0.0046, 0.0158, 0.0197, ..., 0.0439, -0.0191, -0.0295]], + device='cuda:0'), grad: tensor([[ 1.3137e-04, -1.8606e-03, -3.3975e-04, ..., 8.7991e-06, + -5.9395e-03, -8.1921e-04], + [-6.8760e-04, 2.5928e-05, -8.0156e-04, ..., -2.3484e-04, + 5.8746e-04, 6.0678e-05], + [ 2.0850e-04, 1.7846e-04, -5.2261e-04, ..., 9.1672e-05, + -5.2869e-05, 1.6284e-04], + ..., + [ 4.0460e-04, 1.2803e-04, 6.7997e-04, ..., 1.6761e-04, + 3.9697e-04, 1.0473e-04], + [-6.0272e-03, 2.8992e-04, 1.5998e-04, ..., -4.0507e-04, + 2.6369e-04, -5.9586e-03], + [ 2.5082e-04, 5.0497e-04, 1.2481e-04, ..., -3.2008e-05, + 6.5947e-04, 4.8733e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0298, 0.0106, 0.0171, 0.0234, 0.0274, -0.0102, -0.0025, -0.0055, + 0.0139, -0.0270], device='cuda:0'), grad: tensor([-2.0569e-02, 3.0383e-05, 2.4529e-03, 2.8439e-03, 9.0637e-03, + 1.4885e-02, 1.8654e-03, 2.2221e-03, -1.6174e-02, 3.3741e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 219.17, cls_loss 0.0747 cls_loss_mapping 0.0913 cls_loss_causal 0.9916 re_mapping 0.0365 re_causal 0.0920 /// teacc 97.78 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0472, 0.0649, -0.0157, ..., 0.0095, 0.0433, 0.0272], + [ 0.0361, -0.0464, 0.0008, ..., -0.0410, -0.0025, -0.0140], + [-0.0007, -0.0264, 0.0267, ..., -0.0219, 0.0121, -0.0502], + ..., + [-0.0315, -0.0334, 0.0305, ..., -0.0093, -0.0448, -0.0402], + [-0.0136, -0.0200, -0.0101, ..., 0.0034, -0.0500, -0.0119], + [-0.0055, 0.0157, 0.0200, ..., 0.0445, -0.0195, -0.0308]], + device='cuda:0'), grad: tensor([[ 1.2417e-03, 2.2030e-03, 7.8535e-04, ..., 6.6519e-05, + 1.6570e-04, 3.8338e-03], + [-1.3523e-03, 5.2023e-04, -2.0099e-04, ..., 1.9240e-04, + 1.6642e-04, -1.8215e-04], + [ 4.5514e-04, 7.8857e-05, -2.4490e-03, ..., 6.7770e-05, + -1.0004e-03, -1.2369e-03], + ..., + [ 4.1628e-04, 5.3453e-04, 1.2350e-03, ..., 2.6083e-04, + 2.4867e-04, 2.6441e-04], + [-5.9814e-03, -1.2863e-02, -1.5192e-03, ..., -3.0947e-04, + 1.7774e-04, -2.1759e-02], + [ 1.1911e-03, 1.8263e-03, 2.6150e-03, ..., 4.7326e-04, + 1.6057e-04, 2.0766e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0297, 0.0107, 0.0175, 0.0230, 0.0274, -0.0096, -0.0024, -0.0056, + 0.0138, -0.0274], device='cuda:0'), grad: tensor([ 0.0037, 0.0007, -0.0059, 0.0032, -0.0131, 0.0113, 0.0020, 0.0033, + -0.0132, 0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 218.21, cls_loss 0.0687 cls_loss_mapping 0.0769 cls_loss_causal 0.9133 re_mapping 0.0370 re_causal 0.0890 /// teacc 97.73 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0481, 0.0665, -0.0162, ..., 0.0097, 0.0436, 0.0271], + [ 0.0364, -0.0451, 0.0005, ..., -0.0420, -0.0029, -0.0140], + [-0.0014, -0.0270, 0.0264, ..., -0.0225, 0.0129, -0.0515], + ..., + [-0.0324, -0.0338, 0.0312, ..., -0.0096, -0.0451, -0.0407], + [-0.0140, -0.0210, -0.0096, ..., 0.0037, -0.0507, -0.0115], + [-0.0044, 0.0153, 0.0203, ..., 0.0451, -0.0198, -0.0308]], + device='cuda:0'), grad: tensor([[-8.4257e-04, -2.6455e-03, -3.8648e-04, ..., 4.3660e-05, + 5.4911e-06, -4.4918e-04], + [-2.1195e-04, 2.8864e-05, -6.8951e-04, ..., -1.3793e-04, + 4.5039e-06, 5.5254e-05], + [ 2.2221e-04, 3.9935e-04, 8.4591e-04, ..., 2.8181e-04, + 5.4799e-06, 2.1935e-04], + ..., + [ 1.0276e-04, 1.0359e-04, -8.1873e-04, ..., 3.9315e-04, + 1.4596e-05, 2.7847e-04], + [ 1.2159e-04, -3.7265e-04, -1.3180e-03, ..., -7.7057e-04, + 1.3754e-05, -7.2098e-04], + [ 7.0953e-04, 4.3201e-04, 1.6842e-03, ..., 2.0866e-03, + 2.0489e-06, 2.6646e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0297, 0.0103, 0.0177, 0.0230, 0.0277, -0.0098, -0.0028, -0.0055, + 0.0140, -0.0274], device='cuda:0'), grad: tensor([-0.0023, -0.0005, 0.0013, -0.0064, 0.0008, 0.0054, -0.0024, -0.0005, + -0.0020, 0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 218.01, cls_loss 0.0675 cls_loss_mapping 0.0801 cls_loss_causal 0.9360 re_mapping 0.0358 re_causal 0.0889 /// teacc 97.54 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0488, 0.0673, -0.0170, ..., 0.0094, 0.0432, 0.0272], + [ 0.0375, -0.0460, 0.0019, ..., -0.0413, -0.0041, -0.0142], + [-0.0022, -0.0282, 0.0262, ..., -0.0228, 0.0136, -0.0526], + ..., + [-0.0337, -0.0338, 0.0312, ..., -0.0101, -0.0454, -0.0416], + [-0.0142, -0.0217, -0.0093, ..., 0.0039, -0.0511, -0.0114], + [-0.0051, 0.0164, 0.0208, ..., 0.0457, -0.0191, -0.0317]], + device='cuda:0'), grad: tensor([[ 4.0054e-05, -2.5444e-03, -5.3644e-04, ..., -2.3353e-04, + 1.4208e-05, -5.0211e-04], + [ 1.5759e-04, 5.4449e-05, 1.3809e-03, ..., 6.0409e-05, + 1.1759e-03, 4.4107e-05], + [-2.3067e-04, 2.7847e-04, -1.4620e-03, ..., 8.4519e-05, + -1.5059e-03, 8.9705e-05], + ..., + [ 5.1349e-05, 3.5858e-04, -5.0316e-03, ..., 1.4830e-04, + 2.8014e-04, 1.3185e-04], + [ 1.4496e-04, 5.0545e-04, 8.1968e-04, ..., 1.6844e-04, + 1.3244e-04, 3.4988e-05], + [ 5.7638e-05, 6.4325e-04, 4.1389e-03, ..., 2.4354e-04, + 1.4770e-04, 2.0027e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0291, 0.0110, 0.0174, 0.0228, 0.0275, -0.0095, -0.0023, -0.0054, + 0.0138, -0.0273], device='cuda:0'), grad: tensor([-0.0014, 0.0037, -0.0040, 0.0006, -0.0026, 0.0008, 0.0004, -0.0034, + 0.0012, 0.0047], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 218.49, cls_loss 0.0640 cls_loss_mapping 0.0774 cls_loss_causal 0.9610 re_mapping 0.0330 re_causal 0.0837 /// teacc 97.48 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0499, 0.0685, -0.0172, ..., 0.0091, 0.0432, 0.0271], + [ 0.0379, -0.0471, 0.0023, ..., -0.0414, -0.0045, -0.0148], + [-0.0023, -0.0293, 0.0262, ..., -0.0234, 0.0143, -0.0529], + ..., + [-0.0350, -0.0332, 0.0315, ..., -0.0106, -0.0458, -0.0426], + [-0.0151, -0.0216, -0.0089, ..., 0.0045, -0.0519, -0.0114], + [-0.0054, 0.0157, 0.0212, ..., 0.0463, -0.0195, -0.0326]], + device='cuda:0'), grad: tensor([[ 1.3876e-04, -1.3227e-03, -4.2820e-04, ..., 2.3916e-05, + -1.6415e-04, 1.0777e-04], + [ 1.3494e-04, 5.9515e-05, 2.7847e-04, ..., 1.6212e-04, + 1.6354e-06, 1.2445e-04], + [ 2.0301e-04, 6.3038e-04, 7.2145e-04, ..., 3.1805e-04, + 8.7559e-05, 1.5664e-04], + ..., + [ 3.3498e-04, 8.8394e-05, 8.3327e-05, ..., 1.9705e-04, + 1.3739e-05, 3.5214e-04], + [-3.9482e-04, -1.5843e-04, -1.4668e-03, ..., -7.5865e-04, + -7.6115e-05, -1.9550e-04], + [ 1.8549e-04, 2.2709e-04, -4.9561e-05, ..., -2.8610e-04, + 5.2124e-05, 2.1791e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0289, 0.0110, 0.0177, 0.0232, 0.0276, -0.0094, -0.0025, -0.0054, + 0.0135, -0.0276], device='cuda:0'), grad: tensor([-1.0996e-03, 4.5156e-04, 1.2741e-03, 1.2741e-03, -7.0855e-06, + -1.6642e-03, 3.8719e-04, 6.7091e-04, -1.9293e-03, 6.4039e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 219.32, cls_loss 0.0516 cls_loss_mapping 0.0595 cls_loss_causal 0.8927 re_mapping 0.0324 re_causal 0.0813 /// teacc 98.13 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0504, 0.0698, -0.0175, ..., 0.0089, 0.0430, 0.0271], + [ 0.0381, -0.0478, 0.0021, ..., -0.0418, -0.0049, -0.0154], + [-0.0032, -0.0305, 0.0260, ..., -0.0237, 0.0150, -0.0537], + ..., + [-0.0353, -0.0333, 0.0318, ..., -0.0106, -0.0462, -0.0439], + [-0.0152, -0.0212, -0.0082, ..., 0.0045, -0.0520, -0.0108], + [-0.0056, 0.0154, 0.0219, ..., 0.0471, -0.0198, -0.0332]], + device='cuda:0'), grad: tensor([[ 9.5904e-05, -1.1024e-03, -3.8719e-04, ..., 2.2516e-05, + 1.1230e-04, -4.5967e-04], + [-5.7411e-03, 1.8537e-04, -1.4725e-02, ..., -1.2703e-03, + -3.0937e-03, 1.8072e-04], + [ 3.4833e-04, 3.4499e-04, 6.0320e-04, ..., 2.2733e-04, + 7.8440e-05, 4.0960e-04], + ..., + [ 2.0194e-04, -5.5647e-04, -1.1377e-03, ..., 8.1301e-05, + -2.7490e-04, -4.8876e-04], + [ 4.4518e-03, 4.4084e-04, 1.1353e-02, ..., 5.3883e-04, + 2.7580e-03, 1.1730e-04], + [ 1.7309e-04, 2.7037e-04, 5.8460e-04, ..., 7.1600e-06, + 1.1957e-04, 3.1424e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0290, 0.0103, 0.0173, 0.0231, 0.0277, -0.0096, -0.0025, -0.0052, + 0.0141, -0.0274], device='cuda:0'), grad: tensor([-0.0008, -0.0181, 0.0017, -0.0009, 0.0022, 0.0007, 0.0010, -0.0022, + 0.0151, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 218.42, cls_loss 0.0486 cls_loss_mapping 0.0592 cls_loss_causal 0.8967 re_mapping 0.0309 re_causal 0.0828 /// teacc 97.85 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0511, 0.0710, -0.0173, ..., 0.0087, 0.0430, 0.0272], + [ 0.0386, -0.0481, 0.0027, ..., -0.0415, -0.0051, -0.0159], + [-0.0040, -0.0315, 0.0259, ..., -0.0239, 0.0156, -0.0545], + ..., + [-0.0358, -0.0336, 0.0322, ..., -0.0112, -0.0467, -0.0443], + [-0.0154, -0.0213, -0.0080, ..., 0.0048, -0.0526, -0.0105], + [-0.0063, 0.0153, 0.0223, ..., 0.0476, -0.0203, -0.0344]], + device='cuda:0'), grad: tensor([[ 3.0249e-05, -2.1887e-04, -1.2028e-04, ..., 2.9355e-05, + 3.9488e-05, 5.3883e-05], + [-1.1196e-03, 8.6203e-06, -6.5374e-04, ..., 4.7743e-05, + -4.9829e-05, -8.1062e-06], + [ 1.7536e-04, 7.1526e-05, 3.7265e-04, ..., 1.0014e-04, + 2.8700e-05, 1.0127e-04], + ..., + [ 1.2624e-04, 2.3127e-05, 8.3542e-04, ..., 2.4128e-04, + 1.0669e-05, 6.0511e-04], + [-1.3518e-04, -6.2275e-04, -3.2005e-03, ..., -1.2865e-03, + 9.6500e-05, -3.4904e-03], + [ 7.0572e-05, 7.0214e-05, 9.2125e-04, ..., 2.5415e-04, + 1.5289e-05, 6.4611e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0289, 0.0104, 0.0173, 0.0232, 0.0273, -0.0097, -0.0023, -0.0051, + 0.0143, -0.0276], device='cuda:0'), grad: tensor([ 8.0645e-05, -7.6818e-04, 6.0511e-04, 1.4706e-03, -7.5483e-04, + 3.0956e-03, -1.9610e-05, 8.5020e-04, -7.1449e-03, 2.5902e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 218.95, cls_loss 0.0541 cls_loss_mapping 0.0652 cls_loss_causal 0.8551 re_mapping 0.0311 re_causal 0.0784 /// teacc 98.21 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0519, 0.0724, -0.0178, ..., 0.0085, 0.0431, 0.0271], + [ 0.0389, -0.0500, 0.0027, ..., -0.0422, -0.0055, -0.0161], + [-0.0051, -0.0315, 0.0260, ..., -0.0230, 0.0160, -0.0555], + ..., + [-0.0368, -0.0339, 0.0326, ..., -0.0114, -0.0469, -0.0455], + [-0.0155, -0.0215, -0.0077, ..., 0.0049, -0.0532, -0.0102], + [-0.0053, 0.0148, 0.0231, ..., 0.0483, -0.0206, -0.0346]], + device='cuda:0'), grad: tensor([[ 2.5654e-04, -1.1444e-05, -2.0698e-05, ..., 4.4942e-05, + -3.9250e-05, 1.5116e-04], + [ 3.6091e-05, 1.5521e-04, 4.5806e-05, ..., 7.8157e-06, + 8.0347e-05, 1.2338e-04], + [ 2.2864e-04, 7.5579e-04, 2.8896e-04, ..., 8.1301e-05, + -6.8665e-05, 1.4365e-04], + ..., + [ 1.3340e-04, -1.7347e-03, -1.8320e-03, ..., 6.8806e-06, + 1.9491e-05, 9.4235e-05], + [ 8.0872e-04, 2.0206e-04, 3.7998e-05, ..., -2.1070e-05, + 4.1544e-05, 5.5265e-04], + [ 4.1246e-04, 2.2995e-04, 3.3932e-03, ..., 2.2964e-03, + 6.2995e-06, 2.9469e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0288, 0.0105, 0.0175, 0.0228, 0.0273, -0.0101, -0.0024, -0.0047, + 0.0142, -0.0272], device='cuda:0'), grad: tensor([ 0.0014, 0.0008, 0.0021, -0.0243, -0.0024, 0.0274, -0.0042, -0.0071, + 0.0011, 0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 218.70, cls_loss 0.0457 cls_loss_mapping 0.0537 cls_loss_causal 0.8787 re_mapping 0.0292 re_causal 0.0796 /// teacc 98.08 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0524, 0.0739, -0.0175, ..., 0.0085, 0.0436, 0.0272], + [ 0.0392, -0.0502, 0.0026, ..., -0.0423, -0.0060, -0.0166], + [-0.0055, -0.0327, 0.0263, ..., -0.0231, 0.0169, -0.0561], + ..., + [-0.0370, -0.0340, 0.0329, ..., -0.0115, -0.0475, -0.0454], + [-0.0156, -0.0217, -0.0077, ..., 0.0050, -0.0540, -0.0099], + [-0.0051, 0.0144, 0.0233, ..., 0.0488, -0.0212, -0.0347]], + device='cuda:0'), grad: tensor([[ 1.0290e-03, -2.5105e-04, 1.1522e-04, ..., 2.6643e-05, + 3.5954e-04, -1.7196e-05], + [-2.3670e-03, 2.5019e-05, -1.8959e-03, ..., 8.0317e-06, + 1.9431e-05, 2.8402e-05], + [ 1.4076e-03, 2.4509e-04, 1.3342e-03, ..., 6.0529e-05, + 2.0552e-04, 1.5378e-04], + ..., + [ 1.1629e-04, 1.0407e-04, -4.8065e-04, ..., 5.8353e-05, + 9.0241e-05, -1.9327e-05], + [ 5.9128e-04, 8.6069e-05, -7.1347e-05, ..., -1.0985e-04, + 1.0777e-04, 4.9859e-05], + [ 2.2864e-04, 6.1750e-04, 5.7966e-05, ..., -1.2422e-04, + 2.2781e-04, 2.5511e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0290, 0.0100, 0.0180, 0.0227, 0.0273, -0.0097, -0.0025, -0.0046, + 0.0139, -0.0273], device='cuda:0'), grad: tensor([-5.7316e-04, -3.4695e-03, 3.0422e-03, 1.1520e-03, -1.0414e-03, + 7.2479e-04, -1.9369e-03, -3.3408e-05, 3.4356e-04, 1.7948e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 218.25, cls_loss 0.0564 cls_loss_mapping 0.0682 cls_loss_causal 0.8891 re_mapping 0.0278 re_causal 0.0722 /// teacc 98.03 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0535, 0.0753, -0.0178, ..., 0.0083, 0.0438, 0.0271], + [ 0.0399, -0.0492, 0.0032, ..., -0.0424, -0.0068, -0.0166], + [-0.0065, -0.0349, 0.0259, ..., -0.0234, 0.0178, -0.0578], + ..., + [-0.0381, -0.0345, 0.0328, ..., -0.0121, -0.0482, -0.0462], + [-0.0166, -0.0226, -0.0068, ..., 0.0054, -0.0548, -0.0103], + [-0.0049, 0.0148, 0.0243, ..., 0.0496, -0.0218, -0.0352]], + device='cuda:0'), grad: tensor([[ 7.1955e-04, 4.9067e-04, 2.3878e-04, ..., 4.7237e-05, + 9.5367e-07, 5.8126e-04], + [ 8.4579e-05, 1.0121e-04, 3.0918e-03, ..., 5.1916e-05, + 2.8458e-03, 2.0289e-04], + [ 9.5963e-05, 9.9719e-05, -3.7880e-03, ..., 1.9491e-04, + -4.2534e-03, 3.4547e-04], + ..., + [ 4.3780e-05, 3.4660e-05, -2.2590e-04, ..., -6.7890e-05, + 2.6727e-04, 6.3360e-05], + [-1.0109e-02, -5.0049e-03, -5.9776e-03, ..., -3.7727e-03, + 1.8513e-04, -1.1063e-02], + [ 4.5586e-04, 3.1471e-04, 9.9087e-04, ..., 6.1989e-05, + 7.0333e-05, 5.8746e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0290, 0.0100, 0.0177, 0.0226, 0.0275, -0.0092, -0.0027, -0.0049, + 0.0139, -0.0271], device='cuda:0'), grad: tensor([ 0.0011, 0.0067, -0.0094, 0.0011, -0.0011, 0.0099, -0.0008, 0.0002, + -0.0109, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 218.09, cls_loss 0.0450 cls_loss_mapping 0.0528 cls_loss_causal 0.8741 re_mapping 0.0274 re_causal 0.0748 /// teacc 98.20 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0540, 0.0763, -0.0182, ..., 0.0081, 0.0441, 0.0271], + [ 0.0410, -0.0495, 0.0044, ..., -0.0417, -0.0079, -0.0173], + [-0.0077, -0.0357, 0.0254, ..., -0.0241, 0.0190, -0.0588], + ..., + [-0.0391, -0.0347, 0.0332, ..., -0.0124, -0.0491, -0.0461], + [-0.0166, -0.0226, -0.0067, ..., 0.0057, -0.0553, -0.0097], + [-0.0057, 0.0146, 0.0245, ..., 0.0500, -0.0220, -0.0363]], + device='cuda:0'), grad: tensor([[-2.2087e-03, -6.0959e-03, 7.6830e-05, ..., -6.7377e-04, + 6.2585e-05, -2.9736e-03], + [-5.0306e-04, -1.7539e-05, -3.8743e-04, ..., -1.6361e-05, + 4.4942e-05, -3.2097e-05], + [-2.9278e-04, 3.1710e-05, 8.4496e-04, ..., 2.3052e-05, + -1.0347e-03, 4.9680e-05], + ..., + [ 1.2946e-04, 4.3601e-05, -2.3537e-03, ..., 2.9802e-05, + -5.0974e-04, 3.9876e-05], + [ 1.9693e-04, -1.5736e-04, 1.3447e-04, ..., -2.2388e-04, + 4.1294e-04, -1.9038e-04], + [ 5.1355e-04, 1.2302e-03, 2.7657e-04, ..., 1.5986e-04, + 1.4579e-04, 7.3624e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0289, 0.0100, 0.0177, 0.0230, 0.0275, -0.0090, -0.0030, -0.0050, + 0.0141, -0.0276], device='cuda:0'), grad: tensor([-0.0051, -0.0003, -0.0042, 0.0009, 0.0021, 0.0046, 0.0015, -0.0031, + 0.0017, 0.0020], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 218.36, cls_loss 0.0403 cls_loss_mapping 0.0461 cls_loss_causal 0.8619 re_mapping 0.0257 re_causal 0.0708 /// teacc 98.18 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0543, 0.0773, -0.0178, ..., 0.0083, 0.0441, 0.0269], + [ 0.0410, -0.0503, 0.0041, ..., -0.0420, -0.0082, -0.0183], + [-0.0083, -0.0367, 0.0254, ..., -0.0244, 0.0199, -0.0595], + ..., + [-0.0397, -0.0350, 0.0337, ..., -0.0128, -0.0496, -0.0468], + [-0.0169, -0.0225, -0.0065, ..., 0.0059, -0.0560, -0.0094], + [-0.0057, 0.0142, 0.0251, ..., 0.0508, -0.0223, -0.0364]], + device='cuda:0'), grad: tensor([[ 1.8522e-05, 4.8310e-05, 6.9976e-05, ..., 2.4796e-05, + 5.4717e-05, 2.9817e-05], + [-1.5450e-04, 1.2413e-05, 3.5286e-03, ..., 1.4067e-03, + 4.6909e-05, 1.7449e-05], + [ 2.2367e-05, 1.9407e-04, 1.0452e-03, ..., 5.3167e-04, + 4.4703e-05, 1.0669e-04], + ..., + [ 5.2929e-05, -1.4830e-04, -9.4986e-04, ..., 1.6010e-04, + 3.4332e-05, 1.6594e-04], + [ 4.3690e-05, -3.1376e-04, -5.1689e-03, ..., -2.2488e-03, + -3.2806e-04, 7.0810e-05], + [ 3.1948e-05, 1.5354e-04, 4.6873e-04, ..., -1.2577e-05, + 4.2856e-05, 1.3459e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0286, 0.0098, 0.0177, 0.0232, 0.0270, -0.0091, -0.0023, -0.0048, + 0.0137, -0.0272], device='cuda:0'), grad: tensor([ 0.0003, 0.0035, 0.0010, -0.0003, 0.0014, 0.0004, -0.0002, -0.0019, + -0.0053, 0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 218.02, cls_loss 0.0406 cls_loss_mapping 0.0500 cls_loss_causal 0.8431 re_mapping 0.0253 re_causal 0.0686 /// teacc 98.18 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0551, 0.0783, -0.0182, ..., 0.0084, 0.0442, 0.0268], + [ 0.0419, -0.0492, 0.0041, ..., -0.0422, -0.0080, -0.0181], + [-0.0088, -0.0377, 0.0250, ..., -0.0250, 0.0207, -0.0605], + ..., + [-0.0405, -0.0354, 0.0343, ..., -0.0131, -0.0496, -0.0471], + [-0.0176, -0.0227, -0.0062, ..., 0.0061, -0.0570, -0.0095], + [-0.0058, 0.0141, 0.0258, ..., 0.0514, -0.0227, -0.0371]], + device='cuda:0'), grad: tensor([[-1.2743e-04, -1.7452e-03, -2.3293e-04, ..., -6.4913e-07, + -9.7942e-04, -1.3742e-03], + [-1.0055e-04, 1.5467e-05, -1.6785e-04, ..., -5.4359e-05, + 2.4289e-05, 2.1815e-05], + [ 1.9446e-05, 6.5386e-05, 2.6321e-04, ..., 1.2851e-04, + 3.8683e-05, 5.7757e-05], + ..., + [ 6.4790e-05, 3.0637e-05, -4.0102e-04, ..., 8.0168e-06, + 2.2486e-05, 3.1143e-05], + [ 2.0707e-04, 1.7607e-04, 1.3188e-06, ..., -1.1468e-04, + 1.1581e-04, 1.9383e-04], + [ 9.4950e-05, 6.0415e-04, 2.2876e-04, ..., -3.7700e-05, + 4.2295e-04, 5.3310e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0285, 0.0098, 0.0175, 0.0228, 0.0270, -0.0089, -0.0024, -0.0040, + 0.0134, -0.0271], device='cuda:0'), grad: tensor([-3.8033e-03, -5.8711e-05, 5.7030e-04, 3.4642e-04, -1.0586e-03, + 4.3178e-04, 1.8244e-03, -2.1565e-04, 2.0909e-04, 1.7529e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 218.56, cls_loss 0.0396 cls_loss_mapping 0.0443 cls_loss_causal 0.8149 re_mapping 0.0256 re_causal 0.0662 /// teacc 98.11 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0556, 0.0793, -0.0183, ..., 0.0079, 0.0444, 0.0270], + [ 0.0428, -0.0495, 0.0047, ..., -0.0423, -0.0083, -0.0183], + [-0.0090, -0.0387, 0.0250, ..., -0.0251, 0.0214, -0.0614], + ..., + [-0.0413, -0.0356, 0.0343, ..., -0.0135, -0.0506, -0.0473], + [-0.0182, -0.0227, -0.0061, ..., 0.0062, -0.0568, -0.0097], + [-0.0062, 0.0140, 0.0261, ..., 0.0519, -0.0228, -0.0380]], + device='cuda:0'), grad: tensor([[ 5.4836e-04, -6.9380e-05, 5.8603e-04, ..., 1.8466e-04, + 3.2210e-04, 1.8024e-04], + [-7.8201e-03, -2.8992e-03, -9.2850e-03, ..., -2.5024e-03, + -5.0964e-03, -2.3689e-03], + [ 1.5903e-04, 1.0371e-04, 3.9816e-04, ..., 7.4267e-05, + 8.0228e-05, 1.2980e-03], + ..., + [ 1.1802e-04, 4.1187e-05, -5.1308e-04, ..., 3.0637e-04, + 6.1393e-05, 6.4135e-05], + [ 5.1994e-03, 1.9703e-03, 6.3477e-03, ..., 1.6966e-03, + 3.4313e-03, 1.8520e-03], + [ 1.4138e-04, 8.6606e-05, -2.3377e-04, ..., -3.2282e-04, + 6.2287e-05, 1.5640e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0286, 0.0101, 0.0179, 0.0226, 0.0274, -0.0084, -0.0028, -0.0045, + 0.0132, -0.0273], device='cuda:0'), grad: tensor([ 0.0008, -0.0135, 0.0024, -0.0023, -0.0003, 0.0006, 0.0025, -0.0002, + 0.0096, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 218.84, cls_loss 0.0379 cls_loss_mapping 0.0475 cls_loss_causal 0.8359 re_mapping 0.0240 re_causal 0.0669 /// teacc 98.35 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0560, 0.0800, -0.0184, ..., 0.0080, 0.0445, 0.0269], + [ 0.0436, -0.0495, 0.0054, ..., -0.0426, -0.0087, -0.0186], + [-0.0097, -0.0395, 0.0246, ..., -0.0255, 0.0219, -0.0621], + ..., + [-0.0420, -0.0358, 0.0348, ..., -0.0135, -0.0512, -0.0477], + [-0.0185, -0.0225, -0.0053, ..., 0.0066, -0.0571, -0.0095], + [-0.0065, 0.0137, 0.0262, ..., 0.0524, -0.0228, -0.0385]], + device='cuda:0'), grad: tensor([[ 5.1880e-04, -2.7132e-04, 8.3625e-05, ..., 1.4953e-05, + 2.1732e-04, 2.9564e-04], + [ 6.1302e-03, 4.1991e-05, 1.4009e-03, ..., 3.5554e-05, + 1.8167e-03, 3.2215e-03], + [ 6.1226e-04, 9.8825e-05, 3.0398e-04, ..., 2.2575e-05, + 1.5235e-04, 3.6454e-04], + ..., + [ 1.7071e-04, 1.3255e-05, -3.6025e-04, ..., 4.8339e-05, + 1.5366e-04, 8.1122e-05], + [-1.5343e-02, -1.7428e-04, -3.4294e-03, ..., 1.9145e-04, + -4.3526e-03, -8.2321e-03], + [ 2.8896e-04, 9.4175e-05, -5.9986e-04, ..., -7.0620e-04, + 7.7784e-05, 1.3793e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0287, 0.0100, 0.0176, 0.0225, 0.0276, -0.0087, -0.0027, -0.0041, + 0.0135, -0.0276], device='cuda:0'), grad: tensor([ 0.0007, 0.0102, 0.0011, 0.0011, -0.0040, 0.0007, 0.0113, 0.0005, + -0.0220, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 218.43, cls_loss 0.0349 cls_loss_mapping 0.0426 cls_loss_causal 0.8102 re_mapping 0.0246 re_causal 0.0662 /// teacc 98.16 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0564, 0.0812, -0.0185, ..., 0.0077, 0.0448, 0.0266], + [ 0.0441, -0.0486, 0.0062, ..., -0.0420, -0.0088, -0.0185], + [-0.0096, -0.0394, 0.0245, ..., -0.0254, 0.0221, -0.0627], + ..., + [-0.0430, -0.0361, 0.0353, ..., -0.0137, -0.0515, -0.0478], + [-0.0191, -0.0234, -0.0054, ..., 0.0068, -0.0571, -0.0097], + [-0.0068, 0.0134, 0.0265, ..., 0.0530, -0.0232, -0.0390]], + device='cuda:0'), grad: tensor([[ 3.1382e-05, -8.3637e-04, -3.6669e-04, ..., -4.2945e-05, + -2.1309e-06, 1.4417e-05], + [-1.1969e-04, 9.8348e-06, -7.7844e-05, ..., 1.8269e-05, + -1.8515e-06, 5.5254e-05], + [ 1.0443e-04, 6.3539e-05, 1.0830e-04, ..., 1.0639e-05, + 1.8150e-05, 6.8188e-05], + ..., + [ 6.9857e-05, 1.7524e-05, 4.5151e-05, ..., 2.5392e-04, + -1.1273e-05, 1.0923e-05], + [ 3.3236e-04, 1.4627e-04, 2.8300e-04, ..., 3.8296e-05, + 1.8433e-05, 1.9252e-04], + [ 3.4523e-04, 8.9169e-05, -5.0068e-04, ..., -3.9554e-04, + 3.2224e-06, 2.2495e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0288, 0.0103, 0.0178, 0.0225, 0.0277, -0.0083, -0.0035, -0.0038, + 0.0132, -0.0279], device='cuda:0'), grad: tensor([-7.0906e-04, -6.5565e-05, 2.1303e-04, 1.9714e-02, 1.4865e-04, + -2.1729e-02, 1.8749e-03, 1.7726e-04, 5.1832e-04, -1.4305e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 218.47, cls_loss 0.0360 cls_loss_mapping 0.0428 cls_loss_causal 0.8235 re_mapping 0.0239 re_causal 0.0645 /// teacc 98.34 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0568, 0.0820, -0.0185, ..., 0.0075, 0.0446, 0.0264], + [ 0.0447, -0.0489, 0.0062, ..., -0.0421, -0.0090, -0.0188], + [-0.0100, -0.0398, 0.0250, ..., -0.0257, 0.0229, -0.0634], + ..., + [-0.0441, -0.0362, 0.0352, ..., -0.0142, -0.0521, -0.0485], + [-0.0192, -0.0232, -0.0049, ..., 0.0072, -0.0571, -0.0094], + [-0.0069, 0.0129, 0.0271, ..., 0.0536, -0.0234, -0.0393]], + device='cuda:0'), grad: tensor([[ 1.1712e-04, -1.4377e-04, -4.0859e-05, ..., 4.9993e-06, + 1.8865e-05, 7.7784e-06], + [-3.3069e-04, 1.5974e-05, -7.2145e-04, ..., 3.3174e-06, + 7.1704e-05, 9.2089e-06], + [ 2.0611e-04, 6.4135e-05, 6.5279e-04, ..., -1.9312e-05, + -2.7227e-04, 2.0728e-05], + ..., + [ 7.9989e-05, 1.0528e-05, -1.1578e-03, ..., 2.0936e-05, + 4.5419e-05, 8.0094e-06], + [ 4.4703e-04, 1.2946e-04, 1.9228e-04, ..., 8.0228e-05, + 1.7321e-04, 1.0306e-04], + [ 9.6381e-05, 3.6240e-05, -1.4514e-05, ..., -8.2910e-05, + 1.6361e-05, 4.3422e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0292, 0.0103, 0.0183, 0.0224, 0.0273, -0.0083, -0.0035, -0.0043, + 0.0135, -0.0280], device='cuda:0'), grad: tensor([ 0.0003, -0.0009, 0.0008, 0.0007, 0.0013, 0.0029, -0.0053, -0.0025, + 0.0022, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 218.14, cls_loss 0.0385 cls_loss_mapping 0.0422 cls_loss_causal 0.8092 re_mapping 0.0237 re_causal 0.0608 /// teacc 97.89 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0571, 0.0832, -0.0185, ..., 0.0073, 0.0446, 0.0265], + [ 0.0451, -0.0497, 0.0066, ..., -0.0423, -0.0096, -0.0187], + [-0.0107, -0.0409, 0.0245, ..., -0.0257, 0.0238, -0.0643], + ..., + [-0.0452, -0.0366, 0.0354, ..., -0.0146, -0.0526, -0.0496], + [-0.0190, -0.0225, -0.0046, ..., 0.0070, -0.0573, -0.0092], + [-0.0073, 0.0125, 0.0275, ..., 0.0542, -0.0236, -0.0399]], + device='cuda:0'), grad: tensor([[ 5.8264e-05, -1.2617e-03, -1.5926e-04, ..., 7.6964e-06, + -3.2067e-04, -4.4703e-05], + [ 3.0175e-05, 1.5700e-04, 2.1744e-04, ..., 1.3307e-05, + 4.1604e-05, 1.0300e-04], + [ 2.9281e-05, 5.1355e-04, 1.7345e-04, ..., 2.9832e-05, + 6.6161e-05, 1.7786e-04], + ..., + [ 2.0787e-05, 6.4254e-05, -9.1791e-04, ..., -2.5481e-05, + 1.3083e-05, -6.7115e-05], + [ 6.2704e-05, -1.9729e-04, -2.5058e-04, ..., -1.3947e-04, + 1.2481e-04, -5.9605e-04], + [ 3.2276e-05, 1.1516e-04, 1.7583e-04, ..., -1.0604e-04, + 1.9848e-05, 1.0979e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0293, 0.0104, 0.0179, 0.0222, 0.0278, -0.0077, -0.0036, -0.0045, + 0.0135, -0.0282], device='cuda:0'), grad: tensor([-0.0008, 0.0005, 0.0004, 0.0016, -0.0001, -0.0009, 0.0010, -0.0016, + -0.0005, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 218.18, cls_loss 0.0303 cls_loss_mapping 0.0340 cls_loss_causal 0.7888 re_mapping 0.0231 re_causal 0.0616 /// teacc 98.31 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0578, 0.0847, -0.0187, ..., 0.0070, 0.0448, 0.0266], + [ 0.0453, -0.0502, 0.0064, ..., -0.0424, -0.0097, -0.0193], + [-0.0107, -0.0418, 0.0245, ..., -0.0260, 0.0243, -0.0647], + ..., + [-0.0457, -0.0375, 0.0355, ..., -0.0151, -0.0529, -0.0500], + [-0.0193, -0.0221, -0.0040, ..., 0.0073, -0.0579, -0.0089], + [-0.0076, 0.0126, 0.0276, ..., 0.0549, -0.0239, -0.0404]], + device='cuda:0'), grad: tensor([[ 6.8486e-05, -3.7694e-04, -3.7104e-05, ..., 2.0266e-06, + 4.5039e-06, -7.3671e-05], + [-8.9073e-04, 1.3821e-05, -3.9554e-04, ..., -4.4167e-05, + 3.8520e-06, -2.0385e-04], + [ 2.0757e-05, 6.6698e-05, 4.4525e-05, ..., -6.3360e-05, + -1.3745e-04, 3.7670e-05], + ..., + [ 5.4806e-05, 2.9087e-05, 8.2314e-05, ..., 9.9242e-05, + 3.5446e-06, 2.6897e-05], + [ 6.1989e-05, -2.2560e-05, 5.2840e-05, ..., 1.0198e-04, + 1.1712e-04, -7.5698e-05], + [ 1.2922e-04, 7.3195e-05, -6.0701e-04, ..., -6.2609e-04, + 1.7611e-06, 6.5982e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0297, 0.0103, 0.0177, 0.0221, 0.0275, -0.0076, -0.0035, -0.0044, + 0.0136, -0.0283], device='cuda:0'), grad: tensor([-1.6856e-04, -1.3132e-03, -1.0407e-04, 9.8801e-04, 1.0777e-03, + -1.6689e-05, 1.0532e-04, 2.4021e-04, 3.1996e-04, -1.1292e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 218.26, cls_loss 0.0316 cls_loss_mapping 0.0373 cls_loss_causal 0.7781 re_mapping 0.0221 re_causal 0.0598 /// teacc 98.20 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0585, 0.0852, -0.0193, ..., 0.0068, 0.0448, 0.0266], + [ 0.0463, -0.0492, 0.0067, ..., -0.0426, -0.0098, -0.0196], + [-0.0110, -0.0432, 0.0243, ..., -0.0261, 0.0251, -0.0656], + ..., + [-0.0461, -0.0375, 0.0358, ..., -0.0154, -0.0537, -0.0505], + [-0.0200, -0.0221, -0.0038, ..., 0.0075, -0.0583, -0.0085], + [-0.0080, 0.0135, 0.0284, ..., 0.0557, -0.0238, -0.0411]], + device='cuda:0'), grad: tensor([[ 1.8239e-05, -1.7529e-03, -1.9178e-05, ..., 2.6673e-06, + 1.8343e-05, -2.5213e-05], + [ 1.3113e-04, 6.4611e-05, 1.4651e-04, ..., 6.7130e-06, + 3.2377e-04, 5.6326e-05], + [ 6.3539e-05, 3.4547e-04, -1.4744e-03, ..., 1.0207e-05, + -3.9268e-04, -6.0940e-04], + ..., + [ 1.1869e-05, 9.6321e-05, -7.7009e-05, ..., 5.0552e-06, + 2.5228e-05, 2.6894e-04], + [ 8.3327e-05, 1.9479e-04, 2.5916e-04, ..., 6.6876e-05, + 5.3078e-05, 1.7726e-04], + [ 2.2590e-05, 4.3797e-04, 1.4238e-03, ..., 1.0710e-03, + 6.0685e-06, 2.6302e-03]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0294, 0.0105, 0.0175, 0.0223, 0.0277, -0.0075, -0.0043, -0.0040, + 0.0133, -0.0281], device='cuda:0'), grad: tensor([-0.0029, 0.0013, -0.0023, -0.0027, -0.0011, 0.0009, -0.0002, 0.0010, + 0.0010, 0.0049], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 36---------------------------------------------------- +epoch 36, time 218.97, cls_loss 0.0299 cls_loss_mapping 0.0332 cls_loss_causal 0.7722 re_mapping 0.0215 re_causal 0.0585 /// teacc 98.44 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0592, 0.0858, -0.0194, ..., 0.0067, 0.0447, 0.0262], + [ 0.0471, -0.0495, 0.0071, ..., -0.0426, -0.0099, -0.0198], + [-0.0116, -0.0443, 0.0244, ..., -0.0263, 0.0256, -0.0664], + ..., + [-0.0468, -0.0379, 0.0360, ..., -0.0152, -0.0540, -0.0516], + [-0.0205, -0.0221, -0.0036, ..., 0.0076, -0.0587, -0.0084], + [-0.0082, 0.0131, 0.0288, ..., 0.0560, -0.0238, -0.0416]], + device='cuda:0'), grad: tensor([[ 3.4571e-06, -4.1509e-04, 3.1018e-04, ..., 2.2069e-05, + -1.2034e-04, -6.3479e-05], + [ 6.8657e-06, 1.3039e-05, 1.0073e-04, ..., 2.4199e-05, + 1.7926e-05, 5.7518e-06], + [ 1.2554e-06, 6.2346e-05, 1.4973e-04, ..., 9.6709e-06, + -5.7131e-05, 2.6479e-05], + ..., + [ 2.9672e-06, 1.8191e-04, 5.8556e-04, ..., 8.2731e-05, + 8.7097e-06, 7.5065e-06], + [ 1.4700e-05, 8.7097e-06, 6.5267e-05, ..., 1.6153e-05, + 2.2054e-05, -8.7246e-06], + [-2.3022e-05, -2.9087e-04, -1.8396e-03, ..., -2.4629e-04, + 6.8881e-06, 1.4208e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0291, 0.0110, 0.0174, 0.0226, 0.0279, -0.0078, -0.0037, -0.0043, + 0.0131, -0.0282], device='cuda:0'), grad: tensor([ 1.3173e-04, 1.6165e-04, 1.3876e-04, 3.1447e-04, 3.4547e-04, + 3.1829e-04, 5.1409e-05, 9.3985e-04, 1.3328e-04, -2.5330e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 218.01, cls_loss 0.0329 cls_loss_mapping 0.0346 cls_loss_causal 0.7827 re_mapping 0.0211 re_causal 0.0581 /// teacc 98.35 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0599, 0.0868, -0.0193, ..., 0.0065, 0.0446, 0.0260], + [ 0.0478, -0.0497, 0.0075, ..., -0.0428, -0.0095, -0.0202], + [-0.0127, -0.0454, 0.0240, ..., -0.0264, 0.0258, -0.0677], + ..., + [-0.0476, -0.0377, 0.0361, ..., -0.0155, -0.0544, -0.0526], + [-0.0208, -0.0221, -0.0034, ..., 0.0075, -0.0590, -0.0079], + [-0.0084, 0.0124, 0.0294, ..., 0.0564, -0.0240, -0.0425]], + device='cuda:0'), grad: tensor([[ 4.0817e-04, 5.4693e-04, 5.6505e-05, ..., 7.3202e-06, + 1.0914e-04, 6.1083e-04], + [ 1.1706e-04, 1.6010e-04, 1.6367e-04, ..., 1.8567e-05, + 2.5105e-04, 1.3983e-04], + [-9.6679e-05, 1.9193e-04, -7.9441e-04, ..., 2.7433e-05, + 6.9523e-04, 2.1958e-04], + ..., + [ 4.2582e-04, 4.3631e-05, 5.0831e-04, ..., 2.8282e-05, + 5.9795e-04, 2.8300e-04], + [ 5.3596e-04, -8.9526e-05, -4.0317e-04, ..., -7.1526e-05, + 3.2830e-04, 9.1839e-04], + [ 6.9499e-05, 3.0607e-05, 2.8372e-05, ..., -8.7261e-05, + 4.3452e-05, 5.3197e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0290, 0.0110, 0.0173, 0.0225, 0.0276, -0.0071, -0.0035, -0.0042, + 0.0128, -0.0284], device='cuda:0'), grad: tensor([ 0.0016, 0.0008, 0.0012, -0.0042, 0.0001, 0.0022, -0.0048, 0.0018, + 0.0011, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 218.51, cls_loss 0.0305 cls_loss_mapping 0.0328 cls_loss_causal 0.7508 re_mapping 0.0203 re_causal 0.0556 /// teacc 98.35 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0602, 0.0876, -0.0204, ..., 0.0064, 0.0449, 0.0261], + [ 0.0483, -0.0501, 0.0081, ..., -0.0422, -0.0098, -0.0206], + [-0.0132, -0.0465, 0.0237, ..., -0.0266, 0.0264, -0.0688], + ..., + [-0.0483, -0.0382, 0.0368, ..., -0.0158, -0.0551, -0.0535], + [-0.0210, -0.0216, -0.0030, ..., 0.0077, -0.0593, -0.0078], + [-0.0083, 0.0128, 0.0295, ..., 0.0569, -0.0239, -0.0428]], + device='cuda:0'), grad: tensor([[ 4.6045e-05, -9.2173e-04, -2.4462e-04, ..., 3.1777e-06, + 7.1302e-06, -3.7760e-05], + [ 2.6798e-04, 1.8537e-04, 1.2326e-04, ..., 4.6156e-06, + 4.2051e-05, 2.5463e-04], + [ 1.1392e-05, 5.6684e-05, 1.2052e-04, ..., 1.9461e-05, + -1.0476e-05, 4.6754e-04], + ..., + [ 1.0148e-05, 4.0472e-05, 2.7323e-04, ..., 4.5419e-05, + 1.1884e-06, 1.0357e-03], + [ 8.1406e-03, 5.1270e-03, 3.4447e-03, ..., 2.9415e-05, + 1.1711e-03, 4.7874e-03], + [ 7.6175e-05, 2.3794e-04, 2.3770e-04, ..., 1.7568e-05, + 4.9844e-06, 8.2016e-04]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0288, 0.0112, 0.0170, 0.0228, 0.0272, -0.0071, -0.0038, -0.0040, + 0.0129, -0.0283], device='cuda:0'), grad: tensor([-7.8392e-04, 4.3797e-04, -1.2457e-05, -5.7602e-03, 8.5413e-05, + -1.0025e-02, 4.6844e-03, 1.6556e-03, 8.3847e-03, 1.3361e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 218.23, cls_loss 0.0292 cls_loss_mapping 0.0329 cls_loss_causal 0.7693 re_mapping 0.0197 re_causal 0.0553 /// teacc 98.38 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0610, 0.0889, -0.0203, ..., 0.0063, 0.0451, 0.0255], + [ 0.0487, -0.0494, 0.0084, ..., -0.0421, -0.0097, -0.0204], + [-0.0136, -0.0476, 0.0236, ..., -0.0269, 0.0268, -0.0696], + ..., + [-0.0488, -0.0385, 0.0370, ..., -0.0160, -0.0550, -0.0543], + [-0.0217, -0.0225, -0.0033, ..., 0.0076, -0.0603, -0.0078], + [-0.0081, 0.0128, 0.0309, ..., 0.0577, -0.0241, -0.0429]], + device='cuda:0'), grad: tensor([[ 2.7585e-04, 6.2823e-05, 1.5521e-04, ..., 1.3150e-05, + 1.1057e-04, 1.3508e-05], + [-1.4420e-02, -3.7346e-03, -3.3150e-03, ..., 2.1085e-05, + -2.5845e-03, -1.6499e-04], + [-5.0640e-04, 7.0691e-05, -5.6686e-03, ..., 1.7598e-05, + -1.8864e-03, 1.1571e-05], + ..., + [ 2.8729e-04, 1.1042e-05, 4.5547e-03, ..., 1.1432e-04, + 1.0796e-03, 9.2536e-06], + [ 4.6635e-04, -5.5850e-05, 1.6034e-04, ..., -8.1956e-06, + 1.8466e-04, -8.2552e-05], + [ 6.6936e-05, 3.5137e-05, -6.4659e-04, ..., -4.6349e-04, + 4.6402e-05, 3.4243e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0292, 0.0111, 0.0169, 0.0227, 0.0269, -0.0071, -0.0035, -0.0039, + 0.0122, -0.0277], device='cuda:0'), grad: tensor([ 0.0005, -0.0118, -0.0117, 0.0007, 0.0010, 0.0016, 0.0113, 0.0089, + 0.0006, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 219.02, cls_loss 0.0299 cls_loss_mapping 0.0316 cls_loss_causal 0.7992 re_mapping 0.0201 re_causal 0.0550 /// teacc 98.50 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0615, 0.0892, -0.0216, ..., 0.0066, 0.0452, 0.0256], + [ 0.0494, -0.0496, 0.0082, ..., -0.0423, -0.0102, -0.0208], + [-0.0143, -0.0487, 0.0238, ..., -0.0272, 0.0280, -0.0701], + ..., + [-0.0503, -0.0395, 0.0369, ..., -0.0163, -0.0556, -0.0550], + [-0.0212, -0.0214, -0.0025, ..., 0.0079, -0.0606, -0.0069], + [-0.0085, 0.0126, 0.0314, ..., 0.0581, -0.0245, -0.0435]], + device='cuda:0'), grad: tensor([[ 9.5814e-06, -1.5128e-04, -5.5552e-05, ..., 9.7826e-06, + 2.0653e-05, 1.5879e-06], + [-4.8614e-04, -1.1402e-04, -3.4308e-04, ..., -7.6056e-05, + 1.0006e-05, -2.9042e-05], + [ 1.3493e-05, 5.2899e-05, 5.8681e-05, ..., 7.5512e-06, + -1.2323e-05, 8.2105e-06], + ..., + [ 2.4930e-05, 8.2701e-06, -2.1529e-04, ..., 3.9279e-05, + 4.7907e-06, 2.7977e-06], + [ 3.0732e-04, 7.4387e-05, 2.6941e-04, ..., 1.0133e-04, + 2.1741e-05, -2.0444e-05], + [ 3.6687e-05, 6.2048e-05, -6.4790e-05, ..., -2.8276e-04, + 2.8741e-06, 1.8448e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0286, 0.0106, 0.0172, 0.0229, 0.0270, -0.0070, -0.0037, -0.0042, + 0.0130, -0.0278], device='cuda:0'), grad: tensor([-6.7651e-05, -3.8743e-04, 9.4712e-05, 1.8835e-04, -2.5344e-04, + 2.2817e-04, -4.0501e-05, -2.4045e-04, 4.6802e-04, 1.0982e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 218.00, cls_loss 0.0258 cls_loss_mapping 0.0286 cls_loss_causal 0.7488 re_mapping 0.0198 re_causal 0.0536 /// teacc 98.37 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0616, 0.0905, -0.0213, ..., 0.0068, 0.0457, 0.0256], + [ 0.0501, -0.0501, 0.0087, ..., -0.0422, -0.0105, -0.0213], + [-0.0147, -0.0498, 0.0234, ..., -0.0273, 0.0284, -0.0708], + ..., + [-0.0506, -0.0400, 0.0373, ..., -0.0168, -0.0561, -0.0545], + [-0.0218, -0.0208, -0.0019, ..., 0.0080, -0.0610, -0.0067], + [-0.0089, 0.0120, 0.0316, ..., 0.0587, -0.0250, -0.0442]], + device='cuda:0'), grad: tensor([[ 3.0935e-05, 5.4479e-05, 4.0948e-05, ..., 3.4064e-05, + 3.9935e-05, 3.1561e-05], + [-9.2804e-05, 2.1324e-05, 1.4253e-05, ..., 2.2486e-05, + 4.5657e-05, 1.3441e-05], + [ 4.4137e-05, 8.3983e-05, 3.9279e-05, ..., 6.3539e-05, + -3.5143e-04, 4.0293e-05], + ..., + [ 2.8148e-05, 1.9193e-05, -1.5271e-04, ..., 5.0783e-05, + 5.6267e-05, 1.4745e-05], + [ 1.9872e-04, -5.3596e-04, -2.4891e-04, ..., -4.6396e-04, + 9.9301e-05, -2.0802e-05], + [-1.5236e-06, 2.1243e-04, -1.4663e-04, ..., -4.3929e-05, + 1.2726e-05, 5.4002e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0295, 0.0107, 0.0169, 0.0223, 0.0269, -0.0068, -0.0042, -0.0037, + 0.0132, -0.0282], device='cuda:0'), grad: tensor([ 2.7800e-04, 1.8740e-04, -4.5896e-04, 1.8132e-04, 1.7965e-04, + 3.1662e-04, -2.5344e-04, -3.3140e-05, -9.2363e-04, 5.2643e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 218.20, cls_loss 0.0325 cls_loss_mapping 0.0319 cls_loss_causal 0.7622 re_mapping 0.0185 re_causal 0.0515 /// teacc 98.38 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0624, 0.0913, -0.0215, ..., 0.0067, 0.0456, 0.0254], + [ 0.0506, -0.0504, 0.0088, ..., -0.0423, -0.0102, -0.0217], + [-0.0153, -0.0507, 0.0233, ..., -0.0275, 0.0290, -0.0718], + ..., + [-0.0516, -0.0404, 0.0375, ..., -0.0170, -0.0570, -0.0551], + [-0.0225, -0.0208, -0.0016, ..., 0.0082, -0.0614, -0.0066], + [-0.0094, 0.0118, 0.0321, ..., 0.0590, -0.0251, -0.0448]], + device='cuda:0'), grad: tensor([[ 9.7975e-06, 1.4268e-05, 2.8014e-05, ..., 8.0913e-06, + 8.3327e-05, 1.7956e-05], + [-1.5366e-04, 6.7130e-06, -1.2803e-04, ..., 1.5393e-05, + 1.8537e-05, 2.0355e-05], + [ 1.9833e-05, 2.7344e-05, 7.7391e-04, ..., 1.7986e-05, + 3.2330e-04, 2.3633e-05], + ..., + [ 2.6003e-05, 3.5875e-06, -8.6403e-04, ..., 1.8746e-05, + -3.7932e-04, 4.3273e-05], + [ 2.9862e-05, -1.0580e-04, -6.2704e-05, ..., -2.4989e-05, + 5.4330e-05, -7.2420e-05], + [ 2.6941e-05, 5.5432e-05, -1.3895e-03, ..., -1.6661e-03, + 3.6925e-05, 7.2956e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0289, 0.0110, 0.0171, 0.0229, 0.0270, -0.0069, -0.0038, -0.0041, + 0.0128, -0.0282], device='cuda:0'), grad: tensor([ 6.5041e-04, -4.0889e-05, 1.1520e-03, -1.5485e-04, 3.9978e-03, + 3.8862e-04, -2.7771e-03, -1.0424e-03, 1.6022e-04, -2.3289e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 218.03, cls_loss 0.0290 cls_loss_mapping 0.0359 cls_loss_causal 0.7695 re_mapping 0.0193 re_causal 0.0521 /// teacc 98.43 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0629, 0.0920, -0.0218, ..., 0.0064, 0.0457, 0.0252], + [ 0.0513, -0.0505, 0.0086, ..., -0.0426, -0.0107, -0.0221], + [-0.0164, -0.0514, 0.0235, ..., -0.0273, 0.0302, -0.0726], + ..., + [-0.0518, -0.0407, 0.0375, ..., -0.0172, -0.0577, -0.0564], + [-0.0225, -0.0205, -0.0012, ..., 0.0081, -0.0614, -0.0061], + [-0.0090, 0.0117, 0.0324, ..., 0.0594, -0.0250, -0.0453]], + device='cuda:0'), grad: tensor([[ 4.7162e-06, 8.4782e-04, 3.5077e-05, ..., 1.0431e-05, + 8.6948e-06, 1.0662e-05], + [-4.1336e-05, 1.6123e-05, -4.0442e-05, ..., 2.4009e-06, + 3.1441e-05, 4.3437e-06], + [ 1.1712e-05, -8.5890e-05, -1.6019e-05, ..., 2.0221e-05, + -8.0228e-05, 2.6777e-05], + ..., + [ 1.1705e-05, 3.9004e-06, 2.0635e-04, ..., 1.6391e-04, + 1.6361e-05, 4.6678e-06], + [ 3.2783e-05, -3.9041e-05, 1.0714e-05, ..., -2.3425e-05, + 2.1324e-05, -3.9935e-05], + [ 1.5527e-05, 2.0102e-05, -4.6682e-04, ..., -3.9768e-04, + 2.2054e-06, 3.0756e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0289, 0.0108, 0.0180, 0.0229, 0.0275, -0.0070, -0.0044, -0.0044, + 0.0130, -0.0285], device='cuda:0'), grad: tensor([ 1.8473e-03, 1.3232e-04, -6.8140e-04, 1.9050e-04, 3.3665e-04, + 7.6652e-05, -1.8616e-03, 3.8910e-04, 2.4319e-04, -6.7425e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 217.40, cls_loss 0.0253 cls_loss_mapping 0.0257 cls_loss_causal 0.7355 re_mapping 0.0191 re_causal 0.0516 /// teacc 98.34 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0634, 0.0927, -0.0223, ..., 0.0062, 0.0458, 0.0253], + [ 0.0517, -0.0508, 0.0087, ..., -0.0430, -0.0110, -0.0224], + [-0.0167, -0.0519, 0.0233, ..., -0.0275, 0.0306, -0.0733], + ..., + [-0.0530, -0.0411, 0.0378, ..., -0.0173, -0.0577, -0.0570], + [-0.0226, -0.0199, -0.0009, ..., 0.0082, -0.0613, -0.0063], + [-0.0093, 0.0120, 0.0332, ..., 0.0602, -0.0252, -0.0457]], + device='cuda:0'), grad: tensor([[ 1.2338e-05, -1.7750e-04, -7.4744e-05, ..., 1.7323e-06, + 6.7204e-06, 3.0175e-05], + [ 4.4465e-04, 5.8794e-04, 1.5154e-03, ..., -2.4331e-07, + 6.2048e-05, 7.1192e-04], + [ 3.0138e-06, 1.2302e-04, 4.1276e-05, ..., -9.3728e-06, + -9.8944e-05, 2.2054e-05], + ..., + [ 1.1005e-05, 1.7911e-05, -4.4584e-04, ..., 1.8448e-05, + 8.1956e-06, 1.3053e-05], + [-4.4775e-04, -4.4847e-04, -1.0948e-03, ..., 3.5071e-04, + 1.1064e-05, -7.4625e-04], + [ 1.4611e-05, 6.7830e-05, -9.2089e-05, ..., -3.8671e-04, + 3.8967e-06, 1.6421e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0287, 0.0104, 0.0182, 0.0233, 0.0275, -0.0074, -0.0044, -0.0045, + 0.0128, -0.0281], device='cuda:0'), grad: tensor([ 1.6439e-04, 2.6817e-03, -2.1648e-04, 8.4102e-05, 2.2542e-04, + 2.0087e-04, -8.9884e-04, -4.1795e-04, -1.7662e-03, -5.8144e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 218.51, cls_loss 0.0253 cls_loss_mapping 0.0284 cls_loss_causal 0.7152 re_mapping 0.0185 re_causal 0.0497 /// teacc 98.33 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0641, 0.0932, -0.0226, ..., 0.0062, 0.0456, 0.0260], + [ 0.0522, -0.0513, 0.0091, ..., -0.0430, -0.0107, -0.0226], + [-0.0171, -0.0524, 0.0227, ..., -0.0277, 0.0309, -0.0744], + ..., + [-0.0530, -0.0410, 0.0382, ..., -0.0175, -0.0580, -0.0571], + [-0.0235, -0.0203, -0.0005, ..., 0.0084, -0.0616, -0.0066], + [-0.0093, 0.0116, 0.0334, ..., 0.0606, -0.0252, -0.0464]], + device='cuda:0'), grad: tensor([[ 5.6863e-05, 4.2647e-05, 3.3341e-07, ..., 1.4734e-06, + 4.3780e-05, 2.4721e-05], + [-1.4496e-04, 8.5756e-06, -6.4969e-05, ..., 2.2966e-06, + 3.8520e-06, 1.6958e-05], + [ 4.2140e-05, 1.5929e-05, 3.7968e-05, ..., 8.4843e-07, + 3.4785e-07, 2.3916e-05], + ..., + [ 5.0098e-05, -5.1074e-06, -1.7369e-04, ..., 7.0512e-05, + 9.1419e-06, 4.7162e-06], + [ 6.4313e-05, 4.8101e-05, 6.5982e-05, ..., 2.9966e-05, + 2.1681e-05, 3.8654e-05], + [ 3.3796e-05, 4.0829e-05, -1.1259e-04, ..., -2.1076e-04, + 2.6505e-06, 2.5779e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0288, 0.0106, 0.0179, 0.0230, 0.0276, -0.0072, -0.0040, -0.0043, + 0.0124, -0.0282], device='cuda:0'), grad: tensor([ 2.8229e-04, -1.0687e-04, 1.4460e-04, -2.6298e-04, 2.8467e-04, + 4.6563e-04, -7.8726e-04, -1.9586e-04, 2.4796e-04, -7.2539e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 218.30, cls_loss 0.0296 cls_loss_mapping 0.0333 cls_loss_causal 0.7585 re_mapping 0.0181 re_causal 0.0503 /// teacc 98.46 lr 0.00010000 +Epoch 48, weight, value: tensor([[-6.4606e-02, 9.4093e-02, -2.2644e-02, ..., 6.1573e-03, + 4.5534e-02, 2.5863e-02], + [ 5.3295e-02, -5.1590e-02, 9.1473e-03, ..., -4.2985e-02, + -1.0289e-02, -2.3346e-02], + [-1.8046e-02, -5.2534e-02, 2.2706e-02, ..., -2.7645e-02, + 3.1081e-02, -7.4980e-02], + ..., + [-5.3018e-02, -4.1120e-02, 3.8310e-02, ..., -1.7783e-02, + -5.8311e-02, -5.7375e-02], + [-2.4377e-02, -2.0941e-02, -7.2552e-05, ..., 8.4240e-03, + -6.1990e-02, -6.4882e-03], + [-9.6642e-03, 1.1162e-02, 3.3436e-02, ..., 6.0820e-02, + -2.5136e-02, -4.7313e-02]], device='cuda:0'), grad: tensor([[ 4.8208e-04, 4.4584e-04, 1.1826e-04, ..., 7.0512e-05, + 2.6589e-07, 4.3106e-04], + [ 5.8562e-05, 4.3064e-05, 1.1581e-04, ..., 1.6987e-05, + 5.1111e-06, 5.1200e-05], + [ 6.5446e-05, 6.6757e-05, 4.6462e-05, ..., 1.3985e-05, + -1.2621e-05, 6.4135e-05], + ..., + [ 1.6522e-04, 1.5152e-04, -2.8515e-04, ..., 2.0966e-05, + 3.1590e-06, 1.8632e-04], + [ 2.2113e-04, 2.5153e-04, 1.1414e-04, ..., 6.1691e-05, + 1.0170e-06, 1.7726e-04], + [-7.0512e-05, -9.5510e-04, -7.2813e-04, ..., -5.2929e-04, + 3.1851e-07, 3.4362e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0288, 0.0110, 0.0179, 0.0229, 0.0279, -0.0071, -0.0037, -0.0042, + 0.0122, -0.0288], device='cuda:0'), grad: tensor([ 9.9659e-04, 2.2066e-04, 1.4615e-04, 1.3552e-03, -8.2403e-06, + -2.5177e-03, 4.8733e-04, 8.8751e-05, 5.4646e-04, -1.3170e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 218.61, cls_loss 0.0263 cls_loss_mapping 0.0304 cls_loss_causal 0.7030 re_mapping 0.0178 re_causal 0.0470 /// teacc 98.47 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0647, 0.0950, -0.0225, ..., 0.0062, 0.0454, 0.0258], + [ 0.0537, -0.0520, 0.0092, ..., -0.0427, -0.0108, -0.0238], + [-0.0185, -0.0531, 0.0226, ..., -0.0275, 0.0319, -0.0757], + ..., + [-0.0535, -0.0413, 0.0388, ..., -0.0182, -0.0590, -0.0579], + [-0.0243, -0.0205, 0.0001, ..., 0.0082, -0.0620, -0.0058], + [-0.0105, 0.0111, 0.0336, ..., 0.0615, -0.0250, -0.0482]], + device='cuda:0'), grad: tensor([[ 2.2858e-05, -1.9014e-04, -7.3195e-05, ..., 1.7494e-05, + -6.6996e-05, 1.1295e-05], + [-8.4591e-04, 1.9729e-05, -1.1292e-03, ..., 1.1228e-05, + 7.6666e-06, -6.9094e-04], + [ 1.7989e-04, 8.4162e-05, 3.0994e-04, ..., 4.2558e-05, + 1.0066e-05, 1.5593e-04], + ..., + [ 1.0276e-04, 6.4373e-05, 2.4930e-05, ..., 1.8954e-05, + 2.3812e-05, 8.2076e-05], + [ 6.3539e-05, -2.8777e-04, -2.5439e-04, ..., -2.9302e-04, + 1.5825e-05, -1.3256e-04], + [ 2.3134e-06, 1.4770e-04, 8.3208e-05, ..., 4.1425e-05, + 6.4261e-06, 7.1168e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0289, 0.0107, 0.0185, 0.0232, 0.0281, -0.0065, -0.0044, -0.0043, + 0.0122, -0.0295], device='cuda:0'), grad: tensor([-1.5271e-04, -1.5697e-03, 5.0068e-04, 1.4343e-03, 3.1352e-05, + -1.0706e-05, -3.3110e-05, 5.5671e-05, -4.9639e-04, 2.3937e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 48---------------------------------------------------- +epoch 48, time 218.75, cls_loss 0.0228 cls_loss_mapping 0.0245 cls_loss_causal 0.7256 re_mapping 0.0175 re_causal 0.0482 /// teacc 98.51 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0652, 0.0960, -0.0226, ..., 0.0062, 0.0455, 0.0257], + [ 0.0549, -0.0522, 0.0100, ..., -0.0427, -0.0108, -0.0243], + [-0.0195, -0.0542, 0.0223, ..., -0.0277, 0.0324, -0.0764], + ..., + [-0.0545, -0.0414, 0.0391, ..., -0.0184, -0.0593, -0.0581], + [-0.0246, -0.0196, 0.0006, ..., 0.0085, -0.0625, -0.0053], + [-0.0107, 0.0104, 0.0340, ..., 0.0619, -0.0251, -0.0488]], + device='cuda:0'), grad: tensor([[ 1.0855e-05, -5.0485e-05, -2.7731e-05, ..., 1.1977e-06, + 9.1782e-07, 2.7157e-06], + [-2.5463e-04, 3.0939e-06, -2.3031e-04, ..., 4.8354e-06, + 7.1451e-06, 8.9854e-06], + [ 1.5879e-04, 1.0088e-05, 2.0659e-04, ..., 1.1504e-05, + -2.4319e-05, 1.1712e-05], + ..., + [ 3.2544e-05, 4.4368e-06, 2.8253e-05, ..., 1.1504e-05, + 4.9435e-06, 5.4315e-06], + [ 1.0163e-04, 2.1875e-05, -9.8944e-05, ..., -5.9381e-06, + 7.5474e-06, 5.2422e-05], + [ 1.5602e-05, 2.0921e-05, -3.0816e-05, ..., -5.5075e-05, + 1.7695e-07, 9.2462e-06]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0292, 0.0110, 0.0184, 0.0231, 0.0273, -0.0064, -0.0043, -0.0044, + 0.0122, -0.0293], device='cuda:0'), grad: tensor([-2.7120e-05, -3.2449e-04, 2.4390e-04, 1.0842e-04, 7.7128e-05, + 4.6301e-04, -6.4516e-04, 6.4969e-05, 6.7115e-05, -2.8104e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 218.41, cls_loss 0.0253 cls_loss_mapping 0.0286 cls_loss_causal 0.7268 re_mapping 0.0178 re_causal 0.0467 /// teacc 98.41 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0655, 0.0967, -0.0228, ..., 0.0060, 0.0456, 0.0255], + [ 0.0556, -0.0534, 0.0102, ..., -0.0429, -0.0111, -0.0248], + [-0.0201, -0.0547, 0.0221, ..., -0.0278, 0.0332, -0.0770], + ..., + [-0.0555, -0.0415, 0.0390, ..., -0.0189, -0.0598, -0.0585], + [-0.0255, -0.0204, 0.0012, ..., 0.0085, -0.0630, -0.0056], + [-0.0110, 0.0106, 0.0346, ..., 0.0629, -0.0251, -0.0490]], + device='cuda:0'), grad: tensor([[ 1.8142e-06, -5.1451e-04, -1.3089e-04, ..., 1.0118e-05, + -9.1791e-05, 6.9104e-06], + [-7.2241e-05, 5.5172e-06, -3.4481e-05, ..., 1.5004e-06, + 2.3797e-05, 1.9521e-06], + [ 3.1646e-06, 1.3137e-04, -8.3864e-05, ..., 3.6687e-05, + -2.1982e-04, 1.7628e-05], + ..., + [ 1.3381e-05, 2.4244e-05, -5.2899e-06, ..., 1.7300e-05, + 6.2883e-05, 6.6459e-06], + [ 1.2532e-05, -9.6858e-05, 1.6046e-04, ..., -6.5863e-06, + 7.8738e-05, -4.5687e-05], + [ 1.6168e-05, 2.5368e-04, -2.2805e-04, ..., -1.8513e-04, + 5.3018e-05, -7.8678e-06]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0293, 0.0108, 0.0187, 0.0228, 0.0281, -0.0057, -0.0051, -0.0043, + 0.0117, -0.0293], device='cuda:0'), grad: tensor([-9.1171e-04, 5.1260e-06, -2.5940e-04, 2.5821e-04, -1.5587e-05, + 3.5167e-04, 1.5885e-05, 1.7166e-04, 2.4211e-04, 1.4365e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 50---------------------------------------------------- +epoch 50, time 218.98, cls_loss 0.0218 cls_loss_mapping 0.0251 cls_loss_causal 0.7007 re_mapping 0.0170 re_causal 0.0476 /// teacc 98.55 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0662, 0.0973, -0.0226, ..., 0.0059, 0.0456, 0.0249], + [ 0.0560, -0.0534, 0.0098, ..., -0.0437, -0.0113, -0.0250], + [-0.0206, -0.0553, 0.0223, ..., -0.0274, 0.0337, -0.0777], + ..., + [-0.0555, -0.0414, 0.0392, ..., -0.0195, -0.0597, -0.0587], + [-0.0256, -0.0198, 0.0012, ..., 0.0086, -0.0633, -0.0053], + [-0.0108, 0.0102, 0.0349, ..., 0.0633, -0.0251, -0.0494]], + device='cuda:0'), grad: tensor([[ 1.8191e-04, -6.8903e-04, 1.9386e-05, ..., 6.7279e-06, + -2.2697e-04, 6.6042e-05], + [ 2.1172e-04, 1.1259e-04, 2.5916e-04, ..., 5.8502e-05, + 2.1145e-05, 6.5863e-05], + [ 1.0526e-04, 3.2210e-04, 3.2091e-04, ..., 1.0960e-05, + 4.2707e-05, 1.0952e-05], + ..., + [ 6.8545e-05, 6.5207e-05, 7.3135e-05, ..., 5.2243e-05, + 1.2293e-05, 2.2158e-05], + [ 1.0639e-05, -2.6286e-05, -3.2091e-04, ..., 4.6992e-04, + 1.6972e-05, 1.7777e-05], + [ 2.7388e-05, 2.4867e-04, -7.6818e-04, ..., -9.1839e-04, + 5.8234e-05, 3.3289e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0291, 0.0103, 0.0191, 0.0230, 0.0284, -0.0062, -0.0051, -0.0040, + 0.0117, -0.0296], device='cuda:0'), grad: tensor([-0.0014, 0.0010, 0.0005, 0.0014, 0.0004, -0.0017, 0.0004, 0.0003, + -0.0003, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 218.33, cls_loss 0.0231 cls_loss_mapping 0.0237 cls_loss_causal 0.7126 re_mapping 0.0167 re_causal 0.0453 /// teacc 98.46 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0667, 0.0985, -0.0224, ..., 0.0066, 0.0458, 0.0247], + [ 0.0567, -0.0544, 0.0104, ..., -0.0438, -0.0116, -0.0255], + [-0.0214, -0.0562, 0.0222, ..., -0.0281, 0.0338, -0.0780], + ..., + [-0.0565, -0.0417, 0.0395, ..., -0.0199, -0.0597, -0.0588], + [-0.0250, -0.0190, 0.0016, ..., 0.0094, -0.0629, -0.0048], + [-0.0115, 0.0098, 0.0349, ..., 0.0636, -0.0254, -0.0499]], + device='cuda:0'), grad: tensor([[ 2.1183e-04, 2.4840e-05, 1.5600e-06, ..., 1.5318e-05, + 4.5806e-05, 8.3268e-05], + [-9.7882e-07, 4.5002e-05, -3.7879e-05, ..., 8.2748e-07, + 3.3110e-05, 1.5795e-05], + [ 6.1929e-05, -5.3823e-05, 4.5091e-05, ..., 4.4703e-06, + -1.6227e-05, 1.3895e-05], + ..., + [ 1.3418e-05, 2.3723e-05, -1.7583e-05, ..., 3.3528e-06, + 4.8392e-06, 4.3437e-06], + [-1.4086e-03, -1.8377e-03, -1.1063e-03, ..., -6.9952e-04, + -2.1229e-03, 3.1090e-04], + [ 4.4376e-05, 1.1349e-04, 1.0371e-05, ..., -2.6785e-06, + 1.0200e-05, 1.9655e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0292, 0.0106, 0.0188, 0.0230, 0.0287, -0.0063, -0.0054, -0.0038, + 0.0121, -0.0299], device='cuda:0'), grad: tensor([ 2.7061e-05, 9.5010e-05, -4.6873e-04, 2.1732e-04, 3.7694e-04, + -4.5738e-03, 6.8512e-03, 5.5999e-05, -2.7676e-03, 1.8263e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 218.02, cls_loss 0.0174 cls_loss_mapping 0.0199 cls_loss_causal 0.6850 re_mapping 0.0177 re_causal 0.0474 /// teacc 98.41 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0672, 0.0994, -0.0222, ..., 0.0064, 0.0458, 0.0246], + [ 0.0571, -0.0541, 0.0104, ..., -0.0440, -0.0117, -0.0256], + [-0.0218, -0.0579, 0.0216, ..., -0.0283, 0.0340, -0.0785], + ..., + [-0.0570, -0.0425, 0.0401, ..., -0.0201, -0.0598, -0.0590], + [-0.0257, -0.0194, 0.0020, ..., 0.0093, -0.0631, -0.0050], + [-0.0114, 0.0092, 0.0353, ..., 0.0642, -0.0254, -0.0503]], + device='cuda:0'), grad: tensor([[ 1.6242e-06, -4.3958e-05, -2.0683e-05, ..., 6.6543e-07, + 2.1309e-06, -3.1531e-05], + [-3.6120e-05, 4.6492e-06, -2.0087e-05, ..., 6.9384e-08, + 2.7657e-05, 1.7360e-06], + [ 6.8098e-06, -1.5199e-04, 4.3176e-06, ..., 4.0792e-07, + 1.3106e-05, 4.9733e-06], + ..., + [ 1.1943e-05, 6.5975e-06, 1.0528e-05, ..., 8.6501e-06, + 1.2293e-05, 2.4159e-06], + [-6.8903e-05, -1.2910e-04, -6.4433e-05, ..., -3.7819e-05, + 3.4012e-06, -1.2362e-04], + [ 7.3463e-06, 1.4268e-05, -4.0144e-05, ..., -2.4632e-05, + 2.4829e-06, 1.0699e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0295, 0.0102, 0.0183, 0.0232, 0.0287, -0.0060, -0.0053, -0.0034, + 0.0117, -0.0301], device='cuda:0'), grad: tensor([ 1.1015e-04, 5.3585e-05, -6.8045e-04, 4.7302e-04, -3.7384e-04, + 1.7965e-04, 1.0717e-04, 9.7454e-05, -7.0930e-05, 1.0473e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 218.64, cls_loss 0.0199 cls_loss_mapping 0.0212 cls_loss_causal 0.7144 re_mapping 0.0168 re_causal 0.0477 /// teacc 98.44 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0674, 0.0997, -0.0233, ..., 0.0062, 0.0462, 0.0246], + [ 0.0574, -0.0544, 0.0106, ..., -0.0443, -0.0118, -0.0258], + [-0.0223, -0.0586, 0.0216, ..., -0.0285, 0.0342, -0.0789], + ..., + [-0.0580, -0.0426, 0.0396, ..., -0.0201, -0.0600, -0.0593], + [-0.0261, -0.0195, 0.0021, ..., 0.0091, -0.0633, -0.0049], + [-0.0111, 0.0100, 0.0367, ..., 0.0648, -0.0256, -0.0509]], + device='cuda:0'), grad: tensor([[ 3.7611e-05, 4.5037e-04, 5.6171e-04, ..., 2.2864e-07, + -6.1514e-07, 3.5077e-05], + [-6.3777e-05, 9.8497e-06, 3.9011e-05, ..., 1.9744e-07, + 8.5589e-07, 3.1024e-05], + [ 9.8705e-05, 2.8208e-05, 5.1260e-05, ..., 7.3388e-07, + -2.3767e-06, 1.1975e-04], + ..., + [-1.9252e-05, 1.7449e-05, -1.4985e-04, ..., 8.9500e-07, + 3.8790e-07, 6.8694e-06], + [-1.2624e-04, -1.5867e-04, -1.0800e-04, ..., -2.4512e-06, + -1.0012e-07, -4.1199e-04], + [ 4.6670e-05, -5.1022e-04, -6.0892e-04, ..., -1.2599e-05, + 5.1921e-07, 5.4181e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0290, 0.0104, 0.0181, 0.0231, 0.0291, -0.0060, -0.0051, -0.0042, + 0.0117, -0.0293], device='cuda:0'), grad: tensor([ 0.0016, 0.0002, 0.0004, -0.0011, 0.0004, 0.0022, -0.0010, -0.0003, + -0.0008, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 218.38, cls_loss 0.0215 cls_loss_mapping 0.0242 cls_loss_causal 0.7233 re_mapping 0.0168 re_causal 0.0447 /// teacc 98.53 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0687, 0.1006, -0.0233, ..., 0.0063, 0.0464, 0.0244], + [ 0.0581, -0.0546, 0.0108, ..., -0.0444, -0.0121, -0.0261], + [-0.0230, -0.0593, 0.0219, ..., -0.0286, 0.0349, -0.0787], + ..., + [-0.0585, -0.0430, 0.0394, ..., -0.0203, -0.0601, -0.0596], + [-0.0258, -0.0186, 0.0022, ..., 0.0093, -0.0627, -0.0050], + [-0.0118, 0.0094, 0.0377, ..., 0.0652, -0.0259, -0.0519]], + device='cuda:0'), grad: tensor([[ 7.8753e-06, -2.8079e-07, 6.6817e-05, ..., 2.2039e-05, + 4.2468e-06, 4.7907e-06], + [-5.6601e-04, -1.8880e-05, -2.0428e-03, ..., -3.6955e-04, + 9.9652e-08, 1.2435e-05], + [ 1.5393e-05, 1.0580e-05, 5.9575e-05, ..., 1.0990e-05, + -4.4107e-06, 2.1964e-05], + ..., + [ 3.0160e-04, 1.0675e-04, 1.5650e-03, ..., 3.5310e-04, + 2.2464e-06, 7.5027e-06], + [ 3.4660e-05, -4.2915e-05, -1.7315e-05, ..., -1.7658e-06, + 2.9787e-05, -8.8036e-05], + [ 1.8346e-04, -1.2267e-04, 6.8069e-05, ..., -8.5652e-05, + 5.5786e-07, 1.0662e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0288, 0.0102, 0.0184, 0.0230, 0.0287, -0.0056, -0.0056, -0.0040, + 0.0117, -0.0292], device='cuda:0'), grad: tensor([ 8.3089e-05, -1.7920e-03, 4.3601e-05, -8.2888e-08, 1.7214e-04, + 1.2183e-04, -1.5926e-04, 1.4524e-03, 5.9247e-05, 1.9312e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 55---------------------------------------------------- +epoch 55, time 218.85, cls_loss 0.0166 cls_loss_mapping 0.0196 cls_loss_causal 0.6670 re_mapping 0.0156 re_causal 0.0453 /// teacc 98.58 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0689, 0.1012, -0.0233, ..., 0.0062, 0.0463, 0.0244], + [ 0.0578, -0.0550, 0.0104, ..., -0.0444, -0.0122, -0.0263], + [-0.0221, -0.0599, 0.0221, ..., -0.0287, 0.0352, -0.0790], + ..., + [-0.0590, -0.0432, 0.0399, ..., -0.0205, -0.0603, -0.0599], + [-0.0261, -0.0189, 0.0027, ..., 0.0093, -0.0630, -0.0050], + [-0.0120, 0.0091, 0.0378, ..., 0.0657, -0.0262, -0.0522]], + device='cuda:0'), grad: tensor([[ 4.9770e-06, 1.5140e-04, 2.0814e-04, ..., 5.8673e-08, + 8.7842e-06, 4.2981e-07], + [-7.6234e-05, 1.4953e-05, -5.1588e-05, ..., 2.2352e-08, + 8.6846e-07, 4.2561e-07], + [ 2.4602e-05, 5.3085e-06, 5.5909e-05, ..., 3.2643e-07, + 9.8720e-06, 2.5723e-06], + ..., + [ 9.3132e-06, 3.4831e-06, -1.3590e-04, ..., 1.4203e-07, + 1.3318e-06, 2.1793e-06], + [ 8.0228e-05, 1.5944e-05, 3.6299e-05, ..., -6.9849e-07, + 1.6063e-05, -7.8231e-08], + [ 4.8764e-06, 1.6972e-05, 6.1035e-05, ..., -7.4366e-07, + 1.0207e-06, 1.4929e-06]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0289, 0.0093, 0.0191, 0.0230, 0.0287, -0.0057, -0.0057, -0.0034, + 0.0115, -0.0293], device='cuda:0'), grad: tensor([ 1.6794e-03, 3.3826e-05, 1.9467e-04, 1.8686e-05, 2.8610e-03, + 2.1720e-04, -5.6114e-03, -1.9515e-04, 5.1451e-04, 2.8324e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 218.01, cls_loss 0.0166 cls_loss_mapping 0.0205 cls_loss_causal 0.6663 re_mapping 0.0160 re_causal 0.0438 /// teacc 98.57 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0689, 0.1021, -0.0230, ..., 0.0063, 0.0463, 0.0244], + [ 0.0593, -0.0551, 0.0112, ..., -0.0445, -0.0115, -0.0263], + [-0.0229, -0.0608, 0.0221, ..., -0.0289, 0.0355, -0.0795], + ..., + [-0.0605, -0.0434, 0.0396, ..., -0.0206, -0.0613, -0.0604], + [-0.0264, -0.0188, 0.0028, ..., 0.0094, -0.0634, -0.0047], + [-0.0121, 0.0088, 0.0382, ..., 0.0663, -0.0266, -0.0525]], + device='cuda:0'), grad: tensor([[-2.5518e-06, -9.7942e-04, -1.7929e-04, ..., 4.3921e-06, + -1.9324e-04, -8.4281e-05], + [-1.3128e-05, 2.0623e-05, 1.0222e-04, ..., 2.5351e-06, + 9.5516e-06, 1.4767e-05], + [ 9.7752e-06, 1.6582e-04, 1.9588e-03, ..., 1.0943e-04, + 8.9467e-05, 5.9098e-05], + ..., + [ 7.1041e-06, 8.6278e-06, 3.4313e-03, ..., 1.5414e-04, + -1.4529e-06, -9.2685e-05], + [ 2.7791e-05, 5.8651e-05, -6.5422e-03, ..., -3.6049e-04, + 2.4393e-05, -1.0371e-05], + [ 1.6987e-05, 3.2723e-05, 2.2447e-04, ..., 6.2823e-05, + 1.3940e-05, 5.0545e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0289, 0.0103, 0.0186, 0.0230, 0.0285, -0.0057, -0.0053, -0.0037, + 0.0114, -0.0294], device='cuda:0'), grad: tensor([-9.8133e-04, 1.8120e-04, 2.8191e-03, 1.0748e-03, 6.8843e-05, + 1.8167e-04, 5.7268e-04, 4.2458e-03, -8.5983e-03, 4.4155e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 57---------------------------------------------------- +epoch 57, time 218.30, cls_loss 0.0167 cls_loss_mapping 0.0184 cls_loss_causal 0.6749 re_mapping 0.0148 re_causal 0.0426 /// teacc 98.62 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0696, 0.1032, -0.0231, ..., 0.0062, 0.0472, 0.0244], + [ 0.0602, -0.0553, 0.0117, ..., -0.0445, -0.0115, -0.0268], + [-0.0235, -0.0598, 0.0220, ..., -0.0290, 0.0359, -0.0798], + ..., + [-0.0617, -0.0442, 0.0396, ..., -0.0208, -0.0613, -0.0604], + [-0.0269, -0.0190, 0.0034, ..., 0.0094, -0.0640, -0.0050], + [-0.0124, 0.0088, 0.0386, ..., 0.0669, -0.0255, -0.0533]], + device='cuda:0'), grad: tensor([[-3.5405e-05, -4.1389e-04, -7.4282e-06, ..., 1.8515e-06, + -1.1563e-04, -1.7911e-05], + [-1.1101e-05, 1.3284e-05, 5.5790e-05, ..., 2.4401e-06, + 3.5763e-06, 1.0975e-05], + [ 1.3195e-05, 1.9580e-05, 1.6236e-04, ..., 2.0694e-06, + 3.5781e-06, 1.2241e-05], + ..., + [ 1.2375e-05, 2.0236e-05, -3.4142e-04, ..., 1.9088e-05, + 6.0685e-06, 2.3782e-05], + [ 3.0488e-05, 1.5736e-05, 8.2478e-06, ..., -1.0438e-05, + 1.5929e-05, -8.9779e-06], + [ 2.4885e-05, 1.0902e-04, -4.0174e-04, ..., -1.7369e-04, + 2.6688e-05, -3.6448e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0292, 0.0105, 0.0188, 0.0224, 0.0283, -0.0054, -0.0054, -0.0035, + 0.0112, -0.0295], device='cuda:0'), grad: tensor([-7.2289e-04, 2.2590e-04, 5.5265e-04, 6.8140e-04, 6.0654e-04, + 3.6269e-05, 1.5867e-04, -1.2016e-03, 1.4818e-04, -4.8375e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 218.05, cls_loss 0.0160 cls_loss_mapping 0.0194 cls_loss_causal 0.6914 re_mapping 0.0155 re_causal 0.0427 /// teacc 98.52 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0699, 0.1044, -0.0231, ..., 0.0060, 0.0477, 0.0244], + [ 0.0598, -0.0556, 0.0113, ..., -0.0448, -0.0117, -0.0272], + [-0.0229, -0.0611, 0.0214, ..., -0.0291, 0.0360, -0.0809], + ..., + [-0.0625, -0.0451, 0.0401, ..., -0.0209, -0.0614, -0.0607], + [-0.0266, -0.0189, 0.0039, ..., 0.0095, -0.0642, -0.0044], + [-0.0122, 0.0085, 0.0393, ..., 0.0675, -0.0258, -0.0540]], + device='cuda:0'), grad: tensor([[ 7.4267e-05, -6.4254e-05, -4.6432e-05, ..., 1.6112e-06, + 4.0174e-05, 6.9499e-05], + [-1.5426e-04, 1.2524e-05, -1.9324e-04, ..., 2.0582e-07, + -2.6673e-06, 8.7544e-06], + [ 3.8058e-05, 1.1146e-05, 5.2899e-05, ..., 9.4529e-08, + 4.5821e-06, 1.0781e-05], + ..., + [ 4.6700e-05, 9.3430e-06, -2.1845e-05, ..., 1.8496e-06, + 7.2010e-06, 2.5198e-05], + [ 1.0544e-02, 1.6510e-02, 2.2373e-03, ..., 2.3127e-04, + 5.1384e-03, 9.9945e-03], + [ 4.6968e-05, 4.6074e-05, 6.0529e-05, ..., -5.8226e-06, + 1.1414e-05, 3.0816e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0296, 0.0092, 0.0189, 0.0231, 0.0284, -0.0059, -0.0055, -0.0030, + 0.0115, -0.0296], device='cuda:0'), grad: tensor([-5.6392e-07, -2.4283e-04, 1.0365e-04, 3.1114e-05, -6.4039e-04, + 4.8027e-03, -2.0874e-02, -4.1306e-05, 1.6571e-02, 3.0184e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 218.00, cls_loss 0.0212 cls_loss_mapping 0.0239 cls_loss_causal 0.7218 re_mapping 0.0151 re_causal 0.0417 /// teacc 98.57 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0706, 0.1052, -0.0229, ..., 0.0057, 0.0481, 0.0244], + [ 0.0611, -0.0561, 0.0114, ..., -0.0445, -0.0114, -0.0275], + [-0.0241, -0.0617, 0.0209, ..., -0.0289, 0.0357, -0.0817], + ..., + [-0.0635, -0.0461, 0.0401, ..., -0.0211, -0.0612, -0.0608], + [-0.0277, -0.0199, 0.0039, ..., 0.0097, -0.0652, -0.0049], + [-0.0125, 0.0077, 0.0402, ..., 0.0680, -0.0260, -0.0544]], + device='cuda:0'), grad: tensor([[-1.3217e-05, -5.6791e-04, -6.5625e-05, ..., -4.0859e-05, + 7.7160e-07, -2.4378e-05], + [-6.9022e-05, 6.8098e-06, -1.1784e-04, ..., 7.1200e-07, + -1.7751e-06, 8.6352e-06], + [ 3.6031e-05, 5.9038e-05, 6.8784e-05, ..., 7.0967e-06, + 3.5949e-06, 3.0965e-05], + ..., + [ 1.0997e-05, 2.0847e-05, 9.7230e-06, ..., 5.3570e-06, + 3.0454e-07, 1.4015e-05], + [ 4.7386e-05, -5.9366e-05, -5.1588e-05, ..., -1.2264e-05, + 3.3081e-06, -1.2112e-04], + [ 1.9968e-05, 2.4486e-04, 2.4557e-05, ..., 7.6815e-06, + 1.2619e-07, 3.5852e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0297, 0.0102, 0.0182, 0.0225, 0.0291, -0.0053, -0.0053, -0.0032, + 0.0107, -0.0297], device='cuda:0'), grad: tensor([-6.2704e-04, -9.3520e-05, 2.2221e-04, -1.2040e-05, 1.2648e-04, + 1.8764e-04, 8.9586e-05, 6.4015e-05, -2.9516e-04, 3.3832e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 217.73, cls_loss 0.0145 cls_loss_mapping 0.0172 cls_loss_causal 0.6671 re_mapping 0.0153 re_causal 0.0420 /// teacc 98.58 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0710, 0.1062, -0.0229, ..., 0.0056, 0.0483, 0.0243], + [ 0.0614, -0.0566, 0.0118, ..., -0.0454, -0.0113, -0.0277], + [-0.0245, -0.0621, 0.0203, ..., -0.0289, 0.0361, -0.0822], + ..., + [-0.0644, -0.0463, 0.0403, ..., -0.0215, -0.0616, -0.0614], + [-0.0282, -0.0200, 0.0046, ..., 0.0101, -0.0657, -0.0052], + [-0.0121, 0.0073, 0.0402, ..., 0.0688, -0.0264, -0.0548]], + device='cuda:0'), grad: tensor([[ 6.5304e-06, -4.0859e-05, -2.1607e-06, ..., 1.3411e-06, + 6.5193e-06, 2.1495e-06], + [ 1.0335e-04, 7.6443e-06, 5.6934e-04, ..., -1.9185e-07, + 3.3546e-06, 4.2468e-06], + [ 4.2170e-06, 1.4983e-05, 4.7117e-05, ..., -5.1111e-06, + -5.2452e-05, 1.2778e-05], + ..., + [-2.0075e-04, 4.0792e-06, -9.8324e-04, ..., 1.5095e-05, + 1.5646e-05, 1.4290e-05], + [ 7.7724e-05, -7.7533e-07, 9.5904e-05, ..., -2.5667e-06, + 3.5584e-05, 1.4290e-05], + [ 2.8998e-05, 1.8284e-05, 6.0916e-05, ..., -3.5137e-05, + 2.2911e-06, 1.0148e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0302, 0.0103, 0.0177, 0.0227, 0.0291, -0.0056, -0.0047, -0.0032, + 0.0108, -0.0302], device='cuda:0'), grad: tensor([-5.0291e-06, 5.9175e-04, -1.2529e-04, 7.1049e-05, 3.7789e-05, + 6.2823e-05, -1.1975e-04, -8.5449e-04, 1.8179e-04, 1.5974e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 218.12, cls_loss 0.0173 cls_loss_mapping 0.0202 cls_loss_causal 0.6857 re_mapping 0.0149 re_causal 0.0421 /// teacc 98.48 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0712, 0.1069, -0.0228, ..., 0.0057, 0.0481, 0.0241], + [ 0.0624, -0.0570, 0.0124, ..., -0.0455, -0.0116, -0.0279], + [-0.0248, -0.0624, 0.0211, ..., -0.0290, 0.0368, -0.0820], + ..., + [-0.0660, -0.0461, 0.0404, ..., -0.0216, -0.0615, -0.0619], + [-0.0292, -0.0208, 0.0041, ..., 0.0101, -0.0661, -0.0056], + [-0.0119, 0.0071, 0.0402, ..., 0.0693, -0.0266, -0.0550]], + device='cuda:0'), grad: tensor([[ 4.3571e-05, 9.0674e-06, 1.3888e-05, ..., 1.0759e-05, + 5.0068e-05, 3.0458e-05], + [-1.6928e-03, 2.5071e-06, -7.4720e-04, ..., 1.0477e-06, + -8.0919e-04, -9.6941e-04], + [ 1.1463e-03, 4.7460e-06, 5.9462e-04, ..., 4.6566e-06, + 3.9959e-04, 5.1117e-04], + ..., + [ 1.2153e-04, 2.8741e-06, 4.2647e-05, ..., 1.3262e-05, + 7.1943e-05, 8.5413e-05], + [ 1.2791e-04, 4.5747e-05, 2.7761e-05, ..., 1.5765e-05, + 5.1260e-05, 6.4850e-05], + [-2.6512e-04, -2.9087e-04, -3.2425e-04, ..., -4.6015e-04, + -3.1900e-04, 1.9357e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0301, 0.0102, 0.0184, 0.0228, 0.0294, -0.0050, -0.0050, -0.0028, + 0.0099, -0.0308], device='cuda:0'), grad: tensor([ 0.0002, -0.0036, 0.0020, 0.0008, 0.0009, -0.0016, 0.0018, 0.0003, + 0.0002, -0.0010], device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 218.45, cls_loss 0.0179 cls_loss_mapping 0.0215 cls_loss_causal 0.6557 re_mapping 0.0151 re_causal 0.0412 /// teacc 98.48 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0717, 0.1082, -0.0229, ..., 0.0061, 0.0490, 0.0237], + [ 0.0640, -0.0575, 0.0130, ..., -0.0449, -0.0114, -0.0281], + [-0.0255, -0.0629, 0.0209, ..., -0.0290, 0.0367, -0.0830], + ..., + [-0.0665, -0.0467, 0.0406, ..., -0.0220, -0.0611, -0.0622], + [-0.0296, -0.0200, 0.0042, ..., 0.0104, -0.0664, -0.0055], + [-0.0127, 0.0060, 0.0406, ..., 0.0698, -0.0269, -0.0556]], + device='cuda:0'), grad: tensor([[ 2.8089e-05, 7.3195e-04, 9.2536e-06, ..., -2.5332e-06, + 1.2970e-04, 5.4741e-04], + [-6.6102e-05, 9.5069e-06, -7.4208e-05, ..., 1.0394e-06, + 1.0449e-06, 2.1636e-05], + [ 8.7798e-05, 1.8942e-04, 4.8757e-04, ..., 2.1942e-06, + 4.3586e-07, 2.4772e-04], + ..., + [ 2.1800e-05, 1.3098e-05, 2.1055e-05, ..., 9.1791e-06, + 1.5926e-07, 3.6269e-05], + [ 4.1097e-05, -1.7440e-04, -5.1451e-04, ..., -1.7770e-06, + 8.6650e-06, -1.8132e-04], + [ 3.4720e-05, 2.1711e-05, -4.2439e-05, ..., -5.1200e-05, + 1.1111e-06, 5.0545e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0308, 0.0106, 0.0179, 0.0226, 0.0293, -0.0048, -0.0053, -0.0028, + 0.0096, -0.0308], device='cuda:0'), grad: tensor([ 1.3161e-03, 5.8860e-05, 8.5354e-04, -2.1667e-03, -4.6539e-03, + 2.3403e-03, 2.7542e-03, 7.5579e-05, -6.3276e-04, 5.4061e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 218.23, cls_loss 0.0193 cls_loss_mapping 0.0224 cls_loss_causal 0.6715 re_mapping 0.0142 re_causal 0.0397 /// teacc 98.62 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0727, 0.1085, -0.0228, ..., 0.0059, 0.0488, 0.0230], + [ 0.0644, -0.0581, 0.0128, ..., -0.0453, -0.0114, -0.0282], + [-0.0261, -0.0631, 0.0216, ..., -0.0283, 0.0373, -0.0834], + ..., + [-0.0671, -0.0473, 0.0405, ..., -0.0222, -0.0617, -0.0636], + [-0.0299, -0.0197, 0.0040, ..., 0.0102, -0.0668, -0.0047], + [-0.0124, 0.0058, 0.0412, ..., 0.0705, -0.0267, -0.0552]], + device='cuda:0'), grad: tensor([[ 4.4852e-05, 7.6056e-05, -2.7083e-06, ..., 2.0236e-05, + 2.2873e-06, 5.1588e-05], + [ 2.9132e-05, 1.6069e-04, 1.9401e-05, ..., 9.2462e-06, + -6.3218e-06, 6.7055e-05], + [ 1.4400e-04, -1.3390e-03, 1.5044e-04, ..., -3.4237e-04, + 1.0990e-06, -1.0461e-04], + ..., + [ 3.4273e-05, 7.0512e-05, 3.2276e-05, ..., 1.5542e-05, + 1.5553e-06, 2.8849e-05], + [-4.7278e-04, -1.1368e-03, -6.7663e-04, ..., -1.3328e-04, + 2.4457e-06, -5.5361e-04], + [ 9.6321e-05, 2.5344e-04, 9.8765e-05, ..., 7.1704e-05, + 8.3912e-07, 1.5175e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0306, 0.0105, 0.0185, 0.0225, 0.0297, -0.0052, -0.0055, -0.0027, + 0.0097, -0.0310], device='cuda:0'), grad: tensor([ 2.9659e-04, 3.5906e-04, -2.3422e-03, 6.6137e-04, 9.9897e-05, + 2.8381e-03, 2.7990e-04, 3.2783e-04, -2.8896e-03, 3.6860e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 64---------------------------------------------------- +epoch 64, time 219.20, cls_loss 0.0148 cls_loss_mapping 0.0142 cls_loss_causal 0.6731 re_mapping 0.0146 re_causal 0.0399 /// teacc 98.65 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0735, 0.1095, -0.0224, ..., 0.0057, 0.0495, 0.0231], + [ 0.0656, -0.0599, 0.0137, ..., -0.0453, -0.0119, -0.0285], + [-0.0274, -0.0652, 0.0204, ..., -0.0288, 0.0374, -0.0850], + ..., + [-0.0680, -0.0474, 0.0406, ..., -0.0224, -0.0618, -0.0650], + [-0.0302, -0.0192, 0.0046, ..., 0.0103, -0.0673, -0.0041], + [-0.0126, 0.0057, 0.0414, ..., 0.0713, -0.0267, -0.0551]], + device='cuda:0'), grad: tensor([[ 2.9325e-05, 2.9169e-06, 1.8314e-05, ..., 5.3532e-06, + 3.4515e-06, 2.7969e-05], + [-1.1313e-04, 6.3516e-06, -1.9276e-04, ..., 6.6496e-06, + 1.8124e-06, -4.8399e-05], + [ 1.2279e-04, 8.6874e-06, 1.9407e-04, ..., 7.3537e-06, + -4.7654e-05, 1.1420e-04], + ..., + [ 9.8586e-05, 6.1430e-06, -1.7965e-04, ..., 4.4823e-05, + 1.8641e-05, 7.7486e-05], + [ 5.7888e-04, 1.0151e-04, 5.4508e-05, ..., 9.7871e-05, + 1.0505e-06, 5.7268e-04], + [-2.7323e-04, -4.7624e-05, -8.8120e-04, ..., -5.3263e-04, + 4.0792e-07, -2.1577e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0312, 0.0106, 0.0177, 0.0230, 0.0294, -0.0052, -0.0055, -0.0028, + 0.0100, -0.0310], device='cuda:0'), grad: tensor([ 1.3053e-04, -1.4091e-04, -5.6601e-04, 1.9580e-05, 8.1062e-04, + -7.0477e-04, 1.3447e-04, 2.4343e-04, 1.2932e-03, -1.2178e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 218.25, cls_loss 0.0168 cls_loss_mapping 0.0193 cls_loss_causal 0.6673 re_mapping 0.0148 re_causal 0.0394 /// teacc 98.39 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0742, 0.1096, -0.0228, ..., 0.0055, 0.0495, 0.0229], + [ 0.0666, -0.0593, 0.0139, ..., -0.0453, -0.0113, -0.0289], + [-0.0281, -0.0657, 0.0202, ..., -0.0291, 0.0376, -0.0855], + ..., + [-0.0687, -0.0478, 0.0408, ..., -0.0229, -0.0620, -0.0661], + [-0.0304, -0.0191, 0.0051, ..., 0.0107, -0.0675, -0.0038], + [-0.0130, 0.0058, 0.0418, ..., 0.0720, -0.0270, -0.0558]], + device='cuda:0'), grad: tensor([[ 4.8161e-05, 2.6274e-04, 1.0979e-04, ..., 1.0151e-04, + 2.8722e-06, 1.2541e-04], + [-4.1217e-05, 2.8789e-05, -4.0174e-05, ..., 1.0423e-05, + 2.1160e-06, 2.0161e-05], + [ 2.6658e-05, 5.4538e-05, 1.0586e-04, ..., 2.0817e-05, + -3.3947e-07, -3.7241e-04], + ..., + [ 2.2978e-05, 2.4199e-05, -1.3936e-04, ..., 7.0259e-06, + -5.6736e-06, 1.9282e-05], + [-3.3468e-05, -9.2030e-04, -5.0545e-04, ..., -3.9268e-04, + 2.4885e-06, -4.5681e-04], + [ 3.6985e-05, 1.9312e-04, 5.9962e-05, ..., 5.4240e-05, + 4.1537e-07, 8.3983e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0307, 0.0110, 0.0172, 0.0233, 0.0290, -0.0042, -0.0046, -0.0036, + 0.0099, -0.0312], device='cuda:0'), grad: tensor([-1.1833e-02, 5.3585e-05, -2.0199e-03, 2.4452e-03, 8.5220e-03, + 5.6934e-04, 2.3861e-03, 1.9193e-04, -1.0910e-03, 7.8106e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 218.04, cls_loss 0.0142 cls_loss_mapping 0.0136 cls_loss_causal 0.6822 re_mapping 0.0142 re_causal 0.0408 /// teacc 98.50 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0750, 0.1101, -0.0230, ..., 0.0051, 0.0498, 0.0226], + [ 0.0675, -0.0599, 0.0144, ..., -0.0454, -0.0115, -0.0291], + [-0.0289, -0.0665, 0.0201, ..., -0.0291, 0.0378, -0.0865], + ..., + [-0.0696, -0.0484, 0.0405, ..., -0.0234, -0.0620, -0.0669], + [-0.0306, -0.0191, 0.0053, ..., 0.0109, -0.0677, -0.0033], + [-0.0131, 0.0060, 0.0426, ..., 0.0731, -0.0271, -0.0559]], + device='cuda:0'), grad: tensor([[-1.5602e-05, -2.5997e-03, -8.1730e-04, ..., 6.6450e-07, + -6.9571e-04, -5.1022e-04], + [ 1.5534e-06, 1.1225e-03, 5.6553e-04, ..., 3.1311e-06, + 3.4904e-04, 2.5582e-04], + [ 3.9451e-06, 1.3208e-04, 4.5013e-03, ..., 1.0896e-06, + 1.4467e-03, 7.1943e-05], + ..., + [ 3.8408e-06, 8.4415e-06, -5.1155e-03, ..., -1.5259e-05, + -1.4687e-03, 6.3419e-05], + [ 1.1083e-06, 7.3195e-05, 4.1366e-05, ..., -1.9930e-06, + 2.8178e-05, 3.4183e-05], + [ 1.4715e-05, 6.9022e-05, 3.1328e-04, ..., 5.7742e-06, + 4.0568e-06, 3.9965e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0309, 0.0112, 0.0175, 0.0228, 0.0286, -0.0042, -0.0039, -0.0043, + 0.0099, -0.0308], device='cuda:0'), grad: tensor([-4.5738e-03, 2.3956e-03, 6.1951e-03, -2.6321e-04, 1.3971e-04, + 8.3506e-05, 1.9894e-03, -6.6795e-03, 2.2566e-04, 4.8542e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 218.62, cls_loss 0.0171 cls_loss_mapping 0.0188 cls_loss_causal 0.6994 re_mapping 0.0146 re_causal 0.0414 /// teacc 98.59 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0757, 0.1104, -0.0241, ..., 0.0051, 0.0499, 0.0228], + [ 0.0673, -0.0600, 0.0145, ..., -0.0454, -0.0113, -0.0290], + [-0.0279, -0.0670, 0.0197, ..., -0.0294, 0.0375, -0.0872], + ..., + [-0.0700, -0.0490, 0.0407, ..., -0.0236, -0.0616, -0.0673], + [-0.0312, -0.0194, 0.0052, ..., 0.0111, -0.0681, -0.0032], + [-0.0135, 0.0063, 0.0442, ..., 0.0741, -0.0274, -0.0565]], + device='cuda:0'), grad: tensor([[ 2.0635e-04, 1.7190e-04, 3.7003e-04, ..., 3.4988e-05, + 1.5199e-04, 1.1164e-04], + [-1.3971e-03, -1.3037e-03, -2.8267e-03, ..., -2.2626e-04, + -1.1368e-03, -7.7057e-04], + [ 3.0279e-05, 3.1531e-05, 6.8367e-05, ..., 1.0423e-05, + 2.0817e-05, 1.9148e-05], + ..., + [ 5.8919e-05, 4.7952e-05, 4.8661e-04, ..., 1.3733e-04, + 8.0839e-06, 2.6748e-05], + [ 1.0805e-03, 9.7370e-04, 1.8101e-03, ..., 1.6069e-04, + 7.3481e-04, 6.1131e-04], + [ 2.7800e-04, 1.9658e-04, -7.2622e-04, ..., -2.2209e-04, + 3.2634e-05, 1.6308e-04]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0305, 0.0107, 0.0177, 0.0233, 0.0290, -0.0043, -0.0045, -0.0041, + 0.0094, -0.0304], device='cuda:0'), grad: tensor([ 6.0701e-04, -4.3793e-03, 9.9182e-05, 1.0557e-03, 3.5810e-04, + -1.0214e-03, 7.9691e-05, 5.6076e-04, 3.0594e-03, -4.2343e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 68---------------------------------------------------- +epoch 68, time 218.96, cls_loss 0.0154 cls_loss_mapping 0.0174 cls_loss_causal 0.6630 re_mapping 0.0147 re_causal 0.0391 /// teacc 98.66 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0758, 0.1112, -0.0242, ..., 0.0049, 0.0500, 0.0225], + [ 0.0672, -0.0599, 0.0142, ..., -0.0454, -0.0105, -0.0287], + [-0.0270, -0.0676, 0.0191, ..., -0.0296, 0.0375, -0.0879], + ..., + [-0.0703, -0.0493, 0.0416, ..., -0.0242, -0.0603, -0.0683], + [-0.0318, -0.0191, 0.0056, ..., 0.0113, -0.0689, -0.0026], + [-0.0138, 0.0062, 0.0447, ..., 0.0749, -0.0276, -0.0573]], + device='cuda:0'), grad: tensor([[ 1.8656e-05, -7.5698e-05, 9.1940e-06, ..., -7.5363e-06, + 1.1943e-05, 1.5702e-06], + [-3.0899e-04, 3.0473e-06, -3.7360e-04, ..., 8.9593e-07, + -9.8526e-05, 1.5832e-06], + [ 1.1265e-05, 1.8030e-05, 4.7296e-05, ..., 2.7698e-06, + -2.9492e-04, 1.3374e-05], + ..., + [ 3.0726e-05, 2.0452e-06, -2.1207e-04, ..., 3.3174e-06, + -1.2684e-04, 1.7490e-06], + [-3.5428e-06, -1.9789e-05, 3.1769e-05, ..., -5.6475e-06, + 9.2462e-06, -2.7195e-05], + [ 6.3062e-05, 1.7166e-05, 1.4818e-04, ..., -4.2289e-05, + 1.1075e-04, 8.1062e-06]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0307, 0.0099, 0.0177, 0.0230, 0.0289, -0.0042, -0.0046, -0.0030, + 0.0094, -0.0305], device='cuda:0'), grad: tensor([-1.2338e-05, -6.6185e-04, -1.5240e-03, 4.2129e-04, 2.4724e-04, + 4.9710e-05, 1.2913e-03, -2.1100e-04, 5.7906e-05, 3.4022e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 217.86, cls_loss 0.0128 cls_loss_mapping 0.0156 cls_loss_causal 0.6350 re_mapping 0.0141 re_causal 0.0381 /// teacc 98.52 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0764, 0.1117, -0.0240, ..., 0.0049, 0.0499, 0.0223], + [ 0.0682, -0.0600, 0.0139, ..., -0.0453, -0.0110, -0.0297], + [-0.0275, -0.0681, 0.0188, ..., -0.0298, 0.0379, -0.0885], + ..., + [-0.0707, -0.0498, 0.0423, ..., -0.0246, -0.0592, -0.0677], + [-0.0324, -0.0197, 0.0056, ..., 0.0116, -0.0696, -0.0027], + [-0.0141, 0.0060, 0.0452, ..., 0.0755, -0.0277, -0.0580]], + device='cuda:0'), grad: tensor([[ 1.9334e-06, -7.9632e-05, -2.9519e-05, ..., 5.0152e-07, + -9.5516e-06, -1.7926e-05], + [-1.2219e-06, 1.2808e-05, 1.0267e-05, ..., 4.3865e-07, + 1.4074e-05, 2.8722e-06], + [ 9.3319e-07, 8.4192e-06, -8.0764e-06, ..., 7.4180e-07, + 1.1809e-05, 2.6524e-06], + ..., + [-2.4326e-06, 3.9451e-06, 3.4682e-06, ..., 1.3649e-05, + 2.0519e-05, 1.1269e-06], + [ 5.2191e-06, 2.7306e-06, 7.1973e-06, ..., 3.0901e-06, + 4.3474e-06, -3.3118e-06], + [ 2.5444e-06, 9.9912e-06, -6.8665e-05, ..., -4.9591e-05, + 1.3277e-05, 2.9281e-06]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0311, 0.0098, 0.0175, 0.0231, 0.0287, -0.0037, -0.0050, -0.0025, + 0.0090, -0.0308], device='cuda:0'), grad: tensor([-7.6830e-05, 5.5671e-05, 5.0128e-05, 2.6628e-05, -1.0294e-04, + 2.4498e-05, -2.0400e-05, 5.4002e-05, 4.2617e-05, -5.3227e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 218.14, cls_loss 0.0101 cls_loss_mapping 0.0121 cls_loss_causal 0.6248 re_mapping 0.0134 re_causal 0.0380 /// teacc 98.55 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0766, 0.1123, -0.0240, ..., 0.0051, 0.0500, 0.0221], + [ 0.0682, -0.0602, 0.0151, ..., -0.0452, -0.0112, -0.0290], + [-0.0265, -0.0687, 0.0186, ..., -0.0301, 0.0383, -0.0891], + ..., + [-0.0728, -0.0502, 0.0417, ..., -0.0248, -0.0594, -0.0693], + [-0.0329, -0.0195, 0.0064, ..., 0.0121, -0.0697, -0.0025], + [-0.0149, 0.0053, 0.0451, ..., 0.0757, -0.0278, -0.0588]], + device='cuda:0'), grad: tensor([[ 2.6431e-06, -1.0973e-04, -3.9399e-05, ..., 1.2089e-06, + 5.3085e-08, -1.5557e-05], + [-6.3300e-05, 2.5295e-06, -5.1469e-05, ..., -1.9725e-06, + 2.7008e-08, 1.9502e-06], + [ 6.4000e-06, 1.1221e-05, 2.5034e-05, ..., 9.6112e-06, + -2.6356e-07, 1.5125e-05], + ..., + [ 4.5970e-06, 6.5677e-06, -5.8487e-06, ..., 2.5779e-06, + 4.0978e-08, 1.7118e-06], + [-1.9521e-05, -2.6733e-05, -5.4657e-05, ..., -3.9458e-05, + 6.9849e-08, -7.2658e-05], + [ 9.8348e-06, 1.8910e-05, -1.2696e-05, ..., -9.8720e-06, + 6.0536e-09, 8.4117e-06]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0311, 0.0099, 0.0182, 0.0232, 0.0285, -0.0034, -0.0047, -0.0034, + 0.0092, -0.0312], device='cuda:0'), grad: tensor([-1.1605e-04, -8.4579e-05, 4.5657e-05, 4.2617e-05, 1.6704e-05, + 9.3222e-05, 1.0353e-04, 2.3171e-06, -1.2922e-04, 2.6032e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 218.05, cls_loss 0.0134 cls_loss_mapping 0.0141 cls_loss_causal 0.6360 re_mapping 0.0134 re_causal 0.0377 /// teacc 98.55 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0791, 0.1124, -0.0244, ..., 0.0051, 0.0501, 0.0214], + [ 0.0697, -0.0589, 0.0160, ..., -0.0443, -0.0109, -0.0281], + [-0.0271, -0.0693, 0.0186, ..., -0.0305, 0.0387, -0.0896], + ..., + [-0.0730, -0.0508, 0.0416, ..., -0.0251, -0.0601, -0.0709], + [-0.0334, -0.0197, 0.0069, ..., 0.0123, -0.0701, -0.0018], + [-0.0154, 0.0047, 0.0451, ..., 0.0759, -0.0280, -0.0596]], + device='cuda:0'), grad: tensor([[ 1.4538e-06, -2.1443e-05, 2.2620e-05, ..., 3.0845e-06, + 2.5835e-06, 6.7055e-07], + [-3.0361e-06, 1.5739e-06, 8.6069e-05, ..., 1.4566e-06, + 2.8918e-07, 5.6326e-06], + [ 1.0384e-06, 2.6338e-06, 1.1158e-04, ..., 7.4133e-06, + 8.9081e-07, 5.8822e-06], + ..., + [ 8.5682e-07, 2.7884e-06, -4.0102e-04, ..., 5.3227e-05, + 8.3353e-08, -1.4096e-05], + [-2.6692e-06, -1.8644e-04, -5.2512e-05, ..., 1.2647e-06, + 1.9781e-06, -1.8966e-04], + [ 1.0747e-06, 6.6459e-06, -9.3937e-05, ..., -1.0896e-04, + 1.6345e-07, 5.8077e-06]], device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0305, 0.0111, 0.0183, 0.0233, 0.0287, -0.0031, -0.0049, -0.0036, + 0.0091, -0.0318], device='cuda:0'), grad: tensor([ 4.8280e-05, 1.6022e-04, 2.1791e-04, 2.6488e-04, 1.9848e-04, + 3.9101e-04, -3.4738e-04, -7.2765e-04, -1.2839e-04, -7.7963e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 218.81, cls_loss 0.0126 cls_loss_mapping 0.0158 cls_loss_causal 0.6700 re_mapping 0.0131 re_causal 0.0380 /// teacc 98.60 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0793, 0.1132, -0.0243, ..., 0.0050, 0.0501, 0.0216], + [ 0.0697, -0.0588, 0.0159, ..., -0.0443, -0.0108, -0.0283], + [-0.0267, -0.0701, 0.0189, ..., -0.0305, 0.0393, -0.0904], + ..., + [-0.0734, -0.0511, 0.0416, ..., -0.0255, -0.0608, -0.0713], + [-0.0336, -0.0194, 0.0074, ..., 0.0130, -0.0700, -0.0012], + [-0.0156, 0.0051, 0.0459, ..., 0.0767, -0.0273, -0.0602]], + device='cuda:0'), grad: tensor([[ 3.2812e-05, 3.2336e-05, 2.9858e-06, ..., 3.1497e-06, + 7.4040e-08, 1.9357e-05], + [-4.2617e-06, 3.5875e-06, 4.2953e-06, ..., 1.4231e-06, + 4.9826e-08, 2.6524e-06], + [ 5.9009e-06, 3.1982e-06, -3.0380e-06, ..., 1.1912e-06, + -5.0524e-07, 2.8014e-06], + ..., + [ 1.0334e-05, 9.7230e-06, 2.4706e-05, ..., 1.9059e-05, + 3.4925e-08, 5.7071e-06], + [ 2.8998e-05, 2.0295e-05, -8.6278e-06, ..., -1.0565e-05, + 2.6543e-07, 9.8869e-06], + [ 5.9843e-05, 6.7472e-05, -3.5912e-05, ..., -1.2577e-05, + 1.3039e-08, 4.5598e-05]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0305, 0.0107, 0.0189, 0.0230, 0.0287, -0.0033, -0.0054, -0.0039, + 0.0094, -0.0313], device='cuda:0'), grad: tensor([ 5.3465e-05, 1.6376e-05, -1.4350e-05, 2.6870e-04, -1.7822e-05, + -4.6659e-04, 1.1794e-05, 4.6015e-05, 2.8387e-05, 7.3493e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 217.64, cls_loss 0.0122 cls_loss_mapping 0.0133 cls_loss_causal 0.6377 re_mapping 0.0133 re_causal 0.0375 /// teacc 98.60 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0794, 0.1138, -0.0247, ..., 0.0040, 0.0502, 0.0218], + [ 0.0703, -0.0591, 0.0158, ..., -0.0442, -0.0112, -0.0286], + [-0.0270, -0.0709, 0.0185, ..., -0.0307, 0.0397, -0.0909], + ..., + [-0.0739, -0.0516, 0.0419, ..., -0.0262, -0.0609, -0.0713], + [-0.0337, -0.0192, 0.0080, ..., 0.0137, -0.0700, -0.0011], + [-0.0159, 0.0051, 0.0468, ..., 0.0779, -0.0276, -0.0611]], + device='cuda:0'), grad: tensor([[-8.4043e-05, -2.6455e-03, -5.2881e-04, ..., 9.3738e-07, + -2.1973e-03, -6.9571e-04], + [ 5.6207e-05, 1.1873e-03, 2.2805e-04, ..., 1.4361e-06, + 1.0262e-03, 3.4308e-04], + [ 4.9561e-05, 7.6771e-04, 1.5557e-04, ..., 2.3581e-06, + 6.4898e-04, 2.2650e-04], + ..., + [ 2.5302e-05, 2.7671e-05, 1.2182e-05, ..., 5.2266e-06, + 1.7956e-05, 2.2873e-05], + [ 4.1533e-04, 1.7703e-04, 1.7449e-05, ..., 2.1979e-05, + 1.1224e-04, 3.5238e-04], + [-6.5088e-05, 5.1022e-05, -1.2100e-04, ..., -8.9705e-05, + 1.7732e-05, 6.0610e-06]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0302, 0.0104, 0.0187, 0.0229, 0.0283, -0.0034, -0.0055, -0.0033, + 0.0096, -0.0308], device='cuda:0'), grad: tensor([-6.9771e-03, 3.2158e-03, 2.0885e-03, 2.2602e-04, 3.5357e-04, + 1.5712e-04, -9.9242e-05, 1.2201e-04, 1.0605e-03, -1.4305e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 74---------------------------------------------------- +epoch 74, time 218.95, cls_loss 0.0134 cls_loss_mapping 0.0141 cls_loss_causal 0.6296 re_mapping 0.0131 re_causal 0.0367 /// teacc 98.69 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0800, 0.1151, -0.0240, ..., 0.0040, 0.0507, 0.0214], + [ 0.0701, -0.0596, 0.0148, ..., -0.0450, -0.0112, -0.0305], + [-0.0276, -0.0719, 0.0177, ..., -0.0311, 0.0398, -0.0917], + ..., + [-0.0730, -0.0522, 0.0430, ..., -0.0261, -0.0606, -0.0698], + [-0.0343, -0.0198, 0.0080, ..., 0.0136, -0.0705, -0.0012], + [-0.0158, 0.0049, 0.0476, ..., 0.0789, -0.0283, -0.0616]], + device='cuda:0'), grad: tensor([[ 3.2783e-06, 1.8496e-06, 1.8124e-06, ..., 3.8650e-07, + 1.3923e-07, 2.3991e-06], + [-1.7926e-05, 1.3411e-06, -1.8612e-05, ..., -2.8200e-06, + 4.6100e-08, 1.5646e-06], + [ 4.6119e-06, -1.2450e-05, -4.5970e-06, ..., 1.6615e-06, + -2.8107e-06, 1.1250e-06], + ..., + [ 4.8317e-06, 4.4964e-06, -1.7732e-06, ..., 8.6566e-07, + 1.2610e-06, 1.0347e-06], + [ 5.4568e-05, 4.9859e-05, 2.7865e-06, ..., 2.8163e-06, + 2.4354e-07, 2.9892e-05], + [ 7.6592e-06, 5.5097e-06, -1.2174e-05, ..., -9.0227e-06, + 1.5926e-07, 3.2820e-06]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0308, 0.0095, 0.0181, 0.0229, 0.0281, -0.0035, -0.0052, -0.0026, + 0.0091, -0.0305], device='cuda:0'), grad: tensor([ 2.8834e-05, -1.4722e-05, -1.7333e-04, 6.8486e-05, 3.6269e-05, + -2.3973e-04, 1.5616e-04, 4.3005e-05, 1.0002e-04, -5.2154e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 75---------------------------------------------------- +epoch 75, time 219.13, cls_loss 0.0113 cls_loss_mapping 0.0150 cls_loss_causal 0.6452 re_mapping 0.0125 re_causal 0.0360 /// teacc 98.72 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0802, 0.1156, -0.0239, ..., 0.0038, 0.0507, 0.0211], + [ 0.0711, -0.0594, 0.0149, ..., -0.0444, -0.0112, -0.0306], + [-0.0283, -0.0722, 0.0188, ..., -0.0314, 0.0402, -0.0920], + ..., + [-0.0736, -0.0525, 0.0431, ..., -0.0263, -0.0608, -0.0701], + [-0.0346, -0.0199, 0.0078, ..., 0.0136, -0.0707, -0.0011], + [-0.0166, 0.0044, 0.0475, ..., 0.0793, -0.0283, -0.0622]], + device='cuda:0'), grad: tensor([[ 4.1239e-06, -2.3052e-05, -1.0081e-05, ..., -1.8924e-06, + 6.4820e-07, 6.1505e-06], + [-1.8120e-05, 1.3113e-06, -1.8820e-05, ..., 1.0021e-06, + -2.7660e-06, 7.5111e-07], + [ 8.2776e-06, 5.0403e-06, 2.8789e-05, ..., 1.0334e-05, + 8.8755e-07, 5.7332e-06], + ..., + [ 4.9956e-06, 1.9521e-06, -3.2604e-05, ..., 2.5947e-06, + 2.0675e-07, 1.1791e-06], + [ 6.2473e-06, -1.1772e-05, -6.1333e-05, ..., -2.4959e-05, + 7.0734e-07, -4.1306e-05], + [-5.3316e-05, -2.9758e-05, 4.1455e-05, ..., -4.3839e-05, + 9.1270e-08, 1.6931e-06]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0308, 0.0096, 0.0190, 0.0228, 0.0295, -0.0034, -0.0053, -0.0028, + 0.0085, -0.0317], device='cuda:0'), grad: tensor([ 6.4611e-05, -2.1651e-05, -1.3523e-03, 3.7932e-04, 7.4911e-04, + 3.6538e-05, 1.1021e-04, -3.9458e-05, -4.0263e-05, 1.1349e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 217.99, cls_loss 0.0096 cls_loss_mapping 0.0109 cls_loss_causal 0.6376 re_mapping 0.0125 re_causal 0.0360 /// teacc 98.64 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0803, 0.1163, -0.0237, ..., 0.0037, 0.0507, 0.0210], + [ 0.0718, -0.0590, 0.0150, ..., -0.0443, -0.0112, -0.0306], + [-0.0288, -0.0729, 0.0186, ..., -0.0318, 0.0403, -0.0927], + ..., + [-0.0738, -0.0528, 0.0433, ..., -0.0266, -0.0609, -0.0703], + [-0.0348, -0.0198, 0.0080, ..., 0.0140, -0.0708, -0.0008], + [-0.0171, 0.0039, 0.0478, ..., 0.0800, -0.0285, -0.0628]], + device='cuda:0'), grad: tensor([[ 1.5255e-06, -2.7999e-05, 2.4401e-06, ..., 1.0878e-06, + 1.4529e-07, -1.5259e-05], + [ 3.9160e-05, 1.5885e-05, 4.5300e-05, ..., 3.6508e-05, + 1.0384e-07, 7.1041e-06], + [ 6.4634e-06, 2.4512e-06, -7.9349e-06, ..., 4.4517e-06, + -9.0199e-07, 1.2964e-06], + ..., + [ 4.2096e-06, 2.0256e-07, 4.1217e-05, ..., 7.3761e-06, + 1.8300e-07, 6.7335e-07], + [ 3.5048e-05, 1.4298e-05, 3.6478e-05, ..., 2.8148e-05, + 2.5611e-07, 6.8918e-06], + [-1.2112e-04, -3.6150e-05, -3.5143e-04, ..., -1.4353e-04, + 1.7695e-08, -1.5184e-05]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0313, 0.0098, 0.0189, 0.0224, 0.0292, -0.0031, -0.0050, -0.0026, + 0.0084, -0.0321], device='cuda:0'), grad: tensor([ 4.0144e-05, 1.0884e-04, 4.5687e-05, 4.0412e-05, -6.4410e-06, + 6.1452e-05, 2.1800e-05, 6.3062e-05, 7.9572e-05, -4.5538e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 218.13, cls_loss 0.0094 cls_loss_mapping 0.0116 cls_loss_causal 0.6364 re_mapping 0.0123 re_causal 0.0357 /// teacc 98.63 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0805, 0.1168, -0.0237, ..., 0.0035, 0.0510, 0.0209], + [ 0.0722, -0.0592, 0.0151, ..., -0.0448, -0.0117, -0.0308], + [-0.0294, -0.0735, 0.0178, ..., -0.0321, 0.0405, -0.0933], + ..., + [-0.0745, -0.0532, 0.0437, ..., -0.0268, -0.0608, -0.0706], + [-0.0356, -0.0204, 0.0082, ..., 0.0140, -0.0711, -0.0009], + [-0.0171, 0.0038, 0.0482, ..., 0.0807, -0.0287, -0.0633]], + device='cuda:0'), grad: tensor([[ 4.9658e-06, -1.8209e-05, -7.6089e-07, ..., 7.6415e-07, + -1.6289e-06, -9.8348e-07], + [-8.4519e-05, 7.6136e-07, 1.0842e-04, ..., -8.9873e-07, + -7.5903e-08, -5.8860e-06], + [ 4.1574e-05, 7.7859e-06, -2.4021e-04, ..., 1.0859e-06, + 5.9558e-07, 1.0468e-06], + ..., + [ 8.1882e-06, 4.7265e-07, -9.4697e-06, ..., 1.1191e-05, + 5.7742e-08, 7.3528e-07], + [ 4.1634e-05, 5.8822e-06, 6.0081e-05, ..., 1.3046e-05, + 1.0226e-06, 4.9770e-06], + [-4.0674e-04, 1.4901e-06, -3.8648e-04, ..., -2.9278e-04, + 1.7369e-07, -1.1474e-04]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0316, 0.0096, 0.0178, 0.0236, 0.0295, -0.0039, -0.0048, -0.0018, + 0.0082, -0.0324], device='cuda:0'), grad: tensor([ 7.8380e-06, 2.8753e-04, 4.2510e-04, -1.0805e-03, 1.9813e-04, + 8.0585e-04, 4.8399e-05, -5.2571e-05, 1.5461e-04, -7.9441e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 218.06, cls_loss 0.0093 cls_loss_mapping 0.0141 cls_loss_causal 0.6517 re_mapping 0.0127 re_causal 0.0366 /// teacc 98.69 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0810, 0.1171, -0.0237, ..., 0.0033, 0.0509, 0.0203], + [ 0.0726, -0.0594, 0.0152, ..., -0.0450, -0.0119, -0.0311], + [-0.0298, -0.0735, 0.0178, ..., -0.0321, 0.0405, -0.0936], + ..., + [-0.0751, -0.0538, 0.0438, ..., -0.0273, -0.0606, -0.0709], + [-0.0361, -0.0206, 0.0083, ..., 0.0143, -0.0714, -0.0010], + [-0.0173, 0.0038, 0.0488, ..., 0.0819, -0.0288, -0.0638]], + device='cuda:0'), grad: tensor([[ 3.9667e-05, -8.6963e-05, -2.0154e-06, ..., 3.3062e-08, + 5.8301e-07, 2.7614e-07], + [-5.5730e-06, 3.1684e-06, -2.2491e-07, ..., -1.8254e-07, + -3.5483e-07, 1.5553e-07], + [ 3.2373e-06, 2.4382e-06, 3.9674e-06, ..., 4.5169e-08, + 5.0804e-07, 3.2559e-06], + ..., + [ 1.6503e-06, 6.8918e-07, -2.4796e-05, ..., 2.1001e-07, + 9.2667e-08, 3.3248e-07], + [ 2.1666e-05, 2.2039e-05, 4.8727e-06, ..., 2.2305e-07, + 1.0226e-06, 2.8275e-06], + [ 1.1981e-05, 5.6662e-06, 8.2031e-06, ..., -1.3588e-06, + 5.7276e-08, 8.6874e-06]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0314, 0.0094, 0.0179, 0.0233, 0.0286, -0.0034, -0.0040, -0.0015, + 0.0075, -0.0320], device='cuda:0'), grad: tensor([-1.1182e-04, 4.6194e-06, 2.4572e-05, -7.6711e-05, -9.9391e-06, + 1.2648e-04, -3.0965e-05, -2.6092e-05, 5.8889e-05, 4.1008e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 217.92, cls_loss 0.0113 cls_loss_mapping 0.0127 cls_loss_causal 0.6062 re_mapping 0.0125 re_causal 0.0344 /// teacc 98.68 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0813, 0.1175, -0.0238, ..., 0.0031, 0.0510, 0.0197], + [ 0.0723, -0.0597, 0.0153, ..., -0.0456, -0.0125, -0.0320], + [-0.0303, -0.0748, 0.0174, ..., -0.0325, 0.0397, -0.0942], + ..., + [-0.0781, -0.0545, 0.0438, ..., -0.0280, -0.0597, -0.0729], + [-0.0356, -0.0199, 0.0091, ..., 0.0148, -0.0716, -0.0004], + [-0.0174, 0.0035, 0.0495, ..., 0.0832, -0.0282, -0.0644]], + device='cuda:0'), grad: tensor([[ 2.7530e-06, -1.2718e-05, 1.9565e-05, ..., 9.7007e-06, + 1.3284e-05, 2.8312e-06], + [-3.6001e-05, 1.7248e-06, -4.6402e-05, ..., -9.7975e-06, + 1.3495e-06, 7.3621e-07], + [ 5.0738e-06, 3.7849e-06, 5.5981e-04, ..., 1.9968e-04, + -2.3985e-04, 3.3945e-05], + ..., + [ 1.6093e-05, 1.2191e-06, 3.6836e-04, ..., 1.4770e-04, + 4.4942e-05, -4.9081e-07], + [ 8.8871e-05, 3.6657e-05, -1.1339e-03, ..., -3.9244e-04, + 2.0675e-06, -6.9626e-06], + [ 2.3812e-05, 1.5765e-05, 1.0586e-04, ..., 4.0650e-05, + 2.5183e-06, 2.1026e-05]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0311, 0.0091, 0.0167, 0.0234, 0.0283, -0.0024, -0.0038, -0.0024, + 0.0087, -0.0314], device='cuda:0'), grad: tensor([ 9.4712e-05, -2.9117e-05, -5.7697e-04, 2.1350e-04, 7.9107e-04, + -1.3137e-04, 5.3674e-05, 6.6948e-04, -1.2484e-03, 1.6201e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 217.69, cls_loss 0.0115 cls_loss_mapping 0.0135 cls_loss_causal 0.6396 re_mapping 0.0118 re_causal 0.0336 /// teacc 98.59 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0814, 0.1186, -0.0233, ..., 0.0034, 0.0510, 0.0198], + [ 0.0726, -0.0596, 0.0152, ..., -0.0464, -0.0115, -0.0323], + [-0.0314, -0.0749, 0.0170, ..., -0.0327, 0.0400, -0.0947], + ..., + [-0.0789, -0.0548, 0.0441, ..., -0.0282, -0.0604, -0.0745], + [-0.0358, -0.0201, 0.0101, ..., 0.0147, -0.0720, 0.0008], + [-0.0173, 0.0027, 0.0496, ..., 0.0836, -0.0286, -0.0651]], + device='cuda:0'), grad: tensor([[ 6.2585e-06, -2.3270e-04, 2.8964e-07, ..., 7.1106e-07, + 2.6971e-06, -7.4767e-06], + [-1.6347e-05, 1.5602e-05, -3.7253e-05, ..., 6.9663e-06, + -4.0792e-06, 1.6868e-05], + [ 1.4879e-05, 2.8804e-05, 1.0505e-05, ..., 4.9919e-06, + -1.4985e-06, 2.0921e-05], + ..., + [ 1.8567e-05, 7.2382e-06, 3.6567e-05, ..., 3.5111e-06, + 3.9041e-06, 4.6939e-06], + [-2.4065e-05, 9.2328e-05, -8.1658e-05, ..., -5.0396e-05, + 1.7002e-05, -1.5453e-05], + [ 7.5325e-06, 2.3261e-05, -2.7180e-05, ..., -1.1727e-05, + 6.2119e-07, 1.0580e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0311, 0.0090, 0.0164, 0.0235, 0.0290, -0.0022, -0.0036, -0.0025, + 0.0089, -0.0322], device='cuda:0'), grad: tensor([-2.3794e-04, -2.1458e-06, 6.2346e-05, 8.8334e-05, 4.4465e-05, + 7.7963e-05, -4.7112e-04, 7.8857e-05, 3.5024e-04, 1.0103e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 81---------------------------------------------------- +epoch 81, time 218.36, cls_loss 0.0099 cls_loss_mapping 0.0135 cls_loss_causal 0.6664 re_mapping 0.0122 re_causal 0.0356 /// teacc 98.77 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0827, 0.1189, -0.0232, ..., 0.0033, 0.0510, 0.0191], + [ 0.0725, -0.0597, 0.0152, ..., -0.0467, -0.0113, -0.0327], + [-0.0309, -0.0756, 0.0169, ..., -0.0331, 0.0400, -0.0953], + ..., + [-0.0791, -0.0555, 0.0457, ..., -0.0284, -0.0605, -0.0744], + [-0.0365, -0.0200, 0.0097, ..., 0.0151, -0.0722, 0.0008], + [-0.0174, 0.0024, 0.0487, ..., 0.0839, -0.0287, -0.0656]], + device='cuda:0'), grad: tensor([[-1.7077e-05, -2.2173e-04, -2.2948e-05, ..., 5.9120e-06, + 7.1526e-07, 1.6063e-05], + [ 3.1173e-05, 2.4624e-06, 2.3949e-04, ..., 6.1870e-05, + 5.3011e-06, 1.6594e-04], + [ 2.1942e-06, 4.9062e-06, 2.3496e-04, ..., 6.1691e-05, + 1.5553e-07, 1.6558e-04], + ..., + [ 1.8489e-04, 2.1104e-06, 5.6416e-05, ..., 1.3113e-05, + -7.1973e-06, 1.1182e-04], + [-7.7367e-05, -8.9407e-05, -7.5483e-04, ..., -2.3055e-04, + 1.7192e-06, -7.5579e-04], + [-6.4552e-05, 1.7822e-05, -4.6611e-04, ..., -3.5214e-04, + 1.7285e-06, 2.9698e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0310, 0.0087, 0.0168, 0.0234, 0.0294, -0.0018, -0.0035, -0.0015, + 0.0083, -0.0334], device='cuda:0'), grad: tensor([-3.9577e-04, 6.0225e-04, 4.9639e-04, 5.4538e-06, 8.8549e-04, + 3.6716e-05, 4.7040e-04, 4.3893e-04, -1.9207e-03, -6.1989e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 217.64, cls_loss 0.0096 cls_loss_mapping 0.0119 cls_loss_causal 0.5994 re_mapping 0.0121 re_causal 0.0335 /// teacc 98.67 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0828, 0.1195, -0.0231, ..., 0.0032, 0.0510, 0.0190], + [ 0.0740, -0.0601, 0.0162, ..., -0.0467, -0.0113, -0.0325], + [-0.0313, -0.0756, 0.0167, ..., -0.0333, 0.0404, -0.0964], + ..., + [-0.0806, -0.0560, 0.0457, ..., -0.0286, -0.0611, -0.0759], + [-0.0368, -0.0194, 0.0102, ..., 0.0153, -0.0724, 0.0016], + [-0.0186, 0.0021, 0.0485, ..., 0.0844, -0.0289, -0.0664]], + device='cuda:0'), grad: tensor([[ 1.4506e-05, 1.3418e-05, 2.0415e-05, ..., 7.2606e-06, + 5.1595e-06, 6.6794e-06], + [ 1.3590e-04, 1.3077e-04, 6.3598e-05, ..., 1.2159e-05, + 1.7574e-06, 4.5806e-05], + [ 4.8950e-06, 6.4559e-06, 8.2672e-05, ..., 1.0401e-05, + 6.3553e-06, 8.3819e-06], + ..., + [ 3.7402e-06, 2.3209e-06, -1.0067e-04, ..., 1.2638e-06, + -1.4633e-05, 1.4938e-06], + [ 4.5109e-04, 2.6155e-04, -7.2420e-05, ..., -4.8697e-05, + 4.7646e-06, 9.2149e-05], + [ 7.6443e-06, 1.6183e-05, -2.0862e-05, ..., -2.3305e-05, + -1.5408e-05, 1.2003e-05]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0310, 0.0093, 0.0173, 0.0237, 0.0294, -0.0017, -0.0037, -0.0024, + 0.0086, -0.0339], device='cuda:0'), grad: tensor([ 5.6773e-05, 2.0325e-04, 1.4746e-04, -4.7743e-05, 9.6738e-05, + -1.1530e-03, 5.8222e-04, -1.8597e-04, 2.8396e-04, 1.6764e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 217.99, cls_loss 0.0096 cls_loss_mapping 0.0135 cls_loss_causal 0.6323 re_mapping 0.0121 re_causal 0.0340 /// teacc 98.72 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0830, 0.1194, -0.0236, ..., 0.0030, 0.0511, 0.0185], + [ 0.0749, -0.0592, 0.0169, ..., -0.0466, -0.0100, -0.0320], + [-0.0318, -0.0764, 0.0159, ..., -0.0338, 0.0406, -0.0983], + ..., + [-0.0814, -0.0565, 0.0458, ..., -0.0289, -0.0611, -0.0765], + [-0.0371, -0.0193, 0.0105, ..., 0.0157, -0.0735, 0.0018], + [-0.0188, 0.0017, 0.0494, ..., 0.0856, -0.0291, -0.0668]], + device='cuda:0'), grad: tensor([[ 3.3975e-06, -2.2184e-06, 3.0342e-06, ..., 1.9278e-06, + 1.6019e-06, 3.7234e-06], + [-3.4779e-05, 8.6892e-07, -9.7513e-05, ..., -1.1086e-05, + -2.4028e-07, -5.3607e-06], + [ 2.2259e-06, 1.4771e-06, 2.0027e-05, ..., 1.4473e-06, + 2.2724e-07, 7.2896e-05], + ..., + [ 1.1407e-05, 8.9174e-07, 6.9261e-05, ..., 6.9588e-06, + 1.8533e-07, 1.3769e-04], + [ 1.3418e-05, 3.8296e-05, 1.7568e-05, ..., 1.3337e-05, + 1.8716e-05, 1.9431e-05], + [ 1.3418e-05, 1.7975e-06, -1.3612e-05, ..., -3.8713e-05, + 1.6876e-06, 3.3118e-06]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0302, 0.0101, 0.0166, 0.0246, 0.0295, -0.0020, -0.0043, -0.0025, + 0.0083, -0.0330], device='cuda:0'), grad: tensor([ 1.1817e-05, -1.1081e-04, 2.0504e-04, -5.4026e-04, 4.6670e-05, + 6.1464e-04, -7.5483e-04, 4.2748e-04, 1.1206e-04, -1.1794e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 217.56, cls_loss 0.0098 cls_loss_mapping 0.0134 cls_loss_causal 0.6229 re_mapping 0.0113 re_causal 0.0329 /// teacc 98.58 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0831, 0.1201, -0.0231, ..., 0.0031, 0.0514, 0.0185], + [ 0.0754, -0.0590, 0.0176, ..., -0.0470, -0.0091, -0.0317], + [-0.0314, -0.0768, 0.0167, ..., -0.0340, 0.0418, -0.0990], + ..., + [-0.0824, -0.0572, 0.0452, ..., -0.0291, -0.0629, -0.0766], + [-0.0376, -0.0191, 0.0108, ..., 0.0160, -0.0739, 0.0019], + [-0.0188, 0.0011, 0.0495, ..., 0.0866, -0.0304, -0.0676]], + device='cuda:0'), grad: tensor([[ 2.0396e-07, -4.6492e-04, -2.0885e-04, ..., -2.3711e-04, + 8.8010e-08, 2.7828e-06], + [-5.2005e-06, 2.5667e-06, -2.2613e-06, ..., 7.8231e-06, + 7.4506e-08, 2.3879e-06], + [ 5.7183e-07, 3.1054e-05, 2.6867e-05, ..., 2.0608e-05, + -8.8988e-07, 1.0811e-05], + ..., + [ 9.6951e-07, 3.6303e-06, 1.9595e-05, ..., 2.9743e-05, + -2.9430e-07, 2.1569e-06], + [ 1.8537e-05, 2.7871e-04, 7.8797e-05, ..., 1.6820e-04, + 5.1409e-07, -1.9729e-05], + [ 1.3243e-06, 1.0121e-04, 8.6650e-06, ..., 5.1308e-04, + 2.4727e-07, 5.0962e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0306, 0.0103, 0.0178, 0.0239, 0.0301, -0.0018, -0.0043, -0.0030, + 0.0081, -0.0340], device='cuda:0'), grad: tensor([-6.2513e-04, 5.2929e-05, 5.3793e-05, 7.2539e-05, -2.7370e-03, + 4.5031e-05, 8.2970e-05, 1.4806e-04, 3.5882e-04, 2.5501e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 217.92, cls_loss 0.0124 cls_loss_mapping 0.0128 cls_loss_causal 0.6740 re_mapping 0.0115 re_causal 0.0338 /// teacc 98.74 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0834, 0.1206, -0.0230, ..., 0.0029, 0.0519, 0.0185], + [ 0.0764, -0.0597, 0.0174, ..., -0.0480, -0.0095, -0.0321], + [-0.0321, -0.0778, 0.0162, ..., -0.0329, 0.0419, -0.1007], + ..., + [-0.0836, -0.0578, 0.0446, ..., -0.0295, -0.0635, -0.0771], + [-0.0383, -0.0194, 0.0109, ..., 0.0159, -0.0750, 0.0018], + [-0.0190, 0.0009, 0.0515, ..., 0.0875, -0.0293, -0.0683]], + device='cuda:0'), grad: tensor([[ 5.5991e-06, 3.4541e-05, 1.5795e-05, ..., 1.3866e-05, + 7.6555e-07, 2.4065e-05], + [-2.4548e-03, 2.3320e-06, -1.7715e-04, ..., 1.0375e-06, + -3.8981e-04, 1.4398e-06], + [ 1.0699e-04, 1.5390e-04, 7.8082e-05, ..., 6.1214e-05, + 1.4126e-05, 1.0639e-04], + ..., + [ 7.7307e-05, 1.8794e-06, 8.6427e-06, ..., 1.8924e-06, + 1.3478e-05, 1.3486e-06], + [-2.2665e-05, -2.9278e-04, -1.2070e-04, ..., -1.1140e-04, + 1.3206e-06, -2.0099e-04], + [ 1.6212e-05, 2.0768e-06, -1.6019e-05, ..., -1.2696e-05, + 2.7623e-06, 1.9763e-06]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0306, 0.0101, 0.0182, 0.0237, 0.0295, -0.0015, -0.0042, -0.0035, + 0.0074, -0.0329], device='cuda:0'), grad: tensor([ 7.1526e-05, -5.5542e-03, 4.8161e-04, 1.1539e-04, 5.0774e-03, + 7.9751e-05, -6.1452e-05, 1.9443e-04, -4.3082e-04, 3.0339e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 217.65, cls_loss 0.0111 cls_loss_mapping 0.0125 cls_loss_causal 0.6076 re_mapping 0.0124 re_causal 0.0352 /// teacc 98.72 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0836, 0.1209, -0.0246, ..., 0.0006, 0.0523, 0.0188], + [ 0.0770, -0.0601, 0.0179, ..., -0.0471, -0.0096, -0.0316], + [-0.0324, -0.0780, 0.0163, ..., -0.0330, 0.0420, -0.1007], + ..., + [-0.0844, -0.0580, 0.0442, ..., -0.0295, -0.0635, -0.0800], + [-0.0384, -0.0190, 0.0106, ..., 0.0154, -0.0753, 0.0023], + [-0.0190, 0.0018, 0.0519, ..., 0.0884, -0.0300, -0.0685]], + device='cuda:0'), grad: tensor([[ 8.7405e-07, -6.6614e-04, -4.1962e-04, ..., -8.4579e-05, + -9.2089e-05, -2.0802e-04], + [-9.0152e-06, 1.7956e-05, 1.3635e-05, ..., 2.6226e-06, + 4.8280e-06, 6.2250e-06], + [-1.4398e-06, 1.1556e-05, 1.0319e-05, ..., 1.7770e-06, + -6.7651e-06, 9.0227e-06], + ..., + [ 2.6356e-06, 1.2055e-05, -5.6595e-05, ..., 1.6596e-06, + 2.3730e-06, 1.9744e-06], + [ 2.0131e-05, 3.8314e-04, 2.0230e-04, ..., 4.7803e-05, + 8.1897e-05, 1.1975e-04], + [ 8.4043e-06, 1.5962e-04, 1.2803e-04, ..., 1.6078e-05, + 2.4810e-06, 6.2823e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0302, 0.0105, 0.0184, 0.0247, 0.0306, -0.0013, -0.0049, -0.0045, + 0.0073, -0.0335], device='cuda:0'), grad: tensor([-7.8964e-04, 2.8223e-05, -2.4483e-05, 5.0753e-05, 2.1189e-05, + -9.2313e-06, 1.3220e-04, -6.2883e-05, 3.6955e-04, 2.8539e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 217.71, cls_loss 0.0066 cls_loss_mapping 0.0102 cls_loss_causal 0.6146 re_mapping 0.0116 re_causal 0.0342 /// teacc 98.67 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0838, 0.1217, -0.0243, ..., 0.0006, 0.0526, 0.0196], + [ 0.0767, -0.0603, 0.0178, ..., -0.0469, -0.0096, -0.0319], + [-0.0317, -0.0784, 0.0164, ..., -0.0331, 0.0421, -0.1012], + ..., + [-0.0846, -0.0587, 0.0445, ..., -0.0297, -0.0636, -0.0803], + [-0.0386, -0.0191, 0.0111, ..., 0.0156, -0.0757, 0.0028], + [-0.0191, 0.0018, 0.0520, ..., 0.0887, -0.0302, -0.0690]], + device='cuda:0'), grad: tensor([[ 6.3479e-06, -6.0303e-07, 6.9849e-08, ..., 2.4680e-07, + 0.0000e+00, 8.4564e-06], + [-3.1143e-05, 2.6263e-06, -2.9191e-05, ..., 2.5146e-08, + -3.4124e-06, 1.0312e-05], + [ 2.3842e-05, 6.9439e-06, 1.3418e-05, ..., 1.8533e-07, + 9.7603e-07, 4.9710e-05], + ..., + [ 1.1206e-05, 2.4103e-06, -1.2994e-05, ..., 3.9563e-06, + 2.7008e-08, 1.6302e-05], + [ 5.1826e-05, 2.2128e-05, 5.6326e-06, ..., 7.7719e-07, + 3.6927e-07, 7.9274e-05], + [ 1.3880e-05, 5.8599e-06, -2.4080e-05, ..., -2.1592e-05, + 1.0049e-06, 1.7121e-05]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0306, 0.0101, 0.0188, 0.0245, 0.0309, -0.0017, -0.0046, -0.0043, + 0.0073, -0.0339], device='cuda:0'), grad: tensor([ 1.2450e-05, -3.6627e-05, 9.6917e-05, -7.8087e-03, 4.4078e-05, + 7.7133e-03, -1.1486e-04, 6.1020e-06, 1.1039e-04, -2.3752e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 217.56, cls_loss 0.0078 cls_loss_mapping 0.0097 cls_loss_causal 0.6220 re_mapping 0.0112 re_causal 0.0322 /// teacc 98.70 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0840, 0.1221, -0.0244, ..., 0.0006, 0.0526, 0.0192], + [ 0.0783, -0.0604, 0.0188, ..., -0.0471, -0.0093, -0.0311], + [-0.0319, -0.0788, 0.0166, ..., -0.0330, 0.0421, -0.1016], + ..., + [-0.0866, -0.0588, 0.0442, ..., -0.0296, -0.0637, -0.0807], + [-0.0390, -0.0189, 0.0114, ..., 0.0159, -0.0761, 0.0028], + [-0.0198, 0.0014, 0.0527, ..., 0.0894, -0.0305, -0.0702]], + device='cuda:0'), grad: tensor([[ 4.5560e-06, 3.6247e-06, 2.3171e-06, ..., 2.0303e-07, + 2.0452e-06, 8.9556e-06], + [-8.8569e-07, 1.1390e-06, -3.9749e-06, ..., 1.2200e-07, + 5.4110e-07, 8.7693e-06], + [ 7.9513e-05, -9.9242e-06, -1.7360e-05, ..., 2.9057e-07, + -2.0370e-05, 1.1897e-04], + ..., + [ 3.0324e-06, 8.2701e-07, -3.6657e-06, ..., 2.6636e-07, + 1.9670e-06, 5.5581e-06], + [ 1.3523e-05, 5.8822e-06, 1.2219e-05, ..., -5.6159e-07, + 1.2383e-05, 2.8238e-05], + [ 1.0788e-05, 5.3048e-06, 2.7381e-07, ..., -5.2713e-07, + 3.5670e-07, 1.1876e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0306, 0.0110, 0.0190, 0.0244, 0.0305, -0.0017, -0.0046, -0.0046, + 0.0069, -0.0338], device='cuda:0'), grad: tensor([ 5.6624e-05, 2.7508e-05, 5.2595e-04, 8.3387e-05, 3.5465e-06, + -9.7561e-04, 4.9114e-05, 3.3319e-05, 1.6522e-04, 2.9683e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 89---------------------------------------------------- +epoch 89, time 218.29, cls_loss 0.0077 cls_loss_mapping 0.0105 cls_loss_causal 0.6370 re_mapping 0.0112 re_causal 0.0327 /// teacc 98.79 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0842, 0.1230, -0.0244, ..., 0.0004, 0.0530, 0.0202], + [ 0.0784, -0.0608, 0.0185, ..., -0.0476, -0.0092, -0.0312], + [-0.0319, -0.0796, 0.0163, ..., -0.0334, 0.0420, -0.1021], + ..., + [-0.0869, -0.0593, 0.0448, ..., -0.0298, -0.0635, -0.0809], + [-0.0391, -0.0188, 0.0113, ..., 0.0160, -0.0764, 0.0033], + [-0.0197, 0.0014, 0.0534, ..., 0.0904, -0.0308, -0.0706]], + device='cuda:0'), grad: tensor([[ 5.6997e-07, 4.2431e-06, 1.1608e-05, ..., 4.7162e-06, + 8.1491e-07, 6.0685e-06], + [-4.5896e-06, 2.5835e-06, 1.8060e-05, ..., 4.5560e-06, + 1.5646e-07, 2.4848e-06], + [ 4.4983e-07, 7.6517e-06, 3.6806e-05, ..., 8.4043e-06, + 2.8592e-07, 7.5251e-06], + ..., + [ 1.7192e-06, 3.4273e-06, -3.5453e-04, ..., 8.7917e-06, + 1.0151e-07, -6.0260e-05], + [ 3.7700e-06, -3.9667e-05, -1.7130e-04, ..., -7.3791e-05, + 5.8673e-07, -2.7314e-05], + [ 2.7120e-06, 2.3201e-05, 1.0526e-04, ..., 3.7760e-05, + 1.6205e-07, 1.1340e-05]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0312, 0.0109, 0.0190, 0.0247, 0.0301, -0.0023, -0.0052, -0.0041, + 0.0066, -0.0335], device='cuda:0'), grad: tensor([ 2.6003e-05, 2.8521e-05, 5.5015e-05, 6.6423e-04, 2.1942e-06, + 1.1496e-05, -1.2971e-05, -7.1096e-04, -1.8966e-04, 1.2672e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 217.30, cls_loss 0.0089 cls_loss_mapping 0.0127 cls_loss_causal 0.6149 re_mapping 0.0122 re_causal 0.0329 /// teacc 98.75 lr 0.00010000 +Epoch 92, weight, value: tensor([[-8.4600e-02, 1.2208e-01, -2.6762e-02, ..., -4.3078e-05, + 5.2928e-02, 1.9908e-02], + [ 7.9229e-02, -6.0628e-02, 1.9296e-02, ..., -4.7532e-02, + -8.7501e-03, -3.0850e-02], + [-3.2474e-02, -8.0596e-02, 1.5421e-02, ..., -3.3499e-02, + 4.1755e-02, -1.0378e-01], + ..., + [-8.7551e-02, -5.9669e-02, 4.5150e-02, ..., -2.9995e-02, + -6.3422e-02, -8.1266e-02], + [-3.9573e-02, -1.9030e-02, 1.1372e-02, ..., 1.5747e-02, + -7.6937e-02, 2.8958e-03], + [-2.0385e-02, 2.9750e-03, 5.4518e-02, ..., 9.0953e-02, + -3.0393e-02, -7.1479e-02]], device='cuda:0'), grad: tensor([[ 8.4192e-07, -9.5189e-05, 2.9244e-07, ..., 2.0675e-07, + 3.6597e-05, 6.8173e-07], + [-1.3318e-06, 4.9174e-07, 1.7956e-05, ..., 1.8291e-06, + 1.1697e-06, 2.3395e-06], + [ 7.7672e-07, 6.2585e-07, 9.8050e-06, ..., 1.8813e-07, + 1.3404e-05, 1.2238e-06], + ..., + [ 9.0338e-07, 5.9605e-07, -6.2287e-05, ..., 1.0617e-06, + 1.1697e-06, -7.0632e-06], + [ 2.6952e-06, 1.4454e-06, 3.4608e-06, ..., 1.0226e-06, + 1.7621e-06, 1.3374e-06], + [ 3.7588e-06, 8.3894e-06, -6.9499e-05, ..., -4.8339e-05, + 1.6391e-06, 3.4235e-06]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0289, 0.0116, 0.0184, 0.0244, 0.0297, -0.0014, -0.0049, -0.0037, + 0.0059, -0.0324], device='cuda:0'), grad: tensor([ 3.4642e-04, 5.5611e-05, 1.9026e-04, 1.3340e-04, 3.6502e-04, + 5.4061e-05, -9.9373e-04, -1.1969e-04, 3.2693e-05, -6.3419e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 217.41, cls_loss 0.0068 cls_loss_mapping 0.0099 cls_loss_causal 0.6287 re_mapping 0.0113 re_causal 0.0342 /// teacc 98.64 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0849, 0.1236, -0.0265, ..., -0.0003, 0.0535, 0.0195], + [ 0.0799, -0.0626, 0.0196, ..., -0.0477, -0.0090, -0.0313], + [-0.0329, -0.0813, 0.0153, ..., -0.0337, 0.0417, -0.1043], + ..., + [-0.0885, -0.0595, 0.0453, ..., -0.0302, -0.0633, -0.0813], + [-0.0399, -0.0187, 0.0115, ..., 0.0161, -0.0772, 0.0033], + [-0.0209, 0.0027, 0.0545, ..., 0.0913, -0.0303, -0.0720]], + device='cuda:0'), grad: tensor([[ 4.3735e-06, -1.3185e-04, -9.7156e-06, ..., 7.4506e-09, + 1.0990e-07, -1.6112e-06], + [-4.3726e-04, 7.0594e-07, -7.4244e-04, ..., 2.9802e-08, + -3.1665e-07, 6.2771e-07], + [ 1.2182e-05, 2.7698e-06, 2.3097e-05, ..., 1.0058e-07, + 8.7544e-08, 5.0478e-07], + ..., + [ 8.9467e-05, 6.7241e-07, 1.0711e-04, ..., 5.2154e-08, + 1.5460e-07, 3.0734e-07], + [ 4.1187e-05, 5.9381e-06, 5.7787e-05, ..., -3.8743e-07, + 2.6263e-07, 2.4401e-06], + [ 2.0742e-04, 1.6421e-05, 3.7766e-04, ..., -3.5390e-08, + 1.0431e-07, 1.2685e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0300, 0.0112, 0.0183, 0.0234, 0.0297, -0.0008, -0.0055, -0.0030, + 0.0058, -0.0327], device='cuda:0'), grad: tensor([-1.5247e-04, -1.0786e-03, 3.7402e-05, 1.5162e-05, 2.1946e-04, + 3.6180e-05, 1.0365e-04, 1.5247e-04, 1.0037e-04, 5.6648e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 92---------------------------------------------------- +epoch 92, time 218.39, cls_loss 0.0086 cls_loss_mapping 0.0107 cls_loss_causal 0.6353 re_mapping 0.0115 re_causal 0.0314 /// teacc 98.80 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0855, 0.1240, -0.0265, ..., -0.0002, 0.0536, 0.0192], + [ 0.0805, -0.0628, 0.0198, ..., -0.0479, -0.0090, -0.0315], + [-0.0331, -0.0821, 0.0151, ..., -0.0341, 0.0416, -0.1046], + ..., + [-0.0891, -0.0600, 0.0459, ..., -0.0304, -0.0633, -0.0813], + [-0.0404, -0.0184, 0.0116, ..., 0.0162, -0.0777, 0.0036], + [-0.0207, 0.0022, 0.0553, ..., 0.0926, -0.0294, -0.0730]], + device='cuda:0'), grad: tensor([[ 5.4576e-07, -5.9873e-05, 8.0280e-07, ..., 5.6438e-07, + 1.0673e-06, 2.1458e-06], + [ 1.1921e-07, 2.7642e-06, 3.3174e-06, ..., 2.9057e-07, + 1.1306e-06, 2.1253e-06], + [ 5.0850e-07, 6.1877e-06, -4.0978e-08, ..., 1.9558e-06, + 1.2480e-07, 8.9481e-06], + ..., + [ 6.7987e-07, 1.0952e-06, -3.2634e-05, ..., 7.0967e-07, + 2.2352e-08, 7.6368e-07], + [ 2.7735e-06, -2.4572e-05, -1.0245e-07, ..., -1.9923e-05, + 7.4692e-07, -9.4354e-05], + [ 3.6303e-06, 2.2128e-05, 1.4901e-07, ..., 8.5682e-07, + 7.0781e-08, 2.2098e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0297, 0.0112, 0.0186, 0.0235, 0.0303, -0.0009, -0.0068, -0.0030, + 0.0054, -0.0321], device='cuda:0'), grad: tensor([-1.2982e-04, 2.8074e-05, -2.0847e-05, 5.6982e-05, 2.3335e-05, + 3.5852e-05, 1.0198e-04, -8.2195e-05, -9.1076e-05, 7.7367e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 217.86, cls_loss 0.0080 cls_loss_mapping 0.0099 cls_loss_causal 0.6104 re_mapping 0.0115 re_causal 0.0319 /// teacc 98.75 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0857, 0.1246, -0.0267, ..., -0.0003, 0.0537, 0.0197], + [ 0.0810, -0.0626, 0.0202, ..., -0.0477, -0.0088, -0.0316], + [-0.0332, -0.0831, 0.0145, ..., -0.0341, 0.0416, -0.1052], + ..., + [-0.0897, -0.0609, 0.0457, ..., -0.0306, -0.0633, -0.0815], + [-0.0408, -0.0177, 0.0121, ..., 0.0169, -0.0779, 0.0040], + [-0.0213, 0.0018, 0.0559, ..., 0.0928, -0.0296, -0.0741]], + device='cuda:0'), grad: tensor([[ 2.8968e-04, 1.6761e-04, 7.1943e-05, ..., 8.3819e-08, + 4.1351e-07, 1.3864e-04], + [ 4.8801e-06, 2.9542e-06, 1.2834e-06, ..., 1.0245e-07, + 3.9116e-08, 5.2601e-06], + [ 1.0133e-05, 5.8860e-06, -8.4750e-07, ..., 5.1968e-07, + -2.4401e-07, 8.9109e-06], + ..., + [ 4.6566e-06, 2.4624e-06, -2.3097e-05, ..., 1.7509e-07, + -8.4005e-07, 3.1218e-06], + [ 7.2382e-06, 2.7623e-06, 1.2014e-06, ..., 2.6822e-06, + 1.1362e-07, 2.9415e-05], + [ 8.0764e-05, 4.7326e-05, 2.7269e-05, ..., -8.1211e-07, + 3.7812e-07, 4.7475e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0299, 0.0112, 0.0184, 0.0235, 0.0307, -0.0014, -0.0065, -0.0030, + 0.0057, -0.0325], device='cuda:0'), grad: tensor([ 6.5470e-04, 2.7955e-05, -1.7941e-05, 7.0035e-05, 6.0424e-06, + -1.1320e-03, 1.2350e-04, -2.1964e-05, 8.0407e-05, 2.1064e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 94---------------------------------------------------- +epoch 94, time 218.54, cls_loss 0.0068 cls_loss_mapping 0.0108 cls_loss_causal 0.5962 re_mapping 0.0108 re_causal 0.0307 /// teacc 98.84 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0860, 0.1255, -0.0265, ..., -0.0004, 0.0539, 0.0200], + [ 0.0811, -0.0631, 0.0202, ..., -0.0479, -0.0089, -0.0323], + [-0.0335, -0.0831, 0.0142, ..., -0.0343, 0.0416, -0.1058], + ..., + [-0.0907, -0.0620, 0.0460, ..., -0.0308, -0.0630, -0.0826], + [-0.0407, -0.0175, 0.0133, ..., 0.0175, -0.0774, 0.0045], + [-0.0222, 0.0011, 0.0556, ..., 0.0930, -0.0302, -0.0754]], + device='cuda:0'), grad: tensor([[ 1.7732e-06, -7.2457e-07, 1.7919e-06, ..., 6.2585e-07, + 1.4883e-06, 2.2426e-06], + [-2.5798e-06, 2.4699e-06, -2.1011e-05, ..., 6.5751e-07, + 1.2107e-07, 1.2048e-05], + [ 5.9158e-06, -1.4901e-07, 1.5482e-05, ..., 8.6576e-06, + 5.3272e-07, 2.3678e-05], + ..., + [ 9.4026e-06, 5.6438e-07, 9.6634e-06, ..., 7.8976e-06, + 5.5321e-07, 1.1548e-05], + [ 1.6779e-05, 2.3860e-06, 9.3579e-05, ..., 6.8784e-05, + 1.2815e-06, 1.4567e-04], + [ 5.6997e-06, 3.2950e-06, -3.5644e-04, ..., -1.2696e-04, + 1.0803e-06, -2.7761e-05]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0299, 0.0111, 0.0185, 0.0230, 0.0312, -0.0010, -0.0055, -0.0034, + 0.0062, -0.0333], device='cuda:0'), grad: tensor([ 1.5214e-05, -1.2100e-05, 6.3002e-05, -4.2295e-04, 7.2432e-04, + -1.3210e-05, -1.6913e-06, 3.9816e-05, 4.0102e-04, -7.9393e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 217.59, cls_loss 0.0085 cls_loss_mapping 0.0125 cls_loss_causal 0.6230 re_mapping 0.0107 re_causal 0.0311 /// teacc 98.82 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0860, 0.1246, -0.0282, ..., -0.0004, 0.0544, 0.0208], + [ 0.0826, -0.0633, 0.0215, ..., -0.0478, -0.0089, -0.0316], + [-0.0339, -0.0833, 0.0138, ..., -0.0345, 0.0418, -0.1073], + ..., + [-0.0921, -0.0623, 0.0471, ..., -0.0311, -0.0629, -0.0826], + [-0.0413, -0.0183, 0.0119, ..., 0.0178, -0.0775, 0.0041], + [-0.0233, 0.0025, 0.0565, ..., 0.0933, -0.0306, -0.0768]], + device='cuda:0'), grad: tensor([[-1.0958e-03, -2.2411e-03, -9.4175e-04, ..., 7.6368e-08, + 1.1548e-05, -3.6979e-04], + [ 1.8440e-06, 6.7204e-06, 5.7630e-06, ..., 2.0862e-07, + 1.7192e-06, 1.1027e-06], + [ 2.6040e-06, 1.4700e-05, 1.0318e-04, ..., 4.1351e-07, + 5.8889e-05, 1.9874e-06], + ..., + [ 5.8562e-06, 1.2442e-05, -9.7752e-05, ..., 1.1865e-06, + -6.2585e-05, 2.1644e-06], + [ 7.8157e-06, 1.0543e-05, 2.5202e-06, ..., -6.3889e-07, + 1.7695e-06, -3.5763e-07], + [ 7.7200e-04, 1.6079e-03, 6.8283e-04, ..., -5.6028e-06, + 1.9912e-06, 2.6131e-04]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0280, 0.0120, 0.0184, 0.0228, 0.0304, -0.0005, -0.0044, -0.0027, + 0.0047, -0.0327], device='cuda:0'), grad: tensor([-5.2414e-03, 1.3423e-04, 5.9992e-05, 7.6771e-05, 4.4537e-04, + 9.5654e-04, -1.1659e-04, -1.8406e-04, 3.8207e-05, 3.8300e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 217.56, cls_loss 0.0085 cls_loss_mapping 0.0103 cls_loss_causal 0.6439 re_mapping 0.0113 re_causal 0.0315 /// teacc 98.70 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0860, 0.1279, -0.0258, ..., 0.0013, 0.0544, 0.0209], + [ 0.0841, -0.0639, 0.0228, ..., -0.0478, -0.0090, -0.0310], + [-0.0345, -0.0835, 0.0147, ..., -0.0347, 0.0416, -0.1076], + ..., + [-0.0940, -0.0631, 0.0465, ..., -0.0312, -0.0630, -0.0837], + [-0.0421, -0.0191, 0.0118, ..., 0.0178, -0.0776, 0.0038], + [-0.0237, -0.0006, 0.0550, ..., 0.0927, -0.0307, -0.0783]], + device='cuda:0'), grad: tensor([[ 3.0920e-07, 1.3240e-05, 5.5164e-05, ..., 4.2468e-07, + 8.5682e-08, 1.8269e-05], + [-3.4552e-06, 7.8231e-06, 7.1600e-06, ..., 2.1607e-07, + 9.3132e-09, 2.4084e-06], + [ 8.7358e-07, 1.7002e-05, 2.6420e-05, ..., 2.2892e-06, + 9.3132e-09, 8.7321e-06], + ..., + [ 1.5032e-06, 2.1696e-05, 2.7120e-05, ..., 1.4920e-06, + 0.0000e+00, 7.3090e-06], + [ 2.8498e-07, -1.4234e-04, -1.9300e-04, ..., -9.8906e-07, + 3.2224e-07, -5.3406e-05], + [ 2.0321e-06, 5.9277e-05, 2.5600e-05, ..., -1.3128e-05, + 1.4901e-08, 1.3508e-05]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0299, 0.0130, 0.0189, 0.0230, 0.0307, 0.0004, -0.0051, -0.0034, + 0.0039, -0.0343], device='cuda:0'), grad: tensor([ 4.7624e-05, 1.7822e-05, 4.7773e-05, 1.7554e-05, 2.8722e-06, + 2.7418e-05, -2.3972e-06, 5.5522e-05, -2.7847e-04, 6.4492e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 217.68, cls_loss 0.0056 cls_loss_mapping 0.0090 cls_loss_causal 0.6219 re_mapping 0.0112 re_causal 0.0318 /// teacc 98.69 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0862, 0.1283, -0.0256, ..., 0.0012, 0.0546, 0.0208], + [ 0.0843, -0.0644, 0.0216, ..., -0.0481, -0.0091, -0.0324], + [-0.0347, -0.0840, 0.0145, ..., -0.0350, 0.0415, -0.1080], + ..., + [-0.0945, -0.0640, 0.0475, ..., -0.0313, -0.0629, -0.0837], + [-0.0420, -0.0184, 0.0125, ..., 0.0177, -0.0777, 0.0046], + [-0.0239, -0.0004, 0.0554, ..., 0.0933, -0.0306, -0.0780]], + device='cuda:0'), grad: tensor([[ 7.5065e-07, -4.0196e-06, -3.8892e-06, ..., 1.1176e-07, + 5.8115e-07, 1.1362e-06], + [-2.6915e-06, 9.2760e-07, -3.1069e-06, ..., 4.0606e-07, + 2.4959e-07, 2.5798e-06], + [ 1.9576e-06, 1.5855e-05, 2.3395e-05, ..., 1.3374e-05, + 3.8557e-07, 1.9073e-05], + ..., + [ 2.7269e-06, 2.6599e-06, 2.4401e-06, ..., 3.0734e-07, + 2.9989e-07, 3.7253e-08], + [ 3.8128e-06, -1.6108e-05, -2.8685e-05, ..., -1.7166e-05, + 1.1921e-07, -1.2584e-05], + [ 6.1654e-06, 7.6368e-06, 3.6899e-06, ..., 6.4075e-07, + 7.3574e-07, 1.0014e-05]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0301, 0.0118, 0.0191, 0.0230, 0.0307, 0.0002, -0.0052, -0.0027, + 0.0039, -0.0343], device='cuda:0'), grad: tensor([ 6.8173e-07, 3.9302e-07, 4.7863e-05, -7.5281e-05, -1.2167e-05, + 4.9263e-05, -2.2054e-06, 3.7253e-06, -3.5793e-05, 2.3380e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 217.92, cls_loss 0.0064 cls_loss_mapping 0.0080 cls_loss_causal 0.6081 re_mapping 0.0109 re_causal 0.0312 /// teacc 98.76 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0864, 0.1286, -0.0257, ..., 0.0011, 0.0547, 0.0209], + [ 0.0843, -0.0647, 0.0212, ..., -0.0482, -0.0091, -0.0328], + [-0.0346, -0.0848, 0.0148, ..., -0.0352, 0.0414, -0.1090], + ..., + [-0.0945, -0.0643, 0.0476, ..., -0.0315, -0.0629, -0.0837], + [-0.0422, -0.0189, 0.0122, ..., 0.0176, -0.0777, 0.0048], + [-0.0241, 0.0003, 0.0562, ..., 0.0938, -0.0306, -0.0771]], + device='cuda:0'), grad: tensor([[ 1.6186e-06, -4.0174e-05, -2.3320e-05, ..., 8.7544e-08, + 4.0978e-08, -2.9597e-06], + [-1.2964e-05, 1.2629e-05, -7.3686e-06, ..., 1.4901e-08, + 1.4901e-08, 2.1961e-06], + [ 5.3197e-06, 3.6079e-06, 3.6508e-06, ..., 1.5646e-07, + 1.4901e-08, 9.7901e-06], + ..., + [ 9.2201e-07, 2.2221e-06, -1.4734e-06, ..., 1.8068e-07, + 0.0000e+00, 3.5223e-06], + [ 7.8157e-06, -1.4175e-06, -4.0345e-06, ..., -1.3132e-06, + 6.3330e-08, -8.3596e-06], + [ 8.4639e-06, 6.7353e-06, 7.4804e-06, ..., 0.0000e+00, + 3.7253e-09, 1.7926e-05]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0302, 0.0112, 0.0197, 0.0231, 0.0308, -0.0004, -0.0053, -0.0026, + 0.0036, -0.0339], device='cuda:0'), grad: tensor([-5.9903e-05, -7.6443e-06, 2.1234e-05, -1.5867e-04, 4.5076e-06, + 1.2982e-04, 2.9787e-05, 5.8301e-07, -1.8068e-07, 4.0293e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 218.26, cls_loss 0.0069 cls_loss_mapping 0.0088 cls_loss_causal 0.6035 re_mapping 0.0103 re_causal 0.0298 /// teacc 98.70 lr 0.00010000 +Epoch 101, weight, value: tensor([[-8.6576e-02, 1.2912e-01, -2.5564e-02, ..., 1.0502e-03, + 5.4541e-02, 2.1107e-02], + [ 8.4409e-02, -6.4071e-02, 1.9164e-02, ..., -4.8214e-02, + -9.1844e-03, -3.3484e-02], + [-3.4702e-02, -8.4912e-02, 1.5938e-02, ..., -3.5060e-02, + 4.1721e-02, -1.0967e-01], + ..., + [-9.4051e-02, -6.4565e-02, 4.9432e-02, ..., -3.1559e-02, + -6.3075e-02, -8.3385e-02], + [-4.2424e-02, -1.8665e-02, 1.2893e-02, ..., 1.7994e-02, + -7.7767e-02, 5.5522e-03], + [-2.5133e-02, 1.0196e-05, 5.5866e-02, ..., 9.4030e-02, + -3.0692e-02, -7.8095e-02]], device='cuda:0'), grad: tensor([[ 1.6764e-08, -3.6955e-05, -1.1154e-05, ..., 1.4156e-07, + 3.7253e-09, 1.1548e-07], + [-8.7358e-07, 2.0675e-07, 4.4517e-07, ..., 1.6391e-07, + 7.4506e-09, 1.5274e-07], + [ 1.8254e-07, 2.6207e-06, 3.8296e-06, ..., 1.3039e-07, + -3.2224e-07, 3.5223e-06], + ..., + [ 3.1106e-07, 1.8794e-06, 1.1064e-05, ..., 3.6396e-06, + 3.7253e-09, 2.9616e-07], + [ 1.1176e-07, 1.5516e-06, 3.8408e-06, ..., 1.3486e-06, + 3.7253e-09, 1.2275e-06], + [ 9.6858e-08, 1.8775e-05, -3.1978e-05, ..., -1.2793e-05, + 0.0000e+00, -4.0978e-07]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0303, 0.0086, 0.0209, 0.0228, 0.0304, -0.0005, -0.0052, -0.0005, + 0.0039, -0.0346], device='cuda:0'), grad: tensor([-4.1217e-05, 1.1679e-06, 1.5557e-05, -4.9025e-06, 2.2209e-04, + 8.7321e-06, -2.0933e-04, 1.6198e-05, 1.0982e-05, -1.8984e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 218.11, cls_loss 0.0073 cls_loss_mapping 0.0077 cls_loss_causal 0.5786 re_mapping 0.0103 re_causal 0.0290 /// teacc 98.73 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0867, 0.1285, -0.0270, ..., 0.0010, 0.0546, 0.0210], + [ 0.0854, -0.0639, 0.0202, ..., -0.0482, -0.0091, -0.0326], + [-0.0351, -0.0851, 0.0159, ..., -0.0351, 0.0416, -0.1100], + ..., + [-0.0950, -0.0649, 0.0489, ..., -0.0317, -0.0631, -0.0848], + [-0.0427, -0.0186, 0.0132, ..., 0.0186, -0.0778, 0.0057], + [-0.0255, 0.0008, 0.0570, ..., 0.0945, -0.0308, -0.0793]], + device='cuda:0'), grad: tensor([[ 2.7139e-06, -4.0472e-05, -2.9996e-05, ..., 4.2655e-07, + 7.6368e-08, -1.9208e-05], + [-5.9962e-05, 1.1381e-06, -2.7344e-05, ..., 6.9849e-07, + 4.0978e-08, 1.4640e-06], + [ 2.3767e-05, 2.0098e-06, 1.2606e-05, ..., 4.1351e-07, + -3.0547e-07, 1.8794e-06], + ..., + [ 8.4192e-06, 2.8946e-06, 5.8860e-06, ..., 3.8054e-06, + 4.4703e-08, 1.8552e-06], + [ 7.5921e-06, 1.0803e-06, -6.3181e-06, ..., -3.4310e-06, + 5.0478e-07, -2.7493e-06], + [ 1.3866e-05, 2.9519e-05, -1.6555e-05, ..., -1.5825e-05, + 9.3132e-09, 2.3097e-05]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0291, 0.0095, 0.0210, 0.0235, 0.0306, -0.0004, -0.0058, -0.0011, + 0.0039, -0.0341], device='cuda:0'), grad: tensor([-7.6056e-05, -8.5056e-05, 3.8713e-05, 4.7714e-05, -1.4710e-04, + -1.5780e-05, 1.8492e-05, 2.9758e-05, 2.6226e-05, 1.6272e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 217.90, cls_loss 0.0070 cls_loss_mapping 0.0098 cls_loss_causal 0.5925 re_mapping 0.0100 re_causal 0.0300 /// teacc 98.62 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0869, 0.1281, -0.0281, ..., 0.0006, 0.0546, 0.0204], + [ 0.0861, -0.0640, 0.0205, ..., -0.0481, -0.0091, -0.0326], + [-0.0354, -0.0863, 0.0151, ..., -0.0352, 0.0416, -0.1109], + ..., + [-0.0954, -0.0658, 0.0490, ..., -0.0319, -0.0631, -0.0854], + [-0.0430, -0.0173, 0.0146, ..., 0.0198, -0.0778, 0.0070], + [-0.0261, 0.0011, 0.0575, ..., 0.0952, -0.0304, -0.0805]], + device='cuda:0'), grad: tensor([[ 1.2498e-06, -1.4275e-05, -8.0168e-06, ..., 1.5460e-07, + 0.0000e+00, -2.0489e-08], + [-3.0518e-04, 7.9200e-06, -3.3212e-04, ..., 1.0803e-07, + 0.0000e+00, 8.7619e-06], + [ 2.9755e-04, 5.6252e-06, 3.0804e-04, ..., 3.1237e-06, + 0.0000e+00, 8.2925e-06], + ..., + [ 3.6601e-06, 1.8533e-06, 5.7518e-05, ..., 3.0100e-06, + 0.0000e+00, 1.4920e-06], + [ 1.1623e-05, -1.4834e-05, -4.3154e-05, ..., -5.3495e-06, + 0.0000e+00, -2.5615e-05], + [ 7.0035e-07, 9.6336e-06, -1.5691e-05, ..., -3.2354e-06, + 0.0000e+00, 3.0138e-06]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0282, 0.0098, 0.0206, 0.0240, 0.0324, -0.0010, -0.0059, -0.0015, + 0.0050, -0.0347], device='cuda:0'), grad: tensor([-1.0341e-05, -5.7602e-04, 5.0545e-04, 2.1577e-05, 3.6564e-06, + 1.7834e-04, -1.9908e-04, 1.1230e-04, -2.4319e-05, -1.1772e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 217.93, cls_loss 0.0054 cls_loss_mapping 0.0073 cls_loss_causal 0.6279 re_mapping 0.0096 re_causal 0.0289 /// teacc 98.81 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0870, 0.1287, -0.0279, ..., 0.0005, 0.0546, 0.0203], + [ 0.0874, -0.0643, 0.0209, ..., -0.0483, -0.0092, -0.0329], + [-0.0364, -0.0869, 0.0146, ..., -0.0354, 0.0416, -0.1112], + ..., + [-0.0965, -0.0663, 0.0490, ..., -0.0323, -0.0632, -0.0857], + [-0.0440, -0.0177, 0.0146, ..., 0.0202, -0.0785, 0.0074], + [-0.0266, 0.0007, 0.0582, ..., 0.0965, -0.0305, -0.0811]], + device='cuda:0'), grad: tensor([[ 1.5460e-07, -1.6704e-05, -1.1027e-05, ..., 5.0478e-07, + 0.0000e+00, 2.6263e-07], + [-3.4392e-05, 6.1654e-07, -1.8835e-05, ..., 1.2852e-07, + 0.0000e+00, 2.3656e-07], + [ 2.1875e-05, 3.7719e-06, 1.6496e-05, ..., 3.7067e-07, + 0.0000e+00, 1.2945e-06], + ..., + [ 1.1083e-06, 1.0822e-06, 3.1501e-05, ..., 1.2770e-05, + 0.0000e+00, 1.0058e-06], + [ 2.6599e-06, -4.1351e-06, 3.1181e-06, ..., 3.2131e-06, + 0.0000e+00, -8.9407e-06], + [ 3.0007e-06, 6.5640e-06, -5.8979e-05, ..., -2.7567e-05, + 0.0000e+00, 2.1234e-07]], device='cuda:0') +Epoch 104, bias, value: tensor([ 2.8511e-02, 1.0357e-02, 2.0003e-02, 2.3835e-02, 3.1225e-02, + 7.3123e-05, -6.5841e-03, -1.5866e-03, 4.6656e-03, -3.4083e-02], + device='cuda:0'), grad: tensor([-1.8120e-05, -3.7760e-05, 3.1769e-05, 2.3425e-05, -3.4243e-05, + 1.3575e-05, 6.4373e-06, 3.6269e-05, -1.8515e-06, -1.9610e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 217.48, cls_loss 0.0065 cls_loss_mapping 0.0096 cls_loss_causal 0.6101 re_mapping 0.0101 re_causal 0.0308 /// teacc 98.77 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0872, 0.1296, -0.0272, ..., 0.0021, 0.0548, 0.0204], + [ 0.0882, -0.0641, 0.0215, ..., -0.0479, -0.0089, -0.0331], + [-0.0367, -0.0875, 0.0141, ..., -0.0357, 0.0416, -0.1118], + ..., + [-0.0971, -0.0667, 0.0490, ..., -0.0327, -0.0632, -0.0859], + [-0.0440, -0.0161, 0.0162, ..., 0.0209, -0.0786, 0.0093], + [-0.0277, -0.0006, 0.0571, ..., 0.0958, -0.0304, -0.0836]], + device='cuda:0'), grad: tensor([[ 3.1024e-05, -9.3758e-05, 8.1301e-05, ..., 6.0409e-05, + 4.0233e-07, 6.6869e-07], + [ 1.3039e-08, 1.2163e-06, 3.1814e-06, ..., 1.3690e-06, + 8.0094e-08, 8.8662e-07], + [ 6.5193e-06, 1.9193e-05, -9.9763e-06, ..., 1.2748e-05, + 9.6858e-08, 3.2540e-06], + ..., + [ 2.0396e-06, 1.2472e-05, -7.4133e-06, ..., 4.3884e-06, + 3.7253e-09, -1.0610e-05], + [ 1.3541e-06, 5.2489e-06, 2.5168e-05, ..., 1.6853e-05, + 1.7323e-07, -2.8498e-06], + [-6.0350e-05, -7.5251e-06, -1.6868e-04, ..., -1.3721e-04, + 2.4214e-08, 5.9977e-06]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0291, 0.0108, 0.0199, 0.0247, 0.0315, -0.0012, -0.0065, -0.0016, + 0.0061, -0.0354], device='cuda:0'), grad: tensor([-1.2894e-03, 1.8597e-05, -1.1230e-04, 1.5378e-04, 1.4269e-04, + 2.3901e-04, 2.2626e-04, 1.1724e-04, 1.6391e-04, 3.3951e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 217.57, cls_loss 0.0055 cls_loss_mapping 0.0074 cls_loss_causal 0.5752 re_mapping 0.0102 re_causal 0.0286 /// teacc 98.73 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0876, 0.1301, -0.0270, ..., 0.0021, 0.0550, 0.0203], + [ 0.0876, -0.0639, 0.0211, ..., -0.0481, -0.0089, -0.0333], + [-0.0355, -0.0883, 0.0140, ..., -0.0361, 0.0398, -0.1128], + ..., + [-0.0974, -0.0675, 0.0494, ..., -0.0333, -0.0634, -0.0861], + [-0.0444, -0.0159, 0.0165, ..., 0.0214, -0.0792, 0.0100], + [-0.0286, -0.0009, 0.0574, ..., 0.0962, -0.0305, -0.0846]], + device='cuda:0'), grad: tensor([[-2.0824e-06, -4.1187e-05, -1.6212e-05, ..., 1.3039e-07, + 2.0731e-06, 1.6950e-07], + [-3.4273e-07, 5.2527e-07, 8.1584e-07, ..., 2.2538e-07, + 1.7881e-07, 1.2293e-07], + [ 1.3225e-07, -4.1351e-07, 1.8515e-06, ..., 4.8988e-07, + 5.9605e-08, 3.5204e-07], + ..., + [ 1.8626e-07, 9.0152e-07, -2.3078e-06, ..., 4.0121e-06, + 1.3039e-08, 9.1270e-08], + [ 1.9800e-06, 2.0470e-06, -3.1851e-07, ..., -1.0654e-06, + 2.8238e-06, 1.3337e-06], + [ 1.7416e-06, 2.2858e-05, -8.1435e-06, ..., -6.5938e-06, + 7.2643e-08, 7.6741e-07]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0294, 0.0098, 0.0204, 0.0235, 0.0314, -0.0002, -0.0058, -0.0014, + 0.0060, -0.0356], device='cuda:0'), grad: tensor([-5.4300e-05, 2.2054e-06, -4.1574e-06, 6.0275e-06, 1.7643e-05, + 2.2471e-05, -1.3307e-05, -7.0781e-06, 1.2808e-05, 1.7673e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 217.57, cls_loss 0.0084 cls_loss_mapping 0.0110 cls_loss_causal 0.5617 re_mapping 0.0100 re_causal 0.0279 /// teacc 98.74 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0878, 0.1300, -0.0278, ..., 0.0013, 0.0555, 0.0192], + [ 0.0885, -0.0643, 0.0208, ..., -0.0483, -0.0094, -0.0336], + [-0.0366, -0.0911, 0.0140, ..., -0.0365, 0.0400, -0.1140], + ..., + [-0.0981, -0.0686, 0.0498, ..., -0.0339, -0.0648, -0.0865], + [-0.0449, -0.0152, 0.0169, ..., 0.0218, -0.0800, 0.0099], + [-0.0295, -0.0007, 0.0580, ..., 0.0971, -0.0314, -0.0852]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, -4.9919e-07, 1.0040e-06, ..., 4.2841e-08, + 3.8557e-07, 2.1607e-07], + [-1.6093e-06, 1.6950e-07, 3.2466e-06, ..., 1.5832e-07, + 4.2841e-08, 1.9930e-07], + [ 4.4703e-07, 2.1532e-06, 3.9637e-06, ..., 1.8775e-06, + 2.4214e-08, 2.8927e-06], + ..., + [ 9.2015e-07, 9.4995e-08, 2.7180e-04, ..., 6.7055e-07, + 1.8626e-09, 5.1223e-07], + [-2.8498e-07, -2.2147e-06, -4.4405e-06, ..., -3.2522e-06, + 4.6007e-07, -4.9397e-06], + [ 2.4028e-07, 1.2852e-06, 4.8995e-05, ..., -4.2468e-07, + 2.7940e-08, 3.8743e-07]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0288, 0.0093, 0.0203, 0.0238, 0.0315, 0.0004, -0.0050, -0.0012, + 0.0051, -0.0354], device='cuda:0'), grad: tensor([ 4.1276e-06, 2.0146e-05, -2.0504e-05, 8.0615e-06, -9.9468e-04, + 1.8254e-05, -1.3381e-05, 8.2731e-04, 2.2694e-05, 1.2755e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 217.52, cls_loss 0.0080 cls_loss_mapping 0.0091 cls_loss_causal 0.6044 re_mapping 0.0100 re_causal 0.0282 /// teacc 98.79 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0880, 0.1304, -0.0278, ..., 0.0011, 0.0558, 0.0189], + [ 0.0883, -0.0651, 0.0212, ..., -0.0502, -0.0096, -0.0338], + [-0.0367, -0.0914, 0.0144, ..., -0.0366, 0.0400, -0.1141], + ..., + [-0.0987, -0.0694, 0.0495, ..., -0.0341, -0.0649, -0.0867], + [-0.0448, -0.0135, 0.0173, ..., 0.0220, -0.0791, 0.0106], + [-0.0286, -0.0010, 0.0585, ..., 0.0985, -0.0310, -0.0858]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, -3.6526e-06, 1.0356e-05, ..., 1.6019e-07, + 9.6858e-08, 2.4587e-07], + [-8.3819e-08, 6.3330e-08, 6.6049e-06, ..., 2.4959e-07, + 9.3132e-09, 3.9488e-07], + [ 9.4995e-08, -3.6322e-07, -6.8307e-05, ..., 1.6391e-07, + 1.4901e-08, 7.8604e-07], + ..., + [ 7.8231e-08, 8.7544e-08, -8.7246e-06, ..., 2.7362e-06, + 0.0000e+00, 3.0044e-06], + [ 6.1840e-07, 9.1270e-07, 1.8358e-05, ..., 4.1723e-07, + 4.8429e-08, 5.3830e-07], + [ 9.4809e-07, 1.5907e-06, -8.2433e-05, ..., -2.1324e-05, + 3.7253e-09, -2.4289e-05]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0289, 0.0095, 0.0208, 0.0236, 0.0316, -0.0002, -0.0048, -0.0014, + 0.0054, -0.0355], device='cuda:0'), grad: tensor([ 4.0531e-05, 1.6153e-05, -1.9836e-04, 5.5730e-05, 1.2314e-04, + 2.2024e-05, -9.2760e-06, -1.0580e-05, 5.0247e-05, -8.9884e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 217.58, cls_loss 0.0065 cls_loss_mapping 0.0091 cls_loss_causal 0.5835 re_mapping 0.0095 re_causal 0.0273 /// teacc 98.76 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0882, 0.1312, -0.0276, ..., 0.0009, 0.0562, 0.0195], + [ 0.0885, -0.0646, 0.0211, ..., -0.0510, -0.0094, -0.0339], + [-0.0370, -0.0921, 0.0141, ..., -0.0369, 0.0400, -0.1155], + ..., + [-0.0989, -0.0704, 0.0496, ..., -0.0345, -0.0650, -0.0869], + [-0.0454, -0.0140, 0.0169, ..., 0.0217, -0.0796, 0.0106], + [-0.0279, -0.0012, 0.0593, ..., 0.0997, -0.0316, -0.0863]], + device='cuda:0'), grad: tensor([[ 3.3528e-07, -4.4703e-08, 1.1828e-06, ..., 1.4715e-07, + 8.0094e-08, 5.1968e-07], + [-3.7253e-06, 2.2352e-08, 9.4771e-06, ..., 7.8417e-07, + 9.3132e-09, 8.3819e-08], + [ 2.0266e-06, 3.9116e-08, 6.2734e-06, ..., 5.2527e-07, + 9.3132e-09, 1.8422e-06], + ..., + [-1.2387e-06, 3.5390e-08, 1.0155e-05, ..., 6.7353e-06, + 5.5879e-09, 2.9989e-07], + [ 1.0245e-06, 4.7684e-07, 2.1830e-06, ..., 2.1048e-07, + 5.5879e-08, 6.6869e-07], + [ 1.7770e-06, 6.5379e-07, -7.6473e-05, ..., -1.5274e-05, + 2.0489e-08, 2.2165e-07]], device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0296, 0.0091, 0.0193, 0.0232, 0.0317, -0.0003, -0.0041, -0.0005, + 0.0044, -0.0354], device='cuda:0'), grad: tensor([ 4.7013e-06, 2.0966e-05, 1.5110e-05, -6.1989e-06, 5.8591e-05, + 1.5963e-06, 3.2596e-07, 1.2808e-05, 6.0499e-06, -1.1396e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 217.73, cls_loss 0.0064 cls_loss_mapping 0.0088 cls_loss_causal 0.5699 re_mapping 0.0099 re_causal 0.0277 /// teacc 98.74 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0888, 0.1316, -0.0275, ..., 0.0008, 0.0571, 0.0191], + [ 0.0894, -0.0652, 0.0212, ..., -0.0509, -0.0096, -0.0340], + [-0.0372, -0.0933, 0.0135, ..., -0.0357, 0.0400, -0.1168], + ..., + [-0.0997, -0.0709, 0.0498, ..., -0.0350, -0.0652, -0.0885], + [-0.0459, -0.0138, 0.0173, ..., 0.0223, -0.0797, 0.0108], + [-0.0288, -0.0014, 0.0594, ..., 0.0999, -0.0319, -0.0869]], + device='cuda:0'), grad: tensor([[ 2.0135e-06, -4.1306e-05, -5.5730e-06, ..., 2.0172e-06, + -1.9856e-06, 2.8126e-07], + [-1.8626e-05, 4.6045e-06, -1.5169e-05, ..., 1.0747e-06, + 5.1968e-07, 2.1979e-07], + [-2.1505e-04, 4.6715e-06, 1.8969e-05, ..., -1.5986e-04, + -4.1910e-07, 1.6857e-06], + ..., + [ 3.7774e-06, 5.1707e-06, -1.0982e-05, ..., 2.7437e-06, + 5.8487e-07, 8.7731e-07], + [ 2.5667e-06, -6.9067e-06, -9.1456e-07, ..., 1.5832e-06, + 1.6205e-07, -4.8205e-06], + [ 2.1446e-04, 8.3745e-06, -6.8474e-04, ..., -1.7583e-04, + 4.0419e-07, 1.2517e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([ 2.9716e-02, 9.0112e-03, 1.9957e-02, 2.4521e-02, 3.1908e-02, + -3.2295e-06, -4.4256e-03, -1.2520e-03, 4.4769e-03, -3.5926e-02], + device='cuda:0'), grad: tensor([-6.0052e-05, -1.2517e-05, -8.2636e-04, 2.2754e-05, 1.5898e-03, + 4.9621e-05, 1.9804e-05, -1.4521e-05, 2.1100e-05, -7.8964e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 217.67, cls_loss 0.0071 cls_loss_mapping 0.0095 cls_loss_causal 0.5767 re_mapping 0.0102 re_causal 0.0291 /// teacc 98.83 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0885, 0.1329, -0.0268, ..., 0.0022, 0.0572, 0.0190], + [ 0.0914, -0.0652, 0.0216, ..., -0.0509, -0.0088, -0.0342], + [-0.0386, -0.0940, 0.0127, ..., -0.0355, 0.0401, -0.1175], + ..., + [-0.1007, -0.0712, 0.0502, ..., -0.0354, -0.0658, -0.0890], + [-0.0461, -0.0134, 0.0173, ..., 0.0223, -0.0794, 0.0113], + [-0.0292, -0.0025, 0.0591, ..., 0.1004, -0.0322, -0.0868]], + device='cuda:0'), grad: tensor([[ 4.9740e-05, -3.3677e-05, -3.1572e-06, ..., 5.3868e-06, + -7.2494e-06, -2.7474e-06], + [ 5.2482e-05, 6.6124e-07, -8.5235e-06, ..., 7.8231e-08, + -6.6310e-07, 1.3039e-07], + [-6.3848e-04, 1.8943e-06, 3.3062e-06, ..., 2.5146e-07, + 5.2527e-07, 4.6194e-07], + ..., + [ 4.8012e-05, 2.6934e-06, 8.6576e-06, ..., 5.2266e-06, + 2.6431e-06, 3.3528e-07], + [ 3.4291e-06, 4.0084e-06, 6.2212e-06, ..., 6.7428e-07, + 1.4286e-06, 2.1774e-06], + [ 1.2502e-05, 4.4405e-06, -2.5570e-05, ..., -1.4193e-05, + -3.3472e-06, -2.9020e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0309, 0.0097, 0.0194, 0.0243, 0.0304, 0.0002, -0.0046, -0.0008, + 0.0044, -0.0364], device='cuda:0'), grad: tensor([ 9.3043e-05, 1.6141e-04, -1.7462e-03, 1.2302e-03, -3.8910e-04, + 8.6188e-05, 2.6166e-05, 1.5402e-04, 2.9311e-05, 3.5357e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 218.17, cls_loss 0.0074 cls_loss_mapping 0.0088 cls_loss_causal 0.5804 re_mapping 0.0109 re_causal 0.0288 /// teacc 98.72 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0882, 0.1333, -0.0270, ..., 0.0019, 0.0572, 0.0187], + [ 0.0908, -0.0656, 0.0217, ..., -0.0512, -0.0099, -0.0346], + [-0.0396, -0.0947, 0.0117, ..., -0.0358, 0.0401, -0.1183], + ..., + [-0.1016, -0.0729, 0.0503, ..., -0.0358, -0.0658, -0.0892], + [-0.0484, -0.0136, 0.0180, ..., 0.0221, -0.0805, 0.0110], + [-0.0297, -0.0024, 0.0600, ..., 0.1014, -0.0317, -0.0873]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 1.7798e-06, 1.4544e-05, ..., 3.0566e-06, + 1.9092e-07, 2.7753e-06], + [-1.7788e-07, 1.5087e-05, 5.7429e-05, ..., 1.0893e-05, + 7.0781e-08, 9.7379e-06], + [ 2.7940e-09, 9.2611e-06, 3.8058e-05, ..., 6.3777e-06, + -9.3132e-07, 5.9456e-06], + ..., + [ 1.0710e-07, 5.8711e-06, 2.5965e-06, ..., 5.9567e-06, + 6.8918e-08, 4.4629e-06], + [ 6.4727e-07, -8.3864e-05, -2.9635e-04, ..., -5.4866e-05, + 3.4273e-07, -5.4806e-05], + [ 5.6811e-07, 4.1366e-05, 1.1742e-04, ..., 1.8924e-05, + 8.2888e-08, 2.6971e-05]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0310, 0.0092, 0.0189, 0.0231, 0.0319, 0.0011, -0.0048, -0.0008, + 0.0037, -0.0360], device='cuda:0'), grad: tensor([ 2.1294e-05, 8.1539e-05, 4.9353e-05, 5.6148e-05, -9.8571e-06, + 2.8193e-05, 9.4324e-06, -1.9178e-05, -3.9983e-04, 1.8346e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 217.70, cls_loss 0.0046 cls_loss_mapping 0.0065 cls_loss_causal 0.5725 re_mapping 0.0100 re_causal 0.0291 /// teacc 98.69 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0884, 0.1339, -0.0269, ..., 0.0020, 0.0573, 0.0184], + [ 0.0929, -0.0655, 0.0222, ..., -0.0513, -0.0093, -0.0345], + [-0.0420, -0.0948, 0.0103, ..., -0.0360, 0.0411, -0.1189], + ..., + [-0.1018, -0.0733, 0.0501, ..., -0.0361, -0.0666, -0.0893], + [-0.0481, -0.0125, 0.0189, ..., 0.0224, -0.0805, 0.0118], + [-0.0299, -0.0030, 0.0608, ..., 0.1017, -0.0321, -0.0881]], + device='cuda:0'), grad: tensor([[ 7.6368e-08, -4.4983e-07, 2.0042e-06, ..., 3.8184e-08, + -9.7789e-08, 3.8557e-07], + [-3.9823e-06, 3.9116e-08, 2.4885e-06, ..., 5.7742e-08, + 2.3283e-08, 1.3299e-06], + [ 3.7719e-06, 9.1270e-08, 1.6272e-04, ..., 2.9802e-08, + -1.0710e-07, 2.1100e-05], + ..., + [ 1.0431e-07, 2.5146e-08, -4.4489e-04, ..., 1.1409e-06, + 1.1362e-07, -5.4449e-05], + [ 3.6415e-07, -1.4063e-07, 8.0645e-05, ..., -7.5996e-07, + 1.2107e-07, -5.8953e-07], + [ 2.1700e-07, 3.9581e-07, 7.9274e-05, ..., -5.7928e-06, + 7.3574e-08, 9.1121e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0310, 0.0103, 0.0178, 0.0235, 0.0311, 0.0004, -0.0050, -0.0009, + 0.0044, -0.0356], device='cuda:0'), grad: tensor([ 2.5984e-06, 2.6487e-06, 2.0027e-04, 1.2910e-04, 1.1295e-05, + 8.5682e-06, 1.9707e-06, -5.2786e-04, 8.8394e-05, 8.3566e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 217.79, cls_loss 0.0086 cls_loss_mapping 0.0111 cls_loss_causal 0.6142 re_mapping 0.0100 re_causal 0.0294 /// teacc 98.81 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0887, 0.1333, -0.0281, ..., 0.0019, 0.0578, 0.0174], + [ 0.0933, -0.0661, 0.0223, ..., -0.0513, -0.0099, -0.0353], + [-0.0421, -0.0957, 0.0085, ..., -0.0361, 0.0408, -0.1194], + ..., + [-0.1021, -0.0739, 0.0498, ..., -0.0365, -0.0670, -0.0891], + [-0.0485, -0.0115, 0.0192, ..., 0.0223, -0.0795, 0.0122], + [-0.0303, -0.0022, 0.0639, ..., 0.1025, -0.0350, -0.0886]], + device='cuda:0'), grad: tensor([[ 2.4550e-06, -3.2112e-06, 1.8813e-07, ..., 2.7474e-07, + -2.1979e-07, 3.9339e-06], + [ 1.0982e-05, 5.7463e-07, 5.3123e-06, ..., 3.2596e-07, + -1.4342e-07, 2.0027e-05], + [ 4.6164e-05, 2.0210e-07, 2.2963e-05, ..., 2.2817e-07, + 7.2643e-08, 7.2658e-05], + ..., + [ 1.1221e-05, 3.7160e-07, 7.7307e-05, ..., 1.7598e-05, + 6.4261e-08, 2.9728e-05], + [-2.3913e-04, 1.6429e-06, -1.0449e-04, ..., 2.0266e-06, + 2.8592e-07, -3.7479e-04], + [ 9.5144e-06, 4.4480e-06, -6.2180e-04, ..., -2.2388e-04, + 2.5425e-07, -2.7105e-05]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0298, 0.0100, 0.0172, 0.0233, 0.0301, 0.0001, -0.0044, -0.0010, + 0.0043, -0.0335], device='cuda:0'), grad: tensor([ 2.6435e-05, 6.0707e-05, 2.2197e-04, 3.4738e-04, 8.4209e-04, + 2.9373e-04, 1.2189e-04, 2.4915e-04, -1.0977e-03, -1.0653e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 217.19, cls_loss 0.0072 cls_loss_mapping 0.0092 cls_loss_causal 0.5784 re_mapping 0.0097 re_causal 0.0284 /// teacc 98.76 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0887, 0.1336, -0.0286, ..., 0.0009, 0.0582, 0.0167], + [ 0.0934, -0.0664, 0.0222, ..., -0.0514, -0.0093, -0.0372], + [-0.0419, -0.0966, 0.0085, ..., -0.0363, 0.0408, -0.1205], + ..., + [-0.1022, -0.0751, 0.0503, ..., -0.0371, -0.0676, -0.0884], + [-0.0488, -0.0113, 0.0197, ..., 0.0226, -0.0797, 0.0127], + [-0.0311, -0.0025, 0.0641, ..., 0.1035, -0.0354, -0.0893]], + device='cuda:0'), grad: tensor([[ 2.5555e-06, -1.6165e-04, -3.0026e-05, ..., 5.0012e-07, + -1.1899e-05, 1.1520e-06], + [-5.3368e-03, 4.9733e-06, -4.0169e-03, ..., 1.5497e-06, + 1.6168e-06, 3.4925e-07], + [ 5.2719e-03, 1.6391e-05, 3.9864e-03, ..., 7.1619e-07, + -7.8753e-06, 7.1861e-06], + ..., + [ 4.2409e-05, 2.9020e-06, 1.3590e-05, ..., 1.2340e-06, + 2.8890e-06, 1.4529e-06], + [ 5.3197e-06, -1.8217e-06, -6.5155e-06, ..., -7.7300e-08, + 4.9919e-07, -8.7544e-06], + [ 3.9302e-06, 2.5168e-05, -1.9586e-04, ..., -4.7654e-05, + -2.2978e-05, 2.5257e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0293, 0.0098, 0.0180, 0.0233, 0.0301, 0.0002, -0.0042, -0.0010, + 0.0045, -0.0337], device='cuda:0'), grad: tensor([-1.9550e-04, -1.8707e-02, 1.8494e-02, 7.0214e-05, 3.1614e-04, + -1.0774e-05, 1.5497e-04, 1.5569e-04, 9.7975e-06, -2.9254e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 217.43, cls_loss 0.0067 cls_loss_mapping 0.0090 cls_loss_causal 0.5790 re_mapping 0.0094 re_causal 0.0273 /// teacc 98.73 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0893, 0.1336, -0.0286, ..., 0.0006, 0.0583, 0.0158], + [ 0.0945, -0.0664, 0.0229, ..., -0.0515, -0.0082, -0.0377], + [-0.0429, -0.0974, 0.0089, ..., -0.0359, 0.0408, -0.1220], + ..., + [-0.1029, -0.0760, 0.0494, ..., -0.0374, -0.0684, -0.0892], + [-0.0488, -0.0108, 0.0217, ..., 0.0230, -0.0796, 0.0139], + [-0.0325, -0.0029, 0.0636, ..., 0.1036, -0.0359, -0.0901]], + device='cuda:0'), grad: tensor([[ 1.0550e-05, -1.2827e-04, -1.2106e-04, ..., 8.6986e-07, + 1.0595e-05, 1.7695e-08], + [-4.0054e-05, -7.4841e-06, -5.6058e-05, ..., -2.2352e-07, + -3.9190e-05, 7.4506e-09], + [ 1.5041e-06, 6.1318e-06, 1.1958e-05, ..., 1.7481e-06, + 3.6135e-06, 1.6764e-08], + ..., + [ 2.4550e-06, 1.5628e-06, 3.0100e-06, ..., 3.7812e-07, + 2.3860e-06, 1.3970e-08], + [ 1.3672e-05, 5.3681e-06, 2.4050e-05, ..., 1.1632e-06, + 1.3232e-05, 1.8906e-07], + [ 1.7816e-06, 1.0341e-04, 9.7811e-05, ..., -5.7444e-06, + 2.0675e-06, 2.5146e-08]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0292, 0.0109, 0.0182, 0.0237, 0.0310, 0.0006, -0.0048, -0.0021, + 0.0056, -0.0349], device='cuda:0'), grad: tensor([-3.8266e-04, -1.0842e-04, 4.1783e-05, 1.6183e-05, -1.1843e-04, + 1.1325e-05, 5.5432e-05, 9.2924e-05, 5.5879e-05, 3.3617e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 115---------------------------------------------------- +epoch 115, time 218.18, cls_loss 0.0047 cls_loss_mapping 0.0075 cls_loss_causal 0.5779 re_mapping 0.0097 re_causal 0.0285 /// teacc 98.86 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0895, 0.1340, -0.0285, ..., 0.0003, 0.0584, 0.0158], + [ 0.0949, -0.0668, 0.0229, ..., -0.0517, -0.0080, -0.0380], + [-0.0431, -0.0979, 0.0079, ..., -0.0356, 0.0405, -0.1224], + ..., + [-0.1031, -0.0764, 0.0501, ..., -0.0377, -0.0685, -0.0894], + [-0.0503, -0.0114, 0.0217, ..., 0.0231, -0.0805, 0.0142], + [-0.0327, -0.0033, 0.0632, ..., 0.1039, -0.0361, -0.0905]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, -7.3295e-07, 2.8089e-06, ..., 1.3737e-06, + 2.7753e-07, 2.6077e-08], + [-5.9325e-07, 1.1176e-08, 7.1079e-06, ..., 1.6950e-07, + 5.7891e-06, 4.9360e-08], + [ 4.2841e-08, 8.7544e-08, 2.7508e-05, ..., 2.9542e-06, + 1.5926e-07, 3.1944e-07], + ..., + [ 6.1467e-08, 2.5146e-08, -6.0171e-05, ..., 3.7104e-06, + 1.4808e-07, 7.6368e-08], + [ 2.3656e-07, 2.1234e-07, 2.9370e-05, ..., 5.7276e-07, + 7.3574e-08, 3.1758e-07], + [ 1.4622e-07, 4.8336e-07, -1.9252e-05, ..., -1.4886e-05, + 2.7418e-06, 8.1025e-08]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0292, 0.0108, 0.0177, 0.0244, 0.0315, 0.0007, -0.0049, -0.0014, + 0.0052, -0.0357], device='cuda:0'), grad: tensor([ 1.9312e-05, 5.4330e-05, 4.7266e-05, 5.6028e-05, 9.9465e-06, + 1.8865e-05, -1.9930e-06, 8.0615e-06, 5.7667e-05, -2.6965e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 217.53, cls_loss 0.0042 cls_loss_mapping 0.0083 cls_loss_causal 0.5758 re_mapping 0.0094 re_causal 0.0281 /// teacc 98.84 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0896, 0.1352, -0.0282, ..., 0.0004, 0.0594, 0.0157], + [ 0.0955, -0.0667, 0.0229, ..., -0.0519, -0.0079, -0.0379], + [-0.0435, -0.0988, 0.0074, ..., -0.0357, 0.0404, -0.1229], + ..., + [-0.1033, -0.0771, 0.0506, ..., -0.0379, -0.0685, -0.0892], + [-0.0507, -0.0115, 0.0210, ..., 0.0230, -0.0807, 0.0143], + [-0.0327, -0.0036, 0.0637, ..., 0.1051, -0.0364, -0.0908]], + device='cuda:0'), grad: tensor([[ 5.0291e-07, -2.0117e-06, 4.7963e-07, ..., 2.2352e-08, + 4.0531e-06, 3.4086e-07], + [-6.5938e-06, 3.6322e-08, -7.3276e-06, ..., 1.2107e-08, + 5.2303e-06, 2.4214e-08], + [ 1.2005e-06, 6.1654e-07, 4.8056e-06, ..., 5.3551e-07, + 9.5144e-06, 5.2899e-07], + ..., + [ 1.9111e-06, 4.6566e-08, -2.7537e-04, ..., 1.7043e-07, + 3.0175e-07, -1.8068e-07], + [ 7.5623e-07, 8.5682e-08, 4.5300e-06, ..., -7.0129e-07, + 1.0923e-05, -2.3749e-07], + [ 1.1399e-06, 1.5423e-06, 2.6083e-04, ..., -2.0862e-07, + 3.5018e-07, 1.8999e-07]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0298, 0.0109, 0.0173, 0.0239, 0.0310, 0.0006, -0.0052, -0.0008, + 0.0045, -0.0355], device='cuda:0'), grad: tensor([ 2.3603e-05, 1.7032e-05, 5.4061e-05, 1.7345e-05, 5.8860e-05, + 2.8670e-05, -2.5892e-04, -2.8515e-04, 6.8069e-05, 2.7680e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 117---------------------------------------------------- +epoch 117, time 218.42, cls_loss 0.0053 cls_loss_mapping 0.0071 cls_loss_causal 0.5844 re_mapping 0.0092 re_causal 0.0278 /// teacc 98.88 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0901, 0.1355, -0.0283, ..., -0.0002, 0.0596, 0.0154], + [ 0.0960, -0.0669, 0.0228, ..., -0.0519, -0.0076, -0.0380], + [-0.0441, -0.0995, 0.0081, ..., -0.0357, 0.0405, -0.1242], + ..., + [-0.1037, -0.0784, 0.0509, ..., -0.0386, -0.0689, -0.0894], + [-0.0495, -0.0099, 0.0207, ..., 0.0229, -0.0810, 0.0163], + [-0.0329, -0.0037, 0.0642, ..., 0.1065, -0.0365, -0.0921]], + device='cuda:0'), grad: tensor([[ 8.1956e-08, 9.3132e-10, 2.5518e-07, ..., 1.2014e-07, + 6.6590e-07, 1.6419e-06], + [-1.5115e-06, 4.9360e-08, -1.5134e-06, ..., 5.6811e-08, + 1.9558e-08, 1.1828e-07], + [ 2.4401e-07, -4.0326e-07, -7.1060e-07, ..., -9.3970e-07, + 7.5530e-07, 2.0079e-06], + ..., + [ 4.5914e-07, 3.9116e-08, 1.6484e-07, ..., 4.4331e-07, + 3.7253e-09, 7.3574e-08], + [ 3.7998e-07, 7.8231e-08, 1.4510e-06, ..., 1.1856e-06, + 6.3796e-07, 1.6894e-06], + [ 5.2806e-07, 1.3970e-07, -3.2075e-06, ..., -2.0862e-06, + 2.2352e-08, 4.6939e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0296, 0.0106, 0.0177, 0.0242, 0.0309, 0.0002, -0.0061, -0.0008, + 0.0055, -0.0353], device='cuda:0'), grad: tensor([ 2.1547e-05, -1.0692e-06, 1.6674e-05, -5.5060e-06, 1.6302e-05, + 2.0802e-05, -9.4235e-05, 6.3423e-07, 2.6628e-05, -1.7947e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 118---------------------------------------------------- +epoch 118, time 218.51, cls_loss 0.0037 cls_loss_mapping 0.0047 cls_loss_causal 0.5410 re_mapping 0.0089 re_causal 0.0267 /// teacc 98.94 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0905, 0.1361, -0.0281, ..., -0.0002, 0.0598, 0.0153], + [ 0.0971, -0.0670, 0.0233, ..., -0.0517, -0.0076, -0.0381], + [-0.0444, -0.0998, 0.0081, ..., -0.0359, 0.0405, -0.1246], + ..., + [-0.1052, -0.0788, 0.0507, ..., -0.0387, -0.0689, -0.0894], + [-0.0492, -0.0098, 0.0211, ..., 0.0234, -0.0812, 0.0169], + [-0.0339, -0.0041, 0.0641, ..., 0.1066, -0.0365, -0.0930]], + device='cuda:0'), grad: tensor([[ 1.5065e-05, 3.3118e-06, 3.1799e-05, ..., 6.0201e-06, + 1.0453e-05, 2.0675e-07], + [-8.3864e-05, -4.0114e-05, -1.9968e-04, ..., -3.5197e-05, + -5.8174e-05, 2.7753e-07], + [ 4.6194e-07, 4.4238e-07, -7.6257e-06, ..., 2.2911e-07, + -2.3961e-05, 9.5181e-07], + ..., + [ 5.3048e-06, 2.6301e-06, 1.1146e-05, ..., 2.2128e-06, + 3.5912e-06, 4.9174e-07], + [ 4.7117e-05, 2.2665e-05, 1.1361e-04, ..., 1.9595e-05, + 3.2127e-05, 1.3448e-06], + [ 4.8876e-06, 3.7570e-06, 1.2398e-05, ..., 5.8860e-07, + 3.0696e-06, 1.1530e-06]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0299, 0.0110, 0.0178, 0.0241, 0.0311, -0.0003, -0.0059, -0.0007, + 0.0058, -0.0360], device='cuda:0'), grad: tensor([ 5.9336e-05, -2.4140e-04, -4.1246e-04, -5.2005e-05, 2.6631e-04, + 8.2612e-05, 8.8751e-05, 1.8328e-05, 1.6499e-04, 2.5168e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 217.68, cls_loss 0.0048 cls_loss_mapping 0.0068 cls_loss_causal 0.5884 re_mapping 0.0090 re_causal 0.0268 /// teacc 98.82 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0918, 0.1364, -0.0281, ..., -0.0006, 0.0598, 0.0150], + [ 0.0980, -0.0672, 0.0239, ..., -0.0516, -0.0075, -0.0388], + [-0.0439, -0.1007, 0.0074, ..., -0.0363, 0.0404, -0.1257], + ..., + [-0.1070, -0.0796, 0.0505, ..., -0.0390, -0.0685, -0.0893], + [-0.0495, -0.0100, 0.0214, ..., 0.0235, -0.0813, 0.0172], + [-0.0344, -0.0042, 0.0641, ..., 0.1072, -0.0366, -0.0935]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, -2.4700e-04, -4.8548e-05, ..., -1.3620e-05, + 1.0245e-08, -8.2731e-05], + [-6.5099e-07, 1.1791e-06, 7.2643e-08, ..., 1.6391e-07, + 9.3132e-10, 4.0233e-07], + [ 3.0734e-07, 1.5860e-06, 1.6140e-06, ..., 3.3341e-07, + 2.7940e-09, 3.7532e-07], + ..., + [ 9.0338e-08, 2.6487e-06, 1.3404e-05, ..., 5.9679e-06, + 0.0000e+00, 6.4354e-07], + [ 9.3225e-07, 2.4401e-06, 6.5006e-07, ..., 5.3737e-07, + 8.3819e-09, -1.6764e-08], + [ 8.1025e-08, 3.6448e-05, -6.8322e-06, ..., 5.8860e-07, + 9.3132e-10, 2.9430e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0298, 0.0115, 0.0179, 0.0223, 0.0308, 0.0003, -0.0057, -0.0002, + 0.0058, -0.0362], device='cuda:0'), grad: tensor([-2.1374e-04, 7.4953e-06, 1.4327e-05, 2.1353e-05, 1.4138e-04, + 2.3258e-04, -2.5296e-04, 1.9312e-05, 1.0245e-05, 2.0057e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 217.67, cls_loss 0.0049 cls_loss_mapping 0.0058 cls_loss_causal 0.5652 re_mapping 0.0089 re_causal 0.0265 /// teacc 98.80 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0925, 0.1376, -0.0277, ..., -0.0006, 0.0600, 0.0156], + [ 0.0991, -0.0676, 0.0233, ..., -0.0517, -0.0076, -0.0390], + [-0.0453, -0.1025, 0.0059, ..., -0.0364, 0.0404, -0.1261], + ..., + [-0.1064, -0.0810, 0.0509, ..., -0.0395, -0.0686, -0.0894], + [-0.0497, -0.0100, 0.0213, ..., 0.0237, -0.0814, 0.0172], + [-0.0349, -0.0047, 0.0645, ..., 0.1073, -0.0366, -0.0939]], + device='cuda:0'), grad: tensor([[ 4.4610e-07, -6.6757e-05, -3.5405e-05, ..., 4.3772e-08, + 1.6764e-08, -4.2506e-06], + [-3.5644e-05, 9.5926e-07, -9.7632e-05, ..., 1.4901e-08, + 2.0396e-07, 9.4250e-07], + [ 1.5043e-05, 4.5784e-06, 1.1409e-06, ..., 1.1455e-07, + 2.6077e-08, 1.2413e-05], + ..., + [ 2.4259e-05, 1.2806e-06, 6.3181e-05, ..., 1.1548e-07, + 2.2538e-07, 4.8056e-07], + [ 2.0154e-06, 5.8413e-06, 4.3660e-06, ..., -2.6729e-07, + 5.5879e-09, 6.1281e-07], + [ 1.1362e-05, 3.9697e-05, 5.4628e-05, ..., -7.7672e-07, + 2.9709e-07, 8.7731e-07]], device='cuda:0') +Epoch 122, bias, value: tensor([ 3.0598e-02, 1.1425e-02, 1.6771e-02, 2.2175e-02, 3.1160e-02, + 4.9557e-07, -5.2412e-03, 2.3751e-04, 5.4332e-03, -3.6393e-02], + device='cuda:0'), grad: tensor([-8.7857e-05, -9.3043e-05, 9.1136e-05, -1.0574e-04, -2.8446e-05, + 5.3495e-06, 1.4111e-05, 8.3327e-05, 1.8910e-05, 1.0216e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 217.82, cls_loss 0.0044 cls_loss_mapping 0.0083 cls_loss_causal 0.5672 re_mapping 0.0092 re_causal 0.0264 /// teacc 98.83 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0928, 0.1380, -0.0276, ..., -0.0008, 0.0604, 0.0152], + [ 0.0996, -0.0678, 0.0229, ..., -0.0527, -0.0076, -0.0392], + [-0.0457, -0.1029, 0.0074, ..., -0.0366, 0.0403, -0.1265], + ..., + [-0.1067, -0.0809, 0.0512, ..., -0.0399, -0.0683, -0.0902], + [-0.0509, -0.0103, 0.0211, ..., 0.0240, -0.0820, 0.0169], + [-0.0345, -0.0048, 0.0642, ..., 0.1084, -0.0367, -0.0942]], + device='cuda:0'), grad: tensor([[ 1.0058e-07, -7.7114e-06, 1.7658e-06, ..., 9.6485e-07, + 1.4715e-07, 9.3319e-07], + [-4.8801e-07, 9.2201e-07, 1.4827e-06, ..., 8.7731e-07, + 5.5209e-06, 3.3099e-06], + [-1.9260e-06, 4.3362e-06, 1.1072e-05, ..., 6.1244e-06, + -1.0602e-05, 3.4645e-06], + ..., + [ 1.3486e-06, 1.9073e-06, 1.4827e-05, ..., 1.6205e-06, + 4.2319e-06, 1.5542e-05], + [ 6.5565e-07, -1.3418e-05, -4.1366e-05, ..., -1.7241e-05, + 2.4959e-07, -1.0103e-05], + [ 5.8860e-07, 7.7784e-06, -1.4044e-06, ..., 2.9840e-06, + 5.2154e-08, 6.5044e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0306, 0.0111, 0.0176, 0.0225, 0.0314, 0.0007, -0.0055, 0.0002, + 0.0048, -0.0369], device='cuda:0'), grad: tensor([-7.3537e-06, 4.1902e-05, -7.4990e-06, -1.4782e-05, -1.8752e-04, + 3.5465e-05, 1.1563e-05, 1.6284e-04, -7.0453e-05, 3.5465e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 217.51, cls_loss 0.0057 cls_loss_mapping 0.0084 cls_loss_causal 0.5972 re_mapping 0.0082 re_causal 0.0243 /// teacc 98.92 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0940, 0.1365, -0.0277, ..., -0.0010, 0.0606, 0.0153], + [ 0.1006, -0.0665, 0.0229, ..., -0.0531, -0.0073, -0.0388], + [-0.0459, -0.1034, 0.0106, ..., -0.0366, 0.0405, -0.1262], + ..., + [-0.1071, -0.0823, 0.0516, ..., -0.0398, -0.0685, -0.0905], + [-0.0519, -0.0104, 0.0183, ..., 0.0246, -0.0821, 0.0170], + [-0.0348, -0.0051, 0.0642, ..., 0.1088, -0.0368, -0.0948]], + device='cuda:0'), grad: tensor([[ 3.4254e-06, -4.6268e-06, 1.9670e-06, ..., 2.7381e-07, + 1.8626e-09, 1.1735e-06], + [-1.6129e-04, -9.4354e-05, -3.2973e-04, ..., -2.1622e-05, + -3.3528e-08, -9.3997e-05], + [ 2.1234e-06, 1.0788e-05, 4.9442e-05, ..., 4.7460e-06, + 9.3132e-09, 6.0610e-06], + ..., + [ 6.7502e-06, 4.4666e-06, 1.3821e-05, ..., 1.1530e-06, + 9.3132e-09, 1.1414e-05], + [ 1.1462e-04, 5.7936e-05, 1.8418e-04, ..., 1.0565e-05, + 1.8626e-09, 6.7890e-05], + [ 2.8923e-05, 1.8656e-05, 5.4568e-05, ..., 8.0094e-07, + 3.7253e-09, 1.6227e-05]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0292, 0.0111, 0.0206, 0.0231, 0.0318, -0.0003, -0.0040, 0.0002, + 0.0016, -0.0374], device='cuda:0'), grad: tensor([ 1.2770e-05, -5.0449e-04, -8.3983e-05, -9.0748e-06, 2.0444e-05, + 8.5458e-06, -1.1221e-05, 4.0591e-05, 4.3392e-04, 9.1791e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 217.56, cls_loss 0.0036 cls_loss_mapping 0.0069 cls_loss_causal 0.5347 re_mapping 0.0086 re_causal 0.0250 /// teacc 98.86 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0937, 0.1372, -0.0274, ..., -0.0011, 0.0607, 0.0155], + [ 0.1011, -0.0676, 0.0228, ..., -0.0532, -0.0073, -0.0397], + [-0.0463, -0.1047, 0.0104, ..., -0.0372, 0.0405, -0.1273], + ..., + [-0.1075, -0.0838, 0.0517, ..., -0.0406, -0.0685, -0.0907], + [-0.0518, -0.0102, 0.0188, ..., 0.0250, -0.0821, 0.0179], + [-0.0342, -0.0051, 0.0644, ..., 0.1098, -0.0367, -0.0943]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, -2.1290e-06, -1.3784e-06, ..., 7.4506e-09, + 1.1921e-07, 5.7742e-08], + [-5.8487e-07, 3.3528e-08, 7.5065e-07, ..., 3.7253e-08, + 1.8626e-08, 2.3842e-07], + [ 1.8626e-08, 7.7859e-07, -1.6242e-05, ..., 1.6764e-07, + 8.3819e-08, -3.2596e-06], + ..., + [ 3.6880e-07, 1.6950e-07, 5.6252e-06, ..., 1.6578e-06, + 1.8626e-09, 4.4145e-07], + [ 1.6764e-07, -4.6007e-07, 1.0125e-05, ..., 1.5832e-07, + 6.3330e-08, 2.0619e-06], + [ 1.6764e-07, 1.7975e-06, -2.0135e-06, ..., -3.1237e-06, + 1.1176e-08, 1.0990e-07]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0295, 0.0110, 0.0205, 0.0233, 0.0315, -0.0011, -0.0039, 0.0002, + 0.0018, -0.0372], device='cuda:0'), grad: tensor([-9.4995e-07, 4.1015e-06, -7.8619e-05, 1.2539e-05, 2.1346e-06, + -6.2771e-07, -1.3046e-05, 2.7686e-05, 4.1425e-05, 5.3048e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 217.53, cls_loss 0.0051 cls_loss_mapping 0.0074 cls_loss_causal 0.5865 re_mapping 0.0090 re_causal 0.0261 /// teacc 98.80 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0941, 0.1378, -0.0273, ..., -0.0015, 0.0625, 0.0159], + [ 0.1020, -0.0693, 0.0230, ..., -0.0535, -0.0074, -0.0404], + [-0.0468, -0.1058, 0.0103, ..., -0.0373, 0.0406, -0.1283], + ..., + [-0.1092, -0.0850, 0.0514, ..., -0.0418, -0.0689, -0.0908], + [-0.0553, -0.0131, 0.0178, ..., 0.0258, -0.0852, 0.0161], + [-0.0350, -0.0052, 0.0652, ..., 0.1110, -0.0362, -0.0950]], + device='cuda:0'), grad: tensor([[ 2.6450e-07, -6.5826e-06, -9.7603e-07, ..., 1.3784e-06, + 1.1325e-06, 1.9930e-07], + [-1.5348e-06, 2.3283e-07, 1.9744e-06, ..., 4.2841e-08, + 2.5891e-07, 9.1270e-08], + [ 3.0175e-07, 6.5006e-07, 1.9953e-05, ..., 2.3283e-07, + 1.8030e-05, 3.4086e-07], + ..., + [ 2.7940e-07, 8.5495e-07, -2.7239e-05, ..., 1.1511e-06, + -1.1519e-05, 1.8254e-07], + [ 8.0094e-08, -7.4320e-07, 9.1270e-08, ..., -3.7253e-07, + 1.9558e-07, -1.3057e-06], + [ 1.4734e-06, 7.0184e-06, -2.1290e-06, ..., -2.9076e-06, + -3.2596e-07, 1.1865e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0298, 0.0113, 0.0204, 0.0229, 0.0309, 0.0012, -0.0044, -0.0001, + 0.0002, -0.0366], device='cuda:0'), grad: tensor([ 2.5760e-06, 3.9041e-06, 1.0866e-04, 1.6704e-05, 2.2113e-05, + 9.8348e-07, -7.9036e-05, -8.8394e-05, 1.6354e-06, 1.0848e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 217.66, cls_loss 0.0047 cls_loss_mapping 0.0078 cls_loss_causal 0.5673 re_mapping 0.0086 re_causal 0.0253 /// teacc 98.82 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0951, 0.1383, -0.0272, ..., -0.0016, 0.0632, 0.0149], + [ 0.1019, -0.0707, 0.0228, ..., -0.0542, -0.0076, -0.0411], + [-0.0472, -0.1067, 0.0101, ..., -0.0377, 0.0407, -0.1306], + ..., + [-0.1096, -0.0858, 0.0513, ..., -0.0428, -0.0691, -0.0910], + [-0.0552, -0.0130, 0.0181, ..., 0.0254, -0.0852, 0.0165], + [-0.0349, -0.0055, 0.0663, ..., 0.1131, -0.0363, -0.0957]], + device='cuda:0'), grad: tensor([[ 2.0377e-06, 1.1977e-06, 3.4012e-06, ..., 1.1753e-06, + 5.4948e-07, 9.7789e-07], + [ 9.9719e-05, 4.0233e-07, 2.2745e-04, ..., 1.0586e-04, + 1.3970e-07, 1.4529e-07], + [ 4.1798e-06, 6.5193e-08, 8.7172e-06, ..., 3.6806e-06, + 3.7253e-09, 7.7114e-07], + ..., + [ 3.6061e-05, 6.8918e-08, 8.4639e-05, ..., 3.1412e-05, + 1.3039e-08, 2.1979e-07], + [ 2.6867e-05, 1.5065e-05, 5.8472e-05, ..., 2.5779e-05, + 5.2825e-06, 9.0227e-06], + [-1.9884e-04, -1.1716e-06, -4.5681e-04, ..., -1.9431e-04, + 8.0094e-08, -3.7923e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0295, 0.0110, 0.0199, 0.0235, 0.0304, 0.0010, -0.0034, -0.0003, + 0.0004, -0.0360], device='cuda:0'), grad: tensor([ 8.7544e-06, 3.4142e-04, 1.6287e-05, 1.1504e-05, 7.9036e-05, + 5.5504e-04, -5.7602e-04, 1.1843e-04, 1.1343e-04, -6.6900e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 217.55, cls_loss 0.0043 cls_loss_mapping 0.0059 cls_loss_causal 0.6163 re_mapping 0.0083 re_causal 0.0264 /// teacc 98.83 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0953, 0.1390, -0.0271, ..., -0.0016, 0.0636, 0.0152], + [ 0.1024, -0.0708, 0.0225, ..., -0.0548, -0.0074, -0.0414], + [-0.0478, -0.1062, 0.0100, ..., -0.0374, 0.0407, -0.1308], + ..., + [-0.1099, -0.0865, 0.0505, ..., -0.0443, -0.0691, -0.0911], + [-0.0552, -0.0132, 0.0181, ..., 0.0255, -0.0853, 0.0166], + [-0.0352, -0.0059, 0.0681, ..., 0.1137, -0.0363, -0.0965]], + device='cuda:0'), grad: tensor([[ 1.3784e-07, -4.6566e-06, -9.6299e-07, ..., 1.4715e-07, + -6.7055e-08, 1.3039e-07], + [-5.3495e-06, 5.0105e-07, 2.4602e-05, ..., 1.4529e-07, + 7.4506e-09, 1.8626e-07], + [ 2.4773e-07, 1.0915e-06, 8.0764e-06, ..., 3.0361e-07, + -3.3528e-08, 6.4448e-07], + ..., + [ 8.2888e-07, 2.7195e-07, -8.1480e-05, ..., 2.7381e-07, + 2.0489e-08, 1.3039e-07], + [ 1.3616e-06, -2.4997e-06, 5.4240e-06, ..., -1.4920e-06, + 5.5879e-09, -4.2729e-06], + [ 1.9632e-06, 2.2370e-06, 9.7305e-06, ..., -3.0808e-06, + 2.6077e-08, 7.6927e-07]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0299, 0.0105, 0.0198, 0.0234, 0.0307, 0.0007, -0.0030, -0.0008, + 0.0002, -0.0351], device='cuda:0'), grad: tensor([ 2.2408e-06, 6.0350e-05, 1.4074e-05, 3.6091e-05, 1.9789e-05, + 5.6773e-06, -3.8557e-07, -1.7524e-04, 1.6809e-05, 2.0415e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 217.40, cls_loss 0.0050 cls_loss_mapping 0.0062 cls_loss_causal 0.5654 re_mapping 0.0089 re_causal 0.0257 /// teacc 98.75 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0958, 0.1395, -0.0268, ..., -0.0016, 0.0640, 0.0151], + [ 0.1026, -0.0710, 0.0220, ..., -0.0554, -0.0075, -0.0428], + [-0.0475, -0.1071, 0.0099, ..., -0.0375, 0.0408, -0.1316], + ..., + [-0.1104, -0.0878, 0.0512, ..., -0.0445, -0.0692, -0.0908], + [-0.0553, -0.0130, 0.0184, ..., 0.0259, -0.0854, 0.0169], + [-0.0362, -0.0066, 0.0679, ..., 0.1137, -0.0365, -0.0975]], + device='cuda:0'), grad: tensor([[ 2.5462e-06, 7.4506e-09, 1.3411e-07, ..., 1.4901e-08, + 1.8626e-09, 2.6524e-06], + [ 8.9526e-05, 1.8626e-08, -2.5127e-06, ..., 4.4703e-08, + 7.4506e-09, 9.4712e-05], + [ 8.7693e-06, 1.2275e-06, 3.9995e-05, ..., 1.6659e-05, + -4.8429e-08, 1.9133e-05], + ..., + [ 9.9689e-06, 1.6764e-08, 5.4576e-07, ..., 2.9802e-08, + 2.6077e-08, 1.1355e-05], + [ 2.1353e-05, -1.5404e-06, -4.0889e-05, ..., -1.7270e-05, + 3.7253e-09, 1.3322e-05], + [ 4.8399e-05, 2.6077e-08, 5.3085e-07, ..., 1.4901e-08, + 0.0000e+00, 4.9561e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0303, 0.0097, 0.0201, 0.0233, 0.0314, 0.0011, -0.0032, -0.0003, + 0.0003, -0.0362], device='cuda:0'), grad: tensor([ 7.8678e-06, 2.6703e-04, 1.1528e-04, -7.9393e-04, 2.4848e-06, + 2.4962e-04, 3.2447e-06, 2.8759e-05, -2.2933e-05, 1.4234e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 217.79, cls_loss 0.0054 cls_loss_mapping 0.0066 cls_loss_causal 0.5625 re_mapping 0.0087 re_causal 0.0250 /// teacc 98.88 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0962, 0.1407, -0.0263, ..., -0.0016, 0.0659, 0.0156], + [ 0.1044, -0.0728, 0.0221, ..., -0.0557, -0.0084, -0.0437], + [-0.0481, -0.1082, 0.0100, ..., -0.0376, 0.0407, -0.1323], + ..., + [-0.1125, -0.0889, 0.0512, ..., -0.0453, -0.0684, -0.0904], + [-0.0555, -0.0130, 0.0187, ..., 0.0262, -0.0857, 0.0173], + [-0.0358, -0.0070, 0.0678, ..., 0.1146, -0.0370, -0.0994]], + device='cuda:0'), grad: tensor([[ 2.6822e-07, 4.6566e-08, 5.0105e-07, ..., 2.7940e-08, + 3.7253e-09, 2.5518e-07], + [-3.9451e-06, 3.1665e-08, -4.2915e-06, ..., 2.9802e-08, + 0.0000e+00, 8.5682e-08], + [ 2.3283e-07, 2.8312e-07, 7.6070e-06, ..., 6.1467e-08, + 0.0000e+00, 5.4203e-07], + ..., + [ 1.5777e-06, 1.4901e-08, -1.3083e-05, ..., 8.5682e-08, + 0.0000e+00, 1.4529e-07], + [ 5.6624e-07, -6.4634e-07, -7.6555e-07, ..., -9.4995e-08, + 7.4506e-09, -1.6596e-06], + [ 8.9779e-07, 1.0245e-07, 2.2966e-06, ..., -4.9360e-07, + 0.0000e+00, -5.3085e-07]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0317, 0.0096, 0.0202, 0.0223, 0.0313, 0.0016, -0.0041, 0.0002, + 0.0005, -0.0371], device='cuda:0'), grad: tensor([ 5.4315e-06, -1.0449e-06, 2.6733e-05, 4.2841e-06, -1.7598e-05, + 1.0401e-05, -1.2517e-05, -5.6356e-05, 2.4177e-06, 3.8117e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 129---------------------------------------------------- +epoch 129, time 218.45, cls_loss 0.0033 cls_loss_mapping 0.0062 cls_loss_causal 0.5510 re_mapping 0.0088 re_causal 0.0257 /// teacc 98.95 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0965, 0.1410, -0.0262, ..., -0.0017, 0.0660, 0.0155], + [ 0.1051, -0.0755, 0.0217, ..., -0.0558, -0.0084, -0.0465], + [-0.0483, -0.1088, 0.0098, ..., -0.0379, 0.0408, -0.1327], + ..., + [-0.1137, -0.0892, 0.0508, ..., -0.0457, -0.0684, -0.0906], + [-0.0553, -0.0119, 0.0203, ..., 0.0263, -0.0856, 0.0194], + [-0.0349, -0.0071, 0.0680, ..., 0.1152, -0.0371, -0.0995]], + device='cuda:0'), grad: tensor([[ 5.2340e-07, -4.5113e-06, -7.7114e-07, ..., 1.8999e-07, + 1.1176e-08, 8.9407e-07], + [-2.2706e-06, 4.7497e-07, -4.4703e-07, ..., 1.8068e-07, + 3.7253e-09, 5.1409e-07], + [-1.0133e-06, 9.5554e-07, -1.4007e-06, ..., 7.8231e-08, + 0.0000e+00, 6.2771e-07], + ..., + [ 1.3374e-06, 2.2445e-06, 3.7730e-05, ..., 5.3719e-06, + 0.0000e+00, 1.6391e-05], + [ 1.1668e-05, 9.7454e-06, -7.1704e-05, ..., -9.3281e-06, + 3.5949e-07, -1.7568e-05], + [ 3.2652e-06, 2.3916e-06, -3.1106e-07, ..., -8.2925e-06, + 3.7253e-09, 4.8056e-06]], device='cuda:0') +Epoch 131, bias, value: tensor([ 3.1801e-02, 9.1226e-03, 2.0107e-02, 2.2013e-02, 3.1120e-02, + 1.2966e-03, -4.1607e-03, -4.0313e-05, 1.9493e-03, -3.7051e-02], + device='cuda:0'), grad: tensor([-2.8703e-06, 2.0526e-06, -9.6709e-06, -6.3404e-06, 6.9737e-05, + 1.8731e-05, -2.4244e-05, 5.4538e-05, -7.4506e-05, -2.7150e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 217.59, cls_loss 0.0049 cls_loss_mapping 0.0078 cls_loss_causal 0.5889 re_mapping 0.0088 re_causal 0.0256 /// teacc 98.82 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0970, 0.1413, -0.0262, ..., -0.0017, 0.0662, 0.0154], + [ 0.1042, -0.0755, 0.0218, ..., -0.0558, -0.0076, -0.0466], + [-0.0469, -0.1088, 0.0097, ..., -0.0382, 0.0405, -0.1322], + ..., + [-0.1138, -0.0898, 0.0507, ..., -0.0461, -0.0685, -0.0908], + [-0.0554, -0.0116, 0.0199, ..., 0.0266, -0.0856, 0.0194], + [-0.0356, -0.0073, 0.0689, ..., 0.1157, -0.0371, -0.0999]], + device='cuda:0'), grad: tensor([[ 8.1398e-07, 4.5598e-06, -5.7928e-07, ..., -3.7253e-08, + -1.2852e-07, 3.5763e-07], + [-3.2429e-06, 1.5274e-07, -3.5260e-06, ..., 3.7253e-09, + -2.4214e-07, 1.6205e-07], + [ 1.2275e-06, 1.6317e-06, 2.2929e-06, ..., 5.2154e-08, + 1.6950e-07, 1.3150e-06], + ..., + [ 1.0282e-06, 2.0862e-07, 7.6555e-07, ..., 8.5682e-08, + 4.0978e-08, 4.1351e-07], + [ 1.4752e-05, -3.4459e-07, -5.0478e-06, ..., -3.5390e-08, + 1.3225e-07, 2.1726e-05], + [ 3.3341e-06, 2.6654e-06, 7.5996e-07, ..., -3.0547e-07, + 1.3411e-07, 5.8711e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0320, 0.0085, 0.0212, 0.0221, 0.0309, 0.0012, -0.0044, -0.0003, + 0.0014, -0.0366], device='cuda:0'), grad: tensor([ 1.3277e-05, -4.2915e-06, 8.9109e-06, -6.5446e-05, -2.4021e-05, + 3.3945e-05, -2.9698e-05, 2.3711e-06, 3.1263e-05, 3.3557e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 217.48, cls_loss 0.0041 cls_loss_mapping 0.0067 cls_loss_causal 0.5377 re_mapping 0.0083 re_causal 0.0248 /// teacc 98.87 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0972, 0.1414, -0.0262, ..., -0.0018, 0.0662, 0.0151], + [ 0.1046, -0.0752, 0.0221, ..., -0.0559, -0.0074, -0.0466], + [-0.0473, -0.1075, 0.0107, ..., -0.0360, 0.0404, -0.1304], + ..., + [-0.1143, -0.0908, 0.0510, ..., -0.0463, -0.0686, -0.0913], + [-0.0551, -0.0117, 0.0186, ..., 0.0241, -0.0857, 0.0188], + [-0.0360, -0.0073, 0.0692, ..., 0.1166, -0.0369, -0.1003]], + device='cuda:0'), grad: tensor([[ 1.4175e-06, -1.3150e-05, -9.1828e-07, ..., 3.1665e-08, + -3.7253e-08, 2.0675e-06], + [-1.7568e-05, -1.7714e-06, -1.5318e-05, ..., 5.5879e-09, + -6.7018e-06, 8.4750e-07], + [ 1.5706e-05, 3.6694e-06, 1.8075e-05, ..., 1.2983e-06, + 5.8003e-06, 3.7719e-06], + ..., + [-7.5281e-05, 1.4901e-07, -1.2890e-06, ..., 2.6077e-08, + 7.2643e-08, -2.1660e-04], + [ 5.7742e-07, -5.8301e-07, -2.5537e-06, ..., -1.6149e-06, + 1.3784e-07, -4.4666e-06], + [ 3.9637e-06, 1.8235e-06, 4.9591e-05, ..., -1.8068e-07, + 5.4017e-08, 1.0386e-05]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0319, 0.0087, 0.0223, 0.0220, 0.0305, 0.0012, -0.0044, -0.0002, + 0.0004, -0.0365], device='cuda:0'), grad: tensor([-1.0341e-05, -3.3200e-05, 4.5300e-05, 1.9953e-05, 1.3590e-04, + 7.2718e-05, 1.9625e-05, -9.8324e-04, -1.3057e-06, 7.3433e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 132---------------------------------------------------- +epoch 132, time 218.39, cls_loss 0.0056 cls_loss_mapping 0.0080 cls_loss_causal 0.5643 re_mapping 0.0087 re_causal 0.0249 /// teacc 99.06 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0978, 0.1421, -0.0260, ..., -0.0019, 0.0686, 0.0148], + [ 0.1051, -0.0765, 0.0217, ..., -0.0561, -0.0106, -0.0466], + [-0.0478, -0.1086, 0.0102, ..., -0.0361, 0.0399, -0.1307], + ..., + [-0.1148, -0.0921, 0.0518, ..., -0.0468, -0.0681, -0.0932], + [-0.0552, -0.0112, 0.0189, ..., 0.0244, -0.0858, 0.0190], + [-0.0369, -0.0079, 0.0686, ..., 0.1175, -0.0372, -0.1008]], + device='cuda:0'), grad: tensor([[ 2.9746e-06, 1.5777e-06, -1.0487e-06, ..., 1.4901e-08, + 1.4901e-07, 1.3839e-06], + [ 4.6119e-06, 3.4720e-06, 1.4231e-06, ..., 9.6858e-08, + 5.5879e-08, 2.1383e-06], + [-5.4389e-07, 2.1402e-06, 2.7530e-06, ..., 2.0489e-07, + -9.6411e-06, 3.5539e-06], + ..., + [ 4.8839e-06, 3.3472e-06, -1.3579e-06, ..., 4.2841e-08, + 9.1493e-06, 3.1069e-06], + [ 4.9210e-04, 8.7643e-04, 1.8448e-05, ..., -4.7870e-07, + 1.6391e-07, 1.9801e-04], + [ 4.9844e-06, 6.5640e-06, 8.0280e-07, ..., -1.0990e-07, + 1.6764e-08, 3.1237e-06]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0326, 0.0081, 0.0218, 0.0234, 0.0323, 0.0009, -0.0040, -0.0003, + 0.0007, -0.0384], device='cuda:0'), grad: tensor([ 6.7353e-06, 1.4499e-05, -3.2932e-05, 1.9267e-05, 9.6038e-06, + 3.8099e-04, -2.3022e-03, 4.3303e-05, 1.8435e-03, 1.5810e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 217.80, cls_loss 0.0037 cls_loss_mapping 0.0060 cls_loss_causal 0.5914 re_mapping 0.0085 re_causal 0.0248 /// teacc 98.84 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0985, 0.1427, -0.0267, ..., -0.0026, 0.0702, 0.0140], + [ 0.1058, -0.0761, 0.0228, ..., -0.0563, -0.0107, -0.0467], + [-0.0480, -0.1100, 0.0097, ..., -0.0365, 0.0399, -0.1312], + ..., + [-0.1161, -0.0929, 0.0511, ..., -0.0471, -0.0680, -0.0936], + [-0.0552, -0.0105, 0.0196, ..., 0.0249, -0.0857, 0.0195], + [-0.0378, -0.0082, 0.0690, ..., 0.1183, -0.0374, -0.1013]], + device='cuda:0'), grad: tensor([[ 2.6878e-06, 2.2054e-06, 1.1958e-06, ..., 5.1036e-07, + 1.2424e-06, 1.7844e-06], + [ 8.5682e-07, 8.7544e-07, 2.3134e-06, ..., 2.8498e-07, + 8.9779e-07, 7.3574e-07], + [ 2.1942e-06, 9.0748e-06, 2.2650e-05, ..., 1.7323e-07, + 8.6054e-07, 4.4443e-06], + ..., + [ 8.4192e-07, 5.6624e-07, -2.4423e-05, ..., 4.4331e-07, + 3.8929e-07, -5.6438e-07], + [-2.7921e-06, -7.0572e-05, -1.7494e-05, ..., 3.0957e-06, + -1.6794e-05, -1.4722e-05], + [ 2.8033e-06, 1.6838e-06, -1.1839e-05, ..., -1.8049e-06, + 1.6764e-07, 1.5181e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0330, 0.0089, 0.0213, 0.0240, 0.0323, 0.0002, -0.0046, -0.0010, + 0.0013, -0.0384], device='cuda:0'), grad: tensor([ 1.3098e-05, 1.0289e-05, 5.2422e-05, 5.0873e-05, 1.2353e-05, + -1.1241e-04, 1.4818e-04, -3.8415e-05, -9.1672e-05, -4.4703e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 217.52, cls_loss 0.0044 cls_loss_mapping 0.0061 cls_loss_causal 0.5579 re_mapping 0.0087 re_causal 0.0248 /// teacc 98.77 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0989, 0.1420, -0.0271, ..., -0.0030, 0.0703, 0.0135], + [ 0.1068, -0.0762, 0.0236, ..., -0.0564, -0.0106, -0.0468], + [-0.0482, -0.1107, 0.0096, ..., -0.0366, 0.0399, -0.1315], + ..., + [-0.1178, -0.0944, 0.0505, ..., -0.0474, -0.0680, -0.0937], + [-0.0553, -0.0104, 0.0197, ..., 0.0249, -0.0859, 0.0197], + [-0.0372, -0.0078, 0.0693, ..., 0.1192, -0.0375, -0.1014]], + device='cuda:0'), grad: tensor([[ 5.2527e-07, -2.6464e-05, -2.2739e-05, ..., 7.4506e-09, + 4.8429e-08, 9.6858e-08], + [ 7.4506e-09, 2.9430e-07, 5.5730e-06, ..., 3.7253e-09, + 7.3947e-07, 5.5321e-07], + [ 2.7250e-06, 8.3074e-07, 5.0217e-06, ..., 5.5879e-08, + 7.4506e-07, 6.0350e-07], + ..., + [-1.2621e-05, 3.9488e-07, -3.0071e-05, ..., 1.3039e-08, + 2.1234e-07, -1.1399e-06], + [ 5.7891e-06, -1.4901e-08, 5.8673e-07, ..., -2.4214e-07, + 1.6764e-08, 6.8471e-06], + [ 3.3900e-06, 2.3171e-05, 1.9133e-05, ..., -4.0978e-07, + 1.1921e-07, 8.7544e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0320, 0.0094, 0.0212, 0.0238, 0.0324, 0.0004, -0.0033, -0.0012, + 0.0013, -0.0385], device='cuda:0'), grad: tensor([-8.1897e-05, 1.4082e-05, 1.4380e-05, 1.7330e-05, -1.2275e-06, + -2.1588e-06, 4.8354e-06, -5.3734e-05, 1.4961e-05, 7.3433e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 217.79, cls_loss 0.0041 cls_loss_mapping 0.0051 cls_loss_causal 0.5460 re_mapping 0.0084 re_causal 0.0245 /// teacc 98.84 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0996, 0.1418, -0.0273, ..., -0.0032, 0.0707, 0.0132], + [ 0.1071, -0.0762, 0.0234, ..., -0.0564, -0.0106, -0.0469], + [-0.0485, -0.1118, 0.0093, ..., -0.0367, 0.0401, -0.1318], + ..., + [-0.1176, -0.0965, 0.0503, ..., -0.0480, -0.0681, -0.0937], + [-0.0555, -0.0102, 0.0200, ..., 0.0252, -0.0858, 0.0199], + [-0.0372, -0.0085, 0.0709, ..., 0.1199, -0.0379, -0.1000]], + device='cuda:0'), grad: tensor([[ 6.1281e-07, 8.2143e-07, 2.4140e-06, ..., 2.2110e-06, + 4.9919e-07, 1.4380e-06], + [-7.9535e-07, 7.9907e-07, 5.3272e-06, ..., 6.5938e-07, + 3.3528e-08, 9.0152e-07], + [ 4.4703e-07, 3.1829e-05, 1.2302e-04, ..., 3.1501e-05, + -4.6939e-07, 3.5465e-05], + ..., + [ 3.8370e-07, 1.0636e-06, -6.9261e-05, ..., 8.5123e-07, + 3.9302e-07, 1.0412e-06], + [-8.9779e-07, -4.5061e-05, -9.1374e-05, ..., -4.2409e-05, + 1.9185e-07, -5.2780e-05], + [-4.1761e-06, -2.3171e-05, -1.9431e-05, ..., -2.7075e-05, + -9.3356e-06, 1.3076e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0314, 0.0092, 0.0210, 0.0236, 0.0313, 0.0002, -0.0024, -0.0015, + 0.0014, -0.0372], device='cuda:0'), grad: tensor([ 7.8827e-06, 1.2308e-05, 2.5034e-04, 4.9412e-05, 7.0095e-05, + 1.3977e-05, 1.3366e-05, -1.4448e-04, -1.8418e-04, -8.8573e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 217.70, cls_loss 0.0037 cls_loss_mapping 0.0055 cls_loss_causal 0.5703 re_mapping 0.0080 re_causal 0.0248 /// teacc 98.85 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.1003, 0.1421, -0.0272, ..., -0.0034, 0.0708, 0.0127], + [ 0.1072, -0.0764, 0.0236, ..., -0.0566, -0.0106, -0.0471], + [-0.0493, -0.1122, 0.0092, ..., -0.0367, 0.0402, -0.1322], + ..., + [-0.1160, -0.0986, 0.0503, ..., -0.0482, -0.0681, -0.0939], + [-0.0556, -0.0101, 0.0202, ..., 0.0254, -0.0859, 0.0201], + [-0.0386, -0.0088, 0.0708, ..., 0.1200, -0.0379, -0.1006]], + device='cuda:0'), grad: tensor([[ 1.7323e-07, -6.8426e-05, -3.9876e-05, ..., 1.3039e-08, + -3.0361e-07, 3.3341e-07], + [-5.7556e-06, -5.3830e-07, -1.3523e-06, ..., 1.8626e-08, + -5.7369e-07, 5.4389e-07], + [ 3.8594e-06, 2.5649e-06, 7.2606e-06, ..., 4.6939e-07, + 4.9919e-07, 4.4033e-06], + ..., + [ 2.1234e-07, 3.9488e-07, 2.8253e-04, ..., 3.3528e-08, + -8.2701e-07, 1.2374e-04], + [ 4.0233e-07, -1.7025e-06, -3.2449e-04, ..., -1.0487e-06, + 8.5682e-08, -1.3995e-04], + [ 5.0291e-07, 8.4490e-06, 6.5789e-06, ..., 9.6858e-08, + 4.1351e-07, 1.3411e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0314, 0.0088, 0.0207, 0.0241, 0.0316, -0.0004, -0.0026, -0.0005, + 0.0015, -0.0377], device='cuda:0'), grad: tensor([-1.2231e-04, 1.8030e-06, 2.2039e-05, 3.2455e-05, -3.3259e-05, + 7.8976e-06, 1.1230e-04, 3.7479e-04, -4.2892e-04, 3.3289e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 217.87, cls_loss 0.0045 cls_loss_mapping 0.0063 cls_loss_causal 0.6083 re_mapping 0.0081 re_causal 0.0248 /// teacc 98.81 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.1006, 0.1426, -0.0272, ..., -0.0034, 0.0708, 0.0125], + [ 0.1079, -0.0767, 0.0230, ..., -0.0566, -0.0105, -0.0473], + [-0.0500, -0.1129, 0.0095, ..., -0.0368, 0.0405, -0.1325], + ..., + [-0.1158, -0.0999, 0.0507, ..., -0.0484, -0.0687, -0.0943], + [-0.0558, -0.0099, 0.0205, ..., 0.0256, -0.0860, 0.0203], + [-0.0397, -0.0092, 0.0711, ..., 0.1206, -0.0380, -0.1011]], + device='cuda:0'), grad: tensor([[ 3.5688e-06, 8.2888e-07, 9.8720e-08, ..., 2.2352e-08, + 1.8626e-09, 1.8347e-06], + [ 1.0282e-06, 8.2701e-07, 7.0184e-06, ..., 1.8626e-08, + 0.0000e+00, 6.4820e-07], + [ 2.5593e-06, 1.6615e-06, 9.2164e-06, ..., 1.3039e-08, + 0.0000e+00, 1.2480e-06], + ..., + [ 1.5013e-06, 9.2387e-07, 1.5354e-04, ..., 7.8231e-07, + 0.0000e+00, 6.7987e-07], + [ 1.2472e-05, 7.9796e-06, 3.3882e-06, ..., 1.1176e-07, + 7.4506e-09, 6.1095e-06], + [ 2.6494e-05, 1.7822e-05, -2.1005e-04, ..., -2.6841e-06, + 0.0000e+00, 1.2927e-05]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0317, 0.0084, 0.0210, 0.0241, 0.0319, -0.0003, -0.0031, -0.0006, + 0.0016, -0.0378], device='cuda:0'), grad: tensor([ 5.9605e-06, 1.4454e-05, 2.0772e-05, 1.5235e-04, 2.1249e-05, + -2.1434e-04, -2.7940e-08, 1.8108e-04, 2.4155e-05, -2.0599e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 218.00, cls_loss 0.0055 cls_loss_mapping 0.0068 cls_loss_causal 0.5605 re_mapping 0.0079 re_causal 0.0235 /// teacc 98.91 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1032, 0.1435, -0.0266, ..., -0.0033, 0.0714, 0.0121], + [ 0.1098, -0.0773, 0.0235, ..., -0.0566, -0.0102, -0.0473], + [-0.0507, -0.1134, 0.0092, ..., -0.0368, 0.0405, -0.1327], + ..., + [-0.1170, -0.1012, 0.0508, ..., -0.0485, -0.0682, -0.0945], + [-0.0558, -0.0096, 0.0206, ..., 0.0257, -0.0859, 0.0207], + [-0.0417, -0.0102, 0.0712, ..., 0.1209, -0.0388, -0.1025]], + device='cuda:0'), grad: tensor([[ 1.3411e-07, -6.6310e-06, -2.1979e-06, ..., 5.5879e-09, + 0.0000e+00, 1.4342e-07], + [-5.8487e-07, 1.2293e-07, -5.2527e-07, ..., 1.8626e-09, + 1.8626e-09, 8.9407e-08], + [ 2.4214e-07, 3.6135e-07, 1.4734e-06, ..., 1.8626e-09, + 1.8626e-09, 5.5507e-07], + ..., + [ 2.3656e-07, 1.4342e-07, -2.8256e-06, ..., 1.8626e-09, + 1.8626e-09, 5.1036e-07], + [ 4.2282e-07, 1.0226e-06, 2.1607e-07, ..., 7.4506e-09, + 0.0000e+00, -7.0222e-07], + [-1.6708e-06, 1.5981e-06, -8.7544e-07, ..., -2.0303e-07, + 0.0000e+00, -1.6555e-05]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0325, 0.0092, 0.0208, 0.0236, 0.0311, -0.0003, -0.0031, -0.0006, + 0.0016, -0.0379], device='cuda:0'), grad: tensor([ 5.0887e-06, 5.5060e-06, 1.2256e-05, 9.3877e-06, 1.8203e-04, + 9.9316e-06, -3.4928e-05, 8.3819e-08, 7.7188e-06, -1.9717e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 217.71, cls_loss 0.0040 cls_loss_mapping 0.0052 cls_loss_causal 0.5521 re_mapping 0.0085 re_causal 0.0242 /// teacc 98.95 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.1039, 0.1444, -0.0260, ..., -0.0021, 0.0715, 0.0118], + [ 0.1089, -0.0774, 0.0235, ..., -0.0567, -0.0102, -0.0474], + [-0.0498, -0.1150, 0.0087, ..., -0.0371, 0.0405, -0.1332], + ..., + [-0.1171, -0.1030, 0.0506, ..., -0.0485, -0.0683, -0.0947], + [-0.0559, -0.0089, 0.0212, ..., 0.0260, -0.0859, 0.0214], + [-0.0423, -0.0117, 0.0720, ..., 0.1204, -0.0390, -0.1030]], + device='cuda:0'), grad: tensor([[ 5.6848e-06, -8.6501e-06, 2.0880e-06, ..., 3.1665e-08, + 0.0000e+00, 4.0978e-08], + [-2.3656e-07, 5.0105e-07, 6.3404e-06, ..., 2.2352e-08, + 0.0000e+00, 1.6950e-07], + [ 3.0920e-07, 5.5581e-06, 1.1757e-05, ..., 9.1270e-08, + 0.0000e+00, 1.8161e-06], + ..., + [ 9.3132e-08, 4.7870e-07, -2.4116e-04, ..., 3.9116e-08, + 0.0000e+00, 6.0908e-07], + [ 3.2540e-06, 6.8620e-06, 8.0764e-06, ..., -3.8184e-07, + 0.0000e+00, -4.1723e-07], + [ 1.2964e-06, 6.6385e-06, 1.8406e-04, ..., -9.6671e-07, + 0.0000e+00, 1.6261e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0329, 0.0083, 0.0209, 0.0213, 0.0306, 0.0012, -0.0022, -0.0007, + 0.0023, -0.0377], device='cuda:0'), grad: tensor([ 1.0416e-05, 1.1958e-05, 3.9190e-05, 3.7253e-05, 2.5496e-05, + 1.5393e-05, -7.1108e-05, -4.0388e-04, 3.4004e-05, 3.0136e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 218.01, cls_loss 0.0046 cls_loss_mapping 0.0066 cls_loss_causal 0.5620 re_mapping 0.0080 re_causal 0.0233 /// teacc 98.92 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.1043, 0.1445, -0.0270, ..., -0.0032, 0.0715, 0.0111], + [ 0.1090, -0.0778, 0.0225, ..., -0.0569, -0.0102, -0.0477], + [-0.0502, -0.1160, 0.0085, ..., -0.0371, 0.0406, -0.1343], + ..., + [-0.1169, -0.1051, 0.0516, ..., -0.0487, -0.0684, -0.0951], + [-0.0559, -0.0083, 0.0219, ..., 0.0263, -0.0859, 0.0226], + [-0.0423, -0.0117, 0.0726, ..., 0.1211, -0.0391, -0.1040]], + device='cuda:0'), grad: tensor([[ 1.6950e-07, -1.2703e-06, 1.8440e-07, ..., 8.0094e-08, + 0.0000e+00, 4.3772e-07], + [-1.7863e-06, -7.4506e-09, -1.3914e-06, ..., 3.7253e-08, + 0.0000e+00, 2.2165e-07], + [ 1.8608e-06, 5.4017e-08, 2.6926e-05, ..., 1.6987e-06, + 0.0000e+00, 9.3281e-06], + ..., + [ 4.6380e-07, 1.2107e-07, -2.1547e-05, ..., 8.3819e-08, + 0.0000e+00, 4.0419e-07], + [-3.0454e-06, 4.6752e-07, -1.6868e-05, ..., -4.0457e-06, + 0.0000e+00, -2.1875e-05], + [ 1.4193e-06, 3.5390e-07, 6.3814e-06, ..., 6.3702e-07, + 0.0000e+00, 6.2808e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0323, 0.0074, 0.0199, 0.0208, 0.0312, 0.0011, -0.0022, 0.0002, + 0.0037, -0.0380], device='cuda:0'), grad: tensor([ 4.3884e-06, 1.2383e-05, 5.6297e-05, 2.3231e-05, -4.8566e-04, + -2.5220e-06, 5.5730e-06, -3.6091e-05, -7.7069e-05, 4.9925e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 217.59, cls_loss 0.0033 cls_loss_mapping 0.0053 cls_loss_causal 0.5500 re_mapping 0.0076 re_causal 0.0235 /// teacc 98.85 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1051, 0.1449, -0.0274, ..., -0.0035, 0.0715, 0.0107], + [ 0.1092, -0.0781, 0.0227, ..., -0.0569, -0.0101, -0.0479], + [-0.0505, -0.1170, 0.0083, ..., -0.0373, 0.0405, -0.1349], + ..., + [-0.1170, -0.1058, 0.0515, ..., -0.0489, -0.0684, -0.0952], + [-0.0561, -0.0081, 0.0224, ..., 0.0265, -0.0859, 0.0225], + [-0.0418, -0.0118, 0.0728, ..., 0.1218, -0.0382, -0.1044]], + device='cuda:0'), grad: tensor([[ 1.2666e-07, -4.8056e-07, 3.2224e-07, ..., 4.8429e-08, + 9.3132e-10, 1.0058e-07], + [ 1.5842e-06, 1.2107e-08, -4.4797e-07, ..., 1.8626e-09, + 0.0000e+00, 1.7267e-06], + [ 4.5449e-07, -9.5926e-08, 1.7975e-07, ..., 3.9116e-08, + -1.1176e-08, 4.7870e-07], + ..., + [ 4.7497e-07, 2.4214e-08, -4.0047e-06, ..., 2.8871e-08, + 0.0000e+00, 3.2503e-07], + [ 1.1988e-05, 2.3469e-07, 7.4226e-07, ..., 7.4506e-08, + 0.0000e+00, 9.8348e-06], + [ 2.0526e-06, 2.1048e-07, 1.5097e-06, ..., -3.2224e-07, + 0.0000e+00, 1.7220e-06]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0323, 0.0073, 0.0198, 0.0209, 0.0311, 0.0016, -0.0026, 0.0002, + 0.0039, -0.0379], device='cuda:0'), grad: tensor([ 9.4771e-06, 2.8387e-06, 5.5172e-06, -3.6645e-04, -1.8245e-06, + 3.4285e-04, -2.5168e-05, -7.4655e-06, 2.1175e-05, 1.8924e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 218.26, cls_loss 0.0027 cls_loss_mapping 0.0050 cls_loss_causal 0.5722 re_mapping 0.0077 re_causal 0.0231 /// teacc 98.97 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1055, 0.1451, -0.0276, ..., -0.0034, 0.0715, 0.0104], + [ 0.1096, -0.0783, 0.0228, ..., -0.0570, -0.0100, -0.0481], + [-0.0508, -0.1175, 0.0081, ..., -0.0373, 0.0405, -0.1351], + ..., + [-0.1172, -0.1067, 0.0516, ..., -0.0491, -0.0686, -0.0954], + [-0.0564, -0.0079, 0.0227, ..., 0.0264, -0.0860, 0.0225], + [-0.0422, -0.0119, 0.0728, ..., 0.1222, -0.0385, -0.1042]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, 9.5647e-07, 1.4361e-06, ..., 4.6566e-08, + 4.6566e-09, 1.3430e-06], + [-3.5577e-07, 4.7032e-07, 3.9786e-06, ..., 1.6857e-07, + 4.2841e-08, 4.3213e-07], + [ 4.0047e-08, 1.8403e-05, 1.9759e-05, ..., 3.0734e-08, + -1.4901e-07, 2.6420e-05], + ..., + [ 4.5914e-07, 7.3668e-07, -1.3653e-06, ..., 1.8999e-07, + 9.4995e-08, 6.1281e-07], + [ 1.1101e-06, -2.7433e-05, -2.8640e-05, ..., 8.0094e-08, + 2.7940e-09, -3.9130e-05], + [ 7.1712e-06, 6.9812e-06, -3.2753e-05, ..., -3.3006e-06, + 9.3132e-10, 2.6412e-06]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0321, 0.0075, 0.0196, 0.0207, 0.0315, 0.0018, -0.0026, 0.0002, + 0.0040, -0.0382], device='cuda:0'), grad: tensor([ 5.6252e-06, 7.8082e-06, 6.1572e-05, 3.1769e-05, 3.9428e-05, + -1.6987e-05, 4.0755e-06, 1.8859e-06, -1.1277e-04, -2.2426e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 217.80, cls_loss 0.0038 cls_loss_mapping 0.0050 cls_loss_causal 0.5840 re_mapping 0.0076 re_causal 0.0237 /// teacc 98.84 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1054, 0.1458, -0.0276, ..., -0.0035, 0.0716, 0.0099], + [ 0.1113, -0.0789, 0.0234, ..., -0.0570, -0.0100, -0.0485], + [-0.0522, -0.1180, 0.0079, ..., -0.0373, 0.0405, -0.1355], + ..., + [-0.1180, -0.1072, 0.0514, ..., -0.0492, -0.0686, -0.0955], + [-0.0566, -0.0077, 0.0229, ..., 0.0265, -0.0861, 0.0225], + [-0.0434, -0.0126, 0.0731, ..., 0.1229, -0.0388, -0.1048]], + device='cuda:0'), grad: tensor([[ 5.2527e-07, -1.3160e-06, 3.5670e-07, ..., 1.2759e-07, + 4.6566e-09, 5.4203e-07], + [-7.1079e-06, -1.1642e-07, -7.0184e-06, ..., 7.4506e-09, + -6.6124e-08, 1.9353e-06], + [ 7.6275e-07, 1.8924e-06, 2.9951e-06, ..., 5.4482e-07, + 2.7940e-09, 2.0918e-06], + ..., + [ 2.3320e-06, 3.9581e-07, -2.4959e-07, ..., 3.8184e-08, + -4.8429e-08, 1.4221e-06], + [ 1.3985e-05, 5.3681e-06, -8.8196e-07, ..., -9.5367e-07, + 5.3085e-08, 2.4781e-05], + [ 1.9837e-06, 2.1979e-06, 2.0079e-06, ..., 3.7253e-08, + 3.3528e-08, 1.8878e-06]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0325, 0.0083, 0.0180, 0.0211, 0.0311, 0.0019, -0.0030, 0.0011, + 0.0038, -0.0383], device='cuda:0'), grad: tensor([ 4.6846e-07, -6.1318e-06, 8.3521e-06, -1.1581e-04, 1.1958e-06, + 5.2780e-05, 3.9898e-06, 2.1607e-06, 4.5419e-05, 7.4133e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 217.48, cls_loss 0.0041 cls_loss_mapping 0.0064 cls_loss_causal 0.5520 re_mapping 0.0078 re_causal 0.0234 /// teacc 98.78 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1054, 0.1473, -0.0276, ..., -0.0036, 0.0721, 0.0101], + [ 0.1125, -0.0787, 0.0243, ..., -0.0571, -0.0093, -0.0487], + [-0.0529, -0.1191, 0.0082, ..., -0.0374, 0.0411, -0.1358], + ..., + [-0.1189, -0.1090, 0.0507, ..., -0.0494, -0.0697, -0.0957], + [-0.0570, -0.0074, 0.0229, ..., 0.0265, -0.0863, 0.0227], + [-0.0442, -0.0130, 0.0742, ..., 0.1234, -0.0389, -0.1053]], + device='cuda:0'), grad: tensor([[ 3.6322e-07, -1.6503e-06, 2.0191e-06, ..., 9.3132e-10, + 3.0510e-06, 3.5204e-07], + [ 8.9407e-08, 2.3469e-07, 6.5899e-04, ..., 9.3132e-10, + 2.4319e-04, 2.4494e-07], + [ 2.3749e-07, 5.3085e-07, -4.9305e-04, ..., 5.5879e-09, + -2.4867e-04, 2.4401e-07], + ..., + [ 5.4389e-07, 5.3458e-07, -1.7822e-04, ..., 6.5193e-09, + 1.2526e-06, 4.9267e-07], + [ 4.0904e-06, 3.7849e-06, 1.9204e-06, ..., -1.3039e-08, + 1.4836e-06, 3.8221e-06], + [ 8.7731e-07, 2.0433e-06, 7.4040e-07, ..., -1.3970e-08, + 1.6764e-07, 8.7637e-07]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0338, 0.0092, 0.0176, 0.0224, 0.0307, 0.0012, -0.0039, 0.0003, + 0.0036, -0.0379], device='cuda:0'), grad: tensor([ 2.2545e-05, 1.9989e-03, -1.8063e-03, 4.1604e-05, 1.8704e-04, + -4.7714e-05, -1.9407e-04, -2.2614e-04, 1.9088e-05, 4.5113e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 217.35, cls_loss 0.0043 cls_loss_mapping 0.0055 cls_loss_causal 0.5786 re_mapping 0.0074 re_causal 0.0228 /// teacc 98.91 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.1060, 0.1477, -0.0277, ..., -0.0037, 0.0722, 0.0096], + [ 0.1129, -0.0794, 0.0248, ..., -0.0572, -0.0117, -0.0490], + [-0.0526, -0.1198, 0.0078, ..., -0.0375, 0.0427, -0.1362], + ..., + [-0.1199, -0.1104, 0.0505, ..., -0.0494, -0.0696, -0.0961], + [-0.0576, -0.0075, 0.0232, ..., 0.0263, -0.0868, 0.0230], + [-0.0448, -0.0128, 0.0742, ..., 0.1243, -0.0398, -0.1063]], + device='cuda:0'), grad: tensor([[ 7.7020e-07, -1.1455e-06, 7.7300e-07, ..., 2.7940e-09, + 3.7625e-07, 6.5193e-09], + [-3.7402e-06, 2.0489e-08, -1.5134e-06, ..., 7.4506e-09, + -1.3439e-06, 1.0245e-08], + [ 3.1386e-07, 6.5193e-08, 2.2873e-06, ..., 1.6764e-08, + 1.3039e-07, 3.1665e-08], + ..., + [ 4.9919e-07, 2.8871e-08, -9.7528e-06, ..., 1.6764e-08, + 1.5832e-07, 1.7695e-08], + [ 6.4168e-07, 5.0012e-07, 1.1008e-06, ..., -1.2480e-07, + 1.9558e-07, -2.2259e-07], + [ 8.0466e-07, 3.4180e-07, 8.8215e-06, ..., 3.4459e-08, + 2.2631e-07, 1.0990e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([ 3.3713e-02, 9.1661e-03, 1.7731e-02, 2.3006e-02, 3.1059e-02, + 6.9162e-04, -3.2173e-03, 1.7772e-06, 3.4017e-03, -3.8202e-02], + device='cuda:0'), grad: tensor([ 9.0618e-07, -6.8638e-07, 2.8890e-06, 3.0175e-06, -2.2188e-05, + 3.6228e-07, 6.8266e-07, -1.7494e-05, 2.6599e-06, 2.9817e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 217.62, cls_loss 0.0036 cls_loss_mapping 0.0046 cls_loss_causal 0.5508 re_mapping 0.0075 re_causal 0.0230 /// teacc 98.84 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.1063, 0.1480, -0.0277, ..., -0.0038, 0.0723, 0.0090], + [ 0.1126, -0.0798, 0.0248, ..., -0.0572, -0.0113, -0.0492], + [-0.0522, -0.1199, 0.0078, ..., -0.0376, 0.0427, -0.1362], + ..., + [-0.1201, -0.1110, 0.0517, ..., -0.0497, -0.0699, -0.0964], + [-0.0572, -0.0075, 0.0236, ..., 0.0261, -0.0863, 0.0234], + [-0.0452, -0.0123, 0.0726, ..., 0.1256, -0.0413, -0.1069]], + device='cuda:0'), grad: tensor([[ 7.4506e-08, -6.7204e-06, -3.3993e-07, ..., 0.0000e+00, + -1.0617e-07, -2.1700e-07], + [ 9.3132e-09, 5.4017e-08, 3.0734e-07, ..., 0.0000e+00, + 8.2515e-07, 7.3574e-08], + [-4.7404e-07, 4.2748e-07, 1.6857e-07, ..., 0.0000e+00, + -9.5740e-07, 1.3132e-07], + ..., + [ 7.9162e-07, 1.8813e-07, -1.6857e-06, ..., 0.0000e+00, + 9.2201e-08, 7.5437e-08], + [ 2.2817e-07, 1.6596e-06, 2.5239e-07, ..., 9.3132e-10, + 4.7497e-08, 6.5193e-08], + [ 2.1514e-07, 8.2515e-07, 3.3900e-07, ..., -2.7940e-09, + 2.0489e-08, 7.3574e-08]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0337, 0.0088, 0.0182, 0.0230, 0.0314, 0.0004, -0.0030, 0.0005, + 0.0035, -0.0392], device='cuda:0'), grad: tensor([-9.2313e-06, 1.7300e-05, -1.3538e-05, 3.3509e-06, -5.4270e-05, + 7.2829e-07, 1.0148e-05, 2.9393e-06, 2.1860e-05, 2.0653e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 217.43, cls_loss 0.0037 cls_loss_mapping 0.0046 cls_loss_causal 0.5439 re_mapping 0.0076 re_causal 0.0227 /// teacc 98.81 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1069, 0.1469, -0.0301, ..., -0.0061, 0.0724, 0.0088], + [ 0.1130, -0.0799, 0.0247, ..., -0.0572, -0.0110, -0.0493], + [-0.0524, -0.1199, 0.0076, ..., -0.0376, 0.0425, -0.1364], + ..., + [-0.1205, -0.1117, 0.0513, ..., -0.0498, -0.0702, -0.0965], + [-0.0574, -0.0072, 0.0239, ..., 0.0261, -0.0863, 0.0239], + [-0.0449, -0.0111, 0.0746, ..., 0.1274, -0.0415, -0.1082]], + device='cuda:0'), grad: tensor([[ 4.3865e-07, -1.6809e-05, -2.6952e-06, ..., 1.3039e-08, + 1.4156e-07, 1.5646e-07], + [ 5.9046e-07, 1.5832e-07, 5.1439e-05, ..., 1.0245e-08, + 1.5646e-07, 3.0082e-07], + [ 1.7043e-07, 2.2687e-06, 9.1121e-06, ..., 1.4901e-08, + 2.0284e-06, -1.1921e-05], + ..., + [-3.1292e-06, 1.7229e-07, -7.6294e-05, ..., 4.6566e-08, + -2.7288e-06, 3.6694e-07], + [ 1.9334e-06, 1.1055e-06, 9.1344e-06, ..., 6.1281e-07, + 4.2841e-08, 1.0133e-05], + [ 6.3982e-07, 3.0398e-06, -7.3481e-07, ..., -1.4612e-06, + 2.0955e-07, 1.0775e-06]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0313, 0.0086, 0.0181, 0.0226, 0.0314, 0.0006, -0.0021, 0.0003, + 0.0037, -0.0376], device='cuda:0'), grad: tensor([-3.6955e-05, 7.1287e-05, -4.8548e-05, 4.8220e-05, 3.2187e-06, + -1.7226e-05, 2.2724e-05, -1.0520e-04, 5.3734e-05, 8.7023e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 217.73, cls_loss 0.0030 cls_loss_mapping 0.0041 cls_loss_causal 0.5293 re_mapping 0.0075 re_causal 0.0221 /// teacc 98.77 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1074, 0.1473, -0.0301, ..., -0.0063, 0.0723, 0.0086], + [ 0.1136, -0.0799, 0.0248, ..., -0.0573, -0.0109, -0.0494], + [-0.0530, -0.1203, 0.0075, ..., -0.0376, 0.0425, -0.1367], + ..., + [-0.1209, -0.1121, 0.0515, ..., -0.0500, -0.0699, -0.0966], + [-0.0577, -0.0073, 0.0239, ..., 0.0261, -0.0863, 0.0239], + [-0.0455, -0.0115, 0.0748, ..., 0.1287, -0.0416, -0.1087]], + device='cuda:0'), grad: tensor([[ 5.8021e-07, -7.5158e-07, 1.8999e-07, ..., 2.7940e-09, + 2.2352e-08, 5.3551e-07], + [ 3.4459e-07, 1.1362e-07, 8.0280e-07, ..., 4.6566e-09, + 4.7497e-07, 4.0047e-07], + [ 9.9652e-08, 6.1467e-08, -1.7416e-07, ..., 9.3132e-10, + 2.0489e-08, 3.0361e-07], + ..., + [ 2.6450e-07, 7.4506e-08, -1.0077e-06, ..., 8.5682e-08, + -3.9209e-07, 2.6729e-07], + [ 1.7896e-05, 1.1049e-05, 4.7497e-07, ..., 9.5926e-08, + 1.1176e-08, 6.7174e-05], + [ 9.4622e-07, 3.6694e-07, 9.2238e-06, ..., -4.4797e-07, + 2.0489e-07, 8.7544e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([ 3.1420e-02, 8.8596e-03, 1.7808e-02, 2.3150e-02, 3.0879e-02, + -2.0179e-05, -1.9798e-03, 5.6698e-04, 3.4694e-03, -3.7435e-02], + device='cuda:0'), grad: tensor([ 1.9632e-06, 5.2303e-06, -1.1986e-06, -1.9622e-04, -5.5820e-05, + -1.2529e-04, 5.2899e-05, 2.7083e-06, 2.6727e-04, 4.8637e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.70, cls_loss 0.0031 cls_loss_mapping 0.0062 cls_loss_causal 0.5317 re_mapping 0.0080 re_causal 0.0232 /// teacc 98.88 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1077, 0.1477, -0.0302, ..., -0.0066, 0.0722, 0.0084], + [ 0.1154, -0.0796, 0.0263, ..., -0.0573, -0.0102, -0.0493], + [-0.0532, -0.1206, 0.0074, ..., -0.0377, 0.0423, -0.1369], + ..., + [-0.1228, -0.1127, 0.0505, ..., -0.0504, -0.0686, -0.0968], + [-0.0575, -0.0063, 0.0241, ..., 0.0262, -0.0866, 0.0247], + [-0.0459, -0.0122, 0.0761, ..., 0.1307, -0.0419, -0.1105]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, -6.4634e-07, -1.1837e-06, ..., -1.4622e-07, + 7.4506e-09, 1.7872e-06], + [-1.1921e-07, 1.3690e-07, -1.9558e-08, ..., 2.0489e-08, + 2.4214e-08, 9.6858e-08], + [ 2.0489e-08, 9.8720e-07, 5.5321e-07, ..., 5.2154e-08, + -7.7300e-08, 4.3027e-07], + ..., + [ 2.5146e-08, 1.9278e-07, 1.1455e-07, ..., 1.4901e-08, + 8.4750e-08, 1.0058e-07], + [ 8.7544e-08, -2.0787e-05, -5.9269e-06, ..., -1.2107e-07, + 5.5879e-09, -1.4238e-05], + [ 2.3842e-07, 2.3991e-06, 1.1111e-06, ..., 1.1176e-08, + 5.5879e-09, 2.1979e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0314, 0.0107, 0.0178, 0.0232, 0.0292, -0.0003, -0.0024, -0.0007, + 0.0038, -0.0363], device='cuda:0'), grad: tensor([-1.1986e-06, 1.2312e-06, 1.3076e-06, 5.1968e-06, -9.1046e-06, + 1.4372e-05, 8.2478e-06, 1.2610e-06, -2.8193e-05, 6.9141e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 217.88, cls_loss 0.0031 cls_loss_mapping 0.0046 cls_loss_causal 0.5421 re_mapping 0.0078 re_causal 0.0231 /// teacc 98.90 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1081, 0.1480, -0.0301, ..., -0.0065, 0.0720, 0.0078], + [ 0.1151, -0.0796, 0.0266, ..., -0.0573, -0.0097, -0.0492], + [-0.0519, -0.1209, 0.0073, ..., -0.0377, 0.0426, -0.1372], + ..., + [-0.1236, -0.1132, 0.0506, ..., -0.0505, -0.0687, -0.0971], + [-0.0578, -0.0061, 0.0239, ..., 0.0263, -0.0868, 0.0248], + [-0.0462, -0.0127, 0.0757, ..., 0.1310, -0.0419, -0.1112]], + device='cuda:0'), grad: tensor([[-2.8282e-05, -5.0187e-05, -2.5071e-06, ..., 9.3132e-10, + 5.5879e-09, 2.7940e-09], + [-7.0874e-07, 1.6205e-07, -6.4075e-07, ..., 9.3132e-10, + 3.1665e-08, 5.5879e-09], + [ 1.8068e-07, 2.0582e-07, 1.6019e-07, ..., 0.0000e+00, + 3.4459e-08, 1.4901e-08], + ..., + [ 1.8906e-07, 1.4715e-07, 1.6764e-08, ..., 1.2107e-08, + 3.1665e-08, 8.3819e-09], + [ 1.0617e-07, 1.4715e-07, 4.7497e-08, ..., 2.7940e-09, + 1.3970e-08, 2.0489e-08], + [ 1.6736e-06, 3.0175e-06, 1.5739e-07, ..., -7.3574e-08, + 9.2201e-08, 2.0489e-08]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0315, 0.0103, 0.0186, 0.0228, 0.0300, 0.0003, -0.0021, -0.0006, + 0.0034, -0.0372], device='cuda:0'), grad: tensor([-6.9320e-05, -5.0478e-07, -9.1363e-07, 2.2668e-06, -1.0416e-05, + 5.6446e-05, 1.5065e-05, 6.5845e-07, 5.2247e-07, 6.0052e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 217.71, cls_loss 0.0043 cls_loss_mapping 0.0061 cls_loss_causal 0.5704 re_mapping 0.0077 re_causal 0.0225 /// teacc 98.93 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1101, 0.1484, -0.0302, ..., -0.0070, 0.0720, 0.0067], + [ 0.1155, -0.0799, 0.0269, ..., -0.0575, -0.0090, -0.0494], + [-0.0521, -0.1213, 0.0072, ..., -0.0378, 0.0427, -0.1378], + ..., + [-0.1240, -0.1142, 0.0505, ..., -0.0512, -0.0687, -0.0973], + [-0.0581, -0.0059, 0.0238, ..., 0.0262, -0.0874, 0.0249], + [-0.0463, -0.0136, 0.0758, ..., 0.1319, -0.0440, -0.1125]], + device='cuda:0'), grad: tensor([[ 2.6915e-07, -1.0896e-07, 1.8934e-06, ..., 1.1558e-06, + 9.2015e-07, 1.7788e-07], + [-2.5138e-05, 4.0047e-08, -4.6283e-05, ..., 8.8289e-07, + -2.7008e-08, 3.0827e-07], + [ 4.3772e-07, 4.8429e-08, 1.2638e-06, ..., 4.1444e-07, + 3.7532e-07, 3.5204e-07], + ..., + [ 1.3992e-05, 7.6368e-08, 5.8591e-05, ..., 1.1548e-05, + 1.2200e-07, 5.0850e-07], + [ 1.5479e-06, 4.5262e-07, 8.1956e-06, ..., 2.3469e-06, + 7.0594e-07, 1.8459e-06], + [ 8.3819e-06, 1.6764e-07, -1.0252e-04, ..., -8.0943e-05, + -5.6267e-05, 2.0955e-07]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0312, 0.0104, 0.0184, 0.0244, 0.0307, -0.0009, -0.0017, -0.0003, + 0.0032, -0.0381], device='cuda:0'), grad: tensor([ 8.8140e-06, -5.3823e-05, 4.0978e-06, -4.4018e-05, 3.2759e-04, + 4.3988e-05, 5.5581e-06, 7.6473e-05, 2.4974e-05, -3.9411e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 217.82, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5846 re_mapping 0.0076 re_causal 0.0235 /// teacc 98.84 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.1103, 0.1490, -0.0301, ..., -0.0071, 0.0720, 0.0067], + [ 0.1157, -0.0800, 0.0269, ..., -0.0581, -0.0091, -0.0496], + [-0.0522, -0.1215, 0.0070, ..., -0.0378, 0.0425, -0.1380], + ..., + [-0.1241, -0.1151, 0.0509, ..., -0.0515, -0.0687, -0.0973], + [-0.0584, -0.0060, 0.0235, ..., 0.0261, -0.0873, 0.0252], + [-0.0465, -0.0141, 0.0759, ..., 0.1324, -0.0441, -0.1129]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, -5.2899e-07, -9.7789e-08, ..., 3.7253e-09, + 7.4506e-09, 3.3528e-08], + [-5.9046e-07, 1.0338e-07, 1.6699e-06, ..., 9.3132e-10, + 8.1025e-08, 1.4249e-06], + [ 7.8231e-08, 1.6764e-07, -6.0081e-05, ..., 3.7253e-09, + -3.6322e-08, 5.5134e-07], + ..., + [ 2.4121e-07, 6.1374e-07, 6.8247e-05, ..., 7.4506e-09, + -1.4808e-07, 8.7917e-06], + [ 2.8405e-07, -8.7731e-07, -1.1675e-05, ..., 2.5146e-08, + 1.6764e-08, -1.0863e-05], + [ 1.7509e-07, 3.7625e-07, 3.0082e-07, ..., -7.8231e-08, + 4.7497e-08, 3.1572e-07]], device='cuda:0') +Epoch 154, bias, value: tensor([ 3.1544e-02, 1.0238e-02, 1.8396e-02, 2.3986e-02, 3.0981e-02, + -8.9760e-04, -1.6696e-03, -4.6329e-05, 2.9142e-03, -3.8299e-02], + device='cuda:0'), grad: tensor([ 2.4401e-07, 4.9025e-06, -3.0446e-04, -2.1793e-07, 1.2824e-06, + 1.4091e-06, -8.2795e-07, 3.1829e-04, -2.1368e-05, 9.9279e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 218.20, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5249 re_mapping 0.0074 re_causal 0.0224 /// teacc 98.94 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1108, 0.1494, -0.0299, ..., -0.0071, 0.0721, 0.0067], + [ 0.1158, -0.0801, 0.0268, ..., -0.0583, -0.0086, -0.0497], + [-0.0524, -0.1217, 0.0069, ..., -0.0379, 0.0425, -0.1382], + ..., + [-0.1242, -0.1155, 0.0510, ..., -0.0517, -0.0686, -0.0976], + [-0.0586, -0.0061, 0.0236, ..., 0.0261, -0.0874, 0.0254], + [-0.0469, -0.0144, 0.0759, ..., 0.1327, -0.0442, -0.1134]], + device='cuda:0'), grad: tensor([[ 3.2596e-07, -7.4059e-06, -3.1870e-06, ..., 9.3132e-10, + -5.1502e-07, 7.8883e-07], + [-1.2154e-06, 9.4902e-07, 6.8825e-07, ..., 9.3132e-10, + -1.3597e-07, 1.6671e-07], + [ 3.1292e-07, 7.9274e-06, 1.0431e-05, ..., 9.3132e-10, + 1.4529e-07, 7.3723e-06], + ..., + [ 7.0408e-07, 2.5705e-07, -9.5889e-06, ..., 1.0245e-08, + 1.3411e-07, 1.3970e-07], + [ 1.6605e-06, -3.6135e-06, -5.8338e-06, ..., 4.6566e-09, + 5.2713e-07, -7.6219e-06], + [ 1.5199e-06, 2.1253e-06, 4.2655e-06, ..., -6.3330e-08, + 3.2503e-07, 7.2364e-07]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0317, 0.0101, 0.0183, 0.0236, 0.0310, -0.0009, -0.0012, 0.0003, + 0.0027, -0.0385], device='cuda:0'), grad: tensor([-5.9940e-06, 3.6471e-06, 4.4584e-05, 6.2305e-07, -1.1794e-05, + -1.6987e-05, 1.8433e-05, -1.7479e-05, -3.1739e-05, 1.6689e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 217.51, cls_loss 0.0029 cls_loss_mapping 0.0045 cls_loss_causal 0.5492 re_mapping 0.0074 re_causal 0.0222 /// teacc 98.81 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1120, 0.1499, -0.0299, ..., -0.0071, 0.0725, 0.0065], + [ 0.1157, -0.0804, 0.0267, ..., -0.0582, -0.0094, -0.0499], + [-0.0519, -0.1227, 0.0079, ..., -0.0380, 0.0433, -0.1385], + ..., + [-0.1243, -0.1165, 0.0508, ..., -0.0519, -0.0690, -0.0979], + [-0.0589, -0.0058, 0.0237, ..., 0.0263, -0.0875, 0.0258], + [-0.0476, -0.0149, 0.0758, ..., 0.1327, -0.0446, -0.1140]], + device='cuda:0'), grad: tensor([[ 1.0524e-06, 3.3230e-06, 5.4911e-06, ..., 1.3039e-08, + 1.2955e-06, 3.0976e-06], + [-6.2734e-06, 6.4634e-07, 2.1920e-05, ..., 2.0489e-08, + 2.3991e-06, 3.3621e-07], + [ 2.0545e-06, 1.2973e-06, -3.2067e-04, ..., 3.0734e-08, + -5.3197e-05, 9.3970e-07], + ..., + [ 6.1616e-06, 1.7136e-06, 2.8324e-04, ..., 6.2399e-08, + 4.8190e-05, 6.7614e-07], + [ 5.1484e-06, -7.9796e-06, -4.7870e-07, ..., 4.0047e-08, + -1.2917e-06, -8.6054e-06], + [ 2.1048e-06, 1.9837e-06, -2.0452e-06, ..., -8.3353e-07, + 2.0210e-07, 8.2515e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0320, 0.0098, 0.0194, 0.0238, 0.0315, -0.0009, -0.0016, -0.0004, + 0.0029, -0.0389], device='cuda:0'), grad: tensor([ 2.1160e-05, 6.9737e-05, -1.0958e-03, 1.9789e-05, 1.4886e-05, + -1.5333e-05, 3.9190e-06, 9.7656e-04, 3.6303e-06, 1.8515e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 217.24, cls_loss 0.0052 cls_loss_mapping 0.0070 cls_loss_causal 0.5625 re_mapping 0.0077 re_causal 0.0217 /// teacc 98.76 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1124, 0.1508, -0.0296, ..., -0.0073, 0.0724, 0.0062], + [ 0.1140, -0.0835, 0.0260, ..., -0.0591, -0.0108, -0.0519], + [-0.0506, -0.1234, 0.0080, ..., -0.0380, 0.0448, -0.1390], + ..., + [-0.1246, -0.1178, 0.0517, ..., -0.0522, -0.0693, -0.0975], + [-0.0585, -0.0053, 0.0241, ..., 0.0265, -0.0874, 0.0266], + [-0.0507, -0.0163, 0.0755, ..., 0.1332, -0.0453, -0.1156]], + device='cuda:0'), grad: tensor([[ 6.5845e-07, -4.5821e-07, 4.0140e-07, ..., 4.0047e-08, + 0.0000e+00, 1.2107e-06], + [ 3.4459e-08, 1.3504e-07, 7.1339e-07, ..., 3.1665e-08, + 0.0000e+00, 2.8964e-07], + [ 1.6764e-08, 2.3190e-07, 5.1130e-07, ..., 1.3970e-08, + 0.0000e+00, 2.4401e-07], + ..., + [ 1.4901e-08, 8.1211e-06, 4.7028e-05, ..., 2.1867e-06, + 0.0000e+00, 1.8671e-05], + [ 9.1270e-07, 3.1888e-06, 1.7390e-05, ..., 8.2050e-07, + 0.0000e+00, 7.3612e-06], + [ 6.7055e-08, -6.6757e-05, -3.9911e-04, ..., -2.0117e-05, + 0.0000e+00, -1.5342e-04]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0323, 0.0079, 0.0207, 0.0219, 0.0328, 0.0010, -0.0014, 0.0003, + 0.0031, -0.0405], device='cuda:0'), grad: tensor([ 8.6278e-06, 2.4419e-06, 1.0170e-06, 2.8498e-07, 1.1867e-04, + 9.4748e-04, -9.8586e-05, 1.4007e-04, 5.7667e-05, -1.1778e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 217.96, cls_loss 0.0033 cls_loss_mapping 0.0054 cls_loss_causal 0.5445 re_mapping 0.0077 re_causal 0.0218 /// teacc 98.91 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.1127, 0.1514, -0.0297, ..., -0.0074, 0.0723, 0.0064], + [ 0.1146, -0.0859, 0.0275, ..., -0.0596, -0.0113, -0.0528], + [-0.0508, -0.1245, 0.0078, ..., -0.0383, 0.0449, -0.1395], + ..., + [-0.1247, -0.1185, 0.0525, ..., -0.0526, -0.0696, -0.0976], + [-0.0610, -0.0056, 0.0214, ..., 0.0266, -0.0876, 0.0262], + [-0.0513, -0.0163, 0.0758, ..., 0.1336, -0.0453, -0.1159]], + device='cuda:0'), grad: tensor([[ 1.5274e-07, -1.5590e-06, 2.0862e-07, ..., 1.6764e-08, + 0.0000e+00, 1.1176e-07], + [-3.5390e-06, 2.9802e-08, -2.0750e-06, ..., 9.3132e-09, + 0.0000e+00, 1.0058e-07], + [ 9.6112e-07, 2.0117e-07, 1.2144e-05, ..., 5.0291e-08, + 0.0000e+00, 5.7183e-07], + ..., + [ 1.9185e-06, 7.6368e-08, -1.2696e-05, ..., 5.4017e-08, + 0.0000e+00, 1.3225e-07], + [ 3.2485e-06, 4.0233e-06, 8.4192e-07, ..., 1.0245e-07, + 0.0000e+00, 1.7509e-06], + [ 1.1977e-06, 1.6913e-06, -1.0140e-05, ..., -1.6652e-06, + 0.0000e+00, 9.7044e-07]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0325, 0.0090, 0.0209, 0.0219, 0.0332, 0.0017, -0.0022, 0.0010, + 0.0005, -0.0409], device='cuda:0'), grad: tensor([-3.3900e-07, -7.3016e-07, 2.4214e-05, 1.9167e-06, 1.8090e-05, + -5.4203e-06, 6.9849e-07, -2.7806e-05, 7.3649e-06, -1.7956e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 217.57, cls_loss 0.0028 cls_loss_mapping 0.0034 cls_loss_causal 0.5456 re_mapping 0.0075 re_causal 0.0218 /// teacc 98.89 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1128, 0.1519, -0.0295, ..., -0.0074, 0.0723, 0.0073], + [ 0.1150, -0.0859, 0.0275, ..., -0.0597, -0.0111, -0.0530], + [-0.0511, -0.1252, 0.0063, ..., -0.0383, 0.0449, -0.1401], + ..., + [-0.1249, -0.1192, 0.0536, ..., -0.0527, -0.0695, -0.0978], + [-0.0611, -0.0054, 0.0215, ..., 0.0266, -0.0878, 0.0266], + [-0.0514, -0.0166, 0.0756, ..., 0.1341, -0.0454, -0.1162]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, -2.4587e-07, 1.5832e-07, ..., 3.3528e-08, + 7.8417e-07, 1.5460e-07], + [-1.2424e-06, 2.4214e-08, -3.3602e-06, ..., 1.3039e-08, + 9.1083e-07, 1.7136e-07], + [ 1.0245e-07, 1.3225e-07, 6.1840e-07, ..., 5.5879e-09, + -9.6634e-06, 7.6741e-07], + ..., + [ 1.0133e-06, 3.7253e-08, 3.1628e-06, ..., 1.5087e-07, + 4.6194e-06, 2.3097e-07], + [ 3.5390e-07, -2.9802e-07, 1.6764e-07, ..., 3.2410e-07, + 4.2655e-07, -1.3188e-06], + [ 3.1665e-08, -5.5879e-08, -2.3507e-06, ..., -1.3858e-06, + 2.2352e-07, -2.1253e-06]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0326, 0.0089, 0.0197, 0.0218, 0.0334, 0.0016, -0.0026, 0.0023, + 0.0005, -0.0411], device='cuda:0'), grad: tensor([ 4.0084e-06, -2.1048e-07, -4.0859e-05, 8.0764e-06, -6.3479e-05, + 5.1409e-06, -1.6131e-06, 2.6375e-05, 3.1181e-06, 5.9456e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 217.78, cls_loss 0.0036 cls_loss_mapping 0.0048 cls_loss_causal 0.5640 re_mapping 0.0071 re_causal 0.0214 /// teacc 98.80 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.1146, 0.1510, -0.0294, ..., -0.0074, 0.0732, 0.0073], + [ 0.1157, -0.0876, 0.0275, ..., -0.0599, -0.0109, -0.0535], + [-0.0515, -0.1261, 0.0061, ..., -0.0384, 0.0449, -0.1407], + ..., + [-0.1257, -0.1208, 0.0539, ..., -0.0531, -0.0697, -0.0983], + [-0.0611, -0.0049, 0.0215, ..., 0.0266, -0.0882, 0.0274], + [-0.0518, -0.0168, 0.0758, ..., 0.1348, -0.0455, -0.1166]], + device='cuda:0'), grad: tensor([[-4.2841e-07, -1.1176e-06, 7.2531e-06, ..., 2.0508e-06, + 2.6077e-08, 3.2820e-06], + [ 9.5554e-07, 5.7928e-06, 8.6948e-06, ..., 3.4645e-07, + 2.2911e-07, 6.4112e-06], + [-9.2573e-07, 7.3574e-07, 1.6931e-06, ..., 7.4506e-08, + -4.5635e-07, 4.5821e-07], + ..., + [ 7.0222e-07, 1.6913e-06, -3.5670e-06, ..., 7.4133e-07, + -3.1665e-08, 1.4529e-06], + [-5.5283e-06, -8.2552e-06, 5.9530e-06, ..., 5.1185e-06, + 3.9116e-08, -2.0444e-05], + [ 5.1782e-07, -2.0668e-05, -4.1038e-05, ..., -1.0453e-05, + 2.6077e-08, -1.2532e-05]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0313, 0.0087, 0.0196, 0.0216, 0.0337, 0.0018, -0.0013, 0.0023, + 0.0006, -0.0412], device='cuda:0'), grad: tensor([ 1.2264e-05, 2.8893e-05, -2.4930e-05, 1.4000e-05, 5.5656e-06, + 3.7491e-05, 6.9104e-06, -2.8443e-06, -1.1623e-06, -7.6294e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 217.63, cls_loss 0.0031 cls_loss_mapping 0.0043 cls_loss_causal 0.5365 re_mapping 0.0075 re_causal 0.0218 /// teacc 98.82 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1156, 0.1517, -0.0293, ..., -0.0074, 0.0731, 0.0073], + [ 0.1160, -0.0876, 0.0274, ..., -0.0598, -0.0110, -0.0536], + [-0.0518, -0.1264, 0.0064, ..., -0.0385, 0.0449, -0.1407], + ..., + [-0.1262, -0.1218, 0.0540, ..., -0.0544, -0.0696, -0.0993], + [-0.0611, -0.0043, 0.0214, ..., 0.0266, -0.0882, 0.0280], + [-0.0520, -0.0172, 0.0760, ..., 0.1354, -0.0457, -0.1174]], + device='cuda:0'), grad: tensor([[ 3.1292e-07, -9.7789e-07, -3.1479e-07, ..., 5.5879e-09, + 1.8626e-09, 4.8429e-08], + [ 1.1897e-04, 9.1270e-08, -4.6790e-06, ..., 9.3132e-09, + -5.5879e-09, 7.4506e-08], + [-1.3423e-04, -1.2666e-07, 5.1782e-07, ..., 5.5879e-09, + -1.5460e-07, 8.7544e-08], + ..., + [ 3.1516e-06, 2.3097e-07, 2.5090e-06, ..., 3.9116e-08, + 1.5274e-07, 2.2165e-07], + [ 2.2631e-06, -1.6391e-05, -4.5300e-06, ..., -1.3039e-08, + 0.0000e+00, -1.8463e-05], + [ 5.6252e-06, 8.8103e-07, 3.7067e-07, ..., -2.2352e-07, + 1.8626e-09, 1.2293e-07]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0316, 0.0085, 0.0200, 0.0216, 0.0337, 0.0018, -0.0014, 0.0022, + 0.0005, -0.0412], device='cuda:0'), grad: tensor([-1.7509e-07, 1.7965e-04, -2.0707e-04, 2.6263e-06, 2.5183e-06, + 2.7806e-05, 2.6450e-07, 8.7470e-06, -2.2843e-05, 8.8513e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 217.76, cls_loss 0.0029 cls_loss_mapping 0.0056 cls_loss_causal 0.5557 re_mapping 0.0070 re_causal 0.0217 /// teacc 98.93 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1166, 0.1520, -0.0296, ..., -0.0075, 0.0731, 0.0077], + [ 0.1170, -0.0881, 0.0270, ..., -0.0604, -0.0104, -0.0537], + [-0.0524, -0.1268, 0.0063, ..., -0.0386, 0.0448, -0.1412], + ..., + [-0.1264, -0.1228, 0.0550, ..., -0.0546, -0.0685, -0.0996], + [-0.0616, -0.0041, 0.0213, ..., 0.0266, -0.0887, 0.0285], + [-0.0522, -0.0174, 0.0772, ..., 0.1367, -0.0459, -0.1178]], + device='cuda:0'), grad: tensor([[ 2.5146e-07, -2.9758e-05, -1.0341e-05, ..., 1.8254e-07, + 0.0000e+00, 6.1467e-08], + [ 6.1654e-07, 1.1921e-07, 2.8871e-07, ..., 7.2643e-08, + 0.0000e+00, 3.7253e-08], + [ 4.6566e-08, 2.6450e-07, 3.5763e-07, ..., 2.6077e-08, + 0.0000e+00, 9.8720e-08], + ..., + [ 5.7742e-08, 2.9057e-07, 1.4141e-05, ..., 4.1127e-06, + 0.0000e+00, 8.0094e-08], + [ 2.4773e-07, -4.2394e-06, -7.0967e-06, ..., 2.1234e-07, + 0.0000e+00, -3.2187e-06], + [ 1.0431e-07, 8.1882e-06, -1.0408e-05, ..., -5.6997e-06, + 0.0000e+00, 2.6319e-06]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0314, 0.0084, 0.0198, 0.0215, 0.0321, 0.0018, -0.0012, 0.0030, + 0.0004, -0.0401], device='cuda:0'), grad: tensor([-4.5359e-05, 5.9977e-06, 1.0524e-06, 5.2303e-06, 4.1500e-06, + 4.5598e-06, 2.1309e-05, 2.6241e-05, -1.1399e-05, -1.1802e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 217.63, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.5565 re_mapping 0.0076 re_causal 0.0221 /// teacc 98.97 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1168, 0.1532, -0.0293, ..., -0.0078, 0.0731, 0.0074], + [ 0.1168, -0.0884, 0.0270, ..., -0.0610, -0.0104, -0.0539], + [-0.0517, -0.1273, 0.0063, ..., -0.0386, 0.0448, -0.1414], + ..., + [-0.1268, -0.1233, 0.0550, ..., -0.0550, -0.0684, -0.0997], + [-0.0617, -0.0039, 0.0213, ..., 0.0265, -0.0887, 0.0287], + [-0.0523, -0.0183, 0.0777, ..., 0.1375, -0.0459, -0.1183]], + device='cuda:0'), grad: tensor([[ 5.5879e-08, 6.4075e-06, 3.2708e-06, ..., 5.5134e-06, + 0.0000e+00, 7.0781e-08], + [-1.6969e-06, 1.6205e-07, -1.5516e-06, ..., 9.6858e-08, + 0.0000e+00, 5.2154e-08], + [ 2.6450e-07, 6.4075e-07, 2.2911e-06, ..., 3.1851e-07, + 0.0000e+00, 1.3188e-06], + ..., + [ 8.8476e-07, 4.1723e-07, 7.5810e-07, ..., 1.2852e-07, + 0.0000e+00, 4.4703e-08], + [ 1.8813e-07, 5.2340e-06, 5.7742e-07, ..., 3.4366e-06, + 0.0000e+00, -1.4305e-06], + [ 1.1548e-07, -1.8552e-05, -8.7544e-06, ..., -1.2860e-05, + 0.0000e+00, 1.8254e-07]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0322, 0.0081, 0.0205, 0.0217, 0.0316, 0.0015, -0.0019, 0.0029, + 0.0003, -0.0397], device='cuda:0'), grad: tensor([ 1.4089e-05, -2.5872e-06, 5.4091e-06, 7.2271e-06, 9.3132e-07, + 1.9260e-06, -1.1176e-08, 1.6969e-06, 7.6666e-06, -3.6299e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 217.86, cls_loss 0.0023 cls_loss_mapping 0.0035 cls_loss_causal 0.5531 re_mapping 0.0075 re_causal 0.0226 /// teacc 98.87 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1174, 0.1538, -0.0293, ..., -0.0078, 0.0728, 0.0072], + [ 0.1170, -0.0880, 0.0272, ..., -0.0610, -0.0098, -0.0537], + [-0.0516, -0.1285, 0.0061, ..., -0.0388, 0.0448, -0.1421], + ..., + [-0.1269, -0.1240, 0.0551, ..., -0.0553, -0.0685, -0.0998], + [-0.0620, -0.0041, 0.0212, ..., 0.0263, -0.0889, 0.0289], + [-0.0521, -0.0180, 0.0779, ..., 0.1384, -0.0460, -0.1180]], + device='cuda:0'), grad: tensor([[ 2.0862e-07, -4.0904e-06, -1.3206e-06, ..., 1.3411e-07, + 5.5879e-09, 3.7253e-09], + [-1.5460e-06, 2.2352e-08, 3.1404e-06, ..., 2.0526e-06, + -1.3225e-07, 3.7253e-09], + [ 1.1846e-06, 3.6322e-07, 2.6412e-06, ..., 6.1467e-08, + 2.6077e-08, 3.1665e-08], + ..., + [ 1.5106e-06, 7.2643e-08, 5.3793e-06, ..., 9.3877e-07, + 1.1176e-08, 9.3132e-09], + [ 2.6077e-07, 1.8626e-07, 9.1270e-07, ..., 1.4529e-07, + 9.3132e-09, -8.0094e-08], + [-3.1907e-06, 2.9337e-06, -1.9014e-05, ..., -5.1484e-06, + 3.7253e-09, 4.8429e-08]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0324, 0.0080, 0.0205, 0.0216, 0.0315, 0.0016, -0.0022, 0.0032, + 0.0001, -0.0396], device='cuda:0'), grad: tensor([-5.6624e-06, 7.4208e-06, 4.1276e-06, 2.2296e-06, 9.8348e-06, + 1.8161e-06, 8.6799e-07, 9.5963e-06, 2.0452e-06, -3.2306e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 217.94, cls_loss 0.0022 cls_loss_mapping 0.0038 cls_loss_causal 0.5240 re_mapping 0.0072 re_causal 0.0213 /// teacc 98.94 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1178, 0.1541, -0.0293, ..., -0.0079, 0.0728, 0.0069], + [ 0.1185, -0.0885, 0.0280, ..., -0.0614, -0.0097, -0.0542], + [-0.0525, -0.1295, 0.0056, ..., -0.0390, 0.0448, -0.1428], + ..., + [-0.1286, -0.1256, 0.0544, ..., -0.0556, -0.0683, -0.1005], + [-0.0616, -0.0035, 0.0215, ..., 0.0266, -0.0890, 0.0297], + [-0.0522, -0.0182, 0.0781, ..., 0.1391, -0.0460, -0.1187]], + device='cuda:0'), grad: tensor([[ 3.2037e-07, 4.3400e-07, 3.7253e-07, ..., 3.7253e-09, + 9.3132e-09, 5.9977e-07], + [-2.5313e-06, 5.2154e-08, -4.6715e-06, ..., 5.5879e-09, + -1.6019e-07, 6.3330e-08], + [ 7.0035e-07, 7.1339e-07, 2.3134e-06, ..., 5.5879e-09, + 1.6764e-08, 4.3958e-07], + ..., + [ 7.7300e-07, 3.3528e-08, 1.1008e-06, ..., 2.2352e-08, + 8.3819e-08, 2.6077e-08], + [-7.6294e-06, -5.3942e-05, -1.5840e-05, ..., 2.0489e-08, + 1.1176e-08, -3.0085e-05], + [ 3.2969e-07, 7.4133e-07, -4.3027e-07, ..., -4.6194e-07, + 2.4214e-08, 3.5018e-07]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0325, 0.0086, 0.0194, 0.0216, 0.0315, 0.0016, -0.0023, 0.0031, + 0.0005, -0.0397], device='cuda:0'), grad: tensor([ 1.6820e-06, -6.5416e-06, 1.8701e-06, -7.8976e-07, 1.5721e-06, + 2.9579e-05, 5.8711e-05, 1.8161e-06, -8.8334e-05, 4.2655e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 217.87, cls_loss 0.0035 cls_loss_mapping 0.0050 cls_loss_causal 0.5304 re_mapping 0.0071 re_causal 0.0214 /// teacc 98.92 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1180, 0.1542, -0.0295, ..., -0.0082, 0.0723, 0.0058], + [ 0.1181, -0.0887, 0.0267, ..., -0.0605, -0.0118, -0.0546], + [-0.0529, -0.1300, 0.0053, ..., -0.0391, 0.0447, -0.1432], + ..., + [-0.1283, -0.1263, 0.0560, ..., -0.0573, -0.0679, -0.1006], + [-0.0619, -0.0040, 0.0215, ..., 0.0264, -0.0888, 0.0299], + [-0.0537, -0.0183, 0.0781, ..., 0.1391, -0.0461, -0.1191]], + device='cuda:0'), grad: tensor([[-1.6950e-07, -3.0156e-06, 1.4696e-06, ..., 2.8126e-07, + 1.0617e-07, 3.7812e-07], + [-1.1623e-05, 3.7812e-07, -3.0458e-05, ..., 1.6019e-07, + -1.5516e-06, 3.1106e-07], + [ 3.1088e-06, 2.2147e-06, 1.1899e-05, ..., 9.1083e-07, + 1.1753e-06, 2.1495e-06], + ..., + [ 7.1079e-06, 1.2629e-06, 1.6734e-05, ..., 6.9104e-07, + -3.1292e-07, 1.1269e-06], + [-2.1961e-06, -1.0751e-05, -4.8429e-06, ..., -3.9712e-06, + 7.2643e-08, -1.1511e-05], + [ 3.8743e-07, 1.3188e-06, -1.5832e-06, ..., -5.0291e-08, + 1.7881e-07, 7.1898e-07]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0324, 0.0068, 0.0195, 0.0216, 0.0319, 0.0016, -0.0016, 0.0048, + 0.0002, -0.0400], device='cuda:0'), grad: tensor([-6.0722e-07, -5.4926e-05, 2.7329e-05, 1.5825e-05, 4.6417e-06, + 3.8370e-06, 2.0973e-06, 3.0696e-05, -2.7776e-05, -1.1567e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 217.93, cls_loss 0.0026 cls_loss_mapping 0.0037 cls_loss_causal 0.5390 re_mapping 0.0071 re_causal 0.0217 /// teacc 98.92 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1184, 0.1546, -0.0296, ..., -0.0085, 0.0721, 0.0056], + [ 0.1188, -0.0883, 0.0267, ..., -0.0605, -0.0114, -0.0548], + [-0.0533, -0.1307, 0.0051, ..., -0.0389, 0.0448, -0.1434], + ..., + [-0.1286, -0.1281, 0.0563, ..., -0.0576, -0.0681, -0.1003], + [-0.0621, -0.0037, 0.0216, ..., 0.0265, -0.0889, 0.0306], + [-0.0548, -0.0184, 0.0779, ..., 0.1396, -0.0471, -0.1210]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.1735e-07, 9.1270e-08, ..., 4.4703e-08, + 0.0000e+00, 1.0431e-07], + [-1.4342e-07, 8.5682e-08, 1.7881e-07, ..., 5.9605e-08, + 0.0000e+00, 1.7695e-07], + [ 4.4703e-08, 1.2778e-06, 2.1905e-06, ..., 1.0263e-06, + 0.0000e+00, 9.7603e-07], + ..., + [ 8.9407e-08, 6.3330e-08, -6.7987e-07, ..., 1.0617e-07, + 0.0000e+00, 9.4995e-08], + [ 1.0431e-07, -4.6790e-05, -6.8367e-05, ..., -4.1008e-05, + 0.0000e+00, -2.3976e-05], + [ 4.2841e-08, 4.4852e-05, 6.5327e-05, ..., 3.9309e-05, + 0.0000e+00, 2.2992e-05]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0324, 0.0070, 0.0196, 0.0215, 0.0321, 0.0014, -0.0017, 0.0049, + 0.0003, -0.0406], device='cuda:0'), grad: tensor([ 5.4948e-07, 4.1723e-07, 3.5875e-06, -3.1516e-06, 9.7603e-07, + 2.3916e-06, -1.2163e-06, -1.0617e-06, -9.5963e-05, 9.3520e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 218.19, cls_loss 0.0038 cls_loss_mapping 0.0051 cls_loss_causal 0.5352 re_mapping 0.0073 re_causal 0.0200 /// teacc 98.88 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1184, 0.1556, -0.0294, ..., -0.0081, 0.0720, 0.0052], + [ 0.1198, -0.0881, 0.0269, ..., -0.0624, -0.0115, -0.0548], + [-0.0546, -0.1324, 0.0029, ..., -0.0392, 0.0453, -0.1438], + ..., + [-0.1289, -0.1297, 0.0564, ..., -0.0603, -0.0690, -0.1008], + [-0.0623, -0.0032, 0.0215, ..., 0.0269, -0.0890, 0.0314], + [-0.0552, -0.0204, 0.0788, ..., 0.1417, -0.0462, -0.1226]], + device='cuda:0'), grad: tensor([[ 2.4773e-07, -1.1828e-06, -8.9407e-08, ..., -3.1665e-08, + 1.6764e-08, 2.0675e-07], + [-2.4643e-06, 7.8231e-08, -2.2147e-06, ..., 3.7253e-09, + 1.3039e-08, 8.9407e-08], + [ 6.4820e-07, 1.5870e-06, 3.4012e-06, ..., 6.3889e-07, + -1.4901e-08, 1.7732e-06], + ..., + [ 1.3970e-06, 1.7881e-07, -2.4661e-06, ..., 2.7940e-08, + 1.8626e-08, 1.1735e-07], + [ 3.3490e-06, 2.1569e-06, -2.7865e-06, ..., -7.5065e-07, + 5.1782e-07, 8.9221e-07], + [ 1.1213e-06, 2.2687e-06, 2.1961e-06, ..., 5.4017e-08, + 2.0489e-07, 9.2201e-07]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0330, 0.0075, 0.0178, 0.0216, 0.0324, 0.0014, -0.0019, 0.0050, + 0.0002, -0.0405], device='cuda:0'), grad: tensor([ 1.8030e-05, 4.6417e-06, 1.5408e-05, 4.1537e-06, -6.0022e-05, + -1.0766e-05, 1.4126e-05, -5.5172e-06, 1.8943e-06, 1.8120e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 217.82, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5142 re_mapping 0.0071 re_causal 0.0212 /// teacc 98.71 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1185, 0.1560, -0.0294, ..., -0.0081, 0.0720, 0.0052], + [ 0.1201, -0.0894, 0.0272, ..., -0.0631, -0.0143, -0.0550], + [-0.0551, -0.1327, 0.0027, ..., -0.0393, 0.0456, -0.1439], + ..., + [-0.1301, -0.1311, 0.0558, ..., -0.0621, -0.0694, -0.1011], + [-0.0628, -0.0033, 0.0214, ..., 0.0269, -0.0894, 0.0314], + [-0.0553, -0.0208, 0.0792, ..., 0.1427, -0.0460, -0.1230]], + device='cuda:0'), grad: tensor([[ 5.4203e-07, -3.2783e-07, 7.0408e-07, ..., 1.8626e-09, + 1.3225e-07, 1.2666e-07], + [-4.8801e-06, -3.3788e-06, -7.3463e-06, ..., 0.0000e+00, + -8.4750e-07, 2.6822e-07], + [ 1.4398e-06, 3.3155e-07, 1.7248e-06, ..., 7.4506e-09, + 5.9605e-08, 2.6263e-07], + ..., + [ 9.2201e-07, 5.2899e-07, -5.3160e-06, ..., 1.3039e-08, + 1.1362e-07, 2.8871e-07], + [ 4.3772e-06, 5.4613e-06, 3.4608e-06, ..., -1.4901e-08, + 9.2760e-07, 2.2221e-06], + [ 1.4193e-06, 1.6466e-06, 2.3190e-06, ..., -6.1467e-08, + 4.2841e-08, 8.4192e-07]], device='cuda:0') +Epoch 169, bias, value: tensor([ 3.3096e-02, 7.4427e-03, 1.7721e-02, 2.1634e-02, 3.2426e-02, + 1.4159e-03, -3.7978e-04, 4.3448e-03, -8.1988e-05, -4.0342e-02], + device='cuda:0'), grad: tensor([ 1.9055e-06, -9.7603e-06, 4.9360e-06, -7.0967e-07, 1.5423e-06, + -6.8881e-06, 4.8801e-06, -1.7136e-05, 1.1578e-05, 9.6112e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 217.41, cls_loss 0.0032 cls_loss_mapping 0.0043 cls_loss_causal 0.5451 re_mapping 0.0068 re_causal 0.0202 /// teacc 98.89 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1199, 0.1562, -0.0295, ..., -0.0082, 0.0718, 0.0036], + [ 0.1204, -0.0902, 0.0272, ..., -0.0631, -0.0145, -0.0554], + [-0.0556, -0.1345, 0.0022, ..., -0.0393, 0.0459, -0.1449], + ..., + [-0.1303, -0.1309, 0.0549, ..., -0.0623, -0.0687, -0.1013], + [-0.0628, -0.0026, 0.0212, ..., 0.0270, -0.0895, 0.0321], + [-0.0558, -0.0215, 0.0814, ..., 0.1436, -0.0465, -0.1238]], + device='cuda:0'), grad: tensor([[ 5.1409e-07, -1.7621e-06, 8.7544e-08, ..., -6.5193e-08, + 0.0000e+00, 4.8615e-07], + [ 2.2411e-05, 3.0309e-05, 4.5747e-05, ..., 2.0489e-08, + 0.0000e+00, 3.1203e-05], + [ 4.0233e-06, 6.2771e-07, 3.8967e-06, ..., 9.3132e-09, + 0.0000e+00, 5.6997e-07], + ..., + [ 1.9949e-06, 2.1793e-06, -2.6971e-06, ..., 8.5123e-07, + 0.0000e+00, 2.1793e-06], + [-3.2544e-05, -3.7462e-05, -6.2227e-05, ..., 1.3039e-08, + 0.0000e+00, -3.9488e-05], + [ 2.2445e-06, 3.1777e-06, -1.9055e-06, ..., -1.0896e-06, + 0.0000e+00, 1.9632e-06]], device='cuda:0') +Epoch 170, bias, value: tensor([ 3.2900e-02, 7.3967e-03, 1.7533e-02, 2.2091e-02, 3.2841e-02, + 1.0310e-03, -4.9539e-04, 3.4851e-03, -6.0871e-05, -3.9208e-02], + device='cuda:0'), grad: tensor([-1.7304e-06, 8.3148e-05, 6.5453e-06, 5.5820e-05, 1.2033e-06, + -3.9339e-05, 5.7966e-06, -6.7651e-06, -1.0735e-04, 2.6934e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 218.06, cls_loss 0.0024 cls_loss_mapping 0.0041 cls_loss_causal 0.5198 re_mapping 0.0070 re_causal 0.0206 /// teacc 98.92 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1202, 0.1574, -0.0291, ..., -0.0070, 0.0718, 0.0028], + [ 0.1210, -0.0907, 0.0274, ..., -0.0632, -0.0145, -0.0556], + [-0.0558, -0.1345, 0.0022, ..., -0.0393, 0.0459, -0.1452], + ..., + [-0.1308, -0.1322, 0.0548, ..., -0.0625, -0.0687, -0.1015], + [-0.0626, -0.0019, 0.0214, ..., 0.0268, -0.0895, 0.0331], + [-0.0559, -0.0229, 0.0814, ..., 0.1433, -0.0465, -0.1245]], + device='cuda:0'), grad: tensor([[ 9.1270e-08, 7.4506e-08, 1.6205e-07, ..., 5.5879e-09, + 3.7253e-09, 3.9116e-08], + [-8.0615e-06, -1.3970e-07, -1.3173e-05, ..., 3.7253e-09, + -2.7940e-08, 4.6566e-08], + [ 6.8732e-07, -1.3970e-07, 1.9353e-06, ..., 4.6566e-08, + 1.8626e-09, 1.6578e-07], + ..., + [ 7.0147e-06, 8.0094e-08, 9.1642e-06, ..., 2.0489e-08, + 0.0000e+00, 5.5879e-08], + [ 5.3085e-07, 1.8999e-07, 1.0226e-06, ..., -1.8626e-08, + 9.3132e-09, 9.3505e-07], + [ 1.1269e-06, 7.1898e-07, -7.3574e-07, ..., -1.6019e-07, + 3.7253e-09, 7.7486e-07]], device='cuda:0') +Epoch 171, bias, value: tensor([ 3.3400e-02, 7.4328e-03, 1.7879e-02, 2.2054e-02, 3.3023e-02, + 6.0021e-04, -5.6159e-06, 3.3089e-03, -3.2905e-05, -3.9564e-02], + device='cuda:0'), grad: tensor([ 8.1211e-07, -1.6898e-05, 2.1048e-06, -1.6138e-05, 1.0058e-06, + 1.0170e-05, 3.2261e-06, 9.7752e-06, 4.6119e-06, 1.3374e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 217.88, cls_loss 0.0026 cls_loss_mapping 0.0038 cls_loss_causal 0.5146 re_mapping 0.0072 re_causal 0.0207 /// teacc 98.92 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1205, 0.1577, -0.0291, ..., -0.0069, 0.0718, 0.0024], + [ 0.1210, -0.0906, 0.0277, ..., -0.0628, -0.0145, -0.0558], + [-0.0547, -0.1345, 0.0022, ..., -0.0392, 0.0459, -0.1457], + ..., + [-0.1311, -0.1334, 0.0550, ..., -0.0627, -0.0686, -0.1016], + [-0.0635, -0.0021, 0.0213, ..., 0.0267, -0.0895, 0.0330], + [-0.0574, -0.0236, 0.0810, ..., 0.1417, -0.0466, -0.1261]], + device='cuda:0'), grad: tensor([[ 1.3784e-07, -3.0851e-04, -7.4878e-07, ..., 1.8626e-09, + 0.0000e+00, -5.7966e-05], + [ 2.4214e-08, 9.4995e-08, -4.1164e-07, ..., 1.8626e-09, + 0.0000e+00, 5.4389e-07], + [ 4.0419e-07, 4.2655e-07, 2.6636e-07, ..., 1.8626e-09, + 0.0000e+00, 2.6636e-07], + ..., + [ 4.3400e-07, 1.0058e-07, -1.0598e-06, ..., 3.7253e-09, + 0.0000e+00, 5.4017e-08], + [ 1.6540e-06, 1.4119e-06, 4.0233e-07, ..., 3.7253e-09, + 0.0000e+00, 1.5162e-06], + [ 4.7684e-06, 1.8533e-06, -3.8892e-06, ..., -1.6764e-07, + 0.0000e+00, 2.8573e-06]], device='cuda:0') +Epoch 172, bias, value: tensor([ 3.3322e-02, 7.3815e-03, 1.8770e-02, 2.2114e-02, 3.4126e-02, + 6.7332e-04, 1.4388e-05, 3.4360e-03, -4.4353e-04, -4.0905e-02], + device='cuda:0'), grad: tensor([-3.6716e-04, 6.0536e-07, 1.4529e-06, -2.6464e-05, 5.9679e-06, + 3.3647e-05, 3.4833e-04, -1.4678e-06, 5.0962e-06, 1.1735e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 217.31, cls_loss 0.0027 cls_loss_mapping 0.0043 cls_loss_causal 0.5391 re_mapping 0.0070 re_causal 0.0208 /// teacc 98.85 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1203, 0.1585, -0.0297, ..., -0.0075, 0.0718, 0.0025], + [ 0.1214, -0.0898, 0.0280, ..., -0.0629, -0.0144, -0.0559], + [-0.0541, -0.1352, 0.0024, ..., -0.0393, 0.0462, -0.1463], + ..., + [-0.1320, -0.1357, 0.0550, ..., -0.0630, -0.0688, -0.1039], + [-0.0642, -0.0021, 0.0212, ..., 0.0267, -0.0896, 0.0331], + [-0.0586, -0.0240, 0.0815, ..., 0.1424, -0.0466, -0.1271]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -3.1479e-07, -1.4529e-07, ..., 5.5879e-09, + 0.0000e+00, 4.0978e-08], + [-5.0478e-07, 4.8429e-08, -6.2957e-07, ..., 7.4506e-09, + 0.0000e+00, 1.7136e-07], + [ 1.1176e-07, 1.2293e-07, 5.2340e-07, ..., 1.8626e-09, + 0.0000e+00, 5.0105e-07], + ..., + [ 1.0245e-07, 1.8626e-08, 1.7397e-06, ..., 2.0489e-08, + 0.0000e+00, 3.3639e-06], + [ 1.8105e-06, 1.7621e-06, -7.4506e-09, ..., 1.2852e-07, + 0.0000e+00, 1.8533e-06], + [ 1.5832e-07, 1.6764e-08, -3.6135e-07, ..., -3.3155e-07, + 0.0000e+00, 6.6310e-07]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0333, 0.0075, 0.0196, 0.0215, 0.0340, 0.0019, -0.0009, 0.0029, + -0.0006, -0.0407], device='cuda:0'), grad: tensor([ 3.5949e-07, -3.6880e-07, -4.9591e-05, -1.4901e-05, 4.4614e-05, + -1.8626e-06, 5.1521e-06, 1.0714e-05, 4.6268e-06, 1.2573e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 217.25, cls_loss 0.0033 cls_loss_mapping 0.0052 cls_loss_causal 0.5655 re_mapping 0.0071 re_causal 0.0207 /// teacc 99.02 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1244, 0.1580, -0.0301, ..., -0.0081, 0.0719, 0.0016], + [ 0.1220, -0.0905, 0.0281, ..., -0.0630, -0.0144, -0.0562], + [-0.0551, -0.1360, 0.0027, ..., -0.0395, 0.0462, -0.1469], + ..., + [-0.1322, -0.1370, 0.0550, ..., -0.0633, -0.0688, -0.1042], + [-0.0646, -0.0018, 0.0212, ..., 0.0263, -0.0897, 0.0337], + [-0.0582, -0.0230, 0.0817, ..., 0.1435, -0.0466, -0.1273]], + device='cuda:0'), grad: tensor([[ 1.7695e-07, -3.9302e-07, 6.3516e-07, ..., 7.3016e-07, + 5.4017e-08, 1.1362e-07], + [-2.1663e-06, -1.8626e-09, -7.9535e-07, ..., 5.0291e-08, + 3.7812e-07, 1.2480e-07], + [ 4.0233e-07, 3.5390e-07, 1.4618e-05, ..., 6.1467e-08, + 3.0454e-06, 9.4436e-07], + ..., + [ 4.2841e-07, 1.1548e-07, -1.8433e-05, ..., 6.8918e-08, + -4.0904e-06, 6.8918e-08], + [ 7.7859e-07, 3.3695e-06, 3.8967e-06, ..., 3.2764e-06, + 2.0862e-07, 7.5437e-07], + [ 1.1705e-05, 4.3772e-06, -4.0680e-06, ..., -6.0350e-06, + 1.5832e-07, 4.4182e-06]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0323, 0.0076, 0.0199, 0.0213, 0.0344, 0.0023, -0.0007, 0.0029, + -0.0009, -0.0408], device='cuda:0'), grad: tensor([ 2.2911e-07, 2.4028e-07, 3.2634e-05, 1.5274e-06, 1.3355e-06, + -1.6823e-05, 2.0768e-06, -3.7849e-05, 1.0334e-05, 6.2212e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 218.27, cls_loss 0.0026 cls_loss_mapping 0.0040 cls_loss_causal 0.5168 re_mapping 0.0070 re_causal 0.0208 /// teacc 98.90 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1246, 0.1581, -0.0302, ..., -0.0082, 0.0719, 0.0014], + [ 0.1224, -0.0909, 0.0285, ..., -0.0608, -0.0144, -0.0568], + [-0.0545, -0.1374, 0.0021, ..., -0.0396, 0.0461, -0.1479], + ..., + [-0.1330, -0.1382, 0.0550, ..., -0.0645, -0.0688, -0.1048], + [-0.0658, -0.0022, 0.0211, ..., 0.0259, -0.0897, 0.0332], + [-0.0586, -0.0228, 0.0821, ..., 0.1450, -0.0466, -0.1274]], + device='cuda:0'), grad: tensor([[ 8.1025e-07, -7.1898e-06, -1.7956e-06, ..., 2.2352e-08, + 0.0000e+00, 8.9593e-07], + [ 1.3039e-08, 2.8871e-07, -6.8918e-08, ..., 2.5406e-06, + 0.0000e+00, 1.7695e-07], + [ 5.2154e-08, 5.8860e-07, 3.6322e-07, ..., 2.6077e-08, + -3.7253e-09, 5.3272e-07], + ..., + [ 1.1548e-07, 7.0781e-08, 1.6764e-07, ..., 3.4831e-07, + 3.7253e-09, 6.8918e-08], + [ 2.4825e-05, 6.4909e-05, -2.8498e-07, ..., 5.7742e-08, + 0.0000e+00, 3.7819e-05], + [ 4.6194e-07, 8.9705e-06, 1.1567e-06, ..., 1.3281e-06, + 0.0000e+00, 6.7428e-07]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0322, 0.0077, 0.0199, 0.0212, 0.0338, 0.0027, -0.0008, 0.0029, + -0.0013, -0.0405], device='cuda:0'), grad: tensor([-6.3702e-06, 3.2723e-05, 1.8105e-06, 2.0582e-06, -5.5611e-05, + 4.3631e-05, -1.7476e-04, 4.4405e-06, 1.2130e-04, 3.0696e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 217.50, cls_loss 0.0025 cls_loss_mapping 0.0034 cls_loss_causal 0.5629 re_mapping 0.0068 re_causal 0.0207 /// teacc 98.80 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1248, 0.1587, -0.0301, ..., -0.0082, 0.0721, 0.0008], + [ 0.1229, -0.0909, 0.0283, ..., -0.0607, -0.0144, -0.0575], + [-0.0554, -0.1391, 0.0017, ..., -0.0396, 0.0462, -0.1500], + ..., + [-0.1330, -0.1391, 0.0554, ..., -0.0646, -0.0689, -0.1046], + [-0.0660, -0.0020, 0.0214, ..., 0.0259, -0.0897, 0.0339], + [-0.0596, -0.0235, 0.0823, ..., 0.1450, -0.0466, -0.1288]], + device='cuda:0'), grad: tensor([[ 3.5428e-06, -1.8626e-08, 4.8243e-06, ..., 1.7546e-06, + 4.2468e-07, 0.0000e+00], + [-2.5105e-04, 3.7253e-09, -6.4433e-05, ..., 3.3528e-08, + -2.6032e-05, 0.0000e+00], + [ 1.3077e-04, 3.7253e-08, 2.2486e-05, ..., 2.6077e-07, + 1.6123e-05, 1.3970e-07], + ..., + [ 1.0312e-04, -2.6077e-08, 3.8743e-05, ..., 1.8813e-07, + 8.0466e-06, 1.8626e-09], + [ 2.4550e-06, -1.5646e-07, 1.9073e-06, ..., 1.0058e-07, + 1.8068e-07, -4.7684e-07], + [ 1.1381e-06, 2.2352e-08, -1.9148e-05, ..., -6.3628e-06, + 1.1176e-07, 5.5879e-09]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0324, 0.0076, 0.0191, 0.0214, 0.0355, 0.0029, -0.0007, 0.0033, + -0.0013, -0.0422], device='cuda:0'), grad: tensor([ 1.6555e-05, -4.7183e-04, 2.6155e-04, 1.1437e-05, 2.4468e-05, + 3.4869e-06, 5.2117e-06, 1.7774e-04, 5.7444e-06, -3.4571e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 217.87, cls_loss 0.0034 cls_loss_mapping 0.0032 cls_loss_causal 0.5059 re_mapping 0.0069 re_causal 0.0198 /// teacc 98.91 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1277, 0.1575, -0.0301, ..., -0.0085, 0.0722, 0.0004], + [ 0.1237, -0.0909, 0.0283, ..., -0.0607, -0.0143, -0.0578], + [-0.0557, -0.1396, 0.0017, ..., -0.0397, 0.0462, -0.1506], + ..., + [-0.1351, -0.1411, 0.0530, ..., -0.0647, -0.0691, -0.1062], + [-0.0663, -0.0014, 0.0226, ..., 0.0259, -0.0898, 0.0362], + [-0.0568, -0.0243, 0.0850, ..., 0.1453, -0.0467, -0.1316]], + device='cuda:0'), grad: tensor([[-1.3132e-06, -1.3493e-05, -1.1120e-06, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-1.8738e-06, 8.9407e-08, -1.8179e-06, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 3.9116e-07, 4.1574e-06, 1.7304e-06, ..., 1.8626e-09, + 0.0000e+00, 1.6391e-07], + ..., + [ 7.1153e-07, 4.0978e-08, -4.0382e-06, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 1.2852e-07, 9.5740e-07, 2.4773e-07, ..., -1.8626e-09, + 0.0000e+00, -1.8440e-07], + [ 1.3784e-07, 1.0915e-06, 2.7642e-06, ..., -1.8626e-09, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0307, 0.0077, 0.0193, 0.0212, 0.0354, 0.0032, 0.0006, 0.0011, + -0.0005, -0.0403], device='cuda:0'), grad: tensor([-2.9653e-05, -2.3544e-06, 1.1727e-05, 3.5074e-06, 9.3691e-07, + 2.4904e-06, 1.1094e-05, -9.4771e-06, 3.3211e-06, 8.2999e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 217.76, cls_loss 0.0024 cls_loss_mapping 0.0041 cls_loss_causal 0.5442 re_mapping 0.0075 re_causal 0.0214 /// teacc 98.92 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1278, 0.1582, -0.0299, ..., -0.0086, 0.0723, 0.0014], + [ 0.1244, -0.0913, 0.0290, ..., -0.0613, -0.0143, -0.0584], + [-0.0558, -0.1409, 0.0013, ..., -0.0398, 0.0461, -0.1517], + ..., + [-0.1358, -0.1420, 0.0526, ..., -0.0647, -0.0691, -0.1075], + [-0.0661, 0.0003, 0.0233, ..., 0.0258, -0.0899, 0.0386], + [-0.0568, -0.0245, 0.0851, ..., 0.1462, -0.0467, -0.1327]], + device='cuda:0'), grad: tensor([[ 2.3618e-06, 9.0525e-07, 8.0187e-07, ..., 1.3001e-06, + 0.0000e+00, 2.7195e-07], + [-1.7509e-07, 1.0990e-07, 4.7162e-06, ..., 6.7987e-08, + 0.0000e+00, 1.6857e-07], + [ 1.9027e-06, 1.3337e-06, 1.0580e-04, ..., 1.0664e-06, + 0.0000e+00, 1.8906e-07], + ..., + [ 4.6939e-07, 1.5181e-07, -1.1075e-04, ..., 1.0896e-07, + 0.0000e+00, 1.3225e-07], + [ 1.8597e-05, 3.5372e-06, 6.7148e-07, ..., 4.6659e-07, + 0.0000e+00, 1.6987e-05], + [-8.9705e-06, -7.0594e-06, -2.5317e-05, ..., -1.3985e-05, + 0.0000e+00, 6.5565e-07]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0309, 0.0081, 0.0194, 0.0211, 0.0350, 0.0033, -0.0002, 0.0008, + 0.0003, -0.0401], device='cuda:0'), grad: tensor([ 5.4948e-06, 2.2829e-05, 2.2101e-04, 3.7223e-05, 6.5230e-06, + -1.4424e-04, 8.7321e-05, -2.2411e-04, 3.3855e-05, -4.6313e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 217.36, cls_loss 0.0022 cls_loss_mapping 0.0042 cls_loss_causal 0.5519 re_mapping 0.0068 re_causal 0.0204 /// teacc 98.87 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1274, 0.1594, -0.0296, ..., -0.0087, 0.0723, 0.0027], + [ 0.1249, -0.0918, 0.0292, ..., -0.0612, -0.0143, -0.0592], + [-0.0562, -0.1424, 0.0011, ..., -0.0399, 0.0463, -0.1518], + ..., + [-0.1362, -0.1434, 0.0526, ..., -0.0653, -0.0694, -0.1079], + [-0.0662, 0.0007, 0.0234, ..., 0.0257, -0.0901, 0.0388], + [-0.0568, -0.0250, 0.0851, ..., 0.1466, -0.0467, -0.1335]], + device='cuda:0'), grad: tensor([[-6.6124e-08, -1.7965e-06, -6.8732e-07, ..., 2.3283e-08, + 0.0000e+00, 1.6391e-07], + [ 3.0734e-07, 4.2692e-06, 7.5772e-06, ..., 6.1467e-08, + 0.0000e+00, 5.3048e-06], + [-4.4145e-07, 4.9733e-07, 8.7731e-07, ..., 1.0896e-07, + 0.0000e+00, 6.6590e-07], + ..., + [ 1.4156e-07, 5.5507e-07, 6.4541e-07, ..., 9.0338e-08, + 0.0000e+00, 1.0813e-06], + [-9.4436e-07, -1.2070e-05, -1.7881e-05, ..., -5.0515e-06, + 0.0000e+00, -8.2552e-06], + [ 9.1549e-07, 7.0222e-06, 6.9961e-06, ..., 4.1425e-06, + 0.0000e+00, 3.1814e-06]], device='cuda:0') +Epoch 179, bias, value: tensor([ 3.1254e-02, 8.2139e-03, 1.9783e-02, 2.1321e-02, 3.5509e-02, + 2.9895e-03, -4.0581e-04, 7.7312e-04, 9.0251e-05, -4.0533e-02], + device='cuda:0'), grad: tensor([-2.0415e-06, 1.7539e-05, -8.3670e-06, -3.7774e-06, 2.4009e-06, + 2.6505e-06, 2.6077e-07, 3.2596e-06, -2.2382e-05, 1.0461e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.60, cls_loss 0.0031 cls_loss_mapping 0.0054 cls_loss_causal 0.5193 re_mapping 0.0071 re_causal 0.0196 /// teacc 98.83 lr 0.00010000 +Epoch 180, weight, value: tensor([[-1.2775e-01, 1.6028e-01, -2.9261e-02, ..., -8.6730e-03, + 7.2672e-02, 2.6627e-03], + [ 1.2515e-01, -9.2059e-02, 2.8988e-02, ..., -6.2197e-02, + -1.4233e-02, -6.0048e-02], + [-5.6443e-02, -1.4396e-01, 1.2706e-04, ..., -3.9997e-02, + 4.6035e-02, -1.5265e-01], + ..., + [-1.3626e-01, -1.4428e-01, 5.2864e-02, ..., -6.6150e-02, + -6.9440e-02, -1.0752e-01], + [-6.6836e-02, 1.1181e-03, 2.3345e-02, ..., 2.5704e-02, + -9.0155e-02, 3.8912e-02], + [-5.6728e-02, -2.6038e-02, 8.5428e-02, ..., 1.4747e-01, + -4.7039e-02, -1.3453e-01]], device='cuda:0'), grad: tensor([[ 8.8476e-08, -8.0541e-06, -2.0415e-06, ..., 6.8918e-08, + 4.6566e-09, -2.3991e-06], + [ 5.7459e-05, 5.7090e-07, 1.6749e-04, ..., 4.8518e-05, + 9.3132e-10, 1.9092e-07], + [ 2.4121e-07, 3.5483e-07, 1.4976e-06, ..., 2.0117e-07, + 0.0000e+00, 1.6484e-07], + ..., + [ 1.6801e-06, 6.3330e-08, -3.3416e-06, ..., 1.3132e-06, + 0.0000e+00, 4.4703e-08], + [ 3.5781e-06, 4.6417e-06, 1.2591e-05, ..., 2.9765e-06, + 1.5926e-07, 1.5525e-06], + [-6.5744e-05, 6.7614e-07, -1.8573e-04, ..., -5.5462e-05, + 0.0000e+00, 3.0734e-08]], device='cuda:0') +Epoch 180, bias, value: tensor([ 3.1604e-02, 7.7109e-03, 1.8949e-02, 2.1374e-02, 3.5512e-02, + 2.5156e-03, -4.0657e-05, 1.4157e-03, -2.6940e-04, -4.0445e-02], + device='cuda:0'), grad: tensor([-1.3456e-05, 1.9753e-04, 2.2482e-06, 6.7428e-06, 2.2631e-07, + 9.5591e-06, -1.2711e-05, -3.4198e-06, 2.4125e-05, -2.1088e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 217.78, cls_loss 0.0034 cls_loss_mapping 0.0045 cls_loss_causal 0.5583 re_mapping 0.0067 re_causal 0.0192 /// teacc 98.88 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1279, 0.1628, -0.0285, ..., -0.0086, 0.0726, 0.0014], + [ 0.1257, -0.0940, 0.0291, ..., -0.0629, -0.0141, -0.0611], + [-0.0564, -0.1471, -0.0003, ..., -0.0403, 0.0460, -0.1545], + ..., + [-0.1366, -0.1456, 0.0530, ..., -0.0664, -0.0695, -0.1080], + [-0.0674, 0.0011, 0.0234, ..., 0.0258, -0.0903, 0.0397], + [-0.0572, -0.0263, 0.0859, ..., 0.1480, -0.0470, -0.1342]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, -4.0084e-06, -4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 7.1712e-08], + [ 2.7940e-07, 3.5297e-07, 1.8440e-07, ..., 0.0000e+00, + 0.0000e+00, 2.8871e-07], + [ 6.7055e-08, 1.7788e-07, 2.6431e-06, ..., 4.6566e-09, + 0.0000e+00, 6.9849e-08], + ..., + [ 1.0245e-07, 2.0303e-07, -4.6380e-06, ..., 0.0000e+00, + 0.0000e+00, -2.2724e-07], + [ 1.1735e-05, 7.9796e-06, 9.1270e-08, ..., -6.5193e-09, + 0.0000e+00, 7.0557e-06], + [ 4.6194e-07, 1.0263e-06, -3.8184e-08, ..., 0.0000e+00, + 0.0000e+00, 3.6228e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0333, 0.0073, 0.0185, 0.0215, 0.0340, 0.0056, -0.0036, 0.0016, + -0.0007, -0.0396], device='cuda:0'), grad: tensor([-5.0999e-06, 4.4405e-06, 4.6380e-06, 3.8743e-05, -1.8636e-06, + -5.0128e-05, 2.6785e-06, -1.1928e-05, 1.5944e-05, 2.6058e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 217.03, cls_loss 0.0021 cls_loss_mapping 0.0031 cls_loss_causal 0.5164 re_mapping 0.0070 re_causal 0.0205 /// teacc 98.93 lr 0.00010000 +Epoch 182, weight, value: tensor([[-1.2782e-01, 1.6320e-01, -2.8356e-02, ..., -8.6621e-03, + 7.2718e-02, 1.0329e-04], + [ 1.2582e-01, -9.4193e-02, 2.7411e-02, ..., -6.2927e-02, + -1.4140e-02, -6.1301e-02], + [-5.6868e-02, -1.4775e-01, 4.5387e-05, ..., -4.0272e-02, + 4.6003e-02, -1.5456e-01], + ..., + [-1.3639e-01, -1.4614e-01, 5.4162e-02, ..., -6.6421e-02, + -6.9478e-02, -1.0851e-01], + [-6.7894e-02, 1.1510e-03, 2.3325e-02, ..., 2.5731e-02, + -9.0405e-02, 4.0039e-02], + [-5.7360e-02, -2.6582e-02, 8.6097e-02, ..., 1.4870e-01, + -4.7069e-02, -1.3439e-01]], device='cuda:0'), grad: tensor([[ 9.4995e-08, -7.8976e-07, 5.4576e-07, ..., 2.7940e-09, + 0.0000e+00, 1.3411e-07], + [-2.9244e-07, 1.5553e-07, 7.4208e-06, ..., 4.6566e-09, + 0.0000e+00, 2.3749e-07], + [ 8.1956e-08, 1.5739e-07, -3.8534e-05, ..., 3.7253e-09, + 0.0000e+00, 5.5879e-08], + ..., + [ 2.1886e-07, 6.5193e-08, 1.0028e-05, ..., 1.0245e-08, + 0.0000e+00, 1.1362e-07], + [ 2.0117e-07, 3.9209e-07, 1.9610e-05, ..., 6.2399e-08, + 0.0000e+00, 1.7043e-07], + [-3.1777e-06, -2.0713e-06, -4.4793e-05, ..., -9.9018e-06, + 0.0000e+00, -4.9025e-06]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0333, 0.0063, 0.0186, 0.0214, 0.0340, 0.0063, -0.0042, 0.0022, + -0.0010, -0.0395], device='cuda:0'), grad: tensor([ 8.1435e-06, 7.6115e-05, -3.0565e-04, 2.5742e-06, 1.3626e-04, + 1.8880e-05, 3.9265e-06, 7.6294e-05, 1.3387e-04, -1.5008e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 217.67, cls_loss 0.0017 cls_loss_mapping 0.0032 cls_loss_causal 0.5506 re_mapping 0.0064 re_causal 0.0200 /// teacc 98.95 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1286, 0.1628, -0.0284, ..., -0.0087, 0.0727, 0.0002], + [ 0.1262, -0.0940, 0.0274, ..., -0.0633, -0.0141, -0.0616], + [-0.0571, -0.1484, -0.0002, ..., -0.0403, 0.0460, -0.1548], + ..., + [-0.1365, -0.1463, 0.0543, ..., -0.0665, -0.0695, -0.1085], + [-0.0683, 0.0010, 0.0233, ..., 0.0257, -0.0906, 0.0402], + [-0.0574, -0.0268, 0.0861, ..., 0.1489, -0.0471, -0.1348]], + device='cuda:0'), grad: tensor([[ 5.3085e-08, -3.3863e-06, -2.3209e-06, ..., 9.3132e-10, + 0.0000e+00, 4.1910e-08], + [-8.3260e-07, 4.3772e-07, -6.6217e-07, ..., 1.8626e-09, + 0.0000e+00, 6.4261e-08], + [ 3.2503e-07, 3.1944e-07, 9.4436e-07, ..., 0.0000e+00, + 0.0000e+00, 3.5297e-07], + ..., + [ 2.0675e-07, 5.3458e-07, 1.4855e-06, ..., 1.8626e-09, + 0.0000e+00, 7.4506e-08], + [ 1.2293e-07, -1.8626e-08, -2.4121e-07, ..., 2.2352e-08, + 0.0000e+00, -1.3290e-06], + [ 3.7253e-08, 6.5379e-07, -2.3767e-06, ..., -6.7055e-08, + 0.0000e+00, 6.7055e-08]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0325, 0.0063, 0.0187, 0.0213, 0.0341, 0.0062, -0.0036, 0.0024, + -0.0012, -0.0396], device='cuda:0'), grad: tensor([-3.8631e-06, -2.3562e-07, 4.0326e-07, 2.2911e-06, 2.4121e-07, + 8.9128e-07, 1.2135e-06, 4.1351e-06, -8.5216e-07, -4.2357e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 217.27, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.5366 re_mapping 0.0065 re_causal 0.0197 /// teacc 98.99 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1285, 0.1640, -0.0280, ..., -0.0087, 0.0726, 0.0003], + [ 0.1263, -0.0944, 0.0273, ..., -0.0629, -0.0142, -0.0629], + [-0.0572, -0.1492, -0.0006, ..., -0.0403, 0.0460, -0.1549], + ..., + [-0.1366, -0.1472, 0.0534, ..., -0.0668, -0.0695, -0.1116], + [-0.0693, 0.0004, 0.0261, ..., 0.0255, -0.0927, 0.0431], + [-0.0575, -0.0265, 0.0861, ..., 0.1492, -0.0472, -0.1349]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, -2.3991e-06, -1.0431e-06, ..., 0.0000e+00, + 0.0000e+00, 3.4459e-08], + [-2.5239e-07, 2.1327e-07, 2.6356e-07, ..., 9.3132e-09, + 0.0000e+00, 6.8918e-08], + [ 1.6298e-07, 2.3171e-06, 1.1958e-06, ..., 9.3132e-10, + 0.0000e+00, 3.1367e-06], + ..., + [ 9.6858e-08, 2.0768e-07, -4.3958e-07, ..., 3.9116e-08, + 0.0000e+00, 9.6858e-08], + [ 2.4959e-07, -2.3842e-06, -1.1008e-06, ..., 9.3132e-10, + 0.0000e+00, -4.0866e-06], + [ 3.0734e-08, 6.9663e-07, -1.3970e-08, ..., -7.9162e-08, + 0.0000e+00, 6.6124e-08]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0331, 0.0060, 0.0189, 0.0214, 0.0341, 0.0064, -0.0042, 0.0017, + 0.0010, -0.0398], device='cuda:0'), grad: tensor([-3.3937e-06, 6.6403e-07, 6.5789e-06, 4.6372e-05, 3.7067e-07, + -4.4584e-05, 9.7509e-07, -3.9395e-07, -7.3761e-06, 7.3947e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 217.64, cls_loss 0.0024 cls_loss_mapping 0.0030 cls_loss_causal 0.4930 re_mapping 0.0064 re_causal 0.0189 /// teacc 99.01 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1292, 0.1635, -0.0306, ..., -0.0087, 0.0723, -0.0006], + [ 0.1265, -0.0946, 0.0272, ..., -0.0631, -0.0141, -0.0633], + [-0.0566, -0.1514, -0.0023, ..., -0.0404, 0.0458, -0.1566], + ..., + [-0.1369, -0.1491, 0.0536, ..., -0.0672, -0.0697, -0.1116], + [-0.0695, 0.0011, 0.0264, ..., 0.0258, -0.0930, 0.0440], + [-0.0578, -0.0251, 0.0866, ..., 0.1494, -0.0473, -0.1370]], + device='cuda:0'), grad: tensor([[ 1.2666e-07, -9.8705e-05, -9.1851e-05, ..., 7.4506e-09, + 8.3819e-09, 1.9092e-07], + [-8.8848e-07, 2.3097e-07, -5.9977e-07, ..., 1.1176e-08, + -2.0489e-08, 3.1851e-07], + [ 1.7695e-07, 3.8650e-07, 6.6962e-07, ..., 2.2352e-08, + -5.6811e-08, 4.7944e-06], + ..., + [ 2.3283e-07, 9.2201e-08, 8.7451e-07, ..., 7.3574e-08, + 3.2596e-08, 1.8775e-05], + [ 2.3842e-07, 4.1723e-07, 1.5758e-06, ..., 8.2888e-08, + 8.3819e-09, 1.2284e-06], + [ 9.6858e-08, 9.6500e-05, 8.5652e-05, ..., -4.1351e-07, + 5.5879e-09, 3.4180e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0314, 0.0058, 0.0188, 0.0208, 0.0343, 0.0064, -0.0040, 0.0019, + 0.0015, -0.0393], device='cuda:0'), grad: tensor([-2.6774e-04, 2.1514e-07, 9.7975e-06, -4.3988e-05, 3.9712e-06, + 5.5581e-06, -6.4597e-06, 3.2812e-05, 6.6608e-06, 2.5845e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 217.82, cls_loss 0.0026 cls_loss_mapping 0.0039 cls_loss_causal 0.5259 re_mapping 0.0067 re_causal 0.0194 /// teacc 98.93 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1294, 0.1642, -0.0300, ..., -0.0083, 0.0723, -0.0021], + [ 0.1269, -0.0947, 0.0272, ..., -0.0632, -0.0141, -0.0636], + [-0.0568, -0.1522, -0.0027, ..., -0.0404, 0.0458, -0.1573], + ..., + [-0.1373, -0.1513, 0.0538, ..., -0.0674, -0.0698, -0.1120], + [-0.0702, 0.0006, 0.0264, ..., 0.0257, -0.0930, 0.0439], + [-0.0586, -0.0277, 0.0873, ..., 0.1517, -0.0474, -0.1379]], + device='cuda:0'), grad: tensor([[ 1.3039e-07, -8.3633e-07, 9.6858e-08, ..., 2.9802e-08, + 2.0489e-08, 3.1851e-07], + [-8.5756e-06, -1.8254e-07, -1.9133e-05, ..., 1.3039e-08, + -7.2457e-07, 2.6077e-07], + [ 5.5209e-06, 4.0047e-07, 1.3195e-05, ..., 4.0978e-08, + 5.3458e-07, 1.4175e-06], + ..., + [ 1.4585e-06, 1.7881e-07, 3.6396e-06, ..., 2.8685e-07, + 2.4028e-07, 3.0696e-06], + [ 5.6438e-07, 1.8254e-07, 1.0245e-06, ..., 1.3225e-07, + 6.8918e-08, -3.1404e-06], + [ 1.4715e-07, 3.1292e-07, -2.1793e-07, ..., 7.4506e-09, + 4.2841e-08, 9.9838e-07]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0314, 0.0056, 0.0186, 0.0211, 0.0326, 0.0070, -0.0034, 0.0020, + 0.0011, -0.0386], device='cuda:0'), grad: tensor([ 5.3830e-07, -4.3660e-05, 3.4750e-05, -1.6347e-05, -1.2424e-06, + 6.1169e-06, 1.9018e-06, 1.7822e-05, -3.7868e-06, 3.8445e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 217.79, cls_loss 0.0026 cls_loss_mapping 0.0036 cls_loss_causal 0.5278 re_mapping 0.0068 re_causal 0.0194 /// teacc 98.88 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1299, 0.1644, -0.0298, ..., -0.0083, 0.0723, -0.0010], + [ 0.1281, -0.0944, 0.0269, ..., -0.0639, -0.0141, -0.0634], + [-0.0581, -0.1532, -0.0032, ..., -0.0405, 0.0458, -0.1585], + ..., + [-0.1376, -0.1531, 0.0542, ..., -0.0674, -0.0698, -0.1119], + [-0.0707, 0.0005, 0.0265, ..., 0.0257, -0.0930, 0.0441], + [-0.0596, -0.0292, 0.0872, ..., 0.1519, -0.0474, -0.1386]], + device='cuda:0'), grad: tensor([[ 3.3528e-08, -4.6611e-05, -7.6890e-06, ..., 0.0000e+00, + 0.0000e+00, -1.0423e-05], + [-2.6170e-06, 1.8813e-07, 2.0266e-05, ..., 0.0000e+00, + 0.0000e+00, 1.5460e-07], + [ 1.3597e-07, 4.6752e-07, 1.1157e-06, ..., 0.0000e+00, + 0.0000e+00, 2.3693e-06], + ..., + [ 3.0547e-07, 8.8103e-07, -3.4869e-05, ..., 0.0000e+00, + 0.0000e+00, 6.8359e-07], + [ 1.1884e-06, 1.2651e-05, -3.1348e-06, ..., 0.0000e+00, + 0.0000e+00, -5.7742e-08], + [ 8.7544e-08, 3.1032e-06, 3.1125e-06, ..., 0.0000e+00, + 0.0000e+00, 8.4564e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0309, 0.0056, 0.0178, 0.0213, 0.0323, 0.0070, -0.0027, 0.0025, + 0.0010, -0.0389], device='cuda:0'), grad: tensor([-4.5389e-05, 3.1829e-05, 6.3293e-06, 2.0474e-05, -1.8422e-06, + 1.9714e-05, 7.6666e-06, -5.6595e-05, 7.4655e-06, 1.0356e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 217.91, cls_loss 0.0030 cls_loss_mapping 0.0052 cls_loss_causal 0.5564 re_mapping 0.0069 re_causal 0.0206 /// teacc 98.96 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1291, 0.1645, -0.0324, ..., -0.0076, 0.0722, -0.0004], + [ 0.1297, -0.0944, 0.0281, ..., -0.0642, -0.0140, -0.0603], + [-0.0587, -0.1560, -0.0035, ..., -0.0404, 0.0457, -0.1600], + ..., + [-0.1389, -0.1536, 0.0534, ..., -0.0680, -0.0698, -0.1135], + [-0.0711, 0.0009, 0.0266, ..., 0.0258, -0.0931, 0.0447], + [-0.0598, -0.0274, 0.0890, ..., 0.1519, -0.0475, -0.1391]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, -2.4810e-05, -5.7630e-06, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-1.7229e-06, 2.2724e-06, -2.2296e-06, ..., 1.8626e-09, + 0.0000e+00, 2.9802e-08], + [ 1.0245e-07, 5.4277e-06, 1.2722e-06, ..., 0.0000e+00, + 0.0000e+00, 5.9605e-08], + ..., + [ 2.9430e-07, 4.5262e-07, 3.8370e-07, ..., 2.9802e-08, + 0.0000e+00, 2.9802e-08], + [ 3.8184e-07, 3.5409e-06, 1.4454e-06, ..., 1.8626e-09, + 0.0000e+00, 2.0489e-07], + [ 3.9488e-07, 4.9025e-06, 1.9390e-06, ..., -6.1467e-08, + 0.0000e+00, 3.9116e-08]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0292, 0.0066, 0.0172, 0.0213, 0.0324, 0.0064, -0.0030, 0.0018, + 0.0011, -0.0372], device='cuda:0'), grad: tensor([-3.4213e-05, -1.5087e-07, 7.9870e-06, -1.9185e-07, -9.7789e-07, + 9.4436e-07, 7.0706e-06, 4.6343e-06, 6.1914e-06, 8.6576e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 217.74, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.5179 re_mapping 0.0065 re_causal 0.0202 /// teacc 99.00 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1293, 0.1648, -0.0324, ..., -0.0076, 0.0722, -0.0008], + [ 0.1312, -0.0942, 0.0286, ..., -0.0645, -0.0140, -0.0601], + [-0.0598, -0.1561, -0.0040, ..., -0.0403, 0.0457, -0.1603], + ..., + [-0.1396, -0.1544, 0.0533, ..., -0.0680, -0.0698, -0.1136], + [-0.0717, 0.0008, 0.0264, ..., 0.0257, -0.0932, 0.0448], + [-0.0599, -0.0277, 0.0890, ..., 0.1522, -0.0475, -0.1395]], + device='cuda:0'), grad: tensor([[ 6.3330e-08, -8.3447e-06, -3.4869e-06, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [-3.9399e-05, 2.6822e-07, -7.7784e-05, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-08], + [ 5.5879e-08, 9.3132e-07, 5.2154e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1548e-07], + ..., + [ 2.5593e-06, 6.7055e-07, 4.5151e-06, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-08], + [ 2.9802e-07, 2.9802e-07, 5.2154e-08, ..., 3.7253e-09, + 0.0000e+00, -1.6391e-07], + [ 2.0489e-06, 4.3213e-06, 4.6603e-06, ..., -1.1176e-08, + 0.0000e+00, 2.6450e-07]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0293, 0.0078, 0.0165, 0.0211, 0.0328, 0.0063, -0.0029, 0.0014, + 0.0008, -0.0376], device='cuda:0'), grad: tensor([-1.6049e-05, -1.5569e-04, 2.4624e-06, 2.6263e-06, 1.4067e-04, + -9.5367e-07, 1.9744e-06, 8.4937e-06, 1.1399e-06, 1.4991e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 217.87, cls_loss 0.0019 cls_loss_mapping 0.0029 cls_loss_causal 0.5237 re_mapping 0.0067 re_causal 0.0197 /// teacc 98.97 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1293, 0.1652, -0.0323, ..., -0.0077, 0.0725, -0.0011], + [ 0.1326, -0.0944, 0.0289, ..., -0.0644, -0.0140, -0.0601], + [-0.0614, -0.1543, -0.0036, ..., -0.0401, 0.0458, -0.1586], + ..., + [-0.1399, -0.1555, 0.0533, ..., -0.0682, -0.0698, -0.1137], + [-0.0721, 0.0004, 0.0260, ..., 0.0256, -0.0933, 0.0444], + [-0.0600, -0.0278, 0.0889, ..., 0.1525, -0.0476, -0.1398]], + device='cuda:0'), grad: tensor([[ 2.6822e-07, 1.3530e-05, 1.1370e-05, ..., 2.8126e-06, + 0.0000e+00, 1.4044e-06], + [ 1.0407e-04, 4.0233e-05, -6.1840e-07, ..., 2.2352e-08, + 0.0000e+00, 5.4479e-05], + [ 4.9546e-07, 1.3150e-06, 1.1846e-06, ..., 3.3155e-07, + 3.7253e-09, 1.5758e-06], + ..., + [ 5.8115e-07, 1.2293e-07, 1.2293e-07, ..., 3.7253e-09, + 0.0000e+00, 2.0117e-07], + [ 1.9193e-05, 8.5086e-06, 1.3560e-06, ..., 2.8685e-07, + -7.4506e-09, 8.0913e-06], + [ 1.8477e-06, -1.6421e-05, -1.4395e-05, ..., -3.5726e-06, + 0.0000e+00, -5.8487e-07]], device='cuda:0') +Epoch 190, bias, value: tensor([ 2.9465e-02, 8.4648e-03, 1.6531e-02, 2.1288e-02, 3.2799e-02, + 6.1162e-03, -3.0418e-03, 1.3587e-03, 6.8367e-05, -3.7699e-02], + device='cuda:0'), grad: tensor([ 2.7582e-05, 1.1039e-04, 3.2671e-06, 1.2219e-04, 2.1234e-07, + -2.5606e-04, 1.1213e-06, 7.2271e-07, 2.2650e-05, -3.2127e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 217.83, cls_loss 0.0020 cls_loss_mapping 0.0039 cls_loss_causal 0.5227 re_mapping 0.0067 re_causal 0.0187 /// teacc 98.94 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1289, 0.1660, -0.0323, ..., -0.0077, 0.0728, -0.0003], + [ 0.1322, -0.0945, 0.0282, ..., -0.0644, -0.0140, -0.0608], + [-0.0606, -0.1559, -0.0035, ..., -0.0406, 0.0457, -0.1599], + ..., + [-0.1400, -0.1563, 0.0539, ..., -0.0685, -0.0698, -0.1135], + [-0.0724, 0.0009, 0.0261, ..., 0.0268, -0.0934, 0.0448], + [-0.0603, -0.0281, 0.0890, ..., 0.1528, -0.0476, -0.1410]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -4.7684e-07, -1.3411e-07, ..., -7.4506e-09, + 0.0000e+00, -7.4506e-09], + [-6.4075e-07, 1.1176e-08, -7.3016e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.2154e-08, 3.7253e-08, 1.7881e-07, ..., 1.8626e-08, + 0.0000e+00, 4.4703e-08], + ..., + [ 1.3784e-07, 1.4901e-08, 7.4506e-09, ..., 3.7253e-09, + 0.0000e+00, 7.4506e-09], + [ 9.6858e-08, -1.7136e-07, -9.7603e-07, ..., -1.6764e-07, + 0.0000e+00, -3.9861e-07], + [ 1.7136e-07, 5.2899e-07, 1.2517e-06, ..., 1.4529e-07, + 0.0000e+00, 3.7253e-07]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0296, 0.0074, 0.0168, 0.0214, 0.0329, 0.0060, -0.0031, 0.0020, + 0.0001, -0.0378], device='cuda:0'), grad: tensor([ 8.3447e-07, -1.2293e-07, -4.2796e-05, 2.3916e-06, 1.1832e-05, + 2.1979e-07, 9.8348e-07, 2.2471e-05, 5.5879e-07, 3.5726e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 218.01, cls_loss 0.0022 cls_loss_mapping 0.0045 cls_loss_causal 0.5813 re_mapping 0.0068 re_causal 0.0205 /// teacc 98.95 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1291, 0.1661, -0.0323, ..., -0.0079, 0.0726, -0.0002], + [ 0.1323, -0.0946, 0.0283, ..., -0.0639, -0.0140, -0.0611], + [-0.0602, -0.1568, -0.0023, ..., -0.0408, 0.0458, -0.1595], + ..., + [-0.1402, -0.1570, 0.0538, ..., -0.0686, -0.0701, -0.1152], + [-0.0729, 0.0026, 0.0265, ..., 0.0284, -0.0935, 0.0460], + [-0.0604, -0.0285, 0.0887, ..., 0.1522, -0.0472, -0.1430]], + device='cuda:0'), grad: tensor([[ 1.3039e-07, 3.5390e-07, -2.6450e-07, ..., 5.2154e-08, + 0.0000e+00, 1.0505e-06], + [-1.1176e-08, 3.9861e-07, 4.8429e-08, ..., 1.1176e-08, + 3.7253e-09, 2.9430e-07], + [ 2.4326e-06, 2.9474e-05, 4.9137e-06, ..., 1.0319e-06, + 0.0000e+00, 2.0728e-05], + ..., + [ 4.4703e-08, 2.6450e-07, -2.2352e-08, ..., 7.4506e-09, + 0.0000e+00, 1.8999e-07], + [-6.2063e-06, -7.6652e-05, -1.2651e-05, ..., -2.7008e-06, + 3.7253e-09, -5.4151e-05], + [ 2.4959e-07, 3.1665e-06, 7.2271e-07, ..., 8.1956e-08, + 0.0000e+00, 1.7844e-06]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0296, 0.0072, 0.0177, 0.0236, 0.0334, 0.0055, -0.0027, 0.0010, + 0.0007, -0.0383], device='cuda:0'), grad: tensor([ 1.2286e-05, 1.0394e-06, 4.8995e-05, 2.3067e-05, 1.9923e-05, + 4.9770e-05, -3.5107e-05, 6.7055e-07, -1.2612e-04, 5.2974e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 218.10, cls_loss 0.0026 cls_loss_mapping 0.0043 cls_loss_causal 0.5598 re_mapping 0.0068 re_causal 0.0194 /// teacc 98.91 lr 0.00010000 +Epoch 193, weight, value: tensor([[-1.2826e-01, 1.6686e-01, -3.2265e-02, ..., -8.1884e-03, + 7.2637e-02, 1.3726e-04], + [ 1.3308e-01, -9.5010e-02, 2.8872e-02, ..., -6.2466e-02, + -1.3942e-02, -6.0262e-02], + [-6.0504e-02, -1.5818e-01, -2.6647e-03, ..., -4.0893e-02, + 4.5783e-02, -1.6006e-01], + ..., + [-1.4071e-01, -1.5864e-01, 5.3862e-02, ..., -6.8779e-02, + -7.0320e-02, -1.1577e-01], + [-7.4780e-02, 2.7125e-03, 2.6029e-02, ..., 2.8250e-02, + -9.3661e-02, 4.5927e-02], + [-6.1016e-02, -2.8933e-02, 8.8703e-02, ..., 1.5242e-01, + -4.7284e-02, -1.4401e-01]], device='cuda:0'), grad: tensor([[-1.4156e-07, -1.0043e-05, -3.0324e-06, ..., 0.0000e+00, + 1.8626e-08, 2.6822e-07], + [-9.6858e-08, 6.9663e-07, -1.4901e-08, ..., 0.0000e+00, + 7.4506e-09, 1.7881e-07], + [ 4.4703e-07, 9.0152e-07, 3.5018e-07, ..., 0.0000e+00, + -1.1921e-07, 2.5891e-06], + ..., + [ 2.1234e-07, 4.1351e-07, -6.1840e-07, ..., 3.7253e-09, + 7.4506e-09, 1.0058e-07], + [ 3.2410e-07, 1.1586e-06, 3.0920e-07, ..., 3.7253e-09, + 1.4901e-08, 2.8312e-07], + [ 1.2219e-06, 3.6061e-06, 1.3188e-06, ..., -1.8626e-08, + 1.1176e-08, 2.2724e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([ 3.0085e-02, 7.7514e-03, 1.7479e-02, 2.3378e-02, 3.3022e-02, + 5.9898e-03, -2.4470e-03, 8.4071e-04, -4.3467e-05, -3.8558e-02], + device='cuda:0'), grad: tensor([-1.3694e-05, 2.5183e-06, 2.7567e-05, -2.9802e-05, 6.3330e-07, + 7.3612e-05, -7.0751e-05, -1.0282e-06, 3.1628e-06, 7.8678e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 217.51, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5368 re_mapping 0.0065 re_causal 0.0199 /// teacc 98.96 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1280, 0.1673, -0.0323, ..., -0.0085, 0.0724, -0.0008], + [ 0.1339, -0.0952, 0.0293, ..., -0.0624, -0.0136, -0.0604], + [-0.0605, -0.1586, -0.0028, ..., -0.0409, 0.0463, -0.1602], + ..., + [-0.1414, -0.1593, 0.0537, ..., -0.0689, -0.0708, -0.1159], + [-0.0752, 0.0032, 0.0261, ..., 0.0284, -0.0934, 0.0463], + [-0.0614, -0.0293, 0.0887, ..., 0.1526, -0.0475, -0.1448]], + device='cuda:0'), grad: tensor([[ 1.5274e-07, -3.7253e-07, 1.0021e-06, ..., 2.1607e-07, + 0.0000e+00, 3.5763e-07], + [-1.1176e-08, 2.4214e-07, -2.6077e-08, ..., 4.4703e-08, + 0.0000e+00, 7.0035e-07], + [ 1.8626e-07, 1.3784e-07, 3.2410e-07, ..., 6.7055e-08, + 0.0000e+00, 3.7625e-07], + ..., + [ 7.1898e-07, 5.4017e-07, 5.5879e-07, ..., 1.3411e-07, + 0.0000e+00, 1.4827e-06], + [ 3.0875e-05, 2.5705e-05, -1.5423e-06, ..., -1.0431e-07, + 0.0000e+00, 7.0810e-05], + [ 2.1607e-07, 6.0722e-07, -1.8135e-05, ..., -1.1936e-05, + 0.0000e+00, 2.0452e-06]], device='cuda:0') +Epoch 194, bias, value: tensor([ 3.0273e-02, 8.0477e-03, 1.7702e-02, 2.2195e-02, 3.3210e-02, + 6.7446e-03, -2.6052e-03, 6.9272e-04, -1.6843e-05, -3.8841e-02], + device='cuda:0'), grad: tensor([ 1.4752e-06, 1.1772e-06, 1.1399e-06, 9.7603e-07, 3.8683e-05, + -1.2410e-04, 9.6411e-06, 2.9653e-06, 1.0550e-04, -3.7551e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 217.38, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.5186 re_mapping 0.0064 re_causal 0.0188 /// teacc 98.98 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1282, 0.1678, -0.0323, ..., -0.0085, 0.0721, -0.0013], + [ 0.1344, -0.0951, 0.0294, ..., -0.0630, -0.0133, -0.0603], + [-0.0600, -0.1594, -0.0031, ..., -0.0409, 0.0476, -0.1605], + ..., + [-0.1419, -0.1604, 0.0537, ..., -0.0706, -0.0720, -0.1160], + [-0.0756, 0.0033, 0.0258, ..., 0.0282, -0.0939, 0.0463], + [-0.0615, -0.0296, 0.0892, ..., 0.1539, -0.0483, -0.1449]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -5.6885e-06, -3.0138e-06, ..., 0.0000e+00, + 0.0000e+00, 6.7055e-08], + [-1.7136e-07, 1.1921e-07, 2.0452e-06, ..., 0.0000e+00, + -1.1176e-08, 4.0978e-08], + [ 2.2352e-08, 9.2015e-07, 6.2957e-07, ..., 0.0000e+00, + -1.1176e-08, 1.4901e-08], + ..., + [ 1.5646e-07, 1.3784e-07, -4.3586e-06, ..., 0.0000e+00, + 7.4506e-09, 1.7136e-07], + [ 2.3469e-07, 3.7998e-07, 6.6310e-07, ..., 0.0000e+00, + 3.7253e-09, 3.2783e-07], + [ 1.0431e-07, 8.7172e-07, 1.5460e-06, ..., 0.0000e+00, + 0.0000e+00, 1.5274e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0303, 0.0077, 0.0188, 0.0214, 0.0329, 0.0066, -0.0025, 0.0008, + -0.0004, -0.0386], device='cuda:0'), grad: tensor([-1.4842e-05, 4.5337e-06, 2.0899e-06, 2.2277e-06, 8.5682e-08, + -3.3714e-06, 1.0088e-05, -7.8529e-06, 2.1830e-06, 4.7795e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 217.31, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.5372 re_mapping 0.0062 re_causal 0.0189 /// teacc 98.89 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1286, 0.1684, -0.0322, ..., -0.0087, 0.0716, -0.0020], + [ 0.1342, -0.0961, 0.0291, ..., -0.0641, -0.0132, -0.0607], + [-0.0596, -0.1605, -0.0026, ..., -0.0404, 0.0471, -0.1606], + ..., + [-0.1420, -0.1619, 0.0537, ..., -0.0709, -0.0717, -0.1165], + [-0.0767, 0.0029, 0.0258, ..., 0.0284, -0.0949, 0.0463], + [-0.0610, -0.0298, 0.0897, ..., 0.1557, -0.0482, -0.1451]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.1921e-06, -1.0058e-07, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + [-1.7509e-07, 2.6077e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 6.3330e-08], + [ 7.8231e-08, 2.9802e-08, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -8.1211e-07], + ..., + [-5.2154e-08, 1.8626e-08, -5.5507e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7625e-07], + [ 4.4703e-07, 8.1956e-08, -2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, 2.1607e-07], + [ 4.1723e-07, 3.3155e-07, 3.2037e-07, ..., -3.7253e-09, + 0.0000e+00, 3.0547e-07]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0305, 0.0070, 0.0198, 0.0212, 0.0318, 0.0070, -0.0023, 0.0007, + -0.0010, -0.0380], device='cuda:0'), grad: tensor([-1.4827e-06, 1.5348e-06, -2.4587e-05, 7.7188e-06, 1.8626e-07, + 2.7269e-06, 2.2128e-06, 8.9407e-06, 1.2852e-06, 1.4305e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 217.41, cls_loss 0.0026 cls_loss_mapping 0.0037 cls_loss_causal 0.5262 re_mapping 0.0061 re_causal 0.0186 /// teacc 98.91 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1289, 0.1686, -0.0322, ..., -0.0089, 0.0715, -0.0030], + [ 0.1366, -0.0969, 0.0299, ..., -0.0645, -0.0111, -0.0612], + [-0.0605, -0.1619, -0.0037, ..., -0.0406, 0.0457, -0.1612], + ..., + [-0.1438, -0.1633, 0.0536, ..., -0.0712, -0.0739, -0.1168], + [-0.0773, 0.0038, 0.0258, ..., 0.0281, -0.0948, 0.0468], + [-0.0614, -0.0299, 0.0898, ..., 0.1567, -0.0493, -0.1456]], + device='cuda:0'), grad: tensor([[ 1.8999e-07, -3.7253e-08, 1.7509e-07, ..., 0.0000e+00, + 0.0000e+00, 4.6939e-07], + [ 7.1153e-07, 5.9977e-07, 2.0601e-06, ..., 0.0000e+00, + 0.0000e+00, 2.7865e-06], + [ 2.0862e-07, 1.3039e-07, 3.7998e-07, ..., 0.0000e+00, + 0.0000e+00, 6.0350e-07], + ..., + [ 2.5705e-07, 1.0431e-07, -9.6038e-06, ..., 3.7253e-09, + 0.0000e+00, 2.9430e-07], + [-2.6301e-06, -1.4678e-06, -4.3064e-06, ..., 0.0000e+00, + 0.0000e+00, -9.5069e-06], + [ 1.5274e-07, 1.7136e-07, 8.3074e-06, ..., -7.4506e-09, + 0.0000e+00, 2.4959e-07]], device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0305, 0.0075, 0.0192, 0.0216, 0.0315, 0.0066, -0.0023, 0.0007, + -0.0008, -0.0379], device='cuda:0'), grad: tensor([ 8.5682e-07, 6.9626e-06, 1.4082e-06, 3.7588e-06, -1.4901e-08, + 1.4007e-06, 6.0052e-06, -1.2234e-05, -1.9521e-05, 1.1362e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 217.49, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5297 re_mapping 0.0064 re_causal 0.0184 /// teacc 98.95 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1300, 0.1680, -0.0322, ..., -0.0090, 0.0714, -0.0040], + [ 0.1365, -0.0993, 0.0299, ..., -0.0659, -0.0107, -0.0629], + [-0.0601, -0.1635, -0.0035, ..., -0.0408, 0.0453, -0.1617], + ..., + [-0.1445, -0.1650, 0.0535, ..., -0.0717, -0.0746, -0.1169], + [-0.0729, 0.0082, 0.0271, ..., 0.0288, -0.0941, 0.0504], + [-0.0620, -0.0303, 0.0899, ..., 0.1572, -0.0494, -0.1490]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, -2.3469e-07, 0.0000e+00, ..., 2.2352e-08, + 0.0000e+00, 1.8626e-08], + [-1.5199e-06, 7.4506e-09, -1.1660e-06, ..., 1.8626e-08, + 0.0000e+00, 1.4901e-08], + [ 3.2783e-07, 1.8626e-08, 6.3702e-07, ..., 4.0978e-08, + 0.0000e+00, 3.7253e-08], + ..., + [ 4.9919e-07, 0.0000e+00, -2.5555e-06, ..., 1.1176e-08, + 0.0000e+00, 3.7253e-09], + [ 7.1526e-07, 5.2154e-07, 1.2219e-06, ..., 5.5507e-07, + 0.0000e+00, 8.1584e-07], + [ 2.9057e-07, 3.7253e-09, -2.9057e-07, ..., -1.6652e-06, + 0.0000e+00, -1.3523e-06]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0297, 0.0070, 0.0199, 0.0216, 0.0316, 0.0049, -0.0011, 0.0005, + 0.0020, -0.0380], device='cuda:0'), grad: tensor([ 2.7940e-07, -1.2517e-06, 1.4119e-06, 3.9898e-06, -9.7901e-06, + -4.4703e-08, -3.1665e-06, -5.0291e-06, 6.5491e-06, 6.9961e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 217.60, cls_loss 0.0019 cls_loss_mapping 0.0033 cls_loss_causal 0.5286 re_mapping 0.0064 re_causal 0.0190 /// teacc 98.94 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1296, 0.1686, -0.0321, ..., -0.0087, 0.0713, -0.0040], + [ 0.1380, -0.0994, 0.0302, ..., -0.0659, -0.0107, -0.0631], + [-0.0613, -0.1641, -0.0041, ..., -0.0409, 0.0469, -0.1622], + ..., + [-0.1447, -0.1664, 0.0536, ..., -0.0718, -0.0749, -0.1172], + [-0.0742, 0.0075, 0.0269, ..., 0.0287, -0.0947, 0.0497], + [-0.0622, -0.0306, 0.0898, ..., 0.1574, -0.0495, -0.1490]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, -5.7742e-07, 3.9451e-06, ..., 0.0000e+00, + 3.7253e-09, 1.1176e-08], + [-8.5309e-07, 8.9407e-08, -3.6880e-07, ..., 0.0000e+00, + -9.6858e-08, 3.7253e-09], + [ 1.3039e-07, 1.6019e-07, 3.3155e-07, ..., 3.7253e-09, + 0.0000e+00, -3.7253e-08], + ..., + [ 3.3900e-07, -4.4890e-06, -8.3596e-06, ..., 0.0000e+00, + 4.4703e-08, 7.4506e-09], + [ 1.4529e-07, 2.3842e-07, 4.0606e-07, ..., -3.7253e-09, + 2.2352e-08, 3.3528e-08], + [ 1.0058e-07, 3.3379e-06, 2.2985e-06, ..., 0.0000e+00, + 1.4901e-08, 1.1176e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0300, 0.0076, 0.0192, 0.0217, 0.0330, 0.0048, -0.0007, 0.0006, + 0.0010, -0.0389], device='cuda:0'), grad: tensor([ 1.1466e-05, -1.4156e-07, 6.5193e-07, 1.6689e-06, 6.4820e-07, + 5.3160e-06, 1.1921e-06, -4.0352e-05, 2.1942e-06, 1.7300e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 217.38, cls_loss 0.0017 cls_loss_mapping 0.0025 cls_loss_causal 0.4894 re_mapping 0.0063 re_causal 0.0186 /// teacc 98.83 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1297, 0.1694, -0.0324, ..., -0.0113, 0.0713, -0.0037], + [ 0.1384, -0.0997, 0.0302, ..., -0.0661, -0.0106, -0.0629], + [-0.0616, -0.1658, -0.0041, ..., -0.0410, 0.0465, -0.1631], + ..., + [-0.1451, -0.1679, 0.0536, ..., -0.0721, -0.0752, -0.1175], + [-0.0743, 0.0079, 0.0270, ..., 0.0288, -0.0927, 0.0505], + [-0.0623, -0.0309, 0.0900, ..., 0.1582, -0.0499, -0.1496]], + device='cuda:0'), grad: tensor([[-1.8254e-07, -2.1636e-05, -9.9242e-06, ..., 0.0000e+00, + -2.3283e-06, -3.3900e-06], + [-1.8626e-08, 4.2617e-06, 2.1830e-06, ..., 3.7253e-09, + 7.7114e-07, 3.2783e-07], + [ 2.9802e-08, 9.1642e-07, 4.3213e-07, ..., 3.7253e-09, + 4.8429e-08, 2.7195e-07], + ..., + [ 4.4703e-08, 4.7684e-07, 2.3469e-07, ..., 3.7253e-09, + 4.8429e-08, 1.1548e-07], + [ 1.3039e-07, 1.0386e-05, 4.9882e-06, ..., -1.8626e-08, + 1.0952e-06, 1.8701e-06], + [ 7.4506e-08, 2.4140e-06, -1.6391e-07, ..., -1.4901e-08, + 1.4901e-07, 5.9232e-07]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0299, 0.0076, 0.0191, 0.0217, 0.0334, 0.0049, -0.0014, 0.0005, + 0.0015, -0.0390], device='cuda:0'), grad: tensor([-2.9385e-05, 5.8673e-06, 1.2554e-06, -1.1697e-06, 1.7583e-06, + 2.7679e-06, 1.8999e-06, 6.9663e-07, 1.4678e-05, 1.5646e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 217.69, cls_loss 0.0024 cls_loss_mapping 0.0036 cls_loss_causal 0.5190 re_mapping 0.0062 re_causal 0.0182 /// teacc 98.93 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1295, 0.1710, -0.0321, ..., -0.0107, 0.0716, -0.0013], + [ 0.1390, -0.0998, 0.0304, ..., -0.0652, -0.0106, -0.0632], + [-0.0620, -0.1674, -0.0036, ..., -0.0405, 0.0480, -0.1644], + ..., + [-0.1454, -0.1694, 0.0535, ..., -0.0732, -0.0760, -0.1177], + [-0.0742, 0.0085, 0.0271, ..., 0.0288, -0.0924, 0.0514], + [-0.0627, -0.0320, 0.0894, ..., 0.1574, -0.0508, -0.1509]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -4.0978e-08, 1.1176e-08, ..., 7.4506e-09, + 0.0000e+00, 3.7253e-08], + [-7.4506e-08, 3.7253e-09, 5.9605e-08, ..., 1.1176e-08, + 1.8626e-08, 6.7055e-08], + [ 1.1176e-08, 3.7253e-09, 1.8254e-06, ..., 5.9605e-08, + 1.4901e-07, 3.5018e-07], + ..., + [ 4.0978e-08, 3.7253e-09, -2.0340e-06, ..., 2.6077e-08, + -2.1234e-07, 8.9407e-08], + [ 5.2154e-08, 3.3528e-08, 7.8231e-08, ..., 3.7253e-08, + 2.2352e-08, 2.0489e-07], + [ 1.1176e-08, 1.8626e-08, -1.8254e-07, ..., -7.0781e-08, + 7.4506e-09, 1.0803e-07]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0307, 0.0076, 0.0205, 0.0217, 0.0344, 0.0044, -0.0023, -0.0002, + 0.0023, -0.0404], device='cuda:0'), grad: tensor([ 1.3970e-05, 9.0897e-07, -3.4243e-05, 1.2547e-05, 4.2468e-07, + 5.1782e-07, 3.7625e-07, 2.1532e-06, 2.9691e-06, 3.4273e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 217.48, cls_loss 0.0022 cls_loss_mapping 0.0023 cls_loss_causal 0.5013 re_mapping 0.0062 re_causal 0.0184 /// teacc 98.86 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1300, 0.1713, -0.0320, ..., -0.0107, 0.0715, -0.0030], + [ 0.1391, -0.1001, 0.0303, ..., -0.0651, -0.0108, -0.0635], + [-0.0620, -0.1685, -0.0033, ..., -0.0404, 0.0493, -0.1655], + ..., + [-0.1452, -0.1704, 0.0542, ..., -0.0734, -0.0764, -0.1180], + [-0.0744, 0.0085, 0.0270, ..., 0.0286, -0.0927, 0.0516], + [-0.0634, -0.0322, 0.0889, ..., 0.1576, -0.0511, -0.1513]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.4273e-07, -8.5682e-08, ..., 3.7253e-09, + 0.0000e+00, -3.7253e-09], + [-1.0431e-07, 4.0978e-08, 1.9744e-07, ..., 3.7253e-09, + -2.6077e-08, 1.4901e-08], + [ 7.4506e-09, 6.3330e-08, -2.7940e-07, ..., -1.8999e-07, + 0.0000e+00, 1.0058e-07], + ..., + [ 1.1176e-08, 6.3330e-08, -9.8348e-07, ..., 6.3330e-08, + 0.0000e+00, 3.3528e-08], + [ 1.0431e-07, 7.8231e-08, 7.1526e-07, ..., 1.2293e-07, + 1.8626e-08, 7.0781e-08], + [ 1.4901e-08, 7.4506e-08, 1.6391e-07, ..., -3.7253e-09, + 0.0000e+00, 3.3528e-08]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0306, 0.0073, 0.0211, 0.0213, 0.0346, 0.0052, -0.0023, 0.0003, + 0.0021, -0.0410], device='cuda:0'), grad: tensor([ 1.1511e-06, 9.2387e-07, -9.6112e-07, -5.0291e-07, 7.0035e-07, + 5.4389e-07, -4.6156e-06, -2.2799e-06, 4.2692e-06, 7.3761e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 217.88, cls_loss 0.0023 cls_loss_mapping 0.0029 cls_loss_causal 0.4938 re_mapping 0.0064 re_causal 0.0185 /// teacc 98.95 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1300, 0.1717, -0.0320, ..., -0.0108, 0.0723, -0.0024], + [ 0.1412, -0.0996, 0.0300, ..., -0.0630, -0.0125, -0.0636], + [-0.0631, -0.1689, -0.0029, ..., -0.0403, 0.0481, -0.1662], + ..., + [-0.1470, -0.1715, 0.0547, ..., -0.0752, -0.0740, -0.1190], + [-0.0748, 0.0086, 0.0269, ..., 0.0279, -0.0930, 0.0516], + [-0.0639, -0.0323, 0.0888, ..., 0.1581, -0.0523, -0.1516]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.2666e-07, 3.6508e-07, ..., 1.1176e-08, + 0.0000e+00, -1.1176e-08], + [-2.0862e-07, 1.1176e-08, 1.0133e-06, ..., 3.7253e-09, + -4.8429e-08, 1.4901e-08], + [ 2.6077e-08, 3.3528e-08, 2.9057e-07, ..., 1.1176e-08, + 0.0000e+00, 4.0978e-08], + ..., + [ 5.9605e-08, 7.4506e-09, -3.4235e-06, ..., 1.4901e-08, + 1.1176e-08, -2.8312e-07], + [ 1.1176e-08, -1.4901e-08, 1.8626e-08, ..., -3.7253e-08, + 0.0000e+00, -2.2352e-08], + [ 1.0058e-07, 1.1548e-07, -1.0058e-06, ..., -5.9605e-08, + 1.1176e-08, -7.4506e-08]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0306, 0.0071, 0.0211, 0.0209, 0.0348, 0.0059, -0.0023, 0.0003, + 0.0018, -0.0412], device='cuda:0'), grad: tensor([ 4.8429e-07, 1.5572e-06, 5.0664e-07, 3.7588e-06, 1.0803e-06, + 8.1956e-08, 1.1176e-08, -6.1616e-06, 2.3469e-07, -1.5721e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 217.58, cls_loss 0.0018 cls_loss_mapping 0.0041 cls_loss_causal 0.5337 re_mapping 0.0063 re_causal 0.0187 /// teacc 98.93 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1302, 0.1725, -0.0317, ..., -0.0084, 0.0719, -0.0022], + [ 0.1408, -0.0981, 0.0283, ..., -0.0601, -0.0157, -0.0636], + [-0.0633, -0.1697, -0.0026, ..., -0.0406, 0.0474, -0.1669], + ..., + [-0.1459, -0.1730, 0.0564, ..., -0.0761, -0.0711, -0.1194], + [-0.0754, 0.0100, 0.0273, ..., 0.0276, -0.0934, 0.0535], + [-0.0646, -0.0333, 0.0886, ..., 0.1571, -0.0537, -0.1520]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -7.8976e-07, -5.2527e-07, ..., 7.4506e-09, + 0.0000e+00, 2.2352e-08], + [ 3.7253e-09, 2.6077e-08, 1.1921e-07, ..., 1.8626e-08, + 0.0000e+00, 2.2352e-08], + [ 3.7253e-09, 5.6997e-07, 6.0350e-07, ..., 6.7055e-08, + 0.0000e+00, 4.3586e-07], + ..., + [ 7.4506e-09, 1.4901e-08, 3.4161e-06, ..., 5.7369e-07, + 0.0000e+00, 1.1176e-08], + [ 3.2783e-07, -5.2527e-07, -6.1467e-07, ..., -1.3784e-07, + 2.6077e-08, -5.9977e-07], + [ 7.4506e-09, 6.4075e-07, -5.4426e-06, ..., -5.7742e-07, + 0.0000e+00, 1.5646e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0309, 0.0052, 0.0215, 0.0209, 0.0348, 0.0058, -0.0035, 0.0019, + 0.0031, -0.0414], device='cuda:0'), grad: tensor([-1.0915e-06, 5.3868e-06, 1.5423e-06, 8.0839e-07, -5.6475e-05, + 1.0096e-06, 8.2701e-07, 5.0455e-05, -1.2666e-07, -2.4140e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 217.79, cls_loss 0.0028 cls_loss_mapping 0.0046 cls_loss_causal 0.5382 re_mapping 0.0063 re_causal 0.0182 /// teacc 98.88 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1333, 0.1718, -0.0318, ..., -0.0088, 0.0722, -0.0035], + [ 0.1412, -0.0980, 0.0283, ..., -0.0606, -0.0158, -0.0637], + [-0.0643, -0.1724, -0.0037, ..., -0.0421, 0.0443, -0.1708], + ..., + [-0.1462, -0.1739, 0.0566, ..., -0.0776, -0.0702, -0.1196], + [-0.0765, 0.0101, 0.0277, ..., 0.0288, -0.0944, 0.0544], + [-0.0646, -0.0336, 0.0886, ..., 0.1580, -0.0545, -0.1525]], + device='cuda:0'), grad: tensor([[ 8.3074e-07, -6.0350e-07, 3.5018e-07, ..., 0.0000e+00, + 1.0431e-07, -1.1176e-08], + [-1.2410e-04, 2.6077e-08, -9.6858e-05, ..., 3.7253e-08, + -1.6034e-05, 2.2352e-08], + [ 2.0131e-05, 8.1956e-08, 1.5825e-05, ..., 0.0000e+00, + 2.6003e-06, 5.5879e-08], + ..., + [ 9.7156e-05, 4.4703e-08, 6.8665e-05, ..., 2.2352e-08, + 1.2547e-05, 4.4703e-08], + [ 2.8647e-06, 2.4624e-06, 8.4192e-07, ..., 6.3330e-08, + 1.0431e-07, 1.7136e-06], + [ 3.5055e-06, 1.2666e-06, 8.5756e-06, ..., -1.7509e-07, + 3.0175e-07, 8.3447e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0302, 0.0051, 0.0184, 0.0238, 0.0350, 0.0043, -0.0040, 0.0026, + 0.0037, -0.0414], device='cuda:0'), grad: tensor([ 9.3877e-07, -2.3091e-04, 3.0398e-05, 1.1832e-05, 3.8520e-06, + -1.1779e-05, 2.4065e-06, 1.7321e-04, 6.4559e-06, 1.3232e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 217.62, cls_loss 0.0021 cls_loss_mapping 0.0030 cls_loss_causal 0.5263 re_mapping 0.0063 re_causal 0.0189 /// teacc 98.93 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1334, 0.1721, -0.0318, ..., -0.0092, 0.0724, -0.0035], + [ 0.1421, -0.0981, 0.0287, ..., -0.0606, -0.0158, -0.0639], + [-0.0646, -0.1734, -0.0045, ..., -0.0424, 0.0443, -0.1719], + ..., + [-0.1472, -0.1747, 0.0566, ..., -0.0777, -0.0702, -0.1198], + [-0.0771, 0.0099, 0.0277, ..., 0.0289, -0.0946, 0.0544], + [-0.0649, -0.0337, 0.0884, ..., 0.1583, -0.0548, -0.1527]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -2.3060e-06, -1.1735e-06, ..., -7.4506e-09, + 0.0000e+00, -4.3213e-07], + [-2.2352e-07, 3.8370e-07, 3.2410e-07, ..., 0.0000e+00, + -1.1176e-08, 4.4703e-07], + [ 1.4901e-08, 1.5646e-07, 2.7195e-07, ..., 0.0000e+00, + 0.0000e+00, 1.7136e-07], + ..., + [ 4.0978e-08, 7.3388e-07, -2.0452e-06, ..., 3.7253e-09, + 0.0000e+00, 8.9034e-07], + [ 7.8231e-08, -2.7679e-06, -1.9856e-06, ..., 2.6077e-08, + 3.7253e-09, -4.3809e-06], + [ 3.3528e-08, 2.5444e-06, 2.8126e-06, ..., -6.7055e-08, + 3.7253e-09, 2.0638e-06]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0304, 0.0054, 0.0178, 0.0237, 0.0351, 0.0045, -0.0041, 0.0028, + 0.0035, -0.0416], device='cuda:0'), grad: tensor([-2.5779e-06, 1.0580e-06, -7.5810e-06, 1.2890e-06, 2.0675e-06, + 1.1735e-06, 7.7859e-07, -1.6950e-06, -1.1958e-06, 6.6124e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 217.82, cls_loss 0.0018 cls_loss_mapping 0.0039 cls_loss_causal 0.4922 re_mapping 0.0064 re_causal 0.0188 /// teacc 98.92 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1335, 0.1724, -0.0320, ..., -0.0105, 0.0731, -0.0044], + [ 0.1425, -0.0981, 0.0287, ..., -0.0607, -0.0158, -0.0639], + [-0.0648, -0.1741, -0.0051, ..., -0.0426, 0.0441, -0.1722], + ..., + [-0.1474, -0.1759, 0.0578, ..., -0.0779, -0.0701, -0.1199], + [-0.0774, 0.0096, 0.0276, ..., 0.0285, -0.0948, 0.0542], + [-0.0650, -0.0337, 0.0884, ..., 0.1606, -0.0552, -0.1520]], + device='cuda:0'), grad: tensor([[-1.4156e-07, -4.1761e-06, -1.7174e-06, ..., -1.2033e-06, + 3.7253e-09, 1.1176e-08], + [ 1.0803e-07, 2.9802e-08, 1.0245e-06, ..., 3.7253e-09, + -3.7253e-09, 2.2352e-08], + [ 1.8626e-08, 1.5646e-07, 1.7881e-07, ..., 2.6077e-08, + 0.0000e+00, 1.6391e-07], + ..., + [-1.5274e-07, 2.2352e-08, -9.9093e-07, ..., 1.1176e-08, + 3.7253e-09, 1.4901e-08], + [ 1.5646e-07, 1.2033e-06, 5.5879e-07, ..., 4.9546e-07, + 2.9802e-08, -9.3132e-08], + [ 1.8626e-08, 1.8217e-06, 6.5938e-07, ..., 6.3330e-07, + 0.0000e+00, 2.2352e-08]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0303, 0.0054, 0.0176, 0.0237, 0.0342, 0.0045, -0.0041, 0.0040, + 0.0030, -0.0416], device='cuda:0'), grad: tensor([-5.8822e-06, 1.4007e-06, 4.1351e-07, -3.3528e-08, 2.0489e-07, + 2.9430e-07, 6.4075e-07, -1.1846e-06, 2.0079e-06, 2.0936e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 217.59, cls_loss 0.0021 cls_loss_mapping 0.0027 cls_loss_causal 0.5134 re_mapping 0.0062 re_causal 0.0185 /// teacc 98.89 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1336, 0.1729, -0.0321, ..., -0.0114, 0.0730, -0.0045], + [ 0.1441, -0.0982, 0.0295, ..., -0.0611, -0.0152, -0.0641], + [-0.0650, -0.1757, -0.0054, ..., -0.0422, 0.0441, -0.1739], + ..., + [-0.1493, -0.1771, 0.0556, ..., -0.0810, -0.0706, -0.1201], + [-0.0777, 0.0100, 0.0276, ..., 0.0283, -0.0949, 0.0551], + [-0.0651, -0.0343, 0.0908, ..., 0.1629, -0.0556, -0.1524]], + device='cuda:0'), grad: tensor([[ 1.4901e-07, -8.3819e-07, -7.4506e-08, ..., 3.3528e-08, + 3.7253e-08, 3.7253e-08], + [-4.2245e-06, -6.4448e-07, -6.8694e-06, ..., -2.1607e-07, + -7.2271e-07, 4.8429e-08], + [ 3.3528e-07, 5.1782e-07, 2.8685e-07, ..., 2.9802e-08, + 1.8626e-08, 1.3784e-07], + ..., + [ 2.2352e-07, 6.3330e-08, 9.9838e-07, ..., 1.0431e-07, + 3.7253e-09, 9.3132e-08], + [ 5.1297e-06, 9.0152e-07, 5.8189e-06, ..., 5.6252e-07, + 5.4389e-07, 1.2629e-06], + [ 3.8743e-07, 2.0117e-07, -1.5870e-06, ..., -8.9407e-07, + 1.1176e-08, 2.7567e-07]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0301, 0.0061, 0.0174, 0.0235, 0.0340, 0.0045, -0.0033, 0.0018, + 0.0035, -0.0401], device='cuda:0'), grad: tensor([-9.4995e-07, -1.0177e-05, 1.1548e-07, -2.7865e-06, 3.6024e-06, + -1.9670e-05, 1.3463e-05, 3.2969e-06, 1.3545e-05, -5.2899e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 217.51, cls_loss 0.0019 cls_loss_mapping 0.0033 cls_loss_causal 0.5187 re_mapping 0.0063 re_causal 0.0181 /// teacc 98.93 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1337, 0.1734, -0.0320, ..., -0.0116, 0.0730, -0.0049], + [ 0.1444, -0.0983, 0.0294, ..., -0.0610, -0.0152, -0.0642], + [-0.0651, -0.1764, -0.0056, ..., -0.0423, 0.0441, -0.1743], + ..., + [-0.1493, -0.1796, 0.0559, ..., -0.0811, -0.0706, -0.1201], + [-0.0785, 0.0097, 0.0273, ..., 0.0284, -0.0954, 0.0550], + [-0.0664, -0.0349, 0.0906, ..., 0.1631, -0.0558, -0.1526]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -3.7432e-05, -5.5879e-08, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08], + [-3.7812e-07, 1.6950e-07, -1.8254e-07, ..., 0.0000e+00, + 0.0000e+00, 6.3330e-08], + [ 4.6566e-08, 2.3097e-07, 5.1409e-07, ..., 0.0000e+00, + 0.0000e+00, 5.6438e-07], + ..., + [ 3.2969e-07, 3.4273e-07, -2.9150e-06, ..., 1.8626e-09, + 0.0000e+00, -3.1665e-07], + [-5.5321e-07, -1.1642e-06, -8.4750e-07, ..., -1.8626e-09, + 0.0000e+00, -1.5143e-06], + [ 3.7253e-08, 1.8030e-05, 2.2538e-06, ..., -1.4901e-08, + 0.0000e+00, 3.6694e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0300, 0.0059, 0.0177, 0.0234, 0.0340, 0.0047, -0.0030, 0.0021, + 0.0029, -0.0404], device='cuda:0'), grad: tensor([-7.2360e-05, 4.8429e-08, 1.8291e-06, 3.6880e-07, 1.1176e-06, + 8.9779e-06, 2.7776e-05, -4.5411e-06, -2.3581e-06, 3.9011e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 217.84, cls_loss 0.0017 cls_loss_mapping 0.0028 cls_loss_causal 0.5159 re_mapping 0.0060 re_causal 0.0177 /// teacc 98.88 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1333, 0.1748, -0.0319, ..., -0.0116, 0.0730, -0.0030], + [ 0.1456, -0.0982, 0.0302, ..., -0.0602, -0.0152, -0.0635], + [-0.0654, -0.1771, -0.0060, ..., -0.0424, 0.0441, -0.1748], + ..., + [-0.1505, -0.1833, 0.0554, ..., -0.0811, -0.0706, -0.1207], + [-0.0793, 0.0098, 0.0273, ..., 0.0270, -0.0955, 0.0552], + [-0.0665, -0.0354, 0.0907, ..., 0.1634, -0.0559, -0.1536]], + device='cuda:0'), grad: tensor([[ 2.2165e-07, -7.8231e-07, 1.9744e-07, ..., -1.8626e-09, + 0.0000e+00, -2.4214e-08], + [ 2.4289e-05, 8.5682e-08, 5.6565e-05, ..., 1.8626e-09, + 0.0000e+00, 7.8231e-08], + [ 1.3597e-07, 5.7742e-08, 2.0117e-07, ..., 0.0000e+00, + 0.0000e+00, 5.7742e-08], + ..., + [ 2.8349e-06, 5.9605e-08, 6.6236e-06, ..., 4.0978e-08, + 0.0000e+00, 5.0291e-08], + [ 2.4065e-05, 1.7866e-05, -1.0766e-06, ..., 1.8626e-09, + 0.0000e+00, 1.9819e-05], + [-3.2395e-05, 3.0547e-07, -7.6056e-05, ..., -6.3330e-08, + 0.0000e+00, 1.1548e-07]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0306, 0.0067, 0.0176, 0.0233, 0.0345, 0.0047, -0.0030, 0.0015, + 0.0028, -0.0409], device='cuda:0'), grad: tensor([-7.4506e-09, 9.0480e-05, 4.8243e-07, 3.1628e-06, 8.7619e-06, + -5.2154e-05, 1.5318e-05, 1.1005e-05, 4.3303e-05, -1.2028e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 218.01, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5233 re_mapping 0.0061 re_causal 0.0187 /// teacc 98.91 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1334, 0.1754, -0.0319, ..., -0.0123, 0.0733, -0.0032], + [ 0.1462, -0.1010, 0.0301, ..., -0.0593, -0.0143, -0.0667], + [-0.0659, -0.1772, -0.0057, ..., -0.0401, 0.0438, -0.1750], + ..., + [-0.1511, -0.1849, 0.0554, ..., -0.0812, -0.0710, -0.1208], + [-0.0800, 0.0105, 0.0286, ..., 0.0269, -0.0962, 0.0569], + [-0.0681, -0.0359, 0.0905, ..., 0.1632, -0.0592, -0.1550]], + device='cuda:0'), grad: tensor([[-3.8929e-07, -2.0023e-06, -2.2352e-08, ..., 9.3132e-09, + 0.0000e+00, 2.7940e-08], + [-5.9605e-08, 2.2352e-08, 1.0356e-06, ..., 1.7881e-07, + 0.0000e+00, 2.9802e-08], + [ 7.4506e-09, -8.3819e-08, -3.1665e-08, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-09, 1.4901e-08, -6.3963e-06, ..., -1.0319e-06, + 0.0000e+00, -1.4156e-07], + [ 8.5682e-08, 1.1548e-07, 5.3830e-07, ..., 8.1956e-08, + 0.0000e+00, 2.6077e-08], + [ 4.0978e-08, 1.3970e-07, -1.1083e-06, ..., -3.4645e-07, + 0.0000e+00, -3.3528e-07]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0307, 0.0067, 0.0180, 0.0230, 0.0347, 0.0050, -0.0017, 0.0015, + 0.0025, -0.0415], device='cuda:0'), grad: tensor([-2.3413e-06, 1.6261e-06, -8.9966e-07, 6.3814e-06, 4.3437e-06, + 6.2361e-06, -4.4182e-06, -9.4175e-06, 1.8589e-06, -3.4124e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 217.71, cls_loss 0.0027 cls_loss_mapping 0.0046 cls_loss_causal 0.5059 re_mapping 0.0064 re_causal 0.0181 /// teacc 98.96 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1336, 0.1756, -0.0321, ..., -0.0124, 0.0729, -0.0036], + [ 0.1465, -0.1011, 0.0298, ..., -0.0594, -0.0142, -0.0668], + [-0.0660, -0.1776, -0.0061, ..., -0.0401, 0.0439, -0.1753], + ..., + [-0.1510, -0.1855, 0.0533, ..., -0.0839, -0.0710, -0.1211], + [-0.0802, 0.0108, 0.0288, ..., 0.0270, -0.0966, 0.0570], + [-0.0688, -0.0362, 0.0930, ..., 0.1652, -0.0592, -0.1556]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.8217e-06, -1.3001e-06, ..., 1.1176e-08, + 8.5682e-08, 7.4506e-09], + [ 0.0000e+00, 1.3039e-07, 4.1723e-07, ..., 4.4703e-08, + 7.7486e-07, 2.2352e-08], + [-8.9407e-08, 1.1548e-07, 2.7940e-07, ..., 3.7253e-09, + 8.1956e-08, 8.9407e-08], + ..., + [ 7.4506e-09, 4.3213e-07, 1.6801e-06, ..., 5.1409e-07, + 7.8231e-08, 5.5879e-08], + [ 9.6858e-08, 1.5274e-07, 1.9111e-06, ..., 3.9488e-07, + 2.9802e-08, 1.2666e-07], + [ 5.5879e-08, 2.3097e-07, -7.0967e-06, ..., -1.3523e-06, + 2.9057e-07, 1.1548e-07]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0306, 0.0064, 0.0180, 0.0229, 0.0346, 0.0051, -0.0020, -0.0005, + 0.0027, -0.0392], device='cuda:0'), grad: tensor([-2.2501e-06, 9.1344e-06, 9.7230e-07, -4.9546e-07, -1.1131e-05, + 1.3784e-06, 3.2708e-06, 3.6135e-06, 4.7684e-06, -9.3728e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 217.77, cls_loss 0.0026 cls_loss_mapping 0.0035 cls_loss_causal 0.5223 re_mapping 0.0062 re_causal 0.0179 /// teacc 98.87 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1344, 0.1770, -0.0320, ..., -0.0120, 0.0732, -0.0026], + [ 0.1453, -0.1034, 0.0294, ..., -0.0593, -0.0142, -0.0689], + [-0.0654, -0.1788, -0.0057, ..., -0.0401, 0.0439, -0.1756], + ..., + [-0.1513, -0.1866, 0.0530, ..., -0.0843, -0.0710, -0.1216], + [-0.0796, 0.0104, 0.0303, ..., 0.0269, -0.0966, 0.0565], + [-0.0691, -0.0369, 0.0933, ..., 0.1663, -0.0593, -0.1562]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -2.2538e-06, 2.2352e-08, ..., 3.7253e-09, + 0.0000e+00, 5.5879e-08], + [-5.9232e-07, 7.4506e-09, -6.4448e-07, ..., 3.7253e-09, + 0.0000e+00, 1.1176e-08], + [ 1.7509e-07, 1.6764e-07, 2.4587e-07, ..., 0.0000e+00, + 0.0000e+00, 1.6764e-07], + ..., + [ 1.8999e-07, 5.2154e-08, 1.1176e-07, ..., 2.6077e-08, + 0.0000e+00, 2.2352e-08], + [ 1.4976e-06, -2.8312e-07, 7.4506e-09, ..., 5.9605e-08, + 0.0000e+00, 2.3246e-06], + [ 7.8231e-08, 8.8289e-07, -3.3528e-07, ..., -1.3411e-07, + 0.0000e+00, 1.8999e-07]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0314, 0.0051, 0.0193, 0.0229, 0.0341, 0.0057, -0.0025, -0.0008, + 0.0028, -0.0389], device='cuda:0'), grad: tensor([-2.4736e-06, -5.4389e-07, 1.0692e-06, -5.8897e-06, -7.7039e-06, + 2.2985e-06, 8.5384e-06, 8.3447e-07, 1.8924e-06, 1.8552e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 218.08, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5409 re_mapping 0.0062 re_causal 0.0182 /// teacc 98.92 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1355, 0.1778, -0.0320, ..., -0.0120, 0.0732, -0.0030], + [ 0.1458, -0.1033, 0.0297, ..., -0.0594, -0.0142, -0.0688], + [-0.0658, -0.1795, -0.0059, ..., -0.0402, 0.0439, -0.1758], + ..., + [-0.1516, -0.1873, 0.0530, ..., -0.0843, -0.0710, -0.1220], + [-0.0801, 0.0100, 0.0302, ..., 0.0269, -0.0966, 0.0565], + [-0.0694, -0.0372, 0.0933, ..., 0.1665, -0.0593, -0.1568]], + device='cuda:0'), grad: tensor([[ 2.2352e-07, -2.3842e-07, -1.0058e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 5.4061e-05, 2.2352e-08, 3.3192e-06, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [-5.5820e-05, 1.4901e-08, -1.5162e-06, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.5460e-06, 7.4506e-09, -2.0377e-06, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 4.0233e-07, 1.1176e-07, 5.2154e-08, ..., 0.0000e+00, + 0.0000e+00, 1.9744e-07], + [ 1.4529e-07, 1.5274e-07, 2.1234e-07, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0317, 0.0050, 0.0188, 0.0228, 0.0343, 0.0060, -0.0027, -0.0007, + 0.0023, -0.0390], device='cuda:0'), grad: tensor([ 2.4587e-07, 1.1939e-04, -1.2046e-04, 3.6508e-07, -1.5683e-06, + 8.8289e-06, -1.0483e-05, 4.3213e-07, 1.0692e-06, 2.4028e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 217.95, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5105 re_mapping 0.0059 re_causal 0.0176 /// teacc 98.92 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1358, 0.1783, -0.0319, ..., -0.0119, 0.0732, -0.0033], + [ 0.1457, -0.1036, 0.0295, ..., -0.0595, -0.0142, -0.0689], + [-0.0661, -0.1800, -0.0060, ..., -0.0402, 0.0439, -0.1761], + ..., + [-0.1516, -0.1890, 0.0530, ..., -0.0844, -0.0710, -0.1230], + [-0.0818, 0.0081, 0.0297, ..., 0.0270, -0.0967, 0.0560], + [-0.0700, -0.0377, 0.0933, ..., 0.1666, -0.0593, -0.1577]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, -2.1793e-06, -9.2015e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [-1.4529e-07, 1.8626e-08, 4.4703e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 3.7253e-08, 1.7136e-07, 3.9861e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-07], + ..., + [-5.1782e-07, 1.8626e-08, -2.2054e-06, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 9.3132e-08, -1.8626e-07, 1.1176e-07, ..., -3.7253e-09, + 0.0000e+00, -4.6939e-07], + [ 1.5646e-07, 1.6578e-06, 9.6858e-07, ..., -3.7253e-09, + 0.0000e+00, 8.9407e-08]], device='cuda:0') +Epoch 215, bias, value: tensor([ 0.0317, 0.0043, 0.0183, 0.0228, 0.0345, 0.0068, -0.0014, -0.0006, + 0.0011, -0.0391], device='cuda:0'), grad: tensor([-4.0531e-06, 3.2037e-07, 3.6806e-06, 1.2740e-06, 3.2745e-06, + 3.6545e-06, -1.8626e-08, -5.5991e-06, 2.1420e-06, -4.7497e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 218.16, cls_loss 0.0018 cls_loss_mapping 0.0038 cls_loss_causal 0.5449 re_mapping 0.0063 re_causal 0.0189 /// teacc 99.00 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1360, 0.1777, -0.0319, ..., -0.0119, 0.0732, -0.0038], + [ 0.1463, -0.1033, 0.0297, ..., -0.0595, -0.0142, -0.0689], + [-0.0663, -0.1795, -0.0068, ..., -0.0400, 0.0439, -0.1760], + ..., + [-0.1518, -0.1891, 0.0531, ..., -0.0844, -0.0710, -0.1232], + [-0.0828, 0.0072, 0.0296, ..., 0.0267, -0.0967, 0.0556], + [-0.0704, -0.0380, 0.0933, ..., 0.1670, -0.0593, -0.1579]], + device='cuda:0'), grad: tensor([[ 2.0862e-07, -5.5879e-08, 9.3132e-08, ..., 7.4506e-09, + 0.0000e+00, 1.8999e-07], + [-4.5076e-07, 0.0000e+00, 3.7998e-07, ..., 7.4506e-09, + 0.0000e+00, 4.8429e-08], + [ 1.3411e-07, 4.4703e-08, -3.8110e-06, ..., 1.1176e-08, + 0.0000e+00, 2.8312e-07], + ..., + [ 1.2293e-07, 7.4506e-09, 1.1548e-06, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 3.3155e-07, -1.3039e-07, -1.2293e-07, ..., -2.7567e-07, + 0.0000e+00, -2.3581e-06], + [ 4.4703e-08, 3.3528e-08, 1.5870e-06, ..., 1.1548e-07, + 0.0000e+00, 9.7230e-07]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0308, 0.0046, 0.0184, 0.0227, 0.0348, 0.0071, -0.0007, -0.0005, + 0.0004, -0.0392], device='cuda:0'), grad: tensor([ 3.2485e-06, 1.1623e-06, -8.3745e-06, 1.5832e-06, 5.5134e-07, + 1.1522e-04, -1.2457e-04, 2.7232e-06, 3.6322e-06, 4.7497e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 218.27, cls_loss 0.0020 cls_loss_mapping 0.0029 cls_loss_causal 0.5048 re_mapping 0.0062 re_causal 0.0172 /// teacc 98.92 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1356, 0.1773, -0.0319, ..., -0.0129, 0.0734, -0.0040], + [ 0.1470, -0.1035, 0.0297, ..., -0.0596, -0.0142, -0.0690], + [-0.0670, -0.1815, -0.0099, ..., -0.0401, 0.0439, -0.1771], + ..., + [-0.1522, -0.1882, 0.0534, ..., -0.0845, -0.0710, -0.1233], + [-0.0831, 0.0071, 0.0296, ..., 0.0265, -0.0969, 0.0556], + [-0.0705, -0.0387, 0.0933, ..., 0.1677, -0.0593, -0.1583]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.1548e-07, -5.2154e-08, ..., 3.7253e-09, + -3.7253e-09, 3.7253e-09], + [ 3.3155e-07, 5.2154e-08, 5.2154e-08, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08], + [-2.6338e-06, 1.1176e-08, -2.8312e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + ..., + [ 2.2128e-06, 7.4506e-09, 2.6822e-07, ..., 3.7253e-09, + 0.0000e+00, 1.1176e-08], + [-4.5449e-07, -2.3097e-06, -8.9407e-07, ..., 0.0000e+00, + 0.0000e+00, -1.1474e-06], + [ 3.7253e-09, 1.8626e-08, -3.3528e-08, ..., -1.1176e-08, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 217, bias, value: tensor([ 2.9756e-02, 4.7990e-03, 1.5709e-02, 2.2693e-02, 3.4505e-02, + 7.0315e-03, 8.0583e-04, -4.1769e-06, 2.2376e-04, -3.9150e-02], + device='cuda:0'), grad: tensor([ 4.0978e-07, 1.7732e-06, -1.0341e-05, 1.0431e-07, -9.4399e-06, + 3.2112e-06, 8.6129e-06, 8.9929e-06, -3.6061e-06, 2.0862e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 216---------------------------------------------------- +epoch 216, time 217.83, cls_loss 0.0022 cls_loss_mapping 0.0029 cls_loss_causal 0.5412 re_mapping 0.0058 re_causal 0.0178 /// teacc 99.08 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1358, 0.1773, -0.0320, ..., -0.0130, 0.0735, -0.0046], + [ 0.1486, -0.1013, 0.0302, ..., -0.0604, -0.0142, -0.0684], + [-0.0676, -0.1820, -0.0102, ..., -0.0402, 0.0439, -0.1776], + ..., + [-0.1524, -0.1893, 0.0542, ..., -0.0846, -0.0710, -0.1237], + [-0.0847, 0.0063, 0.0283, ..., 0.0264, -0.0969, 0.0553], + [-0.0705, -0.0389, 0.0926, ..., 0.1682, -0.0593, -0.1586]], + device='cuda:0'), grad: tensor([[-1.2219e-06, 5.4312e-04, -1.4529e-07, ..., -8.1956e-08, + 0.0000e+00, -1.0714e-05], + [-1.6600e-05, 3.9488e-07, -1.5706e-05, ..., 3.7253e-09, + 0.0000e+00, -4.2431e-06], + [ 3.7625e-07, 4.2841e-07, 4.3586e-07, ..., 2.9802e-08, + 0.0000e+00, 1.4901e-07], + ..., + [ 1.4111e-05, 1.0282e-06, 1.2800e-05, ..., 7.4506e-09, + 0.0000e+00, 3.9972e-06], + [ 4.0717e-06, 6.1132e-06, 2.9057e-07, ..., -2.1979e-07, + 0.0000e+00, 2.4848e-06], + [ 3.5390e-07, 2.5295e-06, 8.6054e-07, ..., 2.2724e-07, + 0.0000e+00, 5.9605e-07]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0296, 0.0055, 0.0156, 0.0231, 0.0344, 0.0069, 0.0005, 0.0011, + -0.0010, -0.0401], device='cuda:0'), grad: tensor([ 1.8463e-03, -2.9728e-05, 1.6764e-06, 4.1723e-06, 3.6992e-06, + 4.6849e-05, -1.9312e-03, 2.7165e-05, 2.1562e-05, 8.1733e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 217.01, cls_loss 0.0024 cls_loss_mapping 0.0025 cls_loss_causal 0.5401 re_mapping 0.0058 re_causal 0.0170 /// teacc 99.01 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1365, 0.1767, -0.0320, ..., -0.0131, 0.0734, -0.0049], + [ 0.1476, -0.1005, 0.0279, ..., -0.0603, -0.0141, -0.0682], + [-0.0681, -0.1826, -0.0102, ..., -0.0403, 0.0439, -0.1781], + ..., + [-0.1507, -0.1895, 0.0557, ..., -0.0850, -0.0710, -0.1240], + [-0.0851, 0.0061, 0.0279, ..., 0.0262, -0.0971, 0.0552], + [-0.0707, -0.0393, 0.0928, ..., 0.1698, -0.0593, -0.1591]], + device='cuda:0'), grad: tensor([[ 5.9605e-08, -7.4506e-08, 3.3528e-08, ..., 7.4506e-09, + 0.0000e+00, 1.4901e-08], + [-2.3469e-06, 1.4901e-08, -1.2815e-06, ..., 1.1176e-08, + 0.0000e+00, 2.2352e-08], + [ 1.8403e-06, 1.2666e-07, 1.1586e-06, ..., 7.4506e-09, + 0.0000e+00, 1.5646e-07], + ..., + [ 6.3330e-08, 7.4506e-09, 1.5646e-07, ..., 5.2154e-08, + 0.0000e+00, 1.1176e-08], + [ 1.5646e-07, -2.6450e-07, -3.1665e-07, ..., -4.8429e-08, + 0.0000e+00, -2.9802e-07], + [ 6.7055e-08, 1.3784e-07, -3.1665e-07, ..., -1.1548e-07, + 0.0000e+00, 1.1548e-07]], device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0287, 0.0036, 0.0155, 0.0230, 0.0335, 0.0071, 0.0010, 0.0024, + -0.0016, -0.0398], device='cuda:0'), grad: tensor([ 3.6098e-06, 1.4067e-05, 5.3160e-06, 3.5800e-06, -6.0272e-04, + -8.6054e-07, 2.9765e-06, 7.9155e-05, 1.1183e-05, 4.8256e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 218---------------------------------------------------- +epoch 218, time 218.13, cls_loss 0.0024 cls_loss_mapping 0.0034 cls_loss_causal 0.5025 re_mapping 0.0058 re_causal 0.0172 /// teacc 99.09 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1366, 0.1771, -0.0338, ..., -0.0165, 0.0734, -0.0052], + [ 0.1484, -0.1010, 0.0280, ..., -0.0604, -0.0141, -0.0682], + [-0.0684, -0.1835, -0.0103, ..., -0.0409, 0.0439, -0.1784], + ..., + [-0.1513, -0.1892, 0.0569, ..., -0.0853, -0.0710, -0.1240], + [-0.0853, 0.0059, 0.0279, ..., 0.0248, -0.0971, 0.0551], + [-0.0700, -0.0377, 0.0917, ..., 0.1722, -0.0593, -0.1569]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -3.5167e-05, -1.6019e-07, ..., 2.9802e-08, + 0.0000e+00, 1.1176e-08], + [-6.3330e-08, 2.2352e-08, 2.0489e-07, ..., 2.9802e-07, + 0.0000e+00, 4.0978e-08], + [-1.1176e-08, -1.4901e-07, 4.6939e-07, ..., 4.1723e-07, + 0.0000e+00, 5.9605e-08], + ..., + [ 2.9802e-08, 2.9802e-08, 1.1176e-07, ..., 1.0431e-07, + 0.0000e+00, 1.1176e-08], + [ 1.4901e-08, 8.4564e-07, -1.1921e-06, ..., -1.3746e-06, + 0.0000e+00, -1.9744e-07], + [ 7.4506e-09, 4.0606e-07, 7.4506e-08, ..., 7.8231e-08, + 0.0000e+00, 1.4901e-08]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0277, 0.0037, 0.0159, 0.0226, 0.0345, 0.0072, 0.0008, 0.0039, + -0.0019, -0.0413], device='cuda:0'), grad: tensor([-7.7128e-05, 1.5050e-06, 2.1681e-06, 2.9542e-06, -1.8924e-06, + 1.5907e-06, 7.2181e-05, 6.4448e-07, -5.1782e-06, 3.0436e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 217.26, cls_loss 0.0024 cls_loss_mapping 0.0038 cls_loss_causal 0.5381 re_mapping 0.0057 re_causal 0.0171 /// teacc 98.99 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1390, 0.1778, -0.0338, ..., -0.0173, 0.0734, -0.0027], + [ 0.1486, -0.1015, 0.0279, ..., -0.0606, -0.0141, -0.0683], + [-0.0687, -0.1855, -0.0106, ..., -0.0423, 0.0439, -0.1791], + ..., + [-0.1514, -0.1908, 0.0570, ..., -0.0856, -0.0710, -0.1244], + [-0.0855, 0.0056, 0.0279, ..., 0.0227, -0.0969, 0.0552], + [-0.0691, -0.0370, 0.0919, ..., 0.1745, -0.0593, -0.1546]], + device='cuda:0'), grad: tensor([[ 2.6077e-08, 3.1292e-07, 6.0722e-07, ..., 3.7253e-09, + 0.0000e+00, 3.5390e-07], + [-1.4901e-08, 1.1176e-08, 1.7881e-07, ..., 3.7253e-08, + 0.0000e+00, 3.3528e-08], + [ 7.4506e-09, 1.3784e-07, 3.0547e-07, ..., 3.3528e-08, + 0.0000e+00, 2.8685e-07], + ..., + [ 3.3528e-08, 2.2352e-08, -7.9721e-07, ..., 4.4703e-08, + 0.0000e+00, 1.4901e-07], + [ 3.3528e-07, -8.4937e-07, -1.8105e-06, ..., -7.4506e-09, + 0.0000e+00, 1.1548e-07], + [ 1.8142e-06, 1.6838e-06, 1.0766e-06, ..., 1.3709e-06, + 0.0000e+00, 2.5146e-06]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0263, 0.0035, 0.0156, 0.0227, 0.0339, 0.0059, 0.0027, 0.0039, + -0.0022, -0.0410], device='cuda:0'), grad: tensor([ 1.5013e-06, 7.5251e-07, 9.6112e-07, -5.8264e-06, -1.5467e-05, + -2.8126e-06, 6.9290e-07, -5.7742e-07, -1.7136e-06, 2.2396e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 217.79, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5154 re_mapping 0.0058 re_causal 0.0170 /// teacc 98.86 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1391, 0.1784, -0.0336, ..., -0.0172, 0.0734, -0.0029], + [ 0.1489, -0.1021, 0.0278, ..., -0.0613, -0.0141, -0.0685], + [-0.0692, -0.1857, -0.0107, ..., -0.0423, 0.0440, -0.1792], + ..., + [-0.1517, -0.1913, 0.0571, ..., -0.0858, -0.0710, -0.1245], + [-0.0858, 0.0055, 0.0278, ..., 0.0221, -0.0969, 0.0551], + [-0.0696, -0.0374, 0.0919, ..., 0.1753, -0.0593, -0.1547]], + device='cuda:0'), grad: tensor([[ 3.8370e-07, -1.4901e-07, 5.4017e-07, ..., -7.4506e-09, + 0.0000e+00, 2.6077e-08], + [ 9.4995e-07, 1.1176e-08, -1.8626e-08, ..., 3.7253e-09, + 0.0000e+00, 7.4506e-09], + [-1.3039e-06, -7.4506e-08, 2.1607e-07, ..., -5.5879e-08, + 0.0000e+00, 3.7253e-09], + ..., + [ 3.3900e-07, 1.1176e-08, 4.4703e-07, ..., 3.3528e-08, + 0.0000e+00, 1.1176e-08], + [ 4.2096e-07, -6.3330e-08, 3.4273e-07, ..., 1.4901e-08, + 0.0000e+00, -8.5682e-08], + [-2.9892e-05, 1.4529e-07, -5.4926e-05, ..., 5.9530e-06, + 0.0000e+00, -5.5879e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0267, 0.0031, 0.0156, 0.0226, 0.0335, 0.0059, 0.0028, 0.0040, + -0.0026, -0.0410], device='cuda:0'), grad: tensor([ 2.7493e-06, 6.0424e-06, -6.7502e-06, 5.6624e-07, 1.1221e-05, + 4.0442e-05, 1.5163e-04, 2.1420e-06, 2.1234e-06, -2.1017e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 217.25, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5434 re_mapping 0.0060 re_causal 0.0178 /// teacc 98.98 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1391, 0.1789, -0.0335, ..., -0.0172, 0.0734, -0.0032], + [ 0.1491, -0.1020, 0.0278, ..., -0.0611, -0.0140, -0.0685], + [-0.0694, -0.1861, -0.0098, ..., -0.0425, 0.0440, -0.1797], + ..., + [-0.1519, -0.1923, 0.0571, ..., -0.0861, -0.0710, -0.1250], + [-0.0871, 0.0040, 0.0274, ..., 0.0217, -0.0969, 0.0538], + [-0.0697, -0.0377, 0.0920, ..., 0.1755, -0.0593, -0.1549]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, -4.6790e-05, -2.0228e-06, ..., -1.6019e-07, + -2.2352e-07, -1.0431e-06], + [ 4.6194e-07, 1.2442e-06, 3.1888e-06, ..., 6.7055e-08, + 7.0781e-08, 3.5390e-07], + [ 2.2352e-08, 1.6876e-06, 2.1607e-07, ..., 1.8626e-08, + 3.7253e-09, 4.4703e-08], + ..., + [-5.6624e-07, 4.5076e-07, -3.5428e-06, ..., 1.0058e-07, + 3.7253e-09, 4.8429e-08], + [ 2.0862e-07, 9.9093e-07, -1.1846e-06, ..., -5.0664e-07, + 1.1176e-07, -1.4007e-06], + [ 4.0978e-08, 1.4938e-05, 2.2389e-06, ..., 4.5821e-07, + 7.4506e-09, 1.7434e-06]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0268, 0.0031, 0.0167, 0.0228, 0.0348, 0.0068, 0.0024, 0.0039, + -0.0043, -0.0412], device='cuda:0'), grad: tensor([-1.0502e-04, 7.0706e-06, 3.8743e-06, 3.3528e-06, -4.8243e-06, + 2.2113e-05, 3.5912e-05, -4.3698e-06, 1.1250e-06, 4.0621e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 217.44, cls_loss 0.0019 cls_loss_mapping 0.0032 cls_loss_causal 0.5061 re_mapping 0.0057 re_causal 0.0172 /// teacc 99.02 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1390, 0.1802, -0.0328, ..., -0.0168, 0.0734, -0.0016], + [ 0.1492, -0.1022, 0.0276, ..., -0.0646, -0.0140, -0.0686], + [-0.0697, -0.1871, -0.0105, ..., -0.0427, 0.0440, -0.1802], + ..., + [-0.1520, -0.1941, 0.0582, ..., -0.0842, -0.0710, -0.1256], + [-0.0875, 0.0037, 0.0274, ..., 0.0218, -0.0969, 0.0537], + [-0.0686, -0.0391, 0.0908, ..., 0.1746, -0.0593, -0.1556]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -6.2957e-07, -2.5332e-07, ..., 5.5507e-07, + 0.0000e+00, 2.7530e-06], + [-1.0133e-06, 0.0000e+00, -1.1101e-06, ..., -1.1176e-08, + 0.0000e+00, 8.9407e-08], + [ 3.3528e-08, 1.1548e-07, -4.1351e-06, ..., 8.5682e-08, + 0.0000e+00, -9.3877e-07], + ..., + [ 8.5682e-07, 6.3330e-08, 1.0692e-06, ..., 5.2154e-08, + 0.0000e+00, 1.1921e-07], + [ 1.1176e-08, 4.2096e-07, 3.3937e-06, ..., 4.2841e-07, + 0.0000e+00, 3.4645e-06], + [ 5.2154e-08, 2.5705e-07, 3.3155e-07, ..., -3.7253e-08, + 0.0000e+00, 4.2096e-07]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0273, 0.0030, 0.0166, 0.0228, 0.0350, 0.0070, 0.0022, 0.0050, + -0.0046, -0.0424], device='cuda:0'), grad: tensor([ 7.4953e-06, -5.4762e-07, -1.8984e-05, -1.6153e-05, 1.2815e-06, + 1.5378e-05, 1.7546e-06, -3.0488e-05, 2.4378e-05, 1.5780e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 217.36, cls_loss 0.0018 cls_loss_mapping 0.0027 cls_loss_causal 0.5411 re_mapping 0.0058 re_causal 0.0178 /// teacc 98.87 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1391, 0.1808, -0.0326, ..., -0.0168, 0.0741, -0.0019], + [ 0.1499, -0.1027, 0.0278, ..., -0.0645, -0.0140, -0.0689], + [-0.0703, -0.1884, -0.0102, ..., -0.0427, 0.0440, -0.1818], + ..., + [-0.1529, -0.1966, 0.0581, ..., -0.0843, -0.0710, -0.1265], + [-0.0872, 0.0043, 0.0283, ..., 0.0216, -0.0970, 0.0547], + [-0.0688, -0.0393, 0.0908, ..., 0.1747, -0.0593, -0.1557]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.3821e-06, -5.3272e-07, ..., -4.4703e-08, + 0.0000e+00, 0.0000e+00], + [-8.1956e-08, 1.8626e-08, 2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 2.6077e-08, 2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.9802e-08, 2.9802e-08, -5.5507e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-08, 1.4156e-07, 1.6019e-07, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 3.7253e-09, 7.6368e-07, 2.4587e-07, ..., -4.0978e-08, + 0.0000e+00, -2.6077e-08]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0275, 0.0031, 0.0166, 0.0232, 0.0352, 0.0066, 0.0019, 0.0047, + -0.0036, -0.0425], device='cuda:0'), grad: tensor([-1.9334e-06, 1.9372e-07, -9.4995e-07, 9.5740e-07, -4.4703e-07, + 4.4703e-07, 1.1921e-07, -1.5944e-06, 1.1586e-06, 2.0228e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 217.41, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4965 re_mapping 0.0054 re_causal 0.0174 /// teacc 99.00 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1391, 0.1811, -0.0324, ..., -0.0168, 0.0749, -0.0015], + [ 0.1505, -0.1021, 0.0281, ..., -0.0642, -0.0140, -0.0686], + [-0.0704, -0.1888, -0.0106, ..., -0.0427, 0.0440, -0.1823], + ..., + [-0.1531, -0.1971, 0.0582, ..., -0.0844, -0.0710, -0.1268], + [-0.0874, 0.0044, 0.0280, ..., 0.0206, -0.0971, 0.0549], + [-0.0693, -0.0395, 0.0908, ..., 0.1749, -0.0594, -0.1560]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -1.3039e-07, -5.2154e-08, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-07], + [-2.5332e-07, 3.7253e-09, -2.0117e-07, ..., 3.7253e-09, + 0.0000e+00, 2.1234e-07], + [ 1.4901e-08, 1.8626e-08, 7.4506e-08, ..., 3.7253e-09, + 0.0000e+00, 8.7917e-07], + ..., + [ 1.6019e-07, 1.1176e-08, 9.3132e-08, ..., 1.4901e-08, + 0.0000e+00, 3.5018e-07], + [ 8.5682e-08, -1.8626e-08, 2.5332e-07, ..., 5.5879e-08, + 0.0000e+00, 1.7919e-06], + [ 4.4703e-08, 7.8231e-08, -7.1153e-07, ..., -1.8626e-07, + 0.0000e+00, 1.2293e-07]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0275, 0.0034, 0.0165, 0.0230, 0.0354, 0.0066, 0.0018, 0.0048, + -0.0037, -0.0426], device='cuda:0'), grad: tensor([ 8.4937e-07, 4.2096e-07, 2.8238e-06, -1.2308e-05, -2.1271e-06, + 5.1782e-07, 2.3842e-07, 1.5274e-06, 6.2138e-06, 1.7993e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 217.66, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.5172 re_mapping 0.0059 re_causal 0.0183 /// teacc 98.97 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1391, 0.1813, -0.0325, ..., -0.0168, 0.0750, -0.0018], + [ 0.1513, -0.1008, 0.0282, ..., -0.0653, -0.0140, -0.0675], + [-0.0708, -0.1891, -0.0107, ..., -0.0428, 0.0440, -0.1826], + ..., + [-0.1534, -0.1977, 0.0582, ..., -0.0846, -0.0710, -0.1270], + [-0.0876, 0.0043, 0.0280, ..., 0.0205, -0.0971, 0.0549], + [-0.0692, -0.0396, 0.0908, ..., 0.1752, -0.0594, -0.1560]], + device='cuda:0'), grad: tensor([[ 1.7509e-07, 0.0000e+00, 2.1979e-07, ..., 2.6077e-08, + 0.0000e+00, 1.1176e-08], + [-1.7397e-06, 0.0000e+00, -2.3767e-06, ..., -3.6880e-07, + 0.0000e+00, 0.0000e+00], + [ 2.9802e-08, 6.3330e-08, 1.3784e-07, ..., 1.1176e-08, + 0.0000e+00, 3.7253e-08], + ..., + [ 1.9744e-07, 3.7253e-09, 1.5423e-06, ..., 2.8312e-07, + 0.0000e+00, 3.7253e-09], + [ 1.9372e-07, -2.9430e-07, -2.2352e-08, ..., 4.0978e-08, + 0.0000e+00, -1.8626e-07], + [ 2.2724e-07, 1.4901e-08, 6.3330e-08, ..., -5.2154e-08, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0276, 0.0036, 0.0165, 0.0230, 0.0358, 0.0065, 0.0016, 0.0047, + -0.0038, -0.0426], device='cuda:0'), grad: tensor([ 6.0722e-07, -3.3788e-06, 1.7881e-07, 4.4703e-08, 2.7269e-06, + 1.7658e-06, 5.9232e-07, -1.7732e-06, -9.4622e-07, 1.7509e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 217.51, cls_loss 0.0015 cls_loss_mapping 0.0033 cls_loss_causal 0.5335 re_mapping 0.0060 re_causal 0.0181 /// teacc 98.94 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.1391, 0.1818, -0.0324, ..., -0.0168, 0.0752, -0.0024], + [ 0.1525, -0.1008, 0.0288, ..., -0.0654, -0.0140, -0.0676], + [-0.0709, -0.1895, -0.0103, ..., -0.0426, 0.0440, -0.1821], + ..., + [-0.1548, -0.1989, 0.0579, ..., -0.0849, -0.0710, -0.1278], + [-0.0877, 0.0046, 0.0279, ..., 0.0202, -0.0974, 0.0554], + [-0.0692, -0.0398, 0.0909, ..., 0.1756, -0.0594, -0.1561]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -1.8254e-07, 6.3330e-08, ..., -7.4506e-09, + 0.0000e+00, -3.7253e-09], + [-1.4901e-07, 0.0000e+00, 6.4448e-07, ..., 3.7253e-09, + 0.0000e+00, 7.4506e-09], + [-4.0978e-08, 1.1176e-08, 1.6727e-06, ..., 3.7253e-09, + 0.0000e+00, 1.8626e-08], + ..., + [ 6.3330e-08, 3.7253e-09, -1.0364e-05, ..., 2.6077e-08, + 0.0000e+00, 2.2352e-08], + [ 5.9605e-08, -3.7253e-09, 2.0862e-07, ..., -1.8626e-08, + 0.0000e+00, -5.5879e-08], + [ 7.4506e-09, 1.0058e-07, 4.3809e-06, ..., -1.2293e-07, + 0.0000e+00, -2.2352e-08]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0277, 0.0043, 0.0170, 0.0230, 0.0362, 0.0063, 0.0015, 0.0044, + -0.0039, -0.0427], device='cuda:0'), grad: tensor([-3.7253e-09, 1.4640e-06, 1.9930e-06, 4.7088e-06, 5.2527e-07, + 9.6858e-08, 1.2293e-07, -1.6615e-05, 3.5763e-07, 7.3090e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 217.35, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5016 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.96 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.1392, 0.1825, -0.0327, ..., -0.0170, 0.0756, -0.0031], + [ 0.1526, -0.1010, 0.0285, ..., -0.0665, -0.0140, -0.0677], + [-0.0712, -0.1905, -0.0107, ..., -0.0430, 0.0440, -0.1831], + ..., + [-0.1551, -0.2001, 0.0577, ..., -0.0864, -0.0710, -0.1282], + [-0.0877, 0.0048, 0.0281, ..., 0.0207, -0.0975, 0.0558], + [-0.0667, -0.0408, 0.0915, ..., 0.1771, -0.0595, -0.1562]], + device='cuda:0'), grad: tensor([[ 1.1548e-07, -3.0547e-07, -3.2410e-07, ..., -1.5646e-07, + 0.0000e+00, 2.9802e-08], + [-7.9393e-05, 4.4703e-08, -9.9421e-05, ..., 1.4901e-08, + 0.0000e+00, -5.2154e-08], + [ 6.1467e-07, 2.2352e-08, 6.0350e-07, ..., 7.4506e-09, + 0.0000e+00, 1.4901e-08], + ..., + [ 7.7367e-05, 2.6077e-08, 9.6917e-05, ..., 1.4901e-08, + 0.0000e+00, 1.0431e-07], + [ 2.5332e-07, -4.9174e-07, -4.6566e-07, ..., 1.1176e-08, + 0.0000e+00, -1.9781e-06], + [ 1.3411e-06, 4.9546e-07, 1.3411e-06, ..., 1.6019e-07, + 0.0000e+00, 2.2352e-07]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0276, 0.0039, 0.0166, 0.0232, 0.0362, 0.0059, 0.0012, 0.0042, + -0.0035, -0.0421], device='cuda:0'), grad: tensor([-3.9861e-07, -1.3602e-04, 9.8348e-07, 1.4082e-06, -6.1840e-07, + 1.8999e-07, 4.4703e-07, 1.3292e-04, -2.2091e-06, 3.1404e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 217.29, cls_loss 0.0020 cls_loss_mapping 0.0025 cls_loss_causal 0.4974 re_mapping 0.0058 re_causal 0.0166 /// teacc 98.95 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.1393, 0.1839, -0.0317, ..., -0.0170, 0.0752, -0.0030], + [ 0.1530, -0.1031, 0.0284, ..., -0.0666, -0.0139, -0.0680], + [-0.0716, -0.1909, -0.0109, ..., -0.0432, 0.0443, -0.1840], + ..., + [-0.1554, -0.2014, 0.0577, ..., -0.0867, -0.0711, -0.1287], + [-0.0867, 0.0068, 0.0295, ..., 0.0205, -0.0977, 0.0583], + [-0.0674, -0.0414, 0.0916, ..., 0.1780, -0.0596, -0.1572]], + device='cuda:0'), grad: tensor([[-1.1735e-06, -1.0855e-05, -5.9679e-06, ..., 7.4506e-09, + 0.0000e+00, 5.2527e-07], + [-3.5390e-07, 1.8254e-07, -1.4901e-07, ..., 7.4506e-09, + 0.0000e+00, 2.5332e-07], + [ 3.7253e-08, 1.0468e-06, 9.4995e-07, ..., 1.8626e-08, + 0.0000e+00, 1.2331e-06], + ..., + [ 2.3842e-07, 1.4901e-07, 3.6135e-07, ..., 1.4901e-08, + 0.0000e+00, 3.0920e-07], + [ 4.4331e-07, -2.7828e-06, -3.6918e-06, ..., -2.1607e-07, + 0.0000e+00, -3.7067e-06], + [ 3.9861e-07, 1.5236e-06, 1.3635e-06, ..., 9.3132e-08, + 0.0000e+00, 1.4491e-06]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0283, 0.0035, 0.0168, 0.0222, 0.0355, 0.0054, 0.0010, 0.0041, + -0.0016, -0.0419], device='cuda:0'), grad: tensor([-2.5541e-05, 1.0058e-07, 3.3379e-06, -1.9819e-06, -8.5086e-06, + 8.0094e-07, 2.6956e-05, 1.4603e-06, -9.1642e-06, 1.2472e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 217.57, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5017 re_mapping 0.0057 re_causal 0.0165 /// teacc 99.04 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.1392, 0.1844, -0.0316, ..., -0.0170, 0.0748, -0.0017], + [ 0.1549, -0.1031, 0.0294, ..., -0.0672, -0.0139, -0.0681], + [-0.0717, -0.1911, -0.0108, ..., -0.0433, 0.0443, -0.1841], + ..., + [-0.1573, -0.2031, 0.0573, ..., -0.0868, -0.0711, -0.1288], + [-0.0869, 0.0068, 0.0294, ..., 0.0206, -0.0980, 0.0584], + [-0.0680, -0.0420, 0.0916, ..., 0.1783, -0.0596, -0.1577]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -4.9435e-06, -3.2298e-06, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-08], + [-1.3262e-06, 1.7509e-07, -1.7509e-07, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08], + [ 6.3702e-07, 3.3528e-08, 3.2037e-07, ..., 0.0000e+00, + 0.0000e+00, 4.0978e-08], + ..., + [ 1.9744e-07, 1.1176e-08, -5.6997e-07, ..., 0.0000e+00, + 0.0000e+00, -1.2666e-07], + [ 1.5274e-07, 3.2410e-07, 3.5763e-07, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + [ 9.3132e-08, 2.2352e-07, 1.7509e-07, ..., 0.0000e+00, + 0.0000e+00, 4.8429e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0285, 0.0045, 0.0176, 0.0221, 0.0357, 0.0055, 0.0008, 0.0036, + -0.0018, -0.0420], device='cuda:0'), grad: tensor([-7.1488e-06, -9.9093e-07, 1.1176e-06, 5.6624e-07, -5.7742e-07, + 1.4156e-07, 6.0163e-06, -1.2517e-06, 1.3188e-06, 7.4133e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 217.65, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.5371 re_mapping 0.0056 re_causal 0.0171 /// teacc 98.97 lr 0.00010000 +Epoch 232, weight, value: tensor([[-1.3890e-01, 1.8534e-01, -3.1411e-02, ..., -1.6972e-02, + 7.4850e-02, -3.8673e-05], + [ 1.5564e-01, -1.0287e-01, 2.9838e-02, ..., -6.7360e-02, + -1.3871e-02, -6.7861e-02], + [-7.2265e-02, -1.9189e-01, -1.0495e-02, ..., -4.3444e-02, + 4.4265e-02, -1.8347e-01], + ..., + [-1.5781e-01, -2.0634e-01, 5.7087e-02, ..., -8.6872e-02, + -7.1058e-02, -1.3047e-01], + [-8.7106e-02, 7.2244e-03, 2.9683e-02, ..., 2.0566e-02, + -9.7959e-02, 5.8670e-02], + [-6.8655e-02, -4.2586e-02, 9.1630e-02, ..., 1.7845e-01, + -5.9590e-02, -1.5851e-01]], device='cuda:0'), grad: tensor([[ 7.4506e-09, -1.8012e-06, -9.3505e-07, ..., -1.9558e-07, + 0.0000e+00, 3.7253e-09], + [ 3.7253e-09, 8.1956e-08, 1.7695e-07, ..., 5.7742e-08, + 0.0000e+00, 2.6077e-08], + [ 1.8626e-09, 3.0734e-07, 1.8068e-07, ..., 5.0291e-08, + 0.0000e+00, 2.9802e-08], + ..., + [ 7.4506e-09, 1.7881e-07, 6.9663e-07, ..., 2.4401e-07, + 0.0000e+00, 1.0803e-07], + [-2.1048e-07, -1.0990e-07, 2.2352e-08, ..., 2.0303e-07, + 0.0000e+00, -1.5311e-06], + [ 2.0489e-08, 8.1956e-07, -3.5092e-06, ..., -1.4957e-06, + 0.0000e+00, 3.3528e-08]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0288, 0.0051, 0.0180, 0.0219, 0.0361, 0.0053, 0.0005, 0.0033, + -0.0015, -0.0422], device='cuda:0'), grad: tensor([-3.1199e-06, 5.8301e-07, -1.7762e-05, 4.3772e-07, 1.8150e-05, + 2.2240e-06, 9.0152e-07, 5.2974e-06, -9.8534e-07, -5.7593e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 217.29, cls_loss 0.0015 cls_loss_mapping 0.0018 cls_loss_causal 0.4773 re_mapping 0.0058 re_causal 0.0164 /// teacc 99.00 lr 0.00010000 +Epoch 233, weight, value: tensor([[-1.3898e-01, 1.8579e-01, -3.1305e-02, ..., -1.6965e-02, + 7.4955e-02, -1.1184e-04], + [ 1.5645e-01, -1.0264e-01, 3.0053e-02, ..., -6.7749e-02, + -1.3873e-02, -6.7900e-02], + [-7.2741e-02, -1.9261e-01, -1.0735e-02, ..., -4.3509e-02, + 4.4263e-02, -1.8399e-01], + ..., + [-1.5847e-01, -2.0787e-01, 5.7286e-02, ..., -8.6947e-02, + -7.1061e-02, -1.3102e-01], + [-8.7495e-02, 7.2850e-03, 2.9848e-02, ..., 2.0593e-02, + -9.7983e-02, 5.8648e-02], + [-6.9151e-02, -4.2844e-02, 9.1358e-02, ..., 1.7863e-01, + -5.9547e-02, -1.5873e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 2.8498e-07, ..., 0.0000e+00, + 0.0000e+00, 5.9605e-08], + [-1.6764e-08, 9.3132e-09, 3.8482e-06, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 3.7253e-09, -5.5693e-07, 6.0908e-07, ..., 0.0000e+00, + 0.0000e+00, -3.7625e-07], + ..., + [ 5.5879e-09, 3.4273e-07, -9.0897e-06, ..., 0.0000e+00, + 0.0000e+00, 2.4401e-07], + [ 3.7253e-09, 5.7742e-08, 5.4762e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 1.8626e-09, 3.1665e-08, 1.1884e-06, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0289, 0.0054, 0.0179, 0.0217, 0.0361, 0.0057, 0.0002, 0.0034, + -0.0014, -0.0425], device='cuda:0'), grad: tensor([ 1.5069e-06, 8.4490e-06, -5.1111e-06, 5.1744e-06, -4.8801e-06, + 1.8626e-07, 1.6578e-07, -1.5587e-05, 1.9893e-06, 8.1062e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 217.58, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5297 re_mapping 0.0057 re_causal 0.0172 /// teacc 98.99 lr 0.00010000 +Epoch 234, weight, value: tensor([[-1.3901e-01, 1.8618e-01, -3.1215e-02, ..., -1.7016e-02, + 7.4939e-02, -1.5087e-04], + [ 1.5764e-01, -1.0249e-01, 3.0785e-02, ..., -6.7723e-02, + -1.3873e-02, -6.7878e-02], + [-7.2829e-02, -1.9340e-01, -1.1334e-02, ..., -4.3629e-02, + 4.4263e-02, -1.8442e-01], + ..., + [-1.5962e-01, -2.0862e-01, 5.6891e-02, ..., -8.7054e-02, + -7.1061e-02, -1.3220e-01], + [-8.7836e-02, 7.0916e-03, 3.0938e-02, ..., 2.0906e-02, + -9.7985e-02, 5.8621e-02], + [-6.9548e-02, -4.2936e-02, 9.1370e-02, ..., 1.7848e-01, + -5.9530e-02, -1.5995e-01]], device='cuda:0'), grad: tensor([[ 7.4506e-09, -6.5193e-07, -6.7055e-08, ..., 0.0000e+00, + 0.0000e+00, -1.4715e-07], + [-4.6194e-07, -2.6077e-08, -1.9744e-07, ..., -3.7253e-09, + 0.0000e+00, 2.6077e-08], + [ 2.6077e-08, 1.1176e-08, -9.8720e-08, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + ..., + [ 5.4017e-08, 9.3132e-09, -1.0040e-06, ..., 1.8626e-09, + 0.0000e+00, 5.5879e-08], + [ 2.4401e-07, 1.6764e-08, 3.4831e-07, ..., 5.5879e-08, + 0.0000e+00, -1.2107e-07], + [ 2.5332e-07, 1.9558e-07, 3.5390e-08, ..., -6.7055e-08, + 0.0000e+00, 1.5832e-07]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0290, 0.0062, 0.0175, 0.0218, 0.0370, 0.0059, 0.0001, 0.0030, + -0.0011, -0.0428], device='cuda:0'), grad: tensor([-8.8476e-07, 5.4017e-08, -6.2212e-07, 1.1008e-06, -3.5018e-07, + 3.6322e-07, 8.5868e-07, -2.4699e-06, 1.3709e-06, 5.5879e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 217.09, cls_loss 0.0017 cls_loss_mapping 0.0034 cls_loss_causal 0.5035 re_mapping 0.0055 re_causal 0.0165 /// teacc 98.88 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.1390, 0.1868, -0.0310, ..., -0.0170, 0.0750, -0.0003], + [ 0.1577, -0.1026, 0.0307, ..., -0.0685, -0.0139, -0.0683], + [-0.0730, -0.1946, -0.0120, ..., -0.0438, 0.0443, -0.1857], + ..., + [-0.1598, -0.2098, 0.0569, ..., -0.0873, -0.0711, -0.1326], + [-0.0880, 0.0072, 0.0316, ..., 0.0207, -0.0980, 0.0591], + [-0.0685, -0.0433, 0.0914, ..., 0.1782, -0.0595, -0.1607]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.7121e-05, -1.0841e-06, ..., 2.7940e-08, + 0.0000e+00, 1.8626e-09], + [-5.5879e-09, 3.3528e-07, 1.2480e-07, ..., 3.1665e-08, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 3.2410e-07, 1.7323e-07, ..., 1.1176e-08, + 0.0000e+00, 1.8626e-09], + ..., + [ 7.4506e-09, 5.2154e-08, 3.6135e-07, ..., 1.2293e-07, + 0.0000e+00, 1.3039e-08], + [ 3.3528e-08, 2.1793e-07, 3.0547e-07, ..., 9.3132e-08, + 0.0000e+00, 2.2352e-08], + [ 1.3411e-07, 6.2026e-07, -2.6897e-06, ..., -9.9279e-07, + 0.0000e+00, 6.3330e-08]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0292, 0.0060, 0.0172, 0.0230, 0.0389, 0.0046, 0.0002, 0.0030, + -0.0008, -0.0433], device='cuda:0'), grad: tensor([-5.0724e-05, 1.3914e-06, 6.1467e-07, 5.0850e-07, 4.4033e-06, + 5.1036e-07, 4.5210e-05, 1.4659e-06, 1.3523e-06, -4.6827e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 217.56, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.5035 re_mapping 0.0063 re_causal 0.0182 /// teacc 98.87 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.1391, 0.1872, -0.0310, ..., -0.0170, 0.0750, -0.0005], + [ 0.1577, -0.1026, 0.0306, ..., -0.0687, -0.0139, -0.0685], + [-0.0733, -0.1955, -0.0119, ..., -0.0438, 0.0443, -0.1853], + ..., + [-0.1599, -0.2108, 0.0570, ..., -0.0873, -0.0711, -0.1329], + [-0.0881, 0.0072, 0.0315, ..., 0.0206, -0.0980, 0.0590], + [-0.0716, -0.0454, 0.0915, ..., 0.1779, -0.0595, -0.1633]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 2.6450e-07, 2.2538e-07, ..., 1.4901e-07, + 0.0000e+00, 2.6636e-07], + [-9.3132e-09, 2.2352e-08, 1.6764e-08, ..., 2.0489e-08, + 0.0000e+00, 4.2841e-08], + [ 3.7253e-09, 3.9116e-08, 2.6077e-08, ..., 1.1176e-08, + 0.0000e+00, 2.2911e-07], + ..., + [ 2.0489e-08, 1.6764e-08, 8.1956e-08, ..., 1.1176e-08, + 0.0000e+00, 3.7253e-08], + [ 1.1735e-07, -3.9116e-07, -1.6950e-07, ..., 1.3411e-07, + 0.0000e+00, 5.9605e-08], + [ 2.2352e-08, -1.5926e-06, -2.4848e-06, ..., -2.8852e-06, + 0.0000e+00, -3.4198e-06]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0293, 0.0058, 0.0177, 0.0215, 0.0397, 0.0063, 0.0001, 0.0031, + -0.0011, -0.0438], device='cuda:0'), grad: tensor([ 1.0170e-06, 1.6205e-07, 1.3467e-06, 6.9961e-06, 3.2485e-06, + -5.5507e-07, -9.3132e-08, 1.3970e-07, -2.6822e-07, -1.2018e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 217.44, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.5107 re_mapping 0.0059 re_causal 0.0177 /// teacc 98.87 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.1392, 0.1864, -0.0337, ..., -0.0189, 0.0749, -0.0002], + [ 0.1580, -0.1024, 0.0306, ..., -0.0689, -0.0138, -0.0684], + [-0.0734, -0.1961, -0.0118, ..., -0.0439, 0.0442, -0.1855], + ..., + [-0.1601, -0.2120, 0.0569, ..., -0.0874, -0.0711, -0.1337], + [-0.0883, 0.0072, 0.0321, ..., 0.0206, -0.0981, 0.0591], + [-0.0738, -0.0459, 0.0921, ..., 0.1798, -0.0596, -0.1650]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, -2.0117e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.1176e-08], + [ 9.3132e-09, 1.8626e-09, 9.1270e-08, ..., 1.8626e-09, + 0.0000e+00, 2.9802e-08], + [ 1.8626e-09, 2.7940e-08, 1.1176e-08, ..., 1.8626e-09, + 0.0000e+00, 8.0094e-08], + ..., + [-1.1176e-08, 5.5879e-09, -1.9558e-07, ..., 1.3039e-08, + 0.0000e+00, 3.7253e-08], + [ 4.8429e-08, 1.1176e-07, 1.1176e-08, ..., -3.7253e-09, + 0.0000e+00, 5.7742e-08], + [ 2.9802e-08, 2.9802e-08, -4.4703e-07, ..., -7.6368e-08, + 0.0000e+00, 3.9116e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([ 2.7802e-02, 5.7644e-03, 1.8338e-02, 2.0947e-02, 3.9710e-02, + 7.2264e-03, 9.1997e-05, 2.9371e-03, -9.0411e-04, -4.3176e-02], + device='cuda:0'), grad: tensor([-1.4342e-07, 3.2224e-07, 1.9930e-07, 5.5321e-07, 9.1828e-07, + -7.8417e-07, 1.1362e-07, -4.7311e-07, 2.5146e-07, -9.5367e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 218.07, cls_loss 0.0021 cls_loss_mapping 0.0033 cls_loss_causal 0.5234 re_mapping 0.0056 re_causal 0.0173 /// teacc 98.96 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.1402, 0.1870, -0.0336, ..., -0.0189, 0.0749, -0.0009], + [ 0.1582, -0.1024, 0.0308, ..., -0.0691, -0.0138, -0.0685], + [-0.0736, -0.1978, -0.0116, ..., -0.0445, 0.0442, -0.1859], + ..., + [-0.1604, -0.2148, 0.0568, ..., -0.0874, -0.0711, -0.1343], + [-0.0886, 0.0071, 0.0321, ..., 0.0198, -0.0981, 0.0591], + [-0.0739, -0.0462, 0.0922, ..., 0.1805, -0.0596, -0.1656]], + device='cuda:0'), grad: tensor([[ 9.1419e-06, -1.0803e-07, 4.3362e-06, ..., 5.5879e-09, + 0.0000e+00, 7.0781e-08], + [ 9.5892e-04, 7.4506e-09, 4.6086e-04, ..., 0.0000e+00, + 0.0000e+00, 7.6368e-08], + [-1.2197e-03, 1.8626e-08, -5.8651e-04, ..., 1.8626e-09, + 0.0000e+00, 2.6636e-07], + ..., + [ 1.5177e-05, 5.5879e-09, 7.2680e-06, ..., 0.0000e+00, + 0.0000e+00, 1.4156e-07], + [ 5.6118e-05, -5.1595e-07, 2.6435e-05, ..., -6.8918e-08, + 0.0000e+00, -2.5127e-06], + [ 6.3121e-05, 2.9616e-07, 3.0547e-05, ..., 3.1665e-08, + 0.0000e+00, 3.5018e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0277, 0.0057, 0.0186, 0.0172, 0.0395, 0.0107, 0.0002, 0.0029, + -0.0011, -0.0431], device='cuda:0'), grad: tensor([ 2.0355e-05, 2.0905e-03, -2.6608e-03, 4.5091e-05, 8.4877e-05, + 8.5652e-05, 4.2945e-05, 3.3319e-05, 1.1921e-04, 1.3864e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 217.81, cls_loss 0.0020 cls_loss_mapping 0.0028 cls_loss_causal 0.5171 re_mapping 0.0057 re_causal 0.0166 /// teacc 98.95 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.1402, 0.1889, -0.0326, ..., -0.0189, 0.0751, -0.0018], + [ 0.1581, -0.1018, 0.0315, ..., -0.0690, -0.0137, -0.0686], + [-0.0712, -0.1987, -0.0096, ..., -0.0446, 0.0442, -0.1864], + ..., + [-0.1612, -0.2165, 0.0565, ..., -0.0875, -0.0711, -0.1347], + [-0.0893, 0.0067, 0.0316, ..., 0.0195, -0.0985, 0.0591], + [-0.0747, -0.0467, 0.0921, ..., 0.1806, -0.0596, -0.1660]], + device='cuda:0'), grad: tensor([[ 5.2154e-08, -2.3283e-07, 7.4506e-08, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 4.2282e-07, 3.7253e-09, 3.9786e-06, ..., 3.9116e-08, + 0.0000e+00, 7.4506e-09], + [ 3.7625e-07, 1.6764e-08, 7.8604e-07, ..., 1.1176e-08, + 0.0000e+00, 3.5390e-08], + ..., + [-1.0952e-06, 3.7253e-09, -4.5709e-06, ..., 2.4214e-08, + 0.0000e+00, 9.3132e-09], + [ 3.1665e-08, -1.6764e-08, 1.3411e-07, ..., -5.5879e-09, + 0.0000e+00, -1.8626e-09], + [ 2.4214e-08, 1.6764e-08, -4.0978e-08, ..., -3.5390e-08, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0287, 0.0054, 0.0206, 0.0179, 0.0398, 0.0102, -0.0009, 0.0028, + -0.0018, -0.0434], device='cuda:0'), grad: tensor([-2.9244e-07, 8.2403e-06, 1.8962e-06, 4.0978e-08, -5.4464e-06, + -1.2238e-06, 2.2817e-06, -9.3728e-06, 3.2596e-07, 3.5334e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 217.73, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.5027 re_mapping 0.0053 re_causal 0.0161 /// teacc 98.90 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.1403, 0.1895, -0.0323, ..., -0.0187, 0.0747, -0.0019], + [ 0.1583, -0.1014, 0.0315, ..., -0.0699, -0.0134, -0.0688], + [-0.0712, -0.1999, -0.0096, ..., -0.0448, 0.0442, -0.1871], + ..., + [-0.1613, -0.2174, 0.0565, ..., -0.0876, -0.0711, -0.1355], + [-0.0898, 0.0065, 0.0313, ..., 0.0192, -0.1003, 0.0590], + [-0.0745, -0.0474, 0.0922, ..., 0.1809, -0.0597, -0.1662]], + device='cuda:0'), grad: tensor([[ 3.9041e-06, 1.8254e-06, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 1.5963e-06], + [ 4.0047e-07, 3.0920e-07, -2.4773e-07, ..., 0.0000e+00, + 0.0000e+00, 3.3341e-07], + [ 2.9989e-07, 1.3784e-07, 1.0617e-07, ..., 0.0000e+00, + 0.0000e+00, 2.0675e-07], + ..., + [ 2.3898e-06, 1.0505e-06, 1.1735e-07, ..., 1.8626e-09, + 0.0000e+00, 9.2760e-07], + [ 5.2340e-07, 1.4529e-07, -9.5367e-07, ..., 0.0000e+00, + 0.0000e+00, -1.1362e-06], + [ 9.2387e-07, 4.1910e-07, 5.5879e-08, ..., -9.3132e-09, + 0.0000e+00, 3.7253e-07]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0289, 0.0053, 0.0211, 0.0178, 0.0398, 0.0104, -0.0010, 0.0028, + -0.0024, -0.0435], device='cuda:0'), grad: tensor([ 6.6236e-06, 9.1270e-07, 6.0350e-07, 7.3686e-06, 1.2852e-07, + -2.2575e-05, 4.0382e-06, 3.9190e-06, -2.6785e-06, 1.6633e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 217.75, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.5206 re_mapping 0.0055 re_causal 0.0163 /// teacc 98.91 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.1402, 0.1905, -0.0322, ..., -0.0187, 0.0742, -0.0021], + [ 0.1589, -0.1014, 0.0310, ..., -0.0706, -0.0132, -0.0692], + [-0.0723, -0.2010, -0.0102, ..., -0.0449, 0.0441, -0.1884], + ..., + [-0.1613, -0.2188, 0.0568, ..., -0.0876, -0.0711, -0.1365], + [-0.0901, 0.0064, 0.0314, ..., 0.0190, -0.1009, 0.0590], + [-0.0744, -0.0476, 0.0926, ..., 0.1810, -0.0598, -0.1664]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -2.7940e-08, 2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [-7.0781e-08, 3.7253e-09, -3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 7.4506e-09, 1.8626e-09, 2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 3.5390e-08], + ..., + [ 3.3528e-08, 7.4506e-09, -1.0431e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 1.4156e-07, 1.6950e-07, 7.6368e-08, ..., 0.0000e+00, + 0.0000e+00, 1.5832e-07], + [ 1.6391e-07, 2.4028e-07, -3.0734e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-07]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0293, 0.0051, 0.0204, 0.0179, 0.0382, 0.0104, -0.0013, 0.0029, + -0.0026, -0.0429], device='cuda:0'), grad: tensor([ 1.7136e-07, 1.8813e-07, -3.9116e-07, -1.5777e-06, 6.2585e-07, + 1.3374e-06, 5.7742e-08, -1.0803e-07, 6.1467e-07, -9.5181e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 218.04, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.4854 re_mapping 0.0057 re_causal 0.0166 /// teacc 98.98 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.1402, 0.1911, -0.0321, ..., -0.0187, 0.0735, -0.0020], + [ 0.1594, -0.1010, 0.0307, ..., -0.0716, -0.0128, -0.0688], + [-0.0725, -0.2013, -0.0110, ..., -0.0455, 0.0440, -0.1886], + ..., + [-0.1616, -0.2193, 0.0570, ..., -0.0884, -0.0711, -0.1369], + [-0.0904, 0.0062, 0.0312, ..., 0.0189, -0.1015, 0.0588], + [-0.0743, -0.0478, 0.0929, ..., 0.1815, -0.0599, -0.1671]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -6.9588e-06, -4.7907e-06, ..., -6.7241e-07, + 0.0000e+00, 1.6764e-08], + [-1.0803e-07, 1.8813e-07, 2.2352e-08, ..., 2.0489e-08, + 0.0000e+00, 4.0978e-08], + [ 2.2352e-08, 3.0920e-07, 2.3656e-07, ..., 2.9802e-08, + 0.0000e+00, 9.1270e-08], + ..., + [ 1.0245e-07, 4.2282e-07, 3.3900e-07, ..., 4.6566e-08, + 0.0000e+00, 5.9605e-08], + [ 1.2480e-07, 3.0920e-07, 2.2165e-07, ..., 3.3528e-08, + 0.0000e+00, 6.1095e-07], + [ 2.6077e-08, 3.7942e-06, 2.6245e-06, ..., 3.8557e-07, + 0.0000e+00, 7.0781e-08]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0294, 0.0051, 0.0203, 0.0177, 0.0380, 0.0106, -0.0016, 0.0030, + -0.0031, -0.0426], device='cuda:0'), grad: tensor([-1.6659e-05, 5.9307e-06, -3.5048e-05, -1.6335e-06, -2.7776e-05, + 1.9707e-06, 3.1829e-05, 1.6280e-06, 3.2559e-06, 3.6538e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 217.82, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5055 re_mapping 0.0056 re_causal 0.0169 /// teacc 98.94 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.1403, 0.1921, -0.0317, ..., -0.0187, 0.0735, -0.0022], + [ 0.1596, -0.1011, 0.0309, ..., -0.0719, -0.0128, -0.0689], + [-0.0729, -0.2023, -0.0115, ..., -0.0456, 0.0440, -0.1895], + ..., + [-0.1615, -0.2199, 0.0570, ..., -0.0885, -0.0711, -0.1375], + [-0.0906, 0.0063, 0.0312, ..., 0.0190, -0.1015, 0.0589], + [-0.0743, -0.0480, 0.0928, ..., 0.1816, -0.0599, -0.1674]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, -4.2617e-06, -4.2915e-06, ..., 2.0489e-08, + 0.0000e+00, 5.2154e-08], + [-6.9663e-07, 1.8626e-07, -5.4203e-07, ..., 2.2352e-08, + 0.0000e+00, 5.7742e-08], + [ 3.1665e-07, 1.9558e-07, 1.5665e-06, ..., 1.6391e-07, + 0.0000e+00, 5.9605e-07], + ..., + [ 1.0245e-07, 2.5332e-07, 3.3826e-06, ..., 8.7172e-07, + 0.0000e+00, 1.8254e-07], + [ 1.0245e-07, 1.9874e-06, 5.9232e-07, ..., -4.1723e-07, + 0.0000e+00, -1.2200e-06], + [ 1.0058e-07, 5.7183e-07, -2.3860e-06, ..., -7.9162e-07, + 0.0000e+00, 4.4703e-08]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0300, 0.0049, 0.0198, 0.0177, 0.0380, 0.0105, -0.0021, 0.0034, + -0.0031, -0.0429], device='cuda:0'), grad: tensor([-6.6683e-06, -7.9721e-07, 2.2594e-06, 8.9407e-07, 2.7381e-07, + 2.6394e-06, -1.6838e-06, 4.0606e-06, 8.4564e-07, -1.8552e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 217.72, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.4991 re_mapping 0.0056 re_causal 0.0160 /// teacc 99.04 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.1405, 0.1926, -0.0317, ..., -0.0187, 0.0731, -0.0024], + [ 0.1612, -0.1009, 0.0329, ..., -0.0720, -0.0128, -0.0688], + [-0.0731, -0.2028, -0.0097, ..., -0.0457, 0.0440, -0.1892], + ..., + [-0.1630, -0.2205, 0.0558, ..., -0.0885, -0.0712, -0.1378], + [-0.0910, 0.0062, 0.0298, ..., 0.0190, -0.1017, 0.0586], + [-0.0746, -0.0483, 0.0928, ..., 0.1817, -0.0603, -0.1676]], + device='cuda:0'), grad: tensor([[ 1.2666e-07, -1.1288e-06, -6.4448e-07, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-08], + [ 2.3469e-07, 1.3784e-07, -5.0291e-08, ..., 1.8626e-09, + 0.0000e+00, 1.3784e-07], + [ 2.9802e-08, 2.6077e-08, 3.5390e-08, ..., 1.8626e-09, + 0.0000e+00, 2.6077e-08], + ..., + [ 1.1921e-07, 4.2841e-08, 6.3330e-08, ..., 0.0000e+00, + 0.0000e+00, 5.0291e-08], + [ 1.9744e-07, -1.3281e-06, -3.1460e-06, ..., -1.3225e-07, + 0.0000e+00, -1.3411e-06], + [ 4.8243e-07, 1.5739e-06, 3.0212e-06, ..., 1.2480e-07, + 0.0000e+00, 1.5236e-06]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0301, 0.0062, 0.0213, 0.0179, 0.0379, 0.0104, -0.0022, 0.0027, + -0.0050, -0.0429], device='cuda:0'), grad: tensor([-2.7511e-06, 5.3458e-07, -3.2224e-07, 1.7136e-07, -1.1355e-05, + -1.8384e-06, 3.1479e-06, 9.9465e-07, -5.8077e-06, 1.7166e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 218.07, cls_loss 0.0015 cls_loss_mapping 0.0019 cls_loss_causal 0.4943 re_mapping 0.0056 re_causal 0.0166 /// teacc 99.02 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.1406, 0.1931, -0.0319, ..., -0.0187, 0.0732, -0.0031], + [ 0.1613, -0.1010, 0.0331, ..., -0.0720, -0.0126, -0.0689], + [-0.0729, -0.2043, -0.0097, ..., -0.0457, 0.0440, -0.1897], + ..., + [-0.1629, -0.2209, 0.0559, ..., -0.0885, -0.0712, -0.1382], + [-0.0914, 0.0061, 0.0293, ..., 0.0190, -0.1019, 0.0589], + [-0.0748, -0.0487, 0.0928, ..., 0.1817, -0.0603, -0.1681]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -1.6782e-06, -1.6727e-06, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-1.1362e-07, 7.2643e-08, 5.3085e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 7.4506e-09, 5.0291e-08, 4.3027e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + ..., + [ 4.6566e-08, 8.0094e-08, -9.8720e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 3.1479e-07, 1.0058e-06, 8.3819e-07, ..., 1.8626e-09, + 0.0000e+00, 2.7753e-07], + [ 1.1176e-08, 1.9558e-07, 1.9558e-07, ..., -9.3132e-09, + 0.0000e+00, 2.7940e-08]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0300, 0.0061, 0.0216, 0.0180, 0.0386, 0.0103, -0.0022, 0.0029, + -0.0052, -0.0433], device='cuda:0'), grad: tensor([-2.9448e-06, 1.1288e-06, 6.0908e-07, 8.1956e-08, -2.9001e-06, + -3.7439e-07, 1.0356e-06, -6.6683e-07, 2.1439e-06, 1.8403e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 217.78, cls_loss 0.0018 cls_loss_mapping 0.0020 cls_loss_causal 0.5432 re_mapping 0.0054 re_causal 0.0162 /// teacc 98.97 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.1401, 0.1938, -0.0318, ..., -0.0187, 0.0733, -0.0032], + [ 0.1605, -0.1010, 0.0324, ..., -0.0722, -0.0125, -0.0691], + [-0.0731, -0.2054, -0.0100, ..., -0.0457, 0.0440, -0.1905], + ..., + [-0.1622, -0.2242, 0.0565, ..., -0.0886, -0.0712, -0.1400], + [-0.0916, 0.0062, 0.0295, ..., 0.0190, -0.1020, 0.0591], + [-0.0749, -0.0491, 0.0927, ..., 0.1819, -0.0604, -0.1692]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -1.0598e-06, -6.7614e-07, ..., 0.0000e+00, + 0.0000e+00, 1.6764e-08], + [-1.2480e-07, 8.3819e-08, -4.4703e-08, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 1.8626e-08, 1.6205e-07, 2.3097e-07, ..., 1.8626e-09, + 0.0000e+00, 8.0094e-08], + ..., + [ 7.0781e-08, 1.7136e-07, 1.2666e-07, ..., 7.4506e-09, + 0.0000e+00, 2.0489e-08], + [ 2.3656e-07, 2.0489e-08, -3.2596e-07, ..., -1.3039e-08, + 0.0000e+00, -8.9407e-08], + [ 8.3819e-08, 5.6624e-07, 1.5832e-07, ..., -1.6764e-08, + 0.0000e+00, 1.0803e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0301, 0.0050, 0.0215, 0.0181, 0.0386, 0.0103, -0.0023, 0.0038, + -0.0050, -0.0436], device='cuda:0'), grad: tensor([-2.8554e-06, 2.0117e-07, 1.7323e-07, 1.4622e-06, 1.9185e-07, + -7.8045e-07, 5.9418e-07, 1.7323e-07, -4.2841e-07, 1.2852e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 217.67, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5176 re_mapping 0.0054 re_causal 0.0162 /// teacc 99.02 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.1404, 0.1942, -0.0324, ..., -0.0188, 0.0734, -0.0033], + [ 0.1609, -0.1012, 0.0318, ..., -0.0722, -0.0125, -0.0693], + [-0.0737, -0.2064, -0.0104, ..., -0.0459, 0.0438, -0.1911], + ..., + [-0.1623, -0.2248, 0.0569, ..., -0.0886, -0.0712, -0.1402], + [-0.0919, 0.0062, 0.0301, ..., 0.0191, -0.1020, 0.0591], + [-0.0747, -0.0487, 0.0929, ..., 0.1820, -0.0604, -0.1697]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -7.5698e-06, -3.2410e-07, ..., 0.0000e+00, + 0.0000e+00, 7.6368e-08], + [ 1.3877e-06, 1.0859e-06, 6.9849e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3113e-06], + [ 6.9551e-06, 5.3905e-06, 4.3064e-06, ..., 0.0000e+00, + 0.0000e+00, 6.5491e-06], + ..., + [ 6.7055e-08, 7.8231e-08, -1.9409e-06, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-08], + [-1.4663e-05, -1.1235e-05, -6.4857e-06, ..., 0.0000e+00, + 0.0000e+00, -1.3836e-05], + [ 4.0978e-08, 1.9930e-07, 5.0291e-07, ..., 0.0000e+00, + 0.0000e+00, 3.1665e-08]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0298, 0.0045, 0.0213, 0.0182, 0.0384, 0.0101, -0.0025, 0.0042, + -0.0046, -0.0433], device='cuda:0'), grad: tensor([-1.3545e-05, 7.4469e-06, 3.8832e-05, 5.7183e-07, 8.5868e-07, + 2.6584e-05, 1.8150e-05, -3.5465e-06, -7.6830e-05, 1.4957e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 217.60, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4918 re_mapping 0.0055 re_causal 0.0160 /// teacc 99.01 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.1405, 0.1973, -0.0302, ..., -0.0188, 0.0732, -0.0038], + [ 0.1621, -0.0992, 0.0326, ..., -0.0695, -0.0124, -0.0667], + [-0.0744, -0.2075, -0.0111, ..., -0.0460, 0.0434, -0.1918], + ..., + [-0.1623, -0.2287, 0.0571, ..., -0.0886, -0.0718, -0.1422], + [-0.0936, 0.0053, 0.0277, ..., 0.0164, -0.1022, 0.0581], + [-0.0753, -0.0523, 0.0919, ..., 0.1819, -0.0619, -0.1710]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -2.7940e-08, -2.7940e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-3.5185e-06, 7.4506e-09, -1.6224e-06, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-2.2352e-08, 5.5879e-09, 4.6566e-08, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + ..., + [ 3.3658e-06, 1.3039e-08, 1.5125e-06, ..., 1.8626e-09, + 0.0000e+00, 1.4901e-08], + [ 1.6019e-07, 9.3132e-08, 1.8626e-08, ..., 1.1176e-08, + 0.0000e+00, 7.8231e-08], + [ 7.4506e-08, 2.2352e-08, -2.9802e-08, ..., -1.4901e-08, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0313, 0.0055, 0.0207, 0.0183, 0.0381, 0.0101, -0.0024, 0.0044, + -0.0062, -0.0445], device='cuda:0'), grad: tensor([ 4.8243e-07, -3.6508e-06, -6.7987e-07, 1.5460e-07, -5.4203e-06, + -2.1234e-07, 4.7497e-06, 4.0717e-06, 2.9616e-07, 1.9930e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 217.87, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.5300 re_mapping 0.0052 re_causal 0.0159 /// teacc 98.97 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.1408, 0.1978, -0.0298, ..., -0.0187, 0.0756, -0.0042], + [ 0.1623, -0.0992, 0.0320, ..., -0.0695, -0.0123, -0.0668], + [-0.0745, -0.2081, -0.0091, ..., -0.0461, 0.0434, -0.1925], + ..., + [-0.1622, -0.2294, 0.0570, ..., -0.0886, -0.0718, -0.1448], + [-0.0941, 0.0051, 0.0287, ..., 0.0162, -0.1027, 0.0589], + [-0.0750, -0.0526, 0.0921, ..., 0.1820, -0.0624, -0.1710]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, -3.1665e-08, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-1.0431e-07, 3.7253e-09, -7.8231e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.6764e-08, 1.2137e-05, 1.0274e-05, ..., 0.0000e+00, + 0.0000e+00, 8.5086e-06], + ..., + [ 1.6764e-08, 0.0000e+00, 2.4214e-08, ..., 1.8626e-09, + 0.0000e+00, 3.7253e-09], + [ 2.0489e-08, -1.2159e-05, -1.0267e-05, ..., 0.0000e+00, + 0.0000e+00, -8.5160e-06], + [ 5.4017e-08, 3.5390e-08, -4.3958e-07, ..., -4.6566e-08, + 0.0000e+00, 4.4703e-08]], device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0314, 0.0049, 0.0235, 0.0183, 0.0372, 0.0101, -0.0022, 0.0032, + -0.0060, -0.0443], device='cuda:0'), grad: tensor([ 6.5193e-08, -1.4342e-07, 3.3945e-05, -1.2666e-07, 9.7789e-07, + 1.8626e-09, 0.0000e+00, 1.0245e-07, -3.3975e-05, -9.1456e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 218.10, cls_loss 0.0015 cls_loss_mapping 0.0021 cls_loss_causal 0.4908 re_mapping 0.0053 re_causal 0.0158 /// teacc 98.92 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.1408, 0.1995, -0.0291, ..., -0.0187, 0.0755, -0.0042], + [ 0.1624, -0.0992, 0.0320, ..., -0.0700, -0.0122, -0.0668], + [-0.0746, -0.2095, -0.0096, ..., -0.0462, 0.0434, -0.1936], + ..., + [-0.1628, -0.2299, 0.0572, ..., -0.0887, -0.0719, -0.1451], + [-0.0946, 0.0051, 0.0284, ..., 0.0160, -0.1031, 0.0592], + [-0.0751, -0.0533, 0.0919, ..., 0.1825, -0.0622, -0.1714]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, -8.0094e-08, 1.2293e-07, ..., 9.3132e-09, + 1.8626e-09, 1.4901e-08], + [ 3.8177e-05, 6.3330e-08, 2.8014e-04, ..., 1.7509e-07, + 0.0000e+00, 1.1176e-08], + [ 7.3947e-07, 5.5879e-09, 5.3719e-06, ..., 1.1176e-08, + 0.0000e+00, 1.4901e-08], + ..., + [-4.7684e-05, 1.8626e-08, -3.4857e-04, ..., 1.3970e-07, + 3.7253e-09, 3.3528e-08], + [ 2.1048e-07, 2.6077e-08, 1.4808e-06, ..., 2.0117e-07, + 1.1176e-08, -1.1176e-08], + [ 8.4713e-06, 1.3039e-08, 5.3823e-05, ..., -3.7998e-06, + 0.0000e+00, 6.7055e-08]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0327, 0.0043, 0.0234, 0.0182, 0.0371, 0.0101, -0.0021, 0.0034, + -0.0064, -0.0446], device='cuda:0'), grad: tensor([ 4.0792e-07, 3.2687e-04, 6.2883e-06, 2.7753e-07, 1.2398e-05, + -6.1467e-08, -4.8615e-07, -4.0603e-04, 1.9744e-06, 5.8115e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 218.05, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5358 re_mapping 0.0054 re_causal 0.0161 /// teacc 98.98 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.1410, 0.1996, -0.0292, ..., -0.0188, 0.0753, -0.0047], + [ 0.1635, -0.0986, 0.0318, ..., -0.0691, -0.0121, -0.0663], + [-0.0746, -0.2123, -0.0100, ..., -0.0468, 0.0434, -0.1966], + ..., + [-0.1637, -0.2302, 0.0577, ..., -0.0893, -0.0719, -0.1454], + [-0.0955, 0.0051, 0.0279, ..., 0.0148, -0.1039, 0.0596], + [-0.0752, -0.0534, 0.0919, ..., 0.1830, -0.0623, -0.1717]], + device='cuda:0'), grad: tensor([[ 1.3933e-06, 1.6801e-06, 1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 1.0058e-07], + [-5.4203e-07, 7.4506e-09, 2.7008e-07, ..., 0.0000e+00, + 0.0000e+00, 6.7800e-07], + [ 1.2293e-07, 1.1176e-08, 3.2969e-07, ..., 0.0000e+00, + 0.0000e+00, 1.7826e-06], + ..., + [ 2.3842e-07, 5.5879e-09, -9.1456e-07, ..., 0.0000e+00, + 0.0000e+00, 8.3074e-07], + [ 1.5646e-07, 6.8918e-08, 9.1270e-08, ..., 0.0000e+00, + 0.0000e+00, 4.9174e-07], + [ 7.8231e-08, 8.3819e-08, -7.0781e-07, ..., -1.8626e-09, + 0.0000e+00, 9.1270e-08]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0326, 0.0043, 0.0230, 0.0182, 0.0377, 0.0100, -0.0020, 0.0038, + -0.0067, -0.0449], device='cuda:0'), grad: tensor([ 6.2883e-06, 2.4941e-06, 4.9323e-06, -1.0654e-05, -1.7546e-06, + 4.0978e-06, -8.5607e-06, 6.5379e-07, 1.8105e-06, 6.7614e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 217.86, cls_loss 0.0015 cls_loss_mapping 0.0028 cls_loss_causal 0.5003 re_mapping 0.0055 re_causal 0.0159 /// teacc 98.96 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.1410, 0.1998, -0.0295, ..., -0.0188, 0.0745, -0.0059], + [ 0.1635, -0.0988, 0.0316, ..., -0.0694, -0.0121, -0.0667], + [-0.0747, -0.2153, -0.0107, ..., -0.0471, 0.0435, -0.2000], + ..., + [-0.1637, -0.2322, 0.0579, ..., -0.0898, -0.0720, -0.1469], + [-0.0948, 0.0064, 0.0286, ..., 0.0150, -0.1014, 0.0621], + [-0.0753, -0.0536, 0.0922, ..., 0.1836, -0.0623, -0.1719]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -9.6858e-08, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-5.3085e-08, 2.7940e-09, 6.2399e-08, ..., 9.3132e-10, + 0.0000e+00, 1.4901e-08], + [ 1.9558e-08, 8.5682e-08, 2.4866e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3132e-07], + ..., + [ 5.1223e-08, 3.7253e-09, -7.1432e-07, ..., 1.8626e-09, + 0.0000e+00, 4.7497e-08], + [ 4.7497e-08, -6.2399e-08, 5.5879e-08, ..., 0.0000e+00, + 0.0000e+00, -9.7789e-08], + [ 1.8626e-08, 2.9802e-08, 5.0291e-08, ..., -1.1176e-08, + 0.0000e+00, 2.2352e-08]], device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0324, 0.0040, 0.0225, 0.0183, 0.0375, 0.0098, -0.0023, 0.0038, + -0.0046, -0.0446], device='cuda:0'), grad: tensor([ 1.7043e-07, 4.0978e-07, 1.5087e-06, -5.1223e-08, 4.6380e-07, + 5.0943e-07, -3.5763e-07, -3.8594e-06, 5.8301e-07, 6.1374e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 218.00, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5261 re_mapping 0.0053 re_causal 0.0162 /// teacc 98.95 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.1419, 0.2002, -0.0294, ..., -0.0188, 0.0741, -0.0049], + [ 0.1637, -0.0989, 0.0314, ..., -0.0696, -0.0120, -0.0668], + [-0.0750, -0.2164, -0.0111, ..., -0.0474, 0.0433, -0.2008], + ..., + [-0.1638, -0.2331, 0.0581, ..., -0.0899, -0.0721, -0.1472], + [-0.0950, 0.0064, 0.0286, ..., 0.0148, -0.1017, 0.0625], + [-0.0752, -0.0539, 0.0924, ..., 0.1840, -0.0620, -0.1721]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -3.7253e-08, -3.1665e-08, ..., 0.0000e+00, + -1.1176e-08, 4.0978e-08], + [ 7.4506e-09, 1.3970e-08, -1.3039e-08, ..., 0.0000e+00, + 9.3132e-10, 7.3574e-08], + [ 2.6077e-08, 5.5879e-09, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 1.3970e-07], + ..., + [ 6.9849e-08, 3.7253e-08, 5.9605e-08, ..., 0.0000e+00, + 6.5193e-09, 2.4587e-07], + [ 3.4180e-07, 1.6112e-07, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 3.2969e-07], + [ 7.4506e-08, 4.5635e-08, -1.4901e-07, ..., -1.8626e-09, + 1.8626e-09, 6.7987e-08]], device='cuda:0') +Epoch 253, bias, value: tensor([ 0.0324, 0.0039, 0.0223, 0.0184, 0.0373, 0.0098, -0.0024, 0.0039, + -0.0046, -0.0445], device='cuda:0'), grad: tensor([ 9.7789e-08, 3.2224e-07, 2.8312e-07, 2.8498e-06, -7.4983e-05, + -6.3665e-06, 1.4510e-06, 7.5400e-05, 8.0746e-07, 3.9209e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 218.44, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.5259 re_mapping 0.0053 re_causal 0.0159 /// teacc 98.93 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.1428, 0.2002, -0.0295, ..., -0.0189, 0.0699, -0.0080], + [ 0.1639, -0.0991, 0.0311, ..., -0.0699, -0.0120, -0.0672], + [-0.0751, -0.2173, -0.0108, ..., -0.0475, 0.0435, -0.2024], + ..., + [-0.1640, -0.2335, 0.0581, ..., -0.0900, -0.0719, -0.1484], + [-0.0951, 0.0067, 0.0290, ..., 0.0148, -0.1013, 0.0633], + [-0.0753, -0.0540, 0.0927, ..., 0.1842, -0.0633, -0.1724]], + device='cuda:0'), grad: tensor([[ 8.7544e-08, -2.8126e-07, -2.1607e-07, ..., 4.6566e-09, + 3.7253e-09, 6.2399e-08], + [ 1.1176e-08, 5.4948e-08, 2.1793e-07, ..., 9.3132e-10, + -9.3132e-10, 3.0734e-08], + [-7.4506e-09, 1.4622e-07, 8.4750e-08, ..., 0.0000e+00, + 0.0000e+00, 1.7323e-07], + ..., + [ 1.0524e-07, 1.2852e-07, 1.6857e-07, ..., 2.7940e-09, + 0.0000e+00, 1.4435e-07], + [ 7.6648e-07, 3.6880e-07, -9.0338e-08, ..., 4.2841e-08, + 2.7940e-09, 5.5879e-08], + [ 1.7509e-06, 1.4165e-06, 2.2352e-07, ..., 1.0338e-07, + 0.0000e+00, 1.0198e-06]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0320, 0.0035, 0.0223, 0.0196, 0.0371, 0.0088, -0.0023, 0.0040, + -0.0041, -0.0442], device='cuda:0'), grad: tensor([-2.7381e-07, 1.1222e-06, -5.5786e-07, 3.6433e-06, -2.5406e-06, + -6.2101e-06, -8.9779e-07, 1.1045e-06, 8.8383e-07, 3.6880e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 218.53, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5136 re_mapping 0.0054 re_causal 0.0159 /// teacc 99.06 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.1429, 0.2006, -0.0295, ..., -0.0189, 0.0694, -0.0082], + [ 0.1640, -0.0990, 0.0311, ..., -0.0698, -0.0116, -0.0673], + [-0.0751, -0.2180, -0.0108, ..., -0.0476, 0.0434, -0.2031], + ..., + [-0.1641, -0.2342, 0.0575, ..., -0.0900, -0.0718, -0.1521], + [-0.0954, 0.0071, 0.0319, ..., 0.0149, -0.1026, 0.0652], + [-0.0757, -0.0542, 0.0928, ..., 0.1842, -0.0641, -0.1735]], + device='cuda:0'), grad: tensor([[ 2.3004e-07, 8.6613e-08, 1.2852e-07, ..., 9.3132e-09, + 0.0000e+00, 1.4715e-07], + [-8.5123e-07, -1.2293e-07, -1.1455e-06, ..., 7.4506e-09, + -1.8626e-09, 1.5460e-07], + [ 8.5682e-08, 1.3970e-08, 3.2596e-08, ..., 9.3132e-10, + 0.0000e+00, 3.3528e-08], + ..., + [ 9.0804e-07, 1.7602e-07, 6.9570e-07, ..., 9.3132e-10, + 0.0000e+00, 2.7008e-08], + [ 9.1922e-07, 4.8708e-07, 5.1130e-07, ..., 3.8184e-08, + 0.0000e+00, 6.8638e-07], + [-4.3958e-07, -1.0943e-06, -1.9968e-06, ..., -1.9744e-07, + 0.0000e+00, -2.1644e-06]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0321, 0.0033, 0.0224, 0.0196, 0.0349, 0.0086, -0.0022, 0.0032, + -0.0022, -0.0430], device='cuda:0'), grad: tensor([ 5.5134e-07, -1.2647e-06, 2.1979e-07, 2.1532e-06, 3.9116e-08, + -5.5544e-06, 5.4725e-06, 1.1073e-06, 1.9968e-06, -4.6901e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 218.40, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.4905 re_mapping 0.0056 re_causal 0.0157 /// teacc 98.94 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.1409, 0.2033, -0.0274, ..., -0.0189, 0.0686, -0.0086], + [ 0.1644, -0.1002, 0.0313, ..., -0.0701, -0.0123, -0.0691], + [-0.0753, -0.2206, -0.0121, ..., -0.0477, 0.0435, -0.2043], + ..., + [-0.1646, -0.2348, 0.0578, ..., -0.0901, -0.0719, -0.1520], + [-0.0939, 0.0092, 0.0323, ..., 0.0149, -0.0998, 0.0673], + [-0.0756, -0.0543, 0.0927, ..., 0.1844, -0.0643, -0.1738]], + device='cuda:0'), grad: tensor([[-6.6124e-07, -2.7269e-06, 1.6019e-07, ..., 1.8626e-09, + 2.3283e-08, 4.5635e-08], + [-3.1888e-06, -9.0897e-07, -1.9409e-06, ..., 9.3132e-10, + -5.2806e-07, 2.1420e-08], + [ 8.9593e-07, 2.8592e-07, 1.0617e-05, ..., 2.7940e-09, + 1.4901e-07, 4.2003e-07], + ..., + [ 2.6077e-07, 1.0524e-07, -1.1064e-05, ..., 9.3132e-10, + 4.2841e-08, 1.1455e-07], + [ 1.2731e-06, 1.5739e-07, 6.5286e-07, ..., 1.7695e-08, + 2.1700e-07, -1.1967e-06], + [ 3.1572e-07, 2.1700e-07, -6.7689e-06, ..., -2.3860e-06, + 4.1910e-08, -3.5390e-08]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0347, 0.0033, 0.0218, 0.0195, 0.0352, 0.0084, -0.0054, 0.0034, + -0.0010, -0.0432], device='cuda:0'), grad: tensor([-3.5241e-06, -4.3176e-06, 1.6287e-05, 1.4715e-07, 2.3633e-05, + 6.2473e-06, 7.1712e-07, -1.5318e-05, 1.9372e-07, -2.4050e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 218.24, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.5109 re_mapping 0.0056 re_causal 0.0161 /// teacc 99.00 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.1409, 0.2035, -0.0273, ..., -0.0189, 0.0710, -0.0087], + [ 0.1651, -0.1005, 0.0317, ..., -0.0703, -0.0119, -0.0691], + [-0.0759, -0.2210, -0.0129, ..., -0.0477, 0.0410, -0.2046], + ..., + [-0.1649, -0.2351, 0.0582, ..., -0.0901, -0.0714, -0.1528], + [-0.0954, 0.0082, 0.0316, ..., 0.0141, -0.0997, 0.0664], + [-0.0760, -0.0540, 0.0924, ..., 0.1848, -0.0652, -0.1711]], + device='cuda:0'), grad: tensor([[ 1.9558e-08, -6.1467e-08, 9.6858e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.6042e-07, 1.8626e-09, 1.1694e-04, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 1.7509e-07, 9.3132e-10, 6.8638e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 6.1467e-08, 9.3132e-10, -1.2994e-04, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 5.1223e-08, 4.6566e-09, 2.3097e-06, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 3.1665e-08, 2.6077e-08, 9.3505e-06, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0347, 0.0038, 0.0213, 0.0194, 0.0340, 0.0088, -0.0053, 0.0038, + -0.0025, -0.0429], device='cuda:0'), grad: tensor([ 1.6661e-06, 1.3518e-04, 1.6242e-05, -9.8813e-07, 2.4885e-06, + 1.7639e-06, -2.3022e-05, -1.4782e-04, 3.7253e-06, 1.0975e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 218.39, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.5077 re_mapping 0.0054 re_causal 0.0157 /// teacc 98.99 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.1410, 0.2044, -0.0271, ..., -0.0189, 0.0714, -0.0085], + [ 0.1661, -0.1017, 0.0314, ..., -0.0704, -0.0119, -0.0692], + [-0.0769, -0.2235, -0.0134, ..., -0.0478, 0.0410, -0.2058], + ..., + [-0.1655, -0.2357, 0.0584, ..., -0.0902, -0.0717, -0.1530], + [-0.0958, 0.0082, 0.0318, ..., 0.0142, -0.0993, 0.0672], + [-0.0762, -0.0545, 0.0924, ..., 0.1849, -0.0653, -0.1716]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.0245e-07, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [-4.1910e-08, 2.0489e-08, 3.1386e-07, ..., 0.0000e+00, + 3.7253e-09, 3.0734e-08], + [ 8.3819e-09, 1.7229e-07, 1.2573e-07, ..., 0.0000e+00, + 2.7008e-08, 1.6671e-07], + ..., + [ 2.7008e-08, 6.5193e-09, -1.6959e-06, ..., 9.3132e-10, + 0.0000e+00, 6.3330e-08], + [ 1.1921e-07, -6.2678e-07, -2.0489e-07, ..., 9.3132e-10, + -3.8184e-08, -7.6555e-07], + [ 1.8626e-08, 6.0536e-08, 1.2349e-06, ..., -7.4506e-09, + 0.0000e+00, 2.2352e-08]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0355, 0.0035, 0.0206, 0.0191, 0.0330, 0.0090, -0.0057, 0.0044, + -0.0020, -0.0431], device='cuda:0'), grad: tensor([-1.3970e-07, 8.5495e-07, 6.2212e-07, -2.6450e-07, -4.4797e-07, + 9.5647e-07, 8.2888e-08, -3.1125e-06, -1.3905e-06, 2.8424e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 218.29, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.4933 re_mapping 0.0053 re_causal 0.0151 /// teacc 99.00 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.1422, 0.2047, -0.0273, ..., -0.0189, 0.0715, -0.0087], + [ 0.1659, -0.1017, 0.0313, ..., -0.0704, -0.0119, -0.0693], + [-0.0751, -0.2242, -0.0136, ..., -0.0479, 0.0409, -0.2062], + ..., + [-0.1655, -0.2364, 0.0585, ..., -0.0902, -0.0718, -0.1532], + [-0.0960, 0.0082, 0.0319, ..., 0.0142, -0.0991, 0.0675], + [-0.0771, -0.0548, 0.0927, ..., 0.1852, -0.0656, -0.1720]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 7.0781e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-1.9558e-08, 7.4506e-09, 3.5390e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 7.4506e-09, 4.6566e-09, 6.2399e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + ..., + [ 3.7253e-09, 1.8626e-09, -2.0210e-07, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, -9.3132e-09, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, -2.9802e-08], + [ 9.3132e-09, 9.3132e-09, 4.9360e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0353, 0.0031, 0.0217, 0.0187, 0.0335, 0.0093, -0.0057, 0.0044, + -0.0019, -0.0433], device='cuda:0'), grad: tensor([ 7.1153e-07, 8.5682e-08, 6.5193e-08, 8.6613e-08, 3.4459e-08, + 8.1956e-08, -8.4564e-07, -3.2876e-07, 7.4506e-09, 1.1455e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 218.33, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.5232 re_mapping 0.0053 re_causal 0.0160 /// teacc 98.97 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.1419, 0.2052, -0.0273, ..., -0.0189, 0.0713, -0.0074], + [ 0.1635, -0.1021, 0.0307, ..., -0.0702, -0.0115, -0.0693], + [-0.0723, -0.2248, -0.0116, ..., -0.0480, 0.0404, -0.2067], + ..., + [-0.1656, -0.2379, 0.0587, ..., -0.0902, -0.0718, -0.1534], + [-0.0964, 0.0082, 0.0317, ..., 0.0142, -0.0992, 0.0677], + [-0.0778, -0.0549, 0.0927, ..., 0.1852, -0.0658, -0.1723]], + device='cuda:0'), grad: tensor([[ 4.0978e-08, -9.3132e-10, 4.2841e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-3.4198e-06, -4.6566e-09, -3.0622e-06, ..., 0.0000e+00, + -1.2107e-08, 8.3819e-09], + [ 4.4145e-07, 4.6566e-09, 5.4017e-08, ..., 0.0000e+00, + 1.1176e-08, 1.6764e-08], + ..., + [ 2.2911e-07, 9.3132e-10, 8.1956e-08, ..., 4.6566e-09, + 9.3132e-10, 3.5390e-08], + [ 1.6764e-07, -1.1176e-08, 1.6112e-07, ..., -9.3132e-10, + 0.0000e+00, 1.0245e-08], + [ 2.3283e-08, 1.3970e-08, 3.0734e-08, ..., -1.3970e-08, + 0.0000e+00, 2.7008e-08]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0355, 0.0005, 0.0244, 0.0185, 0.0335, 0.0092, -0.0058, 0.0048, + -0.0022, -0.0435], device='cuda:0'), grad: tensor([ 1.1269e-07, -7.5549e-06, 4.8615e-07, -3.9767e-07, 6.2399e-08, + 2.0862e-07, 6.1691e-06, 3.4831e-07, 4.3306e-07, 1.5739e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 218.04, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5221 re_mapping 0.0054 re_causal 0.0162 /// teacc 99.03 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.1419, 0.2052, -0.0273, ..., -0.0189, 0.0715, -0.0077], + [ 0.1635, -0.1021, 0.0306, ..., -0.0702, -0.0115, -0.0694], + [-0.0723, -0.2279, -0.0118, ..., -0.0483, 0.0403, -0.2084], + ..., + [-0.1657, -0.2382, 0.0588, ..., -0.0904, -0.0718, -0.1536], + [-0.0964, 0.0084, 0.0317, ..., 0.0135, -0.0992, 0.0681], + [-0.0779, -0.0548, 0.0928, ..., 0.1856, -0.0661, -0.1723]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, -7.4506e-09, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [-9.4064e-08, -9.3132e-10, -1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + [ 2.1420e-08, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + ..., + [ 2.9802e-08, 1.8626e-09, -2.5146e-08, ..., 1.8626e-09, + 0.0000e+00, 1.4901e-08], + [ 2.1979e-07, 9.6858e-08, 4.5635e-08, ..., 6.5193e-09, + 0.0000e+00, 1.8440e-07], + [ 2.5146e-08, 1.5832e-08, -2.6077e-08, ..., -1.1176e-08, + 0.0000e+00, 1.3039e-08]], device='cuda:0') +Epoch 261, bias, value: tensor([ 0.0354, 0.0004, 0.0242, 0.0185, 0.0336, 0.0090, -0.0058, 0.0049, + -0.0016, -0.0435], device='cuda:0'), grad: tensor([ 1.3411e-07, -9.3132e-10, 2.6822e-07, -2.2165e-07, -6.1467e-06, + -3.7812e-07, 4.4703e-08, 2.5164e-06, 4.9081e-07, 3.2857e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 218.51, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5119 re_mapping 0.0053 re_causal 0.0157 /// teacc 99.01 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.1421, 0.2057, -0.0272, ..., -0.0189, 0.0713, -0.0080], + [ 0.1640, -0.1018, 0.0310, ..., -0.0702, -0.0111, -0.0695], + [-0.0725, -0.2287, -0.0128, ..., -0.0483, 0.0403, -0.2092], + ..., + [-0.1661, -0.2393, 0.0586, ..., -0.0905, -0.0719, -0.1542], + [-0.0968, 0.0084, 0.0317, ..., 0.0135, -0.0994, 0.0682], + [-0.0789, -0.0553, 0.0930, ..., 0.1857, -0.0662, -0.1725]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, -4.2468e-07, -1.0803e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 9.3132e-09, 2.5146e-08, 7.4506e-08, ..., 0.0000e+00, + 0.0000e+00, 3.4459e-08], + [ 3.7253e-09, 8.5682e-08, 8.3819e-08, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 2.1420e-08, 2.3283e-08, -4.5542e-07, ..., 9.3132e-10, + 0.0000e+00, 2.3283e-08], + [ 2.8685e-07, 1.3970e-07, -1.1176e-08, ..., -3.7253e-09, + 0.0000e+00, 3.6322e-08], + [ 3.6787e-07, 4.3120e-07, 3.2503e-07, ..., -9.3132e-10, + 0.0000e+00, 3.1758e-07]], device='cuda:0') +Epoch 262, bias, value: tensor([ 0.0356, 0.0008, 0.0239, 0.0186, 0.0341, 0.0092, -0.0059, 0.0046, + -0.0016, -0.0436], device='cuda:0'), grad: tensor([ 1.4622e-06, 1.0561e-06, 4.7274e-06, 8.9258e-06, 2.5891e-06, + -3.4831e-06, -5.3607e-06, -1.3202e-05, 2.4121e-07, 2.9914e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 218.57, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.5312 re_mapping 0.0051 re_causal 0.0154 /// teacc 98.95 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.1427, 0.2059, -0.0272, ..., -0.0189, 0.0713, -0.0081], + [ 0.1641, -0.1010, 0.0312, ..., -0.0710, -0.0111, -0.0696], + [-0.0725, -0.2289, -0.0126, ..., -0.0483, 0.0403, -0.2094], + ..., + [-0.1663, -0.2407, 0.0585, ..., -0.0906, -0.0719, -0.1565], + [-0.0976, 0.0079, 0.0315, ..., 0.0134, -0.0995, 0.0679], + [-0.0789, -0.0557, 0.0930, ..., 0.1862, -0.0663, -0.1727]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -2.8774e-05, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [-6.1095e-06, 1.9781e-06, -5.5879e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 4.2506e-06, 1.8142e-06, 5.6811e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.6177e-06, 5.2154e-08, -3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 6.5193e-09], + [ 7.9162e-08, 2.4121e-07, -1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-08], + [ 4.6566e-08, 6.2399e-08, -1.6019e-07, ..., -2.7940e-09, + 0.0000e+00, 8.3819e-09]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0357, 0.0008, 0.0240, 0.0188, 0.0345, 0.0094, -0.0058, 0.0042, + -0.0022, -0.0438], device='cuda:0'), grad: tensor([-7.6234e-05, -1.0608e-06, 9.3430e-06, 2.9579e-06, 3.2876e-07, + 4.7497e-08, 6.2346e-05, 1.5870e-06, 6.0443e-07, 9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 218.13, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4643 re_mapping 0.0053 re_causal 0.0151 /// teacc 99.01 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.1428, 0.2061, -0.0277, ..., -0.0189, 0.0715, -0.0078], + [ 0.1646, -0.1010, 0.0312, ..., -0.0711, -0.0111, -0.0698], + [-0.0728, -0.2296, -0.0128, ..., -0.0484, 0.0403, -0.2098], + ..., + [-0.1666, -0.2417, 0.0588, ..., -0.0907, -0.0719, -0.1567], + [-0.0978, 0.0078, 0.0310, ..., 0.0136, -0.0995, 0.0679], + [-0.0791, -0.0558, 0.0934, ..., 0.1864, -0.0663, -0.1729]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, -3.6880e-07, -2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-08], + [ 3.7253e-09, 4.5635e-08, 1.6950e-07, ..., 0.0000e+00, + 0.0000e+00, 3.8184e-08], + [ 8.3819e-09, -4.0326e-07, 2.3283e-08, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-08], + ..., + [ 2.0489e-08, 1.3039e-08, -1.1362e-07, ..., 0.0000e+00, + 0.0000e+00, 1.4715e-07], + [ 3.7812e-07, 3.4831e-07, 7.7300e-08, ..., 0.0000e+00, + 0.0000e+00, 2.9430e-07], + [ 6.3330e-08, 6.7055e-08, 1.5646e-07, ..., 0.0000e+00, + 0.0000e+00, 6.7987e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([ 0.0354, 0.0010, 0.0238, 0.0181, 0.0354, 0.0101, -0.0059, 0.0043, + -0.0026, -0.0438], device='cuda:0'), grad: tensor([-2.9057e-07, 1.5311e-06, -3.4161e-06, 1.2927e-05, -4.5858e-06, + -1.2696e-05, 1.7723e-06, 9.8068e-07, 1.5199e-06, 2.2165e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 218.61, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.5128 re_mapping 0.0056 re_causal 0.0158 /// teacc 98.95 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.1428, 0.2064, -0.0278, ..., -0.0190, 0.0715, -0.0084], + [ 0.1648, -0.0995, 0.0310, ..., -0.0714, -0.0108, -0.0689], + [-0.0729, -0.2301, -0.0139, ..., -0.0485, 0.0404, -0.2110], + ..., + [-0.1666, -0.2420, 0.0592, ..., -0.0907, -0.0719, -0.1569], + [-0.0982, 0.0076, 0.0315, ..., 0.0136, -0.0995, 0.0680], + [-0.0789, -0.0559, 0.0935, ..., 0.1871, -0.0664, -0.1729]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.5483e-07, -2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 1.3318e-07, 4.1910e-08, 2.3190e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-08], + [-1.3225e-07, 2.3283e-08, 6.4261e-08, ..., 0.0000e+00, + 0.0000e+00, 2.8871e-08], + ..., + [ 1.8626e-09, 9.3132e-09, 4.8757e-05, ..., 0.0000e+00, + 0.0000e+00, 1.1973e-05], + [ 3.3993e-07, 2.1793e-07, 5.9605e-08, ..., 0.0000e+00, + 0.0000e+00, 2.0023e-07], + [ 9.1270e-08, 1.2014e-07, -4.9531e-05, ..., 0.0000e+00, + 0.0000e+00, 5.3085e-08]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0354, 0.0010, 0.0235, 0.0176, 0.0355, 0.0103, -0.0063, 0.0050, + -0.0020, -0.0437], device='cuda:0'), grad: tensor([-6.7800e-07, 1.7099e-06, -1.0431e-06, -2.2724e-05, 1.4063e-07, + -3.7346e-07, 1.7043e-07, 1.0449e-04, 6.1933e-07, -8.2254e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 218.71, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5225 re_mapping 0.0053 re_causal 0.0153 /// teacc 99.00 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.1431, 0.2067, -0.0280, ..., -0.0190, 0.0717, -0.0087], + [ 0.1651, -0.0995, 0.0307, ..., -0.0714, -0.0105, -0.0690], + [-0.0728, -0.2291, -0.0139, ..., -0.0485, 0.0401, -0.2113], + ..., + [-0.1672, -0.2446, 0.0591, ..., -0.0908, -0.0720, -0.1576], + [-0.0987, 0.0073, 0.0314, ..., 0.0136, -0.0995, 0.0678], + [-0.0790, -0.0562, 0.0941, ..., 0.1873, -0.0664, -0.1738]], + device='cuda:0'), grad: tensor([[ 1.6391e-07, -1.3039e-08, 1.7229e-07, ..., 0.0000e+00, + 1.8626e-09, 6.5193e-09], + [-1.8943e-06, 6.5193e-09, -1.7695e-06, ..., 0.0000e+00, + 1.8626e-09, 2.4214e-08], + [ 1.0347e-06, 2.7940e-09, 1.1353e-06, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + ..., + [ 7.5437e-08, 2.7940e-09, -8.7544e-08, ..., 0.0000e+00, + 0.0000e+00, 5.2154e-08], + [-1.6484e-07, -5.5786e-07, -6.5565e-07, ..., -1.5832e-08, + 8.3819e-09, -6.9104e-07], + [ 6.0536e-08, 6.7987e-08, -6.0536e-08, ..., 8.3819e-09, + 0.0000e+00, 7.0781e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0353, 0.0006, 0.0239, 0.0183, 0.0357, 0.0096, -0.0064, 0.0049, + -0.0026, -0.0435], device='cuda:0'), grad: tensor([ 5.4017e-07, -5.3644e-06, 3.2540e-06, -3.6322e-07, 4.3865e-07, + 1.2200e-07, 4.0531e-06, 2.6822e-07, -2.8349e-06, -1.2852e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 218.58, cls_loss 0.0012 cls_loss_mapping 0.0027 cls_loss_causal 0.4941 re_mapping 0.0052 re_causal 0.0152 /// teacc 98.95 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.1432, 0.2067, -0.0281, ..., -0.0192, 0.0717, -0.0099], + [ 0.1653, -0.0997, 0.0308, ..., -0.0715, -0.0105, -0.0691], + [-0.0729, -0.2307, -0.0146, ..., -0.0488, 0.0401, -0.2120], + ..., + [-0.1675, -0.2457, 0.0579, ..., -0.0908, -0.0720, -0.1608], + [-0.0990, 0.0073, 0.0341, ..., 0.0135, -0.0995, 0.0700], + [-0.0792, -0.0562, 0.0942, ..., 0.1876, -0.0664, -0.1742]], + device='cuda:0'), grad: tensor([[ 5.4948e-08, -2.4680e-07, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 8.2888e-08], + [ 3.6322e-08, 5.8673e-08, 3.2596e-08, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-08], + [ 1.3970e-08, 2.9709e-07, 4.7125e-07, ..., 0.0000e+00, + 0.0000e+00, 6.8918e-08], + ..., + [ 1.2945e-07, 2.0396e-07, -2.1514e-07, ..., 0.0000e+00, + 0.0000e+00, 1.4380e-06], + [ 1.1455e-07, -2.4680e-07, -7.1991e-07, ..., 0.0000e+00, + 0.0000e+00, -3.0734e-08], + [ 4.0606e-07, 6.0629e-07, 1.7416e-07, ..., 0.0000e+00, + 0.0000e+00, 5.6811e-07]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0351, 0.0007, 0.0237, 0.0185, 0.0358, 0.0097, -0.0064, 0.0038, + -0.0003, -0.0435], device='cuda:0'), grad: tensor([ 4.9826e-07, 2.5705e-07, 1.3588e-06, -2.9616e-07, 7.5903e-07, + -8.7395e-06, -9.6206e-07, 7.1041e-06, -1.4855e-06, 1.5469e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 218.00, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.4851 re_mapping 0.0053 re_causal 0.0152 /// teacc 99.02 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.1432, 0.2069, -0.0280, ..., -0.0192, 0.0719, -0.0097], + [ 0.1651, -0.0999, 0.0308, ..., -0.0715, -0.0105, -0.0693], + [-0.0725, -0.2320, -0.0147, ..., -0.0489, 0.0401, -0.2132], + ..., + [-0.1678, -0.2476, 0.0577, ..., -0.0910, -0.0720, -0.1615], + [-0.0992, 0.0077, 0.0341, ..., 0.0136, -0.0995, 0.0704], + [-0.0796, -0.0562, 0.0948, ..., 0.1879, -0.0665, -0.1741]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, -2.9802e-08, 2.7008e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-3.3528e-07, 9.3132e-10, -2.7195e-07, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-09], + [-2.3935e-07, 3.7253e-09, 6.9849e-08, ..., 0.0000e+00, + 0.0000e+00, 3.4459e-08], + ..., + [ 5.7742e-08, 9.3132e-10, 1.0058e-06, ..., 2.7940e-09, + 0.0000e+00, 2.0489e-08], + [ 6.4261e-08, 2.7940e-09, 8.8476e-08, ..., 9.3132e-10, + 0.0000e+00, 1.2759e-07], + [ 8.6613e-08, 1.7695e-08, -3.5316e-06, ..., -1.3970e-08, + 0.0000e+00, 3.3528e-08]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0351, 0.0004, 0.0241, 0.0191, 0.0351, 0.0094, -0.0062, 0.0032, + -0.0002, -0.0429], device='cuda:0'), grad: tensor([ 1.0524e-07, -1.4799e-06, -1.5646e-07, -1.0496e-06, 6.2287e-06, + 6.9663e-07, 7.6927e-07, 2.5667e-06, 8.1025e-07, -8.4937e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 218.61, cls_loss 0.0016 cls_loss_mapping 0.0023 cls_loss_causal 0.4951 re_mapping 0.0050 re_causal 0.0148 /// teacc 99.03 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.1434, 0.2070, -0.0280, ..., -0.0192, 0.0719, -0.0100], + [ 0.1653, -0.0999, 0.0302, ..., -0.0737, -0.0104, -0.0695], + [-0.0724, -0.2324, -0.0150, ..., -0.0491, 0.0401, -0.2138], + ..., + [-0.1685, -0.2490, 0.0579, ..., -0.0923, -0.0720, -0.1616], + [-0.1004, 0.0071, 0.0340, ..., 0.0134, -0.0997, 0.0702], + [-0.0789, -0.0564, 0.0957, ..., 0.1895, -0.0666, -0.1747]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.4550e-06, -1.1995e-06, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3039e-08, 1.0524e-07, 9.4995e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 2.3004e-07, 1.3225e-07, ..., 0.0000e+00, + -9.3132e-10, 2.7940e-09], + ..., + [ 1.3039e-08, 1.8626e-08, -4.1910e-08, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [-1.4901e-08, 1.3690e-07, 3.0734e-08, ..., -1.7695e-08, + 0.0000e+00, -4.2841e-08], + [ 7.4506e-09, 2.0117e-07, 5.8673e-08, ..., 2.7940e-09, + 0.0000e+00, 1.7695e-08]], device='cuda:0') +Epoch 269, bias, value: tensor([ 3.4854e-02, 9.9217e-05, 2.4134e-02, 1.9115e-02, 3.4884e-02, + 9.2935e-03, -5.8084e-03, 3.1154e-03, -6.2100e-04, -4.2247e-02], + device='cuda:0'), grad: tensor([-4.7535e-06, 2.9244e-07, 4.1351e-07, 1.6857e-07, 4.6566e-08, + 2.1420e-07, 3.0622e-06, -2.1420e-08, 2.5891e-07, 3.1479e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 218.37, cls_loss 0.0020 cls_loss_mapping 0.0032 cls_loss_causal 0.5256 re_mapping 0.0052 re_causal 0.0152 /// teacc 99.04 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.1437, 0.2069, -0.0288, ..., -0.0192, 0.0720, -0.0107], + [ 0.1656, -0.0997, 0.0302, ..., -0.0738, -0.0102, -0.0693], + [-0.0726, -0.2340, -0.0165, ..., -0.0494, 0.0396, -0.2150], + ..., + [-0.1689, -0.2495, 0.0586, ..., -0.0924, -0.0720, -0.1621], + [-0.0993, 0.0087, 0.0340, ..., 0.0133, -0.0998, 0.0716], + [-0.0792, -0.0557, 0.0960, ..., 0.1899, -0.0667, -0.1741]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.2852e-07, -4.9360e-08, ..., 0.0000e+00, + 0.0000e+00, 7.6368e-08], + [-7.0781e-08, 8.3819e-09, -4.4703e-08, ..., 0.0000e+00, + 0.0000e+00, 9.9652e-08], + [ 3.6322e-08, 3.9954e-07, 5.7742e-07, ..., 0.0000e+00, + 0.0000e+00, 1.4529e-06], + ..., + [ 2.2352e-08, 3.7253e-09, -2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 4.2003e-07], + [ 1.1735e-07, -2.8964e-07, -5.2806e-07, ..., 9.3132e-10, + 0.0000e+00, 5.7463e-07], + [ 6.7987e-08, 1.4808e-07, 1.0245e-08, ..., -5.5879e-09, + 0.0000e+00, 7.0501e-07]], device='cuda:0') +Epoch 270, bias, value: tensor([ 3.4455e-02, -2.0974e-05, 2.3412e-02, 1.9136e-02, 3.5503e-02, + 8.3648e-03, -5.8095e-03, 4.0677e-03, 9.4660e-05, -4.1979e-02], + device='cuda:0'), grad: tensor([ 3.2503e-07, 5.7183e-07, 8.2999e-06, -2.2918e-05, 1.7788e-07, + 1.3309e-06, -3.6694e-07, 2.8312e-06, 4.9770e-06, 4.7162e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 218.16, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4852 re_mapping 0.0051 re_causal 0.0155 /// teacc 99.03 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.1438, 0.2069, -0.0292, ..., -0.0192, 0.0719, -0.0110], + [ 0.1660, -0.0998, 0.0304, ..., -0.0739, -0.0101, -0.0695], + [-0.0723, -0.2352, -0.0170, ..., -0.0495, 0.0397, -0.2160], + ..., + [-0.1697, -0.2499, 0.0587, ..., -0.0924, -0.0720, -0.1621], + [-0.0996, 0.0088, 0.0341, ..., 0.0129, -0.0998, 0.0718], + [-0.0797, -0.0553, 0.0962, ..., 0.1905, -0.0667, -0.1736]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, -1.3039e-08, 2.5891e-07, ..., 0.0000e+00, + 0.0000e+00, 1.7695e-08], + [-1.3039e-08, 2.7940e-09, 3.0845e-06, ..., 0.0000e+00, + 0.0000e+00, 1.6112e-07], + [ 4.6566e-09, 2.7940e-09, 1.1828e-07, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 1.1176e-08, 8.3819e-09, 5.9336e-05, ..., 0.0000e+00, + 0.0000e+00, 3.0138e-06], + [ 2.9616e-07, 2.2165e-07, -1.4949e-04, ..., 0.0000e+00, + 0.0000e+00, -7.3463e-06], + [ 8.5589e-07, 6.5099e-07, 8.5592e-05, ..., -1.8626e-09, + 0.0000e+00, 5.0291e-06]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0342, 0.0002, 0.0239, 0.0186, 0.0356, 0.0081, -0.0058, 0.0039, + 0.0002, -0.0417], device='cuda:0'), grad: tensor([ 5.4855e-07, 4.1500e-06, 5.0291e-08, 5.7369e-07, 3.1758e-07, + -1.8505e-06, 2.8126e-07, 7.9513e-05, -1.9944e-04, 1.1575e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 218.51, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.5042 re_mapping 0.0054 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.1439, 0.2068, -0.0292, ..., -0.0192, 0.0721, -0.0112], + [ 0.1670, -0.0998, 0.0313, ..., -0.0755, -0.0101, -0.0695], + [-0.0724, -0.2365, -0.0172, ..., -0.0496, 0.0398, -0.2165], + ..., + [-0.1721, -0.2506, 0.0577, ..., -0.0933, -0.0721, -0.1623], + [-0.0999, 0.0088, 0.0345, ..., 0.0126, -0.0999, 0.0719], + [-0.0785, -0.0554, 0.0967, ..., 0.1916, -0.0668, -0.1739]], + device='cuda:0'), grad: tensor([[ 2.1420e-08, 1.4808e-07, 6.6170e-07, ..., 0.0000e+00, + 0.0000e+00, 2.3702e-07], + [-1.1027e-06, 4.1910e-08, -1.1818e-06, ..., 0.0000e+00, + 0.0000e+00, 6.3330e-08], + [ 1.5460e-07, 1.0710e-07, 3.0529e-06, ..., 0.0000e+00, + 0.0000e+00, 1.6252e-07], + ..., + [ 1.1642e-07, 3.8790e-07, -1.5032e-06, ..., 0.0000e+00, + 0.0000e+00, 5.8208e-07], + [ 7.0129e-07, -9.2154e-07, -2.3767e-06, ..., 0.0000e+00, + 0.0000e+00, -1.4007e-06], + [ 9.7789e-09, 1.9791e-07, 8.3167e-07, ..., 0.0000e+00, + 0.0000e+00, 2.9476e-07]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0337, 0.0008, 0.0238, 0.0186, 0.0356, 0.0082, -0.0053, 0.0030, + 0.0004, -0.0415], device='cuda:0'), grad: tensor([ 1.6084e-06, -1.9353e-06, 4.5300e-06, 4.0047e-07, -7.0035e-06, + 2.8778e-07, 7.1414e-06, -1.3728e-06, -5.8189e-06, 2.1011e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 218.53, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.5113 re_mapping 0.0052 re_causal 0.0159 /// teacc 99.03 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.1439, 0.2073, -0.0291, ..., -0.0192, 0.0721, -0.0109], + [ 0.1670, -0.1001, 0.0311, ..., -0.0757, -0.0100, -0.0697], + [-0.0724, -0.2385, -0.0175, ..., -0.0497, 0.0397, -0.2175], + ..., + [-0.1721, -0.2518, 0.0579, ..., -0.0933, -0.0721, -0.1624], + [-0.1000, 0.0092, 0.0347, ..., 0.0125, -0.0999, 0.0721], + [-0.0786, -0.0556, 0.0966, ..., 0.1918, -0.0668, -0.1743]], + device='cuda:0'), grad: tensor([[ 2.0023e-08, -2.0303e-07, -1.4389e-07, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-09], + [-7.8529e-06, 1.2573e-08, -6.2101e-06, ..., 0.0000e+00, + 4.6566e-10, 8.3819e-09], + [ 3.1665e-08, 2.3283e-09, 2.5611e-08, ..., 0.0000e+00, + 9.3132e-10, 1.1176e-08], + ..., + [ 7.7486e-06, 3.7253e-09, 6.1169e-06, ..., 1.8626e-09, + 0.0000e+00, 9.7789e-09], + [ 6.2399e-08, 1.7695e-08, 2.4680e-08, ..., 4.6566e-10, + 3.7253e-09, 3.7253e-08], + [ 4.5169e-08, 4.2841e-08, 2.5611e-08, ..., -1.0245e-08, + 0.0000e+00, 1.5367e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([ 0.0339, 0.0007, 0.0237, 0.0185, 0.0354, 0.0081, -0.0054, 0.0032, + 0.0007, -0.0416], device='cuda:0'), grad: tensor([-4.5728e-07, -1.2510e-05, 3.3062e-08, -1.3784e-07, 5.7276e-08, + 3.2270e-07, -1.1502e-07, 1.2435e-05, 2.2911e-07, 1.5507e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 218.00, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4828 re_mapping 0.0050 re_causal 0.0151 /// teacc 99.03 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.1439, 0.2077, -0.0307, ..., -0.0192, 0.0720, -0.0107], + [ 0.1670, -0.1006, 0.0307, ..., -0.0757, -0.0099, -0.0704], + [-0.0724, -0.2401, -0.0176, ..., -0.0496, 0.0397, -0.2180], + ..., + [-0.1721, -0.2523, 0.0582, ..., -0.0934, -0.0721, -0.1628], + [-0.1001, 0.0094, 0.0348, ..., 0.0125, -0.1000, 0.0723], + [-0.0790, -0.0554, 0.0975, ..., 0.1920, -0.0669, -0.1749]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 1.5832e-08, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + [-5.3085e-08, 1.1176e-08, -1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, 1.9558e-08], + [ 2.0489e-08, 1.8626e-09, 1.7136e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 3.4459e-08, 1.4901e-08, -7.5065e-07, ..., 3.7253e-09, + 0.0000e+00, 2.8871e-08], + [ 2.7940e-07, 2.1514e-07, -5.4948e-08, ..., 9.3132e-10, + 0.0000e+00, 2.9150e-07], + [ 3.9116e-08, 3.4459e-08, 2.8685e-07, ..., -2.0489e-08, + 0.0000e+00, 5.1223e-08]], device='cuda:0') +Epoch 274, bias, value: tensor([ 0.0336, 0.0002, 0.0236, 0.0191, 0.0358, 0.0073, -0.0053, 0.0035, + 0.0007, -0.0410], device='cuda:0'), grad: tensor([ 1.3970e-07, 2.2165e-07, 3.9116e-07, 1.5115e-06, -8.7693e-06, + -2.3507e-06, 5.4110e-07, -1.5777e-06, 5.6252e-07, 9.3132e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 217.70, cls_loss 0.0011 cls_loss_mapping 0.0024 cls_loss_causal 0.5071 re_mapping 0.0051 re_causal 0.0151 /// teacc 99.03 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.1440, 0.2080, -0.0307, ..., -0.0192, 0.0719, -0.0109], + [ 0.1674, -0.1004, 0.0309, ..., -0.0759, -0.0094, -0.0706], + [-0.0725, -0.2406, -0.0177, ..., -0.0499, 0.0395, -0.2185], + ..., + [-0.1726, -0.2529, 0.0581, ..., -0.0935, -0.0722, -0.1629], + [-0.1007, 0.0093, 0.0350, ..., 0.0125, -0.1001, 0.0722], + [-0.0793, -0.0557, 0.0979, ..., 0.1923, -0.0669, -0.1754]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -5.4017e-08, -1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-3.7253e-09, 2.7940e-09, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-08], + [ 0.0000e+00, 1.8626e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 5.6811e-08], + ..., + [ 2.7940e-09, 1.8626e-09, -4.6566e-08, ..., 0.0000e+00, + 0.0000e+00, 6.6124e-08], + [ 1.6764e-08, 6.5193e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -1.9651e-07], + [ 9.3132e-09, 3.6322e-08, 1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08]], device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0336, 0.0004, 0.0236, 0.0191, 0.0351, 0.0075, -0.0055, 0.0033, + 0.0007, -0.0406], device='cuda:0'), grad: tensor([ 8.9407e-08, 9.3132e-08, -2.3842e-07, -5.8580e-07, 4.9360e-08, + 4.4983e-07, 1.6205e-07, 2.2445e-07, -3.5763e-07, 1.1642e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 218.62, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4845 re_mapping 0.0053 re_causal 0.0149 /// teacc 99.05 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.1443, 0.2081, -0.0308, ..., -0.0192, 0.0718, -0.0116], + [ 0.1680, -0.0983, 0.0319, ..., -0.0759, -0.0087, -0.0694], + [-0.0725, -0.2418, -0.0183, ..., -0.0500, 0.0395, -0.2197], + ..., + [-0.1730, -0.2564, 0.0582, ..., -0.0936, -0.0722, -0.1632], + [-0.1022, 0.0090, 0.0342, ..., 0.0124, -0.1003, 0.0720], + [-0.0806, -0.0561, 0.0974, ..., 0.1907, -0.0670, -0.1769]], + device='cuda:0'), grad: tensor([[ 3.4459e-08, -1.7695e-08, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 3.4459e-08], + [-5.0291e-08, 6.5193e-09, -7.4506e-09, ..., 4.6566e-09, + 0.0000e+00, 1.4901e-08], + [ 3.7253e-09, 1.5832e-08, 2.7008e-08, ..., 0.0000e+00, + 3.7253e-09, 6.4261e-08], + ..., + [ 2.7008e-08, 1.1176e-08, -2.5425e-07, ..., 1.8626e-09, + 0.0000e+00, 1.9558e-08], + [ 7.5065e-07, 2.0117e-07, -1.4808e-07, ..., 9.3132e-10, + -1.2573e-07, -3.1386e-07], + [ 1.8626e-08, 2.6077e-08, 2.7940e-09, ..., -9.3132e-09, + 0.0000e+00, 1.7695e-08]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0335, 0.0010, 0.0234, 0.0194, 0.0365, 0.0061, -0.0041, 0.0035, + -0.0004, -0.0417], device='cuda:0'), grad: tensor([ 1.0710e-07, 5.0291e-08, 1.6764e-08, 3.1032e-06, 2.2911e-06, + -5.4613e-06, 2.3916e-06, 2.3376e-07, 4.8336e-07, -3.2261e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 218.52, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4960 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.01 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.1445, 0.2081, -0.0308, ..., -0.0192, 0.0718, -0.0125], + [ 0.1681, -0.0983, 0.0316, ..., -0.0762, -0.0086, -0.0695], + [-0.0725, -0.2394, -0.0185, ..., -0.0502, 0.0394, -0.2203], + ..., + [-0.1731, -0.2570, 0.0583, ..., -0.0939, -0.0722, -0.1635], + [-0.1028, 0.0088, 0.0342, ..., 0.0124, -0.1003, 0.0718], + [-0.0810, -0.0564, 0.0979, ..., 0.1909, -0.0670, -0.1771]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-5.6811e-08, 9.3132e-10, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 3.1665e-08, 2.7940e-09, 5.2154e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + ..., + [ 1.0245e-08, 0.0000e+00, -1.6298e-07, ..., 0.0000e+00, + 0.0000e+00, 3.4459e-08], + [ 1.0245e-08, -5.4017e-08, -3.4459e-08, ..., 0.0000e+00, + 0.0000e+00, -9.4064e-08], + [ 6.5193e-09, 3.7253e-09, 5.9605e-08, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0332, 0.0008, 0.0239, 0.0194, 0.0373, 0.0060, -0.0038, 0.0035, + -0.0007, -0.0419], device='cuda:0'), grad: tensor([ 6.9849e-08, 1.0151e-07, 1.3411e-07, -9.9279e-07, 1.7695e-08, + 3.2131e-07, -2.6450e-07, 2.9989e-07, 7.4506e-09, 3.1944e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 218.47, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4714 re_mapping 0.0051 re_causal 0.0148 /// teacc 99.02 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.1446, 0.2085, -0.0307, ..., -0.0192, 0.0717, -0.0130], + [ 0.1690, -0.0987, 0.0318, ..., -0.0762, -0.0084, -0.0697], + [-0.0727, -0.2400, -0.0189, ..., -0.0504, 0.0394, -0.2215], + ..., + [-0.1742, -0.2573, 0.0584, ..., -0.0939, -0.0722, -0.1635], + [-0.1033, 0.0088, 0.0340, ..., 0.0124, -0.1006, 0.0718], + [-0.0814, -0.0566, 0.0982, ..., 0.1909, -0.0672, -0.1774]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 1.9893e-06, 4.1276e-05, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-3.6694e-07, 9.3132e-10, -5.7742e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 3.2596e-08, 1.7695e-08, 4.0885e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 1.1176e-08, 5.5879e-09, -2.7567e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.4773e-07, 1.6764e-08, 5.0850e-07, ..., 0.0000e+00, + 0.0000e+00, 2.8871e-08], + [ 6.3330e-08, -2.0228e-06, -4.1634e-05, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09]], device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0334, 0.0013, 0.0236, 0.0191, 0.0370, 0.0063, -0.0038, 0.0033, + -0.0008, -0.0420], device='cuda:0'), grad: tensor([ 6.3717e-05, -8.7917e-07, 6.1560e-07, 1.0803e-07, 2.4866e-07, + -1.3411e-07, 4.7497e-08, -2.2259e-07, 7.8138e-07, -6.4313e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 218.25, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4959 re_mapping 0.0051 re_causal 0.0145 /// teacc 99.02 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.1448, 0.2086, -0.0309, ..., -0.0193, 0.0714, -0.0132], + [ 0.1692, -0.0986, 0.0318, ..., -0.0762, -0.0083, -0.0697], + [-0.0728, -0.2402, -0.0191, ..., -0.0507, 0.0393, -0.2225], + ..., + [-0.1744, -0.2575, 0.0587, ..., -0.0940, -0.0721, -0.1637], + [-0.1038, 0.0088, 0.0337, ..., 0.0106, -0.1006, 0.0717], + [-0.0818, -0.0564, 0.0983, ..., 0.1917, -0.0676, -0.1766]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, -2.7940e-09, 9.1270e-08, ..., 0.0000e+00, + 0.0000e+00, 1.9558e-08], + [-8.9686e-07, 1.1176e-08, -5.8766e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 2.1700e-07, 1.3970e-08, 3.3248e-07, ..., 0.0000e+00, + 0.0000e+00, 1.9558e-08], + ..., + [ 2.4680e-07, 2.7940e-08, -8.8289e-06, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-08], + [ 2.6636e-07, -9.3132e-10, 2.0303e-07, ..., 0.0000e+00, + 0.0000e+00, -3.2596e-08], + [ 1.0291e-06, 7.3295e-07, 8.5086e-06, ..., -9.3132e-10, + 0.0000e+00, 5.6531e-07]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0332, 0.0013, 0.0236, 0.0189, 0.0371, 0.0064, -0.0036, 0.0036, + -0.0013, -0.0420], device='cuda:0'), grad: tensor([ 2.2445e-07, -1.3886e-06, 7.3481e-07, 3.3807e-07, 1.2107e-08, + -1.8589e-06, 3.2224e-07, -1.4625e-05, 5.0105e-07, 1.5736e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 218.54, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4724 re_mapping 0.0052 re_causal 0.0150 /// teacc 98.97 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.1449, 0.2087, -0.0311, ..., -0.0195, 0.0716, -0.0135], + [ 0.1701, -0.0987, 0.0330, ..., -0.0763, -0.0082, -0.0698], + [-0.0728, -0.2404, -0.0195, ..., -0.0509, 0.0390, -0.2233], + ..., + [-0.1759, -0.2581, 0.0581, ..., -0.0941, -0.0721, -0.1640], + [-0.1044, 0.0086, 0.0337, ..., 0.0103, -0.1006, 0.0716], + [-0.0823, -0.0564, 0.0983, ..., 0.1920, -0.0676, -0.1770]], + device='cuda:0'), grad: tensor([[ 8.8476e-08, 0.0000e+00, 1.6205e-07, ..., 3.1665e-08, + 4.6566e-09, 3.3528e-08], + [-4.0699e-07, 1.8161e-07, 9.3132e-10, ..., 1.8626e-09, + -3.9116e-08, 2.2911e-07], + [ 5.5879e-07, 1.8626e-09, 3.1386e-07, ..., 5.6811e-08, + 5.3085e-08, 3.7253e-09], + ..., + [ 8.4750e-08, 5.5879e-09, 2.8871e-08, ..., 2.7940e-09, + 4.6566e-09, 1.3970e-08], + [ 3.1851e-07, 2.2817e-07, 1.8626e-08, ..., 2.7940e-09, + 3.4459e-08, 2.9150e-07], + [ 8.6613e-08, 2.1420e-08, -9.7044e-07, ..., -1.7602e-07, + 9.3132e-10, 3.0734e-08]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0332, 0.0021, 0.0233, 0.0189, 0.0371, 0.0066, -0.0037, 0.0032, + -0.0014, -0.0421], device='cuda:0'), grad: tensor([ 7.3947e-07, -1.9278e-07, 1.6643e-06, 1.1520e-06, 1.4817e-06, + -8.0317e-06, 5.5507e-06, 1.7509e-07, 6.6031e-07, -3.2112e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 218.30, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4845 re_mapping 0.0052 re_causal 0.0149 /// teacc 99.05 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.1450, 0.2091, -0.0309, ..., -0.0194, 0.0716, -0.0138], + [ 0.1703, -0.0991, 0.0330, ..., -0.0768, -0.0082, -0.0706], + [-0.0729, -0.2406, -0.0197, ..., -0.0516, 0.0396, -0.2243], + ..., + [-0.1762, -0.2584, 0.0580, ..., -0.0942, -0.0723, -0.1642], + [-0.1047, 0.0088, 0.0338, ..., 0.0109, -0.1003, 0.0719], + [-0.0827, -0.0569, 0.0984, ..., 0.1921, -0.0675, -0.1777]], + device='cuda:0'), grad: tensor([[ 3.0734e-07, -5.0943e-07, -1.0617e-07, ..., 4.7497e-08, + 0.0000e+00, 3.2503e-07], + [ 2.2445e-07, 2.2072e-07, 4.4797e-07, ..., 5.5879e-09, + 0.0000e+00, 2.7008e-07], + [ 6.7987e-08, 1.0058e-07, 5.3551e-07, ..., 7.1712e-08, + 0.0000e+00, 8.1956e-08], + ..., + [ 2.2724e-07, 1.5739e-07, -1.3486e-06, ..., 3.7253e-09, + 0.0000e+00, 2.1886e-07], + [ 1.3769e-05, 9.6262e-06, -2.5518e-07, ..., -3.7253e-09, + 0.0000e+00, 1.4879e-05], + [ 3.1274e-06, 2.0303e-06, -9.9745e-07, ..., -1.8440e-07, + 0.0000e+00, 2.1700e-06]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0333, 0.0020, 0.0233, 0.0191, 0.0371, 0.0064, -0.0037, 0.0032, + -0.0012, -0.0422], device='cuda:0'), grad: tensor([ 6.5845e-07, 2.4680e-06, 1.9092e-06, 3.4541e-05, 1.4387e-05, + -7.5459e-05, 5.4389e-06, 2.2911e-06, 2.9281e-05, -1.5587e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 218.32, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.5214 re_mapping 0.0051 re_causal 0.0145 /// teacc 99.06 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.1460, 0.2095, -0.0310, ..., -0.0195, 0.0715, -0.0146], + [ 0.1710, -0.0991, 0.0329, ..., -0.0783, -0.0080, -0.0708], + [-0.0730, -0.2410, -0.0223, ..., -0.0540, 0.0397, -0.2252], + ..., + [-0.1773, -0.2594, 0.0580, ..., -0.0965, -0.0722, -0.1644], + [-0.1058, 0.0085, 0.0344, ..., 0.0115, -0.1001, 0.0716], + [-0.0826, -0.0572, 0.0994, ..., 0.1944, -0.0671, -0.1781]], + device='cuda:0'), grad: tensor([[ 1.4808e-07, -9.2201e-08, -1.4901e-08, ..., 1.8626e-09, + 3.7253e-08, 4.0978e-08], + [ 8.4460e-05, 2.5511e-05, 6.3330e-08, ..., 4.9267e-07, + 3.8669e-06, 2.9564e-05], + [ 1.0245e-07, 3.2596e-08, 4.1910e-08, ..., 9.3132e-10, + 4.6566e-09, 3.9116e-08], + ..., + [ 6.9849e-08, 2.2352e-08, 1.8626e-09, ..., 1.2573e-07, + 2.7940e-09, 2.4214e-08], + [ 5.9418e-06, 1.8179e-06, 3.4459e-08, ..., 5.7742e-08, + 3.0268e-07, 2.0415e-06], + [ 1.3849e-06, 4.7963e-07, -2.7753e-07, ..., -1.9092e-07, + 5.7742e-08, 5.2061e-07]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0332, 0.0021, 0.0226, 0.0202, 0.0372, 0.0057, -0.0039, 0.0027, + -0.0010, -0.0413], device='cuda:0'), grad: tensor([ 2.2538e-07, 8.8155e-05, 6.0536e-08, 3.6299e-05, -1.3381e-05, + -1.2934e-04, -1.2470e-06, 1.8999e-07, 6.3442e-06, 1.2636e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 217.81, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.4842 re_mapping 0.0051 re_causal 0.0152 /// teacc 99.08 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.1464, 0.2099, -0.0310, ..., -0.0195, 0.0707, -0.0146], + [ 0.1734, -0.0998, 0.0358, ..., -0.0784, -0.0076, -0.0719], + [-0.0732, -0.2412, -0.0228, ..., -0.0542, 0.0394, -0.2258], + ..., + [-0.1804, -0.2598, 0.0556, ..., -0.0967, -0.0725, -0.1645], + [-0.1063, 0.0086, 0.0343, ..., 0.0116, -0.1004, 0.0716], + [-0.0834, -0.0575, 0.0991, ..., 0.1943, -0.0675, -0.1788]], + device='cuda:0'), grad: tensor([[ 7.5437e-08, -3.2131e-07, -3.2596e-07, ..., 0.0000e+00, + 9.2201e-08, 6.5193e-09], + [-1.1921e-07, 5.5879e-09, 1.2480e-07, ..., -7.4506e-09, + 2.6077e-08, 0.0000e+00], + [-5.6811e-08, 4.6566e-09, 2.2352e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 7.7300e-08, 2.7940e-09, -3.5390e-07, ..., 2.7940e-09, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-08, 1.4901e-08, 3.3528e-08, ..., 9.3132e-10, + 1.5646e-07, 3.7253e-09], + [ 7.1712e-08, 2.7101e-07, 3.9861e-07, ..., 4.6566e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0330, 0.0045, 0.0224, 0.0195, 0.0380, 0.0065, -0.0034, 0.0001, + -0.0011, -0.0419], device='cuda:0'), grad: tensor([-1.3690e-07, 6.2399e-07, -4.1537e-07, 5.6811e-08, -2.7567e-07, + 1.6550e-06, -3.1181e-06, -3.3807e-07, 8.0280e-07, 1.1511e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 217.79, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.5089 re_mapping 0.0053 re_causal 0.0151 /// teacc 99.01 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.1467, 0.2101, -0.0310, ..., -0.0194, 0.0701, -0.0157], + [ 0.1733, -0.0997, 0.0346, ..., -0.0798, -0.0061, -0.0720], + [-0.0733, -0.2414, -0.0229, ..., -0.0550, 0.0385, -0.2271], + ..., + [-0.1802, -0.2601, 0.0560, ..., -0.0990, -0.0728, -0.1646], + [-0.1071, 0.0085, 0.0343, ..., 0.0115, -0.1007, 0.0714], + [-0.0836, -0.0580, 0.1010, ..., 0.1961, -0.0691, -0.1792]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -5.2154e-08, -1.8626e-08, ..., 9.3132e-10, + -3.6322e-08, 3.7253e-08], + [ 1.0692e-06, 3.9116e-08, 2.0303e-06, ..., 0.0000e+00, + 1.3039e-08, 4.3772e-08], + [ 4.6566e-09, 2.4214e-08, 3.4459e-08, ..., 9.3132e-10, + 0.0000e+00, 4.2841e-08], + ..., + [-1.1642e-06, 6.5193e-09, -2.2836e-06, ..., 9.3132e-10, + 9.3132e-10, 7.9162e-08], + [ 4.9360e-08, -6.1467e-08, -1.9092e-07, ..., -3.7253e-09, + 7.4506e-09, -2.5518e-07], + [ 7.0781e-08, 1.3970e-08, 2.7195e-07, ..., -4.6566e-09, + 3.7253e-09, 8.3819e-09]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0329, 0.0035, 0.0224, 0.0196, 0.0380, 0.0067, -0.0034, 0.0006, + -0.0013, -0.0406], device='cuda:0'), grad: tensor([ 5.1223e-08, 3.8147e-06, 1.0617e-07, 1.2107e-07, -1.2610e-06, + -7.2364e-07, 8.2329e-07, -4.1053e-06, -5.5507e-07, 1.7118e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 218.04, cls_loss 0.0016 cls_loss_mapping 0.0023 cls_loss_causal 0.4911 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.07 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.1469, 0.2102, -0.0310, ..., -0.0194, 0.0702, -0.0161], + [ 0.1759, -0.0997, 0.0366, ..., -0.0797, -0.0060, -0.0722], + [-0.0733, -0.2417, -0.0231, ..., -0.0552, 0.0384, -0.2279], + ..., + [-0.1828, -0.2605, 0.0542, ..., -0.0990, -0.0729, -0.1648], + [-0.1078, 0.0083, 0.0336, ..., 0.0088, -0.1007, 0.0714], + [-0.0833, -0.0567, 0.1013, ..., 0.1974, -0.0701, -0.1769]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, -1.5870e-06, -1.0403e-06, ..., 0.0000e+00, + -1.9372e-07, 1.8626e-09], + [-3.1106e-07, 1.0803e-07, 1.0980e-06, ..., 0.0000e+00, + 1.3970e-08, 5.5879e-09], + [ 1.2107e-08, 7.6368e-08, 1.9092e-07, ..., 0.0000e+00, + 9.3132e-09, 6.5193e-09], + ..., + [ 1.4622e-07, 1.0272e-06, -5.9307e-06, ..., 0.0000e+00, + 1.3225e-07, 5.5879e-09], + [ 4.5635e-07, 2.9057e-07, 5.2247e-07, ..., 0.0000e+00, + 6.5193e-09, 4.0513e-07], + [ 4.8429e-08, 1.8440e-07, 3.7886e-06, ..., 0.0000e+00, + 1.5832e-08, 2.5146e-08]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0327, 0.0055, 0.0226, 0.0193, 0.0374, 0.0065, -0.0033, -0.0013, + -0.0020, -0.0399], device='cuda:0'), grad: tensor([-2.5015e-06, 3.0063e-06, -1.0310e-06, 1.1167e-06, 1.0785e-06, + 3.9116e-08, 2.5611e-07, -1.3418e-05, 2.4959e-06, 8.9481e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 217.95, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4865 re_mapping 0.0052 re_causal 0.0149 /// teacc 99.00 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.1471, 0.2108, -0.0308, ..., -0.0194, 0.0693, -0.0166], + [ 0.1762, -0.0998, 0.0366, ..., -0.0803, -0.0073, -0.0724], + [-0.0732, -0.2422, -0.0230, ..., -0.0554, 0.0409, -0.2289], + ..., + [-0.1831, -0.2611, 0.0542, ..., -0.0994, -0.0715, -0.1650], + [-0.1091, 0.0078, 0.0337, ..., 0.0088, -0.1009, 0.0708], + [-0.0838, -0.0574, 0.1013, ..., 0.1977, -0.0742, -0.1774]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.6764e-08, -5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 2.7940e-09], + [-5.5879e-09, 3.7253e-09, 2.0489e-08, ..., 0.0000e+00, + 1.8626e-09, 1.0245e-08], + [ 7.4506e-09, 5.6811e-08, 7.7300e-08, ..., 0.0000e+00, + 5.5879e-09, 7.2643e-08], + ..., + [-3.8464e-07, 1.8626e-09, -9.4995e-08, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-09], + [ 9.3132e-10, -8.1956e-08, -1.2014e-07, ..., 0.0000e+00, + 4.6566e-09, -7.3574e-08], + [ 1.2107e-08, 1.1176e-08, 1.0245e-08, ..., -2.7940e-09, + 9.3132e-10, 7.4506e-09]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0327, 0.0054, 0.0229, 0.0194, 0.0403, 0.0073, -0.0034, -0.0014, + -0.0025, -0.0418], device='cuda:0'), grad: tensor([ 3.9116e-08, 9.7789e-08, -1.5926e-07, -2.6450e-07, 1.6745e-06, + 1.1073e-06, 1.3970e-08, -2.5183e-06, -1.7881e-07, 2.0303e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 217.90, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5054 re_mapping 0.0050 re_causal 0.0148 /// teacc 99.08 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.1476, 0.2119, -0.0306, ..., -0.0194, 0.0698, -0.0172], + [ 0.1762, -0.1000, 0.0365, ..., -0.0807, -0.0068, -0.0727], + [-0.0732, -0.2428, -0.0233, ..., -0.0556, 0.0409, -0.2298], + ..., + [-0.1831, -0.2618, 0.0544, ..., -0.0994, -0.0716, -0.1655], + [-0.1095, 0.0082, 0.0340, ..., 0.0089, -0.1000, 0.0713], + [-0.0839, -0.0583, 0.1011, ..., 0.1979, -0.0757, -0.1783]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.6077e-08, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 8.8476e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-09, 9.3132e-10, -1.5367e-07, ..., 3.7253e-09, + 0.0000e+00, 9.3132e-10], + [ 4.3772e-08, 2.6077e-08, 6.5193e-09, ..., 1.8626e-09, + 0.0000e+00, 3.8184e-08], + [ 2.8871e-08, 3.3528e-08, 9.3132e-10, ..., -1.3039e-08, + 0.0000e+00, 3.1665e-08]], device='cuda:0') +Epoch 287, bias, value: tensor([ 0.0332, 0.0053, 0.0229, 0.0201, 0.0404, 0.0069, -0.0040, -0.0012, + -0.0020, -0.0424], device='cuda:0'), grad: tensor([ 4.9360e-08, 3.8091e-07, -7.4785e-07, 3.8370e-07, -5.6811e-08, + -1.9837e-07, -2.7940e-09, -7.0781e-08, 1.3690e-07, 1.5274e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 218.30, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.5149 re_mapping 0.0051 re_causal 0.0149 /// teacc 99.04 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.1478, 0.2125, -0.0304, ..., -0.0195, 0.0695, -0.0174], + [ 0.1768, -0.1008, 0.0365, ..., -0.0811, -0.0058, -0.0729], + [-0.0751, -0.2432, -0.0243, ..., -0.0557, 0.0379, -0.2305], + ..., + [-0.1832, -0.2623, 0.0545, ..., -0.0995, -0.0716, -0.1659], + [-0.1102, 0.0074, 0.0340, ..., 0.0073, -0.1001, 0.0711], + [-0.0843, -0.0578, 0.1021, ..., 0.1991, -0.0770, -0.1777]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, -3.1386e-07, -2.0955e-07, ..., -2.7940e-09, + 0.0000e+00, 1.4901e-08], + [ 1.8626e-09, 2.8871e-08, 2.1420e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 9.3132e-10, 2.5146e-08, 2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, 3.1665e-08], + ..., + [ 3.7253e-09, 1.0245e-08, -2.5146e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 3.6322e-08, 6.2399e-08, -2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, -4.5635e-08], + [ 4.1910e-08, 1.3132e-07, 9.0338e-08, ..., 2.7940e-09, + 0.0000e+00, 6.1467e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0336, 0.0055, 0.0217, 0.0195, 0.0381, 0.0071, -0.0037, -0.0012, + -0.0024, -0.0405], device='cuda:0'), grad: tensor([ 4.7125e-07, 8.8476e-08, 1.8626e-07, 2.1309e-06, 1.0943e-06, + -2.3413e-06, -1.9595e-06, -2.2352e-08, -5.4017e-08, 3.9209e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 217.94, cls_loss 0.0014 cls_loss_mapping 0.0021 cls_loss_causal 0.5366 re_mapping 0.0049 re_causal 0.0148 /// teacc 99.00 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.1479, 0.2136, -0.0295, ..., -0.0176, 0.0695, -0.0182], + [ 0.1769, -0.1017, 0.0365, ..., -0.0812, -0.0057, -0.0735], + [-0.0751, -0.2435, -0.0245, ..., -0.0560, 0.0379, -0.2313], + ..., + [-0.1832, -0.2629, 0.0547, ..., -0.0997, -0.0717, -0.1661], + [-0.1108, 0.0074, 0.0344, ..., 0.0066, -0.1000, 0.0710], + [-0.0870, -0.0599, 0.1013, ..., 0.1987, -0.0777, -0.1805]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.8184e-08, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 4.6566e-09, 4.6566e-09, 7.4506e-09, ..., 9.3132e-10, + 0.0000e+00, 5.5879e-09], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + ..., + [ 3.7253e-09, 1.8626e-09, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 5.5879e-09], + [ 2.9802e-08, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 4.0047e-08], + [ 9.3132e-10, 1.0245e-08, -1.4994e-07, ..., -2.0489e-08, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0337, 0.0055, 0.0216, 0.0189, 0.0378, 0.0081, -0.0027, -0.0010, + -0.0027, -0.0413], device='cuda:0'), grad: tensor([-1.8626e-08, 1.4529e-07, -6.0443e-07, -3.2820e-06, 5.7090e-07, + 3.0212e-06, 5.0291e-08, 3.7625e-07, 2.2911e-07, -4.7963e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 217.77, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.5185 re_mapping 0.0051 re_causal 0.0150 /// teacc 99.07 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.1484, 0.2138, -0.0294, ..., -0.0176, 0.0718, -0.0185], + [ 0.1769, -0.1020, 0.0362, ..., -0.0814, -0.0057, -0.0740], + [-0.0751, -0.2439, -0.0247, ..., -0.0580, 0.0378, -0.2322], + ..., + [-0.1832, -0.2638, 0.0550, ..., -0.0997, -0.0728, -0.1663], + [-0.1112, 0.0075, 0.0346, ..., 0.0065, -0.0999, 0.0711], + [-0.0873, -0.0601, 0.1012, ..., 0.1990, -0.0753, -0.1811]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.4459e-08, 0.0000e+00, 5.8208e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 1.8626e-09, 0.0000e+00, 9.7789e-08, ..., 0.0000e+00, + 0.0000e+00, 1.6764e-08], + ..., + [ 4.6566e-09, 0.0000e+00, -4.7833e-06, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 1.4901e-08, 0.0000e+00, 8.2888e-08, ..., 0.0000e+00, + 0.0000e+00, -6.6124e-08], + [ 9.3132e-09, 0.0000e+00, 6.5938e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0333, 0.0052, 0.0216, 0.0193, 0.0381, 0.0078, -0.0026, -0.0008, + -0.0026, -0.0416], device='cuda:0'), grad: tensor([ 2.0489e-08, 1.3290e-06, 2.1607e-07, 1.0356e-05, 5.2154e-08, + 1.8626e-08, 4.1910e-08, -1.4029e-05, 1.2852e-07, 1.9046e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 217.59, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4812 re_mapping 0.0052 re_causal 0.0146 /// teacc 98.96 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.1486, 0.2142, -0.0294, ..., -0.0177, 0.0719, -0.0189], + [ 0.1769, -0.1023, 0.0359, ..., -0.0814, -0.0058, -0.0742], + [-0.0749, -0.2440, -0.0245, ..., -0.0581, 0.0380, -0.2326], + ..., + [-0.1832, -0.2649, 0.0559, ..., -0.0998, -0.0730, -0.1664], + [-0.1131, 0.0065, 0.0345, ..., 0.0065, -0.1013, 0.0706], + [-0.0876, -0.0603, 0.1004, ..., 0.1991, -0.0751, -0.1814]], + device='cuda:0'), grad: tensor([[ 2.8871e-08, -1.1176e-08, 5.8673e-08, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [-4.5728e-07, -1.4156e-07, 6.4261e-08, ..., 0.0000e+00, + -1.9930e-07, 1.3039e-08], + [ 3.1665e-08, 9.3132e-10, 6.7055e-08, ..., 0.0000e+00, + 9.3132e-10, 5.1223e-08], + ..., + [ 2.7940e-09, 1.1176e-08, -1.7378e-06, ..., 0.0000e+00, + 9.3132e-10, 1.3597e-07], + [ 1.4808e-07, 2.7008e-08, 1.3225e-07, ..., 0.0000e+00, + 6.2399e-08, -2.2352e-08], + [ 2.8871e-08, 3.0734e-08, 8.3260e-07, ..., 0.0000e+00, + 9.3132e-10, 2.4214e-08]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0334, 0.0049, 0.0221, 0.0191, 0.0410, 0.0082, -0.0027, -0.0017, + -0.0031, -0.0418], device='cuda:0'), grad: tensor([ 2.3562e-07, 2.7567e-07, 3.8091e-07, -8.3912e-07, -3.4511e-05, + 5.1968e-07, 7.9442e-07, 2.9683e-05, 4.4238e-07, 3.0287e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 217.93, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.5123 re_mapping 0.0050 re_causal 0.0154 /// teacc 98.96 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.1489, 0.2140, -0.0295, ..., -0.0177, 0.0708, -0.0213], + [ 0.1769, -0.1022, 0.0355, ..., -0.0815, -0.0059, -0.0743], + [-0.0749, -0.2442, -0.0244, ..., -0.0582, 0.0387, -0.2331], + ..., + [-0.1832, -0.2639, 0.0564, ..., -0.0998, -0.0724, -0.1666], + [-0.1134, 0.0066, 0.0343, ..., 0.0065, -0.1015, 0.0708], + [-0.0878, -0.0604, 0.1001, ..., 0.1992, -0.0756, -0.1815]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 7.4506e-08, 1.6298e-07, ..., 0.0000e+00, + 0.0000e+00, 9.9652e-08], + [-8.7544e-08, 3.2596e-08, 4.1910e-08, ..., 0.0000e+00, + -2.7940e-09, 1.0617e-07], + [ 4.6566e-09, 2.7940e-08, 2.9150e-07, ..., 0.0000e+00, + 9.3132e-10, 6.4261e-08], + ..., + [ 2.3283e-08, 2.3656e-07, 1.2293e-07, ..., 0.0000e+00, + 9.3132e-10, 3.1479e-07], + [ 4.9360e-08, -5.3365e-07, -1.1660e-06, ..., 0.0000e+00, + 9.3132e-10, -6.5379e-07], + [ 3.9116e-08, 1.5274e-07, 3.4459e-07, ..., 0.0000e+00, + 0.0000e+00, 1.6857e-07]], device='cuda:0') +Epoch 292, bias, value: tensor([ 0.0330, 0.0046, 0.0222, 0.0202, 0.0410, 0.0075, -0.0028, -0.0013, + -0.0034, -0.0421], device='cuda:0'), grad: tensor([ 6.0722e-07, 3.4273e-07, 4.7404e-07, -1.2003e-05, 3.0734e-08, + 1.2003e-05, 7.0781e-08, 9.7603e-07, -3.4794e-06, 9.8720e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 218.87, cls_loss 0.0017 cls_loss_mapping 0.0020 cls_loss_causal 0.5146 re_mapping 0.0048 re_causal 0.0138 /// teacc 98.99 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.1496, 0.2148, -0.0294, ..., -0.0177, 0.0706, -0.0222], + [ 0.1771, -0.1020, 0.0355, ..., -0.0816, -0.0045, -0.0744], + [-0.0750, -0.2448, -0.0243, ..., -0.0583, 0.0386, -0.2344], + ..., + [-0.1832, -0.2648, 0.0562, ..., -0.0998, -0.0725, -0.1672], + [-0.1156, 0.0056, 0.0353, ..., 0.0066, -0.1030, 0.0702], + [-0.0882, -0.0610, 0.1006, ..., 0.1992, -0.0757, -0.1836]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.8818e-06, -9.2201e-08, ..., 0.0000e+00, + 0.0000e+00, -1.2927e-06], + [-1.8626e-09, 1.0431e-07, 7.1712e-08, ..., 9.3132e-10, + 1.8626e-09, 3.5390e-08], + [ 0.0000e+00, 5.0291e-08, 3.9116e-08, ..., 0.0000e+00, + -1.6764e-08, 1.0245e-08], + ..., + [ 0.0000e+00, 1.9558e-08, -1.9185e-07, ..., 9.3132e-10, + 1.4901e-08, 3.7253e-09], + [ 0.0000e+00, 1.8813e-07, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 6.2399e-08], + [ 0.0000e+00, 2.7101e-07, -5.5879e-08, ..., -3.9116e-08, + 0.0000e+00, 2.9802e-08]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0331, 0.0046, 0.0224, 0.0205, 0.0405, 0.0077, -0.0024, -0.0014, + -0.0039, -0.0417], device='cuda:0'), grad: tensor([-7.4096e-06, 3.7160e-07, -2.6077e-08, 3.1479e-07, 4.3772e-08, + 5.6904e-07, 5.3234e-06, -1.7509e-07, 3.6880e-07, 6.1374e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 218.12, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4860 re_mapping 0.0051 re_causal 0.0141 /// teacc 99.04 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.1499, 0.2153, -0.0296, ..., -0.0177, 0.0707, -0.0226], + [ 0.1777, -0.1022, 0.0358, ..., -0.0818, -0.0045, -0.0747], + [-0.0751, -0.2452, -0.0249, ..., -0.0582, 0.0386, -0.2360], + ..., + [-0.1838, -0.2658, 0.0557, ..., -0.1000, -0.0725, -0.1678], + [-0.1166, 0.0051, 0.0354, ..., 0.0065, -0.1040, 0.0692], + [-0.0884, -0.0612, 0.1015, ..., 0.1996, -0.0758, -0.1839]], + device='cuda:0'), grad: tensor([[ 2.6450e-07, 4.9733e-07, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.9116e-07], + [ 5.9139e-08, 1.2014e-07, 4.1910e-09, ..., 0.0000e+00, + 0.0000e+00, 9.1735e-08], + [ 1.5078e-06, 2.9951e-06, 2.0955e-08, ..., 0.0000e+00, + 0.0000e+00, 2.2464e-06], + ..., + [ 2.9104e-07, 5.7556e-07, -1.5367e-08, ..., 0.0000e+00, + 0.0000e+00, 4.3353e-07], + [ 2.9914e-06, 5.9307e-06, 3.7719e-08, ..., 0.0000e+00, + 0.0000e+00, 4.4554e-06], + [ 6.9384e-08, 1.3830e-07, -1.2107e-08, ..., -4.6566e-10, + 0.0000e+00, 1.0291e-07]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0330, 0.0049, 0.0224, 0.0210, 0.0410, 0.0077, -0.0024, -0.0021, + -0.0046, -0.0412], device='cuda:0'), grad: tensor([ 1.0189e-06, 2.3516e-07, 5.1819e-06, 2.5928e-05, -5.2154e-08, + -4.6074e-05, 1.5171e-06, 1.0058e-06, 1.0803e-05, 4.5169e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 218.03, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4814 re_mapping 0.0049 re_causal 0.0141 /// teacc 99.06 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.1501, 0.2159, -0.0301, ..., -0.0177, 0.0713, -0.0229], + [ 0.1779, -0.1023, 0.0359, ..., -0.0822, -0.0043, -0.0749], + [-0.0752, -0.2456, -0.0253, ..., -0.0582, 0.0384, -0.2356], + ..., + [-0.1840, -0.2664, 0.0537, ..., -0.1000, -0.0727, -0.1682], + [-0.1168, 0.0050, 0.0355, ..., 0.0065, -0.1041, 0.0694], + [-0.0888, -0.0615, 0.1042, ..., 0.1998, -0.0758, -0.1841]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, -3.0966e-07, -8.3353e-08, ..., 0.0000e+00, + 0.0000e+00, -9.4995e-08], + [-5.7276e-08, 2.7940e-08, 5.8534e-07, ..., 0.0000e+00, + -9.3132e-10, 1.8626e-08], + [ 7.9162e-09, 3.4925e-08, 3.4319e-07, ..., 1.8626e-09, + 0.0000e+00, 2.5146e-08], + ..., + [ 6.9849e-09, 7.4506e-08, -1.5153e-06, ..., 0.0000e+00, + 0.0000e+00, 1.0291e-07], + [ 1.2107e-08, -1.3411e-07, -2.8731e-07, ..., -2.3283e-09, + 4.6566e-10, -1.9744e-07], + [ 6.9849e-09, 6.9384e-08, 7.7393e-07, ..., 0.0000e+00, + 0.0000e+00, 4.9826e-08]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0328, 0.0050, 0.0229, 0.0207, 0.0412, 0.0079, -0.0026, -0.0041, + -0.0051, -0.0386], device='cuda:0'), grad: tensor([-3.3714e-07, 1.1586e-06, 6.5332e-07, 1.6857e-07, -2.9489e-05, + 3.7486e-07, -1.5069e-06, -1.4836e-06, -7.8091e-07, 3.1203e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 218.30, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4836 re_mapping 0.0047 re_causal 0.0139 /// teacc 98.91 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.1504, 0.2164, -0.0302, ..., -0.0177, 0.0673, -0.0231], + [ 0.1786, -0.1024, 0.0363, ..., -0.0824, -0.0054, -0.0751], + [-0.0750, -0.2459, -0.0250, ..., -0.0584, 0.0403, -0.2361], + ..., + [-0.1848, -0.2673, 0.0535, ..., -0.1001, -0.0731, -0.1683], + [-0.1173, 0.0048, 0.0354, ..., 0.0066, -0.1042, 0.0693], + [-0.0894, -0.0617, 0.1041, ..., 0.1999, -0.0721, -0.1844]], + device='cuda:0'), grad: tensor([[ 2.1933e-07, 1.6298e-08, 5.1223e-09, ..., 0.0000e+00, + 0.0000e+00, 6.7055e-08], + [ 9.4064e-08, 4.4703e-08, 1.9092e-08, ..., 0.0000e+00, + 0.0000e+00, 8.6147e-08], + [ 3.2596e-09, 2.2817e-08, 2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-08], + ..., + [ 3.2596e-09, 4.6566e-09, 5.4948e-08, ..., 1.8626e-09, + 0.0000e+00, 1.1642e-08], + [ 1.2871e-06, 2.8964e-07, -1.4808e-07, ..., 0.0000e+00, + 0.0000e+00, 4.7591e-07], + [ 1.1642e-08, 1.2573e-08, -2.2398e-07, ..., -4.6566e-09, + 0.0000e+00, 1.9092e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0327, 0.0054, 0.0236, 0.0208, 0.0412, 0.0077, -0.0028, -0.0043, + -0.0055, -0.0388], device='cuda:0'), grad: tensor([ 1.5125e-06, 5.0105e-07, 1.7369e-07, 2.6496e-07, 3.4273e-07, + 1.0423e-05, -1.6600e-05, 1.2899e-07, 3.6303e-06, -3.9674e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 217.98, cls_loss 0.0014 cls_loss_mapping 0.0015 cls_loss_causal 0.5120 re_mapping 0.0050 re_causal 0.0149 /// teacc 98.99 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.1506, 0.2172, -0.0305, ..., -0.0177, 0.0683, -0.0234], + [ 0.1789, -0.1026, 0.0361, ..., -0.0823, -0.0044, -0.0754], + [-0.0757, -0.2471, -0.0254, ..., -0.0587, 0.0389, -0.2396], + ..., + [-0.1848, -0.2688, 0.0538, ..., -0.1001, -0.0733, -0.1688], + [-0.1178, 0.0044, 0.0359, ..., 0.0067, -0.1044, 0.0692], + [-0.0899, -0.0619, 0.1041, ..., 0.1999, -0.0722, -0.1847]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 1.3970e-08, 2.7008e-08, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-08], + [-1.6438e-07, 9.3132e-10, -2.6589e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3504e-08], + [ 6.0536e-09, 4.6566e-10, 3.3993e-08, ..., 0.0000e+00, + 0.0000e+00, 1.0245e-08], + ..., + [ 2.1886e-08, 1.3504e-08, -1.0114e-06, ..., 0.0000e+00, + 0.0000e+00, 3.3062e-08], + [ 2.1141e-07, 7.4506e-09, 3.9628e-07, ..., 0.0000e+00, + 0.0000e+00, 1.4948e-07], + [ 6.0536e-09, -1.8626e-08, -4.8280e-06, ..., -2.2352e-08, + 0.0000e+00, 1.3039e-08]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0327, 0.0053, 0.0228, 0.0213, 0.0412, 0.0074, -0.0028, -0.0041, + -0.0054, -0.0388], device='cuda:0'), grad: tensor([ 3.5996e-07, -3.5577e-07, -1.0528e-05, -2.5660e-05, 2.8744e-05, + 2.5898e-05, 3.2736e-07, -1.1725e-06, 8.8960e-06, -2.6464e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 218.16, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.5161 re_mapping 0.0049 re_causal 0.0144 /// teacc 99.01 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.1514, 0.2168, -0.0321, ..., -0.0177, 0.0698, -0.0250], + [ 0.1793, -0.1024, 0.0362, ..., -0.0824, -0.0032, -0.0752], + [-0.0761, -0.2474, -0.0249, ..., -0.0588, 0.0387, -0.2411], + ..., + [-0.1849, -0.2711, 0.0548, ..., -0.1002, -0.0734, -0.1690], + [-0.1180, 0.0050, 0.0359, ..., 0.0067, -0.1049, 0.0699], + [-0.0920, -0.0607, 0.1028, ..., 0.2000, -0.0723, -0.1850]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, -3.9116e-08, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + [ 2.7940e-09, 7.6368e-08, 3.0827e-07, ..., 0.0000e+00, + 0.0000e+00, 2.9057e-07], + [ 0.0000e+00, -1.5832e-08, -1.9465e-07, ..., 0.0000e+00, + 0.0000e+00, 2.0117e-07], + ..., + [ 1.8626e-09, 8.3819e-09, -6.6683e-07, ..., 0.0000e+00, + 0.0000e+00, 8.2888e-08], + [ 1.2107e-08, -2.9523e-07, -3.0641e-07, ..., 0.0000e+00, + 0.0000e+00, -1.7295e-06], + [ 3.7253e-09, 2.0489e-08, 6.6217e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0316, 0.0056, 0.0230, 0.0212, 0.0414, 0.0065, -0.0025, -0.0035, + -0.0055, -0.0397], device='cuda:0'), grad: tensor([-1.6764e-08, 1.4370e-06, -6.2305e-07, 2.9877e-06, 2.6077e-08, + 4.9733e-07, 3.6042e-07, -1.0543e-06, -4.9546e-06, 1.3318e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 218.33, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.4801 re_mapping 0.0047 re_causal 0.0137 /// teacc 98.95 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.1520, 0.2178, -0.0318, ..., -0.0172, 0.0698, -0.0258], + [ 0.1798, -0.1005, 0.0352, ..., -0.0824, -0.0006, -0.0768], + [-0.0763, -0.2489, -0.0251, ..., -0.0588, 0.0389, -0.2441], + ..., + [-0.1848, -0.2722, 0.0563, ..., -0.1002, -0.0736, -0.1692], + [-0.1199, 0.0043, 0.0351, ..., 0.0067, -0.1078, 0.0711], + [-0.0934, -0.0619, 0.1019, ..., 0.1998, -0.0725, -0.1857]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, -9.7789e-08, -2.3283e-08, ..., 0.0000e+00, + 0.0000e+00, 4.2841e-08], + [-1.1176e-07, 8.3819e-09, -1.2387e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.6764e-08, 3.1665e-08, 4.7497e-08, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + ..., + [ 1.5832e-08, 1.0245e-08, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 3.4180e-07, -8.5682e-08, -3.8184e-08, ..., 0.0000e+00, + 0.0000e+00, -9.0338e-08], + [ 1.2107e-08, 1.0617e-07, 8.4750e-08, ..., 0.0000e+00, + 0.0000e+00, 2.5146e-08]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0319, 0.0051, 0.0235, 0.0205, 0.0414, 0.0064, -0.0026, -0.0024, + -0.0064, -0.0405], device='cuda:0'), grad: tensor([ 1.0002e-06, 7.4506e-08, 7.7300e-08, 2.2538e-07, 1.7770e-06, + 8.2981e-07, -6.5044e-06, 6.5193e-08, 2.0899e-06, 3.7625e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 217.93, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4886 re_mapping 0.0050 re_causal 0.0144 /// teacc 99.00 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.1522, 0.2182, -0.0317, ..., -0.0172, 0.0697, -0.0263], + [ 0.1800, -0.1006, 0.0349, ..., -0.0824, -0.0005, -0.0767], + [-0.0764, -0.2491, -0.0252, ..., -0.0589, 0.0389, -0.2448], + ..., + [-0.1850, -0.2727, 0.0566, ..., -0.1002, -0.0734, -0.1696], + [-0.1220, 0.0026, 0.0351, ..., 0.0067, -0.1090, 0.0699], + [-0.0935, -0.0623, 0.1019, ..., 0.1998, -0.0725, -0.1862]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 3.0734e-08, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.9558e-08], + [ 1.4901e-08, 6.9849e-08, 1.1735e-07, ..., 0.0000e+00, + 0.0000e+00, 1.2759e-07], + [ 3.0734e-08, 9.3132e-08, 2.7195e-07, ..., 0.0000e+00, + 0.0000e+00, 2.0862e-07], + ..., + [ 2.3283e-08, -4.6566e-07, -5.4389e-07, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-08], + [ 1.2666e-07, -4.0513e-07, -6.2771e-07, ..., 0.0000e+00, + 0.0000e+00, -1.6205e-07], + [ 1.3970e-08, 2.5239e-07, 3.3714e-07, ..., 0.0000e+00, + 0.0000e+00, 1.0151e-07]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0318, 0.0049, 0.0236, 0.0206, 0.0422, 0.0070, -0.0041, -0.0023, + -0.0071, -0.0405], device='cuda:0'), grad: tensor([ 3.3993e-07, 4.4703e-07, 4.2282e-07, -5.9325e-07, 7.6368e-08, + 2.0824e-06, 1.6764e-08, -2.9374e-06, -1.1763e-06, 1.3113e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 217.82, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4783 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.01 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.1534, 0.2193, -0.0315, ..., -0.0172, 0.0698, -0.0270], + [ 0.1800, -0.1007, 0.0350, ..., -0.0824, -0.0005, -0.0770], + [-0.0765, -0.2495, -0.0258, ..., -0.0589, 0.0389, -0.2455], + ..., + [-0.1850, -0.2730, 0.0566, ..., -0.1002, -0.0735, -0.1696], + [-0.1222, 0.0026, 0.0352, ..., 0.0067, -0.1090, 0.0701], + [-0.0938, -0.0625, 0.1020, ..., 0.1999, -0.0725, -0.1873]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -4.8988e-07, -6.0536e-08, ..., 0.0000e+00, + -2.7940e-09, 1.8626e-09], + [ 2.8871e-08, 4.2841e-08, 4.3772e-08, ..., 0.0000e+00, + 0.0000e+00, 4.0047e-08], + [ 2.7940e-09, 7.4506e-09, 2.7940e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 1.0245e-08, 1.3970e-08, -1.1269e-07, ..., 0.0000e+00, + 9.3132e-10, 2.7940e-09], + [ 1.5553e-07, 8.4750e-08, 1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, 1.6391e-07], + [ 9.3132e-09, 3.7253e-08, 3.7253e-08, ..., -9.3132e-10, + 9.3132e-10, 9.3132e-09]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0324, 0.0045, 0.0244, 0.0208, 0.0420, 0.0071, -0.0047, -0.0021, + -0.0072, -0.0405], device='cuda:0'), grad: tensor([-6.6590e-07, 4.1816e-07, 7.1712e-08, 5.9977e-07, -4.5717e-05, + -1.1604e-06, 8.3633e-07, 9.6764e-07, 3.0734e-07, 4.4286e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 218.07, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5014 re_mapping 0.0049 re_causal 0.0142 /// teacc 98.92 lr 0.00010000 +Epoch 302, weight, value: tensor([[-1.5312e-01, 2.2116e-01, -3.1302e-02, ..., -1.7201e-02, + 7.0244e-02, -2.6542e-02], + [ 1.8019e-01, -1.0155e-01, 3.4457e-02, ..., -8.2482e-02, + 2.5871e-04, -7.6983e-02], + [-7.6876e-02, -2.4987e-01, -2.6234e-02, ..., -5.8984e-02, + 3.8308e-02, -2.4620e-01], + ..., + [-1.8496e-01, -2.7386e-01, 5.6951e-02, ..., -1.0024e-01, + -7.3370e-02, -1.6973e-01], + [-1.2226e-01, 2.6398e-03, 3.5397e-02, ..., 6.7204e-03, + -1.0908e-01, 7.0171e-02], + [-9.4310e-02, -6.2986e-02, 1.0199e-01, ..., 1.9992e-01, + -7.2810e-02, -1.8821e-01]], device='cuda:0'), grad: tensor([[ 2.7940e-09, -8.3819e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-2.3283e-08, 0.0000e+00, -1.9558e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 9.3132e-10, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.7695e-08], + ..., + [ 1.4901e-08, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 4.6566e-09, -8.3819e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [ 2.7940e-09, 9.3132e-10, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.1420e-08]], device='cuda:0') +Epoch 302, bias, value: tensor([ 0.0333, 0.0040, 0.0241, 0.0214, 0.0423, 0.0066, -0.0054, -0.0017, + -0.0071, -0.0406], device='cuda:0'), grad: tensor([ 6.7987e-08, 1.8440e-07, -3.8277e-07, -2.9989e-07, 7.6368e-08, + 1.8161e-07, -2.3283e-08, 1.3690e-07, 8.3819e-09, 7.1712e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 218.02, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.5223 re_mapping 0.0050 re_causal 0.0146 /// teacc 98.97 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.1533, 0.2194, -0.0344, ..., -0.0173, 0.0698, -0.0267], + [ 0.1810, -0.1007, 0.0349, ..., -0.0825, 0.0020, -0.0764], + [-0.0774, -0.2495, -0.0267, ..., -0.0590, 0.0375, -0.2463], + ..., + [-0.1854, -0.2751, 0.0569, ..., -0.1003, -0.0735, -0.1699], + [-0.1228, 0.0024, 0.0350, ..., 0.0067, -0.1096, 0.0705], + [-0.0947, -0.0604, 0.1033, ..., 0.2002, -0.0722, -0.1885]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.7695e-08, -2.7940e-09, ..., 0.0000e+00, + -9.3132e-10, 7.4506e-09], + [-2.6636e-07, 5.5879e-09, -3.0361e-07, ..., 0.0000e+00, + 0.0000e+00, 1.7695e-08], + [-3.7253e-09, 9.3132e-10, 1.1642e-07, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-08], + ..., + [ 1.0245e-08, 3.7253e-09, -3.0920e-07, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 2.5239e-07, 4.3865e-07, 6.2399e-08, ..., 0.0000e+00, + 0.0000e+00, 4.5449e-07], + [ 2.3283e-07, 8.3819e-09, 3.8277e-07, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0302, 0.0045, 0.0239, 0.0212, 0.0419, 0.0063, -0.0052, -0.0018, + -0.0077, -0.0390], device='cuda:0'), grad: tensor([ 1.6764e-08, -1.7509e-07, -1.5087e-07, -2.3749e-07, -2.4214e-08, + -5.1875e-07, 1.6764e-07, -4.1164e-07, 7.4599e-07, 5.9046e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 218.04, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4972 re_mapping 0.0049 re_causal 0.0143 /// teacc 99.01 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.1548, 0.2195, -0.0344, ..., -0.0173, 0.0697, -0.0298], + [ 0.1813, -0.1008, 0.0349, ..., -0.0826, 0.0023, -0.0773], + [-0.0777, -0.2518, -0.0271, ..., -0.0591, 0.0368, -0.2486], + ..., + [-0.1856, -0.2769, 0.0570, ..., -0.1005, -0.0727, -0.1703], + [-0.1237, 0.0012, 0.0351, ..., 0.0066, -0.1096, 0.0696], + [-0.0952, -0.0606, 0.1033, ..., 0.2004, -0.0725, -0.1891]], + device='cuda:0'), grad: tensor([[ 1.4901e-07, 9.5926e-07, 2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 9.7789e-08], + [-1.4808e-07, 1.5832e-08, -7.7300e-08, ..., 0.0000e+00, + -1.8626e-09, 7.1712e-08], + [ 1.4901e-08, 1.3970e-08, -2.4214e-08, ..., 0.0000e+00, + -2.9802e-08, 1.4901e-08], + ..., + [ 4.0978e-08, 4.9360e-08, -6.3330e-08, ..., 0.0000e+00, + 7.4506e-09, 5.3085e-08], + [ 9.1642e-07, 1.1027e-06, 9.2201e-08, ..., 0.0000e+00, + 1.5832e-08, 9.4622e-07], + [ 1.8999e-07, 2.3935e-07, 6.3330e-08, ..., 0.0000e+00, + 3.7253e-09, 2.4214e-07]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0301, 0.0045, 0.0238, 0.0211, 0.0416, 0.0077, -0.0055, -0.0016, + -0.0085, -0.0391], device='cuda:0'), grad: tensor([ 4.0680e-06, -5.1223e-08, -1.9278e-07, 8.1807e-06, -7.9162e-07, + -9.0525e-06, -6.6683e-06, 6.7987e-08, 3.0100e-06, 1.4398e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 218.14, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.5033 re_mapping 0.0049 re_causal 0.0138 /// teacc 99.02 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.1554, 0.2198, -0.0344, ..., -0.0176, 0.0702, -0.0307], + [ 0.1808, -0.1005, 0.0338, ..., -0.0827, 0.0028, -0.0775], + [-0.0778, -0.2531, -0.0272, ..., -0.0592, 0.0363, -0.2507], + ..., + [-0.1850, -0.2784, 0.0578, ..., -0.1006, -0.0732, -0.1706], + [-0.1243, 0.0016, 0.0356, ..., 0.0065, -0.1099, 0.0700], + [-0.0965, -0.0610, 0.1039, ..., 0.2010, -0.0730, -0.1903]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, -5.2527e-07, -2.7101e-07, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-2.8405e-07, 3.0734e-08, -1.9092e-07, ..., 0.0000e+00, + 9.3132e-09, 2.4214e-08], + [-5.3085e-08, 1.4901e-08, 7.4506e-08, ..., 0.0000e+00, + -2.0489e-08, 1.4901e-08], + ..., + [ 9.5926e-08, 3.1665e-08, 1.2666e-07, ..., 0.0000e+00, + 4.6566e-09, 4.2841e-08], + [ 5.5879e-09, -1.6298e-07, -5.6438e-07, ..., 0.0000e+00, + 0.0000e+00, -3.9395e-07], + [ 6.4261e-08, 3.3155e-07, 3.1013e-07, ..., 0.0000e+00, + 1.8626e-09, 8.4750e-08]], device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0300, 0.0034, 0.0237, 0.0209, 0.0391, 0.0080, -0.0051, -0.0009, + -0.0084, -0.0379], device='cuda:0'), grad: tensor([-9.6392e-07, 2.8238e-06, 2.0210e-07, 9.7975e-07, -4.5672e-06, + 7.8045e-07, 3.6135e-07, 8.1956e-07, -2.8946e-06, 2.4512e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 217.86, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4755 re_mapping 0.0047 re_causal 0.0137 /// teacc 99.01 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.1556, 0.2207, -0.0343, ..., -0.0176, 0.0703, -0.0306], + [ 0.1815, -0.1003, 0.0338, ..., -0.0827, 0.0043, -0.0776], + [-0.0795, -0.2534, -0.0289, ..., -0.0592, 0.0359, -0.2515], + ..., + [-0.1850, -0.2805, 0.0580, ..., -0.1006, -0.0738, -0.1710], + [-0.1249, 0.0017, 0.0355, ..., 0.0065, -0.1102, 0.0704], + [-0.0970, -0.0615, 0.1039, ..., 0.2011, -0.0732, -0.1907]], + device='cuda:0'), grad: tensor([[ 4.5635e-08, 3.0734e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7008e-08], + [-1.7220e-06, 3.1665e-08, -1.5367e-06, ..., 0.0000e+00, + 0.0000e+00, 3.1665e-08], + [ 5.5879e-09, 1.8626e-09, 2.1420e-08, ..., 0.0000e+00, + 0.0000e+00, 3.1665e-08], + ..., + [ 8.5402e-07, 6.5193e-09, 6.8173e-07, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-08], + [ 1.5832e-07, 1.2387e-07, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 9.1270e-08], + [ 9.6206e-07, 7.3574e-08, 7.8976e-07, ..., 0.0000e+00, + 0.0000e+00, 7.0781e-08]], device='cuda:0') +Epoch 306, bias, value: tensor([ 0.0301, 0.0038, 0.0222, 0.0215, 0.0389, 0.0074, -0.0051, -0.0008, + -0.0088, -0.0380], device='cuda:0'), grad: tensor([ 1.0803e-07, -3.7588e-06, 1.4063e-07, -5.5786e-07, -4.2189e-07, + -1.5780e-05, 1.5393e-05, 1.8422e-06, 3.4366e-07, 2.6859e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 217.91, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4462 re_mapping 0.0048 re_causal 0.0136 /// teacc 99.00 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.1557, 0.2212, -0.0343, ..., -0.0176, 0.0704, -0.0306], + [ 0.1816, -0.1005, 0.0335, ..., -0.0829, 0.0043, -0.0782], + [-0.0797, -0.2557, -0.0292, ..., -0.0593, 0.0359, -0.2540], + ..., + [-0.1850, -0.2817, 0.0579, ..., -0.1006, -0.0739, -0.1728], + [-0.1241, 0.0042, 0.0397, ..., 0.0091, -0.1103, 0.0737], + [-0.0994, -0.0641, 0.1035, ..., 0.1998, -0.0732, -0.1944]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + [ 0.0000e+00, 1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-08], + [ 0.0000e+00, 9.3132e-10, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 2.5425e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 8.2888e-08, ..., 0.0000e+00, + 0.0000e+00, 5.2154e-07], + [ 0.0000e+00, 1.8626e-09, 6.6124e-08, ..., 0.0000e+00, + 0.0000e+00, 6.2399e-08], + [ 1.8626e-09, 1.0245e-08, -2.7940e-07, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-08]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0301, 0.0034, 0.0222, 0.0216, 0.0392, 0.0075, -0.0048, -0.0009, + -0.0043, -0.0387], device='cuda:0'), grad: tensor([ 6.4261e-08, 2.2724e-07, 2.7940e-09, -3.9861e-06, -2.7940e-08, + 1.2293e-07, 2.4401e-07, 3.0287e-06, 4.4610e-07, -1.4342e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 217.91, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.5214 re_mapping 0.0047 re_causal 0.0141 /// teacc 98.97 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.1560, 0.2186, -0.0343, ..., -0.0176, 0.0704, -0.0310], + [ 0.1816, -0.1007, 0.0334, ..., -0.0830, 0.0044, -0.0791], + [-0.0797, -0.2562, -0.0295, ..., -0.0595, 0.0357, -0.2550], + ..., + [-0.1851, -0.2821, 0.0578, ..., -0.1007, -0.0741, -0.1734], + [-0.1242, 0.0043, 0.0400, ..., 0.0090, -0.1103, 0.0738], + [-0.0985, -0.0637, 0.1037, ..., 0.1980, -0.0730, -0.1930]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.5390e-08, -3.5390e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.7497e-08, 3.4459e-08, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 3.9116e-08], + [ 5.5879e-09, -2.1420e-08, 5.6811e-08, ..., 0.0000e+00, + 0.0000e+00, 4.5635e-08], + ..., + [ 1.8626e-09, 8.3819e-09, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 3.9116e-08, 4.0978e-08, -1.0338e-07, ..., 0.0000e+00, + 0.0000e+00, -4.7497e-08], + [ 1.0990e-06, 6.0443e-07, 1.1176e-08, ..., -9.3132e-10, + 0.0000e+00, 8.0839e-07]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0274, 0.0032, 0.0221, 0.0211, 0.0403, 0.0069, -0.0016, -0.0010, + -0.0042, -0.0389], device='cuda:0'), grad: tensor([-5.6811e-08, 2.1327e-07, -2.6356e-07, 2.1160e-06, 2.6077e-08, + -3.7029e-06, 8.9407e-08, 9.2201e-08, -7.0781e-08, 1.5618e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 217.84, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4780 re_mapping 0.0047 re_causal 0.0138 /// teacc 98.98 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.1559, 0.2187, -0.0344, ..., -0.0192, 0.0705, -0.0310], + [ 0.1819, -0.1008, 0.0335, ..., -0.0832, 0.0044, -0.0798], + [-0.0801, -0.2566, -0.0300, ..., -0.0599, 0.0356, -0.2534], + ..., + [-0.1853, -0.2829, 0.0578, ..., -0.1008, -0.0741, -0.1743], + [-0.1247, 0.0040, 0.0400, ..., 0.0088, -0.1103, 0.0738], + [-0.0996, -0.0640, 0.1037, ..., 0.1986, -0.0729, -0.1931]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.4531e-06, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, -5.6811e-08], + [-5.5879e-09, 1.8626e-09, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 2.0489e-08, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 2.7940e-09, 8.3819e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, 7.9162e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 3.7253e-09, 7.0781e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0273, 0.0034, 0.0222, 0.0216, 0.0407, 0.0065, -0.0015, -0.0010, + -0.0045, -0.0391], device='cuda:0'), grad: tensor([-4.8801e-06, 8.1956e-08, 1.1828e-07, 2.0489e-08, -2.1625e-06, + 4.4890e-07, 3.4794e-06, 9.5926e-08, 2.6729e-07, 2.5295e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 218.30, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.5039 re_mapping 0.0048 re_causal 0.0144 /// teacc 99.00 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.1557, 0.2192, -0.0342, ..., -0.0193, 0.0705, -0.0305], + [ 0.1818, -0.1012, 0.0334, ..., -0.0839, 0.0044, -0.0808], + [-0.0802, -0.2571, -0.0300, ..., -0.0601, 0.0356, -0.2539], + ..., + [-0.1851, -0.2839, 0.0581, ..., -0.1008, -0.0741, -0.1747], + [-0.1252, 0.0039, 0.0399, ..., 0.0088, -0.1103, 0.0736], + [-0.1002, -0.0647, 0.1034, ..., 0.1988, -0.0729, -0.1933]], + device='cuda:0'), grad: tensor([[ 2.8871e-08, -5.9512e-07, -3.3155e-07, ..., -1.8626e-09, + 0.0000e+00, 8.3819e-09], + [ 6.3360e-05, 4.2841e-08, 1.2982e-04, ..., 0.0000e+00, + 0.0000e+00, 1.5832e-08], + [ 1.0077e-06, 8.3819e-08, 2.0955e-06, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + ..., + [-9.6738e-05, 4.9360e-08, -1.9801e-04, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + [ 2.7940e-07, 2.5891e-07, 1.3970e-07, ..., 0.0000e+00, + 0.0000e+00, 2.1514e-07], + [ 3.2336e-05, 6.5006e-07, 6.6042e-05, ..., 9.3132e-10, + 0.0000e+00, 2.5053e-07]], device='cuda:0') +Epoch 310, bias, value: tensor([ 0.0275, 0.0031, 0.0227, 0.0215, 0.0407, 0.0064, -0.0014, -0.0007, + -0.0048, -0.0395], device='cuda:0'), grad: tensor([-1.1455e-06, 1.8728e-04, 3.1628e-06, 3.9861e-07, 1.5832e-07, + -1.1055e-06, -4.2841e-08, -2.8563e-04, 6.3423e-07, 9.6142e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 217.70, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.5065 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.00 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.1562, 0.2197, -0.0338, ..., -0.0187, 0.0705, -0.0313], + [ 0.1819, -0.1020, 0.0331, ..., -0.0840, 0.0049, -0.0805], + [-0.0803, -0.2578, -0.0305, ..., -0.0603, 0.0356, -0.2548], + ..., + [-0.1849, -0.2849, 0.0584, ..., -0.1009, -0.0742, -0.1761], + [-0.1253, 0.0040, 0.0399, ..., 0.0088, -0.1103, 0.0739], + [-0.1014, -0.0656, 0.1032, ..., 0.1987, -0.0729, -0.1934]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.7695e-07, -1.5460e-07, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 2.7940e-09, 2.0489e-08, 5.4017e-08, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08], + [ 9.3132e-10, 2.7940e-09, 1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + ..., + [ 1.8626e-09, 7.0781e-08, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-08], + [ 2.1420e-08, -6.3423e-07, -1.1744e-06, ..., 0.0000e+00, + 0.0000e+00, -6.6124e-07], + [ 6.5193e-09, 6.9663e-07, 1.1493e-06, ..., 0.0000e+00, + 0.0000e+00, 5.6811e-07]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0276, 0.0029, 0.0231, 0.0209, 0.0407, 0.0064, -0.0013, -0.0006, + -0.0047, -0.0397], device='cuda:0'), grad: tensor([-3.0827e-07, 1.2293e-07, -9.6858e-08, 8.4750e-07, 1.4808e-07, + -8.1863e-07, 5.4948e-08, 1.3970e-08, -2.1420e-06, 2.1849e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 218.14, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4915 re_mapping 0.0048 re_causal 0.0140 /// teacc 98.98 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.1564, 0.2200, -0.0337, ..., -0.0187, 0.0705, -0.0311], + [ 0.1824, -0.1022, 0.0332, ..., -0.0849, 0.0054, -0.0803], + [-0.0809, -0.2585, -0.0322, ..., -0.0608, 0.0356, -0.2576], + ..., + [-0.1851, -0.2861, 0.0589, ..., -0.1013, -0.0742, -0.1771], + [-0.1258, 0.0040, 0.0401, ..., 0.0087, -0.1103, 0.0743], + [-0.1019, -0.0661, 0.1028, ..., 0.2002, -0.0730, -0.1937]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.5682e-08, -6.0536e-08, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [-1.8626e-09, 9.3132e-10, 6.1467e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 5.5879e-09, 4.3772e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, 1.7798e-06, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1176e-08, -2.8871e-08, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -5.0291e-08], + [ 0.0000e+00, 7.4506e-08, -3.4850e-06, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0277, 0.0029, 0.0224, 0.0213, 0.0401, 0.0066, -0.0014, -0.0001, + -0.0046, -0.0400], device='cuda:0'), grad: tensor([-1.6298e-07, 1.0990e-07, 8.2888e-08, 1.3597e-07, 2.6301e-06, + 1.2107e-07, 9.3132e-09, 3.1888e-06, 3.7253e-09, -6.1020e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 218.14, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4892 re_mapping 0.0047 re_causal 0.0141 /// teacc 98.98 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.1594, 0.2191, -0.0338, ..., -0.0189, 0.0663, -0.0316], + [ 0.1828, -0.1008, 0.0334, ..., -0.0851, 0.0075, -0.0808], + [-0.0813, -0.2592, -0.0344, ..., -0.0611, 0.0355, -0.2597], + ..., + [-0.1849, -0.2872, 0.0594, ..., -0.1014, -0.0744, -0.1775], + [-0.1265, 0.0039, 0.0399, ..., 0.0087, -0.1115, 0.0746], + [-0.1035, -0.0664, 0.1026, ..., 0.2003, -0.0734, -0.1939]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 1.8626e-09, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-2.9802e-08, 2.7940e-09, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, -5.6811e-08, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.5879e-09, 0.0000e+00, -7.4506e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.1712e-08, -5.5879e-08, -5.7742e-08, ..., 0.0000e+00, + 0.0000e+00, -4.3772e-08], + [ 8.3819e-09, 1.0990e-07, 1.3877e-07, ..., 0.0000e+00, + 0.0000e+00, 4.1910e-08]], device='cuda:0') +Epoch 313, bias, value: tensor([ 0.0270, 0.0031, 0.0212, 0.0213, 0.0401, 0.0068, -0.0007, 0.0005, + -0.0048, -0.0404], device='cuda:0'), grad: tensor([ 3.8277e-07, 7.4506e-09, -4.6659e-07, 4.8429e-08, -1.1921e-07, + 9.3132e-10, -1.4156e-06, -1.0990e-07, 1.2647e-06, 4.0978e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 218.24, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4771 re_mapping 0.0047 re_causal 0.0135 /// teacc 98.97 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.1598, 0.2193, -0.0338, ..., -0.0189, 0.0662, -0.0316], + [ 0.1831, -0.1006, 0.0333, ..., -0.0852, 0.0080, -0.0812], + [-0.0813, -0.2595, -0.0351, ..., -0.0612, 0.0355, -0.2602], + ..., + [-0.1850, -0.2882, 0.0598, ..., -0.1014, -0.0743, -0.1784], + [-0.1268, 0.0059, 0.0406, ..., 0.0087, -0.1118, 0.0763], + [-0.1044, -0.0682, 0.1021, ..., 0.2004, -0.0734, -0.1954]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09], + [-8.1025e-07, -1.0710e-07, -5.2620e-07, ..., 0.0000e+00, + -3.8184e-08, 1.7695e-08], + [ 7.4506e-09, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08], + ..., + [ 5.6811e-08, 7.4506e-09, -3.0361e-06, ..., 0.0000e+00, + 2.7940e-09, 6.9849e-08], + [ 2.9523e-07, 4.0978e-08, 2.0582e-07, ..., 0.0000e+00, + 1.3970e-08, 3.5390e-08], + [ 1.3039e-08, 1.8626e-09, 3.0287e-06, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-09]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0270, 0.0030, 0.0213, 0.0212, 0.0399, 0.0066, -0.0008, 0.0010, + -0.0037, -0.0411], device='cuda:0'), grad: tensor([ 3.7253e-08, -1.2759e-06, 3.7253e-09, -3.7253e-07, 3.9581e-07, + 1.1455e-07, 2.6822e-07, -5.9083e-06, 5.9325e-07, 6.1505e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 218.23, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.5059 re_mapping 0.0047 re_causal 0.0140 /// teacc 98.97 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.1599, 0.2203, -0.0336, ..., -0.0189, 0.0662, -0.0316], + [ 0.1832, -0.1009, 0.0334, ..., -0.0861, 0.0082, -0.0814], + [-0.0814, -0.2601, -0.0350, ..., -0.0613, 0.0355, -0.2607], + ..., + [-0.1851, -0.2884, 0.0598, ..., -0.1015, -0.0743, -0.1786], + [-0.1270, 0.0059, 0.0405, ..., 0.0087, -0.1118, 0.0763], + [-0.1045, -0.0687, 0.1021, ..., 0.2006, -0.0736, -0.1956]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0489e-08, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 6.4448e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 3.1665e-07, ..., 0.0000e+00, + 0.0000e+00, 9.4995e-08], + ..., + [ 9.3132e-10, 1.8626e-09, -1.5106e-06, ..., 0.0000e+00, + 0.0000e+00, -1.3784e-07], + [ 1.6764e-08, -1.2014e-07, 2.5705e-07, ..., 0.0000e+00, + 0.0000e+00, -2.0582e-07], + [ 1.8626e-09, 4.6566e-09, 1.5274e-07, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 315, bias, value: tensor([ 0.0274, 0.0030, 0.0217, 0.0211, 0.0400, 0.0067, -0.0011, 0.0009, + -0.0038, -0.0412], device='cuda:0'), grad: tensor([-1.1176e-08, 1.0831e-06, 8.8383e-07, 2.8126e-07, -7.9256e-07, + 2.4121e-07, 9.4995e-08, -2.9244e-06, 2.2911e-07, 9.1642e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 217.86, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4924 re_mapping 0.0049 re_causal 0.0143 /// teacc 99.03 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.1607, 0.2207, -0.0336, ..., -0.0189, 0.0663, -0.0308], + [ 0.1833, -0.1008, 0.0334, ..., -0.0876, 0.0083, -0.0817], + [-0.0814, -0.2601, -0.0353, ..., -0.0615, 0.0352, -0.2613], + ..., + [-0.1851, -0.2890, 0.0598, ..., -0.1020, -0.0744, -0.1788], + [-0.1274, 0.0058, 0.0405, ..., 0.0087, -0.1120, 0.0764], + [-0.1051, -0.0689, 0.1022, ..., 0.2011, -0.0734, -0.1958]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -9.0804e-07, -3.1851e-07, ..., 0.0000e+00, + -9.3132e-10, 2.7940e-09], + [-9.3132e-10, 3.7253e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 1.8626e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + ..., + [ 9.3132e-10, 3.7253e-09, -1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 1.8626e-08, 3.3528e-08, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + [ 6.5193e-09, 4.2841e-08, 9.3132e-09, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0273, 0.0029, 0.0216, 0.0214, 0.0400, 0.0065, -0.0010, 0.0009, + -0.0041, -0.0411], device='cuda:0'), grad: tensor([-3.8743e-07, 1.0338e-07, 1.2107e-08, 2.9057e-07, 7.6834e-07, + -3.4645e-07, -5.4389e-07, -1.7695e-08, 5.9605e-08, 6.4261e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 218.39, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4721 re_mapping 0.0049 re_causal 0.0133 /// teacc 99.01 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.1610, 0.2225, -0.0332, ..., -0.0186, 0.0670, -0.0300], + [ 0.1834, -0.1027, 0.0332, ..., -0.0876, 0.0077, -0.0819], + [-0.0815, -0.2593, -0.0357, ..., -0.0615, 0.0349, -0.2616], + ..., + [-0.1852, -0.2905, 0.0600, ..., -0.1021, -0.0744, -0.1794], + [-0.1280, 0.0052, 0.0400, ..., 0.0084, -0.1122, 0.0761], + [-0.1084, -0.0717, 0.1025, ..., 0.2014, -0.0737, -0.1979]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -5.5507e-07, -1.0245e-07, ..., 0.0000e+00, + 4.6566e-09, -1.0431e-07], + [-7.0035e-07, -1.7416e-07, -5.8953e-07, ..., 0.0000e+00, + -2.6636e-07, 2.7940e-09], + [ 9.3132e-10, -9.6858e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 2.7940e-09, 1.7695e-08, 2.8871e-08, ..., 0.0000e+00, + 9.3132e-10, 3.7253e-09], + [ 4.1258e-07, 1.1548e-07, 3.5018e-07, ..., 0.0000e+00, + 1.5553e-07, 4.9360e-08], + [ 6.5193e-09, 2.0489e-08, -3.0734e-08, ..., 0.0000e+00, + 1.8626e-09, 7.4506e-09]], device='cuda:0') +Epoch 317, bias, value: tensor([ 0.0280, 0.0025, 0.0222, 0.0200, 0.0393, 0.0092, -0.0014, 0.0012, + -0.0051, -0.0413], device='cuda:0'), grad: tensor([-4.2561e-07, -9.2201e-07, -1.3206e-06, 1.9465e-07, -3.1665e-07, + 2.3283e-07, 1.3066e-06, 2.6263e-07, 6.9477e-07, 2.7753e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 217.98, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4822 re_mapping 0.0049 re_causal 0.0138 /// teacc 98.99 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.1609, 0.2232, -0.0335, ..., -0.0184, 0.0662, -0.0293], + [ 0.1835, -0.1028, 0.0328, ..., -0.0877, 0.0078, -0.0833], + [-0.0819, -0.2602, -0.0357, ..., -0.0617, 0.0357, -0.2621], + ..., + [-0.1851, -0.2922, 0.0601, ..., -0.1021, -0.0754, -0.1792], + [-0.1283, 0.0057, 0.0399, ..., 0.0084, -0.1120, 0.0774], + [-0.1085, -0.0719, 0.1032, ..., 0.2013, -0.0733, -0.1981]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, -5.3085e-08, 7.9162e-09, ..., -2.7940e-09, + -3.7253e-09, -4.1910e-09], + [-6.1914e-06, 2.7940e-09, 6.1467e-08, ..., 4.6566e-10, + 4.6566e-10, -3.0082e-07], + [ 3.0920e-06, 3.1665e-08, -5.1502e-07, ..., 0.0000e+00, + 4.6566e-09, 1.6252e-07], + ..., + [ 2.1420e-06, 6.0536e-09, -8.6147e-08, ..., 0.0000e+00, + 4.6566e-10, 9.4529e-08], + [ 1.5367e-08, -3.7253e-08, 4.4098e-07, ..., 0.0000e+00, + -4.6566e-09, -2.2817e-08], + [ 6.5193e-09, 6.0536e-09, 4.8894e-08, ..., 1.3970e-09, + 1.8626e-09, 9.7789e-09]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0280, 0.0020, 0.0224, 0.0192, 0.0388, 0.0095, -0.0017, 0.0012, + -0.0048, -0.0406], device='cuda:0'), grad: tensor([ 1.8114e-07, -9.3952e-06, -1.0524e-06, 1.4883e-06, -2.6882e-05, + 2.2678e-07, 2.3888e-07, 3.0816e-05, 4.0978e-06, 2.5705e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 218.02, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4550 re_mapping 0.0049 re_causal 0.0140 /// teacc 98.88 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.1610, 0.2235, -0.0336, ..., -0.0182, 0.0660, -0.0309], + [ 0.1862, -0.1027, 0.0327, ..., -0.0877, 0.0082, -0.0840], + [-0.0854, -0.2610, -0.0387, ..., -0.0620, 0.0357, -0.2650], + ..., + [-0.1850, -0.2940, 0.0604, ..., -0.1021, -0.0764, -0.1790], + [-0.1287, 0.0059, 0.0398, ..., 0.0085, -0.1126, 0.0774], + [-0.1086, -0.0720, 0.1033, ..., 0.2013, -0.0750, -0.1982]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.0094e-08, -2.7940e-08, ..., 0.0000e+00, + -2.9802e-08, -8.3819e-09], + [-9.3132e-10, 4.4703e-08, 2.7940e-09, ..., 0.0000e+00, + 3.7253e-09, 2.7940e-09], + [ 1.8626e-09, 8.9407e-08, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08], + ..., + [ 1.8626e-09, 1.3039e-08, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.5832e-08, -1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, -3.6322e-08], + [ 0.0000e+00, 3.4459e-08, 5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0280, 0.0034, 0.0194, 0.0193, 0.0390, 0.0096, -0.0017, 0.0015, + -0.0050, -0.0407], device='cuda:0'), grad: tensor([ 1.0328e-06, 3.6135e-07, 5.0757e-07, 3.7253e-08, -2.1849e-06, + 3.4459e-08, -4.6566e-08, 1.0710e-07, -5.0291e-08, 2.2072e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 218.03, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5277 re_mapping 0.0048 re_causal 0.0139 /// teacc 98.95 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.1628, 0.2234, -0.0335, ..., -0.0190, 0.0660, -0.0349], + [ 0.1864, -0.1027, 0.0326, ..., -0.0878, 0.0092, -0.0840], + [-0.0854, -0.2614, -0.0389, ..., -0.0624, 0.0355, -0.2662], + ..., + [-0.1849, -0.2949, 0.0610, ..., -0.1022, -0.0763, -0.1789], + [-0.1296, 0.0058, 0.0398, ..., 0.0084, -0.1118, 0.0769], + [-0.1089, -0.0721, 0.1025, ..., 0.2018, -0.0754, -0.1985]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.9595e-05, 2.4866e-07, ..., 0.0000e+00, + 9.3132e-10, -9.3132e-10], + [-1.6764e-08, 9.3132e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, 1.8626e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 2.7940e-09, 5.1223e-08, 4.3679e-07, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 3.7253e-09, 4.5635e-08, -1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, -3.0734e-08], + [ 3.7253e-09, 2.2352e-08, -1.1660e-06, ..., 0.0000e+00, + -2.9802e-08, 1.5832e-08]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0278, 0.0033, 0.0194, 0.0189, 0.0414, 0.0094, -0.0016, 0.0025, + -0.0053, -0.0429], device='cuda:0'), grad: tensor([-3.0786e-05, 2.7940e-09, 4.0978e-08, 1.2107e-07, 1.1977e-06, + 1.8720e-07, 3.1054e-05, 1.1092e-06, 7.8231e-08, -2.9653e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 217.97, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.4942 re_mapping 0.0046 re_causal 0.0134 /// teacc 99.03 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.1631, 0.2240, -0.0335, ..., -0.0189, 0.0652, -0.0350], + [ 0.1876, -0.1029, 0.0332, ..., -0.0878, 0.0122, -0.0813], + [-0.0855, -0.2638, -0.0392, ..., -0.0643, 0.0347, -0.2686], + ..., + [-0.1846, -0.2971, 0.0611, ..., -0.1023, -0.0785, -0.1793], + [-0.1298, 0.0063, 0.0399, ..., 0.0086, -0.1115, 0.0778], + [-0.1090, -0.0721, 0.1025, ..., 0.2018, -0.0734, -0.1975]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 4.6566e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 2.5146e-08, 6.8918e-08, 1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 1.4063e-07], + [ 1.4808e-07, 8.5123e-07, 2.3469e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1688e-06], + ..., + [-1.4734e-06, 9.3132e-09, -7.6368e-08, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-08], + [-2.4587e-07, -1.1390e-06, -2.9057e-07, ..., 0.0000e+00, + 0.0000e+00, -1.7565e-06], + [ 4.0047e-08, 1.3039e-08, 6.7055e-08, ..., 0.0000e+00, + 0.0000e+00, 4.3772e-08]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0279, 0.0042, 0.0193, 0.0175, 0.0413, 0.0091, -0.0029, 0.0033, + -0.0051, -0.0432], device='cuda:0'), grad: tensor([ 2.1327e-07, 9.5461e-07, 3.2336e-06, 2.2259e-07, -1.4499e-05, + 2.5667e-06, 1.5087e-07, -2.1622e-05, -4.3251e-06, 3.3140e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 217.94, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4934 re_mapping 0.0047 re_causal 0.0140 /// teacc 99.03 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.1649, 0.2241, -0.0334, ..., -0.0190, 0.0655, -0.0359], + [ 0.1895, -0.1031, 0.0348, ..., -0.0884, 0.0125, -0.0811], + [-0.0856, -0.2646, -0.0392, ..., -0.0650, 0.0353, -0.2693], + ..., + [-0.1873, -0.2998, 0.0599, ..., -0.1025, -0.0815, -0.1824], + [-0.1305, 0.0061, 0.0398, ..., 0.0088, -0.1118, 0.0775], + [-0.1094, -0.0723, 0.1024, ..., 0.2020, -0.0738, -0.1978]], + device='cuda:0'), grad: tensor([[ 1.2396e-06, 1.7909e-06, 2.9057e-07, ..., 0.0000e+00, + 3.3714e-07, 1.7220e-06], + [ 2.6077e-08, 5.9605e-08, 1.0151e-07, ..., 0.0000e+00, + 1.2107e-08, 1.1455e-07], + [ 2.7940e-09, 5.5879e-09, 9.3132e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + ..., + [ 2.8871e-08, 3.7253e-08, -4.9826e-07, ..., 0.0000e+00, + 7.4506e-09, 1.6764e-08], + [ 3.4198e-06, 5.1707e-06, 6.9942e-07, ..., 0.0000e+00, + 9.4716e-07, 4.5188e-06], + [ 1.1362e-07, 1.6298e-07, 4.1630e-07, ..., 0.0000e+00, + 1.9558e-08, 1.8068e-07]], device='cuda:0') +Epoch 322, bias, value: tensor([ 0.0279, 0.0060, 0.0193, 0.0178, 0.0414, 0.0097, -0.0030, 0.0016, + -0.0054, -0.0434], device='cuda:0'), grad: tensor([ 3.3677e-06, 1.9483e-06, -5.4017e-06, 2.5090e-06, -2.6636e-07, + -2.0549e-05, 8.0466e-06, 3.0361e-07, 8.8811e-06, 1.1297e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 218.25, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4929 re_mapping 0.0045 re_causal 0.0138 /// teacc 98.91 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.1654, 0.2242, -0.0334, ..., -0.0190, 0.0655, -0.0365], + [ 0.1908, -0.1032, 0.0364, ..., -0.0885, 0.0125, -0.0813], + [-0.0857, -0.2649, -0.0394, ..., -0.0650, 0.0354, -0.2695], + ..., + [-0.1890, -0.3007, 0.0585, ..., -0.1025, -0.0822, -0.1827], + [-0.1325, 0.0046, 0.0397, ..., 0.0089, -0.1118, 0.0755], + [-0.1100, -0.0728, 0.1024, ..., 0.2020, -0.0740, -0.1984]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -6.5193e-09, 4.5635e-08, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-08], + [-1.7695e-08, 8.3819e-09, 2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 2.9802e-08], + [ 2.7940e-09, 5.1223e-08, 2.4121e-07, ..., 0.0000e+00, + 0.0000e+00, 1.7323e-07], + ..., + [ 1.0245e-08, 5.2154e-07, 2.4028e-06, ..., 0.0000e+00, + 0.0000e+00, 1.6820e-06], + [ 1.8626e-09, -1.1148e-06, -5.0850e-06, ..., 0.0000e+00, + 0.0000e+00, -3.5781e-06], + [ 1.8626e-09, 4.0606e-07, 1.7788e-06, ..., 0.0000e+00, + 0.0000e+00, 1.3094e-06]], device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0279, 0.0076, 0.0194, 0.0163, 0.0414, 0.0111, -0.0027, 0.0001, + -0.0065, -0.0436], device='cuda:0'), grad: tensor([ 9.1270e-08, 1.5367e-07, 4.9453e-07, 1.1176e-08, 1.8626e-09, + 1.1073e-06, 7.3574e-08, 5.4948e-06, -1.1511e-05, 4.0680e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 217.79, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4899 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.1655, 0.2243, -0.0334, ..., -0.0194, 0.0655, -0.0356], + [ 0.1909, -0.1036, 0.0361, ..., -0.0900, 0.0125, -0.0816], + [-0.0860, -0.2657, -0.0399, ..., -0.0650, 0.0351, -0.2718], + ..., + [-0.1889, -0.3020, 0.0590, ..., -0.1026, -0.0822, -0.1834], + [-0.1356, 0.0022, 0.0389, ..., 0.0086, -0.1148, 0.0734], + [-0.1099, -0.0726, 0.1023, ..., 0.2031, -0.0741, -0.1985]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.5146e-08, 5.0291e-08, ..., 1.9558e-08, + 0.0000e+00, 1.2107e-08], + [ 2.7940e-09, 5.4017e-08, 1.2545e-06, ..., 5.4017e-08, + 0.0000e+00, 4.6566e-08], + [ 2.7940e-09, 1.3039e-08, -2.3432e-06, ..., 9.3132e-09, + 0.0000e+00, 1.8626e-08], + ..., + [ 4.6566e-09, 1.0245e-08, 4.1071e-07, ..., 7.4506e-09, + 0.0000e+00, 2.4214e-08], + [ 4.4703e-08, 5.5879e-09, 2.5146e-08, ..., 6.5193e-09, + 0.0000e+00, 1.2759e-07], + [ 1.3970e-08, -1.6578e-07, 4.1071e-07, ..., -1.9372e-07, + 0.0000e+00, -3.1665e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0278, 0.0074, 0.0191, 0.0163, 0.0413, 0.0131, -0.0027, 0.0006, + -0.0086, -0.0437], device='cuda:0'), grad: tensor([ 2.1309e-06, 3.4925e-06, -6.3404e-06, -1.4892e-06, 3.2783e-07, + 1.6484e-06, -2.3767e-06, 1.1623e-06, 3.6042e-07, 1.0673e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 218.18, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.5075 re_mapping 0.0044 re_causal 0.0141 /// teacc 98.93 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.1655, 0.2247, -0.0335, ..., -0.0197, 0.0657, -0.0342], + [ 0.1909, -0.1037, 0.0359, ..., -0.0906, 0.0125, -0.0818], + [-0.0860, -0.2664, -0.0400, ..., -0.0643, 0.0351, -0.2726], + ..., + [-0.1889, -0.3029, 0.0592, ..., -0.1029, -0.0821, -0.1841], + [-0.1357, 0.0018, 0.0385, ..., 0.0078, -0.1148, 0.0735], + [-0.1099, -0.0722, 0.1024, ..., 0.2044, -0.0741, -0.1981]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -8.1584e-07, -3.9954e-07, ..., -8.2888e-08, + 9.3132e-10, -1.8626e-09], + [-4.3306e-07, 4.8429e-08, -3.3900e-07, ..., 5.5879e-09, + 0.0000e+00, 9.3132e-10], + [ 1.3039e-08, 1.5553e-07, 1.3597e-07, ..., 1.8626e-08, + -2.7940e-09, 1.8626e-09], + ..., + [ 3.2224e-07, 1.1455e-07, 5.3085e-08, ..., 1.3970e-08, + 9.3132e-10, 3.7253e-09], + [ 3.0734e-08, 4.7497e-08, 4.7497e-08, ..., 3.7253e-09, + 0.0000e+00, -1.3039e-08], + [ 2.0489e-08, 2.4587e-07, 2.8498e-07, ..., 2.8871e-08, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 325, bias, value: tensor([ 0.0277, 0.0073, 0.0192, 0.0166, 0.0412, 0.0127, -0.0026, 0.0007, + -0.0091, -0.0434], device='cuda:0'), grad: tensor([-2.3037e-05, -5.9977e-07, 1.6717e-06, 1.1455e-06, 1.4253e-05, + 7.7300e-08, 4.0606e-06, 2.3376e-07, 4.8801e-07, 1.6615e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 218.18, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.5030 re_mapping 0.0043 re_causal 0.0137 /// teacc 99.07 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.1652, 0.2259, -0.0332, ..., -0.0197, 0.0658, -0.0312], + [ 0.1912, -0.1038, 0.0359, ..., -0.0907, 0.0125, -0.0823], + [-0.0865, -0.2676, -0.0410, ..., -0.0637, 0.0350, -0.2738], + ..., + [-0.1889, -0.3036, 0.0593, ..., -0.1030, -0.0821, -0.1844], + [-0.1362, 0.0015, 0.0382, ..., 0.0061, -0.1148, 0.0725], + [-0.1100, -0.0722, 0.1026, ..., 0.2051, -0.0740, -0.1981]], + device='cuda:0'), grad: tensor([[ 1.0803e-07, 6.0070e-07, 3.4459e-08, ..., 4.6566e-09, + 0.0000e+00, -9.3132e-09], + [ 9.4622e-07, 5.1223e-08, 5.9307e-06, ..., 6.2026e-07, + 4.8149e-07, 6.5193e-09], + [ 5.4017e-08, -1.2014e-07, 1.4901e-06, ..., 5.5879e-09, + 3.7253e-09, -1.0245e-08], + ..., + [-1.6112e-07, 1.0245e-08, -5.7705e-06, ..., 8.3819e-09, + 7.4506e-09, -6.5193e-09], + [ 2.1420e-08, 1.1362e-07, 7.5437e-08, ..., 1.0245e-08, + 7.4506e-09, 1.3039e-08], + [-9.2760e-07, 4.8429e-08, -2.9951e-06, ..., -6.8732e-07, + -5.3365e-07, 2.7940e-09]], device='cuda:0') +Epoch 326, bias, value: tensor([ 0.0283, 0.0073, 0.0187, 0.0171, 0.0426, 0.0133, -0.0032, 0.0002, + -0.0100, -0.0434], device='cuda:0'), grad: tensor([ 7.2755e-06, 1.5467e-05, 3.2671e-06, 3.5949e-06, 3.8967e-06, + 9.3877e-07, -1.2435e-05, -1.8999e-05, 1.0943e-06, -4.1313e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 218.13, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4892 re_mapping 0.0044 re_causal 0.0137 /// teacc 99.00 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.1652, 0.2263, -0.0334, ..., -0.0205, 0.0658, -0.0313], + [ 0.1913, -0.1044, 0.0358, ..., -0.0917, 0.0125, -0.0831], + [-0.0865, -0.2677, -0.0413, ..., -0.0636, 0.0349, -0.2738], + ..., + [-0.1891, -0.3050, 0.0592, ..., -0.1038, -0.0824, -0.1846], + [-0.1360, 0.0017, 0.0383, ..., 0.0055, -0.1147, 0.0729], + [-0.1098, -0.0722, 0.1037, ..., 0.2063, -0.0739, -0.1982]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.3935e-07, -5.1223e-08, ..., 0.0000e+00, + -4.5635e-08, -4.6566e-09], + [-4.6566e-09, 9.3132e-09, -8.3819e-09, ..., 0.0000e+00, + 1.8626e-09, 1.1176e-08], + [ 3.7253e-09, 4.6566e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + ..., + [ 1.7695e-08, 1.8626e-09, 2.6077e-08, ..., 9.3132e-10, + 0.0000e+00, 1.3970e-08], + [ 7.4506e-09, 1.0245e-08, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 2.7940e-09], + [ 4.6566e-09, 3.2596e-08, -1.3039e-08, ..., -9.3132e-10, + 3.7253e-09, 2.7940e-09]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0283, 0.0073, 0.0188, 0.0181, 0.0417, 0.0100, -0.0020, 0.0002, + -0.0098, -0.0426], device='cuda:0'), grad: tensor([-2.8871e-07, 5.8673e-08, 3.2596e-08, -1.3411e-07, 4.3772e-08, + 3.9116e-08, 6.3330e-08, 1.1083e-07, 4.1910e-08, 4.1910e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 217.82, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4981 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.03 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.1657, 0.2278, -0.0331, ..., -0.0205, 0.0659, -0.0319], + [ 0.1915, -0.1047, 0.0359, ..., -0.0920, 0.0125, -0.0833], + [-0.0868, -0.2690, -0.0414, ..., -0.0640, 0.0348, -0.2745], + ..., + [-0.1892, -0.3058, 0.0594, ..., -0.1040, -0.0830, -0.1848], + [-0.1363, 0.0006, 0.0374, ..., 0.0056, -0.1147, 0.0724], + [-0.1101, -0.0718, 0.1037, ..., 0.2064, -0.0740, -0.1981]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -4.6566e-09, 4.8429e-08, ..., 9.3132e-10, + -9.3132e-10, 1.8626e-09], + [-7.4506e-09, 1.8626e-09, 2.7940e-09, ..., 9.3132e-10, + 9.3132e-10, 4.6566e-09], + [-6.5193e-08, 3.7253e-09, 1.0245e-08, ..., -1.5926e-07, + 0.0000e+00, 7.1712e-08], + ..., + [ 6.5193e-09, 4.6566e-09, 2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, 2.0489e-08], + [ 1.8626e-09, -2.6077e-08, -7.4506e-08, ..., 0.0000e+00, + -1.8626e-09, -6.3330e-08], + [ 6.9849e-08, 2.4214e-08, -5.4948e-08, ..., 1.5553e-07, + 2.7940e-09, 4.2841e-08]], device='cuda:0') +Epoch 328, bias, value: tensor([ 0.0290, 0.0073, 0.0187, 0.0179, 0.0416, 0.0112, -0.0031, 0.0002, + -0.0111, -0.0425], device='cuda:0'), grad: tensor([ 1.8999e-07, 8.7544e-08, -2.4643e-06, 2.2352e-08, 3.7439e-07, + -1.1176e-08, 2.4214e-08, 2.4773e-07, -1.0151e-07, 1.6280e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 218.17, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4891 re_mapping 0.0046 re_causal 0.0135 /// teacc 99.02 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.1658, 0.2284, -0.0349, ..., -0.0232, 0.0672, -0.0320], + [ 0.1915, -0.1064, 0.0358, ..., -0.0922, 0.0125, -0.0839], + [-0.0867, -0.2708, -0.0419, ..., -0.0645, 0.0345, -0.2751], + ..., + [-0.1892, -0.3072, 0.0596, ..., -0.1040, -0.0833, -0.1846], + [-0.1364, 0.0007, 0.0377, ..., 0.0060, -0.1147, 0.0727], + [-0.1102, -0.0718, 0.1047, ..., 0.2081, -0.0755, -0.1984]], + device='cuda:0'), grad: tensor([[ 1.6764e-08, 7.4506e-09, 2.9802e-08, ..., 0.0000e+00, + 2.7940e-09, 1.9558e-08], + [-8.9407e-08, -2.6077e-08, -8.1025e-08, ..., 0.0000e+00, + -1.3970e-08, 7.4506e-09], + [ 6.5193e-09, 4.6566e-08, 6.1467e-08, ..., 0.0000e+00, + 9.3132e-10, 3.6322e-08], + ..., + [ 1.5832e-08, 8.3819e-09, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 2.7940e-09], + [ 2.4214e-08, -2.2259e-07, -2.3376e-07, ..., -1.3970e-08, + 3.7253e-09, -2.0396e-07], + [ 1.6764e-08, 9.4064e-08, 1.1083e-07, ..., 8.3819e-09, + 2.7940e-09, 6.5193e-08]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0282, 0.0071, 0.0189, 0.0169, 0.0423, 0.0115, -0.0032, 0.0004, + -0.0110, -0.0423], device='cuda:0'), grad: tensor([ 1.1176e-07, -1.5832e-07, 1.6578e-07, 1.6764e-08, 1.7695e-08, + 3.5390e-08, 2.1793e-07, 8.3819e-09, -7.1898e-07, 2.9802e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 217.94, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.4980 re_mapping 0.0046 re_causal 0.0127 /// teacc 99.00 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.1659, 0.2291, -0.0346, ..., -0.0224, 0.0676, -0.0317], + [ 0.1918, -0.1078, 0.0354, ..., -0.0932, 0.0124, -0.0851], + [-0.0869, -0.2720, -0.0420, ..., -0.0630, 0.0341, -0.2771], + ..., + [-0.1894, -0.3090, 0.0600, ..., -0.1047, -0.0833, -0.1848], + [-0.1365, 0.0022, 0.0389, ..., 0.0080, -0.1139, 0.0743], + [-0.1103, -0.0731, 0.1045, ..., 0.2074, -0.0754, -0.1995]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.3039e-08, 5.5879e-09, ..., 0.0000e+00, + 1.6764e-08, 7.4506e-09], + [-8.0094e-08, 1.8626e-09, 1.8068e-07, ..., 0.0000e+00, + 2.4214e-07, 5.5879e-08], + [ 4.8429e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 4.2841e-08], + ..., + [ 5.0291e-08, 7.4506e-09, -3.9116e-08, ..., 0.0000e+00, + -3.7253e-09, 7.6368e-08], + [ 4.6566e-08, 9.3132e-09, 1.4901e-08, ..., 0.0000e+00, + 2.0489e-08, 3.9116e-08], + [ 3.9116e-08, 1.1176e-08, 3.7253e-09, ..., -1.8626e-09, + 3.7253e-09, 1.1362e-07]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0286, 0.0068, 0.0189, 0.0171, 0.0427, 0.0112, -0.0032, 0.0006, + -0.0098, -0.0430], device='cuda:0'), grad: tensor([ 3.7067e-07, 1.9446e-06, 6.1840e-07, -1.5236e-06, 6.8918e-08, + 5.1968e-07, -3.0249e-06, 2.2538e-07, 3.7812e-07, 4.1537e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 218.19, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4682 re_mapping 0.0045 re_causal 0.0125 /// teacc 99.04 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.1664, 0.2299, -0.0344, ..., -0.0223, 0.0707, -0.0321], + [ 0.1919, -0.1080, 0.0352, ..., -0.0962, 0.0124, -0.0859], + [-0.0869, -0.2711, -0.0426, ..., -0.0622, 0.0338, -0.2771], + ..., + [-0.1894, -0.3109, 0.0602, ..., -0.1058, -0.0838, -0.1851], + [-0.1368, 0.0025, 0.0393, ..., 0.0080, -0.1138, 0.0748], + [-0.1103, -0.0737, 0.1048, ..., 0.2081, -0.0761, -0.1999]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.4506e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-1.8626e-09, 2.4214e-08, 1.1176e-07, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-08], + [ 0.0000e+00, -2.7008e-07, -3.7253e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 0.0000e+00, 1.8626e-09, -3.9190e-06, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-7.4506e-09, 2.1793e-07, 4.6566e-08, ..., 0.0000e+00, + 0.0000e+00, -5.0291e-07], + [ 3.7253e-09, 9.3132e-09, 3.6545e-06, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0288, 0.0064, 0.0200, 0.0167, 0.0426, 0.0111, -0.0033, 0.0007, + -0.0096, -0.0431], device='cuda:0'), grad: tensor([ 3.9116e-08, 5.0478e-07, -1.6522e-06, 4.4331e-07, 1.3970e-07, + 5.4017e-08, 1.1194e-06, -6.0722e-06, -3.3900e-07, 5.7332e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 218.29, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4963 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.06 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.1667, 0.2300, -0.0343, ..., -0.0224, 0.0712, -0.0322], + [ 0.1919, -0.1090, 0.0352, ..., -0.0962, 0.0123, -0.0865], + [-0.0870, -0.2722, -0.0430, ..., -0.0631, 0.0335, -0.2775], + ..., + [-0.1895, -0.3121, 0.0609, ..., -0.1059, -0.0838, -0.1865], + [-0.1370, 0.0026, 0.0396, ..., 0.0080, -0.1140, 0.0751], + [-0.1103, -0.0736, 0.1037, ..., 0.2077, -0.0763, -0.1999]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-08, -3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 3.7253e-09, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -7.4506e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, -9.3132e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 1.1176e-08, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.9802e-08, 4.6566e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 332, bias, value: tensor([ 0.0288, 0.0064, 0.0200, 0.0175, 0.0426, 0.0120, -0.0035, 0.0011, + -0.0095, -0.0439], device='cuda:0'), grad: tensor([-1.0245e-07, 3.1665e-08, -5.0291e-08, 2.9802e-08, -1.9558e-07, + 2.0489e-08, 2.9802e-08, -1.5832e-07, 8.5682e-08, 2.9616e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 218.32, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4845 re_mapping 0.0049 re_causal 0.0134 /// teacc 98.98 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.1669, 0.2307, -0.0338, ..., -0.0224, 0.0717, -0.0323], + [ 0.1919, -0.1105, 0.0341, ..., -0.0984, 0.0125, -0.0867], + [-0.0871, -0.2742, -0.0433, ..., -0.0633, 0.0331, -0.2781], + ..., + [-0.1893, -0.3137, 0.0619, ..., -0.1079, -0.0846, -0.1868], + [-0.1372, 0.0028, 0.0397, ..., 0.0081, -0.1141, 0.0751], + [-0.1104, -0.0740, 0.1036, ..., 0.2089, -0.0765, -0.2002]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 1.3039e-08, 5.2154e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-2.6636e-07, -5.9605e-08, -3.3528e-07, ..., 0.0000e+00, + -1.6764e-08, 0.0000e+00], + [-7.4506e-09, 0.0000e+00, -2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 8.9407e-08, 5.5879e-09, 4.2841e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 7.0781e-08, 4.6566e-08, 8.5682e-08, ..., 0.0000e+00, + 3.7253e-09, 2.9802e-08], + [ 9.4995e-08, 3.1665e-08, 1.4156e-07, ..., 0.0000e+00, + 7.4506e-09, 3.7253e-09]], device='cuda:0') +Epoch 333, bias, value: tensor([ 0.0291, 0.0058, 0.0199, 0.0178, 0.0425, 0.0119, -0.0034, 0.0015, + -0.0097, -0.0440], device='cuda:0'), grad: tensor([ 1.2852e-07, -4.5076e-07, -1.8626e-07, 4.2841e-08, 3.5390e-08, + -1.0990e-07, 5.5879e-09, 1.1176e-07, 1.8068e-07, 2.3842e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 218.41, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4716 re_mapping 0.0045 re_causal 0.0131 /// teacc 99.07 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.1670, 0.2321, -0.0328, ..., -0.0218, 0.0719, -0.0320], + [ 0.1919, -0.1113, 0.0340, ..., -0.0986, 0.0125, -0.0875], + [-0.0871, -0.2750, -0.0434, ..., -0.0635, 0.0330, -0.2790], + ..., + [-0.1894, -0.3155, 0.0626, ..., -0.1082, -0.0850, -0.1868], + [-0.1372, 0.0031, 0.0401, ..., 0.0080, -0.1139, 0.0758], + [-0.1107, -0.0752, 0.1025, ..., 0.2088, -0.0765, -0.2005]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0675e-07, -1.4715e-07, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 5.5879e-09, 3.9116e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 3.7253e-09, 5.5879e-09, -1.0058e-07, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 1.2666e-07, 1.2666e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 334, bias, value: tensor([ 0.0298, 0.0056, 0.0199, 0.0150, 0.0424, 0.0140, -0.0034, 0.0020, + -0.0096, -0.0448], device='cuda:0'), grad: tensor([-2.7940e-07, 9.8720e-08, -6.7055e-08, -9.8348e-07, 0.0000e+00, + 1.0505e-06, 3.9116e-08, -1.4901e-07, 3.9116e-08, 2.5891e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 217.99, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5145 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.06 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.1677, 0.2328, -0.0324, ..., -0.0217, 0.0715, -0.0322], + [ 0.1917, -0.1112, 0.0338, ..., -0.0987, 0.0126, -0.0895], + [-0.0871, -0.2765, -0.0440, ..., -0.0635, 0.0326, -0.2799], + ..., + [-0.1889, -0.3177, 0.0628, ..., -0.1083, -0.0851, -0.1871], + [-0.1375, 0.0034, 0.0402, ..., 0.0080, -0.1141, 0.0763], + [-0.1113, -0.0760, 0.1024, ..., 0.2089, -0.0768, -0.2014]], + device='cuda:0'), grad: tensor([[ 4.2841e-08, -2.2911e-07, -1.8813e-07, ..., 0.0000e+00, + -4.0978e-08, -1.4901e-08], + [-2.9802e-08, 1.0803e-07, 2.4214e-08, ..., 0.0000e+00, + 1.1176e-08, 3.3528e-08], + [ 3.7253e-09, 9.3132e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08], + ..., + [ 7.4506e-08, 7.2643e-08, 3.7253e-08, ..., 0.0000e+00, + 3.7253e-09, 5.5879e-08], + [ 1.8440e-07, 2.9802e-07, 7.6368e-08, ..., 0.0000e+00, + 1.4901e-08, 1.1921e-07], + [ 1.0245e-07, 1.4529e-07, 1.4901e-08, ..., 0.0000e+00, + 3.7253e-09, 6.1467e-08]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0301, 0.0049, 0.0198, 0.0119, 0.0427, 0.0172, -0.0037, 0.0025, + -0.0092, -0.0453], device='cuda:0'), grad: tensor([-3.4831e-07, 2.2165e-07, -1.3039e-08, -2.2352e-08, -1.3225e-07, + -7.9907e-07, 5.9605e-08, 2.2538e-07, 4.8056e-07, 3.0920e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 218.50, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.5048 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.03 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.1691, 0.2330, -0.0324, ..., -0.0217, 0.0705, -0.0323], + [ 0.1920, -0.1101, 0.0339, ..., -0.0990, 0.0137, -0.0901], + [-0.0871, -0.2769, -0.0430, ..., -0.0635, 0.0341, -0.2797], + ..., + [-0.1891, -0.3188, 0.0627, ..., -0.1085, -0.0881, -0.1877], + [-0.1381, 0.0033, 0.0400, ..., 0.0080, -0.1156, 0.0772], + [-0.1115, -0.0762, 0.1026, ..., 0.2091, -0.0771, -0.2016]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.3597e-07, 2.3097e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-07], + [ 7.4506e-08, 8.9407e-08, 8.2701e-07, ..., 0.0000e+00, + 6.8918e-08, 7.4506e-08], + [ 1.8626e-09, 3.5390e-07, 6.8173e-07, ..., 0.0000e+00, + 0.0000e+00, 1.5832e-07], + ..., + [-1.0058e-07, 9.8720e-08, -4.7311e-07, ..., 0.0000e+00, + -9.4995e-08, 8.9407e-08], + [-1.8626e-09, -2.7306e-06, -4.0308e-06, ..., 0.0000e+00, + 0.0000e+00, -2.5984e-06], + [ 5.7742e-08, 1.8869e-06, 2.3413e-06, ..., 0.0000e+00, + 2.6077e-08, 2.0023e-06]], device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0301, 0.0048, 0.0206, 0.0119, 0.0429, 0.0171, -0.0037, 0.0025, + -0.0093, -0.0455], device='cuda:0'), grad: tensor([ 5.7742e-07, 1.1064e-06, 1.4827e-06, 2.3283e-07, 2.6636e-07, + -4.6492e-06, 4.8652e-06, -2.1420e-07, -9.7603e-06, 6.0685e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 218.37, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4846 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.01 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.1710, 0.2333, -0.0323, ..., -0.0218, 0.0700, -0.0332], + [ 0.1920, -0.1097, 0.0332, ..., -0.0994, 0.0140, -0.0910], + [-0.0871, -0.2775, -0.0446, ..., -0.0634, 0.0338, -0.2807], + ..., + [-0.1889, -0.3198, 0.0636, ..., -0.1087, -0.0882, -0.1874], + [-0.1383, 0.0035, 0.0401, ..., 0.0080, -0.1159, 0.0777], + [-0.1117, -0.0764, 0.1026, ..., 0.2092, -0.0772, -0.2020]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0433e-06, -6.5938e-07, ..., 0.0000e+00, + -1.6950e-07, -2.5518e-07], + [-5.2154e-08, 3.9674e-07, 1.8254e-07, ..., 0.0000e+00, + 3.1665e-08, 5.5879e-08], + [ 9.1270e-08, 1.3635e-06, 7.7859e-07, ..., 0.0000e+00, + 7.4506e-09, 1.1958e-06], + ..., + [ 1.6764e-08, 3.3528e-08, -4.0978e-07, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-09], + [-1.3597e-07, -2.1737e-06, -1.3430e-06, ..., 0.0000e+00, + 2.2352e-08, -2.2706e-06], + [ 4.2841e-08, 3.7253e-07, 4.1910e-07, ..., 0.0000e+00, + 2.7940e-08, 8.3819e-08]], device='cuda:0') +Epoch 337, bias, value: tensor([ 0.0301, 0.0044, 0.0204, 0.0119, 0.0429, 0.0171, -0.0038, 0.0030, + -0.0093, -0.0456], device='cuda:0'), grad: tensor([-4.7907e-06, 1.4026e-06, 3.0566e-06, 2.1420e-07, 7.4320e-07, + 2.1271e-06, 1.8273e-06, -2.8871e-07, -5.6066e-06, 1.2890e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 218.28, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4558 re_mapping 0.0044 re_causal 0.0124 /// teacc 98.99 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.1718, 0.2335, -0.0323, ..., -0.0218, 0.0701, -0.0346], + [ 0.1919, -0.1098, 0.0317, ..., -0.0996, 0.0142, -0.0932], + [-0.0872, -0.2784, -0.0455, ..., -0.0635, 0.0346, -0.2823], + ..., + [-0.1886, -0.3219, 0.0650, ..., -0.1088, -0.0894, -0.1857], + [-0.1388, 0.0041, 0.0398, ..., 0.0081, -0.1157, 0.0785], + [-0.1118, -0.0767, 0.1031, ..., 0.2096, -0.0773, -0.2023]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-2.6450e-07, -1.6764e-08, -1.1176e-07, ..., 0.0000e+00, + -7.4506e-09, -2.6077e-08], + [ 7.4506e-08, 9.3132e-09, 4.8429e-08, ..., 0.0000e+00, + 1.8626e-09, 1.6764e-08], + ..., + [ 8.0094e-08, 3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 2.2352e-08, 0.0000e+00, 2.6077e-08, ..., 3.7253e-09, + 1.8626e-09, -1.8626e-09], + [ 9.3132e-09, -4.0978e-08, -2.2911e-07, ..., -4.4703e-08, + 0.0000e+00, -7.4506e-08]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0301, 0.0034, 0.0204, 0.0119, 0.0426, 0.0171, -0.0038, 0.0039, + -0.0094, -0.0454], device='cuda:0'), grad: tensor([ 1.5460e-07, -3.2037e-07, 1.5460e-07, 1.5274e-07, 6.2212e-07, + 5.4017e-08, -1.9185e-07, 6.1467e-08, 5.9605e-08, -7.5996e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 218.24, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4833 re_mapping 0.0044 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.1719, 0.2365, -0.0300, ..., -0.0217, 0.0710, -0.0340], + [ 0.1930, -0.1103, 0.0321, ..., -0.0997, 0.0144, -0.0933], + [-0.0889, -0.2791, -0.0484, ..., -0.0635, 0.0329, -0.2831], + ..., + [-0.1887, -0.3224, 0.0649, ..., -0.1089, -0.0897, -0.1858], + [-0.1390, 0.0043, 0.0413, ..., 0.0081, -0.1158, 0.0813], + [-0.1118, -0.0769, 0.1037, ..., 0.2105, -0.0782, -0.2023]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.0897e-07, -8.5123e-07, ..., 0.0000e+00, + -9.4995e-08, -1.9558e-07], + [ 0.0000e+00, 3.3528e-08, 3.5390e-08, ..., 0.0000e+00, + 5.5879e-09, 1.6764e-08], + [ 0.0000e+00, 9.3132e-09, -4.0978e-08, ..., 0.0000e+00, + 1.8626e-09, -5.5879e-08], + ..., + [ 1.8626e-09, 1.3039e-08, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 1.6764e-07], + [ 0.0000e+00, 1.0617e-07, 1.0245e-07, ..., 0.0000e+00, + 1.1176e-08, 2.7940e-08], + [ 0.0000e+00, 6.2026e-07, 5.6438e-07, ..., 0.0000e+00, + 6.5193e-08, 1.5646e-07]], device='cuda:0') +Epoch 339, bias, value: tensor([ 0.0319, 0.0039, 0.0190, 0.0119, 0.0422, 0.0171, -0.0064, 0.0038, + -0.0075, -0.0449], device='cuda:0'), grad: tensor([-2.1569e-06, 1.6205e-07, -6.8918e-07, -2.7977e-06, 1.5274e-07, + 2.1681e-06, 2.0303e-07, 1.1865e-06, 2.9057e-07, 1.4585e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 217.80, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4840 re_mapping 0.0044 re_causal 0.0129 /// teacc 99.01 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.1723, 0.2371, -0.0296, ..., -0.0217, 0.0713, -0.0343], + [ 0.1931, -0.1112, 0.0320, ..., -0.1005, 0.0140, -0.0938], + [-0.0889, -0.2793, -0.0481, ..., -0.0635, 0.0334, -0.2840], + ..., + [-0.1887, -0.3227, 0.0650, ..., -0.1089, -0.0879, -0.1854], + [-0.1390, 0.0048, 0.0422, ..., 0.0081, -0.1158, 0.0822], + [-0.1117, -0.0771, 0.1037, ..., 0.2107, -0.0786, -0.2025]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.7253e-09, 2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 5.5879e-09, 0.0000e+00, 3.4086e-07, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + [ 1.8626e-09, 0.0000e+00, 6.5193e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.4901e-08, 0.0000e+00, -5.2154e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-08], + [ 1.4901e-08, -1.8626e-09, 6.3330e-08, ..., 0.0000e+00, + 0.0000e+00, 4.0978e-08], + [ 3.7253e-09, 1.8626e-09, -2.4401e-07, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08]], device='cuda:0') +Epoch 340, bias, value: tensor([ 0.0322, 0.0037, 0.0195, 0.0120, 0.0420, 0.0171, -0.0067, 0.0040, + -0.0069, -0.0451], device='cuda:0'), grad: tensor([ 4.8615e-07, 5.6624e-07, -4.4703e-07, -3.9041e-05, -1.9185e-07, + 3.8981e-05, 1.6764e-08, -7.3016e-07, 3.3155e-07, 5.2154e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 218.30, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4844 re_mapping 0.0045 re_causal 0.0128 /// teacc 99.04 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.1735, 0.2373, -0.0295, ..., -0.0219, 0.0714, -0.0354], + [ 0.1930, -0.1119, 0.0318, ..., -0.1007, 0.0139, -0.0940], + [-0.0889, -0.2795, -0.0478, ..., -0.0636, 0.0333, -0.2852], + ..., + [-0.1886, -0.3234, 0.0651, ..., -0.1089, -0.0873, -0.1855], + [-0.1393, 0.0049, 0.0422, ..., 0.0080, -0.1158, 0.0821], + [-0.1118, -0.0771, 0.1038, ..., 0.2111, -0.0789, -0.2029]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.4506e-09, 3.9674e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-2.2352e-08, 1.1176e-08, 5.5879e-09, ..., 0.0000e+00, + 3.7253e-09, 5.0291e-08], + [ 1.8626e-09, 0.0000e+00, -7.2643e-07, ..., 0.0000e+00, + 0.0000e+00, 2.0489e-08], + ..., + [ 5.5879e-09, 3.7253e-09, 3.9674e-07, ..., 0.0000e+00, + 1.8626e-09, 2.4773e-07], + [ 9.3132e-09, -6.1467e-08, 7.4506e-08, ..., 0.0000e+00, + -2.0489e-08, 3.7253e-08], + [ 0.0000e+00, 7.4506e-09, -1.7881e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 341, bias, value: tensor([ 0.0322, 0.0034, 0.0197, 0.0121, 0.0420, 0.0170, -0.0067, 0.0042, + -0.0071, -0.0452], device='cuda:0'), grad: tensor([ 1.0021e-06, 1.1921e-07, -1.8515e-06, -9.7603e-07, 7.2643e-08, + 1.7136e-07, 5.5879e-09, 1.3616e-06, 3.8370e-07, -2.9057e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 340---------------------------------------------------- +epoch 340, time 218.96, cls_loss 0.0012 cls_loss_mapping 0.0027 cls_loss_causal 0.4766 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.11 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.1694, 0.2396, -0.0288, ..., -0.0219, 0.0700, -0.0310], + [ 0.1912, -0.1115, 0.0283, ..., -0.1014, 0.0143, -0.0941], + [-0.0890, -0.2797, -0.0499, ..., -0.0636, 0.0301, -0.2856], + ..., + [-0.1861, -0.3241, 0.0684, ..., -0.1116, -0.0848, -0.1859], + [-0.1398, 0.0047, 0.0421, ..., 0.0079, -0.1161, 0.0821], + [-0.1120, -0.0774, 0.1049, ..., 0.2125, -0.0791, -0.2032]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.0713e-06, -3.2037e-07, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [-1.4901e-08, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, -1.7136e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-09, 0.0000e+00, 6.7055e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 3.7253e-09, 1.1176e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 3.7253e-08, -2.7940e-08, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0336, 0.0009, 0.0188, 0.0121, 0.0419, 0.0167, -0.0069, 0.0066, + -0.0074, -0.0447], device='cuda:0'), grad: tensor([-3.2615e-06, -1.8626e-09, -1.3299e-06, 3.1665e-08, 3.1665e-08, + 4.2841e-08, 3.2764e-06, 3.3341e-07, 8.6240e-07, 1.4901e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 218.03, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4937 re_mapping 0.0044 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.1694, 0.2402, -0.0286, ..., -0.0219, 0.0698, -0.0310], + [ 0.1920, -0.1100, 0.0288, ..., -0.1019, 0.0160, -0.0926], + [-0.0891, -0.2802, -0.0499, ..., -0.0636, 0.0300, -0.2862], + ..., + [-0.1863, -0.3249, 0.0684, ..., -0.1118, -0.0849, -0.1861], + [-0.1403, 0.0045, 0.0417, ..., 0.0078, -0.1166, 0.0822], + [-0.1121, -0.0775, 0.1050, ..., 0.2130, -0.0792, -0.2033]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.6764e-08, -7.4506e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 2.7940e-08, 0.0000e+00, 1.3784e-07, ..., 0.0000e+00, + 3.7253e-09, 5.4017e-08], + [ 6.8918e-08, 0.0000e+00, -1.7323e-07, ..., 0.0000e+00, + -4.2841e-08, 1.1735e-07], + ..., + [ 7.4506e-09, 0.0000e+00, 6.1467e-08, ..., 0.0000e+00, + 3.1665e-08, 5.5879e-09], + [ 1.4901e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.0489e-08], + [ 5.5879e-09, 1.1176e-08, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0340, 0.0015, 0.0188, 0.0122, 0.0419, 0.0166, -0.0078, 0.0066, + -0.0078, -0.0448], device='cuda:0'), grad: tensor([ 3.1665e-08, 3.1851e-07, -3.7253e-07, -1.1940e-06, -1.1176e-08, + 8.2329e-07, -7.4506e-08, 3.4645e-07, 8.3819e-08, 2.4214e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 218.05, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4750 re_mapping 0.0045 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.1695, 0.2402, -0.0300, ..., -0.0221, 0.0697, -0.0309], + [ 0.1923, -0.1101, 0.0289, ..., -0.1026, 0.0161, -0.0927], + [-0.0891, -0.2804, -0.0500, ..., -0.0637, 0.0301, -0.2864], + ..., + [-0.1867, -0.3259, 0.0682, ..., -0.1119, -0.0850, -0.1865], + [-0.1404, 0.0049, 0.0423, ..., 0.0077, -0.1167, 0.0826], + [-0.1127, -0.0777, 0.1063, ..., 0.2144, -0.0786, -0.2035]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -5.5879e-08, -3.9116e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 1.8626e-09, -2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-2.7940e-08, -7.4506e-08, -7.6368e-08, ..., 0.0000e+00, + 0.0000e+00, -6.1467e-08], + [ 3.7253e-09, 3.9116e-08, 3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0332, 0.0015, 0.0188, 0.0121, 0.0416, 0.0167, -0.0076, 0.0064, + -0.0074, -0.0437], device='cuda:0'), grad: tensor([-9.3132e-09, 1.3039e-08, 9.3132e-09, 1.0058e-07, 6.1467e-08, + 1.4901e-08, -1.2107e-07, -5.0291e-08, -1.1362e-07, 9.4995e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 217.82, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4913 re_mapping 0.0044 re_causal 0.0133 /// teacc 99.08 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.1696, 0.2402, -0.0301, ..., -0.0221, 0.0693, -0.0311], + [ 0.1925, -0.1099, 0.0289, ..., -0.1027, 0.0160, -0.0928], + [-0.0891, -0.2805, -0.0491, ..., -0.0638, 0.0304, -0.2875], + ..., + [-0.1868, -0.3277, 0.0682, ..., -0.1120, -0.0853, -0.1867], + [-0.1409, 0.0049, 0.0421, ..., 0.0076, -0.1171, 0.0827], + [-0.1129, -0.0779, 0.1063, ..., 0.2150, -0.0787, -0.2037]], + device='cuda:0'), grad: tensor([[-4.2841e-08, -1.2480e-07, -4.2841e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1176e-08, 1.1176e-08, 2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-08], + [ 1.8626e-09, 5.5879e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 5.5879e-09, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 3.7253e-09, -1.6764e-08, -5.2154e-08, ..., 0.0000e+00, + 0.0000e+00, -2.9802e-08], + [ 1.1176e-08, 2.4214e-08, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0331, 0.0015, 0.0195, 0.0121, 0.0416, 0.0167, -0.0073, 0.0063, + -0.0080, -0.0438], device='cuda:0'), grad: tensor([-1.7509e-07, 8.1956e-08, 4.4703e-08, 1.8626e-09, -1.4901e-07, + 5.7742e-08, 7.4506e-08, 3.7253e-08, -1.4715e-07, 1.8254e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 217.83, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4785 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.10 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.1700, 0.2405, -0.0300, ..., -0.0222, 0.0695, -0.0326], + [ 0.1929, -0.1109, 0.0290, ..., -0.1033, 0.0166, -0.0933], + [-0.0888, -0.2815, -0.0493, ..., -0.0640, 0.0291, -0.2886], + ..., + [-0.1871, -0.3290, 0.0681, ..., -0.1132, -0.0857, -0.1874], + [-0.1412, 0.0060, 0.0424, ..., 0.0076, -0.1172, 0.0831], + [-0.1136, -0.0784, 0.1065, ..., 0.2156, -0.0789, -0.2043]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, -1.1176e-08, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-09], + [ 2.2277e-06, 9.6112e-07, -3.5390e-08, ..., 0.0000e+00, + 2.2724e-07, 6.5379e-07], + [ 1.6764e-08, 1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 2.9802e-08, 3.7253e-09, 3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-07, 7.4506e-08, -1.8626e-09, ..., 0.0000e+00, + 1.8626e-08, 5.0291e-08], + [ 1.1809e-06, 5.1409e-07, -7.4506e-09, ..., 0.0000e+00, + 1.1921e-07, 3.4273e-07]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0331, 0.0015, 0.0206, 0.0120, 0.0421, 0.0166, -0.0070, 0.0060, + -0.0085, -0.0440], device='cuda:0'), grad: tensor([ 1.9930e-07, 2.3525e-06, 1.1176e-08, 4.9807e-06, 1.8626e-08, + -8.4043e-06, -7.4692e-07, 8.7544e-08, 2.7008e-07, 1.2424e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 218.40, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4649 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.08 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.1703, 0.2406, -0.0299, ..., -0.0221, 0.0695, -0.0333], + [ 0.1932, -0.1110, 0.0291, ..., -0.1041, 0.0164, -0.0934], + [-0.0885, -0.2832, -0.0493, ..., -0.0640, 0.0291, -0.2881], + ..., + [-0.1876, -0.3304, 0.0680, ..., -0.1133, -0.0858, -0.1875], + [-0.1415, 0.0060, 0.0424, ..., 0.0075, -0.1173, 0.0831], + [-0.1140, -0.0789, 0.1065, ..., 0.2159, -0.0790, -0.2048]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.2352e-08, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.0489e-08, 0.0000e+00, 1.1548e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 3.7253e-09, 0.0000e+00, -1.0245e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 3.7253e-09, 0.0000e+00, 4.0978e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 2.0489e-08, -2.2911e-07, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-08]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0331, 0.0016, 0.0211, 0.0120, 0.0422, 0.0166, -0.0067, 0.0059, + -0.0087, -0.0443], device='cuda:0'), grad: tensor([-4.2841e-08, 1.6391e-07, -3.3528e-08, 1.1176e-08, 2.6263e-07, + 5.5879e-09, 9.3132e-09, -4.8429e-08, 6.1467e-08, -4.0606e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 217.91, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.5289 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.01 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.1703, 0.2410, -0.0297, ..., -0.0221, 0.0696, -0.0325], + [ 0.1942, -0.1107, 0.0290, ..., -0.1046, 0.0166, -0.0934], + [-0.0891, -0.2845, -0.0503, ..., -0.0641, 0.0288, -0.2885], + ..., + [-0.1881, -0.3321, 0.0683, ..., -0.1135, -0.0858, -0.1877], + [-0.1422, 0.0059, 0.0420, ..., 0.0071, -0.1176, 0.0831], + [-0.1142, -0.0793, 0.1065, ..., 0.2164, -0.0792, -0.2048]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.2352e-08, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1958e-06, 1.8626e-09, -7.5996e-07, ..., 0.0000e+00, + -2.6077e-08, 0.0000e+00], + [ 2.9802e-08, 1.8626e-09, 2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 4.6194e-07, 1.8626e-09, 2.5332e-07, ..., 0.0000e+00, + 9.3132e-09, -5.5879e-09], + [ 1.8626e-09, -5.5879e-09, 1.1176e-08, ..., 0.0000e+00, + -1.8626e-09, -7.4506e-09], + [ 5.0478e-07, 1.3039e-08, 2.7753e-07, ..., 0.0000e+00, + 1.1176e-08, 1.8626e-09]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0332, 0.0019, 0.0205, 0.0121, 0.0424, 0.0166, -0.0068, 0.0058, + -0.0093, -0.0444], device='cuda:0'), grad: tensor([ 3.5949e-07, -1.9502e-06, 1.1735e-07, 6.8918e-08, 3.0920e-07, + 4.4703e-08, -4.5262e-07, 6.7800e-07, 2.4214e-08, 7.8976e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 217.93, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4877 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.1703, 0.2412, -0.0296, ..., -0.0218, 0.0694, -0.0327], + [ 0.1947, -0.1104, 0.0292, ..., -0.1058, 0.0169, -0.0937], + [-0.0893, -0.2864, -0.0510, ..., -0.0646, 0.0288, -0.2891], + ..., + [-0.1887, -0.3371, 0.0682, ..., -0.1137, -0.0864, -0.1883], + [-0.1427, 0.0064, 0.0416, ..., 0.0069, -0.1180, 0.0839], + [-0.1151, -0.0799, 0.1065, ..., 0.2167, -0.0793, -0.2055]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0673e-06, -7.1153e-07, ..., 0.0000e+00, + 2.7940e-08, 1.1176e-08], + [ 0.0000e+00, 1.0058e-07, 2.0303e-07, ..., 0.0000e+00, + 0.0000e+00, 7.8231e-08], + [ 0.0000e+00, 3.1665e-08, 3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 0.0000e+00, 9.1828e-07, 2.4736e-06, ..., 0.0000e+00, + 8.7544e-08, 8.7731e-07], + [ 1.8626e-09, -1.0859e-06, -2.7660e-06, ..., 0.0000e+00, + 0.0000e+00, -1.1679e-06], + [ 0.0000e+00, 6.0536e-07, 3.1106e-07, ..., 0.0000e+00, + -1.1735e-07, 1.3784e-07]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0332, 0.0022, 0.0202, 0.0121, 0.0423, 0.0166, -0.0066, 0.0058, + -0.0095, -0.0446], device='cuda:0'), grad: tensor([-2.7437e-06, 3.5577e-07, 1.1735e-07, 1.7881e-07, 1.8068e-07, + 1.8626e-07, 8.1770e-07, 3.7961e-06, -4.1500e-06, 1.2554e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 218.28, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4991 re_mapping 0.0045 re_causal 0.0129 /// teacc 99.05 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.1728, 0.2400, -0.0289, ..., -0.0218, 0.0694, -0.0332], + [ 0.1949, -0.1100, 0.0293, ..., -0.1067, 0.0170, -0.0938], + [-0.0894, -0.2870, -0.0513, ..., -0.0648, 0.0288, -0.2893], + ..., + [-0.1888, -0.3379, 0.0681, ..., -0.1155, -0.0865, -0.1891], + [-0.1437, 0.0060, 0.0413, ..., 0.0064, -0.1184, 0.0841], + [-0.1152, -0.0815, 0.1065, ..., 0.2182, -0.0795, -0.2059]], + device='cuda:0'), grad: tensor([[ 4.8336e-07, 0.0000e+00, 6.1747e-07, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 1.7695e-08, 9.3132e-10, 1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 6.5193e-09, 1.8626e-09, -2.7940e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 8.7544e-08, 6.7055e-08, 1.0710e-07, ..., 0.0000e+00, + 0.0000e+00, 5.7742e-08], + [ 9.3132e-09, 7.4506e-09, -7.4506e-08, ..., 0.0000e+00, + 0.0000e+00, -2.4214e-08]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0315, 0.0022, 0.0203, 0.0121, 0.0421, 0.0165, -0.0044, 0.0057, + -0.0103, -0.0449], device='cuda:0'), grad: tensor([ 6.5118e-06, 2.2724e-07, -3.9861e-07, 4.4517e-07, -7.9162e-08, + 1.2666e-07, -7.4282e-06, -2.3283e-08, 6.2864e-07, -3.6322e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 218.12, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4907 re_mapping 0.0045 re_causal 0.0133 /// teacc 98.99 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.1732, 0.2399, -0.0288, ..., -0.0218, 0.0694, -0.0334], + [ 0.1954, -0.1098, 0.0294, ..., -0.1070, 0.0173, -0.0939], + [-0.0896, -0.2871, -0.0508, ..., -0.0649, 0.0288, -0.2895], + ..., + [-0.1889, -0.3378, 0.0682, ..., -0.1155, -0.0865, -0.1893], + [-0.1445, 0.0058, 0.0400, ..., 0.0063, -0.1188, 0.0839], + [-0.1160, -0.0820, 0.1068, ..., 0.2183, -0.0796, -0.2058]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.7253e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.1735e-07, 1.3039e-08, -1.8533e-07, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + [ 8.6147e-07, 7.4506e-09, 5.3644e-07, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-08], + ..., + [ 2.8871e-08, 1.8626e-09, 8.1025e-08, ..., 0.0000e+00, + 0.0000e+00, 2.1420e-08], + [-3.7253e-09, -2.8871e-08, -1.4873e-06, ..., 0.0000e+00, + 0.0000e+00, -3.9954e-07], + [ 1.1176e-08, 1.1176e-08, 9.9372e-07, ..., 0.0000e+00, + 0.0000e+00, 2.8871e-07]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0312, 0.0023, 0.0208, 0.0121, 0.0418, 0.0165, -0.0041, 0.0057, + -0.0121, -0.0450], device='cuda:0'), grad: tensor([ 7.2643e-08, -7.4133e-07, 1.2396e-06, -1.6764e-08, 5.5879e-08, + 7.3574e-08, -1.2014e-07, 1.1735e-07, -1.7695e-06, 1.0962e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 217.68, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4989 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.11 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.1733, 0.2402, -0.0288, ..., -0.0217, 0.0695, -0.0335], + [ 0.1955, -0.1099, 0.0293, ..., -0.1071, 0.0173, -0.0940], + [-0.0898, -0.2878, -0.0509, ..., -0.0649, 0.0288, -0.2908], + ..., + [-0.1888, -0.3416, 0.0684, ..., -0.1155, -0.0867, -0.1902], + [-0.1454, 0.0058, 0.0400, ..., 0.0063, -0.1191, 0.0842], + [-0.1164, -0.0821, 0.1065, ..., 0.2183, -0.0798, -0.2058]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3039e-08, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-09, 0.0000e+00, 1.9558e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 9.3132e-10, 0.0000e+00, -5.4948e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0313, 0.0023, 0.0208, 0.0122, 0.0418, 0.0165, -0.0042, 0.0059, + -0.0125, -0.0452], device='cuda:0'), grad: tensor([ 9.4064e-08, 1.5087e-07, 4.2841e-08, 3.0734e-08, -1.0477e-06, + 1.2107e-08, 2.2538e-07, 1.0617e-07, 2.3283e-08, 3.7346e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 217.92, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4833 re_mapping 0.0046 re_causal 0.0135 /// teacc 99.01 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.1734, 0.2404, -0.0281, ..., -0.0218, 0.0692, -0.0338], + [ 0.1970, -0.1098, 0.0300, ..., -0.1075, 0.0175, -0.0942], + [-0.0896, -0.2883, -0.0511, ..., -0.0649, 0.0287, -0.2901], + ..., + [-0.1905, -0.3421, 0.0679, ..., -0.1156, -0.0861, -0.1902], + [-0.1457, 0.0059, 0.0399, ..., 0.0062, -0.1193, 0.0842], + [-0.1167, -0.0831, 0.1060, ..., 0.2185, -0.0801, -0.2060]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 3.7253e-09, 0.0000e+00, -1.7695e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, -1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0316, 0.0030, 0.0211, 0.0122, 0.0421, 0.0165, -0.0042, 0.0053, + -0.0127, -0.0462], device='cuda:0'), grad: tensor([ 1.0245e-08, -5.5879e-09, -2.4214e-08, 0.0000e+00, 3.7253e-09, + 5.5879e-09, 9.3132e-10, -3.2596e-08, -8.3819e-09, 4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 217.73, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.5015 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.02 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.1735, 0.2407, -0.0281, ..., -0.0234, 0.0704, -0.0343], + [ 0.1971, -0.1098, 0.0299, ..., -0.1076, 0.0177, -0.0943], + [-0.0897, -0.2891, -0.0512, ..., -0.0654, 0.0286, -0.2906], + ..., + [-0.1904, -0.3422, 0.0681, ..., -0.1157, -0.0862, -0.1904], + [-0.1461, 0.0057, 0.0400, ..., 0.0061, -0.1194, 0.0842], + [-0.1171, -0.0836, 0.1090, ..., 0.2235, -0.0804, -0.2062]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.5437e-08, -9.3132e-08, ..., 0.0000e+00, + 2.7940e-09, -9.3132e-10], + [ 9.3132e-10, 1.3039e-08, 2.5146e-08, ..., 1.8626e-09, + 3.7253e-09, 1.3039e-08], + [ 0.0000e+00, 3.7253e-09, 9.3132e-09, ..., 9.3132e-10, + -2.0489e-08, 8.3819e-09], + ..., + [ 9.3132e-10, 3.7253e-09, 2.7940e-09, ..., 9.3132e-10, + 2.7940e-09, 1.3970e-08], + [ 0.0000e+00, -9.3132e-08, -2.9802e-07, ..., -4.2841e-08, + -9.2201e-08, -2.1514e-07], + [ 9.3132e-10, 1.0431e-07, 2.7940e-07, ..., 3.6322e-08, + 7.8231e-08, 1.8533e-07]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0316, 0.0029, 0.0211, 0.0122, 0.0392, 0.0164, -0.0042, 0.0054, + -0.0129, -0.0436], device='cuda:0'), grad: tensor([-1.9185e-07, 8.0094e-08, -4.2841e-08, -5.1223e-08, -6.6962e-07, + 8.8476e-08, 2.0023e-07, 4.9360e-08, -6.8266e-07, 1.2368e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 217.90, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4884 re_mapping 0.0042 re_causal 0.0122 /// teacc 99.04 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.1735, 0.2409, -0.0280, ..., -0.0230, 0.0704, -0.0347], + [ 0.1968, -0.1114, 0.0298, ..., -0.1083, 0.0171, -0.0954], + [-0.0898, -0.2898, -0.0514, ..., -0.0654, 0.0286, -0.2912], + ..., + [-0.1904, -0.3424, 0.0682, ..., -0.1157, -0.0864, -0.1905], + [-0.1462, 0.0060, 0.0405, ..., 0.0063, -0.1197, 0.0849], + [-0.1198, -0.0848, 0.1088, ..., 0.2236, -0.0824, -0.2088]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -8.0094e-08, 1.8626e-08, ..., 0.0000e+00, + 9.3132e-10, -9.3132e-10], + [-7.1898e-07, 7.4506e-09, -7.4785e-07, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 2.7940e-09, 5.5879e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + ..., + [ 7.0874e-07, 2.7940e-09, 8.0932e-07, ..., 9.3132e-10, + 9.3132e-10, 2.7940e-09], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 1.3970e-08, 2.7940e-09, -1.5832e-07, ..., -1.8626e-09, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0317, 0.0027, 0.0211, 0.0122, 0.0403, 0.0165, -0.0042, 0.0056, + -0.0127, -0.0449], device='cuda:0'), grad: tensor([-2.0489e-08, -1.0943e-06, 7.4506e-09, -9.3132e-10, 1.8626e-08, + -1.8626e-09, 1.4249e-07, 1.2172e-06, 2.7940e-09, -2.6356e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 217.91, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4653 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.05 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.1739, 0.2407, -0.0284, ..., -0.0254, 0.0708, -0.0351], + [ 0.1967, -0.1128, 0.0297, ..., -0.1085, 0.0162, -0.0960], + [-0.0899, -0.2903, -0.0514, ..., -0.0665, 0.0286, -0.2918], + ..., + [-0.1907, -0.3439, 0.0679, ..., -0.1158, -0.0884, -0.1910], + [-0.1493, 0.0032, 0.0389, ..., 0.0062, -0.1228, 0.0831], + [-0.1187, -0.0852, 0.1095, ..., 0.2241, -0.0786, -0.2088]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, -1.5646e-07, -7.2643e-08, ..., 0.0000e+00, + 0.0000e+00, -1.9558e-08], + [-1.4585e-06, 1.8626e-09, -9.9186e-07, ..., 0.0000e+00, + -1.1176e-08, 1.8626e-09], + [ 9.6112e-07, 1.8626e-09, 6.7148e-07, ..., 0.0000e+00, + 1.2107e-08, 5.5879e-09], + ..., + [ 6.5193e-09, 1.8626e-09, -1.9558e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 1.6764e-08, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-08], + [ 4.6566e-09, 3.0734e-08, 2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0310, 0.0026, 0.0216, 0.0123, 0.0404, 0.0167, -0.0041, 0.0051, + -0.0150, -0.0442], device='cuda:0'), grad: tensor([-1.9930e-07, -3.0473e-06, 2.1122e-06, -8.1956e-08, 2.7940e-09, + 1.0431e-07, 5.3924e-07, -1.8626e-08, 5.1036e-07, 8.6613e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 218.15, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4646 re_mapping 0.0042 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.1739, 0.2408, -0.0282, ..., -0.0254, 0.0706, -0.0354], + [ 0.1977, -0.1132, 0.0299, ..., -0.1088, 0.0173, -0.0964], + [-0.0913, -0.2908, -0.0522, ..., -0.0667, 0.0255, -0.2916], + ..., + [-0.1910, -0.3442, 0.0678, ..., -0.1162, -0.0893, -0.1911], + [-0.1494, 0.0035, 0.0394, ..., 0.0062, -0.1229, 0.0843], + [-0.1192, -0.0860, 0.1097, ..., 0.2241, -0.0776, -0.2098]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., 0.0000e+00, + 2.7940e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 5.5879e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 4.6566e-09, -7.6368e-08, -1.6112e-07, ..., -9.3132e-10, + -2.4308e-07, -2.6170e-07], + [ 4.6566e-09, 3.7253e-09, -4.6566e-09, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0308, 0.0030, 0.0212, 0.0124, 0.0404, 0.0166, -0.0040, 0.0049, + -0.0148, -0.0442], device='cuda:0'), grad: tensor([ 4.2841e-08, 1.3039e-08, -1.8999e-07, 3.0734e-08, 1.4622e-07, + 2.0023e-07, 5.3924e-07, 8.2888e-08, -8.6799e-07, 2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 218.60, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4860 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.03 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.1740, 0.2416, -0.0277, ..., -0.0255, 0.0721, -0.0356], + [ 0.1982, -0.1157, 0.0299, ..., -0.1088, 0.0173, -0.0970], + [-0.0915, -0.2943, -0.0525, ..., -0.0669, 0.0255, -0.2929], + ..., + [-0.1916, -0.3466, 0.0676, ..., -0.1163, -0.0896, -0.1923], + [-0.1496, 0.0037, 0.0403, ..., 0.0064, -0.1231, 0.0856], + [-0.1195, -0.0864, 0.1097, ..., 0.2242, -0.0776, -0.2104]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 2.7940e-09, 4.6566e-09], + [ 2.7940e-09, 9.3132e-09, 1.3225e-07, ..., 0.0000e+00, + -5.5879e-09, 3.0734e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 1.4994e-07, ..., 0.0000e+00, + 2.7940e-09, -2.3283e-08], + [-3.7253e-09, -1.2107e-08, -5.5879e-09, ..., 0.0000e+00, + -9.3132e-10, 2.5146e-08], + [ 9.3132e-10, 2.7940e-09, -2.9895e-07, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0312, 0.0028, 0.0211, 0.0125, 0.0404, 0.0166, -0.0041, 0.0046, + -0.0140, -0.0443], device='cuda:0'), grad: tensor([ 1.7229e-07, 9.0338e-08, 9.8720e-08, 2.7008e-08, 3.0734e-08, + 1.1520e-06, -1.5460e-06, 3.0268e-07, 1.3039e-07, -4.5169e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 218.06, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4559 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.02 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.1740, 0.2408, -0.0278, ..., -0.0245, 0.0720, -0.0357], + [ 0.1984, -0.1157, 0.0300, ..., -0.1087, 0.0174, -0.0971], + [-0.0915, -0.2968, -0.0527, ..., -0.0670, 0.0255, -0.2943], + ..., + [-0.1917, -0.3474, 0.0676, ..., -0.1165, -0.0898, -0.1925], + [-0.1500, 0.0036, 0.0403, ..., 0.0064, -0.1233, 0.0860], + [-0.1195, -0.0870, 0.1096, ..., 0.2241, -0.0776, -0.2106]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [-2.0396e-07, 0.0000e+00, -5.4482e-07, ..., 0.0000e+00, + -1.6764e-08, 6.2399e-08], + [ 3.3528e-08, 6.5193e-09, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-08], + ..., + [ 3.9041e-06, 0.0000e+00, 2.0843e-06, ..., 0.0000e+00, + 1.5832e-08, 1.4994e-07], + [ 2.2352e-08, -5.5879e-09, 6.5193e-09, ..., 9.3132e-10, + 0.0000e+00, 2.7008e-08], + [ 2.2724e-07, -4.6566e-09, -2.7940e-08, ..., -2.7940e-09, + 0.0000e+00, 1.5460e-07]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0305, 0.0029, 0.0210, 0.0125, 0.0406, 0.0166, -0.0033, 0.0044, + -0.0139, -0.0445], device='cuda:0'), grad: tensor([ 2.4214e-08, -1.7229e-07, 2.5798e-07, -1.5780e-05, 4.3772e-08, + 6.5472e-07, -2.7008e-08, 1.4283e-05, 1.4994e-07, 5.8673e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 218.15, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4692 re_mapping 0.0043 re_causal 0.0125 /// teacc 98.91 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.1743, 0.2409, -0.0279, ..., -0.0246, 0.0720, -0.0364], + [ 0.1987, -0.1158, 0.0301, ..., -0.1083, 0.0173, -0.0972], + [-0.0915, -0.2978, -0.0529, ..., -0.0666, 0.0255, -0.2951], + ..., + [-0.1921, -0.3479, 0.0675, ..., -0.1174, -0.0898, -0.1930], + [-0.1498, 0.0046, 0.0409, ..., 0.0066, -0.1233, 0.0876], + [-0.1194, -0.0872, 0.1098, ..., 0.2244, -0.0776, -0.2109]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.3956e-06, -1.6047e-06, ..., 0.0000e+00, + 0.0000e+00, -6.5193e-09], + [-7.4506e-09, 4.0978e-08, 3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 1.2107e-08], + [ 1.8626e-09, 1.2107e-08, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + ..., + [ 0.0000e+00, 1.9558e-08, 2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 2.5146e-08], + [ 9.3132e-10, -7.4506e-08, -1.4808e-07, ..., 0.0000e+00, + 0.0000e+00, -1.3784e-07], + [ 0.0000e+00, 8.6986e-07, 6.3237e-07, ..., 0.0000e+00, + 0.0000e+00, 3.8184e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0304, 0.0029, 0.0211, 0.0126, 0.0404, 0.0165, -0.0032, 0.0043, + -0.0131, -0.0443], device='cuda:0'), grad: tensor([-6.4857e-06, 8.6613e-08, -1.0245e-08, 6.7987e-08, 3.7253e-09, + 2.3097e-07, 4.6492e-06, 9.6858e-08, -4.0326e-07, 1.7816e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 218.06, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.5085 re_mapping 0.0044 re_causal 0.0135 /// teacc 98.99 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.1744, 0.2408, -0.0278, ..., -0.0247, 0.0720, -0.0364], + [ 0.1989, -0.1159, 0.0302, ..., -0.1087, 0.0174, -0.0974], + [-0.0917, -0.2986, -0.0536, ..., -0.0667, 0.0254, -0.2958], + ..., + [-0.1922, -0.3479, 0.0677, ..., -0.1170, -0.0899, -0.1932], + [-0.1498, 0.0049, 0.0414, ..., 0.0065, -0.1232, 0.0883], + [-0.1197, -0.0875, 0.1097, ..., 0.2244, -0.0775, -0.2111]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.0291e-08, -2.5146e-08, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [-5.5879e-09, 2.7940e-09, 1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, -4.6566e-08, ..., 0.0000e+00, + 0.0000e+00, 1.2107e-08], + ..., + [-2.1420e-08, 9.3132e-10, -6.1467e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 4.7497e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 1.8626e-08, 2.3283e-08, 5.5879e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0301, 0.0030, 0.0207, 0.0147, 0.0405, 0.0143, -0.0025, 0.0044, + -0.0128, -0.0445], device='cuda:0'), grad: tensor([-1.0990e-07, 6.6124e-08, -8.1863e-07, -2.4214e-08, -2.7940e-08, + 1.5832e-08, 2.7940e-08, -7.4506e-08, 8.1025e-07, 1.3877e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 218.16, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4922 re_mapping 0.0045 re_causal 0.0124 /// teacc 99.00 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.1744, 0.2412, -0.0279, ..., -0.0249, 0.0718, -0.0365], + [ 0.1995, -0.1163, 0.0303, ..., -0.1117, 0.0177, -0.0982], + [-0.0918, -0.3000, -0.0535, ..., -0.0676, 0.0255, -0.2965], + ..., + [-0.1928, -0.3484, 0.0676, ..., -0.1195, -0.0900, -0.1940], + [-0.1499, 0.0058, 0.0421, ..., 0.0066, -0.1233, 0.0905], + [-0.1207, -0.0882, 0.1100, ..., 0.2248, -0.0777, -0.2118]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.8626e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-5.8673e-08, 0.0000e+00, -4.2841e-08, ..., 0.0000e+00, + 3.7253e-09, 3.6322e-08], + [ 1.0245e-08, 9.3132e-10, 2.1420e-08, ..., 0.0000e+00, + 6.5193e-09, 1.5553e-07], + ..., + [ 3.1665e-08, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 2.7940e-09, 3.6322e-08], + [ 1.1176e-08, -4.6566e-09, -3.7253e-09, ..., 0.0000e+00, + 7.4506e-09, 6.9849e-08], + [ 4.6566e-09, 5.5879e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 362, bias, value: tensor([ 0.0297, 0.0030, 0.0208, 0.0148, 0.0408, 0.0141, -0.0014, 0.0038, + -0.0123, -0.0450], device='cuda:0'), grad: tensor([ 2.8871e-08, -1.9558e-08, 2.3562e-07, -8.0653e-07, 1.4901e-08, + 6.9849e-08, 2.4214e-08, 1.0617e-07, 1.8720e-07, 1.6205e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 218.12, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4881 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.07 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.1744, 0.2413, -0.0279, ..., -0.0250, 0.0719, -0.0368], + [ 0.1998, -0.1167, 0.0301, ..., -0.1130, 0.0176, -0.0987], + [-0.0917, -0.3003, -0.0548, ..., -0.0677, 0.0257, -0.2968], + ..., + [-0.1932, -0.3486, 0.0681, ..., -0.1196, -0.0901, -0.1941], + [-0.1500, 0.0058, 0.0422, ..., 0.0065, -0.1233, 0.0906], + [-0.1209, -0.0884, 0.1105, ..., 0.2256, -0.0778, -0.2125]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, 2.7008e-08, 1.4715e-07, ..., 0.0000e+00, + 6.6124e-08, 7.6368e-08], + [ 1.8924e-06, 1.2461e-06, 6.5491e-06, ..., 0.0000e+00, + 3.0641e-06, 3.5111e-06], + [ 7.3016e-07, 3.7160e-07, 1.9558e-06, ..., 0.0000e+00, + 9.2480e-07, 1.1064e-06], + ..., + [ 9.5926e-08, 3.6322e-08, -3.6787e-07, ..., 0.0000e+00, + 8.9407e-08, 1.1642e-07], + [-5.4538e-06, -2.7828e-06, -1.4618e-05, ..., 9.3132e-10, + -6.8434e-06, -7.8157e-06], + [ 5.4948e-08, 2.3283e-08, 5.2620e-07, ..., -9.3132e-10, + 6.0536e-08, 7.4506e-08]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0296, 0.0028, 0.0202, 0.0168, 0.0404, 0.0122, -0.0013, 0.0044, + -0.0123, -0.0449], device='cuda:0'), grad: tensor([ 6.0443e-07, 2.6539e-05, 8.2403e-06, 1.2636e-05, 8.9686e-07, + 7.5363e-06, 3.1944e-06, -1.7602e-07, -6.0707e-05, 1.3113e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 217.59, cls_loss 0.0009 cls_loss_mapping 0.0010 cls_loss_causal 0.4231 re_mapping 0.0042 re_causal 0.0120 /// teacc 99.06 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.1746, 0.2413, -0.0279, ..., -0.0250, 0.0718, -0.0370], + [ 0.1975, -0.1170, 0.0275, ..., -0.1135, 0.0176, -0.0995], + [-0.0917, -0.3010, -0.0549, ..., -0.0678, 0.0254, -0.2964], + ..., + [-0.1907, -0.3493, 0.0706, ..., -0.1203, -0.0902, -0.1944], + [-0.1501, 0.0059, 0.0430, ..., 0.0064, -0.1228, 0.0910], + [-0.1210, -0.0888, 0.1106, ..., 0.2258, -0.0779, -0.2129]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-08, 5.5879e-08, ..., 9.3132e-10, + 0.0000e+00, 7.0781e-08], + [-1.1083e-07, 3.7253e-09, -3.7253e-08, ..., 1.8626e-09, + -7.4506e-09, 1.8626e-09], + [ 0.0000e+00, 3.0734e-08, 1.2387e-07, ..., 9.3132e-10, + 9.3132e-10, 2.9802e-08], + ..., + [ 1.0710e-07, 3.7253e-09, -1.4231e-06, ..., -1.5087e-07, + 4.6566e-09, 3.7253e-09], + [ 9.3132e-10, -2.1234e-07, -1.2759e-07, ..., 5.5879e-09, + 0.0000e+00, -2.2165e-07], + [ 4.6566e-09, 6.1467e-08, 1.1986e-06, ..., 1.3690e-07, + 1.8626e-09, 5.0291e-08]], device='cuda:0') +Epoch 364, bias, value: tensor([ 0.0296, 0.0003, 0.0204, 0.0167, 0.0403, 0.0122, -0.0012, 0.0067, + -0.0117, -0.0448], device='cuda:0'), grad: tensor([ 1.7416e-07, -4.0047e-08, 2.8592e-07, 2.0210e-07, 1.0803e-07, + 5.4948e-08, 1.6112e-07, -2.6599e-06, -5.1875e-07, 2.2333e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 218.40, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4602 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.08 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.1751, 0.2415, -0.0280, ..., -0.0251, 0.0726, -0.0372], + [ 0.1984, -0.1171, 0.0278, ..., -0.1136, 0.0187, -0.0996], + [-0.0918, -0.3005, -0.0526, ..., -0.0678, 0.0248, -0.2942], + ..., + [-0.1915, -0.3496, 0.0705, ..., -0.1199, -0.0911, -0.1946], + [-0.1502, 0.0056, 0.0409, ..., 0.0063, -0.1228, 0.0902], + [-0.1218, -0.0896, 0.1117, ..., 0.2257, -0.0782, -0.2137]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -1.3039e-08, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 1.6764e-08, 1.9558e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.6764e-08], + [ 0.0000e+00, 1.8626e-09, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 365, bias, value: tensor([ 0.0296, 0.0006, 0.0228, 0.0168, 0.0383, 0.0121, -0.0010, 0.0067, + -0.0150, -0.0436], device='cuda:0'), grad: tensor([-5.5879e-09, 8.3819e-09, -5.5879e-09, -2.4214e-08, -3.5390e-08, + 2.1420e-08, -2.7008e-08, -1.5832e-08, 3.8184e-08, 5.4948e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 218.17, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4947 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.07 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.1753, 0.2417, -0.0278, ..., -0.0252, 0.0724, -0.0369], + [ 0.1986, -0.1172, 0.0278, ..., -0.1136, 0.0190, -0.0997], + [-0.0919, -0.3010, -0.0526, ..., -0.0680, 0.0245, -0.2945], + ..., + [-0.1916, -0.3502, 0.0704, ..., -0.1200, -0.0910, -0.1955], + [-0.1506, 0.0054, 0.0419, ..., 0.0060, -0.1230, 0.0904], + [-0.1222, -0.0898, 0.1118, ..., 0.2256, -0.0784, -0.2139]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 7.2271e-07, 1.1716e-06, ..., 0.0000e+00, + 0.0000e+00, 5.8301e-07], + [-2.8871e-08, 1.0245e-08, -2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 1.0245e-08, 5.4017e-08, 9.6858e-08, ..., 0.0000e+00, + 0.0000e+00, 4.4703e-08], + ..., + [ 2.8871e-08, 2.1420e-08, 4.4703e-08, ..., 0.0000e+00, + 0.0000e+00, 2.0489e-08], + [ 4.7497e-08, -1.5525e-06, -2.5816e-06, ..., 0.0000e+00, + 0.0000e+00, -1.2303e-06], + [ 2.1420e-08, 6.6496e-07, 1.0189e-06, ..., -3.7253e-09, + 0.0000e+00, 5.2713e-07]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0294, 0.0003, 0.0242, 0.0168, 0.0382, 0.0122, -0.0008, 0.0065, + -0.0147, -0.0437], device='cuda:0'), grad: tensor([ 3.3155e-06, -3.2596e-08, 2.6822e-07, 1.0803e-06, -2.4680e-07, + -9.7696e-07, 4.6566e-07, 1.3039e-07, -7.2122e-06, 3.1628e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 218.35, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4726 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.07 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.1754, 0.2429, -0.0274, ..., -0.0251, 0.0721, -0.0392], + [ 0.1987, -0.1187, 0.0278, ..., -0.1139, 0.0191, -0.1000], + [-0.0919, -0.3043, -0.0527, ..., -0.0681, 0.0246, -0.2948], + ..., + [-0.1917, -0.3505, 0.0704, ..., -0.1204, -0.0911, -0.1957], + [-0.1505, 0.0055, 0.0423, ..., 0.0060, -0.1229, 0.0907], + [-0.1224, -0.0900, 0.1119, ..., 0.2257, -0.0781, -0.2143]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-7.0222e-07, 0.0000e+00, -7.0035e-07, ..., -5.1223e-08, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 9.3132e-10, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 6.7428e-07, 0.0000e+00, -1.8068e-07, ..., 4.9360e-08, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, -3.7253e-09, 2.7940e-09, ..., 9.3132e-10, + 0.0000e+00, -5.5879e-09], + [ 8.3819e-09, 9.3132e-10, 7.4226e-07, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 367, bias, value: tensor([ 3.0142e-02, 1.2320e-05, 2.3287e-02, 1.6848e-02, 3.7984e-02, + 1.2152e-02, -9.1098e-04, 6.3871e-03, -1.4432e-02, -4.3572e-02], + device='cuda:0'), grad: tensor([ 1.2107e-08, -9.1828e-07, -1.5367e-07, 2.9616e-07, 7.4506e-09, + 8.3819e-09, 2.2352e-08, -1.0068e-06, 4.9360e-08, 1.6764e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 218.05, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4800 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.03 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.1753, 0.2439, -0.0271, ..., -0.0252, 0.0720, -0.0392], + [ 0.1984, -0.1190, 0.0277, ..., -0.1142, 0.0189, -0.1006], + [-0.0912, -0.3045, -0.0520, ..., -0.0681, 0.0248, -0.2949], + ..., + [-0.1917, -0.3507, 0.0704, ..., -0.1205, -0.0911, -0.1958], + [-0.1515, 0.0050, 0.0423, ..., 0.0058, -0.1233, 0.0906], + [-0.1224, -0.0902, 0.1120, ..., 0.2258, -0.0781, -0.2144]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.3027e-07, 0.0000e+00, -2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.0291e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.7940e-07, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.4901e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0307, -0.0003, 0.0240, 0.0168, 0.0379, 0.0120, -0.0007, 0.0064, + -0.0150, -0.0435], device='cuda:0'), grad: tensor([ 1.8626e-09, -1.0245e-06, 1.2293e-07, 1.5087e-07, 5.2154e-08, + 3.9116e-08, 1.8626e-09, 6.6683e-07, 4.0978e-08, -5.9605e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 218.31, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4712 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.1753, 0.2443, -0.0263, ..., -0.0238, 0.0720, -0.0399], + [ 0.2011, -0.1188, 0.0299, ..., -0.1143, 0.0196, -0.1002], + [-0.0913, -0.3046, -0.0522, ..., -0.0681, 0.0248, -0.2953], + ..., + [-0.1943, -0.3525, 0.0678, ..., -0.1208, -0.0912, -0.1965], + [-0.1516, 0.0056, 0.0438, ..., 0.0058, -0.1233, 0.0911], + [-0.1239, -0.0923, 0.1114, ..., 0.2257, -0.0792, -0.2170]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 5.5879e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [-3.3528e-08, 0.0000e+00, -3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 5.5879e-09, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-08], + ..., + [ 1.4901e-08, 0.0000e+00, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 5.9605e-08, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.3469e-07], + [ 1.1176e-08, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-08]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0310, 0.0020, 0.0239, 0.0179, 0.0380, 0.0120, -0.0006, 0.0028, + -0.0146, -0.0445], device='cuda:0'), grad: tensor([ 6.5193e-08, -4.4703e-08, 8.5682e-08, -6.8732e-07, 1.6764e-08, + 1.7695e-07, -1.0431e-07, 2.7940e-08, 3.7625e-07, 7.2643e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 218.67, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.5034 re_mapping 0.0043 re_causal 0.0128 /// teacc 99.08 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.1755, 0.2447, -0.0260, ..., -0.0237, 0.0717, -0.0400], + [ 0.2013, -0.1207, 0.0299, ..., -0.1152, 0.0190, -0.1008], + [-0.0914, -0.3049, -0.0527, ..., -0.0682, 0.0249, -0.2958], + ..., + [-0.1946, -0.3529, 0.0677, ..., -0.1215, -0.0913, -0.1968], + [-0.1519, 0.0055, 0.0445, ..., 0.0057, -0.1234, 0.0910], + [-0.1246, -0.0928, 0.1119, ..., 0.2256, -0.0790, -0.2174]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -6.8918e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [ 0.0000e+00, 1.8626e-09, 3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0312, 0.0019, 0.0238, 0.0179, 0.0377, 0.0121, -0.0005, 0.0027, + -0.0144, -0.0443], device='cuda:0'), grad: tensor([-7.4506e-09, 4.2841e-08, 9.3132e-09, 4.0978e-08, -6.3665e-06, + 1.4901e-08, 3.3528e-08, -1.2480e-07, -2.0489e-08, 6.3926e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 218.18, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4804 re_mapping 0.0042 re_causal 0.0120 /// teacc 99.00 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.1756, 0.2449, -0.0265, ..., -0.0237, 0.0715, -0.0401], + [ 0.2015, -0.1218, 0.0300, ..., -0.1152, 0.0192, -0.1013], + [-0.0914, -0.3052, -0.0528, ..., -0.0681, 0.0246, -0.2960], + ..., + [-0.1948, -0.3535, 0.0679, ..., -0.1214, -0.0913, -0.1970], + [-0.1524, 0.0053, 0.0448, ..., 0.0057, -0.1236, 0.0907], + [-0.1256, -0.0935, 0.1119, ..., 0.2258, -0.0797, -0.2180]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -8.5682e-08, -5.5879e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.4214e-08, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-09, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.1176e-08, 3.7253e-09, -3.7458e-06, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 3.7253e-09, 5.9605e-08, 3.7290e-06, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0309, 0.0019, 0.0237, 0.0179, 0.0374, 0.0122, -0.0003, 0.0027, + -0.0146, -0.0443], device='cuda:0'), grad: tensor([-1.9558e-07, -5.5879e-09, 1.8626e-08, 6.3330e-08, -9.3132e-09, + -5.4017e-08, 5.4017e-08, -6.0424e-06, 3.1665e-08, 6.1169e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 218.23, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4839 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.00 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.1762, 0.2452, -0.0256, ..., -0.0240, 0.0714, -0.0409], + [ 0.2018, -0.1219, 0.0300, ..., -0.1155, 0.0199, -0.1015], + [-0.0923, -0.3053, -0.0537, ..., -0.0687, 0.0230, -0.2963], + ..., + [-0.1949, -0.3551, 0.0680, ..., -0.1219, -0.0915, -0.1974], + [-0.1540, 0.0025, 0.0449, ..., 0.0057, -0.1236, 0.0871], + [-0.1265, -0.0956, 0.1113, ..., 0.2260, -0.0800, -0.2188]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.1665e-08, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 1.8626e-09, 3.7253e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.8626e-09, 1.8626e-09, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 7.4506e-09, -9.6858e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 372, bias, value: tensor([ 0.0312, 0.0021, 0.0229, 0.0154, 0.0374, 0.0148, -0.0005, 0.0028, + -0.0166, -0.0449], device='cuda:0'), grad: tensor([ 1.8626e-08, 3.7253e-09, 2.7940e-08, -8.5682e-08, 6.8918e-08, + 7.4506e-08, 3.3528e-08, 2.7940e-08, 2.4214e-08, -2.0489e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 217.72, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4601 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.06 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.1764, 0.2457, -0.0255, ..., -0.0241, 0.0713, -0.0409], + [ 0.2022, -0.1220, 0.0289, ..., -0.1155, 0.0201, -0.1016], + [-0.0925, -0.3054, -0.0536, ..., -0.0687, 0.0228, -0.2966], + ..., + [-0.1952, -0.3554, 0.0694, ..., -0.1220, -0.0915, -0.1975], + [-0.1541, 0.0026, 0.0453, ..., 0.0058, -0.1236, 0.0871], + [-0.1279, -0.0966, 0.1104, ..., 0.2260, -0.0801, -0.2194]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0314, 0.0013, 0.0233, 0.0153, 0.0375, 0.0149, -0.0007, 0.0037, + -0.0165, -0.0458], device='cuda:0'), grad: tensor([-3.7253e-09, 0.0000e+00, -1.8626e-08, -1.6764e-08, 0.0000e+00, + -7.4506e-09, 1.8626e-09, 2.0489e-08, 1.4901e-08, 5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 218.06, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4719 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.00 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.1766, 0.2461, -0.0264, ..., -0.0240, 0.0709, -0.0411], + [ 0.2025, -0.1233, 0.0291, ..., -0.1158, 0.0201, -0.1020], + [-0.0931, -0.3051, -0.0550, ..., -0.0689, 0.0227, -0.2975], + ..., + [-0.1953, -0.3558, 0.0694, ..., -0.1220, -0.0916, -0.1977], + [-0.1547, 0.0021, 0.0455, ..., 0.0054, -0.1239, 0.0872], + [-0.1292, -0.0962, 0.1106, ..., 0.2260, -0.0815, -0.2202]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5751e-07, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 2.7940e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 1.8626e-09, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0313, 0.0014, 0.0232, 0.0152, 0.0379, 0.0151, -0.0010, 0.0037, + -0.0168, -0.0461], device='cuda:0'), grad: tensor([ 2.0843e-06, 9.3132e-09, -6.9477e-07, 3.6508e-07, -3.1665e-08, + -3.5390e-08, -2.0098e-06, 2.0862e-07, 1.1176e-08, 9.6858e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 218.06, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4667 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.05 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.1768, 0.2464, -0.0264, ..., -0.0240, 0.0705, -0.0411], + [ 0.2025, -0.1234, 0.0291, ..., -0.1165, 0.0201, -0.1021], + [-0.0932, -0.3053, -0.0551, ..., -0.0691, 0.0250, -0.2981], + ..., + [-0.1953, -0.3554, 0.0694, ..., -0.1222, -0.0941, -0.1979], + [-0.1548, 0.0022, 0.0459, ..., 0.0054, -0.1237, 0.0873], + [-0.1294, -0.0966, 0.1107, ..., 0.2261, -0.0834, -0.2207]], + device='cuda:0'), grad: tensor([[-1.4529e-07, -1.1977e-06, -1.6764e-08, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [-1.3039e-08, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 3.7253e-09, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 7.4506e-09, 2.0489e-08, 1.1176e-08, ..., 1.8626e-09, + 0.0000e+00, 5.5879e-09], + [ 1.8626e-09, -8.5682e-08, -7.6368e-08, ..., -2.0489e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0313, 0.0014, 0.0238, 0.0152, 0.0384, 0.0151, -0.0009, 0.0035, + -0.0165, -0.0466], device='cuda:0'), grad: tensor([-2.0787e-06, -1.8626e-09, 7.4506e-09, 1.8626e-08, 3.0547e-07, + 5.2154e-07, 1.5758e-06, -1.7509e-07, 5.2154e-08, -2.1048e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 218.41, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4892 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.00 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.1769, 0.2471, -0.0260, ..., -0.0241, 0.0702, -0.0412], + [ 0.2027, -0.1234, 0.0291, ..., -0.1180, 0.0203, -0.1023], + [-0.0932, -0.3054, -0.0555, ..., -0.0694, 0.0254, -0.2985], + ..., + [-0.1955, -0.3561, 0.0694, ..., -0.1236, -0.0945, -0.1982], + [-0.1550, 0.0022, 0.0463, ..., 0.0055, -0.1237, 0.0875], + [-0.1291, -0.0980, 0.1106, ..., 0.2266, -0.0836, -0.2213]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 3.7253e-08, 6.7055e-08, ..., 0.0000e+00, + 9.3132e-09, 3.7253e-08], + [-2.9057e-05, -2.9430e-07, -4.1008e-05, ..., 0.0000e+00, + -3.6880e-07, 1.8626e-09], + [ 4.0978e-08, 9.3132e-09, 6.7055e-08, ..., 0.0000e+00, + 1.8626e-09, 1.3039e-08], + ..., + [ 2.8580e-05, 1.2666e-07, 4.0025e-05, ..., 0.0000e+00, + 1.5646e-07, 7.4506e-09], + [ 3.1106e-07, 9.6858e-08, 6.6310e-07, ..., 0.0000e+00, + 1.7136e-07, -4.8429e-08], + [ 4.0978e-08, 1.8626e-08, 7.8231e-08, ..., 0.0000e+00, + 1.6764e-08, 1.3039e-08]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0318, 0.0014, 0.0239, 0.0152, 0.0384, 0.0151, -0.0012, 0.0035, + -0.0164, -0.0468], device='cuda:0'), grad: tensor([ 2.0117e-07, -5.8115e-05, 1.3597e-07, -5.7742e-08, 7.0781e-08, + -1.1362e-07, 8.7544e-08, 5.6595e-05, 9.5181e-07, 1.6950e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 217.86, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4733 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.1770, 0.2466, -0.0261, ..., -0.0249, 0.0701, -0.0416], + [ 0.2029, -0.1235, 0.0291, ..., -0.1180, 0.0203, -0.1026], + [-0.0943, -0.3055, -0.0572, ..., -0.0695, 0.0254, -0.2993], + ..., + [-0.1956, -0.3571, 0.0695, ..., -0.1237, -0.0946, -0.1993], + [-0.1556, 0.0021, 0.0473, ..., 0.0058, -0.1238, 0.0876], + [-0.1297, -0.0992, 0.1107, ..., 0.2267, -0.0835, -0.2227]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-1.8626e-09, 1.8626e-09, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., 0.0000e+00, + -3.7253e-09, -1.4901e-08], + ..., + [ 9.3132e-09, 7.4506e-09, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [ 2.2352e-08, 1.8626e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + [ 1.3039e-08, 1.1176e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0303, 0.0015, 0.0232, 0.0152, 0.0383, 0.0151, 0.0003, 0.0036, + -0.0163, -0.0471], device='cuda:0'), grad: tensor([ 4.4703e-08, 4.4703e-08, -7.1526e-07, 2.1420e-07, 3.7253e-09, + -2.6263e-07, 4.0978e-08, 4.3213e-07, 1.5087e-07, 3.3528e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 218.09, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4544 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.00 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.1771, 0.2469, -0.0261, ..., -0.0250, 0.0700, -0.0419], + [ 0.2026, -0.1235, 0.0285, ..., -0.1197, 0.0204, -0.1028], + [-0.0944, -0.3056, -0.0577, ..., -0.0697, 0.0255, -0.2996], + ..., + [-0.1952, -0.3574, 0.0702, ..., -0.1246, -0.0949, -0.1998], + [-0.1557, 0.0024, 0.0480, ..., 0.0056, -0.1200, 0.0908], + [-0.1300, -0.0999, 0.1107, ..., 0.2271, -0.0835, -0.2237]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.2352e-08, -3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [-3.7253e-09, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 7.4506e-09, 1.4901e-08], + [ 1.6764e-08, 5.5879e-09, 1.8626e-09, ..., 0.0000e+00, + 2.6077e-08, 4.8429e-08], + ..., + [ 3.1665e-08, 0.0000e+00, 1.5646e-07, ..., 0.0000e+00, + 4.0978e-08, 2.4587e-07], + [ 9.3132e-09, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09], + [-1.8626e-08, 5.5879e-09, -1.7136e-07, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09]], device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0305, 0.0008, 0.0233, 0.0152, 0.0385, 0.0151, -0.0007, 0.0043, + -0.0133, -0.0474], device='cuda:0'), grad: tensor([-1.1176e-08, 5.5879e-08, 2.4959e-07, -1.4603e-06, -5.0291e-08, + 4.7497e-07, 2.0489e-08, 9.6671e-07, 4.6566e-08, -2.9057e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 218.68, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4926 re_mapping 0.0041 re_causal 0.0125 /// teacc 98.98 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.1771, 0.2472, -0.0260, ..., -0.0250, 0.0695, -0.0419], + [ 0.2026, -0.1235, 0.0284, ..., -0.1199, 0.0207, -0.1029], + [-0.0946, -0.3057, -0.0579, ..., -0.0697, 0.0257, -0.3000], + ..., + [-0.1952, -0.3576, 0.0704, ..., -0.1247, -0.0955, -0.2004], + [-0.1558, 0.0024, 0.0481, ..., 0.0056, -0.1200, 0.0908], + [-0.1302, -0.1002, 0.1106, ..., 0.2271, -0.0836, -0.2241]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, -6.7241e-07, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, -1.0245e-07], + [-1.6578e-07, 0.0000e+00, -1.2666e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.4017e-08, 6.8918e-08, 5.2154e-08, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + ..., + [ 2.6077e-08, 1.8626e-09, -1.9282e-05, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-09, 3.9116e-08, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 2.1048e-07, 1.7598e-05, ..., 0.0000e+00, + 0.0000e+00, 3.3528e-08]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0306, 0.0007, 0.0233, 0.0152, 0.0386, 0.0151, -0.0007, 0.0044, + -0.0134, -0.0477], device='cuda:0'), grad: tensor([-1.2740e-06, -3.9116e-07, 2.6822e-07, 2.9299e-06, 2.2352e-07, + 4.0792e-07, 6.7614e-07, -3.9279e-05, 1.1921e-07, 3.6299e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 217.91, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4878 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.05 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.1772, 0.2470, -0.0260, ..., -0.0251, 0.0688, -0.0421], + [ 0.2041, -0.1235, 0.0283, ..., -0.1209, 0.0206, -0.1030], + [-0.0954, -0.3058, -0.0584, ..., -0.0699, 0.0256, -0.3003], + ..., + [-0.1966, -0.3578, 0.0705, ..., -0.1261, -0.0955, -0.2004], + [-0.1560, 0.0024, 0.0481, ..., 0.0054, -0.1200, 0.0909], + [-0.1299, -0.1005, 0.1106, ..., 0.2275, -0.0833, -0.2242]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.9116e-08, 2.0117e-07, ..., 5.7742e-08, + 9.3132e-09, 3.9116e-08], + [-1.5646e-07, 0.0000e+00, -1.1176e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.0978e-08, 0.0000e+00, 3.7253e-08, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 7.8231e-08, 1.8626e-09, 6.5193e-08, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 1.0617e-07, 4.0978e-08, 1.9185e-07, ..., 3.9116e-08, + 1.8626e-09, 8.7544e-08], + [ 1.3039e-08, 1.8626e-08, -4.4145e-07, ..., -1.0990e-07, + -7.4506e-09, -5.7742e-08]], device='cuda:0') +Epoch 380, bias, value: tensor([ 2.9929e-02, 1.6560e-03, 2.2982e-02, 1.5181e-02, 3.8590e-02, + 1.5045e-02, 5.3536e-05, 3.7681e-03, -1.3450e-02, -4.7873e-02], + device='cuda:0'), grad: tensor([ 6.1654e-07, -2.7195e-07, 1.2293e-07, 2.2352e-08, -3.7253e-08, + 9.1270e-08, -2.7008e-07, 1.7509e-07, 6.9290e-07, -1.1381e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 218.32, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4541 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.1772, 0.2472, -0.0260, ..., -0.0251, 0.0686, -0.0424], + [ 0.2050, -0.1221, 0.0287, ..., -0.1204, 0.0239, -0.1001], + [-0.0955, -0.3059, -0.0600, ..., -0.0700, 0.0259, -0.3006], + ..., + [-0.1969, -0.3584, 0.0707, ..., -0.1266, -0.0961, -0.2006], + [-0.1565, 0.0024, 0.0479, ..., 0.0050, -0.1201, 0.0906], + [-0.1309, -0.1009, 0.1105, ..., 0.2274, -0.0844, -0.2248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -3.1292e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.8871e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 1.8626e-09, 2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 1.8626e-09, -6.8918e-08, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0299, 0.0023, 0.0224, 0.0151, 0.0388, 0.0150, -0.0007, 0.0040, + -0.0139, -0.0484], device='cuda:0'), grad: tensor([ 1.8440e-07, 4.8429e-08, -1.2815e-06, -6.8918e-08, 5.4017e-08, + -1.3225e-07, 2.0489e-08, 1.2200e-06, 4.2841e-08, -8.5682e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 217.82, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.5104 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.00 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.1778, 0.2476, -0.0262, ..., -0.0251, 0.0670, -0.0426], + [ 0.2051, -0.1220, 0.0287, ..., -0.1205, 0.0240, -0.1001], + [-0.0955, -0.3063, -0.0595, ..., -0.0700, 0.0258, -0.3017], + ..., + [-0.1969, -0.3589, 0.0708, ..., -0.1267, -0.0964, -0.2009], + [-0.1566, 0.0031, 0.0475, ..., 0.0049, -0.1201, 0.0910], + [-0.1310, -0.1012, 0.1104, ..., 0.2274, -0.0844, -0.2249]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 5.5879e-09, 1.8626e-09], + [-2.5891e-07, 0.0000e+00, -2.6263e-07, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 4.6194e-07, 0.0000e+00, 2.8126e-07, ..., 0.0000e+00, + 1.1735e-07, 1.8626e-09], + [ 2.4214e-08, 1.8626e-09, 4.4703e-08, ..., 0.0000e+00, + 1.4901e-08, 3.7253e-09], + [ 1.8626e-09, 0.0000e+00, -9.8720e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0301, 0.0023, 0.0228, 0.0152, 0.0386, 0.0150, -0.0010, 0.0041, + -0.0138, -0.0485], device='cuda:0'), grad: tensor([ 6.8918e-08, 9.5181e-07, 9.3132e-09, 1.8626e-09, 7.8045e-07, + 2.2724e-07, -2.0955e-06, -1.8626e-08, 2.4959e-07, -1.8999e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 218.19, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4624 re_mapping 0.0040 re_causal 0.0115 /// teacc 99.06 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.1773, 0.2484, -0.0261, ..., -0.0254, 0.0661, -0.0428], + [ 0.2058, -0.1217, 0.0289, ..., -0.1205, 0.0250, -0.0990], + [-0.0956, -0.3072, -0.0597, ..., -0.0730, 0.0259, -0.3036], + ..., + [-0.1972, -0.3595, 0.0707, ..., -0.1268, -0.0970, -0.2012], + [-0.1562, 0.0041, 0.0480, ..., 0.0047, -0.1197, 0.0929], + [-0.1316, -0.1018, 0.1109, ..., 0.2277, -0.0841, -0.2252]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, -7.4506e-09, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 5.5879e-09, 3.7253e-09, 5.7742e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-2.4214e-08, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [ 1.3039e-08, 5.5879e-09, -1.2852e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 6.8918e-08, 6.1467e-08, 2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 4.0978e-08], + [ 5.5879e-09, 9.3132e-09, 3.5390e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0304, 0.0027, 0.0228, 0.0152, 0.0383, 0.0149, -0.0015, 0.0039, + -0.0126, -0.0481], device='cuda:0'), grad: tensor([ 2.1234e-07, 1.8813e-07, -7.2271e-07, 6.3330e-08, 1.4901e-08, + -1.5087e-07, 2.6077e-08, 1.4901e-07, 1.3225e-07, 9.1270e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 382---------------------------------------------------- +epoch 382, time 218.58, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4679 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.19 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.1772, 0.2496, -0.0280, ..., -0.0254, 0.0660, -0.0436], + [ 0.2060, -0.1214, 0.0290, ..., -0.1205, 0.0252, -0.0990], + [-0.0960, -0.3077, -0.0597, ..., -0.0733, 0.0265, -0.3041], + ..., + [-0.1973, -0.3610, 0.0708, ..., -0.1268, -0.0979, -0.2020], + [-0.1568, 0.0050, 0.0494, ..., 0.0047, -0.1199, 0.0940], + [-0.1323, -0.1028, 0.1113, ..., 0.2277, -0.0842, -0.2269]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, -2.0489e-08, -2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [-7.4506e-09, 1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + ..., + [ 3.7253e-09, 1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 3.7253e-08, 2.2352e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-08], + [ 1.8626e-09, 1.8626e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0304, 0.0028, 0.0232, 0.0152, 0.0384, 0.0149, -0.0023, 0.0039, + -0.0118, -0.0480], device='cuda:0'), grad: tensor([ 4.8429e-08, 5.5879e-09, 2.9802e-08, -2.2911e-07, 1.6764e-08, + 1.4082e-06, -1.5721e-06, 2.4214e-08, 2.3842e-07, 2.7940e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 218.13, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4896 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.1772, 0.2509, -0.0277, ..., -0.0254, 0.0660, -0.0431], + [ 0.2062, -0.1210, 0.0291, ..., -0.1206, 0.0253, -0.0991], + [-0.0961, -0.3083, -0.0599, ..., -0.0733, 0.0265, -0.3044], + ..., + [-0.1974, -0.3612, 0.0708, ..., -0.1269, -0.0981, -0.2024], + [-0.1577, 0.0039, 0.0490, ..., 0.0046, -0.1200, 0.0938], + [-0.1329, -0.1039, 0.1111, ..., 0.2277, -0.0842, -0.2275]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.0489e-08], + [ 7.6368e-08, 6.7055e-08, -1.6764e-08, ..., 0.0000e+00, + 9.3132e-09, 3.7253e-08], + [ 9.3132e-09, 1.8626e-08, -3.7253e-09, ..., 0.0000e+00, + 3.7253e-09, -2.0489e-08]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0310, 0.0029, 0.0231, 0.0152, 0.0384, 0.0148, -0.0024, 0.0039, + -0.0126, -0.0484], device='cuda:0'), grad: tensor([ 1.3039e-08, 1.8626e-08, -3.7253e-09, 0.0000e+00, 4.6566e-08, + -2.0117e-07, -7.8231e-08, 7.0781e-08, 1.5087e-07, -9.3132e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 217.95, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4633 re_mapping 0.0043 re_causal 0.0123 /// teacc 98.94 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.1772, 0.2522, -0.0267, ..., -0.0254, 0.0659, -0.0428], + [ 0.2062, -0.1210, 0.0289, ..., -0.1208, 0.0254, -0.0991], + [-0.0965, -0.3088, -0.0602, ..., -0.0733, 0.0266, -0.3076], + ..., + [-0.1973, -0.3621, 0.0710, ..., -0.1270, -0.0982, -0.2027], + [-0.1580, 0.0038, 0.0493, ..., 0.0047, -0.1200, 0.0938], + [-0.1337, -0.1066, 0.1110, ..., 0.2278, -0.0848, -0.2289]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 4.2841e-08, 0.0000e+00], + [ 3.7253e-08, 0.0000e+00, 8.3819e-08, ..., 0.0000e+00, + 3.5577e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + -2.1048e-07, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 8.1956e-08, 3.7253e-09], + [-8.9407e-08, 3.7253e-09, -1.8999e-07, ..., 0.0000e+00, + -5.4017e-08, 1.8626e-09]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0318, 0.0027, 0.0222, 0.0153, 0.0386, 0.0148, -0.0029, 0.0040, + -0.0126, -0.0489], device='cuda:0'), grad: tensor([ 2.9430e-07, 2.4457e-06, -3.6322e-07, 1.7136e-06, 5.2899e-07, + 6.2026e-07, -3.7234e-06, -1.5311e-06, 5.3830e-07, -5.1223e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 218.09, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4812 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.06 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.1775, 0.2535, -0.0264, ..., -0.0254, 0.0649, -0.0431], + [ 0.2066, -0.1208, 0.0290, ..., -0.1209, 0.0260, -0.0986], + [-0.0968, -0.3090, -0.0605, ..., -0.0733, 0.0261, -0.3081], + ..., + [-0.1975, -0.3623, 0.0712, ..., -0.1270, -0.0983, -0.2032], + [-0.1601, 0.0025, 0.0494, ..., 0.0047, -0.1202, 0.0927], + [-0.1355, -0.1083, 0.1108, ..., 0.2278, -0.0853, -0.2306]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.6764e-08, 3.7253e-09, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 2.7940e-08, 3.3528e-08, 5.0291e-08, ..., 0.0000e+00, + 5.5879e-09, 3.3528e-08], + ..., + [ 1.8626e-08, 0.0000e+00, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-1.3039e-08, -3.1665e-08, -5.7742e-08, ..., 0.0000e+00, + -3.7253e-09, -3.3528e-08], + [ 3.7253e-09, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0326, 0.0029, 0.0225, 0.0151, 0.0382, 0.0151, -0.0043, 0.0042, + -0.0132, -0.0492], device='cuda:0'), grad: tensor([ 9.3132e-09, -2.4214e-08, 1.2107e-07, -2.4214e-08, -2.6077e-08, + 1.8626e-08, -3.1665e-08, 9.3132e-09, -1.0803e-07, 3.7253e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 217.77, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4690 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.01 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.1777, 0.2540, -0.0260, ..., -0.0254, 0.0656, -0.0433], + [ 0.2068, -0.1213, 0.0291, ..., -0.1209, 0.0260, -0.0988], + [-0.0969, -0.3091, -0.0608, ..., -0.0733, 0.0260, -0.3084], + ..., + [-0.1976, -0.3643, 0.0713, ..., -0.1272, -0.0985, -0.2035], + [-0.1606, 0.0024, 0.0494, ..., 0.0047, -0.1202, 0.0926], + [-0.1368, -0.1099, 0.1103, ..., 0.2279, -0.0855, -0.2315]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0431e-07, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09], + [-5.5879e-09, 1.8626e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.5879e-09, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0324, 0.0030, 0.0226, 0.0162, 0.0381, 0.0140, -0.0038, 0.0043, + -0.0134, -0.0501], device='cuda:0'), grad: tensor([-1.3970e-07, -5.5879e-09, 5.5879e-09, 1.8626e-08, -9.3132e-08, + 9.3132e-08, -9.3132e-09, 1.6764e-08, 2.4214e-08, 9.3132e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 217.63, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4618 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.04 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.1775, 0.2549, -0.0259, ..., -0.0254, 0.0665, -0.0427], + [ 0.2070, -0.1223, 0.0292, ..., -0.1232, 0.0260, -0.0989], + [-0.0970, -0.3093, -0.0611, ..., -0.0733, 0.0260, -0.3085], + ..., + [-0.1977, -0.3646, 0.0713, ..., -0.1273, -0.0985, -0.2038], + [-0.1613, 0.0022, 0.0492, ..., 0.0046, -0.1203, 0.0922], + [-0.1383, -0.1111, 0.1100, ..., 0.2282, -0.0855, -0.2328]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.0803e-07, -2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 1.3039e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + ..., + [ 1.1176e-08, 0.0000e+00, 5.5879e-08, ..., 3.7253e-09, + 0.0000e+00, 9.3132e-09], + [ 0.0000e+00, 8.1956e-08, 2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08], + [-2.0489e-08, 5.5879e-09, -9.4995e-08, ..., 1.8999e-07, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0329, 0.0029, 0.0226, 0.0159, 0.0386, 0.0143, -0.0036, 0.0044, + -0.0142, -0.0510], device='cuda:0'), grad: tensor([ 3.4831e-07, 4.2841e-08, 9.6858e-08, -1.8813e-07, -2.1569e-06, + 1.5087e-07, -5.8860e-07, 2.6263e-07, 2.0117e-07, 1.8124e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 218.04, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4735 re_mapping 0.0040 re_causal 0.0116 /// teacc 98.99 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.1775, 0.2549, -0.0260, ..., -0.0255, 0.0661, -0.0429], + [ 0.2074, -0.1224, 0.0293, ..., -0.1235, 0.0254, -0.0990], + [-0.0982, -0.3094, -0.0616, ..., -0.0733, 0.0256, -0.3088], + ..., + [-0.1979, -0.3657, 0.0715, ..., -0.1275, -0.0986, -0.2046], + [-0.1622, 0.0020, 0.0489, ..., 0.0041, -0.1204, 0.0921], + [-0.1397, -0.1113, 0.1096, ..., 0.2284, -0.0856, -0.2335]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -1.4901e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 1.8626e-09, 3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0328, 0.0029, 0.0222, 0.0157, 0.0386, 0.0145, -0.0031, 0.0045, + -0.0147, -0.0514], device='cuda:0'), grad: tensor([-1.8626e-09, 1.3970e-07, -4.6566e-08, 6.7055e-08, -4.2096e-07, + -3.6508e-07, 3.4273e-07, 1.0803e-07, 1.8626e-08, 1.5087e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 218.17, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4704 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.1777, 0.2550, -0.0273, ..., -0.0255, 0.0656, -0.0439], + [ 0.2075, -0.1225, 0.0293, ..., -0.1240, 0.0255, -0.0990], + [-0.0986, -0.3095, -0.0621, ..., -0.0733, 0.0250, -0.3092], + ..., + [-0.1979, -0.3660, 0.0715, ..., -0.1278, -0.0987, -0.2053], + [-0.1626, 0.0019, 0.0490, ..., 0.0039, -0.1205, 0.0917], + [-0.1396, -0.1113, 0.1103, ..., 0.2285, -0.0855, -0.2336]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3039e-08, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, -5.9605e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0323, 0.0029, 0.0219, 0.0157, 0.0389, 0.0145, -0.0032, 0.0046, + -0.0151, -0.0511], device='cuda:0'), grad: tensor([ 1.3039e-08, 2.9802e-08, -6.5193e-08, 3.7253e-09, 2.0489e-08, + -7.2643e-08, 3.1665e-08, 7.0781e-08, 1.6764e-08, -3.7253e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 218.43, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4772 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.09 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.1777, 0.2556, -0.0273, ..., -0.0255, 0.0654, -0.0431], + [ 0.2077, -0.1225, 0.0276, ..., -0.1243, 0.0255, -0.0991], + [-0.1015, -0.3098, -0.0625, ..., -0.0734, 0.0250, -0.3095], + ..., + [-0.1972, -0.3664, 0.0733, ..., -0.1284, -0.0988, -0.2059], + [-0.1629, 0.0021, 0.0488, ..., 0.0039, -0.1205, 0.0919], + [-0.1399, -0.1115, 0.1103, ..., 0.2288, -0.0854, -0.2333]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.7556e-07, -7.9349e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.6624e-07, 7.8790e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0328, 0.0018, 0.0202, 0.0159, 0.0389, 0.0143, -0.0029, 0.0059, + -0.0155, -0.0513], device='cuda:0'), grad: tensor([-2.1551e-06, 9.3132e-09, 1.3039e-08, -1.8626e-08, 9.3132e-09, + 3.5390e-08, -5.2154e-08, 5.5879e-09, 5.5879e-09, 2.1644e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 218.36, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4902 re_mapping 0.0039 re_causal 0.0111 /// teacc 99.04 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.1779, 0.2558, -0.0274, ..., -0.0256, 0.0650, -0.0429], + [ 0.2081, -0.1224, 0.0276, ..., -0.1244, 0.0255, -0.0992], + [-0.1019, -0.3098, -0.0626, ..., -0.0734, 0.0250, -0.3099], + ..., + [-0.1978, -0.3671, 0.0732, ..., -0.1285, -0.0989, -0.2065], + [-0.1636, 0.0020, 0.0481, ..., 0.0039, -0.1206, 0.0917], + [-0.1384, -0.1118, 0.1110, ..., 0.2285, -0.0855, -0.2335]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.5879e-09, 1.8626e-09, 4.0978e-08, ..., 1.4901e-08, + 0.0000e+00, 1.8626e-09], + [ 5.5879e-09, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 5.5879e-09], + [ 5.5879e-09, 0.0000e+00, -5.5879e-08, ..., -2.0489e-08, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0327, 0.0018, 0.0200, 0.0159, 0.0386, 0.0143, -0.0018, 0.0058, + -0.0163, -0.0514], device='cuda:0'), grad: tensor([ 1.0058e-07, 1.8626e-08, -3.7253e-08, 3.9116e-08, -3.5390e-08, + 1.9185e-07, -3.5390e-07, 1.0803e-07, 2.6077e-08, -6.3330e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 218.24, cls_loss 0.0007 cls_loss_mapping 0.0009 cls_loss_causal 0.4685 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.09 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.1780, 0.2559, -0.0274, ..., -0.0256, 0.0647, -0.0431], + [ 0.2086, -0.1211, 0.0277, ..., -0.1247, 0.0258, -0.0991], + [-0.1020, -0.3100, -0.0625, ..., -0.0734, 0.0248, -0.3102], + ..., + [-0.1982, -0.3699, 0.0732, ..., -0.1287, -0.0990, -0.2069], + [-0.1641, 0.0015, 0.0471, ..., 0.0038, -0.1209, 0.0917], + [-0.1385, -0.1123, 0.1110, ..., 0.2286, -0.0856, -0.2343]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.9802e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 3.7253e-09, 1.1176e-07, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 5.5879e-09, 4.0978e-08, ..., 0.0000e+00, + -2.0489e-08, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -5.8226e-06, ..., 0.0000e+00, + 1.6764e-08, 1.8626e-09], + [ 0.0000e+00, -9.3132e-09, 7.8231e-08, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 0.0000e+00, 5.5879e-09, 5.4315e-06, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0327, 0.0021, 0.0202, 0.0159, 0.0390, 0.0143, -0.0018, 0.0056, + -0.0170, -0.0517], device='cuda:0'), grad: tensor([ 1.6764e-08, 2.4401e-07, -2.8685e-07, -7.4506e-09, 2.0247e-06, + 3.5390e-08, 3.5390e-08, -1.0610e-05, 1.3970e-07, 8.4341e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 218.22, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4679 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.04 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.1782, 0.2562, -0.0287, ..., -0.0256, 0.0631, -0.0433], + [ 0.2106, -0.1211, 0.0290, ..., -0.1251, 0.0258, -0.0992], + [-0.1024, -0.3104, -0.0629, ..., -0.0735, 0.0250, -0.3107], + ..., + [-0.2002, -0.3706, 0.0719, ..., -0.1291, -0.0994, -0.2077], + [-0.1645, 0.0023, 0.0476, ..., 0.0032, -0.1210, 0.0931], + [-0.1387, -0.1128, 0.1120, ..., 0.2291, -0.0847, -0.2350]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.7253e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.5763e-07, 0.0000e+00, -1.2852e-07, ..., 1.8626e-09, + -6.5193e-08, -1.8626e-09], + [ 6.1467e-08, 0.0000e+00, 5.0291e-08, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + ..., + [ 2.3097e-07, 0.0000e+00, 3.5390e-08, ..., 1.6764e-08, + 4.2841e-08, 1.8626e-09], + [ 2.6077e-08, 0.0000e+00, 2.0489e-08, ..., 1.8626e-09, + 5.5879e-09, 1.8626e-09], + [ 1.6764e-08, 1.8626e-09, -4.5896e-06, ..., -3.1851e-06, + 3.7253e-09, 1.8626e-09]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0325, 0.0036, 0.0197, 0.0157, 0.0386, 0.0144, -0.0021, 0.0041, + -0.0163, -0.0508], device='cuda:0'), grad: tensor([ 3.1665e-08, -4.0978e-07, 1.6205e-07, 1.4901e-08, 1.7911e-05, + 1.2107e-07, 1.4901e-08, 1.0617e-07, 5.9605e-08, -1.8016e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 217.97, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4357 re_mapping 0.0040 re_causal 0.0117 /// teacc 98.99 lr 0.00010000 +Epoch 396, weight, value: tensor([[-1.7838e-01, 2.5692e-01, -2.8370e-02, ..., -2.5682e-02, + 6.2559e-02, -4.3259e-02], + [ 2.1095e-01, -1.1940e-01, 2.9174e-02, ..., -1.2280e-01, + 2.6379e-02, -9.8846e-02], + [-1.0282e-01, -3.1084e-01, -6.3113e-02, ..., -7.4177e-02, + 2.4800e-02, -3.1110e-01], + ..., + [-2.0024e-01, -3.7223e-01, 7.2049e-02, ..., -1.2950e-01, + -9.9715e-02, -2.0794e-01], + [-1.6649e-01, 1.1595e-03, 4.6109e-02, ..., -3.4621e-04, + -1.2134e-01, 9.3166e-02], + [-1.3869e-01, -1.1451e-01, 1.1183e-01, ..., 2.3007e-01, + -8.4410e-02, -2.3522e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0327, 0.0039, 0.0194, 0.0158, 0.0383, 0.0144, -0.0019, 0.0042, + -0.0175, -0.0509], device='cuda:0'), grad: tensor([ 2.9802e-08, 3.7253e-09, -2.8685e-07, 2.9802e-08, -1.8626e-09, + -9.3132e-09, 3.7253e-09, 2.6077e-07, 0.0000e+00, -2.9802e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 217.92, cls_loss 0.0010 cls_loss_mapping 0.0010 cls_loss_causal 0.4500 re_mapping 0.0040 re_causal 0.0115 /// teacc 99.02 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.1785, 0.2571, -0.0284, ..., -0.0257, 0.0627, -0.0451], + [ 0.2140, -0.1194, 0.0311, ..., -0.1233, 0.0264, -0.0990], + [-0.1034, -0.3111, -0.0638, ..., -0.0743, 0.0248, -0.3119], + ..., + [-0.2032, -0.3730, 0.0703, ..., -0.1290, -0.0998, -0.2084], + [-0.1672, 0.0015, 0.0463, ..., -0.0004, -0.1214, 0.0934], + [-0.1389, -0.1152, 0.1118, ..., 0.2304, -0.0853, -0.2364]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 2.0489e-08, + 0.0000e+00, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 4.6939e-07, ..., 3.5390e-08, + 0.0000e+00, 6.7055e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -6.3144e-07, ..., 5.5879e-09, + 0.0000e+00, 1.6764e-08], + [ 1.8626e-09, 0.0000e+00, 7.8231e-08, ..., -1.0990e-07, + 0.0000e+00, -1.7323e-07], + [ 1.8626e-09, 1.8626e-09, -9.3132e-09, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0326, 0.0064, 0.0199, 0.0158, 0.0384, 0.0143, -0.0020, 0.0018, + -0.0175, -0.0511], device='cuda:0'), grad: tensor([ 3.1665e-08, 1.2852e-07, 8.4750e-07, 1.3597e-07, 1.3039e-08, + 2.6077e-08, 2.4214e-08, -9.8161e-07, -2.2724e-07, -7.4506e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 217.80, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4695 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.03 lr 0.00010000 +Epoch 398, weight, value: tensor([[-1.7859e-01, 2.5717e-01, -2.8390e-02, ..., -2.5733e-02, + 6.2488e-02, -4.6610e-02], + [ 2.1405e-01, -1.1892e-01, 3.0433e-02, ..., -1.2549e-01, + 2.6267e-02, -9.9152e-02], + [-1.0366e-01, -3.1118e-01, -6.4109e-02, ..., -7.4654e-02, + 2.2704e-02, -3.1243e-01], + ..., + [-2.0324e-01, -3.7483e-01, 7.1050e-02, ..., -1.2676e-01, + -9.9879e-02, -2.0935e-01], + [-1.6742e-01, 1.3527e-03, 4.5874e-02, ..., -3.6184e-04, + -1.2144e-01, 9.3399e-02], + [-1.3953e-01, -1.1568e-01, 1.1174e-01, ..., 2.3036e-01, + -8.5414e-02, -2.3776e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-09, -3.7253e-09, 7.4506e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [-8.3633e-07, -2.9989e-07, -8.7917e-07, ..., -2.4959e-07, + -1.8626e-09, 0.0000e+00], + [ 1.5646e-07, 5.5879e-08, 1.6578e-07, ..., 4.6566e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 5.2899e-07, 1.8813e-07, 5.4203e-07, ..., 1.5646e-07, + 1.8626e-09, -1.8626e-09], + [ 4.4703e-08, 1.8626e-08, 4.8429e-08, ..., 1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 8.1956e-08, 3.3528e-08, 9.3132e-08, ..., 2.4214e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0312, 0.0055, 0.0196, 0.0158, 0.0382, 0.0143, -0.0003, 0.0028, + -0.0181, -0.0514], device='cuda:0'), grad: tensor([ 1.8626e-08, -2.0806e-06, 1.3411e-07, 5.4017e-08, 1.8626e-08, + 5.5879e-09, 2.2352e-08, 1.2927e-06, 1.2107e-07, 4.0233e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 218.05, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4578 re_mapping 0.0041 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +Epoch 399, weight, value: tensor([[-1.7902e-01, 2.5737e-01, -2.8528e-02, ..., -2.5781e-02, + 6.1815e-02, -4.6798e-02], + [ 2.1410e-01, -1.1860e-01, 3.0179e-02, ..., -1.2635e-01, + 2.6442e-02, -9.9240e-02], + [-1.0388e-01, -3.1307e-01, -6.5033e-02, ..., -7.6485e-02, + 2.2565e-02, -3.1432e-01], + ..., + [-2.0317e-01, -3.7560e-01, 7.1391e-02, ..., -1.2594e-01, + -1.0015e-01, -2.1030e-01], + [-1.6804e-01, 2.9610e-03, 4.7254e-02, ..., 1.6661e-04, + -1.2151e-01, 9.4479e-02], + [-1.3960e-01, -1.1601e-01, 1.1175e-01, ..., 2.3065e-01, + -8.5648e-02, -2.3782e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, -1.6764e-08], + [ 0.0000e+00, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0312, 0.0052, 0.0185, 0.0159, 0.0384, 0.0142, -0.0003, 0.0033, + -0.0170, -0.0516], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.8626e-08, -3.7253e-09, 7.4506e-09, -8.7544e-08, + 1.6764e-08, 1.3039e-08, -5.5879e-09, -2.4214e-08, 5.0291e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 217.69, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4645 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.05 lr 0.00010000 +Epoch 400, weight, value: tensor([[-1.7980e-01, 2.5822e-01, -2.8495e-02, ..., -2.5814e-02, + 6.1512e-02, -4.7021e-02], + [ 2.1415e-01, -1.1848e-01, 3.0202e-02, ..., -1.2636e-01, + 2.6456e-02, -9.9332e-02], + [-1.0386e-01, -3.1419e-01, -6.5382e-02, ..., -7.6565e-02, + 2.2637e-02, -3.1634e-01], + ..., + [-2.0319e-01, -3.7605e-01, 7.1399e-02, ..., -1.2599e-01, + -1.0022e-01, -2.1121e-01], + [-1.6844e-01, 3.2970e-03, 4.7844e-02, ..., 2.1424e-04, + -1.2156e-01, 9.4761e-02], + [-1.3982e-01, -1.1679e-01, 1.1180e-01, ..., 2.3089e-01, + -8.6072e-02, -2.3900e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.7253e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, -1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 1.8626e-09, 7.4506e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0315, 0.0051, 0.0184, 0.0154, 0.0386, 0.0147, -0.0006, 0.0033, + -0.0167, -0.0518], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.4214e-08, -3.5390e-08, 1.6205e-07, 5.5879e-09, + -9.6858e-08, -1.6764e-08, -8.9407e-08, 1.1176e-08, 2.2352e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 218.19, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4671 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.07 lr 0.00001000 +Epoch 401, weight, value: tensor([[-1.8026e-01, 2.5852e-01, -2.8951e-02, ..., -2.6327e-02, + 6.1190e-02, -4.7099e-02], + [ 2.1498e-01, -1.2078e-01, 3.0886e-02, ..., -1.2639e-01, + 2.6337e-02, -1.0079e-01], + [-1.0445e-01, -3.1448e-01, -6.5031e-02, ..., -7.7084e-02, + 2.2771e-02, -3.1667e-01], + ..., + [-2.0440e-01, -3.7634e-01, 7.0835e-02, ..., -1.2600e-01, + -1.0313e-01, -2.1207e-01], + [-1.6949e-01, 2.8522e-03, 4.7836e-02, ..., 2.1537e-04, + -1.2179e-01, 9.4547e-02], + [-1.4122e-01, -1.1707e-01, 1.1161e-01, ..., 2.3117e-01, + -8.6317e-02, -2.4012e-01]], device='cuda:0'), grad: tensor([[ 1.1176e-08, 5.5879e-09, 9.3132e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [-2.7008e-07, -1.2666e-07, -1.9372e-07, ..., 0.0000e+00, + -8.1956e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 7.4506e-09, 3.7253e-09, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.4901e-07, 7.0781e-08, 1.0617e-07, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0314, 0.0056, 0.0192, 0.0154, 0.0386, 0.0148, -0.0008, 0.0027, + -0.0174, -0.0521], device='cuda:0'), grad: tensor([ 2.0489e-08, -4.1351e-07, 1.8626e-09, -3.7253e-09, -9.3132e-09, + 2.0489e-08, 1.3597e-07, 1.1176e-08, 2.3283e-07, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 217.87, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4822 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.11 lr 0.00001000 +Epoch 402, weight, value: tensor([[-1.8027e-01, 2.5867e-01, -2.8903e-02, ..., -2.6332e-02, + 6.1394e-02, -4.7105e-02], + [ 2.1500e-01, -1.2096e-01, 3.0890e-02, ..., -1.2639e-01, + 2.6327e-02, -1.0080e-01], + [-1.0449e-01, -3.1453e-01, -6.5074e-02, ..., -7.7090e-02, + 2.2789e-02, -3.1670e-01], + ..., + [-2.0440e-01, -3.7643e-01, 7.0828e-02, ..., -1.2601e-01, + -1.0314e-01, -2.1212e-01], + [-1.6975e-01, 2.6657e-03, 4.7750e-02, ..., 2.1589e-04, + -1.2186e-01, 9.4458e-02], + [-1.4122e-01, -1.1711e-01, 1.1165e-01, ..., 2.3120e-01, + -8.6324e-02, -2.4024e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3225e-07, ..., 6.7055e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.5832e-07, ..., -7.8231e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0315, 0.0056, 0.0192, 0.0154, 0.0386, 0.0149, -0.0008, 0.0027, + -0.0175, -0.0520], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.7253e-09, -9.3132e-09, 3.9116e-08, 0.0000e+00, + 3.7253e-09, 0.0000e+00, 3.5577e-07, 7.4506e-09, -3.9488e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 218.04, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4681 re_mapping 0.0037 re_causal 0.0116 /// teacc 99.11 lr 0.00001000 +Epoch 403, weight, value: tensor([[-1.8027e-01, 2.5887e-01, -2.8853e-02, ..., -2.6332e-02, + 6.1646e-02, -4.7106e-02], + [ 2.1506e-01, -1.2113e-01, 3.0906e-02, ..., -1.2640e-01, + 2.6463e-02, -1.0067e-01], + [-1.0454e-01, -3.1456e-01, -6.5154e-02, ..., -7.7065e-02, + 2.2795e-02, -3.1673e-01], + ..., + [-2.0444e-01, -3.7648e-01, 7.0820e-02, ..., -1.2603e-01, + -1.0325e-01, -2.1216e-01], + [-1.6982e-01, 2.6287e-03, 4.7704e-02, ..., 1.9860e-04, + -1.2189e-01, 9.4414e-02], + [-1.4122e-01, -1.1718e-01, 1.1167e-01, ..., 2.3125e-01, + -8.6332e-02, -2.4027e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09], + [-4.8429e-08, 5.5879e-09, -1.4901e-08, ..., 0.0000e+00, + -1.3039e-08, 7.4506e-09], + [ 1.8626e-09, 2.4214e-08, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 2.4214e-08], + ..., + [ 7.4506e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, -3.1665e-08, -2.7940e-08, ..., 0.0000e+00, + 0.0000e+00, -3.5390e-08], + [ 1.1176e-08, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0316, 0.0056, 0.0192, 0.0154, 0.0386, 0.0149, -0.0008, 0.0026, + -0.0176, -0.0520], device='cuda:0'), grad: tensor([ 3.1665e-08, -5.2154e-08, 6.8918e-08, 3.5390e-08, 5.5879e-09, + -5.5879e-09, -1.4901e-08, 1.4901e-08, -9.8720e-08, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 218.04, cls_loss 0.0007 cls_loss_mapping 0.0008 cls_loss_causal 0.4602 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.11 lr 0.00001000 +Epoch 404, weight, value: tensor([[-1.8028e-01, 2.5891e-01, -2.8845e-02, ..., -2.6370e-02, + 6.1617e-02, -4.7130e-02], + [ 2.1510e-01, -1.2092e-01, 3.0930e-02, ..., -1.2641e-01, + 2.6828e-02, -1.0033e-01], + [-1.0458e-01, -3.1459e-01, -6.5248e-02, ..., -7.7116e-02, + 2.2801e-02, -3.1679e-01], + ..., + [-2.0444e-01, -3.7653e-01, 7.0821e-02, ..., -1.2603e-01, + -1.0328e-01, -2.1224e-01], + [-1.6992e-01, 2.5845e-03, 4.7685e-02, ..., 1.8641e-04, + -1.2193e-01, 9.4400e-02], + [-1.4120e-01, -1.1723e-01, 1.1169e-01, ..., 2.3129e-01, + -8.6347e-02, -2.4030e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.0978e-08, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [-3.7253e-09, 1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0316, 0.0057, 0.0191, 0.0154, 0.0386, 0.0149, -0.0010, 0.0026, + -0.0176, -0.0520], device='cuda:0'), grad: tensor([-9.4995e-08, 3.5390e-08, -2.6077e-08, -1.6205e-07, 1.8626e-09, + 1.0803e-07, 2.0489e-08, 8.3819e-08, 1.8626e-09, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 217.93, cls_loss 0.0007 cls_loss_mapping 0.0006 cls_loss_causal 0.4420 re_mapping 0.0035 re_causal 0.0113 /// teacc 99.10 lr 0.00001000 +Epoch 405, weight, value: tensor([[-1.8028e-01, 2.5894e-01, -2.8839e-02, ..., -2.6375e-02, + 6.1610e-02, -4.7136e-02], + [ 2.1512e-01, -1.2091e-01, 3.0936e-02, ..., -1.2641e-01, + 2.6840e-02, -1.0033e-01], + [-1.0464e-01, -3.1461e-01, -6.5286e-02, ..., -7.7080e-02, + 2.2811e-02, -3.1682e-01], + ..., + [-2.0445e-01, -3.7662e-01, 7.0823e-02, ..., -1.2603e-01, + -1.0331e-01, -2.1232e-01], + [-1.7004e-01, 2.5171e-03, 4.7640e-02, ..., 1.7734e-04, + -1.2197e-01, 9.4367e-02], + [-1.4122e-01, -1.1730e-01, 1.1169e-01, ..., 2.3130e-01, + -8.6370e-02, -2.4035e-01]], device='cuda:0'), grad: tensor([[ 8.3819e-08, 2.0489e-08, 7.2643e-08, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [-1.2144e-06, -3.4273e-07, -8.9779e-07, ..., 0.0000e+00, + -1.3225e-07, 0.0000e+00], + [ 1.0990e-07, 1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-08, 2.9802e-08, 7.8231e-08, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 7.6555e-07, 2.6077e-07, 6.8545e-07, ..., 0.0000e+00, + 1.0058e-07, 0.0000e+00], + [ 5.5879e-09, 3.7253e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0316, 0.0057, 0.0191, 0.0154, 0.0386, 0.0149, -0.0010, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([ 1.3597e-07, -2.0545e-06, 1.3597e-07, 3.1665e-08, -5.5879e-08, + 5.7742e-08, 1.8626e-07, 1.6950e-07, 1.2703e-06, 1.2107e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 217.83, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4727 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.11 lr 0.00001000 +Epoch 406, weight, value: tensor([[-1.8029e-01, 2.5900e-01, -2.8823e-02, ..., -2.6375e-02, + 6.1650e-02, -4.7135e-02], + [ 2.1515e-01, -1.2093e-01, 3.0946e-02, ..., -1.2644e-01, + 2.6852e-02, -1.0033e-01], + [-1.0468e-01, -3.1464e-01, -6.5337e-02, ..., -7.7079e-02, + 2.2800e-02, -3.1686e-01], + ..., + [-2.0447e-01, -3.7674e-01, 7.0808e-02, ..., -1.2609e-01, + -1.0339e-01, -2.1235e-01], + [-1.7007e-01, 2.5664e-03, 4.7654e-02, ..., 1.7221e-04, + -1.2196e-01, 9.4420e-02], + [-1.4117e-01, -1.1734e-01, 1.1173e-01, ..., 2.3144e-01, + -8.6373e-02, -2.4040e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 5.5879e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [-7.4506e-09, 2.0489e-08, 3.3528e-08, ..., 0.0000e+00, + 0.0000e+00, 1.6764e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, -1.0431e-07, -1.9372e-07, ..., 0.0000e+00, + 0.0000e+00, -8.9407e-08], + [ 1.8626e-09, 7.0781e-08, 1.1921e-07, ..., 0.0000e+00, + 0.0000e+00, 6.1467e-08]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0316, 0.0057, 0.0191, 0.0154, 0.0386, 0.0149, -0.0010, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([ 2.0489e-08, 2.4401e-07, -3.5390e-07, -1.0058e-07, 1.3039e-07, + 1.0245e-07, 7.2643e-08, 1.4901e-08, -4.3586e-07, 3.0361e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 218.02, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4277 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.08 lr 0.00001000 +Epoch 407, weight, value: tensor([[-1.8030e-01, 2.5903e-01, -2.8830e-02, ..., -2.6386e-02, + 6.1649e-02, -4.7171e-02], + [ 2.1517e-01, -1.2088e-01, 3.0958e-02, ..., -1.2644e-01, + 2.6939e-02, -1.0025e-01], + [-1.0469e-01, -3.1467e-01, -6.5396e-02, ..., -7.7085e-02, + 2.2794e-02, -3.1689e-01], + ..., + [-2.0447e-01, -3.7676e-01, 7.0809e-02, ..., -1.2609e-01, + -1.0341e-01, -2.1242e-01], + [-1.7011e-01, 2.5545e-03, 4.7638e-02, ..., 1.6076e-04, + -1.2197e-01, 9.4416e-02], + [-1.4119e-01, -1.1737e-01, 1.1175e-01, ..., 2.3145e-01, + -8.6371e-02, -2.4043e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0316, 0.0057, 0.0191, 0.0154, 0.0386, 0.0149, -0.0010, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([ 1.8626e-09, 7.4506e-09, 4.6566e-08, 5.4017e-08, 7.4506e-09, + -4.6566e-08, 1.8626e-09, -7.8231e-08, 1.1176e-08, -7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 218.20, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4533 re_mapping 0.0034 re_causal 0.0111 /// teacc 99.07 lr 0.00001000 +Epoch 408, weight, value: tensor([[-1.8030e-01, 2.5909e-01, -2.8815e-02, ..., -2.6394e-02, + 6.1675e-02, -4.7170e-02], + [ 2.1522e-01, -1.2091e-01, 3.0981e-02, ..., -1.2644e-01, + 2.6946e-02, -1.0026e-01], + [-1.0470e-01, -3.1469e-01, -6.5462e-02, ..., -7.7080e-02, + 2.2806e-02, -3.1691e-01], + ..., + [-2.0452e-01, -3.7682e-01, 7.0792e-02, ..., -1.2610e-01, + -1.0345e-01, -2.1246e-01], + [-1.7014e-01, 2.5462e-03, 4.7635e-02, ..., 1.4931e-04, + -1.2198e-01, 9.4414e-02], + [-1.4123e-01, -1.1739e-01, 1.1176e-01, ..., 2.3146e-01, + -8.6385e-02, -2.4049e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.5018e-07, -1.3784e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 1.8626e-08, -1.5832e-08, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0316, 0.0057, 0.0191, 0.0154, 0.0386, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([-1.4622e-06, 3.1665e-08, 2.8871e-08, 2.1420e-08, 3.5390e-08, + 1.4901e-08, 1.3094e-06, -8.3819e-09, 2.3283e-08, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 217.72, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4456 re_mapping 0.0034 re_causal 0.0110 /// teacc 99.11 lr 0.00001000 +Epoch 409, weight, value: tensor([[-1.8032e-01, 2.5912e-01, -2.8809e-02, ..., -2.6437e-02, + 6.1669e-02, -4.7181e-02], + [ 2.1523e-01, -1.2091e-01, 3.0986e-02, ..., -1.2644e-01, + 2.6949e-02, -1.0028e-01], + [-1.0471e-01, -3.1472e-01, -6.5563e-02, ..., -7.7136e-02, + 2.2809e-02, -3.1695e-01], + ..., + [-2.0453e-01, -3.7688e-01, 7.0794e-02, ..., -1.2611e-01, + -1.0345e-01, -2.1252e-01], + [-1.7020e-01, 2.5214e-03, 4.7626e-02, ..., 1.4920e-04, + -1.2200e-01, 9.4407e-02], + [-1.4126e-01, -1.1742e-01, 1.1176e-01, ..., 2.3149e-01, + -8.6407e-02, -2.4058e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -7.4506e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.0978e-08, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-1.7695e-08, 0.0000e+00, -7.9162e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.1176e-08, 1.8626e-09, 3.9116e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0316, 0.0057, 0.0190, 0.0154, 0.0386, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([ 0.0000e+00, -5.4017e-08, -8.3819e-09, 5.5879e-09, -7.4506e-09, + -2.7940e-09, 8.9407e-08, -1.0524e-07, 2.1420e-08, 7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 217.90, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4433 re_mapping 0.0033 re_causal 0.0112 /// teacc 99.08 lr 0.00001000 +Epoch 410, weight, value: tensor([[-1.8034e-01, 2.5915e-01, -2.8800e-02, ..., -2.6427e-02, + 6.1663e-02, -4.7191e-02], + [ 2.1525e-01, -1.2092e-01, 3.0996e-02, ..., -1.2644e-01, + 2.6952e-02, -1.0030e-01], + [-1.0473e-01, -3.1475e-01, -6.5591e-02, ..., -7.7134e-02, + 2.2813e-02, -3.1699e-01], + ..., + [-2.0455e-01, -3.7693e-01, 7.0781e-02, ..., -1.2615e-01, + -1.0346e-01, -2.1256e-01], + [-1.7024e-01, 2.5250e-03, 4.7623e-02, ..., 1.3539e-04, + -1.2200e-01, 9.4418e-02], + [-1.4126e-01, -1.1746e-01, 1.1180e-01, ..., 2.3156e-01, + -8.6416e-02, -2.4069e-01]], device='cuda:0'), grad: tensor([[-4.6566e-09, -1.2852e-07, -2.5146e-08, ..., 0.0000e+00, + 0.0000e+00, -4.4703e-08], + [-1.1176e-08, 9.3132e-10, -4.6566e-09, ..., 0.0000e+00, + -9.3132e-10, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 1.8626e-09, 6.5193e-09, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0316, 0.0057, 0.0190, 0.0154, 0.0386, 0.0148, -0.0009, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([-2.0396e-07, -8.3819e-09, 4.6566e-09, 2.8871e-08, 1.8626e-08, + 9.0338e-08, 8.6613e-08, 8.3819e-09, -7.4506e-09, -1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 217.67, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4335 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.07 lr 0.00001000 +Epoch 411, weight, value: tensor([[-1.8033e-01, 2.5919e-01, -2.8816e-02, ..., -2.6431e-02, + 6.1657e-02, -4.7193e-02], + [ 2.1527e-01, -1.2091e-01, 3.1003e-02, ..., -1.2644e-01, + 2.6960e-02, -1.0031e-01], + [-1.0474e-01, -3.1477e-01, -6.5631e-02, ..., -7.7137e-02, + 2.2806e-02, -3.1701e-01], + ..., + [-2.0457e-01, -3.7700e-01, 7.0779e-02, ..., -1.2616e-01, + -1.0350e-01, -2.1261e-01], + [-1.7029e-01, 2.5034e-03, 4.7603e-02, ..., 1.3296e-04, + -1.2201e-01, 9.4406e-02], + [-1.4129e-01, -1.1750e-01, 1.1183e-01, ..., 2.3156e-01, + -8.6440e-02, -2.4076e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0315, 0.0057, 0.0190, 0.0154, 0.0386, 0.0148, -0.0009, 0.0026, + -0.0178, -0.0520], device='cuda:0'), grad: tensor([-1.3039e-08, 9.3132e-09, -2.7940e-09, 7.4506e-09, 2.7940e-09, + 2.7940e-09, 4.6566e-09, -4.1910e-08, 1.8626e-09, 4.0047e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 218.10, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4477 re_mapping 0.0033 re_causal 0.0110 /// teacc 99.07 lr 0.00001000 +Epoch 412, weight, value: tensor([[-1.8034e-01, 2.5923e-01, -2.8844e-02, ..., -2.6432e-02, + 6.1644e-02, -4.7212e-02], + [ 2.1531e-01, -1.2085e-01, 3.1012e-02, ..., -1.2649e-01, + 2.7060e-02, -1.0021e-01], + [-1.0478e-01, -3.1481e-01, -6.5664e-02, ..., -7.7128e-02, + 2.2806e-02, -3.1706e-01], + ..., + [-2.0459e-01, -3.7705e-01, 7.0778e-02, ..., -1.2616e-01, + -1.0353e-01, -2.1270e-01], + [-1.7032e-01, 2.5379e-03, 4.7611e-02, ..., 1.3041e-04, + -1.2201e-01, 9.4432e-02], + [-1.4123e-01, -1.1754e-01, 1.1187e-01, ..., 2.3163e-01, + -8.6457e-02, -2.4085e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -7.4506e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 6.5193e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + ..., + [ 4.6566e-09, 3.9116e-08, -7.4506e-09, ..., 0.0000e+00, + 1.8626e-09, 3.3528e-08], + [ 5.5879e-09, 1.4901e-08, -2.0489e-08, ..., -9.3132e-10, + 1.8626e-09, 2.7940e-09], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0315, 0.0057, 0.0190, 0.0154, 0.0386, 0.0148, -0.0010, 0.0026, + -0.0178, -0.0520], device='cuda:0'), grad: tensor([-3.7253e-09, 2.7940e-08, 8.1956e-08, 3.6322e-08, 4.6566e-09, + -8.4750e-08, -5.6811e-08, -6.5193e-09, -2.9802e-08, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 217.55, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4465 re_mapping 0.0033 re_causal 0.0111 /// teacc 99.09 lr 0.00001000 +Epoch 413, weight, value: tensor([[-1.8033e-01, 2.5929e-01, -2.8849e-02, ..., -2.6479e-02, + 6.1641e-02, -4.7193e-02], + [ 2.1531e-01, -1.2085e-01, 3.1013e-02, ..., -1.2649e-01, + 2.7061e-02, -1.0021e-01], + [-1.0478e-01, -3.1483e-01, -6.5689e-02, ..., -7.7168e-02, + 2.2801e-02, -3.1709e-01], + ..., + [-2.0459e-01, -3.7710e-01, 7.0780e-02, ..., -1.2616e-01, + -1.0354e-01, -2.1274e-01], + [-1.7036e-01, 2.5339e-03, 4.7610e-02, ..., 1.2749e-04, + -1.2202e-01, 9.4436e-02], + [-1.4124e-01, -1.1759e-01, 1.1189e-01, ..., 2.3167e-01, + -8.6468e-02, -2.4089e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -4.8429e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.4506e-09, 7.4506e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 1.8626e-09, 1.3039e-08, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0315, 0.0057, 0.0190, 0.0154, 0.0386, 0.0148, -0.0010, 0.0026, + -0.0178, -0.0520], device='cuda:0'), grad: tensor([-4.1910e-08, 6.5193e-09, 1.5832e-08, 1.6764e-08, 9.3132e-10, + -3.4459e-08, 3.0734e-08, -4.3772e-08, 2.7940e-08, 2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 218.38, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4417 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.10 lr 0.00001000 +Epoch 414, weight, value: tensor([[-1.8034e-01, 2.5935e-01, -2.8828e-02, ..., -2.6400e-02, + 6.1640e-02, -4.7195e-02], + [ 2.1532e-01, -1.2086e-01, 3.1014e-02, ..., -1.2650e-01, + 2.7061e-02, -1.0022e-01], + [-1.0478e-01, -3.1486e-01, -6.5743e-02, ..., -7.7175e-02, + 2.2796e-02, -3.1714e-01], + ..., + [-2.0459e-01, -3.7714e-01, 7.0779e-02, ..., -1.2620e-01, + -1.0353e-01, -2.1281e-01], + [-1.7037e-01, 2.5511e-03, 4.7652e-02, ..., 1.2707e-04, + -1.2202e-01, 9.4447e-02], + [-1.4124e-01, -1.1769e-01, 1.1190e-01, ..., 2.3170e-01, + -8.6499e-02, -2.4094e-01]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 3.7253e-09, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, -3.7253e-09, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, -8.3819e-09], + [ 1.8626e-09, 2.7940e-09, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0316, 0.0057, 0.0189, 0.0154, 0.0387, 0.0148, -0.0010, 0.0026, + -0.0178, -0.0521], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.8626e-08, 0.0000e+00, 9.8720e-08, -1.2107e-08, + -1.0990e-07, 5.5879e-09, 7.4506e-09, -2.0489e-08, 1.2107e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 217.86, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4439 re_mapping 0.0033 re_causal 0.0110 /// teacc 99.15 lr 0.00001000 +Epoch 415, weight, value: tensor([[-1.8034e-01, 2.5943e-01, -2.8809e-02, ..., -2.6399e-02, + 6.1634e-02, -4.7179e-02], + [ 2.1533e-01, -1.2086e-01, 3.1012e-02, ..., -1.2650e-01, + 2.7063e-02, -1.0022e-01], + [-1.0469e-01, -3.1491e-01, -6.5719e-02, ..., -7.7166e-02, + 2.2799e-02, -3.1717e-01], + ..., + [-2.0461e-01, -3.7717e-01, 7.0779e-02, ..., -1.2620e-01, + -1.0355e-01, -2.1284e-01], + [-1.7038e-01, 2.6075e-03, 4.7702e-02, ..., 1.2947e-04, + -1.2199e-01, 9.4504e-02], + [-1.4125e-01, -1.1777e-01, 1.1191e-01, ..., 2.3172e-01, + -8.6500e-02, -2.4097e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.9791e-07, -9.6858e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-09, 3.9116e-08, 1.8626e-08, ..., 1.8626e-09, + 0.0000e+00, 1.1176e-08], + [ 0.0000e+00, 1.6578e-07, 3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 6.1467e-08, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.6764e-08, -3.5390e-08, -3.9116e-08, ..., -5.5879e-09, + 0.0000e+00, -2.7940e-08], + [ 1.8626e-09, 1.9185e-07, 3.9116e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0316, 0.0057, 0.0190, 0.0154, 0.0387, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([-1.1809e-06, 9.1270e-08, 2.8685e-07, 1.2852e-07, 2.6077e-08, + 1.4715e-07, 1.0803e-07, 1.1548e-07, -9.6858e-08, 3.6508e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 218.15, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4683 re_mapping 0.0033 re_causal 0.0112 /// teacc 99.12 lr 0.00001000 +Epoch 416, weight, value: tensor([[-1.8035e-01, 2.5949e-01, -2.8790e-02, ..., -2.6430e-02, + 6.1629e-02, -4.7189e-02], + [ 2.1535e-01, -1.2086e-01, 3.1021e-02, ..., -1.2650e-01, + 2.7069e-02, -1.0023e-01], + [-1.0469e-01, -3.1497e-01, -6.5723e-02, ..., -7.7189e-02, + 2.2784e-02, -3.1720e-01], + ..., + [-2.0464e-01, -3.7727e-01, 7.0770e-02, ..., -1.2621e-01, + -1.0358e-01, -2.1290e-01], + [-1.7040e-01, 2.6411e-03, 4.7730e-02, ..., 1.2741e-04, + -1.2200e-01, 9.4530e-02], + [-1.4128e-01, -1.1788e-01, 1.1192e-01, ..., 2.3175e-01, + -8.6536e-02, -2.4104e-01]], device='cuda:0'), grad: tensor([[ 7.4506e-09, -9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + ..., + [-1.8626e-08, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -5.5879e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09], + [ 3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0316, 0.0057, 0.0191, 0.0154, 0.0387, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 6.8918e-08, 1.1176e-08, 1.3039e-08, 1.1176e-08, -7.4506e-09, + 5.5879e-08, 2.4214e-08, -2.3283e-07, -1.6764e-08, 6.5193e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 218.10, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4152 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.12 lr 0.00001000 +Epoch 417, weight, value: tensor([[-1.8036e-01, 2.5952e-01, -2.8787e-02, ..., -2.6435e-02, + 6.1615e-02, -4.7194e-02], + [ 2.1536e-01, -1.2078e-01, 3.1032e-02, ..., -1.2650e-01, + 2.7209e-02, -1.0010e-01], + [-1.0461e-01, -3.1500e-01, -6.5667e-02, ..., -7.7189e-02, + 2.2794e-02, -3.1725e-01], + ..., + [-2.0464e-01, -3.7730e-01, 7.0770e-02, ..., -1.2621e-01, + -1.0359e-01, -2.1300e-01], + [-1.7043e-01, 2.6521e-03, 4.7739e-02, ..., 1.2983e-04, + -1.2201e-01, 9.4550e-02], + [-1.4133e-01, -1.1794e-01, 1.1192e-01, ..., 2.3176e-01, + -8.6554e-02, -2.4113e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, -1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 1.8626e-09, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0316, 0.0057, 0.0192, 0.0154, 0.0387, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.8626e-09, 2.4214e-08, -4.2841e-08, -1.0431e-07, + 3.7253e-09, 2.4214e-08, 2.7940e-08, 2.7940e-08, 3.7253e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 217.72, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4489 re_mapping 0.0032 re_causal 0.0108 /// teacc 99.11 lr 0.00001000 +Epoch 418, weight, value: tensor([[-1.8036e-01, 2.5961e-01, -2.8764e-02, ..., -2.6436e-02, + 6.1615e-02, -4.7198e-02], + [ 2.1537e-01, -1.2078e-01, 3.1031e-02, ..., -1.2650e-01, + 2.7259e-02, -1.0006e-01], + [-1.0461e-01, -3.1503e-01, -6.5653e-02, ..., -7.7189e-02, + 2.2781e-02, -3.1728e-01], + ..., + [-2.0464e-01, -3.7732e-01, 7.0777e-02, ..., -1.2621e-01, + -1.0358e-01, -2.1305e-01], + [-1.7045e-01, 2.6597e-03, 4.7757e-02, ..., 1.3048e-04, + -1.2202e-01, 9.4554e-02], + [-1.4136e-01, -1.1802e-01, 1.1192e-01, ..., 2.3178e-01, + -8.6562e-02, -2.4123e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 5.4017e-08, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 4.0978e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, -8.7544e-08, -1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, -7.6368e-08], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0317, 0.0057, 0.0192, 0.0154, 0.0387, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0522], device='cuda:0'), grad: tensor([ 4.8429e-08, 6.5193e-08, -1.4529e-07, 4.6566e-08, -1.6950e-07, + 4.6566e-08, 1.8626e-08, 2.7940e-07, -2.4401e-07, 5.0291e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 217.87, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4269 re_mapping 0.0032 re_causal 0.0108 /// teacc 99.12 lr 0.00001000 +Epoch 419, weight, value: tensor([[-1.8035e-01, 2.5973e-01, -2.8764e-02, ..., -2.6439e-02, + 6.1624e-02, -4.7186e-02], + [ 2.1540e-01, -1.2085e-01, 3.1043e-02, ..., -1.2651e-01, + 2.7264e-02, -1.0007e-01], + [-1.0462e-01, -3.1505e-01, -6.5678e-02, ..., -7.7181e-02, + 2.2779e-02, -3.1732e-01], + ..., + [-2.0467e-01, -3.7735e-01, 7.0767e-02, ..., -1.2624e-01, + -1.0363e-01, -2.1313e-01], + [-1.7048e-01, 2.6639e-03, 4.7780e-02, ..., 1.2803e-04, + -1.2202e-01, 9.4569e-02], + [-1.4137e-01, -1.1808e-01, 1.1193e-01, ..., 2.3181e-01, + -8.6679e-02, -2.4128e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 2.2352e-08, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + ..., + [ 1.8626e-09, 5.5879e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + [ 0.0000e+00, -5.9605e-08, -7.4506e-08, ..., 0.0000e+00, + 0.0000e+00, -8.0094e-08], + [ 0.0000e+00, 3.7253e-09, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0317, 0.0057, 0.0192, 0.0154, 0.0388, 0.0148, -0.0011, 0.0026, + -0.0177, -0.0522], device='cuda:0'), grad: tensor([ 1.8626e-09, 5.5879e-09, 7.2643e-08, 3.3528e-08, 1.8626e-09, + 1.0058e-07, 5.5879e-09, -9.3132e-09, -2.4773e-07, 4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 218.22, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4584 re_mapping 0.0031 re_causal 0.0111 /// teacc 99.12 lr 0.00001000 +Epoch 420, weight, value: tensor([[-1.8035e-01, 2.5977e-01, -2.8749e-02, ..., -2.6439e-02, + 6.1618e-02, -4.7195e-02], + [ 2.1542e-01, -1.2081e-01, 3.1057e-02, ..., -1.2651e-01, + 2.7345e-02, -9.9991e-02], + [-1.0464e-01, -3.1510e-01, -6.5723e-02, ..., -7.7184e-02, + 2.2773e-02, -3.1737e-01], + ..., + [-2.0467e-01, -3.7739e-01, 7.0768e-02, ..., -1.2624e-01, + -1.0364e-01, -2.1319e-01], + [-1.7049e-01, 2.7051e-03, 4.7833e-02, ..., 1.2703e-04, + -1.2201e-01, 9.4613e-02], + [-1.4139e-01, -1.1813e-01, 1.1193e-01, ..., 2.3182e-01, + -8.6694e-02, -2.4137e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-08, 0.0000e+00, 6.1467e-08, ..., 1.3039e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-5.0291e-08, 0.0000e+00, -1.0617e-07, ..., -3.7253e-08, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0317, 0.0057, 0.0192, 0.0154, 0.0388, 0.0148, -0.0011, 0.0026, + -0.0177, -0.0522], device='cuda:0'), grad: tensor([ 6.5193e-08, 5.1968e-07, 7.4506e-09, 5.5879e-09, -1.0710e-06, + 2.9802e-08, -3.7439e-07, 5.4017e-08, 1.6764e-08, 7.3016e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 218.05, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4403 re_mapping 0.0031 re_causal 0.0106 /// teacc 99.12 lr 0.00001000 +Epoch 421, weight, value: tensor([[-1.8038e-01, 2.5983e-01, -2.8787e-02, ..., -2.6471e-02, + 6.1611e-02, -4.7226e-02], + [ 2.1542e-01, -1.2081e-01, 3.1040e-02, ..., -1.2662e-01, + 2.7349e-02, -1.0000e-01], + [-1.0466e-01, -3.1514e-01, -6.5753e-02, ..., -7.7207e-02, + 2.2772e-02, -3.1740e-01], + ..., + [-2.0468e-01, -3.7744e-01, 7.0773e-02, ..., -1.2624e-01, + -1.0365e-01, -2.1330e-01], + [-1.7052e-01, 2.7416e-03, 4.7875e-02, ..., 1.2256e-04, + -1.2201e-01, 9.4637e-02], + [-1.4124e-01, -1.1820e-01, 1.1201e-01, ..., 2.3198e-01, + -8.6709e-02, -2.4145e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-09, 2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, -9.3132e-09, -9.3132e-09, ..., 0.0000e+00, + 1.8626e-09, -5.5879e-09], + [ 1.8626e-09, 3.7253e-09, -3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0317, 0.0057, 0.0192, 0.0155, 0.0388, 0.0148, -0.0011, 0.0026, + -0.0177, -0.0522], device='cuda:0'), grad: tensor([ 6.3330e-08, 7.4506e-09, 9.3132e-09, 1.4901e-08, 1.1176e-08, + -5.7742e-08, -5.5879e-09, 3.7253e-09, -5.5879e-09, -4.0978e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 218.09, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4321 re_mapping 0.0031 re_causal 0.0107 /// teacc 99.13 lr 0.00001000 +Epoch 422, weight, value: tensor([[-1.8039e-01, 2.5991e-01, -2.8818e-02, ..., -2.6493e-02, + 6.1622e-02, -4.7227e-02], + [ 2.1542e-01, -1.2085e-01, 3.1022e-02, ..., -1.2663e-01, + 2.7343e-02, -1.0002e-01], + [-1.0456e-01, -3.1517e-01, -6.5672e-02, ..., -7.7226e-02, + 2.2777e-02, -3.1745e-01], + ..., + [-2.0469e-01, -3.7749e-01, 7.0786e-02, ..., -1.2624e-01, + -1.0362e-01, -2.1339e-01], + [-1.7055e-01, 2.7444e-03, 4.7866e-02, ..., 1.2068e-04, + -1.2202e-01, 9.4644e-02], + [-1.4129e-01, -1.1825e-01, 1.1205e-01, ..., 2.3201e-01, + -8.6707e-02, -2.4158e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 5.5879e-09, 3.7253e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3039e-08], + [ 3.7253e-09, 1.1176e-08, 2.4214e-08, ..., 0.0000e+00, + -5.5879e-09, 1.3039e-08], + ..., + [ 1.3039e-08, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 2.0489e-08], + [ 1.8626e-09, -5.5879e-08, -1.1548e-07, ..., 0.0000e+00, + 0.0000e+00, -4.2841e-08], + [ 7.4506e-09, 3.7253e-08, 5.4017e-08, ..., 0.0000e+00, + 0.0000e+00, 3.9116e-08]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0317, 0.0057, 0.0193, 0.0155, 0.0388, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0522], device='cuda:0'), grad: tensor([-5.4017e-08, 5.4017e-08, 6.3330e-08, -8.6427e-07, 8.0094e-08, + 7.5996e-07, 8.3819e-08, 5.2154e-08, -2.6636e-07, 8.7544e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 218.26, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4316 re_mapping 0.0031 re_causal 0.0108 /// teacc 99.13 lr 0.00001000 +Epoch 423, weight, value: tensor([[-1.8039e-01, 2.5996e-01, -2.8826e-02, ..., -2.6501e-02, + 6.1617e-02, -4.7253e-02], + [ 2.1539e-01, -1.2086e-01, 3.0995e-02, ..., -1.2683e-01, + 2.7344e-02, -1.0002e-01], + [-1.0456e-01, -3.1520e-01, -6.5691e-02, ..., -7.7221e-02, + 2.2775e-02, -3.1748e-01], + ..., + [-2.0470e-01, -3.7752e-01, 7.0794e-02, ..., -1.2625e-01, + -1.0364e-01, -2.1346e-01], + [-1.7057e-01, 2.7438e-03, 4.7882e-02, ..., 1.0201e-04, + -1.2202e-01, 9.4647e-02], + [-1.4103e-01, -1.1829e-01, 1.1215e-01, ..., 2.3227e-01, + -8.6727e-02, -2.4160e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0317, 0.0056, 0.0193, 0.0155, 0.0388, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 5.5879e-09, 3.7253e-09, -7.4506e-09, 5.5879e-09, -9.3132e-09, + 5.5879e-09, 0.0000e+00, -1.8626e-08, 0.0000e+00, 1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 218.39, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4661 re_mapping 0.0031 re_causal 0.0110 /// teacc 99.15 lr 0.00001000 +Epoch 424, weight, value: tensor([[-1.8041e-01, 2.6000e-01, -2.8827e-02, ..., -2.6507e-02, + 6.1609e-02, -4.7310e-02], + [ 2.1540e-01, -1.2087e-01, 3.0993e-02, ..., -1.2683e-01, + 2.7348e-02, -1.0004e-01], + [-1.0452e-01, -3.1524e-01, -6.5671e-02, ..., -7.7229e-02, + 2.2752e-02, -3.1754e-01], + ..., + [-2.0470e-01, -3.7756e-01, 7.0797e-02, ..., -1.2625e-01, + -1.0366e-01, -2.1349e-01], + [-1.7057e-01, 2.7794e-03, 4.7913e-02, ..., 9.8714e-05, + -1.2201e-01, 9.4670e-02], + [-1.4104e-01, -1.1837e-01, 1.1216e-01, ..., 2.3228e-01, + -8.6757e-02, -2.4162e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.0978e-08, 0.0000e+00, 1.1362e-07, ..., 5.5879e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.4901e-07, 0.0000e+00, 1.7323e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.1048e-07, 1.8626e-09, -3.6508e-07, ..., -9.3132e-09, + -1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0317, 0.0056, 0.0194, 0.0155, 0.0388, 0.0148, -0.0010, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 0.0000e+00, 2.5518e-07, 7.4506e-08, 9.3132e-09, 9.1270e-08, + 5.5879e-09, 1.8626e-09, 4.7125e-07, 9.3132e-09, -9.1828e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 218.39, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4441 re_mapping 0.0031 re_causal 0.0109 /// teacc 99.13 lr 0.00001000 +Epoch 425, weight, value: tensor([[-1.8041e-01, 2.6012e-01, -2.8829e-02, ..., -2.6513e-02, + 6.1636e-02, -4.7315e-02], + [ 2.1543e-01, -1.2087e-01, 3.1006e-02, ..., -1.2688e-01, + 2.7497e-02, -9.9913e-02], + [-1.0449e-01, -3.1527e-01, -6.5658e-02, ..., -7.7234e-02, + 2.2763e-02, -3.1757e-01], + ..., + [-2.0473e-01, -3.7760e-01, 7.0794e-02, ..., -1.2625e-01, + -1.0374e-01, -2.1357e-01], + [-1.7060e-01, 2.7753e-03, 4.7911e-02, ..., 9.8216e-05, + -1.2202e-01, 9.4671e-02], + [-1.4102e-01, -1.1847e-01, 1.1220e-01, ..., 2.3234e-01, + -8.6797e-02, -2.4170e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.5193e-08, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, -1.6764e-08], + [-2.2352e-08, 3.7253e-09, -1.1176e-08, ..., 0.0000e+00, + -5.5879e-09, 1.8626e-09], + [-7.4506e-09, 1.8626e-09, 0.0000e+00, ..., -1.3039e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.8626e-09, 9.3132e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 9.3132e-09, 1.8626e-09, 1.8626e-09, ..., 1.1176e-08, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0317, 0.0056, 0.0195, 0.0155, 0.0388, 0.0148, -0.0011, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([-1.2480e-07, -2.6077e-08, -1.0245e-07, 7.4506e-08, 2.7940e-08, + -5.5879e-09, 1.1176e-08, 0.0000e+00, 2.2352e-08, 1.0990e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 218.24, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4288 re_mapping 0.0030 re_causal 0.0106 /// teacc 99.13 lr 0.00001000 +Epoch 426, weight, value: tensor([[-1.8042e-01, 2.6017e-01, -2.8878e-02, ..., -2.6519e-02, + 6.1626e-02, -4.7328e-02], + [ 2.1544e-01, -1.2087e-01, 3.1005e-02, ..., -1.2688e-01, + 2.7503e-02, -9.9918e-02], + [-1.0449e-01, -3.1531e-01, -6.5681e-02, ..., -7.7200e-02, + 2.2795e-02, -3.1760e-01], + ..., + [-2.0473e-01, -3.7763e-01, 7.0807e-02, ..., -1.2626e-01, + -1.0374e-01, -2.1362e-01], + [-1.7065e-01, 2.7711e-03, 4.7938e-02, ..., 1.0090e-04, + -1.2202e-01, 9.4681e-02], + [-1.4105e-01, -1.1853e-01, 1.1221e-01, ..., 2.3235e-01, + -8.6821e-02, -2.4176e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0317, 0.0056, 0.0195, 0.0155, 0.0388, 0.0148, -0.0011, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.8068e-07, 7.4506e-09, -7.4506e-09, -2.5705e-07, + 9.3132e-09, -7.4506e-09, 1.4901e-08, 0.0000e+00, 5.2154e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 218.27, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4226 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.15 lr 0.00001000 +Epoch 427, weight, value: tensor([[-1.8043e-01, 2.6021e-01, -2.8866e-02, ..., -2.6509e-02, + 6.1616e-02, -4.7338e-02], + [ 2.1545e-01, -1.2086e-01, 3.1014e-02, ..., -1.2689e-01, + 2.7545e-02, -9.9895e-02], + [-1.0449e-01, -3.1533e-01, -6.5699e-02, ..., -7.7199e-02, + 2.2789e-02, -3.1763e-01], + ..., + [-2.0474e-01, -3.7769e-01, 7.0805e-02, ..., -1.2627e-01, + -1.0379e-01, -2.1371e-01], + [-1.7067e-01, 2.8033e-03, 4.7964e-02, ..., 1.0140e-04, + -1.2203e-01, 9.4718e-02], + [-1.4107e-01, -1.1862e-01, 1.1222e-01, ..., 2.3238e-01, + -8.6852e-02, -2.4182e-01]], device='cuda:0'), grad: tensor([[ 7.4506e-09, 3.7253e-09, 1.3039e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.0245e-07, -5.7742e-08, -1.6764e-07, ..., 0.0000e+00, + -3.3528e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.4214e-08, 1.3039e-08, 2.2352e-08, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 5.7742e-08, 2.2352e-08, 9.1270e-08, ..., 0.0000e+00, + 1.8626e-08, -3.7253e-09], + [ 9.3132e-09, 5.5879e-09, 7.4506e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0317, 0.0056, 0.0195, 0.0155, 0.0389, 0.0148, -0.0011, 0.0026, + -0.0177, -0.0522], device='cuda:0'), grad: tensor([ 9.6858e-08, -3.9116e-08, 9.4064e-07, 3.9116e-08, 2.8498e-07, + 7.4506e-08, -1.6987e-06, 4.4703e-08, 2.2724e-07, 2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 218.06, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4220 re_mapping 0.0031 re_causal 0.0106 /// teacc 99.14 lr 0.00001000 +Epoch 428, weight, value: tensor([[-1.8044e-01, 2.6026e-01, -2.8922e-02, ..., -2.6526e-02, + 6.1609e-02, -4.7387e-02], + [ 2.1549e-01, -1.2084e-01, 3.1024e-02, ..., -1.2689e-01, + 2.7589e-02, -9.9871e-02], + [-1.0452e-01, -3.1535e-01, -6.5721e-02, ..., -7.7201e-02, + 2.2783e-02, -3.1768e-01], + ..., + [-2.0476e-01, -3.7774e-01, 7.0806e-02, ..., -1.2628e-01, + -1.0385e-01, -2.1377e-01], + [-1.7069e-01, 2.8184e-03, 4.7980e-02, ..., 1.0258e-04, + -1.2204e-01, 9.4710e-02], + [-1.4108e-01, -1.1867e-01, 1.1226e-01, ..., 2.3240e-01, + -8.6876e-02, -2.4196e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-3.4329e-06, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.4342e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + ..., + [ 3.2503e-06, 0.0000e+00, 1.4715e-07, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-08], + [ 0.0000e+00, -7.4506e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 1.8626e-09, 3.7253e-09, -1.6391e-07, ..., -1.8626e-09, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0317, 0.0056, 0.0195, 0.0155, 0.0388, 0.0147, -0.0011, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 1.1176e-08, -4.2208e-06, 2.3842e-07, -1.0617e-07, 7.4506e-09, + 7.4506e-09, 3.7253e-09, 4.4703e-06, -2.4214e-08, -4.1723e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 218.53, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4225 re_mapping 0.0031 re_causal 0.0107 /// teacc 99.14 lr 0.00001000 +Epoch 429, weight, value: tensor([[-1.8043e-01, 2.6037e-01, -2.8901e-02, ..., -2.6527e-02, + 6.1599e-02, -4.7393e-02], + [ 2.1551e-01, -1.2084e-01, 3.1025e-02, ..., -1.2689e-01, + 2.7595e-02, -9.9878e-02], + [-1.0454e-01, -3.1542e-01, -6.5749e-02, ..., -7.7199e-02, + 2.2777e-02, -3.1772e-01], + ..., + [-2.0478e-01, -3.7782e-01, 7.0804e-02, ..., -1.2629e-01, + -1.0388e-01, -2.1385e-01], + [-1.7072e-01, 2.8142e-03, 4.7964e-02, ..., 1.0052e-04, + -1.2204e-01, 9.4717e-02], + [-1.4109e-01, -1.1879e-01, 1.1230e-01, ..., 2.3242e-01, + -8.6894e-02, -2.4201e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.1176e-08, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + -3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-09, 1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 1.1176e-08, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 3.7253e-09, 3.7253e-09, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0317, 0.0056, 0.0195, 0.0155, 0.0388, 0.0147, -0.0011, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([-4.2841e-08, -1.1176e-08, -2.9802e-08, 2.7381e-07, -7.4506e-09, + -3.2037e-07, 5.2154e-08, 2.7940e-08, 4.8429e-08, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 218.13, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4351 re_mapping 0.0030 re_causal 0.0108 /// teacc 99.12 lr 0.00001000 +Epoch 430, weight, value: tensor([[-1.8043e-01, 2.6047e-01, -2.8878e-02, ..., -2.6527e-02, + 6.1587e-02, -4.7390e-02], + [ 2.1552e-01, -1.2086e-01, 3.1025e-02, ..., -1.2691e-01, + 2.7591e-02, -9.9894e-02], + [-1.0455e-01, -3.1546e-01, -6.5811e-02, ..., -7.7188e-02, + 2.2769e-02, -3.1776e-01], + ..., + [-2.0479e-01, -3.7787e-01, 7.0801e-02, ..., -1.2634e-01, + -1.0389e-01, -2.1398e-01], + [-1.7074e-01, 2.8210e-03, 4.7981e-02, ..., 9.8250e-05, + -1.2204e-01, 9.4736e-02], + [-1.4107e-01, -1.1886e-01, 1.1235e-01, ..., 2.3249e-01, + -8.6912e-02, -2.4205e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 7.4506e-09, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0318, 0.0056, 0.0195, 0.0155, 0.0388, 0.0147, -0.0011, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([-9.3132e-09, 1.3039e-08, 5.5879e-09, 3.7253e-09, -1.3970e-07, + 7.4506e-09, 7.4506e-09, -3.7253e-08, -5.5879e-09, 1.5460e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 218.37, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4534 re_mapping 0.0030 re_causal 0.0111 /// teacc 99.12 lr 0.00001000 +Epoch 431, weight, value: tensor([[-1.8044e-01, 2.6059e-01, -2.8844e-02, ..., -2.6519e-02, + 6.1578e-02, -4.7392e-02], + [ 2.1552e-01, -1.2087e-01, 3.1023e-02, ..., -1.2693e-01, + 2.7592e-02, -9.9903e-02], + [-1.0456e-01, -3.1548e-01, -6.5880e-02, ..., -7.7190e-02, + 2.2752e-02, -3.1778e-01], + ..., + [-2.0480e-01, -3.7791e-01, 7.0805e-02, ..., -1.2638e-01, + -1.0390e-01, -2.1402e-01], + [-1.7076e-01, 2.8245e-03, 4.7994e-02, ..., 9.6210e-05, + -1.2204e-01, 9.4751e-02], + [-1.4105e-01, -1.1897e-01, 1.1237e-01, ..., 2.3255e-01, + -8.6921e-02, -2.4208e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 7.4506e-09, -1.8626e-09, -1.6764e-08, ..., 0.0000e+00, + 1.8626e-09, -1.1176e-08], + [ 1.8626e-09, 9.3132e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0318, 0.0056, 0.0194, 0.0155, 0.0388, 0.0147, -0.0011, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 9.8720e-08, 3.1851e-07, -1.2908e-06, 2.3283e-07, 2.7940e-08, + -1.8626e-09, 1.0803e-07, 3.0920e-07, 1.1176e-07, 8.3819e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 218.06, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4804 re_mapping 0.0031 re_causal 0.0112 /// teacc 99.13 lr 0.00001000 +Epoch 432, weight, value: tensor([[-1.8045e-01, 2.6067e-01, -2.8819e-02, ..., -2.6510e-02, + 6.1574e-02, -4.7473e-02], + [ 2.1550e-01, -1.2089e-01, 3.1000e-02, ..., -1.2701e-01, + 2.7589e-02, -9.9910e-02], + [-1.0448e-01, -3.1552e-01, -6.5817e-02, ..., -7.7193e-02, + 2.2780e-02, -3.1786e-01], + ..., + [-2.0480e-01, -3.7795e-01, 7.0806e-02, ..., -1.2639e-01, + -1.0390e-01, -2.1407e-01], + [-1.7077e-01, 2.8563e-03, 4.8036e-02, ..., 9.4921e-05, + -1.2204e-01, 9.4776e-02], + [-1.4096e-01, -1.1907e-01, 1.1246e-01, ..., 2.3268e-01, + -8.6946e-02, -2.4212e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.2352e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.4214e-08, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 5.5879e-09, 1.8626e-09, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.3039e-08, 0.0000e+00, -1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, -1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 0.0000e+00, 1.1176e-08, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0319, 0.0056, 0.0196, 0.0155, 0.0388, 0.0147, -0.0011, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([-1.4901e-08, 1.6764e-08, -9.4995e-08, 1.8626e-08, -3.7253e-09, + -9.3132e-09, 3.1665e-08, -1.8626e-09, 0.0000e+00, 6.3330e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 217.97, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4250 re_mapping 0.0029 re_causal 0.0103 /// teacc 99.14 lr 0.00001000 +Epoch 433, weight, value: tensor([[-1.8045e-01, 2.6080e-01, -2.8779e-02, ..., -2.6503e-02, + 6.1575e-02, -4.7520e-02], + [ 2.1551e-01, -1.2089e-01, 3.0986e-02, ..., -1.2705e-01, + 2.7584e-02, -9.9916e-02], + [-1.0451e-01, -3.1555e-01, -6.5854e-02, ..., -7.7185e-02, + 2.2842e-02, -3.1791e-01], + ..., + [-2.0481e-01, -3.7800e-01, 7.0821e-02, ..., -1.2640e-01, + -1.0386e-01, -2.1416e-01], + [-1.7079e-01, 2.8806e-03, 4.8057e-02, ..., 9.1855e-05, + -1.2205e-01, 9.4808e-02], + [-1.4093e-01, -1.1918e-01, 1.1248e-01, ..., 2.3275e-01, + -8.6968e-02, -2.4215e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.6764e-08, 1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -7.8231e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -5.5879e-09, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [ 0.0000e+00, 7.4506e-09, 3.5390e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0319, 0.0056, 0.0196, 0.0155, 0.0388, 0.0147, -0.0012, 0.0026, + -0.0177, -0.0520], device='cuda:0'), grad: tensor([ 4.4703e-08, 4.8429e-08, 7.4506e-09, 1.3039e-08, 1.8626e-08, + 1.1176e-08, 2.6077e-08, -2.6077e-07, -1.8626e-08, 1.1362e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 217.74, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0030 re_causal 0.0106 /// teacc 99.12 lr 0.00001000 +Epoch 434, weight, value: tensor([[-1.8044e-01, 2.6088e-01, -2.8741e-02, ..., -2.6556e-02, + 6.1572e-02, -4.7553e-02], + [ 2.1552e-01, -1.2091e-01, 3.0977e-02, ..., -1.2706e-01, + 2.7585e-02, -9.9932e-02], + [-1.0452e-01, -3.1560e-01, -6.5896e-02, ..., -7.7212e-02, + 2.2823e-02, -3.1797e-01], + ..., + [-2.0482e-01, -3.7805e-01, 7.0842e-02, ..., -1.2640e-01, + -1.0386e-01, -2.1423e-01], + [-1.7079e-01, 2.9126e-03, 4.8089e-02, ..., 9.4221e-05, + -1.2204e-01, 9.4848e-02], + [-1.4095e-01, -1.1930e-01, 1.1246e-01, ..., 2.3278e-01, + -8.6980e-02, -2.4224e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0319, 0.0055, 0.0196, 0.0155, 0.0388, 0.0147, -0.0012, 0.0026, + -0.0177, -0.0521], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.1176e-08, -7.4506e-09, -1.1176e-07, -1.4715e-07, + 3.9116e-08, 1.8626e-09, -5.5879e-09, 6.5193e-08, 1.5460e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 217.85, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4243 re_mapping 0.0030 re_causal 0.0107 /// teacc 99.11 lr 0.00001000 +Epoch 435, weight, value: tensor([[-1.8046e-01, 2.6098e-01, -2.8702e-02, ..., -2.6550e-02, + 6.1569e-02, -4.7549e-02], + [ 2.1556e-01, -1.2091e-01, 3.0985e-02, ..., -1.2706e-01, + 2.7584e-02, -9.9947e-02], + [-1.0453e-01, -3.1561e-01, -6.5904e-02, ..., -7.7201e-02, + 2.2824e-02, -3.1801e-01], + ..., + [-2.0485e-01, -3.7808e-01, 7.0843e-02, ..., -1.2641e-01, + -1.0386e-01, -2.1438e-01], + [-1.7083e-01, 2.8960e-03, 4.8059e-02, ..., 9.2770e-05, + -1.2204e-01, 9.4860e-02], + [-1.4097e-01, -1.1944e-01, 1.1245e-01, ..., 2.3278e-01, + -8.7047e-02, -2.4230e-01]], device='cuda:0'), grad: tensor([[ 2.0489e-08, 1.5832e-08, 1.8626e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [-1.1548e-07, 0.0000e+00, -1.3970e-07, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 1.3970e-08, -9.3132e-10, 1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-3.9116e-08, -3.7253e-08, 7.4506e-09, ..., 5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 9.2201e-08, 2.7940e-09, 1.0896e-07, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 1.2107e-08, 1.2107e-08, -2.7940e-09, ..., 1.1362e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0319, 0.0056, 0.0196, 0.0155, 0.0389, 0.0147, -0.0011, 0.0026, + -0.0178, -0.0521], device='cuda:0'), grad: tensor([ 2.5146e-07, -3.2131e-07, 2.3283e-08, 1.6764e-08, -7.6834e-07, + 1.1828e-07, 1.9558e-08, -5.0012e-07, 2.6915e-07, 8.9314e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 218.40, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4313 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.11 lr 0.00001000 +Epoch 436, weight, value: tensor([[-1.8046e-01, 2.6104e-01, -2.8682e-02, ..., -2.6551e-02, + 6.1558e-02, -4.7602e-02], + [ 2.1556e-01, -1.2092e-01, 3.0971e-02, ..., -1.2707e-01, + 2.7581e-02, -9.9957e-02], + [-1.0454e-01, -3.1564e-01, -6.5943e-02, ..., -7.7202e-02, + 2.2828e-02, -3.1807e-01], + ..., + [-2.0485e-01, -3.7812e-01, 7.0868e-02, ..., -1.2640e-01, + -1.0384e-01, -2.1445e-01], + [-1.7087e-01, 2.9152e-03, 4.8061e-02, ..., 9.3022e-05, + -1.2205e-01, 9.4892e-02], + [-1.4098e-01, -1.1955e-01, 1.1246e-01, ..., 2.3279e-01, + -8.7069e-02, -2.4236e-01]], device='cuda:0'), grad: tensor([[-8.3819e-09, -1.8440e-07, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, -1.9558e-08], + [-2.2352e-08, 9.3132e-10, -1.5832e-08, ..., 0.0000e+00, + -1.8626e-09, 1.8626e-09], + [ 1.5832e-08, 9.3132e-09, 2.4214e-08, ..., 2.7940e-09, + 9.3132e-10, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.0338e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, -2.8871e-08, -2.7008e-08, ..., -4.6566e-09, + 0.0000e+00, -1.5832e-08], + [ 1.2107e-08, 1.4901e-08, -1.5553e-07, ..., 1.8626e-09, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0319, 0.0055, 0.0197, 0.0155, 0.0389, 0.0147, -0.0011, 0.0026, + -0.0178, -0.0522], device='cuda:0'), grad: tensor([-3.2503e-07, -3.9116e-08, 7.0781e-08, -9.3132e-10, 1.4529e-07, + 5.8673e-08, 3.0082e-07, 2.1886e-07, -1.0245e-07, -3.1944e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 218.32, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4167 re_mapping 0.0029 re_causal 0.0101 /// teacc 99.10 lr 0.00001000 +Epoch 437, weight, value: tensor([[-1.8049e-01, 2.6114e-01, -2.8669e-02, ..., -2.6554e-02, + 6.1537e-02, -4.7656e-02], + [ 2.1558e-01, -1.2094e-01, 3.0972e-02, ..., -1.2708e-01, + 2.7586e-02, -9.9986e-02], + [-1.0456e-01, -3.1568e-01, -6.5886e-02, ..., -7.7209e-02, + 2.2849e-02, -3.1814e-01], + ..., + [-2.0487e-01, -3.7819e-01, 7.0865e-02, ..., -1.2642e-01, + -1.0391e-01, -2.1453e-01], + [-1.7097e-01, 2.9074e-03, 4.8061e-02, ..., 9.3615e-05, + -1.2207e-01, 9.4918e-02], + [-1.4101e-01, -1.1963e-01, 1.1249e-01, ..., 2.3281e-01, + -8.7059e-02, -2.4246e-01]], device='cuda:0'), grad: tensor([[-3.2596e-08, -2.4308e-07, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -4.1910e-08], + [-9.3132e-10, 9.3132e-09, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 5.5879e-09], + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-9.3132e-10, -2.4214e-08, -1.3970e-08, ..., 0.0000e+00, + -2.7940e-09, -1.7695e-08], + [ 0.0000e+00, 1.8626e-09, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0319, 0.0055, 0.0198, 0.0155, 0.0390, 0.0147, -0.0012, 0.0026, + -0.0178, -0.0522], device='cuda:0'), grad: tensor([-6.4168e-07, 2.2352e-08, 5.5879e-09, 2.8871e-08, -1.1176e-07, + 2.0210e-07, 4.4517e-07, 2.3283e-08, -4.2841e-08, 6.1467e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 218.06, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4262 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.15 lr 0.00001000 +Epoch 438, weight, value: tensor([[-1.8052e-01, 2.6134e-01, -2.8621e-02, ..., -2.6546e-02, + 6.1526e-02, -4.7683e-02], + [ 2.1559e-01, -1.2101e-01, 3.0956e-02, ..., -1.2709e-01, + 2.7585e-02, -9.9999e-02], + [-1.0457e-01, -3.1572e-01, -6.5900e-02, ..., -7.7211e-02, + 2.2876e-02, -3.1819e-01], + ..., + [-2.0487e-01, -3.7825e-01, 7.0887e-02, ..., -1.2644e-01, + -1.0392e-01, -2.1469e-01], + [-1.7102e-01, 2.9233e-03, 4.8076e-02, ..., 9.8418e-05, + -1.2208e-01, 9.4945e-02], + [-1.4104e-01, -1.1988e-01, 1.1248e-01, ..., 2.3284e-01, + -8.7075e-02, -2.4254e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 2.7940e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 9.3132e-10, -1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1176e-08, 1.1176e-08, -3.7253e-09, ..., 0.0000e+00, + 4.6566e-09, 1.3039e-08], + [ 0.0000e+00, 3.7253e-09, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0320, 0.0055, 0.0198, 0.0155, 0.0390, 0.0147, -0.0012, 0.0026, + -0.0179, -0.0523], device='cuda:0'), grad: tensor([-1.5832e-08, 1.4901e-08, 1.6764e-08, 2.2352e-08, 9.3132e-10, + -3.4459e-08, 2.7940e-09, -3.3528e-08, 7.4506e-09, 2.4214e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 218.46, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4225 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.13 lr 0.00001000 +Epoch 439, weight, value: tensor([[-1.8052e-01, 2.6148e-01, -2.8584e-02, ..., -2.6552e-02, + 6.1506e-02, -4.7683e-02], + [ 2.1566e-01, -1.2101e-01, 3.0979e-02, ..., -1.2708e-01, + 2.7667e-02, -9.9929e-02], + [-1.0452e-01, -3.1576e-01, -6.5853e-02, ..., -7.7208e-02, + 2.2860e-02, -3.1825e-01], + ..., + [-2.0494e-01, -3.7836e-01, 7.0872e-02, ..., -1.2645e-01, + -1.0398e-01, -2.1489e-01], + [-1.7107e-01, 2.9058e-03, 4.8072e-02, ..., 9.6733e-05, + -1.2209e-01, 9.4924e-02], + [-1.4110e-01, -1.1999e-01, 1.1250e-01, ..., 2.3286e-01, + -8.7034e-02, -2.4262e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.0245e-07, -6.4261e-08, ..., -2.5146e-08, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-08, 9.3132e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 1.1176e-08, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + ..., + [ 1.8626e-09, 9.3132e-10, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 9.3132e-10, -1.8626e-09, -7.4506e-09, ..., 9.3132e-10, + 9.3132e-10, -7.4506e-09], + [ 9.3132e-10, 7.9162e-08, 5.4017e-08, ..., 2.5146e-08, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0320, 0.0055, 0.0199, 0.0155, 0.0391, 0.0147, -0.0012, 0.0025, + -0.0179, -0.0523], device='cuda:0'), grad: tensor([-2.3190e-07, 4.2841e-08, 3.7253e-09, -6.5193e-09, 8.3819e-09, + 4.6566e-09, 3.7253e-09, 4.6566e-09, -1.2107e-08, 1.9558e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 218.56, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4589 re_mapping 0.0030 re_causal 0.0106 /// teacc 99.13 lr 0.00001000 +Epoch 440, weight, value: tensor([[-1.8056e-01, 2.6158e-01, -2.8635e-02, ..., -2.6572e-02, + 6.1498e-02, -4.7811e-02], + [ 2.1562e-01, -1.2108e-01, 3.0925e-02, ..., -1.2737e-01, + 2.7673e-02, -9.9953e-02], + [-1.0444e-01, -3.1581e-01, -6.5831e-02, ..., -7.7215e-02, + 2.2881e-02, -3.1836e-01], + ..., + [-2.0496e-01, -3.7841e-01, 7.0876e-02, ..., -1.2649e-01, + -1.0402e-01, -2.1499e-01], + [-1.7110e-01, 2.9716e-03, 4.8150e-02, ..., 8.5745e-05, + -1.2209e-01, 9.4950e-02], + [-1.4087e-01, -1.2017e-01, 1.1271e-01, ..., 2.3321e-01, + -8.7047e-02, -2.4282e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [-3.7253e-09, 1.8626e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 0.0000e+00, 1.1176e-08, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 8.3819e-09], + ..., + [ 5.5879e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 9.3132e-10, -3.3528e-08, -1.6764e-08, ..., 0.0000e+00, + -1.8626e-09, -2.1420e-08], + [ 2.7940e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0320, 0.0054, 0.0200, 0.0155, 0.0393, 0.0147, -0.0012, 0.0025, + -0.0179, -0.0523], device='cuda:0'), grad: tensor([ 1.2666e-07, 2.9523e-07, -1.1688e-06, 1.8068e-07, 2.2352e-08, + 4.4703e-08, 7.9162e-08, 3.2037e-07, 4.0978e-08, 7.2643e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 218.41, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4418 re_mapping 0.0029 re_causal 0.0106 /// teacc 99.12 lr 0.00001000 +Epoch 441, weight, value: tensor([[-1.8063e-01, 2.6170e-01, -2.8622e-02, ..., -2.6575e-02, + 6.1468e-02, -4.7857e-02], + [ 2.1563e-01, -1.2107e-01, 3.0892e-02, ..., -1.2745e-01, + 2.7677e-02, -9.9978e-02], + [-1.0423e-01, -3.1584e-01, -6.5616e-02, ..., -7.7213e-02, + 2.2885e-02, -3.1845e-01], + ..., + [-2.0502e-01, -3.7844e-01, 7.0898e-02, ..., -1.2652e-01, + -1.0403e-01, -2.1507e-01], + [-1.7115e-01, 2.9935e-03, 4.8173e-02, ..., 9.5534e-05, + -1.2209e-01, 9.4958e-02], + [-1.4081e-01, -1.2027e-01, 1.1270e-01, ..., 2.3332e-01, + -8.7056e-02, -2.4299e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-08, 0.0000e+00, 3.1013e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-8.6613e-08, 0.0000e+00, -4.3027e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.0489e-08, 9.3132e-10, 1.1362e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0320, 0.0053, 0.0202, 0.0155, 0.0394, 0.0148, -0.0012, 0.0025, + -0.0179, -0.0524], device='cuda:0'), grad: tensor([-5.5879e-09, 4.2189e-07, 9.3132e-09, 5.5879e-09, -9.3132e-10, + 2.7940e-09, 1.8626e-09, -5.9512e-07, 2.7940e-09, 1.6671e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 218.21, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4485 re_mapping 0.0030 re_causal 0.0106 /// teacc 99.12 lr 0.00001000 +Epoch 442, weight, value: tensor([[-1.8067e-01, 2.6175e-01, -2.8702e-02, ..., -2.6582e-02, + 6.1457e-02, -4.7900e-02], + [ 2.1568e-01, -1.2103e-01, 3.0910e-02, ..., -1.2746e-01, + 2.7769e-02, -9.9888e-02], + [-1.0421e-01, -3.1589e-01, -6.5586e-02, ..., -7.7217e-02, + 2.2888e-02, -3.1854e-01], + ..., + [-2.0505e-01, -3.7854e-01, 7.0905e-02, ..., -1.2654e-01, + -1.0404e-01, -2.1517e-01], + [-1.7121e-01, 3.0432e-03, 4.8166e-02, ..., 1.0111e-04, + -1.2211e-01, 9.5033e-02], + [-1.4082e-01, -1.2036e-01, 1.1273e-01, ..., 2.3335e-01, + -8.7066e-02, -2.4306e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 6.5193e-09, 9.3132e-10, 8.3819e-09, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [-9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-4.6566e-09, 1.8626e-09, -8.3819e-09, ..., -9.3132e-10, + 0.0000e+00, 1.8626e-09], + [ 3.7253e-09, 3.7253e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 8.3819e-09, 1.0245e-08, -3.2596e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0320, 0.0053, 0.0203, 0.0155, 0.0395, 0.0148, -0.0012, 0.0025, + -0.0179, -0.0524], device='cuda:0'), grad: tensor([ 4.1910e-08, 4.7497e-08, -4.1910e-08, 8.0094e-08, -2.4121e-07, + -9.0338e-08, 3.0734e-08, 1.5832e-08, 3.3528e-08, 1.2573e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 218.41, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4289 re_mapping 0.0030 re_causal 0.0104 /// teacc 99.13 lr 0.00001000 +Epoch 443, weight, value: tensor([[-1.8075e-01, 2.6188e-01, -2.8706e-02, ..., -2.6589e-02, + 6.1414e-02, -4.7924e-02], + [ 2.1572e-01, -1.2099e-01, 3.0923e-02, ..., -1.2752e-01, + 2.7789e-02, -9.9908e-02], + [-1.0425e-01, -3.1593e-01, -6.5624e-02, ..., -7.7226e-02, + 2.2877e-02, -3.1861e-01], + ..., + [-2.0508e-01, -3.7859e-01, 7.0934e-02, ..., -1.2656e-01, + -1.0410e-01, -2.1532e-01], + [-1.7138e-01, 3.0092e-03, 4.8055e-02, ..., 8.9670e-05, + -1.2214e-01, 9.5042e-02], + [-1.4078e-01, -1.2046e-01, 1.1269e-01, ..., 2.3345e-01, + -8.7060e-02, -2.4312e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.0734e-08, -1.9558e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.4435e-07, 3.4459e-08, 1.3690e-07, ..., 1.4901e-08, + 0.0000e+00, 9.3132e-10], + [ 1.2573e-07, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 1.9558e-08, 9.3132e-10, -1.5646e-07, ..., 9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.3772e-08, -2.0489e-08, -2.7008e-08, ..., -3.7253e-08, + 0.0000e+00, -1.8626e-09]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0320, 0.0053, 0.0203, 0.0155, 0.0398, 0.0147, -0.0012, 0.0025, + -0.0180, -0.0526], device='cuda:0'), grad: tensor([-5.3085e-08, 1.5460e-07, 6.7987e-08, 9.4064e-08, -2.0154e-06, + 6.3330e-08, 6.5193e-09, -2.3656e-07, 1.5832e-08, 1.9064e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 218.23, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4442 re_mapping 0.0029 re_causal 0.0105 /// teacc 99.11 lr 0.00001000 +Epoch 444, weight, value: tensor([[-1.8078e-01, 2.6198e-01, -2.8678e-02, ..., -2.6599e-02, + 6.1398e-02, -4.7955e-02], + [ 2.1577e-01, -1.2106e-01, 3.0924e-02, ..., -1.2753e-01, + 2.7792e-02, -9.9934e-02], + [-1.0421e-01, -3.1602e-01, -6.5603e-02, ..., -7.7216e-02, + 2.2889e-02, -3.1871e-01], + ..., + [-2.0513e-01, -3.7863e-01, 7.0937e-02, ..., -1.2658e-01, + -1.0413e-01, -2.1541e-01], + [-1.7143e-01, 3.0947e-03, 4.8094e-02, ..., 9.6269e-05, + -1.2212e-01, 9.5158e-02], + [-1.4079e-01, -1.2051e-01, 1.1270e-01, ..., 2.3349e-01, + -8.7136e-02, -2.4319e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.7940e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0320, 0.0053, 0.0203, 0.0156, 0.0399, 0.0147, -0.0012, 0.0024, + -0.0180, -0.0527], device='cuda:0'), grad: tensor([ 4.5635e-08, 1.8626e-09, 0.0000e+00, 3.5390e-08, -2.1700e-07, + -2.6077e-08, -1.3039e-08, 6.5193e-09, 1.8626e-09, 1.6391e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 218.74, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4289 re_mapping 0.0029 re_causal 0.0103 /// teacc 99.11 lr 0.00001000 +Epoch 445, weight, value: tensor([[-1.8082e-01, 2.6210e-01, -2.8629e-02, ..., -2.6701e-02, + 6.1385e-02, -4.8000e-02], + [ 2.1582e-01, -1.2108e-01, 3.0944e-02, ..., -1.2763e-01, + 2.7795e-02, -9.9974e-02], + [-1.0424e-01, -3.1606e-01, -6.5633e-02, ..., -7.7250e-02, + 2.2878e-02, -3.1875e-01], + ..., + [-2.0519e-01, -3.7871e-01, 7.0904e-02, ..., -1.2662e-01, + -1.0414e-01, -2.1552e-01], + [-1.7146e-01, 3.1611e-03, 4.8174e-02, ..., 8.8556e-05, + -1.2210e-01, 9.5245e-02], + [-1.4072e-01, -1.2063e-01, 1.1277e-01, ..., 2.3366e-01, + -8.7185e-02, -2.4335e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 3.7253e-09, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0319, 0.0053, 0.0203, 0.0156, 0.0400, 0.0147, -0.0012, 0.0024, + -0.0179, -0.0527], device='cuda:0'), grad: tensor([-2.7940e-09, -9.3132e-10, -3.7253e-09, 7.4506e-09, 2.3283e-08, + -2.7940e-09, -9.3132e-10, 1.3039e-08, 5.5879e-09, -2.1420e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 218.41, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4258 re_mapping 0.0030 re_causal 0.0104 /// teacc 99.11 lr 0.00001000 +Epoch 446, weight, value: tensor([[-1.8087e-01, 2.6222e-01, -2.8660e-02, ..., -2.6712e-02, + 6.1385e-02, -4.8101e-02], + [ 2.1583e-01, -1.2118e-01, 3.0929e-02, ..., -1.2764e-01, + 2.7788e-02, -1.0003e-01], + [-1.0429e-01, -3.1610e-01, -6.5671e-02, ..., -7.7252e-02, + 2.2833e-02, -3.1883e-01], + ..., + [-2.0520e-01, -3.7877e-01, 7.0921e-02, ..., -1.2663e-01, + -1.0415e-01, -2.1572e-01], + [-1.7156e-01, 3.1297e-03, 4.8199e-02, ..., 8.9598e-05, + -1.2211e-01, 9.5192e-02], + [-1.4075e-01, -1.2075e-01, 1.1282e-01, ..., 2.3367e-01, + -8.7239e-02, -2.4364e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-9.3132e-10, 0.0000e+00, -3.1665e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 1.9558e-08, ..., 0.0000e+00, + 7.1712e-08, 0.0000e+00]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0320, 0.0053, 0.0203, 0.0156, 0.0400, 0.0147, -0.0011, 0.0024, + -0.0180, -0.0528], device='cuda:0'), grad: tensor([ 1.3039e-08, 6.3330e-08, -2.1420e-08, 3.7253e-09, -3.8743e-07, + 9.3132e-10, -2.1420e-08, -8.1025e-08, 1.3039e-08, 4.1164e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 218.41, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4571 re_mapping 0.0030 re_causal 0.0106 /// teacc 99.15 lr 0.00001000 +Epoch 447, weight, value: tensor([[-1.8091e-01, 2.6234e-01, -2.8624e-02, ..., -2.6706e-02, + 6.1377e-02, -4.8141e-02], + [ 2.1586e-01, -1.2116e-01, 3.0928e-02, ..., -1.2764e-01, + 2.7831e-02, -1.0002e-01], + [-1.0430e-01, -3.1617e-01, -6.5674e-02, ..., -7.7263e-02, + 2.2833e-02, -3.1895e-01], + ..., + [-2.0522e-01, -3.7885e-01, 7.0929e-02, ..., -1.2664e-01, + -1.0418e-01, -2.1597e-01], + [-1.7159e-01, 3.1486e-03, 4.8206e-02, ..., 7.3255e-05, + -1.2212e-01, 9.5198e-02], + [-1.4076e-01, -1.2087e-01, 1.1284e-01, ..., 2.3369e-01, + -8.7291e-02, -2.4368e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-09, 0.0000e+00, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0320, 0.0053, 0.0203, 0.0156, 0.0400, 0.0147, -0.0011, 0.0024, + -0.0181, -0.0528], device='cuda:0'), grad: tensor([ 7.4506e-09, 4.0047e-08, -1.2107e-08, 7.7300e-08, -6.5193e-09, + -1.1269e-07, 9.3132e-09, -1.4901e-08, -1.3970e-08, 2.8871e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 218.50, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4295 re_mapping 0.0030 re_causal 0.0104 /// teacc 99.13 lr 0.00001000 +Epoch 448, weight, value: tensor([[-1.8095e-01, 2.6254e-01, -2.8587e-02, ..., -2.6725e-02, + 6.1357e-02, -4.8167e-02], + [ 2.1598e-01, -1.2123e-01, 3.0968e-02, ..., -1.2765e-01, + 2.7869e-02, -1.0002e-01], + [-1.0432e-01, -3.1622e-01, -6.5713e-02, ..., -7.7275e-02, + 2.2840e-02, -3.1903e-01], + ..., + [-2.0533e-01, -3.7888e-01, 7.0917e-02, ..., -1.2669e-01, + -1.0419e-01, -2.1614e-01], + [-1.7166e-01, 3.2060e-03, 4.8301e-02, ..., 8.3326e-05, + -1.2211e-01, 9.5242e-02], + [-1.4078e-01, -1.2096e-01, 1.1281e-01, ..., 2.3375e-01, + -8.7295e-02, -2.4375e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.2107e-08], + ..., + [ 9.3132e-10, 9.3132e-10, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [-9.3132e-10, -1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [-9.3132e-10, -2.7940e-09, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0321, 0.0053, 0.0203, 0.0156, 0.0401, 0.0146, -0.0011, 0.0024, + -0.0180, -0.0528], device='cuda:0'), grad: tensor([ 3.0734e-08, 2.1420e-08, 1.2293e-07, -3.2783e-07, 1.0245e-08, + 3.6322e-08, 1.8626e-09, 9.2201e-08, 1.1176e-08, 1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 218.14, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4346 re_mapping 0.0029 re_causal 0.0105 /// teacc 99.11 lr 0.00001000 +Epoch 449, weight, value: tensor([[-1.8098e-01, 2.6268e-01, -2.8550e-02, ..., -2.6794e-02, + 6.1350e-02, -4.8182e-02], + [ 2.1605e-01, -1.2123e-01, 3.0995e-02, ..., -1.2774e-01, + 2.7867e-02, -1.0005e-01], + [-1.0433e-01, -3.1628e-01, -6.5737e-02, ..., -7.7303e-02, + 2.2861e-02, -3.1910e-01], + ..., + [-2.0542e-01, -3.7895e-01, 7.0868e-02, ..., -1.2672e-01, + -1.0420e-01, -2.1626e-01], + [-1.7171e-01, 3.2234e-03, 4.8325e-02, ..., 8.6717e-05, + -1.2212e-01, 9.5259e-02], + [-1.4072e-01, -1.2111e-01, 1.1290e-01, ..., 2.3389e-01, + -8.7298e-02, -2.4383e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 1.8626e-09, 0.0000e+00, -2.0675e-07, ..., 0.0000e+00, + 0.0000e+00, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 1.8533e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0320, 0.0054, 0.0203, 0.0156, 0.0401, 0.0146, -0.0011, 0.0023, + -0.0181, -0.0528], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.4901e-08, 9.3132e-10, 2.9802e-08, 1.8626e-09, + -2.6077e-08, 1.0245e-08, -6.5751e-07, 9.3132e-10, 6.3423e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 218.12, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4352 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.10 lr 0.00001000 +Epoch 450, weight, value: tensor([[-1.8100e-01, 2.6296e-01, -2.8472e-02, ..., -2.6795e-02, + 6.1342e-02, -4.8326e-02], + [ 2.1612e-01, -1.2131e-01, 3.1051e-02, ..., -1.2770e-01, + 2.7894e-02, -1.0004e-01], + [-1.0430e-01, -3.1633e-01, -6.5698e-02, ..., -7.7300e-02, + 2.2864e-02, -3.1922e-01], + ..., + [-2.0550e-01, -3.7904e-01, 7.0816e-02, ..., -1.2678e-01, + -1.0422e-01, -2.1638e-01], + [-1.7175e-01, 3.2565e-03, 4.8381e-02, ..., 8.6592e-05, + -1.2212e-01, 9.5282e-02], + [-1.4075e-01, -1.2134e-01, 1.1290e-01, ..., 2.3389e-01, + -8.7313e-02, -2.4388e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.1665e-08, -1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-9.3132e-10, 2.1420e-08, 1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 3.6322e-08, 2.3283e-08, 5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 2.8871e-08], + [ 1.8626e-09, 9.3132e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0319, 0.0054, 0.0204, 0.0157, 0.0401, 0.0146, -0.0008, 0.0022, + -0.0181, -0.0529], device='cuda:0'), grad: tensor([-3.1665e-08, 4.5635e-08, 3.7253e-09, 7.2736e-07, -6.6124e-08, + -7.9442e-07, -6.7055e-08, 1.6764e-08, 1.0990e-07, 5.5879e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 218.20, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4390 re_mapping 0.0028 re_causal 0.0105 /// teacc 99.09 lr 0.00001000 +Epoch 451, weight, value: tensor([[-1.8099e-01, 2.6308e-01, -2.8433e-02, ..., -2.6875e-02, + 6.1335e-02, -4.8361e-02], + [ 2.1618e-01, -1.2133e-01, 3.1077e-02, ..., -1.2771e-01, + 2.7895e-02, -1.0006e-01], + [-1.0426e-01, -3.1641e-01, -6.5666e-02, ..., -7.7334e-02, + 2.2868e-02, -3.1931e-01], + ..., + [-2.0557e-01, -3.7909e-01, 7.0806e-02, ..., -1.2680e-01, + -1.0423e-01, -2.1646e-01], + [-1.7186e-01, 3.2575e-03, 4.8448e-02, ..., 6.6825e-05, + -1.2214e-01, 9.5303e-02], + [-1.4077e-01, -1.2144e-01, 1.1287e-01, ..., 2.3395e-01, + -8.7315e-02, -2.4396e-01]], device='cuda:0'), grad: tensor([[ 5.7742e-08, -8.0094e-08, -1.7695e-08, ..., -9.3132e-10, + 6.5193e-09, 0.0000e+00], + [-8.1956e-08, -3.4459e-08, -6.0536e-08, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-09, 4.6566e-09, -2.2352e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.0245e-08, 7.4506e-09, 9.3132e-09, ..., 0.0000e+00, + 9.3132e-10, -9.3132e-10], + [ 0.0000e+00, 2.7008e-08, 3.0734e-08, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0319, 0.0054, 0.0205, 0.0156, 0.0402, 0.0146, -0.0008, 0.0022, + -0.0180, -0.0530], device='cuda:0'), grad: tensor([-1.3225e-07, -1.4715e-07, 9.3132e-09, 2.0489e-08, -7.4506e-09, + 1.3970e-08, 1.8440e-07, -7.9162e-08, 2.9802e-08, 1.1828e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 218.08, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4312 re_mapping 0.0029 re_causal 0.0102 /// teacc 99.11 lr 0.00001000 +Epoch 452, weight, value: tensor([[-1.8105e-01, 2.6321e-01, -2.8381e-02, ..., -2.6872e-02, + 6.1328e-02, -4.8362e-02], + [ 2.1623e-01, -1.2134e-01, 3.1103e-02, ..., -1.2771e-01, + 2.7895e-02, -1.0009e-01], + [-1.0426e-01, -3.1647e-01, -6.5689e-02, ..., -7.7342e-02, + 2.2870e-02, -3.1939e-01], + ..., + [-2.0561e-01, -3.7917e-01, 7.0789e-02, ..., -1.2683e-01, + -1.0426e-01, -2.1669e-01], + [-1.7198e-01, 3.2934e-03, 4.8475e-02, ..., 6.6844e-05, + -1.2213e-01, 9.5382e-02], + [-1.4079e-01, -1.2166e-01, 1.1289e-01, ..., 2.3397e-01, + -8.7330e-02, -2.4402e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.7253e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.0489e-08, 0.0000e+00, 4.3772e-08, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 9.3132e-10], + ..., + [ 5.5879e-09, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-3.6322e-08, 3.7253e-09, -8.2888e-08, ..., -9.3132e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0319, 0.0054, 0.0205, 0.0156, 0.0402, 0.0147, -0.0009, 0.0022, + -0.0180, -0.0530], device='cuda:0'), grad: tensor([-5.5879e-09, 1.0338e-07, -2.0489e-08, 2.5146e-08, 4.5635e-08, + -4.0047e-08, 2.5146e-08, 5.3085e-08, 3.7253e-09, -1.8813e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 217.87, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4437 re_mapping 0.0029 re_causal 0.0105 /// teacc 99.10 lr 0.00001000 +Epoch 453, weight, value: tensor([[-1.8102e-01, 2.6356e-01, -2.8447e-02, ..., -2.6888e-02, + 6.1320e-02, -4.8364e-02], + [ 2.1627e-01, -1.2133e-01, 3.1119e-02, ..., -1.2770e-01, + 2.7928e-02, -1.0007e-01], + [-1.0434e-01, -3.1657e-01, -6.5781e-02, ..., -7.7347e-02, + 2.2880e-02, -3.1946e-01], + ..., + [-2.0562e-01, -3.7926e-01, 7.0791e-02, ..., -1.2684e-01, + -1.0426e-01, -2.1681e-01], + [-1.7203e-01, 3.3150e-03, 4.8501e-02, ..., 5.5204e-05, + -1.2214e-01, 9.5425e-02], + [-1.4081e-01, -1.2190e-01, 1.1298e-01, ..., 2.3400e-01, + -8.7328e-02, -2.4408e-01]], device='cuda:0'), grad: tensor([[ 1.4901e-08, -2.9802e-08, -1.2107e-08, ..., 0.0000e+00, + -4.6566e-09, 3.2596e-08], + [ 9.3132e-09, 2.9802e-08, 1.4901e-08, ..., 0.0000e+00, + 2.7940e-09, 1.9558e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 9.3132e-10, 3.0734e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 5.6811e-08, 4.6566e-08, 1.7695e-08, ..., 0.0000e+00, + 1.8626e-09, 1.1642e-07], + [ 1.8626e-09, 2.7940e-09, -4.0326e-07, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0321, 0.0054, 0.0205, 0.0156, 0.0404, 0.0147, -0.0010, 0.0021, + -0.0180, -0.0530], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.3318e-07, -1.5832e-08, 3.1665e-08, 7.4971e-07, + 1.0803e-07, -8.9966e-07, 9.3132e-08, 5.3924e-07, -8.2329e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 218.00, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4014 re_mapping 0.0029 re_causal 0.0100 /// teacc 99.08 lr 0.00001000 +Epoch 454, weight, value: tensor([[-1.8099e-01, 2.6367e-01, -2.8424e-02, ..., -2.6947e-02, + 6.1315e-02, -4.8386e-02], + [ 2.1633e-01, -1.2133e-01, 3.1122e-02, ..., -1.2783e-01, + 2.7945e-02, -1.0008e-01], + [-1.0437e-01, -3.1660e-01, -6.5819e-02, ..., -7.7376e-02, + 2.2885e-02, -3.1953e-01], + ..., + [-2.0569e-01, -3.7931e-01, 7.0760e-02, ..., -1.2693e-01, + -1.0431e-01, -2.1695e-01], + [-1.7207e-01, 3.3794e-03, 4.8561e-02, ..., 5.0486e-05, + -1.2213e-01, 9.5512e-02], + [-1.4072e-01, -1.2199e-01, 1.1313e-01, ..., 2.3420e-01, + -8.7351e-02, -2.4414e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 2.3283e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 9.3132e-10, -5.0291e-08, ..., -1.8626e-09, + 0.0000e+00, 9.3132e-10], + [-9.3132e-10, -2.7940e-09, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, -4.6566e-09], + [ 0.0000e+00, 9.3132e-10, 1.3039e-08, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0321, 0.0054, 0.0205, 0.0156, 0.0404, 0.0146, -0.0010, 0.0021, + -0.0180, -0.0529], device='cuda:0'), grad: tensor([ 0.0000e+00, 4.0978e-08, 9.3132e-09, 2.2352e-08, 2.5146e-08, + -7.4506e-09, 9.3132e-10, -5.4017e-08, -1.1176e-08, -1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 218.07, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4210 re_mapping 0.0028 re_causal 0.0104 /// teacc 99.09 lr 0.00001000 +Epoch 455, weight, value: tensor([[-1.8099e-01, 2.6378e-01, -2.8394e-02, ..., -2.7037e-02, + 6.1305e-02, -4.8506e-02], + [ 2.1632e-01, -1.2134e-01, 3.1096e-02, ..., -1.2794e-01, + 2.7958e-02, -1.0009e-01], + [-1.0440e-01, -3.1666e-01, -6.5877e-02, ..., -7.7402e-02, + 2.2838e-02, -3.1965e-01], + ..., + [-2.0570e-01, -3.7938e-01, 7.0771e-02, ..., -1.2695e-01, + -1.0434e-01, -2.1708e-01], + [-1.7215e-01, 3.4327e-03, 4.8586e-02, ..., 5.3463e-05, + -1.2213e-01, 9.5582e-02], + [-1.4062e-01, -1.2212e-01, 1.1327e-01, ..., 2.3436e-01, + -8.7361e-02, -2.4418e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.2107e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + ..., + [ 4.6566e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 4.6566e-09, 1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0320, 0.0054, 0.0204, 0.0157, 0.0403, 0.0146, -0.0010, 0.0021, + -0.0180, -0.0528], device='cuda:0'), grad: tensor([-2.7940e-09, 5.3085e-08, 1.9558e-08, -2.9895e-07, -7.7300e-08, + 1.3690e-07, 4.6566e-09, 4.7497e-08, 1.4901e-08, 1.1548e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 217.70, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4436 re_mapping 0.0029 re_causal 0.0106 /// teacc 99.07 lr 0.00001000 +Epoch 456, weight, value: tensor([[-1.8103e-01, 2.6388e-01, -2.8371e-02, ..., -2.7045e-02, + 6.1292e-02, -4.8506e-02], + [ 2.1635e-01, -1.2133e-01, 3.1085e-02, ..., -1.2805e-01, + 2.8011e-02, -1.0008e-01], + [-1.0442e-01, -3.1670e-01, -6.5914e-02, ..., -7.7405e-02, + 2.2864e-02, -3.1971e-01], + ..., + [-2.0573e-01, -3.7944e-01, 7.0782e-02, ..., -1.2695e-01, + -1.0441e-01, -2.1714e-01], + [-1.7230e-01, 3.4224e-03, 4.8601e-02, ..., 5.0852e-05, + -1.2214e-01, 9.5647e-02], + [-1.4056e-01, -1.2225e-01, 1.1335e-01, ..., 2.3448e-01, + -8.7394e-02, -2.4422e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-09, 9.4995e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.8184e-08, 3.7253e-09, -2.1420e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-08, 0.0000e+00, 2.7008e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 9.3132e-10, 1.8347e-07, ..., 2.6077e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -9.3132e-10, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-09], + [ 5.5879e-09, 5.5879e-09, -2.9616e-07, ..., -2.7008e-08, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0321, 0.0054, 0.0205, 0.0157, 0.0404, 0.0146, -0.0010, 0.0021, + -0.0180, -0.0528], device='cuda:0'), grad: tensor([ 1.6112e-07, -5.0291e-08, 5.3085e-08, 5.5879e-09, 1.6764e-08, + -3.7253e-09, 1.3039e-08, 3.2876e-07, -1.6764e-08, -5.0757e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 217.92, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4389 re_mapping 0.0030 re_causal 0.0104 /// teacc 99.10 lr 0.00001000 +Epoch 457, weight, value: tensor([[-1.8109e-01, 2.6404e-01, -2.8466e-02, ..., -2.7105e-02, + 6.1287e-02, -4.8555e-02], + [ 2.1649e-01, -1.2135e-01, 3.1087e-02, ..., -1.2810e-01, + 2.8024e-02, -1.0010e-01], + [-1.0449e-01, -3.1681e-01, -6.5986e-02, ..., -7.7429e-02, + 2.2830e-02, -3.1978e-01], + ..., + [-2.0586e-01, -3.7953e-01, 7.0804e-02, ..., -1.2696e-01, + -1.0446e-01, -2.1724e-01], + [-1.7243e-01, 3.3917e-03, 4.8590e-02, ..., 4.9747e-05, + -1.2216e-01, 9.5608e-02], + [-1.4060e-01, -1.2236e-01, 1.1342e-01, ..., 2.3456e-01, + -8.7423e-02, -2.4431e-01]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -3.5390e-08, -9.3132e-10, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [-7.5437e-08, -4.6566e-09, -1.0896e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 3.7253e-08, 3.7253e-09, 5.5879e-08, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 2.1420e-08, 3.7253e-09, 3.0734e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 1.1176e-08, 2.7940e-09, 1.5832e-08, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0320, 0.0054, 0.0204, 0.0156, 0.0404, 0.0147, -0.0009, 0.0021, + -0.0181, -0.0528], device='cuda:0'), grad: tensor([-3.0734e-08, -1.7416e-07, 2.7940e-09, 5.5879e-09, 7.6368e-08, + -5.5879e-09, -3.9116e-08, 9.4995e-08, 5.4948e-08, 2.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 217.90, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4444 re_mapping 0.0029 re_causal 0.0103 /// teacc 99.09 lr 0.00001000 +Epoch 458, weight, value: tensor([[-1.8098e-01, 2.6447e-01, -2.8423e-02, ..., -2.7099e-02, + 6.1287e-02, -4.8648e-02], + [ 2.1663e-01, -1.2137e-01, 3.1166e-02, ..., -1.2810e-01, + 2.8038e-02, -1.0014e-01], + [-1.0455e-01, -3.1695e-01, -6.6081e-02, ..., -7.7433e-02, + 2.2835e-02, -3.1991e-01], + ..., + [-2.0599e-01, -3.7965e-01, 7.0744e-02, ..., -1.2702e-01, + -1.0448e-01, -2.1740e-01], + [-1.7254e-01, 3.4986e-03, 4.8708e-02, ..., 5.0421e-05, + -1.2215e-01, 9.5712e-02], + [-1.4066e-01, -1.2263e-01, 1.1347e-01, ..., 2.3460e-01, + -8.7447e-02, -2.4451e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.2352e-08, -1.6764e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, -9.3132e-10], + [ 0.0000e+00, 8.3819e-09, -6.5193e-09, ..., -2.7940e-09, + 0.0000e+00, -1.0245e-08]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0321, 0.0055, 0.0204, 0.0156, 0.0405, 0.0147, -0.0010, 0.0019, + -0.0180, -0.0530], device='cuda:0'), grad: tensor([-6.4261e-08, 1.3970e-08, -5.5879e-09, 1.3970e-08, -1.0245e-07, + 1.4901e-08, 9.3132e-09, 1.8626e-08, 5.5879e-09, 1.0245e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 217.99, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4396 re_mapping 0.0029 re_causal 0.0103 /// teacc 99.10 lr 0.00001000 +Epoch 459, weight, value: tensor([[-1.8097e-01, 2.6476e-01, -2.8334e-02, ..., -2.7075e-02, + 6.1275e-02, -4.8710e-02], + [ 2.1667e-01, -1.2140e-01, 3.1176e-02, ..., -1.2810e-01, + 2.8071e-02, -1.0015e-01], + [-1.0457e-01, -3.1710e-01, -6.6150e-02, ..., -7.7441e-02, + 2.2824e-02, -3.2007e-01], + ..., + [-2.0602e-01, -3.7977e-01, 7.0748e-02, ..., -1.2706e-01, + -1.0453e-01, -2.1746e-01], + [-1.7269e-01, 3.4762e-03, 4.8815e-02, ..., 6.2958e-05, + -1.2217e-01, 9.5744e-02], + [-1.4069e-01, -1.2285e-01, 1.1347e-01, ..., 2.3463e-01, + -8.7540e-02, -2.4467e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0323, 0.0054, 0.0204, 0.0156, 0.0406, 0.0147, -0.0011, 0.0020, + -0.0180, -0.0531], device='cuda:0'), grad: tensor([ 5.4948e-08, 7.4506e-09, -2.9802e-08, -1.5832e-08, 2.7940e-09, + 1.7695e-08, -1.8161e-07, 4.0047e-08, 1.0524e-07, -9.3132e-10], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 218.02, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4330 re_mapping 0.0028 re_causal 0.0102 /// teacc 99.08 lr 0.00001000 +Epoch 460, weight, value: tensor([[-1.8101e-01, 2.6500e-01, -2.8268e-02, ..., -2.7098e-02, + 6.1258e-02, -4.8751e-02], + [ 2.1672e-01, -1.2144e-01, 3.1171e-02, ..., -1.2811e-01, + 2.8124e-02, -1.0017e-01], + [-1.0464e-01, -3.1726e-01, -6.6251e-02, ..., -7.7459e-02, + 2.2787e-02, -3.2016e-01], + ..., + [-2.0606e-01, -3.7984e-01, 7.0769e-02, ..., -1.2707e-01, + -1.0456e-01, -2.1755e-01], + [-1.7293e-01, 3.3944e-03, 4.8873e-02, ..., 5.6764e-05, + -1.2219e-01, 9.5680e-02], + [-1.4074e-01, -1.2307e-01, 1.1348e-01, ..., 2.3466e-01, + -8.7575e-02, -2.4488e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.3819e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 3.7253e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 9.3132e-10, 9.3132e-10, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0324, 0.0054, 0.0204, 0.0156, 0.0410, 0.0147, -0.0012, 0.0019, + -0.0182, -0.0532], device='cuda:0'), grad: tensor([-1.1176e-08, 8.3819e-09, -2.5146e-08, 1.2107e-08, -1.6764e-08, + -7.0781e-08, 6.9849e-08, 2.4214e-08, 1.2107e-08, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 218.24, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4600 re_mapping 0.0029 re_causal 0.0105 /// teacc 99.08 lr 0.00001000 +Epoch 461, weight, value: tensor([[-1.8107e-01, 2.6506e-01, -2.8235e-02, ..., -2.7103e-02, + 6.1251e-02, -4.8808e-02], + [ 2.1674e-01, -1.2146e-01, 3.1160e-02, ..., -1.2816e-01, + 2.8131e-02, -1.0019e-01], + [-1.0466e-01, -3.1734e-01, -6.6285e-02, ..., -7.7459e-02, + 2.2839e-02, -3.2024e-01], + ..., + [-2.0608e-01, -3.7989e-01, 7.0778e-02, ..., -1.2708e-01, + -1.0457e-01, -2.1763e-01], + [-1.7305e-01, 3.3812e-03, 4.8946e-02, ..., 5.6966e-05, + -1.2220e-01, 9.5697e-02], + [-1.4072e-01, -1.2318e-01, 1.1353e-01, ..., 2.3473e-01, + -8.7599e-02, -2.4504e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.4715e-07, -7.7300e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.9652e-08, 8.3819e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.6566e-09, -4.6566e-09, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, -1.6764e-08], + [ 0.0000e+00, 4.6566e-09, -3.5390e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0324, 0.0054, 0.0204, 0.0156, 0.0410, 0.0147, -0.0012, 0.0019, + -0.0182, -0.0533], device='cuda:0'), grad: tensor([-3.0175e-07, 3.0734e-08, 1.3039e-08, 8.3819e-09, 1.6764e-08, + 2.7940e-08, 6.8918e-08, 2.8033e-07, -5.7742e-08, -8.1025e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 218.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4252 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.08 lr 0.00001000 +Epoch 462, weight, value: tensor([[-1.8113e-01, 2.6524e-01, -2.8234e-02, ..., -2.7111e-02, + 6.1242e-02, -4.8780e-02], + [ 2.1690e-01, -1.2146e-01, 3.1235e-02, ..., -1.2826e-01, + 2.8219e-02, -1.0014e-01], + [-1.0468e-01, -3.1744e-01, -6.6354e-02, ..., -7.7460e-02, + 2.2847e-02, -3.2033e-01], + ..., + [-2.0624e-01, -3.8001e-01, 7.0712e-02, ..., -1.2714e-01, + -1.0462e-01, -2.1770e-01], + [-1.7316e-01, 3.4134e-03, 4.9021e-02, ..., 4.7424e-05, + -1.2221e-01, 9.5756e-02], + [-1.4069e-01, -1.2329e-01, 1.1364e-01, ..., 2.3487e-01, + -8.7626e-02, -2.4510e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.2352e-08, -1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 1.2107e-08, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0325, 0.0055, 0.0205, 0.0155, 0.0411, 0.0148, -0.0013, 0.0018, + -0.0182, -0.0533], device='cuda:0'), grad: tensor([-5.3085e-08, 1.3039e-08, 3.2596e-08, -5.7742e-08, -5.5879e-09, + 2.0489e-08, 0.0000e+00, -9.3132e-09, 9.3132e-10, 5.8673e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 217.94, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4247 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.10 lr 0.00001000 +Epoch 463, weight, value: tensor([[-1.8113e-01, 2.6548e-01, -2.8109e-02, ..., -2.7112e-02, + 6.1227e-02, -4.8853e-02], + [ 2.1693e-01, -1.2147e-01, 3.1245e-02, ..., -1.2830e-01, + 2.8215e-02, -1.0015e-01], + [-1.0471e-01, -3.1757e-01, -6.6460e-02, ..., -7.7465e-02, + 2.2845e-02, -3.2044e-01], + ..., + [-2.0626e-01, -3.8011e-01, 7.0723e-02, ..., -1.2715e-01, + -1.0465e-01, -2.1782e-01], + [-1.7323e-01, 3.4137e-03, 4.8998e-02, ..., 3.3310e-05, + -1.2221e-01, 9.5745e-02], + [-1.4072e-01, -1.2355e-01, 1.1363e-01, ..., 2.3491e-01, + -8.7607e-02, -2.4521e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.7253e-09, 1.1362e-07, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-5.5879e-09, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -2.3283e-08, ..., 0.0000e+00, + 5.5879e-09, 9.3132e-10], + [ 6.5193e-09, 9.3132e-10, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 4.0978e-08, 2.6077e-08, -1.2945e-07, ..., 0.0000e+00, + 1.5832e-08, 2.5146e-08]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0326, 0.0054, 0.0204, 0.0155, 0.0413, 0.0148, -0.0013, 0.0018, + -0.0182, -0.0535], device='cuda:0'), grad: tensor([ 2.0582e-07, 6.7987e-08, -1.1455e-07, 1.0524e-07, -5.2992e-07, + -1.4063e-07, 7.4506e-09, 1.3132e-07, 3.0734e-08, 2.4214e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 218.26, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4375 re_mapping 0.0028 re_causal 0.0102 /// teacc 99.08 lr 0.00001000 +Epoch 464, weight, value: tensor([[-1.8116e-01, 2.6567e-01, -2.8194e-02, ..., -2.7285e-02, + 6.1213e-02, -4.8928e-02], + [ 2.1699e-01, -1.2146e-01, 3.1283e-02, ..., -1.2830e-01, + 2.8288e-02, -1.0010e-01], + [-1.0468e-01, -3.1765e-01, -6.6430e-02, ..., -7.7507e-02, + 2.2826e-02, -3.2054e-01], + ..., + [-2.0630e-01, -3.8023e-01, 7.0728e-02, ..., -1.2717e-01, + -1.0466e-01, -2.1791e-01], + [-1.7333e-01, 3.3981e-03, 4.9026e-02, ..., 2.8039e-05, + -1.2222e-01, 9.5689e-02], + [-1.4080e-01, -1.2366e-01, 1.1364e-01, ..., 2.3499e-01, + -8.7633e-02, -2.4543e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.6566e-09, 1.0431e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -2.3004e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, -1.8626e-09, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, -3.7253e-09], + [ 0.0000e+00, 2.7940e-09, 8.8476e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0326, 0.0054, 0.0205, 0.0155, 0.0414, 0.0148, -0.0014, 0.0018, + -0.0183, -0.0536], device='cuda:0'), grad: tensor([ 1.7229e-07, 1.3970e-08, -3.0734e-08, 8.9407e-08, 1.4901e-08, + 6.5193e-09, 4.6566e-09, -7.9442e-07, -1.1176e-08, 5.3644e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 217.80, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4534 re_mapping 0.0028 re_causal 0.0103 /// teacc 99.10 lr 0.00001000 +Epoch 465, weight, value: tensor([[-1.8118e-01, 2.6587e-01, -2.8101e-02, ..., -2.7332e-02, + 6.1207e-02, -4.8901e-02], + [ 2.1707e-01, -1.2153e-01, 3.1263e-02, ..., -1.2831e-01, + 2.8352e-02, -1.0009e-01], + [-1.0470e-01, -3.1773e-01, -6.6398e-02, ..., -7.7525e-02, + 2.2856e-02, -3.2059e-01], + ..., + [-2.0636e-01, -3.8034e-01, 7.0775e-02, ..., -1.2718e-01, + -1.0467e-01, -2.1801e-01], + [-1.7339e-01, 3.4008e-03, 4.9001e-02, ..., 1.0866e-05, + -1.2223e-01, 9.5719e-02], + [-1.4082e-01, -1.2383e-01, 1.1366e-01, ..., 2.3505e-01, + -8.7690e-02, -2.4559e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-3.7253e-09, -4.6566e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 4.6566e-09, -3.3528e-08, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0327, 0.0054, 0.0206, 0.0156, 0.0414, 0.0147, -0.0014, 0.0018, + -0.0184, -0.0537], device='cuda:0'), grad: tensor([ 1.1362e-07, 4.5635e-08, -5.1036e-07, 4.6566e-08, 5.8673e-08, + -2.7940e-09, 1.0245e-08, 3.0547e-07, 2.0489e-08, -9.2201e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 218.01, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4245 re_mapping 0.0028 re_causal 0.0102 /// teacc 99.09 lr 0.00001000 +Epoch 466, weight, value: tensor([[-1.8120e-01, 2.6601e-01, -2.8033e-02, ..., -2.7338e-02, + 6.1194e-02, -4.8871e-02], + [ 2.1731e-01, -1.2155e-01, 3.1351e-02, ..., -1.2832e-01, + 2.8383e-02, -1.0011e-01], + [-1.0478e-01, -3.1780e-01, -6.6440e-02, ..., -7.7527e-02, + 2.2866e-02, -3.2065e-01], + ..., + [-2.0659e-01, -3.8043e-01, 7.0707e-02, ..., -1.2720e-01, + -1.0478e-01, -2.1815e-01], + [-1.7353e-01, 3.3464e-03, 4.9023e-02, ..., 1.5210e-05, + -1.2225e-01, 9.5707e-02], + [-1.4087e-01, -1.2396e-01, 1.1367e-01, ..., 2.3509e-01, + -8.7715e-02, -2.4565e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [-5.5879e-09, 2.7940e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 5.5879e-09, 9.3132e-10, -1.3039e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, -1.6764e-08, -1.4901e-08, ..., 0.0000e+00, + -9.3132e-10, -1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0328, 0.0056, 0.0207, 0.0155, 0.0414, 0.0148, -0.0015, 0.0017, + -0.0185, -0.0538], device='cuda:0'), grad: tensor([ 8.3819e-09, 2.7940e-09, 1.0245e-08, 5.1223e-08, 1.8626e-09, + 2.0489e-08, -2.7940e-09, -4.4703e-08, -6.4261e-08, 2.4214e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4186 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.07 lr 0.00001000 +Epoch 467, weight, value: tensor([[-1.8118e-01, 2.6612e-01, -2.7979e-02, ..., -2.7329e-02, + 6.1193e-02, -4.8880e-02], + [ 2.1736e-01, -1.2156e-01, 3.1366e-02, ..., -1.2831e-01, + 2.8424e-02, -1.0014e-01], + [-1.0474e-01, -3.1785e-01, -6.6428e-02, ..., -7.7530e-02, + 2.2878e-02, -3.2072e-01], + ..., + [-2.0664e-01, -3.8047e-01, 7.0717e-02, ..., -1.2723e-01, + -1.0492e-01, -2.1832e-01], + [-1.7361e-01, 3.3412e-03, 4.9043e-02, ..., 1.2941e-05, + -1.2226e-01, 9.5714e-02], + [-1.4093e-01, -1.2412e-01, 1.1365e-01, ..., 2.3510e-01, + -8.7772e-02, -2.4576e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.6484e-07, -1.4808e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.2107e-08, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -6.8918e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.5553e-07, 1.7229e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0328, 0.0055, 0.0208, 0.0156, 0.0414, 0.0147, -0.0014, 0.0016, + -0.0185, -0.0539], device='cuda:0'), grad: tensor([-4.4145e-07, 4.9360e-08, 3.1665e-08, 1.2107e-08, -1.5832e-08, + 3.7253e-09, 2.7008e-08, -2.2817e-07, 9.3132e-09, 5.5414e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 218.21, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4107 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.09 lr 0.00001000 +Epoch 468, weight, value: tensor([[-1.8122e-01, 2.6619e-01, -2.7957e-02, ..., -2.7337e-02, + 6.1171e-02, -4.8935e-02], + [ 2.1754e-01, -1.2158e-01, 3.1378e-02, ..., -1.2832e-01, + 2.8444e-02, -1.0018e-01], + [-1.0485e-01, -3.1786e-01, -6.6457e-02, ..., -7.7531e-02, + 2.2862e-02, -3.2078e-01], + ..., + [-2.0680e-01, -3.8053e-01, 7.0838e-02, ..., -1.2723e-01, + -1.0500e-01, -2.1840e-01], + [-1.7369e-01, 3.3612e-03, 4.9114e-02, ..., 8.3337e-06, + -1.2227e-01, 9.5743e-02], + [-1.4100e-01, -1.2423e-01, 1.1345e-01, ..., 2.3512e-01, + -8.7802e-02, -2.4583e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.2352e-08, 2.1420e-08, ..., 1.0710e-07, + 0.0000e+00, 0.0000e+00], + [ 5.2992e-07, 0.0000e+00, 1.7034e-06, ..., 3.8184e-08, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 4.6566e-09, ..., 1.3039e-08, + 0.0000e+00, 0.0000e+00], + ..., + [-5.3458e-07, 1.8626e-09, -1.7164e-06, ..., -2.8871e-08, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 9.3132e-09, ..., 2.7940e-09, + 0.0000e+00, -1.8626e-09], + [ 9.3132e-10, -5.2154e-08, -4.3772e-08, ..., -2.3469e-07, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0328, 0.0056, 0.0208, 0.0155, 0.0415, 0.0147, -0.0014, 0.0017, + -0.0186, -0.0542], device='cuda:0'), grad: tensor([ 4.0699e-07, 2.6152e-06, 5.5879e-08, 2.6450e-07, 4.9360e-08, + 8.3819e-08, 9.3132e-10, -2.5965e-06, 2.1420e-08, -8.9034e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 217.95, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4538 re_mapping 0.0027 re_causal 0.0101 /// teacc 99.09 lr 0.00001000 +Epoch 469, weight, value: tensor([[-1.8123e-01, 2.6642e-01, -2.7927e-02, ..., -2.7397e-02, + 6.1171e-02, -4.8970e-02], + [ 2.1769e-01, -1.2161e-01, 3.1410e-02, ..., -1.2837e-01, + 2.8438e-02, -1.0022e-01], + [-1.0492e-01, -3.1791e-01, -6.6518e-02, ..., -7.7547e-02, + 2.2839e-02, -3.2088e-01], + ..., + [-2.0694e-01, -3.8061e-01, 7.0828e-02, ..., -1.2724e-01, + -1.0502e-01, -2.1853e-01], + [-1.7372e-01, 3.3848e-03, 4.9125e-02, ..., 1.1987e-05, + -1.2227e-01, 9.5762e-02], + [-1.4100e-01, -1.2435e-01, 1.1349e-01, ..., 2.3520e-01, + -8.7813e-02, -2.4597e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-6.5193e-09, 0.0000e+00, -1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 5.5879e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-09], + ..., + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 1.9558e-08, 1.0245e-08, 7.4506e-09, ..., 0.0000e+00, + 4.6566e-09, 1.6764e-08], + [ 1.8626e-09, 9.3132e-10, -4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0329, 0.0056, 0.0208, 0.0156, 0.0415, 0.0147, -0.0014, 0.0016, + -0.0186, -0.0542], device='cuda:0'), grad: tensor([ 8.3819e-09, 2.4587e-07, 2.2352e-08, 2.5705e-07, -3.1572e-07, + -3.6228e-07, 3.8184e-08, 6.3330e-08, 4.1910e-08, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 217.90, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4154 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.09 lr 0.00001000 +Epoch 470, weight, value: tensor([[-1.8127e-01, 2.6649e-01, -2.8042e-02, ..., -2.7467e-02, + 6.1171e-02, -4.9120e-02], + [ 2.1782e-01, -1.2171e-01, 3.1431e-02, ..., -1.2836e-01, + 2.8423e-02, -1.0028e-01], + [-1.0496e-01, -3.1795e-01, -6.6551e-02, ..., -7.7549e-02, + 2.2817e-02, -3.2103e-01], + ..., + [-2.0708e-01, -3.8065e-01, 7.0832e-02, ..., -1.2727e-01, + -1.0503e-01, -2.1872e-01], + [-1.7382e-01, 3.3771e-03, 4.9185e-02, ..., 1.6264e-05, + -1.2229e-01, 9.5736e-02], + [-1.4104e-01, -1.2442e-01, 1.1355e-01, ..., 2.3524e-01, + -8.7834e-02, -2.4609e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -2.4214e-08, -1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 1.7695e-08, 1.5832e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 3.7253e-09, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 3.7253e-09, 7.4506e-09, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0329, 0.0057, 0.0208, 0.0156, 0.0415, 0.0147, -0.0014, 0.0016, + -0.0186, -0.0542], device='cuda:0'), grad: tensor([-4.1910e-08, 4.0978e-08, 4.6566e-09, 6.8825e-07, 1.8626e-09, + -7.1619e-07, 1.6764e-08, -2.2352e-08, 2.7940e-09, 2.5146e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 217.68, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4267 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.10 lr 0.00001000 +Epoch 471, weight, value: tensor([[-1.8127e-01, 2.6673e-01, -2.8095e-02, ..., -2.7528e-02, + 6.1214e-02, -4.9180e-02], + [ 2.1789e-01, -1.2191e-01, 3.1431e-02, ..., -1.2837e-01, + 2.8422e-02, -1.0030e-01], + [-1.0501e-01, -3.1814e-01, -6.6660e-02, ..., -7.7551e-02, + 2.2823e-02, -3.2115e-01], + ..., + [-2.0714e-01, -3.8079e-01, 7.0846e-02, ..., -1.2727e-01, + -1.0506e-01, -2.1884e-01], + [-1.7389e-01, 3.4076e-03, 4.9241e-02, ..., 6.7323e-06, + -1.2227e-01, 9.5832e-02], + [-1.4107e-01, -1.2453e-01, 1.1366e-01, ..., 2.3528e-01, + -8.7895e-02, -2.4615e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.5193e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.7940e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.8626e-08, 3.0734e-08, 2.6077e-08, ..., 0.0000e+00, + 0.0000e+00, 3.2596e-08], + ..., + [ 0.0000e+00, 1.8626e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [-3.1665e-08, -5.3085e-08, -4.5635e-08, ..., -9.3132e-10, + 0.0000e+00, -5.3085e-08], + [ 4.6566e-09, 5.5879e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0329, 0.0056, 0.0207, 0.0156, 0.0416, 0.0148, -0.0013, 0.0015, + -0.0186, -0.0543], device='cuda:0'), grad: tensor([-1.1176e-08, 1.8626e-08, 1.2014e-07, 2.7008e-08, -1.0431e-07, + 1.3970e-08, 9.3132e-09, 9.8720e-08, -1.9278e-07, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 218.02, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4389 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.09 lr 0.00001000 +Epoch 472, weight, value: tensor([[-1.8130e-01, 2.6688e-01, -2.8044e-02, ..., -2.7592e-02, + 6.1198e-02, -4.9284e-02], + [ 2.1811e-01, -1.2193e-01, 3.1509e-02, ..., -1.2837e-01, + 2.8426e-02, -1.0035e-01], + [-1.0511e-01, -3.1822e-01, -6.6740e-02, ..., -7.7564e-02, + 2.2850e-02, -3.2128e-01], + ..., + [-2.0736e-01, -3.8096e-01, 7.0789e-02, ..., -1.2729e-01, + -1.0509e-01, -2.1899e-01], + [-1.7396e-01, 3.4676e-03, 4.9345e-02, ..., 7.7367e-06, + -1.2228e-01, 9.5841e-02], + [-1.4112e-01, -1.2467e-01, 1.1367e-01, ..., 2.3533e-01, + -8.7935e-02, -2.4640e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.2107e-08, 0.0000e+00, 2.7008e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 4.6566e-09, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.4214e-08, 1.8626e-09, -5.6811e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0328, 0.0057, 0.0207, 0.0156, 0.0417, 0.0147, -0.0012, 0.0014, + -0.0187, -0.0544], device='cuda:0'), grad: tensor([-2.7940e-09, 6.2399e-08, -5.3085e-08, 6.5193e-09, 3.9116e-08, + 5.5879e-09, -6.5193e-09, 7.4506e-08, 1.8626e-09, -1.2945e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 217.71, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4483 re_mapping 0.0028 re_causal 0.0101 /// teacc 99.10 lr 0.00001000 +Epoch 473, weight, value: tensor([[-1.8139e-01, 2.6716e-01, -2.8221e-02, ..., -2.7727e-02, + 6.1172e-02, -4.9310e-02], + [ 2.1828e-01, -1.2193e-01, 3.1616e-02, ..., -1.2837e-01, + 2.8445e-02, -1.0040e-01], + [-1.0522e-01, -3.1827e-01, -6.6860e-02, ..., -7.7566e-02, + 2.2833e-02, -3.2138e-01], + ..., + [-2.0751e-01, -3.8107e-01, 7.0699e-02, ..., -1.2733e-01, + -1.0513e-01, -2.1920e-01], + [-1.7404e-01, 3.6216e-03, 4.9581e-02, ..., 5.8138e-06, + -1.2225e-01, 9.6017e-02], + [-1.4120e-01, -1.2491e-01, 1.1383e-01, ..., 2.3541e-01, + -8.7939e-02, -2.4652e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, -4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0329, 0.0059, 0.0207, 0.0156, 0.0418, 0.0147, -0.0013, 0.0012, + -0.0186, -0.0543], device='cuda:0'), grad: tensor([ 1.0245e-08, 8.3819e-09, 9.3132e-10, 5.5879e-09, 3.9116e-08, + -2.7940e-09, -2.6077e-08, -2.5146e-08, 2.7940e-09, -8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 217.85, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4179 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 474, weight, value: tensor([[-1.8136e-01, 2.6750e-01, -2.8088e-02, ..., -2.7728e-02, + 6.1183e-02, -4.9336e-02], + [ 2.1832e-01, -1.2211e-01, 3.1587e-02, ..., -1.2838e-01, + 2.8436e-02, -1.0044e-01], + [-1.0532e-01, -3.1845e-01, -6.6953e-02, ..., -7.7563e-02, + 2.2807e-02, -3.2146e-01], + ..., + [-2.0753e-01, -3.8118e-01, 7.0813e-02, ..., -1.2737e-01, + -1.0514e-01, -2.1927e-01], + [-1.7412e-01, 3.6151e-03, 4.9553e-02, ..., 9.0850e-06, + -1.2225e-01, 9.6053e-02], + [-1.4128e-01, -1.2523e-01, 1.1374e-01, ..., 2.3544e-01, + -8.7978e-02, -2.4670e-01]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -5.8673e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -1.3039e-08], + [ 4.4145e-07, 0.0000e+00, 2.2817e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.7404e-07, 0.0000e+00, -2.3935e-07, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 3.7253e-09, 0.0000e+00, -2.1420e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.1176e-08, 5.5879e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 5.5879e-09, 0.0000e+00, 1.7695e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0331, 0.0058, 0.0207, 0.0156, 0.0418, 0.0148, -0.0014, 0.0013, + -0.0186, -0.0545], device='cuda:0'), grad: tensor([-1.1176e-07, 2.0433e-06, -2.2184e-06, 2.7008e-08, 1.8626e-08, + 3.1665e-08, 1.0338e-07, 5.5879e-09, 4.0978e-08, 6.1467e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 217.73, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4351 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.10 lr 0.00001000 +Epoch 475, weight, value: tensor([[-1.8145e-01, 2.6773e-01, -2.8023e-02, ..., -2.7736e-02, + 6.1158e-02, -4.9354e-02], + [ 2.1837e-01, -1.2213e-01, 3.1614e-02, ..., -1.2840e-01, + 2.8452e-02, -1.0046e-01], + [-1.0526e-01, -3.1852e-01, -6.6933e-02, ..., -7.7567e-02, + 2.2800e-02, -3.2153e-01], + ..., + [-2.0758e-01, -3.8137e-01, 7.0832e-02, ..., -1.2741e-01, + -1.0517e-01, -2.1943e-01], + [-1.7426e-01, 3.7116e-03, 4.9602e-02, ..., -1.6000e-05, + -1.2225e-01, 9.6219e-02], + [-1.4130e-01, -1.2540e-01, 1.1373e-01, ..., 2.3550e-01, + -8.8026e-02, -2.4677e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [-1.1921e-06, 1.0245e-08, -1.0077e-06, ..., 0.0000e+00, + 2.3283e-09, 8.8476e-09], + [ 2.8871e-08, 4.6566e-10, 2.2352e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.1548e-06, 6.5193e-09, 9.6858e-07, ..., 0.0000e+00, + 1.3970e-09, 5.5879e-09], + [ 4.6566e-10, -3.4459e-08, -4.4238e-08, ..., 0.0000e+00, + -9.7789e-09, -2.2817e-08], + [ 2.3283e-09, 9.3132e-10, 8.8476e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 475, bias, value: tensor([ 0.0332, 0.0057, 0.0208, 0.0156, 0.0419, 0.0148, -0.0015, 0.0013, + -0.0186, -0.0546], device='cuda:0'), grad: tensor([ 1.0710e-08, -1.7872e-06, 6.7055e-08, 6.8918e-08, -1.8626e-09, + 3.8650e-08, -1.0710e-08, 1.6652e-06, -7.9162e-08, 2.8871e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 218.22, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4249 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.08 lr 0.00001000 +Epoch 476, weight, value: tensor([[-1.8147e-01, 2.6789e-01, -2.7963e-02, ..., -2.7743e-02, + 6.1128e-02, -4.9390e-02], + [ 2.1845e-01, -1.2227e-01, 3.1659e-02, ..., -1.2840e-01, + 2.8480e-02, -1.0053e-01], + [-1.0527e-01, -3.1858e-01, -6.6996e-02, ..., -7.7573e-02, + 2.2791e-02, -3.2161e-01], + ..., + [-2.0764e-01, -3.8146e-01, 7.0829e-02, ..., -1.2742e-01, + -1.0529e-01, -2.1959e-01], + [-1.7434e-01, 3.7037e-03, 4.9630e-02, ..., -4.3529e-05, + -1.2225e-01, 9.6207e-02], + [-1.4138e-01, -1.2559e-01, 1.1373e-01, ..., 2.3554e-01, + -8.8016e-02, -2.4685e-01]], device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.3970e-08, -2.3283e-09, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [-1.8068e-07, 3.2596e-09, 4.6566e-10, ..., 0.0000e+00, + 9.3132e-10, 2.7940e-09], + [ 1.2806e-07, 4.6566e-10, 1.3970e-09, ..., 0.0000e+00, + 4.6566e-10, 9.3132e-10], + ..., + [ 4.6100e-08, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 6.0536e-09, 3.2596e-09, -3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 6.2399e-08, 4.4238e-08, -5.1223e-09, ..., 0.0000e+00, + 0.0000e+00, 4.3772e-08]], device='cuda:0') +Epoch 476, bias, value: tensor([ 0.0333, 0.0057, 0.0209, 0.0156, 0.0418, 0.0148, -0.0014, 0.0013, + -0.0187, -0.0547], device='cuda:0'), grad: tensor([ 3.4599e-07, -1.9278e-07, 1.5600e-07, 1.8347e-07, 1.6298e-08, + -2.5053e-07, -4.0838e-07, 5.7276e-08, 2.1886e-08, 8.9873e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 217.97, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4622 re_mapping 0.0028 re_causal 0.0103 /// teacc 99.09 lr 0.00001000 +Epoch 477, weight, value: tensor([[-1.8163e-01, 2.6804e-01, -2.7906e-02, ..., -2.7736e-02, + 6.1092e-02, -4.9451e-02], + [ 2.1861e-01, -1.2227e-01, 3.1626e-02, ..., -1.2840e-01, + 2.8478e-02, -1.0059e-01], + [-1.0529e-01, -3.1867e-01, -6.7057e-02, ..., -7.7579e-02, + 2.2741e-02, -3.2170e-01], + ..., + [-2.0780e-01, -3.8165e-01, 7.0932e-02, ..., -1.2742e-01, + -1.0533e-01, -2.1980e-01], + [-1.7449e-01, 3.7787e-03, 4.9758e-02, ..., -3.3800e-05, + -1.2227e-01, 9.6285e-02], + [-1.4148e-01, -1.2596e-01, 1.1365e-01, ..., 2.3555e-01, + -8.8039e-02, -2.4706e-01]], device='cuda:0'), grad: tensor([[ 4.6566e-09, -9.3132e-09, -1.2573e-08, ..., 0.0000e+00, + -4.6566e-10, 3.7253e-09], + [ 2.7940e-09, 6.0536e-09, 3.2596e-09, ..., 0.0000e+00, + 4.6566e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 3.7253e-09, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 5.2154e-08], + [ 1.2107e-07, 1.1874e-07, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 1.0664e-07], + [ 5.5879e-09, 1.1642e-08, 3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, 6.9849e-09]], device='cuda:0') +Epoch 477, bias, value: tensor([ 0.0334, 0.0057, 0.0210, 0.0155, 0.0419, 0.0148, -0.0014, 0.0013, + -0.0187, -0.0549], device='cuda:0'), grad: tensor([ 4.3698e-06, 2.7008e-08, -1.8626e-09, -6.4727e-08, -6.9849e-09, + 1.6997e-07, -5.1595e-06, 1.6671e-07, 4.6240e-07, 3.1199e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 217.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4298 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.09 lr 0.00001000 +Epoch 478, weight, value: tensor([[-1.8166e-01, 2.6819e-01, -2.7922e-02, ..., -2.7785e-02, + 6.1080e-02, -4.9506e-02], + [ 2.1872e-01, -1.2228e-01, 3.1622e-02, ..., -1.2840e-01, + 2.8549e-02, -1.0061e-01], + [-1.0535e-01, -3.1873e-01, -6.7114e-02, ..., -7.7579e-02, + 2.2760e-02, -3.2185e-01], + ..., + [-2.0788e-01, -3.8173e-01, 7.0969e-02, ..., -1.2747e-01, + -1.0536e-01, -2.2001e-01], + [-1.7464e-01, 3.7089e-03, 4.9778e-02, ..., -4.1910e-05, + -1.2230e-01, 9.6243e-02], + [-1.4155e-01, -1.2612e-01, 1.1371e-01, ..., 2.3560e-01, + -8.8085e-02, -2.4715e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -6.5193e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 4.6566e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 478, bias, value: tensor([ 0.0334, 0.0057, 0.0209, 0.0155, 0.0419, 0.0148, -0.0014, 0.0013, + -0.0188, -0.0549], device='cuda:0'), grad: tensor([ 9.3132e-10, 0.0000e+00, 2.7940e-09, 2.1420e-08, 2.7940e-09, + -4.9360e-08, 1.2107e-08, 0.0000e+00, 6.5193e-09, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 218.03, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4055 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.08 lr 0.00001000 +Epoch 479, weight, value: tensor([[-1.8166e-01, 2.6830e-01, -2.7890e-02, ..., -2.7795e-02, + 6.1061e-02, -4.9559e-02], + [ 2.1885e-01, -1.2227e-01, 3.1628e-02, ..., -1.2849e-01, + 2.8692e-02, -1.0053e-01], + [-1.0542e-01, -3.1876e-01, -6.7175e-02, ..., -7.7579e-02, + 2.2747e-02, -3.2196e-01], + ..., + [-2.0798e-01, -3.8179e-01, 7.1002e-02, ..., -1.2748e-01, + -1.0546e-01, -2.2023e-01], + [-1.7483e-01, 3.6618e-03, 4.9827e-02, ..., -3.7798e-05, + -1.2234e-01, 9.6245e-02], + [-1.4157e-01, -1.2630e-01, 1.1375e-01, ..., 2.3570e-01, + -8.8109e-02, -2.4723e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 4.6566e-09, ..., 1.8626e-09, + 0.0000e+00, 2.7940e-08], + [ 0.0000e+00, 1.8626e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + [ 0.0000e+00, 4.6566e-09, 5.5879e-09, ..., 2.7940e-09, + 0.0000e+00, 5.5879e-09], + ..., + [ 9.3132e-10, 1.8626e-09, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 8.3819e-09], + [ 0.0000e+00, 4.9360e-08, 4.3772e-08, ..., 3.4459e-08, + 0.0000e+00, -1.2107e-08], + [ 0.0000e+00, -1.2293e-07, -1.3597e-07, ..., -7.6368e-08, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 479, bias, value: tensor([ 0.0334, 0.0057, 0.0209, 0.0155, 0.0419, 0.0148, -0.0014, 0.0013, + -0.0188, -0.0549], device='cuda:0'), grad: tensor([ 1.0058e-07, 4.4703e-08, 7.4506e-09, -1.3411e-07, 3.2596e-08, + 2.0768e-07, 5.5879e-09, 2.1420e-08, 2.3097e-07, -5.1688e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 217.89, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4101 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.10 lr 0.00001000 +Epoch 480, weight, value: tensor([[-1.8173e-01, 2.6854e-01, -2.7797e-02, ..., -2.7794e-02, + 6.1032e-02, -4.9658e-02], + [ 2.1893e-01, -1.2230e-01, 3.1646e-02, ..., -1.2850e-01, + 2.8716e-02, -1.0057e-01], + [-1.0542e-01, -3.1885e-01, -6.7292e-02, ..., -7.7582e-02, + 2.2730e-02, -3.2203e-01], + ..., + [-2.0804e-01, -3.8195e-01, 7.1021e-02, ..., -1.2757e-01, + -1.0550e-01, -2.2037e-01], + [-1.7500e-01, 3.7011e-03, 5.0029e-02, ..., -5.7463e-05, + -1.2234e-01, 9.6358e-02], + [-1.4164e-01, -1.2650e-01, 1.1380e-01, ..., 2.3580e-01, + -8.8086e-02, -2.4743e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.6858e-08, 0.0000e+00, -9.6858e-08, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 8.6613e-08, 0.0000e+00, 8.7544e-08, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -9.3132e-10, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 7.4506e-09, 9.3132e-10, 7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0335, 0.0056, 0.0210, 0.0155, 0.0418, 0.0149, -0.0015, 0.0013, + -0.0188, -0.0549], device='cuda:0'), grad: tensor([ 2.0489e-08, -1.7229e-07, -1.8626e-09, 3.7253e-09, -3.7253e-09, + 1.8626e-09, -1.5832e-08, 1.5926e-07, -3.7253e-09, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 217.49, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4209 re_mapping 0.0028 re_causal 0.0097 /// teacc 99.12 lr 0.00001000 +Epoch 481, weight, value: tensor([[-1.8183e-01, 2.6868e-01, -2.7773e-02, ..., -2.7802e-02, + 6.1001e-02, -4.9679e-02], + [ 2.1899e-01, -1.2233e-01, 3.1630e-02, ..., -1.2859e-01, + 2.8729e-02, -1.0062e-01], + [-1.0543e-01, -3.1893e-01, -6.7393e-02, ..., -7.7585e-02, + 2.2756e-02, -3.2214e-01], + ..., + [-2.0811e-01, -3.8202e-01, 7.1054e-02, ..., -1.2763e-01, + -1.0554e-01, -2.2053e-01], + [-1.7509e-01, 3.8170e-03, 5.0154e-02, ..., -6.8756e-05, + -1.2230e-01, 9.6531e-02], + [-1.4161e-01, -1.2667e-01, 1.1387e-01, ..., 2.3595e-01, + -8.8084e-02, -2.4755e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.5832e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [-9.3132e-10, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 5.5879e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 7.4506e-09], + ..., + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 6.5193e-09, -1.7695e-08, -1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, -7.4506e-09], + [ 9.3132e-10, 1.6764e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0334, 0.0056, 0.0210, 0.0155, 0.0419, 0.0149, -0.0014, 0.0014, + -0.0187, -0.0549], device='cuda:0'), grad: tensor([-1.2107e-08, 2.3283e-08, 4.0978e-08, -7.6368e-08, -5.1036e-07, + -2.6263e-07, 2.8219e-07, 1.1176e-08, -3.5390e-08, 5.4389e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 218.07, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4097 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.09 lr 0.00001000 +Epoch 482, weight, value: tensor([[-1.8190e-01, 2.6888e-01, -2.7892e-02, ..., -2.7964e-02, + 6.0985e-02, -4.9820e-02], + [ 2.1910e-01, -1.2235e-01, 3.1662e-02, ..., -1.2861e-01, + 2.8737e-02, -1.0068e-01], + [-1.0556e-01, -3.1900e-01, -6.7518e-02, ..., -7.7607e-02, + 2.2823e-02, -3.2233e-01], + ..., + [-2.0818e-01, -3.8208e-01, 7.1135e-02, ..., -1.2769e-01, + -1.0561e-01, -2.2071e-01], + [-1.7516e-01, 3.8532e-03, 5.0163e-02, ..., -1.0318e-04, + -1.2230e-01, 9.6557e-02], + [-1.4167e-01, -1.2685e-01, 1.1388e-01, ..., 2.3611e-01, + -8.8064e-02, -2.4770e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.4901e-08], + [-3.7253e-09, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -2.7940e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0334, 0.0056, 0.0210, 0.0156, 0.0421, 0.0148, -0.0015, 0.0014, + -0.0188, -0.0551], device='cuda:0'), grad: tensor([ 4.5635e-08, 2.4214e-08, -2.4214e-08, -7.5437e-08, -3.2876e-07, + -9.3132e-10, 8.3819e-09, 8.6613e-08, 2.7940e-08, 2.4680e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 217.83, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4301 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 483, weight, value: tensor([[-1.8197e-01, 2.6908e-01, -2.7798e-02, ..., -2.7954e-02, + 6.0967e-02, -4.9849e-02], + [ 2.1921e-01, -1.2238e-01, 3.1701e-02, ..., -1.2872e-01, + 2.8756e-02, -1.0076e-01], + [-1.0560e-01, -3.1910e-01, -6.7659e-02, ..., -7.7612e-02, + 2.2821e-02, -3.2245e-01], + ..., + [-2.0829e-01, -3.8220e-01, 7.1263e-02, ..., -1.2774e-01, + -1.0567e-01, -2.2092e-01], + [-1.7530e-01, 3.9835e-03, 5.0320e-02, ..., -1.1094e-04, + -1.2225e-01, 9.6700e-02], + [-1.4164e-01, -1.2713e-01, 1.1372e-01, ..., 2.3626e-01, + -8.8096e-02, -2.4784e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 1.8626e-09, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.6077e-08, 3.2596e-08, -9.3132e-09, ..., -9.3132e-10, + 9.3132e-09, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0335, 0.0056, 0.0209, 0.0156, 0.0420, 0.0148, -0.0015, 0.0014, + -0.0188, -0.0553], device='cuda:0'), grad: tensor([ 9.3132e-10, 2.3283e-08, -8.1956e-08, 9.3132e-10, -1.8626e-09, + -2.8033e-07, 2.1234e-07, 2.7008e-08, 9.5926e-08, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 218.12, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4257 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 484, weight, value: tensor([[-1.8209e-01, 2.6930e-01, -2.7832e-02, ..., -2.7989e-02, + 6.0929e-02, -4.9885e-02], + [ 2.1925e-01, -1.2238e-01, 3.1702e-02, ..., -1.2878e-01, + 2.8784e-02, -1.0083e-01], + [-1.0564e-01, -3.1922e-01, -6.7765e-02, ..., -7.7613e-02, + 2.2815e-02, -3.2262e-01], + ..., + [-2.0830e-01, -3.8231e-01, 7.1416e-02, ..., -1.2776e-01, + -1.0569e-01, -2.2101e-01], + [-1.7555e-01, 3.8895e-03, 5.0270e-02, ..., -1.2232e-04, + -1.2229e-01, 9.6647e-02], + [-1.4164e-01, -1.2727e-01, 1.1368e-01, ..., 2.3635e-01, + -8.8103e-02, -2.4795e-01]], device='cuda:0'), grad: tensor([[ 1.4901e-08, -1.9558e-08, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, -1.8626e-09], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 6.5193e-09, 3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 1.4901e-08, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 8.3819e-09], + [ 9.3132e-10, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 2.5146e-08]], device='cuda:0') +Epoch 484, bias, value: tensor([ 0.0336, 0.0056, 0.0209, 0.0157, 0.0419, 0.0148, -0.0016, 0.0015, + -0.0190, -0.0554], device='cuda:0'), grad: tensor([-4.3772e-08, 8.3819e-09, -2.7940e-09, -1.0710e-07, 2.2352e-08, + -1.2107e-08, 3.2596e-08, 1.7695e-08, 2.3283e-08, 6.4261e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 217.70, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4119 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.08 lr 0.00001000 +Epoch 485, weight, value: tensor([[-1.8216e-01, 2.6960e-01, -2.7811e-02, ..., -2.8028e-02, + 6.0921e-02, -4.9838e-02], + [ 2.1940e-01, -1.2250e-01, 3.1704e-02, ..., -1.2886e-01, + 2.8802e-02, -1.0085e-01], + [-1.0584e-01, -3.1936e-01, -6.7936e-02, ..., -7.7613e-02, + 2.2782e-02, -3.2272e-01], + ..., + [-2.0841e-01, -3.8240e-01, 7.1454e-02, ..., -1.2779e-01, + -1.0572e-01, -2.2113e-01], + [-1.7568e-01, 3.8841e-03, 5.0301e-02, ..., -1.2937e-04, + -1.2227e-01, 9.6702e-02], + [-1.4166e-01, -1.2741e-01, 1.1372e-01, ..., 2.3647e-01, + -8.8117e-02, -2.4805e-01]], device='cuda:0'), grad: tensor([[ 6.5193e-09, -9.3132e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.8429e-08, 9.3132e-10, -3.6322e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-3.7253e-09, 0.0000e+00, -2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 3.7253e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 9.3132e-10, 2.7940e-09, 3.7253e-09, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 485, bias, value: tensor([ 0.0337, 0.0056, 0.0208, 0.0157, 0.0419, 0.0148, -0.0017, 0.0015, + -0.0190, -0.0554], device='cuda:0'), grad: tensor([-2.7940e-09, -1.0338e-07, 7.4506e-09, 5.1223e-08, -1.5832e-08, + -5.4017e-08, 1.5367e-07, -6.9849e-08, 9.3132e-09, 4.0047e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 218.22, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4303 re_mapping 0.0028 re_causal 0.0100 /// teacc 99.09 lr 0.00001000 +Epoch 486, weight, value: tensor([[-1.8218e-01, 2.6974e-01, -2.7769e-02, ..., -2.8095e-02, + 6.0914e-02, -4.9870e-02], + [ 2.1948e-01, -1.2252e-01, 3.1726e-02, ..., -1.2889e-01, + 2.8810e-02, -1.0092e-01], + [-1.0598e-01, -3.1941e-01, -6.8108e-02, ..., -7.7637e-02, + 2.2780e-02, -3.2291e-01], + ..., + [-2.0844e-01, -3.8246e-01, 7.1488e-02, ..., -1.2786e-01, + -1.0573e-01, -2.2122e-01], + [-1.7578e-01, 3.8978e-03, 5.0331e-02, ..., -1.2914e-04, + -1.2228e-01, 9.6734e-02], + [-1.4169e-01, -1.2752e-01, 1.1373e-01, ..., 2.3659e-01, + -8.8119e-02, -2.4812e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-1.3970e-08, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 486, bias, value: tensor([ 0.0337, 0.0056, 0.0206, 0.0157, 0.0419, 0.0148, -0.0017, 0.0015, + -0.0190, -0.0554], device='cuda:0'), grad: tensor([ 3.7253e-09, 3.7253e-09, 1.8626e-09, -9.3132e-10, 4.5821e-07, + 3.7253e-09, 0.0000e+00, -4.8801e-07, 4.6566e-09, 2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 218.06, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4381 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.09 lr 0.00001000 +Epoch 487, weight, value: tensor([[-1.8227e-01, 2.6995e-01, -2.7704e-02, ..., -2.8089e-02, + 6.0891e-02, -4.9914e-02], + [ 2.1957e-01, -1.2253e-01, 3.1789e-02, ..., -1.2890e-01, + 2.8974e-02, -1.0084e-01], + [-1.0605e-01, -3.1947e-01, -6.8260e-02, ..., -7.7642e-02, + 2.2734e-02, -3.2299e-01], + ..., + [-2.0848e-01, -3.8252e-01, 7.1528e-02, ..., -1.2788e-01, + -1.0574e-01, -2.2131e-01], + [-1.7591e-01, 3.8821e-03, 5.0346e-02, ..., -1.2913e-04, + -1.2230e-01, 9.6696e-02], + [-1.4173e-01, -1.2765e-01, 1.1372e-01, ..., 2.3664e-01, + -8.8162e-02, -2.4829e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.6077e-08, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 6.5193e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0337, 0.0057, 0.0205, 0.0158, 0.0419, 0.0147, -0.0017, 0.0016, + -0.0191, -0.0555], device='cuda:0'), grad: tensor([ 2.7195e-07, 1.7695e-08, 1.6764e-08, 1.4901e-08, 5.5879e-09, + -1.6764e-08, -3.2131e-07, 1.8626e-09, 4.6566e-09, 1.0245e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 218.05, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4322 re_mapping 0.0027 re_causal 0.0101 /// teacc 99.09 lr 0.00001000 +Epoch 488, weight, value: tensor([[-1.8231e-01, 2.7011e-01, -2.7663e-02, ..., -2.8092e-02, + 6.0877e-02, -4.9935e-02], + [ 2.1978e-01, -1.2259e-01, 3.1912e-02, ..., -1.2890e-01, + 2.8987e-02, -1.0087e-01], + [-1.0603e-01, -3.1958e-01, -6.8360e-02, ..., -7.7644e-02, + 2.2735e-02, -3.2304e-01], + ..., + [-2.0870e-01, -3.8258e-01, 7.1483e-02, ..., -1.2790e-01, + -1.0578e-01, -2.2139e-01], + [-1.7605e-01, 3.8580e-03, 5.0347e-02, ..., -1.2453e-04, + -1.2231e-01, 9.6705e-02], + [-1.4179e-01, -1.2778e-01, 1.1374e-01, ..., 2.3669e-01, + -8.8133e-02, -2.4840e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 9.3132e-10, -9.3132e-09, ..., 0.0000e+00, + 2.2352e-08, 9.3132e-10]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0337, 0.0057, 0.0206, 0.0158, 0.0419, 0.0148, -0.0016, 0.0015, + -0.0192, -0.0556], device='cuda:0'), grad: tensor([ 4.6566e-09, 1.6764e-08, -5.5879e-08, 4.9360e-08, -2.1048e-07, + -2.7940e-09, -2.7940e-09, 1.1176e-08, 1.3970e-08, 1.8161e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 217.75, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4379 re_mapping 0.0027 re_causal 0.0102 /// teacc 99.12 lr 0.00001000 +Epoch 489, weight, value: tensor([[-1.8235e-01, 2.7031e-01, -2.7607e-02, ..., -2.8096e-02, + 6.0872e-02, -4.9990e-02], + [ 2.1983e-01, -1.2267e-01, 3.1887e-02, ..., -1.2890e-01, + 2.8987e-02, -1.0089e-01], + [-1.0613e-01, -3.1972e-01, -6.8502e-02, ..., -7.7646e-02, + 2.2729e-02, -3.2310e-01], + ..., + [-2.0872e-01, -3.8266e-01, 7.1596e-02, ..., -1.2792e-01, + -1.0579e-01, -2.2145e-01], + [-1.7618e-01, 3.8679e-03, 5.0381e-02, ..., -1.2557e-04, + -1.2232e-01, 9.6731e-02], + [-1.4187e-01, -1.2795e-01, 1.1370e-01, ..., 2.3673e-01, + -8.8147e-02, -2.4851e-01]], device='cuda:0'), grad: tensor([[0., 0., 0., ..., 0., 0., 0.], + [0., 0., 0., ..., 0., 0., 0.], + [0., 0., 0., ..., 0., 0., 0.], + ..., + [0., 0., 0., ..., 0., 0., 0.], + [0., 0., 0., ..., 0., 0., 0.], + [0., 0., 0., ..., 0., 0., 0.]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0338, 0.0057, 0.0205, 0.0157, 0.0419, 0.0148, -0.0016, 0.0016, + -0.0193, -0.0557], device='cuda:0'), grad: tensor([ 7.4506e-09, 6.9849e-08, 0.0000e+00, 0.0000e+00, -1.0058e-07, + 9.3132e-09, -1.0245e-08, 1.0245e-08, 3.7253e-09, 1.1176e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 217.92, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4472 re_mapping 0.0027 re_causal 0.0101 /// teacc 99.12 lr 0.00001000 +Epoch 490, weight, value: tensor([[-1.8246e-01, 2.7044e-01, -2.7569e-02, ..., -2.8108e-02, + 6.0832e-02, -5.0045e-02], + [ 2.2005e-01, -1.2267e-01, 3.1943e-02, ..., -1.2888e-01, + 2.9011e-02, -1.0094e-01], + [-1.0619e-01, -3.1981e-01, -6.8604e-02, ..., -7.7647e-02, + 2.2718e-02, -3.2320e-01], + ..., + [-2.0893e-01, -3.8275e-01, 7.1597e-02, ..., -1.2797e-01, + -1.0583e-01, -2.2165e-01], + [-1.7633e-01, 3.8346e-03, 5.0333e-02, ..., -1.4913e-04, + -1.2235e-01, 9.6728e-02], + [-1.4191e-01, -1.2805e-01, 1.1376e-01, ..., 2.3678e-01, + -8.8159e-02, -2.4865e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -2.7940e-09, -5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, 9.3132e-10, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0338, 0.0058, 0.0205, 0.0157, 0.0418, 0.0148, -0.0015, 0.0015, + -0.0194, -0.0557], device='cuda:0'), grad: tensor([ 4.6566e-09, 3.7253e-09, 8.3819e-09, 0.0000e+00, -1.9558e-08, + 9.3132e-10, -1.8626e-09, 2.7940e-09, -1.4901e-08, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 217.75, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3956 re_mapping 0.0027 re_causal 0.0094 /// teacc 99.11 lr 0.00001000 +Epoch 491, weight, value: tensor([[-1.8247e-01, 2.7070e-01, -2.7463e-02, ..., -2.8128e-02, + 6.0825e-02, -5.0069e-02], + [ 2.2013e-01, -1.2270e-01, 3.1970e-02, ..., -1.2890e-01, + 2.9019e-02, -1.0101e-01], + [-1.0624e-01, -3.1994e-01, -6.8727e-02, ..., -7.7647e-02, + 2.2740e-02, -3.2332e-01], + ..., + [-2.0899e-01, -3.8284e-01, 7.1619e-02, ..., -1.2802e-01, + -1.0588e-01, -2.2178e-01], + [-1.7639e-01, 3.9234e-03, 5.0434e-02, ..., -1.4913e-04, + -1.2236e-01, 9.6790e-02], + [-1.4197e-01, -1.2825e-01, 1.1374e-01, ..., 2.3686e-01, + -8.8195e-02, -2.4880e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -5.5879e-09, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [-7.4506e-09, -9.3132e-10, -1.2107e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 6.5193e-09], + [ 9.3132e-09, 6.5193e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-09], + [ 2.3283e-08, 3.7253e-08, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.6077e-08]], device='cuda:0') +Epoch 491, bias, value: tensor([ 0.0339, 0.0059, 0.0205, 0.0158, 0.0417, 0.0148, -0.0016, 0.0015, + -0.0194, -0.0558], device='cuda:0'), grad: tensor([-7.4506e-09, -3.7253e-09, 3.7253e-09, -1.0803e-07, 9.3132e-10, + -2.7940e-08, 2.0489e-08, 2.5146e-08, 2.0489e-08, 7.8231e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 217.81, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4061 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.07 lr 0.00001000 +Epoch 492, weight, value: tensor([[-1.8251e-01, 2.7083e-01, -2.7506e-02, ..., -2.8171e-02, + 6.0822e-02, -5.0091e-02], + [ 2.2025e-01, -1.2274e-01, 3.2019e-02, ..., -1.2897e-01, + 2.9031e-02, -1.0104e-01], + [-1.0633e-01, -3.1998e-01, -6.8859e-02, ..., -7.7647e-02, + 2.2720e-02, -3.2340e-01], + ..., + [-2.0908e-01, -3.8286e-01, 7.1616e-02, ..., -1.2803e-01, + -1.0597e-01, -2.2203e-01], + [-1.7646e-01, 3.9324e-03, 5.0496e-02, ..., -1.5057e-04, + -1.2237e-01, 9.6840e-02], + [-1.4199e-01, -1.2839e-01, 1.1381e-01, ..., 2.3696e-01, + -8.8243e-02, -2.4889e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-5.5879e-09, 0.0000e+00, 1.5832e-08, ..., 3.7253e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, -1.8626e-09], + [ 3.7253e-09, 0.0000e+00, -1.3970e-08, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 492, bias, value: tensor([ 0.0339, 0.0059, 0.0206, 0.0158, 0.0417, 0.0148, -0.0016, 0.0015, + -0.0195, -0.0558], device='cuda:0'), grad: tensor([ 1.2107e-08, 1.4901e-08, -6.0536e-08, 3.9116e-08, -1.4901e-08, + 4.9360e-08, -2.7940e-09, -1.6019e-07, 6.5193e-09, 1.1921e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 217.57, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4159 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.08 lr 0.00001000 +Epoch 493, weight, value: tensor([[-1.8249e-01, 2.7095e-01, -2.7476e-02, ..., -2.8175e-02, + 6.0823e-02, -5.0222e-02], + [ 2.2029e-01, -1.2280e-01, 3.2034e-02, ..., -1.2905e-01, + 2.9044e-02, -1.0108e-01], + [-1.0632e-01, -3.2003e-01, -6.8856e-02, ..., -7.7649e-02, + 2.2740e-02, -3.2350e-01], + ..., + [-2.0913e-01, -3.8290e-01, 7.1698e-02, ..., -1.2803e-01, + -1.0600e-01, -2.2215e-01], + [-1.7653e-01, 3.9852e-03, 5.0597e-02, ..., -1.5017e-04, + -1.2236e-01, 9.6887e-02], + [-1.4200e-01, -1.2854e-01, 1.1374e-01, ..., 2.3704e-01, + -8.8272e-02, -2.4900e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0339, 0.0058, 0.0207, 0.0159, 0.0416, 0.0147, -0.0016, 0.0015, + -0.0195, -0.0559], device='cuda:0'), grad: tensor([ 9.3132e-10, 9.3132e-09, -2.0489e-08, 6.5193e-09, -1.8626e-09, + -3.7253e-09, 2.7940e-09, 8.3819e-09, 0.0000e+00, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 217.95, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4440 re_mapping 0.0027 re_causal 0.0102 /// teacc 99.09 lr 0.00001000 +Epoch 494, weight, value: tensor([[-1.8260e-01, 2.7128e-01, -2.7384e-02, ..., -2.8198e-02, + 6.0777e-02, -5.0260e-02], + [ 2.2038e-01, -1.2279e-01, 3.2122e-02, ..., -1.2906e-01, + 2.9232e-02, -1.0100e-01], + [-1.0636e-01, -3.2015e-01, -6.8954e-02, ..., -7.7658e-02, + 2.2721e-02, -3.2358e-01], + ..., + [-2.0915e-01, -3.8302e-01, 7.1761e-02, ..., -1.2805e-01, + -1.0606e-01, -2.2229e-01], + [-1.7672e-01, 3.9223e-03, 5.0504e-02, ..., -1.5103e-04, + -1.2242e-01, 9.6849e-02], + [-1.4207e-01, -1.2873e-01, 1.1371e-01, ..., 2.3709e-01, + -8.8320e-02, -2.4910e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -4.3772e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -4.5635e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0341, 0.0059, 0.0207, 0.0161, 0.0416, 0.0145, -0.0018, 0.0016, + -0.0197, -0.0560], device='cuda:0'), grad: tensor([-8.3819e-08, 5.5879e-09, -5.9605e-08, 1.8626e-08, 8.8476e-08, + 4.7497e-08, 4.9360e-08, -4.6566e-09, 2.7940e-08, -9.7789e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 217.71, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4254 re_mapping 0.0027 re_causal 0.0099 /// teacc 99.09 lr 0.00001000 +Epoch 495, weight, value: tensor([[-1.8272e-01, 2.7143e-01, -2.7353e-02, ..., -2.8197e-02, + 6.0721e-02, -5.0311e-02], + [ 2.2052e-01, -1.2268e-01, 3.2256e-02, ..., -1.2906e-01, + 2.9432e-02, -1.0091e-01], + [-1.0635e-01, -3.2030e-01, -6.9008e-02, ..., -7.7664e-02, + 2.2696e-02, -3.2371e-01], + ..., + [-2.0924e-01, -3.8311e-01, 7.1801e-02, ..., -1.2808e-01, + -1.0612e-01, -2.2248e-01], + [-1.7695e-01, 3.8753e-03, 5.0377e-02, ..., -1.7621e-04, + -1.2251e-01, 9.6826e-02], + [-1.4217e-01, -1.2888e-01, 1.1366e-01, ..., 2.3711e-01, + -8.8361e-02, -2.4920e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 6.7987e-08, ..., 6.5193e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 5.5879e-09, 7.4506e-09, ..., 2.7940e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, -6.5193e-09, -8.0094e-08, ..., -1.0245e-08, + 0.0000e+00, -1.8626e-09]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0341, 0.0060, 0.0207, 0.0162, 0.0417, 0.0145, -0.0018, 0.0016, + -0.0199, -0.0562], device='cuda:0'), grad: tensor([ 3.5390e-08, 5.9605e-08, -1.2107e-08, -8.3819e-09, 1.1176e-08, + 8.3819e-09, -9.2201e-08, 1.1269e-07, 2.5146e-08, -1.3504e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 217.68, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4066 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.08 lr 0.00001000 +Epoch 496, weight, value: tensor([[-1.8274e-01, 2.7166e-01, -2.7235e-02, ..., -2.8196e-02, + 6.0723e-02, -5.0315e-02], + [ 2.2057e-01, -1.2274e-01, 3.2237e-02, ..., -1.2907e-01, + 2.9530e-02, -1.0084e-01], + [-1.0640e-01, -3.2038e-01, -6.9092e-02, ..., -7.7665e-02, + 2.2697e-02, -3.2375e-01], + ..., + [-2.0925e-01, -3.8315e-01, 7.1875e-02, ..., -1.2812e-01, + -1.0613e-01, -2.2253e-01], + [-1.7701e-01, 3.8245e-03, 5.0339e-02, ..., -1.9997e-04, + -1.2252e-01, 9.6808e-02], + [-1.4223e-01, -1.2912e-01, 1.1367e-01, ..., 2.3715e-01, + -8.8362e-02, -2.4931e-01]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1735e-07, 6.0536e-08, 1.8626e-09, ..., 9.3132e-09, + 7.4506e-09, 3.1665e-08], + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + ..., + [ 4.6566e-09, 4.6566e-09, 6.5193e-09, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-09], + [-9.3132e-09, -4.2841e-08, -4.0978e-08, ..., 0.0000e+00, + -8.3819e-09, -3.5390e-08], + [ 1.1176e-08, 4.6566e-09, -9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0343, 0.0059, 0.0207, 0.0162, 0.0418, 0.0145, -0.0019, 0.0017, + -0.0200, -0.0563], device='cuda:0'), grad: tensor([ 9.3132e-10, 1.3504e-07, 8.3819e-09, 1.2107e-07, 2.7940e-09, + -2.2072e-07, 2.2352e-08, 1.5832e-08, -9.9652e-08, 1.8626e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 217.90, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4370 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 497, weight, value: tensor([[-1.8282e-01, 2.7183e-01, -2.7156e-02, ..., -2.8198e-02, + 6.0707e-02, -5.0375e-02], + [ 2.2067e-01, -1.2274e-01, 3.2170e-02, ..., -1.2910e-01, + 2.9620e-02, -1.0081e-01], + [-1.0646e-01, -3.2043e-01, -6.9132e-02, ..., -7.7667e-02, + 2.2740e-02, -3.2389e-01], + ..., + [-2.0933e-01, -3.8323e-01, 7.1944e-02, ..., -1.2818e-01, + -1.0618e-01, -2.2265e-01], + [-1.7713e-01, 3.8222e-03, 5.0376e-02, ..., -2.0069e-04, + -1.2251e-01, 9.6828e-02], + [-1.4227e-01, -1.2927e-01, 1.1375e-01, ..., 2.3721e-01, + -8.8362e-02, -2.4941e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 9.3132e-10, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -2.7940e-09, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0344, 0.0059, 0.0208, 0.0162, 0.0417, 0.0145, -0.0020, 0.0016, + -0.0201, -0.0563], device='cuda:0'), grad: tensor([ 6.5193e-09, 2.5146e-08, -3.4459e-08, 1.2107e-08, 5.5879e-09, + 5.5879e-09, 8.3819e-09, 9.3132e-09, -2.9802e-08, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 217.97, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4167 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.10 lr 0.00001000 +Epoch 498, weight, value: tensor([[-1.8285e-01, 2.7195e-01, -2.7095e-02, ..., -2.8196e-02, + 6.0682e-02, -5.0429e-02], + [ 2.2070e-01, -1.2278e-01, 3.2137e-02, ..., -1.2910e-01, + 2.9618e-02, -1.0084e-01], + [-1.0648e-01, -3.2048e-01, -6.9237e-02, ..., -7.7667e-02, + 2.2752e-02, -3.2405e-01], + ..., + [-2.0934e-01, -3.8329e-01, 7.1995e-02, ..., -1.2820e-01, + -1.0623e-01, -2.2274e-01], + [-1.7719e-01, 3.9264e-03, 5.0474e-02, ..., -2.0493e-04, + -1.2247e-01, 9.6969e-02], + [-1.4228e-01, -1.2943e-01, 1.1380e-01, ..., 2.3724e-01, + -8.8350e-02, -2.4949e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.7229e-08, ..., 0.0000e+00, + 7.9162e-09, 9.3132e-10], + [ 0.0000e+00, 4.6566e-10, 4.8429e-08, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + ..., + [ 4.6566e-10, 0.0000e+00, -3.3062e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -3.5856e-08, ..., 0.0000e+00, + 1.3970e-09, -3.2596e-09], + [ 4.6566e-10, 4.6566e-10, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0344, 0.0058, 0.0207, 0.0162, 0.0418, 0.0144, -0.0020, 0.0016, + -0.0201, -0.0564], device='cuda:0'), grad: tensor([ 1.5367e-08, 1.3225e-07, 1.4808e-07, 3.2596e-08, 3.2596e-09, + 4.7963e-08, -1.9511e-07, -1.1316e-07, -6.8918e-08, 6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 218.00, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4463 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.11 lr 0.00001000 +Epoch 499, weight, value: tensor([[-1.8293e-01, 2.7208e-01, -2.7126e-02, ..., -2.8227e-02, + 6.0675e-02, -5.0551e-02], + [ 2.2071e-01, -1.2280e-01, 3.2116e-02, ..., -1.2911e-01, + 2.9613e-02, -1.0086e-01], + [-1.0650e-01, -3.2065e-01, -6.9363e-02, ..., -7.7667e-02, + 2.2880e-02, -3.2420e-01], + ..., + [-2.0935e-01, -3.8339e-01, 7.2080e-02, ..., -1.2820e-01, + -1.0627e-01, -2.2290e-01], + [-1.7726e-01, 4.0230e-03, 5.0618e-02, ..., -2.0955e-04, + -1.2247e-01, 9.7045e-02], + [-1.4235e-01, -1.2964e-01, 1.1380e-01, ..., 2.3728e-01, + -8.8366e-02, -2.4961e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 1.4901e-08, 0.0000e+00, 4.0513e-08, ..., 0.0000e+00, + 9.3132e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + ..., + [-1.6764e-08, 0.0000e+00, -4.7963e-08, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 2.7940e-09], + [ 2.7940e-09, 1.8626e-09, 7.4506e-09, ..., 0.0000e+00, + 1.3970e-09, 3.7253e-09]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0344, 0.0058, 0.0209, 0.0164, 0.0419, 0.0143, -0.0020, 0.0017, + -0.0200, -0.0565], device='cuda:0'), grad: tensor([ 2.3283e-09, 7.7300e-08, 1.3970e-09, -6.1002e-08, -1.4435e-08, + 1.4901e-08, 9.3132e-10, -6.1467e-08, 7.9162e-09, 3.9581e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 218.19, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4397 re_mapping 0.0026 re_causal 0.0097 /// teacc 99.09 lr 0.00001000 +Epoch 500, weight, value: tensor([[-1.8301e-01, 2.7220e-01, -2.7143e-02, ..., -2.8249e-02, + 6.0661e-02, -5.0606e-02], + [ 2.2093e-01, -1.2283e-01, 3.2164e-02, ..., -1.2912e-01, + 2.9629e-02, -1.0091e-01], + [-1.0660e-01, -3.2079e-01, -6.9533e-02, ..., -7.7666e-02, + 2.2951e-02, -3.2431e-01], + ..., + [-2.0952e-01, -3.8352e-01, 7.2151e-02, ..., -1.2821e-01, + -1.0633e-01, -2.2305e-01], + [-1.7735e-01, 4.1282e-03, 5.0700e-02, ..., -2.0624e-04, + -1.2248e-01, 9.7153e-02], + [-1.4253e-01, -1.2975e-01, 1.1377e-01, ..., 2.3734e-01, + -8.8398e-02, -2.4973e-01]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -3.2596e-09, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 3.2131e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.0268e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-2.0023e-08, 0.0000e+00, -1.5926e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, -9.3132e-10, 1.7229e-08, ..., 0.0000e+00, + 0.0000e+00, -3.2596e-09], + [ 1.8161e-08, 1.3970e-09, 4.9826e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0344, 0.0059, 0.0209, 0.0164, 0.0420, 0.0143, -0.0020, 0.0017, + -0.0200, -0.0566], device='cuda:0'), grad: tensor([-4.1910e-09, 8.8476e-08, 7.9628e-08, 1.0524e-07, 9.7789e-09, + -2.1886e-08, 2.3283e-09, -4.7032e-07, 4.7032e-08, 1.6857e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 218.03, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4336 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.11 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip2', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip2/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.979996 98.989998 ... 77.379173 70.994158 +ShearY 98.830002 98.779999 ... 77.379173 62.507677 +AutoContrast 98.900002 99.089996 ... 77.379173 57.702839 +Invert 98.570000 97.629997 ... 77.379173 63.917725 +Equalize 98.000000 98.009995 ... 77.379173 68.822933 +Solarize 98.019997 97.680000 ... 77.379173 59.182313 +SolarizeAdd 98.229996 97.699997 ... 77.379173 69.875464 +Posterize 99.089996 98.949997 ... 77.379173 71.538998 +Contrast 99.080002 99.150002 ... 77.379173 65.908477 +Color 99.099998 99.190002 ... 77.379173 59.656948 +Brightness 99.049995 99.159996 ... 77.379173 65.517488 +Sharpness 99.049995 99.119995 ... 77.379173 70.164459 +NoiseSalt 99.110001 99.169998 ... 77.379173 53.056127 +NoiseGaussian 99.139999 99.190002 ... 77.379173 56.760730 +w/o do (original x) 99.190000 0.000000 ... 0.000000 72.089808 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.24 67.013675 76.56927 73.317283 81.415047 74.578819 diff --git a/Meta-causal/code-withStyleAttack/66578.error b/Meta-causal/code-withStyleAttack/66578.error new file mode 100644 index 0000000000000000000000000000000000000000..d89a7028ed1f80e49e3a0ae92e1d9741c4616efe --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66578.error @@ -0,0 +1 @@ +run_my_joint_test.sh: line 40: dm}: command not found diff --git a/Meta-causal/code-withStyleAttack/66578.log b/Meta-causal/code-withStyleAttack/66578.log new file mode 100644 index 0000000000000000000000000000000000000000..4879ffceefe38b403416c925372e301af50371ed --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66578.log @@ -0,0 +1,14138 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0306, 0.0119, -0.0199, ..., -0.0066, -0.0307, -0.0255], + [ 0.0228, -0.0176, 0.0280, ..., 0.0063, -0.0039, -0.0037], + [ 0.0095, -0.0040, 0.0156, ..., -0.0134, 0.0029, -0.0208], + ..., + [-0.0270, 0.0054, 0.0070, ..., 0.0302, 0.0263, 0.0274], + [ 0.0165, -0.0206, -0.0018, ..., -0.0238, 0.0132, 0.0112], + [ 0.0169, -0.0250, 0.0047, ..., 0.0038, 0.0017, -0.0093]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0250, 0.0152, 0.0268, 0.0052, 0.0191, -0.0148, -0.0057, -0.0230, + -0.0063, 0.0275], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 222.95, cls_loss 1.5136 cls_loss_mapping 1.9199 cls_loss_causal 2.2362 re_mapping 0.1118 re_causal 0.1147 /// teacc 79.64 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0255, 0.0110, -0.0178, ..., -0.0056, -0.0374, -0.0252], + [ 0.0290, -0.0185, 0.0272, ..., -0.0018, 0.0010, -0.0045], + [ 0.0141, -0.0048, 0.0232, ..., -0.0197, -0.0038, -0.0216], + ..., + [-0.0225, 0.0062, 0.0002, ..., 0.0319, 0.0300, 0.0285], + [ 0.0162, -0.0202, -0.0033, ..., -0.0285, 0.0126, 0.0113], + [ 0.0100, -0.0258, -0.0041, ..., 0.0066, 0.0043, -0.0102]], + device='cuda:0'), grad: tensor([[ 0.0020, 0.0000, 0.0080, ..., 0.0029, 0.0067, 0.0000], + [-0.0033, 0.0000, 0.0051, ..., 0.0006, -0.0120, 0.0000], + [ 0.0002, 0.0000, -0.0071, ..., 0.0036, 0.0172, 0.0000], + ..., + [-0.0038, 0.0000, 0.0002, ..., -0.0090, -0.0126, 0.0000], + [ 0.0153, 0.0000, 0.0369, ..., -0.0058, 0.0131, 0.0000], + [ 0.0034, 0.0000, 0.0078, ..., 0.0266, 0.0378, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0268, 0.0163, 0.0261, 0.0064, 0.0193, -0.0148, -0.0061, -0.0223, + -0.0072, 0.0270], device='cuda:0'), grad: tensor([ 0.0203, -0.0135, 0.0148, -0.0461, -0.0032, -0.0476, -0.0285, -0.0083, + 0.0565, 0.0555], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 222.33, cls_loss 0.5296 cls_loss_mapping 0.8464 cls_loss_causal 1.8996 re_mapping 0.2041 re_causal 0.2492 /// teacc 91.04 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0254, 0.0110, -0.0171, ..., -0.0063, -0.0394, -0.0253], + [ 0.0312, -0.0185, 0.0269, ..., -0.0040, 0.0038, -0.0045], + [ 0.0153, -0.0048, 0.0263, ..., -0.0209, -0.0067, -0.0218], + ..., + [-0.0207, 0.0062, -0.0009, ..., 0.0319, 0.0307, 0.0285], + [ 0.0155, -0.0202, -0.0047, ..., -0.0324, 0.0146, 0.0112], + [ 0.0071, -0.0258, -0.0072, ..., 0.0083, 0.0045, -0.0103]], + device='cuda:0'), grad: tensor([[ 0.0025, 0.0000, 0.0106, ..., 0.0034, 0.0188, 0.0000], + [ 0.0077, 0.0000, 0.0218, ..., 0.0003, 0.0019, 0.0000], + [ 0.0152, 0.0000, 0.0284, ..., 0.0028, 0.0172, 0.0000], + ..., + [-0.0168, 0.0000, -0.0292, ..., -0.0001, -0.0234, 0.0000], + [ 0.0041, 0.0000, 0.0077, ..., 0.0020, -0.0224, 0.0000], + [ 0.0024, 0.0000, 0.0039, ..., 0.0034, 0.0025, 0.0000]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0270, 0.0161, 0.0259, 0.0064, 0.0193, -0.0139, -0.0069, -0.0229, + -0.0069, 0.0275], device='cuda:0'), grad: tensor([ 0.0402, 0.0133, 0.0329, -0.0095, -0.0016, -0.0154, 0.0003, -0.0289, + -0.0329, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 222.25, cls_loss 0.3242 cls_loss_mapping 0.5086 cls_loss_causal 1.6845 re_mapping 0.1569 re_causal 0.2457 /// teacc 92.89 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0253, 0.0110, -0.0159, ..., -0.0068, -0.0402, -0.0253], + [ 0.0322, -0.0185, 0.0264, ..., -0.0051, 0.0056, -0.0045], + [ 0.0163, -0.0048, 0.0279, ..., -0.0211, -0.0081, -0.0218], + ..., + [-0.0204, 0.0062, -0.0012, ..., 0.0312, 0.0325, 0.0285], + [ 0.0145, -0.0202, -0.0055, ..., -0.0351, 0.0159, 0.0112], + [ 0.0063, -0.0258, -0.0081, ..., 0.0088, 0.0034, -0.0103]], + device='cuda:0'), grad: tensor([[ 1.0711e-04, 0.0000e+00, 6.9714e-04, ..., 5.1355e-04, + 1.5554e-03, 0.0000e+00], + [-1.1587e-04, 0.0000e+00, 3.5439e-03, ..., 1.4648e-03, + -2.2030e-03, 0.0000e+00], + [ 7.0496e-03, 0.0000e+00, 2.7370e-04, ..., 5.8317e-04, + 3.3493e-03, 0.0000e+00], + ..., + [-1.0170e-02, 0.0000e+00, -1.3077e-02, ..., 2.3056e-02, + 1.2131e-02, 0.0000e+00], + [ 9.1124e-04, 0.0000e+00, -1.7118e-04, ..., 2.8114e-03, + -1.5202e-03, 0.0000e+00], + [ 4.0126e-04, 0.0000e+00, 2.7256e-03, ..., 9.9182e-05, + -4.7531e-03, 0.0000e+00]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0268, 0.0163, 0.0258, 0.0062, 0.0195, -0.0136, -0.0071, -0.0232, + -0.0072, 0.0278], device='cuda:0'), grad: tensor([ 0.0020, 0.0033, 0.0098, 0.0020, -0.0387, 0.0023, 0.0014, 0.0159, + 0.0039, -0.0019], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 221.79, cls_loss 0.2354 cls_loss_mapping 0.3555 cls_loss_causal 1.5608 re_mapping 0.1203 re_causal 0.2223 /// teacc 95.25 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 0.0254, 0.0110, -0.0148, ..., -0.0074, -0.0408, -0.0245], + [ 0.0330, -0.0185, 0.0264, ..., -0.0075, 0.0078, -0.0047], + [ 0.0170, -0.0048, 0.0306, ..., -0.0234, -0.0100, -0.0250], + ..., + [-0.0199, 0.0062, -0.0016, ..., 0.0305, 0.0339, 0.0284], + [ 0.0139, -0.0202, -0.0068, ..., -0.0367, 0.0168, 0.0077], + [ 0.0058, -0.0258, -0.0092, ..., 0.0098, 0.0024, -0.0115]], + device='cuda:0'), grad: tensor([[ 5.9456e-05, 0.0000e+00, -2.3315e-02, ..., -1.3077e-02, + -7.9269e-03, 0.0000e+00], + [ 1.2197e-03, 0.0000e+00, 1.0170e-02, ..., 1.0414e-03, + 1.2878e-02, 0.0000e+00], + [ 1.7872e-03, 0.0000e+00, -3.4599e-03, ..., 2.2831e-03, + 7.7915e-04, 0.0000e+00], + ..., + [-2.0008e-03, 0.0000e+00, -5.4588e-03, ..., 2.0504e-03, + 4.1056e-04, 0.0000e+00], + [-2.1915e-03, 0.0000e+00, 1.7385e-03, ..., 2.7676e-03, + -1.7456e-02, 0.0000e+00], + [ 1.9097e-04, 0.0000e+00, 1.8368e-03, ..., -7.8487e-04, + -1.2531e-03, 0.0000e+00]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0269, 0.0165, 0.0259, 0.0064, 0.0196, -0.0134, -0.0075, -0.0230, + -0.0076, 0.0276], device='cuda:0'), grad: tensor([-0.0305, 0.0160, 0.0039, 0.0070, -0.0092, 0.0223, 0.0117, 0.0019, + -0.0151, -0.0080], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 221.55, cls_loss 0.1926 cls_loss_mapping 0.2815 cls_loss_causal 1.4285 re_mapping 0.0989 re_causal 0.1999 /// teacc 96.02 lr 0.00010000 +Epoch 6, weight, value: tensor([[ 0.0251, 0.0110, -0.0136, ..., -0.0053, -0.0410, -0.0245], + [ 0.0338, -0.0185, 0.0270, ..., -0.0082, 0.0096, -0.0047], + [ 0.0172, -0.0048, 0.0323, ..., -0.0246, -0.0114, -0.0251], + ..., + [-0.0194, 0.0062, -0.0036, ..., 0.0301, 0.0347, 0.0286], + [ 0.0134, -0.0202, -0.0081, ..., -0.0383, 0.0183, 0.0077], + [ 0.0054, -0.0258, -0.0095, ..., 0.0104, 0.0019, -0.0115]], + device='cuda:0'), grad: tensor([[ 4.9055e-05, 0.0000e+00, -4.7088e-04, ..., -2.1887e-04, + 6.0225e-04, 2.2739e-05], + [ 2.5921e-03, 0.0000e+00, 5.2261e-04, ..., 2.1744e-03, + 2.2675e-02, 2.4498e-05], + [ 6.2990e-04, 0.0000e+00, 2.8877e-03, ..., 6.2418e-04, + 2.3632e-03, 3.7575e-04], + ..., + [-6.2609e-04, 0.0000e+00, -2.5425e-03, ..., 2.3708e-03, + 1.3704e-03, -7.0953e-04], + [ 1.2970e-04, 0.0000e+00, 1.0815e-03, ..., -1.5697e-03, + -1.8096e-04, 1.9833e-05], + [-3.4237e-03, 0.0000e+00, 1.7471e-03, ..., 1.0422e-02, + -2.4887e-02, 4.0084e-05]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0262, 0.0169, 0.0261, 0.0062, 0.0194, -0.0137, -0.0078, -0.0233, + -0.0075, 0.0277], device='cuda:0'), grad: tensor([ 4.8041e-04, 2.8381e-02, 7.3013e-03, 1.8280e-02, -1.6983e-02, + -8.8806e-03, 4.1038e-05, 1.8387e-03, -2.5726e-02, -4.7264e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 222.23, cls_loss 0.1550 cls_loss_mapping 0.2164 cls_loss_causal 1.3587 re_mapping 0.0846 re_causal 0.1903 /// teacc 96.20 lr 0.00010000 +Epoch 7, weight, value: tensor([[ 0.0250, 0.0110, -0.0123, ..., -0.0047, -0.0422, -0.0242], + [ 0.0344, -0.0185, 0.0264, ..., -0.0101, 0.0110, -0.0113], + [ 0.0170, -0.0048, 0.0336, ..., -0.0268, -0.0134, -0.0251], + ..., + [-0.0186, 0.0062, -0.0047, ..., 0.0294, 0.0357, 0.0319], + [ 0.0130, -0.0202, -0.0091, ..., -0.0399, 0.0198, -0.0011], + [ 0.0045, -0.0258, -0.0112, ..., 0.0105, 0.0008, -0.0198]], + device='cuda:0'), grad: tensor([[ 1.5065e-05, 0.0000e+00, -7.1287e-04, ..., 3.2568e-04, + 1.4467e-03, 0.0000e+00], + [-2.2936e-04, 0.0000e+00, 3.9177e-03, ..., 2.4438e-04, + -3.1834e-03, 0.0000e+00], + [ 1.3065e-04, 0.0000e+00, -1.2177e-02, ..., 7.2575e-04, + -2.7122e-03, 0.0000e+00], + ..., + [-6.5899e-04, 0.0000e+00, 1.2302e-03, ..., -5.5733e-03, + -9.9945e-03, 0.0000e+00], + [ 1.1790e-04, 0.0000e+00, 8.4686e-03, ..., 5.0879e-04, + 5.6229e-03, 0.0000e+00], + [ 1.7607e-04, 0.0000e+00, 1.1120e-03, ..., 4.1618e-03, + 7.9727e-03, 0.0000e+00]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0263, 0.0166, 0.0260, 0.0067, 0.0198, -0.0141, -0.0074, -0.0233, + -0.0074, 0.0273], device='cuda:0'), grad: tensor([ 0.0068, 0.0013, -0.0106, -0.0036, 0.0007, 0.0023, 0.0043, -0.0368, + 0.0143, 0.0214], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 221.87, cls_loss 0.1372 cls_loss_mapping 0.1826 cls_loss_causal 1.2902 re_mapping 0.0731 re_causal 0.1732 /// teacc 96.29 lr 0.00010000 +Epoch 8, weight, value: tensor([[ 2.4763e-02, 1.1010e-02, -1.1585e-02, ..., -2.5666e-03, + -4.3285e-02, -2.4056e-02], + [ 3.4629e-02, -1.8465e-02, 2.6861e-02, ..., -1.0430e-02, + 1.2428e-02, -1.1703e-02], + [ 1.7075e-02, -4.7722e-03, 3.5062e-02, ..., -2.8804e-02, + -1.3956e-02, -2.4409e-02], + ..., + [-1.7931e-02, 6.2070e-03, -6.3623e-03, ..., 2.8563e-02, + 3.6762e-02, 3.1173e-02], + [ 1.2703e-02, -2.0200e-02, -1.0620e-02, ..., -4.1443e-02, + 2.1108e-02, -1.3707e-03], + [ 4.2891e-03, -2.5830e-02, -1.1541e-02, ..., 1.0862e-02, + -6.0624e-05, -2.0121e-02]], device='cuda:0'), grad: tensor([[ 2.3976e-05, 0.0000e+00, -1.1024e-02, ..., -1.3229e-02, + -1.2426e-03, 0.0000e+00], + [-1.2338e-04, 0.0000e+00, -1.7715e-04, ..., 4.7231e-04, + -3.0780e-04, 0.0000e+00], + [ 3.0446e-04, 0.0000e+00, 2.7599e-03, ..., 1.3771e-03, + 2.4929e-03, 0.0000e+00], + ..., + [-6.4230e-04, 0.0000e+00, -5.1308e-04, ..., -3.0613e-04, + -1.0895e-02, 0.0000e+00], + [ 1.9979e-04, 0.0000e+00, 2.1667e-03, ..., 3.9520e-03, + 4.8218e-03, 0.0000e+00], + [ 9.6560e-05, 0.0000e+00, 1.4257e-03, ..., 1.9627e-03, + 4.7150e-03, 0.0000e+00]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0260, 0.0170, 0.0261, 0.0064, 0.0196, -0.0142, -0.0080, -0.0235, + -0.0070, 0.0274], device='cuda:0'), grad: tensor([-0.0213, 0.0002, 0.0046, 0.0028, -0.0114, 0.0004, 0.0172, -0.0077, + 0.0104, 0.0048], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 222.21, cls_loss 0.1293 cls_loss_mapping 0.1719 cls_loss_causal 1.2249 re_mapping 0.0630 re_causal 0.1524 /// teacc 96.57 lr 0.00010000 +Epoch 9, weight, value: tensor([[ 0.0245, 0.0110, -0.0113, ..., -0.0002, -0.0441, -0.0241], + [ 0.0348, -0.0185, 0.0273, ..., -0.0117, 0.0136, -0.0117], + [ 0.0172, -0.0048, 0.0360, ..., -0.0305, -0.0154, -0.0244], + ..., + [-0.0173, 0.0062, -0.0080, ..., 0.0279, 0.0382, 0.0312], + [ 0.0124, -0.0202, -0.0112, ..., -0.0427, 0.0218, -0.0014], + [ 0.0037, -0.0258, -0.0121, ..., 0.0111, -0.0005, -0.0201]], + device='cuda:0'), grad: tensor([[-1.1129e-03, 0.0000e+00, -3.6545e-03, ..., -4.7636e-04, + 3.8195e-04, 0.0000e+00], + [-2.7609e-04, 0.0000e+00, 5.0735e-04, ..., 7.4244e-04, + 2.8019e-03, 0.0000e+00], + [ 2.7227e-04, 0.0000e+00, -4.5419e-04, ..., 7.0095e-04, + 1.6756e-03, 0.0000e+00], + ..., + [ 7.3969e-05, 0.0000e+00, 2.8419e-04, ..., 2.8267e-03, + 4.5776e-03, 0.0000e+00], + [ 2.2030e-04, 0.0000e+00, 6.7329e-04, ..., 6.2370e-04, + -9.3918e-03, 0.0000e+00], + [ 1.4746e-04, 0.0000e+00, 5.5790e-04, ..., 7.4387e-04, + 1.4639e-03, 0.0000e+00]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0259, 0.0171, 0.0257, 0.0069, 0.0197, -0.0146, -0.0079, -0.0233, + -0.0070, 0.0271], device='cuda:0'), grad: tensor([ 0.0039, 0.0061, 0.0061, 0.0082, -0.0114, -0.0274, 0.0024, 0.0127, + -0.0063, 0.0057], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 222.27, cls_loss 0.1055 cls_loss_mapping 0.1456 cls_loss_causal 1.1560 re_mapping 0.0592 re_causal 0.1434 /// teacc 96.59 lr 0.00010000 +Epoch 10, weight, value: tensor([[ 0.0244, 0.0110, -0.0105, ..., 0.0010, -0.0444, -0.0241], + [ 0.0351, -0.0185, 0.0269, ..., -0.0129, 0.0145, -0.0121], + [ 0.0171, -0.0048, 0.0376, ..., -0.0321, -0.0158, -0.0245], + ..., + [-0.0166, 0.0062, -0.0093, ..., 0.0275, 0.0392, 0.0317], + [ 0.0123, -0.0202, -0.0119, ..., -0.0433, 0.0231, -0.0016], + [ 0.0036, -0.0258, -0.0131, ..., 0.0112, -0.0011, -0.0203]], + device='cuda:0'), grad: tensor([[ 3.0845e-06, 0.0000e+00, 1.0729e-03, ..., 1.5535e-03, + 4.0960e-04, 0.0000e+00], + [-1.9813e-04, 0.0000e+00, -6.0225e-04, ..., 3.9721e-04, + -1.7099e-03, 0.0000e+00], + [ 6.4492e-05, 0.0000e+00, 3.7403e-03, ..., 5.8603e-04, + 1.1406e-03, 0.0000e+00], + ..., + [ 1.8373e-05, 0.0000e+00, 5.0735e-04, ..., 5.8365e-04, + 4.0293e-04, 0.0000e+00], + [ 1.7017e-05, 0.0000e+00, 4.4594e-03, ..., 2.6531e-03, + -2.5539e-03, 0.0000e+00], + [ 4.3176e-06, 0.0000e+00, 6.5613e-04, ..., -2.5597e-03, + -2.4772e-04, 0.0000e+00]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0257, 0.0168, 0.0260, 0.0065, 0.0197, -0.0145, -0.0081, -0.0230, + -0.0066, 0.0270], device='cuda:0'), grad: tensor([ 0.0059, -0.0008, 0.0026, -0.0083, -0.0188, 0.0061, 0.0035, 0.0018, + 0.0102, -0.0022], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 222.26, cls_loss 0.0976 cls_loss_mapping 0.1364 cls_loss_causal 1.1439 re_mapping 0.0544 re_causal 0.1374 /// teacc 97.35 lr 0.00010000 +Epoch 11, weight, value: tensor([[ 0.0241, 0.0110, -0.0104, ..., 0.0024, -0.0447, -0.0241], + [ 0.0356, -0.0185, 0.0268, ..., -0.0130, 0.0157, -0.0122], + [ 0.0171, -0.0048, 0.0385, ..., -0.0331, -0.0179, -0.0245], + ..., + [-0.0158, 0.0062, -0.0105, ..., 0.0273, 0.0400, 0.0318], + [ 0.0119, -0.0202, -0.0126, ..., -0.0441, 0.0248, -0.0016], + [ 0.0026, -0.0258, -0.0126, ..., 0.0113, -0.0019, -0.0203]], + device='cuda:0'), grad: tensor([[ 2.0131e-05, 0.0000e+00, 1.0908e-04, ..., 2.2423e-04, + 6.2466e-04, 0.0000e+00], + [-7.9632e-04, 0.0000e+00, -1.6057e-04, ..., 1.1575e-04, + -1.9064e-03, 0.0000e+00], + [ 1.2898e-04, 0.0000e+00, 6.0129e-04, ..., 1.1641e-04, + 2.2697e-03, 0.0000e+00], + ..., + [ 2.0266e-04, 0.0000e+00, 2.0504e-04, ..., 6.8235e-04, + 4.5929e-03, 0.0000e+00], + [ 7.1168e-05, 0.0000e+00, -5.8031e-04, ..., 6.1989e-04, + -6.8092e-03, 0.0000e+00], + [ 8.6129e-05, 0.0000e+00, 1.1742e-04, ..., -1.5211e-04, + 2.2392e-03, 0.0000e+00]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0256, 0.0168, 0.0257, 0.0066, 0.0196, -0.0147, -0.0080, -0.0229, + -0.0066, 0.0270], device='cuda:0'), grad: tensor([ 0.0010, -0.0024, 0.0027, 0.0019, -0.0010, 0.0056, -0.0089, 0.0053, + -0.0057, 0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 221.46, cls_loss 0.0974 cls_loss_mapping 0.1288 cls_loss_causal 1.1304 re_mapping 0.0499 re_causal 0.1283 /// teacc 97.32 lr 0.00010000 +Epoch 12, weight, value: tensor([[ 0.0240, 0.0110, -0.0102, ..., 0.0040, -0.0454, -0.0238], + [ 0.0359, -0.0185, 0.0273, ..., -0.0148, 0.0164, -0.0200], + [ 0.0171, -0.0048, 0.0395, ..., -0.0347, -0.0184, -0.0234], + ..., + [-0.0156, 0.0062, -0.0112, ..., 0.0266, 0.0405, 0.0338], + [ 0.0115, -0.0202, -0.0139, ..., -0.0452, 0.0256, -0.0086], + [ 0.0027, -0.0258, -0.0134, ..., 0.0114, -0.0020, -0.0208]], + device='cuda:0'), grad: tensor([[ 9.6709e-06, 0.0000e+00, -1.3704e-03, ..., -2.4395e-03, + 4.4560e-04, 0.0000e+00], + [-8.2076e-05, 0.0000e+00, 2.0046e-03, ..., 1.7500e-04, + 3.5000e-04, 0.0000e+00], + [-2.4170e-05, 0.0000e+00, -2.8473e-02, ..., 3.2663e-04, + -6.1417e-03, 0.0000e+00], + ..., + [-1.9264e-04, 0.0000e+00, 3.5782e-03, ..., 4.7803e-04, + -3.4504e-03, 0.0000e+00], + [ 4.0412e-05, 0.0000e+00, 2.0889e-02, ..., 7.5436e-04, + 4.2000e-03, 0.0000e+00], + [ 1.4234e-04, 0.0000e+00, 1.4620e-03, ..., 3.1662e-03, + 4.6921e-03, 0.0000e+00]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0257, 0.0169, 0.0256, 0.0066, 0.0197, -0.0147, -0.0082, -0.0229, + -0.0064, 0.0271], device='cuda:0'), grad: tensor([-0.0019, 0.0023, -0.0282, -0.0023, -0.0038, 0.0036, -0.0011, -0.0032, + 0.0249, 0.0098], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 222.11, cls_loss 0.0857 cls_loss_mapping 0.1162 cls_loss_causal 1.0856 re_mapping 0.0471 re_causal 0.1214 /// teacc 97.79 lr 0.00010000 +Epoch 13, weight, value: tensor([[ 0.0238, 0.0110, -0.0101, ..., 0.0054, -0.0463, -0.0243], + [ 0.0362, -0.0185, 0.0272, ..., -0.0157, 0.0181, -0.0234], + [ 0.0172, -0.0048, 0.0404, ..., -0.0363, -0.0192, -0.0248], + ..., + [-0.0153, 0.0062, -0.0122, ..., 0.0262, 0.0411, 0.0362], + [ 0.0113, -0.0202, -0.0145, ..., -0.0466, 0.0258, -0.0085], + [ 0.0023, -0.0258, -0.0134, ..., 0.0122, -0.0025, -0.0229]], + device='cuda:0'), grad: tensor([[ 2.0601e-06, 0.0000e+00, 7.5054e-04, ..., -2.6360e-05, + 1.6079e-03, 2.4252e-06], + [ 5.2117e-06, 0.0000e+00, 6.0892e-04, ..., 4.4465e-04, + 4.1771e-04, 1.6931e-06], + [ 2.2709e-05, 0.0000e+00, 2.4509e-04, ..., 3.3927e-04, + 3.8910e-03, 1.7127e-06], + ..., + [-2.0742e-04, 0.0000e+00, -2.9778e-04, ..., -2.0981e-03, + -9.1095e-03, 5.5507e-06], + [ 7.9051e-06, 0.0000e+00, 2.1267e-03, ..., 2.5392e-04, + 1.0738e-03, 1.1891e-05], + [ 9.8705e-05, 0.0000e+00, 7.3099e-04, ..., 2.1011e-02, + 9.2010e-03, 2.1115e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0257, 0.0171, 0.0254, 0.0065, 0.0192, -0.0143, -0.0082, -0.0227, + -0.0067, 0.0273], device='cuda:0'), grad: tensor([ 0.0044, 0.0017, 0.0060, -0.0047, -0.0210, -0.0034, -0.0018, -0.0162, + 0.0127, 0.0222], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 221.91, cls_loss 0.0773 cls_loss_mapping 0.1075 cls_loss_causal 1.0635 re_mapping 0.0439 re_causal 0.1146 /// teacc 97.90 lr 0.00010000 +Epoch 14, weight, value: tensor([[ 0.0237, 0.0110, -0.0103, ..., 0.0065, -0.0470, -0.0250], + [ 0.0361, -0.0185, 0.0270, ..., -0.0173, 0.0186, -0.0253], + [ 0.0171, -0.0048, 0.0415, ..., -0.0381, -0.0191, -0.0268], + ..., + [-0.0147, 0.0062, -0.0130, ..., 0.0261, 0.0419, 0.0415], + [ 0.0111, -0.0202, -0.0147, ..., -0.0477, 0.0271, -0.0118], + [ 0.0020, -0.0258, -0.0135, ..., 0.0122, -0.0034, -0.0228]], + device='cuda:0'), grad: tensor([[ 6.9477e-07, 0.0000e+00, 2.3448e-04, ..., 1.9073e-04, + 6.6698e-05, 2.0340e-06], + [-1.4648e-05, 0.0000e+00, 5.2977e-04, ..., 6.8545e-05, + -5.2500e-04, 8.3596e-06], + [ 8.5309e-06, 0.0000e+00, -3.3760e-03, ..., 5.7507e-04, + 2.8968e-04, 3.4459e-06], + ..., + [-3.3557e-05, 0.0000e+00, 8.9824e-05, ..., 3.9864e-04, + 3.1781e-04, 4.0770e-05], + [ 6.5640e-06, 0.0000e+00, 1.3561e-03, ..., 2.2101e-04, + -1.3971e-04, 6.1631e-05], + [ 1.9208e-05, 0.0000e+00, 4.9621e-05, ..., -4.6682e-04, + -6.2990e-04, -2.9492e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0257, 0.0168, 0.0257, 0.0064, 0.0193, -0.0143, -0.0084, -0.0227, + -0.0065, 0.0271], device='cuda:0'), grad: tensor([ 4.9162e-04, 1.6463e-04, -1.9293e-03, 1.1358e-03, 4.6730e-05, + -2.3079e-04, -2.1017e-04, 1.0834e-03, 1.3399e-03, -1.8911e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 222.09, cls_loss 0.0742 cls_loss_mapping 0.1059 cls_loss_causal 1.0272 re_mapping 0.0404 re_causal 0.1090 /// teacc 97.94 lr 0.00010000 +Epoch 15, weight, value: tensor([[ 0.0235, 0.0110, -0.0102, ..., 0.0080, -0.0481, -0.0256], + [ 0.0361, -0.0185, 0.0273, ..., -0.0184, 0.0193, -0.0312], + [ 0.0171, -0.0048, 0.0421, ..., -0.0395, -0.0205, -0.0320], + ..., + [-0.0138, 0.0062, -0.0133, ..., 0.0255, 0.0427, 0.0383], + [ 0.0110, -0.0202, -0.0150, ..., -0.0489, 0.0281, -0.0198], + [ 0.0018, -0.0258, -0.0135, ..., 0.0124, -0.0043, -0.0196]], + device='cuda:0'), grad: tensor([[ 4.2804e-06, 0.0000e+00, 1.1587e-04, ..., -2.3139e-04, + 1.3745e-04, 7.7665e-05], + [ 1.0133e-05, 0.0000e+00, 4.3464e-04, ..., 4.4465e-05, + 1.2600e-04, 1.0955e-04], + [-1.8179e-06, 0.0000e+00, -1.6356e-03, ..., 6.0588e-05, + -2.0051e-04, 8.8394e-05], + ..., + [-6.9976e-05, 0.0000e+00, 6.6137e-04, ..., 1.4877e-04, + -1.1677e-04, 4.7302e-04], + [ 4.4554e-06, 0.0000e+00, 2.1088e-04, ..., 4.2272e-04, + -3.1166e-03, 9.0981e-04], + [ 3.7521e-05, 0.0000e+00, 3.4356e-04, ..., -5.9013e-03, + -1.7376e-03, -1.3306e-02]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0254, 0.0167, 0.0257, 0.0065, 0.0198, -0.0144, -0.0084, -0.0227, + -0.0065, 0.0268], device='cuda:0'), grad: tensor([ 0.0003, 0.0008, -0.0015, -0.0038, 0.0225, 0.0055, 0.0011, 0.0017, + -0.0014, -0.0252], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 221.38, cls_loss 0.0742 cls_loss_mapping 0.0990 cls_loss_causal 1.0131 re_mapping 0.0400 re_causal 0.1012 /// teacc 97.76 lr 0.00010000 +Epoch 16, weight, value: tensor([[ 0.0232, 0.0110, -0.0100, ..., 0.0095, -0.0488, -0.0322], + [ 0.0360, -0.0185, 0.0273, ..., -0.0196, 0.0195, -0.0371], + [ 0.0171, -0.0048, 0.0432, ..., -0.0404, -0.0202, -0.0353], + ..., + [-0.0132, 0.0062, -0.0140, ..., 0.0251, 0.0434, 0.0390], + [ 0.0108, -0.0202, -0.0161, ..., -0.0500, 0.0292, -0.0237], + [ 0.0010, -0.0258, -0.0141, ..., 0.0124, -0.0047, -0.0162]], + device='cuda:0'), grad: tensor([[ 5.6177e-06, 0.0000e+00, 5.3406e-04, ..., -2.0782e-02, + 5.9795e-04, -1.5244e-02], + [-4.0531e-05, 0.0000e+00, 2.4462e-04, ..., 2.9540e-04, + -2.0313e-03, 2.3651e-04], + [ 5.7220e-05, 0.0000e+00, -2.7027e-03, ..., 3.1815e-03, + 9.6130e-04, 2.6131e-03], + ..., + [-1.1539e-04, 0.0000e+00, 2.4354e-04, ..., 1.0729e-03, + -4.9706e-03, -1.3046e-03], + [ 8.8289e-06, 0.0000e+00, 3.4308e-04, ..., 9.5129e-04, + -6.8932e-03, -1.6699e-03], + [ 4.7445e-05, 0.0000e+00, 9.1553e-05, ..., -1.6232e-03, + 8.2092e-03, 3.9558e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0252, 0.0163, 0.0264, 0.0070, 0.0194, -0.0146, -0.0088, -0.0227, + -0.0066, 0.0268], device='cuda:0'), grad: tensor([-0.0187, -0.0009, 0.0073, 0.0080, 0.0216, 0.0140, 0.0074, -0.0036, + -0.0199, -0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 221.71, cls_loss 0.0574 cls_loss_mapping 0.0772 cls_loss_causal 0.9760 re_mapping 0.0379 re_causal 0.1006 /// teacc 97.75 lr 0.00010000 +Epoch 17, weight, value: tensor([[ 0.0231, 0.0110, -0.0097, ..., 0.0103, -0.0495, -0.0312], + [ 0.0361, -0.0185, 0.0270, ..., -0.0204, 0.0204, -0.0423], + [ 0.0168, -0.0048, 0.0440, ..., -0.0414, -0.0209, -0.0379], + ..., + [-0.0131, 0.0062, -0.0147, ..., 0.0247, 0.0437, 0.0395], + [ 0.0107, -0.0202, -0.0165, ..., -0.0512, 0.0300, -0.0284], + [ 0.0007, -0.0258, -0.0145, ..., 0.0125, -0.0053, -0.0138]], + device='cuda:0'), grad: tensor([[ 3.1181e-06, 0.0000e+00, -2.3627e-04, ..., 5.5027e-04, + 1.3068e-05, 2.2125e-04], + [ 3.1926e-06, 0.0000e+00, 8.9407e-06, ..., 4.4036e-04, + -1.2159e-03, 2.5773e-04], + [ 3.4180e-06, 0.0000e+00, 5.2333e-05, ..., 2.9278e-04, + 2.1505e-04, 7.0989e-05], + ..., + [-4.7922e-05, 0.0000e+00, 1.9014e-05, ..., 4.6825e-04, + -2.2426e-05, 1.9157e-04], + [ 1.6904e-06, 0.0000e+00, 3.8832e-05, ..., 5.2691e-04, + 6.7663e-04, 5.0926e-04], + [ 2.8789e-05, 0.0000e+00, 5.2661e-05, ..., -5.4207e-03, + -6.8712e-04, -5.4932e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0252, 0.0162, 0.0262, 0.0072, 0.0195, -0.0148, -0.0088, -0.0228, + -0.0064, 0.0267], device='cuda:0'), grad: tensor([ 6.1274e-04, -6.7091e-04, 6.5947e-04, -3.2020e-04, 3.9177e-03, + 8.9347e-05, 1.9300e-04, 9.7275e-04, 1.3304e-03, -6.7863e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 221.87, cls_loss 0.0471 cls_loss_mapping 0.0680 cls_loss_causal 0.9607 re_mapping 0.0347 re_causal 0.0947 /// teacc 98.24 lr 0.00010000 +Epoch 18, weight, value: tensor([[ 0.0230, 0.0110, -0.0095, ..., 0.0110, -0.0499, -0.0330], + [ 0.0359, -0.0185, 0.0274, ..., -0.0211, 0.0212, -0.0450], + [ 0.0165, -0.0048, 0.0447, ..., -0.0434, -0.0216, -0.0410], + ..., + [-0.0119, 0.0062, -0.0157, ..., 0.0243, 0.0446, 0.0415], + [ 0.0107, -0.0202, -0.0172, ..., -0.0527, 0.0304, -0.0306], + [-0.0001, -0.0258, -0.0145, ..., 0.0128, -0.0056, -0.0121]], + device='cuda:0'), grad: tensor([[ 1.6868e-04, 0.0000e+00, 2.4629e-04, ..., -1.0614e-03, + 4.0084e-05, 8.4400e-05], + [-4.4078e-05, 0.0000e+00, 7.6199e-04, ..., 1.2207e-04, + 2.0957e-04, 3.8600e-04], + [-3.0112e-04, 0.0000e+00, -6.5498e-03, ..., 2.0444e-04, + 1.3504e-03, 4.7779e-04], + ..., + [-2.0337e-04, 0.0000e+00, 7.8392e-04, ..., 7.5221e-05, + -3.7022e-03, -6.8188e-04], + [ 3.2157e-05, 0.0000e+00, 1.7226e-04, ..., 1.9264e-04, + 4.3583e-04, 5.5218e-04], + [ 1.0282e-04, 0.0000e+00, 6.1989e-04, ..., 1.6129e-04, + 5.1975e-04, -2.6798e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0253, 0.0166, 0.0258, 0.0069, 0.0193, -0.0148, -0.0086, -0.0225, + -0.0064, 0.0270], device='cuda:0'), grad: tensor([-0.0010, 0.0017, -0.0022, 0.0039, 0.0018, 0.0018, -0.0005, -0.0052, + 0.0017, -0.0021], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 221.56, cls_loss 0.0568 cls_loss_mapping 0.0846 cls_loss_causal 1.0054 re_mapping 0.0330 re_causal 0.0943 /// teacc 97.96 lr 0.00010000 +Epoch 19, weight, value: tensor([[ 0.0229, 0.0110, -0.0093, ..., 0.0130, -0.0508, -0.0313], + [ 0.0360, -0.0185, 0.0275, ..., -0.0221, 0.0218, -0.0490], + [ 0.0165, -0.0048, 0.0456, ..., -0.0450, -0.0218, -0.0429], + ..., + [-0.0112, 0.0062, -0.0168, ..., 0.0237, 0.0452, 0.0421], + [ 0.0106, -0.0202, -0.0179, ..., -0.0539, 0.0313, -0.0334], + [-0.0008, -0.0258, -0.0146, ..., 0.0128, -0.0062, -0.0100]], + device='cuda:0'), grad: tensor([[ 4.2953e-06, 0.0000e+00, 3.1686e-04, ..., 9.5940e-04, + 1.4668e-03, 1.9401e-05], + [ 2.8938e-05, 0.0000e+00, 3.4094e-04, ..., -1.1978e-03, + -3.5572e-03, -1.5392e-03], + [ 5.1707e-05, 0.0000e+00, 5.0621e-03, ..., 1.4234e-04, + 1.3412e-02, 1.9085e-04], + ..., + [-1.5604e-04, 0.0000e+00, -5.2452e-03, ..., 1.6153e-04, + -1.2444e-02, 3.6740e-04], + [ 1.0140e-05, 0.0000e+00, -1.0138e-03, ..., 1.3018e-04, + -4.7951e-03, 1.7655e-04], + [ 2.8893e-05, 0.0000e+00, 3.4866e-03, ..., -3.5191e-04, + -2.6083e-04, -1.1806e-03]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0250, 0.0164, 0.0260, 0.0069, 0.0192, -0.0148, -0.0091, -0.0226, + -0.0062, 0.0268], device='cuda:0'), grad: tensor([ 0.0026, -0.0072, 0.0211, -0.0104, 0.0087, 0.0084, -0.0045, -0.0173, + -0.0077, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 221.38, cls_loss 0.0484 cls_loss_mapping 0.0693 cls_loss_causal 0.9107 re_mapping 0.0330 re_causal 0.0878 /// teacc 98.00 lr 0.00010000 +Epoch 20, weight, value: tensor([[ 0.0228, 0.0110, -0.0096, ..., 0.0145, -0.0517, -0.0315], + [ 0.0361, -0.0185, 0.0273, ..., -0.0227, 0.0224, -0.0495], + [ 0.0161, -0.0048, 0.0467, ..., -0.0463, -0.0223, -0.0464], + ..., + [-0.0111, 0.0062, -0.0178, ..., 0.0233, 0.0457, 0.0422], + [ 0.0105, -0.0202, -0.0181, ..., -0.0547, 0.0321, -0.0356], + [-0.0005, -0.0258, -0.0155, ..., 0.0123, -0.0065, -0.0089]], + device='cuda:0'), grad: tensor([[ 2.8927e-06, 0.0000e+00, -1.0502e-04, ..., -7.7248e-05, + 2.4629e-04, 4.0084e-05], + [ 3.9876e-05, 0.0000e+00, 1.6117e-04, ..., 3.6120e-04, + 1.8001e-04, 1.9658e-04], + [ 4.8608e-05, 0.0000e+00, -2.5415e-04, ..., 4.4632e-04, + 6.9904e-04, 6.0678e-05], + ..., + [-1.8144e-04, 0.0000e+00, 2.1923e-04, ..., 1.4055e-04, + -8.0168e-06, 1.9515e-04], + [ 2.4185e-05, 0.0000e+00, 3.2449e-04, ..., 4.4751e-04, + 3.0017e-04, 5.5361e-04], + [ 6.8843e-05, 0.0000e+00, 1.2422e-04, ..., -5.8234e-05, + -2.1303e-04, -4.3068e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0251, 0.0167, 0.0260, 0.0072, 0.0193, -0.0149, -0.0091, -0.0228, + -0.0058, 0.0265], device='cuda:0'), grad: tensor([ 6.4731e-05, 7.6962e-04, 1.0223e-03, 2.3403e-03, -2.5208e-02, + 6.5384e-03, 2.0432e-02, 9.0981e-04, 2.2640e-03, -9.1476e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 221.01, cls_loss 0.0495 cls_loss_mapping 0.0659 cls_loss_causal 0.9269 re_mapping 0.0308 re_causal 0.0829 /// teacc 98.00 lr 0.00010000 +Epoch 21, weight, value: tensor([[ 0.0228, 0.0110, -0.0091, ..., 0.0152, -0.0523, -0.0324], + [ 0.0363, -0.0185, 0.0274, ..., -0.0240, 0.0230, -0.0505], + [ 0.0159, -0.0048, 0.0472, ..., -0.0474, -0.0232, -0.0479], + ..., + [-0.0107, 0.0062, -0.0186, ..., 0.0231, 0.0462, 0.0422], + [ 0.0103, -0.0202, -0.0185, ..., -0.0556, 0.0329, -0.0389], + [-0.0005, -0.0258, -0.0160, ..., 0.0119, -0.0067, -0.0084]], + device='cuda:0'), grad: tensor([[ 2.3171e-06, 0.0000e+00, 7.7546e-05, ..., -5.3501e-04, + 3.7384e-04, 5.8323e-05], + [ 1.5929e-05, 0.0000e+00, -7.7629e-03, ..., 1.2934e-04, + -2.1000e-03, 3.6001e-04], + [ 2.1115e-05, 0.0000e+00, 2.4853e-03, ..., 1.3435e-04, + 1.2741e-03, 2.4700e-04], + ..., + [-8.7857e-05, 0.0000e+00, 9.2030e-04, ..., 1.5855e-04, + -2.6398e-03, -2.1324e-03], + [ 5.3234e-06, 0.0000e+00, 3.6216e-04, ..., 2.8062e-04, + -7.2327e-03, 3.6263e-04], + [ 2.5332e-05, 0.0000e+00, 2.2471e-04, ..., 4.7183e-04, + 3.6025e-04, 3.2735e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0250, 0.0166, 0.0256, 0.0071, 0.0193, -0.0149, -0.0092, -0.0227, + -0.0054, 0.0264], device='cuda:0'), grad: tensor([ 0.0005, -0.0108, 0.0044, 0.0008, 0.0005, 0.0078, 0.0073, -0.0023, + -0.0096, 0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 221.42, cls_loss 0.0518 cls_loss_mapping 0.0732 cls_loss_causal 0.8744 re_mapping 0.0297 re_causal 0.0799 /// teacc 97.99 lr 0.00010000 +Epoch 22, weight, value: tensor([[ 0.0227, 0.0110, -0.0089, ..., 0.0163, -0.0526, -0.0336], + [ 0.0364, -0.0185, 0.0267, ..., -0.0251, 0.0240, -0.0520], + [ 0.0158, -0.0048, 0.0482, ..., -0.0486, -0.0231, -0.0501], + ..., + [-0.0101, 0.0062, -0.0189, ..., 0.0225, 0.0467, 0.0426], + [ 0.0101, -0.0202, -0.0189, ..., -0.0570, 0.0327, -0.0418], + [-0.0011, -0.0258, -0.0167, ..., 0.0122, -0.0071, -0.0065]], + device='cuda:0'), grad: tensor([[ 5.7276e-07, 0.0000e+00, -4.2939e-04, ..., -1.1700e-04, + -4.4703e-05, 7.2904e-06], + [-1.1459e-05, 0.0000e+00, 1.0276e-04, ..., 1.9416e-05, + 1.1837e-04, 4.7162e-06], + [ 6.8396e-06, 0.0000e+00, 5.9557e-04, ..., 4.9412e-05, + 1.7014e-03, 1.2532e-05], + ..., + [-2.1860e-05, 0.0000e+00, 1.1784e-04, ..., 9.5740e-06, + 6.8307e-05, -6.8903e-05], + [ 3.5837e-06, 0.0000e+00, 2.6112e-03, ..., 9.2745e-05, + 1.2573e-02, 8.2999e-06], + [ 8.7097e-06, 0.0000e+00, 1.2958e-04, ..., 1.4627e-04, + 2.4867e-04, 2.2486e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0249, 0.0165, 0.0258, 0.0070, 0.0192, -0.0152, -0.0091, -0.0224, + -0.0057, 0.0267], device='cuda:0'), grad: tensor([-5.5218e-04, 6.2323e-04, 2.6951e-03, 2.0676e-03, 6.3372e-04, + -2.9266e-02, 9.4295e-05, 1.7619e-04, 2.2873e-02, 6.3896e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 222.11, cls_loss 0.0455 cls_loss_mapping 0.0671 cls_loss_causal 0.9136 re_mapping 0.0295 re_causal 0.0832 /// teacc 98.28 lr 0.00010000 +Epoch 23, weight, value: tensor([[ 0.0226, 0.0110, -0.0088, ..., 0.0174, -0.0528, -0.0344], + [ 0.0363, -0.0185, 0.0264, ..., -0.0259, 0.0240, -0.0542], + [ 0.0157, -0.0048, 0.0487, ..., -0.0496, -0.0238, -0.0522], + ..., + [-0.0099, 0.0062, -0.0187, ..., 0.0227, 0.0470, 0.0440], + [ 0.0103, -0.0202, -0.0193, ..., -0.0577, 0.0333, -0.0447], + [-0.0016, -0.0258, -0.0169, ..., 0.0119, -0.0077, -0.0057]], + device='cuda:0'), grad: tensor([[ 2.2296e-06, 0.0000e+00, -1.2732e-04, ..., -7.5698e-05, + 2.1958e-04, 1.7130e-04], + [ 2.1774e-06, 0.0000e+00, 3.1853e-04, ..., 8.3971e-04, + 1.4257e-03, 1.7300e-03], + [ 1.2383e-05, 0.0000e+00, 3.1614e-04, ..., 8.0967e-04, + 1.1425e-03, 1.0462e-03], + ..., + [-3.6478e-05, 0.0000e+00, 1.0490e-04, ..., 3.3379e-04, + 7.1049e-05, 7.2575e-04], + [ 4.4852e-06, 0.0000e+00, 1.5152e-04, ..., 2.0385e-04, + 3.1352e-05, 2.0707e-04], + [ 7.1637e-06, 0.0000e+00, 7.6115e-05, ..., -8.7678e-05, + 2.3782e-04, -4.9305e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0247, 0.0160, 0.0253, 0.0074, 0.0193, -0.0151, -0.0091, -0.0219, + -0.0057, 0.0264], device='cuda:0'), grad: tensor([ 9.8610e-04, 2.5291e-03, 1.9741e-03, 3.7718e-04, -9.9182e-03, + -1.6594e-03, 4.1580e-03, 7.4625e-04, 7.8058e-04, 3.1352e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 221.80, cls_loss 0.0461 cls_loss_mapping 0.0591 cls_loss_causal 0.8728 re_mapping 0.0291 re_causal 0.0764 /// teacc 98.33 lr 0.00010000 +Epoch 24, weight, value: tensor([[ 0.0225, 0.0110, -0.0090, ..., 0.0185, -0.0539, -0.0358], + [ 0.0363, -0.0185, 0.0260, ..., -0.0273, 0.0247, -0.0565], + [ 0.0158, -0.0048, 0.0494, ..., -0.0509, -0.0244, -0.0554], + ..., + [-0.0097, 0.0062, -0.0189, ..., 0.0226, 0.0475, 0.0443], + [ 0.0103, -0.0202, -0.0197, ..., -0.0585, 0.0341, -0.0473], + [-0.0019, -0.0258, -0.0169, ..., 0.0126, -0.0081, -0.0040]], + device='cuda:0'), grad: tensor([[ 1.0515e-06, 0.0000e+00, 2.8057e-03, ..., 4.4518e-03, + 7.5042e-05, 4.2647e-05], + [ 5.1036e-06, 0.0000e+00, 3.6979e-04, ..., 2.5153e-04, + 1.2505e-04, 2.5845e-04], + [ 1.1757e-05, 0.0000e+00, 1.0455e-04, ..., 2.7323e-04, + 2.1112e-04, 5.5462e-05], + ..., + [-9.3162e-05, 0.0000e+00, 5.1498e-05, ..., 1.7977e-04, + 1.9714e-05, 2.3711e-04], + [ 3.9265e-06, 0.0000e+00, -3.2845e-03, ..., -6.1073e-03, + -3.2043e-04, 3.4809e-04], + [ 3.7134e-05, 0.0000e+00, 2.3377e-04, ..., 7.9203e-04, + 4.5514e-04, 3.0375e-04]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0248, 0.0159, 0.0251, 0.0072, 0.0191, -0.0147, -0.0092, -0.0218, + -0.0055, 0.0264], device='cuda:0'), grad: tensor([ 0.0081, 0.0015, 0.0008, -0.0033, -0.0021, 0.0003, 0.0018, 0.0005, + -0.0093, 0.0018], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 221.64, cls_loss 0.0381 cls_loss_mapping 0.0566 cls_loss_causal 0.8845 re_mapping 0.0276 re_causal 0.0775 /// teacc 98.16 lr 0.00010000 +Epoch 25, weight, value: tensor([[ 0.0225, 0.0110, -0.0092, ..., 0.0194, -0.0543, -0.0353], + [ 0.0362, -0.0185, 0.0259, ..., -0.0283, 0.0251, -0.0581], + [ 0.0155, -0.0048, 0.0503, ..., -0.0528, -0.0248, -0.0559], + ..., + [-0.0085, 0.0062, -0.0195, ..., 0.0219, 0.0480, 0.0447], + [ 0.0102, -0.0202, -0.0203, ..., -0.0597, 0.0349, -0.0489], + [-0.0026, -0.0258, -0.0164, ..., 0.0131, -0.0089, -0.0034]], + device='cuda:0'), grad: tensor([[ 2.8193e-05, 0.0000e+00, -1.4520e-04, ..., -7.3254e-05, + 1.6439e-04, 5.8323e-05], + [-1.2207e-03, 0.0000e+00, 1.8108e-04, ..., 2.1338e-05, + -7.1068e-03, 1.1045e-04], + [-1.0424e-03, 0.0000e+00, -3.4866e-03, ..., 5.8591e-05, + -4.3373e-03, 1.6642e-04], + ..., + [ 3.9139e-03, 0.0000e+00, 3.4580e-03, ..., 2.7299e-04, + 8.1482e-03, 1.3115e-02], + [ 6.0034e-04, 0.0000e+00, 3.8445e-05, ..., 5.1856e-05, + 3.1605e-03, 1.6904e-04], + [-2.7885e-03, 0.0000e+00, 3.6091e-05, ..., -4.6325e-04, + -2.9716e-03, -1.4221e-02]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0248, 0.0157, 0.0255, 0.0072, 0.0192, -0.0148, -0.0093, -0.0221, + -0.0054, 0.0265], device='cuda:0'), grad: tensor([ 6.0916e-05, -8.0795e-03, -9.1476e-03, -2.0647e-04, 1.4229e-03, + 8.8978e-04, 1.4219e-03, 3.0411e-02, 3.6373e-03, -2.0416e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 221.24, cls_loss 0.0384 cls_loss_mapping 0.0560 cls_loss_causal 0.8789 re_mapping 0.0270 re_causal 0.0739 /// teacc 98.22 lr 0.00010000 +Epoch 26, weight, value: tensor([[ 0.0224, 0.0110, -0.0088, ..., 0.0208, -0.0553, -0.0360], + [ 0.0362, -0.0185, 0.0260, ..., -0.0294, 0.0255, -0.0594], + [ 0.0154, -0.0048, 0.0514, ..., -0.0541, -0.0252, -0.0585], + ..., + [-0.0076, 0.0062, -0.0207, ..., 0.0217, 0.0486, 0.0455], + [ 0.0100, -0.0202, -0.0206, ..., -0.0605, 0.0353, -0.0497], + [-0.0031, -0.0258, -0.0170, ..., 0.0125, -0.0094, -0.0041]], + device='cuda:0'), grad: tensor([[ 7.7426e-05, 0.0000e+00, -3.3283e-04, ..., -4.3702e-04, + 4.0102e-04, 6.9082e-05], + [ 1.3316e-04, 0.0000e+00, 2.9135e-04, ..., 3.2574e-05, + 1.5819e-04, 1.6415e-04], + [ 1.3912e-04, 0.0000e+00, -4.5919e-04, ..., 7.9930e-05, + 5.8079e-04, 9.7811e-05], + ..., + [-5.1689e-04, 0.0000e+00, 2.8348e-04, ..., 9.2030e-05, + -4.5133e-04, -3.7909e-04], + [ 8.1480e-05, 0.0000e+00, 2.3592e-04, ..., 1.1855e-04, + -2.0084e-03, -2.6393e-04], + [ 2.8872e-04, 0.0000e+00, 1.6880e-04, ..., -8.7261e-05, + 7.5293e-04, -7.3493e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0241, 0.0158, 0.0257, 0.0073, 0.0195, -0.0149, -0.0094, -0.0222, + -0.0056, 0.0259], device='cuda:0'), grad: tensor([ 1.7023e-04, 8.9550e-04, 9.9468e-04, -3.6049e-03, 8.3637e-04, + 1.4544e-03, -1.6451e-03, 9.2328e-05, -7.7868e-04, 1.5879e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 221.45, cls_loss 0.0413 cls_loss_mapping 0.0548 cls_loss_causal 0.8356 re_mapping 0.0257 re_causal 0.0700 /// teacc 98.17 lr 0.00010000 +Epoch 27, weight, value: tensor([[ 0.0221, 0.0110, -0.0088, ..., 0.0220, -0.0561, -0.0362], + [ 0.0360, -0.0185, 0.0264, ..., -0.0305, 0.0258, -0.0621], + [ 0.0150, -0.0048, 0.0521, ..., -0.0549, -0.0263, -0.0606], + ..., + [-0.0070, 0.0062, -0.0211, ..., 0.0220, 0.0492, 0.0465], + [ 0.0100, -0.0202, -0.0209, ..., -0.0611, 0.0363, -0.0510], + [-0.0033, -0.0258, -0.0175, ..., 0.0125, -0.0098, -0.0036]], + device='cuda:0'), grad: tensor([[ 1.0133e-05, 0.0000e+00, -3.8123e-04, ..., -6.4754e-04, + -9.3281e-05, 7.9215e-05], + [-3.3006e-06, 0.0000e+00, 1.6618e-04, ..., 6.3539e-05, + -2.7776e-04, 1.1355e-04], + [ 2.2694e-05, 0.0000e+00, -7.8344e-04, ..., 1.2469e-04, + 7.5459e-05, 1.3876e-04], + ..., + [-2.8086e-04, 0.0000e+00, 9.7811e-05, ..., -1.1933e-04, + -2.6035e-04, -8.3685e-04], + [ 1.4462e-05, 0.0000e+00, 4.4298e-04, ..., 5.4169e-04, + 4.7565e-04, 5.7793e-04], + [ 1.7154e-04, 0.0000e+00, 6.9082e-05, ..., -8.6403e-04, + 2.1100e-04, -1.3142e-03]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0242, 0.0157, 0.0254, 0.0072, 0.0194, -0.0148, -0.0096, -0.0218, + -0.0053, 0.0260], device='cuda:0'), grad: tensor([-6.8378e-04, 2.5585e-05, -3.6716e-04, 9.4557e-04, 4.1580e-03, + -5.9547e-03, 1.8368e-03, -9.9945e-04, 1.8291e-03, -7.8726e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 221.65, cls_loss 0.0360 cls_loss_mapping 0.0507 cls_loss_causal 0.8717 re_mapping 0.0246 re_causal 0.0704 /// teacc 98.18 lr 0.00010000 +Epoch 28, weight, value: tensor([[ 0.0220, 0.0110, -0.0087, ..., 0.0228, -0.0574, -0.0373], + [ 0.0359, -0.0185, 0.0265, ..., -0.0315, 0.0265, -0.0627], + [ 0.0149, -0.0048, 0.0527, ..., -0.0561, -0.0267, -0.0618], + ..., + [-0.0065, 0.0062, -0.0214, ..., 0.0212, 0.0498, 0.0470], + [ 0.0099, -0.0202, -0.0215, ..., -0.0615, 0.0367, -0.0539], + [-0.0034, -0.0258, -0.0181, ..., 0.0128, -0.0104, -0.0023]], + device='cuda:0'), grad: tensor([[ 2.6692e-06, 0.0000e+00, -6.3181e-05, ..., -1.6844e-04, + 8.8692e-05, 6.5386e-05], + [ 1.2584e-05, 0.0000e+00, 5.1880e-03, ..., 5.3793e-05, + 2.1768e-04, 3.1805e-04], + [ 6.5081e-06, 0.0000e+00, -5.2223e-03, ..., 5.1439e-05, + 1.0926e-04, 2.6894e-04], + ..., + [-2.2340e-04, 0.0000e+00, 3.9995e-05, ..., -1.2696e-04, + -4.0703e-03, -4.0092e-03], + [ 3.8266e-05, 0.0000e+00, 4.8459e-05, ..., 1.9968e-04, + 1.2836e-03, 1.5697e-03], + [ 1.5104e-04, 0.0000e+00, 1.3769e-05, ..., 6.0940e-04, + 1.9836e-03, 2.6894e-03]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0245, 0.0160, 0.0255, 0.0071, 0.0193, -0.0146, -0.0100, -0.0216, + -0.0052, 0.0257], device='cuda:0'), grad: tensor([-5.5432e-05, 5.6076e-03, -4.9477e-03, -7.5245e-04, -6.3133e-04, + -1.5235e-04, 5.5373e-05, -5.1918e-03, 2.8763e-03, 3.1910e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 221.48, cls_loss 0.0348 cls_loss_mapping 0.0493 cls_loss_causal 0.8602 re_mapping 0.0252 re_causal 0.0722 /// teacc 98.29 lr 0.00010000 +Epoch 29, weight, value: tensor([[ 0.0219, 0.0110, -0.0087, ..., 0.0236, -0.0578, -0.0376], + [ 0.0358, -0.0185, 0.0265, ..., -0.0327, 0.0270, -0.0632], + [ 0.0148, -0.0048, 0.0535, ..., -0.0566, -0.0274, -0.0631], + ..., + [-0.0058, 0.0062, -0.0223, ..., 0.0209, 0.0507, 0.0475], + [ 0.0097, -0.0202, -0.0223, ..., -0.0628, 0.0366, -0.0554], + [-0.0043, -0.0258, -0.0181, ..., 0.0128, -0.0112, -0.0024]], + device='cuda:0'), grad: tensor([[ 7.2084e-07, 0.0000e+00, -1.2636e-04, ..., 2.0713e-05, + 4.7892e-05, 2.4962e-04], + [ 1.6391e-06, 0.0000e+00, 9.0972e-06, ..., 1.7393e-04, + -6.6519e-04, 2.0611e-04], + [ 4.3847e-06, 0.0000e+00, -1.8668e-04, ..., 6.5088e-05, + 1.6403e-04, -5.4389e-05], + ..., + [-2.4945e-05, 0.0000e+00, 6.3300e-05, ..., -1.1320e-03, + -2.4164e-04, -1.4191e-03], + [ 2.6226e-06, 0.0000e+00, 1.2124e-04, ..., 3.4213e-04, + 6.9809e-04, 7.7057e-04], + [ 1.1876e-05, 0.0000e+00, 1.4037e-05, ..., 4.4441e-04, + -3.2091e-04, 1.6928e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0241, 0.0161, 0.0255, 0.0077, 0.0194, -0.0147, -0.0101, -0.0216, + -0.0059, 0.0257], device='cuda:0'), grad: tensor([ 1.3351e-04, -6.3705e-04, -8.2374e-05, 5.1403e-04, 5.4979e-04, + -7.5102e-04, 5.8079e-04, -1.9360e-03, 1.7624e-03, -1.3447e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 222.04, cls_loss 0.0313 cls_loss_mapping 0.0432 cls_loss_causal 0.8355 re_mapping 0.0242 re_causal 0.0664 /// teacc 98.52 lr 0.00010000 +Epoch 30, weight, value: tensor([[ 0.0218, 0.0110, -0.0083, ..., 0.0244, -0.0586, -0.0387], + [ 0.0357, -0.0185, 0.0265, ..., -0.0334, 0.0275, -0.0639], + [ 0.0148, -0.0048, 0.0541, ..., -0.0569, -0.0278, -0.0626], + ..., + [-0.0062, 0.0062, -0.0235, ..., 0.0213, 0.0509, 0.0465], + [ 0.0095, -0.0202, -0.0226, ..., -0.0637, 0.0369, -0.0567], + [-0.0038, -0.0258, -0.0188, ..., 0.0128, -0.0112, -0.0013]], + device='cuda:0'), grad: tensor([[ 3.5763e-05, 0.0000e+00, 1.2064e-04, ..., -2.0772e-05, + 1.3657e-03, 4.9680e-05], + [ 6.8486e-05, 0.0000e+00, -1.0729e-04, ..., 9.9063e-05, + -6.7368e-03, 2.2542e-04], + [ 1.1835e-03, 0.0000e+00, 3.4180e-03, ..., 7.9155e-05, + 8.5354e-04, 1.8620e-04], + ..., + [-2.2373e-03, 0.0000e+00, -6.0310e-03, ..., 3.5644e-04, + -3.8552e-04, 4.2892e-04], + [ 2.3887e-05, 0.0000e+00, 1.8978e-04, ..., 5.7042e-05, + 3.4447e-03, 1.0610e-04], + [ 4.1157e-05, 0.0000e+00, 3.8266e-05, ..., -3.9756e-05, + 2.6608e-04, -9.6858e-06]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0243, 0.0164, 0.0258, 0.0078, 0.0190, -0.0145, -0.0099, -0.0223, + -0.0060, 0.0260], device='cuda:0'), grad: tensor([ 1.6356e-03, -6.6948e-03, 5.2223e-03, 2.9621e-03, -6.4421e-04, + 7.5638e-05, 8.5497e-04, -6.8970e-03, 3.2711e-03, 2.1088e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 221.58, cls_loss 0.0325 cls_loss_mapping 0.0469 cls_loss_causal 0.8302 re_mapping 0.0234 re_causal 0.0627 /// teacc 98.26 lr 0.00010000 +Epoch 31, weight, value: tensor([[ 0.0217, 0.0110, -0.0081, ..., 0.0256, -0.0594, -0.0396], + [ 0.0357, -0.0185, 0.0263, ..., -0.0343, 0.0282, -0.0649], + [ 0.0147, -0.0048, 0.0550, ..., -0.0586, -0.0288, -0.0641], + ..., + [-0.0054, 0.0062, -0.0237, ..., 0.0205, 0.0517, 0.0460], + [ 0.0095, -0.0202, -0.0232, ..., -0.0644, 0.0374, -0.0565], + [-0.0042, -0.0258, -0.0192, ..., 0.0127, -0.0117, -0.0006]], + device='cuda:0'), grad: tensor([[ 1.1232e-06, 0.0000e+00, -1.0803e-05, ..., 2.2483e-04, + 3.5143e-04, 3.7313e-05], + [ 1.1526e-05, 0.0000e+00, 3.7766e-04, ..., 9.7811e-05, + 1.0741e-04, 9.9182e-05], + [ 5.5507e-06, 0.0000e+00, 8.0442e-04, ..., 2.3246e-05, + 1.8950e-03, 4.2230e-05], + ..., + [-5.5879e-05, 0.0000e+00, -5.8651e-04, ..., 9.1970e-05, + -2.1095e-03, -7.4804e-05], + [ 4.2357e-06, 0.0000e+00, 1.0014e-03, ..., -3.8791e-04, + 3.5620e-04, -2.4363e-05], + [ 1.5408e-05, 0.0000e+00, 3.6240e-05, ..., 1.2102e-03, + 4.7231e-04, 9.9468e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0240, 0.0165, 0.0256, 0.0075, 0.0195, -0.0145, -0.0101, -0.0223, + -0.0060, 0.0259], device='cuda:0'), grad: tensor([ 0.0013, 0.0010, 0.0060, -0.0034, -0.0015, -0.0001, -0.0015, -0.0057, + 0.0010, 0.0030], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 221.57, cls_loss 0.0258 cls_loss_mapping 0.0388 cls_loss_causal 0.8306 re_mapping 0.0225 re_causal 0.0662 /// teacc 98.37 lr 0.00010000 +Epoch 32, weight, value: tensor([[ 0.0216, 0.0110, -0.0082, ..., 0.0255, -0.0604, -0.0407], + [ 0.0358, -0.0185, 0.0260, ..., -0.0354, 0.0286, -0.0658], + [ 0.0145, -0.0048, 0.0556, ..., -0.0594, -0.0294, -0.0651], + ..., + [-0.0049, 0.0062, -0.0239, ..., 0.0198, 0.0523, 0.0458], + [ 0.0097, -0.0202, -0.0238, ..., -0.0655, 0.0378, -0.0573], + [-0.0046, -0.0258, -0.0192, ..., 0.0132, -0.0121, 0.0003]], + device='cuda:0'), grad: tensor([[ 1.4585e-06, 0.0000e+00, -1.5354e-03, ..., -1.2875e-03, + -4.7982e-05, 1.0468e-05], + [ 2.5798e-06, 0.0000e+00, -3.9116e-06, ..., 2.7448e-05, + -1.2913e-03, 7.1526e-06], + [ 1.1362e-05, 0.0000e+00, 6.1369e-04, ..., 2.4021e-04, + 3.3522e-04, 1.6212e-05], + ..., + [-4.1753e-05, 0.0000e+00, 1.4424e-04, ..., 7.4983e-05, + 1.9729e-05, -2.2262e-05], + [ 6.0201e-06, 0.0000e+00, 2.2376e-04, ..., 1.5807e-04, + 4.9019e-04, 7.1645e-05], + [ 9.0227e-06, 0.0000e+00, 9.1016e-05, ..., 1.8924e-05, + 5.6803e-05, -2.8539e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0244, 0.0164, 0.0255, 0.0073, 0.0198, -0.0143, -0.0100, -0.0224, + -0.0060, 0.0260], device='cuda:0'), grad: tensor([-2.7599e-03, -1.2798e-03, 1.2302e-03, 5.4359e-04, 4.4727e-04, + 5.7793e-03, -5.3558e-03, 2.4843e-04, 1.1110e-03, 3.6448e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 221.33, cls_loss 0.0327 cls_loss_mapping 0.0468 cls_loss_causal 0.8291 re_mapping 0.0218 re_causal 0.0600 /// teacc 98.49 lr 0.00010000 +Epoch 33, weight, value: tensor([[ 0.0214, 0.0110, -0.0080, ..., 0.0259, -0.0604, -0.0421], + [ 0.0356, -0.0185, 0.0254, ..., -0.0368, 0.0291, -0.0677], + [ 0.0144, -0.0048, 0.0567, ..., -0.0603, -0.0301, -0.0656], + ..., + [-0.0046, 0.0062, -0.0246, ..., 0.0193, 0.0529, 0.0462], + [ 0.0096, -0.0202, -0.0243, ..., -0.0668, 0.0381, -0.0596], + [-0.0049, -0.0258, -0.0197, ..., 0.0139, -0.0126, 0.0015]], + device='cuda:0'), grad: tensor([[ 1.1221e-05, 0.0000e+00, -5.4836e-04, ..., -1.3323e-03, + -3.9768e-04, 8.7976e-05], + [-6.9320e-05, 0.0000e+00, 1.1110e-04, ..., 3.2246e-05, + -1.8227e-04, 3.5614e-05], + [ 3.3855e-05, 0.0000e+00, -2.9278e-04, ..., 6.9916e-05, + 8.5533e-05, 4.0352e-05], + ..., + [-1.2457e-04, 0.0000e+00, 4.4966e-04, ..., 7.7629e-04, + 3.6407e-04, 1.9073e-04], + [ 1.5289e-05, 0.0000e+00, 1.2958e-04, ..., 1.4901e-04, + -9.8705e-05, 4.7588e-04], + [ 3.7014e-05, 0.0000e+00, 6.2227e-05, ..., -5.9098e-05, + 7.9155e-05, -1.1349e-03]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0244, 0.0159, 0.0259, 0.0074, 0.0196, -0.0148, -0.0100, -0.0222, + -0.0061, 0.0264], device='cuda:0'), grad: tensor([-1.9274e-03, -3.3289e-05, -1.1533e-04, -1.0386e-03, 1.5020e-04, + 1.7385e-03, 1.6558e-04, 1.6489e-03, 7.3481e-04, -1.3266e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 222.22, cls_loss 0.0275 cls_loss_mapping 0.0410 cls_loss_causal 0.8122 re_mapping 0.0222 re_causal 0.0626 /// teacc 98.61 lr 0.00010000 +Epoch 34, weight, value: tensor([[ 0.0212, 0.0110, -0.0077, ..., 0.0267, -0.0606, -0.0434], + [ 0.0360, -0.0185, 0.0256, ..., -0.0376, 0.0293, -0.0681], + [ 0.0142, -0.0048, 0.0570, ..., -0.0614, -0.0308, -0.0668], + ..., + [-0.0043, 0.0062, -0.0256, ..., 0.0189, 0.0534, 0.0462], + [ 0.0095, -0.0202, -0.0247, ..., -0.0677, 0.0385, -0.0609], + [-0.0056, -0.0258, -0.0201, ..., 0.0138, -0.0130, 0.0017]], + device='cuda:0'), grad: tensor([[ 9.9316e-06, 0.0000e+00, 1.1998e-04, ..., 1.9324e-04, + 4.2248e-04, 2.2233e-04], + [ 3.2395e-05, 0.0000e+00, 4.6039e-04, ..., 2.3532e-04, + 1.0773e-02, 3.6359e-04], + [-1.3578e-04, 0.0000e+00, -1.5335e-03, ..., 5.5933e-04, + 2.5845e-04, 6.1369e-04], + ..., + [ 3.0667e-05, 0.0000e+00, 9.5654e-04, ..., -4.3154e-05, + -3.9864e-04, -3.1972e-04], + [ 7.1675e-06, 0.0000e+00, 5.7071e-05, ..., 9.0837e-05, + 2.8191e-03, 1.7476e-04], + [ 3.1203e-05, 0.0000e+00, 6.9141e-05, ..., 5.3787e-04, + 1.9372e-04, 8.2827e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0238, 0.0160, 0.0255, 0.0076, 0.0201, -0.0146, -0.0101, -0.0222, + -0.0062, 0.0259], device='cuda:0'), grad: tensor([ 0.0008, 0.0124, -0.0003, -0.0003, -0.0025, -0.0072, -0.0080, 0.0003, + 0.0035, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 221.35, cls_loss 0.0297 cls_loss_mapping 0.0411 cls_loss_causal 0.8048 re_mapping 0.0213 re_causal 0.0607 /// teacc 98.55 lr 0.00010000 +Epoch 35, weight, value: tensor([[ 0.0211, 0.0110, -0.0076, ..., 0.0274, -0.0610, -0.0443], + [ 0.0359, -0.0185, 0.0250, ..., -0.0384, 0.0296, -0.0684], + [ 0.0141, -0.0048, 0.0580, ..., -0.0619, -0.0313, -0.0684], + ..., + [-0.0038, 0.0062, -0.0259, ..., 0.0188, 0.0537, 0.0461], + [ 0.0096, -0.0202, -0.0251, ..., -0.0693, 0.0387, -0.0612], + [-0.0058, -0.0258, -0.0203, ..., 0.0139, -0.0132, 0.0025]], + device='cuda:0'), grad: tensor([[ 3.3826e-06, 0.0000e+00, -2.4974e-05, ..., -3.4988e-05, + 3.6329e-05, 3.0100e-05], + [ 3.3110e-05, 0.0000e+00, -7.2896e-05, ..., 1.0371e-05, + -4.3058e-04, 4.7028e-05], + [ 1.3605e-05, 0.0000e+00, -1.9640e-05, ..., 6.8620e-06, + 1.6665e-04, 3.3200e-05], + ..., + [-1.2672e-04, 0.0000e+00, 5.1558e-05, ..., 9.2983e-06, + -1.0890e-04, -2.0039e-04], + [ 8.0168e-06, 0.0000e+00, 2.2307e-05, ..., 4.3392e-05, + 1.3280e-04, 7.8142e-05], + [ 4.3243e-05, 0.0000e+00, 2.2743e-06, ..., -6.7726e-06, + 1.0192e-04, 1.7434e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0238, 0.0156, 0.0258, 0.0075, 0.0201, -0.0146, -0.0099, -0.0222, + -0.0063, 0.0261], device='cuda:0'), grad: tensor([ 4.1500e-06, -4.8256e-04, 1.6177e-04, 9.9301e-05, 5.4979e-04, + -5.3644e-05, -4.7350e-04, -1.1027e-04, 2.2221e-04, 8.2076e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 221.46, cls_loss 0.0279 cls_loss_mapping 0.0410 cls_loss_causal 0.7963 re_mapping 0.0206 re_causal 0.0598 /// teacc 98.49 lr 0.00010000 +Epoch 36, weight, value: tensor([[ 0.0210, 0.0110, -0.0080, ..., 0.0273, -0.0615, -0.0457], + [ 0.0356, -0.0185, 0.0248, ..., -0.0400, 0.0305, -0.0698], + [ 0.0139, -0.0048, 0.0587, ..., -0.0624, -0.0317, -0.0701], + ..., + [-0.0037, 0.0062, -0.0264, ..., 0.0186, 0.0543, 0.0475], + [ 0.0094, -0.0202, -0.0256, ..., -0.0700, 0.0385, -0.0623], + [-0.0061, -0.0258, -0.0201, ..., 0.0150, -0.0136, 0.0029]], + device='cuda:0'), grad: tensor([[ 4.1388e-06, 0.0000e+00, 9.4414e-05, ..., -3.0541e-04, + 1.0721e-05, 1.9893e-05], + [ 5.3763e-05, 0.0000e+00, 3.1859e-05, ..., 4.2230e-05, + -1.2219e-04, 2.8419e-04], + [ 1.2986e-05, 0.0000e+00, 2.2888e-05, ..., 8.9645e-05, + 6.4015e-05, 5.5045e-05], + ..., + [-2.4772e-04, 0.0000e+00, 9.1434e-05, ..., 3.0547e-05, + -3.5977e-04, -4.5371e-04], + [ 1.2785e-05, 0.0000e+00, 1.1945e-04, ..., 3.5673e-05, + -1.7095e-04, 1.5914e-04], + [ 8.1837e-05, 0.0000e+00, 7.3814e-04, ..., 1.8156e-04, + 1.7297e-04, 1.2417e-03]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0242, 0.0162, 0.0259, 0.0072, 0.0198, -0.0140, -0.0102, -0.0221, + -0.0067, 0.0261], device='cuda:0'), grad: tensor([ 0.0001, 0.0002, 0.0003, -0.0040, -0.0009, 0.0011, 0.0003, -0.0006, + -0.0003, 0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 221.25, cls_loss 0.0231 cls_loss_mapping 0.0361 cls_loss_causal 0.7591 re_mapping 0.0202 re_causal 0.0576 /// teacc 98.47 lr 0.00010000 +Epoch 37, weight, value: tensor([[ 0.0209, 0.0110, -0.0080, ..., 0.0280, -0.0618, -0.0471], + [ 0.0355, -0.0185, 0.0247, ..., -0.0416, 0.0306, -0.0716], + [ 0.0138, -0.0048, 0.0593, ..., -0.0636, -0.0324, -0.0707], + ..., + [-0.0031, 0.0062, -0.0273, ..., 0.0182, 0.0551, 0.0487], + [ 0.0094, -0.0202, -0.0255, ..., -0.0707, 0.0389, -0.0636], + [-0.0067, -0.0258, -0.0205, ..., 0.0152, -0.0143, 0.0030]], + device='cuda:0'), grad: tensor([[ 3.6601e-06, 0.0000e+00, 1.0663e-04, ..., 1.8871e-04, + 2.4498e-05, 5.2691e-05], + [-3.0935e-05, 0.0000e+00, 2.4453e-05, ..., 2.0528e-04, + -4.4554e-05, 3.5834e-04], + [ 1.3672e-05, 0.0000e+00, -1.7917e-04, ..., 7.6830e-05, + 5.2929e-05, 1.0014e-04], + ..., + [-5.0992e-05, 0.0000e+00, 8.2791e-05, ..., 3.8552e-04, + 5.5969e-05, 6.6805e-04], + [ 1.3210e-05, 0.0000e+00, -1.0008e-04, ..., -2.0146e-04, + -6.5267e-05, 3.1090e-04], + [ 2.0996e-05, 0.0000e+00, 3.3498e-05, ..., -2.2018e-04, + 3.5256e-05, -1.0004e-03]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0242, 0.0155, 0.0260, 0.0075, 0.0196, -0.0144, -0.0096, -0.0217, + -0.0064, 0.0257], device='cuda:0'), grad: tensor([ 8.9741e-04, 2.6846e-04, 7.4029e-05, -4.0174e-04, -6.6996e-04, + 2.2542e-04, 2.3985e-04, 9.4557e-04, -1.0395e-03, -5.3930e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 221.31, cls_loss 0.0217 cls_loss_mapping 0.0318 cls_loss_causal 0.7901 re_mapping 0.0204 re_causal 0.0577 /// teacc 98.53 lr 0.00010000 +Epoch 38, weight, value: tensor([[ 0.0209, 0.0110, -0.0083, ..., 0.0287, -0.0623, -0.0481], + [ 0.0356, -0.0185, 0.0248, ..., -0.0425, 0.0307, -0.0714], + [ 0.0136, -0.0048, 0.0598, ..., -0.0652, -0.0331, -0.0721], + ..., + [-0.0026, 0.0062, -0.0278, ..., 0.0180, 0.0557, 0.0490], + [ 0.0092, -0.0202, -0.0259, ..., -0.0718, 0.0392, -0.0647], + [-0.0070, -0.0258, -0.0197, ..., 0.0155, -0.0146, 0.0032]], + device='cuda:0'), grad: tensor([[ 1.6168e-06, 0.0000e+00, -3.8147e-05, ..., -9.4593e-05, + 6.7532e-05, 9.9242e-06], + [ 6.4597e-06, 0.0000e+00, -1.5348e-05, ..., 1.6883e-05, + -3.9029e-04, 2.4378e-05], + [ 2.9728e-06, 0.0000e+00, 1.4037e-05, ..., 2.8387e-05, + 9.1970e-05, 8.1733e-06], + ..., + [-2.4408e-05, 0.0000e+00, 5.4576e-06, ..., 2.8297e-05, + 2.6241e-05, 1.5736e-05], + [ 9.1968e-07, 0.0000e+00, 1.0163e-05, ..., 3.2687e-04, + 1.3647e-03, 2.9922e-04], + [ 7.0408e-06, 0.0000e+00, 2.1011e-05, ..., 4.9889e-05, + 2.9087e-05, -2.7514e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0244, 0.0156, 0.0259, 0.0074, 0.0198, -0.0140, -0.0099, -0.0213, + -0.0066, 0.0256], device='cuda:0'), grad: tensor([ 1.9252e-05, -4.6992e-04, 1.7154e-04, -1.2386e-04, -2.7701e-05, + 3.0422e-04, -1.9836e-03, 7.4089e-05, 2.1667e-03, -1.2994e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 222.54, cls_loss 0.0223 cls_loss_mapping 0.0316 cls_loss_causal 0.7496 re_mapping 0.0199 re_causal 0.0549 /// teacc 98.63 lr 0.00010000 +Epoch 39, weight, value: tensor([[ 0.0208, 0.0110, -0.0080, ..., 0.0299, -0.0634, -0.0482], + [ 0.0355, -0.0185, 0.0248, ..., -0.0437, 0.0308, -0.0721], + [ 0.0135, -0.0048, 0.0604, ..., -0.0659, -0.0334, -0.0734], + ..., + [-0.0023, 0.0062, -0.0278, ..., 0.0177, 0.0561, 0.0488], + [ 0.0091, -0.0202, -0.0262, ..., -0.0735, 0.0395, -0.0661], + [-0.0073, -0.0258, -0.0209, ..., 0.0155, -0.0147, 0.0040]], + device='cuda:0'), grad: tensor([[ 3.1553e-06, 0.0000e+00, 1.2450e-05, ..., 5.7518e-05, + 6.3896e-05, 5.1945e-05], + [ 5.9992e-05, 0.0000e+00, 2.4661e-06, ..., 7.3791e-05, + 3.2139e-04, 5.0020e-04], + [ 1.2830e-05, 0.0000e+00, 5.1528e-05, ..., 9.3654e-06, + 6.1631e-05, 9.3997e-05], + ..., + [-1.7858e-04, 0.0000e+00, 1.9014e-05, ..., -4.7922e-05, + -7.9107e-04, -9.1743e-04], + [ 6.2510e-06, 0.0000e+00, 3.9428e-05, ..., 4.3005e-05, + 6.0260e-05, 2.0719e-04], + [ 4.2528e-05, 0.0000e+00, 9.2804e-05, ..., -4.1032e-04, + 8.7142e-05, -9.7656e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0240, 0.0155, 0.0258, 0.0077, 0.0196, -0.0146, -0.0094, -0.0212, + -0.0069, 0.0257], device='cuda:0'), grad: tensor([ 2.3484e-04, 6.0558e-04, 3.0184e-04, -5.6362e-04, 8.0252e-04, + 9.3520e-05, -1.1867e-04, -1.8206e-03, 4.6325e-04, 2.8266e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 221.38, cls_loss 0.0190 cls_loss_mapping 0.0329 cls_loss_causal 0.7338 re_mapping 0.0195 re_causal 0.0560 /// teacc 98.49 lr 0.00010000 +Epoch 40, weight, value: tensor([[ 0.0207, 0.0110, -0.0080, ..., 0.0302, -0.0631, -0.0494], + [ 0.0354, -0.0185, 0.0245, ..., -0.0448, 0.0312, -0.0738], + [ 0.0134, -0.0048, 0.0609, ..., -0.0668, -0.0337, -0.0733], + ..., + [-0.0019, 0.0062, -0.0283, ..., 0.0172, 0.0565, 0.0487], + [ 0.0091, -0.0202, -0.0265, ..., -0.0743, 0.0397, -0.0667], + [-0.0076, -0.0258, -0.0208, ..., 0.0160, -0.0150, 0.0044]], + device='cuda:0'), grad: tensor([[ 9.3831e-07, 0.0000e+00, -3.0193e-06, ..., 1.3575e-05, + 3.7313e-05, 2.5958e-05], + [ 7.4208e-06, 0.0000e+00, 1.3463e-05, ..., 1.5676e-05, + -1.5316e-03, -4.3362e-05], + [ 7.4729e-06, 0.0000e+00, -4.4632e-04, ..., 1.2465e-05, + 1.3363e-04, 6.1810e-05], + ..., + [-1.0180e-04, 0.0000e+00, 3.1686e-04, ..., 1.6361e-05, + -1.9383e-04, -2.6941e-04], + [ 6.9924e-06, 0.0000e+00, 3.0965e-05, ..., 2.7925e-05, + 6.2943e-04, 8.1420e-05], + [ 3.9667e-05, 0.0000e+00, 7.8678e-06, ..., 3.0845e-05, + 2.3425e-04, -9.1195e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0239, 0.0151, 0.0259, 0.0078, 0.0198, -0.0146, -0.0097, -0.0214, + -0.0067, 0.0258], device='cuda:0'), grad: tensor([ 1.4639e-04, -1.9283e-03, -1.8048e-04, 5.2214e-04, 7.1859e-04, + -1.5678e-03, 1.1206e-03, -6.1452e-05, 9.8038e-04, 2.4891e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 221.22, cls_loss 0.0224 cls_loss_mapping 0.0321 cls_loss_causal 0.7665 re_mapping 0.0189 re_causal 0.0549 /// teacc 98.62 lr 0.00010000 +Epoch 41, weight, value: tensor([[ 0.0207, 0.0110, -0.0080, ..., 0.0306, -0.0637, -0.0504], + [ 0.0353, -0.0185, 0.0248, ..., -0.0454, 0.0316, -0.0754], + [ 0.0131, -0.0048, 0.0617, ..., -0.0678, -0.0340, -0.0742], + ..., + [-0.0014, 0.0062, -0.0293, ..., 0.0170, 0.0568, 0.0493], + [ 0.0092, -0.0202, -0.0275, ..., -0.0750, 0.0401, -0.0666], + [-0.0082, -0.0258, -0.0207, ..., 0.0161, -0.0154, 0.0051]], + device='cuda:0'), grad: tensor([[ 9.5740e-07, 0.0000e+00, -3.0413e-05, ..., -3.1769e-05, + 4.8906e-05, 1.6138e-05], + [ 1.2763e-05, 0.0000e+00, -2.1145e-05, ..., 7.0408e-06, + 1.2386e-04, 1.4856e-05], + [ 4.7162e-06, 0.0000e+00, 8.6367e-05, ..., 6.9328e-06, + 1.8919e-04, 1.5073e-05], + ..., + [-5.8591e-05, 0.0000e+00, 5.5321e-06, ..., 1.5363e-05, + -1.1587e-03, 4.8459e-05], + [ 1.8060e-05, 0.0000e+00, -5.8681e-05, ..., 4.1723e-05, + 3.0375e-04, 6.1333e-05], + [ 8.1360e-06, 0.0000e+00, 1.2502e-05, ..., 2.5928e-05, + 5.6088e-05, -8.8632e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0243, 0.0152, 0.0261, 0.0080, 0.0197, -0.0144, -0.0095, -0.0217, + -0.0071, 0.0260], device='cuda:0'), grad: tensor([ 5.0902e-05, 5.0813e-05, 4.4274e-04, 3.9530e-04, -3.3349e-05, + 2.1434e-04, 1.0878e-04, -1.9159e-03, 7.0667e-04, -1.9491e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 222.08, cls_loss 0.0166 cls_loss_mapping 0.0273 cls_loss_causal 0.7581 re_mapping 0.0186 re_causal 0.0559 /// teacc 98.54 lr 0.00010000 +Epoch 42, weight, value: tensor([[ 0.0207, 0.0110, -0.0079, ..., 0.0310, -0.0645, -0.0512], + [ 0.0353, -0.0185, 0.0248, ..., -0.0457, 0.0318, -0.0758], + [ 0.0129, -0.0048, 0.0622, ..., -0.0685, -0.0350, -0.0760], + ..., + [-0.0009, 0.0062, -0.0299, ..., 0.0166, 0.0573, 0.0500], + [ 0.0092, -0.0202, -0.0275, ..., -0.0756, 0.0407, -0.0672], + [-0.0085, -0.0258, -0.0211, ..., 0.0159, -0.0156, 0.0053]], + device='cuda:0'), grad: tensor([[ 2.6403e-07, 0.0000e+00, -1.8253e-03, ..., -1.1683e-03, + -1.2894e-03, 1.8790e-05], + [ 4.3362e-06, 0.0000e+00, 7.0751e-05, ..., 1.5771e-04, + -7.7605e-05, 2.9278e-04], + [ 1.5292e-06, 0.0000e+00, 5.2452e-04, ..., 3.6478e-04, + 4.8804e-04, 4.4942e-05], + ..., + [-1.0654e-05, 0.0000e+00, 3.2812e-05, ..., 2.8539e-04, + 3.1900e-04, 1.2455e-03], + [ 9.9652e-07, 0.0000e+00, 1.0948e-03, ..., 6.7949e-04, + 7.5865e-04, 5.6118e-05], + [ 8.3297e-06, 0.0000e+00, 4.0829e-05, ..., 3.2234e-04, + 3.3569e-04, 1.2856e-03]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0244, 0.0150, 0.0262, 0.0079, 0.0200, -0.0145, -0.0096, -0.0215, + -0.0069, 0.0259], device='cuda:0'), grad: tensor([-4.0054e-03, 3.0231e-04, 1.3380e-03, -5.7042e-05, -2.6588e-03, + 1.8287e-04, 2.5034e-04, 1.0500e-03, 2.4166e-03, 1.1768e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 220.86, cls_loss 0.0204 cls_loss_mapping 0.0310 cls_loss_causal 0.7203 re_mapping 0.0173 re_causal 0.0509 /// teacc 98.63 lr 0.00010000 +Epoch 43, weight, value: tensor([[ 0.0206, 0.0110, -0.0075, ..., 0.0317, -0.0646, -0.0522], + [ 0.0358, -0.0185, 0.0246, ..., -0.0465, 0.0322, -0.0756], + [ 0.0125, -0.0048, 0.0629, ..., -0.0692, -0.0355, -0.0775], + ..., + [-0.0006, 0.0062, -0.0305, ..., 0.0162, 0.0578, 0.0504], + [ 0.0091, -0.0202, -0.0281, ..., -0.0764, 0.0418, -0.0674], + [-0.0090, -0.0258, -0.0215, ..., 0.0165, -0.0161, 0.0056]], + device='cuda:0'), grad: tensor([[ 1.7490e-06, 0.0000e+00, 3.2037e-05, ..., 6.3516e-06, + 2.2781e-04, 2.0370e-05], + [ 8.6129e-06, 0.0000e+00, -9.4593e-05, ..., 1.0189e-06, + -8.0824e-04, 2.5928e-05], + [ 5.4836e-06, 0.0000e+00, -7.4327e-05, ..., 1.0803e-06, + 1.0765e-04, 2.9549e-05], + ..., + [-2.2388e-04, 0.0000e+00, 3.7670e-05, ..., 5.1372e-06, + -3.3355e-04, -4.3654e-04], + [ 1.8060e-05, 0.0000e+00, 2.5854e-05, ..., 7.4096e-06, + -4.3678e-04, -7.5293e-04], + [ 1.6963e-04, 0.0000e+00, 8.1360e-06, ..., 4.4852e-06, + 8.2636e-04, 9.7847e-04]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0243, 0.0149, 0.0261, 0.0078, 0.0197, -0.0149, -0.0100, -0.0214, + -0.0060, 0.0257], device='cuda:0'), grad: tensor([ 0.0004, -0.0013, 0.0001, 0.0002, 0.0002, -0.0001, 0.0004, -0.0005, + -0.0010, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 221.76, cls_loss 0.0227 cls_loss_mapping 0.0336 cls_loss_causal 0.7525 re_mapping 0.0178 re_causal 0.0503 /// teacc 98.74 lr 0.00010000 +Epoch 44, weight, value: tensor([[ 2.0485e-02, 1.0998e-02, -7.6998e-03, ..., 3.1724e-02, + -6.4928e-02, -5.3661e-02], + [ 3.5695e-02, -1.8468e-02, 2.4391e-02, ..., -4.7630e-02, + 3.2790e-02, -7.6811e-02], + [ 1.2724e-02, -4.7728e-03, 6.3436e-02, ..., -6.9899e-02, + -3.5511e-02, -7.7433e-02], + ..., + [-1.8292e-06, 6.2061e-03, -3.0748e-02, ..., 1.5783e-02, + 5.7962e-02, 5.0457e-02], + [ 9.1030e-03, -2.0220e-02, -2.8425e-02, ..., -7.7319e-02, + 4.1786e-02, -6.8335e-02], + [-9.6823e-03, -2.5834e-02, -2.1565e-02, ..., 1.6975e-02, + -1.6651e-02, 5.7066e-03]], device='cuda:0'), grad: tensor([[ 9.9279e-07, 0.0000e+00, -2.6059e-04, ..., -3.9124e-04, + -3.8117e-05, 3.1948e-05], + [-5.5507e-06, 0.0000e+00, 2.0102e-05, ..., 1.1362e-05, + -3.6383e-04, -6.4727e-08], + [ 1.0401e-05, 0.0000e+00, -2.0826e-04, ..., 1.2302e-04, + 3.9291e-04, -2.6870e-04], + ..., + [-5.6446e-05, 0.0000e+00, 3.2902e-04, ..., 4.2170e-05, + -3.4451e-05, -9.2983e-05], + [ 6.7949e-06, 0.0000e+00, 4.0591e-05, ..., 3.4958e-05, + -4.1318e-04, 8.7380e-05], + [ 1.5453e-05, 0.0000e+00, 2.6047e-05, ..., 1.0985e-04, + 1.1992e-04, 5.9456e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0247, 0.0147, 0.0263, 0.0079, 0.0204, -0.0144, -0.0099, -0.0214, + -0.0068, 0.0257], device='cuda:0'), grad: tensor([-0.0005, -0.0004, -0.0024, 0.0005, 0.0004, 0.0005, 0.0003, 0.0008, + -0.0002, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 220.96, cls_loss 0.0149 cls_loss_mapping 0.0254 cls_loss_causal 0.7235 re_mapping 0.0176 re_causal 0.0511 /// teacc 98.66 lr 0.00010000 +Epoch 45, weight, value: tensor([[ 0.0204, 0.0110, -0.0074, ..., 0.0325, -0.0656, -0.0550], + [ 0.0357, -0.0185, 0.0242, ..., -0.0481, 0.0328, -0.0773], + [ 0.0125, -0.0048, 0.0640, ..., -0.0706, -0.0361, -0.0779], + ..., + [ 0.0005, 0.0062, -0.0313, ..., 0.0153, 0.0587, 0.0504], + [ 0.0090, -0.0202, -0.0282, ..., -0.0783, 0.0421, -0.0692], + [-0.0099, -0.0258, -0.0223, ..., 0.0170, -0.0170, 0.0063]], + device='cuda:0'), grad: tensor([[ 1.4789e-06, 0.0000e+00, 6.6519e-05, ..., 8.6308e-05, + 1.9103e-05, 5.0753e-05], + [ 1.6009e-06, 0.0000e+00, 2.4036e-05, ..., 1.0483e-05, + 1.6391e-05, 1.5706e-05], + [ 3.4366e-06, 0.0000e+00, -2.5439e-04, ..., 1.7953e-04, + 1.2362e-04, 7.3612e-05], + ..., + [-3.6001e-05, 0.0000e+00, 9.9301e-05, ..., 3.3170e-05, + -5.1677e-05, -3.5048e-05], + [ 9.1689e-07, 0.0000e+00, 1.3900e-04, ..., 2.8297e-05, + -1.2951e-03, -1.7762e-04], + [ 2.4974e-05, 0.0000e+00, -4.7708e-04, ..., -3.5000e-04, + 5.3215e-04, 1.3733e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0245, 0.0145, 0.0265, 0.0076, 0.0203, -0.0144, -0.0098, -0.0215, + -0.0068, 0.0259], device='cuda:0'), grad: tensor([ 2.2173e-04, 7.9572e-05, 1.1265e-04, 3.6621e-04, 8.6904e-05, + 8.5020e-04, 1.3006e-04, 1.2803e-04, -2.3232e-03, 3.4881e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 221.15, cls_loss 0.0153 cls_loss_mapping 0.0254 cls_loss_causal 0.7167 re_mapping 0.0170 re_causal 0.0501 /// teacc 98.53 lr 0.00010000 +Epoch 46, weight, value: tensor([[ 0.0204, 0.0110, -0.0068, ..., 0.0333, -0.0665, -0.0565], + [ 0.0357, -0.0185, 0.0244, ..., -0.0487, 0.0331, -0.0780], + [ 0.0124, -0.0048, 0.0645, ..., -0.0716, -0.0366, -0.0792], + ..., + [ 0.0007, 0.0062, -0.0318, ..., 0.0150, 0.0591, 0.0506], + [ 0.0090, -0.0202, -0.0288, ..., -0.0791, 0.0423, -0.0697], + [-0.0101, -0.0258, -0.0228, ..., 0.0169, -0.0177, 0.0066]], + device='cuda:0'), grad: tensor([[ 4.0838e-07, 0.0000e+00, -7.6830e-05, ..., -1.8930e-04, + 3.5495e-05, 5.8234e-05], + [-1.7121e-05, 0.0000e+00, -9.1934e-04, ..., 2.9922e-05, + -1.1673e-03, 1.7285e-04], + [ 1.2890e-06, 0.0000e+00, -1.1606e-03, ..., 1.2350e-04, + -6.2525e-05, 8.7798e-05], + ..., + [ 5.4650e-06, 0.0000e+00, 1.4858e-03, ..., 2.9638e-05, + 3.3474e-04, -4.5204e-04], + [ 5.8021e-07, 0.0000e+00, 3.0994e-04, ..., 6.3300e-05, + 1.7262e-04, 2.1529e-04], + [ 1.9129e-06, 0.0000e+00, 6.1691e-05, ..., 2.2614e-04, + 2.3019e-04, 8.5878e-04]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0244, 0.0148, 0.0263, 0.0074, 0.0204, -0.0142, -0.0096, -0.0213, + -0.0071, 0.0258], device='cuda:0'), grad: tensor([-4.4912e-05, -2.5501e-03, -1.8139e-03, -1.5936e-03, 1.4365e-04, + 2.6083e-04, 1.7858e-04, 2.6569e-03, 7.8535e-04, 1.9760e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 221.05, cls_loss 0.0172 cls_loss_mapping 0.0281 cls_loss_causal 0.7054 re_mapping 0.0167 re_causal 0.0483 /// teacc 98.56 lr 0.00010000 +Epoch 47, weight, value: tensor([[ 0.0204, 0.0110, -0.0071, ..., 0.0332, -0.0671, -0.0570], + [ 0.0356, -0.0185, 0.0243, ..., -0.0497, 0.0341, -0.0762], + [ 0.0123, -0.0048, 0.0649, ..., -0.0732, -0.0373, -0.0807], + ..., + [ 0.0011, 0.0062, -0.0324, ..., 0.0151, 0.0594, 0.0504], + [ 0.0090, -0.0202, -0.0292, ..., -0.0799, 0.0428, -0.0703], + [-0.0103, -0.0258, -0.0228, ..., 0.0173, -0.0183, 0.0069]], + device='cuda:0'), grad: tensor([[ 7.7672e-07, 0.0000e+00, 8.4341e-05, ..., -7.1645e-05, + 1.4991e-05, 5.6326e-06], + [ 1.7241e-05, 0.0000e+00, 2.8327e-05, ..., 3.5856e-06, + -9.0450e-06, 6.3002e-05], + [ 4.3921e-06, 0.0000e+00, 1.0586e-03, ..., 5.0366e-06, + 5.1469e-05, 1.2553e-04], + ..., + [-4.7237e-05, 0.0000e+00, 9.3699e-05, ..., 5.7034e-06, + -1.4102e-04, -6.5851e-04], + [ 1.7630e-06, 0.0000e+00, 6.6662e-04, ..., 1.0088e-05, + -2.9027e-05, -2.1309e-05], + [ 1.6391e-05, 0.0000e+00, 8.1718e-05, ..., 4.9248e-06, + 8.7321e-05, 1.5998e-04]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0245, 0.0156, 0.0259, 0.0079, 0.0201, -0.0143, -0.0097, -0.0216, + -0.0069, 0.0257], device='cuda:0'), grad: tensor([ 2.0671e-04, 9.5069e-05, 2.5558e-03, 3.5381e-03, 3.3879e-04, + -7.2594e-03, -9.5272e-04, -5.4026e-04, 1.5812e-03, 4.3774e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 221.09, cls_loss 0.0193 cls_loss_mapping 0.0294 cls_loss_causal 0.7114 re_mapping 0.0168 re_causal 0.0463 /// teacc 98.45 lr 0.00010000 +Epoch 48, weight, value: tensor([[ 0.0203, 0.0110, -0.0073, ..., 0.0338, -0.0676, -0.0579], + [ 0.0350, -0.0185, 0.0238, ..., -0.0506, 0.0347, -0.0770], + [ 0.0122, -0.0048, 0.0657, ..., -0.0748, -0.0383, -0.0819], + ..., + [ 0.0019, 0.0062, -0.0325, ..., 0.0150, 0.0602, 0.0510], + [ 0.0090, -0.0202, -0.0297, ..., -0.0804, 0.0426, -0.0714], + [-0.0105, -0.0258, -0.0231, ..., 0.0171, -0.0187, 0.0070]], + device='cuda:0'), grad: tensor([[ 3.4506e-07, 0.0000e+00, 1.3404e-05, ..., -9.7007e-06, + 1.7285e-05, 9.9242e-06], + [ 3.0641e-07, 0.0000e+00, 4.1723e-04, ..., 1.4469e-05, + 2.4843e-04, 1.0830e-04], + [ 1.6345e-07, 0.0000e+00, 4.2558e-04, ..., 4.0568e-06, + 3.8576e-04, 1.4663e-04], + ..., + [-2.7306e-06, 0.0000e+00, -2.1267e-03, ..., 3.5375e-05, + -1.7147e-03, -5.5075e-04], + [ 1.3644e-07, 0.0000e+00, 8.0407e-05, ..., 2.2531e-05, + 6.6638e-05, 6.7890e-05], + [ 1.1930e-06, 0.0000e+00, 6.4850e-05, ..., 1.4048e-03, + 6.2656e-04, 3.9139e-03]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0248, 0.0155, 0.0258, 0.0078, 0.0204, -0.0140, -0.0100, -0.0209, + -0.0074, 0.0256], device='cuda:0'), grad: tensor([ 4.9740e-05, 1.0719e-03, 1.3313e-03, 3.1853e-03, -3.4676e-03, + 1.4949e-04, -1.6361e-05, -6.1417e-03, 2.6703e-04, 3.5763e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 221.18, cls_loss 0.0152 cls_loss_mapping 0.0219 cls_loss_causal 0.7474 re_mapping 0.0166 re_causal 0.0493 /// teacc 98.67 lr 0.00010000 +Epoch 49, weight, value: tensor([[ 0.0203, 0.0110, -0.0071, ..., 0.0344, -0.0679, -0.0587], + [ 0.0350, -0.0185, 0.0229, ..., -0.0512, 0.0345, -0.0779], + [ 0.0121, -0.0048, 0.0671, ..., -0.0758, -0.0383, -0.0826], + ..., + [ 0.0013, 0.0062, -0.0330, ..., 0.0156, 0.0607, 0.0516], + [ 0.0090, -0.0202, -0.0303, ..., -0.0809, 0.0426, -0.0715], + [-0.0107, -0.0258, -0.0235, ..., 0.0169, -0.0191, 0.0070]], + device='cuda:0'), grad: tensor([[ 1.8999e-06, 0.0000e+00, -9.5367e-05, ..., -1.9372e-04, + 6.4492e-05, 5.0962e-06], + [ 6.4671e-06, 0.0000e+00, 3.2991e-05, ..., 6.7651e-06, + -3.1300e-03, -3.9876e-05], + [ 5.1297e-06, 0.0000e+00, -2.0337e-04, ..., 1.2614e-05, + 1.0717e-04, 1.7032e-05], + ..., + [-3.7283e-05, 0.0000e+00, 2.4587e-05, ..., 1.3284e-05, + 6.2644e-05, -3.0726e-05], + [ 8.2050e-07, 0.0000e+00, 5.4270e-05, ..., 1.2957e-05, + 1.6956e-03, 1.5289e-05], + [ 8.7768e-06, 0.0000e+00, 6.8471e-06, ..., 2.7448e-05, + 1.4317e-04, -1.2171e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0249, 0.0146, 0.0265, 0.0079, 0.0204, -0.0138, -0.0097, -0.0205, + -0.0076, 0.0250], device='cuda:0'), grad: tensor([-5.4216e-04, -4.5280e-03, -8.1301e-05, 3.0112e-04, 7.6056e-04, + 2.2328e-04, 9.8991e-04, 1.4985e-04, 2.5311e-03, 1.9729e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 221.80, cls_loss 0.0167 cls_loss_mapping 0.0240 cls_loss_causal 0.7304 re_mapping 0.0166 re_causal 0.0471 /// teacc 98.60 lr 0.00010000 +Epoch 50, weight, value: tensor([[ 0.0202, 0.0110, -0.0068, ..., 0.0352, -0.0681, -0.0593], + [ 0.0349, -0.0185, 0.0234, ..., -0.0523, 0.0355, -0.0786], + [ 0.0122, -0.0048, 0.0671, ..., -0.0766, -0.0397, -0.0831], + ..., + [ 0.0017, 0.0062, -0.0331, ..., 0.0151, 0.0612, 0.0510], + [ 0.0089, -0.0202, -0.0305, ..., -0.0816, 0.0427, -0.0724], + [-0.0110, -0.0258, -0.0237, ..., 0.0171, -0.0196, 0.0067]], + device='cuda:0'), grad: tensor([[ 5.5693e-07, 0.0000e+00, 3.2365e-05, ..., 9.7975e-06, + 9.0003e-06, 6.4149e-06], + [-1.3271e-06, 0.0000e+00, 1.2420e-05, ..., 8.2403e-06, + -5.9068e-05, 7.4953e-06], + [-3.3714e-07, 0.0000e+00, 1.2589e-04, ..., 1.8864e-03, + 3.6955e-05, 8.1897e-05], + ..., + [-1.1742e-05, 0.0000e+00, 3.4362e-05, ..., 1.1578e-05, + -1.5870e-05, -1.9148e-06], + [ 7.4878e-07, 0.0000e+00, 7.3016e-05, ..., 1.1384e-05, + -8.9741e-04, 3.0294e-05], + [ 9.8124e-06, 0.0000e+00, 2.7132e-04, ..., 8.8960e-06, + 2.0862e-05, -1.3220e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0243, 0.0153, 0.0256, 0.0078, 0.0215, -0.0139, -0.0096, -0.0208, + -0.0081, 0.0250], device='cuda:0'), grad: tensor([ 6.1214e-05, -5.6207e-05, 1.4544e-03, 6.4278e-04, -2.6245e-03, + 9.3746e-04, 7.1907e-04, 4.2498e-05, -1.4038e-03, 2.2972e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 220.95, cls_loss 0.0171 cls_loss_mapping 0.0262 cls_loss_causal 0.7369 re_mapping 0.0155 re_causal 0.0479 /// teacc 98.58 lr 0.00010000 +Epoch 51, weight, value: tensor([[ 0.0201, 0.0110, -0.0068, ..., 0.0361, -0.0688, -0.0607], + [ 0.0347, -0.0185, 0.0241, ..., -0.0545, 0.0358, -0.0810], + [ 0.0122, -0.0048, 0.0677, ..., -0.0769, -0.0402, -0.0841], + ..., + [ 0.0016, 0.0062, -0.0343, ..., 0.0146, 0.0622, 0.0517], + [ 0.0086, -0.0202, -0.0307, ..., -0.0824, 0.0427, -0.0735], + [-0.0108, -0.0258, -0.0241, ..., 0.0174, -0.0202, 0.0077]], + device='cuda:0'), grad: tensor([[ 2.4959e-07, 0.0000e+00, -1.6227e-05, ..., -3.2120e-03, + -2.8715e-05, -1.0347e-03], + [ 2.5034e-06, 0.0000e+00, 2.1517e-05, ..., 1.5244e-05, + -8.9884e-05, 6.6459e-05], + [ 4.4238e-07, 0.0000e+00, 4.1342e-04, ..., 9.8801e-04, + 1.1206e-04, 3.3402e-04], + ..., + [-1.0274e-05, 0.0000e+00, 2.2516e-05, ..., 6.3956e-05, + -2.0653e-05, 8.9407e-06], + [ 2.9802e-07, 0.0000e+00, 1.3255e-05, ..., 2.2876e-04, + 2.6417e-04, 1.1885e-04], + [ 4.9919e-06, 0.0000e+00, 5.0962e-06, ..., 1.1402e-04, + 4.5776e-05, -2.2292e-04]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0243, 0.0157, 0.0254, 0.0077, 0.0217, -0.0144, -0.0093, -0.0210, + -0.0081, 0.0250], device='cuda:0'), grad: tensor([-0.0052, -0.0003, 0.0027, 0.0003, 0.0017, -0.0033, 0.0018, 0.0003, + 0.0017, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 221.27, cls_loss 0.0150 cls_loss_mapping 0.0267 cls_loss_causal 0.7046 re_mapping 0.0155 re_causal 0.0467 /// teacc 98.55 lr 0.00010000 +Epoch 52, weight, value: tensor([[ 0.0200, 0.0110, -0.0068, ..., 0.0372, -0.0689, -0.0613], + [ 0.0346, -0.0185, 0.0235, ..., -0.0551, 0.0352, -0.0811], + [ 0.0121, -0.0048, 0.0685, ..., -0.0779, -0.0408, -0.0854], + ..., + [ 0.0019, 0.0062, -0.0347, ..., 0.0139, 0.0626, 0.0510], + [ 0.0086, -0.0202, -0.0308, ..., -0.0829, 0.0435, -0.0735], + [-0.0109, -0.0258, -0.0246, ..., 0.0171, -0.0205, 0.0085]], + device='cuda:0'), grad: tensor([[ 1.1539e-06, 0.0000e+00, 6.4634e-07, ..., -4.8727e-06, + 8.6278e-06, 3.8520e-06], + [ 1.9129e-06, 0.0000e+00, 3.8929e-06, ..., 1.8999e-06, + -1.3399e-04, 8.6427e-07], + [ 5.1297e-06, 0.0000e+00, 2.9474e-05, ..., 1.5739e-06, + 3.7551e-05, 3.0309e-05], + ..., + [-3.8266e-05, 0.0000e+00, -2.3559e-05, ..., 8.5756e-06, + -4.3273e-05, -8.5473e-05], + [ 2.5444e-06, 0.0000e+00, -1.5378e-05, ..., 8.6725e-06, + 5.6416e-05, 1.4730e-05], + [ 1.7852e-05, 0.0000e+00, 6.9216e-06, ..., 8.8513e-05, + 6.3539e-05, 1.2171e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0239, 0.0148, 0.0256, 0.0080, 0.0216, -0.0147, -0.0088, -0.0213, + -0.0078, 0.0251], device='cuda:0'), grad: tensor([ 2.7284e-05, -1.3828e-04, 1.1641e-04, 2.2531e-04, -1.4439e-05, + -7.6532e-04, 2.3830e-04, -1.5008e-04, 2.1410e-04, 2.4700e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 221.27, cls_loss 0.0154 cls_loss_mapping 0.0269 cls_loss_causal 0.7268 re_mapping 0.0153 re_causal 0.0442 /// teacc 98.59 lr 0.00010000 +Epoch 53, weight, value: tensor([[ 0.0199, 0.0110, -0.0063, ..., 0.0376, -0.0697, -0.0627], + [ 0.0347, -0.0185, 0.0236, ..., -0.0560, 0.0359, -0.0815], + [ 0.0118, -0.0048, 0.0691, ..., -0.0788, -0.0416, -0.0871], + ..., + [ 0.0029, 0.0062, -0.0351, ..., 0.0154, 0.0634, 0.0525], + [ 0.0084, -0.0202, -0.0311, ..., -0.0838, 0.0438, -0.0754], + [-0.0112, -0.0258, -0.0250, ..., 0.0164, -0.0211, 0.0083]], + device='cuda:0'), grad: tensor([[ 1.7323e-07, 0.0000e+00, 1.3366e-05, ..., -1.6615e-05, + 3.6299e-05, 4.3400e-06], + [ 1.1586e-06, 0.0000e+00, 6.1989e-05, ..., 2.8864e-05, + 8.4698e-05, 3.1948e-05], + [ 2.7847e-07, 0.0000e+00, -2.6202e-04, ..., 5.1036e-06, + -2.3797e-05, 5.6103e-06], + ..., + [-6.3069e-06, 0.0000e+00, 7.3254e-05, ..., 3.6865e-05, + 5.6066e-06, 1.5450e-04], + [ 5.4017e-07, 0.0000e+00, 5.2959e-05, ..., 1.7837e-05, + -3.0994e-05, 2.0489e-05], + [ 1.5460e-06, 0.0000e+00, 3.5875e-06, ..., -8.5592e-05, + 4.2804e-06, -4.7255e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0240, 0.0157, 0.0252, 0.0069, 0.0216, -0.0138, -0.0098, -0.0199, + -0.0079, 0.0244], device='cuda:0'), grad: tensor([ 5.4628e-05, 2.2602e-04, -2.7657e-04, 5.9277e-05, 5.3316e-05, + 4.6134e-05, -6.0320e-05, 2.0254e-04, 6.4909e-05, -3.7003e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 221.07, cls_loss 0.0150 cls_loss_mapping 0.0208 cls_loss_causal 0.6986 re_mapping 0.0149 re_causal 0.0442 /// teacc 98.66 lr 0.00010000 +Epoch 54, weight, value: tensor([[ 0.0198, 0.0110, -0.0067, ..., 0.0372, -0.0691, -0.0645], + [ 0.0346, -0.0185, 0.0236, ..., -0.0573, 0.0361, -0.0806], + [ 0.0117, -0.0048, 0.0703, ..., -0.0797, -0.0417, -0.0875], + ..., + [ 0.0032, 0.0062, -0.0360, ..., 0.0149, 0.0634, 0.0520], + [ 0.0083, -0.0202, -0.0320, ..., -0.0841, 0.0439, -0.0761], + [-0.0114, -0.0258, -0.0245, ..., 0.0169, -0.0214, 0.0090]], + device='cuda:0'), grad: tensor([[ 5.3924e-07, 0.0000e+00, -8.5115e-05, ..., -1.9503e-04, + -6.2361e-06, 4.8056e-06], + [ 2.4512e-06, 0.0000e+00, 1.3493e-05, ..., 2.3708e-05, + -1.7524e-05, 3.3140e-05], + [ 2.1532e-06, 0.0000e+00, -7.6115e-05, ..., 6.9499e-05, + 1.2428e-05, 9.2387e-06], + ..., + [-2.6315e-05, 0.0000e+00, 7.8797e-05, ..., 1.1131e-05, + -3.5226e-05, -5.3465e-05], + [ 1.2917e-06, 0.0000e+00, 3.3379e-05, ..., 7.0810e-05, + 2.9325e-05, 8.8155e-05], + [ 1.1891e-05, 0.0000e+00, 5.9932e-05, ..., 6.9737e-05, + 3.6627e-05, 8.6725e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0246, 0.0159, 0.0258, 0.0068, 0.0213, -0.0136, -0.0093, -0.0205, + -0.0079, 0.0246], device='cuda:0'), grad: tensor([-2.7013e-04, 2.9415e-05, -3.5435e-05, -2.8992e-03, -9.7454e-05, + 8.7404e-04, 1.7643e-04, 8.8513e-05, 1.5621e-03, 5.7125e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 221.33, cls_loss 0.0169 cls_loss_mapping 0.0255 cls_loss_causal 0.7198 re_mapping 0.0148 re_causal 0.0411 /// teacc 98.65 lr 0.00010000 +Epoch 55, weight, value: tensor([[ 0.0197, 0.0110, -0.0076, ..., 0.0366, -0.0693, -0.0662], + [ 0.0345, -0.0185, 0.0234, ..., -0.0585, 0.0361, -0.0815], + [ 0.0113, -0.0048, 0.0710, ..., -0.0806, -0.0424, -0.0887], + ..., + [ 0.0035, 0.0062, -0.0371, ..., 0.0140, 0.0643, 0.0523], + [ 0.0083, -0.0202, -0.0324, ..., -0.0848, 0.0443, -0.0776], + [-0.0112, -0.0258, -0.0240, ..., 0.0176, -0.0217, 0.0091]], + device='cuda:0'), grad: tensor([[ 7.4133e-07, 0.0000e+00, -5.4687e-06, ..., -1.5378e-05, + 3.8128e-06, 3.8892e-06], + [ 6.7502e-06, 0.0000e+00, 2.9244e-06, ..., 4.5002e-06, + 8.1286e-06, 3.4302e-05], + [ 5.1446e-06, 0.0000e+00, -2.3723e-05, ..., 2.5518e-06, + 1.4432e-05, 1.1683e-05], + ..., + [-4.2677e-05, 0.0000e+00, 8.9034e-06, ..., 5.8636e-06, + -8.1420e-05, -1.2684e-04], + [ 1.5954e-06, 0.0000e+00, 5.7667e-06, ..., 5.5209e-06, + -6.3516e-07, 1.3240e-05], + [ 4.1053e-06, 0.0000e+00, 4.2096e-06, ..., -7.8678e-06, + 1.0222e-05, -2.9340e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0254, 0.0155, 0.0258, 0.0074, 0.0215, -0.0136, -0.0095, -0.0207, + -0.0077, 0.0248], device='cuda:0'), grad: tensor([-9.5069e-06, 4.1604e-05, 1.3590e-05, 2.1243e-04, 1.9185e-06, + -1.3244e-04, 8.5905e-06, -1.9693e-04, 4.7594e-05, 1.3247e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 221.36, cls_loss 0.0137 cls_loss_mapping 0.0219 cls_loss_causal 0.6747 re_mapping 0.0153 re_causal 0.0426 /// teacc 98.54 lr 0.00010000 +Epoch 56, weight, value: tensor([[ 0.0196, 0.0110, -0.0074, ..., 0.0372, -0.0693, -0.0669], + [ 0.0346, -0.0185, 0.0231, ..., -0.0594, 0.0360, -0.0820], + [ 0.0113, -0.0048, 0.0722, ..., -0.0814, -0.0429, -0.0903], + ..., + [ 0.0038, 0.0062, -0.0378, ..., 0.0136, 0.0648, 0.0519], + [ 0.0083, -0.0202, -0.0326, ..., -0.0859, 0.0450, -0.0782], + [-0.0115, -0.0258, -0.0247, ..., 0.0176, -0.0221, 0.0095]], + device='cuda:0'), grad: tensor([[ 5.2806e-07, 0.0000e+00, 4.5039e-06, ..., -2.3060e-06, + 1.7866e-05, 5.3272e-06], + [ 5.5134e-06, 0.0000e+00, 1.1265e-05, ..., 1.1390e-06, + -1.4555e-04, 1.7747e-05], + [ 1.4883e-06, 0.0000e+00, 3.3021e-04, ..., 1.3579e-06, + 1.2256e-05, 3.1531e-05], + ..., + [-1.8045e-05, 0.0000e+00, 1.8209e-05, ..., 1.4035e-06, + -1.6168e-05, -1.5765e-05], + [ 1.7816e-06, 0.0000e+00, 1.8880e-05, ..., 2.9877e-06, + 3.3170e-05, 2.5773e-04], + [ 4.1910e-06, 0.0000e+00, 2.0280e-05, ..., -6.4634e-06, + 1.9297e-05, -3.4571e-04]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0253, 0.0152, 0.0261, 0.0071, 0.0217, -0.0132, -0.0094, -0.0210, + -0.0077, 0.0249], device='cuda:0'), grad: tensor([ 5.1290e-05, -2.1517e-04, 4.9543e-04, -7.5960e-04, 9.5069e-05, + 6.9427e-04, -4.7779e-04, 2.0787e-05, 3.2258e-04, -2.2697e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 221.94, cls_loss 0.0129 cls_loss_mapping 0.0220 cls_loss_causal 0.6722 re_mapping 0.0153 re_causal 0.0428 /// teacc 98.54 lr 0.00010000 +Epoch 57, weight, value: tensor([[ 0.0196, 0.0110, -0.0075, ..., 0.0377, -0.0693, -0.0679], + [ 0.0345, -0.0185, 0.0230, ..., -0.0602, 0.0363, -0.0826], + [ 0.0111, -0.0048, 0.0724, ..., -0.0819, -0.0433, -0.0905], + ..., + [ 0.0044, 0.0062, -0.0387, ..., 0.0152, 0.0662, 0.0547], + [ 0.0083, -0.0202, -0.0330, ..., -0.0870, 0.0453, -0.0787], + [-0.0120, -0.0258, -0.0246, ..., 0.0178, -0.0228, 0.0094]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, 0.0000e+00, 2.3380e-05, ..., 7.8753e-06, + 1.7181e-05, 1.8656e-05], + [ 6.6776e-07, 0.0000e+00, 1.4257e-04, ..., 2.6733e-05, + 1.7798e-04, 4.4018e-05], + [ 2.5332e-07, 0.0000e+00, -1.1044e-03, ..., 1.5229e-05, + -5.0385e-07, 2.3559e-05], + ..., + [-5.4874e-06, 0.0000e+00, 5.9545e-05, ..., 1.4424e-04, + 2.2590e-05, 2.4211e-04], + [ 4.9267e-07, 0.0000e+00, 3.1042e-04, ..., 2.1771e-05, + 1.5574e-03, 7.8738e-05], + [ 2.6319e-06, 0.0000e+00, 8.9854e-06, ..., -1.7494e-05, + -4.5300e-06, -2.1696e-04]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0253, 0.0152, 0.0257, 0.0072, 0.0207, -0.0130, -0.0094, -0.0197, + -0.0079, 0.0246], device='cuda:0'), grad: tensor([ 8.5890e-05, 4.6706e-04, -1.4715e-03, 6.3896e-04, -1.4770e-04, + -2.9011e-03, 5.0402e-04, 3.5262e-04, 2.5520e-03, -7.9393e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 221.22, cls_loss 0.0124 cls_loss_mapping 0.0210 cls_loss_causal 0.7005 re_mapping 0.0141 re_causal 0.0433 /// teacc 98.69 lr 0.00010000 +Epoch 58, weight, value: tensor([[ 0.0195, 0.0110, -0.0073, ..., 0.0384, -0.0697, -0.0669], + [ 0.0347, -0.0185, 0.0229, ..., -0.0622, 0.0368, -0.0827], + [ 0.0110, -0.0048, 0.0733, ..., -0.0825, -0.0439, -0.0907], + ..., + [ 0.0049, 0.0062, -0.0394, ..., 0.0147, 0.0660, 0.0542], + [ 0.0083, -0.0202, -0.0335, ..., -0.0878, 0.0458, -0.0792], + [-0.0126, -0.0258, -0.0256, ..., 0.0181, -0.0230, 0.0097]], + device='cuda:0'), grad: tensor([[ 7.4320e-07, 0.0000e+00, -6.3992e-04, ..., -1.3094e-03, + 1.0747e-06, 7.4565e-05], + [ 6.3609e-07, 0.0000e+00, 2.2173e-05, ..., 1.2249e-05, + -1.1377e-05, 1.1742e-05], + [ 3.1907e-06, 0.0000e+00, -5.3763e-05, ..., 2.3142e-05, + 8.8811e-06, 4.2111e-05], + ..., + [-1.4804e-05, 0.0000e+00, 1.9565e-05, ..., 1.0356e-05, + -1.3150e-05, -1.2062e-05], + [ 1.9819e-06, 0.0000e+00, 9.7752e-05, ..., 2.5964e-04, + -1.1951e-05, 1.7738e-04], + [ 2.3060e-06, 0.0000e+00, 1.3739e-05, ..., 4.7374e-04, + 9.0227e-06, 9.5701e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0249, 0.0154, 0.0258, 0.0074, 0.0207, -0.0133, -0.0095, -0.0205, + -0.0078, 0.0248], device='cuda:0'), grad: tensor([-2.1896e-03, 2.7865e-05, 6.9857e-05, -4.5967e-03, 1.6773e-04, + 6.5088e-04, 1.7176e-03, 2.3857e-05, 8.6355e-04, 3.2597e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 220.96, cls_loss 0.0137 cls_loss_mapping 0.0196 cls_loss_causal 0.6868 re_mapping 0.0132 re_causal 0.0384 /// teacc 98.57 lr 0.00010000 +Epoch 59, weight, value: tensor([[ 0.0193, 0.0110, -0.0078, ..., 0.0381, -0.0700, -0.0676], + [ 0.0345, -0.0185, 0.0229, ..., -0.0632, 0.0368, -0.0834], + [ 0.0112, -0.0048, 0.0738, ..., -0.0835, -0.0441, -0.0905], + ..., + [ 0.0057, 0.0062, -0.0403, ..., 0.0140, 0.0666, 0.0547], + [ 0.0080, -0.0202, -0.0341, ..., -0.0891, 0.0460, -0.0803], + [-0.0132, -0.0258, -0.0243, ..., 0.0189, -0.0234, 0.0096]], + device='cuda:0'), grad: tensor([[ 7.3761e-07, 0.0000e+00, 1.4059e-05, ..., -2.1040e-05, + 1.2204e-05, 2.7940e-05], + [ 2.4829e-06, 0.0000e+00, 1.9848e-05, ..., 5.7295e-06, + -3.8326e-05, 9.0539e-05], + [-2.6729e-06, 0.0000e+00, -1.2887e-04, ..., 8.4490e-06, + 6.3956e-05, 9.7752e-05], + ..., + [-1.0043e-05, 0.0000e+00, 3.6687e-05, ..., 1.1131e-05, + 6.7241e-06, 3.7372e-05], + [ 2.4661e-06, 0.0000e+00, 6.3419e-05, ..., -1.8656e-04, + -1.1711e-03, -3.5248e-03], + [ 1.0915e-06, 0.0000e+00, 1.2815e-05, ..., 1.4484e-04, + 9.5654e-04, 2.8687e-03]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0251, 0.0152, 0.0259, 0.0074, 0.0213, -0.0131, -0.0101, -0.0202, + -0.0079, 0.0248], device='cuda:0'), grad: tensor([ 1.0014e-04, 7.7307e-05, 2.7061e-04, 2.6798e-04, 1.8609e-04, + 2.7061e-04, 4.8131e-06, 2.0015e-04, -8.7204e-03, 7.3395e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 221.73, cls_loss 0.0112 cls_loss_mapping 0.0173 cls_loss_causal 0.6771 re_mapping 0.0147 re_causal 0.0411 /// teacc 98.57 lr 0.00010000 +Epoch 60, weight, value: tensor([[ 0.0192, 0.0110, -0.0076, ..., 0.0385, -0.0704, -0.0684], + [ 0.0341, -0.0185, 0.0231, ..., -0.0638, 0.0369, -0.0843], + [ 0.0111, -0.0048, 0.0744, ..., -0.0842, -0.0447, -0.0916], + ..., + [ 0.0063, 0.0062, -0.0407, ..., 0.0136, 0.0674, 0.0553], + [ 0.0078, -0.0202, -0.0345, ..., -0.0905, 0.0462, -0.0812], + [-0.0131, -0.0258, -0.0246, ..., 0.0194, -0.0239, 0.0103]], + device='cuda:0'), grad: tensor([[ 1.3690e-07, 0.0000e+00, 1.4249e-06, ..., 1.0416e-05, + 1.1012e-05, 1.8748e-06], + [ 1.1800e-06, 0.0000e+00, 4.0606e-06, ..., 1.7583e-06, + -6.7949e-05, 6.5565e-06], + [ 1.5870e-06, 0.0000e+00, -7.4387e-05, ..., 2.7902e-06, + -9.7156e-06, 8.5682e-06], + ..., + [-9.3430e-06, 0.0000e+00, 9.7156e-06, ..., 3.4064e-05, + -2.8297e-05, 3.6974e-06], + [ 5.9884e-07, 0.0000e+00, 4.0710e-05, ..., 1.4886e-05, + 2.1055e-05, 1.7226e-05], + [ 1.3821e-06, 0.0000e+00, 1.2266e-06, ..., 1.0692e-06, + 8.5235e-06, -1.4529e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0252, 0.0152, 0.0258, 0.0074, 0.0213, -0.0135, -0.0101, -0.0199, + -0.0078, 0.0248], device='cuda:0'), grad: tensor([ 4.2439e-05, -1.2696e-04, -7.3075e-05, 2.8110e-04, -4.7460e-06, + -2.6655e-04, 5.6177e-06, 3.2000e-06, 1.1545e-04, 2.3276e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 221.40, cls_loss 0.0144 cls_loss_mapping 0.0254 cls_loss_causal 0.6531 re_mapping 0.0144 re_causal 0.0410 /// teacc 98.62 lr 0.00010000 +Epoch 61, weight, value: tensor([[ 0.0192, 0.0110, -0.0072, ..., 0.0394, -0.0708, -0.0690], + [ 0.0341, -0.0185, 0.0228, ..., -0.0641, 0.0362, -0.0847], + [ 0.0111, -0.0048, 0.0751, ..., -0.0865, -0.0448, -0.0923], + ..., + [ 0.0064, 0.0062, -0.0412, ..., 0.0135, 0.0680, 0.0558], + [ 0.0078, -0.0202, -0.0345, ..., -0.0913, 0.0476, -0.0816], + [-0.0132, -0.0258, -0.0251, ..., 0.0195, -0.0244, 0.0109]], + device='cuda:0'), grad: tensor([[ 1.0058e-07, 0.0000e+00, 1.0980e-06, ..., 1.5974e-05, + 1.6600e-05, 2.5645e-05], + [ 8.2701e-07, 0.0000e+00, 6.1281e-07, ..., 8.5011e-06, + -6.5088e-04, 2.1547e-05], + [ 6.7148e-07, 0.0000e+00, 4.3958e-06, ..., 6.4820e-06, + 1.5251e-05, 1.1757e-05], + ..., + [-3.3043e-06, 0.0000e+00, 1.8813e-06, ..., -2.9728e-06, + -6.2287e-05, -9.5010e-05], + [ 2.4121e-07, 0.0000e+00, 8.9556e-06, ..., 4.7207e-05, + 6.0272e-04, 5.2392e-05], + [ 7.5158e-07, 0.0000e+00, 4.1947e-06, ..., -1.0598e-04, + 4.0770e-05, -7.1287e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0246, 0.0145, 0.0260, 0.0073, 0.0209, -0.0134, -0.0103, -0.0196, + -0.0074, 0.0248], device='cuda:0'), grad: tensor([ 7.1824e-05, -9.3269e-04, 5.6058e-05, -7.9393e-05, 4.5598e-05, + 4.1634e-05, 2.9951e-05, -1.1152e-04, 1.0157e-03, -1.3673e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 60---------------------------------------------------- +epoch 60, time 221.80, cls_loss 0.0117 cls_loss_mapping 0.0188 cls_loss_causal 0.6517 re_mapping 0.0141 re_causal 0.0402 /// teacc 98.76 lr 0.00010000 +Epoch 62, weight, value: tensor([[ 0.0192, 0.0110, -0.0069, ..., 0.0400, -0.0714, -0.0696], + [ 0.0340, -0.0185, 0.0227, ..., -0.0639, 0.0365, -0.0836], + [ 0.0110, -0.0048, 0.0757, ..., -0.0873, -0.0454, -0.0932], + ..., + [ 0.0066, 0.0062, -0.0417, ..., 0.0130, 0.0687, 0.0558], + [ 0.0078, -0.0202, -0.0352, ..., -0.0905, 0.0484, -0.0832], + [-0.0133, -0.0258, -0.0258, ..., 0.0194, -0.0248, 0.0112]], + device='cuda:0'), grad: tensor([[ 1.2200e-07, 0.0000e+00, 9.0152e-06, ..., -2.0933e-04, + -7.1377e-06, 4.7535e-06], + [ 1.3346e-06, 0.0000e+00, 1.0788e-05, ..., 6.3293e-06, + 1.8021e-06, 6.6571e-06], + [ 8.6799e-07, 0.0000e+00, -3.9887e-04, ..., 1.1519e-05, + 4.4666e-06, -2.0817e-05], + ..., + [-7.9498e-06, 0.0000e+00, 3.1042e-04, ..., 1.0535e-05, + -1.7896e-05, 1.6065e-06], + [ 2.8312e-07, 0.0000e+00, 1.2919e-05, ..., 3.6746e-05, + -9.2268e-05, -6.9290e-06], + [ 3.2801e-06, 0.0000e+00, 1.5423e-05, ..., 5.8919e-05, + 2.6420e-05, -3.5316e-06]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0244, 0.0150, 0.0261, 0.0076, 0.0209, -0.0139, -0.0107, -0.0196, + -0.0076, 0.0246], device='cuda:0'), grad: tensor([-2.2268e-04, 2.8804e-05, -6.4182e-04, 6.9976e-05, 9.2685e-06, + 1.0207e-05, 1.2469e-04, 5.0926e-04, -5.4657e-05, 1.6630e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 220.71, cls_loss 0.0092 cls_loss_mapping 0.0137 cls_loss_causal 0.6800 re_mapping 0.0132 re_causal 0.0405 /// teacc 98.75 lr 0.00010000 +Epoch 63, weight, value: tensor([[ 0.0192, 0.0110, -0.0066, ..., 0.0402, -0.0723, -0.0709], + [ 0.0340, -0.0185, 0.0227, ..., -0.0653, 0.0366, -0.0842], + [ 0.0110, -0.0048, 0.0761, ..., -0.0883, -0.0464, -0.0939], + ..., + [ 0.0068, 0.0062, -0.0423, ..., 0.0127, 0.0691, 0.0557], + [ 0.0078, -0.0202, -0.0352, ..., -0.0903, 0.0489, -0.0833], + [-0.0134, -0.0258, -0.0262, ..., 0.0194, -0.0251, 0.0112]], + device='cuda:0'), grad: tensor([[ 1.7229e-07, 0.0000e+00, -7.3574e-07, ..., -1.4091e-06, + 8.9630e-06, 8.7172e-06], + [ 8.6799e-07, 0.0000e+00, 3.3341e-07, ..., 3.2115e-04, + 3.5286e-04, 6.1703e-04], + [ 1.2824e-06, 0.0000e+00, -4.0568e-06, ..., 5.2862e-06, + 2.1651e-05, 2.0325e-05], + ..., + [-1.6496e-05, 0.0000e+00, 2.8200e-06, ..., 2.2307e-05, + -6.9141e-05, -7.5877e-05], + [ 1.2163e-06, 0.0000e+00, 1.9856e-06, ..., 3.2037e-05, + 1.9938e-05, 7.3791e-05], + [ 1.2159e-05, 0.0000e+00, 1.3374e-06, ..., -1.3374e-05, + 7.5102e-05, -1.7667e-04]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0247, 0.0150, 0.0257, 0.0079, 0.0212, -0.0141, -0.0105, -0.0199, + -0.0070, 0.0245], device='cuda:0'), grad: tensor([ 1.7002e-05, 8.4496e-04, 3.9697e-05, -5.2780e-05, -1.0386e-03, + 2.0671e-04, 3.3081e-05, -9.5308e-05, 9.5725e-05, -4.9800e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 220.84, cls_loss 0.0103 cls_loss_mapping 0.0185 cls_loss_causal 0.6683 re_mapping 0.0130 re_causal 0.0395 /// teacc 98.75 lr 0.00010000 +Epoch 64, weight, value: tensor([[ 0.0191, 0.0110, -0.0067, ..., 0.0410, -0.0728, -0.0717], + [ 0.0338, -0.0185, 0.0225, ..., -0.0663, 0.0371, -0.0835], + [ 0.0109, -0.0048, 0.0767, ..., -0.0886, -0.0470, -0.0954], + ..., + [ 0.0073, 0.0062, -0.0430, ..., 0.0125, 0.0692, 0.0551], + [ 0.0076, -0.0202, -0.0354, ..., -0.0915, 0.0494, -0.0842], + [-0.0134, -0.0258, -0.0265, ..., 0.0194, -0.0256, 0.0116]], + device='cuda:0'), grad: tensor([[ 2.9728e-06, 0.0000e+00, 1.0263e-06, ..., 2.2054e-06, + 4.3571e-05, 3.8326e-05], + [ 4.6752e-06, 0.0000e+00, 2.3283e-06, ..., 1.1727e-05, + 8.1062e-04, 4.8399e-04], + [ 2.0397e-04, 0.0000e+00, 2.1899e-04, ..., 7.1079e-06, + 2.0373e-04, 5.9462e-04], + ..., + [ 1.2165e-04, 0.0000e+00, 1.4794e-04, ..., 2.2382e-05, + -1.2550e-03, -2.7275e-04], + [ 1.0036e-05, 0.0000e+00, 4.0010e-06, ..., 3.5405e-05, + -1.8096e-04, 7.4625e-05], + [ 3.9339e-06, 0.0000e+00, 5.7183e-06, ..., 3.5343e-03, + 1.2219e-04, 1.0765e-02]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0245, 0.0153, 0.0257, 0.0075, 0.0215, -0.0138, -0.0104, -0.0203, + -0.0073, 0.0246], device='cuda:0'), grad: tensor([ 0.0002, 0.0016, 0.0017, -0.0095, -0.0083, 0.0075, 0.0001, -0.0014, + -0.0004, 0.0086], device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 220.97, cls_loss 0.0096 cls_loss_mapping 0.0161 cls_loss_causal 0.6065 re_mapping 0.0128 re_causal 0.0361 /// teacc 98.66 lr 0.00010000 +Epoch 65, weight, value: tensor([[ 0.0191, 0.0110, -0.0067, ..., 0.0416, -0.0733, -0.0724], + [ 0.0338, -0.0185, 0.0223, ..., -0.0667, 0.0376, -0.0837], + [ 0.0107, -0.0048, 0.0772, ..., -0.0893, -0.0474, -0.0958], + ..., + [ 0.0077, 0.0062, -0.0430, ..., 0.0123, 0.0698, 0.0554], + [ 0.0075, -0.0202, -0.0357, ..., -0.0921, 0.0494, -0.0853], + [-0.0137, -0.0258, -0.0267, ..., 0.0195, -0.0261, 0.0118]], + device='cuda:0'), grad: tensor([[ 1.2647e-06, 0.0000e+00, -5.4359e-05, ..., -4.3333e-05, + 1.5736e-05, 3.6731e-06], + [ 5.6662e-06, 0.0000e+00, 5.7146e-06, ..., 5.2601e-06, + 1.6689e-05, 1.4052e-05], + [ 3.1330e-06, 0.0000e+00, -1.4484e-04, ..., 1.6481e-05, + 9.3818e-05, 5.9977e-06], + ..., + [-2.1845e-05, 0.0000e+00, 1.7476e-04, ..., 5.5619e-06, + -8.3372e-06, -1.4558e-05], + [ 1.5246e-06, 0.0000e+00, -4.5747e-05, ..., 2.0802e-05, + 4.9502e-05, 9.1493e-06], + [ 5.2936e-06, 0.0000e+00, 1.1571e-05, ..., 1.2569e-05, + 1.3478e-05, 4.1574e-06]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0244, 0.0157, 0.0254, 0.0075, 0.0213, -0.0139, -0.0107, -0.0197, + -0.0074, 0.0243], device='cuda:0'), grad: tensor([-3.9339e-05, 5.0873e-05, -1.8525e-04, 3.7313e-04, -7.6741e-06, + -1.0653e-03, -2.2507e-04, 3.5667e-04, 6.4993e-04, 9.2149e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 64---------------------------------------------------- +epoch 64, time 221.51, cls_loss 0.0106 cls_loss_mapping 0.0175 cls_loss_causal 0.6603 re_mapping 0.0130 re_causal 0.0379 /// teacc 98.87 lr 0.00010000 +Epoch 66, weight, value: tensor([[ 0.0189, 0.0110, -0.0068, ..., 0.0422, -0.0741, -0.0705], + [ 0.0334, -0.0185, 0.0221, ..., -0.0672, 0.0381, -0.0836], + [ 0.0107, -0.0048, 0.0781, ..., -0.0907, -0.0483, -0.0971], + ..., + [ 0.0082, 0.0062, -0.0438, ..., 0.0118, 0.0701, 0.0553], + [ 0.0074, -0.0202, -0.0361, ..., -0.0926, 0.0495, -0.0860], + [-0.0136, -0.0258, -0.0270, ..., 0.0193, -0.0264, 0.0119]], + device='cuda:0'), grad: tensor([[ 4.5076e-07, 0.0000e+00, -2.1569e-06, ..., -3.2708e-06, + 6.3330e-06, 3.2425e-05], + [ 2.9393e-06, 0.0000e+00, -3.3975e-05, ..., 1.5749e-06, + -1.5676e-04, 8.2031e-06], + [ 1.9185e-06, 0.0000e+00, -3.2693e-05, ..., 1.2070e-05, + 9.1195e-05, 4.8041e-05], + ..., + [-1.3337e-05, 0.0000e+00, 4.3690e-05, ..., 4.2915e-06, + -1.4842e-05, -1.4260e-05], + [ 5.1223e-06, 0.0000e+00, 1.8358e-05, ..., 3.8147e-05, + 1.9163e-05, 1.8156e-04], + [ 9.6634e-06, 0.0000e+00, 1.0923e-05, ..., -6.6698e-05, + 2.3365e-05, -3.5906e-04]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0244, 0.0159, 0.0253, 0.0076, 0.0211, -0.0139, -0.0097, -0.0198, + -0.0076, 0.0240], device='cuda:0'), grad: tensor([ 4.8399e-05, -2.5320e-04, 1.6594e-04, 9.5367e-04, 9.6738e-05, + -1.0910e-03, 7.3910e-05, 4.9531e-05, 2.7061e-04, -3.1471e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 221.44, cls_loss 0.0111 cls_loss_mapping 0.0198 cls_loss_causal 0.6015 re_mapping 0.0134 re_causal 0.0378 /// teacc 98.82 lr 0.00010000 +Epoch 67, weight, value: tensor([[ 0.0189, 0.0110, -0.0066, ..., 0.0430, -0.0742, -0.0708], + [ 0.0333, -0.0185, 0.0217, ..., -0.0675, 0.0378, -0.0839], + [ 0.0108, -0.0048, 0.0795, ..., -0.0913, -0.0490, -0.0977], + ..., + [ 0.0081, 0.0062, -0.0443, ..., 0.0116, 0.0714, 0.0553], + [ 0.0070, -0.0202, -0.0359, ..., -0.0935, 0.0506, -0.0870], + [-0.0134, -0.0258, -0.0273, ..., 0.0190, -0.0272, 0.0118]], + device='cuda:0'), grad: tensor([[ 1.2666e-07, 0.0000e+00, -2.2754e-05, ..., -1.4341e-04, + 7.3090e-06, 1.1194e-06], + [ 4.6194e-07, 0.0000e+00, 2.6003e-06, ..., 1.5730e-06, + -1.4096e-05, 8.5495e-07], + [ 7.2923e-07, 0.0000e+00, -1.6078e-05, ..., 6.0052e-06, + 2.3507e-06, 1.5441e-06], + ..., + [-3.9972e-06, 0.0000e+00, 1.7360e-06, ..., 4.2804e-06, + -4.7944e-06, 4.9993e-06], + [ 3.4180e-07, 0.0000e+00, 8.8215e-06, ..., 1.0327e-05, + -8.4471e-07, 6.7540e-06], + [ 9.3039e-07, 0.0000e+00, 6.0201e-06, ..., -9.9018e-06, + 3.3639e-06, -7.8797e-05]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0240, 0.0155, 0.0257, 0.0076, 0.0216, -0.0143, -0.0107, -0.0196, + -0.0070, 0.0235], device='cuda:0'), grad: tensor([-1.2052e-04, -1.8820e-05, -1.5618e-06, 6.8188e-05, 5.2691e-05, + -1.8016e-05, 5.9992e-05, 2.7064e-06, 1.4782e-05, -3.9488e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 221.17, cls_loss 0.0098 cls_loss_mapping 0.0162 cls_loss_causal 0.6314 re_mapping 0.0124 re_causal 0.0372 /// teacc 98.71 lr 0.00010000 +Epoch 68, weight, value: tensor([[ 0.0188, 0.0110, -0.0064, ..., 0.0436, -0.0745, -0.0706], + [ 0.0332, -0.0185, 0.0213, ..., -0.0683, 0.0374, -0.0849], + [ 0.0107, -0.0048, 0.0798, ..., -0.0924, -0.0497, -0.0989], + ..., + [ 0.0083, 0.0062, -0.0449, ..., 0.0123, 0.0721, 0.0560], + [ 0.0069, -0.0202, -0.0360, ..., -0.0940, 0.0508, -0.0866], + [-0.0132, -0.0258, -0.0275, ..., 0.0186, -0.0276, 0.0115]], + device='cuda:0'), grad: tensor([[ 3.5111e-06, 0.0000e+00, -5.2631e-05, ..., -1.8522e-05, + 5.5492e-05, 4.8906e-05], + [ 1.7229e-07, 0.0000e+00, 1.5962e-04, ..., 6.1512e-05, + 1.7077e-05, 3.5673e-05], + [ 1.5469e-06, 0.0000e+00, -4.0555e-04, ..., 3.4392e-05, + -1.4126e-04, 1.8403e-05], + ..., + [-2.3305e-05, 0.0000e+00, 2.2247e-05, ..., 1.9699e-05, + -2.0966e-05, -5.0962e-05], + [ 1.7881e-06, 0.0000e+00, 6.9261e-05, ..., 4.8071e-05, + 7.3552e-05, 4.1008e-05], + [ 1.0751e-05, 0.0000e+00, 3.1859e-05, ..., -5.5492e-05, + 4.2409e-05, -2.7537e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0237, 0.0148, 0.0253, 0.0077, 0.0215, -0.0137, -0.0100, -0.0192, + -0.0072, 0.0231], device='cuda:0'), grad: tensor([ 4.1097e-05, 3.9411e-04, -7.6580e-04, 4.6283e-05, 1.5974e-03, + -1.1379e-04, -1.6537e-03, -2.7224e-05, 4.8161e-04, 2.2203e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 67---------------------------------------------------- +epoch 67, time 221.84, cls_loss 0.0101 cls_loss_mapping 0.0138 cls_loss_causal 0.6174 re_mapping 0.0129 re_causal 0.0369 /// teacc 98.88 lr 0.00010000 +Epoch 69, weight, value: tensor([[ 0.0187, 0.0110, -0.0067, ..., 0.0441, -0.0746, -0.0712], + [ 0.0332, -0.0185, 0.0214, ..., -0.0688, 0.0385, -0.0853], + [ 0.0104, -0.0048, 0.0804, ..., -0.0928, -0.0509, -0.0990], + ..., + [ 0.0085, 0.0062, -0.0460, ..., 0.0112, 0.0721, 0.0552], + [ 0.0074, -0.0202, -0.0360, ..., -0.0951, 0.0509, -0.0881], + [-0.0135, -0.0258, -0.0278, ..., 0.0193, -0.0279, 0.0121]], + device='cuda:0'), grad: tensor([[ 2.3581e-06, 0.0000e+00, 2.0042e-05, ..., 1.9693e-04, + 5.9903e-06, 1.2541e-04], + [ 2.4378e-05, 0.0000e+00, 1.6943e-05, ..., 1.5065e-05, + 4.0263e-05, 1.1033e-04], + [ 2.4930e-05, 0.0000e+00, -8.1778e-05, ..., 1.2808e-05, + 5.7369e-05, 7.9334e-05], + ..., + [-1.3876e-04, 0.0000e+00, 9.8944e-06, ..., -1.6928e-04, + -3.0017e-04, -7.7868e-04], + [ 1.3210e-05, 0.0000e+00, 2.6003e-05, ..., 7.6115e-05, + 1.7002e-05, 1.7142e-04], + [ 2.8506e-05, 0.0000e+00, 3.0380e-06, ..., 1.5646e-05, + 6.9559e-05, 1.1301e-04]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0236, 0.0157, 0.0250, 0.0074, 0.0218, -0.0138, -0.0097, -0.0202, + -0.0075, 0.0236], device='cuda:0'), grad: tensor([ 8.0061e-04, 2.7180e-04, 1.9476e-05, 2.9874e-04, 3.6955e-04, + 2.1708e-04, -8.7452e-04, -1.7967e-03, 2.9349e-04, 4.0030e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 220.92, cls_loss 0.0092 cls_loss_mapping 0.0142 cls_loss_causal 0.6432 re_mapping 0.0120 re_causal 0.0358 /// teacc 98.71 lr 0.00010000 +Epoch 70, weight, value: tensor([[ 0.0186, 0.0110, -0.0067, ..., 0.0445, -0.0750, -0.0717], + [ 0.0331, -0.0185, 0.0209, ..., -0.0694, 0.0384, -0.0855], + [ 0.0117, -0.0048, 0.0822, ..., -0.0937, -0.0508, -0.0984], + ..., + [ 0.0079, 0.0062, -0.0473, ..., 0.0106, 0.0728, 0.0551], + [ 0.0074, -0.0202, -0.0364, ..., -0.0962, 0.0509, -0.0898], + [-0.0138, -0.0258, -0.0283, ..., 0.0202, -0.0284, 0.0129]], + device='cuda:0'), grad: tensor([[ 5.3179e-07, 0.0000e+00, -6.5193e-06, ..., -2.4647e-05, + 8.4192e-06, 6.2436e-06], + [ 2.5257e-06, 0.0000e+00, 7.1600e-06, ..., 4.8429e-06, + 2.2557e-06, 1.6391e-05], + [ 7.6834e-07, 0.0000e+00, -1.0252e-05, ..., 1.1539e-04, + 5.9530e-06, 7.9423e-06], + ..., + [-1.9878e-05, 0.0000e+00, 2.3544e-06, ..., 4.6864e-06, + -6.9737e-05, -7.0393e-05], + [ 5.4203e-07, 0.0000e+00, 1.0645e-04, ..., 9.5516e-06, + 6.1169e-06, -4.1276e-06], + [ 1.0736e-05, 0.0000e+00, 2.7463e-05, ..., 2.5090e-06, + 3.9518e-05, -5.1230e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0239, 0.0155, 0.0264, 0.0072, 0.0217, -0.0145, -0.0097, -0.0205, + -0.0080, 0.0244], device='cuda:0'), grad: tensor([ 1.9386e-05, 2.5541e-05, 1.5342e-04, -3.8028e-05, 2.9874e-04, + 1.1706e-04, -6.9427e-04, -1.0169e-04, 1.9610e-04, 2.2873e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 221.03, cls_loss 0.0085 cls_loss_mapping 0.0144 cls_loss_causal 0.6593 re_mapping 0.0120 re_causal 0.0373 /// teacc 98.79 lr 0.00010000 +Epoch 71, weight, value: tensor([[ 0.0186, 0.0110, -0.0060, ..., 0.0450, -0.0751, -0.0720], + [ 0.0332, -0.0185, 0.0206, ..., -0.0712, 0.0395, -0.0851], + [ 0.0116, -0.0048, 0.0826, ..., -0.0946, -0.0511, -0.0989], + ..., + [ 0.0080, 0.0062, -0.0477, ..., 0.0104, 0.0728, 0.0549], + [ 0.0073, -0.0202, -0.0370, ..., -0.0969, 0.0507, -0.0903], + [-0.0137, -0.0258, -0.0283, ..., 0.0204, -0.0292, 0.0127]], + device='cuda:0'), grad: tensor([[ 4.5635e-08, 0.0000e+00, 4.0531e-06, ..., -4.7326e-05, + 3.3081e-06, 6.1393e-06], + [ 2.0321e-06, 0.0000e+00, -4.0174e-05, ..., 1.2374e-04, + -3.1710e-05, 2.2590e-04], + [ 2.8266e-07, 0.0000e+00, -7.9036e-05, ..., 6.7428e-06, + 2.9430e-06, 2.3171e-05], + ..., + [-4.7423e-06, 0.0000e+00, 1.2025e-05, ..., 7.2196e-06, + 1.9185e-06, 2.2531e-05], + [ 1.6624e-07, 0.0000e+00, 2.8431e-05, ..., 1.9610e-05, + 1.8194e-05, 2.6894e-04], + [ 7.9535e-07, 0.0000e+00, 1.9558e-06, ..., 1.2970e-04, + 1.6332e-05, 6.7912e-06]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0235, 0.0162, 0.0262, 0.0072, 0.0216, -0.0140, -0.0097, -0.0208, + -0.0087, 0.0242], device='cuda:0'), grad: tensor([-4.3064e-05, -1.1432e-04, -4.2468e-05, 5.0366e-05, -2.6274e-04, + -9.6560e-06, 8.5354e-05, 6.2168e-05, 3.2973e-04, -5.6028e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 221.18, cls_loss 0.0079 cls_loss_mapping 0.0139 cls_loss_causal 0.6563 re_mapping 0.0118 re_causal 0.0356 /// teacc 98.88 lr 0.00010000 +Epoch 72, weight, value: tensor([[ 0.0185, 0.0110, -0.0058, ..., 0.0456, -0.0755, -0.0728], + [ 0.0329, -0.0185, 0.0207, ..., -0.0719, 0.0393, -0.0857], + [ 0.0114, -0.0048, 0.0831, ..., -0.0954, -0.0517, -0.0993], + ..., + [ 0.0085, 0.0062, -0.0483, ..., 0.0101, 0.0738, 0.0552], + [ 0.0072, -0.0202, -0.0377, ..., -0.0974, 0.0508, -0.0916], + [-0.0140, -0.0258, -0.0290, ..., 0.0203, -0.0294, 0.0130]], + device='cuda:0'), grad: tensor([[ 4.0885e-07, 0.0000e+00, 8.2254e-06, ..., -6.4075e-06, + 3.0845e-05, 8.0824e-05], + [ 1.2480e-06, 0.0000e+00, 2.4363e-06, ..., -1.2210e-06, + 3.6478e-05, 4.7493e-04], + [ 3.0566e-06, 0.0000e+00, 3.1948e-05, ..., 2.0474e-05, + 5.6952e-05, 2.2840e-04], + ..., + [-8.1360e-06, 0.0000e+00, 1.5302e-06, ..., 2.9467e-06, + 6.2752e-04, 8.0795e-03], + [ 6.0862e-07, 0.0000e+00, 1.4998e-05, ..., -1.6853e-05, + -1.2302e-04, 3.9577e-04], + [ 1.3858e-06, 0.0000e+00, 3.0473e-06, ..., -1.8403e-05, + 1.7118e-04, 1.3189e-03]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0236, 0.0160, 0.0261, 0.0076, 0.0214, -0.0140, -0.0093, -0.0204, + -0.0092, 0.0241], device='cuda:0'), grad: tensor([ 1.6570e-04, 5.1165e-04, 3.5763e-04, 2.0897e-04, -1.2245e-02, + 7.7105e-04, -1.7405e-04, 8.7051e-03, 4.0501e-05, 1.6584e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 220.89, cls_loss 0.0096 cls_loss_mapping 0.0147 cls_loss_causal 0.6650 re_mapping 0.0122 re_causal 0.0369 /// teacc 98.68 lr 0.00010000 +Epoch 73, weight, value: tensor([[ 0.0185, 0.0110, -0.0058, ..., 0.0465, -0.0754, -0.0734], + [ 0.0330, -0.0185, 0.0194, ..., -0.0719, 0.0394, -0.0860], + [ 0.0111, -0.0048, 0.0842, ..., -0.0963, -0.0510, -0.1011], + ..., + [ 0.0088, 0.0062, -0.0484, ..., 0.0104, 0.0740, 0.0557], + [ 0.0071, -0.0202, -0.0382, ..., -0.0982, 0.0510, -0.0925], + [-0.0140, -0.0258, -0.0291, ..., 0.0199, -0.0299, 0.0123]], + device='cuda:0'), grad: tensor([[ 1.4948e-07, 0.0000e+00, 8.3540e-07, ..., 1.1008e-06, + 9.1940e-06, 2.4904e-06], + [-3.4049e-06, 0.0000e+00, -7.5437e-08, ..., -1.3728e-06, + -3.0937e-03, -3.4237e-04], + [-2.3330e-07, 0.0000e+00, 3.3110e-05, ..., 1.1586e-06, + 5.8770e-05, 7.2531e-06], + ..., + [-5.1931e-06, 0.0000e+00, 3.8631e-06, ..., 2.6505e-06, + 2.9984e-03, 3.2759e-04], + [ 1.0133e-06, 0.0000e+00, -4.5717e-05, ..., 9.6634e-06, + -5.0873e-05, 2.3276e-05], + [ 2.2892e-06, 0.0000e+00, 3.5902e-07, ..., -1.2666e-05, + 2.3052e-05, -1.9804e-05]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0232, 0.0153, 0.0266, 0.0076, 0.0222, -0.0132, -0.0112, -0.0198, + -0.0089, 0.0235], device='cuda:0'), grad: tensor([ 2.7597e-05, -7.7629e-03, 2.0158e-04, -2.4891e-04, 5.3734e-05, + 2.9540e-04, -7.8261e-05, 7.5493e-03, -1.1379e-04, 7.7069e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 220.49, cls_loss 0.0091 cls_loss_mapping 0.0157 cls_loss_causal 0.6207 re_mapping 0.0123 re_causal 0.0352 /// teacc 98.71 lr 0.00010000 +Epoch 74, weight, value: tensor([[ 0.0184, 0.0110, -0.0058, ..., 0.0466, -0.0758, -0.0741], + [ 0.0330, -0.0185, 0.0192, ..., -0.0723, 0.0388, -0.0862], + [ 0.0110, -0.0048, 0.0849, ..., -0.0974, -0.0517, -0.1020], + ..., + [ 0.0089, 0.0062, -0.0489, ..., 0.0110, 0.0753, 0.0557], + [ 0.0070, -0.0202, -0.0388, ..., -0.0983, 0.0511, -0.0931], + [-0.0140, -0.0258, -0.0294, ..., 0.0200, -0.0302, 0.0125]], + device='cuda:0'), grad: tensor([[ 2.8089e-06, 0.0000e+00, 4.7311e-06, ..., -1.2375e-05, + 1.1906e-05, 2.3663e-05], + [ 3.6687e-05, 0.0000e+00, 1.0675e-04, ..., 3.4273e-05, + 8.8632e-05, 1.4722e-04], + [ 1.6344e-04, 0.0000e+00, 4.5156e-04, ..., 2.5928e-05, + 3.5453e-04, 4.0197e-04], + ..., + [-2.5702e-04, 0.0000e+00, -6.7949e-04, ..., 3.6389e-05, + -7.8773e-04, -1.8644e-03], + [ 4.4554e-06, 0.0000e+00, -1.6823e-05, ..., 1.9208e-05, + -2.8968e-05, 2.0730e-04], + [ 4.9882e-06, 0.0000e+00, 1.6078e-05, ..., -7.4673e-04, + 2.2602e-04, -1.1473e-03]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0236, 0.0150, 0.0265, 0.0075, 0.0218, -0.0131, -0.0104, -0.0195, + -0.0090, 0.0235], device='cuda:0'), grad: tensor([ 1.4111e-05, 3.1948e-04, 1.1663e-03, 3.9458e-04, 1.0185e-03, + 1.1408e-04, 3.9649e-04, -3.1147e-03, 6.0201e-05, -3.6883e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 220.60, cls_loss 0.0081 cls_loss_mapping 0.0142 cls_loss_causal 0.6488 re_mapping 0.0116 re_causal 0.0358 /// teacc 98.75 lr 0.00010000 +Epoch 75, weight, value: tensor([[ 0.0183, 0.0110, -0.0063, ..., 0.0467, -0.0763, -0.0749], + [ 0.0348, -0.0185, 0.0195, ..., -0.0734, 0.0394, -0.0875], + [ 0.0101, -0.0048, 0.0856, ..., -0.0980, -0.0530, -0.1021], + ..., + [ 0.0086, 0.0062, -0.0496, ..., 0.0104, 0.0754, 0.0560], + [ 0.0069, -0.0202, -0.0394, ..., -0.0990, 0.0512, -0.0941], + [-0.0140, -0.0258, -0.0296, ..., 0.0203, -0.0305, 0.0122]], + device='cuda:0'), grad: tensor([[ 3.2550e-07, 0.0000e+00, -4.9770e-06, ..., -1.1101e-05, + 1.3318e-06, 6.2361e-06], + [ 2.0526e-06, 0.0000e+00, 1.5926e-06, ..., 3.3341e-06, + -5.3719e-06, 7.4767e-06], + [ 2.0787e-06, 0.0000e+00, -1.6615e-05, ..., 6.2622e-06, + 2.7046e-06, 8.1137e-06], + ..., + [-8.7023e-06, 0.0000e+00, 6.8694e-06, ..., 7.0035e-06, + -5.3272e-06, 1.2014e-06], + [ 4.6473e-07, 0.0000e+00, 4.8056e-06, ..., 3.8259e-06, + 1.0338e-06, 4.0978e-06], + [ 1.6531e-06, 0.0000e+00, 2.1420e-06, ..., 4.5925e-05, + 1.9222e-06, 6.1393e-05]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0242, 0.0156, 0.0261, 0.0076, 0.0226, -0.0133, -0.0099, -0.0198, + -0.0092, 0.0232], device='cuda:0'), grad: tensor([-7.8678e-06, 6.0648e-06, -9.3579e-06, -3.1412e-05, -1.1235e-04, + -1.7256e-05, 4.6194e-05, 1.3195e-05, 2.1324e-05, 9.1314e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 220.49, cls_loss 0.0062 cls_loss_mapping 0.0095 cls_loss_causal 0.6319 re_mapping 0.0119 re_causal 0.0354 /// teacc 98.83 lr 0.00010000 +Epoch 76, weight, value: tensor([[ 0.0182, 0.0110, -0.0061, ..., 0.0472, -0.0767, -0.0756], + [ 0.0346, -0.0185, 0.0193, ..., -0.0736, 0.0398, -0.0878], + [ 0.0100, -0.0048, 0.0860, ..., -0.0984, -0.0536, -0.1025], + ..., + [ 0.0093, 0.0062, -0.0500, ..., 0.0102, 0.0756, 0.0559], + [ 0.0068, -0.0202, -0.0397, ..., -0.0997, 0.0512, -0.0950], + [-0.0146, -0.0258, -0.0299, ..., 0.0204, -0.0302, 0.0126]], + device='cuda:0'), grad: tensor([[ 3.7253e-08, 0.0000e+00, -1.6205e-06, ..., -2.2501e-05, + 1.0036e-05, 1.8552e-06], + [ 7.1945e-07, 0.0000e+00, 4.0904e-06, ..., 5.9279e-07, + 1.2517e-04, 5.7332e-06], + [ 3.9442e-07, 0.0000e+00, -3.3617e-05, ..., 2.4289e-06, + 9.0837e-05, 4.3437e-06], + ..., + [-2.4717e-06, 0.0000e+00, 1.0237e-05, ..., 4.1425e-06, + -3.8123e-04, -9.4622e-06], + [ 8.8010e-08, 0.0000e+00, 1.1444e-05, ..., 2.4009e-06, + 4.3154e-05, 2.2575e-06], + [ 7.2550e-07, 0.0000e+00, 4.6119e-06, ..., -2.2650e-06, + 7.1049e-05, -2.7865e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0241, 0.0158, 0.0259, 0.0075, 0.0223, -0.0132, -0.0099, -0.0201, + -0.0094, 0.0236], device='cuda:0'), grad: tensor([-1.2800e-05, 2.4009e-04, 1.5044e-04, 3.2306e-05, 3.1203e-05, + 1.9334e-06, 7.4729e-06, -6.8665e-04, 1.0699e-04, 1.2898e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 220.55, cls_loss 0.0072 cls_loss_mapping 0.0134 cls_loss_causal 0.6175 re_mapping 0.0112 re_causal 0.0341 /// teacc 98.72 lr 0.00010000 +Epoch 77, weight, value: tensor([[ 0.0181, 0.0110, -0.0073, ..., 0.0462, -0.0771, -0.0774], + [ 0.0345, -0.0185, 0.0192, ..., -0.0742, 0.0399, -0.0880], + [ 0.0100, -0.0048, 0.0862, ..., -0.0990, -0.0542, -0.1034], + ..., + [ 0.0100, 0.0062, -0.0503, ..., 0.0100, 0.0760, 0.0562], + [ 0.0067, -0.0202, -0.0397, ..., -0.1002, 0.0517, -0.0952], + [-0.0148, -0.0258, -0.0285, ..., 0.0221, -0.0307, 0.0130]], + device='cuda:0'), grad: tensor([[ 1.6950e-07, 0.0000e+00, 2.5406e-06, ..., 9.5963e-06, + 1.2800e-05, 1.1452e-05], + [ 9.1689e-07, 0.0000e+00, 4.2081e-05, ..., 1.3642e-05, + 7.1347e-05, 1.8016e-05], + [ 1.6401e-06, 0.0000e+00, -3.8648e-04, ..., 5.8562e-06, + -6.4671e-05, 8.1360e-06], + ..., + [-5.9940e-06, 0.0000e+00, 2.8610e-04, ..., 9.1642e-06, + 5.5939e-05, 1.1045e-06], + [ 3.2177e-07, 0.0000e+00, 1.2584e-05, ..., 3.7342e-05, + -1.7023e-04, 5.5134e-05], + [ 1.9893e-06, 0.0000e+00, 4.0457e-06, ..., 1.3363e-04, + 2.2307e-05, 1.5581e-04]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0246, 0.0158, 0.0256, 0.0073, 0.0221, -0.0129, -0.0105, -0.0198, + -0.0091, 0.0241], device='cuda:0'), grad: tensor([ 3.5107e-05, 2.0051e-04, -5.9557e-04, 9.8884e-05, 2.0325e-05, + 3.2246e-05, -2.0301e-04, 4.7112e-04, -2.4748e-04, 1.8740e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 220.57, cls_loss 0.0081 cls_loss_mapping 0.0123 cls_loss_causal 0.6276 re_mapping 0.0112 re_causal 0.0315 /// teacc 98.67 lr 0.00010000 +Epoch 78, weight, value: tensor([[ 0.0179, 0.0110, -0.0069, ..., 0.0466, -0.0779, -0.0779], + [ 0.0349, -0.0185, 0.0190, ..., -0.0753, 0.0403, -0.0881], + [ 0.0098, -0.0048, 0.0871, ..., -0.1008, -0.0546, -0.1040], + ..., + [ 0.0090, 0.0062, -0.0510, ..., 0.0097, 0.0765, 0.0551], + [ 0.0063, -0.0202, -0.0401, ..., -0.1018, 0.0519, -0.0973], + [-0.0141, -0.0258, -0.0287, ..., 0.0223, -0.0310, 0.0139]], + device='cuda:0'), grad: tensor([[ 1.5590e-06, 0.0000e+00, -2.9057e-06, ..., -6.6385e-06, + 4.4331e-06, 1.3977e-05], + [ 4.5635e-07, 0.0000e+00, -5.6103e-06, ..., 1.9241e-06, + -4.3690e-05, 9.5367e-06], + [ 1.6838e-06, 0.0000e+00, -2.0519e-05, ..., 2.6263e-06, + 1.4648e-05, 6.9365e-06], + ..., + [ 4.8876e-05, 0.0000e+00, 1.3202e-05, ..., 5.4538e-05, + 4.0494e-06, 4.7731e-04], + [ 1.3612e-05, 0.0000e+00, 5.0515e-06, ..., 1.3612e-05, + 1.0267e-05, 1.0139e-04], + [-8.7798e-05, 0.0000e+00, 1.5916e-06, ..., -9.7215e-05, + 2.7753e-06, -8.9836e-04]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0243, 0.0160, 0.0256, 0.0075, 0.0222, -0.0129, -0.0105, -0.0206, + -0.0094, 0.0245], device='cuda:0'), grad: tensor([ 4.9174e-05, -3.5793e-05, 2.8566e-05, 4.9543e-04, 2.7418e-04, + -6.9284e-04, -2.6155e-04, 6.7377e-04, 4.1890e-04, -9.5081e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 220.74, cls_loss 0.0083 cls_loss_mapping 0.0129 cls_loss_causal 0.5903 re_mapping 0.0118 re_causal 0.0335 /// teacc 98.58 lr 0.00010000 +Epoch 79, weight, value: tensor([[ 0.0177, 0.0110, -0.0067, ..., 0.0470, -0.0783, -0.0785], + [ 0.0352, -0.0185, 0.0186, ..., -0.0761, 0.0404, -0.0875], + [ 0.0091, -0.0048, 0.0876, ..., -0.1015, -0.0552, -0.1044], + ..., + [ 0.0091, 0.0062, -0.0508, ..., 0.0118, 0.0777, 0.0573], + [ 0.0060, -0.0202, -0.0406, ..., -0.1022, 0.0519, -0.0985], + [-0.0151, -0.0258, -0.0289, ..., 0.0215, -0.0317, 0.0128]], + device='cuda:0'), grad: tensor([[ 1.5926e-07, 0.0000e+00, 7.0967e-07, ..., 2.6785e-06, + 3.1181e-06, 1.0312e-05], + [ 5.9186e-07, 0.0000e+00, 1.2144e-06, ..., 3.1795e-06, + 3.4690e-05, 4.4495e-05], + [ 3.3528e-07, 0.0000e+00, -1.3612e-05, ..., 2.4009e-06, + 1.2629e-05, 9.8348e-06], + ..., + [-3.3565e-06, 0.0000e+00, 4.0196e-06, ..., 9.2313e-06, + -1.1367e-04, -1.0765e-04], + [ 1.9977e-07, 0.0000e+00, 1.3392e-06, ..., 2.9996e-05, + 1.5348e-05, 6.6042e-05], + [ 1.5646e-06, 0.0000e+00, 1.2498e-06, ..., 2.6093e-03, + 8.2478e-06, 4.3373e-03]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0242, 0.0161, 0.0257, 0.0079, 0.0213, -0.0130, -0.0099, -0.0192, + -0.0098, 0.0234], device='cuda:0'), grad: tensor([ 1.6734e-05, 6.4731e-05, 1.5702e-06, 8.3804e-05, -4.0207e-03, + -6.7472e-04, 6.7759e-04, -1.5879e-04, 8.3029e-05, 3.9215e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 220.78, cls_loss 0.0095 cls_loss_mapping 0.0153 cls_loss_causal 0.6416 re_mapping 0.0112 re_causal 0.0338 /// teacc 98.81 lr 0.00010000 +Epoch 80, weight, value: tensor([[ 0.0176, 0.0110, -0.0068, ..., 0.0472, -0.0789, -0.0783], + [ 0.0354, -0.0185, 0.0170, ..., -0.0767, 0.0404, -0.0879], + [ 0.0090, -0.0048, 0.0886, ..., -0.1022, -0.0551, -0.1051], + ..., + [ 0.0094, 0.0062, -0.0500, ..., 0.0114, 0.0782, 0.0574], + [ 0.0060, -0.0202, -0.0411, ..., -0.1033, 0.0521, -0.0997], + [-0.0155, -0.0258, -0.0294, ..., 0.0216, -0.0320, 0.0127]], + device='cuda:0'), grad: tensor([[ 7.8510e-07, 0.0000e+00, -8.8736e-06, ..., -9.3207e-06, + 6.4149e-06, 1.6749e-05], + [ 1.2927e-06, 0.0000e+00, 4.9174e-05, ..., 6.5565e-06, + 7.5735e-06, 2.9773e-05], + [ 1.1832e-05, 0.0000e+00, -1.1170e-04, ..., 4.0792e-06, + 2.4587e-05, 6.4790e-05], + ..., + [-2.2411e-05, 0.0000e+00, 1.8656e-05, ..., 4.9204e-05, + -8.7082e-05, -3.0428e-05], + [ 3.0436e-06, 0.0000e+00, 1.3679e-05, ..., 8.0466e-05, + 1.8626e-05, 3.0041e-04], + [-3.1233e-05, 0.0000e+00, 5.8487e-06, ..., -6.6519e-04, + -7.2360e-05, -2.3899e-03]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0245, 0.0154, 0.0258, 0.0079, 0.0218, -0.0127, -0.0100, -0.0187, + -0.0103, 0.0234], device='cuda:0'), grad: tensor([ 1.8179e-05, 1.0288e-04, 2.3052e-05, -4.4167e-05, 1.9350e-03, + 2.8634e-04, 2.6915e-06, -2.8467e-04, 3.6645e-04, -2.4052e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 220.57, cls_loss 0.0081 cls_loss_mapping 0.0119 cls_loss_causal 0.6217 re_mapping 0.0114 re_causal 0.0322 /// teacc 98.80 lr 0.00010000 +Epoch 81, weight, value: tensor([[ 0.0175, 0.0110, -0.0067, ..., 0.0480, -0.0795, -0.0779], + [ 0.0362, -0.0185, 0.0176, ..., -0.0777, 0.0413, -0.0884], + [ 0.0081, -0.0048, 0.0892, ..., -0.1036, -0.0567, -0.1059], + ..., + [ 0.0097, 0.0062, -0.0512, ..., 0.0112, 0.0784, 0.0579], + [ 0.0073, -0.0202, -0.0415, ..., -0.1041, 0.0531, -0.0995], + [-0.0162, -0.0258, -0.0298, ..., 0.0221, -0.0324, 0.0130]], + device='cuda:0'), grad: tensor([[ 7.7765e-08, 0.0000e+00, 2.0713e-05, ..., -6.0260e-05, + 2.2370e-06, 1.3016e-05], + [ 8.9733e-07, 0.0000e+00, 1.1653e-05, ..., 2.2876e-04, + 3.3998e-04, 6.1321e-04], + [ 1.1455e-07, 0.0000e+00, -1.7762e-04, ..., 7.3984e-06, + 3.9898e-06, -2.7657e-05], + ..., + [-5.5768e-06, 0.0000e+00, 5.3346e-05, ..., 2.3514e-05, + 1.2755e-05, 4.5002e-05], + [ 2.7958e-06, 0.0000e+00, 3.7283e-05, ..., 2.6360e-05, + 1.0234e-04, 1.6257e-05], + [-1.6158e-07, 0.0000e+00, 5.4277e-06, ..., -3.6452e-06, + 7.1339e-06, -7.1526e-06]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0243, 0.0160, 0.0254, 0.0073, 0.0214, -0.0124, -0.0103, -0.0187, + -0.0096, 0.0233], device='cuda:0'), grad: tensor([ 4.6641e-06, 8.2922e-04, -2.6155e-04, 3.2902e-05, -8.3733e-04, + -4.5389e-05, -1.1402e-04, 1.3888e-04, 2.3329e-04, 1.8895e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 220.41, cls_loss 0.0090 cls_loss_mapping 0.0161 cls_loss_causal 0.6287 re_mapping 0.0114 re_causal 0.0322 /// teacc 98.74 lr 0.00010000 +Epoch 82, weight, value: tensor([[ 0.0174, 0.0110, -0.0062, ..., 0.0479, -0.0798, -0.0782], + [ 0.0365, -0.0185, 0.0174, ..., -0.0786, 0.0424, -0.0867], + [ 0.0079, -0.0048, 0.0900, ..., -0.1052, -0.0572, -0.1064], + ..., + [ 0.0100, 0.0062, -0.0518, ..., 0.0109, 0.0779, 0.0576], + [ 0.0074, -0.0202, -0.0422, ..., -0.1049, 0.0533, -0.0994], + [-0.0172, -0.0258, -0.0298, ..., 0.0230, -0.0334, 0.0128]], + device='cuda:0'), grad: tensor([[ 6.2399e-08, 0.0000e+00, 8.9854e-06, ..., -3.2514e-05, + 2.4047e-06, 4.5598e-06], + [ 8.4331e-07, 0.0000e+00, 3.7789e-05, ..., 3.9674e-06, + 2.3827e-05, 3.0220e-05], + [ 4.6985e-07, 0.0000e+00, -3.1352e-04, ..., -6.8769e-06, + 5.1200e-05, 7.2598e-05], + ..., + [-5.0142e-06, 0.0000e+00, 2.9914e-06, ..., -2.5094e-05, + -1.4853e-04, -2.5344e-04], + [ 2.5565e-07, 0.0000e+00, 7.0333e-06, ..., 4.7646e-06, + -3.3170e-05, -2.3454e-05], + [ 3.9395e-07, 0.0000e+00, 2.5742e-06, ..., 5.6736e-06, + 4.3035e-05, 1.4767e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0244, 0.0168, 0.0254, 0.0071, 0.0214, -0.0115, -0.0106, -0.0195, + -0.0093, 0.0228], device='cuda:0'), grad: tensor([-3.5107e-05, 9.7930e-05, -2.7895e-04, 5.4717e-05, 2.5964e-04, + 1.6451e-05, 2.0635e-04, -3.6502e-04, -3.4302e-05, 7.8619e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 220.64, cls_loss 0.0070 cls_loss_mapping 0.0144 cls_loss_causal 0.6264 re_mapping 0.0116 re_causal 0.0333 /// teacc 98.84 lr 0.00010000 +Epoch 83, weight, value: tensor([[ 0.0173, 0.0110, -0.0060, ..., 0.0485, -0.0804, -0.0773], + [ 0.0363, -0.0185, 0.0173, ..., -0.0792, 0.0421, -0.0877], + [ 0.0077, -0.0048, 0.0904, ..., -0.1060, -0.0580, -0.1072], + ..., + [ 0.0100, 0.0062, -0.0521, ..., 0.0109, 0.0790, 0.0579], + [ 0.0072, -0.0202, -0.0423, ..., -0.1050, 0.0540, -0.1005], + [-0.0163, -0.0258, -0.0302, ..., 0.0227, -0.0339, 0.0129]], + device='cuda:0'), grad: tensor([[ 7.9814e-07, 0.0000e+00, 4.1664e-05, ..., -2.8419e-04, + 5.6438e-06, 4.9993e-06], + [ 1.3355e-06, 0.0000e+00, 1.3843e-05, ..., 2.6040e-06, + -1.0580e-06, 6.7763e-06], + [ 1.0833e-05, 0.0000e+00, 1.8001e-04, ..., 2.6636e-06, + 4.3422e-05, 2.7269e-05], + ..., + [-1.9181e-04, 0.0000e+00, -1.9968e-05, ..., -1.6916e-04, + -5.3167e-04, -1.3485e-03], + [ 1.2651e-05, 0.0000e+00, 1.2383e-05, ..., 2.0131e-05, + 4.3035e-05, 8.1480e-05], + [ 1.4973e-04, 0.0000e+00, 9.9763e-06, ..., 3.0255e-04, + 4.0197e-04, 1.1158e-03]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0239, 0.0165, 0.0251, 0.0072, 0.0212, -0.0113, -0.0107, -0.0193, + -0.0087, 0.0225], device='cuda:0'), grad: tensor([-3.5691e-04, 3.2067e-05, 5.5361e-04, -6.8092e-04, 1.3101e-04, + 2.8276e-04, -1.3554e-04, -1.5221e-03, 1.7250e-04, 1.5240e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 221.04, cls_loss 0.0082 cls_loss_mapping 0.0146 cls_loss_causal 0.6263 re_mapping 0.0108 re_causal 0.0324 /// teacc 98.74 lr 0.00010000 +Epoch 84, weight, value: tensor([[ 0.0172, 0.0110, -0.0062, ..., 0.0491, -0.0808, -0.0779], + [ 0.0366, -0.0185, 0.0173, ..., -0.0799, 0.0417, -0.0877], + [ 0.0079, -0.0048, 0.0911, ..., -0.1079, -0.0581, -0.1075], + ..., + [ 0.0091, 0.0062, -0.0528, ..., 0.0106, 0.0799, 0.0578], + [ 0.0075, -0.0202, -0.0432, ..., -0.1056, 0.0536, -0.1009], + [-0.0150, -0.0258, -0.0308, ..., 0.0229, -0.0345, 0.0133]], + device='cuda:0'), grad: tensor([[ 1.1288e-06, 0.0000e+00, 3.1386e-07, ..., -1.0365e-04, + 9.1493e-06, 2.5071e-06], + [ 2.0087e-05, 0.0000e+00, 3.6461e-07, ..., 6.2659e-06, + -7.9751e-05, 3.0026e-05], + [ 7.7859e-06, 0.0000e+00, -6.2771e-06, ..., 1.5199e-05, + 2.8938e-05, 9.6858e-06], + ..., + [ 4.0359e-03, 0.0000e+00, 2.8051e-06, ..., 1.7524e-05, + 1.0544e-02, 2.7485e-03], + [-4.0970e-03, 0.0000e+00, 8.6008e-07, ..., 6.6124e-06, + -1.0841e-02, -2.8172e-03], + [ 6.7838e-06, 0.0000e+00, 5.4343e-07, ..., 4.9174e-05, + 5.2482e-05, 2.1949e-05]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0238, 0.0160, 0.0249, 0.0074, 0.0211, -0.0109, -0.0107, -0.0186, + -0.0095, 0.0226], device='cuda:0'), grad: tensor([-3.6716e-04, -1.6594e-04, 9.5904e-05, 9.5069e-05, -1.2890e-06, + 3.9577e-04, 4.1068e-05, 1.4587e-02, -1.4984e-02, 2.9659e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 83---------------------------------------------------- +epoch 83, time 221.06, cls_loss 0.0077 cls_loss_mapping 0.0119 cls_loss_causal 0.6168 re_mapping 0.0111 re_causal 0.0323 /// teacc 98.94 lr 0.00010000 +Epoch 85, weight, value: tensor([[ 0.0170, 0.0110, -0.0063, ..., 0.0481, -0.0821, -0.0785], + [ 0.0366, -0.0185, 0.0171, ..., -0.0805, 0.0427, -0.0878], + [ 0.0079, -0.0048, 0.0932, ..., -0.1085, -0.0584, -0.1086], + ..., + [ 0.0079, 0.0062, -0.0546, ..., 0.0104, 0.0794, 0.0579], + [ 0.0087, -0.0202, -0.0441, ..., -0.1067, 0.0542, -0.1022], + [-0.0134, -0.0258, -0.0310, ..., 0.0228, -0.0350, 0.0136]], + device='cuda:0'), grad: tensor([[ 1.7928e-07, 0.0000e+00, -2.6643e-05, ..., -1.1021e-04, + 1.4305e-06, 4.4405e-06], + [-1.0692e-06, 0.0000e+00, 9.7007e-06, ..., 1.9111e-06, + -2.9743e-05, -2.0675e-06], + [ 8.0978e-07, 0.0000e+00, -3.7283e-05, ..., 3.3062e-06, + 1.0096e-05, 3.5781e-06], + ..., + [-3.8184e-06, 0.0000e+00, 8.7246e-06, ..., 2.4885e-06, + -7.7710e-06, -1.3188e-05], + [ 1.5963e-06, 0.0000e+00, 1.5140e-05, ..., 1.0662e-05, + -1.0349e-05, 7.7486e-06], + [ 1.6019e-06, 0.0000e+00, 6.4857e-06, ..., 9.5889e-06, + 4.2729e-06, -2.4855e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0256, 0.0170, 0.0259, 0.0061, 0.0212, -0.0108, -0.0094, -0.0194, + -0.0098, 0.0228], device='cuda:0'), grad: tensor([-1.4949e-04, -7.3195e-05, -4.9099e-06, -6.3702e-06, 2.3589e-05, + 5.1022e-05, 1.0979e-04, 4.3511e-06, 2.9907e-05, 1.5512e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 220.48, cls_loss 0.0086 cls_loss_mapping 0.0140 cls_loss_causal 0.5964 re_mapping 0.0113 re_causal 0.0312 /// teacc 98.72 lr 0.00010000 +Epoch 86, weight, value: tensor([[ 0.0170, 0.0110, -0.0053, ..., 0.0499, -0.0816, -0.0787], + [ 0.0364, -0.0185, 0.0150, ..., -0.0837, 0.0416, -0.0877], + [ 0.0077, -0.0048, 0.0953, ..., -0.1101, -0.0579, -0.1092], + ..., + [ 0.0082, 0.0062, -0.0563, ..., 0.0101, 0.0801, 0.0581], + [ 0.0087, -0.0202, -0.0441, ..., -0.1082, 0.0547, -0.1026], + [-0.0134, -0.0258, -0.0311, ..., 0.0225, -0.0357, 0.0133]], + device='cuda:0'), grad: tensor([[ 1.3225e-07, 0.0000e+00, 2.2590e-05, ..., 2.9862e-05, + 2.5958e-05, 1.2510e-05], + [ 3.3807e-07, 0.0000e+00, -6.3276e-04, ..., 6.2212e-06, + -5.1785e-04, 1.8731e-05], + [ 1.0813e-06, 0.0000e+00, 3.7026e-04, ..., 3.1739e-05, + 4.1032e-04, 1.7643e-04], + ..., + [-1.8012e-06, 0.0000e+00, 9.8705e-05, ..., -5.2363e-05, + -1.2910e-04, -3.5048e-04], + [ 9.4995e-08, 0.0000e+00, 3.2037e-05, ..., 2.6777e-05, + 7.9572e-05, 1.1641e-04], + [ 1.3830e-07, 0.0000e+00, 7.1973e-06, ..., -1.6898e-05, + 9.6336e-06, -2.4125e-05]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0240, 0.0159, 0.0270, 0.0058, 0.0215, -0.0113, -0.0092, -0.0196, + -0.0096, 0.0222], device='cuda:0'), grad: tensor([ 1.1557e-04, -1.4992e-03, 1.2960e-03, 1.8179e-04, 1.6022e-04, + 9.5665e-05, -1.9491e-04, -5.1403e-04, 3.4380e-04, 1.3970e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 220.52, cls_loss 0.0058 cls_loss_mapping 0.0104 cls_loss_causal 0.6042 re_mapping 0.0105 re_causal 0.0312 /// teacc 98.91 lr 0.00010000 +Epoch 87, weight, value: tensor([[ 0.0169, 0.0110, -0.0048, ..., 0.0503, -0.0821, -0.0790], + [ 0.0364, -0.0185, 0.0151, ..., -0.0843, 0.0417, -0.0884], + [ 0.0075, -0.0048, 0.0954, ..., -0.1113, -0.0590, -0.1104], + ..., + [ 0.0082, 0.0062, -0.0563, ..., 0.0091, 0.0808, 0.0579], + [ 0.0086, -0.0202, -0.0447, ..., -0.1095, 0.0549, -0.1033], + [-0.0128, -0.0258, -0.0312, ..., 0.0223, -0.0360, 0.0134]], + device='cuda:0'), grad: tensor([[ 5.2154e-07, 0.0000e+00, -2.2110e-06, ..., 1.4957e-06, + 8.5756e-06, 3.5651e-06], + [-1.8403e-06, 0.0000e+00, -3.8482e-06, ..., -6.2734e-06, + -3.1203e-05, 6.6645e-06], + [ 6.1048e-07, 0.0000e+00, 8.6129e-06, ..., 3.4004e-05, + 1.2122e-05, 3.5390e-06], + ..., + [-3.2736e-07, 0.0000e+00, 2.6412e-06, ..., 7.6890e-06, + 1.3877e-06, 1.3277e-05], + [ 4.3446e-07, 0.0000e+00, 4.0606e-06, ..., 2.0534e-05, + 2.7232e-06, 2.9743e-05], + [-1.6196e-06, 0.0000e+00, 5.3011e-06, ..., -1.9282e-05, + 2.0452e-06, -5.5313e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0238, 0.0160, 0.0260, 0.0061, 0.0222, -0.0115, -0.0099, -0.0193, + -0.0097, 0.0223], device='cuda:0'), grad: tensor([ 3.8892e-05, -7.0930e-05, 1.1450e-04, -1.0580e-04, 2.9057e-05, + 3.0547e-05, -9.3937e-05, 2.4259e-05, 6.7055e-05, -3.3826e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 220.97, cls_loss 0.0058 cls_loss_mapping 0.0103 cls_loss_causal 0.6077 re_mapping 0.0106 re_causal 0.0313 /// teacc 98.83 lr 0.00010000 +Epoch 88, weight, value: tensor([[ 0.0167, 0.0110, -0.0045, ..., 0.0508, -0.0826, -0.0794], + [ 0.0365, -0.0185, 0.0152, ..., -0.0847, 0.0422, -0.0890], + [ 0.0074, -0.0048, 0.0960, ..., -0.1123, -0.0599, -0.1112], + ..., + [ 0.0082, 0.0062, -0.0569, ..., 0.0088, 0.0812, 0.0579], + [ 0.0084, -0.0202, -0.0451, ..., -0.1105, 0.0550, -0.1040], + [-0.0128, -0.0258, -0.0314, ..., 0.0222, -0.0365, 0.0135]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, 0.0000e+00, 6.2510e-06, ..., -2.4159e-06, + 5.9279e-07, 1.0982e-05], + [ 3.2131e-08, 0.0000e+00, 1.1437e-06, ..., 3.6024e-06, + 4.0280e-07, 5.8562e-06], + [ 3.4459e-08, 0.0000e+00, 5.3570e-06, ..., 5.9828e-06, + 7.1991e-07, 1.6347e-05], + ..., + [-1.5646e-07, 0.0000e+00, 4.7609e-06, ..., 3.7309e-06, + -1.0412e-06, 1.3605e-05], + [ 5.1223e-09, 0.0000e+00, 3.4459e-06, ..., 2.7828e-06, + -1.0636e-06, 3.9190e-06], + [ 5.1223e-08, 0.0000e+00, 2.2855e-06, ..., -5.1893e-06, + 1.2629e-06, -4.3958e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0237, 0.0163, 0.0257, 0.0062, 0.0226, -0.0119, -0.0098, -0.0193, + -0.0100, 0.0222], device='cuda:0'), grad: tensor([ 2.5541e-05, 8.0392e-06, 3.8832e-05, -7.2896e-05, -1.4335e-05, + 2.9411e-06, 9.9316e-06, 2.0027e-05, 7.4841e-06, -2.5615e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 220.97, cls_loss 0.0064 cls_loss_mapping 0.0104 cls_loss_causal 0.5618 re_mapping 0.0112 re_causal 0.0313 /// teacc 98.74 lr 0.00010000 +Epoch 89, weight, value: tensor([[ 0.0167, 0.0110, -0.0044, ..., 0.0509, -0.0831, -0.0802], + [ 0.0364, -0.0185, 0.0153, ..., -0.0853, 0.0427, -0.0886], + [ 0.0074, -0.0048, 0.0964, ..., -0.1135, -0.0603, -0.1120], + ..., + [ 0.0082, 0.0062, -0.0572, ..., 0.0087, 0.0813, 0.0578], + [ 0.0084, -0.0202, -0.0457, ..., -0.1113, 0.0551, -0.1050], + [-0.0127, -0.0258, -0.0317, ..., 0.0233, -0.0365, 0.0143]], + device='cuda:0'), grad: tensor([[ 1.6554e-07, 0.0000e+00, -1.3858e-06, ..., 3.1926e-06, + 4.0792e-06, 2.1622e-05], + [ 4.2934e-07, 0.0000e+00, 2.3972e-06, ..., 1.5404e-06, + 1.0379e-05, 1.3240e-05], + [ 5.2676e-06, 0.0000e+00, 1.8939e-05, ..., 5.4613e-06, + 4.6551e-05, 4.2915e-05], + ..., + [-9.8497e-06, 0.0000e+00, 8.1211e-06, ..., 1.8656e-05, + -5.0068e-05, -8.2195e-05], + [ 3.2317e-07, 0.0000e+00, 2.0508e-06, ..., 9.8646e-06, + -1.2778e-05, 1.3024e-05], + [ 7.5763e-07, 0.0000e+00, 3.2801e-06, ..., -1.4400e-04, + 5.5283e-06, -1.0127e-04]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0239, 0.0169, 0.0257, 0.0072, 0.0217, -0.0124, -0.0103, -0.0196, + -0.0102, 0.0231], device='cuda:0'), grad: tensor([ 3.0488e-05, 3.7372e-05, 1.9479e-04, 5.1362e-07, 1.3471e-04, + 5.2571e-05, -4.7624e-05, -1.9205e-04, 1.2927e-05, -2.2364e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 220.58, cls_loss 0.0057 cls_loss_mapping 0.0091 cls_loss_causal 0.6211 re_mapping 0.0104 re_causal 0.0316 /// teacc 98.88 lr 0.00010000 +Epoch 90, weight, value: tensor([[ 0.0166, 0.0110, -0.0044, ..., 0.0511, -0.0834, -0.0809], + [ 0.0364, -0.0185, 0.0151, ..., -0.0853, 0.0428, -0.0886], + [ 0.0073, -0.0048, 0.0968, ..., -0.1142, -0.0611, -0.1135], + ..., + [ 0.0076, 0.0062, -0.0575, ..., 0.0080, 0.0815, 0.0573], + [ 0.0083, -0.0202, -0.0461, ..., -0.1122, 0.0557, -0.1055], + [-0.0114, -0.0258, -0.0321, ..., 0.0242, -0.0363, 0.0157]], + device='cuda:0'), grad: tensor([[ 2.9709e-07, 0.0000e+00, 6.7651e-06, ..., -1.4126e-05, + 1.9819e-06, 3.9190e-06], + [ 1.9185e-06, 0.0000e+00, 2.0098e-06, ..., 1.8049e-06, + -1.9260e-06, 1.4454e-05], + [ 8.8383e-07, 0.0000e+00, -1.6034e-04, ..., 2.1961e-06, + 6.1020e-06, -1.3836e-05], + ..., + [-1.1784e-04, 0.0000e+00, 1.3709e-05, ..., -5.0664e-05, + -1.5467e-05, -1.3723e-03], + [ 3.5018e-07, 0.0000e+00, 3.2008e-05, ..., 5.6401e-06, + -4.8056e-06, 1.7714e-06], + [ 1.1128e-04, 0.0000e+00, 1.3217e-05, ..., 3.6806e-05, + 8.4117e-06, 1.2903e-03]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0240, 0.0169, 0.0252, 0.0076, 0.0211, -0.0128, -0.0103, -0.0198, + -0.0101, 0.0241], device='cuda:0'), grad: tensor([ 8.0913e-06, 1.2569e-05, -1.7786e-04, 1.6737e-04, 4.2230e-05, + -1.0288e-04, 2.2769e-05, -1.2169e-03, 4.4793e-05, 1.1997e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 220.75, cls_loss 0.0055 cls_loss_mapping 0.0096 cls_loss_causal 0.5820 re_mapping 0.0103 re_causal 0.0301 /// teacc 98.91 lr 0.00010000 +Epoch 91, weight, value: tensor([[ 0.0165, 0.0110, -0.0041, ..., 0.0516, -0.0837, -0.0814], + [ 0.0365, -0.0185, 0.0143, ..., -0.0856, 0.0429, -0.0889], + [ 0.0072, -0.0048, 0.0972, ..., -0.1151, -0.0615, -0.1137], + ..., + [ 0.0077, 0.0062, -0.0561, ..., 0.0080, 0.0820, 0.0578], + [ 0.0082, -0.0202, -0.0463, ..., -0.1124, 0.0560, -0.1058], + [-0.0116, -0.0258, -0.0325, ..., 0.0231, -0.0369, 0.0147]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 0.0000e+00, 3.1944e-06, ..., -1.5534e-06, + 1.5199e-06, 5.0887e-06], + [ 7.9628e-08, 0.0000e+00, 5.2862e-06, ..., 5.6326e-06, + 7.8827e-06, 5.1051e-05], + [ 1.5600e-08, 0.0000e+00, -7.4029e-05, ..., 3.1013e-07, + 2.1830e-05, 1.1069e-04], + ..., + [-6.2305e-07, 0.0000e+00, 6.8881e-06, ..., 1.7509e-06, + -4.2534e-04, -2.2392e-03], + [ 3.3528e-08, 0.0000e+00, 1.8701e-05, ..., 6.9849e-07, + 3.5256e-05, 3.2067e-05], + [ 4.1770e-07, 0.0000e+00, 3.3416e-06, ..., 7.1526e-06, + 2.1732e-04, 1.1396e-03]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0237, 0.0166, 0.0249, 0.0068, 0.0220, -0.0125, -0.0104, -0.0186, + -0.0100, 0.0228], device='cuda:0'), grad: tensor([ 1.1839e-05, 6.7294e-05, 6.0685e-06, 8.8096e-05, 1.0643e-03, + -1.5879e-04, 6.7949e-05, -2.6665e-03, 1.5128e-04, 1.3685e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 220.46, cls_loss 0.0067 cls_loss_mapping 0.0104 cls_loss_causal 0.6237 re_mapping 0.0098 re_causal 0.0299 /// teacc 98.87 lr 0.00010000 +Epoch 92, weight, value: tensor([[ 0.0165, 0.0110, -0.0037, ..., 0.0521, -0.0841, -0.0814], + [ 0.0365, -0.0185, 0.0142, ..., -0.0864, 0.0433, -0.0889], + [ 0.0071, -0.0048, 0.0975, ..., -0.1165, -0.0620, -0.1143], + ..., + [ 0.0078, 0.0062, -0.0566, ..., 0.0078, 0.0825, 0.0585], + [ 0.0082, -0.0202, -0.0468, ..., -0.1138, 0.0556, -0.1067], + [-0.0117, -0.0258, -0.0329, ..., 0.0229, -0.0379, 0.0141]], + device='cuda:0'), grad: tensor([[ 3.7951e-08, 0.0000e+00, -1.9774e-05, ..., -3.1769e-05, + -3.0041e-05, 2.7809e-06], + [ 1.6019e-07, 0.0000e+00, 2.8089e-06, ..., 3.3225e-07, + -3.0309e-05, 3.1404e-06], + [ 6.5658e-08, 0.0000e+00, -3.0205e-05, ..., 1.6272e-05, + 1.1228e-05, 1.2115e-05], + ..., + [-1.3830e-07, 0.0000e+00, 2.3916e-05, ..., 3.6247e-06, + 3.7793e-06, 3.9749e-06], + [ 5.4017e-08, 0.0000e+00, 2.0117e-05, ..., 2.1189e-05, + 2.4527e-05, 6.5789e-06], + [ 3.8138e-07, 0.0000e+00, 1.8058e-06, ..., 4.0568e-06, + 4.3847e-06, -6.9709e-07]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0232, 0.0168, 0.0246, 0.0068, 0.0222, -0.0127, -0.0098, -0.0181, + -0.0115, 0.0226], device='cuda:0'), grad: tensor([-1.0097e-04, -4.1395e-05, 1.6624e-06, -2.8625e-05, -5.3495e-05, + 1.5289e-05, 5.7995e-05, 4.7654e-05, 9.0599e-05, 1.1459e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 220.88, cls_loss 0.0056 cls_loss_mapping 0.0107 cls_loss_causal 0.6126 re_mapping 0.0103 re_causal 0.0296 /// teacc 98.85 lr 0.00010000 +Epoch 93, weight, value: tensor([[ 0.0164, 0.0110, -0.0035, ..., 0.0525, -0.0847, -0.0818], + [ 0.0363, -0.0185, 0.0141, ..., -0.0870, 0.0434, -0.0896], + [ 0.0070, -0.0048, 0.0980, ..., -0.1174, -0.0631, -0.1147], + ..., + [ 0.0067, 0.0062, -0.0570, ..., 0.0077, 0.0831, 0.0589], + [ 0.0083, -0.0202, -0.0472, ..., -0.1147, 0.0559, -0.1076], + [-0.0117, -0.0258, -0.0330, ..., 0.0228, -0.0388, 0.0138]], + device='cuda:0'), grad: tensor([[ 3.9581e-08, 0.0000e+00, 8.2776e-06, ..., 2.0675e-06, + 3.4664e-06, 1.0580e-06], + [ 1.1176e-07, 0.0000e+00, 5.0105e-07, ..., 1.2284e-06, + -3.5632e-06, 1.2573e-06], + [ 9.3598e-08, 0.0000e+00, -1.5028e-05, ..., 9.4809e-07, + 9.1176e-07, 6.8219e-07], + ..., + [-7.9349e-07, 0.0000e+00, 1.5385e-06, ..., 6.4773e-07, + -8.9174e-07, -3.1684e-06], + [ 7.4273e-08, 0.0000e+00, 2.2221e-06, ..., 1.9614e-06, + 1.2033e-06, 1.7304e-06], + [ 1.0757e-07, 0.0000e+00, 2.0652e-07, ..., 1.4238e-05, + 1.9614e-06, 2.4274e-05]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0232, 0.0167, 0.0243, 0.0073, 0.0224, -0.0125, -0.0089, -0.0184, + -0.0119, 0.0221], device='cuda:0'), grad: tensor([ 2.2262e-05, -2.1923e-06, -2.0534e-05, 8.0541e-06, -1.4365e-05, + -4.3422e-05, 4.0531e-06, 1.0533e-06, 1.8850e-05, 2.6196e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 220.81, cls_loss 0.0053 cls_loss_mapping 0.0109 cls_loss_causal 0.5958 re_mapping 0.0099 re_causal 0.0304 /// teacc 98.77 lr 0.00010000 +Epoch 94, weight, value: tensor([[ 0.0164, 0.0110, -0.0032, ..., 0.0527, -0.0853, -0.0824], + [ 0.0365, -0.0185, 0.0145, ..., -0.0870, 0.0443, -0.0894], + [ 0.0070, -0.0048, 0.0982, ..., -0.1179, -0.0642, -0.1148], + ..., + [ 0.0068, 0.0062, -0.0577, ..., 0.0074, 0.0832, 0.0582], + [ 0.0082, -0.0202, -0.0471, ..., -0.1154, 0.0558, -0.1087], + [-0.0117, -0.0258, -0.0333, ..., 0.0230, -0.0393, 0.0145]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 1.5050e-05, ..., 2.2594e-06, + 3.6769e-06, 2.7996e-06], + [ 1.4668e-08, 0.0000e+00, 7.0110e-06, ..., 1.4268e-06, + 5.3942e-06, 5.2117e-06], + [ 2.3283e-09, 0.0000e+00, -1.9276e-04, ..., -2.0027e-05, + 1.0468e-06, -3.1404e-06], + ..., + [-1.8161e-07, 0.0000e+00, 2.1532e-06, ..., 7.5903e-07, + -1.7181e-05, -2.0087e-05], + [ 9.3132e-10, 0.0000e+00, 1.1706e-04, ..., 1.4752e-05, + 1.8273e-06, 3.2596e-06], + [ 1.4645e-07, 0.0000e+00, 4.2140e-05, ..., 8.1420e-05, + 7.3612e-06, 1.4639e-04]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0234, 0.0176, 0.0238, 0.0068, 0.0223, -0.0124, -0.0085, -0.0189, + -0.0120, 0.0226], device='cuda:0'), grad: tensor([ 3.6091e-05, 2.2978e-05, -3.1447e-04, 6.5029e-05, -1.0324e-04, + -4.9591e-05, -2.1249e-05, -3.8832e-05, 2.0230e-04, 2.0087e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 220.45, cls_loss 0.0063 cls_loss_mapping 0.0107 cls_loss_causal 0.6037 re_mapping 0.0099 re_causal 0.0288 /// teacc 98.78 lr 0.00010000 +Epoch 95, weight, value: tensor([[ 0.0163, 0.0110, -0.0030, ..., 0.0531, -0.0857, -0.0827], + [ 0.0364, -0.0185, 0.0144, ..., -0.0874, 0.0446, -0.0894], + [ 0.0070, -0.0048, 0.0987, ..., -0.1186, -0.0637, -0.1156], + ..., + [ 0.0068, 0.0062, -0.0581, ..., 0.0073, 0.0833, 0.0586], + [ 0.0081, -0.0202, -0.0477, ..., -0.1161, 0.0563, -0.1074], + [-0.0118, -0.0258, -0.0333, ..., 0.0231, -0.0411, 0.0141]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 0.0000e+00, 2.6971e-06, ..., 5.5023e-06, + 4.5262e-06, 2.8051e-06], + [ 1.6368e-07, 0.0000e+00, 8.2552e-06, ..., 4.3511e-06, + 6.3106e-06, 1.0893e-05], + [ 7.7998e-08, 0.0000e+00, -7.8201e-05, ..., 5.8254e-07, + -8.3260e-07, 1.6224e-06], + ..., + [-4.1514e-07, 0.0000e+00, 1.3202e-05, ..., 9.7528e-06, + -8.6986e-07, 2.1741e-05], + [ 2.6077e-08, 0.0000e+00, 7.4729e-06, ..., 1.2368e-05, + -2.5839e-05, 2.6256e-05], + [ 2.7614e-07, 0.0000e+00, 8.3586e-07, ..., -5.3558e-03, + 3.4478e-06, -1.3283e-02]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0234, 0.0177, 0.0242, 0.0062, 0.0222, -0.0115, -0.0083, -0.0189, + -0.0114, 0.0214], device='cuda:0'), grad: tensor([ 5.3167e-05, 3.4571e-05, -9.5904e-05, -6.4087e-04, 1.1024e-02, + 8.9693e-04, -2.5201e-04, 3.6329e-05, 2.1304e-07, -1.1055e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 221.01, cls_loss 0.0060 cls_loss_mapping 0.0089 cls_loss_causal 0.6285 re_mapping 0.0103 re_causal 0.0309 /// teacc 98.80 lr 0.00010000 +Epoch 96, weight, value: tensor([[ 0.0163, 0.0110, -0.0031, ..., 0.0530, -0.0861, -0.0842], + [ 0.0364, -0.0185, 0.0140, ..., -0.0884, 0.0443, -0.0901], + [ 0.0070, -0.0048, 0.1007, ..., -0.1195, -0.0642, -0.1125], + ..., + [ 0.0069, 0.0062, -0.0613, ..., 0.0065, 0.0838, 0.0576], + [ 0.0081, -0.0202, -0.0482, ..., -0.1173, 0.0564, -0.1094], + [-0.0118, -0.0258, -0.0336, ..., 0.0241, -0.0410, 0.0151]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 1.5354e-04, ..., -4.3251e-06, + 9.7677e-06, 2.0079e-06], + [ 3.8883e-08, 0.0000e+00, 2.8163e-06, ..., 1.0073e-05, + -1.3053e-04, 5.2482e-05], + [ 5.6811e-08, 0.0000e+00, -2.6917e-04, ..., 1.0245e-06, + 1.0608e-06, 3.7272e-06], + ..., + [-1.6647e-07, 0.0000e+00, 5.7705e-06, ..., 2.2296e-06, + 1.2374e-04, -6.0908e-07], + [ 2.8638e-08, 0.0000e+00, 4.7088e-05, ..., 6.5491e-06, + 1.1772e-05, 2.3514e-05], + [ 1.4435e-08, 0.0000e+00, 1.5823e-06, ..., -1.5557e-05, + 1.6615e-05, -2.7537e-05]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0237, 0.0171, 0.0252, 0.0060, 0.0225, -0.0116, -0.0082, -0.0195, + -0.0118, 0.0221], device='cuda:0'), grad: tensor([ 1.6117e-04, -1.3530e-04, -2.5415e-04, 6.6042e-05, -1.4496e-04, + 6.4671e-05, 1.5453e-05, 1.7059e-04, 8.6844e-05, -3.1114e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 220.47, cls_loss 0.0060 cls_loss_mapping 0.0097 cls_loss_causal 0.6027 re_mapping 0.0092 re_causal 0.0286 /// teacc 98.78 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 0.0163, 0.0110, -0.0032, ..., 0.0528, -0.0868, -0.0848], + [ 0.0364, -0.0185, 0.0141, ..., -0.0887, 0.0441, -0.0912], + [ 0.0070, -0.0048, 0.1014, ..., -0.1201, -0.0647, -0.1129], + ..., + [ 0.0069, 0.0062, -0.0622, ..., 0.0065, 0.0852, 0.0584], + [ 0.0081, -0.0202, -0.0488, ..., -0.1177, 0.0563, -0.1100], + [-0.0118, -0.0258, -0.0339, ..., 0.0252, -0.0415, 0.0161]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 0.0000e+00, -9.1419e-06, ..., -6.8694e-06, + 1.5087e-05, 1.4104e-05], + [ 3.1851e-07, 0.0000e+00, 6.5416e-06, ..., 2.0340e-06, + 5.2229e-06, 3.8221e-06], + [ 7.6834e-08, 0.0000e+00, -1.0975e-05, ..., 3.3956e-06, + 2.2382e-05, 1.7077e-05], + ..., + [-1.1660e-06, 0.0000e+00, 7.5214e-06, ..., 1.8999e-07, + -3.4243e-05, -4.1246e-05], + [ 8.3819e-09, 0.0000e+00, -5.6950e-07, ..., 1.3560e-06, + -1.9968e-05, 7.8743e-07], + [ 6.3237e-07, 0.0000e+00, 1.1828e-06, ..., 3.4384e-06, + 3.8370e-06, 7.5288e-06]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0240, 0.0167, 0.0254, 0.0062, 0.0211, -0.0124, -0.0073, -0.0187, + -0.0122, 0.0231], device='cuda:0'), grad: tensor([ 5.0992e-05, 2.7671e-05, 2.8685e-05, 4.5598e-05, 4.4614e-05, + -1.1820e-04, -4.1753e-05, -5.2363e-05, -3.3565e-06, 1.8209e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 220.70, cls_loss 0.0058 cls_loss_mapping 0.0096 cls_loss_causal 0.6073 re_mapping 0.0095 re_causal 0.0288 /// teacc 98.86 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0163, 0.0110, -0.0036, ..., 0.0513, -0.0871, -0.0850], + [ 0.0364, -0.0185, 0.0139, ..., -0.0896, 0.0440, -0.0915], + [ 0.0069, -0.0048, 0.1028, ..., -0.1228, -0.0654, -0.1133], + ..., + [ 0.0069, 0.0062, -0.0627, ..., 0.0062, 0.0858, 0.0586], + [ 0.0081, -0.0202, -0.0495, ..., -0.1185, 0.0566, -0.1107], + [-0.0116, -0.0258, -0.0324, ..., 0.0269, -0.0419, 0.0160]], + device='cuda:0'), grad: tensor([[ 2.4401e-07, 0.0000e+00, 1.1623e-06, ..., -2.3544e-06, + 9.4324e-06, 1.6615e-05], + [ 2.1420e-07, 0.0000e+00, 2.4717e-06, ..., 1.1832e-05, + 2.7180e-05, 8.7321e-05], + [ 2.1420e-06, 0.0000e+00, -7.8604e-06, ..., 2.4945e-05, + 4.0793e-04, 3.5644e-04], + ..., + [ 1.1716e-06, 0.0000e+00, -1.8999e-05, ..., -2.9922e-04, + -1.3905e-03, -2.3174e-03], + [ 3.4533e-06, 0.0000e+00, 1.8448e-05, ..., 1.0535e-05, + 5.1796e-05, 9.2089e-05], + [ 3.4738e-07, 0.0000e+00, 1.8235e-06, ..., 2.4819e-04, + 8.5878e-04, 1.7586e-03]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0253, 0.0164, 0.0259, 0.0061, 0.0214, -0.0133, -0.0075, -0.0186, + -0.0125, 0.0244], device='cuda:0'), grad: tensor([ 2.5928e-05, 7.9274e-05, 1.1034e-03, -1.3411e-04, 3.8624e-05, + -4.0698e-04, 3.5167e-04, -3.2787e-03, 2.4867e-04, 1.9703e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 220.52, cls_loss 0.0050 cls_loss_mapping 0.0087 cls_loss_causal 0.6016 re_mapping 0.0096 re_causal 0.0288 /// teacc 98.87 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0163, 0.0110, -0.0022, ..., 0.0520, -0.0876, -0.0855], + [ 0.0364, -0.0185, 0.0136, ..., -0.0906, 0.0442, -0.0916], + [ 0.0068, -0.0048, 0.1043, ..., -0.1236, -0.0658, -0.1114], + ..., + [ 0.0068, 0.0062, -0.0643, ..., 0.0062, 0.0861, 0.0583], + [ 0.0081, -0.0202, -0.0503, ..., -0.1187, 0.0574, -0.1106], + [-0.0115, -0.0258, -0.0331, ..., 0.0265, -0.0431, 0.0154]], + device='cuda:0'), grad: tensor([[ 9.0804e-09, 0.0000e+00, -2.6133e-06, ..., -4.6752e-06, + 1.1288e-05, 2.5295e-06], + [ 3.9348e-08, 0.0000e+00, -1.2713e-07, ..., 3.9376e-06, + 2.9001e-06, 1.4052e-05], + [ 2.1653e-08, 0.0000e+00, 4.3120e-07, ..., 2.4270e-06, + 2.9922e-05, 2.1812e-06], + ..., + [-2.1700e-07, 0.0000e+00, 5.7742e-08, ..., -1.6347e-05, + 6.6757e-05, -9.6083e-05], + [ 1.4203e-08, 0.0000e+00, 1.8766e-07, ..., 3.2037e-06, + -1.5521e-04, 6.3255e-06], + [ 7.5204e-08, 0.0000e+00, 1.2415e-06, ..., -5.1594e-04, + 2.8491e-05, -1.1435e-03]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0244, 0.0163, 0.0269, 0.0052, 0.0218, -0.0125, -0.0086, -0.0190, + -0.0121, 0.0237], device='cuda:0'), grad: tensor([ 3.0488e-05, 1.4812e-05, 5.3555e-05, -4.1336e-05, 9.2125e-04, + 7.3135e-05, -4.3780e-05, 4.2468e-05, -2.2542e-04, -8.2636e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 220.85, cls_loss 0.0043 cls_loss_mapping 0.0072 cls_loss_causal 0.6018 re_mapping 0.0093 re_causal 0.0287 /// teacc 98.82 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0163, 0.0110, -0.0020, ..., 0.0523, -0.0881, -0.0864], + [ 0.0364, -0.0185, 0.0129, ..., -0.0910, 0.0445, -0.0901], + [ 0.0068, -0.0048, 0.1052, ..., -0.1246, -0.0661, -0.1116], + ..., + [ 0.0069, 0.0062, -0.0647, ..., 0.0068, 0.0866, 0.0578], + [ 0.0081, -0.0202, -0.0508, ..., -0.1192, 0.0577, -0.1110], + [-0.0115, -0.0258, -0.0334, ..., 0.0265, -0.0441, 0.0154]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 0.0000e+00, -1.6415e-07, ..., 1.2582e-06, + 2.5854e-06, 3.4906e-06], + [ 1.6531e-08, 0.0000e+00, 6.0871e-06, ..., 1.1154e-05, + -4.7311e-06, 1.2547e-05], + [ 9.0804e-09, 0.0000e+00, 1.4476e-05, ..., 3.4392e-05, + 1.7229e-06, 3.6180e-05], + ..., + [-1.2619e-07, 0.0000e+00, 1.7844e-06, ..., 1.0021e-06, + -1.2241e-05, -2.4647e-05], + [ 1.3039e-08, 0.0000e+00, 6.6729e-07, ..., 1.2498e-06, + -1.1683e-05, 5.0813e-06], + [ 2.0489e-08, 0.0000e+00, 1.2390e-05, ..., 2.2694e-05, + 3.8445e-06, 1.8150e-05]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0244, 0.0168, 0.0273, 0.0053, 0.0215, -0.0123, -0.0087, -0.0193, + -0.0121, 0.0234], device='cuda:0'), grad: tensor([ 8.7470e-06, 1.3113e-05, 6.3479e-05, 3.2634e-05, -1.3673e-04, + 6.7391e-06, 9.9018e-06, -2.1204e-05, -1.8448e-05, 4.1753e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 220.90, cls_loss 0.0055 cls_loss_mapping 0.0090 cls_loss_causal 0.5604 re_mapping 0.0096 re_causal 0.0279 /// teacc 98.82 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0161, 0.0110, -0.0016, ..., 0.0528, -0.0890, -0.0872], + [ 0.0368, -0.0185, 0.0131, ..., -0.0924, 0.0453, -0.0904], + [ 0.0068, -0.0048, 0.1052, ..., -0.1258, -0.0669, -0.1123], + ..., + [ 0.0069, 0.0062, -0.0649, ..., 0.0075, 0.0869, 0.0586], + [ 0.0078, -0.0202, -0.0513, ..., -0.1203, 0.0568, -0.1134], + [-0.0115, -0.0258, -0.0338, ..., 0.0265, -0.0435, 0.0160]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 3.8818e-06, ..., 1.0349e-05, + 6.3218e-06, 3.4738e-06], + [ 1.2224e-07, 0.0000e+00, 1.0375e-06, ..., 5.4296e-07, + -8.0187e-07, 2.2016e-06], + [ 9.7789e-09, 0.0000e+00, 8.8941e-07, ..., 1.7481e-06, + 3.0044e-06, 7.6443e-06], + ..., + [-3.8906e-07, 0.0000e+00, 2.2072e-06, ..., -1.6461e-07, + -4.5151e-06, -2.3827e-05], + [ 1.0012e-08, 0.0000e+00, 4.6715e-06, ..., 8.6054e-07, + 5.4482e-08, 1.3746e-06], + [ 1.5995e-07, 0.0000e+00, 3.4682e-06, ..., 2.0850e-04, + 1.7984e-06, 4.2009e-04]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0242, 0.0177, 0.0267, 0.0053, 0.0213, -0.0126, -0.0085, -0.0191, + -0.0139, 0.0242], device='cuda:0'), grad: tensor([ 3.7193e-05, 2.6133e-06, 2.6926e-05, -4.1847e-03, -3.6764e-04, + 4.1466e-03, -5.4002e-05, -2.7075e-05, 1.9014e-05, 3.9864e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 220.71, cls_loss 0.0064 cls_loss_mapping 0.0096 cls_loss_causal 0.6049 re_mapping 0.0097 re_causal 0.0268 /// teacc 98.77 lr 0.00010000 +Epoch 102, weight, value: tensor([[ 0.0161, 0.0110, -0.0015, ..., 0.0530, -0.0900, -0.0886], + [ 0.0368, -0.0185, 0.0137, ..., -0.0937, 0.0458, -0.0909], + [ 0.0068, -0.0048, 0.1056, ..., -0.1269, -0.0679, -0.1126], + ..., + [ 0.0069, 0.0062, -0.0653, ..., 0.0071, 0.0877, 0.0592], + [ 0.0077, -0.0202, -0.0515, ..., -0.1208, 0.0567, -0.1143], + [-0.0116, -0.0258, -0.0341, ..., 0.0264, -0.0439, 0.0158]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 0.0000e+00, -1.5453e-05, ..., -1.7047e-05, + -4.3586e-06, 4.1239e-06], + [ 1.6997e-08, 0.0000e+00, 4.1351e-06, ..., 5.8748e-06, + 8.7261e-05, 7.1883e-05], + [ 4.6566e-09, 0.0000e+00, 8.9034e-06, ..., 1.7300e-05, + 6.3777e-06, 5.6207e-05], + ..., + [-4.1211e-08, 0.0000e+00, 2.3544e-06, ..., 1.1548e-05, + -1.7560e-04, -7.8797e-05], + [ 1.1642e-09, 0.0000e+00, 5.2415e-06, ..., 1.2830e-05, + 1.3344e-05, 2.3708e-05], + [ 1.2573e-08, 0.0000e+00, 7.9274e-06, ..., 7.6890e-06, + 1.1839e-05, 4.3958e-06]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0244, 0.0182, 0.0263, 0.0056, 0.0214, -0.0133, -0.0091, -0.0186, + -0.0131, 0.0238], device='cuda:0'), grad: tensor([-3.5405e-05, 1.6797e-04, 1.0002e-04, 1.0365e-04, -2.0802e-04, + -3.1199e-06, 1.1146e-05, -2.4581e-04, 4.2170e-05, 6.7234e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 221.18, cls_loss 0.0047 cls_loss_mapping 0.0082 cls_loss_causal 0.5894 re_mapping 0.0091 re_causal 0.0276 /// teacc 98.85 lr 0.00010000 +Epoch 103, weight, value: tensor([[ 0.0161, 0.0110, -0.0028, ..., 0.0520, -0.0912, -0.0904], + [ 0.0368, -0.0185, 0.0134, ..., -0.0941, 0.0462, -0.0910], + [ 0.0067, -0.0048, 0.1067, ..., -0.1286, -0.0684, -0.1127], + ..., + [ 0.0070, 0.0062, -0.0658, ..., 0.0069, 0.0878, 0.0594], + [ 0.0077, -0.0202, -0.0524, ..., -0.1215, 0.0570, -0.1149], + [-0.0116, -0.0258, -0.0324, ..., 0.0274, -0.0443, 0.0153]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -5.6699e-06, ..., -1.4086e-07, + 8.3223e-06, 1.6198e-05], + [ 2.0955e-09, 0.0000e+00, 1.6112e-07, ..., 5.9344e-06, + 1.5469e-06, 1.3597e-05], + [ 1.3970e-09, 0.0000e+00, 1.0496e-06, ..., 4.4256e-06, + 5.3085e-06, 8.1807e-06], + ..., + [-4.6566e-09, 0.0000e+00, 3.2899e-07, ..., 1.2636e-04, + 4.3511e-05, 2.8682e-04], + [ 0.0000e+00, 0.0000e+00, 2.2110e-06, ..., 1.2338e-05, + 7.8902e-06, 2.4214e-05], + [ 6.9849e-10, 0.0000e+00, 9.3412e-07, ..., -4.6158e-04, + -1.5819e-04, -1.0614e-03]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0255, 0.0184, 0.0267, 0.0048, 0.0220, -0.0129, -0.0090, -0.0188, + -0.0130, 0.0238], device='cuda:0'), grad: tensor([ 3.9101e-05, 2.0891e-05, 3.2097e-05, 2.1011e-05, 7.6628e-04, + -2.9802e-05, -2.0063e-04, 2.7680e-04, 5.5969e-05, -9.8228e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 220.75, cls_loss 0.0063 cls_loss_mapping 0.0106 cls_loss_causal 0.6001 re_mapping 0.0093 re_causal 0.0274 /// teacc 98.70 lr 0.00010000 +Epoch 104, weight, value: tensor([[ 0.0161, 0.0110, -0.0025, ..., 0.0525, -0.0922, -0.0904], + [ 0.0367, -0.0185, 0.0136, ..., -0.0951, 0.0448, -0.0918], + [ 0.0067, -0.0048, 0.1068, ..., -0.1305, -0.0691, -0.1139], + ..., + [ 0.0069, 0.0062, -0.0664, ..., 0.0070, 0.0897, 0.0593], + [ 0.0077, -0.0202, -0.0541, ..., -0.1226, 0.0570, -0.1157], + [-0.0116, -0.0258, -0.0326, ..., 0.0286, -0.0441, 0.0167]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 0.0000e+00, -9.2946e-07, ..., 5.5313e-05, + 1.2117e-06, 3.4362e-05], + [ 1.5716e-07, 0.0000e+00, -6.8843e-05, ..., 4.9770e-06, + -1.6034e-04, 7.0147e-06], + [ 1.0477e-08, 0.0000e+00, 4.4316e-05, ..., 2.4885e-06, + 1.4305e-04, 3.1162e-06], + ..., + [-7.3481e-07, 0.0000e+00, 2.1592e-05, ..., 2.3052e-05, + 1.3545e-05, 2.6941e-05], + [ 2.1188e-08, 0.0000e+00, 1.5460e-06, ..., 3.5577e-06, + 1.0310e-06, 4.9137e-06], + [ 5.2247e-07, 0.0000e+00, -5.2191e-06, ..., -2.8044e-05, + 2.1681e-05, 3.0684e-04]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0252, 0.0174, 0.0259, 0.0060, 0.0209, -0.0125, -0.0089, -0.0184, + -0.0138, 0.0247], device='cuda:0'), grad: tensor([ 5.1916e-05, -2.9540e-04, 2.5034e-04, 2.2352e-05, -3.7241e-04, + 2.3142e-05, -3.5241e-06, 8.9347e-05, 1.2413e-05, 2.2185e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 220.72, cls_loss 0.0044 cls_loss_mapping 0.0073 cls_loss_causal 0.5563 re_mapping 0.0088 re_causal 0.0264 /// teacc 98.77 lr 0.00010000 +Epoch 105, weight, value: tensor([[ 0.0161, 0.0110, -0.0023, ..., 0.0522, -0.0938, -0.0911], + [ 0.0367, -0.0185, 0.0134, ..., -0.0958, 0.0448, -0.0923], + [ 0.0067, -0.0048, 0.1075, ..., -0.1314, -0.0696, -0.1142], + ..., + [ 0.0069, 0.0062, -0.0672, ..., 0.0067, 0.0904, 0.0600], + [ 0.0077, -0.0202, -0.0540, ..., -0.1237, 0.0587, -0.1167], + [-0.0115, -0.0258, -0.0330, ..., 0.0274, -0.0447, 0.0154]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3209e-06, ..., -4.4107e-06, + 1.6494e-06, 3.9279e-07], + [ 0.0000e+00, 0.0000e+00, 6.5332e-07, ..., 1.1148e-06, + -2.6003e-06, 1.1884e-06], + [ 0.0000e+00, 0.0000e+00, 1.8133e-06, ..., 1.7546e-06, + 3.0715e-06, 6.7148e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 9.9838e-07, ..., 1.2442e-05, + 7.6368e-06, 2.9013e-05], + [ 0.0000e+00, 0.0000e+00, 1.6928e-05, ..., 1.0664e-06, + 2.0266e-05, 1.4538e-06], + [ 0.0000e+00, 0.0000e+00, 1.6503e-06, ..., -2.1860e-05, + -1.1057e-05, -5.6118e-05]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0257, 0.0174, 0.0258, 0.0065, 0.0224, -0.0132, -0.0096, -0.0181, + -0.0126, 0.0234], device='cuda:0'), grad: tensor([-3.7113e-07, -4.0978e-06, 1.0021e-05, -2.8387e-05, 2.4259e-05, + -8.5533e-06, -5.4479e-05, 3.5405e-05, 7.3373e-05, -4.7147e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 221.19, cls_loss 0.0052 cls_loss_mapping 0.0091 cls_loss_causal 0.5834 re_mapping 0.0091 re_causal 0.0272 /// teacc 98.87 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 0.0161, 0.0110, -0.0014, ..., 0.0529, -0.0943, -0.0916], + [ 0.0364, -0.0185, 0.0123, ..., -0.0970, 0.0462, -0.0903], + [ 0.0067, -0.0048, 0.1091, ..., -0.1339, -0.0692, -0.1147], + ..., + [ 0.0071, 0.0062, -0.0681, ..., 0.0070, 0.0899, 0.0598], + [ 0.0077, -0.0202, -0.0545, ..., -0.1250, 0.0586, -0.1175], + [-0.0116, -0.0258, -0.0335, ..., 0.0265, -0.0459, 0.0141]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.4105e-06, ..., -5.7817e-06, + 1.6317e-06, 1.1008e-06], + [ 0.0000e+00, 0.0000e+00, -2.6569e-03, ..., -5.9046e-07, + -3.3998e-04, 1.4277e-06], + [ 0.0000e+00, 0.0000e+00, 2.5043e-03, ..., 2.3618e-06, + 3.2496e-04, 5.3383e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 2.1055e-05, ..., -3.2224e-06, + -5.0291e-06, 1.6558e-04], + [ 0.0000e+00, 0.0000e+00, 1.1399e-05, ..., 1.2768e-06, + -1.3225e-05, -2.7195e-07], + [ 0.0000e+00, 0.0000e+00, 1.7911e-05, ..., 4.0494e-06, + 1.2405e-05, 7.3127e-06]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0252, 0.0182, 0.0266, 0.0057, 0.0237, -0.0135, -0.0091, -0.0184, + -0.0129, 0.0220], device='cuda:0'), grad: tensor([ 9.4995e-06, -4.4556e-03, 4.2381e-03, -9.1982e-04, 1.0455e-04, + -2.3711e-06, 1.0632e-05, 9.4604e-04, 6.7912e-06, 6.4731e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 220.76, cls_loss 0.0046 cls_loss_mapping 0.0072 cls_loss_causal 0.5668 re_mapping 0.0094 re_causal 0.0273 /// teacc 98.62 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0161, 0.0110, -0.0008, ..., 0.0533, -0.0948, -0.0924], + [ 0.0364, -0.0185, 0.0127, ..., -0.0972, 0.0467, -0.0903], + [ 0.0067, -0.0048, 0.1096, ..., -0.1350, -0.0699, -0.1148], + ..., + [ 0.0071, 0.0062, -0.0686, ..., 0.0069, 0.0897, 0.0600], + [ 0.0077, -0.0202, -0.0556, ..., -0.1258, 0.0592, -0.1174], + [-0.0116, -0.0258, -0.0337, ..., 0.0267, -0.0463, 0.0141]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2457e-05, ..., 4.4680e-04, + 8.3447e-06, 2.4056e-04], + [ 0.0000e+00, 0.0000e+00, 1.1493e-06, ..., 4.0196e-06, + 1.3620e-05, 8.5011e-06], + [ 0.0000e+00, 0.0000e+00, 6.0759e-06, ..., 1.3754e-05, + 2.3507e-06, 7.3835e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4634e-06, ..., 5.0366e-06, + 6.1691e-06, -1.6451e-05], + [ 0.0000e+00, 0.0000e+00, 5.0813e-06, ..., 7.3090e-06, + 1.1939e-04, 3.2820e-06], + [ 0.0000e+00, 0.0000e+00, 1.2796e-06, ..., -5.2166e-04, + 6.4895e-06, -2.8419e-04]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0250, 0.0189, 0.0260, 0.0054, 0.0235, -0.0134, -0.0090, -0.0186, + -0.0127, 0.0220], device='cuda:0'), grad: tensor([ 6.6996e-04, 3.4660e-05, 3.6240e-05, -1.1899e-05, 2.2396e-05, + 1.6177e-04, -4.3869e-04, 1.1124e-05, 2.6894e-04, -7.5626e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 221.07, cls_loss 0.0051 cls_loss_mapping 0.0087 cls_loss_causal 0.5858 re_mapping 0.0092 re_causal 0.0272 /// teacc 98.77 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0161, 0.0110, -0.0003, ..., 0.0539, -0.0946, -0.0931], + [ 0.0363, -0.0185, 0.0128, ..., -0.0985, 0.0469, -0.0907], + [ 0.0067, -0.0048, 0.1104, ..., -0.1377, -0.0694, -0.1142], + ..., + [ 0.0066, 0.0062, -0.0707, ..., 0.0069, 0.0893, 0.0597], + [ 0.0077, -0.0202, -0.0560, ..., -0.1267, 0.0594, -0.1176], + [-0.0116, -0.0258, -0.0340, ..., 0.0271, -0.0466, 0.0146]], + device='cuda:0'), grad: tensor([[ 1.8626e-08, 0.0000e+00, -1.6927e-07, ..., -3.3528e-06, + 2.9569e-07, 3.4273e-07], + [ 3.1013e-07, 0.0000e+00, 4.7982e-06, ..., 4.9081e-07, + -2.7660e-07, 1.8962e-06], + [ 2.1420e-07, 0.0000e+00, -2.0817e-05, ..., -6.5193e-08, + 2.3562e-07, -1.1306e-06], + ..., + [-7.7393e-07, 0.0000e+00, 8.7023e-06, ..., -2.7055e-07, + -2.7381e-06, -3.8017e-06], + [ 2.5379e-08, 0.0000e+00, 3.3341e-06, ..., 2.6869e-07, + -3.2574e-05, 6.9756e-07], + [ 1.6508e-07, 0.0000e+00, 4.0932e-07, ..., 3.3565e-06, + 1.2862e-06, 2.3171e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0245, 0.0191, 0.0261, 0.0057, 0.0232, -0.0123, -0.0097, -0.0194, + -0.0132, 0.0223], device='cuda:0'), grad: tensor([-4.0196e-06, 7.2829e-06, -2.5645e-05, 2.2650e-06, 4.9081e-07, + 4.8995e-05, 1.1928e-05, 6.1691e-06, -5.3257e-05, 5.8562e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 220.63, cls_loss 0.0050 cls_loss_mapping 0.0095 cls_loss_causal 0.5753 re_mapping 0.0089 re_causal 0.0254 /// teacc 98.84 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0160, 0.0110, -0.0002, ..., 0.0536, -0.0951, -0.0937], + [ 0.0361, -0.0185, 0.0126, ..., -0.1002, 0.0462, -0.0921], + [ 0.0065, -0.0048, 0.1109, ..., -0.1390, -0.0700, -0.1147], + ..., + [ 0.0063, 0.0062, -0.0706, ..., 0.0069, 0.0905, 0.0606], + [ 0.0076, -0.0202, -0.0563, ..., -0.1281, 0.0598, -0.1186], + [-0.0118, -0.0258, -0.0345, ..., 0.0267, -0.0469, 0.0142]], + device='cuda:0'), grad: tensor([[ 1.3574e-07, 0.0000e+00, -8.9025e-04, ..., -6.4802e-04, + 2.5146e-07, 3.5483e-07], + [ 9.6392e-08, 0.0000e+00, 9.7901e-06, ..., 9.5069e-06, + 3.1143e-06, 6.6012e-06], + [ 1.8487e-07, 0.0000e+00, 9.3520e-05, ..., 6.8545e-05, + -1.6415e-07, 7.3202e-07], + ..., + [ 1.9954e-07, 0.0000e+00, 5.0068e-06, ..., 5.2564e-06, + 4.5933e-06, 8.0243e-06], + [ 1.7080e-06, 0.0000e+00, 1.7479e-05, ..., 1.7062e-05, + 2.8983e-06, 1.9759e-05], + [ 9.2108e-07, 0.0000e+00, 2.3603e-05, ..., 8.3596e-06, + -2.5555e-06, -1.8954e-05]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0255, 0.0182, 0.0259, 0.0062, 0.0243, -0.0119, -0.0099, -0.0189, + -0.0133, 0.0217], device='cuda:0'), grad: tensor([-2.0618e-03, 3.1829e-05, 2.2268e-04, -6.2943e-05, -1.2293e-05, + 3.4809e-05, 1.7071e-03, 2.1726e-05, 6.6817e-05, 5.0247e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 221.02, cls_loss 0.0039 cls_loss_mapping 0.0079 cls_loss_causal 0.5636 re_mapping 0.0093 re_causal 0.0273 /// teacc 98.86 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 0.0159, 0.0110, 0.0007, ..., 0.0540, -0.0955, -0.0941], + [ 0.0359, -0.0185, 0.0127, ..., -0.1018, 0.0462, -0.0927], + [ 0.0063, -0.0048, 0.1106, ..., -0.1402, -0.0708, -0.1153], + ..., + [ 0.0058, 0.0062, -0.0707, ..., 0.0064, 0.0911, 0.0606], + [ 0.0076, -0.0202, -0.0566, ..., -0.1291, 0.0599, -0.1192], + [-0.0105, -0.0258, -0.0343, ..., 0.0269, -0.0473, 0.0146]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 0.0000e+00, -1.0012e-07, ..., 4.3260e-07, + 1.5479e-06, 1.2200e-06], + [-1.9325e-08, 0.0000e+00, 1.9651e-07, ..., 9.6299e-07, + -1.0002e-04, -1.8384e-06], + [ 1.2107e-08, 0.0000e+00, -5.2005e-06, ..., 3.2852e-07, + 2.3115e-06, 9.2573e-07], + ..., + [-3.2363e-08, 0.0000e+00, 3.8482e-06, ..., 2.0340e-06, + 7.3425e-06, -2.1774e-06], + [ 3.4925e-09, 0.0000e+00, 6.6124e-07, ..., 3.5912e-06, + 3.1173e-05, 7.3202e-06], + [ 1.7695e-08, 0.0000e+00, 1.1851e-07, ..., -2.1040e-04, + -3.9428e-05, -3.5882e-04]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0252, 0.0181, 0.0250, 0.0064, 0.0244, -0.0119, -0.0103, -0.0185, + -0.0134, 0.0220], device='cuda:0'), grad: tensor([ 5.2750e-06, -1.6749e-04, -1.0431e-06, 1.8239e-05, 4.9019e-04, + -1.9407e-04, 2.1625e-04, 1.8284e-05, 6.4313e-05, -4.5013e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 220.51, cls_loss 0.0049 cls_loss_mapping 0.0082 cls_loss_causal 0.5912 re_mapping 0.0087 re_causal 0.0266 /// teacc 98.90 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0159, 0.0110, 0.0011, ..., 0.0541, -0.0960, -0.0949], + [ 0.0358, -0.0185, 0.0120, ..., -0.1039, 0.0460, -0.0932], + [ 0.0063, -0.0048, 0.1112, ..., -0.1423, -0.0707, -0.1162], + ..., + [ 0.0055, 0.0062, -0.0708, ..., 0.0095, 0.0919, 0.0627], + [ 0.0076, -0.0202, -0.0571, ..., -0.1303, 0.0600, -0.1195], + [-0.0099, -0.0258, -0.0345, ..., 0.0265, -0.0484, 0.0142]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.9811e-05, ..., -1.9717e-04, + 2.3842e-06, 4.9993e-06], + [ 0.0000e+00, 0.0000e+00, 5.0664e-07, ..., -4.2543e-06, + -1.7911e-05, -2.0191e-05], + [ 0.0000e+00, 0.0000e+00, -7.8604e-07, ..., 4.8764e-06, + 1.3784e-05, 8.5682e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6641e-06, ..., 9.9093e-06, + -3.4869e-05, -1.2837e-05], + [ 0.0000e+00, 0.0000e+00, 5.6028e-06, ..., 1.3664e-05, + -7.3016e-07, 9.7696e-07], + [ 0.0000e+00, 0.0000e+00, 3.5483e-06, ..., -9.5546e-05, + 2.2709e-05, -1.0562e-04]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0252, 0.0174, 0.0249, 0.0064, 0.0232, -0.0120, -0.0088, -0.0166, + -0.0137, 0.0211], device='cuda:0'), grad: tensor([-3.8505e-04, -6.6876e-05, 4.7415e-05, 1.9953e-05, 1.8561e-04, + 1.1837e-04, 2.3174e-04, -6.8486e-05, 2.5287e-05, -1.0812e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 221.29, cls_loss 0.0050 cls_loss_mapping 0.0112 cls_loss_causal 0.6104 re_mapping 0.0094 re_causal 0.0271 /// teacc 98.89 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0159, 0.0110, 0.0008, ..., 0.0538, -0.0963, -0.0957], + [ 0.0359, -0.0185, 0.0119, ..., -0.1051, 0.0459, -0.0940], + [ 0.0063, -0.0048, 0.1123, ..., -0.1431, -0.0711, -0.1166], + ..., + [ 0.0055, 0.0062, -0.0715, ..., 0.0094, 0.0925, 0.0633], + [ 0.0075, -0.0202, -0.0580, ..., -0.1307, 0.0605, -0.1180], + [-0.0099, -0.0258, -0.0343, ..., 0.0277, -0.0493, 0.0146]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4137e-06, ..., 1.5087e-06, + 2.0459e-05, 6.0759e-06], + [ 0.0000e+00, 0.0000e+00, 6.6608e-06, ..., 1.8971e-06, + 3.7029e-06, 1.0006e-05], + [ 0.0000e+00, 0.0000e+00, -6.8963e-05, ..., 9.1456e-07, + 3.0063e-06, -2.5943e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 4.0382e-05, ..., -9.6932e-06, + -1.0863e-05, 3.4049e-06], + [ 0.0000e+00, 0.0000e+00, 2.3488e-06, ..., 1.3104e-06, + -4.4174e-03, 4.8727e-06], + [ 0.0000e+00, 0.0000e+00, 5.6624e-06, ..., 1.3433e-05, + 1.0751e-05, 2.9504e-05]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0256, 0.0170, 0.0253, 0.0060, 0.0224, -0.0119, -0.0084, -0.0161, + -0.0131, 0.0213], device='cuda:0'), grad: tensor([ 5.4210e-05, 2.5630e-05, -1.0312e-04, 1.5616e-05, -2.6926e-05, + 2.4128e-03, 7.6675e-03, 2.9162e-05, -1.0132e-02, 6.1989e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 220.87, cls_loss 0.0040 cls_loss_mapping 0.0080 cls_loss_causal 0.5863 re_mapping 0.0085 re_causal 0.0262 /// teacc 98.81 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 1.5887e-02, 1.0998e-02, 2.7725e-05, ..., 5.3349e-02, + -9.6547e-02, -9.6051e-02], + [ 3.5917e-02, -1.8468e-02, 1.1904e-02, ..., -1.0585e-01, + 4.6064e-02, -9.4180e-02], + [ 6.2594e-03, -4.7728e-03, 1.1299e-01, ..., -1.4430e-01, + -7.1335e-02, -1.1652e-01], + ..., + [ 5.4894e-03, 6.2061e-03, -7.2306e-02, ..., 9.2758e-03, + 9.2541e-02, 6.3443e-02], + [ 7.5326e-03, -2.0220e-02, -5.9033e-02, ..., -1.3183e-01, + 6.1035e-02, -1.1809e-01], + [-9.9149e-03, -2.5834e-02, -3.3555e-02, ..., 2.8040e-02, + -4.9659e-02, 1.4163e-02]], device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, 4.5705e-07, ..., -1.7192e-06, + 5.3225e-07, 2.4168e-07], + [-1.6298e-09, 0.0000e+00, 8.2469e-07, ..., 4.4098e-07, + -4.4368e-06, 1.6727e-06], + [ 1.6298e-08, 0.0000e+00, -9.0972e-06, ..., 1.8114e-07, + 1.4128e-06, 4.4773e-07], + ..., + [-6.7055e-08, 0.0000e+00, 1.7695e-06, ..., 2.7791e-06, + -2.1774e-06, 1.2647e-06], + [ 8.3819e-09, 0.0000e+00, 4.0680e-06, ..., 3.8906e-07, + 1.5981e-06, 6.7707e-07], + [ 1.3737e-08, 0.0000e+00, 9.9838e-07, ..., 3.3770e-06, + 1.2144e-06, 6.5118e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0260, 0.0170, 0.0253, 0.0056, 0.0230, -0.0114, -0.0092, -0.0159, + -0.0131, 0.0211], device='cuda:0'), grad: tensor([ 2.7865e-06, -1.0513e-05, -2.0936e-06, -9.2316e-04, -1.1951e-05, + 8.9312e-04, 2.1402e-06, 4.6529e-06, 2.5466e-05, 1.9580e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 221.03, cls_loss 0.0037 cls_loss_mapping 0.0065 cls_loss_causal 0.5854 re_mapping 0.0086 re_causal 0.0260 /// teacc 98.89 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0159, 0.0110, 0.0003, ..., 0.0535, -0.0971, -0.0965], + [ 0.0359, -0.0185, 0.0118, ..., -0.1061, 0.0467, -0.0932], + [ 0.0062, -0.0048, 0.1135, ..., -0.1449, -0.0720, -0.1168], + ..., + [ 0.0055, 0.0062, -0.0725, ..., 0.0088, 0.0924, 0.0624], + [ 0.0075, -0.0202, -0.0596, ..., -0.1332, 0.0612, -0.1193], + [-0.0098, -0.0258, -0.0338, ..., 0.0282, -0.0499, 0.0146]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.5858e-06, ..., 2.3127e-05, + 2.6464e-05, 1.4259e-06], + [ 6.9849e-10, 0.0000e+00, 3.4343e-07, ..., 9.5647e-07, + 1.6373e-06, 5.8226e-06], + [ 0.0000e+00, 0.0000e+00, 2.5630e-06, ..., 2.7716e-06, + 3.1050e-06, 1.1057e-05], + ..., + [-3.2596e-09, 0.0000e+00, -1.6131e-06, ..., 4.2990e-06, + -5.2005e-06, -7.7486e-06], + [ 2.3283e-10, 0.0000e+00, 1.4221e-06, ..., 3.3453e-06, + 9.9558e-07, 2.2687e-06], + [ 1.6298e-09, 0.0000e+00, 3.2317e-06, ..., 5.5462e-05, + 6.2119e-07, 1.0687e-04]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0260, 0.0176, 0.0252, 0.0055, 0.0231, -0.0107, -0.0091, -0.0167, + -0.0133, 0.0212], device='cuda:0'), grad: tensor([ 1.2445e-04, 8.3521e-06, 2.2635e-05, 4.9323e-06, -1.0276e-04, + 4.4852e-06, -1.6928e-04, -1.4581e-05, 1.0751e-05, 1.1098e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 220.80, cls_loss 0.0039 cls_loss_mapping 0.0075 cls_loss_causal 0.5811 re_mapping 0.0086 re_causal 0.0258 /// teacc 98.92 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 0.0158, 0.0110, 0.0005, ..., 0.0536, -0.0977, -0.0967], + [ 0.0358, -0.0185, 0.0118, ..., -0.1070, 0.0467, -0.0939], + [ 0.0060, -0.0048, 0.1139, ..., -0.1462, -0.0726, -0.1176], + ..., + [ 0.0054, 0.0062, -0.0733, ..., 0.0086, 0.0927, 0.0622], + [ 0.0074, -0.0202, -0.0603, ..., -0.1333, 0.0617, -0.1193], + [-0.0097, -0.0258, -0.0339, ..., 0.0281, -0.0502, 0.0149]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 0.0000e+00, -5.4389e-07, ..., -4.7777e-07, + 3.1735e-07, 8.3633e-07], + [-2.0815e-07, 0.0000e+00, 3.3528e-08, ..., 8.1258e-08, + -1.0110e-05, -4.7795e-06], + [ 2.1886e-08, 0.0000e+00, -9.2899e-08, ..., 1.5227e-07, + 2.1141e-06, 1.2815e-06], + ..., + [ 4.7265e-08, 0.0000e+00, 1.4901e-07, ..., 3.5693e-07, + 2.8722e-06, 2.2184e-06], + [ 1.0245e-07, 0.0000e+00, 2.2259e-07, ..., 6.9253e-06, + 9.9093e-07, 2.2098e-05], + [-3.9581e-08, 0.0000e+00, 1.0710e-07, ..., -4.7684e-05, + 1.9232e-07, -1.5259e-04]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0260, 0.0172, 0.0249, 0.0056, 0.0232, -0.0109, -0.0087, -0.0169, + -0.0133, 0.0215], device='cuda:0'), grad: tensor([ 6.0443e-07, -3.4362e-05, 7.5623e-06, 1.6582e-04, 1.1347e-05, + -1.4687e-06, 2.2519e-06, 1.1913e-05, 3.2008e-05, -1.9562e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 220.85, cls_loss 0.0036 cls_loss_mapping 0.0079 cls_loss_causal 0.6048 re_mapping 0.0085 re_causal 0.0265 /// teacc 98.88 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0158, 0.0110, 0.0012, ..., 0.0540, -0.0974, -0.0971], + [ 0.0357, -0.0185, 0.0118, ..., -0.1081, 0.0468, -0.0938], + [ 0.0060, -0.0048, 0.1143, ..., -0.1469, -0.0729, -0.1179], + ..., + [ 0.0054, 0.0062, -0.0742, ..., 0.0080, 0.0925, 0.0624], + [ 0.0072, -0.0202, -0.0609, ..., -0.1338, 0.0625, -0.1197], + [-0.0097, -0.0258, -0.0341, ..., 0.0280, -0.0507, 0.0149]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 0.0000e+00, -1.2843e-06, ..., -5.6699e-06, + 1.7006e-06, 3.9637e-06], + [ 1.4668e-08, 0.0000e+00, 1.3262e-05, ..., 9.1828e-07, + 7.6033e-06, 1.0796e-05], + [ 4.4238e-09, 0.0000e+00, -9.8515e-04, ..., 1.2238e-06, + -1.9705e-04, -5.0306e-04], + ..., + [-5.9837e-08, 0.0000e+00, 9.4557e-04, ..., 3.5223e-06, + 1.8251e-04, 4.8184e-04], + [ 3.2596e-09, 0.0000e+00, 1.1511e-05, ..., 1.2919e-05, + 4.2208e-06, 1.3962e-05], + [ 9.5461e-09, 0.0000e+00, 1.5227e-06, ..., 1.6630e-05, + -8.1658e-06, 5.9336e-05]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0256, 0.0174, 0.0247, 0.0049, 0.0233, -0.0108, -0.0084, -0.0172, + -0.0129, 0.0214], device='cuda:0'), grad: tensor([-4.8764e-06, 3.1054e-05, -1.4620e-03, 1.6475e-04, -4.7088e-05, + -2.8920e-04, 1.4603e-05, 1.3914e-03, 1.7095e-04, 3.2336e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 115---------------------------------------------------- +epoch 115, time 221.94, cls_loss 0.0046 cls_loss_mapping 0.0102 cls_loss_causal 0.5448 re_mapping 0.0089 re_causal 0.0261 /// teacc 98.98 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 0.0158, 0.0110, 0.0018, ..., 0.0542, -0.0976, -0.0978], + [ 0.0356, -0.0185, 0.0103, ..., -0.1088, 0.0459, -0.0945], + [ 0.0065, -0.0048, 0.1155, ..., -0.1482, -0.0728, -0.1180], + ..., + [ 0.0047, 0.0062, -0.0745, ..., 0.0077, 0.0934, 0.0626], + [ 0.0070, -0.0202, -0.0635, ..., -0.1348, 0.0616, -0.1196], + [-0.0097, -0.0258, -0.0336, ..., 0.0284, -0.0517, 0.0148]], + device='cuda:0'), grad: tensor([[ 4.0159e-06, 0.0000e+00, -2.9469e-04, ..., -5.1975e-04, + 1.3269e-05, -7.3612e-05], + [ 1.1826e-04, 0.0000e+00, 2.1271e-06, ..., 3.7495e-06, + 3.8600e-04, 4.0561e-05], + [ 1.2880e-06, 0.0000e+00, 6.9141e-06, ..., 1.6376e-05, + 5.0813e-06, 6.1952e-06], + ..., + [-1.5246e-06, 0.0000e+00, 2.8987e-07, ..., -9.1456e-07, + -1.6928e-05, -6.8784e-05], + [-1.4758e-04, 0.0000e+00, 3.8706e-06, ..., 6.9812e-06, + -4.8089e-04, -3.6597e-05], + [ 1.2442e-06, 0.0000e+00, 5.5581e-05, ..., 1.0800e-04, + 4.7795e-06, 2.8193e-05]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0253, 0.0162, 0.0252, 0.0054, 0.0233, -0.0106, -0.0075, -0.0171, + -0.0141, 0.0214], device='cuda:0'), grad: tensor([-8.1062e-04, 8.3780e-04, 3.7462e-05, 1.3125e-04, 8.7917e-06, + 7.6413e-05, 6.4039e-04, -9.3043e-05, -1.0204e-03, 1.9217e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 221.02, cls_loss 0.0033 cls_loss_mapping 0.0058 cls_loss_causal 0.5837 re_mapping 0.0084 re_causal 0.0248 /// teacc 98.81 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 0.0157, 0.0110, 0.0023, ..., 0.0545, -0.0981, -0.0985], + [ 0.0353, -0.0185, 0.0113, ..., -0.1099, 0.0457, -0.0951], + [ 0.0068, -0.0048, 0.1150, ..., -0.1498, -0.0735, -0.1188], + ..., + [ 0.0043, 0.0062, -0.0747, ..., 0.0076, 0.0938, 0.0626], + [ 0.0071, -0.0202, -0.0648, ..., -0.1353, 0.0617, -0.1193], + [-0.0098, -0.0258, -0.0337, ..., 0.0282, -0.0522, 0.0146]], + device='cuda:0'), grad: tensor([[ 1.9325e-08, 0.0000e+00, 3.2294e-07, ..., 3.9339e-06, + 8.8289e-07, 9.8422e-06], + [ 2.3027e-07, 0.0000e+00, 1.0028e-05, ..., 7.1339e-06, + 1.9073e-05, 2.8178e-05], + [ 7.4506e-08, 0.0000e+00, -3.8296e-06, ..., 2.0713e-06, + 2.6956e-05, 1.3486e-05], + ..., + [-8.3167e-07, 0.0000e+00, -7.8082e-06, ..., 8.4221e-05, + -1.5229e-05, 1.7715e-04], + [ 4.3772e-08, 0.0000e+00, -1.2023e-06, ..., 2.9862e-05, + -1.8448e-05, 6.3837e-05], + [ 1.5926e-07, 0.0000e+00, 7.5437e-07, ..., -1.9777e-04, + -4.8757e-05, -4.7064e-04]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0251, 0.0162, 0.0244, 0.0055, 0.0239, -0.0098, -0.0073, -0.0172, + -0.0144, 0.0211], device='cuda:0'), grad: tensor([ 1.5289e-05, 7.6592e-05, 6.6757e-05, 3.8981e-05, 1.6594e-04, + -3.0696e-05, 4.4405e-06, 8.2076e-05, 5.7578e-05, -4.7684e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 221.09, cls_loss 0.0041 cls_loss_mapping 0.0077 cls_loss_causal 0.5565 re_mapping 0.0083 re_causal 0.0246 /// teacc 98.80 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0156, 0.0110, 0.0022, ..., 0.0547, -0.0987, -0.0994], + [ 0.0352, -0.0185, 0.0105, ..., -0.1122, 0.0458, -0.0958], + [ 0.0069, -0.0048, 0.1158, ..., -0.1509, -0.0733, -0.1189], + ..., + [ 0.0042, 0.0062, -0.0749, ..., 0.0069, 0.0941, 0.0623], + [ 0.0069, -0.0202, -0.0656, ..., -0.1364, 0.0615, -0.1200], + [-0.0092, -0.0258, -0.0347, ..., 0.0287, -0.0520, 0.0153]], + device='cuda:0'), grad: tensor([[ 1.4994e-07, 0.0000e+00, -2.9400e-05, ..., 1.3430e-06, + 1.6289e-06, 7.1106e-07], + [-1.3612e-05, 0.0000e+00, -1.0002e-04, ..., 4.2878e-06, + -6.6102e-05, 1.0788e-05], + [ 1.2919e-05, 0.0000e+00, 1.1718e-04, ..., 1.2163e-06, + 6.9618e-05, 4.1649e-06], + ..., + [-1.0245e-07, 0.0000e+00, 1.7621e-06, ..., 8.6799e-06, + -1.3433e-05, 1.6004e-05], + [ 7.4971e-08, 0.0000e+00, 2.3153e-06, ..., 1.4883e-06, + 2.7642e-06, 3.7551e-06], + [ 1.1036e-07, 0.0000e+00, 3.1032e-06, ..., -8.9258e-06, + -1.9353e-06, -1.9073e-05]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0258, 0.0156, 0.0249, 0.0057, 0.0240, -0.0095, -0.0067, -0.0172, + -0.0151, 0.0214], device='cuda:0'), grad: tensor([-4.6372e-05, -3.0398e-04, 3.5405e-04, -2.9683e-05, 9.9912e-06, + 2.1294e-05, -1.1489e-05, 8.3297e-06, 1.4931e-05, -1.6958e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 220.92, cls_loss 0.0034 cls_loss_mapping 0.0078 cls_loss_causal 0.5396 re_mapping 0.0084 re_causal 0.0248 /// teacc 98.92 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0153, 0.0110, 0.0030, ..., 0.0552, -0.0983, -0.0999], + [ 0.0350, -0.0185, 0.0103, ..., -0.1128, 0.0458, -0.0962], + [ 0.0071, -0.0048, 0.1166, ..., -0.1523, -0.0739, -0.1192], + ..., + [ 0.0046, 0.0062, -0.0749, ..., 0.0068, 0.0948, 0.0626], + [ 0.0068, -0.0202, -0.0662, ..., -0.1368, 0.0618, -0.1199], + [-0.0093, -0.0258, -0.0351, ..., 0.0286, -0.0525, 0.0153]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -5.9903e-06, ..., -3.6415e-06, + 3.8603e-07, 1.3132e-07], + [-2.6543e-08, 0.0000e+00, 5.4296e-07, ..., 2.4214e-07, + -2.7753e-07, 1.2526e-07], + [-4.0978e-08, 0.0000e+00, 1.0896e-06, ..., 9.4017e-07, + 2.9821e-06, 1.3784e-07], + ..., + [ 1.3970e-08, 0.0000e+00, 7.8091e-07, ..., 4.4936e-07, + 2.0023e-08, 6.6776e-07], + [ 1.4901e-08, 0.0000e+00, 1.3337e-06, ..., 1.3011e-06, + -6.0573e-06, -6.8406e-07], + [ 9.3132e-09, 0.0000e+00, 1.4734e-06, ..., 7.3574e-08, + 1.8952e-06, -2.0023e-06]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0248, 0.0155, 0.0250, 0.0048, 0.0238, -0.0093, -0.0074, -0.0167, + -0.0150, 0.0212], device='cuda:0'), grad: tensor([-1.2755e-05, 4.5728e-07, 1.1429e-05, 6.6822e-07, 2.6003e-06, + 1.9893e-06, -1.9185e-07, 2.0899e-06, -1.2770e-05, 6.4410e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 221.07, cls_loss 0.0036 cls_loss_mapping 0.0070 cls_loss_causal 0.5399 re_mapping 0.0083 re_causal 0.0247 /// teacc 98.77 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0151, 0.0110, 0.0035, ..., 0.0551, -0.0987, -0.1008], + [ 0.0347, -0.0185, 0.0098, ..., -0.1131, 0.0458, -0.0964], + [ 0.0069, -0.0048, 0.1174, ..., -0.1540, -0.0744, -0.1206], + ..., + [ 0.0046, 0.0062, -0.0755, ..., 0.0066, 0.0951, 0.0625], + [ 0.0073, -0.0202, -0.0667, ..., -0.1373, 0.0617, -0.1204], + [-0.0094, -0.0258, -0.0352, ..., 0.0289, -0.0528, 0.0158]], + device='cuda:0'), grad: tensor([[ 6.1467e-08, 0.0000e+00, -2.3283e-07, ..., -2.4028e-07, + 1.9511e-07, 1.9791e-07], + [-2.9523e-06, 0.0000e+00, 1.4203e-07, ..., 1.7993e-06, + -6.7316e-06, 3.4459e-06], + [ 1.7183e-07, 0.0000e+00, -4.7544e-07, ..., 9.5461e-08, + 6.9616e-07, 4.7823e-07], + ..., + [ 1.4771e-06, 0.0000e+00, 5.9092e-07, ..., 1.1194e-06, + -1.0757e-07, -1.5229e-05], + [ 2.9709e-07, 0.0000e+00, 2.6496e-07, ..., 1.8440e-07, + 8.8336e-07, 5.1875e-07], + [ 2.1188e-07, 0.0000e+00, 2.1048e-07, ..., -3.0119e-06, + 1.4994e-07, -8.1882e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0250, 0.0154, 0.0253, 0.0039, 0.0239, -0.0091, -0.0069, -0.0168, + -0.0154, 0.0215], device='cuda:0'), grad: tensor([ 1.6736e-06, -1.3627e-05, 2.6524e-06, 3.6089e-07, 1.9938e-05, + -4.1872e-05, 7.0892e-06, -4.6343e-06, 3.3230e-05, -4.8131e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 220.74, cls_loss 0.0037 cls_loss_mapping 0.0079 cls_loss_causal 0.5786 re_mapping 0.0086 re_causal 0.0253 /// teacc 98.85 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0150, 0.0110, 0.0040, ..., 0.0554, -0.0988, -0.1012], + [ 0.0349, -0.0185, 0.0096, ..., -0.1135, 0.0462, -0.0967], + [ 0.0068, -0.0048, 0.1179, ..., -0.1545, -0.0747, -0.1213], + ..., + [ 0.0047, 0.0062, -0.0764, ..., 0.0065, 0.0954, 0.0627], + [ 0.0072, -0.0202, -0.0671, ..., -0.1384, 0.0616, -0.1220], + [-0.0095, -0.0258, -0.0353, ..., 0.0289, -0.0532, 0.0163]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -1.6475e-06, ..., -1.7947e-06, + 5.1335e-06, 1.9372e-07], + [-5.7276e-08, 0.0000e+00, 2.9337e-07, ..., 2.7008e-07, + -4.5031e-05, -3.3081e-06], + [ 1.5832e-08, 0.0000e+00, -1.0878e-06, ..., 3.3993e-07, + 5.0887e-06, 3.0920e-07], + ..., + [ 7.9162e-09, 0.0000e+00, 6.3144e-07, ..., 1.3150e-06, + 4.9233e-05, 6.0573e-06], + [ 8.8476e-09, 0.0000e+00, 4.1444e-07, ..., 5.3598e-07, + 4.6313e-05, 8.6427e-07], + [ 6.0536e-09, 0.0000e+00, 3.2410e-07, ..., -8.9360e-07, + 4.8243e-06, -5.0180e-06]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0247, 0.0155, 0.0252, 0.0041, 0.0237, -0.0096, -0.0065, -0.0169, + -0.0160, 0.0219], device='cuda:0'), grad: tensor([ 1.1340e-05, -1.2386e-04, 1.3873e-05, 1.0058e-05, 1.3389e-05, + -2.3289e-03, 2.1172e-03, 1.4102e-04, 1.3542e-04, 1.1340e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 220.87, cls_loss 0.0029 cls_loss_mapping 0.0070 cls_loss_causal 0.5874 re_mapping 0.0082 re_causal 0.0259 /// teacc 98.90 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0149, 0.0110, 0.0045, ..., 0.0557, -0.0993, -0.1018], + [ 0.0350, -0.0185, 0.0093, ..., -0.1143, 0.0462, -0.0969], + [ 0.0066, -0.0048, 0.1188, ..., -0.1550, -0.0748, -0.1194], + ..., + [ 0.0048, 0.0062, -0.0783, ..., 0.0063, 0.0954, 0.0624], + [ 0.0071, -0.0202, -0.0673, ..., -0.1389, 0.0619, -0.1222], + [-0.0096, -0.0258, -0.0359, ..., 0.0287, -0.0537, 0.0161]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, -2.9318e-06, ..., -3.7774e-06, + 3.0221e-07, 4.2142e-07], + [ 3.5390e-08, 0.0000e+00, 1.4901e-07, ..., 1.8999e-07, + -8.8066e-06, -3.8072e-06], + [ 2.2817e-08, 0.0000e+00, 5.2620e-07, ..., 5.4529e-07, + 1.0328e-06, 1.4901e-06], + ..., + [-1.1735e-07, 0.0000e+00, 2.2026e-07, ..., 4.2748e-07, + 1.4994e-06, 1.4985e-06], + [ 5.1223e-09, 0.0000e+00, 7.1665e-07, ..., 2.5239e-07, + -7.2643e-08, 5.5972e-07], + [ 3.0268e-08, 0.0000e+00, 8.3866e-07, ..., 9.7789e-07, + 1.3206e-06, 1.0300e-06]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0244, 0.0154, 0.0258, 0.0042, 0.0239, -0.0094, -0.0063, -0.0173, + -0.0159, 0.0213], device='cuda:0'), grad: tensor([-5.0403e-06, -2.5064e-05, 6.1058e-06, -3.4404e-04, 6.9290e-06, + 3.3498e-04, 6.4671e-06, 6.8694e-06, 4.7907e-06, 7.1600e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 220.33, cls_loss 0.0038 cls_loss_mapping 0.0068 cls_loss_causal 0.6109 re_mapping 0.0083 re_causal 0.0258 /// teacc 98.95 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0148, 0.0110, 0.0037, ..., 0.0560, -0.0999, -0.1024], + [ 0.0348, -0.0185, 0.0090, ..., -0.1146, 0.0462, -0.0971], + [ 0.0060, -0.0048, 0.1194, ..., -0.1565, -0.0760, -0.1197], + ..., + [ 0.0045, 0.0062, -0.0790, ..., 0.0061, 0.0958, 0.0624], + [ 0.0073, -0.0202, -0.0680, ..., -0.1415, 0.0625, -0.1234], + [-0.0098, -0.0258, -0.0363, ..., 0.0289, -0.0540, 0.0162]], + device='cuda:0'), grad: tensor([[ 2.0862e-07, 0.0000e+00, -2.7254e-05, ..., -2.0877e-05, + -4.5374e-06, 1.5311e-06], + [ 4.8475e-07, 0.0000e+00, 3.5539e-06, ..., 9.0748e-06, + 1.2793e-05, 4.3124e-05], + [ 4.2515e-07, 0.0000e+00, 8.7470e-06, ..., 8.0019e-06, + 4.2357e-06, 5.0776e-06], + ..., + [-9.1612e-05, 0.0000e+00, 9.7603e-07, ..., 4.9263e-05, + -8.9854e-06, 1.5593e-04], + [ 5.3011e-06, 0.0000e+00, 1.0774e-05, ..., 7.9647e-06, + 7.8753e-06, 1.5825e-05], + [ 1.4845e-06, 0.0000e+00, 6.8173e-06, ..., 8.3521e-06, + 1.3083e-05, 3.8356e-05]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0251, 0.0152, 0.0255, 0.0049, 0.0240, -0.0090, -0.0064, -0.0176, + -0.0159, 0.0216], device='cuda:0'), grad: tensor([-4.6879e-05, 5.7161e-05, 3.2663e-05, 1.4019e-04, -4.6134e-04, + 1.1230e-04, 1.4044e-05, -2.7165e-05, 7.4685e-05, 1.0419e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 220.76, cls_loss 0.0039 cls_loss_mapping 0.0057 cls_loss_causal 0.5620 re_mapping 0.0084 re_causal 0.0250 /// teacc 98.93 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0148, 0.0110, 0.0042, ..., 0.0559, -0.1004, -0.1037], + [ 0.0347, -0.0185, 0.0098, ..., -0.1156, 0.0463, -0.0968], + [ 0.0056, -0.0048, 0.1195, ..., -0.1573, -0.0777, -0.1207], + ..., + [ 0.0047, 0.0062, -0.0803, ..., 0.0058, 0.0963, 0.0623], + [ 0.0072, -0.0202, -0.0684, ..., -0.1422, 0.0629, -0.1231], + [-0.0099, -0.0258, -0.0358, ..., 0.0294, -0.0545, 0.0164]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, 0.0000e+00, 5.0524e-07, ..., -1.2312e-06, + 6.2818e-07, 6.4354e-07], + [-1.5926e-07, 0.0000e+00, 3.1888e-05, ..., 2.9616e-07, + 1.8422e-06, 2.1216e-06], + [ 4.1444e-08, 0.0000e+00, -1.5199e-04, ..., 2.9802e-07, + -1.2212e-05, 8.1630e-07], + ..., + [-1.2154e-07, 0.0000e+00, 1.8217e-06, ..., 2.5243e-05, + 5.4017e-08, 8.6129e-05], + [ 1.1176e-07, 0.0000e+00, 8.9686e-07, ..., 2.7567e-07, + 1.3597e-06, 1.5581e-06], + [ 1.9558e-08, 0.0000e+00, 8.6101e-07, ..., -3.0234e-05, + -3.3453e-06, -1.1027e-04]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0249, 0.0170, 0.0231, 0.0042, 0.0239, -0.0093, -0.0060, -0.0174, + -0.0154, 0.0217], device='cuda:0'), grad: tensor([ 2.7977e-06, 5.8174e-05, -2.7227e-04, 7.7263e-06, 4.3362e-05, + 6.5006e-06, 1.7571e-04, 9.1791e-05, 3.9265e-06, -1.1748e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 220.84, cls_loss 0.0042 cls_loss_mapping 0.0073 cls_loss_causal 0.5559 re_mapping 0.0082 re_causal 0.0234 /// teacc 98.84 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0147, 0.0110, 0.0040, ..., 0.0562, -0.1007, -0.1041], + [ 0.0346, -0.0185, 0.0074, ..., -0.1168, 0.0447, -0.0974], + [ 0.0043, -0.0048, 0.1215, ..., -0.1590, -0.0775, -0.1215], + ..., + [ 0.0052, 0.0062, -0.0797, ..., 0.0048, 0.0979, 0.0620], + [ 0.0070, -0.0202, -0.0690, ..., -0.1430, 0.0639, -0.1239], + [-0.0099, -0.0258, -0.0361, ..., 0.0297, -0.0543, 0.0164]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, -3.2857e-06, ..., -7.1079e-06, + 2.8647e-06, 3.2522e-06], + [ 8.3819e-09, 0.0000e+00, 1.1008e-06, ..., 9.4762e-07, + 1.3020e-06, 3.1739e-06], + [ 4.1910e-09, 0.0000e+00, -5.4985e-06, ..., 1.6484e-06, + 7.0175e-07, 8.1956e-07], + ..., + [-3.5390e-08, 0.0000e+00, 3.9525e-06, ..., 5.4464e-06, + -3.1199e-06, 5.3532e-06], + [ 2.3749e-08, 0.0000e+00, 2.1696e-05, ..., 2.8610e-06, + 1.2256e-06, 2.2724e-06], + [ 1.7229e-08, 0.0000e+00, 2.5220e-06, ..., -8.1956e-08, + 4.2561e-07, -2.9132e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0247, 0.0145, 0.0242, 0.0040, 0.0243, -0.0097, -0.0066, -0.0164, + -0.0143, 0.0216], device='cuda:0'), grad: tensor([ 3.4086e-06, 6.2808e-06, -1.5926e-06, -5.2512e-05, -9.1642e-06, + -2.2855e-06, -2.0012e-05, 1.0297e-05, 5.5522e-05, 9.9316e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 220.92, cls_loss 0.0039 cls_loss_mapping 0.0064 cls_loss_causal 0.5261 re_mapping 0.0080 re_causal 0.0231 /// teacc 98.89 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0147, 0.0110, 0.0041, ..., 0.0563, -0.1016, -0.1050], + [ 0.0347, -0.0185, 0.0084, ..., -0.1188, 0.0444, -0.0980], + [ 0.0042, -0.0048, 0.1210, ..., -0.1600, -0.0784, -0.1217], + ..., + [ 0.0045, 0.0062, -0.0799, ..., 0.0041, 0.0990, 0.0617], + [ 0.0069, -0.0202, -0.0692, ..., -0.1441, 0.0641, -0.1249], + [-0.0090, -0.0258, -0.0367, ..., 0.0308, -0.0539, 0.0179]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 1.1653e-05, ..., 1.7239e-06, + 1.2524e-05, 3.4608e-06], + [ 4.6566e-10, 0.0000e+00, 9.5321e-07, ..., 3.1991e-07, + 2.2398e-07, 6.5705e-07], + [ 4.6566e-10, 0.0000e+00, -9.8720e-07, ..., 3.2037e-07, + 1.3607e-06, 1.1679e-06], + ..., + [ 4.6566e-10, 0.0000e+00, 2.7427e-07, ..., 2.7306e-06, + -2.2240e-06, 1.9008e-06], + [ 1.3970e-09, 0.0000e+00, 2.8729e-05, ..., 8.6799e-07, + 1.1213e-05, -1.1005e-05], + [ 7.4506e-09, 0.0000e+00, 1.1232e-06, ..., -1.5691e-05, + 1.5959e-05, -1.0200e-05]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0248, 0.0147, 0.0231, 0.0047, 0.0232, -0.0104, -0.0067, -0.0158, + -0.0144, 0.0229], device='cuda:0'), grad: tensor([ 8.5294e-05, 3.8594e-06, 4.1649e-06, 7.6711e-05, -1.2740e-06, + -4.1455e-05, -2.9325e-04, 2.4978e-06, 1.5438e-04, 9.0003e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 220.78, cls_loss 0.0040 cls_loss_mapping 0.0058 cls_loss_causal 0.5815 re_mapping 0.0079 re_causal 0.0245 /// teacc 98.88 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0146, 0.0110, 0.0046, ..., 0.0572, -0.1021, -0.1050], + [ 0.0347, -0.0185, 0.0081, ..., -0.1177, 0.0451, -0.0954], + [ 0.0042, -0.0048, 0.1220, ..., -0.1609, -0.0787, -0.1222], + ..., + [ 0.0043, 0.0062, -0.0805, ..., 0.0037, 0.0988, 0.0605], + [ 0.0069, -0.0202, -0.0698, ..., -0.1451, 0.0641, -0.1254], + [-0.0089, -0.0258, -0.0374, ..., 0.0293, -0.0551, 0.0165]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.6531e-07, ..., 1.2815e-06, + 4.1816e-07, 2.2184e-06], + [ 0.0000e+00, 0.0000e+00, 1.6345e-07, ..., 2.3432e-06, + -9.3207e-06, 3.3341e-06], + [ 0.0000e+00, 0.0000e+00, -9.3728e-06, ..., 1.7881e-06, + 1.3933e-06, 1.6615e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 9.2983e-06, ..., 5.3970e-07, + 2.5220e-06, 1.7416e-06], + [ 0.0000e+00, 0.0000e+00, 4.7823e-07, ..., 1.7248e-06, + 3.7942e-06, 4.7572e-06], + [ 0.0000e+00, 0.0000e+00, 1.3085e-07, ..., -2.7660e-06, + 6.4820e-07, -9.5963e-06]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0241, 0.0157, 0.0235, 0.0040, 0.0244, -0.0113, -0.0055, -0.0168, + -0.0148, 0.0213], device='cuda:0'), grad: tensor([ 6.4299e-06, -1.9848e-05, -2.1160e-06, -3.8333e-06, -4.3654e-04, + -2.8182e-06, 4.2653e-04, 2.1249e-05, 3.0115e-05, -1.9893e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 220.42, cls_loss 0.0031 cls_loss_mapping 0.0046 cls_loss_causal 0.5396 re_mapping 0.0079 re_causal 0.0237 /// teacc 98.97 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0146, 0.0110, 0.0053, ..., 0.0575, -0.1026, -0.1054], + [ 0.0348, -0.0185, 0.0080, ..., -0.1180, 0.0452, -0.0955], + [ 0.0042, -0.0048, 0.1226, ..., -0.1623, -0.0788, -0.1220], + ..., + [ 0.0043, 0.0062, -0.0819, ..., 0.0035, 0.0989, 0.0607], + [ 0.0069, -0.0202, -0.0701, ..., -0.1452, 0.0643, -0.1244], + [-0.0089, -0.0258, -0.0380, ..., 0.0291, -0.0560, 0.0162]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8161e-06, ..., 3.9525e-06, + 4.0866e-06, 1.4007e-05], + [ 2.7940e-09, 0.0000e+00, -1.4275e-05, ..., 2.7474e-07, + -2.8059e-05, -9.9391e-06], + [ 4.6566e-10, 0.0000e+00, -3.6154e-06, ..., 1.0375e-06, + 4.8503e-06, 4.1351e-06], + ..., + [-9.3132e-09, 0.0000e+00, 1.4510e-06, ..., 1.3374e-06, + -2.4602e-05, -5.8115e-06], + [ 9.3132e-10, 0.0000e+00, 2.3805e-06, ..., 7.8045e-07, + 4.6603e-06, 2.5257e-06], + [ 1.8626e-09, 0.0000e+00, 2.3702e-07, ..., -1.7524e-05, + 1.1884e-06, -6.2287e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0241, 0.0157, 0.0236, 0.0040, 0.0244, -0.0108, -0.0049, -0.0168, + -0.0146, 0.0204], device='cuda:0'), grad: tensor([ 4.0203e-05, -1.3423e-04, 1.3411e-05, 2.8372e-05, 6.4790e-05, + 5.3167e-05, 5.7250e-05, -3.3408e-05, 2.0653e-05, -1.1009e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 220.47, cls_loss 0.0033 cls_loss_mapping 0.0061 cls_loss_causal 0.5833 re_mapping 0.0078 re_causal 0.0243 /// teacc 98.84 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0146, 0.0110, 0.0062, ..., 0.0577, -0.1030, -0.1052], + [ 0.0347, -0.0185, 0.0076, ..., -0.1189, 0.0452, -0.0958], + [ 0.0041, -0.0048, 0.1231, ..., -0.1635, -0.0790, -0.1217], + ..., + [ 0.0043, 0.0062, -0.0828, ..., 0.0037, 0.0995, 0.0608], + [ 0.0068, -0.0202, -0.0708, ..., -0.1466, 0.0643, -0.1256], + [-0.0089, -0.0258, -0.0381, ..., 0.0289, -0.0564, 0.0158]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0207e-05, ..., -9.9465e-06, + 2.2352e-07, 2.0973e-06], + [ 1.3970e-08, 0.0000e+00, 3.4133e-07, ..., 9.0823e-06, + 7.6890e-06, 2.0623e-05], + [-1.8161e-08, 0.0000e+00, -1.1874e-06, ..., 1.2042e-06, + 1.2927e-06, 9.6485e-07], + ..., + [-5.0757e-08, 0.0000e+00, 1.9465e-06, ..., -1.9825e-04, + -1.8823e-04, -4.6325e-04], + [ 1.9558e-08, 0.0000e+00, 2.0731e-06, ..., 3.0305e-06, + 1.4529e-07, 3.9712e-06], + [ 8.3819e-09, 0.0000e+00, 6.4960e-07, ..., 1.6415e-04, + 1.5557e-04, 3.7861e-04]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0237, 0.0154, 0.0239, 0.0046, 0.0249, -0.0110, -0.0051, -0.0168, + -0.0152, 0.0202], device='cuda:0'), grad: tensor([-2.5705e-05, 2.6837e-05, 4.7944e-06, 1.2672e-04, 3.4362e-05, + -1.9038e-04, 6.0499e-06, -6.0415e-04, 1.0449e-04, 5.1689e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 220.69, cls_loss 0.0035 cls_loss_mapping 0.0073 cls_loss_causal 0.5712 re_mapping 0.0079 re_causal 0.0241 /// teacc 98.95 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0146, 0.0110, 0.0065, ..., 0.0579, -0.1037, -0.1057], + [ 0.0347, -0.0185, 0.0063, ..., -0.1193, 0.0450, -0.0959], + [ 0.0039, -0.0048, 0.1247, ..., -0.1650, -0.0783, -0.1219], + ..., + [ 0.0044, 0.0062, -0.0831, ..., 0.0037, 0.1001, 0.0611], + [ 0.0068, -0.0202, -0.0714, ..., -0.1484, 0.0659, -0.1267], + [-0.0090, -0.0258, -0.0384, ..., 0.0291, -0.0568, 0.0158]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.1399e-06, ..., -2.2948e-06, + 2.3236e-07, 1.8673e-07], + [ 3.7253e-09, 0.0000e+00, 2.0862e-07, ..., 1.1409e-07, + -5.0664e-06, -1.2629e-06], + [ 2.3283e-09, 0.0000e+00, -3.5558e-06, ..., 2.0163e-07, + 4.3260e-07, 2.8545e-07], + ..., + [-1.0245e-08, 0.0000e+00, 1.2666e-07, ..., 2.5891e-07, + -1.4128e-06, -1.7630e-06], + [ 4.6566e-10, 0.0000e+00, 3.0687e-07, ..., 4.9826e-07, + 1.7174e-06, 1.6354e-06], + [ 1.3970e-09, 0.0000e+00, 5.1595e-07, ..., 1.1884e-06, + 2.3004e-06, 5.7276e-07]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0237, 0.0150, 0.0249, 0.0045, 0.0248, -0.0117, -0.0068, -0.0166, + -0.0141, 0.0204], device='cuda:0'), grad: tensor([-2.6077e-06, -1.4238e-05, -2.0713e-06, -1.0893e-05, 3.2932e-06, + 6.4448e-06, 6.9151e-07, -2.3693e-06, 7.0669e-06, 1.4633e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 220.51, cls_loss 0.0031 cls_loss_mapping 0.0058 cls_loss_causal 0.5728 re_mapping 0.0076 re_causal 0.0234 /// teacc 98.92 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0145, 0.0110, 0.0067, ..., 0.0581, -0.1042, -0.1061], + [ 0.0349, -0.0185, 0.0053, ..., -0.1197, 0.0448, -0.0959], + [ 0.0038, -0.0048, 0.1255, ..., -0.1661, -0.0780, -0.1223], + ..., + [ 0.0044, 0.0062, -0.0834, ..., 0.0037, 0.1000, 0.0612], + [ 0.0068, -0.0202, -0.0720, ..., -0.1496, 0.0661, -0.1270], + [-0.0090, -0.0258, -0.0385, ..., 0.0293, -0.0571, 0.0161]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 4.6566e-10, ..., -5.3644e-06, + 6.1607e-07, 2.7940e-07], + [ 1.8626e-09, 0.0000e+00, 3.9376e-06, ..., 7.8324e-07, + 4.7758e-06, 1.5479e-06], + [ 2.7940e-09, 0.0000e+00, -1.2822e-05, ..., 3.2084e-07, + -3.9726e-05, 1.6689e-06], + ..., + [-1.8626e-09, 0.0000e+00, 4.0025e-05, ..., 9.9652e-07, + 9.9093e-06, 6.4149e-06], + [ 4.1910e-09, 0.0000e+00, 1.5028e-05, ..., 1.3616e-06, + 2.3574e-05, 2.7940e-06], + [ 9.3132e-10, 0.0000e+00, 3.2270e-07, ..., 1.9774e-05, + 1.9325e-07, 3.2663e-05]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0237, 0.0146, 0.0254, 0.0054, 0.0245, -0.0126, -0.0059, -0.0168, + -0.0147, 0.0206], device='cuda:0'), grad: tensor([-2.2128e-05, 2.0236e-05, -4.1902e-05, -2.7609e-04, -4.6730e-05, + 4.1015e-06, 2.3186e-05, 2.2316e-04, 7.1704e-05, 4.4554e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 220.87, cls_loss 0.0059 cls_loss_mapping 0.0079 cls_loss_causal 0.5702 re_mapping 0.0083 re_causal 0.0229 /// teacc 98.97 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0145, 0.0110, 0.0061, ..., 0.0615, -0.1051, -0.1061], + [ 0.0352, -0.0185, 0.0054, ..., -0.1201, 0.0454, -0.0959], + [ 0.0036, -0.0048, 0.1256, ..., -0.1673, -0.0791, -0.1236], + ..., + [ 0.0018, 0.0062, -0.0833, ..., 0.0034, 0.1000, 0.0595], + [ 0.0067, -0.0202, -0.0728, ..., -0.1507, 0.0661, -0.1276], + [-0.0082, -0.0258, -0.0387, ..., 0.0294, -0.0563, 0.0179]], + device='cuda:0'), grad: tensor([[ 1.7043e-07, 0.0000e+00, -5.6485e-07, ..., -1.9148e-06, + 1.9046e-07, 3.9302e-07], + [ 3.6741e-07, 0.0000e+00, 5.2620e-08, ..., 5.9837e-07, + 1.0515e-06, 4.8876e-06], + [ 8.3400e-07, 0.0000e+00, -1.3178e-07, ..., 1.5600e-07, + 1.8300e-07, 3.9302e-07], + ..., + [ 2.5658e-07, 0.0000e+00, 1.4296e-07, ..., 9.9186e-07, + -8.7693e-06, -1.5289e-05], + [ 1.2619e-06, 0.0000e+00, 1.0245e-07, ..., 3.1851e-07, + 1.0077e-06, 6.6636e-07], + [ 1.5777e-06, 0.0000e+00, 1.3039e-07, ..., -4.9137e-06, + 4.5374e-06, -1.5169e-05]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0186, 0.0149, 0.0237, 0.0040, 0.0245, -0.0106, -0.0099, -0.0173, + -0.0153, 0.0222], device='cuda:0'), grad: tensor([-4.0680e-06, 7.3388e-06, 5.2415e-06, -3.3528e-05, 1.6436e-05, + 1.7121e-05, -2.3376e-06, -1.9252e-05, 1.1668e-05, 1.3215e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 220.80, cls_loss 0.0028 cls_loss_mapping 0.0047 cls_loss_causal 0.5613 re_mapping 0.0079 re_causal 0.0229 /// teacc 98.97 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0143, 0.0110, 0.0063, ..., 0.0617, -0.1050, -0.1073], + [ 0.0359, -0.0185, 0.0056, ..., -0.1203, 0.0459, -0.0959], + [ 0.0031, -0.0048, 0.1260, ..., -0.1680, -0.0794, -0.1237], + ..., + [ 0.0019, 0.0062, -0.0845, ..., 0.0028, 0.1001, 0.0591], + [ 0.0066, -0.0202, -0.0733, ..., -0.1514, 0.0661, -0.1280], + [-0.0083, -0.0258, -0.0389, ..., 0.0295, -0.0569, 0.0180]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 0.0000e+00, 1.1316e-07, ..., 7.2181e-05, + 7.3109e-07, 8.7857e-05], + [ 8.8010e-08, 0.0000e+00, 2.1420e-08, ..., 2.1774e-06, + 1.1874e-07, 3.7681e-06], + [ 1.3551e-07, 0.0000e+00, 7.8464e-07, ..., 5.6103e-06, + 1.3467e-06, 7.5288e-06], + ..., + [-4.2561e-07, 0.0000e+00, 2.2817e-08, ..., 4.8988e-07, + -2.7250e-06, -3.7458e-06], + [ 1.7788e-07, 0.0000e+00, -1.3337e-06, ..., 1.6950e-06, + -1.0859e-06, 2.2054e-06], + [ 1.1595e-07, 0.0000e+00, 2.0768e-07, ..., -1.3924e-04, + 1.2591e-06, -1.6809e-04]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0185, 0.0153, 0.0236, 0.0039, 0.0245, -0.0101, -0.0099, -0.0179, + -0.0155, 0.0222], device='cuda:0'), grad: tensor([ 1.6558e-04, 5.6997e-06, 1.8984e-05, 2.4168e-07, 1.1849e-04, + 8.7917e-07, 6.4634e-06, -4.8131e-06, -2.8387e-06, -3.0899e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 220.36, cls_loss 0.0030 cls_loss_mapping 0.0047 cls_loss_causal 0.5779 re_mapping 0.0074 re_causal 0.0231 /// teacc 98.94 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0141, 0.0110, 0.0065, ..., 0.0618, -0.1055, -0.1075], + [ 0.0361, -0.0185, 0.0056, ..., -0.1210, 0.0463, -0.0962], + [ 0.0029, -0.0048, 0.1263, ..., -0.1684, -0.0799, -0.1239], + ..., + [ 0.0019, 0.0062, -0.0850, ..., 0.0026, 0.1006, 0.0595], + [ 0.0064, -0.0202, -0.0741, ..., -0.1520, 0.0662, -0.1287], + [-0.0090, -0.0258, -0.0394, ..., 0.0293, -0.0576, 0.0177]], + device='cuda:0'), grad: tensor([[ 2.0210e-07, 0.0000e+00, -2.7061e-05, ..., -1.3709e-05, + -1.1712e-05, -1.4104e-05], + [ 2.5053e-07, 0.0000e+00, 3.7923e-06, ..., 2.9858e-06, + 3.4589e-06, 5.6773e-06], + [ 3.7439e-07, 0.0000e+00, 1.9837e-06, ..., 2.2743e-06, + 2.6654e-06, 3.5074e-06], + ..., + [ 3.0501e-07, 0.0000e+00, 9.1493e-06, ..., 5.5544e-06, + 4.5672e-06, 1.1884e-05], + [ 1.8040e-06, 0.0000e+00, 1.0103e-05, ..., 4.9807e-06, + 1.8761e-05, 1.1846e-05], + [ 1.7323e-07, 0.0000e+00, 1.8049e-06, ..., -3.6173e-06, + -6.1467e-08, -1.8999e-05]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0185, 0.0154, 0.0235, 0.0033, 0.0252, -0.0111, -0.0100, -0.0176, + -0.0140, 0.0217], device='cuda:0'), grad: tensor([-5.5164e-05, 2.3857e-05, 1.3202e-05, -7.8738e-05, 1.3433e-05, + -9.1314e-05, 1.4164e-05, 4.9114e-05, 1.2815e-04, -1.6719e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 221.08, cls_loss 0.0035 cls_loss_mapping 0.0057 cls_loss_causal 0.5338 re_mapping 0.0080 re_causal 0.0225 /// teacc 98.95 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0139, 0.0110, 0.0068, ..., 0.0614, -0.1065, -0.1105], + [ 0.0351, -0.0185, 0.0055, ..., -0.1217, 0.0465, -0.0966], + [ 0.0030, -0.0048, 0.1275, ..., -0.1694, -0.0800, -0.1240], + ..., + [ 0.0023, 0.0062, -0.0854, ..., 0.0023, 0.1017, 0.0598], + [ 0.0060, -0.0202, -0.0771, ..., -0.1529, 0.0657, -0.1295], + [-0.0095, -0.0258, -0.0398, ..., 0.0298, -0.0584, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4646e-07, ..., -1.3774e-06, + 6.6543e-07, 2.8498e-07], + [ 0.0000e+00, 0.0000e+00, 1.8580e-07, ..., 6.1607e-07, + 1.8999e-07, 1.2368e-06], + [ 0.0000e+00, 0.0000e+00, -1.5944e-06, ..., 3.8231e-07, + 4.2887e-07, 6.5938e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6985e-07, ..., -6.9430e-07, + -3.1479e-06, -7.1302e-06], + [ 0.0000e+00, 0.0000e+00, 6.6031e-07, ..., 2.2585e-07, + -2.4680e-07, 2.4308e-07], + [ 4.6566e-10, 0.0000e+00, 2.4261e-07, ..., 1.5972e-07, + 2.3916e-06, 1.2182e-06]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0191, 0.0155, 0.0241, 0.0025, 0.0257, -0.0101, -0.0099, -0.0172, + -0.0151, 0.0215], device='cuda:0'), grad: tensor([-7.7114e-07, 1.6335e-06, -9.3132e-09, 1.3914e-06, 2.8238e-06, + 4.6985e-07, -9.3924e-07, -9.6932e-06, 2.1560e-07, 4.8503e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 220.88, cls_loss 0.0038 cls_loss_mapping 0.0055 cls_loss_causal 0.5608 re_mapping 0.0074 re_causal 0.0224 /// teacc 98.86 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0137, 0.0110, 0.0061, ..., 0.0611, -0.1082, -0.1109], + [ 0.0346, -0.0185, 0.0054, ..., -0.1233, 0.0457, -0.0970], + [ 0.0023, -0.0048, 0.1298, ..., -0.1702, -0.0803, -0.1248], + ..., + [-0.0002, 0.0062, -0.0858, ..., 0.0021, 0.1018, 0.0586], + [ 0.0058, -0.0202, -0.0801, ..., -0.1527, 0.0661, -0.1289], + [-0.0064, -0.0258, -0.0398, ..., 0.0305, -0.0585, 0.0192]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 0.0000e+00, -1.9856e-06, ..., -5.4240e-06, + 1.4110e-07, 9.8348e-07], + [ 1.3970e-08, 0.0000e+00, 1.9977e-07, ..., 7.9954e-07, + 7.1572e-07, 3.2075e-06], + [ 6.9849e-09, 0.0000e+00, -4.8894e-08, ..., 8.1630e-07, + 2.8592e-07, 2.7269e-06], + ..., + [-9.1270e-08, 0.0000e+00, 6.2911e-07, ..., 8.6501e-06, + -4.8876e-06, 2.3589e-05], + [ 1.3039e-08, 0.0000e+00, 1.0375e-06, ..., 7.0361e-07, + -3.9712e-06, -6.9477e-07], + [ 2.2817e-08, 0.0000e+00, 6.8219e-07, ..., 1.1660e-06, + 6.3777e-06, 6.4820e-06]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0195, 0.0149, 0.0255, 0.0028, 0.0248, -0.0102, -0.0091, -0.0185, + -0.0165, 0.0233], device='cuda:0'), grad: tensor([-9.9167e-06, -3.6645e-04, 8.8632e-05, 3.4124e-05, -4.2081e-05, + 9.3430e-06, 1.0915e-05, 2.5368e-04, -5.4855e-07, 2.2143e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 220.58, cls_loss 0.0023 cls_loss_mapping 0.0052 cls_loss_causal 0.5181 re_mapping 0.0078 re_causal 0.0230 /// teacc 98.96 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0135, 0.0110, 0.0062, ..., 0.0612, -0.1093, -0.1112], + [ 0.0345, -0.0185, 0.0054, ..., -0.1238, 0.0461, -0.0971], + [ 0.0022, -0.0048, 0.1303, ..., -0.1708, -0.0803, -0.1250], + ..., + [-0.0002, 0.0062, -0.0866, ..., 0.0023, 0.1028, 0.0592], + [ 0.0056, -0.0202, -0.0802, ..., -0.1534, 0.0659, -0.1295], + [-0.0064, -0.0258, -0.0409, ..., 0.0303, -0.0603, 0.0187]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, 0.0000e+00, -3.2736e-07, ..., -1.8189e-06, + 3.3434e-06, 2.5937e-07], + [ 5.8673e-08, 0.0000e+00, 1.1921e-06, ..., -6.9384e-08, + -1.7732e-06, 8.6380e-07], + [ 4.3306e-08, 0.0000e+00, -6.0946e-06, ..., 9.9652e-08, + 7.8976e-07, 4.6566e-07], + ..., + [-6.4168e-07, 0.0000e+00, 2.3134e-06, ..., 2.3888e-07, + -5.9232e-06, -1.1899e-05], + [ 4.1118e-07, 0.0000e+00, 7.7626e-07, ..., 1.6810e-07, + 1.6764e-05, 1.3094e-06], + [ 1.7229e-07, 0.0000e+00, 5.2853e-07, ..., 1.6131e-06, + 2.0526e-06, 2.9970e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0197, 0.0152, 0.0258, 0.0025, 0.0248, -0.0099, -0.0090, -0.0180, + -0.0168, 0.0226], device='cuda:0'), grad: tensor([ 1.3776e-05, -3.3546e-06, -6.6385e-06, 8.4564e-06, -3.7719e-07, + 3.2902e-05, -1.2815e-04, -1.4037e-05, 8.7857e-05, 9.4995e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 220.47, cls_loss 0.0029 cls_loss_mapping 0.0046 cls_loss_causal 0.5423 re_mapping 0.0076 re_causal 0.0227 /// teacc 98.92 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 1.3488e-02, 1.0998e-02, 6.1781e-03, ..., 6.1231e-02, + -1.1014e-01, -1.1139e-01], + [ 3.4479e-02, -1.8468e-02, 5.4269e-03, ..., -1.2416e-01, + 4.6188e-02, -9.7214e-02], + [ 1.2695e-03, -4.7728e-03, 1.3049e-01, ..., -1.7136e-01, + -8.0584e-02, -1.2551e-01], + ..., + [-1.0233e-04, 6.2061e-03, -8.7203e-02, ..., 2.0870e-03, + 1.0327e-01, 5.9490e-02], + [ 5.0882e-03, -2.0220e-02, -8.0246e-02, ..., -1.5383e-01, + 6.6086e-02, -1.3004e-01], + [-6.3909e-03, -2.5834e-02, -4.1100e-02, ..., 3.2395e-02, + -6.1235e-02, 2.1199e-02]], device='cuda:0'), grad: tensor([[ 1.6298e-09, 0.0000e+00, -2.3306e-07, ..., -2.6659e-07, + 2.9057e-07, 2.9011e-07], + [ 1.0943e-08, 0.0000e+00, 1.6904e-07, ..., 2.8568e-07, + -2.6189e-06, 6.4261e-07], + [ 6.0536e-09, 0.0000e+00, -8.6008e-07, ..., 1.0291e-07, + 1.3206e-06, 1.1791e-06], + ..., + [-3.8184e-08, 0.0000e+00, 2.9220e-07, ..., 2.1234e-07, + -5.5321e-07, -1.3048e-06], + [ 1.3271e-08, 0.0000e+00, 2.4843e-07, ..., 2.4401e-07, + -1.0114e-06, -1.0114e-06], + [ 9.3132e-09, 0.0000e+00, 1.6764e-07, ..., 1.2517e-05, + 3.0361e-06, 2.2486e-05]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0198, 0.0154, 0.0256, 0.0027, 0.0222, -0.0095, -0.0090, -0.0180, + -0.0168, 0.0244], device='cuda:0'), grad: tensor([ 5.5600e-07, -4.5523e-06, 3.8743e-06, -2.1560e-07, -2.3246e-05, + -4.2841e-07, 9.2248e-07, -2.2929e-06, -1.9278e-06, 2.7284e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 220.41, cls_loss 0.0037 cls_loss_mapping 0.0053 cls_loss_causal 0.5514 re_mapping 0.0073 re_causal 0.0218 /// teacc 98.93 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 1.3350e-02, 1.0998e-02, 6.4693e-03, ..., 6.1342e-02, + -1.1097e-01, -1.1170e-01], + [ 3.4528e-02, -1.8468e-02, 4.8247e-03, ..., -1.2368e-01, + 4.6575e-02, -9.6590e-02], + [ 1.1111e-03, -4.7728e-03, 1.3181e-01, ..., -1.7306e-01, + -8.0198e-02, -1.2558e-01], + ..., + [-9.8335e-05, 6.2061e-03, -8.9080e-02, ..., 1.7592e-03, + 1.0464e-01, 6.1060e-02], + [ 5.0435e-03, -2.0220e-02, -8.0363e-02, ..., -1.5462e-01, + 6.6202e-02, -1.3029e-01], + [-6.4582e-03, -2.5834e-02, -4.2535e-02, ..., 3.1601e-02, + -6.4460e-02, 1.9699e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.5682e-08, ..., 4.6566e-09, + 5.8208e-07, 5.3272e-07], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 4.0652e-07, + -1.2210e-06, 1.9222e-06], + [ 0.0000e+00, 0.0000e+00, -5.0012e-07, ..., 1.6717e-07, + 8.5831e-06, 4.3735e-06], + ..., + [-9.3132e-10, 0.0000e+00, 4.0513e-08, ..., 8.8615e-07, + -2.6405e-05, -1.2703e-05], + [ 0.0000e+00, 0.0000e+00, 7.5903e-08, ..., 7.2783e-07, + 1.0312e-05, 6.5453e-06], + [ 0.0000e+00, 0.0000e+00, 8.2888e-08, ..., 7.4245e-06, + 2.9318e-06, 1.7643e-05]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0198, 0.0155, 0.0264, 0.0026, 0.0226, -0.0089, -0.0086, -0.0172, + -0.0169, 0.0225], device='cuda:0'), grad: tensor([ 2.0321e-06, -4.3772e-06, 2.3425e-05, 2.0057e-05, -2.5913e-05, + -2.9057e-06, 9.3430e-06, -6.4969e-05, 1.5825e-05, 2.7359e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 220.50, cls_loss 0.0034 cls_loss_mapping 0.0049 cls_loss_causal 0.5437 re_mapping 0.0075 re_causal 0.0219 /// teacc 98.90 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 1.3311e-02, 1.0998e-02, 6.5013e-03, ..., 6.0897e-02, + -1.1177e-01, -1.1357e-01], + [ 3.4496e-02, -1.8468e-02, 4.4525e-03, ..., -1.2399e-01, + 4.6960e-02, -9.6689e-02], + [ 7.8872e-04, -4.7728e-03, 1.3218e-01, ..., -1.7421e-01, + -8.1031e-02, -1.2664e-01], + ..., + [ 8.3661e-05, 6.2061e-03, -8.8973e-02, ..., 1.5040e-03, + 1.0510e-01, 5.9703e-02], + [ 4.9633e-03, -2.0220e-02, -8.0436e-02, ..., -1.5539e-01, + 6.6598e-02, -1.3076e-01], + [-6.5966e-03, -2.5834e-02, -4.2803e-02, ..., 3.1851e-02, + -6.5338e-02, 2.0774e-02]], device='cuda:0'), grad: tensor([[ 3.2596e-09, 0.0000e+00, -6.2538e-07, ..., -5.4622e-07, + 8.7544e-08, 3.0128e-07], + [ 1.3504e-08, 0.0000e+00, 5.4017e-08, ..., 5.4017e-08, + 5.6950e-07, 4.2245e-06], + [ 6.5193e-09, 0.0000e+00, 4.1723e-07, ..., 4.0047e-08, + 7.8604e-07, 3.1721e-06], + ..., + [ 1.9558e-08, 0.0000e+00, 1.7183e-07, ..., 7.3574e-08, + -4.8988e-06, -2.7582e-05], + [ 4.6566e-09, 0.0000e+00, 1.3970e-06, ..., 1.3644e-07, + -1.4203e-07, 3.9628e-07], + [ 1.9604e-07, 0.0000e+00, 1.2107e-07, ..., 5.0757e-07, + 3.2056e-06, 1.8775e-05]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0203, 0.0156, 0.0262, 0.0013, 0.0227, -0.0081, -0.0085, -0.0185, + -0.0166, 0.0236], device='cuda:0'), grad: tensor([-1.4305e-06, 6.2436e-06, 6.3404e-06, -7.6517e-06, -1.7229e-07, + 2.0750e-06, 1.4231e-06, -4.1246e-05, 5.5507e-06, 2.8849e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 220.44, cls_loss 0.0033 cls_loss_mapping 0.0060 cls_loss_causal 0.5589 re_mapping 0.0071 re_causal 0.0225 /// teacc 98.85 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 1.3273e-02, 1.0998e-02, 6.4062e-03, ..., 6.0785e-02, + -1.1217e-01, -1.1398e-01], + [ 3.4476e-02, -1.8468e-02, 4.2065e-03, ..., -1.2452e-01, + 4.6997e-02, -9.6733e-02], + [ 7.2692e-04, -4.7728e-03, 1.3301e-01, ..., -1.7607e-01, + -7.9144e-02, -1.2746e-01], + ..., + [ 2.1093e-05, 6.2061e-03, -8.8546e-02, ..., 2.3072e-03, + 1.0601e-01, 6.0164e-02], + [ 4.8926e-03, -2.0220e-02, -8.1623e-02, ..., -1.5531e-01, + 6.4581e-02, -1.2945e-01], + [-6.5884e-03, -2.5834e-02, -4.2704e-02, ..., 3.1570e-02, + -6.6626e-02, 2.0431e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-06, ..., -2.3786e-06, + 1.7975e-06, 3.4831e-07], + [ 0.0000e+00, 0.0000e+00, 1.6578e-07, ..., 1.2154e-07, + 5.6475e-06, 3.9749e-06], + [ 0.0000e+00, 0.0000e+00, 4.3772e-07, ..., 8.6846e-07, + 1.9193e-05, 9.4473e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 7.9628e-08, ..., 8.9873e-08, + -3.2008e-05, -2.1324e-05], + [ 0.0000e+00, 0.0000e+00, 7.1293e-07, ..., 1.8859e-07, + 2.0117e-06, -3.7365e-06], + [ 0.0000e+00, 0.0000e+00, 8.8988e-07, ..., 1.4929e-06, + 4.0643e-06, 8.8066e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0204, 0.0152, 0.0273, 0.0040, 0.0229, -0.0105, -0.0081, -0.0177, + -0.0183, 0.0231], device='cuda:0'), grad: tensor([-9.4436e-07, 2.5824e-05, 7.9870e-05, 1.2711e-05, 5.1446e-06, + 2.1994e-04, -2.4772e-04, -1.2743e-04, 3.8464e-07, 3.1888e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 220.57, cls_loss 0.0026 cls_loss_mapping 0.0051 cls_loss_causal 0.5741 re_mapping 0.0075 re_causal 0.0227 /// teacc 98.97 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 1.3266e-02, 1.0998e-02, 6.9918e-03, ..., 6.0761e-02, + -1.1381e-01, -1.1412e-01], + [ 3.4470e-02, -1.8468e-02, 4.0587e-03, ..., -1.2475e-01, + 4.6834e-02, -9.7002e-02], + [ 7.2522e-04, -4.7728e-03, 1.3304e-01, ..., -1.7856e-01, + -7.9128e-02, -1.2775e-01], + ..., + [-5.7556e-08, 6.2061e-03, -8.8981e-02, ..., 2.0227e-03, + 1.0673e-01, 6.0333e-02], + [ 4.8873e-03, -2.0220e-02, -8.1628e-02, ..., -1.5637e-01, + 6.4542e-02, -1.2952e-01], + [-6.9282e-03, -2.5834e-02, -4.3314e-02, ..., 3.1308e-02, + -6.7198e-02, 1.9878e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2128e-06, ..., -5.9009e-06, + 2.5854e-06, 6.9384e-08], + [ 0.0000e+00, 0.0000e+00, 3.0128e-07, ..., 1.6857e-07, + 4.3586e-07, 2.4028e-07], + [ 0.0000e+00, 0.0000e+00, 3.1758e-07, ..., 4.7404e-07, + 1.2554e-06, 1.9511e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-07, ..., 6.9896e-07, + -4.9314e-07, 2.6170e-07], + [ 0.0000e+00, 0.0000e+00, -3.7104e-06, ..., 4.4005e-07, + -9.8944e-06, 2.0443e-07], + [ 0.0000e+00, 0.0000e+00, 8.7451e-07, ..., 2.4941e-06, + 1.1288e-06, 7.0175e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0205, 0.0150, 0.0271, 0.0037, 0.0236, -0.0102, -0.0079, -0.0173, + -0.0184, 0.0225], device='cuda:0'), grad: tensor([ 1.0081e-05, 3.8892e-06, 6.8694e-06, 2.3752e-05, -1.1679e-06, + 6.0424e-06, 1.4171e-05, 6.0908e-07, -7.7665e-05, 1.3463e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 220.72, cls_loss 0.0028 cls_loss_mapping 0.0044 cls_loss_causal 0.5265 re_mapping 0.0068 re_causal 0.0212 /// teacc 98.92 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0133, 0.0110, 0.0073, ..., 0.0608, -0.1145, -0.1144], + [ 0.0345, -0.0185, 0.0037, ..., -0.1254, 0.0470, -0.0972], + [ 0.0007, -0.0048, 0.1336, ..., -0.1800, -0.0793, -0.1275], + ..., + [-0.0004, 0.0062, -0.0900, ..., 0.0014, 0.1068, 0.0603], + [ 0.0049, -0.0202, -0.0818, ..., -0.1580, 0.0649, -0.1299], + [-0.0070, -0.0258, -0.0437, ..., 0.0317, -0.0669, 0.0200]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.7521e-08, ..., 1.0794e-06, + 5.5227e-07, 1.3867e-06], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 5.4482e-08, + -1.0906e-06, 1.2806e-07], + [ 0.0000e+00, 0.0000e+00, -2.1374e-07, ..., 3.8184e-08, + 4.9360e-07, 1.0431e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0943e-07, ..., 1.2461e-06, + -8.4518e-07, -6.3144e-07], + [ 0.0000e+00, 0.0000e+00, 1.2945e-07, ..., 3.6275e-07, + -8.7544e-08, 3.6415e-07], + [ 0.0000e+00, 0.0000e+00, 1.9465e-07, ..., -3.3490e-06, + 9.0199e-07, -2.0433e-06]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0205, 0.0149, 0.0271, 0.0035, 0.0235, -0.0101, -0.0079, -0.0174, + -0.0181, 0.0226], device='cuda:0'), grad: tensor([ 3.7830e-06, -2.4661e-06, 1.3812e-06, -7.6368e-07, 1.3094e-06, + -4.9686e-07, -4.0559e-07, 4.1071e-07, 1.0710e-06, -3.8184e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 143---------------------------------------------------- +epoch 143, time 221.54, cls_loss 0.0023 cls_loss_mapping 0.0047 cls_loss_causal 0.5335 re_mapping 0.0071 re_causal 0.0224 /// teacc 99.01 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0133, 0.0110, 0.0078, ..., 0.0609, -0.1157, -0.1147], + [ 0.0345, -0.0185, 0.0037, ..., -0.1257, 0.0477, -0.0970], + [ 0.0007, -0.0048, 0.1339, ..., -0.1808, -0.0796, -0.1273], + ..., + [-0.0004, 0.0062, -0.0906, ..., 0.0012, 0.1067, 0.0604], + [ 0.0048, -0.0202, -0.0818, ..., -0.1598, 0.0652, -0.1304], + [-0.0070, -0.0258, -0.0442, ..., 0.0317, -0.0672, 0.0199]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.4075e-07, ..., -7.0920e-07, + 4.1686e-06, 9.0674e-06], + [ 0.0000e+00, 0.0000e+00, 1.5143e-06, ..., 2.1420e-08, + 6.2585e-07, 5.5833e-07], + [ 0.0000e+00, 0.0000e+00, -7.1563e-06, ..., 5.5414e-08, + -2.3711e-06, -2.1793e-07], + ..., + [-4.6566e-10, 0.0000e+00, 4.0606e-06, ..., 8.0559e-08, + -2.8126e-06, -1.1437e-05], + [ 0.0000e+00, 0.0000e+00, 4.3325e-06, ..., 9.3132e-08, + 4.8615e-06, 4.0466e-07], + [-4.6566e-10, 0.0000e+00, 4.2841e-07, ..., 6.1933e-08, + 2.2072e-07, 7.9628e-08]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0209, 0.0155, 0.0269, 0.0035, 0.0236, -0.0102, -0.0075, -0.0175, + -0.0182, 0.0225], device='cuda:0'), grad: tensor([ 1.7077e-05, 3.1348e-06, -1.0774e-05, 2.1420e-06, 2.0973e-06, + 5.0552e-06, -2.1130e-05, -1.2740e-05, 1.2353e-05, 2.7493e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 220.74, cls_loss 0.0024 cls_loss_mapping 0.0049 cls_loss_causal 0.5503 re_mapping 0.0071 re_causal 0.0220 /// teacc 98.95 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0132, 0.0110, 0.0076, ..., 0.0607, -0.1170, -0.1156], + [ 0.0344, -0.0185, 0.0030, ..., -0.1256, 0.0474, -0.0969], + [ 0.0005, -0.0048, 0.1341, ..., -0.1817, -0.0795, -0.1274], + ..., + [-0.0007, 0.0062, -0.0913, ..., 0.0011, 0.1071, 0.0603], + [ 0.0048, -0.0202, -0.0818, ..., -0.1608, 0.0655, -0.1311], + [-0.0066, -0.0258, -0.0445, ..., 0.0318, -0.0675, 0.0199]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 1.0571e-07, ..., -4.9360e-08, + 3.1525e-07, 1.6205e-07], + [ 1.3970e-09, 0.0000e+00, 1.7006e-06, ..., 1.1967e-07, + -2.7474e-07, 7.5949e-07], + [ 1.8626e-09, 0.0000e+00, -4.5002e-06, ..., -1.1735e-07, + 6.7661e-07, 2.7055e-07], + ..., + [ 2.2352e-08, 0.0000e+00, 7.3202e-07, ..., 3.3341e-07, + 1.2755e-05, 9.4026e-06], + [ 1.8626e-09, 0.0000e+00, 6.3190e-07, ..., 1.0757e-07, + -7.7367e-05, -5.0515e-05], + [-2.9337e-08, 0.0000e+00, 1.0431e-07, ..., -7.5810e-07, + 6.0678e-05, 3.6865e-05]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0211, 0.0152, 0.0268, 0.0040, 0.0236, -0.0098, -0.0077, -0.0175, + -0.0182, 0.0224], device='cuda:0'), grad: tensor([ 8.6986e-07, 3.4571e-06, -7.3686e-06, 4.8662e-07, 4.4815e-06, + 2.5891e-06, 9.9279e-07, 2.5421e-05, -1.3530e-04, 1.0425e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 145---------------------------------------------------- +epoch 145, time 220.61, cls_loss 0.0037 cls_loss_mapping 0.0055 cls_loss_causal 0.5438 re_mapping 0.0070 re_causal 0.0204 /// teacc 99.03 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 0.0131, 0.0110, 0.0082, ..., 0.0611, -0.1175, -0.1163], + [ 0.0343, -0.0185, 0.0024, ..., -0.1269, 0.0457, -0.0971], + [ 0.0005, -0.0048, 0.1352, ..., -0.1829, -0.0796, -0.1275], + ..., + [-0.0012, 0.0062, -0.0921, ..., 0.0020, 0.1080, 0.0607], + [ 0.0048, -0.0202, -0.0820, ..., -0.1618, 0.0654, -0.1315], + [-0.0060, -0.0258, -0.0449, ..., 0.0314, -0.0679, 0.0199]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6045e-06, ..., -5.0776e-06, + 2.6636e-07, 1.9968e-06], + [ 0.0000e+00, 0.0000e+00, 1.7229e-07, ..., 3.9535e-07, + -1.3439e-06, 7.2317e-07], + [ 0.0000e+00, 0.0000e+00, 1.6158e-06, ..., 1.2312e-06, + 2.0564e-06, 4.2617e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1735e-07, ..., 1.2340e-07, + -8.8010e-08, -1.2796e-06], + [ 0.0000e+00, 0.0000e+00, 5.9977e-07, ..., 1.0282e-06, + -2.9355e-06, 2.7135e-05], + [ 0.0000e+00, 0.0000e+00, 1.0021e-06, ..., 8.5728e-07, + 5.2107e-07, -1.8537e-04]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0210, 0.0137, 0.0276, 0.0053, 0.0234, -0.0112, -0.0069, -0.0171, + -0.0184, 0.0221], device='cuda:0'), grad: tensor([-6.6459e-06, -1.3858e-06, 2.7910e-05, 4.9162e-04, 1.7226e-05, + 1.7762e-04, 4.2915e-06, 9.2806e-07, 1.1933e-04, -8.3113e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 220.55, cls_loss 0.0033 cls_loss_mapping 0.0046 cls_loss_causal 0.5520 re_mapping 0.0069 re_causal 0.0208 /// teacc 99.00 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0131, 0.0110, 0.0084, ..., 0.0612, -0.1186, -0.1171], + [ 0.0343, -0.0185, 0.0023, ..., -0.1276, 0.0454, -0.0974], + [-0.0009, -0.0048, 0.1357, ..., -0.1841, -0.0800, -0.1280], + ..., + [-0.0017, 0.0062, -0.0925, ..., 0.0016, 0.1096, 0.0607], + [ 0.0047, -0.0202, -0.0822, ..., -0.1631, 0.0652, -0.1327], + [-0.0059, -0.0258, -0.0460, ..., 0.0319, -0.0682, 0.0205]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7474e-08, ..., -7.6368e-08, + 1.0105e-07, 2.8405e-08], + [ 0.0000e+00, 0.0000e+00, 4.9360e-08, ..., 1.9278e-07, + -4.7497e-07, 3.9162e-07], + [ 0.0000e+00, 0.0000e+00, 4.0047e-06, ..., 2.9337e-08, + 8.4331e-07, 7.3109e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.4231e-06, ..., 2.0163e-07, + -3.1618e-07, -6.9663e-07], + [ 0.0000e+00, 0.0000e+00, -6.9290e-07, ..., 1.2573e-07, + -1.0552e-06, 1.2619e-07], + [ 0.0000e+00, 0.0000e+00, 2.1886e-07, ..., 2.1532e-06, + 4.4471e-07, 2.5034e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0213, 0.0135, 0.0274, 0.0051, 0.0227, -0.0111, -0.0068, -0.0167, + -0.0190, 0.0230], device='cuda:0'), grad: tensor([ 7.9954e-07, -1.7192e-06, 7.3165e-06, -3.7625e-06, -4.0010e-06, + -5.5358e-06, 8.8755e-07, 2.4531e-06, -1.4752e-06, 5.0180e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 220.53, cls_loss 0.0027 cls_loss_mapping 0.0050 cls_loss_causal 0.5234 re_mapping 0.0068 re_causal 0.0207 /// teacc 98.94 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0130, 0.0110, 0.0094, ..., 0.0614, -0.1192, -0.1173], + [ 0.0344, -0.0185, 0.0028, ..., -0.1282, 0.0458, -0.0974], + [-0.0018, -0.0048, 0.1358, ..., -0.1856, -0.0802, -0.1291], + ..., + [-0.0024, 0.0062, -0.0934, ..., 0.0010, 0.1097, 0.0606], + [ 0.0046, -0.0202, -0.0823, ..., -0.1638, 0.0655, -0.1326], + [-0.0058, -0.0258, -0.0470, ..., 0.0316, -0.0688, 0.0204]], + device='cuda:0'), grad: tensor([[ 5.6811e-08, 0.0000e+00, 1.6065e-07, ..., 7.2410e-07, + 1.2163e-06, 2.1048e-07], + [-8.7544e-08, 0.0000e+00, -6.2864e-07, ..., 1.9092e-08, + 2.3749e-08, 5.7556e-07], + [ 8.4750e-08, 0.0000e+00, 4.1630e-07, ..., 8.1956e-08, + 1.8049e-06, 4.2515e-07], + ..., + [ 2.8405e-08, 0.0000e+00, 7.6834e-08, ..., 7.6974e-07, + -4.1462e-06, 2.3516e-07], + [-9.0385e-07, 0.0000e+00, 1.8300e-07, ..., -3.3528e-06, + -2.6405e-05, -6.4149e-06], + [ 9.1735e-08, 0.0000e+00, 2.7474e-08, ..., -2.3767e-05, + 2.4781e-05, -5.7191e-05]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0212, 0.0144, 0.0268, 0.0057, 0.0232, -0.0114, -0.0065, -0.0172, + -0.0189, 0.0226], device='cuda:0'), grad: tensor([ 2.1711e-05, 3.6089e-07, 7.9870e-06, 7.4655e-06, 5.8353e-05, + 1.1347e-05, -4.9509e-06, -9.5218e-06, -2.2459e-04, 1.3185e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 220.34, cls_loss 0.0021 cls_loss_mapping 0.0040 cls_loss_causal 0.5367 re_mapping 0.0070 re_causal 0.0219 /// teacc 98.86 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0128, 0.0110, 0.0106, ..., 0.0621, -0.1198, -0.1170], + [ 0.0346, -0.0185, 0.0028, ..., -0.1293, 0.0460, -0.0979], + [-0.0020, -0.0048, 0.1364, ..., -0.1871, -0.0800, -0.1275], + ..., + [-0.0025, 0.0062, -0.0945, ..., 0.0007, 0.1094, 0.0604], + [ 0.0045, -0.0202, -0.0824, ..., -0.1643, 0.0657, -0.1326], + [-0.0060, -0.0258, -0.0489, ..., 0.0312, -0.0693, 0.0203]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.7211e-06, ..., -1.3616e-06, + 2.0955e-08, 2.9802e-08], + [ 4.6566e-09, 0.0000e+00, 1.7695e-07, ..., 1.5600e-07, + 1.0971e-06, 1.5367e-06], + [ 3.2596e-09, 0.0000e+00, 3.1432e-07, ..., 2.6217e-07, + 1.2200e-07, 1.1362e-07], + ..., + [-2.7940e-08, 0.0000e+00, 8.3819e-08, ..., 8.7079e-08, + -3.4031e-06, -4.1313e-06], + [ 3.2596e-09, 0.0000e+00, 5.5088e-07, ..., 3.7206e-07, + -5.5414e-08, 4.2887e-07], + [-9.6392e-08, 0.0000e+00, 2.4820e-07, ..., -2.4885e-06, + 6.8499e-07, -1.2018e-05]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0210, 0.0144, 0.0273, 0.0057, 0.0235, -0.0113, -0.0066, -0.0176, + -0.0188, 0.0221], device='cuda:0'), grad: tensor([-3.4403e-06, 2.8871e-06, 1.1791e-06, 3.0063e-06, 1.2219e-05, + 2.5006e-07, 1.1865e-06, -8.6427e-06, 1.0533e-06, -9.6634e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 220.57, cls_loss 0.0029 cls_loss_mapping 0.0063 cls_loss_causal 0.5331 re_mapping 0.0073 re_causal 0.0214 /// teacc 98.84 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0123, 0.0110, 0.0109, ..., 0.0622, -0.1204, -0.1173], + [ 0.0342, -0.0185, 0.0024, ..., -0.1303, 0.0460, -0.0983], + [-0.0028, -0.0048, 0.1362, ..., -0.1876, -0.0803, -0.1267], + ..., + [-0.0036, 0.0062, -0.0946, ..., 0.0008, 0.1094, 0.0601], + [ 0.0064, -0.0202, -0.0824, ..., -0.1650, 0.0670, -0.1296], + [-0.0065, -0.0258, -0.0492, ..., 0.0311, -0.0714, 0.0200]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -3.0082e-06, ..., -3.2708e-06, + -6.0536e-09, 1.4435e-08], + [ 4.0047e-08, 0.0000e+00, 3.1525e-07, ..., 2.3609e-07, + 8.9128e-07, 1.5832e-06], + [-1.3970e-09, 0.0000e+00, 3.9767e-07, ..., 1.4529e-07, + 9.2573e-07, 2.6403e-07], + ..., + [-3.0594e-07, 0.0000e+00, 1.9465e-07, ..., 7.2643e-08, + -4.1127e-06, -5.1968e-06], + [ 2.3283e-08, 0.0000e+00, -1.2480e-06, ..., 1.0524e-07, + -1.4994e-06, 2.9337e-07], + [ 4.0047e-08, 0.0000e+00, 1.7416e-07, ..., 2.3376e-07, + 9.7044e-07, 1.2778e-06]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0212, 0.0136, 0.0270, 0.0057, 0.0238, -0.0109, -0.0069, -0.0176, + -0.0171, 0.0213], device='cuda:0'), grad: tensor([-5.4315e-06, 3.0268e-06, 2.4550e-06, 6.2510e-06, 6.4494e-07, + -8.0559e-08, 3.6079e-06, -9.6112e-06, -3.5092e-06, 2.6561e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 220.77, cls_loss 0.0023 cls_loss_mapping 0.0032 cls_loss_causal 0.5123 re_mapping 0.0069 re_causal 0.0211 /// teacc 98.94 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0114, 0.0110, 0.0105, ..., 0.0621, -0.1209, -0.1184], + [ 0.0346, -0.0185, 0.0025, ..., -0.1309, 0.0462, -0.0979], + [-0.0024, -0.0048, 0.1361, ..., -0.1896, -0.0806, -0.1271], + ..., + [-0.0038, 0.0062, -0.0949, ..., 0.0004, 0.1098, 0.0602], + [ 0.0066, -0.0202, -0.0824, ..., -0.1662, 0.0674, -0.1304], + [-0.0064, -0.0258, -0.0489, ..., 0.0311, -0.0718, 0.0200]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.8231e-08, ..., -1.8952e-07, + 2.5425e-07, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 1.5460e-07, ..., 3.3062e-08, + -2.0396e-06, 3.7253e-08], + [ 0.0000e+00, 0.0000e+00, -1.4761e-06, ..., 2.1886e-08, + 1.0757e-07, 4.7032e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.9954e-07, ..., 7.4506e-09, + 4.3772e-08, -3.8743e-07], + [ 0.0000e+00, 0.0000e+00, 1.8766e-07, ..., 2.6543e-08, + 5.0664e-07, 2.0489e-08], + [-4.6566e-10, 0.0000e+00, 7.8697e-08, ..., 1.1642e-08, + 1.8673e-07, 1.3597e-07]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0216, 0.0139, 0.0265, 0.0053, 0.0238, -0.0102, -0.0069, -0.0174, + -0.0171, 0.0213], device='cuda:0'), grad: tensor([ 4.9546e-07, -3.5875e-06, -1.0692e-06, -1.7047e-05, 4.4797e-07, + 1.5646e-05, 1.0058e-06, 6.8266e-07, 2.6636e-06, 7.5717e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 220.62, cls_loss 0.0022 cls_loss_mapping 0.0034 cls_loss_causal 0.5317 re_mapping 0.0070 re_causal 0.0215 /// teacc 98.87 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 0.0113, 0.0110, 0.0093, ..., 0.0612, -0.1215, -0.1198], + [ 0.0346, -0.0185, 0.0024, ..., -0.1309, 0.0463, -0.0980], + [-0.0024, -0.0048, 0.1365, ..., -0.1903, -0.0807, -0.1270], + ..., + [-0.0042, 0.0062, -0.0956, ..., 0.0004, 0.1099, 0.0602], + [ 0.0066, -0.0202, -0.0825, ..., -0.1668, 0.0676, -0.1304], + [-0.0064, -0.0258, -0.0476, ..., 0.0317, -0.0723, 0.0200]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -1.4948e-07, ..., -1.0896e-07, + 9.2201e-08, 2.9150e-07], + [ 8.3819e-09, 0.0000e+00, -9.3132e-10, ..., 2.8731e-07, + -2.6543e-07, 5.8394e-07], + [ 2.7940e-09, 0.0000e+00, -2.2352e-08, ..., 8.6613e-08, + 1.8487e-07, 2.1094e-07], + ..., + [ 1.6764e-08, 0.0000e+00, 1.1642e-08, ..., 1.3709e-06, + -9.3132e-09, 1.2489e-06], + [ 1.8626e-09, 0.0000e+00, 3.4925e-08, ..., 7.4133e-07, + -5.1223e-08, 2.1495e-06], + [ 1.2992e-07, 0.0000e+00, 6.6590e-08, ..., 6.1840e-06, + -1.4305e-06, -2.0918e-06]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0221, 0.0139, 0.0266, 0.0049, 0.0239, -0.0097, -0.0071, -0.0176, + -0.0171, 0.0214], device='cuda:0'), grad: tensor([ 6.1793e-07, 3.8650e-08, 7.4692e-07, 5.7695e-07, -1.3649e-05, + 2.0005e-06, -3.4506e-07, -7.6648e-07, 1.5432e-06, 9.2834e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 221.41, cls_loss 0.0020 cls_loss_mapping 0.0041 cls_loss_causal 0.5443 re_mapping 0.0071 re_causal 0.0213 /// teacc 98.92 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0113, 0.0110, 0.0095, ..., 0.0612, -0.1224, -0.1201], + [ 0.0346, -0.0185, 0.0025, ..., -0.1316, 0.0460, -0.0975], + [-0.0024, -0.0048, 0.1369, ..., -0.1916, -0.0808, -0.1270], + ..., + [-0.0044, 0.0062, -0.0968, ..., 0.0002, 0.1102, 0.0600], + [ 0.0067, -0.0202, -0.0824, ..., -0.1678, 0.0682, -0.1302], + [-0.0063, -0.0258, -0.0478, ..., 0.0315, -0.0729, 0.0199]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8219e-07, ..., -1.4156e-07, + 2.7474e-08, 4.7032e-08], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 1.5367e-08, + 4.6566e-09, 6.7987e-08], + [ 0.0000e+00, 0.0000e+00, 2.0443e-07, ..., 4.0047e-08, + 1.1036e-07, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 1.8142e-06, + -1.5274e-07, 6.5565e-06], + [ 0.0000e+00, 0.0000e+00, -2.4177e-06, ..., 8.3353e-08, + -1.9353e-06, 1.0617e-07], + [ 6.5193e-09, 0.0000e+00, 6.4261e-08, ..., -2.1011e-06, + 1.2107e-07, -8.4043e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0223, 0.0139, 0.0266, 0.0046, 0.0241, -0.0095, -0.0066, -0.0177, + -0.0167, 0.0212], device='cuda:0'), grad: tensor([-1.6531e-07, 7.2643e-08, 7.6136e-07, 1.4305e-05, 3.3667e-07, + -8.6008e-07, 1.5553e-07, 7.6890e-06, -1.2770e-05, -9.4995e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 220.52, cls_loss 0.0033 cls_loss_mapping 0.0056 cls_loss_causal 0.5417 re_mapping 0.0074 re_causal 0.0208 /// teacc 98.93 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 0.0112, 0.0110, 0.0080, ..., 0.0594, -0.1230, -0.1206], + [ 0.0345, -0.0185, 0.0024, ..., -0.1326, 0.0454, -0.0977], + [-0.0025, -0.0048, 0.1369, ..., -0.1944, -0.0811, -0.1274], + ..., + [-0.0046, 0.0062, -0.0967, ..., 0.0006, 0.1120, 0.0606], + [ 0.0066, -0.0202, -0.0832, ..., -0.1716, 0.0677, -0.1313], + [-0.0060, -0.0258, -0.0463, ..., 0.0330, -0.0741, 0.0199]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -2.2259e-07, ..., 1.2526e-07, + 5.7742e-08, 5.3085e-07], + [ 5.9605e-08, 0.0000e+00, 5.8208e-08, ..., 7.7346e-07, + 6.9663e-06, 4.3772e-06], + [ 2.0489e-08, 0.0000e+00, -1.1344e-06, ..., 1.4622e-07, + 3.8072e-06, 2.2538e-06], + ..., + [ 8.2841e-07, 0.0000e+00, 7.8464e-07, ..., 5.0291e-06, + -1.7714e-06, 1.4409e-05], + [ 2.7474e-08, 0.0000e+00, 3.0734e-08, ..., 7.4040e-07, + -1.2890e-05, -2.9057e-06], + [ 3.6925e-05, 0.0000e+00, 1.5227e-07, ..., 1.6892e-04, + 5.5134e-07, 7.9441e-04]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0237, 0.0134, 0.0256, 0.0046, 0.0239, -0.0088, -0.0066, -0.0161, + -0.0180, 0.0218], device='cuda:0'), grad: tensor([ 8.7963e-07, 2.0579e-05, 9.5293e-06, 8.9854e-06, -6.7472e-04, + -1.4223e-05, 1.1981e-05, 1.0923e-05, -3.0607e-05, 6.5708e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 220.67, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5723 re_mapping 0.0072 re_causal 0.0227 /// teacc 98.91 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 1.1141e-02, 1.0998e-02, 8.6229e-03, ..., 5.9785e-02, + -1.2337e-01, -1.2086e-01], + [ 3.4576e-02, -1.8468e-02, 2.2358e-04, ..., -1.3378e-01, + 4.5108e-02, -9.6950e-02], + [-2.6007e-03, -4.7728e-03, 1.3831e-01, ..., -1.9711e-01, + -8.0896e-02, -1.2798e-01], + ..., + [-4.6043e-03, 6.2061e-03, -9.7910e-02, ..., 1.5238e-04, + 1.1205e-01, 6.0621e-02], + [ 6.3672e-03, -2.0220e-02, -8.2627e-02, ..., -1.7280e-01, + 6.8500e-02, -1.3159e-01], + [-6.4649e-03, -2.5834e-02, -4.6564e-02, ..., 3.2892e-02, + -7.4868e-02, 1.9866e-02]], device='cuda:0'), grad: tensor([[ 1.7695e-08, 0.0000e+00, 8.7824e-07, ..., -7.6136e-07, + 1.7276e-07, 1.4435e-08], + [ 7.4040e-08, 0.0000e+00, 1.2312e-06, ..., 2.0955e-08, + -1.6950e-07, 8.1025e-08], + [-3.2410e-07, 0.0000e+00, -3.4999e-06, ..., 1.0105e-07, + 1.9148e-06, 2.2352e-07], + ..., + [ 6.5658e-08, 0.0000e+00, 2.0675e-06, ..., 4.7497e-08, + -1.2666e-07, -2.8405e-07], + [ 9.9186e-08, 0.0000e+00, 4.1258e-07, ..., 1.3225e-07, + -2.8275e-06, 1.9418e-07], + [ 8.8476e-09, 0.0000e+00, 1.1958e-06, ..., 2.4959e-07, + 9.0338e-08, -5.4343e-07]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0234, 0.0126, 0.0262, 0.0045, 0.0241, -0.0091, -0.0068, -0.0164, + -0.0169, 0.0216], device='cuda:0'), grad: tensor([ 1.9467e-04, 1.2800e-05, 1.5333e-05, -7.6890e-05, 4.6007e-06, + -4.2129e-04, 7.3373e-05, 1.1569e-04, 4.6760e-05, 3.4541e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 220.69, cls_loss 0.0020 cls_loss_mapping 0.0040 cls_loss_causal 0.5369 re_mapping 0.0072 re_causal 0.0214 /// teacc 98.96 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 1.1064e-02, 1.0998e-02, 9.1262e-03, ..., 6.0165e-02, + -1.2436e-01, -1.2098e-01], + [ 3.4761e-02, -1.8468e-02, 1.4024e-05, ..., -1.3501e-01, + 4.5120e-02, -9.7150e-02], + [-2.6308e-03, -4.7728e-03, 1.3853e-01, ..., -1.9796e-01, + -8.1324e-02, -1.2807e-01], + ..., + [-4.6188e-03, 6.2061e-03, -9.8474e-02, ..., -3.7068e-04, + 1.1277e-01, 6.0596e-02], + [ 6.2731e-03, -2.0220e-02, -8.2426e-02, ..., -1.7344e-01, + 6.9333e-02, -1.3132e-01], + [-6.4774e-03, -2.5834e-02, -4.7011e-02, ..., 3.2750e-02, + -7.5662e-02, 1.9839e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.2608e-07, ..., -1.0347e-06, + 8.1025e-08, 6.4261e-08], + [ 0.0000e+00, 0.0000e+00, 4.5961e-07, ..., 4.0559e-07, + 1.8952e-07, 4.5635e-07], + [ 0.0000e+00, 0.0000e+00, -1.1936e-05, ..., 1.0058e-07, + 2.0023e-07, 1.5600e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5981e-06, ..., 7.3109e-08, + -6.1020e-06, -1.2599e-05], + [ 0.0000e+00, 0.0000e+00, -2.7269e-06, ..., 1.5646e-07, + -1.2979e-05, 5.9465e-07], + [ 0.0000e+00, 0.0000e+00, 7.5810e-07, ..., 5.2387e-07, + 5.4948e-06, 1.0662e-05]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0233, 0.0127, 0.0258, 0.0042, 0.0244, -0.0092, -0.0070, -0.0164, + -0.0161, 0.0213], device='cuda:0'), grad: tensor([-1.2945e-06, 2.0079e-06, -1.4827e-05, 7.0214e-05, 1.1874e-07, + 6.4149e-06, 2.1849e-06, -2.8908e-05, -6.4492e-05, 2.8461e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 220.92, cls_loss 0.0021 cls_loss_mapping 0.0039 cls_loss_causal 0.5382 re_mapping 0.0067 re_causal 0.0203 /// teacc 98.93 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 1.1034e-02, 1.0998e-02, 9.7667e-03, ..., 6.0284e-02, + -1.2541e-01, -1.2088e-01], + [ 3.4847e-02, -1.8468e-02, -4.7932e-05, ..., -1.3576e-01, + 4.5755e-02, -9.6648e-02], + [-2.6542e-03, -4.7728e-03, 1.3883e-01, ..., -1.9846e-01, + -8.1553e-02, -1.2815e-01], + ..., + [-4.6214e-03, 6.2061e-03, -9.8895e-02, ..., -7.1170e-04, + 1.1319e-01, 6.0720e-02], + [ 6.2128e-03, -2.0220e-02, -8.2668e-02, ..., -1.7444e-01, + 6.9444e-02, -1.3181e-01], + [-6.4842e-03, -2.5834e-02, -4.7248e-02, ..., 3.2429e-02, + -7.6542e-02, 1.9408e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.5157e-07, ..., -1.6531e-07, + 6.2352e-07, 4.0513e-08], + [ 0.0000e+00, 0.0000e+00, -8.9919e-07, ..., -3.7253e-07, + -2.5630e-06, 2.0191e-06], + [ 4.6566e-10, 0.0000e+00, -1.5840e-05, ..., 1.1269e-07, + 5.1456e-07, 6.0536e-08], + ..., + [ 4.6566e-10, 0.0000e+00, 1.0394e-05, ..., 1.4063e-07, + -9.9279e-07, -2.8200e-06], + [ 0.0000e+00, 0.0000e+00, 2.1122e-06, ..., 1.0431e-07, + 1.1493e-06, 6.1002e-08], + [ 4.6566e-10, 0.0000e+00, 5.1409e-07, ..., 2.6450e-06, + 5.6066e-07, 3.9712e-06]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0234, 0.0131, 0.0258, 0.0032, 0.0249, -0.0082, -0.0071, -0.0162, + -0.0161, 0.0207], device='cuda:0'), grad: tensor([ 3.7942e-06, -1.2077e-05, -2.2292e-05, -1.5721e-05, 1.4035e-06, + 1.9193e-05, -4.3735e-06, 1.3649e-05, 1.0513e-05, 5.9195e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 220.93, cls_loss 0.0026 cls_loss_mapping 0.0055 cls_loss_causal 0.5459 re_mapping 0.0066 re_causal 0.0198 /// teacc 98.82 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 0.0110, 0.0110, 0.0107, ..., 0.0606, -0.1254, -0.1213], + [ 0.0350, -0.0185, -0.0003, ..., -0.1371, 0.0463, -0.0971], + [-0.0005, -0.0048, 0.1398, ..., -0.1994, -0.0814, -0.1281], + ..., + [-0.0053, 0.0062, -0.1011, ..., -0.0009, 0.1135, 0.0610], + [ 0.0061, -0.0202, -0.0830, ..., -0.1771, 0.0694, -0.1333], + [-0.0065, -0.0258, -0.0471, ..., 0.0325, -0.0778, 0.0195]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.5239e-07, ..., -1.8813e-07, + 7.8557e-07, 3.8277e-07], + [ 0.0000e+00, 0.0000e+00, 2.4706e-05, ..., 8.0094e-08, + 4.6581e-05, 1.1362e-05], + [ 0.0000e+00, 0.0000e+00, -1.2648e-04, ..., 1.8626e-09, + 2.9821e-06, -2.7180e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 5.2154e-05, ..., -9.0338e-08, + -2.3067e-05, -2.0981e-05], + [ 0.0000e+00, 0.0000e+00, 1.2405e-06, ..., 2.5658e-07, + -5.1796e-05, 2.3004e-06], + [ 0.0000e+00, 0.0000e+00, 4.1604e-05, ..., -7.1712e-08, + 1.0356e-05, 2.8998e-05]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0231, 0.0132, 0.0263, 0.0031, 0.0248, -0.0083, -0.0070, -0.0165, + -0.0169, 0.0210], device='cuda:0'), grad: tensor([ 3.4999e-06, 1.6212e-04, -2.3794e-04, 1.3672e-05, 6.0461e-06, + 1.9878e-05, 7.8008e-06, 7.9691e-05, -1.4210e-04, 8.7142e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 220.80, cls_loss 0.0030 cls_loss_mapping 0.0056 cls_loss_causal 0.5348 re_mapping 0.0066 re_causal 0.0200 /// teacc 98.98 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0110, 0.0110, 0.0114, ..., 0.0608, -0.1260, -0.1221], + [ 0.0350, -0.0185, -0.0015, ..., -0.1368, 0.0444, -0.0972], + [-0.0005, -0.0048, 0.1405, ..., -0.2006, -0.0816, -0.1284], + ..., + [-0.0053, 0.0062, -0.1007, ..., -0.0007, 0.1167, 0.0632], + [ 0.0061, -0.0202, -0.0832, ..., -0.1788, 0.0696, -0.1338], + [-0.0065, -0.0258, -0.0469, ..., 0.0326, -0.0800, 0.0185]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.8359e-06, ..., -1.8841e-06, + 4.6729e-07, 4.3549e-06], + [ 0.0000e+00, 0.0000e+00, 3.7625e-07, ..., 1.3672e-06, + 2.2119e-07, 4.0010e-06], + [ 0.0000e+00, 0.0000e+00, 8.1817e-07, ..., 2.1793e-06, + 1.4193e-06, 5.5321e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3947e-07, ..., -4.9174e-07, + -6.3255e-06, -7.9721e-06], + [ 0.0000e+00, 0.0000e+00, 4.2329e-07, ..., 3.2177e-07, + 2.3376e-07, 5.2061e-07], + [ 0.0000e+00, 0.0000e+00, 1.8906e-07, ..., -6.0499e-06, + 2.6468e-06, -1.5251e-05]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0229, 0.0108, 0.0265, 0.0029, 0.0243, -0.0087, -0.0069, -0.0128, + -0.0170, 0.0198], device='cuda:0'), grad: tensor([-1.0416e-05, 6.4522e-06, 2.2396e-05, -1.3739e-05, 7.1153e-06, + 3.6918e-06, 1.4625e-05, -1.6868e-05, 2.9728e-06, -1.6317e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 220.81, cls_loss 0.0023 cls_loss_mapping 0.0032 cls_loss_causal 0.5385 re_mapping 0.0069 re_causal 0.0209 /// teacc 98.93 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0109, 0.0110, 0.0116, ..., 0.0607, -0.1266, -0.1235], + [ 0.0349, -0.0185, -0.0016, ..., -0.1372, 0.0441, -0.0980], + [-0.0004, -0.0048, 0.1408, ..., -0.2015, -0.0821, -0.1290], + ..., + [-0.0053, 0.0062, -0.1011, ..., -0.0017, 0.1172, 0.0632], + [ 0.0059, -0.0202, -0.0833, ..., -0.1795, 0.0698, -0.1340], + [-0.0065, -0.0258, -0.0471, ..., 0.0326, -0.0799, 0.0184]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 2.6869e-07, ..., 1.8962e-06, + 1.2573e-07, 2.2147e-06], + [ 0.0000e+00, 0.0000e+00, 7.8185e-07, ..., 2.0931e-07, + 1.3271e-07, 3.4459e-07], + [ 2.3283e-10, 0.0000e+00, -3.7611e-05, ..., 1.3714e-07, + -1.0245e-05, 2.1071e-07], + ..., + [ 2.3283e-10, 0.0000e+00, 3.4899e-05, ..., 3.1050e-06, + 9.7454e-06, 3.8818e-06], + [ 4.6566e-10, 0.0000e+00, 4.6776e-07, ..., 1.0780e-07, + -5.1223e-07, 3.5297e-07], + [ 4.6566e-10, 0.0000e+00, 1.4622e-07, ..., -9.0525e-06, + -1.0082e-07, -1.4603e-05]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0231, 0.0107, 0.0261, 0.0029, 0.0247, -0.0084, -0.0069, -0.0127, + -0.0169, 0.0198], device='cuda:0'), grad: tensor([ 5.1446e-06, 1.9725e-06, -7.6294e-05, 1.2055e-05, 2.6226e-06, + 9.3132e-09, 4.9919e-07, 7.9453e-05, -2.3493e-07, -2.5064e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 220.33, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5450 re_mapping 0.0066 re_causal 0.0205 /// teacc 98.97 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0109, 0.0110, 0.0115, ..., 0.0606, -0.1271, -0.1241], + [ 0.0349, -0.0185, -0.0015, ..., -0.1373, 0.0439, -0.0981], + [-0.0004, -0.0048, 0.1409, ..., -0.2020, -0.0828, -0.1295], + ..., + [-0.0053, 0.0062, -0.1015, ..., -0.0020, 0.1178, 0.0631], + [ 0.0059, -0.0202, -0.0833, ..., -0.1802, 0.0699, -0.1343], + [-0.0065, -0.0258, -0.0472, ..., 0.0327, -0.0799, 0.0187]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9697e-07, ..., 1.8813e-07, + 1.6093e-06, 2.4959e-06], + [ 4.6566e-10, 0.0000e+00, -1.3215e-06, ..., 1.2759e-07, + -1.2323e-05, 3.9302e-06], + [ 1.3970e-09, 0.0000e+00, -7.9721e-07, ..., 8.3819e-08, + 6.8620e-06, 1.7323e-06], + ..., + [ 1.8626e-09, 0.0000e+00, 1.5460e-06, ..., 2.7986e-07, + -7.6517e-06, -1.5810e-05], + [ 4.6566e-10, 0.0000e+00, -3.1479e-06, ..., 4.8848e-07, + -6.9402e-06, 4.8317e-06], + [ 4.6566e-10, 0.0000e+00, 3.0547e-07, ..., -1.3672e-06, + 1.2619e-06, -1.2748e-05]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0233, 0.0106, 0.0256, 0.0026, 0.0246, -0.0081, -0.0069, -0.0125, + -0.0169, 0.0200], device='cuda:0'), grad: tensor([ 7.6257e-06, -4.1038e-05, 2.1815e-05, 2.6733e-05, 2.8849e-05, + 1.3940e-05, 3.4794e-06, -2.9445e-05, -1.8820e-05, -1.3165e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 220.62, cls_loss 0.0023 cls_loss_mapping 0.0036 cls_loss_causal 0.5354 re_mapping 0.0062 re_causal 0.0192 /// teacc 98.85 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0108, 0.0110, 0.0107, ..., 0.0608, -0.1289, -0.1259], + [ 0.0351, -0.0185, -0.0016, ..., -0.1361, 0.0441, -0.0979], + [-0.0005, -0.0048, 0.1416, ..., -0.2026, -0.0835, -0.1299], + ..., + [-0.0053, 0.0062, -0.1023, ..., -0.0032, 0.1178, 0.0628], + [ 0.0059, -0.0202, -0.0832, ..., -0.1813, 0.0708, -0.1347], + [-0.0065, -0.0258, -0.0459, ..., 0.0324, -0.0804, 0.0186]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.0059e-07, ..., -8.4983e-07, + -4.6566e-10, 1.9092e-08], + [ 0.0000e+00, 0.0000e+00, 2.2398e-07, ..., 6.0536e-08, + -1.4855e-07, 1.5413e-07], + [ 0.0000e+00, 0.0000e+00, -5.5367e-07, ..., 9.5461e-08, + 2.8871e-07, 1.0431e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3109e-07, ..., 2.0536e-07, + -4.4098e-07, -6.3283e-07], + [ 0.0000e+00, 0.0000e+00, 5.7742e-08, ..., 3.8650e-08, + -6.1840e-07, 6.3796e-08], + [ 0.0000e+00, 0.0000e+00, 1.8021e-07, ..., 8.7591e-07, + 1.4575e-07, 1.3262e-06]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0241, 0.0107, 0.0255, 0.0029, 0.0251, -0.0086, -0.0064, -0.0128, + -0.0166, 0.0202], device='cuda:0'), grad: tensor([-1.0952e-06, 2.1420e-08, 4.6100e-08, 6.4494e-07, -1.1493e-06, + 1.0775e-06, 5.9279e-07, -2.4121e-07, -1.5637e-06, 1.6578e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 220.60, cls_loss 0.0027 cls_loss_mapping 0.0054 cls_loss_causal 0.5472 re_mapping 0.0067 re_causal 0.0205 /// teacc 99.03 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0107, 0.0110, 0.0109, ..., 0.0609, -0.1295, -0.1263], + [ 0.0352, -0.0185, -0.0016, ..., -0.1363, 0.0445, -0.0979], + [-0.0006, -0.0048, 0.1419, ..., -0.2033, -0.0839, -0.1303], + ..., + [-0.0053, 0.0062, -0.1029, ..., -0.0018, 0.1168, 0.0626], + [ 0.0058, -0.0202, -0.0833, ..., -0.1801, 0.0735, -0.1327], + [-0.0065, -0.0258, -0.0460, ..., 0.0318, -0.0824, 0.0184]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1222e-06, ..., -1.1735e-06, + 6.5053e-07, 5.7602e-07], + [ 0.0000e+00, 0.0000e+00, 6.7521e-08, ..., 1.8440e-07, + 8.6753e-07, 8.0653e-07], + [ 0.0000e+00, 0.0000e+00, -2.1420e-07, ..., 7.4459e-07, + 1.3839e-06, 1.6131e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5553e-07, ..., -8.8587e-06, + -7.0095e-05, -5.3197e-05], + [ 0.0000e+00, 0.0000e+00, 4.8354e-06, ..., 2.0023e-07, + 3.5996e-07, 5.9791e-07], + [ 0.0000e+00, 0.0000e+00, 7.8324e-07, ..., 7.0222e-06, + 5.4747e-05, 3.9458e-05]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0242, 0.0111, 0.0253, 0.0027, 0.0262, -0.0092, -0.0074, -0.0133, + -0.0144, 0.0197], device='cuda:0'), grad: tensor([-3.9022e-07, 2.0284e-06, 3.3621e-06, -1.1370e-05, 1.4260e-05, + 6.4261e-06, 1.8552e-06, -1.4269e-04, 1.3843e-05, 1.1283e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 220.68, cls_loss 0.0031 cls_loss_mapping 0.0039 cls_loss_causal 0.5062 re_mapping 0.0063 re_causal 0.0192 /// teacc 99.00 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0104, 0.0110, 0.0090, ..., 0.0591, -0.1304, -0.1286], + [ 0.0350, -0.0185, -0.0016, ..., -0.1367, 0.0447, -0.0981], + [-0.0011, -0.0048, 0.1423, ..., -0.2039, -0.0840, -0.1303], + ..., + [-0.0081, 0.0062, -0.1037, ..., -0.0014, 0.1168, 0.0624], + [ 0.0050, -0.0202, -0.0836, ..., -0.1809, 0.0725, -0.1334], + [-0.0066, -0.0258, -0.0441, ..., 0.0334, -0.0839, 0.0180]], + device='cuda:0'), grad: tensor([[ 3.5018e-07, 0.0000e+00, -3.6396e-06, ..., -6.3144e-06, + 6.1188e-07, 6.5612e-07], + [-1.2957e-05, 0.0000e+00, 4.1490e-07, ..., 3.2969e-07, + -1.8254e-05, 2.6636e-07], + [ 3.2187e-06, 0.0000e+00, -2.6841e-06, ..., 1.4435e-07, + 4.2468e-06, 6.6729e-07], + ..., + [ 5.6699e-06, 0.0000e+00, 1.9334e-06, ..., 1.3992e-05, + 1.4901e-05, 3.6955e-05], + [ 7.9069e-07, 0.0000e+00, 2.1514e-07, ..., 7.2923e-07, + 1.4100e-06, 2.7753e-06], + [ 3.3248e-07, 0.0000e+00, 5.0571e-07, ..., -1.6345e-06, + -9.2387e-06, 4.3690e-05]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0260, 0.0113, 0.0252, 0.0050, 0.0260, -0.0093, -0.0055, -0.0143, + -0.0160, 0.0201], device='cuda:0'), grad: tensor([-1.8463e-05, -7.6234e-05, 1.4775e-05, -1.9920e-04, 2.3365e-05, + 1.1057e-05, 1.6198e-05, 7.9751e-05, 1.5497e-05, 1.3292e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 220.62, cls_loss 0.0025 cls_loss_mapping 0.0042 cls_loss_causal 0.5309 re_mapping 0.0068 re_causal 0.0185 /// teacc 98.97 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0102, 0.0110, 0.0089, ..., 0.0589, -0.1315, -0.1294], + [ 0.0358, -0.0185, -0.0016, ..., -0.1378, 0.0456, -0.0958], + [-0.0014, -0.0048, 0.1427, ..., -0.2048, -0.0843, -0.1306], + ..., + [-0.0082, 0.0062, -0.1042, ..., -0.0019, 0.1162, 0.0616], + [ 0.0048, -0.0202, -0.0839, ..., -0.1796, 0.0737, -0.1311], + [-0.0067, -0.0258, -0.0437, ..., 0.0334, -0.0867, 0.0173]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -7.2876e-07, ..., -2.2352e-07, + 1.5181e-07, 1.7229e-08], + [-1.2573e-08, 0.0000e+00, 5.6811e-08, ..., 1.4948e-07, + -1.0729e-05, 2.0443e-07], + [ 1.8626e-09, 0.0000e+00, 6.4122e-07, ..., 3.5670e-07, + 8.6706e-07, 7.5437e-07], + ..., + [ 4.1910e-09, 0.0000e+00, 1.7975e-07, ..., -3.4273e-07, + 8.6129e-06, -1.1493e-06], + [ 1.8626e-09, 0.0000e+00, 4.2189e-07, ..., 1.2061e-07, + -1.3039e-08, 1.3644e-07], + [ 5.5879e-09, 0.0000e+00, 1.9418e-07, ..., 4.6054e-07, + 4.0838e-07, 3.6508e-07]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0264, 0.0124, 0.0251, 0.0052, 0.0258, -0.0092, -0.0046, -0.0151, + -0.0150, 0.0191], device='cuda:0'), grad: tensor([ 1.0561e-06, -1.9908e-05, 3.7886e-06, -3.1516e-06, 1.7053e-06, + 8.8476e-08, -3.7085e-06, 1.6034e-05, 2.4773e-06, 1.6242e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 220.44, cls_loss 0.0028 cls_loss_mapping 0.0047 cls_loss_causal 0.5321 re_mapping 0.0064 re_causal 0.0195 /// teacc 98.98 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0094, 0.0110, 0.0099, ..., 0.0592, -0.1330, -0.1295], + [ 0.0366, -0.0185, -0.0031, ..., -0.1380, 0.0443, -0.0955], + [-0.0019, -0.0048, 0.1443, ..., -0.2063, -0.0823, -0.1311], + ..., + [-0.0085, 0.0062, -0.1046, ..., -0.0019, 0.1162, 0.0612], + [ 0.0042, -0.0202, -0.0841, ..., -0.1803, 0.0737, -0.1314], + [-0.0059, -0.0258, -0.0437, ..., 0.0334, -0.0869, 0.0176]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9446e-06, ..., -5.1782e-06, + 4.9826e-08, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, 6.0722e-07, ..., 8.8010e-08, + -4.9081e-07, 5.2387e-07], + [ 0.0000e+00, 0.0000e+00, -7.0184e-06, ..., 4.4145e-07, + 4.4564e-07, 1.8580e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4168e-07, ..., 1.6484e-07, + -9.5135e-07, -1.4016e-06], + [ 0.0000e+00, 0.0000e+00, 5.4277e-06, ..., 5.2573e-07, + 2.9244e-07, 1.5851e-06], + [ 0.0000e+00, 0.0000e+00, 4.0643e-06, ..., 3.1870e-06, + 3.1851e-07, -1.3793e-06]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0259, 0.0120, 0.0260, 0.0050, 0.0256, -0.0086, -0.0049, -0.0155, + -0.0153, 0.0192], device='cuda:0'), grad: tensor([ 6.3628e-06, -1.0207e-06, -9.2015e-06, 9.2238e-06, 6.2492e-07, + -3.0026e-05, 3.9861e-06, -8.9733e-07, 1.2837e-05, 8.0764e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 220.49, cls_loss 0.0024 cls_loss_mapping 0.0050 cls_loss_causal 0.5330 re_mapping 0.0065 re_causal 0.0192 /// teacc 98.93 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0094, 0.0110, 0.0117, ..., 0.0603, -0.1333, -0.1296], + [ 0.0365, -0.0185, -0.0032, ..., -0.1388, 0.0443, -0.0954], + [-0.0018, -0.0048, 0.1446, ..., -0.2077, -0.0821, -0.1314], + ..., + [-0.0085, 0.0062, -0.1050, ..., -0.0024, 0.1163, 0.0609], + [ 0.0041, -0.0202, -0.0845, ..., -0.1811, 0.0736, -0.1317], + [-0.0059, -0.0258, -0.0449, ..., 0.0322, -0.0870, 0.0172]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.9716e-06, ..., -8.7321e-06, + 2.8182e-06, 4.7032e-07], + [ 0.0000e+00, 0.0000e+00, -1.3085e-07, ..., 1.0803e-07, + 1.6345e-07, 4.6985e-07], + [ 0.0000e+00, 0.0000e+00, 9.8255e-08, ..., 2.9150e-07, + 7.1060e-07, 1.3206e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1548e-07, ..., 1.7742e-07, + -1.7434e-06, -2.4773e-06], + [ 0.0000e+00, 0.0000e+00, 9.8720e-08, ..., -4.4797e-07, + 1.4886e-05, -3.0883e-06], + [ 0.0000e+00, 0.0000e+00, 2.3842e-07, ..., 1.9092e-07, + 1.4845e-06, 1.7565e-06]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0248, 0.0120, 0.0260, 0.0052, 0.0284, -0.0076, -0.0078, -0.0157, + -0.0158, 0.0183], device='cuda:0'), grad: tensor([-1.4082e-05, 7.9628e-07, 4.3064e-06, 2.7176e-06, 1.3737e-06, + 2.2173e-05, -5.7280e-05, -3.2075e-06, 3.8952e-05, 4.2915e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 220.70, cls_loss 0.0021 cls_loss_mapping 0.0031 cls_loss_causal 0.5050 re_mapping 0.0065 re_causal 0.0183 /// teacc 98.97 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0094, 0.0110, 0.0111, ..., 0.0602, -0.1348, -0.1298], + [ 0.0367, -0.0185, -0.0032, ..., -0.1395, 0.0447, -0.0956], + [-0.0019, -0.0048, 0.1447, ..., -0.2088, -0.0824, -0.1317], + ..., + [-0.0085, 0.0062, -0.1057, ..., -0.0028, 0.1166, 0.0609], + [ 0.0040, -0.0202, -0.0847, ..., -0.1818, 0.0732, -0.1320], + [-0.0059, -0.0258, -0.0450, ..., 0.0317, -0.0874, 0.0170]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3568e-05, ..., -9.1270e-08, + 3.6741e-07, 3.4319e-07], + [ 0.0000e+00, 0.0000e+00, 6.5900e-06, ..., 1.5348e-06, + 1.4799e-06, 3.2242e-06], + [ 0.0000e+00, 0.0000e+00, -9.4712e-05, ..., 1.1222e-07, + -5.5581e-06, -8.5542e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 6.2048e-05, ..., 4.8839e-06, + 3.5539e-06, 9.2834e-06], + [ 0.0000e+00, 0.0000e+00, 6.9998e-06, ..., 1.1502e-07, + 1.4454e-06, 2.7753e-07], + [ 0.0000e+00, 0.0000e+00, 1.6950e-06, ..., 6.6087e-06, + 2.5947e-06, 1.2934e-05]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0255, 0.0125, 0.0259, 0.0052, 0.0290, -0.0079, -0.0070, -0.0156, + -0.0167, 0.0179], device='cuda:0'), grad: tensor([ 2.2069e-05, 1.4782e-05, -1.4436e-04, -1.3784e-06, -2.3559e-05, + 3.2801e-06, -1.0945e-05, 1.0538e-04, 1.4648e-05, 2.0146e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 220.80, cls_loss 0.0021 cls_loss_mapping 0.0036 cls_loss_causal 0.5408 re_mapping 0.0065 re_causal 0.0197 /// teacc 98.96 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0094, 0.0110, 0.0114, ..., 0.0604, -0.1363, -0.1300], + [ 0.0367, -0.0185, -0.0032, ..., -0.1399, 0.0449, -0.0957], + [-0.0019, -0.0048, 0.1449, ..., -0.2098, -0.0827, -0.1313], + ..., + [-0.0085, 0.0062, -0.1065, ..., -0.0019, 0.1169, 0.0612], + [ 0.0040, -0.0202, -0.0849, ..., -0.1823, 0.0733, -0.1322], + [-0.0059, -0.0258, -0.0452, ..., 0.0316, -0.0879, 0.0170]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.2201e-08, ..., 1.1083e-07, + 4.9826e-08, 2.7986e-07], + [ 3.3062e-08, 0.0000e+00, 1.5087e-07, ..., 1.4249e-07, + -3.5437e-07, 9.7603e-07], + [-4.6566e-09, 0.0000e+00, 1.5832e-07, ..., 6.1467e-08, + 2.8685e-06, 2.1271e-06], + ..., + [-9.4529e-08, 0.0000e+00, 3.6042e-07, ..., 2.6869e-07, + -4.5672e-06, -3.4515e-06], + [ 1.4435e-08, 0.0000e+00, 1.8859e-07, ..., 1.8440e-06, + -3.3993e-08, 4.8466e-06], + [ 1.8161e-08, 0.0000e+00, 4.6659e-07, ..., -9.3043e-05, + 8.2608e-07, -2.1076e-04]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0254, 0.0126, 0.0258, 0.0052, 0.0287, -0.0083, -0.0065, -0.0155, + -0.0167, 0.0178], device='cuda:0'), grad: tensor([ 7.1339e-07, 6.1514e-07, 1.0900e-05, -2.7437e-06, 2.5296e-04, + 8.8066e-06, 1.0971e-06, -1.1459e-05, 5.9344e-06, -2.6679e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 220.62, cls_loss 0.0027 cls_loss_mapping 0.0052 cls_loss_causal 0.5419 re_mapping 0.0062 re_causal 0.0192 /// teacc 98.97 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0093, 0.0110, 0.0117, ..., 0.0602, -0.1395, -0.1302], + [ 0.0384, -0.0185, -0.0034, ..., -0.1413, 0.0456, -0.0943], + [-0.0027, -0.0048, 0.1453, ..., -0.2107, -0.0830, -0.1316], + ..., + [-0.0086, 0.0062, -0.1068, ..., -0.0020, 0.1168, 0.0609], + [ 0.0039, -0.0202, -0.0851, ..., -0.1834, 0.0737, -0.1327], + [-0.0059, -0.0258, -0.0455, ..., 0.0311, -0.0884, 0.0165]], + device='cuda:0'), grad: tensor([[ 2.0349e-07, 0.0000e+00, 1.0105e-07, ..., 7.7765e-08, + 1.0421e-06, 8.9966e-07], + [-1.0508e-04, 0.0000e+00, -5.2512e-05, ..., 4.0978e-08, + -2.0409e-04, 2.4419e-06], + [ 8.8811e-05, 0.0000e+00, 4.4286e-05, ..., 3.0035e-07, + 1.7941e-04, 3.3751e-06], + ..., + [ 1.0319e-05, 0.0000e+00, 5.6885e-06, ..., -8.7498e-07, + 4.5486e-06, -2.3469e-05], + [ 7.8697e-07, 0.0000e+00, 3.4506e-07, ..., 1.4622e-07, + -2.5611e-07, 3.2336e-06], + [ 3.4086e-07, 0.0000e+00, 6.5193e-08, ..., -1.0198e-07, + 3.6508e-06, 3.1479e-06]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0257, 0.0133, 0.0258, 0.0051, 0.0291, -0.0089, -0.0062, -0.0159, + -0.0164, 0.0173], device='cuda:0'), grad: tensor([ 3.3807e-06, -4.2820e-04, 3.8958e-04, 1.2554e-05, 7.8157e-06, + 5.0813e-06, 3.1497e-06, -1.1899e-05, -4.5523e-06, 2.3529e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 170---------------------------------------------------- +epoch 170, time 221.15, cls_loss 0.0017 cls_loss_mapping 0.0031 cls_loss_causal 0.5458 re_mapping 0.0064 re_causal 0.0200 /// teacc 99.05 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0092, 0.0110, 0.0122, ..., 0.0604, -0.1402, -0.1304], + [ 0.0415, -0.0185, -0.0034, ..., -0.1420, 0.0457, -0.0942], + [-0.0050, -0.0048, 0.1454, ..., -0.2119, -0.0834, -0.1320], + ..., + [-0.0085, 0.0062, -0.1070, ..., -0.0015, 0.1175, 0.0611], + [ 0.0037, -0.0202, -0.0852, ..., -0.1841, 0.0736, -0.1330], + [-0.0059, -0.0258, -0.0457, ..., 0.0311, -0.0888, 0.0166]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.4925e-07, ..., 5.2620e-08, + 3.2596e-07, 8.9407e-08], + [ 0.0000e+00, 0.0000e+00, -2.0236e-05, ..., 3.4459e-08, + -2.9907e-05, 1.2061e-07], + [ 0.0000e+00, 0.0000e+00, 2.1905e-05, ..., 3.3993e-08, + 2.6107e-05, 6.1933e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 6.6534e-06, ..., 3.7765e-07, + 1.4585e-06, 3.6554e-07], + [ 0.0000e+00, 0.0000e+00, 1.2927e-05, ..., 4.2841e-08, + 5.5786e-07, 1.0803e-07], + [ 0.0000e+00, 0.0000e+00, 1.0040e-06, ..., -4.2375e-07, + 1.2051e-06, -1.0105e-06]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0256, 0.0133, 0.0256, 0.0051, 0.0288, -0.0089, -0.0061, -0.0155, + -0.0166, 0.0173], device='cuda:0'), grad: tensor([ 2.4941e-06, -1.1671e-04, 1.2231e-04, -1.1224e-04, 3.1926e-06, + 1.1794e-05, -3.2280e-06, 3.0756e-05, 5.7071e-05, 4.6492e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 220.70, cls_loss 0.0018 cls_loss_mapping 0.0043 cls_loss_causal 0.5283 re_mapping 0.0062 re_causal 0.0191 /// teacc 98.97 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0091, 0.0110, 0.0127, ..., 0.0607, -0.1406, -0.1305], + [ 0.0416, -0.0185, -0.0034, ..., -0.1431, 0.0458, -0.0943], + [-0.0051, -0.0048, 0.1455, ..., -0.2133, -0.0836, -0.1324], + ..., + [-0.0086, 0.0062, -0.1074, ..., -0.0016, 0.1176, 0.0610], + [ 0.0036, -0.0202, -0.0857, ..., -0.1853, 0.0747, -0.1334], + [-0.0059, -0.0258, -0.0458, ..., 0.0310, -0.0888, 0.0167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9092e-08, ..., -7.7300e-08, + 3.4273e-07, 6.8918e-08], + [ 1.3970e-09, 0.0000e+00, -7.1246e-07, ..., 1.1642e-07, + -8.4788e-06, 1.9418e-07], + [-3.2596e-09, 0.0000e+00, -3.6927e-07, ..., 7.8697e-08, + 7.1898e-07, 1.0338e-07], + ..., + [ 2.7940e-09, 0.0000e+00, 4.5123e-07, ..., 8.3912e-07, + 1.7416e-06, 1.6503e-06], + [ 1.8626e-09, 0.0000e+00, 1.3178e-07, ..., 1.5879e-07, + 4.5495e-07, 2.9802e-07], + [ 6.0536e-09, 0.0000e+00, 1.1921e-07, ..., -1.7677e-06, + 1.1623e-06, -5.0776e-06]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0255, 0.0134, 0.0254, 0.0051, 0.0290, -0.0093, -0.0070, -0.0155, + -0.0158, 0.0174], device='cuda:0'), grad: tensor([ 2.6338e-06, -6.6638e-05, 5.3123e-06, 1.9208e-05, 7.1749e-06, + 4.7684e-06, 3.0473e-06, 1.6481e-05, 4.0978e-06, 3.8780e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 220.77, cls_loss 0.0021 cls_loss_mapping 0.0045 cls_loss_causal 0.5210 re_mapping 0.0064 re_causal 0.0189 /// teacc 98.96 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0091, 0.0110, 0.0130, ..., 0.0607, -0.1410, -0.1317], + [ 0.0426, -0.0185, -0.0034, ..., -0.1461, 0.0460, -0.0948], + [-0.0059, -0.0048, 0.1459, ..., -0.2150, -0.0838, -0.1326], + ..., + [-0.0086, 0.0062, -0.1090, ..., -0.0021, 0.1178, 0.0610], + [ 0.0034, -0.0202, -0.0865, ..., -0.1862, 0.0749, -0.1337], + [-0.0059, -0.0258, -0.0460, ..., 0.0313, -0.0888, 0.0171]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -2.7940e-08, ..., 8.1770e-07, + 5.4250e-07, 1.7928e-06], + [-3.0268e-08, 0.0000e+00, -2.2864e-07, ..., 4.8522e-07, + -7.0920e-07, 1.0701e-06], + [ 1.3504e-08, 0.0000e+00, -2.1420e-07, ..., 4.1118e-07, + 7.4273e-07, 8.6334e-07], + ..., + [ 5.5879e-09, 0.0000e+00, 8.8010e-08, ..., 2.2873e-06, + 1.0943e-07, 2.5816e-06], + [ 1.8626e-09, 0.0000e+00, 6.5658e-08, ..., 3.5346e-05, + 9.0078e-06, 5.6118e-05], + [ 4.6566e-10, 0.0000e+00, 2.7940e-08, ..., -1.0777e-04, + -2.7895e-05, -1.7095e-04]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0257, 0.0135, 0.0256, 0.0051, 0.0290, -0.0088, -0.0076, -0.0156, + -0.0160, 0.0177], device='cuda:0'), grad: tensor([ 4.6045e-06, -1.1232e-06, 2.8815e-06, 1.7192e-06, 2.2280e-04, + 7.4245e-06, 1.3169e-06, 6.8285e-06, 1.1909e-04, -3.6526e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 220.85, cls_loss 0.0024 cls_loss_mapping 0.0053 cls_loss_causal 0.5519 re_mapping 0.0064 re_causal 0.0192 /// teacc 99.04 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0090, 0.0110, 0.0129, ..., 0.0609, -0.1420, -0.1319], + [ 0.0423, -0.0185, -0.0036, ..., -0.1472, 0.0461, -0.0942], + [-0.0051, -0.0048, 0.1466, ..., -0.2160, -0.0840, -0.1334], + ..., + [-0.0086, 0.0062, -0.1099, ..., -0.0023, 0.1186, 0.0607], + [ 0.0032, -0.0202, -0.0874, ..., -0.1868, 0.0754, -0.1338], + [-0.0059, -0.0258, -0.0461, ..., 0.0314, -0.0899, 0.0175]], + device='cuda:0'), grad: tensor([[ 4.9360e-08, 0.0000e+00, 6.8452e-08, ..., 1.5553e-07, + 1.5227e-07, 3.4412e-07], + [ 1.1586e-06, 0.0000e+00, 3.6806e-06, ..., 1.2107e-06, + 8.1398e-07, 1.2564e-06], + [-3.8669e-06, 0.0000e+00, -2.1949e-05, ..., 2.5425e-07, + -2.1253e-06, 2.3097e-07], + ..., + [ 1.1101e-06, 0.0000e+00, 1.1988e-05, ..., 5.8450e-06, + 1.1055e-06, 5.8226e-06], + [ 5.4482e-07, 0.0000e+00, 2.0117e-06, ..., 4.1649e-06, + 8.3400e-07, 3.6638e-06], + [ 2.0489e-08, 0.0000e+00, 2.2771e-07, ..., -1.7136e-05, + -1.9874e-06, -1.6913e-05]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0259, 0.0135, 0.0257, 0.0045, 0.0287, -0.0081, -0.0077, -0.0155, + -0.0159, 0.0176], device='cuda:0'), grad: tensor([ 1.5805e-06, 8.7172e-06, -3.5346e-05, -5.3310e-04, 1.9511e-07, + 5.4836e-04, 3.2783e-06, 3.3379e-05, 1.7196e-05, -4.4137e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 220.72, cls_loss 0.0027 cls_loss_mapping 0.0055 cls_loss_causal 0.5582 re_mapping 0.0063 re_causal 0.0186 /// teacc 99.02 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0088, 0.0110, 0.0132, ..., 0.0605, -0.1457, -0.1321], + [ 0.0434, -0.0185, -0.0038, ..., -0.1502, 0.0460, -0.0946], + [-0.0054, -0.0048, 0.1475, ..., -0.2180, -0.0840, -0.1338], + ..., + [-0.0088, 0.0062, -0.1116, ..., -0.0025, 0.1191, 0.0609], + [ 0.0031, -0.0202, -0.0876, ..., -0.1879, 0.0754, -0.1341], + [-0.0061, -0.0258, -0.0463, ..., 0.0338, -0.0877, 0.0198]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -5.3383e-06, ..., -9.0897e-06, + -2.4913e-07, 9.2667e-08], + [ 1.2573e-07, 0.0000e+00, 4.1686e-06, ..., 6.6822e-07, + 1.6261e-06, 8.6287e-07], + [-1.9791e-07, 0.0000e+00, -6.2808e-06, ..., 1.1306e-06, + -6.0443e-07, 1.1325e-06], + ..., + [ 4.3120e-07, 0.0000e+00, 2.1588e-06, ..., -1.3553e-05, + -2.2963e-05, -2.4229e-05], + [ 5.3970e-07, 0.0000e+00, 1.9521e-06, ..., 3.1106e-06, + 1.9409e-06, 3.3472e-06], + [-1.0300e-06, 0.0000e+00, 6.0163e-07, ..., -3.2410e-07, + -7.7393e-07, -6.9737e-06]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0270, 0.0133, 0.0261, 0.0041, 0.0263, -0.0099, -0.0047, -0.0154, + -0.0161, 0.0197], device='cuda:0'), grad: tensor([-1.9133e-05, 8.5011e-06, -6.3814e-06, 4.9233e-05, 8.0988e-06, + -1.2636e-05, 6.9551e-06, -4.8846e-05, 2.0504e-05, -6.1952e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 220.97, cls_loss 0.0024 cls_loss_mapping 0.0042 cls_loss_causal 0.5241 re_mapping 0.0067 re_causal 0.0191 /// teacc 98.98 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0086, 0.0110, 0.0134, ..., 0.0603, -0.1462, -0.1323], + [ 0.0446, -0.0185, -0.0038, ..., -0.1505, 0.0468, -0.0947], + [-0.0054, -0.0048, 0.1478, ..., -0.2195, -0.0844, -0.1339], + ..., + [-0.0090, 0.0062, -0.1125, ..., -0.0029, 0.1187, 0.0609], + [ 0.0029, -0.0202, -0.0871, ..., -0.1872, 0.0757, -0.1344], + [-0.0061, -0.0258, -0.0460, ..., 0.0348, -0.0878, 0.0206]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2726e-05, ..., -1.4573e-05, + -4.5858e-06, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-08, ..., 9.0804e-08, + -3.1618e-07, 3.4925e-08], + [-4.6566e-10, 0.0000e+00, -1.5132e-05, ..., 4.0047e-08, + -4.7050e-06, 5.1223e-09], + ..., + [-4.6566e-10, 0.0000e+00, 1.1045e-06, ..., 3.1386e-07, + 4.2422e-07, 4.0513e-07], + [ 4.6566e-10, 0.0000e+00, 1.4141e-05, ..., 4.0932e-07, + 6.0424e-06, 1.7881e-07], + [ 9.3132e-10, 0.0000e+00, 1.8859e-07, ..., -1.7248e-06, + 1.2759e-07, -5.1484e-06]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0275, 0.0136, 0.0261, 0.0038, 0.0255, -0.0097, -0.0052, -0.0158, + -0.0152, 0.0205], device='cuda:0'), grad: tensor([-6.7651e-05, -4.2375e-08, -2.1964e-05, 3.3349e-05, 9.7901e-06, + -4.5717e-05, 5.7340e-05, 3.4124e-06, 3.4869e-05, -3.5278e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 220.81, cls_loss 0.0021 cls_loss_mapping 0.0046 cls_loss_causal 0.5389 re_mapping 0.0066 re_causal 0.0191 /// teacc 98.98 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0086, 0.0110, 0.0148, ..., 0.0610, -0.1466, -0.1328], + [ 0.0452, -0.0185, -0.0039, ..., -0.1519, 0.0469, -0.0947], + [-0.0059, -0.0048, 0.1483, ..., -0.2203, -0.0844, -0.1337], + ..., + [-0.0091, 0.0062, -0.1135, ..., -0.0038, 0.1188, 0.0609], + [ 0.0028, -0.0202, -0.0881, ..., -0.1879, 0.0760, -0.1346], + [-0.0062, -0.0258, -0.0466, ..., 0.0345, -0.0880, 0.0205]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0338e-07, ..., -5.5181e-07, + 1.0803e-07, 5.3411e-07], + [ 0.0000e+00, 0.0000e+00, 9.1735e-07, ..., 2.1886e-08, + -3.9442e-07, 3.4589e-06], + [ 0.0000e+00, 0.0000e+00, -1.7211e-06, ..., 3.5856e-08, + 1.2778e-06, 6.9067e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6345e-06, ..., 9.9186e-08, + 1.7583e-06, 2.8387e-05], + [ 0.0000e+00, 0.0000e+00, 2.1681e-06, ..., 1.8021e-07, + 8.6101e-07, 6.2510e-06], + [ 0.0000e+00, 0.0000e+00, 9.9093e-07, ..., -7.4040e-08, + -4.3958e-06, -4.9889e-05]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0270, 0.0137, 0.0262, 0.0040, 0.0256, -0.0098, -0.0053, -0.0160, + -0.0154, 0.0202], device='cuda:0'), grad: tensor([ 7.9116e-07, 6.3255e-06, 1.5229e-05, -9.3281e-06, 3.2857e-06, + 2.3209e-06, 8.7265e-07, 6.0827e-05, 1.7568e-05, -9.7871e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 220.49, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5303 re_mapping 0.0062 re_causal 0.0189 /// teacc 98.98 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0086, 0.0110, 0.0134, ..., 0.0597, -0.1468, -0.1330], + [ 0.0452, -0.0185, -0.0038, ..., -0.1526, 0.0468, -0.0950], + [-0.0059, -0.0048, 0.1488, ..., -0.2210, -0.0845, -0.1331], + ..., + [-0.0091, 0.0062, -0.1159, ..., -0.0046, 0.1191, 0.0608], + [ 0.0028, -0.0202, -0.0887, ..., -0.1886, 0.0760, -0.1349], + [-0.0062, -0.0258, -0.0454, ..., 0.0352, -0.0880, 0.0205]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.6252e-06, ..., -7.8455e-06, + 1.6065e-07, 2.7940e-08], + [ 0.0000e+00, 0.0000e+00, 3.2689e-07, ..., 2.5658e-07, + -8.9407e-08, 6.7055e-07], + [ 0.0000e+00, 0.0000e+00, -1.6317e-06, ..., 2.6450e-07, + 2.7521e-07, 1.8626e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.4005e-07, ..., 1.1288e-06, + -5.0431e-07, 2.0657e-06], + [ 0.0000e+00, 0.0000e+00, 6.7754e-07, ..., 3.4040e-07, + 2.4680e-07, 2.0536e-07], + [ 0.0000e+00, 0.0000e+00, 4.5076e-06, ..., 3.9488e-06, + -2.7753e-07, -6.6198e-06]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0286, 0.0138, 0.0265, 0.0041, 0.0258, -0.0097, -0.0053, -0.0162, + -0.0157, 0.0208], device='cuda:0'), grad: tensor([-1.4059e-05, 7.9395e-07, -1.1623e-06, 2.1178e-06, 3.5278e-06, + 1.2740e-06, -2.2687e-06, 2.3358e-06, 2.6952e-06, 4.7348e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 220.68, cls_loss 0.0021 cls_loss_mapping 0.0036 cls_loss_causal 0.5515 re_mapping 0.0059 re_causal 0.0187 /// teacc 98.83 lr 0.00010000 +Epoch 180, weight, value: tensor([[ 0.0086, 0.0110, 0.0144, ..., 0.0602, -0.1470, -0.1331], + [ 0.0452, -0.0185, -0.0040, ..., -0.1539, 0.0473, -0.0950], + [-0.0060, -0.0048, 0.1500, ..., -0.2224, -0.0849, -0.1339], + ..., + [-0.0092, 0.0062, -0.1168, ..., -0.0045, 0.1191, 0.0613], + [ 0.0027, -0.0202, -0.0889, ..., -0.1895, 0.0763, -0.1347], + [-0.0063, -0.0258, -0.0462, ..., 0.0345, -0.0886, 0.0202]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.9651e-07, ..., 1.3551e-07, + 6.4261e-08, 8.3819e-07], + [ 9.7789e-09, 0.0000e+00, 3.6675e-06, ..., 8.7842e-06, + -6.5519e-07, 1.0140e-05], + [ 1.2852e-07, 0.0000e+00, 2.2911e-06, ..., 4.9882e-06, + 1.1465e-06, 6.7316e-06], + ..., + [-1.9930e-07, 0.0000e+00, 1.3877e-07, ..., 5.2061e-07, + -7.7952e-07, -3.4459e-08], + [ 1.6298e-08, 0.0000e+00, 1.5600e-07, ..., 3.5763e-07, + -1.5208e-06, 9.1689e-07], + [ 4.1910e-09, 0.0000e+00, 2.7660e-06, ..., 5.2117e-06, + 1.7229e-07, 5.0105e-06]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0281, 0.0142, 0.0266, 0.0036, 0.0259, -0.0094, -0.0050, -0.0162, + -0.0156, 0.0197], device='cuda:0'), grad: tensor([ 6.0815e-07, 1.5408e-05, 1.5408e-05, 1.1042e-05, -5.6446e-05, + 9.4920e-06, 1.5885e-05, -3.2829e-07, -2.1204e-05, 1.0066e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 179---------------------------------------------------- +epoch 179, time 221.55, cls_loss 0.0032 cls_loss_mapping 0.0045 cls_loss_causal 0.5688 re_mapping 0.0065 re_causal 0.0182 /// teacc 99.06 lr 0.00010000 +Epoch 181, weight, value: tensor([[ 0.0086, 0.0110, 0.0143, ..., 0.0604, -0.1482, -0.1335], + [ 0.0454, -0.0185, -0.0041, ..., -0.1554, 0.0475, -0.0947], + [-0.0061, -0.0048, 0.1508, ..., -0.2234, -0.0852, -0.1327], + ..., + [-0.0092, 0.0062, -0.1173, ..., -0.0055, 0.1192, 0.0596], + [ 0.0027, -0.0202, -0.0893, ..., -0.1911, 0.0777, -0.1353], + [-0.0063, -0.0258, -0.0464, ..., 0.0341, -0.0886, 0.0195]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9600e-06, ..., -6.1877e-06, + 6.8918e-08, -9.9838e-07], + [ 0.0000e+00, 0.0000e+00, -5.2005e-06, ..., 1.5460e-07, + -1.3895e-05, 2.0070e-07], + [ 0.0000e+00, 0.0000e+00, 1.0245e-06, ..., 4.8568e-07, + 3.9376e-06, 1.4249e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 2.6058e-06, ..., 8.1025e-08, + 4.7684e-06, -7.6508e-07], + [ 0.0000e+00, 0.0000e+00, 1.3728e-06, ..., 8.1770e-07, + 4.3027e-07, 1.8999e-07], + [ 0.0000e+00, 0.0000e+00, 6.0443e-07, ..., 6.1467e-07, + 5.9884e-07, 4.7637e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0287, 0.0144, 0.0268, 0.0039, 0.0276, -0.0105, -0.0058, -0.0173, + -0.0145, 0.0194], device='cuda:0'), grad: tensor([-1.2353e-05, -3.8743e-05, 1.0967e-05, 1.4648e-05, 1.8617e-06, + -1.2599e-05, 9.9093e-06, 1.4298e-05, 8.3894e-06, 3.5726e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 220.74, cls_loss 0.0018 cls_loss_mapping 0.0043 cls_loss_causal 0.4770 re_mapping 0.0062 re_causal 0.0178 /// teacc 98.98 lr 0.00010000 +Epoch 182, weight, value: tensor([[ 0.0086, 0.0110, 0.0149, ..., 0.0609, -0.1485, -0.1337], + [ 0.0453, -0.0185, -0.0043, ..., -0.1557, 0.0478, -0.0948], + [-0.0061, -0.0048, 0.1510, ..., -0.2257, -0.0854, -0.1338], + ..., + [-0.0092, 0.0062, -0.1175, ..., -0.0029, 0.1199, 0.0607], + [ 0.0027, -0.0202, -0.0895, ..., -0.1922, 0.0776, -0.1357], + [-0.0063, -0.0258, -0.0467, ..., 0.0337, -0.0892, 0.0192]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., -8.0094e-08, + 9.3132e-09, 4.9826e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 7.0129e-07, + -2.3190e-07, 1.2182e-06], + [ 0.0000e+00, 0.0000e+00, 1.4715e-06, ..., 7.9628e-08, + 3.2783e-07, 3.0361e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6333e-07, ..., 2.6822e-07, + -2.5705e-07, 2.7847e-07], + [ 0.0000e+00, 0.0000e+00, 1.2424e-06, ..., 1.3644e-07, + -1.1176e-08, 3.3528e-07], + [ 0.0000e+00, 0.0000e+00, 3.1339e-07, ..., -2.6310e-07, + 4.1910e-08, -1.1930e-06]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0283, 0.0146, 0.0267, 0.0035, 0.0275, -0.0099, -0.0060, -0.0165, + -0.0150, 0.0189], device='cuda:0'), grad: tensor([ 4.1164e-07, 1.6624e-06, 8.0541e-06, -2.3246e-05, -1.3344e-05, + 5.9493e-06, 1.1317e-05, 2.3209e-06, 7.2904e-06, -4.4191e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 220.58, cls_loss 0.0018 cls_loss_mapping 0.0039 cls_loss_causal 0.5388 re_mapping 0.0061 re_causal 0.0190 /// teacc 98.91 lr 0.00010000 +Epoch 183, weight, value: tensor([[ 0.0086, 0.0110, 0.0152, ..., 0.0612, -0.1487, -0.1341], + [ 0.0454, -0.0185, -0.0043, ..., -0.1560, 0.0481, -0.0950], + [-0.0061, -0.0048, 0.1512, ..., -0.2267, -0.0856, -0.1342], + ..., + [-0.0092, 0.0062, -0.1177, ..., -0.0036, 0.1200, 0.0604], + [ 0.0026, -0.0202, -0.0899, ..., -0.1934, 0.0775, -0.1359], + [-0.0063, -0.0258, -0.0471, ..., 0.0341, -0.0893, 0.0196]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.3085e-08, ..., -8.8476e-08, + 1.8533e-07, 1.0245e-07], + [ 9.3132e-10, 0.0000e+00, 3.4925e-08, ..., 4.4238e-08, + 1.0524e-07, 2.8266e-07], + [ 0.0000e+00, 0.0000e+00, -3.0780e-07, ..., 1.5367e-08, + 1.9465e-07, 3.2689e-07], + ..., + [-2.3283e-09, 0.0000e+00, 1.0571e-07, ..., 1.1940e-06, + -5.4995e-07, 1.9446e-06], + [ 0.0000e+00, 0.0000e+00, 1.0524e-07, ..., 5.4482e-08, + 6.4680e-07, 1.7984e-06], + [ 9.3132e-10, 0.0000e+00, 4.7497e-08, ..., -2.8051e-06, + 2.1607e-07, -1.0677e-05]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0279, 0.0148, 0.0265, 0.0035, 0.0271, -0.0094, -0.0059, -0.0166, + -0.0154, 0.0191], device='cuda:0'), grad: tensor([ 5.7090e-07, 8.8196e-07, 6.8964e-07, 1.1429e-05, 4.8019e-06, + 2.4792e-06, -2.1569e-06, 1.5246e-06, 1.0841e-05, -3.1024e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 220.87, cls_loss 0.0023 cls_loss_mapping 0.0049 cls_loss_causal 0.5358 re_mapping 0.0066 re_causal 0.0193 /// teacc 99.02 lr 0.00010000 +Epoch 184, weight, value: tensor([[ 0.0086, 0.0110, 0.0164, ..., 0.0618, -0.1491, -0.1344], + [ 0.0455, -0.0185, -0.0043, ..., -0.1554, 0.0483, -0.0949], + [-0.0062, -0.0048, 0.1512, ..., -0.2293, -0.0858, -0.1349], + ..., + [-0.0092, 0.0062, -0.1185, ..., -0.0041, 0.1201, 0.0602], + [ 0.0026, -0.0202, -0.0886, ..., -0.1942, 0.0790, -0.1362], + [-0.0063, -0.0258, -0.0480, ..., 0.0337, -0.0895, 0.0180]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.8626e-09, ..., 3.2689e-07, + 3.0268e-08, 5.4482e-07], + [-9.2667e-08, 0.0000e+00, 7.9162e-09, ..., -1.1846e-06, + -3.0622e-06, 6.0070e-08], + [ 1.3970e-08, 0.0000e+00, 1.6019e-07, ..., 1.6224e-06, + 8.6892e-07, 2.4308e-06], + ..., + [ 1.2107e-08, 0.0000e+00, 9.3132e-09, ..., 3.0687e-07, + 3.9721e-07, 2.8312e-07], + [ 1.7695e-08, 0.0000e+00, -2.1234e-07, ..., 1.9092e-06, + 5.2154e-08, 2.0154e-06], + [ 2.3283e-09, 0.0000e+00, 9.7789e-09, ..., -3.3583e-06, + 9.5461e-08, -5.7817e-06]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0277, 0.0150, 0.0262, 0.0059, 0.0290, -0.0118, -0.0072, -0.0168, + -0.0143, 0.0174], device='cuda:0'), grad: tensor([ 1.2368e-06, -8.6203e-06, 7.1749e-06, 1.6898e-05, 4.2357e-06, + -1.6898e-05, 3.9861e-07, 1.5870e-06, 5.6438e-06, -1.1683e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 220.79, cls_loss 0.0018 cls_loss_mapping 0.0027 cls_loss_causal 0.5131 re_mapping 0.0061 re_causal 0.0184 /// teacc 98.92 lr 0.00010000 +Epoch 185, weight, value: tensor([[ 0.0085, 0.0110, 0.0166, ..., 0.0619, -0.1493, -0.1346], + [ 0.0465, -0.0185, -0.0040, ..., -0.1556, 0.0492, -0.0946], + [-0.0067, -0.0048, 0.1516, ..., -0.2302, -0.0867, -0.1351], + ..., + [-0.0093, 0.0062, -0.1189, ..., -0.0045, 0.1200, 0.0604], + [ 0.0023, -0.0202, -0.0882, ..., -0.1948, 0.0785, -0.1364], + [-0.0063, -0.0258, -0.0482, ..., 0.0340, -0.0896, 0.0181]], + device='cuda:0'), grad: tensor([[ 2.7008e-08, 0.0000e+00, -1.8161e-08, ..., 3.3062e-08, + 4.0559e-07, 3.6228e-07], + [ 3.3993e-08, 0.0000e+00, -2.1514e-06, ..., 1.1660e-06, + -3.7067e-06, 6.9588e-06], + [ 7.1246e-08, 0.0000e+00, 1.8151e-06, ..., 3.4273e-07, + 4.3213e-06, 2.0489e-06], + ..., + [ 1.2048e-05, 0.0000e+00, 5.6345e-08, ..., 8.2776e-06, + 1.3039e-07, 5.0098e-05], + [ 8.8289e-07, 0.0000e+00, 6.3796e-08, ..., 4.2561e-07, + 4.0978e-06, 2.2091e-06], + [ 6.9151e-07, 0.0000e+00, 9.3132e-08, ..., 1.5664e-04, + 6.7847e-07, 9.5701e-04]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0277, 0.0154, 0.0261, 0.0050, 0.0287, -0.0110, -0.0056, -0.0167, + -0.0158, 0.0175], device='cuda:0'), grad: tensor([ 1.1772e-06, 5.4948e-07, 1.0580e-05, -2.9564e-05, -9.5606e-04, + -8.0019e-06, -1.2815e-05, 7.9274e-05, 1.1265e-05, 9.0313e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 221.13, cls_loss 0.0018 cls_loss_mapping 0.0026 cls_loss_causal 0.5202 re_mapping 0.0059 re_causal 0.0178 /// teacc 98.94 lr 0.00010000 +Epoch 186, weight, value: tensor([[ 0.0081, 0.0110, 0.0169, ..., 0.0622, -0.1496, -0.1348], + [ 0.0486, -0.0185, -0.0039, ..., -0.1586, 0.0492, -0.0950], + [-0.0071, -0.0048, 0.1518, ..., -0.2310, -0.0868, -0.1358], + ..., + [-0.0097, 0.0062, -0.1197, ..., -0.0050, 0.1200, 0.0602], + [ 0.0010, -0.0202, -0.0887, ..., -0.1953, 0.0785, -0.1367], + [-0.0056, -0.0258, -0.0482, ..., 0.0338, -0.0898, 0.0179]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 2.8312e-05, ..., 1.9092e-08, + 2.7772e-06, 6.0070e-08], + [-3.3993e-08, 0.0000e+00, 2.3935e-07, ..., 6.3330e-08, + -3.6601e-07, 3.7113e-07], + [ 9.3132e-09, 0.0000e+00, -4.2409e-05, ..., 1.8626e-08, + 3.9488e-07, 1.8068e-07], + ..., + [ 2.7940e-09, 0.0000e+00, 5.4901e-07, ..., 3.4925e-08, + -2.2873e-06, -6.1616e-06], + [ 6.0536e-09, 0.0000e+00, 8.8662e-06, ..., -5.5879e-09, + -1.9511e-07, 5.3458e-07], + [ 4.6566e-10, 0.0000e+00, 5.3551e-07, ..., -2.6496e-07, + 1.7341e-06, 3.8147e-06]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0275, 0.0152, 0.0260, 0.0047, 0.0293, -0.0107, -0.0056, -0.0170, + -0.0160, 0.0173], device='cuda:0'), grad: tensor([ 6.2823e-05, -4.6566e-09, -6.6638e-05, 1.4231e-06, 1.0263e-06, + 3.2689e-06, -1.6958e-05, -8.1137e-06, 1.5497e-05, 7.6666e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 220.87, cls_loss 0.0017 cls_loss_mapping 0.0033 cls_loss_causal 0.5073 re_mapping 0.0059 re_causal 0.0177 /// teacc 98.94 lr 0.00010000 +Epoch 187, weight, value: tensor([[ 0.0080, 0.0110, 0.0169, ..., 0.0625, -0.1500, -0.1350], + [ 0.0491, -0.0185, -0.0035, ..., -0.1592, 0.0495, -0.0964], + [-0.0070, -0.0048, 0.1519, ..., -0.2320, -0.0880, -0.1367], + ..., + [-0.0099, 0.0062, -0.1206, ..., -0.0050, 0.1212, 0.0611], + [ 0.0008, -0.0202, -0.0892, ..., -0.1957, 0.0785, -0.1368], + [-0.0055, -0.0258, -0.0486, ..., 0.0334, -0.0904, 0.0176]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, 0.0000e+00, -6.6450e-07, ..., -6.2911e-07, + 1.2619e-07, 1.7928e-07], + [ 1.5600e-07, 0.0000e+00, 1.5413e-07, ..., 8.5682e-08, + 8.5495e-07, 1.2452e-06], + [ 4.3735e-06, 0.0000e+00, 1.0189e-06, ..., 1.9046e-07, + 1.7986e-05, 1.7926e-05], + ..., + [-4.9770e-06, 0.0000e+00, -9.7137e-07, ..., 2.8685e-07, + -2.2665e-05, -2.3678e-05], + [ 2.8452e-07, 0.0000e+00, 5.2014e-07, ..., 4.0885e-07, + 1.5236e-06, 1.9390e-06], + [ 2.5611e-08, 0.0000e+00, 1.3281e-06, ..., 1.0617e-06, + 1.0263e-06, 3.2075e-06]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0276, 0.0154, 0.0251, 0.0048, 0.0296, -0.0106, -0.0056, -0.0160, + -0.0161, 0.0168], device='cuda:0'), grad: tensor([-2.3469e-07, 3.8967e-06, 4.0531e-05, -3.5226e-05, 1.5292e-06, + -2.3544e-05, 1.9446e-05, -5.0098e-05, 1.3366e-05, 3.0398e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 220.65, cls_loss 0.0028 cls_loss_mapping 0.0049 cls_loss_causal 0.5291 re_mapping 0.0060 re_causal 0.0176 /// teacc 98.91 lr 0.00010000 +Epoch 188, weight, value: tensor([[ 0.0079, 0.0110, 0.0169, ..., 0.0626, -0.1510, -0.1354], + [ 0.0493, -0.0185, -0.0035, ..., -0.1598, 0.0495, -0.0968], + [-0.0047, -0.0048, 0.1526, ..., -0.2328, -0.0878, -0.1348], + ..., + [-0.0111, 0.0062, -0.1237, ..., -0.0059, 0.1207, 0.0610], + [ 0.0006, -0.0202, -0.0896, ..., -0.1968, 0.0787, -0.1369], + [-0.0084, -0.0258, -0.0489, ..., 0.0326, -0.0908, 0.0167]], + device='cuda:0'), grad: tensor([[ 6.9849e-09, 0.0000e+00, -9.3598e-08, ..., -1.7602e-07, + 5.3970e-07, 1.8626e-07], + [ 2.1094e-07, 0.0000e+00, 8.3372e-06, ..., 9.6392e-08, + 7.3947e-06, 6.0312e-06], + [ 1.8161e-08, 0.0000e+00, -1.1735e-05, ..., 2.3283e-08, + -3.3192e-06, 3.3667e-07], + ..., + [-6.1514e-07, 0.0000e+00, 3.0734e-06, ..., 9.2201e-08, + -5.7966e-06, -1.6198e-05], + [ 1.5367e-08, 0.0000e+00, 2.0396e-07, ..., 4.1444e-08, + -1.1034e-05, 8.7544e-08], + [ 2.8498e-07, 0.0000e+00, 1.0012e-07, ..., 1.7742e-07, + 6.8434e-06, 7.0408e-06]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0280, 0.0149, 0.0254, 0.0030, 0.0306, -0.0086, -0.0055, -0.0168, + -0.0163, 0.0165], device='cuda:0'), grad: tensor([ 2.6897e-06, 2.8968e-05, -7.3053e-06, 7.1190e-06, 3.5390e-06, + 1.0841e-05, 5.6252e-06, -2.0429e-05, -6.4611e-05, 3.3408e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 220.71, cls_loss 0.0029 cls_loss_mapping 0.0042 cls_loss_causal 0.5540 re_mapping 0.0064 re_causal 0.0183 /// teacc 99.01 lr 0.00010000 +Epoch 189, weight, value: tensor([[ 0.0075, 0.0110, 0.0185, ..., 0.0632, -0.1517, -0.1366], + [ 0.0496, -0.0185, -0.0058, ..., -0.1600, 0.0469, -0.0962], + [-0.0047, -0.0048, 0.1558, ..., -0.2342, -0.0861, -0.1345], + ..., + [-0.0112, 0.0062, -0.1271, ..., -0.0064, 0.1221, 0.0608], + [ 0.0003, -0.0202, -0.0900, ..., -0.1985, 0.0787, -0.1372], + [-0.0083, -0.0258, -0.0499, ..., 0.0328, -0.0913, 0.0168]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -3.5856e-07, ..., 4.6473e-07, + 1.2433e-07, 2.0582e-07], + [-1.0245e-08, 0.0000e+00, 1.4435e-08, ..., 4.9826e-08, + -2.4261e-03, -2.9755e-03], + [ 5.5879e-09, 0.0000e+00, -4.5635e-08, ..., 5.3691e-07, + 7.8138e-07, 6.3144e-07], + ..., + [-6.5193e-09, 0.0000e+00, 2.0489e-08, ..., -4.1444e-07, + 2.3460e-03, 2.8744e-03], + [ 4.1910e-09, 0.0000e+00, 8.1025e-08, ..., 9.1270e-08, + 2.9840e-06, 4.0717e-06], + [ 1.3970e-09, 0.0000e+00, 8.8010e-08, ..., 5.4762e-07, + 6.0767e-05, 7.8380e-05]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0273, 0.0127, 0.0276, 0.0028, 0.0304, -0.0083, -0.0054, -0.0169, + -0.0166, 0.0165], device='cuda:0'), grad: tensor([ 5.3197e-06, -6.3896e-03, 6.8955e-06, 3.1367e-06, 4.8101e-05, + 3.1255e-06, -1.6868e-05, 6.1798e-03, 3.8017e-06, 1.6463e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 220.53, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5097 re_mapping 0.0062 re_causal 0.0180 /// teacc 99.06 lr 0.00010000 +Epoch 190, weight, value: tensor([[ 7.4816e-03, 1.0998e-02, 1.8470e-02, ..., 6.2829e-02, + -1.5214e-01, -1.3809e-01], + [ 4.9818e-02, -1.8468e-02, -5.9027e-03, ..., -1.6031e-01, + 4.8798e-02, -9.3775e-02], + [-4.7799e-03, -4.7728e-03, 1.5678e-01, ..., -2.3489e-01, + -8.5785e-02, -1.3463e-01], + ..., + [-1.1210e-02, 6.2061e-03, -1.2909e-01, ..., -6.5265e-03, + 1.1999e-01, 5.9929e-02], + [-2.9313e-05, -2.0220e-02, -9.0371e-02, ..., -1.9982e-01, + 7.8641e-02, -1.3785e-01], + [-8.2615e-03, -2.5834e-02, -5.0036e-02, ..., 3.2980e-02, + -9.1768e-02, 1.6747e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.7044e-07, ..., -5.8860e-06, + -2.8173e-07, -1.4957e-06], + [ 0.0000e+00, 0.0000e+00, 3.8650e-08, ..., 1.4110e-07, + 3.4496e-06, 1.5832e-06], + [ 0.0000e+00, 0.0000e+00, -6.0536e-09, ..., 1.4044e-06, + 1.8049e-06, 1.3113e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7474e-08, ..., 1.4435e-07, + -7.1786e-06, -3.3230e-06], + [ 0.0000e+00, 0.0000e+00, 6.7987e-08, ..., 1.0282e-06, + 2.0629e-07, 7.4692e-07], + [-4.6566e-10, 0.0000e+00, 2.0349e-07, ..., 1.8440e-06, + 3.4971e-07, 2.2724e-06]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0275, 0.0141, 0.0279, 0.0031, 0.0303, -0.0083, -0.0055, -0.0185, + -0.0170, 0.0165], device='cuda:0'), grad: tensor([-1.4000e-05, 1.0610e-05, 8.2254e-06, 4.8950e-06, -5.0757e-07, + 9.9093e-07, 3.7961e-06, -2.0877e-05, 2.8983e-06, 3.9712e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 220.46, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5387 re_mapping 0.0057 re_causal 0.0178 /// teacc 98.96 lr 0.00010000 +Epoch 191, weight, value: tensor([[ 7.3922e-03, 1.0998e-02, 1.8818e-02, ..., 6.2905e-02, + -1.5245e-01, -1.3855e-01], + [ 4.9907e-02, -1.8468e-02, -5.9654e-03, ..., -1.6050e-01, + 4.8589e-02, -9.3843e-02], + [-4.7904e-03, -4.7728e-03, 1.5678e-01, ..., -2.3595e-01, + -8.6592e-02, -1.3557e-01], + ..., + [-1.1246e-02, 6.2061e-03, -1.2875e-01, ..., -6.7532e-03, + 1.2092e-01, 6.0399e-02], + [-1.4391e-04, -2.0220e-02, -9.0693e-02, ..., -2.0049e-01, + 7.7626e-02, -1.3817e-01], + [-8.5302e-03, -2.5834e-02, -5.0256e-02, ..., 3.3599e-02, + -9.1997e-02, 1.7795e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1921e-07, ..., -1.2852e-07, + 1.4808e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.0151e-07, ..., 2.0023e-08, + 1.2107e-08, 1.1921e-07], + [ 0.0000e+00, 0.0000e+00, -1.3299e-06, ..., 1.4435e-08, + 1.1325e-06, 1.4948e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 8.4704e-07, ..., 1.3504e-08, + -4.1444e-07, -5.4343e-07], + [ 0.0000e+00, 0.0000e+00, 2.2957e-07, ..., 1.4901e-08, + -2.3097e-06, 2.4680e-08], + [ 0.0000e+00, 0.0000e+00, 4.3306e-08, ..., 1.3039e-08, + 6.2119e-07, 8.7079e-08]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0275, 0.0140, 0.0276, 0.0031, 0.0293, -0.0084, -0.0045, -0.0180, + -0.0178, 0.0172], device='cuda:0'), grad: tensor([ 1.5926e-06, 1.4016e-06, 1.6525e-05, 8.9034e-06, 6.0862e-07, + -2.0992e-06, 3.6843e-06, 1.2601e-06, -4.3243e-05, 1.1370e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 220.55, cls_loss 0.0018 cls_loss_mapping 0.0037 cls_loss_causal 0.5339 re_mapping 0.0059 re_causal 0.0174 /// teacc 98.98 lr 0.00010000 +Epoch 192, weight, value: tensor([[ 7.3888e-03, 1.0998e-02, 1.9396e-02, ..., 6.2899e-02, + -1.5306e-01, -1.3919e-01], + [ 4.9909e-02, -1.8468e-02, -5.9361e-03, ..., -1.6085e-01, + 4.8851e-02, -9.3886e-02], + [-4.7899e-03, -4.7728e-03, 1.5686e-01, ..., -2.3704e-01, + -8.6881e-02, -1.3592e-01], + ..., + [-1.1247e-02, 6.2061e-03, -1.2881e-01, ..., -6.9983e-03, + 1.2096e-01, 6.0451e-02], + [-1.4901e-04, -2.0220e-02, -9.1095e-02, ..., -2.0181e-01, + 7.7872e-02, -1.3837e-01], + [-8.5310e-03, -2.5834e-02, -5.0486e-02, ..., 3.3903e-02, + -9.2206e-02, 1.7925e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4760e-05, ..., -5.1549e-07, + 2.3115e-06, 1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 1.7462e-07, ..., 1.0710e-07, + 7.8082e-05, 4.1863e-07], + [ 0.0000e+00, 0.0000e+00, 6.0350e-07, ..., 1.2200e-07, + 1.4879e-05, 4.4610e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 8.8010e-08, ..., 1.7276e-07, + -1.2445e-04, -5.6950e-07], + [ 0.0000e+00, 0.0000e+00, 2.2631e-07, ..., 1.4761e-07, + 2.7254e-05, 4.9779e-07], + [ 0.0000e+00, 0.0000e+00, 2.5099e-07, ..., -8.8941e-08, + 1.8217e-06, -5.4250e-07]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0278, 0.0142, 0.0274, 0.0039, 0.0293, -0.0091, -0.0043, -0.0181, + -0.0180, 0.0174], device='cuda:0'), grad: tensor([ 8.6188e-05, 1.3888e-04, 2.9862e-05, 3.1274e-06, 1.5907e-06, + 2.0713e-06, -9.5546e-05, -2.1899e-04, 4.9442e-05, 3.1386e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 220.53, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5106 re_mapping 0.0060 re_causal 0.0175 /// teacc 99.04 lr 0.00010000 +Epoch 193, weight, value: tensor([[ 7.3867e-03, 1.0998e-02, 1.9735e-02, ..., 6.3190e-02, + -1.5340e-01, -1.3888e-01], + [ 4.9914e-02, -1.8468e-02, -5.9129e-03, ..., -1.6119e-01, + 5.0581e-02, -9.2477e-02], + [-4.7914e-03, -4.7728e-03, 1.5687e-01, ..., -2.3933e-01, + -8.7093e-02, -1.3623e-01], + ..., + [-1.1249e-02, 6.2061e-03, -1.2887e-01, ..., -7.2148e-03, + 1.1949e-01, 5.9925e-02], + [-1.5749e-04, -2.0220e-02, -9.1350e-02, ..., -2.0223e-01, + 7.7694e-02, -1.3857e-01], + [-8.5213e-03, -2.5834e-02, -5.0659e-02, ..., 3.3801e-02, + -9.2439e-02, 1.7821e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 7.4040e-08, + 6.4727e-08, 1.3923e-07], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.9092e-08, + -5.7090e-07, 1.9884e-07], + [ 0.0000e+00, 0.0000e+00, -1.3923e-07, ..., 6.1933e-08, + 2.4820e-07, 1.9325e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.2375e-08, ..., 1.5832e-08, + -2.5220e-06, -3.3788e-06], + [ 0.0000e+00, 0.0000e+00, 6.1002e-08, ..., 1.9092e-08, + 1.5078e-06, 1.7816e-06], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 6.1141e-07, + 9.3831e-07, 1.9521e-06]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0276, 0.0156, 0.0271, 0.0039, 0.0292, -0.0089, -0.0043, -0.0191, + -0.0183, 0.0171], device='cuda:0'), grad: tensor([ 4.4005e-07, -1.8161e-06, 8.1211e-07, -1.3737e-07, -1.4883e-06, + 4.5681e-07, 3.2783e-07, -6.2846e-06, 2.8461e-06, 4.8093e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 220.18, cls_loss 0.0015 cls_loss_mapping 0.0034 cls_loss_causal 0.4951 re_mapping 0.0063 re_causal 0.0182 /// teacc 98.97 lr 0.00010000 +Epoch 194, weight, value: tensor([[ 7.3754e-03, 1.0998e-02, 2.0252e-02, ..., 6.3371e-02, + -1.5361e-01, -1.3884e-01], + [ 5.0152e-02, -1.8468e-02, -5.9264e-03, ..., -1.6170e-01, + 5.1410e-02, -9.2189e-02], + [-4.8933e-03, -4.7728e-03, 1.5716e-01, ..., -2.3977e-01, + -8.7329e-02, -1.3596e-01], + ..., + [-1.1260e-02, 6.2061e-03, -1.2937e-01, ..., -7.3174e-03, + 1.1877e-01, 5.9739e-02], + [-2.0346e-04, -2.0220e-02, -9.1559e-02, ..., -2.0255e-01, + 7.7942e-02, -1.3905e-01], + [-8.4917e-03, -2.5834e-02, -5.0954e-02, ..., 3.3838e-02, + -9.2693e-02, 1.7993e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.6275e-07, ..., -5.9512e-07, + 1.3504e-08, 4.9360e-08], + [ 0.0000e+00, 0.0000e+00, 3.1199e-08, ..., 9.9186e-08, + -1.1176e-08, 1.6624e-07], + [ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., 8.8476e-08, + 1.3672e-06, 2.7847e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4261e-08, ..., 1.3085e-07, + -1.7090e-07, -2.0256e-07], + [ 0.0000e+00, 0.0000e+00, 6.2399e-08, ..., 2.0396e-07, + -1.7229e-06, 3.7299e-07], + [ 0.0000e+00, 0.0000e+00, 3.5437e-07, ..., 4.6730e-05, + 5.4017e-08, 8.8155e-05]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0273, 0.0163, 0.0269, 0.0044, 0.0291, -0.0093, -0.0044, -0.0197, + -0.0183, 0.0172], device='cuda:0'), grad: tensor([-2.2538e-06, 9.1270e-08, 2.8871e-06, 6.5332e-07, -1.0347e-04, + 1.3597e-07, -2.4047e-06, -2.6636e-07, -2.2203e-06, 1.0663e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 220.68, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5166 re_mapping 0.0057 re_causal 0.0177 /// teacc 98.94 lr 0.00010000 +Epoch 195, weight, value: tensor([[ 0.0073, 0.0110, 0.0204, ..., 0.0634, -0.1539, -0.1392], + [ 0.0502, -0.0185, -0.0060, ..., -0.1620, 0.0513, -0.0924], + [-0.0049, -0.0048, 0.1574, ..., -0.2402, -0.0875, -0.1361], + ..., + [-0.0120, 0.0062, -0.1295, ..., -0.0073, 0.1191, 0.0601], + [-0.0005, -0.0202, -0.0916, ..., -0.2028, 0.0781, -0.1394], + [-0.0085, -0.0258, -0.0511, ..., 0.0341, -0.0931, 0.0179]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.5832e-07, ..., 6.2631e-07, + 1.3784e-06, 4.3735e-06], + [ 9.3132e-10, 0.0000e+00, 3.9581e-08, ..., 4.3884e-06, + 5.0943e-07, 5.7817e-06], + [ 0.0000e+00, 0.0000e+00, 1.6065e-07, ..., 9.4483e-07, + 2.8033e-07, 4.7237e-06], + ..., + [ 4.6566e-10, 0.0000e+00, 8.4750e-08, ..., 3.2689e-07, + -7.3723e-06, -4.5113e-06], + [ 4.6566e-10, 0.0000e+00, 6.4261e-07, ..., 1.8673e-07, + 2.1001e-07, 5.6252e-07], + [ 3.3062e-08, 0.0000e+00, 7.6834e-08, ..., 8.9034e-06, + 2.0191e-06, 1.2055e-05]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0274, 0.0162, 0.0266, 0.0062, 0.0291, -0.0111, -0.0042, -0.0193, + -0.0182, 0.0170], device='cuda:0'), grad: tensor([ 4.2588e-05, 1.7300e-05, 4.8369e-05, -3.5148e-06, -3.1918e-05, + 1.5251e-05, -9.8765e-05, -3.1829e-05, 4.6939e-06, 3.7879e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 220.76, cls_loss 0.0017 cls_loss_mapping 0.0027 cls_loss_causal 0.5364 re_mapping 0.0054 re_causal 0.0169 /// teacc 98.98 lr 0.00010000 +Epoch 196, weight, value: tensor([[ 0.0072, 0.0110, 0.0203, ..., 0.0632, -0.1544, -0.1396], + [ 0.0501, -0.0185, -0.0061, ..., -0.1624, 0.0515, -0.0924], + [-0.0050, -0.0048, 0.1578, ..., -0.2405, -0.0876, -0.1358], + ..., + [-0.0120, 0.0062, -0.1300, ..., -0.0075, 0.1190, 0.0601], + [-0.0006, -0.0202, -0.0919, ..., -0.2032, 0.0784, -0.1395], + [-0.0086, -0.0258, -0.0513, ..., 0.0340, -0.0934, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1269e-06, ..., -5.2946e-07, + 7.9721e-06, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 3.3528e-07, ..., 2.8405e-08, + 5.4948e-08, 1.2387e-07], + [ 0.0000e+00, 0.0000e+00, -4.9453e-07, ..., 2.3283e-08, + 7.8464e-07, 2.3749e-08], + ..., + [-4.6566e-10, 0.0000e+00, 7.7765e-08, ..., 5.1223e-08, + -1.0477e-07, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 1.0151e-07, ..., 1.8673e-07, + -1.2502e-05, 6.0583e-07], + [ 0.0000e+00, 0.0000e+00, 8.1956e-08, ..., -3.4226e-07, + 8.1491e-08, -1.4426e-06]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0278, 0.0164, 0.0268, 0.0063, 0.0294, -0.0110, -0.0042, -0.0196, + -0.0181, 0.0167], device='cuda:0'), grad: tensor([ 3.0085e-05, 6.6822e-07, 2.4736e-06, 6.9244e-07, 1.1325e-06, + 1.0002e-06, 1.4596e-05, 4.1397e-07, -4.9263e-05, -1.8086e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 220.24, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5561 re_mapping 0.0058 re_causal 0.0168 /// teacc 98.98 lr 0.00010000 +Epoch 197, weight, value: tensor([[ 0.0072, 0.0110, 0.0208, ..., 0.0633, -0.1558, -0.1404], + [ 0.0500, -0.0185, -0.0061, ..., -0.1631, 0.0516, -0.0922], + [-0.0050, -0.0048, 0.1586, ..., -0.2419, -0.0883, -0.1342], + ..., + [-0.0120, 0.0062, -0.1305, ..., -0.0080, 0.1197, 0.0603], + [-0.0007, -0.0202, -0.0935, ..., -0.2041, 0.0786, -0.1403], + [-0.0085, -0.0258, -0.0525, ..., 0.0340, -0.0941, 0.0172]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1874e-07, ..., -3.2689e-07, + 1.4855e-07, 1.5413e-07], + [ 0.0000e+00, 0.0000e+00, 6.7521e-08, ..., 1.8021e-06, + 1.7330e-05, 2.4289e-05], + [ 0.0000e+00, 0.0000e+00, 9.1270e-08, ..., 7.2177e-08, + 5.1782e-07, 4.1723e-07], + ..., + [ 1.3970e-09, 0.0000e+00, 2.1793e-07, ..., 1.0729e-06, + 1.0617e-05, 1.4432e-05], + [ 1.1642e-08, 0.0000e+00, -9.7416e-07, ..., 4.9472e-06, + 4.6641e-05, 6.7294e-05], + [ 2.7940e-09, 0.0000e+00, 1.7602e-07, ..., 7.8790e-07, + 7.9051e-06, 1.0408e-05]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0280, 0.0164, 0.0269, 0.0062, 0.0294, -0.0111, -0.0041, -0.0189, + -0.0189, 0.0166], device='cuda:0'), grad: tensor([ 5.6950e-07, 6.0052e-05, 1.8366e-06, 1.9908e-05, -2.9397e-04, + -1.9848e-05, 3.9712e-06, 3.7432e-05, 1.6284e-04, 2.7627e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 220.61, cls_loss 0.0024 cls_loss_mapping 0.0053 cls_loss_causal 0.5462 re_mapping 0.0058 re_causal 0.0175 /// teacc 98.93 lr 0.00010000 +Epoch 198, weight, value: tensor([[ 0.0072, 0.0110, 0.0212, ..., 0.0638, -0.1565, -0.1407], + [ 0.0500, -0.0185, -0.0062, ..., -0.1646, 0.0509, -0.0928], + [-0.0050, -0.0048, 0.1587, ..., -0.2427, -0.0887, -0.1343], + ..., + [-0.0120, 0.0062, -0.1305, ..., -0.0081, 0.1210, 0.0607], + [-0.0007, -0.0202, -0.0944, ..., -0.2049, 0.0784, -0.1400], + [-0.0085, -0.0258, -0.0529, ..., 0.0335, -0.0950, 0.0167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0105e-07, ..., 2.3283e-09, + 5.3085e-08, 5.4948e-08], + [ 0.0000e+00, 0.0000e+00, 1.2061e-07, ..., 2.3283e-08, + -6.3330e-08, 2.2911e-07], + [ 0.0000e+00, 0.0000e+00, -1.9848e-05, ..., 4.6566e-09, + 7.2177e-07, -1.0058e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 1.7568e-05, ..., 1.7695e-08, + -1.9707e-06, 8.3968e-06], + [ 0.0000e+00, 0.0000e+00, 1.7630e-06, ..., 5.4017e-08, + 9.9093e-07, 1.2666e-06], + [ 0.0000e+00, 0.0000e+00, 4.1677e-07, ..., -4.8894e-08, + 1.2200e-07, 1.7229e-08]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0283, 0.0159, 0.0264, 0.0068, 0.0299, -0.0111, -0.0042, -0.0179, + -0.0196, 0.0158], device='cuda:0'), grad: tensor([ 4.5309e-07, -5.0524e-07, -3.7789e-05, 4.5821e-07, 1.7323e-07, + -2.0303e-06, 4.4005e-07, 3.1710e-05, 6.1952e-06, 8.5589e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 220.66, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.5084 re_mapping 0.0058 re_causal 0.0174 /// teacc 98.96 lr 0.00010000 +Epoch 199, weight, value: tensor([[ 0.0072, 0.0110, 0.0211, ..., 0.0632, -0.1572, -0.1409], + [ 0.0500, -0.0185, -0.0063, ..., -0.1652, 0.0512, -0.0926], + [-0.0050, -0.0048, 0.1591, ..., -0.2439, -0.0888, -0.1341], + ..., + [-0.0120, 0.0062, -0.1308, ..., -0.0083, 0.1209, 0.0605], + [-0.0007, -0.0202, -0.0949, ..., -0.2056, 0.0783, -0.1409], + [-0.0085, -0.0258, -0.0527, ..., 0.0340, -0.0951, 0.0167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.0524e-07, ..., -1.5087e-07, + 6.5193e-09, 4.1910e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.0010e-06, + 1.2098e-06, 7.0333e-06], + [ 0.0000e+00, 0.0000e+00, 2.7521e-07, ..., 1.2433e-07, + 6.1467e-08, 1.2480e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 2.0722e-07, + 5.4482e-08, 4.7684e-07], + [ 0.0000e+00, 0.0000e+00, 2.0023e-08, ..., 1.0468e-06, + 5.2107e-07, 4.2431e-06], + [ 0.0000e+00, 0.0000e+00, 8.9873e-08, ..., 2.2948e-06, + 1.3318e-07, -1.5227e-06]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0294, 0.0161, 0.0266, 0.0067, 0.0299, -0.0110, -0.0039, -0.0182, + -0.0198, 0.0160], device='cuda:0'), grad: tensor([-7.9023e-07, 1.7464e-05, 8.6194e-07, 1.4249e-07, -3.2485e-05, + -6.8499e-07, 1.7704e-06, 9.3598e-07, 5.9307e-06, 6.8657e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 220.39, cls_loss 0.0014 cls_loss_mapping 0.0031 cls_loss_causal 0.4634 re_mapping 0.0057 re_causal 0.0170 /// teacc 98.94 lr 0.00010000 +Epoch 200, weight, value: tensor([[ 0.0072, 0.0110, 0.0216, ..., 0.0630, -0.1575, -0.1422], + [ 0.0500, -0.0185, -0.0065, ..., -0.1657, 0.0512, -0.0928], + [-0.0050, -0.0048, 0.1595, ..., -0.2448, -0.0890, -0.1342], + ..., + [-0.0120, 0.0062, -0.1310, ..., -0.0082, 0.1212, 0.0610], + [-0.0007, -0.0202, -0.0952, ..., -0.2060, 0.0783, -0.1413], + [-0.0085, -0.0258, -0.0529, ..., 0.0341, -0.0958, 0.0165]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8394e-07, ..., -1.8952e-07, + 1.3039e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 1.3039e-08, + -1.7509e-07, 1.7229e-08], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 3.1199e-08, + 2.0722e-07, 6.0536e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 1.8161e-08, + 2.4214e-08, -1.7229e-08], + [ 0.0000e+00, 0.0000e+00, 2.2817e-08, ..., 9.7789e-09, + -2.8731e-07, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 5.8673e-08, ..., 6.1467e-08, + 1.0058e-07, -1.8161e-08]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0295, 0.0164, 0.0266, 0.0066, 0.0302, -0.0109, -0.0039, -0.0183, + -0.0199, 0.0156], device='cuda:0'), grad: tensor([-3.6880e-07, -5.6904e-07, 6.9104e-07, 8.6753e-07, 1.3504e-07, + -1.3262e-06, 5.7742e-08, 2.0862e-07, -5.1875e-07, 8.1910e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 220.73, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5148 re_mapping 0.0056 re_causal 0.0169 /// teacc 99.00 lr 0.00010000 +Epoch 201, weight, value: tensor([[ 0.0072, 0.0110, 0.0222, ..., 0.0627, -0.1576, -0.1423], + [ 0.0499, -0.0185, -0.0064, ..., -0.1661, 0.0507, -0.0931], + [-0.0050, -0.0048, 0.1596, ..., -0.2457, -0.0894, -0.1345], + ..., + [-0.0126, 0.0062, -0.1312, ..., -0.0087, 0.1215, 0.0612], + [-0.0009, -0.0202, -0.0956, ..., -0.2078, 0.0783, -0.1420], + [-0.0085, -0.0258, -0.0530, ..., 0.0345, -0.0958, 0.0165]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7017e-05, ..., -1.7822e-05, + 3.8557e-07, -5.2303e-06], + [ 0.0000e+00, 0.0000e+00, 2.5667e-06, ..., 2.6710e-06, + 5.3495e-05, 2.6494e-05], + [ 0.0000e+00, 0.0000e+00, 2.1365e-06, ..., 2.3171e-06, + 5.2527e-06, 2.9542e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 8.0615e-06, ..., 8.3521e-06, + -1.0890e-04, -4.9174e-05], + [ 0.0000e+00, 0.0000e+00, 1.4398e-06, ..., 1.5534e-06, + 5.0589e-06, 3.2000e-06], + [ 0.0000e+00, 0.0000e+00, 1.8384e-06, ..., 1.6894e-06, + 4.8168e-06, 3.5558e-06]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0298, 0.0161, 0.0264, 0.0067, 0.0303, -0.0108, -0.0034, -0.0182, + -0.0204, 0.0160], device='cuda:0'), grad: tensor([-5.6088e-05, 1.7929e-04, 2.4378e-05, 1.0335e-04, 4.9956e-06, + 1.6510e-05, 2.6133e-06, -3.1924e-04, 2.2262e-05, 2.1428e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 220.17, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5102 re_mapping 0.0055 re_causal 0.0167 /// teacc 98.96 lr 0.00010000 +Epoch 202, weight, value: tensor([[ 0.0072, 0.0110, 0.0232, ..., 0.0630, -0.1602, -0.1428], + [ 0.0499, -0.0185, -0.0069, ..., -0.1678, 0.0507, -0.0932], + [-0.0050, -0.0048, 0.1602, ..., -0.2470, -0.0903, -0.1348], + ..., + [-0.0133, 0.0062, -0.1313, ..., -0.0096, 0.1218, 0.0610], + [-0.0009, -0.0202, -0.0960, ..., -0.2089, 0.0780, -0.1430], + [-0.0085, -0.0258, -0.0538, ..., 0.0341, -0.0962, 0.0163]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9427e-06, ..., -1.1623e-06, + 2.7195e-07, 1.9488e-07], + [ 0.0000e+00, 0.0000e+00, 2.8452e-07, ..., 7.0548e-08, + -2.5816e-06, 3.0547e-07], + [ 0.0000e+00, 0.0000e+00, -7.0147e-06, ..., 1.0533e-06, + 1.7742e-07, -1.8412e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 4.7125e-06, ..., 6.7288e-08, + -3.4366e-06, -1.0595e-05], + [ 0.0000e+00, 0.0000e+00, 1.4412e-07, ..., 4.4936e-08, + 1.1399e-06, 1.2890e-06], + [ 0.0000e+00, 0.0000e+00, 2.0419e-07, ..., 1.1409e-08, + 3.5297e-07, 3.1367e-06]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0302, 0.0160, 0.0261, 0.0063, 0.0314, -0.0103, -0.0030, -0.0182, + -0.0208, 0.0156], device='cuda:0'), grad: tensor([-7.9582e-07, -5.7407e-06, -1.8373e-05, -1.3426e-05, 1.7554e-05, + 1.1198e-05, -7.2904e-06, -6.7195e-07, 7.7635e-06, 9.7156e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 220.52, cls_loss 0.0015 cls_loss_mapping 0.0033 cls_loss_causal 0.5080 re_mapping 0.0057 re_causal 0.0176 /// teacc 98.99 lr 0.00010000 +Epoch 203, weight, value: tensor([[ 0.0072, 0.0110, 0.0236, ..., 0.0632, -0.1604, -0.1430], + [ 0.0501, -0.0185, -0.0069, ..., -0.1683, 0.0507, -0.0931], + [-0.0050, -0.0048, 0.1605, ..., -0.2475, -0.0906, -0.1348], + ..., + [-0.0134, 0.0062, -0.1316, ..., -0.0127, 0.1206, 0.0586], + [-0.0010, -0.0202, -0.0966, ..., -0.2095, 0.0790, -0.1434], + [-0.0086, -0.0258, -0.0540, ..., 0.0355, -0.0936, 0.0183]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.5460e-06, ..., -4.9435e-06, + 1.3970e-08, 2.7940e-09], + [-6.5193e-09, 0.0000e+00, 5.6857e-07, ..., 3.1199e-08, + -1.2806e-07, 1.1269e-07], + [ 3.7719e-08, 0.0000e+00, -6.0257e-07, ..., 2.0210e-07, + 5.8673e-07, 5.3644e-07], + ..., + [-6.0536e-08, 0.0000e+00, 2.3004e-07, ..., 3.0734e-08, + -8.8941e-07, -9.6485e-07], + [ 6.5193e-09, 0.0000e+00, 1.1967e-07, ..., 1.4901e-07, + 7.4506e-09, 5.0757e-08], + [ 1.0710e-08, 0.0000e+00, 3.7998e-07, ..., 7.2038e-07, + 1.5739e-07, 4.8243e-07]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0302, 0.0159, 0.0260, 0.0063, 0.0312, -0.0104, -0.0026, -0.0198, + -0.0203, 0.0175], device='cuda:0'), grad: tensor([-3.7760e-05, 5.5740e-07, 8.5309e-07, 3.1553e-06, -3.0175e-07, + -3.8594e-06, 3.4809e-05, -1.4659e-06, 7.1106e-07, 3.3751e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 220.55, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4985 re_mapping 0.0058 re_causal 0.0176 /// teacc 99.02 lr 0.00010000 +Epoch 204, weight, value: tensor([[ 0.0071, 0.0110, 0.0240, ..., 0.0637, -0.1609, -0.1433], + [ 0.0507, -0.0185, -0.0069, ..., -0.1687, 0.0505, -0.0933], + [-0.0052, -0.0048, 0.1605, ..., -0.2489, -0.0907, -0.1350], + ..., + [-0.0134, 0.0062, -0.1316, ..., -0.0131, 0.1209, 0.0586], + [-0.0011, -0.0202, -0.0972, ..., -0.2111, 0.0789, -0.1439], + [-0.0086, -0.0258, -0.0544, ..., 0.0354, -0.0936, 0.0184]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.4005e-07, ..., -7.7765e-07, + 1.6764e-08, 8.8476e-09], + [ 9.3132e-10, 0.0000e+00, 2.8685e-07, ..., 3.6787e-08, + -8.4797e-07, -4.8429e-07], + [ 0.0000e+00, 0.0000e+00, -1.0487e-06, ..., 5.4948e-08, + -5.9605e-08, 5.2620e-08], + ..., + [ 4.6566e-10, 0.0000e+00, 6.5425e-07, ..., 4.0513e-08, + 3.3760e-07, -5.4017e-08], + [ 0.0000e+00, 0.0000e+00, 2.5658e-07, ..., 1.3364e-07, + 1.7602e-07, 1.4435e-07], + [ 1.4435e-08, 0.0000e+00, 1.9744e-07, ..., 1.7276e-07, + 2.2911e-07, 1.5227e-07]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0302, 0.0158, 0.0257, 0.0064, 0.0314, -0.0105, -0.0022, -0.0196, + -0.0206, 0.0175], device='cuda:0'), grad: tensor([-2.0564e-06, -1.9558e-06, -1.1232e-06, 1.7869e-04, 3.4552e-07, + -2.1482e-04, 8.3959e-07, 1.8878e-06, 1.5765e-05, 2.1979e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 220.29, cls_loss 0.0019 cls_loss_mapping 0.0039 cls_loss_causal 0.4944 re_mapping 0.0056 re_causal 0.0168 /// teacc 98.94 lr 0.00010000 +Epoch 205, weight, value: tensor([[ 0.0070, 0.0110, 0.0251, ..., 0.0644, -0.1604, -0.1433], + [ 0.0507, -0.0185, -0.0069, ..., -0.1712, 0.0509, -0.0932], + [-0.0050, -0.0048, 0.1613, ..., -0.2497, -0.0914, -0.1340], + ..., + [-0.0135, 0.0062, -0.1330, ..., -0.0123, 0.1209, 0.0598], + [-0.0015, -0.0202, -0.0977, ..., -0.2122, 0.0793, -0.1444], + [-0.0087, -0.0258, -0.0547, ..., 0.0345, -0.0943, 0.0166]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-08, ..., -2.5611e-08, + 2.0489e-08, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-07, ..., 2.0955e-08, + 4.8755e-07, 7.2643e-07], + [ 0.0000e+00, 0.0000e+00, -1.9707e-06, ..., 4.6566e-09, + 5.8534e-07, 6.1002e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 2.0163e-07, ..., 3.8650e-08, + -1.2927e-06, -1.5218e-06], + [ 0.0000e+00, 0.0000e+00, 5.7463e-07, ..., 4.3306e-08, + -1.1176e-08, 1.8906e-07], + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., -1.1455e-07, + 5.6345e-08, -2.0117e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0294, 0.0160, 0.0259, 0.0064, 0.0323, -0.0103, -0.0022, -0.0191, + -0.0210, 0.0156], device='cuda:0'), grad: tensor([ 1.7276e-07, 1.9204e-06, -7.5530e-07, 1.6373e-06, 1.2200e-07, + -2.1160e-06, 6.7800e-07, -3.5204e-06, 1.9390e-06, -8.4285e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 220.69, cls_loss 0.0020 cls_loss_mapping 0.0037 cls_loss_causal 0.5303 re_mapping 0.0058 re_causal 0.0167 /// teacc 98.92 lr 0.00010000 +Epoch 206, weight, value: tensor([[ 0.0069, 0.0110, 0.0259, ..., 0.0647, -0.1603, -0.1438], + [ 0.0504, -0.0185, -0.0073, ..., -0.1729, 0.0511, -0.0933], + [-0.0048, -0.0048, 0.1629, ..., -0.2502, -0.0917, -0.1342], + ..., + [-0.0135, 0.0062, -0.1348, ..., -0.0123, 0.1210, 0.0600], + [-0.0016, -0.0202, -0.0982, ..., -0.2132, 0.0789, -0.1453], + [-0.0087, -0.0258, -0.0549, ..., 0.0341, -0.0946, 0.0163]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.0175e-07, ..., -1.9558e-07, + 2.7940e-08, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-08, ..., 5.1223e-08, + -3.2410e-07, 1.1455e-07], + [ 0.0000e+00, 0.0000e+00, 5.3756e-06, ..., 5.4948e-08, + 2.6263e-07, 6.4261e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.2841e-08, ..., 1.2107e-08, + -1.3970e-08, -2.4214e-08], + [ 0.0000e+00, 0.0000e+00, -9.0338e-08, ..., 1.8626e-08, + -2.1420e-07, 2.7940e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-07, ..., -1.7695e-08, + 8.2888e-08, -3.9116e-08]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0292, 0.0159, 0.0268, 0.0059, 0.0330, -0.0095, -0.0026, -0.0193, + -0.0216, 0.0150], device='cuda:0'), grad: tensor([-1.1735e-07, -5.0012e-07, 1.0498e-05, -9.0152e-06, 6.7987e-08, + 3.0268e-07, 6.6962e-07, 9.3132e-08, -3.3714e-06, 1.3867e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 220.26, cls_loss 0.0021 cls_loss_mapping 0.0046 cls_loss_causal 0.5136 re_mapping 0.0058 re_causal 0.0162 /// teacc 98.97 lr 0.00010000 +Epoch 207, weight, value: tensor([[ 0.0068, 0.0110, 0.0260, ..., 0.0646, -0.1612, -0.1444], + [ 0.0502, -0.0185, -0.0074, ..., -0.1737, 0.0513, -0.0935], + [-0.0049, -0.0048, 0.1630, ..., -0.2524, -0.0921, -0.1344], + ..., + [-0.0143, 0.0062, -0.1349, ..., -0.0126, 0.1209, 0.0599], + [-0.0018, -0.0202, -0.0984, ..., -0.2145, 0.0790, -0.1471], + [-0.0087, -0.0258, -0.0551, ..., 0.0369, -0.0947, 0.0186]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.4124e-06, ..., -5.4277e-06, + -1.0906e-06, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 3.5483e-07, ..., 3.8836e-07, + 1.3039e-08, 1.5460e-07], + [ 0.0000e+00, 0.0000e+00, -1.0710e-06, ..., 2.6077e-07, + 9.9652e-08, 2.6077e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 6.3330e-07, ..., 7.8045e-07, + 1.7602e-07, 9.5926e-08], + [ 0.0000e+00, 0.0000e+00, 1.6298e-07, ..., 5.3272e-07, + -3.7253e-09, 5.5134e-07], + [ 0.0000e+00, 0.0000e+00, 1.2238e-06, ..., 5.2378e-06, + 3.5111e-07, 4.9956e-06]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0296, 0.0161, 0.0263, 0.0059, 0.0301, -0.0093, -0.0022, -0.0195, + -0.0223, 0.0177], device='cuda:0'), grad: tensor([-1.4707e-05, 1.2266e-06, -5.4762e-07, 1.4687e-06, -7.5474e-06, + 2.1327e-06, 1.0254e-06, 2.5779e-06, 1.3290e-06, 1.3039e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 220.77, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.5133 re_mapping 0.0056 re_causal 0.0170 /// teacc 98.96 lr 0.00010000 +Epoch 208, weight, value: tensor([[ 0.0068, 0.0110, 0.0284, ..., 0.0669, -0.1616, -0.1445], + [ 0.0502, -0.0185, -0.0086, ..., -0.1766, 0.0501, -0.0939], + [-0.0049, -0.0048, 0.1641, ..., -0.2534, -0.0924, -0.1346], + ..., + [-0.0153, 0.0062, -0.1350, ..., -0.0129, 0.1214, 0.0599], + [-0.0019, -0.0202, -0.0987, ..., -0.2151, 0.0790, -0.1474], + [-0.0088, -0.0258, -0.0571, ..., 0.0356, -0.0950, 0.0183]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-07, ..., -5.0012e-07, + 4.0047e-08, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 4.4797e-07, ..., 2.8871e-08, + 2.8238e-06, 1.6680e-06], + [ 0.0000e+00, 0.0000e+00, 3.7879e-05, ..., 3.5390e-08, + 5.2154e-07, 5.8919e-05], + ..., + [ 0.0000e+00, 0.0000e+00, -4.0680e-05, ..., 6.3330e-08, + -3.8221e-06, -6.4135e-05], + [ 0.0000e+00, 0.0000e+00, 1.2480e-07, ..., 9.1270e-08, + 7.6368e-08, 1.3132e-07], + [ 0.0000e+00, 0.0000e+00, 1.8720e-07, ..., 2.8219e-07, + 1.3784e-07, 1.0906e-06]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0273, 0.0142, 0.0269, 0.0062, 0.0310, -0.0094, -0.0008, -0.0193, + -0.0223, 0.0167], device='cuda:0'), grad: tensor([-9.6764e-07, 6.7502e-06, 1.6212e-04, 1.5888e-06, 6.3106e-06, + 4.8801e-07, 4.8988e-07, -1.7941e-04, 4.5728e-07, 1.8775e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 220.40, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.5218 re_mapping 0.0052 re_causal 0.0169 /// teacc 98.94 lr 0.00010000 +Epoch 209, weight, value: tensor([[ 0.0068, 0.0110, 0.0285, ..., 0.0670, -0.1621, -0.1447], + [ 0.0502, -0.0185, -0.0109, ..., -0.1767, 0.0500, -0.0941], + [-0.0049, -0.0048, 0.1660, ..., -0.2566, -0.0922, -0.1361], + ..., + [-0.0153, 0.0062, -0.1347, ..., -0.0126, 0.1215, 0.0601], + [-0.0019, -0.0202, -0.0988, ..., -0.2166, 0.0791, -0.1482], + [-0.0088, -0.0258, -0.0576, ..., 0.0356, -0.0949, 0.0184]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 1.1176e-08, + 4.3772e-08, 3.4459e-08], + [-3.7253e-09, 0.0000e+00, 5.5879e-09, ..., 4.8708e-07, + 1.1083e-07, 1.1586e-06], + [ 2.7940e-09, 0.0000e+00, -2.0023e-07, ..., 2.2352e-08, + 4.6380e-07, 6.5379e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.0245e-07, + -1.2591e-06, -1.5954e-06], + [ 0.0000e+00, 0.0000e+00, 6.4261e-08, ..., 3.0734e-08, + 5.8673e-08, 1.9372e-07], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 7.0594e-07, + 5.9605e-08, 1.1418e-06]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0274, 0.0132, 0.0281, 0.0063, 0.0311, -0.0100, -0.0004, -0.0192, + -0.0226, 0.0170], device='cuda:0'), grad: tensor([ 2.1048e-07, 1.8105e-06, 1.4994e-06, 2.1979e-06, -4.1984e-06, + -3.6787e-07, 6.4261e-07, -4.1984e-06, 3.8743e-07, 2.0005e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 220.79, cls_loss 0.0021 cls_loss_mapping 0.0030 cls_loss_causal 0.5299 re_mapping 0.0055 re_causal 0.0164 /// teacc 99.03 lr 0.00010000 +Epoch 210, weight, value: tensor([[ 0.0067, 0.0110, 0.0290, ..., 0.0671, -0.1625, -0.1457], + [ 0.0524, -0.0185, -0.0128, ..., -0.1773, 0.0497, -0.0941], + [-0.0063, -0.0048, 0.1678, ..., -0.2580, -0.0915, -0.1363], + ..., + [-0.0152, 0.0062, -0.1346, ..., -0.0127, 0.1217, 0.0602], + [-0.0020, -0.0202, -0.0988, ..., -0.2175, 0.0797, -0.1474], + [-0.0088, -0.0258, -0.0584, ..., 0.0347, -0.0952, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3962e-05, ..., -9.5740e-06, + -2.2314e-06, 7.0408e-07], + [ 0.0000e+00, 0.0000e+00, 2.0582e-07, ..., 1.8440e-07, + 2.8871e-08, 1.1548e-07], + [ 0.0000e+00, 0.0000e+00, 6.1654e-07, ..., 1.6838e-06, + 1.7509e-07, 4.3139e-06], + ..., + [-9.3132e-10, 0.0000e+00, 3.5018e-07, ..., 3.4925e-07, + 7.4506e-09, 2.6450e-07], + [ 0.0000e+00, 0.0000e+00, 8.4415e-06, ..., 3.8207e-05, + 1.5469e-06, 1.1456e-04], + [ 0.0000e+00, 0.0000e+00, 7.2643e-07, ..., -3.8534e-05, + 1.2852e-07, -1.3804e-04]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0274, 0.0121, 0.0294, 0.0060, 0.0323, -0.0096, -0.0010, -0.0191, + -0.0219, 0.0161], device='cuda:0'), grad: tensor([-3.5524e-05, 8.1677e-07, 1.4849e-05, 1.7639e-06, 1.0684e-05, + 4.3750e-05, 6.7391e-06, 1.5814e-06, 3.6764e-04, -4.1270e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 220.49, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.4993 re_mapping 0.0054 re_causal 0.0164 /// teacc 98.93 lr 0.00010000 +Epoch 211, weight, value: tensor([[ 0.0066, 0.0110, 0.0306, ..., 0.0679, -0.1628, -0.1475], + [ 0.0525, -0.0185, -0.0128, ..., -0.1775, 0.0498, -0.0938], + [-0.0064, -0.0048, 0.1681, ..., -0.2596, -0.0917, -0.1370], + ..., + [-0.0151, 0.0062, -0.1352, ..., -0.0129, 0.1217, 0.0601], + [-0.0021, -0.0202, -0.0994, ..., -0.2198, 0.0799, -0.1486], + [-0.0088, -0.0258, -0.0588, ..., 0.0339, -0.0954, 0.0164]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3027e-07, ..., -2.5239e-07, + 2.8126e-07, 1.8626e-09], + [-9.3132e-09, 0.0000e+00, 2.8126e-07, ..., 5.5879e-09, + 1.3225e-07, -1.3039e-08], + [ 1.8626e-09, 0.0000e+00, -2.3767e-06, ..., 1.9185e-07, + 4.2561e-07, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 7.0874e-07, ..., 1.8626e-08, + -9.3132e-09, -4.2841e-08], + [ 9.3132e-10, 0.0000e+00, 2.1979e-07, ..., 1.5832e-08, + -9.2108e-07, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 4.5635e-08, ..., 4.4703e-08, + 2.7008e-08, 5.9605e-08]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0264, 0.0123, 0.0293, 0.0060, 0.0337, -0.0097, -0.0012, -0.0193, + -0.0222, 0.0151], device='cuda:0'), grad: tensor([ 1.7704e-06, 6.7148e-07, -2.4699e-06, 2.7046e-06, 1.5274e-07, + 1.0356e-06, -2.9430e-06, 1.2685e-06, -2.4140e-06, 2.1141e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 220.36, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.4821 re_mapping 0.0057 re_causal 0.0167 /// teacc 98.99 lr 0.00010000 +Epoch 212, weight, value: tensor([[ 0.0064, 0.0110, 0.0315, ..., 0.0680, -0.1645, -0.1487], + [ 0.0532, -0.0185, -0.0129, ..., -0.1777, 0.0499, -0.0937], + [-0.0067, -0.0048, 0.1682, ..., -0.2603, -0.0919, -0.1373], + ..., + [-0.0151, 0.0062, -0.1354, ..., -0.0131, 0.1218, 0.0601], + [-0.0023, -0.0202, -0.0997, ..., -0.2209, 0.0800, -0.1493], + [-0.0088, -0.0258, -0.0590, ..., 0.0341, -0.0955, 0.0165]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0421e-06, ..., -8.0746e-07, + 2.9597e-06, 1.1176e-08], + [-9.3132e-10, 0.0000e+00, 9.8348e-06, ..., -4.2841e-08, + 2.6617e-06, -4.0978e-08], + [ 0.0000e+00, 0.0000e+00, -1.0408e-05, ..., 6.8918e-08, + -2.9802e-06, 4.1910e-08], + ..., + [-2.7940e-09, 0.0000e+00, 2.1141e-07, ..., 4.7497e-08, + -1.9558e-08, -5.8673e-08], + [ 0.0000e+00, 0.0000e+00, 5.2340e-07, ..., 1.6950e-07, + 2.7847e-07, 1.0235e-06], + [ 1.8626e-09, 0.0000e+00, 6.5099e-07, ..., 5.6997e-07, + -4.9360e-08, -6.4354e-07]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0270, 0.0125, 0.0293, 0.0048, 0.0336, -0.0087, -0.0008, -0.0194, + -0.0223, 0.0152], device='cuda:0'), grad: tensor([ 1.2361e-05, 1.3918e-05, -1.4052e-05, 3.4273e-06, 1.5292e-06, + -2.0131e-05, -1.4096e-05, 5.1409e-07, 1.3404e-05, 3.1237e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 220.43, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.5263 re_mapping 0.0052 re_causal 0.0158 /// teacc 99.03 lr 0.00010000 +Epoch 213, weight, value: tensor([[ 0.0062, 0.0110, 0.0349, ..., 0.0700, -0.1649, -0.1454], + [ 0.0536, -0.0185, -0.0128, ..., -0.1779, 0.0500, -0.0934], + [-0.0069, -0.0048, 0.1682, ..., -0.2619, -0.0922, -0.1379], + ..., + [-0.0153, 0.0062, -0.1360, ..., -0.0132, 0.1221, 0.0606], + [-0.0024, -0.0202, -0.1005, ..., -0.2222, 0.0800, -0.1499], + [-0.0088, -0.0258, -0.0616, ..., 0.0347, -0.0959, 0.0169]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., -7.4506e-09, + 9.3132e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 1.0245e-08, + -1.8626e-07, -1.5832e-08], + [ 0.0000e+00, 0.0000e+00, -7.0781e-08, ..., 1.8626e-09, + 5.7742e-08, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.8918e-08, ..., 1.3970e-08, + 4.2841e-08, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 2.6897e-06, ..., 5.1223e-08, + -8.8476e-08, 9.4995e-08], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., -4.5262e-06, + 1.4901e-08, -9.0376e-06]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0248, 0.0128, 0.0290, 0.0050, 0.0321, -0.0086, -0.0009, -0.0191, + -0.0228, 0.0151], device='cuda:0'), grad: tensor([ 8.1025e-08, -5.5321e-07, 3.7998e-07, -7.0669e-06, 1.2368e-05, + 1.5358e-06, -5.1223e-08, 2.8405e-07, 5.0142e-06, -1.2010e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 212---------------------------------------------------- +epoch 212, time 221.23, cls_loss 0.0016 cls_loss_mapping 0.0033 cls_loss_causal 0.4916 re_mapping 0.0054 re_causal 0.0159 /// teacc 99.15 lr 0.00010000 +Epoch 214, weight, value: tensor([[ 0.0062, 0.0110, 0.0356, ..., 0.0700, -0.1649, -0.1460], + [ 0.0538, -0.0185, -0.0128, ..., -0.1780, 0.0503, -0.0928], + [-0.0069, -0.0048, 0.1682, ..., -0.2630, -0.0926, -0.1382], + ..., + [-0.0153, 0.0062, -0.1363, ..., -0.0131, 0.1223, 0.0608], + [-0.0025, -0.0202, -0.1015, ..., -0.2230, 0.0806, -0.1505], + [-0.0088, -0.0258, -0.0618, ..., 0.0360, -0.0964, 0.0172]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1129e-06, ..., -2.4773e-07, + 1.5134e-06, 2.0675e-06], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 3.3304e-06, + 2.9523e-06, 8.9109e-06], + [-9.3132e-10, 0.0000e+00, -6.9849e-08, ..., 1.3690e-07, + 3.9861e-07, 4.4052e-07], + ..., + [ 9.3132e-10, 0.0000e+00, 5.4017e-08, ..., -2.8205e-04, + -2.9683e-04, -7.4244e-04], + [ 0.0000e+00, 0.0000e+00, 1.5832e-07, ..., 4.5449e-07, + 2.9504e-05, 1.4398e-06], + [ 0.0000e+00, 0.0000e+00, 7.5437e-08, ..., 2.5415e-04, + 2.7633e-04, 6.7854e-04]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0246, 0.0133, 0.0288, 0.0051, 0.0315, -0.0085, -0.0014, -0.0190, + -0.0227, 0.0152], device='cuda:0'), grad: tensor([ 1.5954e-06, 1.4313e-05, 1.2405e-06, 6.6794e-06, 7.0035e-05, + 1.5974e-05, -5.7399e-05, -1.3552e-03, 6.2644e-05, 1.2388e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 220.20, cls_loss 0.0015 cls_loss_mapping 0.0036 cls_loss_causal 0.5394 re_mapping 0.0053 re_causal 0.0164 /// teacc 98.99 lr 0.00010000 +Epoch 215, weight, value: tensor([[ 0.0061, 0.0110, 0.0354, ..., 0.0697, -0.1653, -0.1463], + [ 0.0554, -0.0185, -0.0126, ..., -0.1783, 0.0508, -0.0929], + [-0.0078, -0.0048, 0.1689, ..., -0.2633, -0.0927, -0.1381], + ..., + [-0.0153, 0.0062, -0.1372, ..., -0.0124, 0.1222, 0.0613], + [-0.0026, -0.0202, -0.1024, ..., -0.2234, 0.0809, -0.1505], + [-0.0088, -0.0258, -0.0617, ..., 0.0358, -0.0972, 0.0168]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.1665e-07, ..., 1.4715e-07, + 3.4459e-08, 2.0489e-07], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 4.0978e-08, + -7.0781e-08, 7.2643e-08], + [ 0.0000e+00, 0.0000e+00, -5.9232e-07, ..., 4.3772e-08, + 8.0094e-08, 6.7987e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 1.2666e-07, + -6.5193e-09, 2.3190e-07], + [ 0.0000e+00, 0.0000e+00, 4.0047e-08, ..., 2.4214e-08, + -1.1362e-06, 3.7253e-08], + [ 0.0000e+00, 0.0000e+00, -3.2596e-08, ..., -4.0606e-07, + 1.9558e-08, -8.4378e-07]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0250, 0.0136, 0.0291, 0.0051, 0.0316, -0.0085, -0.0017, -0.0189, + -0.0227, 0.0147], device='cuda:0'), grad: tensor([ 7.9256e-07, -4.6566e-08, -3.4645e-07, 5.3179e-07, 1.6205e-07, + 1.7714e-06, -2.1979e-07, 3.5577e-07, -1.9353e-06, -1.0720e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 220.63, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5145 re_mapping 0.0053 re_causal 0.0166 /// teacc 98.91 lr 0.00010000 +Epoch 216, weight, value: tensor([[ 0.0060, 0.0110, 0.0357, ..., 0.0699, -0.1656, -0.1463], + [ 0.0554, -0.0185, -0.0125, ..., -0.1785, 0.0509, -0.0930], + [-0.0078, -0.0048, 0.1688, ..., -0.2642, -0.0933, -0.1384], + ..., + [-0.0153, 0.0062, -0.1373, ..., -0.0123, 0.1223, 0.0615], + [-0.0027, -0.0202, -0.1025, ..., -0.2239, 0.0813, -0.1506], + [-0.0089, -0.0258, -0.0620, ..., 0.0368, -0.0974, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.5437e-08, ..., 1.7472e-06, + 5.5879e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 2.7008e-08, + -1.0245e-08, 6.9849e-08], + [-9.3132e-10, 0.0000e+00, -1.8654e-06, ..., 1.3970e-08, + 7.4506e-09, 8.3819e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.3772e-08, ..., 1.3970e-08, + -7.7300e-08, -8.3819e-08], + [ 0.0000e+00, 0.0000e+00, 1.3746e-06, ..., 4.9360e-08, + 1.2107e-08, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 6.9849e-08, ..., 2.9150e-07, + 3.4459e-08, 4.0419e-07]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0247, 0.0138, 0.0287, 0.0065, 0.0308, -0.0099, -0.0020, -0.0187, + -0.0225, 0.0152], device='cuda:0'), grad: tensor([ 1.0513e-05, 3.7253e-08, -2.6785e-06, 7.2084e-07, 4.2375e-07, + -4.0978e-07, -1.1727e-05, -7.3574e-08, 2.3916e-06, 7.9162e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 220.69, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.4950 re_mapping 0.0051 re_causal 0.0155 /// teacc 99.01 lr 0.00010000 +Epoch 217, weight, value: tensor([[ 0.0059, 0.0110, 0.0360, ..., 0.0704, -0.1659, -0.1464], + [ 0.0568, -0.0185, -0.0123, ..., -0.1787, 0.0511, -0.0929], + [-0.0086, -0.0048, 0.1690, ..., -0.2645, -0.0935, -0.1385], + ..., + [-0.0153, 0.0062, -0.1374, ..., -0.0124, 0.1225, 0.0620], + [-0.0031, -0.0202, -0.1045, ..., -0.2242, 0.0811, -0.1513], + [-0.0089, -0.0258, -0.0621, ..., 0.0370, -0.0977, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3319e-07, ..., -6.9290e-07, + 5.7742e-08, 3.9116e-08], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 2.1420e-08, + 6.5118e-06, 2.5909e-06], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 1.6764e-08, + 5.1223e-08, 3.0734e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 2.4308e-07, + -9.4101e-06, -2.8722e-06], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 5.4948e-08, + 1.2843e-06, 1.5413e-06], + [ 0.0000e+00, 0.0000e+00, 4.3772e-08, ..., -2.8554e-06, + 8.0466e-07, -8.1807e-06]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0258, 0.0142, 0.0288, 0.0077, 0.0306, -0.0102, -0.0018, -0.0184, + -0.0253, 0.0151], device='cuda:0'), grad: tensor([-2.9318e-06, 1.5378e-05, 3.4645e-07, 1.2275e-06, 1.0334e-05, + -2.3898e-06, 2.4680e-06, -2.0757e-05, 2.7120e-06, -6.4448e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 220.35, cls_loss 0.0022 cls_loss_mapping 0.0041 cls_loss_causal 0.5101 re_mapping 0.0055 re_causal 0.0153 /// teacc 98.91 lr 0.00010000 +Epoch 218, weight, value: tensor([[ 0.0059, 0.0110, 0.0362, ..., 0.0713, -0.1668, -0.1452], + [ 0.0568, -0.0185, -0.0124, ..., -0.1792, 0.0508, -0.0932], + [-0.0087, -0.0048, 0.1696, ..., -0.2650, -0.0939, -0.1376], + ..., + [-0.0153, 0.0062, -0.1375, ..., -0.0126, 0.1230, 0.0627], + [-0.0031, -0.0202, -0.1048, ..., -0.2248, 0.0823, -0.1508], + [-0.0089, -0.0258, -0.0642, ..., 0.0365, -0.0983, 0.0166]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.2655e-07, ..., -7.6927e-07, + 2.9150e-07, 1.0338e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 8.6613e-08, + 4.4052e-07, 1.7136e-06], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 2.7940e-08, + 1.7015e-06, 4.8336e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 1.9465e-07, + -1.3366e-05, -4.5121e-05], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 5.2154e-08, + -2.1696e-05, 3.9116e-07], + [ 0.0000e+00, 0.0000e+00, 6.0536e-08, ..., 2.7660e-07, + 1.7643e-05, 4.0591e-05]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0241, 0.0139, 0.0289, 0.0075, 0.0307, -0.0097, -0.0028, -0.0177, + -0.0247, 0.0137], device='cuda:0'), grad: tensor([-4.3958e-07, 3.4533e-06, 8.4043e-06, 1.3731e-05, 1.5348e-05, + 3.7551e-05, 3.1292e-06, -7.8619e-05, -1.0478e-04, 1.0234e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 220.67, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5102 re_mapping 0.0056 re_causal 0.0167 /// teacc 98.95 lr 0.00010000 +Epoch 219, weight, value: tensor([[ 0.0058, 0.0110, 0.0360, ..., 0.0713, -0.1677, -0.1454], + [ 0.0568, -0.0185, -0.0124, ..., -0.1794, 0.0509, -0.0933], + [-0.0084, -0.0048, 0.1698, ..., -0.2655, -0.0943, -0.1379], + ..., + [-0.0155, 0.0062, -0.1378, ..., -0.0128, 0.1231, 0.0628], + [-0.0032, -0.0202, -0.1047, ..., -0.2253, 0.0840, -0.1497], + [-0.0089, -0.0258, -0.0643, ..., 0.0365, -0.0989, 0.0165]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.3660e-06, ..., 1.3039e-08, + 3.1386e-07, 2.8871e-07], + [ 0.0000e+00, 0.0000e+00, 4.0047e-08, ..., 3.0734e-08, + 8.5961e-07, 3.9209e-07], + [ 0.0000e+00, 0.0000e+00, -8.4043e-06, ..., 6.5193e-09, + 1.4901e-07, 1.1362e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2852e-07, ..., -1.2573e-07, + -2.7083e-06, -3.3304e-06], + [ 0.0000e+00, 0.0000e+00, 8.2888e-08, ..., 3.3826e-06, + 4.1574e-06, 1.0505e-05], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 4.3772e-08, + 1.8142e-06, 2.0899e-06]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0242, 0.0139, 0.0289, 0.0074, 0.0306, -0.0099, -0.0025, -0.0177, + -0.0234, 0.0134], device='cuda:0'), grad: tensor([ 7.0110e-06, 1.8217e-06, -1.1608e-05, 5.6177e-06, -2.1771e-05, + 1.2346e-05, -1.8522e-05, -6.2063e-06, 2.7150e-05, 4.1947e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 220.79, cls_loss 0.0020 cls_loss_mapping 0.0031 cls_loss_causal 0.5388 re_mapping 0.0053 re_causal 0.0166 /// teacc 99.06 lr 0.00010000 +Epoch 220, weight, value: tensor([[ 0.0057, 0.0110, 0.0360, ..., 0.0708, -0.1682, -0.1457], + [ 0.0568, -0.0185, -0.0125, ..., -0.1802, 0.0486, -0.0958], + [-0.0083, -0.0048, 0.1709, ..., -0.2661, -0.0938, -0.1379], + ..., + [-0.0186, 0.0062, -0.1401, ..., -0.0129, 0.1246, 0.0634], + [-0.0039, -0.0202, -0.1050, ..., -0.2269, 0.0845, -0.1513], + [-0.0088, -0.0258, -0.0643, ..., 0.0365, -0.0986, 0.0167]], + device='cuda:0'), grad: tensor([[ 3.8370e-07, 0.0000e+00, 2.1420e-08, ..., 4.9453e-07, + 1.2945e-07, 1.4957e-06], + [ 3.9674e-07, 0.0000e+00, 4.0047e-07, ..., 7.8417e-07, + 3.2410e-07, 2.3991e-06], + [-1.7136e-07, 0.0000e+00, -2.3283e-06, ..., 3.9209e-07, + 8.5682e-07, 1.9651e-06], + ..., + [ 3.1367e-06, 0.0000e+00, 6.4448e-07, ..., 6.6236e-06, + -1.5358e-06, 1.4283e-05], + [ 2.4773e-06, 0.0000e+00, 5.8394e-07, ..., 3.7141e-06, + 2.1420e-07, 9.6858e-06], + [-6.1810e-05, 0.0000e+00, 4.4703e-08, ..., -1.3864e-04, + 8.6892e-07, -3.4285e-04]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0248, 0.0119, 0.0297, 0.0086, 0.0308, -0.0100, -0.0021, -0.0180, + -0.0234, 0.0139], device='cuda:0'), grad: tensor([ 6.3814e-06, 6.5155e-06, 1.5181e-06, 1.1235e-05, 3.9649e-04, + 3.2520e-04, -3.3081e-05, 2.3529e-05, 2.7224e-05, -7.6485e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 220.35, cls_loss 0.0019 cls_loss_mapping 0.0026 cls_loss_causal 0.5160 re_mapping 0.0055 re_causal 0.0156 /// teacc 98.96 lr 0.00010000 +Epoch 221, weight, value: tensor([[ 0.0051, 0.0110, 0.0360, ..., 0.0710, -0.1691, -0.1460], + [ 0.0567, -0.0185, -0.0125, ..., -0.1806, 0.0487, -0.0958], + [-0.0083, -0.0048, 0.1710, ..., -0.2668, -0.0946, -0.1389], + ..., + [-0.0186, 0.0062, -0.1404, ..., -0.0132, 0.1248, 0.0638], + [-0.0047, -0.0202, -0.1043, ..., -0.2278, 0.0850, -0.1520], + [-0.0074, -0.0258, -0.0644, ..., 0.0366, -0.0988, 0.0167]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, 1.2759e-07, ..., -7.5474e-06, + 3.8184e-08, 5.0291e-08], + [ 9.3132e-10, 0.0000e+00, -8.1025e-08, ..., 6.5938e-07, + -4.0978e-07, 1.7909e-06], + [-3.4459e-07, 0.0000e+00, -3.0100e-06, ..., 1.1921e-07, + 3.4273e-07, 1.2107e-07], + ..., + [ 1.1548e-07, 0.0000e+00, 1.7406e-06, ..., 7.0129e-07, + -1.3122e-06, -4.6007e-07], + [ 8.3819e-09, 0.0000e+00, 5.1316e-07, ..., 1.0338e-07, + 2.1420e-07, 2.0862e-07], + [ 6.5193e-09, 0.0000e+00, 2.2165e-07, ..., 1.0729e-05, + 1.1483e-06, 1.3523e-05]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0248, 0.0120, 0.0288, 0.0084, 0.0305, -0.0098, -0.0029, -0.0174, + -0.0229, 0.0140], device='cuda:0'), grad: tensor([-1.7583e-05, 1.7881e-07, -2.7344e-06, -3.9395e-07, -1.7509e-05, + 8.8383e-07, 1.7080e-06, 2.0321e-06, 2.0899e-06, 3.1322e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 220.46, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.5336 re_mapping 0.0055 re_causal 0.0166 /// teacc 98.90 lr 0.00010000 +Epoch 222, weight, value: tensor([[ 0.0049, 0.0110, 0.0360, ..., 0.0712, -0.1703, -0.1461], + [ 0.0566, -0.0185, -0.0124, ..., -0.1812, 0.0486, -0.0961], + [-0.0071, -0.0048, 0.1714, ..., -0.2670, -0.0949, -0.1391], + ..., + [-0.0187, 0.0062, -0.1409, ..., -0.0137, 0.1249, 0.0636], + [-0.0050, -0.0202, -0.1051, ..., -0.2286, 0.0854, -0.1527], + [-0.0067, -0.0258, -0.0645, ..., 0.0366, -0.0991, 0.0163]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., -3.4459e-08, + 6.5193e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -1.3240e-05, ..., 9.3132e-09, + -3.5018e-07, -1.1455e-07], + [ 0.0000e+00, 0.0000e+00, 8.5384e-06, ..., 2.7940e-09, + 1.7881e-07, 8.2888e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.1237e-06, ..., 2.6077e-08, + 3.6322e-08, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.6566e-09, + 1.3970e-08, 5.4017e-08], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 6.5193e-09, + 3.1665e-08, -1.3877e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0250, 0.0120, 0.0287, 0.0083, 0.0315, -0.0098, -0.0032, -0.0176, + -0.0227, 0.0137], device='cuda:0'), grad: tensor([ 4.5635e-08, -7.4029e-05, 4.7863e-05, 8.2105e-06, 3.8091e-07, + 7.4506e-08, -3.5390e-08, 1.7345e-05, 1.9465e-07, 0.0000e+00], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 220.33, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.4795 re_mapping 0.0059 re_causal 0.0162 /// teacc 98.97 lr 0.00010000 +Epoch 223, weight, value: tensor([[ 0.0048, 0.0110, 0.0363, ..., 0.0712, -0.1709, -0.1463], + [ 0.0575, -0.0185, -0.0124, ..., -0.1826, 0.0494, -0.0965], + [-0.0078, -0.0048, 0.1714, ..., -0.2688, -0.0965, -0.1389], + ..., + [-0.0186, 0.0062, -0.1406, ..., -0.0138, 0.1247, 0.0637], + [-0.0054, -0.0202, -0.1050, ..., -0.2300, 0.0858, -0.1536], + [-0.0063, -0.0258, -0.0657, ..., 0.0364, -0.0994, 0.0164]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.1816e-07, ..., 1.8161e-07, + 5.5041e-07, 7.4133e-07], + [ 0.0000e+00, 0.0000e+00, 1.0803e-07, ..., 2.7101e-07, + -1.1310e-05, -4.2394e-06], + [ 0.0000e+00, 0.0000e+00, 1.1520e-06, ..., 4.1071e-07, + 1.1474e-06, 1.4324e-06], + ..., + [ 0.0000e+00, 0.0000e+00, -1.7863e-06, ..., 7.0501e-07, + -2.0899e-06, -1.6401e-06], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 2.8219e-07, + 6.7707e-07, 5.4762e-07], + [-9.3132e-10, 0.0000e+00, 2.1979e-07, ..., 1.7369e-06, + 4.1258e-07, 1.6997e-06]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0256, 0.0122, 0.0271, 0.0080, 0.0321, -0.0098, -0.0020, -0.0171, + -0.0225, 0.0132], device='cuda:0'), grad: tensor([ 2.0787e-06, -2.6256e-05, 5.9083e-06, 1.4342e-07, -8.2552e-06, + 2.7940e-07, 2.7820e-05, -8.0839e-06, 2.2389e-06, 4.1313e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 220.59, cls_loss 0.0021 cls_loss_mapping 0.0026 cls_loss_causal 0.5311 re_mapping 0.0057 re_causal 0.0169 /// teacc 98.91 lr 0.00010000 +Epoch 224, weight, value: tensor([[ 0.0045, 0.0110, 0.0366, ..., 0.0714, -0.1713, -0.1465], + [ 0.0594, -0.0185, -0.0122, ..., -0.1828, 0.0493, -0.0967], + [-0.0094, -0.0048, 0.1717, ..., -0.2703, -0.0973, -0.1384], + ..., + [-0.0186, 0.0062, -0.1407, ..., -0.0143, 0.1252, 0.0635], + [-0.0061, -0.0202, -0.1072, ..., -0.2311, 0.0860, -0.1547], + [-0.0060, -0.0258, -0.0660, ..., 0.0360, -0.0997, 0.0132]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.5751e-07, ..., -3.6322e-08, + 1.9837e-07, 1.3877e-07], + [ 0.0000e+00, 0.0000e+00, 4.9360e-08, ..., 4.6566e-09, + 3.9209e-07, 8.6240e-07], + [ 0.0000e+00, 0.0000e+00, -1.0990e-07, ..., 8.3819e-09, + 2.3004e-07, 3.8929e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.1176e-08, + -1.4398e-06, 7.0743e-06], + [ 0.0000e+00, 0.0000e+00, 2.1420e-07, ..., 7.7300e-08, + -5.2247e-07, 5.5041e-07], + [ 0.0000e+00, 0.0000e+00, 3.0827e-07, ..., -2.1327e-07, + 4.3120e-07, 5.9418e-07]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0256, 0.0120, 0.0259, 0.0079, 0.0353, -0.0096, -0.0023, -0.0161, + -0.0232, 0.0103], device='cuda:0'), grad: tensor([ 6.6869e-07, 1.9819e-06, 2.0228e-06, 3.6433e-06, 1.0297e-05, + -4.2543e-06, -6.9916e-05, 4.8250e-05, 2.6673e-06, 4.5411e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 220.34, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4962 re_mapping 0.0056 re_causal 0.0163 /// teacc 98.94 lr 0.00010000 +Epoch 225, weight, value: tensor([[ 0.0044, 0.0110, 0.0363, ..., 0.0706, -0.1717, -0.1474], + [ 0.0595, -0.0185, -0.0122, ..., -0.1830, 0.0494, -0.0965], + [-0.0094, -0.0048, 0.1720, ..., -0.2707, -0.0975, -0.1386], + ..., + [-0.0186, 0.0062, -0.1409, ..., -0.0144, 0.1253, 0.0635], + [-0.0062, -0.0202, -0.1083, ..., -0.2318, 0.0859, -0.1554], + [-0.0060, -0.0258, -0.0657, ..., 0.0360, -0.1000, 0.0128]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., -1.4901e-08, + 9.3132e-10, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 2.7940e-09, + 2.1420e-08, 3.3528e-08], + [-9.3132e-10, 0.0000e+00, -4.2748e-07, ..., 0.0000e+00, + 3.7253e-09, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 7.7300e-08, ..., 3.0734e-08, + -5.2154e-08, 2.2165e-07], + [ 0.0000e+00, 0.0000e+00, 1.0245e-07, ..., 7.0781e-08, + 0.0000e+00, 6.2492e-07], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -1.6671e-07, + 9.3132e-09, -1.4007e-06]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0264, 0.0121, 0.0260, 0.0078, 0.0358, -0.0095, -0.0022, -0.0160, + -0.0234, 0.0099], device='cuda:0'), grad: tensor([-3.0734e-08, 1.3411e-07, -6.7428e-07, 2.2911e-07, 5.9512e-07, + -1.2573e-07, 1.4156e-07, 3.5670e-07, 1.0272e-06, -1.6615e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 220.68, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4909 re_mapping 0.0053 re_causal 0.0163 /// teacc 98.93 lr 0.00010000 +Epoch 226, weight, value: tensor([[ 0.0044, 0.0110, 0.0363, ..., 0.0702, -0.1722, -0.1478], + [ 0.0595, -0.0185, -0.0146, ..., -0.1832, 0.0486, -0.0966], + [-0.0094, -0.0048, 0.1745, ..., -0.2711, -0.0954, -0.1387], + ..., + [-0.0186, 0.0062, -0.1414, ..., -0.0145, 0.1255, 0.0638], + [-0.0062, -0.0202, -0.1088, ..., -0.2325, 0.0861, -0.1561], + [-0.0060, -0.0258, -0.0659, ..., 0.0369, -0.1002, 0.0131]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -6.8452e-08, ..., -1.8673e-07, + 4.7497e-08, 2.7008e-08], + [-6.0536e-09, 0.0000e+00, 6.9384e-08, ..., 8.6566e-07, + -1.2275e-06, 1.9595e-06], + [ 1.0980e-06, 0.0000e+00, 5.8264e-06, ..., 2.6077e-08, + 3.8045e-07, 2.4121e-07], + ..., + [ 8.6613e-08, 0.0000e+00, 7.8324e-07, ..., 1.8785e-06, + -1.5181e-07, 3.0305e-06], + [ 5.1223e-09, 0.0000e+00, 2.1188e-07, ..., 3.4925e-08, + -2.1420e-08, 8.3819e-08], + [ 1.3970e-09, 0.0000e+00, 6.1467e-08, ..., 1.1707e-06, + 2.9849e-07, 2.7083e-06]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0268, 0.0106, 0.0280, 0.0078, 0.0353, -0.0093, -0.0024, -0.0159, + -0.0236, 0.0103], device='cuda:0'), grad: tensor([-1.4389e-07, 5.6159e-07, 3.1292e-05, -3.3140e-05, -1.2167e-05, + -1.8906e-07, 6.3563e-07, 7.9200e-06, 4.0326e-07, 4.8019e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 220.42, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5187 re_mapping 0.0053 re_causal 0.0155 /// teacc 98.98 lr 0.00010000 +Epoch 227, weight, value: tensor([[ 0.0043, 0.0110, 0.0358, ..., 0.0694, -0.1731, -0.1484], + [ 0.0594, -0.0185, -0.0146, ..., -0.1837, 0.0482, -0.0971], + [-0.0098, -0.0048, 0.1749, ..., -0.2715, -0.0959, -0.1391], + ..., + [-0.0186, 0.0062, -0.1416, ..., -0.0150, 0.1262, 0.0644], + [-0.0064, -0.0202, -0.1092, ..., -0.2337, 0.0870, -0.1560], + [-0.0061, -0.0258, -0.0661, ..., 0.0380, -0.1006, 0.0134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4529e-06, ..., -2.2817e-08, + 3.1665e-08, 5.0105e-07], + [ 0.0000e+00, 0.0000e+00, 5.2620e-08, ..., 1.2154e-07, + -3.8415e-05, -2.0072e-05], + [ 0.0000e+00, 0.0000e+00, -2.8722e-06, ..., 3.2596e-09, + 1.5600e-07, -1.2023e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2610e-06, ..., 8.9407e-08, + 3.4302e-05, 1.8179e-05], + [ 0.0000e+00, 0.0000e+00, 1.5926e-06, ..., 9.3132e-09, + 2.0023e-07, 1.5879e-07], + [ 0.0000e+00, 0.0000e+00, 1.6298e-07, ..., 1.2852e-07, + 3.2447e-06, 1.9986e-06]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0276, 0.0104, 0.0278, 0.0075, 0.0348, -0.0094, -0.0020, -0.0152, + -0.0233, 0.0107], device='cuda:0'), grad: tensor([ 2.4699e-06, -8.4341e-05, 3.0138e-06, 1.3024e-05, 1.5777e-06, + -4.1515e-05, -4.6566e-10, 8.4519e-05, 1.3717e-05, 7.5251e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 220.59, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.5139 re_mapping 0.0054 re_causal 0.0160 /// teacc 98.99 lr 0.00010000 +Epoch 228, weight, value: tensor([[ 0.0038, 0.0110, 0.0359, ..., 0.0693, -0.1735, -0.1487], + [ 0.0607, -0.0185, -0.0146, ..., -0.1840, 0.0483, -0.0972], + [-0.0103, -0.0048, 0.1750, ..., -0.2723, -0.0961, -0.1392], + ..., + [-0.0186, 0.0062, -0.1420, ..., -0.0162, 0.1261, 0.0639], + [-0.0065, -0.0202, -0.1095, ..., -0.2343, 0.0883, -0.1565], + [-0.0061, -0.0258, -0.0662, ..., 0.0382, -0.1008, 0.0136]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 1.3504e-08, + 5.6345e-08, 2.1420e-08], + [ 0.0000e+00, 0.0000e+00, 2.4680e-08, ..., 1.9744e-07, + 6.6636e-07, 4.8522e-07], + [ 0.0000e+00, 0.0000e+00, -1.2200e-07, ..., 1.0710e-08, + 6.4708e-06, 6.1607e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 5.7742e-08, ..., 3.5716e-07, + -8.1584e-06, -2.0256e-07], + [ 0.0000e+00, 0.0000e+00, 1.7229e-08, ..., 1.9837e-07, + 1.4156e-07, 3.3621e-07], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 1.3951e-06, + 5.3784e-07, 2.1756e-06]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0277, 0.0105, 0.0277, 0.0076, 0.0349, -0.0095, -0.0022, -0.0155, + -0.0228, 0.0109], device='cuda:0'), grad: tensor([ 2.8871e-07, 2.9095e-06, 2.1815e-05, 1.6978e-06, -4.7237e-06, + -2.0891e-05, 1.9418e-07, -2.6047e-05, 2.0295e-05, 4.4852e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 220.53, cls_loss 0.0020 cls_loss_mapping 0.0028 cls_loss_causal 0.5197 re_mapping 0.0057 re_causal 0.0157 /// teacc 99.00 lr 0.00010000 +Epoch 229, weight, value: tensor([[ 0.0037, 0.0110, 0.0360, ..., 0.0693, -0.1741, -0.1491], + [ 0.0611, -0.0185, -0.0147, ..., -0.1845, 0.0489, -0.0971], + [-0.0103, -0.0048, 0.1757, ..., -0.2729, -0.0967, -0.1390], + ..., + [-0.0186, 0.0062, -0.1427, ..., -0.0166, 0.1262, 0.0638], + [-0.0074, -0.0202, -0.1089, ..., -0.2344, 0.0885, -0.1569], + [-0.0062, -0.0258, -0.0663, ..., 0.0382, -0.1012, 0.0135]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2573e-08, ..., -2.7940e-09, + 5.5879e-09, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 5.6438e-07, + -4.2375e-08, 5.1083e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 8.9407e-08, + 1.1176e-08, 8.1491e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 1.9930e-07, + 1.1642e-08, 1.8347e-07], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 3.0268e-08, + -1.9558e-08, 4.9360e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 4.8615e-07, + 6.5193e-09, 3.4180e-07]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0278, 0.0113, 0.0278, 0.0072, 0.0354, -0.0092, -0.0046, -0.0157, + -0.0227, 0.0107], device='cuda:0'), grad: tensor([ 1.8161e-08, 9.0571e-07, 2.2119e-07, 8.9500e-07, -2.5295e-06, + -8.6566e-07, 6.5658e-08, 3.9767e-07, 9.4995e-08, 7.9069e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 220.21, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.4832 re_mapping 0.0055 re_causal 0.0155 /// teacc 98.97 lr 0.00010000 +Epoch 230, weight, value: tensor([[ 0.0037, 0.0110, 0.0341, ..., 0.0694, -0.1745, -0.1492], + [ 0.0611, -0.0185, -0.0151, ..., -0.1848, 0.0494, -0.0971], + [-0.0103, -0.0048, 0.1762, ..., -0.2733, -0.0969, -0.1392], + ..., + [-0.0186, 0.0062, -0.1432, ..., -0.0167, 0.1265, 0.0639], + [-0.0074, -0.0202, -0.1096, ..., -0.2351, 0.0869, -0.1575], + [-0.0062, -0.0258, -0.0666, ..., 0.0383, -0.1015, 0.0136]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2852e-07, ..., -9.3132e-08, + 1.4715e-07, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, -3.7998e-06, ..., 2.2817e-08, + -2.3752e-05, -5.1521e-06], + [ 0.0000e+00, 0.0000e+00, 1.7704e-06, ..., 1.9558e-08, + 4.8392e-06, 1.8682e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 2.0172e-06, ..., 5.0291e-08, + 5.7220e-06, 2.3823e-06], + [ 0.0000e+00, 0.0000e+00, -6.3749e-07, ..., 5.7276e-08, + 9.3281e-06, 8.3260e-07], + [ 0.0000e+00, 0.0000e+00, 1.2154e-07, ..., -1.9558e-07, + 3.0221e-07, -5.4296e-07]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0289, 0.0118, 0.0280, 0.0074, 0.0354, -0.0089, -0.0050, -0.0156, + -0.0242, 0.0107], device='cuda:0'), grad: tensor([ 1.8161e-08, -6.3777e-05, 1.4991e-05, 1.0334e-05, 8.3633e-07, + 2.9821e-06, 1.0878e-06, 1.7002e-05, 1.6645e-05, -1.1874e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 220.66, cls_loss 0.0016 cls_loss_mapping 0.0031 cls_loss_causal 0.5351 re_mapping 0.0051 re_causal 0.0160 /// teacc 98.89 lr 0.00010000 +Epoch 231, weight, value: tensor([[ 0.0037, 0.0110, 0.0331, ..., 0.0696, -0.1752, -0.1493], + [ 0.0611, -0.0185, -0.0166, ..., -0.1852, 0.0494, -0.0975], + [-0.0103, -0.0048, 0.1780, ..., -0.2746, -0.0964, -0.1388], + ..., + [-0.0186, 0.0062, -0.1436, ..., -0.0165, 0.1267, 0.0641], + [-0.0074, -0.0202, -0.1108, ..., -0.2392, 0.0855, -0.1598], + [-0.0062, -0.0258, -0.0666, ..., 0.0386, -0.1016, 0.0136]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.2201e-08, ..., -6.1467e-08, + 7.0408e-07, 1.1781e-07], + [ 0.0000e+00, 0.0000e+00, 1.8450e-06, ..., 5.5879e-08, + -3.8855e-06, -1.4622e-07], + [ 0.0000e+00, 0.0000e+00, -6.3330e-06, ..., 3.3062e-08, + 5.7695e-07, 9.3598e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.8452e-07, ..., -2.3283e-09, + 1.7602e-06, -2.0023e-07], + [ 0.0000e+00, 0.0000e+00, 2.3330e-07, ..., 5.0757e-08, + -5.3318e-07, 2.3982e-07], + [ 0.0000e+00, 0.0000e+00, 4.8429e-08, ..., 3.0268e-08, + 4.2329e-07, 4.6100e-08]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0293, 0.0112, 0.0291, 0.0074, 0.0353, -0.0089, -0.0044, -0.0155, + -0.0259, 0.0109], device='cuda:0'), grad: tensor([ 4.1015e-06, -1.1392e-05, -9.9838e-06, 6.7875e-06, 1.3318e-06, + 7.2923e-07, 2.3097e-06, 4.2096e-06, 8.1817e-07, 1.0598e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 220.95, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.5078 re_mapping 0.0052 re_causal 0.0150 /// teacc 99.02 lr 0.00010000 +Epoch 232, weight, value: tensor([[ 0.0037, 0.0110, 0.0331, ..., 0.0699, -0.1760, -0.1497], + [ 0.0610, -0.0185, -0.0171, ..., -0.1855, 0.0499, -0.0973], + [-0.0104, -0.0048, 0.1785, ..., -0.2756, -0.0969, -0.1390], + ..., + [-0.0186, 0.0062, -0.1438, ..., -0.0169, 0.1266, 0.0641], + [-0.0075, -0.0202, -0.1112, ..., -0.2396, 0.0856, -0.1604], + [-0.0058, -0.0258, -0.0667, ..., 0.0384, -0.1019, 0.0135]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -7.4506e-09, + 6.5658e-08, 3.7253e-09], + [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 2.2817e-08, + -2.4900e-05, 5.1223e-08], + [ 4.6566e-10, 0.0000e+00, 1.9139e-07, ..., 3.7253e-09, + 3.2177e-07, 1.2107e-08], + ..., + [-4.6566e-09, 0.0000e+00, 1.2107e-08, ..., 2.0489e-08, + 2.3738e-05, -2.0023e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 3.2596e-09, + 1.1362e-07, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., -2.6543e-08, + 6.1933e-08, -1.1781e-07]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0293, 0.0113, 0.0292, 0.0074, 0.0355, -0.0091, -0.0037, -0.0156, + -0.0261, 0.0108], device='cuda:0'), grad: tensor([ 2.3143e-07, -4.4286e-05, 9.7416e-07, 1.3551e-07, 7.7300e-08, + 1.1856e-06, -9.5135e-07, 4.2349e-05, 3.3434e-07, 2.3283e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 220.54, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.4761 re_mapping 0.0051 re_causal 0.0150 /// teacc 98.97 lr 0.00010000 +Epoch 233, weight, value: tensor([[ 0.0034, 0.0110, 0.0331, ..., 0.0701, -0.1777, -0.1498], + [ 0.0592, -0.0185, -0.0171, ..., -0.1859, 0.0502, -0.0975], + [-0.0110, -0.0048, 0.1788, ..., -0.2777, -0.0970, -0.1395], + ..., + [-0.0185, 0.0062, -0.1446, ..., -0.0148, 0.1267, 0.0654], + [-0.0076, -0.0202, -0.1118, ..., -0.2405, 0.0853, -0.1612], + [-0.0059, -0.0258, -0.0666, ..., 0.0377, -0.1022, 0.0131]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0571e-06, ..., -1.2163e-06, + 1.9558e-08, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 5.1223e-08, ..., 8.5682e-08, + 1.2107e-07, 1.2107e-07], + [ 0.0000e+00, 0.0000e+00, 9.7230e-07, ..., 1.8124e-06, + 4.9360e-08, 2.3022e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 9.9652e-08, ..., 1.3318e-07, + -2.0955e-07, -1.3877e-07], + [ 0.0000e+00, 0.0000e+00, 9.0338e-08, ..., 1.1362e-07, + 5.1223e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 3.7532e-07, ..., 4.5542e-07, + 9.5926e-08, 9.9652e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0297, 0.0115, 0.0292, 0.0072, 0.0356, -0.0091, -0.0032, -0.0150, + -0.0265, 0.0103], device='cuda:0'), grad: tensor([-2.8573e-06, 4.9453e-07, 3.9786e-06, 1.2470e-06, -3.2075e-06, + -4.1313e-06, 1.8142e-06, -1.4622e-07, 1.3588e-06, 1.4305e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 220.76, cls_loss 0.0019 cls_loss_mapping 0.0029 cls_loss_causal 0.4986 re_mapping 0.0053 re_causal 0.0154 /// teacc 98.99 lr 0.00010000 +Epoch 234, weight, value: tensor([[ 0.0032, 0.0110, 0.0333, ..., 0.0702, -0.1787, -0.1500], + [ 0.0579, -0.0185, -0.0171, ..., -0.1873, 0.0496, -0.0980], + [-0.0115, -0.0048, 0.1788, ..., -0.2808, -0.0980, -0.1405], + ..., + [-0.0183, 0.0062, -0.1448, ..., -0.0130, 0.1279, 0.0675], + [-0.0077, -0.0202, -0.1120, ..., -0.2406, 0.0877, -0.1620], + [-0.0058, -0.0258, -0.0668, ..., 0.0367, -0.1043, 0.0124]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9744e-06, ..., -9.3132e-10, + 3.3528e-08, 5.3085e-08], + [ 0.0000e+00, 0.0000e+00, 5.3365e-07, ..., 1.5832e-07, + -6.7055e-08, 3.6415e-07], + [ 0.0000e+00, 0.0000e+00, -1.5807e-04, ..., 5.5879e-08, + 8.5682e-08, 1.7602e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 7.1432e-07, ..., 2.2072e-06, + -2.3004e-07, 3.2671e-06], + [ 0.0000e+00, 0.0000e+00, 6.6645e-06, ..., 3.4459e-08, + 6.5193e-08, 7.0781e-08], + [ 0.0000e+00, 0.0000e+00, 1.4994e-07, ..., 1.7183e-06, + 4.9360e-08, 4.9472e-06]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0297, 0.0111, 0.0286, 0.0066, 0.0359, -0.0088, -0.0032, -0.0136, + -0.0254, 0.0094], device='cuda:0'), grad: tensor([ 3.3155e-06, 9.7323e-07, -2.4986e-04, 2.2793e-04, -1.3627e-05, + 1.2338e-05, -5.4762e-06, 5.8040e-06, 1.0885e-05, 8.0466e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 221.17, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4761 re_mapping 0.0055 re_causal 0.0162 /// teacc 99.01 lr 0.00010000 +Epoch 235, weight, value: tensor([[ 0.0032, 0.0110, 0.0334, ..., 0.0704, -0.1793, -0.1500], + [ 0.0579, -0.0185, -0.0168, ..., -0.1877, 0.0498, -0.0980], + [-0.0116, -0.0048, 0.1790, ..., -0.2814, -0.0994, -0.1416], + ..., + [-0.0183, 0.0062, -0.1452, ..., -0.0127, 0.1284, 0.0679], + [-0.0077, -0.0202, -0.1125, ..., -0.2413, 0.0879, -0.1627], + [-0.0058, -0.0258, -0.0670, ..., 0.0368, -0.1046, 0.0123]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.3330e-08, ..., -6.0536e-08, + 3.7253e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.0489e-08, + 6.7987e-08, 3.8277e-07], + [ 0.0000e+00, 0.0000e+00, 3.5390e-08, ..., 6.5193e-09, + 8.8476e-08, 5.1223e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.0245e-08, + -6.2957e-06, -1.1839e-05], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 2.0489e-08, + -6.9849e-08, 3.0734e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 1.0431e-06, + 2.2091e-06, 5.8189e-06]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0297, 0.0118, 0.0279, 0.0063, 0.0359, -0.0088, -0.0033, -0.0133, + -0.0255, 0.0094], device='cuda:0'), grad: tensor([-1.4715e-07, 5.1409e-07, 4.2003e-07, 1.3821e-05, -1.8692e-06, + 6.1840e-07, 1.2759e-07, -2.3514e-05, -2.3469e-07, 1.0252e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 221.03, cls_loss 0.0016 cls_loss_mapping 0.0040 cls_loss_causal 0.4843 re_mapping 0.0055 re_causal 0.0157 /// teacc 98.92 lr 0.00010000 +Epoch 236, weight, value: tensor([[ 0.0032, 0.0110, 0.0337, ..., 0.0706, -0.1796, -0.1501], + [ 0.0579, -0.0185, -0.0168, ..., -0.1886, 0.0496, -0.0981], + [-0.0116, -0.0048, 0.1791, ..., -0.2823, -0.0997, -0.1419], + ..., + [-0.0183, 0.0062, -0.1454, ..., -0.0130, 0.1286, 0.0678], + [-0.0077, -0.0202, -0.1129, ..., -0.2417, 0.0879, -0.1629], + [-0.0058, -0.0258, -0.0674, ..., 0.0363, -0.1046, 0.0121]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9073e-06, ..., 4.2841e-08, + 7.9907e-07, 1.2349e-06], + [ 0.0000e+00, 0.0000e+00, 1.7975e-07, ..., 2.8871e-08, + 6.7413e-05, 1.0777e-04], + [ 0.0000e+00, 0.0000e+00, 2.3127e-05, ..., 8.3819e-09, + 3.1173e-05, 4.9680e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 5.6811e-08, ..., 3.7253e-08, + 2.6488e-04, 4.2343e-04], + [ 0.0000e+00, 0.0000e+00, -3.5733e-05, ..., 1.2107e-08, + 8.2701e-06, 1.5043e-05], + [ 0.0000e+00, 0.0000e+00, 1.5385e-06, ..., 5.2154e-08, + -3.8314e-04, -6.1321e-04]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0297, 0.0117, 0.0276, 0.0063, 0.0362, -0.0088, -0.0026, -0.0132, + -0.0256, 0.0091], device='cuda:0'), grad: tensor([ 9.9912e-06, 3.9268e-04, 2.3580e-04, 2.9385e-05, 1.5251e-05, + 3.3051e-05, 4.4331e-06, 1.5411e-03, -3.3170e-05, -2.2278e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 220.51, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.4939 re_mapping 0.0051 re_causal 0.0152 /// teacc 98.98 lr 0.00010000 +Epoch 237, weight, value: tensor([[ 0.0032, 0.0110, 0.0336, ..., 0.0705, -0.1800, -0.1507], + [ 0.0579, -0.0185, -0.0168, ..., -0.1892, 0.0493, -0.0987], + [-0.0116, -0.0048, 0.1791, ..., -0.2842, -0.1002, -0.1431], + ..., + [-0.0183, 0.0062, -0.1456, ..., -0.0132, 0.1287, 0.0674], + [-0.0077, -0.0202, -0.1126, ..., -0.2422, 0.0883, -0.1629], + [-0.0058, -0.0258, -0.0670, ..., 0.0364, -0.1033, 0.0124]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.0734e-08, ..., -1.9278e-07, + 1.4901e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + -6.4000e-06, -3.4198e-06], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.7940e-09, + 6.4448e-06, 1.0245e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.7253e-09, + 7.3090e-06, 3.2578e-06], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + -8.2403e-06, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 2.5146e-08, + 2.3842e-07, 9.8720e-08]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0301, 0.0114, 0.0273, 0.0063, 0.0362, -0.0090, -0.0024, -0.0138, + -0.0253, 0.0100], device='cuda:0'), grad: tensor([-7.5996e-07, -1.6570e-05, 1.6108e-05, 7.6462e-07, 2.1793e-07, + -1.6019e-07, 1.3812e-06, 1.8835e-05, -2.0504e-05, 6.7987e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 221.04, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4829 re_mapping 0.0050 re_causal 0.0157 /// teacc 98.97 lr 0.00010000 +Epoch 238, weight, value: tensor([[ 0.0032, 0.0110, 0.0338, ..., 0.0708, -0.1801, -0.1509], + [ 0.0579, -0.0185, -0.0168, ..., -0.1895, 0.0494, -0.0988], + [-0.0116, -0.0048, 0.1792, ..., -0.2845, -0.1006, -0.1431], + ..., + [-0.0183, 0.0062, -0.1458, ..., -0.0135, 0.1287, 0.0673], + [-0.0077, -0.0202, -0.1127, ..., -0.2426, 0.0888, -0.1629], + [-0.0058, -0.0258, -0.0671, ..., 0.0364, -0.1033, 0.0124]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -2.8219e-07, + 8.3819e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 7.4506e-09, + -1.0226e-06, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 3.7253e-09, + 9.3877e-07, 1.4072e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 1.1735e-07, + -8.1584e-07, -2.9579e-06], + [ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., 3.6322e-08, + -2.3376e-07, -4.7497e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -1.7136e-07, + 7.8138e-07, 1.1614e-06]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0299, 0.0114, 0.0271, 0.0062, 0.0362, -0.0090, -0.0027, -0.0138, + -0.0250, 0.0100], device='cuda:0'), grad: tensor([-1.4948e-06, -1.7378e-06, 3.5949e-06, 3.6880e-07, 6.5565e-07, + 4.9453e-07, 7.2829e-07, -5.1856e-06, -4.6194e-07, 3.0398e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 220.35, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.4914 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.00 lr 0.00010000 +Epoch 239, weight, value: tensor([[ 0.0032, 0.0110, 0.0341, ..., 0.0711, -0.1803, -0.1508], + [ 0.0579, -0.0185, -0.0170, ..., -0.1900, 0.0502, -0.0978], + [-0.0116, -0.0048, 0.1807, ..., -0.2850, -0.1010, -0.1429], + ..., + [-0.0183, 0.0062, -0.1478, ..., -0.0137, 0.1280, 0.0666], + [-0.0077, -0.0202, -0.1129, ..., -0.2428, 0.0892, -0.1630], + [-0.0058, -0.0258, -0.0675, ..., 0.0362, -0.1036, 0.0124]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.5926e-07, ..., -1.3039e-08, + 2.0117e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-08, ..., 3.8184e-08, + 9.3132e-09, 6.4261e-08], + [ 0.0000e+00, 0.0000e+00, 3.7905e-07, ..., 4.6566e-09, + 8.8476e-08, 1.5832e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 4.2841e-08, + 2.8871e-08, 8.1956e-08], + [ 0.0000e+00, 0.0000e+00, 1.0813e-06, ..., 1.7509e-07, + 5.5134e-07, 1.1176e-07], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 3.3528e-08, + 2.6077e-08, -8.9407e-08]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0297, 0.0121, 0.0282, 0.0056, 0.0363, -0.0088, -0.0026, -0.0149, + -0.0248, 0.0099], device='cuda:0'), grad: tensor([ 9.9186e-07, 2.3004e-07, 1.6745e-06, -1.3271e-06, 1.3225e-07, + 1.1194e-06, -1.0908e-05, 4.5169e-07, 5.3160e-06, 2.2799e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 221.02, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.4921 re_mapping 0.0050 re_causal 0.0151 /// teacc 98.87 lr 0.00010000 +Epoch 240, weight, value: tensor([[ 0.0032, 0.0110, 0.0344, ..., 0.0707, -0.1805, -0.1513], + [ 0.0579, -0.0185, -0.0170, ..., -0.1910, 0.0498, -0.0983], + [-0.0116, -0.0048, 0.1811, ..., -0.2854, -0.1011, -0.1432], + ..., + [-0.0183, 0.0062, -0.1488, ..., -0.0135, 0.1285, 0.0670], + [-0.0077, -0.0202, -0.1133, ..., -0.2436, 0.0892, -0.1641], + [-0.0058, -0.0258, -0.0676, ..., 0.0363, -0.1038, 0.0124]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3283e-08, ..., -3.1665e-08, + 9.4064e-08, 4.9360e-08], + [ 0.0000e+00, 0.0000e+00, -1.9550e-05, ..., 3.0175e-07, + -2.0042e-05, 1.3569e-06], + [ 0.0000e+00, 0.0000e+00, 1.6078e-05, ..., 3.6322e-08, + 1.7762e-05, 1.5181e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 1.4380e-06, ..., 6.7987e-08, + -4.7404e-07, -4.2319e-06], + [ 0.0000e+00, 0.0000e+00, 1.1520e-06, ..., 1.7323e-07, + 1.7416e-07, 6.7707e-07], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., -1.4724e-06, + 6.0629e-07, -4.4852e-06]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0297, 0.0116, 0.0282, 0.0055, 0.0363, -0.0084, -0.0029, -0.0144, + -0.0253, 0.0099], device='cuda:0'), grad: tensor([ 2.2165e-07, -6.6459e-05, 5.8204e-05, 3.3788e-06, 9.1642e-06, + 2.4121e-06, -1.6037e-06, -3.6135e-06, 3.8780e-06, -5.5395e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 220.76, cls_loss 0.0011 cls_loss_mapping 0.0027 cls_loss_causal 0.5306 re_mapping 0.0047 re_causal 0.0154 /// teacc 99.08 lr 0.00010000 +Epoch 241, weight, value: tensor([[ 0.0032, 0.0110, 0.0353, ..., 0.0707, -0.1810, -0.1518], + [ 0.0579, -0.0185, -0.0170, ..., -0.1916, 0.0499, -0.0985], + [-0.0116, -0.0048, 0.1811, ..., -0.2862, -0.1014, -0.1435], + ..., + [-0.0183, 0.0062, -0.1492, ..., -0.0138, 0.1287, 0.0671], + [-0.0077, -0.0202, -0.1139, ..., -0.2471, 0.0890, -0.1679], + [-0.0058, -0.0258, -0.0678, ..., 0.0369, -0.1037, 0.0126]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.0366e-06, ..., -5.7444e-06, + 5.4017e-08, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 5.1223e-08, + 3.8184e-08, 1.4156e-07], + [ 0.0000e+00, 0.0000e+00, 5.4296e-07, ..., 6.1654e-07, + 2.7940e-08, 5.5879e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 6.3330e-08, + -3.0734e-08, 2.0396e-07], + [ 0.0000e+00, 0.0000e+00, 3.2596e-07, ..., 4.0885e-07, + -4.4890e-07, -2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 3.8892e-06, ..., -1.8910e-05, + 1.9744e-07, -1.0461e-04]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0296, 0.0117, 0.0281, 0.0052, 0.0362, -0.0081, -0.0031, -0.0144, + -0.0268, 0.0104], device='cuda:0'), grad: tensor([-1.1154e-05, 4.7032e-07, 1.4855e-06, 5.1130e-07, 1.0955e-04, + -2.2352e-06, 6.1560e-07, 3.3807e-07, -2.2743e-06, -9.7394e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 220.84, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5046 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.02 lr 0.00010000 +Epoch 242, weight, value: tensor([[ 0.0032, 0.0110, 0.0360, ..., 0.0713, -0.1815, -0.1519], + [ 0.0579, -0.0185, -0.0170, ..., -0.1924, 0.0502, -0.0983], + [-0.0115, -0.0048, 0.1813, ..., -0.2879, -0.1018, -0.1439], + ..., + [-0.0183, 0.0062, -0.1497, ..., -0.0150, 0.1287, 0.0670], + [-0.0077, -0.0202, -0.1142, ..., -0.2467, 0.0896, -0.1686], + [-0.0058, -0.0258, -0.0681, ..., 0.0377, -0.1039, 0.0130]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.4983e-07, ..., -4.3027e-07, + 2.1420e-08, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 4.0047e-08, + 8.1956e-08, 1.0058e-07], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 3.1665e-08, + 2.0117e-07, 2.0489e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., -5.5879e-09, + -1.1176e-07, -1.3504e-07], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 5.0291e-08, + 8.9593e-07, 6.7055e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-08, ..., -2.9523e-07, + 6.3330e-08, -5.5414e-07]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0293, 0.0120, 0.0279, 0.0056, 0.0358, -0.0090, -0.0027, -0.0146, + -0.0262, 0.0108], device='cuda:0'), grad: tensor([-7.5996e-07, 3.0268e-07, 5.3458e-07, -3.1084e-05, 6.4541e-07, + 3.1829e-05, -2.6077e-06, -2.7101e-07, 1.8841e-06, -5.1688e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 220.43, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4973 re_mapping 0.0051 re_causal 0.0153 /// teacc 98.98 lr 0.00010000 +Epoch 243, weight, value: tensor([[ 0.0032, 0.0110, 0.0363, ..., 0.0714, -0.1818, -0.1521], + [ 0.0581, -0.0185, -0.0157, ..., -0.1929, 0.0509, -0.0990], + [-0.0117, -0.0048, 0.1803, ..., -0.2887, -0.1043, -0.1454], + ..., + [-0.0183, 0.0062, -0.1492, ..., -0.0147, 0.1302, 0.0699], + [-0.0078, -0.0202, -0.1165, ..., -0.2468, 0.0888, -0.1690], + [-0.0058, -0.0258, -0.0684, ..., 0.0378, -0.1042, 0.0131]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.5446e-06, ..., -3.2783e-06, + 4.8708e-07, 4.7870e-07], + [-4.6566e-09, 0.0000e+00, 2.9914e-06, ..., 5.6718e-07, + 1.1045e-06, 8.3912e-07], + [ 1.8626e-09, 0.0000e+00, -9.8646e-06, ..., 4.8801e-07, + 1.5218e-06, 4.9174e-07], + ..., + [ 9.3132e-10, 0.0000e+00, 1.7881e-06, ..., 4.1947e-06, + -5.5879e-08, 3.1982e-06], + [ 9.3132e-10, 0.0000e+00, 8.8941e-07, ..., 4.5262e-07, + 1.8356e-06, 6.3051e-07], + [ 0.0000e+00, 0.0000e+00, 8.6240e-07, ..., 1.4296e-06, + 3.0734e-08, 1.0831e-06]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0293, 0.0132, 0.0261, 0.0031, 0.0356, -0.0092, -0.0023, -0.0115, + -0.0274, 0.0109], device='cuda:0'), grad: tensor([-6.5342e-06, 1.0699e-05, -2.1398e-05, -4.3847e-06, -5.7407e-06, + 1.4761e-06, 3.0547e-06, 1.2435e-05, 5.9977e-06, 4.3735e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 220.98, cls_loss 0.0016 cls_loss_mapping 0.0020 cls_loss_causal 0.5068 re_mapping 0.0052 re_causal 0.0144 /// teacc 99.00 lr 0.00010000 +Epoch 244, weight, value: tensor([[ 0.0031, 0.0110, 0.0374, ..., 0.0724, -0.1824, -0.1531], + [ 0.0581, -0.0185, -0.0156, ..., -0.1938, 0.0512, -0.0994], + [-0.0118, -0.0048, 0.1817, ..., -0.2870, -0.1048, -0.1439], + ..., + [-0.0184, 0.0062, -0.1529, ..., -0.0154, 0.1302, 0.0700], + [-0.0079, -0.0202, -0.1177, ..., -0.2472, 0.0893, -0.1694], + [-0.0058, -0.0258, -0.0693, ..., 0.0375, -0.1044, 0.0129]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9185e-07, ..., -1.2945e-07, + 7.4506e-09, 6.6124e-08], + [ 0.0000e+00, 0.0000e+00, -4.4852e-06, ..., 2.2445e-07, + -9.3505e-06, 2.3190e-07], + [ 0.0000e+00, 0.0000e+00, 3.2708e-06, ..., 8.2888e-08, + 6.9924e-06, 8.1956e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0990e-06, ..., 2.6822e-07, + 2.0526e-06, 1.0617e-07], + [ 0.0000e+00, 0.0000e+00, 6.6124e-08, ..., 1.3784e-07, + 4.1910e-08, 1.3504e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-08, ..., 3.9116e-06, + 6.3330e-08, 2.5257e-06]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0285, 0.0134, 0.0267, 0.0028, 0.0358, -0.0090, -0.0027, -0.0117, + -0.0279, 0.0107], device='cuda:0'), grad: tensor([-2.3469e-07, -2.0370e-05, 1.5691e-05, 2.5015e-06, -6.6943e-06, + -9.4101e-06, 3.1367e-06, 4.9546e-06, 4.7311e-06, 5.6475e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 220.71, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4890 re_mapping 0.0048 re_causal 0.0140 /// teacc 98.92 lr 0.00010000 +Epoch 245, weight, value: tensor([[ 0.0031, 0.0110, 0.0396, ..., 0.0751, -0.1831, -0.1527], + [ 0.0562, -0.0185, -0.0155, ..., -0.1949, 0.0508, -0.1006], + [-0.0119, -0.0048, 0.1824, ..., -0.2879, -0.1049, -0.1440], + ..., + [-0.0181, 0.0062, -0.1538, ..., -0.0157, 0.1308, 0.0705], + [-0.0079, -0.0202, -0.1200, ..., -0.2476, 0.0899, -0.1685], + [-0.0059, -0.0258, -0.0703, ..., 0.0374, -0.1049, 0.0129]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 1.5926e-07, ..., -4.5635e-08, + 1.1176e-08, 9.3132e-09], + [-7.0781e-08, 0.0000e+00, 7.0222e-07, ..., 3.7253e-08, + -1.6205e-07, 5.5879e-08], + [ 5.1223e-08, 0.0000e+00, -1.9725e-06, ..., 1.5739e-07, + 4.9081e-07, 5.0385e-07], + ..., + [-1.0245e-08, 0.0000e+00, 8.5123e-07, ..., -1.6484e-07, + -4.8708e-07, -8.1398e-07], + [ 1.2107e-08, 0.0000e+00, 5.4296e-07, ..., 9.3132e-09, + 8.9407e-08, 1.2945e-07], + [ 9.3132e-10, 0.0000e+00, 9.2201e-08, ..., 1.3597e-07, + 2.6077e-08, 2.9895e-07]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0257, 0.0134, 0.0269, 0.0032, 0.0358, -0.0094, -0.0044, -0.0116, + -0.0281, 0.0104], device='cuda:0'), grad: tensor([ 4.4983e-07, 1.1129e-06, -1.3337e-06, -3.8520e-06, 2.2911e-07, + 2.4494e-07, 8.3540e-07, 4.7497e-07, 1.2368e-06, 5.9791e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 220.76, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4707 re_mapping 0.0051 re_causal 0.0147 /// teacc 98.99 lr 0.00010000 +Epoch 246, weight, value: tensor([[ 0.0029, 0.0110, 0.0399, ..., 0.0757, -0.1842, -0.1529], + [ 0.0556, -0.0185, -0.0154, ..., -0.1958, 0.0509, -0.1011], + [-0.0122, -0.0048, 0.1825, ..., -0.2884, -0.1057, -0.1445], + ..., + [-0.0181, 0.0062, -0.1540, ..., -0.0159, 0.1311, 0.0706], + [-0.0092, -0.0202, -0.1210, ..., -0.2479, 0.0902, -0.1693], + [-0.0050, -0.0258, -0.0706, ..., 0.0374, -0.1050, 0.0130]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.5646e-07, ..., -1.8626e-09, + 3.0994e-06, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 3.5204e-07, ..., 1.5274e-07, + -8.4788e-06, 2.5425e-07], + [ 0.0000e+00, 0.0000e+00, -7.9628e-07, ..., 0.0000e+00, + 1.7732e-06, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.0163e-07, ..., -1.9092e-07, + 2.0489e-08, -3.3341e-07], + [ 0.0000e+00, 0.0000e+00, 2.5518e-06, ..., 6.5193e-09, + 3.3248e-07, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.2200e-07, ..., 2.1793e-07, + 1.0431e-07, 3.2224e-07]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0252, 0.0134, 0.0265, 0.0037, 0.0358, -0.0102, -0.0044, -0.0116, + -0.0282, 0.0105], device='cuda:0'), grad: tensor([ 1.3202e-05, -3.2932e-05, 1.0520e-05, -1.4460e-04, 1.1269e-07, + 1.1194e-04, 2.6241e-05, 1.7555e-06, 1.2800e-05, 1.2675e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 221.12, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.4955 re_mapping 0.0051 re_causal 0.0149 /// teacc 98.97 lr 0.00010000 +Epoch 247, weight, value: tensor([[ 0.0028, 0.0110, 0.0380, ..., 0.0733, -0.1853, -0.1560], + [ 0.0556, -0.0185, -0.0148, ..., -0.1981, 0.0509, -0.1014], + [-0.0107, -0.0048, 0.1828, ..., -0.2877, -0.1063, -0.1449], + ..., + [-0.0182, 0.0062, -0.1558, ..., -0.0163, 0.1315, 0.0710], + [-0.0092, -0.0202, -0.1217, ..., -0.2482, 0.0901, -0.1701], + [-0.0050, -0.0258, -0.0688, ..., 0.0397, -0.1054, 0.0137]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5842e-06, ..., -1.1204e-06, + 6.5193e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 6.2399e-08, ..., 1.1828e-07, + -1.2107e-08, 1.0338e-07], + [ 0.0000e+00, 0.0000e+00, -1.8906e-07, ..., 5.2154e-08, + 1.0151e-07, 1.3970e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.4715e-07, ..., 1.5087e-07, + 4.6566e-09, 1.0617e-07], + [ 0.0000e+00, 0.0000e+00, 1.8813e-07, ..., 1.2759e-07, + -1.4901e-07, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 3.0082e-07, ..., 7.2643e-07, + 2.7940e-08, 5.8487e-07]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0277, 0.0139, 0.0262, 0.0038, 0.0361, -0.0104, -0.0046, -0.0117, + -0.0285, 0.0124], device='cuda:0'), grad: tensor([-3.6806e-06, 2.9150e-07, 7.4506e-08, 2.4457e-06, -1.7378e-06, + -2.8126e-06, 2.8778e-06, 4.3120e-07, 2.2538e-07, 1.8533e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 220.89, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.4924 re_mapping 0.0051 re_causal 0.0144 /// teacc 99.01 lr 0.00010000 +Epoch 248, weight, value: tensor([[ 0.0028, 0.0110, 0.0380, ..., 0.0733, -0.1866, -0.1561], + [ 0.0556, -0.0185, -0.0149, ..., -0.1985, 0.0500, -0.1014], + [-0.0106, -0.0048, 0.1830, ..., -0.2883, -0.1067, -0.1451], + ..., + [-0.0182, 0.0062, -0.1560, ..., -0.0164, 0.1315, 0.0709], + [-0.0094, -0.0202, -0.1219, ..., -0.2484, 0.0932, -0.1703], + [-0.0050, -0.0258, -0.0689, ..., 0.0396, -0.1055, 0.0136]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -5.5879e-09, + 1.5832e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 9.3132e-10, + -2.8498e-07, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 8.8476e-08, ..., 0.0000e+00, + 2.6170e-07, 1.3318e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., 2.7940e-09, + 7.9162e-08, -3.8184e-08], + [ 0.0000e+00, 0.0000e+00, -2.6543e-07, ..., 0.0000e+00, + -1.4221e-06, -8.9221e-07], + [ 0.0000e+00, 0.0000e+00, 2.7008e-07, ..., 0.0000e+00, + 1.2787e-06, 7.5251e-07]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0277, 0.0130, 0.0260, 0.0037, 0.0364, -0.0107, -0.0045, -0.0117, + -0.0259, 0.0124], device='cuda:0'), grad: tensor([ 1.0896e-07, -7.9721e-07, 8.6054e-07, -4.6194e-07, 2.9430e-07, + -3.9674e-07, -1.7416e-07, 3.4366e-07, -3.5856e-06, 3.8091e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 221.03, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4677 re_mapping 0.0050 re_causal 0.0149 /// teacc 99.06 lr 0.00010000 +Epoch 249, weight, value: tensor([[ 0.0028, 0.0110, 0.0371, ..., 0.0733, -0.1891, -0.1561], + [ 0.0556, -0.0185, -0.0148, ..., -0.1987, 0.0501, -0.1015], + [-0.0106, -0.0048, 0.1833, ..., -0.2885, -0.1070, -0.1450], + ..., + [-0.0182, 0.0062, -0.1569, ..., -0.0167, 0.1314, 0.0707], + [-0.0094, -0.0202, -0.1226, ..., -0.2484, 0.0935, -0.1701], + [-0.0050, -0.0258, -0.0690, ..., 0.0395, -0.1056, 0.0125]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0245e-08, ..., -8.3819e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -1.2107e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 0.0000e+00, + 1.6764e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 8.3819e-09, + -6.7055e-08, -1.2666e-07], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -3.1665e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -5.5879e-09, + 5.6811e-08, 8.7544e-08]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0282, 0.0131, 0.0259, 0.0036, 0.0375, -0.0098, -0.0032, -0.0119, + -0.0259, 0.0118], device='cuda:0'), grad: tensor([-1.5832e-08, -2.7008e-08, 9.2201e-08, -2.2911e-07, 2.8871e-08, + 1.6578e-07, -2.6077e-08, -1.6857e-07, -3.2596e-08, 2.1700e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 220.91, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.5523 re_mapping 0.0049 re_causal 0.0154 /// teacc 99.03 lr 0.00010000 +Epoch 250, weight, value: tensor([[ 0.0028, 0.0110, 0.0372, ..., 0.0733, -0.1900, -0.1562], + [ 0.0556, -0.0185, -0.0147, ..., -0.1990, 0.0501, -0.1017], + [-0.0106, -0.0048, 0.1832, ..., -0.2889, -0.1074, -0.1452], + ..., + [-0.0182, 0.0062, -0.1570, ..., -0.0168, 0.1317, 0.0708], + [-0.0095, -0.0202, -0.1228, ..., -0.2485, 0.0936, -0.1703], + [-0.0050, -0.0258, -0.0690, ..., 0.0395, -0.1057, 0.0127]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., -1.5832e-08, + 1.1176e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 5.5879e-09, + 5.4017e-08, 1.4994e-07], + [-9.3132e-10, 0.0000e+00, -9.8720e-08, ..., 1.8626e-09, + 2.6077e-08, 5.1223e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 3.6322e-08, ..., 6.4261e-08, + -1.4100e-06, -3.3267e-06], + [ 9.3132e-10, 0.0000e+00, 8.0094e-08, ..., 1.0245e-08, + -4.6566e-09, 5.5879e-08], + [-1.3970e-08, 0.0000e+00, 1.6764e-08, ..., -6.1560e-07, + 4.8522e-07, -1.7202e-06]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0282, 0.0131, 0.0255, 0.0037, 0.0373, -0.0099, -0.0029, -0.0116, + -0.0259, 0.0118], device='cuda:0'), grad: tensor([-1.2107e-08, 4.2375e-07, 6.2399e-08, 2.4494e-07, 8.0913e-06, + 1.4016e-06, 2.1420e-08, -1.0200e-05, -2.1420e-08, -1.3039e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 220.93, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4873 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.06 lr 0.00010000 +Epoch 251, weight, value: tensor([[ 0.0028, 0.0110, 0.0373, ..., 0.0733, -0.1910, -0.1562], + [ 0.0556, -0.0185, -0.0146, ..., -0.1993, 0.0501, -0.1019], + [-0.0106, -0.0048, 0.1833, ..., -0.2891, -0.1076, -0.1453], + ..., + [-0.0182, 0.0062, -0.1573, ..., -0.0169, 0.1317, 0.0709], + [-0.0095, -0.0202, -0.1237, ..., -0.2488, 0.0938, -0.1707], + [-0.0050, -0.0258, -0.0691, ..., 0.0395, -0.1059, 0.0127]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.1654e-07, ..., -2.9150e-07, + 1.6764e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.7008e-07, ..., 1.4994e-07, + 6.7987e-08, 2.8126e-07], + [ 0.0000e+00, 0.0000e+00, 1.9558e-07, ..., 7.7300e-08, + 2.6263e-07, 2.0675e-07], + ..., + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 3.1665e-08, + -5.2713e-07, -3.4831e-07], + [ 0.0000e+00, 0.0000e+00, 1.4529e-07, ..., 6.5193e-08, + -4.1258e-07, -1.3411e-07], + [ 0.0000e+00, 0.0000e+00, 6.2399e-08, ..., 4.6566e-08, + 4.2934e-07, 8.1025e-08]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0283, 0.0131, 0.0255, 0.0036, 0.0372, -0.0099, -0.0018, -0.0116, + -0.0260, 0.0118], device='cuda:0'), grad: tensor([-1.2703e-06, 6.5751e-07, 9.9279e-07, 1.9744e-07, 1.5832e-08, + 3.7253e-08, 4.9360e-08, -9.4343e-07, -5.9046e-07, 8.6706e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 220.75, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5105 re_mapping 0.0046 re_causal 0.0140 /// teacc 98.92 lr 0.00010000 +Epoch 252, weight, value: tensor([[ 0.0028, 0.0110, 0.0375, ..., 0.0733, -0.1915, -0.1562], + [ 0.0556, -0.0185, -0.0143, ..., -0.2006, 0.0502, -0.1022], + [-0.0106, -0.0048, 0.1832, ..., -0.2895, -0.1080, -0.1456], + ..., + [-0.0182, 0.0062, -0.1575, ..., -0.0177, 0.1318, 0.0706], + [-0.0095, -0.0202, -0.1250, ..., -0.2493, 0.0935, -0.1716], + [-0.0050, -0.0258, -0.0692, ..., 0.0394, -0.1061, 0.0126]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.1353e-05, ..., -4.5747e-06, + 7.4506e-09, -1.2862e-06], + [ 0.0000e+00, 0.0000e+00, 1.1824e-05, ..., 1.8068e-06, + -6.5193e-09, 4.4703e-08], + [ 0.0000e+00, 0.0000e+00, 2.0955e-06, ..., 3.5670e-07, + 1.3039e-08, 5.2154e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 8.3819e-08, ..., 1.9558e-08, + -3.2969e-07, -3.5483e-07], + [ 0.0000e+00, 0.0000e+00, 4.4517e-07, ..., 1.6391e-07, + 4.8429e-08, 1.6391e-07], + [ 0.0000e+00, 0.0000e+00, 5.4613e-06, ..., 1.8580e-06, + 2.0582e-07, 1.2424e-06]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0282, 0.0133, 0.0252, 0.0035, 0.0376, -0.0099, -0.0015, -0.0118, + -0.0267, 0.0118], device='cuda:0'), grad: tensor([-4.4256e-05, 2.5719e-05, 4.5076e-06, 7.5623e-07, 2.9895e-07, + 1.2396e-06, 1.4780e-06, -5.8580e-07, 1.1493e-06, 9.7379e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 220.50, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5081 re_mapping 0.0049 re_causal 0.0151 /// teacc 98.98 lr 0.00010000 +Epoch 253, weight, value: tensor([[ 0.0028, 0.0110, 0.0375, ..., 0.0733, -0.1926, -0.1563], + [ 0.0556, -0.0185, -0.0146, ..., -0.2014, 0.0501, -0.1025], + [-0.0106, -0.0048, 0.1839, ..., -0.2897, -0.1078, -0.1459], + ..., + [-0.0182, 0.0062, -0.1577, ..., -0.0178, 0.1319, 0.0707], + [-0.0095, -0.0202, -0.1269, ..., -0.2498, 0.0934, -0.1731], + [-0.0050, -0.0258, -0.0692, ..., 0.0395, -0.1062, 0.0127]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., 4.6566e-09, + 1.8626e-09, 5.9139e-08], + [ 0.0000e+00, 0.0000e+00, -1.3504e-08, ..., 5.1223e-09, + -1.2713e-07, 3.2596e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 6.9849e-09, + 3.1199e-08, 1.8161e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 3.0734e-08, + 5.1223e-08, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 1.7229e-08, + -1.0012e-07, 9.4529e-08], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., -6.0769e-07, + 2.4214e-08, -1.4883e-06]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0282, 0.0130, 0.0254, 0.0036, 0.0375, -0.0102, -0.0006, -0.0117, + -0.0273, 0.0118], device='cuda:0'), grad: tensor([ 3.8650e-08, -3.1246e-07, 1.1269e-07, 2.6962e-07, 1.7099e-06, + -3.5251e-07, 3.0920e-07, 2.0396e-07, 5.1688e-08, -2.0172e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 220.85, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.5097 re_mapping 0.0047 re_causal 0.0146 /// teacc 98.98 lr 0.00010000 +Epoch 254, weight, value: tensor([[ 0.0028, 0.0110, 0.0376, ..., 0.0734, -0.1928, -0.1563], + [ 0.0556, -0.0185, -0.0145, ..., -0.2017, 0.0503, -0.1025], + [-0.0106, -0.0048, 0.1841, ..., -0.2901, -0.1083, -0.1458], + ..., + [-0.0182, 0.0062, -0.1581, ..., -0.0187, 0.1318, 0.0703], + [-0.0095, -0.0202, -0.1274, ..., -0.2505, 0.0935, -0.1752], + [-0.0049, -0.0258, -0.0693, ..., 0.0396, -0.1062, 0.0131]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0955e-08, ..., -5.5414e-08, + 6.5193e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 2.4214e-08, + -2.7474e-08, 4.0047e-08], + [ 0.0000e+00, 0.0000e+00, 1.6065e-07, ..., 5.5879e-09, + 7.0315e-08, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6298e-08, ..., 5.4017e-08, + 2.2352e-08, 1.1222e-07], + [ 0.0000e+00, 0.0000e+00, -3.1153e-07, ..., 1.3039e-08, + -1.8533e-07, 3.5856e-08], + [ 0.0000e+00, 0.0000e+00, 1.3458e-07, ..., -1.4110e-07, + 5.1223e-08, -6.1560e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0281, 0.0133, 0.0253, 0.0035, 0.0371, -0.0097, -0.0011, -0.0120, + -0.0275, 0.0121], device='cuda:0'), grad: tensor([ 9.7789e-09, 1.7695e-08, 6.6310e-07, -8.1956e-08, 4.8336e-07, + 1.9651e-07, 3.1944e-07, 2.1560e-07, -1.6168e-06, -1.9465e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 220.67, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.4781 re_mapping 0.0048 re_causal 0.0137 /// teacc 98.95 lr 0.00010000 +Epoch 255, weight, value: tensor([[ 0.0028, 0.0110, 0.0346, ..., 0.0734, -0.1937, -0.1563], + [ 0.0556, -0.0185, -0.0145, ..., -0.2024, 0.0504, -0.1025], + [-0.0106, -0.0048, 0.1841, ..., -0.2907, -0.1090, -0.1473], + ..., + [-0.0182, 0.0062, -0.1583, ..., -0.0191, 0.1322, 0.0704], + [-0.0095, -0.0202, -0.1287, ..., -0.2511, 0.0938, -0.1761], + [-0.0049, -0.0258, -0.0693, ..., 0.0401, -0.1065, 0.0144]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.5832e-07, ..., -1.3970e-09, + 7.9162e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 4.1910e-09, + 1.9278e-06, 1.6233e-06], + [ 0.0000e+00, 0.0000e+00, -7.6462e-07, ..., 9.3132e-10, + -1.6438e-07, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 7.5996e-07, ..., 1.3039e-08, + -1.9334e-06, -1.7192e-06], + [ 0.0000e+00, 0.0000e+00, 1.4575e-07, ..., 2.1420e-08, + -2.7474e-08, 5.6811e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.0710e-08, + 1.0524e-07, 1.9558e-08]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0307, 0.0135, 0.0245, 0.0060, 0.0359, -0.0084, -0.0020, -0.0118, + -0.0274, 0.0128], device='cuda:0'), grad: tensor([ 2.3022e-06, 4.2096e-06, -9.6858e-07, 7.6741e-07, 5.8208e-08, + 4.0948e-05, -4.5896e-05, -3.4384e-06, 1.4929e-06, 5.0850e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 220.95, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.5021 re_mapping 0.0049 re_causal 0.0147 /// teacc 98.95 lr 0.00010000 +Epoch 256, weight, value: tensor([[ 0.0028, 0.0110, 0.0345, ..., 0.0730, -0.1942, -0.1565], + [ 0.0556, -0.0185, -0.0144, ..., -0.2040, 0.0505, -0.1028], + [-0.0106, -0.0048, 0.1842, ..., -0.2913, -0.1094, -0.1474], + ..., + [-0.0182, 0.0062, -0.1585, ..., -0.0197, 0.1324, 0.0704], + [-0.0095, -0.0202, -0.1292, ..., -0.2514, 0.0942, -0.1759], + [-0.0049, -0.0258, -0.0691, ..., 0.0407, -0.1072, 0.0145]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.2200e-07, ..., 5.1223e-09, + 1.0710e-08, 1.2573e-08], + [ 0.0000e+00, 0.0000e+00, -2.3935e-07, ..., 2.7940e-09, + -7.0268e-07, 9.7789e-09], + [ 0.0000e+00, 0.0000e+00, -5.6764e-07, ..., -2.8871e-08, + 4.5914e-07, -7.3574e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3236e-07, ..., 1.5367e-08, + 1.0245e-08, 1.9558e-08], + [ 0.0000e+00, 0.0000e+00, 1.0105e-07, ..., 1.9092e-08, + 8.3819e-09, 3.6322e-08], + [ 0.0000e+00, 0.0000e+00, 4.2841e-08, ..., 6.5193e-09, + 1.8626e-08, 1.0245e-08]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0310, 0.0136, 0.0242, 0.0060, 0.0359, -0.0082, -0.0019, -0.0119, + -0.0271, 0.0130], device='cuda:0'), grad: tensor([ 2.6729e-07, -1.9874e-06, 7.9628e-08, 4.6520e-07, 6.8592e-07, + -4.5495e-07, -2.1467e-07, 4.5076e-07, 5.7044e-07, 1.4389e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 220.78, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4870 re_mapping 0.0049 re_causal 0.0146 /// teacc 98.91 lr 0.00010000 +Epoch 257, weight, value: tensor([[ 0.0028, 0.0110, 0.0345, ..., 0.0733, -0.1948, -0.1562], + [ 0.0556, -0.0185, -0.0145, ..., -0.2047, 0.0512, -0.1021], + [-0.0106, -0.0048, 0.1844, ..., -0.2919, -0.1098, -0.1479], + ..., + [-0.0182, 0.0062, -0.1587, ..., -0.0199, 0.1326, 0.0708], + [-0.0095, -0.0202, -0.1297, ..., -0.2520, 0.0934, -0.1779], + [-0.0049, -0.0258, -0.0692, ..., 0.0406, -0.1079, 0.0144]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5146e-08, ..., -5.0291e-08, + 1.3970e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.3504e-08, ..., 1.1176e-08, + 1.7136e-07, 1.6764e-07], + [ 0.0000e+00, 0.0000e+00, 5.1688e-08, ..., 1.3970e-09, + 2.9057e-07, 3.2596e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 2.1420e-08, + -2.6077e-07, -1.9511e-07], + [ 0.0000e+00, 0.0000e+00, -1.1176e-07, ..., 2.1374e-07, + -5.3737e-07, 1.0757e-06], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., -3.7486e-07, + -8.7544e-08, -1.5954e-06]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0308, 0.0146, 0.0240, 0.0060, 0.0361, -0.0083, -0.0025, -0.0117, + -0.0285, 0.0127], device='cuda:0'), grad: tensor([-1.0524e-07, 5.2107e-07, 8.5915e-07, 3.7299e-07, 7.7952e-07, + 3.3528e-07, 4.6333e-07, -5.6718e-07, 2.4866e-07, -2.9020e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 220.93, cls_loss 0.0014 cls_loss_mapping 0.0026 cls_loss_causal 0.5040 re_mapping 0.0045 re_causal 0.0138 /// teacc 98.91 lr 0.00010000 +Epoch 258, weight, value: tensor([[ 0.0028, 0.0110, 0.0346, ..., 0.0734, -0.1953, -0.1562], + [ 0.0556, -0.0185, -0.0146, ..., -0.2052, 0.0479, -0.1056], + [-0.0106, -0.0048, 0.1847, ..., -0.2922, -0.1104, -0.1477], + ..., + [-0.0182, 0.0062, -0.1593, ..., -0.0202, 0.1365, 0.0728], + [-0.0095, -0.0202, -0.1300, ..., -0.2525, 0.0928, -0.1796], + [-0.0049, -0.0258, -0.0692, ..., 0.0405, -0.1082, 0.0145]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6915e-07, ..., -1.9837e-07, + 1.4901e-08, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 2.0023e-08, + -4.8894e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 3.0268e-08, + 1.0012e-07, 6.9849e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.1467e-08, ..., -2.7427e-07, + -1.0291e-07, -1.3197e-06], + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 2.1420e-08, + 3.5763e-06, 1.9092e-08], + [ 0.0000e+00, 0.0000e+00, 7.0315e-08, ..., -5.0291e-08, + 3.3993e-08, -3.2363e-07]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0308, 0.0111, 0.0238, 0.0060, 0.0360, -0.0079, -0.0029, -0.0086, + -0.0292, 0.0128], device='cuda:0'), grad: tensor([-5.6531e-07, -3.1199e-07, 4.0000e-07, 1.7323e-07, 2.9169e-06, + -8.5309e-06, 2.3609e-07, -2.0787e-06, 7.9200e-06, -1.6065e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 220.62, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4941 re_mapping 0.0048 re_causal 0.0140 /// teacc 98.94 lr 0.00010000 +Epoch 259, weight, value: tensor([[ 0.0028, 0.0110, 0.0346, ..., 0.0734, -0.1963, -0.1562], + [ 0.0556, -0.0185, -0.0159, ..., -0.2057, 0.0482, -0.1058], + [-0.0106, -0.0048, 0.1863, ..., -0.2927, -0.1109, -0.1483], + ..., + [-0.0182, 0.0062, -0.1599, ..., -0.0227, 0.1364, 0.0720], + [-0.0096, -0.0202, -0.1303, ..., -0.2525, 0.0929, -0.1807], + [-0.0049, -0.0258, -0.0692, ..., 0.0404, -0.1089, 0.0144]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3935e-07, ..., -2.7614e-07, + 1.7695e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.0955e-08, ..., 7.4506e-09, + -1.7975e-07, -3.0268e-08], + [ 0.0000e+00, 0.0000e+00, -8.1118e-07, ..., 2.0955e-08, + 1.9465e-07, 8.1025e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.9430e-07, ..., 1.6298e-08, + 9.9186e-08, 3.3062e-08], + [ 0.0000e+00, 0.0000e+00, 2.7521e-07, ..., 3.7253e-09, + -6.0303e-07, -2.7101e-07], + [ 0.0000e+00, 0.0000e+00, 3.4971e-07, ..., 2.1840e-07, + 9.6392e-08, 6.5193e-08]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0308, 0.0110, 0.0245, 0.0060, 0.0368, -0.0079, -0.0031, -0.0090, + -0.0292, 0.0127], device='cuda:0'), grad: tensor([-4.1770e-07, -5.4436e-07, -1.0515e-06, 2.5239e-07, 1.3784e-07, + 4.7497e-07, 1.8626e-08, 8.5449e-07, -7.2457e-07, 9.9838e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 221.18, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.5057 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.03 lr 0.00010000 +Epoch 260, weight, value: tensor([[ 0.0028, 0.0110, 0.0346, ..., 0.0735, -0.1976, -0.1563], + [ 0.0556, -0.0185, -0.0160, ..., -0.2062, 0.0492, -0.1035], + [-0.0106, -0.0048, 0.1868, ..., -0.2930, -0.1109, -0.1483], + ..., + [-0.0182, 0.0062, -0.1604, ..., -0.0250, 0.1362, 0.0709], + [-0.0096, -0.0202, -0.1308, ..., -0.2528, 0.0940, -0.1798], + [-0.0049, -0.0258, -0.0693, ..., 0.0402, -0.1120, 0.0134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8161e-07, ..., -1.9511e-07, + 1.0710e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 3.4459e-08, ..., 7.9162e-09, + -6.1933e-08, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, -2.8545e-07, ..., 8.8476e-09, + 5.0757e-08, 9.7789e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4261e-08, ..., 1.7695e-08, + 2.3283e-08, 3.5856e-08], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 3.9581e-08, + -6.5425e-07, -5.4110e-07], + [ 0.0000e+00, 0.0000e+00, 2.2724e-07, ..., -1.7323e-07, + 4.0093e-07, -2.4680e-08]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0308, 0.0119, 0.0247, 0.0060, 0.0374, -0.0080, -0.0025, -0.0095, + -0.0287, 0.0116], device='cuda:0'), grad: tensor([-4.3446e-07, -1.1828e-07, -2.3004e-07, -3.4925e-08, 8.5542e-07, + 6.7893e-07, 2.1700e-07, 2.2398e-07, -4.1462e-06, 3.0063e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 220.71, cls_loss 0.0011 cls_loss_mapping 0.0024 cls_loss_causal 0.4864 re_mapping 0.0048 re_causal 0.0141 /// teacc 99.03 lr 0.00010000 +Epoch 261, weight, value: tensor([[ 0.0028, 0.0110, 0.0346, ..., 0.0735, -0.1973, -0.1563], + [ 0.0556, -0.0185, -0.0164, ..., -0.2072, 0.0493, -0.1031], + [-0.0103, -0.0048, 0.1878, ..., -0.2935, -0.1112, -0.1484], + ..., + [-0.0182, 0.0062, -0.1615, ..., -0.0251, 0.1362, 0.0707], + [-0.0096, -0.0202, -0.1312, ..., -0.2531, 0.0945, -0.1799], + [-0.0049, -0.0258, -0.0694, ..., 0.0403, -0.1121, 0.0135]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7008e-08, ..., 7.7300e-08, + 1.8626e-08, 3.1199e-08], + [ 9.3132e-10, 0.0000e+00, 2.5844e-07, ..., 8.5682e-08, + 5.2620e-08, 1.4994e-07], + [ 0.0000e+00, 0.0000e+00, -3.3900e-07, ..., 2.7008e-08, + 4.1910e-08, 2.7474e-08], + ..., + [ 1.8626e-09, 0.0000e+00, 6.5193e-08, ..., 1.0198e-06, + -2.2491e-07, -3.8324e-07], + [ 4.6566e-10, 0.0000e+00, 2.4214e-08, ..., 1.9558e-08, + -1.1176e-08, 3.4925e-08], + [ 2.7940e-08, 0.0000e+00, 1.1176e-08, ..., -1.7956e-06, + 1.8766e-07, 1.4491e-06]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0307, 0.0119, 0.0255, 0.0060, 0.0371, -0.0081, -0.0026, -0.0098, + -0.0284, 0.0117], device='cuda:0'), grad: tensor([ 2.0163e-07, 8.1444e-07, -3.2131e-07, 1.5367e-08, 2.4028e-07, + 3.1944e-07, -5.3830e-07, 3.3295e-07, 1.9325e-07, -1.2610e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 221.16, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.5012 re_mapping 0.0046 re_causal 0.0142 /// teacc 98.99 lr 0.00010000 +Epoch 262, weight, value: tensor([[ 0.0027, 0.0110, 0.0347, ..., 0.0735, -0.1980, -0.1564], + [ 0.0556, -0.0185, -0.0168, ..., -0.2080, 0.0498, -0.1003], + [-0.0103, -0.0048, 0.1883, ..., -0.2940, -0.1122, -0.1502], + ..., + [-0.0182, 0.0062, -0.1622, ..., -0.0253, 0.1360, 0.0683], + [-0.0096, -0.0202, -0.1318, ..., -0.2534, 0.0944, -0.1804], + [-0.0060, -0.0258, -0.0694, ..., 0.0403, -0.1123, 0.0134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -3.3528e-08, + 4.6566e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 1.4435e-08, ..., 8.3819e-09, + -7.4506e-09, 2.9337e-08], + [ 0.0000e+00, 0.0000e+00, -3.9069e-07, ..., 1.8626e-09, + 6.9849e-09, 4.1910e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.1199e-08, ..., 1.3504e-08, + -3.3528e-08, 1.1642e-08], + [ 0.0000e+00, 0.0000e+00, 1.3923e-07, ..., 7.9162e-09, + 4.6566e-10, 3.0268e-08], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., -1.3039e-08, + 2.5146e-08, -2.1933e-07]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0307, 0.0144, 0.0242, 0.0062, 0.0373, -0.0082, -0.0027, -0.0118, + -0.0287, 0.0115], device='cuda:0'), grad: tensor([-4.7963e-08, 1.4435e-08, -5.4250e-07, 2.0629e-07, 2.6170e-07, + 4.6566e-08, 9.2201e-08, -3.8184e-08, 2.8545e-07, -2.7707e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 221.09, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.5146 re_mapping 0.0046 re_causal 0.0142 /// teacc 98.93 lr 0.00010000 +Epoch 263, weight, value: tensor([[ 0.0027, 0.0110, 0.0347, ..., 0.0736, -0.1981, -0.1564], + [ 0.0556, -0.0185, -0.0170, ..., -0.2087, 0.0498, -0.0995], + [-0.0103, -0.0048, 0.1888, ..., -0.2944, -0.1124, -0.1504], + ..., + [-0.0185, 0.0062, -0.1627, ..., -0.0253, 0.1359, 0.0675], + [-0.0098, -0.0202, -0.1323, ..., -0.2538, 0.0953, -0.1790], + [-0.0060, -0.0258, -0.0695, ..., 0.0403, -0.1124, 0.0134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.4238e-07, ..., -4.7917e-07, + 2.9337e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.1688e-07, ..., 9.7323e-08, + -4.7963e-08, 8.6147e-08], + [ 0.0000e+00, 0.0000e+00, -5.1223e-09, ..., 3.0268e-08, + 3.4925e-08, 1.8161e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.7229e-08, ..., 1.3504e-08, + -7.4971e-08, -2.9942e-07], + [ 0.0000e+00, 0.0000e+00, 5.1223e-08, ..., 8.5682e-08, + 1.7229e-08, 3.1199e-08], + [ 0.0000e+00, 0.0000e+00, 9.1270e-08, ..., 8.8476e-08, + 5.0757e-08, 6.0536e-09]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0309, 0.0151, 0.0242, 0.0062, 0.0372, -0.0076, -0.0016, -0.0126, + -0.0283, 0.0113], device='cuda:0'), grad: tensor([-3.0845e-06, 3.5344e-07, 1.7835e-07, 2.2631e-07, 1.5367e-07, + 3.0361e-07, 8.7637e-07, -7.1665e-07, 7.0175e-07, 1.0096e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 220.68, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.5258 re_mapping 0.0047 re_causal 0.0143 /// teacc 98.99 lr 0.00010000 +Epoch 264, weight, value: tensor([[ 0.0027, 0.0110, 0.0347, ..., 0.0736, -0.2002, -0.1565], + [ 0.0556, -0.0185, -0.0165, ..., -0.2098, 0.0501, -0.0995], + [-0.0086, -0.0048, 0.1888, ..., -0.2954, -0.1139, -0.1504], + ..., + [-0.0187, 0.0062, -0.1638, ..., -0.0251, 0.1360, 0.0675], + [-0.0101, -0.0202, -0.1334, ..., -0.2553, 0.0952, -0.1805], + [-0.0060, -0.0258, -0.0697, ..., 0.0402, -0.1127, 0.0134]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 4.6566e-10, + 9.3132e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 2.8871e-08, ..., 2.3283e-09, + 4.0978e-08, 3.5856e-08], + [ 0.0000e+00, 0.0000e+00, -2.0536e-07, ..., 9.3132e-10, + 1.1735e-07, 3.1199e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3644e-07, ..., 4.1910e-09, + -9.3179e-07, -4.3167e-07], + [ 0.0000e+00, 0.0000e+00, -1.0291e-07, ..., 8.8476e-09, + -8.4750e-08, 3.0734e-08], + [ 0.0000e+00, 0.0000e+00, 2.9337e-08, ..., -3.1199e-08, + 4.8429e-08, -5.9139e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0309, 0.0152, 0.0235, 0.0064, 0.0370, -0.0077, -0.0007, -0.0125, + -0.0288, 0.0112], device='cuda:0'), grad: tensor([ 4.2841e-08, 2.0675e-07, 1.0291e-07, 2.5742e-06, 1.0198e-07, + -3.1712e-07, 3.1944e-07, -2.5649e-06, -7.0268e-07, 2.2305e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 220.99, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.5004 re_mapping 0.0048 re_causal 0.0139 /// teacc 99.05 lr 0.00010000 +Epoch 265, weight, value: tensor([[ 0.0024, 0.0110, 0.0348, ..., 0.0737, -0.2007, -0.1566], + [ 0.0555, -0.0185, -0.0171, ..., -0.2123, 0.0497, -0.0996], + [-0.0076, -0.0048, 0.1894, ..., -0.2982, -0.1142, -0.1518], + ..., + [-0.0191, 0.0062, -0.1646, ..., -0.0236, 0.1365, 0.0677], + [-0.0105, -0.0202, -0.1318, ..., -0.2559, 0.0955, -0.1813], + [-0.0039, -0.0258, -0.0698, ..., 0.0402, -0.1129, 0.0136]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., -7.0315e-08, + 9.7323e-08, 3.9581e-08], + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 1.0245e-07, + -9.7230e-07, 3.0315e-07], + [ 0.0000e+00, 0.0000e+00, 5.9512e-07, ..., 1.4435e-08, + 2.7334e-07, 4.0047e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 9.9186e-08, ..., 2.7176e-06, + 4.9826e-08, 8.5980e-06], + [ 0.0000e+00, 0.0000e+00, 2.2305e-07, ..., 4.5169e-08, + 4.8429e-08, 1.2387e-07], + [ 0.0000e+00, 0.0000e+00, 5.5879e-08, ..., -3.3900e-06, + 5.0291e-08, -1.0774e-05]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0308, 0.0149, 0.0233, 0.0064, 0.0363, -0.0080, -0.0003, -0.0122, + -0.0284, 0.0111], device='cuda:0'), grad: tensor([ 1.4473e-06, -3.4310e-06, 2.8461e-06, -1.0747e-06, 2.1607e-06, + 1.2584e-05, -1.3515e-05, 1.3538e-05, 1.3392e-06, -1.5929e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 220.83, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5090 re_mapping 0.0047 re_causal 0.0139 /// teacc 98.98 lr 0.00010000 +Epoch 266, weight, value: tensor([[ 0.0021, 0.0110, 0.0349, ..., 0.0739, -0.2015, -0.1566], + [ 0.0555, -0.0185, -0.0172, ..., -0.2133, 0.0497, -0.0996], + [-0.0054, -0.0048, 0.1903, ..., -0.2987, -0.1144, -0.1516], + ..., + [-0.0194, 0.0062, -0.1666, ..., -0.0241, 0.1367, 0.0676], + [-0.0111, -0.0202, -0.1317, ..., -0.2562, 0.0958, -0.1815], + [-0.0009, -0.0258, -0.0699, ..., 0.0402, -0.1131, 0.0139]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -8.3819e-09, + 1.5274e-07, 3.0734e-08], + [-1.8626e-09, 0.0000e+00, 2.8871e-08, ..., 3.7253e-09, + 3.0726e-05, 1.2387e-07], + [ 9.3132e-10, 0.0000e+00, 7.6462e-07, ..., 1.8626e-09, + 3.6396e-06, 1.6354e-06], + ..., + [ 4.6566e-10, 0.0000e+00, -3.6368e-07, ..., 9.7789e-09, + -3.4630e-05, -1.9372e-06], + [ 0.0000e+00, 0.0000e+00, 4.0978e-08, ..., 8.3819e-09, + 2.0955e-08, 1.6764e-08], + [-6.9849e-09, 0.0000e+00, 3.3062e-08, ..., -9.4995e-08, + 5.6345e-08, -6.7521e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0306, 0.0149, 0.0236, 0.0063, 0.0362, -0.0080, -0.0008, -0.0123, + -0.0282, 0.0112], device='cuda:0'), grad: tensor([ 9.6671e-07, 8.1837e-05, 1.4372e-05, -4.2794e-07, 4.7358e-07, + 4.1304e-07, -1.4324e-06, -9.6679e-05, 1.7602e-07, 1.8720e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 220.89, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4892 re_mapping 0.0050 re_causal 0.0144 /// teacc 99.07 lr 0.00010000 +Epoch 267, weight, value: tensor([[ 0.0019, 0.0110, 0.0350, ..., 0.0741, -0.2017, -0.1567], + [ 0.0554, -0.0185, -0.0173, ..., -0.2160, 0.0495, -0.0997], + [-0.0051, -0.0048, 0.1913, ..., -0.2996, -0.1147, -0.1516], + ..., + [-0.0224, 0.0062, -0.1679, ..., -0.0241, 0.1341, 0.0664], + [-0.0134, -0.0202, -0.1323, ..., -0.2577, 0.0959, -0.1832], + [-0.0013, -0.0258, -0.0701, ..., 0.0403, -0.1132, 0.0140]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.1665e-08, ..., -1.9744e-07, + 7.4506e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.7276e-07, ..., 1.8161e-08, + 1.4156e-07, 4.5169e-08], + [ 0.0000e+00, 0.0000e+00, -9.4296e-07, ..., 3.2596e-09, + 2.3749e-07, 2.3749e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4261e-07, ..., 1.3504e-08, + -7.8417e-07, -5.0617e-07], + [ 0.0000e+00, 0.0000e+00, 7.8231e-08, ..., 2.0489e-08, + -4.8848e-07, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.1623e-06, + 1.4342e-07, 2.8871e-08]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0305, 0.0149, 0.0240, 0.0090, 0.0360, -0.0076, -0.0011, -0.0149, + -0.0287, 0.0113], device='cuda:0'), grad: tensor([-6.5099e-07, 9.3272e-07, -6.3004e-07, 6.5602e-06, 2.5472e-07, + -1.9506e-05, 2.1420e-08, -2.0079e-06, -1.2694e-06, 1.6287e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 221.40, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.5473 re_mapping 0.0048 re_causal 0.0146 /// teacc 98.95 lr 0.00010000 +Epoch 268, weight, value: tensor([[ 0.0018, 0.0110, 0.0350, ..., 0.0741, -0.2025, -0.1567], + [ 0.0553, -0.0185, -0.0174, ..., -0.2183, 0.0493, -0.0997], + [-0.0051, -0.0048, 0.1915, ..., -0.3001, -0.1151, -0.1519], + ..., + [-0.0226, 0.0062, -0.1680, ..., -0.0238, 0.1342, 0.0665], + [-0.0138, -0.0202, -0.1327, ..., -0.2583, 0.0958, -0.1839], + [-0.0017, -0.0258, -0.0702, ..., 0.0400, -0.1141, 0.0136]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., -1.2107e-08, + 1.9511e-07, 6.0536e-08], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 3.4925e-08, + 2.0918e-06, 1.4435e-08], + [ 0.0000e+00, 0.0000e+00, -8.2422e-08, ..., -4.1910e-08, + 2.3842e-07, 6.0536e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 4.0978e-08, + 9.3132e-10, 7.3574e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 8.3819e-08, + 1.8626e-09, 1.5926e-07], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -3.8510e-07, + 1.9558e-08, -8.4192e-07]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0305, 0.0148, 0.0239, 0.0092, 0.0362, -0.0077, -0.0006, -0.0149, + -0.0289, 0.0109], device='cuda:0'), grad: tensor([ 1.3299e-06, 4.6715e-06, 5.3318e-07, 8.5775e-07, 9.9093e-07, + -4.3260e-07, -4.3251e-06, 3.1805e-07, -2.6412e-06, -1.2890e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 220.97, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4740 re_mapping 0.0048 re_causal 0.0140 /// teacc 98.88 lr 0.00010000 +Epoch 269, weight, value: tensor([[ 0.0017, 0.0110, 0.0350, ..., 0.0742, -0.2031, -0.1567], + [ 0.0553, -0.0185, -0.0174, ..., -0.2211, 0.0491, -0.0998], + [-0.0047, -0.0048, 0.1918, ..., -0.3003, -0.1153, -0.1522], + ..., + [-0.0226, 0.0062, -0.1686, ..., -0.0236, 0.1344, 0.0666], + [-0.0139, -0.0202, -0.1332, ..., -0.2595, 0.0957, -0.1862], + [-0.0016, -0.0258, -0.0702, ..., 0.0401, -0.1149, 0.0138]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., -6.9849e-09, + 1.7695e-08, 2.3283e-08], + [-1.0710e-08, 0.0000e+00, 5.1223e-09, ..., 6.0536e-09, + -2.3842e-07, 2.8778e-07], + [ 3.7253e-09, 0.0000e+00, -4.0606e-07, ..., 9.3132e-10, + 5.0338e-07, 3.2503e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3947e-07, ..., 1.0710e-08, + -1.3001e-06, -1.7285e-06], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 2.0489e-08, + 3.1665e-08, 1.4435e-07], + [-4.6566e-10, 0.0000e+00, 3.2596e-08, ..., -1.2945e-07, + 6.1840e-07, 3.3202e-07]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0304, 0.0147, 0.0239, 0.0092, 0.0362, -0.0077, -0.0009, -0.0148, + -0.0293, 0.0110], device='cuda:0'), grad: tensor([ 5.4482e-08, -2.8126e-07, 1.4137e-06, 2.9197e-07, 1.0580e-06, + -5.8534e-07, 5.4762e-07, -5.0254e-06, 6.8545e-07, 1.8459e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 220.82, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4671 re_mapping 0.0051 re_causal 0.0136 /// teacc 98.91 lr 0.00010000 +Epoch 270, weight, value: tensor([[ 0.0017, 0.0110, 0.0350, ..., 0.0741, -0.2048, -0.1569], + [ 0.0553, -0.0185, -0.0176, ..., -0.2226, 0.0480, -0.0998], + [-0.0054, -0.0048, 0.1923, ..., -0.3010, -0.1159, -0.1530], + ..., + [-0.0226, 0.0062, -0.1687, ..., -0.0234, 0.1349, 0.0667], + [-0.0142, -0.0202, -0.1342, ..., -0.2598, 0.0966, -0.1868], + [-0.0016, -0.0258, -0.0703, ..., 0.0402, -0.1151, 0.0139]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.7695e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + -2.2165e-07, 3.9116e-08], + [ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., 0.0000e+00, + 1.0710e-07, 1.0245e-08], + ..., + [-9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 3.7253e-09, + -7.7300e-08, -1.3039e-07], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 6.5193e-09, + 3.2596e-08, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.9116e-08, + 4.0047e-08, -2.7008e-08]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0305, 0.0144, 0.0239, 0.0093, 0.0363, -0.0104, 0.0008, -0.0146, + -0.0307, 0.0111], device='cuda:0'), grad: tensor([ 1.3597e-07, -6.6031e-07, 3.4273e-07, 3.8091e-07, 9.1270e-08, + 5.8860e-07, -8.6520e-07, -2.8219e-07, 2.3469e-07, 2.2352e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 221.49, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.5110 re_mapping 0.0052 re_causal 0.0147 /// teacc 98.82 lr 0.00010000 +Epoch 271, weight, value: tensor([[ 0.0016, 0.0110, 0.0354, ..., 0.0735, -0.2061, -0.1586], + [ 0.0552, -0.0185, -0.0179, ..., -0.2244, 0.0476, -0.0999], + [-0.0054, -0.0048, 0.1942, ..., -0.3032, -0.1166, -0.1527], + ..., + [-0.0226, 0.0062, -0.1706, ..., -0.0237, 0.1352, 0.0667], + [-0.0148, -0.0202, -0.1347, ..., -0.2607, 0.0970, -0.1873], + [-0.0008, -0.0258, -0.0705, ..., 0.0409, -0.1159, 0.0138]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.8580e-07, ..., -1.8626e-08, + 3.4645e-07, 1.2852e-07], + [ 0.0000e+00, 0.0000e+00, -3.2205e-06, ..., 9.3132e-10, + -1.1148e-06, -1.5814e-06], + [ 0.0000e+00, 0.0000e+00, 5.7835e-07, ..., 0.0000e+00, + 6.8825e-07, 9.9465e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3528e-07, ..., 9.3132e-10, + 3.9116e-07, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 1.5739e-07, ..., 5.5879e-09, + 6.3330e-08, 1.4435e-07], + [ 0.0000e+00, 0.0000e+00, 1.2759e-07, ..., -4.6566e-09, + 5.4017e-08, 2.6077e-08]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0304, 0.0142, 0.0239, 0.0092, 0.0368, -0.0108, 0.0009, -0.0146, + -0.0309, 0.0113], device='cuda:0'), grad: tensor([ 3.2298e-06, -1.0237e-05, 6.2138e-06, -3.2187e-06, 5.4613e-06, + 4.7311e-06, -9.8497e-06, 1.7304e-06, 1.1269e-06, 7.7486e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 221.18, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4702 re_mapping 0.0052 re_causal 0.0138 /// teacc 98.97 lr 0.00010000 +Epoch 272, weight, value: tensor([[ 0.0014, 0.0110, 0.0354, ..., 0.0735, -0.2087, -0.1591], + [ 0.0552, -0.0185, -0.0178, ..., -0.2253, 0.0476, -0.0999], + [-0.0054, -0.0048, 0.1946, ..., -0.3038, -0.1174, -0.1532], + ..., + [-0.0226, 0.0062, -0.1711, ..., -0.0239, 0.1354, 0.0668], + [-0.0152, -0.0202, -0.1353, ..., -0.2613, 0.1001, -0.1877], + [ 0.0020, -0.0258, -0.0704, ..., 0.0413, -0.1162, 0.0148]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.6811e-08, ..., -9.2201e-08, + 1.8626e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 5.7742e-08, + -3.4459e-08, 2.5053e-07], + [ 0.0000e+00, 0.0000e+00, -1.4389e-06, ..., 8.3819e-09, + -4.6566e-09, -5.2154e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3011e-06, ..., 1.1176e-08, + -1.7695e-08, 1.2107e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-08, ..., 1.1176e-08, + 1.5832e-08, 5.4948e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-08, ..., 2.7940e-08, + 3.7253e-08, -4.9546e-07]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0305, 0.0142, 0.0239, 0.0092, 0.0359, -0.0115, -0.0015, -0.0145, + -0.0280, 0.0117], device='cuda:0'), grad: tensor([-1.4715e-07, 2.9337e-07, -2.7455e-06, -2.6170e-07, 9.9652e-08, + 7.0129e-07, -4.8243e-07, 2.7120e-06, 4.7684e-07, -6.4168e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 221.11, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4877 re_mapping 0.0048 re_causal 0.0142 /// teacc 98.99 lr 0.00010000 +Epoch 273, weight, value: tensor([[ 0.0013, 0.0110, 0.0357, ..., 0.0757, -0.2092, -0.1581], + [ 0.0551, -0.0185, -0.0177, ..., -0.2260, 0.0481, -0.0999], + [-0.0055, -0.0048, 0.1947, ..., -0.3043, -0.1178, -0.1533], + ..., + [-0.0226, 0.0062, -0.1714, ..., -0.0238, 0.1354, 0.0669], + [-0.0152, -0.0202, -0.1359, ..., -0.2621, 0.0999, -0.1887], + [ 0.0016, -0.0258, -0.0712, ..., 0.0391, -0.1169, 0.0140]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.4459e-08, ..., 1.8626e-09, + 6.3330e-08, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 1.1548e-07, ..., 6.9849e-08, + 1.6019e-07, 2.0023e-07], + [ 0.0000e+00, 0.0000e+00, 2.1141e-07, ..., 1.1176e-08, + 4.2189e-07, 1.7043e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.6764e-08, + -4.0978e-08, -1.4622e-07], + [ 0.0000e+00, 0.0000e+00, 1.4529e-07, ..., 3.7253e-09, + -6.6496e-07, 7.6368e-08], + [ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 1.0524e-07, + 5.0291e-08, 2.4494e-07]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0289, 0.0143, 0.0237, 0.0092, 0.0363, -0.0112, -0.0016, -0.0145, + -0.0282, 0.0095], device='cuda:0'), grad: tensor([ 1.6298e-07, 5.9791e-07, 1.0598e-06, 1.5758e-06, 1.7509e-07, + 1.2107e-07, 5.9512e-07, -2.9337e-07, -4.7758e-06, 7.8697e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 220.97, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4860 re_mapping 0.0046 re_causal 0.0137 /// teacc 99.01 lr 0.00010000 +Epoch 274, weight, value: tensor([[ 0.0011, 0.0110, 0.0357, ..., 0.0758, -0.2103, -0.1581], + [ 0.0546, -0.0185, -0.0180, ..., -0.2269, 0.0489, -0.0996], + [-0.0056, -0.0048, 0.1950, ..., -0.3047, -0.1180, -0.1536], + ..., + [-0.0226, 0.0062, -0.1716, ..., -0.0244, 0.1352, 0.0666], + [-0.0156, -0.0202, -0.1367, ..., -0.2628, 0.1000, -0.1893], + [ 0.0013, -0.0258, -0.0714, ..., 0.0390, -0.1171, 0.0138]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5891e-07, ..., -1.2387e-07, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.3120e-07, ..., 6.5193e-09, + 8.0094e-08, 1.5274e-07], + [ 0.0000e+00, 0.0000e+00, -9.7882e-07, ..., 1.3970e-08, + 1.8626e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7905e-07, ..., 2.5146e-08, + -1.3132e-07, -1.8533e-07], + [ 0.0000e+00, 0.0000e+00, 5.3085e-08, ..., 9.3132e-09, + 3.7253e-09, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 8.4750e-08, ..., -4.4703e-08, + 1.3039e-08, -3.3900e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0289, 0.0147, 0.0236, 0.0092, 0.0366, -0.0098, -0.0018, -0.0148, + -0.0283, 0.0092], device='cuda:0'), grad: tensor([-6.4075e-07, 1.0673e-06, -1.6093e-06, 4.2841e-08, 4.9174e-07, + 1.9092e-07, 2.1700e-07, 2.6915e-07, 1.6671e-07, -1.9092e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 220.49, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4972 re_mapping 0.0043 re_causal 0.0136 /// teacc 98.97 lr 0.00010000 +Epoch 275, weight, value: tensor([[ 0.0011, 0.0110, 0.0358, ..., 0.0759, -0.2119, -0.1581], + [ 0.0544, -0.0185, -0.0180, ..., -0.2283, 0.0488, -0.0996], + [-0.0056, -0.0048, 0.1952, ..., -0.3052, -0.1184, -0.1540], + ..., + [-0.0226, 0.0062, -0.1718, ..., -0.0245, 0.1353, 0.0666], + [-0.0157, -0.0202, -0.1376, ..., -0.2635, 0.0999, -0.1902], + [ 0.0015, -0.0258, -0.0714, ..., 0.0394, -0.1173, 0.0153]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.4064e-08, ..., -4.6566e-09, + 4.4703e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.1828e-06, ..., 8.3819e-09, + -1.0803e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -1.3066e-06, ..., 2.7940e-09, + 3.1665e-08, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0990e-07, ..., 9.3132e-10, + 1.2107e-08, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.8626e-09, + 8.3819e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -1.8626e-09, + 8.3819e-09, -7.4506e-09]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0289, 0.0147, 0.0236, 0.0092, 0.0351, -0.0099, -0.0017, -0.0148, + -0.0285, 0.0099], device='cuda:0'), grad: tensor([ 2.5798e-07, 1.3914e-06, -1.8002e-06, 1.6764e-08, 4.8429e-08, + -4.5635e-08, -2.2631e-07, 2.0582e-07, 9.3132e-08, 5.4017e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 220.86, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4715 re_mapping 0.0046 re_causal 0.0133 /// teacc 98.92 lr 0.00010000 +Epoch 276, weight, value: tensor([[ 0.0011, 0.0110, 0.0357, ..., 0.0751, -0.2134, -0.1592], + [ 0.0544, -0.0185, -0.0180, ..., -0.2294, 0.0486, -0.0996], + [-0.0056, -0.0048, 0.1954, ..., -0.3062, -0.1190, -0.1548], + ..., + [-0.0226, 0.0062, -0.1721, ..., -0.0246, 0.1355, 0.0667], + [-0.0158, -0.0202, -0.1383, ..., -0.2638, 0.1003, -0.1903], + [ 0.0016, -0.0258, -0.0710, ..., 0.0401, -0.1176, 0.0156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.2554e-06, ..., 1.9744e-07, + 9.3132e-09, 1.3728e-06], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 9.1270e-08, + -5.8673e-07, 2.3134e-06], + [ 0.0000e+00, 0.0000e+00, 2.6915e-07, ..., 5.5414e-07, + 5.3644e-07, 5.3138e-05], + ..., + [ 0.0000e+00, 0.0000e+00, 7.1712e-08, ..., 1.1269e-07, + -4.5542e-07, -5.1975e-05], + [ 0.0000e+00, 0.0000e+00, 1.6112e-07, ..., 4.7125e-07, + 9.7789e-08, 1.9222e-06], + [ 0.0000e+00, 0.0000e+00, 1.5181e-07, ..., -5.7593e-06, + 8.9407e-08, -1.5765e-05]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0292, 0.0147, 0.0231, 0.0092, 0.0351, -0.0100, -0.0017, -0.0147, + -0.0284, 0.0103], device='cuda:0'), grad: tensor([-1.8440e-06, 2.0005e-06, 2.0063e-04, 1.8366e-06, 1.5482e-05, + 9.3319e-07, 3.9041e-06, -1.9765e-04, 3.6843e-06, -2.8878e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 220.81, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.4872 re_mapping 0.0048 re_causal 0.0138 /// teacc 98.94 lr 0.00010000 +Epoch 277, weight, value: tensor([[ 0.0011, 0.0110, 0.0358, ..., 0.0751, -0.2153, -0.1594], + [ 0.0544, -0.0185, -0.0180, ..., -0.2323, 0.0492, -0.0997], + [-0.0056, -0.0048, 0.1954, ..., -0.3067, -0.1211, -0.1571], + ..., + [-0.0226, 0.0062, -0.1725, ..., -0.0247, 0.1355, 0.0668], + [-0.0159, -0.0202, -0.1401, ..., -0.2640, 0.1003, -0.1915], + [ 0.0011, -0.0258, -0.0711, ..., 0.0400, -0.1178, 0.0151]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.3062e-06, ..., -4.7162e-06, + 2.2352e-08, -3.4738e-07], + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 1.8533e-07, + -5.4017e-08, 3.6787e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 3.7253e-08, + 5.5879e-08, 8.3819e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 9.5926e-08, + 3.7253e-09, 2.4214e-07], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 6.5193e-09, + 1.2107e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.1814e-06, ..., 4.5523e-06, + 3.7253e-09, 3.6042e-07]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0293, 0.0149, 0.0210, 0.0092, 0.0359, -0.0080, -0.0018, -0.0145, + -0.0299, 0.0100], device='cuda:0'), grad: tensor([-1.2144e-05, 3.4180e-07, 4.1071e-07, 1.9092e-07, -8.5495e-07, + 4.2003e-07, -7.4785e-07, 4.5728e-07, 9.3132e-08, 1.1846e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 220.95, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4897 re_mapping 0.0049 re_causal 0.0141 /// teacc 98.89 lr 0.00010000 +Epoch 278, weight, value: tensor([[ 0.0011, 0.0110, 0.0358, ..., 0.0751, -0.2162, -0.1597], + [ 0.0544, -0.0185, -0.0188, ..., -0.2336, 0.0490, -0.0997], + [-0.0056, -0.0048, 0.1963, ..., -0.3071, -0.1215, -0.1573], + ..., + [-0.0226, 0.0062, -0.1727, ..., -0.0249, 0.1356, 0.0669], + [-0.0159, -0.0202, -0.1407, ..., -0.2647, 0.1009, -0.1940], + [ 0.0012, -0.0258, -0.0713, ..., 0.0400, -0.1180, 0.0152]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5611e-07, ..., -4.7218e-07, + 2.2352e-08, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 4.4238e-07, + 1.1073e-06, 2.1197e-06], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 1.0245e-08, + 3.7253e-09, 2.2352e-08], + ..., + [ 2.1793e-07, 0.0000e+00, 5.5879e-09, ..., 7.9162e-08, + -8.9966e-07, -7.7300e-07], + [ 9.3132e-10, 0.0000e+00, 2.0489e-08, ..., 1.2852e-07, + -2.8405e-07, 6.1374e-07], + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., -3.0268e-07, + -3.4831e-07, -2.4922e-06]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0293, 0.0147, 0.0216, 0.0093, 0.0361, -0.0091, -0.0016, -0.0145, + -0.0297, 0.0100], device='cuda:0'), grad: tensor([-2.2613e-06, 3.6471e-06, 7.9162e-08, -5.3905e-06, -1.2163e-06, + -8.5589e-07, 3.7272e-06, 4.1574e-06, 9.2294e-07, -2.8145e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 220.57, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.5051 re_mapping 0.0049 re_causal 0.0142 /// teacc 98.94 lr 0.00010000 +Epoch 279, weight, value: tensor([[ 0.0010, 0.0110, 0.0358, ..., 0.0750, -0.2181, -0.1600], + [ 0.0544, -0.0185, -0.0190, ..., -0.2346, 0.0496, -0.0997], + [-0.0056, -0.0048, 0.1963, ..., -0.3071, -0.1229, -0.1591], + ..., + [-0.0226, 0.0062, -0.1721, ..., -0.0250, 0.1356, 0.0669], + [-0.0159, -0.0202, -0.1416, ..., -0.2649, 0.1011, -0.1951], + [ 0.0012, -0.0258, -0.0713, ..., 0.0400, -0.1185, 0.0156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9837e-07, ..., -1.2759e-07, + 1.9558e-08, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 3.8184e-07, + -5.3719e-06, 6.8080e-07], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 2.1420e-08, + 1.5181e-07, 3.0734e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 4.9174e-07, + 1.7229e-06, 1.4128e-06], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 2.3283e-08, + 1.5739e-07, 6.4261e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-08, ..., 9.9279e-07, + 3.2447e-06, 1.7649e-06]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0294, 0.0148, 0.0203, 0.0094, 0.0360, -0.0107, -0.0012, -0.0144, + -0.0296, 0.0102], device='cuda:0'), grad: tensor([-7.5903e-07, -1.3642e-05, 5.3458e-07, 2.5053e-07, -6.4820e-06, + 2.6356e-07, 2.5425e-07, 6.9700e-06, 5.8860e-07, 1.2010e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 220.77, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4966 re_mapping 0.0044 re_causal 0.0137 /// teacc 98.99 lr 0.00010000 +Epoch 280, weight, value: tensor([[ 0.0010, 0.0110, 0.0358, ..., 0.0751, -0.2186, -0.1601], + [ 0.0544, -0.0185, -0.0192, ..., -0.2359, 0.0496, -0.0997], + [-0.0056, -0.0048, 0.1967, ..., -0.3072, -0.1232, -0.1593], + ..., + [-0.0226, 0.0062, -0.1723, ..., -0.0254, 0.1357, 0.0669], + [-0.0160, -0.0202, -0.1420, ..., -0.2657, 0.1010, -0.1962], + [ 0.0013, -0.0258, -0.0714, ..., 0.0401, -0.1186, 0.0159]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.0734e-08, ..., -1.4901e-08, + 9.3132e-09, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.3528e-08, + -1.7695e-07, 6.5193e-08], + [ 0.0000e+00, 0.0000e+00, -2.4214e-08, ..., 4.6566e-09, + 2.1420e-08, 1.5832e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 3.1665e-08, + -3.3528e-08, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 6.5193e-09, + 1.5832e-08, 4.2841e-08], + [ 0.0000e+00, 0.0000e+00, 4.3772e-08, ..., 1.3318e-07, + 9.3132e-08, -4.6566e-09]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0294, 0.0148, 0.0202, 0.0094, 0.0356, -0.0115, -0.0006, -0.0144, + -0.0298, 0.0104], device='cuda:0'), grad: tensor([ 8.3819e-09, -6.8918e-07, 9.9652e-08, -9.4064e-07, -1.4715e-07, + 9.5274e-07, 2.5425e-07, 7.4506e-09, 1.6764e-07, 3.0082e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 220.65, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4999 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.08 lr 0.00010000 +Epoch 281, weight, value: tensor([[ 0.0010, 0.0110, 0.0359, ..., 0.0751, -0.2192, -0.1602], + [ 0.0544, -0.0185, -0.0193, ..., -0.2367, 0.0495, -0.0997], + [-0.0057, -0.0048, 0.1979, ..., -0.3064, -0.1235, -0.1579], + ..., + [-0.0226, 0.0062, -0.1741, ..., -0.0253, 0.1358, 0.0670], + [-0.0162, -0.0202, -0.1428, ..., -0.2660, 0.1019, -0.1961], + [ 0.0013, -0.0258, -0.0714, ..., 0.0401, -0.1190, 0.0158]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.7591e-07, ..., -3.3434e-07, + 1.7695e-08, 3.7253e-08], + [ 2.7940e-09, 0.0000e+00, 2.0117e-07, ..., 8.2888e-08, + 4.2841e-08, 2.3935e-07], + [ 7.2643e-08, 0.0000e+00, -5.4389e-07, ..., 6.1467e-08, + 9.5274e-07, 2.0713e-06], + ..., + [-8.1025e-08, 0.0000e+00, -9.4064e-07, ..., 2.7940e-08, + -1.2992e-06, -2.9467e-06], + [ 1.8626e-09, 0.0000e+00, 2.4959e-07, ..., 1.5274e-07, + -9.2201e-08, 2.0489e-08], + [ 9.3132e-10, 0.0000e+00, 8.2888e-08, ..., 7.1712e-08, + 1.6764e-07, 1.3225e-07]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0293, 0.0148, 0.0204, 0.0094, 0.0352, -0.0121, -0.0007, -0.0143, + -0.0294, 0.0103], device='cuda:0'), grad: tensor([-1.4193e-06, 8.0280e-07, 3.2149e-06, 2.9150e-06, -2.3469e-07, + 1.3132e-07, 1.1269e-07, -6.8694e-06, 4.7963e-07, 8.4192e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 220.88, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.5049 re_mapping 0.0048 re_causal 0.0145 /// teacc 98.95 lr 0.00010000 +Epoch 282, weight, value: tensor([[ 0.0010, 0.0110, 0.0359, ..., 0.0747, -0.2206, -0.1611], + [ 0.0544, -0.0185, -0.0195, ..., -0.2375, 0.0499, -0.0994], + [-0.0057, -0.0048, 0.1982, ..., -0.3087, -0.1240, -0.1586], + ..., + [-0.0227, 0.0062, -0.1746, ..., -0.0255, 0.1358, 0.0668], + [-0.0162, -0.0202, -0.1436, ..., -0.2663, 0.1019, -0.1964], + [ 0.0021, -0.0258, -0.0713, ..., 0.0404, -0.1198, 0.0156]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.1956e-08, ..., 0.0000e+00, + 0.0000e+00, 5.8860e-07], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 7.4506e-09, + -2.7940e-09, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 1.5553e-07, ..., 1.8626e-09, + 1.8626e-09, 1.0878e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., 7.4506e-09, + -1.5832e-08, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, -2.5425e-07, ..., 2.7940e-08, + 1.1176e-08, -1.8599e-06]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0296, 0.0151, 0.0202, 0.0094, 0.0355, -0.0124, -0.0002, -0.0144, + -0.0296, 0.0103], device='cuda:0'), grad: tensor([ 7.5717e-07, 5.4948e-08, 1.4557e-06, -5.3160e-06, 1.9558e-08, + 5.0478e-06, 4.0978e-08, 8.8476e-08, 1.4063e-07, -2.2948e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 220.57, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.5064 re_mapping 0.0048 re_causal 0.0147 /// teacc 98.93 lr 0.00010000 +Epoch 283, weight, value: tensor([[ 0.0010, 0.0110, 0.0359, ..., 0.0748, -0.2220, -0.1613], + [ 0.0543, -0.0185, -0.0198, ..., -0.2384, 0.0502, -0.0995], + [-0.0057, -0.0048, 0.1985, ..., -0.3100, -0.1242, -0.1593], + ..., + [-0.0227, 0.0062, -0.1749, ..., -0.0256, 0.1358, 0.0667], + [-0.0162, -0.0202, -0.1440, ..., -0.2669, 0.1020, -0.1972], + [ 0.0026, -0.0258, -0.0716, ..., 0.0404, -0.1200, 0.0161]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -7.4506e-09, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-08, ..., 9.3132e-10, + -2.4939e-04, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -2.0768e-07, ..., 0.0000e+00, + 4.5449e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 7.1712e-08, ..., 1.8626e-09, + 2.4819e-04, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 4.0978e-08, ..., 1.8626e-09, + 8.9966e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 0.0000e+00, + 5.0664e-07, -5.5879e-09]], device='cuda:0') +Epoch 283, bias, value: tensor([-2.9378e-02, 1.5172e-02, 2.0078e-02, 9.3762e-03, 3.5473e-02, + -1.2130e-02, 1.8693e-05, -1.4565e-02, -2.9749e-02, 1.0314e-02], + device='cuda:0'), grad: tensor([ 9.9652e-08, -9.3365e-04, 1.3057e-06, -4.2081e-05, 1.6578e-07, + 3.6448e-05, 3.4813e-06, 9.2936e-04, 2.9393e-06, 1.9278e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 220.97, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4792 re_mapping 0.0049 re_causal 0.0134 /// teacc 98.98 lr 0.00010000 +Epoch 284, weight, value: tensor([[ 0.0009, 0.0110, 0.0360, ..., 0.0749, -0.2225, -0.1615], + [ 0.0529, -0.0185, -0.0199, ..., -0.2406, 0.0539, -0.0991], + [-0.0059, -0.0048, 0.1993, ..., -0.3101, -0.1247, -0.1589], + ..., + [-0.0226, 0.0062, -0.1752, ..., -0.0258, 0.1337, 0.0667], + [-0.0168, -0.0202, -0.1458, ..., -0.2671, 0.1026, -0.1976], + [ 0.0025, -0.0258, -0.0719, ..., 0.0399, -0.1226, 0.0145]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.5193e-08, ..., -4.8429e-08, + 3.1665e-08, 6.2399e-08], + [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 5.6811e-08, + 7.1712e-08, 2.1420e-07], + [-4.6566e-09, 0.0000e+00, -8.3819e-09, ..., 6.5193e-09, + 3.5390e-08, 2.8871e-08], + ..., + [ 9.3132e-10, 0.0000e+00, 9.3132e-09, ..., 1.1548e-07, + -4.8894e-07, -3.7532e-07], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 1.1176e-08, + 9.3132e-09, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., -5.1223e-08, + 2.8778e-07, 1.0151e-07]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0293, 0.0176, 0.0203, 0.0088, 0.0367, -0.0092, -0.0002, -0.0158, + -0.0298, 0.0091], device='cuda:0'), grad: tensor([-3.0734e-08, 4.3306e-07, 9.7789e-08, 8.8476e-08, -1.1642e-07, + 2.9802e-08, 6.3330e-08, -1.7621e-06, 9.2201e-08, 1.1064e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 220.48, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4709 re_mapping 0.0052 re_causal 0.0141 /// teacc 98.88 lr 0.00010000 +Epoch 285, weight, value: tensor([[ 0.0008, 0.0110, 0.0361, ..., 0.0752, -0.2227, -0.1616], + [ 0.0528, -0.0185, -0.0204, ..., -0.2418, 0.0534, -0.0992], + [-0.0059, -0.0048, 0.2002, ..., -0.3103, -0.1273, -0.1607], + ..., + [-0.0226, 0.0062, -0.1753, ..., -0.0260, 0.1344, 0.0670], + [-0.0182, -0.0202, -0.1479, ..., -0.2679, 0.1025, -0.1978], + [ 0.0027, -0.0258, -0.0721, ..., 0.0395, -0.1231, 0.0139]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 8.6613e-08, + 5.8394e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 8.3819e-09, + 3.9116e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 9.3132e-10, + 5.8021e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 4.6566e-09, + 5.5879e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 5.5879e-09, + -1.0245e-06, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 4.6566e-09, + 1.6764e-08, -1.4901e-08]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0291, 0.0173, 0.0195, 0.0089, 0.0374, -0.0095, -0.0003, -0.0155, + -0.0302, 0.0086], device='cuda:0'), grad: tensor([ 2.1234e-06, 2.4680e-07, 2.5108e-06, 1.9278e-07, 4.6566e-07, + 5.1595e-07, -2.3507e-06, 1.1921e-07, -3.9414e-06, 1.0990e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 220.38, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.5085 re_mapping 0.0045 re_causal 0.0136 /// teacc 98.99 lr 0.00010000 +Epoch 286, weight, value: tensor([[ 0.0007, 0.0110, 0.0362, ..., 0.0752, -0.2243, -0.1617], + [ 0.0530, -0.0185, -0.0212, ..., -0.2433, 0.0531, -0.0994], + [-0.0061, -0.0048, 0.2006, ..., -0.3104, -0.1298, -0.1615], + ..., + [-0.0226, 0.0062, -0.1754, ..., -0.0261, 0.1348, 0.0670], + [-0.0184, -0.0202, -0.1486, ..., -0.2683, 0.1025, -0.1985], + [ 0.0026, -0.0258, -0.0722, ..., 0.0395, -0.1233, 0.0142]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.6858e-08, ..., -8.3819e-08, + 2.1420e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 5.1223e-08, + 3.1386e-07, 9.5926e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 7.4506e-09, + 4.8615e-07, 2.7940e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 8.3819e-08, + -8.6799e-07, 1.1735e-07], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 2.2352e-08, + -3.7160e-07, 5.1223e-08], + [-2.7940e-08, 0.0000e+00, 2.7940e-08, ..., 4.3772e-07, + 1.8813e-07, 4.6939e-07]], device='cuda:0') +Epoch 286, bias, value: tensor([-2.8958e-02, 1.7021e-02, 1.9224e-02, 8.7337e-03, 3.7541e-02, + -8.7722e-03, 8.2775e-05, -1.5397e-02, -3.0287e-02, 8.6985e-03], + device='cuda:0'), grad: tensor([-1.9930e-07, 1.4268e-06, 1.7760e-06, 2.7474e-07, -1.2880e-06, + 1.8440e-07, 3.2596e-07, -2.7493e-06, -1.2396e-06, 1.4985e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 220.55, cls_loss 0.0012 cls_loss_mapping 0.0013 cls_loss_causal 0.4665 re_mapping 0.0046 re_causal 0.0133 /// teacc 99.01 lr 0.00010000 +Epoch 287, weight, value: tensor([[ 0.0006, 0.0110, 0.0362, ..., 0.0752, -0.2253, -0.1620], + [ 0.0564, -0.0185, -0.0213, ..., -0.2439, 0.0540, -0.0993], + [-0.0086, -0.0048, 0.2012, ..., -0.3106, -0.1308, -0.1609], + ..., + [-0.0227, 0.0062, -0.1756, ..., -0.0263, 0.1343, 0.0669], + [-0.0185, -0.0202, -0.1490, ..., -0.2700, 0.1030, -0.1994], + [ 0.0027, -0.0258, -0.0725, ..., 0.0396, -0.1236, 0.0143]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1642e-07, ..., -1.0431e-07, + 8.3819e-09, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-08, ..., 1.5832e-08, + 1.2107e-08, 5.8673e-08], + [ 0.0000e+00, 0.0000e+00, -2.6766e-06, ..., 1.0245e-08, + -4.5076e-07, -2.3954e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.4473e-06, ..., 1.1176e-08, + 3.5297e-07, 3.3565e-06], + [ 0.0000e+00, 0.0000e+00, 1.3197e-06, ..., 1.0245e-08, + 2.3469e-07, 1.1325e-06], + [ 0.0000e+00, 0.0000e+00, 1.2787e-06, ..., 2.1420e-08, + -1.6671e-07, -1.9670e-06]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0288, 0.0185, 0.0186, 0.0088, 0.0374, -0.0096, -0.0007, -0.0162, + -0.0302, 0.0088], device='cuda:0'), grad: tensor([-4.1910e-07, 1.7229e-07, -5.3309e-06, -2.9504e-06, 3.7625e-07, + -1.2666e-07, -1.8626e-07, 5.7891e-06, 3.2447e-06, -5.7649e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 220.70, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4622 re_mapping 0.0045 re_causal 0.0137 /// teacc 98.93 lr 0.00010000 +Epoch 288, weight, value: tensor([[ 0.0006, 0.0110, 0.0363, ..., 0.0753, -0.2255, -0.1618], + [ 0.0564, -0.0185, -0.0218, ..., -0.2452, 0.0537, -0.0993], + [-0.0087, -0.0048, 0.2024, ..., -0.3107, -0.1310, -0.1609], + ..., + [-0.0227, 0.0062, -0.1759, ..., -0.0264, 0.1345, 0.0670], + [-0.0187, -0.0202, -0.1497, ..., -0.2702, 0.1031, -0.1994], + [ 0.0029, -0.0258, -0.0729, ..., 0.0395, -0.1242, 0.0142]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.4448e-07, ..., -3.3434e-07, + 3.1665e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 5.5879e-09, + 1.3039e-08, 1.9558e-08], + [ 0.0000e+00, 0.0000e+00, 1.6764e-07, ..., 1.5274e-07, + 9.3132e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 1.8626e-09, + -7.7300e-08, -4.3772e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-07, ..., 6.4261e-08, + 2.7940e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 9.7789e-08, ..., 5.1223e-08, + 2.4214e-08, 2.3283e-08]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0286, 0.0183, 0.0188, 0.0087, 0.0372, -0.0094, -0.0006, -0.0159, + -0.0303, 0.0085], device='cuda:0'), grad: tensor([-1.1465e-06, 7.2643e-08, 4.3958e-07, 8.8569e-07, 2.4214e-08, + -7.2736e-07, -1.6205e-07, -1.6671e-07, 4.0978e-07, 3.6042e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 221.21, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4891 re_mapping 0.0046 re_causal 0.0136 /// teacc 98.94 lr 0.00010000 +Epoch 289, weight, value: tensor([[ 0.0005, 0.0110, 0.0366, ..., 0.0756, -0.2265, -0.1614], + [ 0.0563, -0.0185, -0.0220, ..., -0.2482, 0.0534, -0.1003], + [-0.0086, -0.0048, 0.2031, ..., -0.3106, -0.1316, -0.1609], + ..., + [-0.0227, 0.0062, -0.1763, ..., -0.0259, 0.1350, 0.0674], + [-0.0191, -0.0202, -0.1502, ..., -0.2713, 0.1033, -0.2003], + [ 0.0041, -0.0258, -0.0740, ..., 0.0397, -0.1245, 0.0167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -3.7253e-09, + 1.5832e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 1.8626e-09, + -4.3772e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, -6.4541e-07, ..., 0.0000e+00, + 3.1292e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.1828e-07, ..., 2.7940e-09, + 1.8626e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -4.3772e-07, ..., 0.0000e+00, + -5.6159e-07, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -1.8626e-09, + 1.1455e-07, -2.7008e-08]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0282, 0.0172, 0.0191, 0.0088, 0.0342, -0.0098, -0.0005, -0.0157, + -0.0303, 0.0105], device='cuda:0'), grad: tensor([ 5.2154e-08, -6.9849e-08, 8.5309e-07, 1.1194e-06, 2.7008e-08, + 1.9558e-08, 6.0536e-08, 1.7853e-06, -4.1723e-06, 3.2689e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 220.99, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.4897 re_mapping 0.0048 re_causal 0.0137 /// teacc 99.01 lr 0.00010000 +Epoch 290, weight, value: tensor([[ 2.1439e-04, 1.0998e-02, 3.6620e-02, ..., 7.5673e-02, + -2.2930e-01, -1.6144e-01], + [ 5.6262e-02, -1.8468e-02, -2.3394e-02, ..., -2.4891e-01, + 5.3238e-02, -1.0040e-01], + [-8.8295e-03, -4.7728e-03, 2.0407e-01, ..., -3.1121e-01, + -1.3213e-01, -1.6107e-01], + ..., + [-2.2740e-02, 6.2061e-03, -1.7686e-01, ..., -2.6144e-02, + 1.3504e-01, 6.7378e-02], + [-2.0624e-02, -2.0220e-02, -1.4894e-01, ..., -2.7172e-01, + 1.0426e-01, -2.0103e-01], + [ 5.2439e-03, -2.5834e-02, -7.4102e-02, ..., 3.9734e-02, + -1.2440e-01, 1.7342e-02]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 4.1425e-06, ..., -2.4214e-08, + 3.4459e-08, 8.8476e-09], + [ 3.3667e-07, 0.0000e+00, 2.5555e-06, ..., 6.0536e-09, + 1.5162e-06, 6.8266e-07], + [ 6.4261e-08, 0.0000e+00, -1.4380e-05, ..., 6.5193e-09, + 2.9476e-07, 1.3458e-07], + ..., + [-5.6159e-07, 0.0000e+00, 3.4943e-06, ..., 1.8626e-09, + -2.6040e-06, -1.1921e-06], + [ 6.3796e-08, 0.0000e+00, 2.0266e-06, ..., 8.8476e-09, + 3.6787e-08, 1.2992e-07], + [ 4.1444e-08, 0.0000e+00, 2.3236e-07, ..., 2.3749e-07, + 4.0932e-07, 4.7777e-07]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0284, 0.0170, 0.0173, 0.0087, 0.0334, -0.0096, -0.0007, -0.0154, + -0.0292, 0.0110], device='cuda:0'), grad: tensor([ 7.8902e-06, 8.9854e-06, -3.1978e-05, 2.9337e-06, -3.5856e-08, + 2.8452e-07, 1.4426e-06, 4.0308e-06, 4.1276e-06, 2.2668e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 220.51, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4684 re_mapping 0.0044 re_causal 0.0135 /// teacc 98.97 lr 0.00010000 +Epoch 291, weight, value: tensor([[ 1.0084e-04, 1.0998e-02, 3.6778e-02, ..., 7.5777e-02, + -2.3005e-01, -1.6140e-01], + [ 5.6171e-02, -1.8468e-02, -2.3119e-02, ..., -2.4963e-01, + 5.3226e-02, -1.0045e-01], + [-8.9897e-03, -4.7728e-03, 2.0462e-01, ..., -3.1156e-01, + -1.3370e-01, -1.6121e-01], + ..., + [-2.2739e-02, 6.2061e-03, -1.7715e-01, ..., -2.6300e-02, + 1.3529e-01, 6.7492e-02], + [-2.1068e-02, -2.0220e-02, -1.4981e-01, ..., -2.7200e-01, + 1.0472e-01, -2.0138e-01], + [ 4.6935e-03, -2.5834e-02, -7.4619e-02, ..., 3.9694e-02, + -1.2506e-01, 1.7284e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3912e-07, ..., -4.8289e-07, + 5.1223e-09, 9.3132e-10], + [ 4.6566e-10, 0.0000e+00, 4.6100e-08, ..., 4.7963e-08, + -2.0955e-08, 1.5367e-08], + [ 0.0000e+00, 0.0000e+00, 1.2247e-07, ..., 8.8476e-08, + 3.2596e-09, 4.1910e-09], + ..., + [-3.2596e-09, 0.0000e+00, 4.8429e-08, ..., 4.3772e-08, + 6.0536e-09, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.2201e-08, ..., 7.5903e-08, + 6.0536e-09, 6.5193e-09], + [ 9.3132e-10, 0.0000e+00, 1.5600e-07, ..., 3.2783e-07, + 2.7940e-09, 2.1514e-07]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0282, 0.0170, 0.0169, 0.0088, 0.0334, -0.0104, -0.0008, -0.0152, + -0.0290, 0.0107], device='cuda:0'), grad: tensor([-1.3541e-06, 9.9652e-08, 4.0699e-07, -2.2352e-07, -3.7905e-07, + 2.1700e-07, 1.0710e-08, 1.7323e-07, 2.7195e-07, 7.8790e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 220.74, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4948 re_mapping 0.0044 re_causal 0.0132 /// teacc 98.93 lr 0.00010000 +Epoch 292, weight, value: tensor([[ 6.8219e-05, 1.0998e-02, 3.6802e-02, ..., 7.5664e-02, + -2.3363e-01, -1.6158e-01], + [ 5.6292e-02, -1.8468e-02, -2.3233e-02, ..., -2.5039e-01, + 5.2650e-02, -1.0047e-01], + [-9.0924e-03, -4.7728e-03, 2.0491e-01, ..., -3.1211e-01, + -1.3435e-01, -1.6142e-01], + ..., + [-2.2739e-02, 6.2061e-03, -1.7755e-01, ..., -2.6598e-02, + 1.3551e-01, 6.7509e-02], + [-2.1234e-02, -2.0220e-02, -1.5047e-01, ..., -2.7268e-01, + 1.0601e-01, -2.0164e-01], + [ 4.7603e-03, -2.5834e-02, -7.4729e-02, ..., 3.9858e-02, + -1.2543e-01, 1.7372e-02]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., -2.7940e-09, + 2.9337e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 8.8476e-09, + 2.0489e-08, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., 0.0000e+00, + 4.9826e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 7.4506e-09, + 4.6566e-09, 3.0268e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 4.6566e-10, + -6.4727e-08, 1.3970e-09], + [-2.3283e-09, 0.0000e+00, 5.1223e-09, ..., -1.5367e-08, + 4.1910e-08, -9.4995e-08]], device='cuda:0') +Epoch 292, bias, value: tensor([-2.8331e-02, 1.6652e-02, 1.6853e-02, 8.6544e-03, 3.3372e-02, + -1.0386e-02, 3.2685e-05, -1.5136e-02, -2.8345e-02, 1.0769e-02], + device='cuda:0'), grad: tensor([ 7.9628e-08, 8.5682e-08, 8.2888e-08, 4.9360e-08, -2.3236e-07, + 1.8161e-08, 5.5879e-08, 7.4506e-08, -1.8673e-07, -1.4435e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 220.97, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4932 re_mapping 0.0046 re_causal 0.0132 /// teacc 98.98 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.0004, 0.0110, 0.0368, ..., 0.0757, -0.2347, -0.1617], + [ 0.0562, -0.0185, -0.0235, ..., -0.2515, 0.0519, -0.1005], + [-0.0075, -0.0048, 0.2058, ..., -0.3125, -0.1346, -0.1614], + ..., + [-0.0227, 0.0062, -0.1793, ..., -0.0268, 0.1364, 0.0676], + [-0.0217, -0.0202, -0.1511, ..., -0.2729, 0.1078, -0.2013], + [ 0.0052, -0.0258, -0.0749, ..., 0.0398, -0.1261, 0.0173]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 3.4459e-08, + 1.2713e-06, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 2.3283e-09, + 3.2596e-08, 5.5879e-08], + [ 1.8626e-09, 0.0000e+00, -7.1246e-08, ..., 1.7695e-08, + 1.3039e-08, 2.5146e-08], + ..., + [-4.6566e-09, 0.0000e+00, 4.7497e-08, ..., 1.9092e-08, + -8.9407e-08, -1.3039e-07], + [ 0.0000e+00, 0.0000e+00, 4.0047e-08, ..., 3.7253e-09, + 1.2154e-07, 5.1223e-09], + [ 4.6566e-10, 0.0000e+00, 6.0536e-09, ..., -3.0920e-07, + 7.2177e-08, -2.3469e-07]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0284, 0.0162, 0.0173, 0.0084, 0.0335, -0.0102, -0.0006, -0.0148, + -0.0272, 0.0106], device='cuda:0'), grad: tensor([ 1.9580e-05, 2.5379e-07, -6.5193e-09, 0.0000e+00, 5.0524e-07, + 9.6977e-05, -1.1879e-04, -6.7055e-08, 1.9148e-06, -9.9652e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 221.37, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4653 re_mapping 0.0047 re_causal 0.0134 /// teacc 99.00 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.0004, 0.0110, 0.0369, ..., 0.0758, -0.2364, -0.1617], + [ 0.0562, -0.0185, -0.0258, ..., -0.2522, 0.0517, -0.1005], + [-0.0075, -0.0048, 0.2094, ..., -0.3126, -0.1329, -0.1595], + ..., + [-0.0227, 0.0062, -0.1838, ..., -0.0271, 0.1364, 0.0676], + [-0.0218, -0.0202, -0.1522, ..., -0.2732, 0.1083, -0.2015], + [ 0.0051, -0.0258, -0.0750, ..., 0.0399, -0.1266, 0.0173]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.7521e-08, ..., 4.1910e-09, + 2.2724e-07, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 4.0978e-08, ..., 1.6298e-08, + -7.6648e-07, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, -6.5286e-07, ..., -7.5437e-08, + 8.3353e-08, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.6392e-08, ..., 2.8405e-08, + 2.9337e-08, 2.0023e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-08, ..., 1.5507e-07, + 2.8126e-07, 1.7509e-07], + [ 0.0000e+00, 0.0000e+00, 1.4994e-07, ..., 4.2096e-07, + 2.5611e-08, 4.4750e-07]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0284, 0.0160, 0.0198, 0.0084, 0.0335, -0.0109, -0.0002, -0.0150, + -0.0270, 0.0105], device='cuda:0'), grad: tensor([ 9.7603e-07, -2.6766e-06, -1.0915e-06, 2.2212e-07, -1.1548e-06, + 1.1129e-07, 2.7241e-07, 3.7160e-07, 1.5832e-06, 1.3886e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 220.90, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4632 re_mapping 0.0046 re_causal 0.0135 /// teacc 99.09 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.0004, 0.0110, 0.0369, ..., 0.0758, -0.2378, -0.1618], + [ 0.0562, -0.0185, -0.0260, ..., -0.2528, 0.0521, -0.1005], + [-0.0075, -0.0048, 0.2098, ..., -0.3129, -0.1331, -0.1596], + ..., + [-0.0227, 0.0062, -0.1840, ..., -0.0273, 0.1362, 0.0676], + [-0.0218, -0.0202, -0.1537, ..., -0.2760, 0.1087, -0.2030], + [ 0.0051, -0.0258, -0.0751, ..., 0.0400, -0.1269, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.6927e-07, ..., -4.1910e-07, + 1.8626e-09, -4.0513e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 8.2888e-08, + -2.9802e-08, 7.6834e-08], + [ 0.0000e+00, 0.0000e+00, 7.2177e-08, ..., 8.0559e-08, + 2.3283e-09, 1.7229e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 4.8429e-08, + 3.2596e-09, 6.2864e-08], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.9092e-08, + 1.0245e-08, 2.1886e-08], + [ 0.0000e+00, 0.0000e+00, 2.0629e-07, ..., 2.8359e-07, + 1.8626e-09, 1.1688e-07]], device='cuda:0') +Epoch 295, bias, value: tensor([-2.8444e-02, 1.6286e-02, 1.9898e-02, 8.3496e-03, 3.3004e-02, + -1.1125e-02, 3.4407e-05, -1.5153e-02, -2.7145e-02, 1.0535e-02], + device='cuda:0'), grad: tensor([-1.5339e-06, 9.7323e-08, 2.8778e-07, 1.8300e-07, -4.3167e-07, + -4.7032e-08, 2.3283e-07, 1.9372e-07, 1.4948e-07, 8.8615e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 221.14, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4780 re_mapping 0.0047 re_causal 0.0135 /// teacc 98.97 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.0006, 0.0110, 0.0369, ..., 0.0759, -0.2422, -0.1620], + [ 0.0569, -0.0185, -0.0260, ..., -0.2546, 0.0519, -0.1006], + [-0.0097, -0.0048, 0.2103, ..., -0.3132, -0.1352, -0.1605], + ..., + [-0.0228, 0.0062, -0.1849, ..., -0.0275, 0.1367, 0.0678], + [-0.0220, -0.0202, -0.1554, ..., -0.2768, 0.1090, -0.2030], + [ 0.0051, -0.0258, -0.0752, ..., 0.0400, -0.1274, 0.0174]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.1223e-09, + 8.8476e-09, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.4995e-08, + -3.7253e-09, 2.7381e-07], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 7.9162e-09, + 6.0536e-09, 2.4680e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.5832e-08, + -1.9558e-08, -5.4482e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.7229e-08, + 1.1828e-07, 4.2841e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 2.9095e-06, + 7.9162e-09, 7.9200e-06]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0295, 0.0162, 0.0195, 0.0084, 0.0336, -0.0110, 0.0010, -0.0150, + -0.0271, 0.0105], device='cuda:0'), grad: tensor([ 6.5658e-08, 4.5449e-07, 7.8697e-08, 1.1828e-07, -1.3612e-05, + 3.5437e-07, -1.0040e-06, -1.5879e-07, 6.8452e-07, 1.3016e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 220.64, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.5004 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.04 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.0006, 0.0110, 0.0369, ..., 0.0760, -0.2444, -0.1621], + [ 0.0569, -0.0185, -0.0265, ..., -0.2559, 0.0505, -0.1006], + [-0.0097, -0.0048, 0.2111, ..., -0.3133, -0.1359, -0.1608], + ..., + [-0.0228, 0.0062, -0.1854, ..., -0.0277, 0.1369, 0.0678], + [-0.0223, -0.0202, -0.1560, ..., -0.2773, 0.1121, -0.2035], + [ 0.0050, -0.0258, -0.0753, ..., 0.0395, -0.1281, 0.0167]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.8498e-07, ..., 3.3667e-07, + 5.0152e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 1.3970e-09, + 2.3283e-09, 1.7229e-08], + [ 0.0000e+00, 0.0000e+00, -8.8476e-09, ..., 4.6566e-10, + 5.1223e-09, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.1642e-08, + -5.5879e-08, -3.8650e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 5.1223e-09, + 9.3132e-10, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -5.2154e-08, + 2.0023e-08, -1.9465e-07]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0297, 0.0149, 0.0197, 0.0085, 0.0348, -0.0114, 0.0010, -0.0150, + -0.0243, 0.0100], device='cuda:0'), grad: tensor([ 2.2519e-06, 1.4435e-08, 6.9849e-09, 6.7987e-08, 2.3143e-07, + 7.4506e-09, -2.2985e-06, -6.2864e-08, 3.4925e-08, -2.3842e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 220.57, cls_loss 0.0013 cls_loss_mapping 0.0016 cls_loss_causal 0.5020 re_mapping 0.0044 re_causal 0.0130 /// teacc 98.99 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.0006, 0.0110, 0.0369, ..., 0.0760, -0.2463, -0.1624], + [ 0.0569, -0.0185, -0.0270, ..., -0.2584, 0.0506, -0.1006], + [-0.0097, -0.0048, 0.2122, ..., -0.3138, -0.1360, -0.1603], + ..., + [-0.0228, 0.0062, -0.1863, ..., -0.0291, 0.1364, 0.0676], + [-0.0224, -0.0202, -0.1566, ..., -0.2787, 0.1134, -0.2035], + [ 0.0051, -0.0258, -0.0758, ..., 0.0397, -0.1281, 0.0171]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-09, ..., -1.8161e-08, + 4.1910e-09, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 1.8207e-07, ..., 1.0384e-07, + -7.0315e-08, 2.4540e-07], + [ 0.0000e+00, 0.0000e+00, -8.2562e-07, ..., 3.2596e-09, + 1.8161e-08, -2.9895e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4075e-07, ..., 4.6100e-08, + 5.8673e-08, 3.8277e-07], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.1781e-07, + -1.9092e-08, 3.0128e-07], + [ 0.0000e+00, 0.0000e+00, 7.2643e-08, ..., 6.1514e-07, + 3.7253e-09, 1.4044e-06]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0301, 0.0149, 0.0203, 0.0085, 0.0347, -0.0118, 0.0015, -0.0154, + -0.0235, 0.0103], device='cuda:0'), grad: tensor([ 1.3970e-09, 5.8068e-07, -1.4193e-06, -4.5169e-07, -2.7176e-06, + 1.5320e-07, -2.7334e-07, 1.4231e-06, 5.2713e-07, 2.1979e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 220.76, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4869 re_mapping 0.0044 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.0006, 0.0110, 0.0369, ..., 0.0761, -0.2465, -0.1625], + [ 0.0569, -0.0185, -0.0271, ..., -0.2600, 0.0506, -0.1007], + [-0.0089, -0.0048, 0.2129, ..., -0.3142, -0.1362, -0.1597], + ..., + [-0.0229, 0.0062, -0.1872, ..., -0.0293, 0.1362, 0.0676], + [-0.0228, -0.0202, -0.1585, ..., -0.2794, 0.1137, -0.2031], + [ 0.0051, -0.0258, -0.0759, ..., 0.0397, -0.1284, 0.0172]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., -1.9558e-08, + 1.8626e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.1735e-07, ..., 1.8626e-09, + 1.4901e-07, 4.3772e-08], + [ 0.0000e+00, 0.0000e+00, -2.5425e-07, ..., 9.3132e-10, + 1.0431e-07, -9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 5.0943e-07, ..., 7.4506e-09, + 4.2059e-06, 9.1176e-07], + [ 0.0000e+00, 0.0000e+00, 9.0338e-08, ..., 9.3132e-09, + 7.4506e-08, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., -1.6205e-07, + 9.3132e-08, -4.0326e-07]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0301, 0.0150, 0.0207, 0.0089, 0.0347, -0.0125, 0.0015, -0.0157, + -0.0235, 0.0102], device='cuda:0'), grad: tensor([ 5.1223e-08, 1.1679e-06, 3.2410e-07, -3.7313e-05, 6.2864e-07, + 5.7667e-06, 8.2888e-08, 2.8670e-05, 7.1246e-07, -1.2200e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 221.03, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4533 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.03 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.0007, 0.0110, 0.0369, ..., 0.0761, -0.2470, -0.1625], + [ 0.0568, -0.0185, -0.0271, ..., -0.2603, 0.0508, -0.1007], + [-0.0089, -0.0048, 0.2131, ..., -0.3143, -0.1364, -0.1599], + ..., + [-0.0229, 0.0062, -0.1874, ..., -0.0296, 0.1363, 0.0677], + [-0.0228, -0.0202, -0.1601, ..., -0.2803, 0.1134, -0.2039], + [ 0.0056, -0.0258, -0.0759, ..., 0.0400, -0.1287, 0.0177]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1828e-07, ..., -1.3039e-07, + 3.7253e-08, 6.0536e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 6.5193e-09, + -3.5614e-05, -1.6481e-05], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 1.0245e-08, + 2.8871e-08, 1.1176e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.6566e-09, + 3.2604e-05, 1.5117e-05], + [ 0.0000e+00, 0.0000e+00, 4.5635e-08, ..., 8.3819e-09, + -2.8871e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-08, ..., 2.3283e-08, + 2.2389e-06, 1.1176e-06]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0301, 0.0150, 0.0206, 0.0089, 0.0339, -0.0128, 0.0020, -0.0156, + -0.0239, 0.0105], device='cuda:0'), grad: tensor([-1.7323e-07, -7.5400e-05, 1.6578e-07, 1.2666e-06, 5.4017e-07, + -2.1532e-06, 1.6419e-06, 6.8724e-05, 3.4925e-07, 5.1036e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 220.73, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4753 re_mapping 0.0045 re_causal 0.0132 /// teacc 98.99 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.0007, 0.0110, 0.0370, ..., 0.0762, -0.2471, -0.1626], + [ 0.0568, -0.0185, -0.0277, ..., -0.2627, 0.0539, -0.0991], + [-0.0090, -0.0048, 0.2135, ..., -0.3146, -0.1362, -0.1600], + ..., + [-0.0229, 0.0062, -0.1875, ..., -0.0294, 0.1333, 0.0663], + [-0.0229, -0.0202, -0.1612, ..., -0.2815, 0.1135, -0.2043], + [ 0.0056, -0.0258, -0.0760, ..., 0.0401, -0.1297, 0.0179]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 1.8626e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.4028e-07, ..., 9.3132e-10, + -8.3819e-09, 4.4703e-08], + [ 0.0000e+00, 0.0000e+00, -4.2841e-06, ..., 9.3132e-10, + 1.2107e-08, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 3.7253e-09, + -5.2154e-08, -1.3504e-07], + [ 0.0000e+00, 0.0000e+00, 3.9451e-06, ..., 2.7940e-09, + 2.5146e-08, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., -8.3819e-09, + 2.7940e-08, 4.6566e-08]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0300, 0.0178, 0.0209, 0.0088, 0.0339, -0.0131, 0.0021, -0.0181, + -0.0240, 0.0107], device='cuda:0'), grad: tensor([ 6.6124e-08, 6.9384e-07, -1.1094e-05, -6.5193e-08, 8.5682e-08, + -3.0641e-07, 1.4342e-07, -2.7474e-07, 1.0528e-05, 2.3097e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 220.19, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.5017 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.04 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.0007, 0.0110, 0.0372, ..., 0.0762, -0.2475, -0.1630], + [ 0.0568, -0.0185, -0.0278, ..., -0.2660, 0.0539, -0.0990], + [-0.0090, -0.0048, 0.2137, ..., -0.3150, -0.1365, -0.1600], + ..., + [-0.0229, 0.0062, -0.1877, ..., -0.0303, 0.1334, 0.0661], + [-0.0230, -0.0202, -0.1633, ..., -0.2834, 0.1134, -0.2064], + [ 0.0056, -0.0258, -0.0762, ..., 0.0409, -0.1299, 0.0189]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3225e-07, ..., -1.6298e-07, + 0.0000e+00, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 4.0978e-08, ..., 6.8918e-08, + -7.7300e-08, 1.3784e-07], + [ 0.0000e+00, 0.0000e+00, 1.0338e-07, ..., 2.2352e-08, + 1.5832e-08, 2.1420e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 5.7742e-08, ..., 7.7300e-08, + -1.8626e-09, 9.9652e-08], + [ 0.0000e+00, 0.0000e+00, 1.3225e-07, ..., 1.4137e-06, + 2.7940e-09, 2.5406e-06], + [ 0.0000e+00, 0.0000e+00, 4.1910e-08, ..., -4.4815e-06, + -1.4249e-07, -8.5607e-06]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0299, 0.0179, 0.0207, 0.0088, 0.0329, -0.0134, 0.0020, -0.0182, + -0.0243, 0.0115], device='cuda:0'), grad: tensor([-3.5018e-07, 2.4587e-07, 4.0233e-07, 2.8722e-06, 1.1384e-05, + -5.7481e-06, 1.1595e-06, 3.9488e-07, 6.3889e-06, -1.6809e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 220.98, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4695 re_mapping 0.0044 re_causal 0.0122 /// teacc 99.01 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.0007, 0.0110, 0.0372, ..., 0.0762, -0.2520, -0.1631], + [ 0.0568, -0.0185, -0.0283, ..., -0.2683, 0.0538, -0.0991], + [-0.0090, -0.0048, 0.2147, ..., -0.3158, -0.1365, -0.1605], + ..., + [-0.0229, 0.0062, -0.1886, ..., -0.0309, 0.1336, 0.0662], + [-0.0231, -0.0202, -0.1649, ..., -0.2849, 0.1135, -0.2074], + [ 0.0055, -0.0258, -0.0763, ..., 0.0412, -0.1303, 0.0195]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.0047e-08, ..., -5.6811e-08, + 8.3819e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 4.6566e-09, + 8.2888e-08, 1.0245e-07], + [ 0.0000e+00, 0.0000e+00, 2.8312e-07, ..., 3.7253e-09, + 5.6550e-06, 4.5449e-06], + ..., + [ 0.0000e+00, 0.0000e+00, -1.9092e-07, ..., 1.8626e-09, + -5.5395e-06, -4.6901e-06], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 2.7940e-09, + -1.0543e-06, -2.5053e-07], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 2.7940e-08, + 2.4866e-07, 1.9465e-07]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0305, 0.0177, 0.0212, 0.0087, 0.0326, -0.0146, 0.0032, -0.0180, + -0.0245, 0.0119], device='cuda:0'), grad: tensor([-7.3574e-08, 4.9733e-07, 2.0936e-05, -2.8312e-07, 3.8650e-07, + 4.4797e-07, 6.4913e-07, -2.0206e-05, -3.5912e-06, 1.2117e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 220.27, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.4851 re_mapping 0.0045 re_causal 0.0130 /// teacc 99.01 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.0007, 0.0110, 0.0373, ..., 0.0758, -0.2522, -0.1641], + [ 0.0568, -0.0185, -0.0286, ..., -0.2703, 0.0540, -0.0991], + [-0.0090, -0.0048, 0.2153, ..., -0.3163, -0.1368, -0.1609], + ..., + [-0.0229, 0.0062, -0.1894, ..., -0.0313, 0.1337, 0.0660], + [-0.0231, -0.0202, -0.1652, ..., -0.2859, 0.1132, -0.2092], + [ 0.0055, -0.0258, -0.0764, ..., 0.0413, -0.1306, 0.0194]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., -1.3970e-08, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + 2.0955e-07, 2.0768e-07], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, + 2.7940e-09, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.4680e-07, -2.5146e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.1176e-08, + 2.9802e-08, 4.8429e-08]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0307, 0.0178, 0.0213, 0.0090, 0.0332, -0.0158, 0.0024, -0.0180, + -0.0249, 0.0121], device='cuda:0'), grad: tensor([-2.5146e-08, 4.9360e-07, 1.5832e-08, -1.3039e-08, -1.4901e-08, + 2.8871e-08, -2.4214e-08, -5.7090e-07, 1.3039e-08, 1.0617e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 220.64, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4891 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.02 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.0007, 0.0110, 0.0373, ..., 0.0759, -0.2523, -0.1641], + [ 0.0568, -0.0185, -0.0287, ..., -0.2719, 0.0536, -0.0995], + [-0.0090, -0.0048, 0.2156, ..., -0.3166, -0.1370, -0.1613], + ..., + [-0.0229, 0.0062, -0.1896, ..., -0.0315, 0.1340, 0.0665], + [-0.0231, -0.0202, -0.1655, ..., -0.2863, 0.1134, -0.2097], + [ 0.0055, -0.0258, -0.0765, ..., 0.0412, -0.1309, 0.0194]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.7987e-08, ..., 4.4703e-08, + 2.3935e-07, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 1.4901e-08, + 2.9709e-07, 6.0629e-07], + [ 0.0000e+00, 0.0000e+00, -6.5565e-07, ..., -2.1327e-07, + 1.0710e-07, -5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.4424e-07, ..., 1.4529e-07, + -1.0142e-06, -9.2201e-07], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., -1.9558e-08, + 1.2107e-08, 1.9372e-07], + [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., 1.4901e-08, + 1.0524e-07, 5.0291e-08]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0307, 0.0176, 0.0213, 0.0090, 0.0334, -0.0159, 0.0024, -0.0178, + -0.0249, 0.0120], device='cuda:0'), grad: tensor([ 1.2601e-06, 2.7660e-07, -7.3574e-07, -4.4703e-08, 6.4261e-08, + 1.3039e-07, 3.9954e-07, -1.1837e-06, -3.8743e-07, 2.2259e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 220.68, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.4648 re_mapping 0.0043 re_causal 0.0127 /// teacc 98.98 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.0007, 0.0110, 0.0373, ..., 0.0762, -0.2527, -0.1643], + [ 0.0568, -0.0185, -0.0287, ..., -0.2769, 0.0536, -0.0999], + [-0.0090, -0.0048, 0.2162, ..., -0.3170, -0.1377, -0.1611], + ..., + [-0.0229, 0.0062, -0.1899, ..., -0.0324, 0.1340, 0.0658], + [-0.0231, -0.0202, -0.1667, ..., -0.2886, 0.1138, -0.2110], + [ 0.0055, -0.0258, -0.0770, ..., 0.0442, -0.1291, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5358e-06, ..., -2.9802e-06, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.4657e-07, ..., 1.6205e-06, + 3.6925e-05, 2.8759e-06], + [ 0.0000e+00, 0.0000e+00, 1.2759e-07, ..., 2.5053e-07, + 7.3574e-08, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 4.0978e-08, + -3.9726e-05, -3.0883e-06], + [ 0.0000e+00, 0.0000e+00, 4.7591e-07, ..., 9.0897e-07, + 1.5842e-06, 1.2573e-07], + [ 0.0000e+00, 0.0000e+00, 4.4703e-08, ..., 1.1735e-07, + 1.4342e-07, 1.2107e-08]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0314, 0.0175, 0.0215, 0.0090, 0.0297, -0.0159, 0.0032, -0.0180, + -0.0249, 0.0157], device='cuda:0'), grad: tensor([-7.3910e-06, 6.9976e-05, 8.0187e-07, 1.7118e-06, 2.2352e-08, + 8.2888e-08, 6.8918e-08, -7.0751e-05, 5.0589e-06, 5.2154e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 221.15, cls_loss 0.0011 cls_loss_mapping 0.0023 cls_loss_causal 0.5154 re_mapping 0.0043 re_causal 0.0132 /// teacc 99.00 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.0007, 0.0110, 0.0372, ..., 0.0765, -0.2530, -0.1645], + [ 0.0568, -0.0185, -0.0289, ..., -0.2786, 0.0524, -0.1016], + [-0.0090, -0.0048, 0.2164, ..., -0.3179, -0.1379, -0.1615], + ..., + [-0.0229, 0.0062, -0.1901, ..., -0.0329, 0.1353, 0.0672], + [-0.0231, -0.0202, -0.1678, ..., -0.2909, 0.1137, -0.2117], + [ 0.0055, -0.0258, -0.0773, ..., 0.0444, -0.1300, 0.0244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, + 2.1420e-08, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.0245e-08, + 3.2317e-07, 2.7847e-07], + [ 0.0000e+00, 0.0000e+00, -2.7008e-08, ..., 9.3132e-10, + 8.3819e-09, 1.1176e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + -4.6752e-07, -4.0233e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.0489e-08, + 1.5832e-08, 4.7497e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.8476e-08, + 7.7300e-08, 2.1979e-07]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0314, 0.0164, 0.0213, 0.0101, 0.0298, -0.0185, 0.0029, -0.0170, + -0.0252, 0.0157], device='cuda:0'), grad: tensor([ 3.0920e-07, 6.2026e-07, 8.3819e-09, 1.5367e-07, 1.4342e-07, + 3.1665e-08, -8.8383e-07, -1.0300e-06, 1.6578e-07, 4.6939e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 220.74, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.4840 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.01 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.0007, 0.0110, 0.0373, ..., 0.0767, -0.2532, -0.1646], + [ 0.0568, -0.0185, -0.0290, ..., -0.2795, 0.0524, -0.1016], + [-0.0090, -0.0048, 0.2176, ..., -0.3182, -0.1380, -0.1607], + ..., + [-0.0229, 0.0062, -0.1916, ..., -0.0343, 0.1354, 0.0673], + [-0.0231, -0.0202, -0.1687, ..., -0.2911, 0.1136, -0.2126], + [ 0.0055, -0.0258, -0.0774, ..., 0.0439, -0.1319, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -1.3970e-08, + 1.8626e-09, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., 2.2352e-08, + 1.5832e-08, 6.7055e-08], + [ 0.0000e+00, 0.0000e+00, -4.1816e-07, ..., 9.3132e-10, + 7.4506e-09, -1.6764e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 3.2689e-07, ..., 1.3970e-08, + -8.7544e-08, 4.5635e-08], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 9.3132e-10, + -6.8918e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 8.2888e-08, + 4.3772e-08, 1.4249e-07]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0313, 0.0165, 0.0218, 0.0102, 0.0305, -0.0187, 0.0027, -0.0170, + -0.0252, 0.0150], device='cuda:0'), grad: tensor([-1.8626e-09, 1.6205e-07, -5.6904e-07, 1.1362e-07, -1.8906e-07, + -1.0338e-07, 8.5682e-08, 2.2352e-07, -6.6124e-08, 3.5483e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 220.47, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4853 re_mapping 0.0042 re_causal 0.0122 /// teacc 99.00 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.0007, 0.0110, 0.0373, ..., 0.0767, -0.2534, -0.1647], + [ 0.0568, -0.0185, -0.0292, ..., -0.2812, 0.0524, -0.1017], + [-0.0090, -0.0048, 0.2187, ..., -0.3184, -0.1390, -0.1603], + ..., + [-0.0229, 0.0062, -0.1926, ..., -0.0347, 0.1355, 0.0673], + [-0.0231, -0.0202, -0.1692, ..., -0.2899, 0.1138, -0.2125], + [ 0.0055, -0.0258, -0.0775, ..., 0.0436, -0.1320, 0.0236]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., -1.6764e-08, + 3.7532e-07, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.7008e-08, ..., 1.9558e-08, + -1.3947e-05, 4.0978e-08], + [ 0.0000e+00, 0.0000e+00, 3.8557e-07, ..., 1.8626e-09, + 5.0440e-06, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.0338e-08, ..., 1.3039e-08, + 5.3085e-06, 2.4214e-08], + [ 0.0000e+00, 0.0000e+00, 1.8999e-07, ..., 2.7940e-09, + 1.6391e-06, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 2.7660e-07, + 1.0151e-06, 3.8464e-07]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0313, 0.0164, 0.0223, 0.0097, 0.0309, -0.0176, 0.0023, -0.0170, + -0.0251, 0.0149], device='cuda:0'), grad: tensor([ 8.5216e-07, -2.9817e-05, 1.1660e-05, -1.4259e-06, -3.9395e-07, + 3.2224e-07, 6.5286e-07, 1.1697e-05, 4.0531e-06, 2.3842e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 220.50, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4992 re_mapping 0.0044 re_causal 0.0128 /// teacc 98.90 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.0007, 0.0110, 0.0374, ..., 0.0769, -0.2537, -0.1647], + [ 0.0568, -0.0185, -0.0298, ..., -0.2825, 0.0524, -0.1018], + [-0.0090, -0.0048, 0.2190, ..., -0.3189, -0.1383, -0.1603], + ..., + [-0.0229, 0.0062, -0.1935, ..., -0.0351, 0.1356, 0.0674], + [-0.0231, -0.0202, -0.1670, ..., -0.2902, 0.1142, -0.2128], + [ 0.0055, -0.0258, -0.0775, ..., 0.0434, -0.1322, 0.0235]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 1.6764e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, + 3.1944e-07, 3.4086e-07], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 1.8626e-09, + 1.1809e-06, 1.2098e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + -1.7090e-06, -1.7574e-06], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + -2.4214e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 2.1141e-07, ..., 1.8533e-07, + 9.4995e-08, 6.2212e-07]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0312, 0.0164, 0.0224, 0.0096, 0.0313, -0.0177, 0.0020, -0.0170, + -0.0245, 0.0147], device='cuda:0'), grad: tensor([ 7.4506e-08, 1.6773e-06, 6.2920e-06, 8.4043e-05, -3.4645e-07, + -8.4519e-05, -9.0338e-08, -8.8885e-06, -9.6858e-08, 1.9819e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 220.86, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4721 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.06 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.0007, 0.0110, 0.0380, ..., 0.0782, -0.2539, -0.1648], + [ 0.0568, -0.0185, -0.0276, ..., -0.2837, 0.0527, -0.1019], + [-0.0090, -0.0048, 0.2178, ..., -0.3209, -0.1422, -0.1606], + ..., + [-0.0229, 0.0062, -0.1939, ..., -0.0355, 0.1357, 0.0677], + [-0.0231, -0.0202, -0.1674, ..., -0.2903, 0.1143, -0.2142], + [ 0.0055, -0.0258, -0.0779, ..., 0.0432, -0.1329, 0.0230]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 8.9407e-08, + 1.8626e-09, 6.7055e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-08, + 2.8871e-08, 1.0338e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3039e-08, + 2.7008e-08, 5.3085e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-08, + -1.5739e-07, -1.9837e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 8.3819e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 4.3698e-06, + 8.2888e-08, 2.7940e-06]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0300, 0.0164, 0.0198, 0.0096, 0.0316, -0.0173, 0.0004, -0.0167, + -0.0245, 0.0141], device='cuda:0'), grad: tensor([ 1.2387e-07, 2.1793e-07, 1.3970e-07, 5.2154e-08, -7.4580e-06, + 1.3970e-08, 1.0943e-06, -5.4110e-07, 8.7544e-08, 6.2510e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 220.32, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4787 re_mapping 0.0044 re_causal 0.0131 /// teacc 99.02 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.0007, 0.0110, 0.0381, ..., 0.0784, -0.2540, -0.1650], + [ 0.0568, -0.0185, -0.0276, ..., -0.2840, 0.0529, -0.1017], + [-0.0090, -0.0048, 0.2179, ..., -0.3211, -0.1428, -0.1612], + ..., + [-0.0229, 0.0062, -0.1941, ..., -0.0356, 0.1356, 0.0676], + [-0.0231, -0.0202, -0.1678, ..., -0.2907, 0.1147, -0.2145], + [ 0.0055, -0.0258, -0.0781, ..., 0.0431, -0.1336, 0.0229]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 9.3132e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, + 7.4506e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.8626e-09, + 1.8626e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + -2.7753e-07, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -1.8626e-09, + -1.8626e-09, -5.2154e-08]], device='cuda:0') +Epoch 312, bias, value: tensor([-2.9816e-02, 1.6577e-02, 1.9427e-02, 9.5639e-03, 3.1673e-02, + -1.7365e-02, -5.0478e-05, -1.6780e-02, -2.4282e-02, 1.3892e-02], + device='cuda:0'), grad: tensor([ 3.5390e-08, 3.9116e-08, 1.5832e-08, 1.1176e-08, 4.0047e-08, + 6.4168e-07, -2.3749e-07, 2.8871e-08, -5.1595e-07, -4.2841e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 220.52, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4619 re_mapping 0.0040 re_causal 0.0124 /// teacc 99.01 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.0007, 0.0110, 0.0382, ..., 0.0785, -0.2541, -0.1651], + [ 0.0568, -0.0185, -0.0277, ..., -0.2845, 0.0529, -0.1017], + [-0.0090, -0.0048, 0.2183, ..., -0.3214, -0.1435, -0.1612], + ..., + [-0.0229, 0.0062, -0.1945, ..., -0.0359, 0.1356, 0.0677], + [-0.0232, -0.0202, -0.1684, ..., -0.2914, 0.1148, -0.2154], + [ 0.0055, -0.0258, -0.0784, ..., 0.0431, -0.1340, 0.0229]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 1.8654e-06, 1.9558e-06], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.4901e-08, 9.3132e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + -3.8482e-06, -4.0457e-06], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + -3.8091e-07, -6.7987e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-08, + 1.3728e-06, 1.1958e-06]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0298, 0.0166, 0.0194, 0.0095, 0.0318, -0.0168, -0.0002, -0.0167, + -0.0244, 0.0138], device='cuda:0'), grad: tensor([ 1.3970e-08, 5.3681e-06, 6.2399e-08, 3.9004e-06, 1.6764e-08, + -1.3653e-06, 1.0245e-08, -1.0960e-05, -1.2778e-06, 4.2170e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 220.63, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4757 re_mapping 0.0042 re_causal 0.0123 /// teacc 98.95 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.0007, 0.0110, 0.0383, ..., 0.0787, -0.2543, -0.1652], + [ 0.0568, -0.0185, -0.0279, ..., -0.2855, 0.0533, -0.1006], + [-0.0090, -0.0048, 0.2228, ..., -0.3239, -0.1405, -0.1613], + ..., + [-0.0229, 0.0062, -0.1995, ..., -0.0363, 0.1350, 0.0672], + [-0.0232, -0.0202, -0.1690, ..., -0.2919, 0.1154, -0.2131], + [ 0.0055, -0.0258, -0.0789, ..., 0.0428, -0.1379, 0.0218]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -2.0489e-08, + 9.3132e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 1.2107e-08, + -4.8429e-08, 3.3528e-08], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 1.8626e-09, + 4.6566e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 8.3819e-09, + -2.3283e-08, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 4.6566e-09, + 7.6368e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 4.2841e-08, + 9.3132e-09, 5.0291e-08]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0297, 0.0172, 0.0240, 0.0094, 0.0324, -0.0163, -0.0008, -0.0182, + -0.0242, 0.0122], device='cuda:0'), grad: tensor([ 6.5193e-09, -2.1420e-08, 1.4901e-08, -8.5682e-08, -1.9092e-07, + 3.9116e-08, -6.5193e-08, -1.7695e-08, 1.7229e-07, 1.2945e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 220.72, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4894 re_mapping 0.0042 re_causal 0.0128 /// teacc 98.96 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.0007, 0.0110, 0.0383, ..., 0.0788, -0.2546, -0.1653], + [ 0.0568, -0.0185, -0.0263, ..., -0.2864, 0.0537, -0.1005], + [-0.0090, -0.0048, 0.2224, ..., -0.3244, -0.1423, -0.1630], + ..., + [-0.0229, 0.0062, -0.1995, ..., -0.0365, 0.1351, 0.0675], + [-0.0232, -0.0202, -0.1695, ..., -0.2927, 0.1171, -0.2137], + [ 0.0055, -0.0258, -0.0791, ..., 0.0428, -0.1384, 0.0219]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 9.3132e-09, + 9.3132e-10, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.6764e-08, + -1.1176e-08, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-07, ..., 9.3132e-10, + 4.6566e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 4.2841e-08, + -3.7253e-09, 4.4703e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 1.2107e-08, + -8.3819e-09, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 3.3900e-07, + 5.5879e-09, 4.4890e-07]], device='cuda:0') +Epoch 315, bias, value: tensor([-2.9754e-02, 1.7474e-02, 2.3243e-02, 8.4874e-03, 3.2471e-02, + -1.6712e-02, -5.4720e-05, -1.8074e-02, -2.3015e-02, 1.2154e-02], + device='cuda:0'), grad: tensor([ 1.1642e-07, 2.1420e-08, 2.8126e-07, -4.2561e-07, -3.2596e-07, + 5.2154e-08, -6.0443e-07, 9.7789e-08, 4.4703e-08, 7.5437e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 221.07, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4916 re_mapping 0.0041 re_causal 0.0129 /// teacc 99.00 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.0007, 0.0110, 0.0384, ..., 0.0788, -0.2551, -0.1654], + [ 0.0568, -0.0185, -0.0261, ..., -0.2871, 0.0538, -0.1006], + [-0.0090, -0.0048, 0.2224, ..., -0.3255, -0.1427, -0.1633], + ..., + [-0.0229, 0.0062, -0.1997, ..., -0.0366, 0.1351, 0.0676], + [-0.0232, -0.0202, -0.1698, ..., -0.2931, 0.1178, -0.2134], + [ 0.0055, -0.0258, -0.0792, ..., 0.0428, -0.1391, 0.0220]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0245e-08, ..., -1.5832e-08, + 2.0750e-06, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 9.3132e-10, + -6.5863e-06, 8.3819e-08], + [ 0.0000e+00, 0.0000e+00, -4.2096e-07, ..., 1.8626e-09, + 4.1910e-08, 2.1420e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.4645e-07, ..., 1.8626e-09, + -2.3749e-07, -2.3209e-06], + [ 0.0000e+00, 0.0000e+00, 9.2201e-07, ..., 2.7940e-09, + 3.6620e-06, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 9.3132e-10, + 3.8464e-07, 1.1232e-06]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0299, 0.0174, 0.0229, 0.0084, 0.0323, -0.0164, 0.0002, -0.0180, + -0.0227, 0.0121], device='cuda:0'), grad: tensor([ 5.3048e-06, -1.6809e-05, -6.6869e-07, -3.4049e-06, 1.2247e-06, + 2.3842e-07, 5.3644e-07, -8.2422e-07, 1.2934e-05, 1.4389e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 220.68, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4942 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.00 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.0007, 0.0110, 0.0386, ..., 0.0791, -0.2556, -0.1655], + [ 0.0568, -0.0185, -0.0265, ..., -0.2880, 0.0538, -0.1007], + [-0.0090, -0.0048, 0.2226, ..., -0.3276, -0.1431, -0.1632], + ..., + [-0.0229, 0.0062, -0.1998, ..., -0.0369, 0.1352, 0.0679], + [-0.0233, -0.0202, -0.1712, ..., -0.2939, 0.1182, -0.2129], + [ 0.0055, -0.0258, -0.0796, ..., 0.0429, -0.1406, 0.0220]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -8.3819e-09, + 4.6566e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., -2.7940e-09, + -2.3097e-07, -1.4342e-07], + [ 0.0000e+00, 0.0000e+00, -9.8720e-08, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 9.3132e-10, + -9.3132e-10, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 1.1176e-07, ..., 9.3132e-10, + 4.6566e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-09, + 4.6566e-09, -1.1176e-08]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0297, 0.0172, 0.0223, 0.0090, 0.0322, -0.0180, 0.0003, -0.0176, + -0.0225, 0.0118], device='cuda:0'), grad: tensor([ 2.7940e-09, -6.1560e-07, -2.1234e-07, -9.2201e-08, 5.0105e-07, + -6.0722e-07, 1.1642e-07, 3.9116e-08, 8.4378e-07, 3.6322e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 220.66, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4825 re_mapping 0.0040 re_causal 0.0125 /// teacc 98.91 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.0007, 0.0110, 0.0389, ..., 0.0795, -0.2555, -0.1657], + [ 0.0568, -0.0185, -0.0267, ..., -0.2905, 0.0539, -0.1008], + [-0.0090, -0.0048, 0.2226, ..., -0.3281, -0.1435, -0.1635], + ..., + [-0.0229, 0.0062, -0.1999, ..., -0.0371, 0.1353, 0.0681], + [-0.0233, -0.0202, -0.1719, ..., -0.2947, 0.1178, -0.2132], + [ 0.0055, -0.0258, -0.0798, ..., 0.0424, -0.1410, 0.0217]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6671e-07, ..., -1.1828e-07, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 1.3039e-08, + 2.3283e-08, 2.5146e-08], + [ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 7.0781e-08, + 1.3318e-07, 1.8626e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3597e-07, ..., 2.7940e-09, + -2.2259e-07, -7.9162e-08], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 3.7253e-09, + -3.3528e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 5.6811e-08, + 1.2107e-08, 3.4459e-08]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0292, 0.0173, 0.0221, 0.0089, 0.0326, -0.0176, 0.0002, -0.0175, + -0.0232, 0.0113], device='cuda:0'), grad: tensor([-3.5390e-07, 1.2293e-07, 7.7579e-07, 2.9802e-08, -8.2888e-08, + 6.4261e-08, 7.1712e-08, -3.4645e-07, -4.5076e-07, 1.6671e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 220.79, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4705 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.00 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.0007, 0.0110, 0.0391, ..., 0.0797, -0.2553, -0.1658], + [ 0.0568, -0.0185, -0.0270, ..., -0.2921, 0.0540, -0.1009], + [-0.0090, -0.0048, 0.2206, ..., -0.3314, -0.1436, -0.1661], + ..., + [-0.0229, 0.0062, -0.2000, ..., -0.0385, 0.1353, 0.0682], + [-0.0233, -0.0202, -0.1723, ..., -0.2960, 0.1175, -0.2136], + [ 0.0046, -0.0258, -0.0801, ..., 0.0422, -0.1413, 0.0212]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.4890e-07, ..., -2.0489e-07, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.1176e-08, + 4.6566e-09, 2.1420e-08], + [ 0.0000e+00, 0.0000e+00, -5.4948e-08, ..., 1.8626e-09, + 1.8626e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 4.6566e-09, + -1.3039e-08, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 5.2154e-08, ..., 3.7253e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 1.4901e-08, + 0.0000e+00, 1.1176e-08]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0289, 0.0172, 0.0200, 0.0088, 0.0355, -0.0165, -0.0002, -0.0176, + -0.0235, 0.0108], device='cuda:0'), grad: tensor([-3.1013e-06, 8.6613e-08, -7.7300e-08, 2.7101e-07, -4.3772e-08, + 9.4995e-07, 1.2899e-06, 1.7602e-07, 2.7753e-07, 1.8254e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 221.23, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.4409 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.05 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.0007, 0.0110, 0.0394, ..., 0.0801, -0.2556, -0.1659], + [ 0.0568, -0.0185, -0.0270, ..., -0.2924, 0.0538, -0.1011], + [-0.0090, -0.0048, 0.2205, ..., -0.3316, -0.1438, -0.1668], + ..., + [-0.0229, 0.0062, -0.2001, ..., -0.0386, 0.1359, 0.0687], + [-0.0233, -0.0202, -0.1731, ..., -0.2965, 0.1174, -0.2135], + [ 0.0046, -0.0258, -0.0808, ..., 0.0418, -0.1427, 0.0209]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6019e-07, ..., -2.6263e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 4.6566e-09, + 4.6566e-08, 3.8184e-08], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.9558e-08, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + -6.4261e-08, -4.3772e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + 1.8626e-09, 3.2596e-08], + [ 0.0000e+00, 0.0000e+00, 1.3877e-07, ..., 2.1141e-07, + 2.7940e-09, -4.0047e-08]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0285, 0.0171, 0.0197, 0.0088, 0.0358, -0.0166, -0.0003, -0.0172, + -0.0236, 0.0102], device='cuda:0'), grad: tensor([-4.3306e-07, 1.1735e-07, 6.9849e-08, -1.1455e-07, 2.7008e-08, + 6.7055e-08, 1.3970e-08, -1.1921e-07, 6.3330e-08, 3.1572e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 220.97, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4754 re_mapping 0.0041 re_causal 0.0123 /// teacc 99.02 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.0007, 0.0110, 0.0396, ..., 0.0805, -0.2556, -0.1662], + [ 0.0568, -0.0185, -0.0271, ..., -0.2928, 0.0538, -0.1012], + [-0.0090, -0.0048, 0.2201, ..., -0.3316, -0.1459, -0.1698], + ..., + [-0.0229, 0.0062, -0.1994, ..., -0.0388, 0.1366, 0.0698], + [-0.0233, -0.0202, -0.1740, ..., -0.2968, 0.1171, -0.2145], + [ 0.0046, -0.0258, -0.0815, ..., 0.0419, -0.1433, 0.0211]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.7253e-09, + 2.1420e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 1.0617e-07, 7.7300e-08], + [ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., 0.0000e+00, + 1.4529e-07, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + -8.5682e-08, -7.8231e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, + 3.3528e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -3.7253e-09, + 3.7253e-09, -3.0734e-08]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0282, 0.0167, 0.0187, 0.0086, 0.0356, -0.0166, -0.0010, -0.0162, + -0.0242, 0.0101], device='cuda:0'), grad: tensor([ 1.0524e-07, 4.5635e-07, 8.8196e-07, -1.2387e-07, 3.2783e-07, + 1.2666e-07, -1.7025e-06, -2.1607e-07, 1.6019e-07, -1.6764e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 320---------------------------------------------------- +epoch 320, time 221.68, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4707 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.18 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.0007, 0.0110, 0.0397, ..., 0.0808, -0.2559, -0.1665], + [ 0.0568, -0.0185, -0.0272, ..., -0.2938, 0.0535, -0.1013], + [-0.0091, -0.0048, 0.2206, ..., -0.3316, -0.1460, -0.1690], + ..., + [-0.0229, 0.0062, -0.1996, ..., -0.0389, 0.1367, 0.0699], + [-0.0233, -0.0202, -0.1747, ..., -0.2973, 0.1177, -0.2144], + [ 0.0046, -0.0258, -0.0830, ..., 0.0422, -0.1436, 0.0212]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -4.6566e-09, + 7.4506e-09, 0.0000e+00], + [-5.5879e-09, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, + -3.1665e-08, 6.6124e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.0245e-08, 1.8626e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 3.7253e-09, + -1.2852e-07, -1.7509e-07], + [ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 2.5146e-08, + -1.7695e-08, 1.9092e-07], + [ 9.3132e-10, 0.0000e+00, 9.3132e-09, ..., 2.8871e-08, + 3.2596e-08, -8.6613e-08]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0279, 0.0163, 0.0187, 0.0086, 0.0354, -0.0162, -0.0004, -0.0160, + -0.0240, 0.0099], device='cuda:0'), grad: tensor([ 2.5146e-08, -5.5879e-08, 5.2154e-08, 2.3004e-07, -6.2399e-08, + 8.5682e-08, 3.5390e-08, -5.2247e-07, 2.2352e-07, -2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 220.83, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4764 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.02 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.0008, 0.0110, 0.0398, ..., 0.0811, -0.2562, -0.1665], + [ 0.0568, -0.0185, -0.0273, ..., -0.2948, 0.0534, -0.1016], + [-0.0093, -0.0048, 0.2212, ..., -0.3317, -0.1461, -0.1684], + ..., + [-0.0229, 0.0062, -0.2003, ..., -0.0391, 0.1367, 0.0701], + [-0.0238, -0.0202, -0.1749, ..., -0.2977, 0.1185, -0.2147], + [ 0.0046, -0.0258, -0.0841, ..., 0.0422, -0.1446, 0.0212]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0245e-08, ..., -1.5832e-08, + 1.1176e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -9.3132e-09, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 2.7940e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + -8.3819e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 2.7940e-09, + -9.3132e-10, -2.7008e-08]], device='cuda:0') +Epoch 323, bias, value: tensor([-2.7693e-02, 1.6133e-02, 1.9009e-02, 8.1631e-03, 3.5314e-02, + -1.5647e-02, 4.0765e-05, -1.5918e-02, -2.3684e-02, 9.6611e-03], + device='cuda:0'), grad: tensor([ 3.4459e-08, -9.3132e-09, 4.0978e-08, -7.1712e-08, 5.1223e-08, + -1.9651e-07, -9.3132e-10, 4.2841e-08, 6.7987e-08, 4.4703e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 221.11, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4715 re_mapping 0.0040 re_causal 0.0121 /// teacc 98.99 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.0008, 0.0110, 0.0400, ..., 0.0816, -0.2567, -0.1665], + [ 0.0567, -0.0185, -0.0276, ..., -0.2958, 0.0534, -0.1016], + [-0.0093, -0.0048, 0.2213, ..., -0.3317, -0.1462, -0.1685], + ..., + [-0.0229, 0.0062, -0.2004, ..., -0.0396, 0.1369, 0.0702], + [-0.0238, -0.0202, -0.1754, ..., -0.2985, 0.1184, -0.2149], + [ 0.0048, -0.0258, -0.0845, ..., 0.0417, -0.1450, 0.0207]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., -2.0489e-08, + 0.0000e+00, 9.3132e-10], + [ 9.3132e-10, 0.0000e+00, 2.8871e-08, ..., 9.3132e-10, + 1.8720e-07, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, -9.2201e-08, ..., 5.5879e-09, + 4.1910e-08, 3.7253e-09], + ..., + [ 1.1176e-08, 0.0000e+00, 3.5390e-08, ..., 2.7940e-09, + -3.2876e-07, -5.5879e-09], + [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., 2.7940e-09, + 1.8626e-08, 7.3574e-08], + [ 2.7940e-09, 0.0000e+00, 1.1176e-08, ..., 2.7940e-09, + 7.4506e-09, -9.4995e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0275, 0.0160, 0.0185, 0.0083, 0.0357, -0.0153, 0.0001, -0.0156, + -0.0240, 0.0090], device='cuda:0'), grad: tensor([-2.5146e-08, 3.6508e-07, -2.9802e-08, -2.7381e-07, 1.5832e-07, + 1.9558e-08, 3.1665e-08, -3.5018e-07, 1.9092e-07, -7.7300e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 220.51, cls_loss 0.0011 cls_loss_mapping 0.0027 cls_loss_causal 0.5081 re_mapping 0.0043 re_causal 0.0124 /// teacc 98.81 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.0009, 0.0110, 0.0439, ..., 0.0850, -0.2541, -0.1679], + [ 0.0568, -0.0185, -0.0278, ..., -0.2964, 0.0534, -0.1017], + [-0.0095, -0.0048, 0.2200, ..., -0.3331, -0.1465, -0.1698], + ..., + [-0.0230, 0.0062, -0.2005, ..., -0.0406, 0.1369, 0.0703], + [-0.0240, -0.0202, -0.1762, ..., -0.2997, 0.1183, -0.2147], + [ 0.0051, -0.0258, -0.0849, ..., 0.0419, -0.1453, 0.0210]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., -2.3283e-08, + 2.7940e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 4.6566e-09, + 6.8173e-07, 2.6636e-07], + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 4.6566e-09, + 2.6077e-08, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.6764e-08, + -7.3574e-07, -2.6543e-07], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + -6.5193e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 1.0245e-08, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0229, 0.0160, 0.0172, 0.0080, 0.0366, -0.0170, -0.0030, -0.0156, + -0.0243, 0.0090], device='cuda:0'), grad: tensor([-3.8184e-08, 1.6158e-06, 8.8476e-08, 1.3690e-07, -4.7497e-08, + 3.4459e-08, -2.0489e-08, -1.7826e-06, -1.4901e-08, 4.0978e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 221.06, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4849 re_mapping 0.0045 re_causal 0.0130 /// teacc 98.92 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.0014, 0.0110, 0.0439, ..., 0.0845, -0.2543, -0.1700], + [ 0.0568, -0.0185, -0.0277, ..., -0.2975, 0.0534, -0.1019], + [-0.0100, -0.0048, 0.2197, ..., -0.3332, -0.1469, -0.1701], + ..., + [-0.0231, 0.0062, -0.2004, ..., -0.0414, 0.1371, 0.0700], + [-0.0256, -0.0202, -0.1763, ..., -0.3002, 0.1187, -0.2156], + [ 0.0083, -0.0258, -0.0850, ..., 0.0425, -0.1454, 0.0216]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., -4.0978e-08, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 5.2154e-08, + 2.8871e-08, 7.0781e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-09, + 2.7940e-09, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 2.6077e-08, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 6.5193e-09, + 3.7253e-09, -1.6764e-08]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0231, 0.0159, 0.0166, 0.0081, 0.0369, -0.0185, -0.0028, -0.0155, + -0.0236, 0.0098], device='cuda:0'), grad: tensor([-6.8918e-08, 2.3842e-07, 3.4459e-08, -4.5355e-07, -1.8440e-07, + 3.9767e-07, -1.0896e-07, 6.9849e-08, 5.4017e-08, 3.3528e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 220.60, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4865 re_mapping 0.0042 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.0014, 0.0110, 0.0440, ..., 0.0852, -0.2545, -0.1702], + [ 0.0567, -0.0185, -0.0279, ..., -0.2980, 0.0532, -0.1019], + [-0.0100, -0.0048, 0.2183, ..., -0.3332, -0.1501, -0.1701], + ..., + [-0.0231, 0.0062, -0.1984, ..., -0.0421, 0.1394, 0.0698], + [-0.0257, -0.0202, -0.1765, ..., -0.3009, 0.1198, -0.2161], + [ 0.0089, -0.0258, -0.0858, ..., 0.0420, -0.1455, 0.0221]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 5.1130e-07, 3.3993e-07], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + 8.0094e-08, 9.3132e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 8.3819e-09, + -6.7521e-07, -3.9116e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 6.5193e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.4901e-08, + 4.3772e-08, 3.7253e-09]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0230, 0.0155, 0.0127, 0.0083, 0.0369, -0.0198, -0.0025, -0.0122, + -0.0229, 0.0097], device='cuda:0'), grad: tensor([ 2.2352e-08, 1.3132e-06, 1.9278e-07, 9.4995e-08, 2.1327e-07, + -5.9325e-07, -1.0245e-07, -1.6782e-06, 4.8336e-07, 5.4948e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 220.53, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4981 re_mapping 0.0044 re_causal 0.0126 /// teacc 98.95 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.0014, 0.0110, 0.0439, ..., 0.0853, -0.2550, -0.1703], + [ 0.0564, -0.0185, -0.0283, ..., -0.2984, 0.0533, -0.1021], + [-0.0100, -0.0048, 0.2188, ..., -0.3334, -0.1501, -0.1702], + ..., + [-0.0232, 0.0062, -0.1986, ..., -0.0425, 0.1395, 0.0697], + [-0.0258, -0.0202, -0.1792, ..., -0.3012, 0.1199, -0.2154], + [ 0.0107, -0.0258, -0.0860, ..., 0.0419, -0.1455, 0.0224]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.0839e-07, ..., 1.0245e-08, + 3.7253e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 6.5193e-09, + -5.0291e-08, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, -1.3057e-06, ..., 9.3132e-09, + 6.7987e-08, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 7.4506e-09, + -1.9558e-08, -2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 6.9849e-08, ..., -2.7008e-08, + -6.7987e-08, 4.4703e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -7.5437e-08, + 4.2841e-08, -2.3562e-07]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0230, 0.0155, 0.0127, 0.0085, 0.0370, -0.0198, -0.0025, -0.0122, + -0.0233, 0.0098], device='cuda:0'), grad: tensor([ 1.6922e-06, 1.0245e-08, -1.6429e-06, 2.7902e-06, 1.1176e-07, + -6.6906e-06, 1.7118e-06, 2.7195e-07, 2.0489e-06, -3.2037e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 220.67, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.4761 re_mapping 0.0043 re_causal 0.0119 /// teacc 98.99 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.0015, 0.0110, 0.0436, ..., 0.0851, -0.2560, -0.1708], + [ 0.0562, -0.0185, -0.0287, ..., -0.2989, 0.0519, -0.1021], + [-0.0100, -0.0048, 0.2199, ..., -0.3333, -0.1502, -0.1704], + ..., + [-0.0232, 0.0062, -0.1990, ..., -0.0425, 0.1406, 0.0700], + [-0.0263, -0.0202, -0.1816, ..., -0.3023, 0.1200, -0.2161], + [ 0.0118, -0.0258, -0.0864, ..., 0.0421, -0.1470, 0.0225]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0245e-08, ..., -1.3039e-08, + 2.7940e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.2107e-08, + 9.2201e-08, 1.1642e-07], + [ 0.0000e+00, 0.0000e+00, -1.6578e-07, ..., 5.5879e-09, + 1.8626e-08, 1.4901e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.1910e-08, ..., -6.6124e-08, + -3.8184e-07, -5.2340e-07], + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 1.8626e-09, + -1.0245e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.4901e-08, + 9.9652e-08, 1.8813e-07]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0234, 0.0142, 0.0128, 0.0078, 0.0369, -0.0186, -0.0023, -0.0115, + -0.0238, 0.0091], device='cuda:0'), grad: tensor([ 6.4261e-08, 4.4052e-07, -1.0338e-07, 3.6135e-07, 6.0350e-07, + 1.3970e-08, -6.0908e-07, -1.3523e-06, 7.3574e-08, 5.0850e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 220.50, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4794 re_mapping 0.0042 re_causal 0.0120 /// teacc 98.88 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.0015, 0.0110, 0.0435, ..., 0.0828, -0.2568, -0.1739], + [ 0.0562, -0.0185, -0.0290, ..., -0.2993, 0.0520, -0.1023], + [-0.0101, -0.0048, 0.2204, ..., -0.3333, -0.1502, -0.1707], + ..., + [-0.0235, 0.0062, -0.1993, ..., -0.0427, 0.1406, 0.0705], + [-0.0269, -0.0202, -0.1833, ..., -0.3026, 0.1204, -0.2163], + [ 0.0125, -0.0258, -0.0866, ..., 0.0445, -0.1481, 0.0234]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, + 9.3132e-10, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 2.4214e-07, 1.9092e-07], + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 0.0000e+00, + 1.3039e-08, 2.8871e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + -5.3830e-07, -1.2750e-06], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 6.5193e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 2.7940e-09, + 2.0955e-07, 8.9593e-07]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0241, 0.0142, 0.0129, 0.0081, 0.0365, -0.0184, -0.0021, -0.0115, + -0.0239, 0.0103], device='cuda:0'), grad: tensor([ 2.8871e-08, 4.3493e-07, 1.7229e-07, -4.6100e-07, 1.0431e-07, + 4.4703e-07, 1.4901e-08, -4.3809e-06, 1.0524e-07, 3.5260e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 220.39, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4474 re_mapping 0.0040 re_causal 0.0118 /// teacc 98.90 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.0015, 0.0110, 0.0435, ..., 0.0829, -0.2575, -0.1739], + [ 0.0563, -0.0185, -0.0293, ..., -0.3001, 0.0521, -0.1025], + [-0.0102, -0.0048, 0.2214, ..., -0.3333, -0.1502, -0.1707], + ..., + [-0.0235, 0.0062, -0.2002, ..., -0.0433, 0.1406, 0.0704], + [-0.0273, -0.0202, -0.1851, ..., -0.3036, 0.1210, -0.2163], + [ 0.0125, -0.0258, -0.0867, ..., 0.0445, -0.1485, 0.0231]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.0536e-08, ..., -9.4064e-08, + 7.4506e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.0245e-08, + -5.9605e-07, 1.8626e-08], + [ 0.0000e+00, 0.0000e+00, -5.2154e-08, ..., 1.8626e-09, + 1.0710e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 7.4506e-08, ..., 4.0047e-08, + 5.0478e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 4.6566e-09, + -2.2128e-06, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 2.6077e-08, + 5.2154e-08, -2.7008e-08]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0241, 0.0142, 0.0130, 0.0080, 0.0369, -0.0186, -0.0021, -0.0116, + -0.0235, 0.0100], device='cuda:0'), grad: tensor([-2.0675e-07, -9.3728e-06, 8.6613e-07, 5.8953e-07, -7.4506e-09, + 6.5193e-07, 3.4589e-06, 8.6203e-06, -4.7386e-06, 1.0990e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 220.84, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4784 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.02 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.0017, 0.0110, 0.0436, ..., 0.0832, -0.2578, -0.1740], + [ 0.0565, -0.0185, -0.0296, ..., -0.3010, 0.0521, -0.1026], + [-0.0102, -0.0048, 0.2235, ..., -0.3333, -0.1500, -0.1702], + ..., + [-0.0235, 0.0062, -0.2027, ..., -0.0443, 0.1393, 0.0695], + [-0.0288, -0.0202, -0.1856, ..., -0.3048, 0.1243, -0.2133], + [ 0.0136, -0.0258, -0.0874, ..., 0.0446, -0.1490, 0.0234]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7008e-07, ..., -2.7940e-09, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 6.5193e-09, + 8.3819e-09, 3.8184e-08], + [ 0.0000e+00, 0.0000e+00, 5.9605e-08, ..., 9.3132e-10, + 7.4506e-09, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 8.3819e-09, + -6.1467e-08, -9.8720e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 6.5193e-09, + 1.4901e-08, 1.3970e-08], + [ 4.6566e-09, 0.0000e+00, 3.7253e-09, ..., 7.5810e-07, + 1.9558e-08, 1.3951e-06]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0240, 0.0142, 0.0134, 0.0072, 0.0367, -0.0178, -0.0020, -0.0123, + -0.0208, 0.0103], device='cuda:0'), grad: tensor([ 1.1409e-06, 1.3411e-07, 3.2969e-07, 2.3376e-07, -1.2834e-06, + -4.8429e-08, -2.1029e-06, -5.4855e-07, 8.7544e-08, 2.0638e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 220.82, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4792 re_mapping 0.0041 re_causal 0.0119 /// teacc 98.90 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.0017, 0.0110, 0.0436, ..., 0.0833, -0.2583, -0.1740], + [ 0.0565, -0.0185, -0.0306, ..., -0.3026, 0.0521, -0.1031], + [-0.0098, -0.0048, 0.2225, ..., -0.3334, -0.1501, -0.1694], + ..., + [-0.0235, 0.0062, -0.2028, ..., -0.0418, 0.1395, 0.0707], + [-0.0290, -0.0202, -0.1859, ..., -0.3052, 0.1242, -0.2135], + [ 0.0136, -0.0258, -0.0894, ..., 0.0439, -0.1512, 0.0219]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 5.1223e-08, 3.7253e-08], + [ 0.0000e+00, 0.0000e+00, -5.6811e-08, ..., 0.0000e+00, + -7.4506e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 3.5390e-08, ..., 9.3132e-10, + -4.9360e-08, -3.7253e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 9.3132e-10, + 1.1176e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, + 4.6566e-09, -8.3819e-09]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0241, 0.0140, 0.0131, 0.0087, 0.0370, -0.0182, -0.0016, -0.0121, + -0.0210, 0.0087], device='cuda:0'), grad: tensor([ 2.3283e-08, 2.0210e-07, -1.0803e-07, 2.0489e-08, 2.7008e-08, + -6.3889e-07, -2.6077e-08, -1.2852e-07, 6.3609e-07, -1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 220.55, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4803 re_mapping 0.0041 re_causal 0.0123 /// teacc 98.99 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.0017, 0.0110, 0.0436, ..., 0.0835, -0.2583, -0.1741], + [ 0.0565, -0.0185, -0.0312, ..., -0.3037, 0.0520, -0.1033], + [-0.0098, -0.0048, 0.2229, ..., -0.3334, -0.1501, -0.1694], + ..., + [-0.0235, 0.0062, -0.2030, ..., -0.0422, 0.1397, 0.0710], + [-0.0291, -0.0202, -0.1864, ..., -0.3056, 0.1239, -0.2136], + [ 0.0136, -0.0258, -0.0896, ..., 0.0440, -0.1514, 0.0224]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 8.3819e-09, + 2.7940e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.9116e-08, + -1.0245e-08, 9.2201e-08], + [ 0.0000e+00, 0.0000e+00, -5.5879e-08, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 1.3039e-08, + 5.5879e-09, 2.5146e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 7.4506e-09, + 1.8626e-09, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -2.9802e-08, + 9.3132e-10, -1.4529e-07]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0241, 0.0137, 0.0131, 0.0088, 0.0363, -0.0184, -0.0013, -0.0120, + -0.0216, 0.0090], device='cuda:0'), grad: tensor([ 3.2596e-08, 1.4342e-07, -8.8476e-08, -1.3877e-07, -1.1455e-07, + 1.5832e-08, 7.9162e-08, 1.1455e-07, 3.3528e-08, -8.8476e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 220.58, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4895 re_mapping 0.0041 re_causal 0.0125 /// teacc 99.05 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.0017, 0.0110, 0.0445, ..., 0.0839, -0.2578, -0.1742], + [ 0.0565, -0.0185, -0.0351, ..., -0.3071, 0.0519, -0.1035], + [-0.0098, -0.0048, 0.2235, ..., -0.3334, -0.1500, -0.1695], + ..., + [-0.0235, 0.0062, -0.2033, ..., -0.0424, 0.1398, 0.0713], + [-0.0292, -0.0202, -0.1879, ..., -0.3063, 0.1235, -0.2135], + [ 0.0136, -0.0258, -0.0897, ..., 0.0438, -0.1517, 0.0219]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 6.5193e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + -5.5879e-09, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 8.3819e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0489e-08, + -8.3819e-09, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.4459e-08, + 6.5193e-09, -8.7544e-08]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0236, 0.0131, 0.0132, 0.0088, 0.0364, -0.0181, -0.0011, -0.0120, + -0.0224, 0.0086], device='cuda:0'), grad: tensor([ 4.6566e-08, 8.3819e-08, 6.9849e-08, 6.8918e-07, 1.8533e-07, + -4.6380e-06, 2.6897e-06, 8.7544e-08, 9.6112e-07, -1.8254e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 220.47, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4650 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.0018, 0.0110, 0.0445, ..., 0.0840, -0.2580, -0.1742], + [ 0.0567, -0.0185, -0.0379, ..., -0.3073, 0.0515, -0.1038], + [-0.0101, -0.0048, 0.2252, ..., -0.3335, -0.1501, -0.1696], + ..., + [-0.0236, 0.0062, -0.2036, ..., -0.0427, 0.1401, 0.0716], + [-0.0297, -0.0202, -0.1884, ..., -0.3066, 0.1236, -0.2129], + [ 0.0136, -0.0258, -0.0899, ..., 0.0435, -0.1535, 0.0215]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 9.3132e-10, + 9.3132e-10, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + -8.3819e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -5.2061e-07, ..., 0.0000e+00, + -9.3132e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5134e-07, ..., 9.3132e-10, + -1.4901e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -7.7300e-08, ..., 1.8626e-09, + 2.4214e-08, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -8.3819e-09, + 9.3132e-10, -4.3772e-08]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0236, 0.0123, 0.0134, 0.0088, 0.0366, -0.0178, -0.0011, -0.0118, + -0.0220, 0.0077], device='cuda:0'), grad: tensor([ 6.2399e-08, 4.6566e-09, -1.0123e-06, 5.3924e-07, 5.3085e-08, + -8.0280e-07, 2.7008e-08, 1.1623e-06, -2.4494e-07, 2.1886e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 221.01, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4884 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.00 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.0018, 0.0110, 0.0443, ..., 0.0840, -0.2582, -0.1742], + [ 0.0568, -0.0185, -0.0382, ..., -0.3076, 0.0516, -0.1038], + [-0.0102, -0.0048, 0.2261, ..., -0.3335, -0.1501, -0.1695], + ..., + [-0.0236, 0.0062, -0.2040, ..., -0.0430, 0.1401, 0.0716], + [-0.0297, -0.0202, -0.1885, ..., -0.3074, 0.1237, -0.2130], + [ 0.0136, -0.0258, -0.0902, ..., 0.0436, -0.1538, 0.0217]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, + -1.7695e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, -1.4622e-07, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.1223e-08, ..., 9.3132e-10, + 1.1176e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 1.8626e-09, + -9.3132e-10, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, + 9.3132e-10, -7.4506e-09]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0236, 0.0122, 0.0135, 0.0085, 0.0364, -0.0174, -0.0010, -0.0118, + -0.0224, 0.0078], device='cuda:0'), grad: tensor([ 3.8184e-08, -2.7940e-08, -1.8813e-07, -2.3209e-06, 3.5390e-08, + 2.2687e-06, -4.0047e-08, 1.3970e-07, 9.1270e-08, -2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 220.86, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4911 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.03 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.0018, 0.0110, 0.0442, ..., 0.0841, -0.2587, -0.1742], + [ 0.0568, -0.0185, -0.0383, ..., -0.3080, 0.0517, -0.1039], + [-0.0102, -0.0048, 0.2265, ..., -0.3335, -0.1502, -0.1696], + ..., + [-0.0236, 0.0062, -0.2042, ..., -0.0434, 0.1401, 0.0717], + [-0.0297, -0.0202, -0.1888, ..., -0.3078, 0.1237, -0.2131], + [ 0.0136, -0.0258, -0.0903, ..., 0.0437, -0.1540, 0.0218]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., 7.4506e-09, + 7.4506e-09, 6.7055e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 5.5879e-09, + -3.7253e-09, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, -1.1548e-07, ..., 5.5879e-09, + 3.7253e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 3.1665e-08, + -5.5879e-09, 5.0291e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-08, ..., 7.4506e-09, + 9.3132e-09, 2.7940e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., -6.0350e-07, + 1.8626e-09, -1.2070e-06]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0239, 0.0121, 0.0135, 0.0083, 0.0362, -0.0170, -0.0006, -0.0117, + -0.0224, 0.0078], device='cuda:0'), grad: tensor([ 4.4703e-08, 2.4214e-08, -1.6019e-07, 2.0862e-07, 1.4752e-06, + -2.2911e-07, 9.3132e-09, 1.2666e-07, 1.8813e-07, -1.7118e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 220.83, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4610 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.08 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.0018, 0.0110, 0.0443, ..., 0.0842, -0.2589, -0.1743], + [ 0.0568, -0.0185, -0.0384, ..., -0.3084, 0.0516, -0.1040], + [-0.0102, -0.0048, 0.2272, ..., -0.3335, -0.1502, -0.1696], + ..., + [-0.0236, 0.0062, -0.2050, ..., -0.0442, 0.1402, 0.0717], + [-0.0298, -0.0202, -0.1900, ..., -0.3084, 0.1237, -0.2132], + [ 0.0136, -0.0258, -0.0906, ..., 0.0435, -0.1541, 0.0217]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.6322e-07, ..., 3.7253e-09, + -2.6077e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, -8.2701e-07, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 1.3039e-08, + 1.8626e-08, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + -3.7253e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6764e-08, + 1.1176e-08, 4.0978e-08]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0239, 0.0120, 0.0136, 0.0084, 0.0364, -0.0167, -0.0006, -0.0118, + -0.0228, 0.0077], device='cuda:0'), grad: tensor([ 2.9802e-08, 1.2685e-06, -2.9802e-06, -1.8217e-06, 1.3039e-08, + 1.7639e-06, 1.4659e-06, 1.6205e-07, -2.9802e-08, 1.4156e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 220.36, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4629 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.08 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.0018, 0.0110, 0.0442, ..., 0.0843, -0.2590, -0.1743], + [ 0.0568, -0.0185, -0.0389, ..., -0.3088, 0.0516, -0.1042], + [-0.0102, -0.0048, 0.2278, ..., -0.3335, -0.1506, -0.1711], + ..., + [-0.0236, 0.0062, -0.2050, ..., -0.0451, 0.1405, 0.0722], + [-0.0298, -0.0202, -0.1904, ..., -0.3100, 0.1237, -0.2133], + [ 0.0136, -0.0258, -0.0907, ..., 0.0435, -0.1545, 0.0218]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4214e-08, ..., -1.8626e-09, + 0.0000e+00, 5.4017e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 7.4506e-09, + 0.0000e+00, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 3.9116e-08, + 4.2841e-08, 9.1270e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 1.3039e-08, + -5.5879e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.5146e-07, + 1.8626e-09, 6.1840e-07], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., -3.0547e-07, + 1.8626e-09, -8.1025e-07]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0239, 0.0119, 0.0136, 0.0083, 0.0364, -0.0167, -0.0007, -0.0117, + -0.0230, 0.0079], device='cuda:0'), grad: tensor([ 2.4773e-07, 5.4017e-08, 5.6624e-07, -2.1234e-07, 1.7881e-07, + 4.2841e-08, 2.0489e-08, 1.2293e-07, 3.3118e-06, -4.3474e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 220.43, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4727 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.0018, 0.0110, 0.0443, ..., 0.0844, -0.2590, -0.1744], + [ 0.0568, -0.0185, -0.0420, ..., -0.3090, 0.0497, -0.1048], + [-0.0102, -0.0048, 0.2304, ..., -0.3335, -0.1493, -0.1711], + ..., + [-0.0237, 0.0062, -0.2052, ..., -0.0457, 0.1410, 0.0727], + [-0.0298, -0.0202, -0.1907, ..., -0.3113, 0.1237, -0.2136], + [ 0.0136, -0.0258, -0.0907, ..., 0.0437, -0.1546, 0.0222]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 3.3528e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 4.4145e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., -1.8626e-09, + 5.4017e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 6.7055e-08, ..., 3.7253e-09, + -5.8115e-07, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-08, ..., 0.0000e+00, + -7.4506e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 1.3039e-07, + 3.7253e-09, 5.8673e-07]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0238, 0.0096, 0.0145, 0.0092, 0.0361, -0.0170, -0.0007, -0.0114, + -0.0232, 0.0083], device='cuda:0'), grad: tensor([ 1.7695e-07, 1.1027e-06, 1.3597e-07, 4.0978e-07, -6.1654e-07, + -1.4994e-06, 3.1106e-07, -1.1586e-06, 3.8184e-07, 7.5251e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 220.44, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4605 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.0018, 0.0110, 0.0442, ..., 0.0844, -0.2595, -0.1744], + [ 0.0568, -0.0185, -0.0420, ..., -0.3092, 0.0489, -0.1055], + [-0.0102, -0.0048, 0.2305, ..., -0.3335, -0.1492, -0.1711], + ..., + [-0.0237, 0.0062, -0.2056, ..., -0.0465, 0.1417, 0.0734], + [-0.0298, -0.0202, -0.1917, ..., -0.3132, 0.1237, -0.2140], + [ 0.0136, -0.0258, -0.0909, ..., 0.0455, -0.1547, 0.0257]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -1.3039e-08, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 8.5682e-08, ..., 1.8626e-09, + -1.6764e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -4.0606e-07, ..., 5.5879e-09, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.7742e-08, ..., 0.0000e+00, + 1.8626e-09, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 1.8626e-09, + 1.8626e-09, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 1.8626e-09, + 5.5879e-09, -7.4506e-08]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0238, 0.0089, 0.0145, 0.0094, 0.0327, -0.0170, -0.0007, -0.0111, + -0.0235, 0.0118], device='cuda:0'), grad: tensor([-5.0291e-08, 1.1362e-07, -7.3761e-07, 4.2841e-08, 4.8243e-07, + 1.0990e-07, -3.1665e-08, 1.1548e-07, 1.0245e-07, -1.4715e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 220.58, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4779 re_mapping 0.0042 re_causal 0.0122 /// teacc 98.92 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.0018, 0.0110, 0.0443, ..., 0.0845, -0.2597, -0.1744], + [ 0.0568, -0.0185, -0.0420, ..., -0.3093, 0.0490, -0.1056], + [-0.0102, -0.0048, 0.2306, ..., -0.3336, -0.1493, -0.1712], + ..., + [-0.0239, 0.0062, -0.2059, ..., -0.0470, 0.1418, 0.0734], + [-0.0299, -0.0202, -0.1924, ..., -0.3135, 0.1244, -0.2142], + [ 0.0135, -0.0258, -0.0911, ..., 0.0452, -0.1550, 0.0256]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.4703e-08, ..., -2.0489e-08, + 9.3132e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, + 1.1548e-07, 1.2480e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.1176e-08, + -1.4342e-07, -1.2666e-07], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.1176e-08, + -7.4506e-09, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 2.6077e-08, ..., 2.1420e-07, + 1.8626e-09, 3.0361e-07]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0238, 0.0090, 0.0145, 0.0098, 0.0328, -0.0191, -0.0009, -0.0111, + -0.0218, 0.0117], device='cuda:0'), grad: tensor([-3.1665e-08, 4.3027e-07, 2.6077e-08, 1.8813e-07, -7.3761e-07, + -7.5139e-06, 6.4038e-06, -4.6752e-07, 9.9279e-07, 6.7987e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 220.94, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4883 re_mapping 0.0041 re_causal 0.0122 /// teacc 98.95 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.0018, 0.0110, 0.0443, ..., 0.0846, -0.2604, -0.1745], + [ 0.0568, -0.0185, -0.0423, ..., -0.3095, 0.0465, -0.1038], + [-0.0102, -0.0048, 0.2308, ..., -0.3336, -0.1495, -0.1713], + ..., + [-0.0239, 0.0062, -0.2061, ..., -0.0477, 0.1442, 0.0711], + [-0.0300, -0.0202, -0.1931, ..., -0.3142, 0.1246, -0.2144], + [ 0.0135, -0.0258, -0.0913, ..., 0.0448, -0.1546, 0.0255]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., -7.4506e-09, + 1.1176e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, -5.1595e-07, ..., 0.0000e+00, + -4.4703e-07, 2.0675e-07], + [ 0.0000e+00, 0.0000e+00, 3.3714e-07, ..., 3.7253e-09, + 1.2591e-06, 1.1176e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1735e-07, ..., 0.0000e+00, + -1.2126e-06, -3.1851e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.7253e-09, + 6.8918e-08, 5.0291e-08]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0238, 0.0074, 0.0145, 0.0097, 0.0329, -0.0189, -0.0012, -0.0100, + -0.0216, 0.0117], device='cuda:0'), grad: tensor([ 3.3528e-08, -1.1269e-06, 3.5316e-06, 8.5495e-07, 8.9407e-08, + 7.0781e-08, 4.6566e-08, -3.8706e-06, 6.1467e-08, 2.9616e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 221.09, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4814 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.01 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.0018, 0.0110, 0.0444, ..., 0.0846, -0.2608, -0.1746], + [ 0.0568, -0.0185, -0.0431, ..., -0.3097, 0.0465, -0.1027], + [-0.0102, -0.0048, 0.2317, ..., -0.3336, -0.1492, -0.1714], + ..., + [-0.0239, 0.0062, -0.2063, ..., -0.0479, 0.1441, 0.0701], + [-0.0300, -0.0202, -0.1933, ..., -0.3144, 0.1246, -0.2144], + [ 0.0135, -0.0258, -0.0914, ..., 0.0447, -0.1551, 0.0256]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., -3.5390e-08, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 3.5390e-08, + -4.1164e-07, 2.9802e-08], + [ 0.0000e+00, 0.0000e+00, -2.9802e-08, ..., 5.5879e-09, + -1.4901e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.7055e-08, ..., 3.7253e-09, + 5.0291e-08, -2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 0.0000e+00, + 5.5879e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 2.2352e-08, + 2.6077e-08, 1.4901e-08]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0238, 0.0077, 0.0148, 0.0098, 0.0328, -0.0194, -0.0011, -0.0104, + -0.0216, 0.0117], device='cuda:0'), grad: tensor([-5.5879e-08, -7.3761e-07, 7.4506e-09, -3.5949e-07, -1.5832e-07, + 5.4017e-08, 6.5751e-07, 2.3097e-07, 2.2165e-07, 1.4715e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 220.70, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4724 re_mapping 0.0043 re_causal 0.0124 /// teacc 98.94 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.0018, 0.0110, 0.0443, ..., 0.0836, -0.2611, -0.1757], + [ 0.0568, -0.0185, -0.0433, ..., -0.3102, 0.0464, -0.1028], + [-0.0102, -0.0048, 0.2324, ..., -0.3338, -0.1488, -0.1715], + ..., + [-0.0239, 0.0062, -0.2074, ..., -0.0489, 0.1440, 0.0701], + [-0.0300, -0.0202, -0.1943, ..., -0.3150, 0.1237, -0.2146], + [ 0.0135, -0.0258, -0.0911, ..., 0.0454, -0.1563, 0.0258]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.2107e-07, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.7195e-07, ..., 0.0000e+00, + -2.7940e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.8429e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.6368e-08, ..., 0.0000e+00, + 7.4506e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, -5.5879e-09]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0247, 0.0075, 0.0153, 0.0093, 0.0330, -0.0170, -0.0008, -0.0106, + -0.0232, 0.0122], device='cuda:0'), grad: tensor([ 6.8918e-08, 2.0489e-07, -4.9919e-07, 2.7940e-08, 1.1176e-08, + -9.1270e-08, 2.9802e-08, 9.4995e-08, 1.7136e-07, -7.4506e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 220.99, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4880 re_mapping 0.0040 re_causal 0.0122 /// teacc 98.94 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.0018, 0.0110, 0.0440, ..., 0.0816, -0.2615, -0.1777], + [ 0.0568, -0.0185, -0.0434, ..., -0.3119, 0.0466, -0.1029], + [-0.0102, -0.0048, 0.2340, ..., -0.3339, -0.1476, -0.1718], + ..., + [-0.0239, 0.0062, -0.2101, ..., -0.0495, 0.1435, 0.0702], + [-0.0300, -0.0202, -0.1952, ..., -0.3154, 0.1237, -0.2147], + [ 0.0135, -0.0258, -0.0901, ..., 0.0474, -0.1574, 0.0264]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.0291e-08, ..., -1.3039e-08, + 0.0000e+00, 7.0781e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 1.3039e-08, + 1.6764e-08, 2.4214e-08], + [ 0.0000e+00, 0.0000e+00, -4.6566e-08, ..., 1.4901e-08, + 9.3132e-09, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 2.6077e-08, + -2.9802e-08, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 5.5879e-09, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., -1.3970e-07, + 1.8626e-09, -3.1106e-07]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0262, 0.0076, 0.0167, 0.0086, 0.0330, -0.0165, -0.0009, -0.0116, + -0.0233, 0.0133], device='cuda:0'), grad: tensor([-4.2841e-08, 7.8231e-08, -3.9116e-08, 5.0291e-08, 2.8126e-07, + 9.3132e-09, 5.2154e-08, 7.6368e-08, -1.3039e-08, -4.5449e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 220.60, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4665 re_mapping 0.0040 re_causal 0.0115 /// teacc 98.98 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.0018, 0.0110, 0.0441, ..., 0.0816, -0.2621, -0.1778], + [ 0.0568, -0.0185, -0.0435, ..., -0.3132, 0.0467, -0.1030], + [-0.0102, -0.0048, 0.2341, ..., -0.3344, -0.1477, -0.1720], + ..., + [-0.0239, 0.0062, -0.2102, ..., -0.0508, 0.1434, 0.0700], + [-0.0300, -0.0202, -0.1963, ..., -0.3168, 0.1237, -0.2149], + [ 0.0135, -0.0258, -0.0909, ..., 0.0473, -0.1566, 0.0263]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.3039e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, -1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-09, 9.3132e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + -5.4017e-08, -9.4995e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.8626e-09, + 2.7940e-08, 5.0291e-08]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0262, 0.0076, 0.0166, 0.0089, 0.0335, -0.0164, -0.0010, -0.0117, + -0.0235, 0.0131], device='cuda:0'), grad: tensor([ 3.0920e-07, -8.7544e-08, 1.1735e-07, 3.7253e-08, 9.7416e-07, + 8.1956e-08, -1.4175e-06, -2.1607e-07, 4.6566e-08, 1.4901e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 220.74, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4669 re_mapping 0.0040 re_causal 0.0115 /// teacc 98.99 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.0018, 0.0110, 0.0441, ..., 0.0817, -0.2635, -0.1778], + [ 0.0568, -0.0185, -0.0434, ..., -0.3136, 0.0466, -0.1032], + [-0.0101, -0.0048, 0.2341, ..., -0.3345, -0.1482, -0.1728], + ..., + [-0.0239, 0.0062, -0.2102, ..., -0.0515, 0.1437, 0.0704], + [-0.0300, -0.0202, -0.1956, ..., -0.3173, 0.1244, -0.2145], + [ 0.0135, -0.0258, -0.0913, ..., 0.0472, -0.1572, 0.0263]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., -1.8626e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 1.8626e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., 1.8626e-09, + 1.8626e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 3.7253e-09, + -2.4214e-08, -1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 5.5879e-09, + 0.0000e+00, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.1665e-08, + 1.8626e-08, 2.7940e-08]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0265, 0.0076, 0.0163, 0.0084, 0.0335, -0.0159, -0.0009, -0.0115, + -0.0221, 0.0130], device='cuda:0'), grad: tensor([-2.7940e-08, 7.0781e-08, 0.0000e+00, 9.7789e-07, -9.6858e-08, + -1.4398e-06, 4.8429e-08, -3.5390e-08, 3.5763e-07, 1.4342e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 220.67, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4918 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.07 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.0018, 0.0110, 0.0442, ..., 0.0817, -0.2646, -0.1778], + [ 0.0568, -0.0185, -0.0434, ..., -0.3141, 0.0466, -0.1028], + [-0.0101, -0.0048, 0.2342, ..., -0.3345, -0.1483, -0.1731], + ..., + [-0.0239, 0.0062, -0.2103, ..., -0.0524, 0.1440, 0.0702], + [-0.0300, -0.0202, -0.1983, ..., -0.3178, 0.1239, -0.2148], + [ 0.0135, -0.0258, -0.0913, ..., 0.0471, -0.1585, 0.0261]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., -1.6764e-08, + 5.5879e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 5.5879e-09, + -6.1467e-08, 2.2352e-08], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 1.8626e-09, + 3.3528e-08, 1.8626e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -1.7695e-07, -1.7136e-07], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.1176e-08, + 2.7940e-08, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 0.0000e+00, -1.8626e-08]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0265, 0.0079, 0.0164, 0.0077, 0.0336, -0.0158, -0.0008, -0.0115, + -0.0228, 0.0128], device='cuda:0'), grad: tensor([-3.1665e-08, -8.9407e-08, 1.0617e-07, 5.5321e-07, 1.8626e-09, + 9.3132e-09, 2.4214e-08, -6.5938e-07, 1.3225e-07, -5.7742e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 220.69, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4935 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.04 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.0018, 0.0110, 0.0444, ..., 0.0816, -0.2653, -0.1780], + [ 0.0568, -0.0185, -0.0434, ..., -0.3154, 0.0483, -0.1018], + [-0.0101, -0.0048, 0.2344, ..., -0.3346, -0.1483, -0.1733], + ..., + [-0.0239, 0.0062, -0.2106, ..., -0.0546, 0.1427, 0.0691], + [-0.0300, -0.0202, -0.1989, ..., -0.3186, 0.1228, -0.2150], + [ 0.0135, -0.0258, -0.0915, ..., 0.0473, -0.1590, 0.0262]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., -2.7940e-08, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 7.4506e-09, + 0.0000e+00, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., 5.5879e-09, + 1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 5.5879e-09, + 0.0000e+00, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 7.4506e-09, + 3.7253e-09, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., -9.4995e-08, + 0.0000e+00, -1.6764e-07]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0265, 0.0094, 0.0164, 0.0073, 0.0335, -0.0156, -0.0008, -0.0126, + -0.0244, 0.0128], device='cuda:0'), grad: tensor([-5.4017e-08, 3.9116e-08, 1.2480e-07, -2.6822e-07, 2.6636e-07, + 2.0489e-08, -1.6950e-07, 7.0781e-08, 8.7544e-08, -1.2852e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 220.51, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4770 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.05 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.0018, 0.0110, 0.0445, ..., 0.0817, -0.2656, -0.1780], + [ 0.0568, -0.0185, -0.0435, ..., -0.3167, 0.0484, -0.1016], + [-0.0101, -0.0048, 0.2350, ..., -0.3346, -0.1483, -0.1736], + ..., + [-0.0239, 0.0062, -0.2115, ..., -0.0556, 0.1426, 0.0689], + [-0.0300, -0.0202, -0.1996, ..., -0.3193, 0.1229, -0.2153], + [ 0.0135, -0.0258, -0.0917, ..., 0.0473, -0.1594, 0.0263]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.8243e-07, 3.5949e-07], + [ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., 0.0000e+00, + 3.7253e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + -6.3702e-07, -6.0350e-07], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -5.5879e-09, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.0803e-07, 2.0303e-07]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0265, 0.0096, 0.0169, 0.0073, 0.0335, -0.0157, -0.0009, -0.0130, + -0.0247, 0.0129], device='cuda:0'), grad: tensor([ 1.4901e-08, 1.4286e-06, -5.7742e-08, 6.8918e-08, 7.0781e-08, + 3.1665e-08, -3.7253e-08, -2.0750e-06, 1.1176e-08, 5.3085e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 220.73, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4937 re_mapping 0.0040 re_causal 0.0123 /// teacc 99.04 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.0018, 0.0110, 0.0453, ..., 0.0819, -0.2659, -0.1780], + [ 0.0568, -0.0185, -0.0435, ..., -0.3152, 0.0482, -0.1016], + [-0.0101, -0.0048, 0.2350, ..., -0.3346, -0.1486, -0.1749], + ..., + [-0.0239, 0.0062, -0.2116, ..., -0.0560, 0.1430, 0.0693], + [-0.0300, -0.0202, -0.2001, ..., -0.3200, 0.1228, -0.2154], + [ 0.0135, -0.0258, -0.0928, ..., 0.0471, -0.1598, 0.0262]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., -7.4506e-09, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-07, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8999e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.1176e-08, -5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, + -1.0487e-06, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.3039e-08, + 7.2643e-08, -1.3039e-08]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0262, 0.0095, 0.0169, 0.0069, 0.0334, -0.0157, -0.0009, -0.0128, + -0.0248, 0.0127], device='cuda:0'), grad: tensor([-7.4506e-09, 1.6987e-06, 4.6939e-07, 1.2480e-07, 2.4214e-08, + 1.6764e-08, 1.4901e-08, 2.9802e-08, -2.6803e-06, 3.0920e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 221.00, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4688 re_mapping 0.0039 re_causal 0.0117 /// teacc 98.95 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.0018, 0.0110, 0.0461, ..., 0.0821, -0.2662, -0.1780], + [ 0.0568, -0.0185, -0.0435, ..., -0.3155, 0.0481, -0.1017], + [-0.0101, -0.0048, 0.2351, ..., -0.3347, -0.1486, -0.1751], + ..., + [-0.0239, 0.0062, -0.2118, ..., -0.0581, 0.1433, 0.0696], + [-0.0300, -0.0202, -0.2005, ..., -0.3206, 0.1228, -0.2156], + [ 0.0135, -0.0258, -0.0949, ..., 0.0470, -0.1606, 0.0262]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -3.7253e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -3.0547e-07, ..., 0.0000e+00, + -2.0675e-07, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -3.5390e-07, ..., 0.0000e+00, + 1.2107e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.7323e-07, ..., 0.0000e+00, + 8.0094e-08, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, -1.1176e-08]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0259, 0.0094, 0.0169, 0.0069, 0.0334, -0.0151, -0.0010, -0.0127, + -0.0249, 0.0125], device='cuda:0'), grad: tensor([ 3.7253e-08, -3.3099e-06, 7.5251e-07, 1.0245e-07, 1.0226e-06, + -6.8918e-08, 5.5879e-09, 1.3988e-06, 6.1467e-08, -7.4506e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 220.84, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4908 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.02 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.0018, 0.0110, 0.0466, ..., 0.0823, -0.2662, -0.1780], + [ 0.0573, -0.0185, -0.0435, ..., -0.3160, 0.0479, -0.1019], + [-0.0105, -0.0048, 0.2352, ..., -0.3348, -0.1487, -0.1753], + ..., + [-0.0239, 0.0062, -0.2119, ..., -0.0600, 0.1436, 0.0702], + [-0.0308, -0.0202, -0.2012, ..., -0.3220, 0.1227, -0.2159], + [ 0.0135, -0.0258, -0.0956, ..., 0.0470, -0.1623, 0.0259]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., -3.3528e-08, + 1.4901e-08, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 3.2037e-07, ..., 1.8626e-09, + -2.4214e-07, 2.5332e-07], + [ 0.0000e+00, 0.0000e+00, -1.7621e-06, ..., 0.0000e+00, + 6.8918e-08, -1.0915e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2852e-06, ..., 3.7253e-09, + -1.3784e-07, 6.5379e-07], + [ 0.0000e+00, 0.0000e+00, 6.5193e-08, ..., 0.0000e+00, + 1.8626e-09, 4.2841e-08], + [ 0.0000e+00, 0.0000e+00, 7.6368e-08, ..., 7.4506e-09, + 3.9116e-08, 9.8720e-08]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0257, 0.0092, 0.0169, 0.0063, 0.0333, -0.0149, -0.0012, -0.0124, + -0.0252, 0.0122], device='cuda:0'), grad: tensor([-2.2352e-08, -2.9802e-08, -3.3993e-06, 3.5390e-08, 9.6858e-08, + 2.7940e-08, 6.4448e-07, 2.1383e-06, 1.5832e-07, 3.3341e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 220.41, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4792 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.03 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.0018, 0.0110, 0.0469, ..., 0.0824, -0.2681, -0.1781], + [ 0.0573, -0.0185, -0.0435, ..., -0.3161, 0.0480, -0.1019], + [-0.0105, -0.0048, 0.2354, ..., -0.3350, -0.1487, -0.1754], + ..., + [-0.0239, 0.0062, -0.2121, ..., -0.0613, 0.1436, 0.0702], + [-0.0308, -0.0202, -0.2014, ..., -0.3227, 0.1228, -0.2161], + [ 0.0135, -0.0258, -0.0957, ..., 0.0471, -0.1625, 0.0261]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.2480e-07, ..., 0.0000e+00, + -5.2154e-08, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, -3.9861e-07, ..., 0.0000e+00, + -1.8626e-08, 7.4506e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3155e-07, ..., 0.0000e+00, + 8.0094e-08, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -5.5879e-09, + 0.0000e+00, -1.3039e-08]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0258, 0.0093, 0.0169, 0.0049, 0.0331, -0.0149, -0.0008, -0.0124, + -0.0236, 0.0123], device='cuda:0'), grad: tensor([ 2.7940e-08, 8.9407e-08, -4.6566e-07, -3.5390e-07, 9.3132e-09, + -1.0431e-07, -3.7253e-08, 7.8417e-07, 5.5879e-08, -5.5879e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 220.36, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.5090 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.08 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.0018, 0.0110, 0.0479, ..., 0.0825, -0.2686, -0.1781], + [ 0.0573, -0.0185, -0.0436, ..., -0.3175, 0.0480, -0.1020], + [-0.0105, -0.0048, 0.2355, ..., -0.3354, -0.1488, -0.1756], + ..., + [-0.0239, 0.0062, -0.2122, ..., -0.0619, 0.1437, 0.0702], + [-0.0308, -0.0202, -0.2022, ..., -0.3258, 0.1228, -0.2164], + [ 0.0135, -0.0258, -0.0964, ..., 0.0470, -0.1629, 0.0262]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7509e-07, ..., -3.5763e-07, + 0.0000e+00, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 1.6764e-08, + -2.9802e-08, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 1.8626e-08, + 0.0000e+00, 8.7544e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 5.0291e-08, ..., 1.6764e-08, + 2.0489e-08, 3.9116e-08], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 7.4506e-09, + 0.0000e+00, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, -2.7940e-08, ..., -1.4901e-08, + 3.7253e-09, -3.5949e-07]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0256, 0.0093, 0.0169, 0.0048, 0.0333, -0.0148, -0.0009, -0.0125, + -0.0239, 0.0123], device='cuda:0'), grad: tensor([-6.8359e-07, -1.4901e-08, 3.2224e-07, -1.0058e-07, -4.4703e-08, + 2.7753e-07, 6.4448e-07, 3.3900e-07, 8.0094e-08, -8.3819e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 220.35, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4648 re_mapping 0.0041 re_causal 0.0114 /// teacc 99.10 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.0018, 0.0110, 0.0538, ..., 0.0848, -0.2668, -0.1781], + [ 0.0573, -0.0185, -0.0437, ..., -0.3160, 0.0481, -0.1012], + [-0.0105, -0.0048, 0.2359, ..., -0.3360, -0.1488, -0.1758], + ..., + [-0.0239, 0.0062, -0.2125, ..., -0.0633, 0.1437, 0.0702], + [-0.0310, -0.0202, -0.2070, ..., -0.3303, 0.1225, -0.2165], + [ 0.0134, -0.0258, -0.1015, ..., 0.0465, -0.1643, 0.0269]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 1.7881e-07, 5.0291e-08], + [ 0.0000e+00, 0.0000e+00, -3.0641e-07, ..., -1.8626e-09, + -9.4995e-08, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.4866e-07, ..., 2.7940e-09, + -1.1269e-07, -5.2154e-08], + [ 0.0000e+00, 0.0000e+00, 4.2841e-08, ..., 1.8626e-09, + 2.7940e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.5635e-08, + 8.3819e-09, -1.0617e-07]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0212, 0.0099, 0.0170, 0.0050, 0.0319, -0.0163, -0.0029, -0.0126, + -0.0273, 0.0119], device='cuda:0'), grad: tensor([ 1.1176e-08, 4.6194e-07, -7.3574e-07, 7.7300e-08, 1.5926e-07, + -1.4901e-07, 4.3772e-08, 1.4435e-07, 1.2480e-07, -1.5553e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 220.59, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4821 re_mapping 0.0043 re_causal 0.0120 /// teacc 99.04 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.0018, 0.0110, 0.0540, ..., 0.0849, -0.2671, -0.1782], + [ 0.0573, -0.0185, -0.0437, ..., -0.3160, 0.0482, -0.1011], + [-0.0105, -0.0048, 0.2362, ..., -0.3361, -0.1490, -0.1762], + ..., + [-0.0239, 0.0062, -0.2129, ..., -0.0635, 0.1436, 0.0698], + [-0.0310, -0.0202, -0.2072, ..., -0.3304, 0.1227, -0.2167], + [ 0.0134, -0.0258, -0.1019, ..., 0.0459, -0.1648, 0.0271]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.4063e-07, ..., -3.8184e-08, + 2.7008e-08, 3.2559e-06], + [ 0.0000e+00, 0.0000e+00, 2.7008e-08, ..., 3.7253e-09, + -1.1986e-06, 9.4064e-08], + [ 0.0000e+00, 0.0000e+00, -1.2582e-06, ..., 2.4214e-08, + 2.8871e-08, 6.4224e-06], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5236e-06, ..., 1.4901e-08, + -2.0489e-08, 1.6019e-07], + [ 0.0000e+00, 0.0000e+00, 3.5390e-08, ..., 5.5879e-09, + 3.7253e-08, 8.9407e-08], + [ 0.0000e+00, 0.0000e+00, -8.0839e-07, ..., -4.0978e-08, + 9.3132e-09, -1.4618e-05]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0211, 0.0100, 0.0171, 0.0055, 0.0320, -0.0155, -0.0033, -0.0128, + -0.0274, 0.0118], device='cuda:0'), grad: tensor([ 8.1435e-06, -2.5164e-06, 1.3269e-05, 6.5472e-07, 1.1744e-06, + 9.8869e-06, 2.4997e-06, 3.1628e-06, 3.6042e-07, -3.6567e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 220.81, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4852 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.06 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.0018, 0.0110, 0.0539, ..., 0.0849, -0.2674, -0.1783], + [ 0.0573, -0.0185, -0.0455, ..., -0.3163, 0.0483, -0.1013], + [-0.0105, -0.0048, 0.2385, ..., -0.3361, -0.1490, -0.1768], + ..., + [-0.0239, 0.0062, -0.2141, ..., -0.0668, 0.1436, 0.0689], + [-0.0310, -0.0202, -0.2073, ..., -0.3305, 0.1238, -0.2169], + [ 0.0135, -0.0258, -0.1020, ..., 0.0458, -0.1658, 0.0272]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.2107e-08, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, -9.3132e-08, ..., 2.7940e-09, + 2.4214e-08, -2.8871e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.3039e-08, + -1.1735e-07, -1.1548e-07], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-08, ..., 1.2107e-08, + 3.7253e-09, 7.4506e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0212, 0.0093, 0.0183, 0.0050, 0.0328, -0.0155, -0.0039, -0.0134, + -0.0268, 0.0119], device='cuda:0'), grad: tensor([ 4.0047e-08, 5.8673e-08, -1.0245e-08, 4.1071e-07, 1.1269e-07, + -2.5518e-07, 0.0000e+00, -5.7369e-07, 3.7253e-08, 1.9092e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 221.01, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4540 re_mapping 0.0040 re_causal 0.0113 /// teacc 99.10 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.0018, 0.0110, 0.0539, ..., 0.0850, -0.2678, -0.1784], + [ 0.0573, -0.0185, -0.0457, ..., -0.3165, 0.0482, -0.1015], + [-0.0105, -0.0048, 0.2395, ..., -0.3361, -0.1490, -0.1770], + ..., + [-0.0239, 0.0062, -0.2155, ..., -0.0669, 0.1438, 0.0693], + [-0.0310, -0.0202, -0.2074, ..., -0.3306, 0.1246, -0.2171], + [ 0.0135, -0.0258, -0.1031, ..., 0.0456, -0.1681, 0.0269]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 8.1025e-08, 9.5926e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.2643e-08, 4.9360e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -4.4145e-07, -3.7439e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -1.0245e-08, 5.2154e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, + 2.0396e-07, 1.7043e-07]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0212, 0.0092, 0.0187, 0.0059, 0.0329, -0.0158, -0.0044, -0.0136, + -0.0264, 0.0115], device='cuda:0'), grad: tensor([ 1.0245e-08, 2.1607e-07, 1.8254e-07, 1.7975e-07, 9.3132e-10, + 1.1735e-07, -4.6566e-09, -1.0822e-06, -4.8429e-08, 4.5449e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 220.74, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4581 re_mapping 0.0040 re_causal 0.0116 /// teacc 98.98 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.0018, 0.0110, 0.0539, ..., 0.0850, -0.2683, -0.1784], + [ 0.0578, -0.0185, -0.0458, ..., -0.3167, 0.0480, -0.1016], + [-0.0106, -0.0048, 0.2398, ..., -0.3361, -0.1491, -0.1773], + ..., + [-0.0239, 0.0062, -0.2158, ..., -0.0671, 0.1443, 0.0694], + [-0.0311, -0.0202, -0.2075, ..., -0.3306, 0.1241, -0.2177], + [ 0.0134, -0.0258, -0.1031, ..., 0.0457, -0.1691, 0.0269]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + -7.9162e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 0.0000e+00, + 4.1910e-08, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0212, 0.0090, 0.0189, 0.0051, 0.0329, -0.0151, -0.0044, -0.0134, + -0.0268, 0.0115], device='cuda:0'), grad: tensor([ 3.5390e-08, -2.4308e-07, 1.0151e-07, -1.3411e-07, 0.0000e+00, + 1.1176e-08, -3.5390e-08, 2.0210e-07, 5.5879e-08, 3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 220.41, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4569 re_mapping 0.0040 re_causal 0.0113 /// teacc 98.99 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.0019, 0.0110, 0.0541, ..., 0.0852, -0.2694, -0.1785], + [ 0.0578, -0.0185, -0.0458, ..., -0.3172, 0.0479, -0.1017], + [-0.0106, -0.0048, 0.2399, ..., -0.3362, -0.1492, -0.1774], + ..., + [-0.0239, 0.0062, -0.2160, ..., -0.0673, 0.1445, 0.0695], + [-0.0312, -0.0202, -0.2075, ..., -0.3307, 0.1246, -0.2181], + [ 0.0134, -0.0258, -0.1034, ..., 0.0472, -0.1689, 0.0288]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -9.3132e-10, + 6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + -4.6566e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 2.7940e-09, -9.3132e-10]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0212, 0.0090, 0.0189, 0.0047, 0.0317, -0.0149, -0.0049, -0.0133, + -0.0265, 0.0127], device='cuda:0'), grad: tensor([ 4.6566e-08, -3.7253e-09, 2.0489e-08, 2.1979e-07, -2.7940e-09, + -2.5984e-07, -8.1025e-08, 1.8626e-08, 2.5146e-08, 1.9558e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 220.86, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4811 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.05 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.0019, 0.0110, 0.0543, ..., 0.0852, -0.2699, -0.1785], + [ 0.0578, -0.0185, -0.0458, ..., -0.3173, 0.0479, -0.1018], + [-0.0107, -0.0048, 0.2400, ..., -0.3362, -0.1493, -0.1775], + ..., + [-0.0239, 0.0062, -0.2161, ..., -0.0674, 0.1446, 0.0695], + [-0.0313, -0.0202, -0.2075, ..., -0.3307, 0.1247, -0.2181], + [ 0.0134, -0.0258, -0.1035, ..., 0.0471, -0.1692, 0.0286]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.1944e-07, 3.0734e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.2165e-07, 2.1420e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -8.0094e-08, -2.0675e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.1828e-07, 1.4342e-07]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0210, 0.0089, 0.0188, 0.0027, 0.0318, -0.0127, -0.0050, -0.0133, + -0.0265, 0.0125], device='cuda:0'), grad: tensor([ 5.3085e-08, -2.5444e-06, 1.5544e-06, 2.0489e-08, 3.0734e-07, + 6.5193e-08, -9.2201e-08, -2.2352e-08, 4.7497e-08, 6.2305e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 221.03, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4805 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.05 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.0019, 0.0110, 0.0543, ..., 0.0852, -0.2709, -0.1787], + [ 0.0578, -0.0185, -0.0459, ..., -0.3176, 0.0483, -0.1015], + [-0.0107, -0.0048, 0.2404, ..., -0.3362, -0.1494, -0.1776], + ..., + [-0.0239, 0.0062, -0.2164, ..., -0.0676, 0.1445, 0.0693], + [-0.0313, -0.0202, -0.2076, ..., -0.3307, 0.1243, -0.2184], + [ 0.0134, -0.0258, -0.1034, ..., 0.0470, -0.1704, 0.0285]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -4.6566e-09, + 4.6566e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.0734e-08, + 2.1420e-08, 5.9605e-08], + [ 0.0000e+00, 0.0000e+00, -5.0291e-08, ..., 1.8626e-09, + 7.4506e-09, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.8429e-08, ..., 2.2352e-08, + -2.9802e-08, 2.4214e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 6.5193e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.3190e-07, + 1.8626e-09, 3.3807e-07]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0210, 0.0094, 0.0190, 0.0023, 0.0320, -0.0129, -0.0049, -0.0135, + -0.0269, 0.0124], device='cuda:0'), grad: tensor([-2.7940e-09, 1.6112e-07, -5.7742e-08, -2.2352e-08, -7.0408e-07, + -2.5053e-07, 4.3772e-08, 1.2666e-07, 2.1793e-07, 5.0478e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 220.72, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4281 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.06 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.0019, 0.0110, 0.0544, ..., 0.0853, -0.2710, -0.1788], + [ 0.0576, -0.0185, -0.0459, ..., -0.3181, 0.0483, -0.1016], + [-0.0107, -0.0048, 0.2407, ..., -0.3364, -0.1496, -0.1777], + ..., + [-0.0240, 0.0062, -0.2167, ..., -0.0678, 0.1445, 0.0691], + [-0.0313, -0.0202, -0.2077, ..., -0.3307, 0.1244, -0.2186], + [ 0.0147, -0.0258, -0.1036, ..., 0.0468, -0.1707, 0.0285]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7043e-07, ..., -1.0896e-07, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 1.0710e-07, 6.3330e-08], + [ 0.0000e+00, 0.0000e+00, 1.3039e-07, ..., 6.8918e-08, + 2.5146e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -1.2387e-07, -7.0781e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.1176e-08, + -2.5146e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.8871e-08, ..., 2.7008e-08, + 3.7253e-09, 3.7253e-09]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0210, 0.0095, 0.0190, 0.0023, 0.0322, -0.0130, -0.0051, -0.0137, + -0.0269, 0.0123], device='cuda:0'), grad: tensor([-1.0561e-06, 2.2817e-07, 8.5682e-07, -3.4552e-07, -4.6566e-09, + 2.8498e-07, 2.0675e-07, -2.3842e-07, -7.7300e-08, 1.3877e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 220.97, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4830 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.0019, 0.0110, 0.0545, ..., 0.0853, -0.2716, -0.1788], + [ 0.0576, -0.0185, -0.0460, ..., -0.3182, 0.0480, -0.1020], + [-0.0107, -0.0048, 0.2406, ..., -0.3366, -0.1500, -0.1780], + ..., + [-0.0240, 0.0062, -0.2172, ..., -0.0678, 0.1452, 0.0699], + [-0.0314, -0.0202, -0.2077, ..., -0.3307, 0.1246, -0.2186], + [ 0.0148, -0.0258, -0.1037, ..., 0.0467, -0.1738, 0.0282]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.8626e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + -7.4506e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -9.3132e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 1.8626e-09, + -4.6566e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.8626e-09, + 0.0000e+00, -9.3132e-09]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0210, 0.0093, 0.0187, 0.0031, 0.0322, -0.0127, -0.0051, -0.0133, + -0.0268, 0.0120], device='cuda:0'), grad: tensor([ 1.0245e-08, 0.0000e+00, 7.4506e-09, -1.1176e-07, 2.0489e-08, + 4.7497e-08, -5.7742e-08, 2.7940e-08, 5.9605e-08, -1.2107e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 220.72, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4712 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.07 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.0019, 0.0110, 0.0549, ..., 0.0854, -0.2730, -0.1789], + [ 0.0576, -0.0185, -0.0462, ..., -0.3194, 0.0481, -0.1016], + [-0.0107, -0.0048, 0.2410, ..., -0.3368, -0.1501, -0.1778], + ..., + [-0.0240, 0.0062, -0.2177, ..., -0.0679, 0.1450, 0.0693], + [-0.0314, -0.0202, -0.2078, ..., -0.3307, 0.1248, -0.2191], + [ 0.0148, -0.0258, -0.1038, ..., 0.0467, -0.1746, 0.0284]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.7987e-08, ..., 0.0000e+00, + 2.7940e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, + 8.9407e-08, 6.6124e-08], + [ 0.0000e+00, 0.0000e+00, -1.2666e-07, ..., 0.0000e+00, + 2.3283e-08, 1.3970e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + -1.6764e-07, -1.0524e-07], + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 0.0000e+00, + 4.4703e-08, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 4.6566e-09, 3.7253e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0208, 0.0095, 0.0188, 0.0031, 0.0322, -0.0137, -0.0038, -0.0138, + -0.0267, 0.0121], device='cuda:0'), grad: tensor([ 1.1455e-07, 1.7695e-07, -1.4715e-07, 3.9116e-08, -2.4214e-08, + -1.8626e-09, 1.8626e-09, -3.1386e-07, 1.4529e-07, 1.5832e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 220.85, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4852 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.05 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.0019, 0.0110, 0.0552, ..., 0.0855, -0.2733, -0.1790], + [ 0.0576, -0.0185, -0.0464, ..., -0.3200, 0.0482, -0.1014], + [-0.0107, -0.0048, 0.2412, ..., -0.3369, -0.1503, -0.1787], + ..., + [-0.0240, 0.0062, -0.2182, ..., -0.0681, 0.1449, 0.0687], + [-0.0314, -0.0202, -0.2079, ..., -0.3307, 0.1252, -0.2195], + [ 0.0148, -0.0258, -0.1039, ..., 0.0465, -0.1742, 0.0285]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.1176e-08, ..., -8.3819e-09, + 4.8429e-08, 5.5879e-09], + [ 1.1176e-08, 0.0000e+00, 3.7253e-09, ..., 3.0734e-08, + 5.6438e-07, 4.1910e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-08, ..., 2.7940e-09, + 6.5193e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 1.5832e-08, + -1.1176e-08, 8.3819e-09], + [ 9.3132e-10, 0.0000e+00, 1.8626e-08, ..., 9.3132e-10, + 3.9116e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.3469e-07, + 5.5879e-09, 2.5053e-07]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0205, 0.0100, 0.0183, 0.0031, 0.0324, -0.0137, -0.0039, -0.0143, + -0.0265, 0.0122], device='cuda:0'), grad: tensor([ 2.9150e-07, 2.3805e-06, 2.3283e-07, -2.7940e-09, -5.9325e-07, + 8.7637e-07, -3.7774e-06, 3.8184e-08, -1.8533e-07, 7.4133e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 221.49, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4704 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.05 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.0019, 0.0110, 0.0552, ..., 0.0851, -0.2737, -0.1798], + [ 0.0576, -0.0185, -0.0464, ..., -0.3202, 0.0483, -0.1016], + [-0.0107, -0.0048, 0.2415, ..., -0.3369, -0.1504, -0.1786], + ..., + [-0.0240, 0.0062, -0.2185, ..., -0.0681, 0.1450, 0.0689], + [-0.0314, -0.0202, -0.2081, ..., -0.3308, 0.1252, -0.2200], + [ 0.0148, -0.0258, -0.1044, ..., 0.0464, -0.1759, 0.0281]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 8.4750e-08, 5.6811e-08], + [ 0.0000e+00, 0.0000e+00, -3.0734e-07, ..., 0.0000e+00, + 4.0978e-08, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8161e-07, ..., 9.3132e-10, + -1.2619e-06, -4.9733e-07], + [ 0.0000e+00, 0.0000e+00, 1.0245e-07, ..., 9.3132e-10, + 1.0487e-06, 3.9767e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 2.4214e-08, 1.3039e-08]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0209, 0.0101, 0.0183, 0.0029, 0.0328, -0.0135, -0.0037, -0.0143, + -0.0265, 0.0119], device='cuda:0'), grad: tensor([ 4.5635e-08, 1.1362e-07, -4.0233e-07, 7.8231e-08, 6.4261e-08, + 7.4506e-09, -8.1956e-08, -2.0489e-06, 2.1756e-06, 5.0291e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 220.70, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4775 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.06 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.0019, 0.0110, 0.0552, ..., 0.0851, -0.2739, -0.1798], + [ 0.0577, -0.0185, -0.0464, ..., -0.3204, 0.0482, -0.1017], + [-0.0107, -0.0048, 0.2414, ..., -0.3370, -0.1505, -0.1789], + ..., + [-0.0240, 0.0062, -0.2187, ..., -0.0682, 0.1452, 0.0691], + [-0.0315, -0.0202, -0.2084, ..., -0.3309, 0.1251, -0.2211], + [ 0.0148, -0.0258, -0.1046, ..., 0.0462, -0.1766, 0.0281]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 1.8626e-09, + 2.7940e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, + 1.0245e-08, 3.2596e-08], + [ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 2.5146e-08, + 1.1176e-08, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 3.7253e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.2596e-08, + 1.2107e-08, 1.4901e-08]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0209, 0.0100, 0.0181, 0.0033, 0.0328, -0.0136, -0.0035, -0.0142, + -0.0269, 0.0119], device='cuda:0'), grad: tensor([ 2.4214e-08, 5.9605e-08, -1.2107e-08, -1.2107e-08, -1.8161e-07, + 8.3819e-09, -3.7253e-09, 8.2888e-08, 2.8871e-08, 2.4214e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 220.93, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4815 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.02 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.0019, 0.0110, 0.0554, ..., 0.0852, -0.2747, -0.1801], + [ 0.0578, -0.0185, -0.0464, ..., -0.3207, 0.0484, -0.1018], + [-0.0107, -0.0048, 0.2416, ..., -0.3371, -0.1508, -0.1792], + ..., + [-0.0240, 0.0062, -0.2189, ..., -0.0684, 0.1453, 0.0691], + [-0.0315, -0.0202, -0.2085, ..., -0.3310, 0.1252, -0.2214], + [ 0.0148, -0.0258, -0.1055, ..., 0.0461, -0.1770, 0.0281]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -6.5193e-09, -3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + 9.3132e-10, -3.7253e-09]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0211, 0.0103, 0.0179, 0.0031, 0.0328, -0.0134, -0.0028, -0.0143, + -0.0270, 0.0119], device='cuda:0'), grad: tensor([ 2.2352e-08, 1.2107e-08, -4.6566e-09, 4.6566e-09, 3.7253e-09, + 4.6566e-09, -2.7008e-08, -1.4901e-08, 1.2107e-08, -4.6566e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 221.12, cls_loss 0.0008 cls_loss_mapping 0.0009 cls_loss_causal 0.4574 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.09 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.0019, 0.0110, 0.0554, ..., 0.0853, -0.2750, -0.1801], + [ 0.0579, -0.0185, -0.0464, ..., -0.3210, 0.0481, -0.1018], + [-0.0107, -0.0048, 0.2418, ..., -0.3371, -0.1509, -0.1796], + ..., + [-0.0240, 0.0062, -0.2192, ..., -0.0687, 0.1456, 0.0689], + [-0.0316, -0.0202, -0.2087, ..., -0.3310, 0.1252, -0.2219], + [ 0.0148, -0.0258, -0.1059, ..., 0.0459, -0.1767, 0.0280]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 4.6566e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-09, + 6.6124e-08, 3.0641e-07], + [ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 2.7940e-09, + 1.8626e-08, 2.0489e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.8626e-09, + -1.3039e-07, 9.8068e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 4.2841e-08, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.5146e-08, + 1.3970e-08, -1.3066e-06]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0211, 0.0102, 0.0180, 0.0031, 0.0330, -0.0126, -0.0029, -0.0143, + -0.0273, 0.0117], device='cuda:0'), grad: tensor([ 2.7008e-08, 5.5972e-07, 5.9605e-08, 8.8476e-08, -2.7940e-08, + 3.9823e-06, -4.1798e-06, 1.1940e-06, 1.9092e-07, -1.8952e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 221.05, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4807 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.0019, 0.0110, 0.0553, ..., 0.0853, -0.2747, -0.1803], + [ 0.0581, -0.0185, -0.0466, ..., -0.3213, 0.0475, -0.1017], + [-0.0108, -0.0048, 0.2423, ..., -0.3372, -0.1511, -0.1799], + ..., + [-0.0240, 0.0062, -0.2195, ..., -0.0696, 0.1464, 0.0685], + [-0.0316, -0.0202, -0.2089, ..., -0.3311, 0.1253, -0.2220], + [ 0.0148, -0.0258, -0.1061, ..., 0.0458, -0.1779, 0.0278]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -7.6368e-08, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.4063e-07, ..., 0.0000e+00, + -2.4214e-08, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-07, ..., 0.0000e+00, + 5.2154e-08, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0211, 0.0100, 0.0180, 0.0029, 0.0334, -0.0129, -0.0028, -0.0142, + -0.0274, 0.0115], device='cuda:0'), grad: tensor([ 1.3039e-08, -1.5181e-07, -3.6880e-07, 9.3132e-10, 6.8918e-08, + 8.3819e-09, -3.7253e-08, 4.2282e-07, 4.5635e-08, 3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 220.78, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4813 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.16 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.0019, 0.0110, 0.0555, ..., 0.0854, -0.2754, -0.1804], + [ 0.0581, -0.0185, -0.0466, ..., -0.3215, 0.0479, -0.1014], + [-0.0105, -0.0048, 0.2423, ..., -0.3372, -0.1515, -0.1823], + ..., + [-0.0240, 0.0062, -0.2194, ..., -0.0698, 0.1463, 0.0681], + [-0.0317, -0.0202, -0.2090, ..., -0.3312, 0.1244, -0.2227], + [ 0.0148, -0.0258, -0.1066, ..., 0.0457, -0.1782, 0.0279]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., -9.3132e-10, + 8.4750e-08, 5.1223e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.6154e-06, -2.1495e-06], + [ 0.0000e+00, 0.0000e+00, -3.3528e-08, ..., 0.0000e+00, + 4.6566e-08, 2.8871e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 3.5390e-08, 1.4901e-08], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 3.3453e-06, 1.9912e-06], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, + 8.4750e-08, 5.7742e-08]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0212, 0.0102, 0.0177, 0.0028, 0.0335, -0.0131, -0.0022, -0.0141, + -0.0280, 0.0116], device='cuda:0'), grad: tensor([ 4.3586e-07, -1.7881e-05, 1.8533e-07, 1.1176e-08, 1.7695e-08, + -6.5193e-09, 2.7008e-08, 1.6950e-07, 1.6570e-05, 4.4890e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 220.78, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4762 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.05 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.0019, 0.0110, 0.0559, ..., 0.0862, -0.2759, -0.1804], + [ 0.0582, -0.0185, -0.0463, ..., -0.3218, 0.0495, -0.0999], + [-0.0106, -0.0048, 0.2423, ..., -0.3373, -0.1521, -0.1825], + ..., + [-0.0240, 0.0062, -0.2199, ..., -0.0699, 0.1464, 0.0679], + [-0.0317, -0.0202, -0.2091, ..., -0.3312, 0.1214, -0.2252], + [ 0.0148, -0.0258, -0.1068, ..., 0.0450, -0.1787, 0.0276]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 1.3039e-08, 4.7497e-08], + [ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., 0.0000e+00, + 9.3132e-09, 3.7253e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 9.3132e-10, + -2.8871e-08, -1.0524e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2107e-08, + 1.8626e-09, -3.0734e-08]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0198, 0.0128, 0.0171, 0.0026, 0.0331, -0.0134, -0.0027, -0.0143, + -0.0310, 0.0106], device='cuda:0'), grad: tensor([ 2.5146e-08, 1.3132e-07, 6.5193e-08, 2.1420e-08, 3.5390e-08, + 3.1665e-08, -2.7940e-09, -2.5518e-07, -1.1176e-08, -3.8184e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 220.70, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4568 re_mapping 0.0041 re_causal 0.0112 /// teacc 99.05 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.0019, 0.0110, 0.0559, ..., 0.0862, -0.2762, -0.1805], + [ 0.0582, -0.0185, -0.0466, ..., -0.3220, 0.0491, -0.1003], + [-0.0106, -0.0048, 0.2429, ..., -0.3373, -0.1522, -0.1830], + ..., + [-0.0240, 0.0062, -0.2206, ..., -0.0692, 0.1467, 0.0694], + [-0.0317, -0.0202, -0.2091, ..., -0.3312, 0.1218, -0.2261], + [ 0.0148, -0.0258, -0.1073, ..., 0.0456, -0.1793, 0.0288]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 2.7940e-09, + 0.0000e+00, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-08, ..., 7.4506e-09, + -2.7008e-08, -2.3376e-07], + [ 0.0000e+00, 0.0000e+00, -9.0804e-07, ..., 2.7940e-09, + 9.3132e-10, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 8.5589e-07, ..., 7.4506e-09, + 5.5879e-09, 7.1712e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 4.6566e-09, + 2.7940e-09, 3.9116e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 2.9523e-07, + 1.1176e-08, 6.4727e-07]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0199, 0.0124, 0.0174, 0.0043, 0.0316, -0.0152, -0.0015, -0.0143, + -0.0306, 0.0113], device='cuda:0'), grad: tensor([ 1.3504e-07, -4.3586e-07, -1.3215e-06, 9.9242e-06, -5.7649e-07, + -1.0185e-05, 2.2352e-08, 1.4380e-06, 1.0245e-07, 9.1642e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 220.73, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4361 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.10 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.0019, 0.0110, 0.0560, ..., 0.0862, -0.2767, -0.1807], + [ 0.0582, -0.0185, -0.0465, ..., -0.3226, 0.0490, -0.1007], + [-0.0106, -0.0048, 0.2432, ..., -0.3376, -0.1524, -0.1845], + ..., + [-0.0240, 0.0062, -0.2212, ..., -0.0691, 0.1472, 0.0715], + [-0.0317, -0.0202, -0.2092, ..., -0.3313, 0.1220, -0.2268], + [ 0.0148, -0.0258, -0.1073, ..., 0.0459, -0.1820, 0.0289]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 1.3970e-09, + 4.6566e-10, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.3283e-09, + 4.6566e-09, 1.9092e-08], + [ 0.0000e+00, 0.0000e+00, -1.2573e-08, ..., 9.3132e-10, + 9.3132e-10, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3504e-08, ..., 2.3283e-08, + -7.9162e-09, 1.5832e-07], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 9.3132e-10, + -1.3970e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -3.5809e-07, + 9.3132e-10, -1.5758e-06]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0200, 0.0122, 0.0173, 0.0037, 0.0308, -0.0145, -0.0015, -0.0135, + -0.0305, 0.0111], device='cuda:0'), grad: tensor([ 3.6322e-08, 3.8650e-08, -3.2596e-09, 3.7719e-08, 1.6829e-06, + -1.9046e-07, 1.3970e-09, 1.6624e-07, 1.7695e-08, -1.7677e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 221.16, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4843 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.10 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.0019, 0.0110, 0.0567, ..., 0.0865, -0.2775, -0.1800], + [ 0.0582, -0.0185, -0.0472, ..., -0.3229, 0.0491, -0.1005], + [-0.0106, -0.0048, 0.2441, ..., -0.3377, -0.1525, -0.1844], + ..., + [-0.0240, 0.0062, -0.2226, ..., -0.0692, 0.1473, 0.0714], + [-0.0317, -0.0202, -0.2095, ..., -0.3314, 0.1221, -0.2268], + [ 0.0148, -0.0258, -0.1099, ..., 0.0458, -0.1830, 0.0287]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 4.6566e-10, 1.8161e-08], + [ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 7.0781e-08, ..., 1.3970e-09, + 0.0000e+00, 1.7229e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + -1.3970e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -3.2596e-09, + 9.3132e-10, -3.6322e-08]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0197, 0.0121, 0.0176, 0.0051, 0.0308, -0.0158, -0.0007, -0.0138, + -0.0304, 0.0107], device='cuda:0'), grad: tensor([ 1.8161e-08, 1.0012e-07, 2.3283e-09, -1.0906e-06, 1.3970e-08, + 5.8673e-08, -1.0245e-08, 7.6974e-07, 6.5193e-09, 1.4063e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 220.90, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4560 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.03 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.0019, 0.0110, 0.0593, ..., 0.0886, -0.2780, -0.1776], + [ 0.0582, -0.0185, -0.0475, ..., -0.3235, 0.0489, -0.1013], + [-0.0106, -0.0048, 0.2450, ..., -0.3381, -0.1526, -0.1843], + ..., + [-0.0240, 0.0062, -0.2240, ..., -0.0693, 0.1477, 0.0721], + [-0.0317, -0.0202, -0.2096, ..., -0.3315, 0.1220, -0.2271], + [ 0.0148, -0.0258, -0.1134, ..., 0.0439, -0.1849, 0.0274]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 1.3970e-09, 1.3504e-08], + [ 0.0000e+00, 0.0000e+00, -3.1199e-08, ..., 0.0000e+00, + 2.3283e-09, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 1.3970e-09, + -4.1910e-09, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + -1.3039e-08, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.1223e-09, + 9.3132e-10, -4.4238e-08]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0175, 0.0116, 0.0178, 0.0058, 0.0314, -0.0164, -0.0006, -0.0135, + -0.0305, 0.0086], device='cuda:0'), grad: tensor([ 7.4506e-09, 2.5611e-08, -4.0513e-08, 4.9360e-08, 2.3283e-08, + 2.1420e-08, 6.9849e-09, 4.0513e-08, -5.9605e-08, -6.4261e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 220.95, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4482 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.03 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.0019, 0.0110, 0.0593, ..., 0.0886, -0.2790, -0.1777], + [ 0.0582, -0.0185, -0.0476, ..., -0.3239, 0.0491, -0.1024], + [-0.0106, -0.0048, 0.2462, ..., -0.3381, -0.1528, -0.1815], + ..., + [-0.0240, 0.0062, -0.2261, ..., -0.0695, 0.1478, 0.0722], + [-0.0317, -0.0202, -0.2098, ..., -0.3314, 0.1217, -0.2287], + [ 0.0148, -0.0258, -0.1135, ..., 0.0435, -0.1854, 0.0275]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 1.8626e-09, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 4.1910e-09, 6.9849e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3970e-08, + -5.5879e-09, 2.0955e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.2352e-08, + 0.0000e+00, -4.5169e-08]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0178, 0.0114, 0.0181, 0.0064, 0.0316, -0.0167, -0.0001, -0.0138, + -0.0308, 0.0089], device='cuda:0'), grad: tensor([ 8.3819e-09, 1.5832e-08, 3.3528e-08, -5.0850e-07, 2.1420e-08, + 4.6613e-07, 1.3970e-08, 4.0047e-08, 3.3993e-08, -1.1828e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 220.71, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4521 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.02 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.0019, 0.0110, 0.0594, ..., 0.0886, -0.2792, -0.1778], + [ 0.0582, -0.0185, -0.0476, ..., -0.3241, 0.0485, -0.1031], + [-0.0106, -0.0048, 0.2462, ..., -0.3382, -0.1529, -0.1816], + ..., + [-0.0240, 0.0062, -0.2264, ..., -0.0697, 0.1484, 0.0727], + [-0.0317, -0.0202, -0.2098, ..., -0.3315, 0.1221, -0.2287], + [ 0.0148, -0.0258, -0.1135, ..., 0.0428, -0.1855, 0.0264]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3738e-07, ..., -3.4692e-07, + 0.0000e+00, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 7.0781e-08, ..., 6.9849e-09, + 1.3970e-09, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, -5.5358e-06, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-08, ..., 7.9162e-09, + -1.8626e-09, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 1.6298e-08, ..., 1.3970e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 7.9162e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0177, 0.0111, 0.0180, 0.0064, 0.0326, -0.0166, -0.0004, -0.0137, + -0.0305, 0.0082], device='cuda:0'), grad: tensor([-6.2920e-06, 1.7090e-07, -7.6070e-06, 7.4618e-06, -2.1886e-08, + 6.3330e-08, 5.9605e-06, 1.5367e-07, 4.2375e-08, 7.0315e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 220.81, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4857 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.08 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.0019, 0.0110, 0.0596, ..., 0.0887, -0.2794, -0.1778], + [ 0.0582, -0.0185, -0.0477, ..., -0.3259, 0.0485, -0.1035], + [-0.0106, -0.0048, 0.2467, ..., -0.3384, -0.1530, -0.1819], + ..., + [-0.0240, 0.0062, -0.2267, ..., -0.0699, 0.1484, 0.0726], + [-0.0317, -0.0202, -0.2100, ..., -0.3316, 0.1222, -0.2289], + [ 0.0148, -0.0258, -0.1136, ..., 0.0429, -0.1858, 0.0267]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., -2.7940e-09, + 3.2596e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 7.5437e-08, 1.6997e-07], + [ 0.0000e+00, 0.0000e+00, -7.8231e-08, ..., -1.3039e-08, + 6.0536e-09, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 8.7079e-08, ..., 1.3039e-08, + -3.7998e-07, -8.4704e-07], + [ 0.0000e+00, 0.0000e+00, 4.9826e-08, ..., 0.0000e+00, + 6.5193e-09, 1.1642e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 1.3970e-09, + 2.8871e-07, 6.4494e-07]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0175, 0.0110, 0.0181, 0.0059, 0.0326, -0.0163, -0.0003, -0.0138, + -0.0306, 0.0083], device='cuda:0'), grad: tensor([ 3.5390e-08, 3.1060e-07, -1.9884e-07, -2.0443e-07, 3.4925e-08, + 9.1735e-08, -9.5461e-08, -1.2312e-06, 1.4016e-07, 1.1250e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 220.49, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4744 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.07 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.0019, 0.0110, 0.0596, ..., 0.0887, -0.2798, -0.1780], + [ 0.0582, -0.0185, -0.0476, ..., -0.3265, 0.0480, -0.1040], + [-0.0106, -0.0048, 0.2467, ..., -0.3385, -0.1533, -0.1833], + ..., + [-0.0240, 0.0062, -0.2267, ..., -0.0701, 0.1489, 0.0720], + [-0.0317, -0.0202, -0.2100, ..., -0.3316, 0.1222, -0.2292], + [ 0.0148, -0.0258, -0.1137, ..., 0.0397, -0.1861, 0.0235]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., -1.8626e-08, + 3.2596e-09, 7.9162e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 6.9849e-09, + 2.8545e-07, 6.1933e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 1.4435e-08, 3.7719e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + -4.1910e-08, -2.1979e-07], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + -3.1525e-07, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 1.6764e-08, 8.8476e-08]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0177, 0.0111, 0.0177, 0.0061, 0.0359, -0.0166, 0.0004, -0.0140, + -0.0307, 0.0052], device='cuda:0'), grad: tensor([ 8.8476e-09, 9.8348e-07, 1.2433e-07, 1.0151e-07, -9.3132e-10, + -2.1281e-07, 1.8347e-07, -5.5647e-07, -8.7870e-07, 2.5332e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 220.48, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4757 re_mapping 0.0039 re_causal 0.0112 /// teacc 99.00 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.0019, 0.0110, 0.0596, ..., 0.0886, -0.2802, -0.1783], + [ 0.0582, -0.0185, -0.0477, ..., -0.3268, 0.0476, -0.1047], + [-0.0106, -0.0048, 0.2478, ..., -0.3386, -0.1536, -0.1823], + ..., + [-0.0240, 0.0062, -0.2283, ..., -0.0702, 0.1496, 0.0712], + [-0.0317, -0.0202, -0.2102, ..., -0.3317, 0.1220, -0.2295], + [ 0.0148, -0.0258, -0.1137, ..., 0.0397, -0.1860, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.9873e-08, ..., -2.4214e-08, + 2.3283e-09, -6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.3970e-09, + -2.9337e-08, 7.9162e-09], + [ 0.0000e+00, 0.0000e+00, -2.3097e-07, ..., 1.8626e-09, + 3.2596e-09, -2.1420e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 2.7940e-09, + 1.7229e-08, -4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 4.6566e-10, + 3.2596e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.5832e-08, ..., 9.7789e-09, + 1.8626e-09, -3.2596e-09]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0182, 0.0104, 0.0179, 0.0062, 0.0364, -0.0170, -0.0017, -0.0139, + -0.0309, 0.0054], device='cuda:0'), grad: tensor([ 2.1253e-06, -8.1956e-08, -3.7067e-07, 2.6897e-06, 4.6846e-07, + -2.1104e-06, -2.9467e-06, 1.2061e-07, 4.1910e-08, 4.7497e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 220.69, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4820 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.08 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.0019, 0.0110, 0.0599, ..., 0.0887, -0.2807, -0.1785], + [ 0.0582, -0.0185, -0.0478, ..., -0.3274, 0.0474, -0.1050], + [-0.0106, -0.0048, 0.2484, ..., -0.3389, -0.1537, -0.1820], + ..., + [-0.0240, 0.0062, -0.2292, ..., -0.0721, 0.1499, 0.0709], + [-0.0318, -0.0202, -0.2101, ..., -0.3317, 0.1223, -0.2298], + [ 0.0148, -0.0258, -0.1141, ..., 0.0397, -0.1861, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 5.1223e-09, + -9.3132e-10, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 8.8476e-09]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0184, 0.0102, 0.0181, 0.0059, 0.0364, -0.0144, -0.0016, -0.0146, + -0.0306, 0.0054], device='cuda:0'), grad: tensor([ 3.2596e-09, 2.7940e-09, -3.0268e-08, 6.9849e-09, -1.5832e-08, + 9.3132e-10, -9.3132e-10, 1.9092e-08, 1.2107e-08, 1.1642e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 220.74, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4762 re_mapping 0.0038 re_causal 0.0114 /// teacc 98.97 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.0019, 0.0110, 0.0597, ..., 0.0888, -0.2812, -0.1786], + [ 0.0582, -0.0185, -0.0478, ..., -0.3276, 0.0473, -0.1052], + [-0.0106, -0.0048, 0.2487, ..., -0.3390, -0.1538, -0.1822], + ..., + [-0.0240, 0.0062, -0.2294, ..., -0.0724, 0.1502, 0.0711], + [-0.0318, -0.0202, -0.2101, ..., -0.3317, 0.1224, -0.2300], + [ 0.0148, -0.0258, -0.1142, ..., 0.0397, -0.1863, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.8790e-07, ..., -4.6566e-09, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 4.6566e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -1.0123e-06, ..., 1.3970e-09, + 0.0000e+00, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 8.2888e-08, ..., 9.3132e-10, + 4.6566e-10, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 2.2817e-08, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 4.6566e-10, + 0.0000e+00, -9.3132e-09]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0187, 0.0101, 0.0182, 0.0058, 0.0363, -0.0141, -0.0015, -0.0146, + -0.0305, 0.0055], device='cuda:0'), grad: tensor([ 5.6019e-07, 1.6764e-08, -1.4808e-06, 7.1106e-07, 2.1886e-08, + 2.3283e-09, 9.7789e-09, 1.4575e-07, 3.4459e-08, -2.7940e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 220.59, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4842 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.08 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.0019, 0.0110, 0.0599, ..., 0.0889, -0.2815, -0.1787], + [ 0.0582, -0.0185, -0.0479, ..., -0.3277, 0.0471, -0.1055], + [-0.0106, -0.0048, 0.2488, ..., -0.3392, -0.1540, -0.1824], + ..., + [-0.0240, 0.0062, -0.2295, ..., -0.0725, 0.1509, 0.0710], + [-0.0318, -0.0202, -0.2102, ..., -0.3318, 0.1220, -0.2309], + [ 0.0148, -0.0258, -0.1145, ..., 0.0397, -0.1873, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-09, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.3970e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -5.1223e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 6.0536e-09, ..., 4.6566e-10, + 4.6566e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 1.8626e-09, 5.5879e-09]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0186, 0.0098, 0.0181, 0.0055, 0.0363, -0.0139, -0.0017, -0.0143, + -0.0307, 0.0055], device='cuda:0'), grad: tensor([-1.3039e-08, 3.7253e-09, -5.5879e-09, 1.1176e-08, -5.3551e-08, + 1.8626e-09, 1.5367e-08, 9.3132e-10, 1.9092e-08, 2.1420e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 220.68, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4729 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.05 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.0019, 0.0110, 0.0599, ..., 0.0889, -0.2818, -0.1787], + [ 0.0582, -0.0185, -0.0474, ..., -0.3281, 0.0476, -0.1054], + [-0.0106, -0.0048, 0.2491, ..., -0.3392, -0.1542, -0.1825], + ..., + [-0.0240, 0.0062, -0.2300, ..., -0.0727, 0.1507, 0.0707], + [-0.0318, -0.0202, -0.2107, ..., -0.3319, 0.1220, -0.2315], + [ 0.0148, -0.0258, -0.1148, ..., 0.0397, -0.1885, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 3.3062e-08, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -3.3993e-08, 4.9826e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 4.1910e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.6298e-08, + -6.0536e-09, 8.9873e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.1933e-08, + 4.6566e-09, -4.1677e-07]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0187, 0.0102, 0.0181, 0.0056, 0.0363, -0.0139, -0.0018, -0.0144, + -0.0309, 0.0056], device='cuda:0'), grad: tensor([ 1.6484e-07, -6.1002e-08, 1.8161e-08, 7.4506e-09, 4.3726e-07, + 3.1665e-08, -1.3970e-08, 1.7183e-07, 1.8626e-08, -7.6275e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 220.44, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4508 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.11 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.0019, 0.0110, 0.0600, ..., 0.0890, -0.2821, -0.1789], + [ 0.0582, -0.0185, -0.0487, ..., -0.3283, 0.0473, -0.1058], + [-0.0106, -0.0048, 0.2508, ..., -0.3395, -0.1542, -0.1828], + ..., + [-0.0240, 0.0062, -0.2312, ..., -0.0728, 0.1495, 0.0704], + [-0.0318, -0.0202, -0.2110, ..., -0.3321, 0.1248, -0.2285], + [ 0.0148, -0.0258, -0.1149, ..., 0.0397, -0.1898, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + -9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, -3.3528e-08, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 3.7253e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 5.5879e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -9.3132e-10, + 0.0000e+00, -9.3132e-09]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0187, 0.0096, 0.0188, 0.0059, 0.0362, -0.0142, -0.0016, -0.0152, + -0.0293, 0.0056], device='cuda:0'), grad: tensor([ 1.0245e-08, 4.6566e-09, -6.8918e-08, -6.7987e-08, 7.4506e-09, + -1.1176e-08, 9.3132e-09, 5.6811e-08, 7.1712e-08, -3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 220.59, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4526 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.12 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.0019, 0.0110, 0.0600, ..., 0.0891, -0.2826, -0.1789], + [ 0.0582, -0.0185, -0.0487, ..., -0.3284, 0.0475, -0.1057], + [-0.0106, -0.0048, 0.2511, ..., -0.3395, -0.1547, -0.1831], + ..., + [-0.0240, 0.0062, -0.2318, ..., -0.0730, 0.1496, 0.0703], + [-0.0318, -0.0202, -0.2113, ..., -0.3322, 0.1250, -0.2287], + [ 0.0148, -0.0258, -0.1151, ..., 0.0397, -0.1902, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.0023e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.6578e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0187, 0.0099, 0.0187, 0.0068, 0.0363, -0.0142, -0.0019, -0.0155, + -0.0292, 0.0056], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.5832e-08, -3.0175e-07, -4.6566e-09, 1.8626e-09, + 9.3132e-09, 2.5146e-08, 1.1176e-08, 2.4773e-07, 3.7253e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 220.38, cls_loss 0.0009 cls_loss_mapping 0.0010 cls_loss_causal 0.4775 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.09 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.0019, 0.0110, 0.0598, ..., 0.0892, -0.2833, -0.1790], + [ 0.0582, -0.0185, -0.0489, ..., -0.3287, 0.0477, -0.1060], + [-0.0106, -0.0048, 0.2518, ..., -0.3396, -0.1549, -0.1837], + ..., + [-0.0240, 0.0062, -0.2320, ..., -0.0731, 0.1498, 0.0711], + [-0.0318, -0.0202, -0.2119, ..., -0.3323, 0.1266, -0.2292], + [ 0.0148, -0.0258, -0.1154, ..., 0.0397, -0.1911, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8871e-08, ..., -2.7940e-09, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + -1.8626e-07, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, -7.0781e-08, ..., 0.0000e+00, + 2.7940e-09, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 0.0000e+00, + 1.8068e-07, -2.5146e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + 0.0000e+00, -6.5193e-09]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0189, 0.0099, 0.0188, 0.0056, 0.0363, -0.0160, -0.0020, -0.0151, + -0.0279, 0.0056], device='cuda:0'), grad: tensor([-1.2666e-07, -5.4296e-07, -3.2596e-08, 2.2352e-08, 1.0245e-08, + -2.5053e-07, 3.7625e-07, 5.0943e-07, 4.4703e-08, 0.0000e+00], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 220.66, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4622 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.01 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.0019, 0.0110, 0.0598, ..., 0.0893, -0.2843, -0.1791], + [ 0.0583, -0.0185, -0.0489, ..., -0.3289, 0.0478, -0.1058], + [-0.0106, -0.0048, 0.2527, ..., -0.3396, -0.1551, -0.1848], + ..., + [-0.0240, 0.0062, -0.2330, ..., -0.0735, 0.1503, 0.0715], + [-0.0318, -0.0202, -0.2122, ..., -0.3323, 0.1272, -0.2294], + [ 0.0148, -0.0258, -0.1155, ..., 0.0397, -0.1924, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -9.3132e-10, + 0.0000e+00, -1.8626e-09]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0189, 0.0100, 0.0191, 0.0057, 0.0363, -0.0170, -0.0032, -0.0147, + -0.0271, 0.0055], device='cuda:0'), grad: tensor([ 4.0047e-08, -1.8626e-08, 1.8626e-09, -1.8626e-09, 3.9116e-07, + -2.7940e-09, -4.3586e-07, 1.9558e-08, 1.3039e-08, -1.8626e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 220.92, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4617 re_mapping 0.0036 re_causal 0.0108 /// teacc 98.97 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.0019, 0.0110, 0.0598, ..., 0.0893, -0.2847, -0.1792], + [ 0.0584, -0.0185, -0.0490, ..., -0.3290, 0.0479, -0.1058], + [-0.0106, -0.0048, 0.2547, ..., -0.3397, -0.1552, -0.1849], + ..., + [-0.0240, 0.0062, -0.2337, ..., -0.0736, 0.1504, 0.0715], + [-0.0319, -0.0202, -0.2124, ..., -0.3323, 0.1272, -0.2296], + [ 0.0148, -0.0258, -0.1156, ..., 0.0397, -0.1926, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -9.3132e-10, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + -2.7940e-09, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0190, 0.0101, 0.0198, 0.0043, 0.0362, -0.0167, -0.0032, -0.0148, + -0.0271, 0.0055], device='cuda:0'), grad: tensor([ 2.7940e-09, 4.6566e-09, 9.3132e-09, -1.8626e-08, -3.7253e-09, + 1.8626e-09, 9.3132e-10, -9.3132e-09, 4.6566e-09, 6.5193e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 220.41, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4545 re_mapping 0.0039 re_causal 0.0102 /// teacc 98.99 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.0019, 0.0110, 0.0600, ..., 0.0893, -0.2887, -0.1794], + [ 0.0585, -0.0185, -0.0489, ..., -0.3294, 0.0517, -0.1055], + [-0.0106, -0.0048, 0.2559, ..., -0.3398, -0.1580, -0.1841], + ..., + [-0.0240, 0.0062, -0.2353, ..., -0.0735, 0.1465, 0.0712], + [-0.0320, -0.0202, -0.2125, ..., -0.3324, 0.1265, -0.2299], + [ 0.0147, -0.0258, -0.1166, ..., 0.0397, -0.1944, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.8871e-08, ..., 1.8626e-09, + 2.7940e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 1.8626e-08, + 2.7940e-09, 3.3528e-08], + [ 0.0000e+00, 0.0000e+00, -1.5926e-07, ..., 0.0000e+00, + 2.7940e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 6.4261e-08, ..., 1.2107e-08, + -2.2352e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 2.7008e-08, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.7323e-07, + 9.3132e-10, 2.7288e-07]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0198, 0.0140, 0.0188, 0.0041, 0.0362, -0.0165, -0.0033, -0.0188, + -0.0273, 0.0055], device='cuda:0'), grad: tensor([ 1.6391e-07, 7.0781e-08, -2.4401e-07, -1.3318e-07, -4.5169e-07, + 7.4506e-08, -6.3330e-08, 7.9162e-08, 8.8476e-08, 4.1258e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 220.76, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4653 re_mapping 0.0035 re_causal 0.0102 /// teacc 99.05 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.0021, 0.0110, 0.0599, ..., 0.0895, -0.2890, -0.1796], + [ 0.0586, -0.0185, -0.0495, ..., -0.3318, 0.0524, -0.1052], + [-0.0107, -0.0048, 0.2593, ..., -0.3401, -0.1581, -0.1838], + ..., + [-0.0240, 0.0062, -0.2384, ..., -0.0737, 0.1459, 0.0706], + [-0.0325, -0.0202, -0.2128, ..., -0.3327, 0.1272, -0.2311], + [ 0.0147, -0.0258, -0.1170, ..., 0.0397, -0.1948, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.7742e-08, ..., -9.0338e-08, + 0.0000e+00, -6.1467e-08], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 4.9360e-08, + 9.3132e-10, 1.5367e-07], + [ 0.0000e+00, 0.0000e+00, -2.1420e-08, ..., 2.6077e-08, + 0.0000e+00, 5.4017e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 9.3132e-09, + -3.7253e-09, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.5193e-09, + 0.0000e+00, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 4.5635e-08, ..., 6.1002e-07, + 1.8626e-09, 1.8021e-06]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0197, 0.0149, 0.0203, 0.0037, 0.0364, -0.0183, -0.0052, -0.0198, + -0.0264, 0.0057], device='cuda:0'), grad: tensor([-5.1688e-07, 3.7067e-07, 9.4995e-08, 4.3772e-08, -4.6678e-06, + 4.0047e-08, 2.4401e-07, 5.9605e-08, 4.0047e-08, 4.2841e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 220.38, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4590 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.02 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.0021, 0.0110, 0.0603, ..., 0.0896, -0.2892, -0.1798], + [ 0.0586, -0.0185, -0.0497, ..., -0.3321, 0.0524, -0.1054], + [-0.0108, -0.0048, 0.2597, ..., -0.3403, -0.1584, -0.1838], + ..., + [-0.0240, 0.0062, -0.2387, ..., -0.0738, 0.1459, 0.0708], + [-0.0326, -0.0202, -0.2128, ..., -0.3328, 0.1273, -0.2314], + [ 0.0147, -0.0258, -0.1171, ..., 0.0397, -0.1953, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3690e-07, ..., 0.0000e+00, + 1.0617e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.8184e-08, ..., 0.0000e+00, + 4.6566e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -1.1269e-07, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -3.4273e-07, ..., 9.3132e-10, + -3.0920e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., -3.7253e-09, + 6.5193e-09, -2.4214e-08]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0201, 0.0149, 0.0203, 0.0032, 0.0364, -0.0178, -0.0049, -0.0198, + -0.0264, 0.0057], device='cuda:0'), grad: tensor([ 1.1055e-06, 1.0151e-07, -1.3039e-07, 1.5553e-07, 6.7055e-08, + 1.6754e-06, 1.9558e-08, -3.0752e-06, 5.4948e-08, 1.5832e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 220.38, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4792 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.11 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.0021, 0.0110, 0.0612, ..., 0.0890, -0.2893, -0.1807], + [ 0.0588, -0.0185, -0.0501, ..., -0.3333, 0.0524, -0.1059], + [-0.0109, -0.0048, 0.2601, ..., -0.3406, -0.1585, -0.1839], + ..., + [-0.0240, 0.0062, -0.2388, ..., -0.0741, 0.1459, 0.0707], + [-0.0326, -0.0202, -0.2127, ..., -0.3330, 0.1274, -0.2316], + [ 0.0147, -0.0258, -0.1173, ..., 0.0396, -0.1954, 0.0235]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.0245e-07, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3842e-04, ..., 0.0000e+00, + 1.5832e-08, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, -2.3973e-04, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.4459e-08, ..., 0.0000e+00, + -8.9407e-08, -6.9849e-08], + [ 0.0000e+00, 0.0000e+00, 3.5111e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.0582e-07, ..., 9.3132e-10, + 3.7253e-09, 2.7940e-09]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0202, 0.0148, 0.0203, 0.0032, 0.0371, -0.0199, -0.0035, -0.0198, + -0.0264, 0.0052], device='cuda:0'), grad: tensor([ 1.9372e-07, 4.5466e-04, -4.5705e-04, 7.5810e-07, 1.4808e-07, + 8.1025e-08, 4.3772e-08, -2.3562e-07, 6.0257e-07, 4.0699e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 220.45, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4629 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.07 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.0021, 0.0110, 0.0614, ..., 0.0892, -0.2895, -0.1808], + [ 0.0588, -0.0185, -0.0529, ..., -0.3337, 0.0524, -0.1060], + [-0.0109, -0.0048, 0.2629, ..., -0.3406, -0.1586, -0.1841], + ..., + [-0.0240, 0.0062, -0.2391, ..., -0.0744, 0.1459, 0.0707], + [-0.0326, -0.0202, -0.2129, ..., -0.3331, 0.1273, -0.2318], + [ 0.0147, -0.0258, -0.1177, ..., 0.0396, -0.1958, 0.0232]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 1.3690e-07, 5.2154e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 3.7253e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + -1.5087e-07, -5.8673e-08], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-09, 2.7940e-09]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0204, 0.0145, 0.0228, 0.0059, 0.0372, -0.0201, -0.0032, -0.0198, + -0.0268, 0.0051], device='cuda:0'), grad: tensor([ 1.8626e-08, 3.4552e-07, 6.3330e-08, -2.5239e-07, 9.3132e-09, + -8.3819e-08, 3.7253e-09, -3.6228e-07, 1.4715e-07, 1.1362e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 220.65, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4748 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.00 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.0021, 0.0110, 0.0621, ..., 0.0887, -0.2895, -0.1820], + [ 0.0588, -0.0185, -0.0537, ..., -0.3341, 0.0526, -0.1062], + [-0.0109, -0.0048, 0.2637, ..., -0.3408, -0.1587, -0.1843], + ..., + [-0.0240, 0.0062, -0.2392, ..., -0.0745, 0.1457, 0.0707], + [-0.0326, -0.0202, -0.2133, ..., -0.3333, 0.1276, -0.2307], + [ 0.0147, -0.0258, -0.1182, ..., 0.0397, -0.1966, 0.0233]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, -3.7253e-09, ..., 1.8626e-09, + -2.7940e-09, 1.9558e-08], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 6.5193e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -8.3819e-09, -1.8626e-09], + [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.5879e-09, + 0.0000e+00, -3.5390e-08]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0205, 0.0145, 0.0237, 0.0073, 0.0373, -0.0207, -0.0031, -0.0201, + -0.0267, 0.0051], device='cuda:0'), grad: tensor([ 1.6764e-08, 1.2107e-08, 3.1665e-08, -5.5879e-08, 3.0734e-08, + 1.8626e-09, 9.3132e-10, -1.0245e-08, 2.3283e-08, -5.7742e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 220.24, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4773 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.07 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.0021, 0.0110, 0.0624, ..., 0.0889, -0.2896, -0.1821], + [ 0.0589, -0.0185, -0.0538, ..., -0.3344, 0.0528, -0.1066], + [-0.0109, -0.0048, 0.2638, ..., -0.3410, -0.1595, -0.1853], + ..., + [-0.0240, 0.0062, -0.2393, ..., -0.0746, 0.1456, 0.0713], + [-0.0326, -0.0202, -0.2137, ..., -0.3335, 0.1276, -0.2309], + [ 0.0148, -0.0258, -0.1185, ..., 0.0398, -0.1973, 0.0234]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., -1.3039e-08, + 2.7940e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.3970e-08, + -1.1176e-08, 4.0047e-08], + [ 0.0000e+00, 0.0000e+00, 2.8871e-08, ..., 0.0000e+00, + 1.1176e-08, 2.7940e-09], + ..., + [ 9.3132e-10, 0.0000e+00, 4.9360e-08, ..., 9.5926e-08, + 6.5193e-09, 2.1420e-07], + [ 0.0000e+00, 0.0000e+00, 9.6858e-08, ..., 5.5879e-09, + 3.7253e-09, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., -3.9395e-07, + -1.3970e-07, -1.0105e-06]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0203, 0.0148, 0.0234, 0.0052, 0.0372, -0.0187, -0.0031, -0.0203, + -0.0270, 0.0052], device='cuda:0'), grad: tensor([-4.0978e-08, 4.6566e-08, 2.4028e-07, -8.2981e-07, 1.1213e-06, + 1.8254e-07, 7.9162e-08, 3.8464e-07, 3.9488e-07, -1.5832e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 220.51, cls_loss 0.0012 cls_loss_mapping 0.0012 cls_loss_causal 0.4791 re_mapping 0.0038 re_causal 0.0100 /// teacc 99.13 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.0021, 0.0110, 0.0624, ..., 0.0888, -0.2896, -0.1822], + [ 0.0588, -0.0185, -0.0538, ..., -0.3344, 0.0528, -0.1066], + [-0.0111, -0.0048, 0.2638, ..., -0.3410, -0.1595, -0.1855], + ..., + [-0.0240, 0.0062, -0.2393, ..., -0.0746, 0.1456, 0.0714], + [-0.0326, -0.0202, -0.2137, ..., -0.3335, 0.1276, -0.2309], + [ 0.0149, -0.0258, -0.1185, ..., 0.0398, -0.1973, 0.0234]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -7.0781e-08, + 2.7940e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 4.6566e-09, + -1.5367e-07, -3.6322e-08], + [ 0.0000e+00, 0.0000e+00, -1.0896e-07, ..., 2.7940e-09, + -2.7940e-09, 7.4506e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0896e-07, ..., 9.3132e-10, + 2.1420e-08, -3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 6.7055e-08, 2.5146e-08], + [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 5.7742e-08, + 1.3039e-08, 1.3039e-08]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0204, 0.0148, 0.0234, 0.0048, 0.0371, -0.0186, -0.0031, -0.0202, + -0.0270, 0.0053], device='cuda:0'), grad: tensor([-1.5087e-07, -8.4098e-07, -1.9744e-07, -3.2596e-08, 8.8476e-08, + 1.1548e-07, 1.1921e-07, 3.3434e-07, 3.7812e-07, 1.9465e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 220.19, cls_loss 0.0010 cls_loss_mapping 0.0008 cls_loss_causal 0.4371 re_mapping 0.0035 re_causal 0.0098 /// teacc 99.11 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.0021, 0.0110, 0.0625, ..., 0.0888, -0.2896, -0.1822], + [ 0.0589, -0.0185, -0.0538, ..., -0.3345, 0.0528, -0.1067], + [-0.0111, -0.0048, 0.2638, ..., -0.3410, -0.1596, -0.1855], + ..., + [-0.0240, 0.0062, -0.2394, ..., -0.0746, 0.1456, 0.0715], + [-0.0326, -0.0202, -0.2137, ..., -0.3335, 0.1276, -0.2310], + [ 0.0149, -0.0258, -0.1185, ..., 0.0399, -0.1974, 0.0235]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -1.3970e-08, 4.0047e-08], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 3.7253e-09, + 1.8626e-09, 4.1910e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.5390e-08, + 1.8626e-09, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 3.7253e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.6322e-08, + 1.8626e-09, -9.5926e-08]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0204, 0.0147, 0.0233, 0.0047, 0.0370, -0.0186, -0.0030, -0.0202, + -0.0270, 0.0054], device='cuda:0'), grad: tensor([ 9.3132e-10, 1.8626e-08, 1.2107e-07, -7.7300e-08, -2.7940e-08, + 1.1176e-08, 6.5193e-09, 9.0338e-08, 2.5146e-08, -1.6205e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 220.38, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4460 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.11 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.0021, 0.0110, 0.0625, ..., 0.0888, -0.2897, -0.1823], + [ 0.0589, -0.0185, -0.0538, ..., -0.3346, 0.0527, -0.1068], + [-0.0111, -0.0048, 0.2639, ..., -0.3411, -0.1596, -0.1856], + ..., + [-0.0240, 0.0062, -0.2395, ..., -0.0746, 0.1456, 0.0716], + [-0.0327, -0.0202, -0.2137, ..., -0.3335, 0.1276, -0.2311], + [ 0.0149, -0.0258, -0.1185, ..., 0.0399, -0.1975, 0.0236]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 4.0047e-08, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 3.5390e-08, 3.1665e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + -8.1956e-08, -6.7987e-08], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + -8.3819e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0205, 0.0147, 0.0234, 0.0046, 0.0370, -0.0186, -0.0030, -0.0202, + -0.0270, 0.0054], device='cuda:0'), grad: tensor([ 1.3970e-08, 1.3690e-07, 1.6205e-07, -1.3318e-07, -2.4214e-08, + 1.0245e-08, 1.6764e-08, -2.1327e-07, 2.1420e-08, 1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 220.20, cls_loss 0.0010 cls_loss_mapping 0.0007 cls_loss_causal 0.4271 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.11 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.0021, 0.0110, 0.0625, ..., 0.0888, -0.2897, -0.1823], + [ 0.0585, -0.0185, -0.0538, ..., -0.3347, 0.0528, -0.1069], + [-0.0112, -0.0048, 0.2639, ..., -0.3411, -0.1596, -0.1857], + ..., + [-0.0240, 0.0062, -0.2397, ..., -0.0746, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2138, ..., -0.3337, 0.1276, -0.2313], + [ 0.0150, -0.0258, -0.1186, ..., 0.0400, -0.1976, 0.0236]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.4273e-07, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.3039e-08, + -1.6764e-08, 3.5390e-08], + [ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., 2.3283e-08, + 1.8626e-09, 2.6077e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.2107e-08, + 1.3970e-08, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-08, + 0.0000e+00, -6.4261e-08]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0205, 0.0148, 0.0234, 0.0044, 0.0369, -0.0186, -0.0029, -0.0202, + -0.0270, 0.0055], device='cuda:0'), grad: tensor([ 3.7402e-06, 4.0047e-08, 5.5879e-08, 2.0489e-08, -7.5344e-07, + 6.5193e-08, -3.1143e-06, 8.6613e-08, 7.4506e-09, -1.5181e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 220.36, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4305 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.13 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.0021, 0.0110, 0.0626, ..., 0.0888, -0.2897, -0.1824], + [ 0.0585, -0.0185, -0.0538, ..., -0.3348, 0.0528, -0.1069], + [-0.0112, -0.0048, 0.2639, ..., -0.3411, -0.1596, -0.1858], + ..., + [-0.0241, 0.0062, -0.2397, ..., -0.0747, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2138, ..., -0.3338, 0.1276, -0.2315], + [ 0.0150, -0.0258, -0.1186, ..., 0.0400, -0.1977, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 1.9558e-08, 4.3772e-08], + [ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 0.0000e+00, + 4.6566e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -3.3528e-08, -5.4948e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 6.5193e-09]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0205, 0.0148, 0.0234, 0.0043, 0.0368, -0.0186, -0.0029, -0.0203, + -0.0271, 0.0056], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.2480e-07, 1.2107e-08, 2.2817e-07, 7.4506e-09, + -2.4494e-07, 2.3283e-08, -1.9930e-07, 2.6077e-08, 3.0734e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 220.42, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4192 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.13 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.0021, 0.0110, 0.0626, ..., 0.0888, -0.2897, -0.1824], + [ 0.0585, -0.0185, -0.0538, ..., -0.3349, 0.0528, -0.1069], + [-0.0112, -0.0048, 0.2639, ..., -0.3411, -0.1597, -0.1858], + ..., + [-0.0241, 0.0062, -0.2398, ..., -0.0747, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1276, -0.2315], + [ 0.0150, -0.0258, -0.1186, ..., 0.0400, -0.1977, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -2.2352e-08, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 4.2841e-08, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.7695e-08, + 3.7253e-09, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + -7.8231e-08, -5.4017e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 4.6566e-09, + 1.8626e-09, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -9.3132e-10, + 1.0245e-08, -6.5193e-09]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0206, 0.0148, 0.0233, 0.0043, 0.0368, -0.0186, -0.0029, -0.0202, + -0.0270, 0.0056], device='cuda:0'), grad: tensor([ 1.9558e-08, 1.4622e-07, 5.8673e-08, -3.2876e-07, 1.0245e-08, + 1.8999e-07, -8.3819e-09, -2.3842e-07, 1.0524e-07, 4.8429e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 220.19, cls_loss 0.0006 cls_loss_mapping 0.0006 cls_loss_causal 0.4390 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.13 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.0021, 0.0110, 0.0626, ..., 0.0888, -0.2897, -0.1825], + [ 0.0585, -0.0185, -0.0538, ..., -0.3350, 0.0528, -0.1068], + [-0.0112, -0.0048, 0.2639, ..., -0.3411, -0.1597, -0.1859], + ..., + [-0.0241, 0.0062, -0.2398, ..., -0.0747, 0.1456, 0.0715], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1276, -0.2315], + [ 0.0150, -0.0258, -0.1186, ..., 0.0400, -0.1975, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -3.7253e-09, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 4.6566e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, -6.7055e-08, ..., 0.0000e+00, + 1.8626e-09, 2.7940e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.7497e-08, ..., 1.2107e-08, + 1.1176e-08, 1.5646e-07], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.1176e-08, + -2.3283e-08, -1.9465e-07]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0206, 0.0148, 0.0233, 0.0043, 0.0368, -0.0186, -0.0030, -0.0202, + -0.0270, 0.0056], device='cuda:0'), grad: tensor([ 6.5193e-09, 3.5390e-08, -9.4064e-08, 4.6566e-09, 2.8871e-08, + 2.5146e-08, -5.0291e-08, 2.6729e-07, 2.7940e-08, -2.5611e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 220.54, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4533 re_mapping 0.0031 re_causal 0.0104 /// teacc 99.13 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.0021, 0.0110, 0.0627, ..., 0.0888, -0.2897, -0.1825], + [ 0.0585, -0.0185, -0.0538, ..., -0.3351, 0.0528, -0.1069], + [-0.0112, -0.0048, 0.2640, ..., -0.3412, -0.1597, -0.1859], + ..., + [-0.0241, 0.0062, -0.2398, ..., -0.0747, 0.1456, 0.0715], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1277, -0.2315], + [ 0.0151, -0.0258, -0.1186, ..., 0.0400, -0.1975, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., -3.1665e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 6.5193e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -3.2596e-08, ..., 9.3132e-10, + 5.5879e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -1.2107e-08, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 2.6077e-08, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0206, 0.0148, 0.0233, 0.0042, 0.0368, -0.0187, -0.0029, -0.0202, + -0.0270, 0.0056], device='cuda:0'), grad: tensor([-6.3330e-08, 1.9558e-08, -1.0245e-08, 3.0734e-08, 3.5390e-08, + -1.7695e-08, -6.5193e-09, -2.7940e-08, -3.4459e-08, 7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 220.27, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4232 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.13 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.0021, 0.0110, 0.0627, ..., 0.0887, -0.2897, -0.1825], + [ 0.0585, -0.0185, -0.0538, ..., -0.3351, 0.0528, -0.1069], + [-0.0112, -0.0048, 0.2640, ..., -0.3412, -0.1597, -0.1859], + ..., + [-0.0241, 0.0062, -0.2398, ..., -0.0748, 0.1456, 0.0713], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1277, -0.2315], + [ 0.0151, -0.0258, -0.1186, ..., 0.0400, -0.1973, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3504e-08, + 4.6566e-10, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.7276e-08, + 5.5879e-09, 7.1712e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.0245e-08, + 1.8626e-09, 1.3039e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + -6.5193e-09, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + -1.3970e-08, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -2.7940e-09, + 6.9849e-09, -1.3504e-08]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0207, 0.0148, 0.0233, 0.0042, 0.0368, -0.0187, -0.0029, -0.0202, + -0.0270, 0.0057], device='cuda:0'), grad: tensor([ 2.7474e-08, 1.2061e-07, 2.4680e-08, -6.9849e-09, -1.6019e-07, + 1.6764e-08, 2.5146e-08, -9.7789e-09, -2.9802e-08, -2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 220.53, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4443 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.12 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.0021, 0.0110, 0.0627, ..., 0.0888, -0.2897, -0.1826], + [ 0.0585, -0.0185, -0.0538, ..., -0.3351, 0.0528, -0.1069], + [-0.0112, -0.0048, 0.2640, ..., -0.3412, -0.1598, -0.1860], + ..., + [-0.0241, 0.0062, -0.2398, ..., -0.0748, 0.1456, 0.0714], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1277, -0.2315], + [ 0.0151, -0.0258, -0.1186, ..., 0.0400, -0.1973, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1642e-08, ..., -7.4506e-09, + 0.0000e+00, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 1.1642e-08, ..., 2.1886e-08, + 4.6566e-10, 1.8161e-08], + [ 0.0000e+00, 0.0000e+00, -3.4785e-07, ..., 6.0536e-09, + 0.0000e+00, 5.1223e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3760e-07, ..., 9.7789e-09, + 0.0000e+00, 1.1642e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.3970e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.1420e-08, + 0.0000e+00, -1.8626e-09]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0207, 0.0148, 0.0233, 0.0042, 0.0368, -0.0187, -0.0029, -0.0202, + -0.0269, 0.0057], device='cuda:0'), grad: tensor([-2.2352e-08, 6.1467e-08, -6.5472e-07, 1.0710e-08, -8.1025e-08, + -4.1910e-09, 1.4435e-08, 6.6543e-07, 1.4901e-08, 1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 220.34, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4405 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.14 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.0021, 0.0110, 0.0627, ..., 0.0888, -0.2897, -0.1826], + [ 0.0585, -0.0185, -0.0539, ..., -0.3352, 0.0528, -0.1070], + [-0.0112, -0.0048, 0.2640, ..., -0.3412, -0.1598, -0.1860], + ..., + [-0.0241, 0.0062, -0.2398, ..., -0.0748, 0.1456, 0.0714], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1277, -0.2315], + [ 0.0151, -0.0258, -0.1186, ..., 0.0400, -0.1973, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0023e-08, ..., -7.4506e-09, + 4.6566e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 4.6566e-09, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + -3.2596e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, + 1.8626e-09, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -1.9558e-08, + -1.8626e-09, -9.3132e-08]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0207, 0.0148, 0.0232, 0.0042, 0.0368, -0.0187, -0.0029, -0.0202, + -0.0269, 0.0057], device='cuda:0'), grad: tensor([-4.3306e-08, 2.1886e-08, 9.7789e-09, 1.1502e-07, 1.2992e-07, + -1.2340e-07, 4.6566e-09, 5.5879e-09, 2.1886e-08, -1.3039e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 220.78, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4560 re_mapping 0.0030 re_causal 0.0100 /// teacc 99.14 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1826], + [ 0.0585, -0.0185, -0.0538, ..., -0.3353, 0.0528, -0.1070], + [-0.0112, -0.0048, 0.2640, ..., -0.3412, -0.1598, -0.1861], + ..., + [-0.0241, 0.0062, -0.2399, ..., -0.0748, 0.1456, 0.0714], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1277, -0.2316], + [ 0.0151, -0.0258, -0.1187, ..., 0.0400, -0.1973, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + -5.0180e-06, -3.6657e-06], + [ 0.0000e+00, 0.0000e+00, -1.0245e-08, ..., 1.2107e-08, + 2.0023e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 5.3085e-08, ..., 0.0000e+00, + 4.8354e-06, 3.5428e-06], + [ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -1.3504e-08, + -2.7940e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + 1.4761e-07, 1.0803e-07]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0208, 0.0148, 0.0232, 0.0041, 0.0368, -0.0187, -0.0029, -0.0202, + -0.0269, 0.0057], device='cuda:0'), grad: tensor([ 4.6566e-09, -1.2293e-05, 1.4435e-07, 6.0536e-09, 3.5856e-08, + 3.2596e-09, 3.7253e-09, 1.1921e-05, -1.9837e-07, 3.7067e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 220.85, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4331 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.15 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1828], + [ 0.0584, -0.0185, -0.0538, ..., -0.3353, 0.0528, -0.1070], + [-0.0112, -0.0048, 0.2640, ..., -0.3413, -0.1599, -0.1862], + ..., + [-0.0241, 0.0062, -0.2400, ..., -0.0748, 0.1456, 0.0714], + [-0.0328, -0.0202, -0.2138, ..., -0.3339, 0.1277, -0.2316], + [ 0.0151, -0.0258, -0.1187, ..., 0.0400, -0.1974, 0.0237]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -1.3970e-09, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, + 1.1176e-08, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, -2.6543e-08, ..., 0.0000e+00, + 2.7940e-09, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.5367e-08, ..., 2.3283e-09, + -1.8626e-08, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -4.1910e-09, + 4.1910e-09, -1.4901e-08]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0208, 0.0148, 0.0231, 0.0040, 0.0368, -0.0187, -0.0028, -0.0202, + -0.0270, 0.0057], device='cuda:0'), grad: tensor([ 1.5832e-08, 3.9116e-08, -3.7253e-08, -4.1910e-09, 1.7229e-08, + 3.7253e-09, -1.1176e-08, -1.3039e-08, 9.3132e-09, -1.3504e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 220.44, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4265 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.14 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1828], + [ 0.0584, -0.0185, -0.0538, ..., -0.3354, 0.0528, -0.1071], + [-0.0111, -0.0048, 0.2640, ..., -0.3413, -0.1599, -0.1862], + ..., + [-0.0241, 0.0062, -0.2400, ..., -0.0748, 0.1456, 0.0714], + [-0.0328, -0.0202, -0.2139, ..., -0.3340, 0.1277, -0.2317], + [ 0.0151, -0.0258, -0.1187, ..., 0.0400, -0.1974, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, + -3.7253e-09, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., 0.0000e+00, + 3.7253e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.5390e-08, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 0.0000e+00, 2.7008e-08]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0208, 0.0148, 0.0231, 0.0040, 0.0368, -0.0187, -0.0028, -0.0202, + -0.0270, 0.0057], device='cuda:0'), grad: tensor([ 7.4506e-09, -8.3819e-09, 1.1176e-08, 8.3819e-09, -4.1910e-08, + -4.0140e-07, 1.2759e-07, 9.3132e-09, 2.8312e-07, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 220.73, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4244 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.14 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1828], + [ 0.0584, -0.0185, -0.0539, ..., -0.3354, 0.0528, -0.1072], + [-0.0111, -0.0048, 0.2641, ..., -0.3413, -0.1600, -0.1863], + ..., + [-0.0241, 0.0062, -0.2400, ..., -0.0749, 0.1456, 0.0715], + [-0.0328, -0.0202, -0.2139, ..., -0.3342, 0.1277, -0.2319], + [ 0.0152, -0.0258, -0.1187, ..., 0.0401, -0.1974, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.0734e-08, 1.6764e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.4703e-08, 3.3528e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.4995e-08, -6.2399e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -9.3132e-10, + 5.5879e-09, -1.8626e-09]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0209, 0.0148, 0.0231, 0.0040, 0.0367, -0.0187, -0.0028, -0.0202, + -0.0271, 0.0058], device='cuda:0'), grad: tensor([ 1.2107e-08, 1.0896e-07, 2.3562e-07, 3.7253e-08, 5.1223e-08, + -2.2352e-08, -2.7940e-09, -4.2282e-07, -1.4901e-08, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 220.49, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4542 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.13 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1828], + [ 0.0584, -0.0185, -0.0539, ..., -0.3355, 0.0528, -0.1072], + [-0.0112, -0.0048, 0.2641, ..., -0.3413, -0.1600, -0.1864], + ..., + [-0.0241, 0.0062, -0.2401, ..., -0.0749, 0.1456, 0.0715], + [-0.0328, -0.0202, -0.2139, ..., -0.3342, 0.1277, -0.2320], + [ 0.0152, -0.0258, -0.1187, ..., 0.0401, -0.1974, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.3819e-09, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, + 0.0000e+00, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + -1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0047e-08, + 0.0000e+00, 9.0338e-08]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0209, 0.0148, 0.0231, 0.0039, 0.0367, -0.0187, -0.0028, -0.0202, + -0.0271, 0.0058], device='cuda:0'), grad: tensor([-4.2841e-08, 1.0245e-08, 1.2107e-08, 2.7940e-09, -1.3225e-07, + 3.7253e-09, 5.5879e-09, 2.1420e-08, -2.7940e-09, 1.2945e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 220.55, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4034 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.12 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1828], + [ 0.0584, -0.0185, -0.0539, ..., -0.3356, 0.0528, -0.1073], + [-0.0112, -0.0048, 0.2641, ..., -0.3413, -0.1601, -0.1865], + ..., + [-0.0241, 0.0062, -0.2401, ..., -0.0749, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2140, ..., -0.3342, 0.1277, -0.2320], + [ 0.0152, -0.0258, -0.1187, ..., 0.0401, -0.1975, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, + 1.8626e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, + 1.1176e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 2.6077e-08, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.3283e-08, ..., 2.7940e-09, + -7.0781e-08, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -2.3283e-08, + 1.9558e-08, -5.0291e-08]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0209, 0.0148, 0.0231, 0.0039, 0.0367, -0.0188, -0.0028, -0.0202, + -0.0271, 0.0058], device='cuda:0'), grad: tensor([ 2.1420e-08, 5.4017e-08, 1.5460e-07, 4.1910e-08, 7.3574e-08, + 2.2352e-08, 3.4459e-08, -3.7253e-09, -4.1723e-07, 2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 220.76, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4556 re_mapping 0.0029 re_causal 0.0102 /// teacc 99.13 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2898, -0.1829], + [ 0.0584, -0.0185, -0.0539, ..., -0.3356, 0.0528, -0.1073], + [-0.0111, -0.0048, 0.2641, ..., -0.3413, -0.1601, -0.1865], + ..., + [-0.0241, 0.0062, -0.2401, ..., -0.0750, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2140, ..., -0.3342, 0.1277, -0.2320], + [ 0.0152, -0.0258, -0.1188, ..., 0.0401, -0.1975, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.0047e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 3.7253e-09, + 1.9558e-08, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 3.3528e-08, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 2.7940e-09, + -2.7940e-08, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-08, ..., 1.8626e-09, + 0.0000e+00, 2.7940e-09]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0210, 0.0148, 0.0230, 0.0040, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0271, 0.0058], device='cuda:0'), grad: tensor([ 1.3597e-07, 6.6124e-08, 1.1548e-07, -4.5449e-07, -8.3819e-09, + 8.3819e-09, 2.7940e-09, -9.3132e-09, 7.4506e-08, 7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 220.65, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4521 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.12 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2899, -0.1829], + [ 0.0584, -0.0185, -0.0539, ..., -0.3357, 0.0528, -0.1074], + [-0.0111, -0.0048, 0.2641, ..., -0.3414, -0.1601, -0.1866], + ..., + [-0.0241, 0.0062, -0.2402, ..., -0.0750, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2140, ..., -0.3342, 0.1277, -0.2320], + [ 0.0152, -0.0258, -0.1188, ..., 0.0401, -0.1975, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 5.5879e-09, 5.5879e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -4.9360e-08, -5.8673e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 1.8626e-09, -5.5879e-09]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0210, 0.0148, 0.0231, 0.0040, 0.0367, -0.0188, -0.0027, -0.0203, + -0.0271, 0.0058], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.7253e-08, 1.3970e-08, 4.7497e-08, 7.4506e-08, + 2.7940e-09, 0.0000e+00, -1.6857e-07, 5.5879e-09, -9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 220.37, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4494 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.10 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2899, -0.1829], + [ 0.0584, -0.0185, -0.0539, ..., -0.3358, 0.0528, -0.1074], + [-0.0111, -0.0048, 0.2641, ..., -0.3414, -0.1602, -0.1867], + ..., + [-0.0241, 0.0062, -0.2403, ..., -0.0750, 0.1456, 0.0715], + [-0.0328, -0.0202, -0.2140, ..., -0.3342, 0.1277, -0.2321], + [ 0.0152, -0.0258, -0.1188, ..., 0.0401, -0.1974, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., -3.3528e-08, + 9.3132e-10, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 4.6566e-09, + -4.6566e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, -2.1793e-07, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.1234e-07, ..., 9.3132e-10, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.1420e-08, + 0.0000e+00, 3.2596e-08]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0210, 0.0149, 0.0230, 0.0040, 0.0367, -0.0188, -0.0027, -0.0203, + -0.0271, 0.0059], device='cuda:0'), grad: tensor([-1.8440e-07, 2.4214e-08, -4.2189e-07, -2.7940e-09, -5.9605e-08, + 1.7695e-08, 1.2293e-07, 4.2841e-07, 2.8871e-08, 5.7742e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 220.21, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4301 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.11 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2899, -0.1829], + [ 0.0584, -0.0185, -0.0539, ..., -0.3358, 0.0528, -0.1075], + [-0.0112, -0.0048, 0.2641, ..., -0.3414, -0.1603, -0.1867], + ..., + [-0.0241, 0.0062, -0.2403, ..., -0.0750, 0.1456, 0.0715], + [-0.0328, -0.0202, -0.2141, ..., -0.3342, 0.1277, -0.2321], + [ 0.0152, -0.0258, -0.1189, ..., 0.0401, -0.1974, 0.0238]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -6.5193e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -1.8626e-09, -8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0210, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0203, + -0.0271, 0.0059], device='cuda:0'), grad: tensor([-8.3819e-09, 1.8626e-09, 9.3132e-09, -8.9407e-08, 1.0245e-08, + 5.8673e-08, 2.7940e-09, -4.6566e-09, 7.4506e-09, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 221.05, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4406 re_mapping 0.0028 re_causal 0.0099 /// teacc 99.10 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.0021, 0.0110, 0.0628, ..., 0.0887, -0.2899, -0.1829], + [ 0.0584, -0.0185, -0.0540, ..., -0.3360, 0.0528, -0.1075], + [-0.0111, -0.0048, 0.2643, ..., -0.3414, -0.1603, -0.1868], + ..., + [-0.0241, 0.0062, -0.2403, ..., -0.0751, 0.1456, 0.0716], + [-0.0328, -0.0202, -0.2141, ..., -0.3342, 0.1277, -0.2321], + [ 0.0152, -0.0258, -0.1189, ..., 0.0401, -0.1974, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -4.6566e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0211, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0203, + -0.0272, 0.0059], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.2107e-08, 2.7940e-09, 1.3970e-08, -4.6566e-09, + -5.8487e-07, 1.5832e-08, -5.5879e-09, 4.7125e-07, 7.9162e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 220.36, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4339 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.13 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.0021, 0.0110, 0.0629, ..., 0.0887, -0.2899, -0.1829], + [ 0.0583, -0.0185, -0.0540, ..., -0.3361, 0.0529, -0.1077], + [-0.0111, -0.0048, 0.2643, ..., -0.3415, -0.1604, -0.1869], + ..., + [-0.0241, 0.0062, -0.2404, ..., -0.0751, 0.1455, 0.0716], + [-0.0328, -0.0202, -0.2141, ..., -0.3343, 0.1277, -0.2322], + [ 0.0152, -0.0258, -0.1189, ..., 0.0401, -0.1974, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., -3.7253e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.7602e-07, 1.7509e-07], + [ 0.0000e+00, 0.0000e+00, -1.2107e-08, ..., 1.8626e-08, + 1.2666e-07, 1.1735e-07], + ..., + [ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., 1.8626e-09, + -3.2596e-07, -3.1106e-07], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 7.4506e-09, + 9.3132e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -2.7940e-09, + 1.3039e-08, 1.8626e-09]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0211, 0.0149, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0203, + -0.0272, 0.0059], device='cuda:0'), grad: tensor([-5.6811e-08, 7.3854e-07, 3.3341e-07, -8.1956e-08, 2.9802e-08, + 5.6811e-08, 2.4214e-08, -1.1362e-06, 4.1910e-08, 4.4703e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 220.66, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4466 re_mapping 0.0028 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.0021, 0.0110, 0.0629, ..., 0.0885, -0.2899, -0.1833], + [ 0.0583, -0.0185, -0.0541, ..., -0.3362, 0.0529, -0.1078], + [-0.0111, -0.0048, 0.2644, ..., -0.3415, -0.1604, -0.1870], + ..., + [-0.0241, 0.0062, -0.2404, ..., -0.0751, 0.1456, 0.0717], + [-0.0328, -0.0202, -0.2142, ..., -0.3343, 0.1277, -0.2323], + [ 0.0152, -0.0258, -0.1189, ..., 0.0401, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, + 3.2596e-08, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -7.4506e-08, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 5.0291e-08, 1.4901e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -2.4214e-08, -2.4214e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 3.7253e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -1.5832e-08, + 9.3132e-10, -4.8429e-08]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0212, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0272, 0.0059], device='cuda:0'), grad: tensor([ 1.8254e-07, -3.5204e-07, 2.1048e-07, -5.3085e-08, 8.1025e-08, + 4.0047e-08, -3.6322e-08, -3.7253e-08, 5.3085e-08, -7.9162e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 220.47, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4475 re_mapping 0.0027 re_causal 0.0098 /// teacc 99.10 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.0021, 0.0110, 0.0629, ..., 0.0885, -0.2899, -0.1833], + [ 0.0583, -0.0185, -0.0541, ..., -0.3363, 0.0528, -0.1079], + [-0.0111, -0.0048, 0.2644, ..., -0.3416, -0.1605, -0.1871], + ..., + [-0.0241, 0.0062, -0.2404, ..., -0.0752, 0.1456, 0.0717], + [-0.0328, -0.0202, -0.2143, ..., -0.3344, 0.1277, -0.2324], + [ 0.0152, -0.0258, -0.1189, ..., 0.0401, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.6764e-08, + 0.0000e+00, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, + 9.3132e-10, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 9.3132e-10, + 9.3132e-10, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, + -9.3132e-10, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.5739e-07, + -9.3132e-10, -2.3190e-07]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0212, 0.0148, 0.0231, 0.0039, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0273, 0.0060], device='cuda:0'), grad: tensor([ 6.6124e-08, 2.1420e-08, 9.3132e-09, -3.4459e-08, 4.0513e-07, + 8.3819e-08, 1.8626e-09, 1.3970e-08, 7.6368e-08, -6.3889e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 220.58, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4290 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.09 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.0021, 0.0110, 0.0629, ..., 0.0885, -0.2900, -0.1833], + [ 0.0583, -0.0185, -0.0541, ..., -0.3365, 0.0528, -0.1080], + [-0.0111, -0.0048, 0.2644, ..., -0.3416, -0.1605, -0.1871], + ..., + [-0.0241, 0.0062, -0.2404, ..., -0.0752, 0.1456, 0.0718], + [-0.0328, -0.0202, -0.2143, ..., -0.3344, 0.1277, -0.2324], + [ 0.0152, -0.0258, -0.1189, ..., 0.0401, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.5193e-09, + 6.1467e-08, 3.4459e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 5.3365e-07, 2.1048e-07], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + -6.2212e-07, -2.4308e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, + 5.5879e-09, 8.3819e-09]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0212, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0026, -0.0202, + -0.0273, 0.0060], device='cuda:0'), grad: tensor([ 1.8626e-08, 4.6287e-07, 4.0494e-06, 1.4063e-07, -5.2154e-08, + -3.3528e-08, 1.3970e-08, -4.6827e-06, 4.6566e-09, 6.6124e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 220.33, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4313 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.12 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.0021, 0.0110, 0.0630, ..., 0.0885, -0.2900, -0.1834], + [ 0.0583, -0.0185, -0.0541, ..., -0.3365, 0.0528, -0.1081], + [-0.0111, -0.0048, 0.2644, ..., -0.3416, -0.1606, -0.1872], + ..., + [-0.0241, 0.0062, -0.2404, ..., -0.0752, 0.1456, 0.0719], + [-0.0328, -0.0202, -0.2144, ..., -0.3344, 0.1277, -0.2324], + [ 0.0152, -0.0258, -0.1190, ..., 0.0402, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.7742e-08, ..., -1.0990e-07, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, + -5.5879e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -8.6613e-08, ..., 2.3283e-08, + 1.0245e-08, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 8.7544e-08, ..., 2.7940e-09, + -4.6566e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., 7.0781e-08, + 0.0000e+00, -8.3819e-09]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0213, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0026, -0.0202, + -0.0273, 0.0060], device='cuda:0'), grad: tensor([-2.4121e-07, 4.6566e-09, -1.2945e-07, -8.3819e-09, 6.5193e-09, + 1.8626e-09, 9.3132e-09, 1.8999e-07, 1.1176e-08, 1.4901e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 220.23, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4318 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.10 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.0022, 0.0110, 0.0631, ..., 0.0885, -0.2900, -0.1834], + [ 0.0583, -0.0185, -0.0541, ..., -0.3366, 0.0528, -0.1082], + [-0.0111, -0.0048, 0.2644, ..., -0.3416, -0.1607, -0.1873], + ..., + [-0.0241, 0.0062, -0.2405, ..., -0.0753, 0.1456, 0.0719], + [-0.0328, -0.0202, -0.2144, ..., -0.3344, 0.1277, -0.2325], + [ 0.0152, -0.0258, -0.1190, ..., 0.0402, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -4.6566e-09, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.0245e-08, + -6.5193e-09, 1.3970e-08], + [ 0.0000e+00, 0.0000e+00, -3.8184e-08, ..., 6.5193e-09, + 5.5879e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.8626e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 1.8626e-09]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0212, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0026, -0.0202, + -0.0273, 0.0060], device='cuda:0'), grad: tensor([-1.8626e-09, 2.1420e-08, -4.8429e-08, 5.5879e-09, -8.9407e-08, + -8.2888e-08, 9.4995e-08, 5.2154e-08, 3.8184e-08, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 220.48, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4613 re_mapping 0.0026 re_causal 0.0097 /// teacc 99.11 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.0022, 0.0110, 0.0631, ..., 0.0885, -0.2900, -0.1835], + [ 0.0583, -0.0185, -0.0541, ..., -0.3367, 0.0528, -0.1082], + [-0.0111, -0.0048, 0.2645, ..., -0.3417, -0.1607, -0.1873], + ..., + [-0.0241, 0.0062, -0.2405, ..., -0.0753, 0.1456, 0.0719], + [-0.0329, -0.0202, -0.2144, ..., -0.3344, 0.1277, -0.2325], + [ 0.0153, -0.0258, -0.1191, ..., 0.0402, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 9.3132e-10, + 5.5879e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., 9.3132e-10, + 6.8918e-08, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, -1.7788e-07, ..., 0.0000e+00, + 7.4506e-09, -8.3819e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6391e-07, ..., 0.0000e+00, + -6.0536e-08, -2.2352e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + -1.7695e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.4506e-09, + 4.6566e-09, -4.6566e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0212, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0273, 0.0060], device='cuda:0'), grad: tensor([ 4.0047e-08, 3.0361e-07, -5.1223e-07, 9.9652e-08, 2.6077e-08, + -2.0117e-07, -7.4506e-08, 3.5111e-07, -7.4506e-08, 4.3772e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 220.27, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4235 re_mapping 0.0027 re_causal 0.0094 /// teacc 99.09 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.0022, 0.0110, 0.0632, ..., 0.0884, -0.2900, -0.1836], + [ 0.0583, -0.0185, -0.0541, ..., -0.3369, 0.0528, -0.1083], + [-0.0112, -0.0048, 0.2645, ..., -0.3417, -0.1607, -0.1874], + ..., + [-0.0241, 0.0062, -0.2405, ..., -0.0753, 0.1456, 0.0720], + [-0.0329, -0.0202, -0.2145, ..., -0.3345, 0.1277, -0.2326], + [ 0.0153, -0.0258, -0.1192, ..., 0.0402, -0.1975, 0.0239]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 9.3132e-10, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, + -3.7253e-09, 2.5146e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, + 0.0000e+00, 2.7008e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.5146e-08, + 0.0000e+00, 2.4214e-08]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0213, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0273, 0.0060], device='cuda:0'), grad: tensor([ 1.4901e-08, 5.4017e-08, 9.3132e-10, 2.6077e-08, -1.6298e-07, + -3.1665e-08, 4.6566e-09, 2.7940e-08, 4.7497e-08, 2.3283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 220.46, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4935 re_mapping 0.0027 re_causal 0.0101 /// teacc 99.11 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.0022, 0.0110, 0.0635, ..., 0.0885, -0.2900, -0.1837], + [ 0.0583, -0.0185, -0.0541, ..., -0.3372, 0.0528, -0.1085], + [-0.0112, -0.0048, 0.2645, ..., -0.3418, -0.1608, -0.1875], + ..., + [-0.0241, 0.0062, -0.2405, ..., -0.0754, 0.1456, 0.0720], + [-0.0329, -0.0202, -0.2145, ..., -0.3345, 0.1277, -0.2326], + [ 0.0153, -0.0258, -0.1192, ..., 0.0402, -0.1975, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.9558e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + -9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0212, 0.0148, 0.0230, 0.0039, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0274, 0.0061], device='cuda:0'), grad: tensor([ 3.0734e-08, 2.9802e-08, 5.6811e-08, -2.1234e-07, 5.5879e-09, + 4.8429e-08, -1.9558e-08, 2.2352e-08, 4.5635e-08, 9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 220.56, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4378 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.10 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.0022, 0.0110, 0.0637, ..., 0.0885, -0.2901, -0.1837], + [ 0.0583, -0.0185, -0.0542, ..., -0.3375, 0.0528, -0.1086], + [-0.0112, -0.0048, 0.2645, ..., -0.3419, -0.1608, -0.1876], + ..., + [-0.0241, 0.0062, -0.2406, ..., -0.0755, 0.1456, 0.0720], + [-0.0329, -0.0202, -0.2146, ..., -0.3345, 0.1277, -0.2327], + [ 0.0153, -0.0258, -0.1193, ..., 0.0402, -0.1976, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, + 9.3132e-10, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.5367e-07, 1.8720e-07], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.7940e-09, + 9.3132e-09, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + -2.1234e-07, -2.8685e-07], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -1.8626e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3039e-08, + 1.2107e-08, -3.7253e-09]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0211, 0.0148, 0.0229, 0.0039, 0.0367, -0.0189, -0.0027, -0.0202, + -0.0274, 0.0061], device='cuda:0'), grad: tensor([ 2.4214e-08, 8.4937e-07, 3.8184e-08, 2.8592e-07, 2.5146e-08, + -8.3819e-09, 2.7008e-08, -1.1967e-06, -7.6368e-08, 3.4459e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 220.36, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4382 re_mapping 0.0027 re_causal 0.0097 /// teacc 99.15 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.0022, 0.0110, 0.0638, ..., 0.0885, -0.2901, -0.1837], + [ 0.0583, -0.0185, -0.0542, ..., -0.3376, 0.0528, -0.1087], + [-0.0112, -0.0048, 0.2645, ..., -0.3420, -0.1609, -0.1876], + ..., + [-0.0241, 0.0062, -0.2406, ..., -0.0755, 0.1456, 0.0720], + [-0.0329, -0.0202, -0.2146, ..., -0.3345, 0.1277, -0.2327], + [ 0.0153, -0.0258, -0.1194, ..., 0.0402, -0.1975, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, + -9.3132e-10, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 0.0000e+00, 1.2107e-08]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0210, 0.0147, 0.0229, 0.0039, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0274, 0.0061], device='cuda:0'), grad: tensor([-1.8626e-09, 4.6566e-09, 9.3132e-10, 1.9558e-08, -2.2352e-08, + -3.5390e-08, 1.2107e-08, 3.7253e-09, 1.3039e-08, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 220.68, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4266 re_mapping 0.0027 re_causal 0.0095 /// teacc 99.13 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.0022, 0.0110, 0.0638, ..., 0.0886, -0.2901, -0.1837], + [ 0.0583, -0.0185, -0.0542, ..., -0.3376, 0.0528, -0.1087], + [-0.0112, -0.0048, 0.2645, ..., -0.3420, -0.1610, -0.1876], + ..., + [-0.0241, 0.0062, -0.2406, ..., -0.0755, 0.1456, 0.0720], + [-0.0329, -0.0202, -0.2147, ..., -0.3345, 0.1277, -0.2327], + [ 0.0153, -0.0258, -0.1195, ..., 0.0402, -0.1975, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -2.7940e-09, + 6.9849e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.3970e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.1223e-09, + -4.6566e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 3.7253e-09, + 1.3970e-09, 9.7789e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -6.3330e-08, + 3.2596e-09, -1.2852e-07]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0210, 0.0148, 0.0228, 0.0040, 0.0367, -0.0188, -0.0027, -0.0202, + -0.0274, 0.0061], device='cuda:0'), grad: tensor([ 5.9139e-08, 1.1642e-08, 1.0710e-08, -4.1910e-09, 1.3970e-07, + 1.3877e-07, -1.5553e-07, 0.0000e+00, 6.0536e-08, -2.5006e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 220.77, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4250 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.13 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.0022, 0.0110, 0.0639, ..., 0.0886, -0.2901, -0.1837], + [ 0.0583, -0.0185, -0.0542, ..., -0.3377, 0.0528, -0.1088], + [-0.0112, -0.0048, 0.2645, ..., -0.3421, -0.1610, -0.1877], + ..., + [-0.0241, 0.0062, -0.2407, ..., -0.0756, 0.1456, 0.0721], + [-0.0329, -0.0202, -0.2147, ..., -0.3346, 0.1277, -0.2328], + [ 0.0153, -0.0258, -0.1196, ..., 0.0402, -0.1976, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + 4.6566e-10, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 1.8626e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + -3.2596e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -3.7253e-09, + 9.3132e-10, -1.5367e-08]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0210, 0.0148, 0.0228, 0.0040, 0.0367, -0.0189, -0.0027, -0.0202, + -0.0274, 0.0061], device='cuda:0'), grad: tensor([ 5.5879e-09, 9.7789e-09, 1.0710e-08, -4.7963e-08, 1.4435e-08, + 2.9337e-08, -3.7253e-09, 1.3970e-09, 3.2596e-09, -2.1420e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 220.41, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4628 re_mapping 0.0027 re_causal 0.0100 /// teacc 99.13 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.0022, 0.0110, 0.0640, ..., 0.0886, -0.2901, -0.1838], + [ 0.0583, -0.0185, -0.0543, ..., -0.3378, 0.0528, -0.1089], + [-0.0112, -0.0048, 0.2646, ..., -0.3422, -0.1610, -0.1877], + ..., + [-0.0241, 0.0062, -0.2407, ..., -0.0756, 0.1456, 0.0722], + [-0.0329, -0.0202, -0.2148, ..., -0.3346, 0.1277, -0.2328], + [ 0.0153, -0.0258, -0.1197, ..., 0.0402, -0.1977, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 2.3283e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 0.0000e+00, + 2.7940e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -5.2154e-08, -3.2596e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 5.1223e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 3.7253e-09]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0210, 0.0147, 0.0229, 0.0040, 0.0367, -0.0189, -0.0027, -0.0202, + -0.0275, 0.0061], device='cuda:0'), grad: tensor([ 3.0268e-08, 4.1910e-09, 1.8626e-09, 8.0094e-08, 6.9849e-09, + 3.3993e-08, -8.4285e-08, -1.1874e-07, 3.4925e-08, 1.9092e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 220.40, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4545 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.11 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.0022, 0.0110, 0.0641, ..., 0.0886, -0.2902, -0.1838], + [ 0.0582, -0.0185, -0.0543, ..., -0.3379, 0.0528, -0.1088], + [-0.0112, -0.0048, 0.2647, ..., -0.3423, -0.1611, -0.1878], + ..., + [-0.0241, 0.0062, -0.2408, ..., -0.0757, 0.1456, 0.0721], + [-0.0329, -0.0202, -0.2148, ..., -0.3347, 0.1277, -0.2329], + [ 0.0153, -0.0258, -0.1197, ..., 0.0402, -0.1977, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, + -7.7300e-08, 1.5367e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 1.3504e-08, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + 6.3330e-08, -4.6566e-09], + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 4.6566e-10, + -6.9849e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.9558e-08, + 1.8626e-09, 1.2107e-08]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0210, 0.0148, 0.0229, 0.0040, 0.0367, -0.0189, -0.0027, -0.0202, + -0.0275, 0.0061], device='cuda:0'), grad: tensor([ 6.5193e-09, -2.6776e-07, 4.5169e-08, 1.5832e-08, -4.7032e-08, + -3.7253e-09, -3.7253e-09, 2.4494e-07, -1.2573e-08, 3.3062e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 220.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4378 re_mapping 0.0026 re_causal 0.0098 /// teacc 99.13 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.0022, 0.0110, 0.0642, ..., 0.0887, -0.2902, -0.1839], + [ 0.0582, -0.0185, -0.0543, ..., -0.3380, 0.0528, -0.1089], + [-0.0112, -0.0048, 0.2647, ..., -0.3423, -0.1611, -0.1879], + ..., + [-0.0241, 0.0062, -0.2408, ..., -0.0757, 0.1456, 0.0722], + [-0.0329, -0.0202, -0.2149, ..., -0.3347, 0.1278, -0.2329], + [ 0.0153, -0.0258, -0.1198, ..., 0.0402, -0.1978, 0.0240]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-08, ..., -4.6566e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-09, 6.0536e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -1.0245e-08, -1.3504e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0209, 0.0148, 0.0229, 0.0039, 0.0366, -0.0188, -0.0028, -0.0202, + -0.0275, 0.0061], device='cuda:0'), grad: tensor([-6.0536e-08, -1.2573e-08, 2.5611e-08, 4.0978e-08, 2.3283e-09, + -6.7521e-08, 5.8673e-08, -1.6764e-08, 3.3528e-08, 3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 220.94, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4187 re_mapping 0.0027 re_causal 0.0094 /// teacc 99.11 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.0022, 0.0110, 0.0642, ..., 0.0887, -0.2902, -0.1840], + [ 0.0582, -0.0185, -0.0543, ..., -0.3381, 0.0528, -0.1089], + [-0.0112, -0.0048, 0.2647, ..., -0.3423, -0.1612, -0.1880], + ..., + [-0.0241, 0.0062, -0.2408, ..., -0.0757, 0.1456, 0.0722], + [-0.0329, -0.0202, -0.2149, ..., -0.3347, 0.1278, -0.2330], + [ 0.0153, -0.0258, -0.1198, ..., 0.0403, -0.1978, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -4.6566e-10, + 0.0000e+00, 1.3504e-08], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 2.3283e-09, + 1.8626e-09, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 1.8626e-09, + 4.6566e-10, 1.5367e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-09, + -4.6566e-10, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 1.8626e-09, + 4.6566e-10, 1.2107e-08], + [ 0.0000e+00, 0.0000e+00, -2.1420e-08, ..., -6.9849e-09, + -1.3970e-09, -9.8720e-08]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0209, 0.0148, 0.0229, 0.0039, 0.0366, -0.0188, -0.0028, -0.0202, + -0.0276, 0.0062], device='cuda:0'), grad: tensor([ 3.3993e-08, 3.4459e-08, 8.3819e-08, -2.7474e-08, 2.1886e-08, + 4.2841e-08, 7.9162e-09, 2.1886e-08, 3.6787e-08, -2.4727e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 220.61, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4319 re_mapping 0.0027 re_causal 0.0096 /// teacc 99.12 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.0022, 0.0110, 0.0643, ..., 0.0886, -0.2902, -0.1841], + [ 0.0582, -0.0185, -0.0543, ..., -0.3382, 0.0528, -0.1090], + [-0.0112, -0.0048, 0.2647, ..., -0.3424, -0.1612, -0.1881], + ..., + [-0.0242, 0.0062, -0.2409, ..., -0.0758, 0.1456, 0.0722], + [-0.0329, -0.0202, -0.2150, ..., -0.3348, 0.1278, -0.2330], + [ 0.0153, -0.0258, -0.1199, ..., 0.0403, -0.1979, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 2.7940e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -1.8161e-07, -7.3109e-08], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.1642e-08, 5.1223e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 3.0734e-08, 1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.8871e-08, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.6566e-09, + 1.3970e-09, -1.4435e-08]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0209, 0.0147, 0.0229, 0.0039, 0.0366, -0.0188, -0.0028, -0.0202, + -0.0276, 0.0062], device='cuda:0'), grad: tensor([ 2.0955e-08, -7.6881e-07, 5.4017e-08, -4.0513e-08, 2.9569e-07, + 6.4261e-08, 1.0384e-07, 1.4063e-07, 1.2992e-07, -6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 220.47, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4297 re_mapping 0.0026 re_causal 0.0096 /// teacc 99.13 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.0022, 0.0110, 0.0644, ..., 0.0887, -0.2903, -0.1841], + [ 0.0582, -0.0185, -0.0543, ..., -0.3382, 0.0528, -0.1090], + [-0.0112, -0.0048, 0.2647, ..., -0.3424, -0.1613, -0.1882], + ..., + [-0.0242, 0.0062, -0.2409, ..., -0.0758, 0.1456, 0.0723], + [-0.0329, -0.0202, -0.2151, ..., -0.3348, 0.1278, -0.2331], + [ 0.0153, -0.0258, -0.1199, ..., 0.0403, -0.1980, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 6.9849e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -3.2596e-09, -6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0209, 0.0147, 0.0229, 0.0039, 0.0366, -0.0188, -0.0028, -0.0202, + -0.0276, 0.0062], device='cuda:0'), grad: tensor([ 1.1642e-08, 2.7940e-08, 2.1420e-08, -1.9558e-08, 2.7008e-08, + 1.5832e-08, -3.3528e-08, -1.7695e-08, -3.3528e-08, 3.2596e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 220.81, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4063 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.09 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.0022, 0.0110, 0.0644, ..., 0.0886, -0.2903, -0.1842], + [ 0.0582, -0.0185, -0.0543, ..., -0.3383, 0.0528, -0.1091], + [-0.0112, -0.0048, 0.2647, ..., -0.3425, -0.1613, -0.1882], + ..., + [-0.0242, 0.0062, -0.2410, ..., -0.0758, 0.1456, 0.0723], + [-0.0329, -0.0202, -0.2151, ..., -0.3348, 0.1278, -0.2331], + [ 0.0153, -0.0258, -0.1200, ..., 0.0403, -0.1981, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.3062e-08, ..., -1.4901e-08, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 1.2495e-05, ..., 4.6566e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, -1.2986e-05, ..., 9.3132e-10, + 0.0000e+00, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 0.0000e+00, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.3039e-07, ..., 1.4435e-08, + -1.3970e-09, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, 3.9116e-08, ..., -1.5367e-08, + 0.0000e+00, -3.2596e-08]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0209, 0.0147, 0.0229, 0.0039, 0.0366, -0.0188, -0.0027, -0.0202, + -0.0276, 0.0062], device='cuda:0'), grad: tensor([ 5.2154e-08, 2.4483e-05, -2.5421e-05, 1.7509e-07, 2.0629e-07, + 1.3877e-07, 1.1409e-07, 2.4680e-08, 2.9150e-07, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 220.55, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4327 re_mapping 0.0026 re_causal 0.0094 /// teacc 99.09 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.0022, 0.0110, 0.0646, ..., 0.0887, -0.2903, -0.1842], + [ 0.0582, -0.0185, -0.0544, ..., -0.3384, 0.0528, -0.1091], + [-0.0112, -0.0048, 0.2648, ..., -0.3426, -0.1614, -0.1884], + ..., + [-0.0242, 0.0062, -0.2410, ..., -0.0759, 0.1457, 0.0724], + [-0.0329, -0.0202, -0.2152, ..., -0.3348, 0.1278, -0.2331], + [ 0.0153, -0.0258, -0.1200, ..., 0.0403, -0.1981, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 1.2573e-08, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + -5.0757e-08, -9.1270e-08], + [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 0.0000e+00, + -4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-09, + 2.7940e-09, -5.1223e-09]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0209, 0.0147, 0.0229, 0.0039, 0.0366, -0.0188, -0.0027, -0.0202, + -0.0277, 0.0062], device='cuda:0'), grad: tensor([ 1.0710e-08, 4.2375e-08, -3.7253e-09, 2.3283e-09, 3.4925e-08, + 9.5926e-08, 0.0000e+00, -1.8440e-07, 1.3970e-09, 0.0000e+00], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 220.48, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4332 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.0022, 0.0110, 0.0646, ..., 0.0887, -0.2904, -0.1842], + [ 0.0581, -0.0185, -0.0545, ..., -0.3384, 0.0528, -0.1092], + [-0.0111, -0.0048, 0.2649, ..., -0.3426, -0.1615, -0.1883], + ..., + [-0.0242, 0.0062, -0.2411, ..., -0.0759, 0.1457, 0.0724], + [-0.0329, -0.0202, -0.2152, ..., -0.3349, 0.1278, -0.2333], + [ 0.0153, -0.0258, -0.1201, ..., 0.0403, -0.1982, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.3551e-08, ..., -4.3306e-08, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, + 1.7369e-07, 1.5413e-07], + [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 1.3504e-08, + 3.1665e-08, 2.8871e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + -3.8836e-07, -3.4738e-07], + [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., 5.5879e-09, + 3.2596e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 1.2573e-08, ..., 1.6298e-08, + 4.4238e-08, 4.8894e-08]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0209, 0.0147, 0.0229, 0.0039, 0.0366, -0.0188, -0.0027, -0.0202, + -0.0277, 0.0062], device='cuda:0'), grad: tensor([-1.3458e-07, 5.9325e-07, 1.5879e-07, 5.1176e-07, -6.0536e-09, + -2.3283e-07, 2.1933e-07, -1.3411e-06, 4.8429e-08, 1.9278e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 220.51, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4179 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.09 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.0022, 0.0110, 0.0647, ..., 0.0887, -0.2904, -0.1842], + [ 0.0581, -0.0185, -0.0546, ..., -0.3386, 0.0528, -0.1094], + [-0.0111, -0.0048, 0.2650, ..., -0.3427, -0.1617, -0.1886], + ..., + [-0.0242, 0.0062, -0.2412, ..., -0.0760, 0.1457, 0.0725], + [-0.0329, -0.0202, -0.2153, ..., -0.3350, 0.1279, -0.2333], + [ 0.0153, -0.0258, -0.1203, ..., 0.0403, -0.1983, 0.0241]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.0536e-09, ..., 0.0000e+00, + 9.3132e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 1.8626e-09, + 2.6543e-08, 2.0023e-08], + [ 0.0000e+00, 0.0000e+00, -3.3993e-08, ..., 0.0000e+00, + 5.5879e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 4.6566e-10, + -4.2841e-08, -2.9802e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 4.6566e-10, + -1.3970e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.3283e-09, + 2.3283e-09, -2.3283e-09]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0208, 0.0147, 0.0230, 0.0040, 0.0366, -0.0189, -0.0028, -0.0201, + -0.0277, 0.0062], device='cuda:0'), grad: tensor([ 4.7032e-08, 8.2422e-08, -5.7742e-08, -9.7789e-09, 2.1886e-08, + 2.7940e-09, -2.6077e-08, -7.0315e-08, 4.6566e-10, -3.7253e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 220.60, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4401 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.07 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.0022, 0.0110, 0.0649, ..., 0.0887, -0.2904, -0.1843], + [ 0.0581, -0.0185, -0.0547, ..., -0.3389, 0.0528, -0.1097], + [-0.0111, -0.0048, 0.2651, ..., -0.3429, -0.1617, -0.1888], + ..., + [-0.0242, 0.0062, -0.2412, ..., -0.0760, 0.1457, 0.0728], + [-0.0329, -0.0202, -0.2155, ..., -0.3351, 0.1279, -0.2335], + [ 0.0153, -0.0258, -0.1203, ..., 0.0403, -0.1985, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.3062e-08, ..., -4.7963e-08, + 4.6566e-10, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7695e-08, + 4.6566e-10, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 0.0000e+00, 8.3819e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.4901e-08, + 0.0000e+00, 2.3283e-08], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 9.3132e-09, + 4.6566e-10, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, 1.6298e-08, ..., 2.9337e-08, + 0.0000e+00, 6.0536e-09]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0208, 0.0146, 0.0230, 0.0039, 0.0366, -0.0189, -0.0029, -0.0201, + -0.0278, 0.0063], device='cuda:0'), grad: tensor([-1.0477e-07, 6.2399e-08, 2.0489e-08, 1.1642e-08, -1.5507e-07, + 2.3283e-09, 3.2596e-08, 5.2620e-08, 2.5146e-08, 5.9605e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 220.30, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4428 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.07 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.0022, 0.0110, 0.0650, ..., 0.0888, -0.2905, -0.1844], + [ 0.0581, -0.0185, -0.0547, ..., -0.3390, 0.0527, -0.1100], + [-0.0111, -0.0048, 0.2652, ..., -0.3430, -0.1618, -0.1889], + ..., + [-0.0242, 0.0062, -0.2413, ..., -0.0761, 0.1458, 0.0730], + [-0.0329, -0.0202, -0.2155, ..., -0.3351, 0.1279, -0.2335], + [ 0.0153, -0.0258, -0.1204, ..., 0.0403, -0.1986, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.2573e-08, + 4.6100e-08, 4.1444e-08], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 1.3970e-09, + 1.7229e-08, 1.2107e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -5.7276e-08, + -2.5565e-07, -1.9139e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, + 3.2596e-08, 2.6077e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.0268e-08, + 1.1129e-07, 9.0338e-08]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0207, 0.0146, 0.0230, 0.0039, 0.0366, -0.0189, -0.0029, -0.0201, + -0.0278, 0.0063], device='cuda:0'), grad: tensor([ 3.2596e-09, 1.1548e-07, 4.4238e-08, 3.2596e-08, 4.5635e-08, + 4.1910e-09, 1.2573e-08, -6.0769e-07, 8.3353e-08, 2.7288e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 220.44, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4359 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.08 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.0022, 0.0110, 0.0651, ..., 0.0889, -0.2905, -0.1844], + [ 0.0581, -0.0185, -0.0547, ..., -0.3391, 0.0527, -0.1100], + [-0.0111, -0.0048, 0.2652, ..., -0.3430, -0.1619, -0.1890], + ..., + [-0.0242, 0.0062, -0.2414, ..., -0.0761, 0.1458, 0.0730], + [-0.0329, -0.0202, -0.2155, ..., -0.3352, 0.1280, -0.2335], + [ 0.0153, -0.0258, -0.1205, ..., 0.0403, -0.1987, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 2.8405e-08, 6.8918e-08], + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + 3.7253e-09, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + -3.4459e-08, -5.6345e-08], + [ 0.0000e+00, 0.0000e+00, -2.3283e-09, ..., 0.0000e+00, + -6.9849e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.8626e-09, + 4.6566e-09, -1.0245e-08]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0206, 0.0146, 0.0229, 0.0040, 0.0366, -0.0190, -0.0030, -0.0201, + -0.0278, 0.0063], device='cuda:0'), grad: tensor([ 4.1910e-09, 1.8487e-07, 2.4680e-08, 3.7253e-09, -1.5832e-08, + 3.7253e-09, 3.2596e-09, -1.7742e-07, -2.4214e-08, -1.0710e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 220.19, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4399 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.06 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.0022, 0.0110, 0.0652, ..., 0.0889, -0.2905, -0.1844], + [ 0.0581, -0.0185, -0.0549, ..., -0.3392, 0.0528, -0.1101], + [-0.0111, -0.0048, 0.2654, ..., -0.3431, -0.1620, -0.1891], + ..., + [-0.0242, 0.0062, -0.2415, ..., -0.0762, 0.1458, 0.0731], + [-0.0329, -0.0202, -0.2157, ..., -0.3352, 0.1280, -0.2335], + [ 0.0153, -0.0258, -0.1205, ..., 0.0403, -0.1988, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., -3.3528e-08, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 1.4110e-07, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 0.0000e+00, + 1.2573e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 4.6566e-10, + -1.5320e-07, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, -2.6543e-08, ..., 4.6566e-10, + -1.6764e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 0.0000e+00, + 4.1910e-09, -7.9162e-09]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0205, 0.0146, 0.0230, 0.0040, 0.0366, -0.0190, -0.0031, -0.0200, + -0.0278, 0.0064], device='cuda:0'), grad: tensor([-2.2678e-07, 3.8557e-07, 7.0315e-08, -1.1642e-08, 2.2352e-08, + 3.7253e-09, 2.6822e-07, -4.0699e-07, -1.6065e-07, 6.3796e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 220.33, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4274 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.08 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.0022, 0.0110, 0.0653, ..., 0.0890, -0.2905, -0.1844], + [ 0.0581, -0.0185, -0.0549, ..., -0.3393, 0.0528, -0.1103], + [-0.0111, -0.0048, 0.2654, ..., -0.3432, -0.1621, -0.1892], + ..., + [-0.0242, 0.0062, -0.2415, ..., -0.0763, 0.1458, 0.0732], + [-0.0329, -0.0202, -0.2158, ..., -0.3353, 0.1280, -0.2337], + [ 0.0153, -0.0258, -0.1206, ..., 0.0404, -0.1989, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 0.0000e+00, + 0.0000e+00, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 8.8476e-09, 5.5879e-09], + [ 0.0000e+00, 0.0000e+00, -5.2620e-08, ..., 0.0000e+00, + 2.2352e-08, -1.1642e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 3.7719e-08, ..., 0.0000e+00, + -3.3528e-08, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 1.3970e-09, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.7940e-09, + 9.3132e-10, -3.7253e-09]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0205, 0.0145, 0.0230, 0.0040, 0.0365, -0.0190, -0.0031, -0.0200, + -0.0279, 0.0064], device='cuda:0'), grad: tensor([ 2.1886e-08, 2.1420e-08, -6.7987e-08, -3.8184e-08, 7.9162e-09, + 4.1444e-08, 0.0000e+00, 1.9092e-08, 1.1642e-08, -9.3132e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 220.55, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4227 re_mapping 0.0025 re_causal 0.0091 /// teacc 99.12 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.0022, 0.0110, 0.0653, ..., 0.0890, -0.2906, -0.1845], + [ 0.0581, -0.0185, -0.0550, ..., -0.3395, 0.0527, -0.1105], + [-0.0111, -0.0048, 0.2655, ..., -0.3433, -0.1622, -0.1893], + ..., + [-0.0242, 0.0062, -0.2415, ..., -0.0763, 0.1458, 0.0735], + [-0.0329, -0.0202, -0.2159, ..., -0.3353, 0.1281, -0.2338], + [ 0.0153, -0.0258, -0.1206, ..., 0.0404, -0.1991, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 9.3132e-10, + 1.3970e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 0.0000e+00, + 4.6566e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 0.0000e+00, + -1.3970e-09, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, + 0.0000e+00, 3.7253e-09]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0206, 0.0145, 0.0230, 0.0039, 0.0365, -0.0190, -0.0031, -0.0200, + -0.0279, 0.0064], device='cuda:0'), grad: tensor([ 8.8476e-09, 3.2596e-08, 1.9558e-08, -1.4435e-07, -6.0536e-09, + 3.2131e-08, 1.3970e-09, 1.1642e-08, 4.7032e-08, 1.1642e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 220.40, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4052 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.0022, 0.0110, 0.0654, ..., 0.0890, -0.2907, -0.1845], + [ 0.0581, -0.0185, -0.0550, ..., -0.3396, 0.0527, -0.1108], + [-0.0111, -0.0048, 0.2655, ..., -0.3433, -0.1623, -0.1894], + ..., + [-0.0242, 0.0062, -0.2417, ..., -0.0764, 0.1459, 0.0737], + [-0.0329, -0.0202, -0.2160, ..., -0.3354, 0.1280, -0.2339], + [ 0.0153, -0.0258, -0.1207, ..., 0.0404, -0.1992, 0.0242]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., -3.7253e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -2.3283e-09, ..., 2.7940e-09, + 4.6566e-10, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.3283e-09, + 2.2817e-08, 7.4506e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.4901e-08, + -5.1223e-09, 1.7229e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.3970e-09, + -2.8871e-08, -8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.0757e-07, + 2.3283e-09, 1.4761e-07]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0206, 0.0144, 0.0230, 0.0040, 0.0365, -0.0191, -0.0030, -0.0199, + -0.0279, 0.0064], device='cuda:0'), grad: tensor([-7.4506e-09, 3.2596e-09, 5.8208e-08, 5.9139e-08, -2.2585e-07, + -6.8918e-08, 0.0000e+00, 1.9092e-08, -3.9116e-08, 2.0489e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 220.14, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4268 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.12 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.0022, 0.0110, 0.0654, ..., 0.0890, -0.2907, -0.1845], + [ 0.0580, -0.0185, -0.0550, ..., -0.3396, 0.0527, -0.1109], + [-0.0111, -0.0048, 0.2656, ..., -0.3434, -0.1624, -0.1895], + ..., + [-0.0242, 0.0062, -0.2418, ..., -0.0764, 0.1459, 0.0738], + [-0.0329, -0.0202, -0.2160, ..., -0.3354, 0.1280, -0.2339], + [ 0.0153, -0.0258, -0.1208, ..., 0.0404, -0.1993, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 4.6566e-10, + 1.1642e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, + -2.9337e-08, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 1.4296e-07, ..., 4.6566e-10, + 4.6566e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 2.3283e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 1.0710e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -3.7253e-09, + 1.3970e-09, -9.3132e-09]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0207, 0.0144, 0.0230, 0.0039, 0.0365, -0.0190, -0.0029, -0.0199, + -0.0280, 0.0064], device='cuda:0'), grad: tensor([ 5.4948e-08, -9.2201e-08, 3.5157e-07, -3.6508e-07, 6.5193e-09, + -9.3132e-09, 1.2573e-08, 1.4435e-08, 4.7032e-08, -2.0955e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 220.07, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4417 re_mapping 0.0025 re_causal 0.0094 /// teacc 99.13 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.0022, 0.0110, 0.0655, ..., 0.0889, -0.2908, -0.1847], + [ 0.0580, -0.0185, -0.0550, ..., -0.3398, 0.0526, -0.1110], + [-0.0111, -0.0048, 0.2656, ..., -0.3435, -0.1625, -0.1895], + ..., + [-0.0242, 0.0062, -0.2419, ..., -0.0765, 0.1459, 0.0739], + [-0.0329, -0.0202, -0.2162, ..., -0.3354, 0.1280, -0.2340], + [ 0.0154, -0.0258, -0.1210, ..., 0.0404, -0.1994, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4668e-08, ..., -1.2573e-08, + 2.3283e-10, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 6.9849e-10, + -3.0268e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 5.8208e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + -9.3132e-10, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.9849e-10, + 2.3283e-10, -1.3970e-09]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0207, 0.0144, 0.0230, 0.0040, 0.0365, -0.0191, -0.0029, -0.0199, + -0.0280, 0.0064], device='cuda:0'), grad: tensor([-1.1455e-07, -2.0489e-08, 3.0734e-08, 2.3283e-09, 3.2596e-09, + 1.0710e-08, 9.8720e-08, 4.6566e-10, 3.4925e-09, 3.4925e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 220.42, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4329 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.13 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.0022, 0.0110, 0.0656, ..., 0.0889, -0.2908, -0.1848], + [ 0.0579, -0.0185, -0.0550, ..., -0.3398, 0.0526, -0.1111], + [-0.0111, -0.0048, 0.2656, ..., -0.3435, -0.1626, -0.1896], + ..., + [-0.0242, 0.0062, -0.2420, ..., -0.0765, 0.1460, 0.0740], + [-0.0329, -0.0202, -0.2162, ..., -0.3355, 0.1281, -0.2341], + [ 0.0154, -0.0258, -0.1211, ..., 0.0404, -0.1995, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4680e-08, ..., -3.3993e-08, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 3.0268e-09, ..., 2.3283e-09, + -1.1642e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.5611e-09, ..., 8.3819e-09, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 6.5193e-09, + 9.3132e-10, 8.3819e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 3.7253e-09, + 2.3283e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 3.9581e-09, ..., 4.4238e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0207, 0.0144, 0.0230, 0.0041, 0.0365, -0.0192, -0.0028, -0.0198, + -0.0281, 0.0064], device='cuda:0'), grad: tensor([-1.2456e-07, 5.3551e-09, 3.0734e-08, 2.3516e-08, -5.1223e-09, + 9.0804e-09, 2.1420e-08, 2.4913e-08, 1.3970e-08, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 220.29, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4022 re_mapping 0.0025 re_causal 0.0091 /// teacc 99.13 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.0022, 0.0110, 0.0657, ..., 0.0889, -0.2909, -0.1849], + [ 0.0579, -0.0185, -0.0550, ..., -0.3399, 0.0526, -0.1112], + [-0.0111, -0.0048, 0.2656, ..., -0.3436, -0.1627, -0.1896], + ..., + [-0.0242, 0.0062, -0.2421, ..., -0.0766, 0.1460, 0.0741], + [-0.0329, -0.0202, -0.2163, ..., -0.3355, 0.1281, -0.2341], + [ 0.0154, -0.0258, -0.1211, ..., 0.0404, -0.1996, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 3.4925e-09, + 2.3283e-10, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 1.6298e-09, ..., 2.5611e-09, + 9.5228e-08, 1.0710e-08], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.3970e-09, + 3.0966e-08, 1.1409e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 6.0536e-09, + -1.4412e-07, -1.1874e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.3970e-09, + 6.9849e-10, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 1.8626e-09, + 6.9849e-10, 5.3551e-09]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0208, 0.0144, 0.0230, 0.0041, 0.0365, -0.0192, -0.0028, -0.0198, + -0.0281, 0.0064], device='cuda:0'), grad: tensor([ 2.1188e-08, 3.4412e-07, 1.1944e-07, 8.3819e-09, -4.8196e-08, + 8.8476e-09, 2.7940e-09, -4.7660e-07, 1.1642e-08, 1.6298e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 220.56, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4452 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.17 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.0022, 0.0110, 0.0657, ..., 0.0890, -0.2910, -0.1849], + [ 0.0579, -0.0185, -0.0550, ..., -0.3400, 0.0526, -0.1113], + [-0.0111, -0.0048, 0.2657, ..., -0.3437, -0.1628, -0.1897], + ..., + [-0.0242, 0.0062, -0.2423, ..., -0.0767, 0.1460, 0.0741], + [-0.0329, -0.0202, -0.2164, ..., -0.3355, 0.1281, -0.2342], + [ 0.0154, -0.0258, -0.1212, ..., 0.0404, -0.1997, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 8.8476e-09, + -8.8476e-09, 2.4214e-08], + [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 2.3283e-10, + 1.0710e-08, 6.9849e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.1642e-08, + -6.8219e-08, -5.7276e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, + 1.8626e-09, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 7.2177e-09, + 4.2841e-08, 3.4459e-08]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0208, 0.0144, 0.0229, 0.0041, 0.0365, -0.0191, -0.0028, -0.0198, + -0.0281, 0.0064], device='cuda:0'), grad: tensor([ 5.1688e-08, -6.5193e-08, 3.4925e-08, -3.5390e-08, 6.9849e-10, + 3.3295e-08, 1.2573e-08, -1.2456e-07, 1.0012e-08, 8.5216e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 220.56, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4058 re_mapping 0.0025 re_causal 0.0090 /// teacc 99.15 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.0022, 0.0110, 0.0658, ..., 0.0890, -0.2911, -0.1850], + [ 0.0579, -0.0185, -0.0550, ..., -0.3400, 0.0526, -0.1113], + [-0.0111, -0.0048, 0.2657, ..., -0.3437, -0.1629, -0.1898], + ..., + [-0.0242, 0.0062, -0.2423, ..., -0.0767, 0.1460, 0.0742], + [-0.0329, -0.0202, -0.2165, ..., -0.3356, 0.1281, -0.2342], + [ 0.0154, -0.0258, -0.1213, ..., 0.0404, -0.1998, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., -6.0070e-08, + 0.0000e+00, -3.1898e-08], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.1642e-09, + -1.8626e-09, 1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, + 6.9849e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 4.6566e-10, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 4.6566e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.6065e-08, ..., 5.8906e-08, + 0.0000e+00, 3.2363e-08]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0207, 0.0144, 0.0229, 0.0040, 0.0365, -0.0191, -0.0027, -0.0198, + -0.0282, 0.0064], device='cuda:0'), grad: tensor([-1.4389e-07, 9.3132e-10, 9.7789e-09, -1.1735e-07, -8.1491e-09, + 1.9558e-08, 5.8208e-09, 6.9849e-08, 9.0804e-09, 1.5763e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 220.54, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4360 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.16 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.0022, 0.0110, 0.0658, ..., 0.0891, -0.2911, -0.1850], + [ 0.0579, -0.0185, -0.0551, ..., -0.3401, 0.0525, -0.1116], + [-0.0111, -0.0048, 0.2658, ..., -0.3438, -0.1630, -0.1898], + ..., + [-0.0242, 0.0062, -0.2424, ..., -0.0768, 0.1461, 0.0745], + [-0.0329, -0.0202, -0.2167, ..., -0.3356, 0.1281, -0.2343], + [ 0.0154, -0.0258, -0.1214, ..., 0.0404, -0.2001, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.8022e-08, ..., 0.0000e+00, + 6.7288e-08, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + 3.2596e-09, 2.2817e-08], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 0.0000e+00, + 3.9581e-09, 2.3283e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 6.9849e-10, + -4.6566e-09, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, + 9.3132e-10, 3.0268e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., -9.3132e-10, + -4.6566e-09, -4.6333e-08]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0207, 0.0143, 0.0230, 0.0040, 0.0365, -0.0191, -0.0027, -0.0197, + -0.0283, 0.0064], device='cuda:0'), grad: tensor([ 5.1223e-07, 3.6089e-08, 1.5832e-08, -7.9162e-09, 1.9092e-08, + -1.5832e-08, -4.9965e-07, 9.0804e-09, 1.8626e-08, -8.1491e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 220.67, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4289 re_mapping 0.0026 re_causal 0.0093 /// teacc 99.13 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.0022, 0.0110, 0.0658, ..., 0.0891, -0.2912, -0.1851], + [ 0.0579, -0.0185, -0.0552, ..., -0.3402, 0.0526, -0.1117], + [-0.0111, -0.0048, 0.2659, ..., -0.3438, -0.1631, -0.1898], + ..., + [-0.0242, 0.0062, -0.2427, ..., -0.0768, 0.1461, 0.0746], + [-0.0329, -0.0202, -0.2168, ..., -0.3357, 0.1281, -0.2344], + [ 0.0154, -0.0258, -0.1215, ..., 0.0404, -0.2002, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.5879e-09, ..., -2.4214e-08, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-09, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -4.6566e-10, -4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, + -2.3283e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.3283e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0208, 0.0143, 0.0231, 0.0040, 0.0365, -0.0191, -0.0028, -0.0198, + -0.0283, 0.0064], device='cuda:0'), grad: tensor([-7.2177e-08, 3.7253e-09, 3.0734e-08, 7.4506e-09, 7.9162e-09, + -6.0536e-09, 2.1420e-08, -9.3132e-10, 6.0536e-09, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 220.58, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4224 re_mapping 0.0025 re_causal 0.0091 /// teacc 99.11 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.0022, 0.0110, 0.0659, ..., 0.0891, -0.2912, -0.1851], + [ 0.0579, -0.0185, -0.0552, ..., -0.3403, 0.0526, -0.1118], + [-0.0111, -0.0048, 0.2660, ..., -0.3439, -0.1632, -0.1898], + ..., + [-0.0242, 0.0062, -0.2428, ..., -0.0768, 0.1461, 0.0747], + [-0.0330, -0.0202, -0.2169, ..., -0.3357, 0.1281, -0.2344], + [ 0.0154, -0.0258, -0.1216, ..., 0.0404, -0.2003, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -1.3970e-09, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 4.1910e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 0.0000e+00, + 2.3283e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -1.3970e-08, -9.7789e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, + 5.5879e-09, 4.6566e-09]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0207, 0.0143, 0.0231, 0.0040, 0.0365, -0.0191, -0.0028, -0.0198, + -0.0284, 0.0064], device='cuda:0'), grad: tensor([ 4.6566e-09, 1.2107e-08, -6.0536e-09, -2.7940e-09, 1.7695e-08, + 4.6566e-10, -1.5832e-08, -3.4459e-08, 6.9849e-09, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 220.53, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4298 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.0022, 0.0110, 0.0660, ..., 0.0892, -0.2913, -0.1851], + [ 0.0579, -0.0185, -0.0552, ..., -0.3405, 0.0525, -0.1119], + [-0.0111, -0.0048, 0.2660, ..., -0.3440, -0.1633, -0.1899], + ..., + [-0.0242, 0.0062, -0.2429, ..., -0.0769, 0.1461, 0.0747], + [-0.0330, -0.0202, -0.2170, ..., -0.3357, 0.1281, -0.2345], + [ 0.0154, -0.0258, -0.1216, ..., 0.0404, -0.2004, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., -1.7695e-08, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 8.8476e-09, + 4.1910e-09, 2.0489e-08], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 4.6566e-09, + 2.1420e-08, 1.9558e-08], + ..., + [ 0.0000e+00, 0.0000e+00, -3.2596e-09, ..., 3.2596e-08, + -2.7940e-08, 4.6566e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 9.3132e-10, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., 3.1665e-08, + 4.6566e-10, 3.1199e-08]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0207, 0.0143, 0.0230, 0.0040, 0.0365, -0.0191, -0.0028, -0.0197, + -0.0284, 0.0064], device='cuda:0'), grad: tensor([-2.2352e-08, 1.1206e-05, 4.3288e-06, 1.6578e-07, -2.5146e-07, + -6.7521e-08, 2.6077e-08, -1.5497e-05, 2.0955e-08, 1.0151e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 221.03, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4279 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.11 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.0022, 0.0110, 0.0661, ..., 0.0892, -0.2913, -0.1851], + [ 0.0579, -0.0185, -0.0553, ..., -0.3406, 0.0525, -0.1119], + [-0.0111, -0.0048, 0.2661, ..., -0.3441, -0.1634, -0.1900], + ..., + [-0.0242, 0.0062, -0.2431, ..., -0.0770, 0.1461, 0.0748], + [-0.0330, -0.0202, -0.2171, ..., -0.3358, 0.1281, -0.2346], + [ 0.0154, -0.0258, -0.1217, ..., 0.0404, -0.2005, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7928e-07, ..., -2.0675e-07, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 6.0536e-09, + 1.3970e-09, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., 1.5367e-08, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 3.3993e-08, + 1.3970e-08, 5.0291e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 3.7253e-09, + -4.6566e-10, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 9.7789e-08, ..., 5.9139e-08, + -2.4214e-08, -9.4529e-08]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0207, 0.0142, 0.0231, 0.0042, 0.0365, -0.0192, -0.0029, -0.0197, + -0.0285, 0.0064], device='cuda:0'), grad: tensor([-6.3702e-07, 2.7474e-08, 5.1688e-08, 6.9849e-08, 7.3574e-08, + 2.3749e-08, 1.3830e-07, 1.3923e-07, 0.0000e+00, 1.1828e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 220.74, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4244 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.0022, 0.0110, 0.0662, ..., 0.0893, -0.2914, -0.1852], + [ 0.0579, -0.0185, -0.0553, ..., -0.3407, 0.0525, -0.1120], + [-0.0111, -0.0048, 0.2662, ..., -0.3441, -0.1635, -0.1901], + ..., + [-0.0242, 0.0062, -0.2432, ..., -0.0771, 0.1462, 0.0748], + [-0.0330, -0.0202, -0.2172, ..., -0.3359, 0.1281, -0.2347], + [ 0.0154, -0.0258, -0.1218, ..., 0.0404, -0.2005, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.8476e-09, ..., -1.3504e-08, + 3.2596e-09, -9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + -9.3132e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, -2.3283e-09, ..., 1.3970e-09, + 2.7940e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.2596e-09, + -6.5193e-09, -7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 1.5832e-08, + 0.0000e+00, 6.5193e-09]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0206, 0.0142, 0.0231, 0.0043, 0.0365, -0.0193, -0.0029, -0.0197, + -0.0285, 0.0065], device='cuda:0'), grad: tensor([ 1.3504e-08, -1.3970e-08, 9.3132e-09, 6.5193e-09, -1.3970e-08, + -1.6857e-07, 1.1874e-07, -2.0489e-08, 4.7963e-08, 3.4925e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 220.12, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4168 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.10 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.0022, 0.0110, 0.0663, ..., 0.0893, -0.2915, -0.1853], + [ 0.0579, -0.0185, -0.0553, ..., -0.3408, 0.0526, -0.1121], + [-0.0111, -0.0048, 0.2662, ..., -0.3442, -0.1636, -0.1902], + ..., + [-0.0242, 0.0062, -0.2433, ..., -0.0772, 0.1461, 0.0749], + [-0.0330, -0.0202, -0.2173, ..., -0.3359, 0.1282, -0.2348], + [ 0.0154, -0.0258, -0.1219, ..., 0.0404, -0.2006, 0.0243]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 3.2596e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, -9.7789e-09, ..., 0.0000e+00, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, + -5.1223e-09, -3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 0.0000e+00, + 5.1223e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0207, 0.0142, 0.0231, 0.0042, 0.0365, -0.0192, -0.0029, -0.0197, + -0.0286, 0.0065], device='cuda:0'), grad: tensor([ 6.0536e-09, 1.2107e-08, -1.0710e-08, -3.1665e-08, 2.3283e-09, + 6.9849e-09, -7.9162e-09, -6.9849e-09, 2.8871e-08, 6.5193e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 220.71, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.3986 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.09 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.0022, 0.0110, 0.0663, ..., 0.0893, -0.2915, -0.1853], + [ 0.0579, -0.0185, -0.0553, ..., -0.3408, 0.0525, -0.1122], + [-0.0111, -0.0048, 0.2663, ..., -0.3443, -0.1638, -0.1903], + ..., + [-0.0242, 0.0062, -0.2436, ..., -0.0773, 0.1461, 0.0750], + [-0.0330, -0.0202, -0.2174, ..., -0.3359, 0.1282, -0.2348], + [ 0.0154, -0.0258, -0.1220, ..., 0.0404, -0.2007, 0.0244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., -4.6566e-09, + 0.0000e+00, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 7.9162e-09, + -1.3970e-09, 1.0245e-08], + [ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., 0.0000e+00, + 1.3970e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.1420e-08, + 4.6566e-10, 2.8871e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, + -1.3970e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7474e-08, + 0.0000e+00, 3.6322e-08]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0207, 0.0142, 0.0231, 0.0042, 0.0365, -0.0192, -0.0029, -0.0197, + -0.0286, 0.0065], device='cuda:0'), grad: tensor([-1.2107e-08, 2.1420e-08, -2.0955e-08, -6.6124e-08, -1.5739e-07, + 6.9849e-08, 3.3062e-08, 6.6590e-08, 5.5879e-09, 7.3109e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 221.00, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4078 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.10 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.0022, 0.0110, 0.0665, ..., 0.0894, -0.2916, -0.1854], + [ 0.0579, -0.0185, -0.0554, ..., -0.3410, 0.0525, -0.1123], + [-0.0111, -0.0048, 0.2663, ..., -0.3444, -0.1639, -0.1904], + ..., + [-0.0242, 0.0062, -0.2436, ..., -0.0774, 0.1462, 0.0750], + [-0.0330, -0.0202, -0.2175, ..., -0.3360, 0.1282, -0.2349], + [ 0.0154, -0.0258, -0.1222, ..., 0.0404, -0.2007, 0.0244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.0338e-08, ..., -9.4064e-08, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.0245e-08, + -3.7253e-08, 6.5193e-09], + [ 0.0000e+00, 0.0000e+00, -6.0536e-09, ..., 9.3132e-09, + 7.4506e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 3.7253e-09, + 2.7940e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 1.5832e-08, + 0.0000e+00, 8.8476e-09]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0206, 0.0141, 0.0231, 0.0042, 0.0365, -0.0192, -0.0030, -0.0196, + -0.0286, 0.0065], device='cuda:0'), grad: tensor([-4.5868e-07, -5.6345e-08, 4.4238e-08, 1.0245e-08, -3.7253e-08, + -1.3970e-09, 3.7905e-07, 2.7008e-08, 3.2131e-08, 6.0536e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 220.99, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4040 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.10 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.0022, 0.0110, 0.0666, ..., 0.0894, -0.2916, -0.1854], + [ 0.0579, -0.0185, -0.0554, ..., -0.3410, 0.0525, -0.1124], + [-0.0111, -0.0048, 0.2664, ..., -0.3444, -0.1641, -0.1905], + ..., + [-0.0242, 0.0062, -0.2438, ..., -0.0775, 0.1462, 0.0751], + [-0.0330, -0.0202, -0.2176, ..., -0.3360, 0.1283, -0.2350], + [ 0.0154, -0.0258, -0.1222, ..., 0.0404, -0.2008, 0.0244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 4.6100e-07, 2.4540e-07], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 6.0536e-09, 3.2596e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + -4.8103e-07, -2.5099e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.1176e-08, 3.2596e-09]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0206, 0.0142, 0.0230, 0.0042, 0.0365, -0.0192, -0.0030, -0.0196, + -0.0286, 0.0065], device='cuda:0'), grad: tensor([ 6.5193e-09, 1.1595e-06, 2.0023e-08, 7.4506e-09, 4.1910e-09, + 9.3132e-10, -9.7789e-09, -1.2033e-06, 5.1223e-09, 2.0489e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 220.72, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4350 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.0022, 0.0110, 0.0667, ..., 0.0895, -0.2917, -0.1855], + [ 0.0579, -0.0185, -0.0554, ..., -0.3413, 0.0525, -0.1125], + [-0.0111, -0.0048, 0.2664, ..., -0.3445, -0.1642, -0.1906], + ..., + [-0.0242, 0.0062, -0.2439, ..., -0.0776, 0.1462, 0.0751], + [-0.0330, -0.0202, -0.2177, ..., -0.3361, 0.1283, -0.2350], + [ 0.0154, -0.0258, -0.1223, ..., 0.0404, -0.2009, 0.0244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., -4.1910e-09, + 0.0000e+00, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 1.8626e-09, 1.2573e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + -3.2596e-09, -4.0978e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -4.1910e-09, + 9.3132e-10, 8.8476e-09]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0206, 0.0141, 0.0230, 0.0043, 0.0365, -0.0193, -0.0030, -0.0196, + -0.0286, 0.0065], device='cuda:0'), grad: tensor([-8.8476e-09, 4.0047e-08, 7.9162e-09, 1.0710e-08, 2.0489e-08, + 8.8476e-09, 5.1223e-09, -1.3085e-07, 7.4506e-09, 4.8894e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 220.48, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4502 re_mapping 0.0025 re_causal 0.0095 /// teacc 99.11 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.0022, 0.0110, 0.0668, ..., 0.0895, -0.2917, -0.1855], + [ 0.0579, -0.0185, -0.0554, ..., -0.3413, 0.0525, -0.1126], + [-0.0111, -0.0048, 0.2664, ..., -0.3446, -0.1644, -0.1907], + ..., + [-0.0242, 0.0062, -0.2440, ..., -0.0776, 0.1462, 0.0752], + [-0.0330, -0.0202, -0.2177, ..., -0.3361, 0.1284, -0.2351], + [ 0.0154, -0.0258, -0.1224, ..., 0.0405, -0.2009, 0.0245]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6543e-08, ..., -2.2817e-08, + 0.0000e+00, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 1.3970e-09, + 9.3132e-10, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., 2.7940e-09, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 3.2596e-09, + -1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.8626e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.0710e-08, ..., 1.3039e-08, + 0.0000e+00, 4.1910e-09]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0205, 0.0141, 0.0229, 0.0043, 0.0365, -0.0193, -0.0030, -0.0196, + -0.0286, 0.0065], device='cuda:0'), grad: tensor([-7.0781e-08, 1.2107e-08, 2.4680e-08, -2.6077e-08, -2.3283e-09, + 6.5193e-09, 6.5193e-09, 1.3039e-08, 1.1176e-08, 3.5390e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 220.92, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3993 re_mapping 0.0025 re_causal 0.0090 /// teacc 99.09 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.0022, 0.0110, 0.0669, ..., 0.0896, -0.2918, -0.1856], + [ 0.0579, -0.0185, -0.0554, ..., -0.3415, 0.0525, -0.1127], + [-0.0111, -0.0048, 0.2664, ..., -0.3447, -0.1646, -0.1908], + ..., + [-0.0242, 0.0062, -0.2440, ..., -0.0778, 0.1462, 0.0753], + [-0.0330, -0.0202, -0.2178, ..., -0.3362, 0.1284, -0.2352], + [ 0.0154, -0.0258, -0.1226, ..., 0.0405, -0.2010, 0.0245]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, + 9.7789e-09, 9.7789e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 2.9337e-08, 1.4435e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.3970e-09, + -4.3306e-08, -1.5832e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.3970e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 1.3970e-09]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0204, 0.0141, 0.0228, 0.0044, 0.0365, -0.0193, -0.0030, -0.0195, + -0.0287, 0.0066], device='cuda:0'), grad: tensor([ 5.1223e-09, 4.4703e-08, 1.1874e-07, -3.8091e-07, -1.6764e-08, + 1.3039e-08, 4.6566e-10, 1.9092e-07, 2.3749e-08, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 220.34, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4232 re_mapping 0.0025 re_causal 0.0090 /// teacc 99.12 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.0022, 0.0110, 0.0671, ..., 0.0897, -0.2919, -0.1856], + [ 0.0579, -0.0185, -0.0555, ..., -0.3417, 0.0525, -0.1128], + [-0.0111, -0.0048, 0.2665, ..., -0.3448, -0.1648, -0.1910], + ..., + [-0.0242, 0.0062, -0.2442, ..., -0.0778, 0.1462, 0.0754], + [-0.0330, -0.0202, -0.2179, ..., -0.3362, 0.1284, -0.2353], + [ 0.0154, -0.0258, -0.1226, ..., 0.0405, -0.2011, 0.0245]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8161e-08, ..., -1.1642e-08, + 0.0000e+00, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 1.8626e-09, + 4.6566e-10, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, -2.2352e-08, ..., 3.2596e-09, + 2.7940e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.2596e-09, + -2.3283e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 2.3283e-09, + 0.0000e+00, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 7.9162e-09, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0203, 0.0141, 0.0227, 0.0046, 0.0365, -0.0193, -0.0030, -0.0195, + -0.0288, 0.0066], device='cuda:0'), grad: tensor([-4.1910e-08, 3.2596e-08, -3.1199e-08, 3.7253e-09, -2.3283e-09, + 3.2596e-09, 1.0245e-08, 2.4214e-08, 1.3039e-08, -5.5879e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 220.59, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4556 re_mapping 0.0024 re_causal 0.0094 /// teacc 99.12 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.0022, 0.0110, 0.0671, ..., 0.0898, -0.2919, -0.1856], + [ 0.0578, -0.0185, -0.0555, ..., -0.3418, 0.0525, -0.1129], + [-0.0111, -0.0048, 0.2665, ..., -0.3449, -0.1649, -0.1911], + ..., + [-0.0242, 0.0062, -0.2444, ..., -0.0779, 0.1462, 0.0755], + [-0.0330, -0.0202, -0.2181, ..., -0.3363, 0.1284, -0.2354], + [ 0.0154, -0.0258, -0.1228, ..., 0.0405, -0.2012, 0.0245]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -1.3970e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 2.7940e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, -2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 9.3132e-10, 9.3132e-10]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0204, 0.0140, 0.0227, 0.0047, 0.0365, -0.0193, -0.0030, -0.0195, + -0.0289, 0.0066], device='cuda:0'), grad: tensor([-4.6566e-10, 9.7789e-09, 2.7940e-09, 5.1223e-09, -2.7940e-09, + -1.3039e-08, 7.9162e-09, -7.9162e-09, 1.8626e-09, 5.1223e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 220.54, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4049 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.11 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.0022, 0.0110, 0.0672, ..., 0.0898, -0.2920, -0.1857], + [ 0.0578, -0.0185, -0.0556, ..., -0.3419, 0.0526, -0.1130], + [-0.0111, -0.0048, 0.2667, ..., -0.3449, -0.1652, -0.1912], + ..., + [-0.0242, 0.0062, -0.2445, ..., -0.0780, 0.1462, 0.0755], + [-0.0330, -0.0202, -0.2182, ..., -0.3363, 0.1284, -0.2354], + [ 0.0154, -0.0258, -0.1229, ..., 0.0405, -0.2013, 0.0245]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 2.2352e-08, 3.2131e-08], + [ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 0.0000e+00, + 4.1910e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + -2.7008e-08, -3.0268e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.1176e-08, + 8.8476e-09, 2.5611e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + 4.6566e-10, -1.1176e-08]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0204, 0.0140, 0.0227, 0.0047, 0.0365, -0.0193, -0.0029, -0.0195, + -0.0290, 0.0066], device='cuda:0'), grad: tensor([ 6.9849e-09, 9.9652e-08, 6.5193e-09, 2.3749e-08, -5.5879e-08, + 2.6077e-08, -4.0047e-08, -1.0291e-07, 1.3318e-07, -1.0105e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 220.61, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4562 re_mapping 0.0024 re_causal 0.0093 /// teacc 99.10 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.0022, 0.0110, 0.0673, ..., 0.0897, -0.2921, -0.1859], + [ 0.0578, -0.0185, -0.0556, ..., -0.3420, 0.0526, -0.1130], + [-0.0111, -0.0048, 0.2667, ..., -0.3450, -0.1653, -0.1913], + ..., + [-0.0242, 0.0062, -0.2446, ..., -0.0781, 0.1462, 0.0756], + [-0.0330, -0.0202, -0.2183, ..., -0.3363, 0.1284, -0.2355], + [ 0.0154, -0.0258, -0.1230, ..., 0.0405, -0.2014, 0.0246]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 1.2713e-07, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 0.0000e+00, + 3.9116e-08, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., 0.0000e+00, + -1.8673e-07, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.3970e-09, + 5.5879e-09, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.1910e-09, + 9.3132e-10, -2.8871e-08]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0205, 0.0140, 0.0227, 0.0047, 0.0365, -0.0193, -0.0028, -0.0195, + -0.0290, 0.0066], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.8778e-07, 1.1129e-07, -8.3819e-09, 5.4482e-08, + 2.5146e-08, 0.0000e+00, -3.9395e-07, 5.8673e-08, -1.2061e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 220.51, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4098 re_mapping 0.0024 re_causal 0.0090 /// teacc 99.10 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.0022, 0.0110, 0.0675, ..., 0.0897, -0.2921, -0.1860], + [ 0.0578, -0.0185, -0.0557, ..., -0.3422, 0.0525, -0.1131], + [-0.0111, -0.0048, 0.2668, ..., -0.3451, -0.1655, -0.1914], + ..., + [-0.0242, 0.0062, -0.2447, ..., -0.0781, 0.1463, 0.0756], + [-0.0330, -0.0202, -0.2184, ..., -0.3364, 0.1284, -0.2355], + [ 0.0154, -0.0258, -0.1231, ..., 0.0406, -0.2015, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.1886e-08, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 0.0000e+00, + -4.1910e-09, 1.7695e-08], + [ 0.0000e+00, 0.0000e+00, -9.7789e-08, ..., 0.0000e+00, + 6.0536e-09, -2.3749e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 0.0000e+00, + -2.7940e-09, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 4.6566e-10, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 1.3970e-09, 3.7253e-09]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0205, 0.0139, 0.0227, 0.0047, 0.0365, -0.0194, -0.0027, -0.0194, + -0.0291, 0.0067], device='cuda:0'), grad: tensor([ 2.6543e-08, 5.2620e-08, -1.4529e-07, 2.2817e-08, 7.4506e-09, + 1.3970e-09, 0.0000e+00, 2.3283e-08, 1.3504e-08, 1.8161e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 220.51, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4175 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.0022, 0.0110, 0.0676, ..., 0.0897, -0.2922, -0.1860], + [ 0.0578, -0.0185, -0.0557, ..., -0.3422, 0.0525, -0.1132], + [-0.0111, -0.0048, 0.2668, ..., -0.3452, -0.1657, -0.1915], + ..., + [-0.0242, 0.0062, -0.2447, ..., -0.0782, 0.1463, 0.0757], + [-0.0330, -0.0202, -0.2185, ..., -0.3364, 0.1284, -0.2356], + [ 0.0154, -0.0258, -0.1232, ..., 0.0406, -0.2016, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 6.0536e-09], + [ 0.0000e+00, 0.0000e+00, -1.0990e-07, ..., 0.0000e+00, + 2.7940e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 1.0803e-07, ..., 4.6566e-10, + -3.7253e-09, -6.5193e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 4.6566e-10, 1.8626e-09]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0205, 0.0139, 0.0226, 0.0049, 0.0364, -0.0195, -0.0027, -0.0194, + -0.0292, 0.0067], device='cuda:0'), grad: tensor([ 1.0710e-08, 1.6764e-08, -2.5984e-07, -4.1910e-09, 5.1223e-09, + 8.3819e-09, -2.7940e-09, 2.4820e-07, -1.6298e-08, 4.6566e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 220.71, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4157 re_mapping 0.0024 re_causal 0.0090 /// teacc 99.11 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.0022, 0.0110, 0.0676, ..., 0.0897, -0.2923, -0.1862], + [ 0.0578, -0.0185, -0.0558, ..., -0.3423, 0.0525, -0.1132], + [-0.0111, -0.0048, 0.2668, ..., -0.3453, -0.1658, -0.1917], + ..., + [-0.0242, 0.0062, -0.2448, ..., -0.0783, 0.1463, 0.0758], + [-0.0330, -0.0202, -0.2186, ..., -0.3365, 0.1284, -0.2357], + [ 0.0154, -0.0258, -0.1233, ..., 0.0406, -0.2017, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 7.0781e-08, ..., 0.0000e+00, + 2.5146e-07, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 3.7719e-08, ..., 0.0000e+00, + 1.4622e-07, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, -1.2387e-07, ..., 0.0000e+00, + -4.3679e-07, -1.2573e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 2.3283e-09, 4.6566e-10]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0206, 0.0138, 0.0225, 0.0050, 0.0364, -0.0195, -0.0025, -0.0193, + -0.0293, 0.0067], device='cuda:0'), grad: tensor([ 1.1642e-08, 1.0012e-07, 7.9209e-07, 9.3598e-08, 2.3283e-09, + 1.5367e-08, 4.6566e-10, -1.0207e-06, -4.6566e-10, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 220.72, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4298 re_mapping 0.0024 re_causal 0.0091 /// teacc 99.13 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.0022, 0.0110, 0.0677, ..., 0.0897, -0.2923, -0.1862], + [ 0.0578, -0.0185, -0.0558, ..., -0.3424, 0.0525, -0.1133], + [-0.0111, -0.0048, 0.2669, ..., -0.3454, -0.1660, -0.1918], + ..., + [-0.0242, 0.0062, -0.2450, ..., -0.0783, 0.1464, 0.0759], + [-0.0330, -0.0202, -0.2188, ..., -0.3365, 0.1284, -0.2357], + [ 0.0154, -0.0258, -0.1234, ..., 0.0406, -0.2019, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 2.3283e-09, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + -3.2596e-09, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 4.6566e-10, 2.3283e-09]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0206, 0.0138, 0.0225, 0.0051, 0.0364, -0.0197, -0.0025, -0.0192, + -0.0294, 0.0067], device='cuda:0'), grad: tensor([ 4.1910e-09, 1.0710e-08, 4.6566e-09, -4.6566e-09, -6.5193e-09, + -1.1642e-08, -1.8626e-09, -5.5879e-09, 8.3819e-09, 8.3819e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 220.24, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4207 re_mapping 0.0023 re_causal 0.0091 /// teacc 99.10 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.0022, 0.0110, 0.0678, ..., 0.0898, -0.2924, -0.1863], + [ 0.0578, -0.0185, -0.0558, ..., -0.3425, 0.0525, -0.1133], + [-0.0111, -0.0048, 0.2670, ..., -0.3454, -0.1661, -0.1919], + ..., + [-0.0242, 0.0062, -0.2451, ..., -0.0784, 0.1464, 0.0759], + [-0.0330, -0.0202, -0.2189, ..., -0.3367, 0.1284, -0.2359], + [ 0.0154, -0.0258, -0.1236, ..., 0.0406, -0.2021, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 6.9849e-10, + 2.7940e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, + -6.7521e-09, -4.4238e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 6.9849e-10, 3.4925e-09]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0205, 0.0138, 0.0225, 0.0052, 0.0364, -0.0198, -0.0024, -0.0192, + -0.0297, 0.0067], device='cuda:0'), grad: tensor([ 2.3050e-08, 1.2573e-08, 4.1910e-09, 8.3819e-09, -1.0012e-08, + -1.6298e-08, -2.8173e-08, -1.7695e-08, 1.2573e-08, 1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 220.51, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4114 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.11 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.0022, 0.0110, 0.0679, ..., 0.0899, -0.2925, -0.1863], + [ 0.0578, -0.0185, -0.0559, ..., -0.3425, 0.0525, -0.1134], + [-0.0111, -0.0048, 0.2672, ..., -0.3455, -0.1663, -0.1921], + ..., + [-0.0242, 0.0062, -0.2454, ..., -0.0785, 0.1464, 0.0760], + [-0.0330, -0.0202, -0.2191, ..., -0.3367, 0.1284, -0.2360], + [ 0.0154, -0.0258, -0.1237, ..., 0.0406, -0.2022, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, -2.3283e-10, ..., 0.0000e+00, + -1.7462e-08, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 3.7951e-08, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.0710e-08, -2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 2.3283e-10]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0205, 0.0138, 0.0226, 0.0053, 0.0364, -0.0198, -0.0024, -0.0192, + -0.0298, 0.0067], device='cuda:0'), grad: tensor([ 2.3283e-09, -7.6368e-08, 7.5903e-08, -5.4948e-08, 4.8894e-09, + 4.4238e-09, 1.3970e-09, 4.7497e-08, -9.3132e-10, 1.8626e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 220.70, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4369 re_mapping 0.0024 re_causal 0.0093 /// teacc 99.12 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.0022, 0.0110, 0.0680, ..., 0.0900, -0.2925, -0.1863], + [ 0.0578, -0.0185, -0.0560, ..., -0.3426, 0.0526, -0.1133], + [-0.0111, -0.0048, 0.2673, ..., -0.3456, -0.1664, -0.1922], + ..., + [-0.0242, 0.0062, -0.2454, ..., -0.0785, 0.1463, 0.0760], + [-0.0330, -0.0202, -0.2191, ..., -0.3368, 0.1284, -0.2361], + [ 0.0154, -0.0258, -0.1238, ..., 0.0406, -0.2023, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.0955e-09, + 1.4901e-08, 1.3271e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6298e-09, ..., 1.1409e-08, + -3.6089e-08, -6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, + 2.3283e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.7229e-08, + 1.9092e-08, 4.6799e-08]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0204, 0.0138, 0.0226, 0.0052, 0.0364, -0.0198, -0.0023, -0.0193, + -0.0298, 0.0067], device='cuda:0'), grad: tensor([ 3.4925e-09, 3.4226e-08, 2.7940e-09, 5.8208e-09, -6.5891e-08, + 3.2596e-09, -2.5611e-09, -4.1910e-08, -5.3551e-09, 7.6136e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 221.06, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4346 re_mapping 0.0024 re_causal 0.0093 /// teacc 99.12 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.0022, 0.0110, 0.0681, ..., 0.0899, -0.2926, -0.1864], + [ 0.0578, -0.0185, -0.0561, ..., -0.3427, 0.0526, -0.1133], + [-0.0111, -0.0048, 0.2673, ..., -0.3457, -0.1666, -0.1924], + ..., + [-0.0242, 0.0062, -0.2455, ..., -0.0786, 0.1463, 0.0760], + [-0.0330, -0.0202, -0.2192, ..., -0.3368, 0.1285, -0.2362], + [ 0.0154, -0.0258, -0.1239, ..., 0.0406, -0.2025, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.2573e-08, ..., 0.0000e+00, + 5.5879e-09, 1.1642e-09], + [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 9.7789e-09, + 4.1910e-09, 3.1665e-08], + [ 0.0000e+00, 0.0000e+00, 2.3516e-08, ..., 2.3283e-10, + 2.7940e-09, 2.3283e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + -4.8894e-09, 6.9849e-10], + [ 0.0000e+00, 0.0000e+00, 1.6298e-09, ..., 0.0000e+00, + 6.9849e-10, 2.3283e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 9.3132e-10, 1.8626e-09]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0205, 0.0139, 0.0226, 0.0052, 0.0363, -0.0198, -0.0021, -0.0193, + -0.0299, 0.0067], device='cuda:0'), grad: tensor([ 1.7975e-07, 7.9861e-08, 1.5507e-07, 1.1176e-08, -3.8184e-08, + 5.7509e-08, -4.4750e-07, -8.6147e-09, 2.1653e-08, 9.5461e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 220.85, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4043 re_mapping 0.0024 re_causal 0.0088 /// teacc 99.11 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.0022, 0.0110, 0.0682, ..., 0.0900, -0.2927, -0.1864], + [ 0.0578, -0.0185, -0.0562, ..., -0.3428, 0.0526, -0.1133], + [-0.0111, -0.0048, 0.2674, ..., -0.3458, -0.1668, -0.1925], + ..., + [-0.0242, 0.0062, -0.2456, ..., -0.0787, 0.1463, 0.0760], + [-0.0330, -0.0202, -0.2193, ..., -0.3369, 0.1285, -0.2363], + [ 0.0154, -0.0258, -0.1241, ..., 0.0406, -0.2026, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 0.0000e+00, + 2.7940e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, + 6.7521e-09, 5.8208e-09], + [ 0.0000e+00, 0.0000e+00, -1.0477e-08, ..., 2.3283e-10, + 8.1491e-09, 6.5193e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.0955e-09, ..., 2.3283e-10, + -3.7020e-08, -2.3050e-08], + [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, + -4.6566e-10, 1.6298e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.1642e-09, + 6.5193e-09, 4.4238e-09]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0205, 0.0138, 0.0226, 0.0052, 0.0363, -0.0198, -0.0020, -0.0193, + -0.0300, 0.0067], device='cuda:0'), grad: tensor([ 4.2608e-08, 2.3749e-08, 1.0943e-08, -3.9348e-08, 6.0536e-09, + 6.6124e-08, -7.2177e-09, -1.0361e-07, -1.3504e-08, 3.1432e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 220.45, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4435 re_mapping 0.0024 re_causal 0.0094 /// teacc 99.10 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.0022, 0.0110, 0.0683, ..., 0.0901, -0.2928, -0.1864], + [ 0.0578, -0.0185, -0.0562, ..., -0.3429, 0.0526, -0.1133], + [-0.0111, -0.0048, 0.2674, ..., -0.3459, -0.1670, -0.1926], + ..., + [-0.0242, 0.0062, -0.2457, ..., -0.0788, 0.1464, 0.0761], + [-0.0330, -0.0202, -0.2194, ..., -0.3370, 0.1286, -0.2364], + [ 0.0154, -0.0258, -0.1243, ..., 0.0406, -0.2028, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -3.2596e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, + 5.1223e-09, 2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -4.6566e-09, -1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0205, 0.0138, 0.0224, 0.0054, 0.0363, -0.0200, -0.0020, -0.0192, + -0.0299, 0.0067], device='cuda:0'), grad: tensor([-9.7789e-09, 1.5367e-08, 2.3283e-09, -1.0617e-07, 1.3970e-09, + 1.0571e-07, 2.3283e-09, -9.3132e-09, 3.2596e-09, 2.7940e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 220.37, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4099 re_mapping 0.0024 re_causal 0.0088 /// teacc 99.09 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.0022, 0.0110, 0.0684, ..., 0.0902, -0.2929, -0.1864], + [ 0.0578, -0.0185, -0.0562, ..., -0.3430, 0.0526, -0.1134], + [-0.0111, -0.0048, 0.2675, ..., -0.3459, -0.1672, -0.1929], + ..., + [-0.0242, 0.0062, -0.2457, ..., -0.0789, 0.1464, 0.0762], + [-0.0330, -0.0202, -0.2195, ..., -0.3370, 0.1287, -0.2364], + [ 0.0154, -0.0258, -0.1244, ..., 0.0405, -0.2030, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., -2.0489e-08, + 0.0000e+00, -1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.5193e-09, + 1.3970e-09, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 5.5879e-09, + 0.0000e+00, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.0245e-08, + -1.3970e-09, 1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, + 4.6566e-10, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 1.0245e-08, + -1.8626e-09, 9.3132e-10]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0205, 0.0137, 0.0224, 0.0054, 0.0364, -0.0200, -0.0019, -0.0192, + -0.0299, 0.0066], device='cuda:0'), grad: tensor([-4.9826e-08, 2.1886e-08, 1.4901e-08, 3.7253e-09, -5.8673e-08, + -2.3283e-09, 2.7474e-08, 2.1420e-08, 1.0710e-08, 1.9092e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 220.58, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4241 re_mapping 0.0023 re_causal 0.0088 /// teacc 99.10 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.0022, 0.0110, 0.0685, ..., 0.0902, -0.2930, -0.1865], + [ 0.0578, -0.0185, -0.0562, ..., -0.3432, 0.0526, -0.1136], + [-0.0111, -0.0048, 0.2675, ..., -0.3460, -0.1675, -0.1931], + ..., + [-0.0242, 0.0062, -0.2458, ..., -0.0790, 0.1464, 0.0763], + [-0.0330, -0.0202, -0.2196, ..., -0.3371, 0.1288, -0.2366], + [ 0.0154, -0.0258, -0.1244, ..., 0.0406, -0.2031, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 6.5193e-09, + 9.3132e-10, 8.8476e-09], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 2.7940e-09, + -3.8184e-08, 9.3132e-09], + [ 0.0000e+00, 0.0000e+00, -2.7241e-07, ..., 9.3132e-10, + 2.7940e-08, 1.3970e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 2.3982e-07, ..., 2.7940e-09, + 7.9162e-09, -2.7940e-09], + [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, + -4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, + 2.7940e-09, 6.5193e-09]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0206, 0.0137, 0.0223, 0.0055, 0.0363, -0.0201, -0.0018, -0.0191, + -0.0299, 0.0067], device='cuda:0'), grad: tensor([ 2.1420e-08, -5.1688e-08, -6.1933e-07, 8.3819e-09, -1.0245e-07, + -6.2399e-08, 1.3458e-07, 6.3702e-07, 3.0268e-08, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 220.76, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3947 re_mapping 0.0023 re_causal 0.0087 /// teacc 99.10 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.0022, 0.0110, 0.0686, ..., 0.0903, -0.2931, -0.1865], + [ 0.0578, -0.0185, -0.0563, ..., -0.3433, 0.0526, -0.1136], + [-0.0111, -0.0048, 0.2676, ..., -0.3461, -0.1677, -0.1931], + ..., + [-0.0242, 0.0062, -0.2459, ..., -0.0792, 0.1465, 0.0764], + [-0.0330, -0.0202, -0.2197, ..., -0.3372, 0.1288, -0.2366], + [ 0.0154, -0.0258, -0.1246, ..., 0.0406, -0.2032, 0.0248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.3097e-07, 4.5635e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.9697e-07, 4.6566e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -1.7835e-07, -2.8592e-07], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 3.7253e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 1.9837e-07, 2.2538e-07]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0206, 0.0137, 0.0222, 0.0056, 0.0363, -0.0202, -0.0018, -0.0191, + -0.0300, 0.0067], device='cuda:0'), grad: tensor([ 3.2596e-09, -2.0489e-06, 1.7351e-06, 2.2817e-08, -6.0536e-09, + 5.5879e-09, 1.4435e-08, -4.1677e-07, 1.4901e-08, 6.8219e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 220.79, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4269 re_mapping 0.0024 re_causal 0.0090 /// teacc 99.10 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.0022, 0.0110, 0.0687, ..., 0.0904, -0.2932, -0.1865], + [ 0.0578, -0.0185, -0.0564, ..., -0.3435, 0.0526, -0.1138], + [-0.0111, -0.0048, 0.2677, ..., -0.3462, -0.1680, -0.1932], + ..., + [-0.0242, 0.0062, -0.2460, ..., -0.0793, 0.1465, 0.0765], + [-0.0330, -0.0202, -0.2199, ..., -0.3372, 0.1289, -0.2367], + [ 0.0154, -0.0258, -0.1247, ..., 0.0406, -0.2034, 0.0248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.3970e-09, + 0.0000e+00, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.0536e-09, + 9.3132e-10, 1.3039e-08], + [ 0.0000e+00, 0.0000e+00, -1.2573e-08, ..., 4.6566e-10, + 1.8626e-09, 9.3132e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 1.2573e-08, ..., 1.5832e-08, + -1.0245e-08, 6.9849e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8626e-09, + 9.3132e-10, 3.2596e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.4901e-08, + 7.4506e-09, 3.0268e-08]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0205, 0.0136, 0.0222, 0.0056, 0.0363, -0.0203, -0.0018, -0.0190, + -0.0300, 0.0067], device='cuda:0'), grad: tensor([ 5.5879e-09, 2.5146e-08, -2.2817e-08, 1.9092e-08, -1.0431e-07, + -2.5146e-08, 3.2596e-09, 2.9802e-08, 1.2107e-08, 6.7987e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 220.66, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4132 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.10 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.0022, 0.0110, 0.0688, ..., 0.0905, -0.2933, -0.1865], + [ 0.0578, -0.0185, -0.0564, ..., -0.3436, 0.0526, -0.1139], + [-0.0111, -0.0048, 0.2678, ..., -0.3463, -0.1681, -0.1933], + ..., + [-0.0242, 0.0062, -0.2461, ..., -0.0794, 0.1465, 0.0766], + [-0.0330, -0.0202, -0.2200, ..., -0.3372, 0.1289, -0.2367], + [ 0.0154, -0.0258, -0.1248, ..., 0.0406, -0.2035, 0.0248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -1.3970e-09, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.7789e-09, 7.4506e-09], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, + 4.6566e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.0489e-08, -1.6298e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 9.3132e-10, 1.3970e-09]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0204, 0.0136, 0.0222, 0.0056, 0.0363, -0.0203, -0.0019, -0.0190, + -0.0301, 0.0067], device='cuda:0'), grad: tensor([-8.8476e-09, 2.6543e-08, 1.7229e-08, 1.1176e-08, 4.6566e-10, + 4.6566e-10, 2.3283e-09, -5.8208e-08, 1.8626e-09, 7.9162e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 220.93, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4447 re_mapping 0.0024 re_causal 0.0092 /// teacc 99.08 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.0022, 0.0110, 0.0689, ..., 0.0904, -0.2934, -0.1867], + [ 0.0578, -0.0185, -0.0564, ..., -0.3436, 0.0526, -0.1139], + [-0.0111, -0.0048, 0.2679, ..., -0.3463, -0.1682, -0.1933], + ..., + [-0.0242, 0.0062, -0.2463, ..., -0.0795, 0.1465, 0.0766], + [-0.0330, -0.0202, -0.2202, ..., -0.3373, 0.1289, -0.2368], + [ 0.0154, -0.0258, -0.1249, ..., 0.0406, -0.2036, 0.0248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, + 1.8626e-09, 1.8626e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 8.8476e-09, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, + -1.3970e-09, 4.6566e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0205, 0.0136, 0.0222, 0.0056, 0.0363, -0.0203, -0.0017, -0.0190, + -0.0302, 0.0067], device='cuda:0'), grad: tensor([ 1.8626e-09, 6.9849e-09, 2.1886e-08, -1.3970e-09, -9.7789e-09, + 1.8626e-09, 8.3819e-09, 9.3132e-09, -2.0489e-08, -2.3283e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 220.48, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4308 re_mapping 0.0024 re_causal 0.0092 /// teacc 99.08 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.0022, 0.0110, 0.0689, ..., 0.0904, -0.2934, -0.1868], + [ 0.0577, -0.0185, -0.0564, ..., -0.3437, 0.0525, -0.1141], + [-0.0111, -0.0048, 0.2679, ..., -0.3464, -0.1683, -0.1934], + ..., + [-0.0242, 0.0062, -0.2465, ..., -0.0797, 0.1467, 0.0767], + [-0.0330, -0.0202, -0.2202, ..., -0.3374, 0.1289, -0.2370], + [ 0.0154, -0.0258, -0.1249, ..., 0.0407, -0.2037, 0.0249]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-4.6566e-10, 0.0000e+00, -4.6566e-10, ..., 0.0000e+00, + -4.6566e-09, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 4.6566e-10], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 4.6566e-10]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0206, 0.0134, 0.0221, 0.0056, 0.0362, -0.0203, -0.0016, -0.0187, + -0.0303, 0.0068], device='cuda:0'), grad: tensor([ 1.8626e-09, -9.7789e-09, 2.7940e-09, 5.1223e-09, 1.8626e-09, + -3.2596e-09, -1.8626e-09, 2.3283e-09, 6.0536e-09, 1.3970e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 220.94, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4462 re_mapping 0.0023 re_causal 0.0092 /// teacc 99.08 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.0022, 0.0110, 0.0690, ..., 0.0905, -0.2935, -0.1868], + [ 0.0578, -0.0185, -0.0565, ..., -0.3438, 0.0525, -0.1141], + [-0.0111, -0.0048, 0.2680, ..., -0.3465, -0.1686, -0.1937], + ..., + [-0.0242, 0.0062, -0.2467, ..., -0.0799, 0.1467, 0.0767], + [-0.0330, -0.0202, -0.2203, ..., -0.3374, 0.1290, -0.2370], + [ 0.0154, -0.0258, -0.1251, ..., 0.0406, -0.2038, 0.0249]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -5.5879e-09, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 1.8626e-09, 1.8626e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.7695e-08, -1.9558e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.1910e-09, ..., 5.5879e-09, + 3.7253e-09, 4.1910e-09]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0206, 0.0133, 0.0220, 0.0056, 0.0363, -0.0203, -0.0017, -0.0187, + -0.0304, 0.0068], device='cuda:0'), grad: tensor([-1.4901e-08, 1.8626e-08, 1.1176e-08, 4.1910e-08, 4.6566e-10, + 0.0000e+00, 1.8626e-09, -8.8941e-08, 9.3132e-10, 3.3528e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 220.59, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4416 re_mapping 0.0024 re_causal 0.0092 /// teacc 99.08 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.0022, 0.0110, 0.0691, ..., 0.0905, -0.2936, -0.1869], + [ 0.0578, -0.0185, -0.0566, ..., -0.3439, 0.0525, -0.1142], + [-0.0111, -0.0048, 0.2682, ..., -0.3466, -0.1688, -0.1939], + ..., + [-0.0242, 0.0062, -0.2468, ..., -0.0800, 0.1468, 0.0769], + [-0.0330, -0.0202, -0.2205, ..., -0.3375, 0.1290, -0.2371], + [ 0.0154, -0.0258, -0.1252, ..., 0.0406, -0.2039, 0.0249]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + 0.0000e+00, 2.3283e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 3.4925e-08, 2.3749e-08], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.9802e-08, + 1.8626e-09, 3.1199e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -1.8626e-09, + -4.4703e-08, -4.4238e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + -2.5146e-08, -1.1176e-08], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.3283e-09, + 1.8626e-09, 4.1910e-09]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0206, 0.0133, 0.0221, 0.0056, 0.0363, -0.0203, -0.0017, -0.0187, + -0.0305, 0.0068], device='cuda:0'), grad: tensor([ 8.3819e-09, 1.1083e-07, 8.6613e-08, 2.3749e-08, -7.7300e-08, + 4.6566e-10, 8.0559e-08, -1.7509e-07, -7.4971e-08, 1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 220.31, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4102 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.10 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.0022, 0.0110, 0.0692, ..., 0.0905, -0.2937, -0.1869], + [ 0.0577, -0.0185, -0.0566, ..., -0.3440, 0.0524, -0.1144], + [-0.0111, -0.0048, 0.2682, ..., -0.3468, -0.1691, -0.1941], + ..., + [-0.0242, 0.0062, -0.2468, ..., -0.0801, 0.1468, 0.0771], + [-0.0330, -0.0202, -0.2205, ..., -0.3376, 0.1291, -0.2372], + [ 0.0154, -0.0258, -0.1252, ..., 0.0407, -0.2042, 0.0249]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -1.8626e-09, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + -3.2596e-09, 4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8626e-09, + 4.6566e-10, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -3.2596e-09, -4.1910e-09], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, + 1.3970e-09, 4.1910e-09]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0208, 0.0133, 0.0219, 0.0056, 0.0363, -0.0202, -0.0014, -0.0186, + -0.0306, 0.0068], device='cuda:0'), grad: tensor([ 1.2573e-08, -8.3819e-09, 6.9849e-09, 3.2596e-09, -6.9849e-09, + 4.6566e-10, 0.0000e+00, -1.2573e-08, 7.4506e-09, 1.0710e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 220.23, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4414 re_mapping 0.0023 re_causal 0.0090 /// teacc 99.11 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.0022, 0.0110, 0.0693, ..., 0.0906, -0.2938, -0.1870], + [ 0.0577, -0.0185, -0.0566, ..., -0.3442, 0.0524, -0.1147], + [-0.0111, -0.0048, 0.2683, ..., -0.3468, -0.1693, -0.1942], + ..., + [-0.0242, 0.0062, -0.2470, ..., -0.0803, 0.1469, 0.0772], + [-0.0330, -0.0202, -0.2207, ..., -0.3376, 0.1291, -0.2373], + [ 0.0154, -0.0258, -0.1253, ..., 0.0406, -0.2042, 0.0249]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., 4.6566e-10, + -2.5611e-08, 9.3132e-10], + [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 4.6566e-10, + 2.6077e-08, 4.6566e-10], + ..., + [ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., 1.0245e-08, + 0.0000e+00, 1.2573e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, + 0.0000e+00, 4.6566e-10], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 7.9162e-09, + 0.0000e+00, 9.3132e-09]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0208, 0.0132, 0.0219, 0.0057, 0.0363, -0.0202, -0.0013, -0.0186, + -0.0307, 0.0068], device='cuda:0'), grad: tensor([ 5.5879e-09, -7.6368e-08, 1.0524e-07, -7.9628e-08, -4.2375e-08, + -4.6566e-10, 4.1910e-09, 4.8429e-08, 1.4435e-08, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 219.92, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4042 re_mapping 0.0023 re_causal 0.0088 /// teacc 99.12 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.0022, 0.0110, 0.0694, ..., 0.0907, -0.2941, -0.1870], + [ 0.0577, -0.0185, -0.0566, ..., -0.3443, 0.0523, -0.1148], + [-0.0111, -0.0048, 0.2683, ..., -0.3469, -0.1695, -0.1944], + ..., + [-0.0243, 0.0062, -0.2471, ..., -0.0805, 0.1470, 0.0775], + [-0.0330, -0.0202, -0.2208, ..., -0.3377, 0.1292, -0.2374], + [ 0.0154, -0.0258, -0.1254, ..., 0.0405, -0.2045, 0.0248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., -3.2596e-09, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 3.7253e-09, + 3.2596e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -2.5611e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, + 2.3283e-09, 1.3970e-09]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0208, 0.0131, 0.0218, 0.0057, 0.0364, -0.0202, -0.0013, -0.0185, + -0.0308, 0.0066], device='cuda:0'), grad: tensor([ 7.3435e-07, 1.2573e-08, 2.7008e-08, -1.0245e-08, -4.6566e-10, + 2.0489e-08, -7.0687e-07, 7.9162e-09, -9.1270e-08, 1.2107e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 220.49, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4460 re_mapping 0.0023 re_causal 0.0092 /// teacc 99.11 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.0022, 0.0110, 0.0694, ..., 0.0908, -0.2942, -0.1870], + [ 0.0577, -0.0185, -0.0567, ..., -0.3444, 0.0522, -0.1150], + [-0.0111, -0.0048, 0.2683, ..., -0.3470, -0.1697, -0.1945], + ..., + [-0.0243, 0.0062, -0.2472, ..., -0.0806, 0.1471, 0.0776], + [-0.0330, -0.0202, -0.2209, ..., -0.3377, 0.1292, -0.2374], + [ 0.0154, -0.0258, -0.1255, ..., 0.0405, -0.2047, 0.0248]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -4.1910e-09, + 1.8626e-09, 1.3970e-09], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.9360e-08, 8.1491e-08], + [ 0.0000e+00, 0.0000e+00, -2.6077e-08, ..., 9.3132e-10, + 2.6077e-08, 2.5611e-08], + ..., + [ 0.0000e+00, 0.0000e+00, 1.6298e-08, ..., 4.6566e-10, + -1.6391e-07, -2.1793e-07], + [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, + 2.7940e-09, 5.1223e-09], + [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, + 7.5903e-08, 9.7323e-08]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0209, 0.0131, 0.0217, 0.0058, 0.0365, -0.0203, -0.0012, -0.0184, + -0.0308, 0.0065], device='cuda:0'), grad: tensor([ 9.3132e-10, 2.1467e-07, 8.0094e-08, 5.6811e-08, 1.0710e-08, + -1.3225e-07, 8.8476e-08, -6.9337e-07, 4.3306e-08, 3.3993e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 220.32, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4054 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.12 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.0022, 0.0110, 0.0695, ..., 0.0909, -0.2943, -0.1870], + [ 0.0577, -0.0185, -0.0567, ..., -0.3445, 0.0520, -0.1151], + [-0.0111, -0.0048, 0.2684, ..., -0.3471, -0.1699, -0.1947], + ..., + [-0.0243, 0.0062, -0.2473, ..., -0.0807, 0.1473, 0.0779], + [-0.0330, -0.0202, -0.2210, ..., -0.3378, 0.1292, -0.2375], + [ 0.0154, -0.0258, -0.1256, ..., 0.0405, -0.2049, 0.0247]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.2596e-09, + 4.1910e-09, 1.5832e-08], + [ 0.0000e+00, 0.0000e+00, -4.0978e-08, ..., 0.0000e+00, + 4.1910e-09, 3.7253e-09], + ..., + [ 0.0000e+00, 0.0000e+00, 3.3062e-08, ..., 3.2596e-09, + -1.6764e-08, -1.1642e-08], + [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.9360e-08, + 1.8626e-09, 1.1735e-07]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0208, 0.0129, 0.0217, 0.0060, 0.0365, -0.0204, -0.0012, -0.0181, + -0.0310, 0.0064], device='cuda:0'), grad: tensor([ 2.4214e-08, 3.3062e-08, -5.5879e-08, 3.7253e-08, -2.0768e-07, + 6.0536e-09, -8.8941e-08, 9.7789e-09, 5.5414e-08, 1.9465e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 219.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3970 re_mapping 0.0023 re_causal 0.0086 /// teacc 99.13 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip3', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip3/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.000000 98.989998 ... 82.561035 71.442439 +ShearY 98.750000 98.699997 ... 82.561035 65.832978 +AutoContrast 99.040001 99.029999 ... 82.561035 64.430552 +Invert 98.720001 98.189995 ... 82.561035 65.502023 +Equalize 98.070000 97.930000 ... 82.561035 68.048062 +Solarize 97.979996 97.839996 ... 82.561035 59.778274 +SolarizeAdd 98.269997 97.979996 ... 82.561035 72.102751 +Posterize 98.930000 98.919998 ... 82.561035 69.134236 +Contrast 99.040001 99.110001 ... 82.561035 70.453847 +Color 98.940002 99.180000 ... 82.561035 60.559358 +Brightness 99.010002 99.080002 ... 82.561035 68.488187 +Sharpness 98.979996 99.089996 ... 82.561035 71.397741 +NoiseSalt 99.070000 99.070000 ... 82.561035 58.064358 +NoiseGaussian 98.979996 99.159996 ... 82.561035 57.520034 +w/o do (original x) 99.180000 0.000000 ... 0.000000 75.009668 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.14 68.26598 79.002333 75.829582 84.404584 76.87562 diff --git a/Meta-causal/code-withStyleAttack/66581.error b/Meta-causal/code-withStyleAttack/66581.error new file mode 100644 index 0000000000000000000000000000000000000000..d904956b40cd15b976a8c6306f579672435ab45d --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66581.error @@ -0,0 +1,299 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-ghfkfyr2kyt5 +error libmamba Could not lock non-existing path '/scratch/yuqian_fu/micromamba/pkgs' +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-ghfkfyr2kyt5 + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 h4bc722e_7 conda-forge + + ca-certificates 2024.7.4 hbcca054_0 conda-forge + + certifi 2024.7.4 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + click 8.1.7 unix_pyh707e725_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.1.7 0 nvidia + + libcurand 10.3.6.82 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 14.1.0 h77fa898_0 conda-forge + + libgfortran-ng 14.1.0 h69a702a_0 conda-forge + + libgfortran5 14.1.0 hc5f4f2c_0 /work/conda/cache/conda-forge + + libhwloc 2.11.1 default_hecaa2ac_1000 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 14.1.0 hc0a3c3a_0 /work/conda/cache/conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4bc722e_2 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 /work/conda/cache/conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.3.1 py3.11_cuda12.1_cudnn8.9.2_0 /work/conda/cache/pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 71.0.3 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.13.0 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h434a139_3 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 2.3.1 py311 /work/conda/cache/pytorch + + torchvision 0.18.1 py311_cu121 /work/conda/cache/pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.23.0 py311h5cd10c7_0 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 119 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.7.4-hbcca054_0 +Linking libgcc-ng-14.1.0-h77fa898_0 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4bc722e_2 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-14.1.0-hc5f4f2c_0 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-14.1.0-hc0a3c3a_0 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-h4bc722e_7 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-14.1.0-h69a702a_0 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.11.1-default_hecaa2ac_1000 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h434a139_3 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.82-0 +Linking libcufile-1.10.1.7-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-71.0.3-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.7.4-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking click-8.1.7-unix_pyh707e725_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.23.0-py311h5cd10c7_0 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.13.0-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.3.1-py3.11_cuda12.1_cudnn8.9.2_0 +Linking torchtriton-2.3.1-py311 +Linking torchvision-0.18.1-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-ghfkfyr2kyt5 + +Or to execute a single command in this environment, use: + + mamba run -n auto-ghfkfyr2kyt5 mycommand + +Installing pip packages +WARNING: The candidate selected for download or install is a yanked version: 'opencv-python' candidate (version 4.5.5.62 at https://files.pythonhosted.org/packages/9d/98/36bfcbff30da27dd6922ed73ca7802c37d87f77daf4c569da3dcb87b4296/opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from https://pypi.org/simple/opencv-python/) (requires-python:>=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 diff --git a/Meta-causal/code-withStyleAttack/66581.log b/Meta-causal/code-withStyleAttack/66581.log new file mode 100644 index 0000000000000000000000000000000000000000..bbfd612f06ddd2d645eca7f9b182eb26b7e824d1 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/66581.log @@ -0,0 +1,14226 @@ +Collecting h5py>=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 9.0 MB/s eta 0:00:00 + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.0) +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 22.3 MB/s eta 0:00:00 +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 16.3 MB/s eta 0:00:00 +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (71.0.3) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.3.1) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.18.1) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.24.0-py3-none-any.whl.metadata (13 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 25.5 MB/s eta 0:00:00 +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.13.0) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.7.4) +Requirement already satisfied: mpmath<1.4,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 35.8 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 207.4 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 112.6 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 167.4 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 38.9 MB/s eta 0:00:00 +Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 193.1 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 56.4 MB/s eta 0:00:00 +Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.7 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.7/5.7 MB 234.0 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 42.0 MB/s eta 0:00:00 +Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl (294 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 106.3 MB/s eta 0:00:00 +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 198.6 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 77.7 MB/s eta 0:00:00 +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.24.0-py3-none-any.whl (419 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 419.0/419.0 kB 115.7 MB/s eta 0:00:00 +Downloading packaging-24.1-py3-none-any.whl (53 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 22.4 MB/s eta 0:00:00 +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 179.4 MB/s eta 0:00:00 +Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 177.6/177.6 kB 67.1 MB/s eta 0:00:00 +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 30.4 MB/s eta 0:00:00 +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94507 sha256=7c8cb802a3eb82a53996f1853748864b8cfe2e63acae5deea5995aee9eda1298 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.65.1 h5py-3.11.0 huggingface_hub-0.24.0 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.3 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip4', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0033, 0.0106, -0.0232, ..., -0.0168, -0.0084, -0.0229], + [ 0.0190, -0.0041, -0.0121, ..., 0.0070, 0.0035, -0.0125], + [ 0.0037, -0.0127, -0.0222, ..., -0.0070, 0.0055, 0.0012], + ..., + [ 0.0113, -0.0083, 0.0180, ..., 0.0266, -0.0219, 0.0124], + [-0.0019, 0.0299, -0.0147, ..., 0.0174, 0.0282, 0.0215], + [ 0.0080, -0.0058, 0.0244, ..., 0.0119, 0.0297, -0.0058]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0090, 0.0254, -0.0144, 0.0130, 0.0250, -0.0216, -0.0249, -0.0214, + -0.0198, -0.0243], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 220.89, cls_loss 1.2429 cls_loss_mapping 1.8217 cls_loss_causal 2.2111 re_mapping 0.1623 re_causal 0.1722 /// teacc 84.02 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0096, 0.0050, -0.0253, ..., -0.0166, -0.0116, -0.0229], + [ 0.0196, 0.0030, -0.0149, ..., 0.0039, 0.0046, -0.0125], + [ 0.0075, -0.0160, -0.0209, ..., -0.0105, 0.0097, 0.0012], + ..., + [ 0.0046, -0.0113, 0.0141, ..., 0.0290, -0.0270, 0.0124], + [ 0.0015, 0.0259, -0.0173, ..., 0.0147, 0.0306, 0.0215], + [ 0.0021, -0.0078, 0.0205, ..., 0.0101, 0.0247, -0.0058]], + device='cuda:0'), grad: tensor([[ 0.0097, 0.0004, 0.0005, ..., 0.0000, 0.0117, 0.0000], + [ 0.0275, 0.0021, 0.0007, ..., 0.0000, 0.0295, 0.0000], + [ 0.0309, 0.0035, -0.0082, ..., 0.0000, 0.0452, 0.0000], + ..., + [ 0.0035, 0.0014, 0.0021, ..., 0.0000, 0.0063, 0.0000], + [-0.0222, 0.0022, 0.0024, ..., 0.0000, -0.0240, 0.0000], + [ 0.0080, 0.0008, 0.0030, ..., 0.0000, 0.0024, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0099, 0.0248, -0.0146, 0.0121, 0.0247, -0.0200, -0.0255, -0.0207, + -0.0201, -0.0235], device='cuda:0'), grad: tensor([ 2.9556e-02, 3.3661e-02, 3.9673e-02, -3.7689e-02, 3.3142e-02, + -5.8136e-02, -1.0262e-02, 7.7667e-03, 3.0443e-05, -3.7750e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 221.09, cls_loss 0.4076 cls_loss_mapping 0.7362 cls_loss_causal 1.8887 re_mapping 0.2030 re_causal 0.2662 /// teacc 92.18 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0126, 0.0024, -0.0284, ..., -0.0166, -0.0143, -0.0265], + [ 0.0188, 0.0075, -0.0212, ..., 0.0037, 0.0051, -0.0176], + [ 0.0086, -0.0182, -0.0199, ..., -0.0108, 0.0109, -0.0042], + ..., + [ 0.0026, -0.0137, 0.0106, ..., 0.0294, -0.0286, 0.0079], + [ 0.0038, 0.0225, -0.0261, ..., 0.0145, 0.0332, 0.0164], + [-0.0020, -0.0118, 0.0185, ..., 0.0099, 0.0206, -0.0101]], + device='cuda:0'), grad: tensor([[ 4.0207e-03, 1.2779e-03, 3.2921e-03, ..., 7.3910e-05, + 3.6983e-03, 0.0000e+00], + [ 9.6741e-03, -4.7970e-04, 1.1101e-03, ..., 1.6496e-05, + 1.0933e-02, 0.0000e+00], + [-2.0508e-02, -1.8473e-03, 8.7833e-04, ..., 8.7619e-06, + -3.9520e-02, 0.0000e+00], + ..., + [ 2.9449e-03, 2.4700e-03, -5.3009e-02, ..., -6.3705e-03, + 4.3030e-03, 0.0000e+00], + [ 4.3831e-03, -3.2444e-03, 8.0414e-03, ..., 7.7844e-05, + 1.2493e-03, 0.0000e+00], + [ 3.5343e-03, -2.0390e-03, -1.7365e-02, ..., 3.7313e-04, + 1.1215e-03, 0.0000e+00]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0104, 0.0243, -0.0147, 0.0123, 0.0246, -0.0190, -0.0260, -0.0213, + -0.0200, -0.0229], device='cuda:0'), grad: tensor([ 0.0169, 0.0134, -0.0365, 0.0040, 0.0584, -0.0155, 0.0173, -0.0557, + 0.0110, -0.0133], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 219.85, cls_loss 0.2705 cls_loss_mapping 0.4382 cls_loss_causal 1.6842 re_mapping 0.1473 re_causal 0.2382 /// teacc 93.87 lr 0.00010000 +Epoch 4, weight, value: tensor([[-1.3754e-02, -9.4881e-04, -3.1061e-02, ..., -1.5804e-02, + -1.5001e-02, -2.9382e-02], + [ 1.8693e-02, 1.1509e-02, -2.3888e-02, ..., 1.0280e-07, + 5.4352e-03, -2.1072e-02], + [ 8.6916e-03, -2.1121e-02, -2.2663e-02, ..., -1.3761e-02, + 1.1824e-02, -7.5192e-03], + ..., + [ 9.5631e-04, -1.7752e-02, 1.1693e-02, ..., 3.1799e-02, + -3.0505e-02, 4.7298e-03], + [ 5.0814e-03, 2.2056e-02, -3.1942e-02, ..., 1.2396e-02, + 3.5316e-02, 1.3053e-02], + [-4.6266e-03, -1.2343e-02, 2.0372e-02, ..., 6.6468e-03, + 1.8179e-02, -1.4114e-02]], device='cuda:0'), grad: tensor([[-9.8801e-03, 6.2752e-04, 5.4216e-04, ..., 6.4122e-07, + -6.6757e-03, 0.0000e+00], + [-4.5395e-03, -4.0436e-02, -3.1006e-02, ..., 3.4142e-06, + -2.3956e-02, 0.0000e+00], + [ 2.3329e-04, 4.0855e-03, 9.7084e-04, ..., 2.0228e-06, + 2.6283e-03, 0.0000e+00], + ..., + [ 8.4448e-04, 6.9313e-03, 1.8478e-02, ..., 6.1803e-06, + 6.3095e-03, 0.0000e+00], + [-1.6594e-03, 1.8330e-03, 4.0474e-03, ..., 6.8396e-06, + 5.2032e-03, 0.0000e+00], + [ 2.0695e-03, 3.1548e-03, -8.8577e-03, ..., 9.8825e-05, + -1.0086e-02, 0.0000e+00]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0101, 0.0245, -0.0149, 0.0121, 0.0244, -0.0189, -0.0266, -0.0215, + -0.0196, -0.0226], device='cuda:0'), grad: tensor([-0.0142, -0.0628, 0.0059, 0.0082, 0.0185, 0.0187, 0.0118, 0.0300, + 0.0145, -0.0307], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 220.16, cls_loss 0.2136 cls_loss_mapping 0.3193 cls_loss_causal 1.5109 re_mapping 0.1189 re_causal 0.2136 /// teacc 95.49 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0149, -0.0046, -0.0337, ..., -0.0155, -0.0151, -0.0285], + [ 0.0180, 0.0138, -0.0253, ..., -0.0033, 0.0053, -0.0228], + [ 0.0092, -0.0222, -0.0219, ..., -0.0115, 0.0130, -0.0085], + ..., + [-0.0002, -0.0201, 0.0130, ..., 0.0344, -0.0319, 0.0018], + [ 0.0062, 0.0217, -0.0372, ..., 0.0043, 0.0368, 0.0099], + [-0.0073, -0.0149, 0.0215, ..., 0.0050, 0.0156, -0.0190]], + device='cuda:0'), grad: tensor([[ 1.8597e-04, -1.0557e-03, 2.1720e-04, ..., 5.5097e-06, + 2.7990e-04, 0.0000e+00], + [ 1.2426e-03, -4.5815e-03, 1.4687e-04, ..., 7.3239e-06, + 3.1614e-04, 0.0000e+00], + [ 2.6836e-03, 8.2397e-04, 8.6069e-04, ..., 1.7464e-05, + 2.0447e-03, 0.0000e+00], + ..., + [ 1.0958e-03, 8.1444e-04, 3.9902e-03, ..., 7.4565e-05, + 1.4744e-03, 0.0000e+00], + [-1.5610e-02, 1.8024e-04, -3.2711e-04, ..., 9.2685e-06, + -1.4130e-02, 0.0000e+00], + [ 1.0277e-02, 1.0700e-03, -8.3466e-03, ..., -1.3196e-04, + 1.0246e-02, 0.0000e+00]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0100, 0.0242, -0.0149, 0.0118, 0.0245, -0.0191, -0.0268, -0.0213, + -0.0192, -0.0227], device='cuda:0'), grad: tensor([-0.0090, -0.0016, 0.0060, -0.0071, 0.0093, 0.0050, 0.0015, 0.0084, + -0.0150, 0.0025], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 219.67, cls_loss 0.1685 cls_loss_mapping 0.2558 cls_loss_causal 1.3716 re_mapping 0.0954 re_causal 0.1943 /// teacc 96.71 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0166, -0.0068, -0.0360, ..., -0.0164, -0.0151, -0.0277], + [ 0.0177, 0.0162, -0.0261, ..., -0.0041, 0.0057, -0.0283], + [ 0.0095, -0.0236, -0.0224, ..., -0.0121, 0.0139, -0.0112], + ..., + [-0.0012, -0.0219, 0.0139, ..., 0.0358, -0.0330, -0.0033], + [ 0.0076, 0.0222, -0.0411, ..., 0.0021, 0.0383, 0.0018], + [-0.0094, -0.0163, 0.0227, ..., 0.0050, 0.0139, -0.0240]], + device='cuda:0'), grad: tensor([[ 3.2663e-04, 3.5810e-04, 7.1430e-04, ..., 2.0728e-05, + 5.3787e-04, 6.6400e-05], + [ 1.0884e-04, -9.3079e-03, -8.2321e-03, ..., -2.1040e-04, + 1.5812e-03, 8.9034e-06], + [-3.1948e-04, 6.9141e-04, 1.0786e-03, ..., 4.2953e-06, + -2.1011e-02, 2.5570e-05], + ..., + [ 2.3472e-04, 1.5144e-03, -3.1166e-03, ..., 1.9491e-05, + 1.8206e-03, 5.5701e-05], + [-3.9506e-04, 2.4147e-03, 3.8033e-03, ..., 5.9605e-05, + 1.6022e-03, 6.6102e-05], + [ 2.7347e-04, 3.5954e-03, 1.0090e-03, ..., 6.3181e-05, + 4.4775e-04, -2.6035e-04]], device='cuda:0') +Epoch 6, bias, value: tensor([-0.0098, 0.0245, -0.0150, 0.0117, 0.0243, -0.0194, -0.0272, -0.0215, + -0.0188, -0.0224], device='cuda:0'), grad: tensor([-0.0009, -0.0094, -0.0253, -0.0149, 0.0223, 0.0152, 0.0019, -0.0022, + 0.0091, 0.0042], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 219.77, cls_loss 0.1248 cls_loss_mapping 0.1876 cls_loss_causal 1.2988 re_mapping 0.0819 re_causal 0.1818 /// teacc 96.72 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0184, -0.0087, -0.0373, ..., -0.0176, -0.0162, -0.0281], + [ 0.0173, 0.0181, -0.0272, ..., -0.0049, 0.0058, -0.0315], + [ 0.0096, -0.0258, -0.0229, ..., -0.0130, 0.0148, -0.0116], + ..., + [-0.0022, -0.0236, 0.0149, ..., 0.0366, -0.0342, -0.0003], + [ 0.0088, 0.0224, -0.0449, ..., 0.0005, 0.0398, -0.0016], + [-0.0104, -0.0166, 0.0230, ..., 0.0044, 0.0125, -0.0256]], + device='cuda:0'), grad: tensor([[ 1.7595e-03, 4.4060e-04, 6.8569e-04, ..., 1.7691e-04, + 1.3132e-03, 0.0000e+00], + [ 7.9966e-04, -8.1863e-03, 7.5150e-04, ..., 1.5646e-05, + -3.9024e-03, 0.0000e+00], + [ 5.0116e-04, 4.2686e-03, 7.4387e-04, ..., 3.7640e-05, + 7.9575e-03, 0.0000e+00], + ..., + [ 1.6899e-03, 1.5955e-03, -2.7466e-03, ..., -6.2561e-04, + 4.1542e-03, 0.0000e+00], + [-7.0572e-03, -1.0443e-03, 1.8110e-03, ..., 2.5392e-05, + -6.1913e-03, 0.0000e+00], + [ 1.9989e-03, -2.1210e-03, -7.8087e-03, ..., 1.7023e-04, + -1.4162e-03, 0.0000e+00]], device='cuda:0') +Epoch 7, bias, value: tensor([-0.0095, 0.0246, -0.0151, 0.0119, 0.0245, -0.0199, -0.0272, -0.0215, + -0.0187, -0.0226], device='cuda:0'), grad: tensor([ 0.0037, -0.0027, 0.0110, 0.0199, -0.0086, -0.0009, -0.0025, 0.0048, + -0.0043, -0.0204], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 219.42, cls_loss 0.1314 cls_loss_mapping 0.1853 cls_loss_causal 1.2593 re_mapping 0.0696 re_causal 0.1589 /// teacc 97.23 lr 0.00010000 +Epoch 8, weight, value: tensor([[-1.9604e-02, -9.5689e-03, -3.7476e-02, ..., -1.8186e-02, + -1.6059e-02, -2.7094e-02], + [ 1.6744e-02, 1.9728e-02, -2.7795e-02, ..., -5.5014e-03, + 6.0967e-03, -3.2839e-02], + [ 1.0002e-02, -2.7381e-02, -2.2454e-02, ..., -1.3403e-02, + 1.5827e-02, -1.2867e-02], + ..., + [-3.0162e-03, -2.4251e-02, 1.5811e-02, ..., 3.7565e-02, + -3.5452e-02, 2.7019e-05], + [ 9.8292e-03, 2.2710e-02, -4.6937e-02, ..., -6.9613e-04, + 4.1093e-02, -3.7622e-03], + [-1.2419e-02, -1.7763e-02, 2.3166e-02, ..., 3.9022e-03, + 1.0741e-02, -2.6499e-02]], device='cuda:0'), grad: tensor([[ 9.6226e-04, 5.6076e-04, 2.3603e-04, ..., 9.2089e-06, + -2.1732e-04, 0.0000e+00], + [ 2.7790e-03, -4.2868e-04, 9.5940e-04, ..., 7.1451e-06, + 3.0613e-03, 0.0000e+00], + [-4.7684e-03, -3.0479e-03, 5.8603e-04, ..., 8.4862e-06, + -7.4654e-03, 0.0000e+00], + ..., + [ 3.8433e-04, 1.4305e-03, -4.0169e-03, ..., 5.2266e-06, + 7.4005e-04, 0.0000e+00], + [ 1.3151e-03, 1.1139e-03, 1.1168e-03, ..., 3.3557e-05, + 1.4229e-03, 0.0000e+00], + [ 1.8919e-04, 4.5419e-04, -3.1616e-02, ..., 3.8028e-05, + 2.7585e-04, 0.0000e+00]], device='cuda:0') +Epoch 8, bias, value: tensor([-0.0092, 0.0246, -0.0150, 0.0118, 0.0245, -0.0200, -0.0277, -0.0210, + -0.0184, -0.0229], device='cuda:0'), grad: tensor([-0.0005, 0.0038, -0.0029, 0.0058, 0.0306, 0.0043, -0.0043, -0.0193, + 0.0052, -0.0227], device='cuda:0') +100 +0.0001 +changing lr +epoch 7, time 219.29, cls_loss 0.0951 cls_loss_mapping 0.1423 cls_loss_causal 1.1747 re_mapping 0.0605 re_causal 0.1489 /// teacc 97.17 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0210, -0.0107, -0.0386, ..., -0.0180, -0.0173, -0.0273], + [ 0.0160, 0.0207, -0.0294, ..., -0.0055, 0.0061, -0.0341], + [ 0.0101, -0.0289, -0.0227, ..., -0.0140, 0.0167, -0.0131], + ..., + [-0.0043, -0.0260, 0.0176, ..., 0.0385, -0.0368, 0.0007], + [ 0.0109, 0.0227, -0.0495, ..., -0.0017, 0.0423, -0.0048], + [-0.0137, -0.0184, 0.0239, ..., 0.0038, 0.0093, -0.0273]], + device='cuda:0'), grad: tensor([[ 3.6144e-04, 3.3307e-04, 2.0611e-04, ..., 4.9204e-05, + 5.5933e-04, 0.0000e+00], + [-1.6083e-02, -2.3712e-02, 1.0433e-03, ..., 5.9098e-05, + -6.6071e-03, 0.0000e+00], + [-4.2229e-03, -1.6212e-03, 1.8921e-03, ..., 3.0547e-05, + -1.4429e-03, 0.0000e+00], + ..., + [-6.2513e-04, 4.9877e-04, 5.4893e-03, ..., 1.5154e-03, + -2.1648e-03, 0.0000e+00], + [ 2.4338e-03, 6.0158e-03, 1.1988e-03, ..., 2.1410e-04, + 5.1575e-03, 0.0000e+00], + [ 5.6696e-04, -3.1996e-04, -1.1238e-02, ..., -1.6785e-03, + 1.6375e-03, 0.0000e+00]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0092, 0.0243, -0.0152, 0.0117, 0.0246, -0.0203, -0.0278, -0.0207, + -0.0181, -0.0228], device='cuda:0'), grad: tensor([ 0.0003, -0.0122, 0.0075, -0.0140, 0.0060, 0.0040, 0.0129, -0.0064, + 0.0116, -0.0097], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 219.82, cls_loss 0.0875 cls_loss_mapping 0.1296 cls_loss_causal 1.1137 re_mapping 0.0570 re_causal 0.1382 /// teacc 97.46 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0225, -0.0116, -0.0383, ..., -0.0176, -0.0176, -0.0269], + [ 0.0153, 0.0222, -0.0306, ..., -0.0063, 0.0059, -0.0346], + [ 0.0105, -0.0302, -0.0230, ..., -0.0146, 0.0179, -0.0136], + ..., + [-0.0056, -0.0270, 0.0181, ..., 0.0407, -0.0389, 0.0014], + [ 0.0117, 0.0219, -0.0522, ..., -0.0040, 0.0431, -0.0054], + [-0.0151, -0.0188, 0.0242, ..., 0.0031, 0.0088, -0.0280]], + device='cuda:0'), grad: tensor([[ 7.2575e-04, 2.0719e-04, 1.7667e-04, ..., 7.0453e-05, + 6.4468e-04, 3.3099e-06], + [ 3.6621e-04, -4.0283e-03, -4.1771e-04, ..., 1.8060e-04, + -1.8919e-04, 1.3486e-06], + [ 8.2827e-04, 7.6151e-04, 5.2357e-04, ..., 5.6934e-04, + -2.3613e-03, 1.0729e-05], + ..., + [ 4.1628e-04, 2.5797e-04, -1.8463e-03, ..., 1.7524e-04, + 1.2388e-03, 3.3379e-06], + [ 3.5501e-04, 8.0729e-04, 7.0238e-04, ..., 6.1989e-05, + -4.6301e-04, 3.6228e-06], + [ 8.4829e-04, 4.2415e-04, 1.8568e-03, ..., 3.8624e-04, + 1.3418e-03, 1.0985e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0089, 0.0240, -0.0150, 0.0120, 0.0246, -0.0204, -0.0281, -0.0208, + -0.0182, -0.0227], device='cuda:0'), grad: tensor([ 0.0009, -0.0010, -0.0002, -0.0061, 0.0033, -0.0021, -0.0007, -0.0012, + 0.0023, 0.0046], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 219.70, cls_loss 0.0855 cls_loss_mapping 0.1226 cls_loss_causal 1.0743 re_mapping 0.0512 re_causal 0.1281 /// teacc 97.66 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0237, -0.0129, -0.0377, ..., -0.0189, -0.0181, -0.0255], + [ 0.0150, 0.0237, -0.0316, ..., -0.0075, 0.0065, -0.0349], + [ 0.0106, -0.0317, -0.0242, ..., -0.0160, 0.0182, -0.0145], + ..., + [-0.0068, -0.0274, 0.0196, ..., 0.0418, -0.0402, 0.0017], + [ 0.0127, 0.0222, -0.0547, ..., -0.0053, 0.0441, -0.0059], + [-0.0163, -0.0205, 0.0241, ..., 0.0022, 0.0078, -0.0287]], + device='cuda:0'), grad: tensor([[ 5.5790e-04, 3.1829e-04, 8.4281e-05, ..., 0.0000e+00, + 2.3019e-04, 0.0000e+00], + [ 3.4294e-03, 2.3727e-03, -7.2837e-05, ..., 0.0000e+00, + 1.3266e-03, 0.0000e+00], + [-3.6001e-04, 9.0837e-05, 2.5868e-04, ..., 0.0000e+00, + -1.6661e-03, 0.0000e+00], + ..., + [ 2.9325e-04, 6.8378e-04, 1.1314e-02, ..., 0.0000e+00, + 2.2984e-04, 0.0000e+00], + [ 3.3173e-02, 1.3893e-02, 6.2609e-04, ..., 0.0000e+00, + 9.8953e-03, 0.0000e+00], + [ 7.6437e-04, 2.5630e-04, -1.2169e-02, ..., 0.0000e+00, + 3.5310e-04, 0.0000e+00]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0089, 0.0243, -0.0156, 0.0121, 0.0247, -0.0203, -0.0283, -0.0206, + -0.0178, -0.0232], device='cuda:0'), grad: tensor([ 0.0275, 0.0027, -0.0003, -0.0036, 0.0006, -0.0581, 0.0044, 0.0145, + 0.0249, -0.0125], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 219.96, cls_loss 0.0820 cls_loss_mapping 0.1164 cls_loss_causal 1.0993 re_mapping 0.0455 re_causal 0.1192 /// teacc 97.89 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0250, -0.0141, -0.0387, ..., -0.0199, -0.0187, -0.0253], + [ 0.0141, 0.0243, -0.0316, ..., -0.0056, 0.0066, -0.0358], + [ 0.0110, -0.0322, -0.0237, ..., -0.0176, 0.0193, -0.0149], + ..., + [-0.0078, -0.0284, 0.0199, ..., 0.0425, -0.0416, 0.0021], + [ 0.0133, 0.0216, -0.0582, ..., -0.0067, 0.0449, -0.0066], + [-0.0172, -0.0206, 0.0251, ..., 0.0018, 0.0068, -0.0289]], + device='cuda:0'), grad: tensor([[ 2.7585e-04, 3.0494e-04, 3.2067e-05, ..., 7.5391e-07, + 1.7357e-04, 0.0000e+00], + [ 3.4351e-03, -1.9236e-03, 6.6280e-05, ..., 8.4098e-07, + -3.1395e-03, 0.0000e+00], + [ 7.1239e-04, 1.8644e-03, 7.5006e-04, ..., 2.5262e-07, + 1.8902e-03, 0.0000e+00], + ..., + [ 7.7665e-05, 9.1600e-04, -1.3056e-03, ..., 3.9577e-05, + 7.7057e-04, 0.0000e+00], + [-2.9011e-03, 4.8661e-04, 2.2161e-04, ..., 9.3132e-07, + -5.1689e-03, 0.0000e+00], + [ 2.4533e-04, 1.8632e-04, -6.7253e-03, ..., -6.7055e-05, + 2.4247e-04, 0.0000e+00]], device='cuda:0') +Epoch 12, bias, value: tensor([-0.0089, 0.0246, -0.0155, 0.0119, 0.0248, -0.0200, -0.0288, -0.0208, + -0.0181, -0.0230], device='cuda:0'), grad: tensor([ 0.0002, -0.0047, 0.0039, 0.0047, 0.0070, 0.0049, -0.0065, -0.0003, + -0.0031, -0.0063], device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 219.29, cls_loss 0.0644 cls_loss_mapping 0.0966 cls_loss_causal 1.0539 re_mapping 0.0452 re_causal 0.1130 /// teacc 97.88 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0264, -0.0136, -0.0377, ..., -0.0203, -0.0187, -0.0238], + [ 0.0132, 0.0251, -0.0318, ..., -0.0034, 0.0059, -0.0371], + [ 0.0113, -0.0329, -0.0251, ..., -0.0189, 0.0202, -0.0157], + ..., + [-0.0081, -0.0301, 0.0218, ..., 0.0432, -0.0423, 0.0025], + [ 0.0142, 0.0216, -0.0606, ..., -0.0070, 0.0459, -0.0075], + [-0.0181, -0.0220, 0.0252, ..., 0.0020, 0.0060, -0.0293]], + device='cuda:0'), grad: tensor([[ 6.1846e-04, 6.5947e-04, 1.4710e-04, ..., 1.3009e-05, + -9.0003e-05, 0.0000e+00], + [ 1.4982e-03, 4.1809e-03, 1.2684e-03, ..., 1.7810e-04, + 5.0659e-03, 0.0000e+00], + [ 4.2458e-03, -1.8911e-03, 1.1711e-03, ..., 1.1921e-05, + 4.4107e-04, 0.0000e+00], + ..., + [ 5.6410e-04, 4.2582e-04, -3.9673e-03, ..., -6.0272e-04, + 1.4172e-03, 0.0000e+00], + [ 1.4938e-02, 1.9569e-03, -2.2354e-03, ..., 2.9150e-06, + -7.7391e-04, 0.0000e+00], + [ 2.3258e-04, 6.0081e-04, 1.2493e-03, ..., 3.3474e-04, + 5.6410e-04, 0.0000e+00]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0086, 0.0242, -0.0152, 0.0119, 0.0248, -0.0203, -0.0288, -0.0204, + -0.0180, -0.0231], device='cuda:0'), grad: tensor([-0.0004, 0.0101, 0.0015, -0.0003, 0.0078, -0.0229, 0.0013, -0.0032, + 0.0041, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 217.62, cls_loss 0.0632 cls_loss_mapping 0.0958 cls_loss_causal 1.0086 re_mapping 0.0419 re_causal 0.1067 /// teacc 98.13 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0278, -0.0149, -0.0383, ..., -0.0213, -0.0196, -0.0236], + [ 0.0126, 0.0257, -0.0328, ..., -0.0037, 0.0058, -0.0377], + [ 0.0115, -0.0339, -0.0267, ..., -0.0202, 0.0210, -0.0162], + ..., + [-0.0088, -0.0313, 0.0227, ..., 0.0436, -0.0435, 0.0039], + [ 0.0152, 0.0218, -0.0619, ..., -0.0078, 0.0469, -0.0080], + [-0.0193, -0.0218, 0.0257, ..., 0.0017, 0.0049, -0.0296]], + device='cuda:0'), grad: tensor([[ 1.0097e-04, 5.6535e-05, 6.2287e-05, ..., 9.1968e-09, + 9.1672e-05, -1.0073e-04], + [ 6.6698e-05, 2.1820e-03, 1.6546e-04, ..., 1.3912e-07, + 6.8283e-03, 2.3982e-07], + [-1.8179e-04, -3.5248e-03, -4.7326e-04, ..., 7.7998e-09, + -8.4381e-03, 1.6261e-06], + ..., + [ 1.8620e-04, 7.3671e-04, -2.2011e-03, ..., 3.5693e-07, + 6.3086e-04, 2.4587e-06], + [-3.0804e-03, 4.6968e-04, -7.4816e-04, ..., 1.1094e-07, + -4.8027e-03, 3.4552e-06], + [ 2.8343e-03, -4.9353e-05, 2.2678e-03, ..., -1.3355e-06, + 4.0894e-03, 1.2346e-05]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0088, 0.0238, -0.0157, 0.0124, 0.0248, -0.0203, -0.0291, -0.0204, + -0.0175, -0.0231], device='cuda:0'), grad: tensor([-0.0036, 0.0055, -0.0066, 0.0051, 0.0006, -0.0037, 0.0011, -0.0022, + -0.0033, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 217.08, cls_loss 0.0582 cls_loss_mapping 0.0908 cls_loss_causal 0.9903 re_mapping 0.0403 re_causal 0.1048 /// teacc 98.12 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0294, -0.0160, -0.0381, ..., -0.0215, -0.0198, -0.0236], + [ 0.0122, 0.0271, -0.0339, ..., -0.0036, 0.0058, -0.0383], + [ 0.0119, -0.0354, -0.0263, ..., -0.0210, 0.0220, -0.0160], + ..., + [-0.0098, -0.0320, 0.0237, ..., 0.0440, -0.0450, 0.0043], + [ 0.0157, 0.0216, -0.0642, ..., -0.0083, 0.0477, -0.0084], + [-0.0198, -0.0228, 0.0256, ..., 0.0017, 0.0038, -0.0298]], + device='cuda:0'), grad: tensor([[ 1.3733e-04, 3.3230e-05, 3.4451e-05, ..., 2.6100e-07, + -7.2193e-04, 0.0000e+00], + [ 1.7297e-04, -7.9498e-03, -4.0588e-03, ..., 2.2221e-06, + -1.0786e-03, 0.0000e+00], + [-6.2370e-04, 3.1877e-04, 1.2636e-04, ..., 8.4043e-06, + -1.4725e-03, 0.0000e+00], + ..., + [ 1.1533e-04, 6.7711e-03, 4.5433e-03, ..., 3.2008e-05, + 1.5421e-03, 0.0000e+00], + [-2.8372e-04, -1.5303e-05, 2.6822e-04, ..., 1.8803e-06, + 2.0218e-04, 0.0000e+00], + [ 4.1008e-05, 1.4448e-04, -2.1152e-03, ..., 1.4780e-06, + 2.2840e-04, 0.0000e+00]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0086, 0.0239, -0.0155, 0.0123, 0.0249, -0.0202, -0.0291, -0.0204, + -0.0176, -0.0234], device='cuda:0'), grad: tensor([-0.0033, -0.0068, -0.0002, 0.0012, 0.0012, 0.0007, 0.0003, 0.0078, + 0.0007, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 216.91, cls_loss 0.0555 cls_loss_mapping 0.0857 cls_loss_causal 0.9918 re_mapping 0.0379 re_causal 0.1010 /// teacc 98.06 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0305, -0.0157, -0.0388, ..., -0.0221, -0.0198, -0.0237], + [ 0.0116, 0.0278, -0.0342, ..., -0.0035, 0.0054, -0.0383], + [ 0.0121, -0.0356, -0.0271, ..., -0.0220, 0.0227, -0.0160], + ..., + [-0.0105, -0.0334, 0.0237, ..., 0.0437, -0.0460, 0.0043], + [ 0.0167, 0.0217, -0.0660, ..., -0.0096, 0.0489, -0.0085], + [-0.0205, -0.0230, 0.0263, ..., 0.0015, 0.0023, -0.0298]], + device='cuda:0'), grad: tensor([[ 4.5228e-04, 1.3947e-04, 7.1704e-05, ..., 0.0000e+00, + 1.1593e-04, 0.0000e+00], + [ 3.2592e-04, -3.7909e-04, 5.0974e-04, ..., 0.0000e+00, + 6.6471e-04, 0.0000e+00], + [ 6.4087e-04, 1.9252e-04, -2.1423e-02, ..., 0.0000e+00, + 8.3256e-04, 0.0000e+00], + ..., + [ 2.1029e-04, 7.1704e-05, 2.0264e-02, ..., 0.0000e+00, + 2.1052e-04, 0.0000e+00], + [-7.8201e-04, 2.8515e-04, -8.0013e-04, ..., 0.0000e+00, + -8.6746e-03, 0.0000e+00], + [ 3.8605e-03, 1.6508e-03, 3.3379e-03, ..., 0.0000e+00, + 4.7302e-03, 0.0000e+00]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0085, 0.0238, -0.0155, 0.0126, 0.0247, -0.0202, -0.0291, -0.0210, + -0.0173, -0.0232], device='cuda:0'), grad: tensor([ 0.0003, 0.0012, -0.0272, 0.0057, -0.0015, -0.0169, 0.0044, 0.0271, + -0.0011, 0.0079], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 217.58, cls_loss 0.0501 cls_loss_mapping 0.0776 cls_loss_causal 0.9917 re_mapping 0.0353 re_causal 0.0977 /// teacc 98.27 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0319, -0.0173, -0.0387, ..., -0.0222, -0.0204, -0.0233], + [ 0.0109, 0.0287, -0.0336, ..., -0.0036, 0.0052, -0.0395], + [ 0.0123, -0.0364, -0.0281, ..., -0.0228, 0.0231, -0.0172], + ..., + [-0.0109, -0.0345, 0.0248, ..., 0.0439, -0.0465, 0.0051], + [ 0.0176, 0.0216, -0.0681, ..., -0.0100, 0.0499, -0.0089], + [-0.0213, -0.0235, 0.0264, ..., 0.0015, 0.0017, -0.0304]], + device='cuda:0'), grad: tensor([[ 2.5225e-04, 1.6558e-04, 7.1764e-05, ..., 0.0000e+00, + 7.2777e-05, -3.2115e-04], + [ 4.4131e-04, -2.6822e-04, 1.6856e-04, ..., 0.0000e+00, + 3.3259e-04, 2.4363e-06], + [-5.7602e-04, 3.9983e-04, 2.6011e-04, ..., 0.0000e+00, + -1.9045e-03, 4.0889e-05], + ..., + [ 1.2207e-04, 1.4532e-04, 3.1471e-04, ..., 0.0000e+00, + 1.6546e-04, 1.2182e-05], + [ 2.8610e-03, 1.3351e-03, 1.8616e-03, ..., 0.0000e+00, + 1.3037e-03, 1.5050e-05], + [ 4.9496e-04, 2.1291e-04, -7.1478e-04, ..., 0.0000e+00, + 6.2406e-05, 4.7624e-05]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0086, 0.0236, -0.0155, 0.0125, 0.0245, -0.0204, -0.0293, -0.0205, + -0.0171, -0.0231], device='cuda:0'), grad: tensor([-0.0026, 0.0006, -0.0006, 0.0013, -0.0048, -0.0020, 0.0012, 0.0010, + 0.0062, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 217.60, cls_loss 0.0470 cls_loss_mapping 0.0740 cls_loss_causal 0.9510 re_mapping 0.0342 re_causal 0.0925 /// teacc 98.47 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0327, -0.0181, -0.0388, ..., -0.0224, -0.0207, -0.0231], + [ 0.0104, 0.0291, -0.0347, ..., -0.0037, 0.0053, -0.0411], + [ 0.0127, -0.0370, -0.0289, ..., -0.0231, 0.0238, -0.0178], + ..., + [-0.0112, -0.0347, 0.0253, ..., 0.0440, -0.0477, 0.0048], + [ 0.0181, 0.0218, -0.0690, ..., -0.0101, 0.0507, -0.0101], + [-0.0214, -0.0236, 0.0270, ..., 0.0015, 0.0013, -0.0309]], + device='cuda:0'), grad: tensor([[ 3.1757e-04, 1.9801e-04, 3.2097e-05, ..., 0.0000e+00, + 1.1617e-04, 0.0000e+00], + [-2.4872e-03, -5.6610e-03, 2.3639e-04, ..., 0.0000e+00, + -2.4357e-03, 0.0000e+00], + [-8.4519e-05, 2.6679e-04, 1.0282e-04, ..., 0.0000e+00, + -5.6934e-04, 0.0000e+00], + ..., + [ 1.8060e-04, 6.0463e-04, 7.2861e-04, ..., 0.0000e+00, + 2.1088e-04, 0.0000e+00], + [ 2.5425e-03, 4.7569e-03, 1.3852e-04, ..., 0.0000e+00, + 2.1343e-03, 0.0000e+00], + [ 2.0254e-04, -1.0061e-03, -2.1477e-03, ..., 0.0000e+00, + 1.4770e-04, 0.0000e+00]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0084, 0.0236, -0.0156, 0.0123, 0.0245, -0.0206, -0.0293, -0.0204, + -0.0170, -0.0231], device='cuda:0'), grad: tensor([ 1.9598e-04, -4.3526e-03, 3.4243e-05, 5.5542e-03, 1.5259e-03, + -5.7678e-03, -3.2377e-04, 2.0218e-03, 4.6921e-03, -3.5820e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 217.05, cls_loss 0.0421 cls_loss_mapping 0.0660 cls_loss_causal 0.9220 re_mapping 0.0325 re_causal 0.0881 /// teacc 98.45 lr 0.00010000 +Epoch 19, weight, value: tensor([[-3.3687e-02, -1.8973e-02, -3.9390e-02, ..., -2.2173e-02, + -2.1704e-02, -2.3315e-02], + [ 1.0102e-02, 2.9863e-02, -3.5722e-02, ..., -3.7539e-03, + 4.6851e-03, -4.2936e-02], + [ 1.2609e-02, -3.7429e-02, -2.9608e-02, ..., -2.3292e-02, + 2.4234e-02, -1.9114e-02], + ..., + [-1.1490e-02, -3.5047e-02, 2.6501e-02, ..., 4.3868e-02, + -4.8378e-02, 5.2001e-03], + [ 1.8665e-02, 2.1461e-02, -7.0798e-02, ..., -1.0408e-02, + 5.1416e-02, -8.6810e-03], + [-2.1502e-02, -2.3988e-02, 2.7508e-02, ..., 1.4138e-03, + 5.0486e-05, -3.1226e-02]], device='cuda:0'), grad: tensor([[ 7.6413e-05, -5.0254e-06, 3.1859e-05, ..., 0.0000e+00, + -2.4939e-04, 0.0000e+00], + [ 3.9268e-04, 2.8062e-04, 7.7152e-04, ..., 0.0000e+00, + 3.5019e-03, 0.0000e+00], + [-1.5898e-03, 5.0735e-04, 3.7193e-04, ..., 0.0000e+00, + -6.7062e-03, 0.0000e+00], + ..., + [ 4.5371e-04, -2.1191e-03, -1.5564e-03, ..., 0.0000e+00, + 1.9140e-03, 0.0000e+00], + [ 5.2500e-04, 1.0242e-03, 9.2936e-04, ..., 0.0000e+00, + 5.0163e-04, 0.0000e+00], + [ 2.0778e-04, 1.7560e-04, -5.1403e-04, ..., 0.0000e+00, + 2.1112e-04, 0.0000e+00]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0088, 0.0233, -0.0155, 0.0129, 0.0246, -0.0206, -0.0297, -0.0202, + -0.0169, -0.0232], device='cuda:0'), grad: tensor([-0.0010, 0.0041, -0.0056, 0.0019, -0.0006, -0.0009, 0.0009, -0.0013, + 0.0029, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 216.95, cls_loss 0.0399 cls_loss_mapping 0.0616 cls_loss_causal 0.8926 re_mapping 0.0297 re_causal 0.0823 /// teacc 98.46 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0350, -0.0201, -0.0399, ..., -0.0222, -0.0222, -0.0233], + [ 0.0099, 0.0306, -0.0364, ..., -0.0031, 0.0047, -0.0432], + [ 0.0125, -0.0384, -0.0304, ..., -0.0240, 0.0248, -0.0192], + ..., + [-0.0121, -0.0352, 0.0267, ..., 0.0441, -0.0497, 0.0052], + [ 0.0195, 0.0216, -0.0725, ..., -0.0111, 0.0519, -0.0086], + [-0.0221, -0.0247, 0.0278, ..., 0.0013, -0.0004, -0.0313]], + device='cuda:0'), grad: tensor([[ 5.1737e-04, 6.5744e-05, 3.7730e-05, ..., 0.0000e+00, + 3.0422e-04, 0.0000e+00], + [ 4.0436e-04, -5.8383e-05, 2.3270e-04, ..., 0.0000e+00, + 5.4455e-04, 0.0000e+00], + [-6.8140e-04, 1.5950e-04, 3.5143e-04, ..., 0.0000e+00, + -2.6360e-03, 0.0000e+00], + ..., + [ 2.1219e-04, 2.6211e-05, -1.6804e-03, ..., 0.0000e+00, + 1.5569e-04, 0.0000e+00], + [-4.0169e-03, -1.3447e-03, 2.2209e-04, ..., 0.0000e+00, + -6.7253e-03, 0.0000e+00], + [ 1.4818e-04, 4.2081e-05, -1.8895e-04, ..., 0.0000e+00, + 2.1696e-04, 0.0000e+00]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0089, 0.0231, -0.0155, 0.0129, 0.0247, -0.0208, -0.0294, -0.0203, + -0.0169, -0.0231], device='cuda:0'), grad: tensor([ 5.0774e-03, 1.5125e-03, -6.1512e-04, 4.2000e-03, 1.9627e-03, + -1.5335e-02, 1.1261e-02, 6.0827e-05, -8.6212e-03, 5.0497e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 216.97, cls_loss 0.0393 cls_loss_mapping 0.0622 cls_loss_causal 0.8830 re_mapping 0.0292 re_causal 0.0823 /// teacc 97.98 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0360, -0.0214, -0.0401, ..., -0.0223, -0.0222, -0.0232], + [ 0.0097, 0.0312, -0.0368, ..., -0.0026, 0.0041, -0.0434], + [ 0.0125, -0.0387, -0.0307, ..., -0.0244, 0.0255, -0.0193], + ..., + [-0.0124, -0.0347, 0.0274, ..., 0.0442, -0.0498, 0.0053], + [ 0.0201, 0.0214, -0.0731, ..., -0.0112, 0.0525, -0.0088], + [-0.0224, -0.0251, 0.0281, ..., 0.0013, -0.0012, -0.0314]], + device='cuda:0'), grad: tensor([[ 6.4802e-04, 5.0688e-04, 1.4544e-04, ..., 1.5751e-05, + 4.9257e-04, 0.0000e+00], + [ 1.9207e-03, 1.9646e-03, 4.7255e-04, ..., 5.7369e-05, + 4.0169e-03, 0.0000e+00], + [-1.0025e-02, -4.5433e-03, -5.4932e-03, ..., 3.1209e-04, + -2.0645e-02, 0.0000e+00], + ..., + [ 2.3136e-03, 3.2735e-04, 2.2583e-03, ..., 9.1493e-05, + 5.1422e-03, 0.0000e+00], + [ 3.3760e-03, 2.3899e-03, 2.3174e-03, ..., 4.4137e-05, + 7.8049e-03, 0.0000e+00], + [ 3.3903e-04, 3.2067e-04, 1.7347e-03, ..., 1.4275e-05, + 5.8222e-04, 0.0000e+00]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0089, 0.0230, -0.0154, 0.0128, 0.0244, -0.0210, -0.0293, -0.0201, + -0.0167, -0.0231], device='cuda:0'), grad: tensor([ 2.0256e-03, 5.4359e-03, -1.9470e-02, -8.5068e-03, 7.5758e-05, + 1.9646e-03, -2.2469e-03, 7.8659e-03, 8.4991e-03, 4.3449e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 216.55, cls_loss 0.0378 cls_loss_mapping 0.0593 cls_loss_causal 0.8905 re_mapping 0.0278 re_causal 0.0772 /// teacc 98.26 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0367, -0.0224, -0.0393, ..., -0.0223, -0.0223, -0.0229], + [ 0.0092, 0.0318, -0.0372, ..., -0.0026, 0.0038, -0.0435], + [ 0.0128, -0.0402, -0.0313, ..., -0.0248, 0.0266, -0.0194], + ..., + [-0.0129, -0.0344, 0.0288, ..., 0.0442, -0.0504, 0.0053], + [ 0.0205, 0.0215, -0.0744, ..., -0.0114, 0.0526, -0.0089], + [-0.0224, -0.0256, 0.0281, ..., 0.0013, -0.0015, -0.0318]], + device='cuda:0'), grad: tensor([[ 1.4820e-03, 2.3782e-04, 9.7334e-05, ..., 3.3565e-06, + 8.7833e-04, 0.0000e+00], + [ 1.3888e-04, -5.9557e-04, 4.6998e-05, ..., 2.4983e-07, + 4.9233e-05, 0.0000e+00], + [ 3.1734e-04, 1.6320e-04, 1.3614e-04, ..., 2.4773e-07, + 1.3721e-04, 0.0000e+00], + ..., + [ 1.6022e-04, 2.4748e-04, -6.8903e-05, ..., 8.3819e-07, + 1.0455e-04, 0.0000e+00], + [-3.9749e-03, 2.2948e-04, -7.1883e-05, ..., 1.5423e-06, + -3.2921e-03, 0.0000e+00], + [ 1.8454e-03, 4.4882e-05, -2.0051e-04, ..., 4.1425e-06, + 1.3609e-03, 0.0000e+00]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0085, 0.0227, -0.0153, 0.0130, 0.0243, -0.0212, -0.0294, -0.0197, + -0.0170, -0.0231], device='cuda:0'), grad: tensor([ 0.0032, -0.0002, 0.0011, 0.0014, 0.0003, 0.0009, -0.0015, 0.0004, + -0.0093, 0.0036], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 217.56, cls_loss 0.0357 cls_loss_mapping 0.0550 cls_loss_causal 0.8442 re_mapping 0.0280 re_causal 0.0753 /// teacc 98.55 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0377, -0.0228, -0.0397, ..., -0.0224, -0.0220, -0.0226], + [ 0.0086, 0.0324, -0.0378, ..., -0.0028, 0.0036, -0.0437], + [ 0.0129, -0.0413, -0.0312, ..., -0.0247, 0.0269, -0.0195], + ..., + [-0.0140, -0.0350, 0.0294, ..., 0.0439, -0.0513, 0.0053], + [ 0.0213, 0.0216, -0.0760, ..., -0.0121, 0.0538, -0.0090], + [-0.0236, -0.0258, 0.0283, ..., 0.0012, -0.0026, -0.0321]], + device='cuda:0'), grad: tensor([[ 9.2506e-05, 2.5821e-04, 1.9455e-04, ..., 0.0000e+00, + 8.3303e-04, 1.0314e-07], + [ 4.3583e-04, 1.5335e-03, 1.2426e-03, ..., 0.0000e+00, + 1.3552e-03, 3.9581e-07], + [ 6.2108e-05, -6.8760e-04, 2.1517e-04, ..., 0.0000e+00, + 6.7472e-04, 9.9465e-07], + ..., + [ 9.0241e-05, 1.7605e-03, 2.1629e-03, ..., 0.0000e+00, + 1.1415e-03, 1.0179e-06], + [-1.8225e-03, -6.2065e-03, -7.0686e-03, ..., 0.0000e+00, + -3.8681e-03, 1.0536e-07], + [ 1.7536e-04, 2.2984e-03, 3.0193e-03, ..., 0.0000e+00, + 2.1019e-03, 2.2016e-06]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0084, 0.0226, -0.0153, 0.0132, 0.0243, -0.0210, -0.0294, -0.0199, + -0.0170, -0.0233], device='cuda:0'), grad: tensor([ 0.0020, 0.0044, 0.0014, -0.0041, -0.0036, 0.0024, 0.0007, 0.0056, + -0.0176, 0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 217.04, cls_loss 0.0300 cls_loss_mapping 0.0477 cls_loss_causal 0.8279 re_mapping 0.0264 re_causal 0.0732 /// teacc 98.51 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0384, -0.0236, -0.0401, ..., -0.0223, -0.0225, -0.0221], + [ 0.0082, 0.0330, -0.0379, ..., -0.0006, 0.0031, -0.0447], + [ 0.0130, -0.0417, -0.0312, ..., -0.0265, 0.0276, -0.0196], + ..., + [-0.0147, -0.0353, 0.0300, ..., 0.0439, -0.0519, 0.0050], + [ 0.0216, 0.0216, -0.0776, ..., -0.0123, 0.0542, -0.0094], + [-0.0239, -0.0262, 0.0287, ..., 0.0012, -0.0035, -0.0327]], + device='cuda:0'), grad: tensor([[ 1.9324e-04, 8.3148e-06, 1.1241e-04, ..., 0.0000e+00, + 2.9421e-04, 3.5344e-07], + [ 2.4390e-04, -1.9276e-04, 1.1557e-04, ..., 0.0000e+00, + 3.6359e-04, 1.4680e-07], + [-1.4009e-03, 6.7353e-05, 1.1024e-03, ..., 0.0000e+00, + -1.7328e-03, -5.8562e-06], + ..., + [ 8.9693e-04, 1.0484e-04, 1.9188e-03, ..., 0.0000e+00, + 1.4544e-03, 4.2357e-06], + [-1.1539e-03, 4.0621e-05, 2.0897e-04, ..., 0.0000e+00, + -2.0466e-03, 5.3272e-07], + [ 6.4659e-04, -2.3377e-04, -5.5923e-03, ..., 0.0000e+00, + 9.7370e-04, 3.4971e-07]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0087, 0.0223, -0.0152, 0.0131, 0.0244, -0.0208, -0.0299, -0.0196, + -0.0171, -0.0230], device='cuda:0'), grad: tensor([ 0.0006, 0.0006, -0.0008, -0.0009, 0.0032, -0.0047, 0.0008, 0.0040, + -0.0001, -0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 216.88, cls_loss 0.0301 cls_loss_mapping 0.0520 cls_loss_causal 0.8519 re_mapping 0.0256 re_causal 0.0752 /// teacc 98.48 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0390, -0.0244, -0.0398, ..., -0.0223, -0.0230, -0.0207], + [ 0.0077, 0.0335, -0.0382, ..., -0.0006, 0.0026, -0.0454], + [ 0.0133, -0.0422, -0.0321, ..., -0.0266, 0.0284, -0.0208], + ..., + [-0.0156, -0.0356, 0.0305, ..., 0.0439, -0.0529, 0.0049], + [ 0.0221, 0.0216, -0.0784, ..., -0.0124, 0.0547, -0.0100], + [-0.0245, -0.0272, 0.0283, ..., 0.0013, -0.0045, -0.0331]], + device='cuda:0'), grad: tensor([[ 1.4865e-04, 5.5790e-05, 5.8711e-05, ..., 0.0000e+00, + -1.0651e-04, 0.0000e+00], + [ 6.7770e-05, -3.3677e-06, 1.3399e-04, ..., 0.0000e+00, + 1.0103e-04, 0.0000e+00], + [-1.2217e-03, 1.4544e-03, 8.9741e-04, ..., 0.0000e+00, + 1.4896e-03, 0.0000e+00], + ..., + [ 9.9838e-05, 6.6102e-05, -8.5211e-04, ..., 0.0000e+00, + 1.1873e-04, 0.0000e+00], + [ 9.0122e-04, 3.8028e-04, 2.4080e-04, ..., 0.0000e+00, + 8.8930e-05, 0.0000e+00], + [ 4.3958e-05, 1.1915e-04, -8.9493e-03, ..., 0.0000e+00, + 5.4002e-05, 0.0000e+00]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0085, 0.0220, -0.0151, 0.0133, 0.0247, -0.0207, -0.0299, -0.0197, + -0.0172, -0.0233], device='cuda:0'), grad: tensor([-0.0004, 0.0004, 0.0031, 0.0019, 0.0156, 0.0016, -0.0013, -0.0021, + 0.0013, -0.0200], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 216.77, cls_loss 0.0293 cls_loss_mapping 0.0471 cls_loss_causal 0.8177 re_mapping 0.0254 re_causal 0.0716 /// teacc 98.38 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0398, -0.0252, -0.0387, ..., -0.0222, -0.0234, -0.0193], + [ 0.0071, 0.0341, -0.0381, ..., -0.0006, 0.0022, -0.0474], + [ 0.0135, -0.0430, -0.0317, ..., -0.0267, 0.0290, -0.0218], + ..., + [-0.0162, -0.0368, 0.0307, ..., 0.0439, -0.0534, 0.0048], + [ 0.0228, 0.0221, -0.0799, ..., -0.0125, 0.0560, -0.0114], + [-0.0257, -0.0279, 0.0284, ..., 0.0013, -0.0064, -0.0336]], + device='cuda:0'), grad: tensor([[ 1.4710e-04, 6.2585e-05, 1.8135e-05, ..., 3.9814e-08, + 2.2805e-04, 1.8941e-07], + [-6.5041e-04, -2.0485e-03, 4.4674e-05, ..., 4.4145e-07, + -1.0862e-03, 6.8732e-07], + [-4.2677e-05, 1.3399e-03, 1.5521e-04, ..., 1.3076e-06, + -3.6311e-04, 9.9652e-07], + ..., + [ 1.2028e-04, 2.4080e-04, -1.4174e-04, ..., 3.9884e-07, + 2.3437e-04, 5.5581e-06], + [ 3.3092e-04, 1.0186e-04, 1.6439e-04, ..., 3.0571e-07, + 7.9823e-04, 7.8836e-07], + [ 1.6049e-05, 1.8388e-05, -4.1962e-04, ..., 1.8545e-07, + -3.4547e-04, 2.2016e-06]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0080, 0.0222, -0.0151, 0.0127, 0.0247, -0.0204, -0.0300, -0.0201, + -0.0165, -0.0237], device='cuda:0'), grad: tensor([ 1.1928e-05, -3.2005e-03, 1.6117e-03, 8.8930e-04, 4.2856e-05, + 8.6164e-04, -5.6535e-05, 1.9598e-04, 1.4105e-03, -1.7662e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 216.81, cls_loss 0.0344 cls_loss_mapping 0.0552 cls_loss_causal 0.8086 re_mapping 0.0236 re_causal 0.0681 /// teacc 98.28 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0406, -0.0264, -0.0387, ..., -0.0230, -0.0237, -0.0188], + [ 0.0069, 0.0346, -0.0384, ..., -0.0004, 0.0022, -0.0492], + [ 0.0138, -0.0431, -0.0322, ..., -0.0274, 0.0297, -0.0217], + ..., + [-0.0170, -0.0366, 0.0318, ..., 0.0449, -0.0545, 0.0045], + [ 0.0229, 0.0214, -0.0817, ..., -0.0130, 0.0562, -0.0103], + [-0.0263, -0.0285, 0.0289, ..., 0.0011, -0.0075, -0.0343]], + device='cuda:0'), grad: tensor([[ 3.3593e-04, 1.8013e-04, 2.0996e-05, ..., 0.0000e+00, + 1.0246e-04, -1.8496e-06], + [ 5.5599e-04, -3.9697e-04, -2.6733e-05, ..., 0.0000e+00, + 1.7624e-03, 9.1968e-09], + [-1.6375e-03, -1.1396e-03, -3.6736e-03, ..., 0.0000e+00, + -7.2937e-03, 6.7754e-08], + ..., + [ 2.2149e-04, 1.9336e-04, 2.4056e-04, ..., 0.0000e+00, + 5.1975e-04, 1.7113e-08], + [ 2.4471e-03, 9.3317e-04, 1.8501e-04, ..., 0.0000e+00, + 8.5831e-04, 5.9721e-08], + [ 2.7442e-04, 1.3065e-04, -2.7061e-04, ..., 0.0000e+00, + 2.2268e-04, 6.7847e-07]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0083, 0.0222, -0.0150, 0.0129, 0.0245, -0.0208, -0.0297, -0.0196, + -0.0169, -0.0235], device='cuda:0'), grad: tensor([ 0.0004, 0.0017, -0.0067, -0.0002, 0.0044, -0.0054, 0.0011, 0.0009, + 0.0037, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 217.35, cls_loss 0.0245 cls_loss_mapping 0.0407 cls_loss_causal 0.8231 re_mapping 0.0228 re_causal 0.0693 /// teacc 98.61 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0409, -0.0267, -0.0390, ..., -0.0230, -0.0236, -0.0188], + [ 0.0064, 0.0351, -0.0390, ..., -0.0004, 0.0021, -0.0493], + [ 0.0138, -0.0442, -0.0328, ..., -0.0274, 0.0300, -0.0217], + ..., + [-0.0175, -0.0368, 0.0328, ..., 0.0449, -0.0552, 0.0045], + [ 0.0234, 0.0215, -0.0827, ..., -0.0130, 0.0568, -0.0103], + [-0.0274, -0.0292, 0.0295, ..., 0.0011, -0.0086, -0.0343]], + device='cuda:0'), grad: tensor([[ 9.0361e-05, 1.0705e-04, 2.6807e-05, ..., 3.2666e-07, + 9.8944e-05, 0.0000e+00], + [ 7.3671e-05, -5.2595e-04, 5.5343e-05, ..., 2.3395e-06, + 1.1879e-04, 0.0000e+00], + [-1.9327e-05, 8.4221e-05, 3.5584e-05, ..., 1.6624e-06, + -3.9554e-04, 0.0000e+00], + ..., + [ 9.9421e-05, 2.4021e-04, 3.4857e-04, ..., -1.5557e-04, + 2.2185e-04, 0.0000e+00], + [-5.3310e-04, 8.0943e-05, -7.4804e-06, ..., 1.8897e-06, + -6.0892e-04, 0.0000e+00], + [ 1.2863e-04, 1.2312e-03, 2.6073e-03, ..., 1.1450e-04, + 1.6737e-04, 0.0000e+00]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0078, 0.0221, -0.0154, 0.0128, 0.0244, -0.0202, -0.0300, -0.0194, + -0.0169, -0.0237], device='cuda:0'), grad: tensor([-1.7061e-03, -8.1301e-05, -1.3089e-04, 4.0936e-04, -3.1757e-03, + 6.7186e-04, 7.5996e-05, 1.9722e-03, -5.9891e-04, 2.5654e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 27---------------------------------------------------- +epoch 27, time 217.41, cls_loss 0.0209 cls_loss_mapping 0.0365 cls_loss_causal 0.7677 re_mapping 0.0231 re_causal 0.0672 /// teacc 98.70 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0417, -0.0270, -0.0394, ..., -0.0228, -0.0238, -0.0183], + [ 0.0062, 0.0357, -0.0395, ..., -0.0005, 0.0020, -0.0498], + [ 0.0142, -0.0446, -0.0330, ..., -0.0275, 0.0307, -0.0219], + ..., + [-0.0185, -0.0375, 0.0334, ..., 0.0450, -0.0563, 0.0044], + [ 0.0236, 0.0214, -0.0840, ..., -0.0131, 0.0571, -0.0096], + [-0.0276, -0.0295, 0.0299, ..., 0.0011, -0.0095, -0.0348]], + device='cuda:0'), grad: tensor([[ 6.1989e-05, 2.9206e-05, 4.2319e-05, ..., 0.0000e+00, + 3.0085e-05, 0.0000e+00], + [ 1.1986e-04, -6.8903e-05, 3.5942e-05, ..., 0.0000e+00, + 1.3280e-04, 0.0000e+00], + [ 1.0139e-04, 4.1246e-05, 8.9216e-04, ..., 0.0000e+00, + 3.4666e-04, 0.0000e+00], + ..., + [ 1.2803e-04, 4.4882e-05, -2.1133e-03, ..., 0.0000e+00, + -5.8889e-04, 0.0000e+00], + [ 4.3631e-04, 1.2070e-04, 7.1824e-05, ..., 0.0000e+00, + 2.6655e-04, 0.0000e+00], + [ 5.7787e-05, 4.3094e-05, 1.1641e-04, ..., 0.0000e+00, + 4.8548e-05, 0.0000e+00]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0076, 0.0219, -0.0152, 0.0127, 0.0246, -0.0203, -0.0301, -0.0196, + -0.0168, -0.0236], device='cuda:0'), grad: tensor([ 6.3062e-05, 2.3139e-04, 1.3084e-03, -9.6655e-04, 1.2165e-04, + 1.8311e-04, 1.0055e-04, -2.1610e-03, 8.8692e-04, 2.3305e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 216.97, cls_loss 0.0254 cls_loss_mapping 0.0408 cls_loss_causal 0.8584 re_mapping 0.0210 re_causal 0.0681 /// teacc 98.69 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0425, -0.0272, -0.0398, ..., -0.0228, -0.0241, -0.0185], + [ 0.0054, 0.0358, -0.0399, ..., -0.0005, 0.0012, -0.0502], + [ 0.0142, -0.0451, -0.0333, ..., -0.0275, 0.0311, -0.0217], + ..., + [-0.0190, -0.0374, 0.0344, ..., 0.0451, -0.0567, 0.0049], + [ 0.0237, 0.0218, -0.0852, ..., -0.0131, 0.0574, -0.0097], + [-0.0282, -0.0301, 0.0302, ..., 0.0010, -0.0102, -0.0349]], + device='cuda:0'), grad: tensor([[ 4.4250e-04, 3.7342e-05, 7.8022e-05, ..., 0.0000e+00, + 1.9705e-04, 0.0000e+00], + [ 2.5272e-04, -1.8835e-04, 1.3925e-05, ..., 0.0000e+00, + 2.8801e-04, 0.0000e+00], + [ 6.6900e-04, 3.5256e-05, 9.1219e-04, ..., 0.0000e+00, + 1.4467e-03, 0.0000e+00], + ..., + [ 6.2981e-03, 3.9190e-05, 1.7996e-03, ..., 0.0000e+00, + 9.4299e-03, 0.0000e+00], + [-1.0109e-02, 2.9773e-05, -2.1877e-03, ..., 0.0000e+00, + -1.5594e-02, 0.0000e+00], + [-2.8820e-03, 2.8118e-05, -1.1597e-03, ..., 0.0000e+00, + 6.0034e-04, 0.0000e+00]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0075, 0.0211, -0.0153, 0.0125, 0.0247, -0.0198, -0.0303, -0.0189, + -0.0168, -0.0239], device='cuda:0'), grad: tensor([ 0.0002, 0.0004, 0.0018, 0.0020, 0.0028, 0.0065, 0.0031, 0.0145, + -0.0199, -0.0114], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 216.88, cls_loss 0.0218 cls_loss_mapping 0.0366 cls_loss_causal 0.7505 re_mapping 0.0209 re_causal 0.0604 /// teacc 98.69 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0429, -0.0282, -0.0402, ..., -0.0229, -0.0243, -0.0184], + [ 0.0049, 0.0367, -0.0396, ..., -0.0005, 0.0011, -0.0504], + [ 0.0145, -0.0465, -0.0344, ..., -0.0276, 0.0314, -0.0214], + ..., + [-0.0209, -0.0380, 0.0355, ..., 0.0452, -0.0580, 0.0049], + [ 0.0242, 0.0216, -0.0861, ..., -0.0133, 0.0583, -0.0098], + [-0.0290, -0.0303, 0.0305, ..., 0.0010, -0.0112, -0.0353]], + device='cuda:0'), grad: tensor([[ 1.4290e-05, 3.0309e-05, 5.0187e-05, ..., 0.0000e+00, + 2.6846e-04, -6.5845e-07], + [ 1.3053e-05, 1.8530e-03, 1.3418e-03, ..., 0.0000e+00, + 6.3777e-05, 2.5611e-09], + [-6.6638e-05, 4.4018e-05, -1.0747e-04, ..., 0.0000e+00, + 1.7118e-04, 1.2200e-07], + ..., + [ 2.2173e-05, 1.4806e-04, -2.3186e-05, ..., 0.0000e+00, + -2.4471e-03, 1.0943e-08], + [-6.1989e-05, 1.1647e-04, 9.9182e-05, ..., 0.0000e+00, + 2.3186e-05, 2.4564e-08], + [ 3.6120e-05, -3.4027e-03, -2.1667e-03, ..., 0.0000e+00, + 1.3847e-03, 3.2899e-07]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0074, 0.0216, -0.0157, 0.0129, 0.0241, -0.0201, -0.0300, -0.0189, + -0.0168, -0.0238], device='cuda:0'), grad: tensor([ 0.0003, 0.0027, 0.0043, 0.0006, 0.0013, 0.0003, 0.0001, -0.0066, + 0.0003, -0.0032], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 216.74, cls_loss 0.0180 cls_loss_mapping 0.0328 cls_loss_causal 0.7582 re_mapping 0.0198 re_causal 0.0601 /// teacc 98.66 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0434, -0.0291, -0.0406, ..., -0.0229, -0.0244, -0.0183], + [ 0.0044, 0.0372, -0.0400, ..., -0.0005, 0.0008, -0.0510], + [ 0.0147, -0.0470, -0.0352, ..., -0.0276, 0.0319, -0.0215], + ..., + [-0.0215, -0.0388, 0.0360, ..., 0.0452, -0.0587, 0.0049], + [ 0.0246, 0.0221, -0.0871, ..., -0.0133, 0.0588, -0.0099], + [-0.0299, -0.0308, 0.0308, ..., 0.0010, -0.0122, -0.0355]], + device='cuda:0'), grad: tensor([[ 8.4639e-05, 2.3305e-05, 2.6911e-05, ..., 0.0000e+00, + 7.5400e-05, -6.6614e-04], + [ 3.8981e-05, -1.3697e-04, 5.8472e-05, ..., 0.0000e+00, + 7.4208e-05, 6.9961e-06], + [-5.9748e-04, 1.3816e-04, -1.6439e-04, ..., 0.0000e+00, + -8.2159e-04, 4.1008e-04], + ..., + [ 4.8369e-05, 7.7188e-05, 1.9535e-05, ..., 0.0000e+00, + 3.4285e-04, 3.3766e-05], + [ 4.1276e-05, 9.4950e-05, 2.0623e-04, ..., 0.0000e+00, + 3.3617e-04, 1.7375e-05], + [ 9.3341e-05, 4.9114e-04, -9.5606e-05, ..., 0.0000e+00, + 2.9588e-04, 7.3969e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0077, 0.0213, -0.0157, 0.0130, 0.0242, -0.0201, -0.0297, -0.0189, + -0.0167, -0.0239], device='cuda:0'), grad: tensor([-8.7786e-04, 1.4055e-04, -5.2881e-04, -1.3075e-03, 3.9876e-05, + 3.5119e-04, 4.9734e-04, 4.8876e-04, 1.0605e-03, 1.3673e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 216.83, cls_loss 0.0231 cls_loss_mapping 0.0388 cls_loss_causal 0.7574 re_mapping 0.0200 re_causal 0.0594 /// teacc 98.63 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0444, -0.0285, -0.0404, ..., -0.0229, -0.0242, -0.0176], + [ 0.0039, 0.0379, -0.0399, ..., -0.0004, 0.0005, -0.0521], + [ 0.0149, -0.0479, -0.0358, ..., -0.0276, 0.0326, -0.0222], + ..., + [-0.0221, -0.0403, 0.0363, ..., 0.0452, -0.0595, 0.0058], + [ 0.0251, 0.0220, -0.0883, ..., -0.0133, 0.0595, -0.0103], + [-0.0303, -0.0309, 0.0314, ..., 0.0010, -0.0136, -0.0361]], + device='cuda:0'), grad: tensor([[-5.6744e-04, -1.1091e-03, 1.9789e-04, ..., 0.0000e+00, + -6.6233e-04, 0.0000e+00], + [ 3.9846e-05, -1.3745e-04, 2.4509e-04, ..., 0.0000e+00, + 8.0228e-05, 0.0000e+00], + [-8.7440e-05, 1.2684e-04, 8.3017e-04, ..., 0.0000e+00, + 1.0386e-05, 0.0000e+00], + ..., + [ 1.3483e-04, 8.1003e-05, -2.6913e-03, ..., 0.0000e+00, + 4.9686e-04, 0.0000e+00], + [ 2.9469e-04, -8.5211e-04, -8.7881e-04, ..., 0.0000e+00, + -3.3200e-05, 0.0000e+00], + [ 1.5712e-04, 1.3123e-03, 2.5234e-03, ..., 0.0000e+00, + 8.8739e-04, 0.0000e+00]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0078, 0.0213, -0.0157, 0.0128, 0.0241, -0.0199, -0.0296, -0.0192, + -0.0166, -0.0235], device='cuda:0'), grad: tensor([-0.0030, 0.0005, 0.0011, -0.0018, 0.0004, 0.0008, 0.0018, -0.0042, + -0.0016, 0.0060], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 217.14, cls_loss 0.0197 cls_loss_mapping 0.0324 cls_loss_causal 0.7567 re_mapping 0.0195 re_causal 0.0593 /// teacc 98.63 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0448, -0.0292, -0.0406, ..., -0.0229, -0.0243, -0.0172], + [ 0.0032, 0.0380, -0.0404, ..., -0.0004, 0.0003, -0.0526], + [ 0.0152, -0.0493, -0.0366, ..., -0.0276, 0.0330, -0.0218], + ..., + [-0.0226, -0.0401, 0.0372, ..., 0.0452, -0.0603, 0.0057], + [ 0.0254, 0.0222, -0.0894, ..., -0.0134, 0.0599, -0.0106], + [-0.0310, -0.0317, 0.0315, ..., 0.0010, -0.0147, -0.0365]], + device='cuda:0'), grad: tensor([[ 5.2035e-05, 2.6894e-04, 1.1120e-03, ..., 0.0000e+00, + 1.1218e-04, 0.0000e+00], + [ 4.9055e-05, -6.6853e-04, 1.2010e-04, ..., 0.0000e+00, + -8.8215e-05, 0.0000e+00], + [ 2.7686e-05, 1.4496e-04, 1.2517e-04, ..., 0.0000e+00, + -4.9162e-04, 0.0000e+00], + ..., + [ 4.8816e-05, -2.4870e-05, -2.2984e-03, ..., 0.0000e+00, + 9.7215e-05, 0.0000e+00], + [-3.7527e-04, 1.4031e-04, 9.1970e-05, ..., 0.0000e+00, + -5.7459e-04, 0.0000e+00], + [ 9.5785e-05, 1.3423e-04, -2.8348e-04, ..., 0.0000e+00, + 9.6917e-05, 0.0000e+00]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0081, 0.0210, -0.0159, 0.0129, 0.0246, -0.0200, -0.0293, -0.0190, + -0.0167, -0.0238], device='cuda:0'), grad: tensor([ 0.0021, -0.0001, -0.0002, 0.0018, 0.0005, 0.0007, -0.0002, -0.0038, + -0.0005, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 217.44, cls_loss 0.0173 cls_loss_mapping 0.0298 cls_loss_causal 0.7670 re_mapping 0.0189 re_causal 0.0582 /// teacc 98.72 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0454, -0.0303, -0.0408, ..., -0.0229, -0.0248, -0.0170], + [ 0.0029, 0.0388, -0.0409, ..., -0.0004, 0.0003, -0.0531], + [ 0.0156, -0.0499, -0.0370, ..., -0.0277, 0.0336, -0.0218], + ..., + [-0.0231, -0.0409, 0.0376, ..., 0.0452, -0.0611, 0.0061], + [ 0.0259, 0.0223, -0.0904, ..., -0.0135, 0.0603, -0.0108], + [-0.0316, -0.0319, 0.0312, ..., 0.0010, -0.0156, -0.0373]], + device='cuda:0'), grad: tensor([[ 3.0971e-04, 5.9277e-05, 3.7607e-06, ..., 1.3607e-06, + 2.6321e-04, -2.9638e-05], + [ 6.7532e-05, 3.4481e-05, 2.0981e-05, ..., 9.8813e-07, + 7.2479e-05, 1.6636e-07], + [-2.7885e-03, 5.6863e-05, 4.2349e-05, ..., 1.7718e-07, + -5.0125e-03, 1.9874e-06], + ..., + [ 1.4687e-04, 5.6833e-05, 3.8326e-05, ..., 5.2480e-07, + 2.6608e-04, 5.0757e-07], + [-4.2725e-04, 1.1653e-04, 1.0020e-04, ..., 1.1331e-04, + -2.4402e-04, 8.5542e-07], + [ 6.3419e-05, 1.3202e-05, -1.9848e-04, ..., 5.1185e-06, + 8.9049e-05, 1.4715e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0081, 0.0211, -0.0159, 0.0132, 0.0247, -0.0202, -0.0291, -0.0192, + -0.0167, -0.0240], device='cuda:0'), grad: tensor([ 3.7766e-04, 1.2875e-04, -4.8523e-03, 3.7155e-03, -2.5511e-05, + -1.0890e-04, 6.5136e-04, 3.5834e-04, -2.0540e-04, -3.2663e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 216.79, cls_loss 0.0163 cls_loss_mapping 0.0313 cls_loss_causal 0.7729 re_mapping 0.0179 re_causal 0.0584 /// teacc 98.46 lr 0.00010000 +Epoch 36, weight, value: tensor([[-4.6134e-02, -3.0866e-02, -4.1302e-02, ..., -2.2904e-02, + -2.5376e-02, -1.6938e-02], + [ 2.3908e-03, 3.9476e-02, -4.1275e-02, ..., 4.0723e-05, + 5.1983e-04, -5.3277e-02], + [ 1.6075e-02, -5.0772e-02, -3.6967e-02, ..., -2.7976e-02, + 3.4245e-02, -2.1628e-02], + ..., + [-2.3878e-02, -4.0865e-02, 3.8867e-02, ..., 4.5177e-02, + -6.2036e-02, 6.0404e-03], + [ 2.6227e-02, 2.2156e-02, -9.1449e-02, ..., -1.3834e-02, + 6.0624e-02, -1.0814e-02], + [-3.2237e-02, -3.2428e-02, 3.1464e-02, ..., 1.0267e-03, + -1.6312e-02, -3.7305e-02]], device='cuda:0'), grad: tensor([[-1.4879e-05, 2.4941e-06, 1.3578e-04, ..., 0.0000e+00, + 1.4462e-05, 0.0000e+00], + [ 4.9651e-05, -4.4870e-04, -8.6844e-05, ..., 0.0000e+00, + 1.6332e-04, 0.0000e+00], + [ 1.4281e-04, 1.4234e-04, 1.0204e-04, ..., 0.0000e+00, + 3.4547e-04, 0.0000e+00], + ..., + [ 2.6911e-05, 1.1194e-04, -8.5413e-05, ..., 0.0000e+00, + 4.6551e-05, 0.0000e+00], + [-2.7204e-04, 7.2598e-05, 9.6500e-05, ..., 0.0000e+00, + -4.6420e-04, 0.0000e+00], + [ 2.6524e-05, 1.4782e-05, -5.8270e-04, ..., 0.0000e+00, + -6.1131e-04, 0.0000e+00]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0081, 0.0213, -0.0159, 0.0135, 0.0243, -0.0204, -0.0291, -0.0188, + -0.0169, -0.0239], device='cuda:0'), grad: tensor([ 3.3998e-04, -2.6941e-04, 8.3637e-04, 1.7948e-03, 2.0492e-04, + -1.9562e-04, 1.8489e-04, 6.0111e-05, -6.5565e-05, -2.8915e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 216.59, cls_loss 0.0185 cls_loss_mapping 0.0332 cls_loss_causal 0.7768 re_mapping 0.0182 re_causal 0.0546 /// teacc 98.59 lr 0.00010000 +Epoch 37, weight, value: tensor([[-4.6816e-02, -3.1639e-02, -4.2059e-02, ..., -2.3071e-02, + -2.5756e-02, -1.6679e-02], + [ 1.5957e-03, 4.0093e-02, -4.1226e-02, ..., 2.1183e-04, + 4.6132e-05, -5.3881e-02], + [ 1.6526e-02, -5.1302e-02, -3.7503e-02, ..., -2.8239e-02, + 3.4891e-02, -2.1590e-02], + ..., + [-2.4597e-02, -4.1614e-02, 3.9892e-02, ..., 4.5131e-02, + -6.3124e-02, 5.9929e-03], + [ 2.6572e-02, 2.2203e-02, -9.2494e-02, ..., -1.4103e-02, + 6.1406e-02, -1.1038e-02], + [-3.2968e-02, -3.3237e-02, 3.0910e-02, ..., 1.0046e-03, + -1.7226e-02, -3.8440e-02]], device='cuda:0'), grad: tensor([[ 6.4492e-05, 4.3362e-05, 1.3620e-05, ..., 0.0000e+00, + 5.2214e-05, 0.0000e+00], + [ 4.5156e-04, 1.3113e-04, 9.2164e-06, ..., 0.0000e+00, + 7.1955e-04, 0.0000e+00], + [-8.3447e-04, -3.2067e-04, -6.5975e-06, ..., 0.0000e+00, + -1.4114e-03, 0.0000e+00], + ..., + [ 1.3781e-04, 6.2466e-05, 1.4007e-05, ..., 0.0000e+00, + 2.0564e-04, 0.0000e+00], + [-2.3925e-04, -3.4833e-04, 2.8849e-05, ..., 0.0000e+00, + -8.8453e-05, 0.0000e+00], + [ 2.2757e-04, 2.1327e-04, -7.9632e-05, ..., 0.0000e+00, + 1.7917e-04, 0.0000e+00]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0089, 0.0213, -0.0160, 0.0134, 0.0248, -0.0201, -0.0294, -0.0186, + -0.0169, -0.0241], device='cuda:0'), grad: tensor([ 0.0001, 0.0010, -0.0019, 0.0002, 0.0003, -0.0005, 0.0004, 0.0003, + -0.0003, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 217.02, cls_loss 0.0155 cls_loss_mapping 0.0272 cls_loss_causal 0.7149 re_mapping 0.0176 re_causal 0.0552 /// teacc 98.72 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0472, -0.0316, -0.0422, ..., -0.0230, -0.0258, -0.0165], + [ 0.0011, 0.0402, -0.0418, ..., 0.0007, -0.0005, -0.0540], + [ 0.0167, -0.0519, -0.0376, ..., -0.0284, 0.0354, -0.0215], + ..., + [-0.0252, -0.0410, 0.0399, ..., 0.0451, -0.0638, 0.0059], + [ 0.0269, 0.0224, -0.0932, ..., -0.0142, 0.0618, -0.0111], + [-0.0335, -0.0339, 0.0311, ..., 0.0010, -0.0177, -0.0387]], + device='cuda:0'), grad: tensor([[-7.9691e-05, -3.0231e-04, 7.4655e-06, ..., 0.0000e+00, + 1.3866e-05, 0.0000e+00], + [ 1.7241e-05, -2.7561e-04, 4.0233e-05, ..., 0.0000e+00, + -1.0306e-04, 0.0000e+00], + [-5.7191e-05, 2.6703e-04, -1.9863e-05, ..., 0.0000e+00, + 1.2442e-06, 0.0000e+00], + ..., + [ 5.6386e-05, 6.7532e-05, -4.7445e-05, ..., 0.0000e+00, + 8.1003e-05, 0.0000e+00], + [-8.9169e-05, 6.9618e-05, 9.1076e-05, ..., 0.0000e+00, + -1.1134e-04, 0.0000e+00], + [ 1.1764e-05, 3.1114e-04, 3.4237e-04, ..., 0.0000e+00, + 1.0446e-05, 0.0000e+00]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0082, 0.0206, -0.0157, 0.0131, 0.0250, -0.0199, -0.0292, -0.0185, + -0.0170, -0.0244], device='cuda:0'), grad: tensor([-1.0138e-03, -2.0659e-04, 1.0759e-04, 1.3018e-04, -6.2704e-04, + 1.1325e-04, 7.9918e-04, 7.2300e-05, 8.4877e-05, 5.4026e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 216.82, cls_loss 0.0132 cls_loss_mapping 0.0234 cls_loss_causal 0.7466 re_mapping 0.0168 re_causal 0.0532 /// teacc 98.63 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0479, -0.0316, -0.0426, ..., -0.0229, -0.0262, -0.0164], + [ 0.0008, 0.0410, -0.0421, ..., 0.0007, -0.0007, -0.0542], + [ 0.0174, -0.0522, -0.0381, ..., -0.0284, 0.0363, -0.0213], + ..., + [-0.0258, -0.0417, 0.0408, ..., 0.0451, -0.0648, 0.0059], + [ 0.0271, 0.0224, -0.0943, ..., -0.0143, 0.0619, -0.0111], + [-0.0341, -0.0345, 0.0317, ..., 0.0010, -0.0178, -0.0388]], + device='cuda:0'), grad: tensor([[ 1.5482e-05, 7.7114e-06, -1.5652e-04, ..., 1.7812e-08, + -4.4912e-05, -1.6475e-04], + [ 5.2631e-05, -8.8513e-05, 4.1813e-05, ..., 1.4366e-07, + 9.2804e-05, 8.3540e-07], + [-2.9564e-04, -3.4422e-05, 1.4029e-03, ..., -1.2508e-06, + -6.2990e-04, 1.0484e-04], + ..., + [ 1.7181e-05, 3.9250e-05, -2.1706e-03, ..., 3.7835e-08, + 2.4110e-05, 5.1185e-06], + [ 3.2139e-04, 1.0926e-04, 2.4483e-05, ..., 5.5926e-07, + 5.3310e-04, 1.9725e-06], + [ 3.0041e-05, 1.4913e-04, 4.8113e-04, ..., 1.0943e-08, + 2.5153e-05, 3.5644e-05]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0082, 0.0206, -0.0153, 0.0130, 0.0247, -0.0200, -0.0298, -0.0185, + -0.0172, -0.0238], device='cuda:0'), grad: tensor([-0.0008, 0.0001, 0.0018, 0.0018, -0.0001, -0.0013, 0.0001, -0.0033, + 0.0008, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 217.05, cls_loss 0.0170 cls_loss_mapping 0.0290 cls_loss_causal 0.7540 re_mapping 0.0177 re_causal 0.0537 /// teacc 98.70 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0488, -0.0321, -0.0428, ..., -0.0228, -0.0261, -0.0160], + [ 0.0006, 0.0416, -0.0419, ..., 0.0014, -0.0013, -0.0548], + [ 0.0175, -0.0523, -0.0383, ..., -0.0289, 0.0368, -0.0218], + ..., + [-0.0274, -0.0420, 0.0416, ..., 0.0451, -0.0661, 0.0058], + [ 0.0276, 0.0223, -0.0957, ..., -0.0144, 0.0622, -0.0113], + [-0.0347, -0.0351, 0.0318, ..., 0.0010, -0.0188, -0.0391]], + device='cuda:0'), grad: tensor([[ 1.3709e-05, 1.0975e-05, 9.2313e-06, ..., 0.0000e+00, + 1.7554e-05, 0.0000e+00], + [ 6.9141e-06, -2.0936e-05, 2.3752e-05, ..., 0.0000e+00, + 1.2964e-06, 0.0000e+00], + [-3.7104e-05, -4.4666e-06, 2.6271e-05, ..., 0.0000e+00, + -4.8518e-05, 0.0000e+00], + ..., + [ 3.4887e-06, 1.7449e-05, -7.4053e-04, ..., 0.0000e+00, + 8.9407e-06, 0.0000e+00], + [-1.0234e-04, 1.0461e-05, 4.0591e-05, ..., 0.0000e+00, + -1.0574e-04, 0.0000e+00], + [ 1.6510e-05, 2.1726e-05, 1.6165e-04, ..., 0.0000e+00, + 3.0994e-05, 0.0000e+00]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0082, 0.0205, -0.0152, 0.0135, 0.0244, -0.0200, -0.0296, -0.0183, + -0.0174, -0.0240], device='cuda:0'), grad: tensor([-2.6375e-05, 2.2396e-05, 1.0192e-05, 5.4884e-04, -1.2541e-04, + 1.0878e-04, 1.0604e-04, -8.3256e-04, -5.7995e-05, 2.4557e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 216.72, cls_loss 0.0158 cls_loss_mapping 0.0238 cls_loss_causal 0.7398 re_mapping 0.0171 re_causal 0.0509 /// teacc 98.69 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0501, -0.0331, -0.0433, ..., -0.0230, -0.0266, -0.0160], + [ 0.0002, 0.0416, -0.0427, ..., 0.0024, -0.0016, -0.0557], + [ 0.0179, -0.0522, -0.0387, ..., -0.0289, 0.0378, -0.0221], + ..., + [-0.0281, -0.0411, 0.0424, ..., 0.0445, -0.0670, 0.0061], + [ 0.0278, 0.0221, -0.0970, ..., -0.0158, 0.0626, -0.0102], + [-0.0355, -0.0357, 0.0322, ..., 0.0014, -0.0195, -0.0393]], + device='cuda:0'), grad: tensor([[ 3.1233e-05, -9.2793e-04, 1.4037e-05, ..., 9.5461e-07, + -1.3723e-03, 0.0000e+00], + [ 4.2319e-04, 7.1287e-04, 8.9288e-05, ..., 5.5172e-06, + 1.4706e-03, 0.0000e+00], + [ 1.3769e-04, 4.1652e-04, 1.7333e-04, ..., 1.0110e-05, + 7.0572e-04, 0.0000e+00], + ..., + [ 1.5870e-05, 4.4137e-05, -1.8568e-03, ..., -1.4913e-04, + 3.7342e-05, 0.0000e+00], + [-9.9373e-04, -1.2207e-03, 2.5749e-05, ..., 4.9546e-07, + -2.7885e-03, 0.0000e+00], + [ 7.2777e-05, 8.7917e-05, 3.3045e-04, ..., 3.6538e-05, + 1.7703e-04, 0.0000e+00]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0085, 0.0200, -0.0149, 0.0129, 0.0244, -0.0198, -0.0295, -0.0179, + -0.0175, -0.0239], device='cuda:0'), grad: tensor([-3.2959e-03, 2.3766e-03, 1.4629e-03, 8.2779e-04, 1.0309e-03, + 2.1877e-03, -5.3287e-05, -1.5945e-03, -3.4943e-03, 5.5218e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 217.75, cls_loss 0.0129 cls_loss_mapping 0.0267 cls_loss_causal 0.7035 re_mapping 0.0168 re_causal 0.0497 /// teacc 98.82 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0507, -0.0332, -0.0435, ..., -0.0229, -0.0269, -0.0158], + [-0.0001, 0.0424, -0.0433, ..., 0.0027, -0.0013, -0.0559], + [ 0.0180, -0.0534, -0.0397, ..., -0.0291, 0.0381, -0.0221], + ..., + [-0.0288, -0.0412, 0.0437, ..., 0.0445, -0.0676, 0.0061], + [ 0.0282, 0.0220, -0.0970, ..., -0.0164, 0.0633, -0.0103], + [-0.0363, -0.0360, 0.0316, ..., 0.0013, -0.0206, -0.0396]], + device='cuda:0'), grad: tensor([[ 7.3671e-05, 6.8188e-05, 6.7428e-06, ..., 0.0000e+00, + 1.8403e-05, 2.6543e-08], + [ 1.7360e-05, -8.6308e-05, 2.2143e-05, ..., 0.0000e+00, + -1.4253e-05, 9.9419e-08], + [ 4.1664e-05, 2.5943e-05, 4.7863e-05, ..., 0.0000e+00, + -2.3633e-05, -5.3039e-07], + ..., + [-3.4213e-05, 1.9282e-05, -7.7710e-06, ..., 0.0000e+00, + 5.3614e-05, 1.4133e-07], + [-1.6177e-04, 5.7995e-05, 7.2241e-05, ..., 0.0000e+00, + -1.8668e-04, 4.1444e-08], + [ 1.5080e-04, 2.4214e-05, 9.8825e-05, ..., 0.0000e+00, + 1.2183e-04, 1.3504e-08]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0076, 0.0201, -0.0153, 0.0125, 0.0245, -0.0199, -0.0298, -0.0173, + -0.0173, -0.0244], device='cuda:0'), grad: tensor([-5.3868e-06, -2.5593e-06, 1.1730e-04, 3.2043e-04, -7.7295e-04, + -2.8896e-04, -1.8227e-04, 1.8287e-04, 4.8876e-05, 5.8365e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 216.77, cls_loss 0.0118 cls_loss_mapping 0.0237 cls_loss_causal 0.7285 re_mapping 0.0160 re_causal 0.0494 /// teacc 98.82 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0518, -0.0339, -0.0438, ..., -0.0229, -0.0274, -0.0155], + [-0.0005, 0.0432, -0.0430, ..., 0.0032, -0.0014, -0.0563], + [ 0.0189, -0.0541, -0.0400, ..., -0.0296, 0.0390, -0.0220], + ..., + [-0.0299, -0.0418, 0.0439, ..., 0.0445, -0.0687, 0.0061], + [ 0.0284, 0.0221, -0.0975, ..., -0.0168, 0.0634, -0.0104], + [-0.0369, -0.0365, 0.0320, ..., 0.0013, -0.0213, -0.0398]], + device='cuda:0'), grad: tensor([[ 1.4400e-04, 6.9261e-05, 1.0192e-05, ..., 0.0000e+00, + 6.3062e-05, 0.0000e+00], + [ 3.2139e-04, -4.0841e-04, -7.6711e-05, ..., 0.0000e+00, + 1.9062e-04, 0.0000e+00], + [-1.4696e-03, -7.6354e-05, 3.0279e-05, ..., 0.0000e+00, + -1.3056e-03, 0.0000e+00], + ..., + [ 8.0109e-04, 1.3995e-04, 6.5267e-05, ..., 0.0000e+00, + 5.9462e-04, 0.0000e+00], + [ 6.5756e-04, 6.0749e-04, -5.5879e-05, ..., 0.0000e+00, + 2.8658e-04, 0.0000e+00], + [ 2.6131e-04, 5.5343e-05, 1.6665e-04, ..., 0.0000e+00, + 9.4295e-05, 0.0000e+00]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0076, 0.0204, -0.0150, 0.0122, 0.0244, -0.0198, -0.0298, -0.0178, + -0.0170, -0.0243], device='cuda:0'), grad: tensor([ 2.5892e-04, 7.7188e-05, -4.1885e-03, 4.3464e-04, 2.6822e-04, + 5.4300e-05, -1.2312e-03, 2.7561e-03, 9.0837e-04, 6.6137e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 216.84, cls_loss 0.0171 cls_loss_mapping 0.0298 cls_loss_causal 0.7264 re_mapping 0.0164 re_causal 0.0479 /// teacc 98.73 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0527, -0.0353, -0.0439, ..., -0.0230, -0.0278, -0.0149], + [-0.0015, 0.0435, -0.0434, ..., 0.0035, -0.0016, -0.0565], + [ 0.0190, -0.0548, -0.0408, ..., -0.0298, 0.0394, -0.0229], + ..., + [-0.0304, -0.0422, 0.0436, ..., 0.0445, -0.0693, 0.0060], + [ 0.0287, 0.0223, -0.0986, ..., -0.0171, 0.0639, -0.0106], + [-0.0378, -0.0370, 0.0326, ..., 0.0012, -0.0225, -0.0405]], + device='cuda:0'), grad: tensor([[ 3.9399e-05, 2.5883e-05, 9.5516e-06, ..., 0.0000e+00, + 5.2482e-05, 2.6152e-06], + [ 1.0096e-05, -4.0245e-04, 1.4938e-05, ..., 0.0000e+00, + -1.4699e-04, 1.3923e-07], + [ 1.5944e-05, 3.8028e-05, 8.8215e-06, ..., 0.0000e+00, + -5.3316e-05, -4.3921e-06], + ..., + [ 1.4544e-05, 8.9169e-05, -9.3043e-05, ..., 0.0000e+00, + 4.6998e-05, 2.7474e-07], + [ 1.2147e-04, 1.8907e-04, 7.3135e-05, ..., 0.0000e+00, + 1.3685e-04, 9.3365e-08], + [ 7.0333e-05, 8.4817e-05, 1.2517e-04, ..., 0.0000e+00, + -6.7830e-05, 2.3865e-07]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0081, 0.0200, -0.0155, 0.0134, 0.0246, -0.0203, -0.0295, -0.0182, + -0.0170, -0.0239], device='cuda:0'), grad: tensor([ 2.9802e-04, -4.4823e-04, -5.4032e-05, 7.4482e-04, 1.8165e-05, + -1.6165e-03, 6.2346e-05, 5.0455e-05, 5.9748e-04, 3.4690e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 43---------------------------------------------------- +epoch 43, time 217.39, cls_loss 0.0128 cls_loss_mapping 0.0233 cls_loss_causal 0.7277 re_mapping 0.0159 re_causal 0.0498 /// teacc 98.89 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0536, -0.0360, -0.0442, ..., -0.0230, -0.0279, -0.0147], + [-0.0019, 0.0438, -0.0436, ..., 0.0036, -0.0019, -0.0566], + [ 0.0188, -0.0553, -0.0410, ..., -0.0295, 0.0398, -0.0230], + ..., + [-0.0308, -0.0425, 0.0444, ..., 0.0445, -0.0700, 0.0060], + [ 0.0293, 0.0225, -0.0993, ..., -0.0172, 0.0644, -0.0106], + [-0.0386, -0.0376, 0.0328, ..., 0.0012, -0.0231, -0.0406]], + device='cuda:0'), grad: tensor([[ 1.9401e-05, 8.9267e-07, 5.5470e-06, ..., 0.0000e+00, + 2.0072e-05, -3.4515e-06], + [ 8.3297e-06, -5.6803e-05, 4.5002e-05, ..., 0.0000e+00, + 1.1005e-05, 3.4925e-08], + [ 1.8311e-04, 1.4164e-05, 4.6349e-04, ..., 0.0000e+00, + 1.9908e-04, -9.5181e-07], + ..., + [ 2.5010e-04, 5.3525e-05, -4.1080e-04, ..., 0.0000e+00, + 3.4857e-04, 1.2200e-07], + [-6.6614e-04, 4.1783e-05, -1.5426e-04, ..., 0.0000e+00, + -9.3794e-04, 1.3318e-07], + [ 1.3566e-04, 3.7241e-04, 6.0701e-04, ..., 0.0000e+00, + 1.4281e-04, 5.6904e-07]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0078, 0.0199, -0.0155, 0.0132, 0.0243, -0.0197, -0.0301, -0.0181, + -0.0169, -0.0239], device='cuda:0'), grad: tensor([-9.1553e-05, 1.9103e-05, 9.1362e-04, -4.5586e-04, -1.0185e-03, + 4.8351e-04, 1.9753e-04, -7.5817e-05, -1.2541e-03, 1.2808e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 216.91, cls_loss 0.0122 cls_loss_mapping 0.0206 cls_loss_causal 0.7150 re_mapping 0.0157 re_causal 0.0477 /// teacc 98.83 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0543, -0.0360, -0.0445, ..., -0.0230, -0.0279, -0.0147], + [-0.0021, 0.0449, -0.0438, ..., 0.0040, -0.0015, -0.0568], + [ 0.0188, -0.0562, -0.0413, ..., -0.0298, 0.0400, -0.0228], + ..., + [-0.0314, -0.0428, 0.0452, ..., 0.0445, -0.0705, 0.0060], + [ 0.0301, 0.0224, -0.1000, ..., -0.0173, 0.0650, -0.0107], + [-0.0392, -0.0385, 0.0322, ..., 0.0012, -0.0238, -0.0406]], + device='cuda:0'), grad: tensor([[ 1.7196e-05, 1.8030e-05, 5.1880e-04, ..., 1.0943e-08, + 5.1069e-04, 4.8894e-08], + [-4.8566e-04, -1.7977e-03, -5.1212e-04, ..., 1.2596e-07, + 3.2812e-05, 8.5216e-08], + [ 4.8429e-05, 3.4785e-04, 1.7488e-04, ..., 1.4063e-07, + 3.2410e-06, -5.1688e-07], + ..., + [ 1.1307e-04, 1.9014e-04, -1.1188e-04, ..., 2.2305e-07, + 5.1826e-05, 7.8930e-08], + [-1.0413e-04, 3.8654e-05, 2.2161e-04, ..., 7.5391e-07, + -1.0699e-04, 7.2876e-08], + [ 2.8685e-05, 2.4304e-05, -1.0900e-03, ..., 3.9581e-08, + -8.0776e-04, 7.4506e-09]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0074, 0.0207, -0.0159, 0.0127, 0.0248, -0.0196, -0.0301, -0.0177, + -0.0169, -0.0247], device='cuda:0'), grad: tensor([ 4.9706e-03, -2.0428e-03, 9.2220e-04, 1.9741e-03, 1.5917e-03, + 4.2748e-04, 1.5056e-04, 3.6031e-05, 3.1996e-04, -8.3466e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 216.82, cls_loss 0.0127 cls_loss_mapping 0.0206 cls_loss_causal 0.7091 re_mapping 0.0147 re_causal 0.0455 /// teacc 98.68 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0553, -0.0372, -0.0443, ..., -0.0230, -0.0281, -0.0146], + [-0.0025, 0.0449, -0.0438, ..., 0.0046, -0.0021, -0.0569], + [ 0.0189, -0.0570, -0.0412, ..., -0.0301, 0.0404, -0.0229], + ..., + [-0.0323, -0.0418, 0.0454, ..., 0.0445, -0.0708, 0.0060], + [ 0.0299, 0.0221, -0.1008, ..., -0.0175, 0.0652, -0.0107], + [-0.0399, -0.0390, 0.0325, ..., 0.0011, -0.0244, -0.0407]], + device='cuda:0'), grad: tensor([[ 3.6693e-04, 3.3164e-04, 2.6263e-06, ..., 0.0000e+00, + 1.7375e-05, -2.5883e-05], + [ 8.4400e-05, -2.9773e-05, 8.9873e-07, ..., 0.0000e+00, + 5.3585e-05, 1.7951e-07], + [ 2.7323e-04, 1.2422e-04, 1.3900e-04, ..., 0.0000e+00, + 2.1696e-04, 9.8534e-07], + ..., + [ 5.0962e-05, 2.7791e-05, -1.9836e-04, ..., 0.0000e+00, + -3.5524e-05, 3.4319e-07], + [-9.4557e-04, -1.5879e-04, 2.8133e-05, ..., 0.0000e+00, + -8.8835e-04, 1.5749e-06], + [ 1.9383e-04, 1.6823e-05, -1.5616e-05, ..., 0.0000e+00, + 1.7154e-04, 7.1973e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0072, 0.0205, -0.0158, 0.0128, 0.0245, -0.0196, -0.0296, -0.0175, + -0.0174, -0.0247], device='cuda:0'), grad: tensor([ 0.0006, 0.0001, 0.0015, 0.0009, 0.0002, 0.0006, -0.0012, -0.0014, + -0.0016, 0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 216.77, cls_loss 0.0095 cls_loss_mapping 0.0189 cls_loss_causal 0.6753 re_mapping 0.0146 re_causal 0.0464 /// teacc 98.66 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0558, -0.0381, -0.0445, ..., -0.0230, -0.0282, -0.0146], + [-0.0028, 0.0455, -0.0431, ..., 0.0046, -0.0024, -0.0570], + [ 0.0193, -0.0575, -0.0415, ..., -0.0299, 0.0409, -0.0229], + ..., + [-0.0328, -0.0428, 0.0453, ..., 0.0445, -0.0714, 0.0060], + [ 0.0306, 0.0223, -0.1019, ..., -0.0177, 0.0656, -0.0107], + [-0.0404, -0.0393, 0.0323, ..., 0.0011, -0.0245, -0.0407]], + device='cuda:0'), grad: tensor([[ 1.7032e-05, 9.4771e-06, 1.9997e-05, ..., 0.0000e+00, + 5.8487e-06, 0.0000e+00], + [ 1.2785e-05, -1.0353e-04, 1.4983e-05, ..., 0.0000e+00, + -2.0191e-05, 0.0000e+00], + [-1.2092e-05, 1.7658e-05, 2.2054e-05, ..., 0.0000e+00, + -2.7597e-05, 0.0000e+00], + ..., + [ 7.8455e-06, 1.3389e-05, -1.9336e-04, ..., 0.0000e+00, + 1.1347e-05, 0.0000e+00], + [-6.4909e-05, 3.9250e-05, 1.3687e-05, ..., 0.0000e+00, + -5.4687e-05, 0.0000e+00], + [ 1.9252e-05, 1.7479e-05, 9.8050e-05, ..., 0.0000e+00, + 1.4648e-05, 0.0000e+00]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0072, 0.0206, -0.0158, 0.0127, 0.0248, -0.0197, -0.0296, -0.0178, + -0.0172, -0.0247], device='cuda:0'), grad: tensor([ 7.9393e-05, -4.5925e-05, 3.7372e-05, 2.1195e-04, -9.4712e-05, + -9.1374e-05, 6.2764e-05, -6.0081e-04, 1.4536e-05, 4.2653e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 216.61, cls_loss 0.0122 cls_loss_mapping 0.0252 cls_loss_causal 0.6838 re_mapping 0.0141 re_causal 0.0438 /// teacc 98.75 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0565, -0.0384, -0.0447, ..., -0.0230, -0.0279, -0.0146], + [-0.0033, 0.0460, -0.0436, ..., 0.0049, -0.0028, -0.0570], + [ 0.0197, -0.0583, -0.0416, ..., -0.0301, 0.0415, -0.0228], + ..., + [-0.0335, -0.0430, 0.0455, ..., 0.0445, -0.0723, 0.0060], + [ 0.0302, 0.0213, -0.1030, ..., -0.0179, 0.0659, -0.0107], + [-0.0406, -0.0401, 0.0323, ..., 0.0011, -0.0253, -0.0408]], + device='cuda:0'), grad: tensor([[ 1.2219e-05, -3.6508e-05, 4.6454e-06, ..., 0.0000e+00, + -1.8356e-06, 0.0000e+00], + [ 3.9935e-05, 3.1620e-05, 4.8488e-05, ..., 0.0000e+00, + 1.0318e-04, 0.0000e+00], + [-6.9761e-04, 2.6330e-05, -3.7146e-04, ..., 0.0000e+00, + -1.7710e-03, 0.0000e+00], + ..., + [ 5.7030e-04, 3.7998e-05, 3.1543e-04, ..., 0.0000e+00, + 1.4067e-03, 0.0000e+00], + [ 2.0683e-05, 3.2485e-05, 4.6730e-05, ..., 0.0000e+00, + 7.7665e-05, 0.0000e+00], + [ 1.9208e-05, 4.3154e-05, -1.5244e-05, ..., 0.0000e+00, + 2.8476e-05, 0.0000e+00]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0070, 0.0203, -0.0157, 0.0125, 0.0251, -0.0198, -0.0289, -0.0181, + -0.0175, -0.0246], device='cuda:0'), grad: tensor([-5.8842e-04, 2.1172e-04, -1.9016e-03, 6.6817e-05, -1.7595e-04, + 2.0444e-04, 3.1829e-04, 1.6298e-03, 1.4377e-04, 9.3102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 216.92, cls_loss 0.0110 cls_loss_mapping 0.0209 cls_loss_causal 0.6709 re_mapping 0.0141 re_causal 0.0424 /// teacc 98.86 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0576, -0.0389, -0.0451, ..., -0.0230, -0.0284, -0.0146], + [-0.0035, 0.0464, -0.0439, ..., 0.0049, -0.0023, -0.0571], + [ 0.0198, -0.0588, -0.0420, ..., -0.0300, 0.0417, -0.0227], + ..., + [-0.0343, -0.0432, 0.0463, ..., 0.0445, -0.0732, 0.0059], + [ 0.0306, 0.0215, -0.1037, ..., -0.0181, 0.0663, -0.0108], + [-0.0414, -0.0406, 0.0323, ..., 0.0011, -0.0259, -0.0408]], + device='cuda:0'), grad: tensor([[ 3.2753e-05, 2.8625e-05, 1.2077e-05, ..., 4.7032e-08, + 1.8120e-05, 0.0000e+00], + [ 1.1164e-04, -7.6711e-05, 5.2094e-05, ..., 1.8626e-08, + 1.3363e-04, 0.0000e+00], + [-1.5223e-04, 3.3796e-05, -3.7774e-06, ..., 7.9162e-09, + -3.4451e-04, 0.0000e+00], + ..., + [ 4.3869e-05, 2.7403e-05, -3.6836e-05, ..., 2.1886e-08, + 4.5538e-05, 0.0000e+00], + [-1.5676e-05, 9.9987e-06, 3.3021e-05, ..., 3.1246e-07, + 1.8209e-05, 0.0000e+00], + [ 8.4937e-05, 2.1368e-05, 8.7678e-05, ..., 4.9826e-08, + 6.1691e-06, 0.0000e+00]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0077, 0.0205, -0.0159, 0.0125, 0.0251, -0.0199, -0.0288, -0.0180, + -0.0175, -0.0242], device='cuda:0'), grad: tensor([ 4.9561e-05, 2.3270e-04, -3.1900e-04, 3.1948e-04, 4.4219e-06, + -7.5626e-04, 9.2685e-06, 5.5343e-05, 9.2685e-05, 3.1161e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 216.99, cls_loss 0.0120 cls_loss_mapping 0.0198 cls_loss_causal 0.7202 re_mapping 0.0146 re_causal 0.0441 /// teacc 98.77 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0586, -0.0399, -0.0441, ..., -0.0228, -0.0277, -0.0143], + [-0.0043, 0.0462, -0.0440, ..., 0.0058, -0.0034, -0.0577], + [ 0.0205, -0.0580, -0.0413, ..., -0.0304, 0.0432, -0.0225], + ..., + [-0.0345, -0.0436, 0.0466, ..., 0.0445, -0.0735, 0.0059], + [ 0.0304, 0.0211, -0.1051, ..., -0.0186, 0.0663, -0.0108], + [-0.0421, -0.0411, 0.0324, ..., 0.0009, -0.0266, -0.0411]], + device='cuda:0'), grad: tensor([[ 2.7761e-05, 2.4393e-05, 6.2473e-06, ..., 2.2817e-08, + 4.9442e-05, 0.0000e+00], + [ 1.1951e-05, -7.9334e-05, 1.2927e-05, ..., 5.6345e-08, + -5.0634e-05, 0.0000e+00], + [ 6.5506e-05, 9.2566e-05, 1.8358e-05, ..., 3.7160e-07, + 1.7726e-04, 0.0000e+00], + ..., + [ 8.0764e-06, 5.3734e-05, 8.8289e-06, ..., 7.5903e-08, + 4.0889e-05, 0.0000e+00], + [-3.9101e-05, 6.0380e-05, 6.9082e-05, ..., 3.1525e-07, + 3.5651e-06, 0.0000e+00], + [ 1.0870e-05, 2.8920e-04, 1.6224e-04, ..., 7.8231e-08, + 2.7850e-05, 0.0000e+00]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0070, 0.0197, -0.0146, 0.0121, 0.0247, -0.0195, -0.0288, -0.0181, + -0.0180, -0.0244], device='cuda:0'), grad: tensor([-2.6539e-05, -2.9057e-05, 2.8825e-04, -4.5896e-04, -1.0557e-03, + 1.2290e-04, 1.3435e-04, 1.3852e-04, 2.4557e-04, 6.3992e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 216.91, cls_loss 0.0119 cls_loss_mapping 0.0214 cls_loss_causal 0.7097 re_mapping 0.0144 re_causal 0.0444 /// teacc 98.79 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0593, -0.0405, -0.0445, ..., -0.0228, -0.0281, -0.0141], + [-0.0044, 0.0465, -0.0442, ..., 0.0058, -0.0033, -0.0577], + [ 0.0203, -0.0585, -0.0424, ..., -0.0304, 0.0432, -0.0225], + ..., + [-0.0350, -0.0430, 0.0468, ..., 0.0445, -0.0739, 0.0059], + [ 0.0308, 0.0210, -0.1059, ..., -0.0188, 0.0670, -0.0108], + [-0.0429, -0.0415, 0.0327, ..., 0.0009, -0.0275, -0.0412]], + device='cuda:0'), grad: tensor([[ 6.7770e-05, 2.3097e-05, 3.5912e-05, ..., 0.0000e+00, + 7.9036e-05, 0.0000e+00], + [ 9.9018e-06, -3.3110e-05, 3.0786e-05, ..., 0.0000e+00, + 1.2286e-05, 0.0000e+00], + [ 2.3317e-04, 2.3842e-05, 1.4722e-04, ..., 0.0000e+00, + 3.4261e-04, 0.0000e+00], + ..., + [ 1.6227e-05, 1.9759e-05, -3.0160e-04, ..., 0.0000e+00, + 2.4140e-05, 0.0000e+00], + [-4.4823e-04, 5.4359e-05, -3.4213e-05, ..., 0.0000e+00, + -6.4659e-04, 0.0000e+00], + [ 4.1485e-05, 1.1533e-04, 2.6083e-04, ..., 0.0000e+00, + 6.1154e-05, 0.0000e+00]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0071, 0.0196, -0.0155, 0.0123, 0.0252, -0.0190, -0.0293, -0.0178, + -0.0179, -0.0244], device='cuda:0'), grad: tensor([ 7.4208e-05, 3.3498e-05, 4.8661e-04, 2.2662e-04, -8.6594e-04, + 1.2898e-04, 6.1572e-05, -3.6430e-04, -4.3559e-04, 6.5517e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 51---------------------------------------------------- +epoch 51, time 217.76, cls_loss 0.0101 cls_loss_mapping 0.0179 cls_loss_causal 0.6756 re_mapping 0.0136 re_causal 0.0417 /// teacc 98.90 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0600, -0.0411, -0.0447, ..., -0.0228, -0.0287, -0.0141], + [-0.0049, 0.0470, -0.0444, ..., 0.0053, -0.0033, -0.0577], + [ 0.0202, -0.0591, -0.0417, ..., -0.0307, 0.0432, -0.0225], + ..., + [-0.0362, -0.0430, 0.0475, ..., 0.0449, -0.0746, 0.0059], + [ 0.0318, 0.0209, -0.1070, ..., -0.0195, 0.0685, -0.0109], + [-0.0439, -0.0416, 0.0327, ..., 0.0008, -0.0285, -0.0412]], + device='cuda:0'), grad: tensor([[ 1.7032e-05, 2.2218e-05, 4.6119e-06, ..., 0.0000e+00, + -7.9334e-05, -2.9579e-05], + [ 4.3884e-06, -3.2759e-04, 1.1690e-05, ..., 0.0000e+00, + -9.0897e-05, 8.8010e-08], + [ 1.6063e-05, 1.9276e-04, 6.2704e-05, ..., 0.0000e+00, + 1.2040e-04, 1.1232e-06], + ..., + [ 3.0287e-06, 4.6819e-05, -6.6900e-04, ..., 0.0000e+00, + -1.3828e-04, 4.4331e-07], + [ 1.0375e-06, 5.2541e-05, 3.2753e-05, ..., 0.0000e+00, + 1.8805e-05, 3.6182e-07], + [ 1.9982e-05, 1.6183e-05, 2.9492e-04, ..., 0.0000e+00, + 1.1331e-04, 2.2128e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0074, 0.0197, -0.0157, 0.0118, 0.0249, -0.0192, -0.0291, -0.0175, + -0.0172, -0.0244], device='cuda:0'), grad: tensor([-3.9315e-04, -3.1614e-04, 4.8065e-04, 3.2735e-04, -8.8811e-06, + 3.5143e-04, -4.7922e-05, -1.5659e-03, 1.8799e-04, 9.8324e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 216.84, cls_loss 0.0109 cls_loss_mapping 0.0187 cls_loss_causal 0.6718 re_mapping 0.0132 re_causal 0.0399 /// teacc 98.80 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0608, -0.0415, -0.0447, ..., -0.0228, -0.0284, -0.0141], + [-0.0054, 0.0482, -0.0453, ..., 0.0055, -0.0038, -0.0578], + [ 0.0207, -0.0595, -0.0429, ..., -0.0309, 0.0439, -0.0225], + ..., + [-0.0367, -0.0441, 0.0487, ..., 0.0449, -0.0751, 0.0059], + [ 0.0317, 0.0207, -0.1079, ..., -0.0197, 0.0685, -0.0109], + [-0.0446, -0.0421, 0.0335, ..., 0.0008, -0.0302, -0.0413]], + device='cuda:0'), grad: tensor([[ 3.4124e-05, 3.8356e-05, 9.6411e-06, ..., 0.0000e+00, + 1.2636e-05, 0.0000e+00], + [ 2.7016e-05, 1.8148e-03, 1.1883e-03, ..., 0.0000e+00, + 5.9187e-05, 0.0000e+00], + [-6.5982e-05, 5.7697e-05, 2.8700e-05, ..., 0.0000e+00, + -1.9908e-04, 0.0000e+00], + ..., + [ 2.2173e-05, 2.1994e-04, 1.5020e-04, ..., 0.0000e+00, + 4.8906e-05, 0.0000e+00], + [ 2.9191e-05, 1.1081e-04, 7.9036e-05, ..., 0.0000e+00, + 1.8373e-05, 0.0000e+00], + [ 4.8019e-06, 1.2932e-03, 7.6008e-04, ..., 0.0000e+00, + 2.4214e-06, 0.0000e+00]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0069, 0.0200, -0.0159, 0.0121, 0.0249, -0.0197, -0.0288, -0.0170, + -0.0180, -0.0246], device='cuda:0'), grad: tensor([ 5.4419e-05, 2.7065e-03, -1.4853e-04, 2.2197e-04, -5.0354e-03, + 3.5310e-04, -4.7421e-04, 4.1080e-04, 2.3520e-04, 1.6747e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 53---------------------------------------------------- +epoch 53, time 217.76, cls_loss 0.0084 cls_loss_mapping 0.0160 cls_loss_causal 0.6313 re_mapping 0.0131 re_causal 0.0408 /// teacc 98.95 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0615, -0.0423, -0.0449, ..., -0.0228, -0.0287, -0.0140], + [-0.0056, 0.0484, -0.0455, ..., 0.0055, -0.0041, -0.0578], + [ 0.0211, -0.0591, -0.0423, ..., -0.0309, 0.0450, -0.0224], + ..., + [-0.0375, -0.0445, 0.0494, ..., 0.0449, -0.0758, 0.0059], + [ 0.0317, 0.0204, -0.1084, ..., -0.0197, 0.0685, -0.0109], + [-0.0452, -0.0426, 0.0330, ..., 0.0008, -0.0305, -0.0413]], + device='cuda:0'), grad: tensor([[ 7.6443e-06, 7.0818e-06, 1.0841e-06, ..., 0.0000e+00, + 6.7130e-06, 0.0000e+00], + [ 1.5068e-04, 1.0532e-04, -4.2580e-06, ..., 0.0000e+00, + 2.6870e-04, 0.0000e+00], + [-1.7405e-04, -1.2147e-04, 1.0476e-05, ..., 0.0000e+00, + -1.6260e-04, 0.0000e+00], + ..., + [ 5.6364e-06, 2.6256e-05, -1.4782e-05, ..., 0.0000e+00, + 7.6592e-06, 0.0000e+00], + [ 2.0891e-05, 3.7372e-05, 7.2084e-06, ..., 0.0000e+00, + 5.9634e-05, 0.0000e+00], + [ 3.4571e-05, 3.8981e-05, -1.6794e-05, ..., 0.0000e+00, + 7.4625e-05, 0.0000e+00]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0072, 0.0197, -0.0151, 0.0120, 0.0246, -0.0193, -0.0286, -0.0167, + -0.0183, -0.0249], device='cuda:0'), grad: tensor([ 7.2941e-06, 4.8208e-04, -3.1447e-04, 1.4830e-03, 1.0975e-05, + -2.2202e-03, 2.6274e-04, 8.6799e-06, 1.4079e-04, 1.3888e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 216.92, cls_loss 0.0096 cls_loss_mapping 0.0220 cls_loss_causal 0.6747 re_mapping 0.0131 re_causal 0.0415 /// teacc 98.72 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0618, -0.0426, -0.0451, ..., -0.0228, -0.0286, -0.0140], + [-0.0055, 0.0491, -0.0451, ..., 0.0055, -0.0036, -0.0579], + [ 0.0214, -0.0595, -0.0428, ..., -0.0309, 0.0452, -0.0224], + ..., + [-0.0383, -0.0452, 0.0497, ..., 0.0449, -0.0759, 0.0059], + [ 0.0320, 0.0203, -0.1086, ..., -0.0197, 0.0687, -0.0109], + [-0.0459, -0.0429, 0.0330, ..., 0.0008, -0.0311, -0.0413]], + device='cuda:0'), grad: tensor([[ 3.3200e-05, 7.3537e-06, 1.8820e-05, ..., 0.0000e+00, + 6.5923e-05, 1.6734e-05], + [-1.6069e-04, -1.3790e-03, -1.2481e-04, ..., 0.0000e+00, + -4.6587e-04, 4.5374e-06], + [ 4.8220e-05, 7.6342e-04, 3.0780e-04, ..., 0.0000e+00, + 2.0134e-04, -5.9575e-05], + ..., + [ 5.8264e-05, 3.4475e-04, -3.4523e-03, ..., 0.0000e+00, + 1.3494e-04, 2.6852e-05], + [-2.8923e-05, 5.4866e-05, 4.2588e-05, ..., 0.0000e+00, + -1.1176e-05, 2.8200e-06], + [ 2.0310e-05, 8.6069e-05, 5.9098e-05, ..., 0.0000e+00, + 2.1845e-05, 6.9058e-07]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0069, 0.0201, -0.0152, 0.0122, 0.0244, -0.0194, -0.0287, -0.0171, + -0.0181, -0.0251], device='cuda:0'), grad: tensor([ 1.0216e-04, -1.8520e-03, 1.0376e-03, 3.7998e-07, 3.9368e-03, + 1.0699e-04, 8.7917e-05, -3.7346e-03, 1.1903e-04, 1.9085e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 217.18, cls_loss 0.0060 cls_loss_mapping 0.0137 cls_loss_causal 0.6042 re_mapping 0.0129 re_causal 0.0387 /// teacc 98.75 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0624, -0.0430, -0.0452, ..., -0.0229, -0.0284, -0.0140], + [-0.0060, 0.0494, -0.0454, ..., 0.0062, -0.0037, -0.0581], + [ 0.0215, -0.0601, -0.0431, ..., -0.0312, 0.0455, -0.0222], + ..., + [-0.0388, -0.0449, 0.0501, ..., 0.0448, -0.0764, 0.0058], + [ 0.0319, 0.0203, -0.1092, ..., -0.0198, 0.0686, -0.0109], + [-0.0465, -0.0433, 0.0330, ..., 0.0008, -0.0317, -0.0414]], + device='cuda:0'), grad: tensor([[ 2.5943e-05, 1.6615e-05, 7.5474e-06, ..., 4.6100e-08, + 2.5913e-05, 0.0000e+00], + [ 3.4004e-05, -2.2262e-05, 2.0757e-05, ..., 6.6124e-08, + 4.0323e-05, 0.0000e+00], + [ 2.2221e-04, 5.1171e-05, 2.1175e-05, ..., 9.4529e-08, + 2.3580e-04, 0.0000e+00], + ..., + [ 8.6874e-06, 1.9059e-05, -1.0073e-04, ..., 4.0513e-08, + 1.2100e-05, 0.0000e+00], + [-3.1638e-04, -5.6922e-05, 2.1711e-05, ..., 8.1956e-08, + -3.0661e-04, 0.0000e+00], + [ 7.4387e-05, 1.6725e-04, 4.3035e-04, ..., 3.5111e-07, + 9.3341e-05, 0.0000e+00]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0063, 0.0202, -0.0154, 0.0126, 0.0242, -0.0195, -0.0286, -0.0170, + -0.0185, -0.0252], device='cuda:0'), grad: tensor([ 8.5115e-05, 6.2943e-05, 3.0637e-04, -1.1072e-03, -9.1743e-04, + 5.0831e-04, 1.4102e-04, -8.3268e-05, -3.4302e-05, 1.0376e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 216.83, cls_loss 0.0111 cls_loss_mapping 0.0228 cls_loss_causal 0.6670 re_mapping 0.0134 re_causal 0.0381 /// teacc 98.82 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0635, -0.0442, -0.0455, ..., -0.0230, -0.0297, -0.0139], + [-0.0063, 0.0496, -0.0460, ..., 0.0061, -0.0038, -0.0582], + [ 0.0215, -0.0608, -0.0435, ..., -0.0326, 0.0464, -0.0222], + ..., + [-0.0405, -0.0448, 0.0509, ..., 0.0464, -0.0782, 0.0058], + [ 0.0333, 0.0207, -0.1093, ..., -0.0204, 0.0693, -0.0110], + [-0.0479, -0.0437, 0.0328, ..., 0.0006, -0.0334, -0.0414]], + device='cuda:0'), grad: tensor([[ 2.3648e-05, 2.1577e-05, 4.5836e-05, ..., 0.0000e+00, + 2.9042e-05, 0.0000e+00], + [ 1.0908e-04, -6.3062e-05, 2.9549e-05, ..., 0.0000e+00, + 1.9884e-04, 0.0000e+00], + [ 7.2670e-04, 3.6091e-05, 1.7571e-04, ..., 0.0000e+00, + 1.4248e-03, 0.0000e+00], + ..., + [-1.6260e-03, 4.3690e-05, -4.6229e-04, ..., 0.0000e+00, + -3.1624e-03, 0.0000e+00], + [ 8.8274e-05, 4.2617e-05, 2.2054e-05, ..., 0.0000e+00, + 7.4029e-05, 0.0000e+00], + [ 2.1607e-05, 1.6272e-05, 8.4490e-06, ..., 0.0000e+00, + 3.3677e-05, 0.0000e+00]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0069, 0.0198, -0.0152, 0.0130, 0.0244, -0.0203, -0.0286, -0.0169, + -0.0173, -0.0258], device='cuda:0'), grad: tensor([ 1.8150e-05, 2.8443e-04, 2.4815e-03, 2.1763e-03, -2.4581e-04, + 8.9049e-05, 2.7728e-04, -5.4398e-03, 2.5082e-04, 1.0628e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 217.02, cls_loss 0.0091 cls_loss_mapping 0.0187 cls_loss_causal 0.6815 re_mapping 0.0130 re_causal 0.0401 /// teacc 98.91 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0640, -0.0444, -0.0457, ..., -0.0228, -0.0297, -0.0139], + [-0.0066, 0.0501, -0.0459, ..., 0.0062, -0.0042, -0.0583], + [ 0.0214, -0.0612, -0.0437, ..., -0.0333, 0.0465, -0.0222], + ..., + [-0.0412, -0.0455, 0.0506, ..., 0.0466, -0.0789, 0.0058], + [ 0.0336, 0.0206, -0.1102, ..., -0.0206, 0.0701, -0.0110], + [-0.0487, -0.0444, 0.0332, ..., 0.0006, -0.0338, -0.0415]], + device='cuda:0'), grad: tensor([[ 4.3690e-05, 3.2961e-05, 4.8522e-07, ..., 0.0000e+00, + 7.0259e-06, 0.0000e+00], + [ 1.2375e-05, 9.4175e-06, 2.1964e-05, ..., 0.0000e+00, + 1.4499e-05, 0.0000e+00], + [ 7.0184e-06, 2.3365e-05, 4.8354e-06, ..., 0.0000e+00, + 5.9679e-06, 0.0000e+00], + ..., + [ 6.0797e-06, -1.8820e-05, -7.7069e-05, ..., 0.0000e+00, + 2.4244e-05, 0.0000e+00], + [ 1.9228e-04, 1.3506e-04, 5.6401e-06, ..., 0.0000e+00, + 3.8654e-05, 0.0000e+00], + [ 1.1630e-05, 1.2621e-05, 4.1157e-05, ..., 0.0000e+00, + 3.3319e-05, 0.0000e+00]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0075, 0.0201, -0.0154, 0.0134, 0.0246, -0.0206, -0.0286, -0.0176, + -0.0174, -0.0250], device='cuda:0'), grad: tensor([-1.7500e-04, 9.4771e-05, 9.6738e-05, -3.2043e-04, -6.0759e-06, + 4.0674e-04, -5.4121e-04, -7.5817e-05, 3.3069e-04, 1.9038e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 216.98, cls_loss 0.0112 cls_loss_mapping 0.0198 cls_loss_causal 0.6596 re_mapping 0.0122 re_causal 0.0374 /// teacc 98.69 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0646, -0.0434, -0.0459, ..., -0.0229, -0.0284, -0.0138], + [-0.0074, 0.0507, -0.0462, ..., 0.0060, -0.0046, -0.0583], + [ 0.0219, -0.0620, -0.0443, ..., -0.0337, 0.0470, -0.0223], + ..., + [-0.0420, -0.0462, 0.0511, ..., 0.0465, -0.0799, 0.0058], + [ 0.0343, 0.0201, -0.1107, ..., -0.0215, 0.0699, -0.0110], + [-0.0493, -0.0440, 0.0334, ..., 0.0005, -0.0346, -0.0415]], + device='cuda:0'), grad: tensor([[ 6.6459e-05, 2.6673e-05, 1.6987e-05, ..., 0.0000e+00, + 7.2896e-05, 0.0000e+00], + [ 3.0488e-05, -4.5598e-06, 1.9386e-05, ..., 0.0000e+00, + 1.2964e-05, 0.0000e+00], + [ 1.0329e-04, 4.3303e-05, 1.1936e-05, ..., 0.0000e+00, + 1.4699e-04, 0.0000e+00], + ..., + [ 5.0455e-05, 8.0645e-05, 2.8896e-04, ..., 0.0000e+00, + 2.1189e-05, 0.0000e+00], + [-7.8082e-06, -1.6794e-05, 5.0366e-05, ..., 0.0000e+00, + -6.7592e-05, 0.0000e+00], + [ 3.3498e-05, -1.2982e-04, -8.4019e-04, ..., 0.0000e+00, + 1.4246e-05, 0.0000e+00]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0068, 0.0201, -0.0155, 0.0130, 0.0238, -0.0200, -0.0282, -0.0181, + -0.0175, -0.0245], device='cuda:0'), grad: tensor([ 2.4045e-04, 6.5744e-05, 3.4022e-04, -1.8775e-04, 7.0810e-04, + -6.1321e-04, 4.2892e-04, 7.4816e-04, 5.1677e-05, -1.7834e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 59---------------------------------------------------- +epoch 59, time 217.76, cls_loss 0.0112 cls_loss_mapping 0.0195 cls_loss_causal 0.6962 re_mapping 0.0126 re_causal 0.0370 /// teacc 99.01 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0657, -0.0440, -0.0453, ..., -0.0230, -0.0276, -0.0139], + [-0.0075, 0.0515, -0.0453, ..., 0.0068, -0.0047, -0.0585], + [ 0.0222, -0.0628, -0.0450, ..., -0.0346, 0.0472, -0.0220], + ..., + [-0.0425, -0.0462, 0.0511, ..., 0.0464, -0.0804, 0.0057], + [ 0.0343, 0.0198, -0.1114, ..., -0.0217, 0.0703, -0.0110], + [-0.0501, -0.0452, 0.0330, ..., 0.0005, -0.0357, -0.0415]], + device='cuda:0'), grad: tensor([[ 3.0667e-05, 1.9129e-06, 1.5572e-06, ..., 0.0000e+00, + 6.9082e-05, 0.0000e+00], + [ 1.3635e-05, -1.7121e-05, 9.3356e-06, ..., 0.0000e+00, + 3.2365e-05, 0.0000e+00], + [-2.1338e-04, 5.8003e-06, 5.3495e-06, ..., 0.0000e+00, + -2.8706e-04, 0.0000e+00], + ..., + [ 1.1139e-05, 1.3195e-05, -2.3007e-04, ..., 0.0000e+00, + 2.8282e-05, 0.0000e+00], + [ 4.4632e-04, 6.0201e-06, 1.9506e-05, ..., 0.0000e+00, + 1.1244e-03, 0.0000e+00], + [ 2.8551e-05, 1.4044e-05, 1.9288e-04, ..., 0.0000e+00, + 8.2374e-05, 0.0000e+00]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0068, 0.0208, -0.0160, 0.0127, 0.0242, -0.0190, -0.0280, -0.0184, + -0.0178, -0.0251], device='cuda:0'), grad: tensor([-2.3067e-04, 5.5403e-05, -2.6560e-04, -2.1095e-03, -1.0453e-05, + 3.2687e-04, 2.2411e-04, -2.6202e-04, 1.8005e-03, 4.7112e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 216.89, cls_loss 0.0087 cls_loss_mapping 0.0181 cls_loss_causal 0.6627 re_mapping 0.0126 re_causal 0.0391 /// teacc 98.70 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0662, -0.0444, -0.0456, ..., -0.0230, -0.0277, -0.0139], + [-0.0078, 0.0518, -0.0461, ..., 0.0071, -0.0049, -0.0586], + [ 0.0223, -0.0634, -0.0457, ..., -0.0353, 0.0478, -0.0219], + ..., + [-0.0430, -0.0460, 0.0519, ..., 0.0465, -0.0811, 0.0057], + [ 0.0346, 0.0198, -0.1125, ..., -0.0218, 0.0707, -0.0110], + [-0.0506, -0.0459, 0.0324, ..., 0.0004, -0.0366, -0.0415]], + device='cuda:0'), grad: tensor([[ 7.4506e-06, 7.8604e-06, 4.5523e-06, ..., 0.0000e+00, + -6.6996e-05, 0.0000e+00], + [ 2.4913e-07, -3.4451e-05, -9.8720e-07, ..., 0.0000e+00, + -3.4161e-06, 0.0000e+00], + [ 5.9791e-06, 5.7630e-06, 4.2990e-06, ..., 0.0000e+00, + 1.9848e-05, 0.0000e+00], + ..., + [ 2.8964e-07, 2.5123e-05, -4.3362e-06, ..., 0.0000e+00, + 4.5970e-06, 0.0000e+00], + [-3.7923e-06, 2.3350e-05, 3.0309e-05, ..., 0.0000e+00, + 9.0957e-05, 0.0000e+00], + [ 1.8515e-06, 6.4336e-06, -2.9802e-05, ..., 0.0000e+00, + 6.7241e-06, 0.0000e+00]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0063, 0.0204, -0.0159, 0.0127, 0.0253, -0.0189, -0.0283, -0.0180, + -0.0180, -0.0261], device='cuda:0'), grad: tensor([-1.1420e-04, -2.6524e-05, 4.1217e-05, -6.5982e-05, -1.7524e-05, + 3.9071e-05, -4.4703e-05, 1.0639e-05, 2.1923e-04, -4.1217e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 217.12, cls_loss 0.0076 cls_loss_mapping 0.0149 cls_loss_causal 0.6310 re_mapping 0.0117 re_causal 0.0374 /// teacc 98.82 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0670, -0.0449, -0.0458, ..., -0.0230, -0.0279, -0.0139], + [-0.0083, 0.0522, -0.0463, ..., 0.0072, -0.0054, -0.0586], + [ 0.0227, -0.0645, -0.0463, ..., -0.0354, 0.0484, -0.0218], + ..., + [-0.0434, -0.0459, 0.0520, ..., 0.0465, -0.0813, 0.0057], + [ 0.0348, 0.0201, -0.1137, ..., -0.0219, 0.0712, -0.0110], + [-0.0511, -0.0463, 0.0324, ..., 0.0004, -0.0373, -0.0415]], + device='cuda:0'), grad: tensor([[ 1.1779e-05, 3.2894e-06, 5.6364e-06, ..., 0.0000e+00, + 5.4948e-06, 0.0000e+00], + [ 2.8685e-05, 3.2306e-05, 1.0800e-04, ..., 0.0000e+00, + 3.0115e-05, 0.0000e+00], + [-2.7612e-05, 7.3612e-05, 1.2159e-04, ..., 0.0000e+00, + -3.4243e-05, 0.0000e+00], + ..., + [ 1.4581e-05, -9.1434e-05, -5.7125e-04, ..., 0.0000e+00, + 1.4372e-05, 0.0000e+00], + [ 8.5354e-05, -8.4564e-06, 2.6298e-04, ..., 0.0000e+00, + -1.7151e-05, 0.0000e+00], + [ 1.3195e-05, 8.8885e-06, -4.9263e-05, ..., 0.0000e+00, + 1.8865e-05, 0.0000e+00]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0075, 0.0202, -0.0162, 0.0125, 0.0258, -0.0191, -0.0279, -0.0177, + -0.0179, -0.0257], device='cuda:0'), grad: tensor([-2.4796e-04, 3.1424e-04, 4.1461e-04, -3.5912e-05, 1.5569e-04, + -1.6546e-04, 1.1826e-04, -1.3628e-03, 6.9046e-04, 1.1760e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 217.19, cls_loss 0.0080 cls_loss_mapping 0.0138 cls_loss_causal 0.6243 re_mapping 0.0119 re_causal 0.0341 /// teacc 98.88 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0680, -0.0454, -0.0461, ..., -0.0231, -0.0274, -0.0136], + [-0.0087, 0.0533, -0.0460, ..., 0.0074, -0.0057, -0.0591], + [ 0.0234, -0.0651, -0.0466, ..., -0.0359, 0.0488, -0.0219], + ..., + [-0.0448, -0.0473, 0.0524, ..., 0.0466, -0.0820, 0.0057], + [ 0.0346, 0.0200, -0.1147, ..., -0.0222, 0.0713, -0.0111], + [-0.0519, -0.0465, 0.0324, ..., 0.0004, -0.0379, -0.0416]], + device='cuda:0'), grad: tensor([[ 3.5465e-05, 4.3422e-05, 2.6077e-06, ..., 0.0000e+00, + 1.3329e-05, 0.0000e+00], + [ 1.0812e-04, 4.5240e-05, 5.3607e-06, ..., 0.0000e+00, + 1.6212e-04, 0.0000e+00], + [-1.5998e-04, -4.8488e-05, 1.8626e-05, ..., 0.0000e+00, + -1.5628e-04, 0.0000e+00], + ..., + [ 3.4183e-05, 4.1537e-06, -1.6749e-05, ..., 0.0000e+00, + 1.3983e-04, 0.0000e+00], + [ 1.9446e-05, 8.9183e-06, 1.6227e-05, ..., 0.0000e+00, + 8.1122e-05, 0.0000e+00], + [ 3.5055e-06, 2.1607e-06, 5.2750e-06, ..., 0.0000e+00, + 3.3379e-05, 0.0000e+00]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0070, 0.0207, -0.0162, 0.0126, 0.0255, -0.0185, -0.0280, -0.0181, + -0.0184, -0.0257], device='cuda:0'), grad: tensor([ 8.8274e-05, 2.8181e-04, -1.8680e-04, -9.1887e-04, 1.9357e-05, + 3.2878e-04, -8.0287e-05, 1.9121e-04, 1.9121e-04, 8.4221e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 217.00, cls_loss 0.0073 cls_loss_mapping 0.0137 cls_loss_causal 0.6637 re_mapping 0.0120 re_causal 0.0367 /// teacc 98.76 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0688, -0.0458, -0.0463, ..., -0.0232, -0.0276, -0.0137], + [-0.0091, 0.0538, -0.0464, ..., 0.0074, -0.0058, -0.0593], + [ 0.0241, -0.0655, -0.0471, ..., -0.0360, 0.0496, -0.0215], + ..., + [-0.0457, -0.0475, 0.0531, ..., 0.0467, -0.0828, 0.0056], + [ 0.0346, 0.0198, -0.1156, ..., -0.0223, 0.0715, -0.0112], + [-0.0522, -0.0468, 0.0331, ..., 0.0004, -0.0385, -0.0416]], + device='cuda:0'), grad: tensor([[ 4.5478e-05, 1.7798e-04, 2.0906e-05, ..., 2.2165e-07, + 1.3255e-05, -1.4435e-08], + [ 1.6496e-05, -1.9446e-05, 1.1064e-06, ..., -9.1195e-06, + 1.5587e-05, 0.0000e+00], + [-1.0097e-04, 5.0068e-05, 1.9997e-05, ..., 3.9674e-07, + -1.7071e-04, 2.3283e-09], + ..., + [ 1.3925e-05, 5.3763e-05, 4.5858e-06, ..., 2.3898e-06, + 1.7881e-05, 4.6566e-10], + [-7.2145e-04, -5.0879e-04, 3.9816e-05, ..., 8.5682e-07, + -6.2346e-05, 4.6566e-10], + [ 1.7598e-05, 2.1422e-04, 2.4647e-05, ..., 3.6824e-06, + 1.9252e-05, 6.9849e-09]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0070, 0.0206, -0.0158, 0.0127, 0.0254, -0.0189, -0.0282, -0.0182, + -0.0186, -0.0255], device='cuda:0'), grad: tensor([ 1.3196e-04, 3.8922e-05, -1.3077e-04, 1.0860e-04, -9.1553e-04, + 1.0042e-03, 2.3568e-04, 1.1575e-04, -9.4080e-04, 3.5000e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 217.12, cls_loss 0.0084 cls_loss_mapping 0.0168 cls_loss_causal 0.6403 re_mapping 0.0119 re_causal 0.0356 /// teacc 98.89 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0695, -0.0465, -0.0465, ..., -0.0235, -0.0278, -0.0137], + [-0.0098, 0.0535, -0.0471, ..., 0.0078, -0.0059, -0.0596], + [ 0.0247, -0.0659, -0.0474, ..., -0.0362, 0.0502, -0.0208], + ..., + [-0.0473, -0.0478, 0.0532, ..., 0.0467, -0.0840, 0.0055], + [ 0.0351, 0.0203, -0.1163, ..., -0.0226, 0.0718, -0.0112], + [-0.0528, -0.0480, 0.0331, ..., 0.0003, -0.0392, -0.0417]], + device='cuda:0'), grad: tensor([[ 1.8433e-05, 1.5795e-05, 1.4342e-06, ..., 0.0000e+00, + 1.2413e-05, 0.0000e+00], + [ 9.4831e-05, 2.0266e-04, 3.3844e-06, ..., 0.0000e+00, + 1.3423e-04, 0.0000e+00], + [-1.4222e-04, 2.2337e-05, 2.6617e-06, ..., 0.0000e+00, + -1.8013e-04, 0.0000e+00], + ..., + [ 9.4846e-06, 2.5511e-05, -5.9195e-06, ..., 0.0000e+00, + 1.1891e-05, 0.0000e+00], + [-2.2018e-04, -7.4291e-04, 4.5896e-06, ..., 0.0000e+00, + -4.1008e-04, 0.0000e+00], + [ 8.8751e-05, 3.7217e-04, -4.6313e-05, ..., 0.0000e+00, + 1.4126e-04, 0.0000e+00]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0071, 0.0202, -0.0156, 0.0132, 0.0259, -0.0192, -0.0279, -0.0187, + -0.0184, -0.0255], device='cuda:0'), grad: tensor([ 4.3541e-05, 6.3229e-04, -1.3554e-04, 5.2309e-04, 3.9130e-05, + 2.8498e-06, 4.1097e-05, 4.2111e-05, -2.0638e-03, 8.7500e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 217.03, cls_loss 0.0069 cls_loss_mapping 0.0129 cls_loss_causal 0.6120 re_mapping 0.0115 re_causal 0.0341 /// teacc 98.74 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0705, -0.0475, -0.0466, ..., -0.0234, -0.0282, -0.0137], + [-0.0103, 0.0530, -0.0472, ..., 0.0075, -0.0067, -0.0597], + [ 0.0249, -0.0647, -0.0476, ..., -0.0365, 0.0509, -0.0208], + ..., + [-0.0479, -0.0483, 0.0537, ..., 0.0471, -0.0844, 0.0055], + [ 0.0353, 0.0204, -0.1168, ..., -0.0231, 0.0721, -0.0113], + [-0.0534, -0.0487, 0.0333, ..., 0.0002, -0.0398, -0.0417]], + device='cuda:0'), grad: tensor([[ 1.8224e-05, 2.4676e-05, 3.0380e-06, ..., 3.0315e-07, + 1.3700e-06, 0.0000e+00], + [ 9.3728e-06, 3.3927e-04, 1.0407e-04, ..., 1.8580e-07, + 9.3058e-06, 0.0000e+00], + [ 1.6376e-05, 3.8087e-05, 7.9349e-06, ..., 3.9227e-06, + 2.8536e-05, 0.0000e+00], + ..., + [ 2.2985e-06, 2.4676e-05, 7.8008e-06, ..., -2.3143e-07, + 3.2447e-06, 0.0000e+00], + [-3.6925e-05, 2.3767e-05, 5.6922e-05, ..., 2.6776e-07, + -4.6879e-05, 0.0000e+00], + [ 1.7900e-06, 3.9756e-05, -4.9829e-04, ..., 6.7614e-07, + 2.6412e-06, 0.0000e+00]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0073, 0.0192, -0.0148, 0.0126, 0.0257, -0.0190, -0.0271, -0.0186, + -0.0185, -0.0253], device='cuda:0'), grad: tensor([-2.2259e-07, 4.7684e-04, 1.0180e-04, -4.1187e-05, 1.7655e-04, + 8.3447e-05, 2.0210e-06, 5.6326e-05, 1.1051e-04, -9.6607e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 217.24, cls_loss 0.0069 cls_loss_mapping 0.0138 cls_loss_causal 0.6458 re_mapping 0.0116 re_causal 0.0365 /// teacc 98.95 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0709, -0.0475, -0.0456, ..., -0.0234, -0.0272, -0.0137], + [-0.0106, 0.0533, -0.0477, ..., 0.0076, -0.0066, -0.0597], + [ 0.0251, -0.0654, -0.0479, ..., -0.0366, 0.0510, -0.0208], + ..., + [-0.0487, -0.0484, 0.0542, ..., 0.0471, -0.0852, 0.0055], + [ 0.0353, 0.0201, -0.1177, ..., -0.0231, 0.0725, -0.0113], + [-0.0539, -0.0490, 0.0334, ..., 0.0002, -0.0402, -0.0417]], + device='cuda:0'), grad: tensor([[ 9.3272e-07, 3.8743e-05, 6.1989e-06, ..., 0.0000e+00, + -4.5914e-07, 0.0000e+00], + [ 3.2000e-06, -5.3942e-05, 2.4512e-05, ..., 0.0000e+00, + 4.5449e-06, 0.0000e+00], + [ 5.0254e-06, 3.5703e-05, 1.2487e-05, ..., 0.0000e+00, + 8.0839e-06, 0.0000e+00], + ..., + [ 1.3346e-06, 4.7207e-05, -5.3704e-05, ..., 0.0000e+00, + 1.5656e-06, 0.0000e+00], + [-4.1783e-05, 1.2629e-05, 4.0114e-05, ..., 0.0000e+00, + -4.6790e-05, 0.0000e+00], + [ 2.7977e-06, -9.5785e-05, -7.6294e-04, ..., 0.0000e+00, + 4.9882e-06, 0.0000e+00]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0064, 0.0192, -0.0155, 0.0122, 0.0254, -0.0183, -0.0273, -0.0184, + -0.0190, -0.0249], device='cuda:0'), grad: tensor([-3.9697e-04, 5.9344e-06, 1.3769e-04, -4.8709e-04, 1.2646e-03, + 7.5006e-04, 2.4259e-04, -2.6554e-05, 1.0717e-04, -1.5974e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 217.10, cls_loss 0.0094 cls_loss_mapping 0.0154 cls_loss_causal 0.6116 re_mapping 0.0122 re_causal 0.0338 /// teacc 98.85 lr 0.00010000 +Epoch 69, weight, value: tensor([[-7.2334e-02, -4.8451e-02, -4.5197e-02, ..., -2.2278e-02, + -2.6256e-02, -1.3744e-02], + [-1.1065e-02, 5.3851e-02, -4.7050e-02, ..., 7.6305e-03, + -7.0474e-03, -5.9702e-02], + [ 2.5408e-02, -6.5839e-02, -4.8469e-02, ..., -3.6626e-02, + 5.1362e-02, -2.0723e-02], + ..., + [-4.9505e-02, -4.8887e-02, 5.4603e-02, ..., 4.7106e-02, + -8.5734e-02, 5.4447e-03], + [ 3.6260e-02, 1.9936e-02, -1.1846e-01, ..., -2.3340e-02, + 7.3071e-02, -1.1257e-02], + [-5.4607e-02, -4.9695e-02, 3.2880e-02, ..., 4.4850e-05, + -4.0958e-02, -4.1662e-02]], device='cuda:0'), grad: tensor([[ 4.9546e-06, 1.8561e-06, 1.6820e-06, ..., -4.3097e-07, + 1.3821e-05, 0.0000e+00], + [ 4.9584e-06, -2.7001e-05, 2.3529e-05, ..., 1.7043e-06, + 2.0295e-05, 0.0000e+00], + [-5.6893e-05, 1.4789e-05, 3.2723e-05, ..., -1.1437e-05, + -2.2471e-04, 0.0000e+00], + ..., + [ 6.7912e-06, 1.2234e-05, 1.0803e-05, ..., -6.3777e-06, + 2.2605e-05, 0.0000e+00], + [ 9.1717e-06, 3.0324e-06, 1.4491e-05, ..., 1.6028e-06, + 2.3171e-05, 0.0000e+00], + [ 5.0738e-06, 1.3411e-06, -1.4019e-04, ..., 4.5523e-06, + 5.8413e-06, 0.0000e+00]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0058, 0.0193, -0.0160, 0.0121, 0.0261, -0.0189, -0.0279, -0.0180, + -0.0182, -0.0258], device='cuda:0'), grad: tensor([ 1.7941e-05, 4.0114e-05, -3.8648e-04, 2.4939e-04, 7.0632e-05, + 2.3305e-05, 3.3826e-05, 5.2720e-05, 8.2254e-05, -1.8334e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 217.44, cls_loss 0.0083 cls_loss_mapping 0.0158 cls_loss_causal 0.6460 re_mapping 0.0118 re_causal 0.0334 /// teacc 98.71 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0742, -0.0491, -0.0457, ..., -0.0210, -0.0266, -0.0137], + [-0.0116, 0.0550, -0.0463, ..., 0.0086, -0.0070, -0.0597], + [ 0.0252, -0.0663, -0.0494, ..., -0.0377, 0.0515, -0.0207], + ..., + [-0.0501, -0.0506, 0.0552, ..., 0.0471, -0.0863, 0.0054], + [ 0.0361, 0.0197, -0.1193, ..., -0.0238, 0.0737, -0.0113], + [-0.0550, -0.0496, 0.0334, ..., -0.0003, -0.0417, -0.0417]], + device='cuda:0'), grad: tensor([[-1.5991e-06, -5.0738e-06, 4.5002e-06, ..., 0.0000e+00, + 4.7423e-06, 0.0000e+00], + [ 4.4480e-06, 1.0622e-04, 3.0208e-04, ..., 0.0000e+00, + 3.8594e-06, 0.0000e+00], + [ 1.1057e-05, 8.0988e-06, 8.5831e-06, ..., 0.0000e+00, + 1.4797e-05, 0.0000e+00], + ..., + [ 6.6273e-06, -2.8300e-04, -7.2384e-04, ..., 0.0000e+00, + 8.7023e-06, 0.0000e+00], + [ 5.5015e-05, 1.1206e-05, 5.0664e-06, ..., 0.0000e+00, + 2.0653e-05, 0.0000e+00], + [ 3.0220e-05, 1.4687e-04, 3.5357e-04, ..., 0.0000e+00, + 1.9804e-05, 0.0000e+00]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0067, 0.0201, -0.0166, 0.0119, 0.0255, -0.0189, -0.0270, -0.0181, + -0.0185, -0.0248], device='cuda:0'), grad: tensor([-7.2181e-05, 7.2193e-04, 9.4593e-05, -5.8174e-05, 1.2660e-04, + -5.5647e-04, 3.4571e-04, -1.6851e-03, 1.3638e-04, 9.4557e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 216.98, cls_loss 0.0065 cls_loss_mapping 0.0142 cls_loss_causal 0.6340 re_mapping 0.0115 re_causal 0.0338 /// teacc 98.89 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0750, -0.0495, -0.0461, ..., -0.0211, -0.0269, -0.0137], + [-0.0125, 0.0558, -0.0466, ..., 0.0092, -0.0070, -0.0597], + [ 0.0253, -0.0671, -0.0498, ..., -0.0380, 0.0519, -0.0207], + ..., + [-0.0508, -0.0511, 0.0561, ..., 0.0469, -0.0870, 0.0054], + [ 0.0366, 0.0202, -0.1200, ..., -0.0240, 0.0742, -0.0113], + [-0.0555, -0.0497, 0.0331, ..., -0.0004, -0.0424, -0.0417]], + device='cuda:0'), grad: tensor([[ 8.4266e-06, 2.5686e-06, 2.6170e-06, ..., 7.1712e-08, + 3.9749e-06, 0.0000e+00], + [ 8.9929e-06, -1.5631e-05, 1.0535e-05, ..., 6.8080e-07, + 1.0327e-05, 0.0000e+00], + [-4.5538e-05, 3.3379e-06, 5.1968e-06, ..., 3.3225e-07, + -6.3241e-05, 0.0000e+00], + ..., + [ 2.5511e-05, 6.5863e-06, -4.3988e-05, ..., -4.5896e-06, + 3.2753e-05, 0.0000e+00], + [-3.7849e-05, -1.6928e-05, 6.0648e-06, ..., 4.9593e-08, + -4.1664e-05, 0.0000e+00], + [-5.3905e-06, 6.6832e-06, 4.8988e-07, ..., 1.7518e-06, + 2.9951e-06, 0.0000e+00]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0065, 0.0204, -0.0169, 0.0119, 0.0252, -0.0190, -0.0271, -0.0178, + -0.0183, -0.0251], device='cuda:0'), grad: tensor([ 2.6137e-05, 1.8001e-05, -4.9293e-05, 9.9123e-05, 1.1183e-05, + -3.6303e-06, 7.0870e-05, -4.8637e-05, -5.6535e-05, -6.7294e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 216.89, cls_loss 0.0072 cls_loss_mapping 0.0148 cls_loss_causal 0.6256 re_mapping 0.0108 re_causal 0.0332 /// teacc 98.86 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0754, -0.0498, -0.0463, ..., -0.0211, -0.0271, -0.0137], + [-0.0129, 0.0562, -0.0474, ..., 0.0094, -0.0068, -0.0598], + [ 0.0255, -0.0679, -0.0505, ..., -0.0382, 0.0521, -0.0207], + ..., + [-0.0519, -0.0512, 0.0564, ..., 0.0468, -0.0878, 0.0054], + [ 0.0369, 0.0201, -0.1208, ..., -0.0241, 0.0748, -0.0113], + [-0.0568, -0.0501, 0.0330, ..., -0.0004, -0.0435, -0.0417]], + device='cuda:0'), grad: tensor([[ 3.8259e-06, 4.8429e-06, 8.3726e-07, ..., 0.0000e+00, + 2.4363e-06, 0.0000e+00], + [-2.1264e-05, -1.1864e-03, -3.1447e-04, ..., 0.0000e+00, + -7.7188e-05, 0.0000e+00], + [ 1.7673e-05, 8.8930e-05, 7.6070e-06, ..., 0.0000e+00, + 5.0783e-05, 0.0000e+00], + ..., + [ 5.2378e-06, 9.9564e-04, 2.8920e-04, ..., 0.0000e+00, + 7.3016e-06, 0.0000e+00], + [ 9.1083e-07, 1.6078e-05, 8.2254e-06, ..., 0.0000e+00, + -2.0936e-05, 0.0000e+00], + [ 3.4925e-06, 1.9342e-05, 5.3719e-06, ..., 0.0000e+00, + 4.2692e-06, 0.0000e+00]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0063, 0.0203, -0.0174, 0.0129, 0.0257, -0.0196, -0.0270, -0.0181, + -0.0182, -0.0254], device='cuda:0'), grad: tensor([ 1.0386e-05, -1.7519e-03, 1.3411e-04, 8.2672e-05, -2.6405e-05, + -1.0878e-05, 2.3931e-05, 1.4820e-03, 2.8759e-05, 3.0309e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 216.74, cls_loss 0.0060 cls_loss_mapping 0.0114 cls_loss_causal 0.6291 re_mapping 0.0108 re_causal 0.0333 /// teacc 98.90 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0763, -0.0506, -0.0466, ..., -0.0211, -0.0274, -0.0137], + [-0.0136, 0.0560, -0.0488, ..., 0.0096, -0.0071, -0.0598], + [ 0.0258, -0.0681, -0.0509, ..., -0.0382, 0.0525, -0.0207], + ..., + [-0.0525, -0.0507, 0.0572, ..., 0.0468, -0.0884, 0.0054], + [ 0.0373, 0.0201, -0.1218, ..., -0.0241, 0.0751, -0.0113], + [-0.0573, -0.0511, 0.0332, ..., -0.0005, -0.0440, -0.0417]], + device='cuda:0'), grad: tensor([[ 6.0722e-06, 5.8375e-06, 1.5553e-06, ..., 0.0000e+00, + 7.6443e-06, 0.0000e+00], + [ 5.8934e-06, -2.5649e-06, 1.1884e-05, ..., 0.0000e+00, + 7.9721e-06, 0.0000e+00], + [ 2.6628e-05, 2.8044e-05, 8.1360e-06, ..., 0.0000e+00, + 3.8683e-05, 0.0000e+00], + ..., + [ 2.6897e-06, -3.3695e-06, -5.0873e-05, ..., 0.0000e+00, + 3.1479e-06, 0.0000e+00], + [-1.1671e-04, -1.1551e-04, 8.4564e-06, ..., 0.0000e+00, + -2.0146e-04, 0.0000e+00], + [ 7.8604e-06, 1.6764e-05, 1.4201e-05, ..., 0.0000e+00, + 8.5533e-06, 0.0000e+00]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0064, 0.0196, -0.0173, 0.0130, 0.0261, -0.0197, -0.0268, -0.0179, + -0.0183, -0.0254], device='cuda:0'), grad: tensor([-6.0469e-05, 2.9951e-05, 9.0539e-05, 2.3723e-04, -4.3035e-05, + -3.4273e-05, 5.4628e-05, -6.8903e-05, -2.7943e-04, 7.3493e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 217.14, cls_loss 0.0052 cls_loss_mapping 0.0111 cls_loss_causal 0.6417 re_mapping 0.0103 re_causal 0.0333 /// teacc 98.81 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0770, -0.0509, -0.0472, ..., -0.0209, -0.0274, -0.0137], + [-0.0138, 0.0567, -0.0487, ..., 0.0097, -0.0069, -0.0598], + [ 0.0258, -0.0685, -0.0519, ..., -0.0383, 0.0525, -0.0207], + ..., + [-0.0525, -0.0508, 0.0582, ..., 0.0468, -0.0887, 0.0054], + [ 0.0375, 0.0202, -0.1228, ..., -0.0242, 0.0755, -0.0113], + [-0.0576, -0.0511, 0.0331, ..., -0.0005, -0.0447, -0.0417]], + device='cuda:0'), grad: tensor([[ 2.5313e-06, 7.2308e-06, 2.2382e-05, ..., -4.4005e-08, + 2.5630e-06, 0.0000e+00], + [ 5.4725e-06, 1.9997e-05, 4.7803e-05, ..., 6.9849e-10, + 9.0897e-06, 0.0000e+00], + [ 1.0744e-05, 4.3996e-06, 1.9997e-05, ..., 8.6147e-09, + 2.9519e-05, 0.0000e+00], + ..., + [ 2.7135e-05, 3.9756e-05, 5.0259e-04, ..., 1.6298e-09, + 6.8069e-05, 0.0000e+00], + [-6.0588e-05, 5.4277e-06, 2.7493e-05, ..., 3.9581e-09, + -7.0572e-05, 0.0000e+00], + [ 1.0207e-05, -2.2817e-04, -1.2331e-03, ..., 1.1642e-08, + 1.3687e-05, 0.0000e+00]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0064, 0.0200, -0.0178, 0.0127, 0.0258, -0.0195, -0.0268, -0.0173, + -0.0184, -0.0255], device='cuda:0'), grad: tensor([ 5.1886e-05, 1.1337e-04, 9.7096e-05, -6.8188e-05, 1.2207e-03, + 4.9502e-05, 3.0026e-05, 9.5701e-04, -5.4687e-05, -2.3956e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 216.98, cls_loss 0.0057 cls_loss_mapping 0.0102 cls_loss_causal 0.6054 re_mapping 0.0108 re_causal 0.0326 /// teacc 98.84 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0778, -0.0516, -0.0476, ..., -0.0210, -0.0273, -0.0137], + [-0.0140, 0.0567, -0.0491, ..., 0.0098, -0.0078, -0.0598], + [ 0.0262, -0.0681, -0.0518, ..., -0.0385, 0.0537, -0.0207], + ..., + [-0.0529, -0.0510, 0.0586, ..., 0.0467, -0.0892, 0.0054], + [ 0.0373, 0.0201, -0.1234, ..., -0.0243, 0.0752, -0.0113], + [-0.0582, -0.0524, 0.0328, ..., -0.0005, -0.0453, -0.0417]], + device='cuda:0'), grad: tensor([[ 6.0052e-06, -2.6226e-05, 2.0880e-06, ..., 1.4971e-07, + -2.7373e-05, 0.0000e+00], + [ 1.3262e-06, 4.1649e-06, 6.5863e-06, ..., 1.6182e-07, + 1.7174e-06, 0.0000e+00], + [ 6.8638e-07, 3.2727e-06, 7.5176e-06, ..., 5.7369e-07, + 5.7556e-07, 0.0000e+00], + ..., + [ 7.2364e-07, 6.5751e-06, 3.1926e-06, ..., 1.4091e-06, + 3.3416e-06, 0.0000e+00], + [ 6.0759e-06, 1.6708e-06, 5.0813e-06, ..., 2.1770e-07, + 2.1145e-05, 0.0000e+00], + [ 2.8778e-06, 2.2411e-05, -3.2216e-05, ..., 2.0750e-06, + 4.3586e-06, 0.0000e+00]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0067, 0.0195, -0.0171, 0.0131, 0.0261, -0.0190, -0.0267, -0.0172, + -0.0189, -0.0263], device='cuda:0'), grad: tensor([-4.1795e-04, 2.5898e-05, 6.2227e-05, -5.2541e-05, 1.2405e-05, + 1.5461e-04, 8.0705e-05, 1.5378e-05, 8.8573e-05, 3.0577e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 217.11, cls_loss 0.0061 cls_loss_mapping 0.0142 cls_loss_causal 0.6313 re_mapping 0.0111 re_causal 0.0332 /// teacc 98.80 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0784, -0.0521, -0.0479, ..., -0.0209, -0.0276, -0.0137], + [-0.0144, 0.0564, -0.0497, ..., 0.0098, -0.0081, -0.0598], + [ 0.0266, -0.0679, -0.0518, ..., -0.0386, 0.0548, -0.0207], + ..., + [-0.0534, -0.0506, 0.0590, ..., 0.0467, -0.0899, 0.0054], + [ 0.0372, 0.0199, -0.1242, ..., -0.0244, 0.0751, -0.0113], + [-0.0588, -0.0524, 0.0335, ..., -0.0005, -0.0457, -0.0417]], + device='cuda:0'), grad: tensor([[ 9.1672e-05, 2.9877e-05, 7.7952e-07, ..., 0.0000e+00, + 2.5535e-04, -1.4994e-07], + [ 3.3677e-05, -1.2502e-05, -2.1197e-06, ..., 0.0000e+00, + 1.0043e-04, 6.9849e-10], + [-2.1172e-04, -1.4424e-05, -1.7267e-06, ..., 0.0000e+00, + -7.3576e-04, 5.5879e-09], + ..., + [ 3.5733e-05, 1.2748e-05, -1.8343e-05, ..., 0.0000e+00, + 5.9158e-05, 1.6298e-09], + [ 1.3304e-04, 4.5657e-05, 2.9784e-06, ..., 0.0000e+00, + 1.6582e-04, 4.4238e-09], + [ 2.5734e-05, 1.2949e-05, 4.2878e-06, ..., 0.0000e+00, + 2.9027e-05, 1.1572e-07]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0071, 0.0190, -0.0164, 0.0130, 0.0252, -0.0191, -0.0263, -0.0172, + -0.0194, -0.0253], device='cuda:0'), grad: tensor([ 3.7098e-04, 1.3959e-04, -1.0052e-03, 2.1803e-04, 2.0313e-04, + -1.0233e-03, 5.7125e-04, 7.5698e-05, 3.6120e-04, 8.9169e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 216.91, cls_loss 0.0054 cls_loss_mapping 0.0110 cls_loss_causal 0.6232 re_mapping 0.0102 re_causal 0.0319 /// teacc 98.76 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0792, -0.0526, -0.0483, ..., -0.0209, -0.0279, -0.0137], + [-0.0147, 0.0570, -0.0498, ..., 0.0101, -0.0078, -0.0598], + [ 0.0267, -0.0685, -0.0522, ..., -0.0388, 0.0549, -0.0207], + ..., + [-0.0538, -0.0508, 0.0585, ..., 0.0467, -0.0903, 0.0054], + [ 0.0376, 0.0196, -0.1250, ..., -0.0245, 0.0757, -0.0114], + [-0.0600, -0.0531, 0.0342, ..., -0.0005, -0.0465, -0.0417]], + device='cuda:0'), grad: tensor([[-4.2319e-06, 1.7919e-06, 9.2573e-07, ..., 0.0000e+00, + 1.3653e-06, 0.0000e+00], + [ 4.0489e-07, -6.2823e-05, 1.0710e-07, ..., 0.0000e+00, + 1.7555e-06, 0.0000e+00], + [-3.0585e-06, 3.4928e-05, 3.0454e-06, ..., 4.6566e-10, + -1.0869e-06, 0.0000e+00], + ..., + [ 6.6310e-07, 4.9286e-06, -7.5623e-06, ..., 2.3283e-10, + 1.6373e-06, 0.0000e+00], + [ 4.2357e-06, 2.4401e-06, 5.0385e-07, ..., 4.6566e-10, + 7.9721e-06, 0.0000e+00], + [ 6.1747e-07, 1.1530e-06, 2.4177e-06, ..., 2.3283e-10, + -1.7345e-05, 0.0000e+00]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0073, 0.0195, -0.0168, 0.0129, 0.0257, -0.0190, -0.0262, -0.0181, + -0.0194, -0.0248], device='cuda:0'), grad: tensor([-4.5329e-05, -6.8665e-05, 5.7727e-05, 3.6746e-05, -7.0572e-05, + 1.7792e-05, 1.2422e-04, -1.7090e-06, 4.1008e-05, -9.1314e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 216.96, cls_loss 0.0051 cls_loss_mapping 0.0103 cls_loss_causal 0.6286 re_mapping 0.0103 re_causal 0.0326 /// teacc 98.97 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0799, -0.0529, -0.0487, ..., -0.0200, -0.0281, -0.0136], + [-0.0149, 0.0569, -0.0503, ..., 0.0101, -0.0082, -0.0599], + [ 0.0270, -0.0684, -0.0525, ..., -0.0389, 0.0559, -0.0209], + ..., + [-0.0543, -0.0509, 0.0591, ..., 0.0467, -0.0911, 0.0054], + [ 0.0386, 0.0198, -0.1255, ..., -0.0246, 0.0762, -0.0114], + [-0.0608, -0.0539, 0.0342, ..., -0.0006, -0.0475, -0.0418]], + device='cuda:0'), grad: tensor([[ 3.0249e-06, 2.3395e-06, 6.3609e-07, ..., 0.0000e+00, + 1.1511e-06, 0.0000e+00], + [ 8.7405e-07, -2.5123e-05, 3.4627e-06, ..., 0.0000e+00, + -2.6263e-06, 0.0000e+00], + [-1.4871e-05, 1.3202e-05, 2.6505e-06, ..., 0.0000e+00, + -1.2314e-04, 0.0000e+00], + ..., + [ 2.2613e-06, 7.8082e-06, 1.1645e-05, ..., 0.0000e+00, + 4.8466e-06, 0.0000e+00], + [-1.2733e-05, -3.4925e-08, 1.5274e-05, ..., 0.0000e+00, + -4.5151e-06, 0.0000e+00], + [ 3.2056e-06, 9.5293e-06, -4.2468e-05, ..., 0.0000e+00, + 1.8673e-06, 0.0000e+00]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0080, 0.0190, -0.0164, 0.0123, 0.0259, -0.0185, -0.0273, -0.0177, + -0.0183, -0.0249], device='cuda:0'), grad: tensor([ 8.6650e-06, -3.0249e-05, -1.3447e-04, 1.3141e-06, 1.4031e-04, + -6.2324e-06, 8.1733e-06, 3.7074e-05, 2.1219e-05, -4.5627e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 77---------------------------------------------------- +epoch 77, time 217.93, cls_loss 0.0060 cls_loss_mapping 0.0104 cls_loss_causal 0.6013 re_mapping 0.0100 re_causal 0.0314 /// teacc 99.04 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0805, -0.0530, -0.0489, ..., -0.0194, -0.0281, -0.0135], + [-0.0150, 0.0575, -0.0504, ..., 0.0101, -0.0083, -0.0600], + [ 0.0275, -0.0687, -0.0523, ..., -0.0388, 0.0568, -0.0209], + ..., + [-0.0546, -0.0513, 0.0594, ..., 0.0466, -0.0915, 0.0054], + [ 0.0380, 0.0192, -0.1261, ..., -0.0248, 0.0763, -0.0114], + [-0.0618, -0.0544, 0.0342, ..., -0.0007, -0.0490, -0.0419]], + device='cuda:0'), grad: tensor([[-9.3430e-06, 1.9800e-06, 7.9395e-07, ..., 0.0000e+00, + -2.3440e-05, 0.0000e+00], + [ 1.3430e-06, -1.1069e-04, 3.1218e-06, ..., 0.0000e+00, + 3.3099e-06, 0.0000e+00], + [-5.3763e-05, 3.5763e-06, 6.8173e-06, ..., 0.0000e+00, + -1.4782e-04, 0.0000e+00], + ..., + [ 5.6215e-06, 8.7917e-06, -1.4082e-05, ..., 0.0000e+00, + 1.4961e-05, 0.0000e+00], + [ 2.1756e-05, 2.1428e-05, 6.4149e-06, ..., 0.0000e+00, + 5.3465e-05, 0.0000e+00], + [ 2.0936e-06, 3.7402e-06, -9.4593e-05, ..., 0.0000e+00, + 5.4576e-06, 0.0000e+00]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0086, 0.0192, -0.0161, 0.0118, 0.0254, -0.0184, -0.0264, -0.0175, + -0.0189, -0.0245], device='cuda:0'), grad: tensor([-1.7178e-04, -1.4663e-04, -2.6584e-04, 1.6153e-04, 2.5153e-04, + 2.3961e-05, 4.4852e-05, 2.8640e-05, 2.1386e-04, -1.4031e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 216.88, cls_loss 0.0050 cls_loss_mapping 0.0119 cls_loss_causal 0.6045 re_mapping 0.0103 re_causal 0.0311 /// teacc 98.85 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0807, -0.0528, -0.0478, ..., -0.0191, -0.0271, -0.0134], + [-0.0152, 0.0585, -0.0507, ..., 0.0101, -0.0080, -0.0602], + [ 0.0278, -0.0693, -0.0533, ..., -0.0387, 0.0574, -0.0212], + ..., + [-0.0556, -0.0516, 0.0605, ..., 0.0466, -0.0922, 0.0054], + [ 0.0384, 0.0190, -0.1267, ..., -0.0251, 0.0766, -0.0114], + [-0.0627, -0.0550, 0.0343, ..., -0.0007, -0.0504, -0.0420]], + device='cuda:0'), grad: tensor([[ 3.0324e-05, 2.1324e-05, 1.1995e-05, ..., 2.8638e-08, + 1.9222e-06, 0.0000e+00], + [ 1.4594e-06, 3.0939e-06, 5.8003e-06, ..., 4.4238e-08, + 3.0011e-05, 0.0000e+00], + [-2.0564e-05, -2.0023e-07, 2.0992e-06, ..., 2.6519e-07, + -6.8307e-05, 0.0000e+00], + ..., + [ 1.3197e-06, 8.8662e-06, 2.3991e-06, ..., 1.5600e-07, + 6.4671e-06, 0.0000e+00], + [ 1.8910e-05, 1.2174e-05, 9.1791e-06, ..., 9.1502e-08, + 2.5168e-05, 0.0000e+00], + [ 9.2806e-07, 1.6761e-04, 1.1003e-04, ..., 5.6112e-08, + 1.3029e-06, 0.0000e+00]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0075, 0.0197, -0.0162, 0.0117, 0.0243, -0.0188, -0.0268, -0.0168, + -0.0189, -0.0248], device='cuda:0'), grad: tensor([ 5.1677e-05, 8.0109e-05, -9.7752e-05, 3.1181e-06, -4.3964e-04, + 1.7390e-05, -1.7202e-04, 3.0845e-05, 7.6830e-05, 4.4870e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 217.00, cls_loss 0.0045 cls_loss_mapping 0.0095 cls_loss_causal 0.6149 re_mapping 0.0104 re_causal 0.0320 /// teacc 98.86 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0812, -0.0531, -0.0479, ..., -0.0190, -0.0272, -0.0133], + [-0.0155, 0.0592, -0.0508, ..., 0.0103, -0.0079, -0.0602], + [ 0.0278, -0.0699, -0.0541, ..., -0.0389, 0.0575, -0.0212], + ..., + [-0.0564, -0.0523, 0.0608, ..., 0.0466, -0.0928, 0.0054], + [ 0.0386, 0.0189, -0.1275, ..., -0.0252, 0.0769, -0.0116], + [-0.0631, -0.0553, 0.0335, ..., -0.0008, -0.0509, -0.0420]], + device='cuda:0'), grad: tensor([[ 2.9653e-06, 8.1062e-06, 1.4212e-06, ..., 0.0000e+00, + 1.8496e-06, 0.0000e+00], + [ 1.5236e-06, -3.9697e-05, 1.0759e-05, ..., 0.0000e+00, + -2.4140e-06, 0.0000e+00], + [ 2.2873e-06, 2.1830e-06, 1.0423e-05, ..., 0.0000e+00, + 2.2184e-06, 0.0000e+00], + ..., + [ 2.4978e-06, 5.5619e-06, -1.7643e-05, ..., 0.0000e+00, + 3.3751e-06, 0.0000e+00], + [-1.8939e-05, 1.4789e-05, 2.6710e-06, ..., 0.0000e+00, + -1.9476e-05, 0.0000e+00], + [ 3.3751e-06, 1.6367e-04, 2.2507e-04, ..., 0.0000e+00, + 4.4107e-06, 0.0000e+00]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0071, 0.0202, -0.0167, 0.0119, 0.0258, -0.0190, -0.0265, -0.0171, + -0.0191, -0.0258], device='cuda:0'), grad: tensor([ 1.5378e-05, -2.5362e-05, 3.2097e-05, 1.9252e-05, -5.0926e-04, + 1.3828e-05, 9.1642e-06, -2.8372e-05, -2.1487e-05, 4.9496e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 216.87, cls_loss 0.0058 cls_loss_mapping 0.0106 cls_loss_causal 0.6173 re_mapping 0.0100 re_causal 0.0291 /// teacc 98.85 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0824, -0.0535, -0.0478, ..., -0.0189, -0.0276, -0.0133], + [-0.0160, 0.0603, -0.0516, ..., 0.0103, -0.0070, -0.0603], + [ 0.0291, -0.0708, -0.0538, ..., -0.0389, 0.0581, -0.0212], + ..., + [-0.0572, -0.0527, 0.0614, ..., 0.0466, -0.0934, 0.0054], + [ 0.0380, 0.0188, -0.1285, ..., -0.0253, 0.0769, -0.0117], + [-0.0639, -0.0562, 0.0331, ..., -0.0008, -0.0516, -0.0420]], + device='cuda:0'), grad: tensor([[ 7.0743e-06, -1.7658e-05, -3.4118e-04, ..., -7.2410e-08, + -1.6165e-04, 0.0000e+00], + [ 7.1675e-06, -5.1439e-05, 2.0891e-05, ..., 1.3970e-09, + 2.9534e-05, 0.0000e+00], + [-1.0990e-05, 2.1204e-05, 7.0706e-06, ..., 4.1910e-09, + -1.1688e-04, 0.0000e+00], + ..., + [ 1.3188e-05, 2.7746e-05, 2.1294e-05, ..., 2.0955e-09, + 7.2300e-05, 0.0000e+00], + [ 2.0519e-05, 1.2495e-05, 2.2307e-05, ..., 4.8894e-09, + 1.2420e-05, 0.0000e+00], + [ 8.2105e-06, 3.4180e-06, 3.6985e-05, ..., 2.7474e-08, + 1.8746e-05, 0.0000e+00]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0069, 0.0211, -0.0169, 0.0117, 0.0263, -0.0187, -0.0269, -0.0170, + -0.0199, -0.0264], device='cuda:0'), grad: tensor([-9.0361e-04, 9.0674e-06, -6.9201e-05, 6.3956e-05, 4.5085e-04, + -3.0205e-05, 7.0632e-05, 1.7738e-04, 1.0812e-04, 1.2255e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 217.06, cls_loss 0.0045 cls_loss_mapping 0.0098 cls_loss_causal 0.5931 re_mapping 0.0104 re_causal 0.0326 /// teacc 98.98 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0834, -0.0540, -0.0477, ..., -0.0189, -0.0278, -0.0133], + [-0.0165, 0.0604, -0.0520, ..., 0.0103, -0.0077, -0.0603], + [ 0.0294, -0.0708, -0.0541, ..., -0.0389, 0.0587, -0.0212], + ..., + [-0.0590, -0.0528, 0.0617, ..., 0.0466, -0.0951, 0.0054], + [ 0.0381, 0.0186, -0.1289, ..., -0.0254, 0.0780, -0.0117], + [-0.0640, -0.0563, 0.0329, ..., -0.0008, -0.0520, -0.0420]], + device='cuda:0'), grad: tensor([[ 1.6183e-05, 1.0012e-07, 3.8296e-06, ..., 0.0000e+00, + 1.8165e-05, 0.0000e+00], + [ 8.0168e-06, -5.8003e-06, 2.4121e-06, ..., 0.0000e+00, + 8.4415e-06, 0.0000e+00], + [-3.2401e-04, 3.8184e-06, -7.4685e-05, ..., 0.0000e+00, + -4.0293e-04, 0.0000e+00], + ..., + [ 9.4920e-06, 1.7537e-06, -7.7160e-07, ..., 0.0000e+00, + 1.0617e-05, 0.0000e+00], + [ 3.1680e-05, 5.2415e-06, 8.4266e-06, ..., 0.0000e+00, + 2.2978e-05, 0.0000e+00], + [-5.4501e-06, 1.2526e-06, -7.7039e-06, ..., 0.0000e+00, + 4.6752e-06, 0.0000e+00]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0071, 0.0207, -0.0166, 0.0118, 0.0265, -0.0182, -0.0271, -0.0173, + -0.0202, -0.0262], device='cuda:0'), grad: tensor([ 1.7747e-05, 1.5914e-05, -8.3303e-04, 6.8569e-04, 1.5944e-05, + 3.4064e-05, -1.3700e-06, 2.2471e-05, 8.1837e-05, -3.9726e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 216.99, cls_loss 0.0057 cls_loss_mapping 0.0113 cls_loss_causal 0.5912 re_mapping 0.0100 re_causal 0.0294 /// teacc 98.79 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0852, -0.0549, -0.0481, ..., -0.0188, -0.0289, -0.0128], + [-0.0176, 0.0605, -0.0523, ..., 0.0105, -0.0084, -0.0608], + [ 0.0300, -0.0709, -0.0540, ..., -0.0391, 0.0594, -0.0214], + ..., + [-0.0614, -0.0531, 0.0620, ..., 0.0466, -0.0971, 0.0054], + [ 0.0389, 0.0188, -0.1279, ..., -0.0254, 0.0801, -0.0119], + [-0.0647, -0.0564, 0.0334, ..., -0.0008, -0.0528, -0.0424]], + device='cuda:0'), grad: tensor([[ 1.3206e-06, 1.0908e-05, 4.5486e-06, ..., 0.0000e+00, + 3.6564e-06, -5.5879e-09], + [ 4.3437e-06, 3.4511e-05, 1.4700e-05, ..., 0.0000e+00, + 1.2793e-05, 0.0000e+00], + [ 6.3796e-07, 5.4628e-05, 9.6262e-06, ..., 0.0000e+00, + 8.7693e-06, 2.3283e-10], + ..., + [ 2.8517e-06, 7.3612e-06, 1.2912e-05, ..., 0.0000e+00, + 6.1020e-06, 0.0000e+00], + [ 1.5631e-05, 2.7895e-05, 4.9323e-05, ..., 0.0000e+00, + 3.2395e-05, 2.3283e-10], + [-6.5804e-05, -9.7692e-05, -3.6836e-04, ..., 0.0000e+00, + -1.3733e-04, 4.6566e-09]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0068, 0.0204, -0.0160, 0.0119, 0.0254, -0.0191, -0.0265, -0.0181, + -0.0188, -0.0258], device='cuda:0'), grad: tensor([ 2.0400e-05, 1.0520e-04, 1.2040e-04, 2.0728e-05, 1.6260e-04, + 1.5020e-05, 3.1090e-04, 4.1634e-05, 1.8740e-04, -9.8515e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 216.94, cls_loss 0.0061 cls_loss_mapping 0.0122 cls_loss_causal 0.5688 re_mapping 0.0099 re_causal 0.0300 /// teacc 98.85 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0862, -0.0555, -0.0484, ..., -0.0188, -0.0293, -0.0128], + [-0.0199, 0.0597, -0.0519, ..., 0.0105, -0.0095, -0.0608], + [ 0.0297, -0.0710, -0.0551, ..., -0.0392, 0.0595, -0.0214], + ..., + [-0.0618, -0.0541, 0.0625, ..., 0.0467, -0.0976, 0.0054], + [ 0.0387, 0.0188, -0.1289, ..., -0.0256, 0.0796, -0.0120], + [-0.0658, -0.0565, 0.0336, ..., -0.0008, -0.0538, -0.0424]], + device='cuda:0'), grad: tensor([[ 9.6858e-06, 4.9248e-06, 6.3330e-07, ..., 0.0000e+00, + 1.6481e-05, 0.0000e+00], + [ 1.6332e-05, 1.2957e-05, 4.0345e-06, ..., 0.0000e+00, + 3.0816e-05, 0.0000e+00], + [ 1.8515e-06, 1.2502e-05, 1.2703e-05, ..., 0.0000e+00, + 2.6956e-05, 0.0000e+00], + ..., + [ 1.8058e-06, 3.9227e-06, -9.2462e-06, ..., 0.0000e+00, + 3.0156e-06, 0.0000e+00], + [ 1.1191e-05, -3.5465e-05, 1.2312e-06, ..., 0.0000e+00, + 4.6909e-05, 0.0000e+00], + [ 5.6438e-06, 1.7330e-05, 1.2383e-05, ..., 0.0000e+00, + 1.5587e-05, 0.0000e+00]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0067, 0.0194, -0.0165, 0.0127, 0.0251, -0.0184, -0.0258, -0.0179, + -0.0193, -0.0257], device='cuda:0'), grad: tensor([ 2.9191e-05, 5.3406e-05, 8.4162e-05, -2.4128e-04, -6.9976e-05, + 5.3551e-07, 2.3991e-05, -1.1265e-05, 7.9632e-05, 5.1528e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 216.92, cls_loss 0.0050 cls_loss_mapping 0.0101 cls_loss_causal 0.6091 re_mapping 0.0099 re_causal 0.0305 /// teacc 98.90 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0884, -0.0560, -0.0483, ..., -0.0189, -0.0294, -0.0128], + [-0.0203, 0.0600, -0.0522, ..., 0.0113, -0.0098, -0.0609], + [ 0.0298, -0.0714, -0.0557, ..., -0.0393, 0.0597, -0.0214], + ..., + [-0.0627, -0.0543, 0.0630, ..., 0.0466, -0.0985, 0.0054], + [ 0.0386, 0.0186, -0.1294, ..., -0.0257, 0.0800, -0.0120], + [-0.0669, -0.0573, 0.0337, ..., -0.0009, -0.0543, -0.0424]], + device='cuda:0'), grad: tensor([[ 7.1488e-06, 4.4703e-06, 8.0280e-07, ..., 0.0000e+00, + 3.3937e-06, 0.0000e+00], + [ 5.8651e-05, -4.4227e-05, 3.9227e-06, ..., 0.0000e+00, + -1.3955e-05, 0.0000e+00], + [-3.8855e-06, 5.3912e-05, 1.1623e-06, ..., 0.0000e+00, + 2.1368e-05, 0.0000e+00], + ..., + [ 7.4506e-06, 1.9938e-05, -1.3769e-05, ..., 0.0000e+00, + 1.0721e-05, 0.0000e+00], + [-5.9319e-04, -3.5334e-04, 3.6024e-06, ..., 0.0000e+00, + -2.4867e-04, 0.0000e+00], + [ 9.5144e-06, 1.0967e-05, 5.2806e-07, ..., 0.0000e+00, + 6.3740e-06, 0.0000e+00]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0074, 0.0192, -0.0168, 0.0119, 0.0251, -0.0182, -0.0240, -0.0178, + -0.0194, -0.0258], device='cuda:0'), grad: tensor([-1.3590e-04, -4.0084e-05, 9.0539e-05, 3.1382e-05, 2.1160e-06, + -3.0808e-06, 9.0122e-04, 1.0721e-05, -8.8882e-04, 3.1769e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 217.05, cls_loss 0.0052 cls_loss_mapping 0.0106 cls_loss_causal 0.5962 re_mapping 0.0099 re_causal 0.0287 /// teacc 98.76 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0891, -0.0561, -0.0483, ..., -0.0188, -0.0292, -0.0128], + [-0.0206, 0.0601, -0.0524, ..., 0.0116, -0.0101, -0.0609], + [ 0.0302, -0.0710, -0.0563, ..., -0.0395, 0.0604, -0.0214], + ..., + [-0.0633, -0.0545, 0.0637, ..., 0.0466, -0.0991, 0.0054], + [ 0.0385, 0.0178, -0.1302, ..., -0.0261, 0.0801, -0.0120], + [-0.0677, -0.0580, 0.0336, ..., -0.0010, -0.0548, -0.0424]], + device='cuda:0'), grad: tensor([[ 4.1798e-06, 9.7826e-06, 2.1644e-06, ..., 0.0000e+00, + 4.9248e-06, 0.0000e+00], + [ 1.6928e-05, -9.7215e-05, -6.6124e-07, ..., 0.0000e+00, + 8.5831e-06, 0.0000e+00], + [-2.3991e-05, 5.6118e-05, 4.6343e-05, ..., 4.6566e-10, + -7.1585e-05, 0.0000e+00], + ..., + [ 2.6077e-06, 2.0117e-05, -1.0991e-04, ..., 0.0000e+00, + -7.2420e-05, 0.0000e+00], + [-3.9190e-05, -2.0355e-05, 9.7603e-06, ..., 0.0000e+00, + -2.1920e-05, 0.0000e+00], + [ 8.7023e-06, 4.2766e-06, 7.0147e-06, ..., 0.0000e+00, + 1.0230e-05, 0.0000e+00]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0070, 0.0189, -0.0161, 0.0122, 0.0251, -0.0182, -0.0246, -0.0175, + -0.0200, -0.0260], device='cuda:0'), grad: tensor([-5.4017e-06, -1.3459e-04, 1.0931e-04, 3.3522e-04, 3.5405e-05, + -3.8296e-06, 9.4995e-06, -4.0865e-04, -1.4737e-05, 7.8619e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 217.03, cls_loss 0.0045 cls_loss_mapping 0.0088 cls_loss_causal 0.6042 re_mapping 0.0097 re_causal 0.0292 /// teacc 98.89 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0897, -0.0572, -0.0490, ..., -0.0188, -0.0295, -0.0128], + [-0.0210, 0.0610, -0.0526, ..., 0.0109, -0.0099, -0.0609], + [ 0.0306, -0.0715, -0.0570, ..., -0.0408, 0.0609, -0.0214], + ..., + [-0.0639, -0.0546, 0.0644, ..., 0.0480, -0.0997, 0.0054], + [ 0.0387, 0.0176, -0.1307, ..., -0.0264, 0.0802, -0.0120], + [-0.0686, -0.0589, 0.0335, ..., -0.0010, -0.0558, -0.0424]], + device='cuda:0'), grad: tensor([[ 2.8498e-06, 6.4932e-06, 3.2224e-07, ..., 0.0000e+00, + 2.3376e-06, 0.0000e+00], + [ 1.6719e-05, 2.5295e-06, 1.7975e-06, ..., 0.0000e+00, + 2.0817e-05, 0.0000e+00], + [ 1.3575e-05, 1.1146e-05, 1.2359e-06, ..., 0.0000e+00, + 1.3039e-05, 0.0000e+00], + ..., + [ 1.4581e-05, 6.5826e-06, 1.8170e-06, ..., 0.0000e+00, + 2.0355e-05, 0.0000e+00], + [-3.2097e-05, -2.8357e-05, 1.2871e-06, ..., 0.0000e+00, + -2.3872e-05, 0.0000e+00], + [ 2.1737e-06, 9.0897e-06, 1.7760e-06, ..., 0.0000e+00, + 3.3602e-06, 0.0000e+00]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0076, 0.0194, -0.0162, 0.0113, 0.0249, -0.0173, -0.0249, -0.0171, + -0.0203, -0.0259], device='cuda:0'), grad: tensor([ 1.2517e-05, 4.1276e-05, 4.4376e-05, -3.7122e-04, -3.7283e-05, + 2.1195e-04, 3.6985e-05, 5.6803e-05, -1.0781e-05, 1.5706e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 217.16, cls_loss 0.0044 cls_loss_mapping 0.0083 cls_loss_causal 0.5952 re_mapping 0.0094 re_causal 0.0283 /// teacc 98.95 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0902, -0.0565, -0.0492, ..., -0.0191, -0.0290, -0.0126], + [-0.0213, 0.0611, -0.0531, ..., 0.0110, -0.0101, -0.0610], + [ 0.0310, -0.0720, -0.0574, ..., -0.0411, 0.0614, -0.0215], + ..., + [-0.0645, -0.0546, 0.0647, ..., 0.0481, -0.1004, 0.0053], + [ 0.0391, 0.0180, -0.1311, ..., -0.0265, 0.0805, -0.0120], + [-0.0695, -0.0594, 0.0340, ..., -0.0011, -0.0566, -0.0425]], + device='cuda:0'), grad: tensor([[ 5.1148e-06, 6.8992e-06, 5.1595e-06, ..., -2.2352e-08, + 6.2445e-07, 0.0000e+00], + [ 7.9349e-07, -1.4830e-04, -1.5247e-04, ..., 1.3039e-08, + -1.4633e-05, 0.0000e+00], + [ 2.4810e-06, 1.7464e-05, 1.4573e-05, ..., 2.4680e-08, + -2.6729e-07, 0.0000e+00], + ..., + [ 1.3178e-06, 1.1790e-04, 1.2839e-04, ..., 9.1735e-08, + 1.2621e-05, 0.0000e+00], + [ 5.1633e-06, 5.4836e-06, 4.9807e-06, ..., 2.9802e-08, + 1.1362e-07, 0.0000e+00], + [ 2.6412e-06, 1.0186e-04, 1.1861e-04, ..., 9.4995e-08, + 2.6058e-06, 0.0000e+00]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0069, 0.0191, -0.0163, 0.0114, 0.0251, -0.0174, -0.0254, -0.0172, + -0.0201, -0.0257], device='cuda:0'), grad: tensor([-4.1761e-06, -5.0735e-04, 6.0409e-05, 4.2655e-06, -3.4118e-04, + -1.1370e-05, 1.3232e-05, 4.2343e-04, 2.6494e-05, 3.3617e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 216.85, cls_loss 0.0047 cls_loss_mapping 0.0094 cls_loss_causal 0.5914 re_mapping 0.0098 re_causal 0.0290 /// teacc 98.99 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0909, -0.0576, -0.0496, ..., -0.0191, -0.0293, -0.0126], + [-0.0221, 0.0608, -0.0538, ..., 0.0111, -0.0109, -0.0610], + [ 0.0314, -0.0718, -0.0577, ..., -0.0411, 0.0620, -0.0215], + ..., + [-0.0648, -0.0545, 0.0648, ..., 0.0481, -0.1006, 0.0053], + [ 0.0394, 0.0181, -0.1317, ..., -0.0265, 0.0810, -0.0120], + [-0.0706, -0.0603, 0.0335, ..., -0.0011, -0.0571, -0.0425]], + device='cuda:0'), grad: tensor([[ 5.9940e-06, -1.3271e-07, 2.6450e-06, ..., 0.0000e+00, + 6.5491e-06, 0.0000e+00], + [ 3.2280e-06, -1.1902e-06, 4.1053e-06, ..., 0.0000e+00, + 5.1595e-06, 0.0000e+00], + [-7.4431e-06, 2.7083e-06, 3.6824e-06, ..., 0.0000e+00, + -1.4298e-05, 0.0000e+00], + ..., + [ 6.0489e-07, 2.2240e-06, -1.7211e-05, ..., 0.0000e+00, + 6.9216e-06, 0.0000e+00], + [ 1.3284e-05, 8.1435e-06, 6.5453e-06, ..., 0.0000e+00, + 1.5408e-05, 0.0000e+00], + [ 3.2876e-06, 4.3482e-05, 9.4175e-04, ..., 0.0000e+00, + 1.9938e-05, 0.0000e+00]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0070, 0.0183, -0.0159, 0.0124, 0.0255, -0.0167, -0.0255, -0.0179, + -0.0200, -0.0264], device='cuda:0'), grad: tensor([-3.4153e-05, 2.6420e-05, 4.0382e-06, -4.4403e-03, -2.8267e-03, + 4.3831e-03, 1.7315e-05, -3.3081e-05, 8.7917e-05, 2.8152e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 216.98, cls_loss 0.0039 cls_loss_mapping 0.0083 cls_loss_causal 0.5913 re_mapping 0.0090 re_causal 0.0292 /// teacc 98.89 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0918, -0.0584, -0.0498, ..., -0.0191, -0.0295, -0.0126], + [-0.0229, 0.0606, -0.0545, ..., 0.0115, -0.0108, -0.0610], + [ 0.0321, -0.0724, -0.0582, ..., -0.0417, 0.0623, -0.0215], + ..., + [-0.0660, -0.0539, 0.0654, ..., 0.0481, -0.1015, 0.0053], + [ 0.0399, 0.0181, -0.1321, ..., -0.0266, 0.0816, -0.0120], + [-0.0711, -0.0609, 0.0333, ..., -0.0011, -0.0575, -0.0425]], + device='cuda:0'), grad: tensor([[ 2.4382e-06, 5.7071e-06, 8.2403e-06, ..., 0.0000e+00, + -4.4238e-07, 0.0000e+00], + [ 1.1008e-06, 5.2786e-04, 6.9809e-04, ..., 0.0000e+00, + 6.2957e-07, 0.0000e+00], + [-6.9663e-06, 1.5214e-05, 1.9282e-05, ..., 0.0000e+00, + -1.2934e-05, 0.0000e+00], + ..., + [ 1.6494e-06, 1.8328e-05, 2.3693e-05, ..., 0.0000e+00, + 2.3190e-06, 0.0000e+00], + [ 5.8040e-06, 1.5393e-05, 2.0459e-05, ..., 0.0000e+00, + 5.6289e-06, 0.0000e+00], + [ 6.7148e-07, -9.9182e-04, -1.3447e-03, ..., 0.0000e+00, + 5.7509e-07, 0.0000e+00]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0070, 0.0180, -0.0160, 0.0125, 0.0261, -0.0170, -0.0253, -0.0177, + -0.0198, -0.0268], device='cuda:0'), grad: tensor([ 6.8322e-06, 2.0885e-03, 4.5210e-05, 7.4625e-04, 9.0647e-04, + 1.7479e-05, 9.4026e-06, 7.4863e-05, 7.2658e-05, -3.9673e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 217.03, cls_loss 0.0039 cls_loss_mapping 0.0089 cls_loss_causal 0.6023 re_mapping 0.0088 re_causal 0.0290 /// teacc 99.00 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0925, -0.0587, -0.0501, ..., -0.0191, -0.0296, -0.0126], + [-0.0235, 0.0608, -0.0548, ..., 0.0115, -0.0109, -0.0610], + [ 0.0326, -0.0728, -0.0592, ..., -0.0417, 0.0630, -0.0215], + ..., + [-0.0673, -0.0542, 0.0659, ..., 0.0481, -0.1026, 0.0053], + [ 0.0409, 0.0186, -0.1323, ..., -0.0266, 0.0824, -0.0121], + [-0.0717, -0.0605, 0.0338, ..., -0.0011, -0.0580, -0.0425]], + device='cuda:0'), grad: tensor([[ 2.6114e-06, 1.0878e-06, 2.4354e-07, ..., 0.0000e+00, + 9.8720e-07, 0.0000e+00], + [ 1.9558e-06, -1.7479e-05, 2.3609e-07, ..., 0.0000e+00, + 3.3602e-06, 0.0000e+00], + [-1.1973e-05, 5.3123e-06, 1.3318e-07, ..., 0.0000e+00, + -1.7747e-05, 0.0000e+00], + ..., + [ 1.6000e-06, 2.7083e-06, 1.3404e-05, ..., 0.0000e+00, + 3.0547e-06, 0.0000e+00], + [ 9.4593e-05, 4.1246e-05, 6.8694e-06, ..., 0.0000e+00, + 1.0775e-06, 0.0000e+00], + [ 6.9058e-07, 4.3027e-07, -2.4527e-05, ..., 0.0000e+00, + 3.5530e-07, 0.0000e+00]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0069, 0.0176, -0.0163, 0.0121, 0.0260, -0.0168, -0.0263, -0.0176, + -0.0190, -0.0262], device='cuda:0'), grad: tensor([-2.6822e-07, -1.5087e-05, -2.9832e-05, 6.7018e-06, 1.4223e-05, + 5.2415e-06, -9.1553e-05, 4.0740e-05, 1.2231e-04, -5.2333e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 216.96, cls_loss 0.0047 cls_loss_mapping 0.0101 cls_loss_causal 0.5838 re_mapping 0.0085 re_causal 0.0270 /// teacc 98.85 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0933, -0.0594, -0.0501, ..., -0.0191, -0.0300, -0.0125], + [-0.0239, 0.0613, -0.0550, ..., 0.0116, -0.0114, -0.0612], + [ 0.0329, -0.0731, -0.0596, ..., -0.0417, 0.0638, -0.0217], + ..., + [-0.0676, -0.0547, 0.0660, ..., 0.0481, -0.1033, 0.0053], + [ 0.0412, 0.0186, -0.1331, ..., -0.0266, 0.0826, -0.0121], + [-0.0721, -0.0599, 0.0344, ..., -0.0011, -0.0574, -0.0426]], + device='cuda:0'), grad: tensor([[ 1.4585e-06, 5.3644e-07, 3.2708e-06, ..., 0.0000e+00, + 1.2098e-06, 0.0000e+00], + [ 2.5425e-07, -1.1269e-06, 1.7881e-06, ..., 0.0000e+00, + 1.2629e-06, 0.0000e+00], + [-9.8255e-07, 4.6892e-07, 1.2731e-06, ..., 0.0000e+00, + 3.4254e-06, 0.0000e+00], + ..., + [ 7.8743e-07, 2.5798e-07, -5.3495e-06, ..., 0.0000e+00, + 1.5683e-06, 0.0000e+00], + [ 1.6000e-06, 8.8895e-07, 4.7572e-06, ..., 0.0000e+00, + 1.9260e-06, 0.0000e+00], + [ 1.4696e-06, 3.4366e-07, 5.2713e-07, ..., 0.0000e+00, + 2.3516e-07, 0.0000e+00]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0070, 0.0177, -0.0159, 0.0123, 0.0257, -0.0175, -0.0263, -0.0182, + -0.0194, -0.0249], device='cuda:0'), grad: tensor([ 1.9968e-05, 8.0168e-06, 1.8477e-05, -1.9148e-05, -5.1558e-05, + -7.3910e-06, 2.9169e-06, -4.0326e-07, 3.2216e-05, -3.0473e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 217.05, cls_loss 0.0038 cls_loss_mapping 0.0075 cls_loss_causal 0.6087 re_mapping 0.0093 re_causal 0.0289 /// teacc 98.79 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0938, -0.0601, -0.0493, ..., -0.0191, -0.0293, -0.0125], + [-0.0242, 0.0621, -0.0547, ..., 0.0116, -0.0115, -0.0613], + [ 0.0328, -0.0739, -0.0597, ..., -0.0417, 0.0637, -0.0217], + ..., + [-0.0681, -0.0553, 0.0662, ..., 0.0481, -0.1039, 0.0053], + [ 0.0413, 0.0190, -0.1336, ..., -0.0266, 0.0832, -0.0121], + [-0.0732, -0.0614, 0.0339, ..., -0.0011, -0.0596, -0.0427]], + device='cuda:0'), grad: tensor([[ 1.1958e-06, 1.0915e-06, 1.5795e-06, ..., 0.0000e+00, + -1.0096e-06, 0.0000e+00], + [ 2.9095e-06, -2.2277e-05, 9.4995e-06, ..., 0.0000e+00, + 3.0212e-06, 0.0000e+00], + [-1.2994e-05, -8.3633e-07, 5.6028e-06, ..., 0.0000e+00, + -1.2986e-05, 0.0000e+00], + ..., + [ 7.5400e-06, 5.9381e-06, -2.6211e-05, ..., 0.0000e+00, + 1.0118e-05, 0.0000e+00], + [ 1.1921e-07, 7.0781e-06, 5.6513e-06, ..., 0.0000e+00, + -4.1444e-08, 0.0000e+00], + [ 2.3823e-06, 1.7911e-05, 2.4557e-05, ..., 0.0000e+00, + 3.3285e-06, 0.0000e+00]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0064, 0.0182, -0.0166, 0.0126, 0.0260, -0.0170, -0.0263, -0.0183, + -0.0194, -0.0259], device='cuda:0'), grad: tensor([-8.4490e-06, -2.2557e-06, -2.5615e-05, -3.8415e-05, -3.8683e-05, + 1.9655e-05, 2.5839e-05, -2.2545e-05, 2.3723e-05, 6.6698e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 217.09, cls_loss 0.0052 cls_loss_mapping 0.0092 cls_loss_causal 0.5890 re_mapping 0.0091 re_causal 0.0270 /// teacc 98.87 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0946, -0.0604, -0.0497, ..., -0.0194, -0.0299, -0.0125], + [-0.0249, 0.0627, -0.0541, ..., 0.0119, -0.0117, -0.0613], + [ 0.0333, -0.0743, -0.0599, ..., -0.0418, 0.0643, -0.0217], + ..., + [-0.0683, -0.0564, 0.0674, ..., 0.0480, -0.1041, 0.0053], + [ 0.0418, 0.0188, -0.1344, ..., -0.0267, 0.0841, -0.0121], + [-0.0755, -0.0608, 0.0332, ..., -0.0012, -0.0617, -0.0427]], + device='cuda:0'), grad: tensor([[ 3.6974e-06, 3.0305e-06, 1.4501e-06, ..., 0.0000e+00, + 1.2089e-06, 0.0000e+00], + [ 8.7079e-07, -1.2606e-05, 7.5772e-06, ..., 0.0000e+00, + 3.1758e-06, 0.0000e+00], + [ 4.4070e-06, 9.8944e-06, 1.5453e-05, ..., 0.0000e+00, + 3.9116e-06, 0.0000e+00], + ..., + [-8.4564e-07, 5.6960e-06, -2.8148e-05, ..., 0.0000e+00, + 2.0526e-06, 0.0000e+00], + [ 1.0803e-06, 3.2455e-05, 5.4166e-06, ..., 0.0000e+00, + 1.2636e-05, 0.0000e+00], + [ 4.1462e-06, 8.1658e-05, 1.4514e-05, ..., 0.0000e+00, + 3.9548e-05, 0.0000e+00]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0064, 0.0185, -0.0166, 0.0122, 0.0249, -0.0167, -0.0267, -0.0169, + -0.0194, -0.0265], device='cuda:0'), grad: tensor([ 1.7062e-05, 6.9320e-05, 1.8954e-04, -6.8045e-04, -3.4451e-04, + 6.2847e-04, 4.1604e-05, -2.9850e-04, 1.1313e-04, 2.6464e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 217.05, cls_loss 0.0052 cls_loss_mapping 0.0084 cls_loss_causal 0.5857 re_mapping 0.0098 re_causal 0.0283 /// teacc 98.93 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0954, -0.0622, -0.0500, ..., -0.0194, -0.0301, -0.0124], + [-0.0249, 0.0639, -0.0543, ..., 0.0120, -0.0114, -0.0614], + [ 0.0327, -0.0748, -0.0602, ..., -0.0418, 0.0642, -0.0218], + ..., + [-0.0694, -0.0569, 0.0675, ..., 0.0480, -0.1052, 0.0053], + [ 0.0424, 0.0182, -0.1349, ..., -0.0268, 0.0852, -0.0121], + [-0.0765, -0.0614, 0.0329, ..., -0.0012, -0.0623, -0.0427]], + device='cuda:0'), grad: tensor([[ 8.1658e-06, -2.7437e-06, -4.1304e-07, ..., 0.0000e+00, + 1.7732e-06, 0.0000e+00], + [ 1.9684e-05, 9.9689e-06, 7.2364e-07, ..., 0.0000e+00, + 2.1279e-05, 0.0000e+00], + [-4.7356e-05, -5.4032e-05, 3.5483e-07, ..., 0.0000e+00, + -4.5925e-05, 0.0000e+00], + ..., + [ 1.3132e-06, 5.1335e-06, -2.2464e-06, ..., 0.0000e+00, + 1.9521e-06, 0.0000e+00], + [-1.6257e-05, -7.7719e-07, 1.3504e-07, ..., 0.0000e+00, + -1.3441e-05, 0.0000e+00], + [ 1.7565e-06, 3.5074e-06, 7.8743e-07, ..., 0.0000e+00, + 2.8964e-06, 0.0000e+00]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0072, 0.0188, -0.0173, 0.0131, 0.0249, -0.0173, -0.0263, -0.0164, + -0.0195, -0.0268], device='cuda:0'), grad: tensor([-1.4775e-05, 5.4032e-05, -1.4961e-04, 3.4254e-06, 1.8641e-05, + 6.4857e-06, 8.1480e-05, 5.6885e-06, -1.9923e-05, 1.4625e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 216.67, cls_loss 0.0032 cls_loss_mapping 0.0082 cls_loss_causal 0.6091 re_mapping 0.0087 re_causal 0.0283 /// teacc 98.95 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0961, -0.0623, -0.0502, ..., -0.0194, -0.0301, -0.0124], + [-0.0252, 0.0642, -0.0544, ..., 0.0120, -0.0116, -0.0614], + [ 0.0330, -0.0753, -0.0605, ..., -0.0418, 0.0648, -0.0218], + ..., + [-0.0696, -0.0574, 0.0674, ..., 0.0480, -0.1056, 0.0053], + [ 0.0425, 0.0187, -0.1354, ..., -0.0268, 0.0858, -0.0121], + [-0.0776, -0.0624, 0.0334, ..., -0.0012, -0.0643, -0.0427]], + device='cuda:0'), grad: tensor([[ 5.5879e-07, 5.0757e-07, 5.5088e-07, ..., 0.0000e+00, + 8.3074e-07, 0.0000e+00], + [ 3.7020e-07, -7.8008e-06, 2.0862e-06, ..., 0.0000e+00, + -1.7881e-06, 0.0000e+00], + [-1.2308e-05, 4.8503e-06, 8.3633e-07, ..., 0.0000e+00, + -1.4000e-05, 0.0000e+00], + ..., + [ 1.2806e-06, -1.9418e-07, -6.7502e-06, ..., 0.0000e+00, + 2.1849e-06, 0.0000e+00], + [ 9.2238e-06, 7.6368e-06, 1.4137e-06, ..., 0.0000e+00, + 1.4648e-05, 0.0000e+00], + [ 3.2270e-07, 2.3581e-06, -7.6219e-06, ..., 0.0000e+00, + 1.5581e-06, 0.0000e+00]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0069, 0.0189, -0.0172, 0.0124, 0.0251, -0.0168, -0.0264, -0.0167, + -0.0194, -0.0268], device='cuda:0'), grad: tensor([ 3.5353e-06, 2.4922e-06, -1.1377e-05, -3.8967e-06, 1.0878e-05, + 3.1888e-05, 5.1409e-06, 5.2661e-05, 3.5554e-05, -1.2684e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 216.70, cls_loss 0.0032 cls_loss_mapping 0.0072 cls_loss_causal 0.6177 re_mapping 0.0086 re_causal 0.0269 /// teacc 98.98 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0974, -0.0630, -0.0505, ..., -0.0193, -0.0303, -0.0124], + [-0.0256, 0.0643, -0.0543, ..., 0.0136, -0.0121, -0.0614], + [ 0.0334, -0.0746, -0.0609, ..., -0.0439, 0.0659, -0.0218], + ..., + [-0.0701, -0.0578, 0.0677, ..., 0.0479, -0.1063, 0.0053], + [ 0.0425, 0.0184, -0.1360, ..., -0.0269, 0.0858, -0.0121], + [-0.0780, -0.0627, 0.0335, ..., -0.0012, -0.0650, -0.0427]], + device='cuda:0'), grad: tensor([[ 2.7325e-06, -3.0939e-06, -1.0058e-07, ..., 0.0000e+00, + -1.0282e-05, 0.0000e+00], + [ 4.2934e-07, -1.7285e-05, 6.2631e-07, ..., 0.0000e+00, + 4.2357e-06, 0.0000e+00], + [-8.2236e-07, 4.2059e-06, 9.4716e-07, ..., 0.0000e+00, + -2.3236e-07, 0.0000e+00], + ..., + [ 4.6752e-07, 3.3509e-06, -7.4990e-06, ..., 0.0000e+00, + 1.3635e-06, 0.0000e+00], + [-1.2860e-05, 2.5947e-06, 6.6496e-07, ..., 0.0000e+00, + -1.5318e-05, 0.0000e+00], + [ 3.5986e-06, 2.2408e-06, 3.4999e-06, ..., 0.0000e+00, + 5.8301e-06, 0.0000e+00]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0071, 0.0188, -0.0163, 0.0126, 0.0250, -0.0170, -0.0260, -0.0168, + -0.0199, -0.0269], device='cuda:0'), grad: tensor([-5.4181e-05, -2.0359e-06, 8.9705e-06, 1.0625e-05, 3.6657e-06, + 8.9705e-06, 2.3291e-05, -5.4911e-06, -1.4499e-05, 2.0534e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 216.67, cls_loss 0.0051 cls_loss_mapping 0.0106 cls_loss_causal 0.6081 re_mapping 0.0088 re_causal 0.0266 /// teacc 98.84 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0981, -0.0635, -0.0513, ..., -0.0193, -0.0320, -0.0124], + [-0.0259, 0.0644, -0.0549, ..., 0.0136, -0.0133, -0.0614], + [ 0.0332, -0.0741, -0.0613, ..., -0.0440, 0.0669, -0.0218], + ..., + [-0.0705, -0.0578, 0.0682, ..., 0.0479, -0.1066, 0.0053], + [ 0.0439, 0.0182, -0.1368, ..., -0.0269, 0.0873, -0.0121], + [-0.0791, -0.0635, 0.0338, ..., -0.0012, -0.0649, -0.0427]], + device='cuda:0'), grad: tensor([[ 3.2578e-06, 2.5798e-07, 1.8021e-07, ..., 0.0000e+00, + 4.1872e-06, 0.0000e+00], + [ 1.7257e-06, -3.4720e-06, 6.3283e-07, ..., 0.0000e+00, + 2.1607e-06, 0.0000e+00], + [-4.0084e-05, 1.2591e-06, -2.8731e-07, ..., 0.0000e+00, + -7.8619e-05, 0.0000e+00], + ..., + [ 6.4597e-06, 5.6578e-07, 1.2480e-07, ..., 0.0000e+00, + 9.1195e-06, 0.0000e+00], + [ 2.2367e-05, 1.2703e-06, 7.7859e-07, ..., 0.0000e+00, + 1.9535e-05, 0.0000e+00], + [ 7.4245e-06, 1.1558e-06, -6.1430e-06, ..., 0.0000e+00, + 1.4761e-06, 0.0000e+00]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0085, 0.0177, -0.0151, 0.0128, 0.0247, -0.0175, -0.0262, -0.0167, + -0.0192, -0.0262], device='cuda:0'), grad: tensor([-3.5197e-05, 4.0866e-06, -1.1241e-04, 7.4029e-05, 1.3508e-05, + -6.9439e-05, 1.8418e-05, 2.6777e-05, 7.2718e-05, 7.4692e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 216.91, cls_loss 0.0043 cls_loss_mapping 0.0075 cls_loss_causal 0.5661 re_mapping 0.0091 re_causal 0.0259 /// teacc 98.90 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0989, -0.0623, -0.0516, ..., -0.0192, -0.0322, -0.0124], + [-0.0261, 0.0652, -0.0550, ..., 0.0136, -0.0135, -0.0614], + [ 0.0331, -0.0753, -0.0616, ..., -0.0440, 0.0668, -0.0218], + ..., + [-0.0711, -0.0584, 0.0688, ..., 0.0479, -0.1071, 0.0053], + [ 0.0439, 0.0183, -0.1374, ..., -0.0270, 0.0876, -0.0121], + [-0.0802, -0.0652, 0.0335, ..., -0.0012, -0.0655, -0.0427]], + device='cuda:0'), grad: tensor([[ 6.8871e-07, 7.3854e-07, 3.9600e-06, ..., 1.1493e-06, + 1.3262e-06, 0.0000e+00], + [ 7.5391e-07, -2.1746e-07, 5.5544e-06, ..., 1.4938e-06, + 2.3283e-06, 0.0000e+00], + [-7.3714e-07, 1.7779e-06, 4.4376e-05, ..., 1.2822e-05, + 1.0848e-05, 0.0000e+00], + ..., + [ 3.3248e-07, -1.9092e-06, -9.9480e-05, ..., -2.7791e-05, + 1.9804e-05, 0.0000e+00], + [-8.9034e-06, -6.8024e-06, 2.7027e-06, ..., 5.2340e-07, + -3.9414e-06, 0.0000e+00], + [ 1.3122e-06, 2.2762e-06, 1.5333e-05, ..., 9.4436e-07, + 4.1947e-06, 0.0000e+00]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0077, 0.0183, -0.0159, 0.0130, 0.0249, -0.0170, -0.0266, -0.0166, + -0.0198, -0.0267], device='cuda:0'), grad: tensor([ 2.6464e-05, 4.7296e-05, 2.8014e-04, -2.4867e-04, 7.4446e-05, + 6.2943e-05, 1.5303e-05, -3.6430e-04, 2.1979e-05, 8.4162e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 216.63, cls_loss 0.0041 cls_loss_mapping 0.0075 cls_loss_causal 0.6123 re_mapping 0.0087 re_causal 0.0261 /// teacc 98.91 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0999, -0.0627, -0.0518, ..., -0.0191, -0.0323, -0.0124], + [-0.0263, 0.0651, -0.0554, ..., 0.0136, -0.0142, -0.0615], + [ 0.0333, -0.0759, -0.0619, ..., -0.0441, 0.0673, -0.0218], + ..., + [-0.0714, -0.0577, 0.0696, ..., 0.0480, -0.1066, 0.0053], + [ 0.0441, 0.0186, -0.1380, ..., -0.0271, 0.0882, -0.0122], + [-0.0810, -0.0655, 0.0334, ..., -0.0012, -0.0662, -0.0427]], + device='cuda:0'), grad: tensor([[ 1.5674e-06, 7.3481e-07, 2.1234e-07, ..., 0.0000e+00, + 8.7917e-06, 0.0000e+00], + [ 1.4175e-06, -7.3239e-06, 1.0505e-06, ..., 0.0000e+00, + 6.2808e-06, 0.0000e+00], + [-1.3307e-05, 1.4737e-05, 5.2489e-06, ..., 0.0000e+00, + -1.2922e-04, 0.0000e+00], + ..., + [ 2.0862e-06, 7.8529e-06, 4.5970e-06, ..., 0.0000e+00, + 6.1095e-06, 0.0000e+00], + [ 9.4948e-07, 8.3633e-07, 1.4603e-06, ..., 0.0000e+00, + 6.3106e-06, 0.0000e+00], + [ 1.0040e-06, 4.1425e-06, -9.0152e-07, ..., 0.0000e+00, + -7.2792e-06, 0.0000e+00]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0077, 0.0175, -0.0161, 0.0133, 0.0244, -0.0173, -0.0262, -0.0155, + -0.0198, -0.0269], device='cuda:0'), grad: tensor([-1.8024e-04, 1.2547e-05, -1.2290e-04, 1.5330e-04, -1.1593e-05, + 8.1658e-05, 5.2117e-06, 4.6372e-05, 3.4422e-05, -1.8895e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 216.93, cls_loss 0.0040 cls_loss_mapping 0.0078 cls_loss_causal 0.5769 re_mapping 0.0085 re_causal 0.0259 /// teacc 99.00 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.1004, -0.0632, -0.0521, ..., -0.0190, -0.0330, -0.0124], + [-0.0268, 0.0656, -0.0556, ..., 0.0136, -0.0143, -0.0615], + [ 0.0341, -0.0761, -0.0617, ..., -0.0441, 0.0684, -0.0218], + ..., + [-0.0718, -0.0580, 0.0698, ..., 0.0480, -0.1077, 0.0053], + [ 0.0451, 0.0197, -0.1385, ..., -0.0271, 0.0891, -0.0122], + [-0.0815, -0.0662, 0.0334, ..., -0.0013, -0.0672, -0.0427]], + device='cuda:0'), grad: tensor([[ 1.8943e-06, 2.0452e-06, 1.9930e-07, ..., 0.0000e+00, + 1.1204e-06, 0.0000e+00], + [ 1.3541e-06, -9.5740e-06, 2.3749e-08, ..., 0.0000e+00, + -1.4929e-06, 0.0000e+00], + [-2.2538e-07, 5.7369e-06, 8.9593e-07, ..., 0.0000e+00, + 2.6561e-06, 0.0000e+00], + ..., + [ 7.5484e-07, 3.5409e-06, -9.9838e-07, ..., 0.0000e+00, + 1.1567e-06, 0.0000e+00], + [ 4.2319e-04, 2.5773e-04, 6.1467e-07, ..., 0.0000e+00, + 6.9022e-05, 0.0000e+00], + [ 3.4068e-06, 2.3711e-06, -6.0117e-07, ..., 0.0000e+00, + 4.2953e-06, 0.0000e+00]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0080, 0.0177, -0.0154, 0.0130, 0.0240, -0.0169, -0.0270, -0.0157, + -0.0190, -0.0271], device='cuda:0'), grad: tensor([ 6.1579e-06, -1.2524e-05, 1.1936e-05, -1.1483e-06, -6.2585e-06, + -1.0267e-05, -4.6635e-04, 5.5917e-06, 4.6468e-04, 7.1861e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 101---------------------------------------------------- +epoch 101, time 217.60, cls_loss 0.0030 cls_loss_mapping 0.0082 cls_loss_causal 0.5605 re_mapping 0.0086 re_causal 0.0259 /// teacc 99.06 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.1015, -0.0637, -0.0515, ..., -0.0188, -0.0329, -0.0123], + [-0.0272, 0.0660, -0.0560, ..., 0.0136, -0.0143, -0.0615], + [ 0.0348, -0.0766, -0.0617, ..., -0.0441, 0.0690, -0.0218], + ..., + [-0.0730, -0.0584, 0.0699, ..., 0.0480, -0.1089, 0.0053], + [ 0.0451, 0.0192, -0.1395, ..., -0.0272, 0.0893, -0.0122], + [-0.0820, -0.0663, 0.0333, ..., -0.0013, -0.0678, -0.0427]], + device='cuda:0'), grad: tensor([[ 7.7533e-07, 9.9838e-07, 1.1981e-05, ..., 0.0000e+00, + 2.6688e-05, 0.0000e+00], + [ 1.6447e-06, 2.4755e-06, 7.4916e-06, ..., 0.0000e+00, + 2.0131e-05, 0.0000e+00], + [-3.5726e-06, 4.9062e-06, -6.8724e-05, ..., 0.0000e+00, + -1.6153e-04, 0.0000e+00], + ..., + [ 1.2480e-07, 1.3672e-06, 8.7619e-06, ..., 0.0000e+00, + 2.1979e-05, 0.0000e+00], + [-3.9116e-08, 1.0896e-06, 6.3665e-06, ..., 0.0000e+00, + 1.1012e-05, 0.0000e+00], + [ 8.6613e-07, 2.2426e-06, 3.0734e-06, ..., 0.0000e+00, + 1.0923e-05, 0.0000e+00]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0078, 0.0177, -0.0151, 0.0129, 0.0242, -0.0171, -0.0264, -0.0160, + -0.0195, -0.0270], device='cuda:0'), grad: tensor([ 6.2823e-05, 4.9680e-05, -4.0531e-04, 6.2525e-05, 2.5228e-05, + 2.3901e-05, 7.0572e-05, 6.1989e-05, 4.9353e-05, -5.8860e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 217.29, cls_loss 0.0028 cls_loss_mapping 0.0054 cls_loss_causal 0.5608 re_mapping 0.0081 re_causal 0.0258 /// teacc 98.97 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.1018, -0.0638, -0.0527, ..., -0.0186, -0.0332, -0.0123], + [-0.0280, 0.0672, -0.0548, ..., 0.0136, -0.0145, -0.0615], + [ 0.0353, -0.0769, -0.0618, ..., -0.0442, 0.0695, -0.0218], + ..., + [-0.0733, -0.0601, 0.0695, ..., 0.0480, -0.1093, 0.0053], + [ 0.0452, 0.0195, -0.1399, ..., -0.0272, 0.0895, -0.0122], + [-0.0827, -0.0667, 0.0335, ..., -0.0013, -0.0685, -0.0428]], + device='cuda:0'), grad: tensor([[ 1.8515e-06, 9.8627e-07, 4.1118e-07, ..., 0.0000e+00, + 7.2783e-07, 1.7276e-07], + [ 6.0815e-07, 4.9174e-07, 1.6540e-06, ..., 0.0000e+00, + 5.2340e-07, 3.8650e-08], + [-4.9826e-08, 3.5670e-07, 3.9674e-07, ..., 0.0000e+00, + -5.4948e-08, 4.0047e-08], + ..., + [ 8.2003e-07, 3.5949e-07, -3.9637e-06, ..., 0.0000e+00, + 1.1073e-06, 4.7032e-08], + [ 6.8657e-06, 4.0345e-06, 1.8459e-06, ..., 0.0000e+00, + 6.2818e-07, 8.4657e-07], + [ 6.5416e-06, 4.1574e-06, -1.1511e-06, ..., 0.0000e+00, + 2.3171e-06, 5.9558e-07]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0080, 0.0186, -0.0150, 0.0124, 0.0243, -0.0168, -0.0266, -0.0168, + -0.0195, -0.0268], device='cuda:0'), grad: tensor([ 2.4196e-06, 5.1297e-06, 2.2519e-06, -8.8811e-06, 2.6226e-06, + -7.2896e-05, 4.7952e-05, -1.7835e-06, 1.6868e-05, 6.2548e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 217.12, cls_loss 0.0038 cls_loss_mapping 0.0093 cls_loss_causal 0.5458 re_mapping 0.0085 re_causal 0.0246 /// teacc 98.98 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.1026, -0.0641, -0.0530, ..., -0.0185, -0.0332, -0.0122], + [-0.0275, 0.0684, -0.0550, ..., 0.0136, -0.0137, -0.0616], + [ 0.0358, -0.0778, -0.0620, ..., -0.0442, 0.0698, -0.0219], + ..., + [-0.0748, -0.0606, 0.0707, ..., 0.0480, -0.1106, 0.0053], + [ 0.0454, 0.0192, -0.1401, ..., -0.0273, 0.0898, -0.0124], + [-0.0839, -0.0676, 0.0319, ..., -0.0013, -0.0696, -0.0430]], + device='cuda:0'), grad: tensor([[ 3.3434e-06, 5.8003e-06, 6.1691e-06, ..., 0.0000e+00, + 1.9185e-06, 0.0000e+00], + [ 1.1083e-06, -8.4519e-05, -3.2634e-06, ..., 0.0000e+00, + -5.5462e-05, 0.0000e+00], + [ 4.3362e-06, 2.1338e-05, 3.7581e-05, ..., 0.0000e+00, + 3.3110e-05, 0.0000e+00], + ..., + [-5.2825e-06, 4.0531e-05, -1.0967e-04, ..., 0.0000e+00, + 1.1295e-05, 0.0000e+00], + [ 4.4256e-06, 6.6534e-06, 1.1623e-05, ..., 0.0000e+00, + 1.5106e-06, 0.0000e+00], + [ 1.5590e-06, 5.9046e-06, -1.0595e-05, ..., 0.0000e+00, + 4.2375e-08, 0.0000e+00]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0075, 0.0194, -0.0151, 0.0127, 0.0249, -0.0171, -0.0268, -0.0165, + -0.0196, -0.0280], device='cuda:0'), grad: tensor([ 1.6570e-05, -2.1672e-04, 1.8156e-04, 1.3626e-04, 4.3362e-05, + 1.7211e-06, -2.2307e-05, -1.7893e-04, 4.8518e-05, -1.0334e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 217.03, cls_loss 0.0031 cls_loss_mapping 0.0072 cls_loss_causal 0.5958 re_mapping 0.0085 re_causal 0.0266 /// teacc 98.92 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.1031, -0.0643, -0.0535, ..., -0.0185, -0.0334, -0.0121], + [-0.0281, 0.0686, -0.0552, ..., 0.0136, -0.0138, -0.0619], + [ 0.0366, -0.0782, -0.0623, ..., -0.0442, 0.0705, -0.0221], + ..., + [-0.0758, -0.0608, 0.0714, ..., 0.0480, -0.1119, 0.0053], + [ 0.0456, 0.0195, -0.1411, ..., -0.0274, 0.0897, -0.0125], + [-0.0844, -0.0684, 0.0318, ..., -0.0013, -0.0699, -0.0434]], + device='cuda:0'), grad: tensor([[ 5.2527e-07, 6.7335e-07, 3.5437e-07, ..., 0.0000e+00, + 1.3504e-07, 0.0000e+00], + [-3.5800e-06, -1.7226e-05, 1.2303e-06, ..., 0.0000e+00, + -1.9353e-06, 0.0000e+00], + [ 1.6708e-06, 1.5907e-06, 5.4110e-07, ..., 0.0000e+00, + 1.3039e-06, 0.0000e+00], + ..., + [ 5.5600e-07, 2.2948e-06, 5.2936e-06, ..., 0.0000e+00, + 3.6974e-07, 0.0000e+00], + [-3.5344e-07, 8.7842e-06, 1.0237e-05, ..., 0.0000e+00, + -2.1309e-06, 0.0000e+00], + [ 5.7463e-07, 1.3307e-05, -4.1604e-05, ..., 0.0000e+00, + 5.7742e-07, 0.0000e+00]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0072, 0.0193, -0.0148, 0.0129, 0.0250, -0.0174, -0.0269, -0.0163, + -0.0200, -0.0282], device='cuda:0'), grad: tensor([ 1.6820e-06, -2.0638e-05, 5.4240e-06, -1.7822e-05, -9.2462e-06, + 8.0705e-05, 1.4976e-06, 2.2843e-05, 3.5256e-05, -9.9838e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 216.99, cls_loss 0.0048 cls_loss_mapping 0.0097 cls_loss_causal 0.5687 re_mapping 0.0085 re_causal 0.0241 /// teacc 98.97 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.1045, -0.0649, -0.0540, ..., -0.0185, -0.0333, -0.0120], + [-0.0300, 0.0693, -0.0550, ..., 0.0136, -0.0155, -0.0619], + [ 0.0372, -0.0785, -0.0628, ..., -0.0442, 0.0715, -0.0221], + ..., + [-0.0761, -0.0621, 0.0719, ..., 0.0480, -0.1114, 0.0052], + [ 0.0457, 0.0197, -0.1419, ..., -0.0274, 0.0903, -0.0126], + [-0.0851, -0.0692, 0.0317, ..., -0.0013, -0.0704, -0.0434]], + device='cuda:0'), grad: tensor([[ 5.8189e-06, 1.6168e-06, 2.6636e-07, ..., 0.0000e+00, + 3.9190e-06, 0.0000e+00], + [ 1.1651e-06, -1.0371e-05, 1.8496e-06, ..., 0.0000e+00, + 1.6764e-06, 0.0000e+00], + [-2.5526e-05, 1.8794e-06, -3.6391e-07, ..., 0.0000e+00, + -2.2113e-05, 0.0000e+00], + ..., + [ 1.5981e-06, 1.1109e-05, -2.5681e-07, ..., 0.0000e+00, + 2.5816e-06, 0.0000e+00], + [ 6.0275e-06, -3.1665e-06, 7.5437e-07, ..., 0.0000e+00, + -2.0210e-06, 0.0000e+00], + [ 6.8247e-06, 3.6862e-06, -5.8860e-06, ..., 0.0000e+00, + 8.2627e-06, 0.0000e+00]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0071, 0.0195, -0.0148, 0.0128, 0.0248, -0.0174, -0.0265, -0.0162, + -0.0201, -0.0284], device='cuda:0'), grad: tensor([ 1.2696e-05, -1.1459e-05, -5.0366e-05, 9.9987e-06, 1.2647e-06, + 5.7220e-05, -5.7928e-06, 2.2113e-05, -2.2375e-07, -3.5405e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 216.87, cls_loss 0.0031 cls_loss_mapping 0.0067 cls_loss_causal 0.5814 re_mapping 0.0087 re_causal 0.0257 /// teacc 99.05 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.1055, -0.0656, -0.0543, ..., -0.0186, -0.0335, -0.0118], + [-0.0309, 0.0699, -0.0555, ..., 0.0136, -0.0162, -0.0621], + [ 0.0378, -0.0787, -0.0620, ..., -0.0442, 0.0729, -0.0224], + ..., + [-0.0769, -0.0628, 0.0721, ..., 0.0480, -0.1129, 0.0052], + [ 0.0461, 0.0199, -0.1425, ..., -0.0274, 0.0908, -0.0127], + [-0.0858, -0.0696, 0.0317, ..., -0.0013, -0.0710, -0.0435]], + device='cuda:0'), grad: tensor([[ 2.1458e-06, 1.3392e-06, 2.8592e-06, ..., 1.6298e-09, + 6.7428e-07, 0.0000e+00], + [ 2.9970e-06, 2.2650e-06, 1.5348e-06, ..., 9.3132e-10, + 3.8669e-06, 0.0000e+00], + [ 5.3225e-07, 5.8254e-07, 7.2308e-06, ..., 4.6566e-09, + 4.1327e-07, 0.0000e+00], + ..., + [ 1.0235e-06, 9.7882e-07, -7.2300e-05, ..., 2.5611e-09, + 5.8208e-07, 0.0000e+00], + [-2.8238e-06, -7.2494e-06, 3.8482e-06, ..., 1.8626e-09, + -1.0163e-05, 0.0000e+00], + [ 2.9262e-06, 9.3281e-06, 5.9158e-05, ..., 1.8626e-09, + 3.1665e-06, 0.0000e+00]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0072, 0.0190, -0.0132, 0.0130, 0.0247, -0.0177, -0.0262, -0.0167, + -0.0201, -0.0285], device='cuda:0'), grad: tensor([ 1.1906e-05, 1.2636e-05, 2.0683e-05, 1.8492e-05, -7.5102e-06, + -9.9838e-05, 8.6308e-05, -1.5604e-04, -6.0722e-06, 1.1927e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 216.80, cls_loss 0.0033 cls_loss_mapping 0.0085 cls_loss_causal 0.6105 re_mapping 0.0086 re_causal 0.0258 /// teacc 98.97 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.1064, -0.0664, -0.0547, ..., -0.0186, -0.0338, -0.0118], + [-0.0313, 0.0702, -0.0555, ..., 0.0136, -0.0164, -0.0621], + [ 0.0386, -0.0791, -0.0624, ..., -0.0442, 0.0736, -0.0225], + ..., + [-0.0774, -0.0633, 0.0721, ..., 0.0480, -0.1134, 0.0052], + [ 0.0457, 0.0198, -0.1430, ..., -0.0275, 0.0906, -0.0127], + [-0.0864, -0.0702, 0.0319, ..., -0.0013, -0.0714, -0.0436]], + device='cuda:0'), grad: tensor([[ 9.8273e-06, 3.6135e-06, 1.8114e-07, ..., 0.0000e+00, + 3.9227e-06, 0.0000e+00], + [ 5.4352e-06, 1.6745e-06, 4.0629e-07, ..., 0.0000e+00, + 2.2706e-06, 0.0000e+00], + [ 1.0930e-05, 5.7556e-06, 4.4308e-07, ..., 0.0000e+00, + 4.5151e-06, 0.0000e+00], + ..., + [ 5.7407e-06, 2.2873e-06, -1.2619e-06, ..., 0.0000e+00, + 2.5090e-06, 0.0000e+00], + [ 3.0899e-04, 1.1265e-04, 5.9791e-07, ..., 0.0000e+00, + 1.2994e-04, 0.0000e+00], + [ 6.0648e-06, 2.3022e-06, -1.6512e-06, ..., 0.0000e+00, + 2.6729e-06, 0.0000e+00]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0077, 0.0191, -0.0131, 0.0128, 0.0250, -0.0170, -0.0262, -0.0169, + -0.0207, -0.0283], device='cuda:0'), grad: tensor([ 2.8193e-05, 1.5482e-05, 2.6032e-05, 1.9431e-04, 9.2387e-06, + -7.8297e-04, 8.4162e-05, -1.3769e-04, 5.0211e-04, 6.0320e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 108---------------------------------------------------- +epoch 108, time 217.69, cls_loss 0.0035 cls_loss_mapping 0.0079 cls_loss_causal 0.5749 re_mapping 0.0081 re_causal 0.0244 /// teacc 99.08 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.1073, -0.0670, -0.0548, ..., -0.0186, -0.0342, -0.0118], + [-0.0320, 0.0705, -0.0556, ..., 0.0136, -0.0167, -0.0621], + [ 0.0388, -0.0796, -0.0628, ..., -0.0443, 0.0740, -0.0225], + ..., + [-0.0777, -0.0637, 0.0726, ..., 0.0480, -0.1137, 0.0052], + [ 0.0456, 0.0195, -0.1432, ..., -0.0279, 0.0911, -0.0128], + [-0.0871, -0.0705, 0.0318, ..., -0.0013, -0.0720, -0.0436]], + device='cuda:0'), grad: tensor([[ 6.3926e-06, 6.7987e-06, 3.1404e-06, ..., 9.3132e-10, + 1.8990e-06, 0.0000e+00], + [ 5.4657e-05, 5.7578e-05, 3.3885e-05, ..., 2.3283e-10, + 1.9684e-05, 0.0000e+00], + [ 2.3305e-05, 2.5094e-05, 1.2547e-05, ..., 2.3283e-10, + 6.8806e-06, 0.0000e+00], + ..., + [ 1.4794e-04, 1.7332e-06, 4.8041e-04, ..., 1.1642e-09, + 6.1374e-07, 2.3283e-10], + [ 3.5077e-05, 5.7727e-05, 1.4499e-05, ..., 2.5611e-09, + 3.5129e-06, 2.3283e-10], + [-1.7166e-04, 2.9132e-06, -5.7697e-04, ..., 3.0268e-09, + 1.8831e-06, 2.3283e-10]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0081, 0.0190, -0.0132, 0.0126, 0.0248, -0.0163, -0.0265, -0.0164, + -0.0210, -0.0285], device='cuda:0'), grad: tensor([ 3.2306e-05, 3.1829e-04, 1.2279e-04, 7.3910e-05, 4.5514e-04, + 8.2076e-05, -6.0749e-04, 3.2864e-03, 1.6832e-04, -3.9291e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 217.08, cls_loss 0.0030 cls_loss_mapping 0.0062 cls_loss_causal 0.5546 re_mapping 0.0086 re_causal 0.0248 /// teacc 98.90 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.1080, -0.0673, -0.0530, ..., -0.0186, -0.0330, -0.0116], + [-0.0326, 0.0710, -0.0558, ..., 0.0136, -0.0171, -0.0623], + [ 0.0406, -0.0803, -0.0627, ..., -0.0442, 0.0757, -0.0225], + ..., + [-0.0795, -0.0642, 0.0727, ..., 0.0480, -0.1149, 0.0052], + [ 0.0452, 0.0193, -0.1437, ..., -0.0281, 0.0908, -0.0128], + [-0.0876, -0.0711, 0.0317, ..., -0.0014, -0.0726, -0.0438]], + device='cuda:0'), grad: tensor([[ 5.9232e-07, 4.0466e-07, 6.6357e-08, ..., 4.6566e-10, + 8.1025e-08, 0.0000e+00], + [ 3.9209e-07, -7.8464e-07, 4.4587e-07, ..., 4.6566e-10, + 3.1246e-07, 0.0000e+00], + [-1.7630e-06, 5.0385e-07, 1.8952e-07, ..., 5.3551e-09, + -2.6636e-06, 0.0000e+00], + ..., + [ 1.4924e-07, 5.5227e-07, -1.1772e-06, ..., 6.9849e-10, + 2.1583e-07, 0.0000e+00], + [ 1.8133e-06, 1.6121e-06, 2.6426e-07, ..., 3.4925e-09, + 9.7975e-07, 4.6566e-10], + [ 1.3388e-07, 2.1998e-06, 1.5656e-06, ..., 1.8626e-09, + 2.3609e-07, 6.9849e-10]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0068, 0.0191, -0.0126, 0.0120, 0.0246, -0.0163, -0.0258, -0.0169, + -0.0218, -0.0284], device='cuda:0'), grad: tensor([-3.1646e-06, 1.0021e-06, -1.1018e-06, -1.6708e-06, -7.5884e-06, + 8.5384e-06, -6.9588e-06, -2.0172e-06, 5.2191e-06, 7.7412e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 216.95, cls_loss 0.0028 cls_loss_mapping 0.0051 cls_loss_causal 0.5271 re_mapping 0.0078 re_causal 0.0228 /// teacc 98.97 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.1088, -0.0678, -0.0527, ..., -0.0185, -0.0330, -0.0113], + [-0.0328, 0.0718, -0.0561, ..., 0.0136, -0.0166, -0.0627], + [ 0.0408, -0.0813, -0.0631, ..., -0.0443, 0.0756, -0.0225], + ..., + [-0.0797, -0.0642, 0.0732, ..., 0.0480, -0.1152, 0.0051], + [ 0.0452, 0.0185, -0.1440, ..., -0.0282, 0.0908, -0.0140], + [-0.0886, -0.0714, 0.0314, ..., -0.0014, -0.0732, -0.0440]], + device='cuda:0'), grad: tensor([[ 3.9930e-07, 2.9779e-07, 1.8533e-07, ..., 0.0000e+00, + 3.1060e-07, 1.6298e-09], + [ 1.6019e-07, -7.7393e-07, 1.7625e-07, ..., 0.0000e+00, + 5.5414e-08, 2.3283e-10], + [ 3.2596e-08, 5.6904e-07, 4.4983e-07, ..., 0.0000e+00, + 3.6811e-07, 2.3283e-10], + ..., + [ 1.1991e-07, 5.5321e-07, 1.2992e-07, ..., 0.0000e+00, + 3.0338e-07, 1.8626e-09], + [ 1.2480e-06, 1.1930e-06, 2.9267e-07, ..., 0.0000e+00, + -4.6566e-07, 2.7940e-09], + [ 2.5495e-07, 7.5903e-07, -2.8289e-07, ..., 0.0000e+00, + 3.1549e-07, 1.6298e-09]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0066, 0.0194, -0.0133, 0.0128, 0.0245, -0.0166, -0.0257, -0.0164, + -0.0223, -0.0288], device='cuda:0'), grad: tensor([ 1.1139e-06, 6.1560e-07, 6.1020e-06, -1.4707e-05, -5.5972e-07, + 2.6543e-06, -3.1497e-06, 2.4941e-06, 3.5372e-06, 1.8813e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 216.85, cls_loss 0.0037 cls_loss_mapping 0.0066 cls_loss_causal 0.5833 re_mapping 0.0077 re_causal 0.0244 /// teacc 99.01 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.1097, -0.0685, -0.0530, ..., -0.0185, -0.0332, -0.0113], + [-0.0332, 0.0724, -0.0567, ..., 0.0136, -0.0161, -0.0630], + [ 0.0416, -0.0823, -0.0632, ..., -0.0443, 0.0762, -0.0226], + ..., + [-0.0806, -0.0643, 0.0727, ..., 0.0480, -0.1159, 0.0051], + [ 0.0455, 0.0180, -0.1444, ..., -0.0283, 0.0911, -0.0149], + [-0.0893, -0.0717, 0.0324, ..., -0.0014, -0.0744, -0.0441]], + device='cuda:0'), grad: tensor([[ 3.7670e-05, 2.4021e-05, 8.2888e-08, ..., 0.0000e+00, + 3.3621e-07, 0.0000e+00], + [-9.7416e-07, -1.2167e-05, 1.2014e-07, ..., 0.0000e+00, + -4.0513e-08, 0.0000e+00], + [-2.1439e-06, 1.5730e-06, 2.0582e-07, ..., 0.0000e+00, + -2.2743e-06, 0.0000e+00], + ..., + [ 7.2923e-07, 9.1549e-07, 2.8848e-07, ..., 0.0000e+00, + 8.1770e-07, 0.0000e+00], + [ 2.9713e-05, 2.0489e-05, 2.0233e-07, ..., 0.0000e+00, + 2.5798e-06, 0.0000e+00], + [ 2.5630e-06, 1.8785e-06, -2.8056e-07, ..., 0.0000e+00, + -9.4995e-07, 0.0000e+00]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0069, 0.0193, -0.0131, 0.0121, 0.0245, -0.0164, -0.0255, -0.0173, + -0.0226, -0.0276], device='cuda:0'), grad: tensor([ 6.6757e-05, -1.1310e-05, 4.1388e-06, 5.6699e-06, 2.6841e-06, + 9.9361e-05, -2.1636e-04, 1.0394e-05, 6.1333e-05, -2.2799e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 216.79, cls_loss 0.0024 cls_loss_mapping 0.0052 cls_loss_causal 0.5822 re_mapping 0.0076 re_causal 0.0246 /// teacc 98.78 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.1102, -0.0688, -0.0532, ..., -0.0186, -0.0333, -0.0112], + [-0.0342, 0.0725, -0.0568, ..., 0.0136, -0.0167, -0.0631], + [ 0.0418, -0.0825, -0.0636, ..., -0.0443, 0.0766, -0.0226], + ..., + [-0.0809, -0.0643, 0.0728, ..., 0.0480, -0.1161, 0.0051], + [ 0.0459, 0.0182, -0.1447, ..., -0.0283, 0.0914, -0.0152], + [-0.0895, -0.0723, 0.0324, ..., -0.0014, -0.0746, -0.0441]], + device='cuda:0'), grad: tensor([[ 1.1504e-05, 5.8450e-06, 6.1933e-08, ..., 0.0000e+00, + 3.6620e-06, 8.5449e-08], + [ 7.8045e-07, -6.1560e-07, 2.7963e-07, ..., 0.0000e+00, + 6.9384e-07, 8.3819e-09], + [-2.7284e-05, 5.4110e-07, 1.0268e-07, ..., 0.0000e+00, + -3.7283e-05, 1.2573e-08], + ..., + [ 1.4439e-05, 1.0873e-07, -2.1681e-06, ..., 0.0000e+00, + 2.1592e-05, 3.1898e-08], + [ 5.2340e-06, 1.0394e-06, 7.4506e-08, ..., 0.0000e+00, + 2.8312e-06, 1.9488e-07], + [ 1.0468e-06, 2.4331e-07, 1.0617e-06, ..., 0.0000e+00, + 5.8906e-07, 6.9849e-08]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0074, 0.0188, -0.0132, 0.0121, 0.0248, -0.0165, -0.0253, -0.0171, + -0.0225, -0.0273], device='cuda:0'), grad: tensor([ 1.8865e-05, 2.1700e-06, -7.9393e-05, 1.5020e-05, 2.0023e-06, + -4.8168e-06, -1.3210e-05, 4.2975e-05, 1.0535e-05, 5.9046e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 216.89, cls_loss 0.0035 cls_loss_mapping 0.0062 cls_loss_causal 0.5871 re_mapping 0.0075 re_causal 0.0226 /// teacc 99.07 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.1102, -0.0686, -0.0534, ..., -0.0187, -0.0326, -0.0103], + [-0.0351, 0.0732, -0.0571, ..., 0.0136, -0.0166, -0.0635], + [ 0.0420, -0.0834, -0.0642, ..., -0.0444, 0.0768, -0.0228], + ..., + [-0.0811, -0.0645, 0.0735, ..., 0.0481, -0.1166, 0.0049], + [ 0.0462, 0.0179, -0.1452, ..., -0.0283, 0.0916, -0.0157], + [-0.0909, -0.0729, 0.0320, ..., -0.0014, -0.0753, -0.0444]], + device='cuda:0'), grad: tensor([[-1.4016e-07, 2.2305e-07, 1.8394e-08, ..., 0.0000e+00, + 2.9453e-07, 0.0000e+00], + [ 3.8147e-06, -5.9567e-06, 6.1747e-07, ..., 0.0000e+00, + 2.4065e-06, 0.0000e+00], + [ 3.3770e-06, 3.2093e-06, 6.9663e-07, ..., 0.0000e+00, + 2.6859e-06, 0.0000e+00], + ..., + [-1.2219e-06, 1.6214e-06, -1.9027e-06, ..., 0.0000e+00, + 2.8196e-07, 0.0000e+00], + [-2.4930e-05, -1.7896e-05, 7.1479e-08, ..., 0.0000e+00, + -1.8895e-05, 0.0000e+00], + [ 1.5581e-06, 1.0543e-06, 5.8440e-08, ..., 0.0000e+00, + 1.2899e-06, 0.0000e+00]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0066, 0.0190, -0.0139, 0.0110, 0.0253, -0.0157, -0.0256, -0.0166, + -0.0230, -0.0277], device='cuda:0'), grad: tensor([-4.8935e-05, 3.5949e-07, 1.6853e-05, 5.4017e-06, 2.1327e-06, + 1.5765e-05, 3.3647e-05, -5.9754e-06, -4.7952e-05, 2.8715e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 216.93, cls_loss 0.0032 cls_loss_mapping 0.0064 cls_loss_causal 0.5795 re_mapping 0.0075 re_causal 0.0238 /// teacc 98.93 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.1108, -0.0685, -0.0534, ..., -0.0182, -0.0326, -0.0094], + [-0.0354, 0.0742, -0.0572, ..., 0.0136, -0.0164, -0.0644], + [ 0.0420, -0.0844, -0.0645, ..., -0.0444, 0.0771, -0.0234], + ..., + [-0.0831, -0.0646, 0.0738, ..., 0.0481, -0.1176, 0.0048], + [ 0.0473, 0.0177, -0.1455, ..., -0.0286, 0.0931, -0.0163], + [-0.0941, -0.0736, 0.0319, ..., -0.0015, -0.0769, -0.0451]], + device='cuda:0'), grad: tensor([[ 5.2482e-05, 3.0115e-05, 1.7951e-07, ..., 0.0000e+00, + 1.3085e-07, 0.0000e+00], + [ 5.0850e-06, -5.2035e-05, 6.7800e-07, ..., 0.0000e+00, + 2.6682e-07, 0.0000e+00], + [ 1.3530e-05, 4.6521e-05, 2.7218e-07, ..., 0.0000e+00, + 2.3954e-06, 0.0000e+00], + ..., + [ 1.3188e-06, 2.5500e-06, 7.4040e-07, ..., 0.0000e+00, + 1.5413e-07, 0.0000e+00], + [ 4.1395e-05, 2.6360e-05, 7.0361e-07, ..., 0.0000e+00, + -5.4389e-06, 0.0000e+00], + [ 3.3248e-06, 1.1846e-05, 9.1940e-06, ..., 0.0000e+00, + 3.3318e-07, 0.0000e+00]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0061, 0.0194, -0.0149, 0.0114, 0.0251, -0.0159, -0.0255, -0.0164, + -0.0223, -0.0282], device='cuda:0'), grad: tensor([-1.0663e-04, -8.0049e-05, 8.9586e-05, 9.0823e-06, -4.4820e-07, + 1.2070e-04, -1.4281e-04, 1.0118e-05, 6.5327e-05, 3.5554e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 216.79, cls_loss 0.0035 cls_loss_mapping 0.0075 cls_loss_causal 0.5745 re_mapping 0.0078 re_causal 0.0238 /// teacc 98.91 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.1122, -0.0700, -0.0536, ..., -0.0181, -0.0325, -0.0094], + [-0.0360, 0.0758, -0.0577, ..., 0.0136, -0.0167, -0.0644], + [ 0.0427, -0.0848, -0.0649, ..., -0.0445, 0.0780, -0.0234], + ..., + [-0.0834, -0.0662, 0.0744, ..., 0.0480, -0.1181, 0.0048], + [ 0.0468, 0.0167, -0.1464, ..., -0.0287, 0.0923, -0.0163], + [-0.0943, -0.0748, 0.0311, ..., -0.0015, -0.0772, -0.0451]], + device='cuda:0'), grad: tensor([[-3.8505e-05, 1.1139e-05, 8.4843e-07, ..., 0.0000e+00, + -5.7250e-05, 0.0000e+00], + [ 1.9670e-06, -2.3231e-05, 2.0619e-06, ..., 0.0000e+00, + 1.2264e-05, 0.0000e+00], + [-8.5175e-05, 3.9563e-06, -1.0484e-04, ..., 0.0000e+00, + -2.4104e-04, 0.0000e+00], + ..., + [ 8.5950e-05, 1.4687e-06, 9.8825e-05, ..., 0.0000e+00, + 2.1172e-04, 0.0000e+00], + [ 2.6166e-05, 4.1053e-06, 1.4445e-06, ..., 0.0000e+00, + 4.6372e-05, 0.0000e+00], + [ 9.0823e-06, 8.3372e-06, 4.7721e-06, ..., 0.0000e+00, + 1.5572e-05, 0.0000e+00]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0061, 0.0202, -0.0146, 0.0127, 0.0257, -0.0163, -0.0261, -0.0166, + -0.0236, -0.0288], device='cuda:0'), grad: tensor([ 1.3137e-04, 1.4052e-05, -4.2105e-04, 2.5451e-05, 8.6874e-06, + 9.1940e-06, -1.8597e-05, 3.7074e-04, 1.1367e-04, -2.3365e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 216.84, cls_loss 0.0037 cls_loss_mapping 0.0075 cls_loss_causal 0.5904 re_mapping 0.0073 re_causal 0.0244 /// teacc 98.96 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.1125, -0.0710, -0.0537, ..., -0.0181, -0.0328, -0.0094], + [-0.0365, 0.0746, -0.0595, ..., 0.0136, -0.0188, -0.0648], + [ 0.0427, -0.0857, -0.0652, ..., -0.0445, 0.0783, -0.0234], + ..., + [-0.0837, -0.0644, 0.0758, ..., 0.0481, -0.1164, 0.0048], + [ 0.0470, 0.0166, -0.1464, ..., -0.0288, 0.0927, -0.0165], + [-0.0952, -0.0762, 0.0310, ..., -0.0015, -0.0780, -0.0456]], + device='cuda:0'), grad: tensor([[ 4.0582e-07, 2.2328e-07, 1.4412e-07, ..., 1.6298e-09, + 2.8405e-07, 1.3970e-09], + [ 1.2685e-06, -4.7684e-06, 4.4797e-07, ..., 6.9849e-10, + 1.1940e-06, 4.6566e-10], + [-3.3993e-06, 4.4852e-06, 1.2154e-07, ..., 6.9849e-10, + -5.7630e-06, 4.6566e-10], + ..., + [ 1.3709e-06, 7.8045e-07, 3.2107e-07, ..., 2.3283e-09, + 1.4780e-06, 3.0268e-09], + [-1.1250e-05, -4.0531e-06, 4.4028e-07, ..., 4.4238e-09, + -8.0243e-06, 5.5879e-09], + [ 7.1619e-07, 3.5204e-07, -4.0084e-06, ..., 5.8208e-09, + 6.1374e-07, 2.3283e-09]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0058, 0.0196, -0.0151, 0.0125, 0.0252, -0.0164, -0.0254, -0.0157, + -0.0237, -0.0292], device='cuda:0'), grad: tensor([-5.2387e-07, -2.1365e-06, -2.2701e-07, -1.1377e-05, 8.3596e-06, + 3.4302e-05, 3.8259e-06, 7.1004e-06, -1.9506e-05, -1.9863e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 216.68, cls_loss 0.0030 cls_loss_mapping 0.0065 cls_loss_causal 0.5465 re_mapping 0.0078 re_causal 0.0233 /// teacc 98.94 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.1140, -0.0718, -0.0538, ..., -0.0181, -0.0335, -0.0093], + [-0.0391, 0.0730, -0.0611, ..., 0.0136, -0.0203, -0.0653], + [ 0.0433, -0.0850, -0.0654, ..., -0.0445, 0.0798, -0.0236], + ..., + [-0.0843, -0.0644, 0.0761, ..., 0.0481, -0.1170, 0.0047], + [ 0.0473, 0.0164, -0.1472, ..., -0.0288, 0.0934, -0.0179], + [-0.0954, -0.0744, 0.0315, ..., -0.0015, -0.0786, -0.0447]], + device='cuda:0'), grad: tensor([[ 4.4680e-07, 1.4738e-07, 4.2375e-08, ..., 0.0000e+00, + -4.9360e-07, 0.0000e+00], + [ 1.2144e-06, -2.4159e-06, 8.8941e-08, ..., 0.0000e+00, + 1.1167e-06, 0.0000e+00], + [-3.6001e-05, 7.4692e-07, 4.9826e-08, ..., 0.0000e+00, + -3.1948e-05, 0.0000e+00], + ..., + [ 2.6882e-05, 6.9663e-07, 1.9139e-07, ..., 0.0000e+00, + 2.4423e-05, 0.0000e+00], + [ 3.1386e-06, 1.9697e-07, 2.5984e-07, ..., 0.0000e+00, + 2.4270e-06, 0.0000e+00], + [ 3.3667e-07, 3.2634e-06, -5.8627e-07, ..., 0.0000e+00, + -2.1537e-07, 0.0000e+00]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0062, 0.0182, -0.0142, 0.0125, 0.0256, -0.0168, -0.0244, -0.0157, + -0.0237, -0.0281], device='cuda:0'), grad: tensor([-8.8289e-06, -5.1595e-07, -5.6356e-05, 7.1377e-06, 7.4655e-06, + -1.4668e-08, 1.1930e-06, 4.9859e-05, 7.3016e-06, -7.1898e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 217.03, cls_loss 0.0028 cls_loss_mapping 0.0075 cls_loss_causal 0.5644 re_mapping 0.0077 re_causal 0.0236 /// teacc 98.94 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.1146, -0.0724, -0.0540, ..., -0.0180, -0.0337, -0.0091], + [-0.0394, 0.0735, -0.0610, ..., 0.0136, -0.0197, -0.0654], + [ 0.0432, -0.0857, -0.0671, ..., -0.0445, 0.0801, -0.0237], + ..., + [-0.0847, -0.0645, 0.0766, ..., 0.0481, -0.1173, 0.0047], + [ 0.0468, 0.0155, -0.1481, ..., -0.0288, 0.0934, -0.0182], + [-0.0958, -0.0743, 0.0310, ..., -0.0015, -0.0793, -0.0449]], + device='cuda:0'), grad: tensor([[ 1.0533e-06, 5.0291e-07, 1.6461e-07, ..., 0.0000e+00, + 4.8894e-08, 0.0000e+00], + [ 2.0815e-07, 5.1223e-08, 2.4750e-07, ..., 0.0000e+00, + 3.4133e-07, 0.0000e+00], + [-2.5760e-06, 1.0408e-07, 1.2061e-07, ..., 0.0000e+00, + -4.3288e-06, 0.0000e+00], + ..., + [ 2.5914e-07, 2.7707e-08, 1.4603e-06, ..., 0.0000e+00, + 4.6007e-06, 0.0000e+00], + [ 1.8189e-06, 5.2527e-07, 6.6217e-07, ..., 0.0000e+00, + 2.2817e-06, 0.0000e+00], + [ 1.6158e-07, 2.4866e-07, -4.6194e-07, ..., 0.0000e+00, + 4.0233e-07, 0.0000e+00]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0062, 0.0184, -0.0146, 0.0125, 0.0260, -0.0169, -0.0236, -0.0156, + -0.0248, -0.0283], device='cuda:0'), grad: tensor([-5.0031e-06, 1.9912e-06, -4.3660e-06, -5.1856e-05, -3.9069e-07, + 1.6093e-05, -1.6257e-05, 4.0263e-05, 1.2830e-05, 6.7912e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 216.99, cls_loss 0.0030 cls_loss_mapping 0.0066 cls_loss_causal 0.5491 re_mapping 0.0076 re_causal 0.0232 /// teacc 99.05 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.1155, -0.0709, -0.0513, ..., -0.0180, -0.0306, -0.0089], + [-0.0398, 0.0736, -0.0612, ..., 0.0136, -0.0197, -0.0655], + [ 0.0429, -0.0862, -0.0685, ..., -0.0445, 0.0793, -0.0238], + ..., + [-0.0850, -0.0646, 0.0773, ..., 0.0481, -0.1175, 0.0046], + [ 0.0468, 0.0154, -0.1488, ..., -0.0289, 0.0936, -0.0183], + [-0.0963, -0.0736, 0.0318, ..., -0.0015, -0.0796, -0.0452]], + device='cuda:0'), grad: tensor([[ 1.9884e-07, 4.5216e-07, 1.8976e-07, ..., 4.6566e-10, + 2.5309e-07, 4.6566e-10], + [ 9.7789e-07, -2.7250e-06, 2.6845e-07, ..., 2.3283e-10, + 2.0005e-06, 2.3283e-10], + [-1.7220e-06, 5.0617e-07, 2.5914e-07, ..., 2.3283e-10, + -5.4352e-06, 2.3283e-10], + ..., + [ 7.9582e-07, 1.2852e-06, 2.4717e-06, ..., 9.3132e-10, + 2.0303e-06, 1.1642e-09], + [-2.5928e-06, -1.4026e-06, 1.4971e-07, ..., 5.8208e-09, + -3.2969e-06, 6.7521e-09], + [ 1.8757e-06, 3.5800e-06, -2.1681e-05, ..., 9.3132e-10, + 2.5406e-06, 9.3132e-10]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0033, 0.0184, -0.0158, 0.0128, 0.0230, -0.0167, -0.0240, -0.0153, + -0.0255, -0.0274], device='cuda:0'), grad: tensor([ 3.8184e-07, 2.7157e-06, -1.0476e-05, 8.4266e-06, 1.2061e-06, + 5.0217e-05, 2.6785e-06, 1.4625e-05, -3.7625e-06, -6.6161e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 217.18, cls_loss 0.0022 cls_loss_mapping 0.0047 cls_loss_causal 0.5488 re_mapping 0.0082 re_causal 0.0242 /// teacc 98.91 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.1164, -0.0717, -0.0513, ..., -0.0180, -0.0308, -0.0087], + [-0.0402, 0.0737, -0.0614, ..., 0.0136, -0.0198, -0.0657], + [ 0.0431, -0.0866, -0.0688, ..., -0.0445, 0.0794, -0.0242], + ..., + [-0.0862, -0.0647, 0.0774, ..., 0.0481, -0.1183, 0.0046], + [ 0.0469, 0.0155, -0.1503, ..., -0.0290, 0.0940, -0.0183], + [-0.0951, -0.0737, 0.0320, ..., -0.0016, -0.0789, -0.0453]], + device='cuda:0'), grad: tensor([[ 1.5469e-06, -6.5658e-08, 2.2328e-07, ..., 0.0000e+00, + 2.5965e-06, 0.0000e+00], + [ 1.2610e-06, -5.1968e-06, 3.8743e-07, ..., 0.0000e+00, + 1.5795e-06, 0.0000e+00], + [-1.8036e-04, -5.0008e-05, -4.6402e-05, ..., 0.0000e+00, + -2.9182e-04, 0.0000e+00], + ..., + [ 2.2333e-06, 1.2796e-06, 1.8114e-07, ..., 0.0000e+00, + 3.9004e-06, 0.0000e+00], + [ 1.6880e-04, 4.9114e-05, 4.3899e-05, ..., 0.0000e+00, + 2.7275e-04, 0.0000e+00], + [ 9.0105e-08, 2.4214e-07, -6.3423e-07, ..., 0.0000e+00, + 1.9348e-07, 0.0000e+00]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0035, 0.0184, -0.0159, 0.0130, 0.0228, -0.0169, -0.0234, -0.0155, + -0.0263, -0.0267], device='cuda:0'), grad: tensor([-1.2219e-06, -4.2096e-06, -2.8872e-04, 1.1525e-07, 5.3681e-06, + 4.2766e-06, 7.3016e-06, 4.0382e-06, 2.7370e-04, -3.4249e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 217.06, cls_loss 0.0027 cls_loss_mapping 0.0058 cls_loss_causal 0.5570 re_mapping 0.0077 re_causal 0.0228 /// teacc 98.93 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.1176, -0.0726, -0.0514, ..., -0.0180, -0.0310, -0.0086], + [-0.0407, 0.0739, -0.0616, ..., 0.0136, -0.0198, -0.0657], + [ 0.0432, -0.0871, -0.0690, ..., -0.0445, 0.0798, -0.0242], + ..., + [-0.0869, -0.0648, 0.0777, ..., 0.0481, -0.1189, 0.0046], + [ 0.0480, 0.0152, -0.1507, ..., -0.0291, 0.0954, -0.0184], + [-0.0970, -0.0739, 0.0320, ..., -0.0016, -0.0808, -0.0454]], + device='cuda:0'), grad: tensor([[ 3.9488e-06, -5.5647e-07, 7.0781e-08, ..., 0.0000e+00, + 1.1241e-06, 0.0000e+00], + [ 4.7497e-06, -2.1644e-06, 7.8883e-07, ..., 0.0000e+00, + 3.2969e-06, 0.0000e+00], + [ 4.9248e-06, 5.0813e-06, 2.8173e-07, ..., 0.0000e+00, + 3.6806e-06, 0.0000e+00], + ..., + [ 4.5151e-06, 5.0142e-06, -1.5814e-06, ..., 0.0000e+00, + 4.5374e-06, 0.0000e+00], + [-2.0102e-05, -7.2271e-06, 6.1002e-08, ..., 0.0000e+00, + -3.3349e-05, 0.0000e+00], + [ 3.6843e-06, 5.0999e-06, -7.2224e-07, ..., 0.0000e+00, + 2.3693e-06, 0.0000e+00]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0038, 0.0184, -0.0161, 0.0152, 0.0229, -0.0164, -0.0239, -0.0156, + -0.0246, -0.0289], device='cuda:0'), grad: tensor([-5.3227e-05, 9.8720e-06, 1.7166e-05, 2.7746e-05, 1.0543e-05, + 1.3614e-04, -1.2481e-04, 1.2159e-05, -6.0499e-05, 2.5004e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 216.71, cls_loss 0.0024 cls_loss_mapping 0.0055 cls_loss_causal 0.5488 re_mapping 0.0071 re_causal 0.0227 /// teacc 98.91 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.1183, -0.0729, -0.0514, ..., -0.0180, -0.0311, -0.0086], + [-0.0409, 0.0746, -0.0617, ..., 0.0136, -0.0183, -0.0658], + [ 0.0432, -0.0888, -0.0692, ..., -0.0445, 0.0787, -0.0242], + ..., + [-0.0871, -0.0649, 0.0780, ..., 0.0481, -0.1191, 0.0046], + [ 0.0482, 0.0151, -0.1510, ..., -0.0291, 0.0956, -0.0184], + [-0.0978, -0.0743, 0.0312, ..., -0.0016, -0.0803, -0.0455]], + device='cuda:0'), grad: tensor([[ 1.3635e-06, 1.6708e-06, 2.5891e-07, ..., 0.0000e+00, + 8.5356e-07, 0.0000e+00], + [ 2.8219e-07, -1.7118e-06, 6.7893e-07, ..., 0.0000e+00, + 1.7900e-06, 0.0000e+00], + [ 3.2410e-07, 1.2163e-06, 9.2760e-07, ..., 0.0000e+00, + 9.0450e-06, 0.0000e+00], + ..., + [ 2.2212e-07, 1.1735e-06, -1.5609e-06, ..., 0.0000e+00, + 3.1944e-06, 0.0000e+00], + [ 2.1122e-06, 2.0750e-06, 3.5809e-07, ..., 0.0000e+00, + 1.2331e-05, 0.0000e+00], + [ 1.3327e-06, 8.0187e-07, -5.2191e-06, ..., 0.0000e+00, + 1.8496e-06, 0.0000e+00]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0034, 0.0191, -0.0177, 0.0151, 0.0234, -0.0160, -0.0239, -0.0153, + -0.0249, -0.0296], device='cuda:0'), grad: tensor([ 5.2080e-06, 3.2969e-06, 1.8537e-05, -4.6581e-05, 1.5177e-05, + -5.6177e-06, -7.8157e-06, 1.9781e-06, 2.9013e-05, -1.3143e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 216.85, cls_loss 0.0033 cls_loss_mapping 0.0070 cls_loss_causal 0.5597 re_mapping 0.0076 re_causal 0.0225 /// teacc 98.99 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.1193, -0.0732, -0.0515, ..., -0.0180, -0.0313, -0.0086], + [-0.0414, 0.0747, -0.0619, ..., 0.0136, -0.0185, -0.0658], + [ 0.0431, -0.0892, -0.0698, ..., -0.0445, 0.0784, -0.0242], + ..., + [-0.0876, -0.0650, 0.0784, ..., 0.0481, -0.1196, 0.0046], + [ 0.0477, 0.0144, -0.1515, ..., -0.0292, 0.0956, -0.0184], + [-0.0988, -0.0739, 0.0337, ..., -0.0016, -0.0813, -0.0455]], + device='cuda:0'), grad: tensor([[ 2.3115e-06, 2.6170e-07, 9.3132e-08, ..., 0.0000e+00, + 3.7290e-06, 0.0000e+00], + [ 2.1048e-06, -3.2708e-06, 2.3702e-07, ..., 0.0000e+00, + 3.9376e-06, 0.0000e+00], + [-1.3575e-05, 2.7101e-07, 7.4971e-08, ..., 0.0000e+00, + -2.6241e-05, 0.0000e+00], + ..., + [ 1.3411e-06, 6.0536e-07, -2.1420e-06, ..., 0.0000e+00, + 2.6487e-06, 0.0000e+00], + [ 3.8408e-06, 1.2554e-06, 1.2387e-07, ..., 0.0000e+00, + 6.4038e-06, 0.0000e+00], + [ 7.1805e-07, 3.9628e-07, 1.0710e-06, ..., 0.0000e+00, + 1.0096e-06, 0.0000e+00]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0036, 0.0190, -0.0183, 0.0153, 0.0214, -0.0147, -0.0246, -0.0151, + -0.0258, -0.0281], device='cuda:0'), grad: tensor([ 9.2015e-06, 6.0834e-06, -6.2108e-05, 1.5497e-05, 3.0734e-06, + -1.2666e-05, 1.4044e-05, 1.2349e-06, 1.9923e-05, 5.7518e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 216.86, cls_loss 0.0029 cls_loss_mapping 0.0064 cls_loss_causal 0.5490 re_mapping 0.0072 re_causal 0.0218 /// teacc 99.04 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.1202, -0.0735, -0.0517, ..., -0.0180, -0.0315, -0.0085], + [-0.0417, 0.0751, -0.0623, ..., 0.0136, -0.0186, -0.0658], + [ 0.0437, -0.0894, -0.0700, ..., -0.0445, 0.0792, -0.0242], + ..., + [-0.0884, -0.0652, 0.0793, ..., 0.0481, -0.1203, 0.0046], + [ 0.0485, 0.0134, -0.1522, ..., -0.0294, 0.0966, -0.0185], + [-0.1023, -0.0750, 0.0317, ..., -0.0016, -0.0832, -0.0455]], + device='cuda:0'), grad: tensor([[ 4.7460e-06, 7.5437e-08, 4.3772e-08, ..., 4.6566e-10, + 7.0557e-06, 0.0000e+00], + [ 2.1942e-06, -4.2468e-07, 1.8626e-07, ..., -7.4506e-09, + 6.9514e-06, 0.0000e+00], + [-1.5073e-05, 2.0629e-07, 1.2154e-07, ..., 9.3132e-10, + -2.6882e-05, 0.0000e+00], + ..., + [ 1.8543e-06, 8.8755e-07, 1.0710e-06, ..., 3.2596e-09, + 2.9821e-06, 0.0000e+00], + [ 3.8650e-07, 1.5367e-08, 3.7299e-07, ..., 1.3970e-09, + 1.2126e-06, 0.0000e+00], + [ 2.2678e-07, 3.3155e-07, -8.2096e-07, ..., 4.6566e-10, + 2.5751e-07, 0.0000e+00]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0038, 0.0192, -0.0179, 0.0145, 0.0234, -0.0139, -0.0243, -0.0147, + -0.0255, -0.0304], device='cuda:0'), grad: tensor([ 1.9699e-05, 1.3620e-05, -6.5684e-05, 1.0774e-05, 6.6170e-07, + -8.7731e-07, 1.1280e-05, 8.7023e-06, 4.3921e-06, -2.5034e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 216.82, cls_loss 0.0022 cls_loss_mapping 0.0047 cls_loss_causal 0.5731 re_mapping 0.0068 re_causal 0.0230 /// teacc 99.05 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.1209, -0.0738, -0.0518, ..., -0.0180, -0.0317, -0.0084], + [-0.0417, 0.0750, -0.0623, ..., 0.0136, -0.0190, -0.0659], + [ 0.0447, -0.0884, -0.0694, ..., -0.0446, 0.0803, -0.0244], + ..., + [-0.0887, -0.0653, 0.0799, ..., 0.0481, -0.1205, 0.0046], + [ 0.0483, 0.0132, -0.1538, ..., -0.0294, 0.0963, -0.0186], + [-0.1027, -0.0750, 0.0327, ..., -0.0016, -0.0836, -0.0456]], + device='cuda:0'), grad: tensor([[ 4.8662e-07, 3.7812e-07, 1.3970e-07, ..., 1.1642e-08, + -3.2373e-06, 0.0000e+00], + [ 3.4785e-07, -1.2144e-06, 1.9558e-07, ..., 2.2352e-08, + 1.7192e-06, 0.0000e+00], + [-1.0431e-06, 1.6913e-06, 1.4342e-07, ..., 1.0710e-08, + -2.4848e-06, 0.0000e+00], + ..., + [ 4.4098e-07, 7.8231e-07, 1.3178e-07, ..., 1.1642e-08, + 7.5111e-07, 0.0000e+00], + [ 2.4978e-06, 1.1157e-06, 1.4622e-07, ..., 2.3283e-08, + 1.6429e-06, 0.0000e+00], + [ 3.3854e-07, 3.5716e-07, -5.1688e-07, ..., 2.4214e-08, + 4.2655e-07, 0.0000e+00]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0038, 0.0188, -0.0167, 0.0145, 0.0218, -0.0142, -0.0240, -0.0145, + -0.0257, -0.0293], device='cuda:0'), grad: tensor([-1.3411e-05, 5.3309e-06, 1.6727e-06, -2.1793e-06, 2.9176e-05, + -6.1318e-06, -2.7090e-05, 3.3341e-06, 8.5160e-06, 7.6648e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 216.67, cls_loss 0.0020 cls_loss_mapping 0.0048 cls_loss_causal 0.5456 re_mapping 0.0068 re_causal 0.0219 /// teacc 98.99 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.1228, -0.0747, -0.0518, ..., -0.0180, -0.0318, -0.0084], + [-0.0420, 0.0754, -0.0627, ..., 0.0136, -0.0193, -0.0662], + [ 0.0445, -0.0885, -0.0694, ..., -0.0446, 0.0807, -0.0245], + ..., + [-0.0892, -0.0654, 0.0804, ..., 0.0480, -0.1209, 0.0045], + [ 0.0502, 0.0138, -0.1542, ..., -0.0296, 0.0970, -0.0186], + [-0.1028, -0.0752, 0.0327, ..., -0.0016, -0.0837, -0.0463]], + device='cuda:0'), grad: tensor([[ 1.6298e-07, 2.2119e-07, 7.9628e-08, ..., 0.0000e+00, + -4.3772e-08, 0.0000e+00], + [ 2.1886e-08, -2.1040e-05, 1.8161e-08, ..., 0.0000e+00, + -1.4551e-05, 0.0000e+00], + [ 2.2538e-07, 1.7956e-05, 5.6811e-08, ..., 0.0000e+00, + 1.3120e-05, 0.0000e+00], + ..., + [ 6.5193e-09, 1.5907e-06, 2.4121e-07, ..., 0.0000e+00, + 8.0466e-07, 0.0000e+00], + [ 7.9628e-08, 2.0303e-07, 2.0768e-07, ..., 0.0000e+00, + 1.6158e-07, 4.6566e-10], + [ 1.7229e-08, 3.2363e-07, -2.3702e-07, ..., 0.0000e+00, + 1.3597e-07, 4.6566e-10]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0047, 0.0189, -0.0165, 0.0143, 0.0218, -0.0140, -0.0248, -0.0143, + -0.0246, -0.0294], device='cuda:0'), grad: tensor([-8.3819e-09, -5.2184e-05, 4.6432e-05, 1.4044e-06, -2.0228e-06, + 5.9977e-07, -5.5879e-08, 4.4033e-06, 1.6717e-06, -2.5565e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 216.72, cls_loss 0.0025 cls_loss_mapping 0.0049 cls_loss_causal 0.5354 re_mapping 0.0069 re_causal 0.0208 /// teacc 98.89 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.1238, -0.0757, -0.0520, ..., -0.0180, -0.0319, -0.0083], + [-0.0425, 0.0754, -0.0629, ..., 0.0136, -0.0194, -0.0663], + [ 0.0448, -0.0889, -0.0697, ..., -0.0446, 0.0809, -0.0247], + ..., + [-0.0897, -0.0655, 0.0806, ..., 0.0480, -0.1212, 0.0045], + [ 0.0500, 0.0131, -0.1547, ..., -0.0296, 0.0972, -0.0186], + [-0.1034, -0.0756, 0.0325, ..., -0.0016, -0.0841, -0.0464]], + device='cuda:0'), grad: tensor([[ 9.1502e-07, 1.3923e-07, 5.4715e-07, ..., 0.0000e+00, + 1.0449e-06, 0.0000e+00], + [ 1.3076e-06, -1.9640e-05, -8.4490e-06, ..., 0.0000e+00, + 9.1409e-07, 0.0000e+00], + [-1.0300e-06, 4.5262e-07, -9.8534e-07, ..., 0.0000e+00, + -6.3255e-06, 0.0000e+00], + ..., + [ 2.1100e-05, 7.3761e-06, 1.4544e-05, ..., 0.0000e+00, + 6.8136e-06, 0.0000e+00], + [-2.8491e-05, 4.1863e-07, -1.4871e-05, ..., 0.0000e+00, + -7.9423e-06, 0.0000e+00], + [ 4.3623e-06, 1.0051e-05, 6.5453e-06, ..., 0.0000e+00, + 1.3262e-06, 0.0000e+00]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0049, 0.0187, -0.0168, 0.0143, 0.0222, -0.0151, -0.0228, -0.0143, + -0.0252, -0.0295], device='cuda:0'), grad: tensor([ 4.5002e-06, -4.6968e-05, -8.1509e-06, 2.8107e-06, 4.6343e-06, + 1.9800e-06, 9.0003e-06, 1.1373e-04, -1.2362e-04, 4.2021e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 216.79, cls_loss 0.0032 cls_loss_mapping 0.0056 cls_loss_causal 0.5704 re_mapping 0.0075 re_causal 0.0219 /// teacc 99.00 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.1255, -0.0782, -0.0515, ..., -0.0180, -0.0317, -0.0080], + [-0.0431, 0.0754, -0.0633, ..., 0.0136, -0.0197, -0.0665], + [ 0.0457, -0.0893, -0.0710, ..., -0.0446, 0.0820, -0.0252], + ..., + [-0.0910, -0.0656, 0.0809, ..., 0.0480, -0.1220, 0.0045], + [ 0.0501, 0.0130, -0.1552, ..., -0.0297, 0.0976, -0.0187], + [-0.1040, -0.0757, 0.0326, ..., -0.0016, -0.0846, -0.0465]], + device='cuda:0'), grad: tensor([[ 4.0093e-07, 4.5169e-08, 1.9558e-08, ..., 0.0000e+00, + 3.5716e-07, 0.0000e+00], + [ 2.8871e-07, -1.5236e-06, 4.2375e-08, ..., 0.0000e+00, + 3.2829e-07, 0.0000e+00], + [-3.0315e-07, 1.7276e-07, 1.2107e-08, ..., 0.0000e+00, + -5.7183e-07, 0.0000e+00], + ..., + [ 1.0580e-06, 7.7160e-07, -1.1036e-07, ..., 0.0000e+00, + 1.0915e-06, 0.0000e+00], + [-2.5630e-06, 2.3423e-07, 2.6077e-08, ..., 0.0000e+00, + -3.0082e-06, 0.0000e+00], + [ 8.9779e-07, 2.0815e-07, -6.4727e-08, ..., 0.0000e+00, + 7.3109e-07, 0.0000e+00]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0075, 0.0184, -0.0160, 0.0135, 0.0224, -0.0145, -0.0222, -0.0144, + -0.0253, -0.0283], device='cuda:0'), grad: tensor([-2.3320e-05, -9.6764e-07, 1.1623e-05, 1.9409e-06, 8.7218e-07, + -1.1669e-06, 4.9807e-06, 5.7742e-06, -5.1521e-06, 5.3793e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 216.62, cls_loss 0.0030 cls_loss_mapping 0.0070 cls_loss_causal 0.5318 re_mapping 0.0069 re_causal 0.0213 /// teacc 98.95 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.1264, -0.0786, -0.0515, ..., -0.0180, -0.0320, -0.0072], + [-0.0434, 0.0757, -0.0641, ..., 0.0136, -0.0198, -0.0672], + [ 0.0458, -0.0895, -0.0711, ..., -0.0446, 0.0822, -0.0270], + ..., + [-0.0934, -0.0658, 0.0807, ..., 0.0480, -0.1235, 0.0045], + [ 0.0506, 0.0131, -0.1556, ..., -0.0298, 0.0981, -0.0188], + [-0.1047, -0.0759, 0.0328, ..., -0.0017, -0.0851, -0.0469]], + device='cuda:0'), grad: tensor([[ 2.1653e-07, 2.4214e-07, 2.8498e-07, ..., 4.6566e-10, + 1.7481e-06, 0.0000e+00], + [ 1.1874e-07, 5.4315e-06, 1.8165e-05, ..., 2.3283e-09, + 8.9109e-06, 0.0000e+00], + [-1.1176e-06, 3.5623e-07, 1.8207e-07, ..., 3.2596e-09, + -4.1455e-05, 0.0000e+00], + ..., + [ 1.0198e-07, 1.1232e-06, -1.0990e-06, ..., 1.3970e-09, + 3.3434e-06, 0.0000e+00], + [ 6.4401e-07, 4.9034e-07, 3.8045e-07, ..., 7.9162e-09, + 4.5449e-06, 0.0000e+00], + [ 7.6368e-08, -2.2843e-05, -6.7890e-05, ..., 4.1910e-09, + 1.1278e-06, 0.0000e+00]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0074, 0.0183, -0.0161, 0.0155, 0.0225, -0.0157, -0.0225, -0.0158, + -0.0253, -0.0278], device='cuda:0'), grad: tensor([ 4.1872e-06, 8.3983e-05, -7.5340e-05, 4.7177e-05, 1.5986e-04, + 9.7454e-06, 7.3239e-06, 7.3314e-06, 1.0423e-05, -2.5487e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 216.92, cls_loss 0.0027 cls_loss_mapping 0.0058 cls_loss_causal 0.5431 re_mapping 0.0073 re_causal 0.0219 /// teacc 98.97 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.1270, -0.0790, -0.0515, ..., -0.0180, -0.0321, -0.0071], + [-0.0438, 0.0757, -0.0645, ..., 0.0136, -0.0200, -0.0672], + [ 0.0460, -0.0897, -0.0714, ..., -0.0446, 0.0826, -0.0270], + ..., + [-0.0942, -0.0658, 0.0811, ..., 0.0480, -0.1241, 0.0045], + [ 0.0514, 0.0131, -0.1562, ..., -0.0299, 0.0990, -0.0188], + [-0.1052, -0.0759, 0.0332, ..., -0.0017, -0.0858, -0.0469]], + device='cuda:0'), grad: tensor([[ 3.4183e-05, 3.4899e-05, 1.1502e-07, ..., 0.0000e+00, + 1.3867e-06, 0.0000e+00], + [-6.3106e-06, -1.5247e-04, 7.5577e-07, ..., 0.0000e+00, + -6.9916e-05, 0.0000e+00], + [ 7.7039e-06, 1.3053e-04, 4.4703e-07, ..., 0.0000e+00, + 5.6863e-05, 0.0000e+00], + ..., + [ 8.4983e-07, 4.7684e-06, -1.5404e-06, ..., 0.0000e+00, + 2.4773e-06, 0.0000e+00], + [ 6.0024e-07, 7.9423e-06, 2.6124e-07, ..., 0.0000e+00, + -3.4049e-06, 0.0000e+00], + [ 2.7716e-06, 8.2627e-06, 2.4308e-06, ..., 0.0000e+00, + 2.3171e-06, 0.0000e+00]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0071, 0.0181, -0.0163, 0.0148, 0.0220, -0.0155, -0.0227, -0.0156, + -0.0250, -0.0273], device='cuda:0'), grad: tensor([ 1.1855e-04, -2.9826e-04, 2.6917e-04, 9.1642e-06, -4.8071e-05, + 3.7789e-05, -1.3578e-04, 7.2904e-06, 1.2413e-05, 2.8193e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 216.88, cls_loss 0.0030 cls_loss_mapping 0.0060 cls_loss_causal 0.5564 re_mapping 0.0076 re_causal 0.0221 /// teacc 99.04 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.1277, -0.0791, -0.0518, ..., -0.0179, -0.0324, -0.0071], + [-0.0439, 0.0759, -0.0648, ..., 0.0136, -0.0201, -0.0672], + [ 0.0466, -0.0901, -0.0720, ..., -0.0446, 0.0832, -0.0271], + ..., + [-0.0944, -0.0659, 0.0817, ..., 0.0480, -0.1244, 0.0045], + [ 0.0528, 0.0124, -0.1569, ..., -0.0300, 0.1012, -0.0188], + [-0.1056, -0.0763, 0.0324, ..., -0.0017, -0.0868, -0.0469]], + device='cuda:0'), grad: tensor([[ 1.0896e-07, 3.5996e-07, 1.1967e-06, ..., -7.4506e-09, + 2.8871e-08, 0.0000e+00], + [ 3.6787e-08, -2.2426e-06, 4.3102e-06, ..., 4.6566e-10, + 2.9337e-08, 0.0000e+00], + [ 4.7963e-08, 8.4331e-07, 2.6971e-06, ..., 2.7940e-09, + -7.4040e-08, 0.0000e+00], + ..., + [ 2.2352e-08, 1.2256e-06, 1.5274e-05, ..., 4.6566e-10, + 5.6811e-08, 0.0000e+00], + [ 5.6345e-08, 8.3167e-07, 3.4180e-06, ..., 1.8626e-09, + -3.0128e-07, 0.0000e+00], + [ 1.1548e-07, 7.5102e-06, 1.2326e-04, ..., 3.2596e-09, + 4.8429e-08, 0.0000e+00]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0070, 0.0181, -0.0165, 0.0140, 0.0236, -0.0160, -0.0230, -0.0154, + -0.0237, -0.0282], device='cuda:0'), grad: tensor([ 3.5316e-06, 6.0052e-06, 6.1169e-06, 3.9451e-06, -2.7895e-04, + 5.1558e-06, 1.0319e-05, 6.5804e-05, 8.5384e-06, 1.6975e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 216.94, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.5397 re_mapping 0.0071 re_causal 0.0218 /// teacc 99.04 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.1280, -0.0790, -0.0528, ..., -0.0176, -0.0324, -0.0071], + [-0.0442, 0.0762, -0.0654, ..., 0.0137, -0.0202, -0.0672], + [ 0.0467, -0.0906, -0.0723, ..., -0.0446, 0.0834, -0.0271], + ..., + [-0.0949, -0.0662, 0.0817, ..., 0.0480, -0.1247, 0.0045], + [ 0.0531, 0.0125, -0.1574, ..., -0.0303, 0.1016, -0.0188], + [-0.1063, -0.0765, 0.0326, ..., -0.0020, -0.0876, -0.0469]], + device='cuda:0'), grad: tensor([[ 9.2667e-08, 1.6624e-07, -5.5395e-06, ..., -2.3702e-07, + -1.6671e-06, 0.0000e+00], + [ 5.5414e-08, -2.2408e-06, 3.5064e-07, ..., 6.9849e-09, + 2.1048e-07, 0.0000e+00], + [-6.6264e-07, 2.1933e-07, 1.4016e-07, ..., 2.8405e-08, + -4.3809e-06, 0.0000e+00], + ..., + [ 5.4948e-08, 5.2527e-07, -8.3726e-07, ..., 1.0710e-08, + 2.8638e-07, 0.0000e+00], + [ 2.0023e-08, 5.0897e-07, 2.4773e-06, ..., 5.1223e-09, + 6.6729e-07, 0.0000e+00], + [ 4.5169e-08, 4.3400e-07, 7.3574e-07, ..., 1.5274e-07, + 1.3188e-06, 0.0000e+00]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0078, 0.0180, -0.0168, 0.0141, 0.0239, -0.0162, -0.0232, -0.0158, + -0.0235, -0.0275], device='cuda:0'), grad: tensor([-4.5896e-05, -1.0710e-06, 6.1095e-06, -9.9754e-04, 1.1794e-05, + 9.9277e-04, 5.4389e-07, 1.2973e-06, 1.3970e-05, 1.6674e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 216.86, cls_loss 0.0020 cls_loss_mapping 0.0040 cls_loss_causal 0.5290 re_mapping 0.0068 re_causal 0.0214 /// teacc 99.06 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.1283, -0.0792, -0.0526, ..., -0.0169, -0.0324, -0.0071], + [-0.0448, 0.0763, -0.0657, ..., 0.0137, -0.0207, -0.0672], + [ 0.0474, -0.0907, -0.0726, ..., -0.0446, 0.0844, -0.0271], + ..., + [-0.0952, -0.0664, 0.0823, ..., 0.0480, -0.1249, 0.0045], + [ 0.0529, 0.0122, -0.1584, ..., -0.0305, 0.1014, -0.0188], + [-0.1065, -0.0767, 0.0325, ..., -0.0020, -0.0879, -0.0469]], + device='cuda:0'), grad: tensor([[ 2.0359e-06, 2.1663e-06, 2.0117e-07, ..., 0.0000e+00, + 6.1002e-08, 0.0000e+00], + [ 3.0641e-07, -9.6112e-06, 1.0375e-06, ..., 0.0000e+00, + 1.0384e-07, 0.0000e+00], + [ 3.1618e-07, 2.4457e-06, 3.8557e-07, ..., 0.0000e+00, + -1.8347e-07, 0.0000e+00], + ..., + [ 8.6147e-08, 3.8017e-06, -1.2182e-06, ..., 0.0000e+00, + 1.7323e-07, 0.0000e+00], + [ 8.4713e-06, 5.7705e-06, 2.3423e-07, ..., 0.0000e+00, + 2.1942e-06, 0.0000e+00], + [ 2.3609e-07, 3.2410e-06, 2.4568e-06, ..., 0.0000e+00, + 1.6624e-07, 0.0000e+00]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0070, 0.0179, -0.0164, 0.0141, 0.0241, -0.0165, -0.0236, -0.0158, + -0.0240, -0.0276], device='cuda:0'), grad: tensor([ 4.3549e-06, -1.1176e-05, 4.3511e-06, 1.9036e-06, -1.2524e-05, + 6.0797e-06, -2.2888e-05, 2.9653e-06, 1.7285e-05, 9.5889e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 216.70, cls_loss 0.0021 cls_loss_mapping 0.0052 cls_loss_causal 0.5024 re_mapping 0.0072 re_causal 0.0210 /// teacc 98.92 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.1291, -0.0796, -0.0527, ..., -0.0167, -0.0324, -0.0071], + [-0.0453, 0.0772, -0.0651, ..., 0.0137, -0.0209, -0.0672], + [ 0.0478, -0.0909, -0.0735, ..., -0.0446, 0.0850, -0.0271], + ..., + [-0.0956, -0.0674, 0.0827, ..., 0.0480, -0.1252, 0.0044], + [ 0.0531, 0.0121, -0.1590, ..., -0.0307, 0.1017, -0.0188], + [-0.1076, -0.0771, 0.0324, ..., -0.0021, -0.0891, -0.0469]], + device='cuda:0'), grad: tensor([[ 2.9709e-07, 3.2131e-07, 3.1106e-07, ..., 0.0000e+00, + 5.6345e-07, 0.0000e+00], + [ 5.1875e-07, 8.7544e-08, 7.6788e-07, ..., 0.0000e+00, + 2.3842e-06, 0.0000e+00], + [ 1.0524e-06, 3.0128e-07, 2.5425e-07, ..., 0.0000e+00, + 3.5018e-06, 0.0000e+00], + ..., + [ 2.2957e-07, 7.5903e-08, -2.2836e-06, ..., 0.0000e+00, + 7.3714e-07, 0.0000e+00], + [-8.8587e-06, -4.0345e-06, 1.0412e-06, ..., 0.0000e+00, + -6.4000e-06, 0.0000e+00], + [ 4.0680e-06, 8.0047e-07, -2.0284e-06, ..., 0.0000e+00, + 1.1809e-05, 0.0000e+00]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0070, 0.0185, -0.0163, 0.0140, 0.0243, -0.0166, -0.0238, -0.0161, + -0.0241, -0.0278], device='cuda:0'), grad: tensor([ 1.9744e-07, 1.0967e-05, 1.6630e-05, 9.9659e-05, 7.9274e-06, + -1.7381e-04, 4.2245e-06, -8.7591e-07, -7.1041e-06, 4.2140e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 216.54, cls_loss 0.0025 cls_loss_mapping 0.0065 cls_loss_causal 0.5633 re_mapping 0.0068 re_causal 0.0208 /// teacc 98.87 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.1302, -0.0773, -0.0527, ..., -0.0167, -0.0301, -0.0071], + [-0.0457, 0.0767, -0.0649, ..., 0.0137, -0.0223, -0.0672], + [ 0.0479, -0.0914, -0.0742, ..., -0.0448, 0.0851, -0.0271], + ..., + [-0.0958, -0.0675, 0.0832, ..., 0.0482, -0.1256, 0.0044], + [ 0.0529, 0.0118, -0.1602, ..., -0.0309, 0.1018, -0.0188], + [-0.1080, -0.0777, 0.0324, ..., -0.0021, -0.0897, -0.0469]], + device='cuda:0'), grad: tensor([[ 6.8843e-06, 7.0315e-08, 6.9849e-07, ..., 0.0000e+00, + 8.3148e-06, 0.0000e+00], + [ 3.1572e-07, -2.0675e-07, 6.9570e-07, ..., 0.0000e+00, + 6.3563e-07, 0.0000e+00], + [ 1.8012e-06, 4.2375e-08, 2.6543e-07, ..., 0.0000e+00, + 3.6992e-06, 0.0000e+00], + ..., + [ 9.0618e-07, 6.0536e-08, -2.2743e-06, ..., 0.0000e+00, + 1.4324e-06, 0.0000e+00], + [-3.8743e-05, -2.3236e-07, -2.4084e-06, ..., 0.0000e+00, + -4.5180e-05, 0.0000e+00], + [ 2.5570e-05, 4.7032e-08, 1.1437e-06, ..., 0.0000e+00, + 3.0756e-05, 0.0000e+00]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0051, 0.0175, -0.0167, 0.0136, 0.0244, -0.0156, -0.0245, -0.0160, + -0.0247, -0.0279], device='cuda:0'), grad: tensor([ 2.9594e-05, 3.0436e-06, 1.6555e-05, 7.0858e-04, 4.8280e-06, + -7.1573e-04, 3.0417e-06, 2.2240e-06, -1.6928e-04, 1.1784e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 136---------------------------------------------------- +epoch 136, time 217.34, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5444 re_mapping 0.0065 re_causal 0.0206 /// teacc 99.09 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.1313, -0.0775, -0.0529, ..., -0.0167, -0.0303, -0.0071], + [-0.0467, 0.0768, -0.0650, ..., 0.0137, -0.0228, -0.0673], + [ 0.0495, -0.0916, -0.0746, ..., -0.0448, 0.0865, -0.0271], + ..., + [-0.0962, -0.0676, 0.0826, ..., 0.0482, -0.1274, 0.0044], + [ 0.0526, 0.0120, -0.1610, ..., -0.0310, 0.1016, -0.0188], + [-0.1086, -0.0793, 0.0315, ..., -0.0022, -0.0903, -0.0469]], + device='cuda:0'), grad: tensor([[ 4.1584e-07, 4.9826e-07, 1.8626e-08, ..., 0.0000e+00, + 1.0366e-06, 0.0000e+00], + [-1.8626e-07, -1.0096e-05, 1.0384e-07, ..., 0.0000e+00, + -1.5497e-06, 0.0000e+00], + [ 1.8328e-06, 2.4345e-06, 1.7229e-07, ..., 0.0000e+00, + 4.4666e-06, 0.0000e+00], + ..., + [ 1.0803e-06, 1.9986e-06, 3.4086e-07, ..., 0.0000e+00, + 2.7269e-06, 0.0000e+00], + [ 7.0967e-07, 8.5309e-07, 1.3504e-07, ..., 0.0000e+00, + 1.6140e-06, 4.6566e-10], + [ 1.1269e-07, 2.0675e-06, 2.2817e-07, ..., 0.0000e+00, + -1.5140e-05, 0.0000e+00]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0051, 0.0173, -0.0161, 0.0142, 0.0254, -0.0153, -0.0246, -0.0166, + -0.0249, -0.0286], device='cuda:0'), grad: tensor([ 1.1943e-05, -1.6212e-05, 3.0607e-05, 6.2048e-05, 8.0988e-06, + 5.6267e-05, -1.0980e-06, 1.6659e-05, 1.2934e-05, -1.8120e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 216.71, cls_loss 0.0022 cls_loss_mapping 0.0049 cls_loss_causal 0.5327 re_mapping 0.0065 re_causal 0.0195 /// teacc 98.99 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.1318, -0.0776, -0.0530, ..., -0.0167, -0.0304, -0.0070], + [-0.0475, 0.0781, -0.0646, ..., 0.0137, -0.0227, -0.0673], + [ 0.0498, -0.0921, -0.0752, ..., -0.0449, 0.0870, -0.0272], + ..., + [-0.0964, -0.0681, 0.0831, ..., 0.0483, -0.1277, 0.0044], + [ 0.0527, 0.0120, -0.1618, ..., -0.0310, 0.1020, -0.0188], + [-0.1093, -0.0802, 0.0314, ..., -0.0022, -0.0911, -0.0470]], + device='cuda:0'), grad: tensor([[ 5.1223e-07, 8.6520e-07, 4.7032e-08, ..., 0.0000e+00, + 2.6263e-07, 0.0000e+00], + [ 2.4447e-07, -1.0639e-05, 2.6356e-07, ..., 0.0000e+00, + 1.8813e-07, 0.0000e+00], + [-2.1141e-06, 5.2387e-07, 8.8941e-08, ..., 0.0000e+00, + -1.8599e-06, 0.0000e+00], + ..., + [ 4.5588e-07, 1.4268e-06, -5.1782e-07, ..., 0.0000e+00, + 4.0000e-07, 0.0000e+00], + [ 2.0657e-06, 5.2787e-06, 7.5437e-08, ..., 0.0000e+00, + 4.2794e-07, 0.0000e+00], + [ 2.4727e-07, 4.7451e-07, 8.8057e-07, ..., 0.0000e+00, + -5.5879e-07, 0.0000e+00]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0051, 0.0183, -0.0162, 0.0143, 0.0250, -0.0154, -0.0246, -0.0168, + -0.0250, -0.0289], device='cuda:0'), grad: tensor([ 2.1905e-06, -1.0639e-05, -3.6582e-06, 3.0436e-06, -5.9605e-08, + 4.6715e-06, -2.5127e-06, 2.1653e-07, 8.9854e-06, -2.2575e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 216.64, cls_loss 0.0028 cls_loss_mapping 0.0045 cls_loss_causal 0.5518 re_mapping 0.0067 re_causal 0.0199 /// teacc 99.06 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1321, -0.0779, -0.0530, ..., -0.0163, -0.0305, -0.0070], + [-0.0486, 0.0787, -0.0638, ..., 0.0138, -0.0232, -0.0673], + [ 0.0502, -0.0921, -0.0756, ..., -0.0449, 0.0876, -0.0272], + ..., + [-0.0967, -0.0688, 0.0829, ..., 0.0482, -0.1280, 0.0040], + [ 0.0527, 0.0119, -0.1625, ..., -0.0312, 0.1020, -0.0188], + [-0.1116, -0.0806, 0.0315, ..., -0.0022, -0.0919, -0.0470]], + device='cuda:0'), grad: tensor([[ 8.8587e-06, 2.2259e-07, 1.3737e-07, ..., 0.0000e+00, + 2.6003e-05, 9.3132e-10], + [ 5.3225e-07, -1.7695e-07, 2.2212e-07, ..., 0.0000e+00, + 1.5777e-06, 4.6566e-10], + [-1.3821e-05, 3.7439e-07, 8.1956e-08, ..., 0.0000e+00, + -4.1753e-05, 0.0000e+00], + ..., + [ 6.8638e-07, 9.8255e-08, 8.8476e-09, ..., 0.0000e+00, + 2.2724e-06, 9.3132e-10], + [ 2.3209e-06, 5.7044e-07, 1.5330e-06, ..., 0.0000e+00, + 6.5565e-06, 1.4901e-08], + [ 5.6066e-07, 5.2946e-07, -2.7139e-06, ..., 0.0000e+00, + 1.8366e-06, 3.2596e-09]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0049, 0.0184, -0.0161, 0.0141, 0.0249, -0.0145, -0.0257, -0.0174, + -0.0252, -0.0285], device='cuda:0'), grad: tensor([ 4.6670e-05, 3.7495e-06, -7.4744e-05, -1.1660e-05, 1.6708e-06, + 1.2174e-05, 2.9840e-06, 5.6140e-06, 1.9342e-05, -5.9195e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 216.75, cls_loss 0.0025 cls_loss_mapping 0.0060 cls_loss_causal 0.5292 re_mapping 0.0068 re_causal 0.0203 /// teacc 98.97 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.1340, -0.0780, -0.0530, ..., -0.0151, -0.0300, -0.0070], + [-0.0506, 0.0786, -0.0641, ..., 0.0142, -0.0240, -0.0674], + [ 0.0524, -0.0916, -0.0761, ..., -0.0450, 0.0899, -0.0272], + ..., + [-0.0972, -0.0691, 0.0837, ..., 0.0481, -0.1284, 0.0040], + [ 0.0522, 0.0116, -0.1631, ..., -0.0314, 0.1017, -0.0189], + [-0.1121, -0.0809, 0.0314, ..., -0.0027, -0.0934, -0.0470]], + device='cuda:0'), grad: tensor([[-1.0328e-06, -3.6880e-07, 5.4017e-08, ..., -9.4529e-08, + -1.3020e-06, 0.0000e+00], + [ 6.1467e-08, -3.7858e-07, 2.1560e-07, ..., 2.7940e-09, + 1.0943e-07, 0.0000e+00], + [-8.1956e-08, 3.1432e-07, 2.9663e-07, ..., 9.7789e-09, + -5.7602e-07, 0.0000e+00], + ..., + [ 1.4855e-07, 2.4354e-07, -1.3523e-06, ..., 1.3970e-09, + 2.9756e-07, 0.0000e+00], + [ 3.5856e-07, 5.1549e-07, 1.1362e-07, ..., 2.3283e-09, + 2.5984e-07, 0.0000e+00], + [ 1.8720e-07, 1.4845e-06, 9.3039e-07, ..., 8.8476e-09, + 2.1514e-07, 0.0000e+00]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0047, 0.0176, -0.0140, 0.0145, 0.0248, -0.0153, -0.0257, -0.0169, + -0.0258, -0.0287], device='cuda:0'), grad: tensor([-1.4223e-05, 5.6997e-07, 1.7397e-06, 1.2722e-06, -4.7162e-06, + 5.5181e-07, 7.5325e-06, -3.0547e-06, 3.2149e-06, 7.1339e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 216.74, cls_loss 0.0023 cls_loss_mapping 0.0054 cls_loss_causal 0.5598 re_mapping 0.0065 re_causal 0.0198 /// teacc 98.79 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.1345, -0.0781, -0.0532, ..., -0.0150, -0.0300, -0.0070], + [-0.0512, 0.0789, -0.0639, ..., 0.0143, -0.0242, -0.0674], + [ 0.0526, -0.0918, -0.0770, ..., -0.0451, 0.0901, -0.0272], + ..., + [-0.0980, -0.0695, 0.0839, ..., 0.0480, -0.1287, 0.0040], + [ 0.0526, 0.0112, -0.1638, ..., -0.0323, 0.1021, -0.0189], + [-0.1144, -0.0813, 0.0312, ..., -0.0028, -0.0947, -0.0470]], + device='cuda:0'), grad: tensor([[ 1.5590e-06, 1.5916e-06, 2.7567e-07, ..., 1.8626e-09, + 3.9972e-06, 0.0000e+00], + [ 2.7530e-06, -1.8254e-07, 1.2135e-06, ..., 3.7253e-09, + 7.2420e-06, 0.0000e+00], + [ 1.5408e-05, 1.5318e-05, 7.5437e-07, ..., 9.3132e-10, + 4.0680e-05, 0.0000e+00], + ..., + [ 2.0321e-06, 2.1383e-06, -8.5086e-06, ..., 1.8626e-09, + 5.4166e-06, 0.0000e+00], + [ 3.4925e-06, 3.9749e-06, 2.2352e-07, ..., 1.8626e-08, + 9.8944e-06, 0.0000e+00], + [ 9.0804e-07, 1.0598e-06, 4.1164e-06, ..., 2.7940e-09, + 2.2668e-06, 0.0000e+00]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0046, 0.0176, -0.0143, 0.0143, 0.0250, -0.0150, -0.0259, -0.0168, + -0.0254, -0.0291], device='cuda:0'), grad: tensor([ 1.0408e-05, 1.8790e-05, 1.0163e-04, -2.4843e-04, 8.0466e-05, + 1.0842e-04, 3.3319e-05, 2.5690e-05, 3.0786e-05, -1.6069e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 216.59, cls_loss 0.0024 cls_loss_mapping 0.0041 cls_loss_causal 0.5709 re_mapping 0.0065 re_causal 0.0199 /// teacc 98.92 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1352, -0.0785, -0.0533, ..., -0.0151, -0.0302, -0.0070], + [-0.0525, 0.0790, -0.0643, ..., 0.0144, -0.0248, -0.0674], + [ 0.0531, -0.0925, -0.0778, ..., -0.0454, 0.0908, -0.0272], + ..., + [-0.0982, -0.0696, 0.0843, ..., 0.0478, -0.1290, 0.0040], + [ 0.0528, 0.0107, -0.1646, ..., -0.0335, 0.1024, -0.0189], + [-0.1153, -0.0829, 0.0296, ..., -0.0031, -0.0955, -0.0470]], + device='cuda:0'), grad: tensor([[ 1.4454e-06, 1.9819e-06, 4.0047e-08, ..., 9.3132e-10, + -8.1025e-08, 0.0000e+00], + [ 1.7695e-07, -5.7369e-06, 3.0734e-08, ..., 9.3132e-10, + -4.0233e-07, 0.0000e+00], + [-1.9111e-06, 4.2692e-06, -9.3691e-07, ..., 9.3132e-10, + -2.2240e-06, 0.0000e+00], + ..., + [ 1.7472e-06, 4.3120e-07, 8.1118e-07, ..., 1.8626e-09, + 2.6673e-06, 0.0000e+00], + [ 4.0140e-07, 5.3924e-07, 1.3970e-08, ..., 1.8626e-08, + 4.5635e-08, 0.0000e+00], + [ 1.1642e-07, 1.0710e-07, 1.4901e-08, ..., 3.7253e-09, + 1.1735e-07, 0.0000e+00]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0045, 0.0173, -0.0145, 0.0143, 0.0275, -0.0153, -0.0254, -0.0166, + -0.0254, -0.0308], device='cuda:0'), grad: tensor([-4.3511e-06, -6.8806e-06, 5.0217e-06, -1.8319e-06, 1.5721e-06, + 1.1735e-06, -4.5262e-06, 6.9886e-06, 1.4240e-06, 1.3523e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 216.59, cls_loss 0.0031 cls_loss_mapping 0.0063 cls_loss_causal 0.5457 re_mapping 0.0068 re_causal 0.0190 /// teacc 99.09 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1356, -0.0789, -0.0533, ..., -0.0124, -0.0305, -0.0070], + [-0.0545, 0.0793, -0.0644, ..., 0.0146, -0.0246, -0.0674], + [ 0.0523, -0.0940, -0.0785, ..., -0.0458, 0.0910, -0.0272], + ..., + [-0.0989, -0.0696, 0.0850, ..., 0.0476, -0.1296, 0.0040], + [ 0.0530, 0.0097, -0.1659, ..., -0.0335, 0.1031, -0.0189], + [-0.1158, -0.0829, 0.0286, ..., -0.0033, -0.0962, -0.0470]], + device='cuda:0'), grad: tensor([[ 8.9034e-07, 1.1018e-06, 5.2806e-07, ..., 9.3132e-10, + 2.4773e-07, 0.0000e+00], + [ 5.6904e-07, 1.0148e-05, 3.1143e-05, ..., 6.5193e-09, + 4.2375e-07, 0.0000e+00], + [-5.3868e-06, 3.1963e-06, 3.9637e-06, ..., 2.1420e-08, + -2.2929e-06, 0.0000e+00], + ..., + [ 8.4657e-07, -5.5224e-05, -2.0385e-04, ..., 2.7940e-09, + 5.2154e-07, 0.0000e+00], + [ 1.4519e-06, 2.0452e-06, 5.6550e-06, ..., 9.3132e-10, + 3.3714e-07, 0.0000e+00], + [ 1.2759e-07, 4.3482e-05, 1.3089e-04, ..., 9.3132e-10, + 1.2200e-07, 0.0000e+00]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0042, 0.0174, -0.0152, 0.0137, 0.0269, -0.0153, -0.0240, -0.0167, + -0.0257, -0.0302], device='cuda:0'), grad: tensor([ 3.7476e-06, 1.8263e-04, 9.5442e-06, 1.4508e-04, 1.1027e-05, + 8.8066e-06, 4.2617e-06, -1.1892e-03, 3.5495e-05, 7.8773e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 216.75, cls_loss 0.0026 cls_loss_mapping 0.0044 cls_loss_causal 0.5699 re_mapping 0.0069 re_causal 0.0204 /// teacc 98.96 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1364, -0.0792, -0.0534, ..., -0.0120, -0.0306, -0.0063], + [-0.0545, 0.0800, -0.0646, ..., 0.0148, -0.0245, -0.0677], + [ 0.0522, -0.0953, -0.0792, ..., -0.0461, 0.0912, -0.0274], + ..., + [-0.0993, -0.0699, 0.0856, ..., 0.0477, -0.1299, 0.0040], + [ 0.0534, 0.0091, -0.1669, ..., -0.0338, 0.1035, -0.0191], + [-0.1164, -0.0832, 0.0279, ..., -0.0035, -0.0968, -0.0472]], + device='cuda:0'), grad: tensor([[ 2.1476e-06, 2.4084e-06, 8.9221e-07, ..., 0.0000e+00, + 4.0792e-07, 2.1420e-08], + [-1.4799e-06, -4.3362e-05, -1.8090e-05, ..., 9.3132e-10, + 1.5628e-06, 6.5193e-09], + [-1.4072e-06, 5.9083e-06, 2.9057e-06, ..., 6.5193e-09, + -4.4629e-06, 1.1176e-08], + ..., + [ 3.6396e-06, 3.0294e-05, 8.7097e-06, ..., 2.7940e-09, + 1.2405e-06, 2.8871e-08], + [ 6.5938e-06, 2.5202e-06, 7.9069e-07, ..., 9.3132e-09, + 5.4091e-06, 3.3528e-08], + [ 1.6093e-06, 6.5714e-06, -1.3448e-06, ..., 1.8626e-09, + 5.3458e-07, 2.8871e-08]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0042, 0.0177, -0.0159, 0.0139, 0.0266, -0.0157, -0.0238, -0.0165, + -0.0259, -0.0301], device='cuda:0'), grad: tensor([ 1.0289e-05, -1.5199e-04, 1.7703e-05, 1.0431e-04, 6.2846e-06, + -1.2910e-04, -2.3656e-07, 1.1265e-04, 2.9370e-05, 3.3528e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 216.54, cls_loss 0.0026 cls_loss_mapping 0.0046 cls_loss_causal 0.5330 re_mapping 0.0066 re_causal 0.0196 /// teacc 99.06 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1376, -0.0795, -0.0532, ..., -0.0126, -0.0308, -0.0062], + [-0.0544, 0.0797, -0.0656, ..., 0.0149, -0.0244, -0.0678], + [ 0.0525, -0.0957, -0.0798, ..., -0.0464, 0.0922, -0.0275], + ..., + [-0.1005, -0.0692, 0.0865, ..., 0.0469, -0.1305, 0.0039], + [ 0.0540, 0.0086, -0.1679, ..., -0.0385, 0.1043, -0.0192], + [-0.1171, -0.0846, 0.0272, ..., -0.0042, -0.0975, -0.0473]], + device='cuda:0'), grad: tensor([[ 6.0536e-08, 3.0827e-07, 6.6124e-08, ..., -3.9116e-08, + 5.5041e-07, 0.0000e+00], + [ 5.1223e-08, 6.8359e-06, 1.9558e-07, ..., 7.4506e-09, + 1.7211e-05, 0.0000e+00], + [ 9.3132e-09, -7.7635e-06, 4.8429e-08, ..., 9.3132e-09, + -2.0236e-05, 0.0000e+00], + ..., + [ 1.5832e-08, 3.3714e-07, -8.0094e-08, ..., 6.5193e-09, + 6.3051e-07, 0.0000e+00], + [-3.7253e-08, 2.2724e-07, 1.6298e-07, ..., 2.1420e-08, + 1.5087e-07, 0.0000e+00], + [ 3.9116e-08, 2.5034e-06, 3.2503e-07, ..., 1.7695e-08, + 1.4808e-07, 0.0000e+00]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0040, 0.0172, -0.0157, 0.0133, 0.0270, -0.0160, -0.0239, -0.0158, + -0.0255, -0.0304], device='cuda:0'), grad: tensor([ 3.0883e-06, 6.4790e-05, -6.9976e-05, -1.5438e-05, -4.3958e-06, + 1.6168e-05, 5.0068e-06, 3.0063e-06, 3.1702e-06, -5.5768e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 216.86, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5297 re_mapping 0.0066 re_causal 0.0202 /// teacc 99.01 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.1386, -0.0799, -0.0532, ..., -0.0126, -0.0310, -0.0061], + [-0.0548, 0.0801, -0.0659, ..., 0.0157, -0.0244, -0.0678], + [ 0.0531, -0.0961, -0.0800, ..., -0.0469, 0.0929, -0.0276], + ..., + [-0.1020, -0.0694, 0.0866, ..., 0.0462, -0.1315, 0.0039], + [ 0.0541, 0.0082, -0.1684, ..., -0.0391, 0.1045, -0.0192], + [-0.1173, -0.0850, 0.0273, ..., -0.0045, -0.0990, -0.0474]], + device='cuda:0'), grad: tensor([[ 1.0198e-06, 1.4743e-06, 1.1176e-07, ..., 9.3132e-10, + -6.4261e-08, 0.0000e+00], + [ 1.3690e-07, 1.4454e-06, 1.0962e-06, ..., 2.7940e-09, + 6.2678e-07, 0.0000e+00], + [-5.4948e-07, 9.6112e-07, 4.1816e-07, ..., 8.3819e-09, + -1.1595e-06, 0.0000e+00], + ..., + [ 2.7753e-07, 1.4063e-07, -1.0215e-05, ..., 3.7253e-09, + 8.7731e-07, 0.0000e+00], + [ 1.4612e-06, 1.0692e-06, 7.2643e-08, ..., 1.3970e-08, + 9.9372e-07, 0.0000e+00], + [ 1.8720e-07, 7.5996e-07, 8.8960e-06, ..., 1.8626e-09, + 2.1420e-07, 0.0000e+00]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0039, 0.0173, -0.0151, 0.0133, 0.0269, -0.0159, -0.0234, -0.0164, + -0.0258, -0.0304], device='cuda:0'), grad: tensor([ 2.6263e-06, 5.5358e-06, 1.4547e-06, -5.2564e-06, -4.9174e-06, + 1.6198e-05, -1.7941e-05, -3.3826e-05, 4.2282e-06, 3.1859e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 216.61, cls_loss 0.0020 cls_loss_mapping 0.0052 cls_loss_causal 0.5083 re_mapping 0.0069 re_causal 0.0203 /// teacc 98.80 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.1395, -0.0800, -0.0534, ..., -0.0126, -0.0313, -0.0060], + [-0.0554, 0.0807, -0.0664, ..., 0.0157, -0.0246, -0.0678], + [ 0.0522, -0.0966, -0.0805, ..., -0.0469, 0.0927, -0.0276], + ..., + [-0.1030, -0.0698, 0.0872, ..., 0.0462, -0.1324, 0.0039], + [ 0.0548, 0.0072, -0.1686, ..., -0.0392, 0.1059, -0.0192], + [-0.1182, -0.0856, 0.0273, ..., -0.0045, -0.1002, -0.0474]], + device='cuda:0'), grad: tensor([[ 3.2317e-07, 4.7684e-07, 7.5772e-06, ..., 0.0000e+00, + 1.4249e-07, 0.0000e+00], + [ 3.6322e-08, -1.0459e-06, 1.1399e-06, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [-7.0501e-07, 1.8720e-07, 6.1095e-07, ..., 0.0000e+00, + -4.5914e-07, 0.0000e+00], + ..., + [ 5.3924e-07, 1.8710e-06, 2.0452e-06, ..., 0.0000e+00, + 3.7532e-07, 0.0000e+00], + [-1.0384e-06, 4.9267e-07, 5.0571e-07, ..., 0.0000e+00, + -8.0373e-07, 9.3132e-10], + [ 5.1223e-08, 1.8299e-05, 1.0043e-05, ..., 0.0000e+00, + 9.3132e-08, 0.0000e+00]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0041, 0.0176, -0.0153, 0.0130, 0.0269, -0.0162, -0.0228, -0.0167, + -0.0252, -0.0303], device='cuda:0'), grad: tensor([ 3.0726e-05, 2.0321e-06, 2.3209e-06, 2.0992e-06, -8.3625e-05, + 2.1718e-06, 5.1372e-06, 8.6501e-06, -1.5553e-07, 3.0726e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 216.65, cls_loss 0.0031 cls_loss_mapping 0.0065 cls_loss_causal 0.5393 re_mapping 0.0069 re_causal 0.0203 /// teacc 99.06 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1414, -0.0806, -0.0535, ..., -0.0128, -0.0314, -0.0060], + [-0.0553, 0.0825, -0.0660, ..., 0.0158, -0.0254, -0.0678], + [ 0.0527, -0.0964, -0.0808, ..., -0.0470, 0.0937, -0.0276], + ..., + [-0.1043, -0.0716, 0.0847, ..., 0.0462, -0.1334, 0.0039], + [ 0.0555, 0.0069, -0.1714, ..., -0.0397, 0.1064, -0.0193], + [-0.1206, -0.0867, 0.0292, ..., -0.0047, -0.1017, -0.0475]], + device='cuda:0'), grad: tensor([[ 5.6345e-07, 4.9639e-07, 3.2410e-07, ..., 3.7253e-09, + 4.2003e-07, 0.0000e+00], + [ 9.4324e-06, 8.6725e-06, 2.0396e-07, ..., 9.3132e-10, + 6.8769e-06, 0.0000e+00], + [-2.2445e-07, 4.1816e-07, 2.0210e-07, ..., 9.3132e-09, + 1.9744e-07, 0.0000e+00], + ..., + [ 4.1630e-07, 3.3900e-07, -1.6363e-06, ..., 5.5879e-09, + 5.8394e-07, 0.0000e+00], + [-8.2552e-05, -7.5102e-05, 6.6124e-08, ..., 8.3819e-09, + -5.9336e-05, 0.0000e+00], + [ 1.2200e-06, 1.1260e-06, 3.9767e-07, ..., 1.5832e-08, + 9.6299e-07, 9.3132e-10]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0043, 0.0187, -0.0137, 0.0130, 0.0269, -0.0165, -0.0221, -0.0200, + -0.0252, -0.0284], device='cuda:0'), grad: tensor([ 9.5740e-07, 2.4885e-05, 1.7863e-06, 7.6890e-06, 7.0222e-07, + 4.9211e-06, 1.6630e-04, -7.5847e-06, -2.0576e-04, 6.2026e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 216.76, cls_loss 0.0021 cls_loss_mapping 0.0048 cls_loss_causal 0.5181 re_mapping 0.0067 re_causal 0.0199 /// teacc 99.06 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1431, -0.0815, -0.0536, ..., -0.0128, -0.0315, -0.0059], + [-0.0557, 0.0826, -0.0662, ..., 0.0159, -0.0249, -0.0679], + [ 0.0526, -0.0974, -0.0812, ..., -0.0482, 0.0914, -0.0278], + ..., + [-0.1046, -0.0716, 0.0854, ..., 0.0461, -0.1338, 0.0039], + [ 0.0558, 0.0068, -0.1726, ..., -0.0398, 0.1068, -0.0193], + [-0.1231, -0.0888, 0.0291, ..., -0.0048, -0.1034, -0.0476]], + device='cuda:0'), grad: tensor([[-6.6832e-06, -1.9558e-08, 1.3039e-08, ..., -8.0094e-08, + -1.1414e-05, 0.0000e+00], + [ 5.8673e-08, -1.2191e-06, 6.5193e-08, ..., 2.7940e-09, + 1.1083e-07, 0.0000e+00], + [ 2.9337e-07, 9.5926e-08, 9.3132e-09, ..., 6.5193e-09, + 4.2282e-07, 0.0000e+00], + ..., + [ 2.7940e-08, 5.6066e-07, -9.9652e-08, ..., 1.8626e-09, + 5.9605e-08, 0.0000e+00], + [ 2.3134e-06, 1.5739e-07, 6.5658e-07, ..., 2.7940e-09, + 3.8482e-06, 0.0000e+00], + [ 7.7300e-08, 6.3330e-07, -6.8638e-07, ..., 5.5879e-09, + 6.9849e-08, 0.0000e+00]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0044, 0.0189, -0.0163, 0.0138, 0.0259, -0.0165, -0.0197, -0.0193, + -0.0253, -0.0289], device='cuda:0'), grad: tensor([-3.9220e-05, -6.6124e-07, 1.9334e-06, 1.2061e-06, -6.9477e-07, + 7.1153e-07, 2.3454e-05, 8.2143e-07, 1.3702e-05, -1.1967e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 216.63, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5383 re_mapping 0.0067 re_causal 0.0200 /// teacc 98.90 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1444, -0.0802, -0.0539, ..., -0.0127, -0.0296, -0.0059], + [-0.0562, 0.0826, -0.0663, ..., 0.0159, -0.0252, -0.0679], + [ 0.0526, -0.0980, -0.0811, ..., -0.0482, 0.0916, -0.0278], + ..., + [-0.1050, -0.0716, 0.0855, ..., 0.0461, -0.1347, 0.0039], + [ 0.0564, 0.0072, -0.1732, ..., -0.0399, 0.1076, -0.0194], + [-0.1232, -0.0890, 0.0291, ..., -0.0048, -0.1040, -0.0476]], + device='cuda:0'), grad: tensor([[ 2.8871e-07, 1.2480e-07, 7.0781e-08, ..., 0.0000e+00, + 2.3190e-07, 0.0000e+00], + [ 1.6205e-07, 2.4214e-08, 1.4342e-07, ..., 0.0000e+00, + 4.1444e-07, 0.0000e+00], + [-5.9698e-07, 2.3283e-08, 5.2154e-08, ..., 0.0000e+00, + -9.4902e-07, 0.0000e+00], + ..., + [ 2.5239e-07, 6.2399e-08, -4.8708e-07, ..., 0.0000e+00, + 3.7905e-07, 0.0000e+00], + [ 8.9686e-07, 4.9360e-08, 2.6077e-08, ..., 0.0000e+00, + 8.5402e-07, 0.0000e+00], + [ 1.2014e-07, 3.7346e-07, 2.6170e-07, ..., 0.0000e+00, + 1.4249e-07, 0.0000e+00]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0035, 0.0188, -0.0164, 0.0135, 0.0257, -0.0165, -0.0204, -0.0193, + -0.0248, -0.0288], device='cuda:0'), grad: tensor([ 2.4308e-07, 1.3961e-06, -1.1949e-06, 3.3081e-05, -7.2736e-07, + -3.9548e-05, 7.3016e-07, 3.3993e-07, 3.4291e-06, 2.2538e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 216.66, cls_loss 0.0021 cls_loss_mapping 0.0047 cls_loss_causal 0.5462 re_mapping 0.0062 re_causal 0.0189 /// teacc 99.02 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1452, -0.0804, -0.0542, ..., -0.0126, -0.0299, -0.0060], + [-0.0566, 0.0827, -0.0667, ..., 0.0161, -0.0255, -0.0681], + [ 0.0529, -0.0980, -0.0815, ..., -0.0482, 0.0920, -0.0278], + ..., + [-0.1053, -0.0716, 0.0858, ..., 0.0461, -0.1350, 0.0039], + [ 0.0560, 0.0069, -0.1743, ..., -0.0402, 0.1070, -0.0201], + [-0.1235, -0.0878, 0.0292, ..., -0.0048, -0.1043, -0.0477]], + device='cuda:0'), grad: tensor([[ 3.3900e-07, 1.3404e-05, 2.5686e-06, ..., 0.0000e+00, + 1.1235e-05, 4.7497e-08], + [ 7.3574e-08, -3.8564e-05, -4.2133e-06, ..., 9.3132e-10, + -2.5481e-05, 1.1176e-08], + [ 5.4017e-08, 9.8906e-07, 6.1002e-07, ..., 1.8626e-09, + 3.9954e-07, 1.1176e-08], + ..., + [ 2.4494e-07, 3.4645e-06, 6.9439e-06, ..., 9.3132e-10, + 1.6326e-06, 3.6322e-08], + [ 4.7311e-07, 8.3223e-06, 1.9111e-06, ..., 2.7940e-09, + 7.0222e-06, 6.9849e-08], + [ 4.1258e-07, 3.8520e-06, -1.1355e-04, ..., 9.3132e-10, + 1.1371e-06, 6.3330e-08]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0035, 0.0187, -0.0164, 0.0131, 0.0248, -0.0157, -0.0206, -0.0191, + -0.0257, -0.0281], device='cuda:0'), grad: tensor([ 3.7253e-05, -9.8705e-05, 4.6529e-06, 8.7768e-06, 6.0737e-05, + 1.3006e-04, 1.5453e-05, 1.3709e-05, 2.9683e-05, -2.0182e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 216.87, cls_loss 0.0022 cls_loss_mapping 0.0044 cls_loss_causal 0.5404 re_mapping 0.0062 re_causal 0.0183 /// teacc 99.03 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1467, -0.0807, -0.0544, ..., -0.0127, -0.0306, -0.0061], + [-0.0569, 0.0829, -0.0669, ..., 0.0162, -0.0255, -0.0689], + [ 0.0540, -0.0982, -0.0819, ..., -0.0483, 0.0932, -0.0282], + ..., + [-0.1066, -0.0716, 0.0863, ..., 0.0461, -0.1362, 0.0038], + [ 0.0565, 0.0069, -0.1752, ..., -0.0405, 0.1074, -0.0209], + [-0.1240, -0.0884, 0.0290, ..., -0.0049, -0.1060, -0.0482]], + device='cuda:0'), grad: tensor([[ 1.8626e-07, 4.2561e-07, 1.0431e-07, ..., 0.0000e+00, + 6.4261e-08, -9.3132e-10], + [ 4.4703e-08, 5.6624e-07, 6.5751e-07, ..., 0.0000e+00, + 1.2107e-07, 0.0000e+00], + [-2.7940e-09, 3.0547e-07, 1.7881e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 2.4214e-08, 1.2312e-06, 4.0513e-07, ..., 0.0000e+00, + 1.0245e-07, 0.0000e+00], + [ 1.2200e-07, 6.0070e-07, 5.3737e-07, ..., 0.0000e+00, + 1.3504e-07, 0.0000e+00], + [ 7.5437e-08, 2.9951e-06, 5.8860e-07, ..., 0.0000e+00, + 1.1828e-07, 0.0000e+00]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0038, 0.0186, -0.0155, 0.0131, 0.0250, -0.0157, -0.0210, -0.0187, + -0.0257, -0.0286], device='cuda:0'), grad: tensor([ 1.1586e-06, 3.3118e-06, 1.3085e-06, 4.4610e-07, -1.2890e-05, + -5.6103e-06, 2.3171e-06, 3.7588e-06, 3.7886e-06, 2.3656e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 216.77, cls_loss 0.0023 cls_loss_mapping 0.0049 cls_loss_causal 0.5409 re_mapping 0.0066 re_causal 0.0191 /// teacc 98.89 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.1507, -0.0811, -0.0546, ..., -0.0127, -0.0306, -0.0058], + [-0.0580, 0.0831, -0.0669, ..., 0.0162, -0.0254, -0.0702], + [ 0.0561, -0.0987, -0.0835, ..., -0.0483, 0.0943, -0.0283], + ..., + [-0.1070, -0.0718, 0.0865, ..., 0.0461, -0.1365, 0.0038], + [ 0.0548, 0.0053, -0.1757, ..., -0.0407, 0.1071, -0.0223], + [-0.1245, -0.0886, 0.0289, ..., -0.0049, -0.1073, -0.0488]], + device='cuda:0'), grad: tensor([[ 3.6322e-07, 6.5193e-08, 2.2352e-08, ..., 9.3132e-10, + 6.6496e-07, 0.0000e+00], + [ 1.4137e-06, 6.7241e-07, 5.5879e-08, ..., 9.3132e-10, + 2.0340e-06, 0.0000e+00], + [ 3.2242e-06, 8.5309e-07, 6.2957e-07, ..., 0.0000e+00, + 5.6252e-06, 0.0000e+00], + ..., + [ 3.1292e-06, 8.1956e-08, -5.9418e-07, ..., 9.3132e-10, + 5.1558e-06, 0.0000e+00], + [-2.7195e-06, 1.8952e-06, 6.4261e-08, ..., 9.3132e-10, + 1.1334e-06, 0.0000e+00], + [ 2.0899e-06, 4.6194e-07, -3.8650e-07, ..., 9.3132e-10, + 1.7183e-06, 0.0000e+00]], device='cuda:0') +Epoch 154, bias, value: tensor([-0.0045, 0.0188, -0.0152, 0.0134, 0.0250, -0.0157, -0.0197, -0.0186, + -0.0272, -0.0288], device='cuda:0'), grad: tensor([ 2.4103e-06, 7.6666e-06, 2.0683e-05, -6.7234e-05, 1.3234e-06, + -3.2745e-06, 6.2399e-06, 1.5512e-05, 1.2338e-05, 4.3884e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 216.86, cls_loss 0.0024 cls_loss_mapping 0.0053 cls_loss_causal 0.5473 re_mapping 0.0064 re_causal 0.0201 /// teacc 98.83 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1509, -0.0812, -0.0546, ..., -0.0127, -0.0292, -0.0058], + [-0.0587, 0.0831, -0.0670, ..., 0.0162, -0.0260, -0.0702], + [ 0.0572, -0.0988, -0.0829, ..., -0.0483, 0.0956, -0.0283], + ..., + [-0.1075, -0.0718, 0.0869, ..., 0.0461, -0.1368, 0.0038], + [ 0.0557, 0.0053, -0.1764, ..., -0.0407, 0.1088, -0.0223], + [-0.1252, -0.0900, 0.0278, ..., -0.0049, -0.1099, -0.0488]], + device='cuda:0'), grad: tensor([[ 1.0058e-07, 1.0524e-07, 1.4808e-07, ..., 0.0000e+00, + 8.5682e-08, 0.0000e+00], + [ 1.6298e-07, -4.2841e-08, 6.5099e-07, ..., 0.0000e+00, + 1.9372e-07, 0.0000e+00], + [-2.2352e-07, 5.3085e-08, 4.8429e-08, ..., 0.0000e+00, + -2.6729e-07, 0.0000e+00], + ..., + [ 3.8184e-08, 2.6356e-07, 3.9488e-07, ..., 0.0000e+00, + 5.5879e-08, 0.0000e+00], + [-4.4517e-07, -1.1642e-07, 1.3802e-06, ..., 0.0000e+00, + -8.5868e-07, 0.0000e+00], + [ 1.9558e-08, -2.7753e-06, -2.0429e-05, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0038, 0.0186, -0.0146, 0.0136, 0.0264, -0.0166, -0.0198, -0.0180, + -0.0259, -0.0310], device='cuda:0'), grad: tensor([ 9.4250e-07, 3.8035e-06, -3.0641e-07, 1.3234e-06, 7.3016e-05, + 7.9256e-07, 7.3388e-07, 1.3495e-06, 5.0515e-06, -8.6725e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 216.74, cls_loss 0.0026 cls_loss_mapping 0.0043 cls_loss_causal 0.5153 re_mapping 0.0063 re_causal 0.0197 /// teacc 99.00 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1511, -0.0814, -0.0549, ..., -0.0127, -0.0288, -0.0058], + [-0.0590, 0.0831, -0.0673, ..., 0.0162, -0.0260, -0.0702], + [ 0.0573, -0.0997, -0.0834, ..., -0.0482, 0.0958, -0.0283], + ..., + [-0.1078, -0.0717, 0.0880, ..., 0.0461, -0.1372, 0.0036], + [ 0.0568, 0.0077, -0.1776, ..., -0.0407, 0.1107, -0.0224], + [-0.1277, -0.0921, 0.0279, ..., -0.0049, -0.1130, -0.0488]], + device='cuda:0'), grad: tensor([[ 2.1048e-07, 1.2852e-07, 4.1910e-08, ..., 0.0000e+00, + 2.8219e-07, 9.3132e-10], + [ 1.7695e-07, -2.3507e-06, 5.8673e-08, ..., 0.0000e+00, + 3.7719e-07, 0.0000e+00], + [ 3.3155e-07, 3.0827e-07, 2.1420e-08, ..., 0.0000e+00, + 4.6380e-07, 0.0000e+00], + ..., + [ 3.4180e-07, 6.4261e-07, -8.5682e-08, ..., 0.0000e+00, + 7.7020e-07, 9.3132e-10], + [ 1.5181e-07, 1.3504e-07, 1.9185e-07, ..., 0.0000e+00, + 3.8650e-07, 5.5879e-09], + [ 4.0978e-07, 4.4610e-07, -3.7346e-07, ..., 0.0000e+00, + 4.2468e-07, 9.3132e-10]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0035, 0.0184, -0.0151, 0.0113, 0.0249, -0.0138, -0.0202, -0.0164, + -0.0231, -0.0327], device='cuda:0'), grad: tensor([-2.8033e-07, -1.2927e-06, 3.9339e-06, 3.6329e-05, 1.0952e-06, + -5.3674e-05, 2.2762e-06, 4.9137e-06, 4.1723e-06, 2.4512e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 216.88, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5498 re_mapping 0.0059 re_causal 0.0188 /// teacc 99.00 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1513, -0.0818, -0.0551, ..., -0.0127, -0.0290, -0.0058], + [-0.0591, 0.0826, -0.0677, ..., 0.0177, -0.0267, -0.0702], + [ 0.0576, -0.0995, -0.0838, ..., -0.0489, 0.0969, -0.0283], + ..., + [-0.1084, -0.0714, 0.0882, ..., 0.0455, -0.1381, 0.0036], + [ 0.0566, 0.0073, -0.1781, ..., -0.0408, 0.1106, -0.0225], + [-0.1278, -0.0921, 0.0278, ..., -0.0050, -0.1131, -0.0488]], + device='cuda:0'), grad: tensor([[ 1.0431e-07, -6.7707e-07, 1.4435e-07, ..., 0.0000e+00, + -2.8778e-07, 0.0000e+00], + [ 3.1665e-08, 2.7455e-06, 2.4065e-06, ..., 0.0000e+00, + 1.7509e-07, 0.0000e+00], + [-2.3004e-07, 5.1223e-07, 3.7253e-07, ..., 0.0000e+00, + -1.7509e-07, 0.0000e+00], + ..., + [ 1.0710e-07, -1.9744e-06, -2.4885e-06, ..., 0.0000e+00, + 1.0710e-07, 0.0000e+00], + [-2.4214e-08, 2.4587e-07, 2.1141e-07, ..., 0.0000e+00, + -6.5193e-08, 0.0000e+00], + [ 1.8626e-08, 7.1153e-06, 4.0755e-06, ..., 0.0000e+00, + 5.7742e-08, 0.0000e+00]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0036, 0.0174, -0.0141, 0.0113, 0.0251, -0.0141, -0.0199, -0.0160, + -0.0233, -0.0326], device='cuda:0'), grad: tensor([-4.8503e-06, 1.2033e-05, 1.7639e-06, 1.9222e-06, -3.3706e-05, + 2.7101e-07, 1.4240e-06, -7.9647e-06, 1.0757e-06, 2.8044e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 216.64, cls_loss 0.0018 cls_loss_mapping 0.0044 cls_loss_causal 0.5204 re_mapping 0.0063 re_causal 0.0185 /// teacc 98.95 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.1516, -0.0820, -0.0554, ..., -0.0128, -0.0294, -0.0058], + [-0.0597, 0.0826, -0.0679, ..., 0.0177, -0.0268, -0.0703], + [ 0.0577, -0.0997, -0.0841, ..., -0.0488, 0.0972, -0.0283], + ..., + [-0.1087, -0.0714, 0.0884, ..., 0.0455, -0.1386, 0.0036], + [ 0.0571, 0.0073, -0.1790, ..., -0.0409, 0.1110, -0.0225], + [-0.1280, -0.0924, 0.0278, ..., -0.0050, -0.1133, -0.0489]], + device='cuda:0'), grad: tensor([[ 2.6170e-07, -9.5926e-07, 4.2282e-07, ..., -5.5879e-09, + 8.5682e-07, 0.0000e+00], + [ 1.0617e-07, 5.7090e-07, 2.3376e-06, ..., 0.0000e+00, + 1.5832e-07, 0.0000e+00], + [-2.5090e-06, 1.9744e-07, -8.0001e-07, ..., 0.0000e+00, + -5.5730e-06, 0.0000e+00], + ..., + [ 4.9174e-07, 2.5593e-06, 7.6219e-06, ..., 0.0000e+00, + 6.6403e-07, 0.0000e+00], + [-1.5134e-06, 3.6322e-07, 4.7684e-07, ..., 0.0000e+00, + -9.5647e-07, 0.0000e+00], + [ 2.8089e-06, 3.0473e-06, 1.0654e-05, ..., 9.3132e-10, + 2.2277e-06, 0.0000e+00]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0037, 0.0173, -0.0140, 0.0114, 0.0250, -0.0141, -0.0199, -0.0159, + -0.0232, -0.0327], device='cuda:0'), grad: tensor([-1.6958e-05, 7.2792e-06, -1.0215e-05, 8.2031e-06, -6.7890e-05, + 1.2957e-05, 1.1936e-05, 2.2069e-05, -4.1537e-07, 3.2961e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 216.84, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5213 re_mapping 0.0062 re_causal 0.0182 /// teacc 99.08 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1519, -0.0823, -0.0559, ..., -0.0126, -0.0301, -0.0058], + [-0.0603, 0.0827, -0.0680, ..., 0.0177, -0.0272, -0.0703], + [ 0.0581, -0.0999, -0.0844, ..., -0.0488, 0.0989, -0.0283], + ..., + [-0.1091, -0.0714, 0.0886, ..., 0.0454, -0.1391, 0.0036], + [ 0.0574, 0.0073, -0.1801, ..., -0.0413, 0.1115, -0.0226], + [-0.1281, -0.0926, 0.0277, ..., -0.0051, -0.1134, -0.0489]], + device='cuda:0'), grad: tensor([[ 3.8091e-07, 1.1548e-07, 8.6613e-08, ..., 0.0000e+00, + 1.4715e-07, 0.0000e+00], + [ 5.2266e-06, -3.5390e-07, 1.1828e-06, ..., 0.0000e+00, + 7.4953e-06, 0.0000e+00], + [-1.6451e-05, -6.7800e-07, -2.3954e-06, ..., 0.0000e+00, + -2.4557e-05, 0.0000e+00], + ..., + [ 3.3900e-07, 2.8126e-07, -3.2987e-06, ..., 0.0000e+00, + 9.5740e-07, 0.0000e+00], + [-6.3330e-08, 1.5646e-07, 5.8115e-07, ..., 0.0000e+00, + -6.9384e-07, 0.0000e+00], + [ 4.2655e-07, 2.1234e-07, 1.3690e-06, ..., 0.0000e+00, + 1.1828e-07, 0.0000e+00]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0041, 0.0173, -0.0131, 0.0110, 0.0250, -0.0144, -0.0199, -0.0157, + -0.0232, -0.0326], device='cuda:0'), grad: tensor([ 1.3188e-06, 1.2405e-05, -4.0144e-05, 5.0478e-06, 2.4885e-06, + -6.7838e-06, 2.5257e-05, -5.6475e-06, 1.5246e-06, 4.4927e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 216.89, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5504 re_mapping 0.0061 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.1520, -0.0825, -0.0560, ..., -0.0126, -0.0298, -0.0059], + [-0.0612, 0.0836, -0.0669, ..., 0.0217, -0.0275, -0.0703], + [ 0.0592, -0.1005, -0.0853, ..., -0.0522, 0.1019, -0.0283], + ..., + [-0.1104, -0.0725, 0.0882, ..., 0.0418, -0.1406, 0.0025], + [ 0.0575, 0.0071, -0.1815, ..., -0.0416, 0.1116, -0.0227], + [-0.1283, -0.0925, 0.0281, ..., -0.0052, -0.1136, -0.0489]], + device='cuda:0'), grad: tensor([[ 4.3772e-08, 1.2107e-07, 4.2468e-07, ..., 0.0000e+00, + 4.9546e-07, 0.0000e+00], + [ 4.8429e-08, -3.7812e-07, 2.8498e-07, ..., 0.0000e+00, + 9.0245e-07, 0.0000e+00], + [-7.4506e-08, 1.6764e-07, 2.6114e-06, ..., 0.0000e+00, + 3.3360e-06, 0.0000e+00], + ..., + [ 7.4506e-09, 3.3434e-07, -4.0717e-06, ..., 0.0000e+00, + 4.5169e-07, 0.0000e+00], + [ 4.6566e-09, 1.5460e-07, 6.4261e-08, ..., 0.0000e+00, + 3.0287e-06, 0.0000e+00], + [ 1.3039e-08, 4.1816e-07, 9.0711e-07, ..., 0.0000e+00, + 1.4212e-06, 0.0000e+00]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0032, 0.0180, -0.0106, 0.0097, 0.0248, -0.0145, -0.0198, -0.0164, + -0.0233, -0.0325], device='cuda:0'), grad: tensor([ 3.3975e-06, 2.4252e-06, 2.0742e-05, -2.2918e-05, -2.6152e-06, + 7.2457e-07, 1.8841e-06, -1.8969e-05, 7.6592e-06, 7.6964e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 216.84, cls_loss 0.0020 cls_loss_mapping 0.0045 cls_loss_causal 0.5588 re_mapping 0.0060 re_causal 0.0185 /// teacc 99.01 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1522, -0.0827, -0.0562, ..., -0.0126, -0.0298, -0.0059], + [-0.0612, 0.0841, -0.0667, ..., 0.0219, -0.0268, -0.0703], + [ 0.0592, -0.1018, -0.0900, ..., -0.0525, 0.1018, -0.0283], + ..., + [-0.1100, -0.0729, 0.0883, ..., 0.0417, -0.1411, 0.0025], + [ 0.0575, 0.0070, -0.1825, ..., -0.0417, 0.1114, -0.0227], + [-0.1286, -0.0923, 0.0282, ..., -0.0052, -0.1139, -0.0489]], + device='cuda:0'), grad: tensor([[ 2.4959e-07, 4.1630e-07, 2.2165e-07, ..., 9.3132e-09, + 3.8557e-07, 0.0000e+00], + [ 5.8673e-08, -1.3504e-07, 7.6555e-07, ..., 5.5879e-08, + 3.4831e-07, 0.0000e+00], + [ 2.7847e-07, 5.7742e-07, 1.1912e-06, ..., 2.7493e-06, + 1.3441e-05, 0.0000e+00], + ..., + [ 1.5832e-08, 7.9535e-07, -4.2990e-06, ..., 3.9116e-08, + 2.1793e-07, 0.0000e+00], + [-1.8831e-06, -1.7975e-06, 2.8498e-07, ..., 5.5879e-09, + -3.3379e-06, 0.0000e+00], + [ 4.1071e-07, 2.0228e-06, 3.2000e-06, ..., 1.0245e-08, + 9.2853e-07, 0.0000e+00]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0028, 0.0183, -0.0116, 0.0101, 0.0247, -0.0145, -0.0202, -0.0165, + -0.0235, -0.0323], device='cuda:0'), grad: tensor([ 1.7816e-06, 3.9600e-06, 3.4034e-05, -2.7657e-05, -7.5847e-06, + 3.7476e-06, 2.9784e-06, -2.4915e-05, -4.7609e-06, 1.8388e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 216.82, cls_loss 0.0014 cls_loss_mapping 0.0032 cls_loss_causal 0.4918 re_mapping 0.0063 re_causal 0.0185 /// teacc 99.08 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1527, -0.0832, -0.0562, ..., -0.0126, -0.0308, -0.0059], + [-0.0617, 0.0842, -0.0669, ..., 0.0219, -0.0268, -0.0703], + [ 0.0598, -0.1023, -0.0903, ..., -0.0527, 0.1019, -0.0283], + ..., + [-0.1106, -0.0729, 0.0885, ..., 0.0417, -0.1415, 0.0025], + [ 0.0575, 0.0070, -0.1836, ..., -0.0417, 0.1119, -0.0227], + [-0.1288, -0.0924, 0.0281, ..., -0.0052, -0.1140, -0.0489]], + device='cuda:0'), grad: tensor([[ 6.3237e-07, 3.2317e-07, 2.5425e-07, ..., 0.0000e+00, + 3.9442e-07, 1.3970e-09], + [ 1.6950e-06, 2.4820e-07, 1.6252e-07, ..., 0.0000e+00, + 1.5479e-06, 0.0000e+00], + [ 7.3388e-07, 3.2736e-07, 3.0221e-07, ..., 0.0000e+00, + 7.1619e-07, 0.0000e+00], + ..., + [ 2.9476e-07, 2.6310e-07, -9.9279e-07, ..., 0.0000e+00, + 3.3388e-07, 0.0000e+00], + [-1.6928e-05, -6.4187e-06, 1.2526e-07, ..., 0.0000e+00, + -1.5020e-05, 4.6566e-10], + [ 2.0117e-07, 2.3097e-07, -4.0978e-08, ..., 0.0000e+00, + 2.6682e-07, 0.0000e+00]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0031, 0.0182, -0.0118, 0.0100, 0.0247, -0.0144, -0.0201, -0.0163, + -0.0234, -0.0323], device='cuda:0'), grad: tensor([ 9.4064e-08, 4.7535e-06, 3.4720e-06, 2.9523e-07, 4.4610e-07, + 2.3276e-05, 1.1444e-05, -2.6338e-06, -4.3899e-05, 2.7455e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 216.67, cls_loss 0.0021 cls_loss_mapping 0.0044 cls_loss_causal 0.5491 re_mapping 0.0059 re_causal 0.0181 /// teacc 98.95 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1531, -0.0837, -0.0579, ..., -0.0126, -0.0309, -0.0059], + [-0.0627, 0.0843, -0.0671, ..., 0.0219, -0.0271, -0.0703], + [ 0.0600, -0.1027, -0.0904, ..., -0.0527, 0.1019, -0.0283], + ..., + [-0.1113, -0.0730, 0.0892, ..., 0.0417, -0.1421, 0.0025], + [ 0.0575, 0.0065, -0.1848, ..., -0.0419, 0.1121, -0.0229], + [-0.1289, -0.0928, 0.0280, ..., -0.0052, -0.1141, -0.0489]], + device='cuda:0'), grad: tensor([[ 8.8941e-08, 2.7847e-07, 2.4214e-08, ..., 0.0000e+00, + 2.2855e-06, 0.0000e+00], + [ 7.8231e-08, -4.2357e-06, 2.1281e-07, ..., 0.0000e+00, + 8.2180e-06, 0.0000e+00], + [-1.7928e-07, 4.5309e-07, 3.1898e-07, ..., 0.0000e+00, + 1.5259e-05, 0.0000e+00], + ..., + [ 1.0012e-07, 1.3709e-06, -7.4320e-07, ..., 0.0000e+00, + 2.7135e-05, 0.0000e+00], + [-2.8312e-07, 6.2445e-07, 2.1793e-07, ..., 0.0000e+00, + 3.7625e-06, 4.6566e-10], + [ 1.1269e-07, 5.4296e-07, -3.4645e-07, ..., 0.0000e+00, + 3.1404e-06, 0.0000e+00]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0034, 0.0181, -0.0120, 0.0102, 0.0243, -0.0145, -0.0196, -0.0157, + -0.0236, -0.0324], device='cuda:0'), grad: tensor([ 1.2442e-06, 2.1219e-05, 5.0753e-05, -2.0397e-04, -7.5847e-06, + 3.0696e-05, 7.2606e-06, 7.2062e-05, 1.5661e-05, 1.2614e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 216.77, cls_loss 0.0016 cls_loss_mapping 0.0038 cls_loss_causal 0.4939 re_mapping 0.0060 re_causal 0.0180 /// teacc 99.01 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1536, -0.0844, -0.0576, ..., -0.0126, -0.0314, -0.0059], + [-0.0634, 0.0845, -0.0672, ..., 0.0219, -0.0280, -0.0703], + [ 0.0607, -0.1024, -0.0911, ..., -0.0527, 0.1024, -0.0283], + ..., + [-0.1118, -0.0731, 0.0894, ..., 0.0417, -0.1429, 0.0025], + [ 0.0569, 0.0058, -0.1858, ..., -0.0422, 0.1119, -0.0229], + [-0.1290, -0.0931, 0.0280, ..., -0.0052, -0.1142, -0.0489]], + device='cuda:0'), grad: tensor([[ 3.6322e-08, 4.7032e-08, 9.8720e-08, ..., 0.0000e+00, + 4.7497e-08, 4.6566e-10], + [ 2.6543e-08, -5.7183e-07, 1.0757e-07, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00], + [-1.7602e-07, 1.5227e-07, 1.2061e-07, ..., 0.0000e+00, + -1.8626e-07, 0.0000e+00], + ..., + [ 5.1223e-08, 3.9814e-07, 1.0207e-06, ..., 0.0000e+00, + 8.8010e-08, 0.0000e+00], + [-4.6566e-08, 1.2573e-08, 4.0326e-07, ..., 0.0000e+00, + 1.0710e-08, 4.6566e-10], + [ 3.4925e-08, 1.0338e-07, 1.5870e-05, ..., 0.0000e+00, + 1.5646e-07, 0.0000e+00]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0035, 0.0181, -0.0118, 0.0105, 0.0242, -0.0149, -0.0190, -0.0156, + -0.0239, -0.0323], device='cuda:0'), grad: tensor([ 7.1665e-07, -8.2422e-08, 5.7789e-07, -5.4389e-07, -1.2022e-04, + 3.1525e-07, 5.4715e-07, 7.3351e-06, 2.2426e-06, 1.0890e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 216.80, cls_loss 0.0017 cls_loss_mapping 0.0035 cls_loss_causal 0.5275 re_mapping 0.0061 re_causal 0.0186 /// teacc 99.06 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1543, -0.0852, -0.0578, ..., -0.0126, -0.0318, -0.0059], + [-0.0639, 0.0850, -0.0673, ..., 0.0219, -0.0280, -0.0703], + [ 0.0609, -0.1030, -0.0914, ..., -0.0527, 0.1024, -0.0283], + ..., + [-0.1122, -0.0732, 0.0894, ..., 0.0417, -0.1443, 0.0025], + [ 0.0562, 0.0050, -0.1876, ..., -0.0422, 0.1116, -0.0229], + [-0.1292, -0.0933, 0.0282, ..., -0.0052, -0.1142, -0.0489]], + device='cuda:0'), grad: tensor([[ 4.7684e-05, 5.0329e-06, 4.9826e-08, ..., 0.0000e+00, + 1.1653e-04, 0.0000e+00], + [-1.2979e-05, -3.5256e-05, -8.3223e-06, ..., 0.0000e+00, + -1.5587e-05, 0.0000e+00], + [ 2.0698e-05, 2.5034e-05, 5.9307e-06, ..., 0.0000e+00, + 4.0412e-05, 0.0000e+00], + ..., + [ 4.9621e-06, 8.2850e-06, 1.6233e-06, ..., 0.0000e+00, + 8.3297e-06, 0.0000e+00], + [-9.3520e-05, 1.9670e-06, 8.7079e-08, ..., 0.0000e+00, + -2.5344e-04, 0.0000e+00], + [ 1.0423e-05, 3.7160e-07, -1.2340e-07, ..., 0.0000e+00, + 2.7061e-05, 0.0000e+00]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0036, 0.0184, -0.0120, 0.0109, 0.0240, -0.0141, -0.0196, -0.0158, + -0.0244, -0.0321], device='cuda:0'), grad: tensor([ 2.4366e-04, -6.6102e-05, 1.0705e-04, 1.2839e-04, 1.2860e-05, + 7.0810e-05, -8.2076e-05, 2.4706e-05, -4.9353e-04, 5.4181e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 217.03, cls_loss 0.0019 cls_loss_mapping 0.0052 cls_loss_causal 0.5109 re_mapping 0.0059 re_causal 0.0174 /// teacc 98.98 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1551, -0.0857, -0.0578, ..., -0.0126, -0.0322, -0.0059], + [-0.0642, 0.0861, -0.0674, ..., 0.0219, -0.0300, -0.0703], + [ 0.0608, -0.1035, -0.0917, ..., -0.0527, 0.1023, -0.0283], + ..., + [-0.1129, -0.0733, 0.0895, ..., 0.0417, -0.1446, 0.0025], + [ 0.0558, 0.0035, -0.1886, ..., -0.0422, 0.1127, -0.0230], + [-0.1295, -0.0934, 0.0283, ..., -0.0053, -0.1144, -0.0489]], + device='cuda:0'), grad: tensor([[ 9.0012e-07, 5.1223e-08, 1.1642e-08, ..., 0.0000e+00, + 2.5518e-07, 0.0000e+00], + [ 3.8324e-07, -6.9756e-07, 3.3528e-08, ..., 0.0000e+00, + 3.8883e-07, 0.0000e+00], + [ 1.1809e-06, 1.0431e-07, 1.0245e-08, ..., 0.0000e+00, + 9.8627e-07, 0.0000e+00], + ..., + [ 4.2468e-07, 7.4320e-07, 2.9895e-07, ..., 0.0000e+00, + 3.1618e-07, 0.0000e+00], + [-1.0636e-06, 9.9186e-08, 7.4040e-08, ..., 0.0000e+00, + -1.5022e-06, 0.0000e+00], + [ 1.2051e-06, 2.9337e-07, -3.6601e-07, ..., 0.0000e+00, + 4.1630e-07, 0.0000e+00]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0037, 0.0191, -0.0123, 0.0118, 0.0238, -0.0147, -0.0195, -0.0157, + -0.0252, -0.0320], device='cuda:0'), grad: tensor([ 2.8610e-06, 1.4994e-06, 3.3509e-06, 2.4721e-05, -5.4762e-07, + -5.5194e-05, 2.2855e-06, 5.0701e-06, 2.6040e-06, 1.3344e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 216.49, cls_loss 0.0017 cls_loss_mapping 0.0032 cls_loss_causal 0.5394 re_mapping 0.0060 re_causal 0.0179 /// teacc 99.04 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1553, -0.0856, -0.0580, ..., -0.0126, -0.0318, -0.0059], + [-0.0643, 0.0863, -0.0676, ..., 0.0219, -0.0303, -0.0703], + [ 0.0613, -0.1036, -0.0942, ..., -0.0527, 0.1025, -0.0283], + ..., + [-0.1134, -0.0734, 0.0904, ..., 0.0417, -0.1451, 0.0025], + [ 0.0559, 0.0034, -0.1902, ..., -0.0422, 0.1128, -0.0230], + [-0.1297, -0.0937, 0.0279, ..., -0.0053, -0.1145, -0.0489]], + device='cuda:0'), grad: tensor([[ 4.8727e-06, 1.9912e-06, 3.8650e-08, ..., 0.0000e+00, + 5.4538e-06, 0.0000e+00], + [-1.5441e-06, -2.1502e-05, 3.2131e-08, ..., 0.0000e+00, + 1.3933e-06, 0.0000e+00], + [-1.3843e-05, 1.1418e-06, 8.5216e-08, ..., 0.0000e+00, + -1.1995e-05, 0.0000e+00], + ..., + [ 4.8354e-06, 1.7658e-06, 3.6974e-07, ..., 0.0000e+00, + 4.2394e-06, 0.0000e+00], + [-1.6391e-05, -2.3544e-06, -2.6869e-07, ..., 0.0000e+00, + -2.0206e-05, 0.0000e+00], + [ 1.7611e-06, 1.7360e-06, 1.5581e-06, ..., 0.0000e+00, + 2.1178e-06, 0.0000e+00]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0032, 0.0190, -0.0131, 0.0115, 0.0239, -0.0149, -0.0196, -0.0147, + -0.0253, -0.0324], device='cuda:0'), grad: tensor([ 1.2107e-05, -2.6867e-05, -2.8878e-05, 3.6955e-05, 6.2026e-07, + 1.6034e-05, 1.2487e-05, 1.3791e-05, -4.6253e-05, 1.0081e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 216.84, cls_loss 0.0020 cls_loss_mapping 0.0027 cls_loss_causal 0.5250 re_mapping 0.0061 re_causal 0.0182 /// teacc 98.99 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1560, -0.0863, -0.0586, ..., -0.0126, -0.0320, -0.0059], + [-0.0652, 0.0858, -0.0676, ..., 0.0219, -0.0312, -0.0703], + [ 0.0634, -0.1022, -0.0947, ..., -0.0527, 0.1035, -0.0283], + ..., + [-0.1144, -0.0736, 0.0904, ..., 0.0417, -0.1457, 0.0025], + [ 0.0551, 0.0027, -0.1926, ..., -0.0422, 0.1124, -0.0230], + [-0.1299, -0.0951, 0.0278, ..., -0.0053, -0.1145, -0.0489]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 7.0315e-08, 1.6764e-08, ..., 0.0000e+00, + 2.0489e-07, 0.0000e+00], + [ 1.2107e-08, -4.6007e-07, 8.9873e-08, ..., 9.3132e-10, + 1.5693e-07, 0.0000e+00], + [-1.7695e-08, 1.2433e-07, 2.0023e-08, ..., 2.3283e-09, + 1.4761e-07, 0.0000e+00], + ..., + [ 2.2352e-08, 2.0443e-07, -1.6857e-07, ..., 4.6566e-10, + 1.4435e-07, 0.0000e+00], + [ 2.4680e-07, 1.6112e-07, 3.7719e-08, ..., 1.3970e-09, + 4.3912e-07, 0.0000e+00], + [ 4.4703e-08, 2.1746e-07, -1.9278e-07, ..., 4.6566e-10, + 4.2096e-07, 0.0000e+00]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0037, 0.0184, -0.0118, 0.0112, 0.0246, -0.0151, -0.0186, -0.0147, + -0.0259, -0.0326], device='cuda:0'), grad: tensor([ 1.2722e-06, 5.8999e-07, 1.1679e-06, -3.5971e-05, -4.3306e-07, + 2.9773e-05, -1.2387e-07, -4.7777e-07, 2.2948e-06, 1.8459e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 216.95, cls_loss 0.0019 cls_loss_mapping 0.0039 cls_loss_causal 0.5135 re_mapping 0.0064 re_causal 0.0181 /// teacc 99.01 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1564, -0.0869, -0.0587, ..., -0.0124, -0.0326, -0.0059], + [-0.0653, 0.0868, -0.0678, ..., 0.0219, -0.0309, -0.0704], + [ 0.0639, -0.1043, -0.0952, ..., -0.0528, 0.1035, -0.0283], + ..., + [-0.1155, -0.0737, 0.0905, ..., 0.0417, -0.1468, 0.0024], + [ 0.0553, 0.0028, -0.1937, ..., -0.0427, 0.1126, -0.0232], + [-0.1306, -0.0954, 0.0277, ..., -0.0055, -0.1170, -0.0489]], + device='cuda:0'), grad: tensor([[ 1.3020e-06, 1.0282e-06, 3.2131e-08, ..., 0.0000e+00, + 6.6683e-07, 0.0000e+00], + [ 4.0280e-07, -9.8255e-08, 9.7323e-08, ..., 0.0000e+00, + 5.1735e-07, 0.0000e+00], + [-1.8384e-06, 2.0210e-07, -7.3994e-07, ..., 0.0000e+00, + -4.9844e-06, 0.0000e+00], + ..., + [ 3.9488e-07, 1.1828e-07, -1.8720e-07, ..., 0.0000e+00, + 5.7369e-07, 0.0000e+00], + [-6.7148e-07, -1.0859e-06, 7.9162e-09, ..., 0.0000e+00, + -3.1069e-06, 0.0000e+00], + [ 2.1365e-06, 2.2724e-07, 2.0489e-08, ..., 0.0000e+00, + -1.8207e-07, 0.0000e+00]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0040, 0.0193, -0.0134, 0.0126, 0.0248, -0.0152, -0.0188, -0.0147, + -0.0259, -0.0332], device='cuda:0'), grad: tensor([ 3.7700e-06, 2.4326e-06, -2.9057e-06, 5.5730e-06, 4.8243e-06, + -2.0668e-05, -3.4906e-06, 2.3991e-06, -1.0850e-07, 8.1062e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 216.92, cls_loss 0.0019 cls_loss_mapping 0.0033 cls_loss_causal 0.5550 re_mapping 0.0061 re_causal 0.0180 /// teacc 98.90 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1567, -0.0867, -0.0588, ..., -0.0120, -0.0314, -0.0059], + [-0.0658, 0.0875, -0.0678, ..., 0.0219, -0.0318, -0.0704], + [ 0.0642, -0.1053, -0.0954, ..., -0.0528, 0.1033, -0.0283], + ..., + [-0.1164, -0.0742, 0.0902, ..., 0.0417, -0.1480, 0.0024], + [ 0.0555, 0.0030, -0.1952, ..., -0.0427, 0.1130, -0.0232], + [-0.1311, -0.0956, 0.0279, ..., -0.0055, -0.1173, -0.0489]], + device='cuda:0'), grad: tensor([[ 1.5507e-07, 8.6650e-06, 5.3784e-07, ..., 0.0000e+00, + 1.4435e-08, 0.0000e+00], + [ 1.3411e-07, -2.6450e-07, 1.0310e-06, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [ 4.3772e-08, 3.4366e-07, 1.2759e-07, ..., 0.0000e+00, + -6.0536e-09, 0.0000e+00], + ..., + [-2.1420e-08, 4.8475e-07, 1.5378e-05, ..., 0.0000e+00, + 1.7229e-08, 0.0000e+00], + [-8.3726e-07, 1.2573e-07, 3.5353e-06, ..., 0.0000e+00, + -8.5356e-07, 0.0000e+00], + [ 2.1746e-07, 1.3923e-07, -2.5854e-05, ..., 0.0000e+00, + 1.6345e-07, 0.0000e+00]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0032, 0.0198, -0.0143, 0.0128, 0.0246, -0.0150, -0.0189, -0.0152, + -0.0259, -0.0330], device='cuda:0'), grad: tensor([ 1.6585e-05, 2.5034e-06, 8.6380e-07, 6.3293e-06, 1.2770e-05, + 4.0680e-06, -2.9057e-05, 2.8148e-05, 5.8115e-06, -4.8041e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 216.75, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.4974 re_mapping 0.0061 re_causal 0.0179 /// teacc 98.97 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1573, -0.0872, -0.0597, ..., -0.0120, -0.0320, -0.0059], + [-0.0660, 0.0874, -0.0678, ..., 0.0219, -0.0319, -0.0704], + [ 0.0657, -0.1054, -0.0949, ..., -0.0528, 0.1036, -0.0283], + ..., + [-0.1190, -0.0744, 0.0898, ..., 0.0417, -0.1498, 0.0024], + [ 0.0554, 0.0030, -0.1979, ..., -0.0427, 0.1131, -0.0232], + [-0.1316, -0.0961, 0.0280, ..., -0.0055, -0.1175, -0.0489]], + device='cuda:0'), grad: tensor([[ 3.5856e-08, 2.9802e-08, 4.3772e-08, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00], + [ 5.3234e-06, 6.6916e-07, 6.4448e-07, ..., 0.0000e+00, + 5.8487e-06, 0.0000e+00], + [ 2.4540e-07, 1.2293e-07, 6.7055e-08, ..., 0.0000e+00, + 2.6710e-06, 0.0000e+00], + ..., + [ 4.8429e-08, 4.3772e-08, -9.3551e-07, ..., 0.0000e+00, + 9.9652e-08, 0.0000e+00], + [-5.8040e-06, -6.0722e-07, 1.1595e-07, ..., 0.0000e+00, + -6.2250e-06, 0.0000e+00], + [ 2.3749e-08, 4.6287e-07, -9.1735e-08, ..., 0.0000e+00, + 1.5739e-07, 0.0000e+00]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0036, 0.0195, -0.0140, 0.0117, 0.0247, -0.0136, -0.0186, -0.0155, + -0.0261, -0.0327], device='cuda:0'), grad: tensor([ 3.4273e-07, 2.4110e-05, 5.0925e-06, -1.1250e-05, -6.3665e-06, + 1.4253e-05, 5.5321e-06, -2.2456e-05, -9.7752e-06, 5.8720e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 217.07, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5629 re_mapping 0.0059 re_causal 0.0189 /// teacc 98.93 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1583, -0.0876, -0.0598, ..., -0.0120, -0.0324, -0.0059], + [-0.0668, 0.0873, -0.0680, ..., 0.0219, -0.0318, -0.0704], + [ 0.0660, -0.1056, -0.0951, ..., -0.0528, 0.1039, -0.0283], + ..., + [-0.1206, -0.0745, 0.0898, ..., 0.0417, -0.1507, 0.0024], + [ 0.0554, 0.0029, -0.1995, ..., -0.0427, 0.1134, -0.0233], + [-0.1320, -0.0962, 0.0280, ..., -0.0055, -0.1175, -0.0489]], + device='cuda:0'), grad: tensor([[ 2.5006e-07, 4.8894e-07, 1.5832e-08, ..., 0.0000e+00, + 9.9186e-08, 0.0000e+00], + [-8.6520e-07, -1.0923e-05, 7.1712e-08, ..., 0.0000e+00, + -3.8631e-06, 0.0000e+00], + [ 1.2554e-06, 4.9993e-06, 3.0734e-08, ..., 0.0000e+00, + 2.2352e-06, 0.0000e+00], + ..., + [-1.0757e-07, 6.9290e-07, -2.1094e-07, ..., 0.0000e+00, + 2.6915e-07, 0.0000e+00], + [-2.8089e-06, 2.2203e-06, 1.1036e-07, ..., 0.0000e+00, + -2.1271e-06, 0.0000e+00], + [ 2.2491e-07, 1.6997e-07, -2.9523e-07, ..., 0.0000e+00, + 2.3516e-07, 0.0000e+00]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0039, 0.0193, -0.0142, 0.0119, 0.0248, -0.0139, -0.0175, -0.0158, + -0.0262, -0.0325], device='cuda:0'), grad: tensor([-3.2177e-07, -2.3514e-05, 1.0639e-05, 2.4755e-06, 8.7265e-07, + 2.0936e-06, 6.3702e-06, 1.0505e-06, 6.9337e-07, -3.4599e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 216.92, cls_loss 0.0021 cls_loss_mapping 0.0053 cls_loss_causal 0.5194 re_mapping 0.0056 re_causal 0.0177 /// teacc 99.03 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1584, -0.0869, -0.0600, ..., -0.0120, -0.0304, -0.0059], + [-0.0670, 0.0875, -0.0684, ..., 0.0219, -0.0321, -0.0704], + [ 0.0668, -0.1059, -0.0953, ..., -0.0528, 0.1042, -0.0283], + ..., + [-0.1201, -0.0745, 0.0909, ..., 0.0417, -0.1508, 0.0024], + [ 0.0557, 0.0028, -0.2012, ..., -0.0428, 0.1136, -0.0233], + [-0.1323, -0.0965, 0.0273, ..., -0.0056, -0.1175, -0.0489]], + device='cuda:0'), grad: tensor([[ 4.8429e-08, 1.3690e-07, 4.0978e-07, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 2.9802e-08, -2.7269e-06, 3.8184e-07, ..., 0.0000e+00, + -5.5972e-07, 0.0000e+00], + [ 1.8626e-08, 2.6654e-06, 6.5286e-07, ..., 0.0000e+00, + 5.2433e-07, 0.0000e+00], + ..., + [-6.5193e-09, 4.9081e-07, -2.4177e-06, ..., 0.0000e+00, + 5.6811e-08, 0.0000e+00], + [-1.4529e-07, 1.2107e-07, 1.3039e-07, ..., 0.0000e+00, + -3.8929e-07, 0.0000e+00], + [ 6.5193e-08, 7.5158e-07, 4.2934e-07, ..., 0.0000e+00, + 1.1921e-07, 0.0000e+00]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0027, 0.0190, -0.0144, 0.0125, 0.0248, -0.0146, -0.0184, -0.0146, + -0.0263, -0.0330], device='cuda:0'), grad: tensor([-3.4459e-06, -3.6731e-06, 7.3314e-06, 2.3637e-06, 4.8168e-06, + 3.0287e-06, 1.3383e-06, -6.0573e-06, 1.3029e-06, -7.0110e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 216.93, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.5260 re_mapping 0.0056 re_causal 0.0168 /// teacc 98.99 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1589, -0.0867, -0.0574, ..., -0.0120, -0.0304, -0.0059], + [-0.0670, 0.0880, -0.0685, ..., 0.0218, -0.0315, -0.0704], + [ 0.0659, -0.1065, -0.0966, ..., -0.0528, 0.1037, -0.0283], + ..., + [-0.1201, -0.0746, 0.0910, ..., 0.0416, -0.1512, 0.0024], + [ 0.0553, 0.0025, -0.2025, ..., -0.0434, 0.1144, -0.0234], + [-0.1326, -0.0971, 0.0273, ..., -0.0056, -0.1177, -0.0490]], + device='cuda:0'), grad: tensor([[ 2.4393e-05, 4.5262e-07, 2.4680e-06, ..., 9.3132e-10, + 3.7193e-05, 0.0000e+00], + [ 1.0924e-06, -8.1025e-08, 1.0831e-06, ..., 9.3132e-10, + 1.5106e-06, 0.0000e+00], + [-4.9829e-05, 3.5297e-07, 1.4156e-07, ..., 9.3132e-10, + -6.8545e-05, 0.0000e+00], + ..., + [ 1.0496e-06, -1.2992e-06, -4.6380e-06, ..., 9.3132e-10, + 2.2948e-06, 0.0000e+00], + [ 2.0623e-05, -1.4249e-07, 7.2643e-08, ..., 1.8626e-09, + 2.2173e-05, 0.0000e+00], + [ 4.0717e-06, 1.3663e-06, 2.1793e-07, ..., 0.0000e+00, + 3.1516e-06, 0.0000e+00]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0012, 0.0193, -0.0152, 0.0145, 0.0245, -0.0169, -0.0175, -0.0146, + -0.0266, -0.0329], device='cuda:0'), grad: tensor([ 1.1116e-04, 1.3307e-05, -1.4222e-04, -8.3148e-06, 1.3066e-06, + -4.3869e-05, 1.9819e-05, -3.4690e-05, 5.4479e-05, 2.8819e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 216.72, cls_loss 0.0023 cls_loss_mapping 0.0038 cls_loss_causal 0.5206 re_mapping 0.0055 re_causal 0.0162 /// teacc 98.98 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1593, -0.0876, -0.0576, ..., -0.0121, -0.0309, -0.0059], + [-0.0671, 0.0884, -0.0687, ..., 0.0218, -0.0316, -0.0705], + [ 0.0663, -0.1070, -0.0970, ..., -0.0528, 0.1041, -0.0283], + ..., + [-0.1226, -0.0749, 0.0913, ..., 0.0415, -0.1527, 0.0024], + [ 0.0550, 0.0020, -0.2035, ..., -0.0441, 0.1148, -0.0236], + [-0.1342, -0.1014, 0.0268, ..., -0.0058, -0.1183, -0.0490]], + device='cuda:0'), grad: tensor([[ 3.7253e-07, 3.9022e-07, 2.1700e-07, ..., 0.0000e+00, + 8.3819e-07, 2.7940e-09], + [ 1.3132e-07, -7.7188e-06, 3.9209e-07, ..., 0.0000e+00, + 2.3469e-07, 0.0000e+00], + [-1.4314e-06, 5.2806e-07, 1.1269e-07, ..., 0.0000e+00, + -1.5378e-05, 0.0000e+00], + ..., + [ 1.5367e-07, 5.0515e-06, -3.3341e-07, ..., 0.0000e+00, + 4.4331e-07, 1.8626e-09], + [ 3.1386e-07, 1.6950e-07, 1.7127e-06, ..., 0.0000e+00, + 8.2981e-07, 9.3132e-10], + [ 6.9849e-08, 1.3644e-06, -3.2131e-06, ..., 0.0000e+00, + 5.7742e-08, 9.3132e-10]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0014, 0.0194, -0.0152, 0.0145, 0.0266, -0.0166, -0.0172, -0.0147, + -0.0267, -0.0349], device='cuda:0'), grad: tensor([ 2.5854e-06, -1.7673e-05, -1.8418e-05, 2.7657e-05, 1.3253e-06, + -5.4426e-06, 2.3823e-06, 1.0900e-05, 9.0674e-06, -1.2442e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 174---------------------------------------------------- +epoch 174, time 217.84, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5235 re_mapping 0.0056 re_causal 0.0164 /// teacc 99.15 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1596, -0.0881, -0.0578, ..., -0.0121, -0.0313, -0.0059], + [-0.0672, 0.0889, -0.0687, ..., 0.0218, -0.0307, -0.0705], + [ 0.0666, -0.1073, -0.0972, ..., -0.0528, 0.1042, -0.0283], + ..., + [-0.1248, -0.0753, 0.0913, ..., 0.0415, -0.1541, 0.0024], + [ 0.0549, 0.0015, -0.2049, ..., -0.0444, 0.1152, -0.0237], + [-0.1346, -0.1011, 0.0273, ..., -0.0059, -0.1184, -0.0491]], + device='cuda:0'), grad: tensor([[ 1.3225e-07, 1.6484e-07, 6.5193e-08, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 2.1514e-07, 9.3691e-07, 8.4564e-07, ..., 0.0000e+00, + 7.5437e-08, 0.0000e+00], + [ 4.1910e-08, 5.7742e-08, 1.4901e-08, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + ..., + [-3.3621e-07, 1.6484e-07, -1.0598e-06, ..., 0.0000e+00, + 2.1420e-08, 0.0000e+00], + [-1.3849e-06, -8.8848e-07, 2.0768e-07, ..., 0.0000e+00, + -6.7241e-07, 0.0000e+00], + [ 2.2911e-07, -2.2464e-06, -1.4957e-06, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0016, 0.0199, -0.0154, 0.0141, 0.0264, -0.0163, -0.0171, -0.0150, + -0.0270, -0.0344], device='cuda:0'), grad: tensor([ 4.4517e-07, 1.5140e-05, 2.7195e-07, 4.0792e-06, 1.6153e-05, + 2.7344e-06, 3.9209e-07, -7.2923e-07, -1.9483e-06, -3.6538e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 216.92, cls_loss 0.0023 cls_loss_mapping 0.0049 cls_loss_causal 0.5178 re_mapping 0.0054 re_causal 0.0162 /// teacc 99.04 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1601, -0.0882, -0.0596, ..., -0.0115, -0.0307, -0.0056], + [-0.0674, 0.0892, -0.0689, ..., 0.0218, -0.0309, -0.0708], + [ 0.0646, -0.1075, -0.0981, ..., -0.0534, 0.1027, -0.0284], + ..., + [-0.1249, -0.0754, 0.0900, ..., 0.0419, -0.1547, 0.0023], + [ 0.0564, 0.0013, -0.2058, ..., -0.0451, 0.1172, -0.0239], + [-0.1354, -0.1013, 0.0288, ..., -0.0061, -0.1188, -0.0492]], + device='cuda:0'), grad: tensor([[ 5.0291e-07, 4.9826e-07, 4.4703e-08, ..., 0.0000e+00, + 1.4994e-07, 9.3132e-10], + [ 9.8627e-07, -8.2050e-07, 1.5181e-07, ..., 9.3132e-10, + -1.6391e-07, 0.0000e+00], + [-6.8918e-08, 1.0673e-06, 2.5146e-08, ..., 9.3132e-10, + -3.5390e-08, 0.0000e+00], + ..., + [ 2.3283e-07, 2.0675e-07, -1.1558e-06, ..., 1.8626e-09, + 2.3283e-07, 9.3132e-10], + [ 3.5558e-06, 5.3830e-07, 4.0047e-08, ..., 9.3132e-10, + 9.1270e-07, 4.6566e-09], + [ 3.9861e-07, 1.5274e-07, 6.1374e-07, ..., 0.0000e+00, + 4.8615e-07, 9.3132e-10]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0015, 0.0198, -0.0166, 0.0141, 0.0258, -0.0162, -0.0173, -0.0170, + -0.0267, -0.0323], device='cuda:0'), grad: tensor([-3.0566e-06, -5.3085e-08, 2.1122e-06, 3.2753e-05, 2.0340e-06, + -5.2840e-05, -4.2003e-07, -3.3546e-06, 1.6332e-05, 6.5640e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 216.81, cls_loss 0.0019 cls_loss_mapping 0.0032 cls_loss_causal 0.5179 re_mapping 0.0056 re_causal 0.0168 /// teacc 99.03 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1604, -0.0886, -0.0607, ..., -0.0104, -0.0308, -0.0074], + [-0.0677, 0.0889, -0.0691, ..., 0.0218, -0.0313, -0.0730], + [ 0.0653, -0.1076, -0.0982, ..., -0.0536, 0.1032, -0.0286], + ..., + [-0.1255, -0.0748, 0.0901, ..., 0.0418, -0.1557, 0.0025], + [ 0.0563, 0.0011, -0.2065, ..., -0.0459, 0.1171, -0.0247], + [-0.1356, -0.1015, 0.0290, ..., -0.0063, -0.1189, -0.0523]], + device='cuda:0'), grad: tensor([[ 3.3528e-07, 1.0561e-06, 1.6764e-08, ..., 1.8626e-09, + 4.5076e-07, 0.0000e+00], + [ 2.2165e-07, -1.6307e-06, 8.6613e-08, ..., 0.0000e+00, + 1.0524e-07, 0.0000e+00], + [ 3.3155e-06, 1.0245e-07, 1.0245e-08, ..., 0.0000e+00, + 4.0084e-06, 0.0000e+00], + ..., + [ 3.3528e-08, 3.1851e-07, -1.6857e-07, ..., 9.3132e-10, + 7.5437e-08, 0.0000e+00], + [-5.3085e-06, -4.4703e-07, 2.9244e-07, ..., 1.8626e-09, + -6.5714e-06, 0.0000e+00], + [ 4.6194e-07, 1.9558e-07, -4.2282e-07, ..., 9.3132e-10, + 5.7369e-07, 0.0000e+00]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0019, 0.0193, -0.0164, 0.0154, 0.0253, -0.0175, -0.0159, -0.0167, + -0.0270, -0.0322], device='cuda:0'), grad: tensor([ 2.5239e-07, -2.8051e-06, 8.7023e-06, -3.4552e-07, 7.9349e-07, + 3.5036e-06, 1.3271e-06, 4.9919e-07, -1.0684e-05, -1.2880e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 217.00, cls_loss 0.0018 cls_loss_mapping 0.0032 cls_loss_causal 0.4880 re_mapping 0.0058 re_causal 0.0167 /// teacc 99.05 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1610, -0.0897, -0.0608, ..., -0.0104, -0.0310, -0.0075], + [-0.0677, 0.0893, -0.0694, ..., 0.0219, -0.0316, -0.0731], + [ 0.0663, -0.1079, -0.0975, ..., -0.0537, 0.1039, -0.0286], + ..., + [-0.1268, -0.0749, 0.0902, ..., 0.0417, -0.1568, 0.0024], + [ 0.0567, 0.0012, -0.2074, ..., -0.0472, 0.1174, -0.0251], + [-0.1348, -0.1015, 0.0291, ..., -0.0064, -0.1190, -0.0528]], + device='cuda:0'), grad: tensor([[7.0110e-06, 5.0515e-06, 3.1888e-06, ..., 2.1700e-07, 1.3132e-06, + 0.0000e+00], + [1.3292e-05, 2.3544e-05, 6.2957e-07, ..., 1.6764e-08, 6.5044e-06, + 0.0000e+00], + [2.6766e-06, 3.9861e-06, 4.5728e-07, ..., 3.0734e-08, 9.3691e-07, + 0.0000e+00], + ..., + [6.5230e-06, 4.3958e-07, 4.2804e-06, ..., 3.2596e-07, 8.7544e-08, + 0.0000e+00], + [3.6992e-06, 4.9286e-06, 9.0152e-07, ..., 6.1467e-08, 1.1958e-06, + 0.0000e+00], + [1.7453e-06, 4.7497e-07, 1.3057e-06, ..., 8.1025e-08, 9.8720e-08, + 0.0000e+00]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0020, 0.0193, -0.0160, 0.0154, 0.0247, -0.0175, -0.0169, -0.0166, + -0.0272, -0.0316], device='cuda:0'), grad: tensor([ 1.8403e-05, 4.9084e-05, 8.9705e-06, 5.4799e-06, 5.6475e-06, + -3.0115e-05, -8.6546e-05, 1.2316e-05, 1.2152e-05, 4.5225e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.09, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5425 re_mapping 0.0054 re_causal 0.0172 /// teacc 99.01 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.1630, -0.0907, -0.0610, ..., -0.0104, -0.0300, -0.0075], + [-0.0680, 0.0896, -0.0697, ..., 0.0219, -0.0315, -0.0731], + [ 0.0672, -0.1083, -0.0978, ..., -0.0538, 0.1042, -0.0286], + ..., + [-0.1278, -0.0749, 0.0906, ..., 0.0417, -0.1577, 0.0024], + [ 0.0567, 0.0014, -0.2083, ..., -0.0476, 0.1175, -0.0252], + [-0.1345, -0.1019, 0.0290, ..., -0.0065, -0.1191, -0.0528]], + device='cuda:0'), grad: tensor([[ 7.2643e-08, 1.1455e-07, 3.5390e-08, ..., 0.0000e+00, + 1.7136e-07, 0.0000e+00], + [ 9.3132e-09, -2.6450e-07, 3.5670e-07, ..., 0.0000e+00, + 8.1025e-08, 0.0000e+00], + [ 1.6764e-08, 1.1548e-07, 3.3528e-08, ..., 0.0000e+00, + 2.5406e-06, 0.0000e+00], + ..., + [ 4.3772e-08, 3.1292e-07, 7.7635e-06, ..., 0.0000e+00, + 6.7987e-08, 0.0000e+00], + [ 5.6811e-08, 9.4064e-08, 6.4261e-08, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + [ 4.3772e-08, 6.3330e-07, -8.2180e-06, ..., 0.0000e+00, + 8.7544e-08, 0.0000e+00]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0020, 0.0194, -0.0163, 0.0153, 0.0249, -0.0175, -0.0167, -0.0165, + -0.0275, -0.0316], device='cuda:0'), grad: tensor([ 8.5868e-07, 2.8796e-06, 6.0685e-06, -5.2992e-07, -8.2105e-06, + -4.9248e-06, 9.3691e-07, 6.2466e-05, 7.9349e-07, -6.0439e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 216.91, cls_loss 0.0017 cls_loss_mapping 0.0032 cls_loss_causal 0.5005 re_mapping 0.0059 re_causal 0.0168 /// teacc 99.09 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1632, -0.0910, -0.0614, ..., -0.0101, -0.0300, -0.0075], + [-0.0684, 0.0893, -0.0710, ..., 0.0218, -0.0319, -0.0731], + [ 0.0677, -0.1084, -0.0984, ..., -0.0538, 0.1045, -0.0286], + ..., + [-0.1284, -0.0747, 0.0915, ..., 0.0416, -0.1585, 0.0024], + [ 0.0567, 0.0006, -0.2124, ..., -0.0498, 0.1177, -0.0252], + [-0.1348, -0.1022, 0.0287, ..., -0.0067, -0.1193, -0.0528]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, -2.2817e-07, 3.7253e-09, ..., 0.0000e+00, + 1.8999e-07, 0.0000e+00], + [ 1.3877e-06, 5.0198e-07, 5.4017e-08, ..., 9.3132e-10, + 3.0864e-06, 0.0000e+00], + [ 3.2596e-08, 2.9802e-08, 4.1910e-08, ..., 9.3132e-10, + 2.7101e-06, 0.0000e+00], + ..., + [ 2.3283e-08, 1.9185e-07, -3.4831e-07, ..., 9.3132e-10, + 3.2745e-06, 0.0000e+00], + [-2.2948e-06, -1.3467e-06, 1.6764e-08, ..., 0.0000e+00, + 5.2638e-06, 0.0000e+00], + [ 6.4261e-08, 1.1828e-07, 1.1083e-07, ..., 9.3132e-10, + 2.2538e-07, 0.0000e+00]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0019, 0.0186, -0.0163, 0.0151, 0.0253, -0.0175, -0.0164, -0.0156, + -0.0283, -0.0316], device='cuda:0'), grad: tensor([-1.8235e-06, 7.3425e-06, 6.2250e-06, -3.4839e-05, 3.0827e-07, + 1.5004e-06, 2.7306e-06, 6.8322e-06, 1.0483e-05, 1.2638e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 216.88, cls_loss 0.0018 cls_loss_mapping 0.0034 cls_loss_causal 0.4957 re_mapping 0.0053 re_causal 0.0157 /// teacc 99.11 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.1635, -0.0912, -0.0615, ..., -0.0101, -0.0302, -0.0075], + [-0.0686, 0.0899, -0.0706, ..., 0.0218, -0.0322, -0.0731], + [ 0.0687, -0.1084, -0.0999, ..., -0.0539, 0.1052, -0.0286], + ..., + [-0.1289, -0.0753, 0.0920, ..., 0.0416, -0.1590, 0.0024], + [ 0.0568, 0.0006, -0.2141, ..., -0.0504, 0.1176, -0.0252], + [-0.1354, -0.1023, 0.0290, ..., -0.0069, -0.1198, -0.0528]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, 2.2352e-08, 4.2841e-08, ..., 2.7940e-09, + 2.6356e-07, 0.0000e+00], + [ 1.6112e-07, 3.4459e-08, 1.6242e-06, ..., 1.3970e-08, + 3.8370e-07, 0.0000e+00], + [-6.1467e-08, 3.3528e-08, 2.0675e-07, ..., 8.3819e-09, + -1.7077e-05, 0.0000e+00], + ..., + [ 5.4948e-08, 1.0896e-07, -4.9621e-06, ..., 1.2107e-08, + 4.1164e-07, 0.0000e+00], + [-1.2089e-06, -3.4459e-07, 7.9162e-08, ..., 5.5879e-09, + -1.3616e-06, 0.0000e+00], + [ 1.5553e-07, 1.1362e-07, 2.7567e-06, ..., 2.6077e-08, + 1.4633e-05, 0.0000e+00]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0019, 0.0188, -0.0163, 0.0148, 0.0246, -0.0175, -0.0166, -0.0150, + -0.0288, -0.0313], device='cuda:0'), grad: tensor([-2.5518e-07, 1.0312e-05, -2.8402e-05, 1.9409e-06, -2.6003e-06, + 3.0212e-06, 1.4231e-06, -2.7016e-05, -2.0675e-06, 4.3601e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 217.00, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5025 re_mapping 0.0055 re_causal 0.0166 /// teacc 99.00 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1639, -0.0915, -0.0616, ..., -0.0101, -0.0304, -0.0075], + [-0.0691, 0.0901, -0.0709, ..., 0.0218, -0.0322, -0.0731], + [ 0.0691, -0.1089, -0.1002, ..., -0.0539, 0.1056, -0.0286], + ..., + [-0.1285, -0.0753, 0.0923, ..., 0.0416, -0.1594, 0.0024], + [ 0.0575, 0.0009, -0.2148, ..., -0.0506, 0.1180, -0.0252], + [-0.1362, -0.1024, 0.0289, ..., -0.0070, -0.1202, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.9744e-07, 7.3295e-07, 1.2200e-07, ..., 0.0000e+00, + 1.1362e-07, 0.0000e+00], + [ 1.4901e-08, -1.5110e-05, -8.5961e-07, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [-5.5879e-09, 1.3821e-06, 1.7323e-07, ..., 0.0000e+00, + -4.0606e-07, 0.0000e+00], + ..., + [ 2.0489e-08, 2.3060e-06, 4.5076e-07, ..., 0.0000e+00, + 3.3528e-08, 0.0000e+00], + [ 9.8720e-08, 4.2692e-06, 1.6764e-07, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + [ 2.4214e-08, 1.1306e-06, 6.3121e-05, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0018, 0.0187, -0.0166, 0.0149, 0.0247, -0.0176, -0.0171, -0.0146, + -0.0285, -0.0315], device='cuda:0'), grad: tensor([ 9.7044e-07, -3.3826e-05, 3.0063e-06, 2.4885e-06, -2.5797e-04, + 5.9605e-08, 2.5108e-06, 6.6310e-06, 1.0595e-05, 2.6584e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 217.10, cls_loss 0.0019 cls_loss_mapping 0.0032 cls_loss_causal 0.5289 re_mapping 0.0052 re_causal 0.0158 /// teacc 98.99 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1644, -0.0923, -0.0618, ..., -0.0102, -0.0306, -0.0075], + [-0.0693, 0.0904, -0.0711, ..., 0.0218, -0.0323, -0.0731], + [ 0.0696, -0.1091, -0.1005, ..., -0.0539, 0.1060, -0.0286], + ..., + [-0.1287, -0.0755, 0.0924, ..., 0.0416, -0.1601, 0.0024], + [ 0.0576, 0.0009, -0.2166, ..., -0.0507, 0.1182, -0.0252], + [-0.1365, -0.1025, 0.0287, ..., -0.0071, -0.1204, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.5181e-07, 1.4901e-07, 2.3283e-07, ..., 3.7253e-08, + 4.9360e-08, 0.0000e+00], + [ 1.1735e-07, -3.4541e-05, 1.5646e-07, ..., 1.8626e-08, + 4.0978e-08, 0.0000e+00], + [ 5.3644e-07, 3.2317e-07, 2.0415e-06, ..., 4.0326e-07, + -1.6671e-07, 0.0000e+00], + ..., + [-1.7062e-06, 3.0503e-05, -6.1393e-06, ..., -1.1269e-06, + 5.4017e-08, 0.0000e+00], + [ 2.7101e-07, 3.4831e-07, 1.8161e-07, ..., 1.1176e-08, + 1.8347e-07, 0.0000e+00], + [ 2.2911e-07, 7.6834e-07, -2.5891e-07, ..., 1.5832e-08, + 5.4017e-08, 0.0000e+00]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0021, 0.0188, -0.0166, 0.0145, 0.0249, -0.0173, -0.0174, -0.0147, + -0.0285, -0.0313], device='cuda:0'), grad: tensor([ 7.8697e-07, -7.5042e-05, 3.8818e-06, 6.1654e-06, 2.9728e-06, + 1.8328e-06, 6.0722e-07, 5.5611e-05, 2.0415e-06, 1.2675e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 217.11, cls_loss 0.0014 cls_loss_mapping 0.0038 cls_loss_causal 0.4751 re_mapping 0.0057 re_causal 0.0162 /// teacc 99.06 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1648, -0.0929, -0.0619, ..., -0.0102, -0.0309, -0.0075], + [-0.0695, 0.0906, -0.0711, ..., 0.0218, -0.0325, -0.0731], + [ 0.0697, -0.1093, -0.1007, ..., -0.0540, 0.1062, -0.0286], + ..., + [-0.1293, -0.0758, 0.0924, ..., 0.0416, -0.1609, 0.0024], + [ 0.0585, 0.0010, -0.2175, ..., -0.0513, 0.1190, -0.0252], + [-0.1379, -0.1030, 0.0284, ..., -0.0071, -0.1210, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.8720e-07, 1.6950e-07, 2.7008e-08, ..., 0.0000e+00, + 5.7742e-08, 0.0000e+00], + [ 3.2596e-08, -5.3085e-07, 2.3283e-08, ..., 0.0000e+00, + 9.9652e-08, 0.0000e+00], + [ 1.2107e-07, 1.1548e-07, 4.6566e-09, ..., 0.0000e+00, + 9.3039e-07, 0.0000e+00], + ..., + [ 3.9116e-08, 1.3970e-07, 7.2643e-08, ..., 0.0000e+00, + 1.9092e-07, 0.0000e+00], + [-4.7311e-07, 1.6019e-07, 1.3970e-07, ..., 0.0000e+00, + -4.9639e-07, 0.0000e+00], + [ 2.3842e-07, 5.0291e-08, -6.8359e-07, ..., 0.0000e+00, + 1.4901e-07, 0.0000e+00]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0021, 0.0188, -0.0166, 0.0145, 0.0247, -0.0173, -0.0175, -0.0148, + -0.0282, -0.0312], device='cuda:0'), grad: tensor([ 5.1688e-07, -3.6974e-07, 2.8405e-06, -3.1404e-06, 1.4221e-06, + 3.5390e-07, -8.9314e-07, 1.0347e-06, 2.5611e-07, -2.0135e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 216.99, cls_loss 0.0018 cls_loss_mapping 0.0028 cls_loss_causal 0.4907 re_mapping 0.0053 re_causal 0.0150 /// teacc 98.99 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1653, -0.0933, -0.0643, ..., -0.0103, -0.0313, -0.0075], + [-0.0696, 0.0910, -0.0714, ..., 0.0219, -0.0324, -0.0731], + [ 0.0702, -0.1096, -0.1008, ..., -0.0540, 0.1067, -0.0286], + ..., + [-0.1304, -0.0759, 0.0926, ..., 0.0416, -0.1626, 0.0024], + [ 0.0589, 0.0011, -0.2184, ..., -0.0516, 0.1194, -0.0253], + [-0.1388, -0.1030, 0.0287, ..., -0.0074, -0.1215, -0.0528]], + device='cuda:0'), grad: tensor([[ 2.6822e-07, 3.6787e-07, 1.2573e-07, ..., 0.0000e+00, + 1.2293e-07, 0.0000e+00], + [ 1.7136e-07, -3.6880e-07, 3.3248e-07, ..., 0.0000e+00, + 1.6298e-07, 0.0000e+00], + [ 9.0059e-07, 6.9477e-07, 2.4028e-07, ..., 0.0000e+00, + 1.0338e-06, 0.0000e+00], + ..., + [-6.3330e-08, 1.2573e-07, -9.4902e-07, ..., 0.0000e+00, + 3.6322e-08, 0.0000e+00], + [-1.1334e-06, 9.7789e-08, 2.6822e-07, ..., 0.0000e+00, + -3.2149e-06, 0.0000e+00], + [ 1.1371e-06, 5.6811e-07, -2.1104e-06, ..., 0.0000e+00, + 1.3057e-06, 0.0000e+00]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0034, 0.0191, -0.0167, 0.0146, 0.0247, -0.0173, -0.0182, -0.0149, + -0.0281, -0.0308], device='cuda:0'), grad: tensor([-6.5304e-06, 1.4352e-06, 4.9770e-06, 3.3714e-06, 1.8803e-06, + 5.6326e-06, -2.9523e-06, -3.9190e-06, -1.6559e-06, -2.2929e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 216.90, cls_loss 0.0016 cls_loss_mapping 0.0033 cls_loss_causal 0.5370 re_mapping 0.0053 re_causal 0.0163 /// teacc 98.98 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1662, -0.0939, -0.0650, ..., -0.0104, -0.0317, -0.0075], + [-0.0698, 0.0918, -0.0717, ..., 0.0219, -0.0329, -0.0731], + [ 0.0705, -0.1098, -0.1014, ..., -0.0541, 0.1068, -0.0286], + ..., + [-0.1313, -0.0765, 0.0930, ..., 0.0415, -0.1628, 0.0024], + [ 0.0593, 0.0012, -0.2198, ..., -0.0522, 0.1199, -0.0253], + [-0.1396, -0.1035, 0.0286, ..., -0.0075, -0.1219, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.0245e-07, 3.5390e-08, 2.8126e-07, ..., 0.0000e+00, + 1.9092e-07, 0.0000e+00], + [ 1.2293e-07, -6.0536e-08, 3.6657e-06, ..., 0.0000e+00, + 4.4890e-07, 0.0000e+00], + [-3.5670e-07, 1.3039e-08, 8.1304e-07, ..., 0.0000e+00, + -1.1669e-06, 0.0000e+00], + ..., + [ 6.0536e-08, 8.6613e-08, -1.0200e-05, ..., 0.0000e+00, + 2.4401e-07, 0.0000e+00], + [ 5.7463e-07, 2.3376e-07, 1.1353e-06, ..., 0.0000e+00, + 1.0338e-07, 0.0000e+00], + [ 8.9407e-08, -3.2671e-06, -1.5251e-05, ..., 0.0000e+00, + 2.9337e-07, 0.0000e+00]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0037, 0.0194, -0.0169, 0.0144, 0.0249, -0.0170, -0.0190, -0.0148, + -0.0280, -0.0310], device='cuda:0'), grad: tensor([-6.7148e-07, 1.3947e-05, 1.6028e-06, 9.7901e-06, 8.5890e-05, + -2.5723e-06, 1.7723e-06, -3.5554e-05, 7.1712e-06, -8.1480e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 216.89, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5221 re_mapping 0.0054 re_causal 0.0165 /// teacc 98.98 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1684, -0.0976, -0.0654, ..., -0.0104, -0.0321, -0.0075], + [-0.0697, 0.0927, -0.0720, ..., 0.0219, -0.0322, -0.0731], + [ 0.0706, -0.1101, -0.1017, ..., -0.0541, 0.1070, -0.0286], + ..., + [-0.1320, -0.0767, 0.0933, ..., 0.0415, -0.1639, 0.0024], + [ 0.0591, 0.0008, -0.2203, ..., -0.0524, 0.1186, -0.0253], + [-0.1402, -0.1013, 0.0314, ..., -0.0076, -0.1224, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.1455e-07, 2.6077e-08, -1.6391e-07, ..., 0.0000e+00, + 6.5491e-06, 0.0000e+00], + [ 4.3400e-07, -1.3039e-07, 2.9653e-06, ..., 0.0000e+00, + 8.1360e-06, 0.0000e+00], + [-2.0824e-06, 7.6368e-08, 5.7742e-08, ..., 0.0000e+00, + 7.3351e-06, 0.0000e+00], + ..., + [ 1.6708e-06, 6.3330e-08, -5.0515e-06, ..., 0.0000e+00, + 2.6748e-05, 0.0000e+00], + [-5.6159e-07, -1.9092e-07, 1.2293e-07, ..., 9.3132e-10, + 7.5847e-06, 0.0000e+00], + [ 9.9652e-08, 2.9802e-08, 1.7378e-06, ..., 0.0000e+00, + 2.4185e-05, 0.0000e+00]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0048, 0.0199, -0.0171, 0.0145, 0.0221, -0.0168, -0.0178, -0.0145, + -0.0288, -0.0286], device='cuda:0'), grad: tensor([ 1.9297e-05, 3.6716e-05, 2.4989e-05, -2.6655e-04, 1.4585e-06, + -5.9083e-06, 1.3024e-05, 6.7949e-05, 2.5570e-05, 8.3566e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 216.98, cls_loss 0.0018 cls_loss_mapping 0.0037 cls_loss_causal 0.5097 re_mapping 0.0056 re_causal 0.0161 /// teacc 98.92 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1686, -0.0978, -0.0658, ..., -0.0105, -0.0326, -0.0075], + [-0.0702, 0.0931, -0.0723, ..., 0.0219, -0.0333, -0.0731], + [ 0.0727, -0.1106, -0.1021, ..., -0.0541, 0.1083, -0.0286], + ..., + [-0.1356, -0.0773, 0.0934, ..., 0.0414, -0.1688, 0.0024], + [ 0.0600, 0.0012, -0.2210, ..., -0.0535, 0.1191, -0.0253], + [-0.1409, -0.1014, 0.0314, ..., -0.0078, -0.1228, -0.0528]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 1.5553e-07, 1.7695e-08, ..., 0.0000e+00, + 4.1910e-08, 0.0000e+00], + [-6.6124e-08, -3.5688e-06, 6.2957e-07, ..., 0.0000e+00, + -7.8883e-07, 0.0000e+00], + [-1.9092e-07, 7.1526e-07, 7.1712e-08, ..., 0.0000e+00, + -2.8126e-07, 0.0000e+00], + ..., + [ 4.6566e-08, 7.0781e-07, -1.1604e-06, ..., 0.0000e+00, + 1.0431e-07, 0.0000e+00], + [ 2.5798e-07, 1.5087e-06, 2.5146e-08, ..., 0.0000e+00, + 5.7556e-07, 0.0000e+00], + [ 3.4086e-07, 4.6566e-07, 2.5053e-07, ..., 0.0000e+00, + 1.3132e-07, 0.0000e+00]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0047, 0.0197, -0.0160, 0.0149, 0.0221, -0.0169, -0.0183, -0.0155, + -0.0287, -0.0286], device='cuda:0'), grad: tensor([-8.2329e-07, -3.5074e-06, 1.2135e-06, 2.8946e-06, 5.6811e-07, + -1.7835e-06, 3.6322e-07, -5.7667e-06, 4.0717e-06, 2.7567e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 216.67, cls_loss 0.0019 cls_loss_mapping 0.0043 cls_loss_causal 0.5438 re_mapping 0.0057 re_causal 0.0173 /// teacc 99.00 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1689, -0.0969, -0.0663, ..., -0.0105, -0.0329, -0.0075], + [-0.0680, 0.0944, -0.0730, ..., 0.0219, -0.0299, -0.0731], + [ 0.0718, -0.1119, -0.1036, ..., -0.0541, 0.1079, -0.0286], + ..., + [-0.1362, -0.0776, 0.0943, ..., 0.0414, -0.1692, 0.0024], + [ 0.0582, -0.0012, -0.2216, ..., -0.0537, 0.1171, -0.0253], + [-0.1423, -0.1019, 0.0313, ..., -0.0079, -0.1237, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.0896e-07, 1.3970e-07, 4.3772e-08, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 7.9162e-08, -5.6811e-07, 9.6112e-07, ..., 0.0000e+00, + 4.9360e-08, 0.0000e+00], + [ 1.6764e-08, 2.0955e-07, 5.3085e-08, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + ..., + [-4.3772e-08, 2.5705e-07, -3.1292e-06, ..., 9.3132e-10, + 7.5437e-08, 0.0000e+00], + [ 7.8231e-08, 3.0734e-08, 8.7544e-08, ..., 0.0000e+00, + 2.1420e-08, 0.0000e+00], + [ 1.0598e-06, 1.5087e-07, 1.2182e-06, ..., 0.0000e+00, + 3.6322e-08, 0.0000e+00]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0014, 0.0213, -0.0167, 0.0149, 0.0220, -0.0163, -0.0193, -0.0151, + -0.0308, -0.0296], device='cuda:0'), grad: tensor([ 4.1910e-07, 1.5236e-06, 6.3237e-07, 1.0943e-06, 2.6133e-06, + -2.7753e-06, -1.1437e-06, -6.9961e-06, 5.9605e-07, 4.0010e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 216.86, cls_loss 0.0021 cls_loss_mapping 0.0029 cls_loss_causal 0.4966 re_mapping 0.0056 re_causal 0.0155 /// teacc 99.00 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1698, -0.0973, -0.0668, ..., -0.0106, -0.0327, -0.0075], + [-0.0699, 0.0941, -0.0738, ..., 0.0219, -0.0323, -0.0731], + [ 0.0722, -0.1124, -0.1039, ..., -0.0542, 0.1081, -0.0286], + ..., + [-0.1365, -0.0775, 0.0949, ..., 0.0413, -0.1696, 0.0024], + [ 0.0593, -0.0003, -0.2230, ..., -0.0540, 0.1192, -0.0253], + [-0.1433, -0.1021, 0.0312, ..., -0.0082, -0.1244, -0.0528]], + device='cuda:0'), grad: tensor([[ 4.3772e-08, 6.4261e-08, 9.4529e-07, ..., 1.8626e-09, + 4.7404e-07, 0.0000e+00], + [ 3.7253e-08, 1.3690e-07, 4.0326e-07, ..., 7.4506e-09, + 1.7695e-06, 0.0000e+00], + [ 1.9558e-08, 4.8429e-08, 3.3341e-07, ..., 4.6566e-09, + 1.3895e-06, 0.0000e+00], + ..., + [ 9.3132e-09, -4.5262e-07, -2.1365e-06, ..., 4.6566e-09, + 4.0233e-07, 0.0000e+00], + [ 1.0151e-07, 2.3749e-07, 3.4459e-08, ..., 3.7253e-09, + -2.1327e-07, 0.0000e+00], + [ 2.7660e-07, 2.4959e-07, 1.8999e-07, ..., 1.0245e-08, + 3.5297e-07, 0.0000e+00]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0010, 0.0194, -0.0171, 0.0152, 0.0220, -0.0163, -0.0197, -0.0144, + -0.0295, -0.0299], device='cuda:0'), grad: tensor([ 3.7625e-06, 4.8950e-06, 3.7476e-06, -9.9912e-06, 2.5239e-07, + 1.3560e-06, -2.3749e-07, -6.5044e-06, 4.1537e-07, 2.2873e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 216.86, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4987 re_mapping 0.0054 re_causal 0.0160 /// teacc 99.00 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1706, -0.0976, -0.0669, ..., -0.0107, -0.0338, -0.0075], + [-0.0701, 0.0945, -0.0735, ..., 0.0218, -0.0325, -0.0731], + [ 0.0725, -0.1125, -0.1044, ..., -0.0571, 0.1073, -0.0286], + ..., + [-0.1367, -0.0780, 0.0950, ..., 0.0412, -0.1698, 0.0024], + [ 0.0594, -0.0004, -0.2243, ..., -0.0543, 0.1194, -0.0253], + [-0.1440, -0.1022, 0.0313, ..., -0.0084, -0.1248, -0.0528]], + device='cuda:0'), grad: tensor([[ 3.8445e-06, 2.7046e-06, 5.5879e-09, ..., 0.0000e+00, + 4.1258e-07, 0.0000e+00], + [ 1.7695e-07, -1.4342e-07, 4.3772e-08, ..., 0.0000e+00, + 3.0734e-08, 0.0000e+00], + [-3.3434e-07, 8.6613e-08, 5.5879e-09, ..., 0.0000e+00, + -1.4389e-06, 0.0000e+00], + ..., + [ 2.1514e-07, 2.7474e-07, 1.4435e-07, ..., 0.0000e+00, + 3.1665e-07, 0.0000e+00], + [ 1.0449e-06, 7.4226e-07, 1.3039e-08, ..., 0.0000e+00, + -2.4214e-08, 0.0000e+00], + [ 6.7987e-08, 3.7253e-07, 2.1700e-07, ..., 0.0000e+00, + 5.5879e-08, 0.0000e+00]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0011, 0.0194, -0.0179, 0.0156, 0.0220, -0.0164, -0.0193, -0.0145, + -0.0296, -0.0299], device='cuda:0'), grad: tensor([ 7.3798e-06, 5.9605e-08, -2.7940e-06, 1.9222e-06, -1.1306e-06, + 3.0324e-06, -1.3128e-05, 1.0831e-06, 1.8664e-06, 1.6829e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 217.00, cls_loss 0.0015 cls_loss_mapping 0.0037 cls_loss_causal 0.4904 re_mapping 0.0053 re_causal 0.0159 /// teacc 98.97 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.1715, -0.0980, -0.0671, ..., -0.0107, -0.0344, -0.0075], + [-0.0705, 0.0951, -0.0735, ..., 0.0218, -0.0314, -0.0731], + [ 0.0742, -0.1135, -0.1045, ..., -0.0571, 0.1071, -0.0286], + ..., + [-0.1380, -0.0784, 0.0951, ..., 0.0412, -0.1709, 0.0024], + [ 0.0594, -0.0005, -0.2273, ..., -0.0543, 0.1195, -0.0253], + [-0.1457, -0.1025, 0.0312, ..., -0.0084, -0.1251, -0.0528]], + device='cuda:0'), grad: tensor([[ 6.5193e-08, 3.8184e-08, 2.7940e-08, ..., 0.0000e+00, + 9.6392e-08, 0.0000e+00], + [ 2.1700e-07, -1.5832e-07, 4.8429e-08, ..., 0.0000e+00, + 2.5705e-07, 0.0000e+00], + [ 3.5856e-07, 2.0023e-08, -3.2596e-08, ..., 0.0000e+00, + 4.6287e-07, 0.0000e+00], + ..., + [ 2.1746e-07, 6.0070e-08, 8.9873e-07, ..., 0.0000e+00, + 3.3528e-07, 0.0000e+00], + [-2.3305e-05, -2.7288e-07, 7.3574e-08, ..., 0.0000e+00, + -3.3528e-05, 0.0000e+00], + [ 1.1045e-06, 9.2201e-08, -1.5497e-06, ..., 0.0000e+00, + 7.2876e-07, 0.0000e+00]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0012, 0.0205, -0.0192, 0.0157, 0.0222, -0.0160, -0.0209, -0.0145, + -0.0298, -0.0300], device='cuda:0'), grad: tensor([ 3.2131e-07, 3.4086e-07, 1.0692e-06, 5.9336e-05, 1.6186e-06, + 3.9004e-06, 1.0468e-06, 2.7828e-06, -6.8843e-05, -1.5441e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 216.74, cls_loss 0.0014 cls_loss_mapping 0.0040 cls_loss_causal 0.5230 re_mapping 0.0059 re_causal 0.0172 /// teacc 98.95 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1720, -0.0985, -0.0671, ..., -0.0107, -0.0351, -0.0075], + [-0.0707, 0.0954, -0.0741, ..., 0.0218, -0.0314, -0.0731], + [ 0.0745, -0.1138, -0.1055, ..., -0.0571, 0.1099, -0.0286], + ..., + [-0.1390, -0.0785, 0.0973, ..., 0.0412, -0.1715, 0.0024], + [ 0.0595, -0.0004, -0.2326, ..., -0.0544, 0.1197, -0.0253], + [-0.1449, -0.1026, 0.0306, ..., -0.0084, -0.1253, -0.0528]], + device='cuda:0'), grad: tensor([[ 5.1269e-07, 5.7416e-07, 8.3353e-08, ..., 0.0000e+00, + 6.9849e-09, 0.0000e+00], + [ 5.2620e-08, -3.5437e-07, 8.5682e-08, ..., 0.0000e+00, + 3.1665e-08, 0.0000e+00], + [ 5.1223e-08, 6.6124e-08, 2.0955e-08, ..., 0.0000e+00, + 3.2131e-08, 0.0000e+00], + ..., + [ 1.9558e-08, 3.1246e-07, -2.1979e-07, ..., 0.0000e+00, + 3.1199e-08, 0.0000e+00], + [-6.5239e-07, 6.0070e-08, 3.0268e-08, ..., 0.0000e+00, + -7.7114e-07, 0.0000e+00], + [ 8.2888e-08, 4.3074e-07, 7.6368e-08, ..., 0.0000e+00, + 1.0850e-07, 0.0000e+00]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0011, 0.0204, -0.0169, 0.0143, 0.0222, -0.0161, -0.0204, -0.0132, + -0.0304, -0.0304], device='cuda:0'), grad: tensor([-9.9093e-06, -3.0920e-07, 3.5530e-07, 5.8208e-07, 2.0459e-05, + 7.6927e-07, -2.3976e-05, 6.1002e-08, -1.1763e-06, 1.3128e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 216.89, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4799 re_mapping 0.0056 re_causal 0.0156 /// teacc 99.02 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1723, -0.0987, -0.0672, ..., -0.0107, -0.0353, -0.0075], + [-0.0709, 0.0958, -0.0733, ..., 0.0218, -0.0314, -0.0731], + [ 0.0749, -0.1139, -0.1059, ..., -0.0571, 0.1101, -0.0286], + ..., + [-0.1392, -0.0791, 0.0974, ..., 0.0412, -0.1718, 0.0024], + [ 0.0600, -0.0004, -0.2328, ..., -0.0544, 0.1201, -0.0253], + [-0.1457, -0.1029, 0.0301, ..., -0.0084, -0.1257, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.2573e-07, 5.1130e-07, 2.4680e-08, ..., 0.0000e+00, + 1.2340e-07, 0.0000e+00], + [ 7.0781e-08, -6.0536e-06, 4.3772e-08, ..., 0.0000e+00, + -1.4696e-06, 0.0000e+00], + [-1.5134e-07, 8.3819e-08, 3.7253e-09, ..., 0.0000e+00, + -2.8545e-07, 0.0000e+00], + ..., + [ 3.6322e-08, 1.1604e-06, -1.6298e-08, ..., 0.0000e+00, + 3.7998e-07, 0.0000e+00], + [ 9.3132e-09, 2.1309e-06, 4.5169e-08, ..., 0.0000e+00, + 5.8021e-07, 0.0000e+00], + [ 6.0536e-09, 1.9204e-06, -9.8068e-07, ..., 0.0000e+00, + 5.1875e-07, 0.0000e+00]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0010, 0.0207, -0.0169, 0.0142, 0.0225, -0.0162, -0.0203, -0.0132, + -0.0302, -0.0308], device='cuda:0'), grad: tensor([ 2.2519e-06, -2.8893e-05, -3.2596e-08, 2.3842e-07, 4.7944e-06, + 1.1921e-07, -4.9407e-07, 5.9605e-06, 1.0535e-05, 5.4911e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 216.94, cls_loss 0.0019 cls_loss_mapping 0.0029 cls_loss_causal 0.5285 re_mapping 0.0052 re_causal 0.0153 /// teacc 99.00 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1731, -0.0993, -0.0673, ..., -0.0107, -0.0358, -0.0075], + [-0.0712, 0.0944, -0.0732, ..., 0.0218, -0.0321, -0.0731], + [ 0.0750, -0.1144, -0.1079, ..., -0.0571, 0.1100, -0.0287], + ..., + [-0.1395, -0.0771, 0.0975, ..., 0.0412, -0.1714, 0.0009], + [ 0.0607, -0.0005, -0.2330, ..., -0.0545, 0.1208, -0.0253], + [-0.1468, -0.1032, 0.0297, ..., -0.0085, -0.1264, -0.0528]], + device='cuda:0'), grad: tensor([[ 8.1956e-08, 1.6857e-07, 3.0268e-08, ..., 0.0000e+00, + 9.4995e-08, 0.0000e+00], + [ 3.9116e-08, -6.7987e-07, 8.3819e-09, ..., 0.0000e+00, + 5.7742e-08, 0.0000e+00], + [-8.5961e-07, 4.7963e-08, 8.8476e-09, ..., 0.0000e+00, + -8.7311e-07, 0.0000e+00], + ..., + [ 1.8766e-07, 2.1700e-07, 2.3283e-09, ..., 0.0000e+00, + 2.2026e-07, 0.0000e+00], + [ 4.7032e-08, 5.0291e-08, 4.6100e-08, ..., 0.0000e+00, + 4.2841e-08, 0.0000e+00], + [ 2.1048e-07, 4.0047e-08, -3.3341e-07, ..., 0.0000e+00, + 6.5193e-08, 0.0000e+00]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0012, 0.0192, -0.0172, 0.0141, 0.0228, -0.0157, -0.0203, -0.0120, + -0.0301, -0.0311], device='cuda:0'), grad: tensor([ 6.8033e-07, -1.0580e-06, -2.0955e-06, 1.0449e-06, 7.4646e-07, + 1.6298e-07, 4.0652e-07, 9.9838e-07, 6.9709e-07, -1.5795e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 216.90, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5009 re_mapping 0.0052 re_causal 0.0153 /// teacc 98.96 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1751, -0.1009, -0.0676, ..., -0.0107, -0.0361, -0.0075], + [-0.0714, 0.0947, -0.0731, ..., 0.0218, -0.0321, -0.0731], + [ 0.0750, -0.1145, -0.1082, ..., -0.0571, 0.1101, -0.0287], + ..., + [-0.1405, -0.0774, 0.0974, ..., 0.0412, -0.1718, 0.0008], + [ 0.0611, -0.0005, -0.2333, ..., -0.0545, 0.1212, -0.0253], + [-0.1474, -0.1033, 0.0297, ..., -0.0085, -0.1269, -0.0528]], + device='cuda:0'), grad: tensor([[ 8.3819e-08, 4.0978e-08, 3.7719e-08, ..., 0.0000e+00, + 3.4785e-07, 0.0000e+00], + [ 3.3062e-08, -1.3709e-06, 6.2864e-08, ..., 0.0000e+00, + 2.7288e-07, 0.0000e+00], + [-7.4832e-07, 2.7940e-08, 1.3970e-08, ..., 0.0000e+00, + -2.3432e-06, 0.0000e+00], + ..., + [ 6.0070e-08, 6.9384e-07, 2.1420e-07, ..., 0.0000e+00, + 4.7544e-07, 0.0000e+00], + [-2.9150e-07, 7.3109e-08, 7.4506e-08, ..., 0.0000e+00, + 4.0140e-07, 0.0000e+00], + [ 2.4866e-07, 8.5458e-06, 1.7375e-05, ..., 0.0000e+00, + 7.2224e-07, 0.0000e+00]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0017, 0.0193, -0.0173, 0.0139, 0.0227, -0.0154, -0.0199, -0.0121, + -0.0300, -0.0309], device='cuda:0'), grad: tensor([-1.2480e-05, -9.5088e-07, -3.5819e-06, -5.1688e-08, -7.8499e-05, + 1.2703e-06, 5.3598e-07, 1.4506e-05, 2.3358e-06, 7.6890e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 216.86, cls_loss 0.0017 cls_loss_mapping 0.0032 cls_loss_causal 0.5170 re_mapping 0.0054 re_causal 0.0156 /// teacc 99.07 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1755, -0.1012, -0.0682, ..., -0.0107, -0.0366, -0.0075], + [-0.0719, 0.0954, -0.0730, ..., 0.0218, -0.0324, -0.0731], + [ 0.0765, -0.1147, -0.1083, ..., -0.0571, 0.1108, -0.0287], + ..., + [-0.1412, -0.0780, 0.0977, ..., 0.0412, -0.1728, 0.0008], + [ 0.0612, -0.0005, -0.2335, ..., -0.0545, 0.1212, -0.0254], + [-0.1481, -0.1037, 0.0295, ..., -0.0085, -0.1274, -0.0528]], + device='cuda:0'), grad: tensor([[ 2.9709e-07, 1.5367e-07, 1.2573e-08, ..., 0.0000e+00, + 3.1618e-07, 0.0000e+00], + [ 3.0361e-07, 1.1176e-07, 9.4436e-07, ..., 0.0000e+00, + 3.2270e-07, 0.0000e+00], + [ 9.3132e-07, 5.2899e-07, 2.5611e-08, ..., 0.0000e+00, + 9.7137e-07, 0.0000e+00], + ..., + [ 1.0617e-07, -1.2061e-07, -6.5006e-07, ..., 0.0000e+00, + 1.2619e-07, 0.0000e+00], + [-6.1579e-06, -2.8610e-06, 2.0117e-07, ..., 0.0000e+00, + -6.5640e-06, 0.0000e+00], + [ 3.3714e-07, 1.8859e-07, -6.9924e-06, ..., 0.0000e+00, + 3.5809e-07, 0.0000e+00]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0018, 0.0197, -0.0167, 0.0135, 0.0229, -0.0151, -0.0206, -0.0124, + -0.0301, -0.0312], device='cuda:0'), grad: tensor([-3.8333e-06, 7.2867e-06, 3.2261e-06, 6.9514e-06, 6.4746e-06, + 3.9250e-05, 3.2652e-06, -4.3772e-06, -1.7628e-05, -4.0621e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 216.81, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.5068 re_mapping 0.0054 re_causal 0.0161 /// teacc 99.01 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1759, -0.1018, -0.0682, ..., -0.0107, -0.0366, -0.0075], + [-0.0722, 0.0953, -0.0731, ..., 0.0218, -0.0326, -0.0731], + [ 0.0766, -0.1149, -0.1089, ..., -0.0571, 0.1109, -0.0287], + ..., + [-0.1411, -0.0779, 0.0976, ..., 0.0412, -0.1727, 0.0008], + [ 0.0613, -0.0005, -0.2336, ..., -0.0545, 0.1214, -0.0254], + [-0.1485, -0.1039, 0.0295, ..., -0.0085, -0.1277, -0.0528]], + device='cuda:0'), grad: tensor([[ 6.5658e-08, 6.5193e-08, 6.0536e-09, ..., 0.0000e+00, + 1.5786e-07, 0.0000e+00], + [ 6.5938e-07, -4.2142e-07, 9.3132e-09, ..., 0.0000e+00, + 2.6505e-06, 0.0000e+00], + [-1.0878e-06, 4.9360e-08, 2.3283e-08, ..., 0.0000e+00, + -3.7812e-06, 0.0000e+00], + ..., + [ 1.2526e-07, 1.2573e-07, 3.6787e-08, ..., 0.0000e+00, + 2.4401e-07, 0.0000e+00], + [ 3.6322e-08, 6.8452e-08, 3.2596e-09, ..., 0.0000e+00, + 8.6147e-08, 0.0000e+00], + [ 4.6566e-09, 2.2724e-07, -1.3690e-07, ..., 0.0000e+00, + 3.1199e-08, 0.0000e+00]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0013, 0.0193, -0.0169, 0.0134, 0.0230, -0.0150, -0.0212, -0.0121, + -0.0301, -0.0311], device='cuda:0'), grad: tensor([ 4.1816e-07, 5.3719e-06, -8.3596e-06, 9.8255e-07, -1.3318e-07, + 1.8952e-07, 6.2399e-08, 1.1614e-06, 3.6461e-07, -4.1444e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 217.17, cls_loss 0.0021 cls_loss_mapping 0.0043 cls_loss_causal 0.5211 re_mapping 0.0050 re_causal 0.0155 /// teacc 98.93 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1765, -0.1023, -0.0683, ..., -0.0107, -0.0372, -0.0075], + [-0.0727, 0.0940, -0.0729, ..., 0.0218, -0.0342, -0.0731], + [ 0.0774, -0.1151, -0.1094, ..., -0.0571, 0.1113, -0.0287], + ..., + [-0.1416, -0.0765, 0.0983, ..., 0.0412, -0.1704, 0.0008], + [ 0.0619, -0.0012, -0.2337, ..., -0.0545, 0.1216, -0.0254], + [-0.1503, -0.1048, 0.0291, ..., -0.0085, -0.1289, -0.0528]], + device='cuda:0'), grad: tensor([[ 1.5041e-07, 1.2806e-07, 5.2154e-08, ..., 0.0000e+00, + 1.3504e-07, 0.0000e+00], + [ 6.9384e-08, 1.8580e-07, 2.5947e-06, ..., 4.6566e-10, + 1.2061e-07, 0.0000e+00], + [ 3.0780e-07, 8.8941e-08, 5.8673e-08, ..., 4.6566e-10, + 6.0257e-07, 0.0000e+00], + ..., + [ 5.1688e-08, 1.3877e-07, -1.8597e-05, ..., 9.3132e-10, + 1.8999e-07, 0.0000e+00], + [-2.7418e-06, 2.7008e-08, 2.8405e-08, ..., 0.0000e+00, + -4.8205e-06, 0.0000e+00], + [ 2.7008e-08, 2.1383e-06, 1.4409e-05, ..., 4.6566e-10, + 1.1036e-07, 0.0000e+00]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0013, 0.0186, -0.0174, 0.0134, 0.0236, -0.0150, -0.0209, -0.0111, + -0.0303, -0.0320], device='cuda:0'), grad: tensor([ 8.2050e-07, 7.3537e-06, 1.4836e-06, -1.5562e-06, -4.6939e-06, + 2.9542e-06, 2.8387e-06, -4.5538e-05, -6.6198e-06, 4.2975e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 217.03, cls_loss 0.0018 cls_loss_mapping 0.0030 cls_loss_causal 0.4992 re_mapping 0.0055 re_causal 0.0156 /// teacc 99.00 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1770, -0.1029, -0.0696, ..., -0.0107, -0.0375, -0.0075], + [-0.0748, 0.0933, -0.0754, ..., 0.0217, -0.0351, -0.0731], + [ 0.0784, -0.1154, -0.1108, ..., -0.0571, 0.1116, -0.0287], + ..., + [-0.1427, -0.0765, 0.0989, ..., 0.0412, -0.1707, 0.0008], + [ 0.0629, -0.0008, -0.2332, ..., -0.0546, 0.1222, -0.0254], + [-0.1517, -0.1053, 0.0292, ..., -0.0085, -0.1294, -0.0528]], + device='cuda:0'), grad: tensor([[ 2.4214e-07, 1.8161e-07, 6.7055e-08, ..., 0.0000e+00, + 1.2573e-08, 0.0000e+00], + [ 1.0990e-07, -5.5274e-07, 2.4540e-07, ..., 0.0000e+00, + 3.4459e-08, 0.0000e+00], + [ 2.7940e-08, 5.3830e-07, 1.1874e-07, ..., 0.0000e+00, + -8.9360e-07, 0.0000e+00], + ..., + [ 1.0282e-06, 1.0142e-06, 3.8603e-07, ..., 0.0000e+00, + 1.0021e-06, 0.0000e+00], + [ 3.2363e-07, 1.9325e-07, 9.0804e-08, ..., 0.0000e+00, + 7.6368e-08, 0.0000e+00], + [ 3.6787e-08, 7.6648e-07, 5.7230e-07, ..., 0.0000e+00, + 2.2817e-08, 0.0000e+00]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0016, 0.0181, -0.0173, 0.0136, 0.0237, -0.0150, -0.0206, -0.0110, + -0.0297, -0.0319], device='cuda:0'), grad: tensor([ 5.2806e-07, -1.0803e-07, 3.2596e-09, -5.9092e-07, -7.5512e-06, + 5.6252e-07, -3.1497e-06, 5.7034e-06, 1.0775e-06, 3.5111e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 216.79, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5178 re_mapping 0.0057 re_causal 0.0162 /// teacc 99.05 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1777, -0.1033, -0.0698, ..., -0.0108, -0.0376, -0.0075], + [-0.0750, 0.0935, -0.0745, ..., 0.0215, -0.0351, -0.0731], + [ 0.0789, -0.1157, -0.1115, ..., -0.0573, 0.1117, -0.0287], + ..., + [-0.1454, -0.0768, 0.0979, ..., 0.0406, -0.1711, 0.0008], + [ 0.0625, -0.0008, -0.2336, ..., -0.0566, 0.1219, -0.0254], + [-0.1504, -0.1055, 0.0298, ..., -0.0091, -0.1288, -0.0528]], + device='cuda:0'), grad: tensor([[ 6.7055e-08, 6.1467e-08, 2.4214e-08, ..., 0.0000e+00, + 6.8918e-08, 0.0000e+00], + [ 6.7055e-08, 2.5798e-07, 3.9954e-07, ..., 0.0000e+00, + 1.5460e-07, 0.0000e+00], + [-2.5332e-07, 1.4901e-08, 9.2201e-08, ..., 0.0000e+00, + -3.0827e-07, 0.0000e+00], + ..., + [-2.7940e-09, -1.7881e-07, -5.0850e-07, ..., 0.0000e+00, + 3.8929e-07, 0.0000e+00], + [-1.2014e-07, 2.4214e-08, 2.7940e-08, ..., 0.0000e+00, + 8.5682e-08, 0.0000e+00], + [ 4.9360e-08, 3.2298e-06, 2.1104e-06, ..., 0.0000e+00, + 7.8231e-08, 0.0000e+00]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0014, 0.0181, -0.0174, 0.0138, 0.0228, -0.0146, -0.0209, -0.0113, + -0.0305, -0.0307], device='cuda:0'), grad: tensor([ 6.3051e-07, 2.4512e-06, 2.3283e-08, -1.8794e-06, -1.7107e-05, + -4.9286e-06, 6.1840e-07, -1.8394e-06, 4.7125e-07, 2.1517e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 216.60, cls_loss 0.0021 cls_loss_mapping 0.0045 cls_loss_causal 0.5244 re_mapping 0.0051 re_causal 0.0151 /// teacc 99.02 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1795, -0.1040, -0.0725, ..., -0.0108, -0.0393, -0.0075], + [-0.0753, 0.0937, -0.0747, ..., 0.0215, -0.0339, -0.0731], + [ 0.0769, -0.1161, -0.1121, ..., -0.0573, 0.1100, -0.0287], + ..., + [-0.1460, -0.0769, 0.0986, ..., 0.0406, -0.1717, 0.0008], + [ 0.0653, -0.0007, -0.2343, ..., -0.0568, 0.1247, -0.0254], + [-0.1507, -0.1061, 0.0297, ..., -0.0091, -0.1293, -0.0528]], + device='cuda:0'), grad: tensor([[ 6.3330e-08, 7.7300e-08, 2.1420e-08, ..., 0.0000e+00, + 1.9558e-08, 0.0000e+00], + [ 1.0710e-07, -5.8673e-08, 9.3132e-09, ..., 0.0000e+00, + 1.0710e-07, 0.0000e+00], + [ 4.6566e-08, 1.7695e-08, 4.6566e-09, ..., 0.0000e+00, + 3.9116e-08, 0.0000e+00], + ..., + [ 9.2201e-08, 7.1712e-08, 1.1176e-07, ..., 0.0000e+00, + 1.1176e-07, 0.0000e+00], + [-9.8161e-07, -1.3970e-08, 1.2107e-08, ..., 0.0000e+00, + -1.0338e-06, 0.0000e+00], + [ 2.2538e-07, 9.4157e-07, 4.0457e-06, ..., 0.0000e+00, + 2.2445e-07, 0.0000e+00]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0031, 0.0187, -0.0197, 0.0138, 0.0230, -0.0145, -0.0210, -0.0111, + -0.0291, -0.0309], device='cuda:0'), grad: tensor([-1.6131e-06, 2.3562e-07, 1.6671e-07, 4.5169e-07, -1.2942e-05, + 1.0850e-06, -1.1921e-07, 7.5437e-07, -2.7847e-06, 1.4730e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 217.03, cls_loss 0.0017 cls_loss_mapping 0.0029 cls_loss_causal 0.4707 re_mapping 0.0057 re_causal 0.0149 /// teacc 99.02 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1801, -0.1045, -0.0726, ..., -0.0108, -0.0396, -0.0075], + [-0.0755, 0.0938, -0.0752, ..., 0.0215, -0.0340, -0.0731], + [ 0.0774, -0.1161, -0.1123, ..., -0.0573, 0.1102, -0.0287], + ..., + [-0.1465, -0.0769, 0.1009, ..., 0.0405, -0.1720, 0.0008], + [ 0.0659, -0.0007, -0.2345, ..., -0.0568, 0.1254, -0.0254], + [-0.1517, -0.1065, 0.0286, ..., -0.0092, -0.1297, -0.0528]], + device='cuda:0'), grad: tensor([[ 9.6858e-07, 2.5146e-06, 2.4214e-08, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 5.4017e-08, 2.4214e-07, 3.0361e-07, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 4.9360e-08, 2.4028e-07, 5.4017e-08, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + ..., + [ 6.8918e-08, 1.2480e-06, -4.0792e-07, ..., 0.0000e+00, + 8.7544e-08, 0.0000e+00], + [-3.4552e-07, 8.1956e-08, 1.0245e-07, ..., 0.0000e+00, + -4.0513e-07, 0.0000e+00], + [ 1.7136e-07, 4.3493e-07, 5.0664e-07, ..., 0.0000e+00, + 1.6205e-07, 0.0000e+00]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0029, 0.0186, -0.0197, 0.0138, 0.0233, -0.0147, -0.0213, -0.0106, + -0.0289, -0.0318], device='cuda:0'), grad: tensor([ 5.1856e-06, 1.2917e-06, 6.3609e-07, 3.5297e-07, -3.1441e-06, + 5.8115e-07, -7.0967e-06, 6.9197e-07, -3.0920e-07, 1.7593e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 216.76, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.4988 re_mapping 0.0055 re_causal 0.0147 /// teacc 99.01 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1808, -0.1070, -0.0726, ..., -0.0108, -0.0402, -0.0075], + [-0.0757, 0.0938, -0.0757, ..., 0.0215, -0.0341, -0.0731], + [ 0.0785, -0.1163, -0.1125, ..., -0.0573, 0.1107, -0.0287], + ..., + [-0.1470, -0.0769, 0.1011, ..., 0.0405, -0.1724, 0.0008], + [ 0.0648, -0.0009, -0.2350, ..., -0.0568, 0.1243, -0.0254], + [-0.1520, -0.1067, 0.0286, ..., -0.0092, -0.1298, -0.0529]], + device='cuda:0'), grad: tensor([[ 3.7067e-07, 2.0489e-07, 6.5193e-09, ..., 0.0000e+00, + 4.3772e-07, 0.0000e+00], + [ 1.1269e-07, -5.0291e-08, 8.3819e-09, ..., 0.0000e+00, + 1.4994e-07, 0.0000e+00], + [-8.2552e-06, 1.2107e-08, 2.7940e-09, ..., 0.0000e+00, + -1.0952e-05, 0.0000e+00], + ..., + [ 1.1642e-07, 3.4459e-08, 1.6764e-08, ..., 0.0000e+00, + 1.6578e-07, 0.0000e+00], + [ 6.4261e-06, 9.3132e-10, 1.1176e-08, ..., 0.0000e+00, + 8.6054e-06, 0.0000e+00], + [ 7.1712e-08, 1.4342e-07, -1.1548e-07, ..., 0.0000e+00, + 9.4064e-08, 0.0000e+00]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0028, 0.0186, -0.0196, 0.0141, 0.0234, -0.0146, -0.0209, -0.0106, + -0.0302, -0.0317], device='cuda:0'), grad: tensor([ 1.3988e-06, 2.2445e-07, -1.5616e-05, 2.0191e-06, -1.3942e-06, + 2.1886e-07, 1.0710e-07, 3.9209e-07, 1.2182e-05, 4.9639e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 216.72, cls_loss 0.0021 cls_loss_mapping 0.0052 cls_loss_causal 0.4822 re_mapping 0.0053 re_causal 0.0150 /// teacc 99.11 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1815, -0.1075, -0.0728, ..., -0.0108, -0.0387, -0.0076], + [-0.0774, 0.0935, -0.0758, ..., 0.0215, -0.0346, -0.0732], + [ 0.0792, -0.1167, -0.1128, ..., -0.0573, 0.1109, -0.0287], + ..., + [-0.1460, -0.0766, 0.1019, ..., 0.0405, -0.1723, 0.0008], + [ 0.0650, -0.0015, -0.2354, ..., -0.0568, 0.1248, -0.0254], + [-0.1540, -0.1070, 0.0281, ..., -0.0092, -0.1303, -0.0529]], + device='cuda:0'), grad: tensor([[ 4.7497e-08, 5.7742e-08, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 4.5635e-08, -1.4184e-06, 4.6566e-09, ..., 0.0000e+00, + 5.3085e-08, 0.0000e+00], + [-3.3341e-07, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + -4.5355e-07, 0.0000e+00], + ..., + [ 1.6671e-07, 4.4890e-07, 2.0489e-08, ..., 0.0000e+00, + 1.9092e-07, 0.0000e+00], + [ 6.7987e-08, 1.3970e-08, 9.3132e-10, ..., 0.0000e+00, + 3.0734e-08, 0.0000e+00], + [ 1.9558e-07, 4.0326e-07, 5.5879e-09, ..., 0.0000e+00, + 7.3574e-08, 0.0000e+00]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0025, 0.0180, -0.0197, 0.0123, 0.0236, -0.0130, -0.0206, -0.0098, + -0.0303, -0.0323], device='cuda:0'), grad: tensor([-1.3411e-07, -2.4363e-06, -6.9384e-07, 5.3644e-07, 9.0245e-07, + -1.2256e-06, 1.4901e-07, 1.2554e-06, 2.8405e-07, 1.3411e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 216.62, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5127 re_mapping 0.0050 re_causal 0.0162 /// teacc 99.11 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1821, -0.1078, -0.0729, ..., -0.0108, -0.0390, -0.0076], + [-0.0780, 0.0937, -0.0751, ..., 0.0215, -0.0349, -0.0732], + [ 0.0822, -0.1168, -0.1140, ..., -0.0573, 0.1130, -0.0287], + ..., + [-0.1494, -0.0767, 0.1016, ..., 0.0405, -0.1754, 0.0008], + [ 0.0652, -0.0014, -0.2358, ..., -0.0568, 0.1249, -0.0254], + [-0.1544, -0.1073, 0.0278, ..., -0.0092, -0.1309, -0.0529]], + device='cuda:0'), grad: tensor([[1.6391e-07, 2.2072e-07, 4.0326e-07, ..., 0.0000e+00, 2.0862e-07, + 0.0000e+00], + [1.9465e-07, 2.7120e-06, 5.4687e-06, ..., 0.0000e+00, 4.3213e-07, + 0.0000e+00], + [3.3956e-06, 2.1979e-07, 4.1910e-07, ..., 0.0000e+00, 4.2766e-06, + 0.0000e+00], + ..., + [3.7532e-07, 1.5013e-05, 2.8625e-05, ..., 0.0000e+00, 5.9046e-07, + 0.0000e+00], + [1.4514e-05, 5.3179e-07, 1.1185e-06, ..., 0.0000e+00, 1.9118e-05, + 0.0000e+00], + [4.8708e-07, 4.0494e-06, 7.5214e-06, ..., 0.0000e+00, 6.3144e-07, + 0.0000e+00]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0025, 0.0181, -0.0181, 0.0124, 0.0238, -0.0129, -0.0205, -0.0105, + -0.0302, -0.0325], device='cuda:0'), grad: tensor([ 1.8105e-06, 1.9401e-05, 9.6858e-06, 5.0664e-06, -1.5295e-04, + -5.3823e-05, 3.8482e-06, 9.8646e-05, 4.1485e-05, 2.6733e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 216.81, cls_loss 0.0015 cls_loss_mapping 0.0028 cls_loss_causal 0.5087 re_mapping 0.0048 re_causal 0.0151 /// teacc 99.12 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1825, -0.1083, -0.0729, ..., -0.0108, -0.0393, -0.0076], + [-0.0783, 0.0937, -0.0751, ..., 0.0215, -0.0352, -0.0732], + [ 0.0827, -0.1167, -0.1150, ..., -0.0573, 0.1136, -0.0287], + ..., + [-0.1497, -0.0768, 0.1015, ..., 0.0405, -0.1759, 0.0008], + [ 0.0654, -0.0014, -0.2366, ..., -0.0568, 0.1251, -0.0255], + [-0.1550, -0.1074, 0.0281, ..., -0.0092, -0.1312, -0.0529]], + device='cuda:0'), grad: tensor([[ 1.7444e-06, 9.9186e-07, 1.8626e-09, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00], + [ 2.5053e-07, -7.5717e-07, 1.0245e-08, ..., 0.0000e+00, + 2.8964e-07, 0.0000e+00], + [ 4.5635e-07, 3.3714e-07, 4.6566e-09, ..., 0.0000e+00, + -2.3749e-07, 0.0000e+00], + ..., + [ 1.1828e-07, 4.1910e-07, -6.9849e-08, ..., 0.0000e+00, + 9.1270e-08, 0.0000e+00], + [-9.6634e-06, -1.2359e-06, 1.8626e-09, ..., 0.0000e+00, + -6.6496e-06, 0.0000e+00], + [ 3.4235e-06, 1.4063e-07, 3.6322e-08, ..., 0.0000e+00, + 2.6189e-06, 0.0000e+00]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0025, 0.0180, -0.0176, 0.0121, 0.0236, -0.0126, -0.0203, -0.0107, + -0.0303, -0.0322], device='cuda:0'), grad: tensor([ 3.3081e-06, -6.4448e-07, 8.3912e-07, 1.0535e-05, 5.5134e-07, + 8.9929e-06, -7.3835e-06, 7.9907e-07, -2.8536e-05, 1.1563e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 216.90, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5335 re_mapping 0.0049 re_causal 0.0154 /// teacc 99.01 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1835, -0.1091, -0.0730, ..., -0.0108, -0.0402, -0.0076], + [-0.0788, 0.0938, -0.0751, ..., 0.0215, -0.0358, -0.0732], + [ 0.0836, -0.1169, -0.1153, ..., -0.0573, 0.1144, -0.0287], + ..., + [-0.1504, -0.0769, 0.1012, ..., 0.0405, -0.1768, 0.0008], + [ 0.0659, -0.0012, -0.2368, ..., -0.0569, 0.1261, -0.0255], + [-0.1558, -0.1076, 0.0281, ..., -0.0093, -0.1318, -0.0529]], + device='cuda:0'), grad: tensor([[-1.7416e-07, 3.3528e-08, 2.7940e-09, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + [ 9.3132e-10, -5.7090e-07, 6.5193e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.4901e-08, 6.1467e-07, 2.7940e-09, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + ..., + [ 6.5193e-09, 5.8487e-07, 5.8673e-08, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00], + [ 0.0000e+00, 1.2107e-07, 9.3132e-09, ..., 0.0000e+00, + -5.9605e-08, 0.0000e+00], + [ 9.1270e-08, 2.1141e-07, -5.6811e-08, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0027, 0.0179, -0.0167, 0.0122, 0.0238, -0.0131, -0.0190, -0.0112, + -0.0296, -0.0322], device='cuda:0'), grad: tensor([-8.8569e-07, -1.6140e-06, 1.6149e-06, 1.5646e-07, -2.1476e-06, + 9.7789e-08, 3.1665e-07, 1.4575e-06, 3.2503e-07, 6.7428e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 216.87, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.5391 re_mapping 0.0049 re_causal 0.0162 /// teacc 98.95 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1839, -0.1097, -0.0731, ..., -0.0108, -0.0404, -0.0076], + [-0.0792, 0.0943, -0.0745, ..., 0.0215, -0.0358, -0.0732], + [ 0.0838, -0.1172, -0.1154, ..., -0.0573, 0.1146, -0.0287], + ..., + [-0.1505, -0.0774, 0.1009, ..., 0.0405, -0.1770, 0.0008], + [ 0.0674, -0.0010, -0.2370, ..., -0.0570, 0.1272, -0.0255], + [-0.1566, -0.1077, 0.0281, ..., -0.0094, -0.1325, -0.0529]], + device='cuda:0'), grad: tensor([[ 1.2200e-07, 5.9605e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00], + [ 7.4506e-08, -7.0967e-07, 5.5879e-09, ..., 0.0000e+00, + -5.2154e-08, 0.0000e+00], + [-5.9754e-06, 1.1455e-07, 3.7253e-09, ..., 0.0000e+00, + -1.5602e-05, 0.0000e+00], + ..., + [ 2.0582e-07, 1.3690e-07, -2.6077e-08, ..., 0.0000e+00, + 8.1025e-08, 0.0000e+00], + [ 6.2659e-06, 2.8871e-07, 3.7253e-09, ..., 0.0000e+00, + 1.5184e-05, 0.0000e+00], + [ 6.3051e-07, 4.0047e-08, 9.3132e-10, ..., 0.0000e+00, + 4.2841e-08, 0.0000e+00]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0028, 0.0182, -0.0167, 0.0125, 0.0239, -0.0136, -0.0189, -0.0115, + -0.0290, -0.0322], device='cuda:0'), grad: tensor([ 3.0641e-07, -7.8231e-07, -2.0012e-05, 3.1084e-05, 6.3796e-07, + -3.9488e-05, 3.2037e-06, 8.7544e-07, 2.2069e-05, 2.0936e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 216.61, cls_loss 0.0015 cls_loss_mapping 0.0027 cls_loss_causal 0.4923 re_mapping 0.0050 re_causal 0.0152 /// teacc 99.01 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1841, -0.1098, -0.0731, ..., -0.0101, -0.0405, -0.0069], + [-0.0793, 0.0944, -0.0748, ..., 0.0215, -0.0360, -0.0732], + [ 0.0863, -0.1173, -0.1130, ..., -0.0574, 0.1160, -0.0287], + ..., + [-0.1528, -0.0775, 0.1016, ..., 0.0405, -0.1784, 0.0008], + [ 0.0677, -0.0012, -0.2375, ..., -0.0571, 0.1275, -0.0255], + [-0.1570, -0.1073, 0.0292, ..., -0.0096, -0.1332, -0.0537]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, 2.7940e-08, 3.7253e-08, ..., 9.3132e-10, + 2.9802e-08, 0.0000e+00], + [ 5.3085e-08, -6.7614e-07, 2.1420e-08, ..., 9.3132e-10, + -4.7497e-08, 0.0000e+00], + [-3.5390e-08, 2.2911e-07, 2.1420e-08, ..., 3.7253e-09, + -2.1141e-07, 0.0000e+00], + ..., + [ 4.4703e-08, 1.4901e-07, -4.0978e-08, ..., 2.7940e-09, + 1.4622e-07, 0.0000e+00], + [-4.3306e-07, -1.0151e-07, 4.5635e-08, ..., 7.4506e-09, + -6.6776e-07, 0.0000e+00], + [ 6.2399e-08, 5.1223e-08, -9.4250e-07, ..., 4.6566e-09, + 9.6858e-08, 0.0000e+00]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0026, 0.0181, -0.0155, 0.0125, 0.0217, -0.0137, -0.0185, -0.0119, + -0.0290, -0.0303], device='cuda:0'), grad: tensor([ 1.0245e-07, -7.4692e-07, 2.2165e-07, 2.1160e-06, 5.7109e-06, + -3.7253e-09, 8.1211e-07, 7.0501e-07, -8.3633e-07, -8.1062e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 216.65, cls_loss 0.0015 cls_loss_mapping 0.0024 cls_loss_causal 0.4864 re_mapping 0.0047 re_causal 0.0141 /// teacc 98.99 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1848, -0.1103, -0.0735, ..., -0.0092, -0.0409, -0.0067], + [-0.0800, 0.0945, -0.0751, ..., 0.0215, -0.0363, -0.0732], + [ 0.0862, -0.1175, -0.1133, ..., -0.0575, 0.1161, -0.0287], + ..., + [-0.1529, -0.0775, 0.1018, ..., 0.0405, -0.1786, 0.0008], + [ 0.0698, -0.0003, -0.2378, ..., -0.0571, 0.1290, -0.0255], + [-0.1578, -0.1079, 0.0279, ..., -0.0099, -0.1343, -0.0539]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.0803e-07, 1.8626e-08, ..., 0.0000e+00, + -8.9407e-08, 0.0000e+00], + [ 1.3039e-08, -3.3174e-06, 1.2852e-07, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [-1.0151e-07, 2.4214e-07, 1.7695e-08, ..., 0.0000e+00, + -1.2759e-07, 0.0000e+00], + ..., + [ 1.2107e-07, 2.4363e-06, 3.1944e-07, ..., 0.0000e+00, + 1.7043e-07, 0.0000e+00], + [-1.0533e-06, 4.7218e-07, 4.7497e-08, ..., 0.0000e+00, + -9.9558e-07, 0.0000e+00], + [ 1.3877e-07, 8.4713e-06, 7.9498e-06, ..., 0.0000e+00, + 1.8068e-07, 0.0000e+00]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0026, 0.0181, -0.0155, 0.0126, 0.0222, -0.0139, -0.0190, -0.0119, + -0.0280, -0.0308], device='cuda:0'), grad: tensor([-7.9349e-07, -6.2212e-06, 2.0023e-07, 1.5059e-06, -3.7462e-05, + 1.4352e-06, 5.0757e-07, 6.3479e-06, -1.5590e-06, 3.6031e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 216.78, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.5141 re_mapping 0.0047 re_causal 0.0145 /// teacc 99.01 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1863, -0.1122, -0.0737, ..., -0.0092, -0.0418, -0.0065], + [-0.0802, 0.0953, -0.0731, ..., 0.0215, -0.0369, -0.0732], + [ 0.0864, -0.1180, -0.1136, ..., -0.0575, 0.1164, -0.0287], + ..., + [-0.1530, -0.0781, 0.1013, ..., 0.0405, -0.1788, 0.0008], + [ 0.0703, -0.0004, -0.2383, ..., -0.0571, 0.1295, -0.0255], + [-0.1582, -0.1086, 0.0279, ..., -0.0099, -0.1344, -0.0539]], + device='cuda:0'), grad: tensor([[ 1.2200e-07, 1.0896e-07, 5.3085e-08, ..., 0.0000e+00, + 9.5926e-08, 0.0000e+00], + [ 1.9707e-06, -2.8033e-07, 6.3516e-07, ..., 0.0000e+00, + 2.6636e-06, 0.0000e+00], + [ 3.3062e-06, 5.4017e-08, 9.8441e-07, ..., 0.0000e+00, + 4.4741e-06, 0.0000e+00], + ..., + [-5.9493e-06, 2.2259e-07, -1.6522e-06, ..., 0.0000e+00, + -8.0615e-06, 0.0000e+00], + [-8.8476e-08, 2.8126e-07, 1.6671e-07, ..., 0.0000e+00, + -4.8429e-07, 0.0000e+00], + [ 1.1642e-07, 4.5225e-06, 3.0212e-06, ..., 0.0000e+00, + 1.9092e-07, 0.0000e+00]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0032, 0.0183, -0.0152, 0.0126, 0.0224, -0.0138, -0.0191, -0.0124, + -0.0280, -0.0306], device='cuda:0'), grad: tensor([ 5.7183e-07, 1.0274e-05, 1.8001e-05, 3.7029e-06, -1.2092e-05, + 7.2457e-07, -8.9779e-07, -3.1710e-05, 2.6450e-07, 1.1124e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 216.95, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.5027 re_mapping 0.0045 re_causal 0.0150 /// teacc 99.05 lr 0.00010000 +Epoch 214, weight, value: tensor([[-1.8663e-01, -1.1273e-01, -7.4335e-02, ..., -9.2088e-03, + -4.2467e-02, -6.5127e-03], + [-8.0497e-02, 9.5519e-02, -7.2964e-02, ..., 2.1458e-02, + -3.6959e-02, -7.3216e-02], + [ 8.6466e-02, -1.1796e-01, -1.1389e-01, ..., -5.7499e-02, + 1.1651e-01, -2.8714e-02], + ..., + [-1.5306e-01, -7.8324e-02, 1.0146e-01, ..., 4.0573e-02, + -1.7895e-01, 8.2038e-04], + [ 7.0909e-02, 8.2639e-06, -2.3875e-01, ..., -5.7160e-02, + 1.2999e-01, -2.5547e-02], + [-1.5889e-01, -1.0897e-01, 2.7832e-02, ..., -9.9094e-03, + -1.3428e-01, -5.3909e-02]], device='cuda:0'), grad: tensor([[ 2.8871e-08, 5.2154e-08, 1.3039e-08, ..., 9.3132e-10, + 1.4901e-08, 0.0000e+00], + [ 4.3772e-08, -7.1712e-07, 2.8405e-07, ..., 9.3132e-10, + -4.3772e-08, 0.0000e+00], + [-3.7253e-08, 1.1735e-07, 3.0734e-08, ..., 1.8626e-09, + -5.4948e-08, 0.0000e+00], + ..., + [-6.4261e-08, 2.8126e-07, -3.8836e-07, ..., 6.5193e-09, + 4.2841e-08, 0.0000e+00], + [-3.3714e-07, 1.4342e-07, 1.6578e-07, ..., 9.3132e-10, + -3.8836e-07, 0.0000e+00], + [ 1.3877e-07, 5.0105e-07, 5.0291e-08, ..., 5.5879e-09, + 9.4064e-08, 0.0000e+00]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0038, 0.0184, -0.0152, 0.0126, 0.0224, -0.0137, -0.0195, -0.0125, + -0.0278, -0.0306], device='cuda:0'), grad: tensor([ 1.6298e-07, 2.2724e-07, 2.3004e-07, 4.6194e-07, -7.4226e-07, + -3.9767e-07, 3.0734e-07, -1.4994e-06, 7.1712e-08, 1.1502e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 216.53, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5276 re_mapping 0.0047 re_causal 0.0146 /// teacc 98.98 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1875, -0.1130, -0.0743, ..., -0.0092, -0.0427, -0.0065], + [-0.0815, 0.0957, -0.0730, ..., 0.0214, -0.0373, -0.0732], + [ 0.0874, -0.1183, -0.1143, ..., -0.0576, 0.1174, -0.0287], + ..., + [-0.1531, -0.0784, 0.1017, ..., 0.0406, -0.1792, 0.0008], + [ 0.0710, -0.0003, -0.2392, ..., -0.0577, 0.1302, -0.0255], + [-0.1599, -0.1095, 0.0276, ..., -0.0102, -0.1354, -0.0539]], + device='cuda:0'), grad: tensor([[ 7.8231e-08, 7.6368e-08, 1.8626e-08, ..., 0.0000e+00, + -4.3865e-07, 0.0000e+00], + [ 3.1292e-07, 2.4028e-07, 2.4959e-07, ..., 0.0000e+00, + 5.5134e-07, 0.0000e+00], + [ 9.5926e-08, 6.4261e-08, 8.5682e-08, ..., 0.0000e+00, + 4.4052e-07, 0.0000e+00], + ..., + [ 4.8429e-08, 8.3819e-08, -2.6543e-07, ..., 0.0000e+00, + 2.1141e-07, 0.0000e+00], + [-9.7416e-07, -7.4506e-08, 8.6613e-08, ..., 0.0000e+00, + -1.0375e-06, 0.0000e+00], + [ 2.0489e-07, 1.7406e-06, 7.8510e-07, ..., 0.0000e+00, + 3.5297e-07, 0.0000e+00]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0032, 0.0185, -0.0148, 0.0126, 0.0224, -0.0137, -0.0196, -0.0125, + -0.0281, -0.0309], device='cuda:0'), grad: tensor([-4.7460e-06, 2.0172e-06, 1.8729e-06, -4.5635e-07, -2.8871e-06, + 1.5292e-06, 4.8429e-08, -1.5646e-07, -1.0049e-06, 3.7998e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 216.91, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.4895 re_mapping 0.0047 re_causal 0.0143 /// teacc 98.96 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1882, -0.1134, -0.0761, ..., -0.0092, -0.0430, -0.0065], + [-0.0821, 0.0958, -0.0735, ..., 0.0214, -0.0369, -0.0732], + [ 0.0876, -0.1187, -0.1153, ..., -0.0576, 0.1176, -0.0287], + ..., + [-0.1531, -0.0785, 0.1016, ..., 0.0406, -0.1792, 0.0008], + [ 0.0704, -0.0013, -0.2398, ..., -0.0577, 0.1293, -0.0255], + [-0.1609, -0.1113, 0.0272, ..., -0.0102, -0.1365, -0.0539]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 3.3528e-08, 2.7940e-09, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 2.2352e-08, 5.0105e-07, 1.3039e-08, ..., 0.0000e+00, + 2.4214e-08, 0.0000e+00], + [-3.9022e-07, 2.6077e-08, 6.5193e-09, ..., 0.0000e+00, + -3.9209e-07, 0.0000e+00], + ..., + [ 3.5111e-07, 8.8476e-08, 9.8720e-08, ..., 0.0000e+00, + 3.6880e-07, 0.0000e+00], + [-7.8510e-07, 7.3574e-08, 1.3039e-08, ..., 0.0000e+00, + -7.7114e-07, 0.0000e+00], + [ 3.7253e-08, -1.1604e-06, -2.2631e-07, ..., 0.0000e+00, + 3.4459e-08, 0.0000e+00]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0037, 0.0186, -0.0150, 0.0128, 0.0232, -0.0136, -0.0193, -0.0126, + -0.0294, -0.0314], device='cuda:0'), grad: tensor([-2.9802e-08, 6.5416e-06, -9.8068e-07, 4.4927e-06, 1.6643e-06, + 4.0140e-07, 1.0990e-07, 1.9670e-06, -7.9349e-07, -1.3374e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 216.85, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.5070 re_mapping 0.0050 re_causal 0.0147 /// teacc 99.06 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1885, -0.1137, -0.0761, ..., -0.0091, -0.0428, -0.0065], + [-0.0819, 0.0960, -0.0737, ..., 0.0214, -0.0367, -0.0732], + [ 0.0874, -0.1195, -0.1164, ..., -0.0576, 0.1174, -0.0287], + ..., + [-0.1530, -0.0786, 0.1018, ..., 0.0405, -0.1790, 0.0007], + [ 0.0705, -0.0012, -0.2402, ..., -0.0578, 0.1296, -0.0256], + [-0.1620, -0.1119, 0.0271, ..., -0.0103, -0.1372, -0.0539]], + device='cuda:0'), grad: tensor([[ 6.0536e-08, 5.5879e-09, 2.0303e-07, ..., 0.0000e+00, + 7.0781e-08, 0.0000e+00], + [ 2.1420e-07, -3.1665e-08, 9.7230e-07, ..., 0.0000e+00, + 3.3807e-07, 0.0000e+00], + [-1.9222e-06, 4.6566e-09, 1.5832e-08, ..., 0.0000e+00, + -3.1218e-06, 0.0000e+00], + ..., + [ 4.5169e-07, 2.2352e-08, -1.6065e-06, ..., 0.0000e+00, + 6.9384e-07, 0.0000e+00], + [ 7.6927e-07, 4.1444e-07, 1.2759e-07, ..., 0.0000e+00, + 8.1584e-07, 0.0000e+00], + [ 1.6298e-07, 1.2480e-07, -2.2352e-08, ..., 0.0000e+00, + 8.2888e-08, 0.0000e+00]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0035, 0.0187, -0.0157, 0.0126, 0.0235, -0.0132, -0.0197, -0.0124, + -0.0295, -0.0317], device='cuda:0'), grad: tensor([ 3.1106e-07, 3.3826e-06, -5.6922e-06, -1.4439e-05, -5.4576e-07, + 1.4700e-05, 1.0021e-06, -2.9504e-06, 3.7700e-06, 4.3400e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 216.77, cls_loss 0.0016 cls_loss_mapping 0.0032 cls_loss_causal 0.4849 re_mapping 0.0054 re_causal 0.0146 /// teacc 99.05 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1882, -0.1138, -0.0759, ..., -0.0091, -0.0417, -0.0065], + [-0.0820, 0.0964, -0.0742, ..., 0.0214, -0.0367, -0.0732], + [ 0.0885, -0.1200, -0.1168, ..., -0.0576, 0.1179, -0.0287], + ..., + [-0.1534, -0.0788, 0.1014, ..., 0.0405, -0.1793, 0.0007], + [ 0.0696, -0.0023, -0.2408, ..., -0.0578, 0.1290, -0.0256], + [-0.1628, -0.1124, 0.0272, ..., -0.0103, -0.1377, -0.0539]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 3.1665e-08, -3.7253e-09, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 1.1176e-08, -3.8091e-07, 1.1176e-08, ..., 0.0000e+00, + 1.7695e-08, 0.0000e+00], + [-1.5367e-07, 3.3528e-08, 9.3132e-09, ..., 0.0000e+00, + -2.3562e-07, 0.0000e+00], + ..., + [ 2.2352e-08, 2.2724e-07, 4.0047e-08, ..., 0.0000e+00, + 4.6566e-08, 0.0000e+00], + [ 2.7940e-09, 8.8476e-08, 2.8871e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.0489e-08, 3.7532e-07, 9.1270e-08, ..., 0.0000e+00, + 3.4459e-08, 0.0000e+00]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0028, 0.0186, -0.0158, 0.0126, 0.0237, -0.0131, -0.0204, -0.0124, + -0.0311, -0.0314], device='cuda:0'), grad: tensor([ 2.1420e-08, -5.4948e-07, -3.7812e-07, -2.1607e-06, -1.7053e-06, + 2.1178e-06, 1.6857e-07, 6.6776e-07, 4.0699e-07, 1.4128e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 216.48, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.4788 re_mapping 0.0049 re_causal 0.0138 /// teacc 99.05 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1888, -0.1146, -0.0764, ..., -0.0091, -0.0420, -0.0065], + [-0.0824, 0.0974, -0.0724, ..., 0.0214, -0.0367, -0.0732], + [ 0.0893, -0.1202, -0.1170, ..., -0.0576, 0.1185, -0.0287], + ..., + [-0.1536, -0.0798, 0.1007, ..., 0.0405, -0.1795, 0.0007], + [ 0.0687, -0.0026, -0.2412, ..., -0.0579, 0.1286, -0.0256], + [-0.1631, -0.1128, 0.0272, ..., -0.0103, -0.1379, -0.0539]], + device='cuda:0'), grad: tensor([[ 1.0058e-07, 9.0338e-08, 1.9558e-08, ..., 0.0000e+00, + 7.9162e-08, 0.0000e+00], + [ 2.3600e-06, 1.5274e-07, 9.8720e-08, ..., 0.0000e+00, + 2.1886e-07, 0.0000e+00], + [ 1.2787e-06, 8.8010e-07, 1.8626e-08, ..., 0.0000e+00, + 1.1539e-06, 0.0000e+00], + ..., + [-2.0899e-06, 1.1735e-07, -2.6189e-06, ..., 0.0000e+00, + 1.2200e-07, 0.0000e+00], + [-3.2596e-06, -2.1961e-06, 8.3819e-09, ..., 0.0000e+00, + -3.0436e-06, 0.0000e+00], + [ 2.7474e-07, 8.2701e-07, 2.5965e-06, ..., 0.0000e+00, + 2.5891e-07, 0.0000e+00]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0028, 0.0192, -0.0154, 0.0128, 0.0236, -0.0132, -0.0205, -0.0130, + -0.0321, -0.0312], device='cuda:0'), grad: tensor([ 1.6391e-07, 7.3135e-05, 4.4703e-06, 1.2442e-06, -5.5581e-06, + 1.6289e-06, 9.1083e-07, -8.5890e-05, -7.9274e-06, 1.7911e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 217.11, cls_loss 0.0014 cls_loss_mapping 0.0033 cls_loss_causal 0.4738 re_mapping 0.0052 re_causal 0.0148 /// teacc 99.05 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1896, -0.1153, -0.0764, ..., -0.0091, -0.0424, -0.0067], + [-0.0832, 0.0975, -0.0726, ..., 0.0214, -0.0371, -0.0732], + [ 0.0901, -0.1206, -0.1197, ..., -0.0576, 0.1189, -0.0287], + ..., + [-0.1547, -0.0798, 0.1010, ..., 0.0405, -0.1804, 0.0007], + [ 0.0709, -0.0023, -0.2413, ..., -0.0579, 0.1300, -0.0256], + [-0.1651, -0.1130, 0.0271, ..., -0.0103, -0.1391, -0.0542]], + device='cuda:0'), grad: tensor([[ 1.7695e-08, 2.4214e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-09, -1.5832e-08, 9.3132e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 2.7940e-09, 9.3132e-09, 1.3970e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 2.0489e-08, -6.5193e-08, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [-2.2352e-08, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + -6.7055e-08, 0.0000e+00], + [ 1.7695e-08, 3.5390e-08, 4.9360e-08, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0026, 0.0191, -0.0156, 0.0129, 0.0237, -0.0134, -0.0204, -0.0132, + -0.0308, -0.0314], device='cuda:0'), grad: tensor([ 5.8673e-08, 4.4703e-08, 9.6858e-08, 1.0617e-07, 2.3469e-07, + 1.0617e-07, -5.3179e-07, -3.1106e-07, -9.1270e-08, 2.9989e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 217.15, cls_loss 0.0014 cls_loss_mapping 0.0028 cls_loss_causal 0.4915 re_mapping 0.0050 re_causal 0.0147 /// teacc 99.01 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1902, -0.1156, -0.0765, ..., -0.0091, -0.0429, -0.0067], + [-0.0838, 0.0974, -0.0729, ..., 0.0214, -0.0374, -0.0732], + [ 0.0902, -0.1209, -0.1199, ..., -0.0576, 0.1190, -0.0287], + ..., + [-0.1549, -0.0798, 0.1011, ..., 0.0405, -0.1806, 0.0005], + [ 0.0725, -0.0018, -0.2415, ..., -0.0579, 0.1312, -0.0256], + [-0.1666, -0.1131, 0.0270, ..., -0.0104, -0.1404, -0.0542]], + device='cuda:0'), grad: tensor([[-9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 4.6566e-09, -1.2107e-08, 2.7940e-09, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 9.3132e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 4.1910e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 2.2352e-08, 6.5193e-08, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00], + [-2.8871e-08, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + -2.3283e-08, 0.0000e+00], + [ 7.4506e-09, 1.1176e-08, -6.6124e-08, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0027, 0.0188, -0.0155, 0.0130, 0.0252, -0.0136, -0.0202, -0.0131, + -0.0299, -0.0327], device='cuda:0'), grad: tensor([-1.4333e-06, 1.0710e-07, 2.0675e-07, 1.3690e-07, 5.5879e-09, + -2.2724e-07, 1.7323e-07, 4.2003e-07, 1.1176e-08, 5.9884e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 216.87, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.4636 re_mapping 0.0051 re_causal 0.0144 /// teacc 99.01 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1912, -0.1161, -0.0765, ..., -0.0091, -0.0436, -0.0067], + [-0.0853, 0.0977, -0.0730, ..., 0.0214, -0.0379, -0.0732], + [ 0.0904, -0.1231, -0.1200, ..., -0.0576, 0.1192, -0.0287], + ..., + [-0.1552, -0.0799, 0.1015, ..., 0.0405, -0.1808, 0.0004], + [ 0.0736, -0.0005, -0.2416, ..., -0.0579, 0.1318, -0.0256], + [-0.1676, -0.1141, 0.0267, ..., -0.0104, -0.1414, -0.0542]], + device='cuda:0'), grad: tensor([[ 3.0221e-07, 3.5996e-07, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 2.5285e-07, 3.5297e-07, 1.8626e-09, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 9.7789e-08, 1.7416e-07, 0.0000e+00, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + ..., + [ 1.7695e-08, 3.2131e-08, 7.4506e-09, ..., 0.0000e+00, + 1.2573e-08, 0.0000e+00], + [ 4.4480e-06, 3.4142e-06, 2.3283e-09, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + [ 1.0198e-07, 8.0094e-08, -2.2817e-08, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0028, 0.0187, -0.0159, 0.0132, 0.0259, -0.0136, -0.0202, -0.0124, + -0.0289, -0.0342], device='cuda:0'), grad: tensor([-1.1735e-05, 1.0412e-06, 5.4995e-07, 5.6885e-06, 6.9514e-06, + 2.8446e-05, -4.2677e-05, 2.1607e-07, 1.0662e-05, 8.4471e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 217.01, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5106 re_mapping 0.0047 re_causal 0.0144 /// teacc 98.98 lr 0.00010000 +Epoch 223, weight, value: tensor([[-1.9172e-01, -1.1634e-01, -7.6547e-02, ..., -9.1411e-03, + -4.3811e-02, -6.7061e-03], + [-8.6416e-02, 9.8063e-02, -7.3186e-02, ..., 2.1414e-02, + -3.8688e-02, -7.3246e-02], + [ 9.2028e-02, -1.2350e-01, -1.2047e-01, ..., -5.7849e-02, + 1.2027e-01, -2.8723e-02], + ..., + [-1.5560e-01, -8.0108e-02, 1.0180e-01, ..., 4.1343e-02, + -1.8112e-01, 3.5939e-04], + [ 7.2655e-02, 4.0799e-05, -2.4182e-01, ..., -5.7875e-02, + 1.3152e-01, -2.5795e-02], + [-1.6836e-01, -1.1457e-01, 2.6524e-02, ..., -1.0380e-02, + -1.4203e-01, -5.4188e-02]], device='cuda:0'), grad: tensor([[ 1.1502e-07, 2.5751e-07, 6.0536e-09, ..., 0.0000e+00, + 2.5611e-08, 0.0000e+00], + [ 2.1420e-08, -2.9095e-06, 2.5146e-08, ..., 0.0000e+00, + 6.7055e-08, 0.0000e+00], + [-2.9476e-07, 8.7544e-08, 7.9162e-09, ..., 0.0000e+00, + -1.9325e-07, 0.0000e+00], + ..., + [ 9.7789e-09, 6.8778e-07, -3.4925e-08, ..., 0.0000e+00, + 7.3574e-08, 0.0000e+00], + [ 2.1094e-07, 6.4354e-07, 5.5879e-09, ..., 0.0000e+00, + 3.9628e-07, 0.0000e+00], + [ 2.9337e-08, 3.7253e-07, 7.8697e-08, ..., 0.0000e+00, + 6.5658e-08, 0.0000e+00]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0026, 0.0187, -0.0153, 0.0143, 0.0261, -0.0147, -0.0202, -0.0126, + -0.0290, -0.0344], device='cuda:0'), grad: tensor([ 9.8813e-07, -5.9381e-06, 1.0990e-07, -1.2908e-06, 2.2203e-06, + 5.9512e-07, -4.8336e-07, 1.8310e-06, 2.1383e-06, -1.7975e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 217.12, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4984 re_mapping 0.0046 re_causal 0.0142 /// teacc 99.05 lr 0.00010000 +Epoch 224, weight, value: tensor([[-1.9275e-01, -1.1662e-01, -7.6921e-02, ..., -9.1412e-03, + -4.3634e-02, -6.7061e-03], + [-8.6890e-02, 9.8653e-02, -7.2771e-02, ..., 2.1414e-02, + -3.8974e-02, -7.3249e-02], + [ 9.2500e-02, -1.2391e-01, -1.2088e-01, ..., -5.7849e-02, + 1.2063e-01, -2.8723e-02], + ..., + [-1.5591e-01, -8.0533e-02, 1.0128e-01, ..., 4.1343e-02, + -1.8133e-01, 3.5666e-04], + [ 7.1789e-02, -4.6508e-05, -2.4196e-01, ..., -5.7875e-02, + 1.3119e-01, -2.5822e-02], + [-1.6939e-01, -1.1496e-01, 2.6472e-02, ..., -1.0381e-02, + -1.4276e-01, -5.4191e-02]], device='cuda:0'), grad: tensor([[ 1.6158e-07, 7.4040e-08, 2.3749e-08, ..., 0.0000e+00, + 8.9407e-08, 0.0000e+00], + [ 9.8255e-08, -2.4587e-07, 2.0023e-08, ..., 0.0000e+00, + 1.0896e-07, 0.0000e+00], + [-1.4212e-06, 2.3283e-08, -2.4121e-07, ..., 0.0000e+00, + -8.2795e-07, 0.0000e+00], + ..., + [ 1.6205e-06, 9.7789e-08, 3.8650e-08, ..., 0.0000e+00, + 1.0449e-06, 0.0000e+00], + [-8.8708e-07, 5.5414e-08, 3.5390e-08, ..., 0.0000e+00, + -5.9465e-07, 0.0000e+00], + [ 3.8650e-08, 3.3528e-08, 1.8626e-08, ..., 0.0000e+00, + 3.5996e-07, 0.0000e+00]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0025, 0.0189, -0.0150, 0.0142, 0.0264, -0.0144, -0.0202, -0.0131, + -0.0297, -0.0345], device='cuda:0'), grad: tensor([ 2.0210e-07, 2.3562e-07, -2.3693e-06, -7.6413e-05, 4.6799e-07, + 3.6120e-05, -4.1304e-07, 4.1008e-05, -8.8476e-07, 2.1085e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 216.95, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.4873 re_mapping 0.0049 re_causal 0.0145 /// teacc 99.03 lr 0.00010000 +Epoch 225, weight, value: tensor([[-1.9369e-01, -1.1485e-01, -7.7003e-02, ..., -9.1412e-03, + -4.3972e-02, -6.7079e-03], + [-8.7865e-02, 9.8752e-02, -7.2720e-02, ..., 2.1413e-02, + -3.9117e-02, -7.3249e-02], + [ 9.2882e-02, -1.2418e-01, -1.2092e-01, ..., -5.7850e-02, + 1.2090e-01, -2.8723e-02], + ..., + [-1.5590e-01, -8.0714e-02, 1.0168e-01, ..., 4.1343e-02, + -1.8144e-01, 3.5499e-04], + [ 7.2085e-02, -2.0381e-04, -2.4222e-01, ..., -5.7876e-02, + 1.3136e-01, -2.5834e-02], + [-1.7000e-01, -1.1540e-01, 2.6451e-02, ..., -1.0384e-02, + -1.4327e-01, -5.4193e-02]], device='cuda:0'), grad: tensor([[ 4.3772e-08, 1.4529e-07, 5.1223e-09, ..., 0.0000e+00, + 2.6543e-08, 0.0000e+00], + [ 2.3702e-07, -1.0185e-05, 1.8161e-08, ..., 0.0000e+00, + 2.4308e-07, 0.0000e+00], + [ 7.0641e-07, 1.7555e-07, 9.3132e-10, ..., 0.0000e+00, + 6.7335e-07, 0.0000e+00], + ..., + [ 4.3660e-06, 2.5965e-06, 1.0431e-07, ..., 0.0000e+00, + 4.7050e-06, 0.0000e+00], + [-8.4490e-06, 1.5032e-06, 1.2340e-07, ..., 0.0000e+00, + -8.5756e-06, 0.0000e+00], + [ 3.6741e-07, 3.2037e-07, -1.5087e-05, ..., 0.0000e+00, + 4.0699e-07, 0.0000e+00]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0009, 0.0189, -0.0148, 0.0143, 0.0263, -0.0145, -0.0219, -0.0131, + -0.0298, -0.0346], device='cuda:0'), grad: tensor([ 1.3113e-06, -2.3454e-05, 2.5444e-06, 1.7900e-06, 9.6500e-05, + 6.9737e-06, 1.1493e-06, 2.1398e-05, -2.0117e-05, -8.8215e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 216.76, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.5009 re_mapping 0.0045 re_causal 0.0144 /// teacc 98.97 lr 0.00010000 +Epoch 226, weight, value: tensor([[-1.9546e-01, -1.1545e-01, -7.7046e-02, ..., -9.1376e-03, + -4.4513e-02, -6.7080e-03], + [-8.8140e-02, 9.9137e-02, -7.2634e-02, ..., 2.1413e-02, + -3.9199e-02, -7.3249e-02], + [ 9.3353e-02, -1.2451e-01, -1.2082e-01, ..., -5.7850e-02, + 1.2119e-01, -2.8723e-02], + ..., + [-1.5628e-01, -8.0968e-02, 1.0176e-01, ..., 4.1343e-02, + -1.8174e-01, 3.5491e-04], + [ 7.2746e-02, -2.0172e-04, -2.4237e-01, ..., -5.7876e-02, + 1.3180e-01, -2.5838e-02], + [-1.7053e-01, -1.1573e-01, 2.6105e-02, ..., -1.0386e-02, + -1.4377e-01, -5.4193e-02]], device='cuda:0'), grad: tensor([[ 8.7079e-08, 9.1735e-08, 1.1595e-07, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 3.5856e-08, -6.0536e-08, 3.1386e-07, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [ 2.1886e-08, 6.0536e-08, 1.8673e-07, ..., 0.0000e+00, + -2.1886e-08, 0.0000e+00], + ..., + [ 2.2352e-08, 1.1036e-07, -3.2932e-05, ..., 0.0000e+00, + 1.4435e-08, 0.0000e+00], + [ 1.2387e-07, 1.3877e-07, 1.0943e-07, ..., 0.0000e+00, + -7.1712e-08, 0.0000e+00], + [ 4.5635e-08, 1.2713e-07, 3.1918e-05, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0011, 0.0190, -0.0144, 0.0144, 0.0264, -0.0142, -0.0229, -0.0133, + -0.0295, -0.0348], device='cuda:0'), grad: tensor([ 1.5227e-07, 1.5907e-06, 1.0282e-06, 2.0377e-06, 7.9488e-07, + -2.2687e-06, 2.4633e-07, -1.7726e-04, 1.2796e-06, 1.7214e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 217.01, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4990 re_mapping 0.0046 re_causal 0.0141 /// teacc 98.85 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1968, -0.1170, -0.0771, ..., -0.0091, -0.0448, -0.0067], + [-0.0903, 0.0979, -0.0728, ..., 0.0214, -0.0393, -0.0732], + [ 0.0934, -0.1248, -0.1215, ..., -0.0579, 0.1212, -0.0287], + ..., + [-0.1564, -0.0802, 0.1019, ..., 0.0413, -0.1820, 0.0004], + [ 0.0754, 0.0010, -0.2427, ..., -0.0579, 0.1336, -0.0258], + [-0.1717, -0.1163, 0.0259, ..., -0.0104, -0.1440, -0.0542]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 3.8883e-07, 2.4447e-07, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 8.3819e-09, 1.5832e-08, 5.6904e-07, ..., 0.0000e+00, + 3.2596e-08, 0.0000e+00], + [ 8.8476e-09, 4.7823e-07, 3.5437e-07, ..., 0.0000e+00, + 5.7276e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 1.1884e-06, -9.0804e-07, ..., 0.0000e+00, + 4.3306e-08, 0.0000e+00], + [-9.1270e-08, 1.6578e-07, 1.3318e-07, ..., 0.0000e+00, + -1.1409e-07, 0.0000e+00], + [ 2.4680e-08, 4.7266e-05, 5.2482e-05, ..., 0.0000e+00, + 3.2131e-08, 0.0000e+00]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0019, 0.0181, -0.0147, 0.0142, 0.0266, -0.0148, -0.0204, -0.0127, + -0.0275, -0.0349], device='cuda:0'), grad: tensor([ 2.1104e-06, 2.7809e-06, 2.5127e-06, 2.0526e-06, -3.1734e-04, + -4.2329e-07, 1.7639e-06, -8.4564e-06, 1.1157e-06, 3.1352e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 216.75, cls_loss 0.0026 cls_loss_mapping 0.0043 cls_loss_causal 0.5290 re_mapping 0.0050 re_causal 0.0144 /// teacc 98.93 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.2001, -0.1187, -0.0773, ..., -0.0088, -0.0453, -0.0067], + [-0.0895, 0.0990, -0.0737, ..., 0.0214, -0.0393, -0.0732], + [ 0.0935, -0.1267, -0.1236, ..., -0.0579, 0.1217, -0.0287], + ..., + [-0.1569, -0.0804, 0.1011, ..., 0.0413, -0.1825, 0.0004], + [ 0.0758, -0.0009, -0.2467, ..., -0.0580, 0.1344, -0.0258], + [-0.1707, -0.1192, 0.0261, ..., -0.0105, -0.1451, -0.0542]], + device='cuda:0'), grad: tensor([[ 3.2037e-07, 7.4040e-08, 1.3970e-09, ..., 0.0000e+00, + 3.9814e-07, 0.0000e+00], + [ 8.9873e-08, -1.1921e-07, -1.9092e-08, ..., 0.0000e+00, + 1.2340e-07, 0.0000e+00], + [-7.7337e-06, 3.7719e-08, 7.9162e-09, ..., 0.0000e+00, + -1.0557e-05, 0.0000e+00], + ..., + [ 4.4070e-06, 5.3085e-08, 1.0245e-08, ..., 0.0000e+00, + 5.9791e-06, 0.0000e+00], + [ 2.5630e-06, 8.8476e-09, 9.7789e-09, ..., 0.0000e+00, + 3.4571e-06, 0.0000e+00], + [ 6.7055e-08, 1.0757e-07, 1.0245e-08, ..., 0.0000e+00, + 8.4750e-08, 0.0000e+00]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0004, 0.0183, -0.0153, 0.0131, 0.0282, -0.0149, -0.0219, -0.0142, + -0.0287, -0.0336], device='cuda:0'), grad: tensor([ 1.1381e-06, -1.1176e-08, -2.3201e-05, 9.4622e-07, 1.3597e-07, + 1.2247e-07, -4.7917e-07, 1.3329e-05, 7.7188e-06, 2.7008e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 216.67, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5022 re_mapping 0.0050 re_causal 0.0150 /// teacc 99.08 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.2007, -0.1193, -0.0774, ..., -0.0088, -0.0462, -0.0066], + [-0.0895, 0.0978, -0.0759, ..., 0.0214, -0.0396, -0.0733], + [ 0.0942, -0.1277, -0.1236, ..., -0.0579, 0.1223, -0.0287], + ..., + [-0.1575, -0.0791, 0.1022, ..., 0.0413, -0.1831, 0.0004], + [ 0.0766, -0.0011, -0.2472, ..., -0.0580, 0.1347, -0.0258], + [-0.1737, -0.1203, 0.0259, ..., -0.0105, -0.1459, -0.0542]], + device='cuda:0'), grad: tensor([[ 2.1979e-06, 1.5527e-05, 1.3830e-07, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 8.0559e-08, -6.5193e-09, 6.2864e-08, ..., 0.0000e+00, + 9.0804e-08, 0.0000e+00], + [-1.5227e-07, 3.2736e-07, 9.4529e-08, ..., 0.0000e+00, + -2.4308e-07, 0.0000e+00], + ..., + [ 7.9162e-09, 2.6543e-07, 3.7253e-09, ..., 0.0000e+00, + 3.5856e-08, 0.0000e+00], + [-6.9849e-09, 1.7416e-07, 1.5367e-08, ..., 0.0000e+00, + -1.2107e-08, 0.0000e+00], + [ 2.3283e-08, 2.6636e-06, 1.4259e-06, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0004, 0.0168, -0.0150, 0.0129, 0.0284, -0.0148, -0.0218, -0.0128, + -0.0283, -0.0340], device='cuda:0'), grad: tensor([ 3.4273e-05, 4.2375e-07, 5.0012e-07, 3.1851e-07, -1.2435e-05, + 1.1148e-06, -3.3796e-05, 5.9186e-07, 3.9116e-07, 8.6203e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 216.69, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.4910 re_mapping 0.0050 re_causal 0.0142 /// teacc 99.07 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.2012, -0.1199, -0.0775, ..., -0.0088, -0.0465, -0.0066], + [-0.0918, 0.0984, -0.0763, ..., 0.0214, -0.0388, -0.0733], + [ 0.0944, -0.1285, -0.1244, ..., -0.0579, 0.1226, -0.0287], + ..., + [-0.1577, -0.0791, 0.1027, ..., 0.0413, -0.1834, 0.0003], + [ 0.0774, -0.0012, -0.2476, ..., -0.0581, 0.1342, -0.0259], + [-0.1744, -0.1206, 0.0253, ..., -0.0105, -0.1469, -0.0542]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 1.3039e-08, 1.3039e-08, ..., 0.0000e+00, + -1.2573e-08, 0.0000e+00], + [ 6.9384e-08, -1.0571e-07, 1.5972e-07, ..., 0.0000e+00, + 6.9849e-09, 0.0000e+00], + [-3.2596e-09, 6.9849e-09, 9.5926e-08, ..., 0.0000e+00, + -2.5611e-08, 0.0000e+00], + ..., + [ 6.5193e-09, 9.2667e-08, -7.4413e-07, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 5.8999e-07, 1.1409e-07, 6.5193e-09, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00], + [ 4.7032e-08, 1.0245e-08, -3.7253e-09, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0004, 0.0172, -0.0151, 0.0124, 0.0285, -0.0138, -0.0222, -0.0127, + -0.0296, -0.0343], device='cuda:0'), grad: tensor([ 1.8161e-08, 1.0151e-06, 6.0489e-07, 2.8461e-06, 5.3784e-07, + -1.9930e-06, 1.9092e-07, -3.7886e-06, 1.4501e-06, -8.8802e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 216.68, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5045 re_mapping 0.0050 re_causal 0.0148 /// teacc 99.10 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.2015, -0.1202, -0.0775, ..., -0.0088, -0.0465, -0.0064], + [-0.0920, 0.0987, -0.0764, ..., 0.0214, -0.0385, -0.0733], + [ 0.0945, -0.1290, -0.1251, ..., -0.0580, 0.1226, -0.0288], + ..., + [-0.1578, -0.0792, 0.1031, ..., 0.0413, -0.1833, 0.0003], + [ 0.0778, -0.0017, -0.2479, ..., -0.0582, 0.1342, -0.0259], + [-0.1748, -0.1212, 0.0252, ..., -0.0106, -0.1476, -0.0542]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, 5.8673e-08, 5.1223e-09, ..., 0.0000e+00, + -1.9558e-08, 0.0000e+00], + [ 8.8476e-09, -1.2480e-07, 1.7695e-08, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 2.6543e-08, 6.2864e-08, 4.6566e-09, ..., 0.0000e+00, + 6.0070e-08, 0.0000e+00], + ..., + [ 4.1910e-09, 1.5972e-07, 4.9360e-08, ..., 0.0000e+00, + 1.6298e-08, 0.0000e+00], + [-1.3970e-08, 8.8476e-09, 2.8918e-07, ..., 0.0000e+00, + -4.3772e-08, 0.0000e+00], + [ 2.1886e-08, 6.6590e-08, -5.5134e-07, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0002, 0.0172, -0.0158, 0.0121, 0.0283, -0.0136, -0.0220, -0.0125, + -0.0299, -0.0345], device='cuda:0'), grad: tensor([-5.1688e-08, 5.5879e-08, 2.5379e-07, 2.6543e-07, 1.1642e-07, + 8.5402e-07, -1.8487e-07, 6.9477e-07, 2.8331e-06, -4.8392e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 230---------------------------------------------------- +epoch 230, time 217.75, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4993 re_mapping 0.0049 re_causal 0.0149 /// teacc 99.16 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.2018, -0.1202, -0.0776, ..., -0.0088, -0.0451, -0.0064], + [-0.0921, 0.0989, -0.0764, ..., 0.0214, -0.0386, -0.0733], + [ 0.0948, -0.1292, -0.1254, ..., -0.0580, 0.1229, -0.0288], + ..., + [-0.1579, -0.0793, 0.1032, ..., 0.0413, -0.1835, 0.0003], + [ 0.0778, -0.0017, -0.2483, ..., -0.0582, 0.1342, -0.0259], + [-0.1751, -0.1215, 0.0253, ..., -0.0106, -0.1477, -0.0542]], + device='cuda:0'), grad: tensor([[ 1.4435e-08, 1.5739e-07, 1.8626e-09, ..., 0.0000e+00, + 9.0338e-08, 0.0000e+00], + [ 1.6298e-08, -2.7101e-06, 6.9849e-09, ..., 0.0000e+00, + -1.2182e-06, 0.0000e+00], + [-6.8080e-07, 2.7800e-07, 1.8626e-09, ..., 0.0000e+00, + -9.5926e-07, 0.0000e+00], + ..., + [ 5.0804e-07, 1.0626e-06, 1.5832e-08, ..., 0.0000e+00, + 1.2089e-06, 0.0000e+00], + [ 2.0489e-08, 2.7195e-07, 1.1642e-08, ..., 0.0000e+00, + 1.2619e-07, 0.0000e+00], + [ 2.7940e-09, 5.5656e-06, 4.5029e-07, ..., 0.0000e+00, + 2.7474e-08, 0.0000e+00]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0038, 0.0173, -0.0159, 0.0124, 0.0282, -0.0142, -0.0183, -0.0125, + -0.0302, -0.0343], device='cuda:0'), grad: tensor([ 2.9011e-07, -5.6699e-06, -1.3374e-06, 4.8662e-07, -1.8954e-05, + 6.2864e-08, 1.4100e-06, 3.8128e-06, 7.4599e-07, 1.9163e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 216.69, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.4581 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +Epoch 233, weight, value: tensor([[-2.0201e-01, -1.2037e-01, -7.7646e-02, ..., -8.7844e-03, + -4.5422e-02, -9.0024e-03], + [-9.3186e-02, 9.9006e-02, -7.6731e-02, ..., 2.1394e-02, + -3.9342e-02, -7.4775e-02], + [ 9.5471e-02, -1.2995e-01, -1.2572e-01, ..., -5.7978e-02, + 1.2297e-01, -2.9387e-02], + ..., + [-1.5811e-01, -7.9166e-02, 1.0350e-01, ..., 4.1264e-02, + -1.8370e-01, 4.0327e-05], + [ 7.7648e-02, -1.8562e-03, -2.4880e-01, ..., -5.8184e-02, + 1.3367e-01, -2.6239e-02], + [-1.7543e-01, -1.2281e-01, 2.5206e-02, ..., -1.0605e-02, + -1.4874e-01, -5.7162e-02]], device='cuda:0'), grad: tensor([[ 1.2573e-08, 1.8161e-08, 4.0047e-08, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 7.4506e-09, -4.7963e-08, 1.1874e-07, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 1.3039e-08, 6.0536e-09, 1.0058e-07, ..., 0.0000e+00, + 2.5611e-08, 0.0000e+00], + ..., + [ 1.0710e-08, 7.1712e-08, -9.4017e-07, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 6.5193e-09, 6.5193e-09, 4.7032e-08, ..., 0.0000e+00, + -4.3772e-08, 0.0000e+00], + [ 4.5169e-08, 5.2387e-07, 6.5193e-07, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0039, 0.0171, -0.0164, 0.0125, 0.0284, -0.0138, -0.0182, -0.0123, + -0.0306, -0.0345], device='cuda:0'), grad: tensor([-3.6731e-06, 6.9570e-07, 6.3563e-07, 4.3474e-06, -4.1351e-07, + -3.8296e-06, 3.4226e-07, -4.7423e-06, 7.1479e-07, 5.9120e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 216.66, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.5138 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.06 lr 0.00010000 +Epoch 234, weight, value: tensor([[-2.0226e-01, -1.2351e-01, -8.0257e-02, ..., -8.7699e-03, + -4.5734e-02, -8.9966e-03], + [-9.3510e-02, 9.8953e-02, -7.7203e-02, ..., 2.1389e-02, + -3.9628e-02, -7.4790e-02], + [ 9.5965e-02, -1.3012e-01, -1.2584e-01, ..., -5.7994e-02, + 1.2340e-01, -2.9394e-02], + ..., + [-1.5835e-01, -7.9040e-02, 1.0491e-01, ..., 4.1248e-02, + -1.8400e-01, 1.7539e-05], + [ 7.8122e-02, -1.6745e-03, -2.4931e-01, ..., -5.8231e-02, + 1.3444e-01, -2.6393e-02], + [-1.7704e-01, -1.2352e-01, 2.4784e-02, ..., -1.0625e-02, + -1.5182e-01, -5.7191e-02]], device='cuda:0'), grad: tensor([[ 5.1223e-09, 1.3039e-08, 1.8626e-09, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 5.5879e-09, -1.1548e-07, 7.4506e-09, ..., 0.0000e+00, + 4.8894e-08, 0.0000e+00], + [-1.0710e-08, 7.4506e-09, 9.3132e-09, ..., 0.0000e+00, + -7.6834e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 6.2399e-08, -4.4703e-08, ..., 0.0000e+00, + 4.6100e-08, 0.0000e+00], + [-2.3749e-08, 1.7695e-08, 3.1665e-08, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + [ 1.3970e-09, 7.9162e-09, 7.4273e-07, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0046, 0.0169, -0.0159, 0.0143, 0.0282, -0.0154, -0.0175, -0.0121, + -0.0301, -0.0350], device='cuda:0'), grad: tensor([ 2.3283e-08, 5.8673e-08, 8.3819e-09, -1.6633e-06, -4.6566e-06, + 1.1846e-06, -1.7695e-08, 3.3621e-07, 3.4692e-07, 4.3623e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 216.56, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.5133 re_mapping 0.0045 re_causal 0.0135 /// teacc 99.06 lr 0.00010000 +Epoch 235, weight, value: tensor([[-2.0250e-01, -1.2347e-01, -8.0417e-02, ..., -6.3677e-03, + -4.5952e-02, -9.0058e-03], + [-9.3955e-02, 9.9016e-02, -7.7663e-02, ..., 2.1359e-02, + -3.9723e-02, -7.4799e-02], + [ 9.6321e-02, -1.3017e-01, -1.2754e-01, ..., -5.8051e-02, + 1.2341e-01, -2.9399e-02], + ..., + [-1.5846e-01, -7.9080e-02, 1.0596e-01, ..., 4.1202e-02, + -1.8396e-01, 5.0989e-06], + [ 7.8248e-02, -1.6715e-03, -2.5104e-01, ..., -5.8936e-02, + 1.3462e-01, -2.6856e-02], + [-1.7723e-01, -1.2394e-01, 2.4714e-02, ..., -1.1414e-02, + -1.5239e-01, -5.7205e-02]], device='cuda:0'), grad: tensor([[ 5.1223e-09, 5.5879e-09, 8.8476e-09, ..., 0.0000e+00, + 1.2573e-08, 0.0000e+00], + [ 2.0023e-08, 2.4214e-08, 4.1910e-08, ..., 0.0000e+00, + 3.6787e-08, 0.0000e+00], + [-1.9558e-08, 9.3132e-09, 5.6345e-08, ..., 0.0000e+00, + -6.1467e-08, 0.0000e+00], + ..., + [ 2.1886e-08, 1.6298e-08, -2.0955e-07, ..., 0.0000e+00, + 8.0094e-08, 0.0000e+00], + [ 1.3970e-09, -7.1246e-08, 6.9849e-08, ..., 0.0000e+00, + -1.4761e-07, 0.0000e+00], + [ 1.5832e-08, 3.2037e-07, 1.6904e-07, ..., 0.0000e+00, + 1.1921e-07, 0.0000e+00]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0045, 0.0169, -0.0160, 0.0145, 0.0279, -0.0155, -0.0177, -0.0119, + -0.0305, -0.0347], device='cuda:0'), grad: tensor([ 1.2806e-07, 4.7917e-07, 2.3283e-07, 3.4496e-06, -3.1758e-07, + -6.6012e-06, 1.8347e-07, -4.3539e-07, 7.6601e-07, 2.1160e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 216.61, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4837 re_mapping 0.0044 re_causal 0.0140 /// teacc 99.05 lr 0.00010000 +Epoch 236, weight, value: tensor([[-2.0263e-01, -1.2349e-01, -8.0480e-02, ..., -6.3021e-03, + -4.6115e-02, -8.9909e-03], + [-9.4003e-02, 9.9172e-02, -7.7449e-02, ..., 2.1357e-02, + -3.9836e-02, -7.4799e-02], + [ 9.6463e-02, -1.3035e-01, -1.2781e-01, ..., -5.8055e-02, + 1.2342e-01, -2.9422e-02], + ..., + [-1.5857e-01, -7.9195e-02, 1.0559e-01, ..., 4.1197e-02, + -1.8417e-01, 5.0957e-06], + [ 7.8250e-02, -1.7065e-03, -2.5154e-01, ..., -5.9139e-02, + 1.3476e-01, -2.6907e-02], + [-1.7751e-01, -1.2478e-01, 2.4670e-02, ..., -1.1502e-02, + -1.5271e-01, -5.7208e-02]], device='cuda:0'), grad: tensor([[ 3.3528e-08, 1.3039e-08, 1.3039e-08, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00], + [ 3.5390e-08, -3.2131e-08, 1.7369e-07, ..., 0.0000e+00, + 5.6345e-08, 0.0000e+00], + [-1.7649e-07, 2.8871e-08, 2.3656e-07, ..., 0.0000e+00, + -2.3283e-07, 0.0000e+00], + ..., + [ 2.2817e-08, 7.5903e-08, -7.6890e-06, ..., 0.0000e+00, + 5.4017e-08, 0.0000e+00], + [-7.5437e-08, -2.8871e-08, 1.5367e-08, ..., 0.0000e+00, + -4.8429e-08, 0.0000e+00], + [ 1.3039e-08, 2.5844e-07, 7.1749e-06, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0045, 0.0169, -0.0161, 0.0146, 0.0281, -0.0155, -0.0177, -0.0120, + -0.0307, -0.0347], device='cuda:0'), grad: tensor([-1.2042e-06, 6.5612e-07, 5.6531e-07, 6.7009e-07, -1.2303e-06, + 3.7579e-07, 1.6605e-06, -2.1115e-05, -1.6345e-07, 1.9804e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 216.40, cls_loss 0.0015 cls_loss_mapping 0.0028 cls_loss_causal 0.4984 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.07 lr 0.00010000 +Epoch 237, weight, value: tensor([[-2.0321e-01, -1.2347e-01, -8.2093e-02, ..., -6.3280e-03, + -4.5097e-02, -8.9909e-03], + [-9.4333e-02, 9.9276e-02, -7.7870e-02, ..., 2.1352e-02, + -4.0150e-02, -7.4799e-02], + [ 9.6589e-02, -1.3078e-01, -1.2867e-01, ..., -5.8367e-02, + 1.2354e-01, -2.9422e-02], + ..., + [-1.5877e-01, -7.9261e-02, 1.0308e-01, ..., 4.1712e-02, + -1.8478e-01, 5.0957e-06], + [ 7.9485e-02, -1.6296e-03, -2.5178e-01, ..., -5.9288e-02, + 1.3614e-01, -2.6907e-02], + [-1.7797e-01, -1.2521e-01, 2.7008e-02, ..., -1.1514e-02, + -1.5349e-01, -5.7208e-02]], device='cuda:0'), grad: tensor([[ 2.7940e-09, -2.6673e-06, -3.9227e-06, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 1.0710e-08, 3.6787e-07, 5.6345e-08, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00], + [-3.5856e-08, 2.6077e-07, 2.1094e-07, ..., 0.0000e+00, + -5.4017e-08, 0.0000e+00], + ..., + [ 4.6566e-09, -1.8515e-06, -3.7719e-08, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + [ 5.4017e-08, 9.9279e-07, 6.8452e-08, ..., 0.0000e+00, + 4.3306e-08, 0.0000e+00], + [ 2.7940e-09, 1.1651e-06, 8.1211e-07, ..., 0.0000e+00, + 1.7695e-08, 0.0000e+00]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0044, 0.0168, -0.0162, 0.0152, 0.0283, -0.0163, -0.0181, -0.0129, + -0.0300, -0.0328], device='cuda:0'), grad: tensor([-3.1590e-05, 3.9279e-05, 7.8157e-06, 1.0552e-06, -2.7753e-06, + 1.5832e-07, 3.0503e-05, -9.1851e-05, 3.8594e-05, 8.8736e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 216.40, cls_loss 0.0013 cls_loss_mapping 0.0024 cls_loss_causal 0.4777 re_mapping 0.0046 re_causal 0.0134 /// teacc 98.96 lr 0.00010000 +Epoch 238, weight, value: tensor([[-2.0351e-01, -1.2342e-01, -8.1754e-02, ..., -3.4160e-03, + -4.5143e-02, -8.7210e-03], + [-9.4862e-02, 9.9336e-02, -7.8725e-02, ..., 2.1342e-02, + -4.0585e-02, -7.4801e-02], + [ 9.6773e-02, -1.3153e-01, -1.3067e-01, ..., -5.8373e-02, + 1.2364e-01, -2.9429e-02], + ..., + [-1.5890e-01, -7.9269e-02, 1.0322e-01, ..., 4.1707e-02, + -1.8524e-01, 3.0746e-06], + [ 7.9480e-02, -1.5260e-03, -2.5286e-01, ..., -5.9337e-02, + 1.3664e-01, -2.6986e-02], + [-1.7809e-01, -1.2534e-01, 2.6637e-02, ..., -1.1600e-02, + -1.5409e-01, -5.7228e-02]], device='cuda:0'), grad: tensor([[ 7.7300e-08, 2.3283e-08, 5.5879e-09, ..., 0.0000e+00, + 7.2643e-08, 0.0000e+00], + [ 2.3749e-08, -7.6368e-08, 1.3039e-08, ..., 0.0000e+00, + 4.6100e-08, 0.0000e+00], + [ 1.1409e-07, 8.3819e-09, 1.0245e-08, ..., 0.0000e+00, + -6.1002e-08, 0.0000e+00], + ..., + [ 2.8498e-07, 4.7497e-08, 1.0524e-07, ..., 0.0000e+00, + 3.3900e-07, 0.0000e+00], + [-9.6671e-07, 1.0710e-08, 3.0734e-08, ..., 0.0000e+00, + -1.1688e-06, 0.0000e+00], + [ 2.7707e-07, 5.7742e-08, -3.8557e-06, ..., 0.0000e+00, + 3.1386e-07, 0.0000e+00]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0043, 0.0167, -0.0165, 0.0154, 0.0288, -0.0162, -0.0183, -0.0129, + -0.0301, -0.0329], device='cuda:0'), grad: tensor([-1.3988e-06, 6.8918e-08, 2.1094e-07, 9.4390e-07, -2.9290e-07, + 2.2367e-05, -9.5218e-06, 1.6084e-06, -2.6692e-06, -1.1332e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 216.31, cls_loss 0.0012 cls_loss_mapping 0.0022 cls_loss_causal 0.4996 re_mapping 0.0046 re_causal 0.0138 /// teacc 98.90 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.2039, -0.1234, -0.0817, ..., -0.0034, -0.0451, -0.0118], + [-0.0950, 0.1000, -0.0761, ..., 0.0213, -0.0408, -0.0764], + [ 0.0974, -0.1319, -0.1307, ..., -0.0584, 0.1242, -0.0299], + ..., + [-0.1590, -0.0799, 0.1018, ..., 0.0417, -0.1855, -0.0003], + [ 0.0792, -0.0015, -0.2535, ..., -0.0594, 0.1365, -0.0274], + [-0.1786, -0.1256, 0.0271, ..., -0.0116, -0.1544, -0.0597]], + device='cuda:0'), grad: tensor([[ 4.7497e-08, 8.6147e-08, 5.1223e-09, ..., 0.0000e+00, + 2.4214e-08, 0.0000e+00], + [ 5.6764e-07, 1.3206e-06, 1.7462e-07, ..., 0.0000e+00, + 3.3528e-08, 0.0000e+00], + [ 4.8056e-07, 2.0117e-07, 3.9628e-07, ..., 0.0000e+00, + 1.8114e-07, 0.0000e+00], + ..., + [-4.8196e-07, 8.3353e-08, -6.5379e-07, ..., 0.0000e+00, + 8.8010e-08, 0.0000e+00], + [-4.6100e-07, -5.2154e-08, 9.7789e-09, ..., 0.0000e+00, + -1.0887e-06, 4.6566e-10], + [ 1.1595e-07, 2.1420e-07, -1.9558e-08, ..., 0.0000e+00, + 1.1269e-07, 0.0000e+00]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0042, 0.0168, -0.0161, 0.0154, 0.0286, -0.0163, -0.0183, -0.0132, + -0.0304, -0.0324], device='cuda:0'), grad: tensor([-6.2492e-07, 2.5481e-06, 2.3413e-06, 9.2313e-06, 1.6671e-07, + -8.2403e-06, -1.6000e-06, -2.5108e-06, -1.7621e-06, 4.8848e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 216.48, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5012 re_mapping 0.0048 re_causal 0.0138 /// teacc 98.99 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.2042, -0.1235, -0.0818, ..., -0.0034, -0.0455, -0.0118], + [-0.0950, 0.1003, -0.0763, ..., 0.0213, -0.0407, -0.0764], + [ 0.0976, -0.1324, -0.1313, ..., -0.0584, 0.1243, -0.0299], + ..., + [-0.1592, -0.0802, 0.1019, ..., 0.0417, -0.1858, -0.0003], + [ 0.0791, -0.0018, -0.2546, ..., -0.0594, 0.1365, -0.0274], + [-0.1787, -0.1262, 0.0271, ..., -0.0116, -0.1545, -0.0597]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 8.8476e-09, 2.7474e-08, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 4.6566e-10, -6.2864e-08, 2.5518e-07, ..., 0.0000e+00, + 5.1223e-09, 0.0000e+00], + [-3.7253e-09, 3.6322e-08, 3.3062e-08, ..., 0.0000e+00, + -1.0710e-08, 0.0000e+00], + ..., + [ 1.3970e-09, 1.9930e-07, -6.8173e-07, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00], + [-6.5193e-09, 3.6322e-08, 2.6543e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 3.2736e-07, 6.3842e-07, ..., 0.0000e+00, + 1.6298e-08, 0.0000e+00]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0042, 0.0172, -0.0162, 0.0151, 0.0288, -0.0160, -0.0183, -0.0135, + -0.0309, -0.0323], device='cuda:0'), grad: tensor([ 1.2107e-08, 3.8277e-07, 1.2247e-07, -3.8976e-07, -2.5108e-06, + 4.9826e-07, 2.5984e-07, -7.9954e-07, 1.7369e-07, 2.2594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 216.55, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4839 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.2048, -0.1235, -0.0824, ..., -0.0034, -0.0460, -0.0118], + [-0.0952, 0.1004, -0.0765, ..., 0.0213, -0.0408, -0.0764], + [ 0.0979, -0.1329, -0.1323, ..., -0.0584, 0.1245, -0.0299], + ..., + [-0.1594, -0.0803, 0.1018, ..., 0.0417, -0.1862, -0.0003], + [ 0.0793, -0.0017, -0.2557, ..., -0.0594, 0.1368, -0.0274], + [-0.1793, -0.1266, 0.0272, ..., -0.0116, -0.1551, -0.0597]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 3.2596e-09, -9.3132e-09, 3.1665e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-2.7940e-09, 9.3132e-10, 3.3528e-08, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-09, -2.3236e-07, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [-9.7789e-08, -2.2352e-08, 1.8626e-09, ..., 0.0000e+00, + -1.0105e-07, 0.0000e+00], + [ 2.7940e-09, 6.0536e-09, 1.4016e-07, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0042, 0.0172, -0.0164, 0.0157, 0.0289, -0.0164, -0.0183, -0.0137, + -0.0309, -0.0323], device='cuda:0'), grad: tensor([-7.0315e-08, 7.4925e-07, 6.7474e-07, 8.7079e-08, 2.3283e-08, + 1.5413e-07, 8.2888e-08, -2.2054e-06, 1.3039e-08, 5.0059e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 216.83, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.5266 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.13 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.2054, -0.1238, -0.0825, ..., -0.0034, -0.0465, -0.0118], + [-0.0958, 0.1007, -0.0768, ..., 0.0213, -0.0409, -0.0764], + [ 0.0984, -0.1334, -0.1328, ..., -0.0584, 0.1248, -0.0299], + ..., + [-0.1600, -0.0804, 0.1028, ..., 0.0417, -0.1866, -0.0003], + [ 0.0802, -0.0017, -0.2566, ..., -0.0597, 0.1374, -0.0275], + [-0.1801, -0.1269, 0.0268, ..., -0.0119, -0.1563, -0.0597]], + device='cuda:0'), grad: tensor([[ 9.8255e-08, 1.2014e-07, -3.8184e-08, ..., 0.0000e+00, + 9.3598e-08, 0.0000e+00], + [ 7.1712e-08, -1.6354e-06, 1.3039e-08, ..., 0.0000e+00, + -4.9593e-07, 0.0000e+00], + [ 2.1094e-07, 2.9430e-07, -1.7695e-08, ..., 0.0000e+00, + 3.6275e-07, 0.0000e+00], + ..., + [ 1.6531e-07, 2.3097e-07, 1.1269e-07, ..., 0.0000e+00, + 2.9756e-07, 0.0000e+00], + [-1.8487e-06, -7.1432e-07, 4.6566e-09, ..., 0.0000e+00, + -2.6375e-06, 0.0000e+00], + [ 5.9512e-07, 1.0636e-06, 2.9337e-08, ..., 0.0000e+00, + 1.1362e-06, 0.0000e+00]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0045, 0.0172, -0.0163, 0.0162, 0.0285, -0.0170, -0.0183, -0.0134, + -0.0310, -0.0320], device='cuda:0'), grad: tensor([ 1.0477e-07, -4.1239e-06, 7.2690e-07, 1.2815e-06, 3.4506e-07, + 6.7847e-07, 8.5775e-07, 1.4994e-06, -5.2527e-06, 3.8743e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 216.67, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5095 re_mapping 0.0048 re_causal 0.0143 /// teacc 99.01 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.2063, -0.1238, -0.0835, ..., -0.0036, -0.0458, -0.0118], + [-0.0964, 0.1027, -0.0751, ..., 0.0213, -0.0410, -0.0764], + [ 0.1010, -0.1383, -0.1339, ..., -0.0584, 0.1276, -0.0299], + ..., + [-0.1607, -0.0815, 0.1019, ..., 0.0417, -0.1870, -0.0003], + [ 0.0786, -0.0018, -0.2582, ..., -0.0602, 0.1350, -0.0277], + [-0.1806, -0.1277, 0.0266, ..., -0.0123, -0.1576, -0.0597]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 9.6392e-08, 3.7253e-09, ..., 0.0000e+00, + 2.0023e-08, 0.0000e+00], + [ 2.3283e-09, -3.5111e-07, 2.2817e-08, ..., 0.0000e+00, + -3.4459e-08, 0.0000e+00], + [ 1.9558e-08, 2.3283e-08, 4.1910e-09, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 1.2340e-07, 7.6368e-08, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [-4.7032e-08, 1.3877e-07, 1.5460e-07, ..., 0.0000e+00, + -5.1688e-08, 0.0000e+00], + [ 1.6764e-08, -8.1817e-07, -1.1828e-06, ..., 0.0000e+00, + 2.0955e-08, 0.0000e+00]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0045, 0.0185, -0.0158, 0.0166, 0.0291, -0.0169, -0.0186, -0.0145, + -0.0328, -0.0320], device='cuda:0'), grad: tensor([ 2.0815e-07, -4.8382e-07, 1.3411e-07, -6.3330e-08, 7.0147e-06, + 7.4506e-08, 2.3702e-07, 7.0734e-07, 1.1660e-06, -8.9854e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 242---------------------------------------------------- +epoch 242, time 217.68, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4935 re_mapping 0.0045 re_causal 0.0139 /// teacc 99.20 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.2068, -0.1240, -0.0836, ..., -0.0039, -0.0460, -0.0118], + [-0.0972, 0.1033, -0.0736, ..., 0.0213, -0.0412, -0.0764], + [ 0.1015, -0.1384, -0.1358, ..., -0.0585, 0.1279, -0.0299], + ..., + [-0.1610, -0.0821, 0.1014, ..., 0.0421, -0.1874, -0.0003], + [ 0.0787, -0.0018, -0.2592, ..., -0.0602, 0.1351, -0.0278], + [-0.1814, -0.1287, 0.0275, ..., -0.0124, -0.1582, -0.0597]], + device='cuda:0'), grad: tensor([[ 1.3504e-08, 1.0245e-08, 2.3283e-09, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [ 7.4971e-08, 1.5367e-07, 1.4342e-07, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00], + [ 1.1977e-06, 1.1176e-08, 2.2259e-07, ..., 0.0000e+00, + -4.8894e-08, 0.0000e+00], + ..., + [-1.2582e-06, 6.7987e-08, -2.6077e-07, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 2.0070e-07, 1.2573e-07, 6.2399e-08, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00], + [ 6.8452e-08, 5.5414e-08, 4.3772e-08, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0044, 0.0186, -0.0157, 0.0167, 0.0283, -0.0170, -0.0186, -0.0147, + -0.0328, -0.0316], device='cuda:0'), grad: tensor([-4.9360e-08, 6.3516e-07, 6.1356e-06, 1.8878e-06, -7.5344e-07, + -3.6601e-06, 1.0803e-06, -6.5379e-06, 7.9349e-07, 4.8662e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 216.77, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.4528 re_mapping 0.0045 re_causal 0.0135 /// teacc 99.11 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.2069, -0.1231, -0.0836, ..., -0.0039, -0.0446, -0.0118], + [-0.0975, 0.1033, -0.0737, ..., 0.0213, -0.0413, -0.0764], + [ 0.1016, -0.1385, -0.1361, ..., -0.0585, 0.1282, -0.0299], + ..., + [-0.1618, -0.0821, 0.1016, ..., 0.0421, -0.1877, -0.0003], + [ 0.0793, -0.0017, -0.2581, ..., -0.0603, 0.1352, -0.0279], + [-0.1825, -0.1291, 0.0269, ..., -0.0124, -0.1588, -0.0597]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 9.3132e-09, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 9.3132e-10, -6.2399e-08, 1.2293e-07, ..., 0.0000e+00, + -2.3283e-09, 0.0000e+00], + [-9.3132e-10, 5.1223e-09, 9.3132e-09, ..., 0.0000e+00, + -2.3283e-09, 0.0000e+00], + ..., + [ 1.3970e-09, 2.5611e-08, -4.2422e-07, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 1.8161e-08, 1.1642e-08, 4.1910e-09, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 4.6566e-09, 1.0710e-08, 2.6450e-07, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0038, 0.0186, -0.0154, 0.0163, 0.0284, -0.0167, -0.0192, -0.0147, + -0.0324, -0.0319], device='cuda:0'), grad: tensor([-1.4435e-08, 1.4389e-07, 2.4214e-08, 1.0338e-07, -1.4342e-07, + -4.7032e-08, -1.8626e-09, -7.6368e-07, 6.9849e-08, 6.3749e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 216.67, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.5318 re_mapping 0.0046 re_causal 0.0134 /// teacc 99.11 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.2085, -0.1237, -0.0838, ..., -0.0039, -0.0452, -0.0118], + [-0.0986, 0.1034, -0.0742, ..., 0.0213, -0.0430, -0.0764], + [ 0.1018, -0.1389, -0.1382, ..., -0.0587, 0.1292, -0.0299], + ..., + [-0.1618, -0.0822, 0.1020, ..., 0.0426, -0.1879, -0.0003], + [ 0.0799, -0.0017, -0.2593, ..., -0.0604, 0.1357, -0.0279], + [-0.1841, -0.1299, 0.0268, ..., -0.0126, -0.1612, -0.0597]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 9.3132e-09, 8.3819e-09, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 4.6566e-09, -1.8720e-07, 8.3819e-09, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [-4.5169e-08, 1.9558e-08, 5.1223e-09, ..., 0.0000e+00, + -9.5461e-08, 0.0000e+00], + ..., + [ 4.1910e-09, 6.0536e-08, 1.0710e-08, ..., 0.0000e+00, + 6.9849e-09, 0.0000e+00], + [ 2.1420e-08, 2.4680e-08, 1.8161e-08, ..., 0.0000e+00, + 2.0023e-08, 0.0000e+00], + [ 1.0245e-08, 3.2596e-08, -6.8452e-08, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0039, 0.0179, -0.0141, 0.0192, 0.0292, -0.0196, -0.0192, -0.0145, + -0.0321, -0.0324], device='cuda:0'), grad: tensor([-1.3784e-07, -2.5146e-07, -9.0338e-08, 1.5320e-07, 3.3528e-08, + -3.5856e-08, 1.3970e-08, 2.5053e-07, 1.7742e-07, -1.1874e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 216.53, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4860 re_mapping 0.0049 re_causal 0.0140 /// teacc 99.13 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.2095, -0.1239, -0.0837, ..., -0.0039, -0.0455, -0.0118], + [-0.0989, 0.1035, -0.0743, ..., 0.0213, -0.0431, -0.0764], + [ 0.1017, -0.1391, -0.1399, ..., -0.0587, 0.1293, -0.0299], + ..., + [-0.1617, -0.0822, 0.1021, ..., 0.0424, -0.1880, -0.0003], + [ 0.0803, -0.0017, -0.2593, ..., -0.0605, 0.1360, -0.0279], + [-0.1850, -0.1306, 0.0267, ..., -0.0127, -0.1618, -0.0597]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 4.6566e-09, 1.8626e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 2.6077e-08, -9.1735e-08, 2.1886e-08, ..., 0.0000e+00, + 1.7229e-08, 0.0000e+00], + [-1.0710e-08, 3.1199e-08, 1.2573e-08, ..., 0.0000e+00, + -2.2817e-08, 0.0000e+00], + ..., + [ 3.4925e-08, 5.4017e-08, -7.9162e-08, ..., 0.0000e+00, + 2.1886e-08, 0.0000e+00], + [ 1.1222e-07, -1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + -5.1688e-08, 0.0000e+00], + [ 8.6660e-07, 9.5461e-08, 8.1025e-08, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0041, 0.0179, -0.0143, 0.0189, 0.0292, -0.0192, -0.0189, -0.0145, + -0.0319, -0.0326], device='cuda:0'), grad: tensor([-1.7229e-08, 1.3504e-08, 9.5461e-08, 3.5902e-07, -1.2852e-07, + -4.3884e-06, 2.8592e-07, -6.8452e-08, 4.2934e-07, 3.4161e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 216.76, cls_loss 0.0014 cls_loss_mapping 0.0024 cls_loss_causal 0.5221 re_mapping 0.0044 re_causal 0.0132 /// teacc 98.99 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.2111, -0.1242, -0.0838, ..., -0.0039, -0.0460, -0.0118], + [-0.0993, 0.1037, -0.0745, ..., 0.0213, -0.0433, -0.0764], + [ 0.1015, -0.1394, -0.1413, ..., -0.0587, 0.1293, -0.0299], + ..., + [-0.1616, -0.0824, 0.1024, ..., 0.0424, -0.1882, -0.0003], + [ 0.0810, -0.0016, -0.2599, ..., -0.0608, 0.1367, -0.0280], + [-0.1874, -0.1314, 0.0267, ..., -0.0127, -0.1633, -0.0597]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 1.5832e-07, 2.2352e-08, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 2.2817e-08, -3.9786e-06, 1.2713e-07, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + [ 3.7253e-09, 3.1590e-06, 1.0990e-07, ..., 0.0000e+00, + -8.8476e-09, 0.0000e+00], + ..., + [ 1.6764e-08, 3.2224e-07, -3.5390e-07, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + [-2.1551e-06, -3.9162e-07, 5.5879e-08, ..., 0.0000e+00, + -1.5693e-06, 0.0000e+00], + [-1.1465e-06, 3.1851e-07, -1.4473e-06, ..., 0.0000e+00, + 6.9849e-09, 0.0000e+00]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0041, 0.0179, -0.0146, 0.0186, 0.0291, -0.0193, -0.0184, -0.0145, + -0.0316, -0.0326], device='cuda:0'), grad: tensor([ 4.2468e-07, -8.5235e-06, 7.7263e-06, 4.9211e-06, 5.6326e-06, + -3.6322e-07, 1.1362e-06, -5.8999e-07, -4.3809e-06, -6.0163e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 216.77, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.5111 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.04 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.2118, -0.1244, -0.0833, ..., -0.0039, -0.0470, -0.0118], + [-0.1010, 0.1035, -0.0747, ..., 0.0212, -0.0465, -0.0764], + [ 0.1007, -0.1402, -0.1428, ..., -0.0588, 0.1318, -0.0299], + ..., + [-0.1602, -0.0824, 0.1028, ..., 0.0422, -0.1889, -0.0003], + [ 0.0817, -0.0004, -0.2607, ..., -0.0613, 0.1385, -0.0285], + [-0.1881, -0.1322, 0.0264, ..., -0.0129, -0.1641, -0.0597]], + device='cuda:0'), grad: tensor([[ 2.5611e-08, 1.0710e-08, 1.2107e-08, ..., 0.0000e+00, + 3.4459e-08, 0.0000e+00], + [ 5.1223e-09, -3.8603e-07, 2.5751e-07, ..., 0.0000e+00, + 1.0710e-08, 0.0000e+00], + [-8.5216e-08, 9.3132e-09, 1.7509e-07, ..., 0.0000e+00, + -8.5682e-08, 0.0000e+00], + ..., + [ 4.8429e-08, 8.1491e-08, -7.1852e-07, ..., 0.0000e+00, + 7.3109e-08, 0.0000e+00], + [-3.3528e-08, 3.3528e-08, 3.3528e-08, ..., 0.0000e+00, + -3.7719e-08, 0.0000e+00], + [ 3.7253e-09, 3.9116e-08, 4.6566e-08, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0039, 0.0173, -0.0135, 0.0174, 0.0292, -0.0192, -0.0183, -0.0141, + -0.0302, -0.0328], device='cuda:0'), grad: tensor([ 1.6671e-07, 7.4226e-07, 6.4401e-07, -1.0543e-06, 3.9116e-07, + 1.4529e-07, 5.4948e-08, -1.5823e-06, 1.5972e-07, 3.4599e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 216.76, cls_loss 0.0013 cls_loss_mapping 0.0022 cls_loss_causal 0.5093 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.05 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.2124, -0.1245, -0.0837, ..., -0.0039, -0.0473, -0.0118], + [-0.1019, 0.1037, -0.0748, ..., 0.0212, -0.0467, -0.0764], + [ 0.1017, -0.1407, -0.1466, ..., -0.0589, 0.1322, -0.0299], + ..., + [-0.1609, -0.0824, 0.1031, ..., 0.0421, -0.1895, -0.0004], + [ 0.0816, -0.0004, -0.2624, ..., -0.0616, 0.1385, -0.0308], + [-0.1882, -0.1337, 0.0263, ..., -0.0130, -0.1646, -0.0598]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 2.0023e-08, 1.8626e-09, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 7.4506e-09, -5.7789e-07, 3.7160e-07, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00], + [-5.5879e-09, 5.3458e-07, 1.3504e-08, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + ..., + [-4.1910e-09, -7.7300e-08, -6.3097e-07, ..., 0.0000e+00, + 2.0023e-08, 0.0000e+00], + [-2.9802e-08, 4.6566e-09, 4.6566e-09, ..., 0.0000e+00, + -2.9802e-08, 0.0000e+00], + [ 6.0536e-09, 4.9826e-08, 1.4016e-07, ..., 0.0000e+00, + 2.4680e-08, 0.0000e+00]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0039, 0.0170, -0.0133, 0.0172, 0.0296, -0.0189, -0.0184, -0.0139, + -0.0306, -0.0330], device='cuda:0'), grad: tensor([-2.3283e-09, -1.8161e-07, 7.9256e-07, -1.4529e-07, 2.6915e-07, + 5.6811e-08, -3.3993e-08, -1.2163e-06, -3.7253e-09, 4.6566e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 216.74, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.5031 re_mapping 0.0045 re_causal 0.0135 /// teacc 99.10 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.2128, -0.1246, -0.0837, ..., -0.0039, -0.0477, -0.0119], + [-0.1033, 0.1037, -0.0755, ..., 0.0211, -0.0470, -0.0765], + [ 0.1014, -0.1407, -0.1511, ..., -0.0590, 0.1320, -0.0300], + ..., + [-0.1584, -0.0822, 0.1053, ..., 0.0421, -0.1879, -0.0006], + [ 0.0814, -0.0006, -0.2647, ..., -0.0618, 0.1385, -0.0319], + [-0.1880, -0.1335, 0.0267, ..., -0.0131, -0.1653, -0.0598]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 4.1910e-09, 4.3306e-08, ..., 0.0000e+00, + 5.1223e-09, 0.0000e+00], + [ 1.1176e-08, 2.3469e-07, 3.7672e-07, ..., 0.0000e+00, + 1.2573e-08, 0.0000e+00], + [-7.3109e-08, 4.0513e-08, 5.9139e-08, ..., 0.0000e+00, + -1.1222e-07, 0.0000e+00], + ..., + [ 8.7079e-08, 5.4622e-07, -1.4395e-05, ..., 0.0000e+00, + 6.0536e-08, 0.0000e+00], + [ 9.7789e-09, 1.0012e-07, 1.3690e-07, ..., 0.0000e+00, + 1.3504e-08, 9.3132e-10], + [ 4.9826e-08, 1.0021e-06, 1.5542e-05, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0038, 0.0168, -0.0139, 0.0164, 0.0290, -0.0187, -0.0187, -0.0128, + -0.0310, -0.0325], device='cuda:0'), grad: tensor([-3.6182e-07, 1.4901e-06, 1.8626e-08, 2.2119e-07, -9.0376e-06, + 1.9791e-07, 5.1130e-07, -4.8220e-05, 7.4925e-07, 5.4508e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 216.67, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4656 re_mapping 0.0044 re_causal 0.0125 /// teacc 99.02 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.2133, -0.1247, -0.0842, ..., -0.0040, -0.0483, -0.0120], + [-0.1035, 0.1040, -0.0756, ..., 0.0210, -0.0470, -0.0768], + [ 0.1014, -0.1419, -0.1522, ..., -0.0592, 0.1317, -0.0301], + ..., + [-0.1584, -0.0823, 0.1058, ..., 0.0419, -0.1881, -0.0009], + [ 0.0814, -0.0006, -0.2657, ..., -0.0625, 0.1386, -0.0332], + [-0.1868, -0.1333, 0.0271, ..., -0.0133, -0.1660, -0.0600]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 3.7253e-09, 6.4261e-08, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 7.4506e-09, -5.0664e-07, 2.9756e-07, ..., 0.0000e+00, + -2.8405e-08, 0.0000e+00], + [ 2.7940e-09, 5.0105e-07, 2.8498e-07, ..., 0.0000e+00, + 3.2131e-08, 0.0000e+00], + ..., + [ 1.2107e-08, 4.9826e-08, -2.2110e-06, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 3.0734e-06, 5.5879e-09, 1.4063e-07, ..., 0.0000e+00, + 1.2349e-06, 0.0000e+00], + [ 1.0850e-07, 2.7474e-08, 2.5099e-07, ..., 0.0000e+00, + 4.2375e-08, 0.0000e+00]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0036, 0.0169, -0.0147, 0.0162, 0.0282, -0.0181, -0.0190, -0.0126, + -0.0313, -0.0319], device='cuda:0'), grad: tensor([ 4.5681e-07, 8.9034e-07, 3.1013e-06, 2.5146e-06, 3.7085e-06, + -9.5293e-06, 2.9523e-07, -1.1593e-05, 8.5682e-06, 1.6168e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 216.86, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.4615 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.07 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.2140, -0.1250, -0.0873, ..., -0.0040, -0.0489, -0.0120], + [-0.1038, 0.1040, -0.0756, ..., 0.0210, -0.0471, -0.0768], + [ 0.1017, -0.1421, -0.1529, ..., -0.0593, 0.1321, -0.0301], + ..., + [-0.1585, -0.0824, 0.1057, ..., 0.0418, -0.1886, -0.0009], + [ 0.0805, -0.0012, -0.2665, ..., -0.0630, 0.1388, -0.0332], + [-0.1861, -0.1339, 0.0297, ..., -0.0136, -0.1651, -0.0600]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 5.5879e-09, 4.1910e-09, ..., 6.0536e-09, + 2.3749e-08, 0.0000e+00], + [ 1.4156e-07, 4.1910e-09, 1.2061e-07, ..., 1.0710e-08, + 2.9523e-07, 0.0000e+00], + [-1.7369e-07, 1.0245e-08, 9.3132e-09, ..., 2.2817e-08, + -2.5006e-07, 0.0000e+00], + ..., + [-1.6764e-08, 3.9116e-08, -2.2165e-07, ..., 1.5832e-08, + 6.4261e-08, 0.0000e+00], + [ 6.9849e-09, 6.1467e-08, 4.6100e-08, ..., 1.1642e-08, + 3.5390e-08, 0.0000e+00], + [ 1.2107e-08, 3.4412e-07, 4.1025e-07, ..., 2.0955e-08, + 8.6147e-08, 0.0000e+00]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0060, 0.0168, -0.0146, 0.0158, 0.0278, -0.0209, -0.0161, -0.0130, + -0.0320, -0.0285], device='cuda:0'), grad: tensor([-2.5611e-08, 1.1232e-06, -5.2573e-07, -1.7444e-06, -1.1167e-06, + 5.4948e-07, 1.5600e-07, -5.6392e-07, 3.5670e-07, 1.7826e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 216.45, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4768 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.08 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.2142, -0.1253, -0.0874, ..., -0.0042, -0.0494, -0.0120], + [-0.1042, 0.1045, -0.0748, ..., 0.0209, -0.0471, -0.0768], + [ 0.1019, -0.1429, -0.1550, ..., -0.0594, 0.1323, -0.0301], + ..., + [-0.1587, -0.0827, 0.1058, ..., 0.0395, -0.1888, -0.0009], + [ 0.0804, -0.0015, -0.2675, ..., -0.0651, 0.1389, -0.0332], + [-0.1865, -0.1355, 0.0322, ..., -0.0140, -0.1658, -0.0600]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 1.5367e-08, 7.4506e-09, ..., 0.0000e+00, + 2.4680e-08, 0.0000e+00], + [ 1.4249e-06, 6.0722e-07, 6.4494e-07, ..., 0.0000e+00, + 4.5337e-06, 0.0000e+00], + [ 5.7742e-08, 2.6543e-08, 6.1141e-07, ..., 0.0000e+00, + 1.3458e-07, 0.0000e+00], + ..., + [ 2.0489e-08, -2.4820e-07, -1.3737e-06, ..., 0.0000e+00, + 4.1444e-08, 0.0000e+00], + [-2.9616e-06, -9.3831e-07, 2.6077e-08, ..., 0.0000e+00, + -8.2627e-06, 0.0000e+00], + [ 7.6788e-07, 3.3760e-07, 1.6764e-08, ..., 0.0000e+00, + 1.5125e-06, 0.0000e+00]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0060, 0.0171, -0.0151, 0.0158, 0.0250, -0.0208, -0.0161, -0.0132, + -0.0322, -0.0257], device='cuda:0'), grad: tensor([-2.1867e-06, 1.1601e-05, 2.2277e-06, 1.7583e-06, 1.7649e-07, + 2.9597e-06, 1.3364e-06, -3.5260e-06, -1.9774e-05, 5.3905e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 216.64, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4838 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.02 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.2146, -0.1254, -0.0872, ..., -0.0042, -0.0502, -0.0120], + [-0.1048, 0.1047, -0.0752, ..., 0.0209, -0.0474, -0.0768], + [ 0.1034, -0.1430, -0.1548, ..., -0.0594, 0.1328, -0.0301], + ..., + [-0.1589, -0.0829, 0.1062, ..., 0.0395, -0.1891, -0.0009], + [ 0.0805, -0.0015, -0.2683, ..., -0.0652, 0.1392, -0.0332], + [-0.1871, -0.1357, 0.0326, ..., -0.0141, -0.1676, -0.0600]], + device='cuda:0'), grad: tensor([[ 6.6124e-08, 1.1595e-07, 9.3132e-10, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 8.8941e-08, 8.0094e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8347e-07, 0.0000e+00], + [-7.5903e-08, 5.0291e-08, 9.3132e-10, ..., 0.0000e+00, + -5.9139e-08, 0.0000e+00], + ..., + [ 1.1222e-07, 4.8429e-08, 1.7229e-08, ..., 0.0000e+00, + 1.7788e-07, 0.0000e+00], + [-1.8114e-07, -1.6158e-07, 3.7253e-09, ..., 0.0000e+00, + -5.1269e-07, 0.0000e+00], + [ 2.5099e-07, 2.5192e-07, 1.0384e-07, ..., 0.0000e+00, + 2.3283e-08, 0.0000e+00]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0064, 0.0169, -0.0145, 0.0157, 0.0245, -0.0208, -0.0158, -0.0132, + -0.0320, -0.0254], device='cuda:0'), grad: tensor([-8.7498e-07, 7.3342e-07, -1.3504e-08, 1.6317e-06, -1.9334e-06, + -1.4603e-06, -3.9581e-07, 2.2482e-06, -1.3364e-06, 1.4286e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 216.76, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5075 re_mapping 0.0044 re_causal 0.0135 /// teacc 99.06 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.2148, -0.1254, -0.0900, ..., -0.0041, -0.0503, -0.0132], + [-0.1050, 0.1050, -0.0752, ..., 0.0209, -0.0475, -0.0785], + [ 0.1035, -0.1435, -0.1551, ..., -0.0594, 0.1329, -0.0327], + ..., + [-0.1595, -0.0830, 0.1064, ..., 0.0394, -0.1897, -0.0011], + [ 0.0811, -0.0016, -0.2685, ..., -0.0653, 0.1396, -0.0334], + [-0.1874, -0.1369, 0.0334, ..., -0.0142, -0.1682, -0.0609]], + device='cuda:0'), grad: tensor([[ 6.0536e-09, 1.3970e-09, 9.7789e-09, ..., 0.0000e+00, + 3.2550e-07, 0.0000e+00], + [ 2.3749e-08, -7.1712e-08, 1.3970e-09, ..., 0.0000e+00, + 1.4063e-07, 0.0000e+00], + [ 4.1910e-09, 4.6566e-09, 1.3970e-09, ..., 0.0000e+00, + 3.9535e-07, 0.0000e+00], + ..., + [ 2.1886e-08, 4.3306e-08, 4.6566e-10, ..., 0.0000e+00, + 1.4342e-07, 0.0000e+00], + [-2.5844e-07, 6.5193e-09, 4.6566e-10, ..., 0.0000e+00, + -3.7253e-08, 0.0000e+00], + [ 6.1002e-08, 2.5611e-08, -1.3970e-08, ..., 0.0000e+00, + 2.2585e-07, 0.0000e+00]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0083, 0.0170, -0.0147, 0.0157, 0.0246, -0.0206, -0.0161, -0.0133, + -0.0319, -0.0240], device='cuda:0'), grad: tensor([ 1.1139e-06, 3.2922e-07, 1.4137e-06, -6.2287e-06, 6.1002e-08, + 1.9539e-06, 2.1467e-07, 5.7276e-07, -1.1316e-07, 6.8219e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 216.64, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.4854 re_mapping 0.0046 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.2144, -0.1254, -0.0900, ..., -0.0041, -0.0507, -0.0124], + [-0.1051, 0.1051, -0.0756, ..., 0.0209, -0.0479, -0.0787], + [ 0.1037, -0.1438, -0.1556, ..., -0.0594, 0.1335, -0.0328], + ..., + [-0.1596, -0.0831, 0.1065, ..., 0.0394, -0.1899, -0.0012], + [ 0.0804, -0.0020, -0.2696, ..., -0.0654, 0.1394, -0.0337], + [-0.1884, -0.1395, 0.0333, ..., -0.0142, -0.1707, -0.0631]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 8.8476e-09, 9.3132e-10, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00], + [ 7.9162e-09, -5.9232e-07, 5.1223e-09, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [-1.2573e-08, 1.3970e-08, 1.3970e-09, ..., 0.0000e+00, + -1.7695e-08, 0.0000e+00], + ..., + [ 1.0245e-08, 2.5844e-07, -2.9337e-08, ..., 0.0000e+00, + 1.8161e-08, 0.0000e+00], + [-6.8918e-08, 3.7719e-08, 1.3970e-09, ..., 0.0000e+00, + -1.5274e-07, 0.0000e+00], + [ 2.9802e-08, 9.4529e-08, -3.7253e-08, ..., 0.0000e+00, + 4.2375e-08, 0.0000e+00]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0081, 0.0169, -0.0146, 0.0160, 0.0248, -0.0199, -0.0168, -0.0133, + -0.0330, -0.0242], device='cuda:0'), grad: tensor([-6.9849e-08, -9.7044e-07, 3.7253e-09, 3.0128e-07, 6.6264e-07, + -1.9604e-07, 1.4435e-07, 3.5483e-07, -2.0722e-07, -7.9162e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 217.11, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4962 re_mapping 0.0045 re_causal 0.0135 /// teacc 98.96 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.2148, -0.1256, -0.0900, ..., -0.0041, -0.0509, -0.0124], + [-0.1059, 0.1037, -0.0757, ..., 0.0209, -0.0480, -0.0787], + [ 0.1038, -0.1440, -0.1558, ..., -0.0595, 0.1336, -0.0328], + ..., + [-0.1598, -0.0832, 0.1067, ..., 0.0394, -0.1900, -0.0012], + [ 0.0811, -0.0022, -0.2700, ..., -0.0654, 0.1402, -0.0337], + [-0.1893, -0.1408, 0.0333, ..., -0.0143, -0.1720, -0.0631]], + device='cuda:0'), grad: tensor([[ 4.0513e-08, 1.2824e-06, 1.9558e-08, ..., 0.0000e+00, + 9.4622e-07, 0.0000e+00], + [ 1.2247e-07, -4.2170e-06, 7.4506e-08, ..., 0.0000e+00, + -2.9486e-06, 0.0000e+00], + [ 1.3970e-08, 1.2154e-07, 2.9337e-08, ..., 0.0000e+00, + -5.5414e-08, 0.0000e+00], + ..., + [ 6.1728e-06, 3.0780e-07, 3.2205e-06, ..., 0.0000e+00, + 2.1979e-07, 0.0000e+00], + [-5.8534e-07, 3.9069e-07, 4.2375e-08, ..., 0.0000e+00, + -7.0594e-07, 0.0000e+00], + [-5.9791e-06, 5.8627e-07, -2.9970e-06, ..., 0.0000e+00, + 9.2899e-07, 0.0000e+00]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0081, 0.0160, -0.0146, 0.0172, 0.0249, -0.0206, -0.0164, -0.0130, + -0.0328, -0.0243], device='cuda:0'), grad: tensor([ 3.7104e-06, -9.0599e-06, 1.5702e-06, 1.1614e-06, -1.5534e-06, + 9.9838e-07, 8.2552e-06, 1.5748e-04, 7.6462e-07, -1.6332e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 216.92, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.5001 re_mapping 0.0048 re_causal 0.0137 /// teacc 98.97 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.2151, -0.1259, -0.0900, ..., -0.0041, -0.0513, -0.0125], + [-0.1061, 0.1040, -0.0760, ..., 0.0208, -0.0480, -0.0787], + [ 0.1040, -0.1446, -0.1562, ..., -0.0595, 0.1336, -0.0328], + ..., + [-0.1602, -0.0833, 0.1071, ..., 0.0394, -0.1906, -0.0012], + [ 0.0819, -0.0023, -0.2704, ..., -0.0654, 0.1411, -0.0337], + [-0.1897, -0.1423, 0.0332, ..., -0.0144, -0.1734, -0.0634]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-08, 6.1467e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 6.0536e-09, 5.4669e-07, 1.2368e-06, ..., 0.0000e+00, + 1.4435e-08, 0.0000e+00], + [ 3.4459e-08, 7.5903e-08, 3.2084e-07, ..., 0.0000e+00, + 5.2620e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 1.3225e-07, -9.6299e-07, ..., 0.0000e+00, + 2.0023e-08, 0.0000e+00], + [-3.6787e-08, 1.9791e-07, 1.9697e-07, ..., 0.0000e+00, + -4.0047e-08, 0.0000e+00], + [ 1.4901e-08, 4.3884e-06, 4.4703e-06, ..., 0.0000e+00, + 4.3772e-08, 0.0000e+00]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0082, 0.0161, -0.0148, 0.0167, 0.0250, -0.0203, -0.0163, -0.0131, + -0.0323, -0.0244], device='cuda:0'), grad: tensor([ 1.7146e-06, 1.4067e-05, 5.9791e-06, 1.1779e-05, -1.8477e-05, + -2.6137e-05, 8.2515e-07, -2.1115e-05, 8.3772e-07, 3.0547e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 217.02, cls_loss 0.0015 cls_loss_mapping 0.0032 cls_loss_causal 0.4896 re_mapping 0.0044 re_causal 0.0129 /// teacc 98.93 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.2156, -0.1257, -0.0899, ..., -0.0041, -0.0498, -0.0125], + [-0.1066, 0.1056, -0.0765, ..., 0.0208, -0.0457, -0.0787], + [ 0.1043, -0.1477, -0.1572, ..., -0.0596, 0.1318, -0.0328], + ..., + [-0.1602, -0.0834, 0.1096, ..., 0.0394, -0.1912, -0.0012], + [ 0.0834, -0.0018, -0.2712, ..., -0.0655, 0.1424, -0.0337], + [-0.1921, -0.1459, 0.0324, ..., -0.0144, -0.1773, -0.0634]], + device='cuda:0'), grad: tensor([[ 2.0489e-08, -7.3109e-08, -2.7940e-08, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + [ 1.0738e-06, 2.2212e-07, 2.1420e-08, ..., 0.0000e+00, + 1.1967e-06, 0.0000e+00], + [-2.1886e-06, -9.2527e-07, 1.3970e-08, ..., 0.0000e+00, + -2.4512e-06, 0.0000e+00], + ..., + [ 8.0094e-08, 1.8300e-07, -5.2294e-07, ..., 0.0000e+00, + 1.0477e-07, 0.0000e+00], + [ 2.6915e-07, 1.4296e-07, 1.3504e-08, ..., 0.0000e+00, + 3.0035e-07, 0.0000e+00], + [ 1.3271e-07, 1.2340e-07, 4.0652e-07, ..., 0.0000e+00, + 1.4482e-07, 0.0000e+00]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0080, 0.0184, -0.0179, 0.0171, 0.0253, -0.0206, -0.0164, -0.0121, + -0.0313, -0.0251], device='cuda:0'), grad: tensor([-9.3272e-07, 3.6936e-06, -7.7635e-06, 2.4196e-06, 2.6077e-07, + 8.3353e-08, 3.7346e-07, -3.1814e-06, 1.1334e-06, 3.8929e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 216.58, cls_loss 0.0015 cls_loss_mapping 0.0021 cls_loss_causal 0.4796 re_mapping 0.0045 re_causal 0.0125 /// teacc 99.05 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.2165, -0.1259, -0.0899, ..., -0.0041, -0.0500, -0.0125], + [-0.1085, 0.1053, -0.0770, ..., 0.0208, -0.0461, -0.0787], + [ 0.1047, -0.1478, -0.1582, ..., -0.0596, 0.1326, -0.0328], + ..., + [-0.1599, -0.0829, 0.1108, ..., 0.0394, -0.1914, -0.0012], + [ 0.0840, -0.0021, -0.2719, ..., -0.0655, 0.1430, -0.0337], + [-0.1932, -0.1472, 0.0322, ..., -0.0144, -0.1808, -0.0634]], + device='cuda:0'), grad: tensor([[ 2.0023e-08, 6.0536e-08, -8.3819e-09, ..., 0.0000e+00, + 5.8440e-07, 0.0000e+00], + [ 1.3039e-08, -1.1399e-06, 1.3970e-08, ..., 0.0000e+00, + 3.7067e-07, 0.0000e+00], + [ 5.5879e-09, 1.0757e-07, 9.7789e-09, ..., 0.0000e+00, + 1.2529e-04, 0.0000e+00], + ..., + [ 4.6566e-10, 5.5134e-07, -1.5739e-07, ..., 0.0000e+00, + 9.2201e-08, 0.0000e+00], + [ 4.6566e-09, 1.9604e-07, 2.7940e-09, ..., 0.0000e+00, + 2.2817e-08, 0.0000e+00], + [ 3.2596e-09, 3.4925e-08, 1.1874e-07, ..., 0.0000e+00, + 5.4017e-08, 0.0000e+00]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0080, 0.0177, -0.0175, 0.0168, 0.0254, -0.0205, -0.0165, -0.0110, + -0.0313, -0.0253], device='cuda:0'), grad: tensor([ 1.5302e-06, -7.3202e-07, 3.3545e-04, -3.4475e-04, 3.3574e-07, + 6.0722e-06, -6.8918e-08, 7.5204e-07, 4.0280e-07, 4.4471e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 216.60, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4907 re_mapping 0.0043 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.2168, -0.1260, -0.0899, ..., -0.0041, -0.0502, -0.0125], + [-0.1094, 0.1047, -0.0792, ..., 0.0208, -0.0463, -0.0787], + [ 0.1046, -0.1479, -0.1585, ..., -0.0596, 0.1319, -0.0328], + ..., + [-0.1600, -0.0820, 0.1119, ..., 0.0394, -0.1916, -0.0012], + [ 0.0844, -0.0014, -0.2725, ..., -0.0656, 0.1436, -0.0338], + [-0.1940, -0.1484, 0.0322, ..., -0.0144, -0.1818, -0.0635]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.3970e-09, 1.1642e-09, ..., 0.0000e+00, + 1.7229e-08, 0.0000e+00], + [ 5.3551e-09, -8.3819e-08, 1.1642e-09, ..., 0.0000e+00, + 5.5414e-08, 0.0000e+00], + [-2.0955e-09, 4.6566e-09, 4.6566e-10, ..., 0.0000e+00, + -1.1083e-07, 0.0000e+00], + ..., + [ 1.8626e-09, 6.9616e-08, 2.0722e-08, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00], + [-2.1653e-08, -1.3271e-08, 3.2596e-09, ..., 0.0000e+00, + -2.0023e-08, 0.0000e+00], + [ 3.2596e-09, 7.6834e-09, -5.0990e-08, ..., 0.0000e+00, + 1.0943e-08, 0.0000e+00]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0080, 0.0174, -0.0179, 0.0174, 0.0255, -0.0205, -0.0165, -0.0106, + -0.0307, -0.0254], device='cuda:0'), grad: tensor([-6.1747e-07, 7.2876e-08, -3.0501e-07, 6.0303e-08, 1.2503e-07, + 3.9139e-07, 6.1234e-08, 2.2398e-07, -3.4925e-09, 1.2340e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 216.83, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.5199 re_mapping 0.0041 re_causal 0.0131 /// teacc 98.98 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.2177, -0.1261, -0.0900, ..., -0.0041, -0.0505, -0.0125], + [-0.1100, 0.1048, -0.0794, ..., 0.0208, -0.0465, -0.0787], + [ 0.1078, -0.1480, -0.1587, ..., -0.0596, 0.1348, -0.0328], + ..., + [-0.1633, -0.0823, 0.1118, ..., 0.0394, -0.1949, -0.0012], + [ 0.0842, -0.0016, -0.2732, ..., -0.0656, 0.1436, -0.0338], + [-0.1948, -0.1496, 0.0322, ..., -0.0144, -0.1823, -0.0635]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 8.3819e-09, 6.9849e-10, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 5.5879e-09, -5.8813e-07, 1.3970e-09, ..., 0.0000e+00, + -1.7020e-07, 0.0000e+00], + [ 9.7789e-09, 4.8243e-07, 4.6566e-10, ..., 0.0000e+00, + 2.2701e-07, 0.0000e+00], + ..., + [ 3.4925e-09, 1.1409e-07, 8.6846e-08, ..., 0.0000e+00, + 1.0617e-07, 0.0000e+00], + [-3.3528e-08, -1.7462e-08, 1.3970e-09, ..., 0.0000e+00, + -2.7474e-08, 0.0000e+00], + [ 6.9849e-09, 5.2154e-08, -1.8533e-07, ..., 0.0000e+00, + 2.8638e-08, 0.0000e+00]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0080, 0.0173, -0.0153, 0.0173, 0.0255, -0.0203, -0.0165, -0.0136, + -0.0312, -0.0253], device='cuda:0'), grad: tensor([-8.1258e-08, -1.9632e-06, 1.9241e-06, -1.0105e-06, -6.4494e-08, + 4.8801e-07, 2.7474e-08, 9.4669e-07, 2.1653e-08, -2.7660e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 216.72, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4544 re_mapping 0.0040 re_causal 0.0127 /// teacc 98.99 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.2181, -0.1261, -0.0900, ..., -0.0041, -0.0510, -0.0125], + [-0.1132, 0.1042, -0.0794, ..., 0.0208, -0.0491, -0.0787], + [ 0.1079, -0.1481, -0.1587, ..., -0.0597, 0.1350, -0.0328], + ..., + [-0.1634, -0.0824, 0.1118, ..., 0.0393, -0.1950, -0.0012], + [ 0.0869, 0.0013, -0.2733, ..., -0.0657, 0.1463, -0.0338], + [-0.1951, -0.1504, 0.0323, ..., -0.0144, -0.1826, -0.0635]], + device='cuda:0'), grad: tensor([[ 1.0012e-08, 9.7789e-09, -2.7008e-08, ..., 0.0000e+00, + -1.2247e-06, 0.0000e+00], + [-1.7928e-08, -1.5218e-06, 2.0955e-09, ..., 0.0000e+00, + 9.5693e-08, 0.0000e+00], + [ 1.6647e-07, 1.0077e-06, -2.0489e-08, ..., 0.0000e+00, + 2.7986e-07, 0.0000e+00], + ..., + [ 3.8650e-08, 1.3225e-07, 1.6065e-08, ..., 0.0000e+00, + 2.3656e-07, 0.0000e+00], + [-5.2806e-07, 2.0000e-07, 1.3970e-09, ..., 0.0000e+00, + -5.9418e-07, 0.0000e+00], + [ 6.8219e-08, 3.9581e-08, 1.7928e-08, ..., 0.0000e+00, + 8.9221e-07, 0.0000e+00]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0079, 0.0166, -0.0152, 0.0173, 0.0252, -0.0203, -0.0166, -0.0136, + -0.0286, -0.0253], device='cuda:0'), grad: tensor([-1.1005e-05, -2.4643e-06, 3.6806e-06, -9.6392e-07, 4.9826e-07, + 9.5367e-07, 1.1194e-06, 1.0766e-06, -1.6065e-07, 7.2829e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 217.05, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4784 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.05 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.2184, -0.1264, -0.0900, ..., -0.0042, -0.0509, -0.0125], + [-0.1134, 0.1057, -0.0795, ..., 0.0206, -0.0490, -0.0787], + [ 0.1080, -0.1486, -0.1589, ..., -0.0603, 0.1351, -0.0328], + ..., + [-0.1634, -0.0829, 0.1118, ..., 0.0388, -0.1951, -0.0012], + [ 0.0872, 0.0014, -0.2736, ..., -0.0666, 0.1464, -0.0338], + [-0.1953, -0.1517, 0.0323, ..., -0.0146, -0.1829, -0.0635]], + device='cuda:0'), grad: tensor([[ 4.6799e-08, 6.7754e-08, 4.6566e-10, ..., 0.0000e+00, + 8.8476e-09, 0.0000e+00], + [ 1.4901e-08, -4.0303e-07, 1.9092e-08, ..., 0.0000e+00, + -1.0943e-08, 0.0000e+00], + [ 1.3970e-09, 2.7474e-08, 8.3819e-09, ..., 0.0000e+00, + 9.5461e-09, 0.0000e+00], + ..., + [ 2.5611e-09, 1.9348e-07, -5.1456e-08, ..., 0.0000e+00, + 1.0710e-08, 0.0000e+00], + [ 6.2771e-07, 4.5309e-07, 3.0268e-09, ..., 0.0000e+00, + 1.9325e-08, 0.0000e+00], + [ 1.0245e-08, 5.3318e-08, 1.1176e-08, ..., 0.0000e+00, + 3.4925e-09, 0.0000e+00]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0078, 0.0173, -0.0152, 0.0176, 0.0248, -0.0205, -0.0167, -0.0139, + -0.0286, -0.0252], device='cuda:0'), grad: tensor([ 1.3225e-07, -5.2340e-07, 1.0408e-07, 2.2887e-07, -1.1642e-08, + 1.1455e-06, -2.5406e-06, 1.4505e-07, 1.1241e-06, 1.8557e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 217.04, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4370 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.13 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.2187, -0.1265, -0.0900, ..., -0.0042, -0.0515, -0.0125], + [-0.1135, 0.1058, -0.0796, ..., 0.0206, -0.0492, -0.0787], + [ 0.1081, -0.1487, -0.1591, ..., -0.0603, 0.1351, -0.0328], + ..., + [-0.1635, -0.0829, 0.1119, ..., 0.0388, -0.1952, -0.0012], + [ 0.0873, 0.0014, -0.2737, ..., -0.0666, 0.1466, -0.0338], + [-0.1959, -0.1535, 0.0323, ..., -0.0147, -0.1834, -0.0635]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, 1.3271e-08, 0.0000e+00, ..., 0.0000e+00, + 2.5844e-08, 0.0000e+00], + [ 9.5461e-09, -6.9104e-07, 1.3970e-09, ..., 0.0000e+00, + 3.1618e-07, 0.0000e+00], + [-3.7020e-08, 1.1502e-07, 4.6566e-10, ..., 0.0000e+00, + -9.0338e-07, 0.0000e+00], + ..., + [ 1.6531e-08, 2.3050e-07, -4.6566e-09, ..., 0.0000e+00, + 1.8091e-07, 0.0000e+00], + [ 1.3970e-09, 7.0548e-08, 0.0000e+00, ..., 0.0000e+00, + 7.8697e-08, 0.0000e+00], + [ 2.0955e-09, 2.1653e-08, 2.3283e-09, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0077, 0.0173, -0.0152, 0.0183, 0.0249, -0.0207, -0.0168, -0.0139, + -0.0286, -0.0253], device='cuda:0'), grad: tensor([-4.6566e-10, -2.7311e-07, -1.8058e-06, 4.4634e-07, 3.6927e-07, + 5.5879e-09, 1.6158e-07, 6.9197e-07, 2.8848e-07, 1.4133e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 216.87, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4627 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.03 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.2195, -0.1271, -0.0900, ..., -0.0042, -0.0516, -0.0125], + [-0.1136, 0.1061, -0.0796, ..., 0.0206, -0.0492, -0.0787], + [ 0.1081, -0.1489, -0.1600, ..., -0.0612, 0.1352, -0.0328], + ..., + [-0.1635, -0.0831, 0.1120, ..., 0.0397, -0.1952, -0.0012], + [ 0.0875, 0.0014, -0.2742, ..., -0.0666, 0.1467, -0.0338], + [-0.1967, -0.1545, 0.0323, ..., -0.0147, -0.1842, -0.0635]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 2.0093e-07, 1.4738e-07, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00], + [-2.6310e-08, 1.3411e-05, 1.1623e-05, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [-1.3970e-09, 1.6792e-06, 1.2890e-06, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + ..., + [ 1.8626e-09, -2.5034e-05, -1.9580e-05, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-7.6834e-09, 7.3910e-06, 5.6587e-06, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + [ 1.1642e-09, 8.4564e-07, 6.8359e-07, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0077, 0.0174, -0.0153, 0.0178, 0.0249, -0.0204, -0.0166, -0.0139, + -0.0287, -0.0254], device='cuda:0'), grad: tensor([ 9.4622e-07, 7.0274e-05, 8.1211e-06, 1.3411e-07, 8.4797e-07, + 1.1898e-07, 2.1011e-06, -1.2231e-04, 3.5524e-05, 4.4033e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 216.95, cls_loss 0.0013 cls_loss_mapping 0.0026 cls_loss_causal 0.4584 re_mapping 0.0042 re_causal 0.0120 /// teacc 99.16 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.2200, -0.1270, -0.0900, ..., -0.0042, -0.0516, -0.0125], + [-0.1138, 0.1061, -0.0802, ..., 0.0206, -0.0493, -0.0787], + [ 0.1081, -0.1491, -0.1604, ..., -0.0612, 0.1352, -0.0328], + ..., + [-0.1635, -0.0828, 0.1128, ..., 0.0397, -0.1954, -0.0012], + [ 0.0879, 0.0014, -0.2767, ..., -0.0666, 0.1470, -0.0338], + [-0.1981, -0.1569, 0.0321, ..., -0.0147, -0.1854, -0.0635]], + device='cuda:0'), grad: tensor([[ 9.8487e-08, 9.8720e-08, 4.1910e-09, ..., 0.0000e+00, + 9.0804e-09, 0.0000e+00], + [ 1.6997e-08, 3.5902e-07, 1.0058e-07, ..., 0.0000e+00, + 1.7695e-08, 0.0000e+00], + [-1.1642e-07, 2.6077e-08, 4.6566e-09, ..., 0.0000e+00, + -6.7754e-08, 0.0000e+00], + ..., + [ 1.0058e-07, 2.3912e-07, 8.2888e-08, ..., 0.0000e+00, + 9.1735e-08, 0.0000e+00], + [ 9.6159e-08, 3.4226e-08, 6.7521e-09, ..., 0.0000e+00, + 1.8650e-07, 0.0000e+00], + [ 8.6147e-09, 4.7162e-06, 1.7211e-06, ..., 0.0000e+00, + 1.5367e-08, 0.0000e+00]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0076, 0.0172, -0.0153, 0.0205, 0.0251, -0.0224, -0.0168, -0.0136, + -0.0285, -0.0257], device='cuda:0'), grad: tensor([-5.5041e-07, 9.5926e-07, -1.7835e-07, -8.8522e-07, -1.5497e-05, + 3.6461e-07, 1.8766e-06, 9.6485e-07, 7.7300e-07, 1.2137e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 217.08, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4971 re_mapping 0.0043 re_causal 0.0130 /// teacc 99.09 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.2211, -0.1270, -0.0900, ..., -0.0042, -0.0519, -0.0125], + [-0.1138, 0.1070, -0.0803, ..., 0.0206, -0.0493, -0.0789], + [ 0.1081, -0.1493, -0.1607, ..., -0.0612, 0.1353, -0.0328], + ..., + [-0.1636, -0.0829, 0.1127, ..., 0.0397, -0.1954, -0.0013], + [ 0.0908, 0.0012, -0.2752, ..., -0.0666, 0.1481, -0.0338], + [-0.2016, -0.1595, 0.0319, ..., -0.0147, -0.1886, -0.0636]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 8.1491e-09, 0.0000e+00, ..., 0.0000e+00, + 5.6345e-08, 0.0000e+00], + [ 3.7253e-09, -7.9162e-07, 7.2177e-09, ..., 0.0000e+00, + 3.8184e-08, 0.0000e+00], + [-8.7544e-08, 3.1572e-07, 9.3132e-10, ..., 0.0000e+00, + -7.1013e-07, 0.0000e+00], + ..., + [ 4.8429e-08, 3.1898e-07, -9.5461e-09, ..., 0.0000e+00, + 5.1130e-07, 0.0000e+00], + [ 6.7521e-09, 3.4692e-08, 4.6566e-10, ..., 0.0000e+00, + 3.4692e-08, 0.0000e+00], + [ 4.4238e-09, 7.1479e-08, 4.1910e-09, ..., 0.0000e+00, + 4.8894e-09, 0.0000e+00]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0075, 0.0176, -0.0154, 0.0206, 0.0252, -0.0225, -0.0169, -0.0137, + -0.0266, -0.0266], device='cuda:0'), grad: tensor([ 1.4878e-07, -1.1194e-06, -1.1940e-06, 1.8789e-07, 2.5844e-08, + -6.3097e-08, 4.3074e-08, 1.7369e-06, 1.0850e-07, 1.3667e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 216.83, cls_loss 0.0011 cls_loss_mapping 0.0020 cls_loss_causal 0.4975 re_mapping 0.0044 re_causal 0.0132 /// teacc 99.12 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.2215, -0.1271, -0.0900, ..., -0.0042, -0.0489, -0.0125], + [-0.1139, 0.1075, -0.0802, ..., 0.0206, -0.0493, -0.0794], + [ 0.1082, -0.1494, -0.1616, ..., -0.0612, 0.1354, -0.0328], + ..., + [-0.1637, -0.0833, 0.1114, ..., 0.0397, -0.1955, -0.0013], + [ 0.0908, 0.0011, -0.2756, ..., -0.0666, 0.1482, -0.0338], + [-0.2016, -0.1603, 0.0319, ..., -0.0147, -0.1903, -0.0639]], + device='cuda:0'), grad: tensor([[ 2.6543e-08, 2.8871e-08, 2.5611e-08, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [ 1.8161e-08, 7.4878e-07, 6.7800e-07, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + [-7.0874e-07, 2.2817e-08, 2.0489e-08, ..., 0.0000e+00, + -1.1530e-06, 0.0000e+00], + ..., + [ 4.4703e-08, 1.2312e-06, 1.0841e-06, ..., 0.0000e+00, + 5.8208e-08, 0.0000e+00], + [ 6.3144e-07, 7.2410e-07, 6.3656e-07, ..., 0.0000e+00, + 9.6392e-07, 0.0000e+00], + [ 3.9721e-07, 2.2277e-06, 2.1178e-06, ..., 0.0000e+00, + 6.8452e-08, 0.0000e+00]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0070, 0.0176, -0.0153, 0.0205, 0.0261, -0.0224, -0.0169, -0.0141, + -0.0267, -0.0273], device='cuda:0'), grad: tensor([ 1.7835e-07, 3.4850e-06, -2.2054e-06, 3.0510e-06, -2.4334e-05, + -4.7646e-06, 7.9209e-07, 5.6364e-06, 5.3868e-06, 1.2778e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 216.73, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4744 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.2233, -0.1279, -0.0900, ..., -0.0042, -0.0488, -0.0126], + [-0.1141, 0.1075, -0.0805, ..., 0.0206, -0.0494, -0.0795], + [ 0.1083, -0.1495, -0.1635, ..., -0.0612, 0.1355, -0.0329], + ..., + [-0.1637, -0.0834, 0.1113, ..., 0.0397, -0.1955, -0.0014], + [ 0.0908, 0.0011, -0.2761, ..., -0.0666, 0.1483, -0.0339], + [-0.2017, -0.1610, 0.0318, ..., -0.0147, -0.1904, -0.0640]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 1.5832e-08, -3.2410e-07, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [ 3.2596e-09, -4.0513e-08, 2.5611e-08, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [-7.8697e-08, 3.7719e-08, 5.4482e-08, ..., 0.0000e+00, + -1.4948e-07, 0.0000e+00], + ..., + [ 7.9628e-08, 1.9558e-08, -8.8010e-08, ..., 0.0000e+00, + 1.2666e-07, 0.0000e+00], + [ 3.5390e-08, 1.3039e-08, 1.6298e-08, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.4995e-08, 2.6356e-07, 5.7928e-07, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0071, 0.0175, -0.0153, 0.0203, 0.0264, -0.0221, -0.0168, -0.0141, + -0.0268, -0.0275], device='cuda:0'), grad: tensor([-5.9158e-06, 2.2445e-07, 3.7346e-07, 1.8915e-06, -2.0415e-06, + 3.1712e-07, 2.3935e-07, 2.6403e-07, 2.7521e-07, 4.3735e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 216.69, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.5010 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.11 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.2239, -0.1279, -0.0900, ..., -0.0041, -0.0489, -0.0127], + [-0.1142, 0.1077, -0.0808, ..., 0.0205, -0.0495, -0.0796], + [ 0.1083, -0.1495, -0.1634, ..., -0.0612, 0.1357, -0.0329], + ..., + [-0.1638, -0.0837, 0.1115, ..., 0.0396, -0.1956, -0.0014], + [ 0.0904, 0.0007, -0.2763, ..., -0.0666, 0.1482, -0.0339], + [-0.2018, -0.1613, 0.0318, ..., -0.0147, -0.1909, -0.0649]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, -1.6484e-07, 6.9849e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 4.6566e-09, 1.3970e-09, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + ..., + [ 1.3970e-09, 1.1967e-07, -8.8476e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [-3.3993e-08, 1.5367e-08, 1.3970e-09, ..., 0.0000e+00, + -2.5611e-08, 0.0000e+00], + [ 2.6543e-08, 1.5041e-07, 1.4761e-07, ..., 0.0000e+00, + 2.0023e-08, 0.0000e+00]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0070, 0.0175, -0.0152, 0.0198, 0.0265, -0.0215, -0.0169, -0.0142, + -0.0271, -0.0276], device='cuda:0'), grad: tensor([-2.9802e-08, -1.9232e-07, 1.6298e-08, 1.2107e-08, -8.2236e-07, + -6.9849e-09, 4.6566e-08, 1.3877e-07, -6.8452e-08, 9.2154e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 216.41, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.4802 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.02 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.2258, -0.1287, -0.0900, ..., -0.0041, -0.0490, -0.0127], + [-0.1142, 0.1084, -0.0810, ..., 0.0205, -0.0495, -0.0796], + [ 0.1084, -0.1498, -0.1637, ..., -0.0612, 0.1360, -0.0329], + ..., + [-0.1638, -0.0842, 0.1120, ..., 0.0396, -0.1957, -0.0014], + [ 0.0909, 0.0005, -0.2766, ..., -0.0666, 0.1487, -0.0339], + [-0.2019, -0.1616, 0.0317, ..., -0.0147, -0.1915, -0.0649]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.3970e-08, 1.3970e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.3039e-08, 3.3937e-06, 4.0047e-08, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.0338e-07, 9.9186e-08, 7.4506e-09, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + ..., + [-2.3283e-09, -4.6045e-06, -9.2201e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 3.9814e-07, 7.0035e-07, 2.7940e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 1.8626e-09, 4.3027e-07, 8.9873e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0071, 0.0180, -0.0151, 0.0194, 0.0263, -0.0212, -0.0172, -0.0142, + -0.0269, -0.0276], device='cuda:0'), grad: tensor([-9.3132e-09, 5.3793e-05, 1.1399e-06, 3.8892e-06, -3.5577e-07, + 1.2200e-07, -1.0608e-06, -7.1943e-05, 9.0897e-06, 5.3495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 217.17, cls_loss 0.0011 cls_loss_mapping 0.0021 cls_loss_causal 0.4703 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.10 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.2273, -0.1293, -0.0904, ..., -0.0041, -0.0491, -0.0127], + [-0.1143, 0.1088, -0.0808, ..., 0.0205, -0.0494, -0.0796], + [ 0.1085, -0.1502, -0.1647, ..., -0.0612, 0.1360, -0.0330], + ..., + [-0.1639, -0.0848, 0.1119, ..., 0.0396, -0.1958, -0.0014], + [ 0.0881, -0.0023, -0.2769, ..., -0.0666, 0.1482, -0.0340], + [-0.2020, -0.1622, 0.0318, ..., -0.0147, -0.1916, -0.0649]], + device='cuda:0'), grad: tensor([[4.6566e-10, 7.4506e-09, 3.2596e-09, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 2.9150e-07, 4.9826e-08, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 3.3528e-08, 1.2573e-08, ..., 0.0000e+00, 4.6566e-10, + 0.0000e+00], + ..., + [0.0000e+00, 4.4238e-08, 3.6554e-07, ..., 0.0000e+00, 4.6566e-10, + 0.0000e+00], + [2.7940e-09, 7.9162e-09, 6.9849e-09, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [9.3132e-10, 3.5716e-07, 1.2806e-06, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0072, 0.0181, -0.0153, 0.0193, 0.0271, -0.0209, -0.0169, -0.0142, + -0.0294, -0.0277], device='cuda:0'), grad: tensor([-2.1886e-07, 7.4133e-07, 1.1036e-07, 9.3132e-09, -7.8231e-06, + 2.3283e-09, 1.3364e-07, 9.6112e-07, 5.5414e-08, 6.0350e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 217.03, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.5055 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.06 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.2280, -0.1294, -0.0904, ..., -0.0041, -0.0491, -0.0128], + [-0.1155, 0.1101, -0.0792, ..., 0.0205, -0.0503, -0.0797], + [ 0.1092, -0.1511, -0.1655, ..., -0.0612, 0.1367, -0.0330], + ..., + [-0.1639, -0.0864, 0.1119, ..., 0.0396, -0.1959, -0.0014], + [ 0.0880, -0.0024, -0.2772, ..., -0.0666, 0.1481, -0.0340], + [-0.2020, -0.1632, 0.0314, ..., -0.0147, -0.1917, -0.0651]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 9.3132e-10, 1.3970e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 9.3132e-10, -6.1002e-08, 2.2352e-08, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + [-1.3970e-09, 2.0023e-08, 2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + ..., + [ 3.2596e-09, 2.2352e-08, -2.6543e-08, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 2.2072e-07, 8.3819e-09, 4.1910e-09, ..., 0.0000e+00, + 8.3819e-08, 0.0000e+00], + [ 5.7276e-08, 3.7253e-09, -1.6252e-07, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0071, 0.0184, -0.0150, 0.0192, 0.0274, -0.0210, -0.0170, -0.0144, + -0.0297, -0.0280], device='cuda:0'), grad: tensor([-7.5903e-08, -3.7253e-09, 4.2841e-08, 1.4352e-06, 7.1619e-07, + -2.1718e-06, 4.2375e-08, -2.6543e-08, 5.9931e-07, -5.5553e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 216.91, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.5108 re_mapping 0.0039 re_causal 0.0125 /// teacc 99.09 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.2291, -0.1307, -0.0904, ..., -0.0041, -0.0494, -0.0128], + [-0.1157, 0.1104, -0.0795, ..., 0.0205, -0.0504, -0.0797], + [ 0.1095, -0.1514, -0.1655, ..., -0.0612, 0.1371, -0.0330], + ..., + [-0.1642, -0.0866, 0.1128, ..., 0.0396, -0.1962, -0.0014], + [ 0.0879, -0.0025, -0.2775, ..., -0.0666, 0.1481, -0.0340], + [-0.2021, -0.1642, 0.0311, ..., -0.0147, -0.1920, -0.0651]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 5.1223e-09, 9.3132e-10, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 5.5879e-09, 9.3132e-09, 1.2759e-07, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 4.1910e-09, 1.0245e-08, 3.2596e-09, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + ..., + [ 3.2596e-09, 9.7789e-09, -1.3597e-07, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [-4.1910e-09, 1.9558e-08, 6.0536e-09, ..., 0.0000e+00, + -5.3085e-08, 0.0000e+00], + [ 2.4214e-08, 2.0070e-07, 5.3085e-08, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0075, 0.0185, -0.0147, 0.0189, 0.0275, -0.0207, -0.0165, -0.0145, + -0.0299, -0.0283], device='cuda:0'), grad: tensor([-9.4995e-08, 4.5169e-07, 4.8429e-08, 2.6543e-08, -5.6205e-07, + 1.4435e-07, -6.7055e-08, -4.2561e-07, -8.3819e-08, 5.6438e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 216.56, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4607 re_mapping 0.0038 re_causal 0.0120 /// teacc 99.02 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.2314, -0.1317, -0.0904, ..., -0.0040, -0.0491, -0.0128], + [-0.1156, 0.1107, -0.0798, ..., 0.0205, -0.0505, -0.0797], + [ 0.1102, -0.1516, -0.1657, ..., -0.0614, 0.1374, -0.0330], + ..., + [-0.1648, -0.0869, 0.1130, ..., 0.0398, -0.1965, -0.0014], + [ 0.0874, -0.0029, -0.2777, ..., -0.0667, 0.1482, -0.0341], + [-0.2022, -0.1644, 0.0311, ..., -0.0149, -0.1921, -0.0651]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 4.1910e-08, -9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.1910e-09, -1.4016e-07, 2.0955e-08, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [-5.1223e-09, 2.9802e-07, 2.7940e-09, ..., 0.0000e+00, + -6.9849e-09, 0.0000e+00], + ..., + [ 3.2596e-09, 5.6811e-08, 1.1176e-08, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00], + [-1.1176e-08, 1.3970e-08, 1.4435e-08, ..., 0.0000e+00, + -1.8161e-08, 0.0000e+00], + [ 7.9162e-09, 2.4494e-07, 4.6566e-10, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0075, 0.0187, -0.0143, 0.0192, 0.0274, -0.0204, -0.0163, -0.0153, + -0.0303, -0.0282], device='cuda:0'), grad: tensor([-1.1222e-07, -7.5437e-08, 9.3225e-07, 4.5169e-08, -2.0266e-06, + 9.3132e-09, 5.0012e-07, 1.5320e-07, 6.6590e-08, 5.2061e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 216.82, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.4743 re_mapping 0.0038 re_causal 0.0116 /// teacc 98.99 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.2350, -0.1337, -0.0905, ..., -0.0038, -0.0493, -0.0128], + [-0.1159, 0.1108, -0.0805, ..., 0.0205, -0.0506, -0.0797], + [ 0.1104, -0.1519, -0.1653, ..., -0.0614, 0.1376, -0.0330], + ..., + [-0.1649, -0.0869, 0.1126, ..., 0.0398, -0.1965, -0.0014], + [ 0.0878, -0.0028, -0.2782, ..., -0.0667, 0.1483, -0.0341], + [-0.2024, -0.1645, 0.0314, ..., -0.0150, -0.1924, -0.0651]], + device='cuda:0'), grad: tensor([[ 1.1157e-06, 2.4419e-06, 4.6566e-10, ..., 0.0000e+00, + 2.5611e-08, 0.0000e+00], + [ 1.7695e-08, -1.3784e-06, 4.1910e-09, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 5.3085e-08, 1.2713e-07, 9.3132e-10, ..., 0.0000e+00, + 3.8650e-08, 0.0000e+00], + ..., + [ 1.8626e-08, 8.0001e-07, -2.7940e-09, ..., 0.0000e+00, + 1.5367e-08, 0.0000e+00], + [-3.1199e-07, 1.2619e-07, 6.5193e-09, ..., 0.0000e+00, + -4.5355e-07, 0.0000e+00], + [ 2.9756e-07, 8.8662e-07, 2.9337e-08, ..., 0.0000e+00, + 1.9092e-08, 0.0000e+00]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0076, 0.0184, -0.0142, 0.0168, 0.0273, -0.0180, -0.0162, -0.0156, + -0.0299, -0.0282], device='cuda:0'), grad: tensor([ 1.0684e-05, -2.7958e-06, 4.4191e-07, 1.1446e-06, -4.0792e-06, + -1.0198e-06, -1.0811e-05, 1.9446e-06, -3.1944e-07, 4.8056e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 216.73, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4882 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.08 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.2362, -0.1338, -0.0906, ..., -0.0038, -0.0495, -0.0128], + [-0.1168, 0.1108, -0.0807, ..., 0.0205, -0.0509, -0.0797], + [ 0.1106, -0.1522, -0.1643, ..., -0.0614, 0.1378, -0.0330], + ..., + [-0.1654, -0.0870, 0.1129, ..., 0.0398, -0.1968, -0.0015], + [ 0.0885, -0.0025, -0.2787, ..., -0.0667, 0.1490, -0.0342], + [-0.2025, -0.1650, 0.0313, ..., -0.0151, -0.1927, -0.0651]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.1910e-09, 1.8626e-09, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00], + [ 3.7253e-09, -2.1374e-07, 1.1502e-07, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [-6.0536e-09, 1.4901e-08, 6.0070e-08, ..., 0.0000e+00, + -5.5879e-08, 0.0000e+00], + ..., + [ 3.2922e-07, 1.6857e-07, -2.5053e-07, ..., 0.0000e+00, + 1.5786e-07, 0.0000e+00], + [-3.6880e-07, 5.5879e-08, 2.5611e-08, ..., 0.0000e+00, + -1.6391e-07, 0.0000e+00], + [ 1.1176e-08, 1.2666e-07, 5.8208e-08, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0075, 0.0181, -0.0140, 0.0159, 0.0274, -0.0171, -0.0164, -0.0158, + -0.0293, -0.0284], device='cuda:0'), grad: tensor([-1.3970e-08, 9.5926e-08, 9.9652e-08, 7.1246e-08, -7.6694e-07, + 2.5146e-08, 4.2375e-08, 1.8114e-07, -5.4110e-07, 8.2655e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 216.91, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.5034 re_mapping 0.0038 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.2371, -0.1339, -0.0906, ..., -0.0038, -0.0496, -0.0128], + [-0.1169, 0.1111, -0.0809, ..., 0.0205, -0.0510, -0.0797], + [ 0.1109, -0.1524, -0.1649, ..., -0.0614, 0.1381, -0.0330], + ..., + [-0.1656, -0.0871, 0.1135, ..., 0.0398, -0.1970, -0.0015], + [ 0.0883, -0.0026, -0.2792, ..., -0.0667, 0.1488, -0.0343], + [-0.2027, -0.1656, 0.0312, ..., -0.0151, -0.1930, -0.0651]], + device='cuda:0'), grad: tensor([[ 3.1199e-08, 3.8464e-07, 4.6566e-10, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 3.2596e-08, -8.9407e-06, 5.1223e-09, ..., 0.0000e+00, + 1.7229e-08, 0.0000e+00], + [ 3.7253e-09, 9.9652e-08, 4.6566e-10, ..., 0.0000e+00, + -4.7963e-08, 0.0000e+00], + ..., + [ 6.1467e-08, 2.8871e-06, -3.2596e-09, ..., 0.0000e+00, + 2.2817e-08, 0.0000e+00], + [ 8.4750e-08, 4.7451e-07, 7.9162e-09, ..., 0.0000e+00, + -5.2620e-08, 0.0000e+00], + [ 3.0361e-07, 2.8824e-07, -3.2596e-08, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0073, 0.0182, -0.0139, 0.0150, 0.0274, -0.0161, -0.0165, -0.0157, + -0.0298, -0.0287], device='cuda:0'), grad: tensor([-6.7540e-06, -1.6063e-05, 9.4529e-08, 3.5614e-05, 1.0356e-06, + -4.0084e-05, 1.7121e-05, 5.5395e-06, 1.2666e-06, 2.1458e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 216.58, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4807 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.14 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.2373, -0.1338, -0.0906, ..., -0.0038, -0.0497, -0.0128], + [-0.1170, 0.1121, -0.0801, ..., 0.0205, -0.0511, -0.0797], + [ 0.1109, -0.1529, -0.1654, ..., -0.0614, 0.1382, -0.0330], + ..., + [-0.1656, -0.0883, 0.1133, ..., 0.0398, -0.1971, -0.0015], + [ 0.0884, -0.0027, -0.2795, ..., -0.0667, 0.1488, -0.0344], + [-0.2027, -0.1662, 0.0311, ..., -0.0151, -0.1930, -0.0651]], + device='cuda:0'), grad: tensor([[ 2.3283e-08, 6.3330e-08, 9.3132e-10, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 8.1025e-08, -3.5483e-07, 1.0245e-08, ..., 0.0000e+00, + 1.6112e-07, 0.0000e+00], + [-1.8068e-07, 3.4459e-08, 1.3970e-09, ..., 0.0000e+00, + -3.7579e-07, 0.0000e+00], + ..., + [ 2.8871e-08, 1.2992e-07, -1.3504e-08, ..., 0.0000e+00, + 5.5414e-08, 0.0000e+00], + [ 1.6158e-07, 4.4703e-08, 3.7253e-09, ..., 0.0000e+00, + 1.3923e-07, 0.0000e+00], + [-9.7789e-08, 4.1910e-08, -1.1176e-08, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0071, 0.0186, -0.0141, 0.0150, 0.0274, -0.0161, -0.0170, -0.0158, + -0.0299, -0.0288], device='cuda:0'), grad: tensor([-2.1560e-07, -6.9849e-09, -9.5554e-07, 1.4994e-07, 2.4633e-07, + -3.2596e-09, -7.6834e-08, 3.5297e-07, 2.0526e-06, -1.5562e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 216.90, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4792 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.02 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.2383, -0.1340, -0.0905, ..., -0.0038, -0.0499, -0.0128], + [-0.1148, 0.1140, -0.0805, ..., 0.0205, -0.0510, -0.0798], + [ 0.1099, -0.1555, -0.1658, ..., -0.0614, 0.1384, -0.0330], + ..., + [-0.1657, -0.0886, 0.1137, ..., 0.0398, -0.1972, -0.0015], + [ 0.0884, -0.0034, -0.2800, ..., -0.0667, 0.1489, -0.0345], + [-0.2030, -0.1667, 0.0310, ..., -0.0151, -0.1935, -0.0651]], + device='cuda:0'), grad: tensor([[ 1.7229e-08, 4.2841e-08, 4.1910e-09, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [ 9.7789e-09, -7.5903e-08, 1.0710e-08, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + [-1.4901e-08, 1.1176e-08, -1.8626e-08, ..., 0.0000e+00, + -6.9384e-08, 0.0000e+00], + ..., + [ 5.1223e-09, 1.7695e-08, 3.4925e-08, ..., 0.0000e+00, + 5.8673e-08, 0.0000e+00], + [-9.3132e-09, 3.8184e-08, 2.7940e-09, ..., 0.0000e+00, + 1.8207e-07, 0.0000e+00], + [ 2.7940e-09, 3.4459e-08, -2.3469e-07, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0066, 0.0198, -0.0149, 0.0150, 0.0274, -0.0160, -0.0177, -0.0157, + -0.0305, -0.0290], device='cuda:0'), grad: tensor([-1.0453e-05, 2.8079e-07, 2.4009e-06, -2.0117e-07, 5.8189e-06, + 7.9023e-07, -1.8934e-06, 1.0841e-06, 1.1157e-06, 1.0543e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 216.79, cls_loss 0.0010 cls_loss_mapping 0.0016 cls_loss_causal 0.4863 re_mapping 0.0039 re_causal 0.0123 /// teacc 99.11 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.2388, -0.1340, -0.0905, ..., -0.0038, -0.0500, -0.0128], + [-0.1148, 0.1139, -0.0806, ..., 0.0205, -0.0512, -0.0798], + [ 0.1101, -0.1555, -0.1660, ..., -0.0614, 0.1388, -0.0330], + ..., + [-0.1658, -0.0882, 0.1138, ..., 0.0398, -0.1973, -0.0015], + [ 0.0884, -0.0034, -0.2801, ..., -0.0667, 0.1489, -0.0345], + [-0.2030, -0.1665, 0.0310, ..., -0.0151, -0.1937, -0.0651]], + device='cuda:0'), grad: tensor([[ 1.1642e-08, 7.9162e-09, -5.9139e-08, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 3.4925e-08, -3.8650e-08, 6.6590e-08, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00], + [ 9.3132e-09, 5.5879e-09, 5.7742e-08, ..., 0.0000e+00, + -5.3551e-08, 0.0000e+00], + ..., + [ 2.0955e-08, 2.1886e-08, -5.1921e-07, ..., 0.0000e+00, + 2.7474e-08, 0.0000e+00], + [-1.6717e-07, 6.9849e-09, 1.0245e-08, ..., 0.0000e+00, + -1.2200e-07, 0.0000e+00], + [ 1.8626e-09, 2.3050e-07, 7.1572e-07, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0064, 0.0194, -0.0148, 0.0150, 0.0272, -0.0161, -0.0177, -0.0154, + -0.0306, -0.0289], device='cuda:0'), grad: tensor([-2.0444e-05, 1.1064e-06, 1.3793e-06, 1.9427e-06, -5.4110e-07, + 2.5220e-06, 1.5190e-06, -5.5321e-07, 8.3353e-08, 1.2994e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 216.86, cls_loss 0.0014 cls_loss_mapping 0.0026 cls_loss_causal 0.4885 re_mapping 0.0042 re_causal 0.0118 /// teacc 99.07 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.2414, -0.1345, -0.0904, ..., -0.0037, -0.0502, -0.0128], + [-0.1150, 0.1140, -0.0808, ..., 0.0193, -0.0513, -0.0798], + [ 0.1091, -0.1556, -0.1657, ..., -0.0621, 0.1390, -0.0330], + ..., + [-0.1645, -0.0882, 0.1141, ..., 0.0392, -0.1975, -0.0015], + [ 0.0884, -0.0035, -0.2806, ..., -0.0672, 0.1488, -0.0345], + [-0.2031, -0.1666, 0.0309, ..., -0.0153, -0.1940, -0.0652]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, 1.3970e-09, 2.7940e-08, ..., 0.0000e+00, + 1.3597e-07, 0.0000e+00], + [ 5.5879e-09, -7.4506e-09, 1.3504e-08, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [-9.9186e-08, 9.3132e-10, -5.2154e-08, ..., 0.0000e+00, + -3.5996e-07, 0.0000e+00], + ..., + [ 1.7229e-08, 9.7789e-09, -3.1665e-08, ..., 0.0000e+00, + 3.8184e-08, 0.0000e+00], + [ 2.3749e-08, 5.5879e-09, 6.0536e-09, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00], + [ 1.1642e-08, 1.1921e-07, 1.0245e-08, ..., 0.0000e+00, + 4.3772e-08, 0.0000e+00]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0057, 0.0192, -0.0175, 0.0150, 0.0269, -0.0161, -0.0170, -0.0127, + -0.0311, -0.0291], device='cuda:0'), grad: tensor([-3.1292e-07, 1.4482e-07, -5.0664e-07, 9.0338e-08, -3.9302e-07, + -1.8626e-09, 7.4040e-08, -8.1491e-08, 1.2852e-07, 8.7498e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 216.73, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.5063 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.07 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.2428, -0.1347, -0.0905, ..., -0.0037, -0.0503, -0.0128], + [-0.1151, 0.1142, -0.0813, ..., 0.0193, -0.0516, -0.0798], + [ 0.1093, -0.1558, -0.1659, ..., -0.0621, 0.1395, -0.0330], + ..., + [-0.1646, -0.0889, 0.1145, ..., 0.0392, -0.1977, -0.0015], + [ 0.0884, -0.0036, -0.2814, ..., -0.0672, 0.1487, -0.0345], + [-0.2032, -0.1676, 0.0306, ..., -0.0153, -0.1942, -0.0652]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, -2.9337e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 4.6566e-09, -1.9558e-08, 4.1910e-09, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00], + [-1.8161e-08, 4.1910e-09, -1.3970e-09, ..., 0.0000e+00, + -2.3283e-08, 0.0000e+00], + ..., + [ 5.1223e-09, 1.6298e-08, -4.6566e-09, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [-4.1910e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + [ 1.3970e-09, 4.7032e-08, 8.8476e-09, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0056, 0.0192, -0.0174, 0.0150, 0.0274, -0.0160, -0.0178, -0.0128, + -0.0314, -0.0293], device='cuda:0'), grad: tensor([-3.7206e-07, 5.4948e-08, -2.7940e-08, 9.3132e-10, -1.5274e-07, + 1.7695e-08, 2.1374e-07, -3.0734e-08, 9.7789e-09, 2.8452e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 216.92, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.4709 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.07 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.2452, -0.1352, -0.0905, ..., -0.0037, -0.0507, -0.0128], + [-0.1153, 0.1145, -0.0816, ..., 0.0200, -0.0517, -0.0798], + [ 0.1095, -0.1559, -0.1664, ..., -0.0623, 0.1397, -0.0330], + ..., + [-0.1646, -0.0888, 0.1156, ..., 0.0389, -0.1978, -0.0015], + [ 0.0881, -0.0042, -0.2816, ..., -0.0677, 0.1485, -0.0345], + [-0.2029, -0.1683, 0.0305, ..., -0.0157, -0.1940, -0.0652]], + device='cuda:0'), grad: tensor([[ 1.3225e-07, 5.2759e-07, 1.2573e-08, ..., 0.0000e+00, + 5.7276e-08, 0.0000e+00], + [ 5.1223e-09, -1.4631e-06, 2.7940e-09, ..., 0.0000e+00, + -1.2806e-07, 0.0000e+00], + [ 2.7940e-09, 4.1444e-08, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 4.2655e-07, 4.6566e-08, ..., 0.0000e+00, + 2.1886e-08, 0.0000e+00], + [-3.2596e-09, 1.3830e-07, 1.8626e-09, ..., 0.0000e+00, + 5.1223e-09, 0.0000e+00], + [ 4.5635e-08, 1.4016e-07, -9.6858e-08, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0050, 0.0192, -0.0175, 0.0130, 0.0274, -0.0141, -0.0175, -0.0126, + -0.0332, -0.0291], device='cuda:0'), grad: tensor([ 1.4510e-06, -3.1367e-06, 1.0105e-07, 1.4622e-07, 1.9139e-07, + 9.1717e-06, -9.2164e-06, 1.2470e-06, 3.1013e-07, -2.5658e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 217.00, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4673 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.08 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.2462, -0.1355, -0.0905, ..., -0.0037, -0.0509, -0.0128], + [-0.1154, 0.1151, -0.0808, ..., 0.0183, -0.0516, -0.0799], + [ 0.1096, -0.1560, -0.1671, ..., -0.0633, 0.1399, -0.0330], + ..., + [-0.1647, -0.0896, 0.1154, ..., 0.0380, -0.1980, -0.0015], + [ 0.0879, -0.0043, -0.2817, ..., -0.0680, 0.1484, -0.0345], + [-0.2025, -0.1686, 0.0305, ..., -0.0159, -0.1936, -0.0652]], + device='cuda:0'), grad: tensor([[ 6.2399e-08, 2.7940e-09, 1.3039e-08, ..., 0.0000e+00, + 7.7300e-08, 0.0000e+00], + [ 6.5193e-08, 1.8626e-09, 1.8626e-08, ..., 0.0000e+00, + 6.9849e-08, 0.0000e+00], + [-2.2445e-07, 1.8626e-09, -4.2841e-08, ..., 0.0000e+00, + -3.0175e-07, 0.0000e+00], + ..., + [ 5.9605e-08, 9.3132e-09, -2.1420e-08, ..., 0.0000e+00, + 7.9162e-08, 0.0000e+00], + [-2.9504e-06, -5.8301e-07, 6.5193e-09, ..., 0.0000e+00, + -1.3430e-06, 0.0000e+00], + [ 5.6811e-08, 4.3772e-08, 3.3528e-08, ..., 0.0000e+00, + 6.2399e-08, 0.0000e+00]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0050, 0.0195, -0.0174, 0.0126, 0.0273, -0.0137, -0.0175, -0.0127, + -0.0345, -0.0286], device='cuda:0'), grad: tensor([ 2.1700e-07, 3.6508e-07, -9.4343e-07, -3.2596e-08, -7.8231e-08, + 8.0746e-07, 4.7944e-06, 3.5390e-08, -5.5432e-06, 3.7160e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 216.72, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.4965 re_mapping 0.0041 re_causal 0.0124 /// teacc 99.06 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.2472, -0.1359, -0.0907, ..., -0.0037, -0.0508, -0.0128], + [-0.1156, 0.1151, -0.0809, ..., 0.0183, -0.0517, -0.0799], + [ 0.1095, -0.1563, -0.1673, ..., -0.0634, 0.1397, -0.0330], + ..., + [-0.1649, -0.0897, 0.1156, ..., 0.0380, -0.1989, -0.0015], + [ 0.0884, -0.0043, -0.2819, ..., -0.0681, 0.1492, -0.0345], + [-0.2026, -0.1697, 0.0304, ..., -0.0159, -0.1939, -0.0653]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -2.9616e-07, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 1.8626e-09, 3.7253e-08, 1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-3.7253e-08, 1.6764e-08, -1.8626e-08, ..., 0.0000e+00, + -1.6205e-07, 0.0000e+00], + ..., + [ 3.7253e-09, 5.5879e-09, -3.7253e-09, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 7.4506e-09, 4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 7.4506e-09, 5.0291e-08, 1.8626e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0056, 0.0192, -0.0175, 0.0126, 0.0277, -0.0137, -0.0175, -0.0128, + -0.0342, -0.0283], device='cuda:0'), grad: tensor([-2.6226e-06, 1.3039e-07, -2.1607e-07, 2.6636e-07, -3.7160e-06, + -5.4017e-08, 5.5358e-06, 2.7940e-08, 3.5390e-07, 2.7008e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 216.88, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.5013 re_mapping 0.0038 re_causal 0.0120 /// teacc 98.92 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.2487, -0.1362, -0.0907, ..., -0.0037, -0.0511, -0.0128], + [-0.1157, 0.1151, -0.0810, ..., 0.0183, -0.0518, -0.0799], + [ 0.1096, -0.1564, -0.1672, ..., -0.0634, 0.1399, -0.0330], + ..., + [-0.1650, -0.0899, 0.1158, ..., 0.0380, -0.1991, -0.0015], + [ 0.0884, -0.0044, -0.2822, ..., -0.0681, 0.1493, -0.0346], + [-0.2026, -0.1670, 0.0327, ..., -0.0159, -0.1942, -0.0653]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.1176e-08, 1.8626e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 3.7253e-09, -7.6368e-08, 3.7253e-09, ..., 0.0000e+00, + -1.6764e-08, 0.0000e+00], + [-1.6764e-08, 1.6764e-08, 1.8626e-09, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 2.2352e-08, 6.7055e-08, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + [-1.1176e-08, 2.6077e-08, 1.8626e-08, ..., 0.0000e+00, + 3.3528e-08, 0.0000e+00], + [ 1.8626e-09, 2.6077e-08, -1.6391e-07, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0056, 0.0191, -0.0175, 0.0126, 0.0250, -0.0137, -0.0175, -0.0128, + -0.0343, -0.0258], device='cuda:0'), grad: tensor([ 4.0978e-07, 2.1793e-07, 2.9802e-08, 2.9244e-07, 2.3786e-06, + 2.7027e-06, 8.5682e-08, 6.3330e-06, 1.7397e-06, -1.4223e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 216.67, cls_loss 0.0015 cls_loss_mapping 0.0018 cls_loss_causal 0.5114 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.2515, -0.1389, -0.0908, ..., -0.0037, -0.0523, -0.0128], + [-0.1159, 0.1173, -0.0810, ..., 0.0183, -0.0490, -0.0799], + [ 0.1096, -0.1567, -0.1675, ..., -0.0634, 0.1404, -0.0330], + ..., + [-0.1651, -0.0903, 0.1157, ..., 0.0380, -0.1993, -0.0016], + [ 0.0888, -0.0066, -0.2829, ..., -0.0681, 0.1471, -0.0346], + [-0.2027, -0.1670, 0.0328, ..., -0.0159, -0.1953, -0.0656]], + device='cuda:0'), grad: tensor([[ 5.1409e-07, 4.2841e-06, 0.0000e+00, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 1.1176e-08, -2.6450e-07, 5.5879e-09, ..., 0.0000e+00, + -7.2643e-08, 0.0000e+00], + [ 0.0000e+00, 2.9802e-08, 1.8626e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 5.7742e-08, -2.6077e-08, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [-2.0489e-08, 9.1270e-08, 0.0000e+00, ..., 0.0000e+00, + -9.6858e-08, 0.0000e+00], + [ 1.1176e-08, 7.6368e-08, 2.7940e-08, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0058, 0.0209, -0.0175, 0.0126, 0.0250, -0.0137, -0.0174, -0.0129, + -0.0359, -0.0258], device='cuda:0'), grad: tensor([ 1.6376e-05, -3.9861e-07, 1.0990e-07, -5.4017e-08, 4.6566e-07, + 8.7172e-07, -1.7941e-05, 9.1270e-08, -9.3132e-09, 4.7125e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 216.84, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4667 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.05 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.2533, -0.1395, -0.0908, ..., -0.0037, -0.0528, -0.0138], + [-0.1161, 0.1176, -0.0818, ..., 0.0183, -0.0491, -0.0801], + [ 0.1095, -0.1568, -0.1704, ..., -0.0634, 0.1413, -0.0331], + ..., + [-0.1651, -0.0904, 0.1172, ..., 0.0380, -0.1994, -0.0016], + [ 0.0891, -0.0067, -0.2832, ..., -0.0681, 0.1476, -0.0346], + [-0.2029, -0.1672, 0.0324, ..., -0.0160, -0.1957, -0.0671]], + device='cuda:0'), grad: tensor([[ 3.5390e-08, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [ 9.3877e-07, 1.5274e-07, 1.4901e-08, ..., 0.0000e+00, + 8.1584e-07, 0.0000e+00], + [-2.1514e-06, -3.7812e-07, 1.8626e-09, ..., 0.0000e+00, + -1.8813e-06, 0.0000e+00], + ..., + [ 1.9930e-07, 7.4506e-08, 1.3039e-08, ..., 0.0000e+00, + 1.8999e-07, 0.0000e+00], + [ 4.2841e-08, 1.3039e-08, 5.5879e-09, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 2.4214e-08, 3.5018e-07, 2.4028e-07, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0057, 0.0207, -0.0174, 0.0126, 0.0252, -0.0137, -0.0172, -0.0126, + -0.0358, -0.0261], device='cuda:0'), grad: tensor([ 1.1176e-08, 2.9448e-06, -6.5453e-06, 5.6066e-07, -7.6182e-07, + -2.8498e-07, 8.1770e-07, 9.0711e-07, 2.1607e-07, 2.1234e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 216.84, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5041 re_mapping 0.0043 re_causal 0.0119 /// teacc 99.09 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.2557, -0.1400, -0.0908, ..., -0.0037, -0.0531, -0.0140], + [-0.1162, 0.1186, -0.0821, ..., 0.0183, -0.0490, -0.0803], + [ 0.1112, -0.1571, -0.1678, ..., -0.0634, 0.1442, -0.0331], + ..., + [-0.1660, -0.0906, 0.1184, ..., 0.0379, -0.2006, -0.0017], + [ 0.0891, -0.0073, -0.2840, ..., -0.0681, 0.1474, -0.0347], + [-0.2029, -0.1678, 0.0312, ..., -0.0160, -0.1961, -0.0681]], + device='cuda:0'), grad: tensor([[ 9.4995e-08, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 7.0781e-08, 0.0000e+00], + [ 1.9558e-07, 3.5390e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1362e-07, 0.0000e+00], + [ 6.3330e-08, 1.8626e-08, 0.0000e+00, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00], + ..., + [ 1.1548e-07, 4.6566e-08, 3.7253e-09, ..., 0.0000e+00, + 7.4506e-08, 0.0000e+00], + [-7.8045e-07, -1.3597e-07, 0.0000e+00, ..., 0.0000e+00, + -5.0664e-07, 0.0000e+00], + [ 1.4901e-08, 3.9116e-08, 2.2352e-08, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0056, 0.0211, -0.0166, 0.0125, 0.0263, -0.0139, -0.0155, -0.0125, + -0.0362, -0.0273], device='cuda:0'), grad: tensor([ 2.0675e-07, 9.4809e-07, 3.1851e-07, -6.2212e-07, 9.8720e-08, + 9.6485e-07, 1.9185e-07, 6.0163e-07, -2.9933e-06, 2.7195e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 216.81, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4823 re_mapping 0.0040 re_causal 0.0121 /// teacc 98.94 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.2576, -0.1405, -0.0908, ..., -0.0037, -0.0535, -0.0141], + [-0.1164, 0.1189, -0.0824, ..., 0.0183, -0.0492, -0.0804], + [ 0.1113, -0.1572, -0.1677, ..., -0.0634, 0.1444, -0.0331], + ..., + [-0.1661, -0.0909, 0.1183, ..., 0.0379, -0.2008, -0.0017], + [ 0.0892, -0.0073, -0.2844, ..., -0.0681, 0.1475, -0.0347], + [-0.2030, -0.1678, 0.0312, ..., -0.0160, -0.1962, -0.0684]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 5.7742e-08, -2.9802e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, -2.1979e-07, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-5.5879e-09, 2.0489e-08, 1.8626e-08, ..., 0.0000e+00, + -2.4214e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 1.4529e-07, -1.8626e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-1.8626e-08, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -2.6077e-08, 0.0000e+00], + [ 5.5879e-09, 4.0978e-08, 1.1176e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0056, 0.0211, -0.0165, 0.0126, 0.0263, -0.0140, -0.0154, -0.0126, + -0.0362, -0.0273], device='cuda:0'), grad: tensor([-3.9116e-08, -2.9802e-07, 8.0094e-08, 4.2282e-07, 1.8626e-08, + -1.3039e-08, -1.2480e-07, -1.1735e-07, -2.4214e-08, 8.1956e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 216.48, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4840 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.05 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.2592, -0.1437, -0.0909, ..., -0.0037, -0.0541, -0.0141], + [-0.1177, 0.1202, -0.0815, ..., 0.0183, -0.0506, -0.0804], + [ 0.1110, -0.1574, -0.1678, ..., -0.0634, 0.1442, -0.0331], + ..., + [-0.1658, -0.0915, 0.1182, ..., 0.0379, -0.2001, -0.0017], + [ 0.0900, -0.0072, -0.2851, ..., -0.0681, 0.1495, -0.0347], + [-0.2031, -0.1680, 0.0312, ..., -0.0160, -0.1966, -0.0684]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + [ 9.3132e-09, 7.2643e-08, 4.0978e-08, ..., 0.0000e+00, + 6.3330e-08, 0.0000e+00], + [-7.4506e-09, 3.7253e-09, -5.5879e-08, ..., 0.0000e+00, + -2.6450e-07, 0.0000e+00], + ..., + [ 9.3132e-09, 5.2154e-08, 3.9116e-08, ..., 0.0000e+00, + 2.1420e-07, 0.0000e+00], + [ 9.3132e-09, -2.4214e-08, 1.8626e-09, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00], + [ 3.5390e-08, 4.6194e-07, 1.2852e-07, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0059, 0.0215, -0.0166, 0.0125, 0.0262, -0.0139, -0.0155, -0.0126, + -0.0356, -0.0274], device='cuda:0'), grad: tensor([-9.6858e-08, 3.6880e-07, -6.0908e-07, -2.4214e-08, -1.3877e-06, + -8.3260e-07, 1.0803e-07, 7.5623e-07, 1.3225e-07, 1.5721e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 216.82, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.5012 re_mapping 0.0042 re_causal 0.0126 /// teacc 98.88 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.2615, -0.1447, -0.0906, ..., -0.0037, -0.0550, -0.0141], + [-0.1178, 0.1207, -0.0816, ..., 0.0183, -0.0508, -0.0804], + [ 0.1114, -0.1575, -0.1678, ..., -0.0634, 0.1451, -0.0331], + ..., + [-0.1661, -0.0919, 0.1189, ..., 0.0379, -0.2009, -0.0017], + [ 0.0905, -0.0069, -0.2853, ..., -0.0681, 0.1501, -0.0347], + [-0.2032, -0.1678, 0.0317, ..., -0.0160, -0.1974, -0.0684]], + device='cuda:0'), grad: tensor([[ 6.7055e-08, 8.1956e-08, -1.4901e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 1.6019e-07, 3.3714e-07, 4.4703e-08, ..., 0.0000e+00, + 6.8918e-08, 0.0000e+00], + [-1.1753e-06, 7.4506e-09, -2.0303e-07, ..., 0.0000e+00, + -1.1995e-06, 0.0000e+00], + ..., + [ 1.1791e-06, 9.3132e-09, 3.3341e-07, ..., 0.0000e+00, + 1.3877e-06, 0.0000e+00], + [ 2.8685e-07, 3.9674e-07, 1.3039e-07, ..., 0.0000e+00, + 6.8918e-08, 0.0000e+00], + [ 1.4901e-08, 9.7416e-07, -8.7544e-08, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0057, 0.0217, -0.0163, 0.0125, 0.0258, -0.0139, -0.0155, -0.0127, + -0.0353, -0.0272], device='cuda:0'), grad: tensor([ 2.6077e-08, 1.0990e-06, -2.2352e-06, -7.7859e-07, -1.8142e-06, + -1.9427e-06, 6.1281e-07, 3.4627e-06, 1.9688e-06, -4.0419e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 216.85, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4793 re_mapping 0.0041 re_causal 0.0120 /// teacc 98.99 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.2625, -0.1453, -0.0903, ..., -0.0036, -0.0551, -0.0141], + [-0.1179, 0.1210, -0.0819, ..., 0.0183, -0.0509, -0.0804], + [ 0.1115, -0.1577, -0.1689, ..., -0.0634, 0.1450, -0.0331], + ..., + [-0.1662, -0.0921, 0.1202, ..., 0.0379, -0.2008, -0.0017], + [ 0.0907, -0.0068, -0.2858, ..., -0.0681, 0.1504, -0.0347], + [-0.2033, -0.1642, 0.0364, ..., -0.0161, -0.1978, -0.0684]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, -4.8429e-08, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-2.0489e-08, 5.5879e-09, -3.7253e-09, ..., 0.0000e+00, + -3.7253e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 3.1665e-08, -3.7253e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 3.7253e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 5.5879e-09, 2.9802e-08, 1.6764e-08, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0056, 0.0217, -0.0166, 0.0126, 0.0219, -0.0140, -0.0154, -0.0123, + -0.0352, -0.0234], device='cuda:0'), grad: tensor([ 1.3039e-08, -4.8429e-08, -6.8918e-08, 1.8626e-08, -3.9116e-08, + 1.1176e-08, -3.5390e-08, 4.2841e-08, 1.3039e-08, 8.5682e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 216.81, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4759 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.02 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.2643, -0.1463, -0.0905, ..., -0.0036, -0.0557, -0.0141], + [-0.1182, 0.1211, -0.0818, ..., 0.0183, -0.0509, -0.0804], + [ 0.1115, -0.1578, -0.1692, ..., -0.0634, 0.1450, -0.0331], + ..., + [-0.1664, -0.0923, 0.1206, ..., 0.0379, -0.2012, -0.0017], + [ 0.0912, -0.0066, -0.2862, ..., -0.0681, 0.1513, -0.0347], + [-0.2034, -0.1643, 0.0363, ..., -0.0161, -0.1982, -0.0684]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 1.2666e-07, 0.0000e+00, 7.4506e-09, ..., 0.0000e+00, + 1.0990e-07, 0.0000e+00], + [-1.3970e-07, 7.4506e-09, 1.8626e-09, ..., 0.0000e+00, + -3.3714e-07, 0.0000e+00], + ..., + [ 4.6566e-08, 1.6764e-08, -1.8626e-08, ..., 0.0000e+00, + 1.0058e-07, 0.0000e+00], + [-6.0536e-07, -1.1921e-07, 0.0000e+00, ..., 0.0000e+00, + -4.5076e-07, 0.0000e+00], + [ 5.7742e-08, 2.7940e-08, 1.4901e-08, ..., 0.0000e+00, + 1.0058e-07, 0.0000e+00]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0057, 0.0219, -0.0166, 0.0126, 0.0220, -0.0140, -0.0149, -0.0124, + -0.0349, -0.0235], device='cuda:0'), grad: tensor([-9.6858e-08, 4.1164e-07, -7.8231e-07, 3.9861e-07, -8.1956e-08, + 9.6671e-07, 3.4645e-07, 1.6950e-07, -1.7080e-06, 3.5763e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 216.71, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4774 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.08 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.2672, -0.1477, -0.0908, ..., -0.0037, -0.0563, -0.0141], + [-0.1187, 0.1212, -0.0828, ..., 0.0183, -0.0515, -0.0804], + [ 0.1124, -0.1578, -0.1731, ..., -0.0634, 0.1440, -0.0331], + ..., + [-0.1665, -0.0923, 0.1250, ..., 0.0379, -0.1993, -0.0017], + [ 0.0920, -0.0059, -0.2870, ..., -0.0683, 0.1518, -0.0347], + [-0.2035, -0.1644, 0.0362, ..., -0.0163, -0.1984, -0.0684]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, -1.2554e-06, -7.0781e-08, ..., 0.0000e+00, + -1.2852e-07, 0.0000e+00], + [-1.4901e-08, 4.2468e-07, 3.1665e-08, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 8.0280e-07, 5.5879e-09, ..., 0.0000e+00, + 1.3970e-07, 0.0000e+00], + [ 7.4506e-09, 3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 9.1270e-08, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-08, 0.0000e+00]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0060, 0.0213, -0.0186, 0.0126, 0.0222, -0.0140, -0.0151, -0.0098, + -0.0346, -0.0237], device='cuda:0'), grad: tensor([-3.4645e-07, -4.1910e-06, 1.5292e-06, -1.8198e-06, -1.6764e-08, + 1.2107e-06, 1.6391e-07, 2.7958e-06, 3.6694e-07, 3.0734e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 216.99, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4706 re_mapping 0.0039 re_causal 0.0113 /// teacc 98.95 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.2685, -0.1487, -0.0910, ..., -0.0037, -0.0568, -0.0141], + [-0.1195, 0.1204, -0.0860, ..., 0.0183, -0.0548, -0.0804], + [ 0.1133, -0.1567, -0.1727, ..., -0.0634, 0.1466, -0.0331], + ..., + [-0.1673, -0.0916, 0.1272, ..., 0.0379, -0.2000, -0.0017], + [ 0.0938, -0.0058, -0.2875, ..., -0.0683, 0.1532, -0.0347], + [-0.2037, -0.1645, 0.0361, ..., -0.0164, -0.1988, -0.0684]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.5646e-07, 5.4017e-08, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 3.7253e-09, 1.4715e-07, 7.4506e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-2.4028e-07, 1.4715e-07, 7.0781e-08, ..., 0.0000e+00, + -5.5321e-07, 0.0000e+00], + ..., + [ 1.6764e-08, 5.5879e-08, 1.3039e-08, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00], + [ 1.8626e-09, 1.4715e-07, 2.4214e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 1.8626e-09, 9.1717e-06, 2.9746e-06, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0062, 0.0176, -0.0170, 0.0129, 0.0223, -0.0144, -0.0150, -0.0094, + -0.0335, -0.0239], device='cuda:0'), grad: tensor([ 6.0722e-07, 4.8243e-07, -1.0822e-06, 1.5143e-06, -3.8117e-05, + 2.8126e-07, 6.5379e-07, 1.4715e-07, 5.1223e-07, 3.4958e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 216.68, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4834 re_mapping 0.0038 re_causal 0.0116 /// teacc 98.98 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.2694, -0.1481, -0.0910, ..., -0.0037, -0.0566, -0.0141], + [-0.1217, 0.1200, -0.0858, ..., 0.0183, -0.0564, -0.0804], + [ 0.1148, -0.1551, -0.1728, ..., -0.0634, 0.1478, -0.0331], + ..., + [-0.1674, -0.0921, 0.1275, ..., 0.0379, -0.2001, -0.0017], + [ 0.0939, -0.0055, -0.2880, ..., -0.0683, 0.1536, -0.0347], + [-0.2038, -0.1646, 0.0361, ..., -0.0164, -0.1992, -0.0684]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3970e-07, 0.0000e+00, ..., 0.0000e+00, + 1.1548e-07, 0.0000e+00], + [ 5.5879e-09, -1.7621e-06, 0.0000e+00, ..., 0.0000e+00, + -1.3877e-06, 0.0000e+00], + [ 2.4214e-08, 1.8626e-08, 0.0000e+00, ..., 0.0000e+00, + 6.3330e-08, 0.0000e+00], + ..., + [ 9.3132e-09, 2.2724e-07, 0.0000e+00, ..., 0.0000e+00, + 1.7881e-07, 0.0000e+00], + [ 7.4506e-09, 1.2144e-06, 0.0000e+00, ..., 0.0000e+00, + 1.0263e-06, 0.0000e+00], + [ 0.0000e+00, 3.9116e-08, -5.4017e-08, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0058, 0.0164, -0.0162, 0.0128, 0.0224, -0.0141, -0.0168, -0.0094, + -0.0336, -0.0240], device='cuda:0'), grad: tensor([ 5.6624e-07, -6.9812e-06, 2.4401e-07, -5.2340e-07, 2.4773e-07, + 2.3283e-07, 4.2096e-07, 9.1828e-07, 5.0440e-06, -1.8626e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 216.48, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4663 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.02 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.2708, -0.1454, -0.0911, ..., -0.0037, -0.0538, -0.0141], + [-0.1229, 0.1194, -0.0859, ..., 0.0183, -0.0572, -0.0804], + [ 0.1156, -0.1539, -0.1728, ..., -0.0634, 0.1485, -0.0331], + ..., + [-0.1675, -0.0922, 0.1287, ..., 0.0379, -0.2002, -0.0017], + [ 0.0941, -0.0057, -0.2886, ..., -0.0683, 0.1538, -0.0347], + [-0.2039, -0.1647, 0.0359, ..., -0.0164, -0.1998, -0.0684]], + device='cuda:0'), grad: tensor([[ 3.9116e-07, 1.0375e-06, 1.8626e-09, ..., 0.0000e+00, + 1.1064e-06, 0.0000e+00], + [ 2.8908e-05, 9.4593e-05, 1.6764e-08, ..., 0.0000e+00, + 9.1910e-05, 0.0000e+00], + [-3.0726e-05, -1.0264e-04, 3.7253e-09, ..., 0.0000e+00, + -9.8348e-05, 0.0000e+00], + ..., + [ 6.7055e-08, 4.6380e-07, 2.0489e-08, ..., 0.0000e+00, + 2.1793e-07, 0.0000e+00], + [ 4.5076e-07, 3.8594e-06, 7.4506e-09, ..., 0.0000e+00, + 2.4904e-06, 0.0000e+00], + [ 5.5879e-08, 6.6310e-07, 3.2596e-07, ..., 0.0000e+00, + 1.7323e-07, 0.0000e+00]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0042, 0.0154, -0.0157, 0.0125, 0.0227, -0.0138, -0.0188, -0.0092, + -0.0337, -0.0243], device='cuda:0'), grad: tensor([ 5.3756e-06, 4.7112e-04, -5.0735e-04, 6.1691e-06, 7.7114e-07, + 2.0061e-06, 2.2054e-06, 1.5628e-06, 1.6019e-05, 2.7511e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 216.53, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4761 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.04 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.2715, -0.1456, -0.0910, ..., -0.0037, -0.0541, -0.0141], + [-0.1236, 0.1195, -0.0861, ..., 0.0183, -0.0579, -0.0804], + [ 0.1161, -0.1526, -0.1728, ..., -0.0634, 0.1492, -0.0331], + ..., + [-0.1676, -0.0931, 0.1289, ..., 0.0379, -0.2006, -0.0018], + [ 0.0942, -0.0062, -0.2888, ..., -0.0683, 0.1540, -0.0347], + [-0.2040, -0.1648, 0.0359, ..., -0.0164, -0.2004, -0.0684]], + device='cuda:0'), grad: tensor([[ 4.4703e-08, 6.1467e-08, 3.7253e-09, ..., 0.0000e+00, + 3.3528e-08, 0.0000e+00], + [ 1.8626e-08, -1.2480e-06, 1.8813e-07, ..., 0.0000e+00, + -2.2072e-06, 0.0000e+00], + [ 5.5879e-09, 1.0803e-07, 2.0489e-08, ..., 0.0000e+00, + 1.0245e-07, 0.0000e+00], + ..., + [ 7.2643e-08, -4.0978e-07, -3.0361e-07, ..., 0.0000e+00, + 1.5832e-07, 0.0000e+00], + [-3.4645e-07, 1.3895e-06, 1.6764e-08, ..., 0.0000e+00, + 1.6950e-06, 0.0000e+00], + [ 2.4959e-07, 1.3784e-07, 5.0291e-08, ..., 0.0000e+00, + 1.1921e-07, 0.0000e+00]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0040, 0.0152, -0.0152, 0.0126, 0.0228, -0.0138, -0.0190, -0.0094, + -0.0339, -0.0244], device='cuda:0'), grad: tensor([ 2.7567e-07, -2.9933e-06, 6.1281e-07, 4.0978e-08, 1.9930e-07, + 3.3341e-07, -2.4028e-07, -4.2319e-06, 4.3064e-06, 1.6782e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 216.79, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4960 re_mapping 0.0040 re_causal 0.0115 /// teacc 98.98 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.2748, -0.1459, -0.0911, ..., -0.0037, -0.0544, -0.0141], + [-0.1238, 0.1193, -0.0867, ..., 0.0183, -0.0579, -0.0804], + [ 0.1159, -0.1528, -0.1730, ..., -0.0635, 0.1490, -0.0331], + ..., + [-0.1678, -0.0933, 0.1285, ..., 0.0378, -0.2008, -0.0018], + [ 0.0950, -0.0063, -0.2911, ..., -0.0684, 0.1549, -0.0348], + [-0.2042, -0.1642, 0.0369, ..., -0.0164, -0.2009, -0.0685]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-09, 3.7253e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, -2.1234e-07, 7.4506e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-2.2352e-08, 3.5390e-08, 1.8626e-09, ..., 0.0000e+00, + -6.5193e-08, 0.0000e+00], + ..., + [ 7.4506e-09, 1.4901e-07, 1.4901e-08, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 3.0734e-07, 2.0489e-08, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-07, 0.0000e+00], + [ 4.6566e-08, 2.0824e-06, 1.4398e-06, ..., 0.0000e+00, + 6.7055e-08, 0.0000e+00]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0040, 0.0147, -0.0154, 0.0127, 0.0199, -0.0139, -0.0190, -0.0098, + -0.0337, -0.0215], device='cuda:0'), grad: tensor([-1.8030e-06, -2.5518e-07, 1.6764e-08, 7.2457e-07, -1.0371e-05, + -1.9539e-06, 1.3411e-07, 4.0792e-07, 1.5981e-06, 1.1466e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 216.73, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4559 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.2767, -0.1464, -0.0912, ..., -0.0037, -0.0545, -0.0141], + [-0.1240, 0.1195, -0.0866, ..., 0.0183, -0.0582, -0.0804], + [ 0.1161, -0.1528, -0.1730, ..., -0.0635, 0.1492, -0.0331], + ..., + [-0.1679, -0.0935, 0.1300, ..., 0.0377, -0.2009, -0.0018], + [ 0.0954, -0.0058, -0.2918, ..., -0.0684, 0.1558, -0.0348], + [-0.2043, -0.1642, 0.0366, ..., -0.0164, -0.2013, -0.0685]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.7253e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, 9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 1.8626e-09, 1.4901e-08, -3.1665e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.7136e-07, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 3.2224e-07, 1.4901e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0042, 0.0146, -0.0154, 0.0128, 0.0199, -0.0139, -0.0189, -0.0091, + -0.0333, -0.0216], device='cuda:0'), grad: tensor([-5.5879e-08, 3.1665e-08, 1.1176e-07, 2.6077e-08, -1.0692e-06, + -3.9302e-07, 5.4017e-08, -1.5832e-07, 3.9116e-07, 1.0487e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 216.76, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4760 re_mapping 0.0039 re_causal 0.0124 /// teacc 99.00 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.2773, -0.1464, -0.0912, ..., -0.0037, -0.0545, -0.0141], + [-0.1242, 0.1195, -0.0868, ..., 0.0183, -0.0583, -0.0804], + [ 0.1163, -0.1528, -0.1731, ..., -0.0635, 0.1495, -0.0331], + ..., + [-0.1682, -0.0936, 0.1300, ..., 0.0377, -0.2011, -0.0018], + [ 0.0957, -0.0055, -0.2925, ..., -0.0684, 0.1560, -0.0348], + [-0.2043, -0.1642, 0.0366, ..., -0.0164, -0.2015, -0.0685]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.4901e-08, 7.4506e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 5.5879e-09, 1.7434e-06, 1.3188e-06, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 9.8720e-08, 6.7055e-08, 3.5390e-08, ..., 0.0000e+00, + 1.1176e-07, 0.0000e+00], + ..., + [ 1.1176e-08, 8.8662e-07, 6.0536e-07, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [-1.6764e-07, 1.0617e-07, 1.3970e-07, ..., 0.0000e+00, + -2.2724e-07, 0.0000e+00], + [ 3.7253e-09, 7.7672e-07, 5.4576e-07, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0040, 0.0145, -0.0153, 0.0128, 0.0199, -0.0139, -0.0190, -0.0091, + -0.0331, -0.0216], device='cuda:0'), grad: tensor([ 6.8918e-08, 6.6012e-06, 4.9360e-07, 1.6391e-07, -1.3821e-05, + 6.7055e-08, 1.4529e-07, 3.0994e-06, 1.2107e-07, 3.0082e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 216.61, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4880 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.08 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.2785, -0.1466, -0.0913, ..., -0.0037, -0.0546, -0.0141], + [-0.1243, 0.1197, -0.0861, ..., 0.0183, -0.0583, -0.0804], + [ 0.1165, -0.1531, -0.1732, ..., -0.0635, 0.1497, -0.0331], + ..., + [-0.1683, -0.0939, 0.1297, ..., 0.0377, -0.2014, -0.0019], + [ 0.0962, -0.0053, -0.2945, ..., -0.0684, 0.1568, -0.0348], + [-0.2044, -0.1643, 0.0366, ..., -0.0164, -0.2017, -0.0688]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -2.2352e-08, 0.0000e+00], + [ 3.7253e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 7.4506e-09, 7.4506e-09, -1.8626e-09, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 2.2352e-08, 1.4901e-08, 1.8626e-09, ..., 0.0000e+00, + -3.9116e-08, 0.0000e+00], + [ 3.7253e-09, 8.5682e-08, 1.8626e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0041, 0.0147, -0.0154, 0.0128, 0.0200, -0.0139, -0.0191, -0.0093, + -0.0328, -0.0217], device='cuda:0'), grad: tensor([-1.1176e-07, 6.1467e-08, 1.8626e-08, 2.0675e-07, -3.9302e-07, + -4.4890e-07, 1.3039e-07, 6.7055e-08, 1.0990e-07, 3.6322e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 216.91, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4644 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.05 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.2796, -0.1477, -0.0913, ..., -0.0037, -0.0547, -0.0141], + [-0.1251, 0.1195, -0.0857, ..., 0.0183, -0.0586, -0.0804], + [ 0.1165, -0.1532, -0.1732, ..., -0.0636, 0.1498, -0.0331], + ..., + [-0.1687, -0.0941, 0.1296, ..., 0.0376, -0.2017, -0.0019], + [ 0.0980, -0.0043, -0.2949, ..., -0.0685, 0.1587, -0.0348], + [-0.2045, -0.1643, 0.0366, ..., -0.0164, -0.2020, -0.0688]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-1.0431e-07, -2.0489e-08, -3.7253e-09, ..., 0.0000e+00, + -1.3039e-07, 0.0000e+00], + [ 3.7253e-09, 3.5390e-08, 1.1176e-08, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0044, 0.0145, -0.0153, 0.0127, 0.0201, -0.0139, -0.0187, -0.0096, + -0.0311, -0.0218], device='cuda:0'), grad: tensor([ 0.0000e+00, -2.6077e-08, 5.5879e-09, 2.4028e-07, -1.1176e-07, + 4.4703e-08, 7.4506e-09, 3.5390e-08, -3.2410e-07, 1.2107e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 216.88, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4679 re_mapping 0.0035 re_causal 0.0114 /// teacc 99.09 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.2800, -0.1476, -0.0942, ..., -0.0037, -0.0547, -0.0141], + [-0.1252, 0.1200, -0.0848, ..., 0.0183, -0.0586, -0.0804], + [ 0.1168, -0.1534, -0.1732, ..., -0.0638, 0.1502, -0.0331], + ..., + [-0.1690, -0.0946, 0.1293, ..., 0.0374, -0.2021, -0.0019], + [ 0.0979, -0.0046, -0.2958, ..., -0.0686, 0.1586, -0.0348], + [-0.2046, -0.1645, 0.0363, ..., -0.0165, -0.2029, -0.0688]], + device='cuda:0'), grad: tensor([[ 2.9802e-08, 0.0000e+00, -3.2317e-07, ..., 0.0000e+00, + 4.5635e-08, 0.0000e+00], + [ 6.5193e-09, 9.3132e-10, 1.0245e-08, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [-9.4995e-08, 9.3132e-10, 8.3819e-09, ..., 0.0000e+00, + -1.8254e-07, 0.0000e+00], + ..., + [ 1.7695e-08, 1.8626e-09, -8.7731e-07, ..., 0.0000e+00, + 4.0978e-08, 0.0000e+00], + [ 4.6566e-08, -2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 2.0675e-07, 0.0000e+00], + [ 5.4948e-08, 3.9116e-08, 1.2238e-06, ..., 0.0000e+00, + 1.4249e-07, 0.0000e+00]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0068, 0.0147, -0.0152, 0.0127, 0.0207, -0.0140, -0.0187, -0.0098, + -0.0314, -0.0221], device='cuda:0'), grad: tensor([-4.5076e-06, 8.2888e-08, -4.3586e-07, -4.7535e-05, -1.9465e-07, + 4.6611e-05, 5.4948e-08, -2.1104e-06, 6.5193e-07, 7.3984e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 216.68, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4890 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.04 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.2814, -0.1481, -0.0943, ..., -0.0037, -0.0552, -0.0141], + [-0.1259, 0.1202, -0.0854, ..., 0.0185, -0.0589, -0.0804], + [ 0.1169, -0.1535, -0.1733, ..., -0.0640, 0.1501, -0.0331], + ..., + [-0.1691, -0.0947, 0.1297, ..., 0.0373, -0.2022, -0.0019], + [ 0.0984, -0.0040, -0.2960, ..., -0.0687, 0.1592, -0.0348], + [-0.2047, -0.1646, 0.0362, ..., -0.0166, -0.2054, -0.0688]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -3.4831e-07, 0.0000e+00], + [ 1.8626e-09, 2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 1.8626e-09, -2.7940e-09, ..., 0.0000e+00, + 5.8673e-08, 0.0000e+00], + [-2.1420e-08, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7660e-07, 0.0000e+00], + [ 2.7940e-09, 2.7474e-07, 2.7940e-09, ..., 0.0000e+00, + 3.9116e-08, 0.0000e+00]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0070, 0.0145, -0.0153, 0.0131, 0.0207, -0.0141, -0.0185, -0.0098, + -0.0309, -0.0221], device='cuda:0'), grad: tensor([-4.7870e-06, 1.1548e-07, 1.3225e-07, -3.1665e-07, -3.3021e-05, + 2.2538e-07, 2.0396e-07, 3.6042e-07, 6.3814e-06, 3.0756e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 216.90, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4769 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.2826, -0.1482, -0.0943, ..., -0.0038, -0.0557, -0.0141], + [-0.1261, 0.1206, -0.0856, ..., 0.0184, -0.0589, -0.0804], + [ 0.1175, -0.1535, -0.1733, ..., -0.0644, 0.1508, -0.0331], + ..., + [-0.1692, -0.0948, 0.1327, ..., 0.0369, -0.2023, -0.0019], + [ 0.0980, -0.0047, -0.2964, ..., -0.0690, 0.1585, -0.0348], + [-0.2050, -0.1647, 0.0358, ..., -0.0166, -0.2066, -0.0688]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-7.4506e-09, -1.7975e-07, 9.3132e-10, ..., 0.0000e+00, + -1.5832e-08, 0.0000e+00], + [ 3.7253e-09, 4.8429e-08, -1.8626e-09, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + ..., + [ 4.6566e-09, 6.3330e-08, 3.7253e-09, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [ 2.0489e-08, 8.6613e-08, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 4.6566e-09, 8.3819e-08, 7.6368e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0068, 0.0146, -0.0151, 0.0130, 0.0208, -0.0140, -0.0183, -0.0079, + -0.0318, -0.0226], device='cuda:0'), grad: tensor([-4.1910e-08, -3.4925e-07, 8.0094e-08, 2.0675e-07, -4.0792e-07, + -2.4959e-07, -8.7544e-08, 1.5553e-07, 1.9372e-07, 4.7684e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 216.73, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4370 re_mapping 0.0036 re_causal 0.0114 /// teacc 99.10 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.2835, -0.1484, -0.0943, ..., -0.0038, -0.0558, -0.0141], + [-0.1263, 0.1202, -0.0867, ..., 0.0184, -0.0589, -0.0804], + [ 0.1177, -0.1536, -0.1733, ..., -0.0644, 0.1510, -0.0331], + ..., + [-0.1693, -0.0953, 0.1328, ..., 0.0369, -0.2025, -0.0019], + [ 0.0977, -0.0054, -0.2970, ..., -0.0690, 0.1584, -0.0348], + [-0.2051, -0.1647, 0.0358, ..., -0.0166, -0.2068, -0.0688]], + device='cuda:0'), grad: tensor([[ 1.3970e-08, 2.6077e-08, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 3.7253e-09, 1.1735e-07, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-8.3819e-09, 3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + -1.7229e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 1.0245e-08, 1.8626e-09, ..., 0.0000e+00, + 1.0431e-07, 0.0000e+00], + [ 6.5193e-09, 5.5879e-09, 9.3132e-10, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 9.3132e-10, 2.4308e-07, -5.5879e-09, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0068, 0.0143, -0.0151, 0.0130, 0.0209, -0.0139, -0.0181, -0.0080, + -0.0322, -0.0227], device='cuda:0'), grad: tensor([ 5.0291e-08, 2.7474e-07, -5.0850e-07, 1.5087e-07, -2.7969e-05, + -1.3597e-07, 2.7195e-05, 3.8929e-07, 4.7497e-08, 5.3830e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 216.67, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4838 re_mapping 0.0036 re_causal 0.0119 /// teacc 98.95 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.2843, -0.1488, -0.0944, ..., -0.0038, -0.0558, -0.0141], + [-0.1266, 0.1201, -0.0872, ..., 0.0184, -0.0591, -0.0804], + [ 0.1176, -0.1539, -0.1733, ..., -0.0644, 0.1510, -0.0331], + ..., + [-0.1694, -0.0954, 0.1330, ..., 0.0369, -0.2026, -0.0019], + [ 0.0976, -0.0057, -0.2978, ..., -0.0690, 0.1584, -0.0348], + [-0.2052, -0.1648, 0.0357, ..., -0.0166, -0.2071, -0.0688]], + device='cuda:0'), grad: tensor([[ 2.7008e-08, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + [ 1.5367e-07, -1.5274e-07, 3.7253e-09, ..., 0.0000e+00, + 3.5763e-07, 0.0000e+00], + [-1.4165e-06, 1.7695e-08, 1.8626e-09, ..., 0.0000e+00, + -3.3602e-06, 0.0000e+00], + ..., + [ 9.1270e-07, 1.1455e-07, -4.6566e-09, ..., 0.0000e+00, + 2.1718e-06, 0.0000e+00], + [ 2.1327e-07, 9.3132e-09, 3.7253e-09, ..., 0.0000e+00, + 4.8894e-07, 0.0000e+00], + [ 1.7695e-08, 2.7940e-09, -9.3132e-09, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0075, 0.0140, -0.0151, 0.0129, 0.0209, -0.0138, -0.0181, -0.0079, + -0.0327, -0.0226], device='cuda:0'), grad: tensor([-3.6508e-07, 2.0489e-07, -4.1947e-06, -1.5087e-07, 1.2387e-07, + 2.8964e-07, 5.7742e-08, 2.9914e-06, 7.3668e-07, 3.0175e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 216.76, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4742 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.00 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.2867, -0.1504, -0.0944, ..., -0.0038, -0.0566, -0.0141], + [-0.1269, 0.1204, -0.0875, ..., 0.0184, -0.0596, -0.0804], + [ 0.1179, -0.1540, -0.1732, ..., -0.0644, 0.1515, -0.0331], + ..., + [-0.1697, -0.0957, 0.1331, ..., 0.0369, -0.2030, -0.0019], + [ 0.0978, -0.0064, -0.2985, ..., -0.0690, 0.1588, -0.0348], + [-0.2056, -0.1649, 0.0357, ..., -0.0166, -0.2082, -0.0688]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 2.0489e-08, 6.2399e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 4.2096e-07, 2.0396e-07, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 5.5879e-09, 7.9162e-08, 1.1269e-07, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + ..., + [ 2.7940e-09, -2.5891e-07, 3.1926e-06, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [-4.4703e-08, 4.2841e-08, 1.3784e-07, ..., 0.0000e+00, + -3.2596e-08, 0.0000e+00], + [ 2.7940e-09, 4.4424e-07, 2.5518e-06, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0069, 0.0138, -0.0149, 0.0129, 0.0211, -0.0138, -0.0176, -0.0079, + -0.0330, -0.0229], device='cuda:0'), grad: tensor([ 9.7230e-07, 4.0084e-06, 1.8841e-06, -1.3979e-06, -9.6917e-05, + 2.7604e-06, 1.4780e-06, 4.6611e-05, 2.0228e-06, 3.8505e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 216.87, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4962 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.09 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.2879, -0.1504, -0.0944, ..., -0.0038, -0.0583, -0.0141], + [-0.1270, 0.1207, -0.0876, ..., 0.0184, -0.0596, -0.0804], + [ 0.1182, -0.1542, -0.1732, ..., -0.0644, 0.1514, -0.0331], + ..., + [-0.1701, -0.0959, 0.1330, ..., 0.0369, -0.2033, -0.0019], + [ 0.0978, -0.0067, -0.2990, ..., -0.0690, 0.1588, -0.0348], + [-0.2058, -0.1652, 0.0353, ..., -0.0166, -0.2086, -0.0688]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 2.7940e-09, -9.1828e-07, -1.8626e-08, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-08, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 3.5483e-07, 8.3819e-09, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + [-1.9558e-08, 2.3283e-08, 0.0000e+00, ..., 0.0000e+00, + -2.1420e-08, 0.0000e+00], + [ 0.0000e+00, 5.5227e-07, 9.3132e-10, ..., 0.0000e+00, + 5.0291e-08, 0.0000e+00]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0070, 0.0138, -0.0150, 0.0131, 0.0221, -0.0138, -0.0177, -0.0081, + -0.0332, -0.0238], device='cuda:0'), grad: tensor([-8.4750e-08, -1.7388e-06, 6.9849e-08, -3.1944e-07, -8.7079e-07, + 6.7055e-08, 2.8778e-07, 8.1491e-07, 4.0047e-08, 1.7444e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 217.22, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4944 re_mapping 0.0037 re_causal 0.0116 /// teacc 98.96 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.2902, -0.1508, -0.0944, ..., -0.0038, -0.0585, -0.0141], + [-0.1271, 0.1209, -0.0878, ..., 0.0184, -0.0596, -0.0804], + [ 0.1183, -0.1544, -0.1732, ..., -0.0644, 0.1515, -0.0331], + ..., + [-0.1707, -0.0961, 0.1339, ..., 0.0369, -0.2037, -0.0019], + [ 0.0982, -0.0069, -0.2993, ..., -0.0690, 0.1591, -0.0349], + [-0.2059, -0.1653, 0.0352, ..., -0.0166, -0.2089, -0.0688]], + device='cuda:0'), grad: tensor([[ 1.3039e-07, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.4959e-07, 0.0000e+00], + [ 9.3132e-10, -4.7497e-08, 4.6566e-09, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [-5.7463e-07, 5.5879e-09, 2.7940e-09, ..., 0.0000e+00, + -1.1064e-06, 0.0000e+00], + ..., + [ 3.7253e-09, 2.0489e-08, -2.1420e-08, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [ 4.0419e-07, 1.1176e-08, 9.3132e-10, ..., 0.0000e+00, + 7.8697e-07, 0.0000e+00], + [ 6.5193e-09, 2.7940e-09, 9.3132e-09, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0070, 0.0138, -0.0150, 0.0129, 0.0222, -0.0137, -0.0178, -0.0071, + -0.0330, -0.0241], device='cuda:0'), grad: tensor([ 5.8766e-07, -6.1467e-08, -2.4959e-06, 2.9895e-07, 2.6077e-08, + -2.9989e-07, 5.5879e-09, -1.0245e-08, 1.8664e-06, 8.2888e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 217.11, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4617 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.09 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.2931, -0.1511, -0.0945, ..., -0.0038, -0.0588, -0.0141], + [-0.1271, 0.1212, -0.0879, ..., 0.0184, -0.0597, -0.0804], + [ 0.1197, -0.1546, -0.1731, ..., -0.0644, 0.1531, -0.0331], + ..., + [-0.1710, -0.0962, 0.1340, ..., 0.0369, -0.2041, -0.0019], + [ 0.0984, -0.0071, -0.2996, ..., -0.0690, 0.1592, -0.0349], + [-0.2060, -0.1653, 0.0352, ..., -0.0166, -0.2096, -0.0689]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 8.3819e-09, 9.3132e-10, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 3.7253e-09, 8.9779e-07, 1.0245e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.1176e-08, 1.4901e-08, 4.6566e-09, ..., 0.0000e+00, + -2.6077e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 7.9162e-08, -3.2317e-07, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 1.8626e-09, 6.3330e-08, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 1.3690e-07, 5.7742e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0067, 0.0138, -0.0147, 0.0131, 0.0222, -0.0139, -0.0175, -0.0071, + -0.0333, -0.0241], device='cuda:0'), grad: tensor([-9.1270e-08, 2.3451e-06, 1.9558e-08, 2.2352e-08, -2.3898e-06, + 1.4156e-07, 8.4750e-08, -1.2247e-06, 1.7509e-07, 9.1828e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 216.94, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4630 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.09 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.2938, -0.1513, -0.0945, ..., -0.0038, -0.0589, -0.0141], + [-0.1271, 0.1222, -0.0870, ..., 0.0184, -0.0596, -0.0804], + [ 0.1198, -0.1548, -0.1731, ..., -0.0644, 0.1533, -0.0331], + ..., + [-0.1711, -0.0975, 0.1339, ..., 0.0369, -0.2042, -0.0019], + [ 0.0983, -0.0077, -0.2998, ..., -0.0690, 0.1594, -0.0350], + [-0.2064, -0.1653, 0.0352, ..., -0.0166, -0.2103, -0.0689]], + device='cuda:0'), grad: tensor([[ 2.7940e-08, 2.6077e-08, 1.8626e-09, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 2.7940e-08, 6.1467e-08, 3.9116e-08, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + [-1.4901e-08, 8.3819e-09, 1.0058e-07, ..., 0.0000e+00, + 4.2841e-08, 0.0000e+00], + ..., + [ 6.5193e-09, 3.6322e-08, -1.4715e-07, ..., 0.0000e+00, + -1.0803e-07, 0.0000e+00], + [ 3.2596e-08, 4.8429e-08, 1.8626e-09, ..., 0.0000e+00, + -2.2352e-08, 0.0000e+00], + [ 9.3132e-09, 4.0233e-07, 4.0978e-08, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0065, 0.0143, -0.0147, 0.0152, 0.0221, -0.0159, -0.0173, -0.0073, + -0.0336, -0.0241], device='cuda:0'), grad: tensor([ 1.6764e-08, 4.8336e-07, 4.4424e-07, 9.8720e-08, -3.0808e-06, + 1.5758e-06, -1.8254e-06, -5.1782e-07, 9.7789e-08, 2.6990e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 217.28, cls_loss 0.0011 cls_loss_mapping 0.0022 cls_loss_causal 0.5038 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.03 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.2960, -0.1515, -0.0945, ..., -0.0038, -0.0596, -0.0141], + [-0.1292, 0.1212, -0.0868, ..., 0.0184, -0.0619, -0.0804], + [ 0.1206, -0.1547, -0.1732, ..., -0.0644, 0.1539, -0.0331], + ..., + [-0.1704, -0.0962, 0.1339, ..., 0.0369, -0.2026, -0.0019], + [ 0.1013, -0.0063, -0.3010, ..., -0.0690, 0.1627, -0.0350], + [-0.2066, -0.1654, 0.0350, ..., -0.0166, -0.2112, -0.0689]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-08, 0.0000e+00], + [ 2.7940e-09, -3.8184e-08, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-9.3132e-10, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.2107e-08, -1.8626e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-1.0617e-07, 2.9802e-08, 0.0000e+00, ..., 0.0000e+00, + -1.3318e-07, 0.0000e+00], + [ 8.8476e-08, 7.1712e-08, 1.4901e-08, ..., 0.0000e+00, + 9.2201e-08, 0.0000e+00]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0065, 0.0125, -0.0144, 0.0152, 0.0223, -0.0166, -0.0170, -0.0063, + -0.0308, -0.0243], device='cuda:0'), grad: tensor([-2.5146e-08, -6.5193e-08, 4.5635e-08, 2.0489e-08, -3.2969e-07, + -1.8813e-07, 1.6764e-07, 2.0489e-08, -3.7439e-07, 7.3388e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 216.97, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4706 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.12 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.2962, -0.1515, -0.0945, ..., -0.0038, -0.0595, -0.0141], + [-0.1295, 0.1218, -0.0863, ..., 0.0184, -0.0625, -0.0804], + [ 0.1208, -0.1549, -0.1734, ..., -0.0644, 0.1541, -0.0331], + ..., + [-0.1702, -0.0964, 0.1339, ..., 0.0369, -0.2020, -0.0019], + [ 0.1014, -0.0064, -0.3014, ..., -0.0690, 0.1628, -0.0350], + [-0.2068, -0.1654, 0.0350, ..., -0.0166, -0.2116, -0.0689]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, -1.7975e-07, 2.1420e-08, ..., 0.0000e+00, + -2.1420e-08, 0.0000e+00], + [-1.5832e-08, 4.5635e-08, 9.3132e-10, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + ..., + [ 4.6566e-09, 7.5437e-08, -4.1910e-08, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 1.5832e-08, 2.8871e-08, 1.8626e-09, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 4.6566e-09, 1.7695e-08, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0063, 0.0125, -0.0144, 0.0152, 0.0223, -0.0166, -0.0172, -0.0062, + -0.0308, -0.0244], device='cuda:0'), grad: tensor([ 1.1176e-08, -1.0151e-07, 5.1223e-08, 2.8871e-08, 4.9360e-08, + -8.5682e-08, 3.3528e-08, -1.9372e-07, 9.1270e-08, 1.0803e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 216.88, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4900 re_mapping 0.0036 re_causal 0.0106 /// teacc 98.99 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.2968, -0.1518, -0.0945, ..., -0.0038, -0.0597, -0.0168], + [-0.1299, 0.1220, -0.0864, ..., 0.0184, -0.0627, -0.0818], + [ 0.1218, -0.1552, -0.1734, ..., -0.0644, 0.1551, -0.0337], + ..., + [-0.1712, -0.0970, 0.1339, ..., 0.0369, -0.2029, -0.0021], + [ 0.1015, -0.0063, -0.3019, ..., -0.0690, 0.1629, -0.0356], + [-0.2072, -0.1655, 0.0350, ..., -0.0166, -0.2111, -0.0720]], + device='cuda:0'), grad: tensor([[ 1.2480e-07, 2.2072e-07, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, -4.3772e-08, 5.5879e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 8.3819e-09, 1.4901e-08, 2.7940e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 2.2352e-08, -6.1467e-08, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [-6.5193e-09, 4.6566e-09, 2.7940e-09, ..., 0.0000e+00, + -1.3970e-08, 0.0000e+00], + [ 1.0245e-08, 4.3772e-08, 5.7742e-08, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0062, 0.0124, -0.0140, 0.0152, 0.0225, -0.0166, -0.0174, -0.0068, + -0.0307, -0.0245], device='cuda:0'), grad: tensor([ 6.5193e-07, -4.6566e-08, 9.1270e-08, 6.9849e-08, -7.6368e-08, + -1.9372e-07, -7.4692e-07, -1.0151e-07, 1.4901e-08, 3.2224e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 216.82, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4698 re_mapping 0.0036 re_causal 0.0111 /// teacc 98.97 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.2991, -0.1519, -0.0945, ..., -0.0038, -0.0606, -0.0170], + [-0.1301, 0.1224, -0.0867, ..., 0.0184, -0.0627, -0.0819], + [ 0.1233, -0.1555, -0.1731, ..., -0.0644, 0.1562, -0.0337], + ..., + [-0.1724, -0.0973, 0.1340, ..., 0.0369, -0.2037, -0.0022], + [ 0.1016, -0.0067, -0.3025, ..., -0.0690, 0.1629, -0.0356], + [-0.2073, -0.1656, 0.0349, ..., -0.0166, -0.2117, -0.0723]], + device='cuda:0'), grad: tensor([[ 1.1828e-07, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.5239e-07, 0.0000e+00], + [ 8.9407e-08, -1.7695e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1269e-07, 0.0000e+00], + [-3.4831e-07, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + -7.3295e-07, 0.0000e+00], + ..., + [ 2.5146e-08, 4.5635e-08, 8.3819e-09, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00], + [-1.9558e-07, -1.4156e-07, 0.0000e+00, ..., 0.0000e+00, + -1.5832e-07, 0.0000e+00], + [ 1.2107e-08, 6.7987e-08, 3.7253e-09, ..., 0.0000e+00, + 1.9558e-08, 0.0000e+00]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0063, 0.0126, -0.0135, 0.0152, 0.0225, -0.0166, -0.0178, -0.0071, + -0.0308, -0.0246], device='cuda:0'), grad: tensor([ 5.5879e-07, 1.7695e-07, -1.6214e-06, 3.2783e-07, -1.7136e-07, + 3.8370e-07, 3.9116e-07, 2.4959e-07, -4.8615e-07, 2.0489e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 216.68, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.4982 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.10 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.3035, -0.1542, -0.0945, ..., -0.0038, -0.0619, -0.0171], + [-0.1304, 0.1231, -0.0871, ..., 0.0184, -0.0628, -0.0819], + [ 0.1265, -0.1560, -0.1733, ..., -0.0644, 0.1596, -0.0337], + ..., + [-0.1758, -0.0976, 0.1345, ..., 0.0369, -0.2056, -0.0023], + [ 0.1020, -0.0070, -0.3038, ..., -0.0690, 0.1631, -0.0357], + [-0.2080, -0.1662, 0.0349, ..., -0.0166, -0.2186, -0.0724]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.3039e-08, 2.7940e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 6.5193e-09, -4.0978e-08, 2.7940e-09, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 9.3132e-09, 4.6566e-09, 2.7940e-09, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 1.5832e-08, 2.8871e-08, -3.7253e-09, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [-1.2014e-07, 2.0489e-08, 4.6566e-09, ..., 0.0000e+00, + -1.2852e-07, 0.0000e+00], + [ 7.5437e-08, 5.4017e-08, -6.5193e-09, ..., 0.0000e+00, + 6.7987e-08, 0.0000e+00]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0068, 0.0127, -0.0117, 0.0151, 0.0225, -0.0164, -0.0169, -0.0074, + -0.0309, -0.0247], device='cuda:0'), grad: tensor([-4.4703e-08, -3.3528e-08, 9.4064e-08, 7.4506e-09, -1.8813e-07, + 1.7416e-07, -1.0058e-07, -1.7695e-08, -2.3562e-07, 3.3248e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 216.76, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4779 re_mapping 0.0038 re_causal 0.0108 /// teacc 99.11 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.3031, -0.1542, -0.0945, ..., -0.0038, -0.0616, -0.0171], + [-0.1310, 0.1233, -0.0874, ..., 0.0184, -0.0629, -0.0819], + [ 0.1283, -0.1563, -0.1734, ..., -0.0644, 0.1613, -0.0337], + ..., + [-0.1762, -0.0979, 0.1346, ..., 0.0369, -0.2059, -0.0024], + [ 0.1018, -0.0075, -0.3044, ..., -0.0690, 0.1630, -0.0358], + [-0.2091, -0.1663, 0.0348, ..., -0.0166, -0.2220, -0.0724]], + device='cuda:0'), grad: tensor([[ 6.8918e-08, -5.5879e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1269e-07, 0.0000e+00], + [ 3.3528e-08, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.0338e-07, 0.0000e+00], + [-4.0419e-07, 3.5390e-08, 0.0000e+00, ..., 0.0000e+00, + -5.8115e-07, 0.0000e+00], + ..., + [ 1.7695e-07, 1.0245e-08, -1.8626e-09, ..., 0.0000e+00, + 2.9150e-07, 0.0000e+00], + [ 5.4948e-08, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 8.1956e-08, 0.0000e+00], + [ 1.7695e-08, 7.0781e-08, -9.3132e-10, ..., 0.0000e+00, + 1.2852e-07, 0.0000e+00]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0051, 0.0127, -0.0110, 0.0145, 0.0226, -0.0160, -0.0186, -0.0072, + -0.0311, -0.0249], device='cuda:0'), grad: tensor([-3.0696e-06, 3.0454e-07, -8.4192e-07, -7.5903e-07, -1.0505e-06, + 3.7905e-07, 3.5428e-06, 7.1898e-07, 1.8347e-07, 5.9977e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 216.68, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4551 re_mapping 0.0037 re_causal 0.0109 /// teacc 99.08 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.3044, -0.1545, -0.0945, ..., -0.0038, -0.0619, -0.0171], + [-0.1312, 0.1236, -0.0874, ..., 0.0184, -0.0630, -0.0819], + [ 0.1286, -0.1564, -0.1733, ..., -0.0644, 0.1617, -0.0337], + ..., + [-0.1765, -0.0983, 0.1346, ..., 0.0369, -0.2061, -0.0024], + [ 0.1020, -0.0077, -0.3047, ..., -0.0690, 0.1632, -0.0358], + [-0.2093, -0.1664, 0.0347, ..., -0.0166, -0.2223, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 2.6077e-08, 1.8626e-09, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [-1.0245e-08, 3.6322e-08, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + [ 5.5879e-09, 2.8964e-07, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0051, 0.0127, -0.0109, 0.0144, 0.0228, -0.0159, -0.0185, -0.0075, + -0.0310, -0.0252], device='cuda:0'), grad: tensor([ 1.2107e-08, 5.8673e-08, 3.5390e-08, -1.3504e-07, -1.6866e-06, + 3.7253e-08, 2.0489e-08, 1.8347e-07, 1.2945e-07, 1.3374e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 216.84, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4759 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.05 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.3048, -0.1545, -0.0945, ..., -0.0038, -0.0620, -0.0171], + [-0.1316, 0.1238, -0.0875, ..., 0.0184, -0.0634, -0.0819], + [ 0.1292, -0.1567, -0.1734, ..., -0.0644, 0.1622, -0.0337], + ..., + [-0.1768, -0.0985, 0.1345, ..., 0.0369, -0.2064, -0.0024], + [ 0.1019, -0.0079, -0.3054, ..., -0.0690, 0.1632, -0.0360], + [-0.2095, -0.1667, 0.0346, ..., -0.0166, -0.2225, -0.0724]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -7.1712e-08, 2.7940e-09, ..., 0.0000e+00, + 3.3528e-08, 0.0000e+00], + [ 0.0000e+00, -1.8626e-08, 1.9558e-08, ..., 0.0000e+00, + 3.8184e-08, 0.0000e+00], + [-8.3819e-09, 4.6566e-09, 4.0047e-08, ..., 0.0000e+00, + -6.9849e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 5.5879e-09, -2.9616e-07, ..., 0.0000e+00, + 2.7660e-07, 0.0000e+00], + [-5.5879e-09, 7.4506e-09, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 3.7253e-09, 5.3085e-08, 1.9465e-07, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0049, 0.0124, -0.0107, 0.0144, 0.0232, -0.0158, -0.0189, -0.0074, + -0.0313, -0.0255], device='cuda:0'), grad: tensor([-9.3877e-07, 2.4214e-07, 1.7788e-07, -1.7229e-06, 3.1292e-07, + 4.7870e-07, 2.4587e-07, -2.0489e-07, 1.2387e-07, 1.2852e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 217.02, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4679 re_mapping 0.0037 re_causal 0.0114 /// teacc 98.97 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.3049, -0.1546, -0.0945, ..., -0.0038, -0.0621, -0.0171], + [-0.1319, 0.1241, -0.0876, ..., 0.0184, -0.0635, -0.0819], + [ 0.1299, -0.1569, -0.1734, ..., -0.0644, 0.1626, -0.0337], + ..., + [-0.1775, -0.0986, 0.1345, ..., 0.0369, -0.2069, -0.0024], + [ 0.1019, -0.0080, -0.3060, ..., -0.0690, 0.1632, -0.0360], + [-0.2100, -0.1668, 0.0345, ..., -0.0166, -0.2230, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 2.7008e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 1.3039e-08, -6.2585e-06, 0.0000e+00, ..., 0.0000e+00, + -2.3842e-06, 0.0000e+00], + [-6.9756e-07, 4.7013e-06, 0.0000e+00, ..., 0.0000e+00, + 1.1483e-06, 0.0000e+00], + ..., + [ 6.5099e-07, 8.2888e-08, 1.6764e-08, ..., 0.0000e+00, + 6.5658e-07, 0.0000e+00], + [ 7.4506e-09, 2.9057e-07, 0.0000e+00, ..., 0.0000e+00, + 1.1362e-07, 0.0000e+00], + [ 7.4506e-09, 4.1537e-07, -1.9558e-08, ..., 0.0000e+00, + 1.3225e-07, 0.0000e+00]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0048, 0.0124, -0.0106, 0.0144, 0.0233, -0.0157, -0.0191, -0.0076, + -0.0313, -0.0257], device='cuda:0'), grad: tensor([-1.8626e-09, -2.4185e-05, 1.6063e-05, -5.1223e-08, 8.6334e-07, + 1.7509e-06, 2.2817e-07, 2.5108e-06, 1.1325e-06, 1.6829e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 216.95, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4634 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.05 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.3059, -0.1550, -0.0945, ..., -0.0038, -0.0628, -0.0171], + [-0.1322, 0.1248, -0.0877, ..., 0.0184, -0.0606, -0.0819], + [ 0.1302, -0.1581, -0.1734, ..., -0.0644, 0.1607, -0.0337], + ..., + [-0.1777, -0.0981, 0.1345, ..., 0.0369, -0.2077, -0.0026], + [ 0.1020, -0.0085, -0.3075, ..., -0.0690, 0.1630, -0.0361], + [-0.2103, -0.1669, 0.0345, ..., -0.0166, -0.2244, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.5832e-08, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [ 0.0000e+00, -1.1642e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.0489e-08, 4.4703e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1828e-07, 0.0000e+00], + [-1.5367e-07, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + -1.3597e-07, 0.0000e+00], + [ 3.7253e-09, 4.6566e-09, -9.3132e-10, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0048, 0.0145, -0.0119, 0.0144, 0.0235, -0.0158, -0.0190, -0.0076, + -0.0320, -0.0258], device='cuda:0'), grad: tensor([-3.7253e-09, -2.2352e-07, 2.5891e-07, -5.3924e-07, 7.1712e-08, + 1.8626e-07, 3.7253e-09, 6.8545e-07, -3.3062e-07, -1.0896e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 216.59, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4854 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.07 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.3064, -0.1550, -0.0945, ..., -0.0038, -0.0631, -0.0171], + [-0.1323, 0.1250, -0.0878, ..., 0.0184, -0.0606, -0.0819], + [ 0.1303, -0.1590, -0.1734, ..., -0.0644, 0.1608, -0.0337], + ..., + [-0.1778, -0.0982, 0.1346, ..., 0.0369, -0.2078, -0.0026], + [ 0.1023, -0.0088, -0.3078, ..., -0.0690, 0.1631, -0.0361], + [-0.2119, -0.1671, 0.0345, ..., -0.0166, -0.2267, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.2573e-08, 2.7940e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.3039e-08, -8.5682e-08, 4.6566e-10, ..., 0.0000e+00, + -3.2596e-09, 0.0000e+00], + [ 9.3132e-10, 1.2573e-08, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 8.8476e-09, -1.3970e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 1.6764e-08, 1.0245e-07, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 6.0536e-09, 1.2340e-07, 4.6566e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0046, 0.0145, -0.0120, 0.0143, 0.0236, -0.0158, -0.0185, -0.0074, + -0.0319, -0.0260], device='cuda:0'), grad: tensor([-1.2936e-06, -1.6997e-07, 7.5437e-08, -2.9337e-08, -4.1910e-07, + 4.4238e-08, -2.5751e-07, 3.3062e-08, 6.3516e-07, 1.3877e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 216.89, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4452 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.00 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.3093, -0.1552, -0.0945, ..., -0.0038, -0.0632, -0.0171], + [-0.1326, 0.1251, -0.0879, ..., 0.0184, -0.0606, -0.0819], + [ 0.1307, -0.1592, -0.1734, ..., -0.0644, 0.1610, -0.0337], + ..., + [-0.1781, -0.0983, 0.1347, ..., 0.0369, -0.2080, -0.0026], + [ 0.1019, -0.0091, -0.3082, ..., -0.0690, 0.1631, -0.0361], + [-0.2121, -0.1672, 0.0345, ..., -0.0166, -0.2270, -0.0724]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 1.9558e-08, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 2.8871e-08, 0.0000e+00], + [ 2.7940e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 4.6566e-09, -9.3132e-10, ..., 0.0000e+00, + 5.6811e-08, 0.0000e+00], + [-3.2596e-07, -1.9278e-07, 0.0000e+00, ..., 0.0000e+00, + -3.8557e-07, 0.0000e+00], + [ 2.5239e-07, 1.6950e-07, 0.0000e+00, ..., 0.0000e+00, + 3.1851e-07, 0.0000e+00]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0050, 0.0145, -0.0120, 0.0143, 0.0237, -0.0156, -0.0184, -0.0075, + -0.0325, -0.0261], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.0151e-07, 9.0338e-08, -5.1409e-07, 1.0151e-07, + 2.8964e-07, -7.4506e-09, 8.0466e-07, -1.2564e-06, 3.1665e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 216.88, cls_loss 0.0012 cls_loss_mapping 0.0025 cls_loss_causal 0.4907 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.04 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.3120, -0.1558, -0.0946, ..., -0.0038, -0.0635, -0.0171], + [-0.1328, 0.1253, -0.0877, ..., 0.0184, -0.0604, -0.0819], + [ 0.1301, -0.1595, -0.1735, ..., -0.0644, 0.1607, -0.0337], + ..., + [-0.1773, -0.0985, 0.1347, ..., 0.0369, -0.2083, -0.0026], + [ 0.1022, -0.0088, -0.3083, ..., -0.0690, 0.1633, -0.0362], + [-0.2131, -0.1674, 0.0344, ..., -0.0166, -0.2278, -0.0724]], + device='cuda:0'), grad: tensor([[-1.3970e-08, -3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 6.6124e-08, 1.8626e-08, 9.3132e-10, ..., 0.0000e+00, + 5.6811e-08, 0.0000e+00], + [ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 2.7940e-09, -3.7253e-09, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [-7.4506e-08, -2.7008e-08, 9.3132e-10, ..., 0.0000e+00, + -4.8429e-08, 0.0000e+00], + [ 1.8626e-09, 5.5879e-09, -5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0056, 0.0146, -0.0126, 0.0144, 0.0269, -0.0154, -0.0185, -0.0071, + -0.0324, -0.0293], device='cuda:0'), grad: tensor([-1.4994e-07, 1.8999e-07, 5.4017e-08, -1.8347e-07, 1.4696e-06, + 6.5193e-08, 1.8720e-07, -1.2573e-07, -1.6764e-08, -1.4808e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 217.10, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4828 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.01 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.3141, -0.1570, -0.0946, ..., -0.0038, -0.0640, -0.0171], + [-0.1334, 0.1256, -0.0884, ..., 0.0184, -0.0605, -0.0819], + [ 0.1311, -0.1596, -0.1730, ..., -0.0644, 0.1617, -0.0337], + ..., + [-0.1776, -0.0981, 0.1345, ..., 0.0369, -0.2087, -0.0026], + [ 0.1024, -0.0090, -0.3098, ..., -0.0690, 0.1634, -0.0363], + [-0.2138, -0.1674, 0.0344, ..., -0.0166, -0.2284, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 1.0245e-08, 0.0000e+00, ..., 0.0000e+00, + 3.0734e-08, 0.0000e+00], + [ 2.4494e-07, 2.9523e-07, 0.0000e+00, ..., 0.0000e+00, + 5.3365e-07, 0.0000e+00], + [ 4.6566e-09, 5.5879e-08, 0.0000e+00, ..., 0.0000e+00, + -1.1362e-07, 0.0000e+00], + ..., + [ 3.3528e-08, 3.7253e-08, 9.3132e-10, ..., 0.0000e+00, + 1.3411e-07, 0.0000e+00], + [-3.7625e-07, -6.9663e-07, 0.0000e+00, ..., 0.0000e+00, + -1.0841e-06, 0.0000e+00], + [ 1.8626e-08, 2.2352e-08, -4.6566e-09, ..., 0.0000e+00, + 4.0047e-08, 0.0000e+00]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0059, 0.0143, -0.0122, 0.0144, 0.0269, -0.0154, -0.0177, -0.0083, + -0.0325, -0.0292], device='cuda:0'), grad: tensor([ 1.9558e-08, 1.4286e-06, -9.3132e-08, -2.1644e-06, 1.8440e-07, + -1.4687e-06, 4.4070e-06, 3.6322e-07, -2.6934e-06, 3.5390e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 216.75, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4597 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.03 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.3158, -0.1573, -0.0946, ..., -0.0038, -0.0642, -0.0171], + [-0.1335, 0.1269, -0.0873, ..., 0.0184, -0.0604, -0.0819], + [ 0.1313, -0.1598, -0.1730, ..., -0.0644, 0.1619, -0.0337], + ..., + [-0.1779, -0.0995, 0.1343, ..., 0.0369, -0.2090, -0.0029], + [ 0.1020, -0.0087, -0.3103, ..., -0.0690, 0.1634, -0.0369], + [-0.2142, -0.1675, 0.0344, ..., -0.0166, -0.2287, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.1176e-08, 3.1665e-08, 1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-1.1828e-07, -6.4634e-07, 3.7253e-09, ..., 0.0000e+00, + -6.8918e-08, 0.0000e+00], + [-6.6124e-08, 1.3970e-08, -3.4459e-08, ..., 0.0000e+00, + -2.5611e-07, 0.0000e+00], + ..., + [ 3.2596e-08, 5.4017e-08, 9.3132e-10, ..., 0.0000e+00, + 1.3411e-07, 0.0000e+00], + [ 4.6566e-08, 1.3318e-07, 9.3132e-10, ..., 0.0000e+00, + 3.2596e-08, 0.0000e+00], + [ 2.7940e-09, 1.3039e-08, 8.3819e-09, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0062, 0.0147, -0.0123, 0.0145, 0.0269, -0.0153, -0.0170, -0.0087, + -0.0330, -0.0292], device='cuda:0'), grad: tensor([ 8.7544e-08, -1.0403e-06, -6.7241e-07, 2.3376e-07, 1.8440e-07, + 6.0629e-07, -8.6613e-08, 3.3341e-07, 2.7288e-07, 1.0245e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 216.94, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4686 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.05 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.3164, -0.1572, -0.0946, ..., -0.0038, -0.0640, -0.0171], + [-0.1338, 0.1283, -0.0874, ..., 0.0184, -0.0605, -0.0819], + [ 0.1336, -0.1600, -0.1724, ..., -0.0644, 0.1632, -0.0337], + ..., + [-0.1805, -0.1013, 0.1342, ..., 0.0369, -0.2111, -0.0030], + [ 0.1019, -0.0088, -0.3105, ..., -0.0690, 0.1634, -0.0370], + [-0.2144, -0.1675, 0.0344, ..., -0.0166, -0.2290, -0.0724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 9.3132e-10, -1.5832e-08, 2.7940e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 8.3819e-09, 4.6566e-09, 3.7253e-09, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 6.5193e-09, -2.7940e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.0245e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0061, 0.0154, -0.0116, 0.0148, 0.0269, -0.0156, -0.0168, -0.0099, + -0.0331, -0.0292], device='cuda:0'), grad: tensor([-6.1467e-08, 2.9802e-08, 7.5437e-08, 3.9861e-07, 4.9360e-08, + 4.6566e-08, -3.3528e-08, -5.1223e-07, -2.4214e-08, 4.3772e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 216.83, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4755 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.13 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.3168, -0.1573, -0.0946, ..., -0.0038, -0.0638, -0.0171], + [-0.1341, 0.1287, -0.0875, ..., 0.0184, -0.0605, -0.0819], + [ 0.1343, -0.1606, -0.1725, ..., -0.0644, 0.1638, -0.0337], + ..., + [-0.1810, -0.1015, 0.1342, ..., 0.0369, -0.2116, -0.0030], + [ 0.1017, -0.0091, -0.3107, ..., -0.0690, 0.1633, -0.0370], + [-0.2150, -0.1676, 0.0344, ..., -0.0166, -0.2300, -0.0724]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, -2.0489e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 1.8626e-09, 4.6566e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.4214e-08, 2.1420e-08, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 8.3819e-09, -2.7940e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0057, 0.0154, -0.0115, 0.0149, 0.0270, -0.0156, -0.0173, -0.0101, + -0.0333, -0.0294], device='cuda:0'), grad: tensor([-2.7940e-08, -3.0734e-08, 2.0489e-08, 1.0589e-06, 2.5332e-07, + -1.1576e-06, -2.5239e-07, 1.9558e-08, 7.8231e-08, 2.7940e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 216.64, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4743 re_mapping 0.0036 re_causal 0.0109 /// teacc 99.14 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.3172, -0.1575, -0.0946, ..., -0.0038, -0.0641, -0.0171], + [-0.1346, 0.1293, -0.0876, ..., 0.0184, -0.0606, -0.0819], + [ 0.1345, -0.1609, -0.1726, ..., -0.0644, 0.1641, -0.0337], + ..., + [-0.1811, -0.1020, 0.1342, ..., 0.0369, -0.2121, -0.0030], + [ 0.1018, -0.0092, -0.3111, ..., -0.0690, 0.1635, -0.0370], + [-0.2151, -0.1677, 0.0344, ..., -0.0166, -0.2301, -0.0724]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [ 2.7940e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [-8.3819e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, -7.4506e-09, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + [ 3.8184e-08, 1.8626e-08, 1.5832e-08, ..., 0.0000e+00, + 2.7008e-08, 0.0000e+00], + [ 6.5193e-09, 0.0000e+00, -2.2352e-08, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0056, 0.0155, -0.0115, 0.0149, 0.0270, -0.0156, -0.0174, -0.0106, + -0.0333, -0.0294], device='cuda:0'), grad: tensor([-2.1141e-07, 9.1270e-08, 2.5425e-07, -1.7416e-07, 3.6508e-07, + -3.2596e-08, -5.6811e-08, -4.8243e-07, 3.0734e-07, -6.2399e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 216.80, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4801 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.05 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.3178, -0.1580, -0.0946, ..., -0.0038, -0.0646, -0.0171], + [-0.1354, 0.1297, -0.0875, ..., 0.0184, -0.0608, -0.0819], + [ 0.1381, -0.1610, -0.1727, ..., -0.0644, 0.1670, -0.0337], + ..., + [-0.1840, -0.1022, 0.1342, ..., 0.0369, -0.2148, -0.0030], + [ 0.1018, -0.0093, -0.3116, ..., -0.0690, 0.1636, -0.0370], + [-0.2158, -0.1677, 0.0344, ..., -0.0166, -0.2308, -0.0724]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, -3.9395e-07, 0.0000e+00, ..., 0.0000e+00, + -3.3528e-08, 0.0000e+00], + [ 6.3982e-07, 5.7090e-07, 0.0000e+00, ..., 0.0000e+00, + 8.1211e-07, 0.0000e+00], + [ 5.5879e-09, 2.0489e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 7.4506e-09, 1.3970e-08, -9.3132e-10, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [-1.0226e-06, -4.3586e-07, 0.0000e+00, ..., 0.0000e+00, + -1.2498e-06, 0.0000e+00], + [ 5.9605e-08, 6.7987e-08, 0.0000e+00, ..., 0.0000e+00, + 7.6368e-08, 0.0000e+00]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0059, 0.0154, -0.0092, 0.0150, 0.0270, -0.0155, -0.0171, -0.0121, + -0.0335, -0.0293], device='cuda:0'), grad: tensor([-3.8091e-06, 4.9323e-06, 1.0710e-07, -1.1828e-07, 8.8476e-08, + 1.0217e-06, 3.1851e-07, 8.1025e-08, -3.1404e-06, 5.2899e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 216.60, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4899 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.11 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.3181, -0.1580, -0.0946, ..., -0.0038, -0.0648, -0.0171], + [-0.1363, 0.1316, -0.0875, ..., 0.0184, -0.0606, -0.0819], + [ 0.1384, -0.1641, -0.1728, ..., -0.0644, 0.1671, -0.0337], + ..., + [-0.1842, -0.1026, 0.1342, ..., 0.0369, -0.2152, -0.0030], + [ 0.1025, -0.0078, -0.3117, ..., -0.0690, 0.1645, -0.0370], + [-0.2160, -0.1678, 0.0344, ..., -0.0166, -0.2311, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.0617e-07, 1.4063e-07, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [-6.1467e-08, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + -1.6950e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.6764e-08, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 2.7940e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0059, 0.0164, -0.0097, 0.0153, 0.0269, -0.0158, -0.0181, -0.0124, + -0.0329, -0.0293], device='cuda:0'), grad: tensor([-2.9802e-08, 2.5891e-07, -1.6950e-07, 1.4529e-07, 1.7229e-07, + 1.8282e-06, -2.2054e-06, -3.0734e-08, 8.3819e-08, -4.2841e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 217.16, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4606 re_mapping 0.0037 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.3185, -0.1582, -0.0946, ..., -0.0038, -0.0644, -0.0171], + [-0.1380, 0.1310, -0.0875, ..., 0.0184, -0.0607, -0.0819], + [ 0.1383, -0.1647, -0.1728, ..., -0.0644, 0.1671, -0.0337], + ..., + [-0.1839, -0.1019, 0.1341, ..., 0.0369, -0.2152, -0.0030], + [ 0.1028, -0.0078, -0.3118, ..., -0.0690, 0.1650, -0.0370], + [-0.2162, -0.1678, 0.0344, ..., -0.0166, -0.2314, -0.0724]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 2.7940e-09, -7.0967e-06, -5.9605e-08, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [-6.5193e-09, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + ..., + [ 3.1386e-07, 4.2208e-06, 4.4703e-08, ..., 0.0000e+00, + 3.8743e-07, 0.0000e+00], + [-3.9861e-07, 1.0803e-07, 1.8626e-09, ..., 0.0000e+00, + -4.7870e-07, 0.0000e+00], + [ 1.1176e-08, 1.3523e-06, -3.5390e-08, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0054, 0.0155, -0.0098, 0.0152, 0.0270, -0.0159, -0.0182, -0.0123, + -0.0328, -0.0293], device='cuda:0'), grad: tensor([ 1.0245e-08, -1.2547e-05, 1.6764e-08, -7.6368e-08, 2.7660e-06, + 2.7940e-07, 3.2596e-08, 8.1807e-06, -5.4576e-07, 1.8906e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 216.60, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4670 re_mapping 0.0038 re_causal 0.0110 /// teacc 99.12 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.3196, -0.1585, -0.0946, ..., -0.0038, -0.0648, -0.0171], + [-0.1387, 0.1314, -0.0875, ..., 0.0184, -0.0608, -0.0819], + [ 0.1385, -0.1650, -0.1728, ..., -0.0644, 0.1673, -0.0337], + ..., + [-0.1841, -0.1019, 0.1341, ..., 0.0369, -0.2156, -0.0030], + [ 0.1034, -0.0077, -0.3119, ..., -0.0690, 0.1659, -0.0372], + [-0.2172, -0.1679, 0.0344, ..., -0.0166, -0.2320, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 9.3132e-10, 7.4506e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 8.3819e-09, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-1.2107e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + ..., + [ 1.5646e-07, 5.5879e-09, -9.3132e-10, ..., 0.0000e+00, + 1.5181e-07, 0.0000e+00], + [-1.8254e-07, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + -2.2352e-07, 0.0000e+00], + [ 4.4797e-07, 4.0047e-08, -6.5193e-09, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0053, 0.0150, -0.0097, 0.0151, 0.0270, -0.0158, -0.0187, -0.0122, + -0.0323, -0.0293], device='cuda:0'), grad: tensor([ 2.7660e-06, 1.8626e-08, -2.0489e-08, 5.1036e-06, -6.9849e-08, + -7.5027e-06, 2.8871e-07, 2.9523e-07, -3.4086e-07, -5.6438e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 216.52, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4639 re_mapping 0.0036 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.3220, -0.1588, -0.0946, ..., -0.0038, -0.0651, -0.0171], + [-0.1404, 0.1313, -0.0876, ..., 0.0184, -0.0611, -0.0819], + [ 0.1385, -0.1652, -0.1728, ..., -0.0644, 0.1674, -0.0337], + ..., + [-0.1844, -0.1022, 0.1340, ..., 0.0369, -0.2158, -0.0030], + [ 0.1036, -0.0078, -0.3099, ..., -0.0690, 0.1662, -0.0372], + [-0.2186, -0.1679, 0.0344, ..., -0.0166, -0.2323, -0.0724]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -7.1526e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 6.1374e-07, 0.0000e+00, ..., 0.0000e+00, + -2.2352e-08, 0.0000e+00], + ..., + [ 4.6566e-09, 2.9802e-08, -1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 6.5193e-09, 2.3283e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.1420e-08, 2.3283e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0056, 0.0147, -0.0098, 0.0151, 0.0270, -0.0165, -0.0154, -0.0124, + -0.0325, -0.0293], device='cuda:0'), grad: tensor([ 2.6077e-08, -2.1793e-06, 1.9167e-06, 9.5088e-07, -1.8217e-06, + -8.2050e-07, 7.0781e-08, 4.6287e-07, 1.8720e-07, 1.1930e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 216.52, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4364 re_mapping 0.0037 re_causal 0.0105 /// teacc 99.03 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.3223, -0.1588, -0.0946, ..., -0.0038, -0.0660, -0.0170], + [-0.1437, 0.1299, -0.0873, ..., 0.0184, -0.0637, -0.0819], + [ 0.1387, -0.1658, -0.1728, ..., -0.0644, 0.1674, -0.0337], + ..., + [-0.1817, -0.1008, 0.1340, ..., 0.0369, -0.2135, -0.0030], + [ 0.1038, -0.0076, -0.3100, ..., -0.0690, 0.1667, -0.0372], + [-0.2190, -0.1680, 0.0344, ..., -0.0166, -0.2331, -0.0724]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.5832e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 2.4028e-07, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [-2.6077e-08, 9.3132e-08, 0.0000e+00, ..., 0.0000e+00, + -2.2352e-08, 0.0000e+00], + ..., + [ 1.9558e-08, 2.4494e-07, 0.0000e+00, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [-1.9558e-08, 6.7987e-08, 0.0000e+00, ..., 0.0000e+00, + -2.8871e-08, 0.0000e+00], + [ 9.3132e-10, 1.6401e-06, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0054, 0.0124, -0.0100, 0.0148, 0.0270, -0.0158, -0.0165, -0.0098, + -0.0325, -0.0294], device='cuda:0'), grad: tensor([ 1.1921e-07, 1.8030e-06, 4.4610e-07, -2.1607e-07, -1.6436e-05, + 4.3400e-07, 8.6706e-07, 5.5414e-07, 3.8370e-07, 1.2077e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 216.54, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4711 re_mapping 0.0035 re_causal 0.0105 /// teacc 98.97 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.3230, -0.1590, -0.0946, ..., -0.0038, -0.0670, -0.0170], + [-0.1438, 0.1301, -0.0873, ..., 0.0184, -0.0639, -0.0819], + [ 0.1403, -0.1671, -0.1727, ..., -0.0644, 0.1690, -0.0337], + ..., + [-0.1817, -0.1008, 0.1340, ..., 0.0369, -0.2135, -0.0030], + [ 0.1029, -0.0071, -0.3100, ..., -0.0690, 0.1662, -0.0376], + [-0.2195, -0.1681, 0.0344, ..., -0.0166, -0.2340, -0.0725]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.1665e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.0781e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0056, 0.0125, -0.0096, 0.0147, 0.0271, -0.0154, -0.0175, -0.0099, + -0.0327, -0.0294], device='cuda:0'), grad: tensor([-1.2014e-07, 5.6811e-08, 1.4901e-07, -5.2713e-07, 1.4715e-07, + 1.5926e-07, 3.6322e-08, 1.0701e-06, 4.6380e-07, -1.4212e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 216.68, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4612 re_mapping 0.0036 re_causal 0.0107 /// teacc 98.98 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.3234, -0.1591, -0.0946, ..., -0.0038, -0.0672, -0.0170], + [-0.1432, 0.1305, -0.0873, ..., 0.0184, -0.0636, -0.0819], + [ 0.1394, -0.1702, -0.1728, ..., -0.0644, 0.1687, -0.0337], + ..., + [-0.1817, -0.1009, 0.1340, ..., 0.0369, -0.2136, -0.0030], + [ 0.1049, -0.0052, -0.3101, ..., -0.0690, 0.1679, -0.0376], + [-0.2198, -0.1681, 0.0344, ..., -0.0166, -0.2342, -0.0725]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 8.3819e-09, -1.8626e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 1.4901e-07, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0055, 0.0128, -0.0111, 0.0147, 0.0271, -0.0154, -0.0193, -0.0099, + -0.0309, -0.0294], device='cuda:0'), grad: tensor([ 1.0151e-07, 5.2154e-08, 2.1420e-08, -5.4669e-07, -4.7572e-06, + 5.2247e-07, -1.3039e-08, 1.4622e-07, 6.8732e-07, 3.7849e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 216.67, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4652 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.12 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.3237, -0.1592, -0.0946, ..., -0.0038, -0.0676, -0.0170], + [-0.1433, 0.1305, -0.0874, ..., 0.0184, -0.0636, -0.0819], + [ 0.1395, -0.1703, -0.1728, ..., -0.0644, 0.1691, -0.0337], + ..., + [-0.1820, -0.1009, 0.1336, ..., 0.0369, -0.2137, -0.0030], + [ 0.1056, -0.0047, -0.3102, ..., -0.0690, 0.1684, -0.0376], + [-0.2201, -0.1683, 0.0344, ..., -0.0166, -0.2348, -0.0725]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, -3.4366e-07, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 2.0582e-07, 2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 4.2655e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 1.6578e-07, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-2.3842e-07, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + -4.8801e-07, 0.0000e+00], + [ 2.4214e-08, 1.3597e-07, -9.3132e-10, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0051, 0.0128, -0.0109, 0.0147, 0.0271, -0.0150, -0.0205, -0.0100, + -0.0301, -0.0294], device='cuda:0'), grad: tensor([-1.4901e-07, -6.9849e-07, 8.9966e-07, 3.6322e-08, 4.7497e-08, + 9.3132e-09, 2.1420e-08, 3.4235e-06, -9.2015e-07, -2.6748e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 216.59, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4529 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.04 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.3276, -0.1597, -0.0946, ..., -0.0038, -0.0712, -0.0170], + [-0.1433, 0.1306, -0.0873, ..., 0.0184, -0.0636, -0.0819], + [ 0.1395, -0.1705, -0.1728, ..., -0.0644, 0.1689, -0.0337], + ..., + [-0.1821, -0.1009, 0.1335, ..., 0.0369, -0.2138, -0.0030], + [ 0.1059, -0.0047, -0.3104, ..., -0.0690, 0.1687, -0.0376], + [-0.2228, -0.1684, 0.0344, ..., -0.0166, -0.2377, -0.0725]], + device='cuda:0'), grad: tensor([[ 1.0245e-08, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 8.3819e-09, 3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-8.3819e-09, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 1.5832e-08, 9.3132e-10, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 3.1665e-08, 1.7695e-08, 9.3132e-10, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 1.8626e-09, 3.0734e-06, 9.3132e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0085, 0.0128, -0.0114, 0.0179, 0.0271, -0.0150, -0.0233, -0.0099, + -0.0299, -0.0295], device='cuda:0'), grad: tensor([ 5.0291e-08, 2.2352e-07, -9.3132e-09, -1.0151e-07, -1.8641e-05, + 1.0710e-07, -3.8184e-08, 0.0000e+00, 1.1548e-07, 1.8269e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 216.58, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4577 re_mapping 0.0034 re_causal 0.0104 /// teacc 99.00 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.3276, -0.1598, -0.0946, ..., -0.0038, -0.0712, -0.0170], + [-0.1434, 0.1307, -0.0874, ..., 0.0184, -0.0637, -0.0819], + [ 0.1396, -0.1707, -0.1728, ..., -0.0644, 0.1689, -0.0337], + ..., + [-0.1821, -0.1010, 0.1335, ..., 0.0369, -0.2139, -0.0030], + [ 0.1060, -0.0047, -0.3114, ..., -0.0690, 0.1689, -0.0376], + [-0.2229, -0.1685, 0.0344, ..., -0.0166, -0.2378, -0.0725]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + ..., + [ 4.6566e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-08, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0085, 0.0128, -0.0114, 0.0179, 0.0271, -0.0150, -0.0233, -0.0099, + -0.0299, -0.0295], device='cuda:0'), grad: tensor([-1.1642e-07, 3.8184e-08, 3.0734e-08, -2.4214e-07, -3.3993e-07, + 1.3877e-07, 4.1910e-08, 1.3597e-07, 4.8988e-07, -1.7509e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 216.98, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4447 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.01 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.3276, -0.1600, -0.0946, ..., -0.0038, -0.0712, -0.0170], + [-0.1434, 0.1308, -0.0875, ..., 0.0184, -0.0637, -0.0819], + [ 0.1396, -0.1710, -0.1729, ..., -0.0644, 0.1690, -0.0337], + ..., + [-0.1821, -0.1010, 0.1335, ..., 0.0369, -0.2139, -0.0030], + [ 0.1060, -0.0048, -0.3115, ..., -0.0690, 0.1689, -0.0376], + [-0.2230, -0.1685, 0.0344, ..., -0.0166, -0.2377, -0.0725]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 9.3132e-10, 1.0245e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [-1.6764e-08, 2.8871e-08, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00], + [ 1.8626e-09, 4.9360e-08, -9.3132e-10, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0085, 0.0128, -0.0116, 0.0179, 0.0271, -0.0149, -0.0233, -0.0099, + -0.0300, -0.0295], device='cuda:0'), grad: tensor([ 8.3819e-09, 2.7940e-08, 4.0978e-08, -8.5030e-07, -7.4320e-07, + 8.6986e-07, 9.3132e-10, 8.9407e-08, 1.8347e-07, 3.8277e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 216.75, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4778 re_mapping 0.0035 re_causal 0.0102 /// teacc 98.97 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.3276, -0.1602, -0.0947, ..., -0.0038, -0.0712, -0.0170], + [-0.1435, 0.1309, -0.0876, ..., 0.0183, -0.0636, -0.0819], + [ 0.1397, -0.1712, -0.1730, ..., -0.0644, 0.1689, -0.0337], + ..., + [-0.1821, -0.1010, 0.1337, ..., 0.0369, -0.2141, -0.0030], + [ 0.1060, -0.0049, -0.3121, ..., -0.0690, 0.1687, -0.0376], + [-0.2232, -0.1686, 0.0343, ..., -0.0167, -0.2385, -0.0725]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.8626e-09, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [ 6.5193e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 9.3132e-10, 5.5879e-09, -7.4506e-09, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0085, 0.0128, -0.0119, 0.0180, 0.0271, -0.0149, -0.0233, -0.0099, + -0.0302, -0.0296], device='cuda:0'), grad: tensor([ 4.8429e-08, 1.1642e-07, 1.7416e-07, -8.0280e-07, 1.6205e-07, + 5.0757e-07, -8.3819e-08, -3.9395e-07, 6.5193e-08, 2.1514e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 216.64, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4697 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.03 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.3276, -0.1603, -0.0948, ..., -0.0038, -0.0712, -0.0171], + [-0.1435, 0.1309, -0.0877, ..., 0.0183, -0.0635, -0.0819], + [ 0.1397, -0.1713, -0.1731, ..., -0.0644, 0.1690, -0.0337], + ..., + [-0.1833, -0.1010, 0.1307, ..., 0.0369, -0.2142, -0.0030], + [ 0.1060, -0.0049, -0.3125, ..., -0.0690, 0.1687, -0.0376], + [-0.2235, -0.1688, 0.0341, ..., -0.0167, -0.2388, -0.0726]], + device='cuda:0'), grad: tensor([[ 9.7789e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00], + [ 1.3970e-09, -6.6590e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 1.3970e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4435e-08, 0.0000e+00], + ..., + [ 4.6566e-10, 2.7474e-08, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 2.7008e-08, 3.0734e-08, 9.3132e-10, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 9.3132e-10, 3.2596e-09, -1.8626e-09, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0084, 0.0128, -0.0121, 0.0180, 0.0271, -0.0136, -0.0234, -0.0104, + -0.0303, -0.0297], device='cuda:0'), grad: tensor([ 5.5879e-08, 9.7789e-09, 2.9709e-07, -1.0151e-07, 7.6368e-08, + 3.2317e-07, -2.7986e-07, -4.4797e-07, 1.7555e-07, -8.8476e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 216.65, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.5147 re_mapping 0.0036 re_causal 0.0108 /// teacc 99.03 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.3276, -0.1604, -0.0948, ..., -0.0038, -0.0712, -0.0171], + [-0.1436, 0.1311, -0.0876, ..., 0.0183, -0.0636, -0.0819], + [ 0.1397, -0.1721, -0.1735, ..., -0.0644, 0.1685, -0.0337], + ..., + [-0.1833, -0.1011, 0.1307, ..., 0.0369, -0.2143, -0.0030], + [ 0.1059, -0.0050, -0.3129, ..., -0.0690, 0.1688, -0.0376], + [-0.2212, -0.1685, 0.0341, ..., -0.0167, -0.2391, -0.0726]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-09, 0.0000e+00], + [ 1.3970e-09, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 7.4040e-08, 0.0000e+00], + [-4.6566e-09, 5.1223e-09, 0.0000e+00, ..., 0.0000e+00, + 7.8697e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 3.2596e-09, 9.3132e-10, ..., 0.0000e+00, + 8.4285e-08, 0.0000e+00], + [-4.6566e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 2.8871e-08, 0.0000e+00], + [ 2.7940e-09, 4.2375e-08, -1.3970e-09, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0084, 0.0129, -0.0127, 0.0173, 0.0270, -0.0098, -0.0235, -0.0104, + -0.0304, -0.0296], device='cuda:0'), grad: tensor([ 2.9989e-07, 2.4168e-07, 3.5996e-07, -1.3039e-06, 1.5274e-07, + 6.2259e-07, 2.1420e-08, 2.3581e-06, 1.4994e-07, -2.9113e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 216.75, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4436 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.06 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.3276, -0.1607, -0.0948, ..., -0.0038, -0.0712, -0.0171], + [-0.1438, 0.1312, -0.0877, ..., 0.0183, -0.0643, -0.0819], + [ 0.1399, -0.1728, -0.1736, ..., -0.0645, 0.1693, -0.0337], + ..., + [-0.1834, -0.1010, 0.1307, ..., 0.0369, -0.2144, -0.0030], + [ 0.1060, -0.0049, -0.3129, ..., -0.0690, 0.1689, -0.0376], + [-0.2213, -0.1686, 0.0341, ..., -0.0167, -0.2394, -0.0726]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 2.3283e-09, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [-4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 8.8476e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 5.1223e-09, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, 0.0000e+00], + [-1.4435e-08, 5.6345e-08, 1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0084, 0.0126, -0.0124, 0.0172, 0.0270, -0.0093, -0.0236, -0.0103, + -0.0304, -0.0296], device='cuda:0'), grad: tensor([ 1.0524e-07, -2.3283e-09, -3.7253e-09, 9.7789e-09, -1.1828e-07, + 2.0489e-08, -1.6298e-08, 2.7940e-08, 1.9558e-08, -2.6077e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 216.70, cls_loss 0.0007 cls_loss_mapping 0.0009 cls_loss_causal 0.4471 re_mapping 0.0034 re_causal 0.0109 /// teacc 98.99 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.3276, -0.1608, -0.0948, ..., -0.0038, -0.0712, -0.0171], + [-0.1441, 0.1309, -0.0880, ..., 0.0183, -0.0666, -0.0819], + [ 0.1400, -0.1717, -0.1737, ..., -0.0645, 0.1715, -0.0337], + ..., + [-0.1834, -0.1010, 0.1307, ..., 0.0369, -0.2145, -0.0030], + [ 0.1060, -0.0049, -0.3131, ..., -0.0690, 0.1690, -0.0377], + [-0.2213, -0.1688, 0.0341, ..., -0.0167, -0.2394, -0.0726]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 4.1910e-09, 8.8476e-09, ..., 0.0000e+00, + 2.6077e-08, 0.0000e+00], + [ 4.1910e-09, -1.0477e-07, 4.6566e-10, ..., 0.0000e+00, + -7.9162e-09, 0.0000e+00], + [-4.1910e-09, 2.1886e-08, -2.0023e-08, ..., 0.0000e+00, + -5.3085e-08, 0.0000e+00], + ..., + [ 1.3970e-09, -2.8871e-08, 1.8626e-09, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [-2.3283e-08, 3.3062e-08, 3.2596e-09, ..., 0.0000e+00, + -1.2107e-08, 0.0000e+00], + [ 7.9162e-09, 7.2122e-06, 4.6566e-10, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0084, 0.0120, -0.0104, 0.0172, 0.0270, -0.0093, -0.0235, -0.0102, + -0.0304, -0.0296], device='cuda:0'), grad: tensor([ 7.6834e-08, 9.3132e-09, 8.2888e-08, 3.4040e-07, -6.8307e-05, + 4.0978e-08, -8.8476e-08, -1.0896e-06, 9.2201e-08, 6.8784e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 216.82, cls_loss 0.0009 cls_loss_mapping 0.0016 cls_loss_causal 0.4577 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.01 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.3277, -0.1621, -0.0949, ..., -0.0038, -0.0712, -0.0172], + [-0.1436, 0.1317, -0.0874, ..., 0.0183, -0.0658, -0.0819], + [ 0.1394, -0.1737, -0.1744, ..., -0.0645, 0.1709, -0.0337], + ..., + [-0.1834, -0.1011, 0.1307, ..., 0.0369, -0.2146, -0.0030], + [ 0.1057, -0.0058, -0.3132, ..., -0.0690, 0.1690, -0.0377], + [-0.2213, -0.1689, 0.0341, ..., -0.0167, -0.2395, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, 2.7008e-08, 2.3283e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.7963e-08, 1.2107e-08, 4.6566e-10, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 9.7789e-09, 2.1420e-08, 0.0000e+00, ..., 0.0000e+00, + -8.8476e-09, 0.0000e+00], + ..., + [ 1.3970e-09, 1.6298e-08, 8.8476e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-8.8476e-09, 1.6764e-08, 4.6566e-10, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + [ 5.1223e-09, 7.9162e-09, -5.0757e-08, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0084, 0.0125, -0.0121, 0.0172, 0.0270, -0.0092, -0.0234, -0.0102, + -0.0307, -0.0296], device='cuda:0'), grad: tensor([ 8.1956e-08, 1.0012e-07, 3.1665e-08, 4.7032e-08, 8.7498e-07, + 1.5972e-07, -9.7789e-07, 7.4552e-07, 6.0536e-09, -1.0673e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 216.60, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4609 re_mapping 0.0036 re_causal 0.0103 /// teacc 99.09 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.3277, -0.1644, -0.0949, ..., -0.0038, -0.0712, -0.0172], + [-0.1436, 0.1317, -0.0874, ..., 0.0183, -0.0663, -0.0819], + [ 0.1414, -0.1738, -0.1744, ..., -0.0645, 0.1727, -0.0337], + ..., + [-0.1848, -0.1012, 0.1307, ..., 0.0368, -0.2163, -0.0030], + [ 0.1057, -0.0058, -0.3137, ..., -0.0691, 0.1691, -0.0377], + [-0.2214, -0.1690, 0.0343, ..., -0.0167, -0.2392, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [-1.3970e-09, 9.3132e-10, 1.6764e-08, ..., 0.0000e+00, + -4.1910e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 1.8626e-09, -1.7695e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 4.6566e-10, 2.3283e-08, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0084, 0.0124, -0.0109, 0.0172, 0.0269, -0.0096, -0.0232, -0.0107, + -0.0308, -0.0294], device='cuda:0'), grad: tensor([-2.4568e-06, 1.3039e-08, 6.6916e-07, 3.7253e-09, -3.8650e-08, + 1.0245e-08, 1.3504e-08, -6.7381e-07, 1.6764e-08, 2.4717e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 216.72, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4646 re_mapping 0.0034 re_causal 0.0103 /// teacc 99.02 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.3277, -0.1648, -0.0949, ..., -0.0038, -0.0712, -0.0171], + [-0.1438, 0.1317, -0.0875, ..., 0.0182, -0.0666, -0.0819], + [ 0.1414, -0.1739, -0.1745, ..., -0.0647, 0.1726, -0.0337], + ..., + [-0.1848, -0.1012, 0.1307, ..., 0.0367, -0.2165, -0.0030], + [ 0.1057, -0.0059, -0.3139, ..., -0.0691, 0.1693, -0.0377], + [-0.2215, -0.1691, 0.0343, ..., -0.0167, -0.2397, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.0955e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, -1.4585e-06, 0.0000e+00, ..., 0.0000e+00, + 5.1223e-09, 0.0000e+00], + [-4.6566e-09, 6.8452e-07, 0.0000e+00, ..., 0.0000e+00, + -2.1420e-08, 0.0000e+00], + ..., + [ 2.3283e-09, 5.6624e-07, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 8.3819e-09, 6.9849e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 2.7940e-09, 5.3551e-08, -4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0084, 0.0123, -0.0111, 0.0174, 0.0270, -0.0105, -0.0230, -0.0107, + -0.0308, -0.0295], device='cuda:0'), grad: tensor([ 3.4925e-08, -3.3919e-06, 1.5683e-06, 9.8720e-08, 2.0629e-07, + -5.6811e-08, 3.2596e-08, 1.3215e-06, 3.8650e-08, 1.4342e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 216.65, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.4810 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.12 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.3278, -0.1657, -0.0949, ..., -0.0038, -0.0712, -0.0171], + [-0.1439, 0.1299, -0.0876, ..., 0.0191, -0.0668, -0.0819], + [ 0.1415, -0.1740, -0.1746, ..., -0.0649, 0.1729, -0.0337], + ..., + [-0.1849, -0.0992, 0.1308, ..., 0.0361, -0.2166, -0.0030], + [ 0.1058, -0.0060, -0.3139, ..., -0.0691, 0.1694, -0.0377], + [-0.2217, -0.1695, 0.0343, ..., -0.0167, -0.2400, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -1.0896e-07, 0.0000e+00], + [ 0.0000e+00, -3.8184e-08, 0.0000e+00, ..., 0.0000e+00, + 1.7602e-07, 0.0000e+00], + [-1.4901e-08, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + -2.2445e-07, 0.0000e+00], + ..., + [ 9.3132e-10, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 2.9802e-08, 0.0000e+00], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 2.2352e-08, 0.0000e+00], + [ 7.4506e-09, 1.9558e-08, 1.8626e-09, ..., 0.0000e+00, + 4.1910e-08, 0.0000e+00]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0085, 0.0099, -0.0114, 0.0172, 0.0268, -0.0109, -0.0230, -0.0076, + -0.0309, -0.0296], device='cuda:0'), grad: tensor([-5.1782e-07, 3.7812e-07, -4.1723e-07, 2.3842e-07, -8.5682e-08, + -1.4156e-07, 2.2352e-08, 8.5682e-08, 1.1642e-07, 3.2689e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 216.86, cls_loss 0.0009 cls_loss_mapping 0.0010 cls_loss_causal 0.4506 re_mapping 0.0035 re_causal 0.0101 /// teacc 99.02 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.3278, -0.1661, -0.0949, ..., -0.0038, -0.0712, -0.0171], + [-0.1440, 0.1294, -0.0876, ..., 0.0191, -0.0667, -0.0819], + [ 0.1417, -0.1741, -0.1746, ..., -0.0649, 0.1730, -0.0337], + ..., + [-0.1849, -0.0986, 0.1308, ..., 0.0361, -0.2167, -0.0030], + [ 0.1050, -0.0079, -0.3140, ..., -0.0691, 0.1691, -0.0379], + [-0.2211, -0.1696, 0.0343, ..., -0.0167, -0.2403, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.7742e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -4.9546e-07, 0.0000e+00, ..., 0.0000e+00, + -4.4703e-08, 0.0000e+00], + [ 0.0000e+00, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-9.3132e-10, 5.9605e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 8.0094e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0085, 0.0092, -0.0116, 0.0171, 0.0266, -0.0106, -0.0226, -0.0069, + -0.0320, -0.0296], device='cuda:0'), grad: tensor([ 4.5635e-08, -9.7603e-07, 3.2596e-08, 1.9558e-08, -2.9616e-07, + 3.2596e-08, 6.9477e-07, 3.1665e-08, 1.2480e-07, 2.8219e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 216.80, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4680 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.03 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.3278, -0.1669, -0.0949, ..., -0.0038, -0.0712, -0.0171], + [-0.1440, 0.1295, -0.0876, ..., 0.0191, -0.0666, -0.0819], + [ 0.1417, -0.1742, -0.1746, ..., -0.0649, 0.1731, -0.0337], + ..., + [-0.1850, -0.0986, 0.1308, ..., 0.0361, -0.2169, -0.0030], + [ 0.1050, -0.0080, -0.3140, ..., -0.0691, 0.1690, -0.0379], + [-0.2209, -0.1703, 0.0339, ..., -0.0167, -0.2404, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.1665e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [-8.3819e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -2.7008e-08, 0.0000e+00], + [ 1.8626e-09, 1.2107e-08, -6.5193e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0085, 0.0093, -0.0117, 0.0171, 0.0269, -0.0106, -0.0225, -0.0069, + -0.0321, -0.0299], device='cuda:0'), grad: tensor([ 1.1083e-07, 3.6787e-07, 3.1851e-07, 6.2585e-07, 1.9744e-06, + 1.5460e-07, 9.3132e-09, -4.2506e-06, -2.3283e-08, 7.1060e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 216.55, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4503 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.03 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.3278, -0.1680, -0.0949, ..., -0.0039, -0.0712, -0.0171], + [-0.1441, 0.1301, -0.0876, ..., 0.0191, -0.0659, -0.0819], + [ 0.1418, -0.1743, -0.1746, ..., -0.0650, 0.1734, -0.0337], + ..., + [-0.1850, -0.0990, 0.1308, ..., 0.0360, -0.2177, -0.0030], + [ 0.1050, -0.0082, -0.3140, ..., -0.0691, 0.1690, -0.0379], + [-0.2206, -0.1711, 0.0338, ..., -0.0167, -0.2412, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.8626e-08, -1.2014e-07, 0.0000e+00, ..., 0.0000e+00, + 4.0047e-08, 0.0000e+00], + [ 1.5832e-08, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 3.1665e-08, 0.0000e+00], + ..., + [ 2.7940e-08, 7.9162e-08, 0.0000e+00, ..., 0.0000e+00, + 6.3330e-08, 0.0000e+00], + [-4.3958e-07, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -9.8720e-07, 0.0000e+00], + [ 7.4506e-09, 1.5367e-07, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0085, 0.0094, -0.0118, 0.0171, 0.0270, -0.0105, -0.0224, -0.0070, + -0.0323, -0.0300], device='cuda:0'), grad: tensor([ 1.5832e-08, -1.2666e-07, 6.6124e-08, 3.1944e-07, 4.2319e-06, + 8.0187e-07, 2.7753e-07, 6.9663e-07, -2.6636e-07, -6.0499e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 216.56, cls_loss 0.0009 cls_loss_mapping 0.0012 cls_loss_causal 0.4645 re_mapping 0.0033 re_causal 0.0100 /// teacc 98.99 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.3279, -0.1693, -0.0949, ..., -0.0039, -0.0712, -0.0171], + [-0.1444, 0.1302, -0.0876, ..., 0.0191, -0.0660, -0.0819], + [ 0.1422, -0.1745, -0.1747, ..., -0.0650, 0.1735, -0.0337], + ..., + [-0.1851, -0.0990, 0.1307, ..., 0.0360, -0.2180, -0.0030], + [ 0.1053, -0.0083, -0.3140, ..., -0.0691, 0.1698, -0.0379], + [-0.2208, -0.1714, 0.0338, ..., -0.0167, -0.2420, -0.0731]], + device='cuda:0'), grad: tensor([[ 8.4750e-08, 6.6124e-08, 0.0000e+00, ..., 0.0000e+00, + 2.6450e-07, 0.0000e+00], + [ 6.5193e-09, -1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 2.0154e-06, 0.0000e+00], + [ 4.6566e-09, 4.6566e-09, -1.8626e-09, ..., 0.0000e+00, + 5.8673e-06, 0.0000e+00], + ..., + [ 1.8626e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 3.0518e-05, 0.0000e+00], + [-2.4214e-08, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 1.6652e-06, 0.0000e+00], + [ 1.1176e-08, 1.3039e-07, 0.0000e+00, ..., 0.0000e+00, + 1.8347e-07, 0.0000e+00]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0085, 0.0094, -0.0117, 0.0171, 0.0271, -0.0105, -0.0220, -0.0071, + -0.0322, -0.0301], device='cuda:0'), grad: tensor([ 2.2352e-06, 1.4693e-05, 4.2766e-05, -3.0804e-04, -2.5425e-07, + 1.1265e-05, -4.0606e-07, 2.2340e-04, 1.2457e-05, 1.8524e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 216.72, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4332 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.01 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.3279, -0.1695, -0.0949, ..., -0.0039, -0.0712, -0.0171], + [-0.1449, 0.1304, -0.0851, ..., 0.0191, -0.0670, -0.0819], + [ 0.1413, -0.1745, -0.1768, ..., -0.0650, 0.1722, -0.0337], + ..., + [-0.1852, -0.0992, 0.1307, ..., 0.0360, -0.2192, -0.0030], + [ 0.1059, -0.0084, -0.3140, ..., -0.0691, 0.1717, -0.0379], + [-0.2212, -0.1716, 0.0338, ..., -0.0167, -0.2422, -0.0731]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 1.8626e-09, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 2.1420e-08, 0.0000e+00], + [-2.4587e-07, 1.8626e-09, -1.3039e-08, ..., 0.0000e+00, + -4.0419e-07, 0.0000e+00], + ..., + [ 3.3528e-08, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + 6.9849e-08, 0.0000e+00], + [ 1.8533e-07, 8.3819e-09, 1.0245e-08, ..., 0.0000e+00, + 2.3935e-07, 0.0000e+00], + [ 9.3132e-10, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0085, 0.0094, -0.0126, 0.0174, 0.0271, -0.0124, -0.0205, -0.0072, + -0.0317, -0.0302], device='cuda:0'), grad: tensor([ 2.5146e-08, 1.1735e-07, -8.0001e-07, 1.7043e-07, 2.5146e-08, + 4.0978e-08, 1.8626e-08, 1.9483e-06, 5.3458e-07, -2.0973e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 216.59, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4674 re_mapping 0.0034 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.3279, -0.1697, -0.0950, ..., -0.0039, -0.0712, -0.0171], + [-0.1450, 0.1305, -0.0851, ..., 0.0191, -0.0670, -0.0819], + [ 0.1414, -0.1746, -0.1768, ..., -0.0651, 0.1724, -0.0337], + ..., + [-0.1852, -0.0993, 0.1307, ..., 0.0360, -0.2193, -0.0030], + [ 0.1059, -0.0085, -0.3141, ..., -0.0691, 0.1718, -0.0379], + [-0.2213, -0.1718, 0.0338, ..., -0.0167, -0.2417, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 1.5832e-08, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 7.4506e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 5.5879e-09, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-6.7987e-08, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + -7.7300e-08, 0.0000e+00], + [ 9.3132e-10, 5.4948e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0085, 0.0094, -0.0127, 0.0173, 0.0271, -0.0124, -0.0206, -0.0072, + -0.0318, -0.0301], device='cuda:0'), grad: tensor([-6.1467e-08, 3.7253e-08, 2.7008e-08, 6.5193e-09, -5.1782e-07, + 1.5646e-07, -1.8626e-08, 5.3085e-08, -1.2014e-07, 4.3958e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 216.62, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4786 re_mapping 0.0034 re_causal 0.0103 /// teacc 98.99 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.3280, -0.1703, -0.0951, ..., -0.0039, -0.0712, -0.0171], + [-0.1450, 0.1306, -0.0851, ..., 0.0190, -0.0671, -0.0819], + [ 0.1415, -0.1747, -0.1765, ..., -0.0651, 0.1727, -0.0337], + ..., + [-0.1853, -0.0993, 0.1307, ..., 0.0359, -0.2201, -0.0030], + [ 0.1060, -0.0086, -0.3142, ..., -0.0691, 0.1721, -0.0379], + [-0.2215, -0.1721, 0.0337, ..., -0.0167, -0.2421, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.6764e-07, 1.9092e-07, 0.0000e+00, ..., 0.0000e+00, + 8.6613e-08, 0.0000e+00], + [ 2.2352e-08, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 3.1665e-08, 0.0000e+00], + [ 7.4506e-09, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 1.4901e-08, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 4.9360e-08, 0.0000e+00], + [ 1.1269e-07, 1.8813e-07, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + [ 4.6566e-09, 3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0085, 0.0094, -0.0127, 0.0174, 0.0273, -0.0127, -0.0205, -0.0073, + -0.0318, -0.0302], device='cuda:0'), grad: tensor([ 6.7893e-07, 1.1548e-07, 5.6811e-08, 3.3062e-07, 2.5239e-07, + 5.0478e-07, -2.7288e-06, 2.2165e-07, 5.7556e-07, -1.3970e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 216.84, cls_loss 0.0010 cls_loss_mapping 0.0017 cls_loss_causal 0.4150 re_mapping 0.0035 re_causal 0.0101 /// teacc 98.98 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.3280, -0.1708, -0.0951, ..., -0.0039, -0.0712, -0.0171], + [-0.1452, 0.1306, -0.0851, ..., 0.0190, -0.0672, -0.0819], + [ 0.1414, -0.1747, -0.1765, ..., -0.0654, 0.1728, -0.0337], + ..., + [-0.1853, -0.0993, 0.1307, ..., 0.0358, -0.2202, -0.0030], + [ 0.1060, -0.0087, -0.3144, ..., -0.0691, 0.1724, -0.0379], + [-0.2218, -0.1725, 0.0337, ..., -0.0167, -0.2423, -0.0731]], + device='cuda:0'), grad: tensor([[ 8.7544e-08, 7.7300e-08, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 1.0245e-08, -4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + -2.5146e-08, 0.0000e+00], + [ 2.8871e-08, 6.0536e-08, 0.0000e+00, ..., 0.0000e+00, + -5.7742e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 1.5832e-08, 0.0000e+00, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00], + [ 1.3039e-08, 1.9558e-08, 0.0000e+00, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 5.5879e-09, 9.9652e-08, -9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0086, 0.0094, -0.0130, 0.0171, 0.0279, -0.0116, -0.0205, -0.0072, + -0.0318, -0.0307], device='cuda:0'), grad: tensor([ 3.9674e-07, 1.1921e-07, 3.8631e-06, 1.8440e-07, -4.7684e-07, + 8.1025e-08, -1.2144e-06, -4.1611e-06, 1.2945e-07, 1.0766e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 216.80, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4632 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.04 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.3281, -0.1714, -0.0951, ..., -0.0040, -0.0712, -0.0169], + [-0.1452, 0.1308, -0.0852, ..., 0.0224, -0.0672, -0.0819], + [ 0.1416, -0.1749, -0.1765, ..., -0.0689, 0.1732, -0.0337], + ..., + [-0.1854, -0.0994, 0.1308, ..., 0.0331, -0.2205, -0.0030], + [ 0.1066, -0.0086, -0.3147, ..., -0.0691, 0.1738, -0.0379], + [-0.2222, -0.1733, 0.0335, ..., -0.0168, -0.2434, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -3.3528e-08, 0.0000e+00, ..., 0.0000e+00, + -2.0117e-07, 0.0000e+00], + [ 0.0000e+00, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 1.8626e-09, 1.8626e-08, 0.0000e+00, ..., 0.0000e+00, + 9.1270e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.5390e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 3.1292e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0085, 0.0094, -0.0131, 0.0171, 0.0288, -0.0117, -0.0206, -0.0072, + -0.0314, -0.0317], device='cuda:0'), grad: tensor([-6.1095e-07, 9.6858e-08, 2.8126e-07, 6.3330e-08, -1.0580e-06, + 1.3039e-08, 2.3097e-07, 2.0489e-08, 9.8720e-08, 8.6613e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 216.81, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4807 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.01 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.3281, -0.1721, -0.0952, ..., -0.0040, -0.0712, -0.0169], + [-0.1456, 0.1310, -0.0850, ..., 0.0229, -0.0671, -0.0819], + [ 0.1415, -0.1753, -0.1767, ..., -0.0693, 0.1733, -0.0337], + ..., + [-0.1854, -0.0995, 0.1308, ..., 0.0327, -0.2206, -0.0030], + [ 0.1066, -0.0088, -0.3148, ..., -0.0691, 0.1741, -0.0379], + [-0.2224, -0.1735, 0.0334, ..., -0.0168, -0.2439, -0.0731]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.1176e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0085, 0.0094, -0.0133, 0.0174, 0.0291, -0.0128, -0.0205, -0.0072, + -0.0314, -0.0320], device='cuda:0'), grad: tensor([ 2.4214e-08, 5.5879e-09, 9.3132e-09, -5.5879e-08, -5.2527e-07, + 3.7253e-08, -2.9802e-08, 2.7940e-08, 5.5879e-09, 4.8988e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 216.76, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4698 re_mapping 0.0035 re_causal 0.0102 /// teacc 98.95 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.3281, -0.1732, -0.0953, ..., -0.0040, -0.0712, -0.0169], + [-0.1460, 0.1310, -0.0851, ..., 0.0229, -0.0673, -0.0819], + [ 0.1419, -0.1755, -0.1767, ..., -0.0694, 0.1740, -0.0337], + ..., + [-0.1855, -0.0994, 0.1308, ..., 0.0327, -0.2210, -0.0030], + [ 0.1063, -0.0092, -0.3154, ..., -0.0691, 0.1737, -0.0379], + [-0.2225, -0.1738, 0.0334, ..., -0.0168, -0.2440, -0.0731]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + [ 7.4506e-09, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [-9.1270e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.6950e-07, 0.0000e+00], + ..., + [ 4.8429e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 8.9407e-08, 0.0000e+00], + [ 8.3260e-07, 4.9174e-07, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 5.5879e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0085, 0.0093, -0.0129, 0.0174, 0.0292, -0.0127, -0.0205, -0.0071, + -0.0320, -0.0321], device='cuda:0'), grad: tensor([-5.9605e-08, 5.5879e-08, -4.9919e-07, 7.4506e-08, 2.7940e-08, + 2.1771e-05, -2.2903e-05, 2.5332e-07, 1.1772e-06, 5.5879e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 216.85, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4631 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.01 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.3281, -0.1735, -0.0953, ..., -0.0040, -0.0712, -0.0169], + [-0.1461, 0.1312, -0.0849, ..., 0.0229, -0.0671, -0.0819], + [ 0.1412, -0.1759, -0.1767, ..., -0.0696, 0.1733, -0.0337], + ..., + [-0.1855, -0.0995, 0.1308, ..., 0.0323, -0.2212, -0.0030], + [ 0.1064, -0.0096, -0.3156, ..., -0.0692, 0.1741, -0.0379], + [-0.2229, -0.1741, 0.0333, ..., -0.0168, -0.2460, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -2.0489e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.0489e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 6.5193e-08, -1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0085, 0.0094, -0.0136, 0.0175, 0.0292, -0.0132, -0.0200, -0.0072, + -0.0321, -0.0321], device='cuda:0'), grad: tensor([-4.6566e-08, -2.4214e-08, 1.3039e-08, 2.2352e-08, -3.0547e-07, + 2.7940e-08, 5.5879e-09, -2.2352e-08, 1.1176e-08, 3.0547e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 217.01, cls_loss 0.0010 cls_loss_mapping 0.0018 cls_loss_causal 0.4725 re_mapping 0.0036 re_causal 0.0104 /// teacc 99.02 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.3282, -0.1738, -0.0956, ..., -0.0040, -0.0712, -0.0169], + [-0.1464, 0.1314, -0.0851, ..., 0.0229, -0.0670, -0.0819], + [ 0.1413, -0.1765, -0.1766, ..., -0.0696, 0.1735, -0.0337], + ..., + [-0.1856, -0.0996, 0.1309, ..., 0.0322, -0.2213, -0.0030], + [ 0.1063, -0.0098, -0.3160, ..., -0.0692, 0.1743, -0.0379], + [-0.2231, -0.1746, 0.0333, ..., -0.0168, -0.2471, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 8.1956e-08, 0.0000e+00], + [ 0.0000e+00, -1.1921e-07, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -8.7358e-07, 0.0000e+00], + ..., + [ 5.5879e-09, 5.5879e-08, 0.0000e+00, ..., 0.0000e+00, + 1.7509e-07, 0.0000e+00], + [-7.4506e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 1.8626e-09, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0085, 0.0094, -0.0143, 0.0175, 0.0293, -0.0132, -0.0200, -0.0072, + -0.0323, -0.0322], device='cuda:0'), grad: tensor([ 2.3656e-07, -2.6263e-07, -3.4403e-06, 2.4550e-06, 1.1194e-06, + 5.5879e-08, 2.7940e-08, 8.1211e-07, 9.3132e-09, -1.0207e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 216.78, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4722 re_mapping 0.0035 re_causal 0.0108 /// teacc 98.90 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.3282, -0.1738, -0.0957, ..., -0.0040, -0.0712, -0.0169], + [-0.1464, 0.1334, -0.0851, ..., 0.0229, -0.0670, -0.0819], + [ 0.1414, -0.1766, -0.1766, ..., -0.0696, 0.1738, -0.0337], + ..., + [-0.1856, -0.1012, 0.1309, ..., 0.0322, -0.2215, -0.0030], + [ 0.1065, -0.0099, -0.3162, ..., -0.0692, 0.1747, -0.0379], + [-0.2237, -0.1747, 0.0333, ..., -0.0168, -0.2487, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, -1.6950e-07, 9.3132e-09, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 1.8626e-09, -1.3039e-08, 3.7253e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-5.9605e-08, 1.8626e-09, -1.0431e-07, ..., 0.0000e+00, + -1.6764e-07, 0.0000e+00], + ..., + [ 4.0978e-08, 1.6764e-08, 7.0781e-08, ..., 0.0000e+00, + 1.1921e-07, 0.0000e+00], + [-9.3132e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + [ 5.5879e-09, 2.9430e-07, -1.8626e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0085, 0.0098, -0.0142, 0.0176, 0.0293, -0.0133, -0.0200, -0.0076, + -0.0321, -0.0322], device='cuda:0'), grad: tensor([-7.2084e-07, 2.0489e-08, -7.6368e-07, 5.8115e-07, -1.0040e-06, + -4.9919e-07, 8.0280e-07, 5.3830e-07, 2.4214e-08, 1.0189e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 217.23, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4505 re_mapping 0.0033 re_causal 0.0100 /// teacc 98.97 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.3282, -0.1741, -0.0960, ..., -0.0040, -0.0712, -0.0169], + [-0.1466, 0.1336, -0.0853, ..., 0.0229, -0.0671, -0.0819], + [ 0.1416, -0.1768, -0.1768, ..., -0.0696, 0.1741, -0.0337], + ..., + [-0.1858, -0.1015, 0.1310, ..., 0.0322, -0.2218, -0.0030], + [ 0.1065, -0.0100, -0.3171, ..., -0.0692, 0.1748, -0.0379], + [-0.2238, -0.1747, 0.0334, ..., -0.0168, -0.2491, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.4901e-08, -7.4506e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [-1.8626e-09, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0085, 0.0099, -0.0142, 0.0176, 0.0292, -0.0134, -0.0199, -0.0076, + -0.0322, -0.0321], device='cuda:0'), grad: tensor([ 4.6939e-07, 2.0489e-08, 1.4529e-07, 1.1921e-07, 2.7381e-07, + 1.4529e-07, -1.4901e-08, 4.6566e-08, 7.8231e-08, -1.2908e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 216.90, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4778 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.05 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.3283, -0.1761, -0.0960, ..., -0.0040, -0.0712, -0.0169], + [-0.1466, 0.1353, -0.0860, ..., 0.0230, -0.0670, -0.0819], + [ 0.1416, -0.1772, -0.1771, ..., -0.0697, 0.1741, -0.0337], + ..., + [-0.1858, -0.1029, 0.1295, ..., 0.0322, -0.2221, -0.0030], + [ 0.1065, -0.0102, -0.3188, ..., -0.0692, 0.1749, -0.0379], + [-0.2240, -0.1750, 0.0342, ..., -0.0168, -0.2495, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.1851e-07, 0.0000e+00], + [ 0.0000e+00, -2.9802e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 2.4214e-08, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-4.0978e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 2.7940e-08, 9.3132e-09, -7.4506e-09, ..., 0.0000e+00, + 1.6019e-07, 0.0000e+00]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0085, 0.0103, -0.0144, 0.0176, 0.0291, -0.0137, -0.0198, -0.0081, + -0.0324, -0.0318], device='cuda:0'), grad: tensor([-1.0476e-05, -2.9802e-08, 2.5332e-07, 1.5460e-07, 7.6368e-08, + 2.4028e-07, 3.6266e-06, -1.6764e-08, 1.3858e-06, 4.7870e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 216.52, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4530 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.05 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.3283, -0.1770, -0.0960, ..., -0.0040, -0.0712, -0.0169], + [-0.1468, 0.1360, -0.0862, ..., 0.0230, -0.0671, -0.0819], + [ 0.1418, -0.1775, -0.1776, ..., -0.0698, 0.1748, -0.0337], + ..., + [-0.1859, -0.1036, 0.1295, ..., 0.0321, -0.2227, -0.0030], + [ 0.1065, -0.0105, -0.3219, ..., -0.0692, 0.1750, -0.0379], + [-0.2243, -0.1752, 0.0343, ..., -0.0168, -0.2499, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.3039e-08, -5.5879e-09, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + [ 1.8626e-09, -2.7940e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + ..., + [-9.3132e-09, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 2.0862e-07, 2.4214e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0085, 0.0105, -0.0144, 0.0177, 0.0291, -0.0140, -0.0194, -0.0083, + -0.0328, -0.0317], device='cuda:0'), grad: tensor([-9.1642e-07, 2.9802e-08, 2.7940e-08, 2.3842e-07, -8.4192e-07, + 3.3528e-08, 2.3469e-07, -2.3469e-07, 2.9244e-07, 1.1455e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 216.78, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4554 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.03 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.3283, -0.1770, -0.0962, ..., -0.0040, -0.0712, -0.0169], + [-0.1471, 0.1362, -0.0863, ..., 0.0230, -0.0673, -0.0819], + [ 0.1426, -0.1777, -0.1775, ..., -0.0698, 0.1757, -0.0337], + ..., + [-0.1862, -0.1038, 0.1297, ..., 0.0321, -0.2234, -0.0030], + [ 0.1070, -0.0104, -0.3215, ..., -0.0692, 0.1760, -0.0379], + [-0.2246, -0.1762, 0.0343, ..., -0.0168, -0.2513, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -1.4901e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 6.9290e-07, 3.7253e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0083, 0.0105, -0.0135, 0.0177, 0.0292, -0.0141, -0.0195, -0.0084, + -0.0325, -0.0320], device='cuda:0'), grad: tensor([ 3.7253e-09, 5.5879e-08, 7.4506e-09, 1.8626e-08, -2.4568e-06, + -1.3039e-08, 1.1176e-08, -3.5390e-08, 9.3132e-09, 2.3842e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 216.84, cls_loss 0.0007 cls_loss_mapping 0.0007 cls_loss_causal 0.4706 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.07 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.3284, -0.1772, -0.0962, ..., -0.0040, -0.0712, -0.0169], + [-0.1472, 0.1362, -0.0866, ..., 0.0230, -0.0673, -0.0819], + [ 0.1426, -0.1778, -0.1774, ..., -0.0698, 0.1759, -0.0337], + ..., + [-0.1862, -0.1038, 0.1306, ..., 0.0321, -0.2235, -0.0030], + [ 0.1081, -0.0096, -0.3217, ..., -0.0692, 0.1788, -0.0379], + [-0.2247, -0.1766, 0.0343, ..., -0.0168, -0.2514, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.4156e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.0245e-07, 0.0000e+00], + [ 0.0000e+00, -5.2154e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0083, 0.0105, -0.0137, 0.0175, 0.0291, -0.0139, -0.0195, -0.0082, + -0.0317, -0.0320], device='cuda:0'), grad: tensor([ 3.9116e-08, 5.5879e-08, 4.0233e-07, -1.1250e-06, 6.3516e-07, + 2.9989e-07, 3.7253e-09, 4.4703e-08, 3.0547e-07, -6.7241e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 217.00, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4328 re_mapping 0.0035 re_causal 0.0104 /// teacc 98.96 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.3284, -0.1794, -0.0962, ..., -0.0040, -0.0712, -0.0169], + [-0.1472, 0.1365, -0.0863, ..., 0.0230, -0.0673, -0.0819], + [ 0.1427, -0.1781, -0.1772, ..., -0.0698, 0.1767, -0.0337], + ..., + [-0.1863, -0.1041, 0.1307, ..., 0.0321, -0.2237, -0.0030], + [ 0.1081, -0.0097, -0.3220, ..., -0.0692, 0.1789, -0.0379], + [-0.2251, -0.1770, 0.0342, ..., -0.0168, -0.2524, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -3.4273e-07, 4.6566e-08, ..., 0.0000e+00, + -1.7136e-07, 0.0000e+00], + [ 0.0000e+00, 5.2154e-08, 3.7253e-09, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -5.4017e-08, ..., 0.0000e+00, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 7.8231e-08, -1.8626e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0083, 0.0105, -0.0135, 0.0186, 0.0290, -0.0139, -0.0195, -0.0092, + -0.0317, -0.0320], device='cuda:0'), grad: tensor([ 7.4506e-09, -5.6438e-07, 1.3039e-07, -1.5758e-06, -1.6019e-07, + 2.1141e-06, -1.3039e-08, -1.9185e-07, 5.5879e-08, 1.8254e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 216.84, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4824 re_mapping 0.0036 re_causal 0.0106 /// teacc 99.04 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.3284, -0.1795, -0.0964, ..., -0.0040, -0.0712, -0.0169], + [-0.1475, 0.1367, -0.0863, ..., 0.0230, -0.0674, -0.0819], + [ 0.1432, -0.1787, -0.1775, ..., -0.0698, 0.1767, -0.0337], + ..., + [-0.1863, -0.1042, 0.1307, ..., 0.0321, -0.2242, -0.0030], + [ 0.1083, -0.0097, -0.3225, ..., -0.0692, 0.1793, -0.0379], + [-0.2261, -0.1796, 0.0346, ..., -0.0168, -0.2545, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-09, 2.8685e-07, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 5.5879e-09, -2.3115e-06, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 3.1665e-08, 4.0978e-08, 1.8626e-09, ..., 0.0000e+00, + 3.9116e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 2.2072e-06, 2.4587e-07, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [-1.3970e-07, -1.4901e-07, 0.0000e+00, ..., 0.0000e+00, + -1.6764e-07, 0.0000e+00], + [ 5.4017e-08, 1.3970e-07, -5.4762e-07, ..., 0.0000e+00, + 6.8918e-08, 0.0000e+00]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0083, 0.0105, -0.0134, 0.0189, 0.0304, -0.0150, -0.0194, -0.0092, + -0.0315, -0.0335], device='cuda:0'), grad: tensor([ 2.0694e-06, -5.6699e-06, 2.4587e-07, -1.2480e-07, 5.5134e-07, + 5.8301e-07, 1.0058e-07, 7.3090e-06, -8.4750e-07, -4.2580e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 216.68, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4743 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.09 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.3284, -0.1797, -0.0946, ..., -0.0040, -0.0711, -0.0169], + [-0.1476, 0.1368, -0.0868, ..., 0.0230, -0.0678, -0.0819], + [ 0.1428, -0.1787, -0.1815, ..., -0.0698, 0.1764, -0.0337], + ..., + [-0.1864, -0.1043, 0.1307, ..., 0.0321, -0.2249, -0.0030], + [ 0.1082, -0.0098, -0.3253, ..., -0.0692, 0.1790, -0.0379], + [-0.2263, -0.1797, 0.0346, ..., -0.0168, -0.2576, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 3.7253e-09, -5.5879e-08, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [-1.1176e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-07, 0.0000e+00], + ..., + [ 3.7253e-09, 4.8056e-07, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-08, 0.0000e+00], + [ 0.0000e+00, 2.7940e-08, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -3.3770e-06, -9.4995e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0081, 0.0105, -0.0141, 0.0191, 0.0303, -0.0151, -0.0193, -0.0093, + -0.0319, -0.0336], device='cuda:0'), grad: tensor([ 4.6566e-08, -8.7544e-08, -2.2724e-07, 1.8068e-07, 1.9222e-05, + 8.1956e-08, -7.4506e-08, 3.0007e-06, 1.0617e-07, -2.2262e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 216.86, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4698 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.04 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.3284, -0.1799, -0.0955, ..., -0.0040, -0.0712, -0.0169], + [-0.1477, 0.1374, -0.0866, ..., 0.0230, -0.0677, -0.0819], + [ 0.1428, -0.1788, -0.1819, ..., -0.0698, 0.1763, -0.0337], + ..., + [-0.1864, -0.1048, 0.1305, ..., 0.0321, -0.2252, -0.0030], + [ 0.1079, -0.0103, -0.3264, ..., -0.0692, 0.1787, -0.0379], + [-0.2268, -0.1797, 0.0348, ..., -0.0168, -0.2583, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-5.0291e-08, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + -2.9802e-08, 0.0000e+00], + [ 1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0081, 0.0107, -0.0149, 0.0193, 0.0303, -0.0147, -0.0193, -0.0097, + -0.0325, -0.0334], device='cuda:0'), grad: tensor([-1.4529e-07, 2.6207e-06, 6.5193e-08, 7.9721e-07, 5.9791e-07, + -6.6496e-07, 1.4715e-07, -2.0206e-05, -7.8231e-08, 1.6853e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 216.72, cls_loss 0.0007 cls_loss_mapping 0.0009 cls_loss_causal 0.4584 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.02 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.3284, -0.1801, -0.0954, ..., -0.0040, -0.0712, -0.0169], + [-0.1481, 0.1382, -0.0871, ..., 0.0230, -0.0682, -0.0819], + [ 0.1432, -0.1790, -0.1814, ..., -0.0698, 0.1770, -0.0337], + ..., + [-0.1865, -0.1053, 0.1305, ..., 0.0321, -0.2257, -0.0030], + [ 0.1079, -0.0106, -0.3265, ..., -0.0692, 0.1788, -0.0379], + [-0.2269, -0.1791, 0.0348, ..., -0.0168, -0.2582, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.5193e-08, 6.3330e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-09, -1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 2.6077e-08, 2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 7.4506e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0081, 0.0109, -0.0149, 0.0192, 0.0296, -0.0144, -0.0193, -0.0098, + -0.0327, -0.0328], device='cuda:0'), grad: tensor([ 2.2352e-08, 1.3411e-07, 7.8417e-07, 0.0000e+00, 3.5390e-08, + 2.3283e-07, -4.8056e-07, -4.7870e-07, 7.0781e-08, -3.1292e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 216.87, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4610 re_mapping 0.0034 re_causal 0.0101 /// teacc 98.98 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.3284, -0.1802, -0.0954, ..., -0.0040, -0.0696, -0.0169], + [-0.1482, 0.1384, -0.0873, ..., 0.0230, -0.0686, -0.0819], + [ 0.1433, -0.1791, -0.1812, ..., -0.0698, 0.1757, -0.0337], + ..., + [-0.1866, -0.1054, 0.1304, ..., 0.0321, -0.2263, -0.0030], + [ 0.1079, -0.0107, -0.3269, ..., -0.0692, 0.1786, -0.0379], + [-0.2284, -0.1792, 0.0348, ..., -0.0168, -0.2613, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 3.7253e-09, -4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [-2.9802e-08, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + -9.3132e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 2.9802e-08, 0.0000e+00], + [ 3.7253e-09, 2.2352e-08, 0.0000e+00, ..., 0.0000e+00, + 4.9733e-07, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + -9.0897e-07, 0.0000e+00]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0073, 0.0109, -0.0157, 0.0193, 0.0295, -0.0145, -0.0193, -0.0099, + -0.0328, -0.0335], device='cuda:0'), grad: tensor([ 5.5879e-08, -2.0489e-08, -1.8440e-07, 6.5379e-07, 1.6373e-06, + 5.8673e-07, 2.9802e-08, 1.0058e-07, 3.1982e-06, -6.0722e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 216.90, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4603 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.11 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.3285, -0.1804, -0.0953, ..., -0.0040, -0.0693, -0.0169], + [-0.1482, 0.1385, -0.0880, ..., 0.0239, -0.0680, -0.0819], + [ 0.1434, -0.1796, -0.1811, ..., -0.0703, 0.1742, -0.0337], + ..., + [-0.1867, -0.1054, 0.1306, ..., 0.0313, -0.2269, -0.0030], + [ 0.1080, -0.0108, -0.3283, ..., -0.0692, 0.1784, -0.0379], + [-0.2287, -0.1783, 0.0374, ..., -0.0168, -0.2619, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.5879e-09, 3.7253e-08, 3.7253e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -2.7940e-08, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-3.1665e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-08, 0.0000e+00], + [ 0.0000e+00, 1.4529e-07, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0071, 0.0110, -0.0168, 0.0193, 0.0273, -0.0141, -0.0195, -0.0099, + -0.0335, -0.0314], device='cuda:0'), grad: tensor([ 1.0990e-07, 1.2107e-07, 6.5193e-08, 3.7253e-08, -7.8231e-08, + 1.5832e-07, 1.6764e-08, 1.0245e-07, -8.7544e-08, -4.5076e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 216.70, cls_loss 0.0009 cls_loss_mapping 0.0014 cls_loss_causal 0.4655 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.07 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.3285, -0.1806, -0.0954, ..., -0.0040, -0.0693, -0.0169], + [-0.1484, 0.1416, -0.0885, ..., 0.0239, -0.0679, -0.0819], + [ 0.1437, -0.1798, -0.1812, ..., -0.0704, 0.1747, -0.0337], + ..., + [-0.1870, -0.1084, 0.1304, ..., 0.0312, -0.2281, -0.0030], + [ 0.1081, -0.0108, -0.3310, ..., -0.0692, 0.1786, -0.0379], + [-0.2289, -0.1783, 0.0376, ..., -0.0168, -0.2619, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 4.6566e-08, 0.0000e+00, 1.1176e-08, ..., 0.0000e+00, + 6.6496e-07, 0.0000e+00], + [-5.2154e-08, 0.0000e+00, -2.6077e-08, ..., 0.0000e+00, + -7.0408e-07, 0.0000e+00], + ..., + [ 7.4506e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 2.4214e-08, 0.0000e+00], + [-1.1176e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.4901e-08, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0070, 0.0131, -0.0165, 0.0192, 0.0271, -0.0136, -0.0198, -0.0117, + -0.0335, -0.0313], device='cuda:0'), grad: tensor([ 1.4901e-08, 1.2144e-06, -1.2852e-06, 2.9802e-08, -7.4506e-09, + 2.6077e-08, -1.3039e-08, 1.4901e-08, -3.9116e-08, 3.9116e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 216.60, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4729 re_mapping 0.0031 re_causal 0.0096 /// teacc 99.03 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.3285, -0.1809, -0.0954, ..., -0.0040, -0.0693, -0.0169], + [-0.1484, 0.1417, -0.0893, ..., 0.0246, -0.0678, -0.0819], + [ 0.1431, -0.1800, -0.1823, ..., -0.0709, 0.1736, -0.0337], + ..., + [-0.1870, -0.1084, 0.1313, ..., 0.0308, -0.2271, -0.0030], + [ 0.1085, -0.0112, -0.3313, ..., -0.0693, 0.1792, -0.0379], + [-0.2297, -0.1784, 0.0376, ..., -0.0168, -0.2619, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, -2.6077e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-5.5879e-09, 3.7253e-09, -1.8626e-09, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0070, 0.0131, -0.0183, 0.0190, 0.0271, -0.0131, -0.0197, -0.0116, + -0.0332, -0.0313], device='cuda:0'), grad: tensor([ 1.8626e-09, -2.2352e-08, -7.4506e-09, -7.3835e-06, 1.8440e-07, + 7.2718e-06, 1.1176e-08, 3.1665e-08, 1.3411e-07, -2.4214e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 216.71, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4770 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.01 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.3286, -0.1811, -0.0956, ..., -0.0040, -0.0693, -0.0169], + [-0.1471, 0.1421, -0.0892, ..., 0.0246, -0.0677, -0.0819], + [ 0.1432, -0.1826, -0.1836, ..., -0.0709, 0.1719, -0.0337], + ..., + [-0.1873, -0.1085, 0.1312, ..., 0.0308, -0.2279, -0.0030], + [ 0.1085, -0.0114, -0.3315, ..., -0.0693, 0.1795, -0.0379], + [-0.2302, -0.1785, 0.0377, ..., -0.0168, -0.2620, -0.0731]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 7.4506e-09, -3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 5.5879e-09, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-8.3819e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -1.7695e-08, 0.0000e+00], + ..., + [ 4.6566e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 2.2352e-08, 2.7940e-08, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, -9.3132e-10, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0070, 0.0133, -0.0205, 0.0194, 0.0270, -0.0137, -0.0199, -0.0117, + -0.0333, -0.0312], device='cuda:0'), grad: tensor([-1.2051e-06, 6.5193e-08, -9.3132e-09, 1.2387e-07, -4.1910e-08, + 8.3819e-08, 8.4098e-07, -7.6368e-08, 8.6613e-08, 1.1921e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 216.53, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4346 re_mapping 0.0032 re_causal 0.0097 /// teacc 98.97 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.3286, -0.1802, -0.0955, ..., -0.0040, -0.0693, -0.0169], + [-0.1471, 0.1422, -0.0898, ..., 0.0246, -0.0679, -0.0819], + [ 0.1431, -0.1829, -0.1840, ..., -0.0709, 0.1719, -0.0337], + ..., + [-0.1872, -0.1085, 0.1314, ..., 0.0308, -0.2279, -0.0030], + [ 0.1089, -0.0116, -0.3316, ..., -0.0693, 0.1802, -0.0379], + [-0.2317, -0.1787, 0.0377, ..., -0.0168, -0.2620, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.0245e-08, -2.7940e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [-5.6811e-08, 2.7940e-09, -9.3132e-10, ..., 0.0000e+00, + -7.9162e-08, 0.0000e+00], + ..., + [ 2.9802e-08, 1.1176e-08, 9.3132e-10, ..., 0.0000e+00, + 4.1910e-08, 0.0000e+00], + [-3.9116e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -6.0536e-08, 0.0000e+00], + [ 3.7253e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0069, 0.0133, -0.0210, 0.0194, 0.0270, -0.0163, -0.0174, -0.0116, + -0.0330, -0.0313], device='cuda:0'), grad: tensor([-9.0338e-08, -3.7253e-09, -1.7043e-07, 1.4715e-07, 3.0641e-07, + 4.9360e-08, 4.8429e-08, 1.7602e-07, -1.7043e-07, -2.8405e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 216.96, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4405 re_mapping 0.0033 re_causal 0.0099 /// teacc 98.98 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.3286, -0.1803, -0.0956, ..., -0.0041, -0.0693, -0.0169], + [-0.1472, 0.1424, -0.0896, ..., 0.0246, -0.0681, -0.0819], + [ 0.1434, -0.1831, -0.1840, ..., -0.0709, 0.1720, -0.0337], + ..., + [-0.1874, -0.1085, 0.1314, ..., 0.0308, -0.2276, -0.0030], + [ 0.1089, -0.0116, -0.3317, ..., -0.0693, 0.1803, -0.0379], + [-0.2319, -0.1787, 0.0377, ..., -0.0168, -0.2620, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.6356e-07, 4.6007e-07, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 7.4506e-09, 3.1665e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3970e-08, -1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.8871e-08, 6.4261e-08, 0.0000e+00, ..., 0.0000e+00, + -3.6322e-08, 0.0000e+00], + [ 9.3132e-10, 3.8836e-07, -9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0069, 0.0133, -0.0217, 0.0194, 0.0270, -0.0162, -0.0174, -0.0116, + -0.0330, -0.0313], device='cuda:0'), grad: tensor([ 1.4072e-06, 9.9652e-08, 1.3039e-08, 8.7544e-08, -1.4165e-06, + 5.4948e-08, -1.5860e-06, 2.7008e-08, 6.5193e-08, 1.2564e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 216.74, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4501 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.3287, -0.1807, -0.0957, ..., -0.0041, -0.0693, -0.0169], + [-0.1475, 0.1424, -0.0900, ..., 0.0246, -0.0683, -0.0819], + [ 0.1436, -0.1832, -0.1851, ..., -0.0709, 0.1709, -0.0337], + ..., + [-0.1876, -0.1086, 0.1318, ..., 0.0308, -0.2259, -0.0030], + [ 0.1092, -0.0115, -0.3318, ..., -0.0693, 0.1806, -0.0379], + [-0.2330, -0.1804, 0.0372, ..., -0.0168, -0.2621, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 1.8626e-09, -8.3819e-09, ..., 0.0000e+00, + -5.0850e-07, 0.0000e+00], + [ 2.7940e-09, -2.4214e-07, 9.3132e-10, ..., 0.0000e+00, + 1.5832e-08, 0.0000e+00], + [ 0.0000e+00, 2.7940e-08, 1.8626e-09, ..., 0.0000e+00, + 7.7300e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 1.4063e-07, 2.7940e-09, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00], + [-4.1910e-08, 2.3283e-08, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + [ 1.1176e-08, 8.0653e-07, 1.1735e-07, ..., 0.0000e+00, + 3.9488e-07, 0.0000e+00]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0067, 0.0133, -0.0234, 0.0194, 0.0286, -0.0162, -0.0175, -0.0114, + -0.0327, -0.0329], device='cuda:0'), grad: tensor([-5.9605e-06, -5.1409e-07, 8.0187e-07, -3.0827e-07, -2.1551e-06, + 4.7870e-07, 2.6170e-07, -1.1977e-06, 2.6077e-08, 8.5533e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 216.88, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4696 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.06 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.3287, -0.1809, -0.0956, ..., -0.0041, -0.0693, -0.0169], + [-0.1476, 0.1425, -0.0928, ..., 0.0247, -0.0683, -0.0819], + [ 0.1438, -0.1833, -0.1851, ..., -0.0710, 0.1711, -0.0337], + ..., + [-0.1877, -0.1086, 0.1319, ..., 0.0307, -0.2260, -0.0030], + [ 0.1093, -0.0117, -0.3319, ..., -0.0693, 0.1807, -0.0380], + [-0.2333, -0.1809, 0.0372, ..., -0.0168, -0.2621, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, -1.5832e-08, 0.0000e+00, ..., 0.0000e+00, + -3.0734e-08, 0.0000e+00], + [ 0.0000e+00, -1.9558e-08, 0.0000e+00, ..., 0.0000e+00, + 2.1420e-08, 0.0000e+00], + [-9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 2.3283e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 9.4064e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0064, 0.0134, -0.0234, 0.0193, 0.0292, -0.0161, -0.0175, -0.0114, + -0.0327, -0.0337], device='cuda:0'), grad: tensor([-1.9185e-07, 6.2399e-08, -5.5879e-09, 2.2352e-08, -1.7136e-07, + 7.4506e-09, 4.1910e-08, 4.1910e-08, 2.2352e-08, 1.7602e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 216.55, cls_loss 0.0008 cls_loss_mapping 0.0014 cls_loss_causal 0.4436 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.05 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.3287, -0.1811, -0.0956, ..., -0.0041, -0.0693, -0.0169], + [-0.1477, 0.1427, -0.0935, ..., 0.0247, -0.0681, -0.0819], + [ 0.1446, -0.1834, -0.1850, ..., -0.0710, 0.1719, -0.0337], + ..., + [-0.1881, -0.1087, 0.1319, ..., 0.0307, -0.2267, -0.0030], + [ 0.1095, -0.0122, -0.3321, ..., -0.0693, 0.1810, -0.0380], + [-0.2337, -0.1810, 0.0372, ..., -0.0168, -0.2621, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0063, 0.0134, -0.0230, 0.0192, 0.0292, -0.0160, -0.0177, -0.0115, + -0.0331, -0.0337], device='cuda:0'), grad: tensor([ 1.5739e-07, 1.7602e-07, 1.0245e-08, 7.2643e-08, 4.8243e-06, + 4.8429e-08, -3.9116e-07, 7.6741e-06, 4.4052e-07, -1.3039e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 216.63, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.4597 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.02 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.3288, -0.1818, -0.0956, ..., -0.0041, -0.0693, -0.0169], + [-0.1480, 0.1431, -0.0938, ..., 0.0247, -0.0682, -0.0819], + [ 0.1449, -0.1837, -0.1851, ..., -0.0710, 0.1723, -0.0337], + ..., + [-0.1890, -0.1088, 0.1322, ..., 0.0307, -0.2267, -0.0030], + [ 0.1098, -0.0145, -0.3325, ..., -0.0693, 0.1807, -0.0380], + [-0.2339, -0.1816, 0.0372, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-6.5193e-09, 0.0000e+00, -2.7940e-09, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 5.5879e-09, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-4.6566e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + [ 0.0000e+00, 2.1420e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0063, 0.0136, -0.0242, 0.0190, 0.0294, -0.0160, -0.0177, -0.0106, + -0.0353, -0.0344], device='cuda:0'), grad: tensor([-2.6077e-08, 2.2352e-08, -2.7008e-08, 2.8871e-08, -2.2352e-08, + 8.3819e-09, 2.6077e-08, 1.0245e-08, 8.3819e-09, -2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 216.60, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4484 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.06 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.3288, -0.1821, -0.0956, ..., -0.0041, -0.0693, -0.0169], + [-0.1480, 0.1433, -0.0940, ..., 0.0247, -0.0660, -0.0819], + [ 0.1454, -0.1840, -0.1850, ..., -0.0710, 0.1720, -0.0337], + ..., + [-0.1892, -0.1090, 0.1322, ..., 0.0307, -0.2272, -0.0030], + [ 0.1097, -0.0146, -0.3328, ..., -0.0693, 0.1805, -0.0380], + [-0.2339, -0.1817, 0.0372, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 4.6566e-09, 9.3132e-10, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.9558e-08, 2.7940e-09, 9.3132e-10, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [-1.7695e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -2.8871e-08, 0.0000e+00], + ..., + [-9.1270e-08, -3.7253e-09, 3.7253e-09, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 1.1642e-07, 3.2596e-08, 9.3132e-10, ..., 0.0000e+00, + 4.1910e-08, 0.0000e+00], + [ 1.8626e-09, -4.6566e-09, -4.8429e-08, ..., 0.0000e+00, + 1.9558e-08, 0.0000e+00]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0063, 0.0137, -0.0247, 0.0192, 0.0294, -0.0164, -0.0178, -0.0108, + -0.0360, -0.0342], device='cuda:0'), grad: tensor([ 8.2888e-08, 1.7416e-07, -1.1176e-08, -3.0454e-07, 1.5320e-06, + 2.5425e-07, -1.4715e-07, -3.3807e-07, 6.3144e-07, -1.8636e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 216.81, cls_loss 0.0009 cls_loss_mapping 0.0019 cls_loss_causal 0.4718 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.12 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.3288, -0.1812, -0.0960, ..., -0.0041, -0.0693, -0.0169], + [-0.1484, 0.1435, -0.0943, ..., 0.0247, -0.0655, -0.0819], + [ 0.1477, -0.1856, -0.1850, ..., -0.0710, 0.1730, -0.0337], + ..., + [-0.1893, -0.1090, 0.1326, ..., 0.0307, -0.2275, -0.0030], + [ 0.1097, -0.0146, -0.3333, ..., -0.0693, 0.1804, -0.0380], + [-0.2340, -0.1817, 0.0372, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.3970e-08, -6.5193e-09, 4.6566e-09, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [-2.2817e-07, 9.3132e-10, 1.8626e-08, ..., 0.0000e+00, + -3.3807e-07, 0.0000e+00], + ..., + [ 1.6019e-07, 9.3132e-10, -2.6077e-08, ..., 0.0000e+00, + 2.4401e-07, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0061, 0.0138, -0.0242, 0.0191, 0.0292, -0.0164, -0.0183, -0.0105, + -0.0364, -0.0344], device='cuda:0'), grad: tensor([ 4.7497e-08, 1.8254e-07, -1.0459e-06, 6.7055e-08, 2.3190e-07, + 5.2154e-08, 1.9558e-08, 3.6415e-07, 4.3772e-08, 4.5635e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 216.51, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4516 re_mapping 0.0034 re_causal 0.0099 /// teacc 99.11 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.3289, -0.1816, -0.0960, ..., -0.0041, -0.0693, -0.0169], + [-0.1486, 0.1439, -0.0944, ..., 0.0247, -0.0657, -0.0819], + [ 0.1490, -0.1858, -0.1851, ..., -0.0710, 0.1741, -0.0337], + ..., + [-0.1894, -0.1090, 0.1327, ..., 0.0307, -0.2276, -0.0030], + [ 0.1097, -0.0147, -0.3333, ..., -0.0693, 0.1804, -0.0380], + [-0.2341, -0.1819, 0.0372, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.9558e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-9.3132e-10, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 1.3039e-08, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.2154e-08, 2.7940e-09, -9.3132e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0062, 0.0139, -0.0235, 0.0191, 0.0290, -0.0169, -0.0184, -0.0105, + -0.0366, -0.0341], device='cuda:0'), grad: tensor([ 2.2352e-08, 3.7253e-09, 1.6764e-08, 8.9966e-07, 1.2657e-06, + 3.6322e-08, 4.0978e-08, 6.7055e-08, 8.8476e-08, -2.4531e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 216.94, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4340 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.01 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.3289, -0.1826, -0.0960, ..., -0.0041, -0.0693, -0.0169], + [-0.1487, 0.1439, -0.0943, ..., 0.0247, -0.0657, -0.0819], + [ 0.1491, -0.1860, -0.1851, ..., -0.0710, 0.1742, -0.0337], + ..., + [-0.1896, -0.1091, 0.1328, ..., 0.0307, -0.2279, -0.0030], + [ 0.1097, -0.0148, -0.3333, ..., -0.0693, 0.1806, -0.0380], + [-0.2342, -0.1823, 0.0372, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 5.7556e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0061, 0.0139, -0.0235, 0.0186, 0.0291, -0.0157, -0.0184, -0.0105, + -0.0366, -0.0342], device='cuda:0'), grad: tensor([-1.0151e-07, 5.4948e-08, -6.5193e-09, 1.1083e-07, -1.2731e-06, + -9.1270e-08, 8.6613e-08, 5.0291e-08, 8.3819e-09, 1.1548e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 216.84, cls_loss 0.0007 cls_loss_mapping 0.0011 cls_loss_causal 0.4479 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.3290, -0.1824, -0.0960, ..., -0.0041, -0.0693, -0.0169], + [-0.1500, 0.1437, -0.0945, ..., 0.0247, -0.0659, -0.0819], + [ 0.1495, -0.1869, -0.1851, ..., -0.0710, 0.1744, -0.0337], + ..., + [-0.1889, -0.1087, 0.1328, ..., 0.0307, -0.2282, -0.0030], + [ 0.1097, -0.0150, -0.3334, ..., -0.0693, 0.1808, -0.0380], + [-0.2344, -0.1826, 0.0372, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 1.0151e-07, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 5.7090e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4398e-06, 0.0000e+00], + [-1.8626e-09, 9.4995e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 1.8533e-07, -1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0062, 0.0136, -0.0238, 0.0187, 0.0292, -0.0158, -0.0184, -0.0102, + -0.0367, -0.0342], device='cuda:0'), grad: tensor([ 6.2399e-08, 2.9337e-07, 6.1207e-06, -2.2426e-05, -9.5740e-07, + 6.3609e-07, 3.1665e-08, 1.5527e-05, 2.6263e-07, 4.5821e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 216.79, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4551 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.12 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.3291, -0.1828, -0.0963, ..., -0.0041, -0.0693, -0.0169], + [-0.1502, 0.1437, -0.0949, ..., 0.0247, -0.0659, -0.0819], + [ 0.1500, -0.1875, -0.1849, ..., -0.0710, 0.1748, -0.0337], + ..., + [-0.1892, -0.1087, 0.1326, ..., 0.0307, -0.2287, -0.0030], + [ 0.1097, -0.0151, -0.3341, ..., -0.0693, 0.1808, -0.0380], + [-0.2347, -0.1832, 0.0373, ..., -0.0168, -0.2622, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 2.7940e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 1.7695e-08, 2.7940e-09, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.0245e-08, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-5.4017e-08, 2.1420e-08, -1.3039e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0062, 0.0136, -0.0238, 0.0188, 0.0294, -0.0158, -0.0186, -0.0103, + -0.0368, -0.0343], device='cuda:0'), grad: tensor([ 1.0151e-07, 3.6322e-08, 3.7253e-09, 1.2759e-07, 6.8918e-08, + -2.0675e-07, 9.7789e-08, 1.1642e-07, 5.5879e-08, -4.0699e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 216.56, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4790 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.08 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.3293, -0.1856, -0.0964, ..., -0.0041, -0.0693, -0.0169], + [-0.1499, 0.1439, -0.0947, ..., 0.0247, -0.0652, -0.0819], + [ 0.1503, -0.1877, -0.1838, ..., -0.0710, 0.1755, -0.0337], + ..., + [-0.1894, -0.1088, 0.1325, ..., 0.0307, -0.2291, -0.0030], + [ 0.1097, -0.0160, -0.3368, ..., -0.0693, 0.1801, -0.0380], + [-0.2355, -0.1836, 0.0373, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 3.7253e-09, 1.1176e-08, 9.3132e-10, ..., 0.0000e+00, + 2.1420e-08, 0.0000e+00], + [-3.7253e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 7.3574e-08, -3.1665e-08, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0063, 0.0136, -0.0235, 0.0190, 0.0294, -0.0167, -0.0180, -0.0103, + -0.0375, -0.0345], device='cuda:0'), grad: tensor([ 1.3039e-08, 7.8231e-08, 3.0734e-08, -4.7311e-07, 5.8394e-07, + 2.4959e-07, 2.6077e-08, 1.0338e-07, 3.6322e-08, -6.3423e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 216.86, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4652 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.06 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.3293, -0.1858, -0.0965, ..., -0.0041, -0.0693, -0.0169], + [-0.1533, 0.1438, -0.0949, ..., 0.0247, -0.0675, -0.0819], + [ 0.1491, -0.1883, -0.1838, ..., -0.0710, 0.1749, -0.0337], + ..., + [-0.1882, -0.1087, 0.1330, ..., 0.0307, -0.2291, -0.0030], + [ 0.1120, -0.0157, -0.3375, ..., -0.0693, 0.1826, -0.0380], + [-0.2357, -0.1838, 0.0373, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.4901e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 3.6322e-08, 0.0000e+00], + [-2.7008e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.2841e-08, 0.0000e+00], + ..., + [ 6.5193e-09, 2.7940e-09, 1.8626e-09, ..., 0.0000e+00, + 4.4517e-07, 0.0000e+00], + [-4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + [ 0.0000e+00, 3.1665e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0065, 0.0133, -0.0242, 0.0191, 0.0294, -0.0163, -0.0180, -0.0101, + -0.0353, -0.0345], device='cuda:0'), grad: tensor([-5.8673e-08, 1.2107e-07, 2.5053e-07, -5.2229e-06, -6.0536e-08, + 1.3039e-07, 6.6124e-08, 4.7274e-06, -4.6566e-09, 5.4017e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 216.57, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4677 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.02 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.3294, -0.1859, -0.0965, ..., -0.0041, -0.0693, -0.0169], + [-0.1541, 0.1443, -0.0944, ..., 0.0246, -0.0686, -0.0819], + [ 0.1487, -0.1886, -0.1838, ..., -0.0711, 0.1749, -0.0337], + ..., + [-0.1888, -0.1091, 0.1329, ..., 0.0307, -0.2300, -0.0030], + [ 0.1133, -0.0154, -0.3376, ..., -0.0693, 0.1842, -0.0380], + [-0.2359, -0.1839, 0.0373, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.2107e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 2.6077e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-09, -3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0065, 0.0133, -0.0241, 0.0191, 0.0294, -0.0164, -0.0181, -0.0103, + -0.0339, -0.0346], device='cuda:0'), grad: tensor([ 3.9116e-08, -2.6263e-07, 5.5879e-09, -2.2352e-08, 7.3016e-07, + 6.8918e-08, -1.6950e-07, 1.1921e-07, 4.6566e-08, -5.7369e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 216.67, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4849 re_mapping 0.0034 re_causal 0.0103 /// teacc 98.98 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.3294, -0.1857, -0.0965, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1446, -0.0945, ..., 0.0246, -0.0687, -0.0819], + [ 0.1489, -0.1886, -0.1839, ..., -0.0711, 0.1756, -0.0337], + ..., + [-0.1889, -0.1093, 0.1330, ..., 0.0307, -0.2305, -0.0030], + [ 0.1134, -0.0155, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2363, -0.1809, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 1.8626e-09, 4.6566e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0063, 0.0134, -0.0236, 0.0191, 0.0265, -0.0165, -0.0182, -0.0104, + -0.0340, -0.0317], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.3039e-08, -2.6077e-08, 2.2352e-08, -1.2666e-07, + -1.4342e-07, 1.2480e-07, 1.8626e-08, -1.8626e-09, 1.1921e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 216.76, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4528 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.00 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.3294, -0.1857, -0.0965, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1446, -0.0945, ..., 0.0246, -0.0687, -0.0819], + [ 0.1490, -0.1886, -0.1839, ..., -0.0711, 0.1756, -0.0337], + ..., + [-0.1890, -0.1093, 0.1330, ..., 0.0307, -0.2306, -0.0030], + [ 0.1133, -0.0155, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2363, -0.1809, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0063, 0.0134, -0.0235, 0.0191, 0.0265, -0.0165, -0.0182, -0.0104, + -0.0340, -0.0317], device='cuda:0'), grad: tensor([-6.5193e-08, -1.8626e-09, 5.5879e-09, 1.1176e-08, -1.8626e-09, + -3.7253e-09, 2.2352e-08, 3.7253e-09, 7.4506e-09, 2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 217.01, cls_loss 0.0006 cls_loss_mapping 0.0006 cls_loss_causal 0.4442 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.02 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.3294, -0.1858, -0.0965, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1446, -0.0945, ..., 0.0246, -0.0687, -0.0819], + [ 0.1490, -0.1886, -0.1839, ..., -0.0711, 0.1757, -0.0337], + ..., + [-0.1890, -0.1093, 0.1330, ..., 0.0307, -0.2306, -0.0030], + [ 0.1133, -0.0155, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2363, -0.1809, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.5087e-07, 2.4587e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0063, 0.0134, -0.0235, 0.0191, 0.0265, -0.0165, -0.0182, -0.0104, + -0.0340, -0.0317], device='cuda:0'), grad: tensor([ 7.4506e-09, 4.0978e-08, 1.8626e-09, 0.0000e+00, -9.9465e-07, + 1.1176e-08, 7.4506e-09, -2.9616e-07, 1.1176e-08, 1.2163e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 216.88, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4326 re_mapping 0.0030 re_causal 0.0095 /// teacc 98.99 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.3294, -0.1859, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1446, -0.0945, ..., 0.0246, -0.0687, -0.0819], + [ 0.1490, -0.1886, -0.1839, ..., -0.0711, 0.1758, -0.0337], + ..., + [-0.1890, -0.1093, 0.1330, ..., 0.0307, -0.2307, -0.0030], + [ 0.1133, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2363, -0.1809, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0063, 0.0134, -0.0234, 0.0191, 0.0265, -0.0165, -0.0182, -0.0104, + -0.0341, -0.0317], device='cuda:0'), grad: tensor([-5.2340e-07, 1.6764e-08, -5.7742e-08, 2.7940e-08, 5.2154e-08, + 0.0000e+00, 4.5821e-07, 2.6077e-08, 1.3039e-08, -1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 216.80, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4317 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.01 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.3294, -0.1859, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1446, -0.0945, ..., 0.0246, -0.0687, -0.0819], + [ 0.1491, -0.1887, -0.1839, ..., -0.0711, 0.1759, -0.0337], + ..., + [-0.1890, -0.1093, 0.1330, ..., 0.0307, -0.2308, -0.0030], + [ 0.1133, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2364, -0.1809, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.4995e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.9616e-07, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-08, 0.0000e+00], + [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + [ 0.0000e+00, 2.1048e-07, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0063, 0.0134, -0.0233, 0.0191, 0.0265, -0.0165, -0.0182, -0.0104, + -0.0341, -0.0317], device='cuda:0'), grad: tensor([ 2.8312e-07, 1.9372e-07, -5.6438e-07, 2.1234e-07, -4.9546e-07, + 1.7509e-07, 5.5879e-09, 1.1921e-07, -1.6764e-08, 9.4995e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 216.90, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4228 re_mapping 0.0029 re_causal 0.0094 /// teacc 99.01 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.3294, -0.1859, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1447, -0.0944, ..., 0.0246, -0.0687, -0.0819], + [ 0.1491, -0.1887, -0.1839, ..., -0.0711, 0.1760, -0.0337], + ..., + [-0.1891, -0.1093, 0.1330, ..., 0.0307, -0.2309, -0.0030], + [ 0.1133, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2364, -0.1809, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 7.4506e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0063, 0.0135, -0.0233, 0.0191, 0.0265, -0.0165, -0.0182, -0.0104, + -0.0341, -0.0317], device='cuda:0'), grad: tensor([ 3.7253e-09, 9.3132e-09, -2.0489e-08, 3.7253e-08, 2.9802e-08, + -2.9802e-08, 1.1176e-08, -2.0489e-08, 4.0978e-08, -5.7742e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 216.92, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4601 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.01 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.3294, -0.1859, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1447, -0.0944, ..., 0.0246, -0.0687, -0.0819], + [ 0.1492, -0.1887, -0.1839, ..., -0.0711, 0.1761, -0.0337], + ..., + [-0.1891, -0.1093, 0.1330, ..., 0.0307, -0.2310, -0.0030], + [ 0.1133, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2365, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-5.0291e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3411e-07, 0.0000e+00], + ..., + [ 8.0094e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-08, 0.0000e+00], + [-5.7742e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -8.0094e-08, 0.0000e+00], + [ 1.3039e-08, 9.3132e-09, 1.8626e-09, ..., 0.0000e+00, + 4.8429e-08, 0.0000e+00]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0063, 0.0135, -0.0232, 0.0191, 0.0265, -0.0165, -0.0182, -0.0105, + -0.0341, -0.0317], device='cuda:0'), grad: tensor([-7.0781e-08, 1.6764e-08, -2.9802e-07, 6.7055e-08, -9.3132e-09, + 8.9407e-08, 1.8626e-09, 2.3283e-07, -1.8440e-07, 1.5460e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 216.95, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4415 re_mapping 0.0027 re_causal 0.0093 /// teacc 99.02 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.3294, -0.1860, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1543, 0.1447, -0.0944, ..., 0.0246, -0.0687, -0.0819], + [ 0.1492, -0.1887, -0.1839, ..., -0.0711, 0.1762, -0.0337], + ..., + [-0.1891, -0.1093, 0.1330, ..., 0.0307, -0.2310, -0.0030], + [ 0.1133, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2365, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-2.6077e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -6.1467e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0063, 0.0135, -0.0231, 0.0191, 0.0265, -0.0166, -0.0182, -0.0105, + -0.0341, -0.0317], device='cuda:0'), grad: tensor([ 4.0978e-08, 8.0094e-08, -1.9930e-07, 1.0058e-07, 5.0291e-08, + 6.3330e-08, -1.1176e-08, -1.0617e-07, -1.8626e-09, -1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 216.72, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4232 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.05 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.3294, -0.1861, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1544, 0.1447, -0.0944, ..., 0.0246, -0.0688, -0.0819], + [ 0.1492, -0.1887, -0.1839, ..., -0.0711, 0.1762, -0.0337], + ..., + [-0.1891, -0.1093, 0.1330, ..., 0.0306, -0.2311, -0.0030], + [ 0.1133, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2365, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0063, 0.0134, -0.0231, 0.0191, 0.0265, -0.0166, -0.0182, -0.0104, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([ 1.4901e-08, 9.3132e-09, -3.5390e-08, -7.4506e-09, -6.3330e-08, + 3.9116e-08, -9.3132e-09, -1.6764e-08, 3.7253e-09, 5.9605e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 216.75, cls_loss 0.0006 cls_loss_mapping 0.0005 cls_loss_causal 0.4232 re_mapping 0.0027 re_causal 0.0093 /// teacc 99.04 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.3294, -0.1861, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1544, 0.1447, -0.0945, ..., 0.0246, -0.0688, -0.0819], + [ 0.1493, -0.1887, -0.1839, ..., -0.0711, 0.1763, -0.0337], + ..., + [-0.1891, -0.1093, 0.1330, ..., 0.0306, -0.2311, -0.0030], + [ 0.1134, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2366, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, 3.5390e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0063, 0.0134, -0.0230, 0.0191, 0.0265, -0.0166, -0.0182, -0.0104, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([ 2.2352e-08, 4.6566e-08, -7.4506e-08, -3.7253e-08, -1.0245e-07, + -4.1164e-07, 3.4645e-07, 6.5193e-08, 2.0489e-08, 1.1921e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 216.65, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4414 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.06 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.3294, -0.1862, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1544, 0.1447, -0.0945, ..., 0.0246, -0.0688, -0.0819], + [ 0.1493, -0.1887, -0.1839, ..., -0.0711, 0.1764, -0.0337], + ..., + [-0.1891, -0.1093, 0.1331, ..., 0.0306, -0.2311, -0.0030], + [ 0.1134, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2366, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-08, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.6764e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8813e-07, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.4214e-08, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, -1.8626e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0063, 0.0134, -0.0230, 0.0191, 0.0265, -0.0166, -0.0182, -0.0104, + -0.0341, -0.0318], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.1362e-07, -4.6566e-08, -1.0692e-06, 6.7055e-08, + 8.9407e-08, 0.0000e+00, 7.4133e-07, 1.0990e-07, -3.7253e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 216.77, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4148 re_mapping 0.0026 re_causal 0.0092 /// teacc 99.04 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.3294, -0.1862, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1544, 0.1447, -0.0945, ..., 0.0246, -0.0688, -0.0819], + [ 0.1493, -0.1888, -0.1839, ..., -0.0711, 0.1764, -0.0337], + ..., + [-0.1891, -0.1093, 0.1331, ..., 0.0306, -0.2312, -0.0030], + [ 0.1134, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2366, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0063, 0.0134, -0.0230, 0.0192, 0.0265, -0.0166, -0.0182, -0.0104, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([-1.6764e-08, 3.7253e-09, -5.5879e-09, -3.7253e-08, -2.7940e-08, + 2.0489e-08, 9.3132e-09, 2.6077e-08, 3.7253e-09, 2.4214e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 216.87, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.3975 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.08 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.3294, -0.1862, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1544, 0.1447, -0.0945, ..., 0.0246, -0.0688, -0.0819], + [ 0.1493, -0.1888, -0.1839, ..., -0.0711, 0.1764, -0.0337], + ..., + [-0.1891, -0.1094, 0.1331, ..., 0.0306, -0.2312, -0.0030], + [ 0.1134, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2367, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0062, 0.0134, -0.0229, 0.0192, 0.0265, -0.0166, -0.0182, -0.0105, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([ 0.0000e+00, -2.7940e-08, 9.3132e-09, 5.0291e-08, -4.4703e-08, + -5.4017e-08, 5.5879e-09, 9.3132e-09, 5.5879e-09, 4.2841e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 216.53, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4552 re_mapping 0.0026 re_causal 0.0095 /// teacc 99.05 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.3294, -0.1862, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1544, 0.1447, -0.0947, ..., 0.0246, -0.0688, -0.0819], + [ 0.1494, -0.1888, -0.1839, ..., -0.0711, 0.1765, -0.0337], + ..., + [-0.1892, -0.1094, 0.1331, ..., 0.0306, -0.2312, -0.0030], + [ 0.1134, -0.0156, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2367, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.0489e-08, 0.0000e+00], + [ 3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [-3.7253e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.2293e-07, 0.0000e+00], + [ 1.3039e-08, 4.7684e-07, 0.0000e+00, ..., 0.0000e+00, + 3.5390e-08, 0.0000e+00]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0062, 0.0134, -0.0229, 0.0192, 0.0265, -0.0166, -0.0182, -0.0105, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([ 8.7544e-08, 4.6566e-08, 1.0245e-07, 9.3132e-08, -1.3970e-06, + 3.3528e-08, 2.9802e-08, 7.4506e-08, -3.5949e-07, 1.2740e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 216.59, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4317 re_mapping 0.0027 re_causal 0.0093 /// teacc 99.07 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.3294, -0.1862, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1545, 0.1447, -0.0947, ..., 0.0246, -0.0688, -0.0819], + [ 0.1494, -0.1888, -0.1839, ..., -0.0711, 0.1765, -0.0337], + ..., + [-0.1891, -0.1094, 0.1331, ..., 0.0306, -0.2313, -0.0030], + [ 0.1134, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2367, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, -1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-3.9116e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-08, 0.0000e+00], + ..., + [ 2.0489e-08, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 2.4214e-08, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + [ 1.1176e-08, 2.0489e-08, 0.0000e+00, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0062, 0.0134, -0.0229, 0.0192, 0.0265, -0.0166, -0.0182, -0.0105, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([-4.2841e-08, -2.4214e-08, -1.6391e-07, 4.0978e-08, -9.1270e-08, + 1.3039e-08, 4.0978e-08, 9.8720e-08, -1.3039e-08, 1.4156e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 216.55, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4570 re_mapping 0.0025 re_causal 0.0093 /// teacc 99.07 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.3295, -0.1864, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1545, 0.1447, -0.0948, ..., 0.0246, -0.0688, -0.0819], + [ 0.1494, -0.1888, -0.1839, ..., -0.0711, 0.1766, -0.0337], + ..., + [-0.1891, -0.1094, 0.1331, ..., 0.0306, -0.2313, -0.0030], + [ 0.1133, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2367, -0.1810, 0.0372, ..., -0.0168, -0.2623, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 1.7136e-07, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, -1.9372e-07, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.4703e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0062, 0.0134, -0.0229, 0.0192, 0.0265, -0.0166, -0.0181, -0.0105, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([-2.0489e-08, 9.3319e-07, 1.0245e-07, 3.3528e-08, -1.2852e-07, + -5.5879e-09, 5.5879e-09, -1.0431e-06, 7.4506e-09, 1.2107e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 216.81, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4585 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.09 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.3295, -0.1865, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1545, 0.1448, -0.0949, ..., 0.0246, -0.0688, -0.0819], + [ 0.1494, -0.1889, -0.1839, ..., -0.0711, 0.1766, -0.0337], + ..., + [-0.1891, -0.1094, 0.1332, ..., 0.0306, -0.2314, -0.0030], + [ 0.1134, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2368, -0.1810, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 5.2154e-08, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0062, 0.0134, -0.0229, 0.0192, 0.0265, -0.0166, -0.0181, -0.0105, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([ 4.4703e-08, 1.4901e-08, 1.7509e-07, 1.4901e-08, -1.7323e-07, + 7.4506e-09, 1.8626e-08, -2.6636e-07, -1.1176e-08, 1.8254e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 216.63, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4376 re_mapping 0.0025 re_causal 0.0090 /// teacc 99.09 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.3295, -0.1866, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1545, 0.1448, -0.0949, ..., 0.0246, -0.0688, -0.0819], + [ 0.1494, -0.1889, -0.1839, ..., -0.0711, 0.1766, -0.0337], + ..., + [-0.1891, -0.1094, 0.1332, ..., 0.0306, -0.2314, -0.0030], + [ 0.1133, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2368, -0.1810, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.3039e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + [ 9.3132e-09, 5.5879e-09, -1.8626e-09, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0062, 0.0134, -0.0229, 0.0192, 0.0265, -0.0166, -0.0181, -0.0105, + -0.0342, -0.0318], device='cuda:0'), grad: tensor([ 1.8626e-09, -2.6077e-08, 1.3039e-08, -4.4331e-06, -3.7253e-09, + 4.4182e-06, -3.7253e-09, 3.1665e-08, -4.2841e-08, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 216.79, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4190 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.10 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.3295, -0.1866, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1546, 0.1448, -0.0949, ..., 0.0246, -0.0688, -0.0819], + [ 0.1494, -0.1889, -0.1839, ..., -0.0711, 0.1767, -0.0337], + ..., + [-0.1890, -0.1094, 0.1332, ..., 0.0306, -0.2314, -0.0030], + [ 0.1133, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2369, -0.1810, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -3.7253e-09, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0062, 0.0134, -0.0228, 0.0192, 0.0265, -0.0167, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([-1.8626e-09, 2.2352e-08, -2.9802e-08, -2.6077e-08, -5.5879e-08, + 7.4506e-09, 5.5879e-09, -2.0489e-08, 1.8626e-09, 8.7544e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 216.74, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4188 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.10 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.3295, -0.1868, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1546, 0.1448, -0.0949, ..., 0.0246, -0.0688, -0.0819], + [ 0.1495, -0.1889, -0.1839, ..., -0.0711, 0.1767, -0.0337], + ..., + [-0.1891, -0.1094, 0.1332, ..., 0.0306, -0.2315, -0.0030], + [ 0.1133, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2368, -0.1810, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, -1.0245e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-3.7253e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-8.3819e-09, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 6.5193e-09, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0062, 0.0134, -0.0228, 0.0192, 0.0265, -0.0167, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([ 4.0978e-08, -1.3039e-08, -5.0291e-08, 1.1269e-07, 3.8184e-08, + -5.3085e-08, 6.5193e-09, 2.2352e-08, -6.5193e-09, -9.6858e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 216.96, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4513 re_mapping 0.0025 re_causal 0.0092 /// teacc 99.08 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.3295, -0.1868, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1546, 0.1448, -0.0946, ..., 0.0246, -0.0688, -0.0819], + [ 0.1495, -0.1890, -0.1839, ..., -0.0711, 0.1768, -0.0337], + ..., + [-0.1891, -0.1094, 0.1332, ..., 0.0306, -0.2315, -0.0030], + [ 0.1133, -0.0157, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2367, -0.1810, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, -9.3132e-10, ..., 0.0000e+00, + -6.5193e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.6764e-08, 0.0000e+00], + [-2.3283e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.7008e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0062, 0.0134, -0.0228, 0.0192, 0.0265, -0.0166, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([-4.3772e-08, 1.6764e-08, -2.9802e-07, 1.4994e-07, 1.4901e-08, + 8.3819e-08, 2.5146e-08, 7.6368e-08, -5.8673e-08, 5.0291e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 216.85, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4279 re_mapping 0.0024 re_causal 0.0088 /// teacc 99.08 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.3295, -0.1868, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1547, 0.1448, -0.0946, ..., 0.0246, -0.0688, -0.0819], + [ 0.1495, -0.1890, -0.1839, ..., -0.0711, 0.1768, -0.0337], + ..., + [-0.1891, -0.1095, 0.1332, ..., 0.0306, -0.2316, -0.0030], + [ 0.1133, -0.0158, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2368, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + [ 4.6566e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.4214e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-5.8673e-08, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + -6.4261e-08, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0062, 0.0134, -0.0228, 0.0192, 0.0265, -0.0167, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([ 3.0734e-08, 3.3528e-08, -6.2399e-08, 7.6089e-07, 1.6764e-08, + -6.2492e-07, 7.7300e-08, -1.7695e-08, -2.3842e-07, 3.6322e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 216.88, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4164 re_mapping 0.0024 re_causal 0.0086 /// teacc 99.09 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.3295, -0.1869, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1547, 0.1448, -0.0947, ..., 0.0246, -0.0689, -0.0819], + [ 0.1495, -0.1890, -0.1839, ..., -0.0711, 0.1769, -0.0337], + ..., + [-0.1891, -0.1094, 0.1332, ..., 0.0306, -0.2316, -0.0030], + [ 0.1133, -0.0158, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2368, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-8.3819e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + ..., + [ 9.3132e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-7.4506e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0062, 0.0134, -0.0228, 0.0192, 0.0265, -0.0167, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([-6.5193e-09, 3.7253e-09, -1.7695e-08, 1.7695e-08, 7.4506e-09, + -2.9802e-08, 3.1665e-08, 1.7695e-08, -3.1665e-08, 7.4506e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 216.93, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4251 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.08 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.3295, -0.1869, -0.0966, ..., -0.0041, -0.0693, -0.0169], + [-0.1547, 0.1449, -0.0947, ..., 0.0246, -0.0689, -0.0819], + [ 0.1496, -0.1891, -0.1839, ..., -0.0711, 0.1769, -0.0337], + ..., + [-0.1891, -0.1095, 0.1333, ..., 0.0306, -0.2317, -0.0030], + [ 0.1133, -0.0158, -0.3376, ..., -0.0693, 0.1843, -0.0380], + [-0.2368, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [-7.4506e-09, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + [ 9.3132e-09, 1.2107e-08, 9.3132e-10, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0062, 0.0134, -0.0228, 0.0191, 0.0265, -0.0167, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([-5.5879e-09, 1.8626e-09, -2.3283e-08, 7.3574e-08, -2.7008e-08, + -1.4529e-07, 8.7544e-08, -2.7940e-09, -3.4459e-08, 7.1712e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 216.63, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4220 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.09 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.3295, -0.1870, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1547, 0.1449, -0.0947, ..., 0.0246, -0.0689, -0.0819], + [ 0.1496, -0.1891, -0.1839, ..., -0.0711, 0.1770, -0.0337], + ..., + [-0.1891, -0.1095, 0.1333, ..., 0.0306, -0.2317, -0.0030], + [ 0.1133, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2369, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-2.7940e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.6322e-08, 0.0000e+00], + ..., + [ 1.3970e-08, 9.3132e-10, -9.3132e-10, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 8.3819e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0062, 0.0134, -0.0227, 0.0192, 0.0265, -0.0167, -0.0180, -0.0105, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([ 8.3819e-09, 1.4901e-08, -1.0896e-07, 3.2596e-08, -6.5193e-09, + -9.3132e-09, -1.0245e-08, 1.6764e-08, 5.5879e-09, 6.7055e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 216.70, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4331 re_mapping 0.0024 re_causal 0.0089 /// teacc 99.09 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.3295, -0.1871, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1548, 0.1449, -0.0948, ..., 0.0246, -0.0689, -0.0819], + [ 0.1496, -0.1891, -0.1839, ..., -0.0711, 0.1770, -0.0337], + ..., + [-0.1891, -0.1095, 0.1333, ..., 0.0306, -0.2318, -0.0030], + [ 0.1133, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2369, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.8626e-09, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0062, 0.0134, -0.0227, 0.0192, 0.0265, -0.0167, -0.0179, -0.0104, + -0.0343, -0.0318], device='cuda:0'), grad: tensor([ 3.4459e-08, -6.5193e-09, 1.3039e-08, 6.3330e-08, -8.3819e-09, + -2.2072e-07, 3.7253e-08, 1.5832e-08, 5.2154e-08, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 216.77, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4612 re_mapping 0.0024 re_causal 0.0092 /// teacc 99.08 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.3295, -0.1871, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1548, 0.1449, -0.0948, ..., 0.0246, -0.0689, -0.0819], + [ 0.1496, -0.1891, -0.1839, ..., -0.0711, 0.1770, -0.0337], + ..., + [-0.1890, -0.1094, 0.1333, ..., 0.0306, -0.2318, -0.0030], + [ 0.1133, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2369, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 5.5879e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -1.1176e-08, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0062, 0.0134, -0.0227, 0.0192, 0.0265, -0.0167, -0.0179, -0.0104, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 2.5146e-08, 3.9116e-08, 7.4506e-09, 4.6566e-09, 5.5879e-09, + 9.3132e-09, -3.3528e-08, -4.8429e-08, -2.7940e-09, -1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 216.89, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4605 re_mapping 0.0024 re_causal 0.0090 /// teacc 99.09 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.3296, -0.1873, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1548, 0.1448, -0.0948, ..., 0.0246, -0.0689, -0.0819], + [ 0.1496, -0.1891, -0.1840, ..., -0.0711, 0.1771, -0.0337], + ..., + [-0.1891, -0.1094, 0.1333, ..., 0.0306, -0.2319, -0.0030], + [ 0.1133, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2369, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -2.3283e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -2.0489e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0062, 0.0134, -0.0227, 0.0192, 0.0265, -0.0167, -0.0179, -0.0104, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([-2.5518e-07, 6.5193e-09, -3.7253e-09, 1.9558e-08, 3.9861e-07, + -7.4506e-09, 2.1700e-07, -1.3970e-08, 5.5879e-09, -3.6601e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 216.85, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4352 re_mapping 0.0023 re_causal 0.0087 /// teacc 99.15 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.3296, -0.1873, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1549, 0.1448, -0.0948, ..., 0.0245, -0.0689, -0.0819], + [ 0.1497, -0.1892, -0.1840, ..., -0.0711, 0.1772, -0.0337], + ..., + [-0.1890, -0.1094, 0.1333, ..., 0.0306, -0.2319, -0.0030], + [ 0.1133, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2370, -0.1811, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, -7.4506e-09, 8.3819e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 1.8626e-09, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, -5.1223e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.8626e-08, 2.9802e-08, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0062, 0.0134, -0.0226, 0.0192, 0.0265, -0.0168, -0.0179, -0.0104, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([-5.5879e-09, 4.0978e-08, 1.7695e-08, 1.2107e-08, -5.3830e-07, + 1.3970e-08, 1.2107e-08, 1.4249e-07, 7.4506e-09, 3.1106e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 216.66, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4246 re_mapping 0.0023 re_causal 0.0089 /// teacc 99.09 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.3296, -0.1873, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1549, 0.1449, -0.0947, ..., 0.0245, -0.0690, -0.0819], + [ 0.1497, -0.1892, -0.1840, ..., -0.0711, 0.1772, -0.0337], + ..., + [-0.1890, -0.1095, 0.1333, ..., 0.0306, -0.2320, -0.0030], + [ 0.1134, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2370, -0.1812, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-2.8871e-08, 0.0000e+00, -4.6566e-09, ..., 0.0000e+00, + -2.8871e-08, 0.0000e+00], + ..., + [ 1.3039e-08, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 1.3039e-08, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0062, 0.0134, -0.0226, 0.0192, 0.0265, -0.0168, -0.0179, -0.0104, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([-1.5832e-08, 1.9558e-08, -1.5181e-07, 1.8626e-09, 3.0734e-08, + 3.3528e-08, 1.5832e-08, 8.1025e-08, 1.1176e-08, -1.6764e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 216.87, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4476 re_mapping 0.0023 re_causal 0.0089 /// teacc 99.04 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.3296, -0.1873, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1549, 0.1449, -0.0947, ..., 0.0245, -0.0690, -0.0819], + [ 0.1498, -0.1892, -0.1840, ..., -0.0711, 0.1773, -0.0337], + ..., + [-0.1891, -0.1095, 0.1333, ..., 0.0306, -0.2320, -0.0030], + [ 0.1134, -0.0158, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2371, -0.1812, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, -1.7602e-07, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 1.8626e-09, 8.4750e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 6.0536e-08, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [-6.5193e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + [ 1.8626e-09, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0061, 0.0134, -0.0226, 0.0192, 0.0265, -0.0168, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([-8.3819e-09, -3.4273e-07, 1.8440e-07, -5.0291e-08, 1.9185e-07, + 4.0978e-08, -2.2538e-07, 1.4342e-07, -1.9558e-08, 8.1025e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 216.58, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4391 re_mapping 0.0023 re_causal 0.0087 /// teacc 99.07 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.3296, -0.1873, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1549, 0.1449, -0.0947, ..., 0.0245, -0.0690, -0.0819], + [ 0.1498, -0.1893, -0.1840, ..., -0.0711, 0.1774, -0.0337], + ..., + [-0.1891, -0.1095, 0.1333, ..., 0.0306, -0.2321, -0.0030], + [ 0.1134, -0.0159, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2372, -0.1812, 0.0371, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 3.7253e-09, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-1.0617e-07, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + -1.9185e-07, 0.0000e+00], + ..., + [ 8.5682e-08, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.5274e-07, 0.0000e+00], + [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 9.3132e-10, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0061, 0.0134, -0.0226, 0.0192, 0.0265, -0.0168, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 1.9558e-08, 1.5832e-08, -9.1363e-07, 4.0978e-08, 5.1595e-07, + 0.0000e+00, 1.0245e-08, 8.3819e-07, 1.7695e-08, -5.2620e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 216.97, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4148 re_mapping 0.0023 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.3296, -0.1874, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1550, 0.1449, -0.0949, ..., 0.0245, -0.0690, -0.0819], + [ 0.1498, -0.1893, -0.1840, ..., -0.0711, 0.1774, -0.0337], + ..., + [-0.1890, -0.1095, 0.1334, ..., 0.0306, -0.2322, -0.0030], + [ 0.1134, -0.0159, -0.3377, ..., -0.0693, 0.1844, -0.0380], + [-0.2372, -0.1812, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.2107e-08, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 3.8184e-08, 0.0000e+00], + [ 2.7940e-09, -6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [-1.4249e-07, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -5.2340e-07, 0.0000e+00], + ..., + [ 1.0896e-07, 4.6566e-09, 5.5879e-09, ..., 0.0000e+00, + 4.0047e-07, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, -1.8626e-09, -1.0245e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0061, 0.0134, -0.0225, 0.0192, 0.0265, -0.0168, -0.0179, -0.0104, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 1.0990e-07, 2.8871e-08, -1.3961e-06, 2.1234e-07, 5.7556e-07, + 7.4506e-08, -4.6566e-09, 1.3085e-06, 1.4901e-08, -9.1922e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 216.81, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4275 re_mapping 0.0022 re_causal 0.0088 /// teacc 99.11 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.3296, -0.1874, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1551, 0.1449, -0.0949, ..., 0.0245, -0.0690, -0.0819], + [ 0.1499, -0.1894, -0.1840, ..., -0.0711, 0.1775, -0.0337], + ..., + [-0.1890, -0.1095, 0.1334, ..., 0.0306, -0.2322, -0.0030], + [ 0.1134, -0.0159, -0.3377, ..., -0.0693, 0.1845, -0.0380], + [-0.2373, -0.1812, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 4.0047e-08, 2.2352e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-6.5193e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.0489e-08, 0.0000e+00], + ..., + [ 0.0000e+00, -4.5635e-08, -2.3283e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 2.1420e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0061, 0.0134, -0.0225, 0.0192, 0.0265, -0.0168, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 7.4506e-09, 1.6019e-07, -6.8918e-08, -9.3132e-09, -5.4017e-08, + 6.8918e-08, 9.3132e-10, -1.7136e-07, 2.7940e-09, 6.6124e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 216.77, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4322 re_mapping 0.0022 re_causal 0.0086 /// teacc 99.11 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.3296, -0.1874, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1551, 0.1450, -0.0948, ..., 0.0245, -0.0690, -0.0819], + [ 0.1499, -0.1894, -0.1840, ..., -0.0711, 0.1776, -0.0337], + ..., + [-0.1891, -0.1095, 0.1334, ..., 0.0306, -0.2323, -0.0030], + [ 0.1134, -0.0159, -0.3377, ..., -0.0693, 0.1845, -0.0380], + [-0.2374, -0.1812, 0.0372, ..., -0.0168, -0.2624, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, -6.5193e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0061, 0.0134, -0.0225, 0.0191, 0.0265, -0.0168, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([-3.0734e-08, -9.3132e-10, -4.6566e-09, 1.0245e-08, 2.3283e-08, + 8.3819e-09, 2.6077e-08, 7.4506e-09, -4.6566e-09, -2.9802e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 217.12, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4290 re_mapping 0.0023 re_causal 0.0086 /// teacc 99.11 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.3296, -0.1875, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1551, 0.1450, -0.0949, ..., 0.0245, -0.0690, -0.0819], + [ 0.1500, -0.1894, -0.1841, ..., -0.0711, 0.1777, -0.0337], + ..., + [-0.1891, -0.1095, 0.1335, ..., 0.0306, -0.2324, -0.0030], + [ 0.1134, -0.0159, -0.3377, ..., -0.0693, 0.1845, -0.0380], + [-0.2375, -0.1812, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.2107e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-3.7253e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 2.0489e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0061, 0.0134, -0.0225, 0.0191, 0.0265, -0.0168, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 3.3528e-08, 7.4506e-09, -2.2352e-08, 1.4901e-08, -6.3330e-08, + 1.4901e-08, -3.0734e-08, -1.2107e-08, -3.7253e-09, 6.8918e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 216.86, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4287 re_mapping 0.0022 re_causal 0.0085 /// teacc 99.12 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.3297, -0.1875, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1551, 0.1450, -0.0949, ..., 0.0245, -0.0691, -0.0819], + [ 0.1500, -0.1895, -0.1841, ..., -0.0711, 0.1777, -0.0337], + ..., + [-0.1891, -0.1095, 0.1335, ..., 0.0306, -0.2324, -0.0030], + [ 0.1134, -0.0159, -0.3377, ..., -0.0693, 0.1845, -0.0380], + [-0.2376, -0.1812, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.3039e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.2107e-08, 0.0000e+00], + [ 0.0000e+00, -1.0524e-07, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-2.8871e-08, 2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + -3.5390e-08, 0.0000e+00], + ..., + [ 5.5879e-09, 7.4506e-08, 0.0000e+00, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0061, 0.0134, -0.0225, 0.0191, 0.0265, -0.0167, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([-2.8871e-08, -3.7961e-06, 6.4168e-07, 1.7975e-07, 1.8626e-09, + 5.9605e-08, 1.3970e-08, 2.8051e-06, 7.4506e-09, 1.1362e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 216.79, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4180 re_mapping 0.0022 re_causal 0.0086 /// teacc 99.10 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.3297, -0.1875, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1551, 0.1450, -0.0950, ..., 0.0245, -0.0691, -0.0819], + [ 0.1501, -0.1895, -0.1841, ..., -0.0711, 0.1778, -0.0337], + ..., + [-0.1891, -0.1095, 0.1336, ..., 0.0306, -0.2324, -0.0030], + [ 0.1134, -0.0160, -0.3377, ..., -0.0693, 0.1846, -0.0380], + [-0.2377, -0.1813, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, -6.5193e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0061, 0.0134, -0.0225, 0.0191, 0.0265, -0.0167, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 9.3132e-10, 3.7253e-08, 2.0489e-08, 2.8405e-07, 3.1665e-08, + -2.6729e-07, 9.3132e-09, -2.0117e-07, 1.0245e-08, 7.8231e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 216.83, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4486 re_mapping 0.0023 re_causal 0.0088 /// teacc 99.11 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.3297, -0.1876, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1552, 0.1450, -0.0951, ..., 0.0245, -0.0691, -0.0819], + [ 0.1502, -0.1896, -0.1841, ..., -0.0711, 0.1780, -0.0337], + ..., + [-0.1892, -0.1096, 0.1337, ..., 0.0306, -0.2325, -0.0030], + [ 0.1134, -0.0160, -0.3377, ..., -0.0693, 0.1846, -0.0380], + [-0.2378, -0.1813, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.7253e-09, -3.7253e-09, 9.3132e-10, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.3039e-08, -9.3132e-09, -1.8626e-09, ..., 0.0000e+00, + -1.1176e-08, 0.0000e+00], + [ 1.8626e-09, 5.5879e-09, -4.6566e-09, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0061, 0.0134, -0.0224, 0.0191, 0.0265, -0.0167, -0.0179, -0.0105, + -0.0344, -0.0318], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.4901e-08, -6.5193e-09, 1.0896e-07, 7.3574e-08, + -7.5437e-08, 5.5879e-09, -1.8626e-09, -4.2841e-08, -9.0338e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 216.95, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4410 re_mapping 0.0022 re_causal 0.0085 /// teacc 99.10 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.3297, -0.1876, -0.0967, ..., -0.0041, -0.0693, -0.0169], + [-0.1552, 0.1450, -0.0951, ..., 0.0245, -0.0692, -0.0819], + [ 0.1502, -0.1896, -0.1841, ..., -0.0711, 0.1781, -0.0337], + ..., + [-0.1892, -0.1096, 0.1337, ..., 0.0306, -0.2327, -0.0030], + [ 0.1135, -0.0160, -0.3377, ..., -0.0693, 0.1846, -0.0380], + [-0.2380, -0.1813, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.7789e-09, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [-6.5193e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0061, 0.0134, -0.0223, 0.0190, 0.0266, -0.0166, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([-5.8208e-08, 3.3993e-08, 2.5146e-08, -6.4494e-07, 6.9849e-09, + 9.2667e-08, 3.6322e-08, 5.2201e-07, -1.6764e-08, 1.7229e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 216.55, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4314 re_mapping 0.0023 re_causal 0.0086 /// teacc 99.11 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.3297, -0.1877, -0.0968, ..., -0.0041, -0.0693, -0.0169], + [-0.1552, 0.1451, -0.0952, ..., 0.0245, -0.0692, -0.0819], + [ 0.1503, -0.1897, -0.1841, ..., -0.0711, 0.1782, -0.0337], + ..., + [-0.1893, -0.1096, 0.1337, ..., 0.0306, -0.2328, -0.0030], + [ 0.1135, -0.0160, -0.3377, ..., -0.0693, 0.1847, -0.0380], + [-0.2381, -0.1813, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.3504e-08, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [-1.9092e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -2.7008e-08, 0.0000e+00], + ..., + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [-3.5856e-08, -1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + -3.0268e-08, 0.0000e+00], + [ 1.8626e-09, 1.2573e-08, 2.3283e-09, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0060, 0.0134, -0.0223, 0.0190, 0.0266, -0.0166, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([ 1.4435e-08, 4.7032e-08, -7.2643e-08, 2.0489e-08, 9.8720e-08, + 3.3528e-08, -1.3039e-08, 2.5146e-08, -9.7323e-08, -5.8208e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 216.77, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4152 re_mapping 0.0022 re_causal 0.0085 /// teacc 99.12 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.3297, -0.1877, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1552, 0.1452, -0.0950, ..., 0.0245, -0.0692, -0.0819], + [ 0.1503, -0.1897, -0.1841, ..., -0.0711, 0.1783, -0.0337], + ..., + [-0.1893, -0.1097, 0.1337, ..., 0.0306, -0.2329, -0.0030], + [ 0.1135, -0.0160, -0.3377, ..., -0.0693, 0.1847, -0.0380], + [-0.2382, -0.1814, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 4.6566e-10, -4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.3970e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 5.5879e-09, 1.0710e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0060, 0.0134, -0.0222, 0.0190, 0.0266, -0.0166, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([-6.5193e-09, 2.1420e-08, 9.3132e-09, 1.5367e-08, 4.2375e-08, + -7.4040e-08, 2.9337e-08, -4.3306e-08, 9.3132e-09, 6.0536e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 216.69, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3966 re_mapping 0.0023 re_causal 0.0084 /// teacc 99.14 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.3297, -0.1877, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1554, 0.1451, -0.0950, ..., 0.0245, -0.0692, -0.0819], + [ 0.1503, -0.1898, -0.1841, ..., -0.0711, 0.1783, -0.0337], + ..., + [-0.1891, -0.1097, 0.1337, ..., 0.0306, -0.2329, -0.0030], + [ 0.1136, -0.0160, -0.3377, ..., -0.0693, 0.1848, -0.0380], + [-0.2382, -0.1814, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 6.0536e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -5.5879e-09, -1.3039e-08, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0060, 0.0134, -0.0223, 0.0190, 0.0266, -0.0166, -0.0180, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([ 2.3283e-09, 6.7521e-08, 1.4901e-08, -2.5146e-08, 3.0734e-08, + 1.5367e-08, 9.3132e-10, -1.2293e-07, 2.3283e-09, 2.3749e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 216.87, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4280 re_mapping 0.0022 re_causal 0.0086 /// teacc 99.14 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.3298, -0.1878, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1555, 0.1452, -0.0950, ..., 0.0245, -0.0692, -0.0819], + [ 0.1504, -0.1898, -0.1841, ..., -0.0711, 0.1784, -0.0337], + ..., + [-0.1891, -0.1097, 0.1337, ..., 0.0306, -0.2330, -0.0030], + [ 0.1136, -0.0161, -0.3377, ..., -0.0693, 0.1848, -0.0380], + [-0.2383, -0.1814, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.2107e-08, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + -3.2131e-08, 0.0000e+00], + ..., + [ 6.0536e-09, 4.1910e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4901e-08, 0.0000e+00], + [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 2.7940e-09, 6.7055e-08, 0.0000e+00, ..., 0.0000e+00, + 6.0536e-09, 0.0000e+00]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0060, 0.0134, -0.0222, 0.0190, 0.0266, -0.0166, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([ 1.1642e-08, 9.7789e-09, -5.3551e-08, 6.7987e-08, -5.0431e-07, + -6.3796e-08, 1.1176e-08, 4.5169e-08, 1.2573e-08, 4.8103e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 216.99, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4428 re_mapping 0.0023 re_causal 0.0086 /// teacc 99.13 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.3298, -0.1879, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1555, 0.1452, -0.0950, ..., 0.0245, -0.0693, -0.0819], + [ 0.1504, -0.1899, -0.1841, ..., -0.0711, 0.1786, -0.0337], + ..., + [-0.1892, -0.1097, 0.1337, ..., 0.0306, -0.2332, -0.0030], + [ 0.1136, -0.0161, -0.3378, ..., -0.0693, 0.1849, -0.0380], + [-0.2384, -0.1814, 0.0372, ..., -0.0168, -0.2625, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 5.1223e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -6.6124e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 4.8429e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0060, 0.0134, -0.0221, 0.0190, 0.0266, -0.0166, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([ 6.8452e-08, -1.3039e-07, 1.7695e-08, 3.2596e-09, 2.7008e-08, + 3.7719e-08, -6.8452e-08, 1.0617e-07, 9.7789e-09, -6.2399e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 216.90, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4225 re_mapping 0.0023 re_causal 0.0085 /// teacc 99.10 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.3298, -0.1879, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1555, 0.1453, -0.0951, ..., 0.0245, -0.0693, -0.0819], + [ 0.1505, -0.1900, -0.1841, ..., -0.0711, 0.1786, -0.0337], + ..., + [-0.1892, -0.1098, 0.1337, ..., 0.0306, -0.2333, -0.0030], + [ 0.1136, -0.0161, -0.3378, ..., -0.0693, 0.1849, -0.0380], + [-0.2386, -0.1815, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 0.0000e+00, -5.6345e-08, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [-1.3970e-08, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + -2.6077e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 2.0489e-08, 4.6566e-10, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, -1.8626e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0060, 0.0134, -0.0221, 0.0190, 0.0266, -0.0166, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([ 1.9092e-08, -1.0431e-07, -6.3330e-08, 1.0664e-07, 5.5879e-07, + -5.1223e-08, 1.7229e-08, 8.1491e-08, 5.7276e-08, -6.1234e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 216.77, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4218 re_mapping 0.0022 re_causal 0.0085 /// teacc 99.12 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.3298, -0.1880, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1555, 0.1453, -0.0951, ..., 0.0245, -0.0693, -0.0819], + [ 0.1506, -0.1901, -0.1841, ..., -0.0711, 0.1787, -0.0337], + ..., + [-0.1892, -0.1098, 0.1338, ..., 0.0306, -0.2334, -0.0030], + [ 0.1136, -0.0162, -0.3378, ..., -0.0693, 0.1850, -0.0380], + [-0.2386, -0.1815, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 9.7789e-09, 1.3970e-09, -5.1223e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.7695e-08, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.2107e-08, 1.8161e-08, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0060, 0.0134, -0.0221, 0.0190, 0.0266, -0.0167, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([ 1.7695e-08, 2.6077e-08, 1.8626e-08, 3.5018e-07, -1.1222e-07, + -4.8149e-07, 6.1467e-08, -8.5682e-08, 4.7963e-08, 1.7090e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 216.75, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4305 re_mapping 0.0023 re_causal 0.0087 /// teacc 99.14 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.3298, -0.1880, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1555, 0.1454, -0.0951, ..., 0.0245, -0.0693, -0.0819], + [ 0.1506, -0.1901, -0.1842, ..., -0.0711, 0.1788, -0.0337], + ..., + [-0.1893, -0.1098, 0.1338, ..., 0.0306, -0.2334, -0.0030], + [ 0.1137, -0.0162, -0.3378, ..., -0.0693, 0.1850, -0.0380], + [-0.2387, -0.1815, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + ..., + [ 1.8626e-09, 3.2596e-09, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 2.7940e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0060, 0.0134, -0.0221, 0.0190, 0.0266, -0.0167, -0.0179, -0.0106, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([-1.3039e-08, 1.6764e-08, -1.5367e-08, 2.3609e-07, 5.5879e-09, + -1.6345e-07, 6.5193e-09, -8.1491e-08, 9.3132e-09, 1.3039e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 216.81, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4234 re_mapping 0.0022 re_causal 0.0087 /// teacc 99.13 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.3299, -0.1881, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1556, 0.1454, -0.0951, ..., 0.0245, -0.0694, -0.0819], + [ 0.1506, -0.1901, -0.1842, ..., -0.0711, 0.1788, -0.0337], + ..., + [-0.1893, -0.1098, 0.1339, ..., 0.0306, -0.2334, -0.0030], + [ 0.1137, -0.0163, -0.3378, ..., -0.0693, 0.1851, -0.0380], + [-0.2388, -0.1816, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 5.1223e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.9162e-09, 1.0710e-08, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 3.2596e-09, 5.1223e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.3970e-09, 0.0000e+00], + [ 4.6566e-10, 4.3772e-08, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0060, 0.0134, -0.0222, 0.0190, 0.0266, -0.0168, -0.0179, -0.0105, + -0.0344, -0.0319], device='cuda:0'), grad: tensor([-6.0536e-09, 3.2596e-08, 1.4901e-08, 2.2352e-08, -2.3283e-08, + 1.4901e-08, -1.6671e-07, 1.3039e-08, 3.7253e-09, 1.0012e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 216.88, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4196 re_mapping 0.0022 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.3299, -0.1881, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1556, 0.1454, -0.0951, ..., 0.0245, -0.0694, -0.0819], + [ 0.1507, -0.1902, -0.1842, ..., -0.0711, 0.1789, -0.0337], + ..., + [-0.1894, -0.1098, 0.1339, ..., 0.0306, -0.2335, -0.0030], + [ 0.1137, -0.0163, -0.3378, ..., -0.0693, 0.1851, -0.0380], + [-0.2389, -0.1816, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.1910e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, -1.8161e-07, 0.0000e+00, ..., 0.0000e+00, + -2.5146e-08, 0.0000e+00], + [-9.3132e-10, 9.7789e-08, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 7.3109e-08, 9.3132e-10, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0060, 0.0134, -0.0221, 0.0190, 0.0266, -0.0168, -0.0179, -0.0105, + -0.0343, -0.0320], device='cuda:0'), grad: tensor([ 5.5879e-09, -3.9255e-07, 2.0815e-07, 1.5367e-08, 6.0070e-08, + 1.1176e-08, 2.7940e-09, 2.1467e-07, 1.3970e-09, -1.1176e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 216.58, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4108 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.3299, -0.1882, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1556, 0.1455, -0.0951, ..., 0.0245, -0.0694, -0.0819], + [ 0.1508, -0.1902, -0.1842, ..., -0.0711, 0.1790, -0.0337], + ..., + [-0.1894, -0.1099, 0.1339, ..., 0.0306, -0.2336, -0.0030], + [ 0.1137, -0.0164, -0.3378, ..., -0.0693, 0.1852, -0.0380], + [-0.2389, -0.1816, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0060, 0.0135, -0.0221, 0.0190, 0.0267, -0.0168, -0.0179, -0.0106, + -0.0344, -0.0320], device='cuda:0'), grad: tensor([ 1.2573e-08, 4.8429e-08, -5.5879e-09, -1.4110e-07, 3.7020e-07, + 1.2061e-07, 9.3132e-10, 2.4214e-08, 1.8626e-08, -4.4052e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 216.94, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4198 re_mapping 0.0022 re_causal 0.0086 /// teacc 99.12 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.3299, -0.1882, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1556, 0.1455, -0.0951, ..., 0.0245, -0.0695, -0.0819], + [ 0.1509, -0.1903, -0.1842, ..., -0.0711, 0.1791, -0.0337], + ..., + [-0.1894, -0.1099, 0.1339, ..., 0.0306, -0.2337, -0.0030], + [ 0.1137, -0.0164, -0.3378, ..., -0.0693, 0.1852, -0.0380], + [-0.2390, -0.1817, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 4.6566e-10, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 1.3970e-09, -1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-1.3039e-08, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + -5.2620e-08, 0.0000e+00], + ..., + [ 6.5193e-09, 6.9849e-09, -4.6566e-10, ..., 0.0000e+00, + 2.0955e-08, 0.0000e+00], + [ 2.7940e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 4.6566e-10, 6.0536e-09, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0060, 0.0135, -0.0220, 0.0190, 0.0267, -0.0168, -0.0179, -0.0106, + -0.0344, -0.0320], device='cuda:0'), grad: tensor([ 1.3039e-08, -3.2596e-09, -2.0629e-07, 9.3598e-08, -1.1176e-08, + -1.7276e-07, 1.6764e-07, 8.4285e-08, 1.3970e-08, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 216.65, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.3967 re_mapping 0.0021 re_causal 0.0083 /// teacc 99.12 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.3299, -0.1882, -0.0968, ..., -0.0041, -0.0694, -0.0169], + [-0.1557, 0.1456, -0.0951, ..., 0.0245, -0.0695, -0.0819], + [ 0.1509, -0.1903, -0.1842, ..., -0.0711, 0.1792, -0.0337], + ..., + [-0.1895, -0.1100, 0.1339, ..., 0.0306, -0.2338, -0.0030], + [ 0.1137, -0.0164, -0.3378, ..., -0.0693, 0.1852, -0.0380], + [-0.2390, -0.1817, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.3970e-09, 0.0000e+00], + [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-6.0536e-09, 0.0000e+00, 1.3970e-09, ..., 0.0000e+00, + -1.3504e-08, 0.0000e+00], + ..., + [ 3.7253e-09, 4.6566e-10, -4.6566e-09, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 6.9849e-09, 5.5879e-09, 1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0060, 0.0135, -0.0220, 0.0190, 0.0267, -0.0168, -0.0179, -0.0106, + -0.0345, -0.0320], device='cuda:0'), grad: tensor([-3.2131e-08, 1.8161e-08, -1.1176e-08, 1.4715e-07, -1.5367e-08, + -5.2154e-07, 3.8138e-07, -4.2375e-08, 1.3504e-08, 6.2864e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 216.71, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4062 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.11 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.3300, -0.1883, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1557, 0.1456, -0.0952, ..., 0.0245, -0.0695, -0.0819], + [ 0.1510, -0.1904, -0.1842, ..., -0.0711, 0.1793, -0.0337], + ..., + [-0.1895, -0.1100, 0.1340, ..., 0.0306, -0.2339, -0.0030], + [ 0.1137, -0.0165, -0.3378, ..., -0.0693, 0.1852, -0.0380], + [-0.2391, -0.1817, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.8626e-09, -6.9849e-09, 9.3132e-10, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 9.3132e-10, 1.3970e-09, 4.6566e-10, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 4.1910e-09, -3.7253e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-3.2131e-08, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + -3.6322e-08, 0.0000e+00], + [ 1.8626e-09, 4.6566e-10, 1.8626e-09, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0060, 0.0135, -0.0220, 0.0191, 0.0267, -0.0169, -0.0179, -0.0107, + -0.0345, -0.0320], device='cuda:0'), grad: tensor([-1.3039e-08, 1.9558e-08, 1.5367e-08, 4.1444e-08, 3.6322e-08, + 6.3796e-08, 6.0536e-09, -7.2643e-08, -1.1735e-07, 2.6077e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 216.69, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4108 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.3300, -0.1885, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1557, 0.1457, -0.0952, ..., 0.0245, -0.0695, -0.0819], + [ 0.1510, -0.1905, -0.1843, ..., -0.0711, 0.1793, -0.0337], + ..., + [-0.1895, -0.1101, 0.1340, ..., 0.0306, -0.2340, -0.0030], + [ 0.1137, -0.0165, -0.3379, ..., -0.0693, 0.1852, -0.0380], + [-0.2392, -0.1818, 0.0372, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.4901e-08, 1.9092e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.7940e-09, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 2.3283e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-5.5879e-09, 6.0536e-09, -4.6566e-10, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + [ 4.6566e-10, 1.8161e-08, -9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0060, 0.0135, -0.0220, 0.0191, 0.0267, -0.0170, -0.0178, -0.0107, + -0.0346, -0.0320], device='cuda:0'), grad: tensor([ 5.7742e-08, 2.0955e-08, 1.9092e-08, 2.5146e-08, -5.4482e-08, + 7.5437e-08, -1.7183e-07, -3.6787e-08, -4.6566e-09, 7.4971e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 216.62, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4445 re_mapping 0.0021 re_causal 0.0085 /// teacc 99.10 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.3300, -0.1888, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1558, 0.1459, -0.0950, ..., 0.0245, -0.0695, -0.0819], + [ 0.1511, -0.1906, -0.1843, ..., -0.0711, 0.1794, -0.0337], + ..., + [-0.1895, -0.1103, 0.1339, ..., 0.0306, -0.2341, -0.0030], + [ 0.1137, -0.0166, -0.3379, ..., -0.0693, 0.1853, -0.0380], + [-0.2392, -0.1818, 0.0371, ..., -0.0168, -0.2626, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [-1.2573e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.7940e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 1.3970e-09, -4.6566e-10, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, -9.3132e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0060, 0.0136, -0.0220, 0.0191, 0.0267, -0.0170, -0.0177, -0.0107, + -0.0346, -0.0320], device='cuda:0'), grad: tensor([ 1.1176e-08, 5.1223e-09, -5.8208e-08, 4.9360e-08, 4.8429e-08, + 1.8626e-08, -2.2817e-08, -8.3819e-09, 4.6566e-10, -4.7963e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 216.84, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4074 re_mapping 0.0021 re_causal 0.0081 /// teacc 99.11 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.3300, -0.1889, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1558, 0.1460, -0.0951, ..., 0.0245, -0.0696, -0.0819], + [ 0.1512, -0.1907, -0.1843, ..., -0.0711, 0.1796, -0.0337], + ..., + [-0.1896, -0.1103, 0.1339, ..., 0.0306, -0.2342, -0.0030], + [ 0.1137, -0.0167, -0.3379, ..., -0.0693, 0.1853, -0.0380], + [-0.2393, -0.1819, 0.0372, ..., -0.0168, -0.2627, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.7940e-09, 6.0536e-09, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 4.6566e-10, -1.0245e-08, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [-2.7940e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + -3.2596e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 4.1910e-09, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 9.3132e-10, 5.1223e-09, -4.6566e-10, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0060, 0.0136, -0.0219, 0.0191, 0.0267, -0.0170, -0.0177, -0.0108, + -0.0347, -0.0320], device='cuda:0'), grad: tensor([ 2.0489e-08, -1.1176e-08, -7.9162e-09, -4.9314e-07, 5.1223e-09, + 4.5076e-07, -2.0023e-08, 1.8626e-08, 1.3970e-08, 1.4435e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 216.62, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4238 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.3301, -0.1889, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1558, 0.1460, -0.0951, ..., 0.0245, -0.0696, -0.0819], + [ 0.1513, -0.1907, -0.1843, ..., -0.0711, 0.1797, -0.0337], + ..., + [-0.1897, -0.1104, 0.1340, ..., 0.0306, -0.2344, -0.0030], + [ 0.1137, -0.0167, -0.3379, ..., -0.0693, 0.1853, -0.0380], + [-0.2394, -0.1819, 0.0371, ..., -0.0168, -0.2627, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [-5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.7789e-09, 0.0000e+00], + ..., + [ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 3.2596e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 4.6566e-10, 0.0000e+00, -1.3970e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0060, 0.0136, -0.0219, 0.0191, 0.0267, -0.0170, -0.0177, -0.0108, + -0.0348, -0.0320], device='cuda:0'), grad: tensor([-2.4959e-07, 6.9849e-09, -6.9849e-09, 2.5611e-08, 1.6764e-08, + -4.1910e-09, 6.9849e-09, 2.4214e-08, 1.4435e-08, 1.8207e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 216.66, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.3982 re_mapping 0.0021 re_causal 0.0082 /// teacc 99.10 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.3301, -0.1890, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1559, 0.1461, -0.0951, ..., 0.0245, -0.0696, -0.0819], + [ 0.1514, -0.1908, -0.1843, ..., -0.0711, 0.1799, -0.0337], + ..., + [-0.1898, -0.1104, 0.1340, ..., 0.0306, -0.2345, -0.0030], + [ 0.1137, -0.0168, -0.3379, ..., -0.0693, 0.1854, -0.0380], + [-0.2395, -0.1820, 0.0371, ..., -0.0168, -0.2627, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, -6.9849e-09, 1.3970e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, 2.3283e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, -5.1223e-09, ..., 0.0000e+00, + 1.9558e-08, 0.0000e+00], + [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, 0.0000e+00], + [ 1.3970e-09, 3.1199e-08, -4.6566e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0060, 0.0136, -0.0218, 0.0191, 0.0268, -0.0171, -0.0177, -0.0108, + -0.0348, -0.0320], device='cuda:0'), grad: tensor([-4.6566e-10, 1.3970e-09, 1.5832e-08, -4.7497e-08, -1.1642e-07, + -1.8626e-08, 9.3132e-10, 4.7497e-08, -1.3970e-09, 1.0896e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 217.27, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4024 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.12 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.3302, -0.1890, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1559, 0.1461, -0.0951, ..., 0.0245, -0.0697, -0.0819], + [ 0.1515, -0.1909, -0.1843, ..., -0.0711, 0.1799, -0.0337], + ..., + [-0.1898, -0.1105, 0.1340, ..., 0.0306, -0.2345, -0.0030], + [ 0.1137, -0.0168, -0.3379, ..., -0.0693, 0.1854, -0.0380], + [-0.2396, -0.1821, 0.0371, ..., -0.0168, -0.2627, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 6.9849e-10, 2.3283e-10, ..., 0.0000e+00, + 2.5611e-09, 0.0000e+00], + [ 1.6298e-09, 3.9581e-09, 2.3283e-10, ..., 0.0000e+00, + 4.4238e-09, 0.0000e+00], + [-1.1874e-08, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, + -2.7241e-08, 0.0000e+00], + ..., + [ 4.1910e-09, 7.2177e-09, 2.3283e-10, ..., 0.0000e+00, + 1.1642e-08, 0.0000e+00], + [ 4.6566e-09, 1.6298e-09, 2.3283e-10, ..., 0.0000e+00, + 1.1176e-08, 0.0000e+00], + [ 0.0000e+00, 1.5204e-07, 6.9849e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0060, 0.0137, -0.0218, 0.0191, 0.0268, -0.0172, -0.0177, -0.0108, + -0.0348, -0.0321], device='cuda:0'), grad: tensor([ 4.8894e-09, 5.9139e-08, -5.6811e-08, 1.9791e-08, -5.3737e-07, + 8.5216e-08, 2.7474e-08, -4.2142e-08, 5.3085e-08, 4.0536e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 217.03, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4203 re_mapping 0.0021 re_causal 0.0081 /// teacc 99.11 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.3302, -0.1892, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1559, 0.1462, -0.0952, ..., 0.0245, -0.0697, -0.0819], + [ 0.1515, -0.1910, -0.1843, ..., -0.0711, 0.1800, -0.0337], + ..., + [-0.1898, -0.1105, 0.1341, ..., 0.0306, -0.2346, -0.0030], + [ 0.1137, -0.0169, -0.3379, ..., -0.0693, 0.1855, -0.0380], + [-0.2396, -0.1821, 0.0371, ..., -0.0168, -0.2627, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 1.1642e-09, -9.0804e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, + 6.7521e-09, 0.0000e+00], + [ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + ..., + [ 2.3283e-10, 1.6298e-09, -4.6566e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-3.7253e-09, -9.3132e-10, 9.3132e-10, ..., 0.0000e+00, + -3.0268e-09, 0.0000e+00], + [ 6.9849e-10, 1.4668e-08, 2.3283e-10, ..., 0.0000e+00, + 1.6298e-09, 0.0000e+00]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0060, 0.0136, -0.0218, 0.0191, 0.0268, -0.0172, -0.0176, -0.0108, + -0.0348, -0.0321], device='cuda:0'), grad: tensor([-1.9907e-07, 5.2853e-08, 6.1234e-08, 1.9791e-08, 3.5623e-08, + 1.0012e-07, 2.4447e-08, -1.2922e-07, 3.9581e-09, 3.7719e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 216.83, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.3931 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.12 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.3302, -0.1892, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1560, 0.1463, -0.0952, ..., 0.0245, -0.0698, -0.0819], + [ 0.1516, -0.1910, -0.1844, ..., -0.0711, 0.1801, -0.0337], + ..., + [-0.1899, -0.1106, 0.1342, ..., 0.0306, -0.2348, -0.0030], + [ 0.1137, -0.0169, -0.3380, ..., -0.0693, 0.1855, -0.0380], + [-0.2397, -0.1821, 0.0371, ..., -0.0168, -0.2627, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.0955e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 6.9849e-10, 3.0268e-09, 2.3283e-10, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + -3.4925e-09, 0.0000e+00], + ..., + [ 2.3283e-10, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-09, 0.0000e+00], + [-9.5461e-09, -6.0536e-09, 0.0000e+00, ..., 0.0000e+00, + -8.1491e-09, 0.0000e+00], + [ 1.1642e-09, 9.1968e-08, 1.6298e-09, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0060, 0.0137, -0.0217, 0.0192, 0.0268, -0.0172, -0.0177, -0.0109, + -0.0349, -0.0321], device='cuda:0'), grad: tensor([ 3.4925e-09, 1.5832e-08, -5.1223e-09, 1.4901e-08, 6.8452e-08, + 2.9104e-08, -8.1724e-08, 7.2876e-08, -2.3283e-08, -7.2876e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 216.84, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4043 re_mapping 0.0021 re_causal 0.0081 /// teacc 99.11 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.3303, -0.1893, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1560, 0.1463, -0.0952, ..., 0.0245, -0.0698, -0.0819], + [ 0.1517, -0.1911, -0.1843, ..., -0.0711, 0.1803, -0.0337], + ..., + [-0.1899, -0.1106, 0.1342, ..., 0.0306, -0.2349, -0.0030], + [ 0.1137, -0.0169, -0.3380, ..., -0.0693, 0.1856, -0.0380], + [-0.2398, -0.1822, 0.0371, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 6.9849e-10, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -1.1409e-08, 2.3283e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 6.2864e-09, -4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.9849e-10, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.0268e-09, -3.4925e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0060, 0.0137, -0.0217, 0.0192, 0.0268, -0.0173, -0.0177, -0.0109, + -0.0349, -0.0321], device='cuda:0'), grad: tensor([-8.3819e-09, -1.2573e-08, 6.9849e-09, 7.6834e-09, 5.6811e-08, + 6.9849e-09, 7.2177e-09, 1.6298e-09, 1.1409e-08, -5.5181e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 217.01, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4287 re_mapping 0.0021 re_causal 0.0083 /// teacc 99.13 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.3303, -0.1893, -0.0969, ..., -0.0041, -0.0694, -0.0169], + [-0.1562, 0.1464, -0.0952, ..., 0.0245, -0.0699, -0.0819], + [ 0.1518, -0.1912, -0.1844, ..., -0.0711, 0.1804, -0.0337], + ..., + [-0.1898, -0.1107, 0.1342, ..., 0.0306, -0.2350, -0.0030], + [ 0.1138, -0.0170, -0.3380, ..., -0.0693, 0.1857, -0.0380], + [-0.2399, -0.1822, 0.0371, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + 1.0943e-08, 0.0000e+00], + [ 0.0000e+00, -1.0710e-08, 2.3283e-10, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8859e-08, 6.7521e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 2.3283e-10, 6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00], + [ 2.3283e-10, 5.5879e-09, -2.9569e-08, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0060, 0.0136, -0.0217, 0.0192, 0.0268, -0.0173, -0.0177, -0.0108, + -0.0349, -0.0321], device='cuda:0'), grad: tensor([ 6.4261e-08, -1.3970e-08, 1.1642e-08, -4.9127e-08, 4.8662e-08, + 3.0734e-08, 3.0268e-09, 1.5064e-07, 8.8476e-09, -2.4587e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 216.94, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3987 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.12 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.3304, -0.1894, -0.0970, ..., -0.0041, -0.0694, -0.0169], + [-0.1562, 0.1464, -0.0952, ..., 0.0245, -0.0700, -0.0819], + [ 0.1520, -0.1913, -0.1844, ..., -0.0711, 0.1805, -0.0337], + ..., + [-0.1899, -0.1107, 0.1343, ..., 0.0306, -0.2352, -0.0030], + [ 0.1138, -0.0171, -0.3380, ..., -0.0693, 0.1857, -0.0380], + [-0.2399, -0.1823, 0.0371, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-09, 0.0000e+00], + [ 9.3132e-10, -3.2596e-09, 2.3283e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [-4.6566e-09, 9.3132e-10, -2.3283e-10, ..., 0.0000e+00, + -2.8405e-08, 0.0000e+00], + ..., + [ 6.9849e-10, 3.4925e-09, 0.0000e+00, ..., 0.0000e+00, + 4.8894e-09, 0.0000e+00], + [ 1.2340e-08, 1.0012e-08, 0.0000e+00, ..., 0.0000e+00, + 3.0268e-09, 0.0000e+00], + [ 0.0000e+00, 1.3970e-09, -2.7940e-09, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0060, 0.0136, -0.0216, 0.0192, 0.0268, -0.0173, -0.0177, -0.0109, + -0.0349, -0.0321], device='cuda:0'), grad: tensor([ 1.0943e-08, 6.9849e-10, -8.7079e-08, 9.4529e-08, 2.2352e-08, + 4.4005e-08, -7.8697e-08, -6.2864e-09, 3.7020e-08, -2.0023e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 216.78, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4000 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.13 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.3304, -0.1894, -0.0970, ..., -0.0041, -0.0694, -0.0169], + [-0.1563, 0.1464, -0.0953, ..., 0.0245, -0.0700, -0.0819], + [ 0.1520, -0.1913, -0.1844, ..., -0.0711, 0.1806, -0.0337], + ..., + [-0.1899, -0.1107, 0.1343, ..., 0.0306, -0.2353, -0.0030], + [ 0.1138, -0.0171, -0.3380, ..., -0.0693, 0.1858, -0.0380], + [-0.2400, -0.1823, 0.0371, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [-2.3283e-10, 1.1642e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-7.2177e-09, -4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + -6.0536e-09, 0.0000e+00], + [ 1.1642e-09, 3.4925e-09, -6.9849e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0059, 0.0136, -0.0216, 0.0193, 0.0268, -0.0174, -0.0177, -0.0109, + -0.0350, -0.0321], device='cuda:0'), grad: tensor([ 7.6834e-09, 9.0804e-09, 2.4959e-07, 2.9337e-08, 5.0291e-08, + -1.6065e-08, 6.5193e-09, -2.5728e-07, -1.4901e-08, -6.1933e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.13, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4330 re_mapping 0.0021 re_causal 0.0084 /// teacc 99.15 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.3304, -0.1895, -0.0970, ..., -0.0041, -0.0694, -0.0169], + [-0.1564, 0.1466, -0.0953, ..., 0.0245, -0.0701, -0.0819], + [ 0.1522, -0.1914, -0.1844, ..., -0.0711, 0.1808, -0.0337], + ..., + [-0.1900, -0.1108, 0.1344, ..., 0.0306, -0.2354, -0.0030], + [ 0.1138, -0.0172, -0.3381, ..., -0.0693, 0.1858, -0.0380], + [-0.2401, -0.1824, 0.0372, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 4.4238e-09, 0.0000e+00, ..., 0.0000e+00, + 4.4238e-09, 0.0000e+00], + [-2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 8.3819e-09, 0.0000e+00], + ..., + [ 2.3283e-09, 3.2596e-09, 1.3970e-09, ..., 0.0000e+00, + 8.6147e-09, 0.0000e+00], + [-2.5611e-09, -9.3132e-10, 2.3283e-10, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 4.6566e-10, 3.8184e-08, -8.8476e-09, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0059, 0.0137, -0.0216, 0.0192, 0.0268, -0.0174, -0.0178, -0.0109, + -0.0351, -0.0321], device='cuda:0'), grad: tensor([ 2.6543e-08, 4.0513e-08, 2.9802e-08, -6.6357e-08, -1.8743e-07, + 3.6322e-08, 2.1653e-08, 5.7742e-08, 6.0536e-09, 4.0513e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 216.82, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4040 re_mapping 0.0022 re_causal 0.0082 /// teacc 99.14 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.3305, -0.1895, -0.0970, ..., -0.0041, -0.0694, -0.0169], + [-0.1564, 0.1466, -0.0954, ..., 0.0245, -0.0702, -0.0819], + [ 0.1522, -0.1915, -0.1844, ..., -0.0711, 0.1809, -0.0337], + ..., + [-0.1901, -0.1108, 0.1345, ..., 0.0306, -0.2355, -0.0030], + [ 0.1137, -0.0172, -0.3381, ..., -0.0693, 0.1859, -0.0380], + [-0.2401, -0.1825, 0.0372, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 2.5611e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 1.6298e-09, -2.0955e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 1.8626e-09, 1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + ..., + [ 4.6566e-10, 4.1910e-09, 2.3283e-10, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [-1.5832e-08, -2.0955e-09, 0.0000e+00, ..., 0.0000e+00, + -1.5367e-08, 0.0000e+00], + [ 4.6566e-10, 3.0617e-07, -9.3132e-10, ..., 0.0000e+00, + 1.1642e-09, 0.0000e+00]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0059, 0.0136, -0.0216, 0.0193, 0.0269, -0.0173, -0.0178, -0.0109, + -0.0352, -0.0321], device='cuda:0'), grad: tensor([-8.1491e-09, 1.0943e-08, 1.6764e-08, -1.2806e-08, -9.4203e-07, + 4.0745e-08, 7.2643e-08, 3.9116e-08, -4.5635e-08, 8.4378e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 216.77, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4559 re_mapping 0.0021 re_causal 0.0084 /// teacc 99.12 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.3305, -0.1898, -0.0970, ..., -0.0041, -0.0695, -0.0169], + [-0.1565, 0.1466, -0.0954, ..., 0.0245, -0.0703, -0.0819], + [ 0.1524, -0.1916, -0.1844, ..., -0.0711, 0.1810, -0.0337], + ..., + [-0.1902, -0.1109, 0.1346, ..., 0.0306, -0.2357, -0.0030], + [ 0.1137, -0.0172, -0.3381, ..., -0.0693, 0.1860, -0.0380], + [-0.2403, -0.1826, 0.0372, ..., -0.0168, -0.2628, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-10, 6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, -6.0536e-09, 0.0000e+00, ..., 0.0000e+00, + -1.1642e-09, 0.0000e+00], + [-6.2864e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -1.0477e-08, 0.0000e+00], + ..., + [ 1.3970e-09, 3.4925e-09, 2.3283e-10, ..., 0.0000e+00, + 5.8208e-09, 0.0000e+00], + [-4.1910e-09, 2.5611e-09, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 2.3283e-10, -9.3132e-10, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0059, 0.0136, -0.0215, 0.0194, 0.0269, -0.0173, -0.0179, -0.0110, + -0.0352, -0.0322], device='cuda:0'), grad: tensor([ 8.6147e-09, 2.3283e-09, -7.9162e-09, -2.0000e-07, 1.6764e-08, + 1.8580e-07, 3.2596e-09, 5.3551e-09, 3.4925e-09, -6.9849e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 216.89, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4111 re_mapping 0.0021 re_causal 0.0083 /// teacc 99.10 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.3306, -0.1898, -0.0970, ..., -0.0041, -0.0695, -0.0169], + [-0.1566, 0.1466, -0.0954, ..., 0.0245, -0.0704, -0.0819], + [ 0.1525, -0.1917, -0.1844, ..., -0.0711, 0.1812, -0.0337], + ..., + [-0.1903, -0.1109, 0.1346, ..., 0.0306, -0.2360, -0.0030], + [ 0.1137, -0.0173, -0.3381, ..., -0.0693, 0.1861, -0.0380], + [-0.2404, -0.1827, 0.0371, ..., -0.0168, -0.2629, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, -5.1223e-09, 0.0000e+00, ..., 0.0000e+00, + 4.2375e-08, 0.0000e+00], + [ 9.3132e-10, 1.1874e-08, 1.6298e-09, ..., 0.0000e+00, + 2.9104e-08, 0.0000e+00], + [ 3.8883e-08, 1.2340e-08, 2.3283e-10, ..., 0.0000e+00, + 7.8697e-08, 0.0000e+00], + ..., + [ 2.3283e-10, -4.6566e-10, -2.0955e-09, ..., 0.0000e+00, + 4.0047e-08, 0.0000e+00], + [ 1.6298e-09, 1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + 3.2131e-08, 0.0000e+00], + [ 6.9849e-10, 4.1677e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3749e-08, 0.0000e+00]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0059, 0.0135, -0.0214, 0.0194, 0.0270, -0.0174, -0.0179, -0.0110, + -0.0353, -0.0322], device='cuda:0'), grad: tensor([ 6.2399e-08, 1.2759e-07, 2.9290e-07, -7.1712e-07, -3.8021e-07, + 6.2631e-08, 1.1036e-07, 1.1479e-07, 1.0617e-07, 2.2398e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 217.01, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3916 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.10 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.3307, -0.1899, -0.0970, ..., -0.0041, -0.0695, -0.0169], + [-0.1566, 0.1466, -0.0955, ..., 0.0245, -0.0704, -0.0819], + [ 0.1526, -0.1917, -0.1844, ..., -0.0711, 0.1813, -0.0337], + ..., + [-0.1903, -0.1109, 0.1347, ..., 0.0306, -0.2361, -0.0030], + [ 0.1137, -0.0173, -0.3381, ..., -0.0693, 0.1861, -0.0380], + [-0.2405, -0.1828, 0.0371, ..., -0.0168, -0.2629, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.0804e-09, 0.0000e+00], + [ 0.0000e+00, -3.4925e-09, 1.3970e-09, ..., 0.0000e+00, + 1.1874e-08, 0.0000e+00], + [-6.9849e-10, 6.9849e-10, 4.6566e-10, ..., 0.0000e+00, + -3.1432e-08, 0.0000e+00], + ..., + [ 4.6566e-10, 4.4238e-09, -6.2864e-09, ..., 0.0000e+00, + 9.3132e-09, 0.0000e+00], + [ 5.3551e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 4.6566e-10, 6.7521e-09, 1.1642e-09, ..., 0.0000e+00, + 1.6298e-09, 0.0000e+00]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0059, 0.0135, -0.0214, 0.0193, 0.0270, -0.0173, -0.0179, -0.0110, + -0.0353, -0.0322], device='cuda:0'), grad: tensor([-1.0408e-07, 3.6554e-08, -3.1665e-08, 1.4901e-08, -7.2177e-09, + -5.3551e-08, 7.0781e-08, -6.7521e-09, 2.2817e-08, 7.5204e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 216.89, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3974 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.13 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.3307, -0.1900, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1567, 0.1466, -0.0955, ..., 0.0245, -0.0704, -0.0819], + [ 0.1527, -0.1918, -0.1844, ..., -0.0711, 0.1814, -0.0337], + ..., + [-0.1904, -0.1109, 0.1347, ..., 0.0306, -0.2362, -0.0030], + [ 0.1137, -0.0174, -0.3382, ..., -0.0693, 0.1862, -0.0380], + [-0.2407, -0.1829, 0.0371, ..., -0.0168, -0.2629, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 2.3283e-10, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 4.6566e-09, -1.6298e-09, 0.0000e+00, ..., 0.0000e+00, + 5.3551e-09, 0.0000e+00], + [-1.0524e-07, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, + -1.1642e-07, 0.0000e+00], + ..., + [ 9.3365e-08, 1.3970e-09, 4.6566e-10, ..., 0.0000e+00, + 1.0245e-07, 0.0000e+00], + [ 4.6566e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 3.9581e-09, 3.2596e-09, -2.7940e-09, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0059, 0.0135, -0.0214, 0.0193, 0.0270, -0.0172, -0.0179, -0.0110, + -0.0353, -0.0323], device='cuda:0'), grad: tensor([ 6.7521e-09, 2.2585e-08, -4.4098e-07, 2.9569e-08, -3.7253e-09, + -3.7719e-08, 5.8208e-09, 3.7998e-07, 2.0955e-08, 2.4913e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 216.79, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4315 re_mapping 0.0021 re_causal 0.0082 /// teacc 99.10 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.3308, -0.1901, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1567, 0.1467, -0.0955, ..., 0.0245, -0.0706, -0.0819], + [ 0.1528, -0.1919, -0.1845, ..., -0.0711, 0.1816, -0.0337], + ..., + [-0.1905, -0.1110, 0.1348, ..., 0.0306, -0.2364, -0.0030], + [ 0.1138, -0.0174, -0.3382, ..., -0.0693, 0.1863, -0.0380], + [-0.2409, -0.1830, 0.0371, ..., -0.0168, -0.2630, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.2596e-09, 6.9849e-10, 4.6566e-10, ..., 0.0000e+00, + 5.3551e-09, 0.0000e+00], + [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 8.1491e-09, 0.0000e+00], + [-8.8010e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.7253e-07, 0.0000e+00], + ..., + [ 6.3796e-08, 2.3283e-10, 9.3132e-10, ..., 0.0000e+00, + 1.1665e-07, 0.0000e+00], + [ 1.0477e-08, 4.6566e-10, 2.3283e-10, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 1.1642e-09, -2.7707e-08, -4.8894e-09, ..., 0.0000e+00, + 1.6298e-09, 0.0000e+00]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0059, 0.0135, -0.0214, 0.0193, 0.0271, -0.0171, -0.0179, -0.0110, + -0.0353, -0.0323], device='cuda:0'), grad: tensor([ 3.0966e-08, 4.0978e-08, -7.0548e-07, 1.3900e-07, 6.5612e-07, + -3.9116e-08, 2.6543e-08, 4.6776e-07, 3.7951e-08, -6.3656e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 217.07, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4038 re_mapping 0.0020 re_causal 0.0081 /// teacc 99.13 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.3309, -0.1904, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1570, 0.1468, -0.0956, ..., 0.0245, -0.0707, -0.0819], + [ 0.1530, -0.1920, -0.1845, ..., -0.0711, 0.1818, -0.0337], + ..., + [-0.1905, -0.1111, 0.1349, ..., 0.0306, -0.2366, -0.0030], + [ 0.1138, -0.0174, -0.3382, ..., -0.0693, 0.1863, -0.0380], + [-0.2410, -0.1831, 0.0371, ..., -0.0168, -0.2630, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 6.9849e-10, 0.0000e+00], + [ 0.0000e+00, -1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [-2.5611e-09, 6.9849e-10, 0.0000e+00, ..., 0.0000e+00, + -4.6566e-09, 0.0000e+00], + ..., + [ 4.6566e-10, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.1642e-09, 0.0000e+00], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 2.3283e-10, 1.0710e-08, 0.0000e+00, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0059, 0.0135, -0.0213, 0.0193, 0.0271, -0.0171, -0.0178, -0.0111, + -0.0354, -0.0323], device='cuda:0'), grad: tensor([-1.1409e-08, -2.3283e-10, -6.7521e-09, -2.9569e-08, -2.7241e-08, + 2.3050e-08, 3.4925e-09, 1.5134e-08, 7.6834e-09, 3.3062e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 216.94, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4268 re_mapping 0.0020 re_causal 0.0083 /// teacc 99.12 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.3310, -0.1904, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1570, 0.1468, -0.0956, ..., 0.0245, -0.0708, -0.0819], + [ 0.1531, -0.1921, -0.1845, ..., -0.0711, 0.1820, -0.0337], + ..., + [-0.1905, -0.1111, 0.1349, ..., 0.0306, -0.2367, -0.0030], + [ 0.1138, -0.0175, -0.3382, ..., -0.0693, 0.1864, -0.0380], + [-0.2411, -0.1832, 0.0370, ..., -0.0168, -0.2630, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.3970e-09, 2.3283e-10, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 6.9849e-10, 8.8476e-09, 7.4506e-09, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + [ 4.6566e-10, 1.6298e-09, 1.1642e-09, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00], + ..., + [ 4.6566e-10, -8.8476e-09, -1.1409e-08, ..., 0.0000e+00, + 1.1642e-09, 0.0000e+00], + [-2.4680e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8859e-08, 0.0000e+00], + [ 0.0000e+00, 6.7288e-08, 3.4925e-09, ..., 0.0000e+00, + 2.3283e-10, 0.0000e+00]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0059, 0.0135, -0.0212, 0.0193, 0.0271, -0.0171, -0.0177, -0.0111, + -0.0354, -0.0324], device='cuda:0'), grad: tensor([ 9.7789e-09, 8.3353e-08, 1.4435e-08, 1.1642e-09, -1.3295e-07, + 4.6799e-08, -1.3271e-08, -1.1292e-07, -3.3993e-08, 1.4040e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 217.03, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4239 re_mapping 0.0021 re_causal 0.0083 /// teacc 99.13 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.3311, -0.1906, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1570, 0.1468, -0.0956, ..., 0.0245, -0.0708, -0.0819], + [ 0.1532, -0.1922, -0.1845, ..., -0.0711, 0.1821, -0.0337], + ..., + [-0.1906, -0.1111, 0.1350, ..., 0.0306, -0.2369, -0.0030], + [ 0.1138, -0.0176, -0.3382, ..., -0.0693, 0.1864, -0.0380], + [-0.2411, -0.1833, 0.0370, ..., -0.0168, -0.2630, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 4.1910e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.0710e-08, -5.5879e-09, -1.3970e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 6.0536e-09, 6.9849e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 1.7695e-08, 2.7940e-09, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 1.8626e-09, 1.3970e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 6.0536e-09, -3.2596e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0060, 0.0135, -0.0211, 0.0193, 0.0272, -0.0171, -0.0177, -0.0111, + -0.0355, -0.0324], device='cuda:0'), grad: tensor([ 1.4901e-08, 8.8476e-09, 1.9092e-08, 2.7940e-08, 1.0524e-07, + 5.5879e-09, -1.0850e-07, 6.1002e-08, 1.2573e-08, -1.3271e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 216.86, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4111 re_mapping 0.0020 re_causal 0.0081 /// teacc 99.14 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.3311, -0.1906, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1572, 0.1469, -0.0957, ..., 0.0245, -0.0708, -0.0819], + [ 0.1534, -0.1923, -0.1845, ..., -0.0711, 0.1823, -0.0337], + ..., + [-0.1906, -0.1112, 0.1351, ..., 0.0306, -0.2371, -0.0030], + [ 0.1138, -0.0176, -0.3382, ..., -0.0693, 0.1865, -0.0380], + [-0.2412, -0.1834, 0.0370, ..., -0.0168, -0.2630, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-10, 0.0000e+00], + [ 6.0536e-09, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [-1.3504e-08, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, + -1.7695e-08, 0.0000e+00], + ..., + [ 5.1223e-09, 1.8626e-09, -4.6566e-10, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 2.3283e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0060, 0.0134, -0.0210, 0.0193, 0.0272, -0.0172, -0.0177, -0.0111, + -0.0355, -0.0325], device='cuda:0'), grad: tensor([-2.7940e-09, 3.3993e-08, -1.1176e-08, 2.7940e-09, 2.1886e-08, + 3.2596e-09, -2.7940e-09, -3.3528e-08, 1.0710e-08, -1.3970e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 216.95, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4346 re_mapping 0.0020 re_causal 0.0083 /// teacc 99.08 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.3312, -0.1907, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1573, 0.1469, -0.0958, ..., 0.0245, -0.0708, -0.0819], + [ 0.1535, -0.1924, -0.1845, ..., -0.0711, 0.1825, -0.0337], + ..., + [-0.1906, -0.1112, 0.1353, ..., 0.0306, -0.2373, -0.0030], + [ 0.1138, -0.0177, -0.3382, ..., -0.0693, 0.1865, -0.0380], + [-0.2413, -0.1836, 0.0369, ..., -0.0168, -0.2631, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 2.3283e-09, 1.8626e-09, 4.6566e-10, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00], + [ 1.1642e-08, 1.6298e-08, 0.0000e+00, ..., 0.0000e+00, + 1.0710e-08, 0.0000e+00], + ..., + [ 4.6566e-10, 3.2596e-09, -1.8626e-09, ..., 0.0000e+00, + 3.2596e-09, 0.0000e+00], + [ 6.5193e-09, 6.0536e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [ 4.6566e-09, 3.8650e-08, 2.7008e-08, ..., 0.0000e+00, + 2.3283e-09, 0.0000e+00]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0060, 0.0134, -0.0210, 0.0193, 0.0274, -0.0172, -0.0176, -0.0111, + -0.0357, -0.0326], device='cuda:0'), grad: tensor([ 2.3749e-08, 1.8161e-08, 6.0536e-08, 8.1491e-08, -1.1176e-08, + -1.3039e-07, -1.6764e-07, -1.6298e-08, 3.0268e-08, 1.2014e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 216.96, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4206 re_mapping 0.0020 re_causal 0.0080 /// teacc 99.10 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.3313, -0.1908, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1573, 0.1469, -0.0959, ..., 0.0245, -0.0709, -0.0819], + [ 0.1536, -0.1925, -0.1845, ..., -0.0711, 0.1826, -0.0337], + ..., + [-0.1907, -0.1112, 0.1353, ..., 0.0306, -0.2374, -0.0030], + [ 0.1138, -0.0178, -0.3382, ..., -0.0693, 0.1866, -0.0380], + [-0.2414, -0.1838, 0.0369, ..., -0.0168, -0.2631, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -3.0780e-07, 4.6566e-10, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + [ 4.6566e-10, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 4.6566e-10, 9.1270e-08, -1.3970e-09, ..., 0.0000e+00, + 1.3970e-09, 0.0000e+00], + [ 3.7253e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 9.1735e-08, 4.6566e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0060, 0.0133, -0.0210, 0.0193, 0.0274, -0.0173, -0.0176, -0.0110, + -0.0357, -0.0326], device='cuda:0'), grad: tensor([ 1.4435e-08, -5.9931e-07, 1.4435e-08, 2.2817e-08, 1.7509e-07, + 2.9337e-08, -3.9116e-08, 1.1083e-07, 2.2352e-08, 2.5658e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 216.87, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4028 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.13 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.3314, -0.1910, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1574, 0.1470, -0.0960, ..., 0.0245, -0.0710, -0.0819], + [ 0.1537, -0.1926, -0.1845, ..., -0.0711, 0.1827, -0.0337], + ..., + [-0.1907, -0.1113, 0.1354, ..., 0.0306, -0.2376, -0.0030], + [ 0.1138, -0.0178, -0.3382, ..., -0.0693, 0.1867, -0.0380], + [-0.2415, -0.1841, 0.0369, ..., -0.0168, -0.2632, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.0047e-08, 2.4214e-08, 0.0000e+00, ..., 0.0000e+00, + 1.5367e-08, 0.0000e+00], + [ 3.7253e-09, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + 7.9162e-09, 0.0000e+00], + [-9.0804e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8533e-07, 0.0000e+00], + ..., + [ 2.4680e-08, 4.1910e-09, 4.6566e-10, ..., 0.0000e+00, + 4.4703e-08, 0.0000e+00], + [ 9.3132e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.4435e-08, 0.0000e+00], + [ 1.8626e-09, 3.1199e-08, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0060, 0.0133, -0.0209, 0.0194, 0.0276, -0.0174, -0.0175, -0.0110, + -0.0357, -0.0328], device='cuda:0'), grad: tensor([ 1.2619e-07, 2.5611e-08, -5.3970e-07, 2.0303e-07, -1.1222e-07, + 4.0047e-08, -6.3330e-08, 1.5460e-07, 5.2620e-08, 1.1688e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 216.85, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4165 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.12 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.3315, -0.1911, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1574, 0.1471, -0.0959, ..., 0.0245, -0.0712, -0.0819], + [ 0.1539, -0.1927, -0.1845, ..., -0.0711, 0.1829, -0.0337], + ..., + [-0.1908, -0.1113, 0.1354, ..., 0.0306, -0.2375, -0.0030], + [ 0.1138, -0.0179, -0.3382, ..., -0.0693, 0.1868, -0.0380], + [-0.2416, -0.1842, 0.0369, ..., -0.0168, -0.2632, -0.0731]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -4.0047e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 4.0978e-08, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + ..., + [-9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 8.3819e-09, 7.4506e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 1.1735e-07, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0060, 0.0133, -0.0210, 0.0193, 0.0277, -0.0173, -0.0175, -0.0110, + -0.0358, -0.0329], device='cuda:0'), grad: tensor([-2.1420e-08, -7.7300e-08, 9.0338e-08, 3.7253e-08, -4.2934e-07, + -1.3039e-08, -1.8626e-09, -9.3132e-09, 2.7940e-08, 4.0233e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 216.97, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4368 re_mapping 0.0021 re_causal 0.0084 /// teacc 99.14 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.3316, -0.1911, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1575, 0.1471, -0.0960, ..., 0.0245, -0.0712, -0.0819], + [ 0.1539, -0.1928, -0.1845, ..., -0.0711, 0.1830, -0.0337], + ..., + [-0.1908, -0.1114, 0.1355, ..., 0.0306, -0.2376, -0.0030], + [ 0.1139, -0.0180, -0.3382, ..., -0.0693, 0.1869, -0.0380], + [-0.2417, -0.1844, 0.0369, ..., -0.0168, -0.2632, -0.0731]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -2.9802e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 2.1420e-08, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-2.7940e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + [ 3.7253e-09, 9.3132e-10, -2.7940e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0058, 0.0133, -0.0210, 0.0193, 0.0277, -0.0173, -0.0175, -0.0110, + -0.0357, -0.0330], device='cuda:0'), grad: tensor([ 1.4901e-08, -3.4459e-08, 4.6566e-09, -2.3656e-07, 5.8673e-08, + 2.6263e-07, -5.4948e-08, 4.2841e-08, -7.4506e-09, -4.3772e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 216.68, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4139 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.13 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.3316, -0.1913, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1575, 0.1472, -0.0959, ..., 0.0245, -0.0713, -0.0819], + [ 0.1541, -0.1929, -0.1845, ..., -0.0711, 0.1832, -0.0337], + ..., + [-0.1909, -0.1115, 0.1355, ..., 0.0306, -0.2378, -0.0030], + [ 0.1140, -0.0180, -0.3382, ..., -0.0693, 0.1871, -0.0380], + [-0.2418, -0.1845, 0.0368, ..., -0.0168, -0.2633, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + ..., + [ 0.0000e+00, -9.3132e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.1176e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0058, 0.0133, -0.0208, 0.0192, 0.0278, -0.0174, -0.0174, -0.0110, + -0.0357, -0.0330], device='cuda:0'), grad: tensor([-2.7940e-09, 1.1828e-07, -1.0245e-08, 1.1176e-08, -1.0245e-08, + 7.4506e-09, 7.4506e-09, -8.0094e-08, 5.5879e-09, -4.0047e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 217.13, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4438 re_mapping 0.0021 re_causal 0.0082 /// teacc 99.13 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.3318, -0.1914, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1576, 0.1473, -0.0958, ..., 0.0245, -0.0715, -0.0819], + [ 0.1542, -0.1930, -0.1845, ..., -0.0711, 0.1835, -0.0337], + ..., + [-0.1910, -0.1116, 0.1355, ..., 0.0306, -0.2379, -0.0030], + [ 0.1141, -0.0181, -0.3383, ..., -0.0693, 0.1873, -0.0380], + [-0.2419, -0.1846, 0.0368, ..., -0.0168, -0.2633, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.2107e-08, 0.0000e+00], + ..., + [ 2.7940e-09, 1.8626e-09, 1.8626e-09, ..., 0.0000e+00, + 5.5879e-09, 0.0000e+00], + [-9.3132e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, -5.5879e-09, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0058, 0.0132, -0.0206, 0.0192, 0.0278, -0.0173, -0.0172, -0.0111, + -0.0356, -0.0331], device='cuda:0'), grad: tensor([ 1.8626e-08, 2.7940e-08, -3.1665e-08, -3.8184e-08, 1.3411e-07, + 4.1910e-08, -1.8626e-09, 5.4017e-08, -2.7940e-08, -1.5646e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 216.63, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4014 re_mapping 0.0021 re_causal 0.0079 /// teacc 99.12 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.3319, -0.1917, -0.0971, ..., -0.0041, -0.0695, -0.0169], + [-0.1578, 0.1473, -0.0959, ..., 0.0245, -0.0716, -0.0819], + [ 0.1544, -0.1931, -0.1845, ..., -0.0711, 0.1838, -0.0337], + ..., + [-0.1910, -0.1116, 0.1356, ..., 0.0306, -0.2381, -0.0030], + [ 0.1142, -0.0182, -0.3383, ..., -0.0693, 0.1875, -0.0380], + [-0.2420, -0.1848, 0.0368, ..., -0.0168, -0.2633, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0058, 0.0131, -0.0204, 0.0191, 0.0278, -0.0173, -0.0170, -0.0110, + -0.0357, -0.0331], device='cuda:0'), grad: tensor([ 1.8626e-09, -9.3132e-09, -2.7940e-09, 2.9802e-08, 9.8720e-08, + -1.8626e-09, 9.3132e-10, 9.3132e-10, 3.7253e-09, -1.1921e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 216.78, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3907 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.13 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.3320, -0.1918, -0.0971, ..., -0.0041, -0.0696, -0.0169], + [-0.1578, 0.1474, -0.0958, ..., 0.0245, -0.0717, -0.0819], + [ 0.1546, -0.1932, -0.1845, ..., -0.0711, 0.1840, -0.0337], + ..., + [-0.1911, -0.1117, 0.1355, ..., 0.0306, -0.2383, -0.0030], + [ 0.1142, -0.0182, -0.3383, ..., -0.0693, 0.1876, -0.0380], + [-0.2421, -0.1849, 0.0368, ..., -0.0168, -0.2634, -0.0731]], + device='cuda:0'), grad: tensor([[ 5.5879e-09, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 4.6566e-08, 1.8626e-09, ..., 0.0000e+00, + 3.7253e-09, 0.0000e+00], + [ 1.8626e-09, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 1.3039e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 6.5193e-09, 2.5146e-08, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 1.7509e-07, 5.5879e-09, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0058, 0.0131, -0.0203, 0.0190, 0.0279, -0.0173, -0.0169, -0.0111, + -0.0357, -0.0332], device='cuda:0'), grad: tensor([ 2.6077e-08, 1.4156e-07, 1.7695e-08, -3.1665e-08, -7.1060e-07, + 5.1223e-08, -8.2888e-08, 3.6322e-08, 6.7987e-08, 4.8708e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 216.79, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4238 re_mapping 0.0020 re_causal 0.0082 /// teacc 99.15 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.3321, -0.1919, -0.0971, ..., -0.0041, -0.0696, -0.0169], + [-0.1579, 0.1476, -0.0958, ..., 0.0245, -0.0717, -0.0819], + [ 0.1547, -0.1933, -0.1845, ..., -0.0711, 0.1842, -0.0337], + ..., + [-0.1911, -0.1118, 0.1356, ..., 0.0306, -0.2384, -0.0030], + [ 0.1143, -0.0183, -0.3383, ..., -0.0693, 0.1877, -0.0380], + [-0.2422, -0.1850, 0.0368, ..., -0.0168, -0.2634, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -3.7253e-09, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0058, 0.0132, -0.0202, 0.0190, 0.0279, -0.0173, -0.0168, -0.0111, + -0.0357, -0.0332], device='cuda:0'), grad: tensor([-8.3819e-09, 9.3132e-10, 3.7253e-09, 6.5193e-09, -9.3132e-10, + -3.7253e-09, 1.8626e-09, 9.3132e-10, -4.6566e-09, 1.5832e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 216.66, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4396 re_mapping 0.0021 re_causal 0.0083 /// teacc 99.13 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.3322, -0.1920, -0.0971, ..., -0.0041, -0.0696, -0.0169], + [-0.1580, 0.1477, -0.0957, ..., 0.0245, -0.0718, -0.0819], + [ 0.1549, -0.1933, -0.1845, ..., -0.0711, 0.1844, -0.0337], + ..., + [-0.1912, -0.1119, 0.1356, ..., 0.0306, -0.2386, -0.0030], + [ 0.1144, -0.0183, -0.3383, ..., -0.0693, 0.1879, -0.0380], + [-0.2424, -0.1852, 0.0367, ..., -0.0168, -0.2634, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 1.8626e-09, -6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-9.3132e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -1.3039e-08, 0.0000e+00], + ..., + [ 7.4506e-09, 5.5879e-09, -9.3132e-10, ..., 0.0000e+00, + 1.0245e-08, 0.0000e+00], + [ 5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, -5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0058, 0.0132, -0.0200, 0.0190, 0.0279, -0.0175, -0.0168, -0.0111, + -0.0357, -0.0333], device='cuda:0'), grad: tensor([-1.0524e-07, 8.3819e-09, -2.7940e-08, 2.7847e-07, 1.2200e-07, + -4.5449e-07, 1.7416e-07, 4.6566e-08, 2.6077e-08, -6.8918e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 216.87, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.3920 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.11 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.3323, -0.1921, -0.0971, ..., -0.0041, -0.0696, -0.0169], + [-0.1581, 0.1477, -0.0957, ..., 0.0245, -0.0720, -0.0819], + [ 0.1551, -0.1934, -0.1844, ..., -0.0711, 0.1847, -0.0337], + ..., + [-0.1913, -0.1120, 0.1356, ..., 0.0306, -0.2387, -0.0030], + [ 0.1145, -0.0184, -0.3383, ..., -0.0693, 0.1881, -0.0380], + [-0.2425, -0.1853, 0.0367, ..., -0.0168, -0.2635, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0058, 0.0132, -0.0198, 0.0190, 0.0280, -0.0174, -0.0167, -0.0111, + -0.0356, -0.0333], device='cuda:0'), grad: tensor([-1.8626e-09, 1.8626e-09, 2.7940e-09, 5.5879e-09, 2.2352e-08, + -1.8626e-09, 3.7253e-09, 7.4506e-09, -1.0245e-08, -1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 216.69, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4182 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.13 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.3324, -0.1922, -0.0971, ..., -0.0041, -0.0696, -0.0169], + [-0.1583, 0.1478, -0.0957, ..., 0.0245, -0.0721, -0.0819], + [ 0.1552, -0.1935, -0.1844, ..., -0.0711, 0.1849, -0.0337], + ..., + [-0.1913, -0.1121, 0.1355, ..., 0.0306, -0.2389, -0.0030], + [ 0.1146, -0.0184, -0.3383, ..., -0.0693, 0.1883, -0.0380], + [-0.2426, -0.1855, 0.0367, ..., -0.0168, -0.2635, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, -4.7497e-08, 1.8626e-09, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 0.0000e+00, 1.0245e-08, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + ..., + [ 2.7940e-09, 3.9116e-08, -1.4901e-08, ..., 0.0000e+00, + 6.5193e-09, 0.0000e+00], + [-9.3132e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.0245e-08, 0.0000e+00], + [ 3.7253e-09, 1.6764e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0058, 0.0131, -0.0196, 0.0188, 0.0280, -0.0174, -0.0166, -0.0111, + -0.0355, -0.0333], device='cuda:0'), grad: tensor([ 4.6566e-09, -9.2387e-07, 1.8626e-07, 1.2293e-07, 1.5832e-08, + 6.1467e-08, -7.3574e-08, 5.5879e-07, -2.0489e-08, 8.0094e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 216.93, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4158 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.13 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.3325, -0.1922, -0.0971, ..., -0.0041, -0.0696, -0.0169], + [-0.1583, 0.1478, -0.0958, ..., 0.0245, -0.0723, -0.0819], + [ 0.1554, -0.1937, -0.1844, ..., -0.0711, 0.1852, -0.0337], + ..., + [-0.1914, -0.1121, 0.1357, ..., 0.0306, -0.2390, -0.0030], + [ 0.1147, -0.0184, -0.3383, ..., -0.0693, 0.1885, -0.0380], + [-0.2427, -0.1856, 0.0367, ..., -0.0168, -0.2636, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 1.0245e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0059, 0.0130, -0.0195, 0.0188, 0.0280, -0.0174, -0.0166, -0.0110, + -0.0355, -0.0333], device='cuda:0'), grad: tensor([-3.7253e-09, 7.4506e-09, 3.7253e-09, 8.5682e-08, 2.5146e-08, + -9.1270e-08, 1.4901e-08, -2.2352e-08, 6.5193e-09, -1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 217.03, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4149 re_mapping 0.0020 re_causal 0.0080 /// teacc 99.14 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.3326, -0.1923, -0.0971, ..., -0.0041, -0.0697, -0.0169], + [-0.1586, 0.1479, -0.0957, ..., 0.0245, -0.0724, -0.0819], + [ 0.1556, -0.1938, -0.1844, ..., -0.0711, 0.1854, -0.0337], + ..., + [-0.1913, -0.1121, 0.1357, ..., 0.0306, -0.2391, -0.0030], + [ 0.1147, -0.0186, -0.3383, ..., -0.0693, 0.1886, -0.0380], + [-0.2428, -0.1857, 0.0367, ..., -0.0168, -0.2636, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [ 0.0000e+00, -1.1269e-07, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [ 9.3132e-10, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + -6.5193e-09, 0.0000e+00], + ..., + [ 0.0000e+00, 9.0338e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [-8.3819e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + -8.3819e-09, 0.0000e+00], + [ 9.3132e-10, -1.3970e-08, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0059, 0.0128, -0.0193, 0.0187, 0.0280, -0.0173, -0.0165, -0.0109, + -0.0355, -0.0334], device='cuda:0'), grad: tensor([ 1.2107e-08, -2.0303e-07, -3.7253e-09, 1.3970e-08, 9.6858e-08, + 2.7940e-09, 6.5193e-09, 2.1979e-07, -9.3132e-10, -1.4435e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 216.75, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4593 re_mapping 0.0020 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.3327, -0.1923, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1586, 0.1480, -0.0958, ..., 0.0245, -0.0725, -0.0819], + [ 0.1557, -0.1939, -0.1844, ..., -0.0711, 0.1856, -0.0337], + ..., + [-0.1913, -0.1122, 0.1357, ..., 0.0306, -0.2392, -0.0030], + [ 0.1147, -0.0188, -0.3383, ..., -0.0693, 0.1887, -0.0380], + [-0.2429, -0.1858, 0.0367, ..., -0.0168, -0.2637, -0.0731]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 4.6566e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 2.7940e-09, 0.0000e+00], + [-6.4261e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -5.6811e-08, 0.0000e+00], + ..., + [ 6.5193e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 5.4948e-08, 0.0000e+00], + [ 1.5832e-08, 1.2107e-08, 0.0000e+00, ..., 0.0000e+00, + -1.3970e-08, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0059, 0.0129, -0.0192, 0.0186, 0.0280, -0.0175, -0.0163, -0.0110, + -0.0355, -0.0333], device='cuda:0'), grad: tensor([-6.5193e-09, 2.5146e-08, -1.3877e-07, -2.7940e-08, 2.1420e-08, + 8.8476e-08, -1.3039e-07, 1.3597e-07, 1.2107e-08, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 217.12, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4512 re_mapping 0.0021 re_causal 0.0082 /// teacc 99.16 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.3328, -0.1923, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1587, 0.1481, -0.0958, ..., 0.0245, -0.0728, -0.0819], + [ 0.1558, -0.1941, -0.1845, ..., -0.0711, 0.1857, -0.0337], + ..., + [-0.1914, -0.1123, 0.1360, ..., 0.0306, -0.2391, -0.0030], + [ 0.1147, -0.0188, -0.3383, ..., -0.0693, 0.1889, -0.0380], + [-0.2430, -0.1859, 0.0366, ..., -0.0168, -0.2637, -0.0731]], + device='cuda:0'), grad: tensor([[9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + ..., + [0.0000e+00, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00], + [0.0000e+00, 3.7253e-08, 1.0245e-08, ..., 0.0000e+00, 0.0000e+00, + 0.0000e+00]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0059, 0.0128, -0.0194, 0.0185, 0.0280, -0.0174, -0.0163, -0.0108, + -0.0355, -0.0334], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.1176e-08, 4.6566e-09, 1.3970e-08, -1.8626e-07, + -9.3132e-09, -2.7940e-09, 2.4214e-08, 8.3819e-09, 1.4249e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 216.81, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4203 re_mapping 0.0020 re_causal 0.0077 /// teacc 99.14 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.3329, -0.1926, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1587, 0.1482, -0.0959, ..., 0.0245, -0.0729, -0.0819], + [ 0.1559, -0.1942, -0.1845, ..., -0.0711, 0.1858, -0.0337], + ..., + [-0.1914, -0.1124, 0.1361, ..., 0.0306, -0.2392, -0.0030], + [ 0.1147, -0.0189, -0.3383, ..., -0.0693, 0.1890, -0.0380], + [-0.2431, -0.1861, 0.0366, ..., -0.0168, -0.2637, -0.0731]], + device='cuda:0'), grad: tensor([[-1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, + -8.7544e-08, 0.0000e+00], + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 1.3970e-08, 0.0000e+00], + ..., + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 8.3819e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-08, 0.0000e+00]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0060, 0.0128, -0.0195, 0.0185, 0.0280, -0.0174, -0.0160, -0.0108, + -0.0357, -0.0334], device='cuda:0'), grad: tensor([-5.4017e-07, 1.7695e-08, 6.9849e-08, 8.8476e-08, -9.3132e-10, + 1.1176e-08, 8.4750e-08, -1.1176e-08, 1.5832e-08, 2.6543e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 216.84, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4178 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.12 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.3331, -0.1927, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1588, 0.1484, -0.0960, ..., 0.0245, -0.0730, -0.0819], + [ 0.1560, -0.1945, -0.1846, ..., -0.0711, 0.1859, -0.0337], + ..., + [-0.1915, -0.1126, 0.1364, ..., 0.0306, -0.2391, -0.0030], + [ 0.1147, -0.0191, -0.3383, ..., -0.0693, 0.1891, -0.0380], + [-0.2433, -0.1863, 0.0365, ..., -0.0168, -0.2638, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.6566e-09, 6.5193e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, 5.5879e-09, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 5.8673e-08, -1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0060, 0.0128, -0.0197, 0.0182, 0.0281, -0.0173, -0.0160, -0.0107, + -0.0358, -0.0334], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.0245e-08, 2.6077e-08, 2.0489e-07, -3.5670e-07, + -2.0023e-07, -2.9802e-08, 3.2596e-08, 1.8626e-09, 3.0827e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 216.96, cls_loss 0.0005 cls_loss_mapping 0.0004 cls_loss_causal 0.4223 re_mapping 0.0020 re_causal 0.0079 /// teacc 99.11 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.3332, -0.1928, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1588, 0.1486, -0.0960, ..., 0.0245, -0.0731, -0.0819], + [ 0.1561, -0.1948, -0.1846, ..., -0.0711, 0.1860, -0.0337], + ..., + [-0.1915, -0.1126, 0.1365, ..., 0.0306, -0.2392, -0.0030], + [ 0.1148, -0.0191, -0.3383, ..., -0.0693, 0.1892, -0.0380], + [-0.2434, -0.1865, 0.0364, ..., -0.0168, -0.2638, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 2.7940e-09, 3.7253e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 9.3132e-10, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + ..., + [ 0.0000e+00, -2.7940e-09, -8.3819e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + -9.3132e-10, 0.0000e+00], + [ 9.3132e-10, 1.5832e-08, 1.8626e-09, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0061, 0.0128, -0.0197, 0.0181, 0.0281, -0.0171, -0.0159, -0.0107, + -0.0359, -0.0335], device='cuda:0'), grad: tensor([ 1.8626e-09, 4.4703e-08, 1.9558e-08, 3.4459e-08, -4.0047e-08, + -1.7975e-07, 1.3225e-07, -6.8918e-08, 4.6566e-09, 5.4948e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 217.05, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4198 re_mapping 0.0020 re_causal 0.0080 /// teacc 99.11 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.3333, -0.1929, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1589, 0.1487, -0.0961, ..., 0.0245, -0.0731, -0.0819], + [ 0.1562, -0.1950, -0.1846, ..., -0.0711, 0.1862, -0.0337], + ..., + [-0.1915, -0.1127, 0.1366, ..., 0.0306, -0.2394, -0.0030], + [ 0.1148, -0.0193, -0.3383, ..., -0.0693, 0.1894, -0.0380], + [-0.2435, -0.1866, 0.0365, ..., -0.0168, -0.2639, -0.0731]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + ..., + [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.8626e-09, 2.7940e-09, 0.0000e+00, ..., 0.0000e+00, + -1.8626e-09, 0.0000e+00], + [ 0.0000e+00, 1.4901e-08, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0061, 0.0128, -0.0197, 0.0179, 0.0281, -0.0169, -0.0158, -0.0106, + -0.0359, -0.0335], device='cuda:0'), grad: tensor([-6.5193e-09, 4.6566e-09, 2.7940e-09, 4.0047e-08, -3.8184e-08, + -4.1910e-08, -1.0245e-08, -8.6613e-08, 3.7253e-09, 1.2666e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 217.11, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4015 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.12 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.3335, -0.1930, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1591, 0.1487, -0.0961, ..., 0.0245, -0.0733, -0.0819], + [ 0.1563, -0.1952, -0.1847, ..., -0.0711, 0.1862, -0.0337], + ..., + [-0.1914, -0.1127, 0.1367, ..., 0.0306, -0.2393, -0.0030], + [ 0.1148, -0.0193, -0.3383, ..., -0.0693, 0.1895, -0.0380], + [-0.2436, -0.1868, 0.0365, ..., -0.0168, -0.2639, -0.0731]], + device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 7.4506e-09, 0.0000e+00], + [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-3.0734e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.8673e-08, 0.0000e+00], + ..., + [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 1.8626e-09, 0.0000e+00], + [-1.0245e-08, -3.7253e-09, 9.3132e-10, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + [ 3.2596e-08, 4.6566e-09, -2.7940e-09, ..., 0.0000e+00, + 4.7497e-08, 0.0000e+00]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0062, 0.0127, -0.0198, 0.0177, 0.0281, -0.0167, -0.0157, -0.0104, + -0.0360, -0.0335], device='cuda:0'), grad: tensor([ 2.4214e-08, 6.5193e-09, -1.5181e-07, 3.8184e-08, -6.5193e-09, + -3.6322e-08, 5.5879e-09, 8.3819e-09, -2.1420e-08, 1.2759e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 217.03, cls_loss 0.0005 cls_loss_mapping 0.0003 cls_loss_causal 0.4455 re_mapping 0.0021 re_causal 0.0084 /// teacc 99.13 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.3336, -0.1931, -0.0972, ..., -0.0041, -0.0697, -0.0169], + [-0.1592, 0.1487, -0.0962, ..., 0.0245, -0.0734, -0.0819], + [ 0.1565, -0.1953, -0.1846, ..., -0.0711, 0.1866, -0.0337], + ..., + [-0.1915, -0.1127, 0.1368, ..., 0.0306, -0.2394, -0.0030], + [ 0.1148, -0.0194, -0.3383, ..., -0.0693, 0.1897, -0.0380], + [-0.2437, -0.1869, 0.0364, ..., -0.0168, -0.2640, -0.0731]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.1420e-08, 3.7253e-09, 0.0000e+00, ..., 0.0000e+00, + 4.6566e-09, 0.0000e+00], + [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + -5.5879e-09, 0.0000e+00], + ..., + [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, + 9.3132e-10, 0.0000e+00], + [ 9.8720e-08, 1.7695e-08, 0.0000e+00, ..., 0.0000e+00, + 2.5146e-08, 0.0000e+00], + [ 0.0000e+00, 4.6566e-09, -9.3132e-10, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0061, 0.0126, -0.0196, 0.0174, 0.0281, -0.0165, -0.0157, -0.0104, + -0.0360, -0.0335], device='cuda:0'), grad: tensor([-8.3819e-09, 5.4017e-08, -1.3970e-08, 1.5832e-07, -1.3039e-08, + -5.1130e-07, 5.9605e-08, 7.4506e-09, 2.5239e-07, 2.2352e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 217.00, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4034 re_mapping 0.0021 re_causal 0.0080 /// teacc 99.11 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip4', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_skip4/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.809998 98.959999 ... 83.906326 70.694491 +ShearY 98.699997 98.729996 ... 83.906326 67.395161 +AutoContrast 98.930000 99.119995 ... 83.906326 62.679251 +Invert 98.699997 98.089996 ... 83.906326 62.888086 +Equalize 98.209999 97.939995 ... 83.906326 68.967734 +Solarize 98.070000 97.739998 ... 83.906326 66.233051 +SolarizeAdd 98.419998 97.919998 ... 83.906326 71.701561 +Posterize 99.000000 98.979996 ... 83.906326 73.696783 +Contrast 99.070000 99.129997 ... 83.906326 67.761722 +Color 99.010002 99.199997 ... 83.906326 60.734499 +Brightness 98.979996 99.139999 ... 83.906326 66.071666 +Sharpness 98.919998 99.070000 ... 83.906326 69.373529 +NoiseSalt 98.900002 99.049995 ... 83.906326 56.324764 +NoiseGaussian 98.979996 99.190002 ... 83.906326 58.073993 +w/o do (original x) 99.200000 0.000000 ... 0.000000 75.410608 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.14 66.13783 78.802355 75.034021 84.055805 76.007503 diff --git a/Meta-causal/code-withStyleAttack/71579.error b/Meta-causal/code-withStyleAttack/71579.error new file mode 100644 index 0000000000000000000000000000000000000000..ae0356cda6d135bef0ff92139078c691affce8be --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71579.error @@ -0,0 +1,4 @@ +Solving dependencies +slurmstepd: error: *** JOB 71579 ON gcp-eu-1 CANCELLED AT 2024-07-25T07:37:45 *** +srun: Job step aborted: Waiting up to 32 seconds for job step to finish. +slurmstepd: error: *** STEP 71579.0 ON gcp-eu-1 CANCELLED AT 2024-07-25T07:37:46 *** diff --git a/Meta-causal/code-withStyleAttack/71579.log b/Meta-causal/code-withStyleAttack/71579.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/71580.error b/Meta-causal/code-withStyleAttack/71580.error new file mode 100644 index 0000000000000000000000000000000000000000..07addde1c3ebd8d64d93028d12ef4d0776b5b5f7 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71580.error @@ -0,0 +1,321 @@ +Solving dependencies +Installing conda packages +Empty environment created at prefix: /scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe +error libmamba Could not lock non-existing path '/scratch/yuqian_fu/micromamba/pkgs' +Transaction + + Prefix: /scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe + + + + No specs added or removed. + + Package Version Build Channel Size +───────────────────────────────────────────────────────────────────────────────────────────────────────── + Install: +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + _libgcc_mutex 0.1 conda_forge conda-forge + + _openmp_mutex 4.5 2_kmp_llvm conda-forge + + blas 2.116 mkl conda-forge + + blas-devel 3.9.0 16_linux64_mkl conda-forge + + brotli-python 1.1.0 py311hb755f60_1 conda-forge + + bzip2 1.0.8 h4bc722e_7 conda-forge + + ca-certificates 2024.7.4 hbcca054_0 conda-forge + + certifi 2024.7.4 pyhd8ed1ab_0 conda-forge + + cffi 1.16.0 py311hb3a22ac_0 conda-forge + + charset-normalizer 3.3.2 pyhd8ed1ab_0 conda-forge + + click 8.1.7 unix_pyh707e725_0 conda-forge + + cuda-cudart 12.1.105 0 nvidia + + cuda-cupti 12.1.105 0 /work/conda/cache/nvidia + + cuda-libraries 12.1.0 0 nvidia + + cuda-nvrtc 12.1.105 0 /work/conda/cache/nvidia + + cuda-nvtx 12.1.105 0 nvidia + + cuda-opencl 12.5.39 0 nvidia + + cuda-runtime 12.1.0 0 nvidia + + cuda-version 12.5 3 nvidia + + ffmpeg 4.3 hf484d3e_0 /work/conda/cache/pytorch + + filelock 3.15.4 pyhd8ed1ab_0 conda-forge + + freetype 2.12.1 h267a509_2 conda-forge + + gmp 6.3.0 hac33072_2 conda-forge + + gmpy2 2.1.5 py311hc4f1f91_1 conda-forge + + gnutls 3.6.13 h85f3911_1 /work/conda/cache/conda-forge + + h2 4.1.0 pyhd8ed1ab_0 conda-forge + + hpack 4.0.0 pyh9f0ad1d_0 conda-forge + + hyperframe 6.0.1 pyhd8ed1ab_0 conda-forge + + icu 73.2 h59595ed_0 /work/conda/cache/conda-forge + + idna 3.7 pyhd8ed1ab_0 conda-forge + + jinja2 3.1.4 pyhd8ed1ab_0 conda-forge + + jpeg 9e h166bdaf_2 conda-forge + + lame 3.100 h166bdaf_1003 conda-forge + + lcms2 2.15 hfd0df8a_0 conda-forge + + ld_impl_linux-64 2.40 hf3520f5_7 conda-forge + + lerc 4.0.0 h27087fc_0 conda-forge + + libblas 3.9.0 16_linux64_mkl conda-forge + + libcblas 3.9.0 16_linux64_mkl conda-forge + + libcublas 12.1.0.26 0 /work/conda/cache/nvidia + + libcufft 11.0.2.4 0 /work/conda/cache/nvidia + + libcufile 1.10.1.7 0 nvidia + + libcurand 10.3.6.82 0 nvidia + + libcusolver 11.4.4.55 0 /work/conda/cache/nvidia + + libcusparse 12.0.2.55 0 /work/conda/cache/nvidia + + libdeflate 1.17 h0b41bf4_0 conda-forge + + libexpat 2.6.2 h59595ed_0 conda-forge + + libffi 3.4.2 h7f98852_5 conda-forge + + libgcc-ng 14.1.0 h77fa898_0 conda-forge + + libgfortran-ng 14.1.0 h69a702a_0 conda-forge + + libgfortran5 14.1.0 hc5f4f2c_0 /work/conda/cache/conda-forge + + libhwloc 2.11.1 default_hecaa2ac_1000 conda-forge + + libiconv 1.17 hd590300_2 conda-forge + + libjpeg-turbo 2.0.0 h9bf148f_0 pytorch + + liblapack 3.9.0 16_linux64_mkl conda-forge + + liblapacke 3.9.0 16_linux64_mkl conda-forge + + libnpp 12.0.2.50 0 /work/conda/cache/nvidia + + libnsl 2.0.1 hd590300_0 conda-forge + + libnvjitlink 12.1.105 0 /work/conda/cache/nvidia + + libnvjpeg 12.1.1.14 0 /work/conda/cache/nvidia + + libpng 1.6.43 h2797004_0 conda-forge + + libsqlite 3.46.0 hde9e2c9_0 conda-forge + + libstdcxx-ng 14.1.0 hc0a3c3a_0 /work/conda/cache/conda-forge + + libtiff 4.5.0 h6adf6a1_2 conda-forge + + libuuid 2.38.1 h0b41bf4_0 conda-forge + + libwebp-base 1.4.0 hd590300_0 conda-forge + + libxcb 1.13 h7f98852_1004 conda-forge + + libxcrypt 4.4.36 hd590300_1 conda-forge + + libxml2 2.12.7 hc051c1a_1 conda-forge + + libzlib 1.2.13 h4ab18f5_6 conda-forge + + llvm-openmp 15.0.7 h0cdce71_0 /work/conda/cache/conda-forge + + markupsafe 2.1.5 py311h459d7ec_0 conda-forge + + mkl 2022.1.0 h84fe81f_915 /work/conda/cache/conda-forge + + mkl-devel 2022.1.0 ha770c72_916 conda-forge + + mkl-include 2022.1.0 h84fe81f_915 conda-forge + + mpc 1.3.1 hfe3b2da_0 conda-forge + + mpfr 4.2.1 h9458935_1 conda-forge + + mpmath 1.3.0 pyhd8ed1ab_0 conda-forge + + ncurses 6.5 h59595ed_0 conda-forge + + nettle 3.6 he412f7d_0 /work/conda/cache/conda-forge + + networkx 3.3 pyhd8ed1ab_1 /work/conda/cache/conda-forge + + numpy 2.0.0 py311h1461c94_0 conda-forge + + openh264 2.1.1 h780b84a_0 /work/conda/cache/conda-forge + + openjpeg 2.5.0 hfec8fc6_2 conda-forge + + openssl 3.3.1 h4bc722e_2 conda-forge + + pandas 2.2.2 py311h14de704_1 conda-forge + + pillow 9.4.0 py311h50def17_1 /work/conda/cache/conda-forge + + pip 24.0 pyhd8ed1ab_0 /work/conda/cache/conda-forge + + pthread-stubs 0.4 h36c2ea0_1001 conda-forge + + pycparser 2.22 pyhd8ed1ab_0 conda-forge + + pysocks 1.7.1 pyha2e5f31_6 conda-forge + + python 3.11.9 hb806964_0_cpython /work/conda/cache/conda-forge + + python-dateutil 2.9.0 pyhd8ed1ab_0 conda-forge + + python-tzdata 2024.1 pyhd8ed1ab_0 conda-forge + + python_abi 3.11 4_cp311 conda-forge + + pytorch 2.4.0 py3.11_cuda12.1_cudnn9.1.0_0 pytorch + + pytorch-cuda 12.1 ha16c6d3_5 pytorch + + pytorch-mutex 1.0 cuda pytorch + + pytz 2024.1 pyhd8ed1ab_0 conda-forge + + pyyaml 6.0.1 py311h459d7ec_1 conda-forge + + readline 8.2 h8228510_1 conda-forge + + requests 2.32.3 pyhd8ed1ab_0 conda-forge + + setuptools 71.0.4 pyhd8ed1ab_0 conda-forge + + six 1.16.0 pyh6c4a22f_0 conda-forge + + sympy 1.13.0 pypyh2585a3b_103 conda-forge + + tbb 2021.12.0 h434a139_3 conda-forge + + tk 8.6.13 noxft_h4845f30_101 /work/conda/cache/conda-forge + + torchtriton 3.0.0 py311 pytorch + + torchvision 0.19.0 py311_cu121 pytorch + + typing_extensions 4.12.2 pyha770c72_0 conda-forge + + tzdata 2024a h0c530f3_0 conda-forge + + urllib3 2.2.2 pyhd8ed1ab_1 conda-forge + + wheel 0.43.0 pyhd8ed1ab_1 conda-forge + + xorg-libxau 1.0.11 hd590300_0 conda-forge + + xorg-libxdmcp 1.1.3 h516909a_0 conda-forge + + xz 5.2.6 h166bdaf_0 conda-forge + + yaml 0.2.5 h7f98852_2 conda-forge + + zlib 1.2.13 h4ab18f5_6 conda-forge + + zstandard 0.23.0 py311h5cd10c7_0 conda-forge + + zstd 1.5.6 ha6fb4c9_0 conda-forge + + Summary: + + Install: 119 packages + + Total download: 0 B + +───────────────────────────────────────────────────────────────────────────────────────────────────────── + + + +Transaction starting +Linking libcublas-12.1.0.26-0 +Linking libcufft-11.0.2.4-0 +Linking libcusolver-11.4.4.55-0 +Linking libcusparse-12.0.2.55-0 +Linking libnpp-12.0.2.50-0 +Linking libnvjitlink-12.1.105-0 +Linking cuda-cudart-12.1.105-0 +Linking cuda-nvrtc-12.1.105-0 +Linking libnvjpeg-12.1.1.14-0 +Linking cuda-cupti-12.1.105-0 +Linking cuda-nvtx-12.1.105-0 +Linking pytorch-mutex-1.0-cuda +Linking _libgcc_mutex-0.1-conda_forge +Linking mkl-include-2022.1.0-h84fe81f_915 +Linking python_abi-3.11-4_cp311 +Linking ld_impl_linux-64-2.40-hf3520f5_7 +Linking ca-certificates-2024.7.4-hbcca054_0 +Linking libgcc-ng-14.1.0-h77fa898_0 +Linking libzlib-1.2.13-h4ab18f5_6 +Linking llvm-openmp-15.0.7-h0cdce71_0 +Linking _openmp_mutex-4.5-2_kmp_llvm +Linking xorg-libxdmcp-1.1.3-h516909a_0 +Linking pthread-stubs-0.4-h36c2ea0_1001 +Linking xorg-libxau-1.0.11-hd590300_0 +Linking libwebp-base-1.4.0-hd590300_0 +Linking libdeflate-1.17-h0b41bf4_0 +Linking jpeg-9e-h166bdaf_2 +Linking libffi-3.4.2-h7f98852_5 +Linking tk-8.6.13-noxft_h4845f30_101 +Linking openssl-3.3.1-h4bc722e_2 +Linking libxcrypt-4.4.36-hd590300_1 +Linking libsqlite-3.46.0-hde9e2c9_0 +Linking yaml-0.2.5-h7f98852_2 +Linking ncurses-6.5-h59595ed_0 +Linking libgfortran5-14.1.0-hc5f4f2c_0 +Linking lame-3.100-h166bdaf_1003 +Linking nettle-3.6-he412f7d_0 +Linking zlib-1.2.13-h4ab18f5_6 +Linking libstdcxx-ng-14.1.0-hc0a3c3a_0 +Linking libiconv-1.17-hd590300_2 +Linking bzip2-1.0.8-h4bc722e_7 +Linking libpng-1.6.43-h2797004_0 +Linking xz-5.2.6-h166bdaf_0 +Linking libuuid-2.38.1-h0b41bf4_0 +Linking libnsl-2.0.1-hd590300_0 +Linking libexpat-2.6.2-h59595ed_0 +Linking libxcb-1.13-h7f98852_1004 +Linking readline-8.2-h8228510_1 +Linking libgfortran-ng-14.1.0-h69a702a_0 +Linking icu-73.2-h59595ed_0 +Linking zstd-1.5.6-ha6fb4c9_0 +Linking lerc-4.0.0-h27087fc_0 +Linking openh264-2.1.1-h780b84a_0 +Linking gnutls-3.6.13-h85f3911_1 +Linking gmp-6.3.0-hac33072_2 +Linking freetype-2.12.1-h267a509_2 +Linking libxml2-2.12.7-hc051c1a_1 +Linking libtiff-4.5.0-h6adf6a1_2 +Linking mpfr-4.2.1-h9458935_1 +Linking libhwloc-2.11.1-default_hecaa2ac_1000 +Linking openjpeg-2.5.0-hfec8fc6_2 +Linking lcms2-2.15-hfd0df8a_0 +Linking mpc-1.3.1-hfe3b2da_0 +Linking tbb-2021.12.0-h434a139_3 +Linking mkl-2022.1.0-h84fe81f_915 +Linking mkl-devel-2022.1.0-ha770c72_916 +Linking libblas-3.9.0-16_linux64_mkl +Linking liblapack-3.9.0-16_linux64_mkl +Linking libcblas-3.9.0-16_linux64_mkl +Linking liblapacke-3.9.0-16_linux64_mkl +Linking blas-devel-3.9.0-16_linux64_mkl +Linking blas-2.116-mkl +Linking cuda-version-12.5-3 +Linking tzdata-2024a-h0c530f3_0 +Linking libjpeg-turbo-2.0.0-h9bf148f_0 +warning libmamba [libjpeg-turbo-2.0.0-h9bf148f_0] The following files were already present in the environment: + - bin/cjpeg + - bin/djpeg + - bin/jpegtran + - bin/rdjpgcom + - bin/wrjpgcom + - include/jconfig.h + - include/jerror.h + - include/jmorecfg.h + - include/jpeglib.h + - lib/libjpeg.a + - lib/libjpeg.so + - lib/pkgconfig/libjpeg.pc + - share/man/man1/cjpeg.1 + - share/man/man1/djpeg.1 + - share/man/man1/jpegtran.1 + - share/man/man1/rdjpgcom.1 + - share/man/man1/wrjpgcom.1 +Linking ffmpeg-4.3-hf484d3e_0 +Linking libcurand-10.3.6.82-0 +Linking libcufile-1.10.1.7-0 +Linking cuda-opencl-12.5.39-0 +Linking cuda-libraries-12.1.0-0 +Linking cuda-runtime-12.1.0-0 +Linking python-3.11.9-hb806964_0_cpython +Linking pytorch-cuda-12.1-ha16c6d3_5 +Linking wheel-0.43.0-pyhd8ed1ab_1 +Linking setuptools-71.0.4-pyhd8ed1ab_0 +Linking pip-24.0-pyhd8ed1ab_0 +Linking pycparser-2.22-pyhd8ed1ab_0 +Linking six-1.16.0-pyh6c4a22f_0 +Linking hyperframe-6.0.1-pyhd8ed1ab_0 +Linking pytz-2024.1-pyhd8ed1ab_0 +Linking python-tzdata-2024.1-pyhd8ed1ab_0 +Linking charset-normalizer-3.3.2-pyhd8ed1ab_0 +Linking hpack-4.0.0-pyh9f0ad1d_0 +Linking pysocks-1.7.1-pyha2e5f31_6 +Linking idna-3.7-pyhd8ed1ab_0 +Linking certifi-2024.7.4-pyhd8ed1ab_0 +Linking mpmath-1.3.0-pyhd8ed1ab_0 +Linking typing_extensions-4.12.2-pyha770c72_0 +Linking networkx-3.3-pyhd8ed1ab_1 +Linking filelock-3.15.4-pyhd8ed1ab_0 +Linking click-8.1.7-unix_pyh707e725_0 +Linking python-dateutil-2.9.0-pyhd8ed1ab_0 +Linking h2-4.1.0-pyhd8ed1ab_0 +Linking brotli-python-1.1.0-py311hb755f60_1 +Linking markupsafe-2.1.5-py311h459d7ec_0 +Linking gmpy2-2.1.5-py311hc4f1f91_1 +Linking pyyaml-6.0.1-py311h459d7ec_1 +Linking pillow-9.4.0-py311h50def17_1 +Linking numpy-2.0.0-py311h1461c94_0 +Linking cffi-1.16.0-py311hb3a22ac_0 +Linking pandas-2.2.2-py311h14de704_1 +Linking zstandard-0.23.0-py311h5cd10c7_0 +Linking jinja2-3.1.4-pyhd8ed1ab_0 +Linking sympy-1.13.0-pypyh2585a3b_103 +Linking urllib3-2.2.2-pyhd8ed1ab_1 +Linking requests-2.32.3-pyhd8ed1ab_0 +Linking pytorch-2.4.0-py3.11_cuda12.1_cudnn9.1.0_0 +Linking torchtriton-3.0.0-py311 +Linking torchvision-0.19.0-py311_cu121 + +Transaction finished + +To activate this environment, use: + + mamba activate auto-v5ewbna3m2oe + +Or to execute a single command in this environment, use: + + mamba run -n auto-v5ewbna3m2oe mycommand + +Installing pip packages +WARNING: The candidate selected for download or install is a yanked version: 'opencv-python' candidate (version 4.5.5.62 at https://files.pythonhosted.org/packages/9d/98/36bfcbff30da27dd6922ed73ca7802c37d87f77daf4c569da3dcb87b4296/opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (from https://pypi.org/simple/opencv-python/) (requires-python:>=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:45: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:62: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:72: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:48: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:65: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:75: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) diff --git a/Meta-causal/code-withStyleAttack/71580.log b/Meta-causal/code-withStyleAttack/71580.log new file mode 100644 index 0000000000000000000000000000000000000000..713fb2fbe48114ed0845be2d5b03b7f099f3c514 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71580.log @@ -0,0 +1,13446 @@ +Collecting h5py>=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 9.4 MB/s eta 0:00:00 + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.0) +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 20.4 MB/s eta 0:00:00 +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 7.7 MB/s eta 0:00:00 +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.4-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (71.0.4) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.4.0) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.19.0) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.24.2-py3-none-any.whl.metadata (13 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 23.2 MB/s eta 0:00:00 +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.13.0) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.7.4) +Requirement already satisfied: mpmath<1.4,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 93.3 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 176.7 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 109.8 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 174.1 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 38.9 MB/s eta 0:00:00 +Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 155.9 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 46.0 MB/s eta 0:00:00 +Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.7 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.7/5.7 MB 158.3 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 36.7 MB/s eta 0:00:00 +Downloading protobuf-4.25.4-cp37-abi3-manylinux2014_x86_64.whl (294 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 47.4 MB/s eta 0:00:00 +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 101.1 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 75.6 MB/s eta 0:00:00 +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.24.2-py3-none-any.whl (417 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 417.2/417.2 kB 112.9 MB/s eta 0:00:00 +Downloading packaging-24.1-py3-none-any.whl (53 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 22.6 MB/s eta 0:00:00 +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 172.7 MB/s eta 0:00:00 +Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 177.6/177.6 kB 33.7 MB/s eta 0:00:00 +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 27.3 MB/s eta 0:00:00 +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94506 sha256=f632daf5aa75bd3ea372c8262b8885507277a3546feb9656e476217cb0ff6a79 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.65.1 h5py-3.11.0 huggingface_hub-0.24.2 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.4 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0095, 0.0272, -0.0073, ..., -0.0262, 0.0269, -0.0030], + [ 0.0103, -0.0206, 0.0275, ..., -0.0174, 0.0309, 0.0206], + [ 0.0035, -0.0229, -0.0102, ..., 0.0293, -0.0305, -0.0187], + ..., + [ 0.0102, -0.0112, 0.0231, ..., 0.0112, 0.0303, 0.0193], + [-0.0256, 0.0085, -0.0285, ..., 0.0235, -0.0175, 0.0119], + [-0.0263, -0.0056, -0.0095, ..., -0.0270, 0.0204, 0.0019]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0252, 0.0012, 0.0123, 0.0068, -0.0063, -0.0001, 0.0048, 0.0301, + -0.0290, 0.0285], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 218.83, cls_loss 2.2628 cls_loss_mapping 2.2137 cls_loss_causal 2.2802 re_mapping 0.0098 re_causal 0.0094 /// teacc 48.94 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0077, 0.0274, -0.0079, ..., -0.0266, 0.0241, -0.0030], + [ 0.0101, -0.0251, 0.0287, ..., -0.0187, 0.0286, 0.0206], + [ 0.0051, -0.0237, -0.0112, ..., 0.0295, -0.0319, -0.0187], + ..., + [ 0.0074, -0.0107, 0.0229, ..., 0.0105, 0.0303, 0.0193], + [-0.0250, 0.0089, -0.0277, ..., 0.0215, -0.0188, 0.0119], + [-0.0292, -0.0066, -0.0114, ..., -0.0269, 0.0238, 0.0019]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0023, 0.0001, ..., -0.0020, 0.0053, 0.0000], + [-0.0089, -0.0020, 0.0002, ..., -0.0020, 0.0044, 0.0000], + [-0.0005, 0.0027, -0.0007, ..., 0.0027, 0.0056, 0.0000], + ..., + [-0.0044, -0.0017, 0.0002, ..., -0.0014, 0.0012, 0.0000], + [-0.0068, 0.0015, 0.0001, ..., 0.0029, -0.0018, 0.0000], + [ 0.0110, 0.0048, 0.0002, ..., 0.0024, 0.0081, 0.0000]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0249, 0.0025, 0.0115, 0.0072, -0.0066, -0.0016, 0.0049, 0.0307, + -0.0285, 0.0277], device='cuda:0'), grad: tensor([ 0.0123, -0.0538, 0.0145, 0.0702, 0.0202, -0.0529, -0.0338, -0.0144, + -0.0310, 0.0687], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 217.42, cls_loss 1.9329 cls_loss_mapping 1.1602 cls_loss_causal 1.9270 re_mapping 0.1142 re_causal 0.1214 /// teacc 86.57 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0054, 0.0313, -0.0136, ..., -0.0261, 0.0163, -0.0118], + [ 0.0072, -0.0337, 0.0304, ..., -0.0221, 0.0197, 0.0104], + [ 0.0091, -0.0251, -0.0084, ..., 0.0350, -0.0387, -0.0234], + ..., + [ 0.0013, -0.0176, 0.0242, ..., 0.0065, 0.0306, 0.0275], + [-0.0231, 0.0108, -0.0255, ..., 0.0168, -0.0239, 0.0078], + [-0.0339, -0.0088, -0.0128, ..., -0.0291, 0.0304, 0.0035]], + device='cuda:0'), grad: tensor([[ 0.0080, 0.0030, 0.0046, ..., 0.0006, 0.0029, 0.0078], + [-0.0130, -0.0060, -0.0048, ..., -0.0022, 0.0017, -0.0074], + [ 0.0110, 0.0175, 0.0072, ..., 0.0015, 0.0022, 0.0067], + ..., + [ 0.0060, 0.0102, -0.0039, ..., 0.0005, 0.0017, -0.0014], + [-0.0218, -0.0349, 0.0058, ..., 0.0015, 0.0005, 0.0034], + [ 0.0064, -0.0118, -0.0073, ..., 0.0010, -0.0098, -0.0089]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0231, 0.0026, 0.0107, 0.0078, -0.0073, 0.0002, 0.0051, 0.0302, + -0.0279, 0.0279], device='cuda:0'), grad: tensor([ 0.0482, -0.0417, 0.0839, -0.0696, 0.0262, 0.0281, 0.0073, -0.0024, + -0.0199, -0.0601], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 217.37, cls_loss 1.6248 cls_loss_mapping 0.5084 cls_loss_causal 1.5625 re_mapping 0.1169 re_causal 0.1794 /// teacc 92.03 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0020, 0.0332, -0.0224, ..., -0.0256, 0.0110, -0.0152], + [ 0.0037, -0.0416, 0.0265, ..., -0.0251, 0.0146, 0.0057], + [ 0.0115, -0.0274, -0.0043, ..., 0.0414, -0.0410, -0.0230], + ..., + [-0.0030, -0.0249, 0.0331, ..., 0.0038, 0.0302, 0.0321], + [-0.0223, 0.0116, -0.0251, ..., 0.0125, -0.0266, 0.0068], + [-0.0379, -0.0101, -0.0118, ..., -0.0324, 0.0334, 0.0067]], + device='cuda:0'), grad: tensor([[ 8.7967e-03, 1.1612e-02, 6.0692e-03, ..., 3.8803e-05, + 1.3752e-03, 7.1144e-03], + [ 1.0590e-02, 7.0610e-03, 3.4149e-02, ..., -3.9011e-05, + 1.8454e-03, 2.9816e-02], + [-3.4668e-02, -3.2471e-02, -3.4821e-02, ..., 2.1577e-04, + 1.4334e-03, -3.0350e-02], + ..., + [-4.2305e-03, 1.0307e-02, -3.9062e-02, ..., 8.1837e-05, + 1.8616e-03, -3.1952e-02], + [-3.8719e-03, -1.4496e-02, -2.2934e-02, ..., 1.8227e-04, + -1.2329e-02, -2.6077e-02], + [ 1.1101e-02, 1.4145e-02, 1.6739e-02, ..., 6.6280e-05, + -3.2253e-03, 1.4969e-02]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0227, 0.0016, 0.0112, 0.0080, -0.0066, 0.0002, 0.0055, 0.0302, + -0.0275, 0.0271], device='cuda:0'), grad: tensor([ 0.0322, 0.0464, -0.0427, 0.0150, 0.0161, -0.0079, -0.0164, -0.0410, + -0.0745, 0.0728], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 217.63, cls_loss 1.3867 cls_loss_mapping 0.3279 cls_loss_causal 1.3136 re_mapping 0.0921 re_causal 0.1685 /// teacc 94.16 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0015, 0.0348, -0.0253, ..., -0.0250, 0.0076, -0.0155], + [ 0.0014, -0.0455, 0.0248, ..., -0.0267, 0.0118, 0.0044], + [ 0.0135, -0.0299, -0.0013, ..., 0.0452, -0.0441, -0.0236], + ..., + [-0.0063, -0.0289, 0.0381, ..., 0.0031, 0.0299, 0.0344], + [-0.0215, 0.0125, -0.0261, ..., 0.0080, -0.0295, 0.0050], + [-0.0406, -0.0098, -0.0129, ..., -0.0341, 0.0354, 0.0076]], + device='cuda:0'), grad: tensor([[-0.0070, -0.0242, -0.0029, ..., -0.0107, -0.0019, -0.0060], + [ 0.0079, 0.0094, 0.0124, ..., 0.0056, 0.0030, 0.0181], + [ 0.0079, 0.0175, 0.0050, ..., 0.0014, 0.0022, 0.0093], + ..., + [ 0.0057, 0.0146, 0.0153, ..., 0.0040, 0.0034, 0.0289], + [ 0.0347, 0.0348, -0.0003, ..., 0.0080, -0.0068, 0.0055], + [ 0.0081, 0.0176, -0.0179, ..., 0.0004, 0.0042, -0.0221]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0223, 0.0015, 0.0115, 0.0086, -0.0070, 0.0005, 0.0050, 0.0300, + -0.0281, 0.0277], device='cuda:0'), grad: tensor([-0.0794, 0.0531, 0.0204, -0.0726, 0.0237, -0.0667, 0.0069, 0.0593, + 0.0478, 0.0075], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 226.71, cls_loss 1.2666 cls_loss_mapping 0.2560 cls_loss_causal 1.2034 re_mapping 0.0710 re_causal 0.1461 /// teacc 94.29 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0029, 0.0361, -0.0261, ..., -0.0254, 0.0061, -0.0149], + [-0.0010, -0.0497, 0.0234, ..., -0.0260, 0.0083, 0.0036], + [ 0.0136, -0.0323, 0.0015, ..., 0.0487, -0.0465, -0.0242], + ..., + [-0.0080, -0.0310, 0.0403, ..., 0.0014, 0.0286, 0.0355], + [-0.0205, 0.0139, -0.0262, ..., 0.0049, -0.0336, 0.0035], + [-0.0427, -0.0099, -0.0132, ..., -0.0368, 0.0377, 0.0084]], + device='cuda:0'), grad: tensor([[ 0.0134, 0.0190, 0.0089, ..., 0.0059, 0.0052, 0.0140], + [-0.0009, 0.0052, 0.0072, ..., 0.0007, 0.0034, 0.0076], + [ 0.0093, 0.0117, 0.0036, ..., 0.0014, 0.0030, 0.0083], + ..., + [-0.0027, -0.0126, -0.0267, ..., -0.0100, -0.0082, -0.0247], + [-0.0130, -0.0124, 0.0021, ..., 0.0054, 0.0062, 0.0083], + [ 0.0003, 0.0013, -0.0035, ..., 0.0035, 0.0020, -0.0114]], + device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0220, 0.0007, 0.0112, 0.0092, -0.0072, 0.0004, 0.0054, 0.0303, + -0.0284, 0.0286], device='cuda:0'), grad: tensor([ 0.0711, 0.0070, 0.0405, 0.0262, 0.0407, -0.0315, -0.0495, -0.1036, + 0.0211, -0.0219], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 220.75, cls_loss 1.1997 cls_loss_mapping 0.2134 cls_loss_causal 1.1503 re_mapping 0.0598 re_causal 0.1353 /// teacc 95.64 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0047, 0.0369, -0.0267, ..., -0.0265, 0.0036, -0.0144], + [-0.0031, -0.0526, 0.0220, ..., -0.0277, 0.0066, 0.0029], + [ 0.0146, -0.0335, 0.0025, ..., 0.0527, -0.0494, -0.0256], + ..., + [-0.0091, -0.0330, 0.0420, ..., -0.0005, 0.0274, 0.0362], + [-0.0193, 0.0146, -0.0275, ..., 0.0017, -0.0370, 0.0015], + [-0.0457, -0.0106, -0.0133, ..., -0.0395, 0.0394, 0.0090]], + device='cuda:0'), grad: tensor([[ 0.0053, -0.0075, -0.0006, ..., 0.0002, -0.0013, -0.0126], + [ 0.0055, 0.0205, 0.0138, ..., 0.0104, 0.0069, 0.0194], + [ 0.0042, 0.0123, 0.0093, ..., 0.0055, 0.0027, 0.0084], + ..., + [-0.0013, -0.0050, -0.0138, ..., -0.0020, -0.0009, -0.0031], + [-0.0040, -0.0168, 0.0030, ..., 0.0008, 0.0007, 0.0025], + [ 0.0023, 0.0027, 0.0010, ..., -0.0054, 0.0025, -0.0045]], + device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0215, 0.0002, 0.0111, 0.0102, -0.0069, 0.0007, 0.0062, 0.0303, + -0.0294, 0.0283], device='cuda:0'), grad: tensor([-0.0126, 0.0987, 0.0489, -0.0104, -0.0756, -0.0614, 0.0859, -0.0290, + -0.0336, -0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 6, time 216.67, cls_loss 1.1758 cls_loss_mapping 0.2151 cls_loss_causal 1.1197 re_mapping 0.0481 re_causal 0.1177 /// teacc 95.51 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0071, 0.0373, -0.0278, ..., -0.0267, 0.0017, -0.0146], + [-0.0043, -0.0548, 0.0207, ..., -0.0281, 0.0031, 0.0021], + [ 0.0151, -0.0343, 0.0048, ..., 0.0556, -0.0501, -0.0247], + ..., + [-0.0110, -0.0343, 0.0431, ..., -0.0021, 0.0270, 0.0365], + [-0.0181, 0.0156, -0.0286, ..., -0.0006, -0.0381, 0.0004], + [-0.0473, -0.0107, -0.0128, ..., -0.0404, 0.0408, 0.0102]], + device='cuda:0'), grad: tensor([[ 0.0034, 0.0058, -0.0032, ..., 0.0015, 0.0075, 0.0040], + [ 0.0014, 0.0004, -0.0065, ..., -0.0028, -0.0067, -0.0080], + [-0.0033, 0.0015, -0.0064, ..., -0.0001, 0.0041, -0.0035], + ..., + [ 0.0047, 0.0129, 0.0081, ..., 0.0033, 0.0067, 0.0100], + [ 0.0008, -0.0001, 0.0005, ..., 0.0011, 0.0009, -0.0007], + [ 0.0087, 0.0055, 0.0064, ..., -0.0045, 0.0001, 0.0044]], + device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0213, 0.0002, 0.0117, 0.0100, -0.0072, 0.0011, 0.0057, 0.0302, + -0.0292, 0.0284], device='cuda:0'), grad: tensor([ 0.0112, -0.0307, -0.0153, 0.0024, -0.0254, 0.0297, -0.0230, 0.0467, + -0.0155, 0.0199], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 216.87, cls_loss 1.1054 cls_loss_mapping 0.1779 cls_loss_causal 1.0566 re_mapping 0.0468 re_causal 0.1158 /// teacc 96.52 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0088, 0.0376, -0.0287, ..., -0.0272, 0.0012, -0.0147], + [-0.0066, -0.0573, 0.0195, ..., -0.0294, 0.0008, 0.0013], + [ 0.0157, -0.0350, 0.0052, ..., 0.0581, -0.0508, -0.0251], + ..., + [-0.0127, -0.0367, 0.0454, ..., -0.0029, 0.0261, 0.0376], + [-0.0163, 0.0164, -0.0288, ..., -0.0034, -0.0405, -0.0008], + [-0.0493, -0.0105, -0.0144, ..., -0.0417, 0.0422, 0.0097]], + device='cuda:0'), grad: tensor([[-0.0031, -0.0093, -0.0256, ..., -0.0034, -0.0073, -0.0296], + [ 0.0049, 0.0035, 0.0106, ..., 0.0040, 0.0062, 0.0158], + [-0.0025, 0.0025, -0.0007, ..., -0.0023, 0.0031, 0.0093], + ..., + [ 0.0004, -0.0001, 0.0129, ..., -0.0007, 0.0095, 0.0209], + [-0.0105, -0.0043, 0.0124, ..., 0.0073, 0.0067, 0.0165], + [-0.0189, -0.0245, -0.0086, ..., -0.0050, -0.0145, -0.0171]], + device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0212, -0.0003, 0.0116, 0.0109, -0.0070, 0.0013, 0.0058, 0.0300, + -0.0296, 0.0282], device='cuda:0'), grad: tensor([-0.0415, 0.0263, -0.0097, 0.0704, -0.0694, 0.0310, 0.0201, 0.0145, + 0.0471, -0.0886], device='cuda:0') +100 +0.0001 +changing lr +epoch 8, time 217.02, cls_loss 1.0884 cls_loss_mapping 0.1654 cls_loss_causal 1.0527 re_mapping 0.0405 re_causal 0.1071 /// teacc 96.35 lr 0.00010000 +Epoch 10, weight, value: tensor([[-9.5411e-03, 3.8084e-02, -2.9978e-02, ..., -2.7504e-02, + 8.0280e-05, -1.4423e-02], + [-7.3192e-03, -5.8688e-02, 1.9413e-02, ..., -2.9372e-02, + 1.5264e-03, 1.6764e-03], + [ 1.5923e-02, -3.5433e-02, 5.2620e-03, ..., 5.9611e-02, + -5.2308e-02, -2.6169e-02], + ..., + [-1.3919e-02, -3.8028e-02, 4.6723e-02, ..., -3.3156e-03, + 2.4247e-02, 3.7828e-02], + [-1.6081e-02, 1.6333e-02, -2.8588e-02, ..., -4.1636e-03, + -4.2218e-02, -1.3364e-03], + [-5.0494e-02, -1.0398e-02, -1.4826e-02, ..., -4.3717e-02, + 4.3390e-02, 1.0094e-02]], device='cuda:0'), grad: tensor([[ 0.0040, 0.0039, 0.0046, ..., -0.0005, 0.0033, 0.0014], + [-0.0032, -0.0011, 0.0012, ..., -0.0011, -0.0009, 0.0006], + [ 0.0139, 0.0181, 0.0160, ..., 0.0065, 0.0018, 0.0150], + ..., + [-0.0033, 0.0025, 0.0101, ..., -0.0060, -0.0013, 0.0007], + [ 0.0053, -0.0166, -0.0129, ..., -0.0018, -0.0025, -0.0068], + [-0.0073, -0.0079, -0.0018, ..., 0.0030, -0.0084, -0.0060]], + device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0210, -0.0005, 0.0116, 0.0109, -0.0070, 0.0022, 0.0055, 0.0299, + -0.0296, 0.0281], device='cuda:0'), grad: tensor([ 0.0154, -0.0066, 0.0435, 0.0589, -0.0491, 0.0026, -0.0157, -0.0076, + -0.0282, -0.0133], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 216.41, cls_loss 1.0504 cls_loss_mapping 0.1403 cls_loss_causal 1.0000 re_mapping 0.0377 re_causal 0.1006 /// teacc 96.50 lr 0.00010000 +Epoch 11, weight, value: tensor([[-1.0354e-02, 3.9085e-02, -3.1131e-02, ..., -2.8179e-02, + -1.0718e-03, -1.4528e-02], + [-7.6688e-03, -6.0395e-02, 1.8029e-02, ..., -2.9617e-02, + 7.5836e-05, 3.5740e-04], + [ 1.5929e-02, -3.6129e-02, 6.9685e-03, ..., 6.2759e-02, + -5.2846e-02, -2.5856e-02], + ..., + [-1.4738e-02, -3.9359e-02, 4.7834e-02, ..., -3.7286e-03, + 2.3894e-02, 3.8862e-02], + [-1.5861e-02, 1.7052e-02, -2.8853e-02, ..., -5.8908e-03, + -4.3508e-02, -1.0083e-03], + [-5.1986e-02, -1.0698e-02, -1.4607e-02, ..., -4.4143e-02, + 4.4128e-02, 1.0334e-02]], device='cuda:0'), grad: tensor([[ 0.0105, 0.0072, 0.0056, ..., 0.0015, 0.0026, 0.0003], + [ 0.0028, 0.0063, -0.0164, ..., 0.0009, 0.0005, -0.0056], + [ 0.0155, 0.0090, 0.0220, ..., 0.0020, 0.0007, 0.0131], + ..., + [-0.0017, -0.0044, -0.0351, ..., -0.0015, -0.0024, -0.0258], + [-0.0140, -0.0129, 0.0262, ..., 0.0002, 0.0034, 0.0161], + [ 0.0011, -0.0022, -0.0002, ..., -0.0014, -0.0042, 0.0025]], + device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0214, -0.0008, 0.0117, 0.0115, -0.0070, 0.0016, 0.0050, 0.0296, + -0.0293, 0.0283], device='cuda:0'), grad: tensor([ 0.0293, -0.0260, 0.0507, -0.0420, -0.0053, 0.0445, 0.0079, -0.0313, + -0.0134, -0.0145], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 223.52, cls_loss 1.0304 cls_loss_mapping 0.1483 cls_loss_causal 0.9890 re_mapping 0.0352 re_causal 0.0970 /// teacc 96.71 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0117, 0.0394, -0.0320, ..., -0.0286, -0.0022, -0.0137], + [-0.0078, -0.0608, 0.0174, ..., -0.0295, -0.0016, -0.0002], + [ 0.0162, -0.0368, 0.0074, ..., 0.0652, -0.0544, -0.0264], + ..., + [-0.0162, -0.0409, 0.0485, ..., -0.0051, 0.0236, 0.0393], + [-0.0145, 0.0179, -0.0294, ..., -0.0085, -0.0462, -0.0024], + [-0.0535, -0.0103, -0.0153, ..., -0.0454, 0.0451, 0.0103]], + device='cuda:0'), grad: tensor([[ 6.1913e-03, -6.5460e-03, 2.3785e-03, ..., 6.0768e-03, + 2.8076e-03, 2.5558e-03], + [ 3.0632e-03, 1.7639e-02, 7.4692e-03, ..., 1.0967e-03, + 6.0349e-03, 1.7105e-02], + [-2.2469e-03, -2.6741e-03, -9.1934e-03, ..., -2.8515e-03, + -2.3270e-03, -8.8348e-03], + ..., + [ 3.4943e-03, 1.2482e-02, 2.2087e-03, ..., 2.5826e-03, + 1.7939e-03, 1.0979e-02], + [-1.5366e-02, -1.1017e-02, -5.3825e-03, ..., -3.6373e-03, + 6.9733e-03, -6.9737e-05], + [ 1.1703e-02, 2.7802e-02, 1.7120e-02, ..., 2.4223e-03, + 1.3512e-02, 3.6682e-02]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0210, -0.0006, 0.0116, 0.0111, -0.0068, 0.0019, 0.0055, 0.0296, + -0.0293, 0.0282], device='cuda:0'), grad: tensor([-0.0032, 0.0504, -0.0212, -0.0570, 0.0157, -0.0019, -0.0776, 0.0358, + -0.0076, 0.0667], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 226.52, cls_loss 0.9942 cls_loss_mapping 0.1232 cls_loss_causal 0.9522 re_mapping 0.0341 re_causal 0.0936 /// teacc 97.27 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0125, 0.0398, -0.0335, ..., -0.0293, -0.0030, -0.0135], + [-0.0085, -0.0618, 0.0177, ..., -0.0301, -0.0021, -0.0008], + [ 0.0165, -0.0379, 0.0083, ..., 0.0672, -0.0562, -0.0267], + ..., + [-0.0173, -0.0413, 0.0490, ..., -0.0057, 0.0230, 0.0392], + [-0.0133, 0.0188, -0.0298, ..., -0.0105, -0.0478, -0.0028], + [-0.0545, -0.0105, -0.0156, ..., -0.0456, 0.0456, 0.0102]], + device='cuda:0'), grad: tensor([[ 0.0182, 0.0408, 0.0057, ..., 0.0018, 0.0061, 0.0054], + [ 0.0054, 0.0105, 0.0153, ..., 0.0033, 0.0125, 0.0147], + [ 0.0057, 0.0061, 0.0103, ..., 0.0128, 0.0005, 0.0012], + ..., + [ 0.0035, 0.0055, -0.0318, ..., -0.0015, -0.0149, -0.0262], + [ 0.0093, 0.0205, 0.0073, ..., 0.0028, 0.0067, 0.0090], + [ 0.0009, -0.0082, 0.0050, ..., 0.0016, 0.0046, 0.0063]], + device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0211, -0.0007, 0.0116, 0.0116, -0.0069, 0.0015, 0.0056, 0.0294, + -0.0291, 0.0282], device='cuda:0'), grad: tensor([ 0.0398, 0.0449, 0.0273, -0.1050, -0.0115, -0.0125, 0.0008, -0.0352, + 0.0491, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 225.77, cls_loss 0.9430 cls_loss_mapping 0.1201 cls_loss_causal 0.8963 re_mapping 0.0324 re_causal 0.0904 /// teacc 96.93 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0126, 0.0407, -0.0341, ..., -0.0303, -0.0041, -0.0130], + [-0.0095, -0.0627, 0.0171, ..., -0.0308, -0.0017, -0.0012], + [ 0.0163, -0.0383, 0.0091, ..., 0.0686, -0.0570, -0.0265], + ..., + [-0.0188, -0.0434, 0.0502, ..., -0.0052, 0.0223, 0.0398], + [-0.0129, 0.0191, -0.0301, ..., -0.0116, -0.0486, -0.0025], + [-0.0556, -0.0109, -0.0158, ..., -0.0466, 0.0468, 0.0101]], + device='cuda:0'), grad: tensor([[ 0.0203, 0.0132, -0.0043, ..., 0.0037, -0.0012, -0.0055], + [ 0.0041, 0.0098, 0.0003, ..., 0.0002, -0.0022, 0.0031], + [ 0.0092, 0.0108, 0.0003, ..., 0.0020, -0.0007, 0.0028], + ..., + [-0.0008, -0.0028, -0.0114, ..., -0.0021, -0.0096, -0.0189], + [-0.0107, -0.0190, -0.0123, ..., -0.0106, -0.0052, -0.0091], + [ 0.0057, 0.0100, 0.0154, ..., 0.0057, 0.0081, 0.0170]], + device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0215, -0.0010, 0.0118, 0.0116, -0.0063, 0.0013, 0.0053, 0.0292, + -0.0291, 0.0279], device='cuda:0'), grad: tensor([-1.7197e-02, 1.9135e-02, 2.2293e-02, 1.8539e-02, -1.8311e-02, + 2.2568e-02, -7.8082e-05, -3.4943e-02, -5.2856e-02, 4.0771e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 228.56, cls_loss 0.9522 cls_loss_mapping 0.1339 cls_loss_causal 0.9151 re_mapping 0.0299 re_causal 0.0857 /// teacc 97.62 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0142, 0.0411, -0.0353, ..., -0.0309, -0.0051, -0.0130], + [-0.0101, -0.0637, 0.0172, ..., -0.0304, -0.0016, -0.0021], + [ 0.0170, -0.0381, 0.0095, ..., 0.0713, -0.0573, -0.0271], + ..., + [-0.0195, -0.0446, 0.0511, ..., -0.0055, 0.0216, 0.0401], + [-0.0119, 0.0194, -0.0302, ..., -0.0126, -0.0496, -0.0022], + [-0.0573, -0.0110, -0.0160, ..., -0.0479, 0.0474, 0.0106]], + device='cuda:0'), grad: tensor([[ 0.0035, 0.0056, 0.0004, ..., -0.0097, 0.0037, 0.0042], + [ 0.0058, 0.0114, 0.0077, ..., 0.0022, 0.0029, 0.0072], + [ 0.0040, 0.0125, 0.0073, ..., 0.0058, 0.0023, 0.0087], + ..., + [ 0.0025, -0.0094, -0.0165, ..., -0.0085, -0.0064, -0.0250], + [-0.0030, -0.0268, -0.0011, ..., 0.0024, -0.0028, -0.0055], + [ 0.0053, 0.0165, -0.0008, ..., 0.0029, 0.0046, 0.0106]], + device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0212, -0.0011, 0.0118, 0.0118, -0.0062, 0.0015, 0.0053, 0.0290, + -0.0289, 0.0277], device='cuda:0'), grad: tensor([ 0.0039, 0.0280, 0.0319, 0.0123, 0.0382, -0.0155, -0.0292, -0.0605, + -0.0226, 0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 227.47, cls_loss 0.9330 cls_loss_mapping 0.1255 cls_loss_causal 0.8960 re_mapping 0.0289 re_causal 0.0839 /// teacc 97.44 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0149, 0.0413, -0.0356, ..., -0.0316, -0.0059, -0.0131], + [-0.0104, -0.0643, 0.0165, ..., -0.0314, -0.0028, -0.0034], + [ 0.0169, -0.0388, 0.0104, ..., 0.0731, -0.0575, -0.0267], + ..., + [-0.0210, -0.0458, 0.0520, ..., -0.0056, 0.0216, 0.0410], + [-0.0116, 0.0192, -0.0311, ..., -0.0142, -0.0515, -0.0038], + [-0.0574, -0.0102, -0.0163, ..., -0.0494, 0.0480, 0.0102]], + device='cuda:0'), grad: tensor([[ 0.0052, 0.0082, 0.0057, ..., 0.0028, 0.0018, 0.0071], + [-0.0061, -0.0082, -0.0092, ..., -0.0005, -0.0017, -0.0109], + [-0.0047, -0.0064, -0.0120, ..., 0.0058, -0.0047, -0.0117], + ..., + [ 0.0034, 0.0083, 0.0093, ..., 0.0040, 0.0049, 0.0117], + [ 0.0044, 0.0087, 0.0068, ..., 0.0023, 0.0021, 0.0072], + [-0.0075, -0.0223, -0.0004, ..., 0.0026, -0.0038, -0.0041]], + device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0211, -0.0013, 0.0116, 0.0123, -0.0062, 0.0019, 0.0056, 0.0288, + -0.0292, 0.0276], device='cuda:0'), grad: tensor([ 0.0270, -0.0314, -0.0223, -0.0402, 0.0208, 0.0186, -0.0069, 0.0320, + 0.0285, -0.0262], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 226.42, cls_loss 0.9395 cls_loss_mapping 0.1125 cls_loss_causal 0.9087 re_mapping 0.0284 re_causal 0.0858 /// teacc 97.45 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0149, 0.0424, -0.0370, ..., -0.0337, -0.0066, -0.0135], + [-0.0111, -0.0650, 0.0162, ..., -0.0319, -0.0039, -0.0043], + [ 0.0164, -0.0394, 0.0121, ..., 0.0754, -0.0576, -0.0254], + ..., + [-0.0222, -0.0466, 0.0518, ..., -0.0060, 0.0220, 0.0407], + [-0.0120, 0.0193, -0.0312, ..., -0.0157, -0.0515, -0.0037], + [-0.0582, -0.0102, -0.0158, ..., -0.0491, 0.0482, 0.0105]], + device='cuda:0'), grad: tensor([[ 0.0064, 0.0133, 0.0190, ..., 0.0082, 0.0030, 0.0140], + [ 0.0042, 0.0012, -0.0141, ..., -0.0048, -0.0019, -0.0106], + [ 0.0141, 0.0029, 0.0435, ..., 0.0195, 0.0041, 0.0168], + ..., + [-0.0010, -0.0126, -0.0160, ..., 0.0073, -0.0083, -0.0145], + [-0.0044, 0.0082, -0.0318, ..., -0.0239, 0.0014, -0.0022], + [-0.0239, -0.0244, -0.0199, ..., -0.0089, -0.0065, -0.0188]], + device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0211, -0.0016, 0.0117, 0.0128, -0.0065, 0.0019, 0.0057, 0.0285, + -0.0293, 0.0279], device='cuda:0'), grad: tensor([ 0.0454, -0.0235, 0.0379, -0.0055, 0.0306, 0.0068, 0.0227, -0.0376, + 0.0035, -0.0803], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 226.64, cls_loss 0.8922 cls_loss_mapping 0.1079 cls_loss_causal 0.8550 re_mapping 0.0267 re_causal 0.0747 /// teacc 97.35 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0163, 0.0425, -0.0385, ..., -0.0356, -0.0071, -0.0133], + [-0.0111, -0.0651, 0.0163, ..., -0.0311, -0.0038, -0.0041], + [ 0.0162, -0.0405, 0.0125, ..., 0.0770, -0.0576, -0.0255], + ..., + [-0.0226, -0.0474, 0.0524, ..., -0.0052, 0.0216, 0.0409], + [-0.0105, 0.0203, -0.0322, ..., -0.0164, -0.0525, -0.0039], + [-0.0581, -0.0098, -0.0155, ..., -0.0502, 0.0489, 0.0105]], + device='cuda:0'), grad: tensor([[ 4.0894e-03, 7.3051e-03, 5.5351e-03, ..., 1.2331e-03, + 9.6679e-05, 4.8141e-03], + [ 3.2330e-03, 4.6959e-03, -1.2337e-02, ..., -1.4849e-03, + 1.9777e-04, -2.2869e-03], + [ 5.5847e-03, 1.2085e-02, 1.5640e-02, ..., 1.9951e-03, + 5.7077e-04, 1.5533e-02], + ..., + [-2.7618e-03, -7.1945e-03, 2.5272e-03, ..., 4.1533e-04, + 4.5776e-03, 4.8294e-03], + [-2.0264e-02, -2.3911e-02, 3.1052e-03, ..., 1.3313e-03, + 5.8603e-04, -4.1842e-04], + [ 3.8853e-03, 4.1161e-03, 5.7068e-03, ..., 2.4719e-03, + -1.6375e-03, 4.0207e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0208, -0.0013, 0.0114, 0.0127, -0.0071, 0.0020, 0.0056, 0.0290, + -0.0293, 0.0282], device='cuda:0'), grad: tensor([ 0.0211, -0.0112, 0.0481, -0.0007, -0.0090, 0.0173, -0.0377, -0.0231, + -0.0222, 0.0174], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 226.81, cls_loss 0.8646 cls_loss_mapping 0.1033 cls_loss_causal 0.8408 re_mapping 0.0270 re_causal 0.0784 /// teacc 97.22 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0166, 0.0430, -0.0390, ..., -0.0347, -0.0082, -0.0137], + [-0.0112, -0.0652, 0.0166, ..., -0.0314, -0.0042, -0.0047], + [ 0.0168, -0.0408, 0.0134, ..., 0.0781, -0.0580, -0.0256], + ..., + [-0.0237, -0.0490, 0.0523, ..., -0.0050, 0.0213, 0.0405], + [-0.0106, 0.0204, -0.0315, ..., -0.0171, -0.0524, -0.0034], + [-0.0594, -0.0098, -0.0154, ..., -0.0516, 0.0499, 0.0112]], + device='cuda:0'), grad: tensor([[-2.9259e-03, -1.3229e-02, 9.2864e-05, ..., 1.5287e-03, + 1.1444e-03, -5.7716e-03], + [-1.4048e-03, -6.3744e-03, -5.8899e-03, ..., 6.0558e-04, + -2.2259e-03, -2.0924e-03], + [ 2.9507e-03, 8.2474e-03, 2.4910e-03, ..., -3.0956e-03, + 2.0657e-03, 4.3488e-03], + ..., + [-4.4174e-03, -2.0027e-03, -1.3535e-02, ..., -5.4970e-03, + -8.0347e-04, -8.9951e-03], + [ 2.9259e-03, 4.6501e-03, 6.5346e-03, ..., 2.2221e-03, + 4.9667e-03, 3.1013e-03], + [ 2.7599e-03, 2.5406e-03, 1.4544e-03, ..., 2.4014e-03, + -6.1378e-03, 4.6945e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0208, -0.0009, 0.0112, 0.0126, -0.0066, 0.0020, 0.0050, 0.0286, + -0.0288, 0.0281], device='cuda:0'), grad: tensor([-0.0038, 0.0013, 0.0240, 0.0178, 0.0394, -0.0431, -0.0133, -0.0372, + -0.0026, 0.0175], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 228.00, cls_loss 0.8921 cls_loss_mapping 0.0992 cls_loss_causal 0.8556 re_mapping 0.0253 re_causal 0.0762 /// teacc 97.75 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0173, 0.0432, -0.0389, ..., -0.0348, -0.0092, -0.0140], + [-0.0121, -0.0657, 0.0159, ..., -0.0328, -0.0051, -0.0054], + [ 0.0171, -0.0412, 0.0134, ..., 0.0799, -0.0590, -0.0263], + ..., + [-0.0240, -0.0493, 0.0531, ..., -0.0049, 0.0218, 0.0417], + [-0.0106, 0.0201, -0.0319, ..., -0.0181, -0.0539, -0.0040], + [-0.0606, -0.0100, -0.0159, ..., -0.0535, 0.0512, 0.0109]], + device='cuda:0'), grad: tensor([[-6.9389e-03, -3.0869e-02, -6.8359e-03, ..., 2.8267e-03, + 2.9984e-03, -8.0719e-03], + [ 7.3700e-03, 1.3245e-02, 2.0096e-02, ..., 8.4991e-03, + 6.7520e-03, 1.5106e-02], + [-6.7787e-03, -6.5880e-03, -8.1863e-03, ..., -1.0071e-02, + -7.2050e-04, -3.9940e-03], + ..., + [ 6.8474e-03, 6.6795e-03, 7.7171e-03, ..., 3.4561e-03, + -1.1650e-02, 9.1982e-04], + [ 1.0880e-02, 1.3695e-02, 7.2479e-03, ..., 3.1662e-03, + -2.4080e-05, 2.7599e-03], + [-1.3031e-02, -2.5223e-02, -1.6586e-02, ..., -2.5024e-03, + -1.0941e-02, -2.2964e-02]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0212, -0.0017, 0.0111, 0.0127, -0.0064, 0.0024, 0.0051, 0.0292, + -0.0295, 0.0280], device='cuda:0'), grad: tensor([-0.0350, 0.0636, -0.0220, -0.0195, 0.0234, 0.0222, -0.0275, 0.0199, + 0.0306, -0.0558], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 226.29, cls_loss 0.8621 cls_loss_mapping 0.0878 cls_loss_causal 0.8357 re_mapping 0.0247 re_causal 0.0714 /// teacc 97.48 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0181, 0.0434, -0.0396, ..., -0.0358, -0.0102, -0.0141], + [-0.0132, -0.0666, 0.0157, ..., -0.0328, -0.0051, -0.0055], + [ 0.0179, -0.0417, 0.0141, ..., 0.0816, -0.0590, -0.0268], + ..., + [-0.0231, -0.0496, 0.0530, ..., -0.0059, 0.0209, 0.0418], + [-0.0103, 0.0203, -0.0321, ..., -0.0191, -0.0535, -0.0042], + [-0.0612, -0.0096, -0.0156, ..., -0.0541, 0.0514, 0.0107]], + device='cuda:0'), grad: tensor([[ 2.1992e-03, -9.3002e-03, -1.1650e-02, ..., -1.3819e-03, + -9.9487e-03, -2.0584e-02], + [-1.2733e-02, -1.8265e-02, 4.3064e-05, ..., 1.4341e-04, + -6.8130e-03, 4.4594e-03], + [ 1.2909e-02, 7.2060e-03, 2.8976e-02, ..., 1.8082e-02, + 3.5419e-03, 9.3384e-03], + ..., + [-1.6724e-02, -2.5238e-02, -7.3059e-02, ..., -2.6596e-02, + -1.8494e-02, -3.3600e-02], + [ 1.1269e-02, 2.0782e-02, 9.5215e-03, ..., 3.7632e-03, + 6.3286e-03, 8.1177e-03], + [ 4.5624e-03, 2.4948e-02, 4.1412e-02, ..., 5.7449e-03, + 2.1713e-02, 3.1586e-02]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0211, -0.0016, 0.0110, 0.0133, -0.0065, 0.0027, 0.0049, 0.0289, + -0.0298, 0.0282], device='cuda:0'), grad: tensor([-0.0356, -0.0396, 0.0470, -0.0423, 0.0033, 0.0089, 0.0145, -0.1085, + 0.0559, 0.0963], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 226.67, cls_loss 0.8654 cls_loss_mapping 0.1032 cls_loss_causal 0.8324 re_mapping 0.0232 re_causal 0.0702 /// teacc 97.52 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0190, 0.0439, -0.0400, ..., -0.0367, -0.0104, -0.0140], + [-0.0135, -0.0673, 0.0155, ..., -0.0326, -0.0044, -0.0059], + [ 0.0173, -0.0426, 0.0143, ..., 0.0832, -0.0592, -0.0271], + ..., + [-0.0237, -0.0500, 0.0537, ..., -0.0060, 0.0202, 0.0421], + [-0.0092, 0.0209, -0.0326, ..., -0.0201, -0.0547, -0.0041], + [-0.0622, -0.0102, -0.0161, ..., -0.0555, 0.0521, 0.0103]], + device='cuda:0'), grad: tensor([[ 9.2840e-04, -1.0890e-04, 4.8332e-03, ..., 1.5068e-03, + -1.3340e-04, 1.8778e-03], + [ 4.8790e-03, 4.8103e-03, 7.1907e-03, ..., 3.1548e-03, + 2.2240e-03, 7.1907e-03], + [ 3.2177e-03, -5.6267e-04, -6.7177e-03, ..., -3.7346e-03, + -4.1885e-03, -9.6664e-03], + ..., + [ 3.1776e-03, 1.5732e-02, 1.1436e-02, ..., 3.1757e-03, + 4.6387e-03, 1.0147e-02], + [-1.3863e-02, -2.7008e-02, -3.7231e-03, ..., -7.5579e-05, + -1.1854e-03, -3.1338e-03], + [ 3.2368e-03, -4.3564e-03, -1.8036e-02, ..., -9.3002e-03, + -9.2239e-03, -1.4351e-02]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0212, -0.0020, 0.0108, 0.0132, -0.0065, 0.0026, 0.0055, 0.0289, + -0.0298, 0.0282], device='cuda:0'), grad: tensor([ 0.0101, 0.0344, -0.0279, 0.0219, -0.0088, 0.0154, 0.0239, 0.0567, + -0.0887, -0.0370], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 227.12, cls_loss 0.9032 cls_loss_mapping 0.0922 cls_loss_causal 0.8670 re_mapping 0.0237 re_causal 0.0716 /// teacc 97.72 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0196, 0.0438, -0.0410, ..., -0.0374, -0.0113, -0.0147], + [-0.0136, -0.0686, 0.0149, ..., -0.0323, -0.0044, -0.0072], + [ 0.0171, -0.0426, 0.0149, ..., 0.0844, -0.0596, -0.0267], + ..., + [-0.0240, -0.0507, 0.0541, ..., -0.0062, 0.0197, 0.0427], + [-0.0093, 0.0208, -0.0322, ..., -0.0206, -0.0556, -0.0036], + [-0.0637, -0.0099, -0.0160, ..., -0.0546, 0.0528, 0.0101]], + device='cuda:0'), grad: tensor([[-0.0097, -0.0066, -0.0040, ..., -0.0015, 0.0003, -0.0019], + [-0.0025, -0.0061, -0.0154, ..., -0.0033, 0.0006, -0.0070], + [-0.0054, -0.0035, -0.0194, ..., -0.0116, -0.0003, -0.0053], + ..., + [ 0.0024, -0.0053, 0.0294, ..., 0.0048, 0.0125, 0.0173], + [ 0.0006, -0.0015, 0.0062, ..., 0.0029, 0.0009, 0.0021], + [ 0.0021, 0.0044, -0.0259, ..., 0.0018, -0.0167, -0.0362]], + device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0213, -0.0028, 0.0111, 0.0138, -0.0058, 0.0020, 0.0054, 0.0289, + -0.0298, 0.0281], device='cuda:0'), grad: tensor([-0.0272, -0.0363, -0.0356, 0.0399, 0.0600, 0.0243, -0.0034, 0.0330, + 0.0022, -0.0569], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 229.72, cls_loss 0.8259 cls_loss_mapping 0.0869 cls_loss_causal 0.7883 re_mapping 0.0237 re_causal 0.0675 /// teacc 97.84 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0203, 0.0443, -0.0419, ..., -0.0374, -0.0123, -0.0155], + [-0.0145, -0.0698, 0.0146, ..., -0.0326, -0.0048, -0.0074], + [ 0.0170, -0.0432, 0.0156, ..., 0.0860, -0.0597, -0.0268], + ..., + [-0.0235, -0.0512, 0.0547, ..., -0.0059, 0.0189, 0.0429], + [-0.0089, 0.0212, -0.0327, ..., -0.0206, -0.0561, -0.0041], + [-0.0639, -0.0099, -0.0160, ..., -0.0561, 0.0532, 0.0103]], + device='cuda:0'), grad: tensor([[-1.7061e-03, -1.0284e-02, -3.1013e-03, ..., -5.5885e-04, + 5.6791e-04, -1.5342e-04], + [-3.7708e-03, -5.7144e-03, -1.8250e-02, ..., -8.8425e-03, + -8.2970e-05, -1.1726e-02], + [ 6.2714e-03, 7.9880e-03, 2.8992e-02, ..., 9.4376e-03, + 1.2999e-03, 1.9806e-02], + ..., + [ 5.4628e-05, 1.0895e-02, -1.7900e-03, ..., -2.8825e-04, + 9.2163e-03, 8.7814e-03], + [ 4.1695e-03, 1.3609e-03, 4.4632e-03, ..., 1.4477e-03, + 8.3113e-04, 3.4275e-03], + [ 1.2712e-03, 1.2964e-05, -9.9945e-03, ..., 2.8534e-03, + -9.3613e-03, -1.2466e-02]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0208, -0.0022, 0.0113, 0.0137, -0.0063, 0.0020, 0.0058, 0.0292, + -0.0299, 0.0278], device='cuda:0'), grad: tensor([-0.0138, -0.0447, 0.0674, -0.0774, 0.0284, 0.0137, -0.0066, 0.0156, + 0.0188, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 226.55, cls_loss 0.8239 cls_loss_mapping 0.0745 cls_loss_causal 0.7879 re_mapping 0.0229 re_causal 0.0662 /// teacc 97.84 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0204, 0.0452, -0.0422, ..., -0.0377, -0.0121, -0.0155], + [-0.0153, -0.0704, 0.0148, ..., -0.0323, -0.0048, -0.0068], + [ 0.0174, -0.0442, 0.0157, ..., 0.0872, -0.0599, -0.0273], + ..., + [-0.0238, -0.0519, 0.0545, ..., -0.0068, 0.0182, 0.0421], + [-0.0086, 0.0216, -0.0322, ..., -0.0218, -0.0568, -0.0046], + [-0.0647, -0.0098, -0.0161, ..., -0.0562, 0.0535, 0.0102]], + device='cuda:0'), grad: tensor([[-0.0136, -0.0254, 0.0036, ..., 0.0014, -0.0019, -0.0058], + [ 0.0054, 0.0083, 0.0044, ..., 0.0009, 0.0036, 0.0070], + [-0.0176, -0.0170, -0.0108, ..., -0.0116, -0.0067, -0.0167], + ..., + [-0.0022, -0.0040, -0.0141, ..., -0.0009, -0.0087, -0.0236], + [ 0.0042, 0.0045, -0.0033, ..., 0.0008, -0.0034, -0.0063], + [ 0.0028, 0.0046, 0.0109, ..., 0.0010, 0.0112, 0.0213]], + device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0215, -0.0024, 0.0112, 0.0138, -0.0055, 0.0021, 0.0046, 0.0284, + -0.0296, 0.0280], device='cuda:0'), grad: tensor([-0.0435, 0.0354, -0.0656, 0.0272, 0.0455, -0.0083, 0.0266, -0.0462, + -0.0084, 0.0373], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 229.48, cls_loss 0.8154 cls_loss_mapping 0.0748 cls_loss_causal 0.7841 re_mapping 0.0230 re_causal 0.0638 /// teacc 98.16 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0205, 0.0465, -0.0428, ..., -0.0384, -0.0122, -0.0151], + [-0.0148, -0.0704, 0.0148, ..., -0.0317, -0.0060, -0.0078], + [ 0.0172, -0.0446, 0.0156, ..., 0.0881, -0.0599, -0.0277], + ..., + [-0.0247, -0.0524, 0.0551, ..., -0.0069, 0.0178, 0.0426], + [-0.0081, 0.0217, -0.0323, ..., -0.0234, -0.0579, -0.0049], + [-0.0657, -0.0102, -0.0154, ..., -0.0566, 0.0553, 0.0110]], + device='cuda:0'), grad: tensor([[-2.2297e-03, -7.2441e-03, -4.8332e-03, ..., 5.1880e-04, + -9.4318e-04, -4.0016e-03], + [ 4.0932e-03, 4.9782e-03, 6.8359e-03, ..., 1.2445e-03, + 1.0557e-03, 5.0774e-03], + [-1.5060e-02, -7.4081e-03, -3.3264e-02, ..., -1.1208e-02, + 2.5129e-04, -1.7334e-02], + ..., + [ 1.0277e-02, 5.5733e-03, 1.4114e-02, ..., 8.6899e-03, + -3.6955e-05, 1.5732e-02], + [-2.2078e-04, -3.1090e-03, -2.9316e-03, ..., 2.4757e-03, + -6.3276e-04, -8.0490e-03], + [ 5.4512e-03, 4.9973e-03, 1.0757e-02, ..., 1.0729e-03, + -2.2545e-03, 4.9057e-03]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0212, -0.0023, 0.0114, 0.0136, -0.0060, 0.0020, 0.0049, 0.0289, + -0.0296, 0.0279], device='cuda:0'), grad: tensor([-0.0169, 0.0242, -0.0336, 0.0439, -0.0021, -0.0627, -0.0045, 0.0478, + -0.0260, 0.0301], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 226.62, cls_loss 0.8332 cls_loss_mapping 0.0719 cls_loss_causal 0.8090 re_mapping 0.0223 re_causal 0.0647 /// teacc 97.96 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0211, 0.0463, -0.0426, ..., -0.0392, -0.0133, -0.0150], + [-0.0155, -0.0714, 0.0147, ..., -0.0322, -0.0054, -0.0082], + [ 0.0174, -0.0445, 0.0160, ..., 0.0890, -0.0603, -0.0278], + ..., + [-0.0257, -0.0531, 0.0556, ..., -0.0063, 0.0170, 0.0426], + [-0.0082, 0.0216, -0.0326, ..., -0.0243, -0.0584, -0.0051], + [-0.0666, -0.0104, -0.0149, ..., -0.0578, 0.0562, 0.0113]], + device='cuda:0'), grad: tensor([[ 1.0319e-03, -2.9850e-03, 3.9330e-03, ..., 9.5987e-04, + 1.1444e-03, 4.0169e-03], + [-6.7139e-03, -7.9727e-03, -9.6436e-03, ..., 1.6432e-03, + 5.1212e-04, -9.9564e-03], + [ 2.8000e-03, 2.5921e-03, -6.6261e-03, ..., -5.0850e-03, + 1.0509e-03, -2.4014e-03], + ..., + [ 1.6155e-03, 4.8332e-03, -1.2268e-02, ..., -1.8489e-04, + -1.0986e-02, -1.1139e-02], + [ 2.4586e-03, 4.4632e-03, 1.3247e-03, ..., -4.5085e-04, + 1.1854e-03, 1.1806e-03], + [ 9.2745e-04, -1.1406e-03, 7.9269e-03, ..., 4.7445e-05, + 9.2239e-03, 1.0620e-02]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0207, -0.0023, 0.0117, 0.0140, -0.0060, 0.0017, 0.0050, 0.0287, + -0.0300, 0.0286], device='cuda:0'), grad: tensor([-0.0113, -0.0451, -0.0295, 0.0090, 0.0303, -0.0069, 0.0267, -0.0054, + 0.0153, 0.0169], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 227.40, cls_loss 0.7994 cls_loss_mapping 0.0760 cls_loss_causal 0.7791 re_mapping 0.0229 re_causal 0.0636 /// teacc 98.20 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0214, 0.0468, -0.0424, ..., -0.0394, -0.0138, -0.0145], + [-0.0155, -0.0717, 0.0146, ..., -0.0320, -0.0058, -0.0090], + [ 0.0179, -0.0448, 0.0154, ..., 0.0901, -0.0613, -0.0284], + ..., + [-0.0267, -0.0544, 0.0558, ..., -0.0071, 0.0169, 0.0428], + [-0.0079, 0.0218, -0.0321, ..., -0.0254, -0.0583, -0.0036], + [-0.0681, -0.0111, -0.0150, ..., -0.0594, 0.0562, 0.0108]], + device='cuda:0'), grad: tensor([[ 3.2692e-03, 7.6485e-03, 5.6458e-03, ..., 1.3237e-03, + 1.3247e-03, 5.1689e-03], + [ 1.4029e-03, -1.3666e-03, -7.2403e-03, ..., -1.5259e-03, + -4.7226e-03, -1.7197e-02], + [-2.1347e-02, -1.7120e-02, -2.4612e-02, ..., -1.4130e-02, + -3.8910e-03, -1.4183e-02], + ..., + [ 1.2612e-04, 3.6049e-03, 4.3449e-03, ..., 7.5102e-05, + 2.0905e-03, 5.0049e-03], + [ 1.8520e-03, -1.2708e-04, 2.1992e-03, ..., 3.5725e-03, + 1.9855e-03, 7.0877e-03], + [-2.8744e-03, -1.2283e-02, 1.3285e-03, ..., 4.7350e-04, + -3.8948e-03, 1.0500e-03]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0213, -0.0023, 0.0115, 0.0146, -0.0057, 0.0020, 0.0046, 0.0280, + -0.0300, 0.0280], device='cuda:0'), grad: tensor([ 0.0303, -0.0131, -0.1014, 0.0478, 0.0120, 0.0176, -0.0164, 0.0080, + 0.0163, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 226.95, cls_loss 0.8180 cls_loss_mapping 0.0683 cls_loss_causal 0.7849 re_mapping 0.0214 re_causal 0.0627 /// teacc 97.88 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0221, 0.0469, -0.0431, ..., -0.0401, -0.0140, -0.0144], + [-0.0162, -0.0723, 0.0146, ..., -0.0313, -0.0049, -0.0094], + [ 0.0179, -0.0453, 0.0159, ..., 0.0908, -0.0625, -0.0284], + ..., + [-0.0276, -0.0551, 0.0559, ..., -0.0074, 0.0164, 0.0432], + [-0.0081, 0.0218, -0.0327, ..., -0.0255, -0.0578, -0.0041], + [-0.0686, -0.0112, -0.0146, ..., -0.0593, 0.0560, 0.0109]], + device='cuda:0'), grad: tensor([[ 1.2245e-03, 3.4943e-03, 7.3624e-03, ..., 5.9032e-04, + 2.0542e-03, 7.8430e-03], + [ 4.8790e-03, 6.0310e-03, 8.1253e-03, ..., 7.9489e-04, + 1.9608e-03, 1.4610e-02], + [-8.1956e-05, 3.9520e-03, -4.0512e-03, ..., -5.2147e-03, + 1.0653e-03, -5.2147e-03], + ..., + [ 1.7805e-03, -4.3831e-03, -1.1299e-02, ..., -2.5101e-03, + -1.0666e-02, -1.5808e-02], + [-6.8321e-03, -2.4242e-03, -8.1482e-03, ..., -1.3828e-03, + 1.1711e-03, -3.5095e-03], + [-1.0201e-02, -1.8600e-02, -4.5204e-03, ..., 7.8535e-04, + 2.6894e-03, -1.6203e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0212, -0.0023, 0.0114, 0.0144, -0.0054, 0.0020, 0.0047, 0.0279, + -0.0303, 0.0284], device='cuda:0'), grad: tensor([ 0.0155, 0.0413, -0.0022, 0.0218, 0.0045, -0.0391, 0.0278, -0.0270, + -0.0303, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 226.17, cls_loss 0.7914 cls_loss_mapping 0.0711 cls_loss_causal 0.7628 re_mapping 0.0226 re_causal 0.0643 /// teacc 97.75 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0224, 0.0474, -0.0438, ..., -0.0394, -0.0155, -0.0149], + [-0.0169, -0.0735, 0.0152, ..., -0.0303, -0.0052, -0.0103], + [ 0.0180, -0.0457, 0.0159, ..., 0.0918, -0.0627, -0.0286], + ..., + [-0.0285, -0.0562, 0.0572, ..., -0.0069, 0.0161, 0.0447], + [-0.0074, 0.0218, -0.0329, ..., -0.0254, -0.0581, -0.0042], + [-0.0694, -0.0110, -0.0153, ..., -0.0615, 0.0569, 0.0103]], + device='cuda:0'), grad: tensor([[ 0.0031, 0.0059, 0.0052, ..., 0.0018, 0.0016, 0.0051], + [-0.0076, -0.0100, -0.0247, ..., -0.0075, -0.0026, -0.0199], + [ 0.0036, 0.0079, 0.0092, ..., 0.0027, 0.0018, 0.0100], + ..., + [-0.0049, -0.0095, -0.0141, ..., -0.0029, -0.0027, -0.0103], + [-0.0019, -0.0021, 0.0004, ..., 0.0011, 0.0019, 0.0011], + [ 0.0028, 0.0016, 0.0061, ..., 0.0011, -0.0016, 0.0041]], + device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0208, -0.0020, 0.0113, 0.0142, -0.0056, 0.0021, 0.0050, 0.0281, + -0.0299, 0.0281], device='cuda:0'), grad: tensor([ 0.0258, -0.0979, 0.0434, 0.0533, -0.0159, 0.0296, -0.0174, -0.0352, + -0.0012, 0.0154], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 226.36, cls_loss 0.7956 cls_loss_mapping 0.0745 cls_loss_causal 0.7673 re_mapping 0.0216 re_causal 0.0644 /// teacc 98.25 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0226, 0.0474, -0.0447, ..., -0.0408, -0.0171, -0.0154], + [-0.0177, -0.0743, 0.0142, ..., -0.0312, -0.0057, -0.0119], + [ 0.0190, -0.0453, 0.0163, ..., 0.0929, -0.0619, -0.0290], + ..., + [-0.0291, -0.0570, 0.0577, ..., -0.0069, 0.0157, 0.0460], + [-0.0069, 0.0219, -0.0328, ..., -0.0265, -0.0601, -0.0045], + [-0.0700, -0.0103, -0.0155, ..., -0.0620, 0.0576, 0.0100]], + device='cuda:0'), grad: tensor([[ 0.0026, 0.0036, 0.0049, ..., 0.0022, 0.0019, 0.0048], + [-0.0024, -0.0044, -0.0212, ..., 0.0013, -0.0254, -0.0129], + [ 0.0046, 0.0047, -0.0035, ..., 0.0019, -0.0016, -0.0043], + ..., + [-0.0010, -0.0020, -0.0081, ..., 0.0019, -0.0100, 0.0033], + [-0.0102, -0.0066, 0.0033, ..., 0.0006, 0.0083, 0.0019], + [ 0.0060, 0.0208, 0.0419, ..., 0.0010, 0.0387, 0.0329]], + device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0206, -0.0023, 0.0111, 0.0142, -0.0063, 0.0025, 0.0054, 0.0282, + -0.0298, 0.0285], device='cuda:0'), grad: tensor([ 0.0185, -0.0408, 0.0036, -0.0357, -0.0212, 0.0294, -0.0035, 0.0057, + -0.0279, 0.0719], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 225.92, cls_loss 0.7558 cls_loss_mapping 0.0641 cls_loss_causal 0.7325 re_mapping 0.0210 re_causal 0.0592 /// teacc 98.09 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0233, 0.0476, -0.0443, ..., -0.0418, -0.0178, -0.0154], + [-0.0178, -0.0742, 0.0140, ..., -0.0307, -0.0059, -0.0129], + [ 0.0196, -0.0457, 0.0168, ..., 0.0938, -0.0617, -0.0292], + ..., + [-0.0301, -0.0574, 0.0579, ..., -0.0070, 0.0141, 0.0461], + [-0.0070, 0.0225, -0.0338, ..., -0.0285, -0.0612, -0.0050], + [-0.0704, -0.0107, -0.0151, ..., -0.0637, 0.0585, 0.0107]], + device='cuda:0'), grad: tensor([[ 0.0002, -0.0007, 0.0006, ..., 0.0028, -0.0002, -0.0017], + [-0.0008, 0.0029, -0.0092, ..., 0.0010, -0.0101, 0.0010], + [-0.0024, -0.0068, -0.0155, ..., -0.0069, -0.0046, -0.0124], + ..., + [ 0.0018, 0.0032, 0.0072, ..., 0.0024, 0.0020, -0.0021], + [ 0.0035, 0.0061, 0.0092, ..., 0.0024, 0.0061, 0.0028], + [ 0.0025, 0.0164, -0.0024, ..., 0.0012, 0.0212, 0.0139]], + device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0208, -0.0021, 0.0110, 0.0146, -0.0064, 0.0015, 0.0055, 0.0280, + -0.0299, 0.0287], device='cuda:0'), grad: tensor([ 0.0008, -0.0382, -0.0518, -0.0333, -0.0009, -0.0091, 0.0288, 0.0235, + 0.0388, 0.0414], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 226.04, cls_loss 0.7964 cls_loss_mapping 0.0711 cls_loss_causal 0.7664 re_mapping 0.0209 re_causal 0.0561 /// teacc 97.88 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0243, 0.0477, -0.0453, ..., -0.0427, -0.0189, -0.0158], + [-0.0175, -0.0752, 0.0140, ..., -0.0300, -0.0076, -0.0138], + [ 0.0198, -0.0458, 0.0168, ..., 0.0951, -0.0607, -0.0285], + ..., + [-0.0313, -0.0587, 0.0583, ..., -0.0069, 0.0140, 0.0468], + [-0.0064, 0.0230, -0.0341, ..., -0.0303, -0.0616, -0.0045], + [-0.0694, -0.0103, -0.0153, ..., -0.0640, 0.0594, 0.0095]], + device='cuda:0'), grad: tensor([[ 0.0050, 0.0138, 0.0055, ..., 0.0005, 0.0019, 0.0068], + [ 0.0031, 0.0061, 0.0092, ..., 0.0013, 0.0037, 0.0081], + [-0.0137, -0.0119, -0.0061, ..., -0.0042, -0.0005, -0.0052], + ..., + [ 0.0020, 0.0044, 0.0066, ..., 0.0007, 0.0033, 0.0060], + [ 0.0014, -0.0209, 0.0073, ..., 0.0032, 0.0041, 0.0005], + [ 0.0035, -0.0034, -0.0184, ..., -0.0025, -0.0155, -0.0106]], + device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0202, -0.0025, 0.0109, 0.0147, -0.0061, 0.0017, 0.0054, 0.0288, + -0.0301, 0.0288], device='cuda:0'), grad: tensor([ 0.0333, 0.0316, -0.0235, -0.0114, -0.0118, -0.0146, 0.0587, 0.0231, + -0.0218, -0.0636], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 226.85, cls_loss 0.7760 cls_loss_mapping 0.0590 cls_loss_causal 0.7459 re_mapping 0.0201 re_causal 0.0548 /// teacc 98.30 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0251, 0.0476, -0.0450, ..., -0.0422, -0.0201, -0.0163], + [-0.0176, -0.0758, 0.0144, ..., -0.0303, -0.0069, -0.0138], + [ 0.0196, -0.0457, 0.0170, ..., 0.0964, -0.0603, -0.0279], + ..., + [-0.0313, -0.0596, 0.0585, ..., -0.0069, 0.0135, 0.0474], + [-0.0065, 0.0231, -0.0347, ..., -0.0316, -0.0625, -0.0052], + [-0.0697, -0.0103, -0.0145, ..., -0.0649, 0.0602, 0.0093]], + device='cuda:0'), grad: tensor([[ 0.0059, 0.0109, 0.0051, ..., 0.0052, 0.0026, 0.0074], + [-0.0068, -0.0184, -0.0068, ..., -0.0089, -0.0059, -0.0122], + [ 0.0118, 0.0185, 0.0103, ..., 0.0182, 0.0026, 0.0088], + ..., + [ 0.0022, 0.0050, -0.0025, ..., -0.0046, 0.0026, 0.0032], + [ 0.0063, 0.0069, 0.0036, ..., 0.0022, 0.0024, 0.0044], + [ 0.0028, 0.0049, 0.0042, ..., 0.0010, 0.0026, 0.0044]], + device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0199, -0.0020, 0.0115, 0.0146, -0.0067, 0.0021, 0.0058, 0.0285, + -0.0306, 0.0287], device='cuda:0'), grad: tensor([ 0.0439, -0.0538, 0.0682, -0.0832, -0.0420, 0.0338, -0.0243, 0.0117, + 0.0252, 0.0207], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 225.96, cls_loss 0.7505 cls_loss_mapping 0.0613 cls_loss_causal 0.7287 re_mapping 0.0204 re_causal 0.0558 /// teacc 98.29 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0256, 0.0481, -0.0453, ..., -0.0433, -0.0210, -0.0158], + [-0.0178, -0.0764, 0.0141, ..., -0.0292, -0.0081, -0.0142], + [ 0.0198, -0.0467, 0.0176, ..., 0.0975, -0.0606, -0.0277], + ..., + [-0.0302, -0.0599, 0.0587, ..., -0.0074, 0.0133, 0.0476], + [-0.0064, 0.0237, -0.0345, ..., -0.0325, -0.0631, -0.0050], + [-0.0710, -0.0105, -0.0142, ..., -0.0647, 0.0603, 0.0087]], + device='cuda:0'), grad: tensor([[ 0.0014, 0.0010, 0.0022, ..., 0.0022, 0.0006, 0.0048], + [-0.0055, 0.0041, -0.0007, ..., -0.0187, 0.0006, 0.0025], + [ 0.0115, 0.0041, 0.0142, ..., 0.0273, 0.0010, 0.0082], + ..., + [ 0.0055, 0.0052, -0.0065, ..., -0.0078, 0.0039, -0.0019], + [-0.0059, -0.0105, -0.0130, ..., -0.0025, 0.0009, -0.0136], + [-0.0202, -0.0170, -0.0079, ..., -0.0033, -0.0119, -0.0060]], + device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0200, -0.0019, 0.0112, 0.0149, -0.0066, 0.0017, 0.0060, 0.0286, + -0.0307, 0.0288], device='cuda:0'), grad: tensor([ 0.0067, -0.0114, 0.0673, 0.0040, -0.0054, 0.0349, -0.0075, -0.0028, + -0.0591, -0.0267], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 226.10, cls_loss 0.7531 cls_loss_mapping 0.0660 cls_loss_causal 0.7337 re_mapping 0.0193 re_causal 0.0538 /// teacc 97.94 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0268, 0.0480, -0.0456, ..., -0.0437, -0.0222, -0.0155], + [-0.0175, -0.0767, 0.0146, ..., -0.0285, -0.0072, -0.0141], + [ 0.0203, -0.0468, 0.0182, ..., 0.0986, -0.0605, -0.0281], + ..., + [-0.0304, -0.0605, 0.0588, ..., -0.0085, 0.0138, 0.0476], + [-0.0059, 0.0237, -0.0348, ..., -0.0332, -0.0647, -0.0046], + [-0.0715, -0.0102, -0.0139, ..., -0.0653, 0.0612, 0.0089]], + device='cuda:0'), grad: tensor([[-0.0074, -0.0075, 0.0035, ..., -0.0007, 0.0005, -0.0029], + [ 0.0029, 0.0027, -0.0041, ..., -0.0039, 0.0010, 0.0003], + [-0.0009, 0.0039, -0.0048, ..., -0.0022, 0.0013, -0.0007], + ..., + [ 0.0017, 0.0054, 0.0116, ..., 0.0016, 0.0068, 0.0082], + [-0.0019, -0.0264, -0.0046, ..., 0.0028, -0.0080, -0.0038], + [-0.0004, 0.0027, 0.0094, ..., 0.0012, 0.0148, 0.0046]], + device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0200, -0.0014, 0.0109, 0.0154, -0.0068, 0.0020, 0.0058, 0.0280, + -0.0310, 0.0290], device='cuda:0'), grad: tensor([-0.0142, 0.0098, -0.0050, 0.0230, -0.0037, 0.0177, -0.0055, 0.0414, + -0.0875, 0.0239], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 226.02, cls_loss 0.7611 cls_loss_mapping 0.0472 cls_loss_causal 0.7321 re_mapping 0.0196 re_causal 0.0564 /// teacc 98.09 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0273, 0.0486, -0.0464, ..., -0.0450, -0.0220, -0.0149], + [-0.0178, -0.0772, 0.0143, ..., -0.0303, -0.0069, -0.0145], + [ 0.0203, -0.0477, 0.0192, ..., 0.0998, -0.0613, -0.0279], + ..., + [-0.0316, -0.0610, 0.0585, ..., -0.0094, 0.0137, 0.0473], + [-0.0052, 0.0242, -0.0349, ..., -0.0324, -0.0653, -0.0050], + [-0.0728, -0.0108, -0.0141, ..., -0.0660, 0.0612, 0.0084]], + device='cuda:0'), grad: tensor([[-0.0004, 0.0045, 0.0025, ..., 0.0004, 0.0016, 0.0046], + [ 0.0010, 0.0023, -0.0093, ..., -0.0008, -0.0014, -0.0035], + [-0.0010, 0.0012, -0.0075, ..., -0.0074, 0.0016, -0.0063], + ..., + [ 0.0011, 0.0018, 0.0063, ..., 0.0027, 0.0045, 0.0052], + [ 0.0031, 0.0053, 0.0046, ..., 0.0021, 0.0022, 0.0060], + [ 0.0007, 0.0005, -0.0092, ..., -0.0002, -0.0055, -0.0070]], + device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0197, -0.0019, 0.0113, 0.0155, -0.0065, 0.0021, 0.0059, 0.0283, + -0.0306, 0.0281], device='cuda:0'), grad: tensor([ 0.0271, -0.0133, -0.0178, 0.0379, -0.0090, -0.0157, -0.0180, 0.0220, + 0.0319, -0.0452], device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 226.21, cls_loss 0.7660 cls_loss_mapping 0.0546 cls_loss_causal 0.7394 re_mapping 0.0182 re_causal 0.0532 /// teacc 98.14 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0278, 0.0492, -0.0471, ..., -0.0462, -0.0219, -0.0151], + [-0.0182, -0.0778, 0.0143, ..., -0.0298, -0.0066, -0.0151], + [ 0.0207, -0.0478, 0.0203, ..., 0.1012, -0.0612, -0.0273], + ..., + [-0.0319, -0.0620, 0.0582, ..., -0.0107, 0.0127, 0.0468], + [-0.0056, 0.0241, -0.0346, ..., -0.0330, -0.0641, -0.0045], + [-0.0732, -0.0103, -0.0145, ..., -0.0668, 0.0607, 0.0085]], + device='cuda:0'), grad: tensor([[ 3.1281e-03, 5.7869e-03, 3.7537e-03, ..., 1.6441e-03, + 5.7602e-04, 3.6697e-03], + [-1.0157e-03, -2.9774e-03, -4.6234e-03, ..., 1.1501e-03, + -1.7233e-03, -2.7447e-03], + [ 1.5268e-03, -5.2023e-04, -2.8210e-03, ..., -8.9779e-06, + 1.3733e-03, 9.2649e-04], + ..., + [ 1.8287e-04, 2.8825e-04, 5.6381e-03, ..., -5.9471e-03, + 4.7951e-03, 6.1646e-03], + [ 2.7943e-04, -4.4823e-04, 5.3406e-03, ..., 1.8396e-03, + 1.5297e-03, 2.4529e-03], + [-2.6608e-03, -4.2496e-03, -2.5845e-03, ..., 1.0767e-03, + -1.4782e-03, -1.0025e-02]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0199, -0.0024, 0.0114, 0.0161, -0.0064, 0.0024, 0.0059, 0.0277, + -0.0311, 0.0285], device='cuda:0'), grad: tensor([ 0.0184, -0.0186, -0.0033, 0.0324, -0.0059, -0.0339, 0.0113, 0.0009, + 0.0081, -0.0095], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 226.80, cls_loss 0.7589 cls_loss_mapping 0.0509 cls_loss_causal 0.7280 re_mapping 0.0189 re_causal 0.0525 /// teacc 98.09 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0279, 0.0497, -0.0484, ..., -0.0472, -0.0223, -0.0153], + [-0.0187, -0.0777, 0.0150, ..., -0.0301, -0.0060, -0.0150], + [ 0.0207, -0.0479, 0.0196, ..., 0.1019, -0.0618, -0.0277], + ..., + [-0.0326, -0.0628, 0.0586, ..., -0.0097, 0.0127, 0.0471], + [-0.0057, 0.0239, -0.0342, ..., -0.0346, -0.0652, -0.0049], + [-0.0742, -0.0106, -0.0141, ..., -0.0680, 0.0616, 0.0092]], + device='cuda:0'), grad: tensor([[-0.0072, -0.0156, -0.0039, ..., -0.0003, -0.0011, -0.0023], + [ 0.0013, 0.0038, -0.0072, ..., 0.0005, -0.0018, -0.0012], + [ 0.0094, 0.0058, 0.0049, ..., 0.0002, 0.0010, 0.0007], + ..., + [-0.0025, 0.0017, -0.0055, ..., 0.0003, -0.0071, -0.0043], + [-0.0094, -0.0025, 0.0014, ..., 0.0007, 0.0042, 0.0023], + [ 0.0029, 0.0041, 0.0080, ..., 0.0005, 0.0073, 0.0038]], + device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0196, -0.0017, 0.0108, 0.0160, -0.0064, 0.0028, 0.0054, 0.0278, + -0.0312, 0.0287], device='cuda:0'), grad: tensor([-0.0289, -0.0120, 0.0181, -0.0024, 0.0221, 0.0063, 0.0127, -0.0063, + -0.0359, 0.0263], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 226.81, cls_loss 0.7464 cls_loss_mapping 0.0552 cls_loss_causal 0.7116 re_mapping 0.0196 re_causal 0.0529 /// teacc 98.28 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0287, 0.0497, -0.0481, ..., -0.0479, -0.0228, -0.0142], + [-0.0199, -0.0794, 0.0149, ..., -0.0310, -0.0058, -0.0155], + [ 0.0205, -0.0482, 0.0203, ..., 0.1035, -0.0619, -0.0274], + ..., + [-0.0324, -0.0629, 0.0592, ..., -0.0101, 0.0131, 0.0478], + [-0.0056, 0.0236, -0.0342, ..., -0.0356, -0.0657, -0.0045], + [-0.0744, -0.0103, -0.0147, ..., -0.0679, 0.0627, 0.0088]], + device='cuda:0'), grad: tensor([[-4.2725e-03, -6.9771e-03, -1.4069e-02, ..., -6.8016e-03, + 1.1864e-03, -1.4038e-02], + [-1.3342e-03, 2.0809e-03, -2.8658e-04, ..., 2.2948e-04, + 1.3971e-03, 4.0779e-03], + [ 5.5045e-05, -9.9754e-04, -5.9605e-04, ..., -2.7809e-03, + 1.4544e-04, 3.5934e-03], + ..., + [ 2.5349e-03, 1.8930e-03, 1.2474e-03, ..., 2.3746e-03, + -1.9588e-03, -5.2601e-06], + [ 5.2261e-04, 4.0894e-03, 4.2152e-03, ..., 1.1415e-03, + 3.4046e-03, 5.3482e-03], + [-5.2118e-04, -2.2583e-03, -1.1082e-03, ..., 1.4591e-03, + 3.5596e-04, 3.5191e-03]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0199, -0.0024, 0.0113, 0.0158, -0.0070, 0.0030, 0.0059, 0.0285, + -0.0310, 0.0280], device='cuda:0'), grad: tensor([-0.0727, 0.0122, 0.0009, 0.0391, -0.0082, -0.0296, 0.0258, 0.0168, + 0.0174, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 226.73, cls_loss 0.7457 cls_loss_mapping 0.0515 cls_loss_causal 0.7120 re_mapping 0.0175 re_causal 0.0456 /// teacc 98.19 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0287, 0.0497, -0.0491, ..., -0.0489, -0.0240, -0.0145], + [-0.0210, -0.0798, 0.0149, ..., -0.0324, -0.0060, -0.0157], + [ 0.0209, -0.0484, 0.0202, ..., 0.1041, -0.0625, -0.0274], + ..., + [-0.0333, -0.0623, 0.0595, ..., -0.0095, 0.0129, 0.0480], + [-0.0055, 0.0233, -0.0341, ..., -0.0353, -0.0658, -0.0035], + [-0.0744, -0.0102, -0.0143, ..., -0.0693, 0.0625, 0.0085]], + device='cuda:0'), grad: tensor([[ 0.0014, -0.0034, -0.0029, ..., -0.0028, -0.0020, -0.0053], + [ 0.0011, 0.0036, 0.0028, ..., 0.0017, 0.0002, 0.0031], + [ 0.0027, 0.0049, 0.0053, ..., 0.0029, 0.0031, 0.0086], + ..., + [ 0.0003, -0.0041, -0.0181, ..., 0.0005, -0.0100, -0.0084], + [ 0.0008, 0.0025, 0.0014, ..., 0.0015, -0.0019, -0.0030], + [ 0.0008, 0.0026, 0.0093, ..., 0.0008, 0.0076, 0.0011]], + device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0202, -0.0026, 0.0112, 0.0163, -0.0072, 0.0021, 0.0059, 0.0281, + -0.0307, 0.0285], device='cuda:0'), grad: tensor([-0.0117, 0.0123, 0.0369, -0.0297, 0.0684, -0.0480, -0.0102, -0.0262, + -0.0082, 0.0165], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 225.97, cls_loss 0.6902 cls_loss_mapping 0.0514 cls_loss_causal 0.6573 re_mapping 0.0195 re_causal 0.0523 /// teacc 98.16 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0293, 0.0504, -0.0503, ..., -0.0499, -0.0247, -0.0140], + [-0.0211, -0.0802, 0.0145, ..., -0.0328, -0.0059, -0.0161], + [ 0.0219, -0.0481, 0.0204, ..., 0.1055, -0.0639, -0.0288], + ..., + [-0.0334, -0.0625, 0.0602, ..., -0.0090, 0.0123, 0.0484], + [-0.0056, 0.0232, -0.0346, ..., -0.0359, -0.0655, -0.0037], + [-0.0748, -0.0105, -0.0133, ..., -0.0707, 0.0631, 0.0092]], + device='cuda:0'), grad: tensor([[ 0.0011, 0.0036, 0.0053, ..., 0.0023, 0.0030, 0.0062], + [ 0.0016, 0.0066, 0.0076, ..., 0.0027, 0.0035, 0.0095], + [ 0.0016, 0.0046, 0.0015, ..., -0.0064, 0.0030, 0.0003], + ..., + [ 0.0014, -0.0005, 0.0013, ..., 0.0018, 0.0012, 0.0025], + [ 0.0071, 0.0090, 0.0041, ..., 0.0016, 0.0065, 0.0077], + [-0.0289, -0.0224, -0.0084, ..., -0.0144, -0.0145, 0.0036]], + device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0201, -0.0028, 0.0110, 0.0159, -0.0073, 0.0021, 0.0060, 0.0286, + -0.0307, 0.0288], device='cuda:0'), grad: tensor([ 2.4048e-02, 3.4088e-02, -3.3140e-05, 9.4833e-03, 6.9962e-03, + -1.8311e-02, -3.7689e-02, -5.8670e-03, 1.3138e-02, -2.5879e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 225.86, cls_loss 0.7214 cls_loss_mapping 0.0461 cls_loss_causal 0.6935 re_mapping 0.0186 re_causal 0.0519 /// teacc 98.21 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0306, 0.0503, -0.0504, ..., -0.0498, -0.0256, -0.0140], + [-0.0219, -0.0814, 0.0145, ..., -0.0335, -0.0066, -0.0173], + [ 0.0224, -0.0483, 0.0208, ..., 0.1061, -0.0640, -0.0289], + ..., + [-0.0333, -0.0628, 0.0605, ..., -0.0097, 0.0125, 0.0487], + [-0.0050, 0.0241, -0.0347, ..., -0.0356, -0.0670, -0.0043], + [-0.0741, -0.0093, -0.0141, ..., -0.0716, 0.0633, 0.0090]], + device='cuda:0'), grad: tensor([[ 1.1778e-03, 1.4963e-03, 3.1452e-03, ..., 1.8339e-03, + 2.3842e-03, 3.9482e-03], + [ 2.9068e-03, 5.6190e-03, 5.7945e-03, ..., 3.2368e-03, + 3.0174e-03, 2.2430e-03], + [-3.2024e-03, 9.9087e-04, 5.4741e-04, ..., -2.2850e-03, + -1.9388e-03, -5.2032e-03], + ..., + [ 5.1069e-04, 1.7576e-03, 6.8069e-05, ..., 6.3372e-04, + 3.7785e-03, 2.5673e-03], + [ 4.9706e-03, -1.4450e-02, -1.8005e-02, ..., 2.9049e-03, + -2.4490e-02, -1.9424e-02], + [ 7.6723e-04, 1.8036e-02, 1.4870e-02, ..., 5.6362e-04, + 1.6159e-02, 1.5350e-02]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0199, -0.0031, 0.0109, 0.0158, -0.0072, 0.0019, 0.0062, 0.0289, + -0.0300, 0.0284], device='cuda:0'), grad: tensor([-0.0058, 0.0438, -0.0158, -0.0144, 0.0043, -0.0064, -0.0295, 0.0184, + -0.0366, 0.0421], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 226.69, cls_loss 0.7411 cls_loss_mapping 0.0483 cls_loss_causal 0.7129 re_mapping 0.0180 re_causal 0.0495 /// teacc 98.16 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0310, 0.0501, -0.0499, ..., -0.0506, -0.0272, -0.0146], + [-0.0217, -0.0816, 0.0146, ..., -0.0330, -0.0072, -0.0179], + [ 0.0220, -0.0487, 0.0212, ..., 0.1068, -0.0649, -0.0283], + ..., + [-0.0343, -0.0643, 0.0606, ..., -0.0096, 0.0115, 0.0490], + [-0.0051, 0.0240, -0.0349, ..., -0.0355, -0.0668, -0.0047], + [-0.0749, -0.0093, -0.0142, ..., -0.0719, 0.0636, 0.0088]], + device='cuda:0'), grad: tensor([[ 0.0025, 0.0053, 0.0043, ..., 0.0034, 0.0010, 0.0108], + [ 0.0048, 0.0034, 0.0027, ..., 0.0014, 0.0055, 0.0065], + [-0.0001, -0.0005, 0.0021, ..., -0.0012, -0.0013, 0.0013], + ..., + [ 0.0033, 0.0106, 0.0201, ..., 0.0011, 0.0151, 0.0264], + [ 0.0027, 0.0082, 0.0041, ..., 0.0015, 0.0057, 0.0087], + [ 0.0004, -0.0020, -0.0155, ..., -0.0004, -0.0138, -0.0191]], + device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0195, -0.0028, 0.0103, 0.0158, -0.0068, 0.0021, 0.0064, 0.0288, + -0.0307, 0.0292], device='cuda:0'), grad: tensor([ 0.0277, 0.0210, 0.0114, -0.0831, 0.0266, -0.0152, -0.0339, 0.0458, + 0.0096, -0.0098], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 226.20, cls_loss 0.7163 cls_loss_mapping 0.0514 cls_loss_causal 0.6812 re_mapping 0.0180 re_causal 0.0481 /// teacc 98.02 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0306, 0.0512, -0.0502, ..., -0.0519, -0.0273, -0.0152], + [-0.0215, -0.0807, 0.0143, ..., -0.0330, -0.0078, -0.0183], + [ 0.0223, -0.0493, 0.0212, ..., 0.1071, -0.0642, -0.0292], + ..., + [-0.0348, -0.0652, 0.0608, ..., -0.0095, 0.0107, 0.0495], + [-0.0052, 0.0239, -0.0341, ..., -0.0362, -0.0672, -0.0046], + [-0.0761, -0.0099, -0.0147, ..., -0.0734, 0.0640, 0.0086]], + device='cuda:0'), grad: tensor([[ 2.3117e-03, 3.7060e-03, 4.7722e-03, ..., 1.2951e-03, + 2.0695e-03, 5.2338e-03], + [ 5.3902e-03, 1.1261e-02, 1.1787e-02, ..., 1.7376e-03, + 4.7989e-03, 1.6083e-02], + [-1.3924e-03, -3.0956e-03, -1.9479e-04, ..., 6.6996e-05, + -3.9291e-04, 3.7581e-05], + ..., + [-2.1687e-03, -9.8724e-03, -2.2369e-02, ..., -5.2757e-03, + -4.2229e-03, -1.6281e-02], + [ 4.4098e-03, 1.2207e-03, -5.2490e-03, ..., -3.0918e-03, + -1.1539e-03, -3.4027e-03], + [-2.2144e-03, -3.8910e-03, -1.3704e-03, ..., 9.3126e-04, + 4.0960e-04, -8.9359e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0193, -0.0024, 0.0103, 0.0159, -0.0066, 0.0025, 0.0066, 0.0288, + -0.0304, 0.0278], device='cuda:0'), grad: tensor([ 0.0186, 0.0479, -0.0007, -0.0197, -0.0134, 0.0319, 0.0232, -0.0732, + -0.0078, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 44---------------------------------------------------- +epoch 44, time 228.14, cls_loss 0.7127 cls_loss_mapping 0.0430 cls_loss_causal 0.6778 re_mapping 0.0178 re_causal 0.0492 /// teacc 98.38 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0312, 0.0511, -0.0508, ..., -0.0529, -0.0283, -0.0152], + [-0.0220, -0.0814, 0.0131, ..., -0.0345, -0.0080, -0.0193], + [ 0.0225, -0.0503, 0.0216, ..., 0.1077, -0.0628, -0.0286], + ..., + [-0.0346, -0.0650, 0.0608, ..., -0.0088, 0.0096, 0.0497], + [-0.0047, 0.0242, -0.0344, ..., -0.0365, -0.0674, -0.0048], + [-0.0767, -0.0099, -0.0136, ..., -0.0735, 0.0644, 0.0086]], + device='cuda:0'), grad: tensor([[-0.0047, -0.0221, -0.0025, ..., -0.0008, -0.0084, -0.0137], + [ 0.0015, 0.0028, 0.0068, ..., 0.0017, 0.0022, 0.0070], + [ 0.0020, 0.0053, 0.0077, ..., 0.0009, 0.0029, 0.0074], + ..., + [ 0.0013, 0.0016, -0.0097, ..., -0.0019, 0.0008, -0.0077], + [-0.0061, -0.0005, -0.0047, ..., -0.0008, 0.0022, 0.0010], + [-0.0066, -0.0191, -0.0196, ..., -0.0062, -0.0132, -0.0227]], + device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0193, -0.0026, 0.0107, 0.0160, -0.0064, 0.0023, 0.0064, 0.0288, + -0.0306, 0.0280], device='cuda:0'), grad: tensor([-0.0408, 0.0303, 0.0355, 0.0121, 0.0691, 0.0334, -0.0137, -0.0362, + -0.0171, -0.0724], device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 226.62, cls_loss 0.6903 cls_loss_mapping 0.0410 cls_loss_causal 0.6545 re_mapping 0.0183 re_causal 0.0497 /// teacc 97.94 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0327, 0.0505, -0.0506, ..., -0.0533, -0.0290, -0.0161], + [-0.0228, -0.0818, 0.0127, ..., -0.0359, -0.0079, -0.0190], + [ 0.0228, -0.0509, 0.0217, ..., 0.1090, -0.0627, -0.0287], + ..., + [-0.0348, -0.0657, 0.0612, ..., -0.0080, 0.0083, 0.0494], + [-0.0049, 0.0242, -0.0341, ..., -0.0377, -0.0680, -0.0048], + [-0.0771, -0.0090, -0.0134, ..., -0.0738, 0.0650, 0.0090]], + device='cuda:0'), grad: tensor([[ 0.0034, 0.0088, 0.0059, ..., 0.0006, 0.0020, 0.0054], + [-0.0065, -0.0091, -0.0167, ..., -0.0050, -0.0030, -0.0082], + [ 0.0050, 0.0100, 0.0090, ..., 0.0022, 0.0020, 0.0068], + ..., + [ 0.0017, 0.0038, 0.0025, ..., 0.0010, 0.0003, 0.0021], + [-0.0005, -0.0019, 0.0050, ..., 0.0012, 0.0016, 0.0054], + [-0.0016, -0.0036, -0.0035, ..., 0.0001, -0.0019, -0.0021]], + device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0189, -0.0023, 0.0105, 0.0164, -0.0066, 0.0016, 0.0068, 0.0289, + -0.0307, 0.0283], device='cuda:0'), grad: tensor([ 0.0335, -0.0731, 0.0464, 0.0208, -0.0122, 0.0131, -0.0537, 0.0199, + 0.0226, -0.0171], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 226.22, cls_loss 0.7119 cls_loss_mapping 0.0429 cls_loss_causal 0.6780 re_mapping 0.0170 re_causal 0.0471 /// teacc 98.32 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0334, 0.0508, -0.0505, ..., -0.0532, -0.0288, -0.0156], + [-0.0238, -0.0820, 0.0130, ..., -0.0359, -0.0084, -0.0191], + [ 0.0226, -0.0509, 0.0220, ..., 0.1103, -0.0637, -0.0294], + ..., + [-0.0357, -0.0667, 0.0611, ..., -0.0095, 0.0090, 0.0490], + [-0.0053, 0.0239, -0.0341, ..., -0.0390, -0.0695, -0.0043], + [-0.0765, -0.0085, -0.0132, ..., -0.0738, 0.0661, 0.0090]], + device='cuda:0'), grad: tensor([[ 9.2363e-04, 2.7191e-02, 2.9049e-03, ..., 4.1509e-04, + 1.2293e-03, 1.5961e-02], + [ 7.0143e-04, 3.0384e-03, 1.0624e-03, ..., 3.6836e-04, + 1.9369e-03, 2.5501e-03], + [ 2.3155e-03, 2.3440e-05, -2.1610e-03, ..., -3.0308e-03, + 9.1124e-04, 2.3212e-03], + ..., + [ 1.3561e-03, -4.2419e-03, -6.0577e-03, ..., -2.8515e-04, + -6.8970e-03, -5.4817e-03], + [ 7.7744e-03, 7.1106e-03, 7.5264e-03, ..., 9.4748e-04, + 1.8606e-03, 3.9864e-03], + [ 1.1377e-03, 3.1109e-03, 3.6869e-03, ..., 4.9353e-04, + 1.9169e-03, 2.7161e-03]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0189, -0.0016, 0.0103, 0.0163, -0.0070, 0.0014, 0.0064, 0.0280, + -0.0310, 0.0298], device='cuda:0'), grad: tensor([ 0.0452, -0.0008, -0.0014, -0.0222, -0.0181, -0.0230, -0.0165, -0.0309, + 0.0462, 0.0214], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 226.52, cls_loss 0.7337 cls_loss_mapping 0.0462 cls_loss_causal 0.6959 re_mapping 0.0171 re_causal 0.0481 /// teacc 98.07 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0340, 0.0508, -0.0508, ..., -0.0529, -0.0295, -0.0161], + [-0.0249, -0.0825, 0.0124, ..., -0.0377, -0.0087, -0.0194], + [ 0.0216, -0.0519, 0.0226, ..., 0.1110, -0.0632, -0.0294], + ..., + [-0.0364, -0.0674, 0.0612, ..., -0.0096, 0.0082, 0.0493], + [-0.0050, 0.0240, -0.0340, ..., -0.0400, -0.0695, -0.0041], + [-0.0760, -0.0085, -0.0134, ..., -0.0752, 0.0664, 0.0085]], + device='cuda:0'), grad: tensor([[-0.0059, -0.0140, 0.0016, ..., -0.0006, -0.0034, -0.0067], + [ 0.0005, -0.0111, 0.0025, ..., 0.0012, -0.0053, -0.0020], + [-0.0070, -0.0022, -0.0039, ..., -0.0023, 0.0011, -0.0010], + ..., + [ 0.0014, 0.0025, 0.0064, ..., -0.0014, 0.0027, 0.0043], + [ 0.0020, 0.0039, 0.0027, ..., 0.0013, 0.0019, 0.0030], + [-0.0020, 0.0027, -0.0096, ..., 0.0013, 0.0038, -0.0022]], + device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0183, -0.0016, 0.0100, 0.0163, -0.0069, 0.0014, 0.0067, 0.0278, + -0.0301, 0.0295], device='cuda:0'), grad: tensor([-0.0175, -0.0217, -0.0202, 0.0196, -0.0165, 0.0330, 0.0154, 0.0091, + 0.0074, -0.0086], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 226.65, cls_loss 0.6825 cls_loss_mapping 0.0446 cls_loss_causal 0.6499 re_mapping 0.0168 re_causal 0.0461 /// teacc 98.18 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0342, 0.0509, -0.0514, ..., -0.0539, -0.0295, -0.0166], + [-0.0254, -0.0831, 0.0124, ..., -0.0371, -0.0083, -0.0193], + [ 0.0208, -0.0516, 0.0231, ..., 0.1109, -0.0640, -0.0292], + ..., + [-0.0364, -0.0678, 0.0615, ..., -0.0102, 0.0087, 0.0502], + [-0.0041, 0.0246, -0.0345, ..., -0.0408, -0.0699, -0.0046], + [-0.0770, -0.0092, -0.0132, ..., -0.0755, 0.0669, 0.0087]], + device='cuda:0'), grad: tensor([[ 3.6025e-04, 2.9635e-04, -1.7214e-04, ..., 2.1000e-03, + 1.0920e-03, -5.8222e-04], + [-8.4043e-06, -2.1610e-03, -1.0977e-03, ..., 2.7008e-03, + -8.7881e-04, -3.0994e-03], + [ 6.6452e-03, 4.6921e-03, 1.0399e-02, ..., 1.4160e-02, + 8.6546e-04, 2.8248e-03], + ..., + [-6.0701e-04, -3.3169e-03, -9.5062e-03, ..., -4.2877e-03, + -3.3321e-03, -1.0727e-02], + [ 1.4389e-02, 1.9608e-02, 4.3488e-03, ..., 1.1040e-02, + 1.4181e-03, 6.7291e-03], + [ 1.9574e-04, -2.1667e-03, -1.0529e-02, ..., -2.4529e-03, + 3.9864e-03, 7.6408e-03]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0185, -0.0012, 0.0104, 0.0159, -0.0072, 0.0014, 0.0062, 0.0285, + -0.0302, 0.0293], device='cuda:0'), grad: tensor([ 0.0047, -0.0166, 0.0548, -0.0645, 0.0192, -0.0393, 0.0048, -0.0291, + 0.0589, 0.0071], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 49---------------------------------------------------- +epoch 49, time 228.72, cls_loss 0.6859 cls_loss_mapping 0.0447 cls_loss_causal 0.6509 re_mapping 0.0167 re_causal 0.0443 /// teacc 98.52 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0343, 0.0512, -0.0517, ..., -0.0546, -0.0299, -0.0168], + [-0.0258, -0.0839, 0.0119, ..., -0.0367, -0.0088, -0.0201], + [ 0.0210, -0.0516, 0.0231, ..., 0.1116, -0.0652, -0.0293], + ..., + [-0.0362, -0.0677, 0.0620, ..., -0.0101, 0.0082, 0.0508], + [-0.0041, 0.0249, -0.0342, ..., -0.0420, -0.0704, -0.0045], + [-0.0781, -0.0092, -0.0133, ..., -0.0762, 0.0671, 0.0083]], + device='cuda:0'), grad: tensor([[-2.7885e-03, -1.2123e-02, -6.3820e-03, ..., -1.7214e-04, + -4.4708e-03, -6.8626e-03], + [-2.4050e-05, -1.4353e-04, -7.6141e-03, ..., 6.3658e-05, + 1.7774e-04, -1.9610e-04], + [-1.2253e-02, 3.9444e-03, 1.5182e-03, ..., -7.9880e-03, + -8.9951e-03, -1.5160e-02], + ..., + [ 2.1133e-03, -8.6308e-04, 7.1487e-03, ..., 1.5533e-04, + 5.3482e-03, 6.3171e-03], + [-3.1319e-03, 5.1193e-03, 1.1154e-02, ..., 1.0282e-04, + 3.3894e-03, 3.9215e-03], + [ 2.2049e-03, 6.1646e-03, 3.2330e-03, ..., 1.2076e-04, + 1.0424e-03, 1.0948e-03]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0182, -0.0016, 0.0107, 0.0160, -0.0071, 0.0015, 0.0060, 0.0290, + -0.0300, 0.0288], device='cuda:0'), grad: tensor([-0.0488, -0.0372, -0.0086, 0.0036, 0.0135, 0.0052, -0.0160, 0.0095, + 0.0545, 0.0242], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 226.66, cls_loss 0.6712 cls_loss_mapping 0.0366 cls_loss_causal 0.6333 re_mapping 0.0163 re_causal 0.0423 /// teacc 98.38 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0352, 0.0514, -0.0517, ..., -0.0553, -0.0290, -0.0169], + [-0.0266, -0.0845, 0.0112, ..., -0.0391, -0.0103, -0.0207], + [ 0.0219, -0.0522, 0.0239, ..., 0.1133, -0.0650, -0.0294], + ..., + [-0.0354, -0.0679, 0.0616, ..., -0.0104, 0.0073, 0.0502], + [-0.0040, 0.0246, -0.0336, ..., -0.0421, -0.0699, -0.0036], + [-0.0789, -0.0087, -0.0132, ..., -0.0767, 0.0678, 0.0092]], + device='cuda:0'), grad: tensor([[-0.0092, -0.0251, 0.0036, ..., -0.0009, -0.0015, -0.0066], + [ 0.0010, 0.0020, 0.0049, ..., 0.0012, 0.0036, 0.0031], + [-0.0088, -0.0083, -0.0095, ..., -0.0008, -0.0038, -0.0105], + ..., + [ 0.0021, 0.0037, 0.0046, ..., 0.0012, 0.0040, 0.0048], + [ 0.0048, 0.0076, -0.0058, ..., -0.0013, -0.0107, 0.0028], + [ 0.0043, 0.0080, 0.0044, ..., 0.0008, 0.0040, 0.0056]], + device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0183, -0.0016, 0.0107, 0.0159, -0.0074, 0.0020, 0.0055, 0.0286, + -0.0300, 0.0295], device='cuda:0'), grad: tensor([-0.0065, 0.0223, -0.0439, -0.0007, -0.0422, 0.0358, -0.0065, 0.0236, + -0.0144, 0.0325], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 226.81, cls_loss 0.6661 cls_loss_mapping 0.0409 cls_loss_causal 0.6391 re_mapping 0.0170 re_causal 0.0459 /// teacc 98.28 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0354, 0.0520, -0.0511, ..., -0.0569, -0.0286, -0.0171], + [-0.0274, -0.0850, 0.0114, ..., -0.0382, -0.0110, -0.0211], + [ 0.0220, -0.0520, 0.0237, ..., 0.1137, -0.0657, -0.0293], + ..., + [-0.0355, -0.0682, 0.0617, ..., -0.0101, 0.0068, 0.0505], + [-0.0042, 0.0251, -0.0345, ..., -0.0425, -0.0710, -0.0042], + [-0.0792, -0.0096, -0.0128, ..., -0.0781, 0.0680, 0.0084]], + device='cuda:0'), grad: tensor([[ 0.0015, 0.0043, 0.0025, ..., -0.0006, 0.0015, 0.0070], + [ 0.0009, 0.0022, -0.0040, ..., -0.0005, 0.0007, -0.0053], + [ 0.0181, 0.0119, 0.0288, ..., 0.0286, 0.0011, 0.0114], + ..., + [-0.0117, -0.0109, -0.0216, ..., -0.0206, -0.0038, -0.0165], + [ 0.0019, 0.0020, -0.0062, ..., -0.0030, -0.0006, -0.0101], + [-0.0030, -0.0124, -0.0096, ..., 0.0005, -0.0062, -0.0095]], + device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0184, -0.0013, 0.0100, 0.0161, -0.0072, 0.0017, 0.0061, 0.0286, + -0.0302, 0.0293], device='cuda:0'), grad: tensor([ 0.0205, -0.0089, 0.0682, -0.0128, -0.0026, 0.0395, 0.0297, -0.0662, + -0.0129, -0.0546], device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 226.13, cls_loss 0.6679 cls_loss_mapping 0.0425 cls_loss_causal 0.6361 re_mapping 0.0166 re_causal 0.0432 /// teacc 98.33 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0345, 0.0528, -0.0510, ..., -0.0570, -0.0288, -0.0163], + [-0.0277, -0.0857, 0.0111, ..., -0.0365, -0.0121, -0.0232], + [ 0.0217, -0.0524, 0.0233, ..., 0.1127, -0.0659, -0.0289], + ..., + [-0.0358, -0.0685, 0.0620, ..., -0.0104, 0.0065, 0.0505], + [-0.0029, 0.0253, -0.0346, ..., -0.0426, -0.0711, -0.0040], + [-0.0796, -0.0101, -0.0129, ..., -0.0797, 0.0687, 0.0084]], + device='cuda:0'), grad: tensor([[-0.0056, 0.0018, 0.0075, ..., 0.0041, 0.0006, 0.0050], + [-0.0014, -0.0038, -0.0024, ..., 0.0006, -0.0013, -0.0036], + [ 0.0001, -0.0030, -0.0047, ..., -0.0030, -0.0012, -0.0044], + ..., + [ 0.0026, 0.0046, 0.0046, ..., 0.0009, 0.0017, 0.0046], + [-0.0213, -0.0342, -0.0049, ..., 0.0015, -0.0121, -0.0203], + [ 0.0064, 0.0106, 0.0029, ..., 0.0002, 0.0135, 0.0097]], + device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0183, -0.0023, 0.0101, 0.0159, -0.0068, 0.0017, 0.0064, 0.0283, + -0.0292, 0.0292], device='cuda:0'), grad: tensor([ 0.0288, -0.0157, -0.0209, 0.0493, -0.0677, 0.0279, 0.0364, 0.0258, + -0.0811, 0.0170], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 226.18, cls_loss 0.6747 cls_loss_mapping 0.0370 cls_loss_causal 0.6357 re_mapping 0.0157 re_causal 0.0415 /// teacc 98.37 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0356, 0.0524, -0.0521, ..., -0.0576, -0.0282, -0.0160], + [-0.0278, -0.0863, 0.0111, ..., -0.0371, -0.0116, -0.0228], + [ 0.0214, -0.0521, 0.0241, ..., 0.1141, -0.0658, -0.0285], + ..., + [-0.0359, -0.0681, 0.0623, ..., -0.0107, 0.0073, 0.0509], + [-0.0034, 0.0247, -0.0357, ..., -0.0439, -0.0721, -0.0046], + [-0.0812, -0.0105, -0.0121, ..., -0.0790, 0.0688, 0.0085]], + device='cuda:0'), grad: tensor([[ 0.0017, 0.0037, 0.0028, ..., 0.0006, 0.0007, 0.0056], + [-0.0056, -0.0110, -0.0077, ..., 0.0003, -0.0008, -0.0124], + [-0.0217, -0.0215, 0.0062, ..., 0.0018, 0.0008, -0.0045], + ..., + [ 0.0012, 0.0035, 0.0125, ..., 0.0004, 0.0028, 0.0152], + [-0.0026, -0.0064, -0.0058, ..., -0.0025, -0.0107, -0.0149], + [ 0.0001, -0.0068, -0.0189, ..., 0.0006, -0.0031, -0.0189]], + device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0179, -0.0020, 0.0104, 0.0155, -0.0066, 0.0011, 0.0067, 0.0290, + -0.0299, 0.0295], device='cuda:0'), grad: tensor([ 0.0172, -0.0447, -0.0288, 0.0693, 0.0080, 0.0182, 0.0126, 0.0401, + -0.0438, -0.0482], device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 226.04, cls_loss 0.6712 cls_loss_mapping 0.0365 cls_loss_causal 0.6395 re_mapping 0.0162 re_causal 0.0419 /// teacc 98.27 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0357, 0.0529, -0.0519, ..., -0.0586, -0.0280, -0.0163], + [-0.0287, -0.0859, 0.0111, ..., -0.0380, -0.0110, -0.0230], + [ 0.0219, -0.0519, 0.0240, ..., 0.1150, -0.0666, -0.0297], + ..., + [-0.0368, -0.0696, 0.0625, ..., -0.0103, 0.0069, 0.0509], + [-0.0031, 0.0253, -0.0354, ..., -0.0449, -0.0728, -0.0041], + [-0.0816, -0.0105, -0.0122, ..., -0.0804, 0.0694, 0.0090]], + device='cuda:0'), grad: tensor([[ 0.0014, 0.0058, 0.0044, ..., 0.0005, 0.0012, 0.0069], + [ 0.0015, 0.0018, 0.0027, ..., 0.0013, 0.0014, 0.0031], + [ 0.0046, 0.0071, 0.0045, ..., 0.0010, 0.0016, 0.0047], + ..., + [ 0.0027, 0.0010, -0.0019, ..., 0.0003, 0.0041, -0.0051], + [ 0.0047, 0.0099, 0.0045, ..., 0.0012, 0.0012, 0.0035], + [-0.0135, -0.0258, -0.0221, ..., -0.0064, -0.0102, -0.0215]], + device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0181, -0.0016, 0.0096, 0.0167, -0.0072, 0.0003, 0.0070, 0.0286, + -0.0295, 0.0294], device='cuda:0'), grad: tensor([ 7.0839e-03, 1.5373e-02, 2.4277e-02, 3.2471e-02, 8.2791e-05, + -1.8860e-02, -1.3485e-03, 5.7945e-03, 2.8000e-02, -9.2834e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 225.42, cls_loss 0.6823 cls_loss_mapping 0.0381 cls_loss_causal 0.6443 re_mapping 0.0159 re_causal 0.0399 /// teacc 98.28 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0357, 0.0528, -0.0521, ..., -0.0597, -0.0287, -0.0167], + [-0.0274, -0.0861, 0.0107, ..., -0.0383, -0.0112, -0.0238], + [ 0.0225, -0.0516, 0.0242, ..., 0.1153, -0.0669, -0.0303], + ..., + [-0.0370, -0.0702, 0.0628, ..., -0.0096, 0.0066, 0.0506], + [-0.0029, 0.0255, -0.0349, ..., -0.0441, -0.0729, -0.0033], + [-0.0822, -0.0113, -0.0121, ..., -0.0801, 0.0693, 0.0095]], + device='cuda:0'), grad: tensor([[ 2.0409e-04, -3.1605e-03, 1.6499e-03, ..., -9.0539e-05, + -6.5422e-04, 1.3218e-03], + [ 6.5880e-03, 2.1610e-03, 1.0094e-02, ..., 2.5120e-03, + 2.7580e-03, 1.5427e-02], + [-2.8305e-03, 1.8854e-03, 4.2572e-03, ..., 2.7561e-04, + 1.7586e-03, -3.4065e-03], + ..., + [ 7.9334e-05, -3.0689e-03, 3.6488e-03, ..., 1.1092e-04, + 2.0790e-04, -8.8644e-04], + [-1.0620e-02, -7.4310e-03, -1.2140e-03, ..., 2.2435e-04, + -4.5471e-03, -8.8806e-03], + [-6.2981e-03, -4.1428e-03, 3.7689e-03, ..., 1.0234e-04, + -6.5384e-03, 5.5809e-03]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0181, -0.0020, 0.0095, 0.0162, -0.0068, 0.0005, 0.0063, 0.0288, + -0.0296, 0.0303], device='cuda:0'), grad: tensor([ 0.0103, 0.0757, -0.0004, 0.0129, 0.0306, -0.0754, 0.0052, -0.0086, + -0.0562, 0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 225.88, cls_loss 0.6724 cls_loss_mapping 0.0355 cls_loss_causal 0.6382 re_mapping 0.0163 re_causal 0.0408 /// teacc 98.31 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0371, 0.0531, -0.0529, ..., -0.0613, -0.0293, -0.0169], + [-0.0270, -0.0863, 0.0108, ..., -0.0381, -0.0116, -0.0240], + [ 0.0225, -0.0521, 0.0243, ..., 0.1154, -0.0672, -0.0296], + ..., + [-0.0378, -0.0712, 0.0633, ..., -0.0091, 0.0062, 0.0510], + [-0.0022, 0.0255, -0.0347, ..., -0.0442, -0.0733, -0.0033], + [-0.0814, -0.0108, -0.0120, ..., -0.0796, 0.0698, 0.0094]], + device='cuda:0'), grad: tensor([[ 9.5177e-04, 9.1493e-05, 3.8471e-03, ..., 3.9506e-04, + 1.1787e-03, 6.5899e-04], + [ 5.2929e-04, 2.6035e-03, 5.1155e-03, ..., 6.1214e-05, + 1.2169e-03, 4.2381e-03], + [ 6.6376e-03, 3.5992e-03, -7.5722e-03, ..., 1.0948e-02, + -9.2363e-04, -2.7390e-03], + ..., + [-7.3013e-03, -4.8141e-03, -1.7853e-02, ..., -1.2039e-02, + -3.8280e-03, -1.0094e-02], + [ 5.0812e-03, 9.1324e-03, 8.5449e-03, ..., 1.3514e-03, + 3.6316e-03, 6.2141e-03], + [ 1.8322e-04, -2.2736e-03, 3.1257e-04, ..., -2.8801e-04, + -1.2903e-03, -2.1191e-03]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0180, -0.0023, 0.0095, 0.0163, -0.0071, 0.0009, 0.0065, 0.0284, + -0.0297, 0.0309], device='cuda:0'), grad: tensor([-0.0018, 0.0269, -0.0139, -0.0264, 0.0259, 0.0119, -0.0048, -0.0521, + 0.0383, -0.0040], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 57---------------------------------------------------- +epoch 57, time 228.64, cls_loss 0.6365 cls_loss_mapping 0.0406 cls_loss_causal 0.6092 re_mapping 0.0159 re_causal 0.0407 /// teacc 98.54 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0366, 0.0537, -0.0535, ..., -0.0621, -0.0298, -0.0171], + [-0.0270, -0.0868, 0.0106, ..., -0.0365, -0.0125, -0.0243], + [ 0.0224, -0.0525, 0.0248, ..., 0.1154, -0.0671, -0.0304], + ..., + [-0.0390, -0.0721, 0.0632, ..., -0.0090, 0.0069, 0.0510], + [-0.0019, 0.0263, -0.0352, ..., -0.0459, -0.0723, -0.0039], + [-0.0830, -0.0120, -0.0120, ..., -0.0809, 0.0695, 0.0090]], + device='cuda:0'), grad: tensor([[ 0.0022, 0.0106, 0.0074, ..., 0.0014, 0.0007, 0.0070], + [-0.0012, -0.0026, -0.0035, ..., 0.0005, 0.0001, -0.0026], + [ 0.0010, 0.0052, 0.0008, ..., -0.0038, 0.0003, 0.0005], + ..., + [ 0.0006, 0.0036, 0.0212, ..., 0.0010, 0.0072, 0.0034], + [ 0.0024, 0.0026, 0.0002, ..., 0.0013, 0.0007, -0.0016], + [-0.0068, -0.0139, -0.0307, ..., -0.0006, -0.0082, 0.0017]], + device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0177, -0.0020, 0.0099, 0.0166, -0.0066, 0.0014, 0.0065, 0.0281, + -0.0300, 0.0297], device='cuda:0'), grad: tensor([ 0.0427, -0.0184, 0.0140, -0.0380, -0.0161, -0.0168, 0.0395, 0.0373, + 0.0034, -0.0475], device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 226.68, cls_loss 0.6881 cls_loss_mapping 0.0377 cls_loss_causal 0.6485 re_mapping 0.0147 re_causal 0.0393 /// teacc 98.43 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0381, 0.0538, -0.0543, ..., -0.0638, -0.0292, -0.0168], + [-0.0273, -0.0867, 0.0103, ..., -0.0364, -0.0125, -0.0245], + [ 0.0219, -0.0531, 0.0251, ..., 0.1157, -0.0671, -0.0306], + ..., + [-0.0395, -0.0730, 0.0640, ..., -0.0080, 0.0062, 0.0512], + [-0.0004, 0.0268, -0.0363, ..., -0.0460, -0.0736, -0.0047], + [-0.0828, -0.0122, -0.0110, ..., -0.0809, 0.0702, 0.0097]], + device='cuda:0'), grad: tensor([[-0.0090, -0.0113, -0.0314, ..., -0.0004, 0.0005, -0.0241], + [-0.0005, -0.0023, -0.0052, ..., 0.0001, -0.0018, -0.0109], + [ 0.0026, 0.0032, 0.0050, ..., -0.0018, 0.0006, 0.0042], + ..., + [ 0.0015, 0.0016, -0.0014, ..., 0.0006, 0.0004, -0.0014], + [-0.0065, -0.0066, -0.0010, ..., -0.0013, -0.0033, -0.0034], + [ 0.0024, 0.0026, 0.0063, ..., 0.0002, 0.0015, 0.0053]], + device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0179, -0.0021, 0.0099, 0.0168, -0.0067, 0.0009, 0.0067, 0.0280, + -0.0305, 0.0306], device='cuda:0'), grad: tensor([-0.0433, -0.0425, 0.0190, 0.0196, 0.0495, 0.0256, -0.0205, 0.0062, + -0.0369, 0.0234], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 59---------------------------------------------------- +epoch 59, time 227.65, cls_loss 0.6835 cls_loss_mapping 0.0333 cls_loss_causal 0.6546 re_mapping 0.0155 re_causal 0.0405 /// teacc 98.64 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0387, 0.0539, -0.0538, ..., -0.0632, -0.0289, -0.0163], + [-0.0276, -0.0863, 0.0095, ..., -0.0380, -0.0122, -0.0255], + [ 0.0214, -0.0531, 0.0263, ..., 0.1166, -0.0668, -0.0296], + ..., + [-0.0392, -0.0728, 0.0641, ..., -0.0085, 0.0056, 0.0514], + [-0.0006, 0.0265, -0.0363, ..., -0.0457, -0.0740, -0.0050], + [-0.0831, -0.0124, -0.0107, ..., -0.0822, 0.0704, 0.0103]], + device='cuda:0'), grad: tensor([[ 5.5170e-04, -6.1417e-03, 1.6546e-03, ..., 4.1202e-06, + -1.7899e-02, -9.1782e-03], + [ 8.2433e-05, 2.8114e-03, 6.0234e-03, ..., 1.3962e-03, + 2.3251e-03, 6.1417e-03], + [ 1.9836e-03, 6.3248e-03, -2.4887e-02, ..., -1.9424e-02, + 1.3666e-03, -2.4170e-02], + ..., + [ 1.0853e-03, 6.2895e-04, 6.7482e-03, ..., 2.2106e-03, + 1.2779e-03, 2.0294e-03], + [-5.9128e-04, -4.1885e-03, -1.4580e-02, ..., -3.3975e-04, + 2.4772e-04, -9.6054e-03], + [ 7.6962e-04, 7.1869e-03, -6.0158e-03, ..., -3.9406e-03, + 9.0942e-03, 3.9291e-03]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0183, -0.0021, 0.0099, 0.0164, -0.0074, 0.0002, 0.0067, 0.0284, + -0.0301, 0.0311], device='cuda:0'), grad: tensor([-0.0379, 0.0271, 0.0182, 0.0088, 0.0208, -0.0104, 0.0229, 0.0105, + -0.0547, -0.0054], device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 225.52, cls_loss 0.6773 cls_loss_mapping 0.0337 cls_loss_causal 0.6352 re_mapping 0.0149 re_causal 0.0393 /// teacc 98.39 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0387, 0.0540, -0.0543, ..., -0.0629, -0.0274, -0.0164], + [-0.0274, -0.0867, 0.0099, ..., -0.0379, -0.0125, -0.0259], + [ 0.0219, -0.0536, 0.0269, ..., 0.1179, -0.0663, -0.0302], + ..., + [-0.0378, -0.0721, 0.0636, ..., -0.0088, 0.0059, 0.0513], + [-0.0014, 0.0266, -0.0355, ..., -0.0467, -0.0739, -0.0046], + [-0.0847, -0.0138, -0.0105, ..., -0.0818, 0.0702, 0.0103]], + device='cuda:0'), grad: tensor([[-1.2064e-03, -4.0932e-03, -3.9635e-03, ..., 2.2900e-04, + 3.2353e-04, -1.3771e-03], + [-1.7500e-03, -6.3629e-03, 1.8730e-03, ..., 8.3971e-04, + 5.0926e-04, -2.3613e-03], + [ 1.5621e-03, 6.2218e-03, 6.0310e-03, ..., 2.8439e-03, + 1.3304e-03, 9.0637e-03], + ..., + [ 6.6662e-04, 5.9929e-03, 1.1177e-02, ..., 9.4080e-04, + 4.0321e-03, 9.9716e-03], + [-3.3677e-05, 2.7809e-03, 5.3787e-03, ..., 9.4080e-04, + 2.2030e-03, 7.7171e-03], + [-2.3270e-03, -3.6716e-03, -7.0877e-03, ..., -7.7286e-03, + -9.7427e-03, -1.3519e-02]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0185, -0.0020, 0.0098, 0.0166, -0.0075, 0.0005, 0.0062, 0.0284, + -0.0300, 0.0308], device='cuda:0'), grad: tensor([-0.0248, -0.0038, 0.0413, -0.1012, 0.0020, 0.0222, 0.0177, 0.0454, + 0.0350, -0.0338], device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 225.67, cls_loss 0.6390 cls_loss_mapping 0.0318 cls_loss_causal 0.5948 re_mapping 0.0148 re_causal 0.0373 /// teacc 98.39 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0397, 0.0540, -0.0539, ..., -0.0628, -0.0281, -0.0165], + [-0.0284, -0.0870, 0.0099, ..., -0.0392, -0.0132, -0.0267], + [ 0.0222, -0.0534, 0.0264, ..., 0.1177, -0.0666, -0.0307], + ..., + [-0.0369, -0.0727, 0.0643, ..., -0.0082, 0.0062, 0.0517], + [-0.0015, 0.0266, -0.0364, ..., -0.0471, -0.0746, -0.0049], + [-0.0851, -0.0140, -0.0103, ..., -0.0817, 0.0715, 0.0107]], + device='cuda:0'), grad: tensor([[-0.0006, -0.0044, 0.0001, ..., -0.0006, 0.0014, -0.0010], + [ 0.0003, 0.0040, 0.0039, ..., 0.0002, 0.0038, 0.0087], + [-0.0050, 0.0009, -0.0023, ..., -0.0016, -0.0051, -0.0049], + ..., + [ 0.0005, -0.0022, -0.0056, ..., -0.0010, -0.0002, -0.0093], + [-0.0075, -0.0057, 0.0019, ..., 0.0007, -0.0014, 0.0046], + [ 0.0061, 0.0062, -0.0044, ..., 0.0010, -0.0039, -0.0090]], + device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0184, -0.0017, 0.0091, 0.0168, -0.0073, 0.0007, 0.0059, 0.0280, + -0.0302, 0.0314], device='cuda:0'), grad: tensor([ 0.0065, 0.0336, -0.0253, 0.0053, -0.0055, 0.0256, 0.0030, -0.0451, + 0.0044, -0.0024], device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 225.76, cls_loss 0.6761 cls_loss_mapping 0.0316 cls_loss_causal 0.6399 re_mapping 0.0149 re_causal 0.0416 /// teacc 98.50 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0403, 0.0543, -0.0545, ..., -0.0630, -0.0284, -0.0165], + [-0.0285, -0.0878, 0.0099, ..., -0.0394, -0.0134, -0.0280], + [ 0.0217, -0.0541, 0.0276, ..., 0.1177, -0.0665, -0.0295], + ..., + [-0.0377, -0.0728, 0.0646, ..., -0.0094, 0.0060, 0.0525], + [-0.0014, 0.0264, -0.0359, ..., -0.0470, -0.0744, -0.0044], + [-0.0848, -0.0130, -0.0110, ..., -0.0819, 0.0708, 0.0103]], + device='cuda:0'), grad: tensor([[-0.0071, -0.0098, -0.0007, ..., -0.0021, 0.0011, -0.0010], + [ 0.0004, 0.0011, 0.0030, ..., 0.0009, 0.0004, 0.0021], + [ 0.0031, 0.0066, -0.0009, ..., -0.0031, 0.0014, 0.0042], + ..., + [ 0.0039, 0.0045, 0.0056, ..., 0.0099, -0.0041, 0.0034], + [ 0.0014, -0.0012, -0.0116, ..., 0.0015, -0.0005, -0.0069], + [ 0.0015, -0.0005, 0.0063, ..., 0.0036, 0.0065, -0.0019]], + device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0179, -0.0022, 0.0099, 0.0170, -0.0077, -0.0004, 0.0065, 0.0281, + -0.0304, 0.0324], device='cuda:0'), grad: tensor([-0.0050, 0.0107, -0.0107, 0.0059, 0.0193, 0.0007, -0.0114, 0.0303, + -0.0332, -0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 226.05, cls_loss 0.6449 cls_loss_mapping 0.0339 cls_loss_causal 0.6116 re_mapping 0.0146 re_causal 0.0378 /// teacc 98.47 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0409, 0.0547, -0.0549, ..., -0.0644, -0.0292, -0.0165], + [-0.0289, -0.0877, 0.0113, ..., -0.0386, -0.0126, -0.0276], + [ 0.0224, -0.0551, 0.0281, ..., 0.1194, -0.0685, -0.0304], + ..., + [-0.0388, -0.0740, 0.0645, ..., -0.0100, 0.0064, 0.0520], + [-0.0015, 0.0263, -0.0360, ..., -0.0485, -0.0748, -0.0031], + [-0.0848, -0.0140, -0.0116, ..., -0.0835, 0.0711, 0.0102]], + device='cuda:0'), grad: tensor([[-2.5959e-03, -1.7500e-03, 3.0708e-03, ..., 1.1549e-03, + 1.0004e-03, 3.0079e-03], + [-4.2152e-04, -1.2302e-03, -7.0715e-04, ..., 4.6039e-04, + -6.9022e-05, -8.6212e-04], + [-2.4378e-04, 2.3632e-03, -1.0710e-03, ..., -5.7259e-03, + 6.4278e-04, 2.1515e-03], + ..., + [-7.6151e-04, -9.7322e-04, -4.8470e-04, ..., 1.7414e-03, + 2.2852e-04, -1.2531e-03], + [ 1.0399e-02, 1.1169e-02, 7.3166e-03, ..., 8.2684e-04, + 2.8210e-03, 5.2147e-03], + [-3.7050e-04, -2.7237e-03, -6.4583e-03, ..., 2.9430e-03, + -6.2866e-03, -4.8780e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0177, -0.0011, 0.0098, 0.0169, -0.0071, 0.0005, 0.0059, 0.0268, + -0.0301, 0.0320], device='cuda:0'), grad: tensor([ 0.0082, -0.0071, -0.0075, -0.0064, 0.0299, -0.0768, 0.0347, -0.0115, + 0.0434, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 226.03, cls_loss 0.6046 cls_loss_mapping 0.0328 cls_loss_causal 0.5681 re_mapping 0.0149 re_causal 0.0382 /// teacc 98.64 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0419, 0.0548, -0.0558, ..., -0.0650, -0.0296, -0.0168], + [-0.0293, -0.0875, 0.0108, ..., -0.0385, -0.0135, -0.0281], + [ 0.0230, -0.0554, 0.0283, ..., 0.1197, -0.0686, -0.0308], + ..., + [-0.0396, -0.0750, 0.0653, ..., -0.0105, 0.0059, 0.0524], + [-0.0020, 0.0264, -0.0366, ..., -0.0499, -0.0756, -0.0026], + [-0.0855, -0.0144, -0.0111, ..., -0.0821, 0.0721, 0.0101]], + device='cuda:0'), grad: tensor([[ 0.0033, 0.0100, -0.0013, ..., 0.0012, 0.0005, 0.0051], + [ 0.0011, -0.0029, -0.0038, ..., -0.0015, 0.0009, -0.0078], + [ 0.0039, 0.0047, -0.0005, ..., 0.0019, -0.0018, -0.0013], + ..., + [ 0.0034, 0.0022, 0.0008, ..., -0.0008, 0.0005, 0.0016], + [-0.0155, -0.0079, 0.0030, ..., -0.0075, 0.0014, 0.0035], + [-0.0006, -0.0067, -0.0061, ..., -0.0004, -0.0047, -0.0092]], + device='cuda:0') +Epoch 66, bias, value: tensor([ 1.7603e-02, -1.0856e-03, 9.8284e-03, 1.7770e-02, -7.4962e-03, + -4.2738e-05, 5.9178e-03, 2.6733e-02, -3.0441e-02, 3.2410e-02], + device='cuda:0'), grad: tensor([ 0.0049, -0.0088, -0.0015, 0.0350, 0.0335, -0.0113, 0.0010, 0.0099, + -0.0093, -0.0534], device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 228.32, cls_loss 0.6810 cls_loss_mapping 0.0285 cls_loss_causal 0.6407 re_mapping 0.0137 re_causal 0.0376 /// teacc 98.61 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0425, 0.0543, -0.0561, ..., -0.0635, -0.0309, -0.0166], + [-0.0304, -0.0895, 0.0117, ..., -0.0393, -0.0127, -0.0292], + [ 0.0232, -0.0552, 0.0285, ..., 0.1202, -0.0696, -0.0308], + ..., + [-0.0409, -0.0760, 0.0658, ..., -0.0113, 0.0064, 0.0526], + [-0.0021, 0.0261, -0.0368, ..., -0.0507, -0.0754, -0.0030], + [-0.0858, -0.0135, -0.0122, ..., -0.0831, 0.0720, 0.0098]], + device='cuda:0'), grad: tensor([[-7.4053e-04, -3.8319e-03, -1.2865e-03, ..., 4.4250e-04, + 6.5947e-04, -1.7242e-03], + [ 8.0347e-04, -5.0449e-04, -1.7667e-04, ..., 1.3387e-04, + -9.5415e-04, -1.0366e-03], + [ 2.2106e-03, 1.4791e-03, 9.0361e-04, ..., 1.4353e-03, + -1.0014e-03, -2.4632e-05], + ..., + [ 1.1091e-03, 4.9400e-04, -8.0185e-03, ..., -5.8975e-03, + -7.7188e-06, -3.6831e-03], + [-1.9455e-02, -2.4475e-02, 1.9026e-03, ..., -1.0939e-03, + 9.5940e-04, 2.7580e-03], + [ 1.1721e-03, 5.1765e-03, 4.5662e-03, ..., 2.9874e-04, + 1.5059e-03, 6.5155e-03]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0174, -0.0022, 0.0097, 0.0185, -0.0063, -0.0004, 0.0059, 0.0269, + -0.0303, 0.0320], device='cuda:0'), grad: tensor([-0.0111, -0.0181, -0.0002, -0.0252, 0.0039, 0.0138, 0.0347, 0.0017, + -0.0298, 0.0303], device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 229.06, cls_loss 0.6512 cls_loss_mapping 0.0361 cls_loss_causal 0.6123 re_mapping 0.0145 re_causal 0.0378 /// teacc 98.54 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0431, 0.0547, -0.0559, ..., -0.0643, -0.0313, -0.0151], + [-0.0295, -0.0897, 0.0120, ..., -0.0403, -0.0130, -0.0297], + [ 0.0227, -0.0565, 0.0286, ..., 0.1209, -0.0703, -0.0313], + ..., + [-0.0413, -0.0751, 0.0649, ..., -0.0115, 0.0061, 0.0534], + [-0.0012, 0.0275, -0.0375, ..., -0.0506, -0.0751, -0.0031], + [-0.0869, -0.0147, -0.0112, ..., -0.0832, 0.0723, 0.0091]], + device='cuda:0'), grad: tensor([[ 1.7471e-03, 6.8512e-03, 3.8452e-03, ..., 9.4366e-04, + -7.6914e-04, 1.7986e-03], + [ 1.3962e-03, 5.6190e-03, 9.4986e-03, ..., 1.4343e-03, + 3.7372e-05, 5.2452e-03], + [ 1.6556e-03, 6.2485e-03, 6.9160e-03, ..., 1.5593e-03, + 1.0103e-04, 6.7635e-03], + ..., + [-1.0052e-03, -9.1553e-03, -2.3651e-02, ..., 2.2087e-03, + -4.2796e-04, -8.5068e-03], + [-9.2173e-04, -1.1879e-02, -4.1656e-03, ..., -4.6015e-04, + 2.0313e-04, -6.1455e-03], + [-8.8692e-04, 2.1858e-03, 1.3153e-02, ..., -7.5378e-03, + 3.8481e-04, -7.3814e-03]], device='cuda:0') +Epoch 68, bias, value: tensor([ 1.8258e-02, -1.6316e-03, 8.8844e-03, 1.8368e-02, -6.5816e-03, + 2.6904e-05, 5.5888e-03, 2.7426e-02, -3.0487e-02, 3.1418e-02], + device='cuda:0'), grad: tensor([ 0.0107, 0.0483, 0.0381, 0.0153, -0.0068, 0.0077, 0.0025, -0.0738, + -0.0606, 0.0186], device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 227.05, cls_loss 0.6749 cls_loss_mapping 0.0267 cls_loss_causal 0.6382 re_mapping 0.0146 re_causal 0.0399 /// teacc 98.52 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0434, 0.0547, -0.0562, ..., -0.0643, -0.0321, -0.0156], + [-0.0294, -0.0905, 0.0117, ..., -0.0403, -0.0140, -0.0308], + [ 0.0224, -0.0568, 0.0295, ..., 0.1217, -0.0697, -0.0316], + ..., + [-0.0413, -0.0753, 0.0639, ..., -0.0135, 0.0052, 0.0528], + [-0.0015, 0.0273, -0.0373, ..., -0.0499, -0.0751, -0.0021], + [-0.0874, -0.0146, -0.0107, ..., -0.0834, 0.0726, 0.0106]], + device='cuda:0'), grad: tensor([[ 4.1533e-04, -3.5057e-03, -9.5987e-04, ..., -4.5824e-04, + 1.1992e-04, -2.1591e-03], + [ 1.3380e-03, -9.8133e-04, -8.0566e-03, ..., 1.9896e-04, + -2.6875e-03, -7.0686e-03], + [ 1.5593e-03, 3.2043e-03, 3.9368e-03, ..., 3.1137e-04, + 1.7214e-04, 1.6737e-03], + ..., + [ 2.9874e-04, 2.9049e-03, 2.3823e-03, ..., 1.8597e-05, + 4.8661e-04, 9.1505e-04], + [ 2.7418e-04, 1.8892e-03, -8.3466e-03, ..., 8.2731e-05, + -2.4676e-04, 1.5450e-03], + [ 6.5088e-04, 1.0368e-02, 1.1444e-02, ..., 5.6362e-04, + 1.9016e-03, 1.9388e-03]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0180, -0.0017, 0.0084, 0.0181, -0.0068, -0.0010, 0.0066, 0.0273, + -0.0298, 0.0320], device='cuda:0'), grad: tensor([-0.0122, -0.0420, 0.0236, -0.0612, 0.0295, 0.0200, 0.0297, 0.0178, + -0.0383, 0.0332], device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 225.80, cls_loss 0.6386 cls_loss_mapping 0.0252 cls_loss_causal 0.6106 re_mapping 0.0137 re_causal 0.0366 /// teacc 98.36 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0430, 0.0553, -0.0560, ..., -0.0643, -0.0325, -0.0150], + [-0.0282, -0.0903, 0.0120, ..., -0.0421, -0.0144, -0.0308], + [ 0.0227, -0.0566, 0.0297, ..., 0.1228, -0.0694, -0.0323], + ..., + [-0.0420, -0.0763, 0.0647, ..., -0.0142, 0.0051, 0.0535], + [-0.0018, 0.0267, -0.0382, ..., -0.0503, -0.0754, -0.0026], + [-0.0875, -0.0144, -0.0115, ..., -0.0836, 0.0716, 0.0092]], + device='cuda:0'), grad: tensor([[ 0.0007, 0.0026, -0.0212, ..., -0.0014, 0.0003, -0.0069], + [ 0.0005, 0.0020, 0.0031, ..., -0.0001, 0.0011, 0.0041], + [-0.0009, -0.0017, -0.0036, ..., -0.0045, 0.0003, -0.0009], + ..., + [ 0.0012, 0.0088, 0.0485, ..., 0.0053, 0.0031, 0.0248], + [ 0.0014, 0.0016, -0.0008, ..., 0.0005, -0.0003, -0.0032], + [-0.0016, -0.0050, -0.0044, ..., 0.0003, -0.0041, -0.0065]], + device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0189, -0.0017, 0.0090, 0.0180, -0.0070, -0.0009, 0.0069, 0.0272, + -0.0311, 0.0322], device='cuda:0'), grad: tensor([-0.0145, 0.0287, -0.0231, -0.0825, -0.0432, 0.0334, 0.0121, 0.1119, + -0.0116, -0.0112], device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 226.10, cls_loss 0.6301 cls_loss_mapping 0.0314 cls_loss_causal 0.5936 re_mapping 0.0139 re_causal 0.0362 /// teacc 98.55 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0422, 0.0552, -0.0561, ..., -0.0639, -0.0333, -0.0146], + [-0.0291, -0.0902, 0.0120, ..., -0.0432, -0.0140, -0.0311], + [ 0.0227, -0.0566, 0.0290, ..., 0.1229, -0.0697, -0.0326], + ..., + [-0.0426, -0.0773, 0.0651, ..., -0.0132, 0.0043, 0.0529], + [-0.0022, 0.0265, -0.0373, ..., -0.0499, -0.0743, -0.0012], + [-0.0877, -0.0132, -0.0112, ..., -0.0836, 0.0717, 0.0091]], + device='cuda:0'), grad: tensor([[ 1.8167e-03, 4.1656e-03, -1.2093e-03, ..., 1.5087e-03, + 6.4611e-04, 9.5272e-04], + [ 2.0421e-04, 1.7338e-03, 6.2714e-03, ..., 1.9026e-03, + 1.4658e-03, 5.7297e-03], + [ 8.5068e-04, 2.7370e-03, -2.0866e-03, ..., 4.4584e-04, + -2.1458e-03, -3.8452e-03], + ..., + [ 2.0182e-04, -1.3150e-05, -3.5477e-03, ..., -2.9678e-03, + 1.4725e-03, 1.0204e-03], + [ 1.9646e-03, 7.7105e-04, 2.9526e-03, ..., 7.0906e-04, + 1.2312e-03, 2.8706e-03], + [ 5.2023e-04, 1.9493e-03, -9.5940e-04, ..., 3.5191e-04, + -4.6754e-04, -1.4496e-03]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0194, -0.0008, 0.0083, 0.0176, -0.0073, -0.0001, 0.0064, 0.0267, + -0.0308, 0.0321], device='cuda:0'), grad: tensor([-0.0095, 0.0610, -0.0403, 0.0127, -0.0036, -0.0500, 0.0111, 0.0070, + 0.0179, -0.0063], device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 228.70, cls_loss 0.6502 cls_loss_mapping 0.0295 cls_loss_causal 0.6116 re_mapping 0.0135 re_causal 0.0351 /// teacc 98.47 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0422, 0.0558, -0.0568, ..., -0.0649, -0.0345, -0.0141], + [-0.0291, -0.0902, 0.0116, ..., -0.0436, -0.0147, -0.0316], + [ 0.0223, -0.0567, 0.0294, ..., 0.1230, -0.0699, -0.0327], + ..., + [-0.0418, -0.0773, 0.0650, ..., -0.0138, 0.0036, 0.0525], + [-0.0028, 0.0259, -0.0373, ..., -0.0493, -0.0740, -0.0016], + [-0.0890, -0.0137, -0.0109, ..., -0.0847, 0.0726, 0.0095]], + device='cuda:0'), grad: tensor([[ 7.6294e-05, -1.4048e-03, -2.6436e-03, ..., 3.4630e-05, + 3.7503e-04, -2.8825e-04], + [ 1.5318e-04, 2.4009e-04, -2.4757e-03, ..., 2.7969e-05, + -2.2030e-03, -3.8319e-03], + [ 3.8052e-04, -4.5624e-03, -7.1449e-03, ..., -5.3525e-05, + 2.4307e-04, -4.4518e-03], + ..., + [ 3.7527e-04, -6.9761e-04, -3.1796e-03, ..., -4.1342e-04, + 2.7370e-04, -2.5864e-03], + [ 6.3705e-04, 2.0409e-03, 2.5253e-03, ..., 8.4758e-05, + 3.2139e-04, 2.8610e-03], + [ 2.7061e-04, 2.7847e-03, 6.5269e-03, ..., 4.4775e-04, + 9.0408e-04, 6.6719e-03]], device='cuda:0') +Epoch 72, bias, value: tensor([ 1.9179e-02, -1.5928e-03, 8.4327e-03, 1.8024e-02, -7.2094e-03, + -4.1723e-05, 6.1879e-03, 2.6946e-02, -3.0677e-02, 3.2112e-02], + device='cuda:0'), grad: tensor([-0.0005, 0.0157, -0.0378, 0.0142, 0.0067, -0.0083, -0.0269, -0.0195, + 0.0235, 0.0328], device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 227.74, cls_loss 0.6488 cls_loss_mapping 0.0239 cls_loss_causal 0.6095 re_mapping 0.0135 re_causal 0.0367 /// teacc 98.34 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0427, 0.0560, -0.0574, ..., -0.0642, -0.0348, -0.0144], + [-0.0288, -0.0899, 0.0117, ..., -0.0431, -0.0152, -0.0310], + [ 0.0225, -0.0576, 0.0295, ..., 0.1242, -0.0709, -0.0320], + ..., + [-0.0418, -0.0777, 0.0651, ..., -0.0140, 0.0027, 0.0528], + [-0.0025, 0.0262, -0.0363, ..., -0.0499, -0.0728, -0.0016], + [-0.0896, -0.0139, -0.0109, ..., -0.0857, 0.0727, 0.0092]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0025, 0.0017, ..., 0.0014, 0.0002, 0.0011], + [ 0.0006, -0.0010, 0.0012, ..., 0.0043, 0.0008, 0.0010], + [-0.0030, -0.0008, 0.0049, ..., -0.0089, 0.0001, 0.0012], + ..., + [ 0.0001, 0.0013, 0.0058, ..., 0.0060, 0.0006, 0.0024], + [ 0.0025, 0.0042, 0.0043, ..., 0.0039, 0.0004, -0.0022], + [-0.0008, -0.0028, -0.0124, ..., -0.0046, -0.0021, -0.0072]], + device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0187, -0.0009, 0.0080, 0.0178, -0.0083, 0.0004, 0.0066, 0.0274, + -0.0305, 0.0321], device='cuda:0'), grad: tensor([ 0.0178, 0.0100, -0.0012, -0.0193, 0.0150, 0.0117, -0.0179, 0.0338, + 0.0056, -0.0555], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 72---------------------------------------------------- +epoch 72, time 228.19, cls_loss 0.6241 cls_loss_mapping 0.0260 cls_loss_causal 0.5784 re_mapping 0.0135 re_causal 0.0354 /// teacc 98.66 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0431, 0.0564, -0.0582, ..., -0.0640, -0.0344, -0.0141], + [-0.0290, -0.0905, 0.0113, ..., -0.0429, -0.0154, -0.0318], + [ 0.0218, -0.0572, 0.0295, ..., 0.1245, -0.0702, -0.0317], + ..., + [-0.0423, -0.0782, 0.0661, ..., -0.0133, 0.0022, 0.0527], + [-0.0016, 0.0258, -0.0370, ..., -0.0508, -0.0739, -0.0017], + [-0.0896, -0.0141, -0.0099, ..., -0.0861, 0.0735, 0.0098]], + device='cuda:0'), grad: tensor([[-2.5253e-03, -4.5662e-03, 1.1616e-03, ..., -1.3375e-04, + 1.3769e-04, 1.0548e-03], + [-1.1415e-03, -3.8528e-03, -5.3291e-03, ..., 4.0025e-05, + 1.0166e-03, -6.1378e-03], + [ 4.3845e-04, 1.6460e-03, 2.8076e-03, ..., -2.3937e-04, + 4.6849e-04, 2.3098e-03], + ..., + [ 2.0111e-04, -3.9768e-04, -8.7051e-03, ..., 2.3472e-04, + -3.2063e-03, -3.3913e-03], + [ 6.4182e-04, 2.6894e-03, 2.2259e-03, ..., 1.8156e-04, + 4.5419e-04, 2.2564e-03], + [ 3.6263e-04, 1.7529e-03, -1.8644e-03, ..., -1.4257e-03, + -1.8616e-03, 1.9398e-03]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0186, -0.0015, 0.0088, 0.0181, -0.0084, -0.0006, 0.0066, 0.0274, + -0.0302, 0.0323], device='cuda:0'), grad: tensor([-0.0002, -0.0573, 0.0219, -0.0025, -0.0076, 0.0192, 0.0302, -0.0414, + 0.0228, 0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 226.38, cls_loss 0.6455 cls_loss_mapping 0.0308 cls_loss_causal 0.6098 re_mapping 0.0133 re_causal 0.0347 /// teacc 98.52 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0430, 0.0572, -0.0586, ..., -0.0669, -0.0354, -0.0131], + [-0.0294, -0.0912, 0.0127, ..., -0.0434, -0.0148, -0.0325], + [ 0.0219, -0.0576, 0.0293, ..., 0.1255, -0.0705, -0.0321], + ..., + [-0.0423, -0.0785, 0.0651, ..., -0.0138, 0.0026, 0.0529], + [-0.0017, 0.0256, -0.0362, ..., -0.0504, -0.0747, -0.0022], + [-0.0907, -0.0149, -0.0102, ..., -0.0872, 0.0734, 0.0089]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0034, -0.0017, ..., 0.0003, 0.0003, -0.0017], + [ 0.0011, 0.0020, 0.0012, ..., 0.0007, -0.0012, 0.0007], + [-0.0081, -0.0052, -0.0018, ..., -0.0063, 0.0003, -0.0034], + ..., + [ 0.0065, 0.0070, 0.0086, ..., 0.0010, 0.0002, 0.0054], + [-0.0099, -0.0086, -0.0149, ..., -0.0026, 0.0003, -0.0011], + [ 0.0012, 0.0019, 0.0015, ..., 0.0001, -0.0010, -0.0021]], + device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0191, -0.0012, 0.0079, 0.0186, -0.0074, -0.0004, 0.0066, 0.0267, + -0.0300, 0.0315], device='cuda:0'), grad: tensor([-0.0087, 0.0097, -0.0126, 0.0735, -0.0047, -0.0582, 0.0034, 0.0317, + -0.0278, -0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 226.87, cls_loss 0.6441 cls_loss_mapping 0.0245 cls_loss_causal 0.6149 re_mapping 0.0136 re_causal 0.0348 /// teacc 98.45 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0426, 0.0569, -0.0588, ..., -0.0669, -0.0364, -0.0140], + [-0.0295, -0.0910, 0.0120, ..., -0.0443, -0.0142, -0.0320], + [ 0.0210, -0.0589, 0.0295, ..., 0.1259, -0.0716, -0.0322], + ..., + [-0.0432, -0.0796, 0.0665, ..., -0.0132, 0.0023, 0.0532], + [-0.0013, 0.0261, -0.0362, ..., -0.0515, -0.0758, -0.0019], + [-0.0915, -0.0146, -0.0107, ..., -0.0884, 0.0741, 0.0085]], + device='cuda:0'), grad: tensor([[ 8.6403e-04, -4.1962e-03, -2.8629e-03, ..., 2.5821e-04, + -1.0347e-04, -7.3242e-03], + [ 1.4400e-03, 4.0321e-03, 4.1313e-03, ..., 1.3466e-03, + 2.7132e-04, 3.8605e-03], + [-2.4662e-03, -7.5340e-04, -5.6343e-03, ..., -8.1635e-03, + 1.5414e-04, 5.0926e-03], + ..., + [ 7.4720e-04, 3.0632e-03, 6.1035e-03, ..., 2.8801e-03, + -2.8256e-06, 1.9283e-03], + [-8.2970e-04, -8.4019e-04, -3.7003e-03, ..., -1.5011e-03, + -1.7328e-03, -5.1498e-03], + [-3.0231e-04, -1.6365e-03, -3.7060e-03, ..., -6.8045e-04, + 9.1219e-04, -2.7771e-03]], device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0190, -0.0007, 0.0074, 0.0180, -0.0075, 0.0003, 0.0064, 0.0274, + -0.0303, 0.0313], device='cuda:0'), grad: tensor([-0.0284, 0.0282, 0.0019, -0.0122, 0.0155, -0.0085, 0.0227, 0.0247, + -0.0268, -0.0172], device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 226.61, cls_loss 0.6185 cls_loss_mapping 0.0259 cls_loss_causal 0.5847 re_mapping 0.0137 re_causal 0.0343 /// teacc 98.37 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0425, 0.0571, -0.0588, ..., -0.0676, -0.0375, -0.0139], + [-0.0308, -0.0924, 0.0119, ..., -0.0445, -0.0129, -0.0319], + [ 0.0218, -0.0589, 0.0306, ..., 0.1269, -0.0714, -0.0321], + ..., + [-0.0438, -0.0794, 0.0659, ..., -0.0141, 0.0018, 0.0533], + [-0.0022, 0.0252, -0.0368, ..., -0.0523, -0.0776, -0.0036], + [-0.0905, -0.0137, -0.0110, ..., -0.0899, 0.0751, 0.0087]], + device='cuda:0'), grad: tensor([[ 8.6725e-05, -2.5444e-03, -1.7643e-03, ..., 1.0878e-04, + -2.1589e-04, -1.5182e-03], + [ 1.2839e-04, -2.1362e-03, -1.1864e-03, ..., 3.9744e-04, + 6.8855e-04, -1.6394e-03], + [-6.3002e-05, 1.9102e-03, 1.8167e-03, ..., -6.3133e-04, + 1.0557e-03, 1.4496e-03], + ..., + [-7.3814e-04, 1.0700e-03, -6.8436e-03, ..., -2.3842e-03, + -5.2109e-03, 5.9986e-04], + [ 1.8999e-05, 1.5440e-03, 3.3340e-03, ..., 7.7009e-04, + 1.5802e-03, 1.4248e-03], + [ 3.4523e-04, -2.2259e-03, 4.0131e-03, ..., 2.5616e-03, + 7.0534e-03, 3.3340e-03]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0191, -0.0011, 0.0078, 0.0181, -0.0074, 0.0005, 0.0065, 0.0275, + -0.0307, 0.0312], device='cuda:0'), grad: tensor([-0.0161, -0.0129, 0.0145, -0.0179, 0.0057, 0.0129, 0.0135, -0.0168, + 0.0178, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 226.83, cls_loss 0.6265 cls_loss_mapping 0.0260 cls_loss_causal 0.5866 re_mapping 0.0132 re_causal 0.0336 /// teacc 98.50 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0432, 0.0571, -0.0586, ..., -0.0674, -0.0378, -0.0141], + [-0.0324, -0.0933, 0.0116, ..., -0.0448, -0.0135, -0.0330], + [ 0.0216, -0.0588, 0.0305, ..., 0.1266, -0.0706, -0.0324], + ..., + [-0.0426, -0.0796, 0.0664, ..., -0.0152, 0.0008, 0.0534], + [-0.0006, 0.0255, -0.0372, ..., -0.0519, -0.0789, -0.0041], + [-0.0913, -0.0138, -0.0112, ..., -0.0896, 0.0752, 0.0095]], + device='cuda:0'), grad: tensor([[ 1.6680e-03, 2.8172e-03, -1.1597e-03, ..., 1.8291e-03, + -6.5136e-04, 2.1076e-03], + [ 1.4629e-03, 2.2888e-03, 2.0580e-03, ..., 1.4553e-03, + 9.7752e-04, 2.8076e-03], + [ 1.0559e-02, 5.7564e-03, 1.3481e-02, ..., 7.0801e-03, + -5.3167e-04, 5.5885e-03], + ..., + [ 1.9054e-03, 2.7103e-03, 4.5853e-03, ..., 3.1090e-03, + 1.7862e-03, 2.6798e-03], + [ 1.0696e-02, 2.8267e-03, 2.6779e-03, ..., 2.7905e-03, + 2.1771e-05, -8.5068e-04], + [ 2.4166e-03, 6.3467e-04, -4.7722e-03, ..., -1.0080e-03, + -6.6528e-03, -6.2675e-03]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0191, -0.0016, 0.0083, 0.0190, -0.0076, -0.0005, 0.0075, 0.0271, + -0.0317, 0.0318], device='cuda:0'), grad: tensor([-0.0035, 0.0231, 0.0032, -0.0332, -0.0245, 0.0065, -0.0003, 0.0215, + 0.0081, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 226.71, cls_loss 0.6465 cls_loss_mapping 0.0254 cls_loss_causal 0.6097 re_mapping 0.0140 re_causal 0.0345 /// teacc 98.50 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0433, 0.0580, -0.0591, ..., -0.0670, -0.0383, -0.0147], + [-0.0326, -0.0935, 0.0108, ..., -0.0460, -0.0133, -0.0332], + [ 0.0212, -0.0594, 0.0303, ..., 0.1262, -0.0723, -0.0331], + ..., + [-0.0432, -0.0802, 0.0671, ..., -0.0152, 0.0005, 0.0542], + [ 0.0013, 0.0263, -0.0370, ..., -0.0520, -0.0800, -0.0046], + [-0.0931, -0.0145, -0.0103, ..., -0.0911, 0.0759, 0.0107]], + device='cuda:0'), grad: tensor([[-5.1003e-03, -1.2489e-02, 1.0700e-03, ..., 1.5187e-04, + -3.6240e-04, -3.2825e-03], + [ 8.6129e-05, -3.0231e-03, -6.0701e-04, ..., 5.9080e-04, + -1.6909e-03, -2.7046e-03], + [ 1.1168e-03, 1.6537e-03, 1.1091e-03, ..., -1.3723e-03, + 3.6550e-04, -4.8685e-04], + ..., + [-1.0529e-03, 1.3819e-03, -1.1070e-02, ..., -6.5994e-03, + -8.0299e-04, -2.2564e-03], + [ 1.7345e-04, 1.6069e-03, 1.0300e-03, ..., 1.9252e-04, + 2.6965e-04, 1.1902e-03], + [ 4.7445e-04, 1.7662e-03, 7.5951e-03, ..., 3.8357e-03, + 1.1005e-03, 3.9749e-03]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0192, -0.0016, 0.0085, 0.0176, -0.0086, -0.0011, 0.0080, 0.0278, + -0.0316, 0.0332], device='cuda:0'), grad: tensor([-0.0155, -0.0165, 0.0107, 0.0152, -0.0182, -0.0178, 0.0050, 0.0051, + 0.0135, 0.0186], device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 227.24, cls_loss 0.6620 cls_loss_mapping 0.0257 cls_loss_causal 0.6273 re_mapping 0.0139 re_causal 0.0358 /// teacc 98.46 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0442, 0.0581, -0.0596, ..., -0.0675, -0.0397, -0.0151], + [-0.0333, -0.0942, 0.0104, ..., -0.0463, -0.0143, -0.0343], + [ 0.0224, -0.0592, 0.0294, ..., 0.1273, -0.0734, -0.0337], + ..., + [-0.0435, -0.0804, 0.0677, ..., -0.0147, 0.0015, 0.0541], + [ 0.0016, 0.0268, -0.0365, ..., -0.0529, -0.0800, -0.0031], + [-0.0956, -0.0159, -0.0100, ..., -0.0912, 0.0754, 0.0107]], + device='cuda:0'), grad: tensor([[ 0.0008, 0.0031, 0.0031, ..., 0.0005, 0.0003, 0.0023], + [ 0.0014, 0.0027, 0.0062, ..., 0.0013, 0.0008, 0.0036], + [-0.0035, 0.0015, -0.0077, ..., -0.0091, 0.0003, -0.0007], + ..., + [ 0.0019, -0.0005, 0.0025, ..., 0.0018, 0.0007, -0.0022], + [ 0.0055, 0.0028, -0.0009, ..., 0.0007, 0.0025, -0.0005], + [ 0.0013, 0.0049, 0.0041, ..., 0.0003, 0.0067, 0.0094]], + device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0190, -0.0022, 0.0086, 0.0184, -0.0084, -0.0016, 0.0078, 0.0282, + -0.0306, 0.0319], device='cuda:0'), grad: tensor([ 0.0254, 0.0325, -0.0261, -0.0292, -0.0288, -0.0015, -0.0032, -0.0009, + 0.0060, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 226.69, cls_loss 0.6192 cls_loss_mapping 0.0255 cls_loss_causal 0.5819 re_mapping 0.0137 re_causal 0.0348 /// teacc 98.47 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0445, 0.0584, -0.0595, ..., -0.0673, -0.0402, -0.0152], + [-0.0327, -0.0945, 0.0115, ..., -0.0452, -0.0147, -0.0347], + [ 0.0223, -0.0598, 0.0299, ..., 0.1279, -0.0735, -0.0341], + ..., + [-0.0436, -0.0810, 0.0679, ..., -0.0147, 0.0014, 0.0544], + [ 0.0013, 0.0268, -0.0355, ..., -0.0546, -0.0797, -0.0020], + [-0.0957, -0.0144, -0.0109, ..., -0.0923, 0.0764, 0.0109]], + device='cuda:0'), grad: tensor([[-3.5954e-03, -1.1955e-02, -4.1008e-03, ..., 1.7071e-04, + -5.3978e-03, -1.2306e-02], + [-8.0395e-04, -2.3293e-04, -5.0592e-04, ..., -2.0981e-03, + 1.5039e-03, 3.9363e-04], + [-1.5373e-03, -1.6022e-02, -6.9733e-03, ..., -1.1002e-02, + 1.0633e-03, 3.3779e-03], + ..., + [-1.7128e-03, 1.9461e-05, 6.1264e-03, ..., 2.6798e-03, + 3.0994e-03, 5.2185e-03], + [ 3.3245e-03, 2.3560e-02, 1.1086e-02, ..., 7.6027e-03, + 5.8174e-03, 7.4005e-03], + [ 1.2493e-03, 5.4588e-03, -3.7956e-03, ..., 9.7752e-04, + -6.1188e-03, -8.1787e-03]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0189, -0.0020, 0.0085, 0.0186, -0.0084, -0.0020, 0.0074, 0.0279, + -0.0304, 0.0326], device='cuda:0'), grad: tensor([-0.0739, -0.0110, 0.0056, -0.0341, 0.0227, -0.0147, 0.0272, -0.0026, + 0.0657, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 222.21, cls_loss 0.6297 cls_loss_mapping 0.0220 cls_loss_causal 0.5907 re_mapping 0.0130 re_causal 0.0339 /// teacc 98.62 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0453, 0.0584, -0.0598, ..., -0.0680, -0.0401, -0.0152], + [-0.0326, -0.0947, 0.0115, ..., -0.0457, -0.0153, -0.0355], + [ 0.0211, -0.0606, 0.0298, ..., 0.1281, -0.0737, -0.0351], + ..., + [-0.0438, -0.0814, 0.0680, ..., -0.0135, 0.0018, 0.0548], + [ 0.0016, 0.0261, -0.0351, ..., -0.0548, -0.0800, -0.0026], + [-0.0955, -0.0136, -0.0103, ..., -0.0928, 0.0767, 0.0106]], + device='cuda:0'), grad: tensor([[-2.4719e-03, -6.4163e-03, 2.3708e-03, ..., 3.9506e-04, + 9.5272e-04, 1.2856e-03], + [ 8.0347e-04, 2.7008e-03, -9.0103e-03, ..., 3.2574e-05, + 2.9063e-04, 3.4070e-04], + [-5.4283e-03, -4.7417e-03, -3.9940e-03, ..., -4.4746e-03, + -6.0606e-04, -2.1420e-03], + ..., + [ 1.0071e-03, 3.9368e-03, 1.1971e-02, ..., 2.6321e-04, + 9.6655e-04, 3.0670e-03], + [ 4.6120e-03, -1.7977e-03, -5.4703e-03, ..., 2.0142e-03, + -2.1667e-03, -5.2032e-03], + [ 1.2703e-03, 7.1449e-03, 3.3703e-03, ..., 2.5702e-04, + 7.7200e-04, 3.5915e-03]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0192, -0.0019, 0.0075, 0.0186, -0.0085, -0.0014, 0.0075, 0.0275, + -0.0306, 0.0332], device='cuda:0'), grad: tensor([-0.0097, -0.0114, -0.0278, 0.0007, -0.0138, 0.0106, -0.0010, 0.0427, + -0.0208, 0.0304], device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 223.25, cls_loss 0.6008 cls_loss_mapping 0.0235 cls_loss_causal 0.5609 re_mapping 0.0133 re_causal 0.0328 /// teacc 98.66 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0469, 0.0587, -0.0603, ..., -0.0690, -0.0411, -0.0144], + [-0.0331, -0.0951, 0.0114, ..., -0.0463, -0.0153, -0.0369], + [ 0.0215, -0.0603, 0.0309, ..., 0.1289, -0.0725, -0.0339], + ..., + [-0.0436, -0.0820, 0.0677, ..., -0.0146, 0.0008, 0.0552], + [ 0.0028, 0.0271, -0.0360, ..., -0.0557, -0.0801, -0.0038], + [-0.0958, -0.0135, -0.0098, ..., -0.0932, 0.0770, 0.0105]], + device='cuda:0'), grad: tensor([[ 0.0070, 0.0075, 0.0009, ..., 0.0014, 0.0006, 0.0027], + [ 0.0016, 0.0018, 0.0010, ..., 0.0004, 0.0002, 0.0011], + [ 0.0057, 0.0094, -0.0002, ..., 0.0010, -0.0004, -0.0034], + ..., + [ 0.0015, 0.0010, 0.0035, ..., 0.0006, 0.0034, 0.0009], + [-0.0224, -0.0194, -0.0020, ..., -0.0008, 0.0017, -0.0040], + [-0.0032, -0.0072, -0.0029, ..., -0.0008, -0.0081, -0.0019]], + device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0196, -0.0026, 0.0082, 0.0184, -0.0087, -0.0002, 0.0066, 0.0270, + -0.0309, 0.0337], device='cuda:0'), grad: tensor([ 0.0210, 0.0114, 0.0205, -0.0307, 0.0148, 0.0039, 0.0218, 0.0108, + -0.0632, -0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 220.92, cls_loss 0.6119 cls_loss_mapping 0.0219 cls_loss_causal 0.5715 re_mapping 0.0125 re_causal 0.0333 /// teacc 98.54 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0477, 0.0584, -0.0603, ..., -0.0690, -0.0415, -0.0143], + [-0.0330, -0.0949, 0.0116, ..., -0.0464, -0.0139, -0.0359], + [ 0.0218, -0.0602, 0.0306, ..., 0.1298, -0.0740, -0.0351], + ..., + [-0.0437, -0.0824, 0.0679, ..., -0.0146, -0.0001, 0.0556], + [ 0.0025, 0.0269, -0.0366, ..., -0.0565, -0.0803, -0.0037], + [-0.0969, -0.0143, -0.0096, ..., -0.0938, 0.0775, 0.0101]], + device='cuda:0'), grad: tensor([[ 5.3215e-04, 2.6608e-03, 1.2550e-03, ..., 4.2200e-04, + 3.4142e-04, 2.2392e-03], + [-6.0987e-04, -5.4893e-03, -5.2910e-03, ..., 1.8048e-04, + -9.0837e-05, -4.3564e-03], + [ 6.7978e-03, 6.6566e-03, 2.2526e-03, ..., 3.0003e-03, + 3.9434e-04, 3.1471e-03], + ..., + [ 3.2520e-04, 1.7443e-03, -6.0707e-05, ..., -3.9749e-03, + 1.1435e-03, -1.5297e-03], + [ 9.5940e-04, 3.8433e-03, 3.3989e-03, ..., 5.1737e-04, + 8.3256e-04, 4.5700e-03], + [-1.0500e-03, -4.3449e-03, 4.4899e-03, ..., 2.1343e-03, + 3.6812e-03, 2.6302e-03]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0192, -0.0020, 0.0077, 0.0189, -0.0088, -0.0002, 0.0068, 0.0269, + -0.0314, 0.0339], device='cuda:0'), grad: tensor([ 0.0173, -0.0475, 0.0264, -0.0201, -0.0181, 0.0296, -0.0022, -0.0051, + 0.0288, -0.0089], device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 217.84, cls_loss 0.6318 cls_loss_mapping 0.0212 cls_loss_causal 0.5988 re_mapping 0.0128 re_causal 0.0339 /// teacc 98.32 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0469, 0.0587, -0.0585, ..., -0.0671, -0.0408, -0.0138], + [-0.0336, -0.0950, 0.0118, ..., -0.0464, -0.0138, -0.0361], + [ 0.0219, -0.0607, 0.0315, ..., 0.1310, -0.0733, -0.0351], + ..., + [-0.0445, -0.0833, 0.0679, ..., -0.0157, -0.0013, 0.0560], + [ 0.0029, 0.0277, -0.0373, ..., -0.0581, -0.0803, -0.0027], + [-0.0975, -0.0154, -0.0099, ..., -0.0932, 0.0776, 0.0104]], + device='cuda:0'), grad: tensor([[ 1.0614e-03, 1.2608e-03, 1.7762e-04, ..., 2.4748e-04, + 1.1368e-03, 8.1110e-04], + [ 1.7309e-03, 3.1071e-03, 1.4191e-03, ..., -4.7922e-05, + 1.8322e-04, 1.4515e-03], + [-2.3994e-03, -9.0408e-03, -7.1335e-03, ..., -1.0643e-03, + -3.8624e-03, -6.2637e-03], + ..., + [ 1.9264e-04, -1.3599e-03, -2.3460e-03, ..., 4.5490e-04, + 7.8869e-04, -1.6403e-03], + [-1.2878e-02, -4.6310e-03, 1.6737e-03, ..., 4.9162e-04, + 1.3695e-03, 1.9722e-03], + [-1.3809e-03, -2.7504e-03, -7.2289e-04, ..., 3.2759e-04, + -5.9471e-03, -2.7599e-03]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0200, -0.0021, 0.0073, 0.0187, -0.0083, -0.0019, 0.0070, 0.0271, + -0.0304, 0.0336], device='cuda:0'), grad: tensor([ 0.0004, 0.0193, -0.0638, 0.0115, 0.0252, 0.0511, -0.0014, -0.0193, + -0.0226, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 221.43, cls_loss 0.5719 cls_loss_mapping 0.0207 cls_loss_causal 0.5408 re_mapping 0.0127 re_causal 0.0324 /// teacc 98.59 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0476, 0.0590, -0.0596, ..., -0.0671, -0.0421, -0.0145], + [-0.0340, -0.0955, 0.0134, ..., -0.0452, -0.0147, -0.0369], + [ 0.0225, -0.0599, 0.0315, ..., 0.1301, -0.0735, -0.0351], + ..., + [-0.0448, -0.0834, 0.0683, ..., -0.0157, -0.0022, 0.0564], + [ 0.0033, 0.0279, -0.0378, ..., -0.0586, -0.0812, -0.0019], + [-0.0986, -0.0155, -0.0100, ..., -0.0948, 0.0775, 0.0100]], + device='cuda:0'), grad: tensor([[ 0.0006, 0.0027, 0.0021, ..., 0.0002, 0.0012, 0.0016], + [-0.0008, -0.0027, -0.0039, ..., -0.0002, -0.0007, -0.0028], + [-0.0024, -0.0005, -0.0033, ..., -0.0012, 0.0007, 0.0021], + ..., + [ 0.0023, 0.0021, 0.0047, ..., 0.0022, 0.0014, 0.0019], + [ 0.0031, 0.0011, -0.0041, ..., 0.0007, -0.0072, 0.0014], + [ 0.0011, 0.0040, 0.0068, ..., 0.0005, 0.0051, 0.0045]], + device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0195, -0.0024, 0.0074, 0.0186, -0.0079, -0.0024, 0.0076, 0.0273, + -0.0305, 0.0339], device='cuda:0'), grad: tensor([ 0.0108, -0.0264, 0.0012, -0.0044, -0.0210, 0.0086, -0.0034, 0.0113, + 0.0037, 0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 218.48, cls_loss 0.5577 cls_loss_mapping 0.0209 cls_loss_causal 0.5268 re_mapping 0.0122 re_causal 0.0310 /// teacc 98.53 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0483, 0.0588, -0.0598, ..., -0.0681, -0.0430, -0.0149], + [-0.0337, -0.0956, 0.0140, ..., -0.0458, -0.0154, -0.0376], + [ 0.0218, -0.0605, 0.0310, ..., 0.1307, -0.0749, -0.0351], + ..., + [-0.0452, -0.0840, 0.0682, ..., -0.0164, -0.0016, 0.0568], + [ 0.0025, 0.0275, -0.0382, ..., -0.0588, -0.0819, -0.0029], + [-0.0984, -0.0152, -0.0101, ..., -0.0958, 0.0778, 0.0098]], + device='cuda:0'), grad: tensor([[-7.5758e-05, -2.1725e-03, -7.8249e-04, ..., -2.7865e-05, + 5.4359e-05, 2.7657e-04], + [-4.2686e-03, -6.8169e-03, -2.8801e-03, ..., -4.8137e-04, + -1.8892e-03, -1.3590e-04], + [ 2.5768e-03, -1.7471e-03, -1.9226e-03, ..., 3.8481e-04, + 2.2161e-04, -1.6470e-03], + ..., + [ 7.9775e-04, 1.1883e-03, 8.6403e-04, ..., 1.3435e-04, + 5.9319e-04, 5.0879e-04], + [ 1.6870e-03, 1.8396e-03, 8.5926e-04, ..., 1.9777e-04, + 7.2718e-04, 3.7169e-04], + [-5.4646e-04, 9.5510e-04, -5.4026e-04, ..., -1.0931e-04, + -2.2831e-03, -1.4067e-03]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0192, -0.0027, 0.0071, 0.0194, -0.0074, -0.0021, 0.0082, 0.0272, + -0.0316, 0.0337], device='cuda:0'), grad: tensor([ 0.0046, -0.0372, -0.0102, -0.0008, 0.0141, 0.0157, 0.0186, 0.0077, + 0.0129, -0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 217.94, cls_loss 0.5955 cls_loss_mapping 0.0217 cls_loss_causal 0.5670 re_mapping 0.0130 re_causal 0.0334 /// teacc 98.58 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0487, 0.0591, -0.0604, ..., -0.0701, -0.0434, -0.0137], + [-0.0348, -0.0958, 0.0133, ..., -0.0462, -0.0161, -0.0386], + [ 0.0220, -0.0609, 0.0318, ..., 0.1316, -0.0762, -0.0359], + ..., + [-0.0458, -0.0846, 0.0681, ..., -0.0165, -0.0014, 0.0582], + [ 0.0020, 0.0276, -0.0377, ..., -0.0590, -0.0810, -0.0031], + [-0.0969, -0.0146, -0.0103, ..., -0.0965, 0.0779, 0.0097]], + device='cuda:0'), grad: tensor([[-7.4720e-04, -2.6531e-03, -5.1689e-03, ..., 6.6280e-05, + -1.4696e-03, -5.4550e-03], + [ 8.1348e-04, 8.9502e-04, -2.3422e-03, ..., 2.1115e-05, + -9.6142e-05, -1.2547e-05], + [ 8.0347e-04, 1.0872e-03, 7.5388e-04, ..., -1.5736e-03, + 5.3644e-04, 2.2831e-03], + ..., + [-2.5368e-03, -1.3933e-03, -3.9253e-03, ..., -9.1434e-05, + -5.6534e-03, -4.5052e-03], + [ 5.2643e-03, 2.8419e-03, 7.8917e-04, ..., 8.0109e-05, + 2.9659e-03, 4.4098e-03], + [ 7.8344e-04, 2.6398e-03, 4.5357e-03, ..., 3.8028e-05, + 1.1003e-04, 2.2011e-03]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0196, -0.0027, 0.0061, 0.0192, -0.0081, -0.0026, 0.0091, 0.0278, + -0.0316, 0.0343], device='cuda:0'), grad: tensor([-0.0494, -0.0034, 0.0159, 0.0427, -0.0093, -0.0122, -0.0107, -0.0032, + 0.0068, 0.0228], device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 218.37, cls_loss 0.6062 cls_loss_mapping 0.0226 cls_loss_causal 0.5756 re_mapping 0.0121 re_causal 0.0315 /// teacc 98.57 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0506, 0.0583, -0.0606, ..., -0.0706, -0.0437, -0.0143], + [-0.0362, -0.0960, 0.0132, ..., -0.0459, -0.0162, -0.0397], + [ 0.0234, -0.0604, 0.0313, ..., 0.1312, -0.0756, -0.0356], + ..., + [-0.0453, -0.0849, 0.0682, ..., -0.0165, -0.0019, 0.0585], + [ 0.0022, 0.0276, -0.0366, ..., -0.0583, -0.0815, -0.0026], + [-0.0980, -0.0156, -0.0098, ..., -0.0973, 0.0788, 0.0094]], + device='cuda:0'), grad: tensor([[ 1.0452e-03, 1.0405e-03, 2.2163e-03, ..., 1.7605e-03, + 9.6381e-05, 3.5458e-03], + [ 1.1644e-03, 1.5144e-03, 1.5080e-04, ..., 9.1600e-04, + 1.5020e-04, -3.2082e-03], + [ 3.2578e-03, 2.4109e-03, 2.2125e-03, ..., 1.4038e-03, + 1.0043e-04, 3.7689e-03], + ..., + [ 2.8591e-03, 1.5163e-03, 7.2384e-04, ..., -3.6316e-03, + 3.4595e-04, 4.0722e-04], + [ 4.0340e-04, -1.4524e-03, 2.1076e-04, ..., 1.2789e-03, + -5.9204e-03, -9.3307e-03], + [ 4.2763e-03, 6.5804e-03, 4.1695e-03, ..., 6.7186e-04, + 7.3471e-03, 1.5572e-02]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0192, -0.0033, 0.0066, 0.0203, -0.0086, -0.0030, 0.0088, 0.0277, + -0.0312, 0.0343], device='cuda:0'), grad: tensor([ 0.0190, -0.0142, 0.0236, -0.0778, -0.0082, -0.0243, 0.0280, 0.0089, + 0.0063, 0.0386], device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 217.87, cls_loss 0.5699 cls_loss_mapping 0.0159 cls_loss_causal 0.5314 re_mapping 0.0125 re_causal 0.0339 /// teacc 98.48 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0509, 0.0585, -0.0610, ..., -0.0713, -0.0438, -0.0137], + [-0.0365, -0.0951, 0.0132, ..., -0.0460, -0.0165, -0.0397], + [ 0.0234, -0.0597, 0.0323, ..., 0.1314, -0.0757, -0.0350], + ..., + [-0.0466, -0.0850, 0.0681, ..., -0.0162, -0.0022, 0.0576], + [ 0.0014, 0.0265, -0.0364, ..., -0.0586, -0.0813, -0.0021], + [-0.0985, -0.0158, -0.0097, ..., -0.0979, 0.0792, 0.0093]], + device='cuda:0'), grad: tensor([[ 2.7924e-03, 3.4657e-03, 3.3493e-03, ..., 1.5774e-03, + 5.8031e-04, 2.4014e-03], + [ 4.2844e-04, 1.2188e-03, -4.2114e-03, ..., 9.8991e-04, + -3.0708e-04, -3.0651e-03], + [ 1.4372e-03, 2.1286e-03, 3.9711e-03, ..., 5.5838e-04, + 1.0157e-03, 3.8128e-03], + ..., + [ 5.6267e-04, 1.1835e-03, 4.9019e-03, ..., 2.8038e-04, + 1.9407e-03, 3.7556e-03], + [-7.1373e-03, -8.4763e-03, 2.2089e-04, ..., -4.6196e-03, + 7.5054e-04, -1.0356e-05], + [ 5.1384e-03, 5.5771e-03, 8.1711e-03, ..., -7.3290e-04, + 1.4755e-02, 1.6289e-03]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0191, -0.0034, 0.0070, 0.0192, -0.0086, -0.0024, 0.0089, 0.0274, + -0.0311, 0.0349], device='cuda:0'), grad: tensor([ 0.0249, -0.0313, 0.0262, -0.0347, 0.0185, -0.0081, -0.0142, 0.0321, + -0.0252, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 218.39, cls_loss 0.6179 cls_loss_mapping 0.0240 cls_loss_causal 0.5852 re_mapping 0.0111 re_causal 0.0278 /// teacc 98.66 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0516, 0.0585, -0.0614, ..., -0.0715, -0.0455, -0.0130], + [-0.0378, -0.0958, 0.0123, ..., -0.0472, -0.0172, -0.0399], + [ 0.0227, -0.0613, 0.0319, ..., 0.1306, -0.0753, -0.0353], + ..., + [-0.0477, -0.0858, 0.0689, ..., -0.0157, -0.0027, 0.0581], + [ 0.0027, 0.0271, -0.0369, ..., -0.0582, -0.0820, -0.0023], + [-0.0982, -0.0155, -0.0100, ..., -0.0982, 0.0788, 0.0083]], + device='cuda:0'), grad: tensor([[ 0.0022, 0.0032, 0.0009, ..., 0.0002, 0.0003, 0.0011], + [-0.0017, -0.0032, 0.0002, ..., 0.0002, 0.0011, -0.0037], + [-0.0008, -0.0022, -0.0025, ..., -0.0020, 0.0002, -0.0014], + ..., + [-0.0002, -0.0012, -0.0062, ..., 0.0002, -0.0011, -0.0044], + [ 0.0043, 0.0071, 0.0018, ..., 0.0004, 0.0005, 0.0016], + [ 0.0005, -0.0002, 0.0003, ..., 0.0017, -0.0022, 0.0012]], + device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0191, -0.0036, 0.0059, 0.0197, -0.0078, -0.0019, 0.0084, 0.0274, + -0.0313, 0.0350], device='cuda:0'), grad: tensor([ 0.0135, -0.0126, -0.0254, 0.0066, 0.0076, 0.0066, 0.0107, -0.0440, + 0.0241, 0.0129], device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 218.37, cls_loss 0.6074 cls_loss_mapping 0.0247 cls_loss_causal 0.5776 re_mapping 0.0112 re_causal 0.0298 /// teacc 98.50 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0521, 0.0584, -0.0617, ..., -0.0708, -0.0463, -0.0134], + [-0.0376, -0.0964, 0.0119, ..., -0.0463, -0.0173, -0.0404], + [ 0.0231, -0.0614, 0.0324, ..., 0.1306, -0.0750, -0.0343], + ..., + [-0.0484, -0.0857, 0.0694, ..., -0.0151, -0.0026, 0.0585], + [ 0.0025, 0.0270, -0.0365, ..., -0.0577, -0.0830, -0.0026], + [-0.0993, -0.0157, -0.0099, ..., -0.1001, 0.0796, 0.0081]], + device='cuda:0'), grad: tensor([[ 2.7394e-04, 1.0223e-03, 1.4887e-03, ..., 2.7919e-04, + 3.3879e-04, 1.6489e-03], + [ 7.9632e-04, -2.9316e-03, -3.3498e-04, ..., 1.9550e-03, + -1.3418e-03, -2.9812e-03], + [-6.0987e-04, 2.6155e-04, -2.3994e-03, ..., -5.3864e-03, + -9.9945e-04, -3.2845e-03], + ..., + [ 6.2609e-04, 1.8969e-03, 8.8120e-03, ..., 4.4060e-03, + 1.5354e-03, 5.0926e-03], + [-3.0479e-03, -1.3672e-02, -3.7861e-03, ..., -8.7452e-04, + -6.0797e-05, -2.0084e-03], + [ 4.8590e-04, 1.2054e-03, -6.3896e-03, ..., -5.1460e-03, + 2.8954e-03, -1.9580e-05]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0184, -0.0036, 0.0068, 0.0187, -0.0082, -0.0012, 0.0090, 0.0272, + -0.0313, 0.0353], device='cuda:0'), grad: tensor([ 0.0197, -0.0453, -0.0133, 0.0100, -0.0066, 0.0420, 0.0077, 0.0475, + -0.0596, -0.0020], device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 217.84, cls_loss 0.6034 cls_loss_mapping 0.0231 cls_loss_causal 0.5750 re_mapping 0.0115 re_causal 0.0306 /// teacc 98.39 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0521, 0.0585, -0.0617, ..., -0.0716, -0.0467, -0.0131], + [-0.0372, -0.0971, 0.0120, ..., -0.0445, -0.0160, -0.0408], + [ 0.0232, -0.0619, 0.0323, ..., 0.1306, -0.0744, -0.0337], + ..., + [-0.0495, -0.0862, 0.0696, ..., -0.0144, -0.0032, 0.0576], + [ 0.0036, 0.0278, -0.0359, ..., -0.0585, -0.0838, -0.0019], + [-0.1012, -0.0160, -0.0092, ..., -0.0999, 0.0799, 0.0089]], + device='cuda:0'), grad: tensor([[ 9.7322e-04, 1.2383e-02, 1.5974e-03, ..., 3.6526e-04, + 5.0157e-05, 8.0185e-03], + [ 5.0449e-04, 1.1053e-03, 1.2493e-03, ..., 2.8968e-04, + 1.0774e-05, 1.2159e-03], + [ 1.2695e-02, 5.4626e-03, 6.4354e-03, ..., 7.0343e-03, + 7.0274e-05, 9.1743e-04], + ..., + [-6.8855e-03, -3.0575e-03, -2.0187e-02, ..., -4.8065e-03, + 5.9795e-04, -2.5177e-03], + [-3.7174e-03, -7.2937e-03, -7.4196e-04, ..., -2.8038e-03, + 1.0481e-03, -3.8700e-03], + [ 3.3302e-03, 3.5076e-03, -1.1854e-03, ..., 5.7793e-04, + 4.1056e-04, 3.7823e-03]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0184, -0.0036, 0.0071, 0.0187, -0.0085, -0.0022, 0.0088, 0.0263, + -0.0293, 0.0351], device='cuda:0'), grad: tensor([ 0.0342, 0.0137, 0.0244, 0.0201, 0.0109, -0.0130, -0.0101, -0.0804, + 0.0035, -0.0032], device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 217.99, cls_loss 0.6269 cls_loss_mapping 0.0210 cls_loss_causal 0.5941 re_mapping 0.0131 re_causal 0.0340 /// teacc 98.59 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0510, 0.0591, -0.0616, ..., -0.0719, -0.0470, -0.0129], + [-0.0380, -0.0981, 0.0117, ..., -0.0456, -0.0160, -0.0419], + [ 0.0215, -0.0628, 0.0334, ..., 0.1312, -0.0747, -0.0342], + ..., + [-0.0500, -0.0871, 0.0700, ..., -0.0150, -0.0032, 0.0580], + [ 0.0040, 0.0277, -0.0360, ..., -0.0588, -0.0834, -0.0030], + [-0.1012, -0.0172, -0.0098, ..., -0.0996, 0.0801, 0.0098]], + device='cuda:0'), grad: tensor([[ 7.7200e-04, -3.9520e-03, 9.6738e-05, ..., -9.6411e-06, + 2.6011e-04, -3.5458e-03], + [ 4.4250e-04, 2.7046e-03, -2.2221e-04, ..., 1.4268e-06, + 7.5579e-04, 2.3537e-03], + [ 1.1654e-03, 2.1820e-03, 5.7507e-04, ..., -1.1660e-05, + 1.9598e-04, 2.6474e-03], + ..., + [ 7.8583e-04, -3.6144e-03, -1.5378e-04, ..., 3.3733e-06, + -3.7253e-05, -2.4910e-03], + [ 9.2173e-04, -2.5058e-04, 1.6022e-04, ..., 3.5614e-06, + 2.9540e-04, 7.4148e-04], + [ 1.1396e-03, 2.6512e-03, 8.2397e-04, ..., 7.5102e-06, + 4.7421e-04, 4.0741e-03]], device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0186, -0.0042, 0.0065, 0.0196, -0.0078, -0.0014, 0.0078, 0.0264, + -0.0297, 0.0352], device='cuda:0'), grad: tensor([-0.0229, 0.0036, 0.0215, -0.0137, -0.0076, -0.0090, 0.0301, -0.0323, + 0.0002, 0.0302], device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 218.55, cls_loss 0.5965 cls_loss_mapping 0.0220 cls_loss_causal 0.5625 re_mapping 0.0112 re_causal 0.0280 /// teacc 98.52 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0518, 0.0591, -0.0617, ..., -0.0731, -0.0458, -0.0125], + [-0.0370, -0.0976, 0.0114, ..., -0.0455, -0.0173, -0.0430], + [ 0.0217, -0.0630, 0.0336, ..., 0.1314, -0.0745, -0.0345], + ..., + [-0.0510, -0.0878, 0.0696, ..., -0.0160, -0.0031, 0.0573], + [ 0.0044, 0.0276, -0.0365, ..., -0.0588, -0.0839, -0.0030], + [-0.1019, -0.0168, -0.0094, ..., -0.1000, 0.0802, 0.0111]], + device='cuda:0'), grad: tensor([[ 5.8746e-04, 2.1458e-03, 1.2760e-03, ..., 4.6635e-04, + 1.7428e-04, 1.7309e-03], + [-2.6393e-04, -2.9106e-03, -9.8991e-04, ..., 9.3412e-04, + 7.2718e-05, -7.4272e-03], + [ 1.8940e-03, 2.4681e-03, 3.0861e-03, ..., 3.3531e-03, + 1.0090e-03, 7.2327e-03], + ..., + [ 2.8706e-04, -1.8082e-03, -3.9635e-03, ..., -9.4652e-04, + 3.5214e-04, 2.0657e-03], + [-2.4014e-03, 8.6689e-04, 1.2856e-03, ..., -9.4748e-04, + 2.7204e-04, -9.8648e-03], + [ 1.7214e-03, 7.1259e-03, 8.9216e-04, ..., 9.4366e-04, + 7.0858e-04, 3.2921e-03]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0189, -0.0045, 0.0063, 0.0197, -0.0076, -0.0015, 0.0069, 0.0261, + -0.0293, 0.0358], device='cuda:0'), grad: tensor([ 0.0154, -0.0405, 0.0322, -0.0339, -0.0067, 0.0135, 0.0155, -0.0059, + -0.0208, 0.0312], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 94---------------------------------------------------- +epoch 94, time 218.48, cls_loss 0.6129 cls_loss_mapping 0.0261 cls_loss_causal 0.5824 re_mapping 0.0107 re_causal 0.0273 /// teacc 98.68 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0527, 0.0586, -0.0617, ..., -0.0730, -0.0466, -0.0121], + [-0.0373, -0.0981, 0.0121, ..., -0.0456, -0.0178, -0.0424], + [ 0.0225, -0.0624, 0.0339, ..., 0.1318, -0.0747, -0.0352], + ..., + [-0.0511, -0.0879, 0.0697, ..., -0.0159, -0.0032, 0.0577], + [ 0.0050, 0.0282, -0.0362, ..., -0.0571, -0.0851, -0.0026], + [-0.1013, -0.0163, -0.0095, ..., -0.1014, 0.0809, 0.0112]], + device='cuda:0'), grad: tensor([[ 3.0923e-04, -1.3704e-03, 1.8044e-03, ..., -2.5959e-03, + 1.0090e-03, 2.4395e-03], + [ 3.9101e-04, 2.5253e-03, -1.5998e-04, ..., 2.7585e-04, + 1.3695e-03, 5.5504e-04], + [ 5.5923e-03, 7.0534e-03, -8.1177e-03, ..., 5.5428e-03, + 7.5102e-04, 2.7866e-03], + ..., + [-7.4482e-04, 1.6193e-03, -4.8494e-04, ..., -1.1301e-03, + 6.5422e-04, 4.0054e-05], + [ 9.0790e-04, 2.7943e-03, 2.9144e-03, ..., 9.5081e-04, + -1.5187e-04, 2.6093e-03], + [ 9.7084e-04, -7.1669e-04, -2.0647e-04, ..., 9.2447e-05, + -5.1737e-05, -3.0651e-03]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0192, -0.0039, 0.0064, 0.0194, -0.0083, -0.0022, 0.0063, 0.0266, + -0.0294, 0.0367], device='cuda:0'), grad: tensor([-0.0108, 0.0132, 0.0043, -0.0146, 0.0024, -0.0416, 0.0088, 0.0126, + 0.0235, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 217.81, cls_loss 0.5925 cls_loss_mapping 0.0203 cls_loss_causal 0.5648 re_mapping 0.0115 re_causal 0.0295 /// teacc 98.66 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0529, 0.0589, -0.0623, ..., -0.0729, -0.0457, -0.0112], + [-0.0377, -0.0981, 0.0128, ..., -0.0457, -0.0193, -0.0429], + [ 0.0227, -0.0623, 0.0346, ..., 0.1330, -0.0746, -0.0357], + ..., + [-0.0513, -0.0880, 0.0687, ..., -0.0167, -0.0044, 0.0573], + [ 0.0044, 0.0280, -0.0363, ..., -0.0577, -0.0841, -0.0022], + [-0.1018, -0.0163, -0.0088, ..., -0.1024, 0.0808, 0.0116]], + device='cuda:0'), grad: tensor([[-1.4999e-02, -2.7008e-03, 3.7789e-04, ..., -9.5520e-03, + 1.0405e-03, 2.4128e-03], + [ 3.0303e-04, 1.0128e-03, -5.6076e-03, ..., 4.3988e-04, + -1.0345e-02, -4.8294e-03], + [ 1.1505e-02, 3.0589e-04, -6.2943e-03, ..., 6.0883e-03, + -6.9122e-03, -7.8583e-03], + ..., + [ 8.8513e-05, -1.2255e-03, -2.5988e-04, ..., 1.2255e-03, + 3.2902e-03, -6.4993e-04], + [ 1.0710e-03, 1.4648e-03, 1.9445e-03, ..., 1.0853e-03, + 2.4776e-03, 2.5501e-03], + [ 2.6894e-04, 3.1986e-03, 1.0424e-03, ..., -5.9929e-03, + 5.2414e-03, 5.3062e-03]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0191, -0.0038, 0.0065, 0.0184, -0.0089, -0.0022, 0.0068, 0.0267, + -0.0290, 0.0371], device='cuda:0'), grad: tensor([-0.0201, 0.0007, -0.0297, 0.0086, 0.0234, -0.0469, 0.0126, -0.0121, + 0.0214, 0.0421], device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 218.38, cls_loss 0.6149 cls_loss_mapping 0.0260 cls_loss_causal 0.5832 re_mapping 0.0117 re_causal 0.0283 /// teacc 98.45 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0525, 0.0595, -0.0624, ..., -0.0731, -0.0454, -0.0111], + [-0.0389, -0.0982, 0.0128, ..., -0.0459, -0.0195, -0.0441], + [ 0.0227, -0.0623, 0.0347, ..., 0.1328, -0.0747, -0.0359], + ..., + [-0.0519, -0.0879, 0.0690, ..., -0.0175, -0.0055, 0.0580], + [ 0.0055, 0.0282, -0.0374, ..., -0.0584, -0.0845, -0.0034], + [-0.1033, -0.0178, -0.0087, ..., -0.1021, 0.0810, 0.0122]], + device='cuda:0'), grad: tensor([[ 3.7060e-03, 4.5204e-03, 4.6372e-04, ..., 4.5204e-04, + 4.0722e-04, 3.4618e-03], + [ 7.3481e-04, -4.0746e-04, -5.3120e-04, ..., 3.1495e-04, + 3.6263e-04, -4.0054e-04], + [ 2.0294e-03, 2.5845e-03, 5.6219e-04, ..., -1.2093e-03, + 2.2638e-04, 2.2945e-03], + ..., + [ 6.9904e-04, 1.8139e-03, -4.2915e-06, ..., -2.7871e-04, + 3.3903e-04, 1.6947e-03], + [-5.5820e-05, -8.5068e-04, 2.4796e-03, ..., 1.5945e-03, + -2.0256e-03, 7.9203e-04], + [ 1.6670e-03, 1.3351e-04, 1.2388e-03, ..., 7.1430e-04, + 1.9016e-03, 9.5415e-04]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0192, -0.0041, 0.0060, 0.0189, -0.0083, -0.0030, 0.0072, 0.0274, + -0.0297, 0.0372], device='cuda:0'), grad: tensor([ 0.0066, -0.0078, 0.0173, -0.0053, -0.0356, -0.0046, -0.0253, 0.0149, + 0.0183, 0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 218.12, cls_loss 0.6057 cls_loss_mapping 0.0195 cls_loss_causal 0.5707 re_mapping 0.0117 re_causal 0.0292 /// teacc 98.62 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0530, 0.0588, -0.0628, ..., -0.0741, -0.0460, -0.0108], + [-0.0393, -0.0988, 0.0122, ..., -0.0457, -0.0192, -0.0451], + [ 0.0241, -0.0618, 0.0355, ..., 0.1337, -0.0739, -0.0361], + ..., + [-0.0508, -0.0873, 0.0691, ..., -0.0179, -0.0061, 0.0581], + [ 0.0045, 0.0281, -0.0369, ..., -0.0593, -0.0841, -0.0024], + [-0.1025, -0.0164, -0.0089, ..., -0.1026, 0.0818, 0.0123]], + device='cuda:0'), grad: tensor([[ 2.2869e-03, 5.3902e-03, 1.6651e-03, ..., 1.5211e-04, + 6.4731e-05, 3.0518e-03], + [ 3.8290e-04, -2.2640e-03, -2.5578e-03, ..., 7.8976e-05, + 5.0217e-05, -3.2539e-03], + [ 1.5192e-03, 4.0245e-03, 1.3332e-03, ..., -1.1272e-03, + 1.4937e-04, 2.8534e-03], + ..., + [ 6.0844e-04, 1.9312e-03, 2.3499e-03, ..., 3.2091e-04, + 1.3113e-03, 2.2144e-03], + [ 2.3975e-03, 5.4054e-03, 1.1911e-03, ..., 2.7442e-04, + 2.6226e-03, 1.6861e-03], + [ 6.6423e-04, 1.8234e-03, -5.7459e-04, ..., -1.3900e-04, + -1.6603e-03, 1.5240e-03]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0191, -0.0041, 0.0056, 0.0188, -0.0091, -0.0029, 0.0075, 0.0287, + -0.0295, 0.0366], device='cuda:0'), grad: tensor([ 0.0365, -0.0410, 0.0279, -0.0593, -0.0455, 0.0004, 0.0183, 0.0241, + 0.0262, 0.0124], device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 218.07, cls_loss 0.6235 cls_loss_mapping 0.0199 cls_loss_causal 0.5912 re_mapping 0.0114 re_causal 0.0286 /// teacc 98.50 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0544, 0.0585, -0.0619, ..., -0.0749, -0.0461, -0.0104], + [-0.0394, -0.0991, 0.0117, ..., -0.0442, -0.0189, -0.0459], + [ 0.0243, -0.0625, 0.0356, ..., 0.1336, -0.0753, -0.0367], + ..., + [-0.0505, -0.0874, 0.0693, ..., -0.0179, -0.0056, 0.0584], + [ 0.0040, 0.0283, -0.0365, ..., -0.0599, -0.0843, -0.0016], + [-0.1028, -0.0162, -0.0098, ..., -0.1034, 0.0816, 0.0122]], + device='cuda:0'), grad: tensor([[ 0.0004, -0.0002, 0.0010, ..., 0.0004, 0.0003, -0.0024], + [-0.0016, -0.0019, 0.0005, ..., -0.0003, 0.0009, 0.0005], + [-0.0038, -0.0020, -0.0026, ..., -0.0051, 0.0003, 0.0018], + ..., + [ 0.0006, 0.0010, -0.0024, ..., 0.0009, 0.0012, -0.0019], + [ 0.0013, 0.0025, 0.0011, ..., 0.0005, 0.0006, 0.0022], + [ 0.0004, -0.0026, -0.0022, ..., -0.0007, -0.0058, -0.0092]], + device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0194, -0.0045, 0.0049, 0.0188, -0.0092, -0.0027, 0.0076, 0.0289, + -0.0293, 0.0369], device='cuda:0'), grad: tensor([-0.0143, -0.0353, 0.0056, 0.0188, 0.0213, 0.0224, 0.0177, 0.0026, + 0.0198, -0.0587], device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 218.38, cls_loss 0.5585 cls_loss_mapping 0.0209 cls_loss_causal 0.5284 re_mapping 0.0118 re_causal 0.0297 /// teacc 98.63 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0567, 0.0580, -0.0621, ..., -0.0759, -0.0471, -0.0105], + [-0.0395, -0.0992, 0.0117, ..., -0.0447, -0.0195, -0.0468], + [ 0.0227, -0.0632, 0.0354, ..., 0.1348, -0.0763, -0.0365], + ..., + [-0.0499, -0.0881, 0.0697, ..., -0.0189, -0.0057, 0.0587], + [ 0.0052, 0.0285, -0.0366, ..., -0.0591, -0.0844, -0.0015], + [-0.1038, -0.0169, -0.0094, ..., -0.1036, 0.0820, 0.0117]], + device='cuda:0'), grad: tensor([[ 8.0585e-04, -4.8027e-03, 5.4646e-04, ..., 1.5938e-04, + 1.2932e-03, 2.1973e-03], + [ 8.6117e-04, 1.5087e-03, 1.6298e-03, ..., 3.6389e-05, + 1.0080e-03, 1.6174e-03], + [-8.7738e-03, -2.9011e-03, 7.2765e-04, ..., 4.5627e-05, + 6.8903e-04, 1.2674e-03], + ..., + [ 4.1962e-03, 3.3360e-03, -5.3825e-03, ..., 1.0973e-04, + -6.8521e-04, 6.4898e-04], + [ 6.6452e-03, 5.0774e-03, 1.3952e-03, ..., 1.3924e-04, + 2.8572e-03, 2.6131e-03], + [ 5.7507e-04, -5.5695e-03, -9.4986e-04, ..., -3.2616e-03, + -8.1177e-03, -7.4654e-03]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0191, -0.0042, 0.0047, 0.0192, -0.0093, -0.0030, 0.0077, 0.0287, + -0.0290, 0.0367], device='cuda:0'), grad: tensor([ 0.0040, 0.0145, -0.0081, -0.0229, 0.0335, -0.0215, 0.0200, -0.0082, + 0.0295, -0.0408], device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 218.04, cls_loss 0.5476 cls_loss_mapping 0.0168 cls_loss_causal 0.5120 re_mapping 0.0119 re_causal 0.0313 /// teacc 98.43 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0577, 0.0576, -0.0629, ..., -0.0759, -0.0466, -0.0099], + [-0.0398, -0.0993, 0.0110, ..., -0.0460, -0.0189, -0.0468], + [ 0.0240, -0.0619, 0.0354, ..., 0.1358, -0.0764, -0.0373], + ..., + [-0.0515, -0.0894, 0.0700, ..., -0.0196, -0.0059, 0.0584], + [ 0.0063, 0.0296, -0.0367, ..., -0.0600, -0.0855, -0.0026], + [-0.1031, -0.0159, -0.0095, ..., -0.1044, 0.0816, 0.0120]], + device='cuda:0'), grad: tensor([[ 2.9778e-04, -9.1696e-04, 4.5466e-04, ..., -2.1782e-03, + 1.1492e-03, 1.9951e-03], + [ 5.0783e-04, 8.2874e-04, 7.6962e-04, ..., 1.3638e-04, + 6.9094e-04, 1.4753e-03], + [-2.5511e-04, 1.4963e-03, -5.7459e-04, ..., -6.0081e-04, + -1.2140e-03, -8.6498e-04], + ..., + [-2.1350e-04, -3.8481e-04, -4.3945e-03, ..., -3.2559e-06, + 1.1272e-03, -5.0583e-03], + [ 2.7466e-03, 3.8681e-03, 1.3027e-03, ..., 1.0090e-03, + 9.0742e-04, 3.8700e-03], + [-5.2309e-04, 1.8415e-03, 7.0989e-05, ..., 5.4598e-04, + -1.5345e-03, -4.7350e-04]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0190, -0.0045, 0.0048, 0.0186, -0.0085, -0.0026, 0.0076, 0.0280, + -0.0291, 0.0376], device='cuda:0'), grad: tensor([ 0.0065, -0.0098, -0.0126, 0.0015, -0.0029, -0.0105, 0.0170, -0.0256, + 0.0253, 0.0111], device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 218.30, cls_loss 0.5738 cls_loss_mapping 0.0209 cls_loss_causal 0.5383 re_mapping 0.0120 re_causal 0.0296 /// teacc 98.54 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0585, 0.0578, -0.0622, ..., -0.0747, -0.0454, -0.0101], + [-0.0397, -0.0999, 0.0119, ..., -0.0457, -0.0187, -0.0467], + [ 0.0252, -0.0613, 0.0355, ..., 0.1367, -0.0768, -0.0380], + ..., + [-0.0527, -0.0900, 0.0706, ..., -0.0195, -0.0053, 0.0596], + [ 0.0064, 0.0298, -0.0369, ..., -0.0610, -0.0858, -0.0029], + [-0.1032, -0.0159, -0.0102, ..., -0.1056, 0.0810, 0.0112]], + device='cuda:0'), grad: tensor([[-3.4094e-04, -5.4073e-04, 1.4477e-03, ..., 4.0340e-04, + 2.1815e-04, 6.3801e-04], + [ 5.2357e-04, -1.2140e-03, 3.4790e-03, ..., 1.0162e-02, + 1.5855e-04, 8.2874e-04], + [-1.0239e-02, -2.7523e-03, -7.8678e-04, ..., -9.7752e-04, + -1.9872e-04, -5.8174e-03], + ..., + [-7.9036e-05, -1.6508e-03, -3.7365e-03, ..., 9.0313e-04, + -6.0043e-03, -3.7708e-03], + [ 4.1389e-03, 2.1420e-03, 1.4029e-03, ..., 2.1100e-04, + 3.9148e-04, 3.1090e-03], + [ 3.3402e-04, -2.0046e-03, -5.5504e-04, ..., 5.0068e-04, + -6.6795e-03, -5.3177e-03]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0190, -0.0045, 0.0048, 0.0193, -0.0083, -0.0037, 0.0078, 0.0277, + -0.0293, 0.0379], device='cuda:0'), grad: tensor([ 0.0123, 0.0166, -0.0462, -0.0416, 0.0420, 0.0144, -0.0131, -0.0137, + 0.0231, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 218.59, cls_loss 0.6020 cls_loss_mapping 0.0199 cls_loss_causal 0.5659 re_mapping 0.0114 re_causal 0.0283 /// teacc 98.67 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0588, 0.0578, -0.0627, ..., -0.0761, -0.0464, -0.0097], + [-0.0386, -0.0989, 0.0119, ..., -0.0452, -0.0185, -0.0453], + [ 0.0248, -0.0618, 0.0359, ..., 0.1369, -0.0774, -0.0389], + ..., + [-0.0536, -0.0910, 0.0708, ..., -0.0202, -0.0063, 0.0587], + [ 0.0071, 0.0297, -0.0365, ..., -0.0617, -0.0855, -0.0019], + [-0.1050, -0.0155, -0.0102, ..., -0.1046, 0.0820, 0.0110]], + device='cuda:0'), grad: tensor([[ 3.9876e-05, -8.9169e-04, -1.1673e-02, ..., -5.3520e-03, + -1.0414e-03, 2.1400e-03], + [ 1.6804e-03, 1.8425e-03, 5.9776e-03, ..., 4.4632e-03, + 1.4582e-03, 5.7564e-03], + [-7.3128e-03, -3.2406e-03, -7.0839e-03, ..., -8.7309e-04, + 4.2707e-05, -5.3558e-03], + ..., + [ 2.9564e-03, 2.2888e-03, 8.4925e-04, ..., 3.2449e-04, + 1.4992e-03, 2.6441e-04], + [-3.7060e-03, -4.6463e-03, -4.1389e-03, ..., -9.4995e-06, + -4.3755e-03, -9.6588e-03], + [ 1.2569e-03, -3.7556e-03, -1.5659e-03, ..., 2.6965e-04, + -1.1047e-02, -2.3972e-02]], device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0189, -0.0028, 0.0047, 0.0188, -0.0086, -0.0041, 0.0076, 0.0273, + -0.0289, 0.0376], device='cuda:0'), grad: tensor([-0.0408, 0.0460, -0.0248, -0.0080, 0.0840, 0.0205, 0.0031, 0.0033, + -0.0398, -0.0434], device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 219.07, cls_loss 0.5724 cls_loss_mapping 0.0200 cls_loss_causal 0.5468 re_mapping 0.0111 re_causal 0.0281 /// teacc 98.61 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0591, 0.0581, -0.0631, ..., -0.0772, -0.0474, -0.0111], + [-0.0391, -0.0985, 0.0122, ..., -0.0455, -0.0195, -0.0469], + [ 0.0254, -0.0625, 0.0361, ..., 0.1364, -0.0769, -0.0386], + ..., + [-0.0542, -0.0918, 0.0708, ..., -0.0198, -0.0069, 0.0583], + [ 0.0069, 0.0294, -0.0366, ..., -0.0620, -0.0862, -0.0019], + [-0.1068, -0.0162, -0.0097, ..., -0.1056, 0.0830, 0.0116]], + device='cuda:0'), grad: tensor([[ 0.0005, -0.0004, -0.0046, ..., 0.0004, 0.0004, -0.0047], + [-0.0016, -0.0046, -0.0049, ..., 0.0008, -0.0018, -0.0173], + [ 0.0010, 0.0010, 0.0033, ..., 0.0008, 0.0009, 0.0056], + ..., + [ 0.0011, 0.0022, 0.0012, ..., 0.0012, 0.0020, 0.0091], + [ 0.0010, 0.0018, 0.0001, ..., -0.0005, -0.0018, -0.0028], + [ 0.0018, 0.0022, 0.0053, ..., 0.0010, 0.0030, 0.0112]], + device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0188, -0.0035, 0.0046, 0.0193, -0.0087, -0.0036, 0.0079, 0.0272, + -0.0288, 0.0376], device='cuda:0'), grad: tensor([-0.0170, -0.0350, 0.0280, 0.0217, 0.0290, -0.0024, -0.0726, 0.0224, + 0.0084, 0.0174], device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 219.98, cls_loss 0.5892 cls_loss_mapping 0.0172 cls_loss_causal 0.5507 re_mapping 0.0113 re_causal 0.0270 /// teacc 98.56 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0595, 0.0580, -0.0638, ..., -0.0779, -0.0482, -0.0107], + [-0.0384, -0.0985, 0.0115, ..., -0.0460, -0.0198, -0.0471], + [ 0.0260, -0.0616, 0.0365, ..., 0.1366, -0.0770, -0.0382], + ..., + [-0.0541, -0.0911, 0.0717, ..., -0.0183, -0.0064, 0.0589], + [ 0.0071, 0.0295, -0.0359, ..., -0.0631, -0.0855, -0.0017], + [-0.1069, -0.0162, -0.0103, ..., -0.1068, 0.0829, 0.0104]], + device='cuda:0'), grad: tensor([[ 1.1749e-03, -5.6028e-04, 1.8930e-04, ..., 5.0664e-06, + 2.4676e-04, 7.3290e-04], + [ 7.3624e-04, 1.1921e-03, 2.6817e-03, ..., 1.4055e-04, + 1.3790e-03, 6.1951e-03], + [-1.7297e-04, 1.3244e-04, -1.9150e-03, ..., 8.8811e-05, + -3.3915e-05, -3.4847e-03], + ..., + [ 3.1776e-03, 8.5211e-04, -5.3520e-03, ..., -4.6432e-05, + -4.5891e-03, -8.8348e-03], + [ 7.8964e-04, 1.2894e-03, 1.7939e-03, ..., 5.1558e-05, + 1.0900e-03, 3.9062e-03], + [ 5.3549e-04, 5.9605e-04, 2.4796e-03, ..., 1.9360e-04, + -1.5001e-03, -4.5204e-03]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0184, -0.0037, 0.0052, 0.0191, -0.0093, -0.0037, 0.0078, 0.0284, + -0.0287, 0.0371], device='cuda:0'), grad: tensor([ 0.0062, 0.0338, -0.0129, -0.0233, 0.0407, 0.0177, -0.0494, -0.0209, + 0.0217, -0.0137], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 105---------------------------------------------------- +epoch 105, time 219.74, cls_loss 0.6047 cls_loss_mapping 0.0194 cls_loss_causal 0.5775 re_mapping 0.0107 re_causal 0.0276 /// teacc 98.74 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0592, 0.0583, -0.0634, ..., -0.0765, -0.0482, -0.0099], + [-0.0384, -0.0992, 0.0124, ..., -0.0457, -0.0195, -0.0467], + [ 0.0259, -0.0621, 0.0368, ..., 0.1369, -0.0784, -0.0385], + ..., + [-0.0545, -0.0917, 0.0708, ..., -0.0178, -0.0068, 0.0584], + [ 0.0070, 0.0295, -0.0359, ..., -0.0631, -0.0855, -0.0017], + [-0.1087, -0.0171, -0.0100, ..., -0.1074, 0.0828, 0.0102]], + device='cuda:0'), grad: tensor([[ 1.5650e-03, -9.0551e-04, 2.9993e-04, ..., 1.1230e-04, + 2.0981e-04, 1.1616e-03], + [-2.3067e-04, 1.9705e-04, 2.2948e-04, ..., -6.6328e-04, + -1.6565e-03, -1.6537e-03], + [ 1.5764e-03, 4.5300e-04, -1.7824e-03, ..., -3.7432e-04, + 3.7122e-04, 2.7370e-03], + ..., + [ 1.1396e-03, 6.3038e-04, 1.4868e-03, ..., 2.3639e-04, + 1.0147e-03, 2.1610e-03], + [-4.4847e-04, 3.2745e-06, 1.1644e-03, ..., 1.7929e-04, + 4.3750e-04, 2.1439e-03], + [ 1.1101e-03, -9.6083e-04, -7.6008e-04, ..., 1.9646e-04, + -2.2488e-03, -1.6441e-03]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0188, -0.0037, 0.0051, 0.0191, -0.0091, -0.0034, 0.0077, 0.0275, + -0.0285, 0.0372], device='cuda:0'), grad: tensor([ 0.0193, -0.0184, 0.0019, 0.0154, 0.0178, -0.0197, -0.0662, 0.0234, + 0.0195, 0.0069], device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 219.57, cls_loss 0.5996 cls_loss_mapping 0.0167 cls_loss_causal 0.5687 re_mapping 0.0114 re_causal 0.0289 /// teacc 98.68 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0598, 0.0583, -0.0641, ..., -0.0785, -0.0489, -0.0104], + [-0.0394, -0.1001, 0.0125, ..., -0.0456, -0.0191, -0.0473], + [ 0.0258, -0.0629, 0.0376, ..., 0.1377, -0.0791, -0.0383], + ..., + [-0.0562, -0.0922, 0.0708, ..., -0.0186, -0.0068, 0.0584], + [ 0.0073, 0.0293, -0.0361, ..., -0.0631, -0.0856, -0.0008], + [-0.1086, -0.0154, -0.0112, ..., -0.1084, 0.0830, 0.0104]], + device='cuda:0'), grad: tensor([[ 0.0002, -0.0003, 0.0008, ..., 0.0006, 0.0006, 0.0017], + [-0.0014, -0.0013, -0.0008, ..., 0.0203, -0.0004, -0.0020], + [ 0.0004, 0.0006, -0.0004, ..., -0.0065, -0.0002, -0.0047], + ..., + [ 0.0003, 0.0003, 0.0025, ..., -0.0198, 0.0021, 0.0035], + [ 0.0082, 0.0118, -0.0023, ..., 0.0013, -0.0055, -0.0048], + [ 0.0004, 0.0008, 0.0002, ..., 0.0012, 0.0011, 0.0033]], + device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0175, -0.0026, 0.0044, 0.0188, -0.0089, -0.0033, 0.0079, 0.0278, + -0.0278, 0.0368], device='cuda:0'), grad: tensor([ 0.0089, 0.0126, -0.0121, 0.0097, -0.0068, 0.0031, 0.0105, -0.0082, + -0.0328, 0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 219.67, cls_loss 0.5612 cls_loss_mapping 0.0187 cls_loss_causal 0.5313 re_mapping 0.0120 re_causal 0.0298 /// teacc 98.74 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0607, 0.0587, -0.0636, ..., -0.0798, -0.0487, -0.0107], + [-0.0397, -0.1007, 0.0129, ..., -0.0451, -0.0191, -0.0463], + [ 0.0259, -0.0636, 0.0375, ..., 0.1383, -0.0795, -0.0388], + ..., + [-0.0559, -0.0923, 0.0701, ..., -0.0194, -0.0074, 0.0592], + [ 0.0069, 0.0294, -0.0366, ..., -0.0641, -0.0867, -0.0025], + [-0.1080, -0.0163, -0.0106, ..., -0.1084, 0.0840, 0.0097]], + device='cuda:0'), grad: tensor([[ 0.0073, 0.0107, 0.0013, ..., 0.0006, 0.0005, 0.0029], + [ 0.0014, 0.0011, 0.0033, ..., -0.0007, 0.0012, 0.0031], + [ 0.0016, 0.0014, 0.0023, ..., 0.0021, 0.0006, 0.0037], + ..., + [ 0.0006, 0.0006, -0.0072, ..., -0.0018, -0.0026, -0.0014], + [-0.0084, -0.0026, -0.0026, ..., -0.0020, -0.0036, -0.0126], + [ 0.0019, 0.0015, 0.0018, ..., 0.0003, 0.0009, 0.0033]], + device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0173, -0.0022, 0.0040, 0.0192, -0.0091, -0.0033, 0.0086, 0.0276, + -0.0288, 0.0375], device='cuda:0'), grad: tensor([ 0.0280, 0.0278, 0.0262, -0.0480, 0.0265, 0.0006, 0.0019, -0.0385, + -0.0476, 0.0231], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 108---------------------------------------------------- +epoch 108, time 219.68, cls_loss 0.5787 cls_loss_mapping 0.0170 cls_loss_causal 0.5451 re_mapping 0.0111 re_causal 0.0278 /// teacc 98.76 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0605, 0.0595, -0.0639, ..., -0.0785, -0.0487, -0.0107], + [-0.0406, -0.1010, 0.0124, ..., -0.0446, -0.0197, -0.0466], + [ 0.0254, -0.0646, 0.0377, ..., 0.1386, -0.0786, -0.0385], + ..., + [-0.0581, -0.0940, 0.0704, ..., -0.0190, -0.0077, 0.0590], + [ 0.0074, 0.0300, -0.0367, ..., -0.0638, -0.0866, -0.0025], + [-0.1081, -0.0167, -0.0098, ..., -0.1106, 0.0840, 0.0103]], + device='cuda:0'), grad: tensor([[ 4.9305e-04, -2.7084e-03, -1.1549e-03, ..., -9.3758e-05, + 5.2691e-04, 2.1410e-04], + [-1.0180e-04, -2.9316e-03, -3.8719e-03, ..., 6.9678e-05, + -4.0092e-03, -5.9090e-03], + [ 3.1414e-03, 2.7752e-03, 1.1396e-03, ..., 8.8120e-04, + 1.2274e-03, 2.9068e-03], + ..., + [-2.7447e-03, -8.2541e-04, -7.6332e-03, ..., -1.6575e-03, + -4.3831e-03, -8.7433e-03], + [ 3.6931e-04, 2.1744e-03, 4.8523e-03, ..., 4.3392e-04, + 2.8591e-03, 4.6463e-03], + [-1.9825e-04, -6.7472e-04, 1.6537e-03, ..., 1.1292e-03, + 2.7313e-03, 1.4734e-03]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0177, -0.0027, 0.0044, 0.0194, -0.0093, -0.0038, 0.0091, 0.0276, + -0.0290, 0.0373], device='cuda:0'), grad: tensor([ 0.0028, -0.0440, 0.0236, -0.0003, 0.0068, 0.0173, 0.0205, -0.0553, + 0.0248, 0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 218.86, cls_loss 0.5955 cls_loss_mapping 0.0189 cls_loss_causal 0.5682 re_mapping 0.0104 re_causal 0.0275 /// teacc 98.40 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0609, 0.0596, -0.0635, ..., -0.0779, -0.0492, -0.0108], + [-0.0401, -0.1004, 0.0114, ..., -0.0447, -0.0197, -0.0483], + [ 0.0243, -0.0662, 0.0381, ..., 0.1388, -0.0784, -0.0374], + ..., + [-0.0585, -0.0947, 0.0710, ..., -0.0191, -0.0064, 0.0598], + [ 0.0068, 0.0299, -0.0369, ..., -0.0642, -0.0870, -0.0026], + [-0.1080, -0.0164, -0.0107, ..., -0.1115, 0.0828, 0.0090]], + device='cuda:0'), grad: tensor([[ 2.2817e-04, -7.8321e-05, 7.2145e-04, ..., 4.6015e-04, + 4.9591e-04, 8.7786e-04], + [-1.9073e-03, 2.8801e-04, 3.1738e-03, ..., 1.5535e-03, + -2.1877e-03, 6.9160e-03], + [-3.1033e-03, -5.8556e-03, -3.9024e-03, ..., -6.2981e-03, + -6.2714e-03, 1.2150e-03], + ..., + [ 4.1819e-04, 1.0929e-03, 8.2111e-04, ..., 1.1120e-03, + 5.1641e-04, 2.1782e-03], + [-4.6234e-03, -6.9857e-04, 2.7657e-03, ..., 2.9316e-03, + 3.8986e-03, 2.0885e-03], + [ 8.6594e-04, 1.3580e-03, 1.1196e-03, ..., 1.2951e-03, + 1.4029e-03, 7.5340e-04]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0182, -0.0031, 0.0045, 0.0192, -0.0093, -0.0035, 0.0093, 0.0284, + -0.0294, 0.0364], device='cuda:0'), grad: tensor([ 1.1406e-02, 1.7807e-02, -2.5604e-02, -3.7933e-02, -1.9608e-02, + 2.3727e-02, -5.9545e-05, 2.1393e-02, 1.9806e-02, -1.0918e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 219.12, cls_loss 0.5871 cls_loss_mapping 0.0176 cls_loss_causal 0.5519 re_mapping 0.0104 re_causal 0.0267 /// teacc 98.65 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0603, 0.0600, -0.0639, ..., -0.0788, -0.0495, -0.0105], + [-0.0406, -0.1009, 0.0124, ..., -0.0449, -0.0195, -0.0473], + [ 0.0243, -0.0669, 0.0389, ..., 0.1394, -0.0784, -0.0385], + ..., + [-0.0587, -0.0950, 0.0712, ..., -0.0200, -0.0060, 0.0587], + [ 0.0071, 0.0304, -0.0367, ..., -0.0644, -0.0870, -0.0024], + [-0.1080, -0.0158, -0.0107, ..., -0.1112, 0.0834, 0.0101]], + device='cuda:0'), grad: tensor([[ 1.3943e-03, 1.7080e-03, 1.1940e-03, ..., 6.9475e-04, + 1.2648e-04, 2.2621e-03], + [ 5.8031e-04, -1.5240e-03, -1.4372e-03, ..., 1.2326e-04, + 2.1577e-05, -6.2332e-03], + [-1.0330e-02, -1.0880e-02, 5.2977e-04, ..., -1.2808e-03, + 4.0196e-06, -1.1969e-03], + ..., + [ 9.3079e-04, 1.5011e-03, -3.2227e-02, ..., 8.9312e-04, + -1.6586e-02, 2.6970e-03], + [-2.5349e-03, -1.9550e-03, -1.9503e-03, ..., 3.0637e-04, + 1.3542e-03, -2.1324e-03], + [-1.1158e-03, -2.7504e-03, 2.8992e-02, ..., -2.8286e-03, + 1.3809e-02, -3.1853e-03]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0185, -0.0025, 0.0042, 0.0184, -0.0098, -0.0034, 0.0090, 0.0280, + -0.0293, 0.0376], device='cuda:0'), grad: tensor([ 0.0120, -0.0274, -0.0257, 0.0181, 0.0078, 0.0240, 0.0146, -0.0137, + -0.0173, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 218.97, cls_loss 0.6020 cls_loss_mapping 0.0207 cls_loss_causal 0.5686 re_mapping 0.0110 re_causal 0.0270 /// teacc 98.59 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0607, 0.0596, -0.0627, ..., -0.0791, -0.0500, -0.0097], + [-0.0398, -0.1009, 0.0123, ..., -0.0443, -0.0198, -0.0481], + [ 0.0237, -0.0688, 0.0374, ..., 0.1396, -0.0796, -0.0394], + ..., + [-0.0590, -0.0954, 0.0727, ..., -0.0195, -0.0065, 0.0596], + [ 0.0075, 0.0307, -0.0369, ..., -0.0655, -0.0870, -0.0033], + [-0.1085, -0.0149, -0.0116, ..., -0.1123, 0.0838, 0.0105]], + device='cuda:0'), grad: tensor([[ 5.5933e-04, -3.1929e-03, -2.1493e-04, ..., 7.2718e-05, + -1.0902e-04, -2.9831e-03], + [ 7.0667e-04, 1.3609e-03, 2.2602e-03, ..., 3.9673e-04, + 2.5916e-04, 3.0365e-03], + [-5.2910e-03, -3.8948e-03, -6.9122e-03, ..., -5.8174e-03, + 1.7512e-04, -6.4039e-04], + ..., + [ 1.2941e-03, 1.6270e-03, 3.0441e-03, ..., 5.2643e-04, + 5.5218e-04, 3.6926e-03], + [-2.0981e-03, -2.8992e-03, 3.6049e-04, ..., 4.2367e-04, + 4.5967e-04, 1.8415e-03], + [ 9.4271e-04, 8.2684e-04, -3.3894e-03, ..., 1.1406e-03, + 2.0075e-04, 5.4264e-04]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0191, -0.0025, 0.0036, 0.0183, -0.0092, -0.0032, 0.0087, 0.0283, + -0.0297, 0.0374], device='cuda:0'), grad: tensor([-0.0182, 0.0300, -0.0053, 0.0362, -0.0402, -0.0359, 0.0183, 0.0328, + -0.0151, -0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 217.88, cls_loss 0.5869 cls_loss_mapping 0.0156 cls_loss_causal 0.5511 re_mapping 0.0105 re_causal 0.0269 /// teacc 98.57 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0612, 0.0596, -0.0631, ..., -0.0794, -0.0512, -0.0096], + [-0.0400, -0.1010, 0.0120, ..., -0.0437, -0.0207, -0.0480], + [ 0.0251, -0.0667, 0.0375, ..., 0.1398, -0.0806, -0.0392], + ..., + [-0.0608, -0.0971, 0.0729, ..., -0.0195, -0.0065, 0.0590], + [ 0.0086, 0.0307, -0.0368, ..., -0.0659, -0.0874, -0.0042], + [-0.1092, -0.0155, -0.0116, ..., -0.1121, 0.0844, 0.0105]], + device='cuda:0'), grad: tensor([[ 7.6103e-04, 2.9349e-04, -2.9716e-03, ..., 1.1182e-04, + -3.5524e-04, -7.7171e-03], + [ 2.0131e-05, 8.4925e-04, 1.5240e-03, ..., 3.0041e-04, + 4.6563e-04, 2.5120e-03], + [-8.9188e-03, -1.6159e-02, 6.7568e-04, ..., 1.5938e-04, + 2.9707e-04, -7.3957e-04], + ..., + [-5.5466e-03, -7.7677e-04, -1.8883e-04, ..., 6.1655e-04, + 8.4734e-04, 1.3685e-04], + [-7.7391e-04, -7.1144e-04, 1.2217e-03, ..., 1.8418e-04, + 3.9172e-04, 1.3933e-03], + [ 1.8358e-03, 8.4686e-04, 1.4572e-03, ..., 2.5082e-04, + -3.0589e-04, 1.8349e-03]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0190, -0.0020, 0.0033, 0.0183, -0.0083, -0.0033, 0.0080, 0.0277, + -0.0298, 0.0377], device='cuda:0'), grad: tensor([-0.0200, 0.0070, -0.0164, 0.0459, -0.0122, 0.0113, -0.0212, -0.0254, + 0.0129, 0.0180], device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 217.97, cls_loss 0.5889 cls_loss_mapping 0.0131 cls_loss_causal 0.5463 re_mapping 0.0109 re_causal 0.0271 /// teacc 98.75 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0617, 0.0598, -0.0654, ..., -0.0799, -0.0517, -0.0105], + [-0.0396, -0.1016, 0.0120, ..., -0.0441, -0.0212, -0.0492], + [ 0.0246, -0.0673, 0.0372, ..., 0.1399, -0.0816, -0.0376], + ..., + [-0.0603, -0.0986, 0.0736, ..., -0.0186, -0.0067, 0.0592], + [ 0.0086, 0.0306, -0.0360, ..., -0.0662, -0.0870, -0.0043], + [-0.1084, -0.0151, -0.0117, ..., -0.1118, 0.0849, 0.0107]], + device='cuda:0'), grad: tensor([[ 2.1095e-03, 1.5154e-03, -6.2943e-04, ..., 1.3657e-03, + -7.4923e-05, -1.2159e-03], + [-6.4850e-04, -1.5297e-03, -4.2267e-03, ..., -2.3899e-03, + -5.1022e-04, -3.0613e-03], + [ 4.6730e-03, 6.9046e-03, 3.1185e-03, ..., -1.5078e-03, + -1.4591e-04, 5.2338e-03], + ..., + [-5.5466e-03, -8.1348e-04, -4.4518e-03, ..., 2.6703e-04, + 9.8324e-04, -1.5192e-03], + [ 4.1885e-03, 6.0043e-03, 7.0000e-03, ..., 1.4639e-03, + 1.1606e-03, 4.8790e-03], + [ 4.3750e-04, -1.9236e-03, -1.1768e-03, ..., 1.2589e-03, + -2.1210e-03, -2.5692e-03]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0186, -0.0023, 0.0037, 0.0182, -0.0080, -0.0037, 0.0085, 0.0280, + -0.0301, 0.0378], device='cuda:0'), grad: tensor([-0.0163, -0.0314, 0.0278, 0.0316, 0.0107, -0.0130, -0.0259, -0.0204, + 0.0315, 0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 217.91, cls_loss 0.5692 cls_loss_mapping 0.0117 cls_loss_causal 0.5348 re_mapping 0.0115 re_causal 0.0296 /// teacc 98.44 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0632, 0.0595, -0.0642, ..., -0.0811, -0.0510, -0.0091], + [-0.0407, -0.1026, 0.0117, ..., -0.0449, -0.0221, -0.0499], + [ 0.0248, -0.0680, 0.0373, ..., 0.1406, -0.0813, -0.0373], + ..., + [-0.0602, -0.0996, 0.0732, ..., -0.0192, -0.0069, 0.0589], + [ 0.0088, 0.0315, -0.0354, ..., -0.0654, -0.0863, -0.0039], + [-0.1085, -0.0144, -0.0121, ..., -0.1112, 0.0847, 0.0109]], + device='cuda:0'), grad: tensor([[ 1.6356e-03, 6.6471e-04, -3.2692e-03, ..., -2.2182e-03, + -1.7147e-03, -1.6857e-07], + [-3.2253e-03, 6.4850e-04, -1.0681e-04, ..., 3.6764e-04, + 2.8157e-04, 1.3657e-03], + [ 6.0349e-03, 7.4234e-03, 2.9964e-03, ..., 2.2755e-03, + 4.2367e-04, 6.1417e-03], + ..., + [ 5.0020e-04, -2.8387e-05, 5.7220e-05, ..., 3.6621e-04, + 3.4189e-04, -3.4904e-03], + [-9.7504e-03, -7.3204e-03, -4.3564e-03, ..., -2.8324e-04, + 5.2567e-03, -2.1095e-03], + [-2.6035e-03, -6.5384e-03, -5.3310e-04, ..., 4.8375e-04, + -1.1452e-02, -1.1345e-02]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0198, -0.0022, 0.0039, 0.0185, -0.0081, -0.0038, 0.0081, 0.0270, + -0.0300, 0.0377], device='cuda:0'), grad: tensor([ 0.0050, -0.0117, 0.0476, 0.0474, 0.0309, -0.0541, 0.0517, -0.0391, + -0.0536, -0.0241], device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 216.66, cls_loss 0.5500 cls_loss_mapping 0.0155 cls_loss_causal 0.5188 re_mapping 0.0112 re_causal 0.0278 /// teacc 98.67 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0626, 0.0588, -0.0649, ..., -0.0802, -0.0508, -0.0091], + [-0.0405, -0.1016, 0.0117, ..., -0.0455, -0.0221, -0.0510], + [ 0.0260, -0.0666, 0.0383, ..., 0.1408, -0.0811, -0.0363], + ..., + [-0.0600, -0.0992, 0.0728, ..., -0.0197, -0.0072, 0.0588], + [ 0.0094, 0.0312, -0.0361, ..., -0.0658, -0.0870, -0.0043], + [-0.1095, -0.0144, -0.0111, ..., -0.1115, 0.0845, 0.0107]], + device='cuda:0'), grad: tensor([[ 5.3310e-04, 1.4277e-03, 1.4858e-03, ..., 5.7817e-05, + 5.5218e-04, 1.8492e-03], + [-8.1177e-03, -2.3079e-03, 1.3145e-02, ..., 6.5863e-05, + 1.2770e-03, 4.2191e-03], + [-4.0007e-04, -1.0586e-03, -2.1011e-02, ..., -4.9896e-03, + 4.2582e-04, -1.6708e-02], + ..., + [ 6.5231e-04, 9.4414e-04, 6.4659e-03, ..., -6.5267e-05, + 1.7262e-03, 1.1055e-02], + [ 3.5305e-03, -2.5139e-03, 2.0275e-03, ..., 2.4071e-03, + -4.4365e-03, 3.9697e-04], + [ 1.4296e-03, -6.0797e-04, 4.1008e-03, ..., 1.9455e-04, + 4.2534e-03, 4.6692e-03]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0193, -0.0020, 0.0041, 0.0178, -0.0086, -0.0047, 0.0091, 0.0273, + -0.0298, 0.0381], device='cuda:0'), grad: tensor([-0.0069, 0.0231, -0.0409, 0.0024, 0.0076, 0.0029, -0.0187, 0.0358, + -0.0093, 0.0040], device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 216.53, cls_loss 0.5764 cls_loss_mapping 0.0192 cls_loss_causal 0.5469 re_mapping 0.0109 re_causal 0.0276 /// teacc 98.71 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0622, 0.0592, -0.0657, ..., -0.0797, -0.0519, -0.0085], + [-0.0395, -0.1012, 0.0105, ..., -0.0462, -0.0223, -0.0521], + [ 0.0251, -0.0678, 0.0382, ..., 0.1403, -0.0816, -0.0358], + ..., + [-0.0590, -0.0989, 0.0731, ..., -0.0189, -0.0079, 0.0585], + [ 0.0093, 0.0312, -0.0357, ..., -0.0643, -0.0866, -0.0046], + [-0.1106, -0.0160, -0.0103, ..., -0.1110, 0.0851, 0.0118]], + device='cuda:0'), grad: tensor([[-0.0001, -0.0017, -0.0003, ..., 0.0002, -0.0001, -0.0006], + [ 0.0001, 0.0013, 0.0036, ..., 0.0016, 0.0014, 0.0074], + [ 0.0004, -0.0072, -0.0098, ..., -0.0078, 0.0008, -0.0040], + ..., + [ 0.0003, 0.0024, 0.0061, ..., 0.0028, 0.0012, 0.0074], + [ 0.0013, 0.0037, 0.0023, ..., 0.0011, -0.0012, 0.0027], + [ 0.0021, 0.0066, 0.0014, ..., -0.0007, -0.0042, -0.0097]], + device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0195, -0.0027, 0.0036, 0.0180, -0.0086, -0.0040, 0.0088, 0.0283, + -0.0302, 0.0379], device='cuda:0'), grad: tensor([-0.0105, 0.0375, -0.0258, 0.0087, -0.0526, -0.0069, -0.0098, 0.0389, + 0.0105, 0.0099], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 117---------------------------------------------------- +epoch 117, time 217.40, cls_loss 0.5931 cls_loss_mapping 0.0179 cls_loss_causal 0.5546 re_mapping 0.0107 re_causal 0.0261 /// teacc 98.79 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0629, 0.0587, -0.0666, ..., -0.0802, -0.0529, -0.0091], + [-0.0399, -0.1022, 0.0093, ..., -0.0482, -0.0233, -0.0533], + [ 0.0251, -0.0672, 0.0388, ..., 0.1411, -0.0821, -0.0362], + ..., + [-0.0587, -0.0994, 0.0732, ..., -0.0188, -0.0078, 0.0580], + [ 0.0103, 0.0318, -0.0348, ..., -0.0643, -0.0869, -0.0038], + [-0.1116, -0.0166, -0.0099, ..., -0.1113, 0.0853, 0.0120]], + device='cuda:0'), grad: tensor([[ 3.6383e-04, 1.3285e-03, 1.0324e-04, ..., 9.5427e-05, + 5.2869e-05, -6.5851e-04], + [-1.4591e-04, -6.8474e-04, 6.6185e-04, ..., 7.5459e-05, + 8.1241e-05, 3.0994e-04], + [ 2.0161e-03, 2.2736e-03, -2.1820e-03, ..., -1.4782e-03, + 2.3997e-04, -2.9354e-03], + ..., + [ 1.4381e-03, 3.4122e-03, -2.1267e-03, ..., 9.7847e-04, + -7.4339e-04, -4.2953e-03], + [ 6.0081e-03, 4.7531e-03, -1.5278e-03, ..., 1.0186e-04, + 1.4086e-03, 3.4466e-03], + [ 4.6883e-03, 6.3477e-03, 3.3321e-03, ..., 7.7307e-05, + -6.0129e-04, 6.8588e-03]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0190, -0.0039, 0.0042, 0.0181, -0.0083, -0.0046, 0.0098, 0.0275, + -0.0293, 0.0382], device='cuda:0'), grad: tensor([ 0.0106, -0.0010, -0.0307, -0.0133, 0.0227, -0.0087, -0.0341, 0.0090, + -0.0013, 0.0469], device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 216.98, cls_loss 0.6002 cls_loss_mapping 0.0167 cls_loss_causal 0.5776 re_mapping 0.0102 re_causal 0.0259 /// teacc 98.66 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0628, 0.0593, -0.0682, ..., -0.0806, -0.0537, -0.0095], + [-0.0397, -0.1023, 0.0103, ..., -0.0480, -0.0230, -0.0533], + [ 0.0256, -0.0672, 0.0381, ..., 0.1401, -0.0812, -0.0372], + ..., + [-0.0583, -0.0985, 0.0735, ..., -0.0186, -0.0088, 0.0591], + [ 0.0104, 0.0317, -0.0343, ..., -0.0655, -0.0871, -0.0041], + [-0.1117, -0.0170, -0.0100, ..., -0.1109, 0.0860, 0.0116]], + device='cuda:0'), grad: tensor([[ 2.1744e-03, 3.4924e-03, 2.3193e-03, ..., 5.2547e-04, + 3.7980e-04, 2.2926e-03], + [-1.1975e-04, -8.7082e-05, 4.9667e-03, ..., -8.3351e-04, + 4.4227e-04, -5.2605e-03], + [-1.5869e-03, -1.2062e-02, -2.4548e-03, ..., -1.0500e-03, + -1.6193e-03, -3.8986e-03], + ..., + [ 1.1129e-03, 2.9144e-03, -9.5062e-03, ..., 1.2302e-03, + 7.8201e-04, 2.5439e-04], + [-7.7286e-03, -5.1804e-03, 2.1439e-03, ..., 9.5654e-04, + 4.2820e-04, 2.3670e-03], + [ 1.2140e-03, -8.6784e-04, -3.1342e-02, ..., -3.9902e-03, + -3.2215e-03, -1.4244e-02]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0180, -0.0035, 0.0043, 0.0187, -0.0091, -0.0040, 0.0103, 0.0277, + -0.0295, 0.0378], device='cuda:0'), grad: tensor([ 0.0351, -0.0217, -0.0521, 0.0139, 0.0556, 0.0215, 0.0273, -0.0197, + -0.0034, -0.0565], device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 216.35, cls_loss 0.5646 cls_loss_mapping 0.0135 cls_loss_causal 0.5301 re_mapping 0.0106 re_causal 0.0266 /// teacc 98.62 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0630, 0.0588, -0.0688, ..., -0.0805, -0.0545, -0.0087], + [-0.0395, -0.1023, 0.0096, ..., -0.0501, -0.0223, -0.0528], + [ 0.0250, -0.0673, 0.0402, ..., 0.1412, -0.0816, -0.0378], + ..., + [-0.0588, -0.0991, 0.0736, ..., -0.0183, -0.0092, 0.0601], + [ 0.0110, 0.0323, -0.0330, ..., -0.0658, -0.0860, -0.0033], + [-0.1120, -0.0167, -0.0101, ..., -0.1115, 0.0858, 0.0114]], + device='cuda:0'), grad: tensor([[ 7.0620e-04, 8.2169e-03, 1.7681e-03, ..., 1.0185e-03, + 6.1002e-07, 3.4714e-03], + [ 9.7096e-05, 2.8563e-04, -7.2527e-04, ..., -2.1708e-04, + 4.4368e-06, -1.1368e-03], + [ 6.3515e-04, 1.0815e-03, 5.3368e-03, ..., 5.2910e-03, + 3.8967e-06, 6.1035e-03], + ..., + [ 4.8828e-04, 1.8606e-03, 6.2904e-03, ..., -1.6918e-03, + 2.5332e-05, 1.7776e-03], + [ 5.5361e-04, 1.9855e-03, 1.8234e-03, ..., 1.0252e-03, + 1.2293e-05, 2.4509e-03], + [ 9.4748e-04, 3.8090e-03, -9.1476e-03, ..., 3.0804e-03, + 9.3639e-05, 2.6054e-03]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0184, -0.0037, 0.0042, 0.0179, -0.0104, -0.0040, 0.0102, 0.0282, + -0.0291, 0.0388], device='cuda:0'), grad: tensor([ 0.0346, -0.0335, 0.0428, 0.0231, -0.0676, -0.0705, 0.0263, 0.0229, + 0.0193, 0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 216.77, cls_loss 0.5408 cls_loss_mapping 0.0129 cls_loss_causal 0.5122 re_mapping 0.0109 re_causal 0.0274 /// teacc 98.75 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0630, 0.0586, -0.0685, ..., -0.0790, -0.0552, -0.0083], + [-0.0384, -0.1017, 0.0080, ..., -0.0514, -0.0220, -0.0527], + [ 0.0245, -0.0672, 0.0396, ..., 0.1410, -0.0818, -0.0377], + ..., + [-0.0592, -0.0996, 0.0738, ..., -0.0184, -0.0108, 0.0601], + [ 0.0108, 0.0319, -0.0316, ..., -0.0652, -0.0863, -0.0036], + [-0.1121, -0.0171, -0.0103, ..., -0.1122, 0.0862, 0.0115]], + device='cuda:0'), grad: tensor([[ 2.7637e-03, 4.2191e-03, 1.4687e-03, ..., -1.6963e-04, + 2.1613e-04, 2.7657e-03], + [-5.6696e-04, 2.5940e-04, -7.4625e-05, ..., 2.3935e-06, + -3.5787e-04, -2.4700e-04], + [ 1.6708e-03, 1.6909e-03, 8.8549e-04, ..., 3.6311e-04, + 4.3720e-05, 1.3695e-03], + ..., + [ 4.5371e-04, 8.2397e-04, 1.2598e-03, ..., 1.6764e-05, + 4.3702e-04, 1.4620e-03], + [ 2.7790e-03, 2.6970e-03, 8.1444e-04, ..., 1.7631e-04, + 3.3450e-04, 2.0809e-03], + [-3.2616e-03, -8.3771e-03, -2.5730e-03, ..., 1.9336e-04, + -5.1689e-04, -5.2223e-03]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0184, -0.0032, 0.0046, 0.0170, -0.0098, -0.0043, 0.0102, 0.0282, + -0.0294, 0.0389], device='cuda:0'), grad: tensor([ 0.0172, -0.0003, 0.0130, -0.0239, -0.0209, 0.0327, -0.0377, 0.0120, + 0.0188, -0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 216.62, cls_loss 0.5873 cls_loss_mapping 0.0153 cls_loss_causal 0.5475 re_mapping 0.0108 re_causal 0.0267 /// teacc 98.64 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0634, 0.0587, -0.0684, ..., -0.0780, -0.0570, -0.0084], + [-0.0388, -0.1003, 0.0082, ..., -0.0514, -0.0224, -0.0533], + [ 0.0246, -0.0677, 0.0396, ..., 0.1411, -0.0798, -0.0372], + ..., + [-0.0598, -0.1005, 0.0744, ..., -0.0185, -0.0107, 0.0605], + [ 0.0113, 0.0316, -0.0326, ..., -0.0655, -0.0873, -0.0031], + [-0.1125, -0.0171, -0.0104, ..., -0.1132, 0.0860, 0.0110]], + device='cuda:0'), grad: tensor([[-3.5343e-03, -3.0441e-03, -6.7043e-04, ..., 2.0671e-04, + 5.9068e-05, -5.8594e-03], + [ 1.0653e-03, 1.0662e-03, -3.8872e-03, ..., -1.2466e-02, + 1.1835e-03, -1.6785e-03], + [ 1.9684e-03, 1.0080e-03, 7.6904e-03, ..., 1.0223e-02, + 5.4836e-04, 5.0545e-03], + ..., + [ 7.9632e-04, -1.8203e-04, -1.7223e-03, ..., -3.3212e-04, + 4.7994e-04, -5.0449e-04], + [ 2.1191e-03, 1.6718e-03, 8.3618e-03, ..., 6.1226e-04, + 8.5211e-04, 2.2850e-03], + [ 1.4572e-03, 1.5135e-03, -1.0185e-02, ..., 1.1454e-03, + 2.5177e-03, 4.4098e-03]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0181, -0.0035, 0.0052, 0.0176, -0.0098, -0.0043, 0.0096, 0.0277, + -0.0292, 0.0391], device='cuda:0'), grad: tensor([-0.0428, -0.0110, 0.0449, -0.0108, 0.0117, 0.0201, -0.0290, -0.0145, + 0.0316, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 216.62, cls_loss 0.5489 cls_loss_mapping 0.0138 cls_loss_causal 0.5254 re_mapping 0.0109 re_causal 0.0261 /// teacc 98.73 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0640, 0.0584, -0.0691, ..., -0.0785, -0.0568, -0.0077], + [-0.0390, -0.1009, 0.0078, ..., -0.0513, -0.0224, -0.0539], + [ 0.0246, -0.0684, 0.0392, ..., 0.1424, -0.0817, -0.0375], + ..., + [-0.0613, -0.1012, 0.0745, ..., -0.0187, -0.0110, 0.0608], + [ 0.0112, 0.0314, -0.0339, ..., -0.0655, -0.0871, -0.0038], + [-0.1132, -0.0170, -0.0094, ..., -0.1125, 0.0873, 0.0106]], + device='cuda:0'), grad: tensor([[ 4.7088e-04, 1.1892e-03, 2.6436e-03, ..., -5.1707e-05, + 6.7830e-05, 2.8267e-03], + [ 5.6362e-04, -6.1989e-04, 1.8177e-03, ..., -1.4566e-06, + 4.7255e-04, -2.2945e-03], + [ 1.3523e-03, 2.0847e-03, 2.4071e-03, ..., -1.1444e-05, + 1.2410e-04, 2.1896e-03], + ..., + [ 6.1131e-04, -3.1872e-03, -4.9667e-03, ..., 1.3851e-05, + 5.8794e-04, -3.4466e-03], + [ 8.4000e-03, 8.2626e-03, 2.1687e-03, ..., 1.4186e-05, + 1.1911e-03, 2.5501e-03], + [-2.1896e-03, 9.2363e-04, 2.3327e-03, ..., 1.5870e-05, + -4.4098e-03, -1.0598e-04]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0180, -0.0037, 0.0041, 0.0185, -0.0089, -0.0046, 0.0100, 0.0277, + -0.0296, 0.0390], device='cuda:0'), grad: tensor([-0.0029, 0.0050, 0.0243, -0.0464, -0.0014, 0.0149, -0.0099, -0.0226, + 0.0344, 0.0046], device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 216.35, cls_loss 0.5421 cls_loss_mapping 0.0138 cls_loss_causal 0.5054 re_mapping 0.0113 re_causal 0.0283 /// teacc 98.47 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0631, 0.0591, -0.0684, ..., -0.0785, -0.0581, -0.0061], + [-0.0386, -0.1013, 0.0073, ..., -0.0511, -0.0237, -0.0551], + [ 0.0256, -0.0681, 0.0384, ..., 0.1421, -0.0809, -0.0383], + ..., + [-0.0621, -0.1024, 0.0749, ..., -0.0173, -0.0121, 0.0609], + [ 0.0112, 0.0312, -0.0331, ..., -0.0671, -0.0874, -0.0033], + [-0.1133, -0.0159, -0.0097, ..., -0.1126, 0.0879, 0.0107]], + device='cuda:0'), grad: tensor([[ 3.7575e-03, 6.1836e-03, 2.3518e-03, ..., 1.4019e-03, + 9.9564e-04, 2.2907e-03], + [-1.0824e-03, -6.4373e-04, -3.2063e-03, ..., 3.3402e-04, + 6.9714e-04, -4.1809e-03], + [ 3.8052e-03, 5.5046e-03, 2.8915e-03, ..., 1.2856e-03, + 5.5552e-04, 3.7518e-03], + ..., + [-6.1452e-05, 1.5373e-03, 3.7060e-03, ..., 2.8682e-04, + 1.8349e-03, -2.0084e-03], + [ 3.9902e-03, 3.5248e-03, 3.5667e-03, ..., 5.8365e-04, + 1.9836e-03, 5.3978e-03], + [-3.4294e-03, -6.7558e-03, -1.2604e-02, ..., -3.9935e-04, + -7.8812e-03, -6.3400e-03]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0189, -0.0037, 0.0034, 0.0176, -0.0089, -0.0036, 0.0097, 0.0275, + -0.0295, 0.0394], device='cuda:0'), grad: tensor([ 0.0361, -0.0335, 0.0375, -0.0420, 0.0509, 0.0103, -0.0154, -0.0120, + 0.0475, -0.0794], device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 216.43, cls_loss 0.5840 cls_loss_mapping 0.0144 cls_loss_causal 0.5501 re_mapping 0.0107 re_causal 0.0283 /// teacc 98.63 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0640, 0.0588, -0.0694, ..., -0.0795, -0.0573, -0.0050], + [-0.0394, -0.1017, 0.0074, ..., -0.0508, -0.0234, -0.0554], + [ 0.0251, -0.0692, 0.0391, ..., 0.1432, -0.0809, -0.0389], + ..., + [-0.0618, -0.1019, 0.0741, ..., -0.0179, -0.0131, 0.0611], + [ 0.0119, 0.0314, -0.0333, ..., -0.0677, -0.0880, -0.0041], + [-0.1136, -0.0159, -0.0094, ..., -0.1131, 0.0875, 0.0101]], + device='cuda:0'), grad: tensor([[ 1.2522e-03, 1.2379e-03, -2.1572e-03, ..., 7.3493e-05, + 1.8072e-04, -4.2877e-03], + [-2.3384e-03, -1.5745e-03, -1.7662e-03, ..., 1.7568e-05, + -4.0885e-07, -3.8261e-03], + [ 1.5697e-03, 1.4896e-03, 7.8678e-04, ..., -3.3951e-04, + 2.7075e-05, 1.5240e-03], + ..., + [ 4.3106e-03, 1.0490e-03, 4.9896e-03, ..., -2.1279e-04, + 2.2531e-04, 7.0229e-03], + [-2.4857e-02, -1.4511e-02, -8.8882e-03, ..., 3.8862e-05, + -3.6488e-03, -1.3466e-02], + [ 7.8201e-03, 3.1948e-03, 2.7809e-03, ..., 2.3127e-04, + 1.3437e-03, 4.6844e-03]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0187, -0.0038, 0.0026, 0.0176, -0.0083, -0.0027, 0.0096, 0.0280, + -0.0296, 0.0385], device='cuda:0'), grad: tensor([-0.0196, -0.0185, 0.0115, 0.0375, 0.0179, 0.0143, -0.0126, 0.0294, + -0.0658, 0.0059], device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 216.51, cls_loss 0.5511 cls_loss_mapping 0.0120 cls_loss_causal 0.5215 re_mapping 0.0105 re_causal 0.0261 /// teacc 98.62 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0627, 0.0600, -0.0704, ..., -0.0810, -0.0571, -0.0049], + [-0.0396, -0.1024, 0.0075, ..., -0.0515, -0.0238, -0.0558], + [ 0.0248, -0.0699, 0.0401, ..., 0.1435, -0.0811, -0.0373], + ..., + [-0.0633, -0.1027, 0.0744, ..., -0.0177, -0.0124, 0.0611], + [ 0.0123, 0.0318, -0.0349, ..., -0.0671, -0.0882, -0.0046], + [-0.1127, -0.0159, -0.0088, ..., -0.1127, 0.0881, 0.0104]], + device='cuda:0'), grad: tensor([[ 1.6928e-03, 2.1439e-03, 1.2541e-03, ..., 3.8838e-04, + 5.3501e-04, 1.3971e-03], + [ 3.0351e-04, 8.3065e-04, 5.3835e-04, ..., 5.7578e-05, + 2.9087e-04, 1.1081e-04], + [-1.8728e-04, 6.3658e-04, 7.1096e-04, ..., -6.4421e-04, + 4.2605e-04, 7.0143e-04], + ..., + [-1.1730e-03, -4.6539e-03, -5.9128e-03, ..., -4.4785e-03, + -5.4474e-03, -1.1797e-03], + [-1.2696e-05, 2.0180e-03, -2.1534e-03, ..., 6.5565e-04, + 5.5599e-04, -7.9498e-03], + [ 2.7103e-03, 2.3232e-03, 5.7869e-03, ..., 2.4948e-03, + 2.2984e-03, 6.4087e-03]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0194, -0.0038, 0.0035, 0.0178, -0.0094, -0.0033, 0.0096, 0.0271, + -0.0297, 0.0394], device='cuda:0'), grad: tensor([ 0.0156, 0.0008, 0.0060, 0.0109, 0.0169, -0.0088, -0.0340, -0.0481, + 0.0039, 0.0367], device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 216.37, cls_loss 0.5705 cls_loss_mapping 0.0125 cls_loss_causal 0.5433 re_mapping 0.0108 re_causal 0.0275 /// teacc 98.73 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0633, 0.0603, -0.0726, ..., -0.0814, -0.0580, -0.0064], + [-0.0403, -0.1022, 0.0090, ..., -0.0515, -0.0243, -0.0552], + [ 0.0253, -0.0706, 0.0404, ..., 0.1445, -0.0817, -0.0375], + ..., + [-0.0642, -0.1025, 0.0749, ..., -0.0180, -0.0117, 0.0617], + [ 0.0124, 0.0322, -0.0354, ..., -0.0684, -0.0888, -0.0052], + [-0.1120, -0.0160, -0.0093, ..., -0.1129, 0.0881, 0.0106]], + device='cuda:0'), grad: tensor([[ 7.5626e-04, 1.3933e-03, -7.3612e-05, ..., 6.6280e-04, + 4.5210e-05, 1.0958e-03], + [ 2.5196e-03, 3.8648e-04, -1.0424e-03, ..., 8.9312e-04, + 2.8133e-05, 3.4580e-03], + [ 1.0815e-03, 1.2617e-03, -1.6966e-03, ..., -4.8714e-03, + 1.6809e-04, 1.0242e-03], + ..., + [ 6.6185e-04, 1.6422e-03, 2.3937e-03, ..., 3.1161e-04, + 1.3268e-04, -3.3951e-03], + [ 1.3046e-03, -3.3951e-04, -2.0943e-03, ..., 6.3229e-04, + 6.9761e-04, 4.4899e-03], + [-2.4147e-03, 1.3418e-03, 2.4834e-03, ..., 4.7803e-04, + 1.0788e-04, -7.4806e-03]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0180, -0.0037, 0.0041, 0.0177, -0.0100, -0.0037, 0.0096, 0.0272, + -0.0295, 0.0409], device='cuda:0'), grad: tensor([-0.0134, -0.0012, 0.0021, 0.0465, -0.0323, -0.0216, 0.0269, -0.0025, + -0.0035, -0.0010], device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 216.37, cls_loss 0.5473 cls_loss_mapping 0.0119 cls_loss_causal 0.5111 re_mapping 0.0100 re_causal 0.0241 /// teacc 98.61 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0625, 0.0615, -0.0723, ..., -0.0818, -0.0576, -0.0053], + [-0.0408, -0.1034, 0.0092, ..., -0.0517, -0.0230, -0.0552], + [ 0.0251, -0.0710, 0.0400, ..., 0.1445, -0.0816, -0.0391], + ..., + [-0.0651, -0.1042, 0.0740, ..., -0.0189, -0.0128, 0.0617], + [ 0.0121, 0.0321, -0.0358, ..., -0.0689, -0.0884, -0.0053], + [-0.1128, -0.0163, -0.0077, ..., -0.1140, 0.0884, 0.0103]], + device='cuda:0'), grad: tensor([[-4.4346e-04, 5.2452e-05, -1.1826e-04, ..., -4.0531e-05, + 1.2338e-04, -2.2030e-03], + [-6.2332e-03, -5.1804e-03, -4.8752e-03, ..., -1.8024e-03, + -8.8990e-05, -5.7487e-03], + [ 2.7485e-03, 1.8663e-03, 1.4334e-03, ..., -6.1616e-06, + 9.5546e-05, 3.2272e-03], + ..., + [-1.2957e-05, -7.3051e-04, -1.0347e-03, ..., 3.9744e-04, + 3.0613e-04, -2.2831e-03], + [ 2.8324e-03, 2.5959e-03, 2.7466e-03, ..., 7.8440e-04, + 3.2330e-04, 4.0741e-03], + [ 1.8444e-03, 6.1321e-04, -4.4823e-03, ..., 3.1614e-04, + -5.2185e-03, -2.8019e-03]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0185, -0.0036, 0.0033, 0.0186, -0.0099, -0.0041, 0.0101, 0.0267, + -0.0298, 0.0410], device='cuda:0'), grad: tensor([-0.0064, -0.0563, 0.0241, -0.0567, 0.0464, 0.0202, 0.0256, -0.0096, + 0.0386, -0.0260], device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 216.66, cls_loss 0.5629 cls_loss_mapping 0.0154 cls_loss_causal 0.5415 re_mapping 0.0102 re_causal 0.0257 /// teacc 98.72 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0642, 0.0611, -0.0723, ..., -0.0815, -0.0574, -0.0063], + [-0.0404, -0.1024, 0.0081, ..., -0.0536, -0.0240, -0.0563], + [ 0.0244, -0.0724, 0.0405, ..., 0.1455, -0.0816, -0.0395], + ..., + [-0.0646, -0.1041, 0.0742, ..., -0.0193, -0.0134, 0.0619], + [ 0.0131, 0.0320, -0.0363, ..., -0.0682, -0.0884, -0.0043], + [-0.1141, -0.0158, -0.0076, ..., -0.1142, 0.0890, 0.0100]], + device='cuda:0'), grad: tensor([[ 1.6050e-03, -9.3102e-05, 6.0558e-04, ..., 1.7226e-04, + 2.6011e-04, 7.4387e-04], + [-3.7441e-03, -4.0817e-03, -2.3384e-03, ..., 5.4884e-04, + 7.6962e-04, -3.5934e-03], + [ 1.5652e-04, 1.3037e-03, 8.1730e-04, ..., -7.6723e-04, + 4.8923e-04, 1.5793e-03], + ..., + [ 1.0004e-03, 1.4048e-03, 2.1400e-03, ..., 1.3762e-03, + 1.5697e-03, 1.5574e-03], + [ 1.5211e-03, 1.9073e-03, -3.9101e-04, ..., 2.8324e-04, + 8.6594e-04, 1.3542e-04], + [-2.3746e-03, -3.2520e-03, -1.2169e-02, ..., -8.9798e-03, + -2.1835e-02, -3.4885e-03]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0181, -0.0047, 0.0037, 0.0183, -0.0088, -0.0048, 0.0103, 0.0265, + -0.0293, 0.0412], device='cuda:0'), grad: tensor([ 0.0133, -0.0463, 0.0186, 0.0153, 0.0345, 0.0213, -0.0081, 0.0183, + -0.0067, -0.0602], device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 216.37, cls_loss 0.5746 cls_loss_mapping 0.0136 cls_loss_causal 0.5504 re_mapping 0.0106 re_causal 0.0263 /// teacc 98.59 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0650, 0.0608, -0.0721, ..., -0.0817, -0.0574, -0.0065], + [-0.0397, -0.1021, 0.0068, ..., -0.0539, -0.0240, -0.0564], + [ 0.0252, -0.0731, 0.0409, ..., 0.1463, -0.0817, -0.0398], + ..., + [-0.0658, -0.1038, 0.0743, ..., -0.0186, -0.0144, 0.0637], + [ 0.0130, 0.0319, -0.0349, ..., -0.0678, -0.0875, -0.0049], + [-0.1147, -0.0161, -0.0073, ..., -0.1152, 0.0893, 0.0096]], + device='cuda:0'), grad: tensor([[-0.0217, -0.0273, 0.0010, ..., 0.0003, 0.0004, 0.0004], + [-0.0015, -0.0007, -0.0010, ..., 0.0006, 0.0007, -0.0002], + [-0.0087, 0.0007, -0.0015, ..., -0.0040, -0.0028, 0.0017], + ..., + [-0.0013, -0.0003, -0.0028, ..., 0.0004, -0.0016, -0.0029], + [ 0.0252, 0.0205, 0.0036, ..., 0.0044, 0.0028, 0.0015], + [ 0.0022, 0.0017, 0.0034, ..., 0.0008, 0.0017, 0.0020]], + device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0179, -0.0050, 0.0037, 0.0180, -0.0090, -0.0047, 0.0098, 0.0271, + -0.0292, 0.0418], device='cuda:0'), grad: tensor([-0.0486, -0.0195, -0.0147, -0.0073, -0.0107, 0.0154, 0.0102, -0.0128, + 0.0640, 0.0240], device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 216.47, cls_loss 0.5655 cls_loss_mapping 0.0166 cls_loss_causal 0.5395 re_mapping 0.0097 re_causal 0.0228 /// teacc 98.69 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0650, 0.0605, -0.0720, ..., -0.0826, -0.0587, -0.0055], + [-0.0380, -0.1022, 0.0067, ..., -0.0547, -0.0254, -0.0567], + [ 0.0240, -0.0734, 0.0408, ..., 0.1460, -0.0820, -0.0406], + ..., + [-0.0653, -0.1023, 0.0748, ..., -0.0184, -0.0140, 0.0632], + [ 0.0128, 0.0318, -0.0348, ..., -0.0674, -0.0886, -0.0053], + [-0.1151, -0.0161, -0.0084, ..., -0.1160, 0.0899, 0.0093]], + device='cuda:0'), grad: tensor([[ 2.0087e-04, -9.3997e-05, -1.7033e-03, ..., 1.4544e-04, + -1.6257e-05, -2.7771e-03], + [-3.6073e-04, 5.3465e-05, -1.8921e-03, ..., 1.9625e-05, + 1.1832e-05, -1.5450e-03], + [ 6.8092e-04, 3.8409e-04, 1.5221e-03, ..., 1.9461e-05, + 1.0198e-04, 1.7157e-03], + ..., + [ 2.5916e-04, 7.9215e-05, 2.8858e-03, ..., 9.2015e-06, + 1.1101e-03, 2.7962e-03], + [ 3.3646e-03, 2.1267e-03, 2.0161e-03, ..., 8.6260e-04, + 4.6611e-04, 1.3199e-03], + [ 5.1355e-04, 2.6774e-04, -6.0768e-03, ..., -9.6436e-03, + 2.0123e-03, 2.8934e-03]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0183, -0.0043, 0.0040, 0.0183, -0.0088, -0.0051, 0.0095, 0.0269, + -0.0287, 0.0404], device='cuda:0'), grad: tensor([-0.0173, -0.0196, 0.0169, 0.0064, -0.0199, -0.0511, 0.0570, 0.0212, + 0.0178, -0.0114], device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 216.33, cls_loss 0.6068 cls_loss_mapping 0.0126 cls_loss_causal 0.5726 re_mapping 0.0103 re_causal 0.0252 /// teacc 98.71 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0655, 0.0601, -0.0719, ..., -0.0838, -0.0580, -0.0054], + [-0.0373, -0.1025, 0.0075, ..., -0.0543, -0.0246, -0.0553], + [ 0.0246, -0.0733, 0.0407, ..., 0.1459, -0.0829, -0.0407], + ..., + [-0.0666, -0.1039, 0.0745, ..., -0.0165, -0.0150, 0.0627], + [ 0.0131, 0.0323, -0.0344, ..., -0.0677, -0.0891, -0.0048], + [-0.1155, -0.0159, -0.0086, ..., -0.1182, 0.0896, 0.0082]], + device='cuda:0'), grad: tensor([[ 1.8921e-03, 1.6031e-03, 1.3323e-03, ..., 1.2264e-03, + 6.1083e-04, 1.9293e-03], + [ 3.5024e-04, 1.5247e-04, -1.1101e-03, ..., -1.0500e-03, + 4.2677e-05, 9.8991e-04], + [-6.9046e-03, -6.9733e-03, -1.5106e-03, ..., -1.6155e-03, + -2.3327e-03, -4.5509e-03], + ..., + [ 6.0654e-04, 9.7752e-04, -1.0548e-03, ..., 4.6873e-04, + 7.8583e-04, -1.4782e-03], + [ 9.8133e-04, 1.4992e-03, 2.9240e-03, ..., 1.0700e-03, + 9.6655e-04, 2.5997e-03], + [ 5.8413e-04, -2.0561e-03, -1.9093e-03, ..., 8.3780e-04, + -2.0542e-03, -9.2125e-04]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0181, -0.0030, 0.0043, 0.0174, -0.0080, -0.0049, 0.0092, 0.0258, + -0.0284, 0.0400], device='cuda:0'), grad: tensor([ 0.0137, -0.0025, -0.0313, -0.0158, -0.0162, 0.0109, 0.0138, 0.0093, + 0.0187, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 216.32, cls_loss 0.5845 cls_loss_mapping 0.0157 cls_loss_causal 0.5565 re_mapping 0.0098 re_causal 0.0234 /// teacc 98.56 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0660, 0.0602, -0.0711, ..., -0.0842, -0.0593, -0.0058], + [-0.0388, -0.1034, 0.0072, ..., -0.0542, -0.0237, -0.0564], + [ 0.0243, -0.0738, 0.0403, ..., 0.1455, -0.0841, -0.0417], + ..., + [-0.0648, -0.1032, 0.0744, ..., -0.0159, -0.0152, 0.0629], + [ 0.0128, 0.0329, -0.0349, ..., -0.0692, -0.0888, -0.0043], + [-0.1151, -0.0151, -0.0076, ..., -0.1172, 0.0906, 0.0093]], + device='cuda:0'), grad: tensor([[ 5.4312e-04, 3.2276e-05, -1.5841e-03, ..., 5.0116e-04, + 1.8394e-04, 1.0471e-03], + [ 4.0984e-04, 2.2042e-04, 1.0862e-03, ..., 3.1662e-04, + 1.9443e-04, 1.0519e-03], + [ 2.0218e-03, 3.3426e-04, -1.0562e-04, ..., -3.3703e-03, + 3.2330e-04, 2.3918e-03], + ..., + [ 9.9182e-04, 3.6573e-04, 1.2894e-03, ..., 5.0497e-04, + 4.3011e-04, 1.8892e-03], + [-4.3526e-03, -2.5082e-03, 2.2869e-03, ..., 3.0441e-03, + 4.1223e-04, 1.4791e-03], + [ 3.7479e-04, 1.9038e-04, -2.5005e-03, ..., 8.6248e-05, + -9.8228e-04, -3.0613e-03]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0180, -0.0036, 0.0034, 0.0176, -0.0080, -0.0042, 0.0091, 0.0264, + -0.0285, 0.0403], device='cuda:0'), grad: tensor([-0.0206, 0.0102, 0.0124, -0.0015, 0.0090, -0.0173, 0.0093, 0.0126, + 0.0082, -0.0221], device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 216.23, cls_loss 0.5706 cls_loss_mapping 0.0145 cls_loss_causal 0.5414 re_mapping 0.0098 re_causal 0.0227 /// teacc 98.72 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0649, 0.0601, -0.0719, ..., -0.0848, -0.0622, -0.0059], + [-0.0397, -0.1037, 0.0074, ..., -0.0555, -0.0243, -0.0568], + [ 0.0250, -0.0741, 0.0407, ..., 0.1453, -0.0857, -0.0409], + ..., + [-0.0646, -0.1025, 0.0739, ..., -0.0162, -0.0151, 0.0616], + [ 0.0125, 0.0333, -0.0347, ..., -0.0689, -0.0888, -0.0048], + [-0.1144, -0.0152, -0.0076, ..., -0.1158, 0.0909, 0.0101]], + device='cuda:0'), grad: tensor([[-3.0174e-03, -4.0054e-03, -2.4624e-03, ..., -2.1112e-04, + -2.5821e-04, -3.7060e-03], + [ 1.1101e-03, 1.1873e-03, 8.5115e-05, ..., -1.5581e-04, + 1.0979e-04, 1.3285e-03], + [-3.1261e-03, -1.6432e-03, -4.1656e-03, ..., -3.9291e-03, + -6.6519e-05, -1.7624e-03], + ..., + [ 1.0519e-03, 1.5135e-03, -1.4677e-03, ..., 4.7207e-04, + -1.1082e-03, -4.2439e-04], + [-3.0479e-03, -3.2101e-03, 9.2649e-04, ..., 9.8896e-04, + 1.9634e-04, 2.4211e-04], + [ 3.6144e-03, 4.2305e-03, 3.7441e-03, ..., 1.5478e-03, + 1.3065e-03, 3.8357e-03]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0178, -0.0034, 0.0039, 0.0172, -0.0079, -0.0044, 0.0093, 0.0263, + -0.0285, 0.0402], device='cuda:0'), grad: tensor([-0.0352, 0.0173, -0.0282, 0.0289, -0.0380, 0.0131, -0.0075, 0.0095, + -0.0013, 0.0416], device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 216.25, cls_loss 0.5656 cls_loss_mapping 0.0148 cls_loss_causal 0.5374 re_mapping 0.0104 re_causal 0.0253 /// teacc 98.56 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0636, 0.0614, -0.0730, ..., -0.0842, -0.0629, -0.0074], + [-0.0397, -0.1039, 0.0072, ..., -0.0559, -0.0252, -0.0564], + [ 0.0244, -0.0750, 0.0400, ..., 0.1450, -0.0871, -0.0412], + ..., + [-0.0658, -0.1029, 0.0745, ..., -0.0163, -0.0144, 0.0619], + [ 0.0126, 0.0337, -0.0344, ..., -0.0692, -0.0891, -0.0061], + [-0.1143, -0.0154, -0.0086, ..., -0.1164, 0.0908, 0.0107]], + device='cuda:0'), grad: tensor([[ 5.6953e-03, 7.9651e-03, 1.4935e-03, ..., 7.8812e-03, + 3.2753e-05, 1.7223e-03], + [ 1.4839e-03, 9.2745e-04, 1.8988e-03, ..., 9.3174e-04, + 9.5308e-05, 2.2507e-03], + [ 3.5839e-03, -1.5993e-03, -1.5163e-03, ..., 2.6684e-03, + 5.7757e-05, -3.8815e-04], + ..., + [ 4.3488e-03, 1.1530e-03, 8.7891e-03, ..., 9.0866e-03, + 1.7595e-03, 6.8398e-03], + [ 4.2419e-03, 1.8167e-03, -2.4300e-03, ..., 1.0624e-03, + -3.0637e-05, -6.3896e-04], + [ 7.4120e-03, 1.2789e-03, 6.9618e-04, ..., 5.6686e-03, + 4.1080e-04, 2.1381e-03]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0173, -0.0040, 0.0039, 0.0180, -0.0078, -0.0042, 0.0095, 0.0266, + -0.0292, 0.0402], device='cuda:0'), grad: tensor([ 0.0505, 0.0272, -0.0178, -0.0737, -0.0546, 0.0384, -0.0158, 0.0495, + -0.0311, 0.0273], device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 216.36, cls_loss 0.5639 cls_loss_mapping 0.0138 cls_loss_causal 0.5343 re_mapping 0.0099 re_causal 0.0235 /// teacc 98.68 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0634, 0.0613, -0.0731, ..., -0.0863, -0.0622, -0.0064], + [-0.0403, -0.1050, 0.0078, ..., -0.0547, -0.0258, -0.0560], + [ 0.0255, -0.0745, 0.0409, ..., 0.1468, -0.0870, -0.0411], + ..., + [-0.0668, -0.1033, 0.0741, ..., -0.0169, -0.0150, 0.0619], + [ 0.0126, 0.0329, -0.0351, ..., -0.0707, -0.0889, -0.0064], + [-0.1135, -0.0152, -0.0082, ..., -0.1161, 0.0910, 0.0110]], + device='cuda:0'), grad: tensor([[ 2.3365e-03, 2.7065e-03, 1.5078e-03, ..., 4.0126e-04, + 3.9387e-04, 8.4066e-04], + [ 8.1015e-04, 3.1304e-04, 1.8864e-03, ..., 1.0014e-03, + 7.3290e-04, 1.4105e-03], + [ 6.2981e-03, 6.4230e-04, 6.9847e-03, ..., 6.2752e-03, + 1.4982e-03, 3.2482e-03], + ..., + [ 3.9840e-04, 3.8892e-05, 1.2314e-02, ..., 1.3786e-02, + 3.6964e-03, 4.6539e-03], + [-6.1264e-03, 4.1270e-04, 9.0218e-04, ..., 6.2704e-04, + 7.1716e-04, 8.5545e-04], + [-2.4319e-03, 4.0317e-04, -8.0414e-03, ..., -1.3094e-03, + -6.3934e-03, -9.5320e-04]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0178, -0.0029, 0.0038, 0.0175, -0.0078, -0.0036, 0.0088, 0.0260, + -0.0301, 0.0410], device='cuda:0'), grad: tensor([ 0.0210, 0.0204, 0.0298, 0.0158, -0.0248, -0.0343, -0.0099, 0.0014, + 0.0029, -0.0223], device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 216.17, cls_loss 0.5763 cls_loss_mapping 0.0140 cls_loss_causal 0.5520 re_mapping 0.0100 re_causal 0.0255 /// teacc 98.60 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0631, 0.0612, -0.0732, ..., -0.0851, -0.0622, -0.0068], + [-0.0403, -0.1053, 0.0074, ..., -0.0554, -0.0268, -0.0557], + [ 0.0264, -0.0745, 0.0408, ..., 0.1469, -0.0888, -0.0415], + ..., + [-0.0676, -0.1043, 0.0744, ..., -0.0159, -0.0160, 0.0633], + [ 0.0130, 0.0328, -0.0350, ..., -0.0716, -0.0880, -0.0066], + [-0.1148, -0.0152, -0.0082, ..., -0.1161, 0.0911, 0.0106]], + device='cuda:0'), grad: tensor([[ 7.8964e-04, 5.1212e-04, 1.6832e-03, ..., 3.2926e-04, + 5.8031e-04, 1.0433e-03], + [-1.9798e-03, -3.7694e-04, -5.4741e-03, ..., -1.3494e-04, + -4.5633e-04, -1.5917e-03], + [ 8.0299e-04, 3.9029e-04, 9.5415e-04, ..., -5.4646e-04, + 4.6563e-04, 8.1301e-04], + ..., + [-2.3985e-04, 6.1035e-04, 9.5010e-05, ..., -3.8385e-04, + 5.4884e-04, -1.2074e-03], + [ 1.9913e-03, -6.7139e-04, 1.0834e-03, ..., 1.0653e-03, + -3.3283e-04, 4.3907e-03], + [-1.7490e-03, -2.1400e-03, -6.1913e-03, ..., -1.7433e-03, + -2.3422e-03, -6.8359e-03]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0181, -0.0025, 0.0038, 0.0169, -0.0072, -0.0034, 0.0076, 0.0259, + -0.0296, 0.0409], device='cuda:0'), grad: tensor([ 0.0154, -0.0699, 0.0120, 0.0104, 0.0227, 0.0106, 0.0111, -0.0070, + 0.0124, -0.0177], device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 216.59, cls_loss 0.6004 cls_loss_mapping 0.0131 cls_loss_causal 0.5679 re_mapping 0.0099 re_causal 0.0242 /// teacc 98.59 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0636, 0.0612, -0.0729, ..., -0.0846, -0.0622, -0.0067], + [-0.0409, -0.1057, 0.0070, ..., -0.0572, -0.0285, -0.0561], + [ 0.0258, -0.0760, 0.0400, ..., 0.1468, -0.0898, -0.0414], + ..., + [-0.0670, -0.1051, 0.0746, ..., -0.0153, -0.0169, 0.0635], + [ 0.0130, 0.0325, -0.0350, ..., -0.0721, -0.0885, -0.0073], + [-0.1149, -0.0152, -0.0078, ..., -0.1158, 0.0915, 0.0109]], + device='cuda:0'), grad: tensor([[ 6.7329e-04, 1.8847e-04, 1.5235e-04, ..., 1.4257e-04, + 1.6165e-04, 1.3809e-03], + [ 4.1847e-03, 1.2121e-03, 5.0783e-04, ..., 4.5609e-04, + 3.6311e-04, 3.4676e-03], + [-1.4286e-03, 5.3549e-04, 2.6989e-04, ..., -3.4547e-04, + -6.6996e-04, 6.9365e-06], + ..., + [ 1.4963e-03, 6.2323e-04, 7.5245e-04, ..., 9.9277e-04, + 1.1034e-03, 3.6011e-03], + [-4.0192e-02, -2.1515e-02, 2.0301e-04, ..., -2.0676e-03, + -7.4768e-03, -1.2093e-03], + [ 7.2823e-03, 8.3313e-03, -2.5215e-03, ..., -2.1629e-03, + 7.7324e-03, -1.3695e-03]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0174, -0.0029, 0.0038, 0.0183, -0.0074, -0.0046, 0.0091, 0.0267, + -0.0296, 0.0396], device='cuda:0'), grad: tensor([ 0.0060, 0.0266, -0.0201, 0.0244, -0.0165, -0.0054, 0.0294, 0.0153, + -0.0722, 0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 216.40, cls_loss 0.5361 cls_loss_mapping 0.0163 cls_loss_causal 0.5118 re_mapping 0.0101 re_causal 0.0239 /// teacc 98.67 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0642, 0.0608, -0.0734, ..., -0.0852, -0.0637, -0.0061], + [-0.0418, -0.1064, 0.0066, ..., -0.0570, -0.0282, -0.0571], + [ 0.0252, -0.0765, 0.0404, ..., 0.1471, -0.0906, -0.0417], + ..., + [-0.0672, -0.1063, 0.0748, ..., -0.0158, -0.0181, 0.0643], + [ 0.0129, 0.0325, -0.0354, ..., -0.0732, -0.0876, -0.0068], + [-0.1139, -0.0159, -0.0079, ..., -0.1160, 0.0918, 0.0107]], + device='cuda:0'), grad: tensor([[ 1.2703e-05, -3.9444e-03, 5.7459e-04, ..., 4.2725e-04, + 2.3997e-04, 4.0841e-04], + [ 1.1492e-03, -8.5652e-05, 3.5882e-04, ..., 6.6519e-04, + -7.9250e-04, -1.9855e-03], + [ 4.7302e-04, 5.0974e-04, -7.8106e-04, ..., -3.4122e-03, + 4.9448e-04, 2.6226e-03], + ..., + [-1.2493e-03, 2.7943e-04, -1.2703e-03, ..., -2.2736e-03, + 1.6022e-03, -2.8229e-03], + [ 1.6356e-04, 1.0490e-03, 6.7663e-04, ..., 5.2404e-04, + 6.7663e-04, 2.1400e-03], + [ 2.1446e-04, 8.3160e-04, 2.5902e-03, ..., 8.3923e-04, + 3.1548e-03, 3.7918e-03]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0183, -0.0025, 0.0027, 0.0181, -0.0070, -0.0051, 0.0092, 0.0263, + -0.0292, 0.0395], device='cuda:0'), grad: tensor([ 0.0142, -0.0208, 0.0012, 0.0330, -0.0688, 0.0211, 0.0396, -0.0483, + 0.0276, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 216.47, cls_loss 0.5318 cls_loss_mapping 0.0088 cls_loss_causal 0.5048 re_mapping 0.0098 re_causal 0.0248 /// teacc 98.60 lr 0.00010000 +Epoch 141, weight, value: tensor([[-0.0647, 0.0612, -0.0736, ..., -0.0844, -0.0630, -0.0055], + [-0.0428, -0.1082, 0.0070, ..., -0.0564, -0.0278, -0.0577], + [ 0.0256, -0.0766, 0.0404, ..., 0.1470, -0.0912, -0.0412], + ..., + [-0.0675, -0.1068, 0.0747, ..., -0.0171, -0.0184, 0.0644], + [ 0.0138, 0.0332, -0.0353, ..., -0.0741, -0.0879, -0.0078], + [-0.1143, -0.0161, -0.0081, ..., -0.1172, 0.0919, 0.0115]], + device='cuda:0'), grad: tensor([[ 1.7204e-03, 3.1605e-03, 6.3801e-04, ..., 4.2208e-06, + 7.2336e-04, 8.0013e-04], + [-5.3740e-04, -4.8232e-04, -1.9848e-04, ..., -3.3069e-04, + 2.1291e-04, -1.7691e-04], + [ 2.2233e-04, 3.4094e-04, 7.8201e-04, ..., 2.2392e-03, + -9.1410e-04, 1.3609e-03], + ..., + [ 5.4502e-04, 1.1766e-04, 2.5415e-04, ..., 1.6475e-04, + 8.6308e-05, 5.6553e-04], + [ 1.2497e-02, 1.2672e-02, 1.1482e-03, ..., 9.2447e-05, + 2.6035e-04, 5.7077e-04], + [ 5.6314e-04, -5.9557e-04, 6.3814e-06, ..., 1.3316e-04, + -1.1012e-05, 5.2595e-04]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0184, -0.0028, 0.0029, 0.0176, -0.0070, -0.0048, 0.0101, 0.0259, + -0.0295, 0.0397], device='cuda:0'), grad: tensor([ 0.0193, -0.0301, -0.0113, 0.0131, 0.0117, -0.0068, -0.0566, 0.0142, + 0.0630, -0.0164], device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 216.56, cls_loss 0.5360 cls_loss_mapping 0.0114 cls_loss_causal 0.5033 re_mapping 0.0090 re_causal 0.0225 /// teacc 98.44 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.0651, 0.0611, -0.0740, ..., -0.0853, -0.0626, -0.0062], + [-0.0426, -0.1094, 0.0067, ..., -0.0562, -0.0280, -0.0575], + [ 0.0255, -0.0776, 0.0402, ..., 0.1470, -0.0919, -0.0417], + ..., + [-0.0683, -0.1064, 0.0755, ..., -0.0179, -0.0170, 0.0647], + [ 0.0146, 0.0323, -0.0351, ..., -0.0732, -0.0888, -0.0080], + [-0.1148, -0.0151, -0.0084, ..., -0.1165, 0.0924, 0.0117]], + device='cuda:0'), grad: tensor([[ 7.7772e-04, 3.3684e-03, 4.7183e-04, ..., 1.8346e-04, + 2.2709e-05, 2.2392e-03], + [-1.1292e-03, 6.2084e-04, -4.2725e-04, ..., 2.4772e-04, + 3.2187e-05, 5.2214e-04], + [ 4.6611e-04, 4.1199e-04, 7.1716e-04, ..., 1.3304e-04, + 4.2468e-05, 2.7542e-03], + ..., + [-4.9925e-04, -3.6430e-04, -2.1660e-04, ..., -4.6206e-04, + 4.9859e-05, -4.9706e-03], + [ 3.0732e-04, 4.9639e-04, 4.0722e-04, ..., 7.4625e-05, + 2.6017e-05, 1.2264e-03], + [ 3.0422e-04, 2.7275e-04, -2.2335e-03, ..., -8.4829e-04, + -2.9659e-04, -5.6686e-03]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0182, -0.0025, 0.0033, 0.0171, -0.0066, -0.0050, 0.0092, 0.0263, + -0.0288, 0.0393], device='cuda:0'), grad: tensor([ 0.0334, -0.0128, 0.0253, 0.0193, 0.0183, -0.0157, -0.0237, -0.0451, + 0.0176, -0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 218.07, cls_loss 0.5029 cls_loss_mapping 0.0087 cls_loss_causal 0.4719 re_mapping 0.0099 re_causal 0.0248 /// teacc 98.77 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0654, 0.0604, -0.0736, ..., -0.0849, -0.0625, -0.0040], + [-0.0429, -0.1097, 0.0075, ..., -0.0555, -0.0279, -0.0572], + [ 0.0257, -0.0786, 0.0404, ..., 0.1477, -0.0905, -0.0419], + ..., + [-0.0681, -0.1057, 0.0759, ..., -0.0174, -0.0169, 0.0642], + [ 0.0141, 0.0326, -0.0353, ..., -0.0736, -0.0888, -0.0071], + [-0.1158, -0.0160, -0.0082, ..., -0.1175, 0.0918, 0.0115]], + device='cuda:0'), grad: tensor([[-8.3771e-03, -7.5531e-03, -2.6474e-03, ..., -2.4071e-03, + 1.8072e-04, 4.5514e-04], + [ 2.4486e-04, 1.3161e-04, 7.9727e-04, ..., 1.6475e-04, + 1.3268e-04, 1.4248e-03], + [-3.4294e-03, 2.2995e-04, 8.6880e-04, ..., 6.9499e-05, + 2.2745e-04, -1.3056e-03], + ..., + [ 2.4891e-03, 9.8705e-05, 4.7379e-03, ..., -1.6248e-04, + 1.5650e-03, 2.5101e-03], + [ 1.0128e-03, 1.0985e-04, 8.5974e-04, ..., 1.4976e-05, + 1.3793e-04, 1.3380e-03], + [ 1.0796e-03, 1.2589e-04, -5.8937e-04, ..., 6.1035e-05, + -2.0809e-03, 6.7406e-03]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0191, -0.0019, 0.0030, 0.0169, -0.0068, -0.0054, 0.0096, 0.0260, + -0.0295, 0.0394], device='cuda:0'), grad: tensor([-0.0251, 0.0068, -0.0210, 0.0061, 0.0072, -0.0142, -0.0116, 0.0265, + 0.0109, 0.0143], device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 218.52, cls_loss 0.5548 cls_loss_mapping 0.0123 cls_loss_causal 0.5266 re_mapping 0.0096 re_causal 0.0244 /// teacc 98.68 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0657, 0.0610, -0.0744, ..., -0.0861, -0.0627, -0.0040], + [-0.0426, -0.1099, 0.0080, ..., -0.0543, -0.0287, -0.0572], + [ 0.0263, -0.0801, 0.0401, ..., 0.1467, -0.0901, -0.0417], + ..., + [-0.0695, -0.1069, 0.0756, ..., -0.0169, -0.0178, 0.0636], + [ 0.0144, 0.0323, -0.0345, ..., -0.0719, -0.0880, -0.0067], + [-0.1158, -0.0145, -0.0084, ..., -0.1193, 0.0916, 0.0109]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0003, 0.0006, ..., 0.0004, 0.0005, 0.0017], + [-0.0006, 0.0003, 0.0003, ..., -0.0005, 0.0008, 0.0019], + [ 0.0033, 0.0017, 0.0017, ..., 0.0038, 0.0006, 0.0020], + ..., + [ 0.0001, 0.0002, 0.0008, ..., 0.0009, 0.0007, 0.0016], + [-0.0053, -0.0027, 0.0002, ..., -0.0015, 0.0014, -0.0023], + [ 0.0004, 0.0005, 0.0010, ..., 0.0006, 0.0011, 0.0023]], + device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0195, -0.0024, 0.0033, 0.0165, -0.0073, -0.0050, 0.0104, 0.0258, + -0.0298, 0.0397], device='cuda:0'), grad: tensor([ 0.0170, -0.0115, 0.0323, 0.0164, -0.0358, -0.0151, -0.0081, 0.0013, + -0.0180, 0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 218.36, cls_loss 0.5661 cls_loss_mapping 0.0125 cls_loss_causal 0.5327 re_mapping 0.0097 re_causal 0.0240 /// teacc 98.65 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0659, 0.0611, -0.0760, ..., -0.0864, -0.0634, -0.0027], + [-0.0428, -0.1100, 0.0075, ..., -0.0555, -0.0286, -0.0578], + [ 0.0262, -0.0813, 0.0396, ..., 0.1464, -0.0906, -0.0416], + ..., + [-0.0690, -0.1069, 0.0760, ..., -0.0174, -0.0188, 0.0640], + [ 0.0142, 0.0320, -0.0338, ..., -0.0714, -0.0881, -0.0062], + [-0.1163, -0.0145, -0.0075, ..., -0.1191, 0.0926, 0.0110]], + device='cuda:0'), grad: tensor([[ 1.2512e-03, 2.7275e-03, 4.4518e-03, ..., 2.6627e-03, + 3.4627e-06, 7.1669e-04], + [-7.7581e-04, -6.5947e-04, 1.7655e-04, ..., -5.7983e-04, + 6.2846e-06, 1.4420e-03], + [ 2.8205e-04, 1.1349e-03, 4.0741e-03, ..., 3.1757e-03, + 4.4554e-06, 2.3956e-03], + ..., + [-6.8998e-04, -7.1678e-03, -1.0767e-03, ..., -3.5744e-03, + 7.2050e-04, 7.4081e-03], + [ 2.0027e-03, 2.0046e-03, 2.0866e-03, ..., 1.6890e-03, + 2.7156e-04, 9.0265e-04], + [ 2.1565e-04, 4.0174e-04, -6.1111e-03, ..., -2.5997e-03, + 2.1048e-07, 1.5059e-03]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0193, -0.0019, 0.0037, 0.0160, -0.0086, -0.0051, 0.0107, 0.0260, + -0.0294, 0.0400], device='cuda:0'), grad: tensor([ 0.0202, -0.0084, -0.0043, -0.0188, 0.0076, -0.0156, 0.0009, -0.0047, + 0.0218, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 220.29, cls_loss 0.5604 cls_loss_mapping 0.0093 cls_loss_causal 0.5360 re_mapping 0.0098 re_causal 0.0243 /// teacc 98.68 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0660, 0.0621, -0.0761, ..., -0.0863, -0.0633, -0.0019], + [-0.0430, -0.1098, 0.0075, ..., -0.0545, -0.0274, -0.0583], + [ 0.0264, -0.0820, 0.0400, ..., 0.1456, -0.0900, -0.0416], + ..., + [-0.0694, -0.1069, 0.0759, ..., -0.0168, -0.0202, 0.0631], + [ 0.0141, 0.0316, -0.0324, ..., -0.0713, -0.0872, -0.0055], + [-0.1166, -0.0153, -0.0076, ..., -0.1190, 0.0921, 0.0106]], + device='cuda:0'), grad: tensor([[ 2.5578e-03, 1.8206e-03, 5.5742e-04, ..., 1.2434e-04, + 5.4091e-06, 1.4763e-03], + [ 1.9379e-03, 7.7391e-04, 8.9502e-04, ..., 9.2924e-05, + 1.8179e-06, 1.0262e-03], + [-8.3876e-04, -1.8895e-05, -2.1152e-03, ..., -1.9646e-03, + 1.8075e-05, -2.1529e-04], + ..., + [ 1.7405e-03, 6.3133e-04, 3.1033e-03, ..., 3.9649e-04, + 1.1950e-03, 1.1396e-03], + [ 8.4229e-03, 2.0485e-03, 5.0354e-03, ..., 5.3139e-03, + 1.4257e-03, 2.1229e-03], + [-2.2602e-03, 1.2064e-03, -6.1951e-03, ..., 8.8930e-04, + -3.0994e-03, -4.6425e-03]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0196, -0.0025, 0.0037, 0.0165, -0.0080, -0.0058, 0.0098, 0.0258, + -0.0294, 0.0409], device='cuda:0'), grad: tensor([ 0.0154, 0.0181, -0.0110, 0.0033, 0.0117, 0.0219, -0.0421, 0.0197, + 0.0152, -0.0523], device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 218.22, cls_loss 0.5253 cls_loss_mapping 0.0085 cls_loss_causal 0.5006 re_mapping 0.0094 re_causal 0.0245 /// teacc 98.72 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0655, 0.0627, -0.0768, ..., -0.0867, -0.0646, -0.0026], + [-0.0432, -0.1097, 0.0074, ..., -0.0547, -0.0275, -0.0589], + [ 0.0258, -0.0818, 0.0399, ..., 0.1459, -0.0905, -0.0405], + ..., + [-0.0688, -0.1076, 0.0758, ..., -0.0172, -0.0199, 0.0626], + [ 0.0145, 0.0315, -0.0332, ..., -0.0716, -0.0876, -0.0060], + [-0.1172, -0.0146, -0.0073, ..., -0.1202, 0.0924, 0.0106]], + device='cuda:0'), grad: tensor([[ 6.4564e-04, 1.7905e-04, 6.2227e-04, ..., 3.6788e-04, + 1.2624e-04, 7.6580e-04], + [ 6.9761e-04, 1.2600e-04, 9.4461e-04, ..., 5.4121e-04, + 2.3689e-03, 4.4022e-03], + [ 2.5387e-03, 1.8835e-04, 4.3068e-03, ..., 1.8291e-03, + 9.3460e-05, 3.3035e-03], + ..., + [-1.9188e-03, -2.4548e-03, -1.2436e-03, ..., 4.5371e-04, + 5.7507e-04, -2.4242e-03], + [ 1.7891e-03, 1.0376e-03, 1.0567e-03, ..., 4.3154e-04, + 3.7527e-04, 1.0433e-03], + [-1.1663e-03, 2.7924e-03, -7.9727e-03, ..., -3.1204e-03, + 3.5000e-04, 1.0605e-03]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0187, -0.0025, 0.0030, 0.0165, -0.0064, -0.0057, 0.0097, 0.0258, + -0.0295, 0.0409], device='cuda:0'), grad: tensor([ 0.0064, 0.0248, 0.0278, -0.0104, -0.0236, -0.0105, 0.0212, -0.0491, + 0.0115, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 146---------------------------------------------------- +epoch 146, time 220.63, cls_loss 0.5374 cls_loss_mapping 0.0096 cls_loss_causal 0.5066 re_mapping 0.0096 re_causal 0.0245 /// teacc 98.84 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0653, 0.0629, -0.0776, ..., -0.0863, -0.0648, -0.0017], + [-0.0445, -0.1108, 0.0072, ..., -0.0546, -0.0286, -0.0600], + [ 0.0260, -0.0809, 0.0408, ..., 0.1459, -0.0922, -0.0398], + ..., + [-0.0700, -0.1082, 0.0765, ..., -0.0171, -0.0198, 0.0645], + [ 0.0142, 0.0310, -0.0337, ..., -0.0720, -0.0880, -0.0077], + [-0.1171, -0.0149, -0.0075, ..., -0.1212, 0.0931, 0.0111]], + device='cuda:0'), grad: tensor([[ 0.0038, 0.0039, -0.0011, ..., 0.0019, -0.0083, -0.0066], + [-0.0071, -0.0006, -0.0013, ..., 0.0010, -0.0038, -0.0092], + [ 0.0024, 0.0006, -0.0011, ..., -0.0039, 0.0016, 0.0025], + ..., + [ 0.0038, 0.0009, 0.0059, ..., 0.0032, 0.0039, 0.0042], + [-0.0029, -0.0037, 0.0006, ..., -0.0016, 0.0031, -0.0043], + [ 0.0015, 0.0004, 0.0011, ..., 0.0009, 0.0016, 0.0022]], + device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0187, -0.0034, 0.0027, 0.0166, -0.0064, -0.0049, 0.0103, 0.0255, + -0.0292, 0.0408], device='cuda:0'), grad: tensor([ 0.0025, -0.0596, 0.0144, 0.0194, 0.0453, -0.0033, -0.0281, 0.0417, + -0.0251, -0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 218.24, cls_loss 0.5371 cls_loss_mapping 0.0107 cls_loss_causal 0.5029 re_mapping 0.0095 re_causal 0.0231 /// teacc 98.70 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0661, 0.0633, -0.0769, ..., -0.0853, -0.0633, -0.0024], + [-0.0453, -0.1099, 0.0065, ..., -0.0556, -0.0290, -0.0600], + [ 0.0254, -0.0818, 0.0401, ..., 0.1444, -0.0933, -0.0407], + ..., + [-0.0700, -0.1085, 0.0772, ..., -0.0164, -0.0200, 0.0649], + [ 0.0144, 0.0311, -0.0333, ..., -0.0712, -0.0881, -0.0068], + [-0.1174, -0.0145, -0.0080, ..., -0.1209, 0.0934, 0.0115]], + device='cuda:0'), grad: tensor([[ 1.8167e-03, 8.4209e-04, 1.6518e-03, ..., 4.2105e-04, + 1.0401e-04, -2.6722e-03], + [ 1.0948e-03, 3.2616e-04, 1.0281e-03, ..., 2.7680e-04, + 2.2519e-06, 1.8702e-03], + [-9.3317e-04, -4.2343e-04, 6.9618e-04, ..., -1.5030e-03, + 1.6332e-04, 2.8439e-03], + ..., + [-8.3618e-03, 1.1170e-04, 4.1046e-03, ..., -3.1147e-03, + 2.7103e-03, -3.3684e-03], + [ 2.3861e-03, 6.7663e-04, 3.2043e-03, ..., 5.4646e-04, + 7.7677e-04, 2.9488e-03], + [-1.0335e-04, 4.5061e-04, -5.2376e-03, ..., 4.6539e-04, + 1.0979e-02, 5.0468e-03]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0189, -0.0029, 0.0019, 0.0166, -0.0061, -0.0051, 0.0100, 0.0263, + -0.0296, 0.0404], device='cuda:0'), grad: tensor([-0.0107, 0.0143, 0.0142, 0.0536, -0.0654, 0.0399, -0.0442, -0.0128, + 0.0268, -0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 218.97, cls_loss 0.5451 cls_loss_mapping 0.0116 cls_loss_causal 0.5140 re_mapping 0.0092 re_causal 0.0229 /// teacc 98.56 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.0658, 0.0625, -0.0763, ..., -0.0855, -0.0632, -0.0017], + [-0.0432, -0.1112, 0.0056, ..., -0.0566, -0.0301, -0.0595], + [ 0.0253, -0.0815, 0.0403, ..., 0.1450, -0.0929, -0.0405], + ..., + [-0.0697, -0.1098, 0.0764, ..., -0.0163, -0.0206, 0.0639], + [ 0.0148, 0.0317, -0.0326, ..., -0.0704, -0.0859, -0.0071], + [-0.1184, -0.0141, -0.0077, ..., -0.1215, 0.0932, 0.0119]], + device='cuda:0'), grad: tensor([[ 4.9400e-03, 7.9880e-03, 1.1721e-03, ..., 3.5584e-05, + 7.0333e-05, -1.2077e-02], + [-2.0075e-04, 3.4642e-04, 6.3944e-04, ..., -4.1223e-04, + 1.1437e-05, 1.2865e-03], + [ 3.4404e-04, 5.0592e-04, -3.2163e-04, ..., -5.2035e-05, + 1.5557e-05, -4.1461e-04], + ..., + [ 1.1003e-04, 1.2922e-04, 3.4142e-03, ..., -8.2195e-05, + 1.8284e-05, 1.5011e-03], + [ 4.9114e-04, -3.6182e-03, -1.6365e-03, ..., 1.2445e-04, + -3.9711e-03, -3.7861e-03], + [ 7.4577e-04, 2.0466e-03, -7.0343e-03, ..., -1.4467e-03, + -3.6449e-03, -2.6870e-04]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0202, -0.0025, 0.0019, 0.0161, -0.0056, -0.0057, 0.0094, 0.0258, + -0.0297, 0.0405], device='cuda:0'), grad: tensor([ 0.0111, 0.0109, -0.0101, -0.0147, -0.0137, 0.0293, 0.0228, 0.0338, + -0.0294, -0.0399], device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.57, cls_loss 0.5185 cls_loss_mapping 0.0107 cls_loss_causal 0.4909 re_mapping 0.0098 re_causal 0.0233 /// teacc 98.67 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0673, 0.0623, -0.0770, ..., -0.0867, -0.0647, -0.0008], + [-0.0430, -0.1117, 0.0054, ..., -0.0569, -0.0296, -0.0596], + [ 0.0280, -0.0790, 0.0393, ..., 0.1457, -0.0932, -0.0409], + ..., + [-0.0710, -0.1109, 0.0766, ..., -0.0158, -0.0204, 0.0641], + [ 0.0149, 0.0313, -0.0325, ..., -0.0714, -0.0858, -0.0069], + [-0.1191, -0.0140, -0.0080, ..., -0.1218, 0.0926, 0.0113]], + device='cuda:0'), grad: tensor([[ 3.6259e-03, -6.3419e-04, -2.5487e-04, ..., 4.1053e-06, + -4.1485e-04, -2.4929e-03], + [-7.2632e-03, 4.1509e-04, -5.1498e-04, ..., 7.3127e-06, + 1.7214e-04, -8.1444e-04], + [ 3.4924e-03, 1.7920e-03, -1.2558e-02, ..., -1.4961e-04, + 7.6723e-04, -4.2114e-03], + ..., + [ 5.8746e-04, -2.8725e-03, -9.8610e-04, ..., 5.7071e-05, + -1.1902e-02, -1.6479e-03], + [ 1.1891e-04, 1.9681e-04, -8.0442e-04, ..., 2.2113e-05, + 1.3936e-04, -8.5068e-04], + [ 9.9468e-04, 3.3779e-03, 1.3847e-02, ..., 3.7067e-06, + 1.1917e-02, 7.5607e-03]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0195, -0.0025, 0.0029, 0.0163, -0.0055, -0.0054, 0.0094, 0.0253, + -0.0300, 0.0406], device='cuda:0'), grad: tensor([-0.0038, -0.0235, -0.0354, -0.0025, -0.0105, -0.0041, 0.0285, 0.0127, + -0.0113, 0.0500], device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 218.74, cls_loss 0.5601 cls_loss_mapping 0.0126 cls_loss_causal 0.5309 re_mapping 0.0092 re_causal 0.0238 /// teacc 98.73 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0687, 0.0619, -0.0769, ..., -0.0870, -0.0655, -0.0006], + [-0.0431, -0.1124, 0.0044, ..., -0.0574, -0.0303, -0.0606], + [ 0.0279, -0.0800, 0.0392, ..., 0.1455, -0.0923, -0.0403], + ..., + [-0.0706, -0.1113, 0.0767, ..., -0.0161, -0.0198, 0.0645], + [ 0.0142, 0.0320, -0.0322, ..., -0.0711, -0.0863, -0.0072], + [-0.1179, -0.0138, -0.0077, ..., -0.1221, 0.0929, 0.0111]], + device='cuda:0'), grad: tensor([[ 1.3037e-03, 2.0809e-03, 1.4296e-03, ..., 1.1644e-03, + 3.6716e-04, 3.6964e-03], + [-3.7146e-04, 1.1486e-04, 5.2404e-04, ..., 2.3854e-04, + 4.4852e-05, -6.1750e-04], + [-4.9305e-04, -1.5478e-03, -2.0752e-03, ..., -3.7270e-03, + 1.4174e-04, -4.1151e-04], + ..., + [ 5.9700e-04, 2.5415e-04, 1.0824e-03, ..., 3.8052e-04, + 3.2830e-04, -2.2221e-03], + [ 2.9206e-04, 9.1696e-04, 1.2312e-03, ..., 2.7561e-04, + 4.6396e-04, -9.2077e-04], + [-2.4376e-03, -3.8457e-04, 2.3460e-03, ..., 1.2560e-03, + -6.9809e-04, -2.2964e-03]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0196, -0.0028, 0.0032, 0.0170, -0.0057, -0.0056, 0.0079, 0.0250, + -0.0299, 0.0418], device='cuda:0'), grad: tensor([ 0.0256, -0.0060, -0.0029, -0.0148, 0.0098, 0.0241, -0.0036, -0.0129, + -0.0108, -0.0085], device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 218.31, cls_loss 0.5558 cls_loss_mapping 0.0097 cls_loss_causal 0.5285 re_mapping 0.0097 re_causal 0.0243 /// teacc 98.65 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0688, 0.0615, -0.0776, ..., -0.0875, -0.0652, -0.0013], + [-0.0431, -0.1108, 0.0047, ..., -0.0565, -0.0308, -0.0603], + [ 0.0284, -0.0793, 0.0388, ..., 0.1459, -0.0925, -0.0407], + ..., + [-0.0714, -0.1121, 0.0772, ..., -0.0153, -0.0190, 0.0650], + [ 0.0151, 0.0322, -0.0312, ..., -0.0697, -0.0847, -0.0066], + [-0.1182, -0.0140, -0.0083, ..., -0.1229, 0.0918, 0.0110]], + device='cuda:0'), grad: tensor([[ 5.0926e-04, 2.9182e-04, 6.3896e-04, ..., 7.8022e-05, + 1.3143e-05, 1.2913e-03], + [-5.1230e-05, -1.0926e-04, -1.1358e-03, ..., 4.0913e-04, + 4.8494e-04, -9.3365e-04], + [ 2.3854e-04, 1.5652e-04, 2.1183e-04, ..., 7.2062e-05, + 3.0309e-05, 1.3323e-03], + ..., + [-7.1287e-04, 7.0095e-05, 2.5702e-04, ..., 7.3314e-05, + 2.8268e-05, 3.7346e-03], + [ 3.8109e-03, 3.4142e-03, 6.6090e-04, ..., 9.9897e-05, + 3.9995e-05, 1.1921e-03], + [ 3.9005e-04, 2.0123e-04, -8.7118e-04, ..., 9.5010e-05, + 2.5332e-05, -6.6071e-03]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0196, -0.0016, 0.0026, 0.0173, -0.0055, -0.0055, 0.0078, 0.0247, + -0.0295, 0.0404], device='cuda:0'), grad: tensor([ 0.0199, -0.0334, 0.0156, 0.0249, -0.0388, -0.0126, -0.0100, 0.0316, + 0.0334, -0.0305], device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 219.26, cls_loss 0.5339 cls_loss_mapping 0.0115 cls_loss_causal 0.5028 re_mapping 0.0098 re_causal 0.0237 /// teacc 98.73 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.0688, 0.0620, -0.0784, ..., -0.0877, -0.0650, -0.0016], + [-0.0433, -0.1104, 0.0038, ..., -0.0553, -0.0323, -0.0615], + [ 0.0286, -0.0799, 0.0397, ..., 0.1460, -0.0931, -0.0403], + ..., + [-0.0714, -0.1114, 0.0776, ..., -0.0148, -0.0193, 0.0655], + [ 0.0149, 0.0322, -0.0304, ..., -0.0696, -0.0841, -0.0068], + [-0.1186, -0.0134, -0.0079, ..., -0.1213, 0.0925, 0.0116]], + device='cuda:0'), grad: tensor([[ 8.4209e-04, 2.8305e-03, 5.8651e-04, ..., 2.4378e-04, + 1.8036e-04, 4.9829e-04], + [ 2.9135e-04, 1.6499e-04, 5.8508e-04, ..., 5.5075e-04, + 7.5221e-05, 1.1969e-03], + [ 3.5644e-05, 4.0197e-04, 9.8705e-04, ..., 1.6522e-04, + 6.5207e-05, 1.2226e-03], + ..., + [-5.2564e-06, 1.7977e-04, -3.4485e-03, ..., -2.7351e-03, + -9.6917e-05, -6.8817e-03], + [ 2.1229e-03, 4.7340e-03, 3.5739e-04, ..., 1.2600e-04, + 1.0222e-04, 3.6573e-04], + [ 1.5459e-03, 4.1542e-03, 2.9488e-03, ..., 7.1669e-04, + 2.7695e-03, 1.8368e-03]], device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0194, -0.0026, 0.0033, 0.0168, -0.0055, -0.0057, 0.0089, 0.0250, + -0.0303, 0.0412], device='cuda:0'), grad: tensor([ 0.0137, 0.0090, 0.0098, 0.0107, 0.0066, -0.0080, -0.0601, -0.0218, + 0.0127, 0.0274], device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 218.44, cls_loss 0.5141 cls_loss_mapping 0.0078 cls_loss_causal 0.4911 re_mapping 0.0094 re_causal 0.0237 /// teacc 98.67 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0691, 0.0621, -0.0772, ..., -0.0880, -0.0644, -0.0008], + [-0.0436, -0.1108, 0.0031, ..., -0.0559, -0.0327, -0.0620], + [ 0.0288, -0.0800, 0.0392, ..., 0.1465, -0.0939, -0.0405], + ..., + [-0.0706, -0.1124, 0.0771, ..., -0.0147, -0.0209, 0.0650], + [ 0.0146, 0.0336, -0.0281, ..., -0.0695, -0.0844, -0.0073], + [-0.1179, -0.0145, -0.0091, ..., -0.1226, 0.0927, 0.0126]], + device='cuda:0'), grad: tensor([[ 1.8730e-03, 1.5945e-03, 7.7105e-04, ..., 5.6982e-04, + 5.7316e-04, 2.1057e-03], + [ 2.0194e-04, 7.8976e-05, -1.2497e-02, ..., -1.7654e-02, + -1.3084e-03, -1.4629e-03], + [ 7.6485e-04, 1.6165e-04, 5.2223e-03, ..., 9.4986e-03, + 1.1940e-03, 3.3164e-04], + ..., + [-2.3136e-03, 6.5231e-04, 5.1689e-03, ..., 1.0178e-02, + 8.3303e-04, 1.5078e-03], + [-9.9421e-05, -8.2684e-04, -4.4899e-03, ..., -3.2215e-03, + -2.2049e-03, 1.1539e-03], + [-1.4486e-03, -1.9302e-03, 1.8673e-03, ..., 3.9482e-03, + -8.1730e-04, -8.3923e-03]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0196, -0.0024, 0.0028, 0.0162, -0.0055, -0.0048, 0.0081, 0.0253, + -0.0304, 0.0415], device='cuda:0'), grad: tensor([ 0.0198, -0.0584, -0.0014, -0.0150, 0.0047, -0.0131, 0.0321, 0.0469, + 0.0019, -0.0175], device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 216.62, cls_loss 0.5601 cls_loss_mapping 0.0084 cls_loss_causal 0.5348 re_mapping 0.0094 re_causal 0.0237 /// teacc 98.58 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0684, 0.0620, -0.0767, ..., -0.0891, -0.0635, -0.0006], + [-0.0440, -0.1118, 0.0037, ..., -0.0545, -0.0315, -0.0609], + [ 0.0298, -0.0786, 0.0392, ..., 0.1461, -0.0949, -0.0404], + ..., + [-0.0711, -0.1125, 0.0767, ..., -0.0155, -0.0219, 0.0643], + [ 0.0159, 0.0336, -0.0288, ..., -0.0707, -0.0841, -0.0074], + [-0.1188, -0.0138, -0.0088, ..., -0.1226, 0.0929, 0.0123]], + device='cuda:0'), grad: tensor([[ 1.4763e-03, 9.6893e-04, 2.8825e-04, ..., 9.6858e-06, + 3.7932e-04, -1.3113e-04], + [ 8.3303e-04, 6.5994e-04, -6.5374e-04, ..., 3.1495e-04, + 5.0831e-04, 1.4753e-03], + [ 1.2794e-02, -5.6877e-03, 2.0432e-02, ..., 2.2491e-02, + 3.5167e-04, 1.8826e-03], + ..., + [-2.4902e-02, 2.1183e-04, -2.5848e-02, ..., -2.6611e-02, + 1.8597e-04, -1.1511e-03], + [ 9.5367e-03, 6.2561e-03, 1.5440e-03, ..., 6.2132e-04, + 1.7118e-04, 9.2411e-04], + [ 1.5430e-03, 3.9911e-04, 2.5368e-03, ..., 1.2064e-03, + 1.5039e-03, 2.2392e-03]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0203, -0.0014, 0.0024, 0.0173, -0.0060, -0.0053, 0.0077, 0.0242, + -0.0304, 0.0418], device='cuda:0'), grad: tensor([-0.0111, -0.0139, 0.0411, 0.0025, -0.0092, -0.0127, 0.0215, -0.0431, + 0.0056, 0.0193], device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 216.64, cls_loss 0.5699 cls_loss_mapping 0.0099 cls_loss_causal 0.5400 re_mapping 0.0094 re_causal 0.0240 /// teacc 98.45 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.0696, 0.0614, -0.0764, ..., -0.0896, -0.0645, -0.0004], + [-0.0440, -0.1117, 0.0033, ..., -0.0546, -0.0322, -0.0613], + [ 0.0292, -0.0786, 0.0392, ..., 0.1456, -0.0958, -0.0409], + ..., + [-0.0702, -0.1121, 0.0765, ..., -0.0154, -0.0223, 0.0643], + [ 0.0160, 0.0347, -0.0289, ..., -0.0698, -0.0843, -0.0062], + [-0.1205, -0.0151, -0.0082, ..., -0.1219, 0.0944, 0.0123]], + device='cuda:0'), grad: tensor([[ 1.0815e-03, 4.4733e-05, 3.8028e-04, ..., 1.4658e-03, + 1.4818e-04, 2.4452e-03], + [-1.8787e-03, 1.2428e-05, -1.2913e-03, ..., -8.9979e-04, + -1.2579e-03, -2.5320e-04], + [-5.3501e-04, 2.0444e-04, 3.9864e-04, ..., 2.4748e-04, + 1.6057e-04, 1.7290e-03], + ..., + [-2.5921e-03, 4.1038e-05, 3.7026e-04, ..., -2.9392e-03, + 2.9349e-04, -9.2316e-03], + [ 2.2564e-03, 1.8177e-03, 3.9816e-04, ..., 6.0415e-04, + 1.1616e-03, 1.3151e-03], + [ 2.6646e-03, -7.9203e-04, -4.2272e-04, ..., -7.6008e-04, + 1.1063e-03, -1.4429e-03]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0200, -0.0017, 0.0026, 0.0167, -0.0067, -0.0050, 0.0091, 0.0237, + -0.0293, 0.0411], device='cuda:0'), grad: tensor([ 0.0261, -0.0650, -0.0051, 0.0284, -0.0093, 0.0081, 0.0181, -0.0122, + 0.0248, -0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 216.58, cls_loss 0.5667 cls_loss_mapping 0.0105 cls_loss_causal 0.5379 re_mapping 0.0095 re_causal 0.0240 /// teacc 98.82 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.0692, 0.0617, -0.0769, ..., -0.0904, -0.0650, -0.0008], + [-0.0454, -0.1122, 0.0030, ..., -0.0544, -0.0306, -0.0622], + [ 0.0284, -0.0800, 0.0391, ..., 0.1458, -0.0955, -0.0416], + ..., + [-0.0702, -0.1129, 0.0766, ..., -0.0161, -0.0230, 0.0644], + [ 0.0150, 0.0340, -0.0285, ..., -0.0696, -0.0847, -0.0053], + [-0.1204, -0.0147, -0.0079, ..., -0.1208, 0.0940, 0.0118]], + device='cuda:0'), grad: tensor([[ 1.2512e-02, 1.7715e-02, -3.4404e-04, ..., 3.1147e-03, + 7.7009e-05, -1.6606e-04], + [ 8.0299e-04, 6.1369e-04, 1.6241e-03, ..., -1.6883e-05, + 2.1338e-04, 1.3552e-03], + [-3.0041e-04, 3.4189e-04, 1.2903e-03, ..., -1.3757e-04, + 2.4295e-04, 1.5755e-03], + ..., + [ 7.8821e-04, 4.7350e-04, 3.4828e-03, ..., 1.4007e-04, + 5.6362e-04, 3.9787e-03], + [-1.1292e-02, -1.7090e-02, 1.3580e-03, ..., -3.0479e-03, + -2.9564e-03, 1.0147e-03], + [-1.4524e-03, -4.5753e-04, -1.2337e-02, ..., -8.4639e-05, + 4.1618e-03, -1.4633e-02]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0201, -0.0027, 0.0026, 0.0176, -0.0068, -0.0060, 0.0098, 0.0244, + -0.0296, 0.0411], device='cuda:0'), grad: tensor([ 0.0139, 0.0188, -0.0008, 0.0282, 0.0453, -0.0166, 0.0192, 0.0264, + -0.0400, -0.0944], device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 218.92, cls_loss 0.5431 cls_loss_mapping 0.0114 cls_loss_causal 0.5171 re_mapping 0.0091 re_causal 0.0219 /// teacc 98.80 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0683, 0.0618, -0.0781, ..., -0.0909, -0.0652, -0.0007], + [-0.0448, -0.1129, 0.0022, ..., -0.0543, -0.0319, -0.0631], + [ 0.0300, -0.0779, 0.0391, ..., 0.1452, -0.0967, -0.0417], + ..., + [-0.0703, -0.1143, 0.0775, ..., -0.0150, -0.0235, 0.0645], + [ 0.0140, 0.0338, -0.0277, ..., -0.0691, -0.0851, -0.0050], + [-0.1212, -0.0143, -0.0075, ..., -0.1212, 0.0943, 0.0122]], + device='cuda:0'), grad: tensor([[ 5.7030e-04, 9.7132e-04, 1.1015e-03, ..., 3.6091e-05, + 8.8739e-04, 2.0275e-03], + [ 8.2636e-04, 1.0605e-03, 3.2902e-04, ..., 1.9884e-04, + 1.4067e-03, 2.0523e-03], + [ 3.3307e-04, 1.4317e-04, -1.2608e-03, ..., 5.5432e-05, + -1.7691e-04, -1.1835e-03], + ..., + [ 2.5105e-04, 1.1396e-03, 5.9223e-04, ..., 5.7077e-04, + 1.0681e-03, 2.5787e-03], + [ 3.5882e-04, 7.7534e-04, 1.5068e-03, ..., 1.5569e-04, + 1.4734e-03, 2.5101e-03], + [ 1.1292e-03, 3.7689e-03, 1.6571e-02, ..., 3.8028e-04, + 1.0933e-02, 7.9346e-03]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0208, -0.0024, 0.0027, 0.0152, -0.0067, -0.0057, 0.0100, 0.0248, + -0.0298, 0.0418], device='cuda:0'), grad: tensor([ 0.0188, 0.0134, -0.0178, -0.0408, -0.0386, 0.0123, -0.0091, 0.0180, + 0.0164, 0.0275], device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 219.29, cls_loss 0.5245 cls_loss_mapping 0.0097 cls_loss_causal 0.5013 re_mapping 0.0091 re_causal 0.0220 /// teacc 98.73 lr 0.00010000 +Epoch 160, weight, value: tensor([[-6.8483e-02, 6.1669e-02, -7.7909e-02, ..., -9.0573e-02, + -6.4763e-02, -5.0795e-05], + [-4.5457e-02, -1.1327e-01, 2.4119e-03, ..., -5.3127e-02, + -3.2702e-02, -6.2612e-02], + [ 2.9776e-02, -7.8854e-02, 4.0121e-02, ..., 1.4512e-01, + -9.4742e-02, -4.0965e-02], + ..., + [-7.0818e-02, -1.1364e-01, 7.7740e-02, ..., -1.4546e-02, + -2.4119e-02, 6.3351e-02], + [ 1.4701e-02, 3.3704e-02, -2.7862e-02, ..., -7.0124e-02, + -8.6117e-02, -5.2873e-03], + [-1.2245e-01, -1.4197e-02, -8.0442e-03, ..., -1.2212e-01, + 9.3973e-02, 1.2549e-02]], device='cuda:0'), grad: tensor([[ 0.0051, 0.0048, 0.0010, ..., 0.0003, 0.0004, 0.0008], + [-0.0067, -0.0014, 0.0007, ..., 0.0015, -0.0005, 0.0011], + [ 0.0044, 0.0012, 0.0009, ..., -0.0009, 0.0006, 0.0008], + ..., + [-0.0104, 0.0003, -0.0058, ..., -0.0015, -0.0007, -0.0027], + [-0.0058, -0.0047, -0.0017, ..., 0.0005, -0.0033, -0.0005], + [ 0.0015, 0.0013, 0.0013, ..., 0.0002, 0.0002, 0.0006]], + device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0202, -0.0009, 0.0031, 0.0144, -0.0067, -0.0061, 0.0097, 0.0246, + -0.0294, 0.0416], device='cuda:0'), grad: tensor([ 0.0220, 0.0174, 0.0215, -0.0139, 0.0146, -0.0145, 0.0062, -0.0420, + -0.0281, 0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 216.86, cls_loss 0.5329 cls_loss_mapping 0.0089 cls_loss_causal 0.5042 re_mapping 0.0091 re_causal 0.0228 /// teacc 98.71 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.0694, 0.0620, -0.0792, ..., -0.0914, -0.0661, -0.0004], + [-0.0467, -0.1146, 0.0017, ..., -0.0543, -0.0338, -0.0642], + [ 0.0292, -0.0791, 0.0410, ..., 0.1458, -0.0963, -0.0416], + ..., + [-0.0710, -0.1135, 0.0776, ..., -0.0158, -0.0240, 0.0633], + [ 0.0141, 0.0334, -0.0282, ..., -0.0702, -0.0872, -0.0056], + [-0.1244, -0.0148, -0.0082, ..., -0.1229, 0.0942, 0.0126]], + device='cuda:0'), grad: tensor([[ 4.0674e-04, 7.0333e-04, 3.0088e-04, ..., 4.7266e-05, + 2.1446e-04, 8.1730e-04], + [ 2.5940e-04, 3.0851e-04, 9.7466e-04, ..., 2.1517e-05, + 2.3651e-04, 2.6302e-03], + [ 3.9101e-04, 4.0531e-04, 2.4581e-04, ..., 1.6403e-04, + 1.9467e-04, 6.9189e-04], + ..., + [-2.0492e-04, -4.9353e-04, -5.1308e-03, ..., -9.8705e-04, + -2.7466e-03, -7.4501e-03], + [ 5.0592e-04, 2.4259e-04, 4.5872e-04, ..., 3.6192e-04, + 3.4499e-04, 9.3937e-04], + [ 8.3637e-04, 1.3685e-03, 3.0193e-03, ..., 7.9393e-04, + 2.5864e-03, 3.3760e-03]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0194, -0.0016, 0.0028, 0.0154, -0.0063, -0.0068, 0.0105, 0.0246, + -0.0295, 0.0418], device='cuda:0'), grad: tensor([ 0.0050, 0.0123, 0.0041, 0.0147, 0.0080, 0.0010, -0.0366, -0.0305, + 0.0080, 0.0140], device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 216.85, cls_loss 0.5387 cls_loss_mapping 0.0099 cls_loss_causal 0.5163 re_mapping 0.0095 re_causal 0.0247 /// teacc 98.65 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0695, 0.0615, -0.0795, ..., -0.0910, -0.0656, -0.0011], + [-0.0471, -0.1145, 0.0031, ..., -0.0542, -0.0334, -0.0643], + [ 0.0291, -0.0797, 0.0421, ..., 0.1472, -0.0981, -0.0422], + ..., + [-0.0718, -0.1146, 0.0779, ..., -0.0163, -0.0242, 0.0634], + [ 0.0148, 0.0334, -0.0293, ..., -0.0706, -0.0884, -0.0056], + [-0.1238, -0.0139, -0.0087, ..., -0.1234, 0.0943, 0.0128]], + device='cuda:0'), grad: tensor([[ 5.7650e-04, 4.5705e-04, 2.0289e-04, ..., 2.9469e-04, + -1.6356e-03, -2.4376e-03], + [-3.4422e-05, -2.4307e-04, 3.2997e-04, ..., 5.4359e-04, + 2.6464e-04, 3.6454e-04], + [ 9.4700e-04, 9.8801e-04, -6.9542e-03, ..., -3.3016e-03, + 1.8930e-04, 2.2209e-04], + ..., + [-6.6662e-04, -4.9829e-04, 3.8033e-03, ..., 1.8873e-03, + 6.3801e-04, 1.5509e-04], + [-2.0552e-04, 3.2926e-04, 6.1274e-04, ..., 4.5443e-04, + 1.5640e-04, 1.9860e-04], + [ 2.6250e-04, 3.5691e-04, 1.1272e-03, ..., 5.0735e-04, + 2.3866e-04, 2.2471e-04]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0191, -0.0009, 0.0027, 0.0143, -0.0071, -0.0055, 0.0106, 0.0248, + -0.0298, 0.0423], device='cuda:0'), grad: tensor([-0.0204, -0.0152, -0.0044, 0.0034, 0.0371, -0.0184, 0.0128, -0.0162, + 0.0099, 0.0113], device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 216.65, cls_loss 0.5446 cls_loss_mapping 0.0124 cls_loss_causal 0.5147 re_mapping 0.0088 re_causal 0.0214 /// teacc 98.71 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.0702, 0.0617, -0.0803, ..., -0.0923, -0.0657, -0.0011], + [-0.0474, -0.1145, 0.0038, ..., -0.0545, -0.0319, -0.0647], + [ 0.0277, -0.0806, 0.0431, ..., 0.1472, -0.0983, -0.0412], + ..., + [-0.0722, -0.1153, 0.0784, ..., -0.0155, -0.0247, 0.0633], + [ 0.0159, 0.0329, -0.0297, ..., -0.0699, -0.0883, -0.0060], + [-0.1240, -0.0136, -0.0092, ..., -0.1232, 0.0943, 0.0131]], + device='cuda:0'), grad: tensor([[-1.4257e-03, -9.5825e-03, 1.6820e-04, ..., -2.8563e-04, + 4.0102e-04, -5.7030e-04], + [-1.5173e-03, -8.7404e-04, -3.2902e-04, ..., 3.1680e-05, + -3.5572e-04, 5.4979e-04], + [ 2.3697e-02, 1.1482e-02, -1.5097e-03, ..., 6.1111e-03, + 3.6764e-04, 6.8951e-04], + ..., + [ 4.2033e-04, 5.3501e-04, 2.7680e-04, ..., 2.7418e-04, + 6.1655e-04, 5.7983e-04], + [ 1.6317e-03, 2.5272e-03, 1.9741e-03, ..., 2.1458e-04, + 3.4046e-04, 6.9857e-05], + [ 1.6060e-03, 1.8539e-03, 2.3117e-03, ..., 5.9962e-05, + 1.3657e-03, 1.0414e-03]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0188, -0.0013, 0.0022, 0.0136, -0.0073, -0.0056, 0.0106, 0.0259, + -0.0295, 0.0429], device='cuda:0'), grad: tensor([ 0.0060, -0.0081, 0.0283, -0.0225, 0.0071, -0.0155, -0.0064, 0.0115, + -0.0250, 0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 216.73, cls_loss 0.5395 cls_loss_mapping 0.0104 cls_loss_causal 0.5126 re_mapping 0.0094 re_causal 0.0226 /// teacc 98.77 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0709, 0.0615, -0.0808, ..., -0.0931, -0.0663, -0.0017], + [-0.0483, -0.1164, 0.0043, ..., -0.0537, -0.0305, -0.0641], + [ 0.0277, -0.0812, 0.0425, ..., 0.1464, -0.0981, -0.0408], + ..., + [-0.0731, -0.1171, 0.0789, ..., -0.0147, -0.0255, 0.0641], + [ 0.0150, 0.0325, -0.0295, ..., -0.0694, -0.0881, -0.0062], + [-0.1237, -0.0127, -0.0093, ..., -0.1228, 0.0939, 0.0127]], + device='cuda:0'), grad: tensor([[-0.0101, -0.0087, 0.0009, ..., 0.0002, 0.0003, -0.0039], + [ 0.0006, 0.0005, 0.0008, ..., 0.0085, 0.0006, 0.0024], + [ 0.0028, 0.0020, -0.0007, ..., -0.0002, -0.0010, -0.0009], + ..., + [ 0.0007, 0.0007, -0.0028, ..., 0.0002, 0.0007, -0.0008], + [ 0.0023, 0.0022, 0.0009, ..., 0.0001, 0.0006, 0.0035], + [ 0.0002, 0.0008, 0.0019, ..., 0.0002, -0.0015, -0.0017]], + device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0183, -0.0011, 0.0021, 0.0141, -0.0068, -0.0053, 0.0093, 0.0264, + -0.0294, 0.0427], device='cuda:0'), grad: tensor([-0.0026, 0.0468, -0.0079, 0.0012, 0.0122, 0.0154, -0.0718, 0.0049, + 0.0195, -0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 216.43, cls_loss 0.5199 cls_loss_mapping 0.0110 cls_loss_causal 0.4898 re_mapping 0.0092 re_causal 0.0223 /// teacc 98.76 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0711, 0.0615, -0.0806, ..., -0.0926, -0.0664, -0.0013], + [-0.0494, -0.1171, 0.0045, ..., -0.0546, -0.0306, -0.0638], + [ 0.0291, -0.0809, 0.0430, ..., 0.1470, -0.0989, -0.0398], + ..., + [-0.0741, -0.1171, 0.0776, ..., -0.0154, -0.0246, 0.0637], + [ 0.0148, 0.0326, -0.0297, ..., -0.0689, -0.0891, -0.0075], + [-0.1240, -0.0135, -0.0075, ..., -0.1221, 0.0935, 0.0142]], + device='cuda:0'), grad: tensor([[ 9.1743e-04, 1.0662e-03, 1.1168e-03, ..., 1.8990e-04, + 6.1369e-04, 1.4267e-03], + [ 9.7847e-04, 5.5075e-04, 1.7529e-03, ..., 6.8665e-04, + 8.0729e-04, 8.4066e-04], + [-6.7282e-04, -5.8889e-04, -2.3270e-03, ..., -2.9545e-03, + 4.5633e-04, 3.1710e-04], + ..., + [ 3.7432e-04, 4.2367e-04, 3.4885e-03, ..., 1.2989e-03, + 2.3818e-04, 5.3596e-04], + [-5.1422e-03, 1.6451e-03, 8.9264e-04, ..., 1.3030e-04, + 2.1672e-04, 6.4135e-04], + [ 4.5490e-04, -1.5104e-04, -6.5689e-03, ..., 8.1599e-05, + -1.8692e-03, 1.1959e-03]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0193, -0.0017, 0.0029, 0.0142, -0.0065, -0.0051, 0.0093, 0.0259, + -0.0306, 0.0427], device='cuda:0'), grad: tensor([ 0.0230, 0.0268, -0.0196, 0.0280, -0.0609, -0.0074, 0.0026, 0.0243, + -0.0050, -0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 216.57, cls_loss 0.5489 cls_loss_mapping 0.0087 cls_loss_causal 0.5102 re_mapping 0.0091 re_causal 0.0225 /// teacc 98.79 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0707, 0.0618, -0.0810, ..., -0.0937, -0.0670, -0.0018], + [-0.0505, -0.1173, 0.0039, ..., -0.0560, -0.0300, -0.0645], + [ 0.0292, -0.0816, 0.0430, ..., 0.1481, -0.0996, -0.0400], + ..., + [-0.0750, -0.1181, 0.0775, ..., -0.0161, -0.0240, 0.0645], + [ 0.0151, 0.0327, -0.0296, ..., -0.0680, -0.0903, -0.0071], + [-0.1245, -0.0120, -0.0072, ..., -0.1225, 0.0946, 0.0153]], + device='cuda:0'), grad: tensor([[-6.4611e-04, -7.2908e-04, 5.4264e-04, ..., -1.6558e-04, + 1.2815e-04, -1.5898e-03], + [-1.2350e-03, -3.0575e-03, -7.8888e-03, ..., -5.4245e-03, + 1.4949e-04, -2.7218e-03], + [ 1.0929e-03, 1.9426e-03, 6.4697e-03, ..., 3.6888e-03, + 1.3912e-04, 2.1572e-03], + ..., + [ 8.9874e-03, 1.0977e-03, 4.5433e-03, ..., 2.4724e-04, + 3.3522e-04, 1.3151e-03], + [ 3.7689e-03, 2.3575e-03, -3.8743e-05, ..., 1.3030e-04, + 9.8467e-05, -5.8889e-04], + [ 4.6277e-04, -2.2030e-03, -2.3918e-03, ..., 2.8753e-04, + -2.4357e-03, -3.6354e-03]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0188, -0.0018, 0.0029, 0.0150, -0.0074, -0.0047, 0.0088, 0.0254, + -0.0302, 0.0435], device='cuda:0'), grad: tensor([-0.0121, -0.0804, 0.0479, -0.0323, 0.0293, -0.0011, 0.0247, 0.0488, + -0.0315, 0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 216.89, cls_loss 0.5125 cls_loss_mapping 0.0077 cls_loss_causal 0.4796 re_mapping 0.0097 re_causal 0.0233 /// teacc 98.69 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0699, 0.0631, -0.0811, ..., -0.0932, -0.0687, -0.0016], + [-0.0510, -0.1170, 0.0050, ..., -0.0544, -0.0307, -0.0640], + [ 0.0282, -0.0822, 0.0429, ..., 0.1475, -0.1005, -0.0404], + ..., + [-0.0755, -0.1197, 0.0772, ..., -0.0164, -0.0234, 0.0647], + [ 0.0153, 0.0324, -0.0299, ..., -0.0679, -0.0913, -0.0071], + [-0.1253, -0.0119, -0.0068, ..., -0.1233, 0.0952, 0.0153]], + device='cuda:0'), grad: tensor([[ 0.0007, 0.0007, 0.0003, ..., 0.0003, 0.0003, 0.0007], + [ 0.0006, 0.0006, -0.0035, ..., -0.0011, -0.0056, -0.0067], + [ 0.0036, 0.0012, 0.0011, ..., 0.0016, 0.0008, 0.0019], + ..., + [-0.0008, -0.0032, -0.0030, ..., -0.0016, -0.0005, -0.0034], + [ 0.0065, 0.0028, 0.0009, ..., 0.0023, 0.0013, 0.0017], + [-0.0006, -0.0006, 0.0010, ..., 0.0006, 0.0007, -0.0028]], + device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0186, -0.0015, 0.0022, 0.0162, -0.0068, -0.0056, 0.0088, 0.0248, + -0.0301, 0.0436], device='cuda:0'), grad: tensor([ 0.0100, -0.0319, 0.0195, -0.0168, 0.0083, 0.0032, 0.0286, -0.0419, + 0.0243, -0.0033], device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 217.03, cls_loss 0.5415 cls_loss_mapping 0.0099 cls_loss_causal 0.5131 re_mapping 0.0087 re_causal 0.0218 /// teacc 98.77 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.0699, 0.0630, -0.0823, ..., -0.0944, -0.0709, -0.0027], + [-0.0507, -0.1167, 0.0051, ..., -0.0534, -0.0309, -0.0644], + [ 0.0288, -0.0816, 0.0444, ..., 0.1480, -0.1003, -0.0401], + ..., + [-0.0759, -0.1203, 0.0777, ..., -0.0163, -0.0230, 0.0657], + [ 0.0148, 0.0323, -0.0300, ..., -0.0683, -0.0923, -0.0074], + [-0.1258, -0.0120, -0.0074, ..., -0.1243, 0.0960, 0.0153]], + device='cuda:0'), grad: tensor([[ 4.4727e-04, 1.2903e-03, 1.3940e-05, ..., 6.0701e-04, + 1.9109e-04, 1.6344e-04], + [-1.9302e-03, 1.7395e-03, -2.0695e-03, ..., 1.1683e-03, + 2.3305e-04, 2.8014e-04], + [ 1.2522e-03, 1.5841e-03, 4.8876e-04, ..., 1.5993e-03, + 9.2745e-05, 6.5744e-05], + ..., + [ 5.0688e-04, 5.8413e-04, -3.6210e-05, ..., 2.0683e-04, + 2.0850e-04, 1.6117e-04], + [ 5.3978e-04, -3.7003e-03, 1.3247e-03, ..., -4.9553e-03, + 2.2149e-04, 2.0576e-04], + [ 5.2738e-04, 3.4618e-03, 3.6287e-04, ..., 1.4174e-04, + 5.7335e-03, 7.4615e-03]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0180, -0.0013, 0.0025, 0.0158, -0.0071, -0.0045, 0.0091, 0.0245, + -0.0302, 0.0435], device='cuda:0'), grad: tensor([ 0.0109, -0.0069, 0.0193, -0.0249, -0.0107, 0.0147, -0.0268, 0.0074, + -0.0077, 0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 216.32, cls_loss 0.5262 cls_loss_mapping 0.0087 cls_loss_causal 0.4966 re_mapping 0.0087 re_causal 0.0202 /// teacc 98.81 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0696, 0.0629, -0.0824, ..., -0.0953, -0.0734, -0.0044], + [-0.0509, -0.1166, 0.0045, ..., -0.0541, -0.0304, -0.0652], + [ 0.0278, -0.0811, 0.0438, ..., 0.1483, -0.1017, -0.0408], + ..., + [-0.0757, -0.1205, 0.0767, ..., -0.0161, -0.0226, 0.0670], + [ 0.0143, 0.0318, -0.0295, ..., -0.0692, -0.0911, -0.0057], + [-0.1259, -0.0126, -0.0071, ..., -0.1240, 0.0949, 0.0137]], + device='cuda:0'), grad: tensor([[-2.3413e-04, -1.5602e-03, 2.9826e-04, ..., 3.6657e-05, + -7.6914e-04, -2.0182e-04], + [ 7.3767e-04, 1.2054e-03, 3.2973e-04, ..., 5.6356e-05, + 1.0090e-03, 2.6379e-03], + [-5.2032e-03, -3.2120e-03, -1.5354e-03, ..., -2.0254e-04, + -2.4490e-03, -1.8402e-02], + ..., + [-8.9025e-04, 4.1366e-04, -9.4223e-04, ..., -1.5438e-04, + -8.7357e-04, -7.2632e-03], + [ 9.6226e-04, 1.4849e-03, 3.3021e-04, ..., 9.8407e-05, + 9.6369e-04, 2.0809e-03], + [ 8.5068e-04, 1.0481e-03, 2.7132e-04, ..., 6.1452e-05, + 8.0490e-04, 2.9678e-03]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0180, -0.0014, 0.0020, 0.0150, -0.0071, -0.0054, 0.0100, 0.0251, + -0.0294, 0.0433], device='cuda:0'), grad: tensor([-0.0131, 0.0296, -0.0702, -0.0208, 0.0155, 0.0301, 0.0197, -0.0396, + 0.0237, 0.0250], device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 216.99, cls_loss 0.5649 cls_loss_mapping 0.0093 cls_loss_causal 0.5421 re_mapping 0.0090 re_causal 0.0222 /// teacc 98.73 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0703, 0.0623, -0.0815, ..., -0.0959, -0.0728, -0.0032], + [-0.0497, -0.1162, 0.0031, ..., -0.0551, -0.0314, -0.0649], + [ 0.0277, -0.0814, 0.0447, ..., 0.1490, -0.1018, -0.0396], + ..., + [-0.0756, -0.1209, 0.0781, ..., -0.0155, -0.0217, 0.0670], + [ 0.0141, 0.0320, -0.0309, ..., -0.0694, -0.0928, -0.0077], + [-0.1272, -0.0135, -0.0079, ..., -0.1253, 0.0950, 0.0140]], + device='cuda:0'), grad: tensor([[-2.1493e-04, -1.0335e-04, 1.6558e-04, ..., 5.9217e-05, + 6.5756e-04, 2.3639e-04], + [ 7.9107e-04, 7.9393e-04, 1.7631e-04, ..., 1.3566e-04, + 6.9904e-04, 7.1955e-04], + [ 1.5745e-03, 1.0176e-03, -3.6001e-04, ..., -7.7677e-04, + 4.2796e-04, 2.5654e-04], + ..., + [ 6.2656e-04, 1.3342e-03, 9.3126e-04, ..., 1.1426e-04, + 2.1267e-03, 5.4330e-05], + [ 1.3409e-03, 1.8158e-03, 5.8985e-04, ..., 9.5320e-04, + 5.2500e-04, 2.5201e-04], + [-5.9853e-03, -1.2970e-02, -2.0714e-03, ..., -9.0170e-04, + -2.1152e-03, 9.1124e-04]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0171, -0.0005, 0.0020, 0.0144, -0.0067, -0.0059, 0.0102, 0.0258, + -0.0301, 0.0437], device='cuda:0'), grad: tensor([-0.0179, 0.0210, 0.0165, 0.0209, -0.0090, -0.0292, 0.0145, 0.0138, + -0.0114, -0.0191], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 169---------------------------------------------------- +epoch 169, time 217.07, cls_loss 0.5367 cls_loss_mapping 0.0080 cls_loss_causal 0.5119 re_mapping 0.0090 re_causal 0.0230 /// teacc 98.95 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0695, 0.0633, -0.0825, ..., -0.0966, -0.0737, -0.0042], + [-0.0499, -0.1174, 0.0027, ..., -0.0546, -0.0319, -0.0661], + [ 0.0273, -0.0819, 0.0449, ..., 0.1487, -0.1018, -0.0394], + ..., + [-0.0764, -0.1214, 0.0792, ..., -0.0152, -0.0224, 0.0662], + [ 0.0135, 0.0311, -0.0308, ..., -0.0697, -0.0935, -0.0079], + [-0.1262, -0.0129, -0.0086, ..., -0.1264, 0.0955, 0.0143]], + device='cuda:0'), grad: tensor([[-0.0059, -0.0079, 0.0004, ..., 0.0010, 0.0009, 0.0024], + [-0.0006, 0.0010, -0.0113, ..., -0.0052, 0.0001, -0.0099], + [ 0.0050, 0.0016, 0.0067, ..., 0.0043, -0.0021, 0.0040], + ..., + [-0.0036, 0.0005, -0.0012, ..., 0.0028, 0.0005, 0.0064], + [ 0.0083, 0.0085, 0.0033, ..., 0.0017, 0.0016, 0.0029], + [-0.0006, -0.0020, -0.0016, ..., -0.0004, -0.0046, -0.0011]], + device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0163, -0.0001, 0.0022, 0.0148, -0.0064, -0.0059, 0.0107, 0.0254, + -0.0302, 0.0436], device='cuda:0'), grad: tensor([-0.0048, -0.0514, -0.0113, -0.0313, 0.0182, 0.0283, -0.0125, 0.0250, + 0.0469, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 216.23, cls_loss 0.5484 cls_loss_mapping 0.0075 cls_loss_causal 0.5235 re_mapping 0.0088 re_causal 0.0224 /// teacc 98.73 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0693, 0.0631, -0.0817, ..., -0.0963, -0.0723, -0.0037], + [-0.0501, -0.1168, 0.0028, ..., -0.0542, -0.0319, -0.0654], + [ 0.0274, -0.0818, 0.0445, ..., 0.1503, -0.1010, -0.0416], + ..., + [-0.0758, -0.1202, 0.0791, ..., -0.0152, -0.0233, 0.0667], + [ 0.0134, 0.0316, -0.0316, ..., -0.0684, -0.0936, -0.0073], + [-0.1264, -0.0131, -0.0080, ..., -0.1265, 0.0964, 0.0148]], + device='cuda:0'), grad: tensor([[ 5.2452e-05, 7.2122e-05, 1.8907e-04, ..., -2.7728e-04, + 1.3745e-04, 1.3313e-03], + [ 1.0037e-04, 4.0340e-04, 3.1257e-04, ..., 1.6820e-04, + 5.1212e-04, 1.6813e-03], + [-2.4891e-03, -1.1473e-03, -3.4351e-03, ..., -7.6332e-03, + 7.4983e-05, -7.1406e-05], + ..., + [ 9.5701e-04, -7.9572e-05, 1.3037e-03, ..., 4.1428e-03, + -1.0031e-04, -5.2719e-03], + [ 5.1051e-05, -1.5411e-03, -3.0762e-02, ..., -2.0767e-02, + -2.3575e-02, -1.0460e-02], + [ 1.3924e-03, 4.1161e-03, 3.1677e-02, ..., 2.0401e-02, + 2.7451e-02, 1.2024e-02]], device='cuda:0') +Epoch 172, bias, value: tensor([ 1.7415e-02, 8.6663e-05, 1.1879e-03, 1.5293e-02, -6.2449e-03, + -6.7732e-03, 1.0391e-02, 2.6085e-02, -3.0433e-02, 4.3132e-02], + device='cuda:0'), grad: tensor([ 0.0082, 0.0205, -0.0051, -0.0039, -0.0174, 0.0100, 0.0198, -0.0416, + -0.0463, 0.0557], device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 216.42, cls_loss 0.5544 cls_loss_mapping 0.0097 cls_loss_causal 0.5225 re_mapping 0.0091 re_causal 0.0228 /// teacc 98.63 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0690, 0.0626, -0.0826, ..., -0.0970, -0.0727, -0.0040], + [-0.0504, -0.1173, 0.0022, ..., -0.0530, -0.0330, -0.0667], + [ 0.0276, -0.0817, 0.0441, ..., 0.1502, -0.1025, -0.0416], + ..., + [-0.0754, -0.1200, 0.0789, ..., -0.0158, -0.0239, 0.0658], + [ 0.0128, 0.0308, -0.0309, ..., -0.0679, -0.0933, -0.0073], + [-0.1259, -0.0132, -0.0082, ..., -0.1277, 0.0966, 0.0140]], + device='cuda:0'), grad: tensor([[ 2.6393e-04, 8.3566e-05, 1.8835e-04, ..., 4.0078e-04, + 4.7374e-04, 1.1072e-03], + [ 8.9121e-04, -5.2643e-03, -4.3640e-03, ..., -4.1122e-03, + -1.0239e-02, -4.8637e-03], + [-1.1797e-03, 3.5620e-04, 3.1567e-04, ..., -5.2223e-03, + 2.8276e-04, -2.5406e-03], + ..., + [ 6.8045e-04, 3.2210e-04, 3.0303e-04, ..., 8.4591e-04, + 6.0749e-04, 2.0828e-03], + [-1.0309e-03, 2.2471e-05, 5.8126e-04, ..., 8.7023e-04, + 1.0576e-03, 1.8406e-03], + [ 1.1978e-03, -2.2614e-04, -2.2888e-03, ..., 2.0828e-03, + 1.4791e-03, 4.0283e-03]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0165, 0.0007, 0.0007, 0.0158, -0.0050, -0.0055, 0.0096, 0.0252, + -0.0311, 0.0435], device='cuda:0'), grad: tensor([ 0.0050, -0.0113, -0.0121, -0.0144, 0.0206, 0.0456, -0.0521, 0.0101, + 0.0019, 0.0066], device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 216.43, cls_loss 0.5141 cls_loss_mapping 0.0094 cls_loss_causal 0.4981 re_mapping 0.0087 re_causal 0.0216 /// teacc 98.55 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0695, 0.0628, -0.0833, ..., -0.0971, -0.0733, -0.0047], + [-0.0497, -0.1186, 0.0010, ..., -0.0523, -0.0332, -0.0673], + [ 0.0289, -0.0814, 0.0444, ..., 0.1504, -0.1023, -0.0413], + ..., + [-0.0761, -0.1202, 0.0788, ..., -0.0162, -0.0237, 0.0662], + [ 0.0134, 0.0306, -0.0286, ..., -0.0684, -0.0945, -0.0073], + [-0.1254, -0.0126, -0.0090, ..., -0.1279, 0.0958, 0.0138]], + device='cuda:0'), grad: tensor([[ 1.3634e-02, 5.1910e-02, 4.8184e-04, ..., 1.0014e-04, + 1.9181e-04, 1.2465e-03], + [ 1.0357e-03, 7.4816e-04, 1.1330e-03, ..., 7.3051e-04, + 6.2513e-04, 2.1992e-03], + [ 6.9332e-04, 3.5572e-04, 7.7486e-04, ..., 1.7750e-04, + 2.1529e-04, 1.4133e-03], + ..., + [-1.6298e-03, 3.0060e-03, 1.3666e-03, ..., -8.7786e-04, + -6.9332e-04, 1.8368e-03], + [-2.0826e-04, -1.0252e-03, -9.7752e-04, ..., -3.0309e-05, + 3.0851e-04, -2.3785e-03], + [ 8.5211e-04, -2.6779e-03, -3.1357e-03, ..., 3.1996e-04, + 2.5439e-04, -3.2921e-03]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0170, -0.0003, 0.0008, 0.0161, -0.0050, -0.0065, 0.0106, 0.0250, + -0.0314, 0.0438], device='cuda:0'), grad: tensor([ 0.0493, 0.0223, 0.0187, -0.0115, 0.0081, 0.0177, -0.0682, -0.0055, + -0.0202, -0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 216.47, cls_loss 0.5503 cls_loss_mapping 0.0087 cls_loss_causal 0.5193 re_mapping 0.0085 re_causal 0.0223 /// teacc 98.51 lr 0.00010000 +Epoch 175, weight, value: tensor([[-6.9965e-02, 6.1906e-02, -8.2484e-02, ..., -9.6635e-02, + -7.2429e-02, -3.9269e-03], + [-4.8821e-02, -1.1833e-01, -1.0903e-04, ..., -5.3002e-02, + -3.3274e-02, -6.7328e-02], + [ 2.7929e-02, -8.1985e-02, 4.4894e-02, ..., 1.5076e-01, + -1.0295e-01, -4.2292e-02], + ..., + [-7.6694e-02, -1.2054e-01, 7.8714e-02, ..., -1.6565e-02, + -2.2227e-02, 6.6462e-02], + [ 1.4325e-02, 3.0944e-02, -2.8314e-02, ..., -6.8235e-02, + -9.4947e-02, -5.5364e-03], + [-1.2593e-01, -1.2112e-02, -8.2851e-03, ..., -1.2762e-01, + 9.6167e-02, 1.4126e-02]], device='cuda:0'), grad: tensor([[ 1.0324e-04, 4.8757e-04, 2.7323e-04, ..., 5.7191e-05, + 7.9155e-04, 1.6003e-03], + [ 4.5449e-05, 4.8637e-04, 4.9877e-04, ..., 4.2248e-04, + 1.0424e-03, -3.2825e-03], + [-9.6369e-04, -9.7370e-04, -4.2272e-04, ..., -1.8330e-03, + 5.3024e-04, 1.3409e-03], + ..., + [ 2.3556e-04, 3.2544e-04, 1.0562e-04, ..., 4.6909e-05, + 1.0996e-03, 2.9564e-03], + [ 1.5211e-03, 1.6518e-03, -6.3705e-04, ..., 1.7834e-03, + -1.2789e-03, -3.4285e-04], + [ 2.2516e-05, -9.6917e-05, -6.0606e-04, ..., -7.5531e-04, + -5.4054e-03, -7.6332e-03]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0163, -0.0005, 0.0001, 0.0160, -0.0054, -0.0067, 0.0108, 0.0250, + -0.0296, 0.0440], device='cuda:0'), grad: tensor([ 0.0023, 0.0028, 0.0240, -0.0028, -0.0018, -0.0387, 0.0340, 0.0067, + 0.0066, -0.0330], device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 216.92, cls_loss 0.5278 cls_loss_mapping 0.0090 cls_loss_causal 0.5052 re_mapping 0.0087 re_causal 0.0237 /// teacc 98.57 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0691, 0.0626, -0.0833, ..., -0.0966, -0.0720, -0.0033], + [-0.0498, -0.1194, -0.0003, ..., -0.0547, -0.0339, -0.0667], + [ 0.0278, -0.0821, 0.0449, ..., 0.1514, -0.1047, -0.0431], + ..., + [-0.0769, -0.1217, 0.0786, ..., -0.0175, -0.0227, 0.0655], + [ 0.0148, 0.0319, -0.0282, ..., -0.0695, -0.0932, -0.0053], + [-0.1248, -0.0127, -0.0084, ..., -0.1278, 0.0959, 0.0144]], + device='cuda:0'), grad: tensor([[ 2.3842e-03, 2.6588e-03, 4.3809e-05, ..., 1.1406e-03, + 7.8773e-04, 2.0008e-03], + [ 8.4019e-04, 5.0020e-04, 3.2568e-04, ..., 2.0766e-04, + 1.4229e-03, 3.1071e-03], + [ 1.1396e-03, 2.8992e-04, -4.4370e-04, ..., 2.1305e-03, + -1.4696e-03, 4.0970e-03], + ..., + [ 7.1764e-04, 4.4465e-04, 3.4380e-04, ..., 6.1893e-04, + 1.4296e-03, 7.2908e-04], + [-1.2405e-02, -1.9730e-02, 2.8157e-04, ..., -7.2746e-03, + -4.0984e-04, -5.5580e-03], + [-2.4338e-03, 4.9114e-04, -4.4537e-04, ..., -3.7327e-03, + -4.3321e-04, -9.2697e-03]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0166, -0.0005, 0.0007, 0.0164, -0.0049, -0.0063, 0.0100, 0.0239, + -0.0305, 0.0447], device='cuda:0'), grad: tensor([ 0.0314, 0.0118, -0.0206, 0.0099, 0.0342, -0.0346, 0.0443, 0.0037, + -0.0460, -0.0342], device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 216.81, cls_loss 0.4971 cls_loss_mapping 0.0082 cls_loss_causal 0.4723 re_mapping 0.0086 re_causal 0.0208 /// teacc 98.61 lr 0.00010000 +Epoch 177, weight, value: tensor([[-6.9226e-02, 6.2831e-02, -8.3256e-02, ..., -9.6361e-02, + -7.3134e-02, -3.8521e-03], + [-5.0508e-02, -1.2032e-01, -2.3686e-05, ..., -5.3952e-02, + -3.3701e-02, -6.6643e-02], + [ 2.8754e-02, -8.1108e-02, 4.4591e-02, ..., 1.5136e-01, + -1.0379e-01, -4.1775e-02], + ..., + [-7.7369e-02, -1.2254e-01, 7.8718e-02, ..., -1.9078e-02, + -2.1372e-02, 6.6297e-02], + [ 1.6148e-02, 3.2348e-02, -2.7663e-02, ..., -6.8503e-02, + -9.4267e-02, -6.3687e-03], + [-1.2572e-01, -1.2485e-02, -9.2211e-03, ..., -1.2760e-01, + 9.5618e-02, 1.3852e-02]], device='cuda:0'), grad: tensor([[ 0.0006, -0.0041, 0.0006, ..., 0.0003, 0.0016, 0.0021], + [ 0.0006, 0.0012, 0.0012, ..., 0.0006, 0.0068, 0.0051], + [ 0.0058, 0.0007, 0.0007, ..., 0.0134, 0.0007, 0.0014], + ..., + [-0.0049, 0.0008, 0.0015, ..., -0.0129, -0.0002, 0.0018], + [-0.0024, -0.0055, -0.0152, ..., -0.0119, -0.0145, -0.0186], + [ 0.0003, 0.0007, 0.0086, ..., 0.0082, -0.0003, 0.0009]], + device='cuda:0') +Epoch 177, bias, value: tensor([ 1.6456e-02, 1.5497e-05, 1.5415e-03, 1.6224e-02, -4.8792e-03, + -7.0658e-03, 1.0408e-02, 2.4556e-02, -3.0561e-02, 4.3433e-02], + device='cuda:0'), grad: tensor([-0.0044, 0.0474, 0.0443, 0.0249, 0.0278, -0.0073, 0.0330, -0.0375, + -0.1026, -0.0257], device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 216.53, cls_loss 0.5477 cls_loss_mapping 0.0083 cls_loss_causal 0.5178 re_mapping 0.0077 re_causal 0.0203 /// teacc 98.70 lr 0.00010000 +Epoch 178, weight, value: tensor([[-6.8983e-02, 6.2895e-02, -8.3069e-02, ..., -9.6342e-02, + -7.4834e-02, -4.8973e-03], + [-5.1704e-02, -1.2068e-01, 6.7164e-05, ..., -5.2615e-02, + -3.4823e-02, -6.6787e-02], + [ 2.9235e-02, -8.1604e-02, 4.5308e-02, ..., 1.5170e-01, + -1.0453e-01, -4.1781e-02], + ..., + [-7.9470e-02, -1.2269e-01, 7.9291e-02, ..., -1.8789e-02, + -2.1512e-02, 6.5551e-02], + [ 1.6077e-02, 3.1168e-02, -2.6168e-02, ..., -6.7920e-02, + -9.2890e-02, -5.2729e-03], + [-1.2579e-01, -1.1817e-02, -9.8941e-03, ..., -1.2812e-01, + 9.6352e-02, 1.3815e-02]], device='cuda:0'), grad: tensor([[ 1.9627e-03, -1.2070e-04, 1.3137e-04, ..., -7.3195e-05, + 6.9332e-04, 9.2649e-04], + [-8.6746e-03, -8.0566e-03, -8.7690e-04, ..., 1.2517e-05, + -7.0877e-03, -5.4169e-03], + [ 3.0899e-04, 2.8276e-04, 6.6578e-05, ..., -6.9976e-05, + 6.6423e-04, 1.3037e-03], + ..., + [ 1.8954e-04, 2.2781e-04, 2.9302e-04, ..., 9.3460e-05, + 9.6941e-04, -2.8896e-03], + [ 4.3106e-03, 3.5820e-03, 1.4842e-04, ..., 1.5959e-05, + 7.7581e-04, 9.6416e-04], + [ 7.1347e-05, 2.2316e-04, -3.9220e-04, ..., 2.6155e-04, + 5.2977e-04, 8.8501e-04]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0156, -0.0007, 0.0011, 0.0166, -0.0050, -0.0065, 0.0114, 0.0237, + -0.0303, 0.0442], device='cuda:0'), grad: tensor([ 0.0113, -0.1301, 0.0178, 0.0117, 0.0026, 0.0127, 0.0325, -0.0155, + 0.0246, 0.0324], device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 216.61, cls_loss 0.5509 cls_loss_mapping 0.0122 cls_loss_causal 0.5205 re_mapping 0.0087 re_causal 0.0211 /// teacc 98.77 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0686, 0.0637, -0.0842, ..., -0.0980, -0.0742, -0.0059], + [-0.0521, -0.1204, -0.0009, ..., -0.0533, -0.0360, -0.0661], + [ 0.0295, -0.0818, 0.0454, ..., 0.1518, -0.1053, -0.0427], + ..., + [-0.0785, -0.1244, 0.0793, ..., -0.0180, -0.0215, 0.0655], + [ 0.0157, 0.0309, -0.0255, ..., -0.0673, -0.0925, -0.0058], + [-0.1258, -0.0115, -0.0104, ..., -0.1282, 0.0958, 0.0137]], + device='cuda:0'), grad: tensor([[ 4.9686e-04, 3.8600e-04, 6.1703e-04, ..., 4.3178e-04, + -6.4468e-04, 6.3848e-04], + [ 6.1274e-04, 4.1056e-04, 4.0770e-04, ..., 1.4627e-04, + 4.8637e-04, 5.1785e-04], + [-1.5661e-05, 1.1616e-03, 2.9373e-03, ..., 5.0507e-03, + 2.1915e-03, 2.5406e-03], + ..., + [-5.6982e-04, 3.4595e-04, -2.7275e-03, ..., -2.5940e-03, + -1.8454e-03, 4.8327e-04], + [ 1.3342e-03, 1.7653e-03, 1.3113e-03, ..., 1.2770e-03, + 2.1553e-03, 1.3561e-03], + [ 2.4605e-04, -2.4509e-03, -4.2992e-03, ..., -5.7030e-03, + -2.7771e-03, -5.8594e-03]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0156, -0.0015, 0.0009, 0.0166, -0.0043, -0.0052, 0.0107, 0.0242, + -0.0302, 0.0433], device='cuda:0'), grad: tensor([-1.2779e-02, -2.3270e-02, -2.6710e-06, -1.4252e-02, 2.7664e-02, + 2.5959e-03, -6.1989e-03, 6.1760e-03, 2.5894e-02, -5.8327e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 216.37, cls_loss 0.5181 cls_loss_mapping 0.0072 cls_loss_causal 0.4903 re_mapping 0.0089 re_causal 0.0224 /// teacc 98.60 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0688, 0.0632, -0.0836, ..., -0.0969, -0.0751, -0.0056], + [-0.0517, -0.1213, -0.0018, ..., -0.0544, -0.0361, -0.0665], + [ 0.0297, -0.0810, 0.0450, ..., 0.1524, -0.1046, -0.0435], + ..., + [-0.0795, -0.1255, 0.0804, ..., -0.0175, -0.0221, 0.0657], + [ 0.0155, 0.0311, -0.0255, ..., -0.0674, -0.0919, -0.0062], + [-0.1260, -0.0119, -0.0106, ..., -0.1279, 0.0952, 0.0137]], + device='cuda:0'), grad: tensor([[ 2.2209e-04, 2.4586e-03, 4.5815e-03, ..., 2.4527e-05, + 5.4073e-04, 6.1846e-04], + [-1.6487e-04, -1.7321e-04, 2.4259e-04, ..., 9.1314e-05, + -6.3372e-04, 8.1921e-04], + [-2.4188e-04, 8.0919e-04, 5.7173e-04, ..., -5.7840e-04, + 6.5708e-04, 9.9945e-04], + ..., + [ 8.6212e-04, 5.1537e-03, 3.3684e-03, ..., 7.5586e-06, + 1.4372e-03, 2.2411e-03], + [-3.7956e-03, -2.1606e-02, -1.5030e-02, ..., 3.1382e-05, + -4.2992e-03, -3.8090e-03], + [ 4.0007e-04, 1.1124e-02, 5.9395e-03, ..., 6.3181e-05, + 8.8978e-04, 1.2617e-03]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0153, -0.0016, 0.0012, 0.0180, -0.0043, -0.0050, 0.0101, 0.0235, + -0.0301, 0.0429], device='cuda:0'), grad: tensor([ 0.0115, -0.0178, 0.0147, 0.0251, -0.0316, 0.0199, 0.0164, 0.0266, + -0.0986, 0.0337], device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 216.50, cls_loss 0.5018 cls_loss_mapping 0.0086 cls_loss_causal 0.4725 re_mapping 0.0090 re_causal 0.0230 /// teacc 98.76 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0688, 0.0638, -0.0846, ..., -0.0960, -0.0751, -0.0054], + [-0.0518, -0.1211, -0.0006, ..., -0.0545, -0.0349, -0.0668], + [ 0.0294, -0.0814, 0.0458, ..., 0.1531, -0.1065, -0.0445], + ..., + [-0.0804, -0.1274, 0.0802, ..., -0.0186, -0.0236, 0.0654], + [ 0.0161, 0.0307, -0.0258, ..., -0.0670, -0.0919, -0.0058], + [-0.1267, -0.0120, -0.0095, ..., -0.1271, 0.0973, 0.0147]], + device='cuda:0'), grad: tensor([[ 2.7351e-03, 5.9624e-03, 2.1064e-04, ..., 6.7018e-06, + 2.9111e-04, 6.8569e-04], + [ 1.9050e-04, 3.5286e-04, 1.7750e-04, ..., 4.9826e-07, + 2.9922e-04, 6.4278e-04], + [ 1.5736e-04, -4.2939e-04, 1.2684e-04, ..., -1.0267e-05, + 1.6272e-04, 4.1127e-04], + ..., + [ 1.1116e-05, 3.0088e-04, 1.3471e-04, ..., 1.4426e-06, + 1.9193e-04, 3.9673e-04], + [-3.6945e-03, -7.2212e-03, 1.2708e-04, ..., 2.3246e-06, + 1.5438e-04, 3.7432e-04], + [ 2.0695e-04, 4.1223e-04, 1.0834e-03, ..., 2.8044e-05, + 1.4076e-03, 1.8730e-03]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0155, -0.0017, -0.0001, 0.0180, -0.0045, -0.0044, 0.0107, 0.0229, + -0.0302, 0.0440], device='cuda:0'), grad: tensor([ 0.0170, 0.0111, -0.0210, -0.0262, 0.0103, -0.0126, -0.0118, 0.0079, + 0.0072, 0.0182], device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 216.39, cls_loss 0.5808 cls_loss_mapping 0.0116 cls_loss_causal 0.5517 re_mapping 0.0081 re_causal 0.0210 /// teacc 98.77 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0706, 0.0629, -0.0850, ..., -0.0960, -0.0750, -0.0038], + [-0.0513, -0.1209, -0.0005, ..., -0.0558, -0.0356, -0.0685], + [ 0.0299, -0.0816, 0.0454, ..., 0.1529, -0.1066, -0.0438], + ..., + [-0.0793, -0.1266, 0.0803, ..., -0.0181, -0.0233, 0.0664], + [ 0.0169, 0.0323, -0.0266, ..., -0.0679, -0.0917, -0.0065], + [-0.1283, -0.0112, -0.0094, ..., -0.1270, 0.0964, 0.0136]], + device='cuda:0'), grad: tensor([[-3.1967e-03, -2.1713e-02, 2.2292e-04, ..., 1.9050e-04, + 2.1434e-04, -2.6989e-03], + [ 8.4925e-04, 1.1768e-03, -1.5411e-03, ..., 1.6177e-04, + -1.4365e-04, -1.2684e-03], + [ 4.5586e-04, 8.3017e-04, -6.9199e-03, ..., -8.1968e-04, + -4.7255e-04, -4.8327e-04], + ..., + [ 2.6989e-04, 4.5824e-04, 3.8090e-03, ..., 4.6396e-04, + 1.6289e-03, 1.2369e-03], + [-2.9011e-03, -2.2869e-03, -1.4725e-02, ..., -3.2485e-05, + 2.9594e-05, 6.4802e-04], + [ 8.3780e-04, 1.4811e-03, 1.5564e-02, ..., -6.9857e-04, + -6.4039e-04, -2.7943e-04]], device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0159, -0.0016, 0.0009, 0.0181, -0.0041, -0.0043, 0.0094, 0.0239, + -0.0315, 0.0436], device='cuda:0'), grad: tensor([-0.0418, 0.0031, -0.0372, 0.0202, -0.0018, -0.0002, 0.0602, 0.0269, + -0.0276, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 216.34, cls_loss 0.5296 cls_loss_mapping 0.0077 cls_loss_causal 0.5123 re_mapping 0.0084 re_causal 0.0211 /// teacc 98.75 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0698, 0.0636, -0.0861, ..., -0.0965, -0.0755, -0.0041], + [-0.0524, -0.1213, -0.0013, ..., -0.0570, -0.0354, -0.0683], + [ 0.0298, -0.0816, 0.0455, ..., 0.1528, -0.1059, -0.0422], + ..., + [-0.0790, -0.1269, 0.0805, ..., -0.0172, -0.0237, 0.0658], + [ 0.0186, 0.0331, -0.0268, ..., -0.0673, -0.0911, -0.0067], + [-0.1274, -0.0103, -0.0100, ..., -0.1273, 0.0972, 0.0139]], + device='cuda:0'), grad: tensor([[ 1.8990e-04, 2.5864e-03, 3.7169e-04, ..., 3.4547e-04, + 2.0623e-05, 1.7083e-04], + [ 7.5436e-04, 1.7011e-04, 2.8372e-04, ..., 3.0637e-04, + -6.3610e-04, -1.1911e-03], + [ 5.4169e-04, -5.7030e-04, 1.1091e-03, ..., -2.4014e-03, + 7.8440e-05, 2.4462e-04], + ..., + [-3.2768e-03, 1.1748e-04, -1.3084e-02, ..., -4.7340e-03, + -1.1726e-02, -1.0040e-02], + [ 1.8177e-03, 9.1600e-04, 5.3692e-04, ..., 5.4502e-04, + 2.9540e-04, -2.6560e-04], + [ 1.5106e-02, 1.0824e-03, 7.5798e-03, ..., 3.2883e-03, + 8.0643e-03, 7.1869e-03]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0161, -0.0023, 0.0018, 0.0177, -0.0045, -0.0057, 0.0101, 0.0235, + -0.0307, 0.0442], device='cuda:0'), grad: tensor([ 0.0380, -0.0188, -0.0033, -0.0385, -0.0118, 0.0330, 0.0052, -0.0237, + -0.0150, 0.0350], device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 216.21, cls_loss 0.5577 cls_loss_mapping 0.0087 cls_loss_causal 0.5277 re_mapping 0.0084 re_causal 0.0209 /// teacc 98.62 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0710, 0.0632, -0.0860, ..., -0.0969, -0.0744, -0.0020], + [-0.0519, -0.1215, 0.0003, ..., -0.0571, -0.0353, -0.0678], + [ 0.0303, -0.0803, 0.0449, ..., 0.1531, -0.1059, -0.0410], + ..., + [-0.0796, -0.1274, 0.0811, ..., -0.0177, -0.0246, 0.0644], + [ 0.0181, 0.0325, -0.0265, ..., -0.0684, -0.0908, -0.0078], + [-0.1290, -0.0112, -0.0107, ..., -0.1285, 0.0974, 0.0138]], + device='cuda:0'), grad: tensor([[ 6.3562e-04, 4.0746e-04, -2.8801e-04, ..., 1.0359e-04, + -2.3766e-03, -2.0695e-03], + [ 8.9169e-04, 3.3736e-04, 7.6962e-04, ..., 5.4806e-05, + 3.1681e-03, 4.9133e-03], + [-3.2825e-03, 8.1730e-04, 2.0921e-04, ..., 1.9491e-04, + -8.6117e-04, -1.0437e-02], + ..., + [ 8.6069e-04, 2.2277e-05, 6.6109e-03, ..., 6.7115e-05, + 1.7691e-03, -4.4785e-03], + [-6.1035e-04, -1.2711e-02, 5.2643e-04, ..., 4.0269e-04, + -9.5062e-03, 2.8629e-03], + [ 8.9264e-04, 6.5565e-04, -8.8806e-03, ..., 1.6487e-04, + -6.3133e-03, 2.8000e-03]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0159, -0.0009, 0.0014, 0.0178, -0.0036, -0.0050, 0.0088, 0.0233, + -0.0314, 0.0439], device='cuda:0'), grad: tensor([-0.0292, 0.0395, -0.0434, -0.0021, 0.0232, 0.0155, 0.0194, -0.0145, + -0.0088, 0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 216.85, cls_loss 0.5183 cls_loss_mapping 0.0103 cls_loss_causal 0.4935 re_mapping 0.0084 re_causal 0.0210 /// teacc 98.68 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0710, 0.0640, -0.0862, ..., -0.0964, -0.0751, -0.0018], + [-0.0523, -0.1211, 0.0002, ..., -0.0570, -0.0359, -0.0688], + [ 0.0285, -0.0806, 0.0449, ..., 0.1528, -0.1060, -0.0414], + ..., + [-0.0798, -0.1281, 0.0809, ..., -0.0168, -0.0248, 0.0651], + [ 0.0188, 0.0328, -0.0266, ..., -0.0689, -0.0908, -0.0077], + [-0.1295, -0.0107, -0.0102, ..., -0.1289, 0.0977, 0.0142]], + device='cuda:0'), grad: tensor([[ 3.2377e-04, -1.6010e-04, 1.1629e-04, ..., 6.9261e-05, + 9.2649e-04, 6.9332e-04], + [ 3.5453e-04, 7.1287e-04, 5.1737e-04, ..., 1.8024e-04, + 2.4872e-03, 2.3861e-03], + [-6.3324e-04, 1.5604e-04, 3.9768e-04, ..., -5.0688e-04, + 4.0102e-04, -1.4758e-04], + ..., + [ 4.4680e-04, 5.6553e-04, -4.7569e-03, ..., -1.2417e-03, + 8.3733e-04, -1.1032e-02], + [ 4.1885e-03, 5.6114e-03, 1.0923e-05, ..., 2.7084e-04, + -2.5120e-03, -3.1033e-03], + [-9.0027e-03, -5.9967e-03, 4.1509e-04, ..., 2.7299e-04, + 1.7614e-03, 2.5501e-03]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0164, -0.0014, 0.0013, 0.0175, -0.0035, -0.0060, 0.0099, 0.0229, + -0.0318, 0.0448], device='cuda:0'), grad: tensor([ 0.0282, 0.0336, -0.0109, 0.0173, 0.0374, 0.0182, -0.0299, -0.0221, + -0.0407, -0.0311], device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 216.94, cls_loss 0.5342 cls_loss_mapping 0.0061 cls_loss_causal 0.5082 re_mapping 0.0082 re_causal 0.0214 /// teacc 98.75 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0696, 0.0645, -0.0867, ..., -0.0968, -0.0759, -0.0024], + [-0.0519, -0.1217, 0.0002, ..., -0.0575, -0.0361, -0.0682], + [ 0.0294, -0.0803, 0.0466, ..., 0.1535, -0.1050, -0.0397], + ..., + [-0.0814, -0.1290, 0.0818, ..., -0.0174, -0.0242, 0.0659], + [ 0.0183, 0.0325, -0.0274, ..., -0.0681, -0.0915, -0.0084], + [-0.1297, -0.0114, -0.0099, ..., -0.1271, 0.0976, 0.0142]], + device='cuda:0'), grad: tensor([[ 8.3303e-04, -3.3855e-05, -5.8603e-04, ..., 2.7204e-04, + 4.0889e-04, 8.6355e-04], + [ 4.8780e-04, 2.7752e-04, 1.3494e-03, ..., 5.8222e-04, + 1.7319e-03, 1.6975e-03], + [ 9.0170e-04, 5.0211e-04, 2.4090e-03, ..., 1.0862e-03, + 2.1095e-03, 2.4433e-03], + ..., + [ 4.9162e-04, 2.3055e-04, 1.3809e-03, ..., 4.8065e-04, + 1.4496e-03, 1.7891e-03], + [ 9.9564e-03, 1.1414e-02, 8.2850e-05, ..., 2.2388e-04, + 1.9855e-03, 7.7629e-04], + [ 9.2506e-04, 1.1759e-03, 2.2068e-03, ..., 1.1864e-03, + 3.1490e-03, 2.8362e-03]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0165, -0.0014, 0.0029, 0.0175, -0.0042, -0.0068, 0.0099, 0.0232, + -0.0317, 0.0442], device='cuda:0'), grad: tensor([-0.0468, 0.0216, 0.0267, -0.0113, 0.0167, -0.0298, -0.0475, 0.0184, + 0.0296, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 216.66, cls_loss 0.5646 cls_loss_mapping 0.0094 cls_loss_causal 0.5401 re_mapping 0.0084 re_causal 0.0225 /// teacc 98.83 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0693, 0.0645, -0.0861, ..., -0.0965, -0.0756, -0.0031], + [-0.0542, -0.1233, 0.0013, ..., -0.0572, -0.0358, -0.0688], + [ 0.0305, -0.0786, 0.0456, ..., 0.1533, -0.1056, -0.0389], + ..., + [-0.0808, -0.1295, 0.0817, ..., -0.0183, -0.0241, 0.0656], + [ 0.0167, 0.0318, -0.0274, ..., -0.0684, -0.0914, -0.0082], + [-0.1312, -0.0121, -0.0092, ..., -0.1263, 0.0978, 0.0137]], + device='cuda:0'), grad: tensor([[ 1.4734e-04, 8.6010e-05, 7.6652e-05, ..., -3.1944e-06, + 5.5885e-04, 1.9379e-03], + [ 3.8981e-04, 2.8944e-04, 1.5175e-04, ..., 1.4659e-06, + 1.2455e-03, 3.2921e-03], + [ 7.9966e-04, 3.3569e-04, 1.6582e-04, ..., 3.5703e-05, + 4.7255e-04, 1.8969e-03], + ..., + [-1.1244e-03, 1.7166e-04, 1.5192e-03, ..., 3.1069e-06, + -2.8992e-04, -2.0390e-03], + [ 4.2953e-03, 1.1473e-03, -4.1151e-04, ..., 2.6774e-04, + -2.1038e-03, -3.5591e-03], + [-1.1511e-03, 2.3758e-04, -3.6373e-03, ..., 4.9211e-06, + -2.4357e-03, -7.3624e-03]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0171, -0.0021, 0.0033, 0.0170, -0.0036, -0.0070, 0.0098, 0.0229, + -0.0311, 0.0437], device='cuda:0'), grad: tensor([ 0.0224, 0.0374, 0.0223, 0.0254, -0.0030, 0.0087, 0.0277, -0.0341, + -0.0262, -0.0806], device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 216.91, cls_loss 0.5263 cls_loss_mapping 0.0071 cls_loss_causal 0.4967 re_mapping 0.0078 re_causal 0.0206 /// teacc 98.77 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0704, 0.0633, -0.0877, ..., -0.0962, -0.0771, -0.0038], + [-0.0540, -0.1234, 0.0010, ..., -0.0564, -0.0372, -0.0700], + [ 0.0296, -0.0784, 0.0464, ..., 0.1532, -0.1057, -0.0388], + ..., + [-0.0820, -0.1298, 0.0817, ..., -0.0177, -0.0239, 0.0660], + [ 0.0159, 0.0316, -0.0276, ..., -0.0696, -0.0922, -0.0093], + [-0.1303, -0.0139, -0.0073, ..., -0.1273, 0.0988, 0.0147]], + device='cuda:0'), grad: tensor([[ 3.3545e-04, -2.7771e-03, 1.1568e-03, ..., 7.8440e-04, + 3.6502e-04, 4.2152e-04], + [ 1.0869e-06, -1.1408e-04, -8.9407e-04, ..., 1.8263e-04, + 9.5367e-04, 5.0468e-03], + [ 1.0037e-04, 2.2709e-04, 1.1539e-03, ..., -1.4472e-04, + 3.6526e-04, 7.3910e-04], + ..., + [ 1.4023e-02, 2.5501e-03, 5.6458e-03, ..., -3.0661e-04, + 2.7037e-04, -2.4261e-03], + [-7.4158e-03, -1.7667e-04, -5.1928e-04, ..., 8.6927e-04, + 3.9148e-04, 3.7646e-04], + [ 2.0492e-04, 4.3392e-04, 2.9926e-03, ..., 2.2926e-03, + 2.2662e-04, 5.5218e-04]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0163, -0.0016, 0.0030, 0.0175, -0.0037, -0.0073, 0.0098, 0.0231, + -0.0316, 0.0444], device='cuda:0'), grad: tensor([ 0.0099, 0.0459, 0.0204, -0.0448, -0.0078, -0.0203, -0.0212, 0.0053, + -0.0089, 0.0214], device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 216.58, cls_loss 0.5198 cls_loss_mapping 0.0073 cls_loss_causal 0.4878 re_mapping 0.0087 re_causal 0.0211 /// teacc 98.77 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0711, 0.0629, -0.0892, ..., -0.0972, -0.0784, -0.0051], + [-0.0552, -0.1234, 0.0005, ..., -0.0564, -0.0376, -0.0705], + [ 0.0303, -0.0782, 0.0482, ..., 0.1547, -0.1061, -0.0385], + ..., + [-0.0828, -0.1296, 0.0808, ..., -0.0190, -0.0237, 0.0654], + [ 0.0176, 0.0324, -0.0268, ..., -0.0688, -0.0903, -0.0080], + [-0.1309, -0.0124, -0.0076, ..., -0.1280, 0.0983, 0.0148]], + device='cuda:0'), grad: tensor([[-0.0080, -0.0232, 0.0006, ..., 0.0003, 0.0003, -0.0036], + [-0.0103, -0.0087, 0.0011, ..., -0.0158, 0.0009, 0.0024], + [ 0.0105, 0.0090, 0.0044, ..., 0.0141, 0.0005, 0.0049], + ..., + [ 0.0002, 0.0002, -0.0045, ..., -0.0052, 0.0020, -0.0010], + [ 0.0088, 0.0076, 0.0010, ..., 0.0006, 0.0007, 0.0035], + [-0.0004, 0.0002, -0.0055, ..., 0.0001, -0.0074, -0.0091]], + device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0158, -0.0019, 0.0029, 0.0163, -0.0036, -0.0080, 0.0110, 0.0238, + -0.0305, 0.0442], device='cuda:0'), grad: tensor([-0.0474, -0.0136, 0.0615, -0.0274, 0.0132, 0.0102, 0.0252, -0.0203, + 0.0308, -0.0322], device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 216.44, cls_loss 0.5447 cls_loss_mapping 0.0072 cls_loss_causal 0.5212 re_mapping 0.0081 re_causal 0.0212 /// teacc 98.84 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0712, 0.0627, -0.0899, ..., -0.0971, -0.0783, -0.0046], + [-0.0561, -0.1241, 0.0017, ..., -0.0561, -0.0365, -0.0706], + [ 0.0289, -0.0790, 0.0488, ..., 0.1554, -0.1069, -0.0401], + ..., + [-0.0825, -0.1299, 0.0816, ..., -0.0189, -0.0231, 0.0660], + [ 0.0176, 0.0326, -0.0275, ..., -0.0698, -0.0907, -0.0078], + [-0.1312, -0.0128, -0.0083, ..., -0.1284, 0.0980, 0.0151]], + device='cuda:0'), grad: tensor([[ 1.9535e-05, 1.9598e-04, 4.6778e-04, ..., 9.0778e-05, + 3.3855e-04, 9.8324e-04], + [ 6.1750e-05, 1.5855e-05, 9.8801e-04, ..., 1.2226e-03, + 8.4925e-04, 1.7462e-03], + [ 5.1349e-05, 5.9932e-05, 1.0319e-03, ..., 9.4271e-04, + 8.8358e-04, 1.6222e-03], + ..., + [ 6.4522e-06, 2.7478e-05, -1.7195e-03, ..., 1.1473e-03, + -4.4608e-04, -1.5688e-03], + [ 2.6986e-05, 1.5485e-04, 6.5470e-04, ..., 6.1512e-04, + 4.3678e-04, 1.0328e-03], + [ 2.4781e-05, 2.4453e-05, -5.7191e-05, ..., -3.4332e-03, + -1.5390e-04, -7.5054e-04]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0166, -0.0016, 0.0023, 0.0167, -0.0040, -0.0080, 0.0107, 0.0236, + -0.0312, 0.0446], device='cuda:0'), grad: tensor([ 0.0163, 0.0238, 0.0237, 0.0284, -0.0654, 0.0158, -0.0131, -0.0301, + 0.0214, -0.0208], device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 216.64, cls_loss 0.5371 cls_loss_mapping 0.0071 cls_loss_causal 0.5194 re_mapping 0.0083 re_causal 0.0213 /// teacc 98.82 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.0706, 0.0630, -0.0906, ..., -0.0958, -0.0786, -0.0031], + [-0.0570, -0.1242, 0.0014, ..., -0.0555, -0.0377, -0.0718], + [ 0.0293, -0.0785, 0.0482, ..., 0.1557, -0.1081, -0.0396], + ..., + [-0.0832, -0.1309, 0.0820, ..., -0.0191, -0.0231, 0.0646], + [ 0.0172, 0.0319, -0.0267, ..., -0.0709, -0.0910, -0.0086], + [-0.1315, -0.0132, -0.0081, ..., -0.1287, 0.0978, 0.0159]], + device='cuda:0'), grad: tensor([[ 1.9610e-04, 1.0914e-04, 8.9979e-04, ..., 1.1154e-05, + 2.8396e-04, 7.2622e-04], + [ 2.5654e-04, 1.3328e-04, 1.0223e-03, ..., 7.9811e-05, + 5.4741e-04, 2.3327e-03], + [ 6.3467e-04, 4.9591e-04, 1.3313e-03, ..., -6.9380e-05, + 6.8378e-04, 7.7677e-04], + ..., + [-2.8172e-03, -2.5558e-03, -5.3940e-03, ..., 1.7095e-04, + -4.8409e-03, -3.1624e-03], + [ 1.2445e-03, 1.4887e-03, -1.6421e-05, ..., 7.0751e-05, + 2.0428e-03, -3.2120e-03], + [ 1.0128e-03, 7.6342e-04, 2.4738e-03, ..., 1.4400e-04, + 1.3084e-03, 1.3790e-03]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0161, -0.0022, 0.0025, 0.0165, -0.0037, -0.0067, 0.0108, 0.0232, + -0.0319, 0.0452], device='cuda:0'), grad: tensor([ 0.0227, 0.0333, -0.0011, -0.0203, 0.0199, 0.0092, -0.0012, -0.0657, + -0.0272, 0.0304], device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 216.51, cls_loss 0.5427 cls_loss_mapping 0.0081 cls_loss_causal 0.5140 re_mapping 0.0087 re_causal 0.0215 /// teacc 98.69 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0704, 0.0616, -0.0904, ..., -0.0967, -0.0787, -0.0026], + [-0.0571, -0.1241, 0.0012, ..., -0.0553, -0.0378, -0.0721], + [ 0.0297, -0.0774, 0.0480, ..., 0.1557, -0.1077, -0.0401], + ..., + [-0.0837, -0.1320, 0.0825, ..., -0.0191, -0.0222, 0.0657], + [ 0.0172, 0.0320, -0.0270, ..., -0.0719, -0.0909, -0.0077], + [-0.1320, -0.0137, -0.0087, ..., -0.1286, 0.0973, 0.0154]], + device='cuda:0'), grad: tensor([[ 7.2813e-04, 1.2159e-03, 1.3571e-03, ..., 4.4537e-04, + 5.7906e-05, 2.0390e-03], + [ 1.1492e-04, 1.6475e-04, 1.0910e-03, ..., 4.4727e-04, + 1.9908e-04, 3.6120e-04], + [ 3.1242e-03, 3.7327e-03, 5.2643e-03, ..., 3.2291e-03, + 1.0175e-04, 4.5433e-03], + ..., + [-3.8671e-04, 4.6301e-04, -2.2755e-03, ..., -2.1706e-03, + 2.4199e-04, -5.4550e-03], + [ 6.7616e-04, 1.1511e-03, -1.3533e-03, ..., -1.5192e-03, + 1.0586e-04, -1.5831e-04], + [-2.0373e-04, -3.0460e-03, -3.1052e-03, ..., 6.2847e-04, + 4.0603e-04, -9.0551e-04]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0158, -0.0026, 0.0033, 0.0165, -0.0035, -0.0066, 0.0099, 0.0235, + -0.0311, 0.0448], device='cuda:0'), grad: tensor([ 0.0189, -0.0224, 0.0502, 0.0236, 0.0162, -0.0197, -0.0289, -0.0415, + -0.0038, 0.0073], device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 216.77, cls_loss 0.5321 cls_loss_mapping 0.0077 cls_loss_causal 0.5007 re_mapping 0.0086 re_causal 0.0208 /// teacc 98.74 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0701, 0.0621, -0.0911, ..., -0.0974, -0.0788, -0.0029], + [-0.0572, -0.1236, -0.0007, ..., -0.0560, -0.0363, -0.0716], + [ 0.0308, -0.0775, 0.0476, ..., 0.1559, -0.1093, -0.0405], + ..., + [-0.0837, -0.1321, 0.0842, ..., -0.0180, -0.0225, 0.0649], + [ 0.0168, 0.0324, -0.0273, ..., -0.0730, -0.0913, -0.0082], + [-0.1336, -0.0139, -0.0080, ..., -0.1266, 0.0981, 0.0163]], + device='cuda:0'), grad: tensor([[-9.7990e-05, 8.8453e-05, 2.8849e-04, ..., -6.4313e-05, + 5.3167e-04, 3.1233e-04], + [ 3.2067e-04, 2.4509e-04, 1.5774e-03, ..., 4.2892e-04, + 1.2398e-03, 7.7009e-04], + [-1.1045e-04, 4.0817e-04, -3.8948e-03, ..., -2.8458e-03, + 7.0810e-04, -4.1618e-03], + ..., + [ 9.9659e-04, 5.2309e-04, 8.4610e-03, ..., 3.6621e-03, + 3.9177e-03, 5.1689e-03], + [-2.0754e-04, 1.0338e-03, 9.8991e-04, ..., 6.0320e-04, + 1.5535e-03, 7.5912e-04], + [-7.5626e-04, -1.7052e-03, -8.1253e-03, ..., -2.7618e-03, + -7.7820e-03, -2.6169e-03]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0163, -0.0017, 0.0017, 0.0163, -0.0040, -0.0062, 0.0093, 0.0238, + -0.0313, 0.0456], device='cuda:0'), grad: tensor([ 0.0031, 0.0083, -0.0169, 0.0035, 0.0117, 0.0094, -0.0297, 0.0329, + 0.0003, -0.0228], device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 216.98, cls_loss 0.5266 cls_loss_mapping 0.0058 cls_loss_causal 0.5044 re_mapping 0.0086 re_causal 0.0223 /// teacc 98.91 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0698, 0.0633, -0.0918, ..., -0.0965, -0.0778, -0.0015], + [-0.0578, -0.1240, 0.0008, ..., -0.0553, -0.0369, -0.0719], + [ 0.0308, -0.0781, 0.0467, ..., 0.1555, -0.1096, -0.0397], + ..., + [-0.0835, -0.1338, 0.0825, ..., -0.0193, -0.0234, 0.0639], + [ 0.0154, 0.0315, -0.0270, ..., -0.0743, -0.0913, -0.0074], + [-0.1333, -0.0137, -0.0070, ..., -0.1267, 0.0987, 0.0166]], + device='cuda:0'), grad: tensor([[ 8.8882e-04, 3.1681e-03, 9.8038e-04, ..., -7.0333e-04, + 6.6328e-04, 8.5640e-04], + [ 3.1066e-04, 3.8385e-04, 1.5259e-03, ..., 1.5831e-04, + 9.7942e-04, 1.3037e-03], + [ 3.6812e-04, 5.4121e-04, -2.4891e-04, ..., 6.4325e-04, + -3.9177e-03, -3.6755e-03], + ..., + [ 1.2720e-04, 2.1458e-04, -8.1015e-04, ..., 3.2455e-05, + 5.5361e-04, 4.2915e-04], + [ 2.2531e-04, -5.9175e-04, -1.7071e-03, ..., 4.6849e-05, + 8.2779e-04, -1.5936e-03], + [ 1.4985e-04, 1.0270e-04, 9.9373e-04, ..., 1.4067e-05, + 7.6246e-04, 1.1349e-03]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0165, -0.0010, 0.0016, 0.0160, -0.0040, -0.0066, 0.0093, 0.0230, + -0.0309, 0.0459], device='cuda:0'), grad: tensor([ 0.0311, 0.0349, -0.0034, 0.0198, -0.0981, 0.0213, 0.0436, -0.0105, + -0.0638, 0.0252], device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 217.01, cls_loss 0.5194 cls_loss_mapping 0.0088 cls_loss_causal 0.4913 re_mapping 0.0080 re_causal 0.0195 /// teacc 98.86 lr 0.00010000 +Epoch 195, weight, value: tensor([[-7.0703e-02, 6.2668e-02, -9.2921e-02, ..., -9.7764e-02, + -7.7635e-02, -1.8272e-03], + [-5.7739e-02, -1.2412e-01, -1.5446e-04, ..., -5.5505e-02, + -3.8477e-02, -7.2565e-02], + [ 3.0456e-02, -7.8659e-02, 4.6852e-02, ..., 1.5563e-01, + -1.0945e-01, -3.9049e-02], + ..., + [-8.3024e-02, -1.3367e-01, 8.2073e-02, ..., -1.8765e-02, + -2.3848e-02, 6.4101e-02], + [ 1.5413e-02, 3.1137e-02, -2.6366e-02, ..., -7.5537e-02, + -9.0539e-02, -8.5435e-03], + [-1.3345e-01, -1.3066e-02, -7.4456e-03, ..., -1.2657e-01, + 9.8735e-02, 1.6453e-02]], device='cuda:0'), grad: tensor([[-0.0025, -0.0077, 0.0005, ..., 0.0006, 0.0001, -0.0006], + [ 0.0022, 0.0021, 0.0009, ..., 0.0014, 0.0002, 0.0011], + [ 0.0027, 0.0030, -0.0005, ..., 0.0021, 0.0002, 0.0015], + ..., + [ 0.0004, 0.0003, 0.0006, ..., 0.0005, 0.0001, 0.0007], + [ 0.0014, 0.0032, 0.0011, ..., 0.0008, 0.0001, 0.0009], + [ 0.0005, 0.0015, 0.0005, ..., 0.0004, 0.0002, 0.0008]], + device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0150, -0.0016, 0.0015, 0.0166, -0.0030, -0.0060, 0.0094, 0.0244, + -0.0314, 0.0451], device='cuda:0'), grad: tensor([-0.0010, 0.0493, -0.0003, 0.0277, -0.0443, -0.0286, -0.0389, 0.0160, + 0.0032, 0.0169], device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 216.79, cls_loss 0.5495 cls_loss_mapping 0.0095 cls_loss_causal 0.5190 re_mapping 0.0087 re_causal 0.0219 /// teacc 98.85 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0709, 0.0627, -0.0928, ..., -0.0968, -0.0789, -0.0015], + [-0.0579, -0.1244, -0.0015, ..., -0.0566, -0.0379, -0.0731], + [ 0.0311, -0.0782, 0.0467, ..., 0.1554, -0.1092, -0.0388], + ..., + [-0.0846, -0.1336, 0.0842, ..., -0.0167, -0.0240, 0.0638], + [ 0.0172, 0.0318, -0.0269, ..., -0.0748, -0.0910, -0.0083], + [-0.1342, -0.0135, -0.0077, ..., -0.1289, 0.0984, 0.0156]], + device='cuda:0'), grad: tensor([[ 3.0947e-04, 6.8998e-04, 1.2617e-03, ..., 7.7486e-04, + 6.1655e-04, 1.4372e-03], + [ 2.1368e-05, 7.7128e-05, -3.4499e-04, ..., -6.5422e-04, + -1.3523e-03, 1.3475e-03], + [ 4.7326e-04, 9.0694e-04, 9.6142e-05, ..., 7.9918e-04, + 9.6512e-04, 1.3504e-03], + ..., + [ 1.2946e-04, -5.2273e-05, 1.2253e-02, ..., -1.3018e-03, + 1.1606e-03, 2.2240e-03], + [ 9.0122e-05, -1.0815e-03, 1.3266e-03, ..., 6.8903e-04, + 8.3637e-04, -2.5215e-03], + [ 5.7459e-04, 1.3638e-03, -1.8797e-03, ..., 4.4227e-05, + -3.0403e-03, -2.9202e-03]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0158, -0.0016, 0.0012, 0.0155, -0.0027, -0.0052, 0.0090, 0.0242, + -0.0309, 0.0446], device='cuda:0'), grad: tensor([ 0.0229, -0.0045, -0.0080, -0.0266, -0.0247, 0.0163, 0.0158, 0.0323, + -0.0146, -0.0089], device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 216.77, cls_loss 0.5532 cls_loss_mapping 0.0079 cls_loss_causal 0.5223 re_mapping 0.0084 re_causal 0.0223 /// teacc 98.80 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0710, 0.0627, -0.0958, ..., -0.0969, -0.0795, -0.0020], + [-0.0578, -0.1251, -0.0018, ..., -0.0564, -0.0380, -0.0732], + [ 0.0300, -0.0792, 0.0468, ..., 0.1555, -0.1084, -0.0376], + ..., + [-0.0852, -0.1347, 0.0832, ..., -0.0172, -0.0248, 0.0639], + [ 0.0183, 0.0331, -0.0258, ..., -0.0758, -0.0919, -0.0085], + [-0.1343, -0.0137, -0.0076, ..., -0.1289, 0.0985, 0.0151]], + device='cuda:0'), grad: tensor([[ 2.3544e-05, 4.1395e-05, -2.5539e-03, ..., 3.8147e-04, + 1.2708e-04, 6.0511e-04], + [ 2.4235e-04, 3.2187e-05, 3.5191e-04, ..., 1.3037e-03, + 5.2166e-04, 1.0033e-03], + [ 2.1744e-03, 8.0109e-04, 3.2539e-03, ..., 2.9697e-03, + 7.2336e-04, 2.2449e-03], + ..., + [ 4.1771e-04, 4.3726e-04, 1.7233e-03, ..., 2.3632e-03, + 1.2598e-03, 2.5597e-03], + [ 8.6641e-04, 6.2418e-04, 2.0428e-03, ..., 9.4414e-04, + 2.6178e-04, 1.6556e-03], + [-1.0118e-03, -3.0441e-03, -7.4120e-03, ..., -4.4327e-03, + -1.9627e-03, -6.0844e-03]], device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0145, -0.0007, 0.0020, 0.0164, -0.0037, -0.0056, 0.0085, 0.0235, + -0.0309, 0.0457], device='cuda:0'), grad: tensor([-0.0461, -0.0059, 0.0428, -0.0569, 0.0141, 0.0319, 0.0093, 0.0027, + 0.0302, -0.0222], device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 216.81, cls_loss 0.5131 cls_loss_mapping 0.0067 cls_loss_causal 0.4905 re_mapping 0.0083 re_causal 0.0215 /// teacc 98.74 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0720, 0.0616, -0.0962, ..., -0.0951, -0.0794, -0.0010], + [-0.0580, -0.1258, -0.0022, ..., -0.0575, -0.0374, -0.0719], + [ 0.0285, -0.0800, 0.0476, ..., 0.1558, -0.1083, -0.0376], + ..., + [-0.0835, -0.1343, 0.0829, ..., -0.0179, -0.0254, 0.0639], + [ 0.0173, 0.0320, -0.0265, ..., -0.0757, -0.0926, -0.0087], + [-0.1341, -0.0123, -0.0067, ..., -0.1293, 0.0990, 0.0141]], + device='cuda:0'), grad: tensor([[-1.8021e-02, -2.1210e-02, -5.1231e-03, ..., -1.0880e-02, + -2.3308e-03, -1.7166e-03], + [ 4.1783e-05, 9.3818e-05, 9.3222e-04, ..., -2.1541e-04, + 6.1846e-04, 2.1420e-03], + [ 1.0170e-02, 1.0506e-02, -6.9714e-04, ..., 4.4060e-03, + -5.5408e-04, 9.9480e-05], + ..., + [ 9.0957e-05, 1.8227e-04, -3.0327e-03, ..., 2.5201e-04, + -1.3471e-04, -8.7118e-04], + [-9.4950e-05, 2.8682e-04, 1.5182e-03, ..., 6.3610e-04, + 8.2970e-04, 1.6022e-03], + [ 4.0293e-04, 9.6798e-04, 1.2764e-02, ..., 8.9111e-03, + 7.5035e-03, 8.4152e-03]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0151, -0.0003, 0.0022, 0.0162, -0.0040, -0.0057, 0.0089, 0.0234, + -0.0320, 0.0460], device='cuda:0'), grad: tensor([-0.0426, 0.0168, -0.0229, 0.0330, -0.0370, 0.0098, -0.0064, -0.0186, + 0.0178, 0.0500], device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 217.05, cls_loss 0.5264 cls_loss_mapping 0.0086 cls_loss_causal 0.4933 re_mapping 0.0077 re_causal 0.0189 /// teacc 98.87 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0709, 0.0619, -0.0966, ..., -0.0945, -0.0794, -0.0011], + [-0.0582, -0.1264, -0.0016, ..., -0.0565, -0.0366, -0.0714], + [ 0.0271, -0.0807, 0.0462, ..., 0.1544, -0.1060, -0.0387], + ..., + [-0.0827, -0.1351, 0.0841, ..., -0.0152, -0.0266, 0.0643], + [ 0.0186, 0.0336, -0.0266, ..., -0.0768, -0.0921, -0.0084], + [-0.1352, -0.0130, -0.0066, ..., -0.1307, 0.0989, 0.0144]], + device='cuda:0'), grad: tensor([[ 1.7214e-03, 9.7322e-04, 4.9448e-04, ..., 1.2362e-04, + 2.0504e-04, 1.3676e-03], + [ 9.5129e-04, 4.5490e-04, 2.7347e-04, ..., 4.0436e-03, + 5.6190e-03, 2.2755e-03], + [ 1.0204e-03, 9.3222e-04, 3.3212e-04, ..., -1.4277e-03, + 1.2767e-04, 1.2560e-03], + ..., + [ 5.7030e-04, 2.9469e-04, 1.3924e-04, ..., 5.3972e-05, + 4.0740e-05, 4.8470e-04], + [-5.6877e-03, -3.6678e-03, 4.7946e-04, ..., 1.3857e-03, + -7.5865e-04, 1.1654e-03], + [ 5.8508e-04, 3.2210e-04, 6.9189e-04, ..., 3.0541e-04, + 3.7456e-04, 7.7343e-04]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0155, -0.0003, 0.0012, 0.0162, -0.0039, -0.0065, 0.0095, 0.0238, + -0.0317, 0.0460], device='cuda:0'), grad: tensor([ 0.0269, 0.0269, -0.0101, -0.0430, -0.0373, 0.0337, -0.0356, 0.0180, + 0.0026, 0.0180], device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 216.90, cls_loss 0.5126 cls_loss_mapping 0.0072 cls_loss_causal 0.4849 re_mapping 0.0084 re_causal 0.0204 /// teacc 98.37 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0712, 0.0620, -0.0975, ..., -0.0952, -0.0794, -0.0012], + [-0.0583, -0.1268, -0.0017, ..., -0.0558, -0.0367, -0.0731], + [ 0.0271, -0.0809, 0.0465, ..., 0.1547, -0.1050, -0.0385], + ..., + [-0.0834, -0.1354, 0.0842, ..., -0.0154, -0.0264, 0.0637], + [ 0.0168, 0.0329, -0.0261, ..., -0.0761, -0.0925, -0.0094], + [-0.1347, -0.0134, -0.0073, ..., -0.1314, 0.0989, 0.0149]], + device='cuda:0'), grad: tensor([[ 1.9288e-04, 2.6155e-04, 6.3276e-04, ..., 3.3545e-04, + 5.4693e-04, 1.6890e-03], + [ 2.1505e-04, 2.1684e-04, 1.0824e-03, ..., 8.1253e-04, + 7.2336e-04, 2.8515e-03], + [-8.5449e-04, 4.1753e-05, -4.9782e-03, ..., -4.8637e-03, + -2.6073e-03, -1.8764e-04], + ..., + [ 2.3854e-04, 2.2316e-04, -4.4298e-04, ..., 8.7595e-04, + 8.0395e-04, -6.6605e-03], + [-1.6201e-04, 4.6253e-05, 2.0294e-03, ..., 1.3533e-03, + 2.1896e-03, 2.6627e-03], + [ 1.4839e-03, -1.2122e-05, -4.6654e-03, ..., -2.9278e-03, + -7.1220e-03, -6.7596e-03]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0141, -0.0008, 0.0019, 0.0166, -0.0031, -0.0064, 0.0099, 0.0223, + -0.0313, 0.0464], device='cuda:0'), grad: tensor([ 0.0174, 0.0250, -0.0088, -0.0376, 0.0251, -0.0155, 0.0224, -0.0532, + 0.0240, 0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 217.19, cls_loss 0.5097 cls_loss_mapping 0.0086 cls_loss_causal 0.4880 re_mapping 0.0075 re_causal 0.0184 /// teacc 98.93 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0720, 0.0612, -0.0969, ..., -0.0954, -0.0800, -0.0007], + [-0.0577, -0.1259, -0.0024, ..., -0.0550, -0.0362, -0.0750], + [ 0.0272, -0.0817, 0.0476, ..., 0.1549, -0.1048, -0.0379], + ..., + [-0.0839, -0.1369, 0.0836, ..., -0.0165, -0.0268, 0.0642], + [ 0.0171, 0.0343, -0.0260, ..., -0.0763, -0.0933, -0.0098], + [-0.1348, -0.0142, -0.0079, ..., -0.1320, 0.0989, 0.0152]], + device='cuda:0'), grad: tensor([[ 1.4715e-03, -1.1463e-03, 1.1349e-04, ..., 2.9588e-04, + 6.9380e-05, -2.3975e-03], + [ 8.7070e-04, 2.5058e-04, 1.2046e-04, ..., 9.5725e-05, + 2.7776e-04, 8.5592e-04], + [-1.7548e-03, 1.4019e-03, 6.5184e-04, ..., 5.9700e-04, + 3.0851e-04, -2.1191e-03], + ..., + [ 9.5940e-04, 4.1437e-04, 2.6751e-04, ..., 9.4831e-05, + 5.2929e-04, 1.7433e-03], + [-2.0008e-03, -1.4143e-03, -1.6098e-03, ..., -9.7847e-04, + -3.1328e-04, 1.3227e-03], + [ 1.1616e-03, -9.0480e-05, -3.0804e-04, ..., 2.5916e-04, + -9.0647e-04, 6.4993e-04]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0145, -0.0010, 0.0022, 0.0164, -0.0036, -0.0055, 0.0095, 0.0220, + -0.0317, 0.0470], device='cuda:0'), grad: tensor([ 0.0103, -0.0168, -0.0066, -0.0106, 0.0053, 0.0187, -0.0174, 0.0180, + 0.0097, -0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 217.02, cls_loss 0.5171 cls_loss_mapping 0.0060 cls_loss_causal 0.4910 re_mapping 0.0084 re_causal 0.0204 /// teacc 98.73 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0721, 0.0614, -0.0974, ..., -0.0959, -0.0797, -0.0017], + [-0.0588, -0.1262, -0.0020, ..., -0.0550, -0.0364, -0.0761], + [ 0.0277, -0.0810, 0.0470, ..., 0.1546, -0.1052, -0.0375], + ..., + [-0.0845, -0.1382, 0.0841, ..., -0.0152, -0.0267, 0.0644], + [ 0.0179, 0.0355, -0.0264, ..., -0.0764, -0.0940, -0.0098], + [-0.1351, -0.0136, -0.0077, ..., -0.1319, 0.0994, 0.0156]], + device='cuda:0'), grad: tensor([[ 1.4973e-04, -1.3375e-04, 9.2506e-05, ..., 7.8976e-05, + 1.1563e-04, -3.4657e-03], + [ 8.7881e-04, 8.6799e-06, 2.9755e-04, ..., 1.2102e-03, + 1.6189e-04, 2.1706e-03], + [ 1.1330e-03, 6.6221e-05, 2.7728e-04, ..., 9.5701e-04, + 1.7774e-04, 2.3804e-03], + ..., + [-3.1681e-03, 4.7870e-06, -8.3494e-04, ..., -5.6496e-03, + 2.6083e-04, -5.1422e-03], + [-3.5954e-03, -2.1992e-03, 6.7949e-05, ..., 1.5545e-04, + -1.0214e-03, 1.4668e-03], + [ 1.9693e-04, 2.6926e-05, 2.3949e-04, ..., 1.0312e-04, + 5.0306e-04, 1.6155e-03]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0137, -0.0017, 0.0029, 0.0153, -0.0036, -0.0064, 0.0102, 0.0241, + -0.0311, 0.0463], device='cuda:0'), grad: tensor([-0.0265, 0.0129, 0.0136, 0.0222, 0.0112, -0.0199, 0.0059, -0.0278, + 0.0012, 0.0071], device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 216.96, cls_loss 0.5395 cls_loss_mapping 0.0079 cls_loss_causal 0.5098 re_mapping 0.0080 re_causal 0.0195 /// teacc 98.61 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0714, 0.0620, -0.0973, ..., -0.0967, -0.0802, -0.0018], + [-0.0600, -0.1274, -0.0017, ..., -0.0556, -0.0371, -0.0764], + [ 0.0281, -0.0806, 0.0475, ..., 0.1543, -0.1046, -0.0378], + ..., + [-0.0856, -0.1393, 0.0838, ..., -0.0148, -0.0269, 0.0644], + [ 0.0175, 0.0348, -0.0264, ..., -0.0769, -0.0948, -0.0088], + [-0.1360, -0.0137, -0.0075, ..., -0.1317, 0.1002, 0.0158]], + device='cuda:0'), grad: tensor([[ 7.3195e-05, 1.3423e-04, 5.3215e-04, ..., 2.7567e-05, + 2.7943e-04, 1.0834e-03], + [ 8.6069e-05, 1.7893e-04, 9.9754e-04, ..., 5.8413e-05, + 5.4932e-04, 1.6432e-03], + [ 1.0300e-03, 4.1199e-04, 1.8444e-03, ..., 7.5674e-04, + 2.9421e-04, 1.5001e-03], + ..., + [-2.4509e-03, -1.8129e-03, -8.4076e-03, ..., -2.1076e-03, + -2.0885e-03, -5.7373e-03], + [ 3.2749e-03, 2.0504e-03, 1.4553e-03, ..., 5.6362e-04, + 3.6073e-04, 1.2445e-03], + [ 1.7965e-04, 1.7726e-04, 1.7538e-03, ..., 5.5522e-05, + 1.0624e-03, 1.8816e-03]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0138, -0.0020, 0.0027, 0.0156, -0.0037, -0.0060, 0.0110, 0.0241, + -0.0322, 0.0464], device='cuda:0'), grad: tensor([ 0.0145, 0.0255, 0.0233, -0.0166, 0.0016, 0.0102, -0.0415, -0.0705, + 0.0264, 0.0273], device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 217.09, cls_loss 0.5510 cls_loss_mapping 0.0071 cls_loss_causal 0.5274 re_mapping 0.0081 re_causal 0.0199 /// teacc 98.74 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0713, 0.0629, -0.0973, ..., -0.0957, -0.0802, -0.0024], + [-0.0602, -0.1251, -0.0024, ..., -0.0565, -0.0362, -0.0761], + [ 0.0275, -0.0809, 0.0466, ..., 0.1539, -0.1065, -0.0393], + ..., + [-0.0867, -0.1390, 0.0853, ..., -0.0148, -0.0259, 0.0654], + [ 0.0189, 0.0345, -0.0269, ..., -0.0763, -0.0960, -0.0094], + [-0.1364, -0.0134, -0.0080, ..., -0.1312, 0.1001, 0.0163]], + device='cuda:0'), grad: tensor([[ 2.7609e-04, -4.9448e-04, 3.9041e-05, ..., 9.7901e-06, + 2.8992e-04, 3.5024e-04], + [ 9.0456e-04, 4.4775e-04, 1.7238e-04, ..., 1.4298e-05, + 7.6199e-04, 1.1549e-03], + [-4.2992e-03, 3.9148e-04, 8.2588e-04, ..., -2.6315e-05, + 4.1294e-04, 5.3358e-04], + ..., + [ 6.4850e-04, 1.2058e-04, -6.2256e-03, ..., 1.0632e-05, + -1.0529e-03, 3.6025e-04], + [ 1.5411e-03, 8.3160e-03, 5.3177e-03, ..., 2.2933e-05, + 4.9934e-03, 3.9124e-04], + [-1.2388e-03, -1.0948e-02, -2.0199e-03, ..., -1.9360e-04, + -4.8103e-03, -2.1286e-03]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0143, -0.0017, 0.0014, 0.0158, -0.0040, -0.0061, 0.0120, 0.0242, + -0.0325, 0.0464], device='cuda:0'), grad: tensor([-0.0147, 0.0348, 0.0088, 0.0003, 0.0150, 0.0143, -0.0358, -0.0330, + 0.0459, -0.0356], device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 216.82, cls_loss 0.5056 cls_loss_mapping 0.0071 cls_loss_causal 0.4816 re_mapping 0.0079 re_causal 0.0202 /// teacc 98.89 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0724, 0.0622, -0.0979, ..., -0.0956, -0.0803, -0.0022], + [-0.0595, -0.1251, -0.0024, ..., -0.0567, -0.0371, -0.0774], + [ 0.0287, -0.0807, 0.0477, ..., 0.1541, -0.1063, -0.0398], + ..., + [-0.0871, -0.1402, 0.0854, ..., -0.0146, -0.0259, 0.0669], + [ 0.0193, 0.0343, -0.0260, ..., -0.0768, -0.0955, -0.0085], + [-0.1351, -0.0121, -0.0083, ..., -0.1315, 0.0999, 0.0164]], + device='cuda:0'), grad: tensor([[-2.9373e-03, -2.2049e-03, 2.8324e-04, ..., 2.2137e-04, + 4.4560e-04, -7.0496e-03], + [ 1.2589e-03, -4.2653e-04, -1.2326e-04, ..., 6.9904e-04, + 6.6328e-04, 1.0443e-03], + [ 8.0490e-03, 7.3204e-03, 6.8016e-03, ..., 1.0666e-02, + 1.9426e-03, 3.7880e-03], + ..., + [-4.7684e-03, 2.8586e-04, -3.1128e-03, ..., -2.8610e-03, + 8.7452e-04, -1.0170e-02], + [-4.3124e-05, -3.3398e-03, -1.0185e-03, ..., -2.8915e-03, + 1.3609e-03, 1.5915e-02], + [ 2.6226e-03, 6.5088e-04, 1.3180e-03, ..., 8.1778e-04, + 1.0643e-03, 3.5877e-03]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0145, -0.0026, 0.0013, 0.0158, -0.0036, -0.0065, 0.0119, 0.0238, + -0.0312, 0.0464], device='cuda:0'), grad: tensor([-0.0219, -0.0324, 0.0278, -0.0392, -0.0401, 0.0075, 0.0262, -0.0274, + 0.0583, 0.0412], device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 216.87, cls_loss 0.5628 cls_loss_mapping 0.0078 cls_loss_causal 0.5340 re_mapping 0.0076 re_causal 0.0204 /// teacc 98.70 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0727, 0.0626, -0.0975, ..., -0.0954, -0.0817, -0.0004], + [-0.0600, -0.1257, -0.0023, ..., -0.0572, -0.0364, -0.0786], + [ 0.0285, -0.0820, 0.0476, ..., 0.1534, -0.1072, -0.0399], + ..., + [-0.0880, -0.1414, 0.0850, ..., -0.0141, -0.0257, 0.0669], + [ 0.0189, 0.0346, -0.0259, ..., -0.0758, -0.0958, -0.0076], + [-0.1337, -0.0115, -0.0082, ..., -0.1323, 0.0998, 0.0154]], + device='cuda:0'), grad: tensor([[ 4.5657e-04, 1.0376e-03, 1.5430e-03, ..., 3.4571e-04, + 9.0599e-04, 1.5125e-03], + [ 1.9443e-04, 5.6314e-04, 1.3151e-03, ..., 3.7742e-04, + -1.4029e-03, -2.3594e-03], + [ 6.5088e-04, 1.2503e-03, 6.0730e-03, ..., 4.0588e-03, + 1.2321e-03, 2.3956e-03], + ..., + [ 7.7868e-04, 8.6832e-04, 5.4741e-04, ..., 9.2983e-04, + 1.5268e-03, 1.0777e-03], + [ 7.2098e-04, -2.5177e-04, -1.7166e-03, ..., 6.6280e-05, + 2.1374e-04, 7.9203e-04], + [ 1.0967e-05, 9.9087e-04, -2.6493e-03, ..., -1.0195e-03, + -2.3346e-03, -6.3562e-04]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0149, -0.0026, 0.0018, 0.0158, -0.0026, -0.0058, 0.0105, 0.0232, + -0.0309, 0.0454], device='cuda:0'), grad: tensor([ 0.0271, -0.0317, 0.0444, 0.0107, 0.0278, -0.0020, -0.0563, 0.0305, + -0.0211, -0.0294], device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 216.76, cls_loss 0.5494 cls_loss_mapping 0.0071 cls_loss_causal 0.5205 re_mapping 0.0077 re_causal 0.0192 /// teacc 98.88 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0727, 0.0627, -0.0977, ..., -0.0949, -0.0812, 0.0002], + [-0.0607, -0.1251, -0.0015, ..., -0.0549, -0.0366, -0.0788], + [ 0.0291, -0.0816, 0.0476, ..., 0.1529, -0.1063, -0.0398], + ..., + [-0.0884, -0.1439, 0.0851, ..., -0.0142, -0.0252, 0.0676], + [ 0.0194, 0.0357, -0.0252, ..., -0.0755, -0.0962, -0.0078], + [-0.1343, -0.0113, -0.0080, ..., -0.1329, 0.1001, 0.0159]], + device='cuda:0'), grad: tensor([[ 2.7180e-04, -5.0735e-04, 6.6376e-04, ..., 5.5027e-04, + 1.3864e-04, 1.3132e-03], + [ 1.1702e-03, 7.2181e-05, 2.2566e-04, ..., 2.3258e-04, + 4.5776e-04, 2.4071e-03], + [-3.5686e-03, -8.7798e-05, -5.3596e-03, ..., -2.2755e-03, + 2.4867e-04, -2.0561e-03], + ..., + [ 1.1301e-03, 1.2751e-03, 3.7518e-03, ..., 4.7264e-03, + 2.8396e-04, 2.3746e-03], + [ 2.0809e-03, 1.6594e-03, 2.0332e-03, ..., 6.1369e-04, + 2.2812e-03, 1.8187e-03], + [-2.9159e-04, -5.4646e-04, -1.4925e-04, ..., 1.0139e-04, + -1.8082e-03, 1.5211e-03]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0142, -0.0024, 0.0024, 0.0149, -0.0019, -0.0059, 0.0102, 0.0228, + -0.0307, 0.0460], device='cuda:0'), grad: tensor([-0.0103, 0.0010, -0.0659, -0.0348, 0.0249, -0.0269, 0.0215, 0.0400, + 0.0315, 0.0190], device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 216.35, cls_loss 0.5388 cls_loss_mapping 0.0067 cls_loss_causal 0.5085 re_mapping 0.0077 re_causal 0.0197 /// teacc 98.84 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0742, 0.0626, -0.0984, ..., -0.0948, -0.0815, -0.0008], + [-0.0611, -0.1259, -0.0028, ..., -0.0549, -0.0370, -0.0784], + [ 0.0284, -0.0825, 0.0479, ..., 0.1531, -0.1071, -0.0413], + ..., + [-0.0884, -0.1444, 0.0866, ..., -0.0147, -0.0250, 0.0685], + [ 0.0199, 0.0358, -0.0256, ..., -0.0756, -0.0977, -0.0089], + [-0.1344, -0.0106, -0.0079, ..., -0.1343, 0.0999, 0.0160]], + device='cuda:0'), grad: tensor([[ 4.9448e-04, 3.3689e-04, 4.3541e-05, ..., 1.4663e-05, + 9.2864e-05, 8.8632e-05], + [ 1.9646e-04, 5.8264e-05, -3.6573e-04, ..., 4.8727e-05, + -2.4052e-03, -1.6384e-03], + [-8.1730e-04, -1.6844e-04, -3.9005e-04, ..., -7.0906e-04, + 3.1948e-04, -6.0892e-04], + ..., + [ 7.2718e-04, -5.3197e-05, -2.4681e-03, ..., -2.4929e-03, + -1.5411e-03, -9.6226e-04], + [ 6.2256e-03, 6.1302e-03, 4.1366e-04, ..., 4.6325e-04, + 9.1553e-04, 3.8123e-04], + [ 7.3280e-03, 1.3466e-03, 1.9817e-03, ..., 2.2125e-03, + 3.8662e-03, 1.0910e-03]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0132, -0.0025, 0.0012, 0.0155, -0.0019, -0.0059, 0.0115, 0.0239, + -0.0316, 0.0462], device='cuda:0'), grad: tensor([ 0.0035, -0.0083, -0.0022, -0.0198, 0.0103, 0.0201, -0.0026, -0.0320, + 0.0104, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 216.49, cls_loss 0.5022 cls_loss_mapping 0.0070 cls_loss_causal 0.4746 re_mapping 0.0081 re_causal 0.0198 /// teacc 98.72 lr 0.00010000 +Epoch 209, weight, value: tensor([[-7.4057e-02, 6.2417e-02, -9.7483e-02, ..., -9.4999e-02, + -8.0842e-02, 7.6059e-05], + [-5.9102e-02, -1.2654e-01, -2.6437e-03, ..., -5.5748e-02, + -3.7958e-02, -7.8459e-02], + [ 3.0034e-02, -8.1824e-02, 4.8012e-02, ..., 1.5286e-01, + -1.0770e-01, -4.0497e-02], + ..., + [-8.9937e-02, -1.4599e-01, 8.7592e-02, ..., -1.3495e-02, + -2.3887e-02, 6.8349e-02], + [ 1.9564e-02, 3.4569e-02, -2.4923e-02, ..., -7.5668e-02, + -9.7231e-02, -8.1318e-03], + [-1.3464e-01, -1.0048e-02, -9.7181e-03, ..., -1.3546e-01, + 1.0077e-01, 1.6536e-02]], device='cuda:0'), grad: tensor([[ 1.5497e-04, -7.5798e-03, -7.9870e-04, ..., 1.4591e-04, + 7.2908e-04, 8.2397e-04], + [ 6.0415e-04, 2.4748e-04, 9.3508e-04, ..., 2.8777e-04, + 1.6403e-03, 1.9350e-03], + [ 5.1165e-04, 1.3041e-04, 8.7404e-04, ..., 2.1040e-05, + 6.6233e-04, 1.0509e-03], + ..., + [ 2.3699e-04, 1.8167e-04, -4.7278e-04, ..., 1.3781e-04, + 9.2793e-04, 2.4363e-05], + [ 2.7013e-04, 7.3547e-03, 1.8749e-03, ..., 1.5533e-04, + -9.6846e-04, -1.5535e-03], + [ 1.3199e-03, -1.6367e-04, -1.0128e-03, ..., -1.3571e-03, + -1.9331e-03, 7.9012e-04]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0143, -0.0021, 0.0018, 0.0151, -0.0018, -0.0064, 0.0107, 0.0233, + -0.0311, 0.0458], device='cuda:0'), grad: tensor([-0.0089, 0.0218, 0.0106, -0.0052, -0.0186, -0.0201, 0.0093, 0.0074, + -0.0025, 0.0063], device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 216.60, cls_loss 0.4987 cls_loss_mapping 0.0061 cls_loss_causal 0.4661 re_mapping 0.0080 re_causal 0.0199 /// teacc 98.67 lr 0.00010000 +Epoch 210, weight, value: tensor([[-7.2646e-02, 6.2751e-02, -9.6438e-02, ..., -9.5820e-02, + -8.0766e-02, 6.8615e-05], + [-5.8439e-02, -1.2608e-01, -1.4281e-03, ..., -5.6710e-02, + -3.7786e-02, -7.8263e-02], + [ 2.7939e-02, -8.2531e-02, 4.8889e-02, ..., 1.5314e-01, + -1.0864e-01, -4.0386e-02], + ..., + [-9.0720e-02, -1.4743e-01, 8.7264e-02, ..., -1.4834e-02, + -2.3625e-02, 6.8466e-02], + [ 1.9324e-02, 3.3986e-02, -2.5800e-02, ..., -7.6315e-02, + -9.6993e-02, -8.2106e-03], + [-1.3437e-01, -1.0299e-02, -9.4718e-03, ..., -1.3508e-01, + 1.0134e-01, 1.7188e-02]], device='cuda:0'), grad: tensor([[ 8.1444e-04, 4.8542e-04, 5.4550e-04, ..., 7.4530e-04, + 2.2769e-05, 1.0379e-05], + [ 9.8801e-04, 5.2899e-06, -2.0157e-02, ..., 9.6703e-04, + 3.5390e-06, 2.0824e-06], + [ 3.3054e-03, 2.5228e-05, 7.1096e-04, ..., 3.9139e-03, + 3.4183e-05, 3.9153e-06], + ..., + [ 3.1281e-04, 8.3372e-06, 8.7280e-03, ..., 3.8958e-04, + 3.2043e-04, -7.3090e-06], + [ 6.1274e-04, 1.3089e-04, 3.7651e-03, ..., 6.6853e-04, + 1.0405e-03, 1.7986e-05], + [ 1.2815e-04, -2.4939e-04, -3.0556e-03, ..., -3.3116e-04, + -1.8158e-03, -2.4462e-04]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0152, -0.0017, 0.0018, 0.0157, -0.0023, -0.0071, 0.0104, 0.0225, + -0.0317, 0.0468], device='cuda:0'), grad: tensor([ 0.0064, -0.0164, -0.0042, 0.0060, -0.0105, 0.0077, -0.0155, 0.0196, + 0.0142, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 216.75, cls_loss 0.4989 cls_loss_mapping 0.0072 cls_loss_causal 0.4712 re_mapping 0.0077 re_causal 0.0190 /// teacc 98.82 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0734, 0.0633, -0.0962, ..., -0.0962, -0.0813, -0.0006], + [-0.0597, -0.1264, 0.0009, ..., -0.0561, -0.0376, -0.0777], + [ 0.0293, -0.0816, 0.0493, ..., 0.1535, -0.1082, -0.0403], + ..., + [-0.0891, -0.1486, 0.0867, ..., -0.0152, -0.0246, 0.0690], + [ 0.0194, 0.0341, -0.0255, ..., -0.0770, -0.0972, -0.0072], + [-0.1355, -0.0105, -0.0108, ..., -0.1357, 0.1014, 0.0167]], + device='cuda:0'), grad: tensor([[ 4.6134e-04, -2.4307e-04, 5.9557e-04, ..., 3.5477e-04, + 1.5759e-04, 1.8966e-04], + [ 9.1374e-05, 2.3067e-05, 7.4673e-04, ..., 3.7742e-04, + 9.6202e-05, -9.4414e-04], + [ 3.2158e-03, 2.3727e-03, 1.5669e-03, ..., 1.3628e-03, + 3.8385e-04, 2.2495e-04], + ..., + [ 4.8161e-04, 1.1075e-04, -4.6577e-03, ..., -3.6383e-04, + 1.0157e-03, -6.5660e-04], + [ 1.0910e-03, 9.9754e-04, 3.2959e-03, ..., 1.7729e-03, + 2.9144e-03, 1.1253e-03], + [ 2.2793e-03, 1.2789e-03, -8.2970e-05, ..., 1.2617e-03, + -1.0738e-03, 2.1706e-03]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0151, -0.0010, 0.0011, 0.0164, -0.0025, -0.0061, 0.0110, 0.0221, + -0.0323, 0.0460], device='cuda:0'), grad: tensor([ 0.0086, -0.0446, 0.0234, -0.0192, -0.0254, 0.0122, 0.0173, -0.0124, + 0.0207, 0.0194], device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 216.92, cls_loss 0.5168 cls_loss_mapping 0.0077 cls_loss_causal 0.4919 re_mapping 0.0076 re_causal 0.0192 /// teacc 98.73 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0742, 0.0628, -0.0963, ..., -0.0962, -0.0815, -0.0007], + [-0.0573, -0.1256, 0.0013, ..., -0.0555, -0.0373, -0.0789], + [ 0.0291, -0.0812, 0.0496, ..., 0.1532, -0.1087, -0.0401], + ..., + [-0.0904, -0.1497, 0.0863, ..., -0.0150, -0.0245, 0.0689], + [ 0.0194, 0.0342, -0.0256, ..., -0.0766, -0.0972, -0.0071], + [-0.1371, -0.0100, -0.0104, ..., -0.1365, 0.1017, 0.0162]], + device='cuda:0'), grad: tensor([[ 1.2039e-02, 5.5933e-04, 9.3174e-04, ..., 1.9255e-03, + 5.9366e-04, 1.1711e-03], + [-5.8748e-06, 5.0850e-06, 1.3294e-03, ..., 7.7581e-04, + 7.4387e-04, 2.7714e-03], + [ 2.1954e-03, 5.1409e-05, 1.9026e-03, ..., 1.5802e-03, + 2.4378e-04, 3.6259e-03], + ..., + [ 7.7724e-05, 3.5651e-06, -1.5807e-04, ..., 6.7139e-04, + 5.2118e-04, 5.9557e-04], + [ 6.3467e-04, 1.6336e-03, 2.2163e-03, ..., 1.2980e-03, + 9.2316e-04, 1.8454e-03], + [ 1.4973e-04, 2.2367e-05, -6.5947e-04, ..., 5.9748e-04, + -5.3062e-03, -8.9340e-03]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0156, -0.0005, 0.0003, 0.0165, -0.0024, -0.0057, 0.0106, 0.0221, + -0.0325, 0.0459], device='cuda:0'), grad: tensor([ 0.0059, 0.0315, 0.0201, -0.0185, 0.0174, 0.0084, -0.0163, 0.0072, + 0.0217, -0.0774], device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 217.09, cls_loss 0.5061 cls_loss_mapping 0.0081 cls_loss_causal 0.4782 re_mapping 0.0078 re_causal 0.0191 /// teacc 98.81 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0755, 0.0628, -0.0967, ..., -0.0960, -0.0828, -0.0012], + [-0.0585, -0.1260, 0.0013, ..., -0.0561, -0.0377, -0.0798], + [ 0.0310, -0.0806, 0.0496, ..., 0.1540, -0.1093, -0.0401], + ..., + [-0.0904, -0.1494, 0.0859, ..., -0.0163, -0.0243, 0.0699], + [ 0.0176, 0.0340, -0.0255, ..., -0.0782, -0.0970, -0.0094], + [-0.1357, -0.0101, -0.0112, ..., -0.1377, 0.1009, 0.0164]], + device='cuda:0'), grad: tensor([[-5.8899e-03, -9.5749e-03, -1.5011e-03, ..., -1.4297e-02, + 1.2052e-04, -3.7217e-04], + [ 3.7217e-04, 4.7654e-05, 1.1086e-04, ..., 3.0670e-03, + 2.9159e-04, 4.1533e-04], + [ 1.8883e-03, 3.1300e-03, 6.8092e-04, ..., 1.0004e-03, + 1.4329e-04, 2.7847e-04], + ..., + [ 2.7037e-04, 7.5579e-05, -2.5678e-04, ..., 4.0150e-04, + 1.8919e-04, 2.7776e-04], + [ 2.0409e-03, 1.4484e-04, 1.2279e-04, ..., 2.2583e-03, + 1.6584e-03, 1.8930e-03], + [ 5.3263e-04, 2.5415e-04, 1.8764e-04, ..., 6.8903e-04, + -7.3671e-05, 4.6372e-04]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0143, -0.0005, 0.0014, 0.0168, -0.0032, -0.0049, 0.0116, 0.0220, + -0.0339, 0.0463], device='cuda:0'), grad: tensor([-0.0665, 0.0135, 0.0131, 0.0277, -0.0228, -0.0148, 0.0118, 0.0048, + 0.0251, 0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 217.06, cls_loss 0.4885 cls_loss_mapping 0.0070 cls_loss_causal 0.4738 re_mapping 0.0084 re_causal 0.0200 /// teacc 98.72 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0754, 0.0633, -0.0970, ..., -0.0958, -0.0837, -0.0016], + [-0.0587, -0.1266, 0.0012, ..., -0.0559, -0.0383, -0.0804], + [ 0.0295, -0.0816, 0.0496, ..., 0.1533, -0.1084, -0.0398], + ..., + [-0.0886, -0.1483, 0.0859, ..., -0.0168, -0.0250, 0.0693], + [ 0.0182, 0.0350, -0.0265, ..., -0.0782, -0.0967, -0.0091], + [-0.1367, -0.0106, -0.0097, ..., -0.1375, 0.1018, 0.0163]], + device='cuda:0'), grad: tensor([[-1.6487e-04, -3.3817e-03, 1.6782e-06, ..., -4.6921e-04, + 9.8109e-05, 2.2924e-04], + [ 3.2902e-04, 2.6989e-04, 4.7028e-05, ..., 1.3936e-04, + -3.0136e-04, -1.5125e-03], + [ 3.6359e-04, 2.9278e-04, -7.1004e-06, ..., 5.2869e-05, + 1.3804e-04, 3.2449e-04], + ..., + [ 2.5892e-04, 1.7726e-04, 5.5283e-05, ..., 3.3379e-05, + 3.1519e-04, 5.8079e-04], + [ 1.3742e-03, 7.2670e-04, 7.1049e-05, ..., 1.4853e-04, + 4.9734e-04, 4.7112e-04], + [ 2.5487e-04, 2.5105e-04, -1.1587e-04, ..., 3.8654e-05, + 2.5058e-04, 6.4135e-05]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0138, -0.0007, 0.0016, 0.0176, -0.0032, -0.0058, 0.0122, 0.0222, + -0.0333, 0.0457], device='cuda:0'), grad: tensor([-0.0461, -0.0070, 0.0135, 0.0169, -0.0151, 0.0056, 0.0187, 0.0102, + 0.0194, -0.0162], device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 216.87, cls_loss 0.5122 cls_loss_mapping 0.0074 cls_loss_causal 0.4878 re_mapping 0.0076 re_causal 0.0192 /// teacc 98.77 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0758, 0.0632, -0.0971, ..., -0.0956, -0.0833, -0.0008], + [-0.0588, -0.1271, 0.0012, ..., -0.0549, -0.0380, -0.0802], + [ 0.0297, -0.0811, 0.0499, ..., 0.1530, -0.1082, -0.0393], + ..., + [-0.0888, -0.1493, 0.0858, ..., -0.0159, -0.0262, 0.0698], + [ 0.0170, 0.0344, -0.0258, ..., -0.0785, -0.0958, -0.0085], + [-0.1361, -0.0105, -0.0089, ..., -0.1372, 0.1021, 0.0159]], + device='cuda:0'), grad: tensor([[ 1.7595e-03, 3.9649e-04, 5.5254e-05, ..., 1.1158e-04, + 4.1462e-06, 4.4084e-04], + [ 2.0337e-04, -8.3303e-04, -3.6831e-03, ..., -3.5214e-04, + 5.4657e-05, -1.9817e-03], + [ 6.8045e-04, 2.3037e-05, 5.0831e-04, ..., 1.8096e-04, + 7.2271e-06, 1.6555e-05], + ..., + [ 2.0218e-03, 1.8626e-05, 4.2229e-03, ..., -3.8099e-04, + 1.9360e-04, 2.3174e-03], + [ 1.0509e-03, 8.1837e-05, -1.9598e-04, ..., 2.4498e-05, + 4.6670e-05, 8.0824e-05], + [ 1.2188e-03, 2.0504e-05, 9.1934e-03, ..., 2.4125e-05, + 2.2907e-03, 3.2101e-03]], device='cuda:0') +Epoch 215, bias, value: tensor([ 1.4044e-02, 4.3852e-06, 1.4508e-03, 1.7602e-02, -3.6247e-03, + -5.6474e-03, 1.1832e-02, 2.2295e-02, -3.3486e-02, 4.5416e-02], + device='cuda:0'), grad: tensor([ 0.0181, -0.0211, 0.0020, -0.0530, -0.0598, -0.0096, 0.0435, 0.0396, + 0.0111, 0.0292], device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 217.02, cls_loss 0.5209 cls_loss_mapping 0.0081 cls_loss_causal 0.4954 re_mapping 0.0078 re_causal 0.0191 /// teacc 98.81 lr 0.00010000 +Epoch 216, weight, value: tensor([[-7.6291e-02, 6.3425e-02, -9.6967e-02, ..., -9.5763e-02, + -8.3218e-02, -1.6413e-04], + [-6.0292e-02, -1.2800e-01, -1.1051e-04, ..., -5.6342e-02, + -3.8162e-02, -8.1411e-02], + [ 3.0789e-02, -8.0815e-02, 4.9649e-02, ..., 1.5367e-01, + -1.0790e-01, -3.9853e-02], + ..., + [-8.9448e-02, -1.4985e-01, 8.6237e-02, ..., -1.6696e-02, + -2.6110e-02, 6.9438e-02], + [ 1.6858e-02, 3.4398e-02, -2.6439e-02, ..., -7.8977e-02, + -9.6478e-02, -8.2318e-03], + [-1.3624e-01, -1.1467e-02, -8.7653e-03, ..., -1.3797e-01, + 1.0147e-01, 1.6420e-02]], device='cuda:0'), grad: tensor([[-1.9894e-03, -1.2283e-03, 8.8835e-04, ..., -1.4581e-05, + 4.4274e-04, -1.7300e-03], + [ 1.1170e-04, 8.0884e-05, -1.8473e-03, ..., 2.8324e-04, + 3.7432e-05, -3.2544e-04], + [-2.6455e-03, 8.9169e-04, 8.0490e-04, ..., -2.4624e-03, + 4.0054e-04, 6.2561e-04], + ..., + [ 8.8978e-04, 8.3399e-04, 1.0748e-03, ..., 4.7278e-04, + 8.0204e-04, 1.7347e-03], + [ 1.4124e-03, 1.1530e-03, 4.8652e-06, ..., 1.2197e-03, + 8.6021e-04, -3.7270e-03], + [-2.0218e-03, -4.9133e-03, -1.2451e-02, ..., -2.3708e-03, + -2.0477e-02, -8.8272e-03]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0144, -0.0006, 0.0016, 0.0180, -0.0038, -0.0054, 0.0108, 0.0232, + -0.0330, 0.0446], device='cuda:0'), grad: tensor([-0.0006, -0.0089, -0.0284, 0.0259, 0.0421, 0.0101, -0.0179, 0.0198, + -0.0143, -0.0280], device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 216.84, cls_loss 0.4900 cls_loss_mapping 0.0078 cls_loss_causal 0.4614 re_mapping 0.0081 re_causal 0.0199 /// teacc 98.79 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0770, 0.0630, -0.0961, ..., -0.0965, -0.0831, -0.0007], + [-0.0616, -0.1274, -0.0003, ..., -0.0574, -0.0386, -0.0812], + [ 0.0325, -0.0806, 0.0500, ..., 0.1541, -0.1087, -0.0402], + ..., + [-0.0897, -0.1511, 0.0855, ..., -0.0172, -0.0275, 0.0689], + [ 0.0167, 0.0338, -0.0262, ..., -0.0784, -0.0972, -0.0090], + [-0.1355, -0.0116, -0.0091, ..., -0.1393, 0.1017, 0.0174]], + device='cuda:0'), grad: tensor([[ 0.0004, -0.0017, 0.0006, ..., 0.0004, 0.0006, 0.0010], + [ 0.0012, 0.0002, 0.0008, ..., 0.0005, 0.0013, 0.0029], + [-0.0012, 0.0008, -0.0012, ..., -0.0004, -0.0022, -0.0050], + ..., + [ 0.0010, 0.0004, 0.0020, ..., 0.0005, 0.0012, 0.0003], + [ 0.0021, 0.0009, 0.0019, ..., 0.0007, 0.0018, 0.0020], + [-0.0031, -0.0004, -0.0103, ..., 0.0003, -0.0067, -0.0082]], + device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0133, -0.0006, 0.0021, 0.0167, -0.0040, -0.0051, 0.0112, 0.0233, + -0.0330, 0.0459], device='cuda:0'), grad: tensor([ 0.0117, 0.0392, -0.0401, 0.0046, 0.0038, -0.0143, -0.0020, 0.0157, + 0.0161, -0.0347], device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 216.97, cls_loss 0.4935 cls_loss_mapping 0.0061 cls_loss_causal 0.4657 re_mapping 0.0076 re_causal 0.0189 /// teacc 98.91 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0767, 0.0639, -0.0963, ..., -0.0974, -0.0831, -0.0010], + [-0.0618, -0.1277, -0.0007, ..., -0.0563, -0.0404, -0.0810], + [ 0.0327, -0.0809, 0.0503, ..., 0.1537, -0.1084, -0.0410], + ..., + [-0.0890, -0.1507, 0.0847, ..., -0.0170, -0.0286, 0.0690], + [ 0.0175, 0.0341, -0.0253, ..., -0.0799, -0.0947, -0.0082], + [-0.1367, -0.0108, -0.0077, ..., -0.1376, 0.1021, 0.0175]], + device='cuda:0'), grad: tensor([[-1.5916e-06, 6.1941e-04, -7.9870e-04, ..., 9.1600e-04, + -6.7520e-04, -2.3041e-03], + [ 1.6499e-04, 1.0836e-04, 9.4128e-04, ..., 6.6280e-04, + 1.3094e-03, 1.8253e-03], + [ 1.0777e-03, 1.3340e-04, 5.1308e-03, ..., 4.7417e-03, + 7.2861e-04, 3.4084e-03], + ..., + [-1.8549e-03, 2.7701e-05, -9.0561e-03, ..., -9.6512e-03, + 7.0667e-04, -5.3749e-03], + [ 6.3591e-03, 2.3937e-03, 7.1487e-03, ..., 6.3095e-03, + 9.0504e-04, 5.1231e-03], + [ 2.9826e-04, 1.9383e-04, 5.8460e-04, ..., 1.9407e-03, + 4.6760e-05, -2.9488e-03]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0131, 0.0004, 0.0019, 0.0173, -0.0040, -0.0048, 0.0109, 0.0225, + -0.0331, 0.0457], device='cuda:0'), grad: tensor([-0.0168, -0.0120, 0.0296, 0.0048, 0.0163, -0.0079, -0.0112, -0.0131, + 0.0213, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 216.59, cls_loss 0.5118 cls_loss_mapping 0.0049 cls_loss_causal 0.4758 re_mapping 0.0079 re_causal 0.0194 /// teacc 98.91 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0759, 0.0640, -0.0951, ..., -0.0980, -0.0844, -0.0004], + [-0.0619, -0.1275, -0.0011, ..., -0.0571, -0.0430, -0.0821], + [ 0.0320, -0.0817, 0.0499, ..., 0.1550, -0.1075, -0.0412], + ..., + [-0.0891, -0.1517, 0.0860, ..., -0.0166, -0.0276, 0.0699], + [ 0.0175, 0.0334, -0.0238, ..., -0.0796, -0.0943, -0.0076], + [-0.1381, -0.0102, -0.0085, ..., -0.1383, 0.1029, 0.0171]], + device='cuda:0'), grad: tensor([[-0.0006, -0.0075, -0.0011, ..., -0.0007, -0.0015, -0.0008], + [ 0.0039, 0.0005, 0.0012, ..., 0.0009, 0.0017, 0.0026], + [-0.0098, -0.0021, -0.0021, ..., -0.0010, -0.0030, -0.0015], + ..., + [-0.0061, 0.0004, 0.0004, ..., 0.0003, 0.0007, -0.0076], + [ 0.0033, 0.0008, 0.0009, ..., 0.0002, 0.0019, 0.0015], + [ 0.0011, 0.0006, -0.0022, ..., 0.0003, -0.0021, 0.0020]], + device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0143, 0.0003, 0.0014, 0.0166, -0.0031, -0.0052, 0.0113, 0.0229, + -0.0341, 0.0454], device='cuda:0'), grad: tensor([-0.0205, 0.0325, -0.0380, -0.0001, 0.0246, 0.0176, -0.0086, -0.0286, + 0.0217, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 216.53, cls_loss 0.4898 cls_loss_mapping 0.0075 cls_loss_causal 0.4634 re_mapping 0.0077 re_causal 0.0189 /// teacc 98.93 lr 0.00010000 +Epoch 220, weight, value: tensor([[-7.6528e-02, 6.3376e-02, -9.5938e-02, ..., -9.8430e-02, + -8.5538e-02, 2.0891e-05], + [-6.1817e-02, -1.2835e-01, -1.5504e-03, ..., -5.8066e-02, + -4.3641e-02, -8.4410e-02], + [ 3.2337e-02, -8.1896e-02, 4.9603e-02, ..., 1.5535e-01, + -1.0745e-01, -4.1453e-02], + ..., + [-8.9323e-02, -1.5344e-01, 8.6413e-02, ..., -1.7023e-02, + -2.7817e-02, 7.0138e-02], + [ 1.7315e-02, 3.4587e-02, -2.3796e-02, ..., -8.0280e-02, + -9.4301e-02, -7.2682e-03], + [-1.3855e-01, -9.2106e-03, -9.3394e-03, ..., -1.3879e-01, + 1.0266e-01, 1.7238e-02]], device='cuda:0'), grad: tensor([[ 1.1677e-04, -4.9686e-04, 4.2766e-05, ..., -2.5821e-04, + 7.1144e-04, 4.4703e-04], + [ 9.8324e-04, 2.2662e-04, 1.1578e-03, ..., -2.0351e-03, + -8.1491e-04, 1.9932e-03], + [-3.8643e-03, -2.4681e-03, -3.4809e-03, ..., -3.8204e-03, + 1.2970e-03, -9.8038e-04], + ..., + [ 3.3212e-04, -2.8759e-05, 1.7939e-03, ..., 6.0225e-04, + 3.4637e-03, 1.9989e-03], + [ 6.6185e-04, 3.3796e-05, 1.5230e-03, ..., 1.1854e-03, + 1.7605e-03, 1.3227e-03], + [ 2.1958e-04, 3.1567e-04, -1.6403e-03, ..., 5.4979e-04, + -7.1812e-04, -1.1635e-03]], device='cuda:0') +Epoch 220, bias, value: tensor([ 1.3825e-02, -6.2837e-05, 8.1852e-04, 1.7045e-02, -3.4797e-03, + -4.3526e-03, 1.0905e-02, 2.3613e-02, -3.4609e-02, 4.6139e-02], + device='cuda:0'), grad: tensor([ 0.0119, 0.0271, -0.0426, 0.0050, -0.0634, -0.0126, 0.0323, 0.0302, + 0.0294, -0.0173], device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 216.53, cls_loss 0.5163 cls_loss_mapping 0.0057 cls_loss_causal 0.4854 re_mapping 0.0083 re_causal 0.0209 /// teacc 98.68 lr 0.00010000 +Epoch 221, weight, value: tensor([[-7.6742e-02, 6.3735e-02, -9.7190e-02, ..., -9.8263e-02, + -8.5737e-02, 1.5193e-05], + [-6.1077e-02, -1.2866e-01, -8.6821e-04, ..., -5.5736e-02, + -4.2908e-02, -8.3877e-02], + [ 3.3761e-02, -8.0482e-02, 4.9377e-02, ..., 1.5501e-01, + -1.0585e-01, -4.0966e-02], + ..., + [-9.0139e-02, -1.5342e-01, 8.7181e-02, ..., -1.6901e-02, + -2.7128e-02, 7.0656e-02], + [ 1.7329e-02, 3.4434e-02, -2.4409e-02, ..., -8.1378e-02, + -9.4831e-02, -8.3996e-03], + [-1.3849e-01, -8.9221e-03, -9.3580e-03, ..., -1.3968e-01, + 1.0133e-01, 1.6568e-02]], device='cuda:0'), grad: tensor([[-9.0313e-04, -1.5497e-03, 1.3626e-04, ..., 4.4078e-05, + 2.3723e-04, 6.5422e-04], + [ 1.1911e-03, 9.9719e-05, 1.9586e-04, ..., 1.3068e-05, + 1.0767e-03, 2.3060e-03], + [-1.1139e-02, 5.1689e-04, 1.8525e-04, ..., -6.3477e-03, + -3.8166e-03, -5.7106e-03], + ..., + [ 4.5323e-04, 8.6427e-05, 1.8871e-04, ..., 1.1390e-04, + 6.3324e-04, 1.3952e-03], + [ 4.9782e-04, 2.3460e-04, -2.3973e-04, ..., -3.9315e-04, + 1.3914e-03, 2.4772e-04], + [ 1.3275e-03, -6.6794e-06, -1.5526e-03, ..., 5.3465e-05, + -7.7581e-04, -1.3685e-03]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0133, 0.0006, 0.0019, 0.0160, -0.0018, -0.0052, 0.0115, 0.0233, + -0.0357, 0.0459], device='cuda:0'), grad: tensor([ 0.0064, 0.0153, -0.0378, 0.0232, 0.0237, 0.0098, -0.0152, 0.0123, + -0.0162, -0.0216], device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 216.85, cls_loss 0.5128 cls_loss_mapping 0.0061 cls_loss_causal 0.4867 re_mapping 0.0084 re_causal 0.0209 /// teacc 98.68 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0778, 0.0635, -0.0978, ..., -0.0977, -0.0868, -0.0008], + [-0.0599, -0.1303, 0.0003, ..., -0.0552, -0.0430, -0.0816], + [ 0.0327, -0.0803, 0.0493, ..., 0.1547, -0.1057, -0.0426], + ..., + [-0.0912, -0.1543, 0.0872, ..., -0.0171, -0.0265, 0.0697], + [ 0.0179, 0.0355, -0.0249, ..., -0.0815, -0.0941, -0.0087], + [-0.1384, -0.0086, -0.0094, ..., -0.1395, 0.1014, 0.0174]], + device='cuda:0'), grad: tensor([[ 1.7993e-06, -4.1485e-04, -2.7409e-03, ..., -3.0398e-04, + -1.7414e-03, -3.1128e-03], + [ 1.3106e-05, 1.5843e-04, -5.4407e-04, ..., 5.6535e-05, + -2.2435e-04, -5.3263e-04], + [-1.6260e-04, -2.4605e-03, -4.1504e-03, ..., -7.0381e-04, + 2.6631e-04, -9.6321e-05], + ..., + [ 1.1168e-05, 2.1660e-04, 6.5422e-04, ..., 1.2290e-04, + 2.1148e-04, -1.3435e-04], + [ 5.1670e-06, 2.3520e-04, 4.4250e-03, ..., 5.6314e-04, + 9.2983e-04, 1.9627e-03], + [ 1.6838e-05, 4.7827e-04, -3.2253e-03, ..., -2.2936e-04, + -7.3814e-04, -3.3436e-03]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0134, 0.0018, 0.0013, 0.0174, -0.0019, -0.0058, 0.0109, 0.0223, + -0.0358, 0.0461], device='cuda:0'), grad: tensor([-0.0179, -0.0190, -0.0108, 0.0314, 0.0194, 0.0107, 0.0148, -0.0022, + -0.0052, -0.0212], device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 217.04, cls_loss 0.5078 cls_loss_mapping 0.0059 cls_loss_causal 0.4817 re_mapping 0.0077 re_causal 0.0187 /// teacc 98.72 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0771, 0.0643, -0.0975, ..., -0.0983, -0.0872, -0.0006], + [-0.0600, -0.1309, 0.0011, ..., -0.0551, -0.0431, -0.0829], + [ 0.0318, -0.0811, 0.0504, ..., 0.1544, -0.1056, -0.0418], + ..., + [-0.0900, -0.1539, 0.0859, ..., -0.0170, -0.0273, 0.0704], + [ 0.0183, 0.0354, -0.0252, ..., -0.0819, -0.0941, -0.0083], + [-0.1382, -0.0083, -0.0091, ..., -0.1396, 0.1014, 0.0182]], + device='cuda:0'), grad: tensor([[-3.2120e-03, -5.2567e-03, 1.7571e-04, ..., -4.0460e-04, + 1.2338e-04, -1.4086e-03], + [ 8.0824e-05, 1.1647e-04, 4.8470e-04, ..., 1.8871e-04, + 3.3832e-04, 5.5456e-04], + [ 3.4124e-05, 3.3307e-04, 3.8948e-03, ..., 3.8986e-03, + 3.0637e-04, 6.6042e-04], + ..., + [-1.5929e-05, 8.1599e-05, -3.0460e-03, ..., -3.7365e-03, + 8.8406e-04, 9.2506e-04], + [ 3.8099e-04, 6.0129e-04, -1.7033e-03, ..., -8.5974e-04, + -2.6054e-03, -3.2330e-03], + [ 1.0556e-04, 2.7037e-04, 2.3460e-04, ..., 2.5058e-04, + -4.4899e-03, -1.1971e-02]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0138, 0.0022, 0.0015, 0.0167, -0.0029, -0.0049, 0.0110, 0.0219, + -0.0358, 0.0464], device='cuda:0'), grad: tensor([-0.0378, 0.0300, 0.0247, -0.0201, 0.0447, 0.0108, 0.0015, 0.0033, + -0.0086, -0.0485], device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 216.89, cls_loss 0.5233 cls_loss_mapping 0.0055 cls_loss_causal 0.4955 re_mapping 0.0076 re_causal 0.0192 /// teacc 98.73 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0765, 0.0649, -0.0963, ..., -0.0991, -0.0866, 0.0002], + [-0.0600, -0.1304, 0.0006, ..., -0.0557, -0.0441, -0.0843], + [ 0.0325, -0.0808, 0.0505, ..., 0.1551, -0.1066, -0.0420], + ..., + [-0.0899, -0.1544, 0.0857, ..., -0.0170, -0.0273, 0.0721], + [ 0.0175, 0.0357, -0.0246, ..., -0.0811, -0.0937, -0.0070], + [-0.1387, -0.0082, -0.0087, ..., -0.1401, 0.1019, 0.0179]], + device='cuda:0'), grad: tensor([[ 8.7023e-04, 5.1165e-04, 3.6693e-04, ..., 5.3436e-05, + 4.1890e-04, 7.2336e-04], + [-1.4782e-04, -1.0580e-04, 4.0627e-04, ..., -2.3329e-04, + 1.2836e-03, 4.6086e-04], + [-4.0398e-03, -1.7605e-03, -1.4200e-03, ..., -3.1972e-04, + -2.3499e-03, -4.4823e-03], + ..., + [ 5.0592e-04, 2.6751e-04, 7.8678e-04, ..., 2.3293e-04, + 3.9315e-04, 1.1024e-03], + [ 1.2894e-03, 3.0708e-04, 2.1362e-04, ..., 8.2195e-05, + 5.5027e-04, 5.6696e-04], + [ 5.9700e-04, 3.0351e-04, 2.5582e-04, ..., 6.6578e-05, + 2.4068e-04, 6.5660e-04]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0142, 0.0006, 0.0015, 0.0154, -0.0034, -0.0055, 0.0117, 0.0245, + -0.0358, 0.0465], device='cuda:0'), grad: tensor([ 0.0160, 0.0165, -0.0744, 0.0128, 0.0143, 0.0130, -0.0437, 0.0163, + 0.0173, 0.0120], device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 216.93, cls_loss 0.5051 cls_loss_mapping 0.0051 cls_loss_causal 0.4789 re_mapping 0.0078 re_causal 0.0195 /// teacc 98.61 lr 0.00010000 +Epoch 225, weight, value: tensor([[-7.5705e-02, 6.5057e-02, -9.6651e-02, ..., -9.8695e-02, + -8.7838e-02, 2.5364e-04], + [-5.9702e-02, -1.3142e-01, 7.2629e-05, ..., -5.6327e-02, + -4.4034e-02, -8.5485e-02], + [ 3.2587e-02, -8.1580e-02, 4.9748e-02, ..., 1.5438e-01, + -1.0759e-01, -4.2532e-02], + ..., + [-9.1109e-02, -1.5617e-01, 8.6575e-02, ..., -1.6083e-02, + -2.6857e-02, 7.2959e-02], + [ 1.8029e-02, 3.5479e-02, -2.4660e-02, ..., -8.1157e-02, + -9.3686e-02, -5.9525e-03], + [-1.3963e-01, -9.4385e-03, -8.0830e-03, ..., -1.4093e-01, + 1.0150e-01, 1.7618e-02]], device='cuda:0'), grad: tensor([[ 1.0309e-03, 5.2309e-04, 2.3174e-04, ..., 1.5087e-07, + 2.4755e-06, 8.7082e-05], + [-8.2245e-03, -4.6921e-03, -2.6894e-03, ..., 6.5705e-07, + 8.2672e-05, 1.4579e-04], + [ 3.0565e-04, 1.7822e-04, 4.1866e-04, ..., 4.4554e-06, + 1.3925e-05, -3.5858e-04], + ..., + [ 1.2803e-04, 8.6248e-05, 1.3676e-03, ..., -2.6718e-05, + 6.8426e-04, 5.3549e-04], + [ 1.4114e-03, 8.4686e-04, 1.7333e-04, ..., 4.4294e-06, + 3.2276e-05, 9.0420e-05], + [ 4.4632e-04, 2.6679e-04, -5.3644e-04, ..., 9.4771e-06, + -9.7322e-04, -5.6410e-04]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0141, 0.0003, 0.0007, 0.0172, -0.0041, -0.0054, 0.0122, 0.0247, + -0.0358, 0.0458], device='cuda:0'), grad: tensor([ 0.0115, -0.0480, -0.0201, -0.0189, 0.0099, 0.0107, 0.0191, 0.0164, + 0.0113, 0.0080], device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 217.00, cls_loss 0.5129 cls_loss_mapping 0.0088 cls_loss_causal 0.4918 re_mapping 0.0076 re_causal 0.0196 /// teacc 98.79 lr 0.00010000 +Epoch 226, weight, value: tensor([[-7.6439e-02, 6.4840e-02, -9.6080e-02, ..., -9.9004e-02, + -8.7635e-02, 7.6137e-05], + [-5.9554e-02, -1.3136e-01, -4.8082e-04, ..., -5.7820e-02, + -4.5352e-02, -8.7186e-02], + [ 3.2991e-02, -8.1821e-02, 4.9883e-02, ..., 1.5568e-01, + -1.0774e-01, -4.2622e-02], + ..., + [-9.0588e-02, -1.5698e-01, 8.5661e-02, ..., -1.5900e-02, + -2.7640e-02, 7.2811e-02], + [ 1.7583e-02, 3.5243e-02, -2.3870e-02, ..., -8.1277e-02, + -9.3213e-02, -5.9424e-03], + [-1.4059e-01, -9.7433e-03, -7.9690e-03, ..., -1.4102e-01, + 1.0133e-01, 1.8298e-02]], device='cuda:0'), grad: tensor([[-3.8071e-03, -8.2245e-03, 1.9157e-04, ..., 3.3545e-04, + 5.6887e-04, 8.9502e-04], + [ 1.8120e-04, 2.2769e-04, 5.2118e-04, ..., 2.2185e-04, + 5.6076e-04, 9.4223e-04], + [ 3.5763e-03, 2.0657e-03, 8.2636e-04, ..., 2.8629e-03, + 1.4820e-03, 2.1229e-03], + ..., + [ 1.2999e-03, 8.0538e-04, -8.2254e-04, ..., 9.8991e-04, + 2.4289e-05, 5.6088e-05], + [ 5.5218e-04, 9.0551e-04, 8.9169e-04, ..., 5.8126e-04, + 4.5753e-04, 1.4677e-03], + [-1.0309e-03, -2.6779e-03, -4.2725e-03, ..., -2.9678e-03, + -4.7989e-03, -6.0501e-03]], device='cuda:0') +Epoch 226, bias, value: tensor([ 1.5006e-02, -2.2494e-05, 8.8826e-04, 1.7924e-02, -3.9751e-03, + -6.1819e-03, 1.2592e-02, 2.3800e-02, -3.5788e-02, 4.5523e-02], + device='cuda:0'), grad: tensor([-0.0120, 0.0098, 0.0180, 0.0212, -0.0465, 0.0077, 0.0219, -0.0148, + 0.0125, -0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 216.78, cls_loss 0.5041 cls_loss_mapping 0.0069 cls_loss_causal 0.4726 re_mapping 0.0082 re_causal 0.0200 /// teacc 98.81 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0760, 0.0659, -0.0956, ..., -0.0998, -0.0868, 0.0006], + [-0.0594, -0.1312, -0.0020, ..., -0.0586, -0.0454, -0.0876], + [ 0.0328, -0.0825, 0.0503, ..., 0.1565, -0.1083, -0.0432], + ..., + [-0.0900, -0.1566, 0.0876, ..., -0.0139, -0.0271, 0.0718], + [ 0.0179, 0.0359, -0.0234, ..., -0.0809, -0.0928, -0.0061], + [-0.1401, -0.0093, -0.0083, ..., -0.1408, 0.1012, 0.0182]], + device='cuda:0'), grad: tensor([[ 1.7672e-03, 3.6869e-03, 6.7663e-04, ..., 1.1927e-04, + 9.6262e-05, 6.4659e-04], + [-2.3975e-03, 1.1516e-04, -6.3515e-04, ..., 2.5296e-04, + 1.3262e-05, -9.6359e-03], + [ 3.4118e-04, 5.4121e-04, 1.7977e-03, ..., 4.9782e-04, + 6.2895e-04, 1.6260e-03], + ..., + [ 2.1038e-03, 2.7585e-04, -3.4275e-03, ..., -1.0729e-03, + -1.4582e-03, 3.5400e-03], + [ 1.6966e-03, 1.0023e-03, -3.0594e-03, ..., -5.2071e-04, + 7.5531e-04, 7.8726e-04], + [ 5.6915e-03, 7.6580e-04, 1.9646e-03, ..., 1.2293e-03, + 2.2373e-03, 1.0651e-02]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0151, -0.0007, 0.0006, 0.0165, -0.0035, -0.0051, 0.0132, 0.0230, + -0.0353, 0.0460], device='cuda:0'), grad: tensor([-0.0012, -0.0440, 0.0127, 0.0030, -0.0077, -0.0084, 0.0207, 0.0199, + -0.0317, 0.0367], device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 216.98, cls_loss 0.5045 cls_loss_mapping 0.0054 cls_loss_causal 0.4803 re_mapping 0.0080 re_causal 0.0195 /// teacc 98.78 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0764, 0.0661, -0.0962, ..., -0.1009, -0.0868, 0.0020], + [-0.0605, -0.1320, -0.0021, ..., -0.0589, -0.0450, -0.0870], + [ 0.0331, -0.0828, 0.0505, ..., 0.1563, -0.1094, -0.0441], + ..., + [-0.0903, -0.1574, 0.0872, ..., -0.0149, -0.0287, 0.0711], + [ 0.0185, 0.0364, -0.0233, ..., -0.0806, -0.0932, -0.0063], + [-0.1413, -0.0089, -0.0087, ..., -0.1409, 0.1013, 0.0175]], + device='cuda:0'), grad: tensor([[-2.6584e-04, -1.9817e-03, 4.6849e-04, ..., 2.7609e-04, + 1.1301e-03, 3.4999e-06], + [ 1.3702e-05, 2.8276e-04, -1.1311e-03, ..., 2.1672e-04, + 4.2629e-04, 1.0881e-03], + [ 1.3313e-03, -1.8024e-04, 1.2696e-04, ..., -6.8951e-04, + 2.3222e-04, 3.2330e-04], + ..., + [-7.7820e-03, -3.3379e-04, 1.2894e-03, ..., 2.7251e-04, + 6.4754e-04, 2.7504e-03], + [ 1.7204e-03, 1.1482e-03, 6.8092e-04, ..., 2.7132e-04, + 4.0030e-04, 7.5197e-04], + [ 7.0858e-04, -7.6103e-04, 1.1950e-03, ..., 2.7046e-03, + 1.9531e-03, -9.2447e-05]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0149, -0.0006, 0.0005, 0.0175, -0.0030, -0.0048, 0.0130, 0.0224, + -0.0354, 0.0455], device='cuda:0'), grad: tensor([-0.0044, 0.0033, 0.0166, -0.0022, -0.0456, 0.0135, 0.0147, -0.0083, + -0.0081, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 217.26, cls_loss 0.5128 cls_loss_mapping 0.0052 cls_loss_causal 0.4803 re_mapping 0.0075 re_causal 0.0185 /// teacc 98.79 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0760, 0.0663, -0.0966, ..., -0.0998, -0.0889, 0.0016], + [-0.0616, -0.1334, -0.0023, ..., -0.0593, -0.0445, -0.0858], + [ 0.0327, -0.0838, 0.0505, ..., 0.1559, -0.1098, -0.0453], + ..., + [-0.0914, -0.1569, 0.0873, ..., -0.0144, -0.0301, 0.0712], + [ 0.0186, 0.0357, -0.0240, ..., -0.0813, -0.0936, -0.0078], + [-0.1417, -0.0093, -0.0077, ..., -0.1406, 0.1016, 0.0180]], + device='cuda:0'), grad: tensor([[ 4.8876e-04, 6.0892e-04, 5.8556e-04, ..., 1.0335e-04, + 3.5906e-04, 9.6273e-04], + [ 2.1899e-04, 1.8911e-03, 4.2038e-03, ..., 1.4753e-03, + 6.8951e-04, 3.1738e-03], + [ 7.1383e-04, 2.0814e-04, 1.3380e-03, ..., 4.8637e-04, + 2.5249e-04, 1.2054e-03], + ..., + [ 3.7766e-04, 4.5514e-04, -4.9896e-03, ..., -5.2214e-04, + 7.7343e-04, -6.9618e-04], + [ 1.0834e-03, 9.5177e-04, 7.7820e-04, ..., 1.7929e-04, + 4.9114e-04, 1.6899e-03], + [-6.9261e-05, -1.9321e-03, -2.3422e-03, ..., 2.7490e-04, + -6.5002e-03, -5.7106e-03]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0154, -0.0005, 0.0008, 0.0167, -0.0030, -0.0037, 0.0132, 0.0219, + -0.0348, 0.0442], device='cuda:0'), grad: tensor([ 0.0116, -0.0010, 0.0133, -0.0167, 0.0077, -0.0196, 0.0046, -0.0073, + 0.0150, -0.0077], device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 216.94, cls_loss 0.5197 cls_loss_mapping 0.0064 cls_loss_causal 0.4936 re_mapping 0.0075 re_causal 0.0182 /// teacc 98.91 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0756, 0.0660, -0.0971, ..., -0.1008, -0.0881, 0.0016], + [-0.0602, -0.1318, -0.0016, ..., -0.0598, -0.0448, -0.0860], + [ 0.0323, -0.0845, 0.0521, ..., 0.1577, -0.1117, -0.0452], + ..., + [-0.0914, -0.1584, 0.0874, ..., -0.0152, -0.0296, 0.0718], + [ 0.0174, 0.0350, -0.0250, ..., -0.0817, -0.0929, -0.0073], + [-0.1410, -0.0077, -0.0076, ..., -0.1404, 0.1012, 0.0172]], + device='cuda:0'), grad: tensor([[ 4.1270e-04, 2.8461e-05, 2.6488e-04, ..., 6.3956e-05, + 3.6812e-04, 6.7902e-04], + [-7.1764e-04, 7.9691e-05, 7.2527e-04, ..., 1.4663e-04, + -5.3406e-03, 9.2983e-04], + [ 2.9588e-04, 7.2517e-03, 8.4610e-03, ..., 1.1200e-02, + 3.8767e-04, 1.0071e-03], + ..., + [ 9.2566e-05, -2.0218e-04, -3.9864e-03, ..., 1.2076e-04, + -2.0695e-03, -2.3594e-03], + [ 4.9639e-04, 8.5711e-05, 5.4646e-04, ..., 1.0771e-04, + 4.9925e-04, 1.0462e-03], + [ 1.0675e-04, 1.7405e-04, 3.4332e-03, ..., 1.3304e-04, + 1.3838e-03, 2.1820e-03]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0141, 0.0004, 0.0008, 0.0161, -0.0026, -0.0045, 0.0135, 0.0223, + -0.0341, 0.0441], device='cuda:0'), grad: tensor([ 0.0101, -0.0107, 0.0293, 0.0222, -0.0093, -0.0691, 0.0248, 0.0009, + 0.0141, -0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 217.05, cls_loss 0.5171 cls_loss_mapping 0.0066 cls_loss_causal 0.4935 re_mapping 0.0082 re_causal 0.0205 /// teacc 98.84 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0742, 0.0661, -0.0982, ..., -0.1010, -0.0882, 0.0006], + [-0.0602, -0.1310, -0.0022, ..., -0.0596, -0.0449, -0.0863], + [ 0.0318, -0.0853, 0.0526, ..., 0.1578, -0.1118, -0.0445], + ..., + [-0.0912, -0.1588, 0.0863, ..., -0.0163, -0.0299, 0.0715], + [ 0.0176, 0.0354, -0.0255, ..., -0.0826, -0.0931, -0.0080], + [-0.1422, -0.0089, -0.0068, ..., -0.1400, 0.1017, 0.0171]], + device='cuda:0'), grad: tensor([[-3.0304e-02, -3.9795e-02, 2.5153e-04, ..., 4.2245e-06, + 1.1474e-04, 3.0947e-04], + [ 2.3580e-04, 2.9579e-05, 3.7932e-04, ..., 2.1234e-05, + 2.3091e-04, 5.4407e-04], + [-8.2445e-04, 6.0946e-05, -2.6741e-03, ..., -8.8155e-05, + -2.4910e-03, -4.5357e-03], + ..., + [ 2.8396e-04, 1.0365e-04, 9.0265e-04, ..., 2.1905e-05, + 1.0376e-03, 1.7376e-03], + [ 8.7214e-04, 1.1005e-03, 1.1511e-03, ..., 4.1819e-04, + 1.1110e-03, 1.0471e-03], + [-9.1743e-04, -2.0790e-03, -6.2656e-04, ..., -9.0170e-04, + -1.3428e-03, -1.2836e-03]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0151, 0.0002, 0.0006, 0.0162, -0.0031, -0.0049, 0.0136, 0.0219, + -0.0345, 0.0450], device='cuda:0'), grad: tensor([-0.0091, 0.0208, -0.0452, 0.0152, 0.0238, -0.0108, 0.0262, 0.0173, + -0.0433, 0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 216.69, cls_loss 0.5284 cls_loss_mapping 0.0086 cls_loss_causal 0.5060 re_mapping 0.0075 re_causal 0.0180 /// teacc 98.71 lr 0.00010000 +Epoch 232, weight, value: tensor([[-7.4027e-02, 6.6639e-02, -9.9123e-02, ..., -1.0170e-01, + -8.8190e-02, 1.0178e-03], + [-5.9879e-02, -1.3252e-01, -1.2692e-04, ..., -5.8554e-02, + -4.4243e-02, -8.5391e-02], + [ 3.2622e-02, -8.4710e-02, 5.2322e-02, ..., 1.5747e-01, + -1.0960e-01, -4.3564e-02], + ..., + [-9.3093e-02, -1.5956e-01, 8.6105e-02, ..., -1.6627e-02, + -2.9192e-02, 7.1240e-02], + [ 1.8403e-02, 3.6425e-02, -2.5548e-02, ..., -8.1421e-02, + -9.3490e-02, -8.8011e-03], + [-1.4267e-01, -8.4241e-03, -6.4305e-03, ..., -1.3997e-01, + 1.0093e-01, 1.7119e-02]], device='cuda:0'), grad: tensor([[ 9.5665e-05, 2.7466e-04, 7.4565e-05, ..., 4.8131e-06, + 4.1127e-04, 8.4972e-04], + [ 1.0567e-03, 3.8773e-05, 6.5327e-05, ..., 8.7544e-07, + 1.4591e-03, 1.6069e-03], + [ 3.3593e-04, 8.8692e-05, -9.1568e-06, ..., 3.4952e-04, + 2.6345e-04, -9.6130e-04], + ..., + [ 5.1260e-04, 1.1206e-04, 1.7822e-02, ..., 5.6505e-04, + 1.1276e-02, 1.9474e-03], + [-1.3189e-03, -4.7565e-05, 1.0198e-04, ..., 4.1187e-05, + -3.1738e-03, 6.6137e-04], + [ 3.4481e-05, -8.4162e-05, -2.0432e-02, ..., -4.6700e-05, + -1.3451e-02, -6.8893e-03]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0151, 0.0014, 0.0005, 0.0166, -0.0039, -0.0053, 0.0131, 0.0220, + -0.0347, 0.0452], device='cuda:0'), grad: tensor([ 0.0156, 0.0188, -0.0150, -0.0222, 0.0223, 0.0123, 0.0192, -0.0052, + -0.0134, -0.0325], device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 216.94, cls_loss 0.5010 cls_loss_mapping 0.0076 cls_loss_causal 0.4813 re_mapping 0.0081 re_causal 0.0191 /// teacc 98.91 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.0736, 0.0668, -0.0995, ..., -0.1010, -0.0875, 0.0003], + [-0.0597, -0.1315, 0.0017, ..., -0.0584, -0.0446, -0.0854], + [ 0.0322, -0.0849, 0.0529, ..., 0.1576, -0.1086, -0.0443], + ..., + [-0.0919, -0.1596, 0.0863, ..., -0.0164, -0.0305, 0.0720], + [ 0.0181, 0.0354, -0.0268, ..., -0.0823, -0.0937, -0.0085], + [-0.1442, -0.0087, -0.0063, ..., -0.1412, 0.1015, 0.0169]], + device='cuda:0'), grad: tensor([[ 2.2945e-03, 3.7594e-03, 2.1711e-05, ..., 6.6042e-04, + 2.4354e-04, 8.3745e-05], + [-2.7132e-04, 6.0529e-05, 5.9366e-05, ..., 6.5744e-05, + 5.7840e-04, 2.6274e-04], + [-1.6418e-02, 1.4317e-04, 6.3181e-05, ..., -4.9686e-04, + 2.0838e-04, 2.1470e-04], + ..., + [ 8.7380e-05, 4.1693e-05, 2.6188e-03, ..., 1.8418e-05, + 2.9716e-03, 4.2076e-03], + [ 2.6016e-03, 2.6531e-03, 2.7037e-04, ..., 3.3712e-04, + 1.4181e-03, 6.6996e-04], + [ 1.3745e-04, 1.9252e-04, 7.6637e-03, ..., 3.1292e-05, + 3.1471e-03, 1.0040e-02]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0140, 0.0014, 0.0016, 0.0169, -0.0030, -0.0045, 0.0113, 0.0217, + -0.0345, 0.0451], device='cuda:0'), grad: tensor([ 0.0120, -0.0084, -0.0221, -0.0120, -0.0146, 0.0026, 0.0132, 0.0128, + 0.0172, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 216.74, cls_loss 0.4765 cls_loss_mapping 0.0043 cls_loss_causal 0.4504 re_mapping 0.0080 re_causal 0.0197 /// teacc 98.84 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.0735, 0.0663, -0.0993, ..., -0.1017, -0.0873, 0.0014], + [-0.0595, -0.1316, 0.0012, ..., -0.0579, -0.0453, -0.0872], + [ 0.0327, -0.0847, 0.0523, ..., 0.1578, -0.1077, -0.0446], + ..., + [-0.0923, -0.1597, 0.0869, ..., -0.0153, -0.0300, 0.0706], + [ 0.0180, 0.0351, -0.0274, ..., -0.0822, -0.0957, -0.0092], + [-0.1436, -0.0091, -0.0057, ..., -0.1407, 0.1029, 0.0191]], + device='cuda:0'), grad: tensor([[ 7.4768e-04, 5.7888e-04, 1.9522e-03, ..., 2.8419e-04, + 2.2113e-04, 1.0281e-03], + [ 3.2115e-04, 5.2065e-05, 2.1255e-04, ..., -7.1287e-05, + 3.6144e-04, -1.7471e-03], + [ 4.6959e-03, 2.4948e-03, 3.2558e-03, ..., 7.5378e-03, + -4.3488e-04, -1.4200e-03], + ..., + [ 8.3685e-04, 5.8317e-04, 3.5172e-03, ..., -1.0666e-02, + 1.0147e-03, 1.9274e-03], + [-1.5533e-04, -2.1477e-03, -8.8501e-03, ..., 1.6899e-03, + -1.2608e-03, -1.7033e-03], + [ 7.5531e-04, 2.1756e-04, -1.0681e-04, ..., 4.4060e-03, + -9.0981e-04, 8.2636e-04]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0145, 0.0016, 0.0022, 0.0162, -0.0029, -0.0046, 0.0106, 0.0219, + -0.0354, 0.0457], device='cuda:0'), grad: tensor([ 0.0226, -0.0167, -0.0177, -0.0732, 0.0323, 0.0329, 0.0133, -0.0049, + -0.0173, 0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 216.66, cls_loss 0.5264 cls_loss_mapping 0.0057 cls_loss_causal 0.4980 re_mapping 0.0073 re_causal 0.0179 /// teacc 98.80 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.0738, 0.0660, -0.0976, ..., -0.1022, -0.0849, 0.0036], + [-0.0598, -0.1323, 0.0008, ..., -0.0590, -0.0454, -0.0873], + [ 0.0325, -0.0851, 0.0510, ..., 0.1570, -0.1082, -0.0445], + ..., + [-0.0927, -0.1599, 0.0874, ..., -0.0144, -0.0301, 0.0713], + [ 0.0182, 0.0350, -0.0276, ..., -0.0825, -0.0962, -0.0092], + [-0.1439, -0.0087, -0.0060, ..., -0.1413, 0.1040, 0.0187]], + device='cuda:0'), grad: tensor([[ 2.1350e-04, 2.9516e-04, 3.3051e-05, ..., 4.6706e-04, + 7.2479e-04, 2.9106e-03], + [ 7.8157e-06, 1.4231e-05, 1.3885e-03, ..., 5.4598e-04, + 1.2817e-03, 2.6073e-03], + [-3.1605e-03, -5.7602e-03, -5.1270e-03, ..., -4.7760e-03, + -2.1286e-03, -2.2144e-03], + ..., + [ 2.9349e-04, 5.3453e-04, 3.7041e-03, ..., 9.5558e-04, + 1.0796e-03, 5.5695e-03], + [ 4.7374e-04, 8.6737e-04, -3.9506e-04, ..., 7.0667e-04, + -1.0242e-03, -2.3308e-03], + [ 9.8038e-04, 1.7900e-03, -1.9276e-04, ..., 2.5330e-03, + 2.5444e-03, 1.0939e-03]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0156, 0.0015, 0.0019, 0.0159, -0.0028, -0.0059, 0.0118, 0.0218, + -0.0349, 0.0450], device='cuda:0'), grad: tensor([ 0.0052, 0.0025, 0.0027, 0.0459, -0.0199, -0.0403, -0.0563, 0.0382, + 0.0007, 0.0212], device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 216.39, cls_loss 0.4830 cls_loss_mapping 0.0049 cls_loss_causal 0.4567 re_mapping 0.0080 re_causal 0.0201 /// teacc 98.70 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.0744, 0.0667, -0.0968, ..., -0.1029, -0.0860, 0.0035], + [-0.0603, -0.1329, 0.0023, ..., -0.0595, -0.0453, -0.0869], + [ 0.0328, -0.0863, 0.0517, ..., 0.1579, -0.1084, -0.0454], + ..., + [-0.0936, -0.1602, 0.0877, ..., -0.0138, -0.0306, 0.0708], + [ 0.0185, 0.0357, -0.0289, ..., -0.0832, -0.0966, -0.0106], + [-0.1447, -0.0090, -0.0067, ..., -0.1418, 0.1046, 0.0194]], + device='cuda:0'), grad: tensor([[ 2.4700e-04, 3.9458e-04, 8.6367e-05, ..., 2.6271e-05, + 7.4029e-05, 1.5438e-04], + [ 1.2767e-04, 3.4237e-04, 4.9770e-05, ..., 1.6525e-05, + 4.8137e-04, 6.0654e-04], + [ 4.3464e-04, 5.1451e-04, -2.6762e-05, ..., -2.0909e-04, + 2.8700e-05, 8.5056e-05], + ..., + [ 1.5378e-04, 2.3675e-04, 1.8609e-04, ..., 9.9897e-05, + 2.1243e-04, -8.6260e-04], + [-2.1152e-03, -1.5478e-03, -4.1890e-04, ..., 3.5111e-07, + 6.8140e-04, -4.7064e-04], + [-9.6369e-04, -2.8343e-03, 7.5459e-05, ..., 4.9174e-05, + -5.0316e-03, -4.1580e-03]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0156, 0.0009, 0.0014, 0.0159, -0.0036, -0.0051, 0.0121, 0.0219, + -0.0343, 0.0452], device='cuda:0'), grad: tensor([ 0.0105, 0.0111, 0.0080, -0.0193, -0.0051, 0.0107, 0.0159, 0.0033, + 0.0073, -0.0424], device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 216.43, cls_loss 0.5199 cls_loss_mapping 0.0072 cls_loss_causal 0.4930 re_mapping 0.0071 re_causal 0.0178 /// teacc 98.67 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.0746, 0.0663, -0.0972, ..., -0.1038, -0.0865, 0.0036], + [-0.0598, -0.1327, 0.0023, ..., -0.0601, -0.0448, -0.0862], + [ 0.0331, -0.0871, 0.0519, ..., 0.1583, -0.1098, -0.0456], + ..., + [-0.0936, -0.1596, 0.0875, ..., -0.0149, -0.0310, 0.0705], + [ 0.0188, 0.0369, -0.0295, ..., -0.0837, -0.0978, -0.0114], + [-0.1443, -0.0089, -0.0072, ..., -0.1416, 0.1042, 0.0205]], + device='cuda:0'), grad: tensor([[ 8.0526e-05, 4.5709e-06, 2.2757e-04, ..., 1.6105e-04, + 3.9250e-05, 1.3075e-03], + [ 4.9448e-04, 9.0122e-05, 3.4541e-05, ..., 1.9622e-04, + 3.4541e-05, 2.2678e-03], + [ 1.3605e-05, 1.3185e-04, -1.8854e-03, ..., 2.5299e-02, + 1.4591e-03, 1.9264e-03], + ..., + [ 1.6165e-04, 1.8084e-04, 1.9779e-03, ..., 9.6321e-04, + 1.1644e-03, 2.5501e-03], + [-7.1168e-05, -4.9162e-04, 4.7231e-04, ..., 1.4486e-03, + -7.5912e-04, 8.3637e-04], + [ 1.1945e-04, 8.5771e-05, -1.0614e-03, ..., 1.3340e-04, + -8.6308e-04, 1.4219e-03]], device='cuda:0') +Epoch 237, bias, value: tensor([ 0.0158, 0.0019, 0.0007, 0.0156, -0.0030, -0.0050, 0.0121, 0.0217, + -0.0349, 0.0450], device='cuda:0'), grad: tensor([ 0.0109, -0.0122, 0.0320, -0.0180, -0.0418, 0.0090, -0.0116, 0.0168, + 0.0048, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 216.35, cls_loss 0.5083 cls_loss_mapping 0.0054 cls_loss_causal 0.4881 re_mapping 0.0074 re_causal 0.0182 /// teacc 98.65 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.0750, 0.0660, -0.0969, ..., -0.1033, -0.0860, 0.0047], + [-0.0603, -0.1341, 0.0026, ..., -0.0605, -0.0441, -0.0871], + [ 0.0346, -0.0873, 0.0526, ..., 0.1590, -0.1110, -0.0447], + ..., + [-0.0937, -0.1607, 0.0874, ..., -0.0150, -0.0309, 0.0705], + [ 0.0172, 0.0361, -0.0294, ..., -0.0845, -0.0981, -0.0116], + [-0.1449, -0.0079, -0.0073, ..., -0.1409, 0.1044, 0.0199]], + device='cuda:0'), grad: tensor([[ 3.8838e-04, 1.3914e-03, 1.6183e-05, ..., 2.1851e-04, + 2.6178e-04, 4.6325e-04], + [ 7.6950e-05, 2.2840e-04, -2.6536e-04, ..., 3.3808e-04, + 8.5592e-05, 5.0449e-04], + [ 1.7227e-02, 2.2018e-02, 6.2943e-05, ..., 6.2714e-03, + 3.5971e-05, 3.2234e-04], + ..., + [ 1.7238e-04, 1.8311e-04, 1.6224e-04, ..., 1.4925e-04, + 1.6332e-04, 4.8542e-04], + [ 1.2222e-02, 1.1841e-02, -2.8157e-04, ..., 8.5545e-04, + -3.9315e-04, -5.0735e-04], + [ 5.7077e-04, 9.0647e-04, 4.5156e-04, ..., 2.5892e-04, + 7.8249e-04, -2.4853e-03]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0160, 0.0026, 0.0011, 0.0154, -0.0039, -0.0059, 0.0117, 0.0220, + -0.0347, 0.0457], device='cuda:0'), grad: tensor([ 0.0097, 0.0120, 0.0520, -0.0277, -0.0383, 0.0300, -0.0397, 0.0264, + -0.0205, -0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 216.44, cls_loss 0.5159 cls_loss_mapping 0.0059 cls_loss_causal 0.4848 re_mapping 0.0073 re_causal 0.0184 /// teacc 98.59 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.0759, 0.0650, -0.0962, ..., -0.1041, -0.0851, 0.0053], + [-0.0604, -0.1347, 0.0026, ..., -0.0602, -0.0446, -0.0860], + [ 0.0350, -0.0886, 0.0529, ..., 0.1587, -0.1110, -0.0447], + ..., + [-0.0939, -0.1623, 0.0866, ..., -0.0152, -0.0312, 0.0701], + [ 0.0176, 0.0366, -0.0299, ..., -0.0846, -0.0983, -0.0121], + [-0.1444, -0.0075, -0.0071, ..., -0.1420, 0.1052, 0.0204]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0011, 0.0008, ..., 0.0011, 0.0003, 0.0013], + [ 0.0003, 0.0005, -0.0018, ..., -0.0015, 0.0002, -0.0023], + [ 0.0008, 0.0011, 0.0026, ..., 0.0015, 0.0005, 0.0028], + ..., + [ 0.0002, -0.0018, -0.0064, ..., -0.0033, -0.0046, -0.0066], + [-0.0004, -0.0014, -0.0015, ..., -0.0022, 0.0002, -0.0007], + [ 0.0008, 0.0019, 0.0067, ..., 0.0027, 0.0074, 0.0062]], + device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0164, 0.0024, 0.0003, 0.0156, -0.0038, -0.0067, 0.0119, 0.0224, + -0.0349, 0.0462], device='cuda:0'), grad: tensor([ 0.0183, 0.0023, 0.0236, 0.0166, -0.0039, -0.0021, -0.0211, -0.0067, + -0.0575, 0.0305], device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 216.50, cls_loss 0.4993 cls_loss_mapping 0.0056 cls_loss_causal 0.4659 re_mapping 0.0074 re_causal 0.0189 /// teacc 98.74 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.0765, 0.0643, -0.0960, ..., -0.1045, -0.0852, 0.0060], + [-0.0619, -0.1371, 0.0046, ..., -0.0588, -0.0444, -0.0858], + [ 0.0353, -0.0878, 0.0521, ..., 0.1590, -0.1101, -0.0454], + ..., + [-0.0948, -0.1640, 0.0870, ..., -0.0146, -0.0311, 0.0702], + [ 0.0180, 0.0371, -0.0292, ..., -0.0856, -0.0958, -0.0115], + [-0.1445, -0.0085, -0.0079, ..., -0.1423, 0.1045, 0.0201]], + device='cuda:0'), grad: tensor([[-2.0035e-02, 6.9976e-05, -7.2441e-03, ..., 7.3338e-04, + 2.7895e-04, 1.3237e-03], + [ 5.9032e-04, -1.9205e-04, 6.1703e-04, ..., 2.8777e-04, + 2.1267e-04, 9.4032e-04], + [ 3.4981e-03, -1.2236e-03, 6.6996e-04, ..., -8.3923e-04, + -1.4849e-03, -3.3302e-03], + ..., + [ 2.3193e-03, -3.3140e-04, -1.8954e-05, ..., 4.6945e-04, + -1.0357e-03, -7.8964e-03], + [ 2.3651e-03, 6.3705e-04, 5.1785e-04, ..., 4.5228e-04, + 2.7370e-04, 1.2274e-03], + [ 5.4626e-03, 5.0116e-04, 6.2180e-04, ..., 4.8351e-04, + 3.5834e-04, 1.8349e-03]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0159, 0.0021, 0.0002, 0.0152, -0.0032, -0.0070, 0.0116, 0.0226, + -0.0341, 0.0465], device='cuda:0'), grad: tensor([-0.0134, -0.0053, -0.0038, 0.0216, 0.0304, 0.0013, -0.0094, -0.0658, + 0.0095, 0.0349], device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 216.47, cls_loss 0.5053 cls_loss_mapping 0.0063 cls_loss_causal 0.4764 re_mapping 0.0073 re_causal 0.0174 /// teacc 98.78 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.0749, 0.0646, -0.0971, ..., -0.1053, -0.0865, 0.0051], + [-0.0623, -0.1385, 0.0041, ..., -0.0584, -0.0449, -0.0862], + [ 0.0340, -0.0881, 0.0516, ..., 0.1585, -0.1103, -0.0456], + ..., + [-0.0952, -0.1659, 0.0880, ..., -0.0140, -0.0313, 0.0715], + [ 0.0185, 0.0371, -0.0287, ..., -0.0854, -0.0953, -0.0103], + [-0.1451, -0.0095, -0.0084, ..., -0.1437, 0.1040, 0.0191]], + device='cuda:0'), grad: tensor([[ 1.1630e-05, 6.8665e-04, 2.4211e-04, ..., 1.1967e-06, + 4.6939e-06, 7.4768e-04], + [ 4.8995e-05, 6.2799e-04, 4.8137e-04, ..., 2.2426e-06, + 1.8425e-03, 1.3504e-03], + [ 5.8860e-05, -1.5421e-03, 2.9683e-04, ..., -1.4439e-05, + 1.8284e-05, -7.1573e-04], + ..., + [ 3.4189e-04, 7.7105e-04, -1.2159e-04, ..., -5.5470e-06, + 6.4516e-04, 1.2312e-03], + [ 8.4519e-05, 7.0763e-04, 2.5487e-04, ..., 6.2548e-06, + 2.5582e-04, 7.5340e-04], + [ 2.7433e-05, 7.3528e-04, 4.2105e-04, ..., 3.5670e-06, + 6.7282e-04, -8.2684e-04]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0154, 0.0014, 0.0003, 0.0158, -0.0027, -0.0080, 0.0113, 0.0234, + -0.0334, 0.0460], device='cuda:0'), grad: tensor([ 0.0184, 0.0290, -0.0449, -0.0139, 0.0046, -0.0469, -0.0095, 0.0216, + 0.0195, 0.0220], device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 217.05, cls_loss 0.4999 cls_loss_mapping 0.0047 cls_loss_causal 0.4801 re_mapping 0.0076 re_causal 0.0191 /// teacc 98.87 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.0755, 0.0644, -0.0979, ..., -0.1054, -0.0864, 0.0042], + [-0.0629, -0.1380, 0.0037, ..., -0.0593, -0.0433, -0.0858], + [ 0.0342, -0.0877, 0.0527, ..., 0.1587, -0.1109, -0.0460], + ..., + [-0.0953, -0.1648, 0.0884, ..., -0.0137, -0.0320, 0.0709], + [ 0.0202, 0.0381, -0.0294, ..., -0.0855, -0.0962, -0.0098], + [-0.1446, -0.0082, -0.0087, ..., -0.1441, 0.1049, 0.0209]], + device='cuda:0'), grad: tensor([[ 1.9860e-04, 3.5691e-04, 6.7616e-04, ..., 1.1462e-04, + 3.0303e-04, 1.3018e-03], + [ 4.0889e-04, 6.7830e-05, -7.4482e-04, ..., -1.7917e-04, + 8.0633e-04, -7.3576e-04], + [-2.0409e-03, 8.3685e-05, -1.5507e-03, ..., -1.6069e-03, + 2.8253e-04, -1.0742e-02], + ..., + [ 4.2534e-04, 1.3888e-04, -2.4166e-03, ..., 3.3474e-04, + 8.1682e-04, 3.1681e-03], + [-1.0500e-03, -1.0300e-03, -9.8133e-04, ..., 1.4043e-04, + -4.0398e-03, -4.4136e-03], + [ 1.0405e-03, 8.3637e-04, 3.5496e-03, ..., 1.2189e-04, + 3.7556e-03, 5.2643e-03]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0145, 0.0016, 0.0002, 0.0163, -0.0029, -0.0078, 0.0115, 0.0234, + -0.0329, 0.0456], device='cuda:0'), grad: tensor([ 0.0128, -0.0290, -0.0151, 0.0159, 0.0131, 0.0109, -0.0086, 0.0061, + -0.0389, 0.0327], device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 216.82, cls_loss 0.5156 cls_loss_mapping 0.0047 cls_loss_causal 0.4896 re_mapping 0.0075 re_causal 0.0196 /// teacc 98.88 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.0752, 0.0644, -0.0981, ..., -0.1048, -0.0873, 0.0044], + [-0.0629, -0.1377, 0.0028, ..., -0.0603, -0.0450, -0.0866], + [ 0.0358, -0.0868, 0.0523, ..., 0.1587, -0.1093, -0.0455], + ..., + [-0.0950, -0.1647, 0.0883, ..., -0.0130, -0.0324, 0.0722], + [ 0.0203, 0.0369, -0.0282, ..., -0.0857, -0.0962, -0.0099], + [-0.1441, -0.0066, -0.0089, ..., -0.1447, 0.1057, 0.0202]], + device='cuda:0'), grad: tensor([[-2.0714e-03, -4.6158e-03, 7.1049e-04, ..., 1.8612e-05, + 1.1474e-05, 7.2622e-04], + [ 2.8181e-04, 3.0184e-04, 2.7013e-04, ..., 1.8978e-04, + 6.4373e-05, 6.7425e-04], + [ 2.2762e-06, 1.7014e-03, -2.0428e-03, ..., -1.3142e-03, + -7.0429e-04, -2.2125e-03], + ..., + [ 4.1676e-04, -5.1594e-04, -1.4992e-03, ..., 4.7302e-04, + 7.6473e-05, -8.4496e-04], + [ 1.6525e-02, 8.2779e-03, -4.1127e-04, ..., 2.4164e-04, + 2.6488e-04, 4.6992e-04], + [ 2.8086e-04, 9.3126e-04, 9.7942e-04, ..., 8.5607e-06, + 1.3793e-04, 9.3651e-04]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0141, 0.0012, 0.0011, 0.0156, -0.0033, -0.0070, 0.0112, 0.0246, + -0.0328, 0.0447], device='cuda:0'), grad: tensor([-0.0004, 0.0349, 0.0050, 0.0089, 0.0139, -0.0858, 0.0223, 0.0065, + 0.0360, -0.0414], device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 216.90, cls_loss 0.5048 cls_loss_mapping 0.0056 cls_loss_causal 0.4814 re_mapping 0.0073 re_causal 0.0197 /// teacc 98.94 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.0763, 0.0633, -0.0982, ..., -0.1058, -0.0873, 0.0049], + [-0.0627, -0.1382, 0.0029, ..., -0.0601, -0.0459, -0.0869], + [ 0.0365, -0.0860, 0.0530, ..., 0.1590, -0.1097, -0.0459], + ..., + [-0.0951, -0.1654, 0.0885, ..., -0.0136, -0.0314, 0.0720], + [ 0.0199, 0.0366, -0.0287, ..., -0.0844, -0.0978, -0.0108], + [-0.1442, -0.0064, -0.0088, ..., -0.1451, 0.1062, 0.0198]], + device='cuda:0'), grad: tensor([[-3.0090e-02, -3.3081e-02, 6.2466e-04, ..., 1.1814e-04, + 1.8084e-04, 5.4359e-04], + [-6.7472e-04, -1.4963e-03, -1.9608e-03, ..., -2.7313e-03, + -2.7142e-03, 3.0303e-04], + [ 1.7576e-03, 1.1158e-03, 1.2070e-04, ..., 5.2500e-04, + 9.5844e-04, 7.4482e-04], + ..., + [-9.1970e-05, -1.7338e-03, -1.0101e-02, ..., -3.2368e-03, + -1.9779e-03, -8.1024e-03], + [-2.5330e-02, -1.6647e-02, 2.1248e-03, ..., 3.3054e-03, + 3.2501e-03, 1.9140e-03], + [ 5.8079e-04, -8.2254e-05, 1.5001e-03, ..., 6.9046e-04, + 5.2643e-04, 2.6093e-03]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0144, 0.0007, 0.0010, 0.0155, -0.0022, -0.0067, 0.0109, 0.0237, + -0.0330, 0.0453], device='cuda:0'), grad: tensor([-0.0720, 0.0048, 0.0291, -0.0052, 0.0001, -0.0451, 0.0706, -0.0170, + 0.0415, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 216.66, cls_loss 0.5122 cls_loss_mapping 0.0067 cls_loss_causal 0.4841 re_mapping 0.0073 re_causal 0.0179 /// teacc 98.81 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.0765, 0.0635, -0.0988, ..., -0.1057, -0.0882, 0.0050], + [-0.0634, -0.1385, 0.0028, ..., -0.0589, -0.0459, -0.0867], + [ 0.0354, -0.0860, 0.0523, ..., 0.1584, -0.1100, -0.0454], + ..., + [-0.0962, -0.1656, 0.0878, ..., -0.0138, -0.0321, 0.0726], + [ 0.0208, 0.0369, -0.0291, ..., -0.0843, -0.0989, -0.0116], + [-0.1436, -0.0060, -0.0075, ..., -0.1451, 0.1064, 0.0201]], + device='cuda:0'), grad: tensor([[ 6.5625e-05, -2.5749e-04, 4.4680e-04, ..., 1.8373e-05, + 6.6233e-04, 9.3985e-04], + [ 3.6687e-05, 6.6042e-04, -1.0586e-03, ..., 1.7108e-06, + 1.5676e-04, 7.4053e-04], + [ 4.5389e-05, -6.5422e-04, 2.2769e-04, ..., 6.7540e-06, + 1.3578e-04, -3.1700e-03], + ..., + [ 5.0515e-05, 6.0892e-04, 4.5729e-04, ..., 5.9344e-06, + 3.0422e-04, 6.1417e-04], + [ 6.4015e-05, 7.1764e-04, 5.3596e-04, ..., 4.9323e-05, + 6.4278e-04, 9.8610e-04], + [ 2.4605e-04, 6.4011e-03, 3.7537e-03, ..., 5.7936e-04, + 7.3662e-03, 5.4321e-03]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0147, 0.0010, 0.0010, 0.0149, -0.0028, -0.0059, 0.0108, 0.0235, + -0.0336, 0.0460], device='cuda:0'), grad: tensor([-0.0059, 0.0058, -0.0440, -0.0387, -0.0668, 0.0201, 0.0293, 0.0245, + 0.0154, 0.0604], device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 216.35, cls_loss 0.5142 cls_loss_mapping 0.0070 cls_loss_causal 0.4857 re_mapping 0.0075 re_causal 0.0191 /// teacc 98.64 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.0776, 0.0631, -0.0984, ..., -0.1048, -0.0886, 0.0054], + [-0.0630, -0.1389, 0.0019, ..., -0.0586, -0.0464, -0.0870], + [ 0.0354, -0.0862, 0.0527, ..., 0.1585, -0.1115, -0.0454], + ..., + [-0.0963, -0.1663, 0.0877, ..., -0.0150, -0.0341, 0.0717], + [ 0.0198, 0.0374, -0.0288, ..., -0.0855, -0.0996, -0.0110], + [-0.1445, -0.0068, -0.0077, ..., -0.1451, 0.1071, 0.0200]], + device='cuda:0'), grad: tensor([[ 4.8685e-04, 1.9515e-04, 2.3210e-04, ..., 1.4591e-04, + 3.0804e-04, 7.9727e-04], + [-2.3460e-03, 5.0116e-04, -5.1546e-04, ..., -4.0054e-03, + -3.1090e-03, 2.5597e-03], + [ 1.4534e-03, 1.7047e-04, 2.3052e-05, ..., 1.9205e-04, + 1.0624e-03, 1.4381e-03], + ..., + [ 8.0228e-05, 4.2295e-04, 4.8866e-03, ..., 2.7633e-04, + 2.1152e-03, 2.8248e-03], + [-8.8196e-03, -3.0594e-03, 2.0099e-04, ..., 5.0592e-04, + -1.0201e-02, -1.0017e-02], + [ 1.0262e-03, 3.5644e-04, 7.3576e-04, ..., 3.8528e-04, + 8.5688e-04, 1.8044e-03]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0149, 0.0012, 0.0016, 0.0151, -0.0021, -0.0056, 0.0106, 0.0224, + -0.0340, 0.0457], device='cuda:0'), grad: tensor([ 0.0098, 0.0007, 0.0219, 0.0076, 0.0354, -0.0153, -0.0158, 0.0022, + -0.0642, 0.0177], device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 216.55, cls_loss 0.5412 cls_loss_mapping 0.0060 cls_loss_causal 0.5127 re_mapping 0.0078 re_causal 0.0196 /// teacc 98.81 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.0770, 0.0642, -0.0991, ..., -0.1061, -0.0900, 0.0058], + [-0.0628, -0.1377, 0.0012, ..., -0.0584, -0.0481, -0.0884], + [ 0.0339, -0.0867, 0.0530, ..., 0.1582, -0.1104, -0.0450], + ..., + [-0.0963, -0.1674, 0.0870, ..., -0.0144, -0.0350, 0.0719], + [ 0.0200, 0.0364, -0.0288, ..., -0.0859, -0.1006, -0.0116], + [-0.1455, -0.0069, -0.0075, ..., -0.1444, 0.1073, 0.0194]], + device='cuda:0'), grad: tensor([[ 2.1183e-04, 8.2779e-04, 7.0715e-04, ..., 9.2313e-06, + 2.3115e-04, -9.8228e-04], + [ 3.6740e-04, 4.8304e-04, 1.1024e-03, ..., -4.0680e-06, + 1.8096e-04, -6.2447e-03], + [ 2.5034e-05, 1.0604e-04, 1.2436e-03, ..., 7.6151e-04, + 1.6677e-04, 2.3327e-03], + ..., + [-9.8324e-04, -2.7447e-03, -1.6983e-02, ..., -8.1396e-04, + -6.3591e-03, -1.1269e-02], + [ 1.8704e-04, 4.0960e-04, 9.5654e-04, ..., 1.9744e-05, + 2.4533e-04, 3.0994e-03], + [ 5.8460e-04, 9.1696e-04, 1.2177e-02, ..., 3.1382e-05, + 5.0240e-03, 1.0521e-02]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0150, 0.0008, 0.0021, 0.0152, -0.0028, -0.0060, 0.0113, 0.0222, + -0.0337, 0.0455], device='cuda:0'), grad: tensor([-0.0054, -0.0716, 0.0195, 0.0226, 0.0170, 0.0091, -0.0224, -0.0583, + 0.0230, 0.0665], device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 216.52, cls_loss 0.4956 cls_loss_mapping 0.0047 cls_loss_causal 0.4714 re_mapping 0.0073 re_causal 0.0174 /// teacc 98.67 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.0761, 0.0649, -0.0979, ..., -0.1059, -0.0895, 0.0059], + [-0.0633, -0.1376, 0.0013, ..., -0.0588, -0.0468, -0.0867], + [ 0.0334, -0.0877, 0.0521, ..., 0.1576, -0.1117, -0.0457], + ..., + [-0.0947, -0.1669, 0.0874, ..., -0.0143, -0.0351, 0.0723], + [ 0.0202, 0.0360, -0.0270, ..., -0.0851, -0.1000, -0.0112], + [-0.1464, -0.0079, -0.0081, ..., -0.1450, 0.1072, 0.0190]], + device='cuda:0'), grad: tensor([[ 4.0207e-03, 4.2038e-03, 1.0653e-03, ..., 2.0351e-03, + 3.6049e-04, 8.6117e-04], + [ 1.2207e-03, 5.2404e-04, 3.2115e-04, ..., 2.2018e-04, + 2.9802e-04, 1.2970e-03], + [-7.7744e-03, 3.5596e-04, 1.9484e-03, ..., -4.7073e-03, + 3.0971e-04, 1.1883e-03], + ..., + [ 1.9436e-03, 5.8270e-04, 9.8419e-04, ..., 9.7847e-04, + 8.2541e-04, 1.6737e-03], + [ 2.8248e-03, 4.6501e-03, 3.2940e-03, ..., 9.5129e-04, + 1.1463e-03, 2.3556e-03], + [ 5.4407e-04, 5.9158e-05, -1.3000e-02, ..., -2.6760e-03, + 3.9864e-03, 6.3667e-03]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0151, 0.0014, 0.0013, 0.0152, -0.0023, -0.0062, 0.0117, 0.0230, + -0.0336, 0.0441], device='cuda:0'), grad: tensor([ 0.0256, 0.0215, 0.0010, 0.0017, -0.0342, 0.0451, -0.0353, 0.0197, + 0.0305, -0.0756], device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 216.42, cls_loss 0.4912 cls_loss_mapping 0.0051 cls_loss_causal 0.4596 re_mapping 0.0074 re_causal 0.0179 /// teacc 98.66 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.0766, 0.0645, -0.0989, ..., -0.1071, -0.0911, 0.0055], + [-0.0634, -0.1372, 0.0008, ..., -0.0587, -0.0473, -0.0869], + [ 0.0334, -0.0883, 0.0523, ..., 0.1560, -0.1114, -0.0463], + ..., + [-0.0948, -0.1680, 0.0885, ..., -0.0117, -0.0347, 0.0727], + [ 0.0202, 0.0363, -0.0265, ..., -0.0850, -0.0997, -0.0113], + [-0.1464, -0.0070, -0.0078, ..., -0.1455, 0.1071, 0.0191]], + device='cuda:0'), grad: tensor([[-5.2929e-04, 3.4733e-03, 5.0831e-04, ..., 9.6023e-05, + -1.7443e-03, -4.9820e-03], + [ 1.4746e-04, -8.9359e-04, 1.9360e-04, ..., 3.4976e-04, + -9.4175e-05, -8.2731e-05], + [ 1.0166e-03, 1.0576e-03, -3.6736e-03, ..., -1.2884e-03, + -1.3173e-04, -5.2986e-03], + ..., + [ 9.9754e-04, 9.9277e-04, 2.0657e-03, ..., 1.2083e-03, + 1.0319e-03, 3.4084e-03], + [-6.2828e-03, -4.1351e-03, 2.4680e-06, ..., -2.5978e-03, + 9.2363e-04, 2.5940e-03], + [ 3.9554e-04, 1.7252e-03, 1.8454e-03, ..., 7.2241e-04, + 1.7662e-03, 4.3335e-03]], device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0145, 0.0012, 0.0006, 0.0168, -0.0022, -0.0060, 0.0117, 0.0227, + -0.0336, 0.0439], device='cuda:0'), grad: tensor([-0.0076, -0.0186, -0.0163, 0.0083, 0.0113, 0.0143, -0.0243, 0.0188, + -0.0087, 0.0229], device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 216.47, cls_loss 0.4994 cls_loss_mapping 0.0044 cls_loss_causal 0.4707 re_mapping 0.0071 re_causal 0.0173 /// teacc 98.77 lr 0.00010000 +Epoch 250, weight, value: tensor([[-7.7217e-02, 6.4318e-02, -9.9880e-02, ..., -1.0734e-01, + -9.1987e-02, 4.5442e-03], + [-6.3571e-02, -1.3776e-01, 2.8957e-05, ..., -5.9381e-02, + -4.6915e-02, -8.7548e-02], + [ 3.4441e-02, -8.7237e-02, 5.3367e-02, ..., 1.5760e-01, + -1.1179e-01, -4.3985e-02], + ..., + [-9.4915e-02, -1.6923e-01, 8.9419e-02, ..., -1.1650e-02, + -3.4268e-02, 7.2982e-02], + [ 2.0525e-02, 3.6374e-02, -2.6922e-02, ..., -8.5980e-02, + -9.8245e-02, -1.1492e-02], + [-1.4686e-01, -6.1683e-03, -7.8076e-03, ..., -1.4561e-01, + 1.0737e-01, 1.9321e-02]], device='cuda:0'), grad: tensor([[ 2.2352e-04, 2.0361e-04, 5.1880e-04, ..., 8.8513e-06, + 9.5546e-05, 7.9346e-04], + [ 4.7302e-04, 5.1451e-04, 2.7105e-05, ..., 3.1274e-06, + 2.0182e-04, -4.7040e-04], + [ 1.7643e-03, 8.7929e-04, 6.7663e-04, ..., 7.7677e-04, + 3.3045e-04, -4.2081e-04], + ..., + [-3.8967e-03, -6.9809e-03, -1.6418e-02, ..., -5.2490e-03, + -5.7335e-03, -1.5556e-02], + [ 1.5114e-02, 4.9782e-03, 1.0811e-02, ..., 2.3479e-03, + 4.3373e-03, 1.0025e-02], + [-1.0590e-02, 2.9812e-03, -1.5116e-03, ..., 9.3889e-04, + -1.1009e-02, -3.0079e-03]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0138, 0.0015, 0.0011, 0.0167, -0.0024, -0.0068, 0.0127, 0.0223, + -0.0336, 0.0444], device='cuda:0'), grad: tensor([ 0.0188, -0.0007, -0.0071, -0.0587, 0.0430, -0.0308, 0.0239, -0.0525, + 0.0826, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 216.39, cls_loss 0.5040 cls_loss_mapping 0.0044 cls_loss_causal 0.4820 re_mapping 0.0077 re_causal 0.0194 /// teacc 98.67 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.0767, 0.0648, -0.0998, ..., -0.1080, -0.0923, 0.0040], + [-0.0641, -0.1380, 0.0023, ..., -0.0587, -0.0445, -0.0873], + [ 0.0340, -0.0877, 0.0524, ..., 0.1575, -0.1127, -0.0446], + ..., + [-0.0943, -0.1691, 0.0891, ..., -0.0104, -0.0345, 0.0729], + [ 0.0197, 0.0367, -0.0284, ..., -0.0877, -0.0993, -0.0126], + [-0.1474, -0.0057, -0.0076, ..., -0.1453, 0.1082, 0.0200]], + device='cuda:0'), grad: tensor([[ 2.1800e-05, 1.3065e-04, 8.9598e-04, ..., 9.7036e-05, + 2.9802e-04, 1.8034e-03], + [ 5.3930e-04, 6.3038e-04, 8.1205e-04, ..., 4.5091e-05, + 2.8253e-04, 1.8158e-03], + [ 3.1125e-06, 7.5936e-05, 6.7282e-04, ..., 7.4446e-05, + 2.2697e-04, 1.3905e-03], + ..., + [ 1.3642e-05, 1.7488e-04, -2.3212e-03, ..., -6.9094e-04, + 3.4523e-04, -3.2120e-03], + [ 1.1292e-03, 1.0414e-03, 3.2735e-04, ..., 1.2290e-04, + 4.5872e-04, 3.3641e-04], + [ 2.6188e-03, 7.0453e-05, -4.3869e-04, ..., 1.1927e-04, + 3.0880e-03, -1.7872e-03]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0137, 0.0010, 0.0005, 0.0168, -0.0027, -0.0071, 0.0143, 0.0222, + -0.0342, 0.0451], device='cuda:0'), grad: tensor([ 0.0209, 0.0225, 0.0156, -0.0093, 0.0289, 0.0328, -0.0707, -0.0311, + -0.0079, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 216.31, cls_loss 0.5077 cls_loss_mapping 0.0050 cls_loss_causal 0.4807 re_mapping 0.0069 re_causal 0.0172 /// teacc 98.72 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.0759, 0.0649, -0.0997, ..., -0.1081, -0.0932, 0.0048], + [-0.0651, -0.1376, 0.0030, ..., -0.0580, -0.0442, -0.0877], + [ 0.0340, -0.0884, 0.0529, ..., 0.1579, -0.1130, -0.0453], + ..., + [-0.0951, -0.1711, 0.0884, ..., -0.0100, -0.0351, 0.0730], + [ 0.0198, 0.0367, -0.0281, ..., -0.0890, -0.0998, -0.0122], + [-0.1482, -0.0057, -0.0069, ..., -0.1464, 0.1083, 0.0200]], + device='cuda:0'), grad: tensor([[ 2.3079e-03, 5.6458e-03, 2.8191e-03, ..., 1.9064e-03, + 2.2182e-03, 2.3766e-03], + [ 1.4377e-04, -7.9823e-04, 2.5845e-03, ..., 3.4428e-03, + 2.9316e-03, 2.2640e-03], + [-9.5701e-04, -2.8872e-04, 3.4409e-03, ..., 4.4098e-03, + 4.6616e-03, 3.6736e-03], + ..., + [ 8.6248e-05, 2.9540e-04, -7.5455e-03, ..., 4.5204e-04, + -5.2032e-03, -4.2381e-03], + [ 1.6088e-03, 2.4853e-03, 1.0519e-03, ..., 3.1614e-04, + 5.1498e-04, 6.5422e-04], + [ 3.8958e-04, 9.5129e-04, -2.7351e-03, ..., -1.3046e-02, + -5.8708e-03, -5.0163e-03]], device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0140, 0.0020, 0.0005, 0.0165, -0.0029, -0.0070, 0.0140, 0.0225, + -0.0340, 0.0441], device='cuda:0'), grad: tensor([ 0.0263, -0.0172, 0.0126, -0.0570, 0.0128, 0.0136, 0.0160, -0.0145, + 0.0124, -0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 216.26, cls_loss 0.5473 cls_loss_mapping 0.0052 cls_loss_causal 0.5165 re_mapping 0.0066 re_causal 0.0165 /// teacc 98.73 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.0760, 0.0649, -0.0994, ..., -0.1074, -0.0928, 0.0050], + [-0.0654, -0.1390, 0.0014, ..., -0.0583, -0.0464, -0.0882], + [ 0.0348, -0.0861, 0.0532, ..., 0.1578, -0.1105, -0.0456], + ..., + [-0.0955, -0.1721, 0.0891, ..., -0.0084, -0.0349, 0.0739], + [ 0.0203, 0.0361, -0.0261, ..., -0.0902, -0.0990, -0.0120], + [-0.1474, -0.0060, -0.0092, ..., -0.1476, 0.1087, 0.0203]], + device='cuda:0'), grad: tensor([[ 0.0006, -0.0003, 0.0009, ..., -0.0004, 0.0007, 0.0019], + [-0.0070, 0.0003, -0.0013, ..., -0.0004, 0.0003, -0.0056], + [ 0.0020, 0.0020, 0.0016, ..., 0.0018, 0.0018, 0.0042], + ..., + [ 0.0004, 0.0005, 0.0016, ..., 0.0005, 0.0017, 0.0020], + [ 0.0171, 0.0184, 0.0005, ..., 0.0002, 0.0004, 0.0036], + [ 0.0008, -0.0038, -0.0072, ..., -0.0046, -0.0119, -0.0032]], + device='cuda:0') +Epoch 253, bias, value: tensor([ 0.0135, 0.0018, 0.0013, 0.0166, -0.0033, -0.0063, 0.0136, 0.0222, + -0.0338, 0.0441], device='cuda:0'), grad: tensor([ 0.0082, -0.0298, 0.0287, 0.0350, -0.0459, 0.0054, -0.0555, 0.0199, + 0.0544, -0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 216.44, cls_loss 0.5016 cls_loss_mapping 0.0055 cls_loss_causal 0.4761 re_mapping 0.0071 re_causal 0.0183 /// teacc 98.76 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.0757, 0.0642, -0.1017, ..., -0.1094, -0.0939, 0.0045], + [-0.0644, -0.1384, 0.0022, ..., -0.0581, -0.0468, -0.0879], + [ 0.0350, -0.0848, 0.0536, ..., 0.1572, -0.1089, -0.0454], + ..., + [-0.0961, -0.1733, 0.0887, ..., -0.0085, -0.0349, 0.0739], + [ 0.0193, 0.0349, -0.0251, ..., -0.0879, -0.0987, -0.0118], + [-0.1474, -0.0063, -0.0090, ..., -0.1482, 0.1087, 0.0199]], + device='cuda:0'), grad: tensor([[-5.1165e-04, -4.8304e-04, 3.4261e-04, ..., -2.4068e-04, + 1.7035e-04, 9.2316e-04], + [ 6.0081e-05, -6.6566e-04, -1.5316e-03, ..., 1.8507e-05, + -1.1845e-03, -7.3814e-03], + [-6.6757e-04, -3.0575e-03, -4.7073e-03, ..., -1.7862e-03, + 2.5344e-04, -6.3944e-04], + ..., + [ 1.1271e-04, 7.3099e-04, -2.6207e-03, ..., 1.8632e-04, + -3.1352e-04, 8.1635e-04], + [-1.5926e-03, 7.9250e-04, 1.4229e-03, ..., 4.7946e-04, + 1.5306e-04, 6.5899e-04], + [-2.9774e-03, -8.4229e-03, 1.9522e-03, ..., 2.3580e-04, + -2.9430e-03, 3.7014e-05]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0121, 0.0025, 0.0020, 0.0168, -0.0037, -0.0066, 0.0128, 0.0224, + -0.0334, 0.0444], device='cuda:0'), grad: tensor([ 0.0022, -0.0524, -0.0415, 0.0437, 0.0211, 0.0078, 0.0133, 0.0066, + 0.0126, -0.0135], device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 216.53, cls_loss 0.5367 cls_loss_mapping 0.0055 cls_loss_causal 0.5152 re_mapping 0.0067 re_causal 0.0166 /// teacc 98.71 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.0748, 0.0644, -0.1018, ..., -0.1089, -0.0933, 0.0049], + [-0.0639, -0.1376, 0.0022, ..., -0.0571, -0.0468, -0.0885], + [ 0.0355, -0.0847, 0.0527, ..., 0.1568, -0.1096, -0.0461], + ..., + [-0.0964, -0.1727, 0.0892, ..., -0.0091, -0.0339, 0.0747], + [ 0.0186, 0.0347, -0.0250, ..., -0.0870, -0.0998, -0.0127], + [-0.1473, -0.0079, -0.0087, ..., -0.1476, 0.1096, 0.0198]], + device='cuda:0'), grad: tensor([[ 2.6178e-04, 6.3610e-04, 7.4434e-04, ..., 4.6277e-04, + 3.3164e-04, 1.5192e-03], + [ 5.6684e-05, 3.2282e-04, 2.8777e-04, ..., 2.1279e-05, + 7.0870e-05, 6.2714e-03], + [ 3.5210e-03, 5.2185e-03, 3.1853e-03, ..., 7.4272e-03, + 1.9550e-04, -5.3864e-03], + ..., + [ 1.5087e-03, 3.6144e-03, 6.1493e-03, ..., 1.0055e-04, + 1.6365e-03, 2.9049e-03], + [ 1.0002e-02, 1.3596e-02, -4.1542e-03, ..., 4.3321e-04, + 5.1689e-04, 9.7466e-04], + [ 1.4091e-04, -2.1839e-03, -4.9629e-03, ..., 5.4896e-05, + -1.5354e-03, -8.5592e-04]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0131, 0.0038, 0.0008, 0.0167, -0.0034, -0.0070, 0.0127, 0.0229, + -0.0343, 0.0440], device='cuda:0'), grad: tensor([-0.0041, 0.0228, -0.0259, -0.0320, -0.0329, -0.0150, 0.0253, 0.0411, + 0.0251, -0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 216.32, cls_loss 0.5119 cls_loss_mapping 0.0058 cls_loss_causal 0.4796 re_mapping 0.0071 re_causal 0.0183 /// teacc 98.74 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.0743, 0.0632, -0.1012, ..., -0.1087, -0.0950, 0.0057], + [-0.0651, -0.1374, 0.0015, ..., -0.0568, -0.0468, -0.0900], + [ 0.0355, -0.0856, 0.0533, ..., 0.1572, -0.1115, -0.0466], + ..., + [-0.0961, -0.1730, 0.0883, ..., -0.0095, -0.0343, 0.0747], + [ 0.0193, 0.0352, -0.0264, ..., -0.0873, -0.1005, -0.0119], + [-0.1477, -0.0065, -0.0068, ..., -0.1476, 0.1102, 0.0196]], + device='cuda:0'), grad: tensor([[-3.8513e-02, -2.1759e-02, 3.0923e-04, ..., 1.3952e-03, + -2.1229e-03, -1.8244e-03], + [ 1.0748e-03, 1.4563e-03, 1.9535e-05, ..., 1.8254e-05, + 1.1168e-03, 5.2404e-04], + [ 1.0319e-03, 1.2903e-03, 5.5313e-04, ..., 1.1044e-03, + 9.3651e-04, 2.9802e-04], + ..., + [ 4.5562e-04, 4.8876e-04, 1.0198e-04, ..., 2.0289e-04, + 5.1212e-04, 3.6907e-04], + [-5.1727e-03, -5.9662e-03, -5.9547e-03, ..., -3.8776e-03, + -5.9128e-03, -3.3402e-04], + [ 3.8013e-03, 3.6316e-03, 5.3177e-03, ..., 3.6755e-03, + 5.0926e-03, 1.6451e-03]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0144, 0.0043, 0.0006, 0.0160, -0.0031, -0.0070, 0.0121, 0.0219, + -0.0346, 0.0450], device='cuda:0'), grad: tensor([-0.0435, 0.0192, 0.0128, 0.0169, -0.0222, -0.0058, 0.0453, 0.0085, + -0.0605, 0.0293], device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 216.44, cls_loss 0.5151 cls_loss_mapping 0.0058 cls_loss_causal 0.4919 re_mapping 0.0074 re_causal 0.0189 /// teacc 98.83 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.0734, 0.0636, -0.1001, ..., -0.1079, -0.0948, 0.0065], + [-0.0667, -0.1385, 0.0016, ..., -0.0559, -0.0469, -0.0904], + [ 0.0352, -0.0861, 0.0518, ..., 0.1567, -0.1130, -0.0475], + ..., + [-0.0963, -0.1727, 0.0897, ..., -0.0099, -0.0337, 0.0749], + [ 0.0203, 0.0357, -0.0273, ..., -0.0871, -0.1002, -0.0119], + [-0.1485, -0.0066, -0.0073, ..., -0.1471, 0.1096, 0.0190]], + device='cuda:0'), grad: tensor([[ 6.6280e-04, 3.1757e-04, 9.2697e-04, ..., 1.6534e-04, + 1.1659e-04, 8.8310e-04], + [ 3.8958e-04, 9.2387e-05, 5.4979e-04, ..., 8.1480e-05, + 1.0908e-04, 4.5276e-04], + [-1.5373e-03, 3.8838e-04, -1.6212e-03, ..., -3.4714e-03, + 2.5153e-04, 6.5708e-04], + ..., + [-3.0684e-04, -1.5821e-03, -6.9797e-05, ..., 1.7042e-03, + 9.8705e-04, -2.1973e-03], + [-5.4283e-03, 1.1981e-04, -3.7975e-03, ..., 9.6858e-05, + -5.3596e-04, 6.7139e-04], + [ 6.1893e-04, 2.4819e-04, -4.3831e-03, ..., -4.7588e-04, + -5.2643e-03, -1.3981e-03]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0147, 0.0040, 0.0014, 0.0168, -0.0036, -0.0067, 0.0109, 0.0228, + -0.0345, 0.0438], device='cuda:0'), grad: tensor([ 0.0195, 0.0204, -0.0204, -0.0674, 0.0198, 0.0419, 0.0157, 0.0171, + -0.0217, -0.0248], device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 216.52, cls_loss 0.4910 cls_loss_mapping 0.0042 cls_loss_causal 0.4631 re_mapping 0.0075 re_causal 0.0194 /// teacc 98.59 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.0749, 0.0636, -0.1004, ..., -0.1079, -0.0951, 0.0057], + [-0.0656, -0.1383, 0.0032, ..., -0.0557, -0.0461, -0.0888], + [ 0.0350, -0.0863, 0.0530, ..., 0.1572, -0.1131, -0.0471], + ..., + [-0.0960, -0.1730, 0.0878, ..., -0.0111, -0.0356, 0.0735], + [ 0.0204, 0.0357, -0.0261, ..., -0.0865, -0.0995, -0.0117], + [-0.1481, -0.0058, -0.0084, ..., -0.1471, 0.1097, 0.0192]], + device='cuda:0'), grad: tensor([[ 2.2471e-05, -1.2531e-03, 7.4267e-05, ..., -9.7692e-05, + 2.8944e-04, 3.6573e-04], + [ 1.2573e-06, -4.6015e-05, 1.3361e-03, ..., 1.2740e-06, + 8.2731e-04, 6.6614e-04], + [ 8.3521e-06, 1.0908e-04, 5.7936e-04, ..., 2.4587e-06, + 2.1100e-04, 2.6941e-04], + ..., + [ 3.7462e-05, 1.4532e-04, -1.8139e-03, ..., 5.3318e-07, + -1.3590e-03, 6.3610e-04], + [ 1.8224e-05, 4.9973e-04, -2.0943e-03, ..., 3.2902e-05, + -4.7135e-04, -2.1248e-03], + [-6.4278e-04, -1.6184e-03, -3.7556e-03, ..., 2.1771e-05, + -3.8624e-03, -5.6610e-03]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0146, 0.0041, 0.0016, 0.0161, -0.0038, -0.0070, 0.0115, 0.0220, + -0.0344, 0.0447], device='cuda:0'), grad: tensor([-0.0213, 0.0054, 0.0219, 0.0257, 0.0291, 0.0198, -0.0125, -0.0075, + -0.0360, -0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 217.13, cls_loss 0.4977 cls_loss_mapping 0.0033 cls_loss_causal 0.4728 re_mapping 0.0074 re_causal 0.0184 /// teacc 98.77 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.0743, 0.0646, -0.1011, ..., -0.1093, -0.0961, 0.0067], + [-0.0662, -0.1388, 0.0022, ..., -0.0566, -0.0468, -0.0896], + [ 0.0346, -0.0864, 0.0528, ..., 0.1565, -0.1133, -0.0477], + ..., + [-0.0959, -0.1716, 0.0887, ..., -0.0110, -0.0351, 0.0736], + [ 0.0201, 0.0348, -0.0265, ..., -0.0858, -0.0997, -0.0125], + [-0.1488, -0.0060, -0.0088, ..., -0.1472, 0.1093, 0.0186]], + device='cuda:0'), grad: tensor([[ 6.7532e-05, 4.3839e-05, -6.7997e-04, ..., 7.7820e-04, + 2.4033e-04, 6.1941e-04], + [ 8.8155e-05, 7.1406e-05, 6.6900e-04, ..., 4.1932e-05, + 4.6706e-04, 2.1553e-03], + [ 1.4889e-04, 8.7857e-05, 3.8767e-04, ..., 1.4734e-04, + 5.1451e-04, 2.7637e-03], + ..., + [ 4.0698e-04, -2.2907e-03, -1.1833e-02, ..., 1.6034e-04, + -2.2964e-03, 3.1490e-03], + [ 1.0794e-04, -9.8124e-06, 6.5470e-04, ..., 1.2767e-04, + 7.5197e-04, 2.1400e-03], + [ 3.1471e-04, 2.6474e-03, 1.3489e-02, ..., 3.6389e-05, + 4.2076e-03, -7.6790e-03]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0148, 0.0032, 0.0011, 0.0167, -0.0028, -0.0075, 0.0114, 0.0227, + -0.0350, 0.0447], device='cuda:0'), grad: tensor([-0.0124, 0.0202, -0.0108, -0.0383, 0.0227, 0.0124, -0.0237, -0.0321, + 0.0176, 0.0444], device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 216.39, cls_loss 0.5075 cls_loss_mapping 0.0047 cls_loss_causal 0.4874 re_mapping 0.0074 re_causal 0.0189 /// teacc 98.68 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.0747, 0.0641, -0.1015, ..., -0.1103, -0.0982, 0.0060], + [-0.0674, -0.1391, 0.0019, ..., -0.0567, -0.0455, -0.0888], + [ 0.0343, -0.0860, 0.0515, ..., 0.1563, -0.1128, -0.0482], + ..., + [-0.0964, -0.1724, 0.0896, ..., -0.0102, -0.0343, 0.0745], + [ 0.0200, 0.0343, -0.0269, ..., -0.0863, -0.1006, -0.0151], + [-0.1476, -0.0056, -0.0091, ..., -0.1475, 0.1092, 0.0190]], + device='cuda:0'), grad: tensor([[ 2.6631e-04, 4.7207e-04, 4.1246e-04, ..., 2.4021e-04, + 3.5673e-05, 5.4216e-04], + [ 6.5975e-06, 1.0654e-05, 2.0123e-03, ..., 2.7370e-04, + 4.8220e-05, 4.1389e-04], + [ 4.6968e-05, 5.3614e-05, -2.2373e-03, ..., -3.4499e-04, + 5.6803e-05, -1.7433e-03], + ..., + [ 1.1973e-05, 5.1737e-05, 2.5024e-03, ..., 3.8505e-04, + 6.5613e-04, 1.2350e-03], + [ 4.6462e-05, 1.3828e-04, -2.3842e-03, ..., 2.1422e-04, + 5.4479e-05, 3.7980e-04], + [ 4.0829e-05, 1.0121e-04, -8.9645e-04, ..., 2.5368e-04, + 1.4651e-04, -5.9080e-04]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0143, 0.0037, 0.0007, 0.0162, -0.0037, -0.0063, 0.0119, 0.0240, + -0.0355, 0.0443], device='cuda:0'), grad: tensor([ 0.0229, 0.0433, -0.0333, -0.0100, -0.0085, 0.0179, -0.0071, 0.0090, + -0.0013, -0.0330], device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 216.46, cls_loss 0.5479 cls_loss_mapping 0.0065 cls_loss_causal 0.5216 re_mapping 0.0072 re_causal 0.0177 /// teacc 98.77 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.0742, 0.0641, -0.1027, ..., -0.1106, -0.0982, 0.0069], + [-0.0674, -0.1397, 0.0024, ..., -0.0566, -0.0453, -0.0889], + [ 0.0346, -0.0861, 0.0507, ..., 0.1560, -0.1129, -0.0491], + ..., + [-0.0989, -0.1734, 0.0901, ..., -0.0100, -0.0354, 0.0741], + [ 0.0202, 0.0333, -0.0264, ..., -0.0864, -0.0997, -0.0149], + [-0.1486, -0.0051, -0.0083, ..., -0.1477, 0.1087, 0.0198]], + device='cuda:0'), grad: tensor([[-3.2692e-03, -5.2223e-03, 4.5919e-04, ..., 1.3399e-04, + 2.5138e-05, -3.0804e-04], + [ 2.6774e-04, 1.1116e-04, 1.1520e-03, ..., 6.0225e-04, + 3.7640e-05, 7.8392e-04], + [ 7.3147e-04, 6.9284e-04, -9.7351e-03, ..., -6.1393e-05, + -5.9624e-03, -5.2605e-03], + ..., + [ 3.8314e-04, 2.3055e-04, 6.6996e-04, ..., 1.2197e-03, + 5.1594e-04, 8.8167e-04], + [-4.9925e-04, 5.3138e-05, 2.5921e-03, ..., 2.0492e-04, + 1.2560e-03, 1.8177e-03], + [ 1.0872e-03, 1.1377e-03, 8.5983e-03, ..., 2.1350e-04, + 3.5248e-03, 7.1335e-04]], device='cuda:0') +Epoch 261, bias, value: tensor([ 0.0139, 0.0038, 0.0008, 0.0165, -0.0051, -0.0071, 0.0126, 0.0241, + -0.0355, 0.0452], device='cuda:0'), grad: tensor([-0.0072, -0.0045, -0.0406, -0.0392, 0.0238, 0.0184, -0.0166, 0.0249, + 0.0279, 0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 216.28, cls_loss 0.4879 cls_loss_mapping 0.0056 cls_loss_causal 0.4542 re_mapping 0.0074 re_causal 0.0184 /// teacc 98.62 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.0735, 0.0638, -0.1033, ..., -0.1107, -0.0978, 0.0071], + [-0.0681, -0.1408, 0.0025, ..., -0.0560, -0.0454, -0.0888], + [ 0.0336, -0.0864, 0.0527, ..., 0.1557, -0.1112, -0.0482], + ..., + [-0.1008, -0.1731, 0.0884, ..., -0.0102, -0.0363, 0.0733], + [ 0.0204, 0.0346, -0.0276, ..., -0.0864, -0.0987, -0.0150], + [-0.1476, -0.0063, -0.0080, ..., -0.1486, 0.1073, 0.0199]], + device='cuda:0'), grad: tensor([[-6.0768e-03, -5.3329e-03, 1.0548e-03, ..., 1.8343e-05, + 6.3848e-04, -4.6082e-03], + [ 6.3944e-04, 3.2139e-04, 1.4067e-03, ..., 1.0341e-04, + 1.3065e-03, 1.9026e-03], + [ 1.2875e-03, 8.8644e-04, 1.8969e-03, ..., 1.2779e-04, + 8.4019e-04, 2.0123e-03], + ..., + [ 9.9468e-04, 1.0233e-03, 9.3079e-04, ..., 1.6785e-04, + -6.6710e-04, -1.1230e-04], + [-4.2820e-04, -5.2643e-03, -1.0002e-02, ..., 5.1588e-05, + -8.3313e-03, -5.2490e-03], + [ 9.8896e-04, 4.2877e-03, 7.5417e-03, ..., 3.7402e-05, + 5.3062e-03, 3.6449e-03]], device='cuda:0') +Epoch 262, bias, value: tensor([ 0.0142, 0.0035, 0.0011, 0.0174, -0.0049, -0.0069, 0.0119, 0.0237, + -0.0351, 0.0445], device='cuda:0'), grad: tensor([ 0.0004, 0.0056, 0.0316, -0.0092, -0.0364, 0.0140, -0.0074, -0.0038, + -0.0373, 0.0425], device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 216.55, cls_loss 0.5148 cls_loss_mapping 0.0052 cls_loss_causal 0.4904 re_mapping 0.0072 re_causal 0.0173 /// teacc 98.79 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.0741, 0.0646, -0.1031, ..., -0.1106, -0.0974, 0.0067], + [-0.0675, -0.1409, 0.0020, ..., -0.0558, -0.0464, -0.0898], + [ 0.0333, -0.0866, 0.0522, ..., 0.1559, -0.1117, -0.0484], + ..., + [-0.1013, -0.1735, 0.0883, ..., -0.0105, -0.0369, 0.0725], + [ 0.0200, 0.0344, -0.0270, ..., -0.0866, -0.0986, -0.0143], + [-0.1476, -0.0062, -0.0066, ..., -0.1490, 0.1090, 0.0215]], + device='cuda:0'), grad: tensor([[-4.0436e-03, -7.6637e-03, 1.9588e-03, ..., 1.2417e-03, + 2.9445e-04, 6.8855e-04], + [ 4.7088e-04, 4.6015e-05, -3.9368e-03, ..., 2.6703e-05, + -1.1215e-03, -4.2152e-03], + [ 4.3793e-03, 1.1659e-04, 2.2259e-03, ..., 1.8988e-03, + 5.6314e-04, 7.2193e-04], + ..., + [ 1.2159e-03, 6.3837e-05, 3.4676e-03, ..., 9.1219e-04, + 1.2407e-03, 3.0851e-04], + [ 3.4475e-04, 7.6532e-04, -2.6436e-03, ..., 4.6730e-04, + -7.5340e-04, 2.1725e-03], + [ 1.5240e-03, 4.4966e-04, 8.9035e-03, ..., 8.9121e-04, + 3.3760e-03, 3.0174e-03]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0143, 0.0032, 0.0009, 0.0177, -0.0052, -0.0075, 0.0114, 0.0224, + -0.0342, 0.0460], device='cuda:0'), grad: tensor([-0.0276, -0.0662, 0.0104, -0.0353, -0.0374, 0.0241, 0.0485, 0.0334, + -0.0049, 0.0550], device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 217.20, cls_loss 0.4679 cls_loss_mapping 0.0067 cls_loss_causal 0.4404 re_mapping 0.0070 re_causal 0.0169 /// teacc 98.83 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.0743, 0.0648, -0.1032, ..., -0.1112, -0.0983, 0.0068], + [-0.0678, -0.1409, 0.0020, ..., -0.0549, -0.0464, -0.0897], + [ 0.0340, -0.0864, 0.0530, ..., 0.1566, -0.1134, -0.0484], + ..., + [-0.1008, -0.1720, 0.0890, ..., -0.0117, -0.0356, 0.0725], + [ 0.0205, 0.0353, -0.0285, ..., -0.0880, -0.0988, -0.0143], + [-0.1479, -0.0067, -0.0071, ..., -0.1484, 0.1092, 0.0221]], + device='cuda:0'), grad: tensor([[ 7.1704e-05, -3.6788e-04, 3.7551e-04, ..., -9.6202e-05, + 2.1160e-04, 1.8919e-04], + [ 1.9121e-04, 3.4004e-05, 2.5177e-04, ..., 2.1362e-04, + 1.9884e-04, 2.9325e-04], + [-1.9860e-04, -4.2844e-04, -5.9080e-04, ..., -5.7220e-04, + 2.6679e-04, 3.7479e-04], + ..., + [ 2.3317e-04, 2.0370e-05, 1.8188e-02, ..., 1.8430e-04, + 5.8746e-03, 1.6251e-02], + [ 5.2595e-04, 1.5676e-04, 8.6498e-04, ..., 3.2157e-05, + 6.7234e-04, 7.6532e-04], + [-2.5570e-05, -4.0352e-05, -2.6917e-02, ..., 2.3618e-05, + -2.9922e-02, -3.1891e-02]], device='cuda:0') +Epoch 264, bias, value: tensor([ 0.0142, 0.0029, 0.0010, 0.0171, -0.0042, -0.0073, 0.0110, 0.0232, + -0.0344, 0.0457], device='cuda:0'), grad: tensor([-0.0092, -0.0053, -0.0126, -0.0355, 0.0107, -0.0094, -0.0076, 0.0600, + 0.0253, -0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 217.17, cls_loss 0.4710 cls_loss_mapping 0.0038 cls_loss_causal 0.4488 re_mapping 0.0073 re_causal 0.0183 /// teacc 98.70 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.0742, 0.0643, -0.1031, ..., -0.1096, -0.0984, 0.0068], + [-0.0688, -0.1418, 0.0011, ..., -0.0554, -0.0459, -0.0900], + [ 0.0351, -0.0872, 0.0524, ..., 0.1554, -0.1137, -0.0478], + ..., + [-0.1004, -0.1722, 0.0892, ..., -0.0112, -0.0357, 0.0719], + [ 0.0204, 0.0358, -0.0285, ..., -0.0864, -0.0989, -0.0142], + [-0.1486, -0.0073, -0.0072, ..., -0.1487, 0.1089, 0.0215]], + device='cuda:0'), grad: tensor([[ 7.4005e-03, 1.6373e-02, -1.8101e-03, ..., 1.0461e-05, + -2.9964e-03, -1.4305e-03], + [ 4.2534e-04, 1.0723e-04, -4.1270e-04, ..., 1.3714e-03, + 1.1528e-04, -1.4133e-03], + [-6.6185e-03, 2.8396e-04, -1.1435e-03, ..., 5.6289e-06, + 1.3423e-04, 1.0118e-03], + ..., + [ 1.3390e-03, 5.1594e-04, 3.0632e-03, ..., 1.0785e-06, + 1.2255e-03, 1.4019e-03], + [-6.2408e-03, -1.8524e-02, 1.5669e-03, ..., 6.7689e-06, + 8.0824e-04, 1.0729e-03], + [-8.1968e-04, 8.8310e-04, 2.4700e-03, ..., 3.7160e-07, + 8.6594e-04, 1.0424e-03]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0153, 0.0030, 0.0016, 0.0175, -0.0038, -0.0076, 0.0106, 0.0224, + -0.0350, 0.0454], device='cuda:0'), grad: tensor([ 0.0079, -0.0084, -0.0417, 0.0172, -0.0375, 0.0005, 0.0071, 0.0298, + 0.0006, 0.0244], device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 216.65, cls_loss 0.5110 cls_loss_mapping 0.0045 cls_loss_causal 0.4896 re_mapping 0.0072 re_causal 0.0183 /// teacc 98.74 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.0727, 0.0643, -0.1012, ..., -0.1086, -0.0966, 0.0076], + [-0.0691, -0.1418, 0.0003, ..., -0.0560, -0.0465, -0.0901], + [ 0.0348, -0.0865, 0.0531, ..., 0.1554, -0.1134, -0.0482], + ..., + [-0.1012, -0.1726, 0.0891, ..., -0.0127, -0.0362, 0.0718], + [ 0.0214, 0.0382, -0.0291, ..., -0.0861, -0.0999, -0.0158], + [-0.1491, -0.0083, -0.0078, ..., -0.1481, 0.1095, 0.0217]], + device='cuda:0'), grad: tensor([[-3.0537e-03, -2.0580e-03, -1.6136e-03, ..., -8.4400e-04, + -1.4801e-03, -2.3308e-03], + [ 4.1628e-04, 3.2926e-04, 1.0252e-03, ..., 1.6749e-04, + 3.8075e-04, 9.4461e-04], + [-8.4839e-03, 4.8208e-04, 3.6645e-04, ..., -6.2418e-04, + 2.8300e-04, 9.0933e-04], + ..., + [ 7.1192e-04, -6.7949e-04, -1.0284e-02, ..., 1.8287e-04, + -1.4467e-03, -4.7379e-03], + [ 3.3379e-03, -4.2605e-04, 2.5392e-04, ..., 5.2929e-04, + 8.7261e-05, -3.0565e-04], + [-1.3649e-04, 4.0627e-04, 4.0932e-03, ..., 8.0967e-04, + 1.3027e-03, 2.1114e-03]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0168, 0.0027, 0.0016, 0.0169, -0.0042, -0.0067, 0.0100, 0.0222, + -0.0350, 0.0453], device='cuda:0'), grad: tensor([-0.0457, 0.0179, -0.0096, 0.0203, 0.0172, 0.0293, 0.0156, -0.0386, + -0.0019, -0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 216.87, cls_loss 0.4993 cls_loss_mapping 0.0048 cls_loss_causal 0.4708 re_mapping 0.0072 re_causal 0.0179 /// teacc 98.83 lr 0.00010000 +Epoch 267, weight, value: tensor([[-7.2752e-02, 6.5027e-02, -1.0135e-01, ..., -1.0890e-01, + -9.7826e-02, 7.2279e-03], + [-7.0749e-02, -1.4266e-01, -1.5249e-04, ..., -5.6530e-02, + -4.6439e-02, -8.9667e-02], + [ 3.6627e-02, -8.6513e-02, 5.3497e-02, ..., 1.5643e-01, + -1.1317e-01, -4.7342e-02], + ..., + [-1.0193e-01, -1.7195e-01, 8.8295e-02, ..., -1.3229e-02, + -3.6909e-02, 7.1158e-02], + [ 2.0295e-02, 3.7046e-02, -2.7305e-02, ..., -8.5798e-02, + -1.0029e-01, -1.6967e-02], + [-1.4924e-01, -7.8893e-03, -7.2194e-03, ..., -1.4763e-01, + 1.0989e-01, 2.2084e-02]], device='cuda:0'), grad: tensor([[ 2.8191e-03, 1.2177e-02, 2.8348e-04, ..., 2.9616e-07, + 5.1498e-04, 2.5511e-04], + [-3.2215e-03, -2.0504e-03, 3.5048e-04, ..., 1.7369e-07, + 1.1444e-04, -4.0054e-03], + [ 3.3283e-04, 9.7322e-04, 3.3259e-04, ..., -3.3915e-05, + 3.1805e-04, 2.8133e-04], + ..., + [ 1.8132e-04, 8.0287e-05, 2.2831e-03, ..., 3.1328e-04, + 1.4610e-03, 1.0595e-03], + [ 2.6474e-03, 1.9569e-03, 2.6941e-04, ..., 1.6084e-06, + 6.8474e-04, 3.5477e-03], + [ 2.6464e-04, 3.6764e-04, -2.1877e-03, ..., -3.3832e-04, + -1.3819e-03, -3.9792e-04]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0166, 0.0030, 0.0021, 0.0168, -0.0049, -0.0069, 0.0104, 0.0221, + -0.0346, 0.0450], device='cuda:0'), grad: tensor([ 0.0070, 0.0055, 0.0177, -0.0339, -0.0097, -0.0148, 0.0188, -0.0132, + 0.0388, -0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 216.73, cls_loss 0.5264 cls_loss_mapping 0.0040 cls_loss_causal 0.4996 re_mapping 0.0070 re_causal 0.0178 /// teacc 98.83 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.0727, 0.0660, -0.1007, ..., -0.1086, -0.0971, 0.0072], + [-0.0711, -0.1436, -0.0011, ..., -0.0558, -0.0480, -0.0906], + [ 0.0354, -0.0888, 0.0531, ..., 0.1562, -0.1135, -0.0482], + ..., + [-0.1019, -0.1716, 0.0887, ..., -0.0132, -0.0373, 0.0716], + [ 0.0219, 0.0368, -0.0280, ..., -0.0858, -0.1008, -0.0174], + [-0.1494, -0.0078, -0.0070, ..., -0.1479, 0.1105, 0.0224]], + device='cuda:0'), grad: tensor([[ 3.7842e-03, 6.9466e-03, 9.3746e-04, ..., 1.7195e-03, + 9.0647e-04, 1.0977e-03], + [ 7.4089e-05, -4.1747e-04, -4.6778e-04, ..., 3.2276e-05, + -1.1654e-03, -5.3549e-04], + [-1.1040e-02, -1.0254e-02, 1.9989e-03, ..., 6.3210e-03, + -4.1809e-03, 1.9369e-03], + ..., + [ 1.3590e-03, 8.7738e-04, -3.2997e-03, ..., 3.8087e-05, + 4.4155e-04, -3.0804e-03], + [ 2.9678e-03, 2.4662e-03, 5.4359e-04, ..., 3.4547e-04, + 1.2341e-03, 8.4543e-04], + [ 2.6417e-03, 2.2774e-03, 5.1918e-03, ..., 4.8429e-05, + 1.8682e-03, 4.9324e-03]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0160, 0.0033, 0.0016, 0.0162, -0.0049, -0.0067, 0.0110, 0.0231, + -0.0362, 0.0460], device='cuda:0'), grad: tensor([ 0.0334, -0.0307, 0.0079, -0.0316, -0.0075, -0.0094, -0.0590, 0.0396, + 0.0215, 0.0357], device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 217.06, cls_loss 0.4872 cls_loss_mapping 0.0039 cls_loss_causal 0.4549 re_mapping 0.0071 re_causal 0.0173 /// teacc 98.82 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.0735, 0.0662, -0.1007, ..., -0.1088, -0.0966, 0.0081], + [-0.0720, -0.1440, -0.0008, ..., -0.0559, -0.0473, -0.0906], + [ 0.0363, -0.0882, 0.0537, ..., 0.1570, -0.1130, -0.0486], + ..., + [-0.1015, -0.1736, 0.0881, ..., -0.0131, -0.0371, 0.0728], + [ 0.0230, 0.0361, -0.0283, ..., -0.0852, -0.1014, -0.0171], + [-0.1488, -0.0076, -0.0065, ..., -0.1490, 0.1100, 0.0217]], + device='cuda:0'), grad: tensor([[ 1.0742e-02, 1.5930e-02, 1.1110e-03, ..., -2.3282e-04, + 4.9877e-04, -8.8549e-04], + [-4.9639e-04, 2.4962e-04, 6.0940e-04, ..., 4.5029e-07, + -3.0441e-03, -2.6073e-03], + [ 1.1435e-03, 1.9951e-03, 1.0128e-03, ..., -5.5879e-05, + 6.6328e-04, 1.7643e-03], + ..., + [ 7.6866e-04, 1.3018e-03, -4.4975e-03, ..., 2.5146e-06, + 6.7234e-04, -3.3607e-03], + [ 2.0695e-03, 1.3626e-04, -5.0783e-04, ..., 1.1766e-04, + 3.0947e-04, 9.8705e-04], + [ 1.9445e-03, 3.2024e-03, 2.5444e-03, ..., 1.1874e-06, + 1.8559e-03, 4.0054e-03]], device='cuda:0') +Epoch 269, bias, value: tensor([ 0.0161, 0.0034, 0.0014, 0.0166, -0.0043, -0.0075, 0.0101, 0.0225, + -0.0356, 0.0465], device='cuda:0'), grad: tensor([-0.0027, -0.0030, 0.0202, -0.0003, 0.0200, -0.0345, 0.0102, -0.0398, + -0.0059, 0.0357], device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 216.75, cls_loss 0.4834 cls_loss_mapping 0.0036 cls_loss_causal 0.4585 re_mapping 0.0074 re_causal 0.0194 /// teacc 98.92 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.0743, 0.0652, -0.0995, ..., -0.1080, -0.0977, 0.0077], + [-0.0721, -0.1448, -0.0003, ..., -0.0556, -0.0470, -0.0916], + [ 0.0365, -0.0881, 0.0533, ..., 0.1564, -0.1134, -0.0486], + ..., + [-0.1014, -0.1735, 0.0879, ..., -0.0130, -0.0371, 0.0721], + [ 0.0233, 0.0367, -0.0288, ..., -0.0850, -0.1019, -0.0166], + [-0.1487, -0.0067, -0.0058, ..., -0.1505, 0.1108, 0.0223]], + device='cuda:0'), grad: tensor([[-1.0635e-02, -1.4000e-02, -3.6621e-03, ..., -2.3346e-03, + -7.4434e-04, -1.4477e-03], + [ 6.8069e-05, 2.4390e-04, 1.8001e-04, ..., 2.5511e-05, + -5.3436e-05, 4.9305e-04], + [ 2.9049e-03, 2.7370e-03, 2.7103e-03, ..., 1.3456e-03, + 1.3566e-04, 1.2445e-03], + ..., + [ 3.1567e-03, 5.6791e-04, 5.5923e-03, ..., 8.4639e-04, + 7.0190e-04, 2.0828e-03], + [-8.8043e-03, -5.4550e-03, -7.3204e-03, ..., -4.3750e-04, + -3.6865e-05, -1.6479e-03], + [ 3.2687e-04, 7.4816e-04, -4.8685e-04, ..., 6.6662e-04, + -6.9523e-04, -4.4274e-04]], device='cuda:0') +Epoch 270, bias, value: tensor([ 0.0164, 0.0034, 0.0016, 0.0166, -0.0038, -0.0074, 0.0091, 0.0226, + -0.0348, 0.0456], device='cuda:0'), grad: tensor([-0.0361, 0.0085, 0.0254, 0.0629, 0.0124, -0.0626, 0.0089, 0.0285, + -0.0584, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 217.01, cls_loss 0.4902 cls_loss_mapping 0.0037 cls_loss_causal 0.4628 re_mapping 0.0068 re_causal 0.0179 /// teacc 98.84 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.0741, 0.0661, -0.0995, ..., -0.1081, -0.0978, 0.0084], + [-0.0723, -0.1453, -0.0011, ..., -0.0557, -0.0471, -0.0924], + [ 0.0360, -0.0892, 0.0545, ..., 0.1574, -0.1125, -0.0491], + ..., + [-0.1014, -0.1721, 0.0878, ..., -0.0123, -0.0378, 0.0731], + [ 0.0224, 0.0366, -0.0298, ..., -0.0856, -0.1031, -0.0178], + [-0.1479, -0.0083, -0.0064, ..., -0.1518, 0.1103, 0.0214]], + device='cuda:0'), grad: tensor([[-2.2240e-03, -3.9101e-03, 6.4898e-04, ..., 1.1587e-04, + 1.3506e-04, 5.3787e-04], + [ 1.1218e-04, -3.8791e-04, 4.3321e-04, ..., 1.2314e-04, + 7.4387e-05, 5.5647e-04], + [ 1.6890e-03, 1.8559e-03, -1.6451e-05, ..., 1.3733e-03, + 1.5154e-03, -1.8835e-03], + ..., + [ 6.3181e-04, 5.8603e-04, -5.0116e-04, ..., 3.4785e-04, + 1.4770e-04, 8.1837e-05], + [-9.4299e-03, -6.5002e-03, -2.9469e-03, ..., -2.9163e-03, + -1.2655e-03, -2.4281e-03], + [ 7.6675e-04, 5.0974e-04, 2.3499e-03, ..., 4.1556e-04, + -2.1000e-03, 1.5888e-03]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0152, 0.0028, 0.0014, 0.0162, -0.0035, -0.0065, 0.0108, 0.0232, + -0.0342, 0.0440], device='cuda:0'), grad: tensor([ 0.0013, -0.0217, -0.0176, -0.0031, 0.0083, 0.0204, 0.0185, 0.0118, + -0.0228, 0.0049], device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 216.61, cls_loss 0.5168 cls_loss_mapping 0.0030 cls_loss_causal 0.4858 re_mapping 0.0070 re_causal 0.0179 /// teacc 98.64 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.0745, 0.0659, -0.1000, ..., -0.1096, -0.0966, 0.0084], + [-0.0714, -0.1452, -0.0010, ..., -0.0551, -0.0455, -0.0921], + [ 0.0359, -0.0887, 0.0546, ..., 0.1583, -0.1126, -0.0485], + ..., + [-0.1014, -0.1726, 0.0881, ..., -0.0125, -0.0381, 0.0729], + [ 0.0230, 0.0369, -0.0303, ..., -0.0849, -0.1038, -0.0184], + [-0.1474, -0.0085, -0.0061, ..., -0.1524, 0.1101, 0.0214]], + device='cuda:0'), grad: tensor([[-2.8000e-03, -1.2192e-02, 8.2970e-04, ..., -1.8234e-03, + 5.1355e-04, 1.5116e-03], + [ 6.6996e-05, 2.1017e-04, -2.2278e-03, ..., -7.1526e-04, + -2.5826e-03, -4.4250e-03], + [ 5.2166e-04, 1.5461e-04, 9.9792e-03, ..., 3.7155e-03, + 5.7173e-04, 2.7046e-03], + ..., + [-4.6425e-03, 4.5359e-05, -1.9485e-02, ..., -2.8629e-03, + -3.0785e-03, -1.0010e-02], + [ 7.6294e-04, 1.1845e-03, 1.6079e-03, ..., 4.2605e-04, + 8.2636e-04, 1.5984e-03], + [ 7.8154e-04, 4.0770e-04, 2.3880e-03, ..., 4.5037e-04, + 1.2274e-03, 3.7289e-03]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0141, 0.0035, 0.0007, 0.0157, -0.0033, -0.0058, 0.0114, 0.0229, + -0.0350, 0.0450], device='cuda:0'), grad: tensor([-0.0069, -0.0352, 0.0342, -0.0121, 0.0445, -0.0013, 0.0321, -0.0768, + -0.0078, 0.0292], device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 216.98, cls_loss 0.5328 cls_loss_mapping 0.0046 cls_loss_causal 0.5130 re_mapping 0.0070 re_causal 0.0189 /// teacc 98.73 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.0748, 0.0670, -0.1012, ..., -0.1097, -0.0965, 0.0073], + [-0.0721, -0.1442, -0.0011, ..., -0.0563, -0.0445, -0.0926], + [ 0.0354, -0.0893, 0.0543, ..., 0.1583, -0.1133, -0.0480], + ..., + [-0.1009, -0.1736, 0.0883, ..., -0.0126, -0.0397, 0.0729], + [ 0.0234, 0.0361, -0.0311, ..., -0.0860, -0.1044, -0.0177], + [-0.1473, -0.0088, -0.0061, ..., -0.1522, 0.1099, 0.0213]], + device='cuda:0'), grad: tensor([[ 6.7234e-04, 7.2098e-04, 7.9215e-05, ..., 4.6613e-07, + 3.0637e-05, 1.4887e-03], + [ 1.1367e-04, 7.3910e-05, 2.4959e-05, ..., 1.6196e-06, + 8.1211e-06, 5.6000e-03], + [-2.0920e-02, -3.8025e-02, 1.9705e-04, ..., -4.6287e-07, + 4.7952e-05, 1.4954e-03], + ..., + [ 6.2799e-04, 3.0088e-04, 5.3978e-04, ..., -4.1164e-06, + 9.9182e-05, -8.8501e-03], + [ 1.0042e-03, -1.0979e-02, 3.9411e-04, ..., 4.3176e-06, + 8.3566e-05, 2.3689e-03], + [ 3.8910e-04, 2.6798e-04, -6.3241e-05, ..., 1.9390e-06, + -1.3418e-03, 9.0265e-04]], device='cuda:0') +Epoch 273, bias, value: tensor([ 0.0143, 0.0035, -0.0003, 0.0158, -0.0035, -0.0052, 0.0107, 0.0226, + -0.0343, 0.0457], device='cuda:0'), grad: tensor([ 0.0122, 0.0225, -0.0183, -0.0050, 0.0134, -0.0405, 0.0701, -0.0156, + -0.0421, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 216.94, cls_loss 0.5042 cls_loss_mapping 0.0042 cls_loss_causal 0.4810 re_mapping 0.0071 re_causal 0.0186 /// teacc 98.86 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.0748, 0.0668, -0.1012, ..., -0.1096, -0.0955, 0.0080], + [-0.0715, -0.1443, -0.0003, ..., -0.0553, -0.0447, -0.0929], + [ 0.0356, -0.0885, 0.0541, ..., 0.1579, -0.1124, -0.0482], + ..., + [-0.1010, -0.1743, 0.0893, ..., -0.0126, -0.0406, 0.0726], + [ 0.0237, 0.0367, -0.0319, ..., -0.0865, -0.1057, -0.0194], + [-0.1480, -0.0098, -0.0064, ..., -0.1531, 0.1117, 0.0223]], + device='cuda:0'), grad: tensor([[ 3.3951e-04, 5.8031e-04, 5.1498e-04, ..., 2.5868e-04, + 8.4782e-04, 2.0065e-03], + [ 6.2644e-05, -3.5954e-04, -2.3880e-03, ..., 9.0718e-05, + -2.6870e-04, 3.8552e-04], + [ 3.6508e-05, 1.8990e-04, 1.5628e-04, ..., -9.2888e-04, + 1.2964e-05, 9.8038e-04], + ..., + [ 2.3335e-05, 3.0351e-04, 1.2598e-03, ..., 3.8576e-04, + 7.6723e-04, -3.3531e-03], + [ 3.9744e-04, 6.7043e-04, 4.4942e-04, ..., 5.1826e-05, + 9.7513e-04, 2.0714e-03], + [-1.5092e-04, -1.2484e-03, -3.8290e-04, ..., 1.0614e-03, + -1.7405e-03, -4.9515e-03]], device='cuda:0') +Epoch 274, bias, value: tensor([ 0.0154, 0.0039, -0.0006, 0.0157, -0.0040, -0.0043, 0.0091, 0.0228, + -0.0344, 0.0457], device='cuda:0'), grad: tensor([ 0.0211, -0.0161, 0.0112, 0.0276, -0.0202, 0.0004, 0.0195, -0.0443, + 0.0155, -0.0147], device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 217.14, cls_loss 0.4833 cls_loss_mapping 0.0034 cls_loss_causal 0.4624 re_mapping 0.0065 re_causal 0.0165 /// teacc 98.84 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.0758, 0.0665, -0.1026, ..., -0.1114, -0.0965, 0.0071], + [-0.0710, -0.1428, -0.0004, ..., -0.0558, -0.0444, -0.0935], + [ 0.0356, -0.0882, 0.0539, ..., 0.1576, -0.1134, -0.0491], + ..., + [-0.1012, -0.1764, 0.0899, ..., -0.0125, -0.0407, 0.0722], + [ 0.0232, 0.0367, -0.0311, ..., -0.0862, -0.1050, -0.0191], + [-0.1475, -0.0091, -0.0068, ..., -0.1541, 0.1120, 0.0220]], + device='cuda:0'), grad: tensor([[-0.0014, -0.0035, -0.0025, ..., -0.0030, -0.0018, -0.0003], + [ 0.0003, 0.0004, 0.0003, ..., 0.0002, 0.0003, 0.0006], + [-0.0061, 0.0007, 0.0002, ..., -0.0038, 0.0003, 0.0002], + ..., + [ 0.0001, 0.0002, 0.0003, ..., 0.0002, 0.0004, -0.0001], + [ 0.0049, 0.0018, 0.0017, ..., 0.0034, 0.0016, 0.0014], + [ 0.0002, -0.0003, -0.0115, ..., 0.0004, -0.0131, -0.0149]], + device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0145, 0.0036, -0.0007, 0.0154, -0.0034, -0.0051, 0.0095, 0.0232, + -0.0329, 0.0450], device='cuda:0'), grad: tensor([-0.0088, 0.0157, -0.0325, 0.0167, 0.0319, 0.0117, -0.0251, 0.0119, + 0.0297, -0.0512], device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 216.66, cls_loss 0.5281 cls_loss_mapping 0.0042 cls_loss_causal 0.5011 re_mapping 0.0066 re_causal 0.0169 /// teacc 98.85 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.0764, 0.0668, -0.1020, ..., -0.1115, -0.0965, 0.0076], + [-0.0706, -0.1436, -0.0009, ..., -0.0560, -0.0438, -0.0936], + [ 0.0355, -0.0880, 0.0532, ..., 0.1559, -0.1141, -0.0485], + ..., + [-0.1018, -0.1783, 0.0907, ..., -0.0124, -0.0412, 0.0725], + [ 0.0227, 0.0354, -0.0305, ..., -0.0841, -0.1053, -0.0181], + [-0.1478, -0.0090, -0.0078, ..., -0.1562, 0.1116, 0.0217]], + device='cuda:0'), grad: tensor([[ 3.1352e-05, 6.3848e-04, 1.6079e-03, ..., 2.0397e-04, + 8.8692e-04, 1.7204e-03], + [ 2.6017e-05, 1.6069e-04, 9.3460e-04, ..., 1.8430e-04, + 9.0694e-04, 1.8559e-03], + [-7.5035e-03, -5.1041e-03, 3.4428e-03, ..., -6.9084e-03, + 6.7043e-04, -2.2354e-03], + ..., + [ 2.4915e-04, 1.0357e-03, 1.2598e-03, ..., 7.5579e-04, + 2.2590e-04, 1.1845e-03], + [ 9.9487e-03, 1.1070e-02, 6.0005e-03, ..., 5.5504e-03, + 1.1797e-03, 4.5967e-03], + [ 2.9668e-05, -8.3923e-04, -3.6240e-03, ..., 5.2357e-04, + -4.7150e-03, -7.6256e-03]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0142, 0.0038, -0.0006, 0.0155, -0.0029, -0.0037, 0.0092, 0.0229, + -0.0335, 0.0443], device='cuda:0'), grad: tensor([ 0.0173, 0.0213, 0.0040, -0.0253, 0.0270, -0.0331, -0.0197, -0.0341, + 0.0484, -0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 216.71, cls_loss 0.5221 cls_loss_mapping 0.0046 cls_loss_causal 0.5008 re_mapping 0.0066 re_causal 0.0172 /// teacc 98.85 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.0772, 0.0671, -0.1027, ..., -0.1120, -0.0981, 0.0058], + [-0.0699, -0.1445, -0.0007, ..., -0.0566, -0.0442, -0.0937], + [ 0.0352, -0.0882, 0.0543, ..., 0.1569, -0.1146, -0.0481], + ..., + [-0.1007, -0.1773, 0.0910, ..., -0.0121, -0.0399, 0.0740], + [ 0.0224, 0.0352, -0.0301, ..., -0.0847, -0.1049, -0.0180], + [-0.1490, -0.0097, -0.0094, ..., -0.1570, 0.1109, 0.0204]], + device='cuda:0'), grad: tensor([[ 1.0252e-03, 1.0176e-03, 3.6120e-04, ..., 7.4029e-05, + 5.1832e-04, 1.3399e-03], + [ 5.1260e-05, 1.4019e-04, 7.1478e-04, ..., 2.3972e-06, + 1.5841e-03, 3.7308e-03], + [ 7.0610e-03, 4.7455e-03, 4.0030e-04, ..., 3.3617e-04, + -3.7432e-04, -1.2617e-03], + ..., + [ 4.5896e-05, 5.3072e-04, 1.9207e-03, ..., 6.8367e-05, + 2.5387e-03, 4.2686e-03], + [ 3.1624e-03, 1.6470e-03, -2.1591e-02, ..., -1.4648e-02, + -4.9820e-03, 6.5155e-03], + [ 2.4343e-04, -3.0518e-04, 1.8341e-02, ..., 1.4305e-02, + 5.1842e-03, -9.6359e-03]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0130, 0.0038, -0.0017, 0.0163, -0.0030, -0.0049, 0.0121, 0.0236, + -0.0339, 0.0439], device='cuda:0'), grad: tensor([ 0.0176, 0.0258, -0.0021, -0.0269, -0.0029, 0.0221, -0.0151, 0.0107, + -0.0021, -0.0270], device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 216.62, cls_loss 0.5046 cls_loss_mapping 0.0035 cls_loss_causal 0.4762 re_mapping 0.0075 re_causal 0.0192 /// teacc 98.90 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.0775, 0.0675, -0.1036, ..., -0.1128, -0.0992, 0.0054], + [-0.0693, -0.1442, -0.0009, ..., -0.0577, -0.0449, -0.0937], + [ 0.0359, -0.0879, 0.0540, ..., 0.1567, -0.1144, -0.0471], + ..., + [-0.1016, -0.1784, 0.0908, ..., -0.0117, -0.0410, 0.0740], + [ 0.0223, 0.0351, -0.0294, ..., -0.0839, -0.1044, -0.0187], + [-0.1507, -0.0095, -0.0087, ..., -0.1577, 0.1129, 0.0218]], + device='cuda:0'), grad: tensor([[ 2.9802e-04, -2.9778e-04, 2.9492e-04, ..., 7.7105e-04, + 8.8215e-05, 7.4768e-04], + [ 3.3212e-04, -1.1015e-04, 5.1403e-04, ..., 1.4629e-03, + 1.2839e-04, 1.2894e-03], + [ 5.8126e-04, 5.7936e-04, 4.2009e-04, ..., 1.2913e-03, + 1.4567e-04, 1.2722e-03], + ..., + [ 1.8406e-04, 2.8896e-04, 1.9276e-04, ..., 2.3520e-04, + 9.9659e-05, 2.5082e-04], + [-2.3193e-03, -1.6975e-03, -2.6250e-04, ..., 1.4830e-04, + 5.6505e-05, -4.8804e-04], + [ 4.9019e-04, 6.0368e-04, 7.4625e-05, ..., 1.4365e-04, + 2.1696e-04, 4.9877e-04]], device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0130, 0.0045, -0.0019, 0.0166, -0.0037, -0.0052, 0.0125, 0.0238, + -0.0348, 0.0445], device='cuda:0'), grad: tensor([-0.0096, -0.0008, 0.0220, -0.0469, 0.0150, 0.0137, 0.0200, 0.0147, + -0.0424, 0.0144], device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 216.46, cls_loss 0.4995 cls_loss_mapping 0.0057 cls_loss_causal 0.4783 re_mapping 0.0069 re_causal 0.0172 /// teacc 98.86 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.0772, 0.0670, -0.1030, ..., -0.1110, -0.0986, 0.0047], + [-0.0685, -0.1434, -0.0012, ..., -0.0573, -0.0449, -0.0947], + [ 0.0365, -0.0885, 0.0536, ..., 0.1573, -0.1150, -0.0479], + ..., + [-0.1032, -0.1793, 0.0910, ..., -0.0126, -0.0411, 0.0739], + [ 0.0224, 0.0361, -0.0291, ..., -0.0841, -0.1049, -0.0169], + [-0.1506, -0.0091, -0.0090, ..., -0.1587, 0.1123, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.5612e-03, 6.9475e-04, 7.6914e-04, ..., 5.5838e-04, + 6.5660e-04, 1.0617e-07], + [ 8.5652e-05, 1.3244e-04, -1.0691e-03, ..., -1.1891e-04, + 2.1684e-04, 7.5960e-04], + [-1.6713e-04, -9.6226e-04, -4.2343e-03, ..., -3.6087e-03, + -2.5749e-03, -7.0419e-03], + ..., + [ 1.1486e-04, 3.3569e-04, 1.7872e-03, ..., 1.0290e-03, + 1.2398e-03, 2.6417e-03], + [-2.5892e-04, -6.4945e-04, 2.9516e-04, ..., 2.9588e-04, + -6.7568e-04, 1.1215e-03], + [ 1.6499e-03, 1.3857e-03, -5.9938e-04, ..., 4.4036e-04, + -6.2990e-04, -8.8310e-04]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0126, 0.0036, -0.0020, 0.0168, -0.0032, -0.0058, 0.0128, 0.0241, + -0.0338, 0.0440], device='cuda:0'), grad: tensor([-0.0117, 0.0064, -0.0636, 0.0155, -0.0209, 0.0149, 0.0195, 0.0275, + 0.0136, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 216.56, cls_loss 0.5167 cls_loss_mapping 0.0050 cls_loss_causal 0.4911 re_mapping 0.0068 re_causal 0.0163 /// teacc 98.77 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.0767, 0.0676, -0.1036, ..., -0.1125, -0.0987, 0.0048], + [-0.0686, -0.1427, -0.0017, ..., -0.0570, -0.0449, -0.0951], + [ 0.0365, -0.0887, 0.0529, ..., 0.1575, -0.1154, -0.0468], + ..., + [-0.1029, -0.1779, 0.0907, ..., -0.0122, -0.0414, 0.0731], + [ 0.0217, 0.0342, -0.0284, ..., -0.0837, -0.1056, -0.0163], + [-0.1509, -0.0083, -0.0082, ..., -0.1578, 0.1126, 0.0223]], + device='cuda:0'), grad: tensor([[ 1.0307e-02, 1.6754e-02, 6.4254e-05, ..., 8.4579e-05, + 2.2292e-05, 6.9439e-05], + [ 5.3436e-05, 6.1810e-05, 4.3440e-04, ..., 8.6737e-04, + 2.1592e-05, 4.2486e-04], + [ 1.7810e-04, 1.2600e-04, -1.3371e-03, ..., -4.2686e-03, + 5.4806e-05, -1.3590e-03], + ..., + [-2.2564e-03, -4.1771e-03, -6.7635e-03, ..., 1.6594e-04, + -7.3204e-03, -5.6725e-03], + [-1.5930e-02, -2.8473e-02, 4.7779e-04, ..., 7.6342e-04, + 7.5579e-04, 5.0259e-04], + [ 2.9240e-03, 5.6419e-03, 6.1760e-03, ..., 5.7983e-04, + 6.7902e-03, 5.4855e-03]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0125, 0.0041, -0.0018, 0.0184, -0.0041, -0.0060, 0.0114, 0.0246, + -0.0340, 0.0438], device='cuda:0'), grad: tensor([ 0.0193, 0.0088, -0.0104, 0.0202, 0.0057, 0.0033, -0.0178, -0.0425, + -0.0438, 0.0573], device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 216.76, cls_loss 0.4807 cls_loss_mapping 0.0058 cls_loss_causal 0.4537 re_mapping 0.0071 re_causal 0.0175 /// teacc 98.74 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.0774, 0.0663, -0.1038, ..., -0.1118, -0.0981, 0.0044], + [-0.0685, -0.1453, -0.0021, ..., -0.0582, -0.0441, -0.0950], + [ 0.0372, -0.0888, 0.0549, ..., 0.1598, -0.1166, -0.0460], + ..., + [-0.1032, -0.1778, 0.0898, ..., -0.0136, -0.0414, 0.0735], + [ 0.0226, 0.0349, -0.0284, ..., -0.0839, -0.1060, -0.0171], + [-0.1505, -0.0082, -0.0082, ..., -0.1587, 0.1122, 0.0224]], + device='cuda:0'), grad: tensor([[ 1.3838e-03, 1.4277e-03, 3.4881e-04, ..., 9.9659e-04, + 5.9186e-07, -3.2306e-04], + [ 1.7554e-05, 2.4170e-05, 5.1403e-04, ..., 1.2510e-05, + 5.0497e-04, 1.0195e-03], + [-4.7874e-03, -3.1719e-03, -2.2469e-03, ..., -3.8414e-03, + 5.3681e-06, -2.3246e-04], + ..., + [ 1.2522e-03, 7.0810e-05, -4.9706e-03, ..., 5.8174e-05, + -5.0125e-03, 1.4534e-03], + [ 1.2465e-03, 1.4305e-03, 3.7932e-04, ..., 8.8644e-04, + 8.9630e-06, 5.2071e-04], + [-1.2999e-03, 1.1873e-04, 1.6527e-03, ..., 2.9549e-05, + 1.6546e-03, -7.5989e-03]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0121, 0.0045, -0.0008, 0.0178, -0.0050, -0.0058, 0.0111, 0.0256, + -0.0342, 0.0438], device='cuda:0'), grad: tensor([-0.0170, 0.0160, -0.0289, -0.0149, 0.0194, 0.0106, 0.0119, 0.0115, + 0.0188, -0.0273], device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 216.47, cls_loss 0.4888 cls_loss_mapping 0.0034 cls_loss_causal 0.4668 re_mapping 0.0071 re_causal 0.0177 /// teacc 98.89 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.0760, 0.0670, -0.1034, ..., -0.1120, -0.0972, 0.0050], + [-0.0686, -0.1457, -0.0016, ..., -0.0585, -0.0445, -0.0966], + [ 0.0366, -0.0893, 0.0555, ..., 0.1609, -0.1172, -0.0457], + ..., + [-0.1040, -0.1794, 0.0904, ..., -0.0144, -0.0404, 0.0741], + [ 0.0223, 0.0356, -0.0296, ..., -0.0846, -0.1050, -0.0182], + [-0.1518, -0.0088, -0.0086, ..., -0.1579, 0.1115, 0.0228]], + device='cuda:0'), grad: tensor([[ 1.3423e-04, -1.4400e-04, 1.4412e-04, ..., 4.1556e-04, + 5.3734e-05, 1.6606e-04], + [ 3.1147e-03, 2.2650e-05, -1.3981e-03, ..., -8.7690e-04, + -6.0987e-04, -5.5170e-04], + [-7.5150e-03, 7.3552e-05, 4.8101e-05, ..., -7.4625e-05, + 1.2147e-04, -3.7231e-03], + ..., + [ 6.4325e-04, 3.5077e-05, 9.3603e-04, ..., 4.6730e-05, + 2.5964e-04, 1.2894e-03], + [ 5.1079e-03, 1.0471e-03, 7.7133e-03, ..., 1.4082e-05, + 3.0861e-03, 1.0094e-02], + [ 3.1543e-04, 1.4949e-04, 1.1024e-03, ..., 2.5779e-06, + 2.7013e-04, 1.4095e-03]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0124, 0.0043, -0.0006, 0.0165, -0.0044, -0.0052, 0.0113, 0.0256, + -0.0342, 0.0434], device='cuda:0'), grad: tensor([-0.0065, -0.0997, -0.0133, 0.0117, 0.0005, -0.0010, 0.0300, 0.0187, + 0.0437, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 281---------------------------------------------------- +epoch 281, time 217.34, cls_loss 0.5126 cls_loss_mapping 0.0039 cls_loss_causal 0.4871 re_mapping 0.0068 re_causal 0.0167 /// teacc 98.96 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.0761, 0.0674, -0.1026, ..., -0.1129, -0.0983, 0.0045], + [-0.0690, -0.1451, -0.0012, ..., -0.0581, -0.0449, -0.0973], + [ 0.0358, -0.0892, 0.0556, ..., 0.1602, -0.1181, -0.0458], + ..., + [-0.1029, -0.1801, 0.0896, ..., -0.0153, -0.0417, 0.0752], + [ 0.0217, 0.0347, -0.0301, ..., -0.0837, -0.1061, -0.0192], + [-0.1512, -0.0092, -0.0084, ..., -0.1576, 0.1121, 0.0225]], + device='cuda:0'), grad: tensor([[ 1.3575e-05, -1.0073e-05, 3.6120e-04, ..., 4.7135e-04, + 8.8096e-05, 5.0402e-04], + [-1.5230e-03, 4.6581e-05, -4.6463e-03, ..., -7.8011e-03, + -3.1567e-03, -7.9956e-03], + [ 4.2319e-04, 1.3638e-04, 2.9812e-03, ..., 4.5166e-03, + 9.7132e-04, 4.7150e-03], + ..., + [ 4.9621e-05, 1.1623e-04, 1.4687e-03, ..., 2.9588e-04, + 2.8439e-03, 3.2043e-04], + [-3.2234e-03, -4.4327e-03, -4.7874e-04, ..., 3.1233e-04, + 6.8235e-04, -3.9339e-04], + [ 4.7159e-04, 7.7426e-05, -1.5907e-03, ..., 2.8563e-04, + -2.4757e-03, 1.4007e-04]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0118, 0.0042, -0.0005, 0.0164, -0.0042, -0.0054, 0.0115, 0.0257, + -0.0344, 0.0440], device='cuda:0'), grad: tensor([ 0.0066, -0.0435, 0.0019, 0.0215, 0.0094, 0.0088, -0.0250, 0.0132, + 0.0024, 0.0047], device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 216.62, cls_loss 0.5215 cls_loss_mapping 0.0045 cls_loss_causal 0.4974 re_mapping 0.0068 re_causal 0.0174 /// teacc 98.84 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.0765, 0.0673, -0.1035, ..., -0.1132, -0.0995, 0.0039], + [-0.0686, -0.1449, -0.0025, ..., -0.0584, -0.0436, -0.0978], + [ 0.0357, -0.0893, 0.0546, ..., 0.1596, -0.1188, -0.0456], + ..., + [-0.1026, -0.1789, 0.0907, ..., -0.0159, -0.0426, 0.0753], + [ 0.0215, 0.0337, -0.0292, ..., -0.0844, -0.1062, -0.0200], + [-0.1508, -0.0098, -0.0082, ..., -0.1558, 0.1124, 0.0238]], + device='cuda:0'), grad: tensor([[ 1.6193e-03, 1.9569e-03, 1.7052e-03, ..., 5.0640e-04, + 4.0412e-04, 2.8973e-03], + [ 4.7760e-03, 2.0117e-05, 1.7948e-03, ..., 8.3148e-05, + 9.4509e-04, 7.4272e-03], + [ 2.8019e-03, 3.5191e-03, 2.9850e-03, ..., 6.6233e-04, + 8.6069e-04, 4.0474e-03], + ..., + [ 1.3137e-04, 1.3244e-04, 3.5210e-03, ..., 1.6034e-04, + 1.8730e-03, 2.5806e-03], + [-4.3831e-03, 9.1553e-05, 1.1988e-03, ..., 3.0279e-04, + 2.0123e-03, 4.0550e-03], + [-6.0654e-03, -7.4844e-03, 2.3079e-04, ..., -1.6994e-03, + 4.4785e-03, 5.7716e-03]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0112, 0.0051, -0.0008, 0.0163, -0.0044, -0.0045, 0.0110, 0.0254, + -0.0346, 0.0443], device='cuda:0'), grad: tensor([ 0.0197, 0.0275, 0.0271, -0.0137, -0.0389, -0.0499, 0.0137, -0.0007, + 0.0130, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 216.72, cls_loss 0.4928 cls_loss_mapping 0.0044 cls_loss_causal 0.4707 re_mapping 0.0067 re_causal 0.0165 /// teacc 98.85 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.0776, 0.0681, -0.1041, ..., -0.1124, -0.0987, 0.0060], + [-0.0697, -0.1455, -0.0019, ..., -0.0600, -0.0428, -0.0978], + [ 0.0346, -0.0892, 0.0556, ..., 0.1607, -0.1197, -0.0461], + ..., + [-0.1020, -0.1793, 0.0899, ..., -0.0160, -0.0423, 0.0738], + [ 0.0219, 0.0339, -0.0298, ..., -0.0840, -0.1063, -0.0199], + [-0.1506, -0.0104, -0.0075, ..., -0.1575, 0.1122, 0.0238]], + device='cuda:0'), grad: tensor([[ 8.3017e-04, 2.4853e-03, 3.0971e-04, ..., 8.8811e-05, + 8.8010e-07, 4.5562e-04], + [ 5.3287e-05, 1.6189e-04, 1.1802e-04, ..., 7.4565e-05, + 8.8010e-07, 3.5405e-04], + [-4.4751e-04, -1.9944e-04, 3.3164e-04, ..., -9.0420e-05, + 1.1986e-06, 1.6479e-03], + ..., + [ 7.8082e-05, 2.0313e-04, 8.5926e-04, ..., 5.6076e-04, + 1.9282e-05, 1.5688e-03], + [-2.9302e-04, -2.2188e-05, -3.4561e-03, ..., -2.0561e-03, + 6.9253e-06, -5.9166e-03], + [ 2.5201e-04, 1.2708e-04, 2.5344e-04, ..., 2.0111e-04, + -3.4392e-05, 4.8494e-04]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0117, 0.0053, -0.0014, 0.0165, -0.0036, -0.0051, 0.0112, 0.0247, + -0.0345, 0.0443], device='cuda:0'), grad: tensor([-0.0377, 0.0106, 0.0101, 0.0113, 0.0073, 0.0089, -0.0085, 0.0128, + -0.0231, 0.0082], device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 216.64, cls_loss 0.5002 cls_loss_mapping 0.0057 cls_loss_causal 0.4755 re_mapping 0.0071 re_causal 0.0174 /// teacc 98.77 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.0782, 0.0680, -0.1050, ..., -0.1129, -0.0979, 0.0059], + [-0.0688, -0.1457, -0.0017, ..., -0.0596, -0.0432, -0.0990], + [ 0.0336, -0.0903, 0.0555, ..., 0.1605, -0.1199, -0.0450], + ..., + [-0.1001, -0.1772, 0.0901, ..., -0.0160, -0.0422, 0.0732], + [ 0.0213, 0.0338, -0.0296, ..., -0.0846, -0.1069, -0.0195], + [-0.1533, -0.0105, -0.0076, ..., -0.1583, 0.1126, 0.0227]], + device='cuda:0'), grad: tensor([[-7.8154e-04, -3.0422e-03, 2.8753e-04, ..., -7.9346e-04, + 1.9944e-04, 3.4380e-04], + [-3.7479e-03, -5.1384e-03, -1.4315e-03, ..., -5.1842e-03, + -6.9923e-03, -8.6451e-04], + [ 8.5545e-04, 9.5320e-04, 9.9277e-04, ..., 1.0166e-03, + 8.8406e-04, 1.7700e-03], + ..., + [ 5.0879e-04, -2.7075e-05, -2.2526e-03, ..., 6.8951e-04, + -1.5011e-03, 3.3927e-04], + [-7.3195e-04, -1.0920e-03, -1.7700e-03, ..., 6.3419e-04, + 4.9019e-04, 2.3365e-03], + [-1.6232e-03, 6.5899e-04, 3.0556e-03, ..., 7.5817e-04, + 2.7847e-03, -3.8376e-03]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0127, 0.0048, -0.0011, 0.0166, -0.0041, -0.0057, 0.0115, 0.0249, + -0.0344, 0.0439], device='cuda:0'), grad: tensor([-0.0031, -0.0256, 0.0157, 0.0033, -0.0284, 0.0186, 0.0208, 0.0023, + -0.0116, 0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 216.36, cls_loss 0.4891 cls_loss_mapping 0.0039 cls_loss_causal 0.4683 re_mapping 0.0066 re_causal 0.0158 /// teacc 98.92 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.0781, 0.0677, -0.1055, ..., -0.1132, -0.0989, 0.0064], + [-0.0669, -0.1441, -0.0019, ..., -0.0592, -0.0425, -0.1000], + [ 0.0351, -0.0907, 0.0563, ..., 0.1613, -0.1209, -0.0452], + ..., + [-0.1021, -0.1803, 0.0904, ..., -0.0161, -0.0430, 0.0726], + [ 0.0224, 0.0347, -0.0296, ..., -0.0840, -0.1067, -0.0194], + [-0.1544, -0.0107, -0.0060, ..., -0.1571, 0.1129, 0.0242]], + device='cuda:0'), grad: tensor([[ 8.7967e-03, 3.8967e-03, 1.3742e-03, ..., 1.1373e-04, + 4.7112e-04, 5.8289e-03], + [ 7.0989e-05, 1.8680e-04, 5.0879e-04, ..., -1.9512e-03, + -3.8695e-04, -2.3193e-03], + [ 3.0589e-04, 7.2956e-04, 1.6088e-03, ..., 1.1444e-03, + 3.3116e-04, 1.7376e-03], + ..., + [ 2.0361e-04, 5.9223e-04, -2.9259e-03, ..., 1.2910e-04, + 6.0606e-04, 2.1420e-03], + [ 1.1826e-03, 3.9787e-03, 1.2264e-03, ..., 1.1206e-04, + 5.2595e-04, 1.4744e-03], + [ 3.6716e-04, 4.8351e-04, -7.7343e-04, ..., 4.7982e-05, + -3.2120e-03, -2.1496e-03]], device='cuda:0') +Epoch 287, bias, value: tensor([ 0.0132, 0.0045, -0.0008, 0.0159, -0.0046, -0.0055, 0.0114, 0.0250, + -0.0349, 0.0449], device='cuda:0'), grad: tensor([-0.0104, -0.0053, 0.0248, -0.0080, 0.0237, -0.0445, 0.0176, -0.0029, + 0.0043, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 286---------------------------------------------------- +epoch 286, time 217.02, cls_loss 0.4944 cls_loss_mapping 0.0034 cls_loss_causal 0.4715 re_mapping 0.0072 re_causal 0.0187 /// teacc 99.00 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.0784, 0.0683, -0.1061, ..., -0.1120, -0.0980, 0.0059], + [-0.0673, -0.1429, -0.0010, ..., -0.0596, -0.0415, -0.0993], + [ 0.0344, -0.0914, 0.0560, ..., 0.1615, -0.1208, -0.0453], + ..., + [-0.1009, -0.1797, 0.0898, ..., -0.0170, -0.0439, 0.0720], + [ 0.0233, 0.0353, -0.0305, ..., -0.0831, -0.1063, -0.0199], + [-0.1545, -0.0108, -0.0060, ..., -0.1577, 0.1134, 0.0244]], + device='cuda:0'), grad: tensor([[ 4.2081e-05, -8.9502e-04, -1.2293e-03, ..., 1.1700e-04, + -4.6492e-04, -1.1406e-03], + [ 2.0027e-05, 1.4329e-04, 5.1022e-04, ..., 1.4627e-04, + 2.6283e-03, 6.0511e-04], + [-9.4473e-05, 1.4555e-04, -6.4468e-04, ..., -6.2895e-04, + 7.3910e-04, 4.0007e-04], + ..., + [-1.3332e-03, 9.0957e-05, -1.2604e-02, ..., -3.2163e-04, + -1.1673e-02, -5.9319e-03], + [ 1.8644e-03, 4.8804e-04, 1.8339e-03, ..., 9.8586e-05, + 9.7809e-03, 7.1001e-04], + [ 2.9236e-05, 2.6536e-04, 1.0582e-02, ..., 1.1581e-04, + -3.0403e-03, 3.1147e-03]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0134, 0.0052, -0.0005, 0.0155, -0.0038, -0.0042, 0.0102, 0.0244, + -0.0359, 0.0449], device='cuda:0'), grad: tensor([-0.0200, 0.0021, 0.0096, 0.0138, -0.0030, 0.0106, 0.0147, -0.0455, + 0.0290, -0.0112], device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 217.06, cls_loss 0.5041 cls_loss_mapping 0.0039 cls_loss_causal 0.4791 re_mapping 0.0067 re_causal 0.0173 /// teacc 98.78 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.0780, 0.0682, -0.1050, ..., -0.1129, -0.0999, 0.0038], + [-0.0683, -0.1432, -0.0009, ..., -0.0584, -0.0434, -0.0987], + [ 0.0350, -0.0910, 0.0554, ..., 0.1620, -0.1208, -0.0455], + ..., + [-0.1017, -0.1800, 0.0908, ..., -0.0170, -0.0447, 0.0716], + [ 0.0218, 0.0343, -0.0301, ..., -0.0820, -0.1050, -0.0187], + [-0.1550, -0.0095, -0.0068, ..., -0.1584, 0.1142, 0.0242]], + device='cuda:0'), grad: tensor([[ 1.1796e-04, 1.5583e-03, 4.8041e-04, ..., 5.5218e-04, + 5.6624e-05, 6.1321e-04], + [ 1.4806e-04, 1.0312e-04, 1.0433e-03, ..., 1.8609e-04, + 9.3102e-05, -4.6682e-04], + [ 1.0738e-03, 1.6994e-03, 2.2755e-03, ..., 2.2984e-03, + 5.6267e-04, 1.3180e-03], + ..., + [ 7.5877e-05, 8.1480e-05, -1.0918e-02, ..., 5.3167e-04, + -8.9502e-04, -2.7370e-03], + [ 1.1861e-04, 1.3561e-03, 3.6507e-03, ..., -1.9379e-03, + 5.1689e-04, -4.2963e-04], + [ 9.4950e-05, 1.7357e-04, 3.7746e-03, ..., 3.9244e-04, + 2.8682e-04, 1.4715e-03]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0137, 0.0041, -0.0014, 0.0161, -0.0047, -0.0045, 0.0115, 0.0249, + -0.0353, 0.0446], device='cuda:0'), grad: tensor([ 0.0269, 0.0011, 0.0405, -0.0533, -0.0080, 0.0263, -0.0080, 0.0085, + -0.0298, -0.0042], device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 216.85, cls_loss 0.4799 cls_loss_mapping 0.0057 cls_loss_causal 0.4531 re_mapping 0.0065 re_causal 0.0158 /// teacc 98.83 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.0764, 0.0689, -0.1060, ..., -0.1117, -0.1006, 0.0051], + [-0.0690, -0.1441, 0.0003, ..., -0.0578, -0.0440, -0.0986], + [ 0.0355, -0.0909, 0.0561, ..., 0.1627, -0.1216, -0.0441], + ..., + [-0.1030, -0.1809, 0.0916, ..., -0.0181, -0.0449, 0.0716], + [ 0.0218, 0.0351, -0.0299, ..., -0.0820, -0.1059, -0.0201], + [-0.1542, -0.0100, -0.0077, ..., -0.1582, 0.1144, 0.0245]], + device='cuda:0'), grad: tensor([[ 5.1451e-04, 2.7466e-04, -1.6940e-04, ..., 1.8433e-05, + 1.7732e-05, -1.3542e-03], + [-1.4763e-03, 5.6207e-05, 2.1708e-04, ..., 3.6880e-06, + -1.8954e-04, -8.1062e-04], + [ 2.8872e-04, 2.8658e-04, 1.2672e-04, ..., -5.9319e-04, + 3.2353e-04, 1.5411e-03], + ..., + [ 2.4068e-04, 9.9540e-05, 3.9625e-04, ..., 6.6459e-05, + 3.2806e-04, 1.5097e-03], + [ 1.4496e-03, 1.4563e-03, 3.0041e-04, ..., 2.0608e-05, + 3.2330e-04, 2.5425e-03], + [ 4.6992e-04, 3.6860e-04, 6.2752e-04, ..., 3.8743e-04, + 3.8385e-04, 1.6336e-03]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0132, 0.0036, -0.0010, 0.0159, -0.0051, -0.0056, 0.0114, 0.0256, + -0.0351, 0.0461], device='cuda:0'), grad: tensor([-0.0094, -0.0084, 0.0181, 0.0249, -0.0260, -0.0182, -0.0257, 0.0238, + -0.0028, 0.0237], device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 216.70, cls_loss 0.4690 cls_loss_mapping 0.0033 cls_loss_causal 0.4494 re_mapping 0.0066 re_causal 0.0165 /// teacc 98.88 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.0773, 0.0685, -0.1058, ..., -0.1121, -0.1004, 0.0056], + [-0.0690, -0.1446, 0.0010, ..., -0.0572, -0.0425, -0.0978], + [ 0.0362, -0.0912, 0.0565, ..., 0.1623, -0.1205, -0.0424], + ..., + [-0.1044, -0.1805, 0.0911, ..., -0.0183, -0.0451, 0.0704], + [ 0.0228, 0.0355, -0.0307, ..., -0.0825, -0.1053, -0.0197], + [-0.1530, -0.0099, -0.0070, ..., -0.1576, 0.1154, 0.0259]], + device='cuda:0'), grad: tensor([[-1.3304e-04, -2.0161e-03, 2.4915e-04, ..., 6.7055e-05, + -5.7220e-04, -1.1387e-03], + [ 3.2991e-05, 3.1805e-04, 3.1686e-04, ..., 7.9393e-05, + 5.1212e-04, 6.9332e-04], + [-3.1185e-04, 2.9826e-04, 2.1350e-04, ..., -2.8782e-03, + 4.3344e-04, 9.4700e-04], + ..., + [ 4.5151e-05, 4.5681e-04, 1.6880e-03, ..., 3.8314e-04, + 1.1921e-03, 1.9436e-03], + [-6.7770e-05, -3.4809e-04, 7.8201e-04, ..., 2.6059e-04, + 9.4509e-04, 1.3742e-03], + [ 6.3479e-05, -3.4547e-04, -3.8624e-04, ..., -1.4651e-04, + -1.5984e-03, -1.2434e-04]], device='cuda:0') +Epoch 291, bias, value: tensor([ 1.2857e-02, 4.1527e-03, -8.1386e-05, 1.5289e-02, -5.0530e-03, + -5.2054e-03, 1.1382e-02, 2.4315e-02, -3.4696e-02, 4.6049e-02], + device='cuda:0'), grad: tensor([-0.0203, 0.0103, 0.0054, -0.0132, -0.0402, 0.0059, 0.0189, 0.0161, + 0.0133, 0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 216.58, cls_loss 0.4807 cls_loss_mapping 0.0025 cls_loss_causal 0.4492 re_mapping 0.0065 re_causal 0.0165 /// teacc 98.90 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.0774, 0.0694, -0.1061, ..., -0.1121, -0.1011, 0.0051], + [-0.0684, -0.1457, -0.0015, ..., -0.0562, -0.0434, -0.0983], + [ 0.0374, -0.0911, 0.0565, ..., 0.1619, -0.1213, -0.0426], + ..., + [-0.1054, -0.1813, 0.0912, ..., -0.0180, -0.0452, 0.0694], + [ 0.0220, 0.0347, -0.0306, ..., -0.0827, -0.1038, -0.0184], + [-0.1534, -0.0100, -0.0067, ..., -0.1568, 0.1151, 0.0263]], + device='cuda:0'), grad: tensor([[ 5.4693e-04, 1.1406e-03, 1.0300e-03, ..., 1.9836e-04, + 8.1348e-04, 9.7847e-04], + [ 2.6250e-04, 1.7548e-04, 8.5115e-04, ..., 2.2209e-04, + 9.5844e-04, 2.5463e-03], + [ 2.4223e-04, 7.1526e-04, -5.5408e-04, ..., -1.2417e-03, + 9.3460e-04, 2.5482e-03], + ..., + [ 3.4595e-04, 4.4894e-04, 1.4572e-03, ..., 2.6393e-04, + 1.1549e-03, 3.1319e-03], + [-3.5191e-03, -4.0970e-03, -1.6813e-03, ..., -1.8299e-04, + -2.0370e-03, -5.2795e-03], + [-7.1108e-05, -1.4296e-03, -8.7452e-04, ..., 1.7333e-04, + -6.4969e-05, -1.3828e-03]], device='cuda:0') +Epoch 292, bias, value: tensor([ 1.4339e-02, 3.6475e-03, 5.2438e-05, 1.4124e-02, -4.8172e-03, + -5.2183e-03, 1.0803e-02, 2.5139e-02, -3.4642e-02, 4.5703e-02], + device='cuda:0'), grad: tensor([-0.0080, 0.0217, 0.0187, 0.0343, -0.0383, 0.0193, -0.0435, 0.0249, + -0.0327, 0.0036], device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 216.70, cls_loss 0.4965 cls_loss_mapping 0.0041 cls_loss_causal 0.4724 re_mapping 0.0065 re_causal 0.0167 /// teacc 98.79 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.0759, 0.0705, -0.1055, ..., -0.1124, -0.1016, 0.0041], + [-0.0693, -0.1459, -0.0030, ..., -0.0557, -0.0446, -0.0986], + [ 0.0373, -0.0922, 0.0566, ..., 0.1613, -0.1211, -0.0418], + ..., + [-0.1040, -0.1815, 0.0922, ..., -0.0168, -0.0452, 0.0702], + [ 0.0218, 0.0343, -0.0308, ..., -0.0832, -0.1039, -0.0184], + [-0.1548, -0.0104, -0.0074, ..., -0.1573, 0.1141, 0.0250]], + device='cuda:0'), grad: tensor([[ 4.0150e-04, -1.2457e-04, -1.6379e-04, ..., -1.9073e-05, + 1.3812e-06, 9.1732e-05], + [-2.2560e-05, -9.9003e-05, 1.2377e-06, ..., 1.1828e-07, + 8.3912e-07, 4.3839e-05], + [ 2.1725e-03, 1.5020e-03, 9.9316e-06, ..., 9.2667e-07, + 1.1465e-06, 4.2796e-04], + ..., + [ 8.1241e-05, 5.7667e-05, 3.8952e-05, ..., 4.2003e-07, + 5.0247e-05, 3.2395e-05], + [-5.0621e-03, -3.2101e-03, 2.3544e-05, ..., 1.5004e-06, + 1.9416e-05, -1.0900e-03], + [ 4.5085e-04, 9.6142e-05, 5.8460e-04, ..., 1.7015e-06, + 9.4700e-04, 6.8903e-04]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0150, 0.0041, -0.0003, 0.0143, -0.0051, -0.0051, 0.0110, 0.0252, + -0.0352, 0.0452], device='cuda:0'), grad: tensor([ 0.0051, 0.0090, 0.0131, -0.0232, 0.0051, 0.0053, 0.0070, -0.0263, + -0.0011, 0.0061], device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 216.54, cls_loss 0.5202 cls_loss_mapping 0.0035 cls_loss_causal 0.4916 re_mapping 0.0063 re_causal 0.0162 /// teacc 98.76 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.0757, 0.0693, -0.1063, ..., -0.1123, -0.1010, 0.0046], + [-0.0708, -0.1459, -0.0010, ..., -0.0550, -0.0425, -0.0985], + [ 0.0373, -0.0931, 0.0570, ..., 0.1608, -0.1225, -0.0425], + ..., + [-0.1048, -0.1818, 0.0909, ..., -0.0175, -0.0468, 0.0706], + [ 0.0216, 0.0340, -0.0318, ..., -0.0839, -0.1037, -0.0195], + [-0.1566, -0.0109, -0.0066, ..., -0.1575, 0.1135, 0.0243]], + device='cuda:0'), grad: tensor([[ 2.6569e-03, 1.0662e-03, 2.1756e-04, ..., 5.8365e-04, + 2.7493e-06, 3.7622e-04], + [ 1.5013e-05, 5.4979e-04, 7.0572e-04, ..., 8.2731e-04, + 1.2445e-04, 6.2561e-04], + [ 5.5122e-04, -2.5964e-04, 1.1864e-03, ..., -2.4509e-04, + 1.2070e-05, -1.0262e-03], + ..., + [ 1.9401e-05, 3.3236e-04, 1.1349e-03, ..., 4.2915e-04, + 2.3532e-04, 2.3985e-04], + [-3.6693e-04, 1.7631e-04, -1.8187e-03, ..., -3.1452e-03, + 6.0272e-04, 4.9210e-04], + [ 1.7628e-05, -1.6582e-04, -4.6349e-03, ..., 2.2471e-04, + -9.4032e-04, -4.8876e-04]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0147, 0.0037, -0.0002, 0.0153, -0.0044, -0.0057, 0.0115, 0.0250, + -0.0368, 0.0461], device='cuda:0'), grad: tensor([-0.0023, 0.0116, -0.0012, 0.0066, 0.0089, 0.0039, 0.0056, -0.0223, + -0.0051, -0.0057], device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 216.33, cls_loss 0.4845 cls_loss_mapping 0.0040 cls_loss_causal 0.4622 re_mapping 0.0066 re_causal 0.0165 /// teacc 98.69 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.0755, 0.0697, -0.1052, ..., -0.1106, -0.1006, 0.0051], + [-0.0695, -0.1470, 0.0002, ..., -0.0551, -0.0419, -0.0992], + [ 0.0376, -0.0932, 0.0571, ..., 0.1616, -0.1234, -0.0420], + ..., + [-0.1061, -0.1815, 0.0909, ..., -0.0171, -0.0462, 0.0707], + [ 0.0224, 0.0348, -0.0322, ..., -0.0836, -0.1030, -0.0193], + [-0.1551, -0.0106, -0.0065, ..., -0.1580, 0.1132, 0.0238]], + device='cuda:0'), grad: tensor([[-1.9012e-02, -2.0203e-02, -8.7357e-04, ..., 1.4687e-04, + 2.2483e-04, 2.7227e-04], + [ 1.1593e-04, 2.2435e-04, 3.0065e-04, ..., 1.7166e-04, + 8.8155e-05, 3.0375e-04], + [ 2.1057e-03, 2.3632e-03, 4.6521e-05, ..., -3.2234e-04, + 9.5546e-05, -2.7823e-04], + ..., + [ 1.9598e-04, 1.9515e-04, 1.1120e-03, ..., 3.5262e-04, + 5.9426e-05, -5.0974e-04], + [ 2.3003e-03, 8.4734e-04, 1.6518e-03, ..., 1.2863e-04, + 6.6876e-05, 1.5652e-04], + [-8.2550e-03, 1.2760e-03, -6.0005e-03, ..., 1.4555e-04, + 1.3685e-04, 2.0587e-04]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0146, 0.0037, -0.0002, 0.0155, -0.0050, -0.0054, 0.0117, 0.0252, + -0.0371, 0.0462], device='cuda:0'), grad: tensor([-0.0119, 0.0177, -0.0190, 0.0317, -0.0173, 0.0382, -0.0091, -0.0163, + 0.0251, -0.0390], device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 216.58, cls_loss 0.5080 cls_loss_mapping 0.0047 cls_loss_causal 0.4826 re_mapping 0.0065 re_causal 0.0155 /// teacc 98.73 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.0744, 0.0709, -0.1039, ..., -0.1113, -0.1006, 0.0048], + [-0.0707, -0.1486, 0.0008, ..., -0.0549, -0.0416, -0.0992], + [ 0.0375, -0.0937, 0.0561, ..., 0.1605, -0.1216, -0.0408], + ..., + [-0.1064, -0.1813, 0.0907, ..., -0.0166, -0.0482, 0.0702], + [ 0.0229, 0.0359, -0.0333, ..., -0.0831, -0.1029, -0.0187], + [-0.1528, -0.0115, -0.0052, ..., -0.1571, 0.1146, 0.0249]], + device='cuda:0'), grad: tensor([[ 4.0197e-04, -2.1839e-04, 8.7070e-04, ..., 3.9220e-04, + 4.3297e-04, 7.5054e-04], + [ 6.1035e-04, 2.3711e-04, -1.7729e-03, ..., 3.6597e-04, + 1.4009e-03, 1.3199e-03], + [-1.3876e-04, 3.2449e-04, -3.8922e-05, ..., -3.1185e-04, + 6.8903e-04, 1.0738e-03], + ..., + [ 2.1800e-05, 1.9038e-04, -1.7920e-03, ..., 1.5926e-04, + -4.8089e-04, -1.7014e-03], + [ 1.1943e-05, 4.7013e-06, 1.2856e-03, ..., 6.0177e-04, + 1.0929e-03, 1.5240e-03], + [ 6.1846e-04, 4.5586e-04, -6.1464e-04, ..., -5.7888e-04, + -3.5858e-03, -4.2229e-03]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0137, 0.0038, -0.0007, 0.0154, -0.0049, -0.0068, 0.0119, 0.0247, + -0.0367, 0.0484], device='cuda:0'), grad: tensor([-0.0169, -0.0006, 0.0112, -0.0232, 0.0182, 0.0108, -0.0062, 0.0040, + 0.0162, -0.0134], device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 216.51, cls_loss 0.5086 cls_loss_mapping 0.0042 cls_loss_causal 0.4757 re_mapping 0.0063 re_causal 0.0150 /// teacc 98.69 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.0722, 0.0711, -0.1047, ..., -0.1117, -0.0996, 0.0061], + [-0.0734, -0.1499, 0.0015, ..., -0.0566, -0.0421, -0.0994], + [ 0.0372, -0.0930, 0.0568, ..., 0.1614, -0.1208, -0.0415], + ..., + [-0.1074, -0.1820, 0.0911, ..., -0.0162, -0.0479, 0.0704], + [ 0.0227, 0.0352, -0.0331, ..., -0.0828, -0.1020, -0.0185], + [-0.1534, -0.0111, -0.0055, ..., -0.1583, 0.1138, 0.0245]], + device='cuda:0'), grad: tensor([[ 2.0351e-03, 1.7710e-03, 2.6798e-03, ..., 3.0785e-03, + 1.3269e-05, 1.6689e-05], + [ 5.7846e-05, 4.2059e-06, 3.4254e-06, ..., 3.1758e-06, + 2.4915e-04, 3.2830e-04], + [ 1.3447e-04, 1.2118e-04, 1.6725e-04, ..., 1.8418e-04, + 1.3061e-05, 1.9684e-05], + ..., + [ 4.9734e-04, 2.0638e-05, -8.3566e-05, ..., 2.5600e-05, + 1.2201e-04, 6.3181e-05], + [-6.8741e-03, -3.9444e-03, -5.4817e-03, ..., -6.3133e-03, + 2.1303e-04, 2.6751e-04], + [-1.7586e-03, 5.9700e-04, 6.6471e-04, ..., 5.6553e-04, + -1.0551e-02, -1.3855e-02]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0130, 0.0031, -0.0002, 0.0151, -0.0053, -0.0065, 0.0129, 0.0254, + -0.0365, 0.0478], device='cuda:0'), grad: tensor([ 0.0177, -0.0169, 0.0072, 0.0092, 0.0302, -0.0229, 0.0168, -0.0101, + -0.0228, -0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 216.19, cls_loss 0.4993 cls_loss_mapping 0.0032 cls_loss_causal 0.4671 re_mapping 0.0068 re_causal 0.0172 /// teacc 98.80 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.0723, 0.0708, -0.1040, ..., -0.1115, -0.0993, 0.0053], + [-0.0757, -0.1507, 0.0022, ..., -0.0562, -0.0440, -0.1003], + [ 0.0382, -0.0932, 0.0564, ..., 0.1614, -0.1215, -0.0416], + ..., + [-0.1058, -0.1823, 0.0910, ..., -0.0171, -0.0478, 0.0708], + [ 0.0232, 0.0355, -0.0331, ..., -0.0834, -0.1021, -0.0196], + [-0.1527, -0.0110, -0.0061, ..., -0.1598, 0.1141, 0.0250]], + device='cuda:0'), grad: tensor([[ 9.1121e-06, 4.8971e-04, 6.1846e-04, ..., 4.2844e-04, + 1.3125e-04, 1.5459e-03], + [ 4.4560e-04, 9.8050e-05, -2.3441e-03, ..., -2.0065e-03, + -6.1989e-04, -5.2490e-03], + [ 1.7529e-03, 9.7990e-05, 2.2583e-03, ..., 1.9684e-03, + 1.2410e-04, 2.5101e-03], + ..., + [-1.7462e-03, 5.9509e-04, -1.2903e-03, ..., -2.4357e-03, + 1.5240e-03, 1.0052e-03], + [-8.8596e-04, 3.2878e-04, -2.7485e-03, ..., 3.4285e-04, + -2.7428e-03, 2.0638e-03], + [ 7.5674e-04, 1.1997e-03, 2.2144e-03, ..., 3.9339e-04, + 1.8997e-03, -3.8509e-03]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0142, 0.0024, -0.0013, 0.0152, -0.0049, -0.0060, 0.0126, 0.0256, + -0.0363, 0.0475], device='cuda:0'), grad: tensor([ 0.0241, -0.0604, 0.0265, 0.0191, -0.0070, 0.0103, -0.0172, -0.0115, + 0.0191, -0.0030], device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 216.73, cls_loss 0.4648 cls_loss_mapping 0.0050 cls_loss_causal 0.4418 re_mapping 0.0066 re_causal 0.0160 /// teacc 98.86 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.0718, 0.0713, -0.1040, ..., -0.1123, -0.0996, 0.0052], + [-0.0763, -0.1512, 0.0023, ..., -0.0562, -0.0441, -0.1003], + [ 0.0385, -0.0930, 0.0561, ..., 0.1607, -0.1219, -0.0407], + ..., + [-0.1033, -0.1820, 0.0905, ..., -0.0175, -0.0462, 0.0709], + [ 0.0230, 0.0363, -0.0322, ..., -0.0829, -0.1025, -0.0204], + [-0.1532, -0.0113, -0.0061, ..., -0.1602, 0.1140, 0.0252]], + device='cuda:0'), grad: tensor([[ 1.0681e-03, -4.0865e-04, 5.2929e-04, ..., 5.5879e-08, + 5.3495e-05, -2.0370e-03], + [ 4.2057e-04, 8.6308e-05, 1.8430e-04, ..., 2.6450e-07, + -6.3610e-04, -4.6883e-03], + [ 9.3460e-04, 1.3483e-04, 2.0838e-04, ..., -1.6103e-06, + 1.5676e-04, 2.0065e-03], + ..., + [ 2.1839e-03, 8.9765e-05, -1.5650e-03, ..., 1.2610e-06, + -4.5514e-04, 7.2432e-04], + [ 1.0529e-02, 9.0790e-04, 2.2602e-04, ..., 2.4168e-07, + 1.4806e-04, 1.1463e-03], + [-3.5095e-03, 1.4234e-04, -1.5169e-05, ..., 4.0699e-07, + 1.0386e-05, -4.5853e-03]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0141, 0.0024, -0.0018, 0.0161, -0.0045, -0.0053, 0.0126, 0.0253, + -0.0365, 0.0467], device='cuda:0'), grad: tensor([-0.0051, -0.0327, 0.0201, 0.0239, 0.0183, -0.0241, -0.0090, 0.0061, + 0.0399, -0.0373], device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 216.78, cls_loss 0.5022 cls_loss_mapping 0.0034 cls_loss_causal 0.4781 re_mapping 0.0069 re_causal 0.0171 /// teacc 98.78 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.0720, 0.0709, -0.1049, ..., -0.1129, -0.1008, 0.0043], + [-0.0769, -0.1527, 0.0018, ..., -0.0561, -0.0448, -0.0995], + [ 0.0381, -0.0937, 0.0561, ..., 0.1605, -0.1233, -0.0422], + ..., + [-0.1035, -0.1830, 0.0903, ..., -0.0180, -0.0470, 0.0717], + [ 0.0233, 0.0362, -0.0322, ..., -0.0816, -0.1021, -0.0203], + [-0.1524, -0.0110, -0.0060, ..., -0.1606, 0.1147, 0.0240]], + device='cuda:0'), grad: tensor([[-8.7619e-05, -2.6073e-03, -3.0441e-03, ..., 1.6558e-04, + 3.2020e-04, 4.1652e-04], + [-6.3753e-04, -1.7798e-04, 2.7585e-04, ..., 2.2388e-04, + 5.7191e-05, 4.5991e-04], + [ 2.8157e-04, -4.2953e-03, -8.9188e-03, ..., -5.3406e-03, + -3.4866e-03, -2.5616e-03], + ..., + [ 2.9135e-04, 5.1355e-04, 7.1859e-04, ..., 8.4400e-04, + 9.7513e-04, 1.4610e-03], + [ 1.4234e-04, 3.7384e-03, 8.1863e-03, ..., 4.2229e-03, + 2.9316e-03, 2.6722e-03], + [ 3.8433e-04, 1.6012e-03, 3.2730e-03, ..., 1.8911e-03, + 2.0065e-03, 2.0638e-03]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0141, 0.0030, -0.0017, 0.0161, -0.0045, -0.0046, 0.0123, 0.0252, + -0.0359, 0.0452], device='cuda:0'), grad: tensor([-0.0154, 0.0050, -0.0099, -0.0138, 0.0106, 0.0154, -0.0083, 0.0170, + -0.0235, 0.0229], device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 216.51, cls_loss 0.4940 cls_loss_mapping 0.0042 cls_loss_causal 0.4711 re_mapping 0.0067 re_causal 0.0168 /// teacc 98.84 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.0720, 0.0714, -0.1054, ..., -0.1132, -0.1006, 0.0049], + [-0.0765, -0.1524, 0.0015, ..., -0.0581, -0.0452, -0.0999], + [ 0.0371, -0.0952, 0.0557, ..., 0.1599, -0.1224, -0.0417], + ..., + [-0.1034, -0.1829, 0.0912, ..., -0.0156, -0.0484, 0.0713], + [ 0.0226, 0.0359, -0.0317, ..., -0.0809, -0.1011, -0.0210], + [-0.1532, -0.0114, -0.0057, ..., -0.1609, 0.1152, 0.0246]], + device='cuda:0'), grad: tensor([[ 1.9703e-03, -8.5878e-04, 6.3598e-05, ..., 5.5695e-04, + 2.3603e-04, 9.2268e-05], + [ 7.1859e-04, 2.0802e-05, 2.2328e-04, ..., 2.2769e-04, + 9.2983e-05, 8.4043e-05], + [-7.8735e-03, 6.7353e-05, 3.4027e-03, ..., 3.4332e-03, + -8.1539e-04, 1.2827e-04], + ..., + [ 3.4809e-04, 7.3195e-05, 3.1643e-03, ..., 3.7384e-03, + 9.8050e-05, 2.0218e-04], + [-4.0550e-03, 1.0836e-04, -1.2390e-02, ..., -1.3565e-02, + -2.3270e-04, 1.0008e-04], + [ 1.9526e-04, 2.7442e-04, 8.8596e-04, ..., 8.7786e-04, + -2.2382e-05, -2.1458e-04]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0140, 0.0029, -0.0014, 0.0157, -0.0055, -0.0048, 0.0119, 0.0257, + -0.0351, 0.0456], device='cuda:0'), grad: tensor([ 0.0227, 0.0253, -0.0547, 0.0232, -0.0260, 0.0401, 0.0198, -0.0204, + -0.0528, 0.0227], device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 217.09, cls_loss 0.5455 cls_loss_mapping 0.0057 cls_loss_causal 0.5205 re_mapping 0.0065 re_causal 0.0163 /// teacc 98.84 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.0717, 0.0713, -0.1053, ..., -0.1143, -0.0985, 0.0055], + [-0.0771, -0.1536, 0.0024, ..., -0.0565, -0.0453, -0.0993], + [ 0.0391, -0.0944, 0.0567, ..., 0.1601, -0.1239, -0.0424], + ..., + [-0.1031, -0.1839, 0.0904, ..., -0.0166, -0.0473, 0.0724], + [ 0.0224, 0.0360, -0.0304, ..., -0.0796, -0.1016, -0.0206], + [-0.1545, -0.0117, -0.0052, ..., -0.1619, 0.1154, 0.0253]], + device='cuda:0'), grad: tensor([[ 7.8659e-03, 8.1635e-03, 2.9588e-04, ..., 2.2605e-05, + 1.8513e-04, 6.7406e-03], + [-1.0462e-03, 1.1539e-04, 2.3985e-04, ..., 2.0489e-05, + -2.2268e-04, -5.5218e-04], + [ 5.3263e-04, 3.0994e-04, 2.7442e-04, ..., 2.5705e-05, + 1.3030e-04, -1.2884e-03], + ..., + [ 6.8665e-04, 6.0892e-04, 3.3016e-03, ..., 3.2616e-04, + 2.0862e-04, 2.1267e-03], + [-1.3971e-03, -1.4296e-03, -2.3603e-04, ..., 2.2843e-05, + 2.4700e-04, 4.4203e-04], + [ 7.3338e-04, 3.9291e-04, -1.7273e-02, ..., -2.6672e-02, + -3.4149e-02, -2.5162e-02]], device='cuda:0') +Epoch 302, bias, value: tensor([ 0.0144, 0.0033, -0.0012, 0.0152, -0.0051, -0.0054, 0.0108, 0.0255, + -0.0352, 0.0467], device='cuda:0'), grad: tensor([ 0.0364, 0.0014, -0.0245, 0.0101, 0.0498, -0.0190, -0.0198, 0.0229, + 0.0034, -0.0608], device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 217.03, cls_loss 0.5029 cls_loss_mapping 0.0053 cls_loss_causal 0.4780 re_mapping 0.0067 re_causal 0.0168 /// teacc 98.87 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.0737, 0.0708, -0.1052, ..., -0.1147, -0.1003, 0.0041], + [-0.0783, -0.1542, 0.0029, ..., -0.0556, -0.0450, -0.0989], + [ 0.0381, -0.0950, 0.0568, ..., 0.1596, -0.1229, -0.0429], + ..., + [-0.1039, -0.1853, 0.0903, ..., -0.0168, -0.0472, 0.0726], + [ 0.0222, 0.0369, -0.0309, ..., -0.0801, -0.1012, -0.0212], + [-0.1540, -0.0122, -0.0055, ..., -0.1606, 0.1153, 0.0267]], + device='cuda:0'), grad: tensor([[ 1.3475e-03, 5.0735e-04, 3.1376e-04, ..., 5.7250e-05, + 2.9135e-04, 5.8889e-04], + [ 5.0306e-04, 1.8454e-04, -1.1194e-04, ..., 1.1951e-04, + -2.0771e-03, 4.2415e-04], + [ 9.4223e-04, 2.7704e-04, 5.6124e-04, ..., -1.0520e-04, + 2.9373e-04, 7.9441e-04], + ..., + [ 1.6432e-03, -6.0177e-04, -5.7106e-03, ..., -2.4357e-03, + -1.9665e-03, 7.3910e-04], + [ 1.0700e-03, 1.6427e-04, 4.3082e-04, ..., 9.7394e-05, + 2.7156e-04, 6.3515e-04], + [-9.9487e-03, 4.2057e-04, 4.7493e-03, ..., 1.7786e-03, + 1.9627e-03, -7.9155e-04]], device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0130, 0.0040, -0.0008, 0.0157, -0.0063, -0.0040, 0.0111, 0.0249, + -0.0365, 0.0478], device='cuda:0'), grad: tensor([ 0.0207, -0.0242, 0.0224, 0.0068, -0.0327, 0.0282, 0.0095, 0.0064, + 0.0186, -0.0557], device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 217.00, cls_loss 0.5064 cls_loss_mapping 0.0030 cls_loss_causal 0.4748 re_mapping 0.0071 re_causal 0.0189 /// teacc 98.77 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.0736, 0.0707, -0.1059, ..., -0.1158, -0.1001, 0.0053], + [-0.0802, -0.1564, 0.0030, ..., -0.0557, -0.0438, -0.0976], + [ 0.0383, -0.0938, 0.0561, ..., 0.1605, -0.1239, -0.0435], + ..., + [-0.1041, -0.1856, 0.0917, ..., -0.0171, -0.0471, 0.0720], + [ 0.0216, 0.0360, -0.0305, ..., -0.0811, -0.1005, -0.0204], + [-0.1549, -0.0114, -0.0069, ..., -0.1597, 0.1143, 0.0266]], + device='cuda:0'), grad: tensor([[ 7.6818e-04, 1.7810e-04, 7.8344e-04, ..., 6.0022e-05, + 9.5129e-04, 1.0328e-03], + [-7.4530e-04, 2.4393e-05, 9.9850e-04, ..., 2.5824e-05, + -8.7881e-04, -4.1699e-04], + [-7.1764e-05, 2.3246e-04, -2.7409e-03, ..., -8.3780e-04, + 9.8038e-04, 1.0042e-03], + ..., + [ 2.5616e-03, 2.9993e-04, 1.0319e-03, ..., -6.6876e-05, + 2.1152e-03, 2.3994e-03], + [-1.6983e-02, 2.5768e-03, 1.7338e-03, ..., 5.5361e-04, + 7.4196e-03, 4.1924e-03], + [ 2.9049e-03, 9.0408e-04, -2.7008e-03, ..., 9.8348e-05, + -3.4828e-03, -1.3056e-03]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0131, 0.0042, -0.0015, 0.0154, -0.0059, -0.0049, 0.0115, 0.0249, + -0.0358, 0.0478], device='cuda:0'), grad: tensor([ 0.0201, -0.0255, -0.0164, 0.0373, -0.0039, -0.0404, -0.0085, 0.0298, + 0.0087, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 217.24, cls_loss 0.4586 cls_loss_mapping 0.0032 cls_loss_causal 0.4305 re_mapping 0.0068 re_causal 0.0162 /// teacc 98.93 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.0744, 0.0702, -0.1066, ..., -0.1170, -0.1007, 0.0056], + [-0.0800, -0.1568, 0.0044, ..., -0.0556, -0.0423, -0.0974], + [ 0.0370, -0.0941, 0.0574, ..., 0.1608, -0.1238, -0.0440], + ..., + [-0.1046, -0.1855, 0.0916, ..., -0.0175, -0.0478, 0.0714], + [ 0.0218, 0.0355, -0.0315, ..., -0.0820, -0.1009, -0.0211], + [-0.1553, -0.0116, -0.0076, ..., -0.1598, 0.1139, 0.0274]], + device='cuda:0'), grad: tensor([[ 8.0585e-04, 8.3521e-06, -3.5763e-03, ..., -1.2445e-03, + 1.1599e-04, 1.6861e-02], + [ 3.4380e-04, 2.2585e-07, 1.2960e-03, ..., -4.5052e-03, + 2.2793e-04, -1.0643e-03], + [ 3.0518e-04, 6.7540e-06, 2.4643e-03, ..., 3.8958e-04, + 2.1410e-04, 1.8749e-03], + ..., + [-5.7650e-04, 1.0291e-06, -1.2062e-02, ..., -9.6634e-06, + -4.7340e-03, -2.1729e-02], + [ 2.4796e-04, 1.9431e-05, 3.2635e-03, ..., 1.5774e-03, + 1.7893e-04, 1.0624e-03], + [-8.9765e-05, -2.3112e-05, -6.6032e-03, ..., -8.5068e-04, + -1.0902e-02, -7.2937e-03]], device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0136, 0.0045, -0.0006, 0.0148, -0.0065, -0.0048, 0.0111, 0.0254, + -0.0362, 0.0475], device='cuda:0'), grad: tensor([-0.0150, -0.0115, 0.0239, 0.0206, 0.0176, 0.0095, -0.0213, -0.0515, + 0.0272, 0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 216.91, cls_loss 0.4784 cls_loss_mapping 0.0044 cls_loss_causal 0.4604 re_mapping 0.0066 re_causal 0.0166 /// teacc 98.92 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.0745, 0.0709, -0.1065, ..., -0.1168, -0.1010, 0.0046], + [-0.0809, -0.1577, 0.0045, ..., -0.0550, -0.0422, -0.0982], + [ 0.0369, -0.0944, 0.0569, ..., 0.1612, -0.1248, -0.0441], + ..., + [-0.1055, -0.1858, 0.0918, ..., -0.0182, -0.0480, 0.0715], + [ 0.0228, 0.0356, -0.0311, ..., -0.0834, -0.1005, -0.0203], + [-0.1561, -0.0117, -0.0073, ..., -0.1599, 0.1139, 0.0269]], + device='cuda:0'), grad: tensor([[ 1.5507e-03, 1.1930e-03, 3.6573e-04, ..., 2.6989e-04, + 2.7132e-04, 5.9748e-04], + [-1.4458e-03, 1.5008e-04, 1.6677e-04, ..., 1.8406e-04, + 9.4414e-05, 3.5691e-04], + [-2.4261e-02, -2.7585e-04, -2.3010e-02, ..., -5.2872e-03, + 2.0099e-04, -1.2183e-04], + ..., + [-2.3499e-03, 2.3782e-04, 6.0768e-03, ..., 3.5977e-04, + 4.0207e-03, 3.6850e-03], + [ 1.8959e-03, 1.0862e-03, 5.2118e-04, ..., 2.9182e-04, + 5.0497e-04, 9.5797e-04], + [ 1.3533e-03, 5.6410e-04, -6.9580e-03, ..., 4.1842e-04, + -3.2291e-03, -8.9340e-03]], device='cuda:0') +Epoch 306, bias, value: tensor([ 1.3114e-02, 4.2845e-03, -8.3676e-05, 1.5501e-02, -6.0908e-03, + -3.9465e-03, 1.1738e-02, 2.4172e-02, -3.6074e-02, 4.6386e-02], + device='cuda:0'), grad: tensor([-0.0126, 0.0101, -0.0272, 0.0202, 0.0148, -0.0308, -0.0062, 0.0117, + 0.0230, -0.0031], device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 217.18, cls_loss 0.4877 cls_loss_mapping 0.0035 cls_loss_causal 0.4658 re_mapping 0.0067 re_causal 0.0178 /// teacc 98.84 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.0736, 0.0717, -0.1068, ..., -0.1170, -0.1017, 0.0051], + [-0.0805, -0.1583, 0.0044, ..., -0.0544, -0.0424, -0.0978], + [ 0.0366, -0.0940, 0.0566, ..., 0.1604, -0.1243, -0.0440], + ..., + [-0.1056, -0.1854, 0.0921, ..., -0.0180, -0.0488, 0.0714], + [ 0.0235, 0.0362, -0.0296, ..., -0.0820, -0.0994, -0.0203], + [-0.1562, -0.0120, -0.0064, ..., -0.1610, 0.1136, 0.0274]], + device='cuda:0'), grad: tensor([[ 4.4674e-05, 2.9545e-03, 1.5032e-04, ..., 7.5161e-05, + 1.6272e-04, -3.7719e-06], + [ 3.0458e-05, 9.1374e-05, 2.6464e-04, ..., 2.2900e-04, + 2.1470e-04, 4.8243e-07], + [ 1.2994e-04, -1.1864e-03, 3.8004e-04, ..., 2.5225e-04, + -1.7605e-03, 2.9188e-06], + ..., + [ 5.6982e-05, 1.4126e-04, 6.6698e-05, ..., 2.1183e-04, + 2.3663e-04, 1.8024e-04], + [ 6.2275e-04, 6.9094e-04, 5.7459e-04, ..., 2.7514e-04, + 4.3344e-04, 1.3053e-04], + [ 1.4067e-04, -4.3488e-03, 6.0892e-04, ..., 1.0216e-04, + 2.6178e-04, 2.0480e-04]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0149, 0.0051, -0.0007, 0.0145, -0.0056, -0.0048, 0.0109, 0.0234, + -0.0349, 0.0465], device='cuda:0'), grad: tensor([ 0.0150, 0.0119, -0.0230, -0.0230, -0.0241, 0.0157, 0.0087, 0.0081, + 0.0150, -0.0043], device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 216.85, cls_loss 0.4836 cls_loss_mapping 0.0043 cls_loss_causal 0.4539 re_mapping 0.0063 re_causal 0.0160 /// teacc 98.75 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.0734, 0.0710, -0.1081, ..., -0.1171, -0.1022, 0.0051], + [-0.0796, -0.1585, 0.0063, ..., -0.0529, -0.0421, -0.0982], + [ 0.0365, -0.0941, 0.0567, ..., 0.1591, -0.1240, -0.0430], + ..., + [-0.1060, -0.1860, 0.0907, ..., -0.0182, -0.0500, 0.0708], + [ 0.0244, 0.0371, -0.0306, ..., -0.0827, -0.1008, -0.0222], + [-0.1567, -0.0120, -0.0066, ..., -0.1631, 0.1145, 0.0278]], + device='cuda:0'), grad: tensor([[ 9.1362e-04, 2.5272e-03, 3.4256e-03, ..., 5.3120e-04, + 9.1553e-04, 2.3098e-03], + [ 7.1239e-04, 1.3294e-03, 8.3208e-04, ..., 3.7241e-04, + 1.0920e-03, 2.0161e-03], + [-3.1322e-05, 5.3644e-04, 1.5802e-03, ..., 4.5490e-04, + 1.1988e-03, 3.2692e-03], + ..., + [ 7.5877e-05, 4.1533e-04, 1.8034e-03, ..., 3.9077e-04, + 1.3113e-03, 2.9640e-03], + [-3.0494e-04, -8.0252e-04, 1.4925e-03, ..., 3.2377e-04, + 1.0796e-03, 2.5482e-03], + [ 3.4809e-04, 6.9761e-04, -9.8267e-03, ..., -2.1095e-03, + -1.0658e-02, -1.4778e-02]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0159, 0.0049, -0.0007, 0.0154, -0.0056, -0.0054, 0.0107, 0.0234, + -0.0356, 0.0461], device='cuda:0'), grad: tensor([ 0.0445, 0.0290, 0.0324, -0.0342, -0.0321, 0.0039, 0.0320, 0.0439, + 0.0176, -0.1368], device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 217.19, cls_loss 0.4893 cls_loss_mapping 0.0037 cls_loss_causal 0.4635 re_mapping 0.0063 re_causal 0.0160 /// teacc 98.78 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.0732, 0.0712, -0.1086, ..., -0.1171, -0.1022, 0.0055], + [-0.0793, -0.1585, 0.0064, ..., -0.0534, -0.0424, -0.0979], + [ 0.0367, -0.0943, 0.0565, ..., 0.1595, -0.1241, -0.0435], + ..., + [-0.1066, -0.1875, 0.0918, ..., -0.0166, -0.0496, 0.0703], + [ 0.0242, 0.0378, -0.0314, ..., -0.0830, -0.0995, -0.0221], + [-0.1564, -0.0121, -0.0064, ..., -0.1626, 0.1139, 0.0277]], + device='cuda:0'), grad: tensor([[ 2.1577e-04, 9.2745e-05, 1.5535e-03, ..., 8.4460e-05, + 5.2124e-05, 5.0545e-05], + [ 3.4547e-04, 6.3896e-05, -2.3670e-03, ..., 1.5378e-05, + 1.5244e-05, 1.6138e-05], + [ 2.5964e-04, 3.1090e-04, 1.2565e-04, ..., 2.4483e-05, + 2.5883e-05, 2.4766e-05], + ..., + [ 9.9987e-06, 3.1620e-05, -4.6425e-03, ..., -7.6648e-07, + -1.3721e-04, 4.4107e-06], + [-2.9993e-04, 1.4734e-04, 6.7115e-05, ..., -4.6846e-07, + 4.0196e-06, 2.5436e-05], + [ 6.6817e-05, 1.7464e-04, 4.9858e-03, ..., 2.1100e-05, + 2.2793e-04, 7.3195e-05]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0161, 0.0063, -0.0001, 0.0144, -0.0071, -0.0046, 0.0100, 0.0236, + -0.0351, 0.0457], device='cuda:0'), grad: tensor([ 0.0235, -0.0107, 0.0055, -0.0175, 0.0058, -0.0040, -0.0051, 0.0012, + 0.0031, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 217.23, cls_loss 0.4796 cls_loss_mapping 0.0039 cls_loss_causal 0.4584 re_mapping 0.0062 re_causal 0.0155 /// teacc 98.87 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.0735, 0.0712, -0.1074, ..., -0.1159, -0.1023, 0.0063], + [-0.0792, -0.1566, 0.0054, ..., -0.0546, -0.0430, -0.0980], + [ 0.0388, -0.0932, 0.0563, ..., 0.1600, -0.1251, -0.0436], + ..., + [-0.1079, -0.1887, 0.0913, ..., -0.0179, -0.0497, 0.0704], + [ 0.0227, 0.0376, -0.0305, ..., -0.0818, -0.0992, -0.0231], + [-0.1547, -0.0118, -0.0065, ..., -0.1648, 0.1136, 0.0273]], + device='cuda:0'), grad: tensor([[ 5.3930e-04, 1.2941e-03, 1.8299e-04, ..., 1.1814e-04, + 1.5879e-03, 7.4768e-04], + [ 2.5606e-04, 9.4056e-05, 3.2139e-04, ..., 1.5438e-04, + 1.1420e-04, 2.5487e-04], + [ 3.0479e-03, 7.3481e-04, 1.4663e-04, ..., -1.4174e-04, + 4.3201e-04, -7.2813e-04], + ..., + [ 2.0933e-04, 1.2469e-04, 2.2638e-04, ..., -1.0200e-05, + 2.4891e-04, 2.8276e-04], + [ 5.4270e-05, -7.2336e-04, -1.1616e-03, ..., 2.0668e-05, + -4.8637e-03, -3.7746e-03], + [-1.2846e-03, -9.1629e-03, 5.8937e-04, ..., -7.2861e-04, + -7.6485e-03, -1.8625e-03]], device='cuda:0') +Epoch 310, bias, value: tensor([ 0.0157, 0.0068, -0.0004, 0.0142, -0.0063, -0.0044, 0.0107, 0.0229, + -0.0353, 0.0454], device='cuda:0'), grad: tensor([ 0.0154, -0.0160, -0.0133, 0.0188, 0.0284, 0.0114, -0.0136, 0.0147, + -0.0088, -0.0370], device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 216.82, cls_loss 0.5042 cls_loss_mapping 0.0036 cls_loss_causal 0.4715 re_mapping 0.0064 re_causal 0.0158 /// teacc 98.71 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.0737, 0.0719, -0.1079, ..., -0.1159, -0.1024, 0.0086], + [-0.0780, -0.1565, 0.0054, ..., -0.0556, -0.0426, -0.0970], + [ 0.0375, -0.0942, 0.0562, ..., 0.1597, -0.1262, -0.0441], + ..., + [-0.1088, -0.1891, 0.0918, ..., -0.0175, -0.0495, 0.0710], + [ 0.0222, 0.0380, -0.0310, ..., -0.0826, -0.0982, -0.0217], + [-0.1541, -0.0120, -0.0070, ..., -0.1658, 0.1136, 0.0257]], + device='cuda:0'), grad: tensor([[ 4.0221e-04, 1.6654e-04, 2.7990e-04, ..., 3.1686e-04, + 4.4018e-05, 6.9380e-04], + [ 8.3637e-04, 2.9111e-04, -1.1406e-03, ..., -4.4250e-04, + 2.0814e-04, -1.0366e-03], + [ 8.9121e-04, 3.0160e-04, 1.9140e-03, ..., 1.6184e-03, + 1.7011e-04, 3.1719e-03], + ..., + [ 1.9016e-03, 1.6773e-04, -2.8172e-03, ..., -1.7366e-03, + 4.4870e-04, -5.6207e-05], + [-4.5967e-03, -2.1572e-03, 6.7377e-04, ..., 5.4693e-04, + -1.1654e-03, -2.5597e-03], + [ 3.2043e-04, 2.7671e-05, -9.6893e-04, ..., -6.2418e-04, + -1.0567e-03, 4.0793e-04]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0153, 0.0078, -0.0009, 0.0149, -0.0065, -0.0046, 0.0104, 0.0225, + -0.0345, 0.0449], device='cuda:0'), grad: tensor([ 0.0127, -0.0048, 0.0294, 0.0127, -0.0192, 0.0077, 0.0201, -0.0225, + -0.0146, -0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 216.37, cls_loss 0.5123 cls_loss_mapping 0.0039 cls_loss_causal 0.4847 re_mapping 0.0066 re_causal 0.0164 /// teacc 98.83 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.0740, 0.0720, -0.1078, ..., -0.1156, -0.1013, 0.0092], + [-0.0783, -0.1569, 0.0064, ..., -0.0560, -0.0423, -0.0974], + [ 0.0380, -0.0941, 0.0562, ..., 0.1595, -0.1268, -0.0442], + ..., + [-0.1068, -0.1884, 0.0912, ..., -0.0161, -0.0488, 0.0729], + [ 0.0219, 0.0376, -0.0304, ..., -0.0821, -0.0987, -0.0211], + [-0.1534, -0.0112, -0.0063, ..., -0.1664, 0.1137, 0.0253]], + device='cuda:0'), grad: tensor([[ 2.2084e-05, 3.3081e-05, 3.4761e-04, ..., 9.1255e-05, + 4.0531e-04, 1.9302e-03], + [ 3.9414e-06, 1.0185e-05, 7.2479e-04, ..., -5.4359e-04, + 1.5516e-03, 9.1600e-04], + [ 1.6794e-05, 1.2606e-05, -3.6411e-03, ..., 2.3282e-04, + -5.1231e-03, -8.4534e-03], + ..., + [ 1.8761e-05, 1.4856e-05, 1.5383e-03, ..., 1.1683e-04, + 2.5368e-03, 5.8632e-03], + [ 3.4885e-03, 3.7174e-03, 6.1703e-04, ..., 1.7476e-04, + 3.0041e-03, 4.3182e-03], + [ 1.5235e-04, 1.4842e-04, -2.3899e-03, ..., 1.6606e-04, + -6.6566e-03, 1.6365e-03]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0147, 0.0071, -0.0013, 0.0152, -0.0060, -0.0050, 0.0106, 0.0236, + -0.0348, 0.0452], device='cuda:0'), grad: tensor([ 0.0137, 0.0018, -0.0146, -0.0586, -0.0397, 0.0175, 0.0170, 0.0303, + 0.0356, -0.0030], device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 216.23, cls_loss 0.4802 cls_loss_mapping 0.0029 cls_loss_causal 0.4524 re_mapping 0.0065 re_causal 0.0166 /// teacc 98.88 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.0745, 0.0721, -0.1070, ..., -0.1165, -0.1011, 0.0096], + [-0.0783, -0.1566, 0.0069, ..., -0.0561, -0.0426, -0.0985], + [ 0.0367, -0.0944, 0.0569, ..., 0.1601, -0.1269, -0.0447], + ..., + [-0.1066, -0.1885, 0.0908, ..., -0.0170, -0.0499, 0.0737], + [ 0.0219, 0.0365, -0.0317, ..., -0.0825, -0.0991, -0.0211], + [-0.1541, -0.0114, -0.0066, ..., -0.1667, 0.1146, 0.0251]], + device='cuda:0'), grad: tensor([[ 2.4867e-04, 2.8396e-04, 7.7772e-04, ..., 9.8991e-04, + 4.1842e-04, 1.7195e-03], + [ 3.7456e-04, 4.0841e-04, 3.7265e-04, ..., 7.9536e-04, + 1.4496e-03, 1.9989e-03], + [ 1.7762e-04, 1.9443e-04, 4.1938e-04, ..., 5.1022e-04, + 3.0494e-04, 9.2983e-04], + ..., + [ 9.1970e-05, 3.6407e-04, 3.6221e-03, ..., 2.2113e-04, + 3.5238e-04, 1.6365e-03], + [ 5.6505e-04, 6.4373e-04, 3.4380e-04, ..., 5.7077e-04, + 1.0433e-03, 1.1673e-03], + [ 1.1606e-03, 2.9716e-03, -2.4414e-03, ..., 4.6325e-04, + 3.4084e-03, 1.4830e-03]], device='cuda:0') +Epoch 313, bias, value: tensor([ 0.0146, 0.0069, -0.0012, 0.0148, -0.0056, -0.0043, 0.0113, 0.0239, + -0.0361, 0.0450], device='cuda:0'), grad: tensor([ 0.0173, -0.0106, 0.0125, -0.0539, 0.0206, 0.0246, -0.0422, -0.0130, + 0.0178, 0.0268], device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 216.77, cls_loss 0.4927 cls_loss_mapping 0.0032 cls_loss_causal 0.4690 re_mapping 0.0064 re_causal 0.0160 /// teacc 98.77 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.0746, 0.0720, -0.1066, ..., -0.1171, -0.1013, 0.0094], + [-0.0784, -0.1573, 0.0065, ..., -0.0564, -0.0428, -0.0988], + [ 0.0374, -0.0930, 0.0571, ..., 0.1599, -0.1276, -0.0443], + ..., + [-0.1075, -0.1881, 0.0913, ..., -0.0157, -0.0493, 0.0728], + [ 0.0211, 0.0355, -0.0313, ..., -0.0804, -0.0991, -0.0219], + [-0.1540, -0.0116, -0.0068, ..., -0.1660, 0.1147, 0.0258]], + device='cuda:0'), grad: tensor([[ 1.0824e-03, -8.5974e-04, -2.8324e-03, ..., -3.5820e-03, + -8.9979e-04, -4.0779e-03], + [ 2.3103e-04, 1.1753e-06, 4.7874e-04, ..., 6.4754e-04, + 1.2177e-04, 9.5177e-04], + [ 1.8959e-03, 1.5751e-05, -2.0706e-02, ..., -1.4038e-02, + 9.9361e-05, 9.6369e-04], + ..., + [ 2.8038e-04, 1.8440e-07, 2.1606e-02, ..., 1.5083e-02, + 1.6940e-04, 6.9380e-04], + [ 8.8692e-04, 5.1618e-05, 2.8348e-04, ..., 3.2377e-04, + 1.1235e-04, 6.1083e-04], + [-1.0742e-02, 1.2666e-05, -2.5158e-03, ..., -1.2789e-03, + 1.2994e-04, -2.3289e-03]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0142, 0.0082, -0.0009, 0.0153, -0.0066, -0.0056, 0.0102, 0.0249, + -0.0360, 0.0455], device='cuda:0'), grad: tensor([-0.0191, 0.0240, -0.0431, 0.0112, 0.0295, -0.0214, 0.0201, 0.0121, + 0.0137, -0.0270], device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 217.07, cls_loss 0.4572 cls_loss_mapping 0.0036 cls_loss_causal 0.4257 re_mapping 0.0064 re_causal 0.0162 /// teacc 98.82 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.0753, 0.0717, -0.1061, ..., -0.1173, -0.1014, 0.0098], + [-0.0775, -0.1587, 0.0066, ..., -0.0568, -0.0420, -0.1004], + [ 0.0370, -0.0930, 0.0575, ..., 0.1608, -0.1276, -0.0432], + ..., + [-0.1080, -0.1870, 0.0914, ..., -0.0162, -0.0496, 0.0737], + [ 0.0203, 0.0362, -0.0306, ..., -0.0784, -0.0996, -0.0208], + [-0.1542, -0.0121, -0.0062, ..., -0.1665, 0.1146, 0.0256]], + device='cuda:0'), grad: tensor([[ 4.6349e-03, 2.6274e-04, -7.7105e-04, ..., 7.7057e-04, + 2.6608e-04, 2.0676e-03], + [ 3.1471e-04, 7.2181e-05, -2.3508e-04, ..., -3.1929e-03, + -6.7663e-04, -7.9346e-03], + [ 1.0803e-02, 9.8515e-04, 2.5043e-03, ..., 2.5845e-03, + 6.6280e-04, 1.1536e-02], + ..., + [ 1.1168e-03, 4.8876e-04, 5.5838e-04, ..., 1.4582e-03, + 7.7200e-04, 4.3373e-03], + [ 4.7264e-03, 6.3753e-04, 9.6178e-04, ..., 1.1864e-03, + -9.4891e-04, -4.1733e-03], + [ 6.6376e-03, 8.6021e-04, 9.6703e-04, ..., 5.3215e-04, + 6.8321e-03, 1.4633e-02]], device='cuda:0') +Epoch 315, bias, value: tensor([ 1.3526e-02, 7.0558e-03, -6.3644e-06, 1.5316e-02, -6.6357e-03, + -5.6672e-03, 9.3662e-03, 2.5560e-02, -3.5389e-02, 4.5892e-02], + device='cuda:0'), grad: tensor([-0.0255, -0.0464, 0.0338, -0.0106, 0.0083, 0.0095, -0.0146, 0.0234, + -0.0070, 0.0293], device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 216.96, cls_loss 0.4769 cls_loss_mapping 0.0035 cls_loss_causal 0.4562 re_mapping 0.0061 re_causal 0.0160 /// teacc 98.74 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.0730, 0.0722, -0.1066, ..., -0.1169, -0.1001, 0.0105], + [-0.0771, -0.1594, 0.0073, ..., -0.0551, -0.0427, -0.0997], + [ 0.0363, -0.0936, 0.0575, ..., 0.1609, -0.1286, -0.0444], + ..., + [-0.1090, -0.1881, 0.0911, ..., -0.0160, -0.0506, 0.0735], + [ 0.0199, 0.0358, -0.0319, ..., -0.0801, -0.1007, -0.0222], + [-0.1539, -0.0111, -0.0058, ..., -0.1668, 0.1148, 0.0255]], + device='cuda:0'), grad: tensor([[ 7.0524e-04, -5.5218e-04, 2.4390e-04, ..., -7.6517e-06, + -1.6203e-03, 1.5650e-03], + [ 5.2691e-04, 3.7909e-05, 1.1721e-03, ..., 2.0676e-03, + 3.1322e-05, 5.2595e-04], + [ 1.1511e-03, 5.2881e-04, -1.2522e-03, ..., -3.4256e-03, + 3.5858e-04, 9.5224e-04], + ..., + [ 3.5439e-03, 1.9288e-04, 4.5166e-03, ..., 1.0786e-03, + 1.5316e-03, 3.4122e-03], + [ 8.5592e-04, 6.6137e-04, 9.9087e-04, ..., 8.7881e-04, + 4.9782e-04, 1.0061e-03], + [ 1.9474e-03, 1.1606e-03, -6.0997e-03, ..., 5.5265e-04, + -2.9469e-03, 1.3447e-03]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0134, 0.0079, -0.0002, 0.0155, -0.0078, -0.0048, 0.0095, 0.0251, + -0.0360, 0.0463], device='cuda:0'), grad: tensor([ 0.0083, -0.0352, -0.0025, -0.0023, 0.0193, -0.0355, -0.0087, 0.0288, + 0.0156, 0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 217.08, cls_loss 0.4735 cls_loss_mapping 0.0032 cls_loss_causal 0.4486 re_mapping 0.0065 re_causal 0.0165 /// teacc 98.71 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.0745, 0.0713, -0.1069, ..., -0.1176, -0.1002, 0.0111], + [-0.0771, -0.1607, 0.0064, ..., -0.0559, -0.0422, -0.1000], + [ 0.0368, -0.0931, 0.0580, ..., 0.1594, -0.1271, -0.0434], + ..., + [-0.1093, -0.1876, 0.0896, ..., -0.0163, -0.0510, 0.0736], + [ 0.0204, 0.0366, -0.0317, ..., -0.0798, -0.1009, -0.0230], + [-0.1544, -0.0120, -0.0054, ..., -0.1659, 0.1141, 0.0257]], + device='cuda:0'), grad: tensor([[ 7.0477e-04, -3.0923e-04, 1.6630e-04, ..., -3.4189e-04, + 1.3494e-04, 4.0507e-04], + [ 1.8282e-03, 1.1742e-05, 2.1744e-04, ..., 1.8120e-04, + 1.4627e-04, 2.1100e-04], + [ 1.2722e-03, 1.2550e-03, 7.9751e-05, ..., -1.1021e-04, + 1.0747e-04, 2.7895e-04], + ..., + [ 8.7321e-05, 1.4794e-04, 4.6158e-04, ..., 2.7227e-04, + 1.1426e-04, 3.9339e-04], + [ 1.4505e-03, 3.4447e-03, 6.5422e-04, ..., 9.7227e-04, + 5.6791e-04, 1.1873e-03], + [ 1.2255e-03, 1.9159e-03, 6.7377e-04, ..., 9.9659e-04, + 8.4305e-04, 1.6136e-03]], device='cuda:0') +Epoch 317, bias, value: tensor([ 1.3124e-02, 6.9803e-03, -8.3176e-05, 1.6050e-02, -7.8318e-03, + -4.1140e-03, 9.8540e-03, 2.5551e-02, -3.5425e-02, 4.4936e-02], + device='cuda:0'), grad: tensor([-0.0127, 0.0282, 0.0069, -0.0432, -0.0023, 0.0109, 0.0041, 0.0071, + -0.0116, 0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 217.23, cls_loss 0.5022 cls_loss_mapping 0.0043 cls_loss_causal 0.4828 re_mapping 0.0065 re_causal 0.0158 /// teacc 98.64 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.0743, 0.0720, -0.1079, ..., -0.1181, -0.1013, 0.0102], + [-0.0773, -0.1618, 0.0097, ..., -0.0557, -0.0404, -0.1000], + [ 0.0366, -0.0922, 0.0575, ..., 0.1609, -0.1275, -0.0437], + ..., + [-0.1096, -0.1884, 0.0893, ..., -0.0165, -0.0504, 0.0745], + [ 0.0207, 0.0364, -0.0319, ..., -0.0817, -0.1008, -0.0241], + [-0.1539, -0.0121, -0.0056, ..., -0.1661, 0.1144, 0.0270]], + device='cuda:0'), grad: tensor([[ 2.8062e-04, 7.4387e-05, 8.8120e-04, ..., 7.0381e-04, + 3.7456e-04, 5.0020e-04], + [ 1.1206e-04, 1.3471e-04, 8.5258e-04, ..., 9.2983e-04, + 2.4414e-04, 1.0109e-03], + [ 2.6608e-04, 2.1029e-04, -2.9445e-04, ..., -5.3596e-04, + -1.1265e-04, -1.9360e-03], + ..., + [ 3.0947e-04, 3.1161e-04, 2.4014e-03, ..., 2.5673e-03, + 1.9131e-03, 4.5090e-03], + [ 1.9970e-03, 1.7748e-03, 7.3242e-04, ..., 6.9523e-04, + 1.6565e-03, 3.1223e-03], + [-5.0163e-04, -4.7970e-04, 1.8406e-03, ..., 2.0237e-03, + -9.7370e-04, -1.7166e-03]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0125, 0.0083, -0.0001, 0.0162, -0.0067, -0.0040, 0.0093, 0.0246, + -0.0363, 0.0454], device='cuda:0'), grad: tensor([-0.0453, 0.0198, -0.0132, -0.0347, -0.0115, -0.0081, 0.0223, 0.0320, + 0.0242, 0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 217.25, cls_loss 0.4940 cls_loss_mapping 0.0043 cls_loss_causal 0.4677 re_mapping 0.0063 re_causal 0.0161 /// teacc 98.72 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.0725, 0.0736, -0.1065, ..., -0.1157, -0.1029, 0.0100], + [-0.0778, -0.1616, 0.0083, ..., -0.0569, -0.0405, -0.1001], + [ 0.0358, -0.0929, 0.0570, ..., 0.1614, -0.1276, -0.0439], + ..., + [-0.1065, -0.1885, 0.0897, ..., -0.0168, -0.0515, 0.0753], + [ 0.0208, 0.0372, -0.0301, ..., -0.0786, -0.1002, -0.0242], + [-0.1544, -0.0121, -0.0063, ..., -0.1673, 0.1142, 0.0265]], + device='cuda:0'), grad: tensor([[ 0.0041, 0.0047, 0.0018, ..., 0.0030, 0.0003, -0.0044], + [ 0.0002, 0.0002, 0.0008, ..., 0.0021, 0.0003, 0.0008], + [ 0.0024, 0.0020, -0.0059, ..., 0.0022, -0.0020, -0.0012], + ..., + [ 0.0004, 0.0001, 0.0023, ..., 0.0023, 0.0004, 0.0013], + [ 0.0222, 0.0213, 0.0033, ..., 0.0070, 0.0002, 0.0007], + [-0.0019, -0.0035, -0.0058, ..., -0.0039, -0.0005, 0.0005]], + device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0120, 0.0087, -0.0003, 0.0155, -0.0063, -0.0038, 0.0092, 0.0250, + -0.0355, 0.0446], device='cuda:0'), grad: tensor([-0.0172, 0.0221, -0.0059, -0.0523, 0.0216, 0.0106, -0.0176, 0.0238, + 0.0782, -0.0634], device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 217.23, cls_loss 0.4809 cls_loss_mapping 0.0034 cls_loss_causal 0.4674 re_mapping 0.0065 re_causal 0.0170 /// teacc 98.74 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.0742, 0.0727, -0.1068, ..., -0.1150, -0.1041, 0.0107], + [-0.0784, -0.1619, 0.0084, ..., -0.0576, -0.0400, -0.1016], + [ 0.0364, -0.0931, 0.0564, ..., 0.1612, -0.1270, -0.0432], + ..., + [-0.1074, -0.1886, 0.0904, ..., -0.0162, -0.0519, 0.0748], + [ 0.0204, 0.0362, -0.0303, ..., -0.0784, -0.1000, -0.0227], + [-0.1536, -0.0118, -0.0069, ..., -0.1686, 0.1141, 0.0260]], + device='cuda:0'), grad: tensor([[ 9.6634e-06, -4.7326e-04, 3.2234e-04, ..., 4.6015e-04, + 2.7251e-04, 3.3665e-04], + [ 1.6525e-05, -1.0705e-04, -1.1988e-03, ..., -2.3422e-03, + -1.0777e-03, -2.2755e-03], + [ 1.4849e-05, 3.1978e-05, 2.2411e-03, ..., 1.6756e-03, + 4.5514e-04, 9.1982e-04], + ..., + [ 1.2212e-05, 2.8312e-05, 1.5240e-03, ..., 1.7328e-03, + 2.1839e-03, 1.4687e-03], + [-2.0134e-04, 6.1929e-05, 4.1437e-04, ..., -9.5415e-04, + -1.3676e-03, -3.9902e-03], + [ 1.3374e-05, 8.0287e-05, 2.7695e-03, ..., 2.6035e-03, + 1.8635e-03, 2.5158e-03]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0137, 0.0081, 0.0003, 0.0150, -0.0056, -0.0048, 0.0094, 0.0249, + -0.0355, 0.0438], device='cuda:0'), grad: tensor([ 0.0077, -0.0186, 0.0150, 0.0116, -0.0057, -0.0207, 0.0152, -0.0084, + -0.0167, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 216.53, cls_loss 0.5118 cls_loss_mapping 0.0036 cls_loss_causal 0.4834 re_mapping 0.0061 re_causal 0.0156 /// teacc 98.85 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.0742, 0.0724, -0.1067, ..., -0.1162, -0.1049, 0.0110], + [-0.0784, -0.1618, 0.0083, ..., -0.0574, -0.0395, -0.1016], + [ 0.0361, -0.0942, 0.0564, ..., 0.1604, -0.1274, -0.0427], + ..., + [-0.1078, -0.1896, 0.0917, ..., -0.0168, -0.0509, 0.0748], + [ 0.0203, 0.0369, -0.0308, ..., -0.0804, -0.1009, -0.0227], + [-0.1537, -0.0118, -0.0074, ..., -0.1669, 0.1152, 0.0259]], + device='cuda:0'), grad: tensor([[ 1.4365e-04, -3.0667e-05, 1.1081e-04, ..., 1.6344e-04, + 3.0708e-04, 3.3665e-04], + [ 4.3917e-04, 3.4499e-04, 1.1009e-04, ..., 2.7451e-02, + 1.1663e-03, 9.6846e-04], + [-4.2707e-05, -9.5189e-05, -1.3304e-04, ..., -2.7969e-02, + 3.7766e-04, 4.4417e-04], + ..., + [ 8.2016e-05, 1.1826e-04, 9.5367e-04, ..., 5.3549e-04, + 7.7629e-04, 3.1471e-03], + [-6.9523e-04, 1.0866e-04, 1.5867e-04, ..., 7.0810e-05, + -3.3116e-04, -6.6471e-04], + [ 1.8144e-04, -2.1005e-04, -2.6779e-03, ..., -1.0109e-03, + -7.1239e-04, -3.6602e-03]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0138, 0.0083, 0.0001, 0.0148, -0.0053, -0.0049, 0.0091, 0.0247, + -0.0350, 0.0436], device='cuda:0'), grad: tensor([ 0.0045, 0.0353, -0.0219, -0.0190, -0.0068, -0.0048, 0.0069, 0.0106, + 0.0024, -0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 217.01, cls_loss 0.4999 cls_loss_mapping 0.0025 cls_loss_causal 0.4724 re_mapping 0.0068 re_causal 0.0173 /// teacc 98.87 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.0745, 0.0728, -0.1069, ..., -0.1172, -0.1060, 0.0106], + [-0.0801, -0.1644, 0.0077, ..., -0.0582, -0.0391, -0.1011], + [ 0.0359, -0.0937, 0.0568, ..., 0.1611, -0.1266, -0.0428], + ..., + [-0.1092, -0.1907, 0.0915, ..., -0.0173, -0.0504, 0.0751], + [ 0.0213, 0.0376, -0.0309, ..., -0.0813, -0.1018, -0.0235], + [-0.1547, -0.0118, -0.0069, ..., -0.1665, 0.1153, 0.0260]], + device='cuda:0'), grad: tensor([[ 1.1051e-04, 1.2445e-04, 2.8944e-04, ..., 4.8250e-05, + 2.8729e-04, -2.0790e-03], + [ 2.1386e-04, 4.6164e-05, 1.1146e-05, ..., 2.8983e-06, + 2.0409e-04, 9.4652e-04], + [ 5.9664e-05, 1.6317e-05, -1.0020e-04, ..., -2.0516e-04, + 9.0241e-05, 5.7602e-04], + ..., + [ 2.2554e-04, 4.3797e-04, 6.8665e-04, ..., 9.9421e-05, + 9.4509e-04, 2.2659e-03], + [-9.5272e-04, 2.3842e-04, 1.6832e-04, ..., 1.9103e-05, + -4.4537e-04, -1.5135e-03], + [-2.5535e-04, 2.7800e-04, -4.3607e-04, ..., -2.0802e-04, + -1.7471e-03, 1.9073e-03]], device='cuda:0') +Epoch 322, bias, value: tensor([ 0.0138, 0.0083, -0.0005, 0.0159, -0.0055, -0.0055, 0.0088, 0.0249, + -0.0355, 0.0444], device='cuda:0'), grad: tensor([-0.0133, -0.0086, 0.0153, -0.0405, 0.0215, 0.0150, 0.0169, -0.0004, + 0.0028, -0.0088], device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 216.77, cls_loss 0.5124 cls_loss_mapping 0.0035 cls_loss_causal 0.4812 re_mapping 0.0066 re_causal 0.0169 /// teacc 98.75 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.0748, 0.0733, -0.1068, ..., -0.1172, -0.1068, 0.0110], + [-0.0818, -0.1642, 0.0066, ..., -0.0598, -0.0395, -0.1025], + [ 0.0355, -0.0941, 0.0562, ..., 0.1603, -0.1262, -0.0428], + ..., + [-0.1094, -0.1918, 0.0917, ..., -0.0174, -0.0503, 0.0757], + [ 0.0211, 0.0372, -0.0305, ..., -0.0801, -0.1004, -0.0227], + [-0.1554, -0.0117, -0.0081, ..., -0.1678, 0.1140, 0.0255]], + device='cuda:0'), grad: tensor([[ 6.7520e-04, 5.7258e-06, 6.3956e-05, ..., 1.0049e-04, + 9.3225e-07, 4.7445e-04], + [ 1.9398e-03, 1.9744e-06, 3.4869e-05, ..., 5.6118e-05, + 5.6811e-08, 1.3113e-03], + [-3.8204e-03, -3.1173e-05, -2.2907e-03, ..., -3.7384e-03, + 6.7521e-07, -1.3628e-03], + ..., + [ 1.4842e-04, 1.5637e-06, 1.9753e-04, ..., 3.1829e-04, + 6.6962e-07, -1.6527e-03], + [ 4.5753e-04, 2.6509e-05, 1.0214e-03, ..., 1.6451e-03, + 7.7039e-06, 2.2519e-04], + [ 3.2157e-05, 1.6034e-05, 3.2723e-05, ..., 5.2691e-05, + 7.5661e-06, 2.2995e-04]], device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0142, 0.0073, -0.0007, 0.0161, -0.0049, -0.0065, 0.0091, 0.0255, + -0.0343, 0.0433], device='cuda:0'), grad: tensor([-0.0173, 0.0242, -0.0121, 0.0166, 0.0126, -0.0216, 0.0110, -0.0118, + 0.0170, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 216.91, cls_loss 0.4596 cls_loss_mapping 0.0051 cls_loss_causal 0.4392 re_mapping 0.0062 re_causal 0.0150 /// teacc 98.74 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.0752, 0.0724, -0.1069, ..., -0.1163, -0.1076, 0.0109], + [-0.0830, -0.1651, 0.0067, ..., -0.0595, -0.0396, -0.1018], + [ 0.0364, -0.0944, 0.0577, ..., 0.1616, -0.1263, -0.0430], + ..., + [-0.1095, -0.1920, 0.0892, ..., -0.0187, -0.0516, 0.0757], + [ 0.0219, 0.0378, -0.0305, ..., -0.0804, -0.1001, -0.0234], + [-0.1548, -0.0112, -0.0079, ..., -0.1667, 0.1135, 0.0245]], + device='cuda:0'), grad: tensor([[-1.7080e-03, -2.7637e-03, 1.6857e-06, ..., -8.4341e-06, + 3.3677e-06, -1.1568e-03], + [ 2.6917e-04, 9.3102e-05, 2.2445e-06, ..., 1.0151e-06, + 4.7982e-06, 6.8367e-05], + [-4.0588e-03, 3.6931e-04, 1.6332e-05, ..., 5.5879e-05, + 2.0072e-05, 4.5824e-04], + ..., + [ 7.6532e-04, 1.5330e-04, -8.7261e-05, ..., -1.9342e-05, + -4.6566e-06, 5.1451e-04], + [ 1.9913e-03, 7.5960e-04, 2.1141e-06, ..., -4.6432e-05, + 3.0026e-05, 1.2674e-03], + [ 4.8790e-03, 3.4666e-04, 9.1016e-05, ..., 6.6876e-05, + 1.0854e-04, 1.9989e-03]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0139, 0.0079, 0.0003, 0.0163, -0.0058, -0.0061, 0.0091, 0.0239, + -0.0344, 0.0438], device='cuda:0'), grad: tensor([-0.0505, 0.0077, -0.0203, 0.0104, 0.0086, 0.0083, 0.0102, -0.0152, + 0.0128, 0.0281], device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 216.94, cls_loss 0.4973 cls_loss_mapping 0.0040 cls_loss_causal 0.4717 re_mapping 0.0059 re_causal 0.0150 /// teacc 98.68 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.0759, 0.0717, -0.1075, ..., -0.1181, -0.1085, 0.0106], + [-0.0835, -0.1649, 0.0077, ..., -0.0596, -0.0384, -0.1023], + [ 0.0366, -0.0935, 0.0570, ..., 0.1616, -0.1277, -0.0434], + ..., + [-0.1091, -0.1902, 0.0891, ..., -0.0186, -0.0511, 0.0752], + [ 0.0211, 0.0369, -0.0307, ..., -0.0808, -0.0994, -0.0232], + [-0.1554, -0.0113, -0.0066, ..., -0.1662, 0.1148, 0.0259]], + device='cuda:0'), grad: tensor([[-3.9554e-04, -1.5659e-03, 2.2840e-04, ..., 3.6860e-04, + 2.4557e-04, 3.9554e-04], + [-8.5878e-04, -2.7940e-08, -3.3245e-03, ..., -5.7297e-03, + -1.8454e-04, -2.7237e-03], + [ 7.4577e-04, 1.4830e-04, 2.7847e-03, ..., 4.1428e-03, + 8.4734e-04, 2.5063e-03], + ..., + [ 2.9159e-04, 8.7202e-05, 1.4114e-03, ..., 2.0981e-03, + 8.1348e-04, 1.6184e-03], + [-6.7329e-04, -5.3215e-04, -8.2791e-05, ..., 6.0320e-04, + -1.4763e-03, -9.5177e-04], + [ 1.1683e-04, 4.0126e-04, 3.3307e-04, ..., 5.1737e-04, + 3.6716e-04, 5.9080e-04]], device='cuda:0') +Epoch 325, bias, value: tensor([ 1.2629e-02, 8.2973e-03, 6.6948e-05, 1.5399e-02, -5.8506e-03, + -5.8256e-03, 9.5625e-03, 2.3627e-02, -3.4739e-02, 4.5730e-02], + device='cuda:0'), grad: tensor([ 0.0102, -0.0338, -0.0009, 0.0122, 0.0185, -0.0172, -0.0164, 0.0257, + -0.0165, 0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 217.06, cls_loss 0.4752 cls_loss_mapping 0.0022 cls_loss_causal 0.4493 re_mapping 0.0064 re_causal 0.0166 /// teacc 98.77 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.0778, 0.0708, -0.1078, ..., -0.1185, -0.1091, 0.0109], + [-0.0832, -0.1658, 0.0077, ..., -0.0591, -0.0384, -0.1025], + [ 0.0364, -0.0935, 0.0560, ..., 0.1608, -0.1279, -0.0435], + ..., + [-0.1095, -0.1909, 0.0894, ..., -0.0180, -0.0512, 0.0751], + [ 0.0217, 0.0366, -0.0307, ..., -0.0811, -0.0992, -0.0230], + [-0.1566, -0.0113, -0.0063, ..., -0.1669, 0.1157, 0.0259]], + device='cuda:0'), grad: tensor([[ 1.1027e-05, -2.3198e-04, 1.5163e-04, ..., 1.1444e-04, + 7.9535e-07, 2.5965e-06], + [ 4.2498e-05, -4.0680e-05, 3.7289e-04, ..., 7.0095e-04, + 2.5760e-06, 2.2829e-05], + [ 4.7415e-05, 1.1459e-05, 2.3103e-04, ..., -1.6680e-03, + 1.3644e-06, 3.4660e-05], + ..., + [-4.5090e-03, 1.1206e-05, -1.6565e-03, ..., 5.2795e-03, + 8.8096e-05, 3.4571e-04], + [ 4.2763e-03, 4.6343e-05, 7.0381e-04, ..., 5.2214e-04, + 7.5735e-06, 6.2943e-03], + [ 4.5395e-03, 2.2799e-05, -4.3716e-03, ..., -8.0338e-03, + -8.5831e-04, -3.5095e-04]], device='cuda:0') +Epoch 326, bias, value: tensor([ 0.0121, 0.0092, -0.0003, 0.0149, -0.0045, -0.0051, 0.0084, 0.0244, + -0.0353, 0.0451], device='cuda:0'), grad: tensor([ 0.0140, -0.0103, 0.0134, 0.0167, -0.0028, -0.0175, -0.0471, -0.0166, + 0.0394, 0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 216.74, cls_loss 0.4993 cls_loss_mapping 0.0033 cls_loss_causal 0.4657 re_mapping 0.0065 re_causal 0.0172 /// teacc 98.85 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.0772, 0.0713, -0.1068, ..., -0.1182, -0.1104, 0.0104], + [-0.0851, -0.1653, 0.0075, ..., -0.0585, -0.0397, -0.1034], + [ 0.0368, -0.0932, 0.0555, ..., 0.1600, -0.1275, -0.0418], + ..., + [-0.1090, -0.1920, 0.0890, ..., -0.0174, -0.0518, 0.0742], + [ 0.0215, 0.0371, -0.0302, ..., -0.0817, -0.0989, -0.0233], + [-0.1557, -0.0113, -0.0062, ..., -0.1669, 0.1160, 0.0256]], + device='cuda:0'), grad: tensor([[ 1.0881e-03, 1.8015e-03, 1.0643e-03, ..., 1.6918e-03, + 1.6165e-03, 3.8013e-03], + [-4.7475e-05, 1.7822e-04, 3.5071e-04, ..., 2.5964e-04, + -2.3365e-05, 1.2360e-03], + [ 6.8998e-04, 1.0214e-03, -2.8934e-03, ..., -3.2921e-03, + 1.0805e-03, 3.0136e-03], + ..., + [ 1.1325e-04, 1.1063e-04, 5.6553e-04, ..., 3.4008e-03, + 3.9607e-05, 1.4887e-03], + [ 1.1940e-03, 1.7338e-03, 9.1267e-04, ..., -1.0455e-04, + 2.0485e-03, 3.9711e-03], + [ 2.8944e-04, 3.2854e-04, 2.5082e-03, ..., 1.1644e-03, + 5.8222e-04, 1.7643e-03]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0122, 0.0103, 0.0001, 0.0145, -0.0041, -0.0052, 0.0080, 0.0247, + -0.0358, 0.0442], device='cuda:0'), grad: tensor([ 0.0212, 0.0109, 0.0095, -0.0777, 0.0183, 0.0116, -0.0163, 0.0095, + -0.0054, 0.0184], device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 217.13, cls_loss 0.4773 cls_loss_mapping 0.0043 cls_loss_causal 0.4618 re_mapping 0.0060 re_causal 0.0161 /// teacc 98.77 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.0776, 0.0729, -0.1067, ..., -0.1182, -0.1097, 0.0109], + [-0.0859, -0.1667, 0.0074, ..., -0.0581, -0.0396, -0.1033], + [ 0.0360, -0.0931, 0.0554, ..., 0.1601, -0.1277, -0.0430], + ..., + [-0.1087, -0.1915, 0.0904, ..., -0.0168, -0.0509, 0.0755], + [ 0.0218, 0.0352, -0.0307, ..., -0.0822, -0.0992, -0.0233], + [-0.1546, -0.0099, -0.0068, ..., -0.1675, 0.1154, 0.0244]], + device='cuda:0'), grad: tensor([[ 1.2457e-04, -8.2159e-04, 1.3351e-04, ..., 1.8090e-05, + 5.0038e-05, 1.0414e-03], + [ 4.9621e-05, 1.6484e-06, 9.4950e-05, ..., 5.5358e-06, + -4.8971e-04, 4.2939e-04], + [-4.9965e-07, 2.8200e-06, 2.0742e-04, ..., -2.1482e-04, + 4.7237e-05, 6.4898e-04], + ..., + [-4.9925e-04, 1.1928e-05, -1.3409e-03, ..., 1.3173e-04, + 4.3559e-04, -2.3594e-03], + [ 8.4996e-05, 3.3647e-05, 2.5678e-04, ..., 5.6565e-05, + 4.0865e-04, 1.0519e-03], + [-1.5044e-04, -1.2660e-04, 2.2101e-04, ..., -1.2302e-04, + -1.6384e-03, 3.0541e-04]], device='cuda:0') +Epoch 328, bias, value: tensor([ 1.3299e-02, 9.8099e-03, -1.2428e-05, 1.5741e-02, -4.4775e-03, + -6.7836e-03, 7.8718e-03, 2.4853e-02, -3.4515e-02, 4.3177e-02], + device='cuda:0'), grad: tensor([ 0.0062, 0.0016, 0.0057, 0.0071, -0.0195, 0.0056, -0.0246, -0.0019, + 0.0089, 0.0111], device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 217.48, cls_loss 0.5054 cls_loss_mapping 0.0039 cls_loss_causal 0.4905 re_mapping 0.0055 re_causal 0.0143 /// teacc 98.73 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.0775, 0.0730, -0.1067, ..., -0.1181, -0.1103, 0.0114], + [-0.0871, -0.1678, 0.0064, ..., -0.0593, -0.0389, -0.1033], + [ 0.0372, -0.0931, 0.0550, ..., 0.1602, -0.1290, -0.0430], + ..., + [-0.1087, -0.1915, 0.0910, ..., -0.0165, -0.0505, 0.0752], + [ 0.0208, 0.0349, -0.0312, ..., -0.0822, -0.0992, -0.0239], + [-0.1544, -0.0099, -0.0065, ..., -0.1668, 0.1161, 0.0251]], + device='cuda:0'), grad: tensor([[ 1.3800e-03, 1.5697e-03, 2.1515e-03, ..., 6.4087e-04, + 1.4771e-06, 7.0620e-04], + [ 1.1975e-04, 1.0151e-04, -2.2209e-04, ..., -3.7432e-04, + 2.2491e-07, -7.9584e-04], + [ 1.4610e-03, 9.6941e-04, -1.9943e-02, ..., -3.1776e-03, + 3.4906e-06, 1.0500e-03], + ..., + [ 1.3053e-04, 9.5189e-05, 1.2238e-02, ..., 3.9253e-03, + 3.6880e-06, 1.2712e-03], + [-2.4068e-04, -1.6460e-03, 8.8692e-04, ..., -4.2653e-04, + 2.3376e-06, 4.8089e-04], + [ 5.2929e-04, 5.8985e-04, 1.7624e-03, ..., 3.2377e-04, + 7.0371e-06, 7.0429e-04]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0135, 0.0090, 0.0002, 0.0159, -0.0055, -0.0061, 0.0083, 0.0251, + -0.0354, 0.0440], device='cuda:0'), grad: tensor([-0.0093, -0.0105, -0.0359, -0.0065, 0.0112, 0.0023, -0.0114, 0.0551, + -0.0096, 0.0147], device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 217.19, cls_loss 0.4546 cls_loss_mapping 0.0032 cls_loss_causal 0.4316 re_mapping 0.0060 re_causal 0.0149 /// teacc 98.70 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.0781, 0.0726, -0.1078, ..., -0.1183, -0.1106, 0.0113], + [-0.0874, -0.1689, 0.0059, ..., -0.0589, -0.0376, -0.1037], + [ 0.0371, -0.0934, 0.0549, ..., 0.1602, -0.1309, -0.0436], + ..., + [-0.1085, -0.1933, 0.0911, ..., -0.0160, -0.0511, 0.0755], + [ 0.0205, 0.0338, -0.0323, ..., -0.0827, -0.0988, -0.0241], + [-0.1536, -0.0089, -0.0065, ..., -0.1674, 0.1155, 0.0252]], + device='cuda:0'), grad: tensor([[ 1.3158e-05, 1.5612e-03, 4.6778e-04, ..., 2.1133e-03, + 4.3660e-05, 4.8733e-04], + [ 8.7768e-06, 3.7193e-05, 4.4107e-04, ..., -2.1973e-03, + 1.6403e-03, 2.8973e-03], + [ 2.0428e-03, -1.0042e-03, -2.1172e-03, ..., 7.0763e-04, + -3.8013e-03, -5.9052e-03], + ..., + [ 9.1136e-05, 3.0923e-04, -1.1005e-03, ..., -3.5119e-04, + 6.8188e-04, 1.7395e-03], + [-3.7022e-03, 6.3419e-04, 1.4620e-03, ..., 1.1568e-03, + 2.6417e-03, 2.3804e-03], + [ 1.4156e-05, 2.0981e-04, 8.2397e-04, ..., 3.9530e-04, + 4.9263e-05, 8.2922e-04]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0129, 0.0098, 0.0011, 0.0166, -0.0063, -0.0055, 0.0078, 0.0250, + -0.0363, 0.0437], device='cuda:0'), grad: tensor([ 0.0244, 0.0067, 0.0021, -0.0236, 0.0223, -0.0320, -0.0037, -0.0097, + -0.0054, 0.0189], device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 216.99, cls_loss 0.4674 cls_loss_mapping 0.0026 cls_loss_causal 0.4416 re_mapping 0.0064 re_causal 0.0159 /// teacc 98.73 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.0788, 0.0727, -0.1081, ..., -0.1185, -0.1107, 0.0117], + [-0.0875, -0.1687, 0.0060, ..., -0.0585, -0.0374, -0.1041], + [ 0.0372, -0.0944, 0.0554, ..., 0.1600, -0.1300, -0.0429], + ..., + [-0.1082, -0.1943, 0.0915, ..., -0.0170, -0.0513, 0.0756], + [ 0.0208, 0.0345, -0.0314, ..., -0.0812, -0.0992, -0.0239], + [-0.1531, -0.0097, -0.0072, ..., -0.1675, 0.1152, 0.0247]], + device='cuda:0'), grad: tensor([[-3.6240e-03, -7.0457e-03, 1.0653e-03, ..., 9.5987e-04, + 2.1636e-04, 1.9693e-04], + [ 1.1462e-04, 1.1152e-04, 2.0051e-04, ..., 4.2111e-05, + 1.3435e-04, 1.4532e-04], + [ 1.8132e-04, 4.1604e-04, 5.1928e-04, ..., 4.1032e-04, + 1.7273e-04, 1.6940e-04], + ..., + [ 1.1402e-04, 2.5892e-04, 6.7329e-04, ..., 3.2473e-04, + 4.8685e-04, 4.7326e-04], + [-2.4204e-03, 7.8869e-04, -1.3132e-03, ..., 5.0688e-04, + -1.3208e-03, -1.7567e-03], + [ 2.7609e-04, 6.2227e-04, -1.5564e-03, ..., 3.9887e-04, + -3.4657e-03, -3.9635e-03]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0123, 0.0094, 0.0008, 0.0159, -0.0068, -0.0048, 0.0091, 0.0252, + -0.0358, 0.0436], device='cuda:0'), grad: tensor([ 0.0002, 0.0058, 0.0078, 0.0095, 0.0266, 0.0039, -0.0119, -0.0223, + -0.0241, 0.0045], device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 216.84, cls_loss 0.5300 cls_loss_mapping 0.0037 cls_loss_causal 0.5056 re_mapping 0.0062 re_causal 0.0159 /// teacc 98.73 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.0791, 0.0718, -0.1068, ..., -0.1189, -0.1115, 0.0119], + [-0.0874, -0.1704, 0.0063, ..., -0.0583, -0.0380, -0.1055], + [ 0.0377, -0.0943, 0.0545, ..., 0.1595, -0.1294, -0.0437], + ..., + [-0.1089, -0.1956, 0.0927, ..., -0.0159, -0.0517, 0.0765], + [ 0.0212, 0.0333, -0.0320, ..., -0.0815, -0.0993, -0.0231], + [-0.1538, -0.0092, -0.0082, ..., -0.1685, 0.1151, 0.0242]], + device='cuda:0'), grad: tensor([[ 8.5163e-04, 1.5509e-04, 3.6454e-04, ..., 7.2908e-04, + 3.6144e-04, 2.7924e-03], + [ 8.0490e-04, 1.4031e-04, 1.5354e-03, ..., 6.0158e-03, + -1.6556e-03, -4.0894e-03], + [-2.2564e-03, -2.5082e-04, -1.6060e-03, ..., -2.3329e-04, + -1.1473e-03, -8.5526e-03], + ..., + [ 2.8133e-04, 1.4555e-04, -5.1498e-04, ..., -2.6941e-04, + 6.2275e-04, 1.0614e-03], + [ 1.5926e-04, -1.3943e-03, -2.5415e-04, ..., 1.1474e-04, + 2.3651e-04, 1.0128e-03], + [ 2.1088e-04, -8.4519e-05, 4.2439e-04, ..., 1.7715e-04, + 4.8470e-04, 1.3838e-03]], device='cuda:0') +Epoch 332, bias, value: tensor([ 0.0121, 0.0088, 0.0001, 0.0160, -0.0064, -0.0043, 0.0097, 0.0256, + -0.0356, 0.0431], device='cuda:0'), grad: tensor([-0.0109, 0.0116, -0.0460, -0.0067, 0.0161, 0.0057, -0.0083, 0.0164, + 0.0078, 0.0144], device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 217.34, cls_loss 0.4920 cls_loss_mapping 0.0031 cls_loss_causal 0.4679 re_mapping 0.0063 re_causal 0.0150 /// teacc 98.77 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.0786, 0.0720, -0.1062, ..., -0.1169, -0.1130, 0.0113], + [-0.0870, -0.1701, 0.0074, ..., -0.0589, -0.0375, -0.1034], + [ 0.0382, -0.0941, 0.0549, ..., 0.1591, -0.1310, -0.0430], + ..., + [-0.1098, -0.1963, 0.0915, ..., -0.0162, -0.0521, 0.0761], + [ 0.0222, 0.0342, -0.0325, ..., -0.0808, -0.0999, -0.0226], + [-0.1530, -0.0087, -0.0058, ..., -0.1687, 0.1163, 0.0240]], + device='cuda:0'), grad: tensor([[-2.5902e-03, -4.8141e-03, -3.2825e-03, ..., -8.2159e-04, + -8.8310e-04, -2.2650e-04], + [ 2.3913e-04, 1.6904e-04, -4.8876e-04, ..., -3.7231e-03, + 2.9135e-04, 5.5885e-04], + [ 8.4496e-04, -4.9667e-03, 3.8934e-04, ..., 1.4181e-03, + 2.1231e-04, -8.6288e-03], + ..., + [ 4.2367e-04, 1.0324e-04, 2.4796e-04, ..., 5.8079e-04, + 2.1374e-04, 4.2701e-04], + [ 4.6806e-03, 1.2598e-03, -5.8556e-04, ..., -5.7757e-05, + -1.9474e-03, -3.1185e-03], + [ 2.0809e-03, 1.4699e-04, 1.5557e-04, ..., 6.3479e-05, + 9.9778e-05, 6.5470e-04]], device='cuda:0') +Epoch 333, bias, value: tensor([ 0.0115, 0.0095, 0.0009, 0.0142, -0.0061, -0.0056, 0.0099, 0.0255, + -0.0349, 0.0440], device='cuda:0'), grad: tensor([-0.0461, -0.0280, -0.0079, -0.0081, 0.0164, 0.0571, 0.0211, 0.0156, + -0.0335, 0.0134], device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 217.01, cls_loss 0.5127 cls_loss_mapping 0.0044 cls_loss_causal 0.4854 re_mapping 0.0062 re_causal 0.0156 /// teacc 98.79 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.0790, 0.0707, -0.1057, ..., -0.1161, -0.1134, 0.0115], + [-0.0878, -0.1709, 0.0066, ..., -0.0581, -0.0369, -0.1037], + [ 0.0369, -0.0944, 0.0545, ..., 0.1581, -0.1316, -0.0430], + ..., + [-0.1116, -0.1985, 0.0916, ..., -0.0159, -0.0520, 0.0767], + [ 0.0228, 0.0352, -0.0326, ..., -0.0814, -0.1006, -0.0228], + [-0.1526, -0.0068, -0.0055, ..., -0.1685, 0.1166, 0.0240]], + device='cuda:0'), grad: tensor([[ 1.3332e-03, 1.6298e-03, 3.2997e-04, ..., 1.6582e-04, + 8.3268e-05, 5.6028e-04], + [ 2.2411e-03, 1.2171e-04, 9.0885e-04, ..., 8.3113e-04, + 9.8467e-05, 6.8245e-03], + [ 4.0436e-04, 3.5977e-04, 6.0081e-04, ..., 4.4847e-04, + 2.4283e-04, 1.0471e-03], + ..., + [ 3.4022e-04, 2.7895e-04, -1.6546e-03, ..., -1.4553e-03, + 3.8099e-04, -3.1281e-04], + [ 1.4486e-03, 1.7338e-03, 9.1124e-04, ..., 6.0844e-04, + 4.4489e-04, 2.1496e-03], + [ 3.6383e-04, -1.6146e-03, -4.9515e-03, ..., -4.0436e-03, + -4.0474e-03, -6.8130e-03]], device='cuda:0') +Epoch 334, bias, value: tensor([ 1.1737e-02, 9.0262e-03, -6.0031e-05, 1.3714e-02, -6.5299e-03, + -3.6511e-03, 1.0299e-02, 2.4812e-02, -3.4604e-02, 4.4219e-02], + device='cuda:0'), grad: tensor([ 0.0188, 0.0202, 0.0170, -0.0191, -0.0030, -0.0293, 0.0219, -0.0433, + 0.0260, -0.0092], device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 217.35, cls_loss 0.4616 cls_loss_mapping 0.0040 cls_loss_causal 0.4341 re_mapping 0.0065 re_causal 0.0159 /// teacc 98.67 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.0779, 0.0715, -0.1055, ..., -0.1155, -0.1139, 0.0124], + [-0.0882, -0.1720, 0.0056, ..., -0.0584, -0.0372, -0.1040], + [ 0.0373, -0.0940, 0.0543, ..., 0.1584, -0.1309, -0.0419], + ..., + [-0.1123, -0.1982, 0.0926, ..., -0.0158, -0.0522, 0.0764], + [ 0.0209, 0.0342, -0.0326, ..., -0.0813, -0.1011, -0.0236], + [-0.1540, -0.0073, -0.0064, ..., -0.1696, 0.1164, 0.0222]], + device='cuda:0'), grad: tensor([[-4.5896e-04, -1.3361e-03, 1.0848e-04, ..., 1.3387e-04, + 1.7554e-05, 1.0386e-05], + [-3.6373e-03, 1.9819e-05, 3.1281e-04, ..., 3.1137e-04, + 3.0577e-05, 1.3876e-04], + [ 3.1257e-04, 1.9073e-04, -1.9064e-03, ..., -1.5850e-03, + 1.0848e-05, 1.9640e-05], + ..., + [ 2.4796e-04, 2.9504e-05, 1.6851e-03, ..., 1.4620e-03, + -2.4334e-05, -2.7514e-04], + [ 1.2245e-03, 8.5163e-04, 1.1277e-04, ..., 1.1361e-04, + 3.4690e-04, 1.1645e-05], + [ 4.2224e-04, 5.7936e-04, 3.7241e-04, ..., 4.0627e-04, + 5.7936e-05, 1.1313e-04]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0120, 0.0078, 0.0013, 0.0137, -0.0063, -0.0034, 0.0106, 0.0246, + -0.0350, 0.0437], device='cuda:0'), grad: tensor([ 0.0027, -0.0205, 0.0050, 0.0065, -0.0011, 0.0090, 0.0078, 0.0053, + 0.0107, -0.0253], device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 216.85, cls_loss 0.4804 cls_loss_mapping 0.0031 cls_loss_causal 0.4606 re_mapping 0.0063 re_causal 0.0161 /// teacc 98.77 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.0775, 0.0720, -0.1058, ..., -0.1147, -0.1153, 0.0116], + [-0.0883, -0.1741, 0.0062, ..., -0.0584, -0.0378, -0.1035], + [ 0.0366, -0.0953, 0.0530, ..., 0.1573, -0.1317, -0.0418], + ..., + [-0.1122, -0.1980, 0.0934, ..., -0.0158, -0.0514, 0.0757], + [ 0.0202, 0.0335, -0.0318, ..., -0.0810, -0.1011, -0.0244], + [-0.1535, -0.0066, -0.0064, ..., -0.1695, 0.1168, 0.0245]], + device='cuda:0'), grad: tensor([[ 2.1858e-03, 1.9207e-03, -6.8808e-04, ..., 3.6210e-06, + 2.0843e-06, 3.2926e-04], + [ 4.7565e-05, 1.7011e-04, -3.0208e-04, ..., 4.5133e-04, + 8.8573e-05, -1.1759e-03], + [ 3.6764e-04, 5.8651e-04, 6.3801e-04, ..., 7.6175e-05, + 6.4373e-05, 9.4843e-04], + ..., + [ 1.8194e-05, 1.2255e-04, -2.5902e-03, ..., -6.8617e-04, + -6.4492e-05, -3.3989e-03], + [ 9.7179e-04, 1.0653e-03, 3.3927e-04, ..., 3.8117e-05, + 3.9101e-05, 5.1641e-04], + [ 7.3612e-05, -7.1764e-04, 3.3069e-04, ..., 1.5903e-04, + -9.4700e-04, 5.6458e-04]], device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0121, 0.0080, 0.0007, 0.0129, -0.0073, -0.0043, 0.0115, 0.0254, + -0.0348, 0.0447], device='cuda:0'), grad: tensor([-0.0122, 0.0027, 0.0204, 0.0021, -0.0132, 0.0161, -0.0212, 0.0315, + -0.0142, -0.0119], device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 217.17, cls_loss 0.4494 cls_loss_mapping 0.0031 cls_loss_causal 0.4264 re_mapping 0.0068 re_causal 0.0173 /// teacc 98.87 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.0761, 0.0722, -0.1073, ..., -0.1143, -0.1165, 0.0108], + [-0.0876, -0.1734, 0.0064, ..., -0.0589, -0.0368, -0.1035], + [ 0.0366, -0.0959, 0.0549, ..., 0.1579, -0.1316, -0.0402], + ..., + [-0.1131, -0.1990, 0.0927, ..., -0.0169, -0.0523, 0.0757], + [ 0.0202, 0.0336, -0.0306, ..., -0.0803, -0.0994, -0.0232], + [-0.1541, -0.0078, -0.0064, ..., -0.1703, 0.1164, 0.0246]], + device='cuda:0'), grad: tensor([[ 4.0793e-04, 6.0201e-05, 5.2452e-05, ..., 3.4618e-04, + 2.3954e-06, -3.0923e-04], + [ 7.0190e-04, 5.3644e-05, 4.0746e-04, ..., 5.8937e-04, + 2.9492e-04, 4.3941e-04], + [-2.5986e-02, -1.6373e-02, -3.3447e-02, ..., -8.3466e-03, + -1.4448e-04, -6.2084e-04], + ..., + [ 3.8934e-04, 3.3021e-05, 3.4308e-04, ..., 2.0039e-04, + 4.3607e-04, 7.5579e-05], + [ 2.3627e-04, -1.6861e-03, 5.0545e-04, ..., 7.6818e-04, + -4.0169e-03, -3.5934e-03], + [-4.5151e-05, 3.3706e-05, -5.1051e-05, ..., 1.5700e-04, + -2.5725e-04, 1.6665e-04]], device='cuda:0') +Epoch 337, bias, value: tensor([ 0.0120, 0.0098, 0.0009, 0.0119, -0.0076, -0.0043, 0.0110, 0.0238, + -0.0328, 0.0442], device='cuda:0'), grad: tensor([-0.0418, 0.0376, -0.0439, 0.0224, 0.0284, -0.0408, 0.0007, 0.0173, + 0.0076, 0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 216.83, cls_loss 0.4597 cls_loss_mapping 0.0035 cls_loss_causal 0.4359 re_mapping 0.0066 re_causal 0.0161 /// teacc 98.65 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.0772, 0.0724, -0.1064, ..., -0.1145, -0.1161, 0.0116], + [-0.0884, -0.1732, 0.0060, ..., -0.0592, -0.0360, -0.1031], + [ 0.0354, -0.0969, 0.0555, ..., 0.1573, -0.1315, -0.0410], + ..., + [-0.1128, -0.1993, 0.0936, ..., -0.0159, -0.0517, 0.0762], + [ 0.0209, 0.0354, -0.0319, ..., -0.0816, -0.0999, -0.0235], + [-0.1542, -0.0091, -0.0070, ..., -0.1706, 0.1156, 0.0236]], + device='cuda:0'), grad: tensor([[ 3.1948e-04, -1.6570e-04, 5.9223e-04, ..., -1.2837e-05, + 4.1630e-07, 2.2292e-04], + [ 1.0169e-04, 1.0794e-04, -1.2512e-03, ..., 2.3050e-07, + 3.1432e-07, -3.8128e-03], + [ 3.2139e-04, 3.2902e-04, 3.3989e-03, ..., 3.3248e-07, + 6.7754e-07, 4.4847e-04], + ..., + [-2.5439e-04, -2.2805e-04, 4.1847e-03, ..., 3.6024e-06, + 6.9141e-06, 8.5878e-04], + [ 4.0531e-04, 4.3154e-04, -1.3649e-02, ..., 9.7416e-07, + 2.8774e-05, 3.7026e-04], + [ 3.1137e-04, 3.7527e-04, 2.7714e-03, ..., 3.6329e-05, + 5.0038e-05, 5.4836e-04]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0118, 0.0095, 0.0009, 0.0133, -0.0070, -0.0051, 0.0113, 0.0236, + -0.0332, 0.0439], device='cuda:0'), grad: tensor([ 0.0041, -0.0325, 0.0109, 0.0135, 0.0047, -0.0110, 0.0097, 0.0005, + -0.0096, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 216.71, cls_loss 0.4976 cls_loss_mapping 0.0039 cls_loss_causal 0.4774 re_mapping 0.0061 re_causal 0.0146 /// teacc 98.68 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.0783, 0.0722, -0.1075, ..., -0.1153, -0.1173, 0.0106], + [-0.0878, -0.1726, 0.0069, ..., -0.0595, -0.0369, -0.1039], + [ 0.0369, -0.0954, 0.0561, ..., 0.1576, -0.1313, -0.0400], + ..., + [-0.1123, -0.2006, 0.0927, ..., -0.0163, -0.0522, 0.0761], + [ 0.0204, 0.0351, -0.0326, ..., -0.0816, -0.0999, -0.0232], + [-0.1534, -0.0077, -0.0061, ..., -0.1683, 0.1158, 0.0243]], + device='cuda:0'), grad: tensor([[-1.6624e-06, 1.4503e-02, 4.4465e-04, ..., 2.1410e-04, + 3.5501e-04, 7.8821e-04], + [ 1.0973e-04, 2.1756e-04, 6.5708e-04, ..., -1.4381e-03, + 3.0780e-04, 1.1005e-03], + [ 3.3760e-04, 4.3845e-04, 6.3705e-04, ..., 3.2878e-04, + 6.5374e-04, 1.3504e-03], + ..., + [ 7.7069e-05, 9.1743e-04, 1.1740e-03, ..., 2.5678e-04, + 8.9836e-04, -4.0550e-03], + [-7.4625e-05, -1.4885e-02, -1.5297e-03, ..., -1.0371e-04, + -1.7334e-02, -7.5073e-03], + [ 7.2956e-04, -8.1587e-04, -2.8667e-03, ..., 6.9284e-04, + 1.0643e-02, 7.5760e-03]], device='cuda:0') +Epoch 339, bias, value: tensor([ 0.0118, 0.0089, 0.0024, 0.0126, -0.0082, -0.0053, 0.0111, 0.0241, + -0.0332, 0.0448], device='cuda:0'), grad: tensor([ 0.0457, -0.0434, 0.0240, 0.0186, 0.0045, 0.0212, -0.0105, -0.0022, + -0.0594, 0.0013], device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 216.54, cls_loss 0.4683 cls_loss_mapping 0.0027 cls_loss_causal 0.4496 re_mapping 0.0065 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.0770, 0.0718, -0.1069, ..., -0.1155, -0.1171, 0.0115], + [-0.0887, -0.1710, 0.0068, ..., -0.0587, -0.0365, -0.1043], + [ 0.0372, -0.0944, 0.0569, ..., 0.1576, -0.1314, -0.0400], + ..., + [-0.1130, -0.2020, 0.0928, ..., -0.0163, -0.0528, 0.0757], + [ 0.0203, 0.0347, -0.0337, ..., -0.0815, -0.1000, -0.0242], + [-0.1543, -0.0091, -0.0064, ..., -0.1679, 0.1159, 0.0252]], + device='cuda:0'), grad: tensor([[ 3.1471e-04, -1.5755e-03, 1.1033e-04, ..., 5.3048e-05, + 3.6895e-05, 1.0973e-04], + [ 8.6948e-06, 1.0192e-04, 7.1943e-05, ..., 1.0338e-03, + 7.2956e-04, -1.1623e-04], + [ 8.1158e-04, 4.4608e-04, 4.5031e-05, ..., 4.2462e-04, + 2.9421e-04, 4.1634e-05], + ..., + [ 2.6263e-06, -4.1556e-04, 4.9257e-04, ..., 8.6665e-05, + 6.3658e-04, 2.2566e-04], + [ 5.1022e-04, 1.5087e-03, 2.1076e-04, ..., 2.8396e-04, + 3.4189e-04, 2.1601e-04], + [ 3.1471e-05, -3.5620e-04, -1.2913e-03, ..., -3.0022e-03, + -2.9316e-03, -9.0790e-04]], device='cuda:0') +Epoch 340, bias, value: tensor([ 0.0128, 0.0089, 0.0022, 0.0126, -0.0088, -0.0058, 0.0111, 0.0245, + -0.0330, 0.0445], device='cuda:0'), grad: tensor([-0.0095, -0.0128, 0.0070, 0.0125, 0.0213, -0.0078, 0.0123, -0.0141, + 0.0159, -0.0247], device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 216.77, cls_loss 0.4778 cls_loss_mapping 0.0033 cls_loss_causal 0.4464 re_mapping 0.0064 re_causal 0.0155 /// teacc 98.88 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.0763, 0.0722, -0.1065, ..., -0.1151, -0.1161, 0.0139], + [-0.0882, -0.1710, 0.0070, ..., -0.0586, -0.0364, -0.1038], + [ 0.0376, -0.0940, 0.0568, ..., 0.1574, -0.1324, -0.0407], + ..., + [-0.1124, -0.2029, 0.0933, ..., -0.0165, -0.0528, 0.0761], + [ 0.0208, 0.0353, -0.0335, ..., -0.0814, -0.1002, -0.0233], + [-0.1549, -0.0093, -0.0074, ..., -0.1685, 0.1159, 0.0244]], + device='cuda:0'), grad: tensor([[ 3.2949e-04, 2.0123e-03, 1.3554e-04, ..., 6.2525e-05, + 4.3750e-04, 5.5194e-05], + [ 1.8120e-04, 2.3525e-06, 1.6890e-03, ..., 3.6526e-04, + 4.0460e-04, 8.0347e-04], + [ 2.8896e-04, 2.6062e-05, -1.2035e-03, ..., -2.2259e-03, + 3.6049e-04, 1.5128e-04], + ..., + [-6.6338e-03, 1.6972e-05, -4.1084e-03, ..., 3.6335e-04, + -4.9515e-03, -2.5120e-03], + [ 5.2881e-04, 2.4885e-05, 1.0528e-05, ..., 8.9347e-05, + 5.6934e-04, 1.0872e-04], + [ 3.6311e-04, -6.8331e-04, 6.1321e-04, ..., 3.8767e-04, + -1.0309e-03, 1.8525e-04]], device='cuda:0') +Epoch 341, bias, value: tensor([ 0.0146, 0.0088, 0.0014, 0.0117, -0.0081, -0.0057, 0.0104, 0.0242, + -0.0327, 0.0445], device='cuda:0'), grad: tensor([ 0.0228, 0.0283, 0.0069, -0.0150, -0.0107, 0.0236, -0.0263, -0.0386, + 0.0150, -0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 216.87, cls_loss 0.4670 cls_loss_mapping 0.0024 cls_loss_causal 0.4356 re_mapping 0.0059 re_causal 0.0148 /// teacc 98.72 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.0761, 0.0727, -0.1068, ..., -0.1150, -0.1169, 0.0132], + [-0.0886, -0.1714, 0.0084, ..., -0.0578, -0.0355, -0.1028], + [ 0.0380, -0.0953, 0.0566, ..., 0.1569, -0.1320, -0.0413], + ..., + [-0.1131, -0.2040, 0.0932, ..., -0.0159, -0.0526, 0.0763], + [ 0.0204, 0.0356, -0.0348, ..., -0.0811, -0.0998, -0.0238], + [-0.1550, -0.0095, -0.0073, ..., -0.1685, 0.1156, 0.0249]], + device='cuda:0'), grad: tensor([[ 1.4370e-06, 2.0802e-05, 1.3590e-04, ..., 1.1027e-04, + 4.9025e-05, -2.2182e-03], + [ 2.4199e-05, 1.0002e-04, -5.8289e-03, ..., -8.8549e-04, + 5.4032e-05, 7.9584e-04], + [-8.9407e-05, 6.1333e-05, -4.5800e-04, ..., -5.1451e-04, + 4.0293e-05, -2.7132e-04], + ..., + [ 4.7863e-05, 3.8803e-05, 4.0665e-03, ..., 3.9911e-04, + 3.7432e-05, 9.8419e-04], + [ 6.8367e-05, 1.7273e-04, 1.2293e-03, ..., 5.0545e-04, + -5.1165e-04, -3.9368e-03], + [ 8.3894e-06, 2.6464e-05, 8.9025e-04, ..., 1.6856e-04, + 1.2410e-04, 9.6369e-04]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0125, 0.0095, 0.0017, 0.0116, -0.0081, -0.0058, 0.0107, 0.0245, + -0.0324, 0.0446], device='cuda:0'), grad: tensor([-0.0155, -0.0216, 0.0134, -0.0118, 0.0168, -0.0149, -0.0144, 0.0334, + -0.0059, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 217.04, cls_loss 0.4799 cls_loss_mapping 0.0027 cls_loss_causal 0.4558 re_mapping 0.0062 re_causal 0.0158 /// teacc 98.89 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.0769, 0.0722, -0.1074, ..., -0.1152, -0.1173, 0.0131], + [-0.0881, -0.1739, 0.0081, ..., -0.0578, -0.0357, -0.1021], + [ 0.0381, -0.0944, 0.0563, ..., 0.1566, -0.1320, -0.0418], + ..., + [-0.1140, -0.2040, 0.0936, ..., -0.0162, -0.0521, 0.0764], + [ 0.0208, 0.0354, -0.0342, ..., -0.0812, -0.0996, -0.0239], + [-0.1548, -0.0101, -0.0086, ..., -0.1675, 0.1155, 0.0245]], + device='cuda:0'), grad: tensor([[ 3.8314e-04, -1.1015e-03, -2.4052e-03, ..., 3.1143e-05, + 4.3941e-04, -7.9966e-04], + [ 7.9870e-05, -1.3285e-03, 4.9019e-04, ..., 6.4421e-04, + -3.8886e-04, 6.8188e-04], + [ 1.3647e-03, 8.4686e-04, -3.4142e-04, ..., -1.0176e-03, + 5.4026e-04, -6.2943e-04], + ..., + [-5.3167e-05, 5.3740e-04, 3.1757e-04, ..., 2.9707e-04, + 3.0184e-04, 9.0218e-04], + [ 5.4836e-04, 1.2207e-03, 4.3797e-04, ..., 1.0133e-04, + 9.3269e-04, 1.0529e-03], + [ 4.8018e-04, -5.0068e-04, -1.8895e-04, ..., 8.5711e-05, + -7.8201e-04, -2.9778e-04]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0128, 0.0090, 0.0018, 0.0109, -0.0083, -0.0047, 0.0115, 0.0250, + -0.0332, 0.0442], device='cuda:0'), grad: tensor([ 0.0113, -0.0189, -0.0402, 0.0035, 0.0240, -0.0334, 0.0057, 0.0257, + 0.0324, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 217.18, cls_loss 0.4857 cls_loss_mapping 0.0030 cls_loss_causal 0.4561 re_mapping 0.0063 re_causal 0.0166 /// teacc 98.87 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.0762, 0.0715, -0.1077, ..., -0.1158, -0.1183, 0.0119], + [-0.0871, -0.1740, 0.0086, ..., -0.0583, -0.0362, -0.1036], + [ 0.0369, -0.0955, 0.0571, ..., 0.1572, -0.1324, -0.0418], + ..., + [-0.1136, -0.2037, 0.0932, ..., -0.0163, -0.0510, 0.0767], + [ 0.0197, 0.0355, -0.0348, ..., -0.0812, -0.1017, -0.0221], + [-0.1547, -0.0094, -0.0074, ..., -0.1688, 0.1160, 0.0256]], + device='cuda:0'), grad: tensor([[ 3.1948e-03, -4.9896e-03, -3.3264e-03, ..., -4.0283e-03, + -3.2711e-03, 2.5787e-03], + [ 2.0832e-05, 6.7234e-05, 9.9754e-04, ..., 1.5545e-03, + -1.2417e-03, -6.2103e-03], + [ 3.8099e-04, 3.6097e-04, 5.0735e-04, ..., 4.1747e-04, + 2.3532e-04, 6.6805e-04], + ..., + [ 6.8545e-06, 8.7842e-06, 3.6073e-04, ..., 4.1890e-04, + 7.5483e-04, -1.7366e-03], + [ 5.4073e-04, 1.1740e-03, 2.8419e-04, ..., 2.6631e-04, + 1.9634e-04, 6.1321e-04], + [ 3.1859e-05, 4.2486e-04, 1.3552e-03, ..., 5.4550e-04, + 1.5478e-03, 4.4861e-03]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0125, 0.0095, 0.0031, 0.0105, -0.0084, -0.0045, 0.0113, 0.0239, + -0.0330, 0.0441], device='cuda:0'), grad: tensor([-0.0285, -0.0076, -0.0172, 0.0180, -0.0125, 0.0381, -0.0284, 0.0125, + 0.0114, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 216.85, cls_loss 0.4529 cls_loss_mapping 0.0029 cls_loss_causal 0.4350 re_mapping 0.0065 re_causal 0.0159 /// teacc 98.94 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.0766, 0.0723, -0.1063, ..., -0.1157, -0.1177, 0.0120], + [-0.0875, -0.1744, 0.0071, ..., -0.0589, -0.0345, -0.1037], + [ 0.0368, -0.0954, 0.0577, ..., 0.1575, -0.1301, -0.0405], + ..., + [-0.1118, -0.2026, 0.0925, ..., -0.0160, -0.0518, 0.0768], + [ 0.0208, 0.0356, -0.0360, ..., -0.0820, -0.1018, -0.0222], + [-0.1551, -0.0095, -0.0070, ..., -0.1689, 0.1159, 0.0240]], + device='cuda:0'), grad: tensor([[ 5.9605e-04, 2.2984e-03, 9.0075e-04, ..., 1.0958e-03, + 4.4405e-05, 4.5276e-04], + [ 4.8590e-04, 2.0123e-04, 1.4715e-03, ..., 9.7942e-04, + 7.2718e-04, 1.5965e-03], + [ 7.3719e-04, -6.3934e-03, -2.5349e-03, ..., -2.7657e-03, + 7.5638e-05, 7.1716e-04], + ..., + [ 5.7697e-04, 2.3520e-04, -8.0109e-03, ..., -5.8899e-03, + -4.1885e-03, -5.6763e-03], + [ 1.6708e-03, 9.0885e-04, 4.2953e-03, ..., 3.9635e-03, + 2.0046e-03, 3.7575e-03], + [ 5.1117e-04, 2.5010e-04, 9.9182e-04, ..., 7.6008e-04, + 5.1022e-04, 1.4219e-03]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0125, 0.0089, 0.0030, 0.0103, -0.0073, -0.0049, 0.0119, 0.0228, + -0.0325, 0.0442], device='cuda:0'), grad: tensor([ 0.0190, 0.0224, 0.0095, -0.0103, 0.0256, -0.0367, -0.0449, -0.0067, + 0.0018, 0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 216.85, cls_loss 0.4881 cls_loss_mapping 0.0047 cls_loss_causal 0.4624 re_mapping 0.0057 re_causal 0.0145 /// teacc 98.78 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.0759, 0.0727, -0.1074, ..., -0.1165, -0.1177, 0.0132], + [-0.0878, -0.1751, 0.0072, ..., -0.0584, -0.0357, -0.1035], + [ 0.0359, -0.0970, 0.0571, ..., 0.1567, -0.1298, -0.0406], + ..., + [-0.1115, -0.2022, 0.0916, ..., -0.0155, -0.0505, 0.0769], + [ 0.0219, 0.0356, -0.0344, ..., -0.0814, -0.1007, -0.0215], + [-0.1564, -0.0112, -0.0061, ..., -0.1699, 0.1143, 0.0231]], + device='cuda:0'), grad: tensor([[ 2.8300e-04, 1.9360e-04, 3.6502e-04, ..., 9.9778e-05, + 4.3988e-04, 3.5357e-04], + [ 2.1243e-04, 1.5724e-04, 1.1692e-03, ..., 2.7061e-04, + 1.2302e-03, 1.2665e-03], + [ 3.0518e-04, 2.8396e-04, 1.8139e-03, ..., 9.0981e-04, + 8.5640e-04, 1.0929e-03], + ..., + [-1.0824e-03, -3.8314e-04, 9.8419e-03, ..., 9.7046e-03, + -3.0289e-03, -9.8705e-04], + [-1.9007e-03, -3.4046e-03, 4.7326e-04, ..., 3.4213e-04, + -7.0267e-03, -2.6493e-03], + [ 2.4414e-03, 2.8057e-03, -1.2848e-02, ..., -1.2421e-02, + 6.0349e-03, 2.1076e-03]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0123, 0.0090, 0.0028, 0.0103, -0.0070, -0.0053, 0.0121, 0.0225, + -0.0321, 0.0443], device='cuda:0'), grad: tensor([ 0.0061, 0.0125, 0.0121, 0.0430, 0.0100, -0.0584, -0.0231, 0.0029, + -0.0166, 0.0116], device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 217.32, cls_loss 0.4830 cls_loss_mapping 0.0033 cls_loss_causal 0.4590 re_mapping 0.0060 re_causal 0.0152 /// teacc 98.90 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.0761, 0.0728, -0.1108, ..., -0.1170, -0.1180, 0.0136], + [-0.0879, -0.1755, 0.0061, ..., -0.0595, -0.0362, -0.1034], + [ 0.0361, -0.0971, 0.0570, ..., 0.1571, -0.1302, -0.0405], + ..., + [-0.1114, -0.2023, 0.0933, ..., -0.0152, -0.0506, 0.0773], + [ 0.0225, 0.0364, -0.0349, ..., -0.0815, -0.1022, -0.0236], + [-0.1561, -0.0112, -0.0047, ..., -0.1692, 0.1154, 0.0246]], + device='cuda:0'), grad: tensor([[ 2.4629e-04, 5.7906e-05, 6.2275e-04, ..., 1.1981e-04, + 1.1225e-03, 7.2670e-04], + [ 2.2912e-04, 4.7386e-05, 5.2738e-04, ..., 8.7380e-05, + 1.2646e-03, 8.4066e-04], + [ 4.1962e-04, 1.0830e-04, 9.4891e-04, ..., 7.6115e-05, + 1.4896e-03, 8.3923e-04], + ..., + [ 1.1206e-03, 7.0572e-04, 2.7962e-03, ..., 1.6677e-04, + 1.5335e-03, -3.2768e-03], + [ 5.1641e-04, 6.7592e-05, -3.5038e-03, ..., -1.5011e-03, + -4.0588e-03, -3.9597e-03], + [-1.3137e-04, -8.2111e-04, -1.3666e-03, ..., 3.1352e-04, + -3.6030e-03, 1.3056e-03]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0123, 0.0088, 0.0032, 0.0100, -0.0072, -0.0041, 0.0113, 0.0221, + -0.0328, 0.0455], device='cuda:0'), grad: tensor([ 0.0119, 0.0148, 0.0146, 0.0169, -0.0132, -0.0141, 0.0119, -0.0065, + -0.0154, -0.0208], device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 216.51, cls_loss 0.4865 cls_loss_mapping 0.0039 cls_loss_causal 0.4591 re_mapping 0.0061 re_causal 0.0152 /// teacc 98.89 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.0759, 0.0724, -0.1108, ..., -0.1166, -0.1189, 0.0137], + [-0.0885, -0.1764, 0.0073, ..., -0.0589, -0.0372, -0.1038], + [ 0.0362, -0.0969, 0.0571, ..., 0.1580, -0.1308, -0.0399], + ..., + [-0.1117, -0.2036, 0.0930, ..., -0.0152, -0.0497, 0.0780], + [ 0.0212, 0.0366, -0.0345, ..., -0.0813, -0.1021, -0.0245], + [-0.1560, -0.0098, -0.0043, ..., -0.1692, 0.1156, 0.0260]], + device='cuda:0'), grad: tensor([[ 1.0335e-04, 3.9792e-04, 1.6012e-03, ..., 1.3685e-04, + 7.3195e-04, 9.4652e-04], + [ 5.6744e-04, -1.3542e-03, -2.8515e-03, ..., 1.6189e-04, + 7.6580e-04, 1.4486e-03], + [ 4.5128e-03, 3.1209e-04, 2.0504e-03, ..., 3.8195e-04, + 1.3046e-03, 2.6627e-03], + ..., + [-2.5439e-04, 1.3173e-04, -2.3918e-03, ..., 9.9242e-05, + 6.1417e-04, 1.0490e-03], + [-1.1383e-02, -1.6272e-04, 6.0052e-05, ..., -8.5592e-05, + 3.8242e-04, -2.5158e-03], + [-8.3983e-05, 2.4402e-04, -1.6527e-03, ..., 1.4496e-04, + -5.7297e-03, -1.7242e-03]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0120, 0.0088, 0.0031, 0.0116, -0.0072, -0.0057, 0.0108, 0.0222, + -0.0329, 0.0462], device='cuda:0'), grad: tensor([ 0.0156, -0.0101, 0.0280, -0.0373, -0.0029, 0.0226, 0.0252, 0.0003, + -0.0217, -0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 216.94, cls_loss 0.4598 cls_loss_mapping 0.0030 cls_loss_causal 0.4379 re_mapping 0.0062 re_causal 0.0156 /// teacc 98.75 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.0752, 0.0726, -0.1125, ..., -0.1166, -0.1207, 0.0128], + [-0.0878, -0.1766, 0.0088, ..., -0.0578, -0.0373, -0.1037], + [ 0.0358, -0.0969, 0.0557, ..., 0.1581, -0.1317, -0.0393], + ..., + [-0.1120, -0.2042, 0.0934, ..., -0.0151, -0.0499, 0.0774], + [ 0.0214, 0.0364, -0.0358, ..., -0.0825, -0.1026, -0.0247], + [-0.1571, -0.0094, -0.0046, ..., -0.1699, 0.1155, 0.0271]], + device='cuda:0'), grad: tensor([[ 4.3344e-04, 4.9171e-03, -7.6580e-04, ..., 3.7342e-05, + -2.8667e-03, -2.9507e-03], + [ 9.2536e-06, 9.0599e-05, -5.5275e-03, ..., -5.0888e-03, + -7.7295e-04, -3.5801e-03], + [ 3.2163e-04, 2.8086e-04, 5.8136e-03, ..., 5.2109e-03, + 4.7541e-04, 2.1076e-03], + ..., + [ 1.3418e-05, 7.3731e-05, 5.3978e-04, ..., 5.0640e-04, + 5.7459e-04, 8.8882e-04], + [-1.5080e-04, 1.4281e-04, 1.8573e-04, ..., 1.3769e-05, + 5.1308e-04, 6.6280e-04], + [ 1.0210e-04, 2.0933e-04, 5.3072e-04, ..., 2.8658e-04, + 1.0242e-03, 1.2960e-03]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0119, 0.0091, 0.0023, 0.0128, -0.0069, -0.0071, 0.0114, 0.0219, + -0.0331, 0.0465], device='cuda:0'), grad: tensor([-0.0099, -0.0324, 0.0253, 0.0011, 0.0131, 0.0222, -0.0238, 0.0117, + 0.0087, -0.0161], device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 216.34, cls_loss 0.4767 cls_loss_mapping 0.0035 cls_loss_causal 0.4517 re_mapping 0.0058 re_causal 0.0151 /// teacc 98.78 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.0757, 0.0727, -0.1121, ..., -0.1166, -0.1201, 0.0132], + [-0.0872, -0.1764, 0.0075, ..., -0.0577, -0.0374, -0.1047], + [ 0.0358, -0.0972, 0.0569, ..., 0.1589, -0.1319, -0.0391], + ..., + [-0.1130, -0.2037, 0.0931, ..., -0.0162, -0.0504, 0.0775], + [ 0.0218, 0.0376, -0.0355, ..., -0.0829, -0.1016, -0.0251], + [-0.1560, -0.0087, -0.0048, ..., -0.1700, 0.1162, 0.0276]], + device='cuda:0'), grad: tensor([[-4.8447e-03, -6.3133e-03, 6.0052e-05, ..., 1.0496e-04, + 1.1283e-04, 9.4593e-05], + [ 7.5579e-05, 5.1916e-05, 1.4889e-04, ..., 6.5446e-05, + -3.8147e-04, 3.5739e-04], + [ 6.6662e-04, 7.0763e-04, 6.0606e-04, ..., 5.8126e-04, + 4.1056e-04, 5.4789e-04], + ..., + [ 4.9114e-04, 3.7670e-05, 2.1000e-03, ..., 1.4508e-04, + 2.3937e-04, 1.9407e-04], + [ 9.5940e-04, 2.3327e-03, -3.5667e-03, ..., 2.0969e-04, + 2.6393e-04, 2.2340e-04], + [ 1.6117e-04, 1.7583e-04, 3.7041e-03, ..., 1.2589e-04, + 7.1831e-03, 5.4626e-03]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0126, 0.0086, 0.0029, 0.0128, -0.0074, -0.0064, 0.0121, 0.0217, + -0.0332, 0.0453], device='cuda:0'), grad: tensor([-0.0071, -0.0069, 0.0268, 0.0158, -0.0286, -0.0768, 0.0233, 0.0170, + 0.0123, 0.0241], device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 216.50, cls_loss 0.4933 cls_loss_mapping 0.0029 cls_loss_causal 0.4680 re_mapping 0.0060 re_causal 0.0153 /// teacc 98.82 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.0752, 0.0727, -0.1127, ..., -0.1163, -0.1209, 0.0135], + [-0.0870, -0.1773, 0.0078, ..., -0.0581, -0.0385, -0.1059], + [ 0.0347, -0.0988, 0.0561, ..., 0.1583, -0.1320, -0.0395], + ..., + [-0.1138, -0.2042, 0.0927, ..., -0.0152, -0.0512, 0.0777], + [ 0.0227, 0.0376, -0.0361, ..., -0.0821, -0.1010, -0.0248], + [-0.1560, -0.0075, -0.0028, ..., -0.1702, 0.1168, 0.0271]], + device='cuda:0'), grad: tensor([[ 3.1543e-04, 2.0275e-03, 5.3692e-04, ..., 8.6844e-05, + 2.9349e-04, 1.5278e-03], + [ 3.3116e-04, 6.1369e-04, 1.0815e-03, ..., 5.3978e-04, + 1.5678e-03, 1.2636e-03], + [-2.1076e-03, -1.3800e-03, -3.2558e-03, ..., -3.0575e-03, + -2.9812e-03, -9.3985e-04], + ..., + [ 2.3651e-04, 1.9693e-04, -2.6112e-03, ..., -7.3731e-05, + 2.0123e-04, -2.4872e-03], + [ 1.9512e-03, 2.0474e-05, 3.0575e-03, ..., 2.3365e-03, + 6.7997e-04, 1.8702e-03], + [ 3.7575e-04, 4.1389e-04, 1.2484e-03, ..., 4.1366e-04, + 9.9754e-04, 1.2712e-03]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0122, 0.0081, 0.0032, 0.0129, -0.0084, -0.0053, 0.0122, 0.0213, + -0.0332, 0.0460], device='cuda:0'), grad: tensor([ 0.0167, -0.0042, -0.0073, 0.0231, -0.0043, 0.0105, -0.0031, -0.0620, + 0.0105, 0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 216.44, cls_loss 0.5014 cls_loss_mapping 0.0028 cls_loss_causal 0.4735 re_mapping 0.0063 re_causal 0.0160 /// teacc 98.68 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.0759, 0.0713, -0.1109, ..., -0.1152, -0.1192, 0.0140], + [-0.0868, -0.1777, 0.0083, ..., -0.0586, -0.0388, -0.1072], + [ 0.0329, -0.0990, 0.0570, ..., 0.1592, -0.1330, -0.0397], + ..., + [-0.1140, -0.2045, 0.0923, ..., -0.0148, -0.0518, 0.0785], + [ 0.0222, 0.0380, -0.0373, ..., -0.0834, -0.1015, -0.0240], + [-0.1556, -0.0061, -0.0034, ..., -0.1722, 0.1162, 0.0262]], + device='cuda:0'), grad: tensor([[ 3.5739e-04, 4.9162e-04, 5.0831e-04, ..., 7.2420e-05, + 1.1225e-03, 2.6169e-03], + [ 4.6611e-05, 4.2111e-05, 4.9448e-04, ..., 1.7750e-04, + -7.6437e-04, 2.7313e-03], + [ 5.1260e-04, 5.9128e-04, 7.8011e-04, ..., 2.7132e-04, + 1.3170e-03, 3.4904e-03], + ..., + [ 7.2420e-05, 6.3837e-05, -1.2741e-03, ..., 1.6522e-04, + -2.0695e-03, -2.4452e-03], + [ 2.5105e-04, 3.0613e-04, 7.0333e-04, ..., 1.8585e-04, + 1.6785e-03, 2.6779e-03], + [ 4.7952e-05, -5.3501e-04, 5.6267e-04, ..., 9.7930e-05, + 2.0409e-03, 4.1389e-03]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0116, 0.0082, 0.0028, 0.0140, -0.0068, -0.0060, 0.0114, 0.0218, + -0.0326, 0.0446], device='cuda:0'), grad: tensor([ 0.0162, -0.0010, 0.0257, 0.0280, 0.0170, -0.0416, -0.0847, -0.0066, + 0.0201, 0.0269], device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 216.78, cls_loss 0.4739 cls_loss_mapping 0.0039 cls_loss_causal 0.4490 re_mapping 0.0062 re_causal 0.0159 /// teacc 98.76 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.0760, 0.0723, -0.1104, ..., -0.1149, -0.1180, 0.0142], + [-0.0867, -0.1785, 0.0089, ..., -0.0583, -0.0394, -0.1080], + [ 0.0327, -0.0994, 0.0562, ..., 0.1586, -0.1338, -0.0399], + ..., + [-0.1138, -0.2035, 0.0929, ..., -0.0145, -0.0514, 0.0789], + [ 0.0226, 0.0383, -0.0365, ..., -0.0829, -0.1006, -0.0244], + [-0.1557, -0.0062, -0.0036, ..., -0.1729, 0.1157, 0.0248]], + device='cuda:0'), grad: tensor([[ 4.3321e-04, 1.7777e-05, 1.4400e-04, ..., 2.1946e-04, + 4.5538e-04, 7.1859e-04], + [ 3.0875e-04, 7.9945e-06, 9.8038e-04, ..., 5.6410e-04, + 8.1491e-04, 1.3132e-03], + [ 1.5469e-03, 6.9737e-05, -5.3864e-03, ..., -4.0932e-03, + 9.8896e-04, 1.1444e-03], + ..., + [ 2.7275e-04, 4.4972e-05, 5.1994e-03, ..., 4.1618e-03, + 8.5878e-04, 4.7731e-04], + [ 4.5128e-03, -1.2743e-04, -7.7581e-04, ..., -7.2145e-04, + 3.4428e-04, 3.9148e-04], + [ 2.5010e-04, -3.9458e-05, -1.3018e-04, ..., 6.7520e-04, + 1.1635e-03, -4.4441e-03]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0114, 0.0078, 0.0030, 0.0133, -0.0076, -0.0067, 0.0123, 0.0221, + -0.0326, 0.0458], device='cuda:0'), grad: tensor([-0.0176, 0.0247, -0.0021, 0.0324, -0.0040, -0.0315, -0.0415, 0.0333, + 0.0173, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 216.61, cls_loss 0.4929 cls_loss_mapping 0.0035 cls_loss_causal 0.4705 re_mapping 0.0062 re_causal 0.0158 /// teacc 98.83 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.0749, 0.0718, -0.1113, ..., -0.1155, -0.1187, 0.0152], + [-0.0863, -0.1786, 0.0088, ..., -0.0581, -0.0388, -0.1094], + [ 0.0343, -0.0998, 0.0556, ..., 0.1590, -0.1358, -0.0400], + ..., + [-0.1160, -0.2041, 0.0936, ..., -0.0141, -0.0512, 0.0791], + [ 0.0234, 0.0389, -0.0366, ..., -0.0834, -0.1010, -0.0243], + [-0.1543, -0.0050, -0.0039, ..., -0.1738, 0.1171, 0.0257]], + device='cuda:0'), grad: tensor([[ 1.2436e-03, 8.9073e-04, 1.3542e-03, ..., 4.3559e-04, + 1.0228e-04, 1.2407e-03], + [ 3.3438e-05, 4.9680e-05, 3.0279e-04, ..., 2.9731e-04, + -9.7561e-04, -3.6335e-03], + [ 3.7994e-03, 1.9989e-03, 1.2197e-03, ..., 2.1629e-03, + 1.3340e-04, 1.3351e-03], + ..., + [ 3.7265e-04, 2.9540e-04, 6.1035e-04, ..., 3.1781e-04, + 3.0205e-05, -1.3561e-03], + [ 2.4109e-03, 1.3266e-03, 7.1716e-04, ..., 1.1625e-03, + 1.2219e-04, 1.1530e-03], + [ 4.2486e-04, 4.3368e-04, 4.7612e-04, ..., 3.1447e-04, + 1.1289e-04, 1.4124e-03]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0115, 0.0075, 0.0025, 0.0133, -0.0067, -0.0061, 0.0117, 0.0224, + -0.0331, 0.0458], device='cuda:0'), grad: tensor([ 0.0240, -0.0130, 0.0266, 0.0127, -0.0404, -0.0139, 0.0060, -0.0425, + 0.0214, 0.0191], device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 216.35, cls_loss 0.4677 cls_loss_mapping 0.0036 cls_loss_causal 0.4444 re_mapping 0.0057 re_causal 0.0146 /// teacc 98.85 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.0761, 0.0708, -0.1109, ..., -0.1158, -0.1193, 0.0150], + [-0.0851, -0.1778, 0.0090, ..., -0.0580, -0.0373, -0.1087], + [ 0.0343, -0.1005, 0.0552, ..., 0.1594, -0.1354, -0.0394], + ..., + [-0.1165, -0.2043, 0.0935, ..., -0.0139, -0.0512, 0.0785], + [ 0.0240, 0.0391, -0.0373, ..., -0.0833, -0.1015, -0.0227], + [-0.1540, -0.0050, -0.0045, ..., -0.1742, 0.1157, 0.0257]], + device='cuda:0'), grad: tensor([[-0.0127, -0.0025, 0.0003, ..., 0.0002, 0.0006, 0.0004], + [ 0.0009, 0.0008, 0.0012, ..., 0.0007, 0.0020, 0.0017], + [-0.0007, -0.0016, -0.0041, ..., -0.0023, -0.0052, -0.0049], + ..., + [ 0.0030, 0.0008, 0.0008, ..., -0.0012, 0.0007, -0.0004], + [ 0.0017, 0.0014, 0.0012, ..., 0.0004, 0.0028, 0.0021], + [ 0.0007, -0.0027, -0.0038, ..., 0.0003, -0.0081, -0.0051]], + device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0106, 0.0085, 0.0021, 0.0141, -0.0074, -0.0062, 0.0107, 0.0235, + -0.0330, 0.0457], device='cuda:0'), grad: tensor([-0.0191, 0.0209, -0.0167, -0.0081, 0.0339, 0.0206, -0.0002, 0.0119, + -0.0069, -0.0363], device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 216.82, cls_loss 0.4904 cls_loss_mapping 0.0027 cls_loss_causal 0.4653 re_mapping 0.0060 re_causal 0.0156 /// teacc 98.84 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.0755, 0.0702, -0.1103, ..., -0.1172, -0.1205, 0.0148], + [-0.0855, -0.1766, 0.0085, ..., -0.0583, -0.0389, -0.1101], + [ 0.0336, -0.1014, 0.0558, ..., 0.1593, -0.1352, -0.0396], + ..., + [-0.1162, -0.2053, 0.0934, ..., -0.0135, -0.0517, 0.0795], + [ 0.0234, 0.0383, -0.0368, ..., -0.0831, -0.1008, -0.0237], + [-0.1541, -0.0052, -0.0043, ..., -0.1744, 0.1166, 0.0256]], + device='cuda:0'), grad: tensor([[ 0.0015, -0.0007, -0.0007, ..., 0.0004, 0.0003, 0.0027], + [ 0.0015, -0.0065, 0.0007, ..., 0.0006, 0.0004, -0.0133], + [ 0.0008, 0.0007, 0.0010, ..., 0.0010, 0.0007, 0.0036], + ..., + [ 0.0011, 0.0004, 0.0027, ..., 0.0021, 0.0016, 0.0023], + [-0.0120, -0.0002, 0.0002, ..., -0.0010, -0.0007, -0.0056], + [ 0.0013, 0.0009, 0.0034, ..., 0.0021, 0.0009, 0.0019]], + device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0106, 0.0088, 0.0010, 0.0141, -0.0080, -0.0068, 0.0128, 0.0236, + -0.0332, 0.0458], device='cuda:0'), grad: tensor([ 0.0132, -0.0401, 0.0199, -0.0056, -0.0345, 0.0158, 0.0236, 0.0230, + -0.0384, 0.0230], device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 216.25, cls_loss 0.4997 cls_loss_mapping 0.0034 cls_loss_causal 0.4753 re_mapping 0.0064 re_causal 0.0155 /// teacc 98.81 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.0756, 0.0710, -0.1097, ..., -0.1169, -0.1208, 0.0162], + [-0.0865, -0.1757, 0.0102, ..., -0.0571, -0.0386, -0.1091], + [ 0.0326, -0.1024, 0.0570, ..., 0.1598, -0.1359, -0.0395], + ..., + [-0.1143, -0.2059, 0.0932, ..., -0.0140, -0.0508, 0.0790], + [ 0.0233, 0.0382, -0.0370, ..., -0.0828, -0.1018, -0.0232], + [-0.1536, -0.0050, -0.0046, ..., -0.1755, 0.1180, 0.0256]], + device='cuda:0'), grad: tensor([[-2.4281e-03, -4.2498e-05, -2.7752e-03, ..., -2.7490e-04, + -1.2245e-03, -4.1122e-03], + [ 1.6296e-04, 6.5044e-06, 9.5129e-04, ..., 2.2650e-04, + 7.3814e-04, 1.0509e-03], + [ 4.8089e-04, -4.8965e-05, 8.9264e-04, ..., 7.8902e-06, + 4.3535e-04, 1.0281e-03], + ..., + [ 3.6860e-04, 8.3372e-06, 9.5442e-06, ..., 4.1795e-04, + 1.3723e-03, 2.0638e-03], + [ 3.2711e-04, 3.9816e-05, 8.7166e-04, ..., 1.5616e-04, + 4.5395e-04, 8.5115e-04], + [ 2.9087e-04, 1.5497e-05, 1.5669e-03, ..., -1.1396e-03, + -1.1206e-03, -4.0436e-03]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0109, 0.0086, 0.0013, 0.0140, -0.0072, -0.0059, 0.0117, 0.0239, + -0.0342, 0.0458], device='cuda:0'), grad: tensor([-0.0243, 0.0127, 0.0132, 0.0136, -0.0192, -0.0197, 0.0068, -0.0101, + 0.0120, 0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 216.92, cls_loss 0.4821 cls_loss_mapping 0.0057 cls_loss_causal 0.4654 re_mapping 0.0063 re_causal 0.0158 /// teacc 98.87 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.0744, 0.0722, -0.1108, ..., -0.1179, -0.1199, 0.0166], + [-0.0877, -0.1758, 0.0110, ..., -0.0571, -0.0384, -0.1097], + [ 0.0322, -0.1023, 0.0562, ..., 0.1597, -0.1365, -0.0378], + ..., + [-0.1138, -0.2067, 0.0937, ..., -0.0142, -0.0507, 0.0782], + [ 0.0243, 0.0378, -0.0374, ..., -0.0834, -0.1022, -0.0237], + [-0.1546, -0.0065, -0.0049, ..., -0.1752, 0.1181, 0.0253]], + device='cuda:0'), grad: tensor([[ 1.6761e-04, 6.3038e-04, 5.3734e-05, ..., 7.8604e-06, + 5.2452e-04, 4.9543e-04], + [-3.7861e-03, -2.9697e-03, 6.1750e-05, ..., -7.2690e-07, + 1.0413e-04, 3.2091e-04], + [-8.8990e-05, 2.5749e-04, -1.2314e-04, ..., -3.2902e-04, + 5.1069e-04, 4.6134e-04], + ..., + [ 1.9753e-04, 2.9016e-04, 8.8632e-05, ..., 1.3560e-06, + 5.7697e-04, 4.8470e-04], + [ 7.2813e-04, 1.3475e-03, 3.3164e-04, ..., 2.5487e-04, + 7.1526e-04, 4.3559e-04], + [ 1.8203e-04, 5.3215e-04, -8.3313e-03, ..., 3.6843e-06, + -4.0100e-02, -3.0548e-02]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0122, 0.0078, 0.0019, 0.0138, -0.0070, -0.0068, 0.0114, 0.0239, + -0.0341, 0.0456], device='cuda:0'), grad: tensor([ 0.0161, -0.0281, 0.0173, 0.0141, 0.0149, -0.0468, -0.0117, 0.0183, + 0.0209, -0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 216.70, cls_loss 0.4983 cls_loss_mapping 0.0042 cls_loss_causal 0.4697 re_mapping 0.0056 re_causal 0.0141 /// teacc 98.88 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.0766, 0.0713, -0.1100, ..., -0.1170, -0.1203, 0.0158], + [-0.0872, -0.1753, 0.0111, ..., -0.0564, -0.0388, -0.1094], + [ 0.0322, -0.1022, 0.0568, ..., 0.1599, -0.1368, -0.0382], + ..., + [-0.1139, -0.2060, 0.0924, ..., -0.0156, -0.0505, 0.0791], + [ 0.0237, 0.0375, -0.0384, ..., -0.0843, -0.1035, -0.0246], + [-0.1547, -0.0065, -0.0040, ..., -0.1756, 0.1188, 0.0252]], + device='cuda:0'), grad: tensor([[-4.0855e-03, -5.0449e-04, -2.4319e-04, ..., 2.4235e-04, + 3.1710e-05, 1.3566e-04], + [ 3.6478e-04, 1.0200e-05, 1.2827e-04, ..., 8.4019e-04, + 1.3225e-05, 1.8275e-04], + [-2.5425e-03, 3.5882e-05, 1.8096e-04, ..., -4.3297e-03, + 3.5226e-05, -7.5455e-03], + ..., + [ 6.4516e-04, 6.0946e-05, -6.2180e-04, ..., 1.4043e-04, + -1.5128e-04, 1.5867e-04], + [ 7.3290e-04, 1.2982e-04, 1.0115e-04, ..., 5.3501e-04, + 1.3351e-04, 1.9121e-04], + [-2.5129e-04, -1.2410e-04, -1.4508e-04, ..., -1.5850e-03, + -1.3041e-04, -1.3056e-03]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0119, 0.0086, 0.0019, 0.0143, -0.0075, -0.0064, 0.0105, 0.0241, + -0.0343, 0.0455], device='cuda:0'), grad: tensor([-0.0519, 0.0168, 0.0041, -0.0076, 0.0253, 0.0208, -0.0134, 0.0135, + 0.0138, -0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 216.98, cls_loss 0.4557 cls_loss_mapping 0.0023 cls_loss_causal 0.4337 re_mapping 0.0063 re_causal 0.0159 /// teacc 98.82 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.0763, 0.0714, -0.1106, ..., -0.1172, -0.1195, 0.0163], + [-0.0882, -0.1750, 0.0103, ..., -0.0563, -0.0381, -0.1094], + [ 0.0323, -0.1021, 0.0572, ..., 0.1606, -0.1364, -0.0370], + ..., + [-0.1136, -0.2057, 0.0938, ..., -0.0155, -0.0508, 0.0787], + [ 0.0237, 0.0370, -0.0385, ..., -0.0843, -0.1039, -0.0258], + [-0.1559, -0.0069, -0.0043, ..., -0.1760, 0.1184, 0.0250]], + device='cuda:0'), grad: tensor([[-1.0757e-03, -5.8861e-03, -6.4964e-03, ..., -3.2234e-03, + -1.0815e-03, -1.5503e-02], + [ 2.9087e-04, 1.6034e-04, 9.7132e-04, ..., 1.0786e-03, + 7.0477e-04, 3.1033e-03], + [ 8.2314e-05, 1.6193e-03, -1.2960e-03, ..., -1.6556e-03, + -3.2501e-03, -3.9864e-03], + ..., + [ 4.0245e-04, 1.4424e-04, -8.4734e-04, ..., 1.1292e-03, + 7.5483e-04, 4.0398e-03], + [ 4.9353e-04, 3.5882e-04, 5.6982e-04, ..., 6.0368e-04, + 4.1771e-04, 3.1300e-03], + [ 8.8787e-04, 1.8177e-03, 2.8992e-03, ..., -2.0313e-03, + 6.2943e-04, 2.3060e-03]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0122, 0.0081, 0.0016, 0.0141, -0.0072, -0.0064, 0.0108, 0.0236, + -0.0338, 0.0457], device='cuda:0'), grad: tensor([-0.0612, -0.0074, -0.0022, 0.0285, 0.0314, 0.0162, -0.0080, 0.0201, + 0.0189, -0.0363], device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 216.58, cls_loss 0.4666 cls_loss_mapping 0.0028 cls_loss_causal 0.4437 re_mapping 0.0061 re_causal 0.0148 /// teacc 98.76 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.0769, 0.0723, -0.1099, ..., -0.1177, -0.1209, 0.0162], + [-0.0884, -0.1752, 0.0094, ..., -0.0571, -0.0373, -0.1093], + [ 0.0312, -0.1029, 0.0564, ..., 0.1595, -0.1366, -0.0377], + ..., + [-0.1133, -0.2071, 0.0929, ..., -0.0149, -0.0524, 0.0768], + [ 0.0238, 0.0366, -0.0376, ..., -0.0841, -0.1044, -0.0268], + [-0.1573, -0.0080, -0.0038, ..., -0.1751, 0.1188, 0.0262]], + device='cuda:0'), grad: tensor([[-3.8185e-03, -8.7357e-03, -6.7711e-03, ..., 1.6065e-07, + -1.2648e-04, -3.5305e-03], + [ 2.2620e-05, 1.5936e-03, 4.0169e-03, ..., -6.4898e-04, + 5.3883e-04, 5.5885e-04], + [ 2.7871e-04, 5.2834e-04, 5.3078e-05, ..., 6.3705e-04, + 1.9300e-04, 5.9175e-04], + ..., + [ 9.9018e-06, -1.0691e-03, 1.4544e-03, ..., 1.3644e-06, + 5.5075e-04, 5.6458e-04], + [ 6.0940e-04, 1.2932e-03, 2.1195e-04, ..., 2.4009e-06, + 3.0470e-04, 9.4891e-04], + [ 4.7326e-05, 5.8031e-04, 1.2569e-03, ..., 3.5483e-07, + 1.3332e-03, 1.2627e-03]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0120, 0.0081, 0.0013, 0.0144, -0.0073, -0.0066, 0.0117, 0.0231, + -0.0337, 0.0458], device='cuda:0'), grad: tensor([-0.0615, 0.0410, 0.0098, 0.0096, -0.0022, -0.0213, 0.0112, 0.0104, + -0.0098, 0.0128], device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 216.36, cls_loss 0.4785 cls_loss_mapping 0.0041 cls_loss_causal 0.4486 re_mapping 0.0062 re_causal 0.0146 /// teacc 98.95 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.0756, 0.0730, -0.1091, ..., -0.1183, -0.1198, 0.0177], + [-0.0882, -0.1750, 0.0078, ..., -0.0582, -0.0379, -0.1110], + [ 0.0314, -0.1030, 0.0555, ..., 0.1594, -0.1368, -0.0386], + ..., + [-0.1127, -0.2090, 0.0926, ..., -0.0157, -0.0545, 0.0771], + [ 0.0242, 0.0375, -0.0369, ..., -0.0840, -0.1041, -0.0280], + [-0.1571, -0.0079, -0.0035, ..., -0.1749, 0.1192, 0.0270]], + device='cuda:0'), grad: tensor([[ 1.2338e-04, 7.7188e-05, 8.3148e-05, ..., 1.1867e-04, + 3.6097e-04, 9.5654e-04], + [ 4.5359e-05, -3.1638e-04, 2.7847e-04, ..., 1.2994e-04, + -7.8201e-04, -1.1606e-03], + [-2.3308e-03, -1.1358e-03, -8.5783e-04, ..., -2.4433e-03, + 4.3201e-04, -1.8778e-03], + ..., + [ 8.6498e-04, 1.1635e-03, 2.3766e-03, ..., 7.0047e-04, + 1.1148e-03, 1.4009e-03], + [ 1.7679e-04, 5.4312e-04, 4.2176e-04, ..., 1.4210e-04, + 8.4257e-04, 1.3857e-03], + [ 4.5002e-05, 1.7917e-04, 9.5654e-04, ..., 1.7023e-04, + 1.5144e-03, 2.5940e-03]], device='cuda:0') +Epoch 362, bias, value: tensor([ 0.0128, 0.0079, 0.0001, 0.0138, -0.0076, -0.0060, 0.0124, 0.0232, + -0.0338, 0.0461], device='cuda:0'), grad: tensor([ 0.0066, -0.0206, -0.0124, 0.0115, -0.0103, -0.0245, 0.0060, 0.0204, + 0.0103, 0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 216.73, cls_loss 0.4645 cls_loss_mapping 0.0039 cls_loss_causal 0.4428 re_mapping 0.0059 re_causal 0.0149 /// teacc 98.79 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.0765, 0.0722, -0.1092, ..., -0.1180, -0.1200, 0.0185], + [-0.0881, -0.1756, 0.0083, ..., -0.0569, -0.0390, -0.1116], + [ 0.0318, -0.1036, 0.0566, ..., 0.1586, -0.1347, -0.0386], + ..., + [-0.1137, -0.2090, 0.0916, ..., -0.0163, -0.0541, 0.0763], + [ 0.0258, 0.0393, -0.0375, ..., -0.0848, -0.1036, -0.0275], + [-0.1562, -0.0074, -0.0036, ..., -0.1762, 0.1175, 0.0267]], + device='cuda:0'), grad: tensor([[ 1.1766e-04, 8.7404e-04, 6.6090e-04, ..., 1.9741e-04, + 1.1263e-03, 2.0771e-03], + [ 1.8328e-05, 3.7646e-04, 7.0906e-04, ..., 2.9850e-04, + 1.3342e-03, 1.7204e-03], + [-9.7215e-05, -3.6716e-05, 3.9864e-04, ..., -2.2554e-04, + 5.5742e-04, 1.5020e-03], + ..., + [ 5.4926e-05, -2.3422e-03, 3.8028e-04, ..., -2.9612e-04, + -5.8222e-04, -4.4518e-03], + [ 3.8457e-04, 6.6757e-04, 5.8222e-04, ..., 2.8396e-04, + 8.1205e-04, 1.3742e-03], + [ 3.1257e-04, 6.5851e-04, 5.2881e-04, ..., 3.8314e-04, + 1.2503e-03, 1.5345e-03]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0133, 0.0083, 0.0008, 0.0129, -0.0089, -0.0056, 0.0116, 0.0235, + -0.0336, 0.0466], device='cuda:0'), grad: tensor([ 0.0205, -0.0320, 0.0170, 0.0053, 0.0133, -0.0372, -0.0130, -0.0139, + 0.0225, 0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 216.72, cls_loss 0.4534 cls_loss_mapping 0.0031 cls_loss_causal 0.4276 re_mapping 0.0062 re_causal 0.0150 /// teacc 98.99 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.0768, 0.0715, -0.1094, ..., -0.1169, -0.1190, 0.0189], + [-0.0889, -0.1766, 0.0082, ..., -0.0551, -0.0399, -0.1133], + [ 0.0322, -0.1019, 0.0561, ..., 0.1579, -0.1359, -0.0387], + ..., + [-0.1140, -0.2087, 0.0906, ..., -0.0168, -0.0550, 0.0766], + [ 0.0256, 0.0397, -0.0359, ..., -0.0844, -0.1015, -0.0255], + [-0.1558, -0.0077, -0.0038, ..., -0.1767, 0.1170, 0.0263]], + device='cuda:0'), grad: tensor([[ 1.1018e-06, -1.8759e-03, -2.3117e-03, ..., -6.5947e-04, + -1.3418e-03, -3.7223e-05], + [-7.5054e-04, -3.4695e-03, 4.0174e-04, ..., 1.1981e-04, + 2.3639e-04, -7.5758e-05], + [ 9.3222e-05, 6.5374e-04, 7.8392e-04, ..., 3.1686e-04, + 3.3474e-04, 1.8346e-04], + ..., + [-8.4331e-07, 4.4227e-05, -5.2738e-04, ..., -3.5930e-04, + 1.4901e-04, -2.3901e-04], + [ 9.8288e-05, 1.2798e-03, 3.1204e-03, ..., 1.4484e-04, + 1.0662e-03, 1.0324e-04], + [ 6.1929e-05, -1.0805e-03, -5.8174e-03, ..., -5.8842e-04, + -5.8670e-03, -2.3174e-03]], device='cuda:0') +Epoch 364, bias, value: tensor([ 0.0129, 0.0089, -0.0003, 0.0128, -0.0094, -0.0043, 0.0113, 0.0233, + -0.0336, 0.0472], device='cuda:0'), grad: tensor([ 0.0054, -0.0178, -0.0168, 0.0189, 0.0209, 0.0190, 0.0129, -0.0246, + 0.0152, -0.0331], device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 216.79, cls_loss 0.4632 cls_loss_mapping 0.0027 cls_loss_causal 0.4409 re_mapping 0.0063 re_causal 0.0154 /// teacc 98.92 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.0768, 0.0715, -0.1095, ..., -0.1179, -0.1197, 0.0184], + [-0.0875, -0.1763, 0.0092, ..., -0.0554, -0.0412, -0.1144], + [ 0.0320, -0.1018, 0.0564, ..., 0.1579, -0.1374, -0.0389], + ..., + [-0.1158, -0.2095, 0.0896, ..., -0.0173, -0.0550, 0.0773], + [ 0.0254, 0.0399, -0.0363, ..., -0.0851, -0.1009, -0.0249], + [-0.1552, -0.0076, -0.0032, ..., -0.1757, 0.1184, 0.0271]], + device='cuda:0'), grad: tensor([[ 6.3479e-05, 2.0623e-05, -6.2883e-06, ..., -3.3490e-06, + 6.7711e-05, 4.1910e-09], + [ 1.1295e-04, 8.5533e-05, 2.1625e-06, ..., 9.2760e-06, + -1.1253e-03, 5.5414e-08], + [ 2.9254e-04, 2.5344e-04, -1.6429e-06, ..., 2.9132e-06, + 9.2268e-05, 3.8650e-08], + ..., + [ 4.5627e-05, 7.5054e-04, 1.6079e-03, ..., 2.8033e-06, + 1.8768e-03, 8.7547e-04], + [-5.4932e-03, -5.6458e-03, 5.7332e-06, ..., 1.1057e-04, + 9.4128e-04, 3.2783e-07], + [ 2.1625e-04, -5.1260e-04, -1.6155e-03, ..., 1.4305e-06, + -1.7366e-03, -8.7833e-04]], device='cuda:0') +Epoch 365, bias, value: tensor([ 1.1985e-02, 9.8359e-03, -4.9688e-05, 1.1470e-02, -9.1320e-03, + -4.1009e-03, 1.2912e-02, 2.2462e-02, -3.3616e-02, 4.7048e-02], + device='cuda:0'), grad: tensor([ 0.0115, -0.0237, -0.0503, 0.0157, -0.0173, 0.0129, 0.0181, 0.0166, + 0.0060, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 216.24, cls_loss 0.4570 cls_loss_mapping 0.0028 cls_loss_causal 0.4377 re_mapping 0.0058 re_causal 0.0139 /// teacc 98.82 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.0771, 0.0715, -0.1094, ..., -0.1181, -0.1192, 0.0199], + [-0.0875, -0.1766, 0.0086, ..., -0.0551, -0.0412, -0.1140], + [ 0.0327, -0.1016, 0.0567, ..., 0.1583, -0.1376, -0.0394], + ..., + [-0.1168, -0.2085, 0.0898, ..., -0.0169, -0.0551, 0.0748], + [ 0.0252, 0.0409, -0.0367, ..., -0.0862, -0.1011, -0.0253], + [-0.1561, -0.0081, -0.0032, ..., -0.1766, 0.1187, 0.0282]], + device='cuda:0'), grad: tensor([[-2.1229e-03, 2.2781e-02, -3.2082e-03, ..., 4.5076e-06, + -6.0749e-04, 3.8028e-04], + [ 4.3325e-06, 2.6774e-04, 3.2163e-04, ..., 7.4841e-06, + -2.0564e-04, 2.4128e-04], + [ 1.5414e-04, -1.5044e-04, 6.2418e-04, ..., -2.2873e-05, + -5.8222e-04, 3.2377e-04], + ..., + [ 1.7300e-05, 1.1700e-04, 9.6416e-04, ..., 4.0829e-05, + 6.2466e-04, 5.2404e-04], + [ 9.3651e-04, 1.7996e-03, -6.0921e-03, ..., 1.8924e-05, + -2.1648e-03, -1.7719e-03], + [ 2.5415e-04, -2.6764e-02, 4.5776e-03, ..., -2.6733e-05, + 4.2462e-04, -1.1873e-03]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0130, 0.0107, -0.0002, 0.0121, -0.0086, -0.0055, 0.0128, 0.0211, + -0.0335, 0.0471], device='cuda:0'), grad: tensor([ 0.0267, 0.0013, -0.0473, 0.0160, 0.0134, 0.0158, 0.0135, 0.0166, + -0.0240, -0.0321], device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 216.66, cls_loss 0.4793 cls_loss_mapping 0.0029 cls_loss_causal 0.4543 re_mapping 0.0059 re_causal 0.0150 /// teacc 98.71 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.0768, 0.0718, -0.1097, ..., -0.1176, -0.1195, 0.0195], + [-0.0877, -0.1777, 0.0085, ..., -0.0552, -0.0422, -0.1142], + [ 0.0321, -0.1006, 0.0577, ..., 0.1597, -0.1363, -0.0395], + ..., + [-0.1162, -0.2085, 0.0901, ..., -0.0176, -0.0537, 0.0750], + [ 0.0250, 0.0400, -0.0379, ..., -0.0871, -0.1034, -0.0251], + [-0.1573, -0.0086, -0.0029, ..., -0.1769, 0.1194, 0.0289]], + device='cuda:0'), grad: tensor([[ 5.4419e-05, 2.7597e-05, 9.6619e-05, ..., 1.6391e-04, + 5.2061e-07, 4.5728e-07], + [ 2.4724e-04, 1.1288e-05, 7.7772e-04, ..., 1.4858e-03, + 9.0450e-06, 1.1429e-05], + [-3.0365e-03, -6.8426e-05, -1.2598e-03, ..., -3.1033e-03, + 7.9796e-06, 1.2055e-05], + ..., + [ 2.8920e-04, 1.4536e-05, 6.3553e-06, ..., 3.9363e-04, + 2.4700e-04, 3.1441e-05], + [ 2.1100e-04, 2.8685e-05, 2.9659e-04, ..., 5.5504e-04, + 7.4431e-06, 3.8072e-06], + [ 1.4365e-05, 8.6278e-06, -9.9361e-05, ..., 1.6141e-04, + -8.7798e-05, 1.7929e-04]], device='cuda:0') +Epoch 367, bias, value: tensor([ 0.0131, 0.0110, -0.0001, 0.0126, -0.0090, -0.0052, 0.0119, 0.0208, + -0.0335, 0.0473], device='cuda:0'), grad: tensor([ 0.0092, 0.0212, -0.0091, 0.0131, 0.0095, 0.0075, 0.0131, -0.0875, + 0.0128, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 216.56, cls_loss 0.4561 cls_loss_mapping 0.0028 cls_loss_causal 0.4263 re_mapping 0.0062 re_causal 0.0156 /// teacc 98.63 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.0768, 0.0718, -0.1092, ..., -0.1172, -0.1194, 0.0200], + [-0.0873, -0.1769, 0.0078, ..., -0.0544, -0.0419, -0.1138], + [ 0.0332, -0.1011, 0.0573, ..., 0.1592, -0.1372, -0.0391], + ..., + [-0.1166, -0.2080, 0.0905, ..., -0.0181, -0.0527, 0.0761], + [ 0.0245, 0.0392, -0.0379, ..., -0.0876, -0.1043, -0.0255], + [-0.1568, -0.0082, -0.0038, ..., -0.1766, 0.1189, 0.0275]], + device='cuda:0'), grad: tensor([[ 4.8351e-04, -3.4618e-04, 3.3975e-04, ..., 2.9159e-04, + 4.7326e-04, 8.4448e-04], + [ 2.2030e-04, 2.2686e-04, 8.5771e-05, ..., 1.1933e-04, + 1.7405e-04, 2.8133e-04], + [-4.9257e-04, -9.5654e-04, -8.0109e-04, ..., -1.7309e-03, + 4.0364e-04, 2.4706e-05], + ..., + [ 3.4904e-04, 1.7214e-04, 3.7581e-05, ..., 9.5606e-05, + 3.0017e-04, 3.3593e-04], + [ 1.6851e-03, 2.9392e-03, 9.2173e-04, ..., 9.3269e-04, + 2.0981e-03, 3.8261e-03], + [ 6.6853e-04, -4.1008e-04, 9.2983e-05, ..., 9.3579e-05, + 5.9462e-04, 2.6822e-04]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0125, 0.0106, 0.0003, 0.0126, -0.0087, -0.0064, 0.0123, 0.0221, + -0.0334, 0.0469], device='cuda:0'), grad: tensor([ 0.0095, 0.0188, 0.0122, -0.0054, -0.0099, -0.0082, 0.0201, -0.0169, + -0.0045, -0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 216.65, cls_loss 0.4584 cls_loss_mapping 0.0027 cls_loss_causal 0.4377 re_mapping 0.0059 re_causal 0.0149 /// teacc 98.92 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.0763, 0.0732, -0.1097, ..., -0.1188, -0.1192, 0.0211], + [-0.0877, -0.1776, 0.0076, ..., -0.0538, -0.0417, -0.1145], + [ 0.0339, -0.1016, 0.0567, ..., 0.1604, -0.1378, -0.0383], + ..., + [-0.1164, -0.2082, 0.0911, ..., -0.0177, -0.0528, 0.0758], + [ 0.0238, 0.0386, -0.0384, ..., -0.0885, -0.1044, -0.0256], + [-0.1572, -0.0082, -0.0034, ..., -0.1763, 0.1190, 0.0278]], + device='cuda:0'), grad: tensor([[ 3.3212e-04, 3.5787e-04, 1.0556e-04, ..., 4.2844e-04, + 4.0680e-05, 7.0143e-04], + [ 2.5344e-04, 2.1017e-04, 8.2374e-05, ..., 3.0208e-04, + 7.8976e-05, -2.3880e-03], + [-1.7986e-03, -2.4815e-03, 4.5729e-04, ..., -2.2202e-03, + 1.7452e-04, 5.0640e-04], + ..., + [ 4.9973e-04, 8.3208e-05, 1.1158e-03, ..., 7.8201e-04, + 4.8971e-04, 2.3136e-03], + [-9.2840e-04, 7.0286e-04, -3.2063e-03, ..., -1.6222e-03, + -1.2760e-03, -4.2458e-03], + [ 1.0210e-04, -5.5522e-05, 1.8060e-04, ..., 1.6701e-04, + -3.9376e-06, 8.2064e-04]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0124, 0.0111, 0.0009, 0.0124, -0.0092, -0.0072, 0.0139, 0.0219, + -0.0341, 0.0468], device='cuda:0'), grad: tensor([ 0.0197, -0.0013, -0.0266, 0.0285, 0.0162, -0.0126, 0.0002, -0.0025, + -0.0066, -0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 216.55, cls_loss 0.4608 cls_loss_mapping 0.0023 cls_loss_causal 0.4384 re_mapping 0.0064 re_causal 0.0154 /// teacc 98.73 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.0758, 0.0736, -0.1094, ..., -0.1181, -0.1204, 0.0183], + [-0.0887, -0.1769, 0.0075, ..., -0.0537, -0.0409, -0.1135], + [ 0.0337, -0.1021, 0.0572, ..., 0.1618, -0.1373, -0.0381], + ..., + [-0.1168, -0.2087, 0.0911, ..., -0.0177, -0.0528, 0.0775], + [ 0.0246, 0.0388, -0.0380, ..., -0.0900, -0.1052, -0.0249], + [-0.1580, -0.0074, -0.0035, ..., -0.1771, 0.1187, 0.0266]], + device='cuda:0'), grad: tensor([[ 3.2485e-06, -8.3017e-04, 1.3673e-04, ..., 6.5565e-04, + 1.7667e-04, -1.1176e-05], + [ 8.9705e-06, 4.5031e-05, 6.2943e-04, ..., 1.2236e-03, + 4.9114e-04, 6.5613e-04], + [-7.4387e-03, -1.2074e-03, -1.0986e-02, ..., -9.8572e-03, + 3.8028e-04, -1.8282e-03], + ..., + [ 4.0084e-06, 4.0233e-05, 5.8937e-04, ..., 1.8673e-03, + 1.4019e-04, 1.4889e-04], + [ 6.0350e-06, 4.5002e-05, -4.6272e-03, ..., -1.1650e-02, + 3.0875e-04, 3.6120e-04], + [ 1.1586e-05, 4.5371e-04, 7.6437e-04, ..., 1.4048e-03, + 5.0926e-04, 3.1114e-04]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0123, 0.0105, 0.0007, 0.0118, -0.0082, -0.0064, 0.0132, 0.0218, + -0.0337, 0.0467], device='cuda:0'), grad: tensor([-0.0486, 0.0245, -0.0072, 0.0099, -0.0041, 0.0164, 0.0195, 0.0176, + -0.0159, -0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 216.49, cls_loss 0.4541 cls_loss_mapping 0.0026 cls_loss_causal 0.4279 re_mapping 0.0058 re_causal 0.0143 /// teacc 98.85 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.0759, 0.0731, -0.1095, ..., -0.1192, -0.1202, 0.0177], + [-0.0887, -0.1768, 0.0079, ..., -0.0537, -0.0407, -0.1141], + [ 0.0329, -0.1030, 0.0573, ..., 0.1624, -0.1377, -0.0373], + ..., + [-0.1174, -0.2095, 0.0919, ..., -0.0166, -0.0517, 0.0790], + [ 0.0266, 0.0390, -0.0379, ..., -0.0892, -0.1044, -0.0249], + [-0.1569, -0.0058, -0.0033, ..., -0.1770, 0.1186, 0.0260]], + device='cuda:0'), grad: tensor([[ 1.4305e-04, 1.4460e-04, 2.9540e-04, ..., 4.4608e-04, + 4.0978e-05, 3.2091e-04], + [ 4.5359e-05, 5.5104e-05, 9.2864e-05, ..., -1.9178e-03, + -8.1968e-04, -5.9509e-03], + [ 1.0939e-03, 2.2926e-03, 6.9580e-03, ..., 2.3468e-02, + 1.0741e-04, 1.1797e-03], + ..., + [ 3.9101e-04, -1.5381e-02, -3.0151e-02, ..., 9.5940e-04, + -3.3264e-02, -1.1358e-03], + [ 9.4032e-04, 1.2350e-03, 1.8396e-03, ..., 2.4223e-03, + 1.6820e-04, 7.7820e-04], + [ 5.3406e-05, 1.5457e-02, 2.9831e-02, ..., 3.8481e-04, + 3.3142e-02, 2.6302e-03]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0121, 0.0117, 0.0011, 0.0111, -0.0076, -0.0069, 0.0126, 0.0219, + -0.0332, 0.0460], device='cuda:0'), grad: tensor([ 0.0056, -0.0185, 0.0103, -0.0280, 0.0096, -0.0167, 0.0063, -0.0205, + 0.0154, 0.0364], device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 216.40, cls_loss 0.4689 cls_loss_mapping 0.0023 cls_loss_causal 0.4445 re_mapping 0.0058 re_causal 0.0143 /// teacc 98.94 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.0762, 0.0724, -0.1092, ..., -0.1178, -0.1204, 0.0185], + [-0.0886, -0.1771, 0.0077, ..., -0.0540, -0.0424, -0.1150], + [ 0.0322, -0.1013, 0.0575, ..., 0.1613, -0.1387, -0.0378], + ..., + [-0.1180, -0.2103, 0.0923, ..., -0.0162, -0.0509, 0.0801], + [ 0.0262, 0.0375, -0.0382, ..., -0.0894, -0.1061, -0.0269], + [-0.1574, -0.0061, -0.0039, ..., -0.1772, 0.1183, 0.0261]], + device='cuda:0'), grad: tensor([[ 1.0687e-04, 8.7023e-06, 5.3691e-07, ..., 1.1725e-06, + 5.7649e-07, 7.9632e-04], + [ 1.1063e-04, 1.9465e-06, 1.7388e-06, ..., 5.8534e-07, + 8.3959e-07, 1.1950e-03], + [ 8.7309e-04, 1.4436e-04, -9.1717e-06, ..., 1.3582e-05, + 1.0049e-06, 1.9455e-03], + ..., + [-1.1795e-02, 1.1791e-06, -5.3495e-06, ..., 1.1064e-06, + 5.3495e-06, -4.2694e-02], + [-4.4203e-04, -8.9943e-05, 3.6694e-06, ..., 7.6666e-06, + 4.0770e-05, 5.1117e-03], + [ 1.6928e-03, 1.3329e-05, 1.9944e-04, ..., 3.0696e-06, + 3.3069e-04, 9.9945e-03]], device='cuda:0') +Epoch 372, bias, value: tensor([ 0.0116, 0.0118, 0.0001, 0.0111, -0.0081, -0.0067, 0.0127, 0.0227, + -0.0332, 0.0467], device='cuda:0'), grad: tensor([ 0.0094, 0.0124, 0.0111, -0.0040, -0.0208, -0.0097, 0.0241, -0.0162, + -0.0260, 0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 216.47, cls_loss 0.4678 cls_loss_mapping 0.0020 cls_loss_causal 0.4433 re_mapping 0.0059 re_causal 0.0146 /// teacc 98.94 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.0772, 0.0727, -0.1087, ..., -0.1173, -0.1201, 0.0188], + [-0.0875, -0.1765, 0.0078, ..., -0.0540, -0.0439, -0.1147], + [ 0.0320, -0.1031, 0.0580, ..., 0.1619, -0.1392, -0.0378], + ..., + [-0.1181, -0.2111, 0.0918, ..., -0.0169, -0.0515, 0.0793], + [ 0.0264, 0.0377, -0.0392, ..., -0.0899, -0.1075, -0.0265], + [-0.1576, -0.0060, -0.0037, ..., -0.1777, 0.1192, 0.0270]], + device='cuda:0'), grad: tensor([[ 4.2647e-05, 1.5771e-04, 4.6134e-05, ..., 4.0627e-04, + 6.3229e-04, 4.7731e-04], + [ 3.9786e-05, 1.7059e-04, -1.2922e-03, ..., 1.2779e-04, + -2.3441e-03, -2.0504e-03], + [ 3.8415e-05, 2.4152e-04, 5.8746e-04, ..., 1.5879e-03, + 1.2541e-03, 9.2506e-04], + ..., + [-2.2531e-05, 1.1438e-04, -1.3793e-04, ..., 3.2091e-04, + 7.5579e-04, 5.9128e-04], + [-4.2391e-04, -6.8245e-03, 1.1343e-04, ..., 7.9870e-04, + -1.2636e-04, 8.8739e-04], + [ 5.7369e-05, 2.1958e-04, 6.3848e-04, ..., 5.0020e-04, + 1.7166e-03, 1.3103e-03]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0119, 0.0118, 0.0012, 0.0105, -0.0072, -0.0058, 0.0123, 0.0212, + -0.0333, 0.0463], device='cuda:0'), grad: tensor([ 0.0350, -0.0184, 0.0173, 0.0171, 0.0100, -0.0439, 0.0112, 0.0080, + -0.0480, 0.0116], device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 216.78, cls_loss 0.4885 cls_loss_mapping 0.0034 cls_loss_causal 0.4651 re_mapping 0.0055 re_causal 0.0142 /// teacc 98.91 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.0773, 0.0716, -0.1092, ..., -0.1187, -0.1208, 0.0185], + [-0.0869, -0.1767, 0.0074, ..., -0.0548, -0.0429, -0.1136], + [ 0.0321, -0.1035, 0.0572, ..., 0.1623, -0.1399, -0.0381], + ..., + [-0.1182, -0.2119, 0.0918, ..., -0.0166, -0.0512, 0.0784], + [ 0.0265, 0.0379, -0.0382, ..., -0.0886, -0.1067, -0.0251], + [-0.1593, -0.0063, -0.0044, ..., -0.1772, 0.1181, 0.0270]], + device='cuda:0'), grad: tensor([[-2.0351e-03, -3.0174e-03, -2.5291e-03, ..., -3.6831e-03, + -3.3970e-03, -5.8250e-03], + [-4.4298e-04, -3.3116e-04, 1.3375e-04, ..., -1.4954e-03, + -2.7199e-03, 2.4605e-04], + [-8.1444e-04, 4.3130e-04, 3.4022e-04, ..., 4.8113e-04, + 5.7840e-04, 8.0729e-04], + ..., + [ 2.0874e-04, 2.5630e-04, -3.0577e-05, ..., 3.1948e-04, + 3.7766e-04, -1.7185e-03], + [ 1.0052e-03, 1.2589e-03, 2.7061e-04, ..., 5.3024e-04, + 1.7653e-03, 9.6607e-04], + [-2.5535e-04, -1.4219e-03, 2.8729e-04, ..., 4.6635e-04, + -1.1387e-03, 8.6212e-04]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0111, 0.0120, -0.0001, 0.0105, -0.0071, -0.0068, 0.0124, 0.0225, + -0.0313, 0.0455], device='cuda:0'), grad: tensor([-0.0019, -0.0080, -0.0266, 0.0154, -0.0123, 0.0258, 0.0245, -0.0198, + 0.0224, -0.0197], device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 216.80, cls_loss 0.4934 cls_loss_mapping 0.0039 cls_loss_causal 0.4712 re_mapping 0.0061 re_causal 0.0159 /// teacc 98.84 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.0781, 0.0703, -0.1093, ..., -0.1195, -0.1217, 0.0202], + [-0.0879, -0.1784, 0.0086, ..., -0.0560, -0.0427, -0.1137], + [ 0.0332, -0.1024, 0.0569, ..., 0.1631, -0.1402, -0.0388], + ..., + [-0.1192, -0.2115, 0.0918, ..., -0.0160, -0.0509, 0.0781], + [ 0.0258, 0.0383, -0.0374, ..., -0.0886, -0.1067, -0.0256], + [-0.1584, -0.0036, -0.0046, ..., -0.1767, 0.1195, 0.0281]], + device='cuda:0'), grad: tensor([[ 3.5185e-06, -8.9216e-04, 2.3454e-05, ..., 2.8753e-04, + -9.5520e-03, -6.6566e-03], + [ 3.3855e-05, 3.1561e-05, -1.3399e-03, ..., -1.1162e-02, + -1.1164e-04, -5.9662e-03], + [ 2.9421e-04, 1.6618e-04, 8.8310e-04, ..., 4.4594e-03, + 4.4203e-04, 4.1656e-03], + ..., + [ 6.7800e-06, 3.1471e-05, 9.3207e-06, ..., 7.2289e-04, + 3.2473e-04, 8.8787e-04], + [ 6.1214e-05, 2.9182e-04, 3.9548e-05, ..., 6.5613e-04, + 2.4834e-03, 1.8826e-03], + [ 6.6049e-06, 3.5739e-04, 2.0063e-04, ..., 1.4615e-04, + 3.7365e-03, 2.7084e-03]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0113, 0.0114, -0.0008, 0.0106, -0.0077, -0.0052, 0.0122, 0.0229, + -0.0317, 0.0456], device='cuda:0'), grad: tensor([-0.0202, 0.0016, -0.0321, 0.0121, 0.0246, 0.0095, -0.0175, 0.0132, + -0.0107, 0.0195], device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 216.57, cls_loss 0.4819 cls_loss_mapping 0.0029 cls_loss_causal 0.4496 re_mapping 0.0062 re_causal 0.0150 /// teacc 98.91 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.0775, 0.0704, -0.1089, ..., -0.1204, -0.1193, 0.0210], + [-0.0868, -0.1780, 0.0086, ..., -0.0554, -0.0423, -0.1134], + [ 0.0318, -0.1030, 0.0561, ..., 0.1622, -0.1410, -0.0394], + ..., + [-0.1187, -0.2118, 0.0926, ..., -0.0158, -0.0505, 0.0788], + [ 0.0262, 0.0383, -0.0371, ..., -0.0891, -0.1070, -0.0252], + [-0.1579, -0.0029, -0.0053, ..., -0.1758, 0.1195, 0.0274]], + device='cuda:0'), grad: tensor([[ 4.6611e-05, 2.0206e-04, 3.7193e-04, ..., 7.2777e-05, + 8.6367e-05, 1.7178e-04], + [ 2.0325e-05, 5.9634e-05, -5.5599e-04, ..., 2.1040e-05, + -1.2815e-04, 7.5936e-05], + [-1.9097e-04, 4.2319e-05, -8.1420e-05, ..., -3.2282e-04, + 1.3091e-05, 1.1784e-04], + ..., + [ 4.3511e-05, -1.5903e-04, -3.7217e-04, ..., -1.4710e-04, + -1.1885e-04, -3.1900e-04], + [ 1.0830e-04, 1.8215e-04, 5.0974e-04, ..., 1.3316e-04, + -1.4938e-06, 2.0432e-04], + [ 8.5354e-05, 3.7646e-04, 7.1335e-04, ..., 2.0099e-04, + 4.9162e-04, 5.4884e-04]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0116, 0.0115, -0.0013, 0.0117, -0.0071, -0.0055, 0.0111, 0.0229, + -0.0316, 0.0453], device='cuda:0'), grad: tensor([ 0.0148, -0.0441, 0.0127, 0.0147, -0.0187, 0.0179, -0.0147, 0.0146, + 0.0193, -0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 216.66, cls_loss 0.4622 cls_loss_mapping 0.0028 cls_loss_causal 0.4426 re_mapping 0.0060 re_causal 0.0148 /// teacc 98.95 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.0775, 0.0693, -0.1092, ..., -0.1195, -0.1190, 0.0220], + [-0.0870, -0.1762, 0.0082, ..., -0.0543, -0.0422, -0.1135], + [ 0.0321, -0.1020, 0.0555, ..., 0.1610, -0.1422, -0.0398], + ..., + [-0.1187, -0.2123, 0.0933, ..., -0.0142, -0.0513, 0.0784], + [ 0.0246, 0.0373, -0.0369, ..., -0.0893, -0.1059, -0.0251], + [-0.1576, -0.0030, -0.0059, ..., -0.1758, 0.1193, 0.0268]], + device='cuda:0'), grad: tensor([[ 1.8875e-02, 2.3254e-02, 3.0780e-04, ..., 8.0795e-03, + 2.9635e-04, 2.8133e-04], + [ 3.1054e-05, 1.8641e-05, 5.6362e-04, ..., 2.0072e-05, + -2.5253e-03, -3.2406e-03], + [-1.8951e-02, -2.3849e-02, 2.4390e-04, ..., -8.3466e-03, + 1.9586e-04, 2.4092e-04], + ..., + [ 1.4508e-04, -6.8140e-04, -5.4970e-03, ..., 7.1287e-05, + -2.2144e-03, 5.6839e-04], + [-3.9406e-03, -1.1608e-05, 3.5977e-04, ..., 1.4436e-04, + 5.2929e-04, 5.4932e-04], + [ 1.0711e-04, 7.4959e-04, 5.8022e-03, ..., 4.7475e-05, + 2.7390e-03, 2.8872e-04]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0125, 0.0115, -0.0020, 0.0113, -0.0068, -0.0066, 0.0114, 0.0234, + -0.0324, 0.0464], device='cuda:0'), grad: tensor([ 0.0506, -0.0196, -0.0500, -0.0202, 0.0130, 0.0006, 0.0073, 0.0116, + -0.0132, 0.0200], device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 216.76, cls_loss 0.4533 cls_loss_mapping 0.0022 cls_loss_causal 0.4292 re_mapping 0.0067 re_causal 0.0164 /// teacc 98.94 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.0783, 0.0703, -0.1098, ..., -0.1213, -0.1201, 0.0216], + [-0.0892, -0.1761, 0.0086, ..., -0.0529, -0.0414, -0.1128], + [ 0.0336, -0.1020, 0.0548, ..., 0.1606, -0.1421, -0.0400], + ..., + [-0.1199, -0.2128, 0.0927, ..., -0.0148, -0.0529, 0.0774], + [ 0.0258, 0.0378, -0.0362, ..., -0.0902, -0.1068, -0.0248], + [-0.1581, -0.0028, -0.0053, ..., -0.1759, 0.1192, 0.0272]], + device='cuda:0'), grad: tensor([[-0.0023, -0.0015, 0.0007, ..., 0.0011, -0.0053, -0.0040], + [ 0.0001, 0.0010, 0.0041, ..., 0.0064, 0.0058, 0.0037], + [ 0.0004, 0.0005, 0.0011, ..., 0.0014, 0.0013, 0.0009], + ..., + [ 0.0002, -0.0015, -0.0041, ..., 0.0015, -0.0060, -0.0064], + [ 0.0012, 0.0015, 0.0010, ..., 0.0009, 0.0023, 0.0018], + [ 0.0001, 0.0013, -0.0020, ..., -0.0082, 0.0021, 0.0043]], + device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0120, 0.0110, -0.0023, 0.0119, -0.0067, -0.0061, 0.0112, 0.0231, + -0.0323, 0.0469], device='cuda:0'), grad: tensor([-0.0696, 0.0480, -0.0112, 0.0300, 0.0209, -0.0252, -0.0051, -0.0152, + 0.0240, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 216.71, cls_loss 0.4663 cls_loss_mapping 0.0033 cls_loss_causal 0.4497 re_mapping 0.0057 re_causal 0.0143 /// teacc 98.86 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.0794, 0.0708, -0.1099, ..., -0.1216, -0.1193, 0.0216], + [-0.0898, -0.1769, 0.0082, ..., -0.0538, -0.0416, -0.1130], + [ 0.0329, -0.1018, 0.0553, ..., 0.1614, -0.1432, -0.0409], + ..., + [-0.1204, -0.2126, 0.0921, ..., -0.0160, -0.0533, 0.0764], + [ 0.0269, 0.0380, -0.0363, ..., -0.0916, -0.1072, -0.0236], + [-0.1583, -0.0029, -0.0048, ..., -0.1739, 0.1193, 0.0274]], + device='cuda:0'), grad: tensor([[-1.0736e-05, -3.6925e-05, 8.0943e-05, ..., 2.8729e-04, + 4.3440e-04, 4.5037e-04], + [-1.5521e-04, -1.3475e-03, -8.1539e-04, ..., -2.9011e-03, + -4.2992e-03, -3.5992e-03], + [ 6.9678e-05, 1.0359e-04, 2.3520e-04, ..., 7.4530e-04, + 4.3011e-04, 8.4925e-04], + ..., + [ 3.6824e-06, 1.9956e-04, 6.5765e-03, ..., 6.4774e-03, + 8.5373e-03, 6.5727e-03], + [ 1.6141e-04, 1.1034e-03, 1.6952e-04, ..., 3.5429e-04, + 1.6441e-03, 7.4339e-04], + [ 2.3454e-05, -1.3769e-04, -9.3689e-03, ..., -8.8120e-03, + -1.0628e-02, -8.0643e-03]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0127, 0.0117, -0.0024, 0.0118, -0.0063, -0.0073, 0.0109, 0.0225, + -0.0317, 0.0468], device='cuda:0'), grad: tensor([ 0.0080, -0.0859, 0.0128, 0.0093, 0.0011, 0.0153, 0.0107, 0.0186, + 0.0143, -0.0042], device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 216.80, cls_loss 0.4854 cls_loss_mapping 0.0024 cls_loss_causal 0.4635 re_mapping 0.0058 re_causal 0.0157 /// teacc 98.97 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.0790, 0.0710, -0.1109, ..., -0.1227, -0.1202, 0.0216], + [-0.0904, -0.1760, 0.0082, ..., -0.0534, -0.0399, -0.1123], + [ 0.0329, -0.1011, 0.0554, ..., 0.1614, -0.1450, -0.0419], + ..., + [-0.1199, -0.2122, 0.0919, ..., -0.0163, -0.0533, 0.0769], + [ 0.0262, 0.0375, -0.0360, ..., -0.0918, -0.1063, -0.0249], + [-0.1581, -0.0036, -0.0047, ..., -0.1743, 0.1189, 0.0278]], + device='cuda:0'), grad: tensor([[ 0.0004, -0.0004, 0.0003, ..., 0.0019, 0.0006, 0.0002], + [ 0.0001, 0.0003, 0.0003, ..., 0.0004, 0.0007, 0.0002], + [ 0.0023, 0.0024, 0.0005, ..., 0.0019, -0.0006, 0.0002], + ..., + [ 0.0009, 0.0018, 0.0086, ..., 0.0004, 0.0036, 0.0028], + [-0.0001, 0.0004, 0.0001, ..., 0.0012, -0.0007, -0.0003], + [-0.0003, -0.0002, -0.0080, ..., 0.0010, -0.0015, -0.0022]], + device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0127, 0.0111, -0.0021, 0.0121, -0.0067, -0.0081, 0.0110, 0.0225, + -0.0317, 0.0477], device='cuda:0'), grad: tensor([ 0.0172, 0.0146, -0.0076, -0.0055, 0.0186, -0.0011, -0.0552, 0.0319, + -0.0231, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 216.37, cls_loss 0.5031 cls_loss_mapping 0.0033 cls_loss_causal 0.4794 re_mapping 0.0056 re_causal 0.0150 /// teacc 98.86 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.0788, 0.0703, -0.1114, ..., -0.1250, -0.1219, 0.0213], + [-0.0903, -0.1758, 0.0082, ..., -0.0534, -0.0400, -0.1128], + [ 0.0328, -0.1007, 0.0557, ..., 0.1596, -0.1451, -0.0414], + ..., + [-0.1204, -0.2131, 0.0922, ..., -0.0147, -0.0530, 0.0790], + [ 0.0262, 0.0369, -0.0360, ..., -0.0890, -0.1064, -0.0247], + [-0.1584, -0.0027, -0.0050, ..., -0.1759, 0.1183, 0.0264]], + device='cuda:0'), grad: tensor([[ 1.1081e-04, -2.0587e-04, 8.3065e-04, ..., 5.3358e-04, + 2.8586e-04, 1.0157e-03], + [ 3.7402e-05, 1.5426e-04, 9.9659e-04, ..., 2.2030e-03, + 8.7404e-04, 1.8787e-03], + [ 2.6817e-03, 4.0436e-03, 3.7060e-03, ..., 4.0627e-03, + 4.0703e-03, 4.1275e-03], + ..., + [ 1.4198e-04, 2.4772e-04, 9.7885e-03, ..., -2.8114e-03, + 1.4400e-03, 7.8583e-03], + [ 1.3077e-04, -2.8419e-03, 2.5520e-03, ..., 1.9245e-03, + 3.0689e-03, 3.5381e-03], + [ 3.0828e-04, 6.4516e-04, -8.6136e-03, ..., 1.8282e-03, + 4.0436e-03, -1.5898e-03]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0117, 0.0107, -0.0014, 0.0131, -0.0070, -0.0084, 0.0111, 0.0228, + -0.0313, 0.0471], device='cuda:0'), grad: tensor([ 0.0086, 0.0321, 0.0240, -0.0350, -0.0296, -0.0249, 0.0254, 0.0191, + -0.0037, -0.0161], device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 216.72, cls_loss 0.4469 cls_loss_mapping 0.0034 cls_loss_causal 0.4259 re_mapping 0.0056 re_causal 0.0137 /// teacc 98.96 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.0793, 0.0712, -0.1114, ..., -0.1243, -0.1200, 0.0218], + [-0.0907, -0.1750, 0.0092, ..., -0.0540, -0.0399, -0.1124], + [ 0.0354, -0.1003, 0.0551, ..., 0.1591, -0.1471, -0.0423], + ..., + [-0.1214, -0.2122, 0.0919, ..., -0.0132, -0.0526, 0.0780], + [ 0.0267, 0.0374, -0.0365, ..., -0.0899, -0.1060, -0.0247], + [-0.1591, -0.0029, -0.0028, ..., -0.1755, 0.1191, 0.0263]], + device='cuda:0'), grad: tensor([[ 2.0528e-04, -7.3528e-04, 1.3441e-05, ..., 1.4436e-04, + 3.6812e-04, 9.6023e-05], + [ 1.7893e-04, 6.4909e-05, -1.1444e-04, ..., -4.3941e-04, + -1.2550e-03, -3.1052e-03], + [-5.0354e-03, 1.1406e-03, 1.7881e-05, ..., -2.4509e-03, + 3.7503e-04, 7.4720e-04], + ..., + [ 7.1478e-04, 1.2350e-04, 6.9771e-03, ..., 3.6502e-04, + 4.0627e-03, 6.3515e-03], + [-5.0640e-04, -6.6710e-04, 4.6998e-05, ..., 5.5170e-04, + 2.1915e-03, 1.3428e-03], + [ 1.8060e-04, 1.1005e-03, 1.3838e-03, ..., 1.3196e-04, + 4.6196e-03, 2.4681e-03]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0123, 0.0107, -0.0013, 0.0131, -0.0074, -0.0079, 0.0106, 0.0218, + -0.0315, 0.0481], device='cuda:0'), grad: tensor([ 0.0060, -0.0757, 0.0023, 0.0219, -0.0171, -0.0171, 0.0096, 0.0248, + 0.0193, 0.0261], device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 216.50, cls_loss 0.4658 cls_loss_mapping 0.0039 cls_loss_causal 0.4405 re_mapping 0.0060 re_causal 0.0147 /// teacc 98.88 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.0808, 0.0710, -0.1113, ..., -0.1246, -0.1192, 0.0216], + [-0.0910, -0.1752, 0.0087, ..., -0.0535, -0.0389, -0.1117], + [ 0.0349, -0.1011, 0.0559, ..., 0.1600, -0.1461, -0.0421], + ..., + [-0.1215, -0.2121, 0.0916, ..., -0.0134, -0.0517, 0.0789], + [ 0.0274, 0.0372, -0.0366, ..., -0.0886, -0.1063, -0.0243], + [-0.1583, -0.0031, -0.0034, ..., -0.1768, 0.1177, 0.0249]], + device='cuda:0'), grad: tensor([[ 4.2105e-04, 5.1355e-04, 1.3137e-04, ..., 2.3699e-04, + 3.4308e-04, 8.6874e-06], + [ 1.0449e-04, 1.2350e-04, -6.4552e-05, ..., 1.0288e-04, + 1.6105e-04, 6.4790e-05], + [-6.6643e-03, -3.9291e-03, -7.2021e-03, ..., -7.6790e-03, + -1.2302e-04, 5.5015e-05], + ..., + [ 6.0290e-05, 1.6880e-04, 2.0199e-03, ..., 1.1196e-03, + 8.9216e-04, 5.1355e-04], + [ 2.7585e-04, 3.8552e-04, 1.1158e-03, ..., 8.5592e-04, + 4.2248e-04, 1.4925e-04], + [ 1.0616e-04, -6.8784e-05, -9.2840e-04, ..., 9.2328e-05, + -1.4629e-03, -9.3842e-04]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0128, 0.0114, -0.0006, 0.0126, -0.0072, -0.0069, 0.0092, 0.0214, + -0.0316, 0.0474], device='cuda:0'), grad: tensor([ 0.0038, -0.0004, -0.0275, 0.0244, 0.0029, -0.0326, 0.0244, 0.0074, + 0.0045, -0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 216.46, cls_loss 0.4725 cls_loss_mapping 0.0033 cls_loss_causal 0.4456 re_mapping 0.0061 re_causal 0.0156 /// teacc 98.92 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.0806, 0.0706, -0.1108, ..., -0.1239, -0.1199, 0.0214], + [-0.0908, -0.1756, 0.0092, ..., -0.0541, -0.0399, -0.1131], + [ 0.0340, -0.1025, 0.0559, ..., 0.1603, -0.1437, -0.0411], + ..., + [-0.1231, -0.2127, 0.0905, ..., -0.0147, -0.0518, 0.0784], + [ 0.0282, 0.0383, -0.0364, ..., -0.0887, -0.1061, -0.0249], + [-0.1595, -0.0031, -0.0029, ..., -0.1769, 0.1175, 0.0258]], + device='cuda:0'), grad: tensor([[ 7.2241e-04, 7.2098e-04, 3.5882e-04, ..., 6.2466e-04, + 1.0891e-03, 1.3714e-03], + [-1.6785e-03, -9.0313e-04, 2.5916e-04, ..., -9.2554e-04, + 1.0419e-04, 1.0717e-04], + [ 6.8724e-05, 3.5596e-04, 3.5524e-04, ..., -1.1263e-03, + -9.1028e-04, -1.0414e-03], + ..., + [ 4.2892e-04, 8.1635e-04, 1.0643e-02, ..., 4.6945e-04, + 1.6846e-02, 1.1688e-02], + [ 3.3355e-04, -5.7125e-04, -1.3481e-02, ..., 9.3937e-04, + -1.8799e-02, -1.2764e-02], + [-3.0947e-04, -7.7605e-05, 1.0958e-03, ..., -5.1832e-04, + 4.1466e-03, -4.7779e-04]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0134, 0.0108, -0.0002, 0.0131, -0.0077, -0.0079, 0.0097, 0.0217, + -0.0322, 0.0480], device='cuda:0'), grad: tensor([ 0.0219, -0.0159, 0.0072, 0.0203, 0.0117, 0.0043, -0.0262, 0.0013, + -0.0089, -0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 216.68, cls_loss 0.4758 cls_loss_mapping 0.0030 cls_loss_causal 0.4561 re_mapping 0.0060 re_causal 0.0152 /// teacc 98.73 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.0807, 0.0711, -0.1114, ..., -0.1248, -0.1212, 0.0210], + [-0.0913, -0.1750, 0.0093, ..., -0.0541, -0.0403, -0.1136], + [ 0.0348, -0.1026, 0.0561, ..., 0.1604, -0.1424, -0.0412], + ..., + [-0.1240, -0.2131, 0.0914, ..., -0.0143, -0.0521, 0.0772], + [ 0.0270, 0.0383, -0.0365, ..., -0.0902, -0.1057, -0.0246], + [-0.1595, -0.0029, -0.0035, ..., -0.1767, 0.1181, 0.0269]], + device='cuda:0'), grad: tensor([[ 1.0738e-03, 1.2493e-04, 1.2124e-04, ..., 1.9595e-05, + 1.8704e-04, 5.8502e-05], + [ 1.5759e-04, 3.4332e-05, 1.9729e-04, ..., 6.2704e-05, + 1.0103e-04, 5.0366e-05], + [ 1.8215e-03, 5.0068e-04, 1.6575e-03, ..., 7.0953e-04, + 2.8801e-04, 3.1877e-04], + ..., + [ 5.2547e-04, 6.5029e-05, -2.6340e-03, ..., -1.2016e-03, + -3.9139e-03, -3.1757e-04], + [-1.1253e-02, -1.2341e-03, 8.5771e-05, ..., 9.4116e-05, + 3.9005e-04, 1.4281e-04], + [ 1.6232e-03, 6.8665e-05, 7.7343e-04, ..., 2.8825e-04, + 4.5357e-03, 2.3532e-04]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0122, 0.0113, -0.0009, 0.0137, -0.0084, -0.0085, 0.0103, 0.0220, + -0.0330, 0.0498], device='cuda:0'), grad: tensor([ 0.0055, 0.0018, 0.0133, 0.0103, 0.0012, -0.0005, 0.0003, -0.0294, + -0.0280, 0.0256], device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 216.54, cls_loss 0.4826 cls_loss_mapping 0.0038 cls_loss_causal 0.4607 re_mapping 0.0060 re_causal 0.0151 /// teacc 98.84 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.0801, 0.0711, -0.1121, ..., -0.1250, -0.1213, 0.0208], + [-0.0929, -0.1753, 0.0094, ..., -0.0534, -0.0416, -0.1130], + [ 0.0344, -0.1034, 0.0553, ..., 0.1605, -0.1429, -0.0410], + ..., + [-0.1242, -0.2122, 0.0911, ..., -0.0134, -0.0524, 0.0778], + [ 0.0270, 0.0381, -0.0367, ..., -0.0910, -0.1060, -0.0241], + [-0.1583, -0.0027, -0.0028, ..., -0.1785, 0.1182, 0.0262]], + device='cuda:0'), grad: tensor([[ 1.1292e-02, 1.4315e-03, -1.0031e-04, ..., -1.7822e-05, + 7.7963e-05, 1.7440e-04], + [ 2.6059e-04, 5.7667e-05, -1.2531e-03, ..., -2.5010e-04, + -6.5660e-04, -6.6459e-05], + [-1.2279e-04, 2.3973e-04, 1.8203e-04, ..., 8.1241e-05, + -2.5916e-04, 2.6393e-04], + ..., + [ 2.1458e-04, 5.2333e-05, -1.8263e-03, ..., 7.2241e-05, + -5.8711e-05, 8.5211e-04], + [ 2.0771e-03, 2.7013e-04, 6.4278e-04, ..., 4.1217e-05, + 3.1543e-04, 4.1676e-04], + [ 9.0742e-04, 2.9922e-04, 1.8978e-03, ..., 1.4193e-05, + 1.2283e-03, 9.6607e-04]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0129, 0.0111, -0.0009, 0.0149, -0.0101, -0.0089, 0.0104, 0.0220, + -0.0328, 0.0499], device='cuda:0'), grad: tensor([ 0.0257, 0.0117, -0.0105, -0.0394, -0.0140, 0.0195, -0.0595, 0.0152, + 0.0246, 0.0267], device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 216.49, cls_loss 0.4785 cls_loss_mapping 0.0028 cls_loss_causal 0.4521 re_mapping 0.0059 re_causal 0.0153 /// teacc 98.92 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.0815, 0.0704, -0.1124, ..., -0.1242, -0.1216, 0.0217], + [-0.0934, -0.1763, 0.0089, ..., -0.0541, -0.0423, -0.1136], + [ 0.0348, -0.1031, 0.0546, ..., 0.1604, -0.1442, -0.0417], + ..., + [-0.1233, -0.2129, 0.0914, ..., -0.0135, -0.0532, 0.0778], + [ 0.0263, 0.0386, -0.0369, ..., -0.0916, -0.1056, -0.0251], + [-0.1581, -0.0021, -0.0027, ..., -0.1783, 0.1187, 0.0274]], + device='cuda:0'), grad: tensor([[-6.4621e-03, 6.2103e-03, 3.6678e-03, ..., 3.1452e-03, + 1.8320e-03, -5.2605e-03], + [ 5.0783e-04, 9.6500e-05, 7.6294e-04, ..., 3.2005e-03, + 6.4421e-04, 7.9060e-04], + [-3.6793e-03, -8.2474e-03, -9.7418e-04, ..., -3.5839e-03, + -3.5839e-03, -3.1319e-03], + ..., + [ 4.7989e-03, 1.4246e-04, -5.2757e-03, ..., -2.2125e-03, + 8.9455e-04, 7.2289e-04], + [ 4.1924e-03, 1.4267e-03, 2.1064e-04, ..., 2.8419e-04, + 2.4486e-04, 1.4563e-03], + [ 7.0763e-04, 3.2687e-04, 2.5129e-04, ..., 1.9073e-03, + -9.2626e-05, 6.1083e-04]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0125, 0.0113, -0.0013, 0.0138, -0.0099, -0.0087, 0.0114, 0.0216, + -0.0330, 0.0508], device='cuda:0'), grad: tensor([ 0.0283, 0.0208, -0.0458, -0.0336, 0.0212, 0.0188, -0.0344, -0.0026, + 0.0330, -0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 216.92, cls_loss 0.4762 cls_loss_mapping 0.0014 cls_loss_causal 0.4468 re_mapping 0.0059 re_causal 0.0152 /// teacc 98.88 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.0816, 0.0712, -0.1122, ..., -0.1228, -0.1206, 0.0217], + [-0.0929, -0.1767, 0.0090, ..., -0.0541, -0.0426, -0.1138], + [ 0.0329, -0.1033, 0.0542, ..., 0.1598, -0.1442, -0.0427], + ..., + [-0.1226, -0.2129, 0.0909, ..., -0.0133, -0.0536, 0.0775], + [ 0.0286, 0.0390, -0.0367, ..., -0.0913, -0.1054, -0.0240], + [-0.1578, -0.0015, -0.0021, ..., -0.1786, 0.1185, 0.0277]], + device='cuda:0'), grad: tensor([[ 2.2769e-04, 4.5705e-04, 3.1944e-07, ..., 2.4662e-03, + 6.8235e-04, 1.5211e-04], + [ 8.3089e-05, 1.9026e-04, 1.7509e-07, ..., 2.0158e-04, + -2.1210e-03, 3.5167e-05], + [ 1.0431e-04, -6.2227e-04, 5.1372e-06, ..., 9.1314e-05, + -9.6273e-04, -1.4596e-05], + ..., + [ 1.3745e-04, 1.5163e-04, -1.2040e-05, ..., 8.3566e-05, + 4.9686e-04, 2.5749e-04], + [-5.5075e-04, 3.5024e-04, 9.6764e-07, ..., 4.1342e-04, + 9.6846e-04, -8.0490e-04], + [ 1.5950e-04, 1.0900e-03, 2.1048e-06, ..., 2.8534e-03, + 3.6373e-03, 2.3186e-04]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0125, 0.0112, -0.0019, 0.0144, -0.0099, -0.0083, 0.0100, 0.0208, + -0.0315, 0.0510], device='cuda:0'), grad: tensor([-0.0078, -0.0072, -0.0163, 0.0206, -0.0017, 0.0028, -0.0016, 0.0230, + -0.0416, 0.0299], device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 216.82, cls_loss 0.4764 cls_loss_mapping 0.0024 cls_loss_causal 0.4472 re_mapping 0.0061 re_causal 0.0152 /// teacc 98.86 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.0822, 0.0707, -0.1129, ..., -0.1215, -0.1203, 0.0225], + [-0.0916, -0.1758, 0.0088, ..., -0.0541, -0.0425, -0.1145], + [ 0.0323, -0.1035, 0.0556, ..., 0.1614, -0.1439, -0.0434], + ..., + [-0.1233, -0.2132, 0.0915, ..., -0.0141, -0.0532, 0.0785], + [ 0.0282, 0.0378, -0.0373, ..., -0.0913, -0.1048, -0.0243], + [-0.1589, -0.0015, -0.0030, ..., -0.1802, 0.1177, 0.0281]], + device='cuda:0'), grad: tensor([[-7.9803e-03, -7.0992e-03, 2.4772e-04, ..., -1.9341e-03, + 6.2513e-04, 9.4366e-04], + [ 3.3236e-04, 3.7223e-05, 2.5511e-04, ..., 2.5487e-04, + 5.0926e-04, -2.5773e-04], + [ 3.4428e-03, 2.5368e-03, -3.9077e-04, ..., 2.0254e-04, + 3.5763e-04, 6.7997e-04], + ..., + [ 4.2939e-04, 5.1647e-05, -1.5583e-03, ..., 1.7536e-04, + -4.3259e-03, -5.7831e-03], + [ 1.3971e-03, 9.6798e-04, -6.5565e-05, ..., 4.9925e-04, + -1.9951e-03, -2.2793e-03], + [ 1.6737e-04, 6.5386e-05, 4.2033e-04, ..., 2.9635e-04, + 1.1034e-03, 1.7614e-03]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0124, 0.0108, -0.0011, 0.0154, -0.0094, -0.0090, 0.0093, 0.0221, + -0.0323, 0.0501], device='cuda:0'), grad: tensor([-0.0040, -0.0083, 0.0312, 0.0015, 0.0302, 0.0260, -0.0048, 0.0030, + -0.0052, -0.0697], device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 217.08, cls_loss 0.4728 cls_loss_mapping 0.0023 cls_loss_causal 0.4419 re_mapping 0.0062 re_causal 0.0162 /// teacc 98.92 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.0821, 0.0710, -0.1145, ..., -0.1222, -0.1218, 0.0221], + [-0.0900, -0.1771, 0.0100, ..., -0.0536, -0.0415, -0.1157], + [ 0.0326, -0.1029, 0.0558, ..., 0.1605, -0.1445, -0.0421], + ..., + [-0.1252, -0.2141, 0.0915, ..., -0.0146, -0.0531, 0.0789], + [ 0.0291, 0.0385, -0.0370, ..., -0.0911, -0.1049, -0.0247], + [-0.1584, -0.0011, -0.0030, ..., -0.1809, 0.1175, 0.0276]], + device='cuda:0'), grad: tensor([[ 3.2019e-06, 6.9332e-04, 6.0536e-07, ..., 4.0829e-05, + 5.4091e-05, 4.2057e-04], + [ 5.3421e-06, 7.7903e-05, 8.8289e-06, ..., 9.0480e-05, + 6.5744e-05, 4.4703e-04], + [ 1.7539e-05, 5.7936e-05, 4.0412e-05, ..., 7.5459e-05, + 7.5340e-05, 4.2295e-04], + ..., + [ 8.1360e-05, 1.5044e-04, 7.2098e-04, ..., -3.3689e-04, + 9.2745e-04, 9.0361e-04], + [-1.3530e-04, 2.0456e-04, 6.0987e-04, ..., 7.1585e-05, + 6.9237e-04, 5.3596e-04], + [-8.4519e-05, -9.9838e-05, -1.9588e-03, ..., -2.6870e-04, + -2.2202e-03, -3.6025e-04]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0126, 0.0115, -0.0005, 0.0150, -0.0105, -0.0083, 0.0085, 0.0225, + -0.0323, 0.0499], device='cuda:0'), grad: tensor([ 0.0257, 0.0187, 0.0191, -0.0123, 0.0122, -0.0444, 0.0128, -0.0050, + -0.0358, 0.0090], device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 218.52, cls_loss 0.4915 cls_loss_mapping 0.0027 cls_loss_causal 0.4625 re_mapping 0.0056 re_causal 0.0152 /// teacc 98.98 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.0806, 0.0716, -0.1152, ..., -0.1238, -0.1230, 0.0205], + [-0.0894, -0.1764, 0.0098, ..., -0.0552, -0.0415, -0.1152], + [ 0.0321, -0.1037, 0.0561, ..., 0.1608, -0.1433, -0.0409], + ..., + [-0.1258, -0.2151, 0.0906, ..., -0.0157, -0.0536, 0.0783], + [ 0.0292, 0.0382, -0.0360, ..., -0.0906, -0.1049, -0.0247], + [-0.1585, -0.0012, -0.0034, ..., -0.1810, 0.1171, 0.0285]], + device='cuda:0'), grad: tensor([[ 2.1229e-03, 2.7373e-05, 2.9230e-04, ..., 8.2493e-04, + 5.8651e-04, 4.6945e-04], + [ 7.2837e-05, 6.0320e-05, 4.3035e-04, ..., 9.4843e-04, + 1.1921e-04, 1.6391e-04], + [-4.6110e-04, 1.0246e-04, -4.8370e-03, ..., -1.7120e-02, + -1.2367e-02, -7.9803e-03], + ..., + [ 1.1849e-04, 1.3340e-04, -5.9986e-04, ..., 7.0906e-04, + 3.3855e-04, 3.0017e-04], + [ 6.1417e-04, 6.4039e-04, 6.1703e-04, ..., 1.1625e-03, + 1.5383e-03, 1.8942e-04], + [ 6.9094e-04, 2.0063e-04, 1.9379e-03, ..., 7.1945e-03, + 9.4299e-03, 6.3934e-03]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0123, 0.0111, 0.0003, 0.0149, -0.0091, -0.0085, 0.0085, 0.0214, + -0.0323, 0.0497], device='cuda:0'), grad: tensor([-0.0121, -0.0138, -0.0322, -0.0349, 0.0255, 0.0124, 0.0129, 0.0120, + -0.0084, 0.0386], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 390---------------------------------------------------- +epoch 390, time 217.35, cls_loss 0.4913 cls_loss_mapping 0.0028 cls_loss_causal 0.4704 re_mapping 0.0056 re_causal 0.0151 /// teacc 99.01 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.0799, 0.0720, -0.1152, ..., -0.1228, -0.1224, 0.0210], + [-0.0898, -0.1768, 0.0092, ..., -0.0555, -0.0417, -0.1153], + [ 0.0325, -0.1034, 0.0571, ..., 0.1610, -0.1422, -0.0388], + ..., + [-0.1261, -0.2146, 0.0905, ..., -0.0147, -0.0549, 0.0768], + [ 0.0285, 0.0380, -0.0365, ..., -0.0911, -0.1053, -0.0243], + [-0.1597, -0.0011, -0.0040, ..., -0.1828, 0.1167, 0.0275]], + device='cuda:0'), grad: tensor([[ 7.0477e-04, 1.3113e-04, 1.0675e-04, ..., 5.6446e-05, + 2.0349e-04, 2.1124e-04], + [ 3.0935e-05, 1.0043e-05, 8.3327e-05, ..., 4.3720e-05, + 1.5295e-04, 5.5879e-05], + [ 8.9502e-04, 1.6654e-04, 8.5592e-05, ..., 3.2395e-05, + 1.8811e-04, 2.4486e-04], + ..., + [ 3.3808e-04, 5.7787e-05, 7.4506e-05, ..., 4.8786e-05, + 1.9312e-04, 1.3459e-04], + [-5.1537e-03, -5.4455e-04, 1.4818e-04, ..., 7.7844e-05, + 3.1638e-04, -1.4343e-03], + [ 3.2864e-03, 2.1343e-03, 1.3483e-04, ..., 6.7115e-05, + 4.7989e-03, 5.1832e-04]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0127, 0.0115, -0.0002, 0.0134, -0.0084, -0.0079, 0.0096, 0.0206, + -0.0324, 0.0496], device='cuda:0'), grad: tensor([ 0.0078, 0.0031, 0.0062, 0.0097, 0.0028, -0.0327, 0.0017, 0.0046, + -0.0174, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 217.31, cls_loss 0.4641 cls_loss_mapping 0.0030 cls_loss_causal 0.4418 re_mapping 0.0059 re_causal 0.0146 /// teacc 98.97 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.0802, 0.0726, -0.1151, ..., -0.1246, -0.1237, 0.0199], + [-0.0897, -0.1783, 0.0098, ..., -0.0522, -0.0396, -0.1158], + [ 0.0329, -0.1025, 0.0551, ..., 0.1597, -0.1418, -0.0398], + ..., + [-0.1258, -0.2136, 0.0901, ..., -0.0149, -0.0554, 0.0756], + [ 0.0289, 0.0387, -0.0361, ..., -0.0909, -0.1046, -0.0229], + [-0.1602, -0.0004, -0.0028, ..., -0.1821, 0.1160, 0.0269]], + device='cuda:0'), grad: tensor([[ 1.0413e-04, 1.0170e-05, 3.6638e-06, ..., 9.4995e-08, + 2.1601e-04, 3.0041e-04], + [-2.1112e-04, -4.2009e-04, -3.8356e-05, ..., -1.0617e-06, + -2.6569e-03, -3.3836e-03], + [ 1.2817e-03, 1.3542e-04, 2.5535e-04, ..., -2.6211e-05, + 2.7633e-04, 3.8242e-04], + ..., + [ 2.4867e-04, 3.8981e-05, 8.3596e-06, ..., 9.1344e-06, + 3.1281e-04, 4.1127e-04], + [-1.8701e-05, -6.5982e-05, -2.6298e-04, ..., 8.9556e-06, + 1.9777e-04, 1.4324e-06], + [ 3.5548e-04, 5.4091e-05, 1.0014e-05, ..., -7.8138e-07, + 9.9242e-05, 3.1143e-05]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0116, 0.0117, -0.0009, 0.0143, -0.0078, -0.0074, 0.0097, 0.0206, + -0.0326, 0.0494], device='cuda:0'), grad: tensor([-0.0138, -0.0453, 0.0285, 0.0120, 0.0183, 0.0105, -0.0117, 0.0326, + -0.0210, -0.0099], device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 217.05, cls_loss 0.4699 cls_loss_mapping 0.0030 cls_loss_causal 0.4427 re_mapping 0.0061 re_causal 0.0150 /// teacc 98.58 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.0793, 0.0724, -0.1151, ..., -0.1230, -0.1238, 0.0199], + [-0.0888, -0.1794, 0.0088, ..., -0.0532, -0.0395, -0.1161], + [ 0.0317, -0.1020, 0.0548, ..., 0.1595, -0.1442, -0.0407], + ..., + [-0.1265, -0.2135, 0.0899, ..., -0.0159, -0.0546, 0.0752], + [ 0.0296, 0.0381, -0.0362, ..., -0.0894, -0.1036, -0.0222], + [-0.1602, 0.0003, -0.0025, ..., -0.1839, 0.1152, 0.0266]], + device='cuda:0'), grad: tensor([[7.9679e-04, 7.0953e-04, 1.9097e-04, ..., 1.2147e-04, 6.0463e-04, + 4.0603e-04], + [2.3782e-05, 1.2648e-04, 2.2724e-05, ..., 2.7701e-05, 6.6936e-05, + 1.5116e-04], + [4.4608e-04, 6.1703e-04, 7.3850e-05, ..., 5.3287e-05, 3.1281e-04, + 2.8944e-04], + ..., + [2.6727e-04, 2.7728e-04, 1.4436e-04, ..., 8.7023e-05, 4.1127e-04, + 2.9588e-04], + [4.4727e-04, 4.2510e-04, 1.8013e-04, ..., 6.0141e-05, 4.7350e-04, + 3.4976e-04], + [2.9683e-04, 6.0654e-04, 1.9097e-04, ..., 3.7289e-04, 2.4629e-04, + 5.3406e-04]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0121, 0.0108, -0.0007, 0.0151, -0.0073, -0.0080, 0.0103, 0.0209, + -0.0329, 0.0482], device='cuda:0'), grad: tensor([ 0.0168, -0.0128, 0.0168, 0.0183, -0.0591, 0.0080, -0.0115, 0.0189, + -0.0130, 0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 217.26, cls_loss 0.4580 cls_loss_mapping 0.0031 cls_loss_causal 0.4359 re_mapping 0.0058 re_causal 0.0145 /// teacc 98.93 lr 0.00010000 +Epoch 395, weight, value: tensor([[-7.9330e-02, 7.3517e-02, -1.1472e-01, ..., -1.2181e-01, + -1.2442e-01, 1.9694e-02], + [-8.8200e-02, -1.7949e-01, 8.9559e-03, ..., -5.2675e-02, + -3.9793e-02, -1.1631e-01], + [ 3.1546e-02, -1.0145e-01, 5.5496e-02, ..., 1.6058e-01, + -1.4407e-01, -4.0807e-02], + ..., + [-1.2540e-01, -2.1350e-01, 8.9072e-02, ..., -1.7829e-02, + -5.4547e-02, 7.5606e-02], + [ 2.9496e-02, 3.8487e-02, -3.7155e-02, ..., -8.9470e-02, + -1.0469e-01, -2.2067e-02], + [-1.6049e-01, -1.6982e-04, -1.8611e-03, ..., -1.8366e-01, + 1.1544e-01, 2.6853e-02]], device='cuda:0'), grad: tensor([[ 3.7932e-04, 2.5082e-04, 1.0473e-04, ..., 1.6570e-04, + -2.4261e-03, 2.1911e-04], + [ 6.1728e-06, 7.3090e-06, 5.0068e-04, ..., 1.8740e-03, + 5.0468e-03, 2.7847e-03], + [-1.9896e-04, -3.0726e-05, -4.5943e-04, ..., -2.3019e-04, + 6.0749e-04, 1.9598e-04], + ..., + [ 1.2887e-04, 5.4181e-05, -1.1063e-03, ..., 1.1654e-03, + 1.8301e-03, -1.8177e-03], + [-8.0795e-03, -5.1918e-03, 3.0589e-04, ..., 2.5725e-04, + 3.0804e-04, 4.6301e-04], + [-4.3154e-04, -6.0797e-04, -1.3342e-03, ..., -3.6621e-03, + -5.7220e-03, -3.2692e-03]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0116, 0.0108, -0.0007, 0.0149, -0.0077, -0.0084, 0.0109, 0.0210, + -0.0326, 0.0486], device='cuda:0'), grad: tensor([-0.0143, 0.0075, 0.0167, -0.0115, 0.0246, 0.0202, -0.0300, -0.0008, + 0.0006, -0.0130], device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 217.31, cls_loss 0.4758 cls_loss_mapping 0.0032 cls_loss_causal 0.4593 re_mapping 0.0055 re_causal 0.0133 /// teacc 98.85 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.0795, 0.0730, -0.1148, ..., -0.1222, -0.1244, 0.0217], + [-0.0879, -0.1791, 0.0093, ..., -0.0525, -0.0400, -0.1174], + [ 0.0300, -0.1027, 0.0557, ..., 0.1602, -0.1444, -0.0399], + ..., + [-0.1237, -0.2148, 0.0896, ..., -0.0178, -0.0539, 0.0763], + [ 0.0311, 0.0399, -0.0369, ..., -0.0895, -0.1036, -0.0215], + [-0.1623, -0.0012, -0.0018, ..., -0.1840, 0.1158, 0.0274]], + device='cuda:0'), grad: tensor([[ 2.7239e-05, 1.5914e-04, 1.9316e-06, ..., 9.1934e-04, + 2.4104e-04, 8.2403e-06], + [ 1.0476e-05, 1.2338e-04, 1.2152e-05, ..., 1.2338e-04, + 1.3084e-03, 5.9456e-05], + [-1.8942e-04, -6.3229e-04, -3.8236e-05, ..., -1.4420e-03, + -1.3227e-03, 7.0706e-06], + ..., + [ 3.6120e-05, 2.2054e-04, 5.2303e-06, ..., 1.7595e-04, + -2.8667e-03, -1.3864e-04], + [ 7.7784e-05, -1.3790e-03, 6.7651e-06, ..., 1.1671e-04, + 1.2279e-04, 1.1861e-05], + [ 1.4150e-04, 2.4509e-04, 1.5080e-05, ..., 8.5056e-05, + 6.2323e-04, 5.0962e-05]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0110, 0.0103, -0.0010, 0.0145, -0.0086, -0.0089, 0.0117, 0.0216, + -0.0315, 0.0492], device='cuda:0'), grad: tensor([ 0.0381, 0.0189, -0.0203, -0.0205, 0.0192, 0.0389, -0.0177, -0.0451, + -0.0267, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 216.99, cls_loss 0.4706 cls_loss_mapping 0.0023 cls_loss_causal 0.4525 re_mapping 0.0061 re_causal 0.0149 /// teacc 98.84 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.0792, 0.0723, -0.1154, ..., -0.1235, -0.1245, 0.0228], + [-0.0884, -0.1790, 0.0087, ..., -0.0528, -0.0395, -0.1175], + [ 0.0301, -0.1014, 0.0567, ..., 0.1612, -0.1435, -0.0399], + ..., + [-0.1242, -0.2147, 0.0892, ..., -0.0168, -0.0544, 0.0751], + [ 0.0306, 0.0388, -0.0371, ..., -0.0906, -0.1039, -0.0217], + [-0.1608, 0.0007, -0.0015, ..., -0.1835, 0.1169, 0.0284]], + device='cuda:0'), grad: tensor([[-2.3097e-05, 3.0255e-04, 3.2932e-06, ..., -2.1812e-06, + 9.7871e-05, 9.0694e-04], + [ 2.5555e-06, 1.0967e-04, 1.0826e-05, ..., 8.8587e-06, + 5.7459e-05, 4.0507e-04], + [ 8.0317e-06, 2.7418e-04, 7.2777e-05, ..., -2.8074e-05, + 1.0967e-04, 6.8092e-04], + ..., + [ 1.3039e-06, 5.0640e-04, 1.7762e-04, ..., 1.3532e-06, + 1.8156e-04, 3.7360e-04], + [ 3.1531e-05, 7.2575e-04, 3.0160e-04, ..., 2.4904e-06, + 2.5797e-04, 7.8583e-04], + [ 7.2457e-06, -1.2255e-03, -6.8712e-04, ..., 2.6450e-06, + -2.4390e-04, 7.4053e-04]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0120, 0.0101, -0.0012, 0.0148, -0.0097, -0.0088, 0.0113, 0.0218, + -0.0316, 0.0496], device='cuda:0'), grad: tensor([-0.0078, 0.0168, -0.0132, 0.0217, -0.0741, 0.0168, 0.0192, -0.0136, + 0.0196, 0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 216.98, cls_loss 0.4923 cls_loss_mapping 0.0031 cls_loss_causal 0.4710 re_mapping 0.0059 re_causal 0.0147 /// teacc 98.97 lr 0.00010000 +Epoch 398, weight, value: tensor([[-7.9122e-02, 7.2009e-02, -1.1555e-01, ..., -1.2407e-01, + -1.2480e-01, 2.3701e-02], + [-8.7489e-02, -1.7872e-01, 8.4283e-03, ..., -5.2789e-02, + -4.0200e-02, -1.1851e-01], + [ 3.0251e-02, -1.0188e-01, 5.5779e-02, ..., 1.6209e-01, + -1.4541e-01, -3.9714e-02], + ..., + [-1.2313e-01, -2.1566e-01, 8.9180e-02, ..., -1.6824e-02, + -5.3718e-02, 7.4880e-02], + [ 2.9959e-02, 3.8953e-02, -3.6796e-02, ..., -9.1124e-02, + -1.0394e-01, -2.1862e-02], + [-1.6177e-01, 2.1953e-05, -1.0074e-03, ..., -1.8227e-01, + 1.1648e-01, 2.7929e-02]], device='cuda:0'), grad: tensor([[ 1.1152e-04, 2.1248e-03, 7.9200e-06, ..., 2.5004e-05, + 3.6788e-04, 3.3855e-04], + [ 8.5056e-05, 1.9968e-04, 3.6407e-04, ..., 1.1539e-03, + 1.2290e-04, 8.1897e-05], + [ 2.8114e-03, 1.3943e-03, -9.2793e-04, ..., -2.9488e-03, + 1.0502e-04, 4.3893e-04], + ..., + [ 2.3770e-04, 4.0627e-04, 2.7490e-04, ..., 8.9407e-04, + 2.1040e-04, 4.2191e-03], + [ 1.3390e-02, 6.1378e-03, 6.2764e-05, ..., 1.9836e-04, + 1.5507e-03, 6.1417e-04], + [ 1.0800e-04, 6.3801e-04, 1.4819e-05, ..., 4.2558e-05, + 3.6812e-04, 9.7752e-04]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0119, 0.0109, -0.0028, 0.0142, -0.0081, -0.0084, 0.0118, 0.0214, + -0.0320, 0.0495], device='cuda:0'), grad: tensor([ 0.0094, 0.0088, 0.0081, 0.0062, -0.0225, -0.0226, 0.0083, 0.0270, + 0.0267, -0.0493], device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 217.05, cls_loss 0.4541 cls_loss_mapping 0.0020 cls_loss_causal 0.4298 re_mapping 0.0064 re_causal 0.0166 /// teacc 98.98 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.0797, 0.0726, -0.1156, ..., -0.1245, -0.1251, 0.0224], + [-0.0872, -0.1792, 0.0082, ..., -0.0535, -0.0395, -0.1169], + [ 0.0286, -0.1021, 0.0574, ..., 0.1631, -0.1441, -0.0386], + ..., + [-0.1236, -0.2155, 0.0877, ..., -0.0167, -0.0547, 0.0743], + [ 0.0302, 0.0386, -0.0381, ..., -0.0912, -0.1045, -0.0233], + [-0.1621, -0.0014, -0.0021, ..., -0.1832, 0.1160, 0.0271]], + device='cuda:0'), grad: tensor([[-1.1332e-05, -1.0967e-04, 1.1021e-04, ..., 7.6652e-05, + 2.9874e-04, 3.8671e-04], + [ 1.0971e-06, 7.1013e-07, 3.3522e-04, ..., 1.2755e-04, + -5.6572e-03, -4.3373e-03], + [-4.6825e-04, -1.6689e-04, 8.2684e-04, ..., 5.0396e-05, + 5.5599e-04, 2.6436e-03], + ..., + [ 1.4439e-05, 6.4611e-05, -4.1542e-03, ..., -3.1586e-03, + 1.3123e-03, -4.8370e-03], + [ 2.7180e-05, 2.8536e-05, 4.9561e-05, ..., 3.5381e-04, + 7.0333e-04, 5.2261e-04], + [-1.2249e-05, -2.1800e-05, 4.9770e-05, ..., 1.4973e-04, + 1.8635e-03, 1.9875e-03]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0113, 0.0099, -0.0013, 0.0142, -0.0085, -0.0084, 0.0128, 0.0211, + -0.0324, 0.0496], device='cuda:0'), grad: tensor([ 0.0110, -0.0019, -0.0063, -0.0410, 0.0136, 0.0136, 0.0170, -0.0058, + -0.0203, 0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 217.13, cls_loss 0.4537 cls_loss_mapping 0.0025 cls_loss_causal 0.4289 re_mapping 0.0059 re_causal 0.0153 /// teacc 98.95 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.0798, 0.0729, -0.1162, ..., -0.1242, -0.1249, 0.0223], + [-0.0873, -0.1792, 0.0088, ..., -0.0538, -0.0392, -0.1169], + [ 0.0293, -0.1019, 0.0577, ..., 0.1646, -0.1441, -0.0382], + ..., + [-0.1243, -0.2167, 0.0882, ..., -0.0172, -0.0555, 0.0744], + [ 0.0294, 0.0388, -0.0376, ..., -0.0910, -0.1050, -0.0239], + [-0.1619, -0.0010, -0.0024, ..., -0.1840, 0.1172, 0.0283]], + device='cuda:0'), grad: tensor([[ 2.5768e-03, 5.0497e-04, 3.7169e-04, ..., 3.7861e-04, + 3.8004e-04, 1.9627e-03], + [ 2.0468e-04, 2.0400e-05, -1.0405e-03, ..., -2.0528e-04, + -4.5872e-04, 6.1178e-04], + [ 9.9945e-04, 2.9430e-05, 2.6011e-04, ..., -8.0776e-04, + 2.1219e-04, 8.8787e-04], + ..., + [ 8.8692e-05, 2.5630e-04, 5.4926e-05, ..., 2.0552e-04, + 6.3753e-04, 8.1396e-04], + [ 2.2629e-02, 1.7120e-02, 2.3222e-04, ..., 3.4332e-04, + 3.6526e-04, 8.9836e-04], + [ 6.9904e-04, -7.6199e-04, -1.2035e-03, ..., -5.4359e-04, + -2.7199e-03, -2.7561e-03]], device='cuda:0') +Epoch 400, bias, value: tensor([ 9.9376e-03, 1.1229e-02, 6.8231e-05, 1.3903e-02, -8.9988e-03, + -6.8252e-03, 1.1963e-02, 2.0408e-02, -3.3853e-02, 5.0367e-02], + device='cuda:0'), grad: tensor([ 0.0231, -0.0153, 0.0166, -0.0311, 0.0223, -0.0654, 0.0218, 0.0145, + 0.0309, -0.0175], device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 216.87, cls_loss 0.4663 cls_loss_mapping 0.0026 cls_loss_causal 0.4449 re_mapping 0.0059 re_causal 0.0149 /// teacc 98.89 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.0807, 0.0721, -0.1169, ..., -0.1240, -0.1256, 0.0225], + [-0.0880, -0.1789, 0.0083, ..., -0.0532, -0.0396, -0.1167], + [ 0.0302, -0.1006, 0.0563, ..., 0.1639, -0.1434, -0.0388], + ..., + [-0.1242, -0.2171, 0.0892, ..., -0.0149, -0.0551, 0.0758], + [ 0.0298, 0.0387, -0.0367, ..., -0.0917, -0.1051, -0.0240], + [-0.1627, -0.0016, -0.0018, ..., -0.1836, 0.1178, 0.0295]], + device='cuda:0'), grad: tensor([[-1.3068e-05, -2.0194e-04, 1.9157e-04, ..., 1.5903e-04, + 1.0207e-05, 1.3959e-04], + [-2.4681e-03, -4.4136e-03, 1.7095e-04, ..., 2.1303e-04, + 1.6737e-04, -9.1791e-04], + [-7.0610e-03, 8.1539e-04, -1.7023e-03, ..., -2.7771e-03, + 1.8096e-04, -1.3428e-02], + ..., + [ 3.1519e-04, 1.1122e-04, -9.4318e-04, ..., 1.9681e-04, + -5.7411e-04, -5.2547e-04], + [ 3.0975e-03, 3.2234e-04, 1.5581e-04, ..., 8.2827e-04, + 3.1090e-04, 5.1765e-03], + [-2.6054e-03, -4.3983e-03, 1.0500e-03, ..., 1.4281e-04, + -6.8970e-03, 1.2798e-03]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0104, 0.0109, -0.0008, 0.0140, -0.0086, -0.0064, 0.0110, 0.0212, + -0.0341, 0.0507], device='cuda:0'), grad: tensor([-0.0195, -0.0172, -0.0220, 0.0156, 0.0163, 0.0141, 0.0178, 0.0092, + 0.0231, -0.0376], device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 216.75, cls_loss 0.4600 cls_loss_mapping 0.0020 cls_loss_causal 0.4364 re_mapping 0.0063 re_causal 0.0164 /// teacc 98.99 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.0807, 0.0721, -0.1170, ..., -0.1242, -0.1256, 0.0227], + [-0.0880, -0.1790, 0.0084, ..., -0.0533, -0.0394, -0.1167], + [ 0.0303, -0.1006, 0.0562, ..., 0.1638, -0.1435, -0.0388], + ..., + [-0.1244, -0.2173, 0.0891, ..., -0.0150, -0.0550, 0.0757], + [ 0.0296, 0.0388, -0.0365, ..., -0.0918, -0.1051, -0.0241], + [-0.1627, -0.0015, -0.0016, ..., -0.1833, 0.1179, 0.0295]], + device='cuda:0'), grad: tensor([[ 7.1859e-04, -8.2064e-04, 6.5184e-04, ..., 2.7776e-04, + 4.6992e-04, 5.6839e-04], + [-6.9809e-03, 7.8440e-05, 9.7609e-04, ..., -8.6823e-03, + 1.4496e-04, 1.0147e-03], + [ 1.3895e-03, 1.1015e-03, -2.2566e-04, ..., -2.1172e-03, + -1.0496e-04, 5.7650e-04], + ..., + [ 2.9907e-03, 1.0508e-04, 1.9102e-03, ..., 7.5493e-03, + 8.9931e-04, 4.7531e-03], + [ 2.9469e-03, 3.1328e-04, 2.7924e-03, ..., 3.1586e-03, + 3.5553e-03, 3.3493e-03], + [ 1.0748e-03, 1.8227e-04, 2.5272e-03, ..., 2.9335e-03, + 8.3208e-04, 1.4267e-03]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0103, 0.0109, -0.0008, 0.0142, -0.0084, -0.0066, 0.0110, 0.0211, + -0.0341, 0.0507], device='cuda:0'), grad: tensor([-0.0252, -0.0159, -0.0118, 0.0199, -0.0446, 0.0071, -0.0179, 0.0214, + 0.0426, 0.0244], device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 216.75, cls_loss 0.4561 cls_loss_mapping 0.0016 cls_loss_causal 0.4315 re_mapping 0.0058 re_causal 0.0148 /// teacc 99.00 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.0807, 0.0718, -0.1171, ..., -0.1243, -0.1256, 0.0226], + [-0.0880, -0.1790, 0.0085, ..., -0.0532, -0.0392, -0.1164], + [ 0.0305, -0.1006, 0.0562, ..., 0.1638, -0.1435, -0.0390], + ..., + [-0.1243, -0.2173, 0.0891, ..., -0.0151, -0.0550, 0.0756], + [ 0.0297, 0.0389, -0.0366, ..., -0.0917, -0.1051, -0.0241], + [-0.1627, -0.0013, -0.0016, ..., -0.1833, 0.1180, 0.0297]], + device='cuda:0'), grad: tensor([[-4.9949e-05, -1.6699e-03, -8.6498e-04, ..., -3.1447e-04, + 6.9380e-05, 1.7846e-04], + [ 6.9290e-06, 3.9399e-05, 2.1562e-05, ..., 8.4019e-04, + 7.1935e-06, 1.0866e-04], + [ 3.0413e-05, 3.0613e-04, 2.7800e-04, ..., 6.4433e-05, + 2.0444e-05, -5.0879e-04], + ..., + [ 2.3544e-05, 8.6927e-04, 9.6369e-04, ..., -9.2888e-04, + 2.9802e-04, 5.9032e-04], + [-7.0429e-04, -9.9242e-05, 1.9383e-04, ..., 4.5985e-05, + -2.8801e-04, 1.3459e-04], + [ 1.2493e-04, -6.5422e-04, -1.8444e-03, ..., 1.8013e-04, + -1.5526e-02, -1.1650e-02]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0103, 0.0110, -0.0008, 0.0141, -0.0084, -0.0065, 0.0110, 0.0211, + -0.0343, 0.0507], device='cuda:0'), grad: tensor([-0.0022, -0.0064, -0.0213, 0.0108, 0.0341, 0.0057, -0.0083, 0.0086, + 0.0078, -0.0287], device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 216.84, cls_loss 0.4733 cls_loss_mapping 0.0013 cls_loss_causal 0.4461 re_mapping 0.0054 re_causal 0.0146 /// teacc 98.99 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.0806, 0.0719, -0.1172, ..., -0.1245, -0.1255, 0.0227], + [-0.0879, -0.1790, 0.0088, ..., -0.0532, -0.0390, -0.1162], + [ 0.0306, -0.1006, 0.0563, ..., 0.1639, -0.1436, -0.0389], + ..., + [-0.1244, -0.2174, 0.0892, ..., -0.0150, -0.0550, 0.0756], + [ 0.0296, 0.0389, -0.0367, ..., -0.0919, -0.1051, -0.0242], + [-0.1627, -0.0014, -0.0018, ..., -0.1834, 0.1179, 0.0296]], + device='cuda:0'), grad: tensor([[-1.1347e-05, 6.6185e-04, 1.1230e-04, ..., 1.5640e-03, + 3.6049e-04, 2.4414e-03], + [ 9.0227e-06, 1.3318e-06, 4.0460e-04, ..., 1.8632e-04, + 1.8871e-04, 1.6241e-03], + [-5.2005e-05, -7.4673e-04, -4.5657e-04, ..., -1.7900e-03, + 4.1485e-04, 8.2111e-04], + ..., + [ 3.9458e-05, 6.6906e-06, -9.5987e-04, ..., 4.4197e-05, + 1.4198e-04, 8.1329e-03], + [-3.3706e-05, 3.6031e-05, 6.9916e-05, ..., 4.5478e-05, + -2.1000e-03, -5.6419e-03], + [ 4.4763e-05, 3.3200e-05, 1.9431e-04, ..., -4.6301e-04, + -3.4833e-04, 1.4868e-03]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0102, 0.0111, -0.0007, 0.0141, -0.0084, -0.0065, 0.0111, 0.0209, + -0.0342, 0.0506], device='cuda:0'), grad: tensor([-0.0423, 0.0145, 0.0125, 0.0199, -0.0137, 0.0090, -0.0156, 0.0355, + -0.0104, -0.0095], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 403---------------------------------------------------- +epoch 403, time 217.73, cls_loss 0.4734 cls_loss_mapping 0.0017 cls_loss_causal 0.4506 re_mapping 0.0052 re_causal 0.0142 /// teacc 99.04 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.0807, 0.0718, -0.1172, ..., -0.1245, -0.1255, 0.0226], + [-0.0879, -0.1790, 0.0089, ..., -0.0531, -0.0390, -0.1161], + [ 0.0307, -0.1006, 0.0565, ..., 0.1639, -0.1437, -0.0390], + ..., + [-0.1246, -0.2175, 0.0891, ..., -0.0150, -0.0552, 0.0754], + [ 0.0296, 0.0388, -0.0369, ..., -0.0919, -0.1050, -0.0241], + [-0.1629, -0.0015, -0.0018, ..., -0.1834, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[ 1.8656e-04, 1.2919e-05, 8.8751e-05, ..., 1.0788e-04, + 1.5008e-04, -9.7656e-03], + [ 6.7711e-04, 1.3951e-06, 1.0148e-05, ..., 9.8050e-06, + 2.3746e-03, 6.3782e-03], + [ 8.0109e-05, -1.2302e-04, -9.0599e-04, ..., -1.1263e-03, + 5.6839e-04, 4.1504e-03], + ..., + [ 3.6478e-04, 1.0830e-04, 2.2936e-04, ..., 1.4210e-04, + 4.7660e-04, 1.6022e-03], + [ 1.8377e-03, 1.1757e-05, 5.9992e-05, ..., 6.6400e-05, + 1.5616e-04, 8.0729e-04], + [ 1.0914e-04, -4.6700e-05, 6.0177e-04, ..., 6.9237e-04, + 5.1111e-05, 9.7275e-04]], device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0104, 0.0110, -0.0007, 0.0140, -0.0084, -0.0064, 0.0110, 0.0208, + -0.0341, 0.0506], device='cuda:0'), grad: tensor([-0.0087, 0.0157, 0.0268, -0.0349, -0.0081, -0.0199, -0.0053, 0.0235, + 0.0199, -0.0091], device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 217.04, cls_loss 0.4584 cls_loss_mapping 0.0012 cls_loss_causal 0.4394 re_mapping 0.0050 re_causal 0.0137 /// teacc 99.03 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.0807, 0.0718, -0.1171, ..., -0.1247, -0.1254, 0.0227], + [-0.0879, -0.1791, 0.0090, ..., -0.0531, -0.0390, -0.1161], + [ 0.0306, -0.1005, 0.0566, ..., 0.1639, -0.1437, -0.0390], + ..., + [-0.1245, -0.2175, 0.0890, ..., -0.0152, -0.0552, 0.0755], + [ 0.0294, 0.0388, -0.0369, ..., -0.0919, -0.1050, -0.0243], + [-0.1629, -0.0015, -0.0018, ..., -0.1834, 0.1178, 0.0295]], + device='cuda:0'), grad: tensor([[ 3.2282e-04, 1.9300e-04, -3.3826e-06, ..., 1.5759e-04, + 1.0662e-03, 1.4467e-03], + [ 1.5056e-04, 7.3552e-05, 1.9744e-06, ..., 1.1528e-04, + 4.8232e-04, 1.1005e-03], + [-9.6941e-04, 2.8682e-04, 1.4126e-04, ..., -1.0290e-03, + -5.0201e-03, -1.1414e-02], + ..., + [ 2.0742e-04, 1.5604e-04, -1.5247e-04, ..., -5.1379e-05, + 5.2214e-04, -2.9588e-04], + [ 8.7261e-04, 7.7486e-04, 1.9819e-06, ..., 2.1112e-04, + 1.7996e-03, 3.6526e-03], + [ 7.7820e-04, 8.0013e-04, 2.6263e-06, ..., 1.5950e-04, + 1.3342e-03, 1.9817e-03]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0103, 0.0110, -0.0007, 0.0141, -0.0084, -0.0063, 0.0109, 0.0209, + -0.0340, 0.0505], device='cuda:0'), grad: tensor([ 0.0154, -0.0153, -0.0756, -0.0009, 0.0183, -0.0129, 0.0197, 0.0127, + 0.0202, 0.0184], device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 216.55, cls_loss 0.4670 cls_loss_mapping 0.0013 cls_loss_causal 0.4470 re_mapping 0.0049 re_causal 0.0133 /// teacc 98.99 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.0806, 0.0718, -0.1172, ..., -0.1248, -0.1253, 0.0228], + [-0.0879, -0.1792, 0.0090, ..., -0.0531, -0.0389, -0.1160], + [ 0.0306, -0.1006, 0.0567, ..., 0.1639, -0.1435, -0.0390], + ..., + [-0.1245, -0.2175, 0.0888, ..., -0.0153, -0.0553, 0.0755], + [ 0.0295, 0.0389, -0.0369, ..., -0.0919, -0.1049, -0.0243], + [-0.1629, -0.0015, -0.0017, ..., -0.1835, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[ 2.0778e-04, 7.1478e-04, 8.2850e-05, ..., 1.8418e-05, + 4.5872e-04, 3.4422e-05], + [ 1.3363e-04, 4.0102e-04, 4.7147e-05, ..., 1.0282e-05, + 8.6355e-04, 1.6856e-04], + [ 7.3385e-04, 7.6294e-04, 1.0401e-04, ..., -4.0531e-05, + 4.6897e-04, 4.1038e-05], + ..., + [ 8.2302e-04, 1.5516e-03, 1.8473e-03, ..., 1.1196e-03, + 3.4695e-03, 2.0504e-03], + [-3.6697e-03, 9.9087e-04, -9.1982e-04, ..., 1.9860e-04, + -5.2452e-03, -1.5345e-03], + [ 6.3276e-04, -3.0136e-04, -2.6321e-03, ..., -1.7700e-03, + -4.4365e-03, -4.0474e-03]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0104, 0.0111, -0.0006, 0.0140, -0.0083, -0.0063, 0.0109, 0.0208, + -0.0340, 0.0504], device='cuda:0'), grad: tensor([ 0.0096, 0.0167, 0.0107, -0.0485, 0.0170, 0.0223, -0.0063, 0.0183, + -0.0435, 0.0035], device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 216.53, cls_loss 0.4758 cls_loss_mapping 0.0013 cls_loss_causal 0.4544 re_mapping 0.0048 re_causal 0.0133 /// teacc 98.97 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.0806, 0.0719, -0.1174, ..., -0.1249, -0.1254, 0.0228], + [-0.0878, -0.1792, 0.0091, ..., -0.0531, -0.0389, -0.1160], + [ 0.0306, -0.1006, 0.0567, ..., 0.1639, -0.1435, -0.0390], + ..., + [-0.1245, -0.2173, 0.0887, ..., -0.0154, -0.0552, 0.0755], + [ 0.0296, 0.0389, -0.0370, ..., -0.0919, -0.1049, -0.0243], + [-0.1629, -0.0015, -0.0016, ..., -0.1834, 0.1178, 0.0297]], + device='cuda:0'), grad: tensor([[ 2.5481e-06, 2.9188e-06, 5.7071e-05, ..., 5.8025e-05, + 6.9523e-04, 6.5374e-04], + [-1.4901e-07, 7.0687e-07, 1.0643e-03, ..., 6.0177e-04, + 2.3117e-03, 1.7300e-03], + [ 2.0154e-06, 2.2762e-06, 3.0589e-04, ..., -4.4584e-04, + -6.1188e-03, -5.0201e-03], + ..., + [ 2.9411e-06, 7.1563e-06, -3.2043e-03, ..., -1.0481e-03, + 1.3943e-03, 8.2636e-04], + [ 9.2983e-06, 2.5690e-05, 3.2949e-04, ..., 2.1791e-04, + 1.3676e-03, 1.1320e-03], + [ 2.9895e-07, -2.0161e-05, 1.1444e-03, ..., 9.7132e-04, + 3.9368e-03, 3.4542e-03]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0105, 0.0111, -0.0006, 0.0139, -0.0083, -0.0064, 0.0106, 0.0210, + -0.0340, 0.0504], device='cuda:0'), grad: tensor([ 0.0206, 0.0354, -0.0078, 0.0269, -0.0053, -0.0090, -0.0046, 0.0110, + -0.0069, -0.0604], device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 216.44, cls_loss 0.4767 cls_loss_mapping 0.0014 cls_loss_causal 0.4573 re_mapping 0.0049 re_causal 0.0132 /// teacc 98.97 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.0807, 0.0718, -0.1174, ..., -0.1249, -0.1253, 0.0227], + [-0.0878, -0.1793, 0.0090, ..., -0.0532, -0.0389, -0.1161], + [ 0.0305, -0.1007, 0.0567, ..., 0.1639, -0.1436, -0.0390], + ..., + [-0.1244, -0.2173, 0.0887, ..., -0.0153, -0.0552, 0.0756], + [ 0.0298, 0.0390, -0.0370, ..., -0.0919, -0.1050, -0.0244], + [-0.1630, -0.0015, -0.0016, ..., -0.1834, 0.1178, 0.0298]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0013, 0.0002, ..., 0.0004, 0.0005, 0.0008], + [-0.0035, -0.0014, -0.0002, ..., -0.0025, -0.0034, -0.0033], + [ 0.0010, 0.0003, 0.0005, ..., 0.0006, 0.0009, 0.0018], + ..., + [ 0.0005, 0.0005, 0.0008, ..., 0.0004, 0.0006, 0.0013], + [ 0.0023, -0.0046, 0.0021, ..., 0.0011, 0.0022, 0.0026], + [-0.0019, -0.0020, -0.0021, ..., -0.0008, -0.0023, -0.0012]], + device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0106, 0.0111, -0.0007, 0.0139, -0.0083, -0.0064, 0.0105, 0.0212, + -0.0341, 0.0504], device='cuda:0'), grad: tensor([ 0.0228, -0.0778, 0.0249, -0.0311, -0.0096, 0.0299, 0.0103, 0.0248, + -0.0028, 0.0087], device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 216.45, cls_loss 0.4840 cls_loss_mapping 0.0012 cls_loss_causal 0.4471 re_mapping 0.0047 re_causal 0.0131 /// teacc 98.99 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.0807, 0.0719, -0.1173, ..., -0.1249, -0.1255, 0.0227], + [-0.0878, -0.1794, 0.0090, ..., -0.0532, -0.0389, -0.1161], + [ 0.0306, -0.1005, 0.0567, ..., 0.1638, -0.1437, -0.0392], + ..., + [-0.1243, -0.2174, 0.0887, ..., -0.0153, -0.0552, 0.0756], + [ 0.0299, 0.0391, -0.0371, ..., -0.0920, -0.1048, -0.0242], + [-0.1631, -0.0015, -0.0015, ..., -0.1833, 0.1178, 0.0298]], + device='cuda:0'), grad: tensor([[ 0.0015, 0.0025, 0.0007, ..., 0.0020, 0.0006, 0.0019], + [ 0.0002, 0.0001, 0.0006, ..., 0.0012, 0.0012, 0.0018], + [-0.0107, -0.0092, -0.0008, ..., -0.0014, -0.0026, -0.0012], + ..., + [ 0.0004, 0.0004, 0.0011, ..., 0.0116, 0.0015, 0.0015], + [-0.0106, 0.0019, 0.0005, ..., 0.0013, -0.0058, -0.0070], + [ 0.0137, 0.0004, -0.0010, ..., 0.0004, 0.0066, 0.0087]], + device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0106, 0.0111, -0.0007, 0.0138, -0.0084, -0.0063, 0.0104, 0.0212, + -0.0339, 0.0504], device='cuda:0'), grad: tensor([ 0.0322, -0.0382, -0.0155, -0.0294, -0.0227, 0.0188, -0.0091, 0.0486, + -0.0027, 0.0182], device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 216.74, cls_loss 0.4393 cls_loss_mapping 0.0011 cls_loss_causal 0.4147 re_mapping 0.0047 re_causal 0.0132 /// teacc 99.00 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.0806, 0.0719, -0.1175, ..., -0.1250, -0.1254, 0.0228], + [-0.0879, -0.1794, 0.0091, ..., -0.0531, -0.0389, -0.1161], + [ 0.0307, -0.1006, 0.0569, ..., 0.1640, -0.1437, -0.0391], + ..., + [-0.1244, -0.2175, 0.0887, ..., -0.0153, -0.0553, 0.0756], + [ 0.0299, 0.0391, -0.0372, ..., -0.0921, -0.1048, -0.0241], + [-0.1632, -0.0014, -0.0015, ..., -0.1834, 0.1179, 0.0298]], + device='cuda:0'), grad: tensor([[-6.8098e-06, -2.1264e-05, 4.4703e-08, ..., 7.4506e-08, + 1.7667e-04, 2.3699e-04], + [ 1.4775e-05, 3.2224e-07, 2.4587e-06, ..., 4.3586e-06, + 3.1090e-04, 4.0054e-04], + [ 5.8323e-05, 1.6674e-05, -1.1504e-05, ..., -1.6227e-05, + 1.8191e-04, 2.3675e-04], + ..., + [-2.8496e-03, 1.8310e-06, 1.1474e-05, ..., 2.1562e-05, + -2.7981e-03, 3.5191e-04], + [ 5.5084e-03, 9.0170e-04, 3.5334e-06, ..., 5.2303e-06, + 1.4896e-03, 2.4605e-04], + [ 1.6003e-03, 3.0473e-06, 2.7359e-05, ..., 3.9637e-05, + 2.0409e-03, 3.9458e-04]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0107, 0.0111, -0.0006, 0.0138, -0.0083, -0.0063, 0.0103, 0.0211, + -0.0339, 0.0502], device='cuda:0'), grad: tensor([-0.0141, 0.0320, -0.0002, -0.0247, 0.0155, 0.0174, 0.0178, -0.0285, + 0.0235, -0.0386], device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 217.04, cls_loss 0.4514 cls_loss_mapping 0.0008 cls_loss_causal 0.4212 re_mapping 0.0047 re_causal 0.0132 /// teacc 98.97 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.0806, 0.0720, -0.1176, ..., -0.1251, -0.1255, 0.0227], + [-0.0878, -0.1792, 0.0091, ..., -0.0533, -0.0389, -0.1160], + [ 0.0307, -0.1006, 0.0567, ..., 0.1640, -0.1437, -0.0391], + ..., + [-0.1243, -0.2175, 0.0886, ..., -0.0153, -0.0553, 0.0756], + [ 0.0299, 0.0391, -0.0371, ..., -0.0919, -0.1047, -0.0242], + [-0.1632, -0.0013, -0.0014, ..., -0.1834, 0.1180, 0.0298]], + device='cuda:0'), grad: tensor([[ 5.4017e-06, -3.1758e-06, 3.1460e-06, ..., 5.7966e-06, + 5.5507e-07, 2.0489e-08], + [ 7.5996e-07, 7.8324e-07, 2.2911e-06, ..., 2.2799e-06, + 1.2107e-07, 2.7940e-08], + [-5.3138e-05, 3.2429e-06, 9.1791e-04, ..., 1.0347e-03, + 2.6990e-06, 1.1679e-06], + ..., + [ 3.5297e-06, 1.4929e-06, -1.2445e-03, ..., -1.4391e-03, + 4.2133e-06, 8.0932e-07], + [ 3.5238e-04, 7.7438e-04, 7.4990e-06, ..., 1.0826e-05, + -1.0896e-07, -1.7006e-06], + [ 3.6024e-06, 1.3607e-06, 2.6608e-04, ..., 3.0994e-04, + -7.4022e-06, 7.8790e-07]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0108, 0.0111, -0.0006, 0.0139, -0.0084, -0.0063, 0.0102, 0.0211, + -0.0340, 0.0504], device='cuda:0'), grad: tensor([ 0.0065, 0.0065, -0.0237, 0.0065, 0.0058, 0.0017, 0.0050, 0.0051, + -0.0208, 0.0073], device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 216.93, cls_loss 0.4646 cls_loss_mapping 0.0010 cls_loss_causal 0.4313 re_mapping 0.0048 re_causal 0.0131 /// teacc 98.97 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.0807, 0.0720, -0.1178, ..., -0.1252, -0.1256, 0.0228], + [-0.0878, -0.1791, 0.0092, ..., -0.0532, -0.0388, -0.1160], + [ 0.0307, -0.1006, 0.0567, ..., 0.1640, -0.1437, -0.0392], + ..., + [-0.1242, -0.2175, 0.0886, ..., -0.0154, -0.0554, 0.0755], + [ 0.0299, 0.0390, -0.0371, ..., -0.0917, -0.1048, -0.0244], + [-0.1632, -0.0013, -0.0014, ..., -0.1834, 0.1181, 0.0298]], + device='cuda:0'), grad: tensor([[ 4.3130e-04, -1.0262e-03, 9.0027e-04, ..., 1.1473e-03, + -4.9067e-04, 2.6679e-04], + [ 1.1481e-05, 2.6181e-05, 5.4657e-05, ..., 5.5552e-05, + 3.7742e-04, 1.3101e-04], + [-1.7405e-03, 1.6603e-03, -1.3790e-03, ..., 8.8501e-03, + 4.0555e-04, -3.8052e-04], + ..., + [ 1.0854e-04, 2.3651e-04, 2.0847e-03, ..., 1.4391e-03, + 5.0783e-04, 1.4973e-04], + [ 1.3351e-04, -3.5744e-03, -3.7174e-03, ..., -1.4542e-02, + 3.8052e-04, 1.1778e-04], + [ 6.6710e-04, 9.5963e-05, 4.6301e-04, ..., -1.7989e-04, + -1.6296e-04, 1.4234e-04]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0107, 0.0113, -0.0005, 0.0139, -0.0083, -0.0064, 0.0102, 0.0211, + -0.0340, 0.0503], device='cuda:0'), grad: tensor([ 0.0012, -0.0093, -0.0027, -0.0004, -0.0128, 0.0149, -0.0074, 0.0265, + -0.0092, -0.0008], device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 217.75, cls_loss 0.4522 cls_loss_mapping 0.0011 cls_loss_causal 0.4228 re_mapping 0.0047 re_causal 0.0131 /// teacc 98.97 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.0807, 0.0720, -0.1178, ..., -0.1253, -0.1255, 0.0227], + [-0.0877, -0.1792, 0.0090, ..., -0.0532, -0.0390, -0.1160], + [ 0.0306, -0.1007, 0.0567, ..., 0.1638, -0.1437, -0.0393], + ..., + [-0.1243, -0.2176, 0.0887, ..., -0.0154, -0.0553, 0.0757], + [ 0.0299, 0.0391, -0.0371, ..., -0.0916, -0.1047, -0.0244], + [-0.1632, -0.0012, -0.0013, ..., -0.1833, 0.1181, 0.0298]], + device='cuda:0'), grad: tensor([[ 2.8610e-05, 2.2659e-03, 6.2943e-04, ..., 6.9588e-06, + 1.5736e-04, 1.5274e-05], + [ 6.0272e-04, 1.1654e-03, 3.9577e-04, ..., 1.6189e-04, + 1.0462e-03, 2.5344e-04], + [ 1.8053e-03, 1.6003e-03, 2.3258e-04, ..., 6.8808e-04, + 3.3021e-04, 6.8069e-05], + ..., + [ 4.2915e-05, 7.2002e-04, 4.9019e-04, ..., -1.0408e-05, + -2.2087e-03, -1.6487e-04], + [-1.7567e-03, -9.2173e-04, 4.0841e-04, ..., -8.0347e-04, + 3.4285e-04, 7.3075e-05], + [ 1.1015e-04, -2.9755e-03, -4.0054e-03, ..., 4.6849e-05, + 6.0797e-04, 1.3685e-04]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0106, 0.0112, -0.0007, 0.0141, -0.0083, -0.0064, 0.0102, 0.0211, + -0.0339, 0.0503], device='cuda:0'), grad: tensor([ 0.0171, 0.0160, 0.0126, 0.0047, 0.0132, 0.0104, 0.0016, -0.0522, + 0.0054, -0.0288], device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 217.63, cls_loss 0.4481 cls_loss_mapping 0.0009 cls_loss_causal 0.4189 re_mapping 0.0046 re_causal 0.0132 /// teacc 98.95 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.0807, 0.0721, -0.1177, ..., -0.1253, -0.1256, 0.0228], + [-0.0878, -0.1793, 0.0090, ..., -0.0533, -0.0390, -0.1161], + [ 0.0305, -0.1008, 0.0566, ..., 0.1639, -0.1437, -0.0393], + ..., + [-0.1243, -0.2175, 0.0888, ..., -0.0154, -0.0553, 0.0756], + [ 0.0301, 0.0394, -0.0370, ..., -0.0916, -0.1047, -0.0243], + [-0.1633, -0.0012, -0.0014, ..., -0.1834, 0.1180, 0.0297]], + device='cuda:0'), grad: tensor([[ 6.1607e-04, 1.7178e-04, 2.5615e-05, ..., -1.8626e-09, + 1.1599e-04, 6.8569e-04], + [ 5.2261e-04, 8.0839e-06, 1.1390e-04, ..., 9.3132e-10, + -1.9369e-03, 5.2452e-04], + [ 9.1267e-04, 4.3631e-05, 2.3901e-04, ..., -6.7987e-08, + 1.8626e-05, 3.0589e-04], + ..., + [-4.6802e-04, 2.7716e-05, -5.6791e-04, ..., 1.8626e-08, + 2.7180e-05, 3.0160e-04], + [ 1.5402e-03, 2.3234e-04, 3.6931e-04, ..., 2.7940e-08, + 7.3671e-04, 4.7922e-04], + [ 4.9829e-04, 1.5497e-04, 3.8803e-05, ..., 5.5879e-09, + 1.4126e-04, 5.1832e-04]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0107, 0.0112, -0.0008, 0.0140, -0.0082, -0.0064, 0.0102, 0.0212, + -0.0338, 0.0501], device='cuda:0'), grad: tensor([ 0.0114, -0.0108, 0.0109, 0.0210, -0.0125, 0.0032, -0.0378, 0.0134, + 0.0231, -0.0219], device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 217.23, cls_loss 0.4691 cls_loss_mapping 0.0011 cls_loss_causal 0.4386 re_mapping 0.0045 re_causal 0.0128 /// teacc 98.94 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.0808, 0.0718, -0.1178, ..., -0.1254, -0.1256, 0.0229], + [-0.0878, -0.1792, 0.0089, ..., -0.0532, -0.0390, -0.1161], + [ 0.0305, -0.1009, 0.0566, ..., 0.1639, -0.1437, -0.0394], + ..., + [-0.1244, -0.2176, 0.0889, ..., -0.0155, -0.0550, 0.0757], + [ 0.0299, 0.0393, -0.0370, ..., -0.0917, -0.1049, -0.0244], + [-0.1633, -0.0012, -0.0017, ..., -0.1834, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 4.3154e-05, 1.0830e-04, 2.7609e-04, ..., 3.2902e-04, + 5.3978e-04, 3.0875e-04], + [ 6.8426e-05, 1.2541e-04, 7.0143e-04, ..., 7.5293e-04, + 8.2207e-04, 5.7268e-04], + [ 7.3075e-05, 1.0759e-04, 8.6260e-04, ..., 1.1625e-03, + -5.4884e-04, 6.7854e-04], + ..., + [ 6.0320e-05, 1.5700e-04, 9.1362e-04, ..., 9.2220e-04, + 8.0585e-04, 5.1165e-04], + [-1.1069e-04, 5.7936e-05, -3.6259e-03, ..., -3.3245e-03, + -1.4582e-03, -2.3861e-03], + [ 1.1015e-04, 3.6120e-04, 2.0294e-03, ..., 1.5316e-03, + 3.2177e-03, 2.9526e-03]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0107, 0.0113, -0.0008, 0.0140, -0.0081, -0.0064, 0.0101, 0.0213, + -0.0339, 0.0501], device='cuda:0'), grad: tensor([-0.0101, 0.0164, -0.0021, 0.0274, -0.0005, -0.0100, 0.0165, -0.0349, + -0.0047, 0.0020], device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 216.50, cls_loss 0.4476 cls_loss_mapping 0.0010 cls_loss_causal 0.4159 re_mapping 0.0045 re_causal 0.0126 /// teacc 98.98 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.0809, 0.0718, -0.1178, ..., -0.1252, -0.1256, 0.0229], + [-0.0879, -0.1791, 0.0089, ..., -0.0534, -0.0391, -0.1163], + [ 0.0305, -0.1008, 0.0567, ..., 0.1638, -0.1436, -0.0392], + ..., + [-0.1245, -0.2178, 0.0890, ..., -0.0155, -0.0550, 0.0756], + [ 0.0298, 0.0393, -0.0371, ..., -0.0917, -0.1049, -0.0245], + [-0.1632, -0.0011, -0.0017, ..., -0.1834, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 1.2312e-03, 3.9139e-03, 1.5612e-03, ..., 2.8729e-04, + 4.5090e-03, 7.4911e-04], + [ 1.1063e-04, 1.3225e-05, 1.3745e-04, ..., 1.8513e-04, + 2.5034e-04, 2.9159e-04], + [ 8.9645e-05, 4.3601e-05, 5.2261e-04, ..., 6.9380e-04, + 7.2527e-04, 6.9809e-04], + ..., + [ 9.4652e-05, 8.8215e-05, 6.7806e-04, ..., 8.7595e-04, + 9.4271e-04, 8.5163e-04], + [ 1.0099e-03, -1.1511e-03, 3.4451e-04, ..., 1.8215e-04, + 1.2093e-03, 3.1853e-04], + [-2.8210e-03, -1.0254e-02, -3.0308e-03, ..., 5.5265e-04, + -9.2697e-03, -1.4563e-03]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0107, 0.0112, -0.0007, 0.0139, -0.0081, -0.0065, 0.0102, 0.0213, + -0.0340, 0.0502], device='cuda:0'), grad: tensor([ 0.0263, 0.0178, 0.0203, 0.0210, -0.0417, 0.0182, 0.0237, -0.0102, + -0.0125, -0.0629], device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 216.34, cls_loss 0.4349 cls_loss_mapping 0.0009 cls_loss_causal 0.4006 re_mapping 0.0047 re_causal 0.0127 /// teacc 98.98 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.0809, 0.0719, -0.1178, ..., -0.1252, -0.1256, 0.0230], + [-0.0879, -0.1792, 0.0088, ..., -0.0534, -0.0391, -0.1163], + [ 0.0304, -0.1009, 0.0565, ..., 0.1638, -0.1435, -0.0393], + ..., + [-0.1244, -0.2180, 0.0890, ..., -0.0155, -0.0552, 0.0756], + [ 0.0298, 0.0394, -0.0370, ..., -0.0916, -0.1048, -0.0244], + [-0.1630, -0.0011, -0.0016, ..., -0.1833, 0.1178, 0.0297]], + device='cuda:0'), grad: tensor([[ 5.8636e-06, -1.9653e-02, 1.7077e-05, ..., -6.2026e-06, + 5.2118e-04, 2.6059e-04], + [ 7.3254e-05, 1.6617e-02, 2.2912e-04, ..., -1.0395e-04, + 7.1621e-04, -2.1744e-03], + [-5.9204e-03, -6.8903e-04, 7.8008e-06, ..., -2.4462e-04, + 5.3310e-04, -1.0616e-04], + ..., + [ 7.6175e-05, 1.3723e-03, -3.9995e-05, ..., 4.0919e-05, + 8.2541e-04, 2.9397e-04], + [-2.8825e-04, 1.0614e-03, 1.6439e-04, ..., 1.1939e-04, + -5.1928e-04, 2.0131e-05], + [ 4.5419e-05, -8.5354e-05, -8.3876e-04, ..., -4.1795e-04, + -2.3727e-03, -7.5102e-04]], device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0106, 0.0113, -0.0007, 0.0137, -0.0080, -0.0065, 0.0102, 0.0212, + -0.0340, 0.0502], device='cuda:0'), grad: tensor([-0.0123, 0.0043, 0.0101, 0.0163, -0.0440, 0.0136, 0.0204, 0.0144, + -0.0038, -0.0190], device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 216.84, cls_loss 0.4738 cls_loss_mapping 0.0009 cls_loss_causal 0.4448 re_mapping 0.0046 re_causal 0.0131 /// teacc 98.95 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.0808, 0.0719, -0.1177, ..., -0.1253, -0.1255, 0.0229], + [-0.0880, -0.1794, 0.0087, ..., -0.0534, -0.0389, -0.1162], + [ 0.0305, -0.1010, 0.0565, ..., 0.1637, -0.1436, -0.0392], + ..., + [-0.1244, -0.2181, 0.0890, ..., -0.0156, -0.0552, 0.0757], + [ 0.0300, 0.0395, -0.0370, ..., -0.0916, -0.1046, -0.0243], + [-0.1631, -0.0009, -0.0015, ..., -0.1832, 0.1177, 0.0297]], + device='cuda:0'), grad: tensor([[ 4.4137e-05, -1.9535e-05, 3.7074e-04, ..., 1.1645e-05, + 7.4040e-07, 1.0099e-03], + [ 2.9504e-05, 2.3376e-07, 2.8110e-04, ..., 2.7958e-06, + 8.7637e-07, 1.1911e-03], + [ 2.4772e-04, 2.8182e-06, 1.8997e-03, ..., -9.2685e-05, + 1.3029e-06, 5.4026e-04], + ..., + [ 1.6248e-04, 1.6643e-06, 1.4086e-03, ..., 5.5403e-05, + 4.5598e-06, 2.2483e-04], + [ 1.0318e-04, 2.9411e-06, 8.7023e-04, ..., 3.4235e-06, + 1.3234e-06, 3.0017e-04], + [ 1.7011e-04, 1.0684e-05, 1.3742e-03, ..., 3.6694e-06, + -2.9221e-05, 1.7226e-04]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0106, 0.0115, -0.0007, 0.0137, -0.0080, -0.0067, 0.0102, 0.0213, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0146, 0.0234, 0.0200, -0.0115, -0.0172, -0.0197, 0.0008, -0.0420, + 0.0156, 0.0160], device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 216.89, cls_loss 0.4288 cls_loss_mapping 0.0009 cls_loss_causal 0.3967 re_mapping 0.0046 re_causal 0.0126 /// teacc 98.93 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.0808, 0.0719, -0.1177, ..., -0.1253, -0.1255, 0.0229], + [-0.0880, -0.1793, 0.0087, ..., -0.0535, -0.0388, -0.1161], + [ 0.0305, -0.1009, 0.0565, ..., 0.1639, -0.1437, -0.0393], + ..., + [-0.1244, -0.2180, 0.0888, ..., -0.0156, -0.0552, 0.0757], + [ 0.0301, 0.0396, -0.0370, ..., -0.0915, -0.1048, -0.0244], + [-0.1631, -0.0010, -0.0013, ..., -0.1833, 0.1176, 0.0297]], + device='cuda:0'), grad: tensor([[ 1.7977e-04, 3.0875e-04, 9.5189e-05, ..., 1.5175e-04, + 2.0051e-04, 5.5045e-05], + [-1.6892e-04, 2.9325e-05, 2.4691e-05, ..., -7.6830e-05, + -2.4140e-04, -4.9978e-05], + [ 3.4237e-04, 3.8409e-04, 2.2924e-04, ..., 3.5739e-04, + 3.1257e-04, 2.5463e-04], + ..., + [ 1.8632e-04, 2.4962e-04, 1.0735e-04, ..., 1.5604e-04, + 3.9411e-04, 2.4438e-04], + [ 1.7729e-03, 2.5330e-03, 1.0192e-04, ..., 2.6202e-04, + 1.3790e-03, 3.3355e-04], + [ 4.2367e-04, 6.1560e-04, 3.9625e-04, ..., 6.9237e-04, + 6.0005e-03, 5.8327e-03]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0106, 0.0115, -0.0007, 0.0137, -0.0080, -0.0066, 0.0102, 0.0214, + -0.0336, 0.0499], device='cuda:0'), grad: tensor([-0.0202, 0.0109, -0.0506, 0.0321, -0.0079, 0.0101, -0.0203, 0.0103, + 0.0171, 0.0184], device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 216.49, cls_loss 0.4746 cls_loss_mapping 0.0009 cls_loss_causal 0.4541 re_mapping 0.0045 re_causal 0.0127 /// teacc 98.97 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.0807, 0.0719, -0.1177, ..., -0.1253, -0.1253, 0.0229], + [-0.0881, -0.1792, 0.0088, ..., -0.0534, -0.0388, -0.1162], + [ 0.0305, -0.1010, 0.0565, ..., 0.1638, -0.1437, -0.0393], + ..., + [-0.1245, -0.2182, 0.0888, ..., -0.0157, -0.0552, 0.0757], + [ 0.0300, 0.0395, -0.0372, ..., -0.0915, -0.1049, -0.0243], + [-0.1633, -0.0012, -0.0014, ..., -0.1833, 0.1177, 0.0297]], + device='cuda:0'), grad: tensor([[-0.0043, 0.0016, 0.0010, ..., 0.0002, 0.0009, -0.0136], + [ 0.0005, -0.0012, -0.0030, ..., 0.0002, -0.0015, -0.0011], + [ 0.0006, 0.0001, -0.0007, ..., 0.0003, -0.0033, -0.0013], + ..., + [ 0.0007, 0.0002, 0.0002, ..., 0.0001, 0.0013, 0.0028], + [ 0.0080, 0.0012, 0.0003, ..., 0.0002, 0.0012, 0.0027], + [-0.0064, -0.0008, 0.0008, ..., 0.0001, 0.0015, 0.0052]], + device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0106, 0.0114, -0.0007, 0.0136, -0.0080, -0.0065, 0.0103, 0.0213, + -0.0337, 0.0499], device='cuda:0'), grad: tensor([-0.0296, -0.0276, 0.0024, -0.0210, -0.0292, 0.0392, 0.0348, 0.0163, + 0.0449, -0.0300], device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 216.48, cls_loss 0.4507 cls_loss_mapping 0.0009 cls_loss_causal 0.4269 re_mapping 0.0044 re_causal 0.0122 /// teacc 98.97 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.0808, 0.0718, -0.1177, ..., -0.1253, -0.1254, 0.0228], + [-0.0882, -0.1793, 0.0088, ..., -0.0535, -0.0388, -0.1163], + [ 0.0305, -0.1010, 0.0566, ..., 0.1639, -0.1438, -0.0394], + ..., + [-0.1246, -0.2182, 0.0889, ..., -0.0158, -0.0552, 0.0758], + [ 0.0300, 0.0397, -0.0373, ..., -0.0915, -0.1050, -0.0244], + [-0.1633, -0.0012, -0.0013, ..., -0.1833, 0.1177, 0.0297]], + device='cuda:0'), grad: tensor([[ 2.9579e-05, 1.0943e-04, -1.2450e-05, ..., 6.5923e-05, + 2.5535e-04, 1.5482e-05], + [ 8.3387e-05, 3.7885e-04, 5.9605e-06, ..., 2.5272e-04, + 7.1526e-04, 1.3220e-04], + [-2.4819e-04, 2.5702e-04, 2.8044e-05, ..., -1.5135e-03, + -2.3422e-03, 4.8965e-05], + ..., + [ 3.7253e-05, 1.4172e-03, 4.7398e-04, ..., 2.0945e-04, + 2.9202e-03, 1.7929e-03], + [ 2.6450e-05, 6.4230e-04, -9.3222e-05, ..., 1.1045e-04, + 8.0872e-04, 2.7347e-04], + [ 1.2034e-04, 1.6495e-02, -1.6046e-04, ..., -8.2552e-05, + 1.5228e-02, 4.7150e-03]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0105, 0.0114, -0.0008, 0.0136, -0.0079, -0.0065, 0.0103, 0.0213, + -0.0337, 0.0500], device='cuda:0'), grad: tensor([ 0.0043, 0.0086, -0.0192, -0.0243, -0.0092, -0.0173, 0.0045, 0.0107, + 0.0063, 0.0357], device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 216.77, cls_loss 0.4445 cls_loss_mapping 0.0010 cls_loss_causal 0.4206 re_mapping 0.0045 re_causal 0.0124 /// teacc 98.99 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.0811, 0.0716, -0.1177, ..., -0.1254, -0.1255, 0.0226], + [-0.0882, -0.1791, 0.0087, ..., -0.0536, -0.0387, -0.1162], + [ 0.0307, -0.1008, 0.0567, ..., 0.1640, -0.1439, -0.0393], + ..., + [-0.1248, -0.2183, 0.0888, ..., -0.0158, -0.0550, 0.0758], + [ 0.0301, 0.0397, -0.0370, ..., -0.0915, -0.1050, -0.0245], + [-0.1631, -0.0010, -0.0014, ..., -0.1833, 0.1177, 0.0297]], + device='cuda:0'), grad: tensor([[ 4.6039e-04, 1.3285e-03, 8.9359e-04, ..., 1.4791e-03, + 3.7432e-04, 3.8600e-04], + [ 7.2837e-05, 1.3337e-05, 1.0977e-03, ..., 5.4073e-04, + 1.3304e-03, 1.4658e-03], + [-1.2329e-02, 3.8147e-04, -3.6201e-03, ..., -6.5002e-03, + 5.7268e-04, 1.1587e-03], + ..., + [ 3.9721e-04, 2.1800e-05, -1.6317e-03, ..., 3.6216e-04, + 2.8133e-04, -3.6907e-03], + [ 3.5000e-03, 1.0347e-03, 4.3750e-04, ..., 4.7731e-04, + 5.8937e-04, 6.8140e-04], + [ 1.7536e-04, 5.5283e-05, 9.1314e-04, ..., 6.2418e-04, + -6.7825e-03, -9.7132e-04]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0105, 0.0115, -0.0007, 0.0136, -0.0079, -0.0067, 0.0102, 0.0213, + -0.0338, 0.0501], device='cuda:0'), grad: tensor([ 0.0206, 0.0251, -0.0618, -0.0011, 0.0256, 0.0222, -0.0107, -0.0021, + 0.0219, -0.0396], device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 216.53, cls_loss 0.4490 cls_loss_mapping 0.0010 cls_loss_causal 0.4231 re_mapping 0.0044 re_causal 0.0125 /// teacc 98.91 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.0811, 0.0715, -0.1175, ..., -0.1253, -0.1255, 0.0227], + [-0.0881, -0.1791, 0.0088, ..., -0.0536, -0.0387, -0.1162], + [ 0.0307, -0.1008, 0.0566, ..., 0.1639, -0.1439, -0.0392], + ..., + [-0.1247, -0.2184, 0.0890, ..., -0.0156, -0.0550, 0.0759], + [ 0.0299, 0.0395, -0.0369, ..., -0.0916, -0.1051, -0.0245], + [-0.1630, -0.0010, -0.0016, ..., -0.1834, 0.1176, 0.0295]], + device='cuda:0'), grad: tensor([[ 3.4809e-05, 4.4346e-04, 2.1827e-04, ..., 4.3333e-05, + 3.9101e-04, 2.6631e-04], + [ 6.2168e-05, 1.0557e-05, -8.0299e-04, ..., -6.5470e-04, + -2.4643e-03, -1.4563e-03], + [ 7.7152e-04, 1.7568e-05, -3.1590e-06, ..., -6.3848e-04, + -1.4753e-03, 1.0967e-03], + ..., + [-2.6360e-03, 3.3140e-04, 4.6997e-03, ..., -1.5235e-04, + 7.2517e-03, 7.2289e-03], + [ 4.0436e-04, 3.0565e-04, 1.3485e-03, ..., 2.7037e-04, + 1.2674e-03, 1.3237e-03], + [-3.5691e-04, -9.5606e-04, -3.4332e-03, ..., -2.1946e-04, + -5.5847e-03, -2.2736e-03]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0106, 0.0116, -0.0007, 0.0136, -0.0079, -0.0066, 0.0102, 0.0213, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([ 0.0139, -0.0205, -0.0156, -0.0034, 0.0237, -0.0489, 0.0148, 0.0129, + 0.0200, 0.0032], device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 216.74, cls_loss 0.4554 cls_loss_mapping 0.0010 cls_loss_causal 0.4288 re_mapping 0.0043 re_causal 0.0122 /// teacc 98.96 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.0811, 0.0715, -0.1173, ..., -0.1253, -0.1252, 0.0229], + [-0.0882, -0.1792, 0.0088, ..., -0.0536, -0.0388, -0.1162], + [ 0.0308, -0.1007, 0.0568, ..., 0.1641, -0.1439, -0.0392], + ..., + [-0.1247, -0.2184, 0.0889, ..., -0.0156, -0.0550, 0.0758], + [ 0.0298, 0.0395, -0.0369, ..., -0.0917, -0.1051, -0.0245], + [-0.1630, -0.0010, -0.0017, ..., -0.1835, 0.1175, 0.0294]], + device='cuda:0'), grad: tensor([[-1.2904e-05, -4.5037e-04, 2.2566e-04, ..., 2.2042e-04, + 4.2081e-04, 9.6607e-04], + [ 3.2485e-05, 5.3160e-06, 6.1333e-05, ..., 6.7940e-03, + 5.5847e-03, 1.1711e-03], + [ 9.4235e-05, 1.4198e-04, 3.4714e-04, ..., 7.4530e-04, + 5.4693e-04, 1.8673e-03], + ..., + [ 2.7061e-05, 3.2258e-04, 4.0741e-03, ..., -5.3930e-04, + 1.9932e-03, 2.3823e-03], + [ 1.0071e-02, 5.7831e-03, 4.9305e-04, ..., 2.9302e-04, + 1.1902e-03, 2.1935e-03], + [ 9.4175e-05, 3.6812e-04, 3.1338e-03, ..., 1.0900e-03, + -4.3559e-04, 6.5470e-04]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0106, 0.0116, -0.0007, 0.0133, -0.0078, -0.0066, 0.0102, 0.0213, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0140, 0.0177, 0.0231, -0.0239, 0.0261, -0.0334, -0.0423, 0.0018, + 0.0180, -0.0012], device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 216.53, cls_loss 0.4668 cls_loss_mapping 0.0010 cls_loss_causal 0.4297 re_mapping 0.0043 re_causal 0.0123 /// teacc 98.96 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.0811, 0.0715, -0.1173, ..., -0.1254, -0.1252, 0.0230], + [-0.0882, -0.1792, 0.0089, ..., -0.0536, -0.0389, -0.1163], + [ 0.0307, -0.1007, 0.0568, ..., 0.1641, -0.1440, -0.0392], + ..., + [-0.1247, -0.2185, 0.0889, ..., -0.0157, -0.0551, 0.0758], + [ 0.0297, 0.0393, -0.0370, ..., -0.0917, -0.1053, -0.0246], + [-0.1631, -0.0011, -0.0017, ..., -0.1835, 0.1175, 0.0294]], + device='cuda:0'), grad: tensor([[ 4.9293e-05, 2.7016e-05, 8.3387e-05, ..., 1.2481e-04, + 4.1747e-04, 7.2002e-04], + [ 2.3339e-06, 4.8071e-05, 6.7472e-05, ..., 1.1599e-04, + 5.6791e-04, 9.3174e-04], + [ 8.8155e-05, 7.6711e-05, -5.9605e-04, ..., -1.1320e-03, + 4.1938e-04, 1.9503e-04], + ..., + [ 9.5814e-06, 1.7449e-05, -1.2932e-03, ..., 5.0402e-04, + 2.7704e-04, -5.7173e-04], + [ 1.0866e-04, 1.3053e-05, 9.3758e-05, ..., 1.1700e-04, + 4.4918e-04, 7.4577e-04], + [-7.3051e-03, -8.9035e-03, -5.7554e-04, ..., 3.1382e-05, + -1.1261e-02, -4.6234e-03]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0105, 0.0115, -0.0007, 0.0134, -0.0078, -0.0065, 0.0102, 0.0214, + -0.0339, 0.0501], device='cuda:0'), grad: tensor([ 0.0134, 0.0159, 0.0062, 0.0359, 0.0149, -0.0189, -0.0193, 0.0139, + 0.0123, -0.0743], device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 216.32, cls_loss 0.4466 cls_loss_mapping 0.0011 cls_loss_causal 0.4201 re_mapping 0.0045 re_causal 0.0123 /// teacc 98.99 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.0811, 0.0715, -0.1172, ..., -0.1255, -0.1253, 0.0229], + [-0.0882, -0.1793, 0.0091, ..., -0.0536, -0.0389, -0.1162], + [ 0.0306, -0.1008, 0.0569, ..., 0.1640, -0.1438, -0.0392], + ..., + [-0.1249, -0.2187, 0.0887, ..., -0.0159, -0.0553, 0.0758], + [ 0.0298, 0.0393, -0.0369, ..., -0.0916, -0.1052, -0.0245], + [-0.1632, -0.0012, -0.0017, ..., -0.1835, 0.1176, 0.0294]], + device='cuda:0'), grad: tensor([[ 1.3923e-07, 2.3127e-05, 9.5844e-04, ..., 4.4823e-04, + 7.2908e-04, 7.2384e-04], + [ 1.4997e-04, -7.4196e-04, -9.7466e-04, ..., 7.0035e-05, + -1.9112e-03, -5.2977e-04], + [-6.4430e-03, 1.3888e-04, 1.0338e-03, ..., -3.8986e-03, + 1.0052e-03, -1.1272e-03], + ..., + [ 3.7918e-03, -3.5238e-04, 3.5954e-04, ..., -1.3161e-03, + 1.5049e-03, 3.0861e-03], + [ 3.9368e-03, 1.3161e-04, 1.0643e-03, ..., 3.1967e-03, + 1.1826e-03, 2.2945e-03], + [ 3.7879e-05, 2.5034e-04, 4.1437e-04, ..., 9.3889e-04, + -2.4185e-03, 8.2779e-03]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0104, 0.0116, -0.0007, 0.0137, -0.0079, -0.0066, 0.0102, 0.0212, + -0.0338, 0.0501], device='cuda:0'), grad: tensor([ 0.0217, -0.0464, -0.0003, 0.0169, -0.0049, -0.0078, -0.0274, -0.0180, + 0.0388, 0.0274], device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 216.94, cls_loss 0.4573 cls_loss_mapping 0.0009 cls_loss_causal 0.4329 re_mapping 0.0041 re_causal 0.0120 /// teacc 99.01 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.0809, 0.0715, -0.1172, ..., -0.1256, -0.1252, 0.0230], + [-0.0883, -0.1793, 0.0092, ..., -0.0534, -0.0390, -0.1162], + [ 0.0305, -0.1007, 0.0568, ..., 0.1639, -0.1439, -0.0392], + ..., + [-0.1247, -0.2187, 0.0888, ..., -0.0159, -0.0551, 0.0759], + [ 0.0299, 0.0393, -0.0369, ..., -0.0916, -0.1053, -0.0246], + [-0.1631, -0.0011, -0.0017, ..., -0.1834, 0.1176, 0.0293]], + device='cuda:0'), grad: tensor([[ 2.4033e-04, 3.9041e-05, 7.4923e-05, ..., 6.2943e-04, + 1.0365e-04, 5.6839e-04], + [ 6.2108e-05, -5.2363e-05, 2.1979e-05, ..., 1.6391e-04, + 8.9943e-05, 8.2970e-04], + [ 1.7185e-03, 4.4435e-05, 8.3566e-05, ..., 6.7759e-04, + 9.3102e-05, 1.0710e-03], + ..., + [ 2.0528e-04, 5.2869e-05, 1.0300e-04, ..., 1.9426e-03, + -2.2089e-04, 1.4381e-03], + [ 3.6869e-03, 3.6011e-03, 4.3690e-05, ..., 2.6822e-04, + 2.7514e-04, 1.0080e-03], + [ 2.0618e-03, -1.4296e-03, 1.1081e-04, ..., -1.5125e-03, + -3.8986e-03, -1.7960e-02]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0104, 0.0116, -0.0008, 0.0137, -0.0078, -0.0067, 0.0103, 0.0212, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0188, -0.0128, 0.0224, 0.0010, -0.0032, -0.0050, -0.0120, 0.0003, + 0.0255, -0.0352], device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 216.87, cls_loss 0.4511 cls_loss_mapping 0.0012 cls_loss_causal 0.4261 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.01 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.0808, 0.0716, -0.1173, ..., -0.1257, -0.1252, 0.0230], + [-0.0884, -0.1791, 0.0091, ..., -0.0535, -0.0390, -0.1163], + [ 0.0306, -0.1006, 0.0568, ..., 0.1640, -0.1439, -0.0392], + ..., + [-0.1246, -0.2188, 0.0887, ..., -0.0159, -0.0552, 0.0759], + [ 0.0298, 0.0392, -0.0368, ..., -0.0917, -0.1052, -0.0246], + [-0.1630, -0.0009, -0.0017, ..., -0.1835, 0.1176, 0.0294]], + device='cuda:0'), grad: tensor([[ 5.5097e-06, 5.0962e-05, 5.5850e-05, ..., 2.4930e-05, + 2.2471e-04, 9.4697e-06], + [-1.3001e-05, -5.1069e-04, 2.7447e-03, ..., -2.3818e-04, + -2.3212e-03, -8.5449e-04], + [ 6.3419e-05, 5.7489e-05, -3.9291e-03, ..., 1.0085e-04, + 2.5129e-04, 2.7657e-04], + ..., + [ 3.7968e-05, 2.9635e-04, 5.7030e-03, ..., -7.7772e-04, + 7.8735e-03, -4.7803e-04], + [ 2.1410e-04, 3.2377e-04, 1.2074e-03, ..., 1.2016e-04, + 8.9312e-04, 3.7932e-04], + [-4.8923e-04, -5.4789e-04, -7.1831e-03, ..., -2.3410e-05, + -8.8654e-03, -6.8045e-04]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0103, 0.0115, -0.0009, 0.0137, -0.0078, -0.0066, 0.0103, 0.0212, + -0.0337, 0.0502], device='cuda:0'), grad: tensor([ 1.4801e-02, -4.8248e-02, -1.1330e-02, -2.8223e-05, 1.8066e-02, + 1.6037e-02, -4.3610e-02, 3.0060e-02, 2.5238e-02, -9.8228e-04], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 217.56, cls_loss 0.4440 cls_loss_mapping 0.0010 cls_loss_causal 0.4163 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.01 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.0808, 0.0715, -0.1173, ..., -0.1257, -0.1253, 0.0230], + [-0.0884, -0.1792, 0.0090, ..., -0.0536, -0.0390, -0.1163], + [ 0.0306, -0.1006, 0.0568, ..., 0.1640, -0.1438, -0.0392], + ..., + [-0.1246, -0.2187, 0.0888, ..., -0.0159, -0.0551, 0.0759], + [ 0.0298, 0.0393, -0.0368, ..., -0.0917, -0.1052, -0.0247], + [-0.1631, -0.0011, -0.0017, ..., -0.1836, 0.1176, 0.0294]], + device='cuda:0'), grad: tensor([[ 1.5467e-05, -2.9488e-03, -3.6755e-03, ..., -2.1267e-03, + -2.8267e-03, -2.9564e-03], + [ 2.9206e-05, 9.5701e-04, 4.2462e-04, ..., 4.9019e-04, + 1.3027e-03, 1.3695e-03], + [ 1.0067e-04, 4.9257e-04, 6.4945e-04, ..., 5.9843e-04, + 2.1191e-03, 2.2659e-03], + ..., + [ 1.8865e-05, 2.2507e-04, 1.5745e-03, ..., 5.3978e-04, + 1.9255e-03, 2.2163e-03], + [ 1.1402e-04, 6.0463e-04, -6.9189e-04, ..., -2.1057e-03, + -6.5765e-03, -4.2839e-03], + [ 2.7418e-04, 2.7065e-03, -9.5558e-04, ..., 7.6890e-05, + -2.5387e-03, -4.7417e-03]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0103, 0.0116, -0.0009, 0.0136, -0.0078, -0.0065, 0.0102, 0.0213, + -0.0336, 0.0500], device='cuda:0'), grad: tensor([-0.0211, 0.0185, 0.0187, -0.0211, 0.0394, 0.0149, -0.0154, 0.0203, + -0.0477, -0.0064], device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 217.39, cls_loss 0.4444 cls_loss_mapping 0.0008 cls_loss_causal 0.4175 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.00 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.0808, 0.0715, -0.1173, ..., -0.1256, -0.1253, 0.0231], + [-0.0885, -0.1791, 0.0091, ..., -0.0537, -0.0390, -0.1163], + [ 0.0307, -0.1007, 0.0570, ..., 0.1642, -0.1438, -0.0393], + ..., + [-0.1246, -0.2188, 0.0888, ..., -0.0159, -0.0551, 0.0760], + [ 0.0298, 0.0393, -0.0370, ..., -0.0918, -0.1054, -0.0247], + [-0.1632, -0.0011, -0.0017, ..., -0.1836, 0.1176, 0.0293]], + device='cuda:0'), grad: tensor([[-5.6839e-03, -4.8180e-03, 2.4110e-05, ..., -2.9564e-05, + 3.5501e-04, -8.5449e-04], + [ 6.1655e-04, 4.4894e-04, 2.3317e-04, ..., 2.5225e-04, + 1.6241e-03, 3.1986e-03], + [ 1.5430e-03, 1.2407e-03, -2.0409e-04, ..., 2.4867e-04, + -4.0269e-04, -3.3073e-03], + ..., + [ 1.9646e-04, 1.0294e-04, 1.9789e-04, ..., 2.2495e-04, + -1.5631e-03, -2.1019e-03], + [ 1.1301e-03, 5.3644e-04, -1.5271e-04, ..., 1.1940e-03, + -3.0351e-04, -1.9836e-03], + [ 8.0585e-04, 2.1327e-04, -2.5988e-04, ..., 1.4601e-03, + -7.9095e-05, -2.1477e-03]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0103, 0.0117, -0.0010, 0.0137, -0.0077, -0.0066, 0.0102, 0.0212, + -0.0337, 0.0500], device='cuda:0'), grad: tensor([-0.0067, 0.0276, -0.0125, 0.0191, 0.0079, 0.0143, 0.0025, -0.0440, + -0.0160, 0.0079], device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 216.99, cls_loss 0.4411 cls_loss_mapping 0.0010 cls_loss_causal 0.4128 re_mapping 0.0042 re_causal 0.0116 /// teacc 99.00 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.0808, 0.0715, -0.1174, ..., -0.1257, -0.1255, 0.0231], + [-0.0886, -0.1791, 0.0091, ..., -0.0537, -0.0389, -0.1163], + [ 0.0306, -0.1007, 0.0570, ..., 0.1642, -0.1438, -0.0393], + ..., + [-0.1249, -0.2189, 0.0887, ..., -0.0158, -0.0551, 0.0762], + [ 0.0299, 0.0394, -0.0369, ..., -0.0919, -0.1055, -0.0246], + [-0.1631, -0.0010, -0.0016, ..., -0.1835, 0.1176, 0.0292]], + device='cuda:0'), grad: tensor([[ 1.0896e-04, 3.9744e-04, 3.4571e-04, ..., 1.0529e-03, + 1.7757e-03, 3.2272e-03], + [ 7.2904e-06, 1.0651e-04, 2.3353e-04, ..., 9.0313e-04, + 1.6699e-03, 2.7218e-03], + [ 6.2823e-05, 7.4387e-05, 1.7667e-04, ..., 8.4829e-04, + 1.1721e-03, 2.2106e-03], + ..., + [ 7.9051e-06, -3.7646e-04, -5.7716e-03, ..., 3.1543e-04, + -9.4528e-03, -8.6365e-03], + [ 1.4043e-04, 3.4666e-04, -7.9060e-04, ..., -2.5539e-03, + -3.0746e-03, -7.9422e-03], + [-3.3647e-05, -2.6970e-03, 2.7027e-03, ..., -1.8015e-03, + 3.8357e-03, -4.7264e-03]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0103, 0.0119, -0.0011, 0.0137, -0.0078, -0.0066, 0.0102, 0.0213, + -0.0336, 0.0500], device='cuda:0'), grad: tensor([ 0.0172, 0.0200, 0.0177, -0.0012, 0.0244, 0.0166, -0.0075, -0.0707, + -0.0378, 0.0213], device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 217.02, cls_loss 0.4422 cls_loss_mapping 0.0008 cls_loss_causal 0.4223 re_mapping 0.0040 re_causal 0.0119 /// teacc 98.99 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.0806, 0.0717, -0.1174, ..., -0.1257, -0.1255, 0.0231], + [-0.0885, -0.1791, 0.0092, ..., -0.0536, -0.0389, -0.1164], + [ 0.0306, -0.1007, 0.0568, ..., 0.1639, -0.1438, -0.0393], + ..., + [-0.1249, -0.2189, 0.0888, ..., -0.0157, -0.0551, 0.0763], + [ 0.0297, 0.0393, -0.0369, ..., -0.0919, -0.1054, -0.0245], + [-0.1632, -0.0010, -0.0017, ..., -0.1834, 0.1178, 0.0293]], + device='cuda:0'), grad: tensor([[ 1.7138e-03, 1.6890e-03, 4.9829e-05, ..., 9.8038e-04, + 4.7803e-04, 4.5061e-04], + [ 1.1063e-04, 1.7047e-04, 4.0710e-05, ..., 2.6047e-05, + 4.5776e-04, 4.5228e-04], + [ 1.7061e-03, 4.1437e-04, 3.9291e-03, ..., 2.0313e-03, + 3.0255e-04, -9.6798e-04], + ..., + [-1.9188e-03, 1.6344e-04, -4.8904e-03, ..., -2.6531e-03, + 1.1759e-03, 9.1410e-04], + [ 3.8471e-03, 3.5915e-03, 3.1322e-05, ..., 4.5824e-04, + 3.6502e-04, 4.0889e-04], + [ 1.9872e-04, -7.8630e-04, -7.1259e-03, ..., 2.4843e-04, + -7.3242e-03, -5.0392e-03]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0105, 0.0119, -0.0011, 0.0137, -0.0078, -0.0067, 0.0100, 0.0213, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0168, 0.0116, -0.0474, 0.0002, -0.0220, 0.0008, 0.0156, 0.0035, + 0.0201, 0.0009], device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 216.72, cls_loss 0.4215 cls_loss_mapping 0.0008 cls_loss_causal 0.4031 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.03 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.0806, 0.0718, -0.1175, ..., -0.1258, -0.1255, 0.0230], + [-0.0886, -0.1790, 0.0094, ..., -0.0537, -0.0388, -0.1164], + [ 0.0307, -0.1007, 0.0568, ..., 0.1641, -0.1438, -0.0392], + ..., + [-0.1249, -0.2191, 0.0888, ..., -0.0157, -0.0551, 0.0764], + [ 0.0298, 0.0393, -0.0369, ..., -0.0919, -0.1055, -0.0245], + [-0.1633, -0.0011, -0.0017, ..., -0.1834, 0.1177, 0.0293]], + device='cuda:0'), grad: tensor([[ 2.2072e-06, 6.8009e-05, 1.1653e-04, ..., 6.5207e-05, + 4.4560e-04, 3.4070e-04], + [ 2.2091e-06, 7.3481e-07, -1.0633e-03, ..., 1.7762e-05, + -8.2321e-03, -4.2610e-03], + [ 1.6797e-04, 4.6380e-06, 6.7329e-04, ..., 1.7691e-04, + 2.3460e-04, 3.0351e-04], + ..., + [-1.8585e-04, 3.0287e-06, -2.6894e-04, ..., -1.6189e-04, + 3.1967e-03, 1.6661e-03], + [-3.6843e-06, 1.4707e-05, 5.9068e-05, ..., -1.1128e-04, + 3.5691e-04, -4.7088e-04], + [ 7.3016e-07, -4.6921e-03, -3.8166e-03, ..., -3.0384e-03, + -5.2834e-03, -4.9515e-03]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0105, 0.0118, -0.0010, 0.0138, -0.0078, -0.0067, 0.0100, 0.0213, + -0.0337, 0.0500], device='cuda:0'), grad: tensor([ 0.0073, -0.0198, 0.0089, 0.0077, 0.0226, 0.0064, 0.0090, 0.0147, + -0.0242, -0.0325], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 433---------------------------------------------------- +epoch 433, time 217.76, cls_loss 0.4415 cls_loss_mapping 0.0008 cls_loss_causal 0.4151 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.05 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.0808, 0.0716, -0.1175, ..., -0.1257, -0.1256, 0.0231], + [-0.0886, -0.1790, 0.0093, ..., -0.0538, -0.0388, -0.1166], + [ 0.0307, -0.1005, 0.0568, ..., 0.1641, -0.1437, -0.0392], + ..., + [-0.1248, -0.2190, 0.0888, ..., -0.0156, -0.0553, 0.0764], + [ 0.0298, 0.0393, -0.0369, ..., -0.0919, -0.1056, -0.0246], + [-0.1632, -0.0010, -0.0017, ..., -0.1835, 0.1177, 0.0293]], + device='cuda:0'), grad: tensor([[ 2.8872e-04, 2.5249e-04, 3.3975e-04, ..., 2.3818e-04, + 6.1178e-04, 5.3120e-04], + [ 1.7837e-05, 1.9953e-05, -3.5019e-03, ..., -6.5708e-04, + -1.6203e-03, -2.7981e-03], + [ 2.6727e-04, 2.0266e-04, 3.5977e-04, ..., 8.8394e-05, + 6.8331e-04, 5.0974e-04], + ..., + [ 1.4448e-04, 1.3530e-04, 5.7602e-04, ..., 1.4353e-04, + -3.3512e-03, -1.1930e-03], + [ 1.4567e-04, -1.7004e-03, 4.2033e-04, ..., 9.2924e-05, + 1.2827e-04, 5.3072e-04], + [ 1.9419e-04, 2.0275e-03, 3.6907e-03, ..., 2.9411e-03, + 8.1635e-03, 3.7479e-03]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0104, 0.0119, -0.0010, 0.0137, -0.0076, -0.0068, 0.0101, 0.0213, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([ 0.0169, -0.0058, 0.0151, 0.0094, -0.0070, 0.0160, -0.0156, -0.0768, + 0.0118, 0.0362], device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 216.43, cls_loss 0.4307 cls_loss_mapping 0.0007 cls_loss_causal 0.4014 re_mapping 0.0043 re_causal 0.0125 /// teacc 98.99 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.0808, 0.0715, -0.1175, ..., -0.1255, -0.1255, 0.0231], + [-0.0886, -0.1791, 0.0093, ..., -0.0538, -0.0389, -0.1165], + [ 0.0307, -0.1006, 0.0570, ..., 0.1640, -0.1438, -0.0393], + ..., + [-0.1250, -0.2192, 0.0886, ..., -0.0158, -0.0554, 0.0762], + [ 0.0300, 0.0396, -0.0367, ..., -0.0917, -0.1055, -0.0246], + [-0.1633, -0.0010, -0.0016, ..., -0.1835, 0.1177, 0.0295]], + device='cuda:0'), grad: tensor([[-3.2883e-03, -4.1351e-03, 6.3229e-04, ..., -9.5797e-04, + -4.2629e-04, -2.3708e-03], + [ 2.3162e-04, -1.6665e-04, 5.1785e-04, ..., 6.4278e-04, + 7.0393e-05, 1.0979e-04], + [ 7.0453e-05, -8.4758e-05, 6.9952e-04, ..., -3.1781e-04, + 9.5189e-05, 5.4359e-04], + ..., + [ 3.7241e-04, 2.5317e-05, 5.0575e-05, ..., 4.4632e-04, + 9.3520e-05, 9.4116e-05], + [ 3.1586e-03, 2.1286e-03, 8.8549e-04, ..., 3.0212e-03, + -1.0657e-04, 6.6233e-04], + [ 3.4389e-03, 2.2042e-04, 7.9060e-04, ..., 8.4400e-04, + 8.6725e-05, 3.8509e-03]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0104, 0.0119, -0.0011, 0.0136, -0.0078, -0.0067, 0.0102, 0.0212, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0006, 0.0133, 0.0181, -0.0212, -0.0432, -0.0310, 0.0222, 0.0166, + 0.0225, 0.0021], device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 217.68, cls_loss 0.4292 cls_loss_mapping 0.0008 cls_loss_causal 0.4010 re_mapping 0.0042 re_causal 0.0122 /// teacc 98.97 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.0810, 0.0715, -0.1176, ..., -0.1256, -0.1256, 0.0231], + [-0.0885, -0.1790, 0.0093, ..., -0.0539, -0.0388, -0.1165], + [ 0.0308, -0.1006, 0.0570, ..., 0.1641, -0.1437, -0.0392], + ..., + [-0.1250, -0.2192, 0.0886, ..., -0.0157, -0.0554, 0.0762], + [ 0.0300, 0.0396, -0.0366, ..., -0.0915, -0.1054, -0.0246], + [-0.1633, -0.0011, -0.0017, ..., -0.1835, 0.1177, 0.0295]], + device='cuda:0'), grad: tensor([[ 1.6856e-04, 6.8378e-04, 5.0259e-04, ..., 4.0197e-04, + 1.5802e-03, 2.1000e-03], + [ 3.0294e-05, 1.0931e-04, 6.9618e-05, ..., 7.5877e-05, + 4.5562e-04, 4.0054e-04], + [ 3.7336e-04, 2.5153e-04, 7.6199e-04, ..., 8.0109e-04, + -1.4505e-03, -2.2864e-04], + ..., + [ 4.6802e-04, 6.6900e-04, 2.2774e-03, ..., 1.1024e-03, + 2.4548e-03, 2.6360e-03], + [ 4.7946e-04, 1.0025e-02, -3.4676e-03, ..., -5.1880e-03, + -4.3297e-03, -4.9057e-03], + [-2.5997e-03, -1.2512e-02, 1.2634e-02, ..., 2.3346e-03, + 1.1261e-02, 1.2222e-02]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0104, 0.0120, -0.0010, 0.0137, -0.0077, -0.0068, 0.0101, 0.0212, + -0.0336, 0.0499], device='cuda:0'), grad: tensor([ 0.0142, 0.0084, -0.0209, 0.0157, -0.0082, 0.0105, -0.0234, 0.0167, + -0.0226, 0.0096], device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 216.73, cls_loss 0.4258 cls_loss_mapping 0.0008 cls_loss_causal 0.4048 re_mapping 0.0041 re_causal 0.0117 /// teacc 99.00 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.0810, 0.0716, -0.1175, ..., -0.1256, -0.1256, 0.0231], + [-0.0885, -0.1790, 0.0095, ..., -0.0537, -0.0388, -0.1163], + [ 0.0306, -0.1007, 0.0570, ..., 0.1640, -0.1435, -0.0392], + ..., + [-0.1252, -0.2192, 0.0886, ..., -0.0158, -0.0555, 0.0761], + [ 0.0299, 0.0394, -0.0366, ..., -0.0916, -0.1055, -0.0247], + [-0.1631, -0.0010, -0.0017, ..., -0.1835, 0.1177, 0.0295]], + device='cuda:0'), grad: tensor([[ 2.0945e-04, 3.2224e-06, 7.4625e-05, ..., 3.0026e-05, + 2.7132e-04, 5.7220e-04], + [ 3.3689e-04, 5.0068e-06, 4.9144e-05, ..., -1.4222e-04, + 2.6751e-04, 1.0729e-03], + [ 4.5776e-05, -1.0505e-05, 3.5334e-04, ..., -2.4188e-04, + 4.9257e-04, 1.1902e-03], + ..., + [ 1.7929e-04, 1.2971e-05, -1.2579e-03, ..., 1.3418e-05, + 3.1757e-04, 4.9973e-04], + [ 2.8682e-04, 4.8816e-05, 2.2185e-04, ..., 8.9705e-05, + 4.0984e-04, 5.9462e-04], + [-1.8969e-03, 1.1122e-04, 4.6015e-04, ..., 3.7879e-05, + -1.3561e-03, -5.5809e-03]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0104, 0.0121, -0.0011, 0.0135, -0.0077, -0.0067, 0.0102, 0.0211, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0166, -0.0066, -0.0107, 0.0220, 0.0094, -0.0142, -0.0078, 0.0151, + 0.0179, -0.0416], device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 216.72, cls_loss 0.4529 cls_loss_mapping 0.0008 cls_loss_causal 0.4263 re_mapping 0.0041 re_causal 0.0122 /// teacc 98.99 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.0809, 0.0716, -0.1174, ..., -0.1257, -0.1256, 0.0232], + [-0.0884, -0.1790, 0.0096, ..., -0.0534, -0.0389, -0.1163], + [ 0.0308, -0.1006, 0.0570, ..., 0.1639, -0.1436, -0.0393], + ..., + [-0.1252, -0.2192, 0.0884, ..., -0.0158, -0.0554, 0.0760], + [ 0.0299, 0.0394, -0.0367, ..., -0.0917, -0.1055, -0.0247], + [-0.1633, -0.0010, -0.0017, ..., -0.1835, 0.1176, 0.0296]], + device='cuda:0'), grad: tensor([[ 1.0431e-05, 4.6492e-05, 2.9933e-06, ..., -3.1166e-03, + 9.1434e-05, -2.4738e-03], + [ 7.9423e-06, -9.3937e-04, 2.0251e-05, ..., 2.6011e-04, + 1.5691e-05, 5.6505e-05], + [ 3.0875e-05, 8.2195e-05, -2.6654e-06, ..., 5.4121e-04, + 1.0431e-04, 8.5211e-04], + ..., + [ 9.4697e-06, 3.8356e-05, -1.1489e-05, ..., 1.6718e-03, + 9.2268e-05, 1.3094e-03], + [ 2.2912e-04, 2.2354e-03, 7.0989e-05, ..., 1.8179e-04, + 2.6932e-03, 1.6136e-03], + [-4.2653e-04, -2.3174e-03, 6.0648e-05, ..., 1.3959e-04, + -4.6158e-03, -2.5082e-03]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0104, 0.0120, -0.0010, 0.0135, -0.0078, -0.0067, 0.0103, 0.0212, + -0.0336, 0.0500], device='cuda:0'), grad: tensor([-0.0193, -0.0265, 0.0183, -0.0191, 0.0108, 0.0087, 0.0110, -0.0063, + 0.0176, 0.0048], device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 216.98, cls_loss 0.4523 cls_loss_mapping 0.0009 cls_loss_causal 0.4243 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.00 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.0809, 0.0716, -0.1174, ..., -0.1257, -0.1256, 0.0230], + [-0.0885, -0.1790, 0.0096, ..., -0.0535, -0.0389, -0.1164], + [ 0.0308, -0.1005, 0.0571, ..., 0.1640, -0.1436, -0.0393], + ..., + [-0.1251, -0.2192, 0.0885, ..., -0.0158, -0.0553, 0.0760], + [ 0.0300, 0.0394, -0.0367, ..., -0.0918, -0.1055, -0.0247], + [-0.1633, -0.0011, -0.0019, ..., -0.1835, 0.1175, 0.0295]], + device='cuda:0'), grad: tensor([[ 1.0681e-04, 1.4699e-04, 3.2806e-04, ..., 3.3283e-04, + 4.4012e-04, 7.9393e-04], + [ 2.3007e-04, 4.2343e-03, -5.3406e-04, ..., -1.9045e-03, + -1.0414e-03, -3.7613e-03], + [-1.0166e-03, -3.1441e-05, -2.3994e-03, ..., -9.1568e-06, + 9.3985e-04, 5.9009e-05], + ..., + [ 2.9564e-04, 1.1641e-04, 5.1689e-04, ..., 5.3596e-04, + 5.8031e-04, 1.1616e-03], + [-6.9904e-04, 8.0013e-04, 2.0754e-04, ..., -1.2617e-03, + 8.5783e-04, 8.1825e-04], + [ 4.8518e-05, 2.5773e-04, 5.4657e-05, ..., 1.6665e-04, + 5.3358e-04, 1.1120e-03]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0103, 0.0120, -0.0009, 0.0134, -0.0076, -0.0069, 0.0103, 0.0213, + -0.0337, 0.0498], device='cuda:0'), grad: tensor([ 8.2550e-03, 4.2468e-05, 4.3259e-03, 1.1940e-02, -4.0802e-02, + 7.3280e-03, 1.1307e-02, 1.1261e-02, 6.5536e-03, -2.0203e-02], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 217.04, cls_loss 0.4210 cls_loss_mapping 0.0009 cls_loss_causal 0.3923 re_mapping 0.0042 re_causal 0.0115 /// teacc 98.98 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.0809, 0.0716, -0.1175, ..., -0.1258, -0.1257, 0.0229], + [-0.0886, -0.1792, 0.0096, ..., -0.0535, -0.0388, -0.1165], + [ 0.0307, -0.1005, 0.0570, ..., 0.1638, -0.1437, -0.0394], + ..., + [-0.1251, -0.2194, 0.0886, ..., -0.0157, -0.0553, 0.0761], + [ 0.0300, 0.0395, -0.0367, ..., -0.0919, -0.1055, -0.0248], + [-0.1634, -0.0011, -0.0018, ..., -0.1834, 0.1176, 0.0295]], + device='cuda:0'), grad: tensor([[ 1.0423e-05, 4.1090e-06, 6.9678e-05, ..., 3.9244e-04, + 1.6820e-04, 9.4295e-05], + [ 1.4508e-04, 5.4210e-05, 2.5272e-04, ..., -2.4166e-03, + 1.0324e-04, 6.7711e-05], + [ 1.3900e-04, 1.6332e-04, 4.9973e-04, ..., 7.0477e-04, + 2.5773e-04, 8.0407e-05], + ..., + [ 2.4247e-04, 1.0961e-04, 5.7650e-04, ..., 5.4312e-04, + 1.0500e-03, 4.7445e-04], + [ 1.9083e-03, 9.3699e-04, 3.4542e-03, ..., 2.6531e-03, + 1.2693e-03, 2.1684e-04], + [-4.9171e-03, -2.4567e-03, -7.8964e-03, ..., -5.8937e-03, + -2.7966e-04, 1.1196e-03]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0104, 0.0119, -0.0008, 0.0135, -0.0074, -0.0069, 0.0103, 0.0213, + -0.0337, 0.0497], device='cuda:0'), grad: tensor([ 0.0133, -0.0133, -0.0449, -0.0053, -0.0010, 0.0101, 0.0149, 0.0162, + 0.0235, -0.0135], device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 217.02, cls_loss 0.4721 cls_loss_mapping 0.0009 cls_loss_causal 0.4576 re_mapping 0.0041 re_causal 0.0121 /// teacc 98.96 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.0809, 0.0716, -0.1177, ..., -0.1257, -0.1257, 0.0229], + [-0.0885, -0.1792, 0.0099, ..., -0.0535, -0.0387, -0.1164], + [ 0.0306, -0.1006, 0.0568, ..., 0.1638, -0.1439, -0.0396], + ..., + [-0.1251, -0.2193, 0.0887, ..., -0.0157, -0.0553, 0.0762], + [ 0.0300, 0.0396, -0.0367, ..., -0.0918, -0.1057, -0.0249], + [-0.1630, -0.0011, -0.0017, ..., -0.1832, 0.1177, 0.0295]], + device='cuda:0'), grad: tensor([[-3.1948e-03, -4.5624e-03, -1.5974e-03, ..., 1.1653e-05, + -5.1737e-04, -6.5994e-04], + [ 1.4491e-06, 2.6092e-05, -5.5647e-04, ..., -4.5300e-04, + 3.7104e-05, 1.1320e-03], + [ 2.0771e-03, 2.7580e-03, 2.9635e-04, ..., 1.1045e-04, + 5.9128e-05, 1.5249e-03], + ..., + [ 9.2015e-07, 6.0558e-05, 4.0388e-04, ..., 1.8299e-05, + 1.2827e-04, 3.2864e-03], + [ 1.2791e-04, 2.7657e-04, 2.3866e-04, ..., 4.4525e-05, + 5.8800e-05, 5.6124e-04], + [ 9.8610e-04, 1.3227e-03, 2.4700e-04, ..., 2.2680e-05, + -6.3837e-05, 4.9400e-04]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0104, 0.0120, -0.0010, 0.0134, -0.0073, -0.0068, 0.0101, 0.0213, + -0.0337, 0.0499], device='cuda:0'), grad: tensor([-0.0504, -0.0101, 0.0237, -0.0148, 0.0167, -0.0189, 0.0152, 0.0019, + 0.0167, 0.0199], device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 217.07, cls_loss 0.4526 cls_loss_mapping 0.0009 cls_loss_causal 0.4162 re_mapping 0.0041 re_causal 0.0120 /// teacc 98.95 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.0808, 0.0716, -0.1177, ..., -0.1257, -0.1257, 0.0230], + [-0.0886, -0.1792, 0.0100, ..., -0.0536, -0.0387, -0.1165], + [ 0.0306, -0.1005, 0.0568, ..., 0.1638, -0.1438, -0.0396], + ..., + [-0.1250, -0.2194, 0.0887, ..., -0.0156, -0.0552, 0.0762], + [ 0.0301, 0.0396, -0.0366, ..., -0.0918, -0.1057, -0.0248], + [-0.1630, -0.0010, -0.0017, ..., -0.1832, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[-1.7214e-03, -1.2770e-03, -3.4161e-03, ..., -3.5305e-03, + -5.2414e-03, -6.2408e-03], + [ 6.3241e-05, 3.6180e-05, 7.8678e-04, ..., 1.1244e-03, + 1.7190e-04, 3.5453e-04], + [ 1.3018e-04, 1.0371e-04, -7.4501e-03, ..., -8.0414e-03, + 3.5548e-04, 8.0156e-04], + ..., + [-1.6940e-04, 2.1780e-04, 3.2616e-03, ..., -6.1321e-04, + 9.9468e-04, -2.6360e-03], + [ 6.2132e-04, 2.6584e-04, 3.0270e-03, ..., 3.7746e-03, + 1.1692e-03, 1.7767e-03], + [ 3.1590e-04, 7.8380e-05, 8.4209e-04, ..., 1.5612e-03, + 3.6001e-04, 1.1768e-03]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0105, 0.0120, -0.0010, 0.0133, -0.0074, -0.0066, 0.0100, 0.0212, + -0.0338, 0.0499], device='cuda:0'), grad: tensor([-0.0237, 0.0093, -0.0020, -0.0131, 0.0179, 0.0059, 0.0080, -0.0358, + 0.0173, 0.0162], device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 216.78, cls_loss 0.4471 cls_loss_mapping 0.0009 cls_loss_causal 0.4204 re_mapping 0.0040 re_causal 0.0116 /// teacc 98.99 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.0808, 0.0716, -0.1178, ..., -0.1257, -0.1256, 0.0231], + [-0.0887, -0.1791, 0.0099, ..., -0.0536, -0.0387, -0.1165], + [ 0.0305, -0.1006, 0.0569, ..., 0.1638, -0.1439, -0.0396], + ..., + [-0.1250, -0.2197, 0.0886, ..., -0.0156, -0.0553, 0.0762], + [ 0.0301, 0.0398, -0.0367, ..., -0.0918, -0.1057, -0.0247], + [-0.1631, -0.0011, -0.0016, ..., -0.1833, 0.1178, 0.0297]], + device='cuda:0'), grad: tensor([[ 3.1042e-04, -1.4223e-05, 8.6069e-05, ..., 1.0663e-04, + 2.6917e-04, 1.5509e-04], + [ 1.3423e-04, 2.0400e-05, 1.5450e-04, ..., 1.7834e-04, + 3.9506e-04, 1.8275e-04], + [-9.5725e-05, 9.8884e-05, -8.9340e-03, ..., -1.0073e-05, + -4.2152e-03, 1.9526e-04], + ..., + [ 2.4338e-03, 3.6776e-05, 4.6844e-03, ..., 2.5129e-04, + -3.9940e-03, -2.0752e-03], + [ 2.6474e-03, 6.4313e-05, 2.9659e-04, ..., 2.6917e-04, + 4.8161e-04, 1.8859e-04], + [ 6.3467e-04, 1.2302e-04, 4.0207e-03, ..., 4.0150e-04, + 9.0866e-03, 2.5940e-03]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0106, 0.0120, -0.0010, 0.0134, -0.0074, -0.0066, 0.0099, 0.0213, + -0.0338, 0.0499], device='cuda:0'), grad: tensor([ 0.0143, -0.0179, -0.0190, 0.0095, 0.0056, -0.0223, -0.0217, 0.0176, + -0.0077, 0.0416], device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 216.62, cls_loss 0.4452 cls_loss_mapping 0.0008 cls_loss_causal 0.4187 re_mapping 0.0041 re_causal 0.0122 /// teacc 98.97 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.0808, 0.0715, -0.1179, ..., -0.1258, -0.1257, 0.0230], + [-0.0889, -0.1792, 0.0099, ..., -0.0536, -0.0387, -0.1165], + [ 0.0304, -0.1005, 0.0569, ..., 0.1638, -0.1438, -0.0397], + ..., + [-0.1249, -0.2195, 0.0886, ..., -0.0156, -0.0552, 0.0763], + [ 0.0301, 0.0397, -0.0368, ..., -0.0918, -0.1057, -0.0246], + [-0.1630, -0.0011, -0.0016, ..., -0.1833, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[ 7.0810e-05, -1.9360e-04, -4.2152e-04, ..., 7.9535e-07, + 2.2507e-04, 2.0385e-04], + [ 2.7218e-03, 1.5974e-04, 1.8045e-05, ..., 8.8057e-07, + -1.0185e-03, -3.4070e-04], + [ 7.3051e-04, 4.6396e-04, 4.0770e-05, ..., -1.1760e-04, + 1.4603e-04, 1.9455e-04], + ..., + [ 7.1704e-05, 1.7822e-04, 4.6998e-05, ..., 2.1517e-05, + 3.6621e-04, 4.0340e-04], + [-3.4275e-03, 9.1410e-04, 2.9230e-04, ..., 1.7881e-05, + 9.2411e-04, 5.8603e-04], + [-3.5137e-05, -1.0128e-03, -2.5988e-04, ..., 2.4176e-04, + -1.1597e-03, -5.6791e-04]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0105, 0.0120, -0.0010, 0.0134, -0.0076, -0.0066, 0.0100, 0.0214, + -0.0339, 0.0499], device='cuda:0'), grad: tensor([ 0.0129, -0.0189, 0.0207, -0.0140, 0.0173, -0.0062, -0.0364, -0.0075, + 0.0171, 0.0150], device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 216.83, cls_loss 0.4529 cls_loss_mapping 0.0006 cls_loss_causal 0.4278 re_mapping 0.0041 re_causal 0.0126 /// teacc 99.01 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.0808, 0.0715, -0.1177, ..., -0.1258, -0.1256, 0.0230], + [-0.0890, -0.1792, 0.0098, ..., -0.0536, -0.0388, -0.1166], + [ 0.0304, -0.1006, 0.0570, ..., 0.1638, -0.1439, -0.0398], + ..., + [-0.1249, -0.2195, 0.0885, ..., -0.0155, -0.0553, 0.0762], + [ 0.0301, 0.0397, -0.0368, ..., -0.0919, -0.1057, -0.0245], + [-0.1630, -0.0009, -0.0015, ..., -0.1833, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 1.4253e-05, 4.6521e-05, 2.9564e-04, ..., -9.9754e-04, + -3.6240e-03, -3.2768e-03], + [ 2.5138e-05, 6.6161e-05, 1.7958e-03, ..., 2.8515e-04, + 1.6851e-03, 1.0462e-03], + [ 6.4671e-05, 5.4777e-05, 3.7813e-04, ..., 2.0099e-04, + 1.0242e-03, 6.5804e-04], + ..., + [ 2.9385e-05, 7.8261e-05, -2.8442e-02, ..., -9.9640e-03, + -4.7302e-03, -1.1345e-02], + [ 8.5831e-05, 1.3137e-04, 2.1470e-04, ..., 8.9884e-04, + 1.4973e-03, 9.4604e-04], + [ 9.0957e-05, 4.2260e-05, 2.4292e-02, ..., 9.1400e-03, + 3.6068e-03, 9.8419e-03]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0106, 0.0119, -0.0011, 0.0134, -0.0076, -0.0066, 0.0099, 0.0213, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([-0.0177, 0.0465, 0.0142, 0.0119, -0.0164, -0.0142, 0.0122, -0.0500, + -0.0415, 0.0551], device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 216.54, cls_loss 0.4305 cls_loss_mapping 0.0007 cls_loss_causal 0.4047 re_mapping 0.0043 re_causal 0.0123 /// teacc 99.01 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.0809, 0.0714, -0.1178, ..., -0.1259, -0.1257, 0.0229], + [-0.0891, -0.1791, 0.0098, ..., -0.0535, -0.0385, -0.1166], + [ 0.0304, -0.1005, 0.0570, ..., 0.1639, -0.1439, -0.0398], + ..., + [-0.1250, -0.2196, 0.0887, ..., -0.0154, -0.0552, 0.0763], + [ 0.0302, 0.0398, -0.0369, ..., -0.0919, -0.1058, -0.0245], + [-0.1631, -0.0010, -0.0016, ..., -0.1834, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[ 8.4209e-04, -7.7534e-04, 3.2592e-04, ..., 1.4734e-03, + 4.1771e-04, 1.0371e-04], + [ 1.7595e-04, 1.4377e-04, 6.2275e-04, ..., 1.0748e-03, + 6.2418e-04, 2.4244e-05], + [ 5.0583e-03, -4.6563e-04, -9.8419e-04, ..., -8.3971e-04, + -1.6384e-03, -2.2757e-04], + ..., + [ 7.7300e-08, 1.4544e-03, 1.5841e-03, ..., 7.2336e-04, + 6.3610e-04, 1.4772e-03], + [-8.3847e-03, -9.5215e-03, -5.2567e-03, ..., -7.4816e-04, + -3.9444e-03, -6.4659e-03], + [ 2.2907e-03, 4.6349e-03, 4.4556e-03, ..., 1.0796e-03, + 3.7422e-03, 4.8523e-03]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0104, 0.0119, -0.0009, 0.0134, -0.0076, -0.0067, 0.0099, 0.0214, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([-0.0128, 0.0247, 0.0043, 0.0028, -0.0424, -0.0167, 0.0112, 0.0140, + -0.0213, 0.0363], device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 216.56, cls_loss 0.4601 cls_loss_mapping 0.0007 cls_loss_causal 0.4296 re_mapping 0.0041 re_causal 0.0122 /// teacc 99.00 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.0809, 0.0715, -0.1179, ..., -0.1259, -0.1258, 0.0229], + [-0.0890, -0.1789, 0.0099, ..., -0.0535, -0.0384, -0.1167], + [ 0.0303, -0.1006, 0.0570, ..., 0.1639, -0.1439, -0.0397], + ..., + [-0.1248, -0.2198, 0.0887, ..., -0.0155, -0.0552, 0.0763], + [ 0.0302, 0.0398, -0.0367, ..., -0.0918, -0.1058, -0.0244], + [-0.1633, -0.0010, -0.0017, ..., -0.1834, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[ 3.8236e-05, -2.1696e-04, 1.0127e-04, ..., 6.1035e-04, + 4.3178e-04, 4.8494e-04], + [ 2.1964e-05, 4.6402e-05, 1.0800e-04, ..., 1.1930e-03, + 5.1117e-04, 8.5688e-04], + [ 3.9177e-03, 6.0272e-03, 1.2743e-04, ..., 4.6730e-03, + 7.1573e-04, 1.3170e-03], + ..., + [ 5.2571e-05, 1.3411e-04, 3.9029e-04, ..., 1.1053e-03, + 8.5497e-04, 9.9277e-04], + [-4.3750e-05, 1.1176e-04, 1.0592e-04, ..., 1.0300e-03, + 8.3828e-04, 8.3923e-04], + [ 5.1528e-05, 1.0502e-04, -2.7370e-04, ..., 8.2588e-04, + 2.9802e-04, 5.9223e-04]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0105, 0.0120, -0.0010, 0.0135, -0.0075, -0.0066, 0.0098, 0.0214, + -0.0337, 0.0499], device='cuda:0'), grad: tensor([ 0.0091, -0.0406, 0.0278, 0.0133, -0.0452, 0.0212, -0.0452, 0.0221, + 0.0207, 0.0169], device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 216.48, cls_loss 0.4428 cls_loss_mapping 0.0010 cls_loss_causal 0.4165 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.03 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.0809, 0.0715, -0.1181, ..., -0.1259, -0.1259, 0.0229], + [-0.0890, -0.1789, 0.0098, ..., -0.0536, -0.0384, -0.1167], + [ 0.0306, -0.1005, 0.0572, ..., 0.1639, -0.1439, -0.0397], + ..., + [-0.1249, -0.2198, 0.0885, ..., -0.0157, -0.0552, 0.0763], + [ 0.0302, 0.0398, -0.0369, ..., -0.0918, -0.1059, -0.0246], + [-0.1633, -0.0009, -0.0015, ..., -0.1834, 0.1179, 0.0295]], + device='cuda:0'), grad: tensor([[ 1.6242e-05, 7.0818e-06, 4.5709e-06, ..., 2.3674e-06, + 6.7055e-05, 2.1175e-05], + [-1.0097e-04, 5.0753e-05, 7.8261e-05, ..., 1.5676e-05, + -2.8944e-04, 6.0469e-05], + [ 4.2051e-05, 1.9372e-04, 1.2314e-04, ..., 9.2208e-05, + 1.0026e-04, 7.0930e-05], + ..., + [-8.6904e-05, 1.0140e-05, -6.5947e-04, ..., -3.2973e-04, + 1.1736e-04, -3.1543e-04], + [ 2.6588e-03, 1.2550e-02, 7.4267e-05, ..., 5.6887e-04, + 6.3667e-03, 9.4950e-05], + [ 8.6367e-05, 3.6097e-04, 2.0046e-03, ..., 5.0831e-04, + 4.7951e-03, 4.7302e-03]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0103, 0.0119, -0.0010, 0.0135, -0.0074, -0.0066, 0.0100, 0.0213, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([ 5.7869e-03, -7.4387e-04, -4.1847e-03, -6.9678e-05, -3.8223e-03, + -3.7292e-02, -1.9882e-02, 7.7705e-03, 3.5217e-02, 1.7227e-02], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 216.32, cls_loss 0.4228 cls_loss_mapping 0.0009 cls_loss_causal 0.4000 re_mapping 0.0040 re_causal 0.0115 /// teacc 99.05 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.0809, 0.0717, -0.1181, ..., -0.1260, -0.1259, 0.0230], + [-0.0890, -0.1789, 0.0099, ..., -0.0535, -0.0385, -0.1167], + [ 0.0306, -0.1005, 0.0571, ..., 0.1639, -0.1439, -0.0395], + ..., + [-0.1250, -0.2197, 0.0885, ..., -0.0157, -0.0551, 0.0762], + [ 0.0301, 0.0397, -0.0370, ..., -0.0920, -0.1060, -0.0246], + [-0.1633, -0.0009, -0.0015, ..., -0.1834, 0.1180, 0.0296]], + device='cuda:0'), grad: tensor([[-5.6103e-06, -5.4151e-05, -1.4520e-04, ..., 2.0421e-04, + 1.0884e-04, 1.1587e-03], + [-1.3041e-04, -4.5514e-04, -5.7817e-05, ..., -3.1948e-04, + -7.5722e-04, 1.4436e-04], + [-4.1084e-03, -2.2945e-03, -6.7558e-03, ..., -5.7602e-03, + -2.7847e-03, -7.8506e-03], + ..., + [ 2.8804e-05, 7.9334e-05, 2.0301e-04, ..., 1.3781e-03, + -2.8839e-03, 1.0881e-03], + [ 1.5381e-02, 1.1749e-02, 5.5351e-03, ..., 2.5368e-03, + 4.2000e-03, 1.4286e-03], + [ 3.0594e-03, 1.7700e-03, 7.1640e-03, ..., 9.8877e-03, + 1.9302e-02, 1.0674e-02]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0104, 0.0119, -0.0010, 0.0136, -0.0074, -0.0065, 0.0099, 0.0212, + -0.0340, 0.0501], device='cuda:0'), grad: tensor([-0.0179, -0.0124, -0.0436, 0.0193, -0.0374, 0.0033, -0.0209, 0.0006, + 0.0588, 0.0504], device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 216.47, cls_loss 0.4833 cls_loss_mapping 0.0009 cls_loss_causal 0.4515 re_mapping 0.0040 re_causal 0.0122 /// teacc 99.02 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.0810, 0.0717, -0.1181, ..., -0.1259, -0.1257, 0.0231], + [-0.0889, -0.1789, 0.0098, ..., -0.0536, -0.0386, -0.1167], + [ 0.0308, -0.1003, 0.0571, ..., 0.1638, -0.1439, -0.0396], + ..., + [-0.1250, -0.2198, 0.0885, ..., -0.0156, -0.0552, 0.0761], + [ 0.0302, 0.0397, -0.0369, ..., -0.0920, -0.1056, -0.0243], + [-0.1635, -0.0010, -0.0017, ..., -0.1835, 0.1179, 0.0296]], + device='cuda:0'), grad: tensor([[ 6.0290e-05, 2.0012e-05, -1.0767e-03, ..., -9.5415e-04, + -2.4147e-03, -2.8915e-03], + [-7.7553e-03, 8.5263e-07, 1.8167e-04, ..., 2.6870e-04, + 6.1464e-04, 6.4898e-04], + [-1.4992e-03, 1.6168e-05, -2.6150e-03, ..., -4.9324e-03, + -9.2850e-03, -1.4359e-02], + ..., + [ 1.8239e-04, 1.8645e-06, 9.1171e-04, ..., 6.7902e-04, + 1.3676e-03, 1.5936e-03], + [ 7.2746e-03, 8.4400e-05, 1.4818e-04, ..., 3.1376e-04, + 6.2609e-04, 5.5027e-04], + [ 8.2552e-05, -8.1658e-06, 6.5279e-04, ..., 4.9496e-04, + 8.5783e-04, 1.2350e-03]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0104, 0.0118, -0.0010, 0.0135, -0.0073, -0.0065, 0.0099, 0.0211, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([-0.0102, -0.0666, -0.0101, -0.0094, 0.0201, -0.0419, 0.0193, 0.0266, + 0.0491, 0.0232], device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 216.60, cls_loss 0.4285 cls_loss_mapping 0.0008 cls_loss_causal 0.3956 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.04 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.0810, 0.0716, -0.1182, ..., -0.1259, -0.1257, 0.0230], + [-0.0889, -0.1789, 0.0097, ..., -0.0537, -0.0387, -0.1168], + [ 0.0308, -0.1003, 0.0571, ..., 0.1637, -0.1440, -0.0397], + ..., + [-0.1252, -0.2198, 0.0886, ..., -0.0155, -0.0552, 0.0762], + [ 0.0303, 0.0398, -0.0369, ..., -0.0919, -0.1056, -0.0244], + [-0.1634, -0.0011, -0.0017, ..., -0.1836, 0.1179, 0.0296]], + device='cuda:0'), grad: tensor([[ 6.2525e-05, 5.3585e-05, 2.3097e-05, ..., 8.5175e-05, + 1.7059e-04, 3.9601e-04], + [ 2.4531e-06, 3.8669e-06, 5.4315e-06, ..., 4.4137e-05, + 1.4913e-04, 3.6597e-04], + [ 1.8632e-04, 1.3959e-04, 6.8486e-05, ..., 1.2612e-04, + 2.2602e-04, 4.2605e-04], + ..., + [ 6.6459e-05, 2.9659e-04, 5.4932e-04, ..., 9.9689e-06, + 1.3256e-03, 9.3412e-04], + [-3.9101e-03, -9.6607e-04, 1.7628e-05, ..., 4.9710e-05, + -1.5574e-03, 3.8815e-04], + [ 3.5667e-03, 5.7888e-04, -7.5483e-04, ..., 7.5877e-05, + 1.8787e-04, -4.4012e-04]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0104, 0.0118, -0.0010, 0.0135, -0.0074, -0.0065, 0.0100, 0.0212, + -0.0339, 0.0502], device='cuda:0'), grad: tensor([ 0.0086, 0.0112, 0.0088, 0.0106, -0.0535, 0.0095, 0.0095, -0.0191, + 0.0021, 0.0125], device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 216.32, cls_loss 0.4682 cls_loss_mapping 0.0008 cls_loss_causal 0.4390 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.03 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.0810, 0.0715, -0.1181, ..., -0.1259, -0.1257, 0.0230], + [-0.0889, -0.1790, 0.0097, ..., -0.0537, -0.0387, -0.1168], + [ 0.0310, -0.1002, 0.0572, ..., 0.1638, -0.1439, -0.0395], + ..., + [-0.1253, -0.2199, 0.0886, ..., -0.0155, -0.0552, 0.0760], + [ 0.0301, 0.0399, -0.0369, ..., -0.0919, -0.1056, -0.0244], + [-0.1635, -0.0012, -0.0018, ..., -0.1838, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 7.1228e-05, -3.6550e-04, 1.4818e-04, ..., 4.9210e-04, + 2.9397e-04, 3.7313e-04], + [-8.2910e-05, 6.6221e-05, 5.1975e-04, ..., -6.9666e-04, + 9.9277e-04, -8.1825e-04], + [ 9.9182e-04, 2.5415e-04, 1.0109e-03, ..., 3.9368e-03, + 1.1892e-03, 2.7733e-03], + ..., + [-5.4121e-04, 3.1710e-05, -1.7729e-03, ..., -2.8095e-03, + -3.2578e-03, -6.2408e-03], + [-8.8024e-04, -4.7517e-04, -6.8331e-04, ..., 9.5785e-05, + 5.1171e-05, 7.7343e-04], + [ 1.0967e-04, 1.2082e-04, 1.6665e-04, ..., 4.5466e-04, + 4.1842e-04, 7.0143e-04]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0104, 0.0118, -0.0009, 0.0135, -0.0073, -0.0066, 0.0100, 0.0210, + -0.0340, 0.0502], device='cuda:0'), grad: tensor([ 0.0108, 0.0147, 0.0037, 0.0157, -0.0440, 0.0362, -0.0125, -0.0474, + 0.0087, 0.0140], device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 216.66, cls_loss 0.4724 cls_loss_mapping 0.0008 cls_loss_causal 0.4421 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.04 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.0810, 0.0715, -0.1181, ..., -0.1260, -0.1257, 0.0230], + [-0.0889, -0.1792, 0.0099, ..., -0.0536, -0.0388, -0.1169], + [ 0.0310, -0.1002, 0.0571, ..., 0.1637, -0.1440, -0.0395], + ..., + [-0.1252, -0.2199, 0.0886, ..., -0.0155, -0.0552, 0.0760], + [ 0.0302, 0.0400, -0.0368, ..., -0.0918, -0.1056, -0.0244], + [-0.1633, -0.0011, -0.0017, ..., -0.1837, 0.1180, 0.0298]], + device='cuda:0'), grad: tensor([[ 9.0837e-04, 6.7663e-04, 1.1623e-04, ..., -3.6731e-06, + -1.2960e-03, 7.8917e-04], + [ 5.0783e-04, 2.9778e-04, 1.7381e-04, ..., 6.9849e-09, + -8.6832e-04, 4.2076e-03], + [ 1.0605e-03, 7.4387e-04, 4.3297e-03, ..., 1.8951e-02, + 9.8610e-04, 1.7410e-02], + ..., + [ 9.5701e-04, 9.9421e-05, -5.9624e-03, ..., -1.9577e-02, + 3.0651e-03, -3.2837e-02], + [ 3.2749e-03, 2.5024e-03, 9.5797e-04, ..., 7.0572e-05, + 1.7347e-03, 2.7237e-03], + [-1.6623e-03, 2.0850e-04, -1.7834e-03, ..., 1.8513e-04, + -6.3248e-03, -2.5024e-03]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0106, 0.0118, -0.0009, 0.0137, -0.0073, -0.0066, 0.0098, 0.0209, + -0.0339, 0.0501], device='cuda:0'), grad: tensor([-0.0133, -0.0083, 0.0523, 0.0215, 0.0312, -0.0011, -0.0126, -0.0652, + 0.0002, -0.0046], device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 216.73, cls_loss 0.4634 cls_loss_mapping 0.0009 cls_loss_causal 0.4359 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.01 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.0811, 0.0714, -0.1180, ..., -0.1260, -0.1256, 0.0231], + [-0.0888, -0.1790, 0.0098, ..., -0.0536, -0.0387, -0.1169], + [ 0.0311, -0.1001, 0.0572, ..., 0.1637, -0.1440, -0.0396], + ..., + [-0.1253, -0.2197, 0.0885, ..., -0.0154, -0.0553, 0.0761], + [ 0.0300, 0.0398, -0.0367, ..., -0.0918, -0.1056, -0.0245], + [-0.1633, -0.0012, -0.0018, ..., -0.1837, 0.1180, 0.0296]], + device='cuda:0'), grad: tensor([[ 1.4186e-04, -4.5156e-04, 1.4865e-04, ..., 6.8665e-05, + 7.4482e-04, 4.7040e-04], + [ 1.9407e-04, 1.6832e-04, 2.3448e-04, ..., -7.7486e-04, + 7.4053e-04, -4.9973e-04], + [ 2.6965e-04, 2.8324e-04, 4.3392e-04, ..., 1.1082e-03, + 7.9966e-04, 1.4572e-03], + ..., + [ 3.3236e-04, 3.1686e-04, 4.4417e-04, ..., 3.2043e-04, + 1.1816e-03, 1.0386e-03], + [ 2.5902e-03, -7.2575e-04, -6.8378e-04, ..., -8.1205e-04, + -4.4274e-04, 3.6192e-04], + [ 5.5981e-04, 1.8673e-06, -3.3226e-03, ..., -7.0763e-04, + -2.4624e-03, -2.4872e-03]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0107, 0.0119, -0.0009, 0.0137, -0.0074, -0.0065, 0.0097, 0.0209, + -0.0340, 0.0501], device='cuda:0'), grad: tensor([ 0.0212, 0.0230, 0.0297, -0.0010, -0.0105, -0.1033, 0.0238, 0.0302, + -0.0072, -0.0059], device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 216.65, cls_loss 0.4510 cls_loss_mapping 0.0008 cls_loss_causal 0.4208 re_mapping 0.0039 re_causal 0.0114 /// teacc 99.05 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.0811, 0.0714, -0.1180, ..., -0.1260, -0.1255, 0.0231], + [-0.0887, -0.1789, 0.0096, ..., -0.0536, -0.0388, -0.1168], + [ 0.0310, -0.1001, 0.0572, ..., 0.1638, -0.1440, -0.0396], + ..., + [-0.1254, -0.2197, 0.0885, ..., -0.0154, -0.0553, 0.0761], + [ 0.0298, 0.0398, -0.0367, ..., -0.0920, -0.1057, -0.0245], + [-0.1633, -0.0014, -0.0017, ..., -0.1837, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 7.5150e-04, 1.0498e-05, -8.4591e-04, ..., 1.7369e-04, + 4.3893e-04, 9.9945e-04], + [ 2.0754e-04, 4.0717e-06, 2.2328e-04, ..., -8.5473e-05, + 3.7837e-04, 1.0786e-03], + [ 9.3222e-04, 1.6540e-05, -2.1210e-03, ..., -1.9350e-03, + -1.7653e-03, 1.0366e-03], + ..., + [ 1.0166e-03, 8.8364e-06, -1.3885e-02, ..., -2.6188e-03, + -5.8212e-03, -1.6998e-02], + [ 8.3637e-04, 1.1700e-04, 5.4646e-04, ..., 3.8862e-04, + 7.7295e-04, 9.5797e-04], + [ 5.0354e-04, 2.2650e-05, 1.4000e-03, ..., 4.8470e-04, + -1.7395e-03, 1.9007e-03]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0106, 0.0119, -0.0009, 0.0137, -0.0074, -0.0065, 0.0097, 0.0209, + -0.0339, 0.0501], device='cuda:0'), grad: tensor([ 0.0114, -0.0097, -0.0381, 0.0009, 0.0521, -0.0019, 0.0238, -0.0437, + -0.0025, 0.0076], device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 216.18, cls_loss 0.4095 cls_loss_mapping 0.0007 cls_loss_causal 0.3793 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.04 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.0810, 0.0714, -0.1181, ..., -0.1259, -0.1255, 0.0232], + [-0.0888, -0.1791, 0.0096, ..., -0.0536, -0.0387, -0.1169], + [ 0.0309, -0.1000, 0.0571, ..., 0.1638, -0.1441, -0.0398], + ..., + [-0.1254, -0.2198, 0.0886, ..., -0.0154, -0.0553, 0.0763], + [ 0.0297, 0.0399, -0.0366, ..., -0.0921, -0.1057, -0.0245], + [-0.1633, -0.0015, -0.0018, ..., -0.1837, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 9.3579e-06, -1.6093e-06, 1.3697e-04, ..., 1.1909e-04, + 4.5389e-05, 1.1194e-04], + [ 1.1019e-05, 2.3637e-06, 1.5748e-04, ..., 2.4498e-05, + 7.5817e-05, 2.8253e-04], + [ 1.8820e-05, 1.2353e-05, 2.3115e-04, ..., 1.4334e-03, + 4.5776e-04, 1.3390e-03], + ..., + [ 1.3717e-05, 1.8580e-06, -2.0866e-03, ..., -4.2033e-04, + -1.0643e-03, -4.1866e-04], + [-2.4259e-05, -3.3647e-05, 3.3545e-04, ..., 2.7275e-04, + 9.8288e-05, 2.5868e-04], + [-1.5283e-04, 4.9062e-06, -1.5819e-04, ..., 2.6250e-04, + 6.4671e-05, 1.0198e-04]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0107, 0.0120, -0.0009, 0.0135, -0.0074, -0.0065, 0.0096, 0.0210, + -0.0339, 0.0500], device='cuda:0'), grad: tensor([ 0.0155, 0.0188, 0.0252, -0.0681, 0.0190, 0.0155, -0.0139, -0.0157, + 0.0193, -0.0158], device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 217.25, cls_loss 0.4391 cls_loss_mapping 0.0007 cls_loss_causal 0.4055 re_mapping 0.0040 re_causal 0.0120 /// teacc 98.99 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.0809, 0.0714, -0.1182, ..., -0.1259, -0.1255, 0.0232], + [-0.0888, -0.1791, 0.0094, ..., -0.0536, -0.0387, -0.1169], + [ 0.0308, -0.1001, 0.0569, ..., 0.1636, -0.1441, -0.0397], + ..., + [-0.1255, -0.2198, 0.0887, ..., -0.0152, -0.0553, 0.0764], + [ 0.0298, 0.0400, -0.0364, ..., -0.0919, -0.1056, -0.0245], + [-0.1635, -0.0015, -0.0018, ..., -0.1836, 0.1178, 0.0296]], + device='cuda:0'), grad: tensor([[-1.4126e-04, -6.0921e-03, -8.3237e-03, ..., -5.9357e-03, + -1.5478e-03, -3.5973e-03], + [ 1.3046e-05, 1.1545e-04, 3.1143e-05, ..., -2.2948e-04, + -2.7313e-03, 6.2561e-04], + [-3.7193e-05, 4.4203e-04, 3.7819e-05, ..., -3.8385e-05, + 4.0364e-04, 4.8184e-04], + ..., + [ 2.1279e-05, 2.3806e-04, 2.2113e-04, ..., 2.3580e-04, + 9.1696e-04, 8.1444e-04], + [ 3.6955e-04, -1.3294e-03, 3.9667e-05, ..., 1.1855e-04, + 3.6120e-04, -9.0456e-04], + [ 1.3316e-04, 6.2485e-03, 8.2016e-03, ..., 6.0272e-03, + 1.4334e-03, 1.0729e-03]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0107, 0.0118, -0.0009, 0.0136, -0.0073, -0.0066, 0.0096, 0.0211, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([-0.0542, -0.0143, 0.0069, 0.0079, 0.0083, 0.0052, 0.0085, 0.0110, + -0.0204, 0.0412], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 457---------------------------------------------------- +epoch 457, time 217.69, cls_loss 0.4512 cls_loss_mapping 0.0009 cls_loss_causal 0.4250 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.11 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.0811, 0.0712, -0.1182, ..., -0.1260, -0.1255, 0.0231], + [-0.0888, -0.1791, 0.0094, ..., -0.0535, -0.0386, -0.1169], + [ 0.0308, -0.1000, 0.0570, ..., 0.1636, -0.1440, -0.0398], + ..., + [-0.1254, -0.2198, 0.0887, ..., -0.0151, -0.0553, 0.0764], + [ 0.0297, 0.0399, -0.0365, ..., -0.0920, -0.1055, -0.0246], + [-0.1633, -0.0015, -0.0018, ..., -0.1838, 0.1179, 0.0296]], + device='cuda:0'), grad: tensor([[ 1.5354e-04, 1.0324e-04, 5.2378e-06, ..., 5.8937e-04, + 1.0538e-04, 2.6608e-04], + [ 4.6039e-04, 8.6904e-05, 4.5443e-04, ..., 3.0327e-04, + -8.2111e-04, -2.4242e-03], + [ 2.2774e-03, 6.5565e-04, 1.4901e-04, ..., 3.8738e-03, + 3.8815e-04, 4.3631e-04], + ..., + [-5.4359e-05, -3.8123e-04, -9.7656e-04, ..., 1.2469e-04, + 1.1468e-04, 2.9016e-04], + [-4.8518e-04, 2.2221e-04, 3.8356e-05, ..., 6.9857e-04, + 1.3685e-04, -1.4853e-04], + [ 5.1200e-05, -1.2192e-02, -2.8870e-02, ..., -8.7280e-03, + -4.6600e-02, -2.9434e-02]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0105, 0.0118, -0.0008, 0.0136, -0.0071, -0.0067, 0.0096, 0.0210, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([ 0.0087, -0.0213, 0.0281, 0.0161, 0.0387, -0.0174, -0.0106, -0.0253, + 0.0053, -0.0223], device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 217.04, cls_loss 0.4415 cls_loss_mapping 0.0009 cls_loss_causal 0.4134 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.05 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.0811, 0.0713, -0.1184, ..., -0.1260, -0.1255, 0.0231], + [-0.0887, -0.1790, 0.0093, ..., -0.0533, -0.0387, -0.1169], + [ 0.0308, -0.1002, 0.0568, ..., 0.1633, -0.1442, -0.0399], + ..., + [-0.1253, -0.2198, 0.0888, ..., -0.0150, -0.0553, 0.0765], + [ 0.0299, 0.0399, -0.0365, ..., -0.0921, -0.1054, -0.0247], + [-0.1633, -0.0016, -0.0018, ..., -0.1835, 0.1179, 0.0297]], + device='cuda:0'), grad: tensor([[ 6.8398e-03, 5.6148e-05, 8.3208e-05, ..., 2.0534e-05, + 2.2662e-04, 9.7809e-03], + [-1.4260e-05, 8.7738e-05, 9.4175e-05, ..., 2.2233e-05, + 2.7657e-04, 7.4720e-04], + [ 1.3399e-03, 4.7398e-04, 2.2423e-04, ..., 1.0878e-04, + 4.3416e-04, -3.4976e-04], + ..., + [ 3.5453e-04, 7.6175e-05, -1.8120e-03, ..., 2.5108e-05, + 7.5436e-04, 1.3447e-03], + [-3.0842e-03, -1.3494e-03, 5.2929e-04, ..., 2.5243e-05, + 9.9480e-05, 1.3437e-03], + [-7.0691e-05, -2.1255e-04, 3.8314e-04, ..., 1.8191e-04, + -8.2207e-04, 1.5621e-03]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0105, 0.0118, -0.0010, 0.0137, -0.0071, -0.0066, 0.0096, 0.0210, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([ 0.0354, -0.0156, -0.0117, 0.0226, -0.0154, -0.0116, 0.0189, -0.0008, + -0.0117, -0.0102], device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 217.71, cls_loss 0.4337 cls_loss_mapping 0.0008 cls_loss_causal 0.4055 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.01 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.0810, 0.0715, -0.1182, ..., -0.1259, -0.1254, 0.0233], + [-0.0888, -0.1790, 0.0093, ..., -0.0533, -0.0387, -0.1168], + [ 0.0309, -0.1001, 0.0568, ..., 0.1633, -0.1441, -0.0398], + ..., + [-0.1253, -0.2200, 0.0889, ..., -0.0150, -0.0555, 0.0764], + [ 0.0297, 0.0398, -0.0366, ..., -0.0921, -0.1054, -0.0247], + [-0.1632, -0.0016, -0.0017, ..., -0.1835, 0.1180, 0.0297]], + device='cuda:0'), grad: tensor([[-1.4648e-03, -5.0278e-03, 1.3161e-04, ..., 3.5405e-05, + -2.2912e-04, 9.0599e-05], + [ 8.5413e-05, 1.3900e-04, 6.6662e-04, ..., 7.2777e-05, + 4.8876e-04, 4.9162e-04], + [-2.8801e-04, 1.1253e-04, 6.7472e-04, ..., -7.7426e-05, + 3.8338e-04, -2.2221e-04], + ..., + [ 1.1845e-03, 5.7131e-05, 2.0719e-04, ..., -1.5152e-04, + -3.0651e-03, -4.2610e-03], + [ 5.8317e-04, 8.3590e-04, 5.5408e-04, ..., 7.8738e-05, + 6.8951e-04, 4.6992e-04], + [ 2.6536e-04, 2.1625e-04, -4.7951e-03, ..., -2.1470e-04, + 4.8590e-04, 1.8368e-03]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0106, 0.0118, -0.0009, 0.0136, -0.0071, -0.0065, 0.0095, 0.0210, + -0.0338, 0.0500], device='cuda:0'), grad: tensor([-0.0080, 0.0269, 0.0197, -0.0050, 0.0274, -0.0221, 0.0431, -0.0285, + -0.0390, -0.0145], device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 217.06, cls_loss 0.4530 cls_loss_mapping 0.0008 cls_loss_causal 0.4268 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.04 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.0811, 0.0714, -0.1183, ..., -0.1259, -0.1256, 0.0232], + [-0.0889, -0.1791, 0.0093, ..., -0.0532, -0.0388, -0.1168], + [ 0.0310, -0.0999, 0.0570, ..., 0.1635, -0.1440, -0.0397], + ..., + [-0.1254, -0.2200, 0.0887, ..., -0.0151, -0.0556, 0.0763], + [ 0.0298, 0.0398, -0.0366, ..., -0.0921, -0.1054, -0.0245], + [-0.1632, -0.0015, -0.0016, ..., -0.1836, 0.1182, 0.0298]], + device='cuda:0'), grad: tensor([[-0.0002, -0.0075, 0.0006, ..., -0.0024, 0.0010, 0.0008], + [ 0.0001, 0.0002, 0.0003, ..., 0.0011, -0.0053, -0.0038], + [ 0.0005, 0.0010, 0.0006, ..., 0.0028, 0.0004, 0.0006], + ..., + [-0.0004, -0.0007, -0.0034, ..., -0.0001, -0.0027, 0.0005], + [ 0.0028, 0.0057, -0.0084, ..., 0.0012, 0.0008, -0.0014], + [ 0.0004, 0.0008, 0.0082, ..., 0.0021, 0.0018, 0.0010]], + device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0104, 0.0118, -0.0008, 0.0135, -0.0071, -0.0064, 0.0096, 0.0210, + -0.0337, 0.0500], device='cuda:0'), grad: tensor([ 0.0267, -0.0100, -0.0082, -0.0064, -0.0146, -0.0024, 0.0142, -0.0161, + -0.0286, 0.0453], device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 216.71, cls_loss 0.4449 cls_loss_mapping 0.0008 cls_loss_causal 0.4189 re_mapping 0.0041 re_causal 0.0119 /// teacc 99.03 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.0810, 0.0715, -0.1184, ..., -0.1259, -0.1257, 0.0232], + [-0.0887, -0.1791, 0.0092, ..., -0.0533, -0.0387, -0.1169], + [ 0.0309, -0.1000, 0.0569, ..., 0.1635, -0.1441, -0.0397], + ..., + [-0.1254, -0.2201, 0.0888, ..., -0.0151, -0.0556, 0.0763], + [ 0.0299, 0.0398, -0.0367, ..., -0.0922, -0.1053, -0.0245], + [-0.1632, -0.0017, -0.0016, ..., -0.1836, 0.1182, 0.0299]], + device='cuda:0'), grad: tensor([[-2.6870e-04, -1.3962e-03, 1.3363e-04, ..., -8.0019e-06, + 1.0204e-03, 6.9904e-04], + [ 1.7357e-04, 7.6890e-05, 1.3936e-04, ..., 2.2280e-04, + 5.1641e-04, 7.8773e-04], + [ 1.1616e-03, 5.5075e-04, 1.3828e-03, ..., 6.9809e-04, + 4.4727e-04, 1.1663e-03], + ..., + [ 6.2084e-04, 1.2712e-03, 2.0180e-03, ..., 1.1930e-03, + 1.4219e-03, 2.9106e-03], + [-1.9760e-02, 3.9148e-04, 6.1417e-04, ..., 4.0579e-04, + 1.0252e-03, 2.3232e-03], + [-2.5034e-05, -2.0084e-03, -3.2959e-03, ..., -1.9588e-03, + -1.5659e-03, -8.3084e-03]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0104, 0.0118, -0.0009, 0.0135, -0.0071, -0.0063, 0.0094, 0.0209, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([-0.0077, 0.0186, 0.0339, 0.0028, -0.0174, 0.0372, 0.0342, 0.0019, + 0.0013, -0.1049], device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 217.80, cls_loss 0.4521 cls_loss_mapping 0.0009 cls_loss_causal 0.4247 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.02 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.0809, 0.0717, -0.1183, ..., -0.1257, -0.1256, 0.0232], + [-0.0887, -0.1790, 0.0092, ..., -0.0533, -0.0387, -0.1169], + [ 0.0310, -0.1001, 0.0567, ..., 0.1633, -0.1442, -0.0398], + ..., + [-0.1253, -0.2200, 0.0889, ..., -0.0150, -0.0555, 0.0764], + [ 0.0300, 0.0398, -0.0367, ..., -0.0923, -0.1053, -0.0246], + [-0.1633, -0.0017, -0.0015, ..., -0.1836, 0.1182, 0.0299]], + device='cuda:0'), grad: tensor([[-4.4942e-04, -5.4789e-04, 1.0128e-07, ..., 6.0940e-04, + 5.3197e-06, -4.0627e-04], + [ 7.2327e-03, 3.4199e-03, 1.3672e-06, ..., 8.2003e-07, + 2.3003e-03, 1.0967e-05], + [ 3.1757e-04, 1.5211e-04, 1.2410e-07, ..., 1.7956e-06, + 2.7101e-06, 1.2326e-04], + ..., + [ 1.1422e-05, 1.8522e-05, 1.1645e-05, ..., 1.2964e-06, + 3.0726e-05, 1.6242e-05], + [-7.7477e-03, -3.3569e-03, 2.9188e-06, ..., 1.0524e-06, + -2.2831e-03, 1.4901e-05], + [ 3.9160e-05, 1.8477e-05, -6.3255e-06, ..., 3.4958e-05, + 1.5092e-04, 8.7440e-05]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0106, 0.0118, -0.0011, 0.0136, -0.0071, -0.0064, 0.0093, 0.0209, + -0.0337, 0.0502], device='cuda:0'), grad: tensor([ 0.0053, 0.0042, 0.0064, 0.0068, 0.0034, 0.0059, 0.0078, 0.0070, + -0.0230, -0.0238], device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 217.88, cls_loss 0.4424 cls_loss_mapping 0.0008 cls_loss_causal 0.4202 re_mapping 0.0040 re_causal 0.0116 /// teacc 99.03 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.0812, 0.0715, -0.1182, ..., -0.1258, -0.1257, 0.0233], + [-0.0885, -0.1790, 0.0092, ..., -0.0534, -0.0386, -0.1167], + [ 0.0311, -0.1001, 0.0567, ..., 0.1633, -0.1441, -0.0397], + ..., + [-0.1253, -0.2201, 0.0889, ..., -0.0150, -0.0555, 0.0764], + [ 0.0300, 0.0397, -0.0366, ..., -0.0922, -0.1053, -0.0247], + [-0.1632, -0.0015, -0.0016, ..., -0.1836, 0.1184, 0.0300]], + device='cuda:0'), grad: tensor([[ 3.3879e-04, 1.6708e-03, -3.0413e-05, ..., -1.0500e-03, + -2.8753e-04, 6.1417e-03], + [ 2.4271e-04, 3.8838e-04, 9.4846e-06, ..., 1.8644e-04, + 7.0143e-04, 1.3113e-03], + [ 5.0592e-04, 6.1095e-05, 8.8751e-05, ..., 2.4462e-04, + 6.5374e-04, 1.9722e-03], + ..., + [ 2.0158e-04, 4.2349e-05, -2.1505e-04, ..., 9.0539e-05, + 2.5272e-04, 9.0790e-04], + [ 4.7255e-04, 1.5497e-04, 6.3658e-05, ..., 1.8919e-04, + 6.0701e-04, 2.1381e-03], + [ 3.1352e-04, 1.8692e-04, 5.3614e-05, ..., 1.1593e-04, + 4.0960e-04, 1.7166e-03]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0105, 0.0118, -0.0011, 0.0136, -0.0072, -0.0064, 0.0094, 0.0209, + -0.0336, 0.0502], device='cuda:0'), grad: tensor([-0.0093, 0.0023, 0.0218, -0.0091, -0.0699, 0.0180, -0.0057, 0.0143, + 0.0201, 0.0176], device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 217.88, cls_loss 0.4515 cls_loss_mapping 0.0009 cls_loss_causal 0.4242 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.04 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.0812, 0.0716, -0.1181, ..., -0.1259, -0.1255, 0.0232], + [-0.0887, -0.1792, 0.0092, ..., -0.0533, -0.0386, -0.1167], + [ 0.0310, -0.1000, 0.0566, ..., 0.1632, -0.1444, -0.0400], + ..., + [-0.1254, -0.2202, 0.0888, ..., -0.0150, -0.0555, 0.0763], + [ 0.0301, 0.0398, -0.0366, ..., -0.0922, -0.1054, -0.0248], + [-0.1631, -0.0016, -0.0016, ..., -0.1837, 0.1184, 0.0300]], + device='cuda:0'), grad: tensor([[ 1.3518e-04, 2.8804e-05, 3.0780e-04, ..., 1.4043e-04, + 1.5271e-04, 4.1986e-04], + [ 1.0812e-04, 4.2766e-06, 3.2783e-04, ..., 1.4234e-04, + -1.3981e-03, 4.8828e-04], + [ 2.3317e-04, 8.6486e-05, 4.7898e-04, ..., 1.9479e-04, + 1.8954e-04, 5.8556e-04], + ..., + [-7.8773e-04, 5.6177e-06, -2.7122e-03, ..., -1.1349e-03, + 1.8156e-04, -3.6201e-03], + [ 1.0860e-04, 2.3632e-03, 8.5974e-04, ..., 1.7643e-04, + 3.2082e-03, 7.4625e-04], + [ 1.0556e-04, 4.4861e-03, 1.1024e-03, ..., 8.3089e-05, + 5.9242e-03, 9.2745e-04]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0107, 0.0118, -0.0012, 0.0136, -0.0071, -0.0065, 0.0094, 0.0208, + -0.0336, 0.0502], device='cuda:0'), grad: tensor([ 0.0151, -0.0411, 0.0182, 0.0149, 0.0292, 0.0116, -0.0060, -0.0470, + 0.0177, -0.0126], device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.44, cls_loss 0.4386 cls_loss_mapping 0.0009 cls_loss_causal 0.4102 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.09 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.0812, 0.0716, -0.1182, ..., -0.1259, -0.1254, 0.0234], + [-0.0887, -0.1793, 0.0092, ..., -0.0534, -0.0387, -0.1169], + [ 0.0310, -0.1000, 0.0569, ..., 0.1633, -0.1443, -0.0400], + ..., + [-0.1253, -0.2202, 0.0888, ..., -0.0151, -0.0555, 0.0763], + [ 0.0300, 0.0398, -0.0367, ..., -0.0922, -0.1053, -0.0249], + [-0.1630, -0.0016, -0.0016, ..., -0.1837, 0.1183, 0.0299]], + device='cuda:0'), grad: tensor([[ 8.1584e-06, 9.2089e-06, 1.4435e-07, ..., 2.5928e-06, + 4.1413e-04, 1.0338e-03], + [ 4.3124e-05, 5.0843e-05, 9.3877e-06, ..., 1.1943e-05, + 8.6212e-04, 1.3380e-03], + [-1.5007e-02, 3.1024e-05, -2.5369e-06, ..., -6.2485e-03, + 4.0078e-04, -1.6832e-03], + ..., + [ 1.0687e-04, 1.1139e-05, -1.0721e-05, ..., 4.0650e-05, + 2.6894e-04, -3.8971e-02], + [ 6.6936e-05, 3.6955e-05, 9.2387e-06, ..., 8.7440e-05, + 7.4863e-04, 1.1969e-03], + [-1.3091e-05, -1.2481e-04, -3.0115e-05, ..., 1.2822e-05, + -4.9448e-04, 4.0741e-02]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0107, 0.0118, -0.0011, 0.0135, -0.0071, -0.0064, 0.0093, 0.0208, + -0.0337, 0.0503], device='cuda:0'), grad: tensor([ 0.0173, 0.0217, -0.0309, -0.0285, -0.0414, 0.0171, 0.0246, -0.0433, + 0.0178, 0.0456], device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 217.20, cls_loss 0.4512 cls_loss_mapping 0.0007 cls_loss_causal 0.4160 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.03 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.0811, 0.0717, -0.1183, ..., -0.1260, -0.1256, 0.0233], + [-0.0887, -0.1793, 0.0093, ..., -0.0533, -0.0386, -0.1168], + [ 0.0308, -0.1002, 0.0569, ..., 0.1631, -0.1443, -0.0400], + ..., + [-0.1253, -0.2202, 0.0887, ..., -0.0152, -0.0555, 0.0764], + [ 0.0301, 0.0399, -0.0367, ..., -0.0919, -0.1052, -0.0250], + [-0.1631, -0.0016, -0.0016, ..., -0.1837, 0.1184, 0.0298]], + device='cuda:0'), grad: tensor([[ 8.1122e-05, 9.9838e-05, 2.5518e-06, ..., 6.7711e-05, + 6.7234e-04, 1.5574e-03], + [ 1.5711e-06, 4.6827e-06, 7.8902e-06, ..., 3.9041e-06, + 4.9114e-04, 1.5554e-03], + [ 4.4966e-04, 5.4026e-04, 1.9461e-05, ..., 3.6430e-04, + 1.3247e-03, 1.6699e-03], + ..., + [ 2.2292e-05, 3.7819e-05, -1.4353e-04, ..., -3.3766e-05, + 1.2827e-03, 2.8915e-03], + [ 2.7359e-05, 4.4614e-05, 7.6741e-06, ..., 7.8201e-05, + 5.9271e-04, 1.5440e-03], + [ 5.1260e-05, 1.8373e-05, 8.8215e-05, ..., 1.0371e-04, + -7.2823e-03, -6.3858e-03]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0107, 0.0121, -0.0012, 0.0135, -0.0072, -0.0065, 0.0093, 0.0208, + -0.0336, 0.0502], device='cuda:0'), grad: tensor([ 0.0186, -0.0090, -0.0452, 0.0145, -0.0459, 0.0226, 0.0153, 0.0237, + 0.0159, -0.0106], device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 217.70, cls_loss 0.4428 cls_loss_mapping 0.0007 cls_loss_causal 0.4199 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.04 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.0811, 0.0718, -0.1183, ..., -0.1260, -0.1255, 0.0234], + [-0.0888, -0.1793, 0.0094, ..., -0.0532, -0.0386, -0.1168], + [ 0.0308, -0.1002, 0.0568, ..., 0.1630, -0.1443, -0.0400], + ..., + [-0.1253, -0.2202, 0.0888, ..., -0.0151, -0.0556, 0.0763], + [ 0.0302, 0.0400, -0.0365, ..., -0.0919, -0.1050, -0.0249], + [-0.1629, -0.0015, -0.0016, ..., -0.1837, 0.1184, 0.0299]], + device='cuda:0'), grad: tensor([[ 5.7936e-04, -9.0003e-05, 1.7881e-04, ..., 2.6393e-04, + 2.8625e-05, 1.5316e-03], + [-2.4527e-05, -4.2415e-04, -5.1594e-04, ..., -2.7609e-04, + -2.9755e-03, -9.6970e-03], + [-3.1166e-03, 6.5947e-04, 1.2958e-04, ..., 8.2541e-04, + 9.8407e-05, 1.7042e-03], + ..., + [ 9.5272e-04, 4.5657e-05, 2.6875e-03, ..., 1.2183e-04, + 1.5535e-03, -1.9760e-03], + [ 5.3501e-04, 4.7827e-04, 6.2704e-04, ..., 6.0749e-04, + 2.1420e-03, 3.0766e-03], + [ 4.0817e-04, 4.5872e-04, -1.9474e-03, ..., 1.1225e-03, + -1.1501e-03, 6.1035e-04]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0107, 0.0120, -0.0011, 0.0134, -0.0072, -0.0066, 0.0094, 0.0208, + -0.0336, 0.0502], device='cuda:0'), grad: tensor([-0.0096, -0.0249, -0.0065, -0.0100, 0.0253, 0.0186, -0.0319, 0.0245, + 0.0303, -0.0157], device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 217.14, cls_loss 0.4450 cls_loss_mapping 0.0009 cls_loss_causal 0.4216 re_mapping 0.0038 re_causal 0.0111 /// teacc 99.03 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.0811, 0.0717, -0.1183, ..., -0.1261, -0.1255, 0.0235], + [-0.0887, -0.1792, 0.0095, ..., -0.0533, -0.0384, -0.1166], + [ 0.0307, -0.1003, 0.0566, ..., 0.1628, -0.1442, -0.0401], + ..., + [-0.1253, -0.2203, 0.0890, ..., -0.0149, -0.0557, 0.0764], + [ 0.0302, 0.0400, -0.0366, ..., -0.0918, -0.1051, -0.0250], + [-0.1628, -0.0014, -0.0016, ..., -0.1836, 0.1185, 0.0300]], + device='cuda:0'), grad: tensor([[ 4.9133e-03, 3.2692e-03, 1.1975e-04, ..., 2.3925e-04, + 5.6744e-04, 9.1076e-04], + [ 5.9557e-04, 1.8644e-03, -6.5327e-04, ..., 1.8096e-04, + -1.4019e-04, -2.3155e-03], + [ 2.2972e-04, 1.1176e-04, 1.4138e-04, ..., 3.0112e-04, + 6.3467e-04, 1.1044e-03], + ..., + [-1.0996e-03, 5.9366e-05, -3.8600e-04, ..., -2.0256e-03, + -4.2191e-03, -4.6692e-03], + [ 1.0290e-03, 4.8904e-03, 8.7142e-05, ..., 1.9467e-04, + 1.7595e-03, 6.9332e-04], + [ 2.4271e-04, -2.1877e-03, 2.3031e-04, ..., 1.4353e-04, + -8.7833e-04, 7.1049e-04]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0107, 0.0120, -0.0012, 0.0134, -0.0072, -0.0064, 0.0093, 0.0208, + -0.0336, 0.0502], device='cuda:0'), grad: tensor([ 0.0005, -0.0146, -0.0201, 0.0103, 0.0090, 0.0270, -0.0151, -0.0219, + 0.0179, 0.0069], device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 216.76, cls_loss 0.4299 cls_loss_mapping 0.0009 cls_loss_causal 0.4032 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.04 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.0811, 0.0717, -0.1184, ..., -0.1262, -0.1254, 0.0234], + [-0.0889, -0.1793, 0.0093, ..., -0.0534, -0.0384, -0.1167], + [ 0.0309, -0.1003, 0.0566, ..., 0.1629, -0.1442, -0.0400], + ..., + [-0.1254, -0.2203, 0.0891, ..., -0.0149, -0.0557, 0.0764], + [ 0.0303, 0.0400, -0.0367, ..., -0.0918, -0.1050, -0.0248], + [-0.1629, -0.0013, -0.0016, ..., -0.1836, 0.1186, 0.0299]], + device='cuda:0'), grad: tensor([[-2.2203e-05, 3.6221e-03, 5.2929e-04, ..., 1.1816e-03, + 8.0633e-04, 2.2304e-04], + [ 1.6809e-05, 6.1274e-05, 1.3220e-04, ..., -1.1015e-03, + 4.4465e-04, -6.0558e-04], + [ 4.6521e-05, 3.2377e-04, 1.3041e-04, ..., 4.1580e-04, + 2.7370e-04, 4.0889e-04], + ..., + [-6.6340e-05, 1.3672e-05, -1.0614e-03, ..., -1.2350e-03, + -2.4776e-03, -2.8095e-03], + [ 2.0564e-04, 7.0858e-04, 1.3733e-04, ..., 3.7909e-04, + 3.8815e-04, 3.1567e-04], + [ 4.0084e-05, -5.4092e-03, 9.9182e-03, ..., 3.4981e-03, + 2.0340e-02, 1.4969e-02]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0107, 0.0120, -0.0011, 0.0135, -0.0073, -0.0066, 0.0093, 0.0207, + -0.0335, 0.0503], device='cuda:0'), grad: tensor([ 0.0191, -0.0141, 0.0110, 0.0124, -0.0161, -0.0267, 0.0140, -0.0351, + 0.0118, 0.0237], device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 216.90, cls_loss 0.4655 cls_loss_mapping 0.0009 cls_loss_causal 0.4396 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.03 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.0811, 0.0718, -0.1184, ..., -0.1261, -0.1253, 0.0235], + [-0.0889, -0.1795, 0.0093, ..., -0.0534, -0.0383, -0.1167], + [ 0.0308, -0.1004, 0.0567, ..., 0.1628, -0.1443, -0.0401], + ..., + [-0.1253, -0.2204, 0.0889, ..., -0.0149, -0.0557, 0.0764], + [ 0.0303, 0.0400, -0.0367, ..., -0.0916, -0.1050, -0.0247], + [-0.1631, -0.0014, -0.0017, ..., -0.1838, 0.1184, 0.0298]], + device='cuda:0'), grad: tensor([[ 1.2493e-04, -9.7454e-05, 6.3702e-06, ..., 4.0755e-06, + 8.2135e-05, 1.6183e-05], + [ 7.1228e-05, 1.1563e-04, 2.1517e-05, ..., 2.0102e-05, + -2.9325e-05, 4.1397e-07], + [ 1.2326e-04, 2.3079e-04, -2.7633e-04, ..., -2.6822e-04, + 4.2975e-05, 4.8488e-05], + ..., + [ 1.3149e-04, 1.1367e-04, 5.1618e-05, ..., 7.0572e-05, + 4.0263e-05, -1.6010e-04], + [ 3.3402e-04, 2.3115e-04, 5.0306e-05, ..., 4.7237e-05, + 8.1837e-05, 7.2131e-07], + [ 1.0145e-04, 2.3282e-04, 2.0042e-05, ..., 8.0839e-06, + 4.9770e-05, 8.6010e-05]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0108, 0.0119, -0.0011, 0.0135, -0.0072, -0.0066, 0.0095, 0.0208, + -0.0335, 0.0500], device='cuda:0'), grad: tensor([ 0.0046, 0.0039, -0.0283, 0.0027, 0.0037, -0.0237, 0.0202, 0.0034, + 0.0092, 0.0043], device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 217.07, cls_loss 0.4307 cls_loss_mapping 0.0008 cls_loss_causal 0.3957 re_mapping 0.0039 re_causal 0.0112 /// teacc 99.06 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.0810, 0.0718, -0.1184, ..., -0.1261, -0.1254, 0.0235], + [-0.0889, -0.1793, 0.0094, ..., -0.0533, -0.0384, -0.1168], + [ 0.0309, -0.1001, 0.0567, ..., 0.1629, -0.1440, -0.0401], + ..., + [-0.1254, -0.2206, 0.0888, ..., -0.0151, -0.0558, 0.0762], + [ 0.0303, 0.0401, -0.0368, ..., -0.0918, -0.1051, -0.0249], + [-0.1631, -0.0015, -0.0017, ..., -0.1837, 0.1186, 0.0301]], + device='cuda:0'), grad: tensor([[ 1.1241e-04, -1.6136e-03, -1.7047e-04, ..., 6.0230e-05, + -1.0138e-03, -1.1951e-04], + [-7.2384e-04, 2.7227e-04, 8.2731e-05, ..., 7.1585e-05, + 3.3689e-04, -3.3684e-03], + [-6.5446e-05, 1.1158e-04, -3.1900e-04, ..., -1.0843e-03, + 1.1325e-04, 7.5996e-05], + ..., + [ 1.9646e-04, 5.2810e-05, 2.4438e-04, ..., 2.7108e-04, + 3.0303e-04, 9.3031e-04], + [-4.8409e-03, -6.4011e-03, 3.4046e-04, ..., 8.5175e-05, + -2.5845e-04, 9.4414e-04], + [-4.5013e-03, -6.2168e-05, -4.4060e-03, ..., -1.2131e-03, + -1.3819e-03, 2.4605e-04]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0107, 0.0118, -0.0010, 0.0135, -0.0072, -0.0065, 0.0095, 0.0208, + -0.0336, 0.0501], device='cuda:0'), grad: tensor([ 0.0062, -0.0184, 0.0199, 0.0036, -0.0483, -0.0059, 0.0271, 0.0023, + -0.0045, 0.0178], device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 217.02, cls_loss 0.4514 cls_loss_mapping 0.0008 cls_loss_causal 0.4233 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.03 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.0809, 0.0718, -0.1184, ..., -0.1260, -0.1253, 0.0237], + [-0.0889, -0.1794, 0.0094, ..., -0.0533, -0.0384, -0.1168], + [ 0.0308, -0.1002, 0.0569, ..., 0.1629, -0.1438, -0.0399], + ..., + [-0.1252, -0.2206, 0.0889, ..., -0.0150, -0.0558, 0.0763], + [ 0.0303, 0.0402, -0.0368, ..., -0.0918, -0.1052, -0.0250], + [-0.1632, -0.0015, -0.0017, ..., -0.1837, 0.1186, 0.0301]], + device='cuda:0'), grad: tensor([[-1.0520e-04, -3.7909e-04, 7.7486e-05, ..., 1.0669e-04, + 1.8585e-04, 3.3569e-04], + [ 4.7445e-04, 1.8969e-05, 1.1683e-05, ..., 1.5408e-05, + 2.8551e-05, 1.3552e-03], + [ 4.4370e-04, 3.1352e-04, 8.1241e-05, ..., 9.2685e-05, + 1.8919e-04, 6.9904e-04], + ..., + [ 2.2185e-04, 2.8536e-05, 9.8133e-04, ..., 3.1352e-05, + 7.5865e-04, -1.5612e-03], + [ 3.0899e-04, 8.0526e-05, 5.4061e-05, ..., 6.4552e-05, + 1.2648e-04, 5.7745e-04], + [ 2.1327e-04, 1.2863e-04, -9.0837e-04, ..., 1.2010e-04, + -5.1880e-04, 6.3753e-04]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0109, 0.0118, -0.0009, 0.0135, -0.0073, -0.0066, 0.0094, 0.0210, + -0.0336, 0.0501], device='cuda:0'), grad: tensor([ 0.0061, 0.0189, 0.0082, 0.0089, -0.0301, 0.0044, 0.0113, -0.0424, + 0.0080, 0.0066], device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 217.08, cls_loss 0.4582 cls_loss_mapping 0.0007 cls_loss_causal 0.4270 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.03 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.0807, 0.0719, -0.1185, ..., -0.1261, -0.1254, 0.0236], + [-0.0891, -0.1795, 0.0092, ..., -0.0535, -0.0385, -0.1169], + [ 0.0310, -0.1003, 0.0568, ..., 0.1629, -0.1438, -0.0399], + ..., + [-0.1253, -0.2207, 0.0890, ..., -0.0150, -0.0557, 0.0764], + [ 0.0303, 0.0403, -0.0368, ..., -0.0918, -0.1052, -0.0253], + [-0.1633, -0.0015, -0.0017, ..., -0.1838, 0.1187, 0.0301]], + device='cuda:0'), grad: tensor([[ 1.6737e-04, 1.6463e-04, 1.5616e-04, ..., 1.2696e-04, + 5.5933e-04, 5.1498e-03], + [ 1.5986e-04, 1.1891e-04, -1.6749e-04, ..., 1.1134e-04, + 9.1887e-04, 2.6989e-03], + [ 2.8992e-04, 2.0683e-04, 2.5439e-04, ..., 1.0258e-04, + 4.5824e-04, 2.0790e-03], + ..., + [-8.4496e-04, 6.7472e-05, 9.0714e-03, ..., 1.1625e-03, + 3.3236e-04, 5.5313e-03], + [-4.6825e-04, -8.3542e-04, 2.0400e-05, ..., -1.9431e-04, + -5.3072e-04, 5.2452e-03], + [ 5.9986e-04, 1.1581e-04, -1.0063e-02, ..., -1.0548e-03, + -2.6855e-03, -2.5146e-02]], device='cuda:0') +Epoch 475, bias, value: tensor([ 0.0109, 0.0117, -0.0010, 0.0134, -0.0073, -0.0065, 0.0094, 0.0210, + -0.0336, 0.0501], device='cuda:0'), grad: tensor([ 0.0232, 0.0261, 0.0185, 0.0260, -0.0367, -0.0184, 0.0185, 0.0127, + -0.0164, -0.0536], device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 216.66, cls_loss 0.4195 cls_loss_mapping 0.0007 cls_loss_causal 0.3915 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.06 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.0808, 0.0720, -0.1185, ..., -0.1261, -0.1254, 0.0236], + [-0.0892, -0.1796, 0.0092, ..., -0.0535, -0.0385, -0.1168], + [ 0.0311, -0.1003, 0.0568, ..., 0.1630, -0.1438, -0.0399], + ..., + [-0.1253, -0.2209, 0.0890, ..., -0.0149, -0.0558, 0.0764], + [ 0.0302, 0.0403, -0.0369, ..., -0.0917, -0.1052, -0.0254], + [-0.1633, -0.0013, -0.0016, ..., -0.1838, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 2.7090e-05, 3.3468e-05, 5.4479e-05, ..., 2.3901e-05, + 4.8137e-04, 6.5804e-04], + [ 3.4869e-05, 9.1195e-06, 2.3484e-04, ..., 1.2493e-04, + 2.0599e-03, -2.7847e-03], + [-9.1672e-05, 4.2319e-05, 7.0333e-05, ..., -4.6760e-05, + -9.8572e-03, -6.2981e-03], + ..., + [ 4.3869e-05, 2.9087e-04, -8.1539e-04, ..., -1.1110e-03, + 1.6432e-03, 1.0414e-03], + [ 2.2486e-05, 7.5817e-04, 2.0742e-04, ..., 6.6280e-05, + 2.2373e-03, 9.8419e-04], + [-4.8876e-04, -2.8820e-03, -1.4442e-02, ..., -8.4639e-04, + -2.1469e-02, -1.5244e-02]], device='cuda:0') +Epoch 476, bias, value: tensor([ 0.0109, 0.0118, -0.0010, 0.0135, -0.0072, -0.0065, 0.0093, 0.0210, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0052, -0.0153, -0.0517, 0.0103, 0.0453, 0.0122, 0.0138, 0.0011, + 0.0092, -0.0301], device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 216.97, cls_loss 0.4370 cls_loss_mapping 0.0009 cls_loss_causal 0.4068 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.06 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.0807, 0.0720, -0.1186, ..., -0.1262, -0.1255, 0.0236], + [-0.0892, -0.1795, 0.0093, ..., -0.0535, -0.0385, -0.1169], + [ 0.0312, -0.1002, 0.0568, ..., 0.1630, -0.1437, -0.0398], + ..., + [-0.1254, -0.2209, 0.0892, ..., -0.0149, -0.0558, 0.0764], + [ 0.0302, 0.0405, -0.0369, ..., -0.0919, -0.1052, -0.0255], + [-0.1632, -0.0012, -0.0016, ..., -0.1837, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[-0.0029, -0.0025, 0.0018, ..., 0.0006, 0.0002, 0.0017], + [ 0.0059, 0.0003, 0.0002, ..., 0.0002, 0.0001, 0.0009], + [ 0.0026, 0.0025, 0.0009, ..., 0.0010, 0.0002, 0.0011], + ..., + [-0.0005, -0.0012, -0.0027, ..., -0.0018, -0.0033, -0.0008], + [ 0.0076, 0.0054, 0.0007, ..., 0.0006, 0.0006, 0.0010], + [ 0.0004, 0.0003, 0.0004, ..., 0.0003, 0.0004, -0.0007]], + device='cuda:0') +Epoch 477, bias, value: tensor([ 0.0109, 0.0118, -0.0009, 0.0135, -0.0074, -0.0066, 0.0094, 0.0210, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([ 0.0119, 0.0278, 0.0170, 0.0136, -0.0188, -0.0157, -0.0053, -0.0222, + 0.0145, -0.0229], device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 216.55, cls_loss 0.4665 cls_loss_mapping 0.0008 cls_loss_causal 0.4428 re_mapping 0.0039 re_causal 0.0118 /// teacc 99.08 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.0809, 0.0718, -0.1188, ..., -0.1264, -0.1254, 0.0236], + [-0.0894, -0.1796, 0.0092, ..., -0.0535, -0.0387, -0.1172], + [ 0.0311, -0.1001, 0.0566, ..., 0.1630, -0.1439, -0.0399], + ..., + [-0.1254, -0.2209, 0.0892, ..., -0.0149, -0.0557, 0.0765], + [ 0.0303, 0.0406, -0.0367, ..., -0.0916, -0.1052, -0.0256], + [-0.1631, -0.0012, -0.0016, ..., -0.1837, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0003, 0.0005, ..., 0.0006, 0.0017, 0.0014], + [ 0.0002, 0.0005, 0.0008, ..., 0.0005, 0.0020, 0.0015], + [ 0.0012, 0.0013, 0.0015, ..., 0.0022, 0.0026, 0.0023], + ..., + [ 0.0002, -0.0004, -0.0010, ..., -0.0015, -0.0035, -0.0029], + [ 0.0006, 0.0013, 0.0008, ..., 0.0009, 0.0027, 0.0020], + [-0.0015, -0.0028, -0.0002, ..., -0.0021, -0.0051, -0.0043]], + device='cuda:0') +Epoch 478, bias, value: tensor([ 0.0110, 0.0116, -0.0011, 0.0135, -0.0075, -0.0067, 0.0093, 0.0212, + -0.0335, 0.0503], device='cuda:0'), grad: tensor([ 0.0016, 0.0167, 0.0099, -0.0658, 0.0348, 0.0112, -0.0019, -0.0367, + 0.0444, -0.0141], device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 216.77, cls_loss 0.4505 cls_loss_mapping 0.0008 cls_loss_causal 0.4288 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.05 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.0810, 0.0718, -0.1185, ..., -0.1263, -0.1253, 0.0236], + [-0.0895, -0.1797, 0.0092, ..., -0.0535, -0.0389, -0.1172], + [ 0.0311, -0.1001, 0.0566, ..., 0.1630, -0.1439, -0.0400], + ..., + [-0.1256, -0.2210, 0.0891, ..., -0.0150, -0.0557, 0.0765], + [ 0.0304, 0.0407, -0.0368, ..., -0.0916, -0.1053, -0.0257], + [-0.1629, -0.0012, -0.0015, ..., -0.1839, 0.1189, 0.0301]], + device='cuda:0'), grad: tensor([[-1.1879e-02, -1.1311e-03, 1.5144e-03, ..., 1.6651e-03, + -1.4997e-04, -1.2276e-02], + [ 6.1005e-05, 8.3968e-06, 2.5702e-04, ..., 3.5305e-03, + 9.4354e-05, 6.0415e-04], + [ 3.0479e-03, -1.8275e-04, -3.8280e-03, ..., -7.4654e-03, + 1.7929e-04, 5.3482e-03], + ..., + [ 7.0238e-04, 4.5925e-05, -7.3004e-04, ..., -2.5821e-04, + 1.4019e-03, -2.4147e-03], + [ 1.9083e-03, -5.4588e-03, 6.6996e-05, ..., 3.6087e-03, + -8.0414e-03, -3.6850e-03], + [ 1.5812e-03, 5.7297e-03, -7.3075e-05, ..., -1.7509e-05, + 4.0512e-03, 4.3106e-03]], device='cuda:0') +Epoch 479, bias, value: tensor([ 0.0109, 0.0116, -0.0011, 0.0134, -0.0075, -0.0066, 0.0093, 0.0213, + -0.0337, 0.0504], device='cuda:0'), grad: tensor([-0.0345, 0.0332, -0.0654, -0.0113, 0.0247, 0.0500, 0.0039, -0.0319, + 0.0182, 0.0133], device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 216.96, cls_loss 0.4253 cls_loss_mapping 0.0008 cls_loss_causal 0.4042 re_mapping 0.0039 re_causal 0.0110 /// teacc 99.04 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.0809, 0.0717, -0.1185, ..., -0.1263, -0.1254, 0.0238], + [-0.0895, -0.1798, 0.0092, ..., -0.0536, -0.0387, -0.1172], + [ 0.0311, -0.1002, 0.0566, ..., 0.1630, -0.1438, -0.0400], + ..., + [-0.1256, -0.2210, 0.0889, ..., -0.0150, -0.0558, 0.0765], + [ 0.0303, 0.0407, -0.0367, ..., -0.0916, -0.1052, -0.0257], + [-0.1630, -0.0013, -0.0016, ..., -0.1841, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 7.6828e-03, 1.0025e-02, -2.8229e-04, ..., -4.6659e-04, + -5.3902e-03, -1.2350e-03], + [ 1.7166e-04, 2.5344e-04, 4.8971e-04, ..., 1.4949e-04, + 5.2691e-04, 3.1013e-03], + [-4.2038e-03, 3.5715e-04, -9.3794e-04, ..., -1.6928e-03, + -7.0953e-04, -4.3640e-03], + ..., + [ 2.6679e-04, 2.5415e-04, -1.0014e-03, ..., -7.8440e-05, + -1.2379e-03, -2.2423e-04], + [-8.5220e-03, -1.7075e-02, 3.2711e-04, ..., 1.3266e-03, + 5.1737e-04, -2.5085e-02], + [ 1.9825e-04, 3.7956e-04, 3.4666e-04, ..., 1.2147e-04, + 5.1498e-04, 1.1528e-02]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0108, 0.0116, -0.0010, 0.0135, -0.0075, -0.0067, 0.0094, 0.0212, + -0.0336, 0.0504], device='cuda:0'), grad: tensor([-0.0239, 0.0172, -0.0533, 0.0209, 0.0236, 0.0188, 0.0106, 0.0100, + -0.0479, 0.0242], device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 216.76, cls_loss 0.4455 cls_loss_mapping 0.0008 cls_loss_causal 0.4177 re_mapping 0.0038 re_causal 0.0111 /// teacc 99.09 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.0807, 0.0717, -0.1185, ..., -0.1264, -0.1255, 0.0237], + [-0.0895, -0.1798, 0.0092, ..., -0.0537, -0.0387, -0.1172], + [ 0.0310, -0.1003, 0.0564, ..., 0.1629, -0.1438, -0.0401], + ..., + [-0.1256, -0.2210, 0.0891, ..., -0.0149, -0.0558, 0.0765], + [ 0.0305, 0.0408, -0.0366, ..., -0.0915, -0.1052, -0.0255], + [-0.1631, -0.0012, -0.0016, ..., -0.1841, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 5.0664e-05, 1.5333e-05, 1.9586e-04, ..., 1.7121e-05, + 2.2858e-05, 3.5167e-04], + [ 1.8537e-04, 1.8179e-06, 1.1694e-04, ..., 2.6539e-05, + 1.6749e-05, 2.8610e-04], + [-9.5487e-05, -2.9624e-05, -6.4433e-05, ..., -1.0328e-03, + 3.2574e-05, 3.0088e-04], + ..., + [ 2.5082e-04, 2.7835e-05, 3.8986e-03, ..., 3.7384e-04, + 2.4338e-03, -3.2330e-04], + [ 2.1422e-04, 1.0872e-03, 5.4884e-04, ..., 2.4766e-05, + 6.3705e-04, 8.0776e-04], + [-2.3103e-04, 1.0177e-05, -6.6376e-03, ..., 3.9041e-05, + -2.7847e-03, -2.3041e-03]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0109, 0.0117, -0.0011, 0.0134, -0.0076, -0.0066, 0.0093, 0.0212, + -0.0335, 0.0503], device='cuda:0'), grad: tensor([ 0.0084, 0.0082, 0.0060, 0.0074, 0.0142, 0.0088, -0.0227, 0.0018, + 0.0100, -0.0420], device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 217.10, cls_loss 0.4425 cls_loss_mapping 0.0009 cls_loss_causal 0.4164 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.06 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.0807, 0.0716, -0.1185, ..., -0.1264, -0.1256, 0.0237], + [-0.0895, -0.1799, 0.0092, ..., -0.0537, -0.0389, -0.1173], + [ 0.0309, -0.1005, 0.0563, ..., 0.1628, -0.1439, -0.0401], + ..., + [-0.1256, -0.2211, 0.0891, ..., -0.0148, -0.0558, 0.0765], + [ 0.0304, 0.0407, -0.0368, ..., -0.0916, -0.1053, -0.0255], + [-0.1631, -0.0012, -0.0016, ..., -0.1840, 0.1188, 0.0299]], + device='cuda:0'), grad: tensor([[-1.4853e-04, 6.9094e-04, 3.9786e-06, ..., 9.0748e-06, + 6.1417e-04, 2.1124e-04], + [ 4.4405e-06, 3.0327e-04, 9.1866e-06, ..., 1.3731e-05, + 1.3435e-04, 5.7364e-04], + [-1.1063e-03, 4.3297e-04, -2.1458e-03, ..., -2.5215e-03, + 1.3161e-04, 8.0287e-05], + ..., + [ 2.0161e-05, 6.5863e-05, 1.5640e-04, ..., 3.3498e-05, + 2.0313e-04, 1.0481e-03], + [ 5.1689e-04, 4.3964e-04, 9.9182e-04, ..., 8.4400e-04, + 1.8656e-04, 4.2629e-04], + [ 1.2267e-04, 4.5776e-04, 6.2103e-03, ..., 5.7034e-06, + 6.1073e-03, 1.0735e-02]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0108, 0.0116, -0.0011, 0.0136, -0.0075, -0.0064, 0.0092, 0.0212, + -0.0338, 0.0504], device='cuda:0'), grad: tensor([ 0.0073, 0.0028, -0.0022, -0.0175, -0.0247, 0.0038, 0.0018, 0.0025, + 0.0047, 0.0215], device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 216.81, cls_loss 0.4460 cls_loss_mapping 0.0007 cls_loss_causal 0.4120 re_mapping 0.0038 re_causal 0.0115 /// teacc 98.97 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.0808, 0.0715, -0.1185, ..., -0.1265, -0.1256, 0.0235], + [-0.0895, -0.1797, 0.0092, ..., -0.0538, -0.0388, -0.1173], + [ 0.0310, -0.1006, 0.0562, ..., 0.1628, -0.1438, -0.0401], + ..., + [-0.1254, -0.2211, 0.0891, ..., -0.0148, -0.0557, 0.0766], + [ 0.0303, 0.0407, -0.0367, ..., -0.0915, -0.1053, -0.0256], + [-0.1632, -0.0012, -0.0017, ..., -0.1842, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 3.4070e-04, 1.6248e-04, 2.1875e-04, ..., 3.8958e-04, + 7.2289e-04, 1.2245e-03], + [ 1.3888e-04, 7.6294e-05, 1.1587e-04, ..., 4.0007e-04, + 1.7948e-03, 2.2640e-03], + [ 2.4843e-04, 3.1161e-04, 5.6982e-04, ..., 9.5224e-04, + 6.9380e-04, 1.4362e-03], + ..., + [ 5.8317e-04, 3.7813e-04, 7.9679e-04, ..., 5.7602e-04, + 2.4204e-03, 3.0384e-03], + [ 3.0804e-04, 2.6584e-04, -3.2210e-04, ..., -2.8782e-03, + -3.3927e-04, 4.9353e-04], + [ 3.1662e-04, 1.5640e-03, -2.2984e-03, ..., -2.6536e-04, + 1.5764e-03, -1.8063e-03]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0107, 0.0117, -0.0011, 0.0137, -0.0073, -0.0064, 0.0092, 0.0212, + -0.0339, 0.0504], device='cuda:0'), grad: tensor([-0.0117, 0.0250, 0.0209, -0.0267, 0.0074, -0.0118, -0.0025, 0.0292, + -0.0427, 0.0130], device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 216.96, cls_loss 0.4521 cls_loss_mapping 0.0009 cls_loss_causal 0.4205 re_mapping 0.0037 re_causal 0.0110 /// teacc 98.94 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.0808, 0.0715, -0.1184, ..., -0.1265, -0.1256, 0.0238], + [-0.0893, -0.1796, 0.0093, ..., -0.0538, -0.0389, -0.1175], + [ 0.0311, -0.1006, 0.0564, ..., 0.1630, -0.1439, -0.0401], + ..., + [-0.1254, -0.2212, 0.0890, ..., -0.0149, -0.0558, 0.0766], + [ 0.0301, 0.0406, -0.0368, ..., -0.0916, -0.1054, -0.0256], + [-0.1632, -0.0011, -0.0016, ..., -0.1842, 0.1190, 0.0302]], + device='cuda:0'), grad: tensor([[ 7.7188e-05, 4.5635e-08, 4.1910e-08, ..., 1.2243e-04, + 5.3167e-04, 1.0185e-03], + [ 6.8657e-06, 1.8463e-05, 2.2557e-06, ..., 2.1839e-04, + 5.5456e-04, -2.8954e-03], + [ 3.2687e-04, 4.3988e-05, 1.8030e-05, ..., 4.5562e-04, + 4.7731e-04, 6.8712e-04], + ..., + [ 5.1647e-05, 6.6578e-05, 9.2834e-06, ..., 2.1183e-04, + 8.8882e-04, -4.9553e-03], + [-9.0027e-04, -4.4525e-05, 4.0233e-07, ..., -7.7677e-04, + 5.0449e-04, 1.0309e-03], + [ 6.5751e-06, 5.8055e-05, -4.0710e-05, ..., 7.9823e-04, + 1.3428e-03, 1.9016e-03]], device='cuda:0') +Epoch 484, bias, value: tensor([ 0.0109, 0.0117, -0.0011, 0.0137, -0.0074, -0.0064, 0.0092, 0.0214, + -0.0341, 0.0504], device='cuda:0'), grad: tensor([ 0.0168, -0.0393, -0.0145, -0.0450, 0.0187, 0.0146, 0.0205, -0.0087, + 0.0128, 0.0241], device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 216.53, cls_loss 0.4510 cls_loss_mapping 0.0008 cls_loss_causal 0.4273 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.04 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.0807, 0.0714, -0.1185, ..., -0.1266, -0.1255, 0.0238], + [-0.0892, -0.1797, 0.0092, ..., -0.0538, -0.0388, -0.1174], + [ 0.0311, -0.1006, 0.0564, ..., 0.1630, -0.1439, -0.0403], + ..., + [-0.1256, -0.2212, 0.0889, ..., -0.0149, -0.0559, 0.0766], + [ 0.0301, 0.0406, -0.0367, ..., -0.0916, -0.1053, -0.0255], + [-0.1630, -0.0010, -0.0016, ..., -0.1843, 0.1189, 0.0300]], + device='cuda:0'), grad: tensor([[ 1.7607e-04, 7.1383e-04, 1.8728e-04, ..., 5.7316e-04, + -2.2106e-03, 1.0675e-04], + [ 1.2672e-04, 3.8123e-04, 3.9697e-04, ..., 2.2995e-04, + 4.8733e-04, 7.9334e-05], + [ 9.0694e-04, -1.7939e-03, 1.8704e-04, ..., -1.8005e-03, + 6.6710e-04, 5.3692e-04], + ..., + [ 5.2166e-04, 1.4913e-04, -5.9748e-04, ..., 5.8055e-05, + 4.6754e-04, 2.7680e-04], + [-2.9221e-03, -2.9507e-03, -6.3467e-04, ..., -9.1553e-04, + -2.8763e-03, -1.7271e-03], + [ 2.0611e-04, 3.5834e-04, 3.4523e-04, ..., 2.1887e-04, + 6.7282e-04, -1.0490e-05]], device='cuda:0') +Epoch 485, bias, value: tensor([ 0.0108, 0.0117, -0.0012, 0.0137, -0.0074, -0.0065, 0.0093, 0.0212, + -0.0339, 0.0503], device='cuda:0'), grad: tensor([-0.0107, 0.0185, -0.0104, 0.0203, 0.0170, -0.0320, 0.0153, 0.0186, + -0.0246, -0.0122], device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 216.74, cls_loss 0.4512 cls_loss_mapping 0.0009 cls_loss_causal 0.4287 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.04 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.0808, 0.0715, -0.1186, ..., -0.1267, -0.1256, 0.0238], + [-0.0892, -0.1798, 0.0092, ..., -0.0539, -0.0389, -0.1175], + [ 0.0313, -0.1004, 0.0563, ..., 0.1629, -0.1438, -0.0403], + ..., + [-0.1255, -0.2213, 0.0890, ..., -0.0148, -0.0559, 0.0767], + [ 0.0301, 0.0407, -0.0366, ..., -0.0914, -0.1052, -0.0254], + [-0.1631, -0.0010, -0.0016, ..., -0.1843, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[-2.0733e-03, -2.3232e-03, 5.6458e-04, ..., -5.2376e-03, + 1.9665e-03, 3.9253e-03], + [ 1.0818e-04, 7.8559e-05, 2.8992e-04, ..., 2.7800e-04, + 1.0180e-04, -5.8556e-04], + [-2.4166e-03, 1.5488e-03, -2.2850e-03, ..., -3.3569e-03, + -6.9427e-03, -1.0994e-02], + ..., + [ 4.1771e-04, 6.9857e-04, -1.1473e-03, ..., 7.3671e-04, + -1.1498e-04, 1.2941e-03], + [ 1.0471e-03, 2.7943e-03, 1.6918e-03, ..., 5.7697e-04, + 3.2082e-03, 2.9583e-03], + [ 2.1565e-04, 1.2951e-03, -2.2554e-04, ..., 4.5681e-04, + -1.2102e-03, -3.4637e-03]], device='cuda:0') +Epoch 486, bias, value: tensor([ 0.0107, 0.0116, -0.0012, 0.0136, -0.0075, -0.0064, 0.0093, 0.0213, + -0.0337, 0.0503], device='cuda:0'), grad: tensor([ 0.0227, -0.0575, -0.0313, 0.0190, 0.0052, -0.0102, 0.0518, -0.0249, + 0.0340, -0.0087], device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 217.03, cls_loss 0.4339 cls_loss_mapping 0.0009 cls_loss_causal 0.4030 re_mapping 0.0038 re_causal 0.0111 /// teacc 99.05 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.0810, 0.0713, -0.1187, ..., -0.1268, -0.1258, 0.0237], + [-0.0892, -0.1798, 0.0090, ..., -0.0539, -0.0389, -0.1175], + [ 0.0316, -0.1003, 0.0563, ..., 0.1630, -0.1437, -0.0403], + ..., + [-0.1254, -0.2212, 0.0892, ..., -0.0148, -0.0557, 0.0767], + [ 0.0300, 0.0407, -0.0365, ..., -0.0913, -0.1051, -0.0253], + [-0.1632, -0.0011, -0.0016, ..., -0.1843, 0.1188, 0.0301]], + device='cuda:0'), grad: tensor([[ 2.2936e-04, 2.7514e-04, 6.1512e-05, ..., 1.2619e-07, + 7.2289e-04, 1.5430e-03], + [ 8.3983e-05, 5.7846e-05, 1.3068e-05, ..., 2.8871e-08, + 4.8876e-04, -3.9711e-03], + [ 7.7486e-04, 6.5804e-04, 1.0145e-04, ..., 2.7474e-08, + 7.2908e-04, 1.6747e-03], + ..., + [ 5.8270e-04, 6.3133e-04, 2.5272e-04, ..., 1.8299e-05, + 1.3266e-03, 1.7872e-03], + [-1.9588e-03, -4.2686e-03, -2.3823e-03, ..., 5.0711e-07, + 1.8415e-03, 8.1778e-04], + [ 2.5415e-04, -3.8528e-03, -6.4802e-04, ..., -9.5308e-05, + -3.2330e-03, -2.1343e-03]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0107, 0.0116, -0.0012, 0.0137, -0.0075, -0.0064, 0.0093, 0.0213, + -0.0337, 0.0503], device='cuda:0'), grad: tensor([ 0.0250, -0.0314, 0.0271, 0.0371, 0.0022, -0.0529, 0.0326, -0.0013, + -0.0219, -0.0164], device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 216.91, cls_loss 0.4222 cls_loss_mapping 0.0008 cls_loss_causal 0.3949 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.04 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.0812, 0.0712, -0.1189, ..., -0.1269, -0.1259, 0.0236], + [-0.0891, -0.1796, 0.0091, ..., -0.0539, -0.0391, -0.1176], + [ 0.0316, -0.1003, 0.0561, ..., 0.1629, -0.1438, -0.0403], + ..., + [-0.1254, -0.2213, 0.0892, ..., -0.0148, -0.0557, 0.0768], + [ 0.0298, 0.0407, -0.0367, ..., -0.0915, -0.1052, -0.0253], + [-0.1631, -0.0010, -0.0016, ..., -0.1841, 0.1189, 0.0301]], + device='cuda:0'), grad: tensor([[ 6.3038e-04, 5.0583e-03, 5.2834e-04, ..., 3.8910e-04, + 6.1846e-04, 1.4794e-04], + [-6.6645e-06, -6.8569e-04, -1.3065e-03, ..., -1.1721e-03, + -6.2764e-05, -5.2023e-04], + [ 7.6115e-05, 1.6797e-04, 3.0780e-04, ..., 1.9169e-04, + 2.7180e-04, 1.0502e-04], + ..., + [-2.3150e-04, 1.8156e-04, -9.3365e-04, ..., -4.2987e-04, + 2.6584e-04, -3.0255e-04], + [ 5.6356e-05, 3.7718e-04, 1.3363e-04, ..., 1.1623e-04, + 1.3399e-03, 8.5294e-05], + [ 2.5719e-05, 1.8001e-04, 3.3593e-04, ..., 1.0484e-04, + 5.4359e-04, 2.2367e-05]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0106, 0.0116, -0.0012, 0.0135, -0.0075, -0.0064, 0.0094, 0.0213, + -0.0338, 0.0505], device='cuda:0'), grad: tensor([-0.0390, 0.0031, 0.0123, 0.0145, 0.0133, -0.0158, 0.0035, 0.0055, + -0.0139, 0.0166], device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 216.71, cls_loss 0.4590 cls_loss_mapping 0.0009 cls_loss_causal 0.4320 re_mapping 0.0037 re_causal 0.0109 /// teacc 99.07 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.0812, 0.0711, -0.1190, ..., -0.1269, -0.1258, 0.0237], + [-0.0892, -0.1796, 0.0093, ..., -0.0539, -0.0391, -0.1175], + [ 0.0318, -0.1002, 0.0562, ..., 0.1631, -0.1437, -0.0402], + ..., + [-0.1254, -0.2212, 0.0892, ..., -0.0149, -0.0555, 0.0768], + [ 0.0298, 0.0407, -0.0366, ..., -0.0915, -0.1053, -0.0254], + [-0.1629, -0.0011, -0.0016, ..., -0.1841, 0.1188, 0.0301]], + device='cuda:0'), grad: tensor([[ 7.0035e-05, 4.0078e-04, 2.1172e-04, ..., 9.0313e-04, + 2.8253e-04, 4.6420e-04], + [ 1.0645e-04, 1.0961e-04, 5.4741e-04, ..., 4.0555e-04, + 5.1975e-04, 6.7568e-04], + [ 2.1088e-04, 1.0620e-02, -2.4071e-03, ..., 3.6373e-03, + 2.9755e-04, -4.2458e-03], + ..., + [ 1.1225e-03, 7.8535e-04, 2.9068e-03, ..., -6.0797e-04, + 2.0885e-03, 1.9293e-03], + [-3.8490e-03, -2.4700e-03, -6.9885e-03, ..., -2.3670e-03, + -5.0888e-03, -2.7485e-03], + [ 1.1616e-03, 9.3746e-04, 2.4967e-03, ..., 1.4191e-03, + 3.0594e-03, 1.8816e-03]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0107, 0.0116, -0.0010, 0.0134, -0.0077, -0.0064, 0.0095, 0.0214, + -0.0338, 0.0505], device='cuda:0'), grad: tensor([ 0.0364, 0.0374, -0.0071, 0.0119, -0.0115, 0.0265, -0.0319, -0.0129, + -0.0347, -0.0142], device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 216.81, cls_loss 0.4277 cls_loss_mapping 0.0008 cls_loss_causal 0.4036 re_mapping 0.0037 re_causal 0.0109 /// teacc 99.06 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.0812, 0.0712, -0.1188, ..., -0.1270, -0.1259, 0.0237], + [-0.0892, -0.1795, 0.0094, ..., -0.0538, -0.0390, -0.1174], + [ 0.0318, -0.1004, 0.0562, ..., 0.1631, -0.1437, -0.0403], + ..., + [-0.1254, -0.2214, 0.0892, ..., -0.0148, -0.0555, 0.0768], + [ 0.0300, 0.0409, -0.0365, ..., -0.0916, -0.1054, -0.0254], + [-0.1628, -0.0011, -0.0016, ..., -0.1840, 0.1188, 0.0301]], + device='cuda:0'), grad: tensor([[ 6.1886e-07, -3.8576e-04, 1.8859e-04, ..., 1.9205e-04, + 1.2980e-03, 1.0595e-03], + [ 5.5611e-05, 4.6372e-05, 9.7394e-05, ..., 1.6332e-04, + 1.0223e-03, 1.2131e-03], + [ 2.6870e-04, 2.4796e-04, 1.1563e-04, ..., 1.8358e-04, + 9.8133e-04, 1.0204e-03], + ..., + [ 2.2006e-04, 6.1607e-04, 2.6369e-04, ..., 1.5581e-04, + 1.6909e-03, 1.0891e-03], + [ 8.5831e-04, 1.2932e-03, -9.0313e-04, ..., -1.6489e-03, + -2.1057e-03, -2.9888e-03], + [ 1.7166e-03, -4.1103e-04, -5.2929e-04, ..., 1.7977e-04, + 3.5324e-03, 5.7554e-04]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0108, 0.0117, -0.0010, 0.0133, -0.0077, -0.0064, 0.0094, 0.0214, + -0.0337, 0.0504], device='cuda:0'), grad: tensor([ 0.0165, -0.0117, -0.0153, -0.0127, -0.0132, -0.0320, 0.0433, 0.0192, + -0.0140, 0.0199], device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 216.98, cls_loss 0.4695 cls_loss_mapping 0.0008 cls_loss_causal 0.4395 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.06 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.0812, 0.0713, -0.1188, ..., -0.1271, -0.1259, 0.0237], + [-0.0891, -0.1794, 0.0095, ..., -0.0538, -0.0391, -0.1174], + [ 0.0316, -0.1004, 0.0560, ..., 0.1631, -0.1438, -0.0404], + ..., + [-0.1253, -0.2214, 0.0893, ..., -0.0146, -0.0555, 0.0768], + [ 0.0301, 0.0409, -0.0365, ..., -0.0915, -0.1053, -0.0255], + [-0.1629, -0.0012, -0.0017, ..., -0.1841, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 1.1402e-04, -8.5449e-03, 5.4687e-05, ..., 1.5162e-05, + 8.1956e-05, 3.0398e-04], + [ 1.5192e-05, 3.9786e-05, 1.6320e-04, ..., 5.0247e-05, + 1.0788e-04, 4.1914e-04], + [ 3.4928e-05, 9.8705e-05, 6.9916e-05, ..., 2.3365e-05, + 7.1645e-05, 3.1686e-04], + ..., + [ 1.0744e-05, 3.6508e-05, 2.4433e-03, ..., 7.4005e-04, + 1.2512e-03, 2.2678e-03], + [-2.1529e-04, 3.9482e-03, 7.9346e-04, ..., 2.5272e-04, + -5.1647e-05, 8.1968e-04], + [ 3.1090e-04, 6.6423e-04, -4.5547e-03, ..., -1.4029e-03, + -1.8167e-03, -3.4256e-03]], device='cuda:0') +Epoch 491, bias, value: tensor([ 0.0108, 0.0116, -0.0011, 0.0134, -0.0075, -0.0064, 0.0093, 0.0214, + -0.0336, 0.0503], device='cuda:0'), grad: tensor([-0.0695, 0.0208, -0.0125, 0.0145, 0.0164, 0.0406, -0.0360, 0.0245, + 0.0047, -0.0036], device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 217.02, cls_loss 0.4379 cls_loss_mapping 0.0008 cls_loss_causal 0.4088 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.01 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.0812, 0.0713, -0.1188, ..., -0.1269, -0.1258, 0.0236], + [-0.0891, -0.1793, 0.0095, ..., -0.0538, -0.0392, -0.1174], + [ 0.0315, -0.1004, 0.0561, ..., 0.1630, -0.1438, -0.0402], + ..., + [-0.1253, -0.2214, 0.0894, ..., -0.0148, -0.0555, 0.0769], + [ 0.0300, 0.0408, -0.0364, ..., -0.0914, -0.1053, -0.0256], + [-0.1630, -0.0012, -0.0017, ..., -0.1841, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 2.5421e-05, 1.8370e-04, 4.6158e-04, ..., 2.6107e-04, + 7.7724e-04, 2.1324e-03], + [ 4.6283e-05, 8.0287e-05, 6.6471e-04, ..., 2.5964e-04, + 7.1955e-04, 1.8711e-03], + [-7.6517e-06, 3.7646e-04, -1.5056e-04, ..., -3.6983e-03, + -3.8223e-03, -1.3283e-02], + ..., + [ 1.1861e-04, 3.2496e-04, 1.0672e-03, ..., 2.8682e-04, + 1.1082e-03, 2.3880e-03], + [ 2.1477e-03, 2.4815e-03, 6.2323e-04, ..., 2.6059e-04, + 2.8496e-03, 2.5063e-03], + [-1.8530e-03, -8.5402e-04, 2.3594e-03, ..., 5.2881e-04, + 2.2316e-03, 5.6419e-03]], device='cuda:0') +Epoch 492, bias, value: tensor([ 0.0109, 0.0116, -0.0011, 0.0134, -0.0075, -0.0065, 0.0094, 0.0214, + -0.0337, 0.0502], device='cuda:0'), grad: tensor([ 0.0319, 0.0318, -0.0611, -0.0660, -0.0569, 0.0252, 0.0241, 0.0016, + 0.0388, 0.0305], device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 216.81, cls_loss 0.4414 cls_loss_mapping 0.0008 cls_loss_causal 0.4236 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.03 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.0813, 0.0712, -0.1189, ..., -0.1269, -0.1259, 0.0236], + [-0.0891, -0.1793, 0.0096, ..., -0.0538, -0.0393, -0.1174], + [ 0.0316, -0.1005, 0.0561, ..., 0.1631, -0.1439, -0.0403], + ..., + [-0.1253, -0.2214, 0.0894, ..., -0.0149, -0.0556, 0.0769], + [ 0.0300, 0.0410, -0.0364, ..., -0.0913, -0.1052, -0.0257], + [-0.1630, -0.0012, -0.0018, ..., -0.1841, 0.1187, 0.0301]], + device='cuda:0'), grad: tensor([[ 1.4639e-03, 6.1464e-04, 2.2864e-04, ..., 1.4296e-07, + 2.0170e-04, 2.9778e-04], + [ 1.4153e-03, 1.6606e-04, 1.4110e-06, ..., 1.2005e-06, + 2.4486e-04, 6.6996e-04], + [ 4.9973e-04, 5.5885e-04, 2.0921e-04, ..., -4.8459e-05, + 1.5771e-04, 9.9850e-04], + ..., + [ 4.5276e-04, 1.5259e-04, 9.5293e-06, ..., 2.3004e-06, + 1.9372e-04, 5.7030e-04], + [ 1.1730e-03, 2.4204e-03, 1.0948e-03, ..., 1.9204e-06, + 1.8632e-04, 3.8099e-04], + [-6.6681e-03, 2.1534e-03, 1.1892e-03, ..., 2.8987e-07, + -7.9298e-04, -4.6654e-03]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0109, 0.0114, -0.0012, 0.0135, -0.0074, -0.0064, 0.0094, 0.0214, + -0.0338, 0.0503], device='cuda:0'), grad: tensor([ 0.0263, 0.0054, 0.0261, -0.0285, 0.0013, 0.0113, -0.0007, -0.0073, + 0.0298, -0.0638], device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 216.73, cls_loss 0.4626 cls_loss_mapping 0.0008 cls_loss_causal 0.4391 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.04 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.0813, 0.0712, -0.1190, ..., -0.1270, -0.1259, 0.0235], + [-0.0892, -0.1795, 0.0097, ..., -0.0538, -0.0393, -0.1175], + [ 0.0317, -0.1004, 0.0561, ..., 0.1632, -0.1436, -0.0402], + ..., + [-0.1253, -0.2214, 0.0894, ..., -0.0149, -0.0557, 0.0767], + [ 0.0300, 0.0409, -0.0365, ..., -0.0913, -0.1052, -0.0258], + [-0.1630, -0.0013, -0.0019, ..., -0.1841, 0.1187, 0.0300]], + device='cuda:0'), grad: tensor([[ 3.5495e-05, 9.2268e-05, 2.7871e-04, ..., 4.1872e-05, + 4.5657e-04, 4.6539e-04], + [ 7.8976e-05, 4.0674e-04, -3.1033e-03, ..., -6.3086e-04, + 6.8665e-04, 7.3135e-05], + [ 6.8605e-05, 3.5858e-04, 4.8757e-04, ..., 1.8048e-04, + 3.8624e-04, 4.5037e-04], + ..., + [ 2.2084e-05, 1.3709e-04, 4.9210e-04, ..., 1.0383e-04, + 7.3910e-04, 3.7932e-04], + [ 8.2016e-05, 1.0242e-03, 2.6894e-04, ..., 7.7009e-05, + 5.6696e-04, 5.5218e-04], + [ 5.3257e-05, 2.7800e-04, 2.4939e-04, ..., -1.4448e-04, + -6.2675e-03, -4.1618e-03]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0108, 0.0113, -0.0010, 0.0136, -0.0073, -0.0065, 0.0094, 0.0213, + -0.0338, 0.0502], device='cuda:0'), grad: tensor([ 0.0106, -0.0262, 0.0173, -0.0111, 0.0137, 0.0176, -0.0112, -0.0156, + 0.0185, -0.0137], device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 217.15, cls_loss 0.4223 cls_loss_mapping 0.0007 cls_loss_causal 0.3950 re_mapping 0.0038 re_causal 0.0113 /// teacc 99.04 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.0814, 0.0712, -0.1191, ..., -0.1272, -0.1261, 0.0234], + [-0.0893, -0.1796, 0.0097, ..., -0.0539, -0.0392, -0.1174], + [ 0.0316, -0.1004, 0.0562, ..., 0.1632, -0.1437, -0.0403], + ..., + [-0.1254, -0.2213, 0.0895, ..., -0.0148, -0.0555, 0.0768], + [ 0.0300, 0.0408, -0.0363, ..., -0.0910, -0.1054, -0.0258], + [-0.1630, -0.0013, -0.0020, ..., -0.1841, 0.1187, 0.0300]], + device='cuda:0'), grad: tensor([[ 1.5569e-04, 2.4223e-04, -2.8300e-04, ..., 4.2059e-06, + 1.2708e-04, 2.0218e-04], + [ 5.2266e-06, 9.6560e-06, 1.2767e-04, ..., 2.2240e-06, + -2.9698e-05, -6.4468e-04], + [ 1.5162e-05, 3.6269e-05, 6.3372e-04, ..., 1.2837e-05, + 1.6898e-05, 5.2977e-04], + ..., + [-2.8687e-03, -4.6921e-03, -1.0254e-02, ..., 4.1611e-06, + -1.0681e-02, -8.8120e-03], + [-1.6518e-03, -2.3880e-03, -4.7207e-05, ..., -2.2125e-04, + -9.0981e-04, 6.5565e-04], + [ 3.0918e-03, 4.9744e-03, 7.7782e-03, ..., 4.0308e-06, + 1.0750e-02, 6.5536e-03]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0107, 0.0114, -0.0011, 0.0136, -0.0074, -0.0064, 0.0095, 0.0214, + -0.0337, 0.0502], device='cuda:0'), grad: tensor([-0.0197, -0.0235, 0.0146, 0.0132, 0.0099, -0.0080, 0.0154, -0.0352, + -0.0118, 0.0450], device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 217.22, cls_loss 0.4415 cls_loss_mapping 0.0006 cls_loss_causal 0.4131 re_mapping 0.0040 re_causal 0.0120 /// teacc 99.00 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.0813, 0.0713, -0.1191, ..., -0.1274, -0.1262, 0.0234], + [-0.0892, -0.1796, 0.0096, ..., -0.0540, -0.0392, -0.1174], + [ 0.0317, -0.1002, 0.0563, ..., 0.1634, -0.1438, -0.0403], + ..., + [-0.1254, -0.2215, 0.0895, ..., -0.0148, -0.0556, 0.0767], + [ 0.0299, 0.0408, -0.0363, ..., -0.0912, -0.1054, -0.0259], + [-0.1631, -0.0013, -0.0020, ..., -0.1840, 0.1187, 0.0300]], + device='cuda:0'), grad: tensor([[ 1.7190e-04, 1.1897e-04, 3.9674e-07, ..., 1.1438e-04, + 1.3542e-03, 1.2579e-03], + [ 2.1064e-04, 2.9635e-04, 7.4916e-06, ..., 1.0079e-04, + 1.0872e-03, 9.8610e-04], + [ 5.1308e-04, 9.5725e-05, -1.8001e-05, ..., 1.0854e-04, + 1.0939e-03, 1.3056e-03], + ..., + [ 1.0961e-04, 9.3102e-05, 6.5029e-05, ..., 1.0902e-04, + 8.3780e-04, 8.8644e-04], + [-5.4264e-04, -1.3571e-03, 8.3968e-06, ..., 1.8656e-04, + 2.2163e-03, 2.8038e-03], + [ 1.9264e-04, 1.6761e-04, -1.1086e-04, ..., 1.0037e-04, + 2.4605e-03, 2.0542e-03]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0108, 0.0114, -0.0012, 0.0136, -0.0074, -0.0063, 0.0094, 0.0214, + -0.0337, 0.0501], device='cuda:0'), grad: tensor([-0.0095, 0.0277, 0.0207, -0.0080, -0.0637, -0.0094, 0.0183, -0.0094, + 0.0061, 0.0274], device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 216.67, cls_loss 0.4412 cls_loss_mapping 0.0007 cls_loss_causal 0.4135 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.03 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.0813, 0.0712, -0.1192, ..., -0.1275, -0.1262, 0.0233], + [-0.0893, -0.1797, 0.0098, ..., -0.0539, -0.0392, -0.1175], + [ 0.0317, -0.1001, 0.0563, ..., 0.1635, -0.1436, -0.0402], + ..., + [-0.1254, -0.2216, 0.0893, ..., -0.0149, -0.0557, 0.0766], + [ 0.0299, 0.0409, -0.0364, ..., -0.0911, -0.1056, -0.0258], + [-0.1632, -0.0013, -0.0018, ..., -0.1840, 0.1188, 0.0300]], + device='cuda:0'), grad: tensor([[ 2.0657e-06, 6.4373e-05, 8.5413e-05, ..., 3.5477e-04, + 2.4438e-04, 5.8413e-04], + [ 3.9525e-06, -3.2735e-04, 2.0683e-04, ..., 3.7980e-04, + 5.8603e-04, 1.0834e-03], + [-4.9067e-04, 4.1932e-05, -1.6129e-04, ..., -3.4580e-03, + 4.2367e-04, -2.2507e-03], + ..., + [ 6.1542e-06, 1.2283e-03, 1.5869e-03, ..., 6.5470e-04, + 1.2293e-03, 1.8520e-03], + [ 4.5359e-05, -6.0916e-05, 8.3351e-04, ..., 1.2035e-03, + 2.3270e-03, 3.4161e-03], + [ 1.3402e-06, -1.1520e-03, -3.2425e-03, ..., -1.2054e-03, + -6.2065e-03, -7.5722e-03]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0108, 0.0114, -0.0011, 0.0135, -0.0074, -0.0063, 0.0094, 0.0213, + -0.0338, 0.0502], device='cuda:0'), grad: tensor([ 0.0096, -0.0227, -0.0251, 0.0087, 0.0072, 0.0062, 0.0068, 0.0238, + 0.0182, -0.0327], device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 217.06, cls_loss 0.4157 cls_loss_mapping 0.0006 cls_loss_causal 0.3936 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.07 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.0814, 0.0712, -0.1193, ..., -0.1276, -0.1261, 0.0234], + [-0.0893, -0.1795, 0.0098, ..., -0.0539, -0.0391, -0.1175], + [ 0.0317, -0.1001, 0.0563, ..., 0.1634, -0.1435, -0.0402], + ..., + [-0.1255, -0.2215, 0.0895, ..., -0.0148, -0.0556, 0.0766], + [ 0.0299, 0.0410, -0.0365, ..., -0.0911, -0.1055, -0.0257], + [-0.1633, -0.0015, -0.0019, ..., -0.1841, 0.1187, 0.0300]], + device='cuda:0'), grad: tensor([[ 7.7724e-04, 4.0308e-06, 2.6202e-04, ..., 3.0589e-04, + 6.6137e-04, 8.1253e-04], + [ 2.0657e-03, 1.6224e-04, 2.6941e-04, ..., 3.0041e-04, + 6.1321e-04, 1.3933e-03], + [-1.1986e-02, 1.0425e-04, 2.5105e-04, ..., 3.1185e-04, + 5.5265e-04, -5.1079e-03], + ..., + [ 7.2670e-04, 1.9729e-04, 3.5214e-04, ..., 3.8481e-04, + 8.3113e-04, 9.0647e-04], + [ 4.3182e-03, -4.7755e-04, -8.7023e-04, ..., -1.6327e-03, + -3.4447e-03, -7.3290e-04], + [ 4.7064e-04, 2.9564e-04, 4.7398e-04, ..., 6.1893e-04, + 1.3752e-03, 1.4954e-03]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0108, 0.0114, -0.0011, 0.0136, -0.0074, -0.0063, 0.0095, 0.0213, + -0.0338, 0.0501], device='cuda:0'), grad: tensor([ 0.0196, 0.0331, -0.0700, 0.0259, -0.0157, -0.0440, 0.0195, 0.0242, + -0.0090, 0.0162], device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 217.29, cls_loss 0.4603 cls_loss_mapping 0.0008 cls_loss_causal 0.4342 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.05 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.0813, 0.0713, -0.1193, ..., -0.1277, -0.1262, 0.0234], + [-0.0895, -0.1796, 0.0099, ..., -0.0540, -0.0391, -0.1175], + [ 0.0319, -0.1002, 0.0563, ..., 0.1634, -0.1435, -0.0401], + ..., + [-0.1255, -0.2216, 0.0896, ..., -0.0148, -0.0557, 0.0766], + [ 0.0297, 0.0410, -0.0365, ..., -0.0911, -0.1056, -0.0258], + [-0.1633, -0.0015, -0.0021, ..., -0.1841, 0.1186, 0.0298]], + device='cuda:0'), grad: tensor([[ 6.0081e-04, -9.1410e-04, -1.0914e-04, ..., 1.5929e-05, + 1.8030e-05, -8.3923e-04], + [ 1.7338e-03, 6.3705e-04, -6.5528e-06, ..., 7.7367e-05, + 2.9278e-03, 7.7391e-04], + [ 4.4060e-04, 4.2057e-04, 6.4373e-04, ..., 4.0412e-04, + 5.0879e-04, 9.4891e-04], + ..., + [ 7.3767e-04, 9.4271e-04, -4.9925e-04, ..., -3.9029e-04, + 1.4391e-03, 8.9169e-04], + [ 7.4816e-04, 6.4325e-04, 1.7977e-04, ..., 8.8751e-05, + 1.0424e-03, 8.7595e-04], + [ 4.3344e-04, 6.8760e-04, -9.8586e-05, ..., 7.8917e-05, + 4.1056e-04, 9.6893e-04]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0109, 0.0113, -0.0011, 0.0135, -0.0074, -0.0061, 0.0095, 0.0214, + -0.0339, 0.0500], device='cuda:0'), grad: tensor([-0.0012, 0.0013, 0.0156, -0.0103, -0.0131, -0.0437, 0.0311, 0.0160, + 0.0174, -0.0129], device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 217.09, cls_loss 0.4028 cls_loss_mapping 0.0008 cls_loss_causal 0.3797 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.02 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.0814, 0.0713, -0.1192, ..., -0.1277, -0.1262, 0.0233], + [-0.0896, -0.1795, 0.0099, ..., -0.0543, -0.0391, -0.1175], + [ 0.0318, -0.1003, 0.0563, ..., 0.1635, -0.1434, -0.0400], + ..., + [-0.1255, -0.2216, 0.0894, ..., -0.0149, -0.0559, 0.0765], + [ 0.0297, 0.0410, -0.0364, ..., -0.0911, -0.1056, -0.0257], + [-0.1632, -0.0014, -0.0019, ..., -0.1843, 0.1187, 0.0299]], + device='cuda:0'), grad: tensor([[ 6.9962e-03, -1.0315e-02, 2.1164e-02, ..., -1.7262e-04, + 5.6744e-04, 6.8808e-04], + [ 1.9455e-04, 9.9063e-05, 5.8460e-04, ..., 2.8014e-04, + 4.0746e-04, 8.6260e-04], + [ 6.7186e-04, 3.3855e-04, -5.9843e-04, ..., -1.3762e-03, + -1.0805e-03, -2.9182e-03], + ..., + [ 4.9978e-05, 6.0111e-05, -4.2458e-03, ..., -9.1457e-04, + -7.9727e-04, -3.8567e-03], + [-5.1880e-04, 5.1346e-03, 8.1825e-04, ..., 7.2289e-04, + 6.7997e-04, 7.6723e-04], + [ 6.8235e-04, 6.1750e-05, 1.1806e-03, ..., 6.8808e-04, + -6.6221e-05, 1.0767e-03]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0108, 0.0113, -0.0011, 0.0136, -0.0073, -0.0061, 0.0097, 0.0213, + -0.0339, 0.0499], device='cuda:0'), grad: tensor([ 0.0377, 0.0120, -0.0233, 0.0055, 0.0143, -0.0003, -0.0074, -0.0510, + -0.0055, 0.0179], device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 217.72, cls_loss 0.4445 cls_loss_mapping 0.0007 cls_loss_causal 0.4180 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.02 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.940002 98.889999 ... 89.287498 76.723734 +ShearY 98.930000 98.930000 ... 89.287498 73.559136 +AutoContrast 98.949997 99.019997 ... 89.287498 66.443730 +Invert 99.000000 99.000000 ... 89.287498 72.944757 +Equalize 98.419998 98.409996 ... 89.287498 66.992412 +Solarize 98.449997 98.570000 ... 89.287498 69.236865 +SolarizeAdd 98.680000 98.610001 ... 89.287498 72.770916 +Posterize 98.979996 99.019997 ... 89.287498 76.629822 +Contrast 99.180000 99.190002 ... 89.287498 78.385061 +Color 99.040001 99.080002 ... 89.287498 67.632051 +Brightness 99.119995 99.199997 ... 89.287498 77.591179 +Sharpness 99.040001 99.080002 ... 89.287498 78.275154 +NoiseSalt 98.979996 98.970001 ... 89.287498 70.400456 +NoiseGaussian 98.970001 99.080002 ... 89.287498 64.850480 +w/o do (original x) 99.080000 0.000000 ... 0.000000 79.583664 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.05 70.586202 77.891345 79.681775 89.586447 79.436442 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last +randm: False +stride: 3 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.949997 98.970001 ... 89.486794 76.881126 +ShearY 98.909996 98.909996 ... 89.486794 73.554698 +AutoContrast 98.930000 99.019997 ... 89.486794 66.492076 +Invert 98.979996 99.000000 ... 89.486794 72.618074 +Equalize 98.419998 98.459999 ... 89.486794 67.132525 +Solarize 98.470001 98.580002 ... 89.486794 68.999302 +SolarizeAdd 98.619995 98.659996 ... 89.486794 72.501183 +Posterize 98.949997 99.019997 ... 89.486794 76.556273 +Contrast 99.169998 99.199997 ... 89.486794 78.399466 +Color 99.059998 99.029999 ... 89.486794 67.984040 +Brightness 99.159996 99.190002 ... 89.486794 77.668425 +Sharpness 99.010002 99.049995 ... 89.486794 78.452616 +NoiseSalt 99.049995 99.019997 ... 89.486794 70.660622 +NoiseGaussian 99.019997 99.029999 ... 89.486794 65.199389 +w/o do (original x) 99.030000 0.000000 ... 0.000000 79.652907 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.08 70.816687 77.602489 79.838794 89.78575 79.51093 diff --git a/Meta-causal/code-withStyleAttack/71582.error b/Meta-causal/code-withStyleAttack/71582.error new file mode 100644 index 0000000000000000000000000000000000000000..6299d70f46fb33ff0c8abffa58005ba3dd92ae75 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71582.error @@ -0,0 +1,22 @@ +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:45: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:62: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:72: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:48: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:65: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:75: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) diff --git a/Meta-causal/code-withStyleAttack/71582.log b/Meta-causal/code-withStyleAttack/71582.log new file mode 100644 index 0000000000000000000000000000000000000000..5cbefbea1e2fd98b61c439dfb6d62213d099643d --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71582.log @@ -0,0 +1,13338 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0120, -0.0010, -0.0128, ..., 0.0255, 0.0182, 0.0039], + [-0.0303, 0.0148, -0.0231, ..., 0.0144, -0.0009, -0.0120], + [-0.0132, -0.0048, -0.0112, ..., -0.0214, -0.0085, -0.0015], + ..., + [-0.0015, -0.0011, -0.0129, ..., 0.0289, 0.0076, 0.0227], + [ 0.0211, -0.0194, 0.0239, ..., 0.0201, 0.0085, -0.0168], + [-0.0212, -0.0006, 0.0236, ..., 0.0008, -0.0251, 0.0069]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0239, -0.0221, -0.0014, 0.0055, -0.0262, -0.0104, 0.0130, 0.0054, + 0.0297, -0.0295], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 218.81, cls_loss 2.2794 cls_loss_mapping 2.2514 cls_loss_causal 2.2915 re_mapping 0.0072 re_causal 0.0072 /// teacc 54.17 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0123, -0.0056, -0.0133, ..., 0.0271, 0.0167, 0.0055], + [-0.0298, 0.0210, -0.0236, ..., 0.0114, -0.0003, -0.0132], + [-0.0125, -0.0041, -0.0117, ..., -0.0241, -0.0075, 0.0005], + ..., + [-0.0023, -0.0014, -0.0134, ..., 0.0270, 0.0088, 0.0209], + [ 0.0218, -0.0197, 0.0234, ..., 0.0192, 0.0082, -0.0191], + [-0.0220, -0.0061, 0.0230, ..., 0.0011, -0.0250, 0.0042]], + device='cuda:0'), grad: tensor([[ 0.0000, 0.0032, 0.0000, ..., 0.0035, 0.0021, 0.0042], + [ 0.0000, -0.0105, 0.0000, ..., -0.0030, -0.0035, -0.0079], + [ 0.0000, 0.0007, 0.0000, ..., -0.0040, 0.0025, -0.0085], + ..., + [ 0.0000, 0.0002, 0.0000, ..., -0.0034, -0.0017, -0.0033], + [ 0.0000, -0.0034, 0.0000, ..., 0.0024, -0.0040, 0.0066], + [ 0.0000, 0.0041, 0.0000, ..., 0.0032, 0.0028, 0.0135]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0236, -0.0205, -0.0016, 0.0059, -0.0261, -0.0128, 0.0132, 0.0061, + 0.0273, -0.0298], device='cuda:0'), grad: tensor([ 0.0692, -0.0515, -0.0214, 0.0593, 0.0415, -0.0475, -0.0493, -0.0622, + -0.0302, 0.0920], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 217.25, cls_loss 1.9695 cls_loss_mapping 1.2944 cls_loss_causal 1.9570 re_mapping 0.1093 re_causal 0.1066 /// teacc 84.67 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0203, -0.0097, -0.0147, ..., 0.0343, 0.0105, 0.0094], + [-0.0260, 0.0244, -0.0223, ..., 0.0084, -0.0055, -0.0168], + [-0.0160, 0.0011, -0.0130, ..., -0.0292, -0.0065, 0.0053], + ..., + [-0.0027, -0.0015, -0.0149, ..., 0.0240, 0.0152, 0.0161], + [ 0.0266, -0.0198, 0.0219, ..., 0.0157, 0.0081, -0.0194], + [-0.0194, -0.0099, 0.0215, ..., 0.0016, -0.0215, 0.0012]], + device='cuda:0'), grad: tensor([[ 0.0041, 0.0003, 0.0000, ..., 0.0002, -0.0051, -0.0002], + [-0.0094, -0.0365, 0.0000, ..., 0.0037, -0.0301, -0.0159], + [ 0.0033, 0.0128, 0.0000, ..., 0.0022, 0.0125, 0.0103], + ..., + [ 0.0048, 0.0088, 0.0000, ..., 0.0048, -0.0072, 0.0048], + [ 0.0039, 0.0041, 0.0000, ..., 0.0066, 0.0078, 0.0054], + [ 0.0025, 0.0001, 0.0000, ..., -0.0122, 0.0021, 0.0021]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0231, -0.0206, -0.0020, 0.0055, -0.0263, -0.0116, 0.0121, 0.0071, + 0.0278, -0.0298], device='cuda:0'), grad: tensor([-0.0209, -0.0518, 0.0127, -0.0015, 0.0460, 0.0063, 0.0238, 0.0140, + 0.0218, -0.0504], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 217.20, cls_loss 1.5947 cls_loss_mapping 0.5548 cls_loss_causal 1.5514 re_mapping 0.1244 re_causal 0.1799 /// teacc 91.84 lr 0.00010000 +Epoch 4, weight, value: tensor([[-2.5938e-02, -1.2469e-02, -2.1475e-02, ..., 4.2999e-02, + 6.9269e-03, 1.2125e-02], + [-2.6396e-02, 2.5986e-02, -2.9322e-02, ..., 5.3832e-03, + -9.4213e-03, -2.0880e-02], + [-1.5223e-02, 5.2595e-03, -1.6341e-02, ..., -3.3421e-02, + -2.9382e-03, 1.0018e-02], + ..., + [-6.4098e-03, -3.9228e-04, -2.4613e-02, ..., 2.0228e-02, + 1.6866e-02, 1.3742e-02], + [ 3.0365e-02, -2.1655e-02, 1.8822e-02, ..., 9.8972e-03, + 8.7229e-03, -2.1048e-02], + [-1.5215e-02, -1.4300e-02, 2.0314e-02, ..., -1.0346e-03, + -1.8924e-02, 2.7899e-05]], device='cuda:0'), grad: tensor([[ 6.6681e-03, 1.7033e-03, 9.1612e-05, ..., -3.6888e-03, + 1.6647e-02, -2.0969e-04], + [-7.2174e-03, -4.1504e-03, 6.5756e-04, ..., 8.1205e-04, + -1.4297e-02, -5.7030e-03], + [ 1.7685e-02, 8.0049e-05, 1.3323e-03, ..., 5.2490e-03, + 1.8616e-02, 3.6068e-03], + ..., + [ 3.6564e-03, -1.0033e-03, -7.5874e-03, ..., 2.0638e-03, + -3.4790e-02, 2.4300e-03], + [ 1.3596e-02, -9.4070e-03, 1.7347e-03, ..., 5.2757e-03, + 7.0152e-03, 6.5575e-03], + [ 9.3689e-03, 2.5463e-03, -7.7152e-04, ..., -8.0776e-04, + 1.8997e-02, 3.4695e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0230, -0.0209, -0.0021, 0.0062, -0.0268, -0.0110, 0.0122, 0.0071, + 0.0273, -0.0298], device='cuda:0'), grad: tensor([ 0.0249, -0.0671, 0.0419, -0.0549, 0.0301, 0.0325, -0.0328, -0.0304, + 0.0461, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 217.35, cls_loss 1.3769 cls_loss_mapping 0.3419 cls_loss_causal 1.3131 re_mapping 0.0944 re_causal 0.1656 /// teacc 93.47 lr 0.00010000 +Epoch 5, weight, value: tensor([[-3.0271e-02, -1.3065e-02, -2.4105e-02, ..., 4.6792e-02, + 4.9514e-03, 1.4469e-02], + [-2.9623e-02, 2.9065e-02, -3.0910e-02, ..., 3.9618e-03, + -1.2069e-02, -2.4095e-02], + [-1.2703e-02, 7.2855e-03, -1.9777e-02, ..., -3.4584e-02, + -1.4832e-03, 1.4023e-02], + ..., + [-6.7980e-03, 1.6590e-05, -2.7696e-02, ..., 1.8759e-02, + 1.7525e-02, 1.3227e-02], + [ 3.1546e-02, -2.2049e-02, 1.8277e-02, ..., 8.3874e-03, + 7.6445e-03, -2.2575e-02], + [-1.4415e-02, -1.6883e-02, 2.3732e-02, ..., -3.3791e-03, + -1.7307e-02, -1.3963e-03]], device='cuda:0'), grad: tensor([[ 1.2579e-03, -2.3499e-03, 1.0757e-03, ..., 2.5543e-02, + 3.6335e-03, 6.0349e-03], + [ 3.7599e-04, 6.2227e-05, -1.7834e-03, ..., 1.2722e-03, + -2.4128e-03, 1.7691e-03], + [ 2.9182e-04, -8.5754e-03, 9.6226e-04, ..., 1.0216e-02, + -6.8512e-03, -1.2413e-02], + ..., + [ 1.5249e-03, 2.2964e-03, -2.8539e-04, ..., -4.1795e-04, + -3.5763e-04, 3.8013e-03], + [ 9.6560e-04, -1.2274e-03, 1.5392e-03, ..., -6.9618e-03, + -1.6556e-02, -1.9331e-03], + [-8.1482e-03, 1.7805e-03, 5.3549e-04, ..., 7.4539e-03, + 2.2566e-04, 1.9855e-03]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0230, -0.0211, -0.0016, 0.0059, -0.0274, -0.0112, 0.0126, 0.0073, + 0.0273, -0.0294], device='cuda:0'), grad: tensor([ 0.0732, -0.0196, 0.0040, 0.0591, -0.0646, 0.0346, -0.0478, 0.0014, + -0.0482, 0.0079], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 227.29, cls_loss 1.3115 cls_loss_mapping 0.2684 cls_loss_causal 1.2631 re_mapping 0.0713 re_causal 0.1427 /// teacc 95.05 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0322, -0.0146, -0.0247, ..., 0.0489, 0.0047, 0.0159], + [-0.0300, 0.0305, -0.0323, ..., 0.0046, -0.0129, -0.0256], + [-0.0122, 0.0099, -0.0231, ..., -0.0357, -0.0016, 0.0171], + ..., + [-0.0077, 0.0009, -0.0277, ..., 0.0176, 0.0178, 0.0129], + [ 0.0329, -0.0223, 0.0167, ..., 0.0075, 0.0070, -0.0233], + [-0.0144, -0.0201, 0.0259, ..., -0.0054, -0.0159, -0.0026]], + device='cuda:0'), grad: tensor([[-0.0153, -0.0041, 0.0007, ..., -0.0039, -0.0161, -0.0071], + [ 0.0157, -0.0035, -0.0033, ..., -0.0047, -0.0049, -0.0066], + [ 0.0129, 0.0023, 0.0034, ..., 0.0015, 0.0229, 0.0026], + ..., + [ 0.0057, 0.0012, 0.0016, ..., -0.0009, 0.0138, 0.0025], + [ 0.0302, 0.0023, 0.0037, ..., 0.0045, 0.0209, 0.0028], + [ 0.0242, 0.0018, 0.0132, ..., 0.0038, 0.0393, 0.0031]], + device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0233, -0.0211, -0.0019, 0.0064, -0.0271, -0.0120, 0.0127, 0.0071, + 0.0272, -0.0294], device='cuda:0'), grad: tensor([-0.0734, -0.0536, 0.0490, -0.0058, 0.0039, -0.0195, -0.0358, 0.0335, + 0.0246, 0.0771], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 220.22, cls_loss 1.2216 cls_loss_mapping 0.2157 cls_loss_causal 1.1663 re_mapping 0.0606 re_causal 0.1262 /// teacc 95.23 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0346, -0.0157, -0.0287, ..., 0.0510, 0.0034, 0.0168], + [-0.0318, 0.0333, -0.0350, ..., 0.0045, -0.0142, -0.0282], + [-0.0113, 0.0113, -0.0259, ..., -0.0353, -0.0010, 0.0191], + ..., + [-0.0084, 0.0024, -0.0278, ..., 0.0169, 0.0180, 0.0124], + [ 0.0336, -0.0242, 0.0169, ..., 0.0068, 0.0065, -0.0249], + [-0.0136, -0.0230, 0.0284, ..., -0.0062, -0.0147, -0.0032]], + device='cuda:0'), grad: tensor([[ 0.0219, 0.0033, 0.0008, ..., 0.0063, 0.0163, 0.0225], + [ 0.0018, -0.0050, -0.0089, ..., 0.0017, -0.0070, 0.0055], + [ 0.0159, 0.0068, 0.0100, ..., 0.0052, 0.0288, 0.0246], + ..., + [-0.0022, -0.0059, 0.0026, ..., -0.0035, -0.0173, -0.0099], + [-0.0097, 0.0011, 0.0058, ..., -0.0035, 0.0018, -0.0040], + [ 0.0090, -0.0069, -0.0036, ..., -0.0039, -0.0235, -0.0080]], + device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0234, -0.0212, -0.0016, 0.0060, -0.0271, -0.0113, 0.0122, 0.0067, + 0.0275, -0.0293], device='cuda:0'), grad: tensor([ 0.0349, 0.0010, 0.1001, 0.0396, 0.0006, -0.0239, -0.0413, -0.0488, + 0.0088, -0.0712], device='cuda:0') +100 +0.0001 +changing lr +epoch 6, time 216.53, cls_loss 1.1757 cls_loss_mapping 0.2028 cls_loss_causal 1.1213 re_mapping 0.0520 re_causal 0.1155 /// teacc 94.60 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0341, -0.0159, -0.0325, ..., 0.0518, 0.0024, 0.0179], + [-0.0322, 0.0351, -0.0366, ..., 0.0045, -0.0157, -0.0298], + [-0.0120, 0.0132, -0.0294, ..., -0.0364, -0.0010, 0.0211], + ..., + [-0.0085, 0.0038, -0.0262, ..., 0.0171, 0.0180, 0.0111], + [ 0.0350, -0.0244, 0.0155, ..., 0.0067, 0.0072, -0.0260], + [-0.0141, -0.0248, 0.0299, ..., -0.0073, -0.0140, -0.0029]], + device='cuda:0'), grad: tensor([[ 0.0011, 0.0010, 0.0027, ..., 0.0006, 0.0005, -0.0092], + [-0.0101, -0.0037, -0.0013, ..., 0.0009, -0.0067, -0.0060], + [ 0.0066, 0.0013, 0.0057, ..., 0.0068, 0.0144, 0.0048], + ..., + [ 0.0065, -0.0030, 0.0002, ..., -0.0003, 0.0099, 0.0028], + [ 0.0051, 0.0058, 0.0065, ..., -0.0009, 0.0193, 0.0138], + [-0.0023, 0.0009, 0.0004, ..., -0.0036, -0.0089, -0.0011]], + device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0228, -0.0216, -0.0020, 0.0066, -0.0273, -0.0119, 0.0130, 0.0067, + 0.0279, -0.0290], device='cuda:0'), grad: tensor([-0.0079, -0.0311, 0.0343, 0.0353, -0.0811, 0.0175, 0.0030, 0.0006, + 0.0624, -0.0331], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 217.14, cls_loss 1.1540 cls_loss_mapping 0.1912 cls_loss_causal 1.1180 re_mapping 0.0468 re_causal 0.1153 /// teacc 96.45 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0355, -0.0167, -0.0343, ..., 0.0533, 0.0020, 0.0200], + [-0.0336, 0.0365, -0.0384, ..., 0.0054, -0.0164, -0.0311], + [-0.0118, 0.0142, -0.0322, ..., -0.0362, -0.0012, 0.0233], + ..., + [-0.0081, 0.0057, -0.0266, ..., 0.0155, 0.0183, 0.0109], + [ 0.0358, -0.0258, 0.0155, ..., 0.0054, 0.0069, -0.0271], + [-0.0142, -0.0269, 0.0314, ..., -0.0068, -0.0126, -0.0039]], + device='cuda:0'), grad: tensor([[ 0.0104, 0.0010, 0.0032, ..., 0.0024, 0.0046, 0.0039], + [-0.0137, -0.0048, -0.0004, ..., -0.0035, -0.0047, -0.0033], + [-0.0015, -0.0035, -0.0051, ..., -0.0114, -0.0070, -0.0107], + ..., + [-0.0141, -0.0024, -0.0089, ..., -0.0036, -0.0199, -0.0095], + [ 0.0082, -0.0031, -0.0042, ..., 0.0003, -0.0081, -0.0211], + [-0.0195, 0.0005, -0.0164, ..., -0.0060, -0.0129, -0.0062]], + device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0233, -0.0215, -0.0017, 0.0067, -0.0271, -0.0123, 0.0128, 0.0068, + 0.0274, -0.0292], device='cuda:0'), grad: tensor([ 0.0225, -0.0326, -0.0596, 0.0134, 0.0422, 0.0479, 0.0913, -0.0271, + -0.0580, -0.0400], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 218.09, cls_loss 1.0672 cls_loss_mapping 0.1706 cls_loss_causal 1.0197 re_mapping 0.0429 re_causal 0.1043 /// teacc 96.53 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0361, -0.0177, -0.0351, ..., 0.0544, 0.0012, 0.0204], + [-0.0347, 0.0386, -0.0408, ..., 0.0073, -0.0177, -0.0317], + [-0.0125, 0.0161, -0.0328, ..., -0.0358, -0.0009, 0.0248], + ..., + [-0.0071, 0.0074, -0.0247, ..., 0.0145, 0.0188, 0.0104], + [ 0.0361, -0.0267, 0.0146, ..., 0.0047, 0.0069, -0.0278], + [-0.0151, -0.0285, 0.0317, ..., -0.0076, -0.0125, -0.0038]], + device='cuda:0'), grad: tensor([[ 0.0036, 0.0027, 0.0008, ..., -0.0039, 0.0007, -0.0045], + [ 0.0023, 0.0034, 0.0050, ..., 0.0069, 0.0074, 0.0081], + [ 0.0039, 0.0010, -0.0004, ..., 0.0014, 0.0044, 0.0030], + ..., + [-0.0025, -0.0024, 0.0008, ..., 0.0032, -0.0038, 0.0016], + [-0.0015, -0.0140, -0.0223, ..., -0.0089, -0.0024, -0.0074], + [ 0.0023, 0.0018, 0.0041, ..., 0.0006, 0.0030, -0.0017]], + device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0234, -0.0217, -0.0015, 0.0066, -0.0272, -0.0118, 0.0123, 0.0067, + 0.0273, -0.0290], device='cuda:0'), grad: tensor([ 0.0034, 0.0267, -0.0030, -0.0673, 0.0167, 0.0255, 0.0363, 0.0060, + -0.0528, 0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 9, time 217.11, cls_loss 1.0530 cls_loss_mapping 0.1592 cls_loss_causal 1.0063 re_mapping 0.0394 re_causal 0.1011 /// teacc 96.30 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0359, -0.0182, -0.0366, ..., 0.0547, 0.0015, 0.0210], + [-0.0360, 0.0399, -0.0427, ..., 0.0072, -0.0185, -0.0320], + [-0.0131, 0.0172, -0.0361, ..., -0.0358, -0.0010, 0.0265], + ..., + [-0.0072, 0.0083, -0.0236, ..., 0.0160, 0.0182, 0.0096], + [ 0.0368, -0.0281, 0.0137, ..., 0.0050, 0.0069, -0.0277], + [-0.0149, -0.0312, 0.0325, ..., -0.0091, -0.0117, -0.0046]], + device='cuda:0'), grad: tensor([[ 0.0029, -0.0033, -0.0042, ..., -0.0028, -0.0067, -0.0012], + [ 0.0019, 0.0072, 0.0028, ..., 0.0020, 0.0074, 0.0065], + [-0.0015, -0.0221, 0.0019, ..., -0.0088, -0.0140, -0.0192], + ..., + [-0.0071, -0.0150, -0.0144, ..., -0.0027, -0.0096, -0.0028], + [-0.0001, 0.0048, 0.0035, ..., 0.0034, 0.0033, 0.0037], + [ 0.0018, 0.0086, 0.0045, ..., 0.0039, 0.0035, 0.0027]], + device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0233, -0.0219, -0.0013, 0.0062, -0.0269, -0.0114, 0.0123, 0.0066, + 0.0276, -0.0293], device='cuda:0'), grad: tensor([-0.0107, 0.0264, -0.0624, 0.0223, 0.0152, 0.0233, -0.0048, -0.0468, + 0.0178, 0.0197], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 224.22, cls_loss 1.0390 cls_loss_mapping 0.1514 cls_loss_causal 1.0004 re_mapping 0.0382 re_causal 0.1046 /// teacc 96.74 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0360, -0.0187, -0.0395, ..., 0.0551, 0.0012, 0.0225], + [-0.0362, 0.0411, -0.0436, ..., 0.0082, -0.0192, -0.0331], + [-0.0141, 0.0180, -0.0375, ..., -0.0364, -0.0006, 0.0279], + ..., + [-0.0077, 0.0092, -0.0233, ..., 0.0153, 0.0178, 0.0087], + [ 0.0373, -0.0293, 0.0135, ..., 0.0052, 0.0063, -0.0284], + [-0.0138, -0.0322, 0.0340, ..., -0.0094, -0.0104, -0.0047]], + device='cuda:0'), grad: tensor([[ 2.7409e-03, 4.8141e-03, 3.3131e-03, ..., 5.0659e-03, + 8.4076e-03, 1.1475e-02], + [ 1.0262e-03, -6.1378e-03, -3.0327e-03, ..., -7.2899e-03, + -1.9436e-03, -4.0078e-04], + [-5.2338e-03, -1.1063e-02, 1.3280e-04, ..., -1.6739e-02, + -3.0106e-02, -4.0680e-02], + ..., + [ 1.9970e-03, -9.8646e-05, 1.1772e-02, ..., -3.0098e-03, + 1.3321e-02, -3.3779e-03], + [-1.6899e-03, 5.3482e-03, 1.5020e-03, ..., 5.4016e-03, + 6.4125e-03, 7.2098e-03], + [-8.3923e-03, -4.0512e-03, -1.9577e-02, ..., -1.2856e-03, + -2.4384e-02, -9.3918e-03]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0232, -0.0225, -0.0016, 0.0063, -0.0268, -0.0113, 0.0128, 0.0068, + 0.0278, -0.0294], device='cuda:0'), grad: tensor([ 0.0481, -0.0218, -0.0645, 0.0182, 0.0086, 0.0244, -0.0089, 0.0138, + 0.0408, -0.0586], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 227.86, cls_loss 1.0088 cls_loss_mapping 0.1393 cls_loss_causal 0.9723 re_mapping 0.0333 re_causal 0.0902 /// teacc 97.22 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0365, -0.0186, -0.0388, ..., 0.0556, 0.0008, 0.0226], + [-0.0362, 0.0414, -0.0453, ..., 0.0078, -0.0200, -0.0343], + [-0.0149, 0.0184, -0.0400, ..., -0.0363, -0.0004, 0.0298], + ..., + [-0.0079, 0.0102, -0.0228, ..., 0.0148, 0.0183, 0.0084], + [ 0.0386, -0.0295, 0.0126, ..., 0.0050, 0.0060, -0.0299], + [-0.0145, -0.0332, 0.0347, ..., -0.0100, -0.0100, -0.0045]], + device='cuda:0'), grad: tensor([[-2.7657e-03, -1.2192e-02, -1.6308e-03, ..., -1.2733e-02, + -3.0441e-03, -4.7531e-03], + [-1.5755e-03, 3.6240e-03, 1.6060e-03, ..., 3.1548e-03, + 6.7406e-03, 2.9106e-03], + [ 1.6956e-03, 6.8188e-04, 2.4624e-03, ..., 2.9926e-03, + -2.5024e-03, -3.5648e-03], + ..., + [ 2.5501e-03, 7.7820e-03, 6.5269e-03, ..., 7.6599e-03, + 1.6434e-02, 3.7785e-03], + [ 4.1466e-03, -2.3537e-03, 9.7603e-06, ..., -2.6226e-04, + -2.7394e-04, -1.5879e-03], + [ 7.3662e-03, 5.3291e-03, 2.0027e-03, ..., -3.9101e-03, + -9.4910e-03, 2.5311e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0235, -0.0230, -0.0019, 0.0062, -0.0265, -0.0108, 0.0129, 0.0070, + 0.0276, -0.0295], device='cuda:0'), grad: tensor([-0.0534, 0.0306, 0.0138, 0.0146, -0.0489, 0.0618, -0.0386, 0.0611, + -0.0030, -0.0380], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 226.30, cls_loss 0.9957 cls_loss_mapping 0.1362 cls_loss_causal 0.9603 re_mapping 0.0341 re_causal 0.0938 /// teacc 96.99 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0372, -0.0189, -0.0403, ..., 0.0572, 0.0005, 0.0238], + [-0.0363, 0.0417, -0.0482, ..., 0.0085, -0.0212, -0.0346], + [-0.0147, 0.0205, -0.0416, ..., -0.0352, 0.0003, 0.0315], + ..., + [-0.0076, 0.0098, -0.0216, ..., 0.0139, 0.0182, 0.0074], + [ 0.0384, -0.0297, 0.0128, ..., 0.0043, 0.0064, -0.0295], + [-0.0151, -0.0333, 0.0356, ..., -0.0106, -0.0094, -0.0060]], + device='cuda:0'), grad: tensor([[-8.4457e-03, -7.7324e-03, 2.8744e-03, ..., -1.3313e-02, + -1.9817e-03, -6.8932e-03], + [ 5.5466e-03, -7.4005e-03, -4.9133e-03, ..., -6.5575e-03, + -3.3894e-03, -1.1421e-02], + [ 1.3985e-02, 9.2697e-03, 5.6725e-03, ..., 1.2497e-02, + 1.7471e-02, 4.6883e-03], + ..., + [ 3.0766e-03, 7.4434e-04, -1.7290e-03, ..., 5.1422e-03, + -1.4015e-02, 2.8477e-03], + [ 4.4518e-03, 9.8953e-03, -5.7638e-05, ..., 5.2757e-03, + 2.2934e-02, 3.5782e-03], + [-5.2528e-03, -4.4212e-03, 4.4556e-03, ..., -4.0398e-03, + 4.6425e-03, 2.2240e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0231, -0.0234, -0.0019, 0.0065, -0.0264, -0.0110, 0.0126, 0.0069, + 0.0280, -0.0291], device='cuda:0'), grad: tensor([-0.0309, -0.0472, 0.0699, 0.0070, 0.0033, -0.0710, 0.0076, 0.0069, + 0.0482, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 227.07, cls_loss 0.9521 cls_loss_mapping 0.1235 cls_loss_causal 0.9164 re_mapping 0.0310 re_causal 0.0879 /// teacc 96.92 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0378, -0.0191, -0.0412, ..., 0.0580, 0.0004, 0.0247], + [-0.0372, 0.0428, -0.0489, ..., 0.0089, -0.0215, -0.0352], + [-0.0144, 0.0217, -0.0427, ..., -0.0360, 0.0001, 0.0325], + ..., + [-0.0073, 0.0102, -0.0201, ..., 0.0137, 0.0185, 0.0065], + [ 0.0385, -0.0306, 0.0119, ..., 0.0052, 0.0056, -0.0293], + [-0.0154, -0.0348, 0.0356, ..., -0.0121, -0.0089, -0.0066]], + device='cuda:0'), grad: tensor([[ 0.0061, 0.0147, 0.0106, ..., 0.0080, 0.0118, 0.0014], + [ 0.0316, 0.0413, 0.0088, ..., 0.0143, 0.0214, 0.0052], + [-0.0033, 0.0077, -0.0007, ..., -0.0006, 0.0082, 0.0013], + ..., + [ 0.0021, 0.0098, 0.0086, ..., 0.0053, 0.0143, 0.0043], + [-0.0491, -0.0449, -0.0163, ..., -0.0106, -0.0399, -0.0065], + [ 0.0006, -0.0054, -0.0222, ..., -0.0048, -0.0147, -0.0071]], + device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0235, -0.0233, -0.0025, 0.0066, -0.0265, -0.0107, 0.0121, 0.0073, + 0.0282, -0.0293], device='cuda:0'), grad: tensor([ 0.0541, 0.0702, 0.0065, -0.0316, -0.0043, -0.0299, 0.0477, 0.0434, + -0.1163, -0.0397], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 227.12, cls_loss 0.9381 cls_loss_mapping 0.1244 cls_loss_causal 0.8982 re_mapping 0.0301 re_causal 0.0858 /// teacc 96.85 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0379, -0.0198, -0.0439, ..., 0.0590, -0.0012, 0.0247], + [-0.0374, 0.0436, -0.0497, ..., 0.0098, -0.0220, -0.0354], + [-0.0149, 0.0234, -0.0436, ..., -0.0348, 0.0006, 0.0342], + ..., + [-0.0073, 0.0104, -0.0201, ..., 0.0139, 0.0183, 0.0057], + [ 0.0389, -0.0319, 0.0123, ..., 0.0036, 0.0056, -0.0294], + [-0.0154, -0.0355, 0.0365, ..., -0.0127, -0.0082, -0.0066]], + device='cuda:0'), grad: tensor([[ 0.0026, -0.0023, -0.0016, ..., 0.0066, 0.0118, 0.0074], + [ 0.0002, -0.0085, -0.0015, ..., -0.0057, -0.0097, 0.0004], + [ 0.0003, 0.0082, 0.0046, ..., 0.0052, 0.0110, 0.0124], + ..., + [ 0.0004, -0.0033, -0.0132, ..., -0.0012, -0.0238, -0.0010], + [ 0.0003, -0.0131, -0.0138, ..., -0.0103, -0.0201, -0.0260], + [-0.0037, 0.0033, -0.0021, ..., 0.0049, 0.0034, 0.0055]], + device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0232, -0.0230, -0.0014, 0.0071, -0.0268, -0.0104, 0.0118, 0.0064, + 0.0275, -0.0291], device='cuda:0'), grad: tensor([ 0.0218, -0.0243, 0.0531, 0.0410, 0.0464, -0.0158, -0.0039, -0.0399, + -0.1104, 0.0318], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 227.44, cls_loss 0.9216 cls_loss_mapping 0.1111 cls_loss_causal 0.8931 re_mapping 0.0290 re_causal 0.0830 /// teacc 97.27 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0374, -0.0197, -0.0447, ..., 0.0602, -0.0021, 0.0254], + [-0.0380, 0.0441, -0.0500, ..., 0.0093, -0.0226, -0.0362], + [-0.0166, 0.0247, -0.0454, ..., -0.0347, 0.0006, 0.0349], + ..., + [-0.0074, 0.0105, -0.0191, ..., 0.0145, 0.0186, 0.0053], + [ 0.0400, -0.0330, 0.0127, ..., 0.0034, 0.0058, -0.0295], + [-0.0145, -0.0357, 0.0372, ..., -0.0135, -0.0081, -0.0068]], + device='cuda:0'), grad: tensor([[ 4.2419e-03, -3.9711e-03, 2.7294e-03, ..., -2.4185e-03, + -1.9302e-03, -7.6027e-03], + [ 3.1700e-03, -1.7824e-03, 6.5660e-04, ..., -3.2387e-03, + -8.8751e-05, -2.5234e-03], + [ 3.0308e-03, 4.9896e-03, 8.0109e-03, ..., 1.9608e-03, + 1.3611e-02, 1.2650e-02], + ..., + [ 1.2579e-03, 1.9207e-03, -1.4580e-02, ..., -5.3120e-04, + -2.3365e-03, 2.5597e-03], + [-2.0065e-02, 1.9550e-03, 1.0881e-03, ..., -7.8964e-04, + -2.0103e-03, 1.7996e-03], + [ 1.3237e-02, 8.0919e-04, 4.6806e-03, ..., 1.3714e-03, + 1.0559e-02, 2.3861e-03]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0232, -0.0232, -0.0019, 0.0073, -0.0265, -0.0109, 0.0119, 0.0065, + 0.0279, -0.0290], device='cuda:0'), grad: tensor([-0.0213, -0.0080, 0.0427, 0.0135, -0.0605, 0.0076, 0.0244, -0.0209, + -0.0178, 0.0403], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 230.04, cls_loss 0.9264 cls_loss_mapping 0.1080 cls_loss_causal 0.9001 re_mapping 0.0269 re_causal 0.0803 /// teacc 97.49 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0374, -0.0198, -0.0458, ..., 0.0602, -0.0020, 0.0252], + [-0.0381, 0.0444, -0.0517, ..., 0.0089, -0.0231, -0.0370], + [-0.0167, 0.0252, -0.0453, ..., -0.0349, 0.0004, 0.0355], + ..., + [-0.0069, 0.0115, -0.0183, ..., 0.0143, 0.0193, 0.0047], + [ 0.0402, -0.0327, 0.0131, ..., 0.0040, 0.0059, -0.0291], + [-0.0150, -0.0371, 0.0374, ..., -0.0148, -0.0080, -0.0072]], + device='cuda:0'), grad: tensor([[ 2.1591e-03, 3.4103e-03, 4.7150e-03, ..., 1.7967e-03, + 4.4975e-03, 4.6921e-03], + [-1.5154e-03, 4.5433e-03, 2.9430e-03, ..., 1.4429e-03, + 1.6661e-03, -5.8651e-05], + [ 7.5073e-03, -7.8487e-04, 8.4610e-03, ..., 2.1400e-03, + 8.3389e-03, 2.2972e-04], + ..., + [-2.3251e-03, -1.8387e-02, -4.6616e-03, ..., -8.5907e-03, + -6.8741e-03, -3.3321e-03], + [ 5.3940e-03, 4.0016e-03, 3.1471e-03, ..., 8.2159e-04, + 4.6577e-03, 4.2725e-03], + [ 6.0539e-03, 6.5269e-03, 1.4496e-02, ..., 3.3493e-03, + 1.1932e-02, 8.0872e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0230, -0.0236, -0.0018, 0.0072, -0.0262, -0.0110, 0.0123, 0.0065, + 0.0283, -0.0293], device='cuda:0'), grad: tensor([ 0.0273, -0.0016, 0.0201, -0.0882, 0.0298, -0.0345, 0.0074, -0.0451, + 0.0260, 0.0588], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 227.01, cls_loss 0.8779 cls_loss_mapping 0.0960 cls_loss_causal 0.8401 re_mapping 0.0272 re_causal 0.0792 /// teacc 97.32 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0377, -0.0202, -0.0463, ..., 0.0615, -0.0023, 0.0251], + [-0.0379, 0.0454, -0.0526, ..., 0.0099, -0.0237, -0.0374], + [-0.0169, 0.0256, -0.0450, ..., -0.0354, 0.0011, 0.0371], + ..., + [-0.0073, 0.0118, -0.0188, ..., 0.0138, 0.0191, 0.0045], + [ 0.0399, -0.0329, 0.0118, ..., 0.0042, 0.0054, -0.0298], + [-0.0146, -0.0376, 0.0384, ..., -0.0163, -0.0072, -0.0069]], + device='cuda:0'), grad: tensor([[ 2.7142e-03, -3.0670e-03, -1.7490e-03, ..., -1.1284e-02, + -7.2050e-04, -8.3084e-03], + [ 5.5580e-03, -8.6355e-04, 5.3253e-03, ..., 7.4806e-03, + 1.5306e-04, 4.3335e-03], + [ 2.6913e-03, -9.4986e-04, -6.7115e-05, ..., 6.6805e-04, + 2.8553e-03, 1.3924e-03], + ..., + [ 3.2353e-04, -2.5225e-04, -4.0817e-03, ..., 2.4261e-03, + -5.1575e-03, 8.5831e-04], + [-2.7409e-03, 1.9875e-03, 1.7729e-03, ..., -4.3755e-03, + 3.0918e-03, -3.8242e-03], + [-1.9928e-02, 2.1248e-03, -4.0474e-03, ..., -7.8659e-03, + 2.1725e-03, -6.0415e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0228, -0.0230, -0.0016, 0.0076, -0.0265, -0.0104, 0.0117, 0.0062, + 0.0279, -0.0294], device='cuda:0'), grad: tensor([-0.0386, 0.0388, 0.0058, -0.0323, 0.0366, 0.0414, -0.0032, -0.0114, + 0.0004, -0.0376], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 227.54, cls_loss 0.8909 cls_loss_mapping 0.0976 cls_loss_causal 0.8551 re_mapping 0.0268 re_causal 0.0803 /// teacc 97.63 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0381, -0.0212, -0.0478, ..., 0.0612, -0.0026, 0.0254], + [-0.0375, 0.0458, -0.0530, ..., 0.0098, -0.0238, -0.0379], + [-0.0184, 0.0269, -0.0456, ..., -0.0340, 0.0009, 0.0382], + ..., + [-0.0065, 0.0122, -0.0180, ..., 0.0132, 0.0192, 0.0041], + [ 0.0404, -0.0333, 0.0111, ..., 0.0033, 0.0051, -0.0302], + [-0.0153, -0.0383, 0.0382, ..., -0.0160, -0.0068, -0.0069]], + device='cuda:0'), grad: tensor([[-0.0199, 0.0025, 0.0044, ..., -0.0127, 0.0008, -0.0041], + [ 0.0021, 0.0047, 0.0033, ..., 0.0062, 0.0048, 0.0042], + [ 0.0025, 0.0028, -0.0023, ..., 0.0039, -0.0104, -0.0081], + ..., + [-0.0134, -0.0026, 0.0036, ..., 0.0034, -0.0044, -0.0009], + [ 0.0066, -0.0020, -0.0437, ..., -0.0039, -0.0154, 0.0045], + [ 0.0032, -0.0118, 0.0244, ..., -0.0140, 0.0055, -0.0134]], + device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0227, -0.0232, -0.0015, 0.0073, -0.0262, -0.0102, 0.0115, 0.0060, + 0.0278, -0.0290], device='cuda:0'), grad: tensor([-0.0350, 0.0290, -0.0075, 0.0243, 0.0131, 0.0216, 0.0181, 0.0032, + -0.0409, -0.0258], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 227.58, cls_loss 0.8763 cls_loss_mapping 0.0929 cls_loss_causal 0.8471 re_mapping 0.0260 re_causal 0.0787 /// teacc 97.67 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0379, -0.0215, -0.0481, ..., 0.0608, -0.0024, 0.0255], + [-0.0383, 0.0460, -0.0533, ..., 0.0098, -0.0245, -0.0390], + [-0.0192, 0.0275, -0.0466, ..., -0.0340, 0.0011, 0.0391], + ..., + [-0.0062, 0.0129, -0.0172, ..., 0.0131, 0.0196, 0.0035], + [ 0.0413, -0.0334, 0.0108, ..., 0.0031, 0.0051, -0.0303], + [-0.0161, -0.0391, 0.0382, ..., -0.0153, -0.0063, -0.0075]], + device='cuda:0'), grad: tensor([[ 8.4352e-04, 4.3068e-03, 2.0943e-03, ..., 3.4523e-03, + 1.8425e-03, 3.8662e-03], + [-3.0766e-03, -4.5891e-03, 1.0719e-03, ..., -4.3526e-03, + 1.1406e-03, 6.9618e-05], + [-3.6793e-03, -1.9054e-03, 4.5776e-03, ..., 4.8065e-03, + 3.8795e-03, 1.0986e-03], + ..., + [ 2.7847e-03, 4.7417e-03, 4.8676e-03, ..., 2.9354e-03, + 5.6343e-03, 5.0507e-03], + [ 4.2915e-03, 9.6817e-03, 2.6703e-03, ..., 4.8370e-03, + 4.0398e-03, 7.3586e-03], + [ 1.1091e-03, 3.5057e-03, 6.9160e-03, ..., 2.2545e-03, + 3.7479e-03, 3.5019e-03]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0226, -0.0233, -0.0019, 0.0074, -0.0258, -0.0101, 0.0116, 0.0058, + 0.0283, -0.0291], device='cuda:0'), grad: tensor([ 0.0230, -0.0057, 0.0280, -0.0516, 0.0106, 0.0009, -0.0953, 0.0303, + 0.0317, 0.0280], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 228.95, cls_loss 0.8659 cls_loss_mapping 0.0812 cls_loss_causal 0.8305 re_mapping 0.0265 re_causal 0.0769 /// teacc 97.71 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0380, -0.0212, -0.0498, ..., 0.0612, -0.0028, 0.0261], + [-0.0386, 0.0467, -0.0543, ..., 0.0102, -0.0245, -0.0395], + [-0.0202, 0.0275, -0.0471, ..., -0.0338, 0.0012, 0.0403], + ..., + [-0.0059, 0.0134, -0.0173, ..., 0.0136, 0.0194, 0.0027], + [ 0.0419, -0.0331, 0.0116, ..., 0.0029, 0.0045, -0.0299], + [-0.0166, -0.0405, 0.0384, ..., -0.0159, -0.0055, -0.0080]], + device='cuda:0'), grad: tensor([[ 2.2507e-03, -4.1199e-03, -2.5768e-03, ..., -6.0387e-03, + -2.4662e-03, -7.6637e-03], + [-1.3809e-03, -1.1078e-02, -1.8911e-03, ..., -1.3680e-02, + -3.9220e-05, -4.0665e-03], + [ 7.9346e-03, -4.9210e-03, -1.7185e-03, ..., -3.9330e-03, + 1.1292e-03, -2.7676e-03], + ..., + [-2.5921e-03, 3.2501e-03, -7.2060e-03, ..., 3.3894e-03, + 4.2272e-04, 4.2763e-03], + [-1.2283e-03, 5.8861e-03, 3.8166e-03, ..., 1.2140e-03, + 3.9177e-03, -3.5076e-03], + [ 9.2163e-03, -1.0986e-03, 3.1738e-03, ..., -2.2316e-03, + 4.5280e-03, -4.4403e-03]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0225, -0.0234, -0.0019, 0.0075, -0.0257, -0.0098, 0.0110, 0.0055, + 0.0285, -0.0290], device='cuda:0'), grad: tensor([-0.0236, -0.0307, 0.0010, 0.0062, 0.0306, 0.0283, -0.0353, 0.0073, + 0.0068, 0.0093], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 230.47, cls_loss 0.8307 cls_loss_mapping 0.0841 cls_loss_causal 0.7975 re_mapping 0.0247 re_causal 0.0710 /// teacc 97.81 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0382, -0.0218, -0.0515, ..., 0.0618, -0.0028, 0.0266], + [-0.0395, 0.0472, -0.0544, ..., 0.0112, -0.0246, -0.0393], + [-0.0208, 0.0295, -0.0480, ..., -0.0342, 0.0012, 0.0406], + ..., + [-0.0053, 0.0123, -0.0176, ..., 0.0130, 0.0192, 0.0026], + [ 0.0421, -0.0346, 0.0118, ..., 0.0025, 0.0045, -0.0310], + [-0.0166, -0.0393, 0.0391, ..., -0.0157, -0.0054, -0.0076]], + device='cuda:0'), grad: tensor([[ 1.1005e-03, -2.9526e-03, -1.7385e-03, ..., -2.2869e-03, + 4.7946e-04, -3.7746e-03], + [ 1.7390e-05, -5.1832e-04, -2.5773e-04, ..., -1.1091e-03, + -1.7920e-03, 2.7924e-03], + [ 1.0414e-03, -7.9956e-03, -5.5313e-03, ..., -8.4839e-03, + -1.2140e-03, -5.1384e-03], + ..., + [ 7.2384e-04, 1.0719e-03, 6.6490e-03, ..., 3.2482e-03, + 4.1618e-03, 2.1362e-03], + [-1.0192e-05, 3.6850e-03, 3.8872e-03, ..., 4.2953e-03, + 2.2984e-03, 2.4529e-03], + [ 1.1545e-04, 4.4060e-03, -1.3916e-02, ..., 5.4474e-03, + -1.0445e-02, 3.7994e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0223, -0.0230, -0.0021, 0.0078, -0.0257, -0.0098, 0.0109, 0.0053, + 0.0282, -0.0286], device='cuda:0'), grad: tensor([-0.0003, -0.0126, -0.0307, -0.0155, 0.0279, -0.0653, 0.0466, 0.0238, + 0.0216, 0.0046], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 227.39, cls_loss 0.8332 cls_loss_mapping 0.0831 cls_loss_causal 0.8040 re_mapping 0.0242 re_causal 0.0713 /// teacc 98.05 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0381, -0.0219, -0.0522, ..., 0.0622, -0.0031, 0.0269], + [-0.0393, 0.0478, -0.0543, ..., 0.0112, -0.0247, -0.0400], + [-0.0201, 0.0299, -0.0489, ..., -0.0344, 0.0014, 0.0412], + ..., + [-0.0048, 0.0127, -0.0175, ..., 0.0143, 0.0194, 0.0039], + [ 0.0419, -0.0353, 0.0128, ..., 0.0030, 0.0047, -0.0315], + [-0.0167, -0.0399, 0.0395, ..., -0.0166, -0.0052, -0.0083]], + device='cuda:0'), grad: tensor([[ 0.0042, 0.0069, 0.0064, ..., 0.0117, 0.0041, 0.0077], + [-0.0053, -0.0043, -0.0021, ..., -0.0121, -0.0003, -0.0069], + [ 0.0019, -0.0083, -0.0175, ..., -0.0041, -0.0113, -0.0016], + ..., + [ 0.0010, -0.0014, 0.0025, ..., -0.0009, 0.0037, 0.0007], + [-0.0058, 0.0013, 0.0054, ..., 0.0047, 0.0032, -0.0004], + [ 0.0009, -0.0067, 0.0164, ..., -0.0116, 0.0101, -0.0087]], + device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0223, -0.0230, -0.0025, 0.0080, -0.0260, -0.0097, 0.0109, 0.0054, + 0.0281, -0.0284], device='cuda:0'), grad: tensor([ 0.0471, -0.0551, -0.0335, -0.0011, 0.0135, 0.0018, 0.0187, 0.0145, + 0.0065, -0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 226.36, cls_loss 0.8227 cls_loss_mapping 0.0755 cls_loss_causal 0.7833 re_mapping 0.0242 re_causal 0.0717 /// teacc 97.79 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0387, -0.0226, -0.0536, ..., 0.0621, -0.0032, 0.0269], + [-0.0395, 0.0483, -0.0543, ..., 0.0111, -0.0257, -0.0407], + [-0.0204, 0.0304, -0.0490, ..., -0.0343, 0.0014, 0.0419], + ..., + [-0.0046, 0.0131, -0.0183, ..., 0.0137, 0.0193, 0.0037], + [ 0.0424, -0.0356, 0.0138, ..., 0.0032, 0.0046, -0.0318], + [-0.0179, -0.0401, 0.0395, ..., -0.0169, -0.0050, -0.0081]], + device='cuda:0'), grad: tensor([[-2.9392e-03, 1.4515e-03, 7.5340e-04, ..., 1.8444e-03, + -1.1492e-03, 1.1702e-03], + [ 2.0752e-03, 3.8862e-04, -2.4567e-03, ..., -7.6485e-04, + 2.3327e-03, -1.8225e-03], + [ 1.8034e-03, 5.8441e-03, 1.7166e-03, ..., 1.6289e-03, + 8.9264e-03, -1.1711e-03], + ..., + [ 1.4770e-04, 2.0027e-03, 3.1776e-03, ..., 2.3804e-03, + -7.0286e-04, 1.6603e-03], + [ 2.7790e-03, -2.3117e-03, 1.0853e-03, ..., -7.3493e-05, + -3.7575e-03, 1.5841e-03], + [-5.8126e-04, 2.7180e-03, -7.2861e-04, ..., 1.5249e-03, + 7.2575e-04, 1.0700e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0225, -0.0229, -0.0028, 0.0085, -0.0259, -0.0097, 0.0105, 0.0053, + 0.0279, -0.0283], device='cuda:0'), grad: tensor([-0.0098, 0.0087, 0.0304, -0.0080, 0.0038, -0.0563, 0.0280, 0.0137, + -0.0206, 0.0101], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 226.94, cls_loss 0.8057 cls_loss_mapping 0.0818 cls_loss_causal 0.7743 re_mapping 0.0224 re_causal 0.0634 /// teacc 98.02 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0379, -0.0222, -0.0538, ..., 0.0626, -0.0026, 0.0271], + [-0.0403, 0.0489, -0.0554, ..., 0.0109, -0.0266, -0.0415], + [-0.0215, 0.0307, -0.0494, ..., -0.0345, 0.0019, 0.0434], + ..., + [-0.0048, 0.0137, -0.0175, ..., 0.0136, 0.0195, 0.0027], + [ 0.0433, -0.0354, 0.0131, ..., 0.0043, 0.0045, -0.0308], + [-0.0174, -0.0411, 0.0402, ..., -0.0182, -0.0037, -0.0086]], + device='cuda:0'), grad: tensor([[ 0.0040, -0.0005, 0.0006, ..., -0.0015, 0.0045, 0.0057], + [-0.0116, 0.0012, -0.0001, ..., -0.0036, 0.0032, 0.0015], + [-0.0085, -0.0151, -0.0102, ..., -0.0081, -0.0193, -0.0174], + ..., + [-0.0044, 0.0037, 0.0036, ..., 0.0021, 0.0064, 0.0039], + [ 0.0033, 0.0086, 0.0094, ..., 0.0028, 0.0071, 0.0027], + [-0.0007, 0.0036, 0.0017, ..., 0.0035, 0.0004, 0.0020]], + device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0225, -0.0233, -0.0030, 0.0080, -0.0253, -0.0095, 0.0106, 0.0050, + 0.0285, -0.0283], device='cuda:0'), grad: tensor([ 0.0032, -0.0201, -0.0983, -0.0107, 0.0059, 0.0414, 0.0163, 0.0179, + 0.0479, -0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 226.82, cls_loss 0.7810 cls_loss_mapping 0.0646 cls_loss_causal 0.7430 re_mapping 0.0227 re_causal 0.0666 /// teacc 97.60 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0380, -0.0227, -0.0548, ..., 0.0628, -0.0023, 0.0276], + [-0.0398, 0.0488, -0.0562, ..., 0.0109, -0.0275, -0.0430], + [-0.0221, 0.0311, -0.0495, ..., -0.0339, 0.0018, 0.0444], + ..., + [-0.0054, 0.0141, -0.0171, ..., 0.0140, 0.0196, 0.0023], + [ 0.0430, -0.0348, 0.0142, ..., 0.0047, 0.0045, -0.0312], + [-0.0173, -0.0409, 0.0409, ..., -0.0183, -0.0029, -0.0077]], + device='cuda:0'), grad: tensor([[ 0.0016, -0.0028, -0.0038, ..., -0.0012, 0.0022, -0.0013], + [ 0.0035, -0.0011, -0.0047, ..., -0.0032, -0.0037, -0.0070], + [ 0.0016, 0.0021, 0.0042, ..., 0.0041, 0.0074, 0.0059], + ..., + [ 0.0222, 0.0133, 0.0325, ..., 0.0026, 0.0183, 0.0041], + [ 0.0010, 0.0034, 0.0109, ..., -0.0072, 0.0032, -0.0021], + [-0.0222, -0.0108, -0.0388, ..., 0.0025, -0.0169, 0.0044]], + device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0222, -0.0242, -0.0021, 0.0078, -0.0258, -0.0093, 0.0109, 0.0053, + 0.0283, -0.0277], device='cuda:0'), grad: tensor([-0.0006, -0.0324, 0.0354, -0.0411, -0.0292, 0.0188, -0.0086, 0.0636, + 0.0182, -0.0242], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 226.88, cls_loss 0.7984 cls_loss_mapping 0.0673 cls_loss_causal 0.7585 re_mapping 0.0232 re_causal 0.0668 /// teacc 97.91 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0384, -0.0230, -0.0559, ..., 0.0639, -0.0034, 0.0280], + [-0.0404, 0.0493, -0.0563, ..., 0.0114, -0.0280, -0.0439], + [-0.0230, 0.0314, -0.0500, ..., -0.0341, 0.0021, 0.0452], + ..., + [-0.0048, 0.0143, -0.0165, ..., 0.0134, 0.0200, 0.0020], + [ 0.0439, -0.0359, 0.0139, ..., 0.0039, 0.0046, -0.0310], + [-0.0174, -0.0408, 0.0416, ..., -0.0188, -0.0028, -0.0091]], + device='cuda:0'), grad: tensor([[-0.0042, 0.0009, -0.0103, ..., -0.0028, -0.0089, -0.0021], + [ 0.0007, 0.0097, -0.0030, ..., 0.0116, 0.0002, -0.0027], + [ 0.0056, -0.0145, -0.0013, ..., -0.0254, 0.0036, -0.0010], + ..., + [-0.0030, -0.0059, -0.0060, ..., 0.0037, -0.0303, 0.0004], + [-0.0130, -0.0002, -0.0056, ..., 0.0034, -0.0004, 0.0016], + [ 0.0128, 0.0029, 0.0065, ..., 0.0033, 0.0112, 0.0048]], + device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0218, -0.0236, -0.0028, 0.0083, -0.0256, -0.0090, 0.0111, 0.0050, + 0.0283, -0.0281], device='cuda:0'), grad: tensor([-0.0214, 0.0140, -0.0392, 0.0526, 0.0266, 0.0119, -0.0190, -0.0370, + -0.0251, 0.0367], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 226.04, cls_loss 0.8240 cls_loss_mapping 0.0739 cls_loss_causal 0.7901 re_mapping 0.0222 re_causal 0.0622 /// teacc 97.81 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0380, -0.0238, -0.0554, ..., 0.0643, -0.0039, 0.0278], + [-0.0408, 0.0507, -0.0561, ..., 0.0120, -0.0283, -0.0443], + [-0.0230, 0.0325, -0.0502, ..., -0.0339, 0.0025, 0.0462], + ..., + [-0.0058, 0.0143, -0.0168, ..., 0.0136, 0.0198, 0.0019], + [ 0.0444, -0.0367, 0.0130, ..., 0.0040, 0.0041, -0.0314], + [-0.0171, -0.0420, 0.0419, ..., -0.0199, -0.0023, -0.0108]], + device='cuda:0'), grad: tensor([[ 3.9053e-04, -7.0572e-05, -5.8174e-04, ..., -8.4839e-03, + -1.8702e-03, -1.2184e-02], + [ 2.5463e-03, 7.1287e-04, 3.3398e-03, ..., 5.6496e-03, + 9.4080e-04, 6.6185e-03], + [ 2.4738e-03, -2.5597e-03, -8.1940e-03, ..., -7.6141e-03, + 1.0672e-03, -8.9874e-03], + ..., + [ 2.7790e-03, 9.4795e-04, 3.8166e-03, ..., 3.8891e-03, + 1.5221e-03, 4.5815e-03], + [ 3.3092e-03, 9.5308e-05, 2.6455e-03, ..., 3.8280e-03, + 1.3924e-03, 6.5386e-05], + [-1.8402e-02, 9.1648e-04, -8.3618e-03, ..., 4.7760e-03, + -5.1155e-03, 5.9319e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0219, -0.0235, -0.0022, 0.0080, -0.0255, -0.0096, 0.0114, 0.0050, + 0.0281, -0.0283], device='cuda:0'), grad: tensor([-0.0657, 0.0356, -0.0192, -0.0240, -0.0347, -0.0024, 0.0679, 0.0266, + 0.0323, -0.0164], device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 226.04, cls_loss 0.8109 cls_loss_mapping 0.0719 cls_loss_causal 0.7718 re_mapping 0.0223 re_causal 0.0625 /// teacc 98.00 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0390, -0.0240, -0.0553, ..., 0.0647, -0.0036, 0.0282], + [-0.0410, 0.0505, -0.0574, ..., 0.0111, -0.0290, -0.0446], + [-0.0236, 0.0325, -0.0514, ..., -0.0345, 0.0024, 0.0467], + ..., + [-0.0055, 0.0154, -0.0158, ..., 0.0140, 0.0204, 0.0017], + [ 0.0446, -0.0384, 0.0134, ..., 0.0033, 0.0041, -0.0318], + [-0.0174, -0.0421, 0.0424, ..., -0.0195, -0.0023, -0.0098]], + device='cuda:0'), grad: tensor([[-0.0037, 0.0010, 0.0005, ..., -0.0062, 0.0006, -0.0058], + [ 0.0007, -0.0019, -0.0097, ..., -0.0036, -0.0051, -0.0035], + [-0.0005, -0.0062, -0.0073, ..., -0.0066, -0.0063, -0.0103], + ..., + [ 0.0004, 0.0024, 0.0013, ..., 0.0044, 0.0006, 0.0044], + [ 0.0005, 0.0030, 0.0053, ..., 0.0027, 0.0030, 0.0022], + [ 0.0008, 0.0028, -0.0060, ..., 0.0047, -0.0046, 0.0039]], + device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0221, -0.0235, -0.0030, 0.0083, -0.0254, -0.0094, 0.0113, 0.0050, + 0.0278, -0.0280], device='cuda:0'), grad: tensor([-0.0179, -0.0585, -0.0434, 0.0353, 0.0244, 0.0024, -0.0214, 0.0121, + 0.0401, 0.0270], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 226.76, cls_loss 0.7833 cls_loss_mapping 0.0589 cls_loss_causal 0.7535 re_mapping 0.0220 re_causal 0.0634 /// teacc 97.88 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0382, -0.0245, -0.0560, ..., 0.0651, -0.0036, 0.0283], + [-0.0411, 0.0508, -0.0573, ..., 0.0109, -0.0290, -0.0452], + [-0.0236, 0.0332, -0.0519, ..., -0.0336, 0.0022, 0.0479], + ..., + [-0.0058, 0.0151, -0.0155, ..., 0.0136, 0.0208, 0.0010], + [ 0.0449, -0.0392, 0.0142, ..., 0.0033, 0.0047, -0.0315], + [-0.0174, -0.0426, 0.0421, ..., -0.0202, -0.0028, -0.0094]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0004, 0.0026, ..., 0.0041, 0.0006, 0.0046], + [ 0.0002, -0.0138, 0.0031, ..., 0.0026, -0.0078, -0.0010], + [ 0.0002, 0.0080, -0.0040, ..., -0.0059, 0.0058, -0.0041], + ..., + [-0.0135, 0.0012, -0.0126, ..., -0.0010, -0.0029, 0.0021], + [-0.0011, 0.0010, -0.0026, ..., -0.0086, -0.0004, -0.0078], + [ 0.0149, 0.0012, 0.0236, ..., 0.0008, 0.0037, -0.0010]], + device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0222, -0.0230, -0.0027, 0.0080, -0.0251, -0.0088, 0.0106, 0.0046, + 0.0278, -0.0283], device='cuda:0'), grad: tensor([ 0.0202, -0.0417, 0.0035, 0.0444, -0.0107, 0.0476, -0.0493, 0.0015, + -0.0167, 0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 227.13, cls_loss 0.7900 cls_loss_mapping 0.0649 cls_loss_causal 0.7510 re_mapping 0.0218 re_causal 0.0623 /// teacc 98.23 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0382, -0.0245, -0.0572, ..., 0.0656, -0.0047, 0.0276], + [-0.0405, 0.0507, -0.0573, ..., 0.0114, -0.0284, -0.0461], + [-0.0242, 0.0344, -0.0524, ..., -0.0338, 0.0026, 0.0482], + ..., + [-0.0063, 0.0147, -0.0160, ..., 0.0137, 0.0200, 0.0004], + [ 0.0453, -0.0395, 0.0146, ..., 0.0034, 0.0049, -0.0307], + [-0.0173, -0.0425, 0.0429, ..., -0.0202, -0.0022, -0.0102]], + device='cuda:0'), grad: tensor([[ 0.0012, 0.0018, 0.0020, ..., 0.0038, 0.0002, 0.0018], + [ 0.0012, 0.0027, 0.0039, ..., 0.0032, 0.0002, 0.0043], + [ 0.0012, 0.0005, 0.0042, ..., 0.0048, 0.0025, 0.0047], + ..., + [ 0.0019, 0.0019, 0.0036, ..., 0.0040, -0.0019, 0.0021], + [ 0.0130, 0.0023, 0.0086, ..., 0.0160, 0.0036, 0.0019], + [-0.0015, -0.0008, -0.0077, ..., -0.0004, -0.0022, 0.0013]], + device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0217, -0.0225, -0.0030, 0.0083, -0.0249, -0.0091, 0.0110, 0.0039, + 0.0281, -0.0281], device='cuda:0'), grad: tensor([ 0.0162, 0.0253, 0.0300, -0.0704, -0.0203, -0.0443, -0.0002, 0.0212, + 0.0473, -0.0048], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 226.40, cls_loss 0.7669 cls_loss_mapping 0.0632 cls_loss_causal 0.7349 re_mapping 0.0218 re_causal 0.0613 /// teacc 98.14 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0377, -0.0245, -0.0570, ..., 0.0663, -0.0047, 0.0279], + [-0.0415, 0.0505, -0.0574, ..., 0.0118, -0.0284, -0.0467], + [-0.0247, 0.0351, -0.0536, ..., -0.0342, 0.0024, 0.0482], + ..., + [-0.0048, 0.0147, -0.0158, ..., 0.0132, 0.0204, 0.0005], + [ 0.0448, -0.0399, 0.0152, ..., 0.0038, 0.0041, -0.0312], + [-0.0170, -0.0419, 0.0431, ..., -0.0208, -0.0016, -0.0105]], + device='cuda:0'), grad: tensor([[ 1.4496e-03, -7.9012e-04, -8.2397e-04, ..., -7.4387e-03, + 1.2541e-03, -6.7368e-03], + [ 5.4359e-04, -6.9618e-04, 4.4212e-03, ..., -4.2572e-03, + -4.0948e-05, 4.9496e-04], + [ 6.3324e-04, 1.5421e-03, 3.0727e-03, ..., 4.1389e-03, + 1.3523e-03, -9.8896e-04], + ..., + [ 1.2589e-02, -7.5264e-03, 2.8687e-03, ..., -1.5556e-02, + 5.7755e-03, -8.3694e-03], + [ 1.0214e-03, 2.7771e-03, 5.7487e-03, ..., 9.8114e-03, + 1.9989e-03, 4.2915e-03], + [-2.0081e-02, -2.1992e-03, -2.1301e-02, ..., -7.0229e-03, + -1.4381e-02, 1.3762e-03]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0217, -0.0225, -0.0032, 0.0086, -0.0253, -0.0090, 0.0114, 0.0043, + 0.0276, -0.0283], device='cuda:0'), grad: tensor([-0.0202, -0.0358, 0.0177, 0.0246, 0.0176, 0.0406, 0.0004, -0.0201, + 0.0303, -0.0551], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 226.36, cls_loss 0.7648 cls_loss_mapping 0.0516 cls_loss_causal 0.7234 re_mapping 0.0210 re_causal 0.0606 /// teacc 97.92 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0384, -0.0244, -0.0582, ..., 0.0669, -0.0048, 0.0272], + [-0.0422, 0.0505, -0.0572, ..., 0.0117, -0.0287, -0.0475], + [-0.0243, 0.0354, -0.0541, ..., -0.0334, 0.0028, 0.0492], + ..., + [-0.0038, 0.0154, -0.0154, ..., 0.0135, 0.0207, 0.0006], + [ 0.0450, -0.0407, 0.0160, ..., 0.0027, 0.0043, -0.0312], + [-0.0179, -0.0419, 0.0430, ..., -0.0204, -0.0018, -0.0110]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0051, 0.0022, ..., 0.0068, 0.0023, 0.0055], + [-0.0012, -0.0130, -0.0035, ..., -0.0216, -0.0038, -0.0096], + [ 0.0014, 0.0140, 0.0038, ..., 0.0137, 0.0065, 0.0134], + ..., + [ 0.0010, 0.0058, 0.0108, ..., 0.0004, 0.0133, -0.0005], + [ 0.0008, 0.0008, 0.0033, ..., -0.0032, 0.0063, 0.0013], + [ 0.0039, -0.0033, -0.0160, ..., 0.0064, -0.0157, 0.0047]], + device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0207, -0.0225, -0.0027, 0.0078, -0.0245, -0.0088, 0.0107, 0.0044, + 0.0273, -0.0272], device='cuda:0'), grad: tensor([ 0.0286, -0.0492, 0.0652, -0.0328, -0.0676, 0.0224, 0.0173, 0.0113, + -0.0011, 0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 226.30, cls_loss 0.7819 cls_loss_mapping 0.0634 cls_loss_causal 0.7460 re_mapping 0.0205 re_causal 0.0558 /// teacc 98.16 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0384, -0.0236, -0.0581, ..., 0.0677, -0.0050, 0.0274], + [-0.0431, 0.0513, -0.0577, ..., 0.0124, -0.0297, -0.0481], + [-0.0251, 0.0360, -0.0554, ..., -0.0319, 0.0021, 0.0497], + ..., + [-0.0041, 0.0159, -0.0161, ..., 0.0129, 0.0203, 0.0009], + [ 0.0453, -0.0424, 0.0162, ..., 0.0020, 0.0041, -0.0308], + [-0.0181, -0.0422, 0.0437, ..., -0.0199, -0.0008, -0.0115]], + device='cuda:0'), grad: tensor([[-6.2227e-04, 2.6207e-03, -6.1560e-04, ..., 1.7729e-03, + 1.6947e-03, -4.3368e-04], + [-2.1305e-03, 8.2397e-04, -7.1383e-04, ..., -1.1864e-03, + -2.7637e-03, 3.2501e-03], + [ 5.5695e-03, 2.9202e-03, 4.1580e-03, ..., 2.4071e-03, + 3.3417e-03, 3.5458e-03], + ..., + [-2.3010e-02, -1.0475e-02, -1.3947e-02, ..., -6.8741e-03, + -1.6708e-02, -1.3206e-02], + [ 3.8147e-03, -1.0624e-03, -2.2423e-04, ..., -1.9321e-03, + 1.5850e-03, -8.9884e-04], + [ 3.4637e-03, -4.4847e-04, 6.2275e-04, ..., -2.8467e-04, + 2.4433e-03, 3.6299e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0209, -0.0222, -0.0027, 0.0087, -0.0246, -0.0088, 0.0103, 0.0034, + 0.0273, -0.0271], device='cuda:0'), grad: tensor([-0.0018, 0.0225, 0.0235, 0.0357, 0.0541, 0.0103, -0.0367, -0.0941, + -0.0249, 0.0115], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 225.85, cls_loss 0.7737 cls_loss_mapping 0.0488 cls_loss_causal 0.7405 re_mapping 0.0205 re_causal 0.0591 /// teacc 98.19 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0396, -0.0239, -0.0605, ..., 0.0684, -0.0063, 0.0278], + [-0.0430, 0.0513, -0.0589, ..., 0.0126, -0.0298, -0.0478], + [-0.0243, 0.0362, -0.0564, ..., -0.0317, 0.0024, 0.0503], + ..., + [-0.0041, 0.0165, -0.0153, ..., 0.0123, 0.0211, 0.0002], + [ 0.0455, -0.0420, 0.0167, ..., 0.0022, 0.0046, -0.0306], + [-0.0177, -0.0414, 0.0439, ..., -0.0200, -0.0006, -0.0113]], + device='cuda:0'), grad: tensor([[ 0.0050, 0.0036, 0.0081, ..., 0.0039, 0.0087, 0.0048], + [-0.0132, -0.0116, -0.0196, ..., -0.0124, -0.0076, -0.0139], + [ 0.0015, -0.0005, 0.0052, ..., 0.0010, 0.0013, -0.0003], + ..., + [ 0.0011, 0.0081, 0.0154, ..., -0.0021, 0.0192, -0.0013], + [ 0.0076, 0.0017, 0.0028, ..., 0.0035, 0.0040, 0.0034], + [ 0.0045, -0.0088, -0.0188, ..., 0.0031, -0.0207, 0.0031]], + device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0202, -0.0228, -0.0027, 0.0089, -0.0248, -0.0093, 0.0106, 0.0041, + 0.0276, -0.0267], device='cuda:0'), grad: tensor([ 0.0418, -0.1187, 0.0183, 0.0327, 0.0097, -0.0139, -0.0080, 0.0218, + 0.0343, -0.0180], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 227.73, cls_loss 0.7575 cls_loss_mapping 0.0559 cls_loss_causal 0.7155 re_mapping 0.0199 re_causal 0.0574 /// teacc 98.40 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0402, -0.0244, -0.0611, ..., 0.0688, -0.0056, 0.0276], + [-0.0429, 0.0519, -0.0588, ..., 0.0127, -0.0300, -0.0488], + [-0.0241, 0.0376, -0.0573, ..., -0.0315, 0.0025, 0.0516], + ..., + [-0.0034, 0.0169, -0.0154, ..., 0.0118, 0.0210, 0.0002], + [ 0.0449, -0.0423, 0.0169, ..., 0.0027, 0.0044, -0.0308], + [-0.0179, -0.0422, 0.0440, ..., -0.0198, -0.0009, -0.0122]], + device='cuda:0'), grad: tensor([[-0.0019, -0.0003, -0.0057, ..., -0.0078, -0.0071, -0.0044], + [-0.0003, -0.0035, 0.0033, ..., -0.0079, 0.0014, -0.0004], + [ 0.0026, 0.0031, -0.0097, ..., -0.0040, 0.0017, -0.0030], + ..., + [-0.0016, -0.0030, -0.0059, ..., -0.0060, 0.0042, -0.0058], + [ 0.0039, 0.0035, 0.0070, ..., 0.0083, 0.0037, 0.0043], + [-0.0011, 0.0005, 0.0307, ..., 0.0044, 0.0047, 0.0028]], + device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0199, -0.0225, -0.0030, 0.0090, -0.0248, -0.0091, 0.0107, 0.0039, + 0.0278, -0.0268], device='cuda:0'), grad: tensor([-0.0260, -0.0106, -0.0302, -0.0027, -0.0093, 0.0550, -0.0167, -0.0019, + 0.0450, -0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 226.23, cls_loss 0.7317 cls_loss_mapping 0.0519 cls_loss_causal 0.6881 re_mapping 0.0193 re_causal 0.0555 /// teacc 98.31 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0403, -0.0241, -0.0618, ..., 0.0682, -0.0052, 0.0286], + [-0.0423, 0.0521, -0.0598, ..., 0.0123, -0.0308, -0.0496], + [-0.0239, 0.0377, -0.0568, ..., -0.0308, 0.0026, 0.0517], + ..., + [-0.0034, 0.0176, -0.0150, ..., 0.0124, 0.0207, 0.0004], + [ 0.0449, -0.0424, 0.0165, ..., 0.0024, 0.0044, -0.0305], + [-0.0186, -0.0430, 0.0439, ..., -0.0198, -0.0003, -0.0126]], + device='cuda:0'), grad: tensor([[-0.0002, -0.0044, -0.0012, ..., -0.0026, -0.0088, -0.0060], + [ 0.0005, 0.0002, -0.0005, ..., -0.0011, 0.0017, 0.0005], + [ 0.0006, -0.0011, -0.0052, ..., -0.0006, 0.0031, 0.0035], + ..., + [-0.0009, -0.0091, -0.0027, ..., -0.0042, -0.0064, -0.0054], + [ 0.0007, 0.0049, 0.0032, ..., 0.0054, 0.0039, -0.0002], + [ 0.0006, 0.0046, 0.0034, ..., 0.0050, 0.0040, 0.0046]], + device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0201, -0.0229, -0.0027, 0.0089, -0.0251, -0.0085, 0.0107, 0.0044, + 0.0277, -0.0275], device='cuda:0'), grad: tensor([-0.0124, -0.0061, 0.0095, -0.0003, 0.0218, 0.0242, -0.0319, -0.0308, + -0.0018, 0.0278], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 226.87, cls_loss 0.7270 cls_loss_mapping 0.0505 cls_loss_causal 0.6884 re_mapping 0.0195 re_causal 0.0539 /// teacc 98.34 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0406, -0.0240, -0.0629, ..., 0.0701, -0.0059, 0.0291], + [-0.0429, 0.0537, -0.0591, ..., 0.0134, -0.0299, -0.0489], + [-0.0233, 0.0386, -0.0561, ..., -0.0313, 0.0026, 0.0525], + ..., + [-0.0032, 0.0177, -0.0150, ..., 0.0113, 0.0207, -0.0011], + [ 0.0450, -0.0432, 0.0159, ..., 0.0021, 0.0038, -0.0310], + [-0.0182, -0.0437, 0.0449, ..., -0.0199, 0.0007, -0.0126]], + device='cuda:0'), grad: tensor([[ 0.0011, 0.0030, 0.0037, ..., 0.0034, 0.0011, 0.0017], + [-0.0007, -0.0155, 0.0050, ..., -0.0052, -0.0040, -0.0123], + [-0.0047, 0.0038, -0.0124, ..., -0.0004, 0.0036, 0.0055], + ..., + [ 0.0004, 0.0010, 0.0009, ..., -0.0040, 0.0007, 0.0002], + [ 0.0016, 0.0042, 0.0007, ..., 0.0037, -0.0025, 0.0032], + [-0.0047, -0.0125, -0.0122, ..., -0.0106, -0.0053, -0.0060]], + device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0204, -0.0230, -0.0024, 0.0089, -0.0251, -0.0083, 0.0106, 0.0035, + 0.0275, -0.0270], device='cuda:0'), grad: tensor([ 0.0196, -0.0506, -0.0382, 0.0220, 0.0422, -0.0086, 0.0396, 0.0001, + 0.0245, -0.0505], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 226.43, cls_loss 0.7390 cls_loss_mapping 0.0519 cls_loss_causal 0.6918 re_mapping 0.0194 re_causal 0.0527 /// teacc 98.03 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0402, -0.0239, -0.0638, ..., 0.0705, -0.0061, 0.0296], + [-0.0430, 0.0537, -0.0595, ..., 0.0137, -0.0313, -0.0492], + [-0.0234, 0.0389, -0.0558, ..., -0.0314, 0.0027, 0.0533], + ..., + [-0.0039, 0.0182, -0.0155, ..., 0.0123, 0.0200, -0.0019], + [ 0.0457, -0.0434, 0.0160, ..., 0.0020, 0.0042, -0.0305], + [-0.0177, -0.0446, 0.0454, ..., -0.0199, 0.0011, -0.0128]], + device='cuda:0'), grad: tensor([[ 0.0003, -0.0042, 0.0027, ..., -0.0042, 0.0029, -0.0001], + [ 0.0027, 0.0006, 0.0023, ..., -0.0003, 0.0031, 0.0022], + [ 0.0038, 0.0068, 0.0061, ..., 0.0073, 0.0110, 0.0099], + ..., + [-0.0016, -0.0074, -0.0039, ..., 0.0014, -0.0028, 0.0036], + [ 0.0024, 0.0016, 0.0029, ..., 0.0066, -0.0001, -0.0023], + [ 0.0034, 0.0033, 0.0025, ..., -0.0008, 0.0080, 0.0036]], + device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0208, -0.0231, -0.0027, 0.0091, -0.0254, -0.0089, 0.0108, 0.0039, + 0.0275, -0.0269], device='cuda:0'), grad: tensor([-0.0201, 0.0228, 0.0542, -0.0117, -0.0234, -0.0286, 0.0047, -0.0226, + 0.0116, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 226.41, cls_loss 0.7181 cls_loss_mapping 0.0475 cls_loss_causal 0.6832 re_mapping 0.0195 re_causal 0.0525 /// teacc 98.14 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0412, -0.0236, -0.0651, ..., 0.0708, -0.0068, 0.0297], + [-0.0436, 0.0539, -0.0595, ..., 0.0133, -0.0312, -0.0501], + [-0.0242, 0.0386, -0.0567, ..., -0.0319, 0.0025, 0.0542], + ..., + [-0.0038, 0.0197, -0.0157, ..., 0.0122, 0.0199, -0.0029], + [ 0.0467, -0.0441, 0.0183, ..., 0.0015, 0.0051, -0.0303], + [-0.0178, -0.0450, 0.0459, ..., -0.0204, 0.0015, -0.0125]], + device='cuda:0'), grad: tensor([[ 0.0026, 0.0048, 0.0030, ..., 0.0029, 0.0067, 0.0097], + [-0.0008, -0.0139, 0.0003, ..., -0.0024, -0.0059, -0.0081], + [-0.0003, 0.0026, 0.0046, ..., 0.0026, 0.0025, -0.0020], + ..., + [-0.0007, -0.0009, -0.0020, ..., 0.0023, -0.0027, 0.0011], + [-0.0011, -0.0006, -0.0083, ..., -0.0030, -0.0060, -0.0037], + [ 0.0023, 0.0073, 0.0044, ..., 0.0030, 0.0077, 0.0086]], + device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0207, -0.0231, -0.0033, 0.0094, -0.0252, -0.0085, 0.0105, 0.0038, + 0.0276, -0.0268], device='cuda:0'), grad: tensor([ 0.0490, -0.0558, 0.0203, -0.0062, 0.0104, -0.0534, 0.0032, 0.0206, + -0.0237, 0.0356], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 226.56, cls_loss 0.7253 cls_loss_mapping 0.0567 cls_loss_causal 0.6867 re_mapping 0.0187 re_causal 0.0511 /// teacc 98.24 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0416, -0.0246, -0.0655, ..., 0.0708, -0.0074, 0.0301], + [-0.0428, 0.0539, -0.0599, ..., 0.0131, -0.0313, -0.0512], + [-0.0245, 0.0400, -0.0580, ..., -0.0317, 0.0019, 0.0554], + ..., + [-0.0049, 0.0200, -0.0148, ..., 0.0126, 0.0200, -0.0039], + [ 0.0465, -0.0447, 0.0181, ..., 0.0020, 0.0053, -0.0308], + [-0.0177, -0.0454, 0.0462, ..., -0.0210, 0.0013, -0.0126]], + device='cuda:0'), grad: tensor([[ 2.9755e-03, 4.2610e-03, 1.6031e-03, ..., 9.6512e-03, + 3.7441e-03, 1.4221e-02], + [-2.7790e-03, -8.9722e-03, -4.1542e-03, ..., -1.2161e-02, + -7.3586e-03, -1.7410e-02], + [ 1.4842e-04, 6.6614e-04, 1.9407e-03, ..., 1.0774e-05, + 5.2719e-03, -1.3628e-03], + ..., + [ 2.0523e-03, 5.4359e-04, 9.2850e-03, ..., 2.5902e-03, + 7.0343e-03, 4.3983e-03], + [ 8.0948e-03, 1.1015e-03, -6.9475e-04, ..., 7.9880e-03, + 4.3488e-03, 8.0013e-04], + [-1.0061e-03, -4.3678e-04, -1.1702e-03, ..., -1.2779e-03, + 1.1148e-03, -2.6989e-03]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0203, -0.0229, -0.0036, 0.0096, -0.0253, -0.0084, 0.0109, 0.0038, + 0.0276, -0.0270], device='cuda:0'), grad: tensor([ 0.0470, -0.0646, 0.0125, 0.0083, 0.0084, -0.0055, -0.0314, 0.0268, + 0.0149, -0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 226.57, cls_loss 0.7325 cls_loss_mapping 0.0531 cls_loss_causal 0.6956 re_mapping 0.0184 re_causal 0.0517 /// teacc 98.15 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0421, -0.0247, -0.0647, ..., 0.0719, -0.0076, 0.0303], + [-0.0438, 0.0544, -0.0604, ..., 0.0135, -0.0322, -0.0523], + [-0.0245, 0.0404, -0.0576, ..., -0.0319, 0.0029, 0.0561], + ..., + [-0.0050, 0.0204, -0.0146, ..., 0.0117, 0.0201, -0.0037], + [ 0.0467, -0.0444, 0.0186, ..., 0.0017, 0.0052, -0.0311], + [-0.0178, -0.0453, 0.0460, ..., -0.0211, 0.0018, -0.0129]], + device='cuda:0'), grad: tensor([[-1.6785e-03, 5.2977e-04, 3.4046e-04, ..., -3.3684e-03, + 1.5516e-03, -5.9586e-03], + [ 1.7624e-03, -1.6308e-04, -1.0633e-03, ..., -4.7798e-03, + 1.6565e-03, -9.6703e-04], + [ 9.1493e-05, -5.3024e-03, 1.1606e-03, ..., -3.5954e-03, + -7.5302e-03, -8.8501e-03], + ..., + [-3.0441e-02, -2.0237e-03, -2.6581e-02, ..., -1.4824e-02, + -1.5106e-02, -1.0872e-02], + [ 6.7024e-03, 2.1782e-03, 4.4403e-03, ..., 4.1618e-03, + 3.2482e-03, 6.8779e-03], + [ 4.8645e-02, 1.3676e-03, 5.5481e-02, ..., 2.8458e-03, + 3.8757e-02, 3.2520e-03]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0208, -0.0228, -0.0038, 0.0092, -0.0247, -0.0085, 0.0115, 0.0035, + 0.0276, -0.0276], device='cuda:0'), grad: tensor([-0.0243, -0.0124, -0.0426, 0.0281, 0.0201, 0.0681, -0.0187, -0.0985, + 0.0117, 0.0685], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 227.07, cls_loss 0.6899 cls_loss_mapping 0.0409 cls_loss_causal 0.6543 re_mapping 0.0190 re_causal 0.0503 /// teacc 98.32 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0427, -0.0255, -0.0658, ..., 0.0718, -0.0084, 0.0299], + [-0.0438, 0.0540, -0.0607, ..., 0.0137, -0.0329, -0.0528], + [-0.0244, 0.0411, -0.0569, ..., -0.0307, 0.0031, 0.0569], + ..., + [-0.0037, 0.0202, -0.0146, ..., 0.0110, 0.0202, -0.0047], + [ 0.0461, -0.0452, 0.0196, ..., 0.0017, 0.0059, -0.0308], + [-0.0189, -0.0448, 0.0456, ..., -0.0209, 0.0014, -0.0127]], + device='cuda:0'), grad: tensor([[ 2.5320e-04, 7.4911e-04, 1.2827e-03, ..., 4.3373e-03, + 3.6001e-04, 3.0155e-03], + [ 7.3351e-06, 3.2978e-03, 3.8452e-03, ..., 7.3471e-03, + 1.1435e-03, 3.1357e-03], + [-1.1849e-04, 1.8251e-04, 1.9131e-03, ..., 6.3133e-03, + -6.0129e-04, 2.7485e-03], + ..., + [ 2.5071e-06, 5.7144e-03, 1.2268e-02, ..., 4.3373e-03, + 1.0139e-02, 2.5845e-03], + [ 2.3997e-04, -6.0654e-03, -3.1395e-03, ..., -1.3481e-02, + 2.4738e-03, -3.0556e-03], + [ 1.7524e-05, -8.3847e-03, -1.7033e-03, ..., -6.7558e-03, + 5.7564e-03, -7.3357e-03]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0200, -0.0227, -0.0026, 0.0094, -0.0251, -0.0086, 0.0117, 0.0035, + 0.0275, -0.0280], device='cuda:0'), grad: tensor([ 0.0199, 0.0350, 0.0258, -0.0126, -0.0381, 0.0276, -0.0037, 0.0458, + -0.0385, -0.0612], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 226.59, cls_loss 0.7109 cls_loss_mapping 0.0453 cls_loss_causal 0.6761 re_mapping 0.0182 re_causal 0.0497 /// teacc 98.33 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0421, -0.0259, -0.0676, ..., 0.0718, -0.0092, 0.0293], + [-0.0436, 0.0538, -0.0613, ..., 0.0139, -0.0333, -0.0534], + [-0.0250, 0.0420, -0.0564, ..., -0.0315, 0.0040, 0.0578], + ..., + [-0.0040, 0.0207, -0.0146, ..., 0.0109, 0.0207, -0.0052], + [ 0.0465, -0.0459, 0.0192, ..., 0.0030, 0.0058, -0.0295], + [-0.0192, -0.0452, 0.0462, ..., -0.0219, 0.0024, -0.0125]], + device='cuda:0'), grad: tensor([[-0.0030, 0.0002, -0.0071, ..., -0.0074, 0.0025, 0.0006], + [-0.0015, -0.0020, -0.0024, ..., -0.0053, -0.0003, -0.0048], + [-0.0011, -0.0013, 0.0012, ..., 0.0007, -0.0054, -0.0053], + ..., + [ 0.0052, 0.0009, 0.0039, ..., 0.0020, 0.0041, 0.0018], + [ 0.0032, 0.0011, 0.0048, ..., 0.0046, 0.0026, 0.0036], + [-0.0022, 0.0012, 0.0047, ..., 0.0022, 0.0073, 0.0018]], + device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0200, -0.0232, -0.0031, 0.0091, -0.0251, -0.0083, 0.0118, 0.0040, + 0.0279, -0.0282], device='cuda:0'), grad: tensor([-0.0394, -0.0275, -0.0014, 0.0048, 0.0003, -0.0081, -0.0008, 0.0180, + 0.0401, 0.0141], device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 226.19, cls_loss 0.7191 cls_loss_mapping 0.0435 cls_loss_causal 0.6854 re_mapping 0.0187 re_causal 0.0526 /// teacc 98.28 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0431, -0.0261, -0.0682, ..., 0.0724, -0.0103, 0.0293], + [-0.0441, 0.0545, -0.0624, ..., 0.0149, -0.0337, -0.0531], + [-0.0259, 0.0423, -0.0557, ..., -0.0330, 0.0045, 0.0586], + ..., + [-0.0027, 0.0209, -0.0152, ..., 0.0109, 0.0202, -0.0056], + [ 0.0465, -0.0467, 0.0198, ..., 0.0023, 0.0063, -0.0301], + [-0.0189, -0.0449, 0.0462, ..., -0.0210, 0.0024, -0.0121]], + device='cuda:0'), grad: tensor([[-1.7462e-03, 1.3266e-03, -6.9261e-05, ..., -3.3736e-04, + 1.4639e-03, 2.8725e-03], + [-2.4529e-03, -9.0742e-04, 7.9250e-04, ..., -2.6245e-03, + -1.5945e-03, 1.3542e-03], + [ 1.4896e-03, -1.2672e-04, 2.9545e-03, ..., -1.0473e-04, + 1.8435e-03, -4.8599e-03], + ..., + [-3.0428e-05, -4.7569e-03, -1.1396e-03, ..., -6.4421e-04, + -8.1491e-04, -3.2101e-03], + [ 3.7079e-03, 1.6708e-03, 5.3291e-03, ..., 2.9888e-03, + 4.0588e-03, 3.4771e-03], + [-8.4381e-03, -7.4804e-05, -3.9856e-02, ..., -3.8376e-03, + -1.7059e-02, -3.4428e-03]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0197, -0.0228, -0.0031, 0.0098, -0.0251, -0.0087, 0.0119, 0.0036, + 0.0277, -0.0278], device='cuda:0'), grad: tensor([-0.0093, -0.0048, 0.0103, -0.0180, 0.0437, 0.0145, 0.0193, -0.0130, + 0.0384, -0.0811], device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 226.60, cls_loss 0.7404 cls_loss_mapping 0.0448 cls_loss_causal 0.7016 re_mapping 0.0172 re_causal 0.0496 /// teacc 98.04 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0438, -0.0254, -0.0691, ..., 0.0725, -0.0108, 0.0295], + [-0.0442, 0.0546, -0.0639, ..., 0.0154, -0.0345, -0.0533], + [-0.0276, 0.0427, -0.0559, ..., -0.0331, 0.0047, 0.0588], + ..., + [-0.0029, 0.0215, -0.0142, ..., 0.0095, 0.0213, -0.0052], + [ 0.0478, -0.0473, 0.0187, ..., 0.0017, 0.0062, -0.0310], + [-0.0191, -0.0456, 0.0472, ..., -0.0211, 0.0029, -0.0123]], + device='cuda:0'), grad: tensor([[-0.0165, -0.0089, -0.0202, ..., -0.0036, -0.0033, -0.0017], + [ 0.0032, 0.0005, 0.0033, ..., 0.0013, 0.0004, 0.0010], + [-0.0211, -0.0005, -0.0035, ..., -0.0145, -0.0117, -0.0073], + ..., + [ 0.0028, 0.0028, 0.0036, ..., 0.0034, 0.0026, -0.0002], + [ 0.0151, 0.0021, 0.0078, ..., 0.0062, 0.0007, 0.0013], + [ 0.0030, 0.0011, 0.0049, ..., 0.0034, 0.0017, 0.0012]], + device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0196, -0.0232, -0.0043, 0.0102, -0.0248, -0.0089, 0.0126, 0.0040, + 0.0273, -0.0274], device='cuda:0'), grad: tensor([-0.0963, 0.0105, -0.0734, 0.0444, 0.0363, -0.0265, 0.0252, 0.0035, + 0.0484, 0.0279], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 226.49, cls_loss 0.6946 cls_loss_mapping 0.0458 cls_loss_causal 0.6574 re_mapping 0.0175 re_causal 0.0478 /// teacc 98.12 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0438, -0.0251, -0.0697, ..., 0.0736, -0.0113, 0.0297], + [-0.0453, 0.0559, -0.0639, ..., 0.0157, -0.0344, -0.0529], + [-0.0271, 0.0432, -0.0561, ..., -0.0327, 0.0049, 0.0600], + ..., + [-0.0019, 0.0211, -0.0131, ..., 0.0084, 0.0213, -0.0060], + [ 0.0474, -0.0476, 0.0186, ..., 0.0014, 0.0056, -0.0322], + [-0.0192, -0.0456, 0.0474, ..., -0.0216, 0.0032, -0.0129]], + device='cuda:0'), grad: tensor([[ 0.0006, 0.0022, 0.0025, ..., 0.0033, 0.0012, 0.0025], + [ 0.0012, 0.0028, 0.0031, ..., 0.0026, 0.0026, 0.0031], + [ 0.0048, 0.0039, 0.0019, ..., 0.0043, 0.0018, -0.0012], + ..., + [ 0.0014, -0.0050, 0.0057, ..., -0.0023, 0.0020, 0.0003], + [-0.0123, -0.0060, -0.0052, ..., -0.0012, -0.0086, -0.0031], + [ 0.0026, 0.0019, 0.0190, ..., 0.0028, 0.0062, 0.0021]], + device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0196, -0.0229, -0.0042, 0.0105, -0.0246, -0.0080, 0.0117, 0.0036, + 0.0268, -0.0274], device='cuda:0'), grad: tensor([ 0.0228, 0.0288, 0.0385, -0.0066, -0.0148, -0.0511, 0.0153, -0.0125, + -0.0421, 0.0215], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 227.21, cls_loss 0.7023 cls_loss_mapping 0.0373 cls_loss_causal 0.6725 re_mapping 0.0179 re_causal 0.0495 /// teacc 98.18 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0445, -0.0272, -0.0705, ..., 0.0732, -0.0115, 0.0294], + [-0.0457, 0.0568, -0.0633, ..., 0.0159, -0.0344, -0.0530], + [-0.0267, 0.0435, -0.0561, ..., -0.0326, 0.0049, 0.0604], + ..., + [-0.0026, 0.0214, -0.0136, ..., 0.0079, 0.0211, -0.0064], + [ 0.0483, -0.0477, 0.0188, ..., 0.0022, 0.0059, -0.0313], + [-0.0190, -0.0446, 0.0474, ..., -0.0213, 0.0031, -0.0121]], + device='cuda:0'), grad: tensor([[-0.0009, -0.0148, 0.0006, ..., -0.0180, -0.0047, -0.0072], + [-0.0050, 0.0062, 0.0013, ..., 0.0069, 0.0019, 0.0026], + [ 0.0020, -0.0007, 0.0017, ..., 0.0028, 0.0010, -0.0003], + ..., + [ 0.0008, 0.0017, -0.0070, ..., -0.0009, 0.0017, 0.0001], + [ 0.0170, 0.0065, 0.0167, ..., 0.0092, 0.0090, 0.0031], + [-0.0206, -0.0068, -0.0097, ..., -0.0053, -0.0045, 0.0010]], + device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0192, -0.0234, -0.0038, 0.0104, -0.0244, -0.0083, 0.0118, 0.0031, + 0.0271, -0.0266], device='cuda:0'), grad: tensor([-6.4758e-02, 1.5617e-02, 1.4748e-02, 1.8890e-02, 8.2169e-03, + 8.6278e-06, -5.5733e-03, -6.9237e-03, 7.7698e-02, -5.7953e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 227.34, cls_loss 0.6866 cls_loss_mapping 0.0417 cls_loss_causal 0.6541 re_mapping 0.0171 re_causal 0.0477 /// teacc 98.35 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0440, -0.0265, -0.0721, ..., 0.0736, -0.0116, 0.0296], + [-0.0466, 0.0578, -0.0626, ..., 0.0166, -0.0347, -0.0534], + [-0.0267, 0.0441, -0.0561, ..., -0.0327, 0.0052, 0.0604], + ..., + [-0.0028, 0.0207, -0.0134, ..., 0.0078, 0.0213, -0.0069], + [ 0.0473, -0.0478, 0.0186, ..., 0.0026, 0.0046, -0.0315], + [-0.0182, -0.0448, 0.0479, ..., -0.0216, 0.0032, -0.0119]], + device='cuda:0'), grad: tensor([[ 8.5688e-04, 9.1696e-04, 7.3862e-04, ..., 1.1387e-03, + 5.3978e-04, 1.0624e-03], + [ 1.3232e-04, 4.7951e-03, 2.5406e-03, ..., 3.4866e-03, + 8.1539e-04, 2.6855e-03], + [ 1.6642e-04, -5.5847e-03, 9.7036e-04, ..., 7.8869e-04, + -1.4639e-03, -2.3136e-03], + ..., + [ 1.0383e-04, 1.6813e-03, -7.7057e-03, ..., -4.5395e-04, + -3.4790e-03, 6.2180e-04], + [ 6.9952e-04, -5.1956e-03, -3.6221e-03, ..., -1.0155e-02, + 4.0984e-04, -3.3512e-03], + [ 5.2738e-04, 3.5793e-05, 5.8794e-04, ..., -2.3866e-04, + 1.1482e-03, -9.1887e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0191, -0.0231, -0.0045, 0.0107, -0.0241, -0.0083, 0.0114, 0.0036, + 0.0263, -0.0261], device='cuda:0'), grad: tensor([ 0.0138, 0.0297, -0.0013, 0.0136, -0.0064, 0.0176, 0.0025, -0.0144, + -0.0435, -0.0117], device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 226.56, cls_loss 0.6725 cls_loss_mapping 0.0417 cls_loss_causal 0.6399 re_mapping 0.0171 re_causal 0.0450 /// teacc 98.37 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0446, -0.0268, -0.0722, ..., 0.0738, -0.0113, 0.0297], + [-0.0473, 0.0577, -0.0631, ..., 0.0169, -0.0348, -0.0543], + [-0.0264, 0.0437, -0.0557, ..., -0.0329, 0.0052, 0.0608], + ..., + [-0.0029, 0.0211, -0.0137, ..., 0.0077, 0.0210, -0.0070], + [ 0.0483, -0.0484, 0.0186, ..., 0.0024, 0.0046, -0.0314], + [-0.0179, -0.0454, 0.0487, ..., -0.0223, 0.0037, -0.0126]], + device='cuda:0'), grad: tensor([[ 7.8506e-03, 8.2684e-04, 6.1607e-04, ..., 2.1103e-02, + 1.5732e-02, 2.4445e-02], + [ 1.0834e-03, 2.3022e-03, 1.2798e-03, ..., 2.8515e-03, + 1.4257e-03, 2.9793e-03], + [ 2.0123e-03, 1.5602e-03, 9.6130e-04, ..., 3.0994e-03, + 2.7122e-03, 1.9503e-03], + ..., + [ 1.3056e-03, -1.2140e-03, -4.7226e-03, ..., -5.2338e-03, + 1.4381e-03, -2.6150e-03], + [-7.6714e-03, -2.7542e-03, 2.8629e-03, ..., -1.9798e-03, + -1.3959e-04, -2.3575e-03], + [ 8.3268e-05, -5.8842e-04, 1.8539e-03, ..., -3.2578e-03, + 3.4313e-03, -4.5052e-03]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0191, -0.0235, -0.0041, 0.0106, -0.0240, -0.0083, 0.0117, 0.0034, + 0.0269, -0.0268], device='cuda:0'), grad: tensor([ 0.0457, 0.0284, 0.0124, 0.0118, 0.0160, -0.0114, -0.0076, -0.0261, + -0.0576, -0.0115], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 227.11, cls_loss 0.6809 cls_loss_mapping 0.0388 cls_loss_causal 0.6497 re_mapping 0.0164 re_causal 0.0435 /// teacc 98.22 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0444, -0.0276, -0.0720, ..., 0.0742, -0.0117, 0.0294], + [-0.0473, 0.0584, -0.0635, ..., 0.0170, -0.0344, -0.0542], + [-0.0271, 0.0443, -0.0564, ..., -0.0329, 0.0051, 0.0614], + ..., + [-0.0030, 0.0219, -0.0133, ..., 0.0075, 0.0211, -0.0076], + [ 0.0483, -0.0501, 0.0181, ..., 0.0025, 0.0040, -0.0314], + [-0.0174, -0.0460, 0.0490, ..., -0.0229, 0.0043, -0.0129]], + device='cuda:0'), grad: tensor([[ 1.6298e-03, 5.3930e-04, 2.4242e-03, ..., -1.0857e-02, + -3.4118e-04, 6.4039e-04], + [-3.1433e-03, -1.3351e-03, -7.8535e-04, ..., -5.0354e-03, + -4.9323e-05, -3.8319e-03], + [ 2.7714e-03, -4.0627e-03, 2.6011e-04, ..., -8.4114e-04, + -1.2161e-02, 3.6240e-04], + ..., + [ 2.2278e-03, 1.8835e-03, 7.0419e-03, ..., 3.8147e-03, + 7.0496e-03, 3.0861e-03], + [ 1.0595e-03, 1.4105e-03, 1.7929e-03, ..., 6.2370e-03, + 5.6458e-03, 1.1120e-03], + [-4.6997e-03, -2.5520e-03, -1.7731e-02, ..., -3.4180e-03, + -1.6968e-02, -6.3972e-03]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0187, -0.0232, -0.0040, 0.0103, -0.0242, -0.0080, 0.0119, 0.0037, + 0.0270, -0.0271], device='cuda:0'), grad: tensor([-0.0045, -0.0361, -0.0267, 0.0038, -0.0066, 0.0656, 0.0363, 0.0419, + 0.0193, -0.0930], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 227.34, cls_loss 0.6673 cls_loss_mapping 0.0490 cls_loss_causal 0.6390 re_mapping 0.0175 re_causal 0.0473 /// teacc 98.33 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0446, -0.0291, -0.0736, ..., 0.0748, -0.0121, 0.0299], + [-0.0471, 0.0584, -0.0634, ..., 0.0175, -0.0345, -0.0557], + [-0.0278, 0.0448, -0.0579, ..., -0.0336, 0.0049, 0.0622], + ..., + [-0.0023, 0.0223, -0.0134, ..., 0.0084, 0.0210, -0.0071], + [ 0.0489, -0.0504, 0.0183, ..., 0.0023, 0.0041, -0.0317], + [-0.0177, -0.0458, 0.0491, ..., -0.0232, 0.0040, -0.0126]], + device='cuda:0'), grad: tensor([[-0.0007, -0.0025, -0.0021, ..., -0.0001, -0.0026, 0.0006], + [ 0.0031, 0.0125, 0.0218, ..., 0.0167, 0.0017, 0.0051], + [ 0.0009, 0.0033, 0.0038, ..., 0.0032, 0.0009, 0.0044], + ..., + [ 0.0045, -0.0014, 0.0075, ..., 0.0001, 0.0032, -0.0015], + [-0.0065, -0.0024, 0.0018, ..., -0.0039, -0.0048, -0.0052], + [-0.0042, -0.0123, -0.0290, ..., -0.0141, 0.0005, -0.0003]], + device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0193, -0.0234, -0.0041, 0.0098, -0.0238, -0.0081, 0.0119, 0.0038, + 0.0269, -0.0272], device='cuda:0'), grad: tensor([ 0.0106, 0.0634, 0.0417, -0.0006, 0.0294, -0.0698, 0.0068, 0.0222, + -0.0464, -0.0572], device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 226.90, cls_loss 0.7091 cls_loss_mapping 0.0366 cls_loss_causal 0.6747 re_mapping 0.0164 re_causal 0.0442 /// teacc 98.15 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0447, -0.0297, -0.0739, ..., 0.0748, -0.0129, 0.0302], + [-0.0475, 0.0599, -0.0643, ..., 0.0177, -0.0354, -0.0552], + [-0.0270, 0.0455, -0.0573, ..., -0.0341, 0.0050, 0.0622], + ..., + [-0.0030, 0.0214, -0.0131, ..., 0.0085, 0.0206, -0.0079], + [ 0.0495, -0.0506, 0.0185, ..., 0.0030, 0.0042, -0.0321], + [-0.0181, -0.0462, 0.0497, ..., -0.0238, 0.0054, -0.0124]], + device='cuda:0'), grad: tensor([[ 1.4734e-04, 1.7662e-03, -2.2068e-03, ..., -1.1276e-02, + -2.7637e-03, -1.5076e-02], + [ 2.0695e-04, 4.1127e-05, -4.4656e-04, ..., -1.8396e-03, + 1.1168e-03, 6.2227e-04], + [ 2.9397e-04, 3.7708e-03, 2.5997e-03, ..., 1.8066e-02, + 4.3449e-03, 1.6922e-02], + ..., + [ 5.4932e-04, -3.0041e-04, 1.6565e-03, ..., 3.2120e-03, + 9.2506e-04, 2.4071e-03], + [ 6.9237e-04, -2.1553e-04, -1.1879e-02, ..., 1.9989e-03, + -5.5695e-03, 3.5114e-03], + [ 2.4738e-03, -1.0986e-03, 1.6953e-02, ..., -4.8370e-03, + 1.0620e-02, -2.4490e-03]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0195, -0.0232, -0.0046, 0.0097, -0.0236, -0.0082, 0.0119, 0.0037, + 0.0273, -0.0275], device='cuda:0'), grad: tensor([-0.0281, -0.0111, 0.0605, 0.0167, -0.0353, -0.0028, -0.0342, 0.0186, + 0.0073, 0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 226.82, cls_loss 0.6821 cls_loss_mapping 0.0360 cls_loss_causal 0.6485 re_mapping 0.0163 re_causal 0.0434 /// teacc 98.35 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0450, -0.0302, -0.0749, ..., 0.0749, -0.0137, 0.0297], + [-0.0475, 0.0607, -0.0648, ..., 0.0181, -0.0359, -0.0560], + [-0.0275, 0.0458, -0.0572, ..., -0.0337, 0.0051, 0.0628], + ..., + [-0.0012, 0.0228, -0.0120, ..., 0.0079, 0.0215, -0.0084], + [ 0.0498, -0.0516, 0.0188, ..., 0.0039, 0.0041, -0.0323], + [-0.0188, -0.0464, 0.0493, ..., -0.0243, 0.0055, -0.0123]], + device='cuda:0'), grad: tensor([[ 4.0126e-04, 8.9502e-04, 1.0490e-03, ..., 2.3384e-03, + 5.8794e-04, 1.0643e-03], + [-6.8550e-03, -1.5732e-02, -7.2060e-03, ..., -1.2619e-02, + 2.2333e-06, 1.0357e-03], + [ 9.5034e-04, 6.9695e-03, 1.9274e-03, ..., 5.7335e-03, + 3.3665e-03, 2.4223e-03], + ..., + [ 3.6449e-03, 1.7815e-03, 9.3460e-03, ..., 2.1229e-03, + 2.2163e-03, -3.2024e-03], + [-1.0753e-04, 3.3054e-03, -3.2043e-04, ..., 1.1435e-03, + -1.4715e-03, 1.6963e-04], + [ 1.1854e-03, 7.8917e-04, -5.5161e-03, ..., 2.4624e-03, + -1.2026e-03, 7.5006e-04]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0189, -0.0236, -0.0045, 0.0098, -0.0242, -0.0080, 0.0124, 0.0041, + 0.0279, -0.0280], device='cuda:0'), grad: tensor([ 0.0137, -0.0439, 0.0407, 0.0268, -0.0169, 0.0433, -0.0497, -0.0019, + -0.0212, 0.0090], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 54---------------------------------------------------- +epoch 54, time 229.08, cls_loss 0.6730 cls_loss_mapping 0.0358 cls_loss_causal 0.6422 re_mapping 0.0161 re_causal 0.0416 /// teacc 98.60 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0443, -0.0309, -0.0749, ..., 0.0757, -0.0141, 0.0299], + [-0.0481, 0.0619, -0.0646, ..., 0.0194, -0.0365, -0.0556], + [-0.0282, 0.0455, -0.0581, ..., -0.0339, 0.0054, 0.0632], + ..., + [-0.0009, 0.0224, -0.0124, ..., 0.0069, 0.0212, -0.0098], + [ 0.0506, -0.0512, 0.0183, ..., 0.0027, 0.0038, -0.0329], + [-0.0184, -0.0466, 0.0498, ..., -0.0242, 0.0059, -0.0125]], + device='cuda:0'), grad: tensor([[ 0.0015, 0.0013, 0.0015, ..., 0.0028, 0.0014, 0.0052], + [ 0.0003, 0.0011, 0.0020, ..., 0.0023, 0.0005, 0.0026], + [ 0.0027, 0.0022, -0.0019, ..., 0.0047, 0.0068, 0.0144], + ..., + [-0.0064, 0.0018, -0.0065, ..., 0.0022, -0.0047, 0.0029], + [ 0.0014, 0.0028, -0.0003, ..., -0.0020, 0.0063, -0.0052], + [ 0.0165, 0.0010, 0.0365, ..., 0.0025, 0.0188, 0.0041]], + device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0189, -0.0228, -0.0055, 0.0101, -0.0242, -0.0072, 0.0120, 0.0033, + 0.0276, -0.0272], device='cuda:0'), grad: tensor([ 0.0251, 0.0193, 0.0441, -0.1143, -0.0420, 0.0113, 0.0137, 0.0134, + -0.0309, 0.0601], device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 225.94, cls_loss 0.6639 cls_loss_mapping 0.0393 cls_loss_causal 0.6231 re_mapping 0.0165 re_causal 0.0425 /// teacc 98.56 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0457, -0.0319, -0.0760, ..., 0.0768, -0.0145, 0.0298], + [-0.0479, 0.0620, -0.0654, ..., 0.0196, -0.0371, -0.0557], + [-0.0283, 0.0460, -0.0582, ..., -0.0350, 0.0060, 0.0633], + ..., + [-0.0009, 0.0228, -0.0130, ..., 0.0075, 0.0213, -0.0103], + [ 0.0517, -0.0509, 0.0192, ..., 0.0025, 0.0041, -0.0323], + [-0.0182, -0.0476, 0.0498, ..., -0.0246, 0.0059, -0.0123]], + device='cuda:0'), grad: tensor([[-0.0014, 0.0006, -0.0035, ..., -0.0048, -0.0076, -0.0006], + [ 0.0005, -0.0028, -0.0019, ..., -0.0045, 0.0014, -0.0005], + [ 0.0012, 0.0017, 0.0028, ..., 0.0043, -0.0069, -0.0027], + ..., + [-0.0017, -0.0013, -0.0007, ..., -0.0074, 0.0029, -0.0018], + [ 0.0009, 0.0008, 0.0021, ..., 0.0037, 0.0023, 0.0031], + [ 0.0003, 0.0008, -0.0008, ..., 0.0025, 0.0007, 0.0020]], + device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0194, -0.0230, -0.0058, 0.0100, -0.0239, -0.0078, 0.0119, 0.0036, + 0.0279, -0.0273], device='cuda:0'), grad: tensor([-0.0314, -0.0090, 0.0184, -0.0298, 0.0388, 0.0023, -0.0050, -0.0356, + 0.0311, 0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 226.23, cls_loss 0.6539 cls_loss_mapping 0.0365 cls_loss_causal 0.6139 re_mapping 0.0165 re_causal 0.0426 /// teacc 98.47 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0471, -0.0323, -0.0768, ..., 0.0765, -0.0153, 0.0300], + [-0.0482, 0.0633, -0.0665, ..., 0.0199, -0.0374, -0.0559], + [-0.0280, 0.0468, -0.0587, ..., -0.0353, 0.0062, 0.0645], + ..., + [-0.0016, 0.0226, -0.0123, ..., 0.0082, 0.0214, -0.0107], + [ 0.0514, -0.0511, 0.0197, ..., 0.0030, 0.0035, -0.0325], + [-0.0172, -0.0473, 0.0504, ..., -0.0255, 0.0068, -0.0117]], + device='cuda:0'), grad: tensor([[-0.0006, 0.0002, -0.0003, ..., -0.0023, -0.0010, -0.0045], + [ 0.0004, -0.0015, 0.0019, ..., -0.0004, 0.0007, 0.0007], + [-0.0038, -0.0012, -0.0151, ..., -0.0027, -0.0051, -0.0018], + ..., + [ 0.0187, 0.0056, 0.0374, ..., 0.0016, 0.0205, 0.0017], + [-0.0012, -0.0014, -0.0055, ..., -0.0046, -0.0018, -0.0050], + [-0.0163, -0.0038, -0.0199, ..., 0.0025, -0.0164, 0.0025]], + device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0187, -0.0228, -0.0054, 0.0099, -0.0241, -0.0073, 0.0119, 0.0028, + 0.0280, -0.0268], device='cuda:0'), grad: tensor([-0.0093, 0.0067, -0.0417, 0.0099, 0.0079, -0.0384, 0.0321, 0.0372, + -0.0333, 0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 226.68, cls_loss 0.6652 cls_loss_mapping 0.0330 cls_loss_causal 0.6319 re_mapping 0.0156 re_causal 0.0420 /// teacc 98.45 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0478, -0.0325, -0.0771, ..., 0.0767, -0.0148, 0.0303], + [-0.0472, 0.0629, -0.0652, ..., 0.0202, -0.0365, -0.0572], + [-0.0285, 0.0469, -0.0581, ..., -0.0346, 0.0061, 0.0648], + ..., + [-0.0023, 0.0236, -0.0128, ..., 0.0085, 0.0207, -0.0112], + [ 0.0521, -0.0506, 0.0203, ..., 0.0026, 0.0037, -0.0321], + [-0.0173, -0.0472, 0.0503, ..., -0.0250, 0.0068, -0.0107]], + device='cuda:0'), grad: tensor([[ 0.0001, 0.0048, 0.0034, ..., 0.0059, 0.0023, 0.0190], + [-0.0004, -0.0001, 0.0053, ..., -0.0019, 0.0031, 0.0022], + [-0.0021, -0.0009, -0.0053, ..., -0.0027, -0.0046, -0.0020], + ..., + [ 0.0002, 0.0030, 0.0021, ..., 0.0023, 0.0025, 0.0030], + [ 0.0006, 0.0038, 0.0032, ..., 0.0010, 0.0011, 0.0039], + [ 0.0014, -0.0124, -0.0144, ..., -0.0109, -0.0052, -0.0164]], + device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0184, -0.0231, -0.0054, 0.0100, -0.0245, -0.0077, 0.0119, 0.0030, + 0.0284, -0.0261], device='cuda:0'), grad: tensor([ 0.0702, 0.0263, -0.0108, -0.0135, -0.0218, 0.0421, 0.0027, 0.0137, + 0.0092, -0.1180], device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 226.73, cls_loss 0.6412 cls_loss_mapping 0.0307 cls_loss_causal 0.5981 re_mapping 0.0154 re_causal 0.0405 /// teacc 98.35 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0479, -0.0326, -0.0773, ..., 0.0781, -0.0145, 0.0297], + [-0.0474, 0.0615, -0.0659, ..., 0.0199, -0.0367, -0.0575], + [-0.0289, 0.0474, -0.0585, ..., -0.0337, 0.0063, 0.0655], + ..., + [-0.0014, 0.0241, -0.0129, ..., 0.0072, 0.0203, -0.0120], + [ 0.0523, -0.0504, 0.0215, ..., 0.0035, 0.0036, -0.0326], + [-0.0173, -0.0473, 0.0503, ..., -0.0254, 0.0069, -0.0108]], + device='cuda:0'), grad: tensor([[ 8.9359e-04, 3.3784e-04, 4.8714e-03, ..., 8.6594e-03, + -5.0783e-04, -3.4828e-03], + [ 1.7667e-04, -1.5535e-03, 1.8826e-03, ..., 2.2519e-04, + 1.1311e-03, 1.6088e-03], + [ 8.3542e-04, 1.8482e-03, 2.2125e-03, ..., 3.6201e-03, + 4.6577e-03, 3.1872e-03], + ..., + [-3.6221e-03, -1.4954e-03, -7.9193e-03, ..., -3.7727e-03, + -4.9400e-03, -2.3155e-03], + [ 8.4496e-04, 1.1253e-03, 6.4735e-03, ..., 7.3242e-03, + 4.8943e-03, 5.7316e-04], + [ 1.1168e-03, 9.4831e-05, -2.2003e-02, ..., -6.6795e-03, + -1.4809e-02, -2.4014e-03]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0186, -0.0236, -0.0051, 0.0100, -0.0241, -0.0081, 0.0116, 0.0025, + 0.0293, -0.0263], device='cuda:0'), grad: tensor([ 0.0017, 0.0059, 0.0284, 0.0343, -0.0070, -0.0084, -0.0024, -0.0397, + 0.0224, -0.0353], device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 226.32, cls_loss 0.6865 cls_loss_mapping 0.0356 cls_loss_causal 0.6581 re_mapping 0.0152 re_causal 0.0419 /// teacc 98.41 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0480, -0.0330, -0.0778, ..., 0.0787, -0.0154, 0.0292], + [-0.0477, 0.0628, -0.0662, ..., 0.0203, -0.0373, -0.0575], + [-0.0289, 0.0475, -0.0595, ..., -0.0333, 0.0061, 0.0664], + ..., + [-0.0024, 0.0237, -0.0130, ..., 0.0078, 0.0205, -0.0118], + [ 0.0524, -0.0519, 0.0213, ..., 0.0016, 0.0026, -0.0343], + [-0.0162, -0.0466, 0.0513, ..., -0.0253, 0.0076, -0.0099]], + device='cuda:0'), grad: tensor([[-0.0007, -0.0038, -0.0112, ..., 0.0026, -0.0105, -0.0054], + [-0.0027, -0.0034, -0.0172, ..., -0.0051, -0.0031, 0.0051], + [-0.0020, -0.0051, 0.0040, ..., 0.0005, -0.0030, -0.0053], + ..., + [ 0.0021, 0.0044, 0.0101, ..., 0.0022, 0.0065, 0.0018], + [-0.0009, -0.0025, -0.0119, ..., -0.0163, -0.0023, 0.0002], + [ 0.0003, 0.0028, 0.0036, ..., 0.0084, -0.0002, 0.0018]], + device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0191, -0.0233, -0.0054, 0.0096, -0.0235, -0.0086, 0.0117, 0.0026, + 0.0281, -0.0256], device='cuda:0'), grad: tensor([-0.0180, 0.0036, -0.0356, 0.0650, 0.0108, -0.0241, 0.0217, 0.0152, + -0.0524, 0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 226.14, cls_loss 0.6057 cls_loss_mapping 0.0295 cls_loss_causal 0.5746 re_mapping 0.0152 re_causal 0.0396 /// teacc 98.42 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0487, -0.0329, -0.0780, ..., 0.0787, -0.0158, 0.0295], + [-0.0477, 0.0638, -0.0664, ..., 0.0211, -0.0379, -0.0573], + [-0.0292, 0.0475, -0.0600, ..., -0.0337, 0.0068, 0.0663], + ..., + [-0.0024, 0.0243, -0.0122, ..., 0.0073, 0.0215, -0.0112], + [ 0.0522, -0.0511, 0.0214, ..., 0.0021, 0.0027, -0.0342], + [-0.0164, -0.0464, 0.0517, ..., -0.0256, 0.0068, -0.0102]], + device='cuda:0'), grad: tensor([[ 1.3649e-04, 1.0080e-03, 2.9812e-03, ..., 2.4853e-03, + 3.3212e-04, 1.8520e-03], + [ 3.0565e-04, 2.1896e-03, 5.0697e-03, ..., 5.5733e-03, + 5.1081e-05, 3.9825e-03], + [ 9.2685e-05, -1.1969e-03, -1.6069e-03, ..., -1.7595e-03, + 1.4079e-04, -2.0504e-03], + ..., + [ 1.1230e-04, -6.8712e-04, -9.6464e-04, ..., -1.2226e-03, + 2.8253e-04, -6.3419e-04], + [ 4.3488e-04, -8.8596e-04, 5.4312e-04, ..., -3.1624e-03, + 1.2722e-03, -1.2693e-03], + [-6.8092e-03, 8.2397e-04, -2.2766e-02, ..., 3.3264e-03, + -1.1589e-02, 1.7939e-03]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0193, -0.0233, -0.0053, 0.0092, -0.0236, -0.0085, 0.0114, 0.0027, + 0.0283, -0.0253], device='cuda:0'), grad: tensor([ 0.0198, 0.0395, -0.0188, -0.0175, 0.0423, -0.0668, 0.0305, -0.0113, + -0.0087, -0.0088], device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 226.40, cls_loss 0.7080 cls_loss_mapping 0.0371 cls_loss_causal 0.6767 re_mapping 0.0153 re_causal 0.0414 /// teacc 98.57 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0478, -0.0331, -0.0784, ..., 0.0786, -0.0153, 0.0288], + [-0.0482, 0.0646, -0.0671, ..., 0.0208, -0.0383, -0.0579], + [-0.0294, 0.0479, -0.0585, ..., -0.0344, 0.0081, 0.0670], + ..., + [-0.0033, 0.0251, -0.0117, ..., 0.0080, 0.0209, -0.0114], + [ 0.0526, -0.0521, 0.0212, ..., 0.0017, 0.0025, -0.0341], + [-0.0176, -0.0467, 0.0515, ..., -0.0261, 0.0074, -0.0109]], + device='cuda:0'), grad: tensor([[ 3.6716e-03, 3.5267e-03, 3.0251e-03, ..., 6.1760e-03, + 1.6232e-03, 8.6136e-03], + [ 3.4142e-04, -9.5129e-05, 8.4400e-04, ..., -2.1267e-04, + 2.1040e-04, 2.1000e-03], + [ 2.1820e-03, -4.3958e-05, 1.4496e-03, ..., 3.7169e-04, + 4.2558e-04, 2.6436e-03], + ..., + [ 6.7253e-03, 1.7166e-03, 8.9874e-03, ..., -3.9864e-03, + 4.0131e-03, -4.1199e-03], + [-1.1864e-02, -5.1041e-03, -4.6005e-03, ..., -4.8027e-03, + -4.2038e-03, -1.8707e-02], + [-5.7945e-03, -6.9008e-03, -2.6321e-03, ..., -7.9193e-03, + 8.9931e-04, -6.0997e-03]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0191, -0.0235, -0.0057, 0.0101, -0.0237, -0.0089, 0.0123, 0.0028, + 0.0281, -0.0259], device='cuda:0'), grad: tensor([ 0.0318, 0.0059, 0.0117, 0.0132, 0.0101, 0.0096, 0.0112, -0.0214, + -0.0482, -0.0238], device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 226.75, cls_loss 0.6747 cls_loss_mapping 0.0322 cls_loss_causal 0.6389 re_mapping 0.0153 re_causal 0.0419 /// teacc 98.56 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0482, -0.0331, -0.0799, ..., 0.0787, -0.0139, 0.0296], + [-0.0486, 0.0651, -0.0678, ..., 0.0212, -0.0381, -0.0572], + [-0.0298, 0.0490, -0.0588, ..., -0.0342, 0.0091, 0.0679], + ..., + [-0.0036, 0.0252, -0.0117, ..., 0.0078, 0.0200, -0.0122], + [ 0.0531, -0.0530, 0.0219, ..., 0.0023, 0.0026, -0.0345], + [-0.0170, -0.0454, 0.0522, ..., -0.0260, 0.0079, -0.0107]], + device='cuda:0'), grad: tensor([[ 1.4725e-03, 1.5841e-03, 1.1702e-03, ..., -4.5729e-04, + -1.6851e-03, -4.3564e-03], + [ 8.7500e-04, -1.4465e-02, -1.2150e-03, ..., -1.5099e-02, + 1.2465e-03, -7.9117e-03], + [ 1.7958e-03, 6.4430e-03, 2.5826e-03, ..., 4.1313e-03, + 5.2414e-03, 2.7332e-03], + ..., + [ 1.0128e-03, 4.3564e-03, 1.9016e-03, ..., 5.8784e-03, + -5.5969e-05, 1.5192e-03], + [ 1.7014e-03, 1.9627e-03, 7.3929e-03, ..., 4.0970e-03, + 5.7983e-03, 3.8986e-03], + [ 7.3147e-04, 1.3962e-03, -1.8539e-03, ..., -4.2801e-03, + -2.9125e-03, 1.9264e-03]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0188, -0.0232, -0.0054, 0.0097, -0.0240, -0.0088, 0.0118, 0.0029, + 0.0286, -0.0257], device='cuda:0'), grad: tensor([-0.0079, -0.0663, 0.0359, -0.0396, -0.0086, 0.0312, -0.0165, 0.0219, + 0.0427, 0.0073], device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 226.38, cls_loss 0.6602 cls_loss_mapping 0.0317 cls_loss_causal 0.6214 re_mapping 0.0152 re_causal 0.0407 /// teacc 98.23 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0470, -0.0328, -0.0794, ..., 0.0800, -0.0134, 0.0305], + [-0.0492, 0.0657, -0.0683, ..., 0.0200, -0.0390, -0.0572], + [-0.0293, 0.0488, -0.0584, ..., -0.0338, 0.0093, 0.0675], + ..., + [-0.0042, 0.0262, -0.0123, ..., 0.0088, 0.0203, -0.0119], + [ 0.0541, -0.0534, 0.0226, ..., 0.0016, 0.0038, -0.0351], + [-0.0172, -0.0464, 0.0523, ..., -0.0262, 0.0073, -0.0111]], + device='cuda:0'), grad: tensor([[-0.0121, -0.0017, -0.0079, ..., -0.0124, -0.0083, -0.0029], + [ 0.0003, 0.0292, 0.0035, ..., 0.0325, 0.0002, 0.0024], + [ 0.0008, 0.0019, 0.0012, ..., 0.0041, 0.0009, 0.0028], + ..., + [ 0.0003, -0.0315, 0.0025, ..., -0.0361, 0.0008, -0.0032], + [ 0.0080, 0.0003, 0.0102, ..., 0.0062, 0.0042, -0.0007], + [ 0.0015, 0.0014, 0.0025, ..., 0.0030, 0.0010, 0.0014]], + device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0200, -0.0235, -0.0054, 0.0092, -0.0240, -0.0096, 0.0123, 0.0038, + 0.0282, -0.0262], device='cuda:0'), grad: tensor([-0.0443, 0.0531, 0.0212, 0.0180, -0.0452, -0.0095, 0.0087, -0.0460, + 0.0242, 0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 226.36, cls_loss 0.6584 cls_loss_mapping 0.0346 cls_loss_causal 0.6170 re_mapping 0.0147 re_causal 0.0391 /// teacc 98.22 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0461, -0.0327, -0.0803, ..., 0.0796, -0.0139, 0.0312], + [-0.0503, 0.0653, -0.0679, ..., 0.0202, -0.0390, -0.0573], + [-0.0291, 0.0499, -0.0582, ..., -0.0334, 0.0101, 0.0682], + ..., + [-0.0041, 0.0250, -0.0125, ..., 0.0086, 0.0197, -0.0124], + [ 0.0545, -0.0534, 0.0227, ..., 0.0008, 0.0030, -0.0360], + [-0.0178, -0.0454, 0.0526, ..., -0.0264, 0.0081, -0.0102]], + device='cuda:0'), grad: tensor([[ 5.4502e-04, 8.6117e-04, 1.8444e-03, ..., 2.4738e-03, + 8.7204e-03, 3.3588e-03], + [-4.8876e-04, 1.4257e-03, 3.5763e-03, ..., 4.6196e-03, + 1.7395e-03, 1.9503e-03], + [ 2.0349e-04, -1.2312e-03, 8.8310e-04, ..., -6.7234e-04, + 6.6185e-03, 9.0313e-04], + ..., + [ 6.7472e-05, -6.1646e-03, -1.1032e-02, ..., -1.4099e-02, + -2.5139e-03, -1.2131e-03], + [ 2.9778e-04, 6.4516e-04, -4.3607e-04, ..., -1.3161e-03, + -2.3804e-02, -8.9417e-03], + [ 2.7204e-04, 4.4136e-03, 4.7226e-03, ..., 7.4692e-03, + 3.6678e-03, 2.4319e-03]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0202, -0.0234, -0.0051, 0.0092, -0.0236, -0.0094, 0.0115, 0.0032, + 0.0279, -0.0257], device='cuda:0'), grad: tensor([ 0.0255, 0.0204, -0.0071, -0.0145, -0.0062, 0.0118, 0.0129, -0.0239, + -0.0442, 0.0252], device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 229.30, cls_loss 0.6578 cls_loss_mapping 0.0337 cls_loss_causal 0.6216 re_mapping 0.0139 re_causal 0.0387 /// teacc 98.49 lr 0.00010000 +Epoch 67, weight, value: tensor([[-4.6905e-02, -3.3826e-02, -8.0614e-02, ..., 7.9268e-02, + -1.4147e-02, 3.1453e-02], + [-5.0401e-02, 6.5258e-02, -6.7531e-02, ..., 2.0131e-02, + -3.8745e-02, -5.8191e-02], + [-2.8698e-02, 5.1321e-02, -5.7972e-02, ..., -3.2998e-02, + 1.0082e-02, 6.8395e-02], + ..., + [-4.4422e-03, 2.5881e-02, -1.2292e-02, ..., 8.6484e-03, + 2.0632e-02, -1.2349e-02], + [ 5.4573e-02, -5.4655e-02, 2.2363e-02, ..., 6.7149e-05, + 3.1612e-03, -3.6247e-02], + [-1.7869e-02, -4.6922e-02, 5.2582e-02, ..., -2.7049e-02, + 7.4955e-03, -1.0648e-02]], device='cuda:0'), grad: tensor([[ 0.0002, 0.0008, 0.0017, ..., 0.0018, 0.0043, 0.0024], + [ 0.0003, 0.0005, 0.0011, ..., 0.0013, 0.0013, 0.0020], + [ 0.0003, -0.0008, 0.0003, ..., -0.0049, -0.0002, -0.0047], + ..., + [ 0.0003, -0.0023, -0.0011, ..., 0.0006, 0.0021, 0.0017], + [ 0.0003, 0.0012, -0.0054, ..., 0.0019, 0.0005, 0.0026], + [-0.0023, -0.0031, -0.0024, ..., -0.0066, -0.0124, -0.0093]], + device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0197, -0.0234, -0.0046, 0.0090, -0.0238, -0.0094, 0.0122, 0.0039, + 0.0274, -0.0262], device='cuda:0'), grad: tensor([ 2.5085e-02, 1.4954e-02, -2.7252e-02, 4.5280e-03, 2.0844e-02, + 1.1620e-02, 2.1890e-05, 2.2095e-02, 1.4526e-02, -8.6487e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 229.68, cls_loss 0.6599 cls_loss_mapping 0.0350 cls_loss_causal 0.6306 re_mapping 0.0145 re_causal 0.0392 /// teacc 98.31 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0476, -0.0344, -0.0811, ..., 0.0789, -0.0144, 0.0306], + [-0.0498, 0.0659, -0.0681, ..., 0.0202, -0.0383, -0.0587], + [-0.0286, 0.0510, -0.0582, ..., -0.0335, 0.0092, 0.0692], + ..., + [-0.0055, 0.0267, -0.0120, ..., 0.0087, 0.0205, -0.0126], + [ 0.0551, -0.0540, 0.0231, ..., 0.0007, 0.0030, -0.0357], + [-0.0189, -0.0484, 0.0524, ..., -0.0274, 0.0078, -0.0107]], + device='cuda:0'), grad: tensor([[ 7.8678e-05, 5.8794e-04, 1.0853e-03, ..., 1.1473e-03, + 9.9564e-04, 1.3666e-03], + [ 3.1910e-03, 8.0633e-04, 3.6697e-03, ..., 8.5640e-04, + 1.2426e-03, 1.3180e-05], + [ 5.8365e-04, -9.1362e-04, 1.7214e-03, ..., -2.1038e-03, + 1.5507e-03, -9.2649e-04], + ..., + [-2.3651e-03, -1.5583e-03, 5.3482e-03, ..., 6.5517e-04, + 4.2000e-03, 2.2984e-04], + [ 1.6522e-04, 1.4257e-03, 1.0033e-02, ..., 2.8648e-03, + 5.5389e-03, 2.7084e-03], + [ 7.7629e-04, -1.8625e-03, -2.1469e-02, ..., -3.2368e-03, + -1.3695e-02, -8.8596e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0195, -0.0231, -0.0051, 0.0091, -0.0242, -0.0097, 0.0126, 0.0036, + 0.0281, -0.0261], device='cuda:0'), grad: tensor([ 0.0118, 0.0091, -0.0165, 0.0082, -0.0292, -0.0159, 0.0166, 0.0137, + 0.0340, -0.0316], device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 226.23, cls_loss 0.6545 cls_loss_mapping 0.0291 cls_loss_causal 0.6173 re_mapping 0.0150 re_causal 0.0406 /// teacc 98.51 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0472, -0.0341, -0.0806, ..., 0.0796, -0.0149, 0.0310], + [-0.0512, 0.0648, -0.0690, ..., 0.0198, -0.0393, -0.0603], + [-0.0288, 0.0524, -0.0593, ..., -0.0336, 0.0100, 0.0705], + ..., + [-0.0040, 0.0269, -0.0112, ..., 0.0085, 0.0213, -0.0134], + [ 0.0546, -0.0535, 0.0231, ..., 0.0010, 0.0029, -0.0349], + [-0.0188, -0.0491, 0.0525, ..., -0.0274, 0.0084, -0.0110]], + device='cuda:0'), grad: tensor([[ 8.6874e-06, 8.1968e-04, -2.3174e-03, ..., 1.8606e-03, + 1.5192e-03, -1.7834e-03], + [ 3.0413e-05, 8.4734e-04, 2.9907e-03, ..., 3.3226e-03, + 1.5650e-03, 2.4605e-03], + [-1.4551e-05, 2.1305e-03, 1.9383e-04, ..., 2.2125e-03, + 1.7147e-03, 7.9441e-04], + ..., + [ 3.5954e-04, 6.1941e-04, -4.9543e-04, ..., -1.2827e-03, + 5.6496e-03, -8.9407e-05], + [-7.7486e-04, 3.5782e-03, 4.4403e-03, ..., -2.0790e-03, + 2.0123e-03, 3.4332e-03], + [-3.8600e-04, 2.6894e-03, 1.1482e-03, ..., 4.6682e-04, + -2.3956e-03, 2.0504e-03]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0194, -0.0239, -0.0053, 0.0091, -0.0243, -0.0092, 0.0126, 0.0041, + 0.0282, -0.0259], device='cuda:0'), grad: tensor([ 0.0024, 0.0285, -0.0020, -0.1272, 0.0298, 0.0067, 0.0450, -0.0150, + 0.0235, 0.0082], device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 225.98, cls_loss 0.6531 cls_loss_mapping 0.0285 cls_loss_causal 0.6195 re_mapping 0.0149 re_causal 0.0400 /// teacc 98.27 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0463, -0.0350, -0.0811, ..., 0.0806, -0.0166, 0.0301], + [-0.0511, 0.0649, -0.0681, ..., 0.0197, -0.0380, -0.0601], + [-0.0287, 0.0533, -0.0600, ..., -0.0337, 0.0100, 0.0706], + ..., + [-0.0048, 0.0265, -0.0109, ..., 0.0084, 0.0217, -0.0132], + [ 0.0541, -0.0537, 0.0220, ..., 0.0002, 0.0026, -0.0356], + [-0.0170, -0.0486, 0.0528, ..., -0.0276, 0.0095, -0.0099]], + device='cuda:0'), grad: tensor([[-0.0009, 0.0003, 0.0004, ..., 0.0020, -0.0173, 0.0011], + [ 0.0002, -0.0014, 0.0002, ..., -0.0029, 0.0002, -0.0017], + [ 0.0004, 0.0019, 0.0031, ..., 0.0014, 0.0053, 0.0007], + ..., + [ 0.0057, 0.0293, 0.0137, ..., -0.0004, 0.0208, -0.0013], + [ 0.0011, 0.0003, 0.0010, ..., 0.0032, 0.0007, 0.0012], + [-0.0043, -0.0339, -0.0273, ..., -0.0039, -0.0274, 0.0007]], + device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0204, -0.0236, -0.0053, 0.0095, -0.0241, -0.0100, 0.0130, 0.0042, + 0.0267, -0.0259], device='cuda:0'), grad: tensor([-0.0265, -0.0189, 0.0184, 0.0007, -0.0128, 0.0464, 0.0058, 0.0288, + 0.0179, -0.0597], device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 226.69, cls_loss 0.6430 cls_loss_mapping 0.0250 cls_loss_causal 0.6037 re_mapping 0.0145 re_causal 0.0380 /// teacc 98.45 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0452, -0.0355, -0.0804, ..., 0.0808, -0.0165, 0.0300], + [-0.0501, 0.0653, -0.0668, ..., 0.0198, -0.0377, -0.0599], + [-0.0296, 0.0535, -0.0603, ..., -0.0333, 0.0100, 0.0718], + ..., + [-0.0048, 0.0264, -0.0115, ..., 0.0081, 0.0210, -0.0141], + [ 0.0545, -0.0548, 0.0223, ..., 0.0003, 0.0039, -0.0356], + [-0.0181, -0.0479, 0.0526, ..., -0.0276, 0.0090, -0.0106]], + device='cuda:0'), grad: tensor([[ 0.0029, 0.0014, 0.0011, ..., 0.0032, 0.0019, 0.0019], + [ 0.0003, 0.0024, 0.0020, ..., 0.0024, 0.0015, 0.0009], + [ 0.0009, 0.0016, 0.0017, ..., 0.0020, 0.0017, 0.0014], + ..., + [ 0.0006, -0.0054, -0.0124, ..., -0.0006, -0.0078, 0.0005], + [ 0.0006, -0.0043, 0.0021, ..., -0.0045, 0.0019, -0.0008], + [ 0.0010, 0.0051, 0.0104, ..., 0.0023, 0.0072, 0.0011]], + device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0205, -0.0228, -0.0051, 0.0090, -0.0234, -0.0100, 0.0119, 0.0037, + 0.0267, -0.0256], device='cuda:0'), grad: tensor([ 0.0216, 0.0186, 0.0185, -0.0337, 0.0168, -0.0199, -0.0234, -0.0047, + -0.0186, 0.0248], device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 228.13, cls_loss 0.6143 cls_loss_mapping 0.0260 cls_loss_causal 0.5900 re_mapping 0.0142 re_causal 0.0382 /// teacc 98.53 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0455, -0.0359, -0.0814, ..., 0.0817, -0.0172, 0.0298], + [-0.0508, 0.0663, -0.0665, ..., 0.0192, -0.0367, -0.0599], + [-0.0295, 0.0534, -0.0613, ..., -0.0347, 0.0100, 0.0715], + ..., + [-0.0045, 0.0268, -0.0109, ..., 0.0086, 0.0213, -0.0133], + [ 0.0548, -0.0547, 0.0224, ..., 0.0008, 0.0035, -0.0362], + [-0.0180, -0.0489, 0.0521, ..., -0.0270, 0.0084, -0.0107]], + device='cuda:0'), grad: tensor([[ 4.2367e-04, -3.2845e-03, -1.5812e-03, ..., -1.4374e-02, + 4.2748e-04, -5.7793e-04], + [-1.8513e-04, -9.3126e-04, -7.5645e-03, ..., 1.2960e-03, + 3.2634e-05, 9.6798e-04], + [-7.3051e-03, -3.2291e-03, 1.8206e-03, ..., 5.1041e-03, + -1.4486e-03, 2.6970e-03], + ..., + [ 7.5989e-03, -3.1662e-04, -8.4229e-03, ..., -1.4114e-03, + -1.2672e-02, -2.5787e-03], + [ 1.3161e-03, 1.9188e-03, 2.6016e-03, ..., 5.4436e-03, + 1.6270e-03, 2.7046e-03], + [-1.1673e-03, 7.4148e-04, 4.8370e-03, ..., 2.5272e-03, + 1.2352e-02, 5.9891e-04]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0195, -0.0231, -0.0061, 0.0092, -0.0238, -0.0093, 0.0121, 0.0047, + 0.0271, -0.0255], device='cuda:0'), grad: tensor([-0.0212, -0.0265, 0.0022, 0.0019, -0.0429, -0.0065, 0.0258, 0.0101, + 0.0312, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 227.25, cls_loss 0.6370 cls_loss_mapping 0.0296 cls_loss_causal 0.6017 re_mapping 0.0141 re_causal 0.0373 /// teacc 98.51 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0468, -0.0354, -0.0826, ..., 0.0825, -0.0180, 0.0297], + [-0.0512, 0.0664, -0.0671, ..., 0.0200, -0.0370, -0.0606], + [-0.0289, 0.0545, -0.0620, ..., -0.0354, 0.0102, 0.0723], + ..., + [-0.0040, 0.0271, -0.0104, ..., 0.0085, 0.0219, -0.0139], + [ 0.0542, -0.0551, 0.0226, ..., 0.0005, 0.0031, -0.0362], + [-0.0182, -0.0486, 0.0522, ..., -0.0263, 0.0080, -0.0106]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0018, 0.0044, ..., 0.0038, 0.0065, 0.0103], + [ 0.0007, -0.0049, -0.0051, ..., -0.0049, -0.0049, -0.0016], + [ 0.0005, -0.0025, -0.0048, ..., -0.0066, -0.0011, -0.0079], + ..., + [ 0.0001, 0.0014, 0.0052, ..., 0.0021, 0.0017, 0.0015], + [ 0.0008, -0.0012, 0.0027, ..., -0.0011, 0.0016, 0.0001], + [-0.0013, 0.0002, -0.0108, ..., -0.0011, -0.0103, -0.0100]], + device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0192, -0.0225, -0.0061, 0.0088, -0.0244, -0.0096, 0.0121, 0.0041, + 0.0279, -0.0249], device='cuda:0'), grad: tensor([ 0.0462, -0.0228, -0.0453, -0.0139, 0.0261, 0.0163, 0.0267, 0.0201, + -0.0062, -0.0473], device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 226.96, cls_loss 0.6248 cls_loss_mapping 0.0243 cls_loss_causal 0.5885 re_mapping 0.0140 re_causal 0.0370 /// teacc 98.50 lr 0.00010000 +Epoch 74, weight, value: tensor([[-4.7792e-02, -3.6763e-02, -8.3044e-02, ..., 8.3481e-02, + -1.8326e-02, 2.9715e-02], + [-5.0457e-02, 6.6243e-02, -6.7940e-02, ..., 2.0349e-02, + -3.7924e-02, -6.1128e-02], + [-2.9030e-02, 5.5064e-02, -6.1821e-02, ..., -3.5486e-02, + 1.0169e-02, 7.3595e-02], + ..., + [-4.4262e-03, 2.7171e-02, -1.0446e-02, ..., 8.7757e-03, + 2.2114e-02, -1.5178e-02], + [ 5.3918e-02, -5.4403e-02, 2.2416e-02, ..., 7.5320e-05, + 4.0763e-03, -3.6509e-02], + [-1.8090e-02, -4.8598e-02, 5.2618e-02, ..., -2.6676e-02, + 7.8779e-03, -1.0528e-02]], device='cuda:0'), grad: tensor([[ 0.0005, -0.0005, 0.0003, ..., -0.0011, 0.0004, -0.0008], + [ 0.0002, -0.0011, -0.0009, ..., -0.0012, 0.0001, -0.0019], + [ 0.0008, 0.0020, 0.0026, ..., 0.0028, 0.0012, 0.0029], + ..., + [ 0.0010, -0.0001, -0.0199, ..., 0.0034, -0.0046, 0.0039], + [-0.0230, 0.0030, 0.0115, ..., 0.0049, 0.0089, -0.0012], + [-0.0017, 0.0004, -0.0109, ..., -0.0022, -0.0087, -0.0024]], + device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0192, -0.0223, -0.0057, 0.0088, -0.0244, -0.0093, 0.0119, 0.0039, + 0.0276, -0.0251], device='cuda:0'), grad: tensor([-0.0055, -0.0177, 0.0299, 0.0018, 0.0379, -0.0310, 0.0153, 0.0112, + -0.0034, -0.0386], device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 226.97, cls_loss 0.6301 cls_loss_mapping 0.0308 cls_loss_causal 0.6005 re_mapping 0.0142 re_causal 0.0363 /// teacc 98.52 lr 0.00010000 +Epoch 75, weight, value: tensor([[-0.0473, -0.0370, -0.0841, ..., 0.0841, -0.0186, 0.0303], + [-0.0510, 0.0674, -0.0672, ..., 0.0212, -0.0354, -0.0616], + [-0.0294, 0.0555, -0.0623, ..., -0.0351, 0.0117, 0.0750], + ..., + [-0.0043, 0.0266, -0.0108, ..., 0.0075, 0.0207, -0.0167], + [ 0.0538, -0.0541, 0.0227, ..., -0.0005, 0.0033, -0.0373], + [-0.0175, -0.0493, 0.0540, ..., -0.0281, 0.0091, -0.0111]], + device='cuda:0'), grad: tensor([[ 0.0027, -0.0009, -0.0004, ..., 0.0004, 0.0033, 0.0017], + [ 0.0006, -0.0004, -0.0011, ..., -0.0018, -0.0010, 0.0002], + [ 0.0006, 0.0026, 0.0030, ..., 0.0049, 0.0061, 0.0051], + ..., + [ 0.0004, 0.0021, -0.0045, ..., 0.0012, 0.0021, 0.0039], + [ 0.0002, -0.0018, -0.0017, ..., -0.0019, 0.0007, -0.0025], + [ 0.0004, -0.0001, -0.0023, ..., 0.0009, -0.0052, 0.0006]], + device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0192, -0.0214, -0.0061, 0.0087, -0.0237, -0.0093, 0.0119, 0.0033, + 0.0277, -0.0256], device='cuda:0'), grad: tensor([-0.0003, -0.0036, 0.0336, 0.0075, -0.0266, 0.0210, -0.0070, 0.0206, + -0.0176, -0.0277], device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 226.97, cls_loss 0.6587 cls_loss_mapping 0.0272 cls_loss_causal 0.6297 re_mapping 0.0131 re_causal 0.0349 /// teacc 98.30 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.0475, -0.0358, -0.0846, ..., 0.0849, -0.0183, 0.0308], + [-0.0520, 0.0676, -0.0671, ..., 0.0215, -0.0363, -0.0618], + [-0.0295, 0.0558, -0.0632, ..., -0.0348, 0.0112, 0.0739], + ..., + [-0.0028, 0.0276, -0.0105, ..., 0.0078, 0.0215, -0.0168], + [ 0.0548, -0.0552, 0.0220, ..., -0.0012, 0.0024, -0.0381], + [-0.0187, -0.0493, 0.0542, ..., -0.0287, 0.0096, -0.0102]], + device='cuda:0'), grad: tensor([[ 0.0017, -0.0009, -0.0013, ..., -0.0026, -0.0003, -0.0053], + [ 0.0011, -0.0008, -0.0018, ..., -0.0027, 0.0003, -0.0014], + [ 0.0096, 0.0034, 0.0008, ..., 0.0016, 0.0021, 0.0071], + ..., + [-0.0207, -0.0024, 0.0011, ..., 0.0022, -0.0041, -0.0066], + [ 0.0002, 0.0004, 0.0007, ..., 0.0014, 0.0003, 0.0014], + [ 0.0007, 0.0013, 0.0014, ..., 0.0020, 0.0004, 0.0022]], + device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0196, -0.0216, -0.0064, 0.0084, -0.0233, -0.0092, 0.0118, 0.0037, + 0.0272, -0.0253], device='cuda:0'), grad: tensor([-0.0243, -0.0148, 0.0359, 0.0348, -0.0069, -0.0160, -0.0242, -0.0118, + 0.0096, 0.0177], device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 227.38, cls_loss 0.6007 cls_loss_mapping 0.0286 cls_loss_causal 0.5681 re_mapping 0.0135 re_causal 0.0355 /// teacc 98.36 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.0473, -0.0360, -0.0851, ..., 0.0855, -0.0182, 0.0307], + [-0.0528, 0.0684, -0.0674, ..., 0.0217, -0.0360, -0.0614], + [-0.0292, 0.0553, -0.0636, ..., -0.0356, 0.0110, 0.0739], + ..., + [-0.0028, 0.0277, -0.0102, ..., 0.0080, 0.0225, -0.0167], + [ 0.0550, -0.0550, 0.0224, ..., -0.0009, 0.0022, -0.0370], + [-0.0191, -0.0500, 0.0543, ..., -0.0289, 0.0094, -0.0109]], + device='cuda:0'), grad: tensor([[ 3.0088e-04, 1.7815e-03, 8.5211e-04, ..., 2.1076e-03, + 2.2793e-04, -1.7452e-04], + [ 7.7009e-04, -2.8744e-03, 5.3024e-04, ..., 1.2608e-03, + 1.5366e-04, 4.3899e-05], + [ 8.5831e-04, -4.0054e-03, -5.0354e-04, ..., -1.9274e-03, + 3.3951e-04, -5.4207e-03], + ..., + [-4.1084e-03, 1.7099e-03, -6.0997e-03, ..., 1.8368e-03, + -3.1605e-03, 1.2522e-03], + [ 1.4219e-03, 3.0098e-03, 6.5041e-04, ..., 2.1172e-03, + 4.6754e-04, 3.8395e-03], + [ 6.0768e-03, 1.6527e-03, 8.2932e-03, ..., 1.6747e-03, + 4.8943e-03, 1.5364e-03]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0199, -0.0216, -0.0065, 0.0079, -0.0237, -0.0084, 0.0117, 0.0038, + 0.0272, -0.0255], device='cuda:0'), grad: tensor([ 0.0168, -0.0109, -0.0217, -0.0768, 0.0207, 0.0179, -0.0172, 0.0143, + 0.0259, 0.0311], device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 227.39, cls_loss 0.6338 cls_loss_mapping 0.0309 cls_loss_causal 0.6007 re_mapping 0.0132 re_causal 0.0359 /// teacc 98.40 lr 0.00010000 +Epoch 78, weight, value: tensor([[-0.0478, -0.0360, -0.0852, ..., 0.0853, -0.0183, 0.0300], + [-0.0535, 0.0685, -0.0689, ..., 0.0223, -0.0372, -0.0612], + [-0.0288, 0.0566, -0.0629, ..., -0.0343, 0.0123, 0.0753], + ..., + [-0.0029, 0.0277, -0.0110, ..., 0.0072, 0.0222, -0.0171], + [ 0.0554, -0.0557, 0.0226, ..., -0.0006, 0.0020, -0.0380], + [-0.0191, -0.0502, 0.0546, ..., -0.0287, 0.0096, -0.0109]], + device='cuda:0'), grad: tensor([[ 0.0036, 0.0003, 0.0009, ..., 0.0007, 0.0019, 0.0018], + [-0.0022, -0.0053, -0.0019, ..., -0.0025, -0.0033, -0.0023], + [-0.0115, 0.0029, 0.0035, ..., 0.0028, 0.0012, -0.0005], + ..., + [ 0.0010, -0.0035, -0.0127, ..., -0.0040, -0.0164, -0.0090], + [-0.0028, 0.0012, 0.0016, ..., 0.0013, 0.0038, 0.0026], + [ 0.0016, 0.0011, 0.0041, ..., 0.0017, 0.0043, 0.0022]], + device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0197, -0.0212, -0.0054, 0.0077, -0.0231, -0.0084, 0.0112, 0.0029, + 0.0268, -0.0252], device='cuda:0'), grad: tensor([ 0.0150, -0.0274, -0.0061, 0.0256, -0.0182, -0.0142, 0.0284, -0.0462, + 0.0189, 0.0243], device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 227.23, cls_loss 0.6198 cls_loss_mapping 0.0250 cls_loss_causal 0.5861 re_mapping 0.0140 re_causal 0.0366 /// teacc 98.50 lr 0.00010000 +Epoch 79, weight, value: tensor([[-0.0470, -0.0364, -0.0857, ..., 0.0858, -0.0191, 0.0298], + [-0.0541, 0.0688, -0.0684, ..., 0.0231, -0.0367, -0.0623], + [-0.0283, 0.0566, -0.0636, ..., -0.0343, 0.0122, 0.0756], + ..., + [-0.0037, 0.0286, -0.0112, ..., 0.0071, 0.0222, -0.0169], + [ 0.0555, -0.0555, 0.0225, ..., -0.0007, 0.0020, -0.0375], + [-0.0193, -0.0510, 0.0544, ..., -0.0289, 0.0098, -0.0110]], + device='cuda:0'), grad: tensor([[ 0.0022, 0.0007, -0.0004, ..., -0.0019, 0.0007, 0.0013], + [ 0.0006, 0.0014, 0.0030, ..., 0.0045, 0.0026, 0.0018], + [ 0.0071, 0.0021, 0.0015, ..., 0.0017, 0.0040, 0.0046], + ..., + [ 0.0025, -0.0002, -0.0170, ..., -0.0010, -0.0081, 0.0010], + [-0.0138, 0.0013, 0.0093, ..., 0.0033, 0.0062, 0.0005], + [-0.0057, -0.0025, -0.0548, ..., -0.0021, -0.0375, -0.0002]], + device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0196, -0.0213, -0.0053, 0.0078, -0.0232, -0.0086, 0.0110, 0.0025, + 0.0274, -0.0250], device='cuda:0'), grad: tensor([ 6.5155e-03, 2.2995e-02, 2.6718e-02, -7.3120e-02, 5.8228e-02, + -8.3387e-05, 8.2626e-03, -2.5146e-02, 7.8125e-03, -3.2196e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 226.93, cls_loss 0.6156 cls_loss_mapping 0.0278 cls_loss_causal 0.5814 re_mapping 0.0132 re_causal 0.0343 /// teacc 98.43 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.0475, -0.0375, -0.0866, ..., 0.0864, -0.0198, 0.0298], + [-0.0538, 0.0685, -0.0689, ..., 0.0231, -0.0364, -0.0630], + [-0.0292, 0.0566, -0.0643, ..., -0.0352, 0.0117, 0.0755], + ..., + [-0.0045, 0.0289, -0.0108, ..., 0.0070, 0.0212, -0.0174], + [ 0.0548, -0.0556, 0.0225, ..., -0.0010, 0.0028, -0.0374], + [-0.0187, -0.0515, 0.0550, ..., -0.0288, 0.0105, -0.0110]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0003, 0.0013, ..., 0.0011, 0.0007, 0.0001], + [ 0.0004, 0.0004, 0.0011, ..., 0.0024, 0.0003, 0.0003], + [ 0.0002, -0.0003, -0.0012, ..., 0.0006, -0.0015, -0.0052], + ..., + [-0.0002, -0.0027, -0.0027, ..., -0.0064, -0.0009, -0.0010], + [ 0.0014, 0.0026, 0.0113, ..., 0.0019, 0.0173, 0.0125], + [-0.0024, 0.0005, -0.0075, ..., 0.0021, -0.0064, 0.0006]], + device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0191, -0.0214, -0.0058, 0.0080, -0.0236, -0.0083, 0.0115, 0.0022, + 0.0272, -0.0241], device='cuda:0'), grad: tensor([ 0.0098, 0.0149, -0.0101, -0.0132, 0.0114, -0.0464, 0.0178, -0.0268, + 0.0501, -0.0075], device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 227.04, cls_loss 0.6487 cls_loss_mapping 0.0268 cls_loss_causal 0.6169 re_mapping 0.0126 re_causal 0.0348 /// teacc 98.40 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.0474, -0.0376, -0.0873, ..., 0.0872, -0.0195, 0.0303], + [-0.0543, 0.0679, -0.0695, ..., 0.0233, -0.0371, -0.0629], + [-0.0294, 0.0566, -0.0652, ..., -0.0352, 0.0118, 0.0761], + ..., + [-0.0043, 0.0296, -0.0105, ..., 0.0071, 0.0213, -0.0184], + [ 0.0558, -0.0561, 0.0226, ..., -0.0008, 0.0023, -0.0378], + [-0.0184, -0.0514, 0.0551, ..., -0.0298, 0.0112, -0.0100]], + device='cuda:0'), grad: tensor([[-0.0012, -0.0007, 0.0011, ..., -0.0028, -0.0007, -0.0014], + [ 0.0004, 0.0002, 0.0014, ..., 0.0029, 0.0006, 0.0009], + [ 0.0003, 0.0012, 0.0010, ..., 0.0015, 0.0050, 0.0001], + ..., + [ 0.0006, 0.0012, 0.0048, ..., 0.0014, 0.0002, 0.0007], + [ 0.0004, 0.0004, 0.0002, ..., -0.0022, 0.0038, -0.0012], + [ 0.0002, -0.0004, -0.0097, ..., 0.0021, -0.0023, 0.0008]], + device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0201, -0.0213, -0.0065, 0.0077, -0.0230, -0.0091, 0.0113, 0.0025, + 0.0275, -0.0244], device='cuda:0'), grad: tensor([-0.0226, 0.0142, 0.0105, -0.0275, 0.0403, -0.0213, 0.0425, 0.0193, + -0.0384, -0.0170], device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 221.46, cls_loss 0.6037 cls_loss_mapping 0.0244 cls_loss_causal 0.5714 re_mapping 0.0132 re_causal 0.0340 /// teacc 98.15 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.0475, -0.0384, -0.0883, ..., 0.0868, -0.0203, 0.0304], + [-0.0549, 0.0680, -0.0693, ..., 0.0229, -0.0380, -0.0637], + [-0.0297, 0.0574, -0.0659, ..., -0.0357, 0.0113, 0.0759], + ..., + [-0.0042, 0.0297, -0.0106, ..., 0.0080, 0.0220, -0.0187], + [ 0.0558, -0.0569, 0.0228, ..., -0.0023, 0.0017, -0.0385], + [-0.0179, -0.0518, 0.0552, ..., -0.0295, 0.0107, -0.0100]], + device='cuda:0'), grad: tensor([[ 3.0565e-04, 4.8780e-04, 1.1835e-03, ..., -1.2665e-03, + -2.6122e-05, 1.5283e-04], + [ 1.2760e-03, 2.2087e-03, 4.6158e-03, ..., 2.9011e-03, + 1.1425e-03, 1.6680e-03], + [-1.9646e-03, -4.4870e-04, -8.1863e-03, ..., 1.2579e-03, + -7.2441e-03, -4.9019e-03], + ..., + [-5.4359e-03, -6.6147e-03, -9.1934e-03, ..., -1.4601e-03, + -7.9346e-04, 2.3632e-03], + [ 4.1351e-03, 2.1601e-04, 5.6610e-03, ..., -4.9210e-03, + 2.6131e-03, -1.1787e-03], + [ 9.6703e-04, 1.6260e-03, 2.1896e-03, ..., -8.0347e-04, + 2.1529e-04, 5.9462e-04]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0196, -0.0218, -0.0071, 0.0081, -0.0231, -0.0088, 0.0117, 0.0026, + 0.0272, -0.0238], device='cuda:0'), grad: tensor([ 0.0090, 0.0301, -0.0198, 0.0107, -0.0110, 0.0066, 0.0050, -0.0018, + -0.0325, 0.0038], device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 223.94, cls_loss 0.6375 cls_loss_mapping 0.0214 cls_loss_causal 0.5982 re_mapping 0.0132 re_causal 0.0356 /// teacc 98.16 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.0480, -0.0389, -0.0902, ..., 0.0876, -0.0206, 0.0287], + [-0.0551, 0.0688, -0.0690, ..., 0.0226, -0.0383, -0.0631], + [-0.0301, 0.0573, -0.0662, ..., -0.0364, 0.0110, 0.0768], + ..., + [-0.0054, 0.0305, -0.0107, ..., 0.0078, 0.0220, -0.0191], + [ 0.0566, -0.0573, 0.0224, ..., -0.0015, 0.0015, -0.0389], + [-0.0173, -0.0521, 0.0553, ..., -0.0301, 0.0109, -0.0097]], + device='cuda:0'), grad: tensor([[ 5.5599e-04, 1.3323e-03, 1.8215e-04, ..., 2.2564e-03, + 8.8215e-04, 1.1568e-03], + [ 3.0637e-04, 5.7793e-03, 7.3731e-05, ..., 3.9597e-03, + 2.5010e-04, 2.3327e-03], + [ 4.7684e-04, -3.3627e-03, 1.3936e-04, ..., 1.8435e-03, + 3.5930e-04, -1.2589e-03], + ..., + [-1.9744e-05, -1.1654e-03, -3.8528e-04, ..., -1.4639e-03, + 4.1485e-04, -7.3004e-04], + [ 1.2388e-03, -5.8212e-03, 3.4857e-04, ..., -1.0872e-03, + 1.6499e-03, -4.9067e-04], + [-2.5253e-03, 1.0014e-03, -1.7452e-03, ..., -3.4485e-03, + -1.2779e-04, 8.4162e-04]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0183, -0.0217, -0.0068, 0.0094, -0.0230, -0.0095, 0.0123, 0.0021, + 0.0275, -0.0241], device='cuda:0'), grad: tensor([ 0.0213, 0.0535, -0.0273, -0.0031, -0.0425, 0.0053, 0.0448, -0.0100, + -0.0251, -0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 220.07, cls_loss 0.5820 cls_loss_mapping 0.0210 cls_loss_causal 0.5441 re_mapping 0.0142 re_causal 0.0357 /// teacc 98.38 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.0482, -0.0388, -0.0900, ..., 0.0879, -0.0210, 0.0291], + [-0.0553, 0.0678, -0.0689, ..., 0.0223, -0.0390, -0.0641], + [-0.0291, 0.0568, -0.0675, ..., -0.0353, 0.0108, 0.0777], + ..., + [-0.0051, 0.0307, -0.0108, ..., 0.0073, 0.0217, -0.0198], + [ 0.0570, -0.0563, 0.0229, ..., -0.0021, 0.0014, -0.0395], + [-0.0180, -0.0510, 0.0559, ..., -0.0290, 0.0116, -0.0092]], + device='cuda:0'), grad: tensor([[-1.0010e-02, 4.4560e-04, 1.1911e-03, ..., 1.4048e-03, + 8.1396e-04, 1.0204e-03], + [-1.3100e-02, -2.0332e-03, -4.9744e-03, ..., -5.6763e-03, + -3.2616e-03, -7.3910e-04], + [ 7.7844e-05, -1.7989e-04, 1.7252e-03, ..., 1.3437e-03, + -2.4486e-04, -8.8739e-04], + ..., + [ 1.5697e-03, 1.8275e-04, 1.4687e-03, ..., 1.6823e-03, + 9.1934e-04, 8.8072e-04], + [ 2.0733e-03, 5.0449e-04, 3.1128e-03, ..., 2.2411e-03, + 1.9913e-03, 8.3590e-04], + [-5.6982e-04, 5.6982e-04, -8.2445e-04, ..., -5.1260e-04, + -8.4782e-04, 6.7949e-04]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0188, -0.0216, -0.0059, 0.0089, -0.0233, -0.0101, 0.0126, 0.0013, + 0.0270, -0.0231], device='cuda:0'), grad: tensor([-0.0149, -0.0550, 0.0165, 0.0178, 0.0368, -0.0078, -0.0314, 0.0172, + 0.0202, 0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 218.20, cls_loss 0.6049 cls_loss_mapping 0.0246 cls_loss_causal 0.5694 re_mapping 0.0130 re_causal 0.0340 /// teacc 98.37 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.0477, -0.0382, -0.0909, ..., 0.0871, -0.0207, 0.0297], + [-0.0562, 0.0685, -0.0681, ..., 0.0239, -0.0383, -0.0650], + [-0.0290, 0.0568, -0.0684, ..., -0.0350, 0.0113, 0.0783], + ..., + [-0.0047, 0.0312, -0.0102, ..., 0.0073, 0.0224, -0.0193], + [ 0.0579, -0.0573, 0.0235, ..., -0.0025, 0.0021, -0.0402], + [-0.0176, -0.0513, 0.0559, ..., -0.0289, 0.0106, -0.0093]], + device='cuda:0'), grad: tensor([[ 0.0007, 0.0004, 0.0033, ..., 0.0021, 0.0027, 0.0010], + [ 0.0068, 0.0013, 0.0009, ..., 0.0111, 0.0004, 0.0009], + [ 0.0023, 0.0058, -0.0030, ..., -0.0085, -0.0094, 0.0057], + ..., + [-0.0021, -0.0143, -0.0071, ..., -0.0057, -0.0033, -0.0122], + [-0.0024, 0.0029, 0.0041, ..., 0.0016, 0.0031, 0.0037], + [ 0.0023, 0.0006, -0.0005, ..., -0.0046, -0.0008, -0.0019]], + device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0186, -0.0216, -0.0060, 0.0084, -0.0234, -0.0102, 0.0127, 0.0022, + 0.0273, -0.0234], device='cuda:0'), grad: tensor([ 0.0176, 0.0361, -0.0079, 0.0287, -0.0239, 0.0528, 0.0064, -0.0679, + 0.0019, -0.0437], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 84---------------------------------------------------- +epoch 84, time 222.13, cls_loss 0.5991 cls_loss_mapping 0.0221 cls_loss_causal 0.5743 re_mapping 0.0121 re_causal 0.0313 /// teacc 98.61 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.0467, -0.0381, -0.0918, ..., 0.0867, -0.0213, 0.0296], + [-0.0566, 0.0689, -0.0686, ..., 0.0244, -0.0380, -0.0665], + [-0.0289, 0.0575, -0.0671, ..., -0.0341, 0.0112, 0.0786], + ..., + [-0.0056, 0.0313, -0.0102, ..., 0.0071, 0.0221, -0.0191], + [ 0.0577, -0.0577, 0.0229, ..., -0.0032, 0.0010, -0.0402], + [-0.0182, -0.0508, 0.0558, ..., -0.0290, 0.0104, -0.0087]], + device='cuda:0'), grad: tensor([[-6.9771e-03, 2.3293e-04, 1.3323e-03, ..., -7.0534e-03, + -1.2159e-03, -4.2343e-03], + [ 1.3514e-03, -3.6716e-04, 2.7549e-06, ..., -8.3029e-05, + 8.2111e-04, -1.1530e-03], + [-5.5847e-03, 1.1024e-03, -4.8485e-03, ..., 2.0504e-04, + -8.0261e-03, -7.6485e-03], + ..., + [-1.8501e-03, -4.0841e-04, 2.0301e-04, ..., -3.6488e-03, + -2.6093e-03, -8.4877e-04], + [ 1.4124e-03, 6.2943e-04, 1.8349e-03, ..., 1.8740e-03, + 1.0672e-03, 7.9346e-04], + [ 1.4353e-03, 4.0460e-04, -5.8403e-03, ..., 2.7156e-04, + -1.6441e-03, 1.8370e-04]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0185, -0.0228, -0.0053, 0.0085, -0.0230, -0.0099, 0.0125, 0.0029, + 0.0269, -0.0237], device='cuda:0'), grad: tensor([-0.0304, -0.0234, -0.0132, 0.0032, 0.0290, -0.0106, 0.0353, -0.0107, + 0.0224, -0.0018], device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 218.12, cls_loss 0.6512 cls_loss_mapping 0.0355 cls_loss_causal 0.6198 re_mapping 0.0125 re_causal 0.0340 /// teacc 98.41 lr 0.00010000 +Epoch 87, weight, value: tensor([[-0.0480, -0.0385, -0.0928, ..., 0.0874, -0.0217, 0.0293], + [-0.0561, 0.0688, -0.0677, ..., 0.0235, -0.0382, -0.0671], + [-0.0282, 0.0571, -0.0676, ..., -0.0346, 0.0120, 0.0790], + ..., + [-0.0065, 0.0325, -0.0100, ..., 0.0085, 0.0218, -0.0183], + [ 0.0566, -0.0576, 0.0219, ..., -0.0042, -0.0003, -0.0400], + [-0.0171, -0.0516, 0.0553, ..., -0.0294, 0.0107, -0.0079]], + device='cuda:0'), grad: tensor([[ 0.0012, 0.0014, 0.0003, ..., 0.0024, 0.0005, 0.0034], + [ 0.0012, 0.0038, 0.0014, ..., 0.0028, 0.0017, 0.0008], + [-0.0057, -0.0080, 0.0005, ..., -0.0021, 0.0006, -0.0007], + ..., + [ 0.0039, 0.0053, -0.0014, ..., 0.0018, -0.0008, 0.0031], + [-0.0005, 0.0018, 0.0003, ..., 0.0014, -0.0002, 0.0030], + [ 0.0016, 0.0005, 0.0036, ..., 0.0005, 0.0030, -0.0004]], + device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0189, -0.0224, -0.0050, 0.0087, -0.0239, -0.0097, 0.0130, 0.0030, + 0.0258, -0.0238], device='cuda:0'), grad: tensor([ 0.0285, 0.0152, -0.0501, 0.0042, -0.0412, 0.0252, -0.0447, 0.0427, + 0.0227, -0.0023], device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 218.30, cls_loss 0.6088 cls_loss_mapping 0.0186 cls_loss_causal 0.5766 re_mapping 0.0129 re_causal 0.0347 /// teacc 98.46 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.0484, -0.0390, -0.0934, ..., 0.0877, -0.0220, 0.0281], + [-0.0557, 0.0693, -0.0693, ..., 0.0242, -0.0385, -0.0677], + [-0.0281, 0.0571, -0.0682, ..., -0.0347, 0.0128, 0.0802], + ..., + [-0.0053, 0.0318, -0.0092, ..., 0.0086, 0.0225, -0.0194], + [ 0.0581, -0.0579, 0.0222, ..., -0.0051, -0.0008, -0.0405], + [-0.0178, -0.0513, 0.0556, ..., -0.0303, 0.0112, -0.0076]], + device='cuda:0'), grad: tensor([[ 0.0025, 0.0011, 0.0005, ..., 0.0022, 0.0015, 0.0031], + [-0.0003, -0.0077, -0.0043, ..., -0.0097, -0.0008, -0.0004], + [-0.0008, 0.0002, 0.0006, ..., 0.0015, -0.0007, -0.0021], + ..., + [-0.0144, 0.0031, -0.0058, ..., 0.0042, -0.0042, 0.0021], + [ 0.0198, 0.0038, 0.0034, ..., 0.0012, 0.0067, 0.0086], + [ 0.0098, 0.0019, 0.0064, ..., 0.0009, -0.0051, -0.0020]], + device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0182, -0.0223, -0.0053, 0.0091, -0.0242, -0.0100, 0.0135, 0.0023, + 0.0263, -0.0231], device='cuda:0'), grad: tensor([ 0.0221, -0.0450, 0.0177, -0.0638, 0.0125, -0.0295, 0.0168, 0.0077, + 0.0554, 0.0061], device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 218.56, cls_loss 0.6392 cls_loss_mapping 0.0172 cls_loss_causal 0.5942 re_mapping 0.0122 re_causal 0.0332 /// teacc 98.39 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.0489, -0.0400, -0.0935, ..., 0.0877, -0.0221, 0.0276], + [-0.0563, 0.0697, -0.0700, ..., 0.0242, -0.0387, -0.0678], + [-0.0283, 0.0577, -0.0686, ..., -0.0345, 0.0132, 0.0810], + ..., + [-0.0045, 0.0314, -0.0089, ..., 0.0073, 0.0227, -0.0205], + [ 0.0579, -0.0585, 0.0225, ..., -0.0050, -0.0007, -0.0404], + [-0.0183, -0.0505, 0.0562, ..., -0.0295, 0.0118, -0.0076]], + device='cuda:0'), grad: tensor([[ 1.1575e-04, -1.0529e-03, 4.1556e-04, ..., -6.5851e-04, + -2.3327e-03, -1.9236e-03], + [-1.1787e-03, -4.8409e-03, -2.2869e-03, ..., -2.4036e-05, + -3.8452e-03, -7.2193e-04], + [ 2.5978e-03, 1.0918e-02, 4.0746e-04, ..., -2.8687e-03, + 9.6741e-03, 1.5526e-02], + ..., + [ 1.4210e-03, 1.1225e-03, 7.6914e-04, ..., 1.6518e-03, + 1.2875e-03, 2.2774e-03], + [-1.6083e-02, -2.8706e-04, 6.3705e-04, ..., -1.6613e-03, + 7.2193e-04, -1.3676e-03], + [ 1.5039e-03, -1.8721e-03, -1.2684e-03, ..., -7.9775e-04, + -3.8123e-04, -1.2321e-03]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0169, -0.0223, -0.0054, 0.0094, -0.0243, -0.0091, 0.0132, 0.0020, + 0.0268, -0.0227], device='cuda:0'), grad: tensor([ 0.0085, -0.0145, 0.0092, 0.0062, 0.0283, 0.0100, -0.0131, 0.0259, + -0.0553, -0.0052], device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 218.05, cls_loss 0.5749 cls_loss_mapping 0.0183 cls_loss_causal 0.5481 re_mapping 0.0123 re_causal 0.0323 /// teacc 98.44 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.0494, -0.0398, -0.0938, ..., 0.0882, -0.0220, 0.0284], + [-0.0554, 0.0700, -0.0697, ..., 0.0246, -0.0373, -0.0677], + [-0.0289, 0.0577, -0.0682, ..., -0.0345, 0.0121, 0.0814], + ..., + [-0.0041, 0.0314, -0.0093, ..., 0.0069, 0.0229, -0.0210], + [ 0.0569, -0.0587, 0.0223, ..., -0.0051, -0.0007, -0.0409], + [-0.0189, -0.0512, 0.0559, ..., -0.0290, 0.0116, -0.0080]], + device='cuda:0'), grad: tensor([[-2.6751e-04, -1.6851e-03, 1.8873e-03, ..., -2.8152e-03, + 1.0004e-03, -1.1511e-03], + [ 1.0357e-03, -4.2295e-04, 2.1763e-03, ..., 5.4207e-03, + 1.1559e-03, 1.4997e-04], + [-2.8667e-03, 6.2103e-03, -3.3455e-03, ..., -3.6736e-03, + -1.1587e-03, -2.4748e-04], + ..., + [-3.0708e-03, -1.2306e-02, -2.5650e-02, ..., -1.1803e-02, + -1.6541e-02, -3.0249e-05], + [-6.3419e-05, -1.1921e-03, 2.6250e-04, ..., -3.8776e-03, + -2.4548e-03, -1.3704e-03], + [ 1.1168e-03, 7.3853e-03, 2.8076e-02, ..., 5.3520e-03, + 1.6953e-02, 6.4421e-04]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0177, -0.0215, -0.0062, 0.0099, -0.0242, -0.0097, 0.0133, 0.0018, + 0.0259, -0.0226], device='cuda:0'), grad: tensor([-0.0343, 0.0144, -0.0007, 0.0154, -0.0117, 0.0191, 0.0172, -0.0480, + -0.0238, 0.0523], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 89---------------------------------------------------- +epoch 89, time 218.84, cls_loss 0.6239 cls_loss_mapping 0.0201 cls_loss_causal 0.5853 re_mapping 0.0119 re_causal 0.0319 /// teacc 98.67 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.0502, -0.0410, -0.0935, ..., 0.0879, -0.0218, 0.0283], + [-0.0552, 0.0706, -0.0690, ..., 0.0248, -0.0372, -0.0682], + [-0.0281, 0.0583, -0.0683, ..., -0.0348, 0.0129, 0.0824], + ..., + [-0.0044, 0.0321, -0.0092, ..., 0.0078, 0.0221, -0.0209], + [ 0.0579, -0.0594, 0.0234, ..., -0.0053, 0.0016, -0.0412], + [-0.0192, -0.0511, 0.0560, ..., -0.0283, 0.0114, -0.0070]], + device='cuda:0'), grad: tensor([[-3.4165e-04, -1.1177e-03, -1.0738e-03, ..., 3.7253e-05, + -2.3441e-03, -2.5291e-03], + [-1.1721e-03, 3.0537e-03, 4.5967e-03, ..., 1.1349e-03, + 4.0817e-03, 3.0727e-03], + [ 1.7967e-03, -3.6316e-03, 2.6684e-03, ..., -5.8556e-03, + -1.5268e-03, -2.6531e-03], + ..., + [-9.4376e-03, -3.4657e-03, -1.8768e-02, ..., 2.6264e-03, + -1.2001e-02, 3.1738e-03], + [-6.7291e-03, 1.3342e-03, 9.3918e-03, ..., 5.9776e-03, + 6.8283e-04, 1.6203e-03], + [-3.9148e-04, 4.8943e-03, 1.0010e-02, ..., 5.7907e-03, + 1.0208e-02, 3.0613e-03]], device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0175, -0.0211, -0.0066, 0.0088, -0.0240, -0.0104, 0.0138, 0.0021, + 0.0256, -0.0214], device='cuda:0'), grad: tensor([-0.0199, 0.0109, -0.0048, 0.0121, -0.0464, 0.0240, -0.0221, -0.0184, + 0.0220, 0.0426], device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 217.96, cls_loss 0.5950 cls_loss_mapping 0.0174 cls_loss_causal 0.5661 re_mapping 0.0123 re_causal 0.0335 /// teacc 98.56 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.0502, -0.0404, -0.0933, ..., 0.0896, -0.0209, 0.0295], + [-0.0559, 0.0702, -0.0696, ..., 0.0251, -0.0371, -0.0685], + [-0.0292, 0.0593, -0.0693, ..., -0.0347, 0.0120, 0.0836], + ..., + [-0.0045, 0.0321, -0.0092, ..., 0.0073, 0.0226, -0.0221], + [ 0.0571, -0.0589, 0.0235, ..., -0.0045, 0.0019, -0.0414], + [-0.0188, -0.0515, 0.0564, ..., -0.0298, 0.0108, -0.0080]], + device='cuda:0'), grad: tensor([[ 1.0815e-03, 1.0023e-03, 9.5558e-04, ..., 4.5052e-03, + 7.9489e-04, 2.8267e-03], + [ 9.5654e-04, -1.0544e-04, 6.8235e-04, ..., 8.6498e-04, + 6.1607e-04, -8.1682e-04], + [ 1.3342e-03, 6.1321e-04, 5.6505e-04, ..., -2.1839e-03, + 1.2541e-03, -3.0746e-03], + ..., + [ 1.2445e-03, -1.5106e-03, 3.6011e-03, ..., -4.7722e-03, + 2.9588e-04, 8.3521e-06], + [ 1.0338e-03, 6.0511e-04, 1.2331e-03, ..., -1.7893e-04, + 6.5613e-04, -2.2240e-03], + [ 2.8820e-03, 1.2846e-03, -1.2922e-03, ..., 3.7117e-03, + -8.5497e-04, 1.4086e-03]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0184, -0.0212, -0.0065, 0.0090, -0.0231, -0.0107, 0.0131, 0.0010, + 0.0261, -0.0217], device='cuda:0'), grad: tensor([ 0.0309, -0.0075, -0.0078, 0.0314, -0.0566, -0.0345, 0.0144, -0.0010, + -0.0014, 0.0321], device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 218.09, cls_loss 0.6042 cls_loss_mapping 0.0231 cls_loss_causal 0.5701 re_mapping 0.0125 re_causal 0.0320 /// teacc 98.43 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.0488, -0.0409, -0.0938, ..., 0.0902, -0.0219, 0.0291], + [-0.0566, 0.0702, -0.0695, ..., 0.0245, -0.0367, -0.0684], + [-0.0284, 0.0589, -0.0687, ..., -0.0344, 0.0126, 0.0843], + ..., + [-0.0053, 0.0329, -0.0095, ..., 0.0071, 0.0218, -0.0217], + [ 0.0573, -0.0573, 0.0236, ..., -0.0032, 0.0028, -0.0406], + [-0.0185, -0.0514, 0.0565, ..., -0.0302, 0.0106, -0.0091]], + device='cuda:0'), grad: tensor([[ 8.7452e-04, 1.1873e-03, 7.3576e-04, ..., 2.5997e-03, + 1.9245e-03, 2.5806e-03], + [-9.3889e-04, -2.7981e-03, -1.0948e-03, ..., -2.1706e-03, + 8.2850e-05, 2.2483e-04], + [ 8.0490e-04, -3.2425e-05, 8.2684e-04, ..., -2.7447e-03, + -4.4403e-03, -2.8877e-03], + ..., + [ 1.4048e-03, 1.1559e-03, 1.6479e-03, ..., 3.0231e-03, + 4.4918e-04, 1.3371e-03], + [ 9.5606e-04, -1.7226e-05, 1.7796e-03, ..., -7.3099e-04, + 3.0651e-03, 1.6797e-04], + [ 1.2398e-03, 1.2064e-03, -6.0129e-04, ..., 2.5711e-03, + -2.8515e-03, 2.6655e-04]], device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0181, -0.0211, -0.0064, 0.0089, -0.0228, -0.0110, 0.0125, 0.0014, + 0.0265, -0.0217], device='cuda:0'), grad: tensor([ 0.0244, -0.0291, 0.0004, 0.0242, -0.0683, -0.0149, 0.0114, 0.0299, + 0.0028, 0.0191], device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 218.04, cls_loss 0.6047 cls_loss_mapping 0.0176 cls_loss_causal 0.5698 re_mapping 0.0117 re_causal 0.0319 /// teacc 98.40 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.0486, -0.0417, -0.0939, ..., 0.0909, -0.0226, 0.0278], + [-0.0567, 0.0700, -0.0681, ..., 0.0260, -0.0371, -0.0684], + [-0.0278, 0.0589, -0.0695, ..., -0.0346, 0.0129, 0.0837], + ..., + [-0.0053, 0.0335, -0.0093, ..., 0.0075, 0.0222, -0.0208], + [ 0.0574, -0.0580, 0.0230, ..., -0.0031, 0.0023, -0.0394], + [-0.0176, -0.0520, 0.0570, ..., -0.0313, 0.0111, -0.0097]], + device='cuda:0'), grad: tensor([[-0.0023, -0.0051, 0.0005, ..., -0.0070, -0.0006, -0.0059], + [ 0.0011, 0.0014, 0.0008, ..., 0.0068, 0.0009, 0.0009], + [ 0.0008, 0.0038, 0.0006, ..., 0.0017, 0.0029, 0.0009], + ..., + [ 0.0010, -0.0027, 0.0025, ..., 0.0051, 0.0019, 0.0002], + [ 0.0013, 0.0012, -0.0035, ..., -0.0094, -0.0015, 0.0008], + [ 0.0025, -0.0017, 0.0048, ..., -0.0033, 0.0021, 0.0004]], + device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0180, -0.0205, -0.0066, 0.0083, -0.0227, -0.0102, 0.0122, 0.0015, + 0.0265, -0.0221], device='cuda:0'), grad: tensor([-0.0133, 0.0276, 0.0080, -0.0147, -0.0141, 0.0048, 0.0064, -0.0027, + 0.0008, -0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 218.63, cls_loss 0.6161 cls_loss_mapping 0.0171 cls_loss_causal 0.5757 re_mapping 0.0113 re_causal 0.0310 /// teacc 98.40 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.0490, -0.0417, -0.0943, ..., 0.0919, -0.0225, 0.0284], + [-0.0565, 0.0708, -0.0677, ..., 0.0264, -0.0371, -0.0683], + [-0.0276, 0.0581, -0.0696, ..., -0.0354, 0.0139, 0.0845], + ..., + [-0.0046, 0.0339, -0.0101, ..., 0.0075, 0.0217, -0.0210], + [ 0.0583, -0.0590, 0.0248, ..., -0.0038, 0.0035, -0.0399], + [-0.0184, -0.0517, 0.0573, ..., -0.0323, 0.0112, -0.0098]], + device='cuda:0'), grad: tensor([[ 5.2929e-04, 4.3988e-04, 1.6298e-03, ..., 3.5210e-03, + 1.6232e-03, 3.2187e-04], + [ 1.0729e-03, -8.5545e-04, 5.8317e-04, ..., -9.4843e-04, + 6.1274e-04, -2.5868e-04], + [ 1.4210e-03, 2.2797e-02, 8.8930e-04, ..., 2.6093e-03, + 1.3084e-02, 1.9943e-02], + ..., + [ 9.9754e-04, 1.6489e-03, -1.2848e-02, ..., -1.5701e-02, + -1.2680e-02, 4.6849e-04], + [ 1.4353e-03, -2.7061e-04, 1.4696e-03, ..., 4.8965e-05, + 1.7290e-03, -2.3317e-04], + [ 2.8000e-03, 6.9475e-04, 1.1215e-02, ..., 1.6052e-02, + 1.1147e-02, 6.2799e-04]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0180, -0.0203, -0.0062, 0.0079, -0.0222, -0.0111, 0.0129, 0.0015, + 0.0265, -0.0225], device='cuda:0'), grad: tensor([ 0.0203, -0.0107, 0.0504, -0.0382, 0.0204, -0.0104, -0.0570, -0.0083, + -0.0077, 0.0411], device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 218.48, cls_loss 0.6299 cls_loss_mapping 0.0225 cls_loss_causal 0.5980 re_mapping 0.0122 re_causal 0.0311 /// teacc 98.59 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.0499, -0.0417, -0.0949, ..., 0.0919, -0.0233, 0.0277], + [-0.0576, 0.0722, -0.0677, ..., 0.0263, -0.0368, -0.0678], + [-0.0261, 0.0586, -0.0697, ..., -0.0364, 0.0144, 0.0845], + ..., + [-0.0042, 0.0338, -0.0097, ..., 0.0084, 0.0219, -0.0214], + [ 0.0594, -0.0594, 0.0240, ..., -0.0043, 0.0028, -0.0400], + [-0.0184, -0.0516, 0.0573, ..., -0.0329, 0.0112, -0.0101]], + device='cuda:0'), grad: tensor([[ 0.0028, 0.0009, 0.0037, ..., 0.0083, 0.0052, 0.0053], + [ 0.0007, -0.0018, 0.0002, ..., -0.0023, 0.0005, -0.0006], + [-0.0071, 0.0057, -0.0212, ..., -0.0083, -0.0194, -0.0074], + ..., + [ 0.0024, -0.0020, 0.0044, ..., 0.0034, 0.0017, 0.0007], + [-0.0027, 0.0008, 0.0017, ..., -0.0041, 0.0020, -0.0018], + [ 0.0066, -0.0002, 0.0197, ..., -0.0015, 0.0162, 0.0043]], + device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0170, -0.0199, -0.0055, 0.0069, -0.0217, -0.0110, 0.0131, 0.0019, + 0.0264, -0.0228], device='cuda:0'), grad: tensor([ 0.0341, -0.0160, -0.0341, -0.0040, 0.0040, 0.0286, -0.0175, 0.0177, + -0.0138, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 218.30, cls_loss 0.5955 cls_loss_mapping 0.0220 cls_loss_causal 0.5620 re_mapping 0.0123 re_causal 0.0315 /// teacc 98.54 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.0492, -0.0428, -0.0953, ..., 0.0904, -0.0238, 0.0269], + [-0.0590, 0.0735, -0.0679, ..., 0.0267, -0.0372, -0.0679], + [-0.0265, 0.0588, -0.0707, ..., -0.0363, 0.0138, 0.0848], + ..., + [-0.0049, 0.0337, -0.0095, ..., 0.0075, 0.0223, -0.0220], + [ 0.0606, -0.0599, 0.0248, ..., -0.0035, 0.0024, -0.0398], + [-0.0180, -0.0522, 0.0569, ..., -0.0318, 0.0118, -0.0096]], + device='cuda:0'), grad: tensor([[ 1.9760e-03, 1.0233e-03, 1.8692e-03, ..., -3.7932e-04, + -1.8702e-03, -2.1763e-03], + [ 4.3964e-04, -2.4414e-03, 1.1969e-03, ..., 2.1343e-03, + -1.8179e-04, -7.2765e-04], + [-6.7568e-04, -9.7656e-03, 7.5054e-04, ..., -2.4223e-03, + 8.2791e-05, -1.9684e-03], + ..., + [ 8.9693e-04, -3.7193e-03, -6.7377e-04, ..., 1.2636e-03, + -5.2834e-03, -1.4544e-03], + [-1.0853e-03, 2.5768e-03, -9.5444e-03, ..., 6.2141e-03, + -3.9291e-03, 1.3256e-03], + [-5.5771e-03, 4.6387e-03, -2.2469e-03, ..., -1.1345e-02, + 1.5213e-02, 1.0195e-03]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0173, -0.0211, -0.0058, 0.0074, -0.0222, -0.0102, 0.0127, 0.0015, + 0.0271, -0.0222], device='cuda:0'), grad: tensor([ 0.0017, -0.0020, -0.0802, -0.0083, 0.0337, 0.0052, 0.0479, 0.0002, + 0.0149, -0.0130], device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 218.43, cls_loss 0.5733 cls_loss_mapping 0.0167 cls_loss_causal 0.5449 re_mapping 0.0119 re_causal 0.0303 /// teacc 98.62 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.0488, -0.0430, -0.0952, ..., 0.0915, -0.0241, 0.0275], + [-0.0601, 0.0734, -0.0681, ..., 0.0261, -0.0371, -0.0687], + [-0.0281, 0.0603, -0.0714, ..., -0.0356, 0.0142, 0.0853], + ..., + [-0.0041, 0.0337, -0.0097, ..., 0.0069, 0.0215, -0.0224], + [ 0.0615, -0.0599, 0.0253, ..., -0.0040, 0.0027, -0.0404], + [-0.0184, -0.0527, 0.0570, ..., -0.0315, 0.0122, -0.0098]], + device='cuda:0'), grad: tensor([[ 3.3617e-05, 4.8041e-04, 8.6641e-04, ..., 1.3466e-03, + 2.6155e-04, 6.0797e-04], + [-4.6134e-04, 4.5853e-03, 2.3937e-03, ..., 4.1466e-03, + 1.8334e-04, 4.9305e-04], + [ 1.9932e-03, -1.6909e-03, 5.9814e-03, ..., 5.4626e-03, + 3.7885e-04, 8.5652e-05], + ..., + [-3.7212e-03, -6.8359e-03, -1.3290e-02, ..., -1.0193e-02, + -1.5564e-03, 2.3699e-04], + [ 1.3638e-03, 1.4439e-03, 2.4261e-03, ..., 2.1381e-03, + 5.2452e-04, 5.3310e-04], + [ 4.6229e-04, -3.3331e-04, -1.9445e-03, ..., -1.3475e-03, + 3.5119e-04, -1.5478e-03]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0175, -0.0214, -0.0055, 0.0074, -0.0223, -0.0103, 0.0132, 0.0025, + 0.0261, -0.0226], device='cuda:0'), grad: tensor([ 0.0130, 0.0157, 0.0173, -0.0216, 0.0183, -0.0132, 0.0204, -0.0288, + 0.0176, -0.0386], device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 218.67, cls_loss 0.5851 cls_loss_mapping 0.0203 cls_loss_causal 0.5483 re_mapping 0.0114 re_causal 0.0294 /// teacc 98.54 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.0477, -0.0430, -0.0961, ..., 0.0917, -0.0247, 0.0272], + [-0.0601, 0.0737, -0.0680, ..., 0.0263, -0.0367, -0.0694], + [-0.0287, 0.0606, -0.0717, ..., -0.0345, 0.0144, 0.0862], + ..., + [-0.0049, 0.0346, -0.0097, ..., 0.0067, 0.0216, -0.0230], + [ 0.0605, -0.0600, 0.0248, ..., -0.0040, 0.0025, -0.0398], + [-0.0186, -0.0533, 0.0574, ..., -0.0306, 0.0125, -0.0090]], + device='cuda:0'), grad: tensor([[ 0.0008, -0.0038, 0.0003, ..., -0.0056, -0.0010, -0.0006], + [-0.0123, -0.0028, -0.0015, ..., -0.0069, -0.0005, 0.0007], + [ 0.0015, 0.0029, 0.0009, ..., 0.0029, 0.0021, 0.0010], + ..., + [-0.0008, -0.0043, -0.0004, ..., -0.0019, 0.0014, -0.0004], + [ 0.0026, 0.0013, 0.0006, ..., 0.0026, 0.0009, 0.0004], + [ 0.0046, 0.0037, 0.0013, ..., 0.0030, 0.0023, 0.0005]], + device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0169, -0.0213, -0.0050, 0.0069, -0.0225, -0.0108, 0.0136, 0.0022, + 0.0259, -0.0216], device='cuda:0'), grad: tensor([-0.0288, -0.0629, 0.0260, -0.0250, 0.0234, 0.0178, 0.0152, -0.0131, + 0.0172, 0.0303], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 98---------------------------------------------------- +epoch 98, time 219.17, cls_loss 0.5878 cls_loss_mapping 0.0172 cls_loss_causal 0.5670 re_mapping 0.0121 re_causal 0.0322 /// teacc 98.69 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.0482, -0.0418, -0.0960, ..., 0.0922, -0.0250, 0.0283], + [-0.0609, 0.0742, -0.0676, ..., 0.0265, -0.0374, -0.0698], + [-0.0292, 0.0603, -0.0731, ..., -0.0345, 0.0152, 0.0872], + ..., + [-0.0042, 0.0352, -0.0094, ..., 0.0070, 0.0217, -0.0231], + [ 0.0615, -0.0614, 0.0245, ..., -0.0046, 0.0026, -0.0413], + [-0.0189, -0.0538, 0.0572, ..., -0.0317, 0.0125, -0.0088]], + device='cuda:0'), grad: tensor([[ 3.4928e-04, 6.1369e-04, 2.3258e-04, ..., 1.4248e-03, + 9.1076e-05, 5.4693e-04], + [ 5.6171e-04, 7.5579e-04, 5.3263e-04, ..., 2.0466e-03, + 1.3125e-04, 6.9237e-04], + [ 1.1301e-03, -8.3313e-03, -6.9084e-03, ..., -3.5076e-03, + -1.7670e-02, -6.7368e-03], + ..., + [-9.7561e-04, 1.0595e-03, 1.9760e-03, ..., 2.5253e-03, + 6.2103e-03, 1.6813e-03], + [ 1.6747e-03, 1.6479e-03, 4.7913e-03, ..., 1.5717e-03, + 1.7204e-03, 3.5214e-04], + [-1.6327e-02, 1.7376e-03, -2.2919e-02, ..., 2.7490e-04, + -3.6125e-03, 1.2932e-03]], device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0174, -0.0208, -0.0053, 0.0076, -0.0228, -0.0109, 0.0137, 0.0018, + 0.0253, -0.0216], device='cuda:0'), grad: tensor([ 0.0132, 0.0173, -0.0427, -0.0356, 0.0122, 0.0089, 0.0100, 0.0212, + 0.0233, -0.0277], device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 218.04, cls_loss 0.5984 cls_loss_mapping 0.0147 cls_loss_causal 0.5679 re_mapping 0.0119 re_causal 0.0331 /// teacc 98.64 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.0468, -0.0415, -0.0960, ..., 0.0934, -0.0251, 0.0280], + [-0.0600, 0.0749, -0.0678, ..., 0.0266, -0.0370, -0.0703], + [-0.0289, 0.0601, -0.0731, ..., -0.0347, 0.0151, 0.0877], + ..., + [-0.0051, 0.0351, -0.0084, ..., 0.0051, 0.0220, -0.0241], + [ 0.0610, -0.0616, 0.0246, ..., -0.0047, 0.0027, -0.0406], + [-0.0186, -0.0540, 0.0571, ..., -0.0313, 0.0129, -0.0087]], + device='cuda:0'), grad: tensor([[-2.8397e-02, 9.1171e-04, -1.7059e-02, ..., 1.6174e-03, + 2.0275e-03, 5.6934e-04], + [ 6.5470e-04, -8.0395e-04, 4.2588e-05, ..., -1.4973e-04, + 1.1034e-03, -7.8869e-04], + [ 3.8815e-04, -3.1109e-03, -4.8906e-05, ..., -1.8728e-04, + 1.5106e-03, -2.9469e-03], + ..., + [ 5.1546e-04, 3.1424e-04, 1.5268e-03, ..., -4.2992e-03, + 3.2940e-03, 5.5933e-04], + [ 1.0040e-02, 1.0548e-03, 1.0551e-02, ..., 3.7308e-03, + 1.5602e-02, 3.0231e-03], + [ 1.5688e-03, 6.7806e-04, 1.9379e-03, ..., -7.5388e-04, + 4.5471e-03, 6.3658e-04]], device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0186, -0.0209, -0.0053, 0.0073, -0.0227, -0.0103, 0.0131, 0.0005, + 0.0255, -0.0213], device='cuda:0'), grad: tensor([-0.0077, -0.0096, -0.0159, -0.0296, 0.0521, 0.0030, -0.0096, -0.0365, + 0.0517, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 100---------------------------------------------------- +epoch 100, time 218.69, cls_loss 0.6002 cls_loss_mapping 0.0147 cls_loss_causal 0.5663 re_mapping 0.0114 re_causal 0.0309 /// teacc 98.74 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.0472, -0.0413, -0.0966, ..., 0.0941, -0.0249, 0.0282], + [-0.0611, 0.0754, -0.0686, ..., 0.0273, -0.0382, -0.0701], + [-0.0281, 0.0606, -0.0724, ..., -0.0356, 0.0158, 0.0868], + ..., + [-0.0054, 0.0344, -0.0086, ..., 0.0054, 0.0205, -0.0248], + [ 0.0619, -0.0623, 0.0229, ..., -0.0047, 0.0028, -0.0399], + [-0.0187, -0.0541, 0.0576, ..., -0.0317, 0.0131, -0.0101]], + device='cuda:0'), grad: tensor([[ 8.4829e-04, 3.1042e-04, 1.1778e-03, ..., -8.9722e-03, + -2.0351e-03, -6.4011e-03], + [ 6.8188e-04, -8.7261e-05, 1.9586e-04, ..., -1.9445e-03, + 1.3380e-03, 8.5831e-04], + [ 8.6441e-03, 7.2479e-04, 3.2864e-03, ..., 2.6531e-03, + 3.8376e-03, 4.5433e-03], + ..., + [ 6.4135e-04, -4.5729e-04, 2.0847e-03, ..., 1.3180e-03, + 4.5052e-03, 3.5620e-04], + [-1.5516e-03, 7.4816e-04, -4.7531e-03, ..., 3.2215e-03, + -2.3212e-03, 2.7237e-03], + [ 1.0300e-03, -1.8225e-03, -1.4458e-03, ..., 1.0290e-03, + 2.7618e-03, 8.5735e-04]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0185, -0.0201, -0.0062, 0.0079, -0.0222, -0.0109, 0.0133, 0.0012, + 0.0251, -0.0220], device='cuda:0'), grad: tensor([-0.0115, -0.0043, 0.0302, -0.0140, -0.0260, -0.0152, -0.0120, 0.0399, + 0.0186, -0.0057], device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 218.64, cls_loss 0.6118 cls_loss_mapping 0.0133 cls_loss_causal 0.5710 re_mapping 0.0107 re_causal 0.0276 /// teacc 98.74 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.0478, -0.0414, -0.0967, ..., 0.0947, -0.0257, 0.0290], + [-0.0620, 0.0753, -0.0698, ..., 0.0275, -0.0394, -0.0706], + [-0.0284, 0.0603, -0.0737, ..., -0.0347, 0.0152, 0.0873], + ..., + [-0.0052, 0.0352, -0.0090, ..., 0.0051, 0.0204, -0.0258], + [ 0.0623, -0.0616, 0.0241, ..., -0.0046, 0.0035, -0.0394], + [-0.0188, -0.0543, 0.0577, ..., -0.0324, 0.0134, -0.0106]], + device='cuda:0'), grad: tensor([[ 1.0090e-03, -5.6171e-04, 1.3256e-03, ..., -6.9160e-03, + -1.4572e-03, 1.3695e-03], + [-2.2519e-04, 5.0068e-04, -3.6907e-03, ..., 5.9738e-03, + 5.7459e-04, -1.1845e-03], + [ 9.3555e-04, 1.2445e-04, 9.0504e-04, ..., 2.2793e-04, + 1.5907e-03, -1.3542e-04], + ..., + [ 1.8731e-05, 1.1969e-03, 5.4283e-03, ..., 2.8839e-03, + 5.8632e-03, 1.2388e-03], + [ 2.5272e-03, 7.1669e-04, 3.8700e-03, ..., 3.8891e-03, + 5.2376e-03, 3.0308e-03], + [ 1.1796e-04, -2.3308e-03, -2.0935e-02, ..., -8.7051e-03, + -1.6830e-02, -2.5635e-03]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0175, -0.0205, -0.0069, 0.0085, -0.0226, -0.0097, 0.0133, 0.0011, + 0.0251, -0.0213], device='cuda:0'), grad: tensor([ 0.0056, -0.0127, -0.0189, -0.0003, 0.0260, -0.0013, 0.0010, 0.0259, + 0.0273, -0.0525], device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 218.14, cls_loss 0.6045 cls_loss_mapping 0.0219 cls_loss_causal 0.5718 re_mapping 0.0120 re_causal 0.0313 /// teacc 98.55 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.0472, -0.0429, -0.0972, ..., 0.0943, -0.0269, 0.0275], + [-0.0628, 0.0760, -0.0701, ..., 0.0273, -0.0405, -0.0701], + [-0.0287, 0.0614, -0.0745, ..., -0.0339, 0.0160, 0.0875], + ..., + [-0.0053, 0.0357, -0.0090, ..., 0.0055, 0.0209, -0.0264], + [ 0.0616, -0.0619, 0.0246, ..., -0.0047, 0.0038, -0.0401], + [-0.0183, -0.0551, 0.0579, ..., -0.0329, 0.0140, -0.0106]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0005, 0.0004, ..., 0.0078, 0.0026, 0.0016], + [-0.0003, -0.0012, 0.0003, ..., -0.0061, -0.0060, 0.0003], + [ 0.0006, 0.0005, 0.0005, ..., 0.0042, 0.0030, 0.0015], + ..., + [-0.0072, 0.0006, -0.0078, ..., -0.0067, -0.0238, -0.0044], + [ 0.0003, 0.0002, 0.0052, ..., 0.0041, 0.0061, 0.0010], + [-0.0033, 0.0002, -0.0207, ..., -0.0034, -0.0059, 0.0002]], + device='cuda:0') +Epoch 104, bias, value: tensor([ 0.0173, -0.0211, -0.0063, 0.0078, -0.0220, -0.0097, 0.0133, 0.0013, + 0.0254, -0.0215], device='cuda:0'), grad: tensor([ 0.0327, -0.0329, 0.0209, -0.0009, 0.0490, 0.0259, -0.0190, -0.0609, + 0.0287, -0.0435], device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 218.88, cls_loss 0.6141 cls_loss_mapping 0.0194 cls_loss_causal 0.5822 re_mapping 0.0113 re_causal 0.0317 /// teacc 98.64 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.0474, -0.0428, -0.0970, ..., 0.0941, -0.0264, 0.0273], + [-0.0617, 0.0756, -0.0699, ..., 0.0270, -0.0406, -0.0720], + [-0.0291, 0.0611, -0.0747, ..., -0.0336, 0.0166, 0.0877], + ..., + [-0.0051, 0.0369, -0.0089, ..., 0.0052, 0.0209, -0.0262], + [ 0.0625, -0.0629, 0.0240, ..., -0.0043, 0.0029, -0.0401], + [-0.0199, -0.0555, 0.0586, ..., -0.0329, 0.0152, -0.0098]], + device='cuda:0'), grad: tensor([[-2.7227e-04, 7.0000e-04, 4.5562e-04, ..., 6.1703e-04, + -1.9938e-05, -7.1812e-04], + [ 1.3936e-04, -1.7846e-04, 6.0511e-04, ..., -3.8891e-03, + -3.0732e-04, 7.1907e-04], + [-5.0354e-03, 2.3499e-03, 1.0099e-03, ..., 3.2806e-03, + 1.4315e-03, 7.4625e-04], + ..., + [ 4.9353e-04, -3.3855e-03, -1.1148e-03, ..., -2.5787e-03, + -3.7346e-03, -2.0039e-04], + [ 4.0793e-04, -5.9748e-04, -6.0883e-03, ..., -1.6870e-03, + -2.0256e-03, -2.2411e-03], + [-1.0910e-03, -1.5326e-03, -1.5545e-04, ..., -1.3781e-03, + 8.2827e-04, -6.0225e-04]], device='cuda:0') +Epoch 105, bias, value: tensor([ 0.0165, -0.0210, -0.0064, 0.0070, -0.0219, -0.0095, 0.0135, 0.0011, + 0.0262, -0.0214], device='cuda:0'), grad: tensor([ 0.0105, -0.0149, 0.0142, -0.0075, 0.0292, 0.0120, 0.0171, -0.0040, + -0.0293, -0.0274], device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 219.88, cls_loss 0.5814 cls_loss_mapping 0.0171 cls_loss_causal 0.5524 re_mapping 0.0116 re_causal 0.0293 /// teacc 98.60 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.0482, -0.0431, -0.0970, ..., 0.0941, -0.0258, 0.0283], + [-0.0620, 0.0764, -0.0699, ..., 0.0271, -0.0384, -0.0717], + [-0.0291, 0.0616, -0.0763, ..., -0.0346, 0.0161, 0.0879], + ..., + [-0.0054, 0.0366, -0.0086, ..., 0.0064, 0.0212, -0.0265], + [ 0.0639, -0.0638, 0.0250, ..., -0.0043, 0.0036, -0.0401], + [-0.0193, -0.0555, 0.0579, ..., -0.0340, 0.0148, -0.0088]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0043, -0.0009, ..., 0.0038, 0.0054, 0.0077], + [-0.0002, -0.0080, -0.0011, ..., -0.0106, -0.0036, -0.0049], + [ 0.0005, 0.0006, 0.0003, ..., 0.0028, 0.0024, 0.0029], + ..., + [ 0.0002, 0.0033, 0.0013, ..., 0.0057, 0.0016, 0.0029], + [ 0.0003, -0.0002, -0.0023, ..., -0.0002, -0.0006, -0.0009], + [ 0.0002, 0.0012, -0.0050, ..., 0.0026, -0.0010, 0.0014]], + device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0170, -0.0212, -0.0064, 0.0073, -0.0222, -0.0098, 0.0137, 0.0014, + 0.0263, -0.0217], device='cuda:0'), grad: tensor([ 0.0143, -0.0504, -0.0003, -0.0067, 0.0305, 0.0180, -0.0450, 0.0374, + -0.0133, 0.0155], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 105---------------------------------------------------- +epoch 105, time 220.20, cls_loss 0.5839 cls_loss_mapping 0.0153 cls_loss_causal 0.5568 re_mapping 0.0118 re_causal 0.0307 /// teacc 98.76 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.0470, -0.0442, -0.0981, ..., 0.0928, -0.0253, 0.0282], + [-0.0636, 0.0770, -0.0700, ..., 0.0274, -0.0387, -0.0717], + [-0.0285, 0.0616, -0.0769, ..., -0.0343, 0.0155, 0.0879], + ..., + [-0.0065, 0.0374, -0.0083, ..., 0.0072, 0.0223, -0.0268], + [ 0.0633, -0.0645, 0.0243, ..., -0.0049, 0.0037, -0.0403], + [-0.0187, -0.0558, 0.0586, ..., -0.0355, 0.0144, -0.0099]], + device='cuda:0'), grad: tensor([[ 6.5756e-04, 1.4668e-03, 4.4227e-05, ..., 3.3779e-03, + 8.8573e-05, 3.6907e-03], + [-1.1578e-03, -4.2725e-03, 1.2457e-05, ..., -6.0234e-03, + 1.0043e-04, -3.3607e-03], + [ 6.8235e-04, -1.8301e-03, -4.6577e-03, ..., 1.2646e-03, + -1.2606e-05, 8.3208e-04], + ..., + [ 4.4179e-04, 6.9771e-03, -1.8024e-04, ..., 3.2597e-03, + 1.3878e-02, 8.0643e-03], + [ 1.1635e-04, 2.3251e-03, 4.3845e-04, ..., 2.2850e-03, + 2.6073e-03, 3.9520e-03], + [ 6.4421e-04, 6.7482e-03, 5.1193e-03, ..., 2.1820e-03, + 6.6376e-03, 6.3591e-03]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0172, -0.0214, -0.0057, 0.0076, -0.0227, -0.0103, 0.0139, 0.0016, + 0.0266, -0.0224], device='cuda:0'), grad: tensor([ 0.0254, -0.0124, 0.0024, -0.0752, 0.0165, 0.0374, -0.0688, 0.0361, + 0.0073, 0.0314], device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 219.48, cls_loss 0.5804 cls_loss_mapping 0.0144 cls_loss_causal 0.5507 re_mapping 0.0111 re_causal 0.0298 /// teacc 98.63 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.0463, -0.0443, -0.0996, ..., 0.0929, -0.0266, 0.0276], + [-0.0646, 0.0765, -0.0705, ..., 0.0281, -0.0399, -0.0726], + [-0.0282, 0.0628, -0.0764, ..., -0.0338, 0.0165, 0.0889], + ..., + [-0.0065, 0.0376, -0.0087, ..., 0.0068, 0.0221, -0.0284], + [ 0.0645, -0.0658, 0.0239, ..., -0.0050, 0.0028, -0.0400], + [-0.0179, -0.0568, 0.0591, ..., -0.0347, 0.0149, -0.0092]], + device='cuda:0'), grad: tensor([[ 8.9340e-03, 1.0185e-03, 8.3351e-04, ..., 8.0338e-03, + 1.3100e-02, 3.9330e-03], + [ 6.2883e-05, 4.9019e-04, 4.6158e-04, ..., 1.1282e-03, + 6.8760e-04, 1.1263e-03], + [ 5.7012e-05, 2.2240e-03, 2.9812e-03, ..., 2.3575e-03, + 3.8719e-03, 2.5787e-03], + ..., + [ 2.9731e-04, 2.7714e-03, 1.7176e-03, ..., 2.9850e-03, + 3.9406e-03, 3.5248e-03], + [-5.4979e-04, -2.0523e-03, 6.0034e-04, ..., -1.1559e-02, + -4.3983e-03, -7.4234e-03], + [ 9.2626e-05, 4.0030e-04, 1.5039e-03, ..., 7.7820e-04, + 2.7332e-03, 2.1305e-03]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0168, -0.0214, -0.0059, 0.0071, -0.0233, -0.0104, 0.0141, 0.0011, + 0.0268, -0.0207], device='cuda:0'), grad: tensor([ 0.0370, 0.0112, 0.0307, -0.0123, -0.0102, 0.0284, -0.0762, 0.0321, + -0.0351, -0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 219.17, cls_loss 0.5792 cls_loss_mapping 0.0139 cls_loss_causal 0.5414 re_mapping 0.0105 re_causal 0.0272 /// teacc 98.70 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.0463, -0.0437, -0.0996, ..., 0.0932, -0.0260, 0.0278], + [-0.0639, 0.0771, -0.0705, ..., 0.0283, -0.0388, -0.0722], + [-0.0275, 0.0626, -0.0765, ..., -0.0340, 0.0171, 0.0898], + ..., + [-0.0066, 0.0383, -0.0078, ..., 0.0064, 0.0226, -0.0295], + [ 0.0646, -0.0672, 0.0240, ..., -0.0054, 0.0026, -0.0407], + [-0.0167, -0.0578, 0.0592, ..., -0.0343, 0.0154, -0.0086]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0010, -0.0002, ..., -0.0016, -0.0013, -0.0043], + [ 0.0011, 0.0024, 0.0013, ..., 0.0020, 0.0009, 0.0024], + [-0.0016, -0.0046, -0.0005, ..., -0.0047, -0.0039, -0.0064], + ..., + [ 0.0015, 0.0020, 0.0089, ..., -0.0005, 0.0072, -0.0007], + [ 0.0004, 0.0007, 0.0011, ..., 0.0014, 0.0013, 0.0017], + [ 0.0025, -0.0010, 0.0035, ..., 0.0016, 0.0009, 0.0021]], + device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0178, -0.0209, -0.0062, 0.0066, -0.0227, -0.0102, 0.0142, 0.0006, + 0.0259, -0.0208], device='cuda:0'), grad: tensor([-0.0284, 0.0282, -0.0228, 0.0161, -0.0197, 0.0146, 0.0042, -0.0216, + 0.0166, 0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 218.96, cls_loss 0.6206 cls_loss_mapping 0.0228 cls_loss_causal 0.5859 re_mapping 0.0116 re_causal 0.0312 /// teacc 98.63 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.0452, -0.0455, -0.0991, ..., 0.0927, -0.0262, 0.0271], + [-0.0643, 0.0767, -0.0718, ..., 0.0282, -0.0386, -0.0729], + [-0.0288, 0.0622, -0.0770, ..., -0.0332, 0.0156, 0.0897], + ..., + [-0.0059, 0.0407, -0.0077, ..., 0.0085, 0.0228, -0.0285], + [ 0.0651, -0.0673, 0.0236, ..., -0.0059, 0.0026, -0.0407], + [-0.0165, -0.0586, 0.0593, ..., -0.0353, 0.0154, -0.0100]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0014, 0.0022, ..., 0.0050, 0.0023, 0.0016], + [-0.0003, -0.0059, 0.0003, ..., -0.0122, 0.0005, -0.0002], + [-0.0009, -0.0014, -0.0013, ..., 0.0005, -0.0028, -0.0122], + ..., + [ 0.0002, 0.0017, 0.0016, ..., 0.0005, 0.0019, 0.0014], + [ 0.0004, 0.0009, 0.0023, ..., 0.0090, 0.0033, 0.0019], + [-0.0004, 0.0016, 0.0038, ..., -0.0028, 0.0008, -0.0015]], + device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0178, -0.0214, -0.0065, 0.0063, -0.0236, -0.0110, 0.0151, 0.0024, + 0.0262, -0.0211], device='cuda:0'), grad: tensor([ 0.0343, -0.0551, 0.0041, -0.0344, 0.0190, 0.0263, -0.0214, -0.0083, + 0.0354, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 219.63, cls_loss 0.6063 cls_loss_mapping 0.0164 cls_loss_causal 0.5604 re_mapping 0.0111 re_causal 0.0279 /// teacc 98.35 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.0455, -0.0473, -0.0987, ..., 0.0928, -0.0261, 0.0278], + [-0.0647, 0.0787, -0.0718, ..., 0.0283, -0.0377, -0.0733], + [-0.0295, 0.0625, -0.0778, ..., -0.0327, 0.0156, 0.0910], + ..., + [-0.0061, 0.0401, -0.0074, ..., 0.0073, 0.0228, -0.0294], + [ 0.0660, -0.0666, 0.0239, ..., -0.0046, 0.0037, -0.0406], + [-0.0172, -0.0598, 0.0593, ..., -0.0356, 0.0154, -0.0111]], + device='cuda:0'), grad: tensor([[ 4.8667e-05, -3.2377e-04, 2.7800e-04, ..., -1.6947e-03, + 9.3889e-04, -1.6270e-03], + [ 1.3614e-04, 5.0735e-03, 1.6651e-03, ..., 5.4703e-03, + 3.8700e-03, 1.3113e-03], + [ 1.3435e-04, 2.8381e-03, 9.6655e-04, ..., 2.9545e-03, + 5.1346e-03, 6.6032e-03], + ..., + [ 9.1672e-05, 6.1572e-05, 1.7462e-03, ..., -4.8332e-03, + -2.7084e-03, 8.2636e-04], + [ 1.6844e-04, -1.2070e-02, -4.6959e-03, ..., -7.2632e-03, + -3.9291e-03, -7.9918e-04], + [ 1.3852e-04, 1.4162e-03, 2.3285e-02, ..., 1.2646e-03, + 6.7101e-03, 1.2312e-03]], device='cuda:0') +Epoch 111, bias, value: tensor([ 0.0173, -0.0213, -0.0058, 0.0056, -0.0236, -0.0096, 0.0147, 0.0016, + 0.0271, -0.0217], device='cuda:0'), grad: tensor([-0.0013, 0.0339, 0.0395, 0.0077, -0.0199, -0.0366, -0.0138, -0.0174, + -0.0429, 0.0508], device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 219.49, cls_loss 0.5507 cls_loss_mapping 0.0162 cls_loss_causal 0.5231 re_mapping 0.0110 re_causal 0.0290 /// teacc 98.39 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.0464, -0.0473, -0.0988, ..., 0.0936, -0.0264, 0.0272], + [-0.0646, 0.0789, -0.0715, ..., 0.0289, -0.0380, -0.0740], + [-0.0303, 0.0626, -0.0790, ..., -0.0326, 0.0158, 0.0905], + ..., + [-0.0055, 0.0394, -0.0078, ..., 0.0063, 0.0222, -0.0301], + [ 0.0661, -0.0667, 0.0232, ..., -0.0061, 0.0039, -0.0403], + [-0.0172, -0.0594, 0.0595, ..., -0.0352, 0.0152, -0.0107]], + device='cuda:0'), grad: tensor([[-0.0021, 0.0007, -0.0012, ..., -0.0085, -0.0022, -0.0012], + [ 0.0002, 0.0012, 0.0009, ..., 0.0016, 0.0008, 0.0005], + [ 0.0002, -0.0068, 0.0006, ..., -0.0011, -0.0132, -0.0012], + ..., + [ 0.0007, 0.0013, 0.0021, ..., 0.0030, 0.0020, 0.0007], + [-0.0430, 0.0008, -0.0322, ..., 0.0056, 0.0025, 0.0015], + [ 0.0205, -0.0035, 0.0369, ..., -0.0012, 0.0051, 0.0008]], + device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0163, -0.0207, -0.0062, 0.0062, -0.0223, -0.0096, 0.0141, 0.0012, + 0.0263, -0.0211], device='cuda:0'), grad: tensor([-0.0148, 0.0138, -0.0514, 0.0428, -0.0063, -0.0053, -0.0251, 0.0161, + -0.0037, 0.0339], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 111---------------------------------------------------- +epoch 111, time 220.00, cls_loss 0.6039 cls_loss_mapping 0.0166 cls_loss_causal 0.5630 re_mapping 0.0107 re_causal 0.0277 /// teacc 98.86 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.0474, -0.0466, -0.0991, ..., 0.0945, -0.0262, 0.0266], + [-0.0641, 0.0787, -0.0708, ..., 0.0280, -0.0384, -0.0745], + [-0.0307, 0.0634, -0.0798, ..., -0.0321, 0.0151, 0.0926], + ..., + [-0.0041, 0.0389, -0.0077, ..., 0.0064, 0.0228, -0.0311], + [ 0.0666, -0.0673, 0.0227, ..., -0.0056, 0.0032, -0.0404], + [-0.0171, -0.0585, 0.0598, ..., -0.0346, 0.0154, -0.0106]], + device='cuda:0'), grad: tensor([[-6.0606e-04, 4.2462e-04, 6.2704e-04, ..., -6.3992e-04, + -3.0935e-05, 2.6488e-04], + [ 3.9387e-04, -8.3828e-04, -2.2665e-05, ..., -1.0118e-03, + 3.6567e-05, -1.9188e-03], + [-3.4618e-03, -1.5378e-04, 9.6798e-05, ..., -2.3479e-03, + 3.5167e-04, -2.5043e-03], + ..., + [ 1.4603e-04, 1.0002e-04, -2.1374e-04, ..., 4.9448e-04, + 8.6546e-05, 8.3447e-04], + [ 1.1711e-03, -5.4216e-04, -2.0809e-03, ..., -5.8079e-04, + -2.1973e-03, -1.4811e-03], + [-1.7071e-03, 3.1209e-04, -1.9073e-02, ..., 7.7677e-04, + -1.0872e-02, 7.4625e-04]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0170, -0.0211, -0.0059, 0.0056, -0.0222, -0.0089, 0.0134, 0.0012, + 0.0265, -0.0213], device='cuda:0'), grad: tensor([ 0.0061, -0.0231, -0.0208, 0.0061, 0.0350, 0.0104, 0.0159, 0.0088, + -0.0208, -0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 218.02, cls_loss 0.5557 cls_loss_mapping 0.0159 cls_loss_causal 0.5239 re_mapping 0.0114 re_causal 0.0296 /// teacc 98.53 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.0484, -0.0461, -0.1002, ..., 0.0939, -0.0266, 0.0272], + [-0.0626, 0.0792, -0.0702, ..., 0.0289, -0.0384, -0.0749], + [-0.0307, 0.0641, -0.0790, ..., -0.0314, 0.0147, 0.0933], + ..., + [-0.0041, 0.0389, -0.0078, ..., 0.0065, 0.0228, -0.0319], + [ 0.0672, -0.0685, 0.0221, ..., -0.0057, 0.0033, -0.0410], + [-0.0178, -0.0592, 0.0603, ..., -0.0344, 0.0156, -0.0100]], + device='cuda:0'), grad: tensor([[ 4.7851e-04, 3.6645e-04, 2.3234e-04, ..., 9.2125e-04, + 4.7565e-04, 1.4582e-03], + [ 1.0862e-03, 1.9188e-03, 6.9666e-04, ..., 1.2159e-03, + 5.6887e-04, 1.0042e-03], + [ 5.8460e-04, -1.2989e-03, 4.9591e-04, ..., 1.4519e-06, + -3.8624e-04, -1.6136e-03], + ..., + [ 5.9223e-04, -1.6069e-03, 1.9569e-03, ..., -2.9266e-05, + 2.0618e-03, 4.8280e-04], + [-6.1798e-03, 2.3794e-04, 6.0415e-04, ..., 1.3533e-03, + -9.5558e-04, 2.1088e-04], + [-7.2899e-03, -7.8773e-04, -4.2191e-03, ..., 4.7445e-04, + -7.9651e-03, 4.9621e-05]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0160, -0.0207, -0.0056, 0.0060, -0.0222, -0.0092, 0.0130, 0.0008, + 0.0262, -0.0202], device='cuda:0'), grad: tensor([ 0.0191, 0.0142, -0.0243, 0.0297, -0.0044, 0.0063, -0.0206, 0.0084, + 0.0010, -0.0294], device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 218.60, cls_loss 0.5525 cls_loss_mapping 0.0138 cls_loss_causal 0.5171 re_mapping 0.0104 re_causal 0.0277 /// teacc 98.36 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.0478, -0.0467, -0.1013, ..., 0.0941, -0.0274, 0.0267], + [-0.0630, 0.0794, -0.0698, ..., 0.0290, -0.0388, -0.0738], + [-0.0312, 0.0642, -0.0785, ..., -0.0325, 0.0152, 0.0926], + ..., + [-0.0039, 0.0391, -0.0079, ..., 0.0063, 0.0227, -0.0322], + [ 0.0672, -0.0694, 0.0223, ..., -0.0062, 0.0037, -0.0417], + [-0.0163, -0.0590, 0.0603, ..., -0.0343, 0.0156, -0.0102]], + device='cuda:0'), grad: tensor([[-0.0019, 0.0003, 0.0003, ..., 0.0042, 0.0080, 0.0037], + [ 0.0003, 0.0004, 0.0007, ..., -0.0001, 0.0003, 0.0014], + [ 0.0008, 0.0001, 0.0006, ..., -0.0013, -0.0062, -0.0011], + ..., + [ 0.0005, -0.0004, -0.0005, ..., 0.0014, 0.0003, 0.0009], + [-0.0026, -0.0016, -0.0021, ..., -0.0077, -0.0055, -0.0078], + [ 0.0003, 0.0008, 0.0008, ..., -0.0005, 0.0001, -0.0010]], + device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0158, -0.0196, -0.0062, 0.0062, -0.0226, -0.0083, 0.0136, 0.0001, + 0.0259, -0.0206], device='cuda:0'), grad: tensor([ 0.0254, 0.0221, 0.0025, -0.0024, 0.0009, 0.0201, 0.0168, 0.0137, + -0.0833, -0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 217.60, cls_loss 0.5915 cls_loss_mapping 0.0157 cls_loss_causal 0.5536 re_mapping 0.0111 re_causal 0.0306 /// teacc 98.51 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.0478, -0.0478, -0.0996, ..., 0.0939, -0.0278, 0.0263], + [-0.0633, 0.0811, -0.0706, ..., 0.0296, -0.0396, -0.0726], + [-0.0322, 0.0637, -0.0804, ..., -0.0334, 0.0151, 0.0920], + ..., + [-0.0038, 0.0395, -0.0064, ..., 0.0053, 0.0242, -0.0331], + [ 0.0671, -0.0699, 0.0230, ..., -0.0065, 0.0052, -0.0413], + [-0.0160, -0.0594, 0.0595, ..., -0.0340, 0.0152, -0.0097]], + device='cuda:0'), grad: tensor([[ 2.0123e-04, 3.5024e-04, 3.8266e-04, ..., -7.3862e-04, + -2.7676e-03, -1.4267e-03], + [ 1.7357e-04, 7.2975e-03, 1.3924e-03, ..., 3.1677e-02, + 1.1482e-03, 3.7581e-05], + [ 2.9874e-04, -7.1669e-04, 3.2592e-04, ..., 7.5102e-04, + 2.2173e-04, -1.8764e-04], + ..., + [ 1.0080e-03, 1.5049e-03, 3.1967e-03, ..., 1.8501e-03, + 1.3924e-03, 6.1929e-05], + [-2.8062e-04, -3.0384e-03, 8.8549e-04, ..., 1.7977e-03, + 7.1812e-04, 1.6057e-04], + [ 5.9509e-04, 1.1711e-03, 8.1253e-03, ..., -4.5700e-03, + -4.5490e-04, 3.4618e-04]], device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0167, -0.0198, -0.0072, 0.0068, -0.0220, -0.0091, 0.0138, 0.0002, + 0.0254, -0.0205], device='cuda:0'), grad: tensor([-0.0381, 0.0561, 0.0122, 0.0162, -0.0237, -0.0352, 0.0115, 0.0327, + -0.0119, -0.0197], device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 216.89, cls_loss 0.5615 cls_loss_mapping 0.0115 cls_loss_causal 0.5316 re_mapping 0.0106 re_causal 0.0274 /// teacc 98.59 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.0461, -0.0477, -0.0999, ..., 0.0942, -0.0271, 0.0273], + [-0.0635, 0.0803, -0.0694, ..., 0.0292, -0.0395, -0.0730], + [-0.0332, 0.0650, -0.0813, ..., -0.0332, 0.0151, 0.0920], + ..., + [-0.0029, 0.0397, -0.0075, ..., 0.0062, 0.0234, -0.0334], + [ 0.0669, -0.0709, 0.0233, ..., -0.0064, 0.0051, -0.0412], + [-0.0169, -0.0595, 0.0601, ..., -0.0346, 0.0156, -0.0106]], + device='cuda:0'), grad: tensor([[-4.4443e-06, -3.1805e-04, 3.2568e-04, ..., -3.8218e-04, + 5.7030e-04, 2.7642e-06], + [ 1.3983e-04, -4.6253e-04, 3.4785e-04, ..., -4.2963e-04, + 5.7697e-04, 2.7370e-04], + [ 3.5453e-04, -9.5415e-04, 9.7656e-04, ..., -1.2751e-03, + -6.0606e-04, -9.6750e-04], + ..., + [ 1.1194e-04, -8.4829e-04, 1.5993e-03, ..., -1.7631e-04, + 3.5000e-03, -1.0473e-04], + [ 1.3399e-03, 1.1024e-03, 4.7836e-03, ..., 1.0633e-03, + 7.1373e-03, 1.7881e-03], + [ 7.7820e-04, -1.0958e-03, -4.2839e-03, ..., 6.1512e-04, + -1.2413e-02, -4.6611e-04]], device='cuda:0') +Epoch 117, bias, value: tensor([ 1.7310e-02, -1.9773e-02, -7.5355e-03, 6.7748e-03, -2.1709e-02, + -9.6288e-03, 1.3155e-02, 8.1866e-06, 2.6788e-02, -2.1081e-02], + device='cuda:0'), grad: tensor([-0.0159, -0.0260, 0.0144, -0.0087, 0.0328, 0.0167, -0.0082, -0.0081, + 0.0224, -0.0195], device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 217.03, cls_loss 0.5705 cls_loss_mapping 0.0153 cls_loss_causal 0.5484 re_mapping 0.0100 re_causal 0.0270 /// teacc 98.57 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.0461, -0.0493, -0.1014, ..., 0.0936, -0.0277, 0.0273], + [-0.0634, 0.0803, -0.0692, ..., 0.0294, -0.0392, -0.0719], + [-0.0324, 0.0655, -0.0818, ..., -0.0330, 0.0156, 0.0919], + ..., + [-0.0025, 0.0397, -0.0068, ..., 0.0056, 0.0247, -0.0343], + [ 0.0673, -0.0712, 0.0229, ..., -0.0061, 0.0041, -0.0420], + [-0.0172, -0.0598, 0.0605, ..., -0.0345, 0.0158, -0.0103]], + device='cuda:0'), grad: tensor([[ 3.3474e-04, 1.0958e-03, 5.6171e-04, ..., 1.4963e-03, + 3.3474e-03, 2.1496e-03], + [ 1.7929e-04, -1.1581e-04, 7.2658e-05, ..., 6.8855e-04, + 1.0939e-03, -1.5628e-04], + [ 3.3307e-04, -1.7662e-03, -1.8425e-03, ..., -1.0481e-03, + -1.0040e-02, -2.0943e-03], + ..., + [ 2.9826e-04, -3.6507e-03, 4.7278e-04, ..., -2.0370e-03, + -9.9850e-04, 1.2022e-04], + [ 2.9469e-04, 5.2452e-04, 2.7013e-04, ..., 9.7942e-04, + 1.3428e-03, 7.6580e-04], + [ 4.0627e-04, 7.6866e-04, -1.6785e-02, ..., 1.7986e-03, + -4.6425e-03, 8.2254e-04]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0172, -0.0199, -0.0077, 0.0075, -0.0223, -0.0099, 0.0141, 0.0003, + 0.0259, -0.0209], device='cuda:0'), grad: tensor([ 0.0361, -0.0087, -0.0578, 0.0135, 0.0128, -0.0298, 0.0089, 0.0017, + 0.0136, 0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 216.75, cls_loss 0.5543 cls_loss_mapping 0.0161 cls_loss_causal 0.5203 re_mapping 0.0100 re_causal 0.0269 /// teacc 98.74 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.0460, -0.0482, -0.1009, ..., 0.0958, -0.0272, 0.0272], + [-0.0631, 0.0806, -0.0688, ..., 0.0293, -0.0390, -0.0713], + [-0.0317, 0.0648, -0.0820, ..., -0.0341, 0.0161, 0.0924], + ..., + [-0.0030, 0.0403, -0.0071, ..., 0.0053, 0.0245, -0.0341], + [ 0.0681, -0.0717, 0.0228, ..., -0.0064, 0.0047, -0.0424], + [-0.0185, -0.0600, 0.0610, ..., -0.0348, 0.0160, -0.0109]], + device='cuda:0'), grad: tensor([[-3.9387e-04, -2.3670e-03, 8.3566e-05, ..., -2.8915e-03, + -1.1339e-03, -2.9278e-04], + [ 6.9380e-04, 2.3407e-02, 8.4543e-04, ..., 6.1569e-03, + 8.4305e-04, 1.1671e-04], + [ 2.2995e-04, 1.5144e-03, 1.3649e-04, ..., 4.1270e-04, + 2.0337e-04, 2.0027e-04], + ..., + [-9.4461e-04, -3.4676e-03, -6.5279e-04, ..., 6.2943e-04, + -4.8733e-04, -2.6059e-04], + [ 1.8978e-03, -2.0432e-02, 2.1095e-03, ..., -3.0384e-03, + 1.4820e-03, 6.5660e-04], + [ 3.7460e-03, 4.6515e-04, 6.6185e-03, ..., 4.9877e-04, + 3.3684e-03, 4.8608e-05]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0185, -0.0190, -0.0079, 0.0075, -0.0223, -0.0102, 0.0133, -0.0002, + 0.0259, -0.0213], device='cuda:0'), grad: tensor([-0.0484, 0.0712, 0.0105, 0.0132, -0.0341, -0.0246, -0.0125, 0.0046, + 0.0015, 0.0186], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 118---------------------------------------------------- +epoch 118, time 217.67, cls_loss 0.5684 cls_loss_mapping 0.0156 cls_loss_causal 0.5293 re_mapping 0.0107 re_causal 0.0287 /// teacc 98.88 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.0464, -0.0481, -0.1011, ..., 0.0968, -0.0271, 0.0276], + [-0.0635, 0.0807, -0.0683, ..., 0.0292, -0.0391, -0.0708], + [-0.0320, 0.0656, -0.0818, ..., -0.0328, 0.0161, 0.0936], + ..., + [-0.0038, 0.0410, -0.0076, ..., 0.0050, 0.0241, -0.0345], + [ 0.0689, -0.0716, 0.0232, ..., -0.0061, 0.0049, -0.0431], + [-0.0180, -0.0596, 0.0606, ..., -0.0348, 0.0159, -0.0107]], + device='cuda:0'), grad: tensor([[ 2.2066e-04, -2.5101e-03, -4.7565e-04, ..., -5.8060e-03, + -2.8992e-03, 4.1461e-04], + [ 1.7347e-03, 4.4823e-03, 5.0068e-04, ..., 6.1264e-03, + 3.2425e-03, 6.0797e-04], + [-4.8523e-03, -4.3297e-03, -4.3344e-04, ..., -7.0152e-03, + -4.7836e-03, -3.1929e-03], + ..., + [-4.6349e-03, -7.1564e-03, -1.4885e-02, ..., -3.0823e-03, + -9.7046e-03, 3.4761e-04], + [ 6.4707e-04, 1.8787e-03, 8.5449e-04, ..., 1.0557e-03, + 1.1196e-03, -5.3525e-05], + [ 3.1071e-03, 3.4294e-03, 1.0498e-02, ..., 3.1662e-03, + 6.9008e-03, 1.7583e-04]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0193, -0.0189, -0.0076, 0.0073, -0.0219, -0.0111, 0.0128, -0.0004, + 0.0254, -0.0206], device='cuda:0'), grad: tensor([-0.0094, 0.0388, -0.0649, 0.0124, 0.0090, -0.0231, 0.0403, -0.0208, + -0.0103, 0.0280], device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 217.00, cls_loss 0.5833 cls_loss_mapping 0.0214 cls_loss_causal 0.5512 re_mapping 0.0102 re_causal 0.0267 /// teacc 98.58 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.0471, -0.0483, -0.1028, ..., 0.0966, -0.0290, 0.0266], + [-0.0651, 0.0802, -0.0672, ..., 0.0283, -0.0405, -0.0717], + [-0.0318, 0.0655, -0.0815, ..., -0.0326, 0.0168, 0.0934], + ..., + [-0.0031, 0.0411, -0.0076, ..., 0.0055, 0.0255, -0.0346], + [ 0.0682, -0.0713, 0.0228, ..., -0.0061, 0.0043, -0.0427], + [-0.0186, -0.0590, 0.0607, ..., -0.0346, 0.0162, -0.0103]], + device='cuda:0'), grad: tensor([[ 2.7037e-04, 1.0180e-04, 9.7156e-05, ..., 2.0752e-03, + -1.1234e-03, -1.4410e-03], + [-2.2354e-03, -3.2353e-04, 1.6165e-04, ..., -7.2670e-03, + 4.8089e-04, 8.6784e-05], + [ 2.1350e-04, 2.0158e-04, 3.9053e-04, ..., 4.2343e-04, + 1.7052e-03, 9.5272e-04], + ..., + [ 3.1978e-05, 1.9836e-04, 1.0328e-03, ..., 5.6171e-04, + 1.1873e-03, 1.7250e-04], + [ 4.2415e-04, -8.5831e-04, -2.2564e-03, ..., 8.2016e-04, + -3.7556e-03, -2.6846e-04], + [ 1.3995e-04, 7.1287e-05, 8.7738e-04, ..., 2.7394e-04, + 4.6635e-04, 1.5903e-04]], device='cuda:0') +Epoch 121, bias, value: tensor([ 0.0180, -0.0202, -0.0077, 0.0081, -0.0215, -0.0111, 0.0138, 0.0004, + 0.0249, -0.0203], device='cuda:0'), grad: tensor([ 0.0038, -0.0013, 0.0169, -0.0106, 0.0139, -0.0024, 0.0197, 0.0179, + -0.0410, -0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 216.81, cls_loss 0.5676 cls_loss_mapping 0.0218 cls_loss_causal 0.5380 re_mapping 0.0101 re_causal 0.0255 /// teacc 98.67 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.0483, -0.0477, -0.1037, ..., 0.0972, -0.0303, 0.0267], + [-0.0652, 0.0813, -0.0673, ..., 0.0287, -0.0412, -0.0718], + [-0.0314, 0.0649, -0.0832, ..., -0.0325, 0.0170, 0.0937], + ..., + [-0.0034, 0.0414, -0.0068, ..., 0.0050, 0.0254, -0.0336], + [ 0.0688, -0.0724, 0.0231, ..., -0.0065, 0.0045, -0.0430], + [-0.0189, -0.0598, 0.0602, ..., -0.0347, 0.0162, -0.0116]], + device='cuda:0'), grad: tensor([[ 2.6003e-05, 3.4261e-04, 1.4015e-05, ..., 6.1083e-04, + 1.7631e-04, 6.2764e-05], + [ 5.3674e-05, 7.9727e-04, 8.3089e-05, ..., 3.8505e-04, + 2.2876e-04, 3.5024e-04], + [ 9.2089e-05, -3.9101e-03, 6.1750e-05, ..., -3.0756e-04, + -1.9226e-03, -1.0366e-03], + ..., + [ 1.0765e-02, 7.1764e-04, 1.2489e-02, ..., -4.9782e-04, + 1.3603e-02, 2.0599e-04], + [ 3.5954e-04, 3.6550e-04, 4.7135e-04, ..., 5.1689e-04, + 4.8661e-04, 1.1712e-04], + [-1.1879e-02, -1.0496e-04, -1.4427e-02, ..., -6.6900e-04, + -1.5205e-02, 3.2067e-05]], device='cuda:0') +Epoch 122, bias, value: tensor([ 0.0174, -0.0199, -0.0081, 0.0076, -0.0222, -0.0097, 0.0139, 0.0009, + 0.0247, -0.0201], device='cuda:0'), grad: tensor([ 0.0136, -0.0055, -0.0144, 0.0119, 0.0169, 0.0107, -0.0128, 0.0057, + 0.0146, -0.0407], device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 217.22, cls_loss 0.5583 cls_loss_mapping 0.0120 cls_loss_causal 0.5324 re_mapping 0.0108 re_causal 0.0280 /// teacc 98.77 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.0482, -0.0475, -0.1050, ..., 0.0974, -0.0300, 0.0263], + [-0.0654, 0.0814, -0.0681, ..., 0.0296, -0.0419, -0.0722], + [-0.0313, 0.0653, -0.0841, ..., -0.0321, 0.0173, 0.0943], + ..., + [-0.0044, 0.0412, -0.0062, ..., 0.0045, 0.0256, -0.0346], + [ 0.0687, -0.0718, 0.0228, ..., -0.0067, 0.0047, -0.0428], + [-0.0184, -0.0610, 0.0601, ..., -0.0356, 0.0157, -0.0128]], + device='cuda:0'), grad: tensor([[ 0.0001, 0.0004, 0.0001, ..., 0.0017, 0.0006, 0.0013], + [ 0.0003, 0.0005, 0.0005, ..., 0.0022, 0.0006, 0.0009], + [-0.0009, 0.0017, 0.0007, ..., -0.0079, 0.0132, 0.0115], + ..., + [-0.0004, -0.0006, -0.0030, ..., 0.0008, -0.0070, 0.0005], + [ 0.0002, -0.0032, 0.0005, ..., 0.0013, -0.0147, -0.0179], + [ 0.0001, 0.0003, 0.0005, ..., 0.0007, 0.0009, 0.0002]], + device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0169, -0.0191, -0.0080, 0.0074, -0.0217, -0.0097, 0.0143, 0.0007, + 0.0243, -0.0208], device='cuda:0'), grad: tensor([ 0.0174, 0.0251, -0.0504, -0.0121, -0.0122, -0.0075, 0.0212, 0.0123, + -0.0055, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 216.85, cls_loss 0.5577 cls_loss_mapping 0.0124 cls_loss_causal 0.5184 re_mapping 0.0109 re_causal 0.0287 /// teacc 98.60 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.0475, -0.0478, -0.1049, ..., 0.0986, -0.0294, 0.0257], + [-0.0669, 0.0804, -0.0687, ..., 0.0295, -0.0421, -0.0734], + [-0.0313, 0.0661, -0.0835, ..., -0.0317, 0.0168, 0.0946], + ..., + [-0.0044, 0.0411, -0.0065, ..., 0.0044, 0.0254, -0.0356], + [ 0.0688, -0.0732, 0.0223, ..., -0.0072, 0.0048, -0.0423], + [-0.0174, -0.0611, 0.0609, ..., -0.0365, 0.0163, -0.0123]], + device='cuda:0'), grad: tensor([[ 0.0017, 0.0004, 0.0042, ..., 0.0033, 0.0049, 0.0014], + [ 0.0007, 0.0002, 0.0016, ..., 0.0009, 0.0008, 0.0002], + [ 0.0012, 0.0003, 0.0003, ..., 0.0013, 0.0018, 0.0012], + ..., + [-0.0037, -0.0016, 0.0020, ..., 0.0012, 0.0022, 0.0002], + [ 0.0013, 0.0005, 0.0005, ..., 0.0017, 0.0008, 0.0003], + [ 0.0015, 0.0012, 0.0068, ..., 0.0015, 0.0041, 0.0002]], + device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0179, -0.0201, -0.0069, 0.0071, -0.0221, -0.0100, 0.0143, 0.0004, + 0.0240, -0.0202], device='cuda:0'), grad: tensor([ 0.0322, -0.0078, 0.0188, -0.0568, -0.0331, -0.0122, 0.0146, -0.0020, + 0.0192, 0.0269], device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 216.70, cls_loss 0.5534 cls_loss_mapping 0.0152 cls_loss_causal 0.5215 re_mapping 0.0106 re_causal 0.0263 /// teacc 98.65 lr 0.00010000 +Epoch 125, weight, value: tensor([[-0.0482, -0.0486, -0.1064, ..., 0.0989, -0.0299, 0.0265], + [-0.0657, 0.0819, -0.0685, ..., 0.0294, -0.0421, -0.0729], + [-0.0323, 0.0653, -0.0847, ..., -0.0323, 0.0167, 0.0945], + ..., + [-0.0043, 0.0418, -0.0065, ..., 0.0032, 0.0257, -0.0364], + [ 0.0684, -0.0738, 0.0230, ..., -0.0065, 0.0056, -0.0424], + [-0.0178, -0.0622, 0.0612, ..., -0.0369, 0.0167, -0.0117]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0010, -0.0025, ..., -0.0065, -0.0049, -0.0021], + [-0.0023, 0.0020, 0.0019, ..., -0.0078, 0.0011, -0.0011], + [ 0.0004, -0.0101, -0.0072, ..., -0.0061, -0.0014, 0.0003], + ..., + [ 0.0009, 0.0036, -0.0222, ..., 0.0036, -0.0370, 0.0007], + [ 0.0032, 0.0007, 0.0020, ..., 0.0025, 0.0045, 0.0012], + [-0.0025, -0.0007, 0.0229, ..., -0.0013, 0.0336, 0.0009]], + device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0173, -0.0201, -0.0069, 0.0074, -0.0227, -0.0095, 0.0157, -0.0008, + 0.0242, -0.0202], device='cuda:0'), grad: tensor([ 0.0002, -0.0275, -0.0408, 0.0057, 0.0223, 0.0356, 0.0355, -0.0198, + 0.0078, -0.0188], device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 216.89, cls_loss 0.5570 cls_loss_mapping 0.0161 cls_loss_causal 0.5349 re_mapping 0.0099 re_causal 0.0261 /// teacc 98.70 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.0485, -0.0489, -0.1070, ..., 0.0995, -0.0303, 0.0263], + [-0.0657, 0.0813, -0.0678, ..., 0.0299, -0.0411, -0.0735], + [-0.0328, 0.0666, -0.0849, ..., -0.0326, 0.0180, 0.0951], + ..., + [-0.0021, 0.0410, -0.0054, ..., 0.0024, 0.0264, -0.0373], + [ 0.0672, -0.0728, 0.0227, ..., -0.0062, 0.0048, -0.0423], + [-0.0184, -0.0623, 0.0617, ..., -0.0383, 0.0172, -0.0108]], + device='cuda:0'), grad: tensor([[ 5.2643e-04, 2.7394e-04, 4.1580e-04, ..., 3.2673e-03, + 1.8730e-03, 4.0793e-04], + [ 9.4938e-04, -3.0766e-03, -6.6662e-04, ..., -1.1597e-03, + 1.9419e-04, 8.5711e-05], + [ 4.6120e-03, 1.5936e-03, -5.0116e-04, ..., 1.3714e-03, + -6.0158e-03, -2.3937e-03], + ..., + [ 3.8528e-04, -2.6488e-04, -2.6970e-03, ..., -5.1460e-03, + 1.1282e-03, 6.0654e-04], + [ 1.6832e-03, 9.7752e-04, 2.2316e-03, ..., -8.1558e-03, + -1.9007e-03, 5.0640e-04], + [ 3.7718e-04, 1.6041e-03, -7.5197e-04, ..., 2.0809e-03, + 1.7385e-03, 9.0408e-04]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0169, -0.0195, -0.0066, 0.0078, -0.0227, -0.0093, 0.0155, -0.0008, + 0.0241, -0.0208], device='cuda:0'), grad: tensor([ 0.0210, -0.0037, 0.0074, 0.0228, 0.0091, -0.0024, -0.0301, -0.0226, + -0.0170, 0.0155], device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 216.82, cls_loss 0.5466 cls_loss_mapping 0.0164 cls_loss_causal 0.5177 re_mapping 0.0103 re_causal 0.0268 /// teacc 98.61 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.0500, -0.0487, -0.1071, ..., 0.1001, -0.0306, 0.0268], + [-0.0650, 0.0812, -0.0679, ..., 0.0291, -0.0406, -0.0740], + [-0.0327, 0.0674, -0.0833, ..., -0.0318, 0.0187, 0.0958], + ..., + [-0.0022, 0.0416, -0.0065, ..., 0.0033, 0.0248, -0.0376], + [ 0.0679, -0.0733, 0.0227, ..., -0.0059, 0.0051, -0.0421], + [-0.0183, -0.0634, 0.0617, ..., -0.0399, 0.0178, -0.0117]], + device='cuda:0'), grad: tensor([[-2.5705e-05, 6.0892e-04, 1.2171e-04, ..., 7.7635e-06, + -2.6321e-04, -2.8586e-04], + [ 1.3048e-06, -4.7112e-03, -2.4104e-04, ..., -3.2310e-03, + 5.5343e-05, 1.9240e-04], + [ 3.7956e-04, 5.9271e-04, -1.0328e-03, ..., 9.0866e-03, + 1.3437e-03, 6.4163e-03], + ..., + [-4.2772e-04, 6.1340e-03, 2.3537e-03, ..., -2.6059e-04, + 3.8605e-03, 2.3975e-03], + [ 3.8087e-05, 8.4162e-04, 2.8849e-04, ..., 9.0885e-04, + 4.4394e-04, 5.5218e-04], + [-2.4036e-05, 9.5558e-04, 3.5572e-04, ..., 3.5644e-04, + 3.0518e-04, 2.2960e-04]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0168, -0.0191, -0.0071, 0.0077, -0.0220, -0.0102, 0.0160, -0.0005, + 0.0245, -0.0216], device='cuda:0'), grad: tensor([ 0.0020, -0.0067, 0.0287, -0.0288, 0.0037, 0.0042, 0.0038, -0.0160, + 0.0054, 0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 216.68, cls_loss 0.5993 cls_loss_mapping 0.0116 cls_loss_causal 0.5639 re_mapping 0.0107 re_causal 0.0280 /// teacc 98.57 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.0496, -0.0477, -0.1086, ..., 0.1003, -0.0310, 0.0279], + [-0.0656, 0.0827, -0.0678, ..., 0.0282, -0.0408, -0.0745], + [-0.0327, 0.0668, -0.0826, ..., -0.0321, 0.0189, 0.0956], + ..., + [-0.0027, 0.0407, -0.0062, ..., 0.0019, 0.0250, -0.0387], + [ 0.0680, -0.0737, 0.0234, ..., -0.0043, 0.0058, -0.0428], + [-0.0180, -0.0624, 0.0618, ..., -0.0387, 0.0178, -0.0115]], + device='cuda:0'), grad: tensor([[ 5.0291e-06, 4.3702e-04, 1.8120e-04, ..., 1.2674e-03, + 4.9019e-04, 4.1890e-04], + [ 3.3703e-03, 1.1292e-03, 7.5579e-04, ..., -1.7653e-03, + 4.1795e-04, 2.8515e-04], + [ 1.1086e-04, -1.8573e-04, -2.9125e-03, ..., -7.3586e-03, + -3.7079e-03, -5.7297e-03], + ..., + [-3.8700e-03, -7.8201e-03, 1.1501e-03, ..., -3.9005e-04, + -6.4087e-03, 9.9182e-04], + [ 1.4019e-04, 9.6750e-04, 1.5850e-03, ..., 4.6806e-03, + 2.8458e-03, 1.5802e-03], + [-1.4582e-03, 6.1560e-04, -7.4291e-04, ..., -1.8549e-04, + -5.3501e-04, 4.3440e-04]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0171, -0.0199, -0.0076, 0.0078, -0.0210, -0.0103, 0.0156, -0.0011, + 0.0249, -0.0213], device='cuda:0'), grad: tensor([ 0.0197, 0.0108, -0.0080, 0.0395, -0.0086, -0.0106, 0.0376, -0.0732, + 0.0077, -0.0149], device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 217.19, cls_loss 0.5575 cls_loss_mapping 0.0111 cls_loss_causal 0.5281 re_mapping 0.0103 re_causal 0.0271 /// teacc 98.62 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.0501, -0.0486, -0.1082, ..., 0.1001, -0.0314, 0.0283], + [-0.0647, 0.0835, -0.0687, ..., 0.0290, -0.0407, -0.0748], + [-0.0334, 0.0665, -0.0822, ..., -0.0320, 0.0201, 0.0962], + ..., + [-0.0030, 0.0418, -0.0061, ..., 0.0021, 0.0250, -0.0383], + [ 0.0681, -0.0741, 0.0244, ..., -0.0040, 0.0069, -0.0436], + [-0.0185, -0.0629, 0.0623, ..., -0.0385, 0.0172, -0.0109]], + device='cuda:0'), grad: tensor([[ 7.3433e-05, -2.9936e-05, 2.5177e-03, ..., -7.0620e-04, + 9.6560e-05, -1.1215e-03], + [ 4.9740e-05, 1.4961e-04, 5.4455e-04, ..., 4.3774e-04, + 3.9124e-04, 1.8418e-04], + [ 7.4506e-05, -4.7112e-04, 4.0841e-04, ..., 8.7786e-04, + 6.7472e-04, 4.2415e-04], + ..., + [ 5.2452e-04, 2.3234e-04, 1.8024e-03, ..., 5.1069e-04, + 2.0161e-03, 3.3593e-04], + [ 1.6248e-04, 1.4925e-04, -1.5915e-02, ..., 7.9823e-04, + -9.4681e-03, 3.8671e-04], + [-2.2793e-03, -1.2946e-04, 3.1757e-03, ..., -9.6941e-04, + 7.7486e-04, 7.2360e-05]], device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0171, -0.0190, -0.0084, 0.0069, -0.0212, -0.0102, 0.0159, -0.0008, + 0.0252, -0.0208], device='cuda:0'), grad: tensor([ 0.0111, 0.0069, 0.0102, 0.0124, 0.0172, 0.0352, -0.0503, 0.0161, + -0.0176, -0.0410], device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 216.91, cls_loss 0.5498 cls_loss_mapping 0.0157 cls_loss_causal 0.5177 re_mapping 0.0100 re_causal 0.0253 /// teacc 98.72 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.0505, -0.0483, -0.1090, ..., 0.0994, -0.0318, 0.0278], + [-0.0659, 0.0833, -0.0694, ..., 0.0292, -0.0416, -0.0757], + [-0.0336, 0.0673, -0.0823, ..., -0.0311, 0.0188, 0.0968], + ..., + [-0.0025, 0.0401, -0.0054, ..., 0.0014, 0.0264, -0.0377], + [ 0.0676, -0.0713, 0.0228, ..., -0.0039, 0.0055, -0.0441], + [-0.0185, -0.0625, 0.0628, ..., -0.0392, 0.0172, -0.0119]], + device='cuda:0'), grad: tensor([[ 4.5872e-04, 2.1672e-04, -4.2796e-05, ..., 2.0752e-03, + -2.4700e-04, 2.2888e-04], + [ 2.5845e-04, 6.6614e-04, 3.0017e-04, ..., 1.9140e-03, + 4.5371e-04, 3.1948e-04], + [ 5.1928e-04, 5.1994e-03, 1.8911e-03, ..., 7.1182e-03, + 1.4313e-02, 7.8888e-03], + ..., + [ 5.8460e-04, -4.9438e-03, -4.6844e-03, ..., -1.5860e-03, + -3.6011e-03, -1.9255e-03], + [ 1.8501e-03, -2.2278e-03, -4.1771e-03, ..., -6.1893e-04, + -7.2479e-03, -5.0278e-03], + [ 6.7234e-04, 2.7733e-03, 4.5280e-03, ..., -5.1260e-04, + 3.3054e-03, 3.9053e-04]], device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0166, -0.0190, -0.0083, 0.0070, -0.0223, -0.0103, 0.0163, -0.0011, + 0.0254, -0.0200], device='cuda:0'), grad: tensor([-0.0100, 0.0055, 0.0462, -0.0284, -0.0107, 0.0295, -0.0039, 0.0096, + -0.0377, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 216.91, cls_loss 0.5745 cls_loss_mapping 0.0116 cls_loss_causal 0.5509 re_mapping 0.0104 re_causal 0.0281 /// teacc 98.77 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.0508, -0.0497, -0.1100, ..., 0.0999, -0.0327, 0.0278], + [-0.0662, 0.0839, -0.0701, ..., 0.0294, -0.0419, -0.0755], + [-0.0349, 0.0669, -0.0818, ..., -0.0318, 0.0183, 0.0964], + ..., + [-0.0016, 0.0411, -0.0051, ..., 0.0013, 0.0271, -0.0381], + [ 0.0673, -0.0712, 0.0235, ..., -0.0042, 0.0060, -0.0439], + [-0.0189, -0.0633, 0.0620, ..., -0.0395, 0.0165, -0.0123]], + device='cuda:0'), grad: tensor([[ 9.3356e-06, 3.7622e-04, 2.2924e-04, ..., -2.4378e-04, + 1.1349e-03, 1.2529e-04], + [-1.4699e-04, 1.1921e-03, 1.1015e-04, ..., 1.8959e-03, + 2.4624e-03, 8.6641e-04], + [ 3.1084e-05, 1.2941e-03, 2.1911e-04, ..., 3.5477e-03, + 2.1629e-03, 6.9427e-04], + ..., + [ 2.2739e-05, -8.8072e-04, -2.9316e-03, ..., 4.0793e-04, + -2.6340e-03, -2.2128e-05], + [ 3.0205e-05, 7.9250e-04, 4.7922e-04, ..., -1.0666e-02, + -2.7962e-03, -8.3828e-04], + [ 1.3903e-05, 1.6756e-03, 1.4811e-03, ..., 4.2343e-03, + 3.3741e-03, 7.1955e-04]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0173, -0.0195, -0.0094, 0.0078, -0.0229, -0.0108, 0.0176, -0.0011, + 0.0252, -0.0200], device='cuda:0'), grad: tensor([-1.0948e-02, 2.0355e-02, 2.4246e-02, 8.6823e-03, 1.3245e-02, + -5.1453e-02, -9.9564e-03, -4.9829e-05, -2.5497e-02, 3.1372e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 216.87, cls_loss 0.5722 cls_loss_mapping 0.0118 cls_loss_causal 0.5246 re_mapping 0.0102 re_causal 0.0269 /// teacc 98.77 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.0500, -0.0501, -0.1113, ..., 0.1005, -0.0317, 0.0279], + [-0.0669, 0.0851, -0.0704, ..., 0.0292, -0.0428, -0.0745], + [-0.0353, 0.0659, -0.0812, ..., -0.0331, 0.0182, 0.0957], + ..., + [-0.0015, 0.0419, -0.0054, ..., 0.0018, 0.0273, -0.0381], + [ 0.0669, -0.0712, 0.0224, ..., -0.0030, 0.0048, -0.0427], + [-0.0188, -0.0635, 0.0628, ..., -0.0399, 0.0172, -0.0133]], + device='cuda:0'), grad: tensor([[ 5.0254e-06, 3.4213e-04, -7.1907e-04, ..., -2.5673e-03, + -4.4674e-05, 4.5061e-05], + [ 7.7784e-06, -1.4105e-03, 3.9148e-04, ..., -4.7607e-03, + -4.1556e-04, 2.2411e-05], + [ 3.7774e-06, -2.9540e-04, 1.3065e-04, ..., 8.2350e-04, + -1.1787e-02, -8.3351e-04], + ..., + [ 4.2021e-06, 2.5368e-03, 3.0041e-03, ..., 5.4550e-03, + 4.1199e-03, 2.1303e-04], + [ 5.9754e-06, -7.5388e-04, -3.3131e-03, ..., -1.6613e-03, + 4.3416e-04, 1.6439e-04], + [ 5.3197e-06, 4.8676e-03, 1.6876e-02, ..., 5.1193e-03, + 1.0468e-02, 1.1779e-05]], device='cuda:0') +Epoch 132, bias, value: tensor([ 0.0174, -0.0195, -0.0102, 0.0081, -0.0229, -0.0105, 0.0172, -0.0008, + 0.0253, -0.0197], device='cuda:0'), grad: tensor([-0.0179, -0.0134, -0.0164, -0.0046, -0.0159, -0.0101, -0.0175, 0.0393, + 0.0059, 0.0508], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 131---------------------------------------------------- +epoch 131, time 217.54, cls_loss 0.5743 cls_loss_mapping 0.0126 cls_loss_causal 0.5378 re_mapping 0.0098 re_causal 0.0272 /// teacc 98.89 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.0491, -0.0498, -0.1118, ..., 0.1007, -0.0322, 0.0284], + [-0.0655, 0.0859, -0.0708, ..., 0.0294, -0.0440, -0.0746], + [-0.0365, 0.0659, -0.0813, ..., -0.0330, 0.0193, 0.0969], + ..., + [-0.0018, 0.0419, -0.0055, ..., 0.0010, 0.0265, -0.0388], + [ 0.0672, -0.0730, 0.0228, ..., -0.0033, 0.0057, -0.0428], + [-0.0181, -0.0624, 0.0638, ..., -0.0396, 0.0175, -0.0134]], + device='cuda:0'), grad: tensor([[ 3.1853e-04, 5.3978e-04, 5.9032e-04, ..., 4.4632e-03, + 1.8539e-03, 2.6741e-03], + [ 2.1204e-05, -1.6537e-03, 1.7321e-04, ..., -3.1891e-03, + 4.5705e-04, -3.8300e-03], + [ 9.5189e-05, 4.1080e-04, -6.4516e-04, ..., 1.0061e-03, + -8.9455e-04, -1.6880e-04], + ..., + [-1.1883e-03, -3.2368e-03, -3.2043e-03, ..., -1.1345e-02, + -6.2103e-03, -6.5269e-03], + [ 7.8022e-05, 4.5824e-04, 2.5773e-04, ..., 2.6493e-03, + 8.3303e-04, 1.3962e-03], + [ 4.5753e-04, 9.3842e-04, 1.4744e-03, ..., 3.8280e-03, + 2.0294e-03, 1.6537e-03]], device='cuda:0') +Epoch 133, bias, value: tensor([ 0.0175, -0.0195, -0.0094, 0.0081, -0.0229, -0.0107, 0.0172, -0.0015, + 0.0248, -0.0193], device='cuda:0'), grad: tensor([ 0.0373, -0.0663, 0.0068, -0.0260, -0.0299, 0.0336, 0.0360, -0.0598, + 0.0318, 0.0366], device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 217.17, cls_loss 0.5675 cls_loss_mapping 0.0115 cls_loss_causal 0.5351 re_mapping 0.0099 re_causal 0.0261 /// teacc 98.63 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.0495, -0.0502, -0.1122, ..., 0.1009, -0.0320, 0.0291], + [-0.0656, 0.0869, -0.0703, ..., 0.0296, -0.0445, -0.0742], + [-0.0373, 0.0670, -0.0823, ..., -0.0319, 0.0195, 0.0984], + ..., + [-0.0013, 0.0415, -0.0056, ..., 0.0015, 0.0267, -0.0385], + [ 0.0678, -0.0734, 0.0254, ..., -0.0040, 0.0070, -0.0435], + [-0.0182, -0.0625, 0.0630, ..., -0.0405, 0.0170, -0.0140]], + device='cuda:0'), grad: tensor([[ 5.5981e-04, 1.1826e-03, 2.2316e-03, ..., 3.0422e-03, + 9.9792e-03, 8.3017e-04], + [ 1.8775e-04, 1.3476e-06, 8.8990e-05, ..., 1.3475e-03, + 4.8804e-04, 7.4148e-04], + [ 1.1034e-05, 8.2445e-04, 6.2823e-05, ..., 1.8291e-03, + 1.6232e-03, 8.3923e-04], + ..., + [-7.5006e-04, -5.6114e-03, -5.0163e-04, ..., -5.7106e-03, + -1.6108e-03, -3.2558e-03], + [ 1.0324e-04, 7.6628e-04, 8.9407e-04, ..., 1.6565e-03, + 2.3746e-03, 7.7343e-04], + [ 4.6730e-04, 6.1131e-04, 1.1787e-03, ..., 1.3227e-03, + 2.8248e-03, 4.1771e-04]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0173, -0.0198, -0.0090, 0.0083, -0.0224, -0.0109, 0.0168, -0.0009, + 0.0243, -0.0195], device='cuda:0'), grad: tensor([ 0.0356, 0.0085, 0.0242, 0.0322, -0.0216, -0.0288, 0.0171, -0.1027, + 0.0212, 0.0144], device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 216.46, cls_loss 0.5432 cls_loss_mapping 0.0146 cls_loss_causal 0.5058 re_mapping 0.0110 re_causal 0.0287 /// teacc 98.72 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.0486, -0.0502, -0.1135, ..., 0.1008, -0.0332, 0.0285], + [-0.0655, 0.0867, -0.0698, ..., 0.0293, -0.0448, -0.0742], + [-0.0389, 0.0672, -0.0827, ..., -0.0315, 0.0185, 0.0986], + ..., + [-0.0014, 0.0417, -0.0066, ..., 0.0026, 0.0263, -0.0389], + [ 0.0672, -0.0728, 0.0252, ..., -0.0025, 0.0061, -0.0425], + [-0.0177, -0.0619, 0.0636, ..., -0.0401, 0.0182, -0.0124]], + device='cuda:0'), grad: tensor([[ 6.1572e-05, 1.0324e-04, 2.9278e-04, ..., -1.8227e-04, + -3.6430e-03, -1.0757e-03], + [-4.3660e-05, -3.4481e-05, 7.0667e-04, ..., -4.2305e-03, + 3.2902e-04, 4.2140e-05], + [-2.1572e-03, 7.7784e-05, 3.0589e-04, ..., -4.1542e-03, + -7.2212e-03, 6.7854e-04], + ..., + [ 5.2691e-04, 1.1009e-04, 9.6273e-04, ..., 2.1477e-03, + 2.2545e-03, 1.1688e-04], + [ 9.1219e-04, 1.1426e-04, -5.3215e-04, ..., 4.4632e-04, + 3.3569e-03, 4.3213e-05], + [-4.6062e-04, 7.5281e-05, -1.3641e-02, ..., 1.8463e-03, + -1.1387e-03, 2.8682e-04]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0170, -0.0201, -0.0088, 0.0078, -0.0235, -0.0117, 0.0177, -0.0006, + 0.0246, -0.0181], device='cuda:0'), grad: tensor([-0.0491, 0.0017, -0.0079, -0.0061, 0.0436, -0.0112, 0.0247, 0.0223, + -0.0028, -0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 216.77, cls_loss 0.5426 cls_loss_mapping 0.0097 cls_loss_causal 0.5159 re_mapping 0.0100 re_causal 0.0258 /// teacc 98.81 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.0509, -0.0505, -0.1136, ..., 0.0993, -0.0338, 0.0282], + [-0.0658, 0.0863, -0.0695, ..., 0.0303, -0.0435, -0.0744], + [-0.0377, 0.0673, -0.0829, ..., -0.0313, 0.0199, 0.0997], + ..., + [-0.0013, 0.0420, -0.0060, ..., 0.0021, 0.0262, -0.0400], + [ 0.0691, -0.0719, 0.0253, ..., -0.0014, 0.0069, -0.0419], + [-0.0177, -0.0620, 0.0635, ..., -0.0409, 0.0178, -0.0130]], + device='cuda:0'), grad: tensor([[-5.0354e-03, 8.0049e-05, 1.8224e-05, ..., -8.9874e-03, + -7.6561e-03, 7.8487e-04], + [ 1.5032e-04, -1.4200e-03, -1.1438e-04, ..., 2.7027e-03, + 1.2696e-04, 7.7677e-04], + [ 4.1223e-04, 3.9601e-04, 1.2052e-04, ..., -9.0122e-04, + 1.9491e-04, -4.5891e-03], + ..., + [ 1.4293e-04, 1.2994e-04, 1.3411e-04, ..., -2.7955e-05, + 2.8181e-04, -1.2046e-04], + [ 2.1000e-03, 8.2374e-05, 2.2817e-04, ..., 6.0005e-03, + 3.1929e-03, 1.8559e-03], + [ 3.0637e-04, 1.1045e-04, -7.5293e-04, ..., -1.7233e-03, + 2.1076e-04, -1.4753e-03]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0163, -0.0197, -0.0083, 0.0075, -0.0233, -0.0112, 0.0172, -0.0010, + 0.0255, -0.0189], device='cuda:0'), grad: tensor([-0.0612, 0.0250, 0.0091, 0.0326, -0.0412, 0.0316, 0.0019, 0.0153, + 0.0234, -0.0364], device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 216.71, cls_loss 0.5682 cls_loss_mapping 0.0135 cls_loss_causal 0.5403 re_mapping 0.0098 re_causal 0.0252 /// teacc 98.66 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.0516, -0.0508, -0.1140, ..., 0.1000, -0.0336, 0.0283], + [-0.0660, 0.0864, -0.0684, ..., 0.0292, -0.0427, -0.0751], + [-0.0385, 0.0682, -0.0833, ..., -0.0301, 0.0196, 0.1010], + ..., + [-0.0018, 0.0425, -0.0056, ..., 0.0026, 0.0266, -0.0396], + [ 0.0709, -0.0717, 0.0251, ..., -0.0011, 0.0062, -0.0426], + [-0.0177, -0.0622, 0.0633, ..., -0.0408, 0.0180, -0.0144]], + device='cuda:0'), grad: tensor([[ 2.5272e-04, 2.4962e-04, 6.0654e-04, ..., -6.9809e-04, + 5.0068e-04, 5.5730e-05], + [-3.3226e-03, -8.5402e-04, 1.9586e-04, ..., -8.0795e-03, + 3.2234e-04, 8.6725e-05], + [ 3.1328e-04, 4.1466e-03, 2.3675e-04, ..., 2.9984e-03, + 6.8016e-03, 1.6289e-03], + ..., + [ 6.2418e-04, -7.1716e-03, 4.5061e-04, ..., 9.2077e-04, + -9.4833e-03, -2.8076e-03], + [ 3.9959e-04, 7.5340e-04, 2.5654e-04, ..., 2.8286e-03, + 7.5579e-04, 1.3947e-04], + [-5.4407e-04, 6.9809e-04, 3.3188e-03, ..., 5.1956e-03, + -1.2947e-02, 6.3479e-05]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0166, -0.0208, -0.0092, 0.0070, -0.0223, -0.0115, 0.0171, 0.0007, + 0.0259, -0.0194], device='cuda:0'), grad: tensor([-0.0160, -0.0578, 0.0334, 0.0125, -0.0084, -0.0162, 0.0120, -0.0008, + 0.0202, 0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 216.86, cls_loss 0.5549 cls_loss_mapping 0.0095 cls_loss_causal 0.5266 re_mapping 0.0097 re_causal 0.0259 /// teacc 98.76 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.0510, -0.0522, -0.1153, ..., 0.1005, -0.0348, 0.0285], + [-0.0658, 0.0869, -0.0690, ..., 0.0297, -0.0422, -0.0747], + [-0.0389, 0.0689, -0.0839, ..., -0.0300, 0.0191, 0.1019], + ..., + [-0.0021, 0.0421, -0.0057, ..., 0.0020, 0.0267, -0.0406], + [ 0.0709, -0.0720, 0.0262, ..., -0.0008, 0.0062, -0.0432], + [-0.0184, -0.0619, 0.0634, ..., -0.0418, 0.0184, -0.0141]], + device='cuda:0'), grad: tensor([[ 1.4150e-04, 3.1757e-04, 2.8777e-04, ..., -2.8205e-04, + 7.9393e-04, 3.9196e-04], + [ 1.8287e-04, 6.4278e-04, 2.4147e-03, ..., 1.3514e-03, + 4.1733e-03, 2.7299e-04], + [ 5.3864e-03, 2.4796e-05, 9.6178e-04, ..., 1.8635e-03, + 2.1057e-03, 3.9825e-03], + ..., + [-4.1747e-04, -2.0466e-03, -1.6809e-05, ..., -3.5095e-03, + -3.0804e-04, 8.5115e-04], + [-6.6872e-03, 6.6280e-04, 8.3828e-04, ..., 1.4467e-03, + 2.0084e-03, -4.0016e-03], + [ 7.0632e-05, 8.8215e-04, 2.7065e-03, ..., 1.4162e-03, + 4.2114e-03, 6.3324e-04]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0174, -0.0203, -0.0095, 0.0075, -0.0218, -0.0123, 0.0176, 0.0004, + 0.0252, -0.0200], device='cuda:0'), grad: tensor([-0.0197, 0.0208, 0.0407, -0.0425, 0.0163, -0.0326, 0.0104, -0.0139, + -0.0008, 0.0212], device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 216.91, cls_loss 0.5790 cls_loss_mapping 0.0079 cls_loss_causal 0.5456 re_mapping 0.0098 re_causal 0.0275 /// teacc 98.52 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.0515, -0.0503, -0.1167, ..., 0.1015, -0.0357, 0.0284], + [-0.0651, 0.0859, -0.0691, ..., 0.0297, -0.0427, -0.0751], + [-0.0380, 0.0697, -0.0827, ..., -0.0304, 0.0193, 0.1020], + ..., + [-0.0011, 0.0422, -0.0056, ..., 0.0020, 0.0267, -0.0421], + [ 0.0707, -0.0726, 0.0260, ..., -0.0009, 0.0063, -0.0433], + [-0.0183, -0.0611, 0.0641, ..., -0.0423, 0.0187, -0.0127]], + device='cuda:0'), grad: tensor([[ 6.8140e-04, 2.7437e-06, 4.3654e-04, ..., -1.7593e-02, + -4.7607e-03, -7.0038e-03], + [ 4.0793e-04, 1.0431e-05, 1.6773e-04, ..., 1.4744e-03, + 5.3316e-05, 5.2094e-05], + [ 5.8413e-04, -7.0333e-04, 3.0065e-04, ..., -1.8330e-03, + -6.5231e-03, -3.5267e-03], + ..., + [ 1.0834e-03, -3.4499e-04, 2.1780e-04, ..., 2.2049e-03, + 1.9062e-04, 1.8919e-04], + [-4.3178e-04, 1.0800e-04, 6.9094e-04, ..., -1.1244e-03, + 1.4820e-03, 6.9904e-04], + [-1.3800e-03, 3.2187e-04, -1.7776e-03, ..., -5.6152e-03, + 3.2258e-04, 1.6093e-04]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0170, -0.0201, -0.0095, 0.0070, -0.0220, -0.0112, 0.0178, 0.0005, + 0.0246, -0.0199], device='cuda:0'), grad: tensor([-0.0136, 0.0146, 0.0042, 0.0015, -0.0198, -0.0115, 0.0613, 0.0168, + -0.0108, -0.0426], device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 216.87, cls_loss 0.5368 cls_loss_mapping 0.0140 cls_loss_causal 0.5152 re_mapping 0.0097 re_causal 0.0238 /// teacc 98.75 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.0519, -0.0511, -0.1166, ..., 0.1018, -0.0358, 0.0285], + [-0.0632, 0.0858, -0.0695, ..., 0.0304, -0.0420, -0.0750], + [-0.0391, 0.0707, -0.0826, ..., -0.0287, 0.0197, 0.1036], + ..., + [-0.0006, 0.0421, -0.0056, ..., 0.0015, 0.0263, -0.0435], + [ 0.0720, -0.0723, 0.0254, ..., -0.0023, 0.0064, -0.0440], + [-0.0195, -0.0611, 0.0649, ..., -0.0420, 0.0186, -0.0128]], + device='cuda:0'), grad: tensor([[-0.0005, 0.0032, 0.0003, ..., -0.0034, -0.0002, -0.0005], + [-0.0005, 0.0028, 0.0008, ..., -0.0002, 0.0017, -0.0012], + [ 0.0006, -0.0120, -0.0074, ..., -0.0097, -0.0154, 0.0003], + ..., + [-0.0006, -0.0017, 0.0007, ..., -0.0035, 0.0078, -0.0008], + [-0.0002, 0.0016, 0.0009, ..., 0.0002, -0.0032, 0.0003], + [ 0.0008, 0.0014, 0.0048, ..., 0.0049, 0.0049, 0.0004]], + device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0165, -0.0189, -0.0085, 0.0070, -0.0221, -0.0111, 0.0174, 0.0006, + 0.0234, -0.0199], device='cuda:0'), grad: tensor([-0.0089, 0.0038, -0.0371, 0.0124, -0.0146, 0.0115, 0.0240, -0.0135, + -0.0082, 0.0304], device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 216.97, cls_loss 0.5596 cls_loss_mapping 0.0127 cls_loss_causal 0.5284 re_mapping 0.0100 re_causal 0.0244 /// teacc 98.41 lr 0.00010000 +Epoch 141, weight, value: tensor([[-5.0124e-02, -5.2366e-02, -1.1733e-01, ..., 1.0212e-01, + -3.5703e-02, 2.9217e-02], + [-6.4241e-02, 8.5370e-02, -6.8588e-02, ..., 3.1041e-02, + -4.2077e-02, -7.4141e-02], + [-3.9204e-02, 7.1405e-02, -8.3197e-02, ..., -2.8895e-02, + 2.0302e-02, 1.0395e-01], + ..., + [ 8.2416e-05, 4.0851e-02, -4.6927e-03, ..., -6.8448e-04, + 2.7246e-02, -4.4662e-02], + [ 7.1700e-02, -7.1818e-02, 2.5355e-02, ..., -1.4577e-03, + 6.9669e-03, -4.3133e-02], + [-2.0408e-02, -5.9976e-02, 6.4477e-02, ..., -4.1359e-02, + 1.7415e-02, -1.2823e-02]], device='cuda:0'), grad: tensor([[ 1.4435e-08, 2.2948e-04, 1.1301e-04, ..., 1.6375e-03, + 9.3317e-04, 4.6682e-04], + [ 9.4064e-08, 6.2037e-04, 1.4710e-04, ..., 4.6206e-04, + 4.5872e-04, 4.9114e-04], + [ 4.7917e-07, -4.1580e-03, -3.7117e-03, ..., -3.4485e-03, + -5.7831e-03, -9.5062e-03], + ..., + [ 1.0049e-06, -5.9986e-04, 4.4370e-04, ..., 4.6659e-04, + 5.1212e-04, 3.3998e-04], + [ 6.9737e-06, 1.0653e-03, 2.8210e-03, ..., -4.9734e-04, + 4.2605e-04, 4.4518e-03], + [-3.8408e-06, 2.3782e-04, -3.8338e-04, ..., 4.7565e-04, + 5.0640e-04, 3.6836e-04]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0158, -0.0190, -0.0084, 0.0062, -0.0216, -0.0112, 0.0175, 0.0006, + 0.0239, -0.0195], device='cuda:0'), grad: tensor([ 0.0104, 0.0118, -0.0901, 0.0159, 0.0130, 0.0089, 0.0078, 0.0099, + 0.0027, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 216.73, cls_loss 0.5728 cls_loss_mapping 0.0129 cls_loss_causal 0.5445 re_mapping 0.0096 re_causal 0.0245 /// teacc 98.39 lr 0.00010000 +Epoch 142, weight, value: tensor([[-4.9237e-02, -5.2068e-02, -1.1817e-01, ..., 1.0160e-01, + -3.5626e-02, 2.9327e-02], + [-6.4913e-02, 8.6561e-02, -6.8291e-02, ..., 3.2080e-02, + -4.2003e-02, -7.4834e-02], + [-3.9327e-02, 7.1725e-02, -8.3956e-02, ..., -2.9644e-02, + 2.0731e-02, 1.0467e-01], + ..., + [-8.2696e-04, 4.0072e-02, -5.1943e-03, ..., -5.8640e-05, + 2.7537e-02, -4.5068e-02], + [ 7.2333e-02, -7.3187e-02, 2.5453e-02, ..., -3.1826e-03, + 7.3961e-03, -4.3881e-02], + [-1.9853e-02, -6.1034e-02, 6.5409e-02, ..., -4.2222e-02, + 1.7850e-02, -1.3263e-02]], device='cuda:0'), grad: tensor([[ 1.4067e-04, 5.9271e-04, -9.9421e-05, ..., -1.1597e-03, + -1.0033e-03, -2.1591e-03], + [ 8.0585e-04, -2.7809e-03, -4.2648e-03, ..., -7.0801e-03, + -3.9444e-03, 1.4055e-04], + [ 7.9269e-03, 1.2871e-02, 6.8998e-04, ..., 1.5182e-03, + 8.2779e-04, 1.2529e-04], + ..., + [-1.0941e-02, -1.5640e-02, 9.3222e-04, ..., 1.5068e-03, + 6.8855e-04, 1.0735e-04], + [ 3.2163e-04, 1.1177e-03, 1.2541e-03, ..., 2.2755e-03, + 1.3056e-03, 8.7214e-04], + [ 6.9094e-04, 2.3689e-03, 8.0442e-04, ..., 2.9984e-03, + -1.7948e-03, 2.9707e-04]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0152, -0.0185, -0.0091, 0.0079, -0.0226, -0.0111, 0.0174, 0.0007, + 0.0240, -0.0198], device='cuda:0'), grad: tensor([-0.0120, -0.0454, 0.0381, 0.0261, 0.0076, -0.0155, 0.0201, -0.0464, + 0.0216, 0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 218.25, cls_loss 0.5784 cls_loss_mapping 0.0117 cls_loss_causal 0.5487 re_mapping 0.0100 re_causal 0.0251 /// teacc 98.51 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.0462, -0.0518, -0.1190, ..., 0.1013, -0.0357, 0.0307], + [-0.0645, 0.0864, -0.0685, ..., 0.0320, -0.0428, -0.0756], + [-0.0398, 0.0719, -0.0847, ..., -0.0291, 0.0206, 0.1040], + ..., + [-0.0003, 0.0403, -0.0048, ..., 0.0009, 0.0281, -0.0452], + [ 0.0717, -0.0728, 0.0254, ..., -0.0031, 0.0075, -0.0429], + [-0.0207, -0.0622, 0.0655, ..., -0.0436, 0.0182, -0.0130]], + device='cuda:0'), grad: tensor([[ 2.8172e-03, 9.7811e-05, 3.4237e-04, ..., -1.9245e-03, + 7.9274e-05, 1.1358e-03], + [ 3.5346e-05, 3.1233e-05, 4.9639e-04, ..., 5.0879e-04, + 5.9652e-04, 5.5599e-04], + [-4.4861e-03, -6.2513e-04, 6.1750e-04, ..., 9.5844e-04, + -1.8768e-03, -3.5858e-03], + ..., + [ 9.0003e-06, 2.7701e-05, 5.0354e-04, ..., 3.3236e-04, + 4.6206e-04, 2.7084e-04], + [ 1.2982e-04, -1.9419e-04, -3.3054e-03, ..., -3.5858e-03, + -2.9526e-03, -2.7428e-03], + [ 2.9221e-05, 1.1796e-04, 1.2932e-03, ..., 9.0981e-04, + 1.2693e-03, 8.3590e-04]], device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0147, -0.0184, -0.0093, 0.0070, -0.0228, -0.0111, 0.0188, 0.0011, + 0.0238, -0.0196], device='cuda:0'), grad: tensor([ 0.0091, 0.0083, -0.0069, 0.0096, -0.0024, 0.0108, 0.0105, 0.0057, + -0.0542, 0.0095], device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 218.31, cls_loss 0.5442 cls_loss_mapping 0.0106 cls_loss_causal 0.5176 re_mapping 0.0095 re_causal 0.0243 /// teacc 98.68 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.0470, -0.0507, -0.1194, ..., 0.1015, -0.0365, 0.0308], + [-0.0640, 0.0865, -0.0688, ..., 0.0323, -0.0438, -0.0760], + [-0.0392, 0.0721, -0.0861, ..., -0.0287, 0.0207, 0.1044], + ..., + [-0.0015, 0.0403, -0.0050, ..., 0.0007, 0.0280, -0.0445], + [ 0.0711, -0.0734, 0.0257, ..., -0.0032, 0.0073, -0.0439], + [-0.0206, -0.0629, 0.0647, ..., -0.0436, 0.0177, -0.0126]], + device='cuda:0'), grad: tensor([[ 1.9443e-04, 9.8038e-04, 3.0446e-04, ..., 1.8225e-03, + -2.0719e-04, -4.9210e-04], + [ 4.1544e-05, 9.3918e-03, 1.2846e-03, ..., 1.4175e-02, + 7.8058e-04, -2.5845e-04], + [ 8.7738e-05, -1.0967e-03, -5.1270e-03, ..., -3.7441e-03, + -3.0842e-03, -1.6146e-03], + ..., + [ 1.1616e-03, 3.8242e-04, 5.5351e-03, ..., 2.4834e-03, + 4.3335e-03, 4.6206e-04], + [-2.6207e-03, -1.0780e-02, 1.5364e-03, ..., -1.7883e-02, + 3.5834e-04, -1.7862e-03], + [-1.4448e-03, -4.0817e-04, -6.9160e-03, ..., -1.5392e-03, + -4.3411e-03, 4.5156e-04]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0149, -0.0190, -0.0096, 0.0071, -0.0227, -0.0113, 0.0180, 0.0010, + 0.0247, -0.0191], device='cuda:0'), grad: tensor([-0.0139, -0.0139, -0.0466, 0.0317, 0.0261, 0.0184, 0.0056, 0.0360, + -0.0228, -0.0206], device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 218.92, cls_loss 0.5555 cls_loss_mapping 0.0137 cls_loss_causal 0.5261 re_mapping 0.0097 re_causal 0.0254 /// teacc 98.54 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.0468, -0.0509, -0.1205, ..., 0.1025, -0.0368, 0.0322], + [-0.0642, 0.0864, -0.0689, ..., 0.0318, -0.0440, -0.0768], + [-0.0405, 0.0715, -0.0849, ..., -0.0286, 0.0225, 0.1053], + ..., + [-0.0015, 0.0419, -0.0044, ..., -0.0003, 0.0277, -0.0440], + [ 0.0715, -0.0742, 0.0254, ..., -0.0020, 0.0075, -0.0459], + [-0.0206, -0.0629, 0.0649, ..., -0.0440, 0.0174, -0.0138]], + device='cuda:0'), grad: tensor([[ 1.7440e-04, 2.4891e-04, 6.8665e-05, ..., 3.2253e-03, + 7.3147e-04, 1.1387e-03], + [ 1.3530e-04, -9.7427e-03, -2.4796e-03, ..., -6.5918e-03, + -1.7071e-03, 4.3440e-04], + [ 2.9144e-03, 1.9083e-03, 1.4853e-04, ..., 8.0261e-03, + 1.8501e-03, 5.8403e-03], + ..., + [ 3.1257e-04, 1.0857e-02, 2.9964e-03, ..., 1.4820e-03, + 3.3998e-04, 1.0328e-03], + [-3.4866e-03, -5.4207e-03, -1.0920e-04, ..., 1.9207e-03, + 8.4591e-04, -6.0501e-03], + [ 8.8871e-05, 3.2520e-04, 1.7481e-03, ..., 9.1553e-04, + 1.5802e-03, 2.9135e-04]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0160, -0.0190, -0.0098, 0.0068, -0.0223, -0.0111, 0.0176, 0.0007, + 0.0249, -0.0195], device='cuda:0'), grad: tensor([ 0.0174, -0.0490, 0.0326, -0.0264, -0.0226, 0.0273, -0.0142, 0.0515, + -0.0358, 0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 219.56, cls_loss 0.5563 cls_loss_mapping 0.0121 cls_loss_causal 0.5208 re_mapping 0.0099 re_causal 0.0252 /// teacc 98.73 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.0474, -0.0520, -0.1209, ..., 0.1014, -0.0367, 0.0319], + [-0.0642, 0.0869, -0.0679, ..., 0.0322, -0.0442, -0.0772], + [-0.0413, 0.0721, -0.0852, ..., -0.0279, 0.0227, 0.1054], + ..., + [-0.0015, 0.0414, -0.0039, ..., -0.0016, 0.0282, -0.0466], + [ 0.0715, -0.0748, 0.0269, ..., -0.0021, 0.0072, -0.0473], + [-0.0203, -0.0619, 0.0645, ..., -0.0441, 0.0175, -0.0116]], + device='cuda:0'), grad: tensor([[ 1.4029e-03, 5.1051e-05, 2.6131e-04, ..., 6.3324e-03, + 8.9874e-03, 1.1368e-02], + [-3.5048e-05, 2.3872e-05, 7.0953e-04, ..., 5.0497e-04, + 8.7023e-04, -5.2643e-04], + [ 6.8247e-05, 7.5340e-05, -1.3268e-02, ..., -4.9782e-03, + -1.3451e-02, 2.4533e-04], + ..., + [ 3.4189e-04, 8.0395e-04, 1.1810e-02, ..., 9.9468e-04, + 1.2085e-02, -3.8624e-05], + [-1.2121e-03, -1.5650e-03, -3.0746e-03, ..., -8.5526e-03, + -1.1971e-02, -1.2085e-02], + [ 1.3709e-04, 2.1279e-04, 2.4815e-03, ..., 2.5043e-03, + 2.8534e-03, 3.6263e-04]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0159, -0.0185, -0.0089, 0.0073, -0.0223, -0.0118, 0.0173, -0.0011, + 0.0252, -0.0190], device='cuda:0'), grad: tensor([ 0.0478, -0.0132, -0.0127, 0.0267, 0.0135, -0.0072, 0.0047, 0.0190, + -0.1017, 0.0231], device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 218.44, cls_loss 0.5518 cls_loss_mapping 0.0107 cls_loss_causal 0.5203 re_mapping 0.0096 re_causal 0.0251 /// teacc 98.77 lr 0.00010000 +Epoch 147, weight, value: tensor([[-0.0472, -0.0529, -0.1206, ..., 0.1009, -0.0367, 0.0314], + [-0.0642, 0.0865, -0.0685, ..., 0.0323, -0.0450, -0.0775], + [-0.0413, 0.0728, -0.0830, ..., -0.0275, 0.0248, 0.1061], + ..., + [-0.0027, 0.0416, -0.0049, ..., -0.0026, 0.0276, -0.0467], + [ 0.0735, -0.0736, 0.0254, ..., -0.0021, 0.0060, -0.0468], + [-0.0208, -0.0632, 0.0653, ..., -0.0442, 0.0180, -0.0127]], + device='cuda:0'), grad: tensor([[ 1.5903e-04, 1.5748e-04, 9.7334e-05, ..., 9.9945e-04, + 4.1515e-05, 1.4901e-04], + [-1.5392e-03, -1.6384e-03, 1.2267e-04, ..., -8.7585e-03, + 4.4137e-05, -5.9986e-04], + [ 1.7059e-04, 4.0936e-04, 3.7551e-04, ..., 1.1253e-03, + 2.8920e-04, 2.7156e-04], + ..., + [ 2.4438e-04, -2.0218e-04, 6.3467e-04, ..., -6.1572e-05, + 3.5167e-04, -1.4436e-04], + [ 7.1704e-05, 3.3092e-04, 6.5708e-04, ..., -1.3009e-05, + 4.6206e-04, 1.1057e-04], + [-4.7040e-04, -5.7602e-04, -2.4815e-03, ..., 4.0221e-04, + -1.9007e-03, 9.8050e-05]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0168, -0.0187, -0.0088, 0.0059, -0.0227, -0.0117, 0.0185, -0.0010, + 0.0246, -0.0187], device='cuda:0'), grad: tensor([ 0.0072, -0.0504, 0.0077, 0.0045, -0.0020, 0.0070, 0.0205, 0.0054, + 0.0050, -0.0049], device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 219.65, cls_loss 0.5545 cls_loss_mapping 0.0156 cls_loss_causal 0.5203 re_mapping 0.0090 re_causal 0.0234 /// teacc 98.55 lr 0.00010000 +Epoch 148, weight, value: tensor([[-0.0478, -0.0543, -0.1211, ..., 0.1003, -0.0384, 0.0307], + [-0.0651, 0.0876, -0.0667, ..., 0.0332, -0.0439, -0.0774], + [-0.0427, 0.0725, -0.0839, ..., -0.0279, 0.0245, 0.1066], + ..., + [-0.0012, 0.0411, -0.0053, ..., -0.0021, 0.0270, -0.0471], + [ 0.0745, -0.0734, 0.0245, ..., -0.0018, 0.0061, -0.0451], + [-0.0215, -0.0626, 0.0657, ..., -0.0453, 0.0184, -0.0124]], + device='cuda:0'), grad: tensor([[ 1.5008e-04, 7.8738e-05, 1.1854e-03, ..., 1.2503e-03, + 1.4267e-03, 5.1641e-04], + [ 8.1348e-04, 8.7833e-04, 9.3460e-04, ..., 1.3056e-03, + 7.8392e-04, 1.1740e-03], + [ 3.3259e-04, -8.0338e-03, -7.7782e-03, ..., -4.1389e-03, + -5.0888e-03, 6.6223e-03], + ..., + [ 1.0519e-03, 5.1117e-03, 9.0256e-03, ..., 7.0953e-03, + 8.9417e-03, 2.0351e-03], + [ 6.6185e-04, 9.9182e-05, 1.3342e-03, ..., 1.2522e-03, + 1.4229e-03, -9.1858e-03], + [-9.4528e-03, 3.1543e-04, -7.0381e-03, ..., -9.7084e-04, + -4.2419e-03, -3.4404e-04]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0157, -0.0187, -0.0084, 0.0046, -0.0223, -0.0119, 0.0185, -0.0002, + 0.0255, -0.0188], device='cuda:0'), grad: tensor([ 0.0119, 0.0263, -0.0014, -0.0081, -0.0359, 0.0049, 0.0152, 0.0595, + -0.0339, -0.0386], device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 218.75, cls_loss 0.5467 cls_loss_mapping 0.0136 cls_loss_causal 0.5301 re_mapping 0.0096 re_causal 0.0252 /// teacc 98.73 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.0494, -0.0540, -0.1222, ..., 0.1015, -0.0388, 0.0311], + [-0.0632, 0.0873, -0.0663, ..., 0.0325, -0.0456, -0.0782], + [-0.0429, 0.0739, -0.0855, ..., -0.0278, 0.0242, 0.1061], + ..., + [-0.0012, 0.0411, -0.0040, ..., -0.0028, 0.0275, -0.0485], + [ 0.0745, -0.0744, 0.0258, ..., -0.0023, 0.0069, -0.0446], + [-0.0214, -0.0627, 0.0650, ..., -0.0440, 0.0183, -0.0109]], + device='cuda:0'), grad: tensor([[ 1.2481e-04, 6.1631e-05, -4.6194e-05, ..., 7.5340e-04, + -3.2449e-04, -4.5204e-04], + [ 6.4659e-04, 1.3644e-07, 1.6749e-04, ..., 1.6069e-03, + 1.2302e-04, 1.9908e-05], + [ 1.0920e-03, 6.2704e-05, 5.2547e-04, ..., -1.7509e-03, + 2.4915e-04, -1.3435e-04], + ..., + [ 4.9162e-04, -5.8937e-04, 1.0366e-03, ..., 1.2293e-03, + 5.5265e-04, 8.5890e-05], + [-3.7537e-03, 6.6221e-05, 1.2147e-04, ..., -4.8828e-03, + 7.0751e-05, -3.7163e-05], + [ 9.6977e-05, 1.1224e-04, 7.7629e-04, ..., -4.4131e-04, + -6.9571e-04, 1.4532e-04]], device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0166, -0.0189, -0.0088, 0.0046, -0.0225, -0.0116, 0.0181, -0.0005, + 0.0254, -0.0184], device='cuda:0'), grad: tensor([ 0.0129, 0.0205, -0.0145, 0.0104, 0.0042, 0.0270, -0.0137, 0.0160, + -0.0475, -0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 219.44, cls_loss 0.5687 cls_loss_mapping 0.0111 cls_loss_causal 0.5418 re_mapping 0.0095 re_causal 0.0246 /// teacc 98.73 lr 0.00010000 +Epoch 150, weight, value: tensor([[-5.0625e-02, -5.5483e-02, -1.2251e-01, ..., 1.0156e-01, + -3.9017e-02, 3.0644e-02], + [-6.4196e-02, 8.8790e-02, -6.6445e-02, ..., 3.2618e-02, + -4.5692e-02, -7.6992e-02], + [-4.2810e-02, 7.2717e-02, -8.5510e-02, ..., -2.8157e-02, + 2.4343e-02, 1.0534e-01], + ..., + [ 1.0993e-04, 4.1688e-02, -4.2009e-03, ..., -1.8091e-03, + 2.7494e-02, -4.7318e-02], + [ 7.4449e-02, -7.4823e-02, 2.5418e-02, ..., -2.3711e-03, + 6.9287e-03, -4.3682e-02], + [-2.0378e-02, -6.4456e-02, 6.5157e-02, ..., -4.4950e-02, + 1.8414e-02, -1.1384e-02]], device='cuda:0'), grad: tensor([[ 6.0052e-06, 3.2020e-04, 7.2575e-04, ..., 9.4128e-04, + 5.3358e-04, 7.7248e-04], + [-5.4240e-05, 6.0290e-05, 5.7697e-04, ..., 5.0592e-04, + 3.5691e-04, 5.5218e-04], + [ 7.6741e-06, -1.0562e-04, 8.5402e-04, ..., 8.4019e-04, + 5.9080e-04, 6.4659e-04], + ..., + [ 5.1588e-05, -1.3769e-04, -2.2106e-03, ..., -3.3131e-03, + -2.7866e-03, -2.0561e-03], + [ 2.9013e-05, 4.5872e-04, 1.7405e-03, ..., 1.3227e-03, + 1.0061e-03, 1.6241e-03], + [ 1.4229e-03, 1.3649e-04, 3.0670e-03, ..., 1.0719e-03, + 1.7023e-03, 1.0366e-03]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0166, -0.0198, -0.0094, 0.0045, -0.0223, -0.0122, 0.0194, 0.0005, + 0.0261, -0.0193], device='cuda:0'), grad: tensor([ 0.0074, 0.0049, 0.0058, -0.0274, 0.0050, 0.0033, -0.0190, -0.0081, + 0.0162, 0.0120], device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.70, cls_loss 0.5127 cls_loss_mapping 0.0100 cls_loss_causal 0.4838 re_mapping 0.0098 re_causal 0.0253 /// teacc 98.88 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.0508, -0.0542, -0.1226, ..., 0.1017, -0.0396, 0.0310], + [-0.0640, 0.0882, -0.0655, ..., 0.0325, -0.0447, -0.0773], + [-0.0429, 0.0733, -0.0859, ..., -0.0283, 0.0246, 0.1062], + ..., + [ 0.0006, 0.0414, -0.0040, ..., -0.0010, 0.0278, -0.0473], + [ 0.0743, -0.0756, 0.0239, ..., -0.0029, 0.0060, -0.0435], + [-0.0205, -0.0641, 0.0656, ..., -0.0463, 0.0190, -0.0127]], + device='cuda:0'), grad: tensor([[-2.9755e-04, 6.7568e-04, 2.6837e-05, ..., 1.3456e-03, + 2.5868e-05, -1.7452e-03], + [-1.7281e-03, 4.3564e-03, 3.7849e-05, ..., 5.3596e-03, + 3.3587e-05, 3.7432e-04], + [ 2.7013e-04, 1.2493e-03, 4.0936e-04, ..., 1.4610e-03, + 1.6394e-03, 8.7118e-04], + ..., + [ 1.0347e-03, 1.7345e-04, 2.0199e-03, ..., 4.9067e-04, + 2.7981e-03, 1.9521e-06], + [ 1.9760e-03, 6.2561e-03, 3.9062e-03, ..., 6.0730e-03, + 3.8414e-03, 2.2185e-04], + [ 8.7309e-04, 1.5211e-04, 1.3332e-03, ..., -9.2864e-05, + 5.9166e-03, 9.4295e-05]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0169, -0.0199, -0.0090, 0.0057, -0.0218, -0.0122, 0.0194, 0.0003, + 0.0249, -0.0200], device='cuda:0'), grad: tensor([-0.0021, 0.0118, -0.0135, -0.0428, 0.0125, 0.0103, -0.0464, 0.0187, + 0.0440, 0.0074], device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 218.84, cls_loss 0.5391 cls_loss_mapping 0.0120 cls_loss_causal 0.5069 re_mapping 0.0094 re_causal 0.0230 /// teacc 98.74 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.0518, -0.0542, -0.1224, ..., 0.1032, -0.0393, 0.0307], + [-0.0634, 0.0885, -0.0659, ..., 0.0307, -0.0456, -0.0767], + [-0.0429, 0.0735, -0.0865, ..., -0.0272, 0.0251, 0.1061], + ..., + [ 0.0022, 0.0417, -0.0024, ..., -0.0013, 0.0282, -0.0483], + [ 0.0740, -0.0754, 0.0236, ..., -0.0024, 0.0054, -0.0426], + [-0.0210, -0.0649, 0.0659, ..., -0.0474, 0.0197, -0.0131]], + device='cuda:0'), grad: tensor([[-7.0305e-03, -9.7847e-04, -1.4391e-03, ..., -9.0332e-03, + -3.9978e-03, -3.1776e-03], + [ 9.0525e-06, 1.6642e-04, 9.5177e-04, ..., 1.6727e-03, + 2.4748e-04, 2.3270e-04], + [ 1.5986e-04, 7.5996e-05, 6.9571e-04, ..., 1.1234e-03, + 4.0817e-04, 2.6369e-04], + ..., + [ 1.1563e-04, 2.2733e-04, -1.2970e-03, ..., -9.5987e-04, + -1.8978e-03, -8.3923e-05], + [ 1.6689e-03, 1.2267e-04, 1.0347e-03, ..., 2.5864e-03, + 1.3256e-03, 9.4795e-04], + [-3.4118e-04, 9.9540e-05, -2.9869e-03, ..., -3.7079e-03, + -3.1710e-04, -1.4496e-03]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0171, -0.0200, -0.0091, 0.0056, -0.0202, -0.0124, 0.0191, 0.0003, + 0.0247, -0.0210], device='cuda:0'), grad: tensor([-0.0713, 0.0279, -0.0105, -0.0140, 0.0186, 0.0159, 0.0364, 0.0174, + -0.0057, -0.0147], device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 218.08, cls_loss 0.5130 cls_loss_mapping 0.0128 cls_loss_causal 0.4865 re_mapping 0.0100 re_causal 0.0250 /// teacc 98.41 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.0515, -0.0537, -0.1216, ..., 0.1033, -0.0388, 0.0311], + [-0.0639, 0.0883, -0.0655, ..., 0.0319, -0.0457, -0.0768], + [-0.0423, 0.0734, -0.0866, ..., -0.0280, 0.0250, 0.1055], + ..., + [ 0.0018, 0.0416, -0.0029, ..., -0.0021, 0.0282, -0.0486], + [ 0.0746, -0.0749, 0.0250, ..., -0.0013, 0.0058, -0.0429], + [-0.0222, -0.0631, 0.0662, ..., -0.0476, 0.0195, -0.0124]], + device='cuda:0'), grad: tensor([[ 2.9374e-06, 2.5660e-05, 4.2295e-04, ..., -1.8854e-03, + 1.9777e-04, -2.2650e-05], + [ 7.8559e-05, 7.8354e-03, 8.4734e-04, ..., 1.2913e-03, + 1.4269e-04, 2.8402e-05], + [ 2.3618e-06, 1.0881e-03, 5.0163e-04, ..., -1.5526e-03, + 5.2261e-04, 1.2755e-05], + ..., + [-7.2317e-07, -1.8635e-03, 1.1003e-04, ..., 5.1451e-04, + -6.7282e-04, 2.9191e-05], + [ 6.5506e-05, 9.1434e-05, -6.4201e-03, ..., -7.4310e-03, + -2.6016e-03, -6.6662e-04], + [ 1.3605e-05, -7.5340e-03, 1.0824e-03, ..., 1.4887e-03, + -1.7405e-04, 1.1641e-04]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0179, -0.0198, -0.0090, 0.0054, -0.0198, -0.0122, 0.0192, -0.0016, + 0.0244, -0.0201], device='cuda:0'), grad: tensor([ 0.0050, 0.0399, -0.0239, 0.0054, 0.0017, 0.0252, -0.0159, 0.0017, + -0.0161, -0.0229], device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 220.34, cls_loss 0.5446 cls_loss_mapping 0.0121 cls_loss_causal 0.5082 re_mapping 0.0092 re_causal 0.0226 /// teacc 98.84 lr 0.00010000 +Epoch 154, weight, value: tensor([[-5.2125e-02, -5.4604e-02, -1.2012e-01, ..., 1.0172e-01, + -3.8075e-02, 3.1402e-02], + [-6.4949e-02, 8.7685e-02, -6.4411e-02, ..., 3.2875e-02, + -4.4460e-02, -7.6388e-02], + [-4.1604e-02, 7.3022e-02, -8.7335e-02, ..., -2.8257e-02, + 2.4752e-02, 1.0659e-01], + ..., + [ 1.2217e-03, 4.2745e-02, -3.7406e-03, ..., -2.3695e-03, + 2.7150e-02, -4.8546e-02], + [ 7.4566e-02, -7.5035e-02, 2.5123e-02, ..., -2.1353e-05, + 6.3070e-03, -4.3858e-02], + [-2.2788e-02, -6.1422e-02, 6.5785e-02, ..., -4.7795e-02, + 1.9485e-02, -1.2961e-02]], device='cuda:0'), grad: tensor([[ 0.0001, 0.0012, 0.0007, ..., 0.0047, 0.0023, 0.0014], + [ 0.0011, -0.0030, 0.0012, ..., -0.0099, -0.0072, -0.0045], + [ 0.0002, 0.0017, 0.0007, ..., 0.0068, 0.0037, 0.0003], + ..., + [ 0.0016, 0.0012, 0.0022, ..., 0.0038, 0.0011, 0.0004], + [-0.0124, -0.0006, -0.0453, ..., -0.0068, -0.0173, -0.0015], + [ 0.0039, 0.0005, 0.0131, ..., 0.0023, 0.0045, 0.0003]], + device='cuda:0') +Epoch 154, bias, value: tensor([ 0.0173, -0.0195, -0.0087, 0.0053, -0.0194, -0.0130, 0.0184, -0.0016, + 0.0253, -0.0198], device='cuda:0'), grad: tensor([ 0.0334, -0.0038, 0.0321, 0.0348, 0.0340, 0.0300, -0.0674, -0.0565, + -0.0775, 0.0408], device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 217.99, cls_loss 0.5606 cls_loss_mapping 0.0081 cls_loss_causal 0.5307 re_mapping 0.0092 re_causal 0.0241 /// teacc 98.69 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.0497, -0.0547, -0.1203, ..., 0.1027, -0.0375, 0.0318], + [-0.0646, 0.0878, -0.0651, ..., 0.0335, -0.0443, -0.0768], + [-0.0427, 0.0722, -0.0873, ..., -0.0277, 0.0247, 0.1053], + ..., + [ 0.0010, 0.0436, -0.0039, ..., -0.0021, 0.0267, -0.0472], + [ 0.0748, -0.0753, 0.0256, ..., -0.0011, 0.0061, -0.0443], + [-0.0217, -0.0618, 0.0659, ..., -0.0482, 0.0189, -0.0143]], + device='cuda:0'), grad: tensor([[-1.9445e-03, 9.3508e-04, -1.3344e-02, ..., 1.8673e-03, + -3.4695e-03, 1.2846e-03], + [ 1.2541e-04, -6.3419e-04, -7.6437e-04, ..., -2.3918e-03, + -2.3985e-04, -1.7443e-03], + [ 1.5144e-03, -1.2407e-03, -4.9543e-04, ..., -5.6877e-03, + 1.6890e-03, -8.8263e-04], + ..., + [ 8.7357e-04, -3.9597e-03, 8.9943e-05, ..., 1.7881e-03, + 5.0831e-04, 4.6873e-04], + [ 3.3474e-03, 1.2798e-03, 9.0485e-03, ..., 1.6546e-03, + 5.2643e-03, 1.2789e-03], + [ 1.0920e-03, 2.3098e-03, 2.8934e-03, ..., -2.2519e-04, + 2.1954e-03, 1.2684e-03]], device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0177, -0.0197, -0.0084, 0.0055, -0.0194, -0.0138, 0.0183, -0.0006, + 0.0249, -0.0202], device='cuda:0'), grad: tensor([-0.0065, -0.0136, -0.0665, 0.0129, 0.0318, 0.0113, -0.0145, 0.0032, + 0.0391, 0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 216.56, cls_loss 0.5364 cls_loss_mapping 0.0102 cls_loss_causal 0.5046 re_mapping 0.0088 re_causal 0.0218 /// teacc 98.74 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.0495, -0.0565, -0.1197, ..., 0.1016, -0.0368, 0.0321], + [-0.0651, 0.0874, -0.0646, ..., 0.0333, -0.0438, -0.0771], + [-0.0427, 0.0735, -0.0888, ..., -0.0282, 0.0253, 0.1051], + ..., + [ 0.0003, 0.0435, -0.0032, ..., -0.0008, 0.0270, -0.0467], + [ 0.0757, -0.0759, 0.0258, ..., -0.0013, 0.0060, -0.0455], + [-0.0215, -0.0623, 0.0654, ..., -0.0484, 0.0187, -0.0123]], + device='cuda:0'), grad: tensor([[-0.0012, 0.0008, -0.0028, ..., -0.0012, -0.0137, -0.0006], + [ 0.0008, 0.0014, 0.0006, ..., 0.0014, 0.0003, 0.0003], + [-0.0154, 0.0019, 0.0007, ..., 0.0019, -0.0307, -0.0214], + ..., + [-0.0020, -0.0023, 0.0015, ..., 0.0036, 0.0024, -0.0014], + [ 0.0011, -0.0013, -0.0008, ..., -0.0022, -0.0009, 0.0007], + [ 0.0011, -0.0015, -0.0015, ..., 0.0007, 0.0051, -0.0001]], + device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0167, -0.0194, -0.0079, 0.0050, -0.0205, -0.0135, 0.0185, -0.0001, + 0.0252, -0.0200], device='cuda:0'), grad: tensor([ 0.0030, 0.0303, 0.0044, 0.0643, -0.0345, -0.0037, 0.0147, -0.0192, + -0.0590, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 217.03, cls_loss 0.5340 cls_loss_mapping 0.0120 cls_loss_causal 0.5038 re_mapping 0.0094 re_causal 0.0240 /// teacc 98.82 lr 0.00010000 +Epoch 157, weight, value: tensor([[-4.9805e-02, -5.5961e-02, -1.2005e-01, ..., 1.0231e-01, + -3.7098e-02, 3.2303e-02], + [-6.5050e-02, 8.7422e-02, -6.3582e-02, ..., 3.3190e-02, + -4.3809e-02, -7.7360e-02], + [-4.1121e-02, 7.3126e-02, -8.8857e-02, ..., -2.8688e-02, + 2.4833e-02, 1.0477e-01], + ..., + [ 7.5611e-04, 4.4394e-02, -2.6403e-03, ..., 1.7366e-05, + 2.6599e-02, -4.7002e-02], + [ 7.5037e-02, -7.5511e-02, 2.6145e-02, ..., -5.8573e-04, + 7.0196e-03, -4.4899e-02], + [-2.0853e-02, -6.2832e-02, 6.5807e-02, ..., -4.9737e-02, + 1.9347e-02, -1.3142e-02]], device='cuda:0'), grad: tensor([[-9.2602e-04, -8.8787e-04, 1.6654e-04, ..., -5.0774e-03, + -5.2977e-04, -2.7847e-03], + [ 2.0313e-03, 6.3324e-03, 5.1320e-05, ..., 1.5190e-02, + 1.3149e-04, 4.2701e-04], + [-2.2545e-03, -9.3002e-03, 1.4412e-04, ..., -2.1545e-02, + -4.6587e-04, -2.3136e-03], + ..., + [ 1.6117e-04, 5.9748e-04, -9.8610e-04, ..., 3.5324e-03, + -5.8317e-04, 1.3056e-03], + [ 5.2929e-04, 9.4843e-04, 5.9426e-05, ..., -1.0452e-03, + 3.0565e-04, 1.2712e-03], + [ 2.1362e-04, 6.2132e-04, 8.4162e-04, ..., 2.1038e-03, + 8.4209e-04, 6.4850e-04]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0159, -0.0193, -0.0086, 0.0058, -0.0199, -0.0134, 0.0177, -0.0003, + 0.0260, -0.0201], device='cuda:0'), grad: tensor([-0.0170, 0.0097, -0.0155, -0.0057, 0.0183, 0.0062, -0.0170, 0.0178, + -0.0107, 0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 217.19, cls_loss 0.5593 cls_loss_mapping 0.0100 cls_loss_causal 0.5289 re_mapping 0.0088 re_causal 0.0231 /// teacc 98.80 lr 0.00010000 +Epoch 158, weight, value: tensor([[-5.0096e-02, -5.6710e-02, -1.2016e-01, ..., 1.0233e-01, + -3.6803e-02, 3.2543e-02], + [-6.4083e-02, 8.7536e-02, -6.2770e-02, ..., 3.3331e-02, + -4.4644e-02, -7.9022e-02], + [-3.9689e-02, 7.3949e-02, -8.7432e-02, ..., -2.8299e-02, + 2.5145e-02, 1.0521e-01], + ..., + [ 5.2419e-04, 4.4582e-02, -2.9059e-03, ..., -4.7805e-04, + 2.6414e-02, -4.7600e-02], + [ 7.5638e-02, -7.6667e-02, 2.5605e-02, ..., 9.3757e-05, + 6.8568e-03, -4.5763e-02], + [-2.2021e-02, -6.3172e-02, 6.5821e-02, ..., -4.9458e-02, + 1.9346e-02, -1.3714e-02]], device='cuda:0'), grad: tensor([[ 5.4693e-04, 4.6396e-04, 2.0695e-04, ..., 1.0662e-03, + 1.1647e-04, 5.1785e-04], + [-1.8806e-03, -1.8275e-04, 7.8154e-04, ..., -5.1928e-04, + 4.1389e-04, -1.2655e-03], + [ 9.5844e-05, 5.8937e-04, 4.8184e-04, ..., 1.1063e-03, + 2.1017e-04, -6.4754e-04], + ..., + [-2.5201e-04, -1.7471e-03, -1.5512e-05, ..., 5.9366e-04, + -8.3637e-04, 1.0365e-04], + [ 2.5675e-05, 1.4029e-03, 2.2221e-03, ..., -4.3793e-03, + 9.6750e-04, 1.6081e-04], + [-6.3717e-05, -2.5330e-03, -8.6441e-03, ..., -1.9035e-03, + -3.6297e-03, 5.7876e-05]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0167, -0.0191, -0.0082, 0.0057, -0.0204, -0.0143, 0.0183, -0.0007, + 0.0257, -0.0197], device='cuda:0'), grad: tensor([ 0.0218, -0.0082, 0.0177, -0.0136, 0.0427, 0.0146, 0.0034, -0.0206, + -0.0020, -0.0558], device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 218.82, cls_loss 0.5508 cls_loss_mapping 0.0104 cls_loss_causal 0.5220 re_mapping 0.0090 re_causal 0.0225 /// teacc 98.52 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.0492, -0.0573, -0.1196, ..., 0.1029, -0.0369, 0.0330], + [-0.0653, 0.0867, -0.0634, ..., 0.0327, -0.0459, -0.0787], + [-0.0397, 0.0739, -0.0880, ..., -0.0287, 0.0254, 0.1054], + ..., + [ 0.0002, 0.0459, -0.0032, ..., -0.0007, 0.0262, -0.0466], + [ 0.0763, -0.0762, 0.0249, ..., -0.0006, 0.0070, -0.0468], + [-0.0232, -0.0640, 0.0668, ..., -0.0496, 0.0203, -0.0127]], + device='cuda:0'), grad: tensor([[ 6.7949e-04, 8.7452e-04, 1.6413e-03, ..., 1.4505e-03, + 3.2949e-04, 4.0674e-04], + [ 2.7676e-03, 2.7428e-03, 1.5545e-03, ..., 4.2992e-03, + 3.6573e-04, 5.9509e-04], + [ 1.5841e-03, 1.3742e-03, 8.5258e-04, ..., 1.3123e-03, + 2.1338e-04, 1.0004e-03], + ..., + [-4.6420e-04, -2.9774e-03, -4.8828e-03, ..., -1.3380e-03, + -2.8095e-03, 3.1805e-04], + [ 9.6846e-04, 9.8133e-04, -4.4594e-03, ..., -2.7008e-03, + -1.5984e-03, 2.5773e-04], + [-5.8375e-06, 3.0384e-03, 7.0229e-03, ..., 2.5864e-03, + 3.1605e-03, -2.6360e-03]], device='cuda:0') +Epoch 159, bias, value: tensor([ 0.0169, -0.0186, -0.0084, 0.0057, -0.0203, -0.0136, 0.0173, -0.0006, + 0.0254, -0.0198], device='cuda:0'), grad: tensor([ 0.0259, 0.0530, 0.0313, 0.0249, 0.0226, -0.0162, -0.0881, 0.0112, + -0.0062, -0.0584], device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 219.08, cls_loss 0.5607 cls_loss_mapping 0.0092 cls_loss_causal 0.5292 re_mapping 0.0094 re_causal 0.0236 /// teacc 98.78 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.0496, -0.0574, -0.1196, ..., 0.1034, -0.0368, 0.0335], + [-0.0636, 0.0868, -0.0648, ..., 0.0325, -0.0472, -0.0787], + [-0.0408, 0.0740, -0.0881, ..., -0.0289, 0.0259, 0.1054], + ..., + [-0.0006, 0.0464, -0.0034, ..., -0.0003, 0.0260, -0.0470], + [ 0.0753, -0.0756, 0.0242, ..., -0.0002, 0.0066, -0.0460], + [-0.0220, -0.0649, 0.0666, ..., -0.0495, 0.0201, -0.0121]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0002, 0.0019, ..., 0.0017, 0.0017, 0.0002], + [ 0.0004, 0.0002, 0.0015, ..., 0.0012, 0.0006, 0.0002], + [-0.0001, -0.0007, -0.0147, ..., -0.0110, -0.0163, -0.0019], + ..., + [ 0.0001, -0.0005, 0.0026, ..., 0.0010, 0.0027, 0.0001], + [-0.0035, 0.0002, 0.0021, ..., 0.0009, 0.0036, 0.0001], + [ 0.0005, 0.0002, -0.0159, ..., 0.0032, -0.0009, 0.0001]], + device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0161, -0.0192, -0.0076, 0.0056, -0.0205, -0.0131, 0.0168, -0.0001, + 0.0256, -0.0196], device='cuda:0'), grad: tensor([ 0.0171, 0.0157, -0.0497, -0.0155, 0.0197, 0.0175, -0.0103, 0.0177, + -0.0055, -0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 216.59, cls_loss 0.5568 cls_loss_mapping 0.0092 cls_loss_causal 0.5338 re_mapping 0.0090 re_causal 0.0228 /// teacc 98.82 lr 0.00010000 +Epoch 161, weight, value: tensor([[-4.9964e-02, -5.7572e-02, -1.2064e-01, ..., 1.0325e-01, + -3.6575e-02, 3.3500e-02], + [-6.4164e-02, 8.5925e-02, -6.2537e-02, ..., 3.2721e-02, + -4.6748e-02, -7.8732e-02], + [-4.0277e-02, 7.3879e-02, -8.8294e-02, ..., -2.9618e-02, + 2.6307e-02, 1.0586e-01], + ..., + [-9.7309e-04, 4.7186e-02, -3.3119e-03, ..., -9.1997e-04, + 2.5590e-02, -4.6462e-02], + [ 7.5950e-02, -7.5854e-02, 2.3652e-02, ..., 2.3576e-05, + 6.5415e-03, -4.5078e-02], + [-2.2629e-02, -6.4566e-02, 6.6474e-02, ..., -4.9499e-02, + 2.0422e-02, -1.2749e-02]], device='cuda:0'), grad: tensor([[ 9.6038e-06, 5.5170e-04, -1.0891e-03, ..., 3.3331e-04, + 9.8038e-04, 2.5177e-03], + [ 2.5010e-04, 9.6703e-04, 5.0354e-04, ..., 1.5583e-03, + 8.8882e-04, 1.3180e-03], + [ 1.7128e-03, 6.7749e-03, -2.7990e-04, ..., -3.6030e-03, + 2.3861e-03, -1.4906e-03], + ..., + [-2.2392e-03, -1.2718e-02, -2.5234e-03, ..., -4.7188e-03, + -1.2001e-02, -8.1482e-03], + [ 3.2568e-04, 2.8443e-04, 6.3229e-04, ..., -4.7147e-05, + 1.1263e-03, 1.2722e-03], + [ 8.4877e-05, 7.0333e-04, 1.0099e-03, ..., 1.4963e-03, + 6.9761e-04, 9.1410e-04]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0167, -0.0183, -0.0078, 0.0063, -0.0202, -0.0135, 0.0163, -0.0007, + 0.0257, -0.0204], device='cuda:0'), grad: tensor([ 0.0012, 0.0191, -0.0032, 0.0320, 0.0143, 0.0138, -0.0118, -0.0406, + -0.0409, 0.0162], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 160---------------------------------------------------- +epoch 160, time 217.45, cls_loss 0.5346 cls_loss_mapping 0.0100 cls_loss_causal 0.5083 re_mapping 0.0089 re_causal 0.0226 /// teacc 98.95 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.0501, -0.0584, -0.1204, ..., 0.1041, -0.0357, 0.0338], + [-0.0646, 0.0862, -0.0629, ..., 0.0331, -0.0464, -0.0788], + [-0.0416, 0.0732, -0.0891, ..., -0.0296, 0.0262, 0.1058], + ..., + [-0.0012, 0.0488, -0.0034, ..., -0.0013, 0.0254, -0.0453], + [ 0.0752, -0.0765, 0.0237, ..., -0.0006, 0.0059, -0.0450], + [-0.0215, -0.0653, 0.0669, ..., -0.0494, 0.0205, -0.0130]], + device='cuda:0'), grad: tensor([[ 2.6608e-04, -8.1682e-04, -4.6387e-03, ..., -2.6016e-03, + -8.9216e-04, -8.1825e-04], + [ 5.3835e-04, 2.8934e-03, 6.6185e-04, ..., 3.3340e-03, + 1.8942e-04, 1.5652e-04], + [ 2.3043e-04, -8.5373e-03, -2.2575e-05, ..., 5.3551e-08, + -1.4610e-03, -1.9417e-03], + ..., + [ 2.5201e-04, -2.3918e-03, 1.1629e-04, ..., -1.0384e-02, + 8.8787e-04, 1.1196e-03], + [ 1.0080e-03, 1.7815e-03, 5.8270e-04, ..., 2.3804e-03, + 5.8842e-04, 4.4894e-04], + [ 3.5095e-04, 1.2922e-03, 5.6362e-04, ..., 1.4105e-03, + 3.8505e-04, 1.4842e-04]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0165, -0.0173, -0.0090, 0.0060, -0.0204, -0.0125, 0.0162, -0.0011, + 0.0254, -0.0197], device='cuda:0'), grad: tensor([-0.0073, 0.0016, -0.0118, 0.0206, 0.0162, -0.0085, 0.0178, -0.0409, + 0.0240, -0.0117], device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 215.99, cls_loss 0.5947 cls_loss_mapping 0.0096 cls_loss_causal 0.5629 re_mapping 0.0092 re_causal 0.0252 /// teacc 98.45 lr 0.00010000 +Epoch 163, weight, value: tensor([[-5.0213e-02, -5.7367e-02, -1.2123e-01, ..., 1.0451e-01, + -3.6532e-02, 3.3408e-02], + [-6.5297e-02, 8.6723e-02, -6.2957e-02, ..., 3.2431e-02, + -4.6607e-02, -7.8866e-02], + [-4.0379e-02, 7.4347e-02, -8.8462e-02, ..., -2.8959e-02, + 2.5777e-02, 1.0537e-01], + ..., + [-2.5674e-03, 4.8025e-02, -2.5567e-03, ..., -3.8986e-05, + 2.6235e-02, -4.6650e-02], + [ 7.5850e-02, -7.7078e-02, 2.3677e-02, ..., -5.7740e-04, + 7.0285e-03, -4.3390e-02], + [-2.0221e-02, -6.5053e-02, 6.6873e-02, ..., -5.0605e-02, + 2.0479e-02, -1.2347e-02]], device='cuda:0'), grad: tensor([[ 2.4819e-04, -1.7977e-03, -2.4962e-04, ..., -3.8409e-04, + -2.9564e-04, -2.4796e-03], + [ 1.2231e-04, -1.9312e-05, 6.3324e-04, ..., 1.1206e-03, + 2.1660e-04, 4.3631e-04], + [ 2.7966e-04, 5.3978e-04, 1.3459e-04, ..., 1.5459e-03, + -1.4491e-05, 5.7983e-04], + ..., + [-1.0452e-02, 3.0708e-04, -1.3168e-02, ..., 1.3781e-03, + -9.9106e-03, 7.5197e-04], + [ 1.5020e-03, 4.6825e-04, -5.6076e-03, ..., -1.3275e-03, + -6.8092e-04, 5.2691e-04], + [ 9.2010e-03, 6.2466e-04, 1.2222e-02, ..., 1.2045e-03, + 8.5754e-03, 4.9448e-04]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0168, -0.0178, -0.0081, 0.0062, -0.0205, -0.0126, 0.0151, -0.0012, + 0.0258, -0.0198], device='cuda:0'), grad: tensor([-0.0087, 0.0188, 0.0245, -0.0058, 0.0149, 0.0335, -0.0784, -0.0387, + -0.0035, 0.0434], device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 216.56, cls_loss 0.5495 cls_loss_mapping 0.0102 cls_loss_causal 0.5226 re_mapping 0.0087 re_causal 0.0217 /// teacc 98.72 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.0515, -0.0581, -0.1218, ..., 0.1042, -0.0371, 0.0338], + [-0.0656, 0.0860, -0.0638, ..., 0.0325, -0.0468, -0.0798], + [-0.0410, 0.0749, -0.0885, ..., -0.0294, 0.0272, 0.1060], + ..., + [-0.0018, 0.0497, -0.0031, ..., -0.0011, 0.0247, -0.0456], + [ 0.0766, -0.0773, 0.0241, ..., 0.0006, 0.0079, -0.0423], + [-0.0204, -0.0654, 0.0670, ..., -0.0506, 0.0203, -0.0130]], + device='cuda:0'), grad: tensor([[ 1.3714e-03, 4.1962e-05, 7.7248e-04, ..., 1.5898e-03, + 6.9046e-04, 1.3151e-03], + [-1.0204e-03, 2.5213e-05, 2.9296e-05, ..., -1.9348e-04, + 2.0176e-05, -5.9938e-04], + [ 4.5624e-03, 4.5700e-03, 1.1730e-03, ..., 1.1683e-03, + 2.6855e-03, 1.1559e-02], + ..., + [-1.1482e-03, 3.3766e-05, -4.5471e-03, ..., -1.7138e-03, + -4.3945e-03, -3.9368e-03], + [ 8.9109e-05, 7.3671e-04, -7.4673e-04, ..., 7.2575e-04, + -9.3126e-04, 8.5068e-04], + [-1.5343e-02, 3.3230e-05, -9.6226e-04, ..., -5.6124e-04, + 1.9703e-03, 1.8854e-03]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0168, -0.0184, -0.0085, 0.0058, -0.0200, -0.0124, 0.0150, -0.0007, + 0.0262, -0.0199], device='cuda:0'), grad: tensor([-0.0072, -0.0448, 0.0428, -0.0368, 0.0231, 0.0047, 0.0362, -0.0233, + 0.0111, -0.0058], device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 216.54, cls_loss 0.5566 cls_loss_mapping 0.0127 cls_loss_causal 0.5224 re_mapping 0.0092 re_causal 0.0230 /// teacc 98.72 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.0519, -0.0577, -0.1219, ..., 0.1035, -0.0386, 0.0335], + [-0.0650, 0.0860, -0.0654, ..., 0.0330, -0.0467, -0.0812], + [-0.0405, 0.0744, -0.0883, ..., -0.0296, 0.0266, 0.1062], + ..., + [-0.0017, 0.0507, -0.0027, ..., -0.0004, 0.0256, -0.0454], + [ 0.0769, -0.0785, 0.0254, ..., 0.0008, 0.0083, -0.0432], + [-0.0199, -0.0651, 0.0672, ..., -0.0509, 0.0203, -0.0132]], + device='cuda:0'), grad: tensor([[-1.0002e-02, 2.7227e-04, 6.5744e-05, ..., -5.3329e-03, + -8.1863e-03, -3.2616e-03], + [ 2.9302e-04, 6.9046e-04, 5.0974e-04, ..., 2.2793e-03, + 6.0320e-04, 3.3736e-04], + [ 1.1339e-03, 1.2312e-03, 3.8767e-04, ..., 2.5711e-03, + 1.4906e-03, 7.4005e-04], + ..., + [ 1.5438e-04, -2.1458e-03, 4.8971e-04, ..., 5.4407e-04, + -7.6151e-04, -5.3596e-04], + [ 5.4741e-04, -1.9217e-03, -1.5640e-03, ..., -8.2932e-03, + 9.9945e-04, -4.2558e-04], + [ 4.1270e-04, 4.0460e-04, 1.6460e-03, ..., 2.7981e-03, + 1.0147e-03, 4.0936e-04]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0162, -0.0184, -0.0083, 0.0062, -0.0205, -0.0124, 0.0148, -0.0007, + 0.0266, -0.0196], device='cuda:0'), grad: tensor([-0.0501, 0.0161, 0.0249, 0.0019, 0.0275, -0.0328, 0.0351, -0.0008, + -0.0432, 0.0216], device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 216.53, cls_loss 0.5554 cls_loss_mapping 0.0116 cls_loss_causal 0.5305 re_mapping 0.0092 re_causal 0.0225 /// teacc 98.77 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.0508, -0.0565, -0.1214, ..., 0.1028, -0.0389, 0.0338], + [-0.0662, 0.0859, -0.0669, ..., 0.0330, -0.0470, -0.0826], + [-0.0402, 0.0734, -0.0887, ..., -0.0287, 0.0279, 0.1061], + ..., + [-0.0008, 0.0515, -0.0028, ..., -0.0007, 0.0247, -0.0467], + [ 0.0764, -0.0791, 0.0251, ..., 0.0017, 0.0083, -0.0430], + [-0.0208, -0.0655, 0.0663, ..., -0.0521, 0.0193, -0.0125]], + device='cuda:0'), grad: tensor([[-1.0192e-04, 4.2701e-04, 3.1114e-04, ..., 6.1393e-05, + -6.5029e-05, -1.2708e-04], + [ 1.7390e-05, 4.3154e-04, 1.6499e-04, ..., 1.1339e-03, + 3.0017e-04, 9.2793e-04], + [ 1.8859e-04, 2.3746e-03, 8.6403e-04, ..., 2.8610e-03, + 1.8415e-03, 3.3283e-03], + ..., + [-2.5868e-05, -8.6975e-04, 1.6856e-04, ..., -4.1618e-03, + -5.1594e-04, -6.4850e-04], + [ 1.1415e-03, 7.6962e-04, 1.2102e-03, ..., 2.4853e-03, + 1.1578e-03, 2.1915e-03], + [-1.0723e-04, 1.4534e-03, 3.5172e-03, ..., 2.7637e-03, + 2.6379e-03, 1.8063e-03]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0161, -0.0187, -0.0079, 0.0059, -0.0205, -0.0134, 0.0150, -0.0004, + 0.0271, -0.0194], device='cuda:0'), grad: tensor([ 0.0028, 0.0072, 0.0230, 0.0109, -0.0269, -0.0098, -0.0174, -0.0220, + 0.0202, 0.0120], device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 217.03, cls_loss 0.5434 cls_loss_mapping 0.0078 cls_loss_causal 0.5155 re_mapping 0.0091 re_causal 0.0220 /// teacc 98.68 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.0515, -0.0564, -0.1219, ..., 0.1027, -0.0399, 0.0333], + [-0.0656, 0.0878, -0.0663, ..., 0.0346, -0.0478, -0.0834], + [-0.0397, 0.0737, -0.0884, ..., -0.0290, 0.0281, 0.1066], + ..., + [-0.0006, 0.0504, -0.0027, ..., -0.0014, 0.0247, -0.0472], + [ 0.0761, -0.0802, 0.0254, ..., 0.0011, 0.0088, -0.0420], + [-0.0213, -0.0650, 0.0661, ..., -0.0528, 0.0195, -0.0126]], + device='cuda:0'), grad: tensor([[-7.9012e-04, 9.3555e-04, -4.5180e-04, ..., 2.6751e-04, + -1.2064e-03, -2.7752e-04], + [ 2.3155e-03, 2.4867e-04, -1.9875e-03, ..., -2.4939e-04, + 8.8274e-05, -2.5005e-03], + [ 1.4906e-03, 1.9855e-03, 1.2197e-03, ..., 1.8845e-03, + 9.8324e-04, 9.0122e-04], + ..., + [ 1.3218e-03, -2.4395e-03, -1.7195e-03, ..., -5.7983e-03, + -3.4561e-03, -4.8310e-05], + [-7.7705e-03, 2.6345e-04, 1.2674e-03, ..., 1.5507e-03, + 1.6918e-03, 1.6890e-03], + [ 1.0099e-03, -6.2704e-04, -4.1618e-03, ..., 1.4620e-03, + -1.9640e-05, -2.4986e-04]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0162, -0.0181, -0.0085, 0.0058, -0.0204, -0.0130, 0.0152, -0.0003, + 0.0268, -0.0198], device='cuda:0'), grad: tensor([ 0.0080, 0.0034, -0.0005, -0.0017, 0.0290, -0.0134, -0.0053, -0.0387, + 0.0050, 0.0143], device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 217.13, cls_loss 0.5066 cls_loss_mapping 0.0102 cls_loss_causal 0.4844 re_mapping 0.0090 re_causal 0.0230 /// teacc 98.57 lr 0.00010000 +Epoch 168, weight, value: tensor([[-5.2096e-02, -5.6804e-02, -1.2302e-01, ..., 1.0298e-01, + -4.0624e-02, 3.4111e-02], + [-6.5804e-02, 8.5929e-02, -6.6367e-02, ..., 3.4138e-02, + -4.9299e-02, -8.3382e-02], + [-4.0069e-02, 7.3088e-02, -8.7793e-02, ..., -2.9245e-02, + 2.7887e-02, 1.0697e-01], + ..., + [ 6.3387e-05, 5.1875e-02, -4.0405e-03, ..., -2.2354e-03, + 2.4219e-02, -4.8625e-02], + [ 7.4799e-02, -7.8198e-02, 2.6176e-02, ..., 1.5862e-03, + 8.9775e-03, -4.2728e-02], + [-2.0659e-02, -6.4397e-02, 6.6644e-02, ..., -5.2116e-02, + 2.0024e-02, -1.0966e-02]], device='cuda:0'), grad: tensor([[-1.5688e-03, -2.5253e-03, -3.1776e-03, ..., -9.4795e-04, + 4.0699e-07, -3.4008e-03], + [ 1.4429e-03, 7.3318e-03, 2.0180e-03, ..., 1.1625e-03, + 2.1175e-05, 1.3609e-03], + [ 1.2970e-04, 8.2016e-04, -1.5211e-03, ..., -1.4105e-03, + -1.0717e-04, 2.2459e-04], + ..., + [ 3.4070e-04, -1.2596e-02, -1.8143e-02, ..., -2.3041e-02, + -2.7895e-04, -3.0708e-03], + [ 1.5783e-04, 5.4026e-04, 1.5914e-04, ..., 1.0109e-03, + -4.3654e-04, 1.1196e-03], + [ 2.1887e-04, 1.0471e-03, 1.7786e-03, ..., 1.5831e-03, + 5.4646e-04, 1.3943e-03]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0160, -0.0184, -0.0082, 0.0059, -0.0210, -0.0142, 0.0162, -0.0004, + 0.0270, -0.0192], device='cuda:0'), grad: tensor([-0.0051, 0.0453, -0.0114, -0.0116, 0.0464, 0.0179, -0.0758, -0.0585, + 0.0237, 0.0292], device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 217.26, cls_loss 0.5485 cls_loss_mapping 0.0120 cls_loss_causal 0.5206 re_mapping 0.0090 re_causal 0.0220 /// teacc 98.47 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.0537, -0.0580, -0.1234, ..., 0.1025, -0.0417, 0.0312], + [-0.0650, 0.0868, -0.0662, ..., 0.0332, -0.0493, -0.0832], + [-0.0407, 0.0734, -0.0868, ..., -0.0290, 0.0285, 0.1069], + ..., + [-0.0006, 0.0529, -0.0046, ..., -0.0012, 0.0243, -0.0469], + [ 0.0752, -0.0779, 0.0253, ..., 0.0026, 0.0095, -0.0424], + [-0.0206, -0.0663, 0.0665, ..., -0.0525, 0.0194, -0.0113]], + device='cuda:0'), grad: tensor([[-3.8319e-03, -1.7004e-03, -2.3937e-03, ..., -3.3092e-03, + -1.8101e-03, -2.6493e-03], + [ 2.2678e-03, 7.4043e-03, 3.3498e-04, ..., 7.5531e-03, + 2.2650e-05, -4.1664e-05], + [ 1.2188e-03, -1.5974e-03, -8.4877e-04, ..., 2.7108e-04, + 1.5345e-03, -1.5373e-03], + ..., + [ 5.3930e-04, 2.2161e-04, 4.8375e-04, ..., 8.1825e-04, + -5.7697e-04, 6.1417e-04], + [ 1.4877e-03, 4.9591e-04, 4.8310e-05, ..., -1.9300e-04, + 1.0365e-04, 5.7364e-04], + [ 1.6481e-05, 1.4076e-03, -2.8443e-04, ..., 1.0967e-03, + -6.1214e-05, 1.5965e-03]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0147, -0.0184, -0.0073, 0.0061, -0.0204, -0.0136, 0.0162, -0.0002, + 0.0261, -0.0194], device='cuda:0'), grad: tensor([-0.0253, 0.0189, 0.0039, -0.0388, 0.0167, 0.0433, -0.0082, -0.0089, + -0.0157, 0.0140], device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 216.70, cls_loss 0.5160 cls_loss_mapping 0.0078 cls_loss_causal 0.4881 re_mapping 0.0097 re_causal 0.0252 /// teacc 98.68 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.0527, -0.0569, -0.1220, ..., 0.1047, -0.0403, 0.0326], + [-0.0661, 0.0861, -0.0665, ..., 0.0338, -0.0499, -0.0835], + [-0.0408, 0.0735, -0.0879, ..., -0.0286, 0.0277, 0.1072], + ..., + [ 0.0007, 0.0531, -0.0032, ..., -0.0018, 0.0255, -0.0473], + [ 0.0743, -0.0805, 0.0247, ..., 0.0024, 0.0089, -0.0436], + [-0.0201, -0.0661, 0.0664, ..., -0.0522, 0.0191, -0.0120]], + device='cuda:0'), grad: tensor([[-2.2717e-03, 2.2638e-04, -2.3632e-03, ..., 7.4482e-04, + -3.1719e-03, 5.8442e-05], + [ 4.0293e-05, -1.7738e-04, 2.8872e-04, ..., 2.0046e-03, + -4.1389e-04, 1.6079e-03], + [ 5.4932e-04, -3.9215e-03, -3.6488e-03, ..., -5.2834e-03, + 3.9711e-03, 6.4697e-03], + ..., + [-1.5190e-02, -1.1925e-02, 6.5880e-03, ..., 7.0686e-03, + 2.8992e-03, 2.1820e-03], + [ 8.6212e-04, 5.5885e-04, 1.6870e-03, ..., 3.0746e-03, + 1.6565e-03, -2.0332e-03], + [ 1.5549e-02, 1.3832e-02, -3.1242e-03, ..., 2.4338e-03, + -4.9770e-05, 1.5812e-03]], device='cuda:0') +Epoch 170, bias, value: tensor([ 1.6531e-02, -1.8142e-02, -7.5388e-03, 5.4377e-03, -2.0373e-02, + -1.3327e-02, 1.5934e-02, -6.0533e-05, 2.5263e-02, -1.9824e-02], + device='cuda:0'), grad: tensor([ 0.0200, -0.0074, 0.0156, 0.0079, 0.0034, -0.0515, 0.0028, 0.0131, + -0.0295, 0.0257], device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 216.89, cls_loss 0.5237 cls_loss_mapping 0.0069 cls_loss_causal 0.4897 re_mapping 0.0085 re_causal 0.0212 /// teacc 98.61 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.0525, -0.0595, -0.1213, ..., 0.1050, -0.0400, 0.0321], + [-0.0672, 0.0869, -0.0658, ..., 0.0344, -0.0496, -0.0846], + [-0.0412, 0.0754, -0.0873, ..., -0.0275, 0.0277, 0.1080], + ..., + [-0.0007, 0.0524, -0.0038, ..., -0.0024, 0.0249, -0.0478], + [ 0.0756, -0.0808, 0.0254, ..., 0.0026, 0.0085, -0.0437], + [-0.0201, -0.0667, 0.0658, ..., -0.0523, 0.0199, -0.0130]], + device='cuda:0'), grad: tensor([[ 1.8549e-04, 7.0572e-04, -7.9651e-03, ..., 3.7613e-03, + -9.3918e-03, -5.0592e-04], + [ 7.7295e-04, 8.9073e-04, 1.0900e-03, ..., -9.5606e-05, + 1.4389e-04, 5.3835e-04], + [ 4.7898e-04, 1.9026e-03, 2.9926e-03, ..., 4.3335e-03, + 3.1261e-03, 4.4098e-03], + ..., + [ 6.7568e-04, -2.3003e-03, 9.3079e-03, ..., -1.6212e-03, + 4.3335e-03, -1.2360e-03], + [ 4.7231e-04, 2.2392e-03, -2.6703e-03, ..., -6.7215e-03, + 1.1292e-03, -1.4057e-03], + [ 5.5838e-04, 1.2274e-03, -8.0338e-03, ..., 2.2964e-03, + -5.5275e-03, 3.2845e-03]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0153, -0.0183, -0.0062, 0.0051, -0.0203, -0.0139, 0.0163, -0.0006, + 0.0253, -0.0190], device='cuda:0'), grad: tensor([ 0.0048, -0.0094, 0.0294, 0.0421, -0.0016, -0.0262, -0.0169, -0.0029, + -0.0281, 0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 216.97, cls_loss 0.5175 cls_loss_mapping 0.0084 cls_loss_causal 0.4886 re_mapping 0.0093 re_causal 0.0240 /// teacc 98.64 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.0524, -0.0597, -0.1203, ..., 0.1040, -0.0399, 0.0323], + [-0.0670, 0.0883, -0.0656, ..., 0.0347, -0.0487, -0.0837], + [-0.0403, 0.0748, -0.0887, ..., -0.0261, 0.0279, 0.1075], + ..., + [-0.0004, 0.0526, -0.0046, ..., -0.0019, 0.0245, -0.0480], + [ 0.0756, -0.0805, 0.0252, ..., 0.0034, 0.0073, -0.0431], + [-0.0194, -0.0673, 0.0673, ..., -0.0524, 0.0209, -0.0132]], + device='cuda:0'), grad: tensor([[ 2.0456e-04, -1.5535e-03, 3.0255e-04, ..., -3.4637e-03, + 6.3419e-05, 1.2474e-03], + [ 2.8801e-04, 1.1247e-04, 5.8079e-04, ..., -1.6916e-04, + 4.2975e-05, -5.0211e-04], + [-5.1349e-05, -1.8358e-03, -5.8022e-03, ..., -2.4033e-03, + -9.6970e-03, -1.0880e-02], + ..., + [ 2.1343e-03, 2.9774e-03, 1.5472e-02, ..., 3.5686e-03, + 1.2718e-02, 1.3016e-02], + [-2.5330e-03, 3.4666e-04, -2.3880e-03, ..., -1.9348e-04, + 4.0221e-04, -5.2872e-03], + [ 2.9812e-03, -1.2293e-03, -9.2793e-04, ..., -2.8467e-04, + -7.3051e-04, -8.4448e-04]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0159, -0.0187, -0.0061, 0.0049, -0.0196, -0.0150, 0.0159, -0.0001, + 0.0253, -0.0187], device='cuda:0'), grad: tensor([ 0.0003, 0.0037, -0.0591, 0.0356, 0.0047, -0.0100, -0.0074, 0.0931, + -0.0591, -0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 217.02, cls_loss 0.5711 cls_loss_mapping 0.0066 cls_loss_causal 0.5463 re_mapping 0.0091 re_causal 0.0243 /// teacc 98.72 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.0537, -0.0610, -0.1194, ..., 0.1034, -0.0396, 0.0318], + [-0.0664, 0.0864, -0.0664, ..., 0.0348, -0.0483, -0.0834], + [-0.0406, 0.0761, -0.0892, ..., -0.0270, 0.0272, 0.1079], + ..., + [-0.0002, 0.0532, -0.0042, ..., 0.0003, 0.0252, -0.0489], + [ 0.0757, -0.0808, 0.0271, ..., 0.0034, 0.0082, -0.0439], + [-0.0194, -0.0670, 0.0673, ..., -0.0545, 0.0213, -0.0144]], + device='cuda:0'), grad: tensor([[ 4.8548e-05, 7.2289e-04, 6.2037e-04, ..., 1.6632e-03, + 2.5392e-04, 4.3869e-04], + [ 1.1140e-04, -3.6564e-03, -4.9057e-03, ..., -2.1076e-03, + 3.9864e-04, -2.4211e-04], + [ 1.9205e-04, 1.5612e-03, 1.4248e-03, ..., 2.9202e-03, + 1.0490e-03, 1.0986e-03], + ..., + [ 3.3894e-03, -2.4853e-03, -4.5967e-03, ..., -6.2447e-03, + -1.3771e-03, -3.5429e-04], + [ 1.8656e-04, -4.4708e-03, -2.7390e-03, ..., -1.1543e-02, + -2.8687e-03, -5.0087e-03], + [ 1.3947e-04, 1.8587e-03, 2.1553e-03, ..., 3.5629e-03, + 1.5278e-03, 1.3571e-03]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0145, -0.0184, -0.0062, 0.0036, -0.0194, -0.0138, 0.0166, 0.0008, + 0.0249, -0.0190], device='cuda:0'), grad: tensor([ 0.0094, -0.0169, 0.0157, -0.0365, 0.0170, 0.0179, 0.0289, -0.0177, + -0.0376, 0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 216.51, cls_loss 0.5517 cls_loss_mapping 0.0082 cls_loss_causal 0.5258 re_mapping 0.0088 re_causal 0.0231 /// teacc 98.78 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.0545, -0.0616, -0.1194, ..., 0.1021, -0.0403, 0.0316], + [-0.0660, 0.0868, -0.0676, ..., 0.0346, -0.0484, -0.0832], + [-0.0404, 0.0760, -0.0894, ..., -0.0275, 0.0274, 0.1081], + ..., + [-0.0013, 0.0522, -0.0037, ..., -0.0012, 0.0251, -0.0497], + [ 0.0769, -0.0815, 0.0274, ..., 0.0044, 0.0083, -0.0437], + [-0.0175, -0.0660, 0.0676, ..., -0.0529, 0.0212, -0.0147]], + device='cuda:0'), grad: tensor([[ 0.0012, 0.0007, 0.0001, ..., 0.0006, 0.0005, -0.0003], + [ 0.0013, -0.0033, 0.0005, ..., -0.0130, 0.0005, -0.0004], + [ 0.0028, 0.0021, 0.0012, ..., 0.0015, 0.0011, -0.0003], + ..., + [ 0.0041, 0.0049, 0.0074, ..., -0.0038, 0.0038, 0.0005], + [-0.0081, 0.0044, 0.0020, ..., 0.0108, -0.0003, -0.0030], + [-0.0071, -0.0122, -0.0221, ..., 0.0004, -0.0130, -0.0015]], + device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0132, -0.0188, -0.0061, 0.0053, -0.0192, -0.0142, 0.0159, -0.0003, + 0.0263, -0.0189], device='cuda:0'), grad: tensor([-0.0149, -0.0280, -0.0016, 0.0397, 0.0361, 0.0116, 0.0163, -0.0202, + 0.0090, -0.0481], device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 216.84, cls_loss 0.5201 cls_loss_mapping 0.0094 cls_loss_causal 0.4939 re_mapping 0.0089 re_causal 0.0218 /// teacc 98.78 lr 0.00010000 +Epoch 175, weight, value: tensor([[-5.3786e-02, -6.1600e-02, -1.1907e-01, ..., 1.0277e-01, + -3.9251e-02, 3.2854e-02], + [-6.7441e-02, 8.7620e-02, -6.7223e-02, ..., 3.5422e-02, + -4.8855e-02, -8.3024e-02], + [-3.9831e-02, 7.6429e-02, -9.0263e-02, ..., -2.7734e-02, + 2.6703e-02, 1.0825e-01], + ..., + [ 1.1814e-05, 5.2565e-02, -2.8245e-03, ..., -2.5271e-03, + 2.6064e-02, -4.9779e-02], + [ 7.6373e-02, -8.3185e-02, 2.8016e-02, ..., 4.9174e-03, + 7.9745e-03, -4.4562e-02], + [-1.7434e-02, -6.5576e-02, 6.6810e-02, ..., -5.3230e-02, + 2.0653e-02, -1.5435e-02]], device='cuda:0'), grad: tensor([[ 6.1083e-04, 8.6054e-06, 2.1243e-04, ..., 9.9659e-04, + 4.1938e-04, 1.1358e-03], + [ 8.1158e-04, 3.8236e-05, 3.7980e-04, ..., 1.4305e-03, + 5.1594e-04, 8.2397e-04], + [ 6.4373e-04, 3.3230e-05, 1.0037e-04, ..., 2.6302e-03, + 1.1301e-03, 3.9787e-03], + ..., + [ 6.1369e-04, 4.7892e-05, 3.5400e-03, ..., 2.4338e-03, + 3.7899e-03, 1.0891e-03], + [ 6.6223e-03, -1.9658e-04, -8.2207e-04, ..., -3.7956e-03, + -5.2376e-03, -2.9259e-03], + [ 1.5249e-03, 2.2948e-05, -4.7646e-03, ..., -4.4212e-03, + -4.0741e-03, 8.4162e-04]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0145, -0.0185, -0.0063, 0.0056, -0.0185, -0.0145, 0.0150, -0.0003, + 0.0260, -0.0195], device='cuda:0'), grad: tensor([ 0.0159, 0.0166, 0.0270, 0.0167, -0.0441, -0.0387, 0.0097, 0.0189, + -0.0029, -0.0191], device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 216.72, cls_loss 0.5359 cls_loss_mapping 0.0097 cls_loss_causal 0.5022 re_mapping 0.0089 re_causal 0.0221 /// teacc 98.54 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.0535, -0.0607, -0.1192, ..., 0.1033, -0.0403, 0.0323], + [-0.0678, 0.0869, -0.0691, ..., 0.0349, -0.0494, -0.0832], + [-0.0407, 0.0763, -0.0913, ..., -0.0283, 0.0270, 0.1086], + ..., + [ 0.0003, 0.0526, -0.0034, ..., -0.0038, 0.0247, -0.0507], + [ 0.0762, -0.0828, 0.0286, ..., 0.0070, 0.0078, -0.0442], + [-0.0172, -0.0656, 0.0671, ..., -0.0526, 0.0220, -0.0144]], + device='cuda:0'), grad: tensor([[-1.5783e-03, -7.3016e-05, -1.8148e-03, ..., -2.9068e-03, + -2.2564e-03, -2.6917e-04], + [ 2.7537e-05, -1.2958e-04, 6.8808e-04, ..., 1.1053e-03, + 2.4104e-04, 9.6941e-04], + [-3.0327e-04, 1.1856e-02, 9.3651e-04, ..., 1.2169e-03, + 2.3880e-03, 5.3864e-03], + ..., + [ 5.3501e-04, -9.3002e-03, 7.5188e-03, ..., 1.9064e-03, + -5.5389e-03, -7.1754e-03], + [ 2.6345e-04, 1.7869e-04, -1.4648e-03, ..., -2.4357e-03, + 1.4210e-03, 6.5374e-04], + [ 4.4322e-04, -3.0613e-03, -5.0964e-03, ..., -2.5387e-03, + -7.7820e-04, -3.8071e-03]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0144, -0.0186, -0.0054, 0.0054, -0.0191, -0.0146, 0.0151, -0.0004, + 0.0260, -0.0195], device='cuda:0'), grad: tensor([-0.0077, 0.0156, 0.0300, 0.0117, -0.0357, 0.0217, 0.0188, 0.0209, + -0.0102, -0.0650], device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 216.72, cls_loss 0.5581 cls_loss_mapping 0.0086 cls_loss_causal 0.5346 re_mapping 0.0082 re_causal 0.0206 /// teacc 98.76 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.0538, -0.0613, -0.1195, ..., 0.1040, -0.0397, 0.0317], + [-0.0688, 0.0884, -0.0697, ..., 0.0369, -0.0499, -0.0833], + [-0.0415, 0.0754, -0.0902, ..., -0.0295, 0.0279, 0.1093], + ..., + [ 0.0003, 0.0528, -0.0040, ..., -0.0041, 0.0246, -0.0512], + [ 0.0758, -0.0832, 0.0276, ..., 0.0069, 0.0075, -0.0434], + [-0.0162, -0.0663, 0.0667, ..., -0.0528, 0.0207, -0.0145]], + device='cuda:0'), grad: tensor([[-7.7069e-05, 2.9266e-05, 2.0832e-05, ..., 1.7967e-03, + -9.0301e-05, 7.8964e-04], + [ 4.9882e-06, 4.0561e-05, 2.2209e-04, ..., -1.7862e-03, + 5.6386e-05, -9.2840e-04], + [ 7.3425e-06, 4.4197e-05, 5.4932e-04, ..., 1.2236e-03, + 2.5535e-04, 2.4283e-04], + ..., + [ 1.1206e-05, 1.6674e-05, -9.9487e-03, ..., -6.0158e-03, + -6.3362e-03, -6.5842e-03], + [-1.1407e-05, 3.3617e-05, 7.8106e-04, ..., 2.3823e-03, + 4.0460e-04, 1.3218e-03], + [ 4.9084e-05, 4.2647e-05, 8.4305e-03, ..., 1.4219e-03, + 5.4779e-03, 4.4212e-03]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0140, -0.0174, -0.0064, 0.0051, -0.0195, -0.0137, 0.0152, -0.0007, + 0.0259, -0.0189], device='cuda:0'), grad: tensor([ 0.0144, -0.0055, -0.0123, -0.0163, 0.0141, -0.0131, 0.0023, -0.0321, + 0.0192, 0.0294], device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 216.27, cls_loss 0.5579 cls_loss_mapping 0.0091 cls_loss_causal 0.5280 re_mapping 0.0087 re_causal 0.0232 /// teacc 98.76 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.0544, -0.0607, -0.1195, ..., 0.1047, -0.0390, 0.0319], + [-0.0684, 0.0888, -0.0711, ..., 0.0372, -0.0518, -0.0844], + [-0.0400, 0.0756, -0.0906, ..., -0.0307, 0.0274, 0.1095], + ..., + [ 0.0013, 0.0523, -0.0035, ..., -0.0039, 0.0255, -0.0508], + [ 0.0754, -0.0840, 0.0290, ..., 0.0065, 0.0084, -0.0436], + [-0.0168, -0.0660, 0.0664, ..., -0.0544, 0.0198, -0.0151]], + device='cuda:0'), grad: tensor([[ 2.4866e-06, 5.1451e-04, 1.7729e-03, ..., 3.6163e-03, + -2.3251e-03, 4.1175e-04], + [ 2.9162e-05, -2.8872e-04, -8.3160e-03, ..., -8.7738e-03, + -4.4441e-03, -3.2063e-03], + [ 5.7131e-05, -8.2684e-04, 6.1798e-04, ..., -1.8539e-03, + 9.3269e-04, -1.7910e-03], + ..., + [-3.0212e-06, 8.7261e-04, 2.6989e-03, ..., 1.0910e-03, + 1.4248e-03, 9.1887e-04], + [ 3.4064e-05, 4.7898e-04, 1.4982e-03, ..., 4.3449e-03, + 7.4053e-04, 7.9727e-04], + [-1.3030e-04, 5.9891e-04, -2.2373e-03, ..., 3.0327e-03, + -2.5439e-04, 8.9121e-04]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0146, -0.0173, -0.0066, 0.0050, -0.0198, -0.0124, 0.0146, -0.0003, + 0.0243, -0.0186], device='cuda:0'), grad: tensor([-0.0126, -0.0194, -0.0043, 0.0072, 0.0013, 0.0029, -0.0240, 0.0065, + 0.0284, 0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 216.59, cls_loss 0.5362 cls_loss_mapping 0.0053 cls_loss_causal 0.4976 re_mapping 0.0087 re_causal 0.0225 /// teacc 98.77 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.0556, -0.0609, -0.1203, ..., 0.1064, -0.0389, 0.0335], + [-0.0692, 0.0891, -0.0713, ..., 0.0353, -0.0525, -0.0851], + [-0.0394, 0.0751, -0.0906, ..., -0.0306, 0.0277, 0.1089], + ..., + [ 0.0015, 0.0531, -0.0030, ..., -0.0051, 0.0259, -0.0515], + [ 0.0754, -0.0846, 0.0294, ..., 0.0069, 0.0080, -0.0437], + [-0.0174, -0.0659, 0.0669, ..., -0.0529, 0.0200, -0.0152]], + device='cuda:0'), grad: tensor([[ 1.6856e-04, 1.0930e-05, -4.4918e-04, ..., -2.0838e-04, + -5.2719e-03, -2.0370e-03], + [ 2.9206e-04, -4.7348e-06, 6.1750e-05, ..., -6.2466e-04, + 4.0698e-04, -9.9373e-04], + [ 1.0662e-03, 4.0978e-05, 1.6356e-04, ..., 1.5059e-03, + -2.3384e-03, -1.9062e-04], + ..., + [ 3.8791e-04, -1.5825e-05, 1.3275e-03, ..., 1.4811e-03, + 1.7118e-03, 1.0509e-03], + [-4.2648e-03, -3.7956e-04, -6.6452e-03, ..., -3.7503e-04, + -3.3321e-03, -1.8768e-03], + [ 3.2997e-04, 2.9135e-04, 3.8395e-03, ..., 3.1815e-03, + 4.3983e-03, 2.1286e-03]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0147, -0.0177, -0.0064, 0.0042, -0.0190, -0.0136, 0.0150, -0.0003, + 0.0246, -0.0180], device='cuda:0'), grad: tensor([-0.0122, -0.0146, -0.0031, 0.0289, 0.0190, -0.0406, -0.0146, 0.0213, + -0.0210, 0.0368], device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 216.70, cls_loss 0.5338 cls_loss_mapping 0.0082 cls_loss_causal 0.5108 re_mapping 0.0084 re_causal 0.0220 /// teacc 98.87 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.0574, -0.0610, -0.1211, ..., 0.1068, -0.0397, 0.0346], + [-0.0702, 0.0899, -0.0713, ..., 0.0361, -0.0529, -0.0860], + [-0.0402, 0.0753, -0.0901, ..., -0.0290, 0.0293, 0.1090], + ..., + [ 0.0011, 0.0532, -0.0036, ..., -0.0060, 0.0260, -0.0522], + [ 0.0767, -0.0856, 0.0303, ..., 0.0061, 0.0072, -0.0444], + [-0.0178, -0.0663, 0.0668, ..., -0.0527, 0.0201, -0.0144]], + device='cuda:0'), grad: tensor([[-2.2173e-04, 1.6117e-04, 6.9618e-04, ..., -1.6642e-03, + -2.7084e-03, -6.2523e-03], + [-8.3494e-04, -1.3375e-04, 8.3637e-04, ..., 1.5192e-03, + 6.2323e-04, 1.2712e-03], + [ 2.3627e-04, 2.7895e-04, 4.0627e-04, ..., 1.9896e-04, + 2.8076e-03, 3.7327e-03], + ..., + [ 2.1264e-05, -8.7833e-04, -1.2627e-03, ..., -2.6083e-04, + -8.1873e-04, -1.6689e-04], + [-7.5245e-04, 1.8764e-04, -1.6766e-03, ..., 2.5368e-03, + -5.9605e-04, 2.4104e-04], + [ 1.0538e-03, -6.1274e-04, -1.9875e-03, ..., -1.0399e-02, + -2.8057e-03, -2.1381e-03]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0137, -0.0181, -0.0061, 0.0050, -0.0189, -0.0139, 0.0147, -0.0009, + 0.0245, -0.0167], device='cuda:0'), grad: tensor([ 0.0019, 0.0133, 0.0030, -0.0099, 0.0239, 0.0238, -0.0069, -0.0107, + -0.0044, -0.0340], device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 216.70, cls_loss 0.5505 cls_loss_mapping 0.0084 cls_loss_causal 0.5236 re_mapping 0.0083 re_causal 0.0220 /// teacc 98.79 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.0573, -0.0610, -0.1216, ..., 0.1072, -0.0401, 0.0337], + [-0.0699, 0.0895, -0.0706, ..., 0.0346, -0.0541, -0.0873], + [-0.0405, 0.0763, -0.0914, ..., -0.0296, 0.0294, 0.1095], + ..., + [ 0.0024, 0.0528, -0.0030, ..., -0.0070, 0.0253, -0.0527], + [ 0.0755, -0.0852, 0.0305, ..., 0.0079, 0.0081, -0.0427], + [-0.0170, -0.0658, 0.0672, ..., -0.0535, 0.0203, -0.0142]], + device='cuda:0'), grad: tensor([[ 2.6584e-04, 2.6971e-05, 1.0777e-03, ..., 2.2864e-04, + 9.3365e-04, 1.2994e-04], + [ 8.7738e-04, 1.2436e-03, 3.7813e-04, ..., 3.4630e-05, + 3.9458e-04, 2.7299e-05], + [ 3.6329e-05, -2.6379e-03, 8.6451e-04, ..., -3.9749e-03, + -1.7195e-03, -3.6449e-03], + ..., + [-1.1940e-03, -3.6583e-03, 7.8583e-03, ..., 2.9087e-03, + 3.0537e-03, 2.9802e-04], + [-6.3477e-03, 5.0402e-04, -1.6098e-02, ..., -1.2604e-02, + -1.0586e-04, 5.8556e-04], + [ 5.8289e-03, 3.2501e-03, 9.0866e-03, ..., 1.0361e-02, + -1.5221e-03, -7.7105e-04]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0142, -0.0186, -0.0070, 0.0058, -0.0193, -0.0138, 0.0149, -0.0011, + 0.0246, -0.0165], device='cuda:0'), grad: tensor([ 0.0093, 0.0124, -0.0033, 0.0012, 0.0059, -0.0360, -0.0089, 0.0155, + -0.0076, 0.0116], device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 216.89, cls_loss 0.5061 cls_loss_mapping 0.0071 cls_loss_causal 0.4733 re_mapping 0.0085 re_causal 0.0219 /// teacc 98.75 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.0554, -0.0608, -0.1223, ..., 0.1072, -0.0399, 0.0339], + [-0.0695, 0.0893, -0.0715, ..., 0.0343, -0.0555, -0.0873], + [-0.0402, 0.0768, -0.0909, ..., -0.0274, 0.0310, 0.1106], + ..., + [ 0.0018, 0.0529, -0.0035, ..., -0.0074, 0.0246, -0.0546], + [ 0.0756, -0.0854, 0.0310, ..., 0.0070, 0.0089, -0.0419], + [-0.0166, -0.0664, 0.0677, ..., -0.0547, 0.0205, -0.0150]], + device='cuda:0'), grad: tensor([[ 0.0001, -0.0033, -0.0016, ..., -0.0013, -0.0020, -0.0005], + [ 0.0002, -0.0011, 0.0006, ..., -0.0055, -0.0010, -0.0010], + [ 0.0002, -0.0008, -0.0056, ..., -0.0011, -0.0019, -0.0007], + ..., + [ 0.0001, 0.0003, -0.0155, ..., -0.0200, -0.0215, -0.0087], + [-0.0010, 0.0010, 0.0009, ..., 0.0028, 0.0010, 0.0014], + [ 0.0002, -0.0017, 0.0245, ..., 0.0126, 0.0309, 0.0062]], + device='cuda:0') +Epoch 182, bias, value: tensor([ 0.0155, -0.0178, -0.0071, 0.0056, -0.0195, -0.0136, 0.0144, -0.0018, + 0.0241, -0.0162], device='cuda:0'), grad: tensor([-0.0023, -0.0260, -0.0158, -0.0136, -0.0144, 0.0373, 0.0237, -0.0100, + 0.0167, 0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 216.45, cls_loss 0.5259 cls_loss_mapping 0.0082 cls_loss_causal 0.4944 re_mapping 0.0085 re_causal 0.0223 /// teacc 98.73 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.0565, -0.0609, -0.1228, ..., 0.1073, -0.0400, 0.0343], + [-0.0694, 0.0886, -0.0713, ..., 0.0345, -0.0554, -0.0869], + [-0.0403, 0.0776, -0.0920, ..., -0.0274, 0.0303, 0.1094], + ..., + [ 0.0018, 0.0528, -0.0033, ..., -0.0078, 0.0236, -0.0542], + [ 0.0753, -0.0857, 0.0316, ..., 0.0071, 0.0107, -0.0416], + [-0.0168, -0.0667, 0.0675, ..., -0.0546, 0.0197, -0.0154]], + device='cuda:0'), grad: tensor([[-2.5349e-03, -1.9416e-05, 4.1366e-04, ..., -1.6499e-03, + 2.5320e-04, -5.4054e-03], + [ 6.6900e-04, -1.8275e-04, 7.6723e-04, ..., 6.7091e-04, + 1.4486e-03, 1.3227e-03], + [ 6.9046e-04, -2.6345e-04, 3.9172e-04, ..., -8.9264e-04, + 1.4629e-03, -2.0199e-03], + ..., + [ 4.0388e-04, 7.5626e-04, 2.4090e-03, ..., 7.4053e-04, + 1.9083e-03, -3.7789e-04], + [-2.8362e-03, 1.7416e-04, -5.0659e-03, ..., -1.8990e-04, + -5.8975e-03, 4.6968e-04], + [ 3.5238e-04, -9.6321e-04, -9.8953e-03, ..., -1.5192e-03, + -1.2085e-02, 8.7142e-05]], device='cuda:0') +Epoch 183, bias, value: tensor([ 0.0154, -0.0178, -0.0069, 0.0054, -0.0196, -0.0141, 0.0151, -0.0024, + 0.0244, -0.0161], device='cuda:0'), grad: tensor([-0.0494, 0.0235, -0.0035, 0.0340, 0.0006, -0.0005, 0.0344, -0.0072, + 0.0007, -0.0326], device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 216.39, cls_loss 0.5598 cls_loss_mapping 0.0102 cls_loss_causal 0.5321 re_mapping 0.0083 re_causal 0.0213 /// teacc 98.68 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.0566, -0.0619, -0.1229, ..., 0.1072, -0.0410, 0.0338], + [-0.0701, 0.0880, -0.0720, ..., 0.0342, -0.0574, -0.0874], + [-0.0406, 0.0775, -0.0923, ..., -0.0271, 0.0302, 0.1096], + ..., + [ 0.0010, 0.0540, -0.0030, ..., -0.0074, 0.0241, -0.0539], + [ 0.0763, -0.0857, 0.0337, ..., 0.0055, 0.0118, -0.0410], + [-0.0174, -0.0661, 0.0677, ..., -0.0536, 0.0209, -0.0159]], + device='cuda:0'), grad: tensor([[ 4.6730e-03, 2.5535e-04, 5.9634e-05, ..., 2.9087e-04, + 1.0097e-04, 9.7752e-04], + [ 1.6766e-03, 4.0793e-04, 2.7514e-04, ..., 9.0837e-05, + 1.3809e-03, 2.7943e-04], + [ 2.3003e-03, -4.2648e-03, -9.3555e-04, ..., -9.2077e-04, + -5.2185e-03, -1.2264e-03], + ..., + [ 1.2245e-03, 5.3167e-04, -1.0979e-04, ..., 1.6041e-03, + 1.2627e-03, 4.5896e-04], + [ 1.3664e-02, 6.0034e-04, 2.7752e-04, ..., 1.7290e-03, + 1.0347e-03, 2.2163e-03], + [-1.3618e-02, 6.1846e-04, 3.8815e-04, ..., 1.5631e-03, + 1.2007e-03, 8.0287e-05]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0151, -0.0182, -0.0067, 0.0044, -0.0204, -0.0125, 0.0149, -0.0017, + 0.0241, -0.0156], device='cuda:0'), grad: tensor([ 0.0164, 0.0162, -0.0791, -0.0141, 0.0163, 0.0213, -0.0265, 0.0177, + 0.0391, -0.0073], device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 216.51, cls_loss 0.5316 cls_loss_mapping 0.0085 cls_loss_causal 0.4995 re_mapping 0.0083 re_causal 0.0213 /// teacc 98.73 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.0555, -0.0626, -0.1228, ..., 0.1079, -0.0408, 0.0338], + [-0.0698, 0.0877, -0.0720, ..., 0.0344, -0.0567, -0.0861], + [-0.0406, 0.0773, -0.0919, ..., -0.0273, 0.0314, 0.1095], + ..., + [ 0.0004, 0.0543, -0.0032, ..., -0.0077, 0.0240, -0.0549], + [ 0.0781, -0.0852, 0.0327, ..., 0.0045, 0.0097, -0.0405], + [-0.0192, -0.0660, 0.0672, ..., -0.0536, 0.0207, -0.0169]], + device='cuda:0'), grad: tensor([[ 3.1972e-04, 4.2701e-04, 1.2321e-03, ..., 7.2002e-05, + 8.9264e-04, 1.2779e-03], + [ 4.4870e-04, -2.4471e-03, -3.3436e-03, ..., -4.2877e-03, + -3.5343e-03, -3.4218e-03], + [ 8.8739e-04, 3.5095e-04, 7.3481e-04, ..., 7.7677e-04, + 5.5838e-04, 9.0694e-04], + ..., + [-1.4353e-03, 4.2105e-04, 3.6469e-02, ..., 8.0061e-04, + 2.2537e-02, 6.9094e-04], + [ 1.3599e-03, 1.7464e-04, 8.9788e-04, ..., -4.0817e-04, + 4.9019e-04, 6.7043e-04], + [ 8.4162e-04, -4.2534e-04, -4.1260e-02, ..., -1.3292e-04, + -2.2858e-02, -1.1787e-03]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0143, -0.0180, -0.0059, 0.0047, -0.0197, -0.0129, 0.0150, -0.0021, + 0.0236, -0.0156], device='cuda:0'), grad: tensor([-0.0121, -0.0435, 0.0147, 0.0080, 0.0203, 0.0079, 0.0191, 0.0329, + 0.0166, -0.0637], device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 217.09, cls_loss 0.5298 cls_loss_mapping 0.0089 cls_loss_causal 0.4985 re_mapping 0.0083 re_causal 0.0217 /// teacc 98.82 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.0554, -0.0635, -0.1232, ..., 0.1071, -0.0406, 0.0322], + [-0.0711, 0.0874, -0.0724, ..., 0.0347, -0.0562, -0.0859], + [-0.0394, 0.0782, -0.0924, ..., -0.0269, 0.0310, 0.1090], + ..., + [-0.0004, 0.0537, -0.0032, ..., -0.0086, 0.0243, -0.0542], + [ 0.0797, -0.0851, 0.0331, ..., 0.0051, 0.0096, -0.0401], + [-0.0183, -0.0666, 0.0677, ..., -0.0539, 0.0210, -0.0144]], + device='cuda:0'), grad: tensor([[ 9.7573e-05, 3.0208e-04, 1.8053e-03, ..., 2.8000e-03, + 4.0283e-03, 1.6632e-03], + [ 9.3102e-05, 3.7074e-05, 3.1519e-04, ..., 1.0233e-03, + -1.0723e-04, 1.5342e-04], + [ 3.0732e-04, 3.4785e-04, 1.0872e-03, ..., 2.3632e-03, + 3.1776e-03, 1.7843e-03], + ..., + [ 1.1361e-04, -6.6853e-04, -1.1873e-03, ..., -4.8141e-03, + -5.2338e-03, -5.1842e-03], + [-1.9054e-03, 2.5845e-04, -1.0109e-02, ..., -2.0657e-03, + -5.4245e-03, -1.1568e-03], + [ 3.1424e-04, -2.2640e-03, -1.7033e-03, ..., -7.3967e-03, + -1.2543e-02, 5.6505e-04]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0134, -0.0175, -0.0059, 0.0058, -0.0199, -0.0130, 0.0151, -0.0033, + 0.0238, -0.0150], device='cuda:0'), grad: tensor([ 0.0196, -0.0168, 0.0237, -0.0035, 0.0232, 0.0213, 0.0219, -0.0114, + -0.0052, -0.0729], device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 217.13, cls_loss 0.5238 cls_loss_mapping 0.0063 cls_loss_causal 0.4986 re_mapping 0.0078 re_causal 0.0205 /// teacc 98.80 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.0546, -0.0650, -0.1233, ..., 0.1075, -0.0406, 0.0320], + [-0.0707, 0.0879, -0.0712, ..., 0.0354, -0.0560, -0.0856], + [-0.0401, 0.0783, -0.0930, ..., -0.0280, 0.0301, 0.1085], + ..., + [-0.0009, 0.0534, -0.0028, ..., -0.0072, 0.0259, -0.0534], + [ 0.0794, -0.0859, 0.0328, ..., 0.0055, 0.0084, -0.0399], + [-0.0177, -0.0656, 0.0680, ..., -0.0550, 0.0210, -0.0140]], + device='cuda:0'), grad: tensor([[ 3.0947e-04, 2.7075e-05, -2.2769e-04, ..., 3.5787e-04, + -9.7942e-04, -7.9203e-04], + [ 2.4652e-04, -5.2303e-05, 1.7834e-04, ..., 1.1492e-03, + 1.0085e-04, 1.7405e-04], + [ 6.6090e-04, -1.9765e-04, 2.7561e-04, ..., 4.9174e-05, + 1.5450e-04, -3.2663e-04], + ..., + [ 7.1812e-04, -7.7343e-04, 1.3800e-03, ..., -4.0665e-03, + 2.6093e-03, -1.0643e-03], + [-7.5417e-03, 5.0306e-04, 1.2217e-03, ..., -1.0786e-03, + 2.0866e-03, 8.1062e-04], + [ 3.3319e-05, 3.6895e-05, -2.8172e-03, ..., 1.8826e-03, + -3.6430e-03, 6.2513e-04]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0127, -0.0177, -0.0059, 0.0054, -0.0202, -0.0125, 0.0160, -0.0022, + 0.0231, -0.0153], device='cuda:0'), grad: tensor([-0.0004, 0.0049, 0.0039, -0.0011, 0.0089, -0.0026, 0.0063, -0.0314, + 0.0021, 0.0094], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 186---------------------------------------------------- +epoch 186, time 217.19, cls_loss 0.5459 cls_loss_mapping 0.0088 cls_loss_causal 0.5202 re_mapping 0.0081 re_causal 0.0212 /// teacc 99.01 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.0537, -0.0658, -0.1224, ..., 0.1071, -0.0395, 0.0321], + [-0.0705, 0.0893, -0.0694, ..., 0.0356, -0.0550, -0.0857], + [-0.0408, 0.0782, -0.0940, ..., -0.0265, 0.0300, 0.1088], + ..., + [-0.0010, 0.0532, -0.0027, ..., -0.0075, 0.0259, -0.0529], + [ 0.0798, -0.0857, 0.0336, ..., 0.0051, 0.0086, -0.0403], + [-0.0201, -0.0657, 0.0668, ..., -0.0556, 0.0192, -0.0143]], + device='cuda:0'), grad: tensor([[-5.0783e-04, 8.4460e-05, -6.0415e-04, ..., -8.4400e-05, + -2.0428e-03, 1.7512e-04], + [ 5.2376e-03, 3.9601e-04, 6.5975e-06, ..., 1.2894e-03, + 1.3018e-04, 1.0490e-04], + [ 1.3316e-04, 1.6487e-04, 2.1070e-05, ..., 1.0624e-03, + 3.3641e-04, 2.2125e-04], + ..., + [ 3.6860e-04, 3.6597e-04, -5.0366e-05, ..., 1.2550e-03, + 1.1027e-04, 7.2122e-05], + [ 2.4357e-03, 5.2786e-04, 1.7166e-04, ..., 1.8244e-03, + 7.4530e-04, 1.4353e-04], + [-2.0508e-02, 2.0671e-04, -4.9133e-02, ..., 7.1907e-04, + -2.0538e-02, 9.5665e-05]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0129, -0.0171, -0.0056, 0.0063, -0.0197, -0.0130, 0.0161, -0.0024, + 0.0215, -0.0155], device='cuda:0'), grad: tensor([ 0.0079, -0.0019, 0.0124, -0.0211, 0.0409, -0.0156, 0.0110, -0.0192, + 0.0242, -0.0387], device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 216.63, cls_loss 0.5781 cls_loss_mapping 0.0086 cls_loss_causal 0.5458 re_mapping 0.0078 re_causal 0.0208 /// teacc 98.91 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.0549, -0.0660, -0.1229, ..., 0.1066, -0.0395, 0.0323], + [-0.0704, 0.0900, -0.0695, ..., 0.0356, -0.0550, -0.0869], + [-0.0398, 0.0780, -0.0948, ..., -0.0269, 0.0297, 0.1092], + ..., + [-0.0006, 0.0539, -0.0033, ..., -0.0072, 0.0251, -0.0531], + [ 0.0797, -0.0854, 0.0331, ..., 0.0054, 0.0078, -0.0403], + [-0.0215, -0.0667, 0.0666, ..., -0.0550, 0.0190, -0.0149]], + device='cuda:0'), grad: tensor([[-2.0142e-03, 1.9979e-04, -6.0997e-03, ..., -4.9323e-05, + 8.1635e-04, -1.4267e-03], + [ 5.4073e-04, 2.4676e-04, 1.5697e-03, ..., 2.2149e-04, + 4.6301e-04, 2.3019e-04], + [-3.1776e-03, 3.1352e-04, 9.0313e-04, ..., -3.6883e-04, + -5.5466e-03, 7.2765e-04], + ..., + [ 1.6146e-03, -8.9741e-04, 2.8133e-03, ..., -1.3390e-03, + 1.9989e-03, -1.0854e-04], + [ 3.8266e-04, 2.0540e-04, 7.3576e-04, ..., 1.2426e-03, + 5.9891e-04, 8.8835e-04], + [ 4.1699e-04, 4.7350e-04, 2.5864e-03, ..., 3.5596e-04, + 6.6519e-04, 2.9445e-04]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0128, -0.0180, -0.0051, 0.0061, -0.0196, -0.0128, 0.0152, -0.0016, + 0.0223, -0.0161], device='cuda:0'), grad: tensor([-0.0284, 0.0199, -0.0078, 0.0098, 0.0029, 0.0111, -0.0146, -0.0305, + 0.0173, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 216.70, cls_loss 0.5708 cls_loss_mapping 0.0074 cls_loss_causal 0.5335 re_mapping 0.0082 re_causal 0.0211 /// teacc 98.80 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.0544, -0.0669, -0.1239, ..., 0.1072, -0.0401, 0.0317], + [-0.0716, 0.0914, -0.0700, ..., 0.0359, -0.0551, -0.0859], + [-0.0407, 0.0780, -0.0945, ..., -0.0261, 0.0293, 0.1094], + ..., + [-0.0012, 0.0538, -0.0028, ..., -0.0077, 0.0262, -0.0535], + [ 0.0794, -0.0864, 0.0326, ..., 0.0055, 0.0066, -0.0414], + [-0.0212, -0.0674, 0.0674, ..., -0.0559, 0.0206, -0.0147]], + device='cuda:0'), grad: tensor([[ 2.0719e-04, 7.3767e-04, 3.0354e-05, ..., 2.5177e-03, + 1.3752e-03, 6.5851e-04], + [-7.0953e-04, -2.0523e-03, -1.1122e-04, ..., -1.0178e-02, + 6.3133e-04, -1.5821e-03], + [ 3.5596e-04, 1.5516e-03, 2.7728e-04, ..., -9.7351e-03, + -1.3802e-02, -4.5128e-03], + ..., + [ 2.9182e-04, 1.9569e-03, -5.5224e-05, ..., 7.5989e-03, + 7.1487e-03, 2.0924e-03], + [ 2.3878e-04, 1.0033e-03, 3.8099e-04, ..., 2.0618e-03, + 1.8129e-03, 3.5644e-04], + [ 3.2592e-04, 1.3329e-05, 5.1575e-03, ..., 2.3518e-03, + 2.4872e-03, 4.8661e-04]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0138, -0.0172, -0.0054, 0.0066, -0.0212, -0.0133, 0.0157, -0.0019, + 0.0222, -0.0157], device='cuda:0'), grad: tensor([ 0.0212, -0.0397, 0.0169, -0.0263, -0.0495, -0.0081, 0.0260, 0.0393, + 0.0213, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 216.67, cls_loss 0.5326 cls_loss_mapping 0.0066 cls_loss_causal 0.5107 re_mapping 0.0082 re_causal 0.0210 /// teacc 98.77 lr 0.00010000 +Epoch 191, weight, value: tensor([[-5.5098e-02, -6.7679e-02, -1.2387e-01, ..., 1.0701e-01, + -4.0007e-02, 3.1602e-02], + [-7.2165e-02, 9.1760e-02, -7.1212e-02, ..., 3.6626e-02, + -5.6083e-02, -8.6049e-02], + [-4.0154e-02, 7.8468e-02, -9.5983e-02, ..., -2.6147e-02, + 2.9528e-02, 1.0939e-01], + ..., + [ 2.9341e-06, 5.3700e-02, -2.1607e-03, ..., -7.5489e-03, + 2.5882e-02, -5.3486e-02], + [ 7.9024e-02, -8.5857e-02, 3.2170e-02, ..., 5.7040e-03, + 6.4229e-03, -4.0594e-02], + [-2.0709e-02, -6.8212e-02, 6.7334e-02, ..., -5.5902e-02, + 2.0720e-02, -1.3800e-02]], device='cuda:0'), grad: tensor([[ 3.5691e-04, 5.9080e-04, 3.8296e-05, ..., 2.5520e-03, + 1.2522e-03, 6.5565e-04], + [ 2.2945e-03, 2.9349e-04, -1.6159e-02, ..., -1.4862e-02, + -6.1035e-03, 2.5606e-04], + [ 3.0065e-04, 2.8019e-03, 4.1771e-04, ..., 2.1362e-03, + 6.1989e-03, 9.3508e-04], + ..., + [ 9.0170e-04, 1.4105e-03, 1.7519e-03, ..., 3.5095e-03, + 2.2659e-03, 1.2188e-03], + [ 3.7742e-04, -3.8853e-03, -2.8248e-03, ..., -1.7223e-03, + -2.7065e-03, -2.1229e-03], + [-1.1559e-02, 1.1950e-03, 9.7961e-03, ..., 6.3133e-03, + 6.0806e-03, 1.3180e-03]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0136, -0.0171, -0.0059, 0.0068, -0.0213, -0.0136, 0.0155, -0.0007, + 0.0219, -0.0157], device='cuda:0'), grad: tensor([ 0.0153, -0.0837, 0.0013, -0.0266, 0.0411, -0.0168, 0.0269, 0.0291, + -0.0013, 0.0147], device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 216.27, cls_loss 0.5128 cls_loss_mapping 0.0069 cls_loss_causal 0.4888 re_mapping 0.0084 re_causal 0.0228 /// teacc 98.74 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.0554, -0.0686, -0.1248, ..., 0.1073, -0.0409, 0.0323], + [-0.0713, 0.0919, -0.0707, ..., 0.0367, -0.0555, -0.0850], + [-0.0401, 0.0791, -0.0976, ..., -0.0269, 0.0284, 0.1091], + ..., + [-0.0010, 0.0530, -0.0030, ..., -0.0076, 0.0257, -0.0537], + [ 0.0790, -0.0850, 0.0322, ..., 0.0071, 0.0056, -0.0411], + [-0.0205, -0.0682, 0.0681, ..., -0.0561, 0.0216, -0.0147]], + device='cuda:0'), grad: tensor([[ 3.9864e-04, 1.6987e-04, 1.3447e-03, ..., 1.2102e-03, + 2.6226e-03, 1.1311e-03], + [ 3.3975e-04, 1.2720e-04, 2.5582e-04, ..., 1.9588e-03, + 2.4414e-03, -5.8800e-05], + [ 3.2768e-03, 2.1210e-03, 1.9312e-03, ..., 2.8744e-03, + 1.8902e-03, 1.5764e-03], + ..., + [-6.0616e-03, -8.2855e-03, -2.8634e-04, ..., -9.6283e-03, + -1.4824e-02, -3.2825e-03], + [ 3.6836e-04, 9.9850e-04, -3.1555e-02, ..., 7.3767e-04, + -1.9470e-02, 2.4414e-03], + [ 3.7742e-04, 2.4490e-03, 2.3285e-02, ..., 4.9667e-03, + 2.0615e-02, 3.1328e-04]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0135, -0.0165, -0.0068, 0.0064, -0.0200, -0.0127, 0.0149, -0.0012, + 0.0223, -0.0164], device='cuda:0'), grad: tensor([ 0.0326, -0.0505, 0.0295, 0.0381, 0.0140, -0.0352, 0.0367, -0.0645, + 0.0241, -0.0248], device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 216.36, cls_loss 0.5525 cls_loss_mapping 0.0078 cls_loss_causal 0.5224 re_mapping 0.0078 re_causal 0.0210 /// teacc 98.71 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.0552, -0.0684, -0.1263, ..., 0.1062, -0.0412, 0.0327], + [-0.0727, 0.0924, -0.0687, ..., 0.0369, -0.0542, -0.0855], + [-0.0394, 0.0800, -0.0964, ..., -0.0264, 0.0294, 0.1085], + ..., + [-0.0004, 0.0525, -0.0026, ..., -0.0082, 0.0263, -0.0536], + [ 0.0792, -0.0855, 0.0325, ..., 0.0073, 0.0056, -0.0413], + [-0.0198, -0.0686, 0.0683, ..., -0.0569, 0.0212, -0.0146]], + device='cuda:0'), grad: tensor([[ 3.4952e-04, 1.5593e-04, 3.0470e-04, ..., 5.4741e-04, + 3.6788e-04, 9.5606e-04], + [ 2.2948e-05, 2.5123e-05, -1.2121e-03, ..., -6.5613e-04, + -9.9564e-04, -9.4366e-04], + [ 4.3535e-04, -1.8263e-04, -8.2350e-04, ..., 5.9128e-04, + 6.0940e-04, 7.9498e-06], + ..., + [-4.5166e-03, -1.0624e-03, 7.8821e-04, ..., -4.6134e-04, + 4.4608e-04, -1.5688e-03], + [ 3.6335e-04, 1.9896e-04, 9.3794e-04, ..., 1.5841e-03, + 4.6849e-04, 3.8314e-04], + [ 4.0207e-03, 1.8048e-04, -2.3003e-03, ..., -2.9831e-03, + -8.0633e-04, 7.4673e-04]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0136, -0.0172, -0.0069, 0.0065, -0.0203, -0.0122, 0.0151, -0.0017, + 0.0219, -0.0153], device='cuda:0'), grad: tensor([ 0.0132, -0.0058, 0.0135, -0.0149, 0.0181, 0.0252, -0.0236, -0.0164, + -0.0121, 0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 216.74, cls_loss 0.4908 cls_loss_mapping 0.0063 cls_loss_causal 0.4606 re_mapping 0.0088 re_causal 0.0232 /// teacc 98.82 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.0554, -0.0680, -0.1260, ..., 0.1074, -0.0410, 0.0332], + [-0.0720, 0.0922, -0.0679, ..., 0.0374, -0.0535, -0.0860], + [-0.0404, 0.0789, -0.0981, ..., -0.0274, 0.0287, 0.1090], + ..., + [-0.0007, 0.0547, -0.0039, ..., -0.0080, 0.0254, -0.0527], + [ 0.0804, -0.0864, 0.0330, ..., 0.0070, 0.0063, -0.0414], + [-0.0201, -0.0687, 0.0691, ..., -0.0567, 0.0221, -0.0140]], + device='cuda:0'), grad: tensor([[ 4.6700e-05, 1.5306e-04, 1.1486e-04, ..., 4.5371e-04, + 9.1851e-05, 3.3832e-04], + [ 9.8944e-05, 2.9182e-04, 3.7193e-04, ..., 7.6485e-04, + 2.9635e-04, 4.3106e-04], + [ 1.7083e-04, -1.2076e-04, 3.3593e-04, ..., 5.5075e-04, + 5.3310e-04, 2.5439e-04], + ..., + [ 2.4819e-04, -1.0519e-03, -2.0580e-03, ..., -1.1883e-03, + -1.4925e-03, -1.1325e-04], + [-6.0272e-03, 2.8658e-04, -1.1950e-03, ..., 8.9216e-04, + -4.2009e-04, -9.8133e-04], + [ 9.8896e-04, -4.1270e-04, -6.2180e-03, ..., 1.7214e-03, + -2.3289e-03, -1.8082e-03]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0143, -0.0174, -0.0069, 0.0060, -0.0197, -0.0135, 0.0138, -0.0017, + 0.0235, -0.0151], device='cuda:0'), grad: tensor([ 0.0113, 0.0179, 0.0145, 0.0114, 0.0297, 0.0155, -0.0472, -0.0278, + -0.0100, -0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 216.69, cls_loss 0.5235 cls_loss_mapping 0.0066 cls_loss_causal 0.4969 re_mapping 0.0079 re_causal 0.0206 /// teacc 98.84 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.0565, -0.0678, -0.1263, ..., 0.1067, -0.0424, 0.0333], + [-0.0715, 0.0926, -0.0691, ..., 0.0381, -0.0536, -0.0864], + [-0.0393, 0.0799, -0.0982, ..., -0.0270, 0.0301, 0.1090], + ..., + [ 0.0001, 0.0537, -0.0029, ..., -0.0077, 0.0255, -0.0547], + [ 0.0799, -0.0869, 0.0334, ..., 0.0073, 0.0067, -0.0403], + [-0.0207, -0.0683, 0.0691, ..., -0.0576, 0.0222, -0.0133]], + device='cuda:0'), grad: tensor([[ 8.2180e-06, 1.9848e-04, 3.9577e-04, ..., 2.9731e-04, + 4.0436e-04, 1.7214e-04], + [ 1.9073e-05, 2.1827e-04, 3.7074e-04, ..., 7.8964e-04, + 4.4155e-04, 3.8099e-04], + [-2.8076e-03, -3.4809e-03, -6.4774e-03, ..., -1.9684e-03, + -7.9727e-03, -3.9711e-03], + ..., + [ 1.6346e-03, 5.4693e-04, 2.0905e-03, ..., -2.5425e-03, + 2.9297e-03, -2.6727e-04], + [ 1.8263e-04, 4.7040e-04, 1.0071e-03, ..., 9.5987e-04, + 1.2589e-03, 6.0987e-04], + [ 2.9349e-04, 8.9455e-04, -9.4128e-04, ..., 1.6108e-03, + 3.2377e-04, 9.3794e-04]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0142, -0.0171, -0.0075, 0.0067, -0.0196, -0.0134, 0.0149, -0.0011, + 0.0225, -0.0160], device='cuda:0'), grad: tensor([ 7.6332e-03, 1.3550e-02, -2.8122e-02, -9.2149e-05, 1.6632e-02, + -1.3496e-02, -2.4414e-02, 2.3575e-03, 1.3329e-02, 1.2611e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 216.82, cls_loss 0.4967 cls_loss_mapping 0.0063 cls_loss_causal 0.4691 re_mapping 0.0081 re_causal 0.0216 /// teacc 98.50 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.0565, -0.0682, -0.1262, ..., 0.1072, -0.0424, 0.0340], + [-0.0718, 0.0920, -0.0694, ..., 0.0387, -0.0537, -0.0863], + [-0.0399, 0.0796, -0.0967, ..., -0.0264, 0.0304, 0.1094], + ..., + [ 0.0004, 0.0551, -0.0040, ..., -0.0088, 0.0246, -0.0545], + [ 0.0797, -0.0868, 0.0328, ..., 0.0069, 0.0065, -0.0410], + [-0.0201, -0.0685, 0.0682, ..., -0.0582, 0.0217, -0.0145]], + device='cuda:0'), grad: tensor([[ 1.0222e-04, 2.1100e-04, 2.9683e-04, ..., 4.5967e-03, + 1.2579e-03, 3.3760e-03], + [-8.4019e-04, -6.7234e-04, 4.6706e-04, ..., -3.1815e-03, + 1.7440e-04, -2.3174e-04], + [ 7.1168e-05, 2.4486e-04, 1.6487e-04, ..., 1.3561e-03, + 8.8310e-04, -1.0567e-03], + ..., + [ 8.5652e-05, 4.1723e-04, 2.5821e-04, ..., 2.5196e-03, + 9.3460e-04, 1.3504e-03], + [ 3.1376e-04, 2.8658e-04, 1.5974e-04, ..., -8.6746e-03, + -1.6205e-02, -1.4198e-02], + [-2.8324e-04, -1.5202e-03, -1.1988e-03, ..., -4.1733e-03, + -1.3351e-03, 3.9124e-04]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0156, -0.0171, -0.0080, 0.0069, -0.0185, -0.0138, 0.0152, -0.0014, + 0.0213, -0.0164], device='cuda:0'), grad: tensor([ 0.0316, -0.0314, -0.0124, 0.0637, 0.0213, 0.0282, 0.0045, 0.0291, + -0.1285, -0.0060], device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 216.69, cls_loss 0.5305 cls_loss_mapping 0.0128 cls_loss_causal 0.5026 re_mapping 0.0082 re_causal 0.0194 /// teacc 98.85 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.0567, -0.0688, -0.1267, ..., 0.1068, -0.0422, 0.0332], + [-0.0716, 0.0926, -0.0699, ..., 0.0386, -0.0537, -0.0852], + [-0.0409, 0.0791, -0.0973, ..., -0.0260, 0.0301, 0.1096], + ..., + [-0.0002, 0.0556, -0.0044, ..., -0.0089, 0.0242, -0.0552], + [ 0.0800, -0.0862, 0.0339, ..., 0.0075, 0.0075, -0.0401], + [-0.0195, -0.0697, 0.0679, ..., -0.0583, 0.0211, -0.0142]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0002, 0.0002, ..., 0.0134, 0.0002, 0.0016], + [ 0.0008, 0.0021, 0.0004, ..., 0.0053, 0.0004, 0.0010], + [ 0.0012, 0.0002, 0.0004, ..., 0.0010, -0.0017, -0.0014], + ..., + [ 0.0024, -0.0061, -0.0073, ..., -0.0098, -0.0042, -0.0027], + [ 0.0017, 0.0010, 0.0021, ..., 0.0085, 0.0018, 0.0025], + [ 0.0020, 0.0016, 0.0041, ..., 0.0044, 0.0012, 0.0017]], + device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0151, -0.0159, -0.0076, 0.0064, -0.0199, -0.0141, 0.0150, -0.0011, + 0.0216, -0.0161], device='cuda:0'), grad: tensor([ 0.0220, 0.0215, 0.0115, 0.0169, -0.0196, -0.0022, -0.0400, -0.0699, + 0.0333, 0.0266], device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 216.76, cls_loss 0.5386 cls_loss_mapping 0.0062 cls_loss_causal 0.5094 re_mapping 0.0083 re_causal 0.0218 /// teacc 98.66 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.0563, -0.0687, -0.1287, ..., 0.1065, -0.0441, 0.0325], + [-0.0718, 0.0936, -0.0695, ..., 0.0376, -0.0535, -0.0847], + [-0.0415, 0.0789, -0.0970, ..., -0.0235, 0.0300, 0.1108], + ..., + [-0.0006, 0.0567, -0.0044, ..., -0.0096, 0.0259, -0.0549], + [ 0.0790, -0.0875, 0.0344, ..., 0.0066, 0.0072, -0.0407], + [-0.0194, -0.0702, 0.0687, ..., -0.0589, 0.0212, -0.0150]], + device='cuda:0'), grad: tensor([[ 8.3804e-05, 1.0252e-03, 5.0449e-04, ..., 3.0117e-03, + 1.8110e-03, 1.5049e-03], + [ 5.2780e-05, -1.9932e-03, -2.9850e-03, ..., -1.8568e-03, + -5.3139e-03, 1.6088e-03], + [ 2.4334e-05, 2.9068e-03, 1.1053e-03, ..., 5.6496e-03, + 3.8490e-03, 3.4618e-03], + ..., + [ 7.1883e-05, 3.1242e-03, 1.6050e-03, ..., 4.1466e-03, + 1.2121e-03, 2.1420e-03], + [ 1.8072e-04, -6.1560e-04, 5.5933e-04, ..., -2.4071e-03, + 1.5087e-03, -1.8549e-03], + [ 2.2292e-04, 1.0328e-03, 1.3123e-03, ..., 2.6836e-03, + 2.4376e-03, 1.0376e-03]], device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0152, -0.0163, -0.0068, 0.0052, -0.0185, -0.0155, 0.0155, -0.0013, + 0.0219, -0.0160], device='cuda:0'), grad: tensor([ 0.0184, -0.0089, 0.0383, -0.0428, -0.0173, 0.0281, -0.0099, 0.0242, + -0.0500, 0.0200], device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 217.04, cls_loss 0.5219 cls_loss_mapping 0.0110 cls_loss_causal 0.4923 re_mapping 0.0080 re_causal 0.0204 /// teacc 98.79 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.0543, -0.0697, -0.1288, ..., 0.1079, -0.0444, 0.0333], + [-0.0718, 0.0946, -0.0701, ..., 0.0375, -0.0528, -0.0844], + [-0.0413, 0.0776, -0.0970, ..., -0.0245, 0.0299, 0.1103], + ..., + [-0.0006, 0.0576, -0.0044, ..., -0.0095, 0.0255, -0.0540], + [ 0.0785, -0.0874, 0.0336, ..., 0.0066, 0.0062, -0.0416], + [-0.0200, -0.0705, 0.0695, ..., -0.0589, 0.0218, -0.0150]], + device='cuda:0'), grad: tensor([[ 5.2786e-04, -6.7043e-04, 1.0538e-04, ..., -4.2152e-03, + 5.1880e-04, 2.6631e-04], + [ 3.1414e-03, 1.5678e-03, 1.6779e-05, ..., -4.3602e-03, + 8.3685e-05, -2.8954e-03], + [ 1.3466e-03, 1.9894e-03, 1.4138e-04, ..., 3.9978e-03, + 7.2289e-04, 4.3640e-03], + ..., + [ 5.5408e-04, -1.7099e-03, -7.4959e-04, ..., -1.1406e-03, + -2.3575e-03, -4.4899e-03], + [-6.4774e-03, -4.6387e-03, 2.0623e-05, ..., -3.4237e-03, + 8.6427e-05, -1.6689e-03], + [ 6.0749e-04, 4.3106e-04, 6.8963e-05, ..., 1.1091e-03, + 4.0829e-05, 8.2493e-04]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0148, -0.0166, -0.0071, 0.0055, -0.0191, -0.0153, 0.0158, -0.0011, + 0.0221, -0.0155], device='cuda:0'), grad: tensor([-0.0083, 0.0034, 0.0416, -0.0122, 0.0221, -0.0122, 0.0219, -0.0146, + -0.0549, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 216.67, cls_loss 0.5287 cls_loss_mapping 0.0069 cls_loss_causal 0.4944 re_mapping 0.0079 re_causal 0.0200 /// teacc 98.71 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.0549, -0.0703, -0.1299, ..., 0.1072, -0.0461, 0.0331], + [-0.0719, 0.0948, -0.0687, ..., 0.0390, -0.0533, -0.0856], + [-0.0415, 0.0782, -0.0965, ..., -0.0237, 0.0302, 0.1108], + ..., + [-0.0015, 0.0589, -0.0033, ..., -0.0090, 0.0274, -0.0537], + [ 0.0798, -0.0884, 0.0330, ..., 0.0071, 0.0063, -0.0417], + [-0.0191, -0.0723, 0.0686, ..., -0.0606, 0.0212, -0.0141]], + device='cuda:0'), grad: tensor([[ 6.3229e-04, 5.1975e-04, 2.5415e-04, ..., 2.9964e-03, + 5.0831e-04, 1.7538e-03], + [ 4.3297e-03, 3.8280e-03, 3.7730e-05, ..., -1.5812e-03, + 9.0122e-05, -6.6519e-04], + [ 2.3479e-03, 4.1389e-03, 1.7662e-03, ..., 1.6136e-03, + 3.9673e-03, -4.6420e-04], + ..., + [-7.0229e-03, -1.1215e-02, -2.2590e-05, ..., -6.9351e-03, + 9.8646e-05, 7.2193e-04], + [ 1.8656e-04, 7.4148e-04, 3.9756e-05, ..., 3.0880e-03, + 2.1374e-04, -1.1854e-03], + [ 1.4639e-04, -7.5865e-04, 1.0639e-04, ..., -1.0147e-03, + 1.7333e-04, -1.6613e-03]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0141, -0.0170, -0.0063, 0.0059, -0.0188, -0.0147, 0.0152, -0.0001, + 0.0215, -0.0163], device='cuda:0'), grad: tensor([ 0.0214, -0.0224, 0.0393, -0.0388, 0.0242, -0.0116, 0.0065, -0.0366, + 0.0365, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 216.91, cls_loss 0.5129 cls_loss_mapping 0.0062 cls_loss_causal 0.4736 re_mapping 0.0082 re_causal 0.0209 /// teacc 98.85 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.0553, -0.0699, -0.1297, ..., 0.1069, -0.0464, 0.0313], + [-0.0720, 0.0951, -0.0694, ..., 0.0376, -0.0540, -0.0854], + [-0.0399, 0.0781, -0.0971, ..., -0.0232, 0.0305, 0.1124], + ..., + [-0.0023, 0.0580, -0.0031, ..., -0.0087, 0.0263, -0.0550], + [ 0.0788, -0.0876, 0.0329, ..., 0.0086, 0.0060, -0.0416], + [-0.0193, -0.0720, 0.0688, ..., -0.0606, 0.0217, -0.0137]], + device='cuda:0'), grad: tensor([[-3.4666e-04, 1.5473e-04, -1.8692e-03, ..., -2.3460e-03, + -1.2636e-05, 5.6791e-04], + [ 9.5291e-03, 7.4005e-03, 1.2445e-04, ..., -2.2507e-03, + 2.8896e-04, 4.2844e-04], + [ 3.4738e-04, 1.4293e-04, 9.4831e-05, ..., 8.9550e-04, + 8.1968e-04, 3.9959e-04], + ..., + [ 8.4066e-04, 1.5020e-03, 1.4029e-03, ..., 1.0376e-03, + 9.3126e-04, 9.7275e-04], + [ 8.8692e-04, 4.6396e-04, 9.8610e-04, ..., 8.5354e-04, + 6.6471e-04, 1.8299e-04], + [-3.2306e-04, 4.6873e-04, 1.3123e-03, ..., -1.0271e-03, + -2.4853e-03, -3.8643e-03]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0141, -0.0164, -0.0070, 0.0062, -0.0187, -0.0143, 0.0141, -0.0011, + 0.0227, -0.0162], device='cuda:0'), grad: tensor([ 0.0024, 0.0267, 0.0133, 0.0044, -0.0073, -0.0117, -0.0168, 0.0185, + -0.0157, -0.0138], device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 216.72, cls_loss 0.5086 cls_loss_mapping 0.0066 cls_loss_causal 0.4740 re_mapping 0.0086 re_causal 0.0219 /// teacc 98.69 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.0554, -0.0692, -0.1295, ..., 0.1073, -0.0465, 0.0321], + [-0.0721, 0.0937, -0.0699, ..., 0.0390, -0.0540, -0.0853], + [-0.0397, 0.0783, -0.0970, ..., -0.0238, 0.0301, 0.1127], + ..., + [-0.0029, 0.0587, -0.0018, ..., -0.0087, 0.0274, -0.0554], + [ 0.0786, -0.0892, 0.0342, ..., 0.0072, 0.0067, -0.0412], + [-0.0183, -0.0716, 0.0673, ..., -0.0594, 0.0201, -0.0142]], + device='cuda:0'), grad: tensor([[-0.0005, 0.0003, 0.0195, ..., 0.0004, 0.0260, -0.0002], + [ 0.0003, -0.0007, 0.0002, ..., -0.0027, -0.0003, -0.0004], + [ 0.0002, 0.0003, 0.0006, ..., 0.0004, 0.0005, 0.0005], + ..., + [ 0.0025, 0.0094, 0.0165, ..., 0.0009, 0.0223, 0.0002], + [-0.0023, -0.0034, -0.0045, ..., -0.0017, -0.0005, 0.0030], + [-0.0024, -0.0080, -0.0346, ..., 0.0008, -0.0491, -0.0035]], + device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0150, -0.0162, -0.0073, 0.0056, -0.0194, -0.0140, 0.0139, -0.0007, + 0.0222, -0.0156], device='cuda:0'), grad: tensor([-3.6359e-05, -6.2439e-02, 1.2146e-02, 1.1154e-02, 9.7122e-03, + 1.3866e-03, 2.5620e-02, 5.1086e-02, -6.4621e-03, -4.2175e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 216.63, cls_loss 0.5181 cls_loss_mapping 0.0064 cls_loss_causal 0.4961 re_mapping 0.0082 re_causal 0.0222 /// teacc 98.75 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.0551, -0.0694, -0.1293, ..., 0.1069, -0.0460, 0.0324], + [-0.0721, 0.0936, -0.0702, ..., 0.0400, -0.0540, -0.0863], + [-0.0406, 0.0783, -0.0976, ..., -0.0244, 0.0300, 0.1124], + ..., + [-0.0030, 0.0596, -0.0019, ..., -0.0084, 0.0271, -0.0541], + [ 0.0785, -0.0903, 0.0348, ..., 0.0071, 0.0079, -0.0428], + [-0.0184, -0.0723, 0.0684, ..., -0.0599, 0.0207, -0.0141]], + device='cuda:0'), grad: tensor([[ 5.0187e-05, 3.6407e-04, 3.3498e-05, ..., 5.1546e-04, + 9.6679e-05, 2.5225e-04], + [ 9.5308e-05, 6.6566e-04, 1.5056e-04, ..., -4.3907e-03, + 2.5129e-04, 3.7313e-04], + [ 8.1539e-05, -8.3685e-04, 1.2612e-04, ..., -9.2030e-04, + -1.1616e-03, -1.4944e-03], + ..., + [-1.5383e-03, -4.2763e-03, -6.9237e-04, ..., -1.5478e-03, + -5.2309e-04, -1.2064e-03], + [ 1.9276e-04, 4.8327e-04, 1.5748e-04, ..., 1.1997e-03, + 2.1529e-04, 2.8563e-04], + [ 7.4577e-04, 1.3971e-03, 6.8378e-04, ..., 8.0824e-04, + 3.9768e-04, 2.5010e-04]], device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0149, -0.0156, -0.0076, 0.0057, -0.0194, -0.0147, 0.0141, -0.0006, + 0.0220, -0.0155], device='cuda:0'), grad: tensor([ 0.0146, -0.0016, -0.0099, -0.0049, 0.0136, 0.0171, -0.0118, -0.0589, + 0.0181, 0.0238], device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 216.72, cls_loss 0.5084 cls_loss_mapping 0.0065 cls_loss_causal 0.4754 re_mapping 0.0079 re_causal 0.0199 /// teacc 98.80 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.0566, -0.0686, -0.1302, ..., 0.1065, -0.0471, 0.0315], + [-0.0723, 0.0937, -0.0715, ..., 0.0396, -0.0541, -0.0879], + [-0.0403, 0.0785, -0.0973, ..., -0.0243, 0.0303, 0.1129], + ..., + [-0.0031, 0.0587, -0.0021, ..., -0.0089, 0.0270, -0.0550], + [ 0.0769, -0.0899, 0.0353, ..., 0.0067, 0.0080, -0.0424], + [-0.0175, -0.0719, 0.0691, ..., -0.0596, 0.0210, -0.0144]], + device='cuda:0'), grad: tensor([[ 2.2352e-08, 2.5535e-04, 7.6342e-04, ..., 6.5804e-04, + 6.7139e-04, 1.7281e-03], + [ 3.0965e-05, 1.4086e-03, 1.1170e-04, ..., 3.3587e-05, + 9.1314e-05, 5.4419e-05], + [ 3.3192e-06, 2.8324e-04, 4.1866e-04, ..., -1.4412e-02, + -8.0109e-03, -1.3115e-02], + ..., + [-3.6359e-05, -1.6079e-03, 1.7548e-03, ..., 3.5465e-06, + 1.0929e-03, 2.0874e-04], + [ 5.8673e-07, -5.4979e-04, -6.8903e-04, ..., 1.2970e-02, + 7.7171e-03, 8.8501e-03], + [ 3.5334e-06, -7.2598e-05, -3.3398e-03, ..., 1.9252e-04, + -3.5114e-03, 2.6107e-04]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0143, -0.0162, -0.0078, 0.0058, -0.0184, -0.0139, 0.0145, -0.0010, + 0.0211, -0.0150], device='cuda:0'), grad: tensor([ 0.0104, 0.0023, -0.0247, 0.0062, 0.0034, 0.0068, 0.0014, 0.0039, + 0.0085, -0.0184], device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 216.42, cls_loss 0.5232 cls_loss_mapping 0.0069 cls_loss_causal 0.4965 re_mapping 0.0081 re_causal 0.0203 /// teacc 98.83 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.0567, -0.0683, -0.1302, ..., 0.1072, -0.0469, 0.0315], + [-0.0731, 0.0936, -0.0718, ..., 0.0394, -0.0540, -0.0877], + [-0.0382, 0.0788, -0.0967, ..., -0.0229, 0.0308, 0.1130], + ..., + [-0.0025, 0.0583, -0.0029, ..., -0.0110, 0.0256, -0.0557], + [ 0.0773, -0.0899, 0.0347, ..., 0.0070, 0.0074, -0.0422], + [-0.0183, -0.0728, 0.0689, ..., -0.0609, 0.0215, -0.0152]], + device='cuda:0'), grad: tensor([[-1.0030e-06, 1.5950e-04, -6.7902e-04, ..., 7.4196e-04, + -7.6914e-04, -2.4104e-04], + [ 2.0992e-06, 9.6381e-05, 2.4414e-04, ..., 5.3596e-04, + 2.9016e-04, 4.1771e-04], + [ 2.5146e-07, 1.2082e-04, 1.9324e-04, ..., 1.2999e-03, + 8.1491e-04, 6.0844e-04], + ..., + [ 9.1434e-05, 3.4485e-03, 8.6517e-03, ..., 1.8902e-03, + 2.2221e-03, 7.4291e-04], + [ 4.8950e-06, 1.3018e-04, 3.1209e-04, ..., 3.1233e-04, + 4.4680e-04, 4.0436e-04], + [ 2.0635e-04, -6.1989e-04, -7.6246e-04, ..., 4.4012e-04, + -1.2398e-03, 2.9135e-04]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0151, -0.0161, -0.0087, 0.0059, -0.0184, -0.0150, 0.0147, -0.0012, + 0.0227, -0.0157], device='cuda:0'), grad: tensor([-0.0184, 0.0141, 0.0148, -0.0172, -0.0522, 0.0118, 0.0158, 0.0360, + 0.0121, -0.0170], device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 216.19, cls_loss 0.5225 cls_loss_mapping 0.0073 cls_loss_causal 0.4965 re_mapping 0.0080 re_causal 0.0195 /// teacc 98.54 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.0574, -0.0687, -0.1300, ..., 0.1069, -0.0460, 0.0324], + [-0.0733, 0.0926, -0.0708, ..., 0.0393, -0.0537, -0.0870], + [-0.0376, 0.0795, -0.0972, ..., -0.0226, 0.0304, 0.1121], + ..., + [-0.0024, 0.0565, -0.0036, ..., -0.0125, 0.0250, -0.0563], + [ 0.0769, -0.0878, 0.0342, ..., 0.0065, 0.0067, -0.0431], + [-0.0168, -0.0711, 0.0699, ..., -0.0603, 0.0224, -0.0142]], + device='cuda:0'), grad: tensor([[-1.0960e-05, 2.3222e-04, 2.3961e-04, ..., -2.1040e-04, + -3.0766e-03, -4.0321e-03], + [ 3.1944e-07, -1.1015e-03, 1.1873e-04, ..., -6.6338e-03, + 1.1522e-04, -2.2087e-03], + [ 1.4648e-05, 3.8934e-04, 6.1631e-05, ..., 3.5596e-04, + 6.7329e-04, 2.6627e-03], + ..., + [ 2.7008e-07, -2.1114e-03, 5.2303e-05, ..., 1.9670e-04, + 3.7193e-04, 9.5034e-04], + [-4.2877e-03, 1.4000e-03, 1.9932e-04, ..., 3.3360e-03, + 6.5517e-04, 1.0700e-03], + [ 3.0380e-06, 7.6866e-04, -1.7975e-02, ..., 3.3545e-04, + -3.1036e-02, 1.7500e-03]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0144, -0.0155, -0.0081, 0.0055, -0.0178, -0.0143, 0.0146, -0.0020, + 0.0222, -0.0155], device='cuda:0'), grad: tensor([-0.0369, -0.0236, -0.0049, 0.0156, 0.0212, 0.0086, 0.0255, -0.0214, + 0.0165, -0.0006], device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 216.32, cls_loss 0.5050 cls_loss_mapping 0.0070 cls_loss_causal 0.4875 re_mapping 0.0086 re_causal 0.0208 /// teacc 98.78 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.0567, -0.0698, -0.1298, ..., 0.1065, -0.0455, 0.0332], + [-0.0722, 0.0937, -0.0706, ..., 0.0401, -0.0534, -0.0862], + [-0.0371, 0.0804, -0.0979, ..., -0.0231, 0.0299, 0.1125], + ..., + [-0.0032, 0.0556, -0.0043, ..., -0.0112, 0.0249, -0.0563], + [ 0.0767, -0.0887, 0.0341, ..., 0.0058, 0.0060, -0.0435], + [-0.0175, -0.0708, 0.0702, ..., -0.0610, 0.0228, -0.0152]], + device='cuda:0'), grad: tensor([[ 4.4078e-05, 9.0790e-04, 1.5485e-04, ..., 4.6229e-04, + 1.1177e-03, 1.7319e-03], + [-1.2022e-04, 6.9923e-03, 6.0225e-04, ..., 1.3115e-02, + 1.2083e-03, 2.2583e-03], + [-1.5032e-04, -2.2202e-03, 3.9601e-04, ..., -2.2984e-03, + 3.0212e-03, 3.5191e-03], + ..., + [ 4.6420e-04, -3.8891e-03, 6.3848e-04, ..., -2.0008e-03, + 2.0866e-03, -1.6594e-03], + [-1.1196e-03, 7.9393e-04, -4.4899e-03, ..., -8.1778e-04, + -2.4304e-05, 1.6708e-03], + [ 2.2531e-04, -4.0932e-03, 7.8583e-03, ..., -1.0597e-02, + 3.8013e-03, 2.1076e-03]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0145, -0.0147, -0.0077, 0.0051, -0.0177, -0.0144, 0.0153, -0.0019, + 0.0207, -0.0157], device='cuda:0'), grad: tensor([ 0.0169, 0.0583, 0.0162, 0.0084, -0.0091, 0.0286, -0.0335, -0.0019, + -0.0598, -0.0240], device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 216.75, cls_loss 0.5316 cls_loss_mapping 0.0062 cls_loss_causal 0.4981 re_mapping 0.0081 re_causal 0.0206 /// teacc 98.82 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.0569, -0.0705, -0.1300, ..., 0.1080, -0.0454, 0.0329], + [-0.0720, 0.0939, -0.0712, ..., 0.0391, -0.0543, -0.0864], + [-0.0364, 0.0806, -0.0974, ..., -0.0216, 0.0304, 0.1135], + ..., + [-0.0041, 0.0554, -0.0041, ..., -0.0116, 0.0248, -0.0556], + [ 0.0772, -0.0886, 0.0338, ..., 0.0049, 0.0057, -0.0436], + [-0.0177, -0.0705, 0.0697, ..., -0.0605, 0.0231, -0.0159]], + device='cuda:0'), grad: tensor([[-7.7486e-07, 5.0217e-05, 2.0361e-04, ..., -3.5686e-03, + -3.2578e-03, 6.4945e-04], + [ 1.8626e-09, -3.3617e-04, 4.8429e-05, ..., -1.0103e-04, + 1.0437e-04, 2.1064e-04], + [ 1.0151e-07, 7.0305e-03, 1.6522e-04, ..., 9.5844e-04, + 3.2663e-04, 8.8272e-03], + ..., + [-2.0489e-08, 9.4622e-06, -4.8828e-04, ..., 4.5800e-04, + 1.1414e-05, 3.2306e-04], + [ 3.6322e-08, 1.9348e-04, 1.1986e-04, ..., 5.4169e-04, + 2.1839e-04, 2.9826e-04], + [ 7.9907e-07, 6.3360e-05, 1.0223e-03, ..., 4.0841e-04, + 5.8079e-04, 2.9469e-04]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0147, -0.0155, -0.0066, 0.0050, -0.0181, -0.0140, 0.0155, -0.0014, + 0.0200, -0.0161], device='cuda:0'), grad: tensor([-0.0316, 0.0025, 0.0369, -0.0544, 0.0055, 0.0153, 0.0086, 0.0054, + 0.0055, 0.0065], device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 216.89, cls_loss 0.5123 cls_loss_mapping 0.0092 cls_loss_causal 0.4891 re_mapping 0.0076 re_causal 0.0194 /// teacc 98.76 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.0571, -0.0721, -0.1308, ..., 0.1063, -0.0461, 0.0316], + [-0.0714, 0.0948, -0.0702, ..., 0.0399, -0.0529, -0.0869], + [-0.0368, 0.0812, -0.0959, ..., -0.0230, 0.0302, 0.1123], + ..., + [-0.0041, 0.0543, -0.0046, ..., -0.0107, 0.0249, -0.0560], + [ 0.0764, -0.0893, 0.0345, ..., 0.0053, 0.0062, -0.0435], + [-0.0160, -0.0694, 0.0698, ..., -0.0597, 0.0226, -0.0148]], + device='cuda:0'), grad: tensor([[ 1.2573e-07, 7.2813e-04, 2.0325e-04, ..., 2.9755e-04, + 5.4300e-05, 2.5606e-04], + [ 1.7099e-06, -1.2146e-02, 5.9938e-04, ..., -3.6125e-03, + 4.7493e-04, -3.0136e-03], + [ 3.3155e-07, 8.5297e-03, 6.3515e-04, ..., 2.6379e-03, + -2.3155e-03, -1.8091e-03], + ..., + [ 2.8580e-05, -1.8206e-03, -4.4136e-03, ..., -4.3106e-04, + -6.0120e-03, 1.1253e-03], + [ 3.6228e-07, 1.6985e-03, 1.8177e-03, ..., 6.2943e-04, + 1.5087e-03, 1.4181e-03], + [ 1.2755e-04, 1.1959e-03, -8.3466e-03, ..., -1.1616e-03, + -1.3342e-03, -5.9557e-04]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0143, -0.0143, -0.0073, 0.0049, -0.0187, -0.0138, 0.0156, -0.0015, + 0.0203, -0.0160], device='cuda:0'), grad: tensor([ 0.0040, -0.0563, 0.0292, -0.0069, 0.0052, 0.0272, -0.0113, -0.0033, + 0.0220, -0.0098], device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 216.83, cls_loss 0.5193 cls_loss_mapping 0.0060 cls_loss_causal 0.4858 re_mapping 0.0080 re_causal 0.0193 /// teacc 98.69 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.0570, -0.0704, -0.1297, ..., 0.1076, -0.0457, 0.0316], + [-0.0714, 0.0947, -0.0721, ..., 0.0402, -0.0543, -0.0870], + [-0.0373, 0.0810, -0.0945, ..., -0.0236, 0.0313, 0.1129], + ..., + [-0.0048, 0.0546, -0.0046, ..., -0.0108, 0.0251, -0.0565], + [ 0.0774, -0.0902, 0.0348, ..., 0.0059, 0.0059, -0.0419], + [-0.0161, -0.0698, 0.0696, ..., -0.0600, 0.0224, -0.0144]], + device='cuda:0'), grad: tensor([[ 1.8282e-06, 6.5327e-04, 2.1343e-03, ..., 2.9602e-03, + 2.3785e-03, 5.2643e-04], + [ 2.2724e-07, 7.8440e-04, -2.3842e-03, ..., -9.5367e-04, + -2.6493e-03, 2.0301e-04], + [ 1.5162e-05, -5.2643e-04, 1.5640e-03, ..., -4.3559e-04, + 2.6493e-03, 5.5933e-04], + ..., + [ 6.9290e-07, -1.4896e-03, -7.7152e-04, ..., 5.6267e-04, + -2.3499e-03, -1.1158e-03], + [ 4.1395e-05, -6.9284e-04, -4.6997e-03, ..., -4.5815e-03, + 4.7379e-03, 3.0234e-05], + [ 7.7579e-07, 2.7966e-04, 7.3853e-03, ..., 5.0583e-03, + 3.3417e-03, 2.4700e-04]], device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0165, -0.0152, -0.0078, 0.0059, -0.0182, -0.0141, 0.0145, -0.0024, + 0.0201, -0.0157], device='cuda:0'), grad: tensor([ 0.0238, -0.0191, 0.0077, -0.0122, 0.0218, -0.0419, 0.0141, -0.0228, + -0.0062, 0.0348], device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 216.83, cls_loss 0.5197 cls_loss_mapping 0.0056 cls_loss_causal 0.4911 re_mapping 0.0082 re_causal 0.0203 /// teacc 98.75 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.0573, -0.0711, -0.1288, ..., 0.1074, -0.0460, 0.0318], + [-0.0723, 0.0947, -0.0721, ..., 0.0406, -0.0528, -0.0868], + [-0.0379, 0.0812, -0.0957, ..., -0.0252, 0.0310, 0.1126], + ..., + [-0.0056, 0.0548, -0.0047, ..., -0.0106, 0.0257, -0.0565], + [ 0.0777, -0.0903, 0.0347, ..., 0.0048, 0.0053, -0.0424], + [-0.0147, -0.0705, 0.0697, ..., -0.0612, 0.0224, -0.0150]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.4724e-04, 7.4911e-04, ..., -1.3218e-03, + -4.8876e-04, -4.3774e-04], + [ 4.6566e-09, 4.0665e-03, 3.4580e-03, ..., 5.4436e-03, + 1.0853e-03, 1.2696e-04], + [ 6.5193e-09, 9.9277e-04, -6.1455e-03, ..., -2.9964e-03, + -5.4970e-03, 7.5102e-04], + ..., + [ 9.6709e-06, 9.4223e-04, 5.1460e-03, ..., 3.7518e-03, + 4.2191e-03, 8.3208e-04], + [ 5.6904e-07, -4.3907e-03, -9.2840e-04, ..., -5.4436e-03, + 9.5844e-04, 6.3276e-04], + [-2.4557e-05, -3.3913e-03, -5.7907e-03, ..., -2.0866e-03, + -3.4103e-03, -2.5082e-03]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0165, -0.0149, -0.0087, 0.0056, -0.0181, -0.0134, 0.0143, -0.0014, + 0.0198, -0.0161], device='cuda:0'), grad: tensor([ 0.0022, -0.0071, -0.0156, 0.0210, 0.0048, 0.0200, -0.0157, 0.0278, + -0.0158, -0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 217.12, cls_loss 0.5083 cls_loss_mapping 0.0055 cls_loss_causal 0.4831 re_mapping 0.0079 re_causal 0.0197 /// teacc 98.71 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.0563, -0.0720, -0.1303, ..., 0.1076, -0.0460, 0.0313], + [-0.0722, 0.0928, -0.0731, ..., 0.0413, -0.0536, -0.0873], + [-0.0384, 0.0822, -0.0952, ..., -0.0238, 0.0315, 0.1137], + ..., + [-0.0059, 0.0549, -0.0049, ..., -0.0116, 0.0258, -0.0569], + [ 0.0774, -0.0906, 0.0337, ..., 0.0045, 0.0035, -0.0432], + [-0.0142, -0.0702, 0.0707, ..., -0.0603, 0.0231, -0.0148]], + device='cuda:0'), grad: tensor([[-3.9153e-06, 3.7289e-04, 1.8530e-03, ..., 4.2772e-04, + 1.1377e-03, 4.4556e-03], + [ 7.2736e-07, -1.3971e-03, 2.9564e-04, ..., -1.4992e-03, + -3.2663e-04, -7.6866e-04], + [ 6.4969e-05, -2.8419e-03, -6.2027e-03, ..., -7.7400e-03, + -6.2027e-03, 1.2326e-04], + ..., + [ 2.9914e-06, 1.5936e-03, 2.7046e-03, ..., 5.3673e-03, + 3.7708e-03, 2.7294e-03], + [-9.4175e-05, 9.9003e-05, -1.5373e-03, ..., -2.9449e-03, + -4.4479e-03, -6.4240e-03], + [ 4.2394e-06, 1.0628e-04, 4.3869e-04, ..., 9.5844e-05, + 5.5771e-03, 7.6008e-04]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0160, -0.0152, -0.0083, 0.0056, -0.0184, -0.0132, 0.0142, -0.0013, + 0.0190, -0.0150], device='cuda:0'), grad: tensor([ 0.0323, -0.0105, -0.0276, 0.0217, 0.0123, 0.0191, -0.0064, -0.0231, + -0.0058, -0.0121], device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 216.93, cls_loss 0.4971 cls_loss_mapping 0.0056 cls_loss_causal 0.4696 re_mapping 0.0082 re_causal 0.0210 /// teacc 98.73 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.0563, -0.0719, -0.1299, ..., 0.1078, -0.0464, 0.0321], + [-0.0729, 0.0935, -0.0733, ..., 0.0421, -0.0537, -0.0873], + [-0.0384, 0.0820, -0.0952, ..., -0.0226, 0.0320, 0.1141], + ..., + [-0.0052, 0.0538, -0.0040, ..., -0.0125, 0.0264, -0.0577], + [ 0.0776, -0.0904, 0.0339, ..., 0.0038, 0.0033, -0.0431], + [-0.0140, -0.0691, 0.0708, ..., -0.0605, 0.0229, -0.0153]], + device='cuda:0'), grad: tensor([[ 2.8670e-05, 8.6546e-04, 3.6550e-04, ..., 1.2331e-03, + 5.0575e-05, 4.3058e-04], + [ 6.5845e-07, -4.5013e-03, -2.5043e-03, ..., -6.9656e-03, + 2.5257e-05, -1.0519e-03], + [ 9.7394e-05, 1.7776e-03, 2.7466e-04, ..., 1.8654e-03, + 2.0146e-04, 1.2341e-03], + ..., + [-1.0796e-05, 1.0099e-03, 3.4451e-04, ..., 9.8419e-04, + 1.0943e-04, 5.7888e-04], + [ 3.2353e-04, 1.1501e-03, 6.1321e-04, ..., 1.2321e-03, + 3.0017e-04, 1.5049e-03], + [ 6.3121e-05, 6.3467e-04, -1.1501e-03, ..., 8.3876e-04, + -2.7966e-04, 3.3188e-04]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0155, -0.0150, -0.0082, 0.0055, -0.0200, -0.0124, 0.0145, -0.0017, + 0.0197, -0.0144], device='cuda:0'), grad: tensor([ 0.0136, -0.0619, 0.0210, -0.0524, 0.0097, 0.0232, 0.0095, 0.0192, + -0.0038, 0.0219], device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 217.40, cls_loss 0.5557 cls_loss_mapping 0.0059 cls_loss_causal 0.5235 re_mapping 0.0079 re_causal 0.0206 /// teacc 98.72 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.0578, -0.0711, -0.1309, ..., 0.1095, -0.0473, 0.0319], + [-0.0723, 0.0935, -0.0721, ..., 0.0420, -0.0528, -0.0880], + [-0.0388, 0.0819, -0.0959, ..., -0.0238, 0.0315, 0.1141], + ..., + [-0.0051, 0.0541, -0.0025, ..., -0.0120, 0.0280, -0.0575], + [ 0.0783, -0.0912, 0.0345, ..., 0.0046, 0.0046, -0.0428], + [-0.0144, -0.0694, 0.0704, ..., -0.0623, 0.0217, -0.0156]], + device='cuda:0'), grad: tensor([[ 2.2662e-04, 3.4881e-04, 4.8828e-04, ..., 4.2367e-04, + 5.7554e-04, 1.9372e-05], + [ 1.0147e-03, -2.0142e-03, -9.9182e-04, ..., -5.9128e-03, + 1.0338e-03, -1.4753e-03], + [-3.5381e-03, -7.4081e-03, 4.9973e-04, ..., 4.6802e-04, + -6.5651e-03, -3.4695e-03], + ..., + [-7.9536e-04, -1.3983e-04, 2.0905e-03, ..., 3.3340e-03, + 6.4945e-04, 6.8951e-04], + [ 7.2098e-04, 9.9468e-04, 2.8667e-03, ..., 1.7929e-03, + 3.3474e-03, 6.2990e-04], + [ 2.2469e-03, 2.4300e-03, -8.4839e-03, ..., -2.4338e-03, + -6.8054e-03, -5.5838e-04]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0161, -0.0154, -0.0096, 0.0066, -0.0202, -0.0132, 0.0156, -0.0014, + 0.0196, -0.0145], device='cuda:0'), grad: tensor([ 0.0103, -0.0318, -0.0716, -0.0030, 0.0107, 0.0401, 0.0114, 0.0261, + 0.0224, -0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 216.98, cls_loss 0.5114 cls_loss_mapping 0.0053 cls_loss_causal 0.4887 re_mapping 0.0079 re_causal 0.0199 /// teacc 98.70 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.0594, -0.0713, -0.1309, ..., 0.1097, -0.0472, 0.0325], + [-0.0712, 0.0933, -0.0722, ..., 0.0422, -0.0521, -0.0879], + [-0.0388, 0.0815, -0.0977, ..., -0.0248, 0.0310, 0.1146], + ..., + [-0.0055, 0.0544, -0.0036, ..., -0.0111, 0.0265, -0.0569], + [ 0.0796, -0.0920, 0.0345, ..., 0.0047, 0.0042, -0.0423], + [-0.0151, -0.0691, 0.0713, ..., -0.0621, 0.0234, -0.0161]], + device='cuda:0'), grad: tensor([[ 1.5664e-04, 6.4492e-05, 3.5310e-04, ..., 4.5109e-04, + 2.7800e-04, 3.9387e-04], + [ 7.7903e-05, 1.4365e-04, 1.1247e-04, ..., 6.1798e-04, + 1.5402e-04, 8.3864e-05], + [ 1.0544e-04, 3.0935e-05, 2.1343e-03, ..., 1.1816e-03, + 8.5449e-03, 1.4150e-04], + ..., + [ 2.0117e-05, 1.3089e-04, -3.0365e-03, ..., -7.9775e-04, + -1.7624e-02, 1.7929e-04], + [-2.3041e-03, 1.2165e-04, 2.1877e-03, ..., 1.1921e-03, + 5.1193e-03, -1.6766e-03], + [-2.4331e-04, 1.3983e-04, 5.0402e-04, ..., 1.1024e-03, + 8.5115e-05, 1.7643e-04]], device='cuda:0') +Epoch 215, bias, value: tensor([ 0.0160, -0.0146, -0.0092, 0.0066, -0.0195, -0.0137, 0.0147, -0.0020, + 0.0190, -0.0139], device='cuda:0'), grad: tensor([ 0.0102, -0.0187, 0.0220, 0.0144, 0.0085, -0.0268, 0.0082, -0.0135, + -0.0174, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 217.02, cls_loss 0.4933 cls_loss_mapping 0.0059 cls_loss_causal 0.4642 re_mapping 0.0075 re_causal 0.0187 /// teacc 98.61 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.0595, -0.0719, -0.1305, ..., 0.1107, -0.0473, 0.0327], + [-0.0711, 0.0942, -0.0726, ..., 0.0424, -0.0532, -0.0883], + [-0.0383, 0.0812, -0.0976, ..., -0.0242, 0.0307, 0.1136], + ..., + [-0.0064, 0.0551, -0.0026, ..., -0.0118, 0.0270, -0.0555], + [ 0.0792, -0.0925, 0.0342, ..., 0.0046, 0.0039, -0.0422], + [-0.0152, -0.0706, 0.0710, ..., -0.0631, 0.0237, -0.0175]], + device='cuda:0'), grad: tensor([[ 1.2189e-04, 4.5609e-04, 4.7565e-04, ..., 7.7677e-04, + 3.4857e-04, 1.9395e-04], + [ 6.9961e-06, 4.2200e-04, 1.1665e-04, ..., 9.8228e-04, + 1.1301e-04, 2.1720e-04], + [ 1.6201e-04, 1.6451e-03, 1.6899e-03, ..., 2.8133e-03, + 6.8235e-04, 1.0691e-03], + ..., + [ 2.9892e-05, -1.6689e-03, 1.0948e-03, ..., -4.2610e-03, + 8.2207e-04, -2.4009e-06], + [ 5.5283e-05, 4.7565e-04, 6.2132e-04, ..., 1.7948e-03, + 6.4278e-04, -2.4962e-04], + [ 4.1306e-05, 5.8746e-04, -9.1362e-04, ..., -6.3038e-04, + 4.9496e-04, 3.0327e-04]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0170, -0.0156, -0.0089, 0.0069, -0.0196, -0.0140, 0.0154, -0.0018, + 0.0186, -0.0146], device='cuda:0'), grad: tensor([-0.0065, 0.0208, 0.0336, -0.0068, -0.0050, 0.0278, -0.0432, -0.0314, + 0.0198, -0.0090], device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 216.82, cls_loss 0.4883 cls_loss_mapping 0.0062 cls_loss_causal 0.4641 re_mapping 0.0075 re_causal 0.0196 /// teacc 98.50 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.0584, -0.0726, -0.1309, ..., 0.1105, -0.0476, 0.0315], + [-0.0710, 0.0944, -0.0729, ..., 0.0425, -0.0545, -0.0894], + [-0.0383, 0.0811, -0.0979, ..., -0.0240, 0.0309, 0.1136], + ..., + [-0.0057, 0.0554, -0.0031, ..., -0.0110, 0.0268, -0.0548], + [ 0.0804, -0.0921, 0.0339, ..., 0.0053, 0.0032, -0.0417], + [-0.0160, -0.0708, 0.0710, ..., -0.0637, 0.0236, -0.0179]], + device='cuda:0'), grad: tensor([[ 1.8525e-04, 5.1260e-05, 4.1872e-05, ..., 1.1969e-03, + 1.5459e-03, 6.0959e-03], + [-5.0621e-03, -4.4479e-03, 7.2896e-05, ..., -1.1740e-03, + 3.9250e-05, 1.4591e-04], + [-4.0722e-04, 4.3213e-05, 7.3612e-05, ..., -4.6158e-03, + -2.2011e-03, -8.9722e-03], + ..., + [ 1.1663e-03, 7.7438e-04, 4.1771e-03, ..., 5.8270e-04, + 4.3917e-04, 2.3222e-04], + [ 2.7924e-03, 3.0899e-04, 9.0885e-04, ..., 5.3072e-04, + 2.0850e-04, 3.9005e-04], + [-8.2397e-04, -3.7700e-05, -1.1734e-02, ..., 6.2275e-04, + -1.3771e-03, 6.1810e-05]], device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0167, -0.0152, -0.0089, 0.0061, -0.0193, -0.0144, 0.0166, -0.0010, + 0.0186, -0.0158], device='cuda:0'), grad: tensor([ 0.0226, -0.0099, -0.0416, 0.0151, 0.0333, -0.0539, 0.0230, 0.0229, + 0.0131, -0.0245], device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 217.21, cls_loss 0.5205 cls_loss_mapping 0.0056 cls_loss_causal 0.4929 re_mapping 0.0074 re_causal 0.0189 /// teacc 98.74 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.0602, -0.0732, -0.1312, ..., 0.1103, -0.0476, 0.0307], + [-0.0696, 0.0945, -0.0712, ..., 0.0429, -0.0542, -0.0892], + [-0.0384, 0.0811, -0.0980, ..., -0.0232, 0.0314, 0.1148], + ..., + [-0.0061, 0.0552, -0.0041, ..., -0.0109, 0.0262, -0.0551], + [ 0.0810, -0.0933, 0.0342, ..., 0.0057, 0.0037, -0.0434], + [-0.0163, -0.0708, 0.0712, ..., -0.0639, 0.0235, -0.0176]], + device='cuda:0'), grad: tensor([[ 5.1767e-05, 4.6086e-04, 4.5252e-04, ..., -2.7866e-03, + 4.8590e-04, -4.3602e-03], + [ 1.5676e-05, 2.2364e-04, 2.5725e-04, ..., 6.4039e-04, + 1.2243e-04, 7.6580e-04], + [ 6.0171e-05, -7.4387e-04, 5.4550e-04, ..., 1.2379e-03, + 6.2609e-04, 6.5947e-04], + ..., + [ 5.8144e-05, 1.3329e-02, 3.0899e-03, ..., 8.7051e-03, + 4.9057e-03, 2.6875e-03], + [ 6.2406e-05, -3.2597e-03, -2.9793e-03, ..., -9.9106e-03, + -5.6915e-03, -1.7614e-03], + [-5.2929e-04, -1.6708e-03, -4.2076e-03, ..., -2.5787e-03, + -1.6441e-03, -2.0237e-03]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0156, -0.0146, -0.0087, 0.0060, -0.0198, -0.0145, 0.0168, -0.0012, + 0.0181, -0.0143], device='cuda:0'), grad: tensor([-0.0160, 0.0125, -0.0127, -0.0107, 0.0186, 0.0117, 0.0287, 0.0719, + -0.0443, -0.0597], device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 216.67, cls_loss 0.5022 cls_loss_mapping 0.0060 cls_loss_causal 0.4694 re_mapping 0.0080 re_causal 0.0200 /// teacc 98.77 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.0598, -0.0731, -0.1327, ..., 0.1107, -0.0480, 0.0306], + [-0.0689, 0.0952, -0.0729, ..., 0.0429, -0.0546, -0.0884], + [-0.0390, 0.0812, -0.0972, ..., -0.0226, 0.0322, 0.1155], + ..., + [-0.0060, 0.0541, -0.0033, ..., -0.0135, 0.0262, -0.0561], + [ 0.0800, -0.0930, 0.0337, ..., 0.0053, 0.0032, -0.0438], + [-0.0163, -0.0715, 0.0714, ..., -0.0637, 0.0239, -0.0178]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0070, 0.0002, ..., 0.0013, 0.0053, 0.0048], + [ 0.0001, -0.0057, 0.0001, ..., 0.0009, -0.0052, -0.0055], + [-0.0016, 0.0005, 0.0001, ..., -0.0019, 0.0003, -0.0044], + ..., + [ 0.0007, 0.0035, 0.0018, ..., 0.0068, -0.0025, 0.0007], + [ 0.0002, 0.0012, 0.0005, ..., 0.0038, 0.0006, 0.0006], + [ 0.0002, 0.0012, 0.0008, ..., 0.0021, 0.0006, -0.0018]], + device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0154, -0.0149, -0.0092, 0.0071, -0.0191, -0.0136, 0.0163, -0.0016, + 0.0183, -0.0152], device='cuda:0'), grad: tensor([ 0.0505, -0.0252, -0.0115, 0.0202, -0.0131, -0.0726, 0.0320, 0.0045, + 0.0240, -0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 217.01, cls_loss 0.5014 cls_loss_mapping 0.0062 cls_loss_causal 0.4798 re_mapping 0.0077 re_causal 0.0193 /// teacc 98.79 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.0600, -0.0739, -0.1338, ..., 0.1113, -0.0489, 0.0303], + [-0.0700, 0.0965, -0.0712, ..., 0.0441, -0.0533, -0.0877], + [-0.0378, 0.0805, -0.0982, ..., -0.0241, 0.0312, 0.1152], + ..., + [-0.0057, 0.0537, -0.0026, ..., -0.0142, 0.0265, -0.0559], + [ 0.0813, -0.0920, 0.0344, ..., 0.0053, 0.0037, -0.0429], + [-0.0170, -0.0719, 0.0713, ..., -0.0646, 0.0241, -0.0182]], + device='cuda:0'), grad: tensor([[ 9.5725e-05, 2.2268e-04, -2.2185e-04, ..., -7.1430e-04, + -2.2626e-04, -1.3030e-04], + [ 2.3210e-04, 5.9128e-04, 3.5191e-04, ..., 3.1052e-03, + 2.8896e-04, 5.5933e-04], + [ 1.5259e-04, 3.0208e-04, 5.5218e-04, ..., -1.1158e-03, + 5.5981e-04, -1.3990e-03], + ..., + [ 4.0936e-04, 8.3113e-04, 1.2732e-03, ..., 5.6610e-03, + 4.9400e-04, 4.8280e-04], + [ 2.9516e-04, 6.0987e-04, -6.1560e-04, ..., -3.0537e-03, + 1.2140e-03, 5.4550e-04], + [-2.3413e-04, -1.0061e-03, -2.6340e-03, ..., -1.7996e-03, + -3.3607e-03, -4.4417e-04]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0153, -0.0135, -0.0099, 0.0068, -0.0189, -0.0144, 0.0155, -0.0011, + 0.0186, -0.0153], device='cuda:0'), grad: tensor([-0.0268, 0.0224, -0.0156, 0.0120, 0.0110, -0.0105, 0.0139, 0.0305, + -0.0055, -0.0312], device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 216.83, cls_loss 0.5217 cls_loss_mapping 0.0068 cls_loss_causal 0.4897 re_mapping 0.0079 re_causal 0.0199 /// teacc 98.67 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.0613, -0.0749, -0.1333, ..., 0.1117, -0.0488, 0.0311], + [-0.0702, 0.0949, -0.0705, ..., 0.0435, -0.0526, -0.0882], + [-0.0381, 0.0817, -0.0979, ..., -0.0241, 0.0309, 0.1155], + ..., + [-0.0054, 0.0536, -0.0026, ..., -0.0131, 0.0265, -0.0567], + [ 0.0809, -0.0924, 0.0349, ..., 0.0053, 0.0041, -0.0435], + [-0.0167, -0.0709, 0.0709, ..., -0.0647, 0.0239, -0.0182]], + device='cuda:0'), grad: tensor([[-1.7519e-03, 4.7350e-04, 3.3975e-04, ..., 9.0075e-04, + -3.6788e-04, 2.3767e-06], + [ 2.4295e-04, -8.2016e-03, -2.6965e-04, ..., 4.5323e-04, + -3.4409e-03, 1.5867e-04], + [-1.8549e-04, 5.9853e-03, 1.0481e-03, ..., 1.4043e-04, + 3.5725e-03, -7.8535e-04], + ..., + [-2.1267e-03, 1.0176e-03, -1.1024e-03, ..., -1.0780e-02, + -1.3638e-03, 3.4165e-04], + [ 3.0947e-04, -2.8104e-05, 3.1424e-04, ..., 2.5654e-03, + 2.0075e-04, 2.0289e-04], + [ 1.1072e-03, 8.8406e-04, 2.7275e-03, ..., 4.7226e-03, + 2.1858e-03, 4.3726e-04]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0151, -0.0137, -0.0101, 0.0080, -0.0184, -0.0154, 0.0152, -0.0021, + 0.0194, -0.0146], device='cuda:0'), grad: tensor([ 0.0093, -0.0359, 0.0219, -0.0041, 0.0268, -0.0160, 0.0175, -0.0274, + -0.0063, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 217.07, cls_loss 0.4846 cls_loss_mapping 0.0069 cls_loss_causal 0.4611 re_mapping 0.0077 re_causal 0.0190 /// teacc 98.84 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.0607, -0.0735, -0.1340, ..., 0.1116, -0.0485, 0.0313], + [-0.0705, 0.0945, -0.0707, ..., 0.0429, -0.0524, -0.0891], + [-0.0395, 0.0839, -0.0981, ..., -0.0231, 0.0312, 0.1155], + ..., + [-0.0047, 0.0524, -0.0025, ..., -0.0139, 0.0260, -0.0569], + [ 0.0802, -0.0923, 0.0355, ..., 0.0051, 0.0049, -0.0435], + [-0.0172, -0.0708, 0.0707, ..., -0.0656, 0.0243, -0.0192]], + device='cuda:0'), grad: tensor([[ 2.7418e-04, 3.1471e-05, 8.2350e-04, ..., 1.5945e-03, + 3.6931e-04, 6.1464e-04], + [ 1.3905e-03, -1.4929e-06, 2.0242e-04, ..., 5.4502e-04, + 2.3976e-05, 2.4676e-04], + [-2.7237e-03, 3.1400e-04, 1.8895e-04, ..., -5.4588e-03, + 6.7186e-04, -1.0853e-03], + ..., + [ 5.4932e-04, -1.0614e-03, -8.0032e-03, ..., 1.2531e-03, + -2.1458e-03, 1.7226e-04], + [ 1.4362e-03, 1.7270e-05, 4.3488e-04, ..., 3.4904e-03, + 6.2764e-05, 7.4577e-04], + [ 1.0986e-03, 3.8058e-05, 7.7362e-03, ..., 2.3746e-03, + 9.5987e-04, 8.0204e-04]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0147, -0.0136, -0.0109, 0.0081, -0.0174, -0.0156, 0.0168, -0.0015, + 0.0188, -0.0162], device='cuda:0'), grad: tensor([ 0.0196, -0.0068, -0.0465, 0.0319, 0.0203, -0.0815, 0.0225, -0.0361, + 0.0276, 0.0490], device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 216.96, cls_loss 0.5129 cls_loss_mapping 0.0089 cls_loss_causal 0.4905 re_mapping 0.0083 re_causal 0.0206 /// teacc 98.72 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.0611, -0.0724, -0.1349, ..., 0.1120, -0.0494, 0.0314], + [-0.0706, 0.0944, -0.0702, ..., 0.0427, -0.0528, -0.0894], + [-0.0379, 0.0842, -0.0984, ..., -0.0231, 0.0312, 0.1156], + ..., + [-0.0042, 0.0525, -0.0028, ..., -0.0140, 0.0260, -0.0575], + [ 0.0792, -0.0922, 0.0347, ..., 0.0054, 0.0048, -0.0424], + [-0.0167, -0.0707, 0.0709, ..., -0.0662, 0.0244, -0.0191]], + device='cuda:0'), grad: tensor([[ 1.8291e-03, 5.3453e-04, 8.5878e-04, ..., 1.7529e-03, + 1.9813e-04, 8.0442e-04], + [ 1.6773e-04, 3.8314e-04, 5.3978e-04, ..., 7.7581e-04, + 1.1235e-04, 3.7384e-04], + [ 3.3035e-03, 6.2180e-04, 5.1403e-04, ..., 5.8651e-04, + 7.2384e-04, 2.5678e-04], + ..., + [-1.0231e-02, 4.7684e-04, 8.0442e-04, ..., -4.3373e-03, + 3.7694e-04, 1.5104e-04], + [ 3.9411e-04, 5.3263e-04, 5.7697e-04, ..., 1.2197e-03, + 5.6124e-04, 4.0841e-04], + [ 8.7070e-04, 1.3189e-03, -3.3474e-03, ..., 4.7970e-04, + 2.0885e-03, 6.9678e-05]], device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0161, -0.0141, -0.0111, 0.0075, -0.0170, -0.0160, 0.0164, -0.0010, + 0.0196, -0.0172], device='cuda:0'), grad: tensor([ 0.0344, 0.0220, 0.0292, 0.0138, 0.0038, -0.0074, 0.0132, -0.0683, + -0.0392, -0.0014], device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 216.88, cls_loss 0.4912 cls_loss_mapping 0.0044 cls_loss_causal 0.4672 re_mapping 0.0078 re_causal 0.0196 /// teacc 98.79 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.0621, -0.0745, -0.1346, ..., 0.1122, -0.0495, 0.0310], + [-0.0716, 0.0944, -0.0711, ..., 0.0418, -0.0534, -0.0887], + [-0.0381, 0.0849, -0.0982, ..., -0.0223, 0.0320, 0.1162], + ..., + [-0.0043, 0.0522, -0.0033, ..., -0.0146, 0.0251, -0.0572], + [ 0.0794, -0.0915, 0.0366, ..., 0.0059, 0.0050, -0.0424], + [-0.0163, -0.0704, 0.0708, ..., -0.0664, 0.0248, -0.0200]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.7011e-04, 5.8532e-05, ..., -3.3379e-04, + -2.0278e-04, -5.8699e-04], + [ 0.0000e+00, -1.1981e-05, 8.5592e-04, ..., 6.0987e-04, + 6.0606e-04, 2.3410e-05], + [ 7.1712e-08, 1.4579e-04, 6.7949e-04, ..., 9.8705e-04, + 4.0793e-04, 4.6754e-04], + ..., + [ 5.1223e-08, -3.1257e-04, -1.9760e-02, ..., 8.8739e-04, + -6.0844e-04, 2.7370e-04], + [-1.6019e-07, -1.7953e-04, -9.1324e-03, ..., -9.7513e-04, + -1.3199e-02, -2.9659e-03], + [-2.7008e-08, 6.2585e-05, 2.0508e-02, ..., -3.0899e-03, + 3.8147e-03, 4.9162e-04]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0155, -0.0138, -0.0100, 0.0079, -0.0175, -0.0167, 0.0171, -0.0023, + 0.0201, -0.0170], device='cuda:0'), grad: tensor([ 0.0061, -0.0130, -0.0149, 0.0260, -0.0115, 0.0049, 0.0131, -0.0123, + -0.0251, 0.0267], device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 217.08, cls_loss 0.5409 cls_loss_mapping 0.0078 cls_loss_causal 0.5114 re_mapping 0.0066 re_causal 0.0177 /// teacc 98.93 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.0625, -0.0762, -0.1356, ..., 0.1127, -0.0500, 0.0303], + [-0.0714, 0.0951, -0.0718, ..., 0.0417, -0.0541, -0.0894], + [-0.0381, 0.0853, -0.0988, ..., -0.0231, 0.0324, 0.1160], + ..., + [-0.0050, 0.0516, -0.0034, ..., -0.0156, 0.0251, -0.0593], + [ 0.0803, -0.0924, 0.0362, ..., 0.0053, 0.0039, -0.0420], + [-0.0161, -0.0707, 0.0715, ..., -0.0663, 0.0256, -0.0200]], + device='cuda:0'), grad: tensor([[ 7.7561e-06, 2.2184e-06, 2.4354e-04, ..., 2.5439e-04, + 3.1114e-04, 2.2864e-04], + [ 6.5472e-07, 1.7816e-06, 2.3246e-04, ..., 3.9649e-04, + 3.2830e-04, 1.5974e-04], + [ 8.9183e-06, 1.3185e-04, 4.2295e-04, ..., 5.9128e-04, + 9.6083e-04, 3.3164e-04], + ..., + [ 1.7844e-06, -1.3435e-04, -5.3825e-03, ..., 7.5400e-05, + -1.6068e-02, -1.7872e-03], + [ 6.5684e-05, -1.6943e-05, -1.5383e-03, ..., -2.2411e-04, + -2.1744e-03, -1.9588e-03], + [ 4.4368e-06, 1.7166e-05, -1.7586e-03, ..., -4.5624e-03, + -2.6360e-03, -7.2050e-04]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0152, -0.0142, -0.0106, 0.0088, -0.0165, -0.0165, 0.0174, -0.0031, + 0.0197, -0.0169], device='cuda:0'), grad: tensor([ 0.0060, 0.0073, 0.0077, 0.0379, 0.0099, 0.0050, 0.0226, -0.0262, + -0.0170, -0.0533], device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 216.45, cls_loss 0.5291 cls_loss_mapping 0.0066 cls_loss_causal 0.5031 re_mapping 0.0074 re_causal 0.0201 /// teacc 98.72 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.0621, -0.0777, -0.1356, ..., 0.1128, -0.0512, 0.0303], + [-0.0726, 0.0952, -0.0708, ..., 0.0409, -0.0528, -0.0910], + [-0.0384, 0.0848, -0.0988, ..., -0.0227, 0.0321, 0.1160], + ..., + [-0.0040, 0.0512, -0.0027, ..., -0.0126, 0.0262, -0.0595], + [ 0.0804, -0.0906, 0.0347, ..., 0.0050, 0.0038, -0.0417], + [-0.0155, -0.0711, 0.0718, ..., -0.0676, 0.0255, -0.0210]], + device='cuda:0'), grad: tensor([[ 2.0814e-04, -1.0834e-03, -7.7009e-05, ..., 9.3079e-04, + -4.2605e-04, -8.4877e-04], + [ 1.2798e-03, -1.1330e-03, 1.3971e-04, ..., -2.2602e-03, + 5.3596e-04, -1.0405e-03], + [ 7.3385e-04, -3.2825e-03, -1.7948e-03, ..., -4.6611e-04, + -3.2997e-03, -1.9264e-03], + ..., + [-8.9788e-04, -3.7136e-03, -8.2092e-03, ..., 2.7227e-04, + -8.3618e-03, 2.0504e-03], + [ 6.7520e-04, 9.7036e-04, 1.8013e-04, ..., 5.0735e-04, + 7.8344e-04, 1.2493e-03], + [ 9.5673e-03, 5.8823e-03, 2.8519e-02, ..., 2.7504e-03, + 1.8616e-02, 1.5297e-03]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0147, -0.0140, -0.0105, 0.0084, -0.0173, -0.0165, 0.0188, -0.0019, + 0.0194, -0.0178], device='cuda:0'), grad: tensor([-0.0426, -0.0303, 0.0028, -0.0482, -0.0597, 0.0385, 0.0264, 0.0192, + 0.0216, 0.0723], device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 217.14, cls_loss 0.5115 cls_loss_mapping 0.0089 cls_loss_causal 0.4859 re_mapping 0.0074 re_causal 0.0194 /// teacc 98.82 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.0621, -0.0776, -0.1353, ..., 0.1134, -0.0525, 0.0295], + [-0.0723, 0.0952, -0.0710, ..., 0.0398, -0.0529, -0.0915], + [-0.0379, 0.0848, -0.0978, ..., -0.0232, 0.0325, 0.1171], + ..., + [-0.0031, 0.0518, -0.0029, ..., -0.0125, 0.0263, -0.0593], + [ 0.0802, -0.0896, 0.0351, ..., 0.0057, 0.0035, -0.0418], + [-0.0161, -0.0712, 0.0716, ..., -0.0686, 0.0261, -0.0223]], + device='cuda:0'), grad: tensor([[ 3.3587e-05, 2.6560e-04, 6.2847e-04, ..., -2.0046e-03, + 1.2131e-03, -2.6550e-03], + [-9.2745e-04, -3.2406e-03, 1.3494e-04, ..., -6.8016e-03, + -1.9035e-03, -1.6489e-03], + [ 3.3188e-04, 1.5001e-03, 8.4102e-05, ..., 3.8681e-03, + 1.5907e-03, 2.4776e-03], + ..., + [ 4.3780e-05, -1.2083e-03, -5.3453e-04, ..., -1.8015e-03, + -1.5764e-03, -8.8024e-04], + [ 5.5164e-05, 5.2261e-04, 2.7895e-04, ..., 1.6031e-03, + 1.1902e-03, 1.3151e-03], + [ 1.9297e-05, 3.8576e-04, 1.2522e-03, ..., 2.7485e-03, + -1.6046e-04, -4.5204e-04]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0154, -0.0142, -0.0103, 0.0091, -0.0169, -0.0167, 0.0187, -0.0022, + 0.0184, -0.0178], device='cuda:0'), grad: tensor([-0.0107, -0.0410, 0.0337, 0.0187, 0.0055, 0.0233, 0.0015, -0.0141, + 0.0212, -0.0382], device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 216.95, cls_loss 0.5210 cls_loss_mapping 0.0042 cls_loss_causal 0.4903 re_mapping 0.0075 re_causal 0.0204 /// teacc 98.78 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.0617, -0.0783, -0.1350, ..., 0.1132, -0.0525, 0.0292], + [-0.0734, 0.0958, -0.0727, ..., 0.0415, -0.0539, -0.0907], + [-0.0386, 0.0848, -0.0964, ..., -0.0229, 0.0328, 0.1159], + ..., + [-0.0025, 0.0524, -0.0027, ..., -0.0130, 0.0269, -0.0590], + [ 0.0802, -0.0895, 0.0355, ..., 0.0058, 0.0030, -0.0417], + [-0.0157, -0.0709, 0.0719, ..., -0.0685, 0.0273, -0.0210]], + device='cuda:0'), grad: tensor([[ 1.5287e-03, 1.7607e-04, 9.5606e-05, ..., 1.1969e-03, + 3.1185e-04, 3.1948e-04], + [ 6.4945e-04, -1.5478e-03, 1.8075e-05, ..., -6.2599e-03, + 4.9448e-04, -3.6335e-04], + [ 2.1687e-03, 6.2256e-03, 3.9978e-03, ..., 2.1572e-03, + 6.4812e-03, 5.2185e-03], + ..., + [-2.6123e-02, 1.1559e-03, 5.2299e-03, ..., 1.0500e-03, + 2.0752e-03, 8.0681e-04], + [ 2.3823e-03, 8.5115e-04, 4.8542e-04, ..., 4.8027e-03, + -5.4092e-03, 8.9598e-04], + [ 1.1444e-02, -7.7209e-03, -1.0399e-02, ..., -2.0924e-03, + -7.3128e-03, -8.3542e-03]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0154, -0.0147, -0.0102, 0.0078, -0.0174, -0.0162, 0.0180, -0.0016, + 0.0189, -0.0168], device='cuda:0'), grad: tensor([ 0.0162, -0.0187, 0.0405, -0.0088, 0.0260, 0.0117, -0.0334, 0.0203, + -0.0002, -0.0537], device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 216.37, cls_loss 0.5350 cls_loss_mapping 0.0048 cls_loss_causal 0.5051 re_mapping 0.0077 re_causal 0.0210 /// teacc 98.68 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.0597, -0.0778, -0.1357, ..., 0.1137, -0.0527, 0.0300], + [-0.0736, 0.0962, -0.0715, ..., 0.0410, -0.0536, -0.0913], + [-0.0383, 0.0851, -0.0969, ..., -0.0234, 0.0319, 0.1165], + ..., + [-0.0025, 0.0520, -0.0022, ..., -0.0138, 0.0280, -0.0611], + [ 0.0810, -0.0885, 0.0345, ..., 0.0058, 0.0033, -0.0409], + [-0.0162, -0.0711, 0.0717, ..., -0.0677, 0.0267, -0.0199]], + device='cuda:0'), grad: tensor([[ 0.0019, 0.0004, 0.0005, ..., 0.0010, 0.0006, 0.0005], + [-0.0033, 0.0028, 0.0034, ..., 0.0022, 0.0014, 0.0005], + [ 0.0007, -0.0056, 0.0002, ..., -0.0049, -0.0037, -0.0008], + ..., + [ 0.0005, -0.0003, -0.0016, ..., 0.0014, 0.0013, 0.0004], + [ 0.0013, 0.0009, 0.0011, ..., 0.0011, 0.0010, -0.0002], + [-0.0004, -0.0001, -0.0143, ..., -0.0023, -0.0067, 0.0002]], + device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0161, -0.0142, -0.0101, 0.0081, -0.0172, -0.0165, 0.0177, -0.0025, + 0.0194, -0.0174], device='cuda:0'), grad: tensor([ 0.0342, 0.0371, -0.0218, -0.0259, -0.0205, 0.0232, 0.0345, -0.0022, + -0.0544, -0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 216.80, cls_loss 0.4837 cls_loss_mapping 0.0059 cls_loss_causal 0.4579 re_mapping 0.0076 re_causal 0.0202 /// teacc 98.86 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.0614, -0.0790, -0.1340, ..., 0.1128, -0.0520, 0.0310], + [-0.0733, 0.0969, -0.0716, ..., 0.0414, -0.0540, -0.0917], + [-0.0392, 0.0856, -0.0964, ..., -0.0233, 0.0316, 0.1156], + ..., + [-0.0029, 0.0517, -0.0023, ..., -0.0144, 0.0276, -0.0599], + [ 0.0804, -0.0903, 0.0336, ..., 0.0053, 0.0038, -0.0416], + [-0.0152, -0.0711, 0.0726, ..., -0.0671, 0.0273, -0.0205]], + device='cuda:0'), grad: tensor([[ 4.0555e-04, 2.0933e-04, 4.6343e-05, ..., -8.8978e-04, + 3.5954e-04, 1.2541e-04], + [ 2.2259e-03, -1.7614e-03, 5.9903e-05, ..., -1.8873e-03, + 4.3720e-05, -3.7122e-04], + [ 3.9315e-04, 6.3229e-04, 6.6876e-05, ..., 9.6226e-04, + 1.1768e-03, 1.0281e-03], + ..., + [-1.2379e-03, 1.7405e-04, -1.9050e-04, ..., 7.5340e-04, + -9.1457e-04, 2.5964e-04], + [ 1.3199e-03, 3.3665e-04, 2.0237e-03, ..., 6.9714e-04, + 2.4223e-03, 3.7813e-04], + [ 2.9898e-04, 5.3406e-04, -1.3939e-02, ..., -3.9101e-03, + -1.4366e-02, 2.3198e-04]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0152, -0.0145, -0.0105, 0.0095, -0.0180, -0.0161, 0.0164, -0.0020, + 0.0197, -0.0163], device='cuda:0'), grad: tensor([ 0.0074, -0.0336, 0.0226, -0.0361, -0.0207, 0.0319, 0.0031, 0.0136, + 0.0165, -0.0049], device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 216.47, cls_loss 0.5051 cls_loss_mapping 0.0068 cls_loss_causal 0.4786 re_mapping 0.0073 re_causal 0.0190 /// teacc 98.85 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.0608, -0.0797, -0.1348, ..., 0.1127, -0.0532, 0.0313], + [-0.0745, 0.0990, -0.0723, ..., 0.0420, -0.0550, -0.0910], + [-0.0405, 0.0850, -0.0968, ..., -0.0233, 0.0299, 0.1144], + ..., + [-0.0019, 0.0523, -0.0014, ..., -0.0136, 0.0284, -0.0592], + [ 0.0818, -0.0922, 0.0323, ..., 0.0050, 0.0033, -0.0423], + [-0.0146, -0.0706, 0.0730, ..., -0.0676, 0.0288, -0.0198]], + device='cuda:0'), grad: tensor([[ 0.0002, -0.0041, 0.0004, ..., -0.0029, -0.0005, -0.0026], + [ 0.0022, 0.0095, 0.0005, ..., 0.0019, 0.0008, 0.0007], + [ 0.0005, -0.0056, 0.0004, ..., 0.0011, 0.0009, 0.0011], + ..., + [-0.0057, -0.0037, 0.0010, ..., 0.0012, 0.0007, 0.0003], + [ 0.0009, 0.0028, 0.0013, ..., -0.0037, 0.0024, -0.0012], + [ 0.0012, 0.0029, 0.0020, ..., 0.0026, 0.0025, 0.0012]], + device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0149, -0.0142, -0.0107, 0.0084, -0.0183, -0.0168, 0.0165, -0.0019, + 0.0202, -0.0151], device='cuda:0'), grad: tensor([-0.0605, 0.0297, 0.0090, 0.0347, -0.0568, 0.0114, 0.0100, -0.0237, + 0.0075, 0.0387], device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 216.94, cls_loss 0.5249 cls_loss_mapping 0.0062 cls_loss_causal 0.4971 re_mapping 0.0077 re_causal 0.0194 /// teacc 98.79 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.0605, -0.0801, -0.1347, ..., 0.1127, -0.0539, 0.0313], + [-0.0753, 0.0988, -0.0719, ..., 0.0424, -0.0538, -0.0912], + [-0.0400, 0.0859, -0.0977, ..., -0.0231, 0.0291, 0.1138], + ..., + [-0.0019, 0.0525, -0.0016, ..., -0.0142, 0.0290, -0.0592], + [ 0.0820, -0.0928, 0.0325, ..., 0.0056, 0.0032, -0.0425], + [-0.0141, -0.0716, 0.0732, ..., -0.0674, 0.0286, -0.0192]], + device='cuda:0'), grad: tensor([[ 1.4579e-04, 2.6405e-05, 2.9159e-04, ..., 4.7755e-04, + 1.8334e-04, 6.7294e-05], + [ 1.1225e-03, -9.7007e-06, -3.7632e-03, ..., -4.4327e-03, + -1.0157e-03, -8.7070e-04], + [ 3.8528e-04, -2.1553e-04, 1.6642e-04, ..., 1.7738e-04, + -1.3514e-03, -1.1768e-03], + ..., + [ 3.6478e-04, -4.7064e-04, -6.6452e-03, ..., -1.5898e-03, + -7.5989e-03, -8.5652e-05], + [-4.0894e-03, 2.4247e-04, 1.2312e-03, ..., -1.5268e-03, + 1.5020e-03, 9.6846e-04], + [ 7.6532e-04, 2.4939e-04, 5.9586e-03, ..., 2.7943e-03, + 5.8289e-03, 2.0349e-04]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0149, -0.0135, -0.0108, 0.0091, -0.0178, -0.0177, 0.0160, -0.0018, + 0.0197, -0.0149], device='cuda:0'), grad: tensor([ 0.0071, -0.0120, 0.0059, 0.0105, 0.0174, 0.0108, 0.0047, -0.0353, + -0.0031, -0.0061], device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 216.58, cls_loss 0.4991 cls_loss_mapping 0.0060 cls_loss_causal 0.4645 re_mapping 0.0079 re_causal 0.0186 /// teacc 98.72 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.0601, -0.0803, -0.1349, ..., 0.1129, -0.0528, 0.0321], + [-0.0755, 0.0988, -0.0715, ..., 0.0424, -0.0523, -0.0913], + [-0.0373, 0.0845, -0.0980, ..., -0.0225, 0.0282, 0.1141], + ..., + [-0.0018, 0.0545, -0.0033, ..., -0.0137, 0.0287, -0.0581], + [ 0.0814, -0.0930, 0.0332, ..., 0.0049, 0.0039, -0.0425], + [-0.0131, -0.0702, 0.0723, ..., -0.0675, 0.0280, -0.0194]], + device='cuda:0'), grad: tensor([[ 1.3056e-03, 3.2592e-04, 4.0382e-06, ..., 1.5616e-04, + 2.3937e-04, 2.1482e-04], + [ 5.4398e-03, 7.1602e-03, 1.6376e-05, ..., 8.0967e-04, + 6.0654e-04, 5.3930e-04], + [ 2.8439e-03, -3.6144e-03, 1.3840e-04, ..., -2.4261e-03, + -1.5697e-03, -3.0556e-03], + ..., + [ 2.3341e-04, 7.6580e-04, -7.9751e-05, ..., 2.7514e-04, + 3.1400e-04, 4.7088e-04], + [-2.1759e-02, -7.3700e-03, 1.7142e-04, ..., -1.8907e-04, + 1.0681e-03, 8.7547e-04], + [ 8.5678e-03, 7.9584e-04, 1.6556e-03, ..., 3.4189e-04, + -2.1243e-04, 2.8348e-04]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0161, -0.0143, -0.0103, 0.0091, -0.0172, -0.0182, 0.0154, -0.0015, + 0.0190, -0.0151], device='cuda:0'), grad: tensor([ 0.0070, 0.0408, -0.0199, 0.0172, -0.0190, -0.0072, -0.0043, 0.0110, + -0.0076, -0.0180], device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 216.40, cls_loss 0.5014 cls_loss_mapping 0.0050 cls_loss_causal 0.4745 re_mapping 0.0080 re_causal 0.0206 /// teacc 98.61 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.0595, -0.0806, -0.1361, ..., 0.1124, -0.0532, 0.0315], + [-0.0754, 0.0987, -0.0712, ..., 0.0425, -0.0522, -0.0914], + [-0.0392, 0.0850, -0.0986, ..., -0.0229, 0.0271, 0.1140], + ..., + [-0.0013, 0.0548, -0.0024, ..., -0.0134, 0.0300, -0.0588], + [ 0.0824, -0.0941, 0.0348, ..., 0.0056, 0.0045, -0.0418], + [-0.0150, -0.0699, 0.0714, ..., -0.0683, 0.0274, -0.0194]], + device='cuda:0'), grad: tensor([[ 1.9407e-03, 7.3862e-04, 2.1911e-04, ..., 1.3816e-04, + 2.9106e-03, 2.2626e-04], + [ 6.9559e-05, -3.4313e-03, 6.4373e-05, ..., -2.4834e-03, + -1.9217e-04, -1.9445e-03], + [ 2.6751e-04, -7.8888e-03, 5.3120e-04, ..., -6.5851e-04, + -7.9203e-04, -3.4485e-03], + ..., + [ 2.2340e-04, 7.6370e-03, -9.8228e-04, ..., 2.0771e-03, + 5.9414e-04, 3.8719e-03], + [-1.0750e-02, 7.1764e-04, 3.5930e-04, ..., 2.2638e-04, + -8.4381e-03, 2.2054e-04], + [ 1.9150e-03, 1.0605e-03, 1.3113e-03, ..., 2.1982e-04, + 3.4161e-03, 2.7609e-04]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0159, -0.0140, -0.0104, 0.0087, -0.0174, -0.0181, 0.0152, -0.0013, + 0.0196, -0.0153], device='cuda:0'), grad: tensor([ 0.0142, -0.0199, 0.0025, -0.0239, 0.0116, -0.0166, 0.0214, 0.0050, + -0.0167, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 216.93, cls_loss 0.5011 cls_loss_mapping 0.0061 cls_loss_causal 0.4730 re_mapping 0.0073 re_causal 0.0183 /// teacc 98.68 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.0579, -0.0806, -0.1357, ..., 0.1124, -0.0525, 0.0318], + [-0.0750, 0.0979, -0.0709, ..., 0.0421, -0.0512, -0.0919], + [-0.0401, 0.0855, -0.0999, ..., -0.0226, 0.0271, 0.1149], + ..., + [-0.0033, 0.0543, -0.0032, ..., -0.0133, 0.0294, -0.0595], + [ 0.0827, -0.0942, 0.0369, ..., 0.0062, 0.0061, -0.0423], + [-0.0155, -0.0691, 0.0713, ..., -0.0685, 0.0271, -0.0172]], + device='cuda:0'), grad: tensor([[ 9.4399e-06, 1.0262e-03, 1.2159e-04, ..., 4.3368e-04, + 1.2767e-04, 2.7180e-04], + [ 9.2089e-06, 1.6670e-03, 4.9210e-04, ..., 1.0061e-03, + 1.2755e-04, 4.2820e-04], + [ 3.4976e-04, -8.1253e-03, 1.6365e-03, ..., 4.8518e-04, + 8.8406e-04, 3.1400e-04], + ..., + [ 9.1612e-05, 1.4467e-03, 1.3542e-03, ..., 2.3997e-04, + 8.2684e-04, 5.2601e-05], + [ 1.7700e-03, 8.6546e-04, 7.0419e-03, ..., 2.0313e-03, + 2.8687e-03, 9.5701e-04], + [ 2.3282e-04, 3.2020e-04, -1.1349e-03, ..., -1.8282e-03, + -2.1744e-03, -2.7828e-03]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0166, -0.0143, -0.0106, 0.0081, -0.0178, -0.0187, 0.0156, -0.0019, + 0.0199, -0.0140], device='cuda:0'), grad: tensor([ 0.0109, -0.0062, -0.0158, -0.0188, -0.0127, -0.0083, 0.0025, 0.0191, + 0.0401, -0.0108], device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 216.72, cls_loss 0.4890 cls_loss_mapping 0.0038 cls_loss_causal 0.4538 re_mapping 0.0078 re_causal 0.0202 /// teacc 98.70 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.0589, -0.0825, -0.1360, ..., 0.1129, -0.0527, 0.0322], + [-0.0752, 0.0985, -0.0714, ..., 0.0416, -0.0499, -0.0916], + [-0.0404, 0.0857, -0.0993, ..., -0.0230, 0.0276, 0.1158], + ..., + [-0.0033, 0.0543, -0.0046, ..., -0.0140, 0.0289, -0.0591], + [ 0.0818, -0.0928, 0.0376, ..., 0.0071, 0.0071, -0.0419], + [-0.0162, -0.0688, 0.0723, ..., -0.0698, 0.0270, -0.0168]], + device='cuda:0'), grad: tensor([[ 1.3970e-06, 1.2696e-04, 1.4775e-05, ..., 3.0976e-06, + -2.4494e-07, -1.6146e-03], + [-3.5893e-06, 1.0424e-03, 1.2137e-05, ..., -2.5555e-05, + 1.0198e-06, 2.6751e-04], + [-1.2970e-04, 5.3453e-04, 1.9729e-05, ..., 2.1346e-06, + 1.4547e-06, 1.7083e-04], + ..., + [ 3.2514e-05, -1.3227e-03, 7.7391e-04, ..., 4.1306e-05, + 5.2452e-04, 2.1422e-04], + [ 4.9740e-05, -1.0643e-03, 1.4627e-04, ..., 5.4911e-06, + 5.6118e-05, -7.1907e-04], + [-1.2934e-04, -7.1824e-06, -1.2712e-03, ..., -3.2961e-05, + -7.9346e-04, 2.6846e-04]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0162, -0.0142, -0.0097, 0.0073, -0.0171, -0.0174, 0.0144, -0.0030, + 0.0209, -0.0143], device='cuda:0'), grad: tensor([-0.0230, 0.0075, 0.0059, 0.0058, -0.0119, 0.0061, 0.0261, 0.0041, + -0.0241, 0.0036], device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 216.76, cls_loss 0.5263 cls_loss_mapping 0.0053 cls_loss_causal 0.5006 re_mapping 0.0071 re_causal 0.0187 /// teacc 98.69 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.0585, -0.0828, -0.1364, ..., 0.1135, -0.0524, 0.0324], + [-0.0738, 0.0983, -0.0697, ..., 0.0424, -0.0495, -0.0919], + [-0.0404, 0.0857, -0.0980, ..., -0.0233, 0.0289, 0.1158], + ..., + [-0.0043, 0.0533, -0.0048, ..., -0.0147, 0.0277, -0.0607], + [ 0.0831, -0.0928, 0.0363, ..., 0.0066, 0.0060, -0.0418], + [-0.0155, -0.0671, 0.0728, ..., -0.0706, 0.0271, -0.0170]], + device='cuda:0'), grad: tensor([[ 4.3353e-07, -2.3689e-03, 2.1517e-04, ..., -4.9973e-03, + 1.0568e-04, -2.6016e-03], + [-1.4948e-07, 9.3818e-05, 5.0783e-04, ..., 5.4121e-04, + 2.7370e-04, 1.0127e-04], + [ 4.8690e-06, 7.0076e-03, 3.4142e-04, ..., 1.3023e-02, + 3.3569e-03, 2.8477e-03], + ..., + [ 2.2165e-07, 1.2517e-04, 4.1199e-04, ..., 2.7027e-03, + 6.3419e-04, 3.0947e-04], + [ 7.7009e-04, 1.3628e-03, 2.8777e-04, ..., 3.2921e-03, + 3.9697e-04, 6.7997e-04], + [ 7.4767e-06, 4.6463e-03, 7.7553e-03, ..., 1.2789e-03, + 3.3498e-04, 1.3506e-04]], device='cuda:0') +Epoch 237, bias, value: tensor([ 0.0168, -0.0137, -0.0098, 0.0074, -0.0161, -0.0160, 0.0125, -0.0042, + 0.0201, -0.0139], device='cuda:0'), grad: tensor([-0.0137, 0.0107, 0.0407, -0.0130, -0.0081, -0.0255, -0.0139, 0.0046, + 0.0134, 0.0049], device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 216.23, cls_loss 0.4865 cls_loss_mapping 0.0048 cls_loss_causal 0.4582 re_mapping 0.0073 re_causal 0.0178 /// teacc 98.81 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.0591, -0.0827, -0.1379, ..., 0.1139, -0.0527, 0.0332], + [-0.0721, 0.0973, -0.0699, ..., 0.0402, -0.0493, -0.0917], + [-0.0409, 0.0867, -0.0966, ..., -0.0242, 0.0299, 0.1154], + ..., + [-0.0042, 0.0534, -0.0037, ..., -0.0135, 0.0288, -0.0583], + [ 0.0819, -0.0940, 0.0357, ..., 0.0078, 0.0053, -0.0428], + [-0.0152, -0.0679, 0.0718, ..., -0.0713, 0.0271, -0.0166]], + device='cuda:0'), grad: tensor([[ 2.7120e-05, 3.5739e-04, -2.8477e-03, ..., -4.8714e-03, + -4.3755e-03, -8.2731e-04], + [-1.9016e-03, -9.3079e-03, 3.2520e-04, ..., -3.6240e-03, + -6.6757e-04, -4.2000e-03], + [ 1.2569e-03, 5.4703e-03, 1.7519e-03, ..., 8.4162e-04, + 2.3365e-03, 3.1414e-03], + ..., + [ 1.9395e-04, -4.6849e-04, 5.2595e-04, ..., 1.5926e-03, + 3.2496e-04, 5.0735e-04], + [-6.8321e-03, -8.5020e-04, 1.2178e-03, ..., 1.9503e-03, + 1.2808e-03, -2.9240e-03], + [ 2.1172e-04, 5.0402e-04, 3.7918e-03, ..., 1.4963e-03, + 3.9215e-03, 3.9577e-04]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0164, -0.0132, -0.0098, 0.0084, -0.0171, -0.0168, 0.0126, -0.0025, + 0.0185, -0.0136], device='cuda:0'), grad: tensor([-0.0077, -0.0419, 0.0307, 0.0319, -0.0267, 0.0284, 0.0151, -0.0091, + -0.0140, -0.0066], device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 216.87, cls_loss 0.5116 cls_loss_mapping 0.0039 cls_loss_causal 0.4790 re_mapping 0.0073 re_causal 0.0185 /// teacc 98.93 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.0580, -0.0829, -0.1380, ..., 0.1128, -0.0530, 0.0325], + [-0.0722, 0.0971, -0.0712, ..., 0.0404, -0.0497, -0.0913], + [-0.0413, 0.0860, -0.0966, ..., -0.0242, 0.0297, 0.1147], + ..., + [-0.0044, 0.0535, -0.0036, ..., -0.0138, 0.0292, -0.0585], + [ 0.0828, -0.0949, 0.0355, ..., 0.0077, 0.0049, -0.0418], + [-0.0151, -0.0684, 0.0722, ..., -0.0704, 0.0273, -0.0172]], + device='cuda:0'), grad: tensor([[-3.3736e-04, 1.1390e-04, 6.8617e-04, ..., 3.5357e-04, + 3.7060e-03, 2.5940e-03], + [-4.1723e-06, -4.4882e-05, 4.7952e-05, ..., -3.2093e-06, + -1.5283e-04, 7.0870e-05], + [ 7.0855e-06, 7.5936e-05, 4.0483e-04, ..., 1.9538e-04, + 2.0828e-03, 1.4420e-03], + ..., + [-2.8044e-05, 6.7616e-04, 3.2368e-03, ..., 1.1131e-05, + 2.6741e-03, 6.8426e-05], + [ 7.8231e-06, 1.0677e-05, 5.5027e-04, ..., 5.4151e-05, + 9.1219e-04, 3.6836e-04], + [ 3.7253e-05, -7.1096e-04, -3.9339e-04, ..., 2.0802e-05, + 1.0079e-04, 1.4520e-04]], device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0181, -0.0135, -0.0102, 0.0082, -0.0167, -0.0168, 0.0131, -0.0034, + 0.0180, -0.0135], device='cuda:0'), grad: tensor([ 0.0233, -0.0208, -0.0138, 0.0154, -0.0099, -0.0519, 0.0124, 0.0154, + 0.0114, 0.0184], device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 216.79, cls_loss 0.4948 cls_loss_mapping 0.0048 cls_loss_causal 0.4653 re_mapping 0.0073 re_causal 0.0181 /// teacc 98.92 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.0567, -0.0818, -0.1357, ..., 0.1126, -0.0509, 0.0323], + [-0.0740, 0.0978, -0.0709, ..., 0.0413, -0.0499, -0.0913], + [-0.0393, 0.0856, -0.0981, ..., -0.0249, 0.0284, 0.1151], + ..., + [-0.0032, 0.0544, -0.0043, ..., -0.0146, 0.0299, -0.0578], + [ 0.0826, -0.0957, 0.0362, ..., 0.0082, 0.0062, -0.0424], + [-0.0157, -0.0697, 0.0718, ..., -0.0700, 0.0264, -0.0178]], + device='cuda:0'), grad: tensor([[ 6.0463e-04, 3.1567e-04, 7.7868e-04, ..., 6.4707e-04, + 1.1625e-03, 3.7408e-04], + [-6.6261e-03, -2.6870e-04, 7.9346e-04, ..., 1.2608e-03, + 1.7080e-03, 9.0659e-05], + [ 2.6345e-04, 8.4448e-04, 3.4866e-03, ..., 3.0384e-03, + 5.6458e-03, 5.1022e-04], + ..., + [ 1.1683e-04, -2.5654e-03, -5.0850e-03, ..., -4.3945e-03, + -4.8757e-04, 8.7380e-05], + [ 9.7036e-04, 3.9768e-04, -1.2054e-02, ..., -4.0703e-03, + -1.6846e-02, -8.6746e-03], + [ 1.6642e-04, 9.6369e-04, 9.6893e-03, ..., 1.4877e-03, + 1.5152e-02, 7.0572e-03]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0178, -0.0124, -0.0102, 0.0067, -0.0172, -0.0168, 0.0132, -0.0032, + 0.0186, -0.0135], device='cuda:0'), grad: tensor([ 0.0199, 0.0116, -0.0010, -0.0218, -0.0396, 0.0010, 0.0091, 0.0092, + -0.0385, 0.0501], device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 217.16, cls_loss 0.5066 cls_loss_mapping 0.0046 cls_loss_causal 0.4791 re_mapping 0.0070 re_causal 0.0176 /// teacc 98.79 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.0561, -0.0816, -0.1368, ..., 0.1111, -0.0506, 0.0325], + [-0.0739, 0.0990, -0.0709, ..., 0.0410, -0.0505, -0.0915], + [-0.0405, 0.0853, -0.0990, ..., -0.0245, 0.0274, 0.1154], + ..., + [-0.0045, 0.0542, -0.0040, ..., -0.0146, 0.0301, -0.0581], + [ 0.0823, -0.0966, 0.0363, ..., 0.0090, 0.0070, -0.0415], + [-0.0149, -0.0703, 0.0714, ..., -0.0706, 0.0257, -0.0198]], + device='cuda:0'), grad: tensor([[ 5.0515e-05, -7.4339e-04, 2.9278e-04, ..., -2.5539e-03, + -3.0708e-03, -6.0225e-04], + [ 1.1307e-04, 6.1631e-05, 3.5596e-04, ..., 9.0790e-04, + 3.7384e-04, 3.5834e-04], + [ 6.5863e-05, 2.7800e-04, -1.8656e-04, ..., 1.0099e-03, + 3.0136e-03, 1.5819e-04], + ..., + [-9.1629e-03, -2.5501e-03, 4.4327e-03, ..., -1.4153e-03, + -4.9162e-04, -9.1982e-04], + [ 5.4300e-05, 2.0802e-04, 6.9284e-04, ..., 6.1083e-04, + 6.6948e-04, 2.2554e-04], + [ 6.9022e-05, -1.6584e-03, 8.6899e-03, ..., -1.3647e-03, + 7.7744e-03, -2.3532e-04]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0179, -0.0127, -0.0113, 0.0061, -0.0166, -0.0174, 0.0149, -0.0029, + 0.0194, -0.0141], device='cuda:0'), grad: tensor([-0.0284, 0.0229, -0.0123, 0.0174, -0.0051, 0.0107, 0.0178, -0.0609, + 0.0152, 0.0228], device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 216.95, cls_loss 0.5073 cls_loss_mapping 0.0064 cls_loss_causal 0.4783 re_mapping 0.0072 re_causal 0.0183 /// teacc 98.78 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.0555, -0.0824, -0.1374, ..., 0.1105, -0.0505, 0.0313], + [-0.0740, 0.0982, -0.0711, ..., 0.0406, -0.0516, -0.0923], + [-0.0411, 0.0850, -0.0990, ..., -0.0245, 0.0273, 0.1166], + ..., + [-0.0044, 0.0541, -0.0043, ..., -0.0144, 0.0301, -0.0599], + [ 0.0812, -0.0964, 0.0379, ..., 0.0091, 0.0092, -0.0417], + [-0.0150, -0.0698, 0.0721, ..., -0.0699, 0.0249, -0.0183]], + device='cuda:0'), grad: tensor([[ 1.0818e-05, 4.1723e-06, 4.0817e-04, ..., 6.0014e-06, + 1.3256e-04, 1.9920e-04], + [ 1.4579e-04, 2.0862e-04, 4.7278e-04, ..., 6.1132e-06, + 1.7226e-04, 2.5773e-04], + [-5.9462e-04, -1.0023e-03, 3.6311e-04, ..., 5.8860e-06, + 2.9281e-05, -2.5368e-04], + ..., + [ 5.1826e-05, 8.2552e-05, -2.0850e-04, ..., 5.7258e-06, + 1.6046e-04, 1.8930e-04], + [ 4.6790e-05, 8.5056e-05, 4.8327e-04, ..., -9.8467e-05, + 3.3903e-04, 3.1567e-04], + [ 4.6529e-06, 1.1005e-05, -1.2604e-02, ..., 1.0282e-05, + -7.6599e-03, -1.8930e-03]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0163, -0.0135, -0.0110, 0.0058, -0.0163, -0.0172, 0.0159, -0.0021, + 0.0199, -0.0148], device='cuda:0'), grad: tensor([ 0.0132, -0.0163, 0.0107, 0.0200, 0.0294, 0.0123, -0.0107, -0.0174, + -0.0105, -0.0307], device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 217.00, cls_loss 0.5070 cls_loss_mapping 0.0108 cls_loss_causal 0.4748 re_mapping 0.0079 re_causal 0.0190 /// teacc 98.90 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.0569, -0.0833, -0.1375, ..., 0.1101, -0.0508, 0.0315], + [-0.0746, 0.0987, -0.0705, ..., 0.0402, -0.0529, -0.0951], + [-0.0404, 0.0845, -0.0977, ..., -0.0240, 0.0279, 0.1166], + ..., + [-0.0049, 0.0545, -0.0039, ..., -0.0143, 0.0294, -0.0602], + [ 0.0820, -0.0975, 0.0378, ..., 0.0089, 0.0096, -0.0425], + [-0.0158, -0.0701, 0.0720, ..., -0.0709, 0.0248, -0.0189]], + device='cuda:0'), grad: tensor([[ 1.7002e-05, 4.6110e-04, -1.1490e-02, ..., 8.2195e-05, + -2.8534e-03, -1.6737e-03], + [ 1.1757e-05, 3.1090e-04, 2.3484e-04, ..., -1.9150e-03, + -1.1808e-04, -6.5279e-04], + [ 1.9252e-05, 1.0977e-03, 9.9897e-05, ..., 2.8658e-04, + 4.1890e-04, 1.7852e-05], + ..., + [ 4.5508e-05, -3.9215e-03, -6.8817e-03, ..., -5.2810e-05, + -4.3182e-03, 6.7294e-05], + [ 1.6153e-05, 2.7466e-04, 3.9864e-03, ..., 2.2542e-04, + 3.2406e-03, 3.7026e-04], + [-1.0765e-02, 5.4932e-04, -1.0605e-02, ..., 1.7822e-04, + -4.7874e-03, 1.6522e-04]], device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0159, -0.0138, -0.0119, 0.0070, -0.0168, -0.0160, 0.0160, -0.0020, + 0.0206, -0.0159], device='cuda:0'), grad: tensor([-0.0153, -0.0173, 0.0171, 0.0079, 0.0286, 0.0119, 0.0154, -0.0375, + 0.0144, -0.0253], device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 216.79, cls_loss 0.5254 cls_loss_mapping 0.0063 cls_loss_causal 0.5000 re_mapping 0.0078 re_causal 0.0195 /// teacc 98.70 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.0573, -0.0848, -0.1375, ..., 0.1108, -0.0512, 0.0307], + [-0.0745, 0.0982, -0.0711, ..., 0.0399, -0.0546, -0.0955], + [-0.0390, 0.0845, -0.0992, ..., -0.0247, 0.0278, 0.1174], + ..., + [-0.0045, 0.0551, -0.0028, ..., -0.0139, 0.0299, -0.0599], + [ 0.0824, -0.0976, 0.0388, ..., 0.0094, 0.0101, -0.0417], + [-0.0155, -0.0702, 0.0712, ..., -0.0714, 0.0241, -0.0194]], + device='cuda:0'), grad: tensor([[ 3.8967e-03, 2.3127e-04, -1.3866e-03, ..., 4.0084e-05, + 5.3120e-04, 2.1610e-03], + [ 4.6976e-06, -3.8586e-03, -1.8024e-04, ..., -8.2111e-04, + -4.5633e-04, 1.8215e-04], + [-3.9864e-04, 3.5834e-04, -9.9182e-04, ..., 2.4647e-05, + -1.9064e-03, -2.6417e-03], + ..., + [ 2.8915e-03, 5.5790e-04, 1.2276e-02, ..., 8.8394e-05, + 9.3536e-03, 4.6062e-04], + [ 3.0446e-04, 5.5408e-04, 3.3512e-03, ..., 2.2030e-04, + 2.0351e-03, 1.2369e-03], + [ 1.0443e-03, 3.2854e-04, -6.4964e-03, ..., -1.4514e-05, + -1.5049e-03, -1.9264e-03]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0158, -0.0146, -0.0124, 0.0062, -0.0171, -0.0155, 0.0162, -0.0017, + 0.0217, -0.0154], device='cuda:0'), grad: tensor([ 0.0143, -0.0277, -0.0019, -0.0077, -0.0038, 0.0001, -0.0202, 0.0430, + 0.0300, -0.0261], device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 216.99, cls_loss 0.5227 cls_loss_mapping 0.0074 cls_loss_causal 0.4929 re_mapping 0.0073 re_causal 0.0176 /// teacc 98.68 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.0566, -0.0846, -0.1355, ..., 0.1116, -0.0498, 0.0335], + [-0.0756, 0.1000, -0.0711, ..., 0.0412, -0.0549, -0.0937], + [-0.0404, 0.0845, -0.1005, ..., -0.0245, 0.0266, 0.1155], + ..., + [-0.0044, 0.0546, -0.0030, ..., -0.0149, 0.0295, -0.0611], + [ 0.0823, -0.0981, 0.0381, ..., 0.0082, 0.0093, -0.0418], + [-0.0166, -0.0703, 0.0707, ..., -0.0719, 0.0238, -0.0204]], + device='cuda:0'), grad: tensor([[ 1.7288e-02, -3.5572e-04, 2.7405e-02, ..., 5.8174e-04, + 2.0187e-02, 1.6159e-02], + [ 1.3709e-04, 5.8621e-05, 4.6754e-04, ..., 7.6103e-04, + 3.5119e-04, 7.3099e-04], + [ 3.3937e-06, 3.4869e-05, 4.1652e-04, ..., 6.4135e-04, + 2.9969e-04, 6.9380e-04], + ..., + [ 2.0242e-04, 1.0490e-04, 1.4982e-03, ..., 9.1839e-04, + 7.1049e-04, 1.0319e-03], + [-1.7532e-02, 3.6275e-07, -2.7710e-02, ..., -5.9547e-03, + -2.2964e-02, -1.9669e-02], + [-1.3361e-03, 1.9953e-05, -4.3106e-03, ..., 7.7009e-04, + -1.3895e-03, 9.5320e-04]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0176, -0.0136, -0.0136, 0.0076, -0.0170, -0.0169, 0.0149, -0.0020, + 0.0216, -0.0158], device='cuda:0'), grad: tensor([ 0.0440, 0.0099, 0.0085, -0.0231, 0.0235, 0.0096, -0.0079, 0.0139, + -0.0802, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 216.84, cls_loss 0.5290 cls_loss_mapping 0.0051 cls_loss_causal 0.5006 re_mapping 0.0073 re_causal 0.0187 /// teacc 98.63 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.0563, -0.0839, -0.1365, ..., 0.1124, -0.0503, 0.0326], + [-0.0745, 0.1000, -0.0697, ..., 0.0411, -0.0541, -0.0945], + [-0.0404, 0.0847, -0.0997, ..., -0.0241, 0.0270, 0.1149], + ..., + [-0.0054, 0.0539, -0.0042, ..., -0.0134, 0.0294, -0.0607], + [ 0.0840, -0.0970, 0.0374, ..., 0.0068, 0.0083, -0.0391], + [-0.0179, -0.0699, 0.0715, ..., -0.0728, 0.0252, -0.0204]], + device='cuda:0'), grad: tensor([[-2.6202e-04, -3.3522e-04, -5.0201e-03, ..., -1.8473e-03, + -3.8986e-03, -1.3208e-03], + [ 2.4962e-04, -1.5271e-04, -1.1694e-04, ..., 1.3189e-03, + -9.5320e-04, -9.3794e-04], + [-1.1415e-03, 3.1185e-04, 4.6806e-03, ..., 5.9366e-04, + 3.9005e-03, -2.5368e-03], + ..., + [ 1.0109e-04, 1.3571e-03, 1.2569e-03, ..., -3.7918e-03, + 9.6560e-04, -7.7486e-05], + [-5.2547e-04, -5.5456e-04, 4.5657e-04, ..., 6.0797e-04, + 6.1941e-04, 6.3848e-04], + [ 5.1498e-04, 1.1358e-03, 2.3193e-03, ..., 1.4610e-03, + 2.0084e-03, 1.2016e-03]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0171, -0.0122, -0.0137, 0.0072, -0.0171, -0.0163, 0.0149, -0.0016, + 0.0204, -0.0157], device='cuda:0'), grad: tensor([-0.0091, 0.0039, -0.0124, 0.0124, 0.0096, 0.0196, -0.0121, 0.0045, + -0.0189, 0.0024], device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 216.40, cls_loss 0.5100 cls_loss_mapping 0.0047 cls_loss_causal 0.4838 re_mapping 0.0070 re_causal 0.0184 /// teacc 98.75 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.0567, -0.0847, -0.1363, ..., 0.1127, -0.0501, 0.0329], + [-0.0749, 0.0998, -0.0709, ..., 0.0421, -0.0542, -0.0962], + [-0.0377, 0.0849, -0.1004, ..., -0.0245, 0.0275, 0.1157], + ..., + [-0.0052, 0.0545, -0.0045, ..., -0.0138, 0.0288, -0.0606], + [ 0.0831, -0.0976, 0.0375, ..., 0.0066, 0.0084, -0.0407], + [-0.0187, -0.0701, 0.0732, ..., -0.0731, 0.0259, -0.0203]], + device='cuda:0'), grad: tensor([[ 1.3925e-05, 1.4521e-05, 4.1223e-04, ..., 1.3685e-03, + 2.8777e-04, 5.2595e-04], + [ 2.6170e-07, 3.7700e-05, 6.2287e-05, ..., 2.4796e-04, + 4.0203e-05, 5.0449e-04], + [-1.4544e-05, -9.5558e-04, -1.9989e-03, ..., 2.2304e-04, + -6.6490e-03, -5.1117e-03], + ..., + [ 9.0990e-07, 1.7285e-05, 1.1283e-04, ..., -7.6890e-05, + 3.2878e-04, -3.1052e-03], + [ 2.0862e-06, 6.1083e-04, 4.5419e-04, ..., 2.8095e-03, + -1.8328e-05, 2.0676e-03], + [ 6.8871e-07, 1.2517e-05, 1.8911e-03, ..., 3.9387e-04, + 6.0883e-03, 3.0327e-03]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0164, -0.0125, -0.0122, 0.0076, -0.0173, -0.0161, 0.0145, -0.0028, + 0.0199, -0.0145], device='cuda:0'), grad: tensor([ 0.0166, -0.0156, -0.0169, -0.0344, 0.0130, 0.0289, 0.0016, -0.0181, + 0.0184, 0.0065], device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 216.65, cls_loss 0.5051 cls_loss_mapping 0.0056 cls_loss_causal 0.4745 re_mapping 0.0073 re_causal 0.0182 /// teacc 98.82 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.0559, -0.0852, -0.1364, ..., 0.1132, -0.0495, 0.0333], + [-0.0755, 0.0998, -0.0713, ..., 0.0421, -0.0547, -0.0945], + [-0.0372, 0.0848, -0.1007, ..., -0.0244, 0.0272, 0.1162], + ..., + [-0.0049, 0.0542, -0.0034, ..., -0.0135, 0.0299, -0.0606], + [ 0.0826, -0.0969, 0.0372, ..., 0.0074, 0.0075, -0.0407], + [-0.0189, -0.0700, 0.0729, ..., -0.0738, 0.0265, -0.0212]], + device='cuda:0'), grad: tensor([[ 3.2829e-07, 1.3165e-05, 1.0633e-03, ..., -3.2288e-02, + 1.0672e-03, 4.6182e-04], + [ 2.0492e-04, 4.5204e-04, 1.1787e-03, ..., -1.5604e-04, + 9.5463e-04, 4.0793e-04], + [ 2.2464e-06, 2.5585e-05, 7.3290e-04, ..., 2.4939e-04, + 2.6169e-03, 2.5120e-03], + ..., + [-7.1955e-04, -3.1643e-03, -4.1046e-03, ..., -1.7385e-03, + -1.9569e-03, -7.6115e-05], + [-1.9148e-05, 5.3644e-04, 3.8261e-03, ..., 9.2030e-04, + 4.2381e-03, 5.0497e-04], + [ 3.6168e-04, 1.6108e-03, -2.4509e-03, ..., 9.6655e-04, + -6.1646e-03, 3.4451e-04]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0164, -0.0122, -0.0124, 0.0078, -0.0177, -0.0165, 0.0140, -0.0023, + 0.0201, -0.0144], device='cuda:0'), grad: tensor([-0.0189, -0.0089, 0.0292, -0.0020, -0.0121, -0.0132, 0.0137, -0.0229, + 0.0299, 0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 216.64, cls_loss 0.5056 cls_loss_mapping 0.0041 cls_loss_causal 0.4734 re_mapping 0.0070 re_causal 0.0180 /// teacc 98.82 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.0571, -0.0856, -0.1372, ..., 0.1136, -0.0509, 0.0326], + [-0.0756, 0.0998, -0.0719, ..., 0.0422, -0.0555, -0.0948], + [-0.0379, 0.0846, -0.1020, ..., -0.0240, 0.0262, 0.1158], + ..., + [-0.0047, 0.0556, -0.0033, ..., -0.0117, 0.0302, -0.0600], + [ 0.0836, -0.0972, 0.0382, ..., 0.0062, 0.0087, -0.0407], + [-0.0189, -0.0701, 0.0728, ..., -0.0727, 0.0264, -0.0201]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0009, -0.0009, ..., 0.0002, -0.0011, -0.0013], + [ 0.0002, 0.0015, 0.0026, ..., 0.0001, 0.0003, 0.0010], + [ 0.0020, 0.0042, 0.0045, ..., 0.0003, 0.0047, 0.0014], + ..., + [-0.0031, -0.0134, -0.0128, ..., -0.0007, -0.0148, -0.0094], + [ 0.0013, 0.0013, 0.0010, ..., 0.0001, 0.0011, 0.0015], + [ 0.0006, 0.0064, 0.0096, ..., 0.0001, 0.0071, 0.0049]], + device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0167, -0.0122, -0.0130, 0.0077, -0.0176, -0.0165, 0.0146, -0.0020, + 0.0202, -0.0149], device='cuda:0'), grad: tensor([-0.0028, -0.0464, 0.0339, 0.0134, -0.0027, -0.0104, -0.0279, -0.0247, + 0.0295, 0.0380], device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 216.56, cls_loss 0.4991 cls_loss_mapping 0.0062 cls_loss_causal 0.4791 re_mapping 0.0073 re_causal 0.0189 /// teacc 98.75 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.0576, -0.0847, -0.1390, ..., 0.1127, -0.0516, 0.0324], + [-0.0752, 0.0996, -0.0709, ..., 0.0429, -0.0539, -0.0957], + [-0.0395, 0.0848, -0.1020, ..., -0.0233, 0.0263, 0.1168], + ..., + [-0.0037, 0.0562, -0.0034, ..., -0.0127, 0.0303, -0.0593], + [ 0.0836, -0.0981, 0.0387, ..., 0.0065, 0.0089, -0.0414], + [-0.0181, -0.0712, 0.0728, ..., -0.0741, 0.0262, -0.0212]], + device='cuda:0'), grad: tensor([[ 3.7813e-04, 2.8655e-05, -3.5167e-05, ..., 3.4481e-05, + -5.0831e-04, -3.4794e-06], + [ 1.4037e-05, 1.5461e-04, 7.6342e-04, ..., 1.1611e-04, + 7.0047e-04, 2.4214e-08], + [ 2.1732e-04, 6.1274e-05, 3.7813e-04, ..., 8.8096e-05, + 3.5834e-04, -6.0536e-06], + ..., + [ 1.6257e-05, -5.4884e-04, -2.1706e-03, ..., -9.2936e-04, + -9.8801e-04, 4.7032e-08], + [ 4.4870e-04, 3.5077e-05, 2.9325e-04, ..., 4.6164e-05, + 3.2568e-04, -5.9744e-07], + [ 1.9169e-04, -9.2363e-04, -5.4741e-03, ..., -3.7432e-05, + -4.8676e-03, 6.7288e-07]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0163, -0.0117, -0.0139, 0.0083, -0.0177, -0.0186, 0.0149, -0.0013, + 0.0204, -0.0140], device='cuda:0'), grad: tensor([-0.0482, 0.0177, 0.0133, 0.0139, 0.0281, 0.0124, -0.0214, -0.0068, + 0.0130, -0.0220], device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 216.54, cls_loss 0.5031 cls_loss_mapping 0.0047 cls_loss_causal 0.4693 re_mapping 0.0069 re_causal 0.0167 /// teacc 98.63 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.0588, -0.0858, -0.1392, ..., 0.1133, -0.0518, 0.0315], + [-0.0756, 0.0998, -0.0714, ..., 0.0431, -0.0544, -0.0962], + [-0.0401, 0.0860, -0.1016, ..., -0.0216, 0.0271, 0.1181], + ..., + [-0.0042, 0.0555, -0.0021, ..., -0.0129, 0.0299, -0.0601], + [ 0.0832, -0.0980, 0.0390, ..., 0.0063, 0.0091, -0.0403], + [-0.0165, -0.0718, 0.0726, ..., -0.0757, 0.0263, -0.0219]], + device='cuda:0'), grad: tensor([[ 9.1732e-05, 6.3801e-04, 1.6713e-04, ..., 4.6730e-04, + 1.1683e-03, 6.4802e-04], + [ 9.7275e-04, 1.1784e-04, -1.5717e-03, ..., -2.7537e-04, + -5.1651e-03, 4.5848e-04], + [ 1.6365e-03, 3.7174e-03, 4.1175e-04, ..., 1.2407e-03, + 1.2951e-03, 3.3207e-03], + ..., + [ 5.1928e-04, -3.0613e-03, 1.1196e-03, ..., -4.0703e-03, + 8.9121e-04, 4.1771e-04], + [ 7.4196e-04, 9.8991e-04, 5.3835e-04, ..., 4.9400e-04, + 1.0948e-03, 1.0481e-03], + [ 1.0958e-03, 6.1655e-04, 1.7538e-03, ..., 4.6563e-04, + -3.0651e-03, -1.2684e-03]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0153, -0.0120, -0.0128, 0.0090, -0.0182, -0.0186, 0.0150, -0.0008, + 0.0200, -0.0141], device='cuda:0'), grad: tensor([-0.0131, -0.0044, 0.0085, 0.0244, -0.0432, 0.0157, 0.0140, -0.0085, + 0.0203, -0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 216.82, cls_loss 0.5172 cls_loss_mapping 0.0048 cls_loss_causal 0.4888 re_mapping 0.0073 re_causal 0.0187 /// teacc 98.70 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.0594, -0.0869, -0.1381, ..., 0.1134, -0.0504, 0.0315], + [-0.0755, 0.0992, -0.0713, ..., 0.0437, -0.0550, -0.0970], + [-0.0411, 0.0849, -0.1025, ..., -0.0206, 0.0268, 0.1187], + ..., + [-0.0042, 0.0574, -0.0030, ..., -0.0121, 0.0296, -0.0610], + [ 0.0835, -0.0983, 0.0398, ..., 0.0053, 0.0092, -0.0410], + [-0.0171, -0.0727, 0.0712, ..., -0.0764, 0.0258, -0.0217]], + device='cuda:0'), grad: tensor([[ 3.8236e-05, -6.7329e-04, 7.6771e-05, ..., -1.1177e-03, + -9.4175e-04, -5.3253e-03], + [-1.2839e-04, -7.1049e-05, -6.1333e-05, ..., -6.4325e-04, + -1.7252e-03, -3.5610e-03], + [ 6.0648e-05, 9.4533e-05, 8.3625e-05, ..., 4.3035e-04, + 7.8058e-04, 2.2907e-03], + ..., + [-1.3399e-04, -5.8383e-05, 1.1188e-04, ..., 1.5652e-04, + 2.1124e-04, 3.8457e-04], + [ 1.9569e-03, 1.6749e-04, -3.5024e-04, ..., 2.7394e-04, + 6.8285e-06, 1.0767e-03], + [ 1.3733e-04, 1.3387e-04, 1.7273e-04, ..., 1.3018e-04, + 3.1161e-04, 5.1117e-04]], device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0159, -0.0121, -0.0128, 0.0078, -0.0179, -0.0183, 0.0150, -0.0004, + 0.0196, -0.0140], device='cuda:0'), grad: tensor([-0.0174, -0.0253, 0.0196, 0.0234, 0.0107, -0.0002, 0.0151, -0.0220, + -0.0145, 0.0107], device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 216.75, cls_loss 0.5022 cls_loss_mapping 0.0052 cls_loss_causal 0.4734 re_mapping 0.0069 re_causal 0.0181 /// teacc 98.66 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.0602, -0.0886, -0.1370, ..., 0.1144, -0.0498, 0.0325], + [-0.0760, 0.1002, -0.0727, ..., 0.0438, -0.0559, -0.0955], + [-0.0410, 0.0841, -0.1036, ..., -0.0205, 0.0270, 0.1191], + ..., + [-0.0050, 0.0577, -0.0023, ..., -0.0121, 0.0311, -0.0614], + [ 0.0836, -0.0985, 0.0384, ..., 0.0038, 0.0079, -0.0409], + [-0.0170, -0.0726, 0.0720, ..., -0.0788, 0.0263, -0.0228]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0006, 0.0014, ..., 0.0022, 0.0023, 0.0017], + [-0.0004, -0.0023, 0.0007, ..., -0.0022, -0.0013, -0.0016], + [ 0.0007, 0.0010, 0.0004, ..., 0.0010, 0.0010, 0.0006], + ..., + [ 0.0004, 0.0008, 0.0021, ..., 0.0007, 0.0023, 0.0002], + [-0.0043, 0.0007, -0.0103, ..., 0.0008, -0.0101, 0.0004], + [ 0.0011, 0.0009, 0.0109, ..., 0.0012, 0.0089, 0.0008]], + device='cuda:0') +Epoch 253, bias, value: tensor([ 1.5789e-02, -1.0751e-02, -1.3441e-02, 7.6038e-03, -1.7726e-02, + -1.9254e-02, 1.4453e-02, 6.9520e-05, 2.0338e-02, -1.4395e-02], + device='cuda:0'), grad: tensor([-0.0053, -0.0089, 0.0131, 0.0136, -0.0611, 0.0163, -0.0061, 0.0231, + -0.0209, 0.0364], device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 216.64, cls_loss 0.4951 cls_loss_mapping 0.0047 cls_loss_causal 0.4659 re_mapping 0.0067 re_causal 0.0174 /// teacc 98.80 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.0605, -0.0878, -0.1361, ..., 0.1153, -0.0502, 0.0321], + [-0.0772, 0.0996, -0.0748, ..., 0.0437, -0.0567, -0.0958], + [-0.0419, 0.0842, -0.1035, ..., -0.0190, 0.0279, 0.1201], + ..., + [-0.0040, 0.0576, -0.0023, ..., -0.0119, 0.0307, -0.0613], + [ 0.0841, -0.0986, 0.0388, ..., 0.0028, 0.0088, -0.0407], + [-0.0167, -0.0714, 0.0719, ..., -0.0778, 0.0260, -0.0218]], + device='cuda:0'), grad: tensor([[-1.9436e-03, 9.8161e-07, 3.3760e-04, ..., -6.2675e-03, + -1.0663e-04, -4.7207e-04], + [ 6.5565e-04, -2.3469e-05, 6.6233e-04, ..., 1.5936e-03, + 1.1690e-05, 2.1219e-04], + [ 4.1914e-04, 2.0400e-05, 3.8910e-04, ..., 9.3985e-04, + 9.0301e-05, 3.5429e-04], + ..., + [ 2.5439e-04, -8.1837e-05, 5.8889e-04, ..., 9.8515e-04, + -2.9758e-05, 1.1408e-04], + [-1.4334e-03, 1.3091e-05, -4.5357e-03, ..., -1.5628e-04, + 1.1921e-04, -7.4863e-04], + [ 3.5048e-04, 3.5316e-05, 5.9366e-04, ..., 8.4782e-04, + -5.4300e-05, 1.8048e-04]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0155, -0.0099, -0.0136, 0.0081, -0.0172, -0.0197, 0.0134, -0.0004, + 0.0196, -0.0132], device='cuda:0'), grad: tensor([-0.0103, 0.0266, -0.0450, 0.0083, -0.0118, 0.0232, 0.0275, -0.0117, + -0.0235, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 216.74, cls_loss 0.5312 cls_loss_mapping 0.0046 cls_loss_causal 0.5034 re_mapping 0.0068 re_causal 0.0187 /// teacc 98.63 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.0617, -0.0872, -0.1370, ..., 0.1144, -0.0509, 0.0321], + [-0.0766, 0.0988, -0.0740, ..., 0.0439, -0.0557, -0.0960], + [-0.0402, 0.0850, -0.1015, ..., -0.0170, 0.0292, 0.1198], + ..., + [-0.0042, 0.0579, -0.0026, ..., -0.0137, 0.0298, -0.0626], + [ 0.0827, -0.0989, 0.0383, ..., 0.0025, 0.0092, -0.0389], + [-0.0158, -0.0725, 0.0724, ..., -0.0786, 0.0265, -0.0215]], + device='cuda:0'), grad: tensor([[ 7.8321e-05, 1.2755e-05, 8.7547e-04, ..., -9.7046e-03, + -2.4948e-03, -4.2114e-03], + [ 1.0884e-04, 7.6666e-06, -1.5602e-03, ..., 1.7416e-04, + 3.0446e-04, 1.7226e-04], + [ 1.7214e-04, 3.8362e-04, 9.0265e-04, ..., 4.4212e-03, + 2.1114e-03, 1.3475e-03], + ..., + [ 1.5450e-04, 4.2289e-05, 5.8174e-04, ..., 5.3704e-05, + 5.0211e-04, 2.2411e-04], + [-1.5411e-03, -6.0129e-04, -4.4174e-03, ..., -4.0779e-03, + -1.8291e-03, -1.9627e-03], + [-1.0624e-03, 1.0237e-05, 4.0317e-04, ..., 1.6677e-04, + 3.5400e-03, 4.1413e-04]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0149, -0.0107, -0.0124, 0.0072, -0.0172, -0.0191, 0.0146, -0.0007, + 0.0184, -0.0123], device='cuda:0'), grad: tensor([-0.0092, -0.0223, 0.0170, -0.0348, 0.0197, 0.0171, 0.0014, 0.0078, + -0.0135, 0.0169], device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 216.73, cls_loss 0.4915 cls_loss_mapping 0.0050 cls_loss_causal 0.4612 re_mapping 0.0071 re_causal 0.0187 /// teacc 98.54 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.0624, -0.0861, -0.1378, ..., 0.1142, -0.0512, 0.0322], + [-0.0762, 0.0987, -0.0743, ..., 0.0435, -0.0568, -0.0973], + [-0.0410, 0.0854, -0.1017, ..., -0.0178, 0.0302, 0.1192], + ..., + [-0.0035, 0.0564, -0.0021, ..., -0.0137, 0.0294, -0.0635], + [ 0.0831, -0.0977, 0.0389, ..., 0.0038, 0.0092, -0.0381], + [-0.0165, -0.0715, 0.0726, ..., -0.0792, 0.0273, -0.0212]], + device='cuda:0'), grad: tensor([[ 5.6887e-04, 1.0997e-05, -6.2227e-05, ..., 7.7486e-04, + -1.4126e-04, -6.7377e-04], + [ 1.7881e-04, -1.5764e-03, 1.0121e-04, ..., -1.7487e-02, + 1.5318e-04, -6.7139e-03], + [ 3.4523e-04, 6.5029e-05, 9.2328e-05, ..., 8.6308e-04, + 2.0206e-04, 3.6812e-04], + ..., + [ 2.9564e-03, 1.2922e-03, 3.6979e-04, ..., 1.7366e-03, + 2.8086e-04, 8.0287e-05], + [ 1.1120e-03, 6.7234e-05, -4.4918e-04, ..., -5.5428e-03, + -1.2808e-03, -5.3644e-04], + [ 3.5000e-03, 7.0632e-05, -8.3256e-04, ..., 8.1015e-04, + -2.1696e-04, 3.0565e-04]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0133, -0.0108, -0.0130, 0.0071, -0.0161, -0.0188, 0.0150, -0.0010, + 0.0194, -0.0125], device='cuda:0'), grad: tensor([ 0.0043, -0.0376, 0.0080, -0.0179, 0.0085, 0.0092, 0.0375, 0.0185, + -0.0154, -0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 216.79, cls_loss 0.4791 cls_loss_mapping 0.0050 cls_loss_causal 0.4569 re_mapping 0.0070 re_causal 0.0180 /// teacc 98.72 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.0612, -0.0866, -0.1382, ..., 0.1141, -0.0515, 0.0333], + [-0.0763, 0.1005, -0.0746, ..., 0.0457, -0.0564, -0.0939], + [-0.0403, 0.0849, -0.1033, ..., -0.0190, 0.0293, 0.1195], + ..., + [-0.0050, 0.0571, -0.0018, ..., -0.0136, 0.0285, -0.0643], + [ 0.0829, -0.0987, 0.0409, ..., 0.0050, 0.0109, -0.0380], + [-0.0171, -0.0728, 0.0716, ..., -0.0791, 0.0269, -0.0233]], + device='cuda:0'), grad: tensor([[ 6.3591e-06, 1.7846e-04, 5.9414e-04, ..., 4.2558e-04, + 6.3086e-04, 3.5357e-04], + [ 2.3603e-04, 2.3234e-04, -5.8085e-05, ..., 2.8849e-04, + 4.6706e-04, 2.3627e-04], + [ 2.4304e-05, 2.2662e-04, 9.8324e-04, ..., 1.0700e-03, + 1.1644e-03, 7.5674e-04], + ..., + [ 7.0274e-05, -2.1229e-03, 1.4104e-05, ..., -1.6232e-03, + -1.4143e-03, -1.6441e-03], + [-9.7847e-04, -1.6344e-04, -3.9411e-04, ..., -5.0507e-03, + -1.1415e-03, -2.6226e-03], + [ 1.0860e-04, 3.1424e-04, 1.8425e-03, ..., 5.8699e-04, + 1.9445e-03, 6.5327e-04]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0136, -0.0108, -0.0135, 0.0068, -0.0170, -0.0182, 0.0147, -0.0002, + 0.0203, -0.0130], device='cuda:0'), grad: tensor([ 0.0133, -0.0184, 0.0200, -0.0010, -0.0108, 0.0196, 0.0142, -0.0064, + -0.0534, 0.0228], device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 216.68, cls_loss 0.5290 cls_loss_mapping 0.0060 cls_loss_causal 0.5032 re_mapping 0.0070 re_causal 0.0186 /// teacc 98.84 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.0610, -0.0881, -0.1393, ..., 0.1135, -0.0518, 0.0323], + [-0.0770, 0.1006, -0.0748, ..., 0.0468, -0.0558, -0.0939], + [-0.0404, 0.0845, -0.1031, ..., -0.0189, 0.0295, 0.1198], + ..., + [-0.0043, 0.0585, -0.0028, ..., -0.0138, 0.0278, -0.0637], + [ 0.0830, -0.1007, 0.0409, ..., 0.0030, 0.0114, -0.0392], + [-0.0155, -0.0732, 0.0736, ..., -0.0801, 0.0279, -0.0251]], + device='cuda:0'), grad: tensor([[ 3.4261e-04, 5.0354e-04, 6.0171e-05, ..., 1.0529e-02, + 1.0056e-02, 8.1491e-04], + [-1.5192e-05, 1.9464e-03, 8.9034e-06, ..., 4.1885e-03, + 6.0415e-04, 2.0676e-03], + [ 3.7479e-04, -9.5129e-04, 6.6900e-04, ..., 1.7147e-03, + 1.1644e-03, 9.1171e-04], + ..., + [ 2.6867e-05, 2.9869e-03, 3.2163e-04, ..., -2.2644e-02, + -2.4490e-02, 9.6035e-04], + [-7.1373e-03, -3.8385e-04, 4.9210e-03, ..., -1.7176e-03, + 7.6752e-03, -1.3533e-03], + [ 1.0580e-04, -2.7523e-03, -8.4000e-03, ..., 1.0979e-02, + 1.3375e-04, -2.7294e-03]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0134, -0.0108, -0.0135, 0.0068, -0.0182, -0.0185, 0.0151, -0.0003, + 0.0209, -0.0124], device='cuda:0'), grad: tensor([ 0.0270, 0.0321, -0.0070, 0.0307, -0.0104, -0.0324, 0.0231, -0.0198, + -0.0190, -0.0242], device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 216.70, cls_loss 0.4695 cls_loss_mapping 0.0042 cls_loss_causal 0.4451 re_mapping 0.0071 re_causal 0.0183 /// teacc 98.75 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.0618, -0.0891, -0.1396, ..., 0.1138, -0.0505, 0.0332], + [-0.0760, 0.1003, -0.0745, ..., 0.0464, -0.0553, -0.0926], + [-0.0402, 0.0853, -0.1038, ..., -0.0199, 0.0289, 0.1195], + ..., + [-0.0049, 0.0586, -0.0027, ..., -0.0137, 0.0282, -0.0624], + [ 0.0830, -0.1002, 0.0400, ..., 0.0050, 0.0108, -0.0397], + [-0.0160, -0.0733, 0.0731, ..., -0.0807, 0.0275, -0.0246]], + device='cuda:0'), grad: tensor([[ 1.1986e-06, 4.4912e-05, 2.3804e-03, ..., 3.0220e-05, + 2.3403e-03, 3.6089e-07], + [-5.2661e-05, -7.2861e-04, -2.2304e-04, ..., -5.6601e-04, + -4.9733e-07, 2.0955e-08], + [ 2.6431e-06, 1.1760e-04, 2.8133e-04, ..., 6.8188e-05, + 2.3210e-04, 1.4491e-06], + ..., + [ 8.1658e-06, -8.0988e-06, 1.6603e-03, ..., 6.4492e-05, + 1.3218e-03, 3.4086e-07], + [ 2.0452e-06, 1.1516e-04, 1.1039e-04, ..., 1.0592e-04, + 9.0659e-05, 1.4817e-06], + [ 1.9046e-06, 5.3763e-05, 1.1683e-03, ..., 2.9564e-05, + 1.5326e-03, 4.9099e-06]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0125, -0.0104, -0.0123, 0.0064, -0.0183, -0.0183, 0.0157, -0.0002, + 0.0202, -0.0127], device='cuda:0'), grad: tensor([ 0.0171, -0.0243, -0.0201, -0.0105, 0.0089, 0.0087, 0.0108, 0.0154, + -0.0213, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 216.43, cls_loss 0.5092 cls_loss_mapping 0.0056 cls_loss_causal 0.4933 re_mapping 0.0072 re_causal 0.0192 /// teacc 98.61 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.0616, -0.0894, -0.1390, ..., 0.1150, -0.0506, 0.0323], + [-0.0757, 0.1003, -0.0754, ..., 0.0465, -0.0568, -0.0944], + [-0.0403, 0.0850, -0.1044, ..., -0.0185, 0.0282, 0.1200], + ..., + [-0.0050, 0.0587, -0.0024, ..., -0.0132, 0.0289, -0.0622], + [ 0.0829, -0.1001, 0.0409, ..., 0.0039, 0.0116, -0.0406], + [-0.0159, -0.0730, 0.0725, ..., -0.0817, 0.0271, -0.0226]], + device='cuda:0'), grad: tensor([[ 4.3035e-04, -8.7547e-04, 9.3269e-04, ..., 4.3941e-04, + 6.7377e-04, 9.8801e-04], + [ 1.9416e-05, -3.1680e-05, 5.5170e-04, ..., 4.3559e-04, + 3.0136e-04, 5.0831e-04], + [-3.3283e-03, 3.7909e-05, -4.1389e-03, ..., -3.7193e-04, + -2.2564e-03, -6.3171e-03], + ..., + [ 8.4862e-06, 1.2898e-04, 1.4595e-02, ..., -3.8482e-06, + 4.6082e-03, 2.8110e-04], + [-1.2362e-04, 3.0160e-04, 1.7345e-04, ..., -3.0708e-04, + -1.2789e-03, 8.2076e-05], + [ 4.6611e-05, 2.2638e-04, 2.6054e-03, ..., -9.4175e-04, + 5.1081e-05, -3.0174e-03]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0134, -0.0112, -0.0118, 0.0074, -0.0172, -0.0187, 0.0148, -0.0005, + 0.0195, -0.0131], device='cuda:0'), grad: tensor([-0.0107, 0.0128, -0.0505, -0.0192, -0.0165, 0.0157, 0.0395, 0.0438, + -0.0017, -0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 216.77, cls_loss 0.4951 cls_loss_mapping 0.0055 cls_loss_causal 0.4781 re_mapping 0.0068 re_causal 0.0170 /// teacc 98.85 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.0621, -0.0893, -0.1391, ..., 0.1151, -0.0509, 0.0332], + [-0.0760, 0.1007, -0.0750, ..., 0.0476, -0.0568, -0.0938], + [-0.0416, 0.0850, -0.1038, ..., -0.0185, 0.0298, 0.1200], + ..., + [-0.0055, 0.0587, -0.0024, ..., -0.0137, 0.0288, -0.0628], + [ 0.0845, -0.1005, 0.0418, ..., 0.0029, 0.0113, -0.0417], + [-0.0152, -0.0726, 0.0721, ..., -0.0807, 0.0276, -0.0230]], + device='cuda:0'), grad: tensor([[ 2.4170e-05, 7.3850e-05, -6.5947e-04, ..., 4.3750e-04, + -2.0552e-04, 1.0890e-04], + [-3.6693e-04, -3.2395e-05, 4.5109e-04, ..., -2.6245e-03, + 7.9989e-05, -1.0557e-03], + [ 6.6817e-05, -1.2159e-04, -7.1466e-05, ..., 3.4046e-04, + 6.1631e-05, 1.2517e-05], + ..., + [ 7.6175e-05, 1.2646e-03, 4.8790e-03, ..., 4.3106e-04, + 2.0909e-04, 1.2422e-04], + [-1.8525e-04, 1.6153e-05, 4.0650e-04, ..., 3.9935e-04, + 8.2731e-05, 1.4782e-04], + [ 7.9751e-05, -1.2922e-03, -4.6844e-03, ..., 3.3689e-04, + -7.6234e-05, 1.1688e-04]], device='cuda:0') +Epoch 261, bias, value: tensor([ 1.3633e-02, -1.1078e-02, -1.2826e-02, 6.7171e-03, -1.7709e-02, + -1.8478e-02, 1.4992e-02, -7.7504e-05, 2.0514e-02, -1.3138e-02], + device='cuda:0'), grad: tensor([-0.0141, -0.0089, -0.0141, 0.0192, -0.0394, 0.0151, 0.0173, -0.0029, + 0.0135, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 217.12, cls_loss 0.5009 cls_loss_mapping 0.0042 cls_loss_causal 0.4695 re_mapping 0.0072 re_causal 0.0182 /// teacc 98.79 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.0606, -0.0887, -0.1394, ..., 0.1167, -0.0515, 0.0331], + [-0.0771, 0.1005, -0.0743, ..., 0.0472, -0.0577, -0.0932], + [-0.0416, 0.0830, -0.1040, ..., -0.0192, 0.0298, 0.1201], + ..., + [-0.0063, 0.0601, -0.0011, ..., -0.0145, 0.0301, -0.0619], + [ 0.0846, -0.1007, 0.0406, ..., 0.0018, 0.0107, -0.0428], + [-0.0145, -0.0726, 0.0722, ..., -0.0799, 0.0276, -0.0238]], + device='cuda:0'), grad: tensor([[-5.7650e-04, -1.7862e-03, -2.2024e-05, ..., -2.9640e-03, + -1.3828e-03, -2.3174e-03], + [ 5.4181e-05, 3.3826e-05, -4.2096e-06, ..., 8.3828e-04, + 1.0341e-04, 5.1403e-04], + [ 6.1274e-05, 3.4475e-04, 2.1744e-04, ..., 9.0361e-04, + 1.9646e-04, -1.5812e-03], + ..., + [ 4.4286e-05, 5.8450e-06, -2.1720e-04, ..., 2.3866e-04, + -5.7459e-05, 3.6931e-04], + [ 5.5999e-05, 1.4365e-04, 1.1303e-05, ..., 6.0177e-04, + 2.0826e-04, 5.4932e-04], + [ 3.8773e-05, 1.2398e-04, 1.4520e-04, ..., -3.6926e-03, + 1.4496e-04, -6.1877e-06]], device='cuda:0') +Epoch 262, bias, value: tensor([ 0.0135, -0.0102, -0.0131, 0.0065, -0.0182, -0.0178, 0.0155, -0.0003, + 0.0197, -0.0130], device='cuda:0'), grad: tensor([-0.0185, 0.0227, -0.0186, 0.0189, -0.0154, 0.0151, 0.0169, -0.0167, + 0.0134, -0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 216.45, cls_loss 0.4824 cls_loss_mapping 0.0040 cls_loss_causal 0.4509 re_mapping 0.0070 re_causal 0.0177 /// teacc 98.73 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.0586, -0.0884, -0.1403, ..., 0.1165, -0.0523, 0.0319], + [-0.0775, 0.1009, -0.0741, ..., 0.0470, -0.0569, -0.0933], + [-0.0388, 0.0824, -0.1041, ..., -0.0182, 0.0302, 0.1202], + ..., + [-0.0068, 0.0596, -0.0014, ..., -0.0165, 0.0311, -0.0617], + [ 0.0847, -0.1008, 0.0408, ..., 0.0027, 0.0105, -0.0427], + [-0.0155, -0.0709, 0.0732, ..., -0.0798, 0.0286, -0.0239]], + device='cuda:0'), grad: tensor([[ 2.3410e-05, 3.4294e-03, 1.1692e-03, ..., 4.8447e-03, + 1.7862e-03, 3.5763e-04], + [ 1.0061e-04, 9.4080e-04, 2.9278e-03, ..., 1.7986e-03, + 6.2943e-04, 3.3569e-04], + [ 8.0243e-06, -1.5278e-03, -2.0742e-05, ..., -4.7989e-03, + -1.3390e-03, -2.2125e-03], + ..., + [ 1.1520e-06, -4.6005e-03, -8.9550e-04, ..., -5.4207e-03, + -2.2888e-03, 3.8433e-04], + [ 7.8773e-04, 1.4591e-04, -6.6566e-03, ..., 5.6887e-04, + -2.4304e-05, 1.8561e-04], + [ 1.2919e-05, 8.9312e-04, 1.1387e-03, ..., 1.5078e-03, + 7.9918e-04, -3.0446e-04]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0145, -0.0119, -0.0115, 0.0073, -0.0180, -0.0177, 0.0151, -0.0013, + 0.0188, -0.0127], device='cuda:0'), grad: tensor([ 0.0281, 0.0304, -0.0642, 0.0159, 0.0233, -0.0327, 0.0264, 0.0021, + -0.0157, -0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 216.53, cls_loss 0.4739 cls_loss_mapping 0.0054 cls_loss_causal 0.4480 re_mapping 0.0071 re_causal 0.0182 /// teacc 98.67 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.0586, -0.0892, -0.1395, ..., 0.1165, -0.0517, 0.0305], + [-0.0781, 0.1010, -0.0732, ..., 0.0466, -0.0568, -0.0924], + [-0.0393, 0.0818, -0.1056, ..., -0.0173, 0.0300, 0.1203], + ..., + [-0.0059, 0.0605, -0.0011, ..., -0.0170, 0.0307, -0.0623], + [ 0.0852, -0.1006, 0.0402, ..., 0.0015, 0.0101, -0.0407], + [-0.0157, -0.0732, 0.0723, ..., -0.0792, 0.0280, -0.0244]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5041e-04, 1.3666e-03, ..., 2.6741e-03, + 1.8377e-03, 1.4544e-03], + [ 0.0000e+00, -1.7462e-03, 1.2417e-03, ..., 6.6566e-04, + 1.1053e-03, 7.0286e-04], + [ 0.0000e+00, 6.3038e-04, 9.1934e-04, ..., 1.2999e-03, + 1.3828e-03, 1.1749e-02], + ..., + [ 1.2271e-05, 4.5681e-04, -3.5000e-03, ..., 1.1820e-04, + -2.1019e-03, 6.0129e-04], + [ 1.3970e-09, -1.2569e-03, -1.5163e-04, ..., -5.6124e-04, + -8.7118e-04, -2.8944e-04], + [ 1.8752e-04, 1.7557e-03, -3.0842e-03, ..., -6.1188e-03, + -3.5763e-03, -2.6550e-03]], device='cuda:0') +Epoch 264, bias, value: tensor([ 0.0130, -0.0115, -0.0117, 0.0079, -0.0181, -0.0171, 0.0152, -0.0013, + 0.0189, -0.0127], device='cuda:0'), grad: tensor([ 0.0209, 0.0069, 0.0232, -0.0297, 0.0075, 0.0098, 0.0193, 0.0039, + -0.0276, -0.0341], device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 216.46, cls_loss 0.5260 cls_loss_mapping 0.0059 cls_loss_causal 0.4990 re_mapping 0.0066 re_causal 0.0166 /// teacc 98.78 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.0602, -0.0897, -0.1390, ..., 0.1149, -0.0522, 0.0308], + [-0.0784, 0.0984, -0.0746, ..., 0.0450, -0.0576, -0.0927], + [-0.0406, 0.0826, -0.1052, ..., -0.0160, 0.0295, 0.1199], + ..., + [-0.0066, 0.0614, -0.0005, ..., -0.0161, 0.0318, -0.0621], + [ 0.0860, -0.1004, 0.0406, ..., 0.0035, 0.0108, -0.0394], + [-0.0163, -0.0732, 0.0717, ..., -0.0778, 0.0263, -0.0252]], + device='cuda:0'), grad: tensor([[ 1.2004e-04, 3.4237e-04, 2.8419e-04, ..., 5.8222e-04, + 4.5800e-04, 4.9639e-04], + [ 6.6795e-03, 5.0926e-03, 2.0957e-04, ..., 1.2672e-04, + 4.5824e-04, 4.0531e-04], + [-5.5265e-04, -6.6042e-04, 4.3124e-05, ..., 1.2035e-03, + 2.3305e-04, -1.4343e-03], + ..., + [-6.4468e-04, 2.0587e-04, 4.2343e-03, ..., 4.1924e-03, + 6.7749e-03, 8.3351e-04], + [ 5.6839e-04, 2.7180e-03, 1.0662e-03, ..., 2.1133e-03, + 1.8482e-03, 3.0270e-03], + [ 2.4557e-04, 9.3269e-04, 6.3658e-04, ..., 1.1473e-03, + 1.2789e-03, 4.8971e-04]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0130, -0.0126, -0.0111, 0.0072, -0.0173, -0.0163, 0.0148, -0.0017, + 0.0190, -0.0124], device='cuda:0'), grad: tensor([ 0.0082, -0.0030, 0.0034, -0.0150, -0.0220, -0.0197, -0.0102, 0.0244, + 0.0229, 0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 216.73, cls_loss 0.4883 cls_loss_mapping 0.0047 cls_loss_causal 0.4646 re_mapping 0.0079 re_causal 0.0200 /// teacc 98.79 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.0608, -0.0881, -0.1390, ..., 0.1151, -0.0518, 0.0308], + [-0.0780, 0.0996, -0.0732, ..., 0.0436, -0.0585, -0.0919], + [-0.0405, 0.0826, -0.1036, ..., -0.0149, 0.0298, 0.1194], + ..., + [-0.0070, 0.0600, -0.0012, ..., -0.0169, 0.0312, -0.0628], + [ 0.0858, -0.1003, 0.0412, ..., 0.0037, 0.0110, -0.0401], + [-0.0158, -0.0723, 0.0719, ..., -0.0783, 0.0269, -0.0243]], + device='cuda:0'), grad: tensor([[ 2.1886e-08, -2.4527e-05, -3.6931e-04, ..., -1.8847e-04, + 3.1519e-04, -5.3108e-05], + [ 6.2523e-03, 5.2118e-04, 3.1328e-04, ..., -7.0229e-03, + 1.2474e-02, 2.3749e-08], + [ 4.1211e-07, 1.1616e-03, 3.0518e-04, ..., 3.1242e-03, + 9.8133e-04, 9.6038e-06], + ..., + [ 1.1045e-06, 4.3869e-03, -4.3702e-04, ..., 7.1945e-03, + 3.7909e-04, 1.1036e-07], + [-2.9549e-05, 5.2595e-04, -8.7595e-04, ..., -5.3749e-03, + -7.6637e-03, 5.0152e-07], + [ 3.6150e-05, -3.0422e-03, 6.2704e-04, ..., -4.4823e-03, + 1.6775e-03, 2.1104e-06]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0138, -0.0106, -0.0120, 0.0072, -0.0180, -0.0172, 0.0143, -0.0021, + 0.0182, -0.0109], device='cuda:0'), grad: tensor([-0.0064, 0.0310, 0.0221, 0.0011, -0.0178, -0.0055, 0.0195, 0.0022, + -0.0068, -0.0394], device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 216.54, cls_loss 0.5154 cls_loss_mapping 0.0056 cls_loss_causal 0.4873 re_mapping 0.0071 re_causal 0.0178 /// teacc 98.81 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.0607, -0.0878, -0.1399, ..., 0.1153, -0.0526, 0.0309], + [-0.0789, 0.0992, -0.0734, ..., 0.0439, -0.0606, -0.0929], + [-0.0403, 0.0820, -0.1034, ..., -0.0153, 0.0302, 0.1194], + ..., + [-0.0059, 0.0608, -0.0005, ..., -0.0160, 0.0322, -0.0629], + [ 0.0877, -0.1004, 0.0407, ..., 0.0039, 0.0102, -0.0399], + [-0.0162, -0.0739, 0.0722, ..., -0.0773, 0.0271, -0.0240]], + device='cuda:0'), grad: tensor([[ 9.5987e-04, 1.1735e-07, -4.2200e-04, ..., 2.1744e-04, + -2.1667e-03, 8.5592e-05], + [ 3.1710e-05, 5.5321e-06, 1.1392e-05, ..., 9.2462e-06, + 4.0680e-05, 1.5333e-05], + [ 4.6641e-05, 6.3837e-05, 1.5289e-05, ..., 3.2365e-05, + 1.5354e-04, 7.9691e-05], + ..., + [ 9.6321e-05, -8.9169e-05, 1.1146e-04, ..., 1.1735e-05, + 1.6034e-04, 1.9580e-05], + [ 5.6219e-04, -3.3975e-05, 3.7956e-04, ..., 1.4365e-05, + 8.3065e-04, 8.9824e-05], + [ 1.0052e-03, 7.8529e-06, -1.4153e-03, ..., 1.1611e-04, + -1.0681e-03, 2.0459e-05]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0138, -0.0107, -0.0121, 0.0084, -0.0193, -0.0175, 0.0146, -0.0019, + 0.0180, -0.0108], device='cuda:0'), grad: tensor([-0.0283, 0.0103, 0.0087, 0.0173, 0.0050, -0.0092, -0.0075, 0.0077, + 0.0144, -0.0182], device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 216.56, cls_loss 0.5042 cls_loss_mapping 0.0057 cls_loss_causal 0.4798 re_mapping 0.0069 re_causal 0.0166 /// teacc 98.78 lr 0.00010000 +Epoch 268, weight, value: tensor([[-6.0039e-02, -8.8452e-02, -1.3959e-01, ..., 1.1533e-01, + -5.1627e-02, 3.0010e-02], + [-7.9783e-02, 9.9850e-02, -7.3150e-02, ..., 4.3850e-02, + -5.9882e-02, -9.2902e-02], + [-4.0005e-02, 8.1687e-02, -1.0496e-01, ..., -1.5609e-02, + 2.9863e-02, 1.2048e-01], + ..., + [-4.8130e-03, 6.0823e-02, -6.2102e-05, ..., -1.6696e-02, + 3.2087e-02, -6.3894e-02], + [ 8.7535e-02, -1.0000e-01, 4.1606e-02, ..., 5.9855e-03, + 1.1577e-02, -3.9706e-02], + [-1.6531e-02, -7.3718e-02, 7.0869e-02, ..., -7.7165e-02, + 2.5671e-02, -2.3182e-02]], device='cuda:0'), grad: tensor([[-8.1301e-04, 4.8280e-04, -1.8442e-04, ..., -6.2418e-04, + 2.0480e-04, 3.0017e-04], + [ 1.9297e-05, -8.8196e-03, 1.7393e-04, ..., -4.6844e-03, + 2.4629e-04, -1.5961e-02], + [ 5.3123e-06, 7.8201e-03, 2.5606e-04, ..., -3.0670e-03, + 2.4748e-04, 1.1124e-02], + ..., + [ 5.1856e-05, 8.4543e-04, 2.5749e-03, ..., 2.3632e-03, + 1.2903e-03, 4.7255e-04], + [ 8.2064e-04, 4.6229e-04, 1.3641e-02, ..., 7.7248e-03, + 1.9226e-03, 5.5275e-03], + [-9.9564e-04, 4.1819e-04, -1.6815e-02, ..., 1.3266e-03, + -2.7542e-03, 3.2949e-04]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0141, -0.0110, -0.0124, 0.0085, -0.0182, -0.0171, 0.0141, -0.0016, + 0.0180, -0.0115], device='cuda:0'), grad: tensor([ 5.8085e-05, -4.4739e-02, -2.0676e-02, 5.2834e-03, 1.1620e-02, + -4.5898e-02, 2.3895e-02, 2.2476e-02, 5.3833e-02, -5.8975e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 216.49, cls_loss 0.4884 cls_loss_mapping 0.0041 cls_loss_causal 0.4671 re_mapping 0.0074 re_causal 0.0186 /// teacc 98.80 lr 0.00010000 +Epoch 269, weight, value: tensor([[-6.0782e-02, -8.8248e-02, -1.3969e-01, ..., 1.1505e-01, + -5.1991e-02, 2.9857e-02], + [-8.0513e-02, 9.9811e-02, -7.4126e-02, ..., 4.4167e-02, + -6.1020e-02, -9.3273e-02], + [-4.0539e-02, 8.1702e-02, -1.0500e-01, ..., -1.5719e-02, + 3.0462e-02, 1.2136e-01], + ..., + [-4.8789e-03, 6.0601e-02, -8.8508e-05, ..., -1.6470e-02, + 3.1741e-02, -6.4198e-02], + [ 8.7241e-02, -1.0003e-01, 4.1184e-02, ..., 6.5154e-03, + 1.1091e-02, -4.1137e-02], + [-1.5486e-02, -7.3411e-02, 7.0678e-02, ..., -7.6334e-02, + 2.5682e-02, -2.3497e-02]], device='cuda:0'), grad: tensor([[ 1.6928e-04, 1.1361e-04, 3.0184e-04, ..., 6.8092e-04, + 4.2343e-04, 7.1812e-04], + [ 1.6272e-05, 1.3351e-04, 3.9029e-04, ..., -7.8869e-04, + -6.2227e-04, -1.7834e-03], + [ 1.2720e-04, 2.2960e-04, 4.7994e-04, ..., 1.2636e-03, + 6.0177e-04, 1.1396e-03], + ..., + [ 1.0186e-04, -9.3877e-05, 2.6913e-03, ..., 2.1732e-04, + 9.5272e-04, 2.2745e-04], + [ 4.7445e-04, -9.0790e-04, 1.3695e-03, ..., -3.2063e-03, + 1.2064e-03, -3.4695e-03], + [ 3.8266e-04, 1.0335e-04, -1.2054e-02, ..., -6.9809e-04, + -2.8400e-03, 5.3346e-06]], device='cuda:0') +Epoch 269, bias, value: tensor([ 0.0136, -0.0115, -0.0121, 0.0080, -0.0196, -0.0170, 0.0141, -0.0010, + 0.0184, -0.0105], device='cuda:0'), grad: tensor([ 0.0118, -0.0197, 0.0167, -0.0118, 0.0058, 0.0341, 0.0211, 0.0107, + -0.0238, -0.0448], device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 216.64, cls_loss 0.5217 cls_loss_mapping 0.0046 cls_loss_causal 0.4989 re_mapping 0.0072 re_causal 0.0181 /// teacc 98.85 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.0609, -0.0880, -0.1393, ..., 0.1141, -0.0521, 0.0298], + [-0.0813, 0.1002, -0.0734, ..., 0.0447, -0.0607, -0.0947], + [-0.0408, 0.0818, -0.1058, ..., -0.0172, 0.0305, 0.1209], + ..., + [-0.0064, 0.0605, -0.0003, ..., -0.0159, 0.0327, -0.0632], + [ 0.0875, -0.0999, 0.0409, ..., 0.0070, 0.0113, -0.0402], + [-0.0161, -0.0740, 0.0714, ..., -0.0755, 0.0260, -0.0239]], + device='cuda:0'), grad: tensor([[ 2.4939e-04, 1.1969e-04, 4.2939e-04, ..., 1.1082e-03, + 1.4102e-04, 2.0540e-04], + [ 6.7949e-04, 1.0002e-04, 2.7537e-04, ..., -2.2488e-03, + 9.8765e-05, 1.4341e-04], + [ 1.4362e-03, 1.6677e-04, 7.7963e-04, ..., 8.5211e-04, + 1.4305e-04, 1.6689e-04], + ..., + [-8.8730e-03, -1.5986e-04, -3.9172e-04, ..., 6.5374e-04, + 1.2236e-03, -2.4724e-04], + [ 1.0290e-03, 1.5056e-04, 7.5483e-04, ..., 1.3981e-03, + 2.1529e-04, 1.6046e-04], + [ 1.0986e-03, 2.9659e-04, 6.7215e-03, ..., 1.5411e-03, + 6.2943e-04, 7.0715e-04]], device='cuda:0') +Epoch 270, bias, value: tensor([ 0.0138, -0.0124, -0.0119, 0.0074, -0.0192, -0.0162, 0.0137, 0.0006, + 0.0171, -0.0105], device='cuda:0'), grad: tensor([ 0.0145, 0.0130, 0.0159, -0.0545, 0.0202, -0.0380, 0.0172, -0.0207, + 0.0217, 0.0107], device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 216.60, cls_loss 0.4897 cls_loss_mapping 0.0048 cls_loss_causal 0.4654 re_mapping 0.0077 re_causal 0.0185 /// teacc 98.91 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.0609, -0.0885, -0.1389, ..., 0.1137, -0.0518, 0.0307], + [-0.0817, 0.1006, -0.0732, ..., 0.0447, -0.0594, -0.0940], + [-0.0403, 0.0818, -0.1061, ..., -0.0177, 0.0305, 0.1204], + ..., + [-0.0057, 0.0607, -0.0014, ..., -0.0162, 0.0324, -0.0626], + [ 0.0873, -0.1003, 0.0406, ..., 0.0076, 0.0110, -0.0413], + [-0.0152, -0.0740, 0.0725, ..., -0.0761, 0.0267, -0.0221]], + device='cuda:0'), grad: tensor([[ 5.4359e-03, 2.9683e-04, 6.1083e-04, ..., 1.1253e-03, + 6.9571e-04, 3.3116e-04], + [ 2.5909e-06, -9.7322e-04, 1.1473e-03, ..., -1.5192e-03, + 6.8665e-05, 2.4366e-04], + [ 6.6400e-05, 2.2602e-04, 1.7726e-04, ..., 9.5749e-04, + -2.4423e-05, -3.5435e-05], + ..., + [ 3.5495e-05, -1.8644e-04, -2.2526e-03, ..., -3.7498e-03, + -4.8599e-03, -3.2520e-04], + [ 4.4632e-03, 6.4230e-04, 2.1477e-03, ..., 1.9102e-03, + 1.2856e-03, 4.5657e-04], + [-1.4656e-02, 1.0803e-05, 8.6451e-04, ..., 2.1877e-03, + 1.9236e-03, -9.8705e-04]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0136, -0.0123, -0.0112, 0.0075, -0.0197, -0.0160, 0.0137, 0.0002, + 0.0163, -0.0097], device='cuda:0'), grad: tensor([ 0.0302, 0.0178, 0.0160, 0.0182, -0.0310, -0.0357, 0.0195, 0.0006, + 0.0289, -0.0642], device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 216.93, cls_loss 0.4725 cls_loss_mapping 0.0043 cls_loss_causal 0.4459 re_mapping 0.0075 re_causal 0.0185 /// teacc 98.80 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.0604, -0.0869, -0.1386, ..., 0.1139, -0.0505, 0.0303], + [-0.0808, 0.1007, -0.0727, ..., 0.0446, -0.0589, -0.0938], + [-0.0404, 0.0815, -0.1060, ..., -0.0165, 0.0320, 0.1216], + ..., + [-0.0065, 0.0607, -0.0015, ..., -0.0154, 0.0324, -0.0627], + [ 0.0885, -0.1010, 0.0418, ..., 0.0080, 0.0113, -0.0411], + [-0.0142, -0.0741, 0.0724, ..., -0.0781, 0.0252, -0.0236]], + device='cuda:0'), grad: tensor([[-2.2352e-08, 1.7679e-04, 4.4346e-05, ..., 5.9223e-04, + -2.4438e-05, 1.3292e-04], + [-4.9826e-08, -2.0733e-03, 7.5161e-05, ..., -4.1389e-03, + 4.5806e-05, 2.6178e-04], + [ 1.8626e-08, 2.5916e-04, 9.8228e-05, ..., 1.5364e-03, + 1.3649e-04, 6.2084e-04], + ..., + [ 9.3132e-09, 5.1689e-04, 5.4026e-04, ..., 2.9831e-03, + 6.2048e-05, 7.7915e-04], + [ 1.2433e-07, -4.9973e-04, 7.9651e-03, ..., 4.6921e-03, + -3.4904e-04, -2.5005e-03], + [ 2.4680e-08, 1.3697e-04, 2.0504e-03, ..., 2.0905e-03, + 4.8208e-04, 1.7345e-04]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0143, -0.0117, -0.0099, 0.0070, -0.0191, -0.0171, 0.0132, 0.0008, + 0.0161, -0.0108], device='cuda:0'), grad: tensor([ 0.0149, 0.0071, -0.0110, 0.0152, 0.0120, -0.0516, 0.0143, 0.0214, + -0.0089, -0.0134], device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 216.61, cls_loss 0.4892 cls_loss_mapping 0.0053 cls_loss_causal 0.4619 re_mapping 0.0073 re_causal 0.0191 /// teacc 98.67 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.0607, -0.0851, -0.1379, ..., 0.1138, -0.0504, 0.0304], + [-0.0813, 0.1018, -0.0730, ..., 0.0450, -0.0599, -0.0948], + [-0.0399, 0.0822, -0.1045, ..., -0.0159, 0.0328, 0.1219], + ..., + [-0.0068, 0.0604, -0.0020, ..., -0.0159, 0.0318, -0.0631], + [ 0.0889, -0.1021, 0.0416, ..., 0.0076, 0.0119, -0.0403], + [-0.0153, -0.0740, 0.0723, ..., -0.0782, 0.0255, -0.0224]], + device='cuda:0'), grad: tensor([[ 7.6485e-04, 2.1875e-04, 9.3508e-04, ..., 7.2956e-04, + 1.0729e-03, 9.4748e-04], + [ 2.6245e-06, -1.6613e-03, 2.2376e-04, ..., -1.1168e-03, + 4.5443e-04, 6.7043e-04], + [ 2.0349e-04, 6.2513e-04, 3.6049e-04, ..., 7.5340e-04, + 6.8569e-04, 5.0020e-04], + ..., + [ 8.1062e-05, -3.2020e-04, -4.9057e-03, ..., -1.3809e-03, + -1.2293e-03, -2.0466e-03], + [ 5.3177e-03, -6.3467e-04, 6.5575e-03, ..., -4.2844e-04, + 2.0428e-03, 6.8903e-05], + [ 3.3855e-04, 4.0817e-04, 2.6340e-03, ..., 5.3120e-04, + 6.2895e-04, 3.4976e-04]], device='cuda:0') +Epoch 273, bias, value: tensor([ 0.0134, -0.0110, -0.0091, 0.0071, -0.0180, -0.0187, 0.0134, 0.0002, + 0.0161, -0.0111], device='cuda:0'), grad: tensor([ 0.0191, 0.0090, -0.0145, 0.0048, 0.0144, -0.0135, -0.0067, -0.0372, + -0.0015, 0.0261], device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 216.74, cls_loss 0.4923 cls_loss_mapping 0.0040 cls_loss_causal 0.4652 re_mapping 0.0072 re_causal 0.0175 /// teacc 98.79 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.0622, -0.0844, -0.1386, ..., 0.1148, -0.0507, 0.0318], + [-0.0832, 0.1019, -0.0742, ..., 0.0446, -0.0603, -0.0940], + [-0.0398, 0.0811, -0.1047, ..., -0.0163, 0.0326, 0.1204], + ..., + [-0.0073, 0.0612, -0.0027, ..., -0.0159, 0.0311, -0.0634], + [ 0.0912, -0.1010, 0.0414, ..., 0.0083, 0.0109, -0.0403], + [-0.0149, -0.0739, 0.0725, ..., -0.0794, 0.0262, -0.0232]], + device='cuda:0'), grad: tensor([[ 5.0402e-04, 1.7977e-04, 3.1561e-05, ..., 7.2384e-04, + 1.9205e-04, 9.4995e-08], + [ 4.9448e-04, 1.1110e-04, 9.8228e-04, ..., 1.0271e-03, + 1.6725e-04, 6.5193e-08], + [ 1.3816e-04, 1.9467e-04, 4.9561e-05, ..., 8.4352e-04, + 3.6860e-04, 2.3795e-07], + ..., + [ 1.2815e-04, 1.0729e-03, 1.0347e-03, ..., 1.4572e-03, + 1.0004e-03, 9.3132e-09], + [ 3.9005e-04, 7.6830e-05, 1.0204e-04, ..., 6.7234e-04, + -2.9945e-04, 3.5902e-07], + [ 1.2946e-04, 3.0494e-04, 4.9286e-03, ..., 1.0004e-03, + 1.2970e-03, 9.6858e-08]], device='cuda:0') +Epoch 274, bias, value: tensor([ 1.3869e-02, -1.1711e-02, -9.6624e-03, 7.8124e-03, -1.7853e-02, + -1.8147e-02, 1.3611e-02, 8.5542e-05, 1.6201e-02, -1.1656e-02], + device='cuda:0'), grad: tensor([ 0.0244, -0.0335, -0.0076, -0.0401, -0.0320, 0.0478, -0.0154, 0.0059, + 0.0043, 0.0461], device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 216.66, cls_loss 0.5127 cls_loss_mapping 0.0056 cls_loss_causal 0.4824 re_mapping 0.0069 re_causal 0.0175 /// teacc 98.71 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.0594, -0.0851, -0.1389, ..., 0.1156, -0.0514, 0.0325], + [-0.0840, 0.1017, -0.0731, ..., 0.0451, -0.0606, -0.0949], + [-0.0401, 0.0812, -0.1040, ..., -0.0170, 0.0337, 0.1208], + ..., + [-0.0065, 0.0621, -0.0028, ..., -0.0168, 0.0317, -0.0639], + [ 0.0914, -0.1011, 0.0419, ..., 0.0085, 0.0111, -0.0401], + [-0.0150, -0.0742, 0.0720, ..., -0.0806, 0.0267, -0.0247]], + device='cuda:0'), grad: tensor([[ 1.8394e-04, 1.0240e-04, 3.4541e-05, ..., 3.0518e-04, + 1.3304e-04, 3.0470e-04], + [-1.4009e-03, -6.4611e-05, 5.6934e-04, ..., 5.2422e-05, + 2.1601e-04, -2.5616e-03], + [ 3.9625e-04, 5.7297e-03, 9.4652e-04, ..., 5.1308e-04, + 5.0888e-03, 8.0261e-03], + ..., + [ 1.3781e-04, -1.7595e-03, -1.8072e-03, ..., -2.7256e-03, + -9.9850e-04, 4.9067e-04], + [ 1.0452e-03, -4.9400e-03, -5.1804e-03, ..., -4.1175e-04, + -5.6419e-03, -7.3547e-03], + [ 1.2529e-04, 6.3515e-04, 5.3406e-03, ..., 2.1305e-03, + 1.8873e-03, 2.4235e-04]], device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0137, -0.0115, -0.0100, 0.0084, -0.0177, -0.0188, 0.0137, 0.0008, + 0.0168, -0.0128], device='cuda:0'), grad: tensor([ 0.0122, -0.0228, 0.0098, -0.0265, 0.0139, 0.0412, -0.0173, -0.0233, + -0.0260, 0.0387], device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 216.70, cls_loss 0.4961 cls_loss_mapping 0.0042 cls_loss_causal 0.4722 re_mapping 0.0071 re_causal 0.0184 /// teacc 98.76 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.0592, -0.0859, -0.1403, ..., 0.1158, -0.0520, 0.0314], + [-0.0842, 0.1020, -0.0729, ..., 0.0449, -0.0608, -0.0946], + [-0.0395, 0.0801, -0.1033, ..., -0.0161, 0.0348, 0.1209], + ..., + [-0.0048, 0.0625, -0.0033, ..., -0.0165, 0.0307, -0.0645], + [ 0.0901, -0.0999, 0.0421, ..., 0.0068, 0.0106, -0.0399], + [-0.0152, -0.0744, 0.0729, ..., -0.0800, 0.0272, -0.0244]], + device='cuda:0'), grad: tensor([[ 1.3089e-04, 3.7408e-04, -4.3154e-04, ..., 1.7567e-03, + 1.8692e-04, 1.3504e-03], + [ 2.6774e-04, 1.0281e-03, 2.4204e-03, ..., 8.1940e-03, + 2.5253e-03, 4.6206e-04], + [ 1.0958e-03, -1.3304e-03, 4.2856e-05, ..., -3.9368e-03, + -8.6260e-04, -3.4027e-03], + ..., + [ 4.8294e-03, 1.2550e-03, -9.0265e-04, ..., -2.1915e-03, + -1.2102e-03, -3.1686e-04], + [ 2.7919e-04, 1.2321e-03, 2.7633e-04, ..., -4.1580e-03, + 3.2635e-03, 2.4261e-03], + [ 3.5024e-04, 1.1301e-03, 1.8339e-03, ..., 2.9106e-03, + 2.7790e-03, 1.1559e-03]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0127, -0.0109, -0.0105, 0.0093, -0.0172, -0.0200, 0.0140, 0.0003, + 0.0166, -0.0119], device='cuda:0'), grad: tensor([-0.0009, 0.0507, -0.0523, -0.0506, 0.0109, -0.0258, 0.0026, 0.0212, + 0.0309, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 216.92, cls_loss 0.5075 cls_loss_mapping 0.0043 cls_loss_causal 0.4817 re_mapping 0.0069 re_causal 0.0172 /// teacc 98.77 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.0594, -0.0872, -0.1406, ..., 0.1153, -0.0520, 0.0316], + [-0.0839, 0.1016, -0.0731, ..., 0.0442, -0.0622, -0.0946], + [-0.0393, 0.0818, -0.1031, ..., -0.0161, 0.0345, 0.1214], + ..., + [-0.0042, 0.0614, -0.0027, ..., -0.0155, 0.0315, -0.0647], + [ 0.0899, -0.0996, 0.0426, ..., 0.0073, 0.0112, -0.0403], + [-0.0149, -0.0736, 0.0731, ..., -0.0818, 0.0271, -0.0256]], + device='cuda:0'), grad: tensor([[ 3.3832e-04, 1.6963e-04, 2.2209e-04, ..., 8.7643e-04, + 6.2637e-03, 2.3918e-03], + [-1.3275e-03, -4.9531e-05, 1.3560e-05, ..., 7.3338e-04, + 1.6952e-04, 1.4257e-04], + [ 4.5776e-04, 4.5133e-04, 3.1042e-04, ..., 1.4572e-03, + 1.0262e-03, 2.6464e-04], + ..., + [ 1.9693e-04, -1.1702e-03, -4.3182e-03, ..., -1.5056e-04, + -1.7014e-03, 1.4710e-04], + [ 5.3644e-04, 3.3402e-04, 1.1444e-02, ..., 2.8667e-03, + 1.5099e-02, -4.2572e-03], + [-4.7982e-05, -1.6129e-04, -1.0056e-02, ..., -3.2997e-03, + -2.5391e-02, 6.3705e-04]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0123, -0.0115, -0.0096, 0.0092, -0.0182, -0.0204, 0.0147, 0.0002, + 0.0174, -0.0119], device='cuda:0'), grad: tensor([ 0.0224, -0.0161, 0.0172, -0.0405, 0.0071, 0.0189, -0.0002, 0.0092, + 0.0142, -0.0321], device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 216.58, cls_loss 0.4929 cls_loss_mapping 0.0046 cls_loss_causal 0.4651 re_mapping 0.0067 re_causal 0.0176 /// teacc 98.84 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.0590, -0.0874, -0.1422, ..., 0.1152, -0.0534, 0.0320], + [-0.0831, 0.1016, -0.0735, ..., 0.0438, -0.0626, -0.0949], + [-0.0407, 0.0813, -0.1027, ..., -0.0161, 0.0369, 0.1214], + ..., + [-0.0048, 0.0613, -0.0038, ..., -0.0153, 0.0304, -0.0650], + [ 0.0894, -0.0996, 0.0423, ..., 0.0067, 0.0111, -0.0408], + [-0.0145, -0.0740, 0.0744, ..., -0.0833, 0.0270, -0.0272]], + device='cuda:0'), grad: tensor([[ 6.6310e-07, 1.2875e-04, 1.0949e-04, ..., 1.1101e-03, + 1.5459e-03, 4.9925e-04], + [ 1.6997e-08, 6.5327e-05, 1.1319e-04, ..., -6.3777e-05, + 1.3561e-03, 4.7755e-04], + [ 2.8638e-08, -2.6011e-04, -3.4761e-04, ..., 2.1541e-04, + 2.4567e-03, -2.8000e-03], + ..., + [ 2.3283e-10, -2.0826e-04, -6.2847e-04, ..., -8.8573e-05, + -7.9298e-04, -1.1520e-03], + [ 2.0443e-07, 8.6486e-05, 2.6894e-04, ..., 1.1301e-03, + 1.7347e-03, 5.3406e-04], + [ 1.3970e-09, 8.9586e-05, -4.2486e-04, ..., -5.9166e-03, + -1.3107e-02, 5.1308e-04]], device='cuda:0') +Epoch 278, bias, value: tensor([ 1.2119e-02, -1.1668e-02, -1.0603e-02, 9.4247e-03, -1.7762e-02, + -1.9249e-02, 1.4272e-02, 4.2316e-05, 1.7728e-02, -1.1989e-02], + device='cuda:0'), grad: tensor([ 0.0148, -0.0140, -0.0170, -0.0165, 0.0153, 0.0155, 0.0143, -0.0128, + 0.0146, -0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 216.59, cls_loss 0.4879 cls_loss_mapping 0.0039 cls_loss_causal 0.4545 re_mapping 0.0068 re_causal 0.0178 /// teacc 98.78 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.0595, -0.0875, -0.1414, ..., 0.1160, -0.0541, 0.0323], + [-0.0829, 0.1008, -0.0730, ..., 0.0437, -0.0642, -0.0955], + [-0.0413, 0.0820, -0.1022, ..., -0.0164, 0.0370, 0.1219], + ..., + [-0.0057, 0.0618, -0.0038, ..., -0.0152, 0.0302, -0.0652], + [ 0.0888, -0.0999, 0.0420, ..., 0.0077, 0.0110, -0.0408], + [-0.0135, -0.0735, 0.0748, ..., -0.0832, 0.0282, -0.0267]], + device='cuda:0'), grad: tensor([[ 0.0001, -0.0024, -0.0005, ..., 0.0012, -0.0027, 0.0002], + [ 0.0001, 0.0002, 0.0006, ..., 0.0015, 0.0005, 0.0003], + [ 0.0003, 0.0012, 0.0005, ..., 0.0019, 0.0020, 0.0019], + ..., + [ 0.0001, 0.0004, 0.0016, ..., 0.0050, 0.0028, 0.0002], + [ 0.0007, 0.0004, 0.0017, ..., -0.0023, 0.0016, 0.0014], + [ 0.0001, -0.0010, -0.0052, ..., -0.0077, -0.0048, 0.0003]], + device='cuda:0') +Epoch 279, bias, value: tensor([ 1.1731e-02, -1.1451e-02, -1.0940e-02, 1.0407e-02, -1.7771e-02, + -1.9125e-02, 1.4092e-02, 3.1270e-05, 1.6842e-02, -1.1581e-02], + device='cuda:0'), grad: tensor([-0.0573, 0.0240, 0.0097, 0.0301, 0.0233, -0.0306, -0.0038, 0.0037, + 0.0097, -0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 216.75, cls_loss 0.5254 cls_loss_mapping 0.0040 cls_loss_causal 0.5006 re_mapping 0.0068 re_causal 0.0183 /// teacc 98.77 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.0594, -0.0884, -0.1421, ..., 0.1147, -0.0557, 0.0322], + [-0.0841, 0.1011, -0.0707, ..., 0.0441, -0.0613, -0.0958], + [-0.0411, 0.0808, -0.1035, ..., -0.0164, 0.0359, 0.1221], + ..., + [-0.0051, 0.0619, -0.0032, ..., -0.0146, 0.0301, -0.0639], + [ 0.0889, -0.1005, 0.0415, ..., 0.0083, 0.0110, -0.0417], + [-0.0140, -0.0720, 0.0745, ..., -0.0828, 0.0278, -0.0277]], + device='cuda:0'), grad: tensor([[ 3.5822e-05, 2.2277e-05, 1.2684e-03, ..., 3.6573e-04, + 1.8139e-03, 9.2936e-04], + [ 6.7472e-05, 3.5119e-04, 1.1959e-03, ..., 1.0824e-03, + 1.3247e-03, 4.9734e-04], + [ 1.3006e-04, -1.8959e-03, -1.8063e-03, ..., -1.6747e-03, + -3.2959e-03, -7.1526e-03], + ..., + [-2.7537e-04, -1.0386e-03, -1.3475e-03, ..., -3.8528e-03, + 2.7199e-03, -7.1573e-04], + [-4.1080e-04, 1.7786e-03, 2.0447e-03, ..., 5.9223e-04, + 5.2567e-03, 2.7981e-03], + [ 6.3086e-04, 3.9840e-04, 6.5117e-03, ..., 1.5793e-03, + 6.0005e-03, 1.4114e-03]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0127, -0.0119, -0.0110, 0.0110, -0.0182, -0.0184, 0.0133, -0.0004, + 0.0173, -0.0121], device='cuda:0'), grad: tensor([ 0.0172, 0.0172, -0.0327, 0.0131, -0.0502, 0.0229, -0.0124, -0.0366, + 0.0309, 0.0305], device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 216.62, cls_loss 0.4954 cls_loss_mapping 0.0040 cls_loss_causal 0.4716 re_mapping 0.0064 re_causal 0.0159 /// teacc 98.84 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.0588, -0.0883, -0.1421, ..., 0.1145, -0.0544, 0.0330], + [-0.0848, 0.1009, -0.0717, ..., 0.0440, -0.0615, -0.0966], + [-0.0407, 0.0808, -0.1042, ..., -0.0161, 0.0351, 0.1226], + ..., + [-0.0051, 0.0620, -0.0030, ..., -0.0149, 0.0303, -0.0645], + [ 0.0881, -0.1005, 0.0421, ..., 0.0089, 0.0114, -0.0409], + [-0.0144, -0.0726, 0.0735, ..., -0.0814, 0.0279, -0.0270]], + device='cuda:0'), grad: tensor([[ 9.5248e-05, 6.1214e-05, 2.7609e-04, ..., 6.3038e-04, + 3.7241e-04, 9.2316e-04], + [ 2.5344e-04, 1.6987e-04, 6.5804e-04, ..., 1.0700e-03, + 7.9775e-04, 1.3905e-03], + [ 1.4126e-04, 7.3433e-05, 3.4809e-04, ..., 6.1750e-04, + 6.1560e-04, 1.3800e-03], + ..., + [ 2.5344e-04, -6.8855e-04, 3.2783e-04, ..., 7.9107e-04, + 6.2275e-04, 1.1883e-03], + [ 1.4770e-04, 9.9957e-05, 4.0674e-04, ..., 3.4618e-04, + 7.9811e-05, 5.3406e-04], + [ 2.5787e-03, 8.9979e-04, 2.8400e-03, ..., 9.3937e-04, + 2.8729e-04, 1.0653e-03]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0132, -0.0121, -0.0117, 0.0111, -0.0167, -0.0202, 0.0129, -0.0002, + 0.0179, -0.0120], device='cuda:0'), grad: tensor([ 0.0125, 0.0186, 0.0147, -0.0453, -0.0164, 0.0128, -0.0178, 0.0165, + 0.0130, -0.0085], device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 216.43, cls_loss 0.4935 cls_loss_mapping 0.0033 cls_loss_causal 0.4642 re_mapping 0.0068 re_causal 0.0184 /// teacc 98.85 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.0585, -0.0885, -0.1418, ..., 0.1145, -0.0541, 0.0329], + [-0.0850, 0.1008, -0.0719, ..., 0.0447, -0.0605, -0.0954], + [-0.0399, 0.0812, -0.1031, ..., -0.0168, 0.0363, 0.1231], + ..., + [-0.0054, 0.0620, -0.0031, ..., -0.0141, 0.0289, -0.0635], + [ 0.0881, -0.1016, 0.0425, ..., 0.0088, 0.0130, -0.0416], + [-0.0147, -0.0724, 0.0734, ..., -0.0816, 0.0286, -0.0281]], + device='cuda:0'), grad: tensor([[ 2.1625e-04, 2.5082e-04, 2.1112e-04, ..., 4.3464e-04, + 4.1223e-04, 6.3801e-04], + [ 6.2995e-06, 1.9956e-04, 1.6427e-04, ..., 5.9891e-04, + 2.7823e-04, 5.6839e-04], + [-1.1581e-04, -2.1763e-03, -4.3225e-04, ..., -8.9073e-04, + -5.9605e-04, -3.9673e-03], + ..., + [ 1.8448e-05, 7.4387e-03, 1.3626e-02, ..., 6.1302e-03, + 1.0414e-02, 2.7924e-03], + [ 7.9060e-04, 4.5919e-04, 4.3774e-04, ..., -2.5864e-03, + 6.7711e-04, -9.9850e-04], + [ 2.7254e-05, -6.9542e-03, -1.7731e-02, ..., -6.1302e-03, + -1.6769e-02, -1.3714e-03]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0129, -0.0110, -0.0121, 0.0114, -0.0161, -0.0208, 0.0136, -0.0003, + 0.0170, -0.0123], device='cuda:0'), grad: tensor([ 0.0124, -0.0203, -0.0168, -0.0130, 0.0097, 0.0153, 0.0076, 0.0392, + -0.0087, -0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 216.81, cls_loss 0.5007 cls_loss_mapping 0.0041 cls_loss_causal 0.4802 re_mapping 0.0067 re_causal 0.0172 /// teacc 98.79 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.0580, -0.0884, -0.1416, ..., 0.1161, -0.0529, 0.0341], + [-0.0850, 0.1019, -0.0707, ..., 0.0439, -0.0609, -0.0964], + [-0.0393, 0.0802, -0.1042, ..., -0.0161, 0.0355, 0.1226], + ..., + [-0.0059, 0.0620, -0.0027, ..., -0.0149, 0.0296, -0.0634], + [ 0.0879, -0.1013, 0.0431, ..., 0.0082, 0.0124, -0.0432], + [-0.0141, -0.0722, 0.0722, ..., -0.0820, 0.0275, -0.0279]], + device='cuda:0'), grad: tensor([[ 5.4017e-08, -4.6659e-04, -3.4485e-03, ..., -2.4147e-03, + -5.2643e-03, -4.4403e-03], + [ 1.4901e-07, 3.2282e-04, 2.8992e-04, ..., 2.5773e-04, + 3.9053e-04, 4.0460e-04], + [ 4.4191e-07, 5.0011e-03, 1.4954e-03, ..., 1.9970e-03, + 4.0894e-03, 8.6975e-03], + ..., + [ 9.3746e-04, 1.0994e-02, 8.9884e-04, ..., 4.4012e-04, + 1.5554e-03, 3.2120e-03], + [ 4.5542e-07, 3.1996e-04, 6.0892e-04, ..., -1.7130e-04, + 1.2274e-03, -2.0008e-03], + [-9.4128e-04, -1.5450e-02, 1.4484e-04, ..., 1.3714e-03, + 8.3733e-04, -6.1264e-03]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0139, -0.0116, -0.0130, 0.0121, -0.0160, -0.0203, 0.0139, 0.0006, + 0.0154, -0.0125], device='cuda:0'), grad: tensor([-0.0170, 0.0098, 0.0370, 0.0106, -0.0540, 0.0093, 0.0141, 0.0458, + -0.0228, -0.0328], device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 216.90, cls_loss 0.4940 cls_loss_mapping 0.0040 cls_loss_causal 0.4655 re_mapping 0.0069 re_causal 0.0177 /// teacc 98.84 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.0579, -0.0884, -0.1431, ..., 0.1170, -0.0535, 0.0344], + [-0.0862, 0.1019, -0.0709, ..., 0.0445, -0.0615, -0.0966], + [-0.0385, 0.0814, -0.1047, ..., -0.0156, 0.0361, 0.1234], + ..., + [-0.0060, 0.0612, -0.0032, ..., -0.0152, 0.0291, -0.0638], + [ 0.0883, -0.1014, 0.0435, ..., 0.0083, 0.0121, -0.0442], + [-0.0143, -0.0715, 0.0728, ..., -0.0818, 0.0280, -0.0266]], + device='cuda:0'), grad: tensor([[ 1.1349e-04, 1.4806e-04, 1.7011e-04, ..., 3.4904e-04, + 1.7393e-04, 2.3139e-04], + [ 2.4629e-04, -1.5959e-05, -3.8624e-05, ..., 1.2493e-04, + 8.1635e-04, 3.0971e-04], + [-5.8365e-04, -7.8297e-04, -1.4076e-03, ..., -4.2439e-04, + -5.9853e-03, -1.6193e-03], + ..., + [ 8.4162e-05, 1.6367e-04, 7.5996e-05, ..., -1.7595e-03, + 6.6233e-04, 2.8110e-04], + [ 1.5945e-03, 8.4698e-05, 1.2863e-04, ..., 3.1447e-04, + 2.4629e-04, 1.8799e-04], + [ 9.2328e-05, 1.9461e-05, 7.4089e-05, ..., 2.8563e-04, + 1.9872e-04, 3.0303e-04]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0142, -0.0110, -0.0135, 0.0117, -0.0149, -0.0207, 0.0133, -0.0005, + 0.0159, -0.0121], device='cuda:0'), grad: tensor([-0.0133, -0.0047, 0.0064, -0.0355, -0.0068, 0.0289, 0.0181, -0.0109, + -0.0030, 0.0207], device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 216.99, cls_loss 0.5095 cls_loss_mapping 0.0041 cls_loss_causal 0.4748 re_mapping 0.0070 re_causal 0.0187 /// teacc 98.83 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.0590, -0.0880, -0.1427, ..., 0.1185, -0.0528, 0.0344], + [-0.0866, 0.1016, -0.0704, ..., 0.0437, -0.0613, -0.0974], + [-0.0370, 0.0821, -0.1055, ..., -0.0162, 0.0352, 0.1225], + ..., + [-0.0058, 0.0615, -0.0041, ..., -0.0154, 0.0293, -0.0640], + [ 0.0881, -0.0998, 0.0423, ..., 0.0088, 0.0124, -0.0439], + [-0.0141, -0.0721, 0.0741, ..., -0.0815, 0.0281, -0.0264]], + device='cuda:0'), grad: tensor([[ 2.3991e-05, 3.7003e-04, -3.0684e-04, ..., -2.6836e-03, + 1.3676e-03, -2.1771e-05], + [ 3.3915e-05, -5.3673e-03, 3.5429e-04, ..., -1.3676e-03, + 4.6325e-04, 6.9666e-04], + [ 1.8537e-05, 7.9193e-03, 7.0286e-04, ..., 2.4834e-03, + 1.4992e-03, 8.8501e-03], + ..., + [-7.5865e-04, 3.1204e-03, 2.7618e-03, ..., 4.8523e-03, + 3.0937e-03, 2.8763e-03], + [ 1.2290e-04, -5.0812e-03, 1.3657e-03, ..., -2.5578e-03, + 2.4319e-03, -9.9182e-03], + [ 9.0637e-03, -2.0676e-03, -2.7657e-03, ..., -4.4174e-03, + -2.0905e-03, -2.0523e-03]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0147, -0.0112, -0.0142, 0.0112, -0.0152, -0.0210, 0.0144, 0.0004, + 0.0155, -0.0121], device='cuda:0'), grad: tensor([-0.0030, -0.0252, 0.0661, 0.0098, -0.0192, -0.0080, -0.0594, 0.0535, + -0.0446, 0.0299], device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 217.00, cls_loss 0.4853 cls_loss_mapping 0.0058 cls_loss_causal 0.4574 re_mapping 0.0064 re_causal 0.0164 /// teacc 98.81 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.0593, -0.0898, -0.1426, ..., 0.1171, -0.0527, 0.0338], + [-0.0871, 0.1033, -0.0711, ..., 0.0440, -0.0609, -0.0976], + [-0.0368, 0.0813, -0.1062, ..., -0.0166, 0.0358, 0.1225], + ..., + [-0.0057, 0.0614, -0.0052, ..., -0.0162, 0.0281, -0.0648], + [ 0.0890, -0.0997, 0.0427, ..., 0.0109, 0.0123, -0.0431], + [-0.0147, -0.0719, 0.0744, ..., -0.0826, 0.0275, -0.0266]], + device='cuda:0'), grad: tensor([[-9.0981e-04, -1.8320e-03, 9.0338e-07, ..., -4.5738e-03, + -1.8082e-03, -5.3711e-03], + [-6.6910e-03, -7.4959e-03, 3.6154e-06, ..., 5.3883e-04, + 7.5281e-05, 9.9003e-05], + [ 1.7633e-03, 5.1422e-03, -1.6105e-04, ..., 6.3658e-04, + -2.2686e-04, 8.2731e-04], + ..., + [ 1.0643e-03, -2.7580e-03, 2.0161e-05, ..., 1.0071e-03, + 8.7559e-05, 1.4520e-04], + [ 9.5892e-04, 1.3056e-03, 1.9276e-04, ..., 1.8530e-03, + 6.4707e-04, 5.4598e-04], + [ 5.6171e-04, 5.9700e-04, 2.4188e-04, ..., 8.4877e-04, + 2.3437e-04, 1.6725e-04]], device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0148, -0.0102, -0.0141, 0.0110, -0.0157, -0.0215, 0.0142, -0.0003, + 0.0174, -0.0131], device='cuda:0'), grad: tensor([-0.0342, -0.0209, 0.0273, 0.0125, 0.0117, -0.0202, 0.0262, -0.0269, + 0.0141, 0.0103], device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 216.86, cls_loss 0.4613 cls_loss_mapping 0.0031 cls_loss_causal 0.4422 re_mapping 0.0067 re_causal 0.0179 /// teacc 98.59 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.0599, -0.0900, -0.1430, ..., 0.1170, -0.0532, 0.0329], + [-0.0860, 0.1033, -0.0721, ..., 0.0436, -0.0611, -0.0980], + [-0.0370, 0.0812, -0.1064, ..., -0.0160, 0.0355, 0.1232], + ..., + [-0.0051, 0.0620, -0.0048, ..., -0.0168, 0.0294, -0.0645], + [ 0.0903, -0.1001, 0.0423, ..., 0.0104, 0.0113, -0.0434], + [-0.0155, -0.0727, 0.0746, ..., -0.0823, 0.0276, -0.0272]], + device='cuda:0'), grad: tensor([[ 5.5283e-05, 1.4938e-05, -9.8571e-06, ..., 1.5354e-03, + -9.7305e-06, 9.8586e-05], + [ 7.1645e-05, -1.5802e-03, 3.6880e-07, ..., -7.2289e-04, + 3.8780e-06, 2.6256e-05], + [ 1.0834e-03, 1.0422e-02, 1.9372e-06, ..., 1.9002e-04, + 1.5223e-04, 1.7703e-04], + ..., + [ 1.2505e-04, 1.9073e-03, 1.4789e-05, ..., 1.1120e-03, + 1.7226e-04, -1.7071e-04], + [ 9.9719e-05, 8.6248e-05, 3.2216e-05, ..., 2.9898e-04, + 1.1367e-04, 2.2009e-05], + [ 7.6234e-05, 2.3529e-05, -5.8919e-05, ..., 1.1396e-04, + 2.7394e-04, 8.2180e-06]], device='cuda:0') +Epoch 287, bias, value: tensor([ 1.4320e-02, -1.0130e-02, -1.3777e-02, 1.1180e-02, -1.5622e-02, + -2.1610e-02, 1.4105e-02, -6.7699e-05, 1.7889e-02, -1.3922e-02], + device='cuda:0'), grad: tensor([ 0.0178, 0.0091, 0.0290, -0.0823, 0.0075, -0.0242, 0.0143, 0.0110, + 0.0090, 0.0088], device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 216.94, cls_loss 0.4817 cls_loss_mapping 0.0042 cls_loss_causal 0.4493 re_mapping 0.0068 re_causal 0.0175 /// teacc 98.81 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.0616, -0.0894, -0.1428, ..., 0.1163, -0.0528, 0.0346], + [-0.0859, 0.1038, -0.0717, ..., 0.0447, -0.0614, -0.0974], + [-0.0364, 0.0812, -0.1037, ..., -0.0163, 0.0366, 0.1231], + ..., + [-0.0058, 0.0624, -0.0051, ..., -0.0165, 0.0296, -0.0639], + [ 0.0899, -0.1010, 0.0411, ..., 0.0095, 0.0118, -0.0431], + [-0.0155, -0.0740, 0.0740, ..., -0.0820, 0.0267, -0.0286]], + device='cuda:0'), grad: tensor([[ 8.8476e-09, 1.0304e-05, -8.7738e-03, ..., 4.9782e-04, + -1.2611e-02, 2.4354e-04], + [ 1.3970e-09, 2.6628e-05, -3.2520e-03, ..., -3.4618e-03, + 2.7599e-03, 1.4044e-05], + [ 2.2817e-08, -2.7776e-04, 1.8015e-03, ..., 3.6335e-04, + 1.9817e-03, 1.0796e-02], + ..., + [ 9.3132e-10, -2.2545e-05, 1.8826e-03, ..., 8.2827e-04, + 1.2989e-03, 1.8865e-05], + [ 7.2923e-07, 2.7776e-05, 1.1530e-03, ..., -1.8339e-03, + 8.4591e-04, -1.1345e-02], + [ 3.3993e-08, 2.9266e-05, 1.2884e-03, ..., 9.1267e-04, + 2.7466e-04, 9.7379e-06]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0135, -0.0098, -0.0141, 0.0115, -0.0161, -0.0210, 0.0142, 0.0004, + 0.0166, -0.0130], device='cuda:0'), grad: tensor([-0.0163, -0.0635, 0.0456, 0.0173, 0.0217, 0.0158, 0.0212, -0.0122, + -0.0456, 0.0160], device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 216.81, cls_loss 0.5012 cls_loss_mapping 0.0041 cls_loss_causal 0.4754 re_mapping 0.0063 re_causal 0.0168 /// teacc 98.65 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.0620, -0.0889, -0.1427, ..., 0.1149, -0.0524, 0.0343], + [-0.0850, 0.1026, -0.0695, ..., 0.0450, -0.0594, -0.0977], + [-0.0370, 0.0820, -0.1040, ..., -0.0160, 0.0373, 0.1224], + ..., + [-0.0059, 0.0628, -0.0061, ..., -0.0158, 0.0284, -0.0649], + [ 0.0903, -0.1004, 0.0420, ..., 0.0098, 0.0123, -0.0416], + [-0.0149, -0.0743, 0.0744, ..., -0.0807, 0.0263, -0.0295]], + device='cuda:0'), grad: tensor([[-2.9135e-04, -1.0719e-03, 4.0457e-06, ..., -2.6264e-03, + -1.3046e-02, -2.4967e-03], + [ 2.9945e-04, 7.7934e-03, -4.1294e-04, ..., -8.2350e-04, + 3.7503e-04, -8.2064e-04], + [ 6.0701e-04, 1.9627e-03, 3.8362e-04, ..., 1.2436e-03, + 2.8419e-03, 1.0653e-03], + ..., + [ 1.2798e-03, 1.8797e-03, 7.0429e-04, ..., 7.0858e-04, + 1.2150e-03, 2.3639e-04], + [ 4.0269e-04, 8.3399e-04, 2.1374e-04, ..., 7.2765e-04, + 8.9359e-04, 6.8283e-04], + [-7.6485e-04, 9.2506e-04, 1.1139e-05, ..., 9.0981e-04, + 3.7479e-03, -6.2764e-05]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0132, -0.0106, -0.0136, 0.0114, -0.0157, -0.0214, 0.0143, 0.0004, + 0.0171, -0.0129], device='cuda:0'), grad: tensor([-0.0849, 0.0288, 0.0129, 0.0213, 0.0345, -0.0075, -0.0092, 0.0036, + 0.0349, -0.0343], device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 216.69, cls_loss 0.4902 cls_loss_mapping 0.0058 cls_loss_causal 0.4570 re_mapping 0.0062 re_causal 0.0157 /// teacc 98.83 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.0622, -0.0882, -0.1440, ..., 0.1156, -0.0516, 0.0350], + [-0.0837, 0.1031, -0.0687, ..., 0.0457, -0.0587, -0.0978], + [-0.0371, 0.0833, -0.1044, ..., -0.0162, 0.0367, 0.1220], + ..., + [-0.0066, 0.0617, -0.0040, ..., -0.0163, 0.0292, -0.0649], + [ 0.0907, -0.1012, 0.0422, ..., 0.0094, 0.0128, -0.0418], + [-0.0161, -0.0744, 0.0733, ..., -0.0806, 0.0264, -0.0301]], + device='cuda:0'), grad: tensor([[ 1.0836e-04, 2.3115e-04, -7.2978e-06, ..., 9.2268e-04, + 4.1556e-04, 3.0613e-04], + [-3.5167e-04, 7.7820e-04, 2.7552e-05, ..., -4.2081e-04, + 3.4857e-04, 3.5596e-04], + [ 1.6403e-04, 3.3627e-03, -1.1152e-04, ..., 1.3676e-03, + 2.4796e-04, 4.7064e-04], + ..., + [ 1.2910e-04, -5.8708e-03, 1.8692e-04, ..., 8.1360e-05, + 5.9175e-04, 2.5463e-04], + [ 6.0892e-04, 3.5596e-04, -5.2357e-04, ..., 1.8740e-03, + 2.5597e-03, 1.3285e-03], + [-1.6987e-05, -1.3256e-04, 1.1575e-04, ..., 5.7697e-04, + -3.2753e-05, -1.0490e-03]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0131, -0.0100, -0.0135, 0.0113, -0.0161, -0.0204, 0.0142, -0.0006, + 0.0170, -0.0128], device='cuda:0'), grad: tensor([ 0.0190, 0.0215, 0.0321, -0.0243, -0.0124, -0.0346, 0.0145, 0.0020, + 0.0206, -0.0383], device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 216.77, cls_loss 0.4708 cls_loss_mapping 0.0042 cls_loss_causal 0.4447 re_mapping 0.0069 re_causal 0.0175 /// teacc 98.55 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.0624, -0.0884, -0.1439, ..., 0.1157, -0.0523, 0.0344], + [-0.0834, 0.1040, -0.0693, ..., 0.0460, -0.0588, -0.0979], + [-0.0367, 0.0836, -0.1042, ..., -0.0173, 0.0364, 0.1233], + ..., + [-0.0074, 0.0615, -0.0045, ..., -0.0171, 0.0302, -0.0652], + [ 0.0905, -0.1018, 0.0418, ..., 0.0098, 0.0122, -0.0426], + [-0.0159, -0.0744, 0.0742, ..., -0.0798, 0.0268, -0.0303]], + device='cuda:0'), grad: tensor([[ 6.4820e-07, 7.6711e-05, 3.2973e-04, ..., 1.1969e-04, + 1.3959e-04, 1.2994e-05], + [ 3.0279e-04, 7.6199e-04, 7.0512e-05, ..., 2.3842e-04, + 8.4788e-06, 1.0622e-04], + [-2.0008e-03, -1.2417e-03, 4.4942e-05, ..., 1.5950e-04, + 1.6606e-04, -5.1498e-04], + ..., + [ 1.6174e-03, 1.7014e-03, -8.3447e-04, ..., -1.5223e-04, + -6.5923e-05, 5.9938e-04], + [ 5.8115e-05, 3.2425e-04, -1.3571e-03, ..., -1.1808e-04, + -2.3079e-03, -7.5054e-04], + [-4.7445e-05, 2.2447e-04, -1.4257e-04, ..., 5.5981e-04, + -3.0696e-05, 9.4354e-05]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0135, -0.0095, -0.0143, 0.0109, -0.0162, -0.0198, 0.0134, -0.0005, + 0.0171, -0.0124], device='cuda:0'), grad: tensor([ 0.0048, 0.0107, 0.0035, -0.0162, 0.0041, -0.0175, 0.0055, 0.0081, + -0.0105, 0.0074], device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 216.51, cls_loss 0.4664 cls_loss_mapping 0.0034 cls_loss_causal 0.4399 re_mapping 0.0063 re_causal 0.0158 /// teacc 98.69 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.0629, -0.0866, -0.1441, ..., 0.1169, -0.0516, 0.0352], + [-0.0825, 0.1045, -0.0659, ..., 0.0462, -0.0563, -0.0969], + [-0.0380, 0.0842, -0.1048, ..., -0.0180, 0.0363, 0.1229], + ..., + [-0.0073, 0.0601, -0.0059, ..., -0.0171, 0.0297, -0.0645], + [ 0.0914, -0.1020, 0.0418, ..., 0.0088, 0.0123, -0.0425], + [-0.0158, -0.0735, 0.0751, ..., -0.0803, 0.0269, -0.0319]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3888e-04, -4.5180e-05, ..., 1.1504e-04, + -1.0753e-04, 4.2391e-04], + [ 4.6566e-10, -2.4891e-03, -1.0997e-04, ..., -1.9283e-03, + 2.5611e-08, -6.3086e-04], + [ 1.2107e-08, -1.4313e-05, 8.9183e-06, ..., 2.4354e-04, + 2.8638e-07, 2.7037e-04], + ..., + [-1.0431e-07, 7.8726e-04, 2.6870e-04, ..., 4.7421e-04, + 6.2548e-06, 3.7646e-04], + [ 3.2596e-09, 2.3508e-04, 6.3121e-05, ..., 1.6582e-04, + 1.6335e-06, 3.2473e-04], + [ 8.9454e-07, 1.5986e-04, 9.5963e-05, ..., 1.1921e-04, + 8.2850e-05, 3.7742e-04]], device='cuda:0') +Epoch 292, bias, value: tensor([ 1.5171e-02, -9.3187e-03, -1.3921e-02, 9.9335e-03, -1.5742e-02, + -1.9475e-02, 1.3338e-02, 6.2332e-05, 1.5981e-02, -1.3471e-02], + device='cuda:0'), grad: tensor([ 0.0156, -0.0536, 0.0154, 0.0158, -0.0151, 0.0154, -0.0089, -0.0137, + 0.0141, 0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 216.60, cls_loss 0.4990 cls_loss_mapping 0.0049 cls_loss_causal 0.4738 re_mapping 0.0067 re_causal 0.0169 /// teacc 98.75 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.0634, -0.0874, -0.1443, ..., 0.1170, -0.0527, 0.0358], + [-0.0829, 0.1044, -0.0668, ..., 0.0465, -0.0566, -0.0964], + [-0.0363, 0.0840, -0.1046, ..., -0.0184, 0.0361, 0.1227], + ..., + [-0.0073, 0.0604, -0.0061, ..., -0.0172, 0.0289, -0.0637], + [ 0.0909, -0.1023, 0.0425, ..., 0.0093, 0.0125, -0.0424], + [-0.0168, -0.0739, 0.0751, ..., -0.0798, 0.0268, -0.0315]], + device='cuda:0'), grad: tensor([[ 4.4847e-04, -1.4229e-03, 3.2991e-05, ..., -1.8251e-04, + -7.4911e-04, -9.5129e-04], + [ 8.3160e-04, 3.6168e-04, 1.2442e-06, ..., -3.7060e-03, + -1.6727e-03, -9.9850e-04], + [ 4.4155e-04, 2.6345e-04, 1.8701e-06, ..., 6.9046e-04, + 7.0143e-04, 5.1641e-04], + ..., + [-8.7881e-04, -3.3307e-04, 7.7020e-07, ..., -1.8823e-04, + -7.0076e-03, 2.4867e-04], + [ 3.1872e-03, 4.7302e-04, 2.7623e-06, ..., 1.3628e-03, + 1.6537e-03, 2.3293e-04], + [ 2.5921e-03, 2.2376e-04, 2.3857e-05, ..., 4.9353e-04, + 4.0550e-03, 1.7881e-04]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0152, -0.0090, -0.0147, 0.0107, -0.0162, -0.0199, 0.0131, 0.0003, + 0.0165, -0.0134], device='cuda:0'), grad: tensor([-0.0223, -0.0154, 0.0147, -0.0147, 0.0138, -0.0105, 0.0188, -0.0386, + 0.0270, 0.0272], device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 217.02, cls_loss 0.4886 cls_loss_mapping 0.0050 cls_loss_causal 0.4622 re_mapping 0.0071 re_causal 0.0177 /// teacc 98.76 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.0642, -0.0871, -0.1449, ..., 0.1172, -0.0530, 0.0362], + [-0.0828, 0.1047, -0.0655, ..., 0.0476, -0.0568, -0.0966], + [-0.0382, 0.0837, -0.1050, ..., -0.0178, 0.0364, 0.1232], + ..., + [-0.0055, 0.0610, -0.0066, ..., -0.0178, 0.0283, -0.0642], + [ 0.0907, -0.1022, 0.0417, ..., 0.0091, 0.0126, -0.0417], + [-0.0160, -0.0741, 0.0748, ..., -0.0796, 0.0266, -0.0312]], + device='cuda:0'), grad: tensor([[ 2.1327e-04, 5.1886e-05, 3.6263e-04, ..., 3.4124e-05, + 5.1975e-04, -4.5824e-04], + [ 2.1064e-04, -3.6407e-04, 3.7909e-04, ..., -7.7188e-05, + 5.6219e-04, -7.9441e-04], + [ 3.2878e-04, -1.0389e-04, 1.1027e-04, ..., 9.9361e-05, + 1.5342e-04, 8.6403e-04], + ..., + [ 1.6916e-04, 1.7142e-04, 4.8518e-04, ..., 1.3781e-04, + 4.3488e-04, 3.2520e-04], + [-1.2836e-03, 3.0369e-05, 9.8610e-04, ..., 1.8501e-04, + 7.2098e-04, 6.9618e-04], + [ 3.9005e-04, 4.0770e-05, 2.3315e-02, ..., 3.2902e-04, + 9.8267e-03, 5.6362e-04]], device='cuda:0') +Epoch 294, bias, value: tensor([ 1.5412e-02, -8.5694e-03, -1.4477e-02, 1.0728e-02, -1.6554e-02, + -1.9478e-02, 1.2414e-02, -6.0537e-05, 1.5804e-02, -1.2698e-02], + device='cuda:0'), grad: tensor([ 0.0042, -0.0136, -0.0094, -0.0302, -0.0387, -0.0065, 0.0342, 0.0143, + -0.0019, 0.0475], device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 216.50, cls_loss 0.5101 cls_loss_mapping 0.0035 cls_loss_causal 0.4818 re_mapping 0.0067 re_causal 0.0174 /// teacc 98.77 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.0640, -0.0875, -0.1454, ..., 0.1186, -0.0523, 0.0357], + [-0.0830, 0.1053, -0.0645, ..., 0.0482, -0.0559, -0.0973], + [-0.0375, 0.0836, -0.1048, ..., -0.0176, 0.0362, 0.1248], + ..., + [-0.0071, 0.0612, -0.0061, ..., -0.0164, 0.0295, -0.0641], + [ 0.0902, -0.1019, 0.0407, ..., 0.0091, 0.0122, -0.0414], + [-0.0143, -0.0744, 0.0748, ..., -0.0786, 0.0271, -0.0311]], + device='cuda:0'), grad: tensor([[-1.0604e-04, 4.2200e-04, -5.4032e-05, ..., 2.9445e-04, + 6.5327e-05, 7.1192e-04], + [ 1.1456e-04, -1.4424e-04, 1.4079e-04, ..., -5.0879e-04, + 4.0627e-04, 1.2083e-03], + [ 5.2303e-05, 5.7727e-05, 2.7490e-04, ..., 5.7602e-04, + 8.0347e-04, -4.9820e-03], + ..., + [ 7.0155e-05, -1.1158e-03, -1.3895e-03, ..., 1.2301e-05, + -1.8587e-03, 1.0948e-03], + [-6.3896e-04, 4.2176e-04, 1.7369e-04, ..., -3.4070e-04, + 2.9206e-04, 2.1112e-04], + [ 6.1631e-05, -2.3842e-03, 6.3992e-04, ..., -3.0727e-03, + -2.0065e-03, -2.6379e-03]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0145, -0.0078, -0.0140, 0.0113, -0.0176, -0.0195, 0.0114, 0.0013, + 0.0156, -0.0129], device='cuda:0'), grad: tensor([ 0.0212, 0.0105, -0.0155, 0.0416, -0.0120, -0.0064, 0.0211, -0.0147, + -0.0090, -0.0368], device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 216.82, cls_loss 0.4672 cls_loss_mapping 0.0048 cls_loss_causal 0.4407 re_mapping 0.0071 re_causal 0.0172 /// teacc 98.57 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.0628, -0.0880, -0.1452, ..., 0.1178, -0.0530, 0.0347], + [-0.0815, 0.1051, -0.0636, ..., 0.0475, -0.0554, -0.0975], + [-0.0379, 0.0840, -0.1067, ..., -0.0186, 0.0356, 0.1259], + ..., + [-0.0069, 0.0614, -0.0065, ..., -0.0159, 0.0284, -0.0638], + [ 0.0896, -0.1025, 0.0408, ..., 0.0108, 0.0129, -0.0404], + [-0.0147, -0.0746, 0.0756, ..., -0.0791, 0.0283, -0.0303]], + device='cuda:0'), grad: tensor([[ 9.5546e-05, 2.3410e-05, 1.9479e-04, ..., 7.5102e-04, + 6.0654e-04, 9.7656e-04], + [ 5.0068e-05, 3.2216e-05, 4.5085e-04, ..., 2.3067e-04, + 3.0708e-04, 3.0708e-04], + [ 5.5838e-04, 1.6248e-04, 2.0981e-05, ..., 1.0443e-03, + 2.0733e-03, 1.3943e-03], + ..., + [ 6.6400e-05, -1.3280e-04, -5.2243e-05, ..., 9.9063e-05, + 1.8764e-04, 1.9121e-04], + [-8.2731e-04, 1.6499e-04, 1.9264e-04, ..., 5.6763e-03, + 1.2074e-03, 8.2703e-03], + [-1.2598e-03, -9.3222e-04, 5.2065e-05, ..., 1.7548e-04, + -4.7379e-03, -1.6861e-03]], device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0137, -0.0081, -0.0138, 0.0102, -0.0169, -0.0202, 0.0126, 0.0017, + 0.0164, -0.0131], device='cuda:0'), grad: tensor([ 1.9089e-02, 2.7893e-02, 2.3514e-02, -2.7359e-02, -1.2886e-02, + -5.8746e-03, -8.8425e-03, 2.0142e-02, 5.2482e-05, -3.5706e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 216.83, cls_loss 0.4878 cls_loss_mapping 0.0055 cls_loss_causal 0.4661 re_mapping 0.0066 re_causal 0.0166 /// teacc 98.76 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.0635, -0.0892, -0.1459, ..., 0.1166, -0.0536, 0.0339], + [-0.0820, 0.1053, -0.0636, ..., 0.0494, -0.0555, -0.0984], + [-0.0369, 0.0836, -0.1069, ..., -0.0188, 0.0362, 0.1262], + ..., + [-0.0069, 0.0618, -0.0058, ..., -0.0166, 0.0295, -0.0639], + [ 0.0898, -0.1033, 0.0413, ..., 0.0103, 0.0127, -0.0407], + [-0.0152, -0.0757, 0.0744, ..., -0.0783, 0.0262, -0.0313]], + device='cuda:0'), grad: tensor([[ 3.1561e-05, 1.0967e-05, 1.1824e-05, ..., 3.8147e-05, + -3.7169e-04, -1.4601e-03], + [ 2.2370e-06, -5.6982e-04, 9.3603e-04, ..., -8.3160e-04, + 5.3868e-06, 3.4124e-06], + [ 5.7109e-06, 5.8562e-05, 3.3915e-05, ..., 9.3400e-05, + 2.5049e-05, 3.3021e-05], + ..., + [ 2.3305e-04, -9.4855e-07, 1.1586e-05, ..., 5.5522e-05, + 9.9465e-06, 9.6112e-06], + [-1.9455e-03, 4.6313e-05, 3.7074e-04, ..., 8.0585e-05, + 2.0003e-04, 2.4244e-05], + [ 1.2360e-03, 3.2961e-05, -8.9359e-04, ..., 4.4614e-05, + -3.2711e-04, 3.7909e-05]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0132, -0.0081, -0.0141, 0.0105, -0.0166, -0.0199, 0.0136, 0.0014, + 0.0158, -0.0135], device='cuda:0'), grad: tensor([-0.0262, -0.0266, 0.0129, -0.0086, 0.0148, 0.0150, 0.0203, 0.0123, + 0.0012, -0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 216.51, cls_loss 0.5139 cls_loss_mapping 0.0057 cls_loss_causal 0.4842 re_mapping 0.0064 re_causal 0.0161 /// teacc 98.88 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.0635, -0.0899, -0.1443, ..., 0.1182, -0.0526, 0.0329], + [-0.0822, 0.1047, -0.0639, ..., 0.0485, -0.0571, -0.0989], + [-0.0365, 0.0846, -0.1077, ..., -0.0182, 0.0366, 0.1257], + ..., + [-0.0061, 0.0622, -0.0051, ..., -0.0170, 0.0288, -0.0650], + [ 0.0897, -0.1031, 0.0420, ..., 0.0108, 0.0134, -0.0399], + [-0.0148, -0.0755, 0.0743, ..., -0.0785, 0.0261, -0.0318]], + device='cuda:0'), grad: tensor([[ 4.1771e-04, 2.0742e-04, 1.6844e-04, ..., 1.8942e-04, + 1.9872e-04, 2.6155e-04], + [-2.3155e-03, 2.2888e-04, 1.0471e-03, ..., 5.8591e-05, + 1.7405e-03, -8.5306e-04], + [ 5.4026e-04, -4.3602e-03, 2.3472e-04, ..., 2.4211e-04, + 1.1021e-04, 9.5463e-04], + ..., + [ 1.5557e-04, 3.1338e-03, 5.1041e-03, ..., 9.5673e-03, + -2.4471e-03, 1.6701e-04], + [ 2.5010e-04, 1.2147e-04, -4.0245e-03, ..., 3.5191e-03, + -5.7793e-04, -2.8286e-03], + [ 1.1361e-04, -3.6945e-03, -5.0049e-03, ..., -1.1658e-02, + -6.8855e-04, 4.4274e-04]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0129, -0.0084, -0.0141, 0.0100, -0.0160, -0.0207, 0.0129, 0.0005, + 0.0166, -0.0117], device='cuda:0'), grad: tensor([ 0.0109, -0.0699, 0.0015, 0.0365, 0.0319, -0.0377, 0.0157, 0.0194, + -0.0113, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 216.95, cls_loss 0.5009 cls_loss_mapping 0.0045 cls_loss_causal 0.4738 re_mapping 0.0068 re_causal 0.0177 /// teacc 98.63 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.0638, -0.0910, -0.1449, ..., 0.1176, -0.0524, 0.0322], + [-0.0826, 0.1062, -0.0640, ..., 0.0496, -0.0578, -0.0992], + [-0.0366, 0.0859, -0.1078, ..., -0.0176, 0.0380, 0.1267], + ..., + [-0.0069, 0.0600, -0.0051, ..., -0.0200, 0.0294, -0.0668], + [ 0.0901, -0.1032, 0.0414, ..., 0.0114, 0.0134, -0.0404], + [-0.0143, -0.0759, 0.0752, ..., -0.0783, 0.0259, -0.0316]], + device='cuda:0'), grad: tensor([[ 2.8396e-04, 1.6069e-04, 2.1954e-03, ..., 2.8229e-03, + 2.7924e-03, 2.7966e-04], + [ 3.6025e-04, 1.6193e-03, 1.4603e-04, ..., 2.1210e-03, + 3.7670e-04, 4.3720e-05], + [ 2.6202e-04, 1.5438e-04, 2.5320e-04, ..., -3.3436e-03, + 9.4032e-04, -2.3155e-03], + ..., + [ 4.1008e-04, -4.1542e-03, 1.4238e-05, ..., -5.2681e-03, + -8.9979e-04, 4.1270e-04], + [ 3.8171e-04, 3.9983e-04, 1.1883e-03, ..., 3.9139e-03, + 1.7719e-03, 1.2951e-03], + [ 3.5429e-04, 9.7942e-04, -4.7989e-03, ..., -3.3455e-03, + -5.3825e-03, 3.8207e-05]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0127, -0.0070, -0.0140, 0.0100, -0.0159, -0.0201, 0.0131, -0.0001, + 0.0155, -0.0119], device='cuda:0'), grad: tensor([ 0.0252, 0.0288, -0.0238, 0.0146, 0.0142, 0.0174, -0.0310, -0.0585, + 0.0242, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 216.84, cls_loss 0.4996 cls_loss_mapping 0.0060 cls_loss_causal 0.4732 re_mapping 0.0067 re_causal 0.0169 /// teacc 98.81 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.0641, -0.0895, -0.1451, ..., 0.1177, -0.0523, 0.0324], + [-0.0819, 0.1056, -0.0638, ..., 0.0475, -0.0581, -0.0995], + [-0.0356, 0.0852, -0.1093, ..., -0.0188, 0.0382, 0.1264], + ..., + [-0.0088, 0.0616, -0.0045, ..., -0.0186, 0.0293, -0.0646], + [ 0.0905, -0.1050, 0.0411, ..., 0.0115, 0.0135, -0.0410], + [-0.0133, -0.0761, 0.0746, ..., -0.0776, 0.0260, -0.0318]], + device='cuda:0'), grad: tensor([[ 4.6015e-05, 5.5075e-05, 6.1214e-05, ..., 2.1279e-04, + 2.5511e-04, 6.8521e-04], + [ 2.1651e-05, -1.2589e-03, 1.4696e-03, ..., -2.1229e-03, + -2.2373e-03, -9.4271e-04], + [-6.1512e-05, 5.3930e-04, 3.6764e-04, ..., 9.7561e-04, + 9.8991e-04, 1.2150e-03], + ..., + [ 2.4483e-05, -2.3484e-04, -1.9913e-03, ..., 1.0407e-04, + 6.5994e-04, 8.0585e-04], + [ 7.6652e-05, -4.6277e-04, -4.1342e-04, ..., -6.5613e-04, + -6.1131e-04, -9.5129e-04], + [ 2.0444e-05, 2.3937e-04, 1.1360e-02, ..., 3.7766e-04, + 7.7629e-04, 6.7329e-04]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0126, -0.0082, -0.0135, 0.0096, -0.0162, -0.0206, 0.0144, -0.0002, + 0.0167, -0.0124], device='cuda:0'), grad: tensor([ 0.0153, -0.0421, 0.0242, -0.0236, 0.0204, -0.0420, 0.0072, 0.0169, + 0.0179, 0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 216.58, cls_loss 0.4890 cls_loss_mapping 0.0028 cls_loss_causal 0.4649 re_mapping 0.0065 re_causal 0.0170 /// teacc 99.00 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.0641, -0.0892, -0.1436, ..., 0.1183, -0.0526, 0.0329], + [-0.0812, 0.1054, -0.0644, ..., 0.0475, -0.0576, -0.0998], + [-0.0360, 0.0855, -0.1093, ..., -0.0192, 0.0389, 0.1263], + ..., + [-0.0087, 0.0615, -0.0044, ..., -0.0183, 0.0292, -0.0650], + [ 0.0903, -0.1050, 0.0408, ..., 0.0103, 0.0126, -0.0408], + [-0.0143, -0.0767, 0.0738, ..., -0.0779, 0.0260, -0.0320]], + device='cuda:0'), grad: tensor([[ 1.7524e-04, 8.4750e-08, 2.6584e-04, ..., -3.6359e-04, + 1.7710e-03, 1.5295e-04], + [ 3.2306e-04, 4.7177e-05, -8.3923e-04, ..., -2.0485e-03, + 8.2135e-05, 3.0641e-06], + [ 1.4699e-04, 9.1362e-04, 1.4675e-04, ..., 2.3723e-04, + 4.4847e-04, 5.1230e-05], + ..., + [ 3.0112e-04, 1.2584e-05, 1.7796e-03, ..., 3.0184e-04, + 2.7447e-03, 3.0458e-05], + [ 6.0511e-04, -9.8515e-04, 1.6327e-03, ..., 1.1787e-03, + 3.7861e-03, 1.0853e-03], + [-3.0918e-03, 4.2841e-06, -6.0463e-03, ..., -1.8282e-03, + -2.0370e-02, -3.1319e-03]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0130, -0.0075, -0.0141, 0.0097, -0.0160, -0.0199, 0.0139, -0.0013, + 0.0171, -0.0126], device='cuda:0'), grad: tensor([ 0.0096, -0.0196, -0.0184, 0.0162, 0.0193, 0.0163, 0.0080, 0.0131, + 0.0151, -0.0598], device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 216.75, cls_loss 0.4899 cls_loss_mapping 0.0033 cls_loss_causal 0.4623 re_mapping 0.0069 re_causal 0.0176 /// teacc 98.87 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.0637, -0.0893, -0.1443, ..., 0.1208, -0.0542, 0.0309], + [-0.0805, 0.1062, -0.0640, ..., 0.0464, -0.0577, -0.1015], + [-0.0359, 0.0851, -0.1093, ..., -0.0189, 0.0382, 0.1269], + ..., + [-0.0083, 0.0623, -0.0045, ..., -0.0187, 0.0281, -0.0649], + [ 0.0917, -0.1062, 0.0410, ..., 0.0092, 0.0127, -0.0418], + [-0.0137, -0.0775, 0.0741, ..., -0.0777, 0.0270, -0.0306]], + device='cuda:0'), grad: tensor([[-4.6611e-04, 1.9646e-04, -1.6766e-03, ..., -5.8594e-03, + -1.9665e-03, -3.1719e-03], + [ 8.5950e-05, 7.3013e-03, 2.8300e-04, ..., 2.7809e-03, + 6.0797e-04, 3.3474e-04], + [ 6.2525e-05, -1.0204e-04, 1.8632e-04, ..., 1.0757e-03, + 2.8580e-05, 3.0935e-05], + ..., + [ 8.3148e-05, -7.3624e-03, 2.9492e-04, ..., 7.4530e-04, + 5.6171e-04, 3.4046e-04], + [-3.9935e-04, 1.6701e-04, 1.3840e-04, ..., -1.6556e-03, + -3.5095e-03, 1.7881e-04], + [ 1.1814e-04, 1.6141e-04, 2.1994e-04, ..., 9.8801e-04, + 5.1880e-04, 3.0780e-04]], device='cuda:0') +Epoch 302, bias, value: tensor([ 1.3077e-02, -8.3388e-03, -1.3799e-02, 8.9888e-03, -1.5679e-02, + -2.0570e-02, 1.3349e-02, 9.2770e-05, 1.7157e-02, -1.2048e-02], + device='cuda:0'), grad: tensor([-0.0685, 0.0411, 0.0112, 0.0240, -0.0131, 0.0342, -0.0064, -0.0041, + -0.0373, 0.0189], device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 216.62, cls_loss 0.4753 cls_loss_mapping 0.0039 cls_loss_causal 0.4461 re_mapping 0.0064 re_causal 0.0157 /// teacc 98.85 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.0635, -0.0901, -0.1443, ..., 0.1206, -0.0542, 0.0310], + [-0.0797, 0.1066, -0.0628, ..., 0.0462, -0.0572, -0.1013], + [-0.0365, 0.0849, -0.1086, ..., -0.0198, 0.0387, 0.1269], + ..., + [-0.0086, 0.0623, -0.0044, ..., -0.0179, 0.0276, -0.0648], + [ 0.0924, -0.1070, 0.0419, ..., 0.0087, 0.0127, -0.0419], + [-0.0134, -0.0763, 0.0739, ..., -0.0770, 0.0271, -0.0298]], + device='cuda:0'), grad: tensor([[ 2.1820e-03, 4.4793e-05, 1.6415e-04, ..., 1.8692e-04, + 2.3766e-03, 3.4466e-03], + [ 1.5056e-04, -3.9482e-03, -4.8676e-03, ..., -1.2497e-02, + 9.7007e-06, -2.9445e-04], + [ 6.4754e-04, 2.8515e-04, 1.1015e-04, ..., 3.8910e-04, + 2.6917e-04, 1.6069e-04], + ..., + [ 4.5347e-04, 3.8586e-03, 7.3776e-03, ..., 1.5007e-02, + 1.5656e-02, 5.5504e-03], + [-5.4359e-03, 1.0145e-04, 1.2064e-03, ..., 6.2656e-04, + -2.7752e-03, -4.2534e-03], + [ 2.9445e-04, 2.6321e-04, -6.4201e-03, ..., 2.0180e-03, + -3.3264e-03, 1.0198e-04]], device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0130, -0.0085, -0.0133, 0.0092, -0.0167, -0.0219, 0.0142, -0.0005, + 0.0176, -0.0109], device='cuda:0'), grad: tensor([ 0.0277, -0.0289, 0.0102, 0.0009, 0.0111, 0.0138, 0.0060, 0.0790, + -0.0912, -0.0285], device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 216.73, cls_loss 0.4891 cls_loss_mapping 0.0041 cls_loss_causal 0.4636 re_mapping 0.0068 re_causal 0.0167 /// teacc 98.82 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.0647, -0.0905, -0.1442, ..., 0.1203, -0.0536, 0.0318], + [-0.0793, 0.1077, -0.0632, ..., 0.0476, -0.0567, -0.1020], + [-0.0370, 0.0838, -0.1098, ..., -0.0199, 0.0383, 0.1280], + ..., + [-0.0087, 0.0633, -0.0049, ..., -0.0180, 0.0280, -0.0625], + [ 0.0926, -0.1070, 0.0414, ..., 0.0082, 0.0120, -0.0425], + [-0.0131, -0.0773, 0.0748, ..., -0.0782, 0.0271, -0.0313]], + device='cuda:0'), grad: tensor([[ 2.2009e-05, 5.3436e-05, 4.2558e-05, ..., 2.4244e-05, + 1.7509e-05, 4.5402e-07], + [-9.1248e-03, -1.6413e-03, -2.2354e-03, ..., -8.6365e-03, + 1.4588e-05, 3.4831e-07], + [ 7.6666e-06, 3.0661e-04, 1.2106e-04, ..., 3.8415e-05, + 8.6904e-05, -1.8207e-06], + ..., + [ 9.0694e-04, 2.2144e-03, 1.4143e-03, ..., 7.7009e-04, + -3.7479e-04, 8.7544e-08], + [ 1.7822e-04, 6.0648e-05, 2.3663e-04, ..., 1.6630e-04, + 1.9625e-05, 2.3525e-06], + [ 3.6774e-03, 3.6263e-04, 1.1921e-04, ..., 3.5038e-03, + 1.8048e-04, 2.0768e-07]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0122, -0.0069, -0.0137, 0.0097, -0.0167, -0.0218, 0.0136, -0.0001, + 0.0175, -0.0119], device='cuda:0'), grad: tensor([ 0.0048, -0.0105, 0.0063, 0.0068, 0.0049, -0.0029, -0.0244, -0.0151, + 0.0147, 0.0154], device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 216.64, cls_loss 0.5148 cls_loss_mapping 0.0042 cls_loss_causal 0.4857 re_mapping 0.0062 re_causal 0.0162 /// teacc 98.70 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.0628, -0.0906, -0.1443, ..., 0.1204, -0.0545, 0.0307], + [-0.0795, 0.1061, -0.0636, ..., 0.0475, -0.0576, -0.1025], + [-0.0363, 0.0826, -0.1094, ..., -0.0189, 0.0402, 0.1273], + ..., + [-0.0087, 0.0640, -0.0055, ..., -0.0187, 0.0263, -0.0624], + [ 0.0921, -0.1058, 0.0425, ..., 0.0088, 0.0123, -0.0426], + [-0.0124, -0.0762, 0.0749, ..., -0.0777, 0.0278, -0.0310]], + device='cuda:0'), grad: tensor([[ 0.0012, -0.0007, -0.0005, ..., 0.0004, -0.0031, -0.0009], + [-0.0036, -0.0031, -0.0011, ..., -0.0155, -0.0046, -0.0012], + [-0.0028, 0.0002, 0.0001, ..., 0.0005, 0.0010, 0.0002], + ..., + [ 0.0007, 0.0007, 0.0003, ..., 0.0006, 0.0019, 0.0002], + [ 0.0056, 0.0003, 0.0010, ..., 0.0106, 0.0008, 0.0007], + [ 0.0009, 0.0003, 0.0004, ..., 0.0006, 0.0007, 0.0002]], + device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0131, -0.0075, -0.0132, 0.0091, -0.0170, -0.0216, 0.0151, -0.0008, + 0.0166, -0.0115], device='cuda:0'), grad: tensor([-0.0435, -0.0765, 0.0096, -0.0053, 0.0199, -0.0131, 0.0049, 0.0287, + 0.0520, 0.0233], device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 216.86, cls_loss 0.4903 cls_loss_mapping 0.0029 cls_loss_causal 0.4656 re_mapping 0.0063 re_causal 0.0165 /// teacc 98.78 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.0611, -0.0911, -0.1446, ..., 0.1202, -0.0551, 0.0307], + [-0.0802, 0.1059, -0.0649, ..., 0.0484, -0.0580, -0.1027], + [-0.0361, 0.0829, -0.1093, ..., -0.0196, 0.0405, 0.1276], + ..., + [-0.0085, 0.0640, -0.0051, ..., -0.0184, 0.0260, -0.0626], + [ 0.0912, -0.1055, 0.0428, ..., 0.0086, 0.0129, -0.0422], + [-0.0126, -0.0760, 0.0756, ..., -0.0781, 0.0285, -0.0326]], + device='cuda:0'), grad: tensor([[ 1.8096e-06, 2.6729e-06, 4.4405e-05, ..., -7.9721e-06, + 1.5748e-04, 5.4032e-05], + [-4.7266e-05, -1.2684e-04, 1.2144e-05, ..., -6.9976e-05, + 1.1843e-04, 1.7560e-04], + [ 3.1620e-05, 1.7002e-05, 2.0474e-05, ..., 9.2238e-06, + 2.5487e-04, 1.2994e-04], + ..., + [-5.8323e-05, 1.6943e-05, 4.6468e-04, ..., 5.1707e-05, + 5.4359e-04, -1.0862e-03], + [ 1.2150e-03, 1.6853e-05, 3.5024e-04, ..., 7.9945e-06, + -3.6411e-03, 2.4104e-04], + [ 1.6928e-04, 5.5045e-05, -2.7657e-03, ..., 1.1064e-06, + -9.9182e-04, 7.5996e-05]], device='cuda:0') +Epoch 306, bias, value: tensor([ 0.0133, -0.0078, -0.0132, 0.0101, -0.0169, -0.0213, 0.0146, -0.0010, + 0.0158, -0.0113], device='cuda:0'), grad: tensor([-0.0140, 0.0281, -0.0704, -0.0020, 0.0224, 0.0209, -0.0078, -0.0019, + 0.0042, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 216.51, cls_loss 0.4929 cls_loss_mapping 0.0051 cls_loss_causal 0.4695 re_mapping 0.0061 re_causal 0.0158 /// teacc 98.87 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.0618, -0.0915, -0.1443, ..., 0.1209, -0.0539, 0.0317], + [-0.0795, 0.1057, -0.0652, ..., 0.0477, -0.0584, -0.1033], + [-0.0372, 0.0832, -0.1089, ..., -0.0184, 0.0407, 0.1275], + ..., + [-0.0068, 0.0636, -0.0046, ..., -0.0195, 0.0264, -0.0631], + [ 0.0920, -0.1068, 0.0434, ..., 0.0077, 0.0132, -0.0423], + [-0.0135, -0.0759, 0.0751, ..., -0.0774, 0.0278, -0.0328]], + device='cuda:0'), grad: tensor([[ 4.5588e-07, 1.0297e-05, 1.6659e-05, ..., 5.0735e-04, + 1.1122e-04, 2.1696e-04], + [ 5.4762e-06, 2.1422e-04, 2.5854e-05, ..., 9.2411e-04, + 9.7215e-05, 5.1117e-04], + [ 6.4727e-07, 7.5340e-04, 3.3259e-05, ..., 6.5088e-04, + 9.5367e-05, 2.8610e-04], + ..., + [ 1.2800e-05, -1.0624e-03, -1.5182e-03, ..., -2.0351e-03, + 6.8903e-05, -1.4362e-03], + [ 2.6003e-06, 1.1154e-05, 1.9228e-04, ..., 5.8174e-04, + 1.1832e-04, 2.6846e-04], + [ 1.0155e-05, 2.7835e-05, -2.0409e-04, ..., 4.5538e-04, + 8.6904e-05, 2.0134e-04]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0133, -0.0076, -0.0125, 0.0108, -0.0167, -0.0219, 0.0134, 0.0005, + 0.0151, -0.0123], device='cuda:0'), grad: tensor([-0.0199, -0.0088, 0.0161, 0.0084, -0.0122, -0.0140, 0.0196, -0.0202, + 0.0181, 0.0129], device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 216.53, cls_loss 0.5174 cls_loss_mapping 0.0034 cls_loss_causal 0.5012 re_mapping 0.0063 re_causal 0.0165 /// teacc 98.77 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.0623, -0.0923, -0.1450, ..., 0.1196, -0.0540, 0.0322], + [-0.0803, 0.1058, -0.0654, ..., 0.0482, -0.0584, -0.1021], + [-0.0376, 0.0832, -0.1093, ..., -0.0173, 0.0398, 0.1274], + ..., + [-0.0069, 0.0640, -0.0044, ..., -0.0194, 0.0270, -0.0640], + [ 0.0920, -0.1076, 0.0440, ..., 0.0082, 0.0135, -0.0419], + [-0.0146, -0.0756, 0.0752, ..., -0.0779, 0.0277, -0.0339]], + device='cuda:0'), grad: tensor([[ 3.7622e-04, 9.0170e-04, 5.9575e-05, ..., 1.7822e-02, + 3.6987e-02, 1.3412e-02], + [ 8.3447e-05, 1.2267e-04, 1.2028e-04, ..., 1.7023e-03, + 9.5654e-04, 1.1183e-05], + [ 1.1301e-04, 2.8872e-04, 1.6260e-04, ..., 1.6632e-03, + 1.1644e-03, 1.4124e-03], + ..., + [-7.7391e-04, -1.7481e-03, 1.1740e-03, ..., 1.9855e-03, + 2.0046e-03, 3.6985e-05], + [ 1.2222e-02, 7.7784e-05, -2.2831e-03, ..., 3.0823e-03, + -5.8975e-03, -3.1166e-03], + [ 7.7546e-05, 9.2363e-04, 1.3206e-02, ..., -3.1738e-02, + -2.8168e-02, -1.3237e-02]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0131, -0.0072, -0.0127, 0.0107, -0.0174, -0.0225, 0.0137, 0.0004, + 0.0161, -0.0120], device='cuda:0'), grad: tensor([ 0.0442, 0.0101, 0.0140, 0.0322, -0.0180, -0.0171, -0.0216, 0.0085, + 0.0002, -0.0525], device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 216.58, cls_loss 0.4888 cls_loss_mapping 0.0039 cls_loss_causal 0.4609 re_mapping 0.0064 re_causal 0.0168 /// teacc 98.80 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.0623, -0.0931, -0.1426, ..., 0.1203, -0.0554, 0.0315], + [-0.0804, 0.1058, -0.0650, ..., 0.0510, -0.0589, -0.1034], + [-0.0379, 0.0837, -0.1097, ..., -0.0182, 0.0404, 0.1277], + ..., + [-0.0070, 0.0641, -0.0043, ..., -0.0190, 0.0274, -0.0647], + [ 0.0928, -0.1082, 0.0440, ..., 0.0077, 0.0131, -0.0430], + [-0.0151, -0.0764, 0.0747, ..., -0.0777, 0.0290, -0.0301]], + device='cuda:0'), grad: tensor([[ 9.0539e-05, 3.4332e-05, -2.5439e-04, ..., -6.6280e-04, + 3.0875e-04, 1.2045e-03], + [ 5.7787e-05, -6.1572e-05, 5.8556e-04, ..., 1.2388e-03, + 1.9324e-04, 5.4979e-04], + [ 9.3639e-05, -2.1890e-05, 1.3709e-04, ..., 9.2983e-04, + 2.3842e-04, -5.2810e-05], + ..., + [-6.1846e-04, 2.5183e-05, 9.8228e-04, ..., 1.2579e-03, + 1.8120e-04, -1.6451e-04], + [-1.2383e-05, 1.7494e-05, 2.3150e-04, ..., 8.4543e-04, + 2.0361e-04, 6.7377e-04], + [ 4.3297e-04, 6.5684e-05, 1.2941e-03, ..., -1.0526e-04, + 2.1291e-04, -3.1681e-03]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0133, -0.0077, -0.0133, 0.0109, -0.0179, -0.0219, 0.0136, 0.0006, + 0.0162, -0.0119], device='cuda:0'), grad: tensor([ 0.0300, 0.0056, 0.0019, -0.0161, 0.0170, -0.0268, -0.0271, -0.0184, + 0.0318, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 216.69, cls_loss 0.5108 cls_loss_mapping 0.0049 cls_loss_causal 0.4886 re_mapping 0.0062 re_causal 0.0160 /// teacc 98.63 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.0629, -0.0923, -0.1433, ..., 0.1193, -0.0554, 0.0316], + [-0.0820, 0.1061, -0.0652, ..., 0.0500, -0.0589, -0.1037], + [-0.0379, 0.0834, -0.1098, ..., -0.0189, 0.0403, 0.1289], + ..., + [-0.0059, 0.0650, -0.0039, ..., -0.0191, 0.0278, -0.0638], + [ 0.0930, -0.1075, 0.0432, ..., 0.0078, 0.0129, -0.0427], + [-0.0147, -0.0772, 0.0739, ..., -0.0780, 0.0284, -0.0318]], + device='cuda:0'), grad: tensor([[-1.1530e-03, 5.8651e-05, -3.7823e-03, ..., -5.4884e-04, + -2.0485e-03, 3.5858e-04], + [ 4.3011e-04, 1.3981e-03, 9.7215e-05, ..., 1.1463e-03, + 5.1379e-05, 4.4298e-04], + [ 2.3520e-04, 4.3416e-04, 2.1070e-05, ..., 3.3855e-04, + -2.3067e-05, 7.6056e-04], + ..., + [ 3.9530e-04, 8.3542e-03, 5.0038e-05, ..., 1.1311e-03, + 4.2945e-05, 3.9124e-04], + [-1.7147e-03, -3.4962e-03, 3.6740e-04, ..., -6.4316e-03, + 2.0194e-04, 4.5133e-04], + [ 2.4772e-04, 3.5572e-04, 1.0595e-03, ..., 5.3358e-04, + 2.1720e-04, 3.8052e-04]], device='cuda:0') +Epoch 310, bias, value: tensor([ 0.0124, -0.0081, -0.0134, 0.0105, -0.0186, -0.0211, 0.0142, 0.0013, + 0.0174, -0.0125], device='cuda:0'), grad: tensor([-0.0194, -0.0075, -0.0145, -0.0042, 0.0117, -0.0464, 0.0331, 0.0417, + -0.0101, 0.0156], device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 216.80, cls_loss 0.4949 cls_loss_mapping 0.0043 cls_loss_causal 0.4642 re_mapping 0.0065 re_causal 0.0163 /// teacc 98.69 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.0627, -0.0914, -0.1444, ..., 0.1196, -0.0542, 0.0321], + [-0.0823, 0.1066, -0.0652, ..., 0.0504, -0.0597, -0.1038], + [-0.0372, 0.0843, -0.1102, ..., -0.0188, 0.0405, 0.1294], + ..., + [-0.0056, 0.0648, -0.0041, ..., -0.0192, 0.0266, -0.0638], + [ 0.0932, -0.1093, 0.0434, ..., 0.0085, 0.0128, -0.0436], + [-0.0148, -0.0779, 0.0735, ..., -0.0788, 0.0288, -0.0319]], + device='cuda:0'), grad: tensor([[-2.4382e-06, 9.6858e-05, 1.2875e-04, ..., 8.5926e-04, + 2.7418e-04, 6.9761e-04], + [ 1.2014e-07, 7.3671e-05, 5.5522e-05, ..., 4.4584e-04, + 1.2362e-04, 2.9898e-04], + [ 1.2154e-07, 9.8038e-04, 8.3447e-05, ..., 4.6992e-04, + 3.2558e-03, 2.2774e-03], + ..., + [-9.3937e-04, -4.0936e-04, 4.2707e-05, ..., 1.3985e-05, + 5.9462e-04, 4.5419e-04], + [ 1.6928e-05, 1.5759e-04, 4.4513e-04, ..., -1.9288e-04, + 8.2350e-04, -7.8008e-06], + [ 8.3923e-04, 5.6028e-04, 8.3876e-04, ..., 9.5749e-04, + 8.2922e-04, 4.6039e-04]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0126, -0.0086, -0.0140, 0.0095, -0.0182, -0.0197, 0.0152, 0.0020, + 0.0160, -0.0127], device='cuda:0'), grad: tensor([-0.0127, 0.0151, 0.0289, -0.0077, -0.0732, 0.0524, -0.0046, 0.0183, + -0.0092, -0.0074], device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 216.76, cls_loss 0.4573 cls_loss_mapping 0.0035 cls_loss_causal 0.4312 re_mapping 0.0065 re_causal 0.0157 /// teacc 98.79 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.0638, -0.0910, -0.1451, ..., 0.1201, -0.0544, 0.0313], + [-0.0834, 0.1075, -0.0654, ..., 0.0501, -0.0601, -0.1031], + [-0.0380, 0.0837, -0.1107, ..., -0.0188, 0.0399, 0.1299], + ..., + [-0.0059, 0.0662, -0.0031, ..., -0.0198, 0.0284, -0.0643], + [ 0.0943, -0.1100, 0.0437, ..., 0.0080, 0.0132, -0.0427], + [-0.0144, -0.0793, 0.0731, ..., -0.0786, 0.0279, -0.0311]], + device='cuda:0'), grad: tensor([[ 1.2910e-04, 2.2966e-06, 9.5606e-05, ..., 8.3590e-04, + 6.6185e-04, 1.1148e-03], + [ 7.9811e-05, 5.2834e-04, 6.3610e-04, ..., 1.5283e-04, + 4.3440e-04, 5.2166e-04], + [-1.6308e-03, 1.8415e-03, -1.8778e-03, ..., 6.4182e-04, + -4.3030e-03, -6.4659e-04], + ..., + [-1.3018e-03, -2.1286e-03, -1.4944e-03, ..., -3.8090e-03, + -1.8406e-03, -9.4271e-04], + [ 9.8515e-04, 2.1487e-05, 6.9332e-04, ..., 1.4181e-03, + 4.7951e-03, 3.1624e-03], + [ 5.2500e-04, 2.2364e-04, 1.9045e-03, ..., 4.8494e-04, + 9.3174e-04, 5.0688e-04]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0124, -0.0080, -0.0131, 0.0086, -0.0176, -0.0203, 0.0143, 0.0021, + 0.0163, -0.0126], device='cuda:0'), grad: tensor([ 0.0182, 0.0188, -0.0095, 0.0218, 0.0107, -0.0113, 0.0031, -0.0742, + 0.0526, -0.0302], device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 217.04, cls_loss 0.4748 cls_loss_mapping 0.0043 cls_loss_causal 0.4400 re_mapping 0.0064 re_causal 0.0164 /// teacc 98.84 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.0656, -0.0918, -0.1449, ..., 0.1202, -0.0534, 0.0321], + [-0.0832, 0.1081, -0.0655, ..., 0.0508, -0.0590, -0.1037], + [-0.0395, 0.0835, -0.1110, ..., -0.0178, 0.0388, 0.1300], + ..., + [-0.0045, 0.0657, -0.0030, ..., -0.0200, 0.0282, -0.0640], + [ 0.0936, -0.1102, 0.0445, ..., 0.0076, 0.0136, -0.0430], + [-0.0140, -0.0792, 0.0733, ..., -0.0791, 0.0277, -0.0316]], + device='cuda:0'), grad: tensor([[-1.7059e-04, 1.7926e-05, 1.7321e-04, ..., -1.0719e-03, + 1.4007e-04, -1.0099e-03], + [ 9.9182e-05, 1.2003e-05, 7.1585e-05, ..., -2.0385e-04, + 2.6870e-04, 2.1207e-04], + [ 9.5546e-05, -3.4332e-04, 1.0961e-04, ..., 9.0003e-05, + -4.9133e-03, 2.2769e-04], + ..., + [-1.2617e-03, -1.8311e-03, -1.9207e-03, ..., 2.1195e-04, + -7.3738e-03, 2.0897e-04], + [ 1.4725e-03, 1.1581e-04, 1.4949e-04, ..., 9.8801e-04, + 1.0127e-04, -9.3579e-05], + [-2.1534e-03, 2.8968e-04, 5.0163e-04, ..., 1.3661e-04, + 1.9608e-03, 1.5354e-04]], device='cuda:0') +Epoch 313, bias, value: tensor([ 0.0129, -0.0081, -0.0123, 0.0104, -0.0177, -0.0210, 0.0133, 0.0021, + 0.0151, -0.0126], device='cuda:0'), grad: tensor([-0.0151, 0.0168, -0.0331, 0.0168, 0.0138, 0.0156, 0.0130, -0.0354, + -0.0120, 0.0197], device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 216.47, cls_loss 0.4993 cls_loss_mapping 0.0037 cls_loss_causal 0.4727 re_mapping 0.0061 re_causal 0.0157 /// teacc 98.84 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.0666, -0.0924, -0.1451, ..., 0.1199, -0.0545, 0.0316], + [-0.0802, 0.1069, -0.0652, ..., 0.0501, -0.0595, -0.1040], + [-0.0387, 0.0843, -0.1111, ..., -0.0160, 0.0399, 0.1303], + ..., + [-0.0044, 0.0657, -0.0029, ..., -0.0181, 0.0288, -0.0645], + [ 0.0929, -0.1103, 0.0440, ..., 0.0084, 0.0133, -0.0415], + [-0.0143, -0.0774, 0.0738, ..., -0.0799, 0.0283, -0.0313]], + device='cuda:0'), grad: tensor([[ 1.5736e-04, 8.1480e-05, 8.1122e-05, ..., -1.6680e-03, + -1.5898e-03, -9.5081e-04], + [ 2.1019e-03, 2.0370e-03, 4.0078e-04, ..., 5.3644e-04, + 2.0874e-04, 1.0498e-05], + [-8.5592e-04, -3.2711e-03, 7.5281e-05, ..., -6.4611e-04, + 2.0981e-05, -8.4043e-05], + ..., + [ 1.0777e-04, 8.2397e-04, 8.5402e-04, ..., -8.0681e-04, + 4.8208e-04, 3.9451e-06], + [ 6.2656e-04, 8.5056e-05, 1.1086e-04, ..., 2.7466e-04, + 1.3185e-04, 5.8204e-05], + [ 8.5711e-05, 3.8594e-05, -6.3599e-02, ..., 5.0640e-04, + -3.7598e-02, 1.4238e-05]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0124, -0.0071, -0.0120, 0.0098, -0.0186, -0.0204, 0.0137, 0.0021, + 0.0147, -0.0125], device='cuda:0'), grad: tensor([-0.0926, 0.0304, 0.0126, 0.0071, 0.0249, -0.0162, 0.0288, -0.0076, + 0.0178, -0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 216.85, cls_loss 0.4768 cls_loss_mapping 0.0042 cls_loss_causal 0.4533 re_mapping 0.0064 re_causal 0.0162 /// teacc 98.72 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.0672, -0.0921, -0.1459, ..., 0.1207, -0.0553, 0.0314], + [-0.0809, 0.1067, -0.0639, ..., 0.0506, -0.0586, -0.1038], + [-0.0373, 0.0845, -0.1120, ..., -0.0155, 0.0386, 0.1311], + ..., + [-0.0051, 0.0653, -0.0036, ..., -0.0170, 0.0302, -0.0639], + [ 0.0938, -0.1101, 0.0447, ..., 0.0085, 0.0134, -0.0418], + [-0.0144, -0.0761, 0.0746, ..., -0.0805, 0.0282, -0.0314]], + device='cuda:0'), grad: tensor([[ 5.8937e-04, 2.1651e-05, 7.8557e-07, ..., -2.3127e-04, + -3.8433e-04, -4.0550e-03], + [ 4.0126e-04, 2.5539e-03, 2.4028e-06, ..., 4.4751e-04, + 1.2279e-04, 7.8821e-04], + [ 2.6870e-04, -2.8133e-03, 2.0247e-06, ..., -2.0099e-04, + 2.6536e-04, 1.2989e-03], + ..., + [ 6.1452e-05, 5.5492e-05, -1.6198e-05, ..., 1.1706e-04, + -2.7776e-04, 6.4659e-04], + [ 2.1973e-03, 1.7419e-05, 2.5351e-06, ..., 1.2910e-04, + 9.2447e-05, 6.3133e-04], + [ 3.7909e-04, 4.9084e-05, 4.6754e-04, ..., 1.6451e-04, + 4.2653e-04, 6.2895e-04]], device='cuda:0') +Epoch 315, bias, value: tensor([ 0.0118, -0.0063, -0.0124, 0.0101, -0.0178, -0.0209, 0.0123, 0.0020, + 0.0158, -0.0127], device='cuda:0'), grad: tensor([-0.0168, 0.0339, -0.0118, -0.0102, 0.0102, -0.0124, -0.0315, 0.0200, + -0.0036, 0.0223], device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 216.89, cls_loss 0.4751 cls_loss_mapping 0.0031 cls_loss_causal 0.4526 re_mapping 0.0065 re_causal 0.0171 /// teacc 98.84 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.0675, -0.0927, -0.1465, ..., 0.1214, -0.0557, 0.0319], + [-0.0811, 0.1077, -0.0643, ..., 0.0505, -0.0596, -0.1057], + [-0.0374, 0.0840, -0.1125, ..., -0.0149, 0.0397, 0.1306], + ..., + [-0.0050, 0.0654, -0.0042, ..., -0.0175, 0.0290, -0.0641], + [ 0.0939, -0.1100, 0.0466, ..., 0.0086, 0.0144, -0.0417], + [-0.0150, -0.0755, 0.0733, ..., -0.0810, 0.0270, -0.0310]], + device='cuda:0'), grad: tensor([[ 5.6982e-05, 2.0992e-06, -8.3968e-06, ..., 1.1625e-03, + 3.9458e-04, 8.3685e-04], + [ 1.8501e-03, -4.0866e-06, 1.5602e-05, ..., -1.2245e-02, + 3.1781e-04, 5.4455e-04], + [-2.2068e-03, 6.3956e-05, 1.5998e-04, ..., 3.5934e-03, + 9.9468e-04, 1.2026e-03], + ..., + [ 8.9824e-05, -2.3878e-04, -5.8222e-04, ..., 1.2131e-03, + -3.1304e-04, 2.4557e-04], + [ 3.8314e-04, 6.4373e-05, 1.3804e-04, ..., 3.7689e-03, + 6.7282e-04, 8.9645e-04], + [ 7.2122e-05, 1.3970e-05, -6.8951e-04, ..., 7.6437e-04, + 4.3583e-04, 3.8505e-04]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0119, -0.0068, -0.0126, 0.0098, -0.0175, -0.0201, 0.0115, 0.0017, + 0.0156, -0.0117], device='cuda:0'), grad: tensor([ 0.0108, 0.0045, -0.0090, -0.0410, 0.0087, -0.0165, 0.0098, 0.0062, + 0.0190, 0.0075], device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 216.89, cls_loss 0.4790 cls_loss_mapping 0.0027 cls_loss_causal 0.4558 re_mapping 0.0065 re_causal 0.0171 /// teacc 98.78 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.0681, -0.0920, -0.1474, ..., 0.1208, -0.0568, 0.0313], + [-0.0814, 0.1074, -0.0650, ..., 0.0491, -0.0606, -0.1050], + [-0.0374, 0.0850, -0.1127, ..., -0.0149, 0.0399, 0.1302], + ..., + [-0.0044, 0.0653, -0.0047, ..., -0.0159, 0.0293, -0.0642], + [ 0.0939, -0.1100, 0.0466, ..., 0.0081, 0.0148, -0.0409], + [-0.0151, -0.0761, 0.0744, ..., -0.0807, 0.0275, -0.0307]], + device='cuda:0'), grad: tensor([[ 3.2596e-08, 1.3137e-04, 1.6618e-04, ..., 1.2751e-03, + 3.6645e-04, 4.9993e-06], + [ 9.3132e-10, -4.7994e-04, 3.0115e-05, ..., -2.7370e-03, + 4.0263e-05, 1.9837e-07], + [ 2.3283e-09, 3.9712e-06, 2.3156e-05, ..., 1.8194e-05, + 3.5077e-05, 8.5076e-07], + ..., + [ 4.6566e-10, 1.6654e-04, 4.2686e-03, ..., 9.6130e-04, + 3.8071e-03, 1.8626e-07], + [ 5.4017e-08, 1.7202e-04, 3.0651e-03, ..., 1.0544e-02, + 6.8779e-03, 5.2720e-05], + [ 0.0000e+00, 2.3395e-06, -4.5471e-03, ..., 1.8761e-05, + -4.0054e-03, 7.4273e-07]], device='cuda:0') +Epoch 317, bias, value: tensor([ 0.0120, -0.0059, -0.0132, 0.0101, -0.0175, -0.0204, 0.0123, 0.0014, + 0.0147, -0.0116], device='cuda:0'), grad: tensor([-0.0190, -0.0139, -0.0202, -0.0180, 0.0118, 0.0434, -0.0077, -0.0041, + 0.0306, -0.0030], device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 216.59, cls_loss 0.4903 cls_loss_mapping 0.0054 cls_loss_causal 0.4665 re_mapping 0.0066 re_causal 0.0163 /// teacc 98.77 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.0679, -0.0929, -0.1483, ..., 0.1200, -0.0567, 0.0317], + [-0.0809, 0.1097, -0.0648, ..., 0.0480, -0.0603, -0.1060], + [-0.0376, 0.0840, -0.1130, ..., -0.0142, 0.0406, 0.1307], + ..., + [-0.0038, 0.0645, -0.0032, ..., -0.0171, 0.0290, -0.0646], + [ 0.0931, -0.1111, 0.0467, ..., 0.0094, 0.0143, -0.0410], + [-0.0150, -0.0748, 0.0740, ..., -0.0804, 0.0270, -0.0292]], + device='cuda:0'), grad: tensor([[ 1.1533e-05, 1.0705e-04, 1.7273e-04, ..., 1.2407e-03, + 6.0558e-04, 1.0834e-03], + [ 1.0028e-05, 6.6566e-04, 2.0599e-03, ..., -4.2839e-03, + -1.8430e-04, -2.0561e-03], + [ 2.6766e-06, 4.0174e-04, 3.9220e-04, ..., 2.8515e-03, + 2.1992e-03, 2.6531e-03], + ..., + [ 3.2894e-06, 8.6641e-04, 2.4204e-03, ..., 1.1616e-03, + 7.5035e-03, 8.2445e-04], + [ 1.0622e-04, 2.2697e-04, 7.6914e-04, ..., 9.8801e-04, + 1.5316e-03, 8.0347e-04], + [ 8.1584e-06, -2.0943e-03, -9.5444e-03, ..., 1.5898e-03, + -9.8114e-03, 9.8133e-04]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0125, -0.0070, -0.0125, 0.0098, -0.0177, -0.0205, 0.0129, 0.0013, + 0.0146, -0.0112], device='cuda:0'), grad: tensor([ 0.0218, -0.0492, 0.0365, -0.0076, 0.0400, -0.0414, -0.0391, 0.0370, + 0.0253, -0.0233], device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 216.73, cls_loss 0.4648 cls_loss_mapping 0.0044 cls_loss_causal 0.4418 re_mapping 0.0062 re_causal 0.0156 /// teacc 98.74 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.0678, -0.0924, -0.1480, ..., 0.1199, -0.0562, 0.0322], + [-0.0807, 0.1090, -0.0651, ..., 0.0477, -0.0604, -0.1040], + [-0.0387, 0.0842, -0.1129, ..., -0.0133, 0.0420, 0.1305], + ..., + [-0.0047, 0.0657, -0.0030, ..., -0.0180, 0.0294, -0.0646], + [ 0.0935, -0.1113, 0.0460, ..., 0.0087, 0.0140, -0.0416], + [-0.0153, -0.0750, 0.0733, ..., -0.0805, 0.0271, -0.0297]], + device='cuda:0'), grad: tensor([[ 3.9116e-06, 1.2410e-04, 6.7592e-05, ..., 1.2379e-03, + 1.1450e-04, 2.6369e-04], + [ 4.6566e-08, 1.8847e-04, -4.3368e-04, ..., -2.8057e-03, + 1.6260e-04, -3.3617e-04], + [-6.0606e-04, -8.5211e-04, 7.5817e-05, ..., -8.6546e-05, + -4.0150e-04, -2.0008e-03], + ..., + [ 3.4762e-07, -6.4230e-04, -6.0797e-04, ..., 3.6120e-04, + -2.3115e-04, 3.0923e-04], + [ 1.0125e-05, 8.3089e-05, -2.0957e-04, ..., -8.8310e-04, + -6.1893e-04, -9.1410e-04], + [ 1.0692e-06, 9.0981e-04, 7.8964e-04, ..., 5.1880e-04, + 4.3607e-04, 3.2449e-04]], device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0140, -0.0065, -0.0130, 0.0090, -0.0175, -0.0211, 0.0135, 0.0013, + 0.0142, -0.0115], device='cuda:0'), grad: tensor([ 0.0222, -0.0355, 0.0089, -0.0371, 0.0061, 0.0167, 0.0091, -0.0061, + -0.0121, 0.0277], device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 216.98, cls_loss 0.5052 cls_loss_mapping 0.0041 cls_loss_causal 0.4691 re_mapping 0.0060 re_causal 0.0156 /// teacc 98.56 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.0677, -0.0930, -0.1486, ..., 0.1197, -0.0567, 0.0316], + [-0.0817, 0.1084, -0.0637, ..., 0.0497, -0.0590, -0.1036], + [-0.0383, 0.0855, -0.1137, ..., -0.0138, 0.0409, 0.1297], + ..., + [-0.0047, 0.0651, -0.0021, ..., -0.0186, 0.0300, -0.0636], + [ 0.0945, -0.1102, 0.0469, ..., 0.0084, 0.0137, -0.0420], + [-0.0154, -0.0739, 0.0722, ..., -0.0815, 0.0262, -0.0297]], + device='cuda:0'), grad: tensor([[ 2.0087e-05, 1.5748e-04, 7.1704e-05, ..., 4.0233e-05, + -3.4291e-06, 2.9039e-04], + [ 1.1194e-04, 4.7565e-04, 2.6211e-05, ..., 1.4067e-03, + 8.6939e-07, 3.0708e-04], + [ 4.3482e-05, -8.5211e-04, -5.7936e-04, ..., -2.6703e-04, + 1.3253e-06, -1.1168e-03], + ..., + [ 1.4877e-04, 2.1470e-04, 1.2314e-04, ..., 5.4151e-05, + -7.2969e-07, 4.5562e-04], + [ 5.4884e-04, 1.4770e-04, 1.2660e-04, ..., 2.2751e-02, + 8.0094e-07, 2.3544e-04], + [ 4.7607e-03, -9.4748e-04, 6.6137e-04, ..., -1.5724e-04, + -1.4639e-04, -1.5974e-03]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0133, -0.0060, -0.0131, 0.0100, -0.0176, -0.0211, 0.0134, 0.0003, + 0.0148, -0.0120], device='cuda:0'), grad: tensor([-0.0140, 0.0039, 0.0343, 0.0216, 0.0251, -0.0278, -0.0124, 0.0217, + 0.0457, -0.0980], device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 216.78, cls_loss 0.4693 cls_loss_mapping 0.0031 cls_loss_causal 0.4447 re_mapping 0.0064 re_causal 0.0167 /// teacc 98.68 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.0684, -0.0924, -0.1486, ..., 0.1189, -0.0568, 0.0318], + [-0.0822, 0.1085, -0.0630, ..., 0.0501, -0.0583, -0.1043], + [-0.0382, 0.0860, -0.1143, ..., -0.0135, 0.0399, 0.1286], + ..., + [-0.0047, 0.0649, -0.0032, ..., -0.0178, 0.0292, -0.0640], + [ 0.0955, -0.1101, 0.0462, ..., 0.0080, 0.0141, -0.0415], + [-0.0156, -0.0750, 0.0729, ..., -0.0826, 0.0270, -0.0295]], + device='cuda:0'), grad: tensor([[ 1.7866e-05, 1.5080e-05, 2.4056e-04, ..., 3.0208e-04, + 2.7680e-04, 9.5248e-05], + [ 5.1409e-07, -1.3673e-04, 1.7428e-04, ..., 3.2663e-04, + 2.0409e-04, 1.8740e-04], + [ 1.0826e-05, -5.1111e-05, 9.3699e-04, ..., 5.3406e-04, + -3.7861e-03, -2.1782e-03], + ..., + [ 1.2256e-06, -2.5630e-05, 6.7558e-03, ..., 5.2500e-04, + 7.7324e-03, 2.0480e-04], + [-1.2398e-04, 9.3937e-05, 1.4143e-03, ..., 2.8825e-04, + 3.2139e-03, 9.1648e-04], + [ 4.3124e-05, -8.5652e-05, -1.6541e-02, ..., 4.8733e-04, + -1.8341e-02, 9.7632e-05]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0132, -0.0054, -0.0132, 0.0101, -0.0179, -0.0212, 0.0141, 0.0007, + 0.0148, -0.0131], device='cuda:0'), grad: tensor([ 0.0115, 0.0169, -0.0221, 0.0260, -0.0106, -0.0165, 0.0120, 0.0249, + 0.0036, -0.0457], device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 216.64, cls_loss 0.4972 cls_loss_mapping 0.0044 cls_loss_causal 0.4718 re_mapping 0.0059 re_causal 0.0158 /// teacc 98.78 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.0694, -0.0920, -0.1478, ..., 0.1189, -0.0575, 0.0311], + [-0.0820, 0.1088, -0.0624, ..., 0.0505, -0.0585, -0.1042], + [-0.0376, 0.0859, -0.1138, ..., -0.0133, 0.0405, 0.1279], + ..., + [-0.0048, 0.0652, -0.0033, ..., -0.0175, 0.0294, -0.0633], + [ 0.0955, -0.1100, 0.0448, ..., 0.0085, 0.0123, -0.0411], + [-0.0152, -0.0756, 0.0745, ..., -0.0837, 0.0280, -0.0296]], + device='cuda:0'), grad: tensor([[ 9.5367e-05, 7.2233e-06, 4.6134e-04, ..., 6.3300e-05, + 5.7697e-04, 1.3471e-04], + [ 2.3675e-04, 3.4124e-06, 3.5572e-04, ..., 2.0310e-05, + 5.2643e-04, 6.3539e-05], + [ 2.9778e-04, 7.7561e-06, 1.4524e-03, ..., 7.9274e-05, + 2.3289e-03, 3.6144e-04], + ..., + [-2.9850e-03, -2.9683e-05, -8.7204e-03, ..., 6.8367e-05, + -7.1793e-03, 2.5940e-04], + [ 3.6478e-04, 7.6108e-06, 4.2267e-03, ..., 2.0817e-05, + 2.5940e-03, 2.8276e-04], + [ 1.0595e-03, 6.1452e-05, 6.1302e-03, ..., 3.6526e-04, + 6.1264e-03, 3.8218e-04]], device='cuda:0') +Epoch 322, bias, value: tensor([ 1.2812e-02, -5.0562e-03, -1.2922e-02, 1.0429e-02, -1.7921e-02, + -2.1455e-02, 1.4118e-02, -7.0244e-05, 1.4877e-02, -1.2803e-02], + device='cuda:0'), grad: tensor([ 0.0090, -0.0193, 0.0147, -0.0108, -0.0095, 0.0315, -0.0365, -0.0079, + -0.0011, 0.0299], device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 216.92, cls_loss 0.4914 cls_loss_mapping 0.0040 cls_loss_causal 0.4718 re_mapping 0.0058 re_causal 0.0156 /// teacc 98.76 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.0704, -0.0929, -0.1476, ..., 0.1175, -0.0586, 0.0306], + [-0.0816, 0.1086, -0.0630, ..., 0.0516, -0.0587, -0.1047], + [-0.0376, 0.0863, -0.1151, ..., -0.0131, 0.0408, 0.1284], + ..., + [-0.0054, 0.0643, -0.0042, ..., -0.0185, 0.0296, -0.0634], + [ 0.0964, -0.1101, 0.0449, ..., 0.0083, 0.0123, -0.0415], + [-0.0155, -0.0748, 0.0754, ..., -0.0840, 0.0281, -0.0289]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0003, 0.0004, ..., 0.0013, 0.0089, 0.0001], + [ 0.0020, -0.0011, 0.0007, ..., 0.0039, 0.0003, 0.0001], + [ 0.0002, -0.0035, -0.0063, ..., -0.0136, -0.0163, -0.0079], + ..., + [ 0.0001, 0.0030, -0.0031, ..., 0.0073, -0.0022, 0.0037], + [-0.0037, 0.0007, 0.0016, ..., 0.0058, 0.0074, 0.0037], + [ 0.0006, 0.0003, 0.0068, ..., 0.0012, -0.0004, 0.0003]], + device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0129, -0.0057, -0.0138, 0.0102, -0.0168, -0.0215, 0.0142, 0.0002, + 0.0155, -0.0130], device='cuda:0'), grad: tensor([ 0.0395, -0.0083, -0.0356, -0.0132, 0.0335, -0.0142, -0.0296, 0.0115, + -0.0143, 0.0308], device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 216.70, cls_loss 0.4931 cls_loss_mapping 0.0027 cls_loss_causal 0.4714 re_mapping 0.0058 re_causal 0.0158 /// teacc 98.83 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.0718, -0.0935, -0.1462, ..., 0.1177, -0.0558, 0.0301], + [-0.0820, 0.1083, -0.0629, ..., 0.0518, -0.0595, -0.1047], + [-0.0385, 0.0860, -0.1144, ..., -0.0127, 0.0403, 0.1288], + ..., + [-0.0052, 0.0651, -0.0050, ..., -0.0193, 0.0275, -0.0648], + [ 0.0957, -0.1096, 0.0448, ..., 0.0081, 0.0133, -0.0408], + [-0.0140, -0.0733, 0.0751, ..., -0.0831, 0.0276, -0.0292]], + device='cuda:0'), grad: tensor([[ 1.0365e-04, 3.9876e-05, -1.4620e-03, ..., 1.9369e-03, + 1.1110e-03, 1.1295e-04], + [ 7.9498e-06, 5.6553e-04, 6.4325e-04, ..., 4.6492e-04, + 4.8494e-04, 7.5161e-05], + [ 6.1207e-06, -1.3304e-03, 2.7180e-04, ..., -1.3447e-04, + 1.1617e-04, 1.7965e-04], + ..., + [ 7.7784e-06, 1.4210e-04, -2.6520e-02, ..., -4.7913e-03, + -2.1393e-02, -2.3413e-04], + [ 1.0006e-05, 5.7936e-05, 3.3226e-03, ..., 2.8086e-04, + 4.8332e-03, 1.0592e-04], + [ 1.7077e-05, 2.7776e-05, 2.1851e-02, ..., 1.0214e-03, + 1.3985e-02, 3.9428e-05]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0123, -0.0059, -0.0130, 0.0109, -0.0176, -0.0210, 0.0143, -0.0008, + 0.0147, -0.0121], device='cuda:0'), grad: tensor([-0.0220, 0.0064, -0.0012, -0.0017, 0.0056, 0.0082, 0.0061, -0.0409, + 0.0107, 0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 216.76, cls_loss 0.4787 cls_loss_mapping 0.0029 cls_loss_causal 0.4566 re_mapping 0.0059 re_causal 0.0155 /// teacc 98.74 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.0719, -0.0964, -0.1467, ..., 0.1178, -0.0562, 0.0308], + [-0.0791, 0.1079, -0.0639, ..., 0.0516, -0.0609, -0.1049], + [-0.0380, 0.0862, -0.1157, ..., -0.0130, 0.0398, 0.1274], + ..., + [-0.0062, 0.0651, -0.0043, ..., -0.0193, 0.0281, -0.0637], + [ 0.0950, -0.1104, 0.0449, ..., 0.0095, 0.0136, -0.0397], + [-0.0146, -0.0743, 0.0748, ..., -0.0832, 0.0275, -0.0283]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.9833e-05, -4.7607e-03, ..., -3.2902e-04, + -8.3618e-03, 5.4628e-05], + [ 0.0000e+00, 6.3539e-05, 8.2970e-05, ..., 7.5221e-05, + 1.4842e-04, 7.0155e-05], + [ 0.0000e+00, 7.8619e-05, 2.9850e-04, ..., 2.8205e-04, + 5.3453e-04, -5.0116e-04], + ..., + [ 4.6566e-10, -5.8842e-04, -6.2895e-04, ..., -9.3412e-04, + -1.2302e-03, 5.2691e-05], + [ 0.0000e+00, 5.8264e-05, 5.9795e-04, ..., 1.0407e-04, + 8.9121e-04, 4.9084e-05], + [ 0.0000e+00, 8.5533e-05, 9.4175e-04, ..., 1.8632e-04, + 1.9817e-03, 4.5121e-05]], device='cuda:0') +Epoch 325, bias, value: tensor([ 0.0130, -0.0057, -0.0130, 0.0096, -0.0165, -0.0203, 0.0129, -0.0010, + 0.0156, -0.0126], device='cuda:0'), grad: tensor([-0.0120, 0.0163, -0.0483, 0.0152, -0.0472, 0.0066, 0.0210, 0.0081, + 0.0228, 0.0177], device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 216.92, cls_loss 0.4860 cls_loss_mapping 0.0041 cls_loss_causal 0.4627 re_mapping 0.0057 re_causal 0.0148 /// teacc 98.69 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.0716, -0.0980, -0.1477, ..., 0.1186, -0.0565, 0.0314], + [-0.0782, 0.1078, -0.0647, ..., 0.0514, -0.0610, -0.1059], + [-0.0381, 0.0861, -0.1164, ..., -0.0134, 0.0402, 0.1273], + ..., + [-0.0067, 0.0654, -0.0045, ..., -0.0197, 0.0285, -0.0634], + [ 0.0949, -0.1098, 0.0452, ..., 0.0101, 0.0138, -0.0395], + [-0.0149, -0.0735, 0.0743, ..., -0.0833, 0.0273, -0.0288]], + device='cuda:0'), grad: tensor([[ 3.7432e-05, 1.4240e-06, 2.4680e-08, ..., -8.8692e-04, + -1.5819e-04, -3.0454e-07], + [ 1.6034e-05, 4.3702e-04, 3.4133e-07, ..., 4.1813e-05, + 1.3761e-05, 6.5658e-08], + [ 3.7074e-05, 1.3125e-04, 1.1129e-07, ..., 1.9789e-05, + 1.8477e-05, 1.1036e-07], + ..., + [ 1.1645e-05, -8.8835e-04, 5.7332e-06, ..., 1.1647e-04, + 2.1115e-05, 1.4435e-08], + [ 6.7115e-05, 8.2999e-06, 3.5875e-06, ..., 2.2018e-04, + 2.8640e-05, 1.1854e-05], + [ 2.6345e-05, 2.0123e-04, 3.1628e-06, ..., 1.7512e-04, + 3.3200e-05, 1.1036e-07]], device='cuda:0') +Epoch 326, bias, value: tensor([ 1.3754e-02, -6.2315e-03, -1.3279e-02, 9.4605e-03, -1.6676e-02, + -1.9688e-02, 1.2999e-02, -4.5535e-05, 1.5291e-02, -1.3381e-02], + device='cuda:0'), grad: tensor([ 0.0060, 0.0085, -0.0235, -0.0189, 0.0100, 0.0037, -0.0081, 0.0067, + 0.0087, 0.0071], device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 216.86, cls_loss 0.5050 cls_loss_mapping 0.0037 cls_loss_causal 0.4789 re_mapping 0.0063 re_causal 0.0162 /// teacc 98.88 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.0690, -0.0973, -0.1466, ..., 0.1185, -0.0553, 0.0321], + [-0.0784, 0.1076, -0.0656, ..., 0.0517, -0.0608, -0.1069], + [-0.0368, 0.0857, -0.1167, ..., -0.0131, 0.0400, 0.1278], + ..., + [-0.0059, 0.0656, -0.0054, ..., -0.0207, 0.0285, -0.0630], + [ 0.0960, -0.1102, 0.0448, ..., 0.0104, 0.0126, -0.0396], + [-0.0158, -0.0744, 0.0741, ..., -0.0824, 0.0276, -0.0296]], + device='cuda:0'), grad: tensor([[ 2.0456e-04, 1.9197e-03, -3.0446e-04, ..., 9.5444e-03, + 1.4248e-03, 3.1624e-03], + [ 6.8545e-07, 1.1358e-03, 8.3089e-05, ..., -7.0267e-03, + 2.6965e-04, 7.3576e-04], + [ 2.7671e-05, 1.4868e-03, 1.8096e-04, ..., 1.6832e-03, + 1.6708e-03, -5.0116e-04], + ..., + [ 3.1898e-07, 3.4475e-04, -2.5183e-05, ..., 1.0884e-04, + 4.5967e-04, 5.7697e-04], + [-4.2000e-03, -3.7632e-03, 4.4018e-05, ..., -5.1575e-03, + -2.1763e-03, -3.1223e-03], + [ 7.0632e-06, 1.4248e-03, 1.7776e-02, ..., 7.9250e-04, + 9.4223e-03, 5.3835e-04]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0143, -0.0065, -0.0137, 0.0095, -0.0160, -0.0214, 0.0140, -0.0006, + 0.0163, -0.0138], device='cuda:0'), grad: tensor([ 0.0023, 0.0125, 0.0167, -0.0148, -0.0080, 0.0123, -0.0158, 0.0109, + -0.0599, 0.0439], device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 217.02, cls_loss 0.4670 cls_loss_mapping 0.0037 cls_loss_causal 0.4437 re_mapping 0.0062 re_causal 0.0165 /// teacc 98.73 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.0697, -0.0982, -0.1464, ..., 0.1193, -0.0569, 0.0317], + [-0.0788, 0.1091, -0.0663, ..., 0.0525, -0.0604, -0.1063], + [-0.0371, 0.0844, -0.1161, ..., -0.0134, 0.0391, 0.1273], + ..., + [-0.0053, 0.0657, -0.0043, ..., -0.0205, 0.0278, -0.0634], + [ 0.0956, -0.1111, 0.0438, ..., 0.0106, 0.0112, -0.0399], + [-0.0151, -0.0737, 0.0745, ..., -0.0831, 0.0290, -0.0293]], + device='cuda:0'), grad: tensor([[ 3.1944e-07, 3.3528e-05, 2.7224e-05, ..., 5.5015e-05, + -2.8741e-06, 2.6718e-05], + [ 2.5797e-04, -4.9067e-04, 2.1696e-05, ..., -3.6359e-04, + -1.0735e-04, 1.7315e-05], + [ 4.1537e-06, 1.9014e-04, 2.9147e-05, ..., 2.9469e-04, + 4.9740e-05, 1.8626e-05], + ..., + [-4.4227e-04, -5.9032e-04, 4.4560e-04, ..., 2.2888e-04, + 2.6369e-04, 8.2135e-05], + [ 3.6368e-07, -1.3638e-04, 2.6077e-05, ..., -8.9121e-04, + 4.8876e-05, 2.8923e-05], + [ 2.0707e-04, 6.7329e-04, -4.1771e-04, ..., 2.9278e-04, + -3.5882e-04, 1.9550e-04]], device='cuda:0') +Epoch 328, bias, value: tensor([ 0.0148, -0.0064, -0.0136, 0.0093, -0.0166, -0.0221, 0.0140, -0.0003, + 0.0162, -0.0131], device='cuda:0'), grad: tensor([-0.0153, 0.0077, -0.0198, 0.0128, -0.0276, 0.0174, 0.0089, -0.0151, + 0.0122, 0.0187], device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 217.13, cls_loss 0.5025 cls_loss_mapping 0.0036 cls_loss_causal 0.4769 re_mapping 0.0062 re_causal 0.0169 /// teacc 98.71 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.0699, -0.0983, -0.1460, ..., 0.1196, -0.0562, 0.0317], + [-0.0799, 0.1094, -0.0650, ..., 0.0518, -0.0601, -0.1064], + [-0.0374, 0.0836, -0.1157, ..., -0.0126, 0.0395, 0.1272], + ..., + [-0.0039, 0.0656, -0.0042, ..., -0.0206, 0.0279, -0.0644], + [ 0.0960, -0.1086, 0.0429, ..., 0.0104, 0.0118, -0.0398], + [-0.0165, -0.0743, 0.0747, ..., -0.0843, 0.0278, -0.0288]], + device='cuda:0'), grad: tensor([[ 2.6155e-04, 3.9101e-04, 8.2076e-05, ..., 4.9171e-03, + 8.1491e-04, 1.7071e-03], + [-1.5392e-03, -5.2834e-03, -6.3248e-03, ..., -7.5817e-05, + -4.4899e-03, 6.8617e-04], + [-9.4271e-04, -8.7738e-04, -9.6485e-06, ..., 8.7070e-04, + -2.3222e-04, -4.4966e-04], + ..., + [ 8.8024e-04, 1.7567e-03, 1.1463e-03, ..., 1.2941e-03, + 1.3628e-03, 7.1049e-04], + [ 1.0681e-03, 8.8072e-04, 6.9916e-05, ..., 1.5430e-03, + 3.6693e-04, 8.9598e-04], + [ 5.1880e-04, 2.3422e-03, 3.2768e-03, ..., 8.8787e-04, + 2.9392e-03, 5.4741e-04]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0157, -0.0062, -0.0134, 0.0095, -0.0158, -0.0223, 0.0129, -0.0008, + 0.0154, -0.0127], device='cuda:0'), grad: tensor([ 0.0407, -0.0280, -0.0440, 0.0257, 0.0389, -0.0558, -0.0221, 0.0357, + 0.0273, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 217.03, cls_loss 0.4776 cls_loss_mapping 0.0031 cls_loss_causal 0.4526 re_mapping 0.0065 re_causal 0.0171 /// teacc 98.75 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.0700, -0.1004, -0.1454, ..., 0.1191, -0.0562, 0.0316], + [-0.0803, 0.1093, -0.0653, ..., 0.0534, -0.0601, -0.1044], + [-0.0381, 0.0841, -0.1154, ..., -0.0125, 0.0397, 0.1270], + ..., + [-0.0043, 0.0661, -0.0046, ..., -0.0208, 0.0276, -0.0661], + [ 0.0953, -0.1090, 0.0432, ..., 0.0099, 0.0123, -0.0409], + [-0.0154, -0.0754, 0.0732, ..., -0.0843, 0.0267, -0.0264]], + device='cuda:0'), grad: tensor([[-4.2343e-03, -1.2144e-05, 8.6576e-06, ..., 2.0561e-03, + 1.0910e-03, 1.0004e-03], + [ 1.1489e-05, 7.0520e-06, 3.6024e-06, ..., 4.8399e-04, + 3.7074e-04, 2.0087e-05], + [-2.1839e-04, -2.1458e-03, 1.7500e-04, ..., 8.5735e-04, + -1.2369e-03, -1.1406e-03], + ..., + [ 4.6134e-04, 2.0847e-03, 1.8254e-05, ..., -5.0278e-03, + -1.7452e-03, 1.2407e-03], + [ 5.4359e-04, 6.3293e-06, 3.9488e-05, ..., 2.5215e-03, + 1.2684e-03, 1.2290e-04], + [-2.6181e-05, 1.8626e-05, -3.3398e-03, ..., 9.2745e-04, + -8.2922e-04, 1.6224e-04]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0163, -0.0057, -0.0133, 0.0100, -0.0160, -0.0224, 0.0134, -0.0012, + 0.0148, -0.0135], device='cuda:0'), grad: tensor([-0.0074, -0.0406, 0.0176, 0.0363, 0.0183, -0.0782, 0.0377, -0.0341, + 0.0292, 0.0212], device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 216.86, cls_loss 0.4952 cls_loss_mapping 0.0022 cls_loss_causal 0.4682 re_mapping 0.0066 re_causal 0.0186 /// teacc 98.70 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.0719, -0.1007, -0.1467, ..., 0.1187, -0.0570, 0.0313], + [-0.0796, 0.1099, -0.0661, ..., 0.0534, -0.0608, -0.1045], + [-0.0390, 0.0846, -0.1156, ..., -0.0125, 0.0396, 0.1266], + ..., + [-0.0044, 0.0656, -0.0033, ..., -0.0206, 0.0290, -0.0656], + [ 0.0965, -0.1084, 0.0422, ..., 0.0099, 0.0111, -0.0412], + [-0.0154, -0.0755, 0.0739, ..., -0.0828, 0.0274, -0.0259]], + device='cuda:0'), grad: tensor([[-4.7350e-04, 4.3571e-05, 2.6393e-04, ..., 6.8367e-05, + 2.6131e-04, 2.7623e-06], + [ 1.7548e-04, 9.2030e-05, 4.5562e-04, ..., 3.1501e-05, + 3.6836e-04, 3.1367e-06], + [-6.0320e-04, -1.6832e-03, 3.2783e-04, ..., -5.0724e-05, + 3.5024e-04, -1.2040e-04], + ..., + [ 5.2834e-04, 1.7920e-03, 3.4485e-03, ..., 2.0294e-03, + 7.9498e-03, 8.5592e-05], + [ 4.2129e-04, 6.1214e-05, 1.1787e-03, ..., 6.5947e-04, + 2.7046e-03, 8.1863e-07], + [ 2.6894e-04, -6.1750e-05, -4.9438e-03, ..., -3.2692e-03, + -1.3329e-02, 5.8627e-07]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0169, -0.0053, -0.0138, 0.0091, -0.0155, -0.0222, 0.0136, -0.0024, + 0.0148, -0.0128], device='cuda:0'), grad: tensor([ 0.0048, -0.0293, 0.0033, 0.0051, -0.0045, 0.0030, 0.0087, 0.0206, + 0.0087, -0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 216.80, cls_loss 0.4442 cls_loss_mapping 0.0035 cls_loss_causal 0.4215 re_mapping 0.0062 re_causal 0.0145 /// teacc 98.91 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.0721, -0.0999, -0.1450, ..., 0.1193, -0.0567, 0.0313], + [-0.0795, 0.1091, -0.0665, ..., 0.0527, -0.0594, -0.1047], + [-0.0378, 0.0847, -0.1165, ..., -0.0118, 0.0403, 0.1270], + ..., + [-0.0044, 0.0662, -0.0036, ..., -0.0212, 0.0284, -0.0655], + [ 0.0962, -0.1091, 0.0433, ..., 0.0095, 0.0108, -0.0425], + [-0.0153, -0.0767, 0.0736, ..., -0.0837, 0.0272, -0.0258]], + device='cuda:0'), grad: tensor([[-8.8336e-07, 7.6368e-08, 6.6698e-05, ..., 8.7309e-04, + -1.1339e-03, -3.6049e-03], + [ 1.3495e-06, 3.1162e-06, 6.2287e-05, ..., 1.1110e-03, + 1.3673e-04, 3.8338e-04], + [ 4.1462e-06, 9.3505e-06, 5.0217e-05, ..., -4.8981e-03, + 2.3210e-04, 5.6696e-04], + ..., + [-7.6517e-06, -1.7524e-05, 4.9400e-04, ..., 2.4915e-04, + 5.1212e-04, 3.2544e-04], + [ 1.2945e-07, 7.9162e-08, -7.8201e-05, ..., 3.3879e-04, + -2.5392e-04, -1.1802e-04], + [ 2.2296e-06, 4.2841e-06, -2.4509e-04, ..., 3.4833e-04, + -6.1655e-04, 4.9114e-04]], device='cuda:0') +Epoch 332, bias, value: tensor([ 0.0180, -0.0064, -0.0134, 0.0094, -0.0153, -0.0210, 0.0121, -0.0024, + 0.0150, -0.0135], device='cuda:0'), grad: tensor([-0.0141, -0.0154, -0.0181, 0.0133, 0.0115, -0.0025, 0.0140, 0.0119, + 0.0098, -0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 216.58, cls_loss 0.4881 cls_loss_mapping 0.0035 cls_loss_causal 0.4660 re_mapping 0.0058 re_causal 0.0156 /// teacc 98.84 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.0724, -0.1009, -0.1449, ..., 0.1188, -0.0569, 0.0304], + [-0.0802, 0.1086, -0.0670, ..., 0.0524, -0.0597, -0.1041], + [-0.0375, 0.0846, -0.1163, ..., -0.0111, 0.0409, 0.1269], + ..., + [-0.0063, 0.0671, -0.0027, ..., -0.0209, 0.0290, -0.0648], + [ 0.0955, -0.1086, 0.0422, ..., 0.0086, 0.0095, -0.0420], + [-0.0144, -0.0786, 0.0731, ..., -0.0849, 0.0268, -0.0259]], + device='cuda:0'), grad: tensor([[-7.1973e-06, 8.2731e-05, 4.1910e-08, ..., 1.4343e-02, + 2.2163e-03, 5.9128e-04], + [-1.8144e-04, -2.4242e-03, 2.2165e-07, ..., -2.1423e-02, + -3.0727e-03, 5.2229e-06], + [ 1.5691e-05, 6.5327e-04, 1.3784e-07, ..., 7.2908e-04, + 1.2941e-03, 1.2994e-04], + ..., + [ 2.5496e-05, 3.9029e-04, 1.8179e-05, ..., 8.5068e-03, + -1.1808e-04, 4.4405e-06], + [-5.1155e-03, 2.8825e-04, 1.4435e-06, ..., 1.7996e-03, + 3.2634e-05, -1.3673e-04], + [ 1.3165e-05, 2.9063e-04, -3.2306e-05, ..., -2.8992e-03, + 2.6536e-04, 8.5831e-06]], device='cuda:0') +Epoch 333, bias, value: tensor([ 0.0176, -0.0065, -0.0131, 0.0095, -0.0142, -0.0222, 0.0114, -0.0019, + 0.0161, -0.0143], device='cuda:0'), grad: tensor([ 0.0352, -0.0468, 0.0164, -0.0201, 0.0169, 0.0427, -0.0428, 0.0202, + -0.0047, -0.0171], device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 216.90, cls_loss 0.4991 cls_loss_mapping 0.0036 cls_loss_causal 0.4692 re_mapping 0.0062 re_causal 0.0160 /// teacc 98.89 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.0729, -0.0999, -0.1457, ..., 0.1204, -0.0564, 0.0291], + [-0.0791, 0.1084, -0.0673, ..., 0.0514, -0.0602, -0.1041], + [-0.0364, 0.0848, -0.1159, ..., -0.0097, 0.0415, 0.1277], + ..., + [-0.0052, 0.0677, -0.0036, ..., -0.0213, 0.0293, -0.0635], + [ 0.0948, -0.1083, 0.0428, ..., 0.0074, 0.0092, -0.0430], + [-0.0141, -0.0780, 0.0736, ..., -0.0854, 0.0270, -0.0258]], + device='cuda:0'), grad: tensor([[ 3.4925e-08, 5.9128e-04, 1.8549e-04, ..., 5.9986e-04, + 7.3528e-04, 7.4625e-04], + [ 4.1910e-09, 3.2592e-04, 3.4779e-05, ..., 3.4189e-04, + 1.2994e-04, 5.5885e-04], + [ 6.9849e-09, 9.6226e-04, 2.5535e-04, ..., 1.0929e-03, + 1.1234e-03, 1.2474e-03], + ..., + [ 6.2864e-08, 8.7440e-05, 8.1444e-04, ..., -4.8012e-05, + 1.5602e-03, 1.2627e-03], + [ 1.1437e-06, 4.7398e-04, 2.6679e-04, ..., 3.2926e-04, + 2.1136e-04, 5.3024e-04], + [ 1.3504e-08, -3.0537e-03, -8.2550e-03, ..., -3.6812e-03, + -7.1259e-03, -4.6654e-03]], device='cuda:0') +Epoch 334, bias, value: tensor([ 0.0171, -0.0054, -0.0116, 0.0086, -0.0155, -0.0214, 0.0102, -0.0016, + 0.0158, -0.0139], device='cuda:0'), grad: tensor([ 0.0147, 0.0131, 0.0212, 0.0098, 0.0253, -0.0105, -0.0030, -0.0354, + 0.0122, -0.0473], device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 216.73, cls_loss 0.4613 cls_loss_mapping 0.0046 cls_loss_causal 0.4367 re_mapping 0.0061 re_causal 0.0165 /// teacc 98.84 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.0740, -0.1008, -0.1457, ..., 0.1211, -0.0572, 0.0284], + [-0.0794, 0.1095, -0.0670, ..., 0.0509, -0.0603, -0.1035], + [-0.0372, 0.0832, -0.1149, ..., -0.0109, 0.0409, 0.1277], + ..., + [-0.0038, 0.0683, -0.0036, ..., -0.0201, 0.0297, -0.0629], + [ 0.0950, -0.1071, 0.0436, ..., 0.0075, 0.0095, -0.0418], + [-0.0146, -0.0803, 0.0741, ..., -0.0864, 0.0279, -0.0246]], + device='cuda:0'), grad: tensor([[ 2.5725e-04, 2.6226e-04, 2.1231e-04, ..., 5.1641e-04, + 2.4509e-04, 3.0175e-05], + [ 3.7694e-04, 5.3167e-04, 6.6423e-04, ..., 6.4039e-04, + 3.7169e-04, 1.9282e-05], + [-4.0054e-03, -1.3031e-02, 4.0460e-04, ..., 5.5695e-04, + 3.7599e-04, 1.5780e-05], + ..., + [ 5.7526e-03, 1.2550e-02, -1.6868e-04, ..., -1.6832e-04, + -1.7338e-03, 2.6256e-05], + [ 2.0206e-04, 1.4901e-04, 1.1909e-04, ..., 1.6189e-04, + 6.0034e-04, -7.9393e-05], + [ 7.9632e-04, 6.3944e-04, 1.3494e-03, ..., 1.3952e-03, + 7.9727e-04, 4.0650e-05]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0164, -0.0056, -0.0126, 0.0084, -0.0154, -0.0219, 0.0114, -0.0009, + 0.0157, -0.0132], device='cuda:0'), grad: tensor([ 0.0164, 0.0258, -0.0021, -0.0105, -0.0962, -0.0038, -0.0169, 0.0345, + 0.0238, 0.0290], device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 216.91, cls_loss 0.4702 cls_loss_mapping 0.0047 cls_loss_causal 0.4417 re_mapping 0.0057 re_causal 0.0152 /// teacc 98.76 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.0721, -0.1009, -0.1444, ..., 0.1206, -0.0578, 0.0282], + [-0.0799, 0.1091, -0.0671, ..., 0.0524, -0.0590, -0.1043], + [-0.0359, 0.0835, -0.1159, ..., -0.0118, 0.0405, 0.1279], + ..., + [-0.0059, 0.0686, -0.0041, ..., -0.0203, 0.0292, -0.0630], + [ 0.0957, -0.1076, 0.0465, ..., 0.0077, 0.0109, -0.0409], + [-0.0150, -0.0810, 0.0739, ..., -0.0870, 0.0277, -0.0247]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0002, 0.0015, ..., 0.0004, 0.0023, 0.0006], + [-0.0008, -0.0007, 0.0002, ..., -0.0013, 0.0003, -0.0002], + [ 0.0003, 0.0006, -0.0097, ..., 0.0006, -0.0059, -0.0048], + ..., + [ 0.0006, 0.0019, 0.0025, ..., 0.0020, 0.0053, 0.0009], + [ 0.0006, 0.0003, 0.0034, ..., 0.0011, 0.0056, 0.0003], + [-0.0007, -0.0002, -0.0194, ..., -0.0008, -0.0144, 0.0009]], + device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0163, -0.0055, -0.0129, 0.0090, -0.0146, -0.0210, 0.0116, -0.0009, + 0.0145, -0.0142], device='cuda:0'), grad: tensor([ 0.0126, 0.0020, -0.0492, 0.0085, 0.0056, 0.0076, 0.0048, 0.0347, + 0.0142, -0.0408], device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 216.85, cls_loss 0.4918 cls_loss_mapping 0.0038 cls_loss_causal 0.4686 re_mapping 0.0063 re_causal 0.0171 /// teacc 98.73 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.0701, -0.0998, -0.1434, ..., 0.1202, -0.0596, 0.0296], + [-0.0797, 0.1093, -0.0675, ..., 0.0522, -0.0589, -0.1030], + [-0.0366, 0.0831, -0.1158, ..., -0.0122, 0.0413, 0.1283], + ..., + [-0.0058, 0.0681, -0.0051, ..., -0.0206, 0.0288, -0.0638], + [ 0.0957, -0.1075, 0.0453, ..., 0.0080, 0.0098, -0.0414], + [-0.0161, -0.0805, 0.0742, ..., -0.0871, 0.0291, -0.0245]], + device='cuda:0'), grad: tensor([[ 3.3641e-04, 2.8104e-05, 3.0060e-03, ..., 3.1042e-04, + -5.0735e-04, 2.3925e-04], + [ 1.3363e-04, 1.3001e-05, 4.1485e-04, ..., 5.5695e-04, + 2.5916e-04, 9.5487e-05], + [-2.9392e-03, -3.4833e-04, -9.4032e-04, ..., 7.8201e-04, + -2.0847e-03, -2.3041e-03], + ..., + [ 9.2804e-05, 1.0476e-05, 2.7370e-04, ..., 5.7268e-04, + 1.6034e-04, 7.0512e-05], + [ 1.3189e-03, 1.5044e-04, 8.8978e-04, ..., 2.8396e-04, + 1.5364e-03, 1.0138e-03], + [ 1.6546e-04, 1.6987e-05, 3.2692e-03, ..., -2.5719e-05, + 9.8610e-04, 1.2255e-04]], device='cuda:0') +Epoch 337, bias, value: tensor([ 1.6065e-02, -6.0051e-03, -1.1745e-02, 8.0642e-03, -1.4190e-02, + -2.1090e-02, 1.1452e-02, -7.0563e-05, 1.3367e-02, -1.3616e-02], + device='cuda:0'), grad: tensor([ 0.0294, -0.0300, 0.0070, 0.0259, -0.0274, -0.0166, 0.0015, -0.0307, + 0.0330, 0.0080], device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 216.85, cls_loss 0.4838 cls_loss_mapping 0.0032 cls_loss_causal 0.4605 re_mapping 0.0060 re_causal 0.0154 /// teacc 98.84 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.0707, -0.0991, -0.1435, ..., 0.1212, -0.0592, 0.0301], + [-0.0799, 0.1080, -0.0655, ..., 0.0499, -0.0584, -0.1036], + [-0.0366, 0.0836, -0.1162, ..., -0.0127, 0.0419, 0.1282], + ..., + [-0.0064, 0.0687, -0.0038, ..., -0.0203, 0.0294, -0.0641], + [ 0.0958, -0.1081, 0.0453, ..., 0.0091, 0.0099, -0.0410], + [-0.0152, -0.0814, 0.0740, ..., -0.0880, 0.0289, -0.0244]], + device='cuda:0'), grad: tensor([[-5.1688e-08, 9.9372e-07, -2.1001e-07, ..., 7.2736e-07, + 2.5645e-05, 1.1632e-06], + [ 3.8475e-05, -1.9357e-05, 8.8383e-07, ..., -4.6611e-05, + 4.4703e-05, 5.9139e-08], + [ 5.9698e-07, 2.0176e-05, 6.0024e-07, ..., 3.2857e-06, + 4.9095e-03, -1.3504e-07], + ..., + [ 5.1335e-06, -6.6698e-05, -4.3632e-07, ..., 1.8865e-05, + -9.2697e-03, 9.4064e-08], + [-6.7651e-05, 3.8967e-06, 1.7518e-06, ..., 1.0379e-05, + 2.0561e-03, 3.3807e-07], + [ 1.2822e-05, 3.9130e-05, 3.1710e-04, ..., 7.0333e-06, + 1.2505e-04, 3.4459e-07]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0153, -0.0061, -0.0126, 0.0077, -0.0146, -0.0215, 0.0123, 0.0001, + 0.0139, -0.0124], device='cuda:0'), grad: tensor([ 0.0110, 0.0313, 0.0230, -0.0134, -0.0154, 0.0122, 0.0143, -0.0338, + -0.0144, -0.0149], device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 216.77, cls_loss 0.4675 cls_loss_mapping 0.0038 cls_loss_causal 0.4430 re_mapping 0.0061 re_causal 0.0157 /// teacc 98.94 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.0708, -0.0994, -0.1442, ..., 0.1214, -0.0603, 0.0311], + [-0.0798, 0.1081, -0.0658, ..., 0.0494, -0.0586, -0.1048], + [-0.0356, 0.0834, -0.1162, ..., -0.0137, 0.0420, 0.1272], + ..., + [-0.0075, 0.0694, -0.0037, ..., -0.0198, 0.0294, -0.0630], + [ 0.0960, -0.1083, 0.0462, ..., 0.0099, 0.0096, -0.0396], + [-0.0156, -0.0815, 0.0741, ..., -0.0873, 0.0295, -0.0233]], + device='cuda:0'), grad: tensor([[ 1.7583e-05, 3.4988e-05, 1.1034e-05, ..., 6.1846e-04, + 1.1669e-06, 1.6820e-04], + [ 4.2409e-05, -3.8195e-04, 5.1558e-05, ..., -4.4346e-05, + 4.3660e-06, 8.5652e-05], + [ 3.1143e-05, 3.8147e-05, 3.7611e-05, ..., 2.6131e-04, + 2.2694e-05, 5.7876e-05], + ..., + [ 7.9727e-04, 3.5793e-05, 2.3384e-03, ..., 1.8911e-03, + -1.5751e-05, 8.8736e-06], + [ 8.9228e-05, 3.4481e-05, 9.2983e-05, ..., 3.4714e-04, + 4.4060e-04, 8.9705e-05], + [ 3.9711e-03, 5.6982e-05, 2.2087e-03, ..., -1.8196e-03, + 5.1141e-05, 2.0131e-05]], device='cuda:0') +Epoch 339, bias, value: tensor([ 0.0151, -0.0051, -0.0125, 0.0072, -0.0159, -0.0207, 0.0127, -0.0003, + 0.0134, -0.0118], device='cuda:0'), grad: tensor([ 0.0092, -0.0207, 0.0069, -0.0232, -0.0058, 0.0081, 0.0241, 0.0140, + -0.0238, 0.0112], device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 216.50, cls_loss 0.4845 cls_loss_mapping 0.0039 cls_loss_causal 0.4583 re_mapping 0.0060 re_causal 0.0156 /// teacc 98.97 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.0713, -0.1006, -0.1451, ..., 0.1205, -0.0602, 0.0309], + [-0.0790, 0.1074, -0.0655, ..., 0.0496, -0.0570, -0.1051], + [-0.0358, 0.0825, -0.1180, ..., -0.0134, 0.0417, 0.1271], + ..., + [-0.0086, 0.0717, -0.0030, ..., -0.0201, 0.0300, -0.0634], + [ 0.0946, -0.1085, 0.0459, ..., 0.0092, 0.0088, -0.0396], + [-0.0148, -0.0821, 0.0742, ..., -0.0866, 0.0297, -0.0239]], + device='cuda:0'), grad: tensor([[ 4.2230e-05, 2.2456e-05, -2.2635e-05, ..., 3.8123e-04, + -1.3852e-04, -3.4046e-04], + [-3.6168e-04, -2.2840e-04, 1.4007e-04, ..., 3.9577e-04, + 6.0606e-04, 2.8443e-04], + [ 3.6316e-03, 1.1377e-03, 1.7262e-04, ..., 5.3644e-04, + 2.6588e-03, 2.6379e-03], + ..., + [ 4.1924e-03, 4.6879e-05, 2.4223e-03, ..., 4.0364e-04, + 2.5387e-03, 2.3448e-04], + [ 3.1872e-03, 9.3317e-04, 3.2353e-04, ..., 6.3038e-04, + 3.3379e-03, 1.0490e-03], + [-3.5187e-02, 1.1432e-04, -6.3095e-03, ..., 3.6716e-04, + -1.1192e-02, 1.9467e-04]], device='cuda:0') +Epoch 340, bias, value: tensor([ 1.4479e-02, -5.3744e-03, -1.2296e-02, 7.1329e-03, -1.4824e-02, + -2.1448e-02, 1.2626e-02, 5.5390e-05, 1.4164e-02, -1.2481e-02], + device='cuda:0'), grad: tensor([ 0.0060, 0.0111, 0.0292, -0.0084, 0.0370, -0.0159, -0.0361, 0.0057, + -0.0124, -0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 216.65, cls_loss 0.4706 cls_loss_mapping 0.0026 cls_loss_causal 0.4552 re_mapping 0.0065 re_causal 0.0174 /// teacc 98.90 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.0710, -0.1013, -0.1451, ..., 0.1203, -0.0611, 0.0304], + [-0.0794, 0.1074, -0.0659, ..., 0.0500, -0.0578, -0.1048], + [-0.0349, 0.0830, -0.1191, ..., -0.0136, 0.0396, 0.1268], + ..., + [-0.0085, 0.0709, -0.0008, ..., -0.0207, 0.0314, -0.0635], + [ 0.0944, -0.1075, 0.0456, ..., 0.0100, 0.0102, -0.0391], + [-0.0151, -0.0811, 0.0731, ..., -0.0862, 0.0295, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.3553e-05, 8.6799e-06, 1.7571e-04, ..., 6.0606e-04, + 2.7323e-04, 3.8356e-05], + [ 5.2601e-05, 1.2815e-04, 2.1000e-03, ..., 7.7133e-03, + 1.5945e-03, 5.5552e-05], + [-4.9400e-03, -2.9874e-04, 7.7367e-05, ..., 3.1686e-04, + 3.2787e-03, -1.1339e-03], + ..., + [ 1.6665e-04, -1.4603e-05, 1.2016e-03, ..., 1.0815e-03, + 8.9836e-04, 3.9876e-05], + [ 1.7271e-03, 2.3139e-04, -8.5754e-03, ..., -3.8223e-03, + -1.6571e-02, -1.9073e-04], + [ 1.8239e-05, 8.6844e-05, -3.4637e-03, ..., 2.7084e-03, + 1.2236e-03, 1.4983e-05]], device='cuda:0') +Epoch 341, bias, value: tensor([ 0.0140, -0.0054, -0.0129, 0.0073, -0.0140, -0.0217, 0.0132, -0.0004, + 0.0145, -0.0126], device='cuda:0'), grad: tensor([-0.0202, 0.0360, -0.0216, 0.0525, 0.0398, -0.0358, -0.0106, 0.0178, + -0.0315, -0.0264], device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 216.64, cls_loss 0.4685 cls_loss_mapping 0.0043 cls_loss_causal 0.4412 re_mapping 0.0064 re_causal 0.0167 /// teacc 98.72 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.0697, -0.1015, -0.1451, ..., 0.1202, -0.0614, 0.0302], + [-0.0803, 0.1078, -0.0640, ..., 0.0488, -0.0580, -0.1057], + [-0.0341, 0.0838, -0.1217, ..., -0.0134, 0.0395, 0.1279], + ..., + [-0.0096, 0.0703, -0.0007, ..., -0.0213, 0.0322, -0.0632], + [ 0.0944, -0.1071, 0.0455, ..., 0.0106, 0.0101, -0.0395], + [-0.0145, -0.0818, 0.0734, ..., -0.0863, 0.0292, -0.0238]], + device='cuda:0'), grad: tensor([[ 3.7789e-05, 3.2711e-04, 7.5960e-04, ..., 2.1243e-04, + 4.5013e-04, 7.1001e-04], + [ 2.4261e-03, 8.6188e-05, 9.5010e-05, ..., 4.1425e-05, + 2.5660e-05, 8.7738e-04], + [-8.4686e-03, 3.9339e-04, 3.9577e-04, ..., 2.1434e-04, + 8.9705e-05, -1.6832e-03], + ..., + [ 3.5839e-03, 1.9109e-04, 3.2020e-04, ..., 6.5029e-05, + 2.2292e-04, 1.2350e-03], + [ 5.8699e-04, 1.4031e-04, 5.0163e-04, ..., 2.0802e-04, + 2.0635e-04, 6.8951e-04], + [ 3.7217e-04, 1.0014e-04, -1.1158e-03, ..., 6.4611e-05, + -1.1683e-03, 3.2640e-04]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0135, -0.0053, -0.0114, 0.0063, -0.0134, -0.0216, 0.0133, -0.0010, + 0.0145, -0.0131], device='cuda:0'), grad: tensor([ 0.0088, 0.0103, -0.0089, 0.0052, -0.0193, 0.0048, -0.0248, 0.0129, + 0.0093, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 216.79, cls_loss 0.4663 cls_loss_mapping 0.0034 cls_loss_causal 0.4352 re_mapping 0.0063 re_causal 0.0155 /// teacc 98.85 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.0691, -0.1020, -0.1462, ..., 0.1213, -0.0618, 0.0302], + [-0.0814, 0.1074, -0.0639, ..., 0.0483, -0.0589, -0.1043], + [-0.0347, 0.0837, -0.1203, ..., -0.0129, 0.0403, 0.1276], + ..., + [-0.0093, 0.0710, -0.0008, ..., -0.0203, 0.0309, -0.0654], + [ 0.0955, -0.1082, 0.0455, ..., 0.0101, 0.0095, -0.0384], + [-0.0160, -0.0822, 0.0730, ..., -0.0864, 0.0310, -0.0248]], + device='cuda:0'), grad: tensor([[ 7.0572e-04, 2.3693e-06, 1.4675e-04, ..., 4.7541e-04, + -2.0771e-03, 4.8608e-05], + [ 3.9697e-04, -1.5162e-05, 3.9983e-04, ..., 9.9087e-04, + 6.0558e-04, 2.8372e-05], + [-1.1759e-03, 8.7395e-06, 3.4237e-04, ..., -2.0599e-03, + -1.3566e-04, -1.8865e-05], + ..., + [ 3.1710e-04, -5.4628e-05, 1.0481e-03, ..., 1.7452e-03, + 8.3494e-04, 2.4259e-05], + [ 3.4308e-04, 3.9022e-07, -2.4235e-04, ..., -1.6165e-03, + 1.2732e-03, 2.6241e-05], + [ 4.1652e-04, 5.3763e-05, -4.4346e-05, ..., 1.6785e-03, + 4.9829e-04, 9.2760e-06]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0123, -0.0045, -0.0101, 0.0073, -0.0136, -0.0221, 0.0132, -0.0020, + 0.0135, -0.0122], device='cuda:0'), grad: tensor([ 0.0015, 0.0212, -0.0165, -0.0117, -0.0134, 0.0239, -0.0024, -0.0025, + -0.0317, 0.0315], device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 217.05, cls_loss 0.4761 cls_loss_mapping 0.0034 cls_loss_causal 0.4423 re_mapping 0.0059 re_causal 0.0144 /// teacc 98.83 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.0700, -0.1022, -0.1472, ..., 0.1208, -0.0630, 0.0310], + [-0.0817, 0.1081, -0.0637, ..., 0.0490, -0.0597, -0.1027], + [-0.0352, 0.0827, -0.1208, ..., -0.0131, 0.0401, 0.1279], + ..., + [-0.0076, 0.0714, -0.0015, ..., -0.0197, 0.0304, -0.0642], + [ 0.0975, -0.1094, 0.0463, ..., 0.0098, 0.0105, -0.0389], + [-0.0171, -0.0808, 0.0723, ..., -0.0874, 0.0306, -0.0259]], + device='cuda:0'), grad: tensor([[ 1.1864e-03, 3.9972e-06, 4.5800e-04, ..., -2.1381e-03, + -2.3186e-04, -4.7833e-06], + [ 2.1915e-03, -6.9499e-05, 2.1839e-04, ..., 1.2255e-04, + 2.3746e-04, 1.5581e-06], + [ 6.1083e-04, -1.7866e-05, 4.3488e-04, ..., 1.1212e-04, + 2.3949e-04, -2.1309e-05], + ..., + [ 5.2595e-04, 8.5607e-06, 1.9684e-03, ..., 2.1708e-04, + 1.8940e-03, 5.2415e-06], + [ 3.6926e-03, 1.0341e-05, -7.0686e-03, ..., 2.6393e-04, + -5.9052e-03, -9.0450e-06], + [ 5.2071e-04, 3.8743e-06, 3.0637e-04, ..., 1.7416e-04, + 3.1924e-04, 4.1611e-06]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0132, -0.0042, -0.0104, 0.0064, -0.0134, -0.0232, 0.0142, -0.0008, + 0.0132, -0.0132], device='cuda:0'), grad: tensor([-0.0155, -0.0103, -0.0176, -0.0118, 0.0191, -0.0103, 0.0172, 0.0231, + -0.0078, 0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 216.97, cls_loss 0.5007 cls_loss_mapping 0.0023 cls_loss_causal 0.4825 re_mapping 0.0056 re_causal 0.0151 /// teacc 98.75 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.0695, -0.1045, -0.1483, ..., 0.1200, -0.0633, 0.0321], + [-0.0824, 0.1096, -0.0639, ..., 0.0499, -0.0595, -0.1031], + [-0.0342, 0.0826, -0.1220, ..., -0.0132, 0.0397, 0.1281], + ..., + [-0.0077, 0.0717, -0.0024, ..., -0.0204, 0.0295, -0.0661], + [ 0.0990, -0.1101, 0.0472, ..., 0.0105, 0.0107, -0.0394], + [-0.0164, -0.0818, 0.0732, ..., -0.0877, 0.0310, -0.0255]], + device='cuda:0'), grad: tensor([[ 2.0005e-06, 1.6654e-04, 5.5504e-04, ..., 4.3064e-05, + 1.2178e-03, 4.4084e-04], + [ 6.8732e-06, 4.8923e-04, 1.8704e-04, ..., 3.0565e-04, + 3.3998e-04, 1.3232e-04], + [ 3.6545e-06, 2.2620e-05, 1.5965e-03, ..., 6.3002e-05, + 3.3417e-03, 4.5967e-03], + ..., + [ 1.0710e-03, -4.9591e-04, -1.1539e-03, ..., -6.7949e-04, + -6.9141e-04, 5.8460e-04], + [ 5.8003e-06, -8.4352e-04, 1.1969e-03, ..., 1.3971e-04, + 2.3918e-03, -4.4937e-03], + [-1.3838e-03, -7.1287e-04, -7.2594e-03, ..., -1.0020e-04, + -1.3130e-02, -1.3752e-03]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0129, -0.0032, -0.0113, 0.0069, -0.0137, -0.0227, 0.0140, -0.0012, + 0.0130, -0.0129], device='cuda:0'), grad: tensor([-0.0166, 0.0006, 0.0367, -0.0096, 0.0206, 0.0148, 0.0140, 0.0139, + -0.0414, -0.0331], device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 216.67, cls_loss 0.4546 cls_loss_mapping 0.0029 cls_loss_causal 0.4304 re_mapping 0.0063 re_causal 0.0162 /// teacc 98.81 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.0709, -0.1053, -0.1489, ..., 0.1204, -0.0626, 0.0317], + [-0.0818, 0.1092, -0.0642, ..., 0.0495, -0.0604, -0.1033], + [-0.0337, 0.0828, -0.1226, ..., -0.0138, 0.0404, 0.1290], + ..., + [-0.0086, 0.0725, -0.0025, ..., -0.0196, 0.0297, -0.0670], + [ 0.0996, -0.1097, 0.0469, ..., 0.0119, 0.0100, -0.0399], + [-0.0152, -0.0832, 0.0734, ..., -0.0894, 0.0307, -0.0254]], + device='cuda:0'), grad: tensor([[ 4.5562e-04, 6.2864e-07, 6.8620e-06, ..., -2.8229e-04, + 3.4809e-04, -3.5000e-04], + [ 3.1590e-04, -1.2897e-05, 6.7987e-06, ..., 4.0263e-05, + 3.3259e-04, 2.6062e-05], + [ 2.3818e-04, -1.4246e-04, 3.5793e-05, ..., -6.0380e-05, + 1.6041e-03, -2.0102e-05], + ..., + [ 4.2558e-05, 2.8014e-06, 1.6718e-03, ..., 2.1644e-06, + -7.5436e-04, 8.2672e-05], + [-5.7945e-03, 5.1893e-06, 2.3186e-05, ..., -4.1872e-05, + 9.6750e-04, 7.7128e-05], + [ 1.1510e-04, 2.1216e-06, -2.5120e-03, ..., 8.3297e-06, + -1.8349e-03, 5.2124e-05]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0121, -0.0037, -0.0123, 0.0081, -0.0138, -0.0235, 0.0151, -0.0009, + 0.0136, -0.0129], device='cuda:0'), grad: tensor([ 0.0036, 0.0126, 0.0173, 0.0131, 0.0119, 0.0172, -0.0121, 0.0133, + -0.0802, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 216.80, cls_loss 0.5150 cls_loss_mapping 0.0032 cls_loss_causal 0.4838 re_mapping 0.0059 re_causal 0.0155 /// teacc 98.85 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.0732, -0.1062, -0.1502, ..., 0.1213, -0.0617, 0.0328], + [-0.0808, 0.1094, -0.0647, ..., 0.0488, -0.0607, -0.1047], + [-0.0344, 0.0830, -0.1234, ..., -0.0128, 0.0401, 0.1288], + ..., + [-0.0075, 0.0728, -0.0018, ..., -0.0196, 0.0305, -0.0665], + [ 0.0979, -0.1103, 0.0469, ..., 0.0117, 0.0092, -0.0400], + [-0.0137, -0.0832, 0.0742, ..., -0.0898, 0.0309, -0.0258]], + device='cuda:0'), grad: tensor([[ 1.8720e-07, -3.6106e-03, 2.7567e-07, ..., -7.5111e-03, + -3.6430e-03, 5.8794e-04], + [-1.3327e-06, -4.4060e-04, 3.1851e-06, ..., -2.7561e-04, + 3.3998e-04, 6.1333e-05], + [ 2.2184e-06, 3.1452e-03, 2.2948e-05, ..., 1.6775e-03, + 7.1764e-04, 4.0588e-03], + ..., + [ 1.2231e-04, -3.5572e-03, 1.2865e-03, ..., 7.2956e-05, + 5.2500e-04, -6.2103e-03], + [ 1.8356e-06, 3.5834e-04, 5.2713e-06, ..., 1.9705e-04, + 5.7667e-05, -7.4387e-05], + [ 2.1935e-05, 1.6296e-04, 2.3925e-04, ..., 1.5390e-04, + 1.3304e-04, 2.1446e-04]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0125, -0.0037, -0.0122, 0.0088, -0.0149, -0.0233, 0.0154, -0.0013, + 0.0129, -0.0123], device='cuda:0'), grad: tensor([-0.0086, -0.0385, 0.0355, 0.0113, -0.0527, 0.0263, 0.0269, -0.0200, + 0.0091, 0.0106], device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 216.48, cls_loss 0.4858 cls_loss_mapping 0.0037 cls_loss_causal 0.4614 re_mapping 0.0056 re_causal 0.0144 /// teacc 98.80 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.0720, -0.1057, -0.1504, ..., 0.1208, -0.0627, 0.0317], + [-0.0806, 0.1093, -0.0655, ..., 0.0483, -0.0616, -0.1053], + [-0.0343, 0.0832, -0.1225, ..., -0.0120, 0.0412, 0.1280], + ..., + [-0.0079, 0.0724, -0.0016, ..., -0.0209, 0.0300, -0.0666], + [ 0.0977, -0.1084, 0.0459, ..., 0.0152, 0.0093, -0.0393], + [-0.0137, -0.0842, 0.0740, ..., -0.0895, 0.0317, -0.0255]], + device='cuda:0'), grad: tensor([[ 4.4227e-04, 1.5891e-04, -7.8058e-04, ..., 2.7156e-04, + -1.0977e-03, 6.0272e-04], + [ 2.6178e-04, 5.4073e-04, 5.1767e-05, ..., 1.6212e-04, + 1.6570e-04, 3.4690e-04], + [ 4.0174e-04, 8.4782e-04, 7.7784e-05, ..., 2.1291e-04, + 2.2995e-04, 4.5347e-04], + ..., + [-1.2884e-03, -3.9749e-03, 1.5488e-03, ..., 3.0065e-04, + 1.3437e-03, 4.3797e-04], + [-3.3245e-03, 6.8617e-04, 1.2887e-04, ..., 1.8156e-04, + 2.3842e-04, 3.6550e-04], + [ 2.4915e-04, -3.1650e-05, -6.2656e-04, ..., 9.3937e-05, + 4.4274e-04, 4.2367e-04]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0141, -0.0044, -0.0119, 0.0078, -0.0144, -0.0239, 0.0153, -0.0018, + 0.0127, -0.0117], device='cuda:0'), grad: tensor([ 0.0103, 0.0160, 0.0143, -0.0164, 0.0067, 0.0077, 0.0163, -0.0081, + -0.0266, -0.0201], device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 216.67, cls_loss 0.4757 cls_loss_mapping 0.0034 cls_loss_causal 0.4522 re_mapping 0.0059 re_causal 0.0153 /// teacc 98.80 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.0708, -0.1042, -0.1512, ..., 0.1203, -0.0631, 0.0329], + [-0.0798, 0.1109, -0.0656, ..., 0.0477, -0.0617, -0.1060], + [-0.0345, 0.0820, -0.1234, ..., -0.0123, 0.0422, 0.1286], + ..., + [-0.0082, 0.0731, -0.0018, ..., -0.0211, 0.0297, -0.0669], + [ 0.0982, -0.1092, 0.0477, ..., 0.0165, 0.0106, -0.0410], + [-0.0143, -0.0845, 0.0750, ..., -0.0889, 0.0310, -0.0248]], + device='cuda:0'), grad: tensor([[-8.9931e-04, -3.4356e-04, 6.0558e-05, ..., -1.5793e-03, + 1.1712e-04, -5.5466e-03], + [ 2.4274e-05, -2.6599e-05, 1.2806e-06, ..., -9.5427e-05, + 2.8480e-06, 4.4703e-04], + [ 1.6916e-04, -6.6471e-04, 3.3021e-04, ..., 3.2640e-04, + 6.7949e-04, 9.7418e-04], + ..., + [ 8.9109e-06, 6.3144e-06, 4.4137e-05, ..., 8.5771e-05, + 1.1760e-04, 2.3925e-04], + [ 6.8843e-05, 7.2241e-05, 7.8261e-05, ..., -6.1083e-04, + 1.6856e-04, 5.0354e-04], + [ 9.9957e-05, 3.0518e-05, -1.4620e-03, ..., 5.6952e-05, + -3.1757e-03, 2.8610e-04]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0145, -0.0053, -0.0119, 0.0074, -0.0143, -0.0228, 0.0147, -0.0018, + 0.0128, -0.0114], device='cuda:0'), grad: tensor([-0.0634, 0.0172, 0.0208, 0.0191, -0.0161, 0.0571, -0.0656, 0.0178, + 0.0127, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 216.72, cls_loss 0.4608 cls_loss_mapping 0.0030 cls_loss_causal 0.4393 re_mapping 0.0058 re_causal 0.0153 /// teacc 98.73 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.0704, -0.1044, -0.1521, ..., 0.1201, -0.0632, 0.0315], + [-0.0792, 0.1108, -0.0662, ..., 0.0493, -0.0627, -0.1036], + [-0.0352, 0.0819, -0.1239, ..., -0.0122, 0.0423, 0.1296], + ..., + [-0.0089, 0.0729, -0.0032, ..., -0.0211, 0.0291, -0.0674], + [ 0.0981, -0.1086, 0.0481, ..., 0.0155, 0.0101, -0.0411], + [-0.0157, -0.0843, 0.0748, ..., -0.0874, 0.0304, -0.0243]], + device='cuda:0'), grad: tensor([[ 1.3924e-04, 7.4040e-08, 4.8256e-04, ..., 8.8644e-04, + 1.2290e-04, 1.1498e-04], + [ 7.7295e-04, -5.6531e-07, 6.2895e-04, ..., 1.9388e-03, + 1.0103e-04, 5.5742e-04], + [ 2.7680e-04, 2.3078e-06, -4.9057e-03, ..., -2.5196e-03, + 1.4901e-04, 2.0969e-04], + ..., + [ 1.2982e-04, 7.1526e-04, 1.9522e-03, ..., 1.0090e-03, + 8.2016e-04, 1.1784e-04], + [-2.5997e-03, 4.4964e-06, 8.1825e-04, ..., -1.7195e-03, + 2.8849e-04, -1.6804e-03], + [ 1.5330e-04, -8.5974e-04, -7.7868e-04, ..., 9.9754e-04, + -5.8699e-04, 1.4138e-04]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0143, -0.0045, -0.0129, 0.0072, -0.0142, -0.0220, 0.0149, -0.0026, + 0.0132, -0.0115], device='cuda:0'), grad: tensor([ 0.0190, 0.0043, -0.0428, 0.0113, 0.0263, -0.0048, -0.0582, 0.0229, + -0.0039, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 216.66, cls_loss 0.4499 cls_loss_mapping 0.0040 cls_loss_causal 0.4283 re_mapping 0.0064 re_causal 0.0164 /// teacc 98.83 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.0696, -0.1055, -0.1527, ..., 0.1203, -0.0636, 0.0340], + [-0.0804, 0.1107, -0.0677, ..., 0.0489, -0.0643, -0.1038], + [-0.0347, 0.0827, -0.1218, ..., -0.0115, 0.0426, 0.1304], + ..., + [-0.0085, 0.0727, -0.0018, ..., -0.0215, 0.0297, -0.0678], + [ 0.0990, -0.1079, 0.0472, ..., 0.0159, 0.0095, -0.0423], + [-0.0148, -0.0847, 0.0742, ..., -0.0880, 0.0302, -0.0234]], + device='cuda:0'), grad: tensor([[ 1.5438e-04, 6.4261e-07, -6.6757e-04, ..., -1.8787e-03, + -1.3533e-03, -5.1785e-04], + [-2.1076e-03, -5.4576e-06, -2.5183e-06, ..., -1.7843e-03, + -1.0773e-02, -6.9199e-03], + [ 1.2469e-04, 2.5487e-04, 1.0055e-04, ..., 3.2735e-04, + 1.9159e-03, 1.1358e-03], + ..., + [ 3.6621e-04, -2.5201e-04, -7.3910e-05, ..., 7.4744e-05, + -6.4969e-06, 7.6413e-05], + [ 2.7108e-04, -1.9982e-05, 1.1945e-04, ..., 6.3467e-04, + 1.8358e-03, 1.1406e-03], + [ 2.6965e-04, 6.4448e-06, 1.2958e-04, ..., 5.3692e-04, + 5.5313e-04, 2.8467e-04]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0140, -0.0051, -0.0124, 0.0073, -0.0139, -0.0235, 0.0151, -0.0030, + 0.0138, -0.0107], device='cuda:0'), grad: tensor([-0.0007, -0.0339, -0.0167, 0.0215, -0.0252, 0.0082, 0.0239, 0.0048, + 0.0113, 0.0069], device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 216.64, cls_loss 0.4843 cls_loss_mapping 0.0048 cls_loss_causal 0.4545 re_mapping 0.0060 re_causal 0.0150 /// teacc 98.74 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.0695, -0.1070, -0.1539, ..., 0.1193, -0.0639, 0.0334], + [-0.0809, 0.1101, -0.0664, ..., 0.0486, -0.0632, -0.1031], + [-0.0366, 0.0827, -0.1194, ..., -0.0117, 0.0422, 0.1296], + ..., + [-0.0072, 0.0731, -0.0021, ..., -0.0203, 0.0312, -0.0663], + [ 0.0994, -0.1093, 0.0474, ..., 0.0164, 0.0099, -0.0436], + [-0.0142, -0.0831, 0.0744, ..., -0.0891, 0.0303, -0.0226]], + device='cuda:0'), grad: tensor([[ 9.6709e-06, 3.6359e-06, 4.2915e-05, ..., 1.0109e-04, + 1.0395e-04, 4.0591e-05], + [-1.3649e-04, 6.2725e-07, 4.1924e-03, ..., -1.6489e-03, + 2.3425e-04, -4.1342e-04], + [ 4.5113e-06, -5.9336e-05, -8.2627e-06, ..., 2.6655e-04, + 3.9506e-04, 1.0341e-04], + ..., + [ 1.0349e-05, -9.9123e-05, 2.8253e-04, ..., 1.8203e-04, + 2.9302e-04, 4.8697e-05], + [ 2.6450e-05, 2.6181e-05, 4.4441e-04, ..., 6.4135e-04, + 3.0208e-04, 4.1276e-05], + [ 3.1777e-06, 2.4348e-05, -2.7204e-04, ..., 1.3316e-04, + -2.2739e-05, 4.4644e-05]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0130, -0.0046, -0.0127, 0.0074, -0.0134, -0.0225, 0.0145, -0.0022, + 0.0127, -0.0106], device='cuda:0'), grad: tensor([ 6.9733e-03, -5.9624e-03, 1.3313e-02, -7.1869e-03, -1.1444e-02, + 9.2926e-03, 8.6746e-03, -2.1774e-02, -5.6148e-05, 8.1635e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 216.66, cls_loss 0.4827 cls_loss_mapping 0.0036 cls_loss_causal 0.4598 re_mapping 0.0059 re_causal 0.0154 /// teacc 98.85 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.0708, -0.1054, -0.1535, ..., 0.1198, -0.0631, 0.0336], + [-0.0813, 0.1102, -0.0661, ..., 0.0479, -0.0648, -0.1029], + [-0.0374, 0.0836, -0.1196, ..., -0.0119, 0.0400, 0.1282], + ..., + [-0.0073, 0.0718, -0.0028, ..., -0.0213, 0.0306, -0.0661], + [ 0.0994, -0.1099, 0.0477, ..., 0.0152, 0.0102, -0.0446], + [-0.0137, -0.0818, 0.0744, ..., -0.0896, 0.0300, -0.0220]], + device='cuda:0'), grad: tensor([[-1.2529e-04, -3.8052e-04, -2.6627e-03, ..., -1.5717e-03, + -3.1147e-03, -2.4567e-03], + [ 1.7453e-06, -3.8242e-03, 1.0896e-04, ..., -1.8234e-03, + -1.3838e-03, 8.5652e-05], + [ 2.5153e-05, 4.1351e-03, 1.0118e-03, ..., 1.3227e-03, + 1.5926e-03, -1.3304e-04], + ..., + [ 3.9898e-06, -2.8172e-03, -2.5058e-04, ..., -1.4944e-03, + 1.2660e-04, 2.4652e-04], + [ 2.4796e-05, 4.6086e-04, 4.2534e-04, ..., -9.7466e-04, + 5.7173e-04, 6.0511e-04], + [ 4.6283e-05, 5.0211e-04, 2.5678e-04, ..., 8.1205e-04, + 4.0579e-04, 2.2340e-04]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0126, -0.0046, -0.0130, 0.0078, -0.0139, -0.0222, 0.0158, -0.0022, + 0.0120, -0.0104], device='cuda:0'), grad: tensor([-0.0446, -0.0470, 0.0345, 0.0284, 0.0285, 0.0206, -0.0072, -0.0323, + -0.0068, 0.0260], device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 216.58, cls_loss 0.4820 cls_loss_mapping 0.0029 cls_loss_causal 0.4621 re_mapping 0.0059 re_causal 0.0155 /// teacc 98.70 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.0712, -0.1049, -0.1519, ..., 0.1196, -0.0630, 0.0336], + [-0.0824, 0.1104, -0.0662, ..., 0.0485, -0.0649, -0.1010], + [-0.0365, 0.0838, -0.1184, ..., -0.0119, 0.0404, 0.1275], + ..., + [-0.0075, 0.0716, -0.0031, ..., -0.0219, 0.0306, -0.0669], + [ 0.0998, -0.1104, 0.0487, ..., 0.0151, 0.0116, -0.0442], + [-0.0133, -0.0815, 0.0745, ..., -0.0895, 0.0295, -0.0230]], + device='cuda:0'), grad: tensor([[ 1.3970e-09, 1.4722e-04, 6.5207e-05, ..., 2.4605e-04, + 1.5080e-04, 2.4214e-08], + [ 0.0000e+00, 3.9787e-03, 2.0385e-04, ..., 7.3891e-03, + 1.2398e-03, 3.1199e-08], + [ 0.0000e+00, 7.7152e-04, 3.9458e-05, ..., 1.9226e-03, + 3.0303e-04, -4.0652e-07], + ..., + [ 0.0000e+00, -1.2550e-03, -2.0905e-03, ..., 1.3514e-03, + 6.1226e-04, 5.1688e-08], + [ 3.2596e-09, -3.7122e-04, 4.6104e-05, ..., -4.6120e-03, + 1.4520e-04, 6.7987e-08], + [ 0.0000e+00, 2.2812e-03, 1.3626e-04, ..., 1.1759e-03, + -1.4944e-03, 6.9849e-09]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0122, -0.0046, -0.0127, 0.0069, -0.0139, -0.0222, 0.0156, -0.0020, + 0.0122, -0.0099], device='cuda:0'), grad: tensor([ 0.0160, 0.0363, -0.0157, -0.0203, -0.0157, 0.0089, 0.0023, 0.0068, + -0.0352, 0.0165], device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 216.78, cls_loss 0.4775 cls_loss_mapping 0.0045 cls_loss_causal 0.4525 re_mapping 0.0060 re_causal 0.0148 /// teacc 98.85 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.0718, -0.1049, -0.1523, ..., 0.1200, -0.0634, 0.0339], + [-0.0838, 0.1113, -0.0659, ..., 0.0489, -0.0638, -0.1015], + [-0.0367, 0.0837, -0.1191, ..., -0.0115, 0.0401, 0.1277], + ..., + [-0.0070, 0.0714, -0.0026, ..., -0.0210, 0.0305, -0.0670], + [ 0.1004, -0.1108, 0.0484, ..., 0.0147, 0.0110, -0.0439], + [-0.0124, -0.0820, 0.0746, ..., -0.0908, 0.0302, -0.0239]], + device='cuda:0'), grad: tensor([[ 4.8423e-04, 4.0627e-04, 4.5240e-05, ..., 1.0473e-04, + 1.6394e-03, 1.5984e-03], + [-1.0719e-03, -1.1368e-03, 3.0808e-06, ..., -1.9875e-03, + -2.2066e-04, 9.0972e-06], + [ 3.4499e-04, -1.8883e-03, 5.9158e-06, ..., 9.6619e-05, + -1.5686e-02, -1.3344e-02], + ..., + [ 6.3658e-04, 2.9445e-04, 6.7055e-06, ..., 6.2752e-04, + 1.4544e-04, 3.9876e-05], + [ 5.3406e-04, 1.6606e-04, -8.7118e-04, ..., 3.9530e-04, + -4.7417e-03, -6.5851e-04], + [ 3.0184e-04, 8.9049e-05, 8.2791e-05, ..., 1.8597e-04, + 3.8624e-04, 5.5403e-05]], device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0127, -0.0042, -0.0134, 0.0067, -0.0137, -0.0227, 0.0155, -0.0015, + 0.0129, -0.0104], device='cuda:0'), grad: tensor([-0.0069, 0.0008, -0.0138, 0.0339, 0.0100, 0.0129, -0.0080, -0.0175, + -0.0231, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 216.92, cls_loss 0.4868 cls_loss_mapping 0.0055 cls_loss_causal 0.4623 re_mapping 0.0058 re_causal 0.0149 /// teacc 98.89 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.0720, -0.1060, -0.1489, ..., 0.1202, -0.0647, 0.0334], + [-0.0837, 0.1112, -0.0650, ..., 0.0486, -0.0647, -0.1010], + [-0.0377, 0.0848, -0.1192, ..., -0.0101, 0.0397, 0.1283], + ..., + [-0.0077, 0.0704, -0.0034, ..., -0.0223, 0.0304, -0.0662], + [ 0.1003, -0.1104, 0.0476, ..., 0.0150, 0.0101, -0.0440], + [-0.0127, -0.0815, 0.0741, ..., -0.0912, 0.0316, -0.0233]], + device='cuda:0'), grad: tensor([[ 9.9468e-04, -5.1737e-04, 1.5509e-04, ..., -2.9888e-03, + -1.9817e-03, -2.0885e-03], + [ 3.2210e-04, 2.9731e-04, 5.6326e-05, ..., 1.0567e-03, + 8.8155e-05, 5.5790e-04], + [ 5.5599e-04, 5.1975e-04, 2.1887e-04, ..., 8.0967e-04, + 2.4104e-04, 1.1663e-03], + ..., + [-2.5063e-03, -5.3585e-05, -9.2363e-04, ..., 4.3821e-04, + -2.4662e-03, 7.3862e-04], + [ 8.7833e-04, 2.8062e-04, 1.8966e-04, ..., 7.3051e-04, + 6.4564e-04, 1.0681e-03], + [ 6.6280e-04, 5.8460e-04, 2.3448e-04, ..., 4.8113e-04, + 2.6321e-04, 8.4209e-04]], device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0128, -0.0046, -0.0148, 0.0072, -0.0118, -0.0230, 0.0158, -0.0010, + 0.0126, -0.0115], device='cuda:0'), grad: tensor([-0.0041, 0.0341, 0.0278, 0.0123, -0.0654, 0.0208, -0.0293, -0.0062, + 0.0179, -0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 216.73, cls_loss 0.4908 cls_loss_mapping 0.0052 cls_loss_causal 0.4640 re_mapping 0.0058 re_causal 0.0142 /// teacc 98.84 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.0725, -0.1065, -0.1487, ..., 0.1199, -0.0650, 0.0335], + [-0.0831, 0.1110, -0.0649, ..., 0.0485, -0.0641, -0.1019], + [-0.0380, 0.0847, -0.1195, ..., -0.0103, 0.0408, 0.1270], + ..., + [-0.0056, 0.0702, -0.0041, ..., -0.0229, 0.0294, -0.0660], + [ 0.0989, -0.1104, 0.0475, ..., 0.0143, 0.0101, -0.0434], + [-0.0137, -0.0812, 0.0749, ..., -0.0912, 0.0331, -0.0229]], + device='cuda:0'), grad: tensor([[ 1.3256e-03, -2.0885e-04, -3.2157e-05, ..., -1.3142e-03, + -2.0516e-04, -1.9197e-03], + [ 3.6478e-05, 2.9898e-04, 1.7807e-05, ..., 7.9870e-04, + 2.8396e-04, 2.5487e-04], + [ 1.2910e-04, 6.9046e-04, 3.6098e-06, ..., -8.8120e-04, + 2.4152e-04, 2.2411e-04], + ..., + [ 3.1432e-07, -1.4133e-03, -2.3112e-05, ..., -3.8171e-04, + -3.6097e-04, 2.1780e-04], + [ 4.4537e-04, 1.2553e-04, 1.8096e-06, ..., 3.1519e-04, + 5.5641e-05, 2.4629e-04], + [ 3.4380e-04, 1.2589e-04, -2.7490e-04, ..., 2.8825e-04, + -1.6356e-03, 1.6785e-04]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0125, -0.0029, -0.0152, 0.0074, -0.0137, -0.0223, 0.0157, -0.0010, + 0.0123, -0.0111], device='cuda:0'), grad: tensor([-0.0512, 0.0152, -0.0170, 0.0139, 0.0115, 0.0209, -0.0208, 0.0070, + 0.0111, 0.0095], device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 216.81, cls_loss 0.4952 cls_loss_mapping 0.0029 cls_loss_causal 0.4693 re_mapping 0.0060 re_causal 0.0156 /// teacc 98.85 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.0723, -0.1071, -0.1494, ..., 0.1196, -0.0654, 0.0328], + [-0.0830, 0.1111, -0.0662, ..., 0.0496, -0.0649, -0.1020], + [-0.0380, 0.0841, -0.1196, ..., -0.0118, 0.0411, 0.1265], + ..., + [-0.0045, 0.0699, -0.0034, ..., -0.0228, 0.0295, -0.0676], + [ 0.0983, -0.1104, 0.0468, ..., 0.0152, 0.0091, -0.0420], + [-0.0146, -0.0800, 0.0746, ..., -0.0900, 0.0339, -0.0222]], + device='cuda:0'), grad: tensor([[ 1.0014e-04, 5.7280e-05, 3.0115e-05, ..., 1.3709e-04, + -2.0057e-05, 1.8805e-05], + [ 1.2007e-03, -3.8815e-03, 1.1218e-04, ..., -2.7695e-03, + 3.2991e-05, 7.7868e-04], + [ 2.0766e-04, 1.3697e-04, 7.3791e-05, ..., 5.5790e-04, + 3.2830e-04, 9.6679e-05], + ..., + [-5.0621e-03, 5.8975e-03, -2.4090e-03, ..., 6.6948e-03, + -8.1587e-04, 8.2970e-05], + [ 6.1464e-04, 8.7917e-05, 2.7728e-04, ..., -2.2850e-03, + -8.4862e-06, 6.5506e-05], + [ 1.1883e-03, 5.9754e-05, 5.6410e-04, ..., 1.8418e-04, + 2.6464e-04, 2.6122e-05]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0120, -0.0042, -0.0147, 0.0077, -0.0135, -0.0225, 0.0158, -0.0009, + 0.0132, -0.0112], device='cuda:0'), grad: tensor([ 0.0055, 0.0239, -0.0196, 0.0135, 0.0172, -0.0024, -0.0193, -0.0324, + 0.0018, 0.0118], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 357---------------------------------------------------- +epoch 357, time 217.44, cls_loss 0.4993 cls_loss_mapping 0.0021 cls_loss_causal 0.4720 re_mapping 0.0056 re_causal 0.0148 /// teacc 99.02 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.0709, -0.1069, -0.1485, ..., 0.1208, -0.0644, 0.0336], + [-0.0826, 0.1106, -0.0678, ..., 0.0492, -0.0653, -0.1034], + [-0.0384, 0.0845, -0.1203, ..., -0.0117, 0.0415, 0.1271], + ..., + [-0.0046, 0.0698, -0.0023, ..., -0.0240, 0.0299, -0.0683], + [ 0.0989, -0.1109, 0.0474, ..., 0.0148, 0.0097, -0.0412], + [-0.0136, -0.0805, 0.0744, ..., -0.0889, 0.0332, -0.0220]], + device='cuda:0'), grad: tensor([[ 1.1623e-04, 7.4387e-05, 1.1861e-05, ..., 3.3545e-04, + 3.4999e-06, 5.0575e-05], + [-1.0128e-03, -1.5724e-04, 1.9640e-05, ..., -2.6073e-03, + 1.3225e-05, -4.1223e-04], + [ 2.3758e-04, 4.8071e-05, 1.1645e-05, ..., 4.0674e-04, + 4.0978e-06, 1.0687e-04], + ..., + [ 2.0618e-03, 1.7655e-04, 5.1308e-03, ..., 1.4353e-03, + 3.4771e-03, 2.0301e-04], + [ 6.3300e-05, 9.7573e-05, 1.4675e-04, ..., 5.3835e-04, + 1.0276e-04, 1.4126e-04], + [-1.9875e-03, 9.1493e-05, -5.5237e-03, ..., -6.4182e-04, + -3.7441e-03, 1.2136e-04]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0138, -0.0037, -0.0145, 0.0066, -0.0141, -0.0221, 0.0154, -0.0014, + 0.0129, -0.0110], device='cuda:0'), grad: tensor([ 0.0189, -0.0331, 0.0203, -0.0048, 0.0252, -0.0174, -0.0068, -0.0414, + 0.0220, 0.0172], device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 216.96, cls_loss 0.4924 cls_loss_mapping 0.0025 cls_loss_causal 0.4649 re_mapping 0.0057 re_causal 0.0145 /// teacc 98.95 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.0712, -0.1067, -0.1487, ..., 0.1218, -0.0649, 0.0323], + [-0.0834, 0.1109, -0.0687, ..., 0.0498, -0.0646, -0.1043], + [-0.0372, 0.0854, -0.1202, ..., -0.0115, 0.0412, 0.1278], + ..., + [-0.0041, 0.0697, -0.0029, ..., -0.0235, 0.0299, -0.0679], + [ 0.0987, -0.1112, 0.0472, ..., 0.0160, 0.0103, -0.0413], + [-0.0126, -0.0810, 0.0748, ..., -0.0889, 0.0331, -0.0227]], + device='cuda:0'), grad: tensor([[ 6.2287e-05, 1.6332e-05, 3.1948e-04, ..., 1.4009e-03, + 8.5545e-04, 1.4534e-03], + [ 7.6056e-05, 1.1425e-03, 2.7847e-04, ..., 4.8470e-04, + 3.6573e-04, 2.2697e-04], + [ 8.3089e-05, -7.0839e-03, -5.3644e-04, ..., -1.7786e-04, + 1.8139e-03, 1.8940e-03], + ..., + [-2.7733e-03, 6.9923e-03, -1.4553e-03, ..., 5.0011e-03, + -5.3940e-03, 5.3024e-04], + [ 2.4052e-03, 1.8501e-04, 2.8172e-03, ..., 9.7513e-04, + 7.2365e-03, 1.5926e-03], + [-2.2590e-04, -1.7357e-03, 3.0956e-03, ..., 3.0470e-04, + 5.4359e-04, 5.9986e-04]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0134, -0.0039, -0.0142, 0.0082, -0.0151, -0.0226, 0.0148, -0.0015, + 0.0127, -0.0102], device='cuda:0'), grad: tensor([ 0.0179, 0.0100, -0.0204, -0.0365, -0.0196, 0.0011, -0.0133, 0.0138, + 0.0345, 0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 216.78, cls_loss 0.4690 cls_loss_mapping 0.0048 cls_loss_causal 0.4483 re_mapping 0.0060 re_causal 0.0150 /// teacc 98.80 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.0714, -0.1063, -0.1489, ..., 0.1219, -0.0641, 0.0315], + [-0.0841, 0.1113, -0.0686, ..., 0.0503, -0.0649, -0.1035], + [-0.0374, 0.0855, -0.1221, ..., -0.0119, 0.0414, 0.1280], + ..., + [-0.0038, 0.0700, -0.0029, ..., -0.0222, 0.0294, -0.0677], + [ 0.0988, -0.1128, 0.0491, ..., 0.0147, 0.0111, -0.0416], + [-0.0127, -0.0814, 0.0740, ..., -0.0897, 0.0329, -0.0230]], + device='cuda:0'), grad: tensor([[ 1.5736e-04, 2.2542e-04, 2.9969e-04, ..., 1.0433e-03, + 4.2844e-04, 6.0654e-04], + [ 1.1969e-03, -7.1716e-04, -1.5783e-03, ..., -3.4733e-03, + -7.8535e-04, -7.7248e-04], + [-2.7633e-04, 5.4866e-05, 9.7418e-04, ..., -3.0422e-03, + -6.8808e-04, -2.7580e-03], + ..., + [-1.6113e-02, 1.7290e-03, 2.8305e-03, ..., 1.1379e-04, + 2.3899e-03, 1.3666e-03], + [ 4.4870e-04, 2.8443e-04, 3.0398e-04, ..., 1.2703e-03, + 6.1417e-04, 1.0595e-03], + [ 1.2802e-02, -3.8109e-03, -5.2299e-03, ..., 3.0756e-04, + -5.1994e-03, -2.9678e-03]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0140, -0.0030, -0.0159, 0.0095, -0.0160, -0.0229, 0.0153, -0.0015, + 0.0122, -0.0102], device='cuda:0'), grad: tensor([ 0.0044, -0.0099, -0.0140, 0.0089, 0.0100, -0.0004, 0.0043, -0.0035, + 0.0071, -0.0069], device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 216.92, cls_loss 0.4892 cls_loss_mapping 0.0037 cls_loss_causal 0.4668 re_mapping 0.0056 re_causal 0.0146 /// teacc 98.58 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.0715, -0.1053, -0.1493, ..., 0.1221, -0.0638, 0.0321], + [-0.0844, 0.1104, -0.0674, ..., 0.0500, -0.0651, -0.1031], + [-0.0366, 0.0856, -0.1226, ..., -0.0128, 0.0398, 0.1285], + ..., + [-0.0038, 0.0709, -0.0036, ..., -0.0207, 0.0296, -0.0675], + [ 0.0993, -0.1136, 0.0498, ..., 0.0147, 0.0113, -0.0419], + [-0.0131, -0.0814, 0.0737, ..., -0.0903, 0.0336, -0.0235]], + device='cuda:0'), grad: tensor([[ 2.8491e-04, 1.4380e-05, 1.4091e-04, ..., 8.1241e-05, + 2.9182e-04, 1.3271e-07], + [ 1.3523e-05, 4.7497e-06, 4.5967e-04, ..., 7.2598e-05, + 2.6679e-04, 5.0291e-08], + [ 9.7930e-05, 2.5034e-04, 1.1978e-03, ..., 2.4509e-04, + 1.1730e-03, -8.8960e-06], + ..., + [ 6.2644e-05, -7.0095e-04, -9.0866e-03, ..., -8.2731e-04, + -3.4561e-03, 3.4785e-07], + [-9.0027e-04, 4.2111e-05, -2.4738e-03, ..., -2.6321e-04, + -8.5783e-04, 3.3993e-06], + [ 1.7607e-04, 3.3951e-04, 3.9406e-03, ..., 3.4809e-04, + 6.9475e-04, 2.3283e-08]], device='cuda:0') +Epoch 362, bias, value: tensor([ 0.0127, -0.0041, -0.0148, 0.0095, -0.0157, -0.0225, 0.0155, -0.0011, + 0.0123, -0.0102], device='cuda:0'), grad: tensor([ 0.0188, 0.0152, 0.0154, 0.0203, 0.0166, 0.0201, 0.0118, -0.0952, + -0.0538, 0.0309], device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 216.68, cls_loss 0.4793 cls_loss_mapping 0.0041 cls_loss_causal 0.4525 re_mapping 0.0062 re_causal 0.0154 /// teacc 98.88 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.0696, -0.1055, -0.1509, ..., 0.1226, -0.0640, 0.0338], + [-0.0844, 0.1108, -0.0674, ..., 0.0490, -0.0655, -0.1038], + [-0.0370, 0.0867, -0.1241, ..., -0.0117, 0.0389, 0.1271], + ..., + [-0.0038, 0.0707, -0.0039, ..., -0.0224, 0.0297, -0.0680], + [ 0.0994, -0.1150, 0.0499, ..., 0.0148, 0.0123, -0.0409], + [-0.0121, -0.0829, 0.0732, ..., -0.0902, 0.0330, -0.0243]], + device='cuda:0'), grad: tensor([[ 2.6240e-07, 7.9811e-05, 1.4439e-05, ..., -5.5161e-03, + -5.7745e-04, 2.1877e-03], + [ 1.6019e-07, 6.2287e-05, 3.9972e-06, ..., 7.6580e-04, + 6.0976e-05, 9.2685e-06], + [ 6.4494e-08, -5.7554e-04, 1.6699e-03, ..., 6.1989e-04, + 2.4738e-03, 6.0171e-05], + ..., + [ 1.1642e-09, 3.7819e-05, 1.8406e-04, ..., 1.5461e-04, + 3.0494e-04, 2.4065e-06], + [ 4.8429e-07, 7.3373e-05, -2.4872e-03, ..., 6.1464e-04, + -3.3188e-03, 2.9370e-05], + [ 2.1420e-08, 7.1108e-05, 4.0555e-04, ..., 7.8535e-04, + 5.9700e-04, 4.7684e-06]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0135, -0.0050, -0.0143, 0.0080, -0.0157, -0.0213, 0.0150, -0.0015, + 0.0134, -0.0104], device='cuda:0'), grad: tensor([ 0.0010, 0.0135, -0.0124, 0.0125, 0.0112, 0.0167, -0.0171, -0.0501, + 0.0110, 0.0135], device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 216.80, cls_loss 0.4628 cls_loss_mapping 0.0021 cls_loss_causal 0.4367 re_mapping 0.0062 re_causal 0.0160 /// teacc 98.88 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.0693, -0.1067, -0.1507, ..., 0.1232, -0.0637, 0.0339], + [-0.0845, 0.1119, -0.0690, ..., 0.0489, -0.0652, -0.1039], + [-0.0375, 0.0861, -0.1251, ..., -0.0127, 0.0395, 0.1273], + ..., + [-0.0032, 0.0706, -0.0033, ..., -0.0225, 0.0301, -0.0689], + [ 0.1003, -0.1152, 0.0501, ..., 0.0151, 0.0122, -0.0421], + [-0.0128, -0.0827, 0.0736, ..., -0.0897, 0.0330, -0.0232]], + device='cuda:0'), grad: tensor([[ 6.7043e-04, 1.0006e-05, 3.7122e-04, ..., 6.2084e-04, + 5.2595e-04, 1.5974e-03], + [ 3.9887e-04, -1.1273e-05, 2.1279e-04, ..., 3.3998e-04, + 3.8123e-04, 4.9734e-04], + [ 5.0211e-04, 9.4175e-05, 3.3784e-04, ..., 1.3769e-04, + 1.1358e-03, -6.0005e-03], + ..., + [-1.8940e-03, 4.0665e-03, 3.0060e-03, ..., -1.4944e-03, + 6.3038e-04, -7.0870e-05], + [ 7.5006e-04, 6.5386e-05, 5.8317e-04, ..., 6.0272e-04, + 7.8583e-04, 1.2531e-03], + [ 1.6212e-03, 6.2370e-04, 4.9934e-03, ..., 3.3927e-04, + 4.6196e-03, 4.8113e-04]], device='cuda:0') +Epoch 364, bias, value: tensor([ 0.0135, -0.0044, -0.0150, 0.0077, -0.0144, -0.0215, 0.0149, -0.0021, + 0.0145, -0.0114], device='cuda:0'), grad: tensor([ 0.0213, 0.0144, -0.0145, -0.0173, -0.0031, 0.0088, -0.0143, -0.0416, + 0.0211, 0.0252], device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 216.82, cls_loss 0.4614 cls_loss_mapping 0.0028 cls_loss_causal 0.4343 re_mapping 0.0059 re_causal 0.0149 /// teacc 98.89 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.0691, -0.1068, -0.1505, ..., 0.1233, -0.0635, 0.0341], + [-0.0841, 0.1130, -0.0683, ..., 0.0496, -0.0665, -0.1027], + [-0.0376, 0.0847, -0.1257, ..., -0.0130, 0.0387, 0.1281], + ..., + [-0.0029, 0.0718, -0.0025, ..., -0.0230, 0.0302, -0.0696], + [ 0.0992, -0.1167, 0.0507, ..., 0.0149, 0.0126, -0.0428], + [-0.0124, -0.0829, 0.0735, ..., -0.0904, 0.0326, -0.0227]], + device='cuda:0'), grad: tensor([[-3.0565e-04, 1.8537e-04, -5.1117e-03, ..., -7.7200e-04, + -1.7090e-03, -9.0933e-04], + [ 5.3549e-04, 3.8958e-04, 1.2808e-03, ..., 2.4891e-03, + 1.8358e-04, 9.8896e-04], + [ 6.2704e-04, 2.1434e-04, 2.9969e-04, ..., 9.1982e-04, + 2.9847e-05, 3.1805e-04], + ..., + [-1.9913e-03, 1.8740e-04, -3.2940e-03, ..., -1.1301e-03, + 8.1897e-05, -1.3132e-03], + [ 2.0523e-03, -1.2894e-03, 2.6455e-03, ..., -1.1969e-03, + 4.1485e-04, -7.7307e-05], + [ 1.6320e-04, -3.8671e-04, -1.3227e-03, ..., -4.9324e-03, + -7.9012e-04, -9.3746e-04]], device='cuda:0') +Epoch 365, bias, value: tensor([ 0.0130, -0.0035, -0.0143, 0.0075, -0.0145, -0.0211, 0.0151, -0.0032, + 0.0140, -0.0112], device='cuda:0'), grad: tensor([-0.0110, -0.0041, 0.0149, -0.0182, 0.0369, 0.0148, -0.0054, -0.0167, + 0.0097, -0.0210], device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 216.86, cls_loss 0.4841 cls_loss_mapping 0.0026 cls_loss_causal 0.4579 re_mapping 0.0059 re_causal 0.0163 /// teacc 98.97 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.0696, -0.1074, -0.1504, ..., 0.1236, -0.0636, 0.0341], + [-0.0834, 0.1126, -0.0678, ..., 0.0504, -0.0661, -0.1018], + [-0.0377, 0.0854, -0.1253, ..., -0.0136, 0.0397, 0.1288], + ..., + [-0.0019, 0.0714, -0.0027, ..., -0.0212, 0.0298, -0.0694], + [ 0.0986, -0.1152, 0.0516, ..., 0.0150, 0.0128, -0.0438], + [-0.0133, -0.0818, 0.0747, ..., -0.0911, 0.0326, -0.0232]], + device='cuda:0'), grad: tensor([[ 7.2360e-05, 1.2457e-04, 2.4557e-05, ..., 5.6744e-05, + -1.1027e-05, 6.3896e-04], + [ 1.2212e-05, -7.0534e-03, -5.0774e-03, ..., -8.6517e-03, + -1.7672e-03, -1.6937e-03], + [ 4.4197e-05, -4.1924e-03, -6.5727e-03, ..., 6.0272e-04, + -1.2367e-02, 1.5974e-03], + ..., + [ 6.3324e-04, 7.9193e-03, 6.1913e-03, ..., 2.2049e-03, + 1.2695e-02, 3.9864e-04], + [-3.1066e-04, 2.3103e-04, 7.8630e-04, ..., 8.7643e-04, + 2.3329e-04, -2.6550e-03], + [-5.9128e-04, 1.4257e-03, -9.5673e-03, ..., 2.6665e-03, + -7.1983e-03, 4.4751e-04]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0124, -0.0033, -0.0138, 0.0083, -0.0141, -0.0215, 0.0140, -0.0025, + 0.0144, -0.0121], device='cuda:0'), grad: tensor([ 0.0094, -0.0424, -0.0032, 0.0090, 0.0261, -0.0240, 0.0060, 0.0357, + -0.0175, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 216.96, cls_loss 0.4716 cls_loss_mapping 0.0040 cls_loss_causal 0.4480 re_mapping 0.0059 re_causal 0.0154 /// teacc 98.80 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.0699, -0.1067, -0.1509, ..., 0.1237, -0.0630, 0.0348], + [-0.0840, 0.1122, -0.0665, ..., 0.0509, -0.0647, -0.1005], + [-0.0378, 0.0869, -0.1252, ..., -0.0137, 0.0402, 0.1286], + ..., + [-0.0016, 0.0707, -0.0035, ..., -0.0229, 0.0288, -0.0700], + [ 0.0975, -0.1168, 0.0515, ..., 0.0145, 0.0128, -0.0429], + [-0.0134, -0.0819, 0.0736, ..., -0.0924, 0.0322, -0.0249]], + device='cuda:0'), grad: tensor([[ 6.8188e-05, 2.0778e-04, -5.6458e-04, ..., -7.5483e-04, + 6.9141e-05, -6.0797e-05], + [ 7.1637e-06, -9.5308e-05, 3.1233e-05, ..., -7.1287e-04, + 2.1055e-05, -1.1892e-03], + [ 8.7023e-06, 3.1209e-04, 3.4660e-05, ..., 8.2302e-04, + 3.2449e-04, 3.7217e-04], + ..., + [-5.5408e-04, -1.4410e-03, -2.3460e-04, ..., -3.9458e-04, + -8.3590e-04, 1.7285e-04], + [ 2.9027e-05, -8.0168e-05, 8.8155e-05, ..., -9.3603e-04, + -3.1090e-04, -2.0635e-04], + [ 5.0306e-04, 8.6308e-04, -5.2404e-04, ..., 5.6982e-04, + 2.9945e-04, 1.2732e-04]], device='cuda:0') +Epoch 367, bias, value: tensor([ 0.0125, -0.0038, -0.0145, 0.0086, -0.0133, -0.0210, 0.0126, -0.0017, + 0.0138, -0.0115], device='cuda:0'), grad: tensor([ 0.0022, -0.0240, 0.0182, 0.0108, 0.0085, 0.0108, 0.0094, 0.0020, + -0.0119, -0.0260], device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 217.07, cls_loss 0.4731 cls_loss_mapping 0.0022 cls_loss_causal 0.4494 re_mapping 0.0059 re_causal 0.0160 /// teacc 98.87 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.0697, -0.1067, -0.1496, ..., 0.1227, -0.0630, 0.0348], + [-0.0838, 0.1136, -0.0679, ..., 0.0506, -0.0651, -0.1018], + [-0.0375, 0.0862, -0.1256, ..., -0.0141, 0.0403, 0.1282], + ..., + [-0.0016, 0.0711, -0.0015, ..., -0.0230, 0.0302, -0.0704], + [ 0.0986, -0.1180, 0.0512, ..., 0.0147, 0.0129, -0.0426], + [-0.0136, -0.0828, 0.0729, ..., -0.0923, 0.0322, -0.0230]], + device='cuda:0'), grad: tensor([[ 7.7844e-05, 1.0765e-04, 2.0444e-04, ..., 5.1260e-04, + 9.1612e-05, 9.2745e-05], + [-1.1759e-03, 7.4148e-05, 1.4210e-04, ..., 4.2496e-03, + 6.4708e-06, 7.3850e-05], + [ 5.2363e-05, -6.8855e-04, 1.6451e-04, ..., 1.1069e-04, + 9.6738e-05, -7.1573e-04], + ..., + [ 6.3944e-04, 2.2686e-04, 1.3971e-03, ..., 7.9823e-04, + 6.3992e-04, 1.5152e-04], + [ 8.2970e-05, 3.2020e-04, 3.8576e-04, ..., 4.4465e-04, + 3.0160e-04, 2.5773e-04], + [-8.7917e-05, -1.7035e-04, -2.1992e-03, ..., 5.0688e-04, + -1.7853e-03, 1.0979e-04]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0129, -0.0047, -0.0129, 0.0072, -0.0146, -0.0220, 0.0125, -0.0010, + 0.0145, -0.0103], device='cuda:0'), grad: tensor([ 0.0112, 0.0031, 0.0068, 0.0109, 0.0119, -0.0175, -0.0349, 0.0205, + 0.0111, -0.0231], device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 216.95, cls_loss 0.4771 cls_loss_mapping 0.0025 cls_loss_causal 0.4552 re_mapping 0.0059 re_causal 0.0158 /// teacc 98.97 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.0703, -0.1071, -0.1503, ..., 0.1223, -0.0628, 0.0347], + [-0.0838, 0.1135, -0.0675, ..., 0.0511, -0.0648, -0.1017], + [-0.0376, 0.0854, -0.1274, ..., -0.0145, 0.0397, 0.1291], + ..., + [-0.0018, 0.0715, -0.0014, ..., -0.0231, 0.0302, -0.0706], + [ 0.0983, -0.1193, 0.0520, ..., 0.0141, 0.0128, -0.0425], + [-0.0144, -0.0822, 0.0727, ..., -0.0921, 0.0325, -0.0232]], + device='cuda:0'), grad: tensor([[-1.9085e-04, 2.5686e-06, -3.9959e-04, ..., 1.2094e-04, + 3.1769e-05, -1.8282e-06], + [ 3.3832e-04, -5.1200e-05, 3.4142e-04, ..., -1.4806e-04, + 1.2231e-04, 7.5949e-07], + [-2.6455e-03, -3.0518e-04, -5.2834e-03, ..., -3.3112e-03, + -2.0027e-03, -1.8144e-04], + ..., + [ 5.7793e-04, -4.3869e-03, -4.1466e-03, ..., 6.2275e-04, + -4.6120e-03, 3.1618e-07], + [ 2.1005e-04, 1.3061e-05, 1.9646e-03, ..., 7.9393e-04, + 7.0667e-04, 2.6915e-07], + [ 5.8126e-04, 4.3144e-03, 6.9923e-03, ..., 3.6168e-04, + 5.9471e-03, 5.9232e-07]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0128, -0.0038, -0.0143, 0.0081, -0.0140, -0.0212, 0.0118, -0.0005, + 0.0133, -0.0106], device='cuda:0'), grad: tensor([-0.0164, -0.0019, -0.0052, 0.0173, -0.0003, 0.0162, -0.0053, -0.0210, + -0.0150, 0.0316], device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 217.28, cls_loss 0.4585 cls_loss_mapping 0.0024 cls_loss_causal 0.4325 re_mapping 0.0061 re_causal 0.0162 /// teacc 98.76 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.0717, -0.1075, -0.1515, ..., 0.1233, -0.0630, 0.0343], + [-0.0841, 0.1135, -0.0656, ..., 0.0519, -0.0639, -0.1016], + [-0.0379, 0.0859, -0.1276, ..., -0.0138, 0.0406, 0.1288], + ..., + [-0.0015, 0.0718, -0.0016, ..., -0.0241, 0.0297, -0.0716], + [ 0.0998, -0.1204, 0.0515, ..., 0.0126, 0.0117, -0.0420], + [-0.0152, -0.0827, 0.0724, ..., -0.0917, 0.0333, -0.0223]], + device='cuda:0'), grad: tensor([[ 1.6298e-08, 4.3392e-05, -6.5804e-03, ..., 8.5309e-06, + -4.2648e-03, 3.4499e-04], + [ 9.0338e-08, 5.7077e-04, 7.4482e-04, ..., 3.4779e-05, + 5.1212e-04, 4.7386e-06], + [ 4.3306e-08, -1.5087e-03, -1.5297e-03, ..., -1.8239e-04, + -1.1568e-03, 3.1090e-03], + ..., + [ 2.9411e-06, 9.3758e-05, 3.0422e-03, ..., 1.5333e-05, + 1.9798e-03, 1.8766e-06], + [ 2.7508e-05, 1.9324e-04, 5.3310e-04, ..., 3.3170e-05, + 3.5763e-04, 2.6398e-03], + [-5.5134e-05, 1.6308e-04, 1.4315e-03, ..., 3.4243e-05, + 1.0519e-03, 1.4976e-05]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0119, -0.0038, -0.0137, 0.0082, -0.0136, -0.0214, 0.0123, -0.0014, + 0.0131, -0.0101], device='cuda:0'), grad: tensor([-0.0477, 0.0384, -0.0225, 0.0217, -0.0067, -0.0111, -0.0356, -0.0007, + 0.0383, 0.0257], device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 217.07, cls_loss 0.4920 cls_loss_mapping 0.0022 cls_loss_causal 0.4754 re_mapping 0.0060 re_causal 0.0156 /// teacc 98.88 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.0710, -0.1067, -0.1520, ..., 0.1235, -0.0632, 0.0345], + [-0.0833, 0.1128, -0.0646, ..., 0.0524, -0.0638, -0.1014], + [-0.0375, 0.0862, -0.1274, ..., -0.0142, 0.0414, 0.1292], + ..., + [-0.0009, 0.0729, -0.0027, ..., -0.0238, 0.0303, -0.0725], + [ 0.0994, -0.1205, 0.0520, ..., 0.0125, 0.0116, -0.0409], + [-0.0154, -0.0838, 0.0723, ..., -0.0918, 0.0323, -0.0234]], + device='cuda:0'), grad: tensor([[ 1.2636e-04, 1.3873e-05, 5.8591e-05, ..., 7.9060e-04, + 9.7752e-05, 2.8777e-04], + [ 3.3617e-05, 4.4899e-03, 1.6525e-05, ..., 2.6760e-03, + 6.6571e-06, 1.4901e-04], + [ 2.1648e-04, 6.9284e-04, 1.6892e-04, ..., 6.1131e-04, + 1.7965e-04, 1.5533e-04], + ..., + [ 3.2466e-06, -5.7526e-03, 6.8378e-04, ..., -2.4281e-03, + 4.8327e-04, 9.6679e-05], + [-2.9621e-03, 1.1200e-04, 5.8889e-04, ..., 3.0518e-04, + -3.1834e-03, 1.0371e-04], + [ 1.2636e-05, 3.3319e-05, -2.2850e-03, ..., 2.3448e-04, + -1.8883e-03, -3.1531e-05]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0124, -0.0028, -0.0139, 0.0081, -0.0141, -0.0214, 0.0118, -0.0008, + 0.0126, -0.0104], device='cuda:0'), grad: tensor([ 0.0102, 0.0317, 0.0107, 0.0010, 0.0037, -0.0236, -0.0142, -0.0273, + 0.0025, 0.0052], device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 217.32, cls_loss 0.4716 cls_loss_mapping 0.0024 cls_loss_causal 0.4387 re_mapping 0.0063 re_causal 0.0160 /// teacc 98.87 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.0717, -0.1074, -0.1523, ..., 0.1247, -0.0626, 0.0353], + [-0.0827, 0.1132, -0.0640, ..., 0.0515, -0.0645, -0.0994], + [-0.0381, 0.0869, -0.1278, ..., -0.0138, 0.0410, 0.1266], + ..., + [ 0.0019, 0.0724, -0.0026, ..., -0.0249, 0.0293, -0.0732], + [ 0.0986, -0.1214, 0.0514, ..., 0.0135, 0.0127, -0.0404], + [-0.0152, -0.0837, 0.0729, ..., -0.0923, 0.0327, -0.0229]], + device='cuda:0'), grad: tensor([[ 9.1672e-05, 9.2745e-05, 1.3838e-03, ..., 4.8876e-04, + 8.9645e-05, 5.9813e-05], + [ 1.6665e-04, -1.5507e-03, -1.1625e-03, ..., -1.4763e-03, + 1.6296e-04, 1.1003e-04], + [ 2.1946e-04, 1.3995e-04, 2.0742e-04, ..., 1.0462e-03, + 1.7798e-04, 9.1314e-05], + ..., + [ 6.1607e-04, 1.0757e-03, 1.1187e-03, ..., 2.1973e-03, + 6.6948e-04, 4.2415e-04], + [ 3.5024e-04, 5.5504e-04, 7.8201e-04, ..., 8.8978e-04, + 3.6788e-04, 2.0897e-04], + [ 2.7037e-04, 8.1301e-04, 4.3182e-03, ..., 1.2465e-03, + 1.9121e-04, 1.8120e-04]], device='cuda:0') +Epoch 372, bias, value: tensor([ 0.0120, -0.0031, -0.0143, 0.0082, -0.0139, -0.0219, 0.0130, -0.0016, + 0.0139, -0.0107], device='cuda:0'), grad: tensor([ 0.0149, 0.0043, 0.0188, -0.0147, -0.0313, -0.0328, -0.0438, 0.0378, + 0.0171, 0.0297], device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 217.01, cls_loss 0.4586 cls_loss_mapping 0.0022 cls_loss_causal 0.4308 re_mapping 0.0060 re_causal 0.0157 /// teacc 98.84 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.0713, -0.1066, -0.1513, ..., 0.1246, -0.0637, 0.0356], + [-0.0831, 0.1124, -0.0639, ..., 0.0516, -0.0636, -0.0992], + [-0.0373, 0.0879, -0.1278, ..., -0.0122, 0.0407, 0.1262], + ..., + [ 0.0016, 0.0718, -0.0025, ..., -0.0255, 0.0294, -0.0743], + [ 0.0991, -0.1197, 0.0505, ..., 0.0136, 0.0128, -0.0397], + [-0.0136, -0.0838, 0.0733, ..., -0.0922, 0.0337, -0.0235]], + device='cuda:0'), grad: tensor([[ 5.3376e-05, 1.7643e-05, 4.6229e-04, ..., 6.6519e-05, + 2.2659e-02, -2.6310e-07], + [ 1.3933e-03, 8.5688e-04, -1.9226e-03, ..., 3.6102e-02, + 2.9802e-07, 6.6729e-07], + [ 9.4652e-05, 2.0771e-03, 3.6693e-04, ..., 5.7161e-05, + 1.4111e-05, 6.6310e-06], + ..., + [ 1.3359e-05, -4.9400e-03, -4.3344e-04, ..., 2.6655e-04, + -6.3367e-06, 2.7055e-07], + [ 8.9836e-04, 6.8128e-05, 1.0815e-03, ..., -3.6926e-02, + 4.4331e-06, 3.9414e-06], + [ 5.1737e-05, 1.0896e-04, 1.7481e-03, ..., 1.4806e-04, + -2.2675e-02, 1.0738e-06]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0137, -0.0028, -0.0142, 0.0082, -0.0146, -0.0223, 0.0121, -0.0016, + 0.0129, -0.0098], device='cuda:0'), grad: tensor([ 0.0456, 0.0009, 0.0150, -0.0086, -0.0008, -0.0409, 0.0120, -0.0255, + -0.0011, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 217.39, cls_loss 0.4808 cls_loss_mapping 0.0049 cls_loss_causal 0.4581 re_mapping 0.0056 re_causal 0.0146 /// teacc 98.85 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.0722, -0.1071, -0.1507, ..., 0.1262, -0.0644, 0.0366], + [-0.0843, 0.1124, -0.0638, ..., 0.0512, -0.0633, -0.0996], + [-0.0364, 0.0876, -0.1293, ..., -0.0120, 0.0405, 0.1261], + ..., + [ 0.0014, 0.0724, -0.0037, ..., -0.0267, 0.0293, -0.0748], + [ 0.0981, -0.1186, 0.0502, ..., 0.0153, 0.0131, -0.0396], + [-0.0137, -0.0838, 0.0735, ..., -0.0932, 0.0335, -0.0241]], + device='cuda:0'), grad: tensor([[ 8.7857e-05, 4.1604e-05, 3.4356e-04, ..., 6.0654e-04, + -1.8161e-08, 1.6737e-04], + [ 1.3041e-04, -8.6451e-04, 1.0033e-03, ..., 2.5439e-04, + 2.5183e-06, 9.9599e-05], + [ 9.3341e-05, 4.3035e-04, 3.4952e-04, ..., 1.0271e-03, + -3.4589e-06, 2.7493e-05], + ..., + [ 1.3542e-04, 6.0987e-04, 1.2188e-03, ..., 1.2970e-03, + 1.5451e-06, 1.1927e-04], + [-1.2720e-04, -6.0844e-04, -8.0185e-03, ..., -2.7800e-04, + 2.6915e-07, 3.3283e-04], + [ 1.7297e-04, 1.7524e-04, 3.9330e-03, ..., -2.9106e-03, + -1.8924e-05, 1.5342e-04]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0133, -0.0030, -0.0139, 0.0086, -0.0149, -0.0220, 0.0120, -0.0015, + 0.0129, -0.0100], device='cuda:0'), grad: tensor([-0.0135, 0.0239, 0.0194, -0.0113, 0.0162, -0.0102, -0.0104, 0.0268, + -0.0235, -0.0173], device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 217.35, cls_loss 0.4653 cls_loss_mapping 0.0043 cls_loss_causal 0.4444 re_mapping 0.0057 re_causal 0.0147 /// teacc 98.78 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.0732, -0.1056, -0.1507, ..., 0.1267, -0.0635, 0.0372], + [-0.0841, 0.1121, -0.0643, ..., 0.0513, -0.0637, -0.0994], + [-0.0377, 0.0876, -0.1284, ..., -0.0124, 0.0411, 0.1270], + ..., + [ 0.0025, 0.0724, -0.0041, ..., -0.0274, 0.0291, -0.0748], + [ 0.0981, -0.1185, 0.0510, ..., 0.0153, 0.0130, -0.0402], + [-0.0139, -0.0840, 0.0731, ..., -0.0943, 0.0331, -0.0248]], + device='cuda:0'), grad: tensor([[ 1.9580e-05, 9.8467e-05, 1.3285e-03, ..., 1.6975e-03, + 5.0211e-04, 9.7322e-04], + [ 4.4443e-06, 7.3731e-05, 9.6858e-05, ..., 4.4405e-05, + 2.2545e-05, 4.2051e-05], + [ 6.2995e-06, -2.0981e-03, 3.4738e-04, ..., 3.1090e-04, + 1.2481e-04, 1.8811e-04], + ..., + [ 2.8595e-05, 1.1034e-03, 4.5657e-04, ..., 1.8609e-04, + 7.6830e-05, 2.0623e-04], + [ 5.2124e-05, 2.6870e-04, -2.0294e-03, ..., -3.6945e-03, + -9.0694e-04, -1.9550e-03], + [-1.7762e-04, -4.1342e-04, -7.4863e-04, ..., 4.9639e-04, + 5.5933e-04, -2.7370e-04]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0136, -0.0026, -0.0132, 0.0088, -0.0151, -0.0217, 0.0125, -0.0015, + 0.0117, -0.0108], device='cuda:0'), grad: tensor([ 0.0098, 0.0046, -0.0339, 0.0084, 0.0046, -0.0015, 0.0059, 0.0110, + -0.0037, -0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 217.06, cls_loss 0.4781 cls_loss_mapping 0.0039 cls_loss_causal 0.4551 re_mapping 0.0059 re_causal 0.0155 /// teacc 98.90 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.0741, -0.1057, -0.1499, ..., 0.1264, -0.0629, 0.0379], + [-0.0840, 0.1117, -0.0645, ..., 0.0516, -0.0637, -0.0992], + [-0.0375, 0.0884, -0.1301, ..., -0.0127, 0.0407, 0.1263], + ..., + [ 0.0017, 0.0722, -0.0052, ..., -0.0266, 0.0291, -0.0755], + [ 0.0992, -0.1185, 0.0519, ..., 0.0151, 0.0128, -0.0413], + [-0.0143, -0.0842, 0.0728, ..., -0.0952, 0.0324, -0.0248]], + device='cuda:0'), grad: tensor([[-2.0683e-05, 1.3128e-05, 2.7442e-04, ..., 1.4961e-04, + 2.5388e-06, 2.6468e-06], + [-4.1342e-04, -4.4778e-06, 4.9305e-04, ..., -4.6268e-06, + 1.3247e-05, 2.0582e-06], + [ 5.3287e-05, 1.2493e-04, 1.9350e-03, ..., 4.8637e-04, + 9.1791e-06, 2.2918e-05], + ..., + [ 4.6778e-04, 1.8626e-05, 2.4776e-03, ..., 5.2357e-04, + 9.7847e-04, 1.0483e-05], + [ 3.3307e-04, 7.5877e-05, 1.4658e-03, ..., 9.5010e-05, + 1.3793e-04, 2.3544e-05], + [-4.5090e-03, 2.5019e-05, -3.5496e-03, ..., -3.0651e-03, + -1.2808e-03, 7.1526e-06]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0124, -0.0027, -0.0139, 0.0094, -0.0153, -0.0220, 0.0127, -0.0011, + 0.0132, -0.0112], device='cuda:0'), grad: tensor([ 0.0045, 0.0131, 0.0122, -0.0217, 0.0133, 0.0018, -0.0045, 0.0135, + -0.0011, -0.0312], device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 217.32, cls_loss 0.4794 cls_loss_mapping 0.0031 cls_loss_causal 0.4605 re_mapping 0.0057 re_causal 0.0148 /// teacc 98.74 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.0737, -0.1048, -0.1506, ..., 0.1256, -0.0632, 0.0376], + [-0.0841, 0.1113, -0.0636, ..., 0.0524, -0.0635, -0.0987], + [-0.0384, 0.0887, -0.1304, ..., -0.0134, 0.0414, 0.1268], + ..., + [ 0.0013, 0.0723, -0.0052, ..., -0.0263, 0.0289, -0.0762], + [ 0.1002, -0.1199, 0.0511, ..., 0.0132, 0.0123, -0.0420], + [-0.0150, -0.0834, 0.0730, ..., -0.0957, 0.0325, -0.0255]], + device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.1498e-04, 4.3392e-04, ..., -8.6164e-04, + 1.1614e-06, 1.7695e-08], + [ 1.4156e-07, 9.9361e-05, 1.7822e-04, ..., 2.0714e-03, + 4.1723e-05, 2.7940e-09], + [ 2.1420e-08, 2.9526e-03, 4.9412e-05, ..., 4.1428e-03, + -7.3671e-04, 2.5192e-02], + ..., + [ 4.3035e-05, -5.6791e-04, 7.8249e-04, ..., 3.4733e-03, + 5.2309e-04, 1.8999e-07], + [ 7.1339e-06, 2.9013e-05, 1.2386e-04, ..., 1.6937e-03, + 7.2755e-06, 4.1910e-08], + [-1.6830e-02, 2.2411e-04, -2.1515e-02, ..., 1.9569e-03, + -1.3864e-04, 2.7940e-09]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0120, -0.0018, -0.0157, 0.0110, -0.0148, -0.0223, 0.0125, -0.0010, + 0.0122, -0.0107], device='cuda:0'), grad: tensor([ 0.0124, 0.0002, 0.0188, -0.0035, -0.0164, -0.0008, 0.0229, 0.0050, + -0.0099, -0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 217.00, cls_loss 0.4835 cls_loss_mapping 0.0028 cls_loss_causal 0.4579 re_mapping 0.0057 re_causal 0.0147 /// teacc 98.90 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.0749, -0.1058, -0.1501, ..., 0.1259, -0.0640, 0.0366], + [-0.0843, 0.1114, -0.0638, ..., 0.0524, -0.0633, -0.0983], + [-0.0382, 0.0879, -0.1303, ..., -0.0136, 0.0422, 0.1266], + ..., + [ 0.0016, 0.0730, -0.0046, ..., -0.0269, 0.0283, -0.0776], + [ 0.1012, -0.1202, 0.0509, ..., 0.0129, 0.0117, -0.0412], + [-0.0137, -0.0836, 0.0745, ..., -0.0958, 0.0331, -0.0256]], + device='cuda:0'), grad: tensor([[ 1.6892e-04, 6.2048e-05, 6.4611e-04, ..., 1.5078e-03, + 1.5869e-03, -4.4136e-03], + [-1.8291e-03, 2.6846e-04, 2.5940e-04, ..., 7.6027e-03, + 2.8706e-03, 3.2210e-04], + [ 1.8024e-04, -1.0653e-03, 6.6996e-04, ..., 3.5610e-03, + 2.5558e-03, 1.9588e-03], + ..., + [ 2.6345e-04, 2.8157e-04, 3.0923e-04, ..., 8.4782e-04, + 6.7329e-04, 3.4142e-04], + [ 2.3818e-04, 1.2493e-04, 6.8474e-04, ..., -1.5717e-02, + -3.2120e-03, 2.5291e-03], + [ 5.0621e-03, 7.8440e-05, -5.7888e-04, ..., 6.1941e-04, + -9.1600e-04, 8.8882e-04]], device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0122, -0.0019, -0.0147, 0.0115, -0.0160, -0.0219, 0.0117, -0.0024, + 0.0122, -0.0095], device='cuda:0'), grad: tensor([ 0.0014, 0.0113, 0.0320, -0.0216, -0.0052, 0.0273, 0.0023, -0.0378, + 0.0122, -0.0219], device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 216.90, cls_loss 0.4886 cls_loss_mapping 0.0028 cls_loss_causal 0.4644 re_mapping 0.0057 re_causal 0.0149 /// teacc 98.76 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.0739, -0.1064, -0.1503, ..., 0.1249, -0.0643, 0.0372], + [-0.0855, 0.1109, -0.0620, ..., 0.0508, -0.0637, -0.0984], + [-0.0386, 0.0880, -0.1321, ..., -0.0144, 0.0415, 0.1263], + ..., + [ 0.0009, 0.0726, -0.0042, ..., -0.0263, 0.0295, -0.0776], + [ 0.1015, -0.1183, 0.0496, ..., 0.0141, 0.0108, -0.0424], + [-0.0135, -0.0839, 0.0759, ..., -0.0960, 0.0344, -0.0255]], + device='cuda:0'), grad: tensor([[ 7.4804e-05, 1.3363e-04, 7.0706e-06, ..., 1.0306e-04, + 2.5332e-07, 6.8188e-05], + [ 1.0484e-04, -1.4305e-03, 5.8621e-05, ..., -1.7071e-03, + 4.6566e-09, 6.3598e-05], + [ 7.6711e-05, 2.0766e-04, 1.3582e-05, ..., 1.5342e-04, + 9.2201e-08, 9.2864e-05], + ..., + [-7.2479e-04, 6.8322e-06, -8.7929e-04, ..., -8.5068e-04, + 4.3958e-07, 3.4332e-05], + [ 2.4629e-04, 1.5354e-04, 2.1935e-05, ..., 1.4555e-04, + 9.5833e-07, 6.1274e-05], + [-1.1612e-02, 9.9599e-05, 7.3290e-04, ..., 8.4782e-04, + 6.6590e-07, 5.3346e-05]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0130, -0.0020, -0.0146, 0.0116, -0.0156, -0.0218, 0.0118, -0.0033, + 0.0123, -0.0100], device='cuda:0'), grad: tensor([-0.0151, 0.0074, 0.0181, 0.0178, 0.0293, -0.0178, -0.0390, 0.0034, + 0.0153, -0.0193], device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 216.69, cls_loss 0.4972 cls_loss_mapping 0.0025 cls_loss_causal 0.4697 re_mapping 0.0056 re_causal 0.0142 /// teacc 98.83 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.0735, -0.1066, -0.1511, ..., 0.1241, -0.0646, 0.0365], + [-0.0863, 0.1114, -0.0619, ..., 0.0509, -0.0631, -0.0982], + [-0.0388, 0.0863, -0.1332, ..., -0.0141, 0.0419, 0.1268], + ..., + [ 0.0014, 0.0740, -0.0030, ..., -0.0264, 0.0298, -0.0779], + [ 0.1016, -0.1184, 0.0500, ..., 0.0152, 0.0103, -0.0418], + [-0.0143, -0.0841, 0.0751, ..., -0.0955, 0.0340, -0.0259]], + device='cuda:0'), grad: tensor([[ 1.3161e-04, 2.6274e-04, 3.8671e-04, ..., 2.1124e-04, + 2.3186e-04, 2.4462e-04], + [-3.9744e-04, -1.9875e-03, -4.2343e-03, ..., -2.5635e-03, + -2.3918e-03, -8.4686e-04], + [-8.4686e-04, 3.4261e-04, 5.8365e-04, ..., 4.5371e-04, + 4.2939e-04, -1.3561e-03], + ..., + [ 4.3303e-05, -9.1028e-04, -3.6645e-04, ..., -1.0890e-04, + -1.3173e-04, -5.8025e-05], + [ 4.1294e-04, 6.1655e-04, 2.9993e-04, ..., 5.1785e-04, + 2.6822e-04, 7.0000e-04], + [ 3.0726e-05, 4.3392e-04, 4.9210e-04, ..., 1.9217e-04, + -1.3685e-04, 1.2350e-04]], device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0140, -0.0026, -0.0147, 0.0118, -0.0152, -0.0220, 0.0118, -0.0024, + 0.0106, -0.0099], device='cuda:0'), grad: tensor([ 0.0136, -0.0373, -0.0257, 0.0128, 0.0475, -0.0490, -0.0032, 0.0103, + 0.0193, 0.0116], device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 217.34, cls_loss 0.5081 cls_loss_mapping 0.0025 cls_loss_causal 0.4810 re_mapping 0.0058 re_causal 0.0151 /// teacc 98.70 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.0744, -0.1067, -0.1509, ..., 0.1237, -0.0631, 0.0362], + [-0.0880, 0.1118, -0.0626, ..., 0.0514, -0.0632, -0.0977], + [-0.0378, 0.0865, -0.1323, ..., -0.0143, 0.0415, 0.1271], + ..., + [ 0.0005, 0.0733, -0.0022, ..., -0.0261, 0.0290, -0.0796], + [ 0.1008, -0.1172, 0.0497, ..., 0.0150, 0.0108, -0.0402], + [-0.0150, -0.0835, 0.0755, ..., -0.0956, 0.0333, -0.0277]], + device='cuda:0'), grad: tensor([[ 8.0317e-06, 1.0425e-04, 3.0971e-04, ..., 8.3983e-05, + 4.8113e-04, 9.5189e-05], + [ 1.2350e-04, 5.7936e-04, 1.1379e-04, ..., 1.8787e-04, + 1.7405e-04, 5.4502e-04], + [ 1.0891e-03, 2.1591e-03, 9.1982e-04, ..., -1.4715e-03, + 1.4267e-03, 2.1610e-03], + ..., + [ 1.9670e-04, 3.5858e-04, 8.7118e-04, ..., 9.7692e-05, + 3.5119e-04, 7.1812e-04], + [ 6.3360e-05, 5.6362e-04, 1.0878e-04, ..., 3.7861e-04, + 1.2791e-04, 5.1117e-04], + [-1.6537e-03, -4.9400e-03, -2.3556e-03, ..., 6.0737e-05, + -2.9316e-03, -5.1155e-03]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0148, -0.0025, -0.0147, 0.0115, -0.0151, -0.0221, 0.0106, -0.0021, + 0.0113, -0.0102], device='cuda:0'), grad: tensor([-0.0155, 0.0026, 0.0163, 0.0128, 0.0149, 0.0113, 0.0055, 0.0120, + 0.0218, -0.0816], device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 216.82, cls_loss 0.4808 cls_loss_mapping 0.0039 cls_loss_causal 0.4593 re_mapping 0.0059 re_causal 0.0153 /// teacc 98.75 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.0743, -0.1075, -0.1520, ..., 0.1237, -0.0631, 0.0373], + [-0.0871, 0.1115, -0.0638, ..., 0.0510, -0.0627, -0.0981], + [-0.0374, 0.0865, -0.1318, ..., -0.0134, 0.0409, 0.1269], + ..., + [ 0.0008, 0.0725, -0.0025, ..., -0.0269, 0.0292, -0.0812], + [ 0.1008, -0.1177, 0.0496, ..., 0.0146, 0.0123, -0.0404], + [-0.0162, -0.0823, 0.0758, ..., -0.0923, 0.0325, -0.0287]], + device='cuda:0'), grad: tensor([[ 4.1914e-04, 3.0920e-07, 7.1096e-04, ..., 9.9373e-04, + 2.1398e-04, 8.5258e-04], + [ 2.2364e-04, -4.3869e-05, -4.1962e-03, ..., -1.0605e-03, + -1.9073e-04, 1.5929e-05], + [ 2.7984e-05, 6.6608e-06, 6.4850e-04, ..., 2.6393e-04, + 8.4381e-03, 8.9645e-03], + ..., + [-8.2016e-05, -5.6446e-05, 9.9754e-04, ..., 1.1462e-04, + 2.2316e-04, 5.7638e-05], + [ 4.2295e-04, 3.4161e-06, 2.9316e-03, ..., 3.9792e-04, + 5.7316e-04, 3.4642e-04], + [-3.4580e-03, 6.5506e-05, -1.0704e-02, ..., 3.5119e-04, + -5.4598e-04, 2.0826e-04]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0148, -0.0022, -0.0138, 0.0107, -0.0150, -0.0218, 0.0102, -0.0023, + 0.0107, -0.0098], device='cuda:0'), grad: tensor([-0.0169, -0.0256, 0.0203, -0.0239, 0.0228, 0.0154, -0.0016, 0.0105, + 0.0150, -0.0161], device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 217.30, cls_loss 0.4632 cls_loss_mapping 0.0036 cls_loss_causal 0.4452 re_mapping 0.0062 re_causal 0.0155 /// teacc 98.90 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.0749, -0.1072, -0.1520, ..., 0.1243, -0.0635, 0.0379], + [-0.0871, 0.1123, -0.0625, ..., 0.0518, -0.0623, -0.0984], + [-0.0353, 0.0873, -0.1325, ..., -0.0141, 0.0396, 0.1282], + ..., + [ 0.0020, 0.0716, -0.0018, ..., -0.0270, 0.0299, -0.0815], + [ 0.1000, -0.1184, 0.0500, ..., 0.0139, 0.0120, -0.0407], + [-0.0171, -0.0826, 0.0754, ..., -0.0918, 0.0332, -0.0300]], + device='cuda:0'), grad: tensor([[ 1.0014e-03, 1.1146e-04, 9.7847e-04, ..., 8.2397e-04, + 8.4043e-05, 2.1935e-03], + [ 2.7924e-03, 2.2697e-03, 2.7943e-04, ..., 5.1308e-03, + 4.0345e-06, 5.7526e-03], + [ 2.5821e-04, 1.6823e-03, 2.7251e-04, ..., 1.6909e-03, + 7.2978e-06, 2.1973e-03], + ..., + [-3.9520e-03, -2.2316e-03, 4.6563e-04, ..., -4.6844e-03, + 2.1651e-05, -5.1727e-03], + [ 1.9944e-04, 1.1832e-04, 3.2997e-04, ..., 1.6701e-04, + 1.2860e-05, 5.2929e-04], + [-4.1275e-03, 6.4492e-05, -3.4981e-03, ..., 1.2684e-04, + -7.5161e-05, -7.5817e-04]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0157, -0.0009, -0.0134, 0.0109, -0.0150, -0.0229, 0.0100, -0.0023, + 0.0092, -0.0100], device='cuda:0'), grad: tensor([ 0.0278, 0.0508, 0.0017, 0.0159, 0.0010, -0.0531, 0.0141, -0.0508, + 0.0129, -0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 217.31, cls_loss 0.4738 cls_loss_mapping 0.0030 cls_loss_causal 0.4481 re_mapping 0.0064 re_causal 0.0161 /// teacc 99.02 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.0751, -0.1076, -0.1533, ..., 0.1250, -0.0647, 0.0365], + [-0.0875, 0.1130, -0.0620, ..., 0.0517, -0.0625, -0.0988], + [-0.0360, 0.0874, -0.1327, ..., -0.0142, 0.0396, 0.1286], + ..., + [ 0.0007, 0.0712, -0.0021, ..., -0.0264, 0.0291, -0.0811], + [ 0.1009, -0.1182, 0.0500, ..., 0.0137, 0.0125, -0.0418], + [-0.0170, -0.0821, 0.0758, ..., -0.0922, 0.0344, -0.0285]], + device='cuda:0'), grad: tensor([[ 2.8804e-05, 2.7046e-06, 9.9558e-07, ..., 5.5879e-09, + -6.9849e-08, 8.3876e-04], + [-3.6216e-04, -5.0831e-04, 1.3504e-07, ..., 2.7940e-08, + 8.3819e-09, 6.0844e-04], + [ 3.8600e-04, 4.6420e-04, 3.2559e-06, ..., -1.3411e-07, + 9.0338e-08, 1.2407e-03], + ..., + [ 3.2514e-05, -1.5542e-05, 1.4938e-05, ..., 3.8184e-08, + 6.5193e-06, -1.9970e-03], + [ 3.9041e-05, 2.7224e-05, 1.6633e-06, ..., 1.0245e-08, + 1.0990e-07, -1.4458e-03], + [ 1.1379e-04, 1.6063e-05, 2.1005e-04, ..., 2.7940e-09, + -7.2308e-06, 7.3099e-04]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0157, -0.0013, -0.0140, 0.0114, -0.0156, -0.0217, 0.0101, -0.0016, + 0.0085, -0.0100], device='cuda:0'), grad: tensor([-0.0225, 0.0354, 0.0468, 0.0232, -0.0297, -0.0124, 0.0058, -0.0192, + -0.0337, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 217.16, cls_loss 0.5021 cls_loss_mapping 0.0039 cls_loss_causal 0.4763 re_mapping 0.0056 re_causal 0.0140 /// teacc 98.98 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.0737, -0.1080, -0.1536, ..., 0.1243, -0.0652, 0.0349], + [-0.0876, 0.1131, -0.0618, ..., 0.0527, -0.0622, -0.0983], + [-0.0364, 0.0881, -0.1327, ..., -0.0139, 0.0393, 0.1288], + ..., + [ 0.0005, 0.0709, -0.0024, ..., -0.0278, 0.0296, -0.0810], + [ 0.1005, -0.1190, 0.0495, ..., 0.0133, 0.0122, -0.0415], + [-0.0166, -0.0837, 0.0764, ..., -0.0911, 0.0342, -0.0292]], + device='cuda:0'), grad: tensor([[-3.9983e-04, 4.7594e-05, -1.8234e-03, ..., 2.6464e-04, + -2.1973e-03, -1.0958e-03], + [-4.3225e-04, 4.6551e-05, 2.2137e-04, ..., -1.9610e-04, + 3.2395e-05, 7.6175e-05], + [ 3.5214e-04, 1.3697e-04, 1.2493e-03, ..., 7.4530e-04, + 9.9659e-04, 7.7868e-04], + ..., + [ 4.0919e-05, 7.2145e-04, 8.6594e-04, ..., 2.9907e-03, + -5.4896e-05, 7.0190e-04], + [-7.2658e-05, 1.0997e-04, -5.0201e-03, ..., 7.0953e-04, + -2.8954e-03, -4.2915e-04], + [ 5.0020e-04, 2.5368e-04, 4.9782e-03, ..., 1.2999e-03, + 2.3651e-03, 3.5501e-04]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0153, -0.0006, -0.0153, 0.0117, -0.0149, -0.0219, 0.0100, -0.0014, + 0.0089, -0.0103], device='cuda:0'), grad: tensor([ 0.0024, -0.0469, 0.0192, 0.0341, 0.0147, -0.0116, 0.0207, -0.0327, + -0.0393, 0.0393], device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 216.76, cls_loss 0.5062 cls_loss_mapping 0.0033 cls_loss_causal 0.4844 re_mapping 0.0059 re_causal 0.0152 /// teacc 98.80 lr 0.00010000 +Epoch 386, weight, value: tensor([[-7.4435e-02, -1.0911e-01, -1.5397e-01, ..., 1.2447e-01, + -6.5565e-02, 3.4722e-02], + [-8.8311e-02, 1.1355e-01, -6.2640e-02, ..., 5.1881e-02, + -6.2637e-02, -9.8109e-02], + [-3.8327e-02, 8.7799e-02, -1.3413e-01, ..., -1.3662e-02, + 3.8771e-02, 1.2844e-01], + ..., + [-2.9486e-05, 7.0302e-02, -6.5808e-04, ..., -2.8064e-02, + 3.0147e-02, -8.1130e-02], + [ 1.0075e-01, -1.2003e-01, 4.8691e-02, ..., 1.3821e-02, + 1.2368e-02, -4.1766e-02], + [-1.6309e-02, -8.4928e-02, 7.5769e-02, ..., -9.1804e-02, + 3.4829e-02, -2.9158e-02]], device='cuda:0'), grad: tensor([[-5.6744e-04, 2.8517e-06, -3.9482e-03, ..., 3.2043e-04, + -2.4281e-03, -2.9445e-04], + [ 3.1665e-08, -8.3625e-05, 7.2539e-05, ..., 2.0134e-04, + 6.9559e-05, 1.1390e-04], + [ 3.3733e-06, 1.9670e-05, 1.5335e-03, ..., 7.3433e-04, + 1.0748e-03, 4.3988e-04], + ..., + [ 3.7923e-06, -7.5936e-05, 1.1110e-03, ..., 6.7425e-04, + 5.9414e-04, 5.0402e-04], + [ 3.2857e-06, 1.5497e-05, -1.4286e-03, ..., 9.9659e-04, + 1.2755e-04, 1.7619e-04], + [ 5.3596e-04, 7.6592e-06, 4.1962e-03, ..., 6.2752e-04, + 2.7351e-03, 6.4516e-04]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0144, -0.0002, -0.0154, 0.0107, -0.0156, -0.0207, 0.0108, -0.0012, + 0.0101, -0.0113], device='cuda:0'), grad: tensor([ 0.0012, 0.0200, -0.0351, -0.0443, -0.0107, 0.0049, 0.0196, 0.0241, + -0.0121, 0.0323], device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 217.33, cls_loss 0.4697 cls_loss_mapping 0.0027 cls_loss_causal 0.4458 re_mapping 0.0059 re_causal 0.0154 /// teacc 98.93 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.0757, -0.1100, -0.1539, ..., 0.1245, -0.0657, 0.0347], + [-0.0891, 0.1133, -0.0621, ..., 0.0518, -0.0625, -0.0980], + [-0.0386, 0.0881, -0.1333, ..., -0.0157, 0.0390, 0.1289], + ..., + [ 0.0013, 0.0709, 0.0004, ..., -0.0273, 0.0306, -0.0806], + [ 0.1014, -0.1212, 0.0478, ..., 0.0161, 0.0110, -0.0426], + [-0.0176, -0.0857, 0.0742, ..., -0.0924, 0.0338, -0.0298]], + device='cuda:0'), grad: tensor([[ 1.2405e-05, 1.2094e-04, 1.3506e-04, ..., 2.5071e-06, + 1.1599e-04, 2.4152e-04], + [ 4.2081e-05, 8.4698e-05, 9.4101e-06, ..., 6.3051e-07, + 7.2718e-06, -2.3251e-03], + [ 6.6459e-05, 1.7762e-04, 1.1563e-04, ..., 1.0237e-05, + 9.8169e-05, 4.8018e-04], + ..., + [-3.9148e-04, 8.8167e-04, 2.2964e-03, ..., 1.8728e-04, + 1.9703e-03, 1.0532e-04], + [ 3.5495e-05, 1.7738e-04, 1.5950e-04, ..., 9.4473e-06, + 1.1098e-04, 5.0497e-04], + [ 4.3660e-05, -2.1973e-03, -3.3531e-03, ..., -2.6417e-04, + -2.8324e-03, -9.2602e-04]], device='cuda:0') +Epoch 387, bias, value: tensor([ 1.4304e-02, 2.2290e-05, -1.6008e-02, 1.1506e-02, -1.5963e-02, + -2.0002e-02, 1.0761e-02, -1.8602e-03, 1.0100e-02, -1.1318e-02], + device='cuda:0'), grad: tensor([ 0.0128, -0.0411, -0.0111, 0.0227, 0.0173, 0.0155, 0.0122, 0.0051, + 0.0115, -0.0448], device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 217.07, cls_loss 0.4911 cls_loss_mapping 0.0036 cls_loss_causal 0.4609 re_mapping 0.0056 re_causal 0.0138 /// teacc 98.88 lr 0.00010000 +Epoch 388, weight, value: tensor([[-7.5924e-02, -1.0982e-01, -1.5347e-01, ..., 1.2613e-01, + -6.4998e-02, 3.6935e-02], + [-9.0225e-02, 1.1278e-01, -6.2113e-02, ..., 5.2499e-02, + -6.2688e-02, -9.8052e-02], + [-3.7567e-02, 8.8071e-02, -1.3109e-01, ..., -1.5686e-02, + 3.9017e-02, 1.2742e-01], + ..., + [ 2.3222e-03, 7.1120e-02, -1.3644e-04, ..., -2.8234e-02, + 3.0370e-02, -8.0105e-02], + [ 1.0167e-01, -1.2088e-01, 4.7437e-02, ..., 1.6723e-02, + 1.1504e-02, -4.2615e-02], + [-1.8041e-02, -8.6039e-02, 7.4021e-02, ..., -9.2388e-02, + 3.3018e-02, -2.8891e-02]], device='cuda:0'), grad: tensor([[ 1.2815e-04, 4.7743e-05, 3.1781e-04, ..., 1.0568e-04, + -3.7122e-04, -2.8629e-03], + [ 8.1837e-05, 4.3488e-04, -3.3784e-04, ..., -5.3525e-05, + -4.3583e-04, 4.0078e-04], + [ 8.4221e-05, 3.4404e-04, 2.5439e-04, ..., 1.2958e-04, + -1.2989e-03, 5.2404e-04], + ..., + [ 2.0421e-04, 5.2500e-04, 1.5116e-04, ..., 1.7440e-04, + 1.4246e-04, 3.0518e-04], + [ 6.8521e-04, 2.4092e-04, 3.2845e-03, ..., 1.2374e-04, + 1.9360e-03, 3.0208e-04], + [ 1.7738e-04, 3.6192e-04, 1.1225e-03, ..., 1.3363e-04, + 6.0272e-04, 5.1308e-04]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0151, -0.0010, -0.0160, 0.0115, -0.0155, -0.0210, 0.0106, -0.0010, + 0.0114, -0.0126], device='cuda:0'), grad: tensor([-0.0043, 0.0039, -0.0044, -0.0084, -0.0303, 0.0462, -0.0580, 0.0309, + -0.0070, 0.0315], device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 216.75, cls_loss 0.4949 cls_loss_mapping 0.0031 cls_loss_causal 0.4788 re_mapping 0.0062 re_causal 0.0156 /// teacc 98.86 lr 0.00010000 +Epoch 389, weight, value: tensor([[-7.6336e-02, -1.1020e-01, -1.5359e-01, ..., 1.2608e-01, + -6.4614e-02, 3.7344e-02], + [-8.9270e-02, 1.1295e-01, -6.0165e-02, ..., 5.2818e-02, + -6.2148e-02, -9.7870e-02], + [-3.8720e-02, 8.7815e-02, -1.3154e-01, ..., -1.6759e-02, + 3.9403e-02, 1.2759e-01], + ..., + [ 4.4196e-03, 7.1327e-02, 7.5988e-05, ..., -2.7929e-02, + 3.0946e-02, -8.0567e-02], + [ 1.0233e-01, -1.1957e-01, 4.7371e-02, ..., 1.6835e-02, + 1.1071e-02, -4.2295e-02], + [-1.8296e-02, -8.6190e-02, 7.4249e-02, ..., -9.2815e-02, + 3.2876e-02, -3.0063e-02]], device='cuda:0'), grad: tensor([[ 2.7156e-04, 7.6182e-07, -2.0415e-05, ..., 5.0926e-04, + -1.8835e-05, 1.9813e-04], + [ 3.1519e-04, -1.3530e-05, 3.8855e-06, ..., 4.7374e-04, + 1.5348e-05, 1.4353e-04], + [ 1.8537e-04, 2.0601e-06, 4.8093e-06, ..., 2.8443e-04, + 1.0967e-05, 1.5557e-04], + ..., + [ 2.1636e-04, 6.1318e-06, 3.2783e-05, ..., 5.3883e-04, + 4.8399e-05, 1.5700e-04], + [-2.4567e-03, 4.3772e-08, 1.4305e-05, ..., -3.3998e-04, + 2.0832e-05, 1.7762e-04], + [ 1.4031e-04, 0.0000e+00, -1.1086e-05, ..., 5.7459e-04, + 2.6636e-06, 1.5926e-04]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0145, -0.0007, -0.0148, 0.0119, -0.0150, -0.0221, 0.0100, -0.0016, + 0.0119, -0.0125], device='cuda:0'), grad: tensor([-0.0027, -0.0020, 0.0265, -0.0025, -0.0068, 0.0051, -0.0047, 0.0272, + -0.0660, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 217.28, cls_loss 0.4482 cls_loss_mapping 0.0019 cls_loss_causal 0.4297 re_mapping 0.0060 re_causal 0.0157 /// teacc 98.84 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.0785, -0.1113, -0.1539, ..., 0.1264, -0.0649, 0.0380], + [-0.0898, 0.1147, -0.0603, ..., 0.0530, -0.0616, -0.0966], + [-0.0388, 0.0874, -0.1313, ..., -0.0164, 0.0397, 0.1269], + ..., + [ 0.0043, 0.0705, 0.0002, ..., -0.0279, 0.0312, -0.0822], + [ 0.1022, -0.1188, 0.0468, ..., 0.0165, 0.0117, -0.0414], + [-0.0169, -0.0867, 0.0741, ..., -0.0928, 0.0325, -0.0294]], + device='cuda:0'), grad: tensor([[ 7.1943e-05, 3.2306e-05, 5.3585e-05, ..., 5.4151e-05, + 4.8466e-06, 2.1402e-06], + [-1.7776e-03, 4.6501e-03, 2.5368e-03, ..., 1.9491e-04, + 2.8044e-05, 1.6205e-07], + [-4.7493e-04, -8.4610e-03, -3.8509e-03, ..., -3.5954e-04, + 1.4439e-05, 5.1782e-06], + ..., + [ 6.1989e-04, -6.0707e-05, 1.7281e-03, ..., 8.3399e-04, + 2.6536e-04, 8.1398e-07], + [ 1.0366e-03, 5.0163e-04, 4.1366e-04, ..., 1.3387e-04, + 1.2524e-05, 9.7379e-06], + [ 1.9014e-04, -2.2173e-04, -4.4060e-03, ..., -1.5011e-03, + -5.1785e-04, 3.6042e-07]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0144, -0.0009, -0.0148, 0.0122, -0.0154, -0.0215, 0.0100, -0.0017, + 0.0119, -0.0126], device='cuda:0'), grad: tensor([ 0.0154, -0.0165, -0.0962, 0.0173, 0.0264, 0.0143, -0.0096, 0.0254, + 0.0258, -0.0024], device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 219.15, cls_loss 0.4834 cls_loss_mapping 0.0032 cls_loss_causal 0.4585 re_mapping 0.0057 re_causal 0.0141 /// teacc 98.63 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.0755, -0.1090, -0.1530, ..., 0.1263, -0.0653, 0.0390], + [-0.0898, 0.1149, -0.0604, ..., 0.0527, -0.0622, -0.0959], + [-0.0388, 0.0867, -0.1323, ..., -0.0182, 0.0398, 0.1265], + ..., + [ 0.0043, 0.0706, -0.0006, ..., -0.0272, 0.0315, -0.0813], + [ 0.1009, -0.1177, 0.0468, ..., 0.0174, 0.0116, -0.0427], + [-0.0159, -0.0882, 0.0748, ..., -0.0936, 0.0331, -0.0293]], + device='cuda:0'), grad: tensor([[ 8.7142e-05, 2.1756e-06, 5.7101e-05, ..., 1.7233e-03, + 5.9080e-04, 1.3483e-04], + [ 3.3945e-05, -7.0989e-05, 3.1963e-06, ..., 1.8282e-03, + 6.3276e-04, 1.4949e-04], + [ 1.5600e-06, -3.2812e-05, 5.5507e-06, ..., 1.0605e-03, + 3.1972e-04, 6.7115e-05], + ..., + [ 2.6189e-06, 1.1072e-05, 4.0442e-05, ..., 2.5692e-03, + 8.0776e-04, 1.9109e-04], + [-3.6502e-04, 1.8343e-05, -1.9157e-04, ..., 1.2398e-03, + 2.5368e-04, 7.8142e-05], + [ 1.0622e-04, 7.9945e-06, -1.3463e-05, ..., 5.6601e-04, + 2.2626e-04, 4.8965e-05]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0156, -0.0011, -0.0158, 0.0115, -0.0158, -0.0221, 0.0106, -0.0022, + 0.0124, -0.0115], device='cuda:0'), grad: tensor([ 0.0296, 0.0224, 0.0099, 0.0324, -0.0489, -0.0466, 0.0147, 0.0212, + -0.0196, -0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 216.99, cls_loss 0.4611 cls_loss_mapping 0.0036 cls_loss_causal 0.4280 re_mapping 0.0060 re_causal 0.0146 /// teacc 98.89 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.0728, -0.1093, -0.1523, ..., 0.1257, -0.0646, 0.0394], + [-0.0903, 0.1137, -0.0598, ..., 0.0512, -0.0627, -0.0967], + [-0.0396, 0.0862, -0.1325, ..., -0.0199, 0.0387, 0.1272], + ..., + [ 0.0046, 0.0715, 0.0008, ..., -0.0246, 0.0325, -0.0827], + [ 0.1001, -0.1160, 0.0461, ..., 0.0185, 0.0111, -0.0408], + [-0.0163, -0.0880, 0.0741, ..., -0.0946, 0.0331, -0.0292]], + device='cuda:0'), grad: tensor([[ 1.8847e-04, 1.1168e-05, 3.4261e-04, ..., -2.4815e-03, + 1.3840e-04, -2.7161e-03], + [ 1.5569e-04, -7.0000e-04, 6.1369e-04, ..., -1.9455e-04, + 3.5310e-04, -1.6844e-04], + [ 4.5300e-04, 6.0701e-04, 9.3699e-04, ..., 1.2302e-03, + 5.7364e-04, 5.3072e-04], + ..., + [ 1.5478e-03, 3.6508e-06, 1.3800e-03, ..., 5.1498e-04, + 7.6199e-04, 1.8239e-05], + [ 5.2147e-03, 1.2720e-04, 1.6260e-03, ..., 8.2111e-04, + 1.0796e-03, 4.1556e-04], + [-8.9569e-03, -1.4663e-04, 8.5678e-03, ..., 5.7697e-04, + 3.5839e-03, 6.0737e-05]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0158, -0.0017, -0.0150, 0.0113, -0.0156, -0.0208, 0.0107, -0.0028, + 0.0120, -0.0121], device='cuda:0'), grad: tensor([-0.0112, 0.0114, 0.0204, -0.0146, -0.0004, -0.0152, 0.0201, 0.0171, + 0.0052, -0.0328], device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 217.24, cls_loss 0.4544 cls_loss_mapping 0.0038 cls_loss_causal 0.4387 re_mapping 0.0056 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +Epoch 393, weight, value: tensor([[-7.1968e-02, -1.1018e-01, -1.5150e-01, ..., 1.2666e-01, + -6.3999e-02, 3.8967e-02], + [-9.1310e-02, 1.1176e-01, -5.9780e-02, ..., 5.1142e-02, + -6.2314e-02, -9.6533e-02], + [-4.0806e-02, 8.8398e-02, -1.3271e-01, ..., -1.8990e-02, + 4.0475e-02, 1.2694e-01], + ..., + [ 4.5757e-03, 7.1498e-02, 8.7226e-05, ..., -2.6237e-02, + 3.1653e-02, -8.4794e-02], + [ 1.0034e-01, -1.1617e-01, 4.5690e-02, ..., 1.7403e-02, + 1.0255e-02, -4.0070e-02], + [-1.6125e-02, -8.7793e-02, 7.4263e-02, ..., -9.3914e-02, + 3.3370e-02, -2.8214e-02]], device='cuda:0'), grad: tensor([[ 7.1526e-04, 5.4240e-05, -3.4714e-04, ..., -1.5574e-03, + -1.4448e-03, -3.8457e-04], + [ 1.0757e-02, 3.8357e-03, 2.6703e-04, ..., 9.5963e-05, + -3.4869e-05, 1.9348e-04], + [ 1.0170e-02, 8.1177e-03, 1.3943e-03, ..., 4.2152e-04, + 1.5678e-03, -5.3787e-03], + ..., + [ 3.8643e-03, 2.5291e-03, 8.7051e-03, ..., 8.5754e-03, + 8.7433e-03, 2.4366e-04], + [ 1.9693e-04, 4.7493e-04, 1.2436e-03, ..., -4.0474e-03, + -1.5163e-03, 4.4136e-03], + [ 2.8591e-03, -5.3101e-03, -6.0043e-03, ..., -8.3160e-03, + -6.6681e-03, 2.4366e-04]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0157, -0.0026, -0.0141, 0.0112, -0.0157, -0.0220, 0.0105, -0.0023, + 0.0116, -0.0109], device='cuda:0'), grad: tensor([-0.0053, 0.0153, 0.0079, 0.0500, -0.0366, -0.0403, 0.0136, 0.0370, + -0.0244, -0.0173], device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 216.43, cls_loss 0.4525 cls_loss_mapping 0.0026 cls_loss_causal 0.4308 re_mapping 0.0055 re_causal 0.0133 /// teacc 99.01 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.0712, -0.1108, -0.1523, ..., 0.1265, -0.0652, 0.0374], + [-0.0928, 0.1116, -0.0601, ..., 0.0509, -0.0623, -0.0962], + [-0.0427, 0.0892, -0.1324, ..., -0.0190, 0.0404, 0.1256], + ..., + [ 0.0040, 0.0715, 0.0004, ..., -0.0271, 0.0325, -0.0873], + [ 0.1001, -0.1163, 0.0462, ..., 0.0189, 0.0109, -0.0368], + [-0.0138, -0.0888, 0.0736, ..., -0.0935, 0.0326, -0.0278]], + device='cuda:0'), grad: tensor([[ 4.5991e-04, 7.2956e-05, 9.1410e-04, ..., 2.7466e-04, + -1.4178e-05, 9.8133e-04], + [ 4.3964e-04, -7.0152e-03, -1.9169e-04, ..., -5.3711e-03, + -3.9787e-03, 2.1350e-04], + [-7.7515e-03, -1.4138e-04, -2.1801e-03, ..., 2.7680e-04, + 2.0623e-04, 1.4057e-03], + ..., + [ 2.5291e-03, 1.0185e-02, 9.0332e-03, ..., 4.3983e-03, + 2.9694e-02, 2.0587e-04], + [ 2.0313e-03, 1.4699e-04, 4.9305e-04, ..., 6.8474e-04, + 7.2122e-05, 3.5071e-04], + [ 1.0080e-03, -3.4523e-03, -7.8125e-03, ..., 6.1846e-04, + -2.5803e-02, 2.8181e-04]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0155, -0.0024, -0.0142, 0.0108, -0.0157, -0.0222, 0.0103, -0.0032, + 0.0125, -0.0099], device='cuda:0'), grad: tensor([ 0.0141, -0.0445, -0.0077, -0.0392, 0.0078, 0.0072, 0.0010, 0.0532, + 0.0144, -0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 217.20, cls_loss 0.4609 cls_loss_mapping 0.0025 cls_loss_causal 0.4385 re_mapping 0.0057 re_causal 0.0148 /// teacc 98.90 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.0714, -0.1104, -0.1519, ..., 0.1275, -0.0650, 0.0360], + [-0.0926, 0.1115, -0.0606, ..., 0.0510, -0.0626, -0.0952], + [-0.0424, 0.0889, -0.1333, ..., -0.0202, 0.0385, 0.1244], + ..., + [ 0.0035, 0.0727, 0.0005, ..., -0.0270, 0.0318, -0.0865], + [ 0.0999, -0.1169, 0.0458, ..., 0.0194, 0.0116, -0.0361], + [-0.0133, -0.0893, 0.0739, ..., -0.0938, 0.0326, -0.0276]], + device='cuda:0'), grad: tensor([[ 1.4486e-03, 5.9992e-05, 3.9673e-04, ..., 1.0319e-03, + 4.5052e-03, 5.9471e-03], + [-7.6180e-03, -7.2250e-03, -4.1275e-03, ..., 1.4806e-04, + 9.1612e-05, 2.8706e-04], + [ 6.4850e-04, 6.3562e-04, 5.5218e-04, ..., 1.6892e-04, + 2.5153e-04, -9.0647e-04], + ..., + [ 9.1095e-03, 5.4779e-03, 3.7422e-03, ..., -7.4625e-04, + 6.4135e-05, -7.7295e-04], + [-8.5144e-03, 4.5747e-05, 1.4193e-05, ..., -2.4891e-03, + -4.9324e-03, -7.0801e-03], + [-8.0261e-03, 3.1495e-04, -6.9847e-03, ..., 1.1694e-04, + -2.0142e-03, 5.9032e-04]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0154, -0.0020, -0.0151, 0.0099, -0.0155, -0.0221, 0.0106, -0.0020, + 0.0119, -0.0097], device='cuda:0'), grad: tensor([ 0.0024, -0.0042, -0.0228, 0.0036, 0.0068, 0.0316, 0.0387, 0.0061, + -0.0498, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 217.31, cls_loss 0.4641 cls_loss_mapping 0.0026 cls_loss_causal 0.4460 re_mapping 0.0053 re_causal 0.0144 /// teacc 98.99 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.0733, -0.1104, -0.1531, ..., 0.1273, -0.0651, 0.0361], + [-0.0933, 0.1114, -0.0611, ..., 0.0509, -0.0637, -0.0956], + [-0.0413, 0.0883, -0.1342, ..., -0.0199, 0.0380, 0.1236], + ..., + [ 0.0018, 0.0731, 0.0021, ..., -0.0270, 0.0323, -0.0858], + [ 0.1008, -0.1168, 0.0473, ..., 0.0201, 0.0123, -0.0348], + [-0.0125, -0.0887, 0.0727, ..., -0.0938, 0.0319, -0.0288]], + device='cuda:0'), grad: tensor([[ 3.6764e-04, 1.7032e-05, 4.3392e-05, ..., 1.2481e-04, + 4.5729e-04, 2.1255e-04], + [ 1.3196e-04, -8.5297e-03, 3.9005e-04, ..., -5.7297e-03, + 2.1684e-04, 3.0518e-04], + [ 1.6916e-04, -4.4098e-03, -5.3291e-03, ..., -3.3474e-03, + 2.7394e-04, -3.2425e-03], + ..., + [-4.0936e-04, 1.1742e-02, 3.1834e-03, ..., 8.5831e-03, + 3.3379e-04, 1.8654e-03], + [ 1.7297e-04, 8.4698e-05, 2.6274e-04, ..., 4.3845e-04, + 2.6989e-04, 4.9448e-04], + [ 8.4591e-04, 3.4976e-04, 1.9875e-03, ..., 4.8804e-04, + 9.9468e-04, 4.5562e-04]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0161, -0.0016, -0.0144, 0.0086, -0.0150, -0.0215, 0.0096, -0.0023, + 0.0128, -0.0107], device='cuda:0'), grad: tensor([-0.0044, -0.0672, 0.0010, 0.0261, 0.0279, -0.0390, 0.0286, 0.0674, + -0.0388, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 217.29, cls_loss 0.4359 cls_loss_mapping 0.0016 cls_loss_causal 0.4087 re_mapping 0.0058 re_causal 0.0153 /// teacc 98.84 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.0752, -0.1084, -0.1541, ..., 0.1266, -0.0652, 0.0362], + [-0.0931, 0.1106, -0.0605, ..., 0.0526, -0.0631, -0.0961], + [-0.0412, 0.0873, -0.1336, ..., -0.0219, 0.0379, 0.1253], + ..., + [ 0.0018, 0.0744, 0.0011, ..., -0.0260, 0.0318, -0.0862], + [ 0.1013, -0.1161, 0.0478, ..., 0.0201, 0.0122, -0.0346], + [-0.0128, -0.0888, 0.0737, ..., -0.0938, 0.0336, -0.0296]], + device='cuda:0'), grad: tensor([[ 2.5854e-05, 6.5453e-06, 6.4559e-06, ..., 6.2697e-06, + 5.0589e-06, 3.2596e-08], + [ 3.1918e-05, 4.9286e-03, 2.1496e-03, ..., 1.2146e-02, + 1.5154e-03, 5.5462e-05], + [ 1.1343e-04, 5.0621e-03, 1.7002e-05, ..., 3.0503e-05, + 5.1231e-03, -6.3062e-05], + ..., + [ 2.6188e-03, -1.0902e-02, 4.1161e-03, ..., -1.2367e-02, + -7.4730e-03, 1.6745e-06], + [-5.2643e-02, 2.2089e-04, 4.2229e-03, ..., 1.8511e-03, + 2.1243e-04, 2.4736e-06], + [ 5.4169e-02, 1.3733e-04, 6.2943e-03, ..., 1.9944e-04, + 1.2815e-04, 1.5832e-08]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0144, -0.0009, -0.0134, 0.0088, -0.0153, -0.0215, 0.0093, -0.0018, + 0.0132, -0.0114], device='cuda:0'), grad: tensor([ 0.0006, 0.0346, 0.0149, -0.0404, 0.0046, 0.0024, 0.0005, -0.0420, + -0.0118, 0.0365], device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 216.49, cls_loss 0.4488 cls_loss_mapping 0.0017 cls_loss_causal 0.4215 re_mapping 0.0053 re_causal 0.0143 /// teacc 98.72 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.0748, -0.1090, -0.1533, ..., 0.1266, -0.0645, 0.0375], + [-0.0933, 0.1109, -0.0607, ..., 0.0523, -0.0640, -0.0961], + [-0.0406, 0.0870, -0.1329, ..., -0.0225, 0.0384, 0.1259], + ..., + [ 0.0008, 0.0749, 0.0006, ..., -0.0259, 0.0316, -0.0862], + [ 0.1022, -0.1168, 0.0458, ..., 0.0207, 0.0099, -0.0362], + [-0.0121, -0.0882, 0.0743, ..., -0.0945, 0.0339, -0.0286]], + device='cuda:0'), grad: tensor([[ 6.6710e-04, 9.1553e-05, 1.4055e-04, ..., 2.8944e-04, + -1.7233e-03, -5.0354e-04], + [ 2.4748e-04, -2.1667e-03, 8.0299e-04, ..., -8.0032e-03, + 3.6687e-05, -2.5225e-04], + [ 8.3208e-04, 3.1567e-04, 5.7173e-04, ..., 1.4315e-03, + 3.6049e-04, 1.0738e-03], + ..., + [ 5.5790e-04, 2.7943e-04, 1.4696e-03, ..., 4.7417e-03, + 2.3496e-04, 7.4267e-05], + [-3.7937e-03, 2.6393e-04, -1.7807e-02, ..., -2.2430e-03, + -6.2599e-03, 2.1720e-04], + [ 1.2054e-03, 6.6805e-04, -3.0766e-03, ..., 8.6355e-04, + -1.3847e-03, 3.8314e-04]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0152, -0.0013, -0.0132, 0.0096, -0.0147, -0.0218, 0.0094, -0.0028, + 0.0130, -0.0119], device='cuda:0'), grad: tensor([-0.0160, -0.0542, 0.0349, -0.0306, 0.0497, -0.0058, 0.0266, 0.0008, + -0.0211, 0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 217.02, cls_loss 0.4452 cls_loss_mapping 0.0020 cls_loss_causal 0.4213 re_mapping 0.0052 re_causal 0.0135 /// teacc 98.97 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.0755, -0.1092, -0.1541, ..., 0.1268, -0.0657, 0.0374], + [-0.0938, 0.1112, -0.0611, ..., 0.0538, -0.0643, -0.0959], + [-0.0398, 0.0876, -0.1341, ..., -0.0226, 0.0382, 0.1256], + ..., + [ 0.0014, 0.0749, 0.0014, ..., -0.0260, 0.0324, -0.0866], + [ 0.1023, -0.1179, 0.0466, ..., 0.0199, 0.0104, -0.0371], + [-0.0127, -0.0890, 0.0745, ..., -0.0957, 0.0330, -0.0290]], + device='cuda:0'), grad: tensor([[-9.5725e-05, -4.7356e-05, 1.4043e-04, ..., -2.8819e-05, + -1.4663e-04, -5.5999e-05], + [-4.2152e-03, 1.8820e-05, -3.7270e-03, ..., 7.9155e-04, + -2.1019e-03, 2.6718e-05], + [ 1.8090e-05, -3.3665e-04, 8.6248e-05, ..., 1.6761e-04, + 3.4302e-05, -4.8590e-04], + ..., + [ 2.5872e-06, 4.3243e-05, 5.4121e-04, ..., 9.0301e-05, + 3.4308e-04, 6.1035e-05], + [ 3.8967e-03, 3.0115e-05, 3.9177e-03, ..., 4.2379e-05, + 2.0695e-03, 4.2528e-05], + [ 9.6619e-05, 2.7925e-05, -1.4133e-03, ..., 9.5725e-05, + -8.6212e-04, 3.6895e-05]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0151, -0.0027, -0.0134, 0.0104, -0.0160, -0.0214, 0.0105, -0.0027, + 0.0130, -0.0113], device='cuda:0'), grad: tensor([-0.0195, -0.0444, 0.0110, 0.0169, 0.0110, 0.0128, 0.0123, -0.0122, + 0.0100, 0.0021], device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 216.73, cls_loss 0.4659 cls_loss_mapping 0.0022 cls_loss_causal 0.4385 re_mapping 0.0055 re_causal 0.0150 /// teacc 98.87 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.0746, -0.1092, -0.1520, ..., 0.1264, -0.0659, 0.0364], + [-0.0931, 0.1108, -0.0599, ..., 0.0523, -0.0644, -0.0962], + [-0.0403, 0.0884, -0.1328, ..., -0.0223, 0.0404, 0.1258], + ..., + [ 0.0020, 0.0737, 0.0011, ..., -0.0254, 0.0318, -0.0866], + [ 0.1022, -0.1184, 0.0461, ..., 0.0201, 0.0096, -0.0367], + [-0.0122, -0.0882, 0.0743, ..., -0.0959, 0.0344, -0.0287]], + device='cuda:0'), grad: tensor([[ 7.7486e-06, 2.4270e-06, 8.2254e-06, ..., 2.2039e-05, + -1.0597e-02, 3.3639e-06], + [-2.5496e-05, -7.6830e-05, -3.5614e-06, ..., -4.2963e-04, + 2.7064e-06, 2.1812e-06], + [ 5.1223e-06, 8.4862e-06, 5.9381e-06, ..., 5.3704e-05, + 1.1339e-03, 2.8324e-04], + ..., + [ 2.0787e-06, 1.1660e-06, 8.2254e-06, ..., 1.8060e-04, + 9.7603e-06, 3.3490e-06], + [ 1.5092e-04, 1.8716e-05, 4.2915e-06, ..., 6.2287e-05, + -5.4893e-03, -4.4403e-03], + [ 6.0871e-06, 1.0841e-05, 5.0440e-06, ..., 2.7552e-05, + 8.8196e-03, 1.9103e-05]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0144, -0.0027, -0.0137, 0.0100, -0.0149, -0.0217, 0.0110, -0.0036, + 0.0130, -0.0104], device='cuda:0'), grad: tensor([-0.0266, 0.0157, 0.0205, -0.0059, 0.0153, 0.0132, 0.0148, -0.0152, + -0.0259, -0.0060], device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 216.93, cls_loss 0.4413 cls_loss_mapping 0.0024 cls_loss_causal 0.4190 re_mapping 0.0060 re_causal 0.0157 /// teacc 98.81 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.0745, -0.1092, -0.1526, ..., 0.1272, -0.0650, 0.0365], + [-0.0925, 0.1118, -0.0595, ..., 0.0525, -0.0649, -0.0961], + [-0.0402, 0.0875, -0.1329, ..., -0.0218, 0.0420, 0.1256], + ..., + [ 0.0015, 0.0744, 0.0019, ..., -0.0252, 0.0332, -0.0872], + [ 0.1021, -0.1183, 0.0456, ..., 0.0202, 0.0090, -0.0375], + [-0.0124, -0.0875, 0.0743, ..., -0.0973, 0.0340, -0.0287]], + device='cuda:0'), grad: tensor([[ 2.4867e-04, 8.6054e-06, 8.7070e-04, ..., 2.7728e-04, + 1.9464e-03, 1.8978e-04], + [ 7.7105e-04, 6.5947e-04, 2.0771e-03, ..., 9.6440e-05, + 2.2328e-04, 1.1736e-04], + [ 1.7858e-04, 9.7132e-04, 3.5095e-04, ..., 1.8835e-04, + 3.0155e-03, 1.6403e-03], + ..., + [ 4.4680e-04, -1.5652e-04, 1.2674e-03, ..., 8.5473e-05, + 4.3449e-03, 1.6642e-04], + [ 4.5657e-04, 1.6406e-05, 2.6083e-04, ..., 3.5977e-04, + 1.0490e-03, 1.1605e-04], + [-6.1512e-04, 5.0592e-04, -5.1384e-03, ..., 1.1063e-04, + -9.5081e-04, -2.3115e-04]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0143, -0.0024, -0.0140, 0.0097, -0.0148, -0.0219, 0.0104, -0.0023, + 0.0130, -0.0108], device='cuda:0'), grad: tensor([ 0.0202, 0.0209, 0.0236, 0.0046, -0.0420, -0.0753, 0.0049, 0.0207, + 0.0203, 0.0020], device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 217.16, cls_loss 0.4816 cls_loss_mapping 0.0018 cls_loss_causal 0.4638 re_mapping 0.0056 re_causal 0.0155 /// teacc 98.78 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.0746, -0.1093, -0.1526, ..., 0.1272, -0.0649, 0.0365], + [-0.0925, 0.1117, -0.0595, ..., 0.0523, -0.0649, -0.0961], + [-0.0402, 0.0876, -0.1329, ..., -0.0216, 0.0420, 0.1255], + ..., + [ 0.0016, 0.0744, 0.0019, ..., -0.0251, 0.0331, -0.0870], + [ 0.1022, -0.1182, 0.0456, ..., 0.0204, 0.0089, -0.0375], + [-0.0123, -0.0875, 0.0744, ..., -0.0973, 0.0341, -0.0287]], + device='cuda:0'), grad: tensor([[ 1.2326e-04, 2.2352e-06, 1.4110e-07, ..., 2.1219e-04, + -7.2233e-06, 3.4499e-04], + [ 1.2875e-03, -1.7273e-04, 1.0384e-07, ..., 3.0842e-03, + 1.2480e-06, 1.2867e-05], + [ 7.6234e-05, 5.1641e-04, 1.5631e-05, ..., 3.7909e-04, + 1.6198e-05, 1.0633e-04], + ..., + [-2.8804e-05, -4.8184e-04, -1.2076e-04, ..., 4.5347e-04, + -1.1581e-04, 3.9414e-06], + [ 1.9515e-04, 6.3062e-05, 1.6704e-05, ..., 6.1083e-04, + 1.3731e-05, 2.6062e-05], + [-2.2564e-03, 2.7597e-05, -2.4676e-05, ..., -5.1117e-03, + 2.8431e-05, 1.5706e-05]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0144, -0.0024, -0.0140, 0.0096, -0.0148, -0.0220, 0.0104, -0.0022, + 0.0131, -0.0108], device='cuda:0'), grad: tensor([-0.0134, 0.0001, -0.0133, 0.0345, 0.0170, -0.0161, 0.0079, 0.0011, + 0.0183, -0.0363], device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 217.30, cls_loss 0.4454 cls_loss_mapping 0.0017 cls_loss_causal 0.4106 re_mapping 0.0054 re_causal 0.0146 /// teacc 98.87 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.0747, -0.1092, -0.1527, ..., 0.1270, -0.0651, 0.0365], + [-0.0925, 0.1120, -0.0595, ..., 0.0522, -0.0650, -0.0962], + [-0.0402, 0.0875, -0.1330, ..., -0.0217, 0.0417, 0.1252], + ..., + [ 0.0016, 0.0742, 0.0018, ..., -0.0251, 0.0331, -0.0868], + [ 0.1021, -0.1182, 0.0456, ..., 0.0206, 0.0094, -0.0374], + [-0.0122, -0.0876, 0.0744, ..., -0.0972, 0.0339, -0.0288]], + device='cuda:0'), grad: tensor([[ 5.9336e-05, 5.9426e-05, 4.2081e-04, ..., -1.3611e-02, + -1.1740e-03, 4.5627e-05], + [ 2.0027e-04, -4.7989e-03, 2.2817e-04, ..., 3.0346e-03, + 2.7061e-04, 9.6977e-05], + [ 7.9751e-05, 4.2305e-03, 8.4698e-05, ..., -5.9938e-04, + 3.9220e-05, -5.5027e-04], + ..., + [ 4.8220e-05, -8.5235e-05, 1.3752e-03, ..., 2.8396e-04, + 6.5136e-04, 1.3721e-04], + [ 1.2684e-04, 1.8978e-04, 1.9398e-03, ..., 2.3861e-03, + 1.0080e-03, 1.5330e-04], + [ 1.2314e-04, 5.9009e-05, 6.8474e-04, ..., 1.5812e-03, + 5.6076e-04, 1.1182e-04]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0144, -0.0024, -0.0141, 0.0095, -0.0150, -0.0219, 0.0105, -0.0022, + 0.0131, -0.0107], device='cuda:0'), grad: tensor([-0.0427, -0.0018, 0.0235, 0.0168, 0.0087, -0.0150, 0.0324, -0.0172, + -0.0143, 0.0096], device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 217.00, cls_loss 0.4653 cls_loss_mapping 0.0009 cls_loss_causal 0.4383 re_mapping 0.0051 re_causal 0.0149 /// teacc 98.87 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.0748, -0.1093, -0.1528, ..., 0.1271, -0.0653, 0.0366], + [-0.0927, 0.1119, -0.0596, ..., 0.0520, -0.0651, -0.0964], + [-0.0402, 0.0876, -0.1329, ..., -0.0214, 0.0417, 0.1251], + ..., + [ 0.0016, 0.0743, 0.0018, ..., -0.0252, 0.0330, -0.0868], + [ 0.1023, -0.1183, 0.0459, ..., 0.0205, 0.0095, -0.0374], + [-0.0121, -0.0877, 0.0743, ..., -0.0972, 0.0339, -0.0288]], + device='cuda:0'), grad: tensor([[ 6.1178e-04, 1.9944e-04, 3.1982e-06, ..., 9.1076e-05, + -1.8686e-05, -6.8903e-05], + [ 6.4049e-03, 3.5458e-03, -8.0884e-05, ..., 1.0180e-04, + 1.2927e-06, 9.0122e-05], + [ 2.0027e-03, 7.4625e-04, 1.8617e-06, ..., 9.8050e-05, + 6.9261e-05, 1.4591e-04], + ..., + [-1.2520e-02, -9.7275e-03, -5.0402e-04, ..., -5.0604e-05, + 4.0263e-05, 1.2290e-04], + [-1.6083e-02, -3.5515e-03, 1.6645e-05, ..., -3.3283e-04, + -1.1034e-03, 1.2314e-04], + [ 8.6594e-03, 6.5956e-03, 5.6028e-05, ..., 2.4676e-04, + -3.1853e-04, -8.8406e-04]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0143, -0.0025, -0.0140, 0.0097, -0.0149, -0.0220, 0.0105, -0.0022, + 0.0132, -0.0108], device='cuda:0'), grad: tensor([ 0.0128, 0.0245, 0.0211, 0.0148, -0.0121, 0.0109, 0.0100, -0.0016, + -0.0426, -0.0379], device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 217.10, cls_loss 0.4555 cls_loss_mapping 0.0012 cls_loss_causal 0.4287 re_mapping 0.0051 re_causal 0.0143 /// teacc 98.86 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.0749, -0.1093, -0.1529, ..., 0.1272, -0.0653, 0.0366], + [-0.0927, 0.1118, -0.0594, ..., 0.0519, -0.0650, -0.0965], + [-0.0401, 0.0876, -0.1329, ..., -0.0214, 0.0416, 0.1250], + ..., + [ 0.0017, 0.0743, 0.0017, ..., -0.0253, 0.0330, -0.0868], + [ 0.1024, -0.1184, 0.0461, ..., 0.0207, 0.0096, -0.0373], + [-0.0122, -0.0877, 0.0743, ..., -0.0972, 0.0338, -0.0287]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0002, 0.0011, ..., 0.0010, 0.0014, -0.0002], + [ 0.0003, 0.0004, 0.0003, ..., 0.0009, 0.0005, 0.0006], + [ 0.0009, 0.0026, 0.0036, ..., 0.0006, 0.0136, 0.0048], + ..., + [-0.0011, -0.0028, -0.0017, ..., -0.0019, -0.0117, -0.0021], + [-0.0007, 0.0003, 0.0009, ..., 0.0011, 0.0012, 0.0004], + [ 0.0003, -0.0019, -0.0055, ..., -0.0041, -0.0078, -0.0002]], + device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0145, -0.0026, -0.0140, 0.0098, -0.0151, -0.0221, 0.0105, -0.0022, + 0.0134, -0.0109], device='cuda:0'), grad: tensor([-0.0374, -0.0054, 0.0745, 0.0179, 0.0026, -0.0021, -0.0215, -0.0306, + 0.0261, -0.0240], device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 216.70, cls_loss 0.4643 cls_loss_mapping 0.0011 cls_loss_causal 0.4317 re_mapping 0.0048 re_causal 0.0136 /// teacc 98.88 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.0750, -0.1094, -0.1530, ..., 0.1272, -0.0654, 0.0366], + [-0.0927, 0.1118, -0.0594, ..., 0.0519, -0.0650, -0.0965], + [-0.0400, 0.0877, -0.1330, ..., -0.0214, 0.0414, 0.1247], + ..., + [ 0.0017, 0.0743, 0.0018, ..., -0.0252, 0.0331, -0.0868], + [ 0.1023, -0.1184, 0.0461, ..., 0.0207, 0.0097, -0.0372], + [-0.0121, -0.0876, 0.0743, ..., -0.0971, 0.0338, -0.0286]], + device='cuda:0'), grad: tensor([[ 2.4214e-05, 2.3544e-05, 1.0952e-06, ..., 7.3195e-04, + -3.1447e-04, 1.9217e-04], + [ 1.5080e-04, 8.3297e-06, 1.1884e-06, ..., 2.0733e-03, + 8.2433e-05, 9.4950e-05], + [ 2.1979e-05, -1.3828e-04, 1.2882e-05, ..., 1.5182e-03, + 2.9898e-04, 4.9782e-04], + ..., + [-1.9148e-05, -3.7163e-05, -8.4534e-03, ..., 8.5592e-04, + -4.5433e-03, 1.3733e-04], + [-3.7432e-04, 1.9237e-05, 9.7416e-07, ..., -3.7365e-03, + 2.7871e-04, 5.3263e-04], + [ 9.7990e-05, 1.4149e-05, 8.4000e-03, ..., 1.2321e-03, + 4.6310e-03, 1.4675e-04]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0145, -0.0026, -0.0143, 0.0098, -0.0151, -0.0222, 0.0106, -0.0022, + 0.0135, -0.0109], device='cuda:0'), grad: tensor([-0.0214, 0.0167, 0.0166, 0.0300, -0.0199, 0.0156, -0.0153, 0.0002, + -0.0508, 0.0284], device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 216.63, cls_loss 0.4718 cls_loss_mapping 0.0010 cls_loss_causal 0.4452 re_mapping 0.0045 re_causal 0.0131 /// teacc 98.88 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.0749, -0.1094, -0.1531, ..., 0.1271, -0.0655, 0.0366], + [-0.0927, 0.1117, -0.0595, ..., 0.0519, -0.0649, -0.0965], + [-0.0399, 0.0877, -0.1329, ..., -0.0215, 0.0415, 0.1248], + ..., + [ 0.0018, 0.0744, 0.0017, ..., -0.0252, 0.0331, -0.0866], + [ 0.1022, -0.1185, 0.0460, ..., 0.0206, 0.0096, -0.0372], + [-0.0121, -0.0876, 0.0743, ..., -0.0971, 0.0337, -0.0287]], + device='cuda:0'), grad: tensor([[ 7.1335e-04, 1.0210e-04, 1.7667e-04, ..., 6.1226e-04, + 5.2071e-04, 5.2340e-06], + [ 5.5361e-04, -1.8549e-04, -7.5912e-04, ..., 1.2636e-03, + -1.2316e-05, 3.7067e-07], + [ 7.6723e-04, 1.8179e-04, 4.7040e-04, ..., 5.5695e-04, + 8.2493e-04, 4.8876e-05], + ..., + [ 1.1816e-03, 2.5582e-04, 8.4305e-04, ..., 7.4100e-04, + 1.0681e-03, 3.3993e-07], + [ 6.0940e-04, 4.4376e-05, 1.6809e-04, ..., 2.1327e-04, + 1.3351e-04, -1.4022e-05], + [-9.1028e-04, 9.3400e-05, -1.1243e-05, ..., 4.8256e-04, + -1.4992e-03, 4.3735e-06]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0145, -0.0024, -0.0142, 0.0098, -0.0149, -0.0223, 0.0106, -0.0022, + 0.0134, -0.0110], device='cuda:0'), grad: tensor([ 0.0145, 0.0189, 0.0149, 0.0101, -0.0281, -0.0246, -0.0434, 0.0188, + 0.0131, 0.0057], device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 217.16, cls_loss 0.4376 cls_loss_mapping 0.0009 cls_loss_causal 0.4133 re_mapping 0.0047 re_causal 0.0131 /// teacc 98.89 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.0751, -0.1094, -0.1532, ..., 0.1270, -0.0655, 0.0366], + [-0.0927, 0.1119, -0.0594, ..., 0.0521, -0.0649, -0.0965], + [-0.0400, 0.0876, -0.1329, ..., -0.0216, 0.0414, 0.1247], + ..., + [ 0.0017, 0.0744, 0.0016, ..., -0.0252, 0.0331, -0.0864], + [ 0.1023, -0.1185, 0.0460, ..., 0.0206, 0.0096, -0.0372], + [-0.0120, -0.0875, 0.0744, ..., -0.0970, 0.0337, -0.0288]], + device='cuda:0'), grad: tensor([[ 1.1301e-04, 1.3220e-04, 2.2125e-04, ..., 4.5300e-04, + -3.6806e-05, -2.2307e-05], + [ 3.1018e-04, 2.5463e-03, 6.4230e-04, ..., 6.6996e-04, + 7.6741e-06, 3.3760e-07], + [ 1.3745e-04, -6.8130e-03, 2.7204e-04, ..., 4.6611e-05, + 8.2850e-06, -1.3612e-05], + ..., + [ 1.0843e-03, 8.2850e-05, 2.3403e-03, ..., -1.5297e-03, + -6.0886e-05, -2.6356e-06], + [ 6.5947e-04, 1.3924e-03, 1.4114e-03, ..., 6.8283e-04, + 2.3901e-05, 5.2676e-06], + [-5.6343e-03, 7.2300e-05, -1.4671e-02, ..., 7.5054e-04, + -7.7963e-04, 3.4031e-06]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0144, -0.0024, -0.0143, 0.0097, -0.0149, -0.0222, 0.0106, -0.0022, + 0.0135, -0.0109], device='cuda:0'), grad: tensor([ 0.0100, -0.0159, 0.0008, -0.0179, 0.0275, -0.0148, 0.0143, -0.0068, + 0.0178, -0.0151], device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 217.18, cls_loss 0.4374 cls_loss_mapping 0.0011 cls_loss_causal 0.4186 re_mapping 0.0046 re_causal 0.0130 /// teacc 98.93 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.0751, -0.1096, -0.1533, ..., 0.1271, -0.0656, 0.0367], + [-0.0927, 0.1120, -0.0594, ..., 0.0521, -0.0649, -0.0964], + [-0.0400, 0.0876, -0.1329, ..., -0.0216, 0.0413, 0.1246], + ..., + [ 0.0018, 0.0744, 0.0017, ..., -0.0252, 0.0334, -0.0862], + [ 0.1022, -0.1186, 0.0458, ..., 0.0205, 0.0095, -0.0373], + [-0.0120, -0.0877, 0.0744, ..., -0.0968, 0.0338, -0.0288]], + device='cuda:0'), grad: tensor([[ 2.6131e-03, 1.6761e-04, 9.4995e-08, ..., 2.9349e-04, + -1.4175e-06, 2.3860e-06], + [ 1.8377e-03, 1.0443e-03, 2.0154e-06, ..., 1.6842e-03, + 1.6671e-06, 1.2591e-06], + [ 5.1346e-03, -2.6679e-04, 5.4110e-07, ..., 2.0695e-04, + 1.2405e-05, 1.3351e-05], + ..., + [ 4.0855e-03, 2.3782e-04, 5.4245e-03, ..., -1.8177e-03, + 1.8463e-03, 4.8755e-07], + [ 2.7943e-03, 3.3569e-04, 6.7102e-07, ..., 6.0034e-04, + 7.7561e-06, 8.2105e-06], + [ 4.2915e-03, 1.1909e-04, -5.6686e-03, ..., 1.5278e-03, + -1.9741e-03, 1.4305e-06]], device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0144, -0.0023, -0.0144, 0.0095, -0.0149, -0.0221, 0.0106, -0.0021, + 0.0134, -0.0109], device='cuda:0'), grad: tensor([ 0.0147, 0.0275, 0.0153, -0.0182, -0.0470, -0.0146, 0.0076, -0.0224, + 0.0170, 0.0202], device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 217.08, cls_loss 0.4466 cls_loss_mapping 0.0012 cls_loss_causal 0.4184 re_mapping 0.0046 re_causal 0.0132 /// teacc 98.91 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.0750, -0.1096, -0.1532, ..., 0.1272, -0.0655, 0.0367], + [-0.0927, 0.1121, -0.0594, ..., 0.0520, -0.0649, -0.0963], + [-0.0400, 0.0876, -0.1329, ..., -0.0216, 0.0414, 0.1248], + ..., + [ 0.0018, 0.0744, 0.0016, ..., -0.0252, 0.0334, -0.0862], + [ 0.1022, -0.1187, 0.0458, ..., 0.0205, 0.0093, -0.0373], + [-0.0120, -0.0876, 0.0744, ..., -0.0968, 0.0338, -0.0290]], + device='cuda:0'), grad: tensor([[ 1.9825e-04, 2.1622e-05, 1.4019e-04, ..., 6.7759e-04, + 5.7173e-04, 5.3905e-06], + [ 2.2662e-04, -3.9506e-04, 2.6420e-05, ..., -5.3692e-04, + 3.7122e-04, 2.8446e-05], + [ 2.1005e-04, 4.8733e-04, -3.0923e-04, ..., -4.0169e-03, + -8.5297e-03, 9.9087e-04], + ..., + [ 4.3488e-04, 6.0797e-05, 1.3405e-02, ..., 1.4324e-03, + 4.2038e-03, 1.5557e-04], + [ 1.8311e-04, -4.8923e-04, -1.6966e-03, ..., -1.1969e-03, + 9.3699e-04, -1.4153e-03], + [ 1.0462e-03, 1.5348e-05, 3.0041e-03, ..., 1.1253e-03, + -1.9207e-03, 3.9160e-05]], device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0146, -0.0025, -0.0144, 0.0095, -0.0150, -0.0221, 0.0107, -0.0020, + 0.0135, -0.0109], device='cuda:0'), grad: tensor([ 0.0222, -0.0127, -0.0244, -0.0041, -0.0249, -0.0030, 0.0214, 0.0452, + -0.0143, -0.0054], device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 217.54, cls_loss 0.4233 cls_loss_mapping 0.0009 cls_loss_causal 0.3938 re_mapping 0.0046 re_causal 0.0131 /// teacc 98.90 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.0751, -0.1096, -0.1533, ..., 0.1272, -0.0655, 0.0368], + [-0.0928, 0.1120, -0.0593, ..., 0.0522, -0.0649, -0.0964], + [-0.0400, 0.0877, -0.1329, ..., -0.0216, 0.0414, 0.1248], + ..., + [ 0.0018, 0.0744, 0.0016, ..., -0.0251, 0.0333, -0.0862], + [ 0.1021, -0.1187, 0.0460, ..., 0.0204, 0.0095, -0.0372], + [-0.0119, -0.0876, 0.0743, ..., -0.0969, 0.0337, -0.0291]], + device='cuda:0'), grad: tensor([[ 2.9087e-05, 5.4613e-06, 1.0699e-05, ..., 7.7367e-05, + 1.1280e-05, 8.5533e-05], + [-1.2598e-03, -4.1151e-04, 5.4836e-05, ..., -8.3089e-05, + 2.9311e-05, 7.1943e-05], + [ 1.1975e-04, 3.4899e-05, 5.5254e-05, ..., -2.1591e-03, + 1.3685e-04, -7.2193e-04], + ..., + [ 5.0336e-05, -5.6171e-04, -1.3800e-03, ..., -1.8282e-03, + -8.9455e-04, 6.7711e-05], + [ 1.0319e-03, 3.6526e-04, 4.4060e-04, ..., 2.6932e-03, + 2.0421e-04, 1.1339e-03], + [ 4.1962e-05, 4.3344e-04, 4.1032e-04, ..., 1.4133e-03, + 2.3353e-04, 6.8724e-05]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0144, -0.0023, -0.0144, 0.0095, -0.0150, -0.0221, 0.0107, -0.0021, + 0.0135, -0.0109], device='cuda:0'), grad: tensor([ 8.2855e-03, 8.2245e-03, -1.6815e-02, -1.9394e-02, 1.1108e-02, + -2.4872e-02, -2.2690e-02, -1.1876e-05, 4.4342e-02, 1.1803e-02], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 216.63, cls_loss 0.4717 cls_loss_mapping 0.0010 cls_loss_causal 0.4419 re_mapping 0.0045 re_causal 0.0135 /// teacc 98.91 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.0752, -0.1096, -0.1534, ..., 0.1271, -0.0656, 0.0368], + [-0.0929, 0.1120, -0.0593, ..., 0.0523, -0.0651, -0.0964], + [-0.0400, 0.0877, -0.1331, ..., -0.0218, 0.0412, 0.1247], + ..., + [ 0.0017, 0.0744, 0.0017, ..., -0.0252, 0.0334, -0.0861], + [ 0.1020, -0.1187, 0.0461, ..., 0.0205, 0.0096, -0.0373], + [-0.0120, -0.0877, 0.0743, ..., -0.0968, 0.0338, -0.0291]], + device='cuda:0'), grad: tensor([[-4.6387e-03, 1.8746e-05, -6.5708e-04, ..., 1.2565e-04, + -4.2686e-03, -1.2627e-03], + [ 2.1505e-04, -1.5249e-03, -1.9026e-03, ..., -2.4204e-03, + -1.4696e-03, 3.7998e-05], + [ 5.5504e-04, -5.4264e-04, 2.9397e-04, ..., -3.3927e-04, + 1.2188e-03, 3.6216e-04], + ..., + [ 4.8327e-04, 1.0481e-03, 1.4439e-03, ..., 1.0538e-03, + 1.4057e-03, 6.0260e-05], + [ 6.7282e-04, 4.6611e-05, 4.9114e-04, ..., 1.1331e-04, + 1.6088e-03, -3.2157e-05], + [ 5.4455e-04, 7.2670e-04, 2.2621e-03, ..., 1.2379e-03, + 1.5326e-03, 7.0620e-04]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0144, -0.0022, -0.0147, 0.0094, -0.0149, -0.0220, 0.0109, -0.0021, + 0.0133, -0.0108], device='cuda:0'), grad: tensor([-0.0226, 0.0084, 0.0115, -0.0050, -0.0376, 0.0168, 0.0307, -0.0125, + -0.0132, 0.0236], device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 217.73, cls_loss 0.4317 cls_loss_mapping 0.0012 cls_loss_causal 0.4071 re_mapping 0.0043 re_causal 0.0123 /// teacc 98.93 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.0751, -0.1094, -0.1535, ..., 0.1271, -0.0656, 0.0367], + [-0.0929, 0.1120, -0.0592, ..., 0.0523, -0.0650, -0.0964], + [-0.0399, 0.0878, -0.1332, ..., -0.0216, 0.0411, 0.1248], + ..., + [ 0.0018, 0.0744, 0.0017, ..., -0.0254, 0.0334, -0.0860], + [ 0.1021, -0.1187, 0.0460, ..., 0.0205, 0.0095, -0.0373], + [-0.0119, -0.0879, 0.0744, ..., -0.0968, 0.0340, -0.0291]], + device='cuda:0'), grad: tensor([[ 4.8923e-04, 3.9864e-04, 8.0156e-04, ..., 2.8038e-03, + 4.5657e-04, 6.8009e-05], + [ 7.9489e-04, 1.6031e-03, 6.1178e-04, ..., 1.1147e-02, + 2.9659e-04, 8.4519e-05], + [ 4.3392e-04, 1.5330e-04, 7.5912e-03, ..., 1.2026e-03, + 3.0613e-04, 6.5498e-03], + ..., + [-1.5736e-03, -1.1272e-03, -5.4817e-03, ..., -7.1564e-03, + -3.0708e-03, -1.4009e-03], + [ 6.3276e-04, 5.7364e-04, 6.3515e-04, ..., 3.9711e-03, + 3.7599e-04, 1.8537e-04], + [ 5.2691e-04, 2.0051e-04, -7.2174e-03, ..., 1.2550e-03, + -5.1613e-03, 2.4700e-04]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0144, -0.0023, -0.0147, 0.0095, -0.0149, -0.0221, 0.0109, -0.0021, + 0.0134, -0.0110], device='cuda:0'), grad: tensor([-0.0131, 0.0027, 0.0385, -0.0109, 0.0218, 0.0143, 0.0122, -0.0951, + 0.0220, 0.0076], device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 217.78, cls_loss 0.4647 cls_loss_mapping 0.0009 cls_loss_causal 0.4330 re_mapping 0.0043 re_causal 0.0128 /// teacc 98.94 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.0749, -0.1091, -0.1534, ..., 0.1271, -0.0656, 0.0366], + [-0.0930, 0.1120, -0.0593, ..., 0.0522, -0.0651, -0.0965], + [-0.0399, 0.0879, -0.1333, ..., -0.0216, 0.0412, 0.1249], + ..., + [ 0.0018, 0.0744, 0.0017, ..., -0.0253, 0.0333, -0.0859], + [ 0.1022, -0.1186, 0.0460, ..., 0.0205, 0.0095, -0.0374], + [-0.0121, -0.0880, 0.0743, ..., -0.0967, 0.0340, -0.0291]], + device='cuda:0'), grad: tensor([[ 5.3078e-05, 4.4167e-05, 1.1593e-04, ..., 1.2767e-04, + 2.0182e-04, 1.1230e-04], + [ 5.5104e-05, 3.8534e-05, 5.7220e-05, ..., -5.4646e-04, + 9.3222e-05, 5.1618e-05], + [ 7.0751e-05, -6.3229e-04, 4.0913e-04, ..., 4.1910e-07, + 8.1968e-04, 1.0568e-04], + ..., + [-1.2503e-03, -3.1519e-04, -9.5558e-04, ..., -6.5088e-04, + -1.5850e-03, -5.0926e-04], + [ 1.8883e-04, 1.5342e-04, 3.1972e-04, ..., 8.1396e-04, + 5.0402e-04, 2.4045e-04], + [ 5.8079e-04, 4.3225e-04, 9.7036e-04, ..., 5.7840e-04, + 8.7118e-04, 4.0293e-04]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0146, -0.0024, -0.0144, 0.0095, -0.0150, -0.0220, 0.0110, -0.0021, + 0.0132, -0.0110], device='cuda:0'), grad: tensor([ 0.0049, 0.0009, -0.0238, -0.0080, 0.0020, 0.0055, 0.0072, -0.0083, + 0.0095, 0.0103], device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 218.35, cls_loss 0.4807 cls_loss_mapping 0.0011 cls_loss_causal 0.4559 re_mapping 0.0042 re_causal 0.0126 /// teacc 98.99 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.0748, -0.1090, -0.1534, ..., 0.1271, -0.0656, 0.0367], + [-0.0930, 0.1119, -0.0593, ..., 0.0522, -0.0651, -0.0965], + [-0.0400, 0.0878, -0.1334, ..., -0.0216, 0.0412, 0.1250], + ..., + [ 0.0018, 0.0743, 0.0016, ..., -0.0253, 0.0332, -0.0860], + [ 0.1023, -0.1185, 0.0461, ..., 0.0206, 0.0096, -0.0374], + [-0.0121, -0.0880, 0.0744, ..., -0.0967, 0.0340, -0.0293]], + device='cuda:0'), grad: tensor([[ 2.2423e-04, 1.1707e-06, 3.1263e-05, ..., 1.6832e-04, + 1.0721e-05, -6.0946e-06], + [ 3.2163e-04, 8.3148e-06, -1.9050e-04, ..., 4.5747e-05, + 1.2694e-06, 3.0734e-08], + [ 2.4962e-04, 2.2769e-05, 2.1979e-05, ..., 1.6439e-04, + 1.5214e-05, 4.0233e-07], + ..., + [ 3.2473e-04, -3.8376e-03, 1.3151e-03, ..., 2.0206e-04, + 1.0490e-03, 2.9709e-07], + [-2.0466e-03, 6.4932e-06, 5.4121e-05, ..., 1.6916e-04, + 2.8774e-05, 8.0373e-07], + [ 2.1744e-02, 3.6373e-03, -1.8349e-03, ..., -1.1003e-04, + -1.2541e-03, 1.8468e-06]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0147, -0.0024, -0.0144, 0.0093, -0.0151, -0.0219, 0.0109, -0.0020, + 0.0133, -0.0110], device='cuda:0'), grad: tensor([ 1.3054e-02, 1.5221e-02, 1.2199e-02, -2.1988e-02, -3.7903e-02, + 2.8076e-02, -1.6907e-02, 1.8388e-05, -7.1144e-03, 1.5350e-02], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 217.48, cls_loss 0.4195 cls_loss_mapping 0.0009 cls_loss_causal 0.3962 re_mapping 0.0042 re_causal 0.0123 /// teacc 98.96 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.0749, -0.1089, -0.1534, ..., 0.1271, -0.0656, 0.0368], + [-0.0929, 0.1120, -0.0593, ..., 0.0524, -0.0652, -0.0965], + [-0.0398, 0.0878, -0.1332, ..., -0.0215, 0.0412, 0.1250], + ..., + [ 0.0018, 0.0743, 0.0017, ..., -0.0254, 0.0334, -0.0861], + [ 0.1022, -0.1187, 0.0461, ..., 0.0206, 0.0096, -0.0375], + [-0.0121, -0.0881, 0.0743, ..., -0.0968, 0.0340, -0.0291]], + device='cuda:0'), grad: tensor([[ 2.8682e-04, 4.5486e-06, 5.9426e-05, ..., 2.2650e-04, + 3.7491e-05, 5.9232e-06], + [ 6.1846e-04, -2.3201e-05, 5.0128e-05, ..., -1.1120e-03, + 2.7895e-05, 1.0654e-06], + [ 1.4811e-03, 1.5426e-04, 1.5102e-05, ..., 2.9297e-03, + 5.2376e-03, 2.2392e-03], + ..., + [ 2.6083e-04, 2.5593e-06, 2.7790e-03, ..., 2.1136e-04, + 1.4763e-03, 3.5949e-07], + [ 1.4744e-03, 8.3029e-05, 5.6237e-05, ..., 1.6441e-03, + 2.7466e-03, 1.1654e-03], + [ 2.4092e-04, 1.1958e-06, 8.5144e-03, ..., 2.0790e-04, + 4.5547e-03, 1.6829e-06]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0147, -0.0023, -0.0143, 0.0095, -0.0150, -0.0221, 0.0108, -0.0020, + 0.0132, -0.0111], device='cuda:0'), grad: tensor([ 0.0115, -0.0153, 0.0132, -0.0463, -0.0220, 0.0135, -0.0442, 0.0202, + 0.0332, 0.0361], device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 217.19, cls_loss 0.4562 cls_loss_mapping 0.0010 cls_loss_causal 0.4299 re_mapping 0.0043 re_causal 0.0130 /// teacc 98.95 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.0751, -0.1090, -0.1534, ..., 0.1269, -0.0657, 0.0367], + [-0.0929, 0.1119, -0.0594, ..., 0.0525, -0.0650, -0.0966], + [-0.0397, 0.0881, -0.1331, ..., -0.0212, 0.0413, 0.1251], + ..., + [ 0.0018, 0.0743, 0.0017, ..., -0.0255, 0.0334, -0.0860], + [ 0.1022, -0.1187, 0.0459, ..., 0.0205, 0.0095, -0.0377], + [-0.0122, -0.0881, 0.0743, ..., -0.0968, 0.0339, -0.0291]], + device='cuda:0'), grad: tensor([[ 2.6178e-04, -8.4460e-05, 4.2510e-04, ..., 1.7381e-04, + 1.1301e-03, 6.3705e-04], + [ 1.5366e-04, 9.5654e-04, 8.6260e-04, ..., -3.1376e-04, + 1.2341e-03, 4.5323e-04], + [ 1.8075e-05, 9.0027e-04, -1.0004e-03, ..., 1.6704e-05, + -3.3150e-03, -6.8855e-04], + ..., + [ 5.8079e-04, -4.7722e-03, -2.2221e-03, ..., -3.2425e-03, + -2.6188e-03, 4.6706e-04], + [-1.6820e-04, 1.6558e-04, -6.6102e-05, ..., 5.3120e-04, + 4.8375e-04, 6.5899e-04], + [ 1.2846e-03, 9.5272e-04, 2.1381e-03, ..., 8.7452e-04, + 2.1706e-03, 4.5156e-04]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0146, -0.0023, -0.0141, 0.0095, -0.0150, -0.0220, 0.0109, -0.0021, + 0.0129, -0.0112], device='cuda:0'), grad: tensor([ 0.0106, 0.0122, 0.0112, -0.0285, 0.0195, -0.0107, -0.0022, -0.0489, + 0.0245, 0.0124], device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 217.10, cls_loss 0.4689 cls_loss_mapping 0.0011 cls_loss_causal 0.4464 re_mapping 0.0043 re_causal 0.0128 /// teacc 98.98 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.0751, -0.1091, -0.1535, ..., 0.1270, -0.0656, 0.0367], + [-0.0929, 0.1119, -0.0594, ..., 0.0525, -0.0650, -0.0967], + [-0.0398, 0.0880, -0.1332, ..., -0.0214, 0.0413, 0.1249], + ..., + [ 0.0018, 0.0743, 0.0018, ..., -0.0254, 0.0336, -0.0860], + [ 0.1022, -0.1186, 0.0459, ..., 0.0205, 0.0095, -0.0377], + [-0.0123, -0.0881, 0.0742, ..., -0.0969, 0.0339, -0.0291]], + device='cuda:0'), grad: tensor([[-0.0011, -0.0014, 0.0006, ..., -0.0013, 0.0014, 0.0003], + [-0.0038, 0.0031, -0.0019, ..., 0.0028, 0.0051, 0.0011], + [ 0.0005, -0.0007, 0.0006, ..., 0.0019, 0.0024, -0.0004], + ..., + [-0.0030, -0.0008, -0.0108, ..., 0.0008, -0.0073, 0.0001], + [-0.0030, 0.0007, 0.0010, ..., 0.0017, 0.0021, 0.0006], + [ 0.0043, -0.0024, 0.0147, ..., -0.0019, 0.0134, -0.0027]], + device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0146, -0.0023, -0.0142, 0.0096, -0.0150, -0.0218, 0.0109, -0.0021, + 0.0128, -0.0112], device='cuda:0'), grad: tensor([-0.0123, 0.0172, 0.0242, -0.0335, 0.0388, -0.0413, 0.0202, -0.0289, + 0.0136, 0.0021], device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 216.96, cls_loss 0.4502 cls_loss_mapping 0.0009 cls_loss_causal 0.4220 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.00 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.0751, -0.1091, -0.1536, ..., 0.1270, -0.0658, 0.0367], + [-0.0928, 0.1119, -0.0593, ..., 0.0524, -0.0649, -0.0968], + [-0.0397, 0.0881, -0.1332, ..., -0.0214, 0.0412, 0.1249], + ..., + [ 0.0018, 0.0742, 0.0017, ..., -0.0254, 0.0335, -0.0859], + [ 0.1022, -0.1184, 0.0459, ..., 0.0206, 0.0095, -0.0376], + [-0.0123, -0.0881, 0.0742, ..., -0.0969, 0.0339, -0.0291]], + device='cuda:0'), grad: tensor([[ 7.3537e-06, 9.2238e-06, 1.2971e-05, ..., 1.1390e-04, + 2.4724e-04, 1.2666e-07], + [-2.7084e-03, 7.0222e-06, 1.6165e-04, ..., -2.0809e-03, + 7.0524e-04, 3.2596e-07], + [ 7.1824e-06, -3.0327e-04, 1.5393e-05, ..., -1.2141e-04, + 1.9610e-04, -8.4758e-05], + ..., + [ 9.0525e-06, 1.3936e-04, 1.6403e-03, ..., -2.0599e-03, + -4.7493e-03, 5.6446e-05], + [ 7.3910e-05, 6.3777e-05, 1.5986e-04, ..., 6.1178e-04, + 1.1702e-03, 1.9819e-05], + [-2.3678e-05, 2.4736e-05, -4.1046e-03, ..., 3.8171e-04, + 8.7678e-05, 7.4599e-07]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0145, -0.0024, -0.0142, 0.0097, -0.0150, -0.0218, 0.0108, -0.0021, + 0.0129, -0.0112], device='cuda:0'), grad: tensor([-0.0159, -0.0301, -0.0133, 0.0307, 0.0184, 0.0105, 0.0123, -0.0068, + -0.0154, 0.0095], device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 217.34, cls_loss 0.4533 cls_loss_mapping 0.0011 cls_loss_causal 0.4231 re_mapping 0.0042 re_causal 0.0125 /// teacc 99.02 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.0749, -0.1092, -0.1536, ..., 0.1268, -0.0659, 0.0367], + [-0.0928, 0.1121, -0.0591, ..., 0.0526, -0.0649, -0.0968], + [-0.0397, 0.0880, -0.1333, ..., -0.0215, 0.0412, 0.1249], + ..., + [ 0.0018, 0.0743, 0.0016, ..., -0.0254, 0.0337, -0.0860], + [ 0.1022, -0.1184, 0.0460, ..., 0.0206, 0.0096, -0.0376], + [-0.0123, -0.0882, 0.0743, ..., -0.0969, 0.0340, -0.0292]], + device='cuda:0'), grad: tensor([[-4.9744e-03, 8.5294e-05, 5.3138e-05, ..., -9.8991e-04, + -2.5600e-05, 2.5225e-04], + [ 3.0446e-04, 1.1349e-03, 4.6998e-05, ..., 1.6270e-03, + 6.6638e-05, 9.6512e-04], + [ 3.5739e-04, -2.8591e-03, 3.6693e-04, ..., -2.4853e-03, + 2.2864e-04, 1.3857e-03], + ..., + [ 2.6202e-04, 8.1015e-04, -1.0147e-03, ..., 1.1311e-03, + -5.8794e-04, 5.3596e-04], + [ 2.0552e-04, 4.0627e-04, 9.0718e-05, ..., 6.8903e-04, + 6.7532e-05, -6.5079e-03], + [ 3.3522e-04, 1.2922e-04, 6.1333e-05, ..., 5.5224e-05, + -1.7107e-05, 1.9968e-04]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0145, -0.0024, -0.0141, 0.0097, -0.0152, -0.0216, 0.0108, -0.0021, + 0.0129, -0.0111], device='cuda:0'), grad: tensor([-0.0254, 0.0228, 0.0025, -0.0004, 0.0110, 0.0193, -0.0184, 0.0130, + -0.0079, -0.0166], device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 216.50, cls_loss 0.4407 cls_loss_mapping 0.0009 cls_loss_causal 0.4195 re_mapping 0.0042 re_causal 0.0123 /// teacc 99.00 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.0748, -0.1091, -0.1536, ..., 0.1269, -0.0658, 0.0368], + [-0.0929, 0.1120, -0.0591, ..., 0.0525, -0.0650, -0.0969], + [-0.0397, 0.0881, -0.1329, ..., -0.0214, 0.0415, 0.1248], + ..., + [ 0.0020, 0.0743, 0.0016, ..., -0.0253, 0.0337, -0.0860], + [ 0.1022, -0.1184, 0.0459, ..., 0.0206, 0.0095, -0.0376], + [-0.0122, -0.0882, 0.0743, ..., -0.0970, 0.0339, -0.0292]], + device='cuda:0'), grad: tensor([[ 7.0190e-04, 3.0518e-05, 3.0786e-05, ..., 2.6188e-03, + 2.5797e-04, -1.9703e-03], + [-4.8676e-03, 9.7394e-05, 1.4627e-04, ..., -4.6844e-03, + 1.1331e-04, 6.8724e-05], + [ 4.8041e-04, 5.1707e-05, 4.8667e-05, ..., 8.0204e-04, + 2.8944e-04, 1.8883e-04], + ..., + [ 2.8181e-04, -3.8370e-07, 4.9353e-05, ..., 9.5844e-04, + -4.3392e-05, 3.2932e-05], + [ 1.8358e-03, -4.9877e-04, 7.7438e-04, ..., -2.0580e-03, + 8.2922e-04, 2.7776e-04], + [ 2.4402e-04, 8.1897e-05, 8.6069e-05, ..., 8.9216e-04, + 2.7823e-04, 1.3471e-04]], device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0147, -0.0024, -0.0141, 0.0097, -0.0153, -0.0217, 0.0108, -0.0021, + 0.0129, -0.0113], device='cuda:0'), grad: tensor([ 0.0081, 0.0058, -0.0035, 0.0031, 0.0250, -0.0212, -0.0199, -0.0031, + -0.0231, 0.0288], device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 217.01, cls_loss 0.4684 cls_loss_mapping 0.0010 cls_loss_causal 0.4342 re_mapping 0.0042 re_causal 0.0128 /// teacc 99.02 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.0750, -0.1092, -0.1536, ..., 0.1269, -0.0659, 0.0369], + [-0.0930, 0.1120, -0.0590, ..., 0.0524, -0.0648, -0.0969], + [-0.0397, 0.0881, -0.1330, ..., -0.0214, 0.0414, 0.1248], + ..., + [ 0.0020, 0.0743, 0.0015, ..., -0.0253, 0.0336, -0.0859], + [ 0.1022, -0.1185, 0.0459, ..., 0.0206, 0.0095, -0.0375], + [-0.0122, -0.0882, 0.0744, ..., -0.0970, 0.0340, -0.0293]], + device='cuda:0'), grad: tensor([[ 1.0663e-04, 9.9391e-06, 6.3705e-04, ..., 3.8695e-04, + 2.0351e-03, 1.5736e-04], + [ 1.8120e-05, -1.0264e-04, -3.0130e-05, ..., 2.3508e-04, + 1.2338e-04, 1.9550e-05], + [ 6.3181e-05, -1.9348e-04, 2.3115e-04, ..., 3.1257e-04, + 8.3637e-04, 5.3495e-05], + ..., + [-1.8775e-04, -2.3544e-04, -1.5259e-03, ..., -5.3883e-05, + 1.8403e-05, 8.5533e-05], + [-1.0681e-03, 4.3869e-05, 4.6754e-04, ..., 5.3215e-04, + 2.0599e-03, 3.2616e-04], + [-1.7083e-04, 3.3450e-04, -2.2411e-03, ..., -2.2519e-04, + -1.0040e-02, -7.9918e-04]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0147, -0.0025, -0.0141, 0.0098, -0.0152, -0.0217, 0.0109, -0.0020, + 0.0127, -0.0113], device='cuda:0'), grad: tensor([ 0.0099, 0.0063, 0.0092, -0.0075, 0.0112, -0.0457, 0.0128, 0.0055, + 0.0025, -0.0042], device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 216.80, cls_loss 0.4330 cls_loss_mapping 0.0009 cls_loss_causal 0.4081 re_mapping 0.0041 re_causal 0.0120 /// teacc 99.02 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.0751, -0.1092, -0.1538, ..., 0.1269, -0.0659, 0.0370], + [-0.0929, 0.1119, -0.0590, ..., 0.0524, -0.0648, -0.0968], + [-0.0398, 0.0884, -0.1331, ..., -0.0214, 0.0414, 0.1247], + ..., + [ 0.0020, 0.0742, 0.0016, ..., -0.0252, 0.0335, -0.0859], + [ 0.1022, -0.1184, 0.0459, ..., 0.0206, 0.0094, -0.0375], + [-0.0123, -0.0882, 0.0745, ..., -0.0969, 0.0341, -0.0291]], + device='cuda:0'), grad: tensor([[ 2.8038e-04, 5.2977e-04, 1.0246e-04, ..., 4.9973e-03, + 1.1343e-04, 1.6320e-04], + [ 2.2469e-03, -7.8678e-04, 2.2144e-03, ..., -4.1695e-03, + 1.0443e-04, 2.6441e-04], + [ 6.1214e-05, -9.7179e-04, -8.7976e-04, ..., -1.5965e-03, + -1.2197e-03, -4.2629e-04], + ..., + [ 1.5914e-04, -1.4290e-05, 7.2360e-05, ..., 7.3850e-05, + 5.5403e-05, 7.7307e-05], + [-4.1504e-03, 3.2926e-04, 1.1897e-04, ..., -2.6226e-03, + 7.2718e-05, -1.6680e-03], + [ 2.5916e-04, 8.1420e-05, 1.0026e-04, ..., 1.6570e-04, + 7.0333e-05, 1.1098e-04]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0147, -0.0024, -0.0141, 0.0097, -0.0153, -0.0216, 0.0109, -0.0020, + 0.0127, -0.0113], device='cuda:0'), grad: tensor([ 0.0364, -0.0478, 0.0030, 0.0320, -0.0667, 0.0226, 0.0468, -0.0374, + -0.0150, 0.0261], device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 216.93, cls_loss 0.4475 cls_loss_mapping 0.0009 cls_loss_causal 0.4173 re_mapping 0.0043 re_causal 0.0123 /// teacc 98.99 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.0752, -0.1092, -0.1539, ..., 0.1270, -0.0659, 0.0370], + [-0.0932, 0.1119, -0.0590, ..., 0.0524, -0.0648, -0.0969], + [-0.0395, 0.0885, -0.1331, ..., -0.0214, 0.0414, 0.1248], + ..., + [ 0.0020, 0.0741, 0.0017, ..., -0.0253, 0.0335, -0.0858], + [ 0.1022, -0.1185, 0.0458, ..., 0.0206, 0.0095, -0.0376], + [-0.0124, -0.0883, 0.0745, ..., -0.0968, 0.0342, -0.0291]], + device='cuda:0'), grad: tensor([[ 1.5087e-05, 2.5272e-05, 7.4729e-06, ..., -1.5812e-03, + 1.1772e-05, 3.2216e-05], + [ 3.1348e-06, 4.7493e-03, -3.8567e-03, ..., 2.0576e-04, + 8.0559e-07, 8.0168e-06], + [-3.9005e-04, -1.3123e-02, 3.1829e-05, ..., 2.5296e-04, + 3.1203e-05, -1.2159e-03], + ..., + [ 3.1561e-05, 4.8846e-05, -3.6125e-03, ..., 1.0347e-04, + -3.4904e-03, 9.6321e-05], + [ 8.1122e-05, 2.8419e-04, 2.4962e-04, ..., 1.9026e-04, + 1.7941e-04, 2.5845e-04], + [ 1.9580e-05, 2.4706e-05, 3.3894e-03, ..., 1.1015e-04, + 3.2902e-03, 2.4095e-05]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0147, -0.0024, -0.0141, 0.0097, -0.0152, -0.0215, 0.0109, -0.0021, + 0.0127, -0.0113], device='cuda:0'), grad: tensor([-0.0185, -0.0160, -0.0291, -0.0124, 0.0131, 0.0181, 0.0415, 0.0104, + 0.0052, -0.0123], device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 217.42, cls_loss 0.4070 cls_loss_mapping 0.0012 cls_loss_causal 0.3778 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.00 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.0752, -0.1092, -0.1540, ..., 0.1268, -0.0661, 0.0371], + [-0.0932, 0.1119, -0.0590, ..., 0.0525, -0.0649, -0.0970], + [-0.0391, 0.0885, -0.1332, ..., -0.0215, 0.0414, 0.1248], + ..., + [ 0.0020, 0.0741, 0.0017, ..., -0.0253, 0.0336, -0.0858], + [ 0.1022, -0.1186, 0.0457, ..., 0.0205, 0.0094, -0.0376], + [-0.0124, -0.0883, 0.0746, ..., -0.0968, 0.0343, -0.0291]], + device='cuda:0'), grad: tensor([[ 4.7505e-05, 1.0657e-04, -4.0740e-05, ..., -3.6983e-03, + -7.6580e-04, 3.3155e-07], + [ 8.1420e-05, 3.4475e-04, 3.5238e-04, ..., 2.8110e-04, + 9.6336e-06, 2.1886e-07], + [ 5.1528e-05, -1.3588e-02, -6.7353e-05, ..., -1.0017e-02, + -1.6994e-03, 1.5348e-06], + ..., + [ 1.4353e-04, 1.3115e-02, 6.5804e-05, ..., 9.7427e-03, + 1.9035e-03, 2.1234e-07], + [ 2.2278e-03, 7.2658e-05, 1.1671e-04, ..., 1.4334e-03, + 1.8191e-04, 6.5677e-06], + [ 7.3552e-05, 5.1528e-05, -2.5702e-04, ..., 6.8140e-04, + 1.2589e-04, 2.6822e-06]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0146, -0.0023, -0.0141, 0.0099, -0.0152, -0.0216, 0.0108, -0.0022, + 0.0126, -0.0112], device='cuda:0'), grad: tensor([-0.0125, 0.0106, -0.0363, 0.0038, 0.0121, -0.0013, 0.0041, 0.0219, + 0.0176, -0.0201], device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 217.22, cls_loss 0.4463 cls_loss_mapping 0.0009 cls_loss_causal 0.4234 re_mapping 0.0042 re_causal 0.0125 /// teacc 98.97 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.0753, -0.1092, -0.1541, ..., 0.1269, -0.0662, 0.0372], + [-0.0932, 0.1119, -0.0590, ..., 0.0525, -0.0649, -0.0971], + [-0.0392, 0.0886, -0.1333, ..., -0.0215, 0.0414, 0.1248], + ..., + [ 0.0020, 0.0742, 0.0017, ..., -0.0253, 0.0335, -0.0857], + [ 0.1023, -0.1187, 0.0458, ..., 0.0205, 0.0093, -0.0377], + [-0.0124, -0.0883, 0.0747, ..., -0.0967, 0.0344, -0.0290]], + device='cuda:0'), grad: tensor([[ 2.7061e-05, 1.9953e-05, 1.0386e-03, ..., 2.8396e-04, + 4.0507e-04, 1.5652e-04], + [ 4.6372e-05, -2.4366e-04, -3.5267e-03, ..., -2.2564e-03, + -8.1205e-04, 3.9428e-05], + [-2.6584e-05, -1.3485e-03, 1.3685e-03, ..., 2.7108e-04, + -2.3437e-04, -4.9877e-04], + ..., + [ 5.9098e-05, 1.2989e-03, 2.2717e-03, ..., 5.8031e-04, + 1.4448e-03, 5.0879e-04], + [-2.2739e-05, 9.1076e-05, 4.8375e-04, ..., -1.9670e-04, + -3.7861e-04, -5.3692e-04], + [-2.6655e-04, -4.4537e-04, 1.9484e-03, ..., 2.9278e-04, + 3.1996e-04, 2.0361e-04]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0145, -0.0023, -0.0140, 0.0099, -0.0153, -0.0215, 0.0107, -0.0022, + 0.0126, -0.0111], device='cuda:0'), grad: tensor([-0.0003, -0.0330, 0.0244, -0.0333, -0.0176, -0.0086, 0.0314, 0.0405, + -0.0077, 0.0042], device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 216.81, cls_loss 0.4412 cls_loss_mapping 0.0007 cls_loss_causal 0.4066 re_mapping 0.0041 re_causal 0.0121 /// teacc 98.95 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.0753, -0.1091, -0.1541, ..., 0.1268, -0.0661, 0.0373], + [-0.0934, 0.1117, -0.0591, ..., 0.0523, -0.0649, -0.0972], + [-0.0392, 0.0886, -0.1332, ..., -0.0215, 0.0415, 0.1248], + ..., + [ 0.0022, 0.0742, 0.0018, ..., -0.0252, 0.0335, -0.0856], + [ 0.1023, -0.1188, 0.0458, ..., 0.0207, 0.0092, -0.0377], + [-0.0123, -0.0882, 0.0747, ..., -0.0967, 0.0344, -0.0290]], + device='cuda:0'), grad: tensor([[ 4.1580e-04, 6.3665e-06, 2.2352e-08, ..., 7.4768e-04, + 1.2303e-06, 6.4516e-04], + [ 4.5681e-04, 2.2926e-03, 1.1455e-07, ..., 1.2894e-03, + 8.5115e-04, 7.0906e-04], + [ 1.0118e-03, -2.5978e-03, 4.7684e-05, ..., -3.1471e-05, + -9.5654e-04, 1.3533e-03], + ..., + [ 3.0661e-04, 1.5008e-04, 1.2033e-06, ..., 1.5891e-04, + 5.5373e-05, 4.7851e-04], + [-1.3153e-02, 1.9610e-05, 2.5760e-06, ..., -2.4033e-03, + 5.7593e-06, 8.2922e-04], + [ 2.6131e-04, 2.3976e-05, -9.2108e-07, ..., 2.0719e-04, + 6.6161e-06, 4.0483e-04]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0145, -0.0023, -0.0140, 0.0098, -0.0153, -0.0216, 0.0108, -0.0021, + 0.0126, -0.0111], device='cuda:0'), grad: tensor([ 0.0245, -0.0021, -0.0088, 0.0220, -0.0116, -0.0069, -0.0075, 0.0191, + -0.0183, -0.0103], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 426---------------------------------------------------- +epoch 426, time 217.70, cls_loss 0.4497 cls_loss_mapping 0.0009 cls_loss_causal 0.4197 re_mapping 0.0041 re_causal 0.0121 /// teacc 99.03 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.0753, -0.1091, -0.1539, ..., 0.1267, -0.0663, 0.0373], + [-0.0934, 0.1116, -0.0589, ..., 0.0523, -0.0646, -0.0973], + [-0.0391, 0.0886, -0.1333, ..., -0.0214, 0.0415, 0.1248], + ..., + [ 0.0021, 0.0743, 0.0018, ..., -0.0252, 0.0336, -0.0857], + [ 0.1023, -0.1189, 0.0460, ..., 0.0206, 0.0093, -0.0376], + [-0.0125, -0.0882, 0.0746, ..., -0.0967, 0.0343, -0.0292]], + device='cuda:0'), grad: tensor([[ 8.3923e-03, 9.2363e-04, 1.2410e-04, ..., 9.4771e-05, + 6.9695e-03, 7.1239e-04], + [ 7.1466e-05, -4.4629e-06, 5.5701e-05, ..., -1.1330e-03, + 4.5896e-05, 1.2815e-04], + [ 1.0052e-03, -2.1982e-04, 1.2839e-04, ..., 1.0365e-04, + 3.4928e-04, 1.5271e-04], + ..., + [-8.6746e-03, -7.2861e-04, 1.1617e-04, ..., 1.9503e-04, + -7.1869e-03, -3.6478e-04], + [ 9.2459e-04, -2.4185e-05, 2.3899e-03, ..., -2.8515e-04, + 1.2836e-03, -1.0290e-03], + [ 4.4084e-04, 2.8208e-05, -7.0858e-04, ..., 3.1185e-04, + -6.9189e-04, 4.1270e-04]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0145, -0.0023, -0.0141, 0.0098, -0.0151, -0.0215, 0.0108, -0.0020, + 0.0125, -0.0113], device='cuda:0'), grad: tensor([ 0.0052, -0.0104, 0.0175, -0.0372, 0.0350, 0.0074, -0.0071, 0.0020, + -0.0380, 0.0255], device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 216.56, cls_loss 0.4468 cls_loss_mapping 0.0009 cls_loss_causal 0.4227 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.97 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.0753, -0.1090, -0.1539, ..., 0.1269, -0.0661, 0.0374], + [-0.0934, 0.1116, -0.0589, ..., 0.0522, -0.0648, -0.0974], + [-0.0390, 0.0886, -0.1332, ..., -0.0212, 0.0417, 0.1249], + ..., + [ 0.0022, 0.0744, 0.0018, ..., -0.0252, 0.0336, -0.0856], + [ 0.1024, -0.1190, 0.0460, ..., 0.0206, 0.0093, -0.0378], + [-0.0126, -0.0884, 0.0745, ..., -0.0969, 0.0343, -0.0292]], + device='cuda:0'), grad: tensor([[ 4.0621e-05, 1.5211e-04, 1.2863e-04, ..., 2.5463e-04, + 3.0375e-04, 5.1945e-05], + [ 3.7372e-05, 1.8466e-04, 1.7786e-04, ..., 3.1376e-04, + 4.1723e-04, 1.3041e-04], + [ 8.4758e-05, 6.3095e-03, 5.2071e-04, ..., 9.0714e-03, + 9.8991e-04, -1.0386e-03], + ..., + [ 6.9499e-05, 1.2035e-03, 3.8195e-04, ..., 7.0152e-03, + 9.5177e-04, 3.0208e-04], + [ 1.1110e-04, -6.8512e-03, 5.5885e-04, ..., -1.5465e-02, + 1.4172e-03, 4.9257e-04], + [ 4.7803e-05, 2.3842e-04, 2.5082e-04, ..., 4.0627e-04, + 5.9414e-04, 1.9479e-04]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0145, -0.0024, -0.0142, 0.0099, -0.0152, -0.0215, 0.0108, -0.0020, + 0.0127, -0.0114], device='cuda:0'), grad: tensor([ 0.0148, 0.0133, -0.0015, -0.0557, 0.0103, 0.0090, 0.0086, -0.0062, + -0.0056, 0.0130], device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 217.59, cls_loss 0.4290 cls_loss_mapping 0.0009 cls_loss_causal 0.3977 re_mapping 0.0040 re_causal 0.0117 /// teacc 98.97 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.0753, -0.1090, -0.1537, ..., 0.1269, -0.0659, 0.0374], + [-0.0934, 0.1116, -0.0589, ..., 0.0522, -0.0648, -0.0973], + [-0.0389, 0.0885, -0.1333, ..., -0.0211, 0.0417, 0.1247], + ..., + [ 0.0022, 0.0745, 0.0019, ..., -0.0253, 0.0336, -0.0856], + [ 0.1023, -0.1191, 0.0460, ..., 0.0206, 0.0093, -0.0377], + [-0.0126, -0.0884, 0.0744, ..., -0.0970, 0.0340, -0.0293]], + device='cuda:0'), grad: tensor([[ 1.2420e-05, 2.2333e-06, 3.0255e-04, ..., 1.5843e-04, + 9.0003e-05, 3.7160e-07], + [ 6.1467e-07, 7.5623e-06, 4.7493e-04, ..., 2.4676e-04, + 2.2978e-05, 1.9185e-06], + [ 1.6810e-06, 7.2765e-04, 6.4707e-04, ..., 5.0592e-04, + 1.0052e-03, 1.4710e-04], + ..., + [-4.2498e-05, -8.3017e-04, -2.4843e-04, ..., -3.0231e-04, + -1.3199e-03, 1.8999e-05], + [-9.0332e-03, 1.1545e-04, -3.0518e-03, ..., 1.9300e-04, + -8.6486e-05, 3.3498e-05], + [ 5.3436e-05, 2.7753e-06, 4.7531e-03, ..., 2.3766e-03, + 4.7374e-04, 2.6776e-07]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0144, -0.0023, -0.0141, 0.0100, -0.0152, -0.0214, 0.0108, -0.0021, + 0.0127, -0.0114], device='cuda:0'), grad: tensor([ 0.0076, 0.0096, 0.0087, 0.0058, -0.0232, -0.0241, 0.0156, 0.0043, + -0.0023, -0.0019], device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 217.07, cls_loss 0.4322 cls_loss_mapping 0.0008 cls_loss_causal 0.4041 re_mapping 0.0039 re_causal 0.0118 /// teacc 98.90 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.0754, -0.1090, -0.1536, ..., 0.1269, -0.0658, 0.0375], + [-0.0934, 0.1118, -0.0588, ..., 0.0523, -0.0646, -0.0974], + [-0.0390, 0.0885, -0.1333, ..., -0.0212, 0.0416, 0.1246], + ..., + [ 0.0022, 0.0745, 0.0019, ..., -0.0253, 0.0337, -0.0855], + [ 0.1023, -0.1192, 0.0460, ..., 0.0207, 0.0091, -0.0378], + [-0.0126, -0.0885, 0.0744, ..., -0.0971, 0.0339, -0.0293]], + device='cuda:0'), grad: tensor([[ 4.4048e-05, 1.0863e-05, 1.4496e-04, ..., -2.6017e-05, + 8.1778e-05, -4.9442e-05], + [ 4.8578e-05, 2.1112e-04, 8.9109e-05, ..., 4.8667e-05, + 1.5199e-05, 9.1255e-05], + [ 4.2945e-05, -2.8014e-04, -2.8553e-03, ..., 1.1981e-05, + -4.4785e-03, -1.4460e-04], + ..., + [ 1.0958e-03, -1.7047e-04, 4.8561e-03, ..., 6.2561e-04, + 2.3937e-03, 4.6879e-05], + [ 1.0723e-04, 5.6505e-05, 3.0160e-04, ..., 8.8274e-05, + 1.4031e-04, 2.1920e-05], + [-1.9646e-03, -1.0830e-04, -3.9330e-03, ..., -1.1768e-03, + 1.2913e-03, -3.1352e-05]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0143, -0.0024, -0.0141, 0.0101, -0.0153, -0.0213, 0.0108, -0.0020, + 0.0126, -0.0114], device='cuda:0'), grad: tensor([ 0.0047, -0.0073, -0.0212, 0.0054, -0.0200, 0.0042, 0.0047, 0.0319, + 0.0079, -0.0103], device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 216.89, cls_loss 0.4411 cls_loss_mapping 0.0008 cls_loss_causal 0.4100 re_mapping 0.0040 re_causal 0.0121 /// teacc 98.92 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.0754, -0.1090, -0.1535, ..., 0.1269, -0.0657, 0.0373], + [-0.0934, 0.1119, -0.0589, ..., 0.0523, -0.0647, -0.0974], + [-0.0390, 0.0885, -0.1334, ..., -0.0212, 0.0414, 0.1246], + ..., + [ 0.0023, 0.0745, 0.0018, ..., -0.0253, 0.0338, -0.0854], + [ 0.1025, -0.1193, 0.0459, ..., 0.0207, 0.0091, -0.0378], + [-0.0127, -0.0885, 0.0744, ..., -0.0971, 0.0339, -0.0292]], + device='cuda:0'), grad: tensor([[ 1.2755e-04, 3.8696e-07, 2.6599e-05, ..., 2.0351e-03, + 9.2554e-04, 4.7159e-04], + [ 6.7711e-05, 1.1325e-06, 1.5333e-05, ..., 1.9848e-04, + 2.9355e-05, 5.7727e-05], + [ 8.3625e-05, -8.3223e-06, 1.7345e-05, ..., 4.8399e-04, + 1.8520e-03, 2.2805e-04], + ..., + [ 2.0161e-05, -6.4230e-04, -8.3466e-03, ..., -1.5039e-03, + -4.0855e-03, 1.8224e-05], + [ 3.5644e-05, 3.6899e-06, 2.8729e-05, ..., 2.1589e-04, + 1.7509e-03, 4.4107e-05], + [ 5.2691e-05, 6.3658e-04, 8.3008e-03, ..., 1.8482e-03, + 7.1144e-03, 6.1750e-05]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0144, -0.0024, -0.0141, 0.0099, -0.0153, -0.0212, 0.0108, -0.0017, + 0.0125, -0.0115], device='cuda:0'), grad: tensor([ 0.0166, 0.0091, 0.0148, -0.0125, -0.0521, -0.0084, -0.0055, -0.0071, + 0.0123, 0.0327], device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 216.99, cls_loss 0.4284 cls_loss_mapping 0.0007 cls_loss_causal 0.4013 re_mapping 0.0041 re_causal 0.0127 /// teacc 98.97 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.0754, -0.1091, -0.1536, ..., 0.1269, -0.0657, 0.0372], + [-0.0935, 0.1119, -0.0589, ..., 0.0522, -0.0647, -0.0974], + [-0.0390, 0.0886, -0.1334, ..., -0.0213, 0.0414, 0.1247], + ..., + [ 0.0024, 0.0745, 0.0016, ..., -0.0253, 0.0338, -0.0855], + [ 0.1024, -0.1193, 0.0459, ..., 0.0208, 0.0091, -0.0376], + [-0.0127, -0.0885, 0.0745, ..., -0.0971, 0.0341, -0.0293]], + device='cuda:0'), grad: tensor([[ 2.7776e-04, 3.5942e-05, 2.7084e-04, ..., -1.6003e-03, + 1.0616e-04, -1.9989e-03], + [-6.5651e-03, -5.0211e-04, -1.5144e-03, ..., -1.5230e-03, + 4.7714e-05, 3.1734e-04], + [ 8.5068e-04, 4.7296e-05, 4.1008e-04, ..., 7.1192e-04, + 2.3103e-04, 6.9809e-04], + ..., + [ 3.0270e-03, 1.2815e-04, 4.9248e-03, ..., 5.7316e-04, + 1.5652e-04, 2.8801e-04], + [ 3.5896e-03, 8.3804e-05, 1.0376e-03, ..., -3.4213e-05, + 1.3733e-03, 1.1415e-03], + [-1.2341e-03, 3.5262e-04, -2.8896e-03, ..., 1.3075e-03, + 4.5085e-04, 4.8876e-04]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0146, -0.0025, -0.0142, 0.0098, -0.0154, -0.0212, 0.0109, -0.0017, + 0.0125, -0.0115], device='cuda:0'), grad: tensor([-0.0426, -0.0356, 0.0217, 0.0534, 0.0408, -0.0274, -0.0336, 0.0156, + 0.0055, 0.0022], device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 216.85, cls_loss 0.4186 cls_loss_mapping 0.0008 cls_loss_causal 0.3978 re_mapping 0.0042 re_causal 0.0127 /// teacc 98.96 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.0755, -0.1092, -0.1536, ..., 0.1270, -0.0658, 0.0373], + [-0.0936, 0.1118, -0.0589, ..., 0.0521, -0.0646, -0.0975], + [-0.0390, 0.0885, -0.1335, ..., -0.0213, 0.0413, 0.1247], + ..., + [ 0.0024, 0.0747, 0.0016, ..., -0.0253, 0.0339, -0.0855], + [ 0.1021, -0.1195, 0.0462, ..., 0.0207, 0.0093, -0.0378], + [-0.0127, -0.0885, 0.0744, ..., -0.0972, 0.0339, -0.0291]], + device='cuda:0'), grad: tensor([[ 3.9792e-04, 9.0122e-05, 4.3996e-06, ..., 3.1805e-07, + 3.0875e-05, 1.9409e-06], + [-1.2863e-02, -2.9469e-03, -3.4451e-05, ..., -4.4525e-05, + 2.8566e-05, 1.1120e-06], + [ 7.3147e-04, 1.6272e-04, 2.1338e-05, ..., -7.5577e-07, + 1.3685e-03, 2.6822e-05], + ..., + [ 1.9293e-03, 4.7803e-04, 4.8971e-04, ..., 3.2216e-05, + -1.2708e-04, 2.7642e-05], + [ 1.0614e-03, 3.1304e-04, 2.7269e-05, ..., 2.3637e-06, + 4.7445e-04, 4.2289e-05], + [ 3.3989e-03, 7.4816e-04, -5.5408e-04, ..., 2.2352e-06, + -2.7332e-03, 1.4501e-06]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0147, -0.0025, -0.0143, 0.0099, -0.0154, -0.0212, 0.0109, -0.0017, + 0.0124, -0.0114], device='cuda:0'), grad: tensor([ 0.0073, -0.0149, -0.0188, -0.0212, 0.0148, 0.0073, 0.0070, 0.0235, + 0.0106, -0.0156], device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 216.98, cls_loss 0.4417 cls_loss_mapping 0.0008 cls_loss_causal 0.4123 re_mapping 0.0040 re_causal 0.0127 /// teacc 98.95 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.0755, -0.1092, -0.1536, ..., 0.1270, -0.0658, 0.0374], + [-0.0936, 0.1119, -0.0587, ..., 0.0521, -0.0645, -0.0976], + [-0.0388, 0.0885, -0.1334, ..., -0.0212, 0.0414, 0.1246], + ..., + [ 0.0023, 0.0747, 0.0017, ..., -0.0253, 0.0339, -0.0855], + [ 0.1021, -0.1197, 0.0463, ..., 0.0207, 0.0094, -0.0377], + [-0.0127, -0.0887, 0.0743, ..., -0.0972, 0.0339, -0.0291]], + device='cuda:0'), grad: tensor([[ 1.4591e-04, 1.0943e-04, 3.3481e-07, ..., 2.0584e-02, + 2.1942e-02, -3.2574e-05], + [ 2.2948e-04, -9.1672e-05, 1.6671e-06, ..., -3.6812e-04, + 1.5065e-05, 3.4319e-07], + [ 1.8537e-04, 8.3256e-04, 1.3404e-05, ..., -2.6443e-02, + -2.8351e-02, 1.2375e-05], + ..., + [ 4.5866e-05, -3.1650e-05, -4.0245e-03, ..., 4.6082e-03, + 2.1000e-03, 1.9427e-06], + [-4.4465e-04, 2.5257e-05, 1.4019e-04, ..., 8.4758e-05, + 1.1168e-05, -9.0078e-06], + [ 7.7784e-05, 9.7454e-06, 3.8967e-03, ..., 1.1368e-03, + 3.7174e-03, 1.2470e-06]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0146, -0.0026, -0.0144, 0.0100, -0.0155, -0.0211, 0.0109, -0.0018, + 0.0124, -0.0113], device='cuda:0'), grad: tensor([ 0.0370, 0.0159, 0.0008, -0.0029, 0.0132, 0.0183, -0.0645, -0.0147, + 0.0113, -0.0145], device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 216.94, cls_loss 0.4491 cls_loss_mapping 0.0008 cls_loss_causal 0.4274 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.95 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.0754, -0.1092, -0.1535, ..., 0.1271, -0.0658, 0.0373], + [-0.0937, 0.1120, -0.0586, ..., 0.0521, -0.0645, -0.0974], + [-0.0389, 0.0885, -0.1336, ..., -0.0212, 0.0415, 0.1245], + ..., + [ 0.0022, 0.0747, 0.0017, ..., -0.0252, 0.0339, -0.0853], + [ 0.1022, -0.1195, 0.0464, ..., 0.0207, 0.0096, -0.0376], + [-0.0128, -0.0888, 0.0742, ..., -0.0973, 0.0338, -0.0291]], + device='cuda:0'), grad: tensor([[ 0.0007, 0.0004, 0.0003, ..., 0.0013, 0.0004, 0.0004], + [ 0.0005, 0.0004, 0.0004, ..., 0.0018, 0.0005, 0.0004], + [ 0.0002, -0.0004, -0.0003, ..., -0.0008, -0.0005, -0.0007], + ..., + [ 0.0001, 0.0001, 0.0003, ..., 0.0005, 0.0002, 0.0001], + [ 0.0001, 0.0002, 0.0003, ..., 0.0007, 0.0002, 0.0002], + [ 0.0004, 0.0001, -0.0019, ..., 0.0006, -0.0009, 0.0001]], + device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0146, -0.0025, -0.0143, 0.0101, -0.0155, -0.0213, 0.0108, -0.0017, + 0.0125, -0.0114], device='cuda:0'), grad: tensor([-0.0044, -0.0047, -0.0244, 0.0245, 0.0225, -0.0185, -0.0108, 0.0164, + -0.0134, 0.0128], device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 218.01, cls_loss 0.4658 cls_loss_mapping 0.0011 cls_loss_causal 0.4381 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.00 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.0754, -0.1092, -0.1534, ..., 0.1272, -0.0659, 0.0372], + [-0.0936, 0.1121, -0.0585, ..., 0.0521, -0.0642, -0.0971], + [-0.0389, 0.0883, -0.1335, ..., -0.0212, 0.0415, 0.1244], + ..., + [ 0.0021, 0.0748, 0.0018, ..., -0.0251, 0.0339, -0.0851], + [ 0.1023, -0.1196, 0.0464, ..., 0.0206, 0.0097, -0.0376], + [-0.0126, -0.0888, 0.0742, ..., -0.0973, 0.0338, -0.0291]], + device='cuda:0'), grad: tensor([[-1.1057e-05, 3.7253e-08, 2.6250e-04, ..., 1.8418e-04, + 4.5109e-04, -3.0845e-05], + [ 2.9802e-08, -1.6578e-06, 2.4986e-04, ..., 4.7421e-04, + 4.8423e-04, 7.4506e-09], + [ 1.7975e-07, 4.0513e-08, 1.1903e-04, ..., 2.4343e-04, + 4.0412e-04, 3.2783e-07], + ..., + [ 7.3910e-05, -3.1665e-08, 6.5994e-04, ..., 3.8600e-04, + 7.9346e-04, 4.9826e-08], + [ 1.3867e-06, 2.5285e-07, 4.1342e-04, ..., 2.3997e-04, + 7.2336e-04, 7.1665e-07], + [-1.3673e-04, 2.7893e-07, 1.5125e-03, ..., 3.2997e-04, + 3.1700e-03, 6.5155e-06]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0145, -0.0023, -0.0144, 0.0099, -0.0155, -0.0213, 0.0109, -0.0016, + 0.0124, -0.0113], device='cuda:0'), grad: tensor([ 0.0157, -0.0065, 0.0163, -0.0105, -0.0507, 0.0153, 0.0213, 0.0222, + -0.0458, 0.0227], device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 216.61, cls_loss 0.4450 cls_loss_mapping 0.0010 cls_loss_causal 0.4217 re_mapping 0.0039 re_causal 0.0118 /// teacc 98.97 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.0754, -0.1091, -0.1534, ..., 0.1271, -0.0660, 0.0372], + [-0.0938, 0.1121, -0.0585, ..., 0.0521, -0.0641, -0.0970], + [-0.0388, 0.0885, -0.1336, ..., -0.0212, 0.0414, 0.1245], + ..., + [ 0.0020, 0.0746, 0.0017, ..., -0.0251, 0.0338, -0.0852], + [ 0.1024, -0.1197, 0.0463, ..., 0.0206, 0.0095, -0.0377], + [-0.0125, -0.0885, 0.0743, ..., -0.0973, 0.0340, -0.0291]], + device='cuda:0'), grad: tensor([[ 3.0268e-06, 9.8944e-06, 4.2170e-05, ..., 1.0347e-04, + 7.7546e-05, 2.4855e-05], + [ 3.0175e-06, -2.2888e-04, 6.6221e-05, ..., 2.7999e-05, + 7.9453e-05, 4.1693e-05], + [ 1.8049e-06, 8.2910e-05, -1.7334e-02, ..., -6.1913e-03, + -1.2787e-02, 7.9250e-04], + ..., + [ 5.6505e-04, 1.1224e-04, 1.7746e-02, ..., 6.1798e-03, + 1.2878e-02, 2.1696e-05], + [ 9.9912e-06, 3.3885e-05, 2.2507e-04, ..., 8.0824e-04, + 8.5640e-04, 1.5268e-03], + [ 1.8549e-04, -9.0539e-05, -1.7258e-02, ..., 2.3425e-04, + -3.6454e-04, 4.5449e-05]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0144, -0.0024, -0.0142, 0.0100, -0.0155, -0.0214, 0.0108, -0.0017, + 0.0125, -0.0112], device='cuda:0'), grad: tensor([ 0.0007, -0.0003, -0.0235, -0.0215, 0.0152, -0.0035, 0.0015, 0.0340, + 0.0152, -0.0179], device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 216.81, cls_loss 0.4536 cls_loss_mapping 0.0012 cls_loss_causal 0.4269 re_mapping 0.0037 re_causal 0.0114 /// teacc 98.96 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.0753, -0.1092, -0.1534, ..., 0.1270, -0.0661, 0.0372], + [-0.0938, 0.1121, -0.0586, ..., 0.0521, -0.0641, -0.0969], + [-0.0388, 0.0885, -0.1335, ..., -0.0210, 0.0415, 0.1246], + ..., + [ 0.0020, 0.0745, 0.0016, ..., -0.0253, 0.0336, -0.0853], + [ 0.1024, -0.1196, 0.0463, ..., 0.0206, 0.0096, -0.0377], + [-0.0126, -0.0884, 0.0744, ..., -0.0971, 0.0341, -0.0291]], + device='cuda:0'), grad: tensor([[ 7.4506e-07, 1.7333e-04, 5.3085e-06, ..., -2.7695e-03, + -1.4696e-03, 1.3243e-06], + [ 1.0252e-05, -7.7438e-03, 1.3318e-06, ..., 2.4283e-04, + -1.9484e-03, 2.1577e-04], + [ 8.2143e-07, 1.2178e-03, 2.8729e-04, ..., -1.2455e-03, + 1.2941e-03, -7.2956e-04], + ..., + [ 2.5257e-06, 4.7646e-03, 1.1480e-04, ..., 2.8467e-04, + 7.9651e-03, 1.3626e-04], + [-2.4676e-05, 4.2224e-04, -4.7469e-04, ..., 2.4319e-04, + 3.2306e-04, -1.6308e-04], + [ 1.2860e-05, -2.8634e-04, -1.9348e-04, ..., 3.3259e-04, + -8.8196e-03, 1.6797e-04]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0144, -0.0023, -0.0142, 0.0100, -0.0155, -0.0213, 0.0106, -0.0018, + 0.0125, -0.0112], device='cuda:0'), grad: tensor([-0.0029, -0.0018, 0.0051, -0.0412, 0.0138, 0.0227, 0.0168, 0.0512, + -0.0181, -0.0456], device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 216.61, cls_loss 0.4430 cls_loss_mapping 0.0009 cls_loss_causal 0.4109 re_mapping 0.0037 re_causal 0.0112 /// teacc 98.98 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.0753, -0.1092, -0.1535, ..., 0.1270, -0.0661, 0.0372], + [-0.0938, 0.1120, -0.0586, ..., 0.0520, -0.0640, -0.0970], + [-0.0388, 0.0886, -0.1335, ..., -0.0208, 0.0416, 0.1247], + ..., + [ 0.0020, 0.0744, 0.0016, ..., -0.0254, 0.0335, -0.0854], + [ 0.1024, -0.1195, 0.0464, ..., 0.0207, 0.0096, -0.0377], + [-0.0125, -0.0883, 0.0742, ..., -0.0972, 0.0341, -0.0291]], + device='cuda:0'), grad: tensor([[ 3.7923e-06, 3.8445e-06, 6.5714e-06, ..., 8.4734e-04, + 1.0166e-03, 7.4878e-07], + [-4.9055e-05, -9.1791e-06, 2.5295e-06, ..., 2.0695e-04, + 2.1827e-04, 7.3668e-07], + [ 1.3840e-04, 3.9749e-06, 9.5516e-06, ..., 2.3441e-03, + 3.5496e-03, 2.3738e-05], + ..., + [-2.8834e-05, -8.6963e-05, 5.6505e-05, ..., -6.3019e-03, + -7.3280e-03, 1.0356e-05], + [ 1.2177e-04, 3.3498e-05, 9.5308e-05, ..., 1.8473e-03, + 2.2030e-03, 2.8744e-05], + [ 3.9339e-05, 2.9594e-05, -2.7222e-02, ..., 7.0477e-04, + -5.7373e-03, -5.4777e-05]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0143, -0.0024, -0.0141, 0.0101, -0.0157, -0.0214, 0.0106, -0.0017, + 0.0127, -0.0112], device='cuda:0'), grad: tensor([ 0.0083, -0.0248, -0.0089, -0.0131, 0.0368, 0.0126, 0.0050, -0.0099, + 0.0200, -0.0259], device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 216.87, cls_loss 0.4333 cls_loss_mapping 0.0010 cls_loss_causal 0.4059 re_mapping 0.0038 re_causal 0.0115 /// teacc 98.97 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.0753, -0.1092, -0.1535, ..., 0.1269, -0.0663, 0.0371], + [-0.0936, 0.1121, -0.0585, ..., 0.0519, -0.0640, -0.0971], + [-0.0389, 0.0885, -0.1336, ..., -0.0208, 0.0417, 0.1246], + ..., + [ 0.0020, 0.0745, 0.0016, ..., -0.0253, 0.0335, -0.0854], + [ 0.1022, -0.1194, 0.0463, ..., 0.0207, 0.0096, -0.0377], + [-0.0125, -0.0884, 0.0743, ..., -0.0973, 0.0342, -0.0291]], + device='cuda:0'), grad: tensor([[ 4.5586e-04, 9.5546e-05, 2.4755e-06, ..., 8.4266e-06, + 1.4734e-06, 7.4089e-05], + [ 2.0218e-03, -3.2330e-04, 1.8016e-05, ..., -1.1837e-04, + 5.0943e-07, 9.6202e-05], + [ 1.0147e-03, 3.2902e-03, 8.0317e-06, ..., 2.0587e-04, + 3.0473e-06, 2.2221e-04], + ..., + [-1.0719e-02, 3.3545e-04, 4.9877e-04, ..., 4.3988e-05, + 1.0252e-04, 1.3280e-04], + [ 1.6518e-03, -3.8548e-03, 1.2612e-04, ..., -1.9825e-04, + -5.9700e-04, 7.2837e-05], + [ 2.3060e-03, 9.3043e-05, -9.9087e-04, ..., 6.3889e-06, + -3.8099e-04, 7.6652e-05]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0143, -0.0022, -0.0141, 0.0103, -0.0156, -0.0214, 0.0106, -0.0018, + 0.0126, -0.0113], device='cuda:0'), grad: tensor([ 0.0127, -0.0426, 0.0428, 0.0134, 0.0157, -0.0109, -0.0177, -0.0101, + -0.0190, 0.0156], device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 216.85, cls_loss 0.4678 cls_loss_mapping 0.0009 cls_loss_causal 0.4405 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.93 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.0752, -0.1092, -0.1535, ..., 0.1270, -0.0662, 0.0371], + [-0.0938, 0.1120, -0.0585, ..., 0.0519, -0.0639, -0.0969], + [-0.0388, 0.0886, -0.1336, ..., -0.0207, 0.0417, 0.1246], + ..., + [ 0.0019, 0.0744, 0.0015, ..., -0.0255, 0.0335, -0.0854], + [ 0.1021, -0.1193, 0.0464, ..., 0.0206, 0.0096, -0.0376], + [-0.0125, -0.0883, 0.0744, ..., -0.0972, 0.0342, -0.0292]], + device='cuda:0'), grad: tensor([[-6.8588e-03, 2.6751e-04, -9.5797e-04, ..., 3.2276e-05, + 1.4424e-04, 1.0710e-08], + [ 2.0301e-04, 6.6328e-04, 2.7633e-04, ..., 4.0144e-05, + 4.1761e-06, 1.8626e-09], + [ 7.8869e-04, 2.9602e-03, 1.1981e-04, ..., 3.3712e-04, + 1.7181e-05, 4.2608e-07], + ..., + [ 5.7173e-04, -7.0953e-03, 3.6955e-04, ..., -7.6437e-04, + 7.8201e-04, 3.2596e-09], + [ 1.6785e-03, 4.3488e-04, 1.8811e-04, ..., 5.1111e-05, + 3.6716e-05, -1.4799e-06], + [ 2.5997e-03, 4.0793e-04, 2.2526e-03, ..., 4.9502e-05, + 1.4977e-02, 3.9861e-07]], device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0142, -0.0022, -0.0140, 0.0102, -0.0156, -0.0213, 0.0106, -0.0018, + 0.0124, -0.0114], device='cuda:0'), grad: tensor([-0.0426, 0.0165, 0.0222, 0.0158, -0.0864, 0.0155, 0.0166, 0.0047, + 0.0170, 0.0209], device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 216.51, cls_loss 0.4485 cls_loss_mapping 0.0009 cls_loss_causal 0.4208 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.96 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.0753, -0.1091, -0.1536, ..., 0.1270, -0.0662, 0.0371], + [-0.0940, 0.1120, -0.0585, ..., 0.0518, -0.0638, -0.0970], + [-0.0387, 0.0887, -0.1334, ..., -0.0206, 0.0417, 0.1247], + ..., + [ 0.0020, 0.0744, 0.0015, ..., -0.0255, 0.0335, -0.0854], + [ 0.1021, -0.1194, 0.0464, ..., 0.0206, 0.0097, -0.0376], + [-0.0124, -0.0883, 0.0743, ..., -0.0973, 0.0341, -0.0293]], + device='cuda:0'), grad: tensor([[ 1.3518e-04, 1.0291e-06, 2.0981e-04, ..., 1.4174e-04, + 3.7342e-05, 5.0336e-05], + [ 3.1853e-04, -3.0156e-06, 2.3556e-04, ..., 7.6234e-05, + 3.7044e-05, 1.9241e-06], + [ 1.5390e-04, -3.1907e-06, 1.9526e-04, ..., 7.3373e-05, + 1.1283e-04, 1.3900e-04], + ..., + [ 1.5604e-04, -1.2673e-05, 9.3079e-03, ..., 3.4273e-05, + -9.1648e-04, 1.6782e-06], + [ 2.2447e-04, 6.0163e-06, 7.8678e-04, ..., -1.2856e-03, + 1.2684e-04, -3.3528e-05], + [ 1.3685e-04, 3.5707e-06, -1.6815e-02, ..., 5.8264e-05, + 1.9979e-04, 7.6741e-06]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0143, -0.0023, -0.0139, 0.0101, -0.0156, -0.0213, 0.0107, -0.0019, + 0.0124, -0.0114], device='cuda:0'), grad: tensor([ 0.0168, 0.0233, 0.0168, -0.0585, 0.0315, -0.0056, 0.0400, 0.0084, + -0.0688, -0.0038], device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 216.74, cls_loss 0.4355 cls_loss_mapping 0.0008 cls_loss_causal 0.4128 re_mapping 0.0040 re_causal 0.0123 /// teacc 98.99 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.0753, -0.1091, -0.1536, ..., 0.1270, -0.0662, 0.0372], + [-0.0941, 0.1120, -0.0586, ..., 0.0519, -0.0639, -0.0970], + [-0.0386, 0.0885, -0.1337, ..., -0.0207, 0.0415, 0.1246], + ..., + [ 0.0021, 0.0746, 0.0015, ..., -0.0254, 0.0335, -0.0853], + [ 0.1021, -0.1195, 0.0463, ..., 0.0206, 0.0095, -0.0376], + [-0.0124, -0.0884, 0.0744, ..., -0.0973, 0.0342, -0.0294]], + device='cuda:0'), grad: tensor([[ 5.8502e-05, 5.5097e-06, 9.2447e-05, ..., 3.0923e-04, + 2.0373e-04, 1.0687e-04], + [ 6.2883e-05, 4.8714e-03, 2.2964e-03, ..., 4.6730e-03, + 1.9388e-03, 5.1171e-05], + [ 2.2972e-04, 4.4525e-05, 5.4979e-04, ..., 6.1464e-04, + 7.0953e-04, 4.7445e-05], + ..., + [ 4.2191e-03, 6.6185e-03, 1.2579e-03, ..., 1.3456e-03, + 1.0195e-03, 3.4682e-06], + [ 4.5433e-03, 1.4579e-04, 2.9030e-03, ..., -2.9421e-04, + 2.9182e-03, -2.7037e-04], + [-3.7937e-03, -5.3368e-03, -7.6752e-03, ..., 5.0926e-04, + -7.3090e-03, 2.5332e-05]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0141, -0.0023, -0.0138, 0.0102, -0.0156, -0.0214, 0.0109, -0.0018, + 0.0123, -0.0114], device='cuda:0'), grad: tensor([ 0.0021, 0.0248, 0.0035, -0.0275, 0.0062, -0.0343, -0.0173, 0.0208, + 0.0376, -0.0158], device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 217.12, cls_loss 0.4477 cls_loss_mapping 0.0009 cls_loss_causal 0.4182 re_mapping 0.0038 re_causal 0.0118 /// teacc 98.96 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.0753, -0.1092, -0.1536, ..., 0.1268, -0.0661, 0.0374], + [-0.0941, 0.1123, -0.0585, ..., 0.0519, -0.0638, -0.0970], + [-0.0386, 0.0885, -0.1339, ..., -0.0205, 0.0414, 0.1246], + ..., + [ 0.0020, 0.0744, 0.0015, ..., -0.0255, 0.0334, -0.0852], + [ 0.1021, -0.1195, 0.0463, ..., 0.0207, 0.0094, -0.0377], + [-0.0124, -0.0884, 0.0744, ..., -0.0974, 0.0343, -0.0295]], + device='cuda:0'), grad: tensor([[ 5.4264e-04, 6.8508e-06, 1.1837e-04, ..., 4.3321e-04, + 3.9399e-05, 0.0000e+00], + [-3.5381e-03, -2.8357e-05, 1.9407e-04, ..., -4.1466e-03, + 3.0249e-05, 0.0000e+00], + [ 3.4308e-04, 6.5117e-03, 1.5581e-04, ..., 3.0518e-04, + 9.1851e-05, 0.0000e+00], + ..., + [ 4.0817e-04, 1.7118e-04, 3.2139e-04, ..., 4.4465e-04, + 2.3639e-04, 4.6566e-10], + [ 1.2655e-03, 1.7628e-05, 4.6492e-04, ..., 9.5654e-04, + 2.7919e-04, 4.6566e-10], + [ 5.3787e-04, 1.7390e-05, -3.7479e-04, ..., 6.2656e-04, + -9.5665e-06, 9.3132e-10]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0140, -0.0024, -0.0138, 0.0102, -0.0156, -0.0213, 0.0108, -0.0019, + 0.0124, -0.0113], device='cuda:0'), grad: tensor([ 0.0245, -0.0954, 0.0325, -0.0082, 0.0066, 0.0079, 0.0091, 0.0159, + 0.0230, -0.0159], device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 216.63, cls_loss 0.4414 cls_loss_mapping 0.0008 cls_loss_causal 0.4098 re_mapping 0.0039 re_causal 0.0117 /// teacc 98.94 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.0754, -0.1090, -0.1536, ..., 0.1268, -0.0660, 0.0374], + [-0.0942, 0.1123, -0.0584, ..., 0.0518, -0.0636, -0.0970], + [-0.0385, 0.0885, -0.1339, ..., -0.0205, 0.0414, 0.1245], + ..., + [ 0.0021, 0.0744, 0.0016, ..., -0.0254, 0.0335, -0.0852], + [ 0.1020, -0.1194, 0.0462, ..., 0.0207, 0.0093, -0.0378], + [-0.0124, -0.0885, 0.0744, ..., -0.0975, 0.0343, -0.0295]], + device='cuda:0'), grad: tensor([[ 1.3673e-04, 4.0507e-04, 6.2323e-04, ..., 1.8435e-03, + 1.3590e-03, 7.8058e-04], + [ 2.8539e-04, 1.3554e-04, 3.0041e-04, ..., 5.9891e-04, + 3.5667e-04, 1.4675e-04], + [ 2.8205e-04, 4.5128e-03, 2.1019e-03, ..., 3.4962e-03, + 3.7575e-03, 2.3994e-03], + ..., + [ 1.5008e-04, -5.3749e-03, -3.5858e-03, ..., -2.6760e-03, + -4.1542e-03, -2.3746e-03], + [ 1.3838e-03, 3.1137e-04, 1.0881e-03, ..., 1.9207e-03, + 2.3060e-03, 6.0034e-04], + [-1.9872e-04, 1.0502e-04, 7.6234e-05, ..., -9.0933e-04, + 5.3978e-04, 1.8489e-04]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0140, -0.0024, -0.0138, 0.0100, -0.0156, -0.0212, 0.0109, -0.0019, + 0.0124, -0.0112], device='cuda:0'), grad: tensor([ 0.0322, 0.0230, 0.0223, -0.0247, -0.0067, -0.0205, 0.0148, -0.0377, + 0.0041, -0.0068], device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 216.88, cls_loss 0.4439 cls_loss_mapping 0.0007 cls_loss_causal 0.4177 re_mapping 0.0040 re_causal 0.0126 /// teacc 98.96 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.0753, -0.1090, -0.1536, ..., 0.1267, -0.0660, 0.0374], + [-0.0942, 0.1123, -0.0584, ..., 0.0519, -0.0635, -0.0969], + [-0.0387, 0.0884, -0.1339, ..., -0.0206, 0.0413, 0.1246], + ..., + [ 0.0020, 0.0745, 0.0017, ..., -0.0254, 0.0335, -0.0852], + [ 0.1019, -0.1194, 0.0462, ..., 0.0206, 0.0094, -0.0378], + [-0.0123, -0.0885, 0.0743, ..., -0.0975, 0.0342, -0.0296]], + device='cuda:0'), grad: tensor([[ 1.4389e-04, 8.8274e-05, 1.8072e-04, ..., 3.6716e-04, + 1.9896e-04, 4.2319e-05], + [-4.0323e-05, 9.4414e-05, 3.6907e-04, ..., 1.4293e-04, + 2.5272e-04, 6.5923e-05], + [-3.3913e-03, 1.1504e-04, 2.3162e-04, ..., 6.3610e-04, + 3.9840e-04, 5.2035e-05], + ..., + [ 1.5423e-05, 1.2994e-04, 6.1512e-04, ..., 7.2336e-04, + 3.7360e-04, 5.0664e-05], + [ 1.7083e-04, 1.0794e-04, 3.2306e-04, ..., 9.0265e-04, + 6.0081e-04, 4.7952e-05], + [ 9.1642e-06, 1.1063e-04, 1.7395e-03, ..., -1.0509e-03, + -1.5430e-03, 4.2915e-05]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0140, -0.0022, -0.0139, 0.0101, -0.0155, -0.0212, 0.0108, -0.0019, + 0.0123, -0.0112], device='cuda:0'), grad: tensor([ 0.0110, 0.0098, 0.0063, -0.0271, 0.0105, -0.0167, -0.0141, 0.0145, + 0.0182, -0.0124], device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 217.05, cls_loss 0.4460 cls_loss_mapping 0.0008 cls_loss_causal 0.4177 re_mapping 0.0039 re_causal 0.0120 /// teacc 98.95 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.0754, -0.1090, -0.1539, ..., 0.1266, -0.0662, 0.0374], + [-0.0943, 0.1122, -0.0584, ..., 0.0519, -0.0636, -0.0969], + [-0.0386, 0.0885, -0.1340, ..., -0.0206, 0.0416, 0.1248], + ..., + [ 0.0021, 0.0746, 0.0015, ..., -0.0253, 0.0333, -0.0851], + [ 0.1019, -0.1194, 0.0462, ..., 0.0206, 0.0094, -0.0378], + [-0.0123, -0.0886, 0.0743, ..., -0.0976, 0.0342, -0.0298]], + device='cuda:0'), grad: tensor([[ 4.6641e-06, 1.2314e-04, 1.7035e-04, ..., 3.6299e-05, + 7.4878e-06, 6.6698e-05], + [-6.3562e-04, -9.3758e-05, 3.5596e-04, ..., 2.2399e-04, + 2.1076e-04, 7.3493e-05], + [ 9.6023e-05, -1.5008e-04, 2.4366e-04, ..., -1.1152e-04, + 9.9361e-05, -8.1110e-04], + ..., + [ 1.8334e-04, -1.3704e-03, 9.9123e-05, ..., -7.3957e-04, + -4.9305e-04, 1.4114e-04], + [ 8.6904e-05, 3.3259e-04, 4.3249e-04, ..., 1.5390e-04, + 1.6451e-04, 1.5473e-04], + [-8.3208e-05, 3.3736e-04, -7.3862e-04, ..., 4.1544e-05, + -4.6659e-04, -7.2479e-05]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0139, -0.0022, -0.0138, 0.0101, -0.0154, -0.0211, 0.0108, -0.0020, + 0.0123, -0.0113], device='cuda:0'), grad: tensor([ 0.0081, 0.0158, 0.0082, 0.0129, 0.0164, 0.0048, -0.0150, -0.0104, + 0.0127, -0.0536], device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 216.66, cls_loss 0.4397 cls_loss_mapping 0.0008 cls_loss_causal 0.4199 re_mapping 0.0038 re_causal 0.0120 /// teacc 98.95 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.0755, -0.1090, -0.1538, ..., 0.1265, -0.0660, 0.0374], + [-0.0944, 0.1121, -0.0585, ..., 0.0518, -0.0635, -0.0971], + [-0.0387, 0.0885, -0.1339, ..., -0.0206, 0.0416, 0.1248], + ..., + [ 0.0020, 0.0746, 0.0015, ..., -0.0253, 0.0333, -0.0850], + [ 0.1020, -0.1194, 0.0461, ..., 0.0208, 0.0094, -0.0377], + [-0.0124, -0.0886, 0.0743, ..., -0.0976, 0.0342, -0.0298]], + device='cuda:0'), grad: tensor([[ 2.9460e-05, 2.6989e-04, 8.8930e-05, ..., 1.0788e-02, + 1.8263e-06, -2.2501e-05], + [ 2.9802e-05, -4.2534e-04, 1.1319e-04, ..., -9.6560e-05, + 2.9430e-05, 1.3597e-07], + [ 3.6150e-05, 8.2612e-05, -5.3883e-04, ..., -1.3514e-03, + 7.1526e-05, -1.2899e-07], + ..., + [-9.7334e-05, -5.0497e-04, 2.0587e-04, ..., 5.4121e-04, + 4.5776e-04, 6.2818e-07], + [-4.3335e-03, 4.2737e-05, 2.9635e-04, ..., 9.9945e-04, + 2.9564e-04, 3.7700e-06], + [ 3.4738e-04, 4.1890e-04, 1.6680e-03, ..., -1.8478e-02, + 1.3981e-03, 9.2611e-06]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0139, -0.0024, -0.0137, 0.0102, -0.0154, -0.0210, 0.0108, -0.0021, + 0.0123, -0.0114], device='cuda:0'), grad: tensor([ 0.0233, 0.0066, -0.0234, 0.0105, 0.0084, -0.0018, 0.0150, 0.0093, + -0.0365, -0.0117], device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 216.73, cls_loss 0.4552 cls_loss_mapping 0.0008 cls_loss_causal 0.4207 re_mapping 0.0039 re_causal 0.0121 /// teacc 98.93 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.0756, -0.1091, -0.1538, ..., 0.1266, -0.0661, 0.0374], + [-0.0944, 0.1122, -0.0586, ..., 0.0519, -0.0637, -0.0971], + [-0.0387, 0.0885, -0.1340, ..., -0.0208, 0.0415, 0.1248], + ..., + [ 0.0022, 0.0745, 0.0016, ..., -0.0254, 0.0334, -0.0850], + [ 0.1022, -0.1195, 0.0462, ..., 0.0207, 0.0094, -0.0377], + [-0.0124, -0.0886, 0.0744, ..., -0.0975, 0.0343, -0.0297]], + device='cuda:0'), grad: tensor([[ 5.5599e-04, 2.1791e-04, 4.8494e-04, ..., 4.5323e-04, + 1.5950e-04, 3.6284e-06], + [-1.5030e-03, -1.9350e-03, -9.7752e-04, ..., 1.1301e-04, + 2.0635e-04, 2.0891e-05], + [-9.3765e-03, 2.1708e-04, -1.4732e-02, ..., 1.3714e-03, + 4.4990e-04, -1.1343e-04], + ..., + [ 2.3575e-03, 1.6284e-04, 2.1362e-03, ..., 9.4461e-04, + 2.7847e-04, -1.8597e-05], + [ 1.3418e-03, 2.7418e-04, -1.0757e-03, ..., 1.0061e-03, + 5.8413e-04, 4.0948e-05], + [ 1.9341e-03, 3.8266e-04, 2.5330e-03, ..., 6.7949e-04, + 4.5514e-04, 3.3021e-05]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0140, -0.0023, -0.0139, 0.0101, -0.0154, -0.0209, 0.0108, -0.0021, + 0.0123, -0.0113], device='cuda:0'), grad: tensor([ 0.0175, -0.0565, -0.0068, -0.0059, 0.0052, 0.0362, 0.0194, 0.0007, + -0.0365, 0.0266], device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 216.42, cls_loss 0.4365 cls_loss_mapping 0.0009 cls_loss_causal 0.4127 re_mapping 0.0039 re_causal 0.0117 /// teacc 98.94 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.0755, -0.1092, -0.1537, ..., 0.1267, -0.0660, 0.0374], + [-0.0944, 0.1122, -0.0587, ..., 0.0518, -0.0636, -0.0972], + [-0.0386, 0.0885, -0.1337, ..., -0.0209, 0.0415, 0.1248], + ..., + [ 0.0023, 0.0746, 0.0017, ..., -0.0253, 0.0335, -0.0851], + [ 0.1022, -0.1194, 0.0461, ..., 0.0208, 0.0094, -0.0376], + [-0.0126, -0.0887, 0.0743, ..., -0.0975, 0.0343, -0.0297]], + device='cuda:0'), grad: tensor([[ 9.8407e-05, 9.5654e-04, 1.6940e-04, ..., 1.5306e-03, + 1.0567e-03, 6.8045e-04], + [-7.7820e-04, 4.8220e-05, 4.0030e-04, ..., 3.2825e-03, + 3.3474e-04, 3.0375e-04], + [ 1.2505e-04, 1.8167e-03, 2.0957e-04, ..., 3.0575e-03, + 1.7729e-03, 9.6321e-04], + ..., + [-1.5128e-04, -2.2411e-03, -2.6760e-03, ..., -8.3466e-03, + -4.4403e-03, -1.8921e-03], + [-3.0861e-03, -1.5175e-04, 2.7847e-04, ..., -3.2291e-03, + 6.9046e-04, -3.2282e-04], + [ 1.8275e-04, 2.5344e-04, 6.8760e-04, ..., 1.1234e-03, + 6.6423e-04, 1.0103e-04]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0139, -0.0023, -0.0139, 0.0101, -0.0153, -0.0210, 0.0109, -0.0020, + 0.0123, -0.0114], device='cuda:0'), grad: tensor([ 0.0128, -0.0008, 0.0186, -0.0095, 0.0109, -0.0131, 0.0112, -0.0402, + -0.0041, 0.0144], device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 217.01, cls_loss 0.4348 cls_loss_mapping 0.0008 cls_loss_causal 0.4135 re_mapping 0.0038 re_causal 0.0117 /// teacc 98.94 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.0756, -0.1092, -0.1537, ..., 0.1268, -0.0661, 0.0374], + [-0.0943, 0.1123, -0.0587, ..., 0.0519, -0.0636, -0.0973], + [-0.0388, 0.0884, -0.1337, ..., -0.0210, 0.0414, 0.1245], + ..., + [ 0.0022, 0.0746, 0.0017, ..., -0.0253, 0.0336, -0.0850], + [ 0.1024, -0.1196, 0.0461, ..., 0.0206, 0.0094, -0.0376], + [-0.0126, -0.0887, 0.0743, ..., -0.0975, 0.0343, -0.0295]], + device='cuda:0'), grad: tensor([[ 6.1870e-05, 9.4891e-05, 2.4843e-04, ..., 7.5400e-05, + 9.7871e-05, -9.2149e-05], + [ 6.1691e-05, -9.8419e-04, -8.6641e-04, ..., -1.3428e-03, + 2.3198e-04, 1.1832e-04], + [ 2.2125e-03, 7.1812e-04, 1.2188e-03, ..., 2.8992e-04, + 4.6039e-04, 1.1530e-03], + ..., + [ 7.1621e-04, 5.0926e-04, 2.4338e-03, ..., 2.1172e-03, + 1.3762e-03, 2.1362e-04], + [ 4.9496e-04, 6.8140e-04, 1.2703e-03, ..., 1.0881e-03, + 3.9506e-04, 1.5593e-04], + [-7.0143e-04, 6.0987e-04, -9.7513e-04, ..., -1.1635e-03, + -1.0900e-03, 1.9002e-04]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0137, -0.0022, -0.0139, 0.0101, -0.0153, -0.0210, 0.0107, -0.0019, + 0.0123, -0.0112], device='cuda:0'), grad: tensor([ 0.0119, 0.0060, 0.0322, -0.0138, -0.0060, 0.0106, -0.0472, 0.0339, + -0.0089, -0.0187], device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 216.70, cls_loss 0.4500 cls_loss_mapping 0.0009 cls_loss_causal 0.4216 re_mapping 0.0038 re_causal 0.0118 /// teacc 98.95 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.0756, -0.1091, -0.1538, ..., 0.1269, -0.0661, 0.0375], + [-0.0942, 0.1123, -0.0588, ..., 0.0519, -0.0637, -0.0974], + [-0.0388, 0.0884, -0.1339, ..., -0.0208, 0.0415, 0.1245], + ..., + [ 0.0023, 0.0746, 0.0017, ..., -0.0253, 0.0337, -0.0849], + [ 0.1023, -0.1196, 0.0462, ..., 0.0206, 0.0094, -0.0376], + [-0.0124, -0.0886, 0.0743, ..., -0.0976, 0.0342, -0.0295]], + device='cuda:0'), grad: tensor([[ 1.1638e-05, 9.5248e-05, 2.7370e-03, ..., 1.8096e-04, + 3.6144e-03, 3.0494e-04], + [ 1.6153e-05, -2.2399e-04, -8.7976e-04, ..., -6.2799e-04, + -3.9368e-03, 7.7114e-06], + [ 3.0613e-04, 9.7394e-05, 1.8799e-04, ..., 2.1601e-04, + 8.5115e-04, 4.3474e-06], + ..., + [ 5.4121e-05, 2.9325e-04, 4.4670e-03, ..., 1.2791e-04, + 2.3956e-03, 7.1786e-06], + [-1.4591e-03, 1.2201e-04, 1.2569e-03, ..., 2.2328e-04, + 2.0466e-03, 1.8001e-05], + [-1.4820e-03, -1.4174e-04, -1.5099e-02, ..., 1.9693e-04, + -1.1971e-02, -6.1512e-04]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0139, -0.0023, -0.0140, 0.0103, -0.0154, -0.0210, 0.0106, -0.0020, + 0.0123, -0.0112], device='cuda:0'), grad: tensor([ 0.0274, -0.0278, 0.0184, 0.0229, 0.0226, -0.0121, -0.0092, 0.0383, + -0.0149, -0.0655], device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 216.70, cls_loss 0.4123 cls_loss_mapping 0.0007 cls_loss_causal 0.3876 re_mapping 0.0039 re_causal 0.0117 /// teacc 98.93 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.0757, -0.1092, -0.1538, ..., 0.1267, -0.0662, 0.0375], + [-0.0941, 0.1124, -0.0587, ..., 0.0520, -0.0636, -0.0975], + [-0.0387, 0.0885, -0.1338, ..., -0.0207, 0.0416, 0.1246], + ..., + [ 0.0023, 0.0746, 0.0017, ..., -0.0253, 0.0335, -0.0848], + [ 0.1022, -0.1197, 0.0462, ..., 0.0205, 0.0094, -0.0376], + [-0.0125, -0.0887, 0.0745, ..., -0.0975, 0.0342, -0.0296]], + device='cuda:0'), grad: tensor([[-5.3482e-03, 3.3751e-06, 8.1587e-04, ..., -3.5210e-03, + 6.1035e-04, 2.9355e-05], + [ 3.9177e-03, -3.4332e-04, 2.2240e-03, ..., 2.4548e-03, + 1.0948e-03, 1.9029e-05], + [ 2.7347e-04, 9.3341e-05, 1.6975e-04, ..., 4.2820e-04, + 1.1854e-03, 1.5855e-04], + ..., + [ 6.6137e-04, 1.5032e-04, 3.1900e-04, ..., 5.5647e-04, + 7.8917e-04, 1.1128e-04], + [ 1.0681e-03, 1.2591e-05, 1.1911e-03, ..., 3.3116e-04, + 7.6818e-04, -5.0925e-06], + [ 9.1648e-04, 2.4363e-05, 5.6744e-04, ..., 2.3985e-04, + 5.0974e-04, 2.7999e-05]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0139, -0.0021, -0.0140, 0.0103, -0.0154, -0.0211, 0.0106, -0.0019, + 0.0122, -0.0113], device='cuda:0'), grad: tensor([-0.0237, 0.0340, 0.0085, -0.0118, -0.0440, -0.0077, 0.0060, 0.0089, + 0.0184, 0.0114], device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 216.52, cls_loss 0.4579 cls_loss_mapping 0.0008 cls_loss_causal 0.4299 re_mapping 0.0040 re_causal 0.0124 /// teacc 98.91 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.0757, -0.1090, -0.1539, ..., 0.1268, -0.0662, 0.0376], + [-0.0942, 0.1124, -0.0588, ..., 0.0519, -0.0637, -0.0976], + [-0.0387, 0.0885, -0.1338, ..., -0.0207, 0.0417, 0.1245], + ..., + [ 0.0022, 0.0746, 0.0015, ..., -0.0254, 0.0335, -0.0848], + [ 0.1022, -0.1198, 0.0465, ..., 0.0206, 0.0098, -0.0373], + [-0.0124, -0.0888, 0.0744, ..., -0.0976, 0.0341, -0.0296]], + device='cuda:0'), grad: tensor([[ 1.0294e-04, 7.4625e-05, 1.4436e-04, ..., 3.1567e-04, + 8.9884e-05, 2.5004e-05], + [-3.7193e-03, -2.7466e-03, -3.6983e-03, ..., 3.2687e-04, + 3.0804e-04, 1.4603e-05], + [ 2.8634e-04, 1.0185e-03, 3.1614e-04, ..., 8.7214e-04, + 1.5211e-04, 1.4015e-05], + ..., + [ 9.3412e-04, 1.1887e-02, 1.3170e-03, ..., 3.4924e-03, + 3.4285e-04, 7.8902e-06], + [ 2.5535e-04, 1.3208e-03, 1.9150e-03, ..., 6.5899e-04, + 3.2544e-04, 1.0115e-04], + [ 3.4404e-04, -1.2054e-02, 4.1151e-04, ..., -3.7632e-03, + 2.8777e-04, 2.4617e-05]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0138, -0.0021, -0.0141, 0.0103, -0.0153, -0.0210, 0.0107, -0.0020, + 0.0124, -0.0114], device='cuda:0'), grad: tensor([ 0.0212, 0.0143, -0.0060, -0.0354, 0.0014, -0.0529, 0.0284, 0.0370, + 0.0321, -0.0400], device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 216.76, cls_loss 0.4736 cls_loss_mapping 0.0010 cls_loss_causal 0.4415 re_mapping 0.0038 re_causal 0.0119 /// teacc 98.92 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.0757, -0.1090, -0.1538, ..., 0.1268, -0.0663, 0.0377], + [-0.0943, 0.1124, -0.0588, ..., 0.0519, -0.0637, -0.0976], + [-0.0387, 0.0885, -0.1339, ..., -0.0207, 0.0416, 0.1244], + ..., + [ 0.0025, 0.0746, 0.0015, ..., -0.0253, 0.0336, -0.0848], + [ 0.1023, -0.1198, 0.0464, ..., 0.0206, 0.0100, -0.0372], + [-0.0126, -0.0886, 0.0743, ..., -0.0976, 0.0341, -0.0297]], + device='cuda:0'), grad: tensor([[ 1.0097e-04, 5.8770e-05, 4.4197e-05, ..., 4.1986e-04, + 1.0508e-04, 2.2388e-04], + [ 2.6777e-05, -1.8349e-03, 6.4909e-05, ..., -2.3537e-03, + -3.4409e-03, 5.6952e-05], + [ 8.7261e-05, -8.7261e-04, 1.2982e-04, ..., 3.8409e-04, + 1.6105e-04, 1.2386e-04], + ..., + [-1.0157e-03, 1.2302e-04, 4.7088e-04, ..., -2.3251e-03, + 5.0020e-04, -2.2297e-03], + [ 1.1927e-04, 1.5182e-03, -1.1854e-03, ..., 2.0447e-03, + 1.5211e-03, 1.5163e-04], + [-2.8872e-04, 1.1808e-04, -1.1492e-04, ..., 3.2663e-04, + 3.4523e-04, 5.0038e-05]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0136, -0.0021, -0.0139, 0.0103, -0.0154, -0.0210, 0.0107, -0.0019, + 0.0125, -0.0115], device='cuda:0'), grad: tensor([-0.0155, -0.0075, -0.0371, 0.0232, 0.0297, 0.0206, -0.0148, -0.0426, + 0.0293, 0.0147], device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 216.48, cls_loss 0.4403 cls_loss_mapping 0.0008 cls_loss_causal 0.4146 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.90 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.0758, -0.1090, -0.1538, ..., 0.1267, -0.0664, 0.0377], + [-0.0944, 0.1125, -0.0587, ..., 0.0520, -0.0636, -0.0976], + [-0.0386, 0.0885, -0.1340, ..., -0.0206, 0.0416, 0.1244], + ..., + [ 0.0025, 0.0746, 0.0016, ..., -0.0253, 0.0338, -0.0847], + [ 0.1023, -0.1199, 0.0465, ..., 0.0207, 0.0099, -0.0371], + [-0.0127, -0.0886, 0.0743, ..., -0.0978, 0.0340, -0.0298]], + device='cuda:0'), grad: tensor([[ 8.8802e-07, 1.2696e-04, 8.2888e-08, ..., 1.1772e-04, + 2.0832e-05, 7.2177e-07], + [-8.3745e-06, 1.7762e-04, 1.4808e-07, ..., 1.6940e-04, + 4.2081e-05, 6.5155e-06], + [ 1.2852e-07, 7.4029e-05, 1.6484e-07, ..., 9.0659e-05, + 2.4676e-05, -6.4015e-05], + ..., + [ 7.8902e-06, 1.6570e-04, 1.1787e-05, ..., 1.9813e-04, + 4.2915e-05, 5.9186e-07], + [ 6.4401e-07, 2.0778e-04, 1.2919e-05, ..., 1.7667e-04, + 3.4899e-05, 5.1171e-05], + [ 8.9547e-07, 1.8656e-04, -3.9041e-05, ..., 1.5247e-04, + -9.3356e-06, 3.1432e-07]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0136, -0.0022, -0.0139, 0.0105, -0.0154, -0.0209, 0.0107, -0.0019, + 0.0123, -0.0116], device='cuda:0'), grad: tensor([ 0.0134, 0.0175, 0.0128, 0.0147, -0.0457, -0.0194, 0.0199, -0.0166, + 0.0178, -0.0142], device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 217.78, cls_loss 0.4448 cls_loss_mapping 0.0008 cls_loss_causal 0.4166 re_mapping 0.0037 re_causal 0.0116 /// teacc 98.95 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.0757, -0.1091, -0.1539, ..., 0.1267, -0.0664, 0.0378], + [-0.0945, 0.1125, -0.0587, ..., 0.0520, -0.0636, -0.0977], + [-0.0386, 0.0887, -0.1339, ..., -0.0206, 0.0418, 0.1246], + ..., + [ 0.0026, 0.0746, 0.0015, ..., -0.0253, 0.0338, -0.0846], + [ 0.1024, -0.1200, 0.0464, ..., 0.0207, 0.0098, -0.0374], + [-0.0128, -0.0888, 0.0743, ..., -0.0977, 0.0340, -0.0298]], + device='cuda:0'), grad: tensor([[ 5.4389e-05, 2.3954e-06, 1.6320e-04, ..., 2.6751e-04, + 1.1957e-04, 9.4473e-05], + [ 9.5963e-05, 9.5218e-06, 4.6343e-05, ..., 1.5581e-04, + 4.1753e-05, 4.0621e-05], + [ 7.6234e-05, -8.3596e-06, -5.1498e-05, ..., -3.4475e-04, + -7.8726e-04, -2.0695e-04], + ..., + [-6.0034e-04, -1.4484e-05, 8.4341e-05, ..., 2.2757e-04, + 1.1843e-04, 6.1035e-05], + [ 5.7757e-05, -4.4882e-05, 3.0804e-04, ..., 5.3406e-04, + 7.1955e-04, 2.3246e-04], + [ 3.7551e-05, 3.1292e-05, 3.0804e-04, ..., 1.4114e-04, + 3.0446e-04, 1.7035e-04]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0137, -0.0021, -0.0139, 0.0105, -0.0154, -0.0209, 0.0106, -0.0018, + 0.0123, -0.0117], device='cuda:0'), grad: tensor([ 0.0184, 0.0229, 0.0161, -0.0138, 0.0217, 0.0256, -0.0508, -0.0436, + -0.0149, 0.0184], device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 217.40, cls_loss 0.4728 cls_loss_mapping 0.0008 cls_loss_causal 0.4367 re_mapping 0.0038 re_causal 0.0118 /// teacc 98.99 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.0758, -0.1092, -0.1541, ..., 0.1268, -0.0665, 0.0377], + [-0.0946, 0.1126, -0.0586, ..., 0.0520, -0.0636, -0.0978], + [-0.0385, 0.0886, -0.1338, ..., -0.0207, 0.0417, 0.1245], + ..., + [ 0.0025, 0.0745, 0.0015, ..., -0.0253, 0.0338, -0.0846], + [ 0.1024, -0.1200, 0.0463, ..., 0.0206, 0.0098, -0.0373], + [-0.0128, -0.0887, 0.0742, ..., -0.0977, 0.0340, -0.0299]], + device='cuda:0'), grad: tensor([[ 3.8324e-07, 5.5265e-04, 6.3086e-04, ..., 1.5574e-03, + 6.6805e-04, 9.1553e-04], + [ 1.9874e-06, 3.0041e-03, 2.0194e-04, ..., 3.7556e-03, + 7.9453e-05, 4.5705e-04], + [ 5.5075e-05, 3.2806e-04, 1.5616e-04, ..., 9.6607e-04, + 7.1096e-04, 5.3263e-04], + ..., + [ 2.6003e-05, 3.1233e-04, 1.7715e-04, ..., 7.9536e-04, + 3.0160e-04, 4.0579e-04], + [ 1.1343e-04, -3.6926e-03, -1.2457e-04, ..., -3.5210e-03, + 1.1616e-03, -6.8760e-04], + [ 1.3046e-05, 2.4724e-04, 2.4319e-04, ..., 4.9639e-04, + 2.7490e-04, 3.1471e-04]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0137, -0.0022, -0.0140, 0.0106, -0.0153, -0.0209, 0.0105, -0.0018, + 0.0123, -0.0117], device='cuda:0'), grad: tensor([ 0.0229, 0.0034, 0.0191, -0.0535, -0.0155, 0.0236, -0.0155, 0.0206, + -0.0203, 0.0152], device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 216.86, cls_loss 0.4747 cls_loss_mapping 0.0009 cls_loss_causal 0.4447 re_mapping 0.0036 re_causal 0.0113 /// teacc 98.99 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.0759, -0.1092, -0.1541, ..., 0.1268, -0.0666, 0.0377], + [-0.0946, 0.1129, -0.0588, ..., 0.0518, -0.0635, -0.0979], + [-0.0385, 0.0884, -0.1339, ..., -0.0206, 0.0417, 0.1246], + ..., + [ 0.0027, 0.0746, 0.0015, ..., -0.0252, 0.0338, -0.0845], + [ 0.1025, -0.1201, 0.0463, ..., 0.0206, 0.0098, -0.0372], + [-0.0129, -0.0887, 0.0744, ..., -0.0977, 0.0341, -0.0300]], + device='cuda:0'), grad: tensor([[ 1.1005e-03, 3.2783e-05, 1.3518e-04, ..., 1.5173e-03, + 1.9388e-03, 2.4486e-04], + [ 1.1766e-04, 9.5218e-06, 4.8661e-04, ..., 2.6536e-04, + 5.9381e-06, 1.6168e-05], + [ 1.1837e-04, 8.9228e-05, 9.4354e-05, ..., 3.3903e-04, + 5.6885e-06, 1.3262e-05], + ..., + [-3.5691e-04, -5.4598e-04, 3.4237e-04, ..., -3.3245e-03, + -3.3200e-05, 3.0577e-05], + [ 4.9496e-04, 1.0148e-05, 2.7704e-04, ..., 3.7670e-04, + 3.1829e-05, 9.0003e-05], + [ 9.3365e-04, 4.0197e-04, 7.6389e-04, ..., 5.7459e-04, + 2.8625e-05, 1.7154e-04]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0136, -0.0021, -0.0141, 0.0106, -0.0155, -0.0209, 0.0104, -0.0017, + 0.0125, -0.0117], device='cuda:0'), grad: tensor([ 0.0142, 0.0083, 0.0085, -0.0226, -0.0080, 0.0098, 0.0078, -0.0127, + -0.0205, 0.0151], device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 217.27, cls_loss 0.4546 cls_loss_mapping 0.0010 cls_loss_causal 0.4255 re_mapping 0.0036 re_causal 0.0108 /// teacc 98.94 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.0759, -0.1092, -0.1542, ..., 0.1268, -0.0666, 0.0378], + [-0.0946, 0.1129, -0.0588, ..., 0.0518, -0.0635, -0.0979], + [-0.0386, 0.0884, -0.1338, ..., -0.0206, 0.0417, 0.1246], + ..., + [ 0.0026, 0.0746, 0.0014, ..., -0.0251, 0.0338, -0.0845], + [ 0.1026, -0.1203, 0.0463, ..., 0.0206, 0.0096, -0.0374], + [-0.0129, -0.0887, 0.0743, ..., -0.0978, 0.0341, -0.0299]], + device='cuda:0'), grad: tensor([[ 1.5205e-02, 5.6297e-05, 5.0545e-04, ..., 4.8971e-04, + 3.8362e-04, 6.3753e-04], + [-1.3876e-03, -4.7660e-04, -7.8917e-04, ..., -1.9417e-03, + -6.3038e-04, -1.9684e-03], + [-4.4861e-03, -6.5651e-03, 3.8242e-04, ..., 6.8283e-04, + -5.2071e-03, -1.5480e-02], + ..., + [ 3.7599e-04, -3.1888e-05, 9.0313e-04, ..., 5.5838e-04, + 6.8665e-04, 5.4836e-04], + [ 3.3188e-03, 5.8860e-05, 3.8075e-04, ..., 4.3559e-04, + 2.9516e-04, 4.3511e-04], + [ 1.4515e-03, 7.1704e-05, 4.7684e-04, ..., 4.6802e-04, + 3.3450e-04, 4.5228e-04]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0136, -0.0021, -0.0141, 0.0106, -0.0155, -0.0209, 0.0105, -0.0017, + 0.0125, -0.0117], device='cuda:0'), grad: tensor([-0.0079, 0.0002, -0.0824, -0.0021, -0.0006, 0.0231, -0.0021, 0.0017, + 0.0371, 0.0329], device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 216.89, cls_loss 0.4678 cls_loss_mapping 0.0008 cls_loss_causal 0.4429 re_mapping 0.0037 re_causal 0.0116 /// teacc 98.95 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.0759, -0.1092, -0.1543, ..., 0.1268, -0.0667, 0.0378], + [-0.0946, 0.1129, -0.0588, ..., 0.0517, -0.0635, -0.0978], + [-0.0385, 0.0886, -0.1338, ..., -0.0205, 0.0417, 0.1248], + ..., + [ 0.0026, 0.0745, 0.0014, ..., -0.0253, 0.0339, -0.0846], + [ 0.1026, -0.1204, 0.0463, ..., 0.0206, 0.0096, -0.0374], + [-0.0129, -0.0887, 0.0743, ..., -0.0979, 0.0341, -0.0300]], + device='cuda:0'), grad: tensor([[ 2.9707e-04, 5.8040e-06, 1.1418e-06, ..., 4.1056e-04, + 1.0813e-06, 5.9247e-05], + [ 2.2006e-04, 7.2598e-05, 5.6058e-05, ..., 6.8283e-04, + 3.1292e-05, 9.6858e-05], + [-2.0924e-03, 1.0605e-02, 3.1203e-05, ..., 7.1526e-03, + -1.5147e-05, 4.5166e-03], + ..., + [ 6.1631e-05, -1.0780e-02, 5.9992e-05, ..., -8.5754e-03, + 1.4499e-05, -4.5509e-03], + [ 6.8855e-04, 4.0531e-05, -2.0087e-04, ..., 2.4533e-04, + -1.3506e-04, 9.2626e-05], + [ 5.7578e-05, 6.0871e-06, -1.2153e-04, ..., 3.2997e-04, + -1.4983e-05, 4.3154e-05]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0135, -0.0020, -0.0140, 0.0108, -0.0154, -0.0210, 0.0104, -0.0018, + 0.0125, -0.0118], device='cuda:0'), grad: tensor([ 0.0158, 0.0205, 0.0041, -0.0166, 0.0116, 0.0135, -0.0117, -0.0712, + 0.0231, 0.0110], device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 216.90, cls_loss 0.4594 cls_loss_mapping 0.0008 cls_loss_causal 0.4330 re_mapping 0.0038 re_causal 0.0119 /// teacc 98.95 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.0760, -0.1090, -0.1545, ..., 0.1268, -0.0668, 0.0376], + [-0.0945, 0.1129, -0.0587, ..., 0.0517, -0.0634, -0.0980], + [-0.0386, 0.0886, -0.1339, ..., -0.0205, 0.0418, 0.1248], + ..., + [ 0.0026, 0.0745, 0.0014, ..., -0.0251, 0.0338, -0.0847], + [ 0.1024, -0.1205, 0.0463, ..., 0.0207, 0.0097, -0.0373], + [-0.0130, -0.0888, 0.0743, ..., -0.0980, 0.0340, -0.0299]], + device='cuda:0'), grad: tensor([[ 8.3804e-05, 1.1778e-04, 2.0874e-04, ..., -3.2043e-04, + 1.6367e-04, 1.8501e-04], + [ 1.2279e-04, 1.6665e-04, 3.8356e-05, ..., 1.3247e-05, + 2.1636e-05, 3.6764e-04], + [ 3.8099e-04, 6.0177e-04, 1.4699e-04, ..., 8.6129e-05, + 7.3493e-05, 1.0662e-03], + ..., + [-1.3714e-03, -2.3117e-03, -5.6534e-03, ..., -3.0937e-03, + -4.9591e-03, -2.3289e-03], + [ 2.1172e-04, 2.7800e-04, -1.4400e-03, ..., 3.8886e-04, + 3.6454e-04, 4.2009e-04], + [ 1.0166e-03, 1.8692e-03, 5.8823e-03, ..., 2.3537e-03, + 3.7708e-03, 1.8415e-03]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0134, -0.0021, -0.0140, 0.0109, -0.0155, -0.0211, 0.0107, -0.0017, + 0.0124, -0.0118], device='cuda:0'), grad: tensor([-0.0146, -0.0083, 0.0292, -0.0081, 0.0170, 0.0154, -0.0453, -0.0406, + 0.0111, 0.0442], device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 217.23, cls_loss 0.4573 cls_loss_mapping 0.0008 cls_loss_causal 0.4301 re_mapping 0.0038 re_causal 0.0120 /// teacc 98.95 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.0761, -0.1091, -0.1546, ..., 0.1268, -0.0669, 0.0375], + [-0.0946, 0.1129, -0.0588, ..., 0.0517, -0.0633, -0.0980], + [-0.0384, 0.0885, -0.1339, ..., -0.0206, 0.0418, 0.1248], + ..., + [ 0.0025, 0.0747, 0.0014, ..., -0.0248, 0.0339, -0.0846], + [ 0.1024, -0.1206, 0.0463, ..., 0.0206, 0.0096, -0.0374], + [-0.0131, -0.0888, 0.0743, ..., -0.0980, 0.0340, -0.0299]], + device='cuda:0'), grad: tensor([[ 4.2844e-04, 3.3665e-04, 6.1655e-04, ..., 6.7139e-04, + -6.7472e-04, -5.7757e-05], + [-2.1439e-03, -2.2907e-03, 1.0939e-03, ..., -6.9084e-03, + 5.6171e-04, 7.4327e-05], + [ 3.1471e-04, 6.2895e-04, 1.9093e-03, ..., 5.5981e-04, + 9.6817e-03, 3.2940e-03], + ..., + [ 6.3837e-05, 6.1893e-04, 1.1444e-03, ..., 2.8496e-03, + -1.2421e-02, 5.4240e-05], + [ 1.4842e-05, 3.1352e-04, 8.6212e-04, ..., 1.1177e-02, + 1.8845e-02, 6.5384e-03], + [ 3.9911e-04, 3.2115e-04, -3.3264e-03, ..., 5.0497e-04, + 4.7922e-04, 1.0049e-04]], device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0135, -0.0021, -0.0141, 0.0109, -0.0155, -0.0212, 0.0108, -0.0017, + 0.0124, -0.0119], device='cuda:0'), grad: tensor([-0.0114, -0.0622, -0.0231, 0.0161, -0.0011, 0.0240, 0.0013, 0.0013, + 0.0401, 0.0150], device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 218.00, cls_loss 0.4478 cls_loss_mapping 0.0007 cls_loss_causal 0.4242 re_mapping 0.0039 re_causal 0.0120 /// teacc 98.97 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.0761, -0.1091, -0.1546, ..., 0.1268, -0.0668, 0.0376], + [-0.0949, 0.1128, -0.0590, ..., 0.0517, -0.0634, -0.0979], + [-0.0386, 0.0886, -0.1337, ..., -0.0205, 0.0420, 0.1248], + ..., + [ 0.0027, 0.0747, 0.0013, ..., -0.0249, 0.0338, -0.0846], + [ 0.1026, -0.1206, 0.0462, ..., 0.0206, 0.0096, -0.0374], + [-0.0131, -0.0888, 0.0743, ..., -0.0981, 0.0340, -0.0301]], + device='cuda:0'), grad: tensor([[-7.1585e-05, -2.7447e-03, 9.8610e-04, ..., 4.7755e-04, + 7.7248e-03, 5.0201e-03], + [-8.1730e-04, 3.8986e-03, 3.9411e-04, ..., 6.5660e-04, + 5.1689e-04, 4.3869e-05], + [ 2.2149e-04, -3.7527e-04, -3.3512e-03, ..., -3.4332e-04, + 5.2810e-05, -2.7199e-03], + ..., + [-2.2113e-04, -3.3531e-03, -1.3838e-03, ..., -2.1572e-03, + -5.2547e-04, 6.2943e-04], + [-9.6750e-04, 3.3116e-04, -7.1716e-04, ..., -1.7185e-03, + -9.4604e-03, -4.8599e-03], + [ 5.2738e-04, 4.6134e-04, 6.2370e-04, ..., 1.1559e-03, + 5.0974e-04, 8.6784e-04]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0136, -0.0020, -0.0141, 0.0107, -0.0154, -0.0212, 0.0107, -0.0016, + 0.0125, -0.0121], device='cuda:0'), grad: tensor([-0.0148, -0.0068, -0.0140, -0.0079, 0.0195, 0.0190, 0.0030, 0.0178, + -0.0412, 0.0253], device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 217.68, cls_loss 0.4420 cls_loss_mapping 0.0007 cls_loss_causal 0.4120 re_mapping 0.0037 re_causal 0.0116 /// teacc 98.99 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.0762, -0.1091, -0.1546, ..., 0.1264, -0.0670, 0.0375], + [-0.0948, 0.1130, -0.0591, ..., 0.0519, -0.0634, -0.0975], + [-0.0386, 0.0886, -0.1338, ..., -0.0204, 0.0420, 0.1248], + ..., + [ 0.0028, 0.0746, 0.0013, ..., -0.0250, 0.0337, -0.0848], + [ 0.1026, -0.1206, 0.0464, ..., 0.0208, 0.0097, -0.0373], + [-0.0132, -0.0887, 0.0742, ..., -0.0982, 0.0341, -0.0301]], + device='cuda:0'), grad: tensor([[ 5.1546e-04, 6.4969e-05, 2.6393e-04, ..., 7.1812e-04, + 3.6895e-05, 3.1982e-06], + [ 1.3769e-04, -7.6246e-04, 7.2062e-05, ..., 4.3945e-03, + -4.3559e-04, 1.4141e-05], + [ 6.7949e-05, 4.9305e-04, 7.9691e-05, ..., 7.3767e-04, + 2.7966e-04, 4.5836e-05], + ..., + [ 2.0623e-04, -8.3542e-04, 2.9907e-03, ..., 4.9084e-05, + 2.6989e-03, -1.2827e-04], + [ 2.2488e-03, 1.6546e-04, 1.2665e-03, ..., 1.0405e-03, + 1.9383e-04, 1.0774e-05], + [-8.0347e-04, 1.8537e-04, -5.5580e-03, ..., 5.3644e-04, + -3.7384e-03, 1.3947e-05]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0137, -0.0019, -0.0141, 0.0106, -0.0153, -0.0211, 0.0108, -0.0016, + 0.0124, -0.0122], device='cuda:0'), grad: tensor([ 0.0114, -0.0031, 0.0135, -0.0193, 0.0171, -0.0207, -0.0039, 0.0100, + 0.0202, -0.0251], device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.05, cls_loss 0.4641 cls_loss_mapping 0.0008 cls_loss_causal 0.4383 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.01 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.0763, -0.1089, -0.1546, ..., 0.1265, -0.0669, 0.0376], + [-0.0950, 0.1130, -0.0591, ..., 0.0518, -0.0634, -0.0976], + [-0.0385, 0.0886, -0.1338, ..., -0.0206, 0.0419, 0.1248], + ..., + [ 0.0028, 0.0746, 0.0013, ..., -0.0249, 0.0338, -0.0848], + [ 0.1026, -0.1206, 0.0465, ..., 0.0210, 0.0099, -0.0373], + [-0.0133, -0.0888, 0.0743, ..., -0.0982, 0.0339, -0.0302]], + device='cuda:0'), grad: tensor([[ 3.1900e-04, 2.7823e-04, 1.4341e-04, ..., 7.0715e-04, + 2.8396e-04, 4.2152e-04], + [ 5.6505e-04, -1.4937e-04, 9.3699e-05, ..., -6.5708e-04, + 9.6917e-05, 5.6148e-05], + [ 1.6761e-04, 2.0826e-04, 1.5068e-04, ..., 6.6662e-04, + 1.9097e-04, 2.7084e-04], + ..., + [ 1.4915e-03, -3.1203e-05, 1.8597e-03, ..., -2.6741e-03, + 9.0694e-04, 3.0756e-05], + [ 6.9237e-04, 3.6454e-04, -5.1832e-04, ..., 1.0939e-03, + -2.4509e-04, 2.1446e-04], + [-1.1780e-02, -4.4084e-04, -2.7161e-03, ..., -3.7169e-04, + -1.4744e-03, 1.0967e-04]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0138, -0.0020, -0.0143, 0.0107, -0.0152, -0.0212, 0.0109, -0.0016, + 0.0124, -0.0122], device='cuda:0'), grad: tensor([-0.0091, 0.0157, -0.0125, 0.0218, -0.0028, 0.0069, 0.0035, 0.0050, + 0.0216, -0.0500], device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 216.56, cls_loss 0.4194 cls_loss_mapping 0.0007 cls_loss_causal 0.3969 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.01 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.0764, -0.1089, -0.1547, ..., 0.1266, -0.0670, 0.0376], + [-0.0950, 0.1130, -0.0590, ..., 0.0517, -0.0634, -0.0975], + [-0.0385, 0.0887, -0.1338, ..., -0.0206, 0.0419, 0.1249], + ..., + [ 0.0027, 0.0746, 0.0012, ..., -0.0249, 0.0337, -0.0849], + [ 0.1026, -0.1208, 0.0466, ..., 0.0209, 0.0100, -0.0372], + [-0.0133, -0.0889, 0.0742, ..., -0.0984, 0.0339, -0.0301]], + device='cuda:0'), grad: tensor([[ 6.7997e-04, 1.5891e-04, -4.1351e-03, ..., -6.2599e-03, + -5.0087e-03, -2.0847e-03], + [ 3.2368e-03, 1.1454e-03, 2.1720e-04, ..., 1.9445e-03, + 2.5225e-04, 1.0037e-04], + [ 6.8903e-04, 1.2231e-04, 5.9652e-04, ..., 7.4339e-04, + 1.1253e-03, 2.1577e-04], + ..., + [ 1.8387e-03, 8.8632e-05, 2.8896e-04, ..., 1.3037e-03, + -6.0034e-04, 3.3951e-04], + [-1.2260e-02, 7.4720e-04, 1.2159e-03, ..., 2.1629e-03, + 1.7872e-03, 1.6844e-04], + [ 4.2229e-03, 9.7752e-05, 4.3154e-04, ..., 2.6093e-03, + 4.6015e-04, 8.4066e-04]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0138, -0.0020, -0.0142, 0.0106, -0.0153, -0.0211, 0.0108, -0.0016, + 0.0124, -0.0122], device='cuda:0'), grad: tensor([-0.0086, -0.0053, 0.0138, -0.0183, 0.0187, -0.0206, 0.0134, -0.0170, + 0.0013, 0.0224], device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 217.61, cls_loss 0.4222 cls_loss_mapping 0.0008 cls_loss_causal 0.3929 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.01 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.0764, -0.1089, -0.1548, ..., 0.1267, -0.0671, 0.0376], + [-0.0951, 0.1128, -0.0591, ..., 0.0516, -0.0634, -0.0975], + [-0.0386, 0.0887, -0.1340, ..., -0.0203, 0.0419, 0.1250], + ..., + [ 0.0026, 0.0746, 0.0011, ..., -0.0251, 0.0335, -0.0849], + [ 0.1027, -0.1208, 0.0466, ..., 0.0208, 0.0100, -0.0373], + [-0.0131, -0.0889, 0.0743, ..., -0.0986, 0.0340, -0.0302]], + device='cuda:0'), grad: tensor([[ 5.8394e-07, 3.7670e-05, 2.1148e-04, ..., 2.7561e-04, + 2.1428e-05, 2.7940e-07], + [ 4.6566e-10, 6.4187e-06, 1.2422e-04, ..., 1.7047e-04, + 1.5855e-05, 1.0384e-07], + [ 3.7719e-08, 2.6684e-03, 4.5836e-05, ..., 2.4939e-04, + 4.3821e-04, -2.0787e-06], + ..., + [ 1.1642e-08, -4.8561e-03, -2.3087e-02, ..., 1.5318e-04, + -6.4392e-03, 8.2329e-07], + [ 1.0170e-05, 4.6939e-05, 1.1247e-04, ..., 2.2936e-04, + 3.6299e-05, 3.4086e-07], + [ 4.3772e-06, 1.7061e-03, 2.3087e-02, ..., 2.7657e-04, + 5.7297e-03, 2.3749e-08]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0138, -0.0020, -0.0142, 0.0105, -0.0152, -0.0211, 0.0108, -0.0016, + 0.0122, -0.0120], device='cuda:0'), grad: tensor([ 0.0108, -0.0177, 0.0143, -0.0206, -0.0342, 0.0097, 0.0162, -0.0217, + 0.0091, 0.0340], device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 216.93, cls_loss 0.4632 cls_loss_mapping 0.0009 cls_loss_causal 0.4290 re_mapping 0.0037 re_causal 0.0114 /// teacc 98.96 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.0765, -0.1089, -0.1549, ..., 0.1268, -0.0671, 0.0377], + [-0.0949, 0.1129, -0.0590, ..., 0.0516, -0.0634, -0.0975], + [-0.0387, 0.0888, -0.1340, ..., -0.0205, 0.0419, 0.1248], + ..., + [ 0.0025, 0.0745, 0.0012, ..., -0.0249, 0.0337, -0.0850], + [ 0.1027, -0.1208, 0.0467, ..., 0.0207, 0.0101, -0.0373], + [-0.0130, -0.0889, 0.0742, ..., -0.0987, 0.0339, -0.0301]], + device='cuda:0'), grad: tensor([[ 1.4746e-04, 1.8820e-05, 4.9362e-03, ..., 2.1267e-03, + 5.0163e-03, 1.1116e-04], + [-1.4519e-02, -4.7951e-03, -3.4142e-03, ..., -4.5128e-03, + 9.2506e-05, -7.8154e-04], + [ 7.8261e-05, -1.8813e-07, 4.6194e-05, ..., 2.6059e-04, + -1.9276e-04, -1.3089e-04], + ..., + [ 9.0897e-05, 4.8801e-06, 2.0194e-04, ..., 1.2636e-04, + 1.0216e-04, 1.1998e-04], + [ 9.7513e-05, 2.2769e-05, 6.1893e-04, ..., 3.2616e-04, + 4.9210e-04, 5.9903e-05], + [ 7.1824e-05, 4.6045e-06, 6.0616e-03, ..., -4.8943e-03, + 7.7477e-03, 9.7096e-05]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0138, -0.0019, -0.0143, 0.0106, -0.0151, -0.0209, 0.0106, -0.0017, + 0.0120, -0.0120], device='cuda:0'), grad: tensor([ 0.0218, -0.0462, -0.0239, 0.0038, 0.0131, -0.0217, 0.0531, 0.0086, + 0.0072, -0.0157], device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 217.00, cls_loss 0.4335 cls_loss_mapping 0.0008 cls_loss_causal 0.4046 re_mapping 0.0038 re_causal 0.0116 /// teacc 98.97 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.0766, -0.1090, -0.1547, ..., 0.1267, -0.0670, 0.0375], + [-0.0950, 0.1129, -0.0591, ..., 0.0517, -0.0634, -0.0975], + [-0.0387, 0.0888, -0.1341, ..., -0.0203, 0.0418, 0.1249], + ..., + [ 0.0025, 0.0746, 0.0012, ..., -0.0250, 0.0338, -0.0850], + [ 0.1027, -0.1208, 0.0467, ..., 0.0207, 0.0100, -0.0372], + [-0.0130, -0.0889, 0.0741, ..., -0.0987, 0.0338, -0.0301]], + device='cuda:0'), grad: tensor([[ 7.8559e-05, 7.1824e-05, 6.2323e-04, ..., 7.9751e-05, + 8.6689e-04, 1.1988e-05], + [ 4.1902e-05, -1.4555e-04, 2.0194e-04, ..., -1.6975e-04, + 3.8832e-05, 1.4484e-05], + [ 6.3181e-04, 4.3839e-05, 1.2171e-04, ..., 5.8562e-05, + 3.8576e-04, 1.3661e-04], + ..., + [ 3.0115e-05, 1.7419e-05, 7.7295e-04, ..., 9.2983e-06, + 4.3124e-05, 6.3106e-06], + [-3.1948e-03, 3.5260e-06, 3.3426e-04, ..., 2.3949e-04, + -1.1396e-03, -6.9809e-04], + [ 4.4465e-05, 6.7241e-06, -3.4070e-04, ..., 6.8955e-06, + 3.6550e-04, 9.6038e-06]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0136, -0.0018, -0.0141, 0.0105, -0.0152, -0.0210, 0.0106, -0.0015, + 0.0121, -0.0120], device='cuda:0'), grad: tensor([-0.0165, 0.0158, 0.0142, 0.0208, -0.0179, -0.0195, 0.0040, 0.0159, + 0.0011, -0.0179], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 470---------------------------------------------------- +epoch 470, time 217.50, cls_loss 0.4385 cls_loss_mapping 0.0007 cls_loss_causal 0.4112 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.04 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.0766, -0.1090, -0.1547, ..., 0.1267, -0.0670, 0.0375], + [-0.0949, 0.1129, -0.0589, ..., 0.0516, -0.0634, -0.0975], + [-0.0387, 0.0889, -0.1341, ..., -0.0202, 0.0418, 0.1250], + ..., + [ 0.0026, 0.0746, 0.0013, ..., -0.0249, 0.0337, -0.0849], + [ 0.1028, -0.1209, 0.0466, ..., 0.0207, 0.0099, -0.0373], + [-0.0131, -0.0890, 0.0742, ..., -0.0989, 0.0340, -0.0301]], + device='cuda:0'), grad: tensor([[ 6.6340e-05, 3.0428e-05, -4.3983e-03, ..., -2.3937e-03, + -1.8549e-03, 2.5421e-05], + [-2.2483e-04, -8.4639e-04, 4.6825e-04, ..., -5.3310e-04, + 2.3353e-04, 1.4149e-05], + [ 2.0218e-04, 2.8467e-04, 2.6679e-04, ..., 2.4533e-04, + 2.6369e-04, -1.7250e-04], + ..., + [ 1.8716e-04, -1.5574e-03, 1.1463e-03, ..., 6.6185e-04, + -1.3924e-03, -1.2531e-03], + [ 9.4509e-04, 9.7871e-05, 1.8444e-03, ..., 6.2323e-04, + 1.9464e-03, 6.6936e-05], + [-2.0428e-03, -7.0930e-05, 2.5291e-03, ..., -3.4928e-04, + -1.7033e-03, -1.1170e-04]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0136, -0.0018, -0.0141, 0.0105, -0.0152, -0.0210, 0.0105, -0.0015, + 0.0120, -0.0119], device='cuda:0'), grad: tensor([-0.0205, 0.0100, -0.0185, 0.0162, 0.0138, 0.0125, -0.0145, 0.0064, + -0.0073, 0.0020], device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 217.18, cls_loss 0.4284 cls_loss_mapping 0.0007 cls_loss_causal 0.4002 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.00 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.0765, -0.1091, -0.1547, ..., 0.1267, -0.0669, 0.0374], + [-0.0951, 0.1129, -0.0589, ..., 0.0515, -0.0633, -0.0976], + [-0.0386, 0.0889, -0.1341, ..., -0.0203, 0.0419, 0.1252], + ..., + [ 0.0025, 0.0746, 0.0011, ..., -0.0248, 0.0337, -0.0851], + [ 0.1029, -0.1211, 0.0465, ..., 0.0208, 0.0099, -0.0374], + [-0.0132, -0.0890, 0.0742, ..., -0.0990, 0.0340, -0.0301]], + device='cuda:0'), grad: tensor([[ 2.3773e-02, 1.4555e-04, 1.9813e-04, ..., -3.0308e-03, + -4.4274e-04, 1.4544e-04], + [ 1.1368e-03, -1.3542e-03, -1.6289e-03, ..., -2.1534e-03, + 3.0488e-05, 1.3614e-04], + [-3.5954e-03, -4.9639e-04, 1.4353e-04, ..., -1.2770e-03, + 1.3280e-04, 2.0981e-04], + ..., + [ 2.7885e-03, 2.9159e-04, 3.7813e-04, ..., -2.7671e-05, + -4.3064e-05, -7.5436e-04], + [ 5.0201e-03, 2.5749e-04, 5.7793e-04, ..., 2.1896e-03, + 6.1893e-04, 2.9230e-04], + [ 2.5291e-03, 1.8716e-04, -6.2752e-04, ..., 5.7459e-04, + -3.2640e-04, -1.9372e-04]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0136, -0.0019, -0.0140, 0.0106, -0.0151, -0.0210, 0.0104, -0.0017, + 0.0120, -0.0119], device='cuda:0'), grad: tensor([ 0.0014, -0.0216, 0.0002, -0.0055, -0.0055, -0.0150, 0.0046, -0.0073, + 0.0362, 0.0126], device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 216.44, cls_loss 0.4280 cls_loss_mapping 0.0007 cls_loss_causal 0.3978 re_mapping 0.0038 re_causal 0.0116 /// teacc 98.96 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.0764, -0.1091, -0.1546, ..., 0.1267, -0.0669, 0.0373], + [-0.0952, 0.1127, -0.0588, ..., 0.0515, -0.0631, -0.0976], + [-0.0385, 0.0890, -0.1340, ..., -0.0203, 0.0420, 0.1252], + ..., + [ 0.0027, 0.0747, 0.0013, ..., -0.0249, 0.0337, -0.0851], + [ 0.1029, -0.1211, 0.0467, ..., 0.0209, 0.0101, -0.0372], + [-0.0132, -0.0890, 0.0742, ..., -0.0991, 0.0338, -0.0301]], + device='cuda:0'), grad: tensor([[-9.3269e-04, 2.7609e-04, 3.2043e-04, ..., 3.0537e-03, + 3.3760e-03, 4.3368e-04], + [ 8.8043e-03, 3.0947e-04, 3.1304e-04, ..., 1.1396e-03, + 2.2173e-04, 2.4939e-04], + [ 5.7936e-04, -7.7581e-04, 2.0814e-04, ..., 1.7433e-03, + 1.8738e-02, 5.8222e-04], + ..., + [ 1.7273e-04, 2.5010e-04, 4.0817e-03, ..., 2.0885e-03, + 4.3602e-03, 3.7837e-04], + [ 7.1144e-04, 9.9361e-05, -7.3433e-03, ..., -7.4692e-03, + -6.8817e-03, -2.7523e-03], + [-4.4785e-03, 1.4031e-04, -3.1357e-03, ..., 3.4485e-03, + -1.2865e-03, 4.4656e-04]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0137, -0.0019, -0.0140, 0.0106, -0.0151, -0.0211, 0.0105, -0.0016, + 0.0120, -0.0120], device='cuda:0'), grad: tensor([ 0.0195, 0.0222, -0.0140, 0.0193, -0.0051, -0.0072, 0.0038, 0.0036, + -0.0088, -0.0334], device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 216.48, cls_loss 0.4671 cls_loss_mapping 0.0008 cls_loss_causal 0.4395 re_mapping 0.0037 re_causal 0.0119 /// teacc 98.94 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.0765, -0.1090, -0.1545, ..., 0.1266, -0.0670, 0.0373], + [-0.0951, 0.1127, -0.0588, ..., 0.0516, -0.0633, -0.0974], + [-0.0386, 0.0889, -0.1340, ..., -0.0202, 0.0421, 0.1253], + ..., + [ 0.0027, 0.0748, 0.0012, ..., -0.0249, 0.0336, -0.0852], + [ 0.1030, -0.1210, 0.0467, ..., 0.0209, 0.0101, -0.0374], + [-0.0131, -0.0891, 0.0742, ..., -0.0990, 0.0338, -0.0301]], + device='cuda:0'), grad: tensor([[ 5.1737e-05, 1.1349e-04, 5.5218e-04, ..., -2.9621e-03, + -3.2623e-02, 4.1336e-05], + [ 2.2256e-04, -1.3626e-04, 2.7084e-04, ..., -4.9858e-03, + 8.0824e-05, 3.3081e-05], + [-2.4109e-02, -1.1272e-03, -6.8626e-03, ..., -9.9087e-04, + 9.1255e-05, -2.2614e-04], + ..., + [ 3.5610e-03, 3.0136e-04, 9.3555e-04, ..., 1.8005e-03, + -9.3937e-05, 1.0389e-04], + [ 2.8920e-04, 2.0719e-04, 1.6463e-04, ..., 2.9945e-03, + 1.3924e-02, 5.2869e-05], + [ 1.1963e-02, -6.9714e-04, 3.9768e-04, ..., 7.5960e-04, + 7.7667e-03, 2.4185e-05]], device='cuda:0') +Epoch 475, bias, value: tensor([ 0.0135, -0.0017, -0.0140, 0.0105, -0.0150, -0.0210, 0.0104, -0.0016, + 0.0121, -0.0121], device='cuda:0'), grad: tensor([-0.0097, -0.0193, -0.0199, -0.0517, 0.0389, 0.0355, -0.0242, 0.0195, + 0.0239, 0.0068], device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 216.67, cls_loss 0.4345 cls_loss_mapping 0.0010 cls_loss_causal 0.4165 re_mapping 0.0037 re_causal 0.0110 /// teacc 98.99 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.0765, -0.1090, -0.1546, ..., 0.1266, -0.0669, 0.0374], + [-0.0952, 0.1127, -0.0589, ..., 0.0516, -0.0634, -0.0973], + [-0.0384, 0.0890, -0.1340, ..., -0.0201, 0.0422, 0.1252], + ..., + [ 0.0027, 0.0748, 0.0011, ..., -0.0250, 0.0333, -0.0853], + [ 0.1030, -0.1212, 0.0468, ..., 0.0210, 0.0101, -0.0374], + [-0.0130, -0.0891, 0.0743, ..., -0.0991, 0.0339, -0.0302]], + device='cuda:0'), grad: tensor([[ 1.2159e-05, 1.9401e-05, 2.4235e-04, ..., -7.9727e-04, + 1.4806e-04, -3.5930e-04], + [-1.5972e-06, -3.2969e-07, 5.1355e-04, ..., 1.0414e-03, + 2.0766e-04, 4.2349e-05], + [ 9.4795e-04, 3.2711e-03, 5.9986e-04, ..., 7.9155e-04, + 3.2282e-04, 1.3483e-04], + ..., + [-9.3269e-04, -3.2635e-03, -3.0594e-03, ..., -4.8370e-03, + -1.5020e-03, 3.1769e-05], + [ 2.1815e-05, 3.3259e-05, 1.8396e-03, ..., 1.2226e-03, + 2.6751e-04, 4.4966e-04], + [ 7.4245e-06, 1.1072e-05, -3.1338e-03, ..., 4.1938e-04, + -1.7052e-03, 5.7399e-05]], device='cuda:0') +Epoch 476, bias, value: tensor([ 0.0135, -0.0017, -0.0140, 0.0106, -0.0149, -0.0209, 0.0104, -0.0017, + 0.0120, -0.0121], device='cuda:0'), grad: tensor([ 0.0015, -0.0173, 0.0163, 0.0081, 0.0248, -0.0142, 0.0232, -0.0477, + 0.0180, -0.0128], device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 216.97, cls_loss 0.4248 cls_loss_mapping 0.0007 cls_loss_causal 0.3986 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.01 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.0766, -0.1090, -0.1545, ..., 0.1267, -0.0669, 0.0375], + [-0.0951, 0.1128, -0.0589, ..., 0.0515, -0.0633, -0.0972], + [-0.0384, 0.0889, -0.1340, ..., -0.0201, 0.0421, 0.1251], + ..., + [ 0.0029, 0.0750, 0.0011, ..., -0.0250, 0.0333, -0.0853], + [ 0.1030, -0.1214, 0.0468, ..., 0.0209, 0.0101, -0.0374], + [-0.0130, -0.0893, 0.0743, ..., -0.0992, 0.0338, -0.0303]], + device='cuda:0'), grad: tensor([[ 1.5795e-04, 1.4222e-04, -1.4544e-05, ..., -1.9646e-04, + -1.7655e-04, -5.5969e-05], + [ 3.9029e-04, -4.6682e-04, 7.1466e-05, ..., 2.5511e-04, + 2.6798e-04, 8.6129e-05], + [ 8.2493e-04, 2.8229e-03, 8.1205e-04, ..., 3.6430e-03, + 4.2343e-03, 1.3580e-03], + ..., + [ 2.2495e-04, 8.3160e-04, 3.0756e-04, ..., 6.0606e-04, + 6.2561e-04, 1.8752e-04], + [ 6.0081e-04, 7.2670e-04, 2.6155e-04, ..., 1.0576e-03, + 1.1787e-03, 3.6502e-04], + [-1.6113e-02, -4.3373e-03, -1.4830e-03, ..., -5.1842e-03, + -6.4087e-03, -2.0790e-03]], device='cuda:0') +Epoch 477, bias, value: tensor([ 0.0134, -0.0016, -0.0140, 0.0106, -0.0149, -0.0209, 0.0104, -0.0017, + 0.0119, -0.0120], device='cuda:0'), grad: tensor([ 0.0064, 0.0081, 0.0235, 0.0324, 0.0069, 0.0045, -0.0246, 0.0108, + -0.0204, -0.0476], device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 216.64, cls_loss 0.4620 cls_loss_mapping 0.0008 cls_loss_causal 0.4410 re_mapping 0.0037 re_causal 0.0118 /// teacc 98.95 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.0766, -0.1091, -0.1546, ..., 0.1268, -0.0669, 0.0374], + [-0.0952, 0.1128, -0.0588, ..., 0.0514, -0.0633, -0.0973], + [-0.0384, 0.0889, -0.1341, ..., -0.0200, 0.0421, 0.1249], + ..., + [ 0.0028, 0.0750, 0.0012, ..., -0.0251, 0.0333, -0.0852], + [ 0.1030, -0.1214, 0.0468, ..., 0.0208, 0.0101, -0.0374], + [-0.0131, -0.0894, 0.0743, ..., -0.0992, 0.0338, -0.0304]], + device='cuda:0'), grad: tensor([[ 3.2401e-04, 2.0468e-04, 2.2197e-04, ..., -9.5825e-03, + -2.9349e-04, 2.1124e-04], + [-5.2166e-04, -2.6965e-04, 6.6876e-05, ..., 1.8513e-04, + 3.1441e-05, 5.8949e-05], + [ 5.3912e-05, -1.2565e-04, 1.6959e-06, ..., 1.4370e-06, + -3.8934e-04, -4.6921e-04], + ..., + [ 1.0008e-04, 6.5744e-05, 2.7753e-06, ..., 8.7738e-03, + 2.5711e-03, -8.9550e-04], + [ 5.5939e-05, -6.2287e-05, 2.9922e-04, ..., 1.6534e-04, + 3.7479e-03, 9.8348e-05], + [ 1.6853e-05, 1.1481e-05, -4.8637e-04, ..., 2.0099e-04, + -6.1226e-03, 9.6202e-05]], device='cuda:0') +Epoch 478, bias, value: tensor([ 0.0133, -0.0016, -0.0140, 0.0107, -0.0149, -0.0209, 0.0103, -0.0015, + 0.0118, -0.0121], device='cuda:0'), grad: tensor([ 0.0015, 0.0131, 0.0094, 0.0129, 0.0114, -0.0483, 0.0133, -0.0352, + 0.0115, 0.0105], device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 216.88, cls_loss 0.4565 cls_loss_mapping 0.0008 cls_loss_causal 0.4330 re_mapping 0.0038 re_causal 0.0117 /// teacc 98.97 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.0766, -0.1091, -0.1547, ..., 0.1269, -0.0669, 0.0373], + [-0.0952, 0.1128, -0.0588, ..., 0.0514, -0.0633, -0.0973], + [-0.0383, 0.0889, -0.1341, ..., -0.0198, 0.0422, 0.1250], + ..., + [ 0.0029, 0.0748, 0.0011, ..., -0.0252, 0.0333, -0.0853], + [ 0.1029, -0.1214, 0.0467, ..., 0.0207, 0.0101, -0.0374], + [-0.0133, -0.0892, 0.0741, ..., -0.0995, 0.0336, -0.0304]], + device='cuda:0'), grad: tensor([[ 7.7057e-04, 2.1038e-03, 1.1182e-04, ..., 5.3062e-03, + 1.2636e-04, 2.9489e-05], + [ 1.5535e-03, 1.6632e-03, 1.1218e-04, ..., -4.5929e-03, + 1.0270e-04, -1.6242e-05], + [ 2.4128e-03, -4.7531e-03, 1.0180e-04, ..., -9.0942e-03, + 3.2711e-04, 1.1139e-03], + ..., + [ 3.3684e-03, 5.2786e-04, 3.6263e-04, ..., 1.0824e-03, + 1.4389e-04, 4.5151e-06], + [-1.4015e-02, -1.6022e-03, 1.5903e-04, ..., 2.7695e-03, + -8.8513e-05, 8.1956e-05], + [ 3.1204e-03, 4.3416e-04, -6.1655e-04, ..., 7.5531e-04, + -2.4581e-04, 4.1947e-06]], device='cuda:0') +Epoch 479, bias, value: tensor([ 0.0133, -0.0017, -0.0139, 0.0107, -0.0147, -0.0210, 0.0104, -0.0016, + 0.0117, -0.0121], device='cuda:0'), grad: tensor([-0.0019, -0.0310, -0.0187, -0.0163, 0.0042, -0.0172, 0.0294, 0.0269, + 0.0029, 0.0216], device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 217.01, cls_loss 0.4187 cls_loss_mapping 0.0007 cls_loss_causal 0.3959 re_mapping 0.0037 re_causal 0.0112 /// teacc 98.97 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.0767, -0.1093, -0.1548, ..., 0.1268, -0.0670, 0.0372], + [-0.0951, 0.1127, -0.0587, ..., 0.0514, -0.0630, -0.0971], + [-0.0382, 0.0890, -0.1341, ..., -0.0197, 0.0421, 0.1251], + ..., + [ 0.0027, 0.0748, 0.0011, ..., -0.0251, 0.0333, -0.0853], + [ 0.1031, -0.1213, 0.0468, ..., 0.0207, 0.0101, -0.0374], + [-0.0133, -0.0892, 0.0740, ..., -0.0996, 0.0335, -0.0305]], + device='cuda:0'), grad: tensor([[ 7.3314e-05, 1.3375e-04, 7.1488e-06, ..., -3.9673e-04, + -3.3450e-04, -3.7527e-04], + [ 2.0313e-03, 1.1320e-03, 7.6151e-04, ..., 9.3985e-04, + 1.5247e-04, 2.6509e-05], + [ 9.4032e-04, -7.0763e-04, 8.0776e-04, ..., 4.8566e-04, + 4.5252e-04, 1.1396e-04], + ..., + [-9.4986e-03, -3.8853e-03, -4.0054e-03, ..., -4.4594e-03, + -1.0157e-03, 1.3256e-04], + [ 1.0757e-03, 7.6866e-04, 1.2197e-03, ..., 1.0786e-03, + 2.5362e-05, 1.1754e-04], + [ 9.6416e-04, 4.7255e-04, -2.2907e-03, ..., -8.4829e-04, + -1.7233e-03, -1.0214e-03]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0132, -0.0017, -0.0137, 0.0106, -0.0147, -0.0210, 0.0103, -0.0015, + 0.0116, -0.0120], device='cuda:0'), grad: tensor([-0.0238, 0.0259, -0.0387, 0.0192, 0.0307, 0.0133, -0.0154, -0.0309, + 0.0180, 0.0016], device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 216.81, cls_loss 0.4221 cls_loss_mapping 0.0007 cls_loss_causal 0.3848 re_mapping 0.0037 re_causal 0.0113 /// teacc 98.99 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.0767, -0.1094, -0.1548, ..., 0.1267, -0.0670, 0.0371], + [-0.0950, 0.1129, -0.0585, ..., 0.0515, -0.0628, -0.0972], + [-0.0382, 0.0890, -0.1342, ..., -0.0198, 0.0421, 0.1251], + ..., + [ 0.0026, 0.0747, 0.0012, ..., -0.0251, 0.0333, -0.0853], + [ 0.1031, -0.1213, 0.0469, ..., 0.0209, 0.0102, -0.0373], + [-0.0134, -0.0890, 0.0740, ..., -0.0994, 0.0334, -0.0306]], + device='cuda:0'), grad: tensor([[ 9.9182e-04, 3.1137e-04, 1.2150e-03, ..., 2.3956e-03, + 1.0366e-03, 8.3017e-04], + [-5.8556e-03, -6.7253e-03, -1.4267e-03, ..., -1.9503e-03, + 1.8156e-04, 2.3234e-04], + [ 3.6106e-03, 2.5997e-03, 5.9271e-04, ..., 1.8797e-03, + 5.3465e-05, -1.4973e-03], + ..., + [ 7.4768e-04, 1.1806e-03, 4.7135e-04, ..., 5.7268e-04, + -2.3613e-03, 4.3821e-04], + [ 9.6178e-04, 7.6914e-04, 1.3447e-03, ..., 1.7138e-03, + 9.3985e-04, 1.1721e-03], + [-2.6207e-03, 2.4939e-04, -5.3711e-03, ..., 4.5091e-05, + -3.5858e-03, -1.2569e-03]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0132, -0.0018, -0.0137, 0.0107, -0.0148, -0.0211, 0.0104, -0.0013, + 0.0117, -0.0121], device='cuda:0'), grad: tensor([ 0.0280, -0.0081, 0.0223, -0.0287, -0.0003, -0.0323, 0.0287, 0.0203, + 0.0225, -0.0525], device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 216.92, cls_loss 0.4190 cls_loss_mapping 0.0007 cls_loss_causal 0.3968 re_mapping 0.0037 re_causal 0.0112 /// teacc 99.00 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.0767, -0.1095, -0.1548, ..., 0.1267, -0.0670, 0.0372], + [-0.0950, 0.1129, -0.0583, ..., 0.0515, -0.0627, -0.0971], + [-0.0382, 0.0890, -0.1344, ..., -0.0198, 0.0422, 0.1252], + ..., + [ 0.0026, 0.0747, 0.0011, ..., -0.0252, 0.0334, -0.0853], + [ 0.1031, -0.1215, 0.0468, ..., 0.0209, 0.0102, -0.0375], + [-0.0135, -0.0890, 0.0742, ..., -0.0993, 0.0334, -0.0307]], + device='cuda:0'), grad: tensor([[-8.3113e-04, 1.1891e-04, -1.5297e-03, ..., -1.0757e-03, + -2.5425e-03, -1.3494e-03], + [-2.1529e-04, -6.2370e-04, -1.7858e-04, ..., -9.7942e-04, + 5.9139e-07, 3.5344e-07], + [-4.4022e-03, -3.1166e-03, 3.5614e-05, ..., -1.1368e-03, + -9.2447e-05, -9.4056e-05], + ..., + [ 1.9131e-03, 7.6056e-04, -1.2093e-03, ..., 1.1330e-03, + 1.0198e-04, 7.4089e-05], + [-1.3504e-03, 4.2939e-04, 2.2244e-04, ..., 1.7190e-04, + 2.3052e-05, 2.1741e-05], + [ 1.2627e-03, 1.2465e-03, -2.1482e-04, ..., 2.5201e-04, + -4.4703e-04, 1.9632e-06]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0133, -0.0017, -0.0139, 0.0108, -0.0147, -0.0210, 0.0103, -0.0013, + 0.0116, -0.0122], device='cuda:0'), grad: tensor([-0.0015, 0.0223, -0.0486, 0.0018, 0.0208, 0.0162, -0.0064, -0.0230, + -0.0155, 0.0338], device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 217.17, cls_loss 0.4294 cls_loss_mapping 0.0007 cls_loss_causal 0.4053 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.01 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.0766, -0.1095, -0.1547, ..., 0.1266, -0.0669, 0.0373], + [-0.0950, 0.1128, -0.0583, ..., 0.0514, -0.0626, -0.0971], + [-0.0381, 0.0889, -0.1343, ..., -0.0197, 0.0423, 0.1252], + ..., + [ 0.0025, 0.0748, 0.0012, ..., -0.0251, 0.0335, -0.0853], + [ 0.1031, -0.1214, 0.0469, ..., 0.0210, 0.0101, -0.0377], + [-0.0136, -0.0890, 0.0740, ..., -0.0995, 0.0333, -0.0306]], + device='cuda:0'), grad: tensor([[ 3.5018e-07, 1.4365e-04, -1.9610e-05, ..., 1.3125e-04, + 3.2115e-04, -2.2519e-04], + [-1.3761e-05, 3.8552e-04, 2.2209e-04, ..., 2.2392e-03, + 3.7766e-03, 7.0751e-05], + [ 3.2689e-06, 1.8251e-04, -1.1414e-04, ..., -4.5433e-03, + -9.4147e-03, 1.1986e-04], + ..., + [ 1.7546e-06, -2.7180e-04, -6.5267e-06, ..., 5.5742e-04, + 9.8801e-04, 6.1393e-05], + [ 4.6678e-06, -1.2999e-03, 1.1700e-04, ..., -1.1950e-03, + 6.3896e-04, -3.3545e-04], + [ 3.3285e-06, 2.3293e-04, 3.3498e-05, ..., 7.2002e-04, + 7.5531e-04, 5.5641e-05]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0133, -0.0017, -0.0138, 0.0107, -0.0148, -0.0211, 0.0104, -0.0013, + 0.0117, -0.0123], device='cuda:0'), grad: tensor([ 0.0078, 0.0199, -0.0152, -0.0190, 0.0173, 0.0074, -0.0182, 0.0101, + -0.0210, 0.0110], device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 217.19, cls_loss 0.4416 cls_loss_mapping 0.0007 cls_loss_causal 0.4131 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.01 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.0766, -0.1096, -0.1547, ..., 0.1265, -0.0669, 0.0374], + [-0.0951, 0.1127, -0.0585, ..., 0.0513, -0.0627, -0.0971], + [-0.0380, 0.0890, -0.1342, ..., -0.0196, 0.0423, 0.1251], + ..., + [ 0.0025, 0.0748, 0.0012, ..., -0.0251, 0.0334, -0.0854], + [ 0.1031, -0.1214, 0.0470, ..., 0.0211, 0.0102, -0.0377], + [-0.0135, -0.0889, 0.0741, ..., -0.0994, 0.0333, -0.0306]], + device='cuda:0'), grad: tensor([[ 6.0892e-04, 3.0063e-06, 8.2111e-04, ..., 2.7561e-04, + 1.1864e-03, -6.9737e-06], + [-2.7256e-03, 3.1137e-04, -8.2493e-05, ..., 4.6778e-04, + -2.7142e-03, -6.5565e-04], + [ 2.6181e-05, -7.4673e-04, -3.8035e-06, ..., -3.5973e-03, + 6.2513e-04, -1.7843e-03], + ..., + [-8.9169e-04, 4.9057e-03, 3.4275e-03, ..., 5.6314e-04, + 2.3003e-03, 6.3992e-04], + [ 1.6394e-03, 2.7537e-05, 8.3065e-04, ..., 5.4264e-04, + -2.5711e-03, 2.7847e-04], + [-2.1057e-03, -4.6120e-03, 4.9686e-04, ..., -1.2624e-04, + 9.2125e-04, 2.5177e-04]], device='cuda:0') +Epoch 484, bias, value: tensor([ 0.0134, -0.0018, -0.0137, 0.0106, -0.0148, -0.0210, 0.0103, -0.0013, + 0.0117, -0.0122], device='cuda:0'), grad: tensor([-0.0070, -0.0099, -0.0136, 0.0435, -0.0093, -0.0143, -0.0135, 0.0185, + -0.0023, 0.0079], device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 217.06, cls_loss 0.4524 cls_loss_mapping 0.0007 cls_loss_causal 0.4251 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.00 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.0767, -0.1094, -0.1546, ..., 0.1265, -0.0669, 0.0372], + [-0.0950, 0.1128, -0.0584, ..., 0.0513, -0.0626, -0.0970], + [-0.0380, 0.0889, -0.1342, ..., -0.0198, 0.0422, 0.1251], + ..., + [ 0.0024, 0.0749, 0.0011, ..., -0.0251, 0.0334, -0.0855], + [ 0.1032, -0.1215, 0.0469, ..., 0.0211, 0.0102, -0.0378], + [-0.0136, -0.0890, 0.0741, ..., -0.0994, 0.0333, -0.0305]], + device='cuda:0'), grad: tensor([[ 3.4022e-04, 8.4117e-06, 2.5317e-05, ..., 5.5265e-04, + 2.6330e-05, 2.7514e-04], + [ 2.4033e-04, -9.2149e-05, -2.1048e-07, ..., 3.8981e-04, + 6.7651e-06, 1.9228e-04], + [ 1.3018e-04, -8.9034e-06, 8.4415e-06, ..., -1.5163e-03, + 1.5214e-05, -1.7655e-04], + ..., + [ 1.3638e-04, -9.1195e-05, 6.2250e-06, ..., 2.9278e-04, + 6.8434e-06, 1.2302e-04], + [ 1.7178e-04, 3.3975e-05, -4.5037e-04, ..., 2.4796e-04, + -2.0361e-04, 4.6879e-05], + [ 1.3220e-04, 4.9055e-05, 2.1374e-04, ..., 3.9458e-04, + 1.4913e-04, 1.8513e-04]], device='cuda:0') +Epoch 485, bias, value: tensor([ 0.0134, -0.0017, -0.0137, 0.0106, -0.0148, -0.0210, 0.0103, -0.0013, + 0.0116, -0.0122], device='cuda:0'), grad: tensor([ 0.0117, 0.0103, -0.0234, -0.0275, 0.0075, 0.0121, 0.0152, 0.0081, + 0.0080, -0.0220], device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 217.09, cls_loss 0.4348 cls_loss_mapping 0.0007 cls_loss_causal 0.4069 re_mapping 0.0036 re_causal 0.0116 /// teacc 98.99 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.0767, -0.1094, -0.1546, ..., 0.1265, -0.0669, 0.0374], + [-0.0950, 0.1128, -0.0583, ..., 0.0513, -0.0625, -0.0972], + [-0.0379, 0.0889, -0.1343, ..., -0.0199, 0.0421, 0.1251], + ..., + [ 0.0024, 0.0749, 0.0010, ..., -0.0251, 0.0335, -0.0856], + [ 0.1031, -0.1217, 0.0470, ..., 0.0210, 0.0103, -0.0378], + [-0.0136, -0.0889, 0.0741, ..., -0.0993, 0.0332, -0.0307]], + device='cuda:0'), grad: tensor([[-2.4433e-03, 2.9874e-04, -6.6683e-06, ..., -3.0160e-04, + 1.9991e-04, 3.2949e-04], + [ 4.8113e-04, 2.0182e-04, -8.4972e-04, ..., 8.9884e-04, + 1.7035e-04, 3.1447e-04], + [ 4.1270e-04, 1.4334e-03, 1.0777e-04, ..., 2.4872e-03, + 9.6416e-04, 2.0561e-03], + ..., + [ 4.1962e-04, 1.6365e-03, -5.1346e-03, ..., 1.5430e-03, + -6.4087e-03, 2.9945e-03], + [ 2.7704e-04, 2.6965e-04, 5.2214e-05, ..., 6.6805e-04, + 2.0635e-04, 4.0674e-04], + [ 4.2748e-04, 2.3592e-04, 3.7346e-03, ..., -1.3037e-03, + 5.0240e-03, 2.9135e-04]], device='cuda:0') +Epoch 486, bias, value: tensor([ 0.0136, -0.0018, -0.0137, 0.0106, -0.0148, -0.0210, 0.0102, -0.0014, + 0.0115, -0.0121], device='cuda:0'), grad: tensor([-0.0237, -0.0153, 0.0197, 0.0249, 0.0318, -0.0530, 0.0062, 0.0046, + 0.0080, -0.0032], device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 216.70, cls_loss 0.4412 cls_loss_mapping 0.0007 cls_loss_causal 0.4085 re_mapping 0.0037 re_causal 0.0114 /// teacc 98.94 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.0767, -0.1095, -0.1546, ..., 0.1265, -0.0669, 0.0373], + [-0.0950, 0.1129, -0.0583, ..., 0.0513, -0.0626, -0.0970], + [-0.0379, 0.0889, -0.1342, ..., -0.0198, 0.0423, 0.1252], + ..., + [ 0.0024, 0.0748, 0.0011, ..., -0.0251, 0.0334, -0.0857], + [ 0.1031, -0.1217, 0.0472, ..., 0.0210, 0.0103, -0.0379], + [-0.0136, -0.0889, 0.0740, ..., -0.0992, 0.0331, -0.0308]], + device='cuda:0'), grad: tensor([[-8.4257e-04, 9.7677e-06, -5.0783e-04, ..., 3.9911e-04, + 5.5790e-05, -1.0433e-03], + [ 2.5797e-04, 3.3021e-05, 1.6212e-04, ..., 5.9271e-04, + 1.0741e-04, 1.0744e-05], + [-1.3866e-03, -4.0007e-04, -1.5230e-03, ..., -2.0599e-03, + -1.0290e-03, -2.8126e-06], + ..., + [ 3.1686e-04, -2.3091e-04, 2.5225e-04, ..., -6.9904e-04, + 2.9683e-04, 1.6183e-05], + [ 6.0558e-04, 1.0538e-04, 4.0746e-04, ..., 8.7214e-04, + 2.7442e-04, 1.4615e-04], + [-1.1492e-03, 2.6536e-04, -1.6749e-04, ..., 5.1165e-04, + -2.4581e-04, 8.9049e-05]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0136, -0.0019, -0.0136, 0.0107, -0.0149, -0.0209, 0.0102, -0.0014, + 0.0115, -0.0122], device='cuda:0'), grad: tensor([-0.0069, 0.0127, -0.0128, -0.0077, -0.0157, 0.0185, 0.0229, -0.0154, + 0.0201, -0.0155], device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 217.20, cls_loss 0.4590 cls_loss_mapping 0.0009 cls_loss_causal 0.4301 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.02 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.0767, -0.1095, -0.1547, ..., 0.1264, -0.0669, 0.0374], + [-0.0950, 0.1129, -0.0583, ..., 0.0513, -0.0625, -0.0971], + [-0.0381, 0.0890, -0.1343, ..., -0.0199, 0.0422, 0.1251], + ..., + [ 0.0024, 0.0747, 0.0013, ..., -0.0250, 0.0335, -0.0858], + [ 0.1031, -0.1218, 0.0471, ..., 0.0209, 0.0103, -0.0379], + [-0.0136, -0.0888, 0.0739, ..., -0.0993, 0.0331, -0.0307]], + device='cuda:0'), grad: tensor([[ 5.8889e-05, 1.9693e-04, 5.3549e-04, ..., 1.1129e-03, + 7.1144e-04, 4.2772e-04], + [ 2.2042e-04, 6.9952e-04, 6.0034e-04, ..., 1.2703e-03, + 2.2709e-04, 5.5075e-04], + [ 5.6028e-04, 1.7014e-03, 1.3580e-03, ..., 3.3092e-03, + 6.5088e-04, 1.5240e-03], + ..., + [-3.3784e-04, 9.3412e-04, -6.0539e-03, ..., -2.3575e-03, + -2.5406e-03, -1.1234e-03], + [-1.3762e-03, -3.8738e-03, 3.2864e-03, ..., -5.8022e-03, + 7.8278e-03, -2.8839e-03], + [ 2.0993e-04, 2.9469e-04, -4.5891e-03, ..., 6.5470e-04, + -1.0582e-02, 2.7251e-04]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0136, -0.0020, -0.0138, 0.0107, -0.0148, -0.0208, 0.0102, -0.0013, + 0.0115, -0.0122], device='cuda:0'), grad: tensor([ 0.0102, 0.0086, 0.0205, 0.0096, 0.0147, -0.0139, 0.0118, -0.0541, + -0.0197, 0.0123], device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 216.65, cls_loss 0.4613 cls_loss_mapping 0.0008 cls_loss_causal 0.4362 re_mapping 0.0036 re_causal 0.0119 /// teacc 99.01 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.0770, -0.1094, -0.1548, ..., 0.1266, -0.0671, 0.0374], + [-0.0952, 0.1127, -0.0583, ..., 0.0510, -0.0626, -0.0971], + [-0.0380, 0.0889, -0.1345, ..., -0.0197, 0.0423, 0.1251], + ..., + [ 0.0024, 0.0749, 0.0014, ..., -0.0250, 0.0336, -0.0857], + [ 0.1033, -0.1217, 0.0470, ..., 0.0208, 0.0102, -0.0377], + [-0.0137, -0.0887, 0.0739, ..., -0.0992, 0.0331, -0.0306]], + device='cuda:0'), grad: tensor([[-4.1389e-04, 5.7042e-05, -1.2522e-03, ..., -3.6508e-05, + 2.2456e-05, 8.8632e-05], + [ 2.3723e-04, 1.1611e-04, 1.7309e-04, ..., 1.9467e-04, + 7.6413e-05, 5.1081e-05], + [ 2.6989e-04, 1.4953e-05, 6.9261e-05, ..., 3.8409e-04, + -1.1978e-03, 2.0480e-04], + ..., + [-5.5170e-04, -1.4153e-03, -6.9678e-05, ..., -5.3406e-04, + -4.1437e-04, 7.5996e-05], + [-1.3008e-03, 6.6328e-04, 1.5676e-04, ..., 4.1080e-04, + 1.2779e-04, 3.0446e-04], + [ 3.5977e-04, 2.9969e-04, 2.2018e-04, ..., 3.4142e-04, + 9.7394e-05, 1.0502e-04]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0135, -0.0021, -0.0138, 0.0107, -0.0148, -0.0208, 0.0103, -0.0013, + 0.0117, -0.0122], device='cuda:0'), grad: tensor([-0.0280, 0.0047, 0.0038, -0.0196, 0.0057, 0.0145, 0.0046, -0.0012, + 0.0083, 0.0073], device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 217.03, cls_loss 0.4089 cls_loss_mapping 0.0006 cls_loss_causal 0.3778 re_mapping 0.0037 re_causal 0.0111 /// teacc 98.99 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.0769, -0.1094, -0.1549, ..., 0.1265, -0.0672, 0.0375], + [-0.0953, 0.1127, -0.0584, ..., 0.0510, -0.0626, -0.0971], + [-0.0379, 0.0889, -0.1344, ..., -0.0197, 0.0423, 0.1251], + ..., + [ 0.0024, 0.0748, 0.0015, ..., -0.0250, 0.0337, -0.0858], + [ 0.1032, -0.1218, 0.0471, ..., 0.0209, 0.0103, -0.0376], + [-0.0137, -0.0886, 0.0739, ..., -0.0991, 0.0331, -0.0305]], + device='cuda:0'), grad: tensor([[ 1.0424e-03, 9.7215e-05, 6.2895e-04, ..., 1.6537e-03, + 2.1305e-03, 3.3951e-04], + [ 2.4128e-03, -2.1648e-03, 2.5845e-04, ..., -9.4175e-04, + 9.0694e-04, 1.7679e-04], + [ 1.8158e-03, 2.9826e-04, 7.4339e-04, ..., -6.6605e-03, + -2.0714e-03, -1.9817e-03], + ..., + [-1.9562e-02, 2.3913e-04, 5.3558e-03, ..., 1.3089e-04, + 1.4679e-02, 2.3866e-04], + [ 5.2490e-03, 2.1100e-04, 1.7347e-03, ..., 1.7252e-03, + 4.4823e-03, 4.5204e-04], + [ 2.0161e-03, 7.0095e-05, -1.0422e-02, ..., -3.7422e-03, + -3.2410e-02, -1.1406e-03]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0134, -0.0021, -0.0138, 0.0106, -0.0147, -0.0209, 0.0104, -0.0014, + 0.0117, -0.0120], device='cuda:0'), grad: tensor([ 0.0116, 0.0030, 0.0036, 0.0130, 0.0202, 0.0131, -0.0431, -0.0365, + 0.0270, -0.0119], device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 216.90, cls_loss 0.4760 cls_loss_mapping 0.0007 cls_loss_causal 0.4479 re_mapping 0.0036 re_causal 0.0120 /// teacc 98.96 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.0769, -0.1096, -0.1549, ..., 0.1265, -0.0673, 0.0375], + [-0.0955, 0.1127, -0.0583, ..., 0.0512, -0.0626, -0.0971], + [-0.0379, 0.0889, -0.1345, ..., -0.0198, 0.0424, 0.1251], + ..., + [ 0.0028, 0.0749, 0.0015, ..., -0.0250, 0.0336, -0.0859], + [ 0.1029, -0.1219, 0.0471, ..., 0.0208, 0.0103, -0.0377], + [-0.0136, -0.0886, 0.0739, ..., -0.0993, 0.0331, -0.0305]], + device='cuda:0'), grad: tensor([[ 2.3469e-07, 1.4566e-06, 1.8311e-04, ..., 6.0987e-04, + 1.1176e-07, 7.9162e-09], + [-1.0826e-05, -6.8426e-05, 1.1301e-04, ..., 3.3736e-04, + 4.6566e-07, 4.1910e-09], + [ 1.5004e-06, 8.6352e-06, 6.9886e-06, ..., 2.1204e-05, + 8.0168e-06, 4.5961e-07], + ..., + [ 2.7809e-06, 1.9610e-05, 1.9923e-05, ..., 5.2422e-05, + -2.8871e-07, 3.7253e-09], + [ 1.5162e-06, 1.3433e-05, 1.1019e-05, ..., 3.4660e-05, + 2.9802e-08, 4.0513e-08], + [ 5.4901e-07, 2.6114e-06, 9.4175e-05, ..., 1.4186e-04, + 1.0421e-06, 2.5146e-08]], device='cuda:0') +Epoch 491, bias, value: tensor([ 0.0135, -0.0020, -0.0138, 0.0105, -0.0146, -0.0207, 0.0102, -0.0014, + 0.0116, -0.0121], device='cuda:0'), grad: tensor([ 0.0073, 0.0088, 0.0057, -0.0247, 0.0315, -0.0268, 0.0075, -0.0223, + 0.0066, 0.0064], device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 216.92, cls_loss 0.4402 cls_loss_mapping 0.0007 cls_loss_causal 0.4108 re_mapping 0.0036 re_causal 0.0113 /// teacc 98.97 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.0769, -0.1096, -0.1550, ..., 0.1265, -0.0674, 0.0374], + [-0.0956, 0.1127, -0.0584, ..., 0.0511, -0.0626, -0.0971], + [-0.0377, 0.0890, -0.1347, ..., -0.0199, 0.0423, 0.1251], + ..., + [ 0.0026, 0.0750, 0.0016, ..., -0.0250, 0.0336, -0.0858], + [ 0.1030, -0.1220, 0.0471, ..., 0.0209, 0.0104, -0.0376], + [-0.0135, -0.0887, 0.0738, ..., -0.0993, 0.0330, -0.0306]], + device='cuda:0'), grad: tensor([[ 1.1511e-03, 2.6941e-05, 7.8142e-05, ..., 1.2856e-03, + -1.8206e-03, -2.7132e-04], + [ 1.1997e-03, 5.2166e-04, 3.1829e-04, ..., 4.0245e-04, + 2.4706e-05, 1.1867e-04], + [ 1.0550e-04, 1.5482e-05, 7.2122e-05, ..., 1.6422e-03, + 5.6219e-04, 2.4204e-03], + ..., + [ 7.1108e-05, -3.9029e-04, -1.5783e-03, ..., 3.3689e-04, + -1.2390e-02, 2.4724e-04], + [ 1.5316e-03, 1.7774e-04, 4.7922e-04, ..., 1.0490e-03, + 1.3313e-03, 5.9223e-04], + [-4.0442e-05, 2.3678e-05, 1.5097e-03, ..., 2.7847e-04, + 1.2367e-02, 1.5163e-04]], device='cuda:0') +Epoch 492, bias, value: tensor([ 0.0135, -0.0021, -0.0140, 0.0106, -0.0146, -0.0205, 0.0101, -0.0015, + 0.0117, -0.0120], device='cuda:0'), grad: tensor([ 0.0175, 0.0187, 0.0267, -0.0469, -0.0164, 0.0403, -0.0555, -0.0225, + 0.0174, 0.0207], device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 216.97, cls_loss 0.4593 cls_loss_mapping 0.0007 cls_loss_causal 0.4275 re_mapping 0.0036 re_causal 0.0114 /// teacc 98.99 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.0769, -0.1097, -0.1549, ..., 0.1265, -0.0674, 0.0375], + [-0.0954, 0.1127, -0.0583, ..., 0.0513, -0.0626, -0.0972], + [-0.0377, 0.0890, -0.1347, ..., -0.0198, 0.0423, 0.1251], + ..., + [ 0.0025, 0.0749, 0.0015, ..., -0.0251, 0.0336, -0.0859], + [ 0.1030, -0.1220, 0.0471, ..., 0.0208, 0.0104, -0.0376], + [-0.0135, -0.0887, 0.0737, ..., -0.0994, 0.0329, -0.0307]], + device='cuda:0'), grad: tensor([[ 6.7091e-04, 1.3125e-04, -1.7798e-04, ..., 1.8716e-04, + -1.3316e-04, 6.2466e-04], + [ 2.3976e-05, 8.7452e-04, 1.4091e-04, ..., 5.1880e-04, + 6.7651e-05, 2.7013e-04], + [ 6.2943e-05, 7.0381e-04, 2.1815e-04, ..., 4.7421e-04, + 2.5725e-04, 2.0230e-04], + ..., + [ 3.9600e-06, -5.7373e-03, -1.7462e-03, ..., -3.0746e-03, + -2.2869e-03, -1.4277e-03], + [ 1.3387e-04, 1.3220e-04, 1.3983e-04, ..., 1.5008e-04, + 1.1206e-04, 1.0753e-04], + [ 9.4473e-05, 6.6710e-04, 3.3355e-04, ..., 4.3797e-04, + 5.4836e-04, 2.4164e-04]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0136, -0.0020, -0.0139, 0.0108, -0.0147, -0.0206, 0.0101, -0.0016, + 0.0117, -0.0121], device='cuda:0'), grad: tensor([ 0.0195, 0.0218, -0.0777, 0.0288, 0.0274, 0.0040, -0.0009, -0.0534, + 0.0124, 0.0181], device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 216.78, cls_loss 0.4248 cls_loss_mapping 0.0007 cls_loss_causal 0.3947 re_mapping 0.0037 re_causal 0.0113 /// teacc 98.98 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.0769, -0.1098, -0.1549, ..., 0.1266, -0.0674, 0.0377], + [-0.0955, 0.1128, -0.0582, ..., 0.0514, -0.0625, -0.0972], + [-0.0376, 0.0891, -0.1347, ..., -0.0197, 0.0423, 0.1251], + ..., + [ 0.0026, 0.0749, 0.0016, ..., -0.0251, 0.0336, -0.0860], + [ 0.1031, -0.1220, 0.0472, ..., 0.0207, 0.0104, -0.0376], + [-0.0135, -0.0887, 0.0737, ..., -0.0994, 0.0330, -0.0306]], + device='cuda:0'), grad: tensor([[ 2.2794e-07, 9.6679e-05, 1.1854e-05, ..., 1.4055e-04, + 6.9290e-07, 6.7294e-05], + [ 2.9709e-06, 3.6788e-04, 4.7863e-05, ..., 7.5579e-04, + 1.2824e-06, 3.5262e-04], + [ 2.6822e-06, -1.8301e-03, -1.9014e-04, ..., -2.9812e-03, + 4.3847e-06, -1.4534e-03], + ..., + [-9.5461e-07, 6.2771e-06, 2.9415e-05, ..., -1.9744e-07, + 5.5097e-06, 5.8383e-05], + [ 6.4587e-07, 2.4748e-04, 2.5734e-05, ..., 3.6597e-04, + 2.7288e-06, 1.7214e-04], + [-1.1474e-05, 1.2517e-04, -4.5866e-05, ..., 1.7238e-04, + -3.5435e-05, 5.9068e-05]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0137, -0.0018, -0.0138, 0.0107, -0.0149, -0.0206, 0.0099, -0.0016, + 0.0116, -0.0120], device='cuda:0'), grad: tensor([ 0.0012, 0.0068, -0.0281, 0.0015, 0.0027, 0.0020, 0.0082, 0.0009, + 0.0035, 0.0013], device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 216.75, cls_loss 0.4298 cls_loss_mapping 0.0007 cls_loss_causal 0.3951 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.01 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.0770, -0.1098, -0.1549, ..., 0.1266, -0.0674, 0.0377], + [-0.0956, 0.1127, -0.0583, ..., 0.0513, -0.0626, -0.0973], + [-0.0375, 0.0892, -0.1346, ..., -0.0198, 0.0424, 0.1251], + ..., + [ 0.0025, 0.0749, 0.0016, ..., -0.0252, 0.0336, -0.0861], + [ 0.1032, -0.1219, 0.0472, ..., 0.0208, 0.0104, -0.0377], + [-0.0136, -0.0887, 0.0737, ..., -0.0995, 0.0331, -0.0305]], + device='cuda:0'), grad: tensor([[ 3.8624e-04, -2.8877e-03, -1.6365e-03, ..., 2.5654e-04, + 4.8018e-04, -4.7913e-03], + [ 1.0198e-04, 2.6989e-04, 3.8242e-04, ..., 4.1395e-05, + 9.5665e-05, 4.6277e-04], + [-5.7125e-04, 2.1648e-03, 1.8330e-03, ..., -7.7057e-04, + -8.2254e-04, 3.5629e-03], + ..., + [ 1.6201e-04, 4.9695e-06, 1.1826e-03, ..., 6.2346e-05, + 1.1474e-04, 9.8348e-06], + [-3.4370e-03, 1.9741e-04, -7.2527e-04, ..., -1.1616e-03, + 3.9177e-03, 3.1781e-04], + [ 2.0647e-04, 1.7464e-05, 1.0548e-03, ..., 7.8678e-05, + 1.9729e-04, 4.1723e-05]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0138, -0.0020, -0.0138, 0.0108, -0.0147, -0.0207, 0.0100, -0.0017, + 0.0116, -0.0121], device='cuda:0'), grad: tensor([-0.0130, 0.0122, 0.0033, 0.0258, 0.0131, 0.0280, -0.0132, -0.0141, + -0.0265, -0.0155], device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 217.23, cls_loss 0.4063 cls_loss_mapping 0.0007 cls_loss_causal 0.3774 re_mapping 0.0037 re_causal 0.0110 /// teacc 98.99 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.0771, -0.1097, -0.1550, ..., 0.1266, -0.0674, 0.0376], + [-0.0955, 0.1127, -0.0583, ..., 0.0514, -0.0626, -0.0974], + [-0.0375, 0.0894, -0.1348, ..., -0.0198, 0.0423, 0.1253], + ..., + [ 0.0024, 0.0748, 0.0015, ..., -0.0253, 0.0336, -0.0861], + [ 0.1032, -0.1219, 0.0472, ..., 0.0209, 0.0104, -0.0377], + [-0.0138, -0.0888, 0.0738, ..., -0.0996, 0.0331, -0.0304]], + device='cuda:0'), grad: tensor([[-6.5422e-04, 2.8200e-06, -1.7653e-03, ..., -3.0975e-03, + -2.5139e-03, -9.8991e-04], + [ 9.4473e-05, -8.1730e-04, 1.0720e-06, ..., -8.6486e-05, + 8.4043e-06, 7.8678e-06], + [ 1.3661e-04, 7.3004e-04, -2.7299e-04, ..., 2.6441e-04, + 3.1561e-05, -6.7949e-05], + ..., + [ 1.1139e-03, 1.6260e-04, 6.9284e-04, ..., 4.8399e-04, + 5.6601e-04, 4.0680e-05], + [ 4.7326e-04, 2.0385e-05, 2.0647e-04, ..., 2.7704e-04, + 2.0027e-04, 9.4235e-05], + [-2.9354e-03, -1.9491e-04, -8.3256e-04, ..., 4.7040e-04, + -2.8062e-04, 2.6441e-04]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0138, -0.0020, -0.0138, 0.0109, -0.0148, -0.0208, 0.0099, -0.0017, + 0.0116, -0.0121], device='cuda:0'), grad: tensor([-0.0278, -0.0083, 0.0164, 0.0231, -0.0007, 0.0226, -0.0445, 0.0354, + 0.0203, -0.0366], device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 216.80, cls_loss 0.4374 cls_loss_mapping 0.0006 cls_loss_causal 0.4127 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.00 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.0770, -0.1097, -0.1548, ..., 0.1265, -0.0675, 0.0374], + [-0.0955, 0.1128, -0.0582, ..., 0.0514, -0.0626, -0.0974], + [-0.0375, 0.0894, -0.1346, ..., -0.0198, 0.0423, 0.1253], + ..., + [ 0.0023, 0.0747, 0.0015, ..., -0.0254, 0.0335, -0.0860], + [ 0.1032, -0.1218, 0.0472, ..., 0.0209, 0.0105, -0.0377], + [-0.0138, -0.0887, 0.0737, ..., -0.0997, 0.0331, -0.0305]], + device='cuda:0'), grad: tensor([[ 1.2517e-04, 1.2106e-04, 1.6613e-03, ..., 2.0771e-03, + 1.6575e-03, 2.7990e-04], + [-7.5340e-04, -8.8835e-04, 2.0218e-04, ..., -8.7509e-03, + 1.6510e-04, -1.8053e-03], + [ 4.2129e-04, 3.9434e-04, 1.2457e-04, ..., 4.2076e-03, + 1.0616e-04, 8.7595e-04], + ..., + [-1.7667e-04, 1.9222e-05, 1.2924e-02, ..., -4.1656e-03, + 5.8098e-03, 1.7631e-04], + [ 8.9705e-05, 1.0449e-04, 2.3899e-03, ..., 1.4763e-03, + 1.9932e-03, 1.8942e-04], + [ 1.1188e-04, 9.7394e-05, 3.8055e-02, ..., 1.2541e-03, + 1.7319e-02, 1.3161e-04]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0137, -0.0019, -0.0139, 0.0108, -0.0147, -0.0207, 0.0099, -0.0018, + 0.0116, -0.0120], device='cuda:0'), grad: tensor([ 0.0289, -0.0620, 0.0299, 0.0275, -0.1019, -0.0002, -0.0088, 0.0133, + 0.0262, 0.0472], device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 216.86, cls_loss 0.4354 cls_loss_mapping 0.0007 cls_loss_causal 0.4045 re_mapping 0.0036 re_causal 0.0114 /// teacc 99.02 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.0770, -0.1098, -0.1550, ..., 0.1266, -0.0675, 0.0374], + [-0.0954, 0.1129, -0.0583, ..., 0.0514, -0.0626, -0.0975], + [-0.0375, 0.0893, -0.1347, ..., -0.0199, 0.0423, 0.1253], + ..., + [ 0.0025, 0.0747, 0.0016, ..., -0.0254, 0.0335, -0.0862], + [ 0.1032, -0.1219, 0.0472, ..., 0.0210, 0.0104, -0.0375], + [-0.0138, -0.0888, 0.0736, ..., -0.0996, 0.0331, -0.0304]], + device='cuda:0'), grad: tensor([[ 1.0309e-03, 1.3962e-05, 4.2582e-04, ..., 3.9363e-04, + 5.6416e-05, 1.3041e-04], + [ 3.9363e-04, 6.0081e-04, 7.5400e-05, ..., 5.5361e-04, + 6.6519e-05, 9.4652e-05], + [-7.6389e-04, -6.7425e-04, -1.9526e-04, ..., 1.0538e-03, + 1.7014e-03, 1.6510e-04], + ..., + [ 3.4308e-04, -1.5366e-04, 2.2087e-03, ..., -1.0691e-03, + -2.2564e-03, 9.4891e-05], + [ 2.4915e-04, 1.2890e-05, 2.6369e-04, ..., 2.9421e-04, + 2.0659e-04, 8.0705e-05], + [ 2.7871e-04, -2.9132e-05, 1.6190e-02, ..., -1.5774e-03, + -1.6651e-03, -9.7656e-04]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0137, -0.0019, -0.0138, 0.0108, -0.0149, -0.0206, 0.0100, -0.0017, + 0.0116, -0.0121], device='cuda:0'), grad: tensor([ 0.0010, 0.0270, -0.0221, -0.0269, -0.0132, 0.0292, -0.0132, 0.0067, + -0.0036, 0.0151], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 497---------------------------------------------------- +epoch 497, time 217.27, cls_loss 0.4177 cls_loss_mapping 0.0007 cls_loss_causal 0.3929 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.05 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.0768, -0.1097, -0.1550, ..., 0.1265, -0.0677, 0.0373], + [-0.0953, 0.1128, -0.0580, ..., 0.0517, -0.0624, -0.0974], + [-0.0376, 0.0892, -0.1347, ..., -0.0200, 0.0423, 0.1253], + ..., + [ 0.0024, 0.0748, 0.0015, ..., -0.0254, 0.0334, -0.0862], + [ 0.1032, -0.1220, 0.0475, ..., 0.0211, 0.0106, -0.0375], + [-0.0137, -0.0888, 0.0736, ..., -0.0998, 0.0331, -0.0305]], + device='cuda:0'), grad: tensor([[ 1.7223e-03, 3.6955e-04, 4.5371e-04, ..., 1.6356e-03, + 4.6134e-04, 1.1641e-04], + [-1.5001e-03, -7.5006e-04, 2.1374e-04, ..., -3.1757e-03, + -1.7941e-04, 2.0806e-06], + [ 9.1457e-04, 2.5940e-04, 9.2745e-05, ..., 5.2500e-04, + 2.5368e-04, 9.1344e-06], + ..., + [-6.9160e-03, -3.3550e-03, -2.4796e-03, ..., -9.5701e-04, + -2.1887e-04, 9.1409e-07], + [ 1.1005e-03, 3.1662e-04, 2.7633e-04, ..., -1.0395e-03, + 2.9039e-04, 2.3231e-05], + [ 3.5357e-04, 2.1820e-03, 3.8767e-04, ..., 9.0933e-04, + -2.4910e-03, 1.5028e-05]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0137, -0.0018, -0.0138, 0.0108, -0.0149, -0.0205, 0.0101, -0.0019, + 0.0117, -0.0121], device='cuda:0'), grad: tensor([ 0.0267, -0.0505, 0.0127, 0.0184, 0.0197, -0.0179, 0.0148, -0.0458, + 0.0115, 0.0104], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 498---------------------------------------------------- +epoch 498, time 217.80, cls_loss 0.4471 cls_loss_mapping 0.0006 cls_loss_causal 0.4126 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.08 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.0767, -0.1097, -0.1550, ..., 0.1265, -0.0676, 0.0373], + [-0.0954, 0.1128, -0.0580, ..., 0.0517, -0.0625, -0.0975], + [-0.0373, 0.0893, -0.1346, ..., -0.0199, 0.0423, 0.1254], + ..., + [ 0.0024, 0.0749, 0.0015, ..., -0.0254, 0.0335, -0.0863], + [ 0.1032, -0.1220, 0.0475, ..., 0.0211, 0.0106, -0.0376], + [-0.0138, -0.0888, 0.0737, ..., -0.0998, 0.0333, -0.0304]], + device='cuda:0'), grad: tensor([[ 1.7738e-04, 7.3075e-05, 3.4899e-05, ..., 1.3578e-04, + -1.0125e-05, -2.5891e-07], + [ 4.1056e-04, 1.2255e-04, 8.7857e-05, ..., 2.1720e-04, + 1.2899e-07, 9.7789e-09], + [ 4.1509e-04, 9.0182e-05, 4.8310e-05, ..., 1.7345e-04, + 5.4296e-07, 7.9535e-07], + ..., + [-4.5717e-05, 1.2338e-04, 2.3975e-03, ..., 2.3627e-04, + 7.1645e-05, 6.0536e-09], + [ 7.1573e-04, -7.1907e-04, 2.4748e-04, ..., -1.3876e-03, + 1.7554e-05, -4.4815e-06], + [ 3.2749e-03, 5.7250e-05, 3.2425e-03, ..., 1.1986e-04, + 2.3675e-04, 1.2480e-07]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0137, -0.0019, -0.0137, 0.0108, -0.0150, -0.0206, 0.0100, -0.0018, + 0.0117, -0.0120], device='cuda:0'), grad: tensor([ 0.0096, 0.0131, 0.0141, -0.0019, 0.0072, -0.0065, -0.0220, -0.0092, + -0.0186, 0.0142], device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 218.12, cls_loss 0.4205 cls_loss_mapping 0.0006 cls_loss_causal 0.3901 re_mapping 0.0039 re_causal 0.0116 /// teacc 99.03 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.849998 98.860001 ... 89.287498 76.723734 +ShearY 98.879997 98.889999 ... 89.287498 73.559136 +AutoContrast 98.949997 99.019997 ... 89.287498 66.443730 +Invert 99.000000 99.000000 ... 89.287498 72.944757 +Equalize 98.419998 98.409996 ... 89.287498 66.992412 +Solarize 98.449997 98.570000 ... 89.287498 69.236865 +SolarizeAdd 98.680000 98.610001 ... 89.287498 72.770916 +Posterize 98.979996 99.019997 ... 89.287498 76.629822 +Contrast 99.180000 99.190002 ... 89.287498 78.385061 +Color 99.040001 99.080002 ... 89.287498 67.632051 +Brightness 99.119995 99.199997 ... 89.287498 77.591179 +Sharpness 99.040001 99.080002 ... 89.287498 78.275154 +NoiseSalt 99.029999 99.070000 ... 89.287498 70.400456 +NoiseGaussian 98.970001 99.080002 ... 89.287498 64.850480 +w/o do (original x) 99.080000 0.000000 ... 0.000000 79.583664 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.05 70.605409 77.891345 79.629436 89.586447 79.428159 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last +randm: False +stride: 3 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.909996 98.949997 ... 89.486794 76.881126 +ShearY 98.949997 98.889999 ... 89.486794 73.554698 +AutoContrast 98.930000 99.019997 ... 89.486794 66.492076 +Invert 98.979996 99.000000 ... 89.486794 72.618074 +Equalize 98.419998 98.459999 ... 89.486794 67.132525 +Solarize 98.470001 98.580002 ... 89.486794 68.999302 +SolarizeAdd 98.619995 98.659996 ... 89.486794 72.501183 +Posterize 98.949997 99.019997 ... 89.486794 76.556273 +Contrast 99.169998 99.199997 ... 89.486794 78.399466 +Color 99.059998 99.029999 ... 89.486794 67.984040 +Brightness 99.159996 99.190002 ... 89.486794 77.668425 +Sharpness 99.010002 99.049995 ... 89.486794 78.452616 +NoiseSalt 99.099998 99.049995 ... 89.486794 70.660622 +NoiseGaussian 99.010002 99.019997 ... 89.486794 65.199389 +w/o do (original x) 99.030000 0.000000 ... 0.000000 79.652907 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.08 70.85126 77.602489 79.828326 89.78575 79.516956 diff --git a/Meta-causal/code-withStyleAttack/71585.error b/Meta-causal/code-withStyleAttack/71585.error new file mode 100644 index 0000000000000000000000000000000000000000..6299d70f46fb33ff0c8abffa58005ba3dd92ae75 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71585.error @@ -0,0 +1,22 @@ +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:45: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:62: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:72: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:48: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:65: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:75: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) +/scratch/yuqian_fu/micromamba/envs/auto-v5ewbna3m2oe/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) diff --git a/Meta-causal/code-withStyleAttack/71585.log b/Meta-causal/code-withStyleAttack/71585.log new file mode 100644 index 0000000000000000000000000000000000000000..896a7cc7fab4377d308fec25468918d5059fe434 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71585.log @@ -0,0 +1,13385 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0216, -0.0169, -0.0161, ..., 0.0227, -0.0073, 0.0201], + [ 0.0072, -0.0098, -0.0046, ..., -0.0208, -0.0281, 0.0310], + [ 0.0020, 0.0019, -0.0231, ..., -0.0306, -0.0189, -0.0239], + ..., + [-0.0107, 0.0013, 0.0097, ..., -0.0293, 0.0198, -0.0112], + [-0.0206, -0.0148, -0.0202, ..., -0.0044, -0.0050, -0.0301], + [-0.0254, 0.0144, -0.0179, ..., 0.0231, 0.0164, 0.0167]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0171, -0.0200, 0.0085, -0.0257, 0.0185, 0.0054, -0.0096, -0.0275, + 0.0104, -0.0105], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 218.29, cls_loss 2.2478 cls_loss_mapping 2.1626 cls_loss_causal 2.2689 re_mapping 0.0186 re_causal 0.0185 /// teacc 56.55 lr 0.00010000 +Epoch 2, weight, value: tensor([[-2.1272e-02, -1.5845e-02, -2.1736e-02, ..., 2.1055e-02, + -5.7673e-03, 1.7309e-02], + [ 5.2374e-03, -1.1090e-02, 1.3981e-03, ..., -2.0834e-02, + -3.0799e-02, 3.5579e-02], + [-9.6785e-04, 2.6062e-03, -2.4732e-02, ..., -3.2260e-02, + -1.9268e-02, -2.2415e-02], + ..., + [-8.4400e-03, -5.6490e-05, 6.3655e-03, ..., -3.1020e-02, + 1.9608e-02, -1.2059e-02], + [-2.2661e-02, -1.5072e-02, -2.0060e-02, ..., -6.1237e-03, + -6.0846e-03, -2.9732e-02], + [-2.5021e-02, 1.0841e-02, -1.9531e-02, ..., 2.1407e-02, + 1.4514e-02, 1.2631e-02]], device='cuda:0'), grad: tensor([[-0.0026, -0.0016, 0.0003, ..., 0.0000, -0.0142, -0.0223], + [-0.0005, -0.0006, -0.0178, ..., 0.0000, 0.0040, -0.0086], + [-0.0008, -0.0001, 0.0002, ..., 0.0000, -0.0048, -0.0081], + ..., + [ 0.0052, 0.0019, 0.0110, ..., 0.0000, 0.0147, 0.0194], + [ 0.0021, 0.0005, 0.0067, ..., 0.0000, 0.0088, 0.0123], + [-0.0240, 0.0006, -0.0012, ..., 0.0000, -0.0199, -0.0082]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0165, -0.0174, 0.0074, -0.0254, 0.0176, 0.0044, -0.0094, -0.0267, + 0.0097, -0.0114], device='cuda:0'), grad: tensor([-0.0413, -0.0797, 0.0073, 0.0466, 0.0408, 0.0884, -0.0475, 0.0526, + -0.0008, -0.0666], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 216.86, cls_loss 1.8955 cls_loss_mapping 1.0402 cls_loss_causal 1.8900 re_mapping 0.1264 re_causal 0.1424 /// teacc 88.63 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0184, -0.0097, -0.0291, ..., 0.0191, -0.0016, 0.0161], + [ 0.0009, -0.0184, 0.0041, ..., -0.0228, -0.0347, 0.0384], + [-0.0038, 0.0043, -0.0242, ..., -0.0311, -0.0198, -0.0178], + ..., + [-0.0089, -0.0017, 0.0022, ..., -0.0330, 0.0176, -0.0112], + [-0.0257, -0.0158, -0.0198, ..., -0.0081, -0.0081, -0.0328], + [-0.0221, 0.0067, -0.0260, ..., 0.0196, 0.0163, 0.0106]], + device='cuda:0'), grad: tensor([[ 0.0135, 0.0115, 0.0106, ..., 0.0000, 0.0192, 0.0116], + [ 0.0048, 0.0059, -0.0017, ..., 0.0000, -0.0061, -0.0155], + [ 0.0107, 0.0018, 0.0019, ..., 0.0000, 0.0175, 0.0075], + ..., + [-0.0157, -0.0072, -0.0093, ..., 0.0000, -0.0295, -0.0225], + [ 0.0039, 0.0002, -0.0144, ..., 0.0000, 0.0112, 0.0130], + [-0.0113, 0.0040, 0.0046, ..., 0.0000, -0.0090, -0.0066]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0169, -0.0167, 0.0075, -0.0250, 0.0165, 0.0052, -0.0094, -0.0271, + 0.0093, -0.0118], device='cuda:0'), grad: tensor([ 0.0581, -0.0248, 0.0615, -0.0936, 0.0560, 0.0130, 0.0324, -0.0864, + 0.0248, -0.0410], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 216.87, cls_loss 1.5661 cls_loss_mapping 0.4825 cls_loss_causal 1.5157 re_mapping 0.1189 re_causal 0.1853 /// teacc 91.49 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0163, -0.0069, -0.0343, ..., 0.0184, 0.0006, 0.0149], + [-0.0046, -0.0234, 0.0049, ..., -0.0229, -0.0370, 0.0410], + [-0.0083, 0.0040, -0.0266, ..., -0.0313, -0.0221, -0.0146], + ..., + [-0.0102, -0.0009, -0.0022, ..., -0.0332, 0.0163, -0.0078], + [-0.0313, -0.0177, -0.0195, ..., -0.0094, -0.0108, -0.0340], + [-0.0188, 0.0052, -0.0278, ..., 0.0191, 0.0179, 0.0091]], + device='cuda:0'), grad: tensor([[ 3.2825e-03, -2.4548e-03, 1.9512e-03, ..., 0.0000e+00, + 2.0542e-03, 3.3493e-03], + [-1.0300e-02, 3.5477e-03, -1.2741e-03, ..., 0.0000e+00, + -1.5152e-02, -3.1891e-02], + [ 7.4272e-03, -2.4155e-02, -1.2970e-02, ..., 0.0000e+00, + -1.1566e-02, -8.7833e-04], + ..., + [ 1.7059e-02, 7.7591e-03, 2.1801e-03, ..., 0.0000e+00, + 1.7679e-04, -6.6032e-03], + [-1.0521e-02, -4.7379e-03, -4.3899e-05, ..., 0.0000e+00, + -2.3162e-04, 6.1417e-03], + [-2.4902e-02, -8.6451e-04, 2.2869e-03, ..., 0.0000e+00, + -1.6375e-03, 9.7046e-03]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0169, -0.0172, 0.0072, -0.0249, 0.0156, 0.0052, -0.0100, -0.0271, + 0.0093, -0.0097], device='cuda:0'), grad: tensor([ 0.0070, -0.0759, -0.0084, 0.0700, 0.0441, -0.0322, -0.0231, 0.0092, + 0.0183, -0.0090], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 218.74, cls_loss 1.3811 cls_loss_mapping 0.3622 cls_loss_causal 1.3383 re_mapping 0.0919 re_causal 0.1674 /// teacc 93.71 lr 0.00010000 +Epoch 5, weight, value: tensor([[-0.0147, -0.0042, -0.0367, ..., 0.0163, 0.0021, 0.0152], + [-0.0084, -0.0265, 0.0061, ..., -0.0243, -0.0383, 0.0425], + [-0.0113, 0.0026, -0.0297, ..., -0.0353, -0.0236, -0.0108], + ..., + [-0.0115, -0.0007, -0.0054, ..., -0.0377, 0.0151, -0.0034], + [-0.0339, -0.0185, -0.0204, ..., -0.0106, -0.0137, -0.0371], + [-0.0171, 0.0051, -0.0290, ..., 0.0192, 0.0188, 0.0064]], + device='cuda:0'), grad: tensor([[-5.9624e-03, -1.5850e-03, 3.9902e-03, ..., 1.6892e-04, + -2.1439e-02, 4.1656e-03], + [-1.2321e-02, -8.1360e-05, 2.9945e-03, ..., 1.1331e-04, + -2.7924e-03, -1.5762e-02], + [-1.5640e-02, 1.2436e-03, 5.2223e-03, ..., 3.8218e-04, + -6.0158e-03, -1.9522e-03], + ..., + [ 2.9278e-03, -6.7024e-03, -6.5842e-03, ..., 7.5996e-05, + -2.7115e-02, -8.9493e-03], + [-1.1568e-03, 1.7996e-03, 2.9774e-03, ..., 2.2948e-04, + 1.7118e-03, 1.7862e-03], + [ 1.6220e-02, 7.4615e-03, 5.4817e-03, ..., 6.9380e-05, + 3.5431e-02, 1.3947e-02]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0168, -0.0172, 0.0074, -0.0239, 0.0152, 0.0044, -0.0106, -0.0268, + 0.0093, -0.0096], device='cuda:0'), grad: tensor([ 0.0197, -0.0647, -0.0373, 0.0230, 0.0162, -0.0134, 0.0316, -0.0243, + -0.0094, 0.0586], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 226.58, cls_loss 1.3236 cls_loss_mapping 0.2895 cls_loss_causal 1.2855 re_mapping 0.0684 re_causal 0.1410 /// teacc 94.78 lr 0.00010000 +Epoch 6, weight, value: tensor([[-1.3532e-02, -2.2119e-03, -3.9380e-02, ..., 1.5567e-02, + 3.7612e-03, 1.5809e-02], + [-1.1712e-02, -2.9554e-02, 6.6768e-03, ..., -2.3649e-02, + -4.0111e-02, 4.3700e-02], + [-1.4374e-02, 1.2228e-03, -3.0154e-02, ..., -3.6499e-02, + -2.5640e-02, -9.1345e-03], + ..., + [-1.2040e-02, -7.0142e-04, -7.8608e-03, ..., -3.8825e-02, + 1.4733e-02, 3.9212e-05], + [-3.6277e-02, -1.8502e-02, -2.1491e-02, ..., -1.0483e-02, + -1.5148e-02, -3.8012e-02], + [-1.5682e-02, 5.3608e-03, -3.0244e-02, ..., 1.9863e-02, + 1.9645e-02, 4.0173e-03]], device='cuda:0'), grad: tensor([[ 1.9897e-02, -7.4685e-05, 1.6737e-03, ..., 0.0000e+00, + 1.0246e-02, 9.5444e-03], + [ 2.1133e-02, 2.6226e-03, 1.1787e-03, ..., 0.0000e+00, + 1.5076e-02, 1.2077e-02], + [-6.0225e-04, 1.3704e-03, 8.1787e-03, ..., 0.0000e+00, + 5.5962e-03, -6.2523e-03], + ..., + [-2.8549e-02, 3.5667e-03, 1.1978e-03, ..., 0.0000e+00, + -1.0498e-02, -1.4984e-02], + [ 2.6016e-02, 6.2714e-03, 4.2343e-03, ..., 0.0000e+00, + 1.4153e-02, 1.1543e-02], + [-2.5757e-02, -7.9956e-03, 2.7394e-04, ..., 0.0000e+00, + -2.2705e-02, -4.7150e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0168, -0.0176, 0.0072, -0.0234, 0.0148, 0.0038, -0.0101, -0.0268, + 0.0099, -0.0095], device='cuda:0'), grad: tensor([ 0.0565, 0.0844, -0.0160, -0.0020, -0.0271, -0.0300, -0.0302, -0.0707, + 0.0781, -0.0431], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 217.90, cls_loss 1.2016 cls_loss_mapping 0.2316 cls_loss_causal 1.1527 re_mapping 0.0605 re_causal 0.1292 /// teacc 95.01 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0126, -0.0006, -0.0421, ..., 0.0113, 0.0051, 0.0170], + [-0.0139, -0.0321, 0.0075, ..., -0.0226, -0.0417, 0.0449], + [-0.0156, 0.0005, -0.0323, ..., -0.0386, -0.0256, -0.0063], + ..., + [-0.0122, -0.0014, -0.0101, ..., -0.0444, 0.0142, 0.0020], + [-0.0389, -0.0195, -0.0231, ..., -0.0126, -0.0169, -0.0390], + [-0.0145, 0.0048, -0.0314, ..., 0.0165, 0.0204, 0.0027]], + device='cuda:0'), grad: tensor([[ 6.9199e-03, -1.4191e-03, -1.3437e-03, ..., 8.4162e-05, + 6.1264e-03, -6.9189e-04], + [-3.5172e-03, 9.5308e-05, -4.4098e-03, ..., 1.5926e-04, + -7.1526e-03, -8.7738e-03], + [-4.6768e-03, 4.7188e-03, 8.6823e-03, ..., 3.4571e-04, + -1.6571e-02, 7.0114e-03], + ..., + [-2.0828e-03, -3.8662e-03, -4.9210e-03, ..., -7.9346e-04, + 8.2541e-04, 8.1110e-04], + [ 1.4694e-02, -2.2907e-03, -2.4235e-04, ..., -6.5386e-05, + 7.2746e-03, 6.9580e-03], + [ 3.3073e-03, 3.8967e-03, 9.1171e-03, ..., 2.1636e-04, + 3.4599e-03, -3.6488e-03]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0167, -0.0179, 0.0075, -0.0231, 0.0150, 0.0037, -0.0104, -0.0268, + 0.0096, -0.0093], device='cuda:0'), grad: tensor([ 0.0173, -0.0233, 0.0130, -0.0310, -0.0104, -0.0260, 0.0551, -0.0343, + 0.0294, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 216.75, cls_loss 1.1780 cls_loss_mapping 0.2158 cls_loss_causal 1.1464 re_mapping 0.0506 re_causal 0.1158 /// teacc 96.01 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0124, 0.0006, -0.0437, ..., 0.0043, 0.0057, 0.0162], + [-0.0152, -0.0331, 0.0094, ..., -0.0227, -0.0429, 0.0451], + [-0.0164, -0.0011, -0.0344, ..., -0.0393, -0.0251, -0.0045], + ..., + [-0.0122, -0.0015, -0.0120, ..., -0.0469, 0.0131, 0.0040], + [-0.0414, -0.0206, -0.0239, ..., -0.0128, -0.0182, -0.0406], + [-0.0138, 0.0056, -0.0325, ..., 0.0111, 0.0213, 0.0024]], + device='cuda:0'), grad: tensor([[ 0.0262, 0.0037, 0.0068, ..., 0.0038, 0.0185, 0.0164], + [ 0.0138, 0.0023, 0.0075, ..., 0.0021, 0.0073, 0.0181], + [-0.0111, -0.0021, -0.0075, ..., -0.0028, -0.0080, -0.0139], + ..., + [-0.0006, 0.0026, 0.0080, ..., 0.0012, 0.0004, -0.0195], + [-0.0080, 0.0001, -0.0035, ..., -0.0017, -0.0051, -0.0005], + [ 0.0046, 0.0015, -0.0036, ..., 0.0024, 0.0018, -0.0068]], + device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0165, -0.0180, 0.0076, -0.0227, 0.0148, 0.0038, -0.0104, -0.0270, + 0.0099, -0.0095], device='cuda:0'), grad: tensor([ 0.0858, 0.0553, -0.0441, 0.0030, 0.0107, -0.0159, -0.0180, -0.0116, + -0.0378, -0.0274], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 216.41, cls_loss 1.1000 cls_loss_mapping 0.1992 cls_loss_causal 1.0685 re_mapping 0.0482 re_causal 0.1136 /// teacc 96.06 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0120, 0.0020, -0.0469, ..., -0.0034, 0.0065, 0.0160], + [-0.0169, -0.0335, 0.0103, ..., -0.0215, -0.0442, 0.0453], + [-0.0180, -0.0032, -0.0350, ..., -0.0388, -0.0259, -0.0033], + ..., + [-0.0127, -0.0009, -0.0135, ..., -0.0490, 0.0126, 0.0055], + [-0.0427, -0.0218, -0.0260, ..., -0.0153, -0.0192, -0.0405], + [-0.0118, 0.0052, -0.0328, ..., 0.0060, 0.0229, 0.0022]], + device='cuda:0'), grad: tensor([[ 0.0135, 0.0019, 0.0058, ..., 0.0014, 0.0105, 0.0065], + [ 0.0010, 0.0006, 0.0042, ..., 0.0024, -0.0026, 0.0004], + [-0.0354, 0.0026, -0.0004, ..., 0.0015, -0.0160, -0.0315], + ..., + [ 0.0127, 0.0004, 0.0035, ..., 0.0002, 0.0081, 0.0065], + [ 0.0197, 0.0049, 0.0112, ..., 0.0031, 0.0058, 0.0113], + [ 0.0107, 0.0007, 0.0029, ..., 0.0008, 0.0111, 0.0078]], + device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0162, -0.0178, 0.0074, -0.0227, 0.0145, 0.0037, -0.0104, -0.0274, + 0.0103, -0.0090], device='cuda:0'), grad: tensor([ 0.0156, -0.0084, -0.0817, -0.0279, 0.0190, -0.0438, 0.0323, 0.0334, + 0.0410, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 216.78, cls_loss 1.0795 cls_loss_mapping 0.1518 cls_loss_causal 1.0444 re_mapping 0.0431 re_causal 0.1020 /// teacc 96.52 lr 0.00010000 +Epoch 10, weight, value: tensor([[-0.0110, 0.0038, -0.0470, ..., -0.0045, 0.0078, 0.0153], + [-0.0170, -0.0348, 0.0113, ..., -0.0212, -0.0444, 0.0455], + [-0.0195, -0.0041, -0.0355, ..., -0.0400, -0.0269, -0.0024], + ..., + [-0.0131, -0.0018, -0.0151, ..., -0.0508, 0.0126, 0.0068], + [-0.0444, -0.0224, -0.0262, ..., -0.0169, -0.0212, -0.0414], + [-0.0108, 0.0048, -0.0344, ..., 0.0032, 0.0237, 0.0016]], + device='cuda:0'), grad: tensor([[-0.0133, -0.0144, -0.0164, ..., 0.0010, -0.0164, -0.0031], + [-0.0266, -0.0037, -0.0142, ..., -0.0008, -0.0120, -0.0214], + [ 0.0099, 0.0029, 0.0033, ..., 0.0027, 0.0083, 0.0016], + ..., + [ 0.0277, 0.0023, 0.0046, ..., 0.0005, 0.0160, 0.0193], + [-0.0089, 0.0017, 0.0010, ..., 0.0016, -0.0099, -0.0166], + [-0.0078, 0.0013, 0.0028, ..., 0.0007, -0.0111, -0.0011]], + device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0165, -0.0175, 0.0069, -0.0226, 0.0145, 0.0041, -0.0108, -0.0275, + 0.0103, -0.0090], device='cuda:0'), grad: tensor([-0.0313, -0.0902, 0.0240, 0.0376, 0.0197, 0.0261, 0.0094, 0.0840, + -0.0724, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 216.30, cls_loss 1.0639 cls_loss_mapping 0.1551 cls_loss_causal 1.0356 re_mapping 0.0408 re_causal 0.1043 /// teacc 96.78 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0113, 0.0048, -0.0483, ..., -0.0072, 0.0080, 0.0156], + [-0.0185, -0.0379, 0.0121, ..., -0.0226, -0.0457, 0.0453], + [-0.0192, -0.0043, -0.0365, ..., -0.0420, -0.0252, -0.0015], + ..., + [-0.0128, -0.0018, -0.0159, ..., -0.0508, 0.0124, 0.0078], + [-0.0464, -0.0238, -0.0269, ..., -0.0186, -0.0226, -0.0410], + [-0.0110, 0.0051, -0.0356, ..., -0.0010, 0.0236, 0.0011]], + device='cuda:0'), grad: tensor([[ 5.4855e-03, 1.1921e-03, 3.3226e-03, ..., 1.2770e-03, + 4.3983e-03, 4.1733e-03], + [ 7.4158e-03, 6.8188e-04, 3.7313e-05, ..., 3.8695e-04, + -6.3171e-03, -2.1019e-03], + [ 1.7195e-03, -9.0485e-03, -1.5097e-03, ..., -7.4577e-03, + 9.6436e-03, -1.2100e-02], + ..., + [ 1.2764e-02, 3.9902e-03, 3.8910e-03, ..., 6.1369e-04, + 2.0370e-02, 5.9471e-03], + [ 1.0939e-03, 6.2256e-03, -1.7883e-02, ..., 5.4245e-03, + 3.8223e-03, -4.4518e-03], + [-7.5317e-02, -6.8626e-03, -7.3357e-03, ..., -1.3199e-03, + -7.8491e-02, -1.1536e-02]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0166, -0.0178, 0.0075, -0.0223, 0.0146, 0.0039, -0.0111, -0.0280, + 0.0106, -0.0092], device='cuda:0'), grad: tensor([ 0.0214, 0.0058, -0.0446, 0.0125, 0.0708, -0.0020, 0.0164, 0.0314, + -0.0127, -0.0990], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 227.69, cls_loss 1.0407 cls_loss_mapping 0.1593 cls_loss_causal 1.0077 re_mapping 0.0381 re_causal 0.0955 /// teacc 97.13 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0106, 0.0058, -0.0502, ..., -0.0095, 0.0091, 0.0159], + [-0.0197, -0.0390, 0.0123, ..., -0.0239, -0.0471, 0.0451], + [-0.0204, -0.0045, -0.0377, ..., -0.0442, -0.0258, -0.0005], + ..., + [-0.0128, -0.0021, -0.0159, ..., -0.0516, 0.0121, 0.0095], + [-0.0474, -0.0239, -0.0270, ..., -0.0184, -0.0232, -0.0423], + [-0.0099, 0.0053, -0.0353, ..., -0.0033, 0.0241, -0.0004]], + device='cuda:0'), grad: tensor([[ 3.4332e-02, -3.7098e-03, 1.3742e-03, ..., 5.7564e-03, + 1.2863e-02, 1.3603e-02], + [-3.1300e-03, 3.7899e-03, -7.0810e-04, ..., 5.5373e-05, + -1.2789e-03, 5.9509e-03], + [ 4.4975e-03, -1.5503e-02, -1.1292e-02, ..., -1.7872e-03, + -1.7748e-03, 1.5282e-02], + ..., + [ 6.1569e-03, 4.3221e-03, 7.7820e-04, ..., 1.8513e-04, + 7.0457e-03, 9.9945e-04], + [-2.0782e-02, 3.0384e-03, 1.0414e-03, ..., -5.2795e-03, + -7.1678e-03, -2.6825e-02], + [-1.6373e-02, -8.0261e-03, 4.5013e-04, ..., -8.4019e-04, + -1.5457e-02, -1.0391e-02]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0166, -0.0179, 0.0071, -0.0221, 0.0146, 0.0040, -0.0115, -0.0276, + 0.0106, -0.0092], device='cuda:0'), grad: tensor([ 0.0474, -0.0064, 0.0134, 0.0070, 0.0254, 0.0170, -0.0158, 0.0210, + -0.0765, -0.0326], device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 226.12, cls_loss 0.9999 cls_loss_mapping 0.1389 cls_loss_causal 0.9711 re_mapping 0.0353 re_causal 0.0938 /// teacc 96.95 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0101, 0.0067, -0.0517, ..., -0.0128, 0.0094, 0.0151], + [-0.0206, -0.0400, 0.0128, ..., -0.0245, -0.0474, 0.0452], + [-0.0212, -0.0046, -0.0381, ..., -0.0455, -0.0262, 0.0002], + ..., + [-0.0132, -0.0034, -0.0172, ..., -0.0529, 0.0123, 0.0112], + [-0.0487, -0.0259, -0.0279, ..., -0.0208, -0.0243, -0.0425], + [-0.0095, 0.0055, -0.0351, ..., -0.0067, 0.0248, -0.0011]], + device='cuda:0'), grad: tensor([[ 0.0060, -0.0022, 0.0011, ..., 0.0017, 0.0114, 0.0014], + [ 0.0078, 0.0009, 0.0024, ..., 0.0024, 0.0031, -0.0019], + [ 0.0038, 0.0017, 0.0009, ..., 0.0024, -0.0010, 0.0012], + ..., + [ 0.0178, 0.0024, 0.0016, ..., 0.0014, 0.0144, 0.0129], + [-0.0121, 0.0042, 0.0034, ..., 0.0023, -0.0085, -0.0130], + [-0.0164, -0.0064, 0.0016, ..., -0.0014, -0.0219, -0.0019]], + device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0165, -0.0181, 0.0074, -0.0219, 0.0150, 0.0045, -0.0122, -0.0277, + 0.0106, -0.0095], device='cuda:0'), grad: tensor([ 0.0276, 0.0116, -0.0294, 0.0057, -0.0071, 0.0033, -0.0267, 0.0536, + -0.0042, -0.0344], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 225.85, cls_loss 0.9804 cls_loss_mapping 0.1408 cls_loss_causal 0.9432 re_mapping 0.0331 re_causal 0.0884 /// teacc 97.19 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0092, 0.0082, -0.0525, ..., -0.0139, 0.0111, 0.0153], + [-0.0216, -0.0406, 0.0135, ..., -0.0242, -0.0479, 0.0454], + [-0.0225, -0.0047, -0.0384, ..., -0.0466, -0.0264, 0.0011], + ..., + [-0.0133, -0.0046, -0.0176, ..., -0.0528, 0.0121, 0.0116], + [-0.0497, -0.0262, -0.0282, ..., -0.0201, -0.0250, -0.0437], + [-0.0082, 0.0061, -0.0355, ..., -0.0071, 0.0257, -0.0012]], + device='cuda:0'), grad: tensor([[-0.0121, -0.0036, 0.0004, ..., 0.0019, -0.0093, 0.0002], + [ 0.0077, 0.0015, 0.0020, ..., 0.0014, 0.0108, 0.0054], + [-0.0014, 0.0008, -0.0030, ..., -0.0048, 0.0001, -0.0158], + ..., + [-0.0009, 0.0008, 0.0015, ..., 0.0002, 0.0016, -0.0009], + [ 0.0101, 0.0018, 0.0062, ..., 0.0024, 0.0123, 0.0095], + [-0.0101, -0.0014, 0.0004, ..., 0.0004, -0.0029, -0.0081]], + device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0163, -0.0180, 0.0077, -0.0220, 0.0148, 0.0051, -0.0124, -0.0282, + 0.0108, -0.0093], device='cuda:0'), grad: tensor([-0.0164, 0.0374, -0.0414, -0.0790, 0.0343, 0.0408, 0.0089, -0.0123, + 0.0558, -0.0281], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 226.43, cls_loss 0.9499 cls_loss_mapping 0.1187 cls_loss_causal 0.9250 re_mapping 0.0330 re_causal 0.0892 /// teacc 97.03 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0091, 0.0101, -0.0532, ..., -0.0166, 0.0115, 0.0156], + [-0.0231, -0.0424, 0.0136, ..., -0.0233, -0.0496, 0.0453], + [-0.0233, -0.0058, -0.0391, ..., -0.0478, -0.0275, 0.0015], + ..., + [-0.0132, -0.0045, -0.0186, ..., -0.0531, 0.0125, 0.0132], + [-0.0491, -0.0258, -0.0284, ..., -0.0197, -0.0242, -0.0439], + [-0.0083, 0.0054, -0.0364, ..., -0.0118, 0.0254, -0.0023]], + device='cuda:0'), grad: tensor([[-0.0249, -0.0055, 0.0011, ..., -0.0050, -0.0187, -0.0068], + [-0.0070, 0.0006, -0.0014, ..., -0.0007, -0.0046, -0.0086], + [-0.0135, -0.0036, -0.0118, ..., 0.0014, -0.0033, 0.0018], + ..., + [ 0.0106, 0.0006, 0.0010, ..., 0.0011, 0.0099, 0.0058], + [ 0.0150, 0.0035, 0.0031, ..., 0.0029, 0.0102, 0.0066], + [ 0.0095, 0.0030, 0.0043, ..., 0.0015, 0.0083, 0.0037]], + device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0168, -0.0187, 0.0070, -0.0217, 0.0158, 0.0047, -0.0126, -0.0279, + 0.0112, -0.0098], device='cuda:0'), grad: tensor([-0.0636, -0.0247, -0.0142, 0.0323, 0.0056, -0.0378, 0.0071, 0.0345, + 0.0445, 0.0162], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 228.66, cls_loss 0.9656 cls_loss_mapping 0.1237 cls_loss_causal 0.9339 re_mapping 0.0310 re_causal 0.0862 /// teacc 97.21 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0086, 0.0122, -0.0543, ..., -0.0188, 0.0126, 0.0157], + [-0.0234, -0.0441, 0.0142, ..., -0.0227, -0.0504, 0.0457], + [-0.0231, -0.0070, -0.0398, ..., -0.0495, -0.0271, 0.0024], + ..., + [-0.0130, -0.0041, -0.0191, ..., -0.0541, 0.0129, 0.0136], + [-0.0513, -0.0263, -0.0283, ..., -0.0209, -0.0256, -0.0453], + [-0.0082, 0.0050, -0.0370, ..., -0.0118, 0.0256, -0.0026]], + device='cuda:0'), grad: tensor([[-1.4687e-02, -6.0730e-03, 1.3447e-03, ..., 4.6325e-04, + -7.9498e-03, -8.9417e-03], + [-2.0962e-03, 6.1798e-04, -6.4731e-05, ..., 3.7599e-04, + -1.9445e-03, -1.3227e-03], + [ 3.5267e-03, 2.1801e-03, 1.8654e-03, ..., 9.8038e-04, + 6.8893e-03, 3.7823e-03], + ..., + [ 7.1030e-03, -6.9714e-04, 7.2384e-04, ..., 1.6439e-04, + 4.4746e-03, 8.5144e-03], + [ 5.1765e-03, 1.9255e-03, 6.7091e-04, ..., -1.2541e-04, + 5.4474e-03, -2.8648e-03], + [-3.9978e-03, 5.6458e-04, -3.6087e-03, ..., -1.1816e-03, + -4.0436e-03, -8.7357e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0165, -0.0183, 0.0073, -0.0215, 0.0159, 0.0053, -0.0133, -0.0280, + 0.0108, -0.0100], device='cuda:0'), grad: tensor([-0.0335, 0.0053, 0.0127, 0.0046, -0.0044, -0.0013, 0.0278, 0.0244, + -0.0038, -0.0318], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 228.95, cls_loss 0.8734 cls_loss_mapping 0.1091 cls_loss_causal 0.8427 re_mapping 0.0316 re_causal 0.0847 /// teacc 97.65 lr 0.00010000 +Epoch 17, weight, value: tensor([[-0.0084, 0.0136, -0.0556, ..., -0.0195, 0.0130, 0.0155], + [-0.0235, -0.0444, 0.0146, ..., -0.0229, -0.0514, 0.0459], + [-0.0234, -0.0076, -0.0400, ..., -0.0515, -0.0275, 0.0027], + ..., + [-0.0128, -0.0039, -0.0195, ..., -0.0530, 0.0134, 0.0147], + [-0.0524, -0.0268, -0.0288, ..., -0.0206, -0.0260, -0.0458], + [-0.0081, 0.0034, -0.0386, ..., -0.0150, 0.0251, -0.0022]], + device='cuda:0'), grad: tensor([[-0.0062, 0.0024, -0.0019, ..., 0.0006, -0.0007, 0.0011], + [ 0.0032, -0.0005, 0.0071, ..., 0.0015, -0.0054, -0.0022], + [ 0.0085, -0.0005, -0.0041, ..., 0.0016, 0.0041, -0.0084], + ..., + [ 0.0052, 0.0031, 0.0058, ..., 0.0010, 0.0038, 0.0096], + [-0.0083, 0.0009, -0.0227, ..., -0.0103, -0.0061, -0.0011], + [-0.0046, 0.0025, 0.0058, ..., 0.0023, -0.0031, -0.0079]], + device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0165, -0.0180, 0.0075, -0.0215, 0.0159, 0.0056, -0.0134, -0.0280, + 0.0105, -0.0104], device='cuda:0'), grad: tensor([ 0.0082, -0.0262, 0.0054, -0.0158, -0.0110, 0.0141, 0.0378, 0.0382, + -0.0649, 0.0143], device='cuda:0') +100 +0.0001 +changing lr +epoch 16, time 225.75, cls_loss 0.9449 cls_loss_mapping 0.1156 cls_loss_causal 0.9195 re_mapping 0.0298 re_causal 0.0817 /// teacc 97.40 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0086, 0.0140, -0.0566, ..., -0.0218, 0.0135, 0.0148], + [-0.0245, -0.0446, 0.0150, ..., -0.0223, -0.0520, 0.0455], + [-0.0237, -0.0067, -0.0398, ..., -0.0523, -0.0281, 0.0037], + ..., + [-0.0120, -0.0044, -0.0202, ..., -0.0532, 0.0135, 0.0156], + [-0.0524, -0.0273, -0.0296, ..., -0.0211, -0.0259, -0.0467], + [-0.0075, 0.0038, -0.0388, ..., -0.0153, 0.0261, -0.0027]], + device='cuda:0'), grad: tensor([[ 0.0121, 0.0026, 0.0003, ..., 0.0011, 0.0085, 0.0116], + [-0.0002, 0.0007, -0.0026, ..., 0.0008, 0.0021, 0.0042], + [-0.0014, 0.0005, 0.0009, ..., -0.0013, -0.0029, -0.0098], + ..., + [ 0.0068, -0.0046, 0.0021, ..., 0.0005, 0.0050, -0.0067], + [-0.0013, 0.0014, 0.0029, ..., 0.0018, 0.0011, 0.0001], + [ 0.0127, 0.0029, 0.0046, ..., 0.0021, 0.0035, 0.0091]], + device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0156, -0.0187, 0.0078, -0.0213, 0.0156, 0.0055, -0.0127, -0.0275, + 0.0107, -0.0104], device='cuda:0'), grad: tensor([ 0.0415, 0.0187, -0.0221, -0.0016, -0.0402, -0.0228, -0.0041, -0.0112, + 0.0062, 0.0356], device='cuda:0') +100 +0.0001 +changing lr +epoch 17, time 225.27, cls_loss 0.9093 cls_loss_mapping 0.1049 cls_loss_causal 0.8926 re_mapping 0.0297 re_causal 0.0816 /// teacc 97.43 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0080, 0.0146, -0.0576, ..., -0.0238, 0.0142, 0.0151], + [-0.0246, -0.0446, 0.0147, ..., -0.0240, -0.0523, 0.0459], + [-0.0243, -0.0063, -0.0405, ..., -0.0540, -0.0285, 0.0038], + ..., + [-0.0129, -0.0041, -0.0214, ..., -0.0532, 0.0127, 0.0154], + [-0.0529, -0.0283, -0.0299, ..., -0.0208, -0.0264, -0.0466], + [-0.0068, 0.0029, -0.0398, ..., -0.0179, 0.0270, -0.0034]], + device='cuda:0'), grad: tensor([[ 0.0073, -0.0035, -0.0037, ..., 0.0044, 0.0016, 0.0024], + [-0.0115, -0.0003, 0.0011, ..., -0.0008, -0.0125, -0.0478], + [-0.0046, 0.0059, 0.0037, ..., 0.0022, 0.0023, -0.0075], + ..., + [-0.0131, -0.0049, 0.0051, ..., 0.0023, -0.0063, 0.0369], + [-0.0042, 0.0066, 0.0081, ..., 0.0084, 0.0034, -0.0014], + [ 0.0305, 0.0116, 0.0034, ..., 0.0019, 0.0214, 0.0168]], + device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0159, -0.0180, 0.0079, -0.0212, 0.0157, 0.0056, -0.0133, -0.0282, + 0.0109, -0.0107], device='cuda:0'), grad: tensor([ 0.0170, -0.0806, -0.0154, 0.0252, 0.0003, -0.0368, -0.0540, 0.0687, + 0.0070, 0.0687], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 225.91, cls_loss 0.8699 cls_loss_mapping 0.1098 cls_loss_causal 0.8492 re_mapping 0.0300 re_causal 0.0798 /// teacc 97.12 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0079, 0.0159, -0.0590, ..., -0.0249, 0.0149, 0.0149], + [-0.0256, -0.0454, 0.0149, ..., -0.0248, -0.0533, 0.0457], + [-0.0249, -0.0065, -0.0402, ..., -0.0557, -0.0290, 0.0035], + ..., + [-0.0122, -0.0047, -0.0211, ..., -0.0514, 0.0128, 0.0164], + [-0.0533, -0.0292, -0.0305, ..., -0.0206, -0.0271, -0.0466], + [-0.0067, 0.0024, -0.0404, ..., -0.0200, 0.0272, -0.0031]], + device='cuda:0'), grad: tensor([[ 0.0105, 0.0017, 0.0006, ..., 0.0029, 0.0095, 0.0044], + [-0.0026, -0.0011, -0.0004, ..., -0.0004, 0.0019, -0.0042], + [-0.0127, 0.0006, 0.0005, ..., -0.0013, -0.0136, -0.0092], + ..., + [ 0.0011, 0.0010, 0.0005, ..., -0.0015, -0.0117, -0.0016], + [-0.0005, 0.0039, 0.0037, ..., 0.0017, -0.0008, -0.0008], + [-0.0033, 0.0005, 0.0014, ..., 0.0015, 0.0099, 0.0002]], + device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0158, -0.0185, 0.0078, -0.0212, 0.0159, 0.0054, -0.0130, -0.0275, + 0.0108, -0.0108], device='cuda:0'), grad: tensor([ 0.0370, -0.0248, -0.0338, 0.0138, 0.0138, -0.0146, -0.0113, -0.0061, + 0.0220, 0.0041], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 225.88, cls_loss 0.8763 cls_loss_mapping 0.1055 cls_loss_causal 0.8490 re_mapping 0.0282 re_causal 0.0795 /// teacc 97.45 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0078, 0.0167, -0.0599, ..., -0.0258, 0.0151, 0.0147], + [-0.0260, -0.0460, 0.0158, ..., -0.0234, -0.0536, 0.0452], + [-0.0248, -0.0066, -0.0404, ..., -0.0563, -0.0286, 0.0038], + ..., + [-0.0130, -0.0050, -0.0211, ..., -0.0526, 0.0121, 0.0175], + [-0.0533, -0.0306, -0.0315, ..., -0.0215, -0.0273, -0.0467], + [-0.0066, 0.0024, -0.0413, ..., -0.0195, 0.0270, -0.0037]], + device='cuda:0'), grad: tensor([[-0.0035, -0.0003, 0.0003, ..., 0.0020, 0.0023, -0.0117], + [ 0.0013, 0.0003, 0.0002, ..., 0.0004, 0.0021, -0.0008], + [ 0.0046, 0.0005, 0.0006, ..., -0.0006, -0.0016, 0.0154], + ..., + [-0.0050, 0.0034, 0.0038, ..., 0.0002, -0.0050, 0.0006], + [ 0.0020, 0.0019, 0.0028, ..., -0.0047, 0.0020, -0.0074], + [-0.0041, 0.0017, -0.0008, ..., 0.0005, 0.0006, -0.0108]], + device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0155, -0.0186, 0.0083, -0.0212, 0.0160, 0.0053, -0.0133, -0.0280, + 0.0111, -0.0105], device='cuda:0'), grad: tensor([-0.0264, -0.0042, 0.0605, -0.0080, -0.0212, 0.0308, 0.0315, -0.0431, + -0.0253, 0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 225.63, cls_loss 0.8962 cls_loss_mapping 0.1004 cls_loss_causal 0.8683 re_mapping 0.0268 re_causal 0.0761 /// teacc 97.56 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0075, 0.0179, -0.0602, ..., -0.0265, 0.0153, 0.0151], + [-0.0257, -0.0468, 0.0165, ..., -0.0239, -0.0536, 0.0458], + [-0.0246, -0.0071, -0.0420, ..., -0.0572, -0.0287, 0.0045], + ..., + [-0.0133, -0.0053, -0.0210, ..., -0.0529, 0.0119, 0.0176], + [-0.0541, -0.0310, -0.0327, ..., -0.0214, -0.0272, -0.0476], + [-0.0068, 0.0019, -0.0427, ..., -0.0206, 0.0271, -0.0035]], + device='cuda:0'), grad: tensor([[ 0.0124, 0.0019, 0.0014, ..., 0.0015, 0.0094, 0.0049], + [ 0.0063, 0.0005, -0.0003, ..., 0.0007, 0.0039, -0.0093], + [-0.0110, -0.0030, 0.0010, ..., -0.0004, -0.0071, -0.0123], + ..., + [ 0.0099, 0.0016, 0.0007, ..., 0.0005, 0.0076, 0.0152], + [-0.0119, -0.0015, -0.0037, ..., -0.0023, -0.0027, -0.0061], + [ 0.0246, 0.0090, 0.0012, ..., 0.0020, 0.0175, 0.0014]], + device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0161, -0.0182, 0.0082, -0.0211, 0.0160, 0.0054, -0.0134, -0.0280, + 0.0110, -0.0113], device='cuda:0'), grad: tensor([ 0.0624, -0.0184, -0.0512, 0.0252, -0.0371, 0.0562, -0.0449, 0.0522, + -0.0662, 0.0216], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 229.23, cls_loss 0.8133 cls_loss_mapping 0.0794 cls_loss_causal 0.7811 re_mapping 0.0266 re_causal 0.0722 /// teacc 97.75 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0070, 0.0201, -0.0609, ..., -0.0267, 0.0159, 0.0148], + [-0.0262, -0.0476, 0.0170, ..., -0.0246, -0.0545, 0.0458], + [-0.0249, -0.0080, -0.0414, ..., -0.0572, -0.0289, 0.0059], + ..., + [-0.0131, -0.0061, -0.0228, ..., -0.0546, 0.0119, 0.0183], + [-0.0544, -0.0310, -0.0327, ..., -0.0217, -0.0272, -0.0481], + [-0.0068, 0.0008, -0.0434, ..., -0.0217, 0.0273, -0.0039]], + device='cuda:0'), grad: tensor([[ 1.6449e-02, 4.8876e-04, 5.0753e-05, ..., 9.2983e-04, + 3.0422e-03, 2.7161e-03], + [ 1.0117e-02, 5.9319e-04, -1.0524e-06, ..., 7.9489e-04, + 1.1894e-02, 2.0004e-02], + [-5.1003e-03, -6.5079e-03, 4.7731e-04, ..., -5.5618e-03, + 6.8703e-03, -7.7095e-03], + ..., + [-4.7150e-02, 1.9407e-03, -8.2874e-04, ..., 4.1676e-04, + -3.8025e-02, -4.0131e-02], + [ 1.6880e-03, -3.5477e-03, -9.0301e-05, ..., 1.6603e-03, + 3.5419e-03, -3.4084e-03], + [ 1.3504e-02, 5.0116e-04, 1.9493e-03, ..., 2.3282e-04, + 9.8953e-03, 2.1408e-02]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0161, -0.0184, 0.0083, -0.0209, 0.0160, 0.0048, -0.0134, -0.0279, + 0.0111, -0.0112], device='cuda:0'), grad: tensor([ 0.0168, 0.0624, -0.0001, -0.0170, 0.0097, 0.0236, -0.0256, -0.1230, + 0.0002, 0.0531], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 226.60, cls_loss 0.8335 cls_loss_mapping 0.0865 cls_loss_causal 0.8031 re_mapping 0.0257 re_causal 0.0729 /// teacc 97.91 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0071, 0.0210, -0.0619, ..., -0.0271, 0.0160, 0.0142], + [-0.0262, -0.0468, 0.0170, ..., -0.0249, -0.0535, 0.0455], + [-0.0252, -0.0076, -0.0403, ..., -0.0582, -0.0291, 0.0067], + ..., + [-0.0132, -0.0064, -0.0238, ..., -0.0560, 0.0108, 0.0186], + [-0.0557, -0.0323, -0.0339, ..., -0.0221, -0.0284, -0.0483], + [-0.0064, 0.0010, -0.0432, ..., -0.0229, 0.0278, -0.0039]], + device='cuda:0'), grad: tensor([[ 0.0081, 0.0045, 0.0030, ..., 0.0004, 0.0055, 0.0032], + [ 0.0127, 0.0004, 0.0006, ..., 0.0007, 0.0073, 0.0178], + [ 0.0037, 0.0008, 0.0004, ..., 0.0002, 0.0026, 0.0003], + ..., + [-0.0048, -0.0013, -0.0022, ..., 0.0003, -0.0028, -0.0090], + [-0.0137, -0.0101, -0.0061, ..., -0.0061, -0.0112, -0.0061], + [ 0.0119, 0.0076, 0.0046, ..., 0.0053, 0.0106, 0.0031]], + device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0167, -0.0186, 0.0081, -0.0206, 0.0164, 0.0048, -0.0137, -0.0281, + 0.0107, -0.0111], device='cuda:0'), grad: tensor([ 0.0226, 0.0577, 0.0050, -0.0125, -0.0341, 0.0157, -0.0097, -0.0264, + -0.0468, 0.0285], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 226.02, cls_loss 0.8285 cls_loss_mapping 0.0836 cls_loss_causal 0.7926 re_mapping 0.0253 re_causal 0.0703 /// teacc 97.92 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0066, 0.0220, -0.0613, ..., -0.0263, 0.0169, 0.0137], + [-0.0275, -0.0482, 0.0167, ..., -0.0248, -0.0546, 0.0457], + [-0.0253, -0.0079, -0.0413, ..., -0.0600, -0.0294, 0.0073], + ..., + [-0.0136, -0.0061, -0.0245, ..., -0.0576, 0.0103, 0.0191], + [-0.0560, -0.0325, -0.0338, ..., -0.0216, -0.0291, -0.0484], + [-0.0057, 0.0025, -0.0426, ..., -0.0233, 0.0285, -0.0047]], + device='cuda:0'), grad: tensor([[ 8.1177e-03, 1.0657e-04, 1.1311e-03, ..., 3.4790e-03, + 9.9869e-03, 3.2749e-03], + [ 4.0474e-03, 6.1131e-04, 4.5252e-04, ..., 1.4019e-03, + 3.5934e-03, 2.0278e-04], + [-1.0624e-03, 8.3256e-04, 1.3199e-03, ..., 4.1223e-04, + 8.2254e-04, -3.3112e-03], + ..., + [-7.7486e-04, 8.9884e-04, 7.2813e-04, ..., -8.2064e-04, + -7.0524e-04, 4.5872e-04], + [-2.4548e-03, -2.0370e-03, -5.0240e-03, ..., -9.2077e-04, + -7.0801e-03, -1.5140e-05], + [-6.0539e-03, -1.2684e-03, -7.7724e-04, ..., -4.2653e-04, + -5.7220e-03, -2.0447e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0167, -0.0189, 0.0082, -0.0207, 0.0159, 0.0052, -0.0133, -0.0284, + 0.0110, -0.0112], device='cuda:0'), grad: tensor([ 3.3936e-02, 6.0558e-04, 1.8845e-02, 2.2156e-02, -2.9541e-02, + 4.0283e-02, -4.8096e-02, 1.5717e-03, -3.9734e-02, -9.3102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 225.62, cls_loss 0.8252 cls_loss_mapping 0.0758 cls_loss_causal 0.7962 re_mapping 0.0254 re_causal 0.0687 /// teacc 97.50 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0064, 0.0226, -0.0624, ..., -0.0260, 0.0170, 0.0136], + [-0.0273, -0.0475, 0.0178, ..., -0.0232, -0.0547, 0.0461], + [-0.0257, -0.0083, -0.0420, ..., -0.0608, -0.0297, 0.0077], + ..., + [-0.0137, -0.0066, -0.0255, ..., -0.0592, 0.0107, 0.0198], + [-0.0567, -0.0328, -0.0345, ..., -0.0231, -0.0294, -0.0500], + [-0.0055, 0.0017, -0.0435, ..., -0.0239, 0.0286, -0.0054]], + device='cuda:0'), grad: tensor([[ 1.6432e-03, -2.9707e-04, -6.6757e-04, ..., -2.0161e-03, + 3.2654e-03, 6.0425e-03], + [ 5.2567e-03, 1.1969e-03, -2.8682e-04, ..., 1.9445e-03, + 3.4885e-03, 4.9477e-03], + [-5.6305e-03, 5.8699e-04, -2.7442e-04, ..., -1.6794e-03, + -2.0523e-03, -1.4137e-02], + ..., + [ 1.5129e-02, 3.8738e-03, 2.6951e-03, ..., 4.0283e-03, + 8.3389e-03, 2.2079e-02], + [-1.3420e-02, 1.0223e-03, 1.3094e-03, ..., -5.7727e-05, + -2.9640e-03, -1.5541e-02], + [-2.8553e-03, -2.2411e-03, -3.2177e-03, ..., -1.1168e-03, + 1.1473e-03, 1.1673e-03]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0161, -0.0182, 0.0084, -0.0209, 0.0161, 0.0054, -0.0128, -0.0285, + 0.0106, -0.0115], device='cuda:0'), grad: tensor([ 0.0096, 0.0167, -0.0147, -0.0398, 0.0183, -0.0039, -0.0103, 0.0487, + -0.0207, -0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 225.37, cls_loss 0.8155 cls_loss_mapping 0.0903 cls_loss_causal 0.7812 re_mapping 0.0239 re_causal 0.0633 /// teacc 97.87 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0067, 0.0243, -0.0636, ..., -0.0266, 0.0178, 0.0143], + [-0.0272, -0.0479, 0.0185, ..., -0.0237, -0.0544, 0.0461], + [-0.0263, -0.0092, -0.0427, ..., -0.0617, -0.0305, 0.0072], + ..., + [-0.0139, -0.0069, -0.0256, ..., -0.0589, 0.0100, 0.0206], + [-0.0570, -0.0334, -0.0351, ..., -0.0240, -0.0298, -0.0506], + [-0.0051, 0.0017, -0.0436, ..., -0.0247, 0.0287, -0.0051]], + device='cuda:0'), grad: tensor([[-0.0207, -0.0075, 0.0004, ..., -0.0007, -0.0127, -0.0198], + [-0.0033, 0.0007, -0.0169, ..., -0.0058, -0.0011, -0.0104], + [ 0.0065, 0.0017, 0.0184, ..., 0.0064, 0.0056, 0.0182], + ..., + [ 0.0201, 0.0083, 0.0015, ..., 0.0004, 0.0172, 0.0178], + [-0.0004, -0.0118, -0.0267, ..., -0.0361, -0.0098, -0.0065], + [-0.0078, -0.0065, 0.0010, ..., 0.0003, -0.0089, 0.0003]], + device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0164, -0.0180, 0.0078, -0.0208, 0.0157, 0.0060, -0.0132, -0.0284, + 0.0108, -0.0115], device='cuda:0'), grad: tensor([-0.0526, -0.0471, 0.0571, 0.0060, 0.0337, 0.0634, -0.0573, 0.0620, + -0.0698, 0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 225.59, cls_loss 0.7922 cls_loss_mapping 0.0682 cls_loss_causal 0.7592 re_mapping 0.0249 re_causal 0.0685 /// teacc 97.76 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0061, 0.0244, -0.0652, ..., -0.0273, 0.0186, 0.0145], + [-0.0276, -0.0492, 0.0184, ..., -0.0247, -0.0548, 0.0458], + [-0.0264, -0.0083, -0.0435, ..., -0.0627, -0.0296, 0.0072], + ..., + [-0.0144, -0.0062, -0.0253, ..., -0.0586, 0.0097, 0.0210], + [-0.0573, -0.0328, -0.0355, ..., -0.0232, -0.0298, -0.0504], + [-0.0047, 0.0005, -0.0437, ..., -0.0253, 0.0282, -0.0046]], + device='cuda:0'), grad: tensor([[ 1.4658e-03, 2.2247e-05, 6.7186e-04, ..., 9.4843e-04, + -5.6982e-04, 2.8057e-03], + [ 1.2894e-03, 9.1410e-04, 9.7096e-05, ..., -6.5842e-03, + 1.9245e-03, 6.7101e-03], + [-1.1368e-02, 6.3181e-04, 1.5974e-04, ..., 2.6727e-04, + -2.8351e-02, -7.2060e-03], + ..., + [ 2.3994e-03, 6.4850e-03, 8.4066e-04, ..., 7.0477e-04, + 1.8311e-03, 4.5280e-03], + [ 3.9940e-03, -4.2839e-03, -5.3368e-03, ..., -1.2268e-02, + 1.2579e-03, 7.0534e-03], + [ 1.0239e-02, -7.7934e-03, -2.3580e-04, ..., -3.9339e-04, + 5.3406e-03, -1.6510e-02]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0163, -0.0180, 0.0076, -0.0210, 0.0155, 0.0058, -0.0131, -0.0281, + 0.0107, -0.0109], device='cuda:0'), grad: tensor([ 0.0210, 0.0060, -0.0276, 0.0422, -0.0042, -0.0419, 0.0494, 0.0031, + -0.0339, -0.0141], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 27---------------------------------------------------- +epoch 27, time 225.95, cls_loss 0.8272 cls_loss_mapping 0.0813 cls_loss_causal 0.7961 re_mapping 0.0245 re_causal 0.0681 /// teacc 97.94 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0061, 0.0249, -0.0661, ..., -0.0283, 0.0188, 0.0143], + [-0.0277, -0.0494, 0.0196, ..., -0.0244, -0.0554, 0.0463], + [-0.0269, -0.0083, -0.0442, ..., -0.0632, -0.0298, 0.0075], + ..., + [-0.0143, -0.0077, -0.0276, ..., -0.0593, 0.0094, 0.0220], + [-0.0578, -0.0340, -0.0358, ..., -0.0232, -0.0295, -0.0510], + [-0.0046, 0.0004, -0.0439, ..., -0.0264, 0.0280, -0.0047]], + device='cuda:0'), grad: tensor([[-0.0138, -0.0098, -0.0056, ..., 0.0003, -0.0134, 0.0023], + [-0.0171, -0.0027, 0.0004, ..., -0.0002, -0.0106, -0.0248], + [-0.0005, 0.0052, 0.0016, ..., 0.0044, -0.0018, -0.0039], + ..., + [ 0.0051, 0.0010, 0.0007, ..., 0.0005, 0.0036, 0.0059], + [-0.0058, 0.0027, 0.0040, ..., 0.0033, -0.0020, -0.0029], + [ 0.0131, 0.0018, 0.0008, ..., 0.0005, 0.0115, 0.0045]], + device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0160, -0.0176, 0.0073, -0.0204, 0.0151, 0.0054, -0.0129, -0.0279, + 0.0104, -0.0107], device='cuda:0'), grad: tensor([-0.0389, -0.0920, -0.0158, 0.0122, 0.0142, 0.0376, 0.0251, 0.0260, + -0.0116, 0.0433], device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 225.54, cls_loss 0.8088 cls_loss_mapping 0.0799 cls_loss_causal 0.7746 re_mapping 0.0244 re_causal 0.0643 /// teacc 97.68 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0061, 0.0260, -0.0667, ..., -0.0291, 0.0194, 0.0146], + [-0.0283, -0.0502, 0.0195, ..., -0.0258, -0.0564, 0.0460], + [-0.0269, -0.0102, -0.0453, ..., -0.0635, -0.0294, 0.0083], + ..., + [-0.0144, -0.0082, -0.0284, ..., -0.0601, 0.0093, 0.0229], + [-0.0579, -0.0351, -0.0361, ..., -0.0244, -0.0290, -0.0514], + [-0.0047, 0.0011, -0.0441, ..., -0.0269, 0.0279, -0.0055]], + device='cuda:0'), grad: tensor([[-0.0027, -0.0052, 0.0004, ..., -0.0015, -0.0004, 0.0007], + [-0.0288, 0.0007, -0.0033, ..., -0.0027, -0.0113, -0.0249], + [ 0.0098, 0.0022, 0.0061, ..., 0.0061, 0.0107, 0.0131], + ..., + [ 0.0117, 0.0053, 0.0040, ..., 0.0052, 0.0090, 0.0210], + [ 0.0070, -0.0059, -0.0126, ..., -0.0091, 0.0045, -0.0083], + [ 0.0016, 0.0035, 0.0033, ..., 0.0026, 0.0011, 0.0098]], + device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0159, -0.0179, 0.0070, -0.0204, 0.0158, 0.0056, -0.0133, -0.0282, + 0.0109, -0.0108], device='cuda:0'), grad: tensor([ 0.0020, -0.0900, 0.0465, -0.0049, -0.0489, 0.0517, -0.0304, 0.0629, + -0.0129, 0.0239], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 225.77, cls_loss 0.7933 cls_loss_mapping 0.0776 cls_loss_causal 0.7471 re_mapping 0.0245 re_causal 0.0664 /// teacc 97.91 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0058, 0.0270, -0.0674, ..., -0.0302, 0.0193, 0.0134], + [-0.0290, -0.0515, 0.0193, ..., -0.0261, -0.0576, 0.0460], + [-0.0273, -0.0100, -0.0452, ..., -0.0633, -0.0301, 0.0089], + ..., + [-0.0138, -0.0092, -0.0297, ..., -0.0608, 0.0097, 0.0233], + [-0.0582, -0.0365, -0.0376, ..., -0.0239, -0.0288, -0.0509], + [-0.0044, 0.0012, -0.0449, ..., -0.0270, 0.0275, -0.0065]], + device='cuda:0'), grad: tensor([[-0.0116, -0.0043, -0.0020, ..., -0.0001, -0.0160, -0.0028], + [ 0.0033, 0.0009, 0.0021, ..., 0.0006, 0.0022, 0.0002], + [ 0.0060, 0.0021, -0.0035, ..., -0.0024, 0.0053, -0.0019], + ..., + [ 0.0042, 0.0014, 0.0008, ..., 0.0002, 0.0033, 0.0058], + [-0.0084, -0.0026, -0.0074, ..., 0.0003, -0.0040, -0.0150], + [ 0.0024, 0.0008, 0.0005, ..., 0.0006, 0.0020, 0.0030]], + device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0153, -0.0178, 0.0067, -0.0198, 0.0158, 0.0057, -0.0132, -0.0277, + 0.0107, -0.0112], device='cuda:0'), grad: tensor([-0.0161, -0.0075, 0.0094, 0.0463, -0.0387, 0.0186, 0.0238, 0.0221, + -0.0705, 0.0126], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 225.80, cls_loss 0.7820 cls_loss_mapping 0.0686 cls_loss_causal 0.7498 re_mapping 0.0237 re_causal 0.0646 /// teacc 97.94 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0046, 0.0265, -0.0692, ..., -0.0308, 0.0215, 0.0136], + [-0.0293, -0.0521, 0.0200, ..., -0.0266, -0.0583, 0.0464], + [-0.0277, -0.0095, -0.0456, ..., -0.0636, -0.0308, 0.0102], + ..., + [-0.0139, -0.0094, -0.0306, ..., -0.0597, 0.0097, 0.0234], + [-0.0590, -0.0372, -0.0376, ..., -0.0243, -0.0302, -0.0513], + [-0.0041, 0.0014, -0.0440, ..., -0.0275, 0.0276, -0.0068]], + device='cuda:0'), grad: tensor([[ 4.0779e-03, 1.1277e-04, 1.4746e-04, ..., 1.0389e-04, + 4.1351e-03, 3.6259e-03], + [ 3.7575e-03, 6.4492e-05, 4.2343e-04, ..., 6.4993e-04, + 4.9629e-03, 6.3972e-03], + [-1.2680e-02, -6.3171e-03, -3.3970e-03, ..., 2.3866e-04, + -1.3603e-02, -1.3252e-02], + ..., + [ 3.3092e-03, -1.9550e-03, 6.4230e-04, ..., 5.5313e-04, + 1.1627e-02, -3.1776e-03], + [ 2.8744e-03, 1.1091e-03, 1.3256e-03, ..., 1.3533e-03, + 3.0174e-03, 4.2725e-03], + [-1.9140e-03, 1.7624e-03, 3.5787e-04, ..., 1.5793e-03, + -1.1475e-02, 2.9774e-03]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0157, -0.0175, 0.0066, -0.0200, 0.0152, 0.0064, -0.0136, -0.0282, + 0.0107, -0.0107], device='cuda:0'), grad: tensor([ 0.0128, 0.0213, -0.0682, 0.0133, 0.0346, 0.0056, -0.0500, 0.0100, + 0.0182, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 31---------------------------------------------------- +epoch 31, time 226.23, cls_loss 0.7772 cls_loss_mapping 0.0667 cls_loss_causal 0.7337 re_mapping 0.0238 re_causal 0.0652 /// teacc 98.02 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0044, 0.0271, -0.0698, ..., -0.0327, 0.0216, 0.0151], + [-0.0300, -0.0527, 0.0204, ..., -0.0267, -0.0591, 0.0459], + [-0.0289, -0.0101, -0.0459, ..., -0.0647, -0.0316, 0.0099], + ..., + [-0.0129, -0.0098, -0.0307, ..., -0.0600, 0.0102, 0.0245], + [-0.0596, -0.0373, -0.0380, ..., -0.0253, -0.0307, -0.0514], + [-0.0041, 0.0013, -0.0443, ..., -0.0286, 0.0278, -0.0076]], + device='cuda:0'), grad: tensor([[-0.0057, -0.0106, -0.0024, ..., -0.0066, -0.0021, 0.0033], + [ 0.0087, 0.0006, 0.0020, ..., 0.0033, 0.0118, 0.0091], + [-0.0116, 0.0004, 0.0006, ..., 0.0013, -0.0053, -0.0182], + ..., + [-0.0162, 0.0001, -0.0042, ..., -0.0034, -0.0140, -0.0125], + [-0.0031, -0.0027, 0.0005, ..., -0.0100, -0.0090, -0.0017], + [ 0.0056, 0.0002, 0.0004, ..., -0.0025, 0.0008, 0.0059]], + device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0163, -0.0175, 0.0058, -0.0198, 0.0150, 0.0063, -0.0137, -0.0275, + 0.0106, -0.0110], device='cuda:0'), grad: tensor([-0.0060, 0.0501, -0.0617, 0.0133, 0.0076, 0.0260, 0.0471, -0.0594, + -0.0145, -0.0026], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 225.92, cls_loss 0.7243 cls_loss_mapping 0.0626 cls_loss_causal 0.6857 re_mapping 0.0222 re_causal 0.0574 /// teacc 98.25 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0041, 0.0287, -0.0698, ..., -0.0333, 0.0221, 0.0152], + [-0.0302, -0.0525, 0.0215, ..., -0.0261, -0.0598, 0.0455], + [-0.0290, -0.0108, -0.0461, ..., -0.0649, -0.0319, 0.0106], + ..., + [-0.0132, -0.0118, -0.0321, ..., -0.0609, 0.0101, 0.0245], + [-0.0590, -0.0376, -0.0379, ..., -0.0248, -0.0300, -0.0514], + [-0.0043, 0.0018, -0.0442, ..., -0.0280, 0.0277, -0.0082]], + device='cuda:0'), grad: tensor([[ 0.0041, 0.0016, 0.0003, ..., 0.0002, 0.0059, 0.0029], + [-0.0012, -0.0093, -0.0012, ..., -0.0035, -0.0025, -0.0030], + [-0.0052, 0.0015, 0.0001, ..., 0.0009, -0.0055, -0.0068], + ..., + [-0.0054, 0.0002, 0.0002, ..., 0.0003, -0.0031, -0.0023], + [-0.0312, 0.0058, 0.0019, ..., 0.0009, -0.0200, -0.0015], + [-0.0261, -0.0077, -0.0038, ..., -0.0015, -0.0210, -0.0046]], + device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0160, -0.0176, 0.0058, -0.0196, 0.0156, 0.0063, -0.0139, -0.0280, + 0.0113, -0.0113], device='cuda:0'), grad: tensor([ 0.0027, -0.0438, -0.0206, 0.0305, 0.0936, 0.0198, 0.0185, -0.0105, + -0.0299, -0.0604], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 224.88, cls_loss 0.7680 cls_loss_mapping 0.0642 cls_loss_causal 0.7380 re_mapping 0.0210 re_causal 0.0575 /// teacc 98.09 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0036, 0.0296, -0.0701, ..., -0.0336, 0.0231, 0.0152], + [-0.0311, -0.0519, 0.0224, ..., -0.0263, -0.0600, 0.0454], + [-0.0289, -0.0117, -0.0469, ..., -0.0653, -0.0316, 0.0111], + ..., + [-0.0125, -0.0112, -0.0322, ..., -0.0621, 0.0105, 0.0248], + [-0.0600, -0.0383, -0.0390, ..., -0.0253, -0.0306, -0.0515], + [-0.0041, 0.0012, -0.0438, ..., -0.0286, 0.0276, -0.0087]], + device='cuda:0'), grad: tensor([[-0.0317, -0.0069, -0.0040, ..., -0.0067, -0.0256, -0.0102], + [ 0.0034, -0.0048, 0.0012, ..., 0.0024, -0.0030, 0.0013], + [ 0.0093, 0.0040, 0.0017, ..., 0.0016, 0.0088, 0.0041], + ..., + [ 0.0079, 0.0012, 0.0006, ..., 0.0004, 0.0053, 0.0035], + [ 0.0027, -0.0037, -0.0019, ..., -0.0016, -0.0012, 0.0020], + [ 0.0205, 0.0039, 0.0022, ..., 0.0011, 0.0118, 0.0052]], + device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0153, -0.0179, 0.0063, -0.0196, 0.0158, 0.0063, -0.0136, -0.0283, + 0.0112, -0.0112], device='cuda:0'), grad: tensor([-0.0543, -0.0015, 0.0439, -0.0051, -0.0599, -0.0221, 0.0282, 0.0331, + -0.0211, 0.0590], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 225.00, cls_loss 0.7714 cls_loss_mapping 0.0624 cls_loss_causal 0.7344 re_mapping 0.0213 re_causal 0.0560 /// teacc 98.24 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0033, 0.0308, -0.0704, ..., -0.0342, 0.0239, 0.0147], + [-0.0318, -0.0534, 0.0229, ..., -0.0263, -0.0610, 0.0456], + [-0.0292, -0.0128, -0.0475, ..., -0.0651, -0.0318, 0.0111], + ..., + [-0.0118, -0.0115, -0.0334, ..., -0.0640, 0.0103, 0.0251], + [-0.0604, -0.0387, -0.0396, ..., -0.0259, -0.0308, -0.0524], + [-0.0043, 0.0017, -0.0434, ..., -0.0288, 0.0282, -0.0087]], + device='cuda:0'), grad: tensor([[-1.0338e-02, -7.8125e-03, -3.2539e-03, ..., -4.6463e-03, + -1.3573e-02, -1.2207e-04], + [ 3.0117e-03, 1.8799e-04, -7.7367e-05, ..., 3.4571e-04, + 2.2888e-03, 1.0948e-03], + [-5.4817e-03, 1.2188e-03, 1.2617e-03, ..., 1.1387e-03, + -2.3689e-03, -5.6610e-03], + ..., + [-5.1956e-03, 2.9254e-04, -6.2037e-04, ..., 4.2820e-04, + -6.2141e-03, -3.7060e-03], + [ 3.5362e-03, 8.4972e-04, 3.1471e-04, ..., 4.4155e-04, + 2.7084e-03, 1.1358e-03], + [-7.9536e-04, 7.0620e-04, -2.2602e-03, ..., -3.1281e-03, + 5.0507e-03, -7.6628e-04]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0153, -0.0177, 0.0065, -0.0193, 0.0153, 0.0064, -0.0130, -0.0281, + 0.0109, -0.0117], device='cuda:0'), grad: tensor([-0.0172, 0.0145, -0.0085, 0.0205, -0.0084, 0.0116, 0.0177, -0.0285, + 0.0152, -0.0169], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 225.17, cls_loss 0.7618 cls_loss_mapping 0.0704 cls_loss_causal 0.7228 re_mapping 0.0212 re_causal 0.0575 /// teacc 98.15 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0028, 0.0314, -0.0712, ..., -0.0339, 0.0246, 0.0143], + [-0.0320, -0.0531, 0.0237, ..., -0.0267, -0.0610, 0.0452], + [-0.0296, -0.0118, -0.0482, ..., -0.0671, -0.0319, 0.0110], + ..., + [-0.0118, -0.0121, -0.0339, ..., -0.0643, 0.0101, 0.0259], + [-0.0609, -0.0392, -0.0400, ..., -0.0260, -0.0310, -0.0532], + [-0.0042, 0.0015, -0.0438, ..., -0.0290, 0.0281, -0.0087]], + device='cuda:0'), grad: tensor([[ 1.9470e-02, 3.9101e-03, 2.2926e-03, ..., 3.5629e-03, + 5.9738e-03, 5.8823e-03], + [ 4.5037e-04, 4.8566e-04, 1.8680e-04, ..., 6.5029e-05, + 1.3628e-03, -2.5864e-03], + [ 7.6180e-03, 5.8413e-04, 3.7098e-04, ..., 3.1853e-04, + 1.6165e-03, 4.6463e-03], + ..., + [ 1.9730e-02, 1.3294e-03, 4.5800e-04, ..., 1.3506e-04, + 6.6071e-03, 1.3252e-02], + [-5.8289e-03, 1.7443e-03, 1.0424e-03, ..., 3.3307e-04, + 7.9751e-05, -9.9869e-03], + [ 2.5063e-03, -6.6376e-03, -5.9128e-03, ..., -2.3327e-03, + 4.1842e-04, 4.4022e-03]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0154, -0.0182, 0.0066, -0.0194, 0.0153, 0.0064, -0.0131, -0.0274, + 0.0108, -0.0117], device='cuda:0'), grad: tensor([ 0.0547, -0.0019, 0.0210, 0.0098, -0.0249, -0.0014, -0.0874, 0.0580, + -0.0333, 0.0053], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 36---------------------------------------------------- +epoch 36, time 226.70, cls_loss 0.7695 cls_loss_mapping 0.0572 cls_loss_causal 0.7248 re_mapping 0.0201 re_causal 0.0547 /// teacc 98.39 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0028, 0.0322, -0.0720, ..., -0.0343, 0.0244, 0.0132], + [-0.0316, -0.0540, 0.0249, ..., -0.0255, -0.0602, 0.0463], + [-0.0300, -0.0122, -0.0496, ..., -0.0700, -0.0321, 0.0118], + ..., + [-0.0122, -0.0123, -0.0344, ..., -0.0655, 0.0101, 0.0257], + [-0.0611, -0.0397, -0.0402, ..., -0.0260, -0.0314, -0.0538], + [-0.0043, 0.0015, -0.0439, ..., -0.0295, 0.0279, -0.0089]], + device='cuda:0'), grad: tensor([[-1.3748e-02, -2.3425e-04, -3.9279e-05, ..., 4.7255e-04, + -4.3411e-03, -1.5755e-03], + [-8.4698e-05, 1.2722e-03, 1.5106e-03, ..., -5.0392e-03, + 3.1548e-03, 6.3019e-03], + [ 4.0703e-03, -7.9107e-04, -5.3358e-04, ..., -3.1757e-04, + 7.1411e-03, -6.0201e-05], + ..., + [ 4.3221e-03, 7.8630e-04, 2.8000e-03, ..., 1.3132e-03, + -4.3035e-04, -1.1629e-04], + [ 6.9885e-03, 8.3113e-04, 1.2293e-03, ..., 1.4505e-03, + 3.8929e-03, 6.6376e-03], + [ 3.5095e-03, 1.4567e-04, 1.2970e-03, ..., 6.7520e-04, + -6.4240e-03, 5.3835e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0148, -0.0174, 0.0060, -0.0189, 0.0158, 0.0060, -0.0131, -0.0277, + 0.0107, -0.0117], device='cuda:0'), grad: tensor([-0.0197, 0.0012, 0.0370, -0.0153, -0.0024, -0.0675, 0.0312, 0.0058, + 0.0327, -0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 225.79, cls_loss 0.7430 cls_loss_mapping 0.0592 cls_loss_causal 0.7026 re_mapping 0.0201 re_causal 0.0532 /// teacc 98.10 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0022, 0.0326, -0.0722, ..., -0.0347, 0.0252, 0.0135], + [-0.0313, -0.0543, 0.0279, ..., -0.0249, -0.0601, 0.0465], + [-0.0293, -0.0132, -0.0511, ..., -0.0705, -0.0320, 0.0124], + ..., + [-0.0131, -0.0129, -0.0350, ..., -0.0667, 0.0091, 0.0255], + [-0.0618, -0.0399, -0.0411, ..., -0.0269, -0.0312, -0.0532], + [-0.0033, 0.0016, -0.0449, ..., -0.0295, 0.0282, -0.0080]], + device='cuda:0'), grad: tensor([[ 0.0054, 0.0038, 0.0016, ..., 0.0006, 0.0031, 0.0022], + [ 0.0035, 0.0005, 0.0003, ..., 0.0005, 0.0018, 0.0017], + [ 0.0059, 0.0029, 0.0029, ..., 0.0030, 0.0070, 0.0043], + ..., + [ 0.0108, 0.0023, 0.0032, ..., 0.0002, 0.0072, 0.0091], + [ 0.0052, 0.0019, 0.0020, ..., 0.0010, 0.0028, 0.0025], + [ 0.0012, 0.0074, 0.0062, ..., 0.0007, 0.0054, -0.0141]], + device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0147, -0.0168, 0.0062, -0.0189, 0.0154, 0.0057, -0.0135, -0.0279, + 0.0103, -0.0107], device='cuda:0'), grad: tensor([ 0.0177, 0.0134, 0.0366, -0.0025, -0.0501, -0.0311, -0.0173, 0.0395, + 0.0411, -0.0473], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 225.58, cls_loss 0.7677 cls_loss_mapping 0.0598 cls_loss_causal 0.7347 re_mapping 0.0212 re_causal 0.0578 /// teacc 98.15 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0027, 0.0327, -0.0751, ..., -0.0362, 0.0249, 0.0136], + [-0.0323, -0.0553, 0.0271, ..., -0.0264, -0.0605, 0.0458], + [-0.0309, -0.0127, -0.0489, ..., -0.0703, -0.0335, 0.0128], + ..., + [-0.0125, -0.0124, -0.0339, ..., -0.0669, 0.0102, 0.0262], + [-0.0615, -0.0403, -0.0417, ..., -0.0269, -0.0302, -0.0540], + [-0.0035, 0.0011, -0.0454, ..., -0.0307, 0.0275, -0.0082]], + device='cuda:0'), grad: tensor([[ 0.0043, 0.0003, 0.0009, ..., 0.0010, 0.0058, 0.0036], + [ 0.0005, 0.0002, 0.0052, ..., 0.0067, -0.0086, 0.0030], + [ 0.0050, 0.0009, 0.0014, ..., 0.0009, 0.0020, 0.0039], + ..., + [-0.0094, -0.0025, -0.0051, ..., 0.0005, -0.0001, -0.0121], + [-0.0002, -0.0118, -0.0104, ..., -0.0217, 0.0018, -0.0027], + [ 0.0028, 0.0023, 0.0043, ..., 0.0010, 0.0010, 0.0047]], + device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0145, -0.0179, 0.0060, -0.0184, 0.0155, 0.0052, -0.0127, -0.0273, + 0.0106, -0.0110], device='cuda:0'), grad: tensor([ 0.0321, -0.0007, 0.0279, 0.0447, -0.0114, 0.0288, -0.0474, -0.0421, + -0.0436, 0.0117], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 225.75, cls_loss 0.7612 cls_loss_mapping 0.0577 cls_loss_causal 0.7210 re_mapping 0.0201 re_causal 0.0522 /// teacc 98.33 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0023, 0.0338, -0.0752, ..., -0.0364, 0.0256, 0.0143], + [-0.0330, -0.0554, 0.0279, ..., -0.0272, -0.0610, 0.0458], + [-0.0310, -0.0129, -0.0490, ..., -0.0696, -0.0329, 0.0133], + ..., + [-0.0127, -0.0126, -0.0350, ..., -0.0670, 0.0098, 0.0258], + [-0.0620, -0.0407, -0.0424, ..., -0.0271, -0.0308, -0.0551], + [-0.0039, 0.0012, -0.0463, ..., -0.0311, 0.0278, -0.0073]], + device='cuda:0'), grad: tensor([[-0.0010, -0.0008, -0.0006, ..., 0.0028, 0.0005, 0.0022], + [ 0.0052, -0.0007, -0.0030, ..., -0.0008, 0.0010, 0.0026], + [ 0.0059, 0.0029, 0.0030, ..., 0.0025, 0.0044, -0.0030], + ..., + [ 0.0201, 0.0016, 0.0011, ..., 0.0015, 0.0061, 0.0142], + [ 0.0063, 0.0020, 0.0025, ..., 0.0018, 0.0025, 0.0051], + [-0.0251, -0.0075, -0.0065, ..., 0.0016, -0.0070, -0.0101]], + device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0145, -0.0177, 0.0058, -0.0175, 0.0155, 0.0051, -0.0127, -0.0275, + 0.0100, -0.0110], device='cuda:0'), grad: tensor([ 0.0093, 0.0278, 0.0132, -0.0050, 0.0186, -0.0554, -0.0304, 0.0610, + 0.0333, -0.0724], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 225.67, cls_loss 0.7558 cls_loss_mapping 0.0498 cls_loss_causal 0.7188 re_mapping 0.0202 re_causal 0.0547 /// teacc 98.35 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0012, 0.0347, -0.0752, ..., -0.0372, 0.0265, 0.0150], + [-0.0326, -0.0551, 0.0280, ..., -0.0267, -0.0614, 0.0459], + [-0.0323, -0.0133, -0.0496, ..., -0.0705, -0.0345, 0.0131], + ..., + [-0.0121, -0.0135, -0.0358, ..., -0.0683, 0.0095, 0.0270], + [-0.0620, -0.0407, -0.0421, ..., -0.0266, -0.0304, -0.0564], + [-0.0035, 0.0017, -0.0463, ..., -0.0311, 0.0283, -0.0081]], + device='cuda:0'), grad: tensor([[ 9.1171e-03, 4.4918e-04, 8.2541e-04, ..., 7.1001e-04, + 3.2253e-03, 5.4893e-03], + [ 5.1117e-03, 1.4305e-04, 6.4254e-05, ..., 2.7704e-04, + 2.3613e-03, 3.6182e-03], + [ 3.4962e-03, 2.4140e-04, 5.0735e-04, ..., -2.0523e-03, + 1.7452e-03, -5.7755e-03], + ..., + [-2.7733e-03, -9.9659e-05, -4.0245e-04, ..., 2.8539e-04, + -1.9474e-03, -8.3494e-04], + [-1.0765e-02, 5.8508e-04, 9.8515e-04, ..., 1.3285e-03, + -5.4932e-03, -1.0971e-02], + [-9.5081e-04, 3.7432e-04, 8.2588e-04, ..., 3.1066e-04, + -1.7452e-03, -1.2293e-03]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0147, -0.0175, 0.0057, -0.0180, 0.0149, 0.0053, -0.0123, -0.0271, + 0.0102, -0.0113], device='cuda:0'), grad: tensor([ 0.0567, 0.0269, -0.0069, -0.0057, 0.0025, 0.0253, -0.0074, 0.0035, + -0.0793, -0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 225.04, cls_loss 0.7385 cls_loss_mapping 0.0597 cls_loss_causal 0.7056 re_mapping 0.0185 re_causal 0.0504 /// teacc 98.39 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0015, 0.0352, -0.0762, ..., -0.0383, 0.0265, 0.0146], + [-0.0338, -0.0560, 0.0283, ..., -0.0264, -0.0622, 0.0462], + [-0.0321, -0.0134, -0.0490, ..., -0.0693, -0.0338, 0.0135], + ..., + [-0.0125, -0.0139, -0.0338, ..., -0.0693, 0.0092, 0.0275], + [-0.0615, -0.0420, -0.0422, ..., -0.0274, -0.0306, -0.0568], + [-0.0025, 0.0015, -0.0470, ..., -0.0307, 0.0285, -0.0082]], + device='cuda:0'), grad: tensor([[-3.0422e-03, -3.2043e-03, 4.0770e-04, ..., -2.2030e-03, + -5.1117e-03, -4.7455e-03], + [ 3.6755e-03, 3.1710e-04, 4.3774e-04, ..., 9.8896e-04, + 2.2697e-03, 3.4103e-03], + [ 4.3106e-03, 7.0620e-04, 2.0962e-03, ..., 3.8967e-03, + 5.1422e-03, 7.4844e-03], + ..., + [ 4.9858e-03, 2.5773e-04, 9.5427e-05, ..., 3.8934e-04, + 4.0970e-03, -1.4477e-03], + [ 1.2169e-03, 7.0381e-04, 5.7459e-04, ..., 1.3189e-03, + 1.4629e-03, 3.6774e-03], + [-1.9897e-02, -9.3384e-03, 6.0737e-05, ..., 4.1676e-04, + -4.3335e-02, 2.8210e-03]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0141, -0.0175, 0.0059, -0.0175, 0.0150, 0.0052, -0.0129, -0.0272, + 0.0104, -0.0110], device='cuda:0'), grad: tensor([-0.0696, 0.0226, 0.0562, -0.0015, 0.0107, -0.0063, -0.0212, 0.0141, + 0.0224, -0.0273], device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 225.55, cls_loss 0.7251 cls_loss_mapping 0.0543 cls_loss_causal 0.6886 re_mapping 0.0190 re_causal 0.0519 /// teacc 98.12 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0016, 0.0351, -0.0772, ..., -0.0380, 0.0266, 0.0145], + [-0.0339, -0.0554, 0.0296, ..., -0.0273, -0.0627, 0.0463], + [-0.0318, -0.0138, -0.0504, ..., -0.0701, -0.0339, 0.0147], + ..., + [-0.0129, -0.0140, -0.0341, ..., -0.0697, 0.0096, 0.0272], + [-0.0626, -0.0427, -0.0440, ..., -0.0271, -0.0313, -0.0577], + [-0.0024, 0.0018, -0.0461, ..., -0.0304, 0.0288, -0.0063]], + device='cuda:0'), grad: tensor([[-2.6718e-02, 1.5283e-04, -2.0528e-04, ..., 6.0654e-04, + -4.6021e-02, -3.6097e-04], + [ 5.9547e-03, 1.0341e-04, 2.5734e-05, ..., 4.3654e-04, + 7.6828e-03, 1.4410e-03], + [-4.4274e-04, 5.2977e-04, 8.1301e-04, ..., 1.1730e-03, + -1.7185e-03, -3.8528e-03], + ..., + [ 6.1493e-03, 4.6873e-04, 8.3923e-04, ..., 6.2895e-04, + 4.7607e-03, 3.2330e-03], + [ 3.3512e-03, 1.1021e-04, -7.5417e-03, ..., 7.8201e-04, + 1.4839e-03, 2.0084e-03], + [-8.4381e-03, -1.6689e-03, -6.5136e-04, ..., 6.6328e-04, + -1.4677e-03, -7.6370e-03]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0138, -0.0175, 0.0057, -0.0176, 0.0145, 0.0061, -0.0134, -0.0274, + 0.0101, -0.0099], device='cuda:0'), grad: tensor([-0.0635, 0.0260, -0.0121, 0.0131, 0.0211, -0.0116, 0.0088, 0.0264, + 0.0118, -0.0200], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 225.39, cls_loss 0.7073 cls_loss_mapping 0.0509 cls_loss_causal 0.6729 re_mapping 0.0190 re_causal 0.0497 /// teacc 98.29 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0010, 0.0355, -0.0778, ..., -0.0377, 0.0277, 0.0146], + [-0.0347, -0.0558, 0.0307, ..., -0.0269, -0.0633, 0.0464], + [-0.0316, -0.0134, -0.0508, ..., -0.0710, -0.0344, 0.0153], + ..., + [-0.0125, -0.0147, -0.0352, ..., -0.0703, 0.0089, 0.0278], + [-0.0628, -0.0430, -0.0447, ..., -0.0278, -0.0320, -0.0573], + [-0.0019, 0.0020, -0.0457, ..., -0.0305, 0.0287, -0.0065]], + device='cuda:0'), grad: tensor([[ 3.6678e-03, -3.6359e-05, 8.9693e-04, ..., -3.8123e-04, + 4.9019e-04, 4.6959e-03], + [-2.9254e-04, 8.9228e-05, 2.3365e-03, ..., 2.8572e-03, + -9.9957e-05, -1.0605e-03], + [ 5.7983e-03, 7.3290e-04, 2.1667e-03, ..., 2.1019e-03, + 3.9330e-03, 5.1575e-03], + ..., + [-1.5160e-02, 2.6727e-04, -2.6550e-03, ..., -5.0688e-04, + -2.3403e-03, -1.7868e-02], + [-5.6152e-03, 4.7994e-04, -8.8806e-03, ..., -1.1002e-02, + -2.0778e-04, -2.2030e-04], + [-1.3371e-03, -2.2197e-04, 8.6451e-04, ..., 5.1546e-04, + -2.9297e-03, -7.9918e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0142, -0.0176, 0.0058, -0.0185, 0.0141, 0.0058, -0.0127, -0.0265, + 0.0097, -0.0097], device='cuda:0'), grad: tensor([ 0.0161, -0.0032, 0.0218, 0.0146, -0.0175, 0.0132, 0.0323, -0.0465, + -0.0398, 0.0091], device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 225.47, cls_loss 0.7125 cls_loss_mapping 0.0564 cls_loss_causal 0.6741 re_mapping 0.0179 re_causal 0.0480 /// teacc 98.15 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0016, 0.0365, -0.0783, ..., -0.0375, 0.0271, 0.0133], + [-0.0355, -0.0568, 0.0314, ..., -0.0273, -0.0648, 0.0463], + [-0.0318, -0.0137, -0.0509, ..., -0.0721, -0.0338, 0.0165], + ..., + [-0.0118, -0.0153, -0.0357, ..., -0.0706, 0.0087, 0.0274], + [-0.0628, -0.0425, -0.0445, ..., -0.0279, -0.0321, -0.0576], + [-0.0021, 0.0016, -0.0469, ..., -0.0315, 0.0294, -0.0072]], + device='cuda:0'), grad: tensor([[ 4.9734e-04, -2.0199e-03, -2.4748e-04, ..., 5.0259e-04, + -5.3883e-04, 1.6117e-03], + [ 4.8828e-03, 3.5143e-04, 4.6730e-05, ..., 6.2227e-04, + 4.4594e-03, 4.8180e-03], + [ 3.4847e-03, 4.0817e-04, 1.0700e-03, ..., 1.5402e-03, + 3.1376e-03, 3.6144e-03], + ..., + [ 1.3809e-03, 2.7895e-04, -1.1539e-03, ..., -2.7313e-03, + 2.0485e-03, 3.4790e-03], + [-1.5354e-03, 9.0265e-04, -5.5265e-04, ..., -4.1237e-03, + -5.2147e-03, -9.3842e-03], + [-9.7122e-03, -1.3685e-04, -2.6436e-03, ..., 1.8559e-03, + -4.9744e-03, -8.3389e-03]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0139, -0.0175, 0.0057, -0.0182, 0.0144, 0.0057, -0.0125, -0.0267, + 0.0098, -0.0101], device='cuda:0'), grad: tensor([ 0.0123, 0.0241, 0.0141, 0.0043, 0.0242, 0.0349, -0.0269, 0.0031, + -0.0406, -0.0497], device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 225.57, cls_loss 0.7316 cls_loss_mapping 0.0531 cls_loss_causal 0.6984 re_mapping 0.0188 re_causal 0.0499 /// teacc 98.39 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0010, 0.0375, -0.0799, ..., -0.0376, 0.0280, 0.0133], + [-0.0361, -0.0565, 0.0322, ..., -0.0285, -0.0639, 0.0475], + [-0.0320, -0.0140, -0.0498, ..., -0.0726, -0.0342, 0.0168], + ..., + [-0.0108, -0.0165, -0.0368, ..., -0.0704, 0.0094, 0.0272], + [-0.0636, -0.0437, -0.0454, ..., -0.0289, -0.0326, -0.0576], + [-0.0026, 0.0010, -0.0471, ..., -0.0336, 0.0291, -0.0078]], + device='cuda:0'), grad: tensor([[ 0.0053, 0.0003, 0.0007, ..., 0.0013, 0.0046, 0.0038], + [ 0.0066, 0.0003, 0.0015, ..., 0.0017, 0.0058, 0.0070], + [-0.0056, 0.0006, -0.0008, ..., 0.0019, -0.0069, -0.0110], + ..., + [ 0.0048, 0.0008, 0.0010, ..., 0.0010, 0.0045, 0.0040], + [-0.0142, 0.0005, -0.0026, ..., -0.0006, -0.0070, -0.0056], + [-0.0004, -0.0009, 0.0007, ..., 0.0008, -0.0085, -0.0031]], + device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0145, -0.0165, 0.0055, -0.0180, 0.0148, 0.0057, -0.0136, -0.0266, + 0.0093, -0.0106], device='cuda:0'), grad: tensor([ 0.0285, 0.0391, -0.0538, 0.0226, -0.0114, 0.0191, 0.0040, 0.0239, + -0.0456, -0.0265], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 46---------------------------------------------------- +epoch 46, time 228.51, cls_loss 0.7483 cls_loss_mapping 0.0534 cls_loss_causal 0.7146 re_mapping 0.0186 re_causal 0.0488 /// teacc 98.44 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0013, 0.0376, -0.0818, ..., -0.0391, 0.0282, 0.0130], + [-0.0357, -0.0576, 0.0328, ..., -0.0283, -0.0643, 0.0473], + [-0.0320, -0.0148, -0.0505, ..., -0.0723, -0.0345, 0.0171], + ..., + [-0.0112, -0.0158, -0.0366, ..., -0.0699, 0.0094, 0.0282], + [-0.0650, -0.0439, -0.0451, ..., -0.0307, -0.0338, -0.0578], + [-0.0021, 0.0012, -0.0467, ..., -0.0332, 0.0297, -0.0084]], + device='cuda:0'), grad: tensor([[-0.0014, -0.0022, 0.0015, ..., 0.0009, -0.0060, 0.0021], + [ 0.0008, -0.0001, 0.0004, ..., -0.0057, -0.0008, -0.0079], + [ 0.0035, 0.0005, 0.0075, ..., 0.0042, 0.0018, 0.0049], + ..., + [-0.0164, -0.0024, 0.0020, ..., 0.0004, -0.0031, -0.0125], + [ 0.0004, 0.0007, -0.0027, ..., -0.0006, 0.0020, 0.0002], + [ 0.0092, 0.0003, -0.0035, ..., 0.0006, -0.0017, 0.0125]], + device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0144, -0.0166, 0.0058, -0.0181, 0.0149, 0.0062, -0.0139, -0.0267, + 0.0086, -0.0102], device='cuda:0'), grad: tensor([ 0.0036, -0.0272, 0.0426, 0.0113, 0.0028, 0.0206, -0.0374, -0.0305, + -0.0027, 0.0169], device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 225.16, cls_loss 0.6906 cls_loss_mapping 0.0500 cls_loss_causal 0.6568 re_mapping 0.0184 re_causal 0.0466 /// teacc 98.42 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0008, 0.0370, -0.0839, ..., -0.0409, 0.0292, 0.0127], + [-0.0358, -0.0593, 0.0327, ..., -0.0290, -0.0643, 0.0471], + [-0.0325, -0.0149, -0.0510, ..., -0.0724, -0.0353, 0.0171], + ..., + [-0.0100, -0.0150, -0.0355, ..., -0.0691, 0.0105, 0.0284], + [-0.0647, -0.0440, -0.0452, ..., -0.0302, -0.0326, -0.0577], + [-0.0021, 0.0010, -0.0469, ..., -0.0334, 0.0294, -0.0087]], + device='cuda:0'), grad: tensor([[-2.1915e-03, -2.3413e-04, 6.8247e-05, ..., -3.4857e-04, + 1.1787e-02, -5.2605e-03], + [ 3.9864e-03, 4.4018e-05, 7.4096e-06, ..., 7.2718e-05, + 2.6245e-03, 3.0022e-03], + [ 3.2578e-03, 3.6931e-04, 2.5797e-04, ..., 3.0899e-04, + 4.4479e-03, 4.3755e-03], + ..., + [ 2.7084e-03, 2.0301e-04, 2.5630e-04, ..., 2.5439e-04, + 2.3079e-03, 2.3460e-03], + [ 2.1992e-03, 4.8429e-05, -1.2502e-05, ..., 2.5725e-04, + -7.2708e-03, 4.4632e-03], + [-1.3039e-02, 1.1951e-04, 1.3769e-04, ..., 1.3685e-04, + -2.2568e-02, 1.8682e-03]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0146, -0.0170, 0.0055, -0.0179, 0.0147, 0.0058, -0.0132, -0.0267, + 0.0091, -0.0103], device='cuda:0'), grad: tensor([-0.0215, 0.0143, 0.0160, -0.0078, -0.0056, -0.0190, 0.0257, 0.0116, + 0.0039, -0.0177], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 48---------------------------------------------------- +epoch 48, time 228.34, cls_loss 0.7514 cls_loss_mapping 0.0458 cls_loss_causal 0.7087 re_mapping 0.0180 re_causal 0.0483 /// teacc 98.50 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0007, 0.0377, -0.0842, ..., -0.0412, 0.0295, 0.0120], + [-0.0360, -0.0600, 0.0329, ..., -0.0283, -0.0645, 0.0464], + [-0.0321, -0.0154, -0.0512, ..., -0.0729, -0.0356, 0.0181], + ..., + [-0.0109, -0.0151, -0.0365, ..., -0.0698, 0.0090, 0.0288], + [-0.0649, -0.0441, -0.0461, ..., -0.0308, -0.0321, -0.0573], + [-0.0017, 0.0008, -0.0463, ..., -0.0330, 0.0293, -0.0083]], + device='cuda:0'), grad: tensor([[-1.8173e-02, -1.5430e-03, 4.4912e-05, ..., -2.7466e-03, + -8.6060e-03, -8.9798e-03], + [-6.1655e-04, 2.3925e-04, -3.6776e-05, ..., 1.4086e-03, + 8.9979e-04, -2.4453e-05], + [ 1.9989e-03, 2.0385e-04, 6.4671e-05, ..., 1.2455e-03, + 8.4305e-04, 1.1345e-02], + ..., + [-2.4242e-03, -3.0766e-03, 1.1301e-04, ..., 6.1512e-04, + -2.6493e-03, -1.6876e-02], + [ 1.6451e-03, 2.6345e-04, 1.6272e-04, ..., 1.5726e-03, + 1.0767e-03, -1.1511e-03], + [ 6.9466e-03, 1.8206e-03, 1.7166e-04, ..., -1.0590e-02, + 3.7727e-03, 6.4735e-03]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0141, -0.0172, 0.0055, -0.0181, 0.0156, 0.0057, -0.0135, -0.0274, + 0.0098, -0.0099], device='cuda:0'), grad: tensor([-0.0534, -0.0052, 0.0441, -0.0424, 0.0391, 0.0054, 0.0349, -0.0204, + -0.0039, 0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 225.59, cls_loss 0.7141 cls_loss_mapping 0.0451 cls_loss_causal 0.6840 re_mapping 0.0179 re_causal 0.0466 /// teacc 98.46 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0013, 0.0377, -0.0852, ..., -0.0422, 0.0291, 0.0119], + [-0.0366, -0.0595, 0.0343, ..., -0.0268, -0.0651, 0.0467], + [-0.0327, -0.0146, -0.0510, ..., -0.0737, -0.0366, 0.0177], + ..., + [-0.0103, -0.0150, -0.0376, ..., -0.0688, 0.0093, 0.0299], + [-0.0648, -0.0441, -0.0469, ..., -0.0317, -0.0320, -0.0572], + [-0.0018, 0.0005, -0.0459, ..., -0.0328, 0.0299, -0.0087]], + device='cuda:0'), grad: tensor([[-1.9135e-02, -1.1176e-04, 7.2289e-04, ..., 3.1590e-04, + -6.9847e-03, -1.6113e-02], + [-1.3535e-02, 1.1027e-04, -1.0006e-05, ..., 2.0742e-04, + -9.8114e-03, -5.5962e-03], + [ 2.3460e-03, -4.0793e-04, 2.5010e-04, ..., 7.6103e-04, + 4.9629e-03, 1.6174e-03], + ..., + [-6.8665e-04, -3.8218e-04, 3.4380e-04, ..., 6.6185e-04, + -6.0225e-04, -4.7722e-03], + [-2.6131e-03, -2.1027e-02, -3.3386e-02, ..., -1.7288e-02, + -6.4354e-03, -2.0676e-03], + [ 6.2752e-03, 1.2980e-03, 8.2588e-04, ..., 4.0507e-04, + 4.4174e-03, 1.1497e-02]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0139, -0.0167, 0.0053, -0.0184, 0.0149, 0.0058, -0.0131, -0.0272, + 0.0098, -0.0098], device='cuda:0'), grad: tensor([-0.0972, -0.0208, 0.0168, 0.0339, 0.0013, 0.0818, -0.0074, 0.0030, + -0.0615, 0.0501], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 225.62, cls_loss 0.7170 cls_loss_mapping 0.0448 cls_loss_causal 0.6856 re_mapping 0.0178 re_causal 0.0476 /// teacc 98.48 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0008, 0.0393, -0.0847, ..., -0.0410, 0.0301, 0.0127], + [-0.0361, -0.0609, 0.0346, ..., -0.0269, -0.0650, 0.0462], + [-0.0334, -0.0152, -0.0515, ..., -0.0754, -0.0374, 0.0180], + ..., + [-0.0101, -0.0151, -0.0374, ..., -0.0694, 0.0092, 0.0295], + [-0.0655, -0.0445, -0.0468, ..., -0.0303, -0.0321, -0.0572], + [-0.0022, -0.0007, -0.0467, ..., -0.0343, 0.0288, -0.0091]], + device='cuda:0'), grad: tensor([[-6.8665e-04, -1.0239e-02, -9.5520e-03, ..., -6.3858e-03, + -2.6741e-03, -4.9438e-03], + [ 4.8561e-03, 1.5438e-04, 7.8321e-05, ..., 1.3113e-04, + 5.0850e-03, 7.5722e-03], + [ 3.7408e-04, 1.5564e-03, 2.7299e-04, ..., 1.5812e-03, + 7.2403e-03, 2.3193e-03], + ..., + [-4.1351e-03, 6.6757e-04, 5.3692e-04, ..., 5.2977e-04, + -6.6414e-03, -1.1185e-02], + [-1.3721e-04, 5.5847e-03, 5.2223e-03, ..., 4.0665e-03, + 2.4605e-03, 5.0468e-03], + [ 2.5845e-03, 1.7385e-03, 1.5316e-03, ..., 1.1683e-03, + 3.2520e-03, 5.2071e-03]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0146, -0.0172, 0.0047, -0.0182, 0.0155, 0.0055, -0.0128, -0.0270, + 0.0103, -0.0107], device='cuda:0'), grad: tensor([-0.0332, 0.0435, -0.0133, -0.0171, 0.0290, 0.0139, -0.0142, -0.0603, + 0.0202, 0.0314], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 226.11, cls_loss 0.6927 cls_loss_mapping 0.0469 cls_loss_causal 0.6510 re_mapping 0.0174 re_causal 0.0456 /// teacc 98.21 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0004, 0.0403, -0.0852, ..., -0.0421, 0.0311, 0.0122], + [-0.0366, -0.0609, 0.0352, ..., -0.0262, -0.0659, 0.0462], + [-0.0326, -0.0143, -0.0505, ..., -0.0748, -0.0366, 0.0185], + ..., + [-0.0103, -0.0161, -0.0380, ..., -0.0709, 0.0087, 0.0297], + [-0.0655, -0.0458, -0.0482, ..., -0.0309, -0.0321, -0.0569], + [-0.0024, -0.0007, -0.0471, ..., -0.0337, 0.0284, -0.0093]], + device='cuda:0'), grad: tensor([[ 3.8147e-03, 1.1473e-03, 4.6992e-04, ..., 9.2459e-04, + 3.5572e-03, 5.3062e-03], + [-8.0490e-04, -7.8058e-04, -1.8448e-05, ..., 7.2575e-04, + -2.2278e-03, -1.2543e-02], + [-3.5973e-03, 9.4128e-04, 2.5535e-04, ..., 7.2765e-04, + -4.9114e-04, -1.7681e-03], + ..., + [ 1.6270e-03, 9.2745e-04, 2.9802e-04, ..., 3.9339e-04, + -1.3244e-04, 4.6158e-03], + [-2.6722e-03, -6.0463e-04, 1.8988e-03, ..., -5.1928e-04, + -1.2960e-03, -3.9749e-03], + [ 2.3308e-03, 2.4776e-03, 1.4296e-03, ..., 1.0729e-03, + -5.4741e-03, 1.6670e-03]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0145, -0.0172, 0.0052, -0.0183, 0.0147, 0.0060, -0.0133, -0.0269, + 0.0103, -0.0104], device='cuda:0'), grad: tensor([ 0.0323, -0.0107, -0.0183, -0.0083, 0.0036, 0.0158, 0.0311, -0.0087, + -0.0174, -0.0193], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 52---------------------------------------------------- +epoch 52, time 226.31, cls_loss 0.6633 cls_loss_mapping 0.0380 cls_loss_causal 0.6271 re_mapping 0.0175 re_causal 0.0462 /// teacc 98.52 lr 0.00010000 +Epoch 54, weight, value: tensor([[-0.0002, 0.0411, -0.0862, ..., -0.0429, 0.0310, 0.0118], + [-0.0368, -0.0613, 0.0369, ..., -0.0263, -0.0663, 0.0466], + [-0.0331, -0.0148, -0.0512, ..., -0.0743, -0.0363, 0.0190], + ..., + [-0.0112, -0.0167, -0.0381, ..., -0.0713, 0.0074, 0.0306], + [-0.0656, -0.0456, -0.0485, ..., -0.0306, -0.0321, -0.0571], + [-0.0021, -0.0013, -0.0477, ..., -0.0339, 0.0287, -0.0096]], + device='cuda:0'), grad: tensor([[-0.0061, -0.0051, -0.0059, ..., 0.0009, -0.0074, 0.0015], + [ 0.0017, 0.0010, -0.0033, ..., -0.0050, 0.0013, -0.0097], + [ 0.0004, 0.0003, 0.0008, ..., 0.0018, 0.0022, 0.0035], + ..., + [-0.0073, 0.0001, -0.0012, ..., -0.0034, -0.0022, -0.0059], + [ 0.0047, 0.0004, 0.0059, ..., 0.0204, 0.0047, 0.0093], + [ 0.0085, 0.0025, 0.0026, ..., 0.0017, 0.0111, 0.0058]], + device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0135, -0.0166, 0.0056, -0.0180, 0.0141, 0.0058, -0.0135, -0.0266, + 0.0104, -0.0101], device='cuda:0'), grad: tensor([-0.0196, -0.0507, 0.0309, -0.0266, 0.0293, 0.0237, -0.0773, -0.0454, + 0.0850, 0.0505], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 225.59, cls_loss 0.6968 cls_loss_mapping 0.0452 cls_loss_causal 0.6624 re_mapping 0.0173 re_causal 0.0476 /// teacc 98.30 lr 0.00010000 +Epoch 55, weight, value: tensor([[ 0.0005, 0.0421, -0.0868, ..., -0.0443, 0.0320, 0.0125], + [-0.0371, -0.0633, 0.0364, ..., -0.0277, -0.0674, 0.0469], + [-0.0341, -0.0153, -0.0511, ..., -0.0743, -0.0370, 0.0196], + ..., + [-0.0113, -0.0165, -0.0387, ..., -0.0722, 0.0066, 0.0302], + [-0.0660, -0.0463, -0.0501, ..., -0.0318, -0.0323, -0.0571], + [-0.0020, -0.0015, -0.0470, ..., -0.0324, 0.0283, -0.0103]], + device='cuda:0'), grad: tensor([[ 0.0026, 0.0015, 0.0016, ..., 0.0027, 0.0034, 0.0053], + [ 0.0118, 0.0024, 0.0041, ..., 0.0035, 0.0120, 0.0105], + [ 0.0013, 0.0043, 0.0016, ..., 0.0043, 0.0038, 0.0065], + ..., + [ 0.0020, -0.0033, 0.0005, ..., -0.0015, -0.0073, -0.0163], + [ 0.0023, -0.0003, 0.0011, ..., -0.0078, -0.0020, 0.0057], + [-0.0102, 0.0001, -0.0023, ..., 0.0004, -0.0023, -0.0051]], + device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0134, -0.0163, 0.0053, -0.0181, 0.0144, 0.0056, -0.0129, -0.0266, + 0.0099, -0.0102], device='cuda:0'), grad: tensor([ 0.0269, 0.0571, 0.0228, 0.0592, -0.1021, -0.0146, 0.0159, -0.0177, + -0.0208, -0.0267], device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 225.29, cls_loss 0.6599 cls_loss_mapping 0.0382 cls_loss_causal 0.6225 re_mapping 0.0168 re_causal 0.0450 /// teacc 98.49 lr 0.00010000 +Epoch 56, weight, value: tensor([[ 5.7066e-04, 4.1984e-02, -8.6625e-02, ..., -4.4390e-02, + 3.2981e-02, 1.2333e-02], + [-3.7800e-02, -6.4175e-02, 3.6200e-02, ..., -2.8336e-02, + -6.8009e-02, 4.7212e-02], + [-3.3859e-02, -1.5319e-02, -5.1762e-02, ..., -7.6218e-02, + -3.6601e-02, 1.9409e-02], + ..., + [-1.2191e-02, -1.7102e-02, -3.9077e-02, ..., -7.2232e-02, + 5.5924e-03, 3.1239e-02], + [-6.6348e-02, -4.5007e-02, -5.0654e-02, ..., -3.2035e-02, + -3.1807e-02, -5.7615e-02], + [-1.1237e-03, -1.6215e-05, -4.6481e-02, ..., -3.2075e-02, + 2.8670e-02, -9.6121e-03]], device='cuda:0'), grad: tensor([[-2.2621e-03, 3.1972e-04, 2.2054e-04, ..., 6.8378e-04, + -7.9250e-04, 3.5501e-04], + [ 2.4319e-03, 3.6168e-04, 1.6654e-04, ..., 4.3631e-04, + 2.8114e-03, 3.1033e-03], + [-4.6577e-03, 5.7697e-04, 5.3215e-04, ..., 2.5019e-05, + -1.0519e-03, -1.3399e-04], + ..., + [-3.8013e-03, -1.7548e-03, -3.0947e-04, ..., -3.0518e-04, + -3.2387e-03, -1.3367e-02], + [ 2.4815e-03, 4.9448e-04, 3.9744e-04, ..., -4.2763e-03, + 2.8553e-03, 6.4278e-03], + [ 3.1261e-03, 7.6818e-04, 4.9639e-04, ..., 9.0170e-04, + 3.3875e-03, 5.5199e-03]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0148, -0.0166, 0.0053, -0.0181, 0.0141, 0.0047, -0.0136, -0.0265, + 0.0103, -0.0100], device='cuda:0'), grad: tensor([-0.0090, 0.0236, -0.0062, 0.0050, 0.0168, -0.0026, 0.0090, -0.0476, + 0.0100, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 225.22, cls_loss 0.6950 cls_loss_mapping 0.0344 cls_loss_causal 0.6596 re_mapping 0.0165 re_causal 0.0430 /// teacc 98.44 lr 0.00010000 +Epoch 57, weight, value: tensor([[ 0.0007, 0.0429, -0.0871, ..., -0.0437, 0.0328, 0.0122], + [-0.0382, -0.0645, 0.0365, ..., -0.0282, -0.0688, 0.0477], + [-0.0339, -0.0162, -0.0523, ..., -0.0769, -0.0367, 0.0199], + ..., + [-0.0124, -0.0177, -0.0396, ..., -0.0726, 0.0051, 0.0317], + [-0.0665, -0.0453, -0.0518, ..., -0.0328, -0.0320, -0.0576], + [-0.0016, -0.0015, -0.0473, ..., -0.0326, 0.0273, -0.0102]], + device='cuda:0'), grad: tensor([[ 1.4791e-03, 9.2745e-05, 1.4579e-04, ..., 1.6623e-03, + 1.5116e-03, 2.7905e-03], + [ 2.9068e-03, 8.2791e-05, 1.0037e-04, ..., 7.4339e-04, + 2.5959e-03, 5.3902e-03], + [-8.7976e-04, 1.1759e-03, 2.2900e-04, ..., 1.7509e-03, + 5.4693e-04, -2.2717e-03], + ..., + [ 9.5215e-03, 3.5977e-04, 2.1112e-04, ..., 5.2309e-04, + 7.9956e-03, 2.5024e-02], + [-1.9569e-03, -4.3106e-03, -3.0060e-03, ..., 1.7595e-03, + -4.7226e-03, 4.3607e-04], + [-4.4861e-03, -4.1676e-04, 2.7657e-04, ..., 7.3862e-04, + -4.5624e-03, -1.9821e-02]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0149, -0.0166, 0.0056, -0.0175, 0.0145, 0.0044, -0.0143, -0.0270, + 0.0102, -0.0096], device='cuda:0'), grad: tensor([-0.0171, 0.0219, -0.0029, 0.0082, -0.0202, -0.0098, -0.0024, 0.0507, + -0.0183, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 224.66, cls_loss 0.6957 cls_loss_mapping 0.0429 cls_loss_causal 0.6616 re_mapping 0.0172 re_causal 0.0462 /// teacc 98.34 lr 0.00010000 +Epoch 58, weight, value: tensor([[ 0.0010, 0.0421, -0.0878, ..., -0.0441, 0.0329, 0.0121], + [-0.0379, -0.0652, 0.0379, ..., -0.0265, -0.0689, 0.0474], + [-0.0343, -0.0167, -0.0526, ..., -0.0780, -0.0360, 0.0202], + ..., + [-0.0116, -0.0170, -0.0403, ..., -0.0727, 0.0050, 0.0320], + [-0.0673, -0.0468, -0.0523, ..., -0.0338, -0.0327, -0.0573], + [-0.0017, -0.0007, -0.0477, ..., -0.0318, 0.0279, -0.0097]], + device='cuda:0'), grad: tensor([[ 3.9711e-03, 1.6308e-03, 9.6798e-04, ..., 8.0299e-04, + 1.6575e-03, 4.9248e-03], + [ 1.8759e-03, 6.0749e-04, 5.6446e-05, ..., 8.2016e-05, + -1.6129e-04, 1.9855e-03], + [ 4.3259e-03, 1.4534e-03, 2.0444e-04, ..., 5.8222e-04, + -6.3479e-05, 5.0392e-03], + ..., + [ 2.8877e-03, 9.0933e-04, 2.8706e-04, ..., 2.6226e-04, + 4.0531e-04, 6.1188e-03], + [-9.9487e-03, -1.0033e-02, -4.6043e-03, ..., -3.5133e-03, + -7.0953e-03, -1.8875e-02], + [-4.7226e-03, 1.5011e-03, 8.0967e-04, ..., 6.4135e-04, + -2.5043e-03, 2.7394e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0155, -0.0161, 0.0059, -0.0184, 0.0143, 0.0049, -0.0150, -0.0266, + 0.0098, -0.0098], device='cuda:0'), grad: tensor([ 0.0290, 0.0035, 0.0287, -0.0478, 0.0494, 0.0076, -0.0023, 0.0180, + -0.0723, -0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 225.03, cls_loss 0.7097 cls_loss_mapping 0.0378 cls_loss_causal 0.6698 re_mapping 0.0163 re_causal 0.0411 /// teacc 98.27 lr 0.00010000 +Epoch 59, weight, value: tensor([[ 0.0012, 0.0425, -0.0879, ..., -0.0439, 0.0330, 0.0124], + [-0.0382, -0.0655, 0.0389, ..., -0.0262, -0.0694, 0.0471], + [-0.0345, -0.0164, -0.0531, ..., -0.0775, -0.0362, 0.0209], + ..., + [-0.0120, -0.0179, -0.0415, ..., -0.0738, 0.0051, 0.0319], + [-0.0673, -0.0465, -0.0518, ..., -0.0341, -0.0323, -0.0575], + [-0.0012, -0.0012, -0.0477, ..., -0.0317, 0.0281, -0.0094]], + device='cuda:0'), grad: tensor([[-9.9030e-03, 2.0504e-04, 1.3545e-05, ..., 2.2531e-05, + -4.3392e-04, 2.8286e-03], + [ 3.8700e-03, 7.5531e-04, -1.7229e-08, ..., 3.3248e-07, + 2.5978e-03, 3.9139e-03], + [-5.3635e-03, -1.0624e-03, 1.5043e-05, ..., 3.3267e-06, + -2.5330e-03, -7.0457e-03], + ..., + [-7.8011e-04, -7.3719e-04, 3.0413e-05, ..., 8.6650e-06, + 2.1124e-04, -3.4218e-03], + [ 3.8338e-03, 4.9877e-04, 1.2197e-05, ..., 1.1764e-05, + 1.2426e-03, 3.6621e-03], + [ 1.0468e-02, -1.5962e-04, 3.4690e-05, ..., -7.5817e-05, + 1.3447e-03, 2.2564e-03]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0160, -0.0158, 0.0062, -0.0180, 0.0151, 0.0047, -0.0158, -0.0280, + 0.0098, -0.0097], device='cuda:0'), grad: tensor([-0.0106, 0.0346, -0.0682, -0.0362, -0.0095, 0.0202, -0.0114, -0.0098, + 0.0501, 0.0408], device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 225.92, cls_loss 0.6736 cls_loss_mapping 0.0398 cls_loss_causal 0.6411 re_mapping 0.0166 re_causal 0.0412 /// teacc 98.26 lr 0.00010000 +Epoch 60, weight, value: tensor([[ 0.0015, 0.0438, -0.0880, ..., -0.0448, 0.0336, 0.0123], + [-0.0382, -0.0667, 0.0388, ..., -0.0263, -0.0695, 0.0469], + [-0.0353, -0.0169, -0.0547, ..., -0.0786, -0.0365, 0.0209], + ..., + [-0.0122, -0.0189, -0.0431, ..., -0.0738, 0.0050, 0.0323], + [-0.0684, -0.0462, -0.0521, ..., -0.0342, -0.0335, -0.0569], + [-0.0009, -0.0023, -0.0483, ..., -0.0318, 0.0271, -0.0089]], + device='cuda:0'), grad: tensor([[ 3.2139e-03, 9.4986e-03, 3.7823e-03, ..., 3.4084e-03, + 1.1070e-02, 2.6321e-03], + [ 5.2223e-03, 6.7902e-04, -4.0970e-03, ..., -5.3358e-04, + 5.9853e-03, 1.8415e-03], + [-5.5618e-03, -9.5978e-03, 1.5163e-03, ..., 4.2826e-05, + -2.3155e-03, -2.3327e-03], + ..., + [-2.7866e-03, -3.3112e-03, -3.2592e-04, ..., -9.2506e-05, + -1.0178e-02, -1.1345e-02], + [-1.5182e-03, 1.7869e-04, -3.4885e-03, ..., -8.7690e-04, + 2.5330e-03, -5.0011e-03], + [ 2.9945e-03, 1.9283e-03, 1.3742e-03, ..., 5.9366e-04, + -1.1787e-03, 2.2793e-03]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0156, -0.0158, 0.0060, -0.0170, 0.0156, 0.0037, -0.0158, -0.0280, + 0.0098, -0.0098], device='cuda:0'), grad: tensor([ 0.0193, 0.0220, -0.0404, -0.0093, 0.0053, -0.0070, -0.0113, -0.0065, + 0.0206, 0.0073], device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 224.15, cls_loss 0.6635 cls_loss_mapping 0.0436 cls_loss_causal 0.6285 re_mapping 0.0165 re_causal 0.0424 /// teacc 98.45 lr 0.00010000 +Epoch 61, weight, value: tensor([[ 0.0015, 0.0443, -0.0879, ..., -0.0449, 0.0338, 0.0117], + [-0.0387, -0.0676, 0.0379, ..., -0.0271, -0.0704, 0.0475], + [-0.0363, -0.0172, -0.0533, ..., -0.0785, -0.0368, 0.0211], + ..., + [-0.0119, -0.0189, -0.0430, ..., -0.0734, 0.0057, 0.0324], + [-0.0678, -0.0464, -0.0527, ..., -0.0348, -0.0325, -0.0560], + [-0.0011, -0.0026, -0.0476, ..., -0.0319, 0.0272, -0.0090]], + device='cuda:0'), grad: tensor([[ 0.0021, -0.0013, -0.0023, ..., -0.0016, 0.0035, -0.0015], + [-0.0022, 0.0007, 0.0019, ..., 0.0006, 0.0011, -0.0030], + [ 0.0015, 0.0010, 0.0013, ..., 0.0006, 0.0018, 0.0003], + ..., + [-0.0041, -0.0008, 0.0009, ..., -0.0012, -0.0064, -0.0092], + [ 0.0023, 0.0102, 0.0087, ..., 0.0009, 0.0135, 0.0025], + [ 0.0032, 0.0009, 0.0014, ..., 0.0004, 0.0073, 0.0045]], + device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0153, -0.0161, 0.0055, -0.0178, 0.0150, 0.0041, -0.0151, -0.0276, + 0.0110, -0.0099], device='cuda:0'), grad: tensor([-0.0118, -0.0023, 0.0156, -0.0526, 0.0107, 0.0042, 0.0151, -0.0499, + 0.0448, 0.0261], device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 224.36, cls_loss 0.6757 cls_loss_mapping 0.0297 cls_loss_causal 0.6426 re_mapping 0.0164 re_causal 0.0429 /// teacc 98.43 lr 0.00010000 +Epoch 62, weight, value: tensor([[ 0.0020, 0.0445, -0.0883, ..., -0.0450, 0.0345, 0.0120], + [-0.0392, -0.0675, 0.0385, ..., -0.0272, -0.0711, 0.0476], + [-0.0363, -0.0171, -0.0539, ..., -0.0785, -0.0373, 0.0216], + ..., + [-0.0121, -0.0206, -0.0441, ..., -0.0755, 0.0056, 0.0327], + [-0.0677, -0.0467, -0.0549, ..., -0.0356, -0.0320, -0.0556], + [-0.0009, -0.0026, -0.0473, ..., -0.0314, 0.0279, -0.0089]], + device='cuda:0'), grad: tensor([[ 0.0032, 0.0019, 0.0008, ..., 0.0017, 0.0046, 0.0051], + [ 0.0009, -0.0097, -0.0149, ..., -0.0049, -0.0147, -0.0107], + [ 0.0008, 0.0084, 0.0088, ..., 0.0025, 0.0117, 0.0079], + ..., + [-0.0065, -0.0091, -0.0008, ..., -0.0002, -0.0039, -0.0187], + [ 0.0037, 0.0003, 0.0009, ..., 0.0005, 0.0046, 0.0035], + [-0.0075, 0.0032, 0.0004, ..., -0.0014, -0.0027, 0.0068]], + device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0163, -0.0160, 0.0046, -0.0178, 0.0146, 0.0033, -0.0146, -0.0275, + 0.0117, -0.0103], device='cuda:0'), grad: tensor([ 0.0355, -0.0187, 0.0278, 0.0453, -0.0336, -0.0002, 0.0364, -0.0847, + 0.0169, -0.0249], device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 225.31, cls_loss 0.6754 cls_loss_mapping 0.0306 cls_loss_causal 0.6414 re_mapping 0.0161 re_causal 0.0422 /// teacc 98.44 lr 0.00010000 +Epoch 63, weight, value: tensor([[ 0.0017, 0.0445, -0.0900, ..., -0.0465, 0.0341, 0.0107], + [-0.0394, -0.0686, 0.0389, ..., -0.0256, -0.0715, 0.0472], + [-0.0354, -0.0165, -0.0535, ..., -0.0790, -0.0356, 0.0222], + ..., + [-0.0120, -0.0198, -0.0450, ..., -0.0743, 0.0053, 0.0341], + [-0.0687, -0.0472, -0.0551, ..., -0.0360, -0.0323, -0.0560], + [-0.0001, -0.0029, -0.0461, ..., -0.0312, 0.0282, -0.0091]], + device='cuda:0'), grad: tensor([[ 3.0518e-05, 4.0483e-04, 2.0072e-05, ..., -5.1041e-03, + 1.7738e-03, 1.1161e-05], + [ 8.0204e-04, 3.5524e-04, -6.6471e-04, ..., 7.4625e-04, + 1.5669e-03, 2.2244e-04], + [ 3.1796e-03, 1.4496e-04, 8.9073e-04, ..., 3.8433e-03, + 8.5373e-03, 2.2461e-02], + ..., + [ 1.0977e-03, 1.3514e-03, -9.9373e-04, ..., 2.7313e-03, + -4.8370e-03, -1.7014e-02], + [ 2.6054e-03, 3.6478e-04, 1.0223e-03, ..., 7.8106e-04, + 2.4033e-03, -2.7866e-03], + [-3.3169e-03, -1.0214e-03, 8.5068e-04, ..., -1.8206e-03, + -1.3990e-03, 8.8959e-03]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0155, -0.0165, 0.0055, -0.0173, 0.0148, 0.0036, -0.0153, -0.0267, + 0.0111, -0.0103], device='cuda:0'), grad: tensor([-0.0008, 0.0100, 0.0685, 0.0077, -0.0390, -0.0166, -0.0096, -0.0189, + -0.0180, 0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 225.02, cls_loss 0.6388 cls_loss_mapping 0.0326 cls_loss_causal 0.6041 re_mapping 0.0156 re_causal 0.0409 /// teacc 98.50 lr 0.00010000 +Epoch 64, weight, value: tensor([[ 2.4275e-03, 4.4616e-02, -8.9562e-02, ..., -4.7088e-02, + 3.4766e-02, 1.1245e-02], + [-4.0169e-02, -6.9452e-02, 3.8923e-02, ..., -2.6323e-02, + -7.1997e-02, 4.7112e-02], + [-3.5910e-02, -1.6920e-02, -5.4046e-02, ..., -8.0639e-02, + -3.6268e-02, 2.2695e-02], + ..., + [-1.2366e-02, -1.9815e-02, -4.6078e-02, ..., -7.4803e-02, + 5.9358e-03, 3.4208e-02], + [-6.9503e-02, -4.8270e-02, -5.5267e-02, ..., -3.6966e-02, + -3.3415e-02, -5.7272e-02], + [-6.5580e-05, -3.5525e-03, -4.6799e-02, ..., -3.1321e-02, + 2.7720e-02, -9.4830e-03]], device='cuda:0'), grad: tensor([[ 1.3151e-03, 2.7418e-05, 1.1599e-04, ..., 3.2449e-04, + 1.5020e-03, 1.8492e-03], + [-4.3526e-03, 2.4378e-05, -1.9255e-03, ..., -1.6508e-03, + -1.8282e-03, -1.2901e-02], + [ 1.0014e-03, 3.7122e-04, 2.7227e-04, ..., 1.8060e-04, + 1.1617e-04, 1.1930e-03], + ..., + [ 9.5367e-04, 7.8499e-05, 8.4937e-05, ..., 3.7766e-04, + 4.7445e-04, 1.3533e-03], + [-1.6618e-04, 5.3263e-04, 9.6989e-04, ..., 1.0719e-03, + 2.9526e-03, 4.8828e-03], + [ 9.1887e-04, 2.0838e-04, 6.2168e-05, ..., 3.2520e-04, + -6.2525e-05, 1.1559e-03]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0165, -0.0165, 0.0056, -0.0181, 0.0151, 0.0047, -0.0152, -0.0275, + 0.0101, -0.0102], device='cuda:0'), grad: tensor([ 0.0116, -0.0523, 0.0093, -0.0109, -0.0193, 0.0202, 0.0051, 0.0083, + 0.0195, 0.0086], device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 224.87, cls_loss 0.6830 cls_loss_mapping 0.0390 cls_loss_causal 0.6536 re_mapping 0.0145 re_causal 0.0392 /// teacc 98.44 lr 0.00010000 +Epoch 65, weight, value: tensor([[ 0.0026, 0.0467, -0.0881, ..., -0.0460, 0.0348, 0.0114], + [-0.0407, -0.0697, 0.0405, ..., -0.0258, -0.0727, 0.0469], + [-0.0361, -0.0178, -0.0547, ..., -0.0798, -0.0369, 0.0224], + ..., + [-0.0124, -0.0199, -0.0473, ..., -0.0757, 0.0053, 0.0342], + [-0.0697, -0.0488, -0.0561, ..., -0.0378, -0.0340, -0.0573], + [-0.0005, -0.0037, -0.0476, ..., -0.0318, 0.0284, -0.0095]], + device='cuda:0'), grad: tensor([[-6.9427e-04, -8.7814e-03, -1.3062e-02, ..., -6.6643e-03, + -5.1155e-03, -1.4715e-03], + [-1.2589e-02, -8.6308e-04, 7.8964e-04, ..., -2.5692e-03, + -1.3580e-02, -9.6436e-03], + [ 4.3678e-03, 3.4351e-03, 1.1549e-03, ..., 5.8708e-03, + 9.6512e-03, 8.2397e-03], + ..., + [ 4.0092e-03, 1.1396e-03, 3.9983e-04, ..., 1.1606e-03, + 3.5286e-03, -8.6308e-04], + [-2.1210e-03, 1.1787e-03, 2.0695e-03, ..., 1.6870e-03, + 2.2812e-03, -1.1692e-03], + [ 2.4281e-03, -2.4438e-05, -1.4150e-04, ..., 8.8501e-04, + 2.6932e-03, 5.0125e-03]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0161, -0.0160, 0.0049, -0.0171, 0.0150, 0.0041, -0.0157, -0.0275, + 0.0103, -0.0097], device='cuda:0'), grad: tensor([-0.0393, -0.0820, 0.0501, 0.0548, 0.0207, 0.0015, -0.0375, 0.0030, + -0.0034, 0.0321], device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 225.45, cls_loss 0.6500 cls_loss_mapping 0.0267 cls_loss_causal 0.6066 re_mapping 0.0144 re_causal 0.0371 /// teacc 98.41 lr 0.00010000 +Epoch 66, weight, value: tensor([[ 0.0024, 0.0467, -0.0881, ..., -0.0469, 0.0353, 0.0121], + [-0.0410, -0.0704, 0.0417, ..., -0.0261, -0.0721, 0.0465], + [-0.0366, -0.0184, -0.0554, ..., -0.0801, -0.0375, 0.0231], + ..., + [-0.0122, -0.0205, -0.0472, ..., -0.0760, 0.0047, 0.0347], + [-0.0701, -0.0490, -0.0556, ..., -0.0378, -0.0329, -0.0578], + [ 0.0003, -0.0032, -0.0475, ..., -0.0321, 0.0291, -0.0100]], + device='cuda:0'), grad: tensor([[ 4.1771e-03, 1.1349e-03, 6.9952e-04, ..., 1.6146e-03, + -6.0120e-03, 7.5493e-03], + [ 7.8535e-04, 6.6519e-04, 7.6175e-05, ..., 7.2718e-04, + 1.0357e-03, -2.5349e-03], + [-2.4490e-03, -7.5006e-04, -2.6150e-03, ..., -4.9257e-04, + 6.8617e-04, -1.5388e-02], + ..., + [-4.7874e-03, 7.3957e-04, 1.5306e-03, ..., 1.0033e-03, + -2.1896e-03, 7.4806e-03], + [ 2.3422e-03, 5.8413e-04, -2.1877e-03, ..., 5.6356e-05, + 3.3607e-03, 6.0349e-03], + [-4.9019e-04, 3.2926e-04, 5.0592e-04, ..., 5.7745e-04, + 2.8858e-03, -6.5231e-04]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0165, -0.0156, 0.0047, -0.0175, 0.0142, 0.0039, -0.0150, -0.0271, + 0.0101, -0.0097], device='cuda:0'), grad: tensor([ 0.0283, -0.0044, -0.0410, 0.0393, -0.0450, -0.0522, 0.0300, 0.0172, + 0.0242, 0.0036], device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 227.44, cls_loss 0.6575 cls_loss_mapping 0.0367 cls_loss_causal 0.6267 re_mapping 0.0155 re_causal 0.0400 /// teacc 98.47 lr 0.00010000 +Epoch 67, weight, value: tensor([[ 0.0028, 0.0472, -0.0880, ..., -0.0478, 0.0355, 0.0117], + [-0.0410, -0.0708, 0.0427, ..., -0.0258, -0.0719, 0.0471], + [-0.0368, -0.0176, -0.0547, ..., -0.0797, -0.0384, 0.0232], + ..., + [-0.0128, -0.0214, -0.0488, ..., -0.0771, 0.0047, 0.0346], + [-0.0707, -0.0495, -0.0566, ..., -0.0369, -0.0320, -0.0573], + [-0.0002, -0.0025, -0.0466, ..., -0.0326, 0.0288, -0.0105]], + device='cuda:0'), grad: tensor([[-1.6766e-03, 4.2057e-04, 1.0605e-03, ..., 6.3562e-04, + -5.3940e-03, 3.9444e-03], + [ 1.9989e-03, 3.1376e-04, -3.3212e-04, ..., 2.6965e-04, + 2.0275e-03, 3.0422e-03], + [ 2.8968e-04, 2.8362e-03, 2.3804e-03, ..., 1.5364e-03, + 3.7842e-03, 7.3242e-03], + ..., + [ 2.0814e-04, 2.5082e-04, 3.1400e-04, ..., 1.6987e-04, + 1.4353e-03, 2.4557e-04], + [ 4.5700e-03, 9.0361e-04, 1.1930e-03, ..., 8.6260e-04, + 4.3106e-03, -8.9884e-05], + [ 2.4567e-03, 5.8174e-04, 2.5558e-04, ..., 2.8348e-04, + 2.4948e-03, 2.5673e-03]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0165, -0.0153, 0.0047, -0.0176, 0.0143, 0.0042, -0.0154, -0.0273, + 0.0105, -0.0103], device='cuda:0'), grad: tensor([ 0.0058, 0.0206, 0.0280, -0.1130, -0.0385, 0.0221, 0.0361, -0.0024, + 0.0221, 0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 229.32, cls_loss 0.6643 cls_loss_mapping 0.0392 cls_loss_causal 0.6345 re_mapping 0.0150 re_causal 0.0385 /// teacc 98.47 lr 0.00010000 +Epoch 68, weight, value: tensor([[ 2.3588e-03, 4.7824e-02, -8.9415e-02, ..., -4.8016e-02, + 3.5947e-02, 1.1280e-02], + [-4.1629e-02, -7.1511e-02, 4.3136e-02, ..., -2.5728e-02, + -7.1830e-02, 4.6428e-02], + [-3.7065e-02, -1.8793e-02, -5.5638e-02, ..., -8.0089e-02, + -3.8290e-02, 2.4786e-02], + ..., + [-1.3074e-02, -2.0846e-02, -4.9333e-02, ..., -7.6993e-02, + 4.2430e-03, 3.4946e-02], + [-7.1460e-02, -5.0773e-02, -5.8150e-02, ..., -3.7796e-02, + -3.3249e-02, -5.7191e-02], + [-9.9651e-05, -2.8808e-03, -4.6023e-02, ..., -3.3223e-02, + 2.7450e-02, -1.1071e-02]], device='cuda:0'), grad: tensor([[-6.1913e-03, 1.3256e-03, -2.3308e-03, ..., 1.4772e-03, + -8.4534e-03, -5.4502e-04], + [ 1.3294e-03, 8.9586e-05, 5.1451e-04, ..., -9.0179e-03, + 9.9564e-04, 2.7990e-04], + [ 2.3067e-04, 8.6117e-04, -3.7460e-03, ..., -1.2846e-03, + -6.5851e-04, 2.9993e-04], + ..., + [ 2.4395e-03, 2.9778e-04, 7.1430e-04, ..., 4.4465e-04, + 1.4315e-03, 1.1787e-03], + [-1.3672e-02, -6.2675e-03, -4.9210e-03, ..., -7.4196e-03, + -6.6261e-03, -2.6054e-03], + [ 7.6790e-03, 1.2760e-03, 4.4060e-03, ..., 3.5477e-03, + 5.4703e-03, 5.3644e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0159, -0.0155, 0.0044, -0.0169, 0.0145, 0.0043, -0.0154, -0.0268, + 0.0099, -0.0101], device='cuda:0'), grad: tensor([-0.0093, -0.0223, -0.0149, 0.0260, 0.0011, 0.0202, 0.0307, 0.0147, + -0.0890, 0.0428], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 67---------------------------------------------------- +epoch 67, time 227.35, cls_loss 0.6543 cls_loss_mapping 0.0296 cls_loss_causal 0.6228 re_mapping 0.0154 re_causal 0.0391 /// teacc 98.53 lr 0.00010000 +Epoch 69, weight, value: tensor([[ 2.9902e-03, 4.8225e-02, -8.9906e-02, ..., -4.9170e-02, + 3.6698e-02, 1.0973e-02], + [-4.2294e-02, -7.2576e-02, 4.4069e-02, ..., -2.5119e-02, + -7.2937e-02, 4.6766e-02], + [-3.7350e-02, -1.8475e-02, -5.4803e-02, ..., -8.0363e-02, + -3.8325e-02, 2.4504e-02], + ..., + [-1.3524e-02, -2.1903e-02, -5.0529e-02, ..., -7.8673e-02, + 3.9988e-03, 3.4956e-02], + [-7.1014e-02, -5.2119e-02, -5.8450e-02, ..., -3.8427e-02, + -3.3056e-02, -5.7326e-02], + [ 7.2411e-05, -3.6497e-03, -4.7548e-02, ..., -3.4007e-02, + 2.7502e-02, -1.1708e-02]], device='cuda:0'), grad: tensor([[-0.0017, -0.0092, -0.0056, ..., -0.0033, 0.0001, 0.0014], + [-0.0028, 0.0009, 0.0020, ..., 0.0014, -0.0012, -0.0026], + [ 0.0049, 0.0054, 0.0073, ..., 0.0045, 0.0055, 0.0084], + ..., + [ 0.0030, 0.0010, 0.0012, ..., 0.0008, 0.0015, -0.0002], + [ 0.0107, 0.0086, 0.0114, ..., 0.0077, 0.0047, 0.0095], + [-0.0070, 0.0013, 0.0014, ..., -0.0016, -0.0041, -0.0063]], + device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0160, -0.0153, 0.0040, -0.0175, 0.0151, 0.0039, -0.0152, -0.0264, + 0.0101, -0.0102], device='cuda:0'), grad: tensor([-0.0657, 0.0157, 0.0338, -0.0115, -0.0109, 0.0180, -0.0084, 0.0161, + 0.0712, -0.0583], device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 224.90, cls_loss 0.6615 cls_loss_mapping 0.0301 cls_loss_causal 0.6189 re_mapping 0.0150 re_causal 0.0390 /// teacc 98.46 lr 0.00010000 +Epoch 70, weight, value: tensor([[ 0.0032, 0.0491, -0.0897, ..., -0.0487, 0.0373, 0.0110], + [-0.0421, -0.0743, 0.0440, ..., -0.0257, -0.0736, 0.0460], + [-0.0381, -0.0188, -0.0554, ..., -0.0800, -0.0382, 0.0244], + ..., + [-0.0136, -0.0222, -0.0509, ..., -0.0794, 0.0045, 0.0358], + [-0.0712, -0.0527, -0.0578, ..., -0.0386, -0.0338, -0.0579], + [ 0.0002, -0.0039, -0.0481, ..., -0.0343, 0.0270, -0.0118]], + device='cuda:0'), grad: tensor([[ 6.5184e-04, -3.2444e-03, -4.1809e-03, ..., -7.2670e-03, + -1.7290e-03, 5.7220e-04], + [-2.0027e-05, 2.6951e-03, 2.5425e-03, ..., 6.1913e-03, + 1.3256e-03, -9.2030e-04], + [ 3.4008e-03, 1.2751e-03, 8.7976e-04, ..., 1.1578e-03, + 2.0046e-03, 2.9449e-03], + ..., + [ 5.0926e-03, 8.0633e-04, 9.5129e-04, ..., 1.0939e-03, + 2.8076e-03, 1.7319e-03], + [ 3.4561e-03, 9.1457e-04, 3.0212e-03, ..., 2.8057e-03, + 1.7891e-03, 3.1662e-03], + [-2.1545e-02, -8.2932e-03, -6.0425e-03, ..., -8.9264e-03, + -1.3290e-02, -1.5587e-02]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0165, -0.0152, 0.0036, -0.0176, 0.0151, 0.0039, -0.0158, -0.0264, + 0.0104, -0.0101], device='cuda:0'), grad: tensor([ 0.0054, -0.0148, 0.0162, 0.0470, 0.0390, -0.0148, -0.0335, 0.0211, + 0.0201, -0.0858], device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 225.54, cls_loss 0.6001 cls_loss_mapping 0.0302 cls_loss_causal 0.5677 re_mapping 0.0144 re_causal 0.0374 /// teacc 98.21 lr 0.00010000 +Epoch 71, weight, value: tensor([[ 0.0022, 0.0490, -0.0901, ..., -0.0495, 0.0367, 0.0094], + [-0.0422, -0.0752, 0.0444, ..., -0.0254, -0.0728, 0.0461], + [-0.0368, -0.0189, -0.0562, ..., -0.0810, -0.0368, 0.0250], + ..., + [-0.0142, -0.0216, -0.0504, ..., -0.0782, 0.0038, 0.0370], + [-0.0709, -0.0530, -0.0578, ..., -0.0380, -0.0338, -0.0584], + [ 0.0007, -0.0045, -0.0480, ..., -0.0335, 0.0268, -0.0114]], + device='cuda:0'), grad: tensor([[ 3.8986e-03, 1.7462e-03, 1.1787e-03, ..., 6.7558e-03, + 5.5122e-03, 6.0196e-03], + [ 4.3411e-03, 1.3447e-03, 3.3283e-03, ..., 1.0193e-02, + 1.3838e-03, 7.6294e-03], + [ 2.0161e-03, 4.1270e-04, 1.2159e-03, ..., -7.1907e-04, + -3.6030e-03, 4.0703e-03], + ..., + [ 1.3647e-03, 8.7643e-04, 1.4198e-04, ..., 1.1597e-03, + 1.4915e-03, -3.4447e-03], + [-5.2795e-03, -2.7828e-03, -1.6994e-03, ..., -1.2589e-04, + 1.7494e-05, -2.8400e-03], + [-9.6054e-03, -2.1667e-03, -2.3918e-03, ..., -1.4486e-03, + -4.6959e-03, -1.2939e-02]], device='cuda:0') +Epoch 71, bias, value: tensor([ 0.0151, -0.0143, 0.0041, -0.0177, 0.0152, 0.0032, -0.0158, -0.0262, + 0.0101, -0.0094], device='cuda:0'), grad: tensor([ 0.0337, 0.0539, 0.0181, 0.0197, 0.0041, -0.0097, -0.0334, 0.0049, + -0.0350, -0.0564], device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 227.52, cls_loss 0.6537 cls_loss_mapping 0.0274 cls_loss_causal 0.6132 re_mapping 0.0144 re_causal 0.0357 /// teacc 98.41 lr 0.00010000 +Epoch 72, weight, value: tensor([[ 0.0022, 0.0498, -0.0902, ..., -0.0494, 0.0359, 0.0085], + [-0.0428, -0.0762, 0.0444, ..., -0.0254, -0.0736, 0.0459], + [-0.0368, -0.0190, -0.0564, ..., -0.0806, -0.0363, 0.0243], + ..., + [-0.0144, -0.0224, -0.0509, ..., -0.0797, 0.0036, 0.0374], + [-0.0711, -0.0529, -0.0574, ..., -0.0379, -0.0331, -0.0585], + [ 0.0006, -0.0044, -0.0480, ..., -0.0348, 0.0268, -0.0120]], + device='cuda:0'), grad: tensor([[ 2.7523e-03, 2.0921e-05, 4.2009e-04, ..., 9.7513e-04, + 1.1063e-03, 3.1548e-03], + [ 8.1348e-04, 1.9240e-04, 7.5388e-04, ..., 1.2236e-03, + 1.0777e-03, -2.4199e-04], + [-1.5926e-03, 6.3515e-04, 6.8092e-04, ..., 3.0022e-03, + 2.4140e-05, -2.5330e-03], + ..., + [ 1.4553e-03, 1.1387e-03, 1.3809e-03, ..., 2.7275e-03, + 1.2608e-03, 3.7975e-03], + [-1.3596e-02, 1.3268e-04, 1.1616e-03, ..., -7.4615e-03, + -8.9264e-03, -1.3474e-02], + [-1.4095e-03, -1.5850e-03, -7.2174e-03, ..., -3.2806e-03, + 1.8406e-03, -2.6131e-03]], device='cuda:0') +Epoch 72, bias, value: tensor([ 0.0150, -0.0146, 0.0041, -0.0176, 0.0149, 0.0039, -0.0164, -0.0262, + 0.0103, -0.0093], device='cuda:0'), grad: tensor([ 0.0186, -0.0056, 0.0048, 0.0130, -0.0178, 0.0446, 0.0371, 0.0096, + -0.0708, -0.0335], device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 226.78, cls_loss 0.6589 cls_loss_mapping 0.0299 cls_loss_causal 0.6160 re_mapping 0.0145 re_causal 0.0379 /// teacc 98.52 lr 0.00010000 +Epoch 73, weight, value: tensor([[ 0.0034, 0.0515, -0.0912, ..., -0.0505, 0.0381, 0.0085], + [-0.0439, -0.0764, 0.0448, ..., -0.0250, -0.0743, 0.0461], + [-0.0366, -0.0190, -0.0574, ..., -0.0823, -0.0371, 0.0249], + ..., + [-0.0142, -0.0229, -0.0514, ..., -0.0786, 0.0025, 0.0379], + [-0.0707, -0.0531, -0.0584, ..., -0.0379, -0.0330, -0.0590], + [ 0.0009, -0.0052, -0.0483, ..., -0.0354, 0.0276, -0.0129]], + device='cuda:0'), grad: tensor([[ 0.0087, 0.0033, 0.0003, ..., 0.0017, 0.0069, 0.0057], + [ 0.0021, 0.0002, -0.0024, ..., -0.0006, 0.0010, 0.0020], + [ 0.0011, -0.0004, 0.0012, ..., 0.0012, -0.0017, -0.0056], + ..., + [ 0.0039, 0.0027, 0.0004, ..., 0.0005, 0.0004, 0.0005], + [-0.0162, -0.0128, -0.0002, ..., -0.0006, -0.0060, -0.0135], + [ 0.0051, -0.0015, -0.0002, ..., 0.0008, -0.0012, 0.0060]], + device='cuda:0') +Epoch 73, bias, value: tensor([ 0.0151, -0.0147, 0.0035, -0.0180, 0.0158, 0.0038, -0.0167, -0.0265, + 0.0109, -0.0092], device='cuda:0'), grad: tensor([ 0.0379, 0.0142, 0.0014, 0.0213, -0.0690, 0.0523, -0.0143, 0.0240, + -0.0955, 0.0278], device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 225.85, cls_loss 0.6055 cls_loss_mapping 0.0385 cls_loss_causal 0.5686 re_mapping 0.0141 re_causal 0.0357 /// teacc 98.51 lr 0.00010000 +Epoch 74, weight, value: tensor([[ 0.0037, 0.0505, -0.0926, ..., -0.0516, 0.0386, 0.0092], + [-0.0442, -0.0778, 0.0444, ..., -0.0266, -0.0748, 0.0459], + [-0.0378, -0.0189, -0.0580, ..., -0.0821, -0.0376, 0.0253], + ..., + [-0.0145, -0.0238, -0.0519, ..., -0.0792, 0.0026, 0.0386], + [-0.0717, -0.0538, -0.0580, ..., -0.0377, -0.0343, -0.0603], + [ 0.0009, -0.0054, -0.0496, ..., -0.0362, 0.0282, -0.0125]], + device='cuda:0'), grad: tensor([[-1.0483e-02, -6.5756e-04, 2.5415e-04, ..., -6.1616e-06, + -9.0561e-03, 3.2959e-03], + [-1.0538e-03, 5.4270e-05, -3.6359e-04, ..., 7.5758e-05, + -1.3266e-03, -1.2312e-03], + [ 2.6474e-03, -4.9204e-05, 2.1207e-04, ..., 2.9349e-04, + 1.6279e-03, 6.7940e-03], + ..., + [-2.8655e-05, 8.0228e-05, 1.1891e-04, ..., 7.4387e-05, + 7.5817e-04, -2.1027e-02], + [ 4.8752e-03, 2.1305e-03, 3.1738e-03, ..., 2.8629e-03, + 2.9240e-03, 6.4659e-03], + [ 8.0338e-03, 6.8545e-05, 5.5408e-04, ..., 9.6369e-04, + 5.2452e-03, 5.7220e-03]], device='cuda:0') +Epoch 74, bias, value: tensor([ 0.0156, -0.0146, 0.0038, -0.0176, 0.0151, 0.0031, -0.0162, -0.0265, + 0.0106, -0.0090], device='cuda:0'), grad: tensor([-0.0034, 0.0142, 0.0268, -0.0548, -0.0248, 0.0308, -0.0236, -0.0467, + 0.0399, 0.0414], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 73---------------------------------------------------- +epoch 73, time 226.00, cls_loss 0.6491 cls_loss_mapping 0.0288 cls_loss_causal 0.6117 re_mapping 0.0141 re_causal 0.0377 /// teacc 98.57 lr 0.00010000 +Epoch 75, weight, value: tensor([[ 0.0043, 0.0517, -0.0928, ..., -0.0512, 0.0398, 0.0091], + [-0.0452, -0.0789, 0.0447, ..., -0.0276, -0.0759, 0.0471], + [-0.0383, -0.0195, -0.0579, ..., -0.0819, -0.0375, 0.0249], + ..., + [-0.0139, -0.0247, -0.0524, ..., -0.0796, 0.0024, 0.0394], + [-0.0716, -0.0521, -0.0572, ..., -0.0373, -0.0340, -0.0613], + [ 0.0003, -0.0053, -0.0490, ..., -0.0366, 0.0274, -0.0125]], + device='cuda:0'), grad: tensor([[ 5.0116e-04, -3.9840e-04, -3.6471e-06, ..., -6.6328e-04, + -7.5674e-04, 8.8930e-04], + [-1.4067e-03, 3.1382e-05, -2.0325e-04, ..., 4.8685e-04, + -1.1091e-03, -1.2321e-03], + [ 1.0004e-03, 1.1957e-04, 1.0139e-04, ..., 4.4823e-04, + 2.3341e-04, -1.8448e-02], + ..., + [ 2.3003e-03, 6.1214e-05, 4.6968e-05, ..., 4.4751e-04, + 1.6098e-03, 1.2650e-02], + [ 1.9531e-03, 1.0151e-04, 9.6798e-05, ..., 1.0071e-03, + 9.7036e-04, 7.1907e-03], + [ 6.1646e-03, 5.9903e-05, 3.7581e-05, ..., 4.5156e-04, + 1.1740e-03, 2.1706e-03]], device='cuda:0') +Epoch 75, bias, value: tensor([ 0.0150, -0.0148, 0.0033, -0.0174, 0.0153, 0.0032, -0.0160, -0.0258, + 0.0106, -0.0093], device='cuda:0'), grad: tensor([ 0.0041, -0.0125, -0.0207, -0.0147, -0.0186, -0.0118, 0.0213, 0.0269, + 0.0198, 0.0061], device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 226.02, cls_loss 0.6470 cls_loss_mapping 0.0289 cls_loss_causal 0.6064 re_mapping 0.0146 re_causal 0.0362 /// teacc 98.52 lr 0.00010000 +Epoch 76, weight, value: tensor([[ 0.0040, 0.0516, -0.0929, ..., -0.0509, 0.0394, 0.0093], + [-0.0454, -0.0787, 0.0457, ..., -0.0261, -0.0766, 0.0476], + [-0.0386, -0.0193, -0.0583, ..., -0.0821, -0.0379, 0.0258], + ..., + [-0.0139, -0.0263, -0.0542, ..., -0.0806, 0.0024, 0.0390], + [-0.0722, -0.0527, -0.0574, ..., -0.0376, -0.0339, -0.0623], + [ 0.0003, -0.0055, -0.0496, ..., -0.0379, 0.0279, -0.0128]], + device='cuda:0'), grad: tensor([[-0.0054, 0.0007, 0.0004, ..., -0.0006, -0.0027, -0.0086], + [ 0.0028, 0.0003, 0.0003, ..., 0.0021, 0.0009, 0.0040], + [ 0.0028, -0.0005, 0.0007, ..., 0.0021, 0.0014, 0.0034], + ..., + [ 0.0029, 0.0004, 0.0003, ..., 0.0014, 0.0019, 0.0042], + [-0.0074, -0.0007, 0.0035, ..., 0.0014, -0.0031, -0.0123], + [ 0.0036, 0.0012, 0.0004, ..., 0.0015, 0.0035, 0.0028]], + device='cuda:0') +Epoch 76, bias, value: tensor([ 0.0150, -0.0145, 0.0032, -0.0174, 0.0152, 0.0034, -0.0154, -0.0257, + 0.0098, -0.0094], device='cuda:0'), grad: tensor([-0.0402, 0.0179, 0.0231, 0.0224, 0.0099, -0.0202, -0.0079, 0.0245, + -0.0578, 0.0283], device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 225.89, cls_loss 0.6221 cls_loss_mapping 0.0255 cls_loss_causal 0.5842 re_mapping 0.0149 re_causal 0.0369 /// teacc 98.36 lr 0.00010000 +Epoch 77, weight, value: tensor([[ 0.0042, 0.0523, -0.0934, ..., -0.0514, 0.0395, 0.0096], + [-0.0462, -0.0789, 0.0466, ..., -0.0261, -0.0783, 0.0469], + [-0.0397, -0.0198, -0.0585, ..., -0.0831, -0.0382, 0.0262], + ..., + [-0.0137, -0.0270, -0.0553, ..., -0.0830, 0.0023, 0.0388], + [-0.0723, -0.0536, -0.0582, ..., -0.0365, -0.0349, -0.0619], + [ 0.0007, -0.0059, -0.0501, ..., -0.0378, 0.0279, -0.0120]], + device='cuda:0'), grad: tensor([[ 1.1797e-03, 2.2686e-04, 5.8079e-04, ..., 2.1210e-03, + -9.8324e-04, -1.9169e-04], + [ 1.5986e-04, 1.4126e-04, 7.0594e-06, ..., -1.5521e-04, + 5.5265e-04, -7.1287e-04], + [-3.5648e-03, -1.8263e-04, 8.6308e-05, ..., -1.3866e-03, + -3.1052e-03, -3.5400e-03], + ..., + [-1.0252e-03, 2.1136e-04, 2.6703e-05, ..., -3.2306e-04, + -5.7459e-04, -3.7327e-03], + [ 3.7694e-04, -7.0632e-05, 1.1235e-04, ..., 2.6836e-03, + 9.2936e-04, -5.7316e-04], + [-9.9487e-03, 9.1124e-04, 9.3281e-05, ..., 1.4153e-03, + -1.6602e-02, 4.2839e-03]], device='cuda:0') +Epoch 77, bias, value: tensor([ 0.0150, -0.0149, 0.0035, -0.0182, 0.0144, 0.0031, -0.0149, -0.0259, + 0.0102, -0.0081], device='cuda:0'), grad: tensor([-0.0041, -0.0065, -0.0101, -0.0045, 0.0214, -0.0222, 0.0293, -0.0163, + 0.0091, 0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 225.83, cls_loss 0.6368 cls_loss_mapping 0.0339 cls_loss_causal 0.5969 re_mapping 0.0147 re_causal 0.0357 /// teacc 98.19 lr 0.00010000 +Epoch 78, weight, value: tensor([[ 0.0028, 0.0521, -0.0955, ..., -0.0523, 0.0386, 0.0094], + [-0.0463, -0.0789, 0.0470, ..., -0.0253, -0.0779, 0.0466], + [-0.0401, -0.0202, -0.0595, ..., -0.0828, -0.0379, 0.0257], + ..., + [-0.0131, -0.0270, -0.0560, ..., -0.0840, 0.0023, 0.0397], + [-0.0717, -0.0546, -0.0585, ..., -0.0380, -0.0337, -0.0614], + [ 0.0009, -0.0065, -0.0498, ..., -0.0368, 0.0273, -0.0114]], + device='cuda:0'), grad: tensor([[-2.1820e-03, -7.0915e-03, 5.9724e-05, ..., 1.1969e-03, + -3.5362e-03, 2.5959e-03], + [ 3.2387e-03, 4.5681e-04, -8.6546e-05, ..., 2.6207e-03, + 2.9583e-03, 4.6005e-03], + [-3.1796e-03, 1.0185e-03, -1.8215e-04, ..., -1.2560e-03, + -7.9679e-04, -9.8114e-03], + ..., + [ 5.3406e-03, 6.3038e-04, 2.2233e-04, ..., 1.9398e-03, + 2.9335e-03, 6.6795e-03], + [-3.1376e-03, 5.0426e-05, 4.3988e-04, ..., -4.0398e-03, + -1.0162e-02, -6.8398e-03], + [-1.4887e-03, 8.5402e-04, 9.9778e-05, ..., -1.3781e-03, + 6.8998e-04, 2.7847e-03]], device='cuda:0') +Epoch 78, bias, value: tensor([ 0.0145, -0.0150, 0.0032, -0.0179, 0.0149, 0.0030, -0.0154, -0.0259, + 0.0107, -0.0081], device='cuda:0'), grad: tensor([ 0.0122, 0.0335, -0.0086, -0.0668, 0.0294, 0.0111, 0.0130, 0.0399, + -0.0778, 0.0141], device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 225.58, cls_loss 0.6026 cls_loss_mapping 0.0301 cls_loss_causal 0.5699 re_mapping 0.0145 re_causal 0.0356 /// teacc 98.41 lr 0.00010000 +Epoch 79, weight, value: tensor([[ 0.0032, 0.0530, -0.0966, ..., -0.0518, 0.0387, 0.0101], + [-0.0463, -0.0785, 0.0469, ..., -0.0262, -0.0781, 0.0471], + [-0.0417, -0.0210, -0.0580, ..., -0.0821, -0.0380, 0.0260], + ..., + [-0.0139, -0.0277, -0.0570, ..., -0.0837, 0.0020, 0.0398], + [-0.0724, -0.0544, -0.0586, ..., -0.0387, -0.0340, -0.0616], + [ 0.0012, -0.0056, -0.0492, ..., -0.0358, 0.0273, -0.0124]], + device='cuda:0'), grad: tensor([[ 6.0463e-04, -2.0254e-04, 5.5972e-07, ..., 2.8992e-04, + 3.4761e-04, 1.3590e-03], + [-1.4496e-03, -2.1038e-03, -2.6356e-06, ..., -3.4523e-04, + -7.2174e-03, -1.3626e-02], + [-6.3744e-03, 1.0433e-03, 1.5320e-07, ..., 2.1923e-04, + -8.3494e-04, 4.2458e-03], + ..., + [ 1.1339e-03, 1.7333e-04, 2.4363e-06, ..., 2.0969e-04, + 1.6236e-04, 6.0272e-04], + [ 4.0770e-04, 3.7932e-04, 5.1916e-05, ..., -3.2825e-03, + 1.1520e-03, 3.1929e-03], + [ 6.7177e-03, 7.2479e-03, 4.6939e-05, ..., -4.3750e-04, + 6.4201e-03, 1.4277e-03]], device='cuda:0') +Epoch 79, bias, value: tensor([ 0.0148, -0.0148, 0.0033, -0.0176, 0.0153, 0.0029, -0.0160, -0.0259, + 0.0110, -0.0089], device='cuda:0'), grad: tensor([ 0.0154, -0.0442, 0.0071, 0.0376, -0.0287, -0.0518, 0.0189, 0.0144, + 0.0129, 0.0184], device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 225.29, cls_loss 0.6236 cls_loss_mapping 0.0263 cls_loss_causal 0.5914 re_mapping 0.0136 re_causal 0.0351 /// teacc 98.51 lr 0.00010000 +Epoch 80, weight, value: tensor([[ 0.0031, 0.0529, -0.0976, ..., -0.0520, 0.0390, 0.0099], + [-0.0470, -0.0779, 0.0478, ..., -0.0254, -0.0782, 0.0469], + [-0.0414, -0.0201, -0.0579, ..., -0.0820, -0.0375, 0.0262], + ..., + [-0.0137, -0.0285, -0.0571, ..., -0.0832, 0.0014, 0.0405], + [-0.0735, -0.0559, -0.0602, ..., -0.0396, -0.0342, -0.0625], + [ 0.0016, -0.0056, -0.0497, ..., -0.0371, 0.0279, -0.0119]], + device='cuda:0'), grad: tensor([[ 2.5902e-03, 2.4095e-05, 4.3362e-05, ..., 9.4128e-04, + 1.5097e-03, 4.0474e-03], + [-4.7326e-04, 4.5747e-05, -9.0313e-04, ..., -9.7656e-04, + 4.0650e-04, -9.9468e-04], + [ 3.1128e-03, 8.6288e-03, 3.2883e-03, ..., -8.0228e-05, + 2.2106e-03, 7.8812e-03], + ..., + [-2.7828e-03, 2.5916e-04, 6.5136e-04, ..., 6.1333e-05, + -2.2488e-03, -1.7303e-02], + [ 1.7297e-04, 7.2479e-05, 9.4354e-05, ..., -7.2908e-04, + 3.7742e-04, 1.3571e-03], + [-3.1548e-03, 4.2230e-05, 3.3665e-04, ..., 1.0576e-03, + -2.1572e-03, 3.2101e-03]], device='cuda:0') +Epoch 80, bias, value: tensor([ 0.0152, -0.0149, 0.0036, -0.0177, 0.0155, 0.0019, -0.0155, -0.0253, + 0.0100, -0.0087], device='cuda:0'), grad: tensor([ 0.0246, -0.0092, 0.0599, -0.0625, -0.0020, 0.0147, 0.0262, -0.0573, + -0.0019, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 225.49, cls_loss 0.6389 cls_loss_mapping 0.0263 cls_loss_causal 0.6054 re_mapping 0.0131 re_causal 0.0332 /// teacc 98.55 lr 0.00010000 +Epoch 81, weight, value: tensor([[ 0.0037, 0.0528, -0.0970, ..., -0.0521, 0.0389, 0.0107], + [-0.0469, -0.0782, 0.0484, ..., -0.0249, -0.0788, 0.0473], + [-0.0417, -0.0199, -0.0580, ..., -0.0832, -0.0370, 0.0262], + ..., + [-0.0138, -0.0297, -0.0572, ..., -0.0830, 0.0021, 0.0406], + [-0.0733, -0.0570, -0.0605, ..., -0.0401, -0.0344, -0.0621], + [ 0.0018, -0.0043, -0.0510, ..., -0.0374, 0.0280, -0.0118]], + device='cuda:0'), grad: tensor([[ 1.6479e-03, -2.4271e-04, 1.4651e-04, ..., 4.2701e-04, + 9.0408e-04, 2.4185e-03], + [-2.8205e-04, 2.4867e-04, 1.2720e-04, ..., 2.5675e-05, + 4.6253e-04, 1.0900e-05], + [-3.5610e-03, 1.7815e-03, 2.0123e-03, ..., 8.7881e-04, + -2.2793e-03, -7.7591e-03], + ..., + [-3.7632e-03, -2.2614e-04, 9.4235e-05, ..., 1.6153e-04, + -1.6813e-03, -3.3550e-03], + [ 2.6016e-03, 9.3746e-04, 7.7724e-04, ..., 1.3266e-03, + 1.5802e-03, 3.1528e-03], + [ 3.2253e-03, 2.1350e-04, 1.6451e-04, ..., 9.0742e-04, + 1.7881e-03, 2.0466e-03]], device='cuda:0') +Epoch 81, bias, value: tensor([ 0.0154, -0.0146, 0.0036, -0.0177, 0.0149, 0.0017, -0.0160, -0.0252, + 0.0103, -0.0084], device='cuda:0'), grad: tensor([ 0.0150, -0.0113, -0.0194, 0.0143, -0.0313, 0.0368, -0.0137, -0.0287, + 0.0215, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 221.55, cls_loss 0.6437 cls_loss_mapping 0.0243 cls_loss_causal 0.6121 re_mapping 0.0141 re_causal 0.0356 /// teacc 98.16 lr 0.00010000 +Epoch 82, weight, value: tensor([[ 0.0048, 0.0531, -0.0969, ..., -0.0522, 0.0401, 0.0102], + [-0.0456, -0.0781, 0.0487, ..., -0.0244, -0.0788, 0.0468], + [-0.0420, -0.0204, -0.0582, ..., -0.0841, -0.0375, 0.0274], + ..., + [-0.0144, -0.0307, -0.0584, ..., -0.0839, 0.0013, 0.0400], + [-0.0746, -0.0568, -0.0604, ..., -0.0403, -0.0351, -0.0619], + [ 0.0018, -0.0040, -0.0514, ..., -0.0380, 0.0271, -0.0112]], + device='cuda:0'), grad: tensor([[-1.6546e-03, -1.8797e-03, 5.3912e-05, ..., -4.1389e-04, + -3.4833e-04, -1.8053e-03], + [ 5.0087e-03, 6.5517e-04, 8.5592e-05, ..., 8.6975e-04, + 1.9646e-03, 5.6000e-03], + [-9.8877e-03, -8.8358e-04, 1.8167e-04, ..., 9.5177e-04, + -4.8828e-04, -9.7351e-03], + ..., + [-2.4872e-03, 3.0732e-04, -2.5868e-04, ..., 2.1088e-04, + -2.5997e-03, 3.1567e-04], + [ 2.6970e-03, 8.5373e-03, 1.3745e-04, ..., 5.1880e-04, + 1.4067e-03, 2.9430e-03], + [-1.5993e-03, 2.9397e-04, 1.4269e-04, ..., 2.4021e-04, + -9.2697e-04, -6.7377e-04]], device='cuda:0') +Epoch 82, bias, value: tensor([ 0.0157, -0.0142, 0.0038, -0.0180, 0.0153, 0.0025, -0.0164, -0.0260, + 0.0097, -0.0083], device='cuda:0'), grad: tensor([-0.0186, 0.0387, -0.0581, 0.0115, 0.0228, -0.0101, -0.0297, 0.0151, + 0.0640, -0.0356], device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 221.84, cls_loss 0.6295 cls_loss_mapping 0.0250 cls_loss_causal 0.5952 re_mapping 0.0136 re_causal 0.0361 /// teacc 98.23 lr 0.00010000 +Epoch 83, weight, value: tensor([[ 0.0045, 0.0527, -0.0976, ..., -0.0517, 0.0400, 0.0095], + [-0.0450, -0.0765, 0.0488, ..., -0.0248, -0.0770, 0.0466], + [-0.0424, -0.0196, -0.0583, ..., -0.0842, -0.0377, 0.0283], + ..., + [-0.0137, -0.0314, -0.0591, ..., -0.0844, 0.0019, 0.0397], + [-0.0743, -0.0581, -0.0605, ..., -0.0403, -0.0350, -0.0619], + [ 0.0018, -0.0042, -0.0513, ..., -0.0381, 0.0273, -0.0107]], + device='cuda:0'), grad: tensor([[-2.1191e-03, -7.3891e-03, -1.1749e-03, ..., 1.8191e-04, + -6.6376e-03, -2.2907e-03], + [-1.4353e-03, -1.5078e-03, -4.6635e-04, ..., 1.4162e-04, + 3.6550e-04, -4.4022e-03], + [ 2.4452e-03, 1.7567e-03, 7.5674e-04, ..., 1.1415e-03, + 1.0843e-03, 5.4016e-03], + ..., + [ 2.2488e-03, 9.5844e-04, 4.5872e-04, ..., 6.2943e-04, + 4.8661e-04, 3.7823e-03], + [ 2.6302e-03, 1.4210e-03, -6.3229e-04, ..., -6.1417e-04, + 1.1187e-03, 3.0689e-03], + [-6.6795e-03, 6.9904e-04, 5.6662e-06, ..., 9.1171e-04, + -1.1015e-03, -8.3771e-03]], device='cuda:0') +Epoch 83, bias, value: tensor([ 0.0143, -0.0140, 0.0042, -0.0180, 0.0143, 0.0024, -0.0155, -0.0256, + 0.0096, -0.0079], device='cuda:0'), grad: tensor([-0.0640, -0.0206, 0.0396, 0.0474, -0.0311, 0.0184, 0.0287, 0.0317, + 0.0261, -0.0760], device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 220.34, cls_loss 0.6274 cls_loss_mapping 0.0272 cls_loss_causal 0.5940 re_mapping 0.0129 re_causal 0.0335 /// teacc 98.46 lr 0.00010000 +Epoch 84, weight, value: tensor([[ 0.0040, 0.0532, -0.0979, ..., -0.0532, 0.0396, 0.0104], + [-0.0453, -0.0754, 0.0488, ..., -0.0249, -0.0771, 0.0458], + [-0.0426, -0.0205, -0.0583, ..., -0.0843, -0.0382, 0.0285], + ..., + [-0.0141, -0.0317, -0.0592, ..., -0.0845, 0.0015, 0.0393], + [-0.0743, -0.0594, -0.0610, ..., -0.0411, -0.0359, -0.0613], + [ 0.0025, -0.0046, -0.0516, ..., -0.0386, 0.0282, -0.0112]], + device='cuda:0'), grad: tensor([[ 9.2010e-03, 3.6316e-03, 7.4565e-05, ..., 4.3321e-04, + 7.5302e-03, 9.9468e-04], + [ 8.5354e-05, 6.1274e-04, 1.0461e-04, ..., 1.2560e-03, + 1.9753e-04, -7.4911e-04], + [ 3.8261e-03, 1.8425e-03, 4.1389e-04, ..., 2.6493e-03, + 2.2507e-03, 5.3787e-03], + ..., + [-4.8294e-03, 4.3440e-04, 3.3331e-04, ..., 1.4858e-03, + -4.9782e-03, -1.4853e-04], + [-2.9526e-03, -3.1204e-03, -1.1258e-05, ..., 6.7902e-04, + -4.4727e-04, -1.0735e-02], + [ 1.2161e-02, 4.3297e-03, 7.6485e-04, ..., 2.5654e-03, + 8.7814e-03, 1.3702e-02]], device='cuda:0') +Epoch 84, bias, value: tensor([ 0.0148, -0.0138, 0.0036, -0.0168, 0.0150, 0.0011, -0.0154, -0.0261, + 0.0101, -0.0085], device='cuda:0'), grad: tensor([ 0.0404, 0.0041, 0.0282, 0.0127, -0.0054, -0.0136, -0.0673, -0.0090, + -0.0649, 0.0748], device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 217.17, cls_loss 0.6361 cls_loss_mapping 0.0267 cls_loss_causal 0.6014 re_mapping 0.0127 re_causal 0.0330 /// teacc 98.47 lr 0.00010000 +Epoch 85, weight, value: tensor([[ 0.0037, 0.0539, -0.0980, ..., -0.0527, 0.0402, 0.0101], + [-0.0453, -0.0764, 0.0492, ..., -0.0261, -0.0777, 0.0463], + [-0.0429, -0.0209, -0.0595, ..., -0.0855, -0.0379, 0.0285], + ..., + [-0.0152, -0.0327, -0.0593, ..., -0.0853, 0.0009, 0.0396], + [-0.0746, -0.0604, -0.0612, ..., -0.0402, -0.0362, -0.0615], + [ 0.0031, -0.0032, -0.0519, ..., -0.0384, 0.0288, -0.0110]], + device='cuda:0'), grad: tensor([[-2.0683e-04, 1.9562e-04, 2.1362e-04, ..., 4.3011e-04, + 1.0061e-03, -4.2534e-03], + [ 1.9217e-03, 5.0545e-04, 4.8971e-04, ..., 9.0063e-05, + 1.5316e-03, -1.7838e-02], + [ 1.5678e-03, 3.0684e-04, 5.0306e-04, ..., 6.5041e-04, + 1.4114e-03, 2.0905e-02], + ..., + [-5.8594e-03, -1.3804e-04, -2.6245e-03, ..., -3.9482e-03, + -5.0392e-03, 9.3794e-04], + [-2.2280e-04, -3.1853e-03, -6.7949e-05, ..., -4.5824e-04, + -1.8101e-03, -1.1566e-02], + [ 1.8644e-03, 1.8044e-03, 4.2868e-04, ..., 8.6975e-04, + 2.4948e-03, 5.7564e-03]], device='cuda:0') +Epoch 85, bias, value: tensor([ 0.0155, -0.0135, 0.0032, -0.0164, 0.0146, 0.0016, -0.0164, -0.0268, + 0.0101, -0.0079], device='cuda:0'), grad: tensor([-0.0114, -0.0029, 0.0311, 0.0186, 0.0053, 0.0143, -0.0159, -0.0334, + -0.0318, 0.0262], device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 220.89, cls_loss 0.6031 cls_loss_mapping 0.0220 cls_loss_causal 0.5674 re_mapping 0.0136 re_causal 0.0348 /// teacc 98.46 lr 0.00010000 +Epoch 86, weight, value: tensor([[ 0.0035, 0.0557, -0.0983, ..., -0.0523, 0.0398, 0.0101], + [-0.0462, -0.0767, 0.0488, ..., -0.0269, -0.0790, 0.0456], + [-0.0439, -0.0223, -0.0602, ..., -0.0861, -0.0387, 0.0287], + ..., + [-0.0138, -0.0337, -0.0590, ..., -0.0852, 0.0015, 0.0402], + [-0.0737, -0.0610, -0.0612, ..., -0.0416, -0.0360, -0.0616], + [ 0.0029, -0.0031, -0.0517, ..., -0.0386, 0.0288, -0.0107]], + device='cuda:0'), grad: tensor([[ 1.2493e-03, -1.2665e-03, 5.4622e-07, ..., -4.1342e-04, + 2.7943e-04, 1.5488e-03], + [ 1.4906e-03, 3.2276e-05, 1.8999e-07, ..., 1.2732e-04, + 8.6784e-04, 1.2541e-03], + [ 1.2627e-03, 3.9935e-05, 1.0151e-06, ..., -3.7879e-05, + 9.1696e-04, 8.8882e-04], + ..., + [ 2.4147e-03, 2.0897e-04, 7.6741e-07, ..., 1.2839e-04, + 1.7014e-03, 1.6232e-03], + [-1.8053e-03, 7.9453e-05, 6.1318e-06, ..., 1.5068e-04, + -1.5287e-03, -1.6289e-03], + [ 2.1515e-02, 3.0304e-02, 2.0325e-05, ..., -4.8614e-04, + 3.0624e-02, 8.0156e-04]], device='cuda:0') +Epoch 86, bias, value: tensor([ 0.0159, -0.0144, 0.0030, -0.0172, 0.0149, 0.0017, -0.0161, -0.0266, + 0.0103, -0.0075], device='cuda:0'), grad: tensor([ 0.0146, 0.0135, 0.0135, -0.0789, -0.0280, -0.0140, 0.0231, 0.0215, + -0.0135, 0.0482], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 85---------------------------------------------------- +epoch 85, time 217.80, cls_loss 0.6152 cls_loss_mapping 0.0256 cls_loss_causal 0.5798 re_mapping 0.0131 re_causal 0.0326 /// teacc 98.60 lr 0.00010000 +Epoch 87, weight, value: tensor([[ 0.0032, 0.0554, -0.0987, ..., -0.0532, 0.0393, 0.0099], + [-0.0460, -0.0767, 0.0481, ..., -0.0281, -0.0783, 0.0456], + [-0.0444, -0.0225, -0.0607, ..., -0.0862, -0.0383, 0.0290], + ..., + [-0.0140, -0.0346, -0.0591, ..., -0.0841, 0.0012, 0.0413], + [-0.0738, -0.0629, -0.0617, ..., -0.0412, -0.0367, -0.0623], + [ 0.0027, -0.0036, -0.0515, ..., -0.0388, 0.0285, -0.0108]], + device='cuda:0'), grad: tensor([[ 3.6774e-03, -1.1665e-02, -1.3189e-03, ..., -3.6106e-03, + -1.0483e-02, 4.5891e-03], + [-4.9019e-03, 1.2314e-04, 9.1851e-05, ..., 3.5548e-04, + -9.5520e-03, -4.0550e-03], + [-3.0112e-04, 6.7520e-04, 2.1708e-04, ..., 2.2531e-04, + 8.4400e-04, 5.7640e-03], + ..., + [ 1.9550e-03, 1.7858e-04, 1.2624e-04, ..., 4.2248e-04, + 1.2159e-03, -2.4090e-03], + [ 1.1415e-03, -1.9255e-03, -2.3384e-03, ..., -6.0797e-04, + -8.4829e-04, 3.3360e-03], + [ 1.6327e-03, 7.5874e-03, 2.8362e-03, ..., 3.3360e-03, + 1.0109e-02, 5.0125e-03]], device='cuda:0') +Epoch 87, bias, value: tensor([ 0.0154, -0.0138, 0.0030, -0.0179, 0.0146, 0.0018, -0.0165, -0.0258, + 0.0108, -0.0076], device='cuda:0'), grad: tensor([ 0.0153, -0.0347, -0.0061, 0.0277, -0.0399, 0.0094, -0.0449, 0.0223, + 0.0106, 0.0403], device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 217.21, cls_loss 0.6105 cls_loss_mapping 0.0266 cls_loss_causal 0.5785 re_mapping 0.0127 re_causal 0.0328 /// teacc 98.50 lr 0.00010000 +Epoch 88, weight, value: tensor([[ 0.0035, 0.0554, -0.0997, ..., -0.0536, 0.0402, 0.0106], + [-0.0462, -0.0773, 0.0480, ..., -0.0280, -0.0780, 0.0469], + [-0.0443, -0.0233, -0.0612, ..., -0.0861, -0.0386, 0.0294], + ..., + [-0.0146, -0.0350, -0.0581, ..., -0.0837, 0.0003, 0.0412], + [-0.0742, -0.0624, -0.0623, ..., -0.0419, -0.0359, -0.0625], + [ 0.0037, -0.0039, -0.0531, ..., -0.0388, 0.0289, -0.0102]], + device='cuda:0'), grad: tensor([[-8.0204e-04, 1.0692e-06, 4.2820e-04, ..., -9.6035e-04, + -2.7537e-05, -1.5955e-03], + [ 1.0710e-03, 3.0899e-04, -2.0237e-03, ..., 1.6797e-04, + 1.8990e-04, -4.4136e-03], + [-6.6233e-04, 1.3196e-04, -1.8203e-04, ..., 5.3835e-04, + -2.3603e-04, -5.5275e-03], + ..., + [-5.2452e-04, 2.5678e-04, 8.6975e-04, ..., 7.5626e-04, + 1.1587e-04, -3.5896e-03], + [ 1.0128e-03, 5.7077e-04, 1.2922e-03, ..., 1.2283e-03, + 2.6727e-04, 5.4893e-03], + [ 9.2649e-04, 2.5702e-04, 7.8249e-04, ..., 7.0286e-04, + 1.2255e-04, 2.6207e-03]], device='cuda:0') +Epoch 88, bias, value: tensor([ 0.0161, -0.0131, 0.0035, -0.0184, 0.0145, 0.0014, -0.0176, -0.0263, + 0.0107, -0.0069], device='cuda:0'), grad: tensor([-0.0190, -0.0072, -0.0499, 0.0154, -0.0151, -0.0102, 0.0140, 0.0248, + 0.0288, 0.0184], device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 217.17, cls_loss 0.6363 cls_loss_mapping 0.0210 cls_loss_causal 0.5993 re_mapping 0.0132 re_causal 0.0342 /// teacc 98.48 lr 0.00010000 +Epoch 89, weight, value: tensor([[ 0.0041, 0.0563, -0.0995, ..., -0.0541, 0.0412, 0.0111], + [-0.0474, -0.0778, 0.0488, ..., -0.0291, -0.0789, 0.0468], + [-0.0447, -0.0227, -0.0627, ..., -0.0869, -0.0391, 0.0298], + ..., + [-0.0147, -0.0362, -0.0590, ..., -0.0843, 0.0002, 0.0408], + [-0.0751, -0.0629, -0.0622, ..., -0.0413, -0.0358, -0.0628], + [ 0.0044, -0.0040, -0.0538, ..., -0.0397, 0.0285, -0.0104]], + device='cuda:0'), grad: tensor([[-9.4032e-04, 9.4473e-05, 1.5199e-04, ..., -4.6921e-04, + 1.0341e-04, -1.8864e-03], + [ 1.8024e-03, 1.9538e-04, 6.9237e-04, ..., 7.0667e-04, + 1.1778e-03, 1.2560e-03], + [ 1.9264e-03, 5.5361e-04, 3.4404e-04, ..., 8.4877e-04, + 1.4305e-03, 1.5459e-03], + ..., + [ 1.6241e-03, -1.7488e-04, 2.7180e-04, ..., 7.7915e-04, + 5.5075e-04, 4.5443e-04], + [ 1.3485e-03, 4.2129e-04, -2.7370e-04, ..., 7.2813e-04, + 8.5545e-04, 8.9884e-04], + [-8.0719e-03, 5.1117e-04, 1.4582e-03, ..., -1.2074e-03, + -6.2675e-03, -2.1515e-03]], device='cuda:0') +Epoch 89, bias, value: tensor([ 0.0162, -0.0141, 0.0039, -0.0175, 0.0152, 0.0012, -0.0167, -0.0269, + 0.0103, -0.0076], device='cuda:0'), grad: tensor([-0.0155, 0.0225, 0.0146, 0.0088, 0.0484, 0.0139, -0.0597, 0.0128, + 0.0004, -0.0462], device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 217.36, cls_loss 0.6075 cls_loss_mapping 0.0221 cls_loss_causal 0.5735 re_mapping 0.0126 re_causal 0.0324 /// teacc 98.60 lr 0.00010000 +Epoch 90, weight, value: tensor([[ 0.0042, 0.0568, -0.1008, ..., -0.0548, 0.0416, 0.0101], + [-0.0472, -0.0778, 0.0500, ..., -0.0283, -0.0790, 0.0460], + [-0.0447, -0.0230, -0.0629, ..., -0.0872, -0.0387, 0.0302], + ..., + [-0.0154, -0.0371, -0.0601, ..., -0.0845, 0.0001, 0.0413], + [-0.0748, -0.0644, -0.0624, ..., -0.0413, -0.0367, -0.0630], + [ 0.0043, -0.0034, -0.0527, ..., -0.0387, 0.0292, -0.0106]], + device='cuda:0'), grad: tensor([[-6.4049e-03, -6.6338e-03, -3.9139e-03, ..., -7.8735e-03, + -4.9782e-03, 1.1015e-03], + [ 1.1797e-03, 3.7193e-04, -3.3913e-03, ..., 6.6233e-04, + 7.4100e-04, 2.8610e-05], + [ 7.6103e-04, 2.8172e-03, 9.3639e-05, ..., 2.4223e-03, + 2.6035e-03, 2.5806e-03], + ..., + [ 2.2011e-03, 1.9848e-04, 1.7309e-04, ..., 4.1223e-04, + 5.3501e-04, 4.9210e-03], + [-3.9139e-03, -8.7051e-03, -1.6851e-03, ..., -1.2016e-02, + -6.6261e-03, -5.8594e-03], + [-4.1962e-03, 1.4839e-03, 2.1152e-03, ..., 2.4967e-03, + -1.0719e-03, 1.8625e-03]], device='cuda:0') +Epoch 90, bias, value: tensor([ 0.0157, -0.0134, 0.0038, -0.0168, 0.0142, 0.0014, -0.0170, -0.0279, + 0.0106, -0.0068], device='cuda:0'), grad: tensor([-0.0215, 0.0107, 0.0009, 0.0046, 0.0340, 0.0323, -0.0218, 0.0296, + -0.0602, -0.0086], device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 217.55, cls_loss 0.6000 cls_loss_mapping 0.0214 cls_loss_causal 0.5592 re_mapping 0.0129 re_causal 0.0329 /// teacc 98.49 lr 0.00010000 +Epoch 91, weight, value: tensor([[ 0.0051, 0.0577, -0.0997, ..., -0.0556, 0.0429, 0.0094], + [-0.0479, -0.0782, 0.0504, ..., -0.0293, -0.0798, 0.0460], + [-0.0457, -0.0245, -0.0631, ..., -0.0877, -0.0398, 0.0310], + ..., + [-0.0154, -0.0367, -0.0604, ..., -0.0844, 0.0004, 0.0410], + [-0.0756, -0.0645, -0.0624, ..., -0.0404, -0.0373, -0.0627], + [ 0.0055, -0.0034, -0.0533, ..., -0.0385, 0.0294, -0.0117]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0021, 0.0011, ..., 0.0021, 0.0017, 0.0020], + [-0.0005, 0.0002, 0.0002, ..., -0.0030, 0.0007, -0.0009], + [ 0.0017, -0.0021, 0.0002, ..., 0.0010, 0.0005, -0.0012], + ..., + [ 0.0014, 0.0006, 0.0003, ..., 0.0008, 0.0009, -0.0020], + [ 0.0017, 0.0010, 0.0003, ..., 0.0009, 0.0010, 0.0017], + [ 0.0029, 0.0002, -0.0025, ..., 0.0012, -0.0011, 0.0032]], + device='cuda:0') +Epoch 91, bias, value: tensor([ 0.0167, -0.0137, 0.0033, -0.0163, 0.0147, 0.0005, -0.0167, -0.0277, + 0.0106, -0.0074], device='cuda:0'), grad: tensor([ 0.0202, -0.0138, 0.0250, -0.0046, -0.0283, -0.0281, 0.0092, -0.0051, + 0.0195, 0.0059], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 90---------------------------------------------------- +epoch 90, time 217.85, cls_loss 0.6132 cls_loss_mapping 0.0237 cls_loss_causal 0.5793 re_mapping 0.0126 re_causal 0.0317 /// teacc 98.71 lr 0.00010000 +Epoch 92, weight, value: tensor([[ 0.0047, 0.0578, -0.1012, ..., -0.0574, 0.0425, 0.0093], + [-0.0484, -0.0779, 0.0522, ..., -0.0282, -0.0801, 0.0462], + [-0.0460, -0.0249, -0.0632, ..., -0.0874, -0.0394, 0.0307], + ..., + [-0.0151, -0.0368, -0.0618, ..., -0.0851, 0.0008, 0.0416], + [-0.0753, -0.0649, -0.0615, ..., -0.0394, -0.0369, -0.0632], + [ 0.0060, -0.0038, -0.0527, ..., -0.0396, 0.0292, -0.0121]], + device='cuda:0'), grad: tensor([[-8.0729e-04, 2.0275e-03, -5.3978e-04, ..., -1.2970e-03, + -1.4515e-03, 1.5821e-03], + [-3.7336e-04, 2.7013e-04, -2.1982e-04, ..., 2.1350e-04, + 4.0352e-05, 1.9388e-03], + [ 3.4027e-03, -5.1069e-04, 3.0667e-05, ..., -6.4468e-04, + -2.4796e-04, 4.2000e-03], + ..., + [-2.6283e-03, -1.0788e-02, 3.0115e-05, ..., 2.3019e-04, + 9.2149e-05, -7.3051e-03], + [ 6.1226e-04, 2.7351e-03, 1.4544e-04, ..., 3.6359e-04, + 7.6771e-05, -1.6060e-03], + [ 1.5421e-03, 4.1733e-03, 1.2093e-03, ..., 3.8147e-04, + 8.9586e-05, 1.9550e-03]], device='cuda:0') +Epoch 92, bias, value: tensor([ 0.0160, -0.0130, 0.0028, -0.0167, 0.0149, 0.0005, -0.0164, -0.0273, + 0.0105, -0.0073], device='cuda:0'), grad: tensor([ 0.0122, 0.0071, -0.0081, -0.0031, -0.0230, 0.0289, 0.0069, -0.0330, + -0.0060, 0.0181], device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 217.04, cls_loss 0.6280 cls_loss_mapping 0.0236 cls_loss_causal 0.5899 re_mapping 0.0123 re_causal 0.0317 /// teacc 98.50 lr 0.00010000 +Epoch 93, weight, value: tensor([[ 0.0052, 0.0587, -0.1015, ..., -0.0576, 0.0432, 0.0091], + [-0.0485, -0.0784, 0.0534, ..., -0.0285, -0.0802, 0.0469], + [-0.0464, -0.0256, -0.0638, ..., -0.0874, -0.0397, 0.0301], + ..., + [-0.0140, -0.0359, -0.0624, ..., -0.0849, 0.0009, 0.0424], + [-0.0756, -0.0654, -0.0618, ..., -0.0394, -0.0370, -0.0636], + [ 0.0054, -0.0052, -0.0529, ..., -0.0402, 0.0293, -0.0126]], + device='cuda:0'), grad: tensor([[-0.0062, -0.0005, 0.0001, ..., -0.0020, -0.0027, -0.0036], + [ 0.0015, 0.0003, 0.0001, ..., 0.0010, 0.0004, 0.0084], + [ 0.0013, 0.0003, 0.0002, ..., 0.0015, 0.0004, 0.0032], + ..., + [ 0.0031, 0.0010, 0.0003, ..., 0.0008, 0.0007, 0.0028], + [-0.0018, -0.0011, -0.0012, ..., -0.0013, -0.0001, -0.0160], + [-0.0040, -0.0018, -0.0002, ..., 0.0004, -0.0003, -0.0014]], + device='cuda:0') +Epoch 93, bias, value: tensor([ 0.0156, -0.0137, 0.0024, -0.0165, 0.0150, 0.0009, -0.0163, -0.0257, + 0.0104, -0.0081], device='cuda:0'), grad: tensor([-0.0462, 0.0102, 0.0228, -0.0161, 0.0057, 0.0158, 0.0525, 0.0241, + -0.0603, -0.0084], device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 217.28, cls_loss 0.5822 cls_loss_mapping 0.0210 cls_loss_causal 0.5488 re_mapping 0.0124 re_causal 0.0307 /// teacc 98.62 lr 0.00010000 +Epoch 94, weight, value: tensor([[ 0.0050, 0.0593, -0.1024, ..., -0.0581, 0.0437, 0.0087], + [-0.0483, -0.0798, 0.0541, ..., -0.0288, -0.0806, 0.0465], + [-0.0465, -0.0259, -0.0645, ..., -0.0875, -0.0394, 0.0298], + ..., + [-0.0142, -0.0380, -0.0635, ..., -0.0850, 0.0005, 0.0428], + [-0.0760, -0.0667, -0.0624, ..., -0.0394, -0.0374, -0.0634], + [ 0.0056, -0.0039, -0.0513, ..., -0.0403, 0.0295, -0.0117]], + device='cuda:0'), grad: tensor([[ 0.0013, 0.0008, 0.0003, ..., 0.0008, 0.0002, 0.0018], + [-0.0055, -0.0022, 0.0005, ..., 0.0007, -0.0003, -0.0121], + [ 0.0032, 0.0017, -0.0076, ..., -0.0012, -0.0020, 0.0067], + ..., + [ 0.0031, 0.0026, 0.0003, ..., 0.0002, 0.0012, -0.0007], + [ 0.0015, 0.0011, 0.0003, ..., 0.0009, 0.0006, -0.0003], + [ 0.0027, 0.0021, 0.0002, ..., 0.0006, 0.0016, 0.0092]], + device='cuda:0') +Epoch 94, bias, value: tensor([ 0.0148, -0.0136, 0.0027, -0.0170, 0.0150, 0.0013, -0.0162, -0.0264, + 0.0104, -0.0071], device='cuda:0'), grad: tensor([ 0.0108, -0.0434, 0.0313, -0.0247, 0.0018, -0.0067, -0.0097, 0.0100, + -0.0118, 0.0424], device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 217.70, cls_loss 0.6134 cls_loss_mapping 0.0191 cls_loss_causal 0.5813 re_mapping 0.0134 re_causal 0.0341 /// teacc 98.60 lr 0.00010000 +Epoch 95, weight, value: tensor([[ 0.0049, 0.0591, -0.1024, ..., -0.0577, 0.0440, 0.0077], + [-0.0484, -0.0802, 0.0540, ..., -0.0289, -0.0802, 0.0464], + [-0.0467, -0.0253, -0.0646, ..., -0.0877, -0.0403, 0.0301], + ..., + [-0.0152, -0.0398, -0.0649, ..., -0.0853, -0.0007, 0.0429], + [-0.0762, -0.0667, -0.0620, ..., -0.0399, -0.0377, -0.0636], + [ 0.0061, -0.0031, -0.0509, ..., -0.0402, 0.0298, -0.0112]], + device='cuda:0'), grad: tensor([[ 1.9741e-03, 8.6355e-04, 5.1111e-05, ..., 4.6587e-04, + 1.4067e-03, 1.4315e-03], + [-1.8835e-03, 1.0854e-04, -3.9816e-04, ..., -6.8932e-03, + -3.8280e-03, 7.3719e-04], + [-9.7580e-03, -4.2648e-03, 1.1981e-04, ..., 4.7541e-04, + -7.7133e-03, -5.6458e-03], + ..., + [ 1.8940e-03, 2.9659e-04, -7.3314e-05, ..., 2.7180e-04, + 9.6226e-04, 1.0567e-03], + [ 9.5069e-05, 9.1028e-04, 1.3089e-04, ..., 2.0790e-04, + 1.3132e-03, -8.2922e-04], + [ 1.3161e-03, 4.9114e-04, -1.6427e-04, ..., 5.7554e-04, + 6.9332e-04, 1.8787e-03]], device='cuda:0') +Epoch 95, bias, value: tensor([ 0.0144, -0.0131, 0.0024, -0.0173, 0.0146, 0.0014, -0.0161, -0.0262, + 0.0096, -0.0059], device='cuda:0'), grad: tensor([ 0.0255, -0.0173, -0.1020, -0.0096, -0.0121, 0.0328, 0.0317, 0.0243, + 0.0035, 0.0231], device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 217.12, cls_loss 0.5818 cls_loss_mapping 0.0238 cls_loss_causal 0.5532 re_mapping 0.0124 re_causal 0.0313 /// teacc 98.53 lr 0.00010000 +Epoch 96, weight, value: tensor([[ 0.0053, 0.0595, -0.1027, ..., -0.0576, 0.0442, 0.0079], + [-0.0491, -0.0801, 0.0536, ..., -0.0293, -0.0807, 0.0469], + [-0.0465, -0.0260, -0.0647, ..., -0.0875, -0.0405, 0.0303], + ..., + [-0.0164, -0.0412, -0.0654, ..., -0.0862, -0.0018, 0.0435], + [-0.0764, -0.0655, -0.0615, ..., -0.0397, -0.0368, -0.0635], + [ 0.0066, -0.0025, -0.0508, ..., -0.0404, 0.0294, -0.0123]], + device='cuda:0'), grad: tensor([[ 1.9045e-03, 1.0014e-03, 1.0514e-04, ..., 1.2465e-03, + 1.0777e-03, 2.5730e-03], + [ 1.5717e-03, 3.3927e-04, -2.9507e-03, ..., -1.7099e-03, + 6.7568e-04, -6.3858e-03], + [ 2.0599e-03, 3.0041e-03, 2.7618e-03, ..., 3.2654e-03, + 6.9809e-04, 5.8556e-03], + ..., + [-1.2074e-03, 1.7095e-04, 1.9431e-04, ..., -3.4451e-05, + -7.9775e-04, -1.1625e-03], + [ 1.9588e-03, -2.1057e-03, -1.3840e-04, ..., 3.3826e-05, + 1.0691e-03, 4.8904e-03], + [-5.5656e-03, -1.2102e-03, 2.4402e-04, ..., 5.1785e-04, + -2.2202e-03, 1.5249e-03]], device='cuda:0') +Epoch 96, bias, value: tensor([ 0.0153, -0.0131, 0.0022, -0.0177, 0.0147, 0.0011, -0.0154, -0.0261, + 0.0096, -0.0067], device='cuda:0'), grad: tensor([ 0.0337, -0.0019, 0.0399, 0.0207, -0.0011, -0.0388, -0.0466, -0.0144, + 0.0363, -0.0277], device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 217.23, cls_loss 0.6149 cls_loss_mapping 0.0190 cls_loss_causal 0.5828 re_mapping 0.0112 re_causal 0.0296 /// teacc 98.66 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 0.0066, 0.0597, -0.1032, ..., -0.0577, 0.0437, 0.0077], + [-0.0510, -0.0804, 0.0537, ..., -0.0294, -0.0811, 0.0466], + [-0.0464, -0.0261, -0.0645, ..., -0.0878, -0.0400, 0.0303], + ..., + [-0.0167, -0.0412, -0.0650, ..., -0.0851, -0.0023, 0.0438], + [-0.0760, -0.0664, -0.0622, ..., -0.0405, -0.0367, -0.0642], + [ 0.0073, -0.0023, -0.0506, ..., -0.0412, 0.0304, -0.0118]], + device='cuda:0'), grad: tensor([[-4.1342e-04, -1.7118e-04, 1.9574e-04, ..., 1.6880e-04, + 8.8406e-04, -1.1377e-03], + [ 1.6670e-03, 4.2510e-04, 6.4354e-07, ..., 2.5902e-03, + 1.1559e-03, 6.8474e-04], + [-2.7294e-03, 1.6870e-03, 1.5173e-03, ..., -3.0994e-05, + -2.1744e-03, -3.5229e-03], + ..., + [ 1.8311e-03, 4.4584e-04, 2.2590e-04, ..., 9.5034e-04, + 1.2236e-03, 2.3499e-03], + [ 7.1573e-04, -2.9106e-03, -3.7384e-03, ..., -8.8120e-03, + -2.7442e-04, -2.1040e-04], + [ 3.5114e-03, 8.5497e-04, 4.7421e-04, ..., 1.0653e-03, + 2.1973e-03, 1.8311e-03]], device='cuda:0') +Epoch 97, bias, value: tensor([ 0.0151, -0.0134, 0.0017, -0.0172, 0.0145, 0.0013, -0.0157, -0.0260, + 0.0094, -0.0059], device='cuda:0'), grad: tensor([-0.0082, 0.0346, -0.0355, 0.0502, 0.0103, -0.0348, 0.0069, -0.0029, + -0.0591, 0.0384], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 96---------------------------------------------------- +epoch 96, time 217.57, cls_loss 0.5738 cls_loss_mapping 0.0179 cls_loss_causal 0.5351 re_mapping 0.0121 re_causal 0.0302 /// teacc 98.73 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0064, 0.0604, -0.1040, ..., -0.0583, 0.0437, 0.0079], + [-0.0514, -0.0785, 0.0560, ..., -0.0290, -0.0810, 0.0464], + [-0.0454, -0.0255, -0.0661, ..., -0.0887, -0.0381, 0.0298], + ..., + [-0.0173, -0.0423, -0.0648, ..., -0.0841, -0.0034, 0.0445], + [-0.0764, -0.0672, -0.0624, ..., -0.0407, -0.0371, -0.0641], + [ 0.0075, -0.0017, -0.0501, ..., -0.0412, 0.0311, -0.0117]], + device='cuda:0'), grad: tensor([[ 1.3828e-03, 1.1134e-04, 6.9976e-05, ..., 2.5105e-04, + 2.3060e-03, 3.3112e-03], + [-4.4289e-03, 3.5286e-05, -1.4687e-04, ..., 1.1724e-04, + -6.7863e-03, 3.9649e-04], + [ 1.4000e-03, -1.3018e-03, 8.1837e-05, ..., 3.5238e-04, + 2.3556e-03, -1.3580e-03], + ..., + [-2.0370e-03, 6.5231e-04, 1.2040e-04, ..., -9.7418e-04, + -2.9736e-03, -2.2034e-02], + [-7.1220e-03, -1.3053e-04, -4.5319e-03, ..., -5.8899e-03, + -8.3733e-04, 2.5063e-03], + [-1.5430e-03, 1.0505e-05, 8.9264e-04, ..., 1.1148e-03, + -2.5940e-03, 5.5771e-03]], device='cuda:0') +Epoch 98, bias, value: tensor([ 0.0146, -0.0139, 0.0024, -0.0182, 0.0153, 0.0012, -0.0159, -0.0256, + 0.0093, -0.0054], device='cuda:0'), grad: tensor([ 0.0234, -0.0415, -0.0286, 0.0276, 0.0235, 0.0351, 0.0209, -0.0633, + -0.0190, 0.0218], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 97---------------------------------------------------- +epoch 97, time 217.87, cls_loss 0.5863 cls_loss_mapping 0.0192 cls_loss_causal 0.5576 re_mapping 0.0114 re_causal 0.0303 /// teacc 98.77 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0063, 0.0600, -0.1042, ..., -0.0590, 0.0444, 0.0081], + [-0.0511, -0.0784, 0.0573, ..., -0.0271, -0.0809, 0.0471], + [-0.0452, -0.0251, -0.0666, ..., -0.0897, -0.0384, 0.0300], + ..., + [-0.0175, -0.0437, -0.0660, ..., -0.0853, -0.0038, 0.0441], + [-0.0766, -0.0674, -0.0636, ..., -0.0405, -0.0377, -0.0631], + [ 0.0066, -0.0028, -0.0506, ..., -0.0419, 0.0304, -0.0124]], + device='cuda:0'), grad: tensor([[ 8.6308e-04, 5.9128e-05, 4.4179e-04, ..., 8.1396e-04, + 3.9935e-04, 1.0157e-03], + [ 1.8311e-03, 1.9979e-04, 1.0300e-03, ..., 9.3794e-04, + 9.1553e-04, 5.7678e-03], + [ 8.4543e-04, 4.5109e-04, 2.1017e-04, ..., 3.2687e-04, + 4.1962e-04, 2.4300e-03], + ..., + [ 1.4706e-03, 1.1486e-04, 6.3467e-04, ..., 5.3930e-04, + 7.2145e-04, -5.6267e-03], + [-1.5554e-03, 4.5323e-04, 6.8426e-04, ..., 7.6056e-04, + -2.6679e-04, 1.6651e-03], + [ 2.6882e-05, 1.4015e-05, 3.1543e-04, ..., 5.7125e-04, + -7.8249e-04, -2.2755e-03]], device='cuda:0') +Epoch 99, bias, value: tensor([ 0.0145, -0.0126, 0.0022, -0.0188, 0.0151, 0.0018, -0.0162, -0.0264, + 0.0099, -0.0057], device='cuda:0'), grad: tensor([ 0.0110, 0.0268, 0.0148, -0.0514, -0.0119, 0.0144, 0.0152, 0.0038, + -0.0063, -0.0165], device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 216.89, cls_loss 0.5778 cls_loss_mapping 0.0214 cls_loss_causal 0.5445 re_mapping 0.0123 re_causal 0.0314 /// teacc 98.71 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0059, 0.0598, -0.1048, ..., -0.0591, 0.0452, 0.0073], + [-0.0510, -0.0812, 0.0573, ..., -0.0281, -0.0820, 0.0459], + [-0.0460, -0.0264, -0.0677, ..., -0.0911, -0.0390, 0.0311], + ..., + [-0.0170, -0.0437, -0.0669, ..., -0.0848, -0.0040, 0.0446], + [-0.0769, -0.0680, -0.0651, ..., -0.0413, -0.0378, -0.0635], + [ 0.0071, -0.0029, -0.0512, ..., -0.0419, 0.0302, -0.0120]], + device='cuda:0'), grad: tensor([[ 0.0046, 0.0022, 0.0006, ..., 0.0003, 0.0024, 0.0024], + [-0.0018, -0.0002, -0.0031, ..., -0.0005, -0.0028, -0.0068], + [ 0.0019, 0.0007, 0.0008, ..., 0.0003, 0.0013, 0.0033], + ..., + [ 0.0032, 0.0005, 0.0009, ..., 0.0002, 0.0020, 0.0023], + [-0.0042, 0.0084, 0.0104, ..., 0.0064, -0.0016, -0.0035], + [-0.0032, -0.0005, -0.0003, ..., -0.0006, -0.0011, -0.0033]], + device='cuda:0') +Epoch 100, bias, value: tensor([ 0.0142, -0.0129, 0.0014, -0.0182, 0.0152, 0.0017, -0.0167, -0.0261, + 0.0100, -0.0050], device='cuda:0'), grad: tensor([ 0.0295, -0.0320, 0.0230, 0.0118, -0.0180, -0.0275, 0.0071, 0.0231, + 0.0072, -0.0243], device='cuda:0') +100 +0.0001 +changing lr +epoch 99, time 217.72, cls_loss 0.6186 cls_loss_mapping 0.0204 cls_loss_causal 0.5807 re_mapping 0.0116 re_causal 0.0304 /// teacc 98.65 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0060, 0.0601, -0.1045, ..., -0.0593, 0.0457, 0.0064], + [-0.0509, -0.0824, 0.0569, ..., -0.0280, -0.0820, 0.0456], + [-0.0467, -0.0276, -0.0680, ..., -0.0917, -0.0386, 0.0315], + ..., + [-0.0168, -0.0447, -0.0671, ..., -0.0838, -0.0047, 0.0449], + [-0.0766, -0.0677, -0.0649, ..., -0.0416, -0.0373, -0.0624], + [ 0.0064, -0.0035, -0.0525, ..., -0.0437, 0.0293, -0.0128]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0010, 0.0000, ..., -0.0002, 0.0029, 0.0012], + [-0.0024, 0.0004, 0.0000, ..., -0.0010, -0.0025, -0.0068], + [ 0.0025, 0.0019, 0.0000, ..., 0.0019, 0.0043, 0.0063], + ..., + [ 0.0036, 0.0004, 0.0000, ..., 0.0009, 0.0031, 0.0069], + [-0.0008, -0.0040, 0.0000, ..., 0.0006, -0.0029, -0.0024], + [ 0.0011, 0.0011, 0.0000, ..., 0.0008, 0.0028, -0.0031]], + device='cuda:0') +Epoch 101, bias, value: tensor([ 0.0145, -0.0129, 0.0015, -0.0188, 0.0146, 0.0011, -0.0155, -0.0259, + 0.0110, -0.0059], device='cuda:0'), grad: tensor([ 0.0017, -0.0244, 0.0418, 0.0417, -0.0473, -0.0511, 0.0289, 0.0329, + -0.0329, 0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 217.85, cls_loss 0.5789 cls_loss_mapping 0.0155 cls_loss_causal 0.5488 re_mapping 0.0118 re_causal 0.0306 /// teacc 98.67 lr 0.00010000 +Epoch 102, weight, value: tensor([[ 0.0057, 0.0610, -0.1047, ..., -0.0603, 0.0461, 0.0071], + [-0.0513, -0.0828, 0.0571, ..., -0.0268, -0.0828, 0.0466], + [-0.0469, -0.0271, -0.0690, ..., -0.0924, -0.0380, 0.0321], + ..., + [-0.0173, -0.0457, -0.0671, ..., -0.0841, -0.0047, 0.0444], + [-0.0771, -0.0680, -0.0649, ..., -0.0415, -0.0377, -0.0619], + [ 0.0072, -0.0045, -0.0530, ..., -0.0447, 0.0282, -0.0122]], + device='cuda:0'), grad: tensor([[-7.0915e-03, -1.8616e-03, -1.6534e-04, ..., -2.6584e-05, + -1.0195e-03, -8.6823e-03], + [-8.0967e-04, 1.7858e-04, 7.7667e-03, ..., -5.4283e-03, + -4.0817e-04, -2.5082e-03], + [ 1.1024e-03, 2.1248e-03, 6.7377e-04, ..., 4.9734e-04, + 1.7567e-03, 1.6060e-03], + ..., + [ 1.0796e-03, 1.0008e-04, 1.4710e-04, ..., 2.3770e-04, + -5.2243e-05, 4.4036e-04], + [ 1.5068e-03, 2.5558e-03, 6.2447e-03, ..., 7.1259e-03, + 1.2245e-03, 3.0289e-03], + [ 1.7443e-03, 4.8208e-04, 4.7994e-04, ..., 6.0272e-04, + 4.8161e-04, 2.4796e-03]], device='cuda:0') +Epoch 102, bias, value: tensor([ 0.0149, -0.0126, 0.0015, -0.0180, 0.0147, -0.0001, -0.0159, -0.0267, + 0.0105, -0.0047], device='cuda:0'), grad: tensor([-0.0527, -0.0396, 0.0244, 0.0086, -0.0039, -0.0209, -0.0056, 0.0127, + 0.0564, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 218.19, cls_loss 0.6125 cls_loss_mapping 0.0180 cls_loss_causal 0.5786 re_mapping 0.0118 re_causal 0.0311 /// teacc 98.50 lr 0.00010000 +Epoch 103, weight, value: tensor([[ 0.0063, 0.0613, -0.1051, ..., -0.0604, 0.0461, 0.0075], + [-0.0521, -0.0838, 0.0572, ..., -0.0271, -0.0842, 0.0463], + [-0.0483, -0.0270, -0.0694, ..., -0.0924, -0.0391, 0.0319], + ..., + [-0.0159, -0.0456, -0.0669, ..., -0.0850, -0.0026, 0.0449], + [-0.0771, -0.0685, -0.0655, ..., -0.0417, -0.0381, -0.0628], + [ 0.0049, -0.0053, -0.0543, ..., -0.0448, 0.0272, -0.0130]], + device='cuda:0'), grad: tensor([[ 2.5120e-03, 2.2202e-03, 2.3878e-04, ..., 1.4019e-04, + 2.3613e-03, 3.0689e-03], + [ 1.3618e-03, 3.1734e-04, 7.7963e-04, ..., 3.2783e-05, + 3.7336e-04, 2.9068e-03], + [ 7.3576e-04, 4.5991e-04, -1.3075e-03, ..., -1.7405e-03, + 5.6601e-04, -3.2496e-04], + ..., + [ 1.5507e-03, 7.2145e-04, 2.7657e-04, ..., 1.2302e-04, + 6.4182e-04, 6.2103e-03], + [-2.3727e-03, -2.4376e-03, -2.5940e-04, ..., 4.2486e-04, + 1.0262e-03, -5.9471e-03], + [-2.3823e-03, 2.1286e-03, -4.1032e-04, ..., 1.4150e-04, + -8.5163e-04, -1.3054e-02]], device='cuda:0') +Epoch 103, bias, value: tensor([ 0.0157, -0.0125, 0.0010, -0.0186, 0.0152, 0.0007, -0.0159, -0.0262, + 0.0097, -0.0053], device='cuda:0'), grad: tensor([ 0.0252, 0.0162, 0.0043, -0.0151, -0.0071, 0.0218, 0.0129, 0.0282, + -0.0285, -0.0579], device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 217.01, cls_loss 0.6416 cls_loss_mapping 0.0203 cls_loss_causal 0.6140 re_mapping 0.0110 re_causal 0.0299 /// teacc 98.68 lr 0.00010000 +Epoch 104, weight, value: tensor([[ 0.0063, 0.0614, -0.1055, ..., -0.0607, 0.0460, 0.0076], + [-0.0519, -0.0832, 0.0572, ..., -0.0275, -0.0838, 0.0475], + [-0.0490, -0.0269, -0.0686, ..., -0.0908, -0.0374, 0.0327], + ..., + [-0.0157, -0.0456, -0.0670, ..., -0.0852, -0.0031, 0.0438], + [-0.0773, -0.0693, -0.0654, ..., -0.0420, -0.0379, -0.0637], + [ 0.0048, -0.0052, -0.0544, ..., -0.0441, 0.0271, -0.0115]], + device='cuda:0'), grad: tensor([[ 3.7518e-03, 2.2411e-03, 1.1950e-03, ..., 3.3932e-03, + -5.4779e-03, -2.3975e-03], + [ 1.0216e-04, 5.9271e-04, -3.5896e-03, ..., 3.1900e-04, + 3.6597e-04, -6.7949e-04], + [-2.5730e-03, 7.6485e-04, -1.2569e-03, ..., 2.9063e-04, + -2.1577e-05, -1.4582e-03], + ..., + [-1.9503e-04, -2.9182e-03, 9.6369e-04, ..., 2.6298e-04, + 1.4317e-04, -3.1109e-03], + [ 5.4512e-03, 4.2877e-03, 2.4719e-03, ..., 2.1858e-03, + 4.1389e-03, 4.1656e-03], + [ 3.3112e-03, 1.4524e-03, 1.6117e-03, ..., 7.9012e-04, + 2.3670e-03, 6.2256e-03]], device='cuda:0') +Epoch 104, bias, value: tensor([ 1.4758e-02, -1.1438e-02, 2.6310e-03, -1.8840e-02, 1.4827e-02, + -1.1703e-05, -1.6996e-02, -2.6508e-02, 9.1363e-03, -4.0164e-03], + device='cuda:0'), grad: tensor([ 0.0020, -0.0112, -0.0049, 0.0086, 0.0215, -0.0601, -0.0250, 0.0117, + 0.0235, 0.0340], device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 217.97, cls_loss 0.5616 cls_loss_mapping 0.0162 cls_loss_causal 0.5276 re_mapping 0.0119 re_causal 0.0308 /// teacc 98.68 lr 0.00010000 +Epoch 105, weight, value: tensor([[ 0.0065, 0.0614, -0.1056, ..., -0.0616, 0.0461, 0.0075], + [-0.0519, -0.0839, 0.0564, ..., -0.0274, -0.0833, 0.0471], + [-0.0502, -0.0270, -0.0689, ..., -0.0920, -0.0383, 0.0319], + ..., + [-0.0170, -0.0460, -0.0663, ..., -0.0858, -0.0043, 0.0449], + [-0.0785, -0.0694, -0.0659, ..., -0.0413, -0.0385, -0.0642], + [ 0.0052, -0.0057, -0.0538, ..., -0.0436, 0.0263, -0.0126]], + device='cuda:0'), grad: tensor([[-7.1030e-03, -5.7459e-04, 1.9252e-05, ..., -2.7237e-03, + -5.1880e-03, -1.8787e-03], + [ 3.0689e-03, 8.5592e-04, 4.5747e-05, ..., 6.6996e-04, + 2.2373e-03, 2.5139e-03], + [-1.4715e-03, -2.1648e-03, 6.8426e-05, ..., -3.1203e-05, + -1.1005e-03, 2.0771e-03], + ..., + [-7.0457e-03, 1.4818e-04, 7.3254e-05, ..., 4.1437e-04, + -6.3438e-03, -7.7286e-03], + [ 2.4700e-03, 1.1511e-03, 1.5998e-04, ..., 6.0654e-04, + 1.8492e-03, 2.2793e-03], + [ 3.1185e-03, -2.5673e-03, 5.9873e-05, ..., 7.9250e-04, + 2.2717e-03, 2.4300e-03]], device='cuda:0') +Epoch 105, bias, value: tensor([ 1.4335e-02, -1.1521e-02, 1.8202e-03, -1.8584e-02, 1.4787e-02, + -5.6482e-06, -1.6407e-02, -2.6200e-02, 9.9489e-03, -4.6352e-03], + device='cuda:0'), grad: tensor([-0.0309, 0.0269, 0.0047, -0.0002, 0.0309, 0.0444, -0.0334, -0.0626, + 0.0255, -0.0053], device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 219.14, cls_loss 0.5946 cls_loss_mapping 0.0221 cls_loss_causal 0.5585 re_mapping 0.0115 re_causal 0.0294 /// teacc 98.50 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 0.0074, 0.0614, -0.1058, ..., -0.0621, 0.0467, 0.0084], + [-0.0529, -0.0839, 0.0575, ..., -0.0277, -0.0837, 0.0455], + [-0.0509, -0.0287, -0.0696, ..., -0.0925, -0.0388, 0.0317], + ..., + [-0.0161, -0.0470, -0.0663, ..., -0.0866, -0.0032, 0.0454], + [-0.0794, -0.0694, -0.0668, ..., -0.0421, -0.0387, -0.0646], + [ 0.0046, -0.0057, -0.0536, ..., -0.0442, 0.0262, -0.0134]], + device='cuda:0'), grad: tensor([[ 2.7752e-04, 9.1791e-04, 2.0161e-05, ..., 2.7180e-04, + 2.1072e-02, -7.5798e-03], + [-4.3221e-03, 1.2684e-04, 2.9832e-05, ..., 9.0003e-05, + 7.5293e-04, 1.9741e-03], + [ 7.0906e-04, 3.9792e-04, 7.2300e-05, ..., -2.9831e-03, + -5.7554e-04, -5.8823e-03], + ..., + [ 2.6855e-03, 1.4782e-04, 5.4836e-06, ..., 7.6437e-04, + 3.8795e-03, 8.4000e-03], + [ 2.3575e-03, -6.8843e-05, 1.0395e-03, ..., 1.6356e-03, + 1.0710e-03, 3.2654e-03], + [-4.4327e-03, -2.0065e-03, 4.8846e-05, ..., 5.6803e-05, + -6.0577e-03, -8.0490e-03]], device='cuda:0') +Epoch 106, bias, value: tensor([ 0.0147, -0.0124, 0.0016, -0.0186, 0.0158, 0.0002, -0.0165, -0.0258, + 0.0094, -0.0050], device='cuda:0'), grad: tensor([-0.0153, -0.0085, -0.0116, -0.0137, 0.0222, 0.0199, 0.0004, 0.0347, + 0.0205, -0.0487], device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 218.39, cls_loss 0.6031 cls_loss_mapping 0.0216 cls_loss_causal 0.5611 re_mapping 0.0123 re_causal 0.0306 /// teacc 98.51 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0085, 0.0629, -0.1047, ..., -0.0625, 0.0469, 0.0077], + [-0.0521, -0.0831, 0.0565, ..., -0.0280, -0.0841, 0.0456], + [-0.0521, -0.0282, -0.0696, ..., -0.0909, -0.0390, 0.0318], + ..., + [-0.0158, -0.0481, -0.0661, ..., -0.0864, -0.0035, 0.0462], + [-0.0793, -0.0704, -0.0681, ..., -0.0432, -0.0390, -0.0649], + [ 0.0037, -0.0053, -0.0533, ..., -0.0439, 0.0257, -0.0144]], + device='cuda:0'), grad: tensor([[ 1.7967e-03, 2.7323e-04, 3.5596e-04, ..., 4.9877e-04, + 1.8063e-03, 1.7395e-03], + [ 8.8739e-04, 2.7370e-04, 6.3248e-03, ..., 4.5466e-04, + 6.3181e-04, 1.8930e-03], + [-3.1052e-03, 3.5429e-04, 4.0054e-04, ..., 6.2418e-04, + -3.8948e-03, 4.5466e-04], + ..., + [-1.5989e-05, 1.4699e-04, 2.8586e-04, ..., 1.8919e-04, + 5.1975e-04, -2.4104e-04], + [ 1.3571e-03, 8.1444e-04, -7.3891e-03, ..., 1.6232e-03, + 1.3552e-03, 1.8387e-03], + [ 7.9823e-04, 1.5259e-04, 4.1962e-04, ..., 1.7273e-04, + 3.6478e-04, 9.6369e-04]], device='cuda:0') +Epoch 107, bias, value: tensor([ 0.0153, -0.0121, 0.0011, -0.0187, 0.0152, 0.0010, -0.0159, -0.0255, + 0.0093, -0.0060], device='cuda:0'), grad: tensor([ 0.0147, 0.0271, -0.0169, 0.0125, 0.0120, -0.0275, -0.0190, 0.0030, + -0.0117, 0.0059], device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 218.77, cls_loss 0.5630 cls_loss_mapping 0.0262 cls_loss_causal 0.5401 re_mapping 0.0110 re_causal 0.0296 /// teacc 98.35 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0084, 0.0629, -0.1051, ..., -0.0626, 0.0476, 0.0061], + [-0.0510, -0.0847, 0.0553, ..., -0.0275, -0.0836, 0.0464], + [-0.0531, -0.0279, -0.0703, ..., -0.0919, -0.0386, 0.0315], + ..., + [-0.0158, -0.0482, -0.0659, ..., -0.0862, -0.0039, 0.0464], + [-0.0801, -0.0716, -0.0689, ..., -0.0437, -0.0387, -0.0653], + [ 0.0056, -0.0049, -0.0542, ..., -0.0434, 0.0269, -0.0141]], + device='cuda:0'), grad: tensor([[ 9.3365e-04, 1.7083e-04, 1.9968e-05, ..., 3.8743e-05, + 3.1805e-04, 1.7242e-03], + [-1.1673e-03, 2.0676e-03, 5.3177e-03, ..., 5.7727e-05, + -8.9693e-04, 5.2643e-03], + [ 1.5450e-04, -1.1683e-03, -4.9820e-03, ..., -3.4571e-04, + 1.4186e-04, -5.6114e-03], + ..., + [ 2.9678e-03, 1.6356e-03, 2.2335e-03, ..., 3.0088e-04, + 5.6934e-04, 3.0518e-03], + [ 6.2485e-03, 3.9902e-03, 9.6178e-04, ..., 5.4300e-05, + 1.8225e-03, 9.3842e-03], + [ 2.8343e-03, -2.0504e-03, -1.8511e-03, ..., 4.5586e-04, + 3.2883e-03, -8.7967e-03]], device='cuda:0') +Epoch 108, bias, value: tensor([ 0.0145, -0.0117, 0.0002, -0.0183, 0.0153, 0.0013, -0.0165, -0.0253, + 0.0094, -0.0053], device='cuda:0'), grad: tensor([ 0.0084, 0.0134, -0.0314, 0.0058, -0.0012, -0.0474, 0.0096, 0.0111, + 0.0323, -0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 218.69, cls_loss 0.6219 cls_loss_mapping 0.0270 cls_loss_causal 0.5874 re_mapping 0.0111 re_causal 0.0288 /// teacc 98.61 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0085, 0.0631, -0.1063, ..., -0.0616, 0.0474, 0.0064], + [-0.0516, -0.0844, 0.0547, ..., -0.0264, -0.0837, 0.0459], + [-0.0515, -0.0270, -0.0703, ..., -0.0914, -0.0374, 0.0315], + ..., + [-0.0168, -0.0491, -0.0666, ..., -0.0888, -0.0042, 0.0476], + [-0.0804, -0.0718, -0.0686, ..., -0.0438, -0.0377, -0.0660], + [ 0.0057, -0.0060, -0.0542, ..., -0.0428, 0.0264, -0.0142]], + device='cuda:0'), grad: tensor([[-0.0099, -0.0040, -0.0005, ..., -0.0086, -0.0037, -0.0099], + [-0.0008, 0.0028, 0.0019, ..., -0.0004, 0.0014, -0.0007], + [-0.0017, 0.0028, 0.0034, ..., 0.0029, 0.0009, -0.0036], + ..., + [ 0.0025, 0.0013, 0.0004, ..., 0.0023, 0.0016, 0.0019], + [ 0.0008, -0.0049, -0.0016, ..., -0.0014, -0.0009, 0.0037], + [ 0.0026, 0.0010, 0.0002, ..., 0.0016, 0.0018, 0.0027]], + device='cuda:0') +Epoch 109, bias, value: tensor([ 0.0147, -0.0119, 0.0006, -0.0181, 0.0152, 0.0004, -0.0162, -0.0253, + 0.0096, -0.0053], device='cuda:0'), grad: tensor([-0.0721, -0.0169, -0.0064, 0.0318, 0.0184, 0.0174, 0.0167, 0.0254, + -0.0350, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 218.73, cls_loss 0.5932 cls_loss_mapping 0.0176 cls_loss_causal 0.5614 re_mapping 0.0113 re_causal 0.0294 /// teacc 98.47 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 0.0097, 0.0635, -0.1078, ..., -0.0614, 0.0480, 0.0072], + [-0.0516, -0.0859, 0.0545, ..., -0.0273, -0.0830, 0.0464], + [-0.0511, -0.0263, -0.0706, ..., -0.0912, -0.0360, 0.0314], + ..., + [-0.0166, -0.0496, -0.0682, ..., -0.0894, -0.0049, 0.0473], + [-0.0818, -0.0718, -0.0683, ..., -0.0440, -0.0384, -0.0669], + [ 0.0063, -0.0059, -0.0534, ..., -0.0419, 0.0264, -0.0128]], + device='cuda:0'), grad: tensor([[-7.8201e-03, 2.6875e-03, -5.0306e-04, ..., -3.7651e-03, + -1.0214e-03, -6.6338e-03], + [-5.0468e-03, -4.4882e-05, 7.2420e-05, ..., -5.2185e-03, + -1.4553e-03, -4.9896e-03], + [-5.7640e-03, -7.7248e-05, 4.9561e-05, ..., 1.2007e-03, + -4.5815e-03, -8.8959e-03], + ..., + [-3.8395e-03, 3.9244e-04, 9.2089e-05, ..., -6.9737e-05, + 4.1890e-04, -5.4054e-03], + [ 2.9259e-03, 4.4847e-04, -4.3899e-05, ..., 1.1320e-03, + 1.3065e-03, 2.5692e-03], + [ 1.1986e-02, 4.5433e-03, 1.1241e-04, ..., 1.3342e-03, + 7.3471e-03, 8.7509e-03]], device='cuda:0') +Epoch 110, bias, value: tensor([ 0.0151, -0.0119, 0.0001, -0.0184, 0.0149, 0.0006, -0.0161, -0.0253, + 0.0085, -0.0039], device='cuda:0'), grad: tensor([-0.0526, -0.0327, -0.0216, 0.0053, 0.0544, 0.0072, -0.0111, -0.0177, + -0.0031, 0.0720], device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 218.26, cls_loss 0.5856 cls_loss_mapping 0.0186 cls_loss_causal 0.5527 re_mapping 0.0117 re_causal 0.0296 /// teacc 98.55 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0096, 0.0644, -0.1080, ..., -0.0620, 0.0485, 0.0068], + [-0.0530, -0.0861, 0.0550, ..., -0.0272, -0.0840, 0.0470], + [-0.0513, -0.0267, -0.0709, ..., -0.0914, -0.0353, 0.0304], + ..., + [-0.0161, -0.0487, -0.0682, ..., -0.0900, -0.0046, 0.0482], + [-0.0823, -0.0715, -0.0682, ..., -0.0429, -0.0391, -0.0674], + [ 0.0054, -0.0077, -0.0544, ..., -0.0429, 0.0257, -0.0128]], + device='cuda:0'), grad: tensor([[ 1.0881e-03, 3.8087e-05, 1.5929e-05, ..., 1.4150e-04, + 5.7173e-04, 8.5592e-04], + [ 7.7152e-04, 1.6093e-05, 1.3649e-05, ..., 8.9526e-05, + 5.3358e-04, 9.8419e-04], + [ 2.0294e-03, 1.0926e-04, 8.4460e-05, ..., 1.1271e-04, + 4.9067e-04, -5.8651e-04], + ..., + [-2.2526e-03, 2.2382e-05, 1.9044e-05, ..., -1.2350e-03, + -2.3193e-03, -3.0861e-03], + [ 1.1892e-03, 1.9801e-04, 8.1778e-05, ..., 9.1374e-05, + 5.3310e-04, 8.8406e-04], + [ 2.9068e-03, 7.3433e-05, 5.5492e-05, ..., 4.0102e-04, + 1.2941e-03, 2.3308e-03]], device='cuda:0') +Epoch 111, bias, value: tensor([ 1.5071e-02, -1.1897e-02, 7.1030e-05, -1.7955e-02, 1.4337e-02, + 1.3873e-03, -1.6225e-02, -2.5024e-02, 8.4318e-03, -4.5945e-03], + device='cuda:0'), grad: tensor([ 0.0058, 0.0058, 0.0025, -0.0033, 0.0055, 0.0051, -0.0234, -0.0139, + 0.0103, 0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 218.99, cls_loss 0.5853 cls_loss_mapping 0.0201 cls_loss_causal 0.5541 re_mapping 0.0109 re_causal 0.0277 /// teacc 98.54 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0094, 0.0644, -0.1080, ..., -0.0636, 0.0480, 0.0075], + [-0.0551, -0.0869, 0.0556, ..., -0.0285, -0.0863, 0.0477], + [-0.0518, -0.0266, -0.0695, ..., -0.0903, -0.0349, 0.0317], + ..., + [-0.0161, -0.0501, -0.0701, ..., -0.0887, -0.0038, 0.0472], + [-0.0813, -0.0724, -0.0686, ..., -0.0431, -0.0390, -0.0672], + [ 0.0051, -0.0075, -0.0545, ..., -0.0430, 0.0248, -0.0136]], + device='cuda:0'), grad: tensor([[-5.9700e-04, -5.1231e-03, -4.9591e-04, ..., -1.8728e-04, + -4.9553e-03, 9.2411e-04], + [ 1.1005e-03, 2.4121e-06, -1.2660e-04, ..., -5.9396e-05, + 4.9448e-04, 1.6344e-04], + [ 1.0567e-03, 3.0905e-05, 2.3976e-05, ..., 5.8375e-06, + 5.0306e-04, 2.5177e-04], + ..., + [-5.3749e-03, 4.7833e-06, 2.2948e-06, ..., 3.0026e-06, + -2.7580e-03, -2.1687e-03], + [ 1.4496e-03, -2.4152e-04, 7.1228e-06, ..., 3.7938e-05, + 6.7949e-04, 1.5755e-03], + [ 3.1910e-03, 2.8572e-03, 2.8133e-04, ..., 3.7607e-06, + 4.0207e-03, 1.8740e-03]], device='cuda:0') +Epoch 112, bias, value: tensor([ 0.0143, -0.0127, 0.0003, -0.0176, 0.0138, 0.0006, -0.0154, -0.0245, + 0.0091, -0.0044], device='cuda:0'), grad: tensor([ 0.0061, 0.0068, 0.0087, 0.0222, -0.0189, -0.0505, 0.0140, -0.0196, + 0.0088, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 218.29, cls_loss 0.5954 cls_loss_mapping 0.0209 cls_loss_causal 0.5633 re_mapping 0.0111 re_causal 0.0277 /// teacc 98.38 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 0.0092, 0.0651, -0.1086, ..., -0.0640, 0.0491, 0.0079], + [-0.0553, -0.0872, 0.0563, ..., -0.0280, -0.0877, 0.0483], + [-0.0516, -0.0263, -0.0701, ..., -0.0910, -0.0343, 0.0316], + ..., + [-0.0165, -0.0500, -0.0699, ..., -0.0885, -0.0038, 0.0475], + [-0.0810, -0.0718, -0.0688, ..., -0.0438, -0.0385, -0.0672], + [ 0.0054, -0.0081, -0.0549, ..., -0.0429, 0.0242, -0.0140]], + device='cuda:0'), grad: tensor([[ 3.6240e-03, -9.2983e-04, 6.5231e-04, ..., -1.5793e-03, + 1.9779e-03, 4.0855e-03], + [ 3.5191e-03, -2.8629e-03, -1.8034e-03, ..., 5.4073e-04, + 7.0572e-04, 3.0823e-03], + [ 3.2349e-03, 1.4997e-04, 9.5010e-05, ..., -7.5006e-04, + 2.3499e-03, 5.9319e-04], + ..., + [-7.2060e-03, 1.7624e-03, 2.2125e-03, ..., -1.8282e-03, + -6.4926e-03, 1.1215e-03], + [ 2.2850e-03, 1.4925e-03, 9.5892e-04, ..., 9.0599e-04, + 2.0027e-03, 3.0556e-03], + [-1.5251e-02, -2.1839e-03, -2.9125e-03, ..., -1.5888e-03, + -8.4534e-03, -1.9470e-02]], device='cuda:0') +Epoch 113, bias, value: tensor([ 0.0141, -0.0123, 0.0003, -0.0175, 0.0142, 0.0006, -0.0159, -0.0244, + 0.0093, -0.0049], device='cuda:0'), grad: tensor([ 0.0284, 0.0196, -0.0068, 0.0263, 0.0356, 0.0131, -0.0089, -0.0411, + 0.0277, -0.0938], device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 217.54, cls_loss 0.5725 cls_loss_mapping 0.0136 cls_loss_causal 0.5377 re_mapping 0.0115 re_causal 0.0289 /// teacc 98.54 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0091, 0.0661, -0.1092, ..., -0.0642, 0.0485, 0.0084], + [-0.0549, -0.0881, 0.0572, ..., -0.0280, -0.0881, 0.0479], + [-0.0520, -0.0264, -0.0705, ..., -0.0914, -0.0350, 0.0320], + ..., + [-0.0156, -0.0517, -0.0710, ..., -0.0888, -0.0040, 0.0480], + [-0.0820, -0.0725, -0.0683, ..., -0.0417, -0.0373, -0.0684], + [ 0.0046, -0.0078, -0.0550, ..., -0.0433, 0.0249, -0.0142]], + device='cuda:0'), grad: tensor([[ 1.7338e-03, 6.4707e-04, 7.9691e-05, ..., 6.4278e-04, + 1.0624e-03, 1.9226e-03], + [-3.3665e-03, 4.5133e-04, 1.8148e-03, ..., 2.1458e-03, + -1.7643e-03, -5.7793e-03], + [ 2.0180e-03, 5.9891e-04, 3.9005e-04, ..., 9.1791e-04, + 1.0843e-03, 5.5656e-03], + ..., + [-4.5052e-03, -1.0271e-03, -3.9637e-05, ..., -2.1687e-03, + -2.2829e-04, -6.2218e-03], + [ 1.6890e-03, 1.9526e-04, 1.2245e-03, ..., 1.6174e-03, + 7.6151e-04, 1.0452e-03], + [ 1.2369e-03, -1.1575e-04, 1.7059e-04, ..., 9.2411e-04, + 1.8978e-04, 1.2503e-03]], device='cuda:0') +Epoch 114, bias, value: tensor([ 0.0142, -0.0118, 0.0007, -0.0182, 0.0140, 0.0006, -0.0156, -0.0243, + 0.0090, -0.0051], device='cuda:0'), grad: tensor([ 0.0199, -0.0362, 0.0277, -0.0091, -0.0042, -0.0046, 0.0271, -0.0474, + 0.0063, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 217.63, cls_loss 0.6105 cls_loss_mapping 0.0204 cls_loss_causal 0.5754 re_mapping 0.0113 re_causal 0.0287 /// teacc 98.53 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 0.0091, 0.0663, -0.1108, ..., -0.0658, 0.0483, 0.0089], + [-0.0563, -0.0891, 0.0583, ..., -0.0280, -0.0888, 0.0478], + [-0.0519, -0.0282, -0.0725, ..., -0.0925, -0.0355, 0.0326], + ..., + [-0.0150, -0.0529, -0.0706, ..., -0.0872, -0.0042, 0.0472], + [-0.0822, -0.0725, -0.0678, ..., -0.0425, -0.0371, -0.0688], + [ 0.0043, -0.0084, -0.0560, ..., -0.0435, 0.0238, -0.0127]], + device='cuda:0'), grad: tensor([[-2.4033e-04, -1.2236e-03, -1.2505e-04, ..., 1.9622e-04, + -2.5406e-03, 4.2610e-03], + [-1.0872e-03, 1.4520e-04, -5.1498e-03, ..., -4.8485e-03, + -3.0398e-04, -6.3934e-03], + [ 3.2997e-04, -3.4218e-03, -1.3380e-03, ..., 1.8597e-04, + 1.3244e-04, 2.1229e-03], + ..., + [ 3.2878e-04, 6.2466e-04, 1.2922e-03, ..., 8.7309e-04, + -9.8571e-06, 8.4114e-04], + [ 2.7866e-03, 2.3479e-03, -5.7554e-04, ..., 3.2730e-03, + -1.9932e-04, -2.1267e-03], + [ 2.6970e-03, 1.2846e-03, 3.6373e-03, ..., 1.1921e-03, + 2.1477e-03, 3.8681e-03]], device='cuda:0') +Epoch 115, bias, value: tensor([ 0.0141, -0.0120, 0.0009, -0.0179, 0.0134, 0.0004, -0.0163, -0.0239, + 0.0090, -0.0044], device='cuda:0'), grad: tensor([ 0.0192, -0.0325, -0.0081, 0.0255, -0.0049, -0.0201, -0.0254, 0.0107, + -0.0151, 0.0507], device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 217.45, cls_loss 0.5798 cls_loss_mapping 0.0195 cls_loss_causal 0.5506 re_mapping 0.0111 re_causal 0.0292 /// teacc 98.32 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0097, 0.0667, -0.1122, ..., -0.0663, 0.0492, 0.0081], + [-0.0569, -0.0893, 0.0588, ..., -0.0278, -0.0894, 0.0485], + [-0.0525, -0.0278, -0.0722, ..., -0.0918, -0.0341, 0.0324], + ..., + [-0.0153, -0.0518, -0.0700, ..., -0.0874, -0.0046, 0.0468], + [-0.0824, -0.0726, -0.0678, ..., -0.0433, -0.0365, -0.0686], + [ 0.0054, -0.0077, -0.0556, ..., -0.0428, 0.0242, -0.0124]], + device='cuda:0'), grad: tensor([[ 0.0022, 0.0051, 0.0032, ..., 0.0020, 0.0025, 0.0028], + [-0.0002, 0.0002, 0.0051, ..., 0.0087, 0.0005, -0.0043], + [ 0.0014, 0.0013, 0.0007, ..., 0.0005, 0.0006, 0.0025], + ..., + [ 0.0012, 0.0008, -0.0047, ..., -0.0085, 0.0006, 0.0024], + [ 0.0055, 0.0079, 0.0038, ..., 0.0010, 0.0009, 0.0018], + [-0.0062, -0.0069, -0.0030, ..., 0.0027, 0.0008, -0.0056]], + device='cuda:0') +Epoch 116, bias, value: tensor([ 0.0143, -0.0122, 0.0009, -0.0176, 0.0129, -0.0006, -0.0157, -0.0245, + 0.0096, -0.0040], device='cuda:0'), grad: tensor([ 0.0139, 0.0270, 0.0217, -0.0129, -0.0124, -0.0067, -0.0315, -0.0103, + 0.0348, -0.0235], device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 216.51, cls_loss 0.5750 cls_loss_mapping 0.0187 cls_loss_causal 0.5455 re_mapping 0.0108 re_causal 0.0282 /// teacc 98.54 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 0.0115, 0.0665, -0.1134, ..., -0.0660, 0.0504, 0.0075], + [-0.0576, -0.0891, 0.0598, ..., -0.0279, -0.0904, 0.0485], + [-0.0527, -0.0280, -0.0721, ..., -0.0929, -0.0348, 0.0322], + ..., + [-0.0165, -0.0524, -0.0708, ..., -0.0884, -0.0039, 0.0474], + [-0.0831, -0.0736, -0.0691, ..., -0.0445, -0.0371, -0.0689], + [ 0.0061, -0.0066, -0.0555, ..., -0.0428, 0.0242, -0.0129]], + device='cuda:0'), grad: tensor([[ 1.2445e-03, -5.6648e-04, 1.8716e-04, ..., 1.3995e-04, + 5.2738e-04, 1.5306e-03], + [ 2.4376e-03, -7.2908e-04, -3.1071e-03, ..., -1.7872e-03, + 9.0027e-04, 2.7828e-03], + [-3.7403e-03, -8.1348e-04, -1.8749e-03, ..., -2.8496e-03, + 6.2084e-04, -2.3785e-03], + ..., + [ 1.5020e-03, 3.1543e-04, 5.1832e-04, ..., 3.8743e-04, + -2.6226e-05, 1.1520e-03], + [ 1.3580e-03, -7.7057e-04, -3.1662e-03, ..., 1.1368e-03, + 6.0701e-04, -3.1734e-04], + [ 2.3384e-03, -2.6226e-04, 5.0735e-04, ..., 5.2452e-04, + 6.1321e-04, 3.4676e-03]], device='cuda:0') +Epoch 117, bias, value: tensor([ 0.0145, -0.0131, -0.0005, -0.0169, 0.0132, -0.0002, -0.0153, -0.0240, + 0.0097, -0.0043], device='cuda:0'), grad: tensor([-0.0134, 0.0447, -0.0143, 0.0358, -0.0352, -0.0154, -0.0123, 0.0100, + -0.0313, 0.0314], device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 216.33, cls_loss 0.6182 cls_loss_mapping 0.0167 cls_loss_causal 0.5858 re_mapping 0.0107 re_causal 0.0278 /// teacc 98.39 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 0.0119, 0.0666, -0.1134, ..., -0.0669, 0.0504, 0.0075], + [-0.0590, -0.0897, 0.0610, ..., -0.0285, -0.0895, 0.0486], + [-0.0518, -0.0284, -0.0736, ..., -0.0932, -0.0337, 0.0337], + ..., + [-0.0164, -0.0526, -0.0721, ..., -0.0865, -0.0036, 0.0478], + [-0.0826, -0.0740, -0.0698, ..., -0.0441, -0.0372, -0.0687], + [ 0.0065, -0.0054, -0.0550, ..., -0.0423, 0.0249, -0.0133]], + device='cuda:0'), grad: tensor([[ 2.8954e-03, 8.5688e-04, 2.6375e-05, ..., 2.3270e-04, + 1.6603e-03, 2.0008e-03], + [ 6.3801e-04, 1.9282e-05, 1.3091e-05, ..., 1.2696e-04, + 5.3835e-04, -7.0724e-03], + [ 2.8496e-03, 7.3969e-05, 2.0057e-05, ..., 1.2660e-04, + 7.2002e-04, 2.7466e-03], + ..., + [ 6.1150e-03, 8.8736e-06, 2.0206e-05, ..., 1.2648e-04, + 8.6784e-04, 1.4236e-02], + [-9.1934e-03, 5.8317e-04, 1.2234e-05, ..., -1.3781e-03, + 1.9860e-04, 6.6042e-04], + [-9.0408e-04, 2.2039e-05, 1.6004e-05, ..., 1.8370e-04, + -1.4944e-03, 7.2956e-05]], device='cuda:0') +Epoch 118, bias, value: tensor([ 0.0145, -0.0137, 0.0011, -0.0178, 0.0135, -0.0005, -0.0160, -0.0240, + 0.0099, -0.0039], device='cuda:0'), grad: tensor([ 0.0249, 0.0039, 0.0260, 0.0429, -0.0170, -0.0505, -0.0300, 0.0668, + -0.0380, -0.0291], device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 216.17, cls_loss 0.5714 cls_loss_mapping 0.0197 cls_loss_causal 0.5328 re_mapping 0.0110 re_causal 0.0278 /// teacc 98.60 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0112, 0.0670, -0.1140, ..., -0.0663, 0.0512, 0.0067], + [-0.0594, -0.0898, 0.0592, ..., -0.0293, -0.0900, 0.0486], + [-0.0522, -0.0291, -0.0743, ..., -0.0922, -0.0330, 0.0340], + ..., + [-0.0167, -0.0526, -0.0721, ..., -0.0868, -0.0037, 0.0481], + [-0.0812, -0.0739, -0.0685, ..., -0.0437, -0.0365, -0.0687], + [ 0.0056, -0.0056, -0.0547, ..., -0.0430, 0.0244, -0.0136]], + device='cuda:0'), grad: tensor([[ 1.6463e-04, -2.2774e-03, 2.7746e-05, ..., 6.5193e-08, + 1.3053e-04, 6.0892e-04], + [ 2.5864e-03, -6.6090e-04, -3.7432e-04, ..., -5.6550e-06, + 7.8487e-04, 9.4795e-04], + [ 1.5602e-03, 1.7262e-04, 2.0564e-05, ..., 2.3749e-07, + 5.3740e-04, -2.0295e-05], + ..., + [ 1.7719e-03, 2.4164e-04, 1.8775e-05, ..., 3.0268e-08, + -6.4015e-05, 5.7220e-04], + [ 1.4467e-03, 9.0408e-04, 2.5797e-04, ..., 5.8487e-06, + -6.6519e-04, 2.5711e-03], + [-1.0780e-02, 4.0627e-04, 1.6555e-05, ..., 1.3318e-07, + -4.4775e-04, -8.7128e-03]], device='cuda:0') +Epoch 119, bias, value: tensor([ 0.0135, -0.0141, 0.0011, -0.0178, 0.0140, -0.0010, -0.0157, -0.0237, + 0.0110, -0.0044], device='cuda:0'), grad: tensor([-0.0016, 0.0095, 0.0131, -0.0095, 0.0271, -0.0167, 0.0168, 0.0007, + 0.0311, -0.0703], device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 216.95, cls_loss 0.5949 cls_loss_mapping 0.0147 cls_loss_causal 0.5617 re_mapping 0.0113 re_causal 0.0293 /// teacc 98.72 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0104, 0.0673, -0.1155, ..., -0.0668, 0.0502, 0.0067], + [-0.0606, -0.0902, 0.0588, ..., -0.0292, -0.0904, 0.0486], + [-0.0524, -0.0281, -0.0743, ..., -0.0928, -0.0317, 0.0333], + ..., + [-0.0162, -0.0532, -0.0725, ..., -0.0866, -0.0037, 0.0477], + [-0.0822, -0.0738, -0.0679, ..., -0.0439, -0.0371, -0.0674], + [ 0.0054, -0.0083, -0.0551, ..., -0.0437, 0.0240, -0.0122]], + device='cuda:0'), grad: tensor([[ 8.7595e-04, 1.1879e-04, 1.0902e-04, ..., 3.4285e-04, + 4.5705e-04, -1.6201e-04], + [-1.5345e-03, 1.1645e-05, -1.2007e-03, ..., -2.6836e-03, + -1.5469e-03, -1.4896e-03], + [-1.6518e-03, 2.6584e-04, 3.8290e-04, ..., 6.2418e-04, + -3.8075e-04, -6.8569e-04], + ..., + [-7.9727e-03, -4.1795e-04, 1.3065e-04, ..., 2.7442e-04, + -7.0457e-03, -1.0597e-02], + [ 1.0567e-03, -1.8835e-04, -1.6734e-05, ..., 2.5368e-04, + 5.2118e-04, 1.1501e-03], + [ 1.6785e-03, 2.5845e-04, 1.5867e-04, ..., 4.3225e-04, + 1.4448e-03, 2.4986e-03]], device='cuda:0') +Epoch 120, bias, value: tensor([ 0.0132, -0.0147, 0.0006, -0.0181, 0.0141, 0.0001, -0.0152, -0.0239, + 0.0104, -0.0034], device='cuda:0'), grad: tensor([-0.0161, -0.0072, -0.0017, 0.0183, 0.0422, 0.0241, 0.0275, -0.0987, + -0.0123, 0.0238], device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 216.40, cls_loss 0.6020 cls_loss_mapping 0.0119 cls_loss_causal 0.5700 re_mapping 0.0100 re_causal 0.0276 /// teacc 98.73 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0110, 0.0670, -0.1157, ..., -0.0666, 0.0508, 0.0065], + [-0.0603, -0.0901, 0.0591, ..., -0.0303, -0.0899, 0.0490], + [-0.0533, -0.0273, -0.0754, ..., -0.0936, -0.0316, 0.0333], + ..., + [-0.0169, -0.0529, -0.0733, ..., -0.0878, -0.0045, 0.0475], + [-0.0825, -0.0741, -0.0678, ..., -0.0434, -0.0378, -0.0683], + [ 0.0052, -0.0091, -0.0552, ..., -0.0439, 0.0241, -0.0116]], + device='cuda:0'), grad: tensor([[ 2.5234e-03, 9.7334e-05, 3.6025e-04, ..., 2.4366e-04, + -2.5272e-04, -2.8095e-03], + [ 1.8272e-03, 1.0020e-04, 4.0740e-05, ..., 1.5080e-04, + 7.9870e-04, 2.9278e-04], + [ 1.1530e-03, -5.9515e-05, 1.0467e-04, ..., 9.3281e-05, + 5.7697e-04, 1.7667e-04], + ..., + [-8.8577e-03, 1.9681e-04, 2.3425e-04, ..., -1.0281e-03, + -1.1978e-03, -4.6806e-03], + [ 2.5673e-03, 2.3866e-04, 3.5262e-04, ..., 3.3355e-04, + 1.2550e-03, 1.1911e-03], + [-1.8740e-03, -1.1082e-03, -4.1771e-04, ..., -2.9397e-04, + -9.5320e-04, 5.1117e-04]], device='cuda:0') +Epoch 121, bias, value: tensor([ 1.3246e-02, -1.4051e-02, 3.8063e-05, -1.7904e-02, 1.4824e-02, + 3.8939e-04, -1.5240e-02, -2.5350e-02, 1.0912e-02, -3.7545e-03], + device='cuda:0'), grad: tensor([-0.0111, -0.0113, -0.0209, 0.0226, -0.0005, 0.0236, 0.0134, -0.0174, + 0.0355, -0.0339], device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 216.35, cls_loss 0.5988 cls_loss_mapping 0.0174 cls_loss_causal 0.5677 re_mapping 0.0102 re_causal 0.0266 /// teacc 98.74 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0108, 0.0666, -0.1157, ..., -0.0674, 0.0502, 0.0073], + [-0.0606, -0.0899, 0.0603, ..., -0.0301, -0.0916, 0.0486], + [-0.0533, -0.0267, -0.0761, ..., -0.0939, -0.0319, 0.0329], + ..., + [-0.0159, -0.0538, -0.0730, ..., -0.0866, -0.0038, 0.0479], + [-0.0836, -0.0740, -0.0688, ..., -0.0430, -0.0381, -0.0678], + [ 0.0046, -0.0101, -0.0552, ..., -0.0446, 0.0233, -0.0113]], + device='cuda:0'), grad: tensor([[-6.9275e-03, -2.5406e-03, -1.9920e-04, ..., -2.9850e-03, + -3.4256e-03, 1.4763e-03], + [ 1.7939e-03, 2.3752e-05, 3.1978e-05, ..., 5.8860e-05, + 1.2140e-03, 8.8739e-04], + [-1.6266e-02, -6.1226e-03, -1.0857e-02, ..., -3.2120e-03, + -6.3553e-03, -1.7843e-03], + ..., + [ 3.9597e-03, 2.4199e-04, 4.0078e-04, ..., 7.9572e-05, + 2.0866e-03, 1.4009e-03], + [-2.6398e-03, 2.0468e-04, 9.4712e-05, ..., 2.0254e-04, + 1.4629e-03, -2.0580e-03], + [ 4.5280e-03, 2.4045e-04, 2.5439e-04, ..., 4.2009e-04, + 2.6207e-03, 1.9779e-03]], device='cuda:0') +Epoch 122, bias, value: tensor([ 1.3088e-02, -1.4289e-02, 2.3874e-04, -1.8175e-02, 1.4739e-02, + -7.6129e-05, -1.5324e-02, -2.4469e-02, 1.0878e-02, -3.6010e-03], + device='cuda:0'), grad: tensor([-0.0282, 0.0099, -0.0569, -0.0767, 0.0294, 0.0375, 0.0568, 0.0194, + -0.0179, 0.0266], device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 216.27, cls_loss 0.5938 cls_loss_mapping 0.0177 cls_loss_causal 0.5589 re_mapping 0.0101 re_causal 0.0268 /// teacc 98.59 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0109, 0.0674, -0.1150, ..., -0.0690, 0.0503, 0.0072], + [-0.0598, -0.0900, 0.0612, ..., -0.0293, -0.0918, 0.0486], + [-0.0521, -0.0271, -0.0766, ..., -0.0932, -0.0312, 0.0328], + ..., + [-0.0160, -0.0546, -0.0737, ..., -0.0867, -0.0038, 0.0488], + [-0.0840, -0.0742, -0.0698, ..., -0.0439, -0.0374, -0.0687], + [ 0.0054, -0.0101, -0.0541, ..., -0.0448, 0.0244, -0.0117]], + device='cuda:0'), grad: tensor([[ 5.8031e-04, -1.1569e-04, 5.4210e-05, ..., 7.8869e-04, + 1.0023e-03, -8.3733e-04], + [ 2.8210e-03, 4.3631e-05, 4.4680e-04, ..., 2.9063e-04, + 1.5545e-03, 1.6800e-02], + [ 1.7109e-03, 2.2069e-05, 1.0312e-04, ..., 1.0586e-03, + 8.2493e-04, 1.8539e-03], + ..., + [ 5.3139e-03, -7.2908e-04, -2.3613e-03, ..., -2.7823e-04, + -1.9445e-03, 4.0970e-03], + [ 2.7828e-03, 1.0037e-04, 2.3103e-04, ..., 5.3787e-04, + 1.2712e-03, -1.4145e-02], + [-1.2062e-02, 7.3910e-05, 1.7107e-04, ..., 2.0349e-04, + -1.1188e-04, -9.5062e-03]], device='cuda:0') +Epoch 123, bias, value: tensor([ 0.0123, -0.0141, 0.0013, -0.0176, 0.0136, -0.0002, -0.0154, -0.0243, + 0.0106, -0.0033], device='cuda:0'), grad: tensor([-0.0086, 0.0219, 0.0243, 0.0068, -0.0092, 0.0329, -0.0369, 0.0066, + -0.0044, -0.0334], device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 216.40, cls_loss 0.5694 cls_loss_mapping 0.0152 cls_loss_causal 0.5369 re_mapping 0.0111 re_causal 0.0285 /// teacc 98.46 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0109, 0.0687, -0.1150, ..., -0.0690, 0.0501, 0.0075], + [-0.0597, -0.0909, 0.0611, ..., -0.0289, -0.0916, 0.0486], + [-0.0519, -0.0275, -0.0774, ..., -0.0937, -0.0305, 0.0335], + ..., + [-0.0159, -0.0539, -0.0729, ..., -0.0870, -0.0042, 0.0483], + [-0.0837, -0.0747, -0.0712, ..., -0.0439, -0.0369, -0.0685], + [ 0.0054, -0.0104, -0.0533, ..., -0.0436, 0.0240, -0.0123]], + device='cuda:0'), grad: tensor([[ 3.3545e-04, -6.1572e-05, 4.5538e-05, ..., 7.7486e-05, + 7.1812e-04, 2.0180e-03], + [ 4.8780e-04, 1.3041e-04, 1.2815e-04, ..., 4.9293e-05, + 5.9938e-04, 1.0948e-03], + [-7.1144e-04, -2.8944e-04, -9.2745e-05, ..., -2.2757e-04, + -2.4548e-03, -3.7422e-03], + ..., + [-3.2735e-04, 7.5459e-05, 7.3791e-05, ..., 1.8880e-05, + -1.4448e-03, -1.5755e-03], + [ 6.8092e-04, 5.2541e-05, -2.6170e-06, ..., -5.2392e-05, + 1.0080e-03, -1.4000e-03], + [-5.3520e-03, 1.9920e-04, 2.4021e-04, ..., 2.7239e-05, + -4.5204e-03, -2.5129e-04]], device='cuda:0') +Epoch 124, bias, value: tensor([ 0.0124, -0.0133, 0.0015, -0.0173, 0.0137, -0.0005, -0.0157, -0.0249, + 0.0113, -0.0041], device='cuda:0'), grad: tensor([ 0.0178, -0.0014, -0.0163, 0.0152, 0.0089, 0.0131, -0.0114, -0.0180, + 0.0112, -0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 216.31, cls_loss 0.6000 cls_loss_mapping 0.0198 cls_loss_causal 0.5685 re_mapping 0.0102 re_causal 0.0259 /// teacc 98.51 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0116, 0.0691, -0.1151, ..., -0.0685, 0.0503, 0.0078], + [-0.0603, -0.0908, 0.0597, ..., -0.0289, -0.0927, 0.0482], + [-0.0523, -0.0281, -0.0770, ..., -0.0940, -0.0299, 0.0338], + ..., + [-0.0165, -0.0548, -0.0728, ..., -0.0874, -0.0054, 0.0489], + [-0.0834, -0.0747, -0.0712, ..., -0.0440, -0.0366, -0.0689], + [ 0.0063, -0.0104, -0.0534, ..., -0.0455, 0.0256, -0.0125]], + device='cuda:0'), grad: tensor([[ 1.5793e-03, -2.6751e-04, 1.7786e-04, ..., 1.2827e-03, + 9.2840e-04, 1.4496e-03], + [ 2.5368e-03, 4.1544e-05, -1.6189e-04, ..., 2.8496e-03, + 1.5268e-03, 4.0817e-03], + [-3.1319e-03, 1.3387e-04, 5.2547e-04, ..., -1.7796e-03, + -1.7395e-03, -1.5669e-03], + ..., + [ 1.7605e-03, 1.4913e-04, 2.3878e-04, ..., -3.5439e-03, + 1.4400e-03, -4.4861e-03], + [ 3.1872e-03, -1.5631e-03, -1.4915e-03, ..., -1.0691e-03, + 1.8549e-03, 5.2834e-03], + [ 9.2864e-05, 2.9135e-04, 5.0306e-04, ..., -5.0735e-04, + -1.7958e-03, -2.2566e-04]], device='cuda:0') +Epoch 125, bias, value: tensor([ 0.0127, -0.0134, 0.0019, -0.0180, 0.0129, 0.0006, -0.0165, -0.0247, + 0.0110, -0.0035], device='cuda:0'), grad: tensor([ 0.0224, 0.0413, -0.0298, 0.0020, 0.0456, -0.0413, -0.0471, -0.0302, + 0.0362, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 215.99, cls_loss 0.5421 cls_loss_mapping 0.0185 cls_loss_causal 0.5154 re_mapping 0.0108 re_causal 0.0269 /// teacc 98.72 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0123, 0.0699, -0.1159, ..., -0.0676, 0.0501, 0.0080], + [-0.0607, -0.0913, 0.0592, ..., -0.0305, -0.0922, 0.0483], + [-0.0529, -0.0285, -0.0771, ..., -0.0937, -0.0294, 0.0339], + ..., + [-0.0172, -0.0550, -0.0734, ..., -0.0858, -0.0063, 0.0497], + [-0.0838, -0.0750, -0.0719, ..., -0.0442, -0.0361, -0.0681], + [ 0.0062, -0.0101, -0.0532, ..., -0.0460, 0.0254, -0.0130]], + device='cuda:0'), grad: tensor([[ 3.6454e-04, 7.2250e-03, 4.3154e-05, ..., -8.9550e-04, + -1.2875e-03, 1.5717e-03], + [ 8.9741e-04, 6.7174e-05, 4.1693e-05, ..., 3.8767e-04, + 5.1069e-04, 1.1110e-03], + [-2.6464e-04, 7.7248e-04, 1.1939e-04, ..., 1.1797e-03, + 2.5153e-04, -5.8746e-04], + ..., + [-2.2545e-03, 3.9673e-04, 8.0168e-05, ..., 5.0592e-04, + -5.1975e-04, -4.7150e-03], + [ 5.7487e-03, 3.0594e-03, 1.5697e-03, ..., 2.7394e-04, + 3.6469e-03, -1.9188e-03], + [ 7.3738e-03, -7.4615e-03, 4.0102e-04, ..., 1.0948e-03, + 3.5191e-03, 2.9926e-03]], device='cuda:0') +Epoch 126, bias, value: tensor([ 0.0123, -0.0134, 0.0025, -0.0179, 0.0137, 0.0004, -0.0168, -0.0246, + 0.0110, -0.0041], device='cuda:0'), grad: tensor([ 0.0313, 0.0133, -0.0120, -0.0382, 0.0102, -0.0082, -0.0031, -0.0322, + 0.0177, 0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 216.64, cls_loss 0.5718 cls_loss_mapping 0.0138 cls_loss_causal 0.5400 re_mapping 0.0100 re_causal 0.0260 /// teacc 98.69 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0126, 0.0690, -0.1161, ..., -0.0671, 0.0499, 0.0081], + [-0.0596, -0.0912, 0.0594, ..., -0.0301, -0.0904, 0.0490], + [-0.0531, -0.0273, -0.0785, ..., -0.0945, -0.0292, 0.0337], + ..., + [-0.0168, -0.0555, -0.0738, ..., -0.0872, -0.0063, 0.0502], + [-0.0841, -0.0744, -0.0722, ..., -0.0440, -0.0363, -0.0679], + [ 0.0059, -0.0109, -0.0535, ..., -0.0473, 0.0253, -0.0133]], + device='cuda:0'), grad: tensor([[ 1.8778e-03, -3.5686e-03, -2.1309e-05, ..., 1.9836e-04, + 1.8063e-03, -5.0087e-03], + [-2.3384e-03, 1.4048e-03, 4.7870e-06, ..., 2.2674e-04, + -1.8578e-03, -1.1148e-03], + [-1.6832e-03, 3.0965e-05, 2.6170e-06, ..., -1.3132e-03, + -3.1471e-03, 1.1292e-03], + ..., + [-9.5177e-04, 9.8407e-05, 1.8496e-06, ..., 2.2697e-04, + -3.9959e-04, -9.3031e-04], + [ 1.4858e-03, 2.0733e-03, 1.6928e-05, ..., -1.6272e-05, + 1.5812e-03, 4.9896e-03], + [ 1.5602e-03, -2.1076e-04, 1.1587e-04, ..., 4.3130e-04, + 6.8426e-04, 8.4305e-04]], device='cuda:0') +Epoch 127, bias, value: tensor([ 0.0123, -0.0122, 0.0016, -0.0179, 0.0131, 0.0007, -0.0166, -0.0248, + 0.0114, -0.0045], device='cuda:0'), grad: tensor([-0.0053, 0.0100, -0.0454, -0.0097, -0.0287, 0.0149, 0.0329, -0.0128, + 0.0264, 0.0179], device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 215.96, cls_loss 0.6059 cls_loss_mapping 0.0139 cls_loss_causal 0.5697 re_mapping 0.0100 re_causal 0.0267 /// teacc 98.65 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0129, 0.0696, -0.1160, ..., -0.0684, 0.0505, 0.0076], + [-0.0589, -0.0920, 0.0599, ..., -0.0291, -0.0891, 0.0489], + [-0.0533, -0.0277, -0.0775, ..., -0.0952, -0.0296, 0.0348], + ..., + [-0.0170, -0.0568, -0.0744, ..., -0.0875, -0.0059, 0.0498], + [-0.0842, -0.0754, -0.0733, ..., -0.0453, -0.0371, -0.0688], + [ 0.0062, -0.0102, -0.0533, ..., -0.0472, 0.0256, -0.0125]], + device='cuda:0'), grad: tensor([[-2.4033e-03, 6.4898e-04, 2.0266e-05, ..., -8.2541e-04, + -4.2152e-03, -2.0889e-02], + [-1.6479e-03, 5.9456e-05, 1.9064e-06, ..., 5.7364e-04, + -7.3242e-04, 1.5745e-03], + [ 1.3227e-03, 5.6887e-04, 5.8365e-04, ..., 1.0900e-03, + 1.4744e-03, 3.3073e-03], + ..., + [ 2.1305e-03, 9.8407e-05, 3.2187e-05, ..., 6.7902e-04, + 3.6736e-03, 1.3535e-02], + [ 1.2970e-04, 4.3893e-04, 2.3460e-04, ..., 8.5735e-04, + -1.3771e-03, 6.9666e-04], + [ 3.0804e-03, 9.2125e-04, 1.4293e-04, ..., 2.1458e-03, + 2.5368e-03, 2.5635e-03]], device='cuda:0') +Epoch 128, bias, value: tensor([ 0.0119, -0.0114, 0.0011, -0.0182, 0.0132, 0.0019, -0.0172, -0.0249, + 0.0108, -0.0042], device='cuda:0'), grad: tensor([-0.0803, 0.0040, 0.0177, -0.0011, 0.0123, 0.0359, -0.0478, 0.0399, + 0.0013, 0.0182], device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 216.90, cls_loss 0.5682 cls_loss_mapping 0.0148 cls_loss_causal 0.5435 re_mapping 0.0102 re_causal 0.0270 /// teacc 98.73 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0128, 0.0699, -0.1154, ..., -0.0687, 0.0502, 0.0079], + [-0.0586, -0.0912, 0.0607, ..., -0.0292, -0.0889, 0.0481], + [-0.0535, -0.0293, -0.0791, ..., -0.0952, -0.0279, 0.0349], + ..., + [-0.0170, -0.0567, -0.0754, ..., -0.0877, -0.0060, 0.0502], + [-0.0848, -0.0762, -0.0742, ..., -0.0467, -0.0388, -0.0688], + [ 0.0063, -0.0124, -0.0536, ..., -0.0479, 0.0252, -0.0127]], + device='cuda:0'), grad: tensor([[ 0.0017, 0.0004, 0.0005, ..., 0.0014, 0.0017, 0.0011], + [ 0.0022, 0.0005, 0.0005, ..., 0.0020, 0.0022, 0.0018], + [-0.0003, 0.0016, 0.0012, ..., 0.0020, -0.0017, 0.0007], + ..., + [-0.0024, -0.0026, -0.0020, ..., -0.0087, -0.0035, -0.0026], + [ 0.0029, 0.0028, 0.0020, ..., 0.0029, 0.0029, 0.0017], + [-0.0025, 0.0004, 0.0004, ..., -0.0009, -0.0023, -0.0026]], + device='cuda:0') +Epoch 129, bias, value: tensor([ 0.0130, -0.0109, 0.0006, -0.0189, 0.0139, 0.0017, -0.0170, -0.0250, + 0.0106, -0.0048], device='cuda:0'), grad: tensor([ 0.0128, 0.0186, -0.0023, -0.0147, 0.0043, 0.0143, 0.0267, -0.0196, + 0.0208, -0.0609], device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 216.76, cls_loss 0.5622 cls_loss_mapping 0.0125 cls_loss_causal 0.5316 re_mapping 0.0099 re_causal 0.0271 /// teacc 98.55 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0140, 0.0711, -0.1171, ..., -0.0689, 0.0507, 0.0076], + [-0.0595, -0.0921, 0.0609, ..., -0.0308, -0.0885, 0.0480], + [-0.0533, -0.0295, -0.0783, ..., -0.0945, -0.0275, 0.0347], + ..., + [-0.0163, -0.0546, -0.0759, ..., -0.0874, -0.0056, 0.0507], + [-0.0852, -0.0773, -0.0742, ..., -0.0452, -0.0401, -0.0696], + [ 0.0059, -0.0134, -0.0541, ..., -0.0494, 0.0250, -0.0133]], + device='cuda:0'), grad: tensor([[ 0.0005, -0.0003, 0.0003, ..., 0.0008, -0.0004, 0.0005], + [ 0.0014, 0.0014, 0.0020, ..., 0.0034, 0.0015, 0.0016], + [ 0.0002, 0.0009, 0.0009, ..., 0.0005, 0.0010, 0.0033], + ..., + [-0.0011, 0.0014, -0.0004, ..., -0.0009, -0.0012, -0.0004], + [ 0.0014, 0.0011, 0.0004, ..., 0.0007, 0.0010, 0.0016], + [-0.0009, -0.0004, -0.0008, ..., -0.0017, -0.0065, -0.0092]], + device='cuda:0') +Epoch 130, bias, value: tensor([ 0.0126, -0.0106, 0.0004, -0.0186, 0.0144, 0.0019, -0.0172, -0.0249, + 0.0102, -0.0050], device='cuda:0'), grad: tensor([ 0.0128, 0.0263, 0.0064, 0.0289, 0.0320, 0.0309, -0.0866, -0.0104, + 0.0267, -0.0671], device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 216.14, cls_loss 0.5512 cls_loss_mapping 0.0123 cls_loss_causal 0.5130 re_mapping 0.0096 re_causal 0.0247 /// teacc 98.74 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0147, 0.0722, -0.1166, ..., -0.0675, 0.0515, 0.0074], + [-0.0604, -0.0944, 0.0602, ..., -0.0318, -0.0887, 0.0481], + [-0.0531, -0.0303, -0.0794, ..., -0.0956, -0.0279, 0.0362], + ..., + [-0.0174, -0.0553, -0.0763, ..., -0.0879, -0.0062, 0.0504], + [-0.0865, -0.0765, -0.0727, ..., -0.0449, -0.0403, -0.0698], + [ 0.0071, -0.0115, -0.0545, ..., -0.0504, 0.0263, -0.0141]], + device='cuda:0'), grad: tensor([[ 3.7975e-03, 2.0332e-03, 7.2670e-04, ..., 2.4967e-03, + 3.4943e-03, 2.0218e-03], + [ 1.1787e-03, -4.8614e-04, 1.1520e-03, ..., 1.0147e-03, + -6.9275e-03, -2.1454e-02], + [ 4.2992e-03, 3.2825e-03, 1.1768e-03, ..., 1.0643e-03, + 1.4496e-02, 2.2110e-02], + ..., + [ 3.3531e-03, 9.0408e-04, -2.2926e-03, ..., 9.9540e-05, + 2.8915e-03, 8.5449e-04], + [ 2.8000e-03, 1.4467e-03, 7.6914e-04, ..., 8.7833e-04, + 2.0580e-03, 1.8873e-03], + [ 1.9369e-03, 1.8721e-03, 9.1982e-04, ..., 5.1349e-05, + -1.5485e-04, -2.1248e-03]], device='cuda:0') +Epoch 131, bias, value: tensor([ 0.0142, -0.0114, 0.0008, -0.0181, 0.0149, 0.0004, -0.0174, -0.0244, + 0.0099, -0.0055], device='cuda:0'), grad: tensor([ 0.0248, -0.0246, 0.0670, -0.0428, -0.0948, 0.0166, 0.0118, 0.0251, + 0.0173, -0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 216.21, cls_loss 0.5641 cls_loss_mapping 0.0129 cls_loss_causal 0.5325 re_mapping 0.0101 re_causal 0.0252 /// teacc 98.64 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0140, 0.0716, -0.1170, ..., -0.0679, 0.0506, 0.0073], + [-0.0609, -0.0939, 0.0600, ..., -0.0310, -0.0891, 0.0496], + [-0.0537, -0.0303, -0.0797, ..., -0.0956, -0.0284, 0.0356], + ..., + [-0.0188, -0.0586, -0.0766, ..., -0.0891, -0.0079, 0.0494], + [-0.0875, -0.0773, -0.0730, ..., -0.0456, -0.0403, -0.0695], + [ 0.0081, -0.0094, -0.0542, ..., -0.0510, 0.0272, -0.0132]], + device='cuda:0'), grad: tensor([[ 2.0103e-03, 7.4148e-04, 3.0994e-04, ..., 2.9159e-04, + 7.0572e-04, 1.2426e-03], + [ 2.7809e-03, 1.2369e-03, 4.9019e-04, ..., 9.3460e-04, + 1.3523e-03, 1.4706e-03], + [-5.3883e-04, -1.7452e-03, 2.4068e-04, ..., 3.0398e-04, + -1.2016e-03, -2.6302e-03], + ..., + [ 4.6616e-03, 6.6233e-04, 7.8559e-05, ..., 2.9516e-04, + 1.6022e-03, 2.2984e-03], + [-7.2174e-03, 7.3814e-04, 4.7064e-04, ..., 5.6410e-04, + 1.0204e-03, 1.2159e-03], + [-1.5869e-03, 4.4882e-05, 1.1139e-03, ..., -2.2469e-03, + -1.7328e-03, -1.7233e-03]], device='cuda:0') +Epoch 132, bias, value: tensor([ 1.3573e-02, -1.0839e-02, 9.5948e-05, -1.8297e-02, 1.4488e-02, + 6.5863e-05, -1.6638e-02, -2.4721e-02, 9.5707e-03, -4.0979e-03], + device='cuda:0'), grad: tensor([ 0.0104, 0.0134, -0.0046, 0.0256, 0.0262, -0.0507, -0.0133, 0.0179, + -0.0176, -0.0074], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 131---------------------------------------------------- +epoch 131, time 216.72, cls_loss 0.5455 cls_loss_mapping 0.0137 cls_loss_causal 0.5122 re_mapping 0.0103 re_causal 0.0267 /// teacc 98.81 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0147, 0.0716, -0.1182, ..., -0.0679, 0.0517, 0.0060], + [-0.0617, -0.0946, 0.0610, ..., -0.0322, -0.0895, 0.0487], + [-0.0545, -0.0295, -0.0788, ..., -0.0959, -0.0289, 0.0364], + ..., + [-0.0190, -0.0597, -0.0779, ..., -0.0898, -0.0076, 0.0505], + [-0.0876, -0.0777, -0.0735, ..., -0.0458, -0.0400, -0.0691], + [ 0.0078, -0.0104, -0.0552, ..., -0.0504, 0.0263, -0.0127]], + device='cuda:0'), grad: tensor([[-1.1282e-03, -3.6025e-04, 1.2243e-04, ..., 1.0002e-04, + -9.7370e-04, -2.4414e-04], + [ 2.5940e-04, 2.4390e-04, 2.1648e-03, ..., 3.4428e-03, + 8.5115e-04, 3.7651e-03], + [-7.3700e-03, -5.9700e-03, -2.9831e-03, ..., -4.9362e-03, + -6.2675e-03, -1.3878e-02], + ..., + [ 1.6251e-03, 1.0729e-03, 1.6785e-04, ..., 1.2118e-04, + 1.5287e-03, -1.2379e-03], + [ 5.4398e-03, 1.4715e-03, 9.6941e-04, ..., 8.1873e-04, + 1.6565e-03, 6.6833e-03], + [-1.7624e-02, -7.7343e-04, -5.9271e-04, ..., 2.2128e-05, + -5.8403e-03, -1.1353e-02]], device='cuda:0') +Epoch 133, bias, value: tensor([ 1.2919e-02, -1.1298e-02, 8.8552e-05, -1.8257e-02, 1.4161e-02, + -2.5965e-04, -1.6212e-02, -2.4441e-02, 1.0537e-02, -4.1448e-03], + device='cuda:0'), grad: tensor([-0.0115, 0.0068, -0.0357, 0.0316, 0.0300, 0.0189, 0.0227, -0.0032, + 0.0352, -0.0948], device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 215.98, cls_loss 0.5852 cls_loss_mapping 0.0132 cls_loss_causal 0.5559 re_mapping 0.0101 re_causal 0.0265 /// teacc 98.72 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0154, 0.0726, -0.1176, ..., -0.0679, 0.0524, 0.0074], + [-0.0627, -0.0946, 0.0614, ..., -0.0314, -0.0895, 0.0490], + [-0.0555, -0.0291, -0.0789, ..., -0.0960, -0.0296, 0.0350], + ..., + [-0.0191, -0.0601, -0.0787, ..., -0.0900, -0.0074, 0.0511], + [-0.0872, -0.0770, -0.0737, ..., -0.0457, -0.0404, -0.0688], + [ 0.0082, -0.0127, -0.0559, ..., -0.0519, 0.0255, -0.0135]], + device='cuda:0'), grad: tensor([[ 5.4855e-03, 1.8740e-03, 5.5027e-04, ..., 4.9925e-04, + 3.0766e-03, 2.7962e-03], + [-2.1801e-03, -5.8670e-03, -2.0103e-03, ..., -1.9379e-03, + -1.1330e-03, 5.8708e-03], + [ 2.9812e-03, 7.2479e-04, 4.6730e-04, ..., 2.0611e-04, + 1.9150e-03, 1.8435e-03], + ..., + [-2.7728e-04, 4.1771e-04, 1.6940e-04, ..., 8.4996e-05, + 1.1435e-03, -1.9093e-03], + [-2.8057e-03, 9.3985e-04, 2.3627e-04, ..., 2.2992e-05, + -8.1015e-04, -2.2221e-03], + [-1.0071e-03, 6.7139e-04, 2.1303e-04, ..., 1.6248e-04, + -1.2703e-03, -1.3762e-03]], device='cuda:0') +Epoch 134, bias, value: tensor([ 0.0136, -0.0112, -0.0004, -0.0171, 0.0147, -0.0008, -0.0163, -0.0252, + 0.0103, -0.0044], device='cuda:0'), grad: tensor([ 0.0391, 0.0225, 0.0250, 0.0242, -0.0525, -0.0328, 0.0098, 0.0032, + -0.0387, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 133---------------------------------------------------- +epoch 133, time 217.03, cls_loss 0.5615 cls_loss_mapping 0.0120 cls_loss_causal 0.5317 re_mapping 0.0099 re_causal 0.0258 /// teacc 98.87 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0149, 0.0718, -0.1180, ..., -0.0675, 0.0522, 0.0070], + [-0.0631, -0.0923, 0.0616, ..., -0.0310, -0.0898, 0.0484], + [-0.0555, -0.0289, -0.0796, ..., -0.0968, -0.0289, 0.0351], + ..., + [-0.0186, -0.0600, -0.0791, ..., -0.0897, -0.0070, 0.0518], + [-0.0872, -0.0782, -0.0741, ..., -0.0459, -0.0413, -0.0687], + [ 0.0086, -0.0134, -0.0562, ..., -0.0527, 0.0254, -0.0138]], + device='cuda:0'), grad: tensor([[ 9.9659e-04, -1.3857e-03, 1.0580e-04, ..., 6.8331e-04, + 1.0462e-03, 2.4986e-03], + [ 4.0245e-03, -1.1021e-04, -3.8457e-04, ..., 9.3126e-04, + 2.6913e-03, -3.1223e-03], + [-1.5511e-02, 1.3724e-05, 8.6129e-05, ..., -3.8509e-03, + -1.0567e-02, -1.6891e-02], + ..., + [ 1.5993e-03, -1.2529e-04, 2.2984e-04, ..., 7.6771e-04, + 1.3971e-03, 5.0850e-03], + [-2.0332e-03, 1.2197e-03, 1.2016e-04, ..., -1.1879e-04, + 2.4021e-04, -1.7643e-03], + [ 4.0512e-03, 5.1785e-04, 3.4720e-05, ..., 4.4227e-04, + 2.5539e-03, 4.5090e-03]], device='cuda:0') +Epoch 135, bias, value: tensor([ 0.0138, -0.0111, -0.0002, -0.0173, 0.0136, -0.0012, -0.0158, -0.0247, + 0.0106, -0.0046], device='cuda:0'), grad: tensor([ 0.0143, 0.0013, -0.1035, 0.0276, 0.0288, -0.0522, 0.0309, 0.0309, + -0.0045, 0.0264], device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 216.22, cls_loss 0.5361 cls_loss_mapping 0.0133 cls_loss_causal 0.5083 re_mapping 0.0098 re_causal 0.0254 /// teacc 98.62 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0156, 0.0717, -0.1195, ..., -0.0681, 0.0529, 0.0074], + [-0.0624, -0.0924, 0.0620, ..., -0.0309, -0.0901, 0.0502], + [-0.0555, -0.0290, -0.0804, ..., -0.0971, -0.0291, 0.0346], + ..., + [-0.0183, -0.0601, -0.0788, ..., -0.0899, -0.0067, 0.0524], + [-0.0878, -0.0784, -0.0739, ..., -0.0458, -0.0416, -0.0699], + [ 0.0080, -0.0131, -0.0547, ..., -0.0514, 0.0252, -0.0137]], + device='cuda:0'), grad: tensor([[ 8.2626e-03, 6.8207e-03, 1.0157e-03, ..., 1.5678e-03, + 6.1569e-03, 5.3749e-03], + [ 2.2945e-03, 1.6093e-04, -1.1116e-05, ..., 1.5572e-05, + 1.7557e-03, -1.7142e-04], + [ 3.1452e-03, 1.3485e-03, 1.8740e-04, ..., 7.4029e-05, + 2.3193e-03, 1.8244e-03], + ..., + [-2.1820e-03, -5.1641e-04, 3.4153e-05, ..., 9.7156e-05, + 2.7866e-03, 8.7678e-05], + [-5.5933e-04, 6.5231e-04, 2.1577e-04, ..., 1.1283e-04, + -4.3182e-03, -1.0195e-03], + [-6.2904e-03, -9.6970e-03, -1.4896e-03, ..., -2.1667e-03, + -5.4474e-03, -2.2068e-03]], device='cuda:0') +Epoch 136, bias, value: tensor([ 0.0140, -0.0104, -0.0005, -0.0181, 0.0138, -0.0009, -0.0165, -0.0241, + 0.0102, -0.0045], device='cuda:0'), grad: tensor([ 0.0589, 0.0098, 0.0159, -0.0073, 0.0255, -0.0356, 0.0120, -0.0099, + -0.0592, -0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 216.19, cls_loss 0.5576 cls_loss_mapping 0.0149 cls_loss_causal 0.5265 re_mapping 0.0090 re_causal 0.0235 /// teacc 98.71 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0147, 0.0715, -0.1204, ..., -0.0675, 0.0526, 0.0068], + [-0.0623, -0.0929, 0.0618, ..., -0.0320, -0.0907, 0.0506], + [-0.0542, -0.0284, -0.0802, ..., -0.0984, -0.0281, 0.0356], + ..., + [-0.0191, -0.0620, -0.0805, ..., -0.0920, -0.0077, 0.0521], + [-0.0885, -0.0791, -0.0749, ..., -0.0463, -0.0427, -0.0715], + [ 0.0094, -0.0119, -0.0534, ..., -0.0516, 0.0265, -0.0137]], + device='cuda:0'), grad: tensor([[-1.6565e-03, -2.7199e-03, 9.1195e-05, ..., 2.7514e-04, + -1.0252e-03, -5.1832e-04], + [-3.2482e-03, 1.7202e-04, -4.0317e-04, ..., -1.5125e-03, + -2.1896e-03, -1.7052e-03], + [ 2.5654e-03, 5.4312e-04, 3.3092e-04, ..., 8.5068e-04, + 2.0485e-03, 1.4238e-03], + ..., + [ 3.4275e-03, 2.7084e-04, 1.1510e-04, ..., 2.1517e-04, + 1.6451e-03, 1.4582e-03], + [ 3.7384e-04, 1.0881e-03, 4.3750e-04, ..., 8.2302e-04, + 4.2915e-05, -2.5826e-03], + [ 3.1033e-03, 8.4686e-04, 1.8513e-04, ..., 3.3236e-04, + 2.5444e-03, 2.1591e-03]], device='cuda:0') +Epoch 137, bias, value: tensor([ 0.0128, -0.0095, 0.0004, -0.0182, 0.0131, -0.0013, -0.0160, -0.0245, + 0.0100, -0.0037], device='cuda:0'), grad: tensor([-0.0264, -0.0108, 0.0254, -0.0412, -0.0057, 0.0186, 0.0206, 0.0224, + -0.0206, 0.0177], device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 216.54, cls_loss 0.5567 cls_loss_mapping 0.0204 cls_loss_causal 0.5262 re_mapping 0.0097 re_causal 0.0241 /// teacc 98.46 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0148, 0.0716, -0.1213, ..., -0.0676, 0.0522, 0.0082], + [-0.0616, -0.0931, 0.0622, ..., -0.0322, -0.0901, 0.0497], + [-0.0553, -0.0285, -0.0811, ..., -0.0988, -0.0283, 0.0363], + ..., + [-0.0196, -0.0629, -0.0809, ..., -0.0906, -0.0078, 0.0518], + [-0.0888, -0.0792, -0.0740, ..., -0.0461, -0.0437, -0.0721], + [ 0.0100, -0.0119, -0.0531, ..., -0.0514, 0.0270, -0.0143]], + device='cuda:0'), grad: tensor([[-4.0970e-03, -4.1885e-03, 5.6550e-06, ..., -8.1863e-03, + -1.1833e-02, -2.1477e-03], + [ 5.3215e-04, -8.3074e-06, -4.2987e-04, ..., 4.4060e-04, + 1.5602e-03, 9.2316e-04], + [ 6.8808e-04, 3.6669e-04, 7.3552e-05, ..., 9.1076e-04, + 2.5520e-03, 2.2602e-03], + ..., + [ 6.2037e-04, 1.5867e-04, 2.0400e-05, ..., 2.6703e-04, + 1.3456e-03, 1.1425e-03], + [ 8.2684e-04, 1.8120e-04, 1.9181e-04, ..., 1.1616e-03, + 2.0332e-03, 6.2513e-04], + [-5.4979e-04, 5.0545e-04, -9.7632e-05, ..., 4.4703e-04, + 7.4100e-04, 6.1226e-04]], device='cuda:0') +Epoch 138, bias, value: tensor([ 0.0139, -0.0096, -0.0003, -0.0178, 0.0132, -0.0017, -0.0156, -0.0242, + 0.0091, -0.0038], device='cuda:0'), grad: tensor([-0.0681, 0.0103, 0.0210, -0.0006, 0.0140, -0.0131, 0.0159, 0.0148, + 0.0169, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 216.11, cls_loss 0.5654 cls_loss_mapping 0.0128 cls_loss_causal 0.5333 re_mapping 0.0105 re_causal 0.0276 /// teacc 98.75 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 0.0148, 0.0725, -0.1216, ..., -0.0671, 0.0531, 0.0091], + [-0.0618, -0.0932, 0.0615, ..., -0.0321, -0.0894, 0.0498], + [-0.0563, -0.0281, -0.0801, ..., -0.0992, -0.0298, 0.0355], + ..., + [-0.0198, -0.0645, -0.0807, ..., -0.0918, -0.0095, 0.0519], + [-0.0888, -0.0793, -0.0751, ..., -0.0457, -0.0440, -0.0726], + [ 0.0103, -0.0115, -0.0539, ..., -0.0518, 0.0283, -0.0139]], + device='cuda:0'), grad: tensor([[ 2.3003e-03, 8.6927e-04, 1.6081e-04, ..., 2.4986e-04, + 1.2846e-03, 1.7118e-03], + [-1.0204e-03, -1.2150e-03, -6.2990e-04, ..., -5.0879e-04, + -1.2293e-03, -1.0614e-03], + [ 1.5402e-03, 8.5545e-04, 3.6240e-04, ..., 3.9053e-04, + -1.4362e-03, -7.8087e-03], + ..., + [-5.2595e-04, 9.6941e-04, 3.0065e-04, ..., -1.4172e-03, + 6.7997e-04, 9.3699e-04], + [ 2.7542e-03, 1.9207e-03, 9.5654e-04, ..., 2.0657e-03, + 2.2964e-03, 2.3479e-03], + [ 1.1002e-02, 1.1757e-02, 4.0855e-03, ..., 2.0714e-03, + 1.0201e-02, -4.9174e-05]], device='cuda:0') +Epoch 139, bias, value: tensor([ 0.0145, -0.0092, -0.0013, -0.0182, 0.0141, -0.0015, -0.0163, -0.0249, + 0.0089, -0.0030], device='cuda:0'), grad: tensor([ 0.0206, -0.0121, 0.0054, 0.0106, -0.1218, 0.0247, 0.0146, -0.0603, + 0.0411, 0.0774], device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 216.16, cls_loss 0.5405 cls_loss_mapping 0.0124 cls_loss_causal 0.5168 re_mapping 0.0098 re_causal 0.0267 /// teacc 98.59 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 0.0142, 0.0732, -0.1231, ..., -0.0676, 0.0526, 0.0087], + [-0.0618, -0.0936, 0.0623, ..., -0.0320, -0.0891, 0.0502], + [-0.0567, -0.0298, -0.0815, ..., -0.0997, -0.0302, 0.0367], + ..., + [-0.0189, -0.0644, -0.0807, ..., -0.0913, -0.0084, 0.0523], + [-0.0879, -0.0788, -0.0754, ..., -0.0457, -0.0436, -0.0731], + [ 0.0101, -0.0128, -0.0542, ..., -0.0505, 0.0276, -0.0142]], + device='cuda:0'), grad: tensor([[-8.2922e-04, -3.5229e-03, 9.5189e-05, ..., 8.1718e-05, + 7.1168e-05, -4.4632e-03], + [ 9.9659e-04, 4.9263e-05, -6.1369e-04, ..., -4.3678e-04, + 5.1355e-04, 5.3024e-04], + [ 1.3103e-03, 4.5471e-03, 8.1873e-04, ..., 6.7854e-04, + 1.0300e-03, 3.4828e-03], + ..., + [-4.6921e-03, 7.8619e-05, 1.2100e-04, ..., 1.1718e-04, + -2.6474e-03, -2.0046e-03], + [-1.0538e-03, -6.3019e-03, -5.8403e-03, ..., -4.3154e-04, + 1.8406e-04, 1.2693e-03], + [ 1.5793e-03, 1.2026e-03, 1.2426e-03, ..., 9.6202e-05, + 6.4278e-04, 1.2493e-03]], device='cuda:0') +Epoch 140, bias, value: tensor([ 0.0147, -0.0086, -0.0007, -0.0183, 0.0133, -0.0015, -0.0169, -0.0246, + 0.0089, -0.0032], device='cuda:0'), grad: tensor([-0.0318, 0.0057, 0.0361, 0.0127, 0.0131, 0.0289, -0.0243, -0.0438, + -0.0148, 0.0181], device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 216.19, cls_loss 0.5521 cls_loss_mapping 0.0135 cls_loss_causal 0.5214 re_mapping 0.0097 re_causal 0.0262 /// teacc 98.63 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 0.0150, 0.0727, -0.1247, ..., -0.0670, 0.0531, 0.0095], + [-0.0614, -0.0945, 0.0629, ..., -0.0333, -0.0890, 0.0506], + [-0.0564, -0.0312, -0.0823, ..., -0.1022, -0.0300, 0.0368], + ..., + [-0.0193, -0.0643, -0.0798, ..., -0.0911, -0.0088, 0.0526], + [-0.0885, -0.0769, -0.0760, ..., -0.0459, -0.0452, -0.0736], + [ 0.0101, -0.0147, -0.0559, ..., -0.0516, 0.0275, -0.0147]], + device='cuda:0'), grad: tensor([[ 3.3073e-03, 2.7752e-03, 3.2845e-03, ..., -1.1545e-04, + 2.0752e-03, 2.9602e-03], + [-1.5144e-03, -7.8440e-04, -1.5278e-03, ..., -2.0542e-03, + -9.3603e-04, -1.6565e-03], + [-9.4461e-04, 3.3712e-04, 9.2924e-05, ..., 9.6512e-04, + -5.0640e-04, -2.4104e-04], + ..., + [-1.3723e-03, 5.6362e-04, 1.0271e-03, ..., 5.2786e-04, + -9.4986e-04, -6.2141e-03], + [ 2.9469e-03, 4.0970e-03, 4.1733e-03, ..., 3.4122e-03, + 1.4515e-03, 1.7824e-03], + [ 1.4887e-03, 1.0958e-03, 2.1496e-03, ..., 1.1415e-03, + 1.0004e-03, 1.8797e-03]], device='cuda:0') +Epoch 141, bias, value: tensor([ 0.0155, -0.0083, -0.0006, -0.0190, 0.0125, -0.0023, -0.0161, -0.0245, + 0.0090, -0.0030], device='cuda:0'), grad: tensor([ 0.0157, -0.0344, -0.0068, -0.0163, -0.0114, -0.0448, 0.0501, -0.0059, + 0.0323, 0.0214], device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 216.02, cls_loss 0.5831 cls_loss_mapping 0.0113 cls_loss_causal 0.5500 re_mapping 0.0094 re_causal 0.0255 /// teacc 98.68 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 0.0168, 0.0723, -0.1265, ..., -0.0683, 0.0536, 0.0093], + [-0.0621, -0.0948, 0.0638, ..., -0.0324, -0.0897, 0.0495], + [-0.0574, -0.0322, -0.0839, ..., -0.1028, -0.0292, 0.0375], + ..., + [-0.0197, -0.0648, -0.0777, ..., -0.0904, -0.0103, 0.0523], + [-0.0891, -0.0769, -0.0770, ..., -0.0464, -0.0459, -0.0730], + [ 0.0107, -0.0153, -0.0567, ..., -0.0519, 0.0273, -0.0138]], + device='cuda:0'), grad: tensor([[ 1.0467e-04, 7.8869e-04, -1.6761e-04, ..., -3.2878e-04, + 4.9019e-04, -4.8161e-04], + [ 3.4165e-04, 1.9372e-04, -1.4949e-04, ..., -3.2067e-04, + 2.1019e-03, 1.7920e-03], + [ 2.2392e-03, 1.7490e-03, 9.8288e-05, ..., 2.8586e-04, + 2.4815e-03, 2.5940e-03], + ..., + [-1.8225e-03, -2.3060e-03, 1.9312e-04, ..., 3.2520e-04, + -9.3613e-03, -1.6647e-02], + [-4.5991e-04, 1.0071e-03, -1.8620e-04, ..., -2.1970e-04, + -2.0163e-07, -4.0102e-04], + [ 1.6632e-02, 3.5038e-03, 6.7520e-04, ..., 3.0160e-04, + 1.5129e-02, 1.2421e-02]], device='cuda:0') +Epoch 142, bias, value: tensor([ 0.0158, -0.0093, -0.0004, -0.0174, 0.0125, -0.0026, -0.0171, -0.0242, + 0.0088, -0.0031], device='cuda:0'), grad: tensor([-0.0019, 0.0060, 0.0284, 0.0223, -0.0834, -0.0112, 0.0200, -0.0578, + -0.0051, 0.0826], device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 216.78, cls_loss 0.5478 cls_loss_mapping 0.0131 cls_loss_causal 0.5163 re_mapping 0.0091 re_causal 0.0235 /// teacc 98.46 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 0.0163, 0.0727, -0.1262, ..., -0.0694, 0.0536, 0.0105], + [-0.0630, -0.0952, 0.0638, ..., -0.0303, -0.0905, 0.0491], + [-0.0566, -0.0321, -0.0847, ..., -0.1021, -0.0281, 0.0369], + ..., + [-0.0175, -0.0654, -0.0781, ..., -0.0921, -0.0102, 0.0539], + [-0.0898, -0.0773, -0.0769, ..., -0.0467, -0.0459, -0.0738], + [ 0.0099, -0.0146, -0.0552, ..., -0.0501, 0.0273, -0.0147]], + device='cuda:0'), grad: tensor([[-0.0100, -0.0012, 0.0002, ..., 0.0004, -0.0087, -0.0043], + [-0.0025, -0.0006, -0.0012, ..., -0.0018, -0.0029, -0.0044], + [ 0.0015, 0.0019, 0.0003, ..., 0.0012, 0.0015, 0.0014], + ..., + [ 0.0022, 0.0008, 0.0009, ..., 0.0006, 0.0022, 0.0018], + [ 0.0010, 0.0032, 0.0003, ..., 0.0020, 0.0011, 0.0007], + [ 0.0010, 0.0002, 0.0001, ..., 0.0002, 0.0009, 0.0008]], + device='cuda:0') +Epoch 143, bias, value: tensor([ 0.0160, -0.0091, -0.0003, -0.0180, 0.0126, -0.0027, -0.0174, -0.0233, + 0.0084, -0.0032], device='cuda:0'), grad: tensor([-0.0229, -0.0508, 0.0126, 0.0058, 0.0109, 0.0058, 0.0061, 0.0159, + 0.0095, 0.0069], device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 217.87, cls_loss 0.5517 cls_loss_mapping 0.0130 cls_loss_causal 0.5177 re_mapping 0.0098 re_causal 0.0253 /// teacc 98.70 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0160, 0.0723, -0.1258, ..., -0.0702, 0.0527, 0.0111], + [-0.0627, -0.0937, 0.0639, ..., -0.0296, -0.0897, 0.0495], + [-0.0567, -0.0325, -0.0857, ..., -0.1018, -0.0277, 0.0369], + ..., + [-0.0184, -0.0655, -0.0786, ..., -0.0934, -0.0110, 0.0527], + [-0.0887, -0.0782, -0.0765, ..., -0.0467, -0.0449, -0.0737], + [ 0.0107, -0.0142, -0.0551, ..., -0.0503, 0.0276, -0.0142]], + device='cuda:0'), grad: tensor([[ 1.2417e-03, 2.3711e-04, 1.5765e-05, ..., 2.0254e-04, + 1.1282e-03, 1.0891e-03], + [ 1.0157e-03, 2.0540e-04, -2.1338e-04, ..., 5.5671e-05, + 1.0796e-03, 7.5293e-04], + [ 7.9966e-04, 2.3139e-04, 4.1604e-05, ..., 5.8264e-05, + 7.6199e-04, 2.0294e-03], + ..., + [-5.1785e-04, 5.4216e-04, 1.1843e-04, ..., 7.4506e-05, + -2.0599e-03, -2.0618e-03], + [ 1.4849e-03, 4.4155e-04, 1.1331e-04, ..., 2.4295e-04, + 1.0452e-03, 8.5258e-04], + [ 1.1702e-03, 1.0443e-03, -4.1902e-05, ..., 1.4782e-04, + 2.1267e-03, 1.8587e-03]], device='cuda:0') +Epoch 144, bias, value: tensor([ 0.0162, -0.0082, -0.0006, -0.0180, 0.0128, -0.0024, -0.0183, -0.0242, + 0.0087, -0.0030], device='cuda:0'), grad: tensor([ 0.0095, 0.0084, 0.0087, 0.0105, -0.0160, -0.0312, 0.0068, -0.0211, + 0.0098, 0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 217.54, cls_loss 0.5662 cls_loss_mapping 0.0135 cls_loss_causal 0.5280 re_mapping 0.0090 re_causal 0.0233 /// teacc 98.61 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0153, 0.0738, -0.1258, ..., -0.0706, 0.0529, 0.0096], + [-0.0629, -0.0932, 0.0648, ..., -0.0276, -0.0906, 0.0495], + [-0.0569, -0.0326, -0.0859, ..., -0.1026, -0.0272, 0.0375], + ..., + [-0.0184, -0.0673, -0.0798, ..., -0.0953, -0.0125, 0.0531], + [-0.0895, -0.0786, -0.0783, ..., -0.0470, -0.0445, -0.0732], + [ 0.0117, -0.0145, -0.0552, ..., -0.0492, 0.0280, -0.0142]], + device='cuda:0'), grad: tensor([[-1.1463e-03, -9.5367e-03, -1.1902e-03, ..., -7.0477e-04, + -3.9029e-04, 6.9189e-04], + [-8.1539e-05, -1.6594e-03, -6.3944e-04, ..., -2.2907e-03, + 2.3003e-03, 9.0218e-04], + [ 5.7526e-03, 9.1934e-04, 2.0862e-04, ..., 8.7214e-04, + 7.2708e-03, 3.4428e-03], + ..., + [ 3.6278e-03, 6.2847e-04, 2.2328e-04, ..., 6.0797e-04, + 3.5782e-03, 1.2655e-03], + [ 7.2956e-05, 1.0712e-02, 6.8703e-03, ..., 2.0390e-03, + 1.7121e-05, -2.2335e-03], + [-3.4389e-03, 1.3437e-03, 1.3041e-04, ..., -5.7697e-04, + -3.9101e-03, -1.3387e-04]], device='cuda:0') +Epoch 145, bias, value: tensor([ 0.0150, -0.0083, -0.0010, -0.0182, 0.0125, -0.0022, -0.0181, -0.0245, + 0.0090, -0.0015], device='cuda:0'), grad: tensor([-0.0029, -0.0099, 0.0477, -0.0151, -0.0213, -0.0577, 0.0270, 0.0287, + 0.0216, -0.0181], device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 219.45, cls_loss 0.5503 cls_loss_mapping 0.0139 cls_loss_causal 0.5195 re_mapping 0.0093 re_causal 0.0243 /// teacc 98.62 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0156, 0.0735, -0.1269, ..., -0.0715, 0.0530, 0.0086], + [-0.0635, -0.0930, 0.0656, ..., -0.0284, -0.0904, 0.0488], + [-0.0576, -0.0324, -0.0870, ..., -0.1040, -0.0271, 0.0379], + ..., + [-0.0178, -0.0659, -0.0799, ..., -0.0948, -0.0120, 0.0528], + [-0.0900, -0.0791, -0.0779, ..., -0.0477, -0.0445, -0.0723], + [ 0.0125, -0.0138, -0.0545, ..., -0.0489, 0.0277, -0.0143]], + device='cuda:0'), grad: tensor([[ 2.4986e-04, -8.3590e-04, 7.5006e-04, ..., 3.9995e-05, + -5.1994e-03, 3.9315e-04], + [ 1.1120e-03, 7.2746e-03, 6.8893e-03, ..., 1.0780e-02, + 2.0275e-03, 2.4166e-03], + [ 2.1577e-04, 1.1247e-04, 5.2834e-04, ..., 4.2766e-05, + 5.7268e-04, 1.5121e-02], + ..., + [ 1.1854e-03, 5.0163e-04, 1.0214e-03, ..., 2.2531e-04, + 3.1338e-03, -1.4046e-02], + [-1.8919e-04, 3.7270e-03, 6.2103e-03, ..., 8.9979e-04, + 3.8261e-03, 4.5180e-04], + [ 2.7466e-04, 7.3767e-04, 9.2220e-04, ..., 2.2769e-04, + 2.6035e-03, 4.5347e-04]], device='cuda:0') +Epoch 146, bias, value: tensor([ 0.0148, -0.0090, -0.0011, -0.0176, 0.0113, -0.0021, -0.0157, -0.0247, + 0.0082, -0.0012], device='cuda:0'), grad: tensor([-0.0055, 0.0339, 0.0169, -0.0131, -0.0151, -0.0280, -0.0246, 0.0020, + 0.0230, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 217.41, cls_loss 0.5395 cls_loss_mapping 0.0093 cls_loss_causal 0.5110 re_mapping 0.0093 re_causal 0.0244 /// teacc 98.65 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 0.0158, 0.0733, -0.1289, ..., -0.0724, 0.0533, 0.0078], + [-0.0635, -0.0936, 0.0665, ..., -0.0280, -0.0904, 0.0496], + [-0.0579, -0.0331, -0.0879, ..., -0.1038, -0.0273, 0.0382], + ..., + [-0.0173, -0.0663, -0.0816, ..., -0.0947, -0.0118, 0.0530], + [-0.0909, -0.0793, -0.0787, ..., -0.0493, -0.0453, -0.0725], + [ 0.0120, -0.0143, -0.0539, ..., -0.0487, 0.0270, -0.0138]], + device='cuda:0'), grad: tensor([[ 2.1152e-03, 5.8556e-04, 4.5300e-04, ..., 6.9571e-04, + 1.3027e-03, 1.8463e-03], + [-1.5068e-03, 2.2471e-04, -2.6131e-03, ..., 3.7575e-04, + -6.0701e-04, 7.9422e-03], + [ 3.0823e-03, -9.4175e-04, -2.2106e-03, ..., 1.0424e-03, + 2.6531e-03, 2.6073e-03], + ..., + [ 1.0462e-03, 2.1350e-04, 8.4209e-04, ..., 3.7265e-04, + 7.5102e-04, -1.1574e-02], + [ 7.8154e-04, 2.0087e-04, 6.0177e-04, ..., 3.0756e-04, + -5.0545e-03, -6.9916e-05], + [-5.3167e-04, 1.1730e-04, -2.0969e-04, ..., -6.7139e-04, + 1.3542e-03, 1.6565e-03]], device='cuda:0') +Epoch 147, bias, value: tensor([ 0.0142, -0.0087, -0.0009, -0.0179, 0.0104, -0.0019, -0.0153, -0.0240, + 0.0074, -0.0005], device='cuda:0'), grad: tensor([ 0.0173, 0.0092, 0.0176, 0.0249, -0.0462, -0.0152, 0.0244, -0.0188, + -0.0123, -0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 219.15, cls_loss 0.5510 cls_loss_mapping 0.0098 cls_loss_causal 0.5193 re_mapping 0.0092 re_causal 0.0236 /// teacc 98.63 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0163, 0.0738, -0.1286, ..., -0.0717, 0.0534, 0.0069], + [-0.0642, -0.0924, 0.0681, ..., -0.0264, -0.0912, 0.0494], + [-0.0574, -0.0332, -0.0890, ..., -0.1040, -0.0267, 0.0392], + ..., + [-0.0181, -0.0668, -0.0823, ..., -0.0936, -0.0122, 0.0532], + [-0.0904, -0.0797, -0.0786, ..., -0.0498, -0.0448, -0.0720], + [ 0.0118, -0.0147, -0.0534, ..., -0.0497, 0.0264, -0.0135]], + device='cuda:0'), grad: tensor([[ 1.8816e-03, 1.8370e-04, -1.3196e-04, ..., 2.7866e-03, + 6.7091e-04, 1.0757e-03], + [ 1.5335e-03, 1.2763e-05, 4.8351e-04, ..., 6.2418e-04, + 6.6328e-04, 1.8110e-03], + [ 1.5049e-03, 4.6802e-04, 4.9973e-04, ..., 1.4629e-03, + 7.7438e-04, 5.1460e-03], + ..., + [-1.8835e-03, 2.3991e-05, -1.8287e-04, ..., -4.1318e-04, + -8.0395e-04, -3.2520e-04], + [ 5.4741e-04, -1.0061e-03, -4.6659e-04, ..., 1.0071e-03, + 2.0659e-04, -4.7302e-03], + [ 2.8172e-03, 9.4235e-05, 7.1239e-04, ..., 9.0790e-04, + 1.0872e-03, 1.4410e-03]], device='cuda:0') +Epoch 148, bias, value: tensor([ 0.0145, -0.0088, -0.0003, -0.0185, 0.0100, -0.0018, -0.0154, -0.0250, + 0.0081, 0.0001], device='cuda:0'), grad: tensor([ 0.0099, 0.0117, 0.0222, -0.0196, -0.0064, 0.0099, -0.0306, -0.0184, + 0.0026, 0.0186], device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 218.16, cls_loss 0.5595 cls_loss_mapping 0.0123 cls_loss_causal 0.5348 re_mapping 0.0091 re_causal 0.0237 /// teacc 98.64 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0167, 0.0743, -0.1293, ..., -0.0708, 0.0534, 0.0066], + [-0.0656, -0.0931, 0.0680, ..., -0.0264, -0.0911, 0.0491], + [-0.0568, -0.0335, -0.0881, ..., -0.1041, -0.0268, 0.0390], + ..., + [-0.0182, -0.0674, -0.0829, ..., -0.0938, -0.0120, 0.0536], + [-0.0903, -0.0786, -0.0777, ..., -0.0493, -0.0457, -0.0716], + [ 0.0114, -0.0146, -0.0544, ..., -0.0505, 0.0260, -0.0149]], + device='cuda:0'), grad: tensor([[ 0.0022, 0.0016, 0.0008, ..., 0.0020, -0.0014, 0.0017], + [-0.0060, -0.0005, -0.0004, ..., -0.0005, -0.0064, -0.0049], + [-0.0033, -0.0010, -0.0007, ..., -0.0012, -0.0014, -0.0026], + ..., + [ 0.0020, 0.0008, 0.0005, ..., 0.0007, 0.0016, 0.0009], + [ 0.0028, 0.0017, 0.0009, ..., 0.0018, 0.0021, 0.0014], + [-0.0015, -0.0017, 0.0002, ..., -0.0045, 0.0004, -0.0007]], + device='cuda:0') +Epoch 149, bias, value: tensor([ 0.0148, -0.0093, -0.0004, -0.0182, 0.0112, -0.0022, -0.0158, -0.0250, + 0.0084, -0.0007], device='cuda:0'), grad: tensor([ 0.0243, -0.0434, -0.0363, -0.0088, 0.0092, 0.0203, 0.0151, 0.0189, + 0.0267, -0.0259], device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 218.28, cls_loss 0.5444 cls_loss_mapping 0.0112 cls_loss_causal 0.5083 re_mapping 0.0091 re_causal 0.0233 /// teacc 98.61 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0167, 0.0741, -0.1290, ..., -0.0718, 0.0541, 0.0061], + [-0.0652, -0.0915, 0.0692, ..., -0.0259, -0.0917, 0.0497], + [-0.0566, -0.0347, -0.0902, ..., -0.1038, -0.0267, 0.0392], + ..., + [-0.0178, -0.0674, -0.0820, ..., -0.0926, -0.0108, 0.0531], + [-0.0909, -0.0776, -0.0776, ..., -0.0490, -0.0466, -0.0709], + [ 0.0121, -0.0136, -0.0558, ..., -0.0511, 0.0264, -0.0154]], + device='cuda:0'), grad: tensor([[-1.7357e-03, -1.6022e-03, -1.2932e-03, ..., -6.4373e-05, + -1.5316e-03, -2.1667e-03], + [ 2.7065e-03, 4.4298e-04, 2.1005e-04, ..., 5.0211e-04, + 1.1778e-03, -2.0361e-04], + [ 3.4504e-03, 2.4109e-03, 1.3647e-03, ..., 9.5892e-04, + 3.2043e-03, 3.2673e-03], + ..., + [ 1.1272e-03, -1.7071e-03, 8.3447e-04, ..., -1.8024e-03, + 4.1723e-04, -8.2016e-04], + [-2.3994e-03, 1.6699e-03, 5.4264e-04, ..., 1.6670e-03, + -3.7980e-04, -4.1351e-03], + [-1.3107e-02, -9.3937e-04, -4.5128e-03, ..., -1.4172e-03, + -1.0193e-02, -2.7090e-05]], device='cuda:0') +Epoch 150, bias, value: tensor([ 0.0145, -0.0078, -0.0007, -0.0188, 0.0102, -0.0023, -0.0157, -0.0242, + 0.0088, -0.0011], device='cuda:0'), grad: tensor([-0.0050, 0.0171, 0.0369, 0.0328, 0.0240, 0.0123, -0.0055, -0.0214, + -0.0227, -0.0685], device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 217.93, cls_loss 0.5573 cls_loss_mapping 0.0124 cls_loss_causal 0.5282 re_mapping 0.0090 re_causal 0.0233 /// teacc 98.71 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0161, 0.0750, -0.1306, ..., -0.0720, 0.0539, 0.0052], + [-0.0663, -0.0917, 0.0705, ..., -0.0253, -0.0925, 0.0494], + [-0.0565, -0.0357, -0.0915, ..., -0.1040, -0.0270, 0.0390], + ..., + [-0.0161, -0.0681, -0.0811, ..., -0.0930, -0.0103, 0.0540], + [-0.0913, -0.0795, -0.0791, ..., -0.0508, -0.0469, -0.0708], + [ 0.0124, -0.0129, -0.0549, ..., -0.0522, 0.0267, -0.0135]], + device='cuda:0'), grad: tensor([[-1.1765e-02, -1.2465e-03, -6.8474e-03, ..., -2.0466e-03, + -5.5161e-03, -4.2763e-03], + [ 1.1272e-03, -1.5914e-05, 1.1969e-03, ..., -5.5161e-03, + 6.6948e-04, -2.8744e-03], + [-1.4172e-03, -3.4779e-05, -2.5973e-05, ..., 7.6580e-04, + 3.3188e-04, -2.5024e-03], + ..., + [-1.1168e-03, 5.8317e-04, -5.1804e-03, ..., -8.4972e-04, + 3.7813e-04, -2.1248e-03], + [ 3.3875e-03, 7.3767e-04, 2.5864e-03, ..., 1.9436e-03, + 1.2188e-03, 4.7874e-03], + [ 3.6602e-03, 5.5313e-04, 1.7252e-03, ..., 1.4486e-03, + 1.0862e-03, 3.3684e-03]], device='cuda:0') +Epoch 151, bias, value: tensor([ 0.0141, -0.0077, -0.0001, -0.0191, 0.0099, -0.0026, -0.0151, -0.0241, + 0.0081, -0.0007], device='cuda:0'), grad: tensor([-0.0757, -0.0203, -0.0037, -0.0120, 0.0022, 0.0259, 0.0418, -0.0055, + 0.0334, 0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 217.53, cls_loss 0.5416 cls_loss_mapping 0.0096 cls_loss_causal 0.5114 re_mapping 0.0097 re_causal 0.0254 /// teacc 98.57 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0171, 0.0751, -0.1301, ..., -0.0719, 0.0539, 0.0050], + [-0.0656, -0.0901, 0.0721, ..., -0.0249, -0.0917, 0.0500], + [-0.0562, -0.0346, -0.0911, ..., -0.1042, -0.0254, 0.0393], + ..., + [-0.0170, -0.0687, -0.0815, ..., -0.0929, -0.0121, 0.0534], + [-0.0913, -0.0797, -0.0798, ..., -0.0507, -0.0466, -0.0706], + [ 0.0121, -0.0136, -0.0556, ..., -0.0533, 0.0269, -0.0138]], + device='cuda:0'), grad: tensor([[-5.2986e-03, -3.2120e-03, 2.0564e-04, ..., 2.8133e-04, + -2.8019e-03, -5.4321e-03], + [-1.5211e-03, 9.2983e-05, 1.3227e-03, ..., 3.5787e-04, + -2.7199e-03, -6.2132e-04], + [ 1.6232e-03, 7.2336e-04, -5.7564e-03, ..., -6.4774e-03, + 2.6588e-03, 2.1477e-03], + ..., + [-8.1587e-04, 9.1136e-05, -4.4167e-05, ..., 1.5402e-04, + -1.3542e-04, -2.5520e-03], + [ 1.7290e-03, 4.1175e-04, 1.0796e-03, ..., 1.2722e-03, + 1.2579e-03, 2.2125e-03], + [ 1.7633e-03, 2.1887e-04, 2.4045e-04, ..., 2.7633e-04, + 1.0614e-03, 2.2144e-03]], device='cuda:0') +Epoch 152, bias, value: tensor([ 0.0145, -0.0070, 0.0005, -0.0204, 0.0108, -0.0027, -0.0152, -0.0243, + 0.0074, -0.0008], device='cuda:0'), grad: tensor([-0.0422, -0.0019, 0.0044, -0.0170, -0.0064, 0.0174, 0.0219, -0.0151, + 0.0225, 0.0164], device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 218.63, cls_loss 0.5691 cls_loss_mapping 0.0112 cls_loss_causal 0.5426 re_mapping 0.0090 re_causal 0.0232 /// teacc 98.41 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 0.0171, 0.0743, -0.1308, ..., -0.0726, 0.0539, 0.0056], + [-0.0663, -0.0903, 0.0714, ..., -0.0244, -0.0899, 0.0499], + [-0.0572, -0.0348, -0.0911, ..., -0.1053, -0.0267, 0.0396], + ..., + [-0.0179, -0.0693, -0.0820, ..., -0.0932, -0.0125, 0.0535], + [-0.0910, -0.0794, -0.0785, ..., -0.0500, -0.0472, -0.0708], + [ 0.0124, -0.0136, -0.0553, ..., -0.0536, 0.0267, -0.0141]], + device='cuda:0'), grad: tensor([[-1.2693e-03, -8.4043e-06, -6.0415e-04, ..., -9.5272e-04, + -4.2295e-04, -9.1314e-04], + [ 3.1338e-03, 2.9945e-03, 2.3289e-03, ..., 1.7490e-03, + 2.3956e-03, 3.7174e-03], + [-3.1924e-04, 1.2465e-03, 7.8249e-04, ..., 6.7329e-04, + 7.5436e-04, 1.7385e-03], + ..., + [ 2.2736e-03, 3.8743e-04, 1.6270e-03, ..., 2.6417e-04, + 3.6240e-04, 2.3499e-03], + [ 1.7414e-03, 1.3733e-03, 1.2388e-03, ..., 7.5531e-04, + 1.3580e-03, 1.5659e-03], + [ 2.3384e-03, 1.2779e-03, 8.3113e-04, ..., 5.0545e-04, + 2.2373e-03, 2.4509e-03]], device='cuda:0') +Epoch 153, bias, value: tensor([ 0.0148, -0.0076, -0.0004, -0.0194, 0.0125, -0.0025, -0.0154, -0.0252, + 0.0071, -0.0009], device='cuda:0'), grad: tensor([-0.0095, 0.0527, -0.0063, -0.0701, 0.0298, -0.0340, -0.0220, 0.0308, + 0.0287, -0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 217.33, cls_loss 0.5607 cls_loss_mapping 0.0106 cls_loss_causal 0.5209 re_mapping 0.0090 re_causal 0.0228 /// teacc 98.40 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0167, 0.0745, -0.1309, ..., -0.0722, 0.0541, 0.0054], + [-0.0664, -0.0907, 0.0716, ..., -0.0250, -0.0897, 0.0500], + [-0.0573, -0.0351, -0.0913, ..., -0.1053, -0.0271, 0.0392], + ..., + [-0.0166, -0.0684, -0.0829, ..., -0.0933, -0.0118, 0.0534], + [-0.0909, -0.0792, -0.0782, ..., -0.0495, -0.0479, -0.0722], + [ 0.0113, -0.0144, -0.0546, ..., -0.0550, 0.0265, -0.0145]], + device='cuda:0'), grad: tensor([[ 5.5695e-03, -7.9155e-05, 4.1962e-04, ..., 4.9210e-04, + 2.0676e-03, -1.9894e-03], + [ 4.0474e-03, -2.0933e-04, -1.6928e-03, ..., 4.2653e-04, + 1.8654e-03, 4.1656e-03], + [-4.1656e-03, -4.4584e-04, -4.3750e-04, ..., -3.2973e-04, + -2.7790e-03, -1.1816e-03], + ..., + [-1.0414e-02, -3.7174e-03, -3.0766e-03, ..., -1.5144e-03, + -7.1793e-03, -7.7858e-03], + [-2.3766e-03, 6.7854e-04, 6.6090e-04, ..., 2.4056e-04, + 6.4659e-04, -9.4223e-03], + [-1.4629e-03, -2.3270e-04, -1.7262e-03, ..., -1.7529e-03, + -8.7738e-04, 6.7616e-04]], device='cuda:0') +Epoch 154, bias, value: tensor([ 1.5274e-02, -8.6273e-03, -7.5855e-05, -1.9340e-02, 1.2457e-02, + -1.7459e-03, -1.4180e-02, -2.5467e-02, 6.4565e-03, -1.6352e-03], + device='cuda:0'), grad: tensor([ 0.0112, -0.0090, -0.0222, 0.0331, 0.0549, 0.0388, 0.0120, -0.0452, + -0.0417, -0.0318], device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 219.40, cls_loss 0.5585 cls_loss_mapping 0.0099 cls_loss_causal 0.5266 re_mapping 0.0093 re_causal 0.0244 /// teacc 98.46 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 0.0176, 0.0744, -0.1322, ..., -0.0728, 0.0547, 0.0055], + [-0.0666, -0.0910, 0.0715, ..., -0.0254, -0.0901, 0.0502], + [-0.0584, -0.0348, -0.0911, ..., -0.1057, -0.0284, 0.0390], + ..., + [-0.0161, -0.0681, -0.0835, ..., -0.0940, -0.0119, 0.0532], + [-0.0905, -0.0800, -0.0783, ..., -0.0511, -0.0470, -0.0707], + [ 0.0112, -0.0151, -0.0538, ..., -0.0539, 0.0257, -0.0149]], + device='cuda:0'), grad: tensor([[ 0.0009, -0.0001, 0.0004, ..., 0.0006, 0.0001, 0.0011], + [ 0.0015, 0.0004, 0.0010, ..., 0.0005, 0.0010, 0.0021], + [-0.0015, 0.0007, -0.0028, ..., -0.0035, -0.0009, -0.0054], + ..., + [ 0.0034, 0.0006, 0.0003, ..., 0.0001, 0.0007, 0.0023], + [ 0.0002, 0.0004, -0.0020, ..., 0.0027, -0.0020, 0.0011], + [ 0.0021, 0.0005, 0.0001, ..., 0.0001, 0.0013, 0.0011]], + device='cuda:0') +Epoch 155, bias, value: tensor([ 0.0147, -0.0081, -0.0007, -0.0182, 0.0124, -0.0034, -0.0139, -0.0248, + 0.0077, -0.0028], device='cuda:0'), grad: tensor([ 0.0154, 0.0304, -0.0741, 0.0068, -0.0133, 0.0148, -0.0135, 0.0240, + -0.0056, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 215.98, cls_loss 0.5756 cls_loss_mapping 0.0099 cls_loss_causal 0.5454 re_mapping 0.0088 re_causal 0.0244 /// teacc 98.68 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 0.0186, 0.0752, -0.1319, ..., -0.0732, 0.0550, 0.0052], + [-0.0665, -0.0905, 0.0718, ..., -0.0249, -0.0904, 0.0499], + [-0.0589, -0.0350, -0.0916, ..., -0.1059, -0.0283, 0.0397], + ..., + [-0.0153, -0.0694, -0.0836, ..., -0.0925, -0.0116, 0.0531], + [-0.0920, -0.0816, -0.0785, ..., -0.0516, -0.0487, -0.0717], + [ 0.0111, -0.0146, -0.0544, ..., -0.0537, 0.0266, -0.0140]], + device='cuda:0'), grad: tensor([[ 8.7261e-04, -4.1753e-05, 8.1491e-07, ..., -1.3189e-03, + -6.5279e-04, 3.0117e-03], + [-1.6575e-03, 3.0905e-05, 1.5525e-06, ..., 1.6344e-04, + -1.7214e-03, -4.3068e-03], + [ 1.5011e-03, 1.8368e-03, 6.3539e-05, ..., 1.3504e-03, + 6.6710e-04, 4.9934e-03], + ..., + [-1.6556e-03, -9.8801e-04, 3.8967e-06, ..., 1.4806e-04, + 3.1109e-03, -3.9101e-03], + [ 8.6880e-04, 4.3035e-04, 2.3797e-05, ..., 2.5439e-04, + 6.2370e-04, 1.6928e-03], + [ 8.0729e-04, 2.6178e-04, 2.1663e-06, ..., 6.0797e-05, + -4.0817e-03, 1.5144e-03]], device='cuda:0') +Epoch 156, bias, value: tensor([ 0.0145, -0.0077, -0.0004, -0.0182, 0.0116, -0.0032, -0.0134, -0.0250, + 0.0074, -0.0027], device='cuda:0'), grad: tensor([ 0.0114, -0.0217, 0.0301, 0.0048, 0.0129, -0.0149, -0.0125, -0.0344, + 0.0157, 0.0085], device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 216.48, cls_loss 0.5468 cls_loss_mapping 0.0133 cls_loss_causal 0.5190 re_mapping 0.0087 re_causal 0.0234 /// teacc 98.77 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 0.0187, 0.0750, -0.1330, ..., -0.0734, 0.0542, 0.0061], + [-0.0659, -0.0904, 0.0722, ..., -0.0259, -0.0897, 0.0499], + [-0.0587, -0.0352, -0.0923, ..., -0.1044, -0.0287, 0.0404], + ..., + [-0.0153, -0.0703, -0.0834, ..., -0.0926, -0.0119, 0.0533], + [-0.0911, -0.0811, -0.0792, ..., -0.0518, -0.0481, -0.0725], + [ 0.0109, -0.0144, -0.0530, ..., -0.0532, 0.0262, -0.0138]], + device='cuda:0'), grad: tensor([[-3.2616e-04, 2.8343e-03, 1.7121e-05, ..., 8.5354e-04, + 8.4972e-04, -8.4639e-04], + [ 1.4029e-03, 2.4164e-04, 3.4392e-05, ..., 4.2558e-04, + 9.5129e-04, 3.1834e-03], + [ 1.3752e-03, 3.0556e-03, -2.3258e-04, ..., 3.1853e-04, + 1.6050e-03, 3.9101e-03], + ..., + [-9.0408e-03, 1.0319e-03, 7.0810e-05, ..., 2.4164e-04, + -2.6722e-03, 8.8644e-04], + [ 2.8629e-03, 1.2856e-03, 9.0361e-05, ..., 1.3351e-03, + 2.5940e-03, 1.0004e-03], + [-6.1607e-04, 1.5461e-04, 3.6925e-05, ..., 6.7091e-04, + -5.4026e-04, -3.9053e-04]], device='cuda:0') +Epoch 157, bias, value: tensor([ 0.0142, -0.0077, -0.0005, -0.0184, 0.0118, -0.0039, -0.0137, -0.0250, + 0.0082, -0.0022], device='cuda:0'), grad: tensor([-0.0051, 0.0247, 0.0401, -0.0036, -0.0264, 0.0167, -0.0254, -0.0172, + 0.0210, -0.0248], device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 216.09, cls_loss 0.5179 cls_loss_mapping 0.0095 cls_loss_causal 0.4936 re_mapping 0.0088 re_causal 0.0234 /// teacc 98.68 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 0.0187, 0.0753, -0.1328, ..., -0.0741, 0.0544, 0.0069], + [-0.0668, -0.0896, 0.0720, ..., -0.0265, -0.0891, 0.0495], + [-0.0576, -0.0356, -0.0911, ..., -0.1037, -0.0281, 0.0414], + ..., + [-0.0155, -0.0706, -0.0840, ..., -0.0931, -0.0120, 0.0532], + [-0.0915, -0.0819, -0.0796, ..., -0.0514, -0.0495, -0.0732], + [ 0.0113, -0.0133, -0.0529, ..., -0.0526, 0.0264, -0.0130]], + device='cuda:0'), grad: tensor([[-7.9870e-04, -1.1559e-03, 8.0615e-06, ..., -4.5586e-03, + 5.3215e-04, -3.0499e-03], + [ 1.4639e-03, 5.9605e-05, 1.3625e-06, ..., 1.8990e-04, + 8.8978e-04, 2.3365e-03], + [ 9.9564e-04, 8.2135e-05, 2.9773e-05, ..., 2.5916e-04, + 1.0319e-06, 1.2341e-03], + ..., + [-3.7785e-03, 2.1443e-05, 1.5542e-05, ..., 1.8203e-04, + -8.9073e-04, -4.3945e-03], + [ 1.4334e-03, 4.4775e-04, -1.4567e-04, ..., 1.3218e-03, + 5.6410e-04, 2.5291e-03], + [ 1.4515e-03, 1.3983e-04, 3.0756e-05, ..., 3.2210e-04, + 3.8838e-04, 3.8738e-03]], device='cuda:0') +Epoch 158, bias, value: tensor([ 0.0145, -0.0082, 0.0004, -0.0179, 0.0110, -0.0041, -0.0135, -0.0252, + 0.0079, -0.0021], device='cuda:0'), grad: tensor([-3.2501e-02, 1.7654e-02, 1.7059e-02, -5.4240e-05, -6.6528e-02, + 6.7253e-03, 2.7435e-02, -2.3972e-02, 2.2583e-02, 3.1586e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 217.33, cls_loss 0.5579 cls_loss_mapping 0.0126 cls_loss_causal 0.5294 re_mapping 0.0088 re_causal 0.0235 /// teacc 98.48 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 0.0192, 0.0747, -0.1340, ..., -0.0743, 0.0550, 0.0066], + [-0.0680, -0.0894, 0.0719, ..., -0.0264, -0.0898, 0.0502], + [-0.0577, -0.0365, -0.0912, ..., -0.1029, -0.0282, 0.0409], + ..., + [-0.0158, -0.0717, -0.0855, ..., -0.0939, -0.0129, 0.0535], + [-0.0903, -0.0820, -0.0801, ..., -0.0509, -0.0489, -0.0746], + [ 0.0113, -0.0124, -0.0522, ..., -0.0534, 0.0267, -0.0130]], + device='cuda:0'), grad: tensor([[ 8.3148e-06, 5.2643e-04, -2.0385e-04, ..., 3.0175e-06, + 8.1897e-05, -1.2589e-03], + [ 2.3594e-03, -3.4828e-03, -1.4404e-02, ..., -5.2299e-03, + 7.5674e-04, 1.4372e-03], + [-2.1801e-03, 4.8971e-04, 4.8757e-04, ..., 2.5249e-04, + -1.3046e-03, -3.5114e-03], + ..., + [ 7.7188e-05, -4.7708e-04, -1.8746e-05, ..., 5.8115e-05, + 3.8028e-04, -1.0548e-03], + [ 2.5845e-03, -3.7599e-04, 6.5384e-03, ..., 2.3270e-03, + 8.5497e-04, 2.0370e-03], + [-1.2035e-03, -9.1124e-04, 2.0850e-04, ..., 1.6201e-04, + -8.4257e-04, -3.6049e-04]], device='cuda:0') +Epoch 159, bias, value: tensor([ 1.4270e-02, -7.8300e-03, 2.7432e-06, -1.8102e-02, 1.1726e-02, + -4.4926e-03, -1.4041e-02, -2.5257e-02, 8.4514e-03, -2.0198e-03], + device='cuda:0'), grad: tensor([-0.0091, 0.0015, -0.0437, -0.0225, 0.0227, 0.0337, 0.0144, -0.0237, + 0.0195, 0.0072], device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 218.69, cls_loss 0.5591 cls_loss_mapping 0.0110 cls_loss_causal 0.5317 re_mapping 0.0091 re_causal 0.0241 /// teacc 98.44 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0193, 0.0763, -0.1349, ..., -0.0729, 0.0556, 0.0053], + [-0.0681, -0.0882, 0.0726, ..., -0.0259, -0.0886, 0.0495], + [-0.0586, -0.0375, -0.0911, ..., -0.1030, -0.0290, 0.0411], + ..., + [-0.0166, -0.0730, -0.0858, ..., -0.0941, -0.0134, 0.0537], + [-0.0901, -0.0825, -0.0809, ..., -0.0506, -0.0489, -0.0748], + [ 0.0114, -0.0128, -0.0532, ..., -0.0541, 0.0265, -0.0127]], + device='cuda:0'), grad: tensor([[ 1.2312e-03, -5.6863e-05, 4.7684e-05, ..., 6.7806e-04, + 9.4604e-04, 1.2274e-03], + [ 1.1263e-03, -5.4359e-05, -1.1225e-03, ..., 1.7977e-04, + 5.6458e-04, 1.5936e-03], + [ 1.0471e-03, 3.1114e-05, 6.4969e-05, ..., 2.2042e-04, + 5.4646e-04, 7.0333e-04], + ..., + [-7.9498e-03, 3.1799e-05, 7.8976e-05, ..., 1.2130e-04, + -4.0512e-03, -2.2621e-03], + [ 2.7599e-03, 2.0790e-04, 9.2268e-04, ..., 6.9952e-04, + 7.6294e-04, 1.2360e-03], + [ 4.1168e-02, 2.6774e-04, 3.3951e-04, ..., 4.2796e-04, + 6.3744e-03, 5.8174e-03]], device='cuda:0') +Epoch 160, bias, value: tensor([ 0.0141, -0.0075, -0.0004, -0.0191, 0.0126, -0.0051, -0.0131, -0.0249, + 0.0082, -0.0022], device='cuda:0'), grad: tensor([ 0.0119, 0.0091, 0.0101, -0.0225, -0.0235, -0.0172, -0.0437, -0.0019, + 0.0144, 0.0633], device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 217.13, cls_loss 0.5141 cls_loss_mapping 0.0104 cls_loss_causal 0.4827 re_mapping 0.0095 re_causal 0.0236 /// teacc 98.54 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0197, 0.0774, -0.1345, ..., -0.0719, 0.0560, 0.0045], + [-0.0685, -0.0881, 0.0746, ..., -0.0261, -0.0888, 0.0486], + [-0.0594, -0.0382, -0.0921, ..., -0.1033, -0.0296, 0.0410], + ..., + [-0.0162, -0.0737, -0.0872, ..., -0.0947, -0.0125, 0.0549], + [-0.0904, -0.0832, -0.0817, ..., -0.0505, -0.0488, -0.0748], + [ 0.0111, -0.0127, -0.0539, ..., -0.0542, 0.0272, -0.0117]], + device='cuda:0'), grad: tensor([[ 1.0767e-03, 1.1683e-05, 5.1022e-05, ..., 4.5389e-05, + 1.1148e-03, 2.1610e-03], + [ 1.0386e-03, -2.2399e-04, -3.9721e-04, ..., -1.0127e-04, + 1.1644e-03, 1.9989e-03], + [-7.0333e-04, -3.7849e-05, 1.0338e-03, ..., 1.0223e-03, + 1.1234e-03, 3.3226e-03], + ..., + [ 1.1454e-03, 2.0355e-05, 6.6280e-05, ..., 4.3422e-05, + 1.1415e-03, 1.9855e-03], + [-5.9547e-03, 8.4698e-05, -5.8365e-03, ..., -5.9052e-03, + 9.7179e-04, -1.3056e-03], + [ 8.2731e-04, 5.3734e-05, 1.5807e-04, ..., 9.3520e-05, + 8.0776e-04, 1.7204e-03]], device='cuda:0') +Epoch 161, bias, value: tensor([ 0.0136, -0.0068, -0.0003, -0.0185, 0.0126, -0.0047, -0.0136, -0.0250, + 0.0078, -0.0022], device='cuda:0'), grad: tensor([ 0.0119, 0.0117, 0.0031, -0.0363, -0.0231, 0.0269, 0.0157, 0.0124, + -0.0320, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 216.13, cls_loss 0.5441 cls_loss_mapping 0.0086 cls_loss_causal 0.5086 re_mapping 0.0092 re_causal 0.0248 /// teacc 98.70 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0189, 0.0773, -0.1352, ..., -0.0730, 0.0551, 0.0034], + [-0.0673, -0.0888, 0.0737, ..., -0.0269, -0.0879, 0.0491], + [-0.0597, -0.0394, -0.0926, ..., -0.1036, -0.0289, 0.0429], + ..., + [-0.0160, -0.0741, -0.0860, ..., -0.0950, -0.0130, 0.0532], + [-0.0909, -0.0841, -0.0822, ..., -0.0505, -0.0491, -0.0747], + [ 0.0113, -0.0125, -0.0531, ..., -0.0537, 0.0274, -0.0108]], + device='cuda:0'), grad: tensor([[-2.0676e-03, -2.6169e-03, 2.9862e-05, ..., -7.7248e-03, + -4.5466e-04, -1.8573e-04], + [-2.2106e-03, 2.1601e-04, -6.2943e-05, ..., 3.0661e-04, + -2.3663e-04, -1.7118e-04], + [ 9.2888e-04, 3.7098e-04, 8.7917e-05, ..., 2.9778e-04, + -2.4462e-04, 2.7161e-03], + ..., + [ 5.1880e-04, 3.5477e-04, 7.2956e-05, ..., 1.4985e-04, + -3.6359e-04, -7.1335e-03], + [-5.1498e-03, -4.2191e-03, -4.5967e-04, ..., -1.3661e-04, + -3.6087e-03, -1.9145e-04], + [ 2.5845e-03, 1.5631e-03, 1.8537e-04, ..., 8.2445e-04, + 1.8148e-03, 2.3689e-03]], device='cuda:0') +Epoch 162, bias, value: tensor([ 0.0126, -0.0065, 0.0003, -0.0184, 0.0129, -0.0047, -0.0139, -0.0249, + 0.0076, -0.0024], device='cuda:0'), grad: tensor([-0.0351, -0.0072, 0.0440, -0.0040, -0.0358, 0.0146, 0.0308, -0.0238, + -0.0116, 0.0281], device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 216.02, cls_loss 0.5431 cls_loss_mapping 0.0093 cls_loss_causal 0.5133 re_mapping 0.0095 re_causal 0.0240 /// teacc 98.54 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0199, 0.0768, -0.1369, ..., -0.0747, 0.0550, 0.0032], + [-0.0687, -0.0882, 0.0747, ..., -0.0265, -0.0889, 0.0489], + [-0.0601, -0.0397, -0.0933, ..., -0.1042, -0.0288, 0.0422], + ..., + [-0.0181, -0.0747, -0.0860, ..., -0.0962, -0.0135, 0.0535], + [-0.0901, -0.0834, -0.0824, ..., -0.0518, -0.0493, -0.0748], + [ 0.0122, -0.0127, -0.0542, ..., -0.0538, 0.0286, -0.0115]], + device='cuda:0'), grad: tensor([[-1.7996e-03, -5.5313e-04, 1.5771e-04, ..., -4.5371e-04, + -2.5578e-03, -5.9700e-04], + [ 8.5115e-04, 1.7643e-04, -1.0643e-03, ..., 2.9945e-03, + 5.9128e-04, 1.2350e-03], + [-2.5249e-04, 7.3862e-04, 3.1495e-04, ..., 2.2030e-03, + -2.3401e-04, -2.6989e-04], + ..., + [ 1.0891e-03, 1.4853e-04, 4.2230e-05, ..., 1.5998e-04, + 8.0252e-04, -7.0667e-04], + [ 1.1892e-03, -1.6916e-04, 1.4296e-03, ..., -8.1873e-04, + 3.8385e-04, -4.7708e-04], + [-3.3188e-04, -1.2004e-04, -2.0962e-03, ..., 4.2391e-04, + 1.0633e-03, 3.4904e-03]], device='cuda:0') +Epoch 163, bias, value: tensor([ 0.0128, -0.0077, -0.0004, -0.0177, 0.0133, -0.0046, -0.0133, -0.0252, + 0.0075, -0.0020], device='cuda:0'), grad: tensor([-0.0226, 0.0237, 0.0051, 0.0002, -0.0529, -0.0089, 0.0280, 0.0196, + -0.0163, 0.0241], device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 216.08, cls_loss 0.5288 cls_loss_mapping 0.0104 cls_loss_causal 0.5022 re_mapping 0.0094 re_causal 0.0245 /// teacc 98.66 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0195, 0.0772, -0.1375, ..., -0.0747, 0.0555, 0.0019], + [-0.0692, -0.0880, 0.0752, ..., -0.0275, -0.0896, 0.0491], + [-0.0597, -0.0401, -0.0932, ..., -0.1053, -0.0281, 0.0416], + ..., + [-0.0181, -0.0737, -0.0868, ..., -0.0965, -0.0123, 0.0535], + [-0.0900, -0.0830, -0.0819, ..., -0.0516, -0.0493, -0.0748], + [ 0.0126, -0.0127, -0.0543, ..., -0.0549, 0.0282, -0.0110]], + device='cuda:0'), grad: tensor([[-1.5011e-03, -7.0839e-03, -1.9252e-05, ..., -5.0316e-03, + 3.4261e-04, 6.6328e-04], + [ 1.3809e-03, 2.8849e-04, 1.5032e-04, ..., 2.1935e-03, + 2.9373e-04, 1.1206e-03], + [ 6.6698e-05, 2.6202e-04, 2.0790e-04, ..., 9.8038e-04, + -5.9557e-04, -2.7943e-03], + ..., + [ 7.5293e-04, 1.8859e-04, 1.6439e-04, ..., 3.7503e-04, + 3.4523e-04, 1.0319e-03], + [ 1.2569e-03, 3.4833e-04, 3.3808e-04, ..., 1.4067e-03, + 2.9969e-04, 9.5272e-04], + [ 1.0929e-03, 4.7445e-04, 3.9816e-04, ..., 8.7214e-04, + 7.9489e-04, 8.2970e-04]], device='cuda:0') +Epoch 164, bias, value: tensor([ 0.0125, -0.0074, -0.0008, -0.0182, 0.0126, -0.0048, -0.0125, -0.0251, + 0.0081, -0.0018], device='cuda:0'), grad: tensor([-0.0052, 0.0135, -0.0160, -0.0189, -0.0037, 0.0093, -0.0170, 0.0126, + 0.0132, 0.0121], device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 216.10, cls_loss 0.5435 cls_loss_mapping 0.0094 cls_loss_causal 0.5169 re_mapping 0.0091 re_causal 0.0240 /// teacc 98.80 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0190, 0.0777, -0.1375, ..., -0.0738, 0.0562, 0.0015], + [-0.0692, -0.0875, 0.0742, ..., -0.0286, -0.0903, 0.0486], + [-0.0593, -0.0404, -0.0931, ..., -0.1056, -0.0272, 0.0411], + ..., + [-0.0188, -0.0747, -0.0865, ..., -0.0962, -0.0133, 0.0545], + [-0.0904, -0.0832, -0.0805, ..., -0.0505, -0.0483, -0.0743], + [ 0.0132, -0.0118, -0.0530, ..., -0.0552, 0.0278, -0.0117]], + device='cuda:0'), grad: tensor([[ 1.4544e-03, -3.2723e-05, 4.1395e-05, ..., 3.9250e-05, + 8.0824e-04, 4.3221e-03], + [ 2.1954e-03, 9.4250e-06, 1.0815e-03, ..., 5.4312e-04, + 1.7061e-03, -3.2544e-04], + [ 1.6279e-03, -7.2457e-06, 2.8920e-04, ..., 4.6939e-06, + 1.2922e-03, 2.4738e-03], + ..., + [-4.1890e-04, -7.9489e-04, -4.0340e-04, ..., -1.1730e-03, + -6.1836e-03, -4.2496e-03], + [-1.0967e-03, -4.3144e-03, -9.0256e-03, ..., -3.0651e-03, + -2.4853e-03, -3.5267e-03], + [-5.4131e-03, 2.9057e-05, 3.5167e-04, ..., 1.7989e-04, + -1.2188e-03, -3.9864e-03]], device='cuda:0') +Epoch 165, bias, value: tensor([ 0.0128, -0.0073, -0.0002, -0.0192, 0.0123, -0.0036, -0.0127, -0.0263, + 0.0083, -0.0013], device='cuda:0'), grad: tensor([ 0.0282, 0.0144, 0.0225, 0.0555, 0.0297, -0.0136, -0.0394, -0.0102, + -0.0467, -0.0405], device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 216.23, cls_loss 0.5310 cls_loss_mapping 0.0131 cls_loss_causal 0.5044 re_mapping 0.0092 re_causal 0.0235 /// teacc 98.59 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0187, 0.0779, -0.1365, ..., -0.0732, 0.0549, 0.0026], + [-0.0692, -0.0865, 0.0747, ..., -0.0293, -0.0899, 0.0477], + [-0.0593, -0.0411, -0.0930, ..., -0.1069, -0.0284, 0.0408], + ..., + [-0.0200, -0.0756, -0.0878, ..., -0.0977, -0.0136, 0.0544], + [-0.0914, -0.0837, -0.0814, ..., -0.0504, -0.0481, -0.0751], + [ 0.0134, -0.0127, -0.0531, ..., -0.0561, 0.0273, -0.0119]], + device='cuda:0'), grad: tensor([[-1.8797e-03, 1.1740e-03, 3.7980e-04, ..., 2.0180e-03, + -3.1681e-03, 5.3644e-04], + [-1.0805e-03, -1.0195e-03, 1.3694e-05, ..., -5.8794e-04, + -7.8726e-04, -2.4700e-04], + [-2.2526e-03, -8.4457e-03, -1.1816e-03, ..., -5.6877e-03, + 3.4285e-04, 1.1833e-02], + ..., + [ 7.2002e-04, 1.1998e-04, -2.1820e-03, ..., -5.3072e-04, + 3.4189e-04, -1.2764e-02], + [ 2.7490e-04, 1.9875e-03, 7.3051e-04, ..., 1.7052e-03, + -1.7786e-04, -6.4087e-04], + [ 2.0111e-04, 1.2910e-04, 3.4273e-05, ..., 6.0797e-05, + 3.0518e-04, 2.6059e-04]], device='cuda:0') +Epoch 166, bias, value: tensor([ 0.0139, -0.0068, -0.0009, -0.0183, 0.0123, -0.0034, -0.0137, -0.0272, + 0.0087, -0.0019], device='cuda:0'), grad: tensor([-0.0079, -0.0076, 0.0069, -0.0363, 0.0195, 0.0256, 0.0115, -0.0107, + -0.0107, 0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 216.11, cls_loss 0.5548 cls_loss_mapping 0.0138 cls_loss_causal 0.5340 re_mapping 0.0094 re_causal 0.0232 /// teacc 98.74 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0189, 0.0776, -0.1379, ..., -0.0743, 0.0550, 0.0023], + [-0.0701, -0.0860, 0.0756, ..., -0.0281, -0.0900, 0.0479], + [-0.0586, -0.0403, -0.0932, ..., -0.1060, -0.0277, 0.0414], + ..., + [-0.0193, -0.0771, -0.0876, ..., -0.0988, -0.0140, 0.0542], + [-0.0908, -0.0846, -0.0816, ..., -0.0505, -0.0494, -0.0748], + [ 0.0130, -0.0121, -0.0529, ..., -0.0550, 0.0285, -0.0123]], + device='cuda:0'), grad: tensor([[ 7.1859e-04, 4.7255e-04, -4.2856e-05, ..., 6.6519e-04, + 6.0177e-04, 9.1219e-04], + [-1.8969e-03, -1.0900e-03, 5.4687e-05, ..., 4.5443e-04, + -1.3781e-03, -2.3899e-03], + [ 1.4277e-03, -1.8132e-04, 8.7142e-05, ..., -4.7231e-04, + 1.2970e-03, -2.9388e-02], + ..., + [ 1.0681e-03, 7.2575e-04, 1.1444e-04, ..., 3.8576e-04, + 8.8024e-04, 3.0380e-02], + [ 1.5917e-03, 1.6155e-03, 8.5771e-05, ..., 2.4962e-04, + 1.3790e-03, 2.0142e-03], + [ 1.2636e-03, 1.2894e-03, -1.6602e-02, ..., 6.5708e-04, + 1.3905e-03, 1.6165e-03]], device='cuda:0') +Epoch 167, bias, value: tensor([ 0.0136, -0.0067, -0.0003, -0.0183, 0.0125, -0.0033, -0.0156, -0.0274, + 0.0096, -0.0015], device='cuda:0'), grad: tensor([ 0.0120, -0.0146, -0.0479, -0.0056, 0.0074, 0.0084, -0.0227, 0.0288, + 0.0258, 0.0083], device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 216.30, cls_loss 0.5733 cls_loss_mapping 0.0105 cls_loss_causal 0.5388 re_mapping 0.0084 re_causal 0.0211 /// teacc 98.61 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0192, 0.0775, -0.1389, ..., -0.0742, 0.0556, 0.0032], + [-0.0697, -0.0846, 0.0763, ..., -0.0277, -0.0893, 0.0477], + [-0.0588, -0.0405, -0.0933, ..., -0.1058, -0.0269, 0.0414], + ..., + [-0.0184, -0.0772, -0.0885, ..., -0.0995, -0.0143, 0.0549], + [-0.0915, -0.0859, -0.0821, ..., -0.0522, -0.0498, -0.0752], + [ 0.0124, -0.0119, -0.0512, ..., -0.0559, 0.0289, -0.0125]], + device='cuda:0'), grad: tensor([[ 4.2419e-03, 4.1618e-03, 7.2479e-05, ..., 6.3717e-05, + 4.7989e-03, 1.5192e-03], + [ 7.0620e-04, 5.5313e-05, -4.2915e-06, ..., 9.6932e-06, + 4.8614e-04, 8.4734e-04], + [ 1.3266e-03, 3.4833e-04, 1.0687e-04, ..., 1.3717e-05, + 8.3447e-04, 4.5662e-03], + ..., + [ 8.2970e-04, 2.7776e-04, 2.6330e-05, ..., 4.3586e-07, + 6.8760e-04, 1.3237e-03], + [-1.3914e-03, 3.1042e-04, 2.1887e-04, ..., 1.9026e-04, + -7.0190e-04, -6.0997e-03], + [-2.4738e-03, 7.7534e-04, 9.1672e-05, ..., 9.8795e-06, + -1.9836e-03, 9.0647e-04]], device='cuda:0') +Epoch 168, bias, value: tensor([ 0.0144, -0.0065, -0.0002, -0.0191, 0.0119, -0.0039, -0.0158, -0.0255, + 0.0094, -0.0020], device='cuda:0'), grad: tensor([ 0.0329, 0.0187, 0.0300, 0.0021, -0.0151, -0.0173, -0.0123, 0.0245, + -0.0734, 0.0099], device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 216.72, cls_loss 0.5616 cls_loss_mapping 0.0083 cls_loss_causal 0.5313 re_mapping 0.0089 re_causal 0.0228 /// teacc 98.57 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0188, 0.0782, -0.1400, ..., -0.0743, 0.0554, 0.0025], + [-0.0695, -0.0850, 0.0760, ..., -0.0278, -0.0891, 0.0472], + [-0.0600, -0.0402, -0.0934, ..., -0.1057, -0.0275, 0.0411], + ..., + [-0.0183, -0.0780, -0.0900, ..., -0.1003, -0.0139, 0.0554], + [-0.0918, -0.0857, -0.0825, ..., -0.0521, -0.0497, -0.0736], + [ 0.0126, -0.0133, -0.0505, ..., -0.0555, 0.0277, -0.0127]], + device='cuda:0'), grad: tensor([[ 5.1451e-04, -5.9813e-05, 3.6001e-04, ..., 8.2493e-05, + -8.3389e-03, 1.2484e-03], + [ 7.7200e-04, 1.0204e-04, 5.4061e-05, ..., 7.2002e-04, + 3.6597e-04, 6.5041e-04], + [ 1.0996e-03, 1.5554e-03, 2.2316e-03, ..., 8.8215e-04, + 1.2455e-03, 2.4796e-03], + ..., + [-6.8426e-04, 1.3041e-04, 1.1549e-03, ..., 3.8922e-05, + -1.2314e-04, 4.5633e-04], + [ 5.7793e-04, 2.1315e-04, 7.1764e-04, ..., 3.7980e-04, + 1.3475e-03, 1.4706e-03], + [ 7.1239e-04, 1.8442e-04, 7.2098e-04, ..., 4.5419e-05, + 4.8137e-04, 1.9121e-03]], device='cuda:0') +Epoch 169, bias, value: tensor([ 0.0127, -0.0059, -0.0006, -0.0192, 0.0122, -0.0041, -0.0148, -0.0258, + 0.0098, -0.0015], device='cuda:0'), grad: tensor([ 0.0049, 0.0196, 0.0287, -0.0546, 0.0149, -0.0419, -0.0082, -0.0068, + 0.0221, 0.0211], device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 216.28, cls_loss 0.5542 cls_loss_mapping 0.0089 cls_loss_causal 0.5274 re_mapping 0.0085 re_causal 0.0215 /// teacc 98.44 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0195, 0.0792, -0.1401, ..., -0.0745, 0.0552, 0.0024], + [-0.0694, -0.0844, 0.0748, ..., -0.0292, -0.0894, 0.0475], + [-0.0608, -0.0403, -0.0916, ..., -0.1055, -0.0279, 0.0398], + ..., + [-0.0182, -0.0790, -0.0898, ..., -0.1001, -0.0128, 0.0562], + [-0.0929, -0.0863, -0.0843, ..., -0.0524, -0.0499, -0.0727], + [ 0.0125, -0.0142, -0.0513, ..., -0.0552, 0.0274, -0.0128]], + device='cuda:0'), grad: tensor([[-7.7677e-04, 2.1505e-04, 1.6630e-05, ..., 2.6315e-05, + -1.6010e-04, -1.3244e-04], + [-3.0684e-04, 6.8784e-05, 5.9843e-04, ..., 1.0643e-03, + 6.0707e-05, -4.7714e-05], + [-6.8741e-03, -2.6679e-04, -3.6097e-04, ..., -5.3406e-04, + -3.7918e-03, 7.3290e-04], + ..., + [ 2.9926e-03, 2.8110e-04, 1.3912e-04, ..., 9.4891e-05, + 1.2178e-03, 1.8024e-03], + [ 4.2629e-04, 1.0902e-04, -1.3475e-03, ..., -2.6512e-03, + 6.8998e-04, 1.2789e-03], + [ 5.1308e-03, 2.1100e-04, 2.4939e-04, ..., 2.8825e-04, + 2.7485e-03, 3.8548e-03]], device='cuda:0') +Epoch 170, bias, value: tensor([ 0.0126, -0.0055, -0.0012, -0.0191, 0.0122, -0.0030, -0.0157, -0.0252, + 0.0095, -0.0019], device='cuda:0'), grad: tensor([-0.0171, -0.0259, -0.0315, 0.0272, -0.0382, 0.0284, -0.0011, 0.0254, + -0.0021, 0.0349], device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 216.17, cls_loss 0.5143 cls_loss_mapping 0.0077 cls_loss_causal 0.4776 re_mapping 0.0088 re_causal 0.0215 /// teacc 98.72 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0199, 0.0795, -0.1391, ..., -0.0747, 0.0567, 0.0021], + [-0.0692, -0.0847, 0.0757, ..., -0.0291, -0.0898, 0.0480], + [-0.0616, -0.0402, -0.0916, ..., -0.1058, -0.0282, 0.0398], + ..., + [-0.0177, -0.0803, -0.0900, ..., -0.1009, -0.0121, 0.0561], + [-0.0937, -0.0865, -0.0844, ..., -0.0524, -0.0490, -0.0733], + [ 0.0121, -0.0132, -0.0515, ..., -0.0547, 0.0265, -0.0124]], + device='cuda:0'), grad: tensor([[ 4.5705e-04, 8.1658e-05, 2.5010e-04, ..., 4.0007e-04, + 2.6321e-04, -2.6875e-03], + [ 4.0245e-04, 7.7844e-05, -5.8591e-05, ..., 4.6086e-04, + 4.6635e-04, 1.0994e-02], + [ 5.0449e-04, 2.8872e-04, 5.4264e-04, ..., 6.0940e-04, + 5.3644e-04, 1.3666e-03], + ..., + [-1.2274e-03, 1.8811e-04, 4.8232e-04, ..., 1.6487e-04, + 8.6164e-04, -1.9467e-04], + [ 3.7646e-04, 6.0844e-04, -7.3719e-04, ..., -5.5580e-03, + -1.9245e-03, -6.0701e-04], + [ 1.0767e-03, 4.1938e-04, 7.8726e-04, ..., 5.4979e-04, + 2.0657e-03, 1.7061e-03]], device='cuda:0') +Epoch 171, bias, value: tensor([ 0.0135, -0.0052, -0.0007, -0.0197, 0.0110, -0.0034, -0.0155, -0.0249, + 0.0092, -0.0018], device='cuda:0'), grad: tensor([-0.0115, 0.0314, 0.0153, -0.0073, 0.0095, -0.0188, 0.0060, -0.0122, + -0.0373, 0.0249], device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 216.58, cls_loss 0.5412 cls_loss_mapping 0.0081 cls_loss_causal 0.5108 re_mapping 0.0087 re_causal 0.0226 /// teacc 98.81 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0205, 0.0793, -0.1393, ..., -0.0747, 0.0572, 0.0029], + [-0.0701, -0.0858, 0.0756, ..., -0.0288, -0.0904, 0.0467], + [-0.0619, -0.0399, -0.0917, ..., -0.1064, -0.0281, 0.0395], + ..., + [-0.0173, -0.0807, -0.0898, ..., -0.1006, -0.0127, 0.0567], + [-0.0925, -0.0859, -0.0833, ..., -0.0527, -0.0491, -0.0731], + [ 0.0124, -0.0125, -0.0505, ..., -0.0551, 0.0268, -0.0117]], + device='cuda:0'), grad: tensor([[ 8.6517e-03, 3.0708e-03, 2.8086e-04, ..., -4.5568e-05, + 9.4452e-03, 7.1192e-04], + [-6.5279e-04, 3.3110e-05, 1.9625e-05, ..., -3.3450e-04, + 5.9986e-04, 8.2111e-04], + [-3.8338e-03, 2.2840e-04, 1.5306e-04, ..., 3.2723e-05, + -1.7529e-03, -1.2302e-03], + ..., + [-3.9787e-03, 6.4194e-05, -3.4094e-04, ..., 2.2426e-05, + -4.6223e-05, -1.9321e-03], + [ 2.0294e-03, 1.7138e-03, 1.0900e-03, ..., 1.1486e-04, + 1.3609e-03, 4.4203e-04], + [ 2.0313e-03, 4.2528e-05, 8.9467e-05, ..., 2.6569e-05, + -6.5899e-04, 7.2050e-04]], device='cuda:0') +Epoch 172, bias, value: tensor([ 0.0141, -0.0051, -0.0017, -0.0204, 0.0115, -0.0036, -0.0156, -0.0246, + 0.0099, -0.0018], device='cuda:0'), grad: tensor([ 0.0345, -0.0391, -0.0207, 0.0200, 0.0102, -0.0041, -0.0116, -0.0056, + 0.0179, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 216.31, cls_loss 0.5218 cls_loss_mapping 0.0082 cls_loss_causal 0.4966 re_mapping 0.0084 re_causal 0.0221 /// teacc 98.60 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0191, 0.0788, -0.1407, ..., -0.0759, 0.0568, 0.0019], + [-0.0695, -0.0856, 0.0763, ..., -0.0291, -0.0907, 0.0472], + [-0.0613, -0.0404, -0.0922, ..., -0.1070, -0.0280, 0.0403], + ..., + [-0.0176, -0.0809, -0.0902, ..., -0.0999, -0.0133, 0.0559], + [-0.0924, -0.0860, -0.0838, ..., -0.0528, -0.0490, -0.0722], + [ 0.0122, -0.0126, -0.0510, ..., -0.0555, 0.0273, -0.0120]], + device='cuda:0'), grad: tensor([[-2.4462e-04, -6.3562e-04, 4.3720e-05, ..., 1.0794e-04, + -1.8501e-03, 6.3896e-04], + [ 2.5711e-03, -2.8076e-03, -8.0013e-04, ..., 7.3947e-06, + 1.6594e-03, -1.6522e-04], + [ 8.7404e-04, 6.7472e-04, 2.6059e-04, ..., 2.7016e-05, + 4.9973e-04, 1.4315e-03], + ..., + [-2.2869e-03, 3.2854e-04, 1.5259e-04, ..., 1.9774e-05, + -8.4209e-04, -4.1161e-03], + [ 1.1053e-03, -4.8752e-03, -7.5455e-03, ..., -1.5945e-03, + -8.1491e-04, 9.9277e-04], + [ 8.2254e-04, 1.4400e-03, 1.8072e-03, ..., 4.1986e-04, + 1.2016e-03, 8.3208e-04]], device='cuda:0') +Epoch 173, bias, value: tensor([ 0.0128, -0.0044, -0.0018, -0.0191, 0.0114, -0.0046, -0.0149, -0.0246, + 0.0104, -0.0027], device='cuda:0'), grad: tensor([-0.0032, -0.0004, 0.0148, -0.0008, -0.0177, 0.0054, 0.0188, -0.0186, + -0.0010, 0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 216.39, cls_loss 0.5384 cls_loss_mapping 0.0081 cls_loss_causal 0.5032 re_mapping 0.0092 re_causal 0.0242 /// teacc 98.71 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0196, 0.0800, -0.1402, ..., -0.0764, 0.0577, 0.0024], + [-0.0713, -0.0868, 0.0749, ..., -0.0302, -0.0924, 0.0469], + [-0.0617, -0.0398, -0.0920, ..., -0.1055, -0.0271, 0.0403], + ..., + [-0.0182, -0.0824, -0.0910, ..., -0.1005, -0.0146, 0.0562], + [-0.0931, -0.0858, -0.0834, ..., -0.0542, -0.0501, -0.0729], + [ 0.0131, -0.0126, -0.0518, ..., -0.0557, 0.0274, -0.0122]], + device='cuda:0'), grad: tensor([[-2.3937e-03, -1.4079e-04, 7.1168e-05, ..., 2.1076e-04, + -6.2523e-03, 8.9979e-04], + [-3.3498e-04, -1.4353e-04, -3.6627e-05, ..., -2.1820e-03, + 9.1970e-05, -5.9929e-03], + [-2.6608e-03, 1.2058e-04, 5.3316e-05, ..., 1.4555e-04, + -2.8744e-03, -4.1161e-03], + ..., + [ 2.3961e-04, 1.2827e-04, 5.7250e-05, ..., 3.8457e-04, + 1.0719e-03, 1.1091e-03], + [ 1.1282e-03, 1.5044e-04, 5.9545e-05, ..., 5.9509e-04, + 8.0061e-04, 2.0771e-03], + [ 1.3285e-03, 2.8205e-04, 1.0979e-04, ..., 3.6693e-04, + 9.1505e-04, 2.3079e-03]], device='cuda:0') +Epoch 174, bias, value: tensor([ 0.0130, -0.0055, -0.0018, -0.0191, 0.0112, -0.0042, -0.0142, -0.0241, + 0.0101, -0.0027], device='cuda:0'), grad: tensor([-0.0010, -0.0555, -0.0421, -0.0103, 0.0234, 0.0155, 0.0262, 0.0142, + 0.0012, 0.0284], device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 216.14, cls_loss 0.5288 cls_loss_mapping 0.0111 cls_loss_causal 0.4991 re_mapping 0.0087 re_causal 0.0214 /// teacc 98.70 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0189, 0.0801, -0.1397, ..., -0.0776, 0.0573, 0.0021], + [-0.0712, -0.0873, 0.0738, ..., -0.0295, -0.0919, 0.0472], + [-0.0615, -0.0402, -0.0924, ..., -0.1056, -0.0271, 0.0408], + ..., + [-0.0182, -0.0826, -0.0917, ..., -0.0994, -0.0152, 0.0559], + [-0.0937, -0.0847, -0.0829, ..., -0.0550, -0.0505, -0.0728], + [ 0.0130, -0.0129, -0.0519, ..., -0.0566, 0.0276, -0.0122]], + device='cuda:0'), grad: tensor([[-6.9904e-04, 2.6846e-04, 1.8191e-04, ..., 5.6314e-04, + -1.0357e-03, -2.4915e-04], + [ 1.1730e-03, 8.4221e-05, 2.6441e-04, ..., 2.4509e-04, + 5.0545e-04, 6.4802e-04], + [ 1.1063e-03, 3.6025e-04, 1.8644e-04, ..., -1.1158e-04, + 6.4516e-04, 6.8235e-04], + ..., + [-5.7840e-04, 7.5436e-04, 4.9019e-04, ..., 2.2626e-04, + -3.8171e-04, -2.2564e-03], + [ 1.1768e-03, 4.2915e-04, 4.1008e-04, ..., 4.6682e-04, + 5.0259e-04, 5.5552e-04], + [ 5.4893e-03, 3.0160e-04, 2.8343e-03, ..., 6.2656e-04, + 4.9877e-04, 9.0647e-04]], device='cuda:0') +Epoch 175, bias, value: tensor([ 0.0123, -0.0050, -0.0012, -0.0176, 0.0112, -0.0059, -0.0139, -0.0239, + 0.0096, -0.0033], device='cuda:0'), grad: tensor([-0.0183, 0.0130, 0.0204, -0.0095, -0.0432, 0.0024, 0.0065, -0.0078, + 0.0148, 0.0218], device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 216.18, cls_loss 0.5056 cls_loss_mapping 0.0097 cls_loss_causal 0.4772 re_mapping 0.0087 re_causal 0.0223 /// teacc 98.67 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0190, 0.0801, -0.1399, ..., -0.0784, 0.0574, 0.0028], + [-0.0699, -0.0852, 0.0738, ..., -0.0294, -0.0907, 0.0466], + [-0.0622, -0.0418, -0.0930, ..., -0.1052, -0.0278, 0.0406], + ..., + [-0.0186, -0.0838, -0.0905, ..., -0.0997, -0.0157, 0.0566], + [-0.0922, -0.0842, -0.0835, ..., -0.0543, -0.0502, -0.0722], + [ 0.0130, -0.0127, -0.0525, ..., -0.0563, 0.0279, -0.0122]], + device='cuda:0'), grad: tensor([[-3.3607e-03, -3.0499e-03, 5.4715e-07, ..., -1.0195e-03, + -9.3746e-04, 8.8513e-05], + [-1.7920e-03, -7.2527e-04, 4.4471e-07, ..., -1.8196e-03, + -1.1625e-03, -3.3474e-03], + [ 8.8310e-04, 1.4865e-04, 1.0803e-06, ..., 8.4519e-05, + 3.5882e-04, -1.7285e-04], + ..., + [ 7.9918e-04, 1.4567e-04, 5.7463e-07, ..., 1.0073e-04, + 4.1366e-04, 1.3275e-03], + [ 1.2379e-03, 2.5749e-04, 1.8135e-05, ..., 2.2817e-04, + 4.5848e-04, 3.1471e-04], + [ 2.2316e-03, 3.9244e-04, 2.0415e-06, ..., 5.4693e-04, + 1.0052e-03, 2.2831e-03]], device='cuda:0') +Epoch 176, bias, value: tensor([ 0.0129, -0.0039, -0.0018, -0.0187, 0.0117, -0.0063, -0.0143, -0.0244, + 0.0104, -0.0031], device='cuda:0'), grad: tensor([-0.0533, -0.0352, 0.0020, 0.0081, -0.0149, 0.0162, 0.0328, 0.0102, + 0.0095, 0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 216.16, cls_loss 0.5238 cls_loss_mapping 0.0063 cls_loss_causal 0.4916 re_mapping 0.0086 re_causal 0.0219 /// teacc 98.60 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0195, 0.0799, -0.1408, ..., -0.0795, 0.0581, 0.0031], + [-0.0709, -0.0864, 0.0745, ..., -0.0300, -0.0919, 0.0466], + [-0.0613, -0.0417, -0.0932, ..., -0.1057, -0.0272, 0.0411], + ..., + [-0.0186, -0.0844, -0.0910, ..., -0.0994, -0.0159, 0.0573], + [-0.0935, -0.0844, -0.0842, ..., -0.0540, -0.0510, -0.0733], + [ 0.0124, -0.0140, -0.0523, ..., -0.0561, 0.0281, -0.0132]], + device='cuda:0'), grad: tensor([[ 1.2980e-03, 8.8358e-04, -1.6832e-04, ..., 1.4770e-04, + 4.6045e-05, 8.4734e-04], + [ 1.8520e-03, 5.9700e-04, -5.4985e-05, ..., 3.0470e-04, + 4.7755e-04, 6.1798e-04], + [-3.3417e-03, -1.0176e-03, 2.6718e-05, ..., -9.4414e-04, + -5.9462e-04, -1.6613e-03], + ..., + [ 1.0614e-03, 1.5628e-04, 9.4622e-06, ..., 9.0539e-05, + 3.7050e-04, 1.3418e-03], + [ 1.7433e-03, 7.1621e-04, 2.6792e-05, ..., 2.9206e-04, + 6.7568e-04, 7.9823e-04], + [ 1.6518e-03, 3.5286e-04, 3.0369e-05, ..., 2.4724e-04, + 7.5960e-04, 7.5912e-04]], device='cuda:0') +Epoch 177, bias, value: tensor([ 0.0132, -0.0036, -0.0023, -0.0179, 0.0123, -0.0057, -0.0149, -0.0243, + 0.0096, -0.0039], device='cuda:0'), grad: tensor([ 0.0191, 0.0151, 0.0048, -0.0107, -0.0105, 0.0051, -0.0354, -0.0090, + 0.0060, 0.0154], device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 216.27, cls_loss 0.5479 cls_loss_mapping 0.0086 cls_loss_causal 0.5112 re_mapping 0.0087 re_causal 0.0227 /// teacc 98.48 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0202, 0.0801, -0.1403, ..., -0.0792, 0.0580, 0.0030], + [-0.0712, -0.0881, 0.0740, ..., -0.0298, -0.0898, 0.0469], + [-0.0607, -0.0417, -0.0950, ..., -0.1058, -0.0259, 0.0408], + ..., + [-0.0185, -0.0836, -0.0905, ..., -0.1003, -0.0159, 0.0574], + [-0.0921, -0.0834, -0.0850, ..., -0.0544, -0.0496, -0.0736], + [ 0.0125, -0.0147, -0.0524, ..., -0.0568, 0.0282, -0.0141]], + device='cuda:0'), grad: tensor([[ 1.0529e-03, 3.1948e-05, 3.3677e-06, ..., 2.9862e-05, + 6.3276e-04, 2.5916e-04], + [-1.3123e-03, 1.8865e-05, 4.2580e-06, ..., 2.1267e-04, + -7.8869e-04, 2.1923e-04], + [-4.8027e-03, 5.1928e-04, 1.1313e-04, ..., -1.6680e-03, + -2.6855e-03, -8.2626e-03], + ..., + [ 1.1654e-03, 2.3437e-04, 3.8832e-05, ..., 1.2579e-03, + 7.5436e-04, 6.6032e-03], + [-1.8272e-03, -6.8331e-04, 1.4342e-05, ..., 9.1717e-06, + -3.3360e-03, -7.8344e-04], + [ 1.2894e-03, 2.8968e-04, 1.6540e-05, ..., 4.0740e-05, + 8.2350e-04, 7.4387e-04]], device='cuda:0') +Epoch 178, bias, value: tensor([ 0.0137, -0.0033, -0.0019, -0.0175, 0.0128, -0.0069, -0.0158, -0.0237, + 0.0106, -0.0056], device='cuda:0'), grad: tensor([ 0.0136, -0.0245, -0.0751, 0.0172, 0.0177, 0.0405, 0.0128, 0.0382, + -0.0618, 0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 216.11, cls_loss 0.5348 cls_loss_mapping 0.0098 cls_loss_causal 0.5102 re_mapping 0.0080 re_causal 0.0210 /// teacc 98.80 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0210, 0.0798, -0.1402, ..., -0.0798, 0.0574, 0.0028], + [-0.0714, -0.0886, 0.0740, ..., -0.0299, -0.0897, 0.0481], + [-0.0605, -0.0419, -0.0951, ..., -0.1061, -0.0258, 0.0418], + ..., + [-0.0174, -0.0838, -0.0917, ..., -0.1002, -0.0152, 0.0565], + [-0.0931, -0.0841, -0.0841, ..., -0.0547, -0.0500, -0.0731], + [ 0.0131, -0.0139, -0.0543, ..., -0.0569, 0.0292, -0.0145]], + device='cuda:0'), grad: tensor([[ 2.4624e-03, 8.6566e-07, 4.5836e-05, ..., 6.9427e-04, + 1.0910e-03, 5.5170e-04], + [-1.6117e-03, 2.6654e-06, 5.5408e-04, ..., 3.0251e-03, + 1.0639e-05, 5.3482e-03], + [ 1.7357e-03, -4.7374e-04, 7.5161e-05, ..., 5.0211e-04, + 7.7486e-04, 1.3876e-03], + ..., + [ 1.3723e-03, 1.5095e-05, 6.8903e-05, ..., 4.0770e-04, + 7.0000e-04, -7.5951e-03], + [ 2.2335e-03, 8.8930e-05, 7.3075e-05, ..., 5.4073e-04, + 7.7391e-04, 1.0691e-03], + [ 5.2338e-03, 1.1191e-05, 6.9380e-04, ..., 9.5749e-04, + 1.8682e-03, 1.3769e-04]], device='cuda:0') +Epoch 179, bias, value: tensor([ 0.0142, -0.0028, -0.0009, -0.0184, 0.0119, -0.0073, -0.0156, -0.0239, + 0.0106, -0.0054], device='cuda:0'), grad: tensor([ 0.0179, 0.0084, 0.0133, 0.0183, -0.0311, 0.0098, -0.0204, -0.0273, + 0.0195, -0.0085], device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 216.17, cls_loss 0.5726 cls_loss_mapping 0.0092 cls_loss_causal 0.5432 re_mapping 0.0086 re_causal 0.0229 /// teacc 98.67 lr 0.00010000 +Epoch 180, weight, value: tensor([[ 0.0207, 0.0803, -0.1408, ..., -0.0806, 0.0579, 0.0032], + [-0.0721, -0.0884, 0.0745, ..., -0.0293, -0.0902, 0.0482], + [-0.0600, -0.0422, -0.0944, ..., -0.1058, -0.0254, 0.0423], + ..., + [-0.0186, -0.0845, -0.0930, ..., -0.1007, -0.0166, 0.0557], + [-0.0931, -0.0850, -0.0852, ..., -0.0545, -0.0492, -0.0741], + [ 0.0128, -0.0139, -0.0544, ..., -0.0579, 0.0294, -0.0142]], + device='cuda:0'), grad: tensor([[-5.9217e-05, 3.9309e-05, 2.9337e-06, ..., 6.2847e-04, + 1.1120e-03, -1.2708e-04], + [ 2.0390e-03, 1.1042e-05, -9.9719e-05, ..., -1.1182e-04, + 1.1883e-03, 1.0532e-04], + [-2.6131e-03, 1.3244e-04, -1.2264e-05, ..., -2.9049e-03, + -2.8343e-03, -1.4019e-03], + ..., + [-1.4286e-03, 1.9670e-05, 2.6256e-05, ..., 1.3018e-04, + -3.9607e-05, -7.8154e-04], + [ 3.2902e-03, 1.2350e-03, 1.2815e-05, ..., 5.5933e-04, + 1.6365e-03, 5.0831e-04], + [-1.6623e-03, 8.7118e-04, 5.0366e-06, ..., -4.0436e-04, + -3.8128e-03, 3.4833e-04]], device='cuda:0') +Epoch 180, bias, value: tensor([ 0.0138, -0.0036, -0.0002, -0.0179, 0.0122, -0.0066, -0.0157, -0.0239, + 0.0105, -0.0060], device='cuda:0'), grad: tensor([-0.0083, 0.0198, -0.0083, 0.0182, -0.0350, -0.0051, 0.0252, -0.0184, + 0.0219, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 215.96, cls_loss 0.5354 cls_loss_mapping 0.0095 cls_loss_causal 0.5037 re_mapping 0.0083 re_causal 0.0206 /// teacc 98.66 lr 0.00010000 +Epoch 181, weight, value: tensor([[ 0.0210, 0.0796, -0.1420, ..., -0.0811, 0.0585, 0.0035], + [-0.0717, -0.0878, 0.0758, ..., -0.0289, -0.0899, 0.0480], + [-0.0596, -0.0420, -0.0945, ..., -0.1057, -0.0258, 0.0426], + ..., + [-0.0169, -0.0837, -0.0942, ..., -0.1008, -0.0157, 0.0560], + [-0.0940, -0.0869, -0.0867, ..., -0.0561, -0.0506, -0.0745], + [ 0.0121, -0.0135, -0.0547, ..., -0.0575, 0.0284, -0.0142]], + device='cuda:0'), grad: tensor([[ 1.7176e-03, 2.0142e-03, 1.4631e-06, ..., 3.3230e-05, + 9.2010e-03, 1.5802e-03], + [ 6.7806e-04, 2.2352e-06, -4.7743e-05, ..., 2.6360e-05, + 2.0349e-04, 1.3714e-03], + [ 1.2026e-03, 1.3494e-04, 3.2157e-05, ..., 5.5671e-05, + 6.1369e-04, -2.1896e-03], + ..., + [ 1.4601e-03, 3.2663e-05, 3.0398e-04, ..., 3.6180e-05, + 1.2398e-04, 2.0618e-03], + [ 4.0960e-04, 4.4078e-05, -3.8862e-05, ..., 3.0845e-05, + 1.2338e-04, 4.8089e-04], + [-3.3550e-03, -2.7905e-03, -6.0177e-04, ..., 2.4095e-05, + -9.9182e-03, -1.4801e-03]], device='cuda:0') +Epoch 181, bias, value: tensor([ 0.0135, -0.0035, -0.0001, -0.0178, 0.0112, -0.0059, -0.0159, -0.0225, + 0.0100, -0.0064], device='cuda:0'), grad: tensor([ 0.0007, 0.0112, 0.0084, -0.0103, 0.0108, -0.0228, 0.0080, 0.0070, + 0.0065, -0.0194], device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 215.90, cls_loss 0.5509 cls_loss_mapping 0.0129 cls_loss_causal 0.5216 re_mapping 0.0085 re_causal 0.0218 /// teacc 98.63 lr 0.00010000 +Epoch 182, weight, value: tensor([[ 0.0206, 0.0799, -0.1422, ..., -0.0808, 0.0587, 0.0016], + [-0.0722, -0.0876, 0.0753, ..., -0.0295, -0.0895, 0.0488], + [-0.0604, -0.0425, -0.0945, ..., -0.1058, -0.0264, 0.0431], + ..., + [-0.0160, -0.0836, -0.0939, ..., -0.0990, -0.0157, 0.0559], + [-0.0950, -0.0856, -0.0863, ..., -0.0567, -0.0507, -0.0747], + [ 0.0119, -0.0129, -0.0537, ..., -0.0570, 0.0288, -0.0139]], + device='cuda:0'), grad: tensor([[-1.0414e-03, -9.0778e-05, 2.4036e-05, ..., 2.2709e-04, + -2.8172e-03, 6.6471e-04], + [ 9.2793e-04, 7.1451e-06, -7.9572e-05, ..., -8.8835e-04, + 1.9050e-04, -1.1263e-03], + [ 3.2482e-03, 2.2042e-04, 3.5453e-04, ..., 3.3879e-04, + 6.9427e-04, 1.4849e-03], + ..., + [ 3.5019e-03, 6.8665e-05, 3.5214e-04, ..., 1.6892e-04, + 2.6512e-04, 1.1806e-03], + [ 2.9411e-03, 3.2574e-05, 1.3888e-04, ..., 6.4135e-04, + 7.7963e-04, 1.8044e-03], + [-4.0970e-03, -1.5676e-04, -1.3218e-03, ..., -9.0981e-04, + 9.0301e-06, -4.2839e-03]], device='cuda:0') +Epoch 182, bias, value: tensor([ 1.2485e-02, -3.7579e-03, -6.0627e-05, -1.8954e-02, 1.0987e-02, + -5.5898e-03, -1.5346e-02, -2.1773e-02, 1.0248e-02, -5.8389e-03], + device='cuda:0'), grad: tensor([ 0.0062, -0.0139, 0.0273, -0.0123, 0.0353, -0.0085, 0.0176, 0.0232, + -0.0244, -0.0504], device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 216.32, cls_loss 0.5489 cls_loss_mapping 0.0078 cls_loss_causal 0.5276 re_mapping 0.0085 re_causal 0.0224 /// teacc 98.82 lr 0.00010000 +Epoch 183, weight, value: tensor([[ 0.0197, 0.0795, -0.1431, ..., -0.0817, 0.0593, 0.0009], + [-0.0717, -0.0873, 0.0757, ..., -0.0307, -0.0887, 0.0498], + [-0.0603, -0.0432, -0.0948, ..., -0.1058, -0.0263, 0.0431], + ..., + [-0.0163, -0.0835, -0.0954, ..., -0.0997, -0.0160, 0.0550], + [-0.0954, -0.0863, -0.0865, ..., -0.0559, -0.0514, -0.0761], + [ 0.0117, -0.0131, -0.0535, ..., -0.0569, 0.0280, -0.0132]], + device='cuda:0'), grad: tensor([[ 8.5545e-04, 3.4738e-04, 5.0354e-04, ..., 1.6487e-04, + 1.2922e-04, 6.3515e-04], + [-2.8286e-03, -4.8828e-04, -7.0620e-04, ..., -5.0688e-04, + -1.4915e-03, -1.2369e-03], + [-3.4466e-03, 9.9838e-05, 2.5570e-05, ..., 1.9789e-04, + -9.8896e-04, -3.3436e-03], + ..., + [ 9.0551e-04, 6.6566e-04, 4.1068e-05, ..., 6.1333e-05, + 3.0923e-04, 8.9025e-04], + [-6.2084e-04, 2.6017e-05, -3.1281e-04, ..., 3.6097e-04, + 2.0373e-04, 3.6216e-04], + [ 6.7139e-04, 6.3658e-05, 1.1802e-04, ..., 7.8440e-05, + 2.6464e-04, 4.6182e-04]], device='cuda:0') +Epoch 183, bias, value: tensor([ 1.1367e-02, -2.7596e-03, 3.4431e-05, -1.9674e-02, 1.1644e-02, + -3.8947e-03, -1.5151e-02, -2.2966e-02, 8.9731e-03, -5.1283e-03], + device='cuda:0'), grad: tensor([ 0.0141, -0.0292, -0.0441, 0.0226, 0.0110, 0.0100, 0.0168, 0.0143, + -0.0247, 0.0092], device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 215.88, cls_loss 0.5020 cls_loss_mapping 0.0085 cls_loss_causal 0.4810 re_mapping 0.0085 re_causal 0.0215 /// teacc 98.52 lr 0.00010000 +Epoch 184, weight, value: tensor([[ 0.0205, 0.0804, -0.1441, ..., -0.0813, 0.0596, 0.0008], + [-0.0732, -0.0873, 0.0760, ..., -0.0303, -0.0889, 0.0491], + [-0.0616, -0.0443, -0.0940, ..., -0.1059, -0.0268, 0.0431], + ..., + [-0.0158, -0.0841, -0.0969, ..., -0.1005, -0.0157, 0.0557], + [-0.0970, -0.0869, -0.0873, ..., -0.0555, -0.0530, -0.0761], + [ 0.0122, -0.0138, -0.0547, ..., -0.0564, 0.0272, -0.0133]], + device='cuda:0'), grad: tensor([[ 6.3972e-03, 1.0548e-03, 1.8144e-04, ..., 2.4757e-03, + 5.3329e-03, 1.6766e-03], + [-1.1282e-03, 2.9945e-04, 3.6240e-04, ..., 4.4098e-03, + 1.9729e-04, 2.0516e-04], + [ 1.4472e-04, 1.4210e-03, 4.1819e-04, ..., 1.5287e-03, + 2.1400e-03, -2.7218e-03], + ..., + [-2.4090e-03, 8.4043e-05, 2.3723e-04, ..., 5.2601e-05, + -1.0214e-03, -2.1935e-03], + [ 1.0900e-03, 4.1056e-04, 7.1430e-04, ..., 5.6887e-04, + 7.1096e-04, 7.3719e-04], + [ 1.1492e-03, 2.4343e-04, 2.2221e-04, ..., 1.2767e-04, + 7.4005e-04, 4.7445e-04]], device='cuda:0') +Epoch 184, bias, value: tensor([ 0.0121, -0.0036, -0.0003, -0.0202, 0.0118, -0.0039, -0.0150, -0.0227, + 0.0092, -0.0050], device='cuda:0'), grad: tensor([ 0.0403, 0.0001, -0.0059, -0.0219, 0.0083, 0.0172, -0.0369, -0.0174, + 0.0172, -0.0010], device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 216.47, cls_loss 0.5221 cls_loss_mapping 0.0085 cls_loss_causal 0.4886 re_mapping 0.0093 re_causal 0.0240 /// teacc 98.55 lr 0.00010000 +Epoch 185, weight, value: tensor([[ 0.0214, 0.0810, -0.1453, ..., -0.0813, 0.0605, 0.0009], + [-0.0722, -0.0876, 0.0763, ..., -0.0291, -0.0883, 0.0497], + [-0.0627, -0.0452, -0.0926, ..., -0.1079, -0.0275, 0.0435], + ..., + [-0.0151, -0.0840, -0.0948, ..., -0.1007, -0.0158, 0.0549], + [-0.0973, -0.0870, -0.0869, ..., -0.0561, -0.0533, -0.0758], + [ 0.0126, -0.0138, -0.0562, ..., -0.0569, 0.0276, -0.0136]], + device='cuda:0'), grad: tensor([[-1.5135e-03, -8.4496e-04, 2.8655e-05, ..., 1.7965e-04, + -2.4166e-03, 3.3450e-04], + [ 9.0408e-04, 6.2644e-05, 6.5207e-05, ..., 2.2006e-04, + 4.7636e-04, 6.4850e-04], + [ 1.8129e-03, 5.4979e-04, 4.0936e-04, ..., 6.7759e-04, + 1.7023e-03, -1.6155e-03], + ..., + [ 1.1053e-03, 2.7299e-05, -3.6389e-05, ..., 5.4646e-04, + 9.5606e-04, 2.2087e-03], + [-2.5978e-03, 1.4651e-04, 4.9162e-04, ..., 4.5228e-04, + 2.4796e-04, 5.5265e-04], + [-1.2688e-05, 5.3465e-05, 3.6031e-05, ..., 1.2189e-04, + 8.4877e-05, 4.2129e-04]], device='cuda:0') +Epoch 185, bias, value: tensor([ 0.0123, -0.0029, -0.0014, -0.0199, 0.0119, -0.0045, -0.0150, -0.0233, + 0.0097, -0.0044], device='cuda:0'), grad: tensor([-0.0103, 0.0136, 0.0111, 0.0159, -0.0498, 0.0089, 0.0138, 0.0213, + -0.0327, 0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 216.04, cls_loss 0.5583 cls_loss_mapping 0.0084 cls_loss_causal 0.5348 re_mapping 0.0085 re_causal 0.0229 /// teacc 98.64 lr 0.00010000 +Epoch 186, weight, value: tensor([[ 0.0205, 0.0807, -0.1457, ..., -0.0825, 0.0605, 0.0006], + [-0.0737, -0.0880, 0.0766, ..., -0.0294, -0.0889, 0.0495], + [-0.0614, -0.0451, -0.0937, ..., -0.1078, -0.0275, 0.0436], + ..., + [-0.0138, -0.0841, -0.0945, ..., -0.1004, -0.0171, 0.0546], + [-0.0958, -0.0870, -0.0867, ..., -0.0563, -0.0520, -0.0749], + [ 0.0115, -0.0128, -0.0563, ..., -0.0571, 0.0283, -0.0133]], + device='cuda:0'), grad: tensor([[ 3.8552e-04, 2.3812e-05, 7.5400e-05, ..., 3.3927e-04, + 2.4402e-04, 5.1785e-04], + [-1.4486e-03, 8.5533e-06, 1.2350e-04, ..., 1.8251e-04, + -6.3276e-04, -1.1749e-03], + [-2.3711e-04, 1.7321e-04, 2.4486e-04, ..., 4.7350e-04, + -3.4285e-04, -9.7227e-04], + ..., + [ 1.0099e-03, 6.7540e-06, 6.3837e-05, ..., 2.1958e-04, + 4.9973e-04, -4.4703e-04], + [-1.4486e-03, 1.1069e-04, -1.1950e-03, ..., 4.5204e-03, + -5.6028e-04, -1.2932e-03], + [-1.9836e-03, 1.6201e-04, 1.0920e-04, ..., 1.0204e-03, + -2.9564e-05, 1.0338e-03]], device='cuda:0') +Epoch 186, bias, value: tensor([ 0.0115, -0.0036, -0.0011, -0.0193, 0.0118, -0.0055, -0.0155, -0.0229, + 0.0113, -0.0044], device='cuda:0'), grad: tensor([ 0.0110, -0.0137, -0.0146, 0.0055, 0.0231, -0.0122, -0.0216, 0.0136, + 0.0015, 0.0073], device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 216.65, cls_loss 0.5419 cls_loss_mapping 0.0090 cls_loss_causal 0.5166 re_mapping 0.0081 re_causal 0.0213 /// teacc 98.64 lr 0.00010000 +Epoch 187, weight, value: tensor([[ 0.0215, 0.0812, -0.1463, ..., -0.0828, 0.0622, 0.0009], + [-0.0746, -0.0881, 0.0766, ..., -0.0295, -0.0890, 0.0491], + [-0.0606, -0.0443, -0.0926, ..., -0.1074, -0.0274, 0.0437], + ..., + [-0.0143, -0.0846, -0.0952, ..., -0.1007, -0.0175, 0.0556], + [-0.0958, -0.0865, -0.0870, ..., -0.0576, -0.0521, -0.0752], + [ 0.0117, -0.0138, -0.0574, ..., -0.0575, 0.0280, -0.0138]], + device='cuda:0'), grad: tensor([[-9.9182e-04, -1.1730e-03, 1.0192e-04, ..., -7.2527e-04, + -3.1357e-03, 4.0841e-04], + [ 9.8801e-04, 4.4227e-05, 6.9916e-05, ..., -1.3046e-03, + 1.2083e-03, 6.9141e-04], + [ 1.2989e-03, 3.7909e-04, 2.3174e-04, ..., 2.3067e-04, + -9.4652e-04, -9.8705e-04], + ..., + [ 1.6584e-03, 1.9383e-04, 1.7099e-03, ..., 2.0146e-04, + 2.0275e-03, 3.7098e-03], + [ 3.2597e-03, 2.5654e-03, 2.2392e-03, ..., 1.0147e-03, + 1.7185e-03, 1.0242e-03], + [-3.7937e-03, -6.0827e-05, -2.1896e-03, ..., 2.0695e-04, + -1.3113e-03, -6.0310e-03]], device='cuda:0') +Epoch 187, bias, value: tensor([ 0.0115, -0.0041, -0.0007, -0.0195, 0.0124, -0.0055, -0.0150, -0.0226, + 0.0112, -0.0054], device='cuda:0'), grad: tensor([-0.0212, 0.0098, -0.0104, -0.0177, 0.0314, 0.0014, -0.0038, 0.0287, + 0.0387, -0.0568], device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 216.12, cls_loss 0.5479 cls_loss_mapping 0.0054 cls_loss_causal 0.5161 re_mapping 0.0083 re_causal 0.0226 /// teacc 98.36 lr 0.00010000 +Epoch 188, weight, value: tensor([[ 0.0225, 0.0810, -0.1465, ..., -0.0823, 0.0622, 0.0011], + [-0.0748, -0.0880, 0.0773, ..., -0.0296, -0.0895, 0.0494], + [-0.0613, -0.0447, -0.0934, ..., -0.1079, -0.0281, 0.0422], + ..., + [-0.0153, -0.0849, -0.0949, ..., -0.1012, -0.0181, 0.0567], + [-0.0945, -0.0854, -0.0871, ..., -0.0572, -0.0507, -0.0760], + [ 0.0121, -0.0146, -0.0582, ..., -0.0567, 0.0287, -0.0140]], + device='cuda:0'), grad: tensor([[-2.1515e-03, 3.3796e-05, 4.3184e-05, ..., 1.9658e-04, + -1.4286e-03, -7.6711e-05], + [ 2.9016e-04, 2.7061e-05, -2.0707e-04, ..., 8.3876e-04, + 9.0313e-04, -4.1699e-04], + [ 3.3283e-04, 1.1051e-04, -4.5562e-04, ..., -1.8473e-03, + -8.4915e-03, -3.4428e-03], + ..., + [ 2.1229e-03, 1.5306e-03, 3.7527e-04, ..., 1.1187e-03, + 2.4815e-03, 5.2528e-03], + [ 4.8399e-04, 3.1590e-04, 4.6849e-05, ..., 2.4116e-04, + 6.2065e-03, 1.9627e-03], + [ 6.6376e-04, 2.3413e-04, 5.9307e-06, ..., 2.5582e-04, + 6.3372e-04, 8.6260e-04]], device='cuda:0') +Epoch 188, bias, value: tensor([ 0.0122, -0.0041, -0.0012, -0.0186, 0.0119, -0.0062, -0.0157, -0.0227, + 0.0118, -0.0050], device='cuda:0'), grad: tensor([-0.0237, 0.0082, -0.0502, -0.0155, 0.0035, 0.0054, 0.0060, 0.0294, + 0.0565, -0.0196], device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 216.43, cls_loss 0.5613 cls_loss_mapping 0.0086 cls_loss_causal 0.5278 re_mapping 0.0079 re_causal 0.0213 /// teacc 98.64 lr 0.00010000 +Epoch 189, weight, value: tensor([[ 0.0220, 0.0815, -0.1466, ..., -0.0821, 0.0612, 0.0011], + [-0.0743, -0.0879, 0.0784, ..., -0.0299, -0.0881, 0.0492], + [-0.0608, -0.0457, -0.0937, ..., -0.1070, -0.0275, 0.0431], + ..., + [-0.0153, -0.0850, -0.0949, ..., -0.1015, -0.0184, 0.0569], + [-0.0950, -0.0852, -0.0869, ..., -0.0568, -0.0515, -0.0768], + [ 0.0116, -0.0153, -0.0579, ..., -0.0575, 0.0288, -0.0147]], + device='cuda:0'), grad: tensor([[ 1.2808e-03, -7.3051e-04, 1.3009e-05, ..., -2.2784e-05, + 8.6975e-04, 2.2984e-04], + [ 2.0695e-03, 1.7166e-05, -5.0545e-04, ..., 1.7837e-05, + 1.7147e-03, 3.4666e-04], + [-6.2637e-03, 7.3135e-05, 3.7074e-05, ..., 5.6252e-06, + -5.5656e-03, -7.5579e-04], + ..., + [ 3.8004e-04, 1.4901e-05, 6.9439e-05, ..., 6.8769e-06, + -8.5592e-04, 6.3562e-04], + [ 2.2068e-03, 3.9196e-04, 9.1434e-05, ..., 2.7612e-05, + 2.4090e-03, -2.4819e-04], + [-7.2212e-03, 5.5850e-05, 4.5121e-05, ..., 2.1443e-05, + -1.4248e-03, 8.7976e-05]], device='cuda:0') +Epoch 189, bias, value: tensor([ 0.0122, -0.0036, -0.0013, -0.0184, 0.0123, -0.0064, -0.0158, -0.0230, + 0.0116, -0.0052], device='cuda:0'), grad: tensor([ 0.0164, 0.0218, -0.0384, 0.0224, 0.0374, -0.0580, 0.0331, -0.0080, + 0.0058, -0.0323], device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 216.32, cls_loss 0.5432 cls_loss_mapping 0.0105 cls_loss_causal 0.5123 re_mapping 0.0081 re_causal 0.0204 /// teacc 98.84 lr 0.00010000 +Epoch 190, weight, value: tensor([[ 0.0215, 0.0820, -0.1475, ..., -0.0827, 0.0602, 0.0019], + [-0.0741, -0.0872, 0.0780, ..., -0.0303, -0.0875, 0.0496], + [-0.0607, -0.0463, -0.0932, ..., -0.1082, -0.0279, 0.0435], + ..., + [-0.0161, -0.0864, -0.0965, ..., -0.1018, -0.0194, 0.0566], + [-0.0955, -0.0856, -0.0872, ..., -0.0568, -0.0527, -0.0776], + [ 0.0129, -0.0154, -0.0579, ..., -0.0573, 0.0295, -0.0154]], + device='cuda:0'), grad: tensor([[ 8.8692e-04, 5.2261e-04, 9.8896e-04, ..., 1.1044e-03, + -3.1494e-02, 5.5599e-04], + [-3.7789e-04, 1.5080e-04, -8.5728e-07, ..., 5.0497e-04, + 1.7917e-04, 2.7299e-04], + [ 2.8877e-03, 1.0262e-03, 1.5516e-03, ..., 6.8760e-04, + 1.6079e-03, 5.1651e-03], + ..., + [-8.6451e-04, 2.6655e-04, 3.9601e-04, ..., 1.7190e-04, + 1.1673e-03, -6.2828e-03], + [ 1.0643e-03, -3.7651e-03, -8.8272e-03, ..., -1.7862e-03, + 2.8038e-03, 2.6321e-04], + [ 2.6054e-03, 1.0500e-03, 1.0157e-03, ..., 2.4152e-04, + 2.0027e-03, 1.5774e-03]], device='cuda:0') +Epoch 190, bias, value: tensor([ 0.0121, -0.0043, -0.0006, -0.0178, 0.0113, -0.0057, -0.0163, -0.0234, + 0.0110, -0.0043], device='cuda:0'), grad: tensor([-0.0098, -0.0167, 0.0384, -0.0152, -0.0044, 0.0511, 0.0152, -0.0007, + -0.0215, -0.0363], device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 216.29, cls_loss 0.5715 cls_loss_mapping 0.0066 cls_loss_causal 0.5442 re_mapping 0.0078 re_causal 0.0208 /// teacc 98.83 lr 0.00010000 +Epoch 191, weight, value: tensor([[ 0.0223, 0.0825, -0.1468, ..., -0.0832, 0.0616, 0.0025], + [-0.0753, -0.0868, 0.0771, ..., -0.0307, -0.0879, 0.0472], + [-0.0618, -0.0463, -0.0927, ..., -0.1091, -0.0280, 0.0430], + ..., + [-0.0162, -0.0864, -0.0971, ..., -0.1022, -0.0189, 0.0593], + [-0.0963, -0.0857, -0.0874, ..., -0.0570, -0.0536, -0.0769], + [ 0.0126, -0.0156, -0.0581, ..., -0.0582, 0.0289, -0.0167]], + device='cuda:0'), grad: tensor([[-8.3618e-03, 5.6893e-05, 3.8147e-05, ..., -1.0691e-03, + -7.6714e-03, -3.9902e-03], + [ 2.9354e-03, 5.1230e-05, 2.0564e-05, ..., 1.1712e-04, + 2.1152e-03, 1.8768e-03], + [-1.0101e-02, -8.8453e-04, -3.0327e-04, ..., -3.0499e-03, + -8.0872e-03, -2.1515e-03], + ..., + [ 1.2608e-03, 1.1802e-04, 1.4472e-04, ..., 1.4162e-04, + 4.1351e-03, 5.7487e-03], + [ 3.2578e-03, 8.2791e-05, 9.5606e-05, ..., 4.2272e-04, + 2.5959e-03, 1.9064e-03], + [-3.6163e-03, -4.2796e-04, -1.6956e-03, ..., 2.4116e-04, + -4.5280e-03, -8.8043e-03]], device='cuda:0') +Epoch 191, bias, value: tensor([ 0.0119, -0.0056, -0.0008, -0.0182, 0.0117, -0.0062, -0.0152, -0.0225, + 0.0114, -0.0042], device='cuda:0'), grad: tensor([-0.0256, 0.0242, -0.1035, 0.0316, 0.0057, 0.0273, 0.0137, 0.0156, + 0.0273, -0.0164], device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 216.48, cls_loss 0.5425 cls_loss_mapping 0.0081 cls_loss_causal 0.5218 re_mapping 0.0084 re_causal 0.0218 /// teacc 98.76 lr 0.00010000 +Epoch 192, weight, value: tensor([[ 0.0233, 0.0835, -0.1458, ..., -0.0828, 0.0622, 0.0029], + [-0.0758, -0.0860, 0.0777, ..., -0.0306, -0.0889, 0.0478], + [-0.0611, -0.0476, -0.0934, ..., -0.1099, -0.0272, 0.0431], + ..., + [-0.0160, -0.0873, -0.0989, ..., -0.1022, -0.0201, 0.0588], + [-0.0963, -0.0848, -0.0871, ..., -0.0575, -0.0533, -0.0766], + [ 0.0123, -0.0158, -0.0577, ..., -0.0585, 0.0284, -0.0165]], + device='cuda:0'), grad: tensor([[ 3.1490e-03, 1.4343e-03, 5.5456e-04, ..., 5.1355e-04, + 2.7924e-03, 3.1614e-04], + [ 1.9646e-03, 1.7059e-04, 5.6028e-05, ..., 3.5763e-05, + 1.0386e-03, -1.5144e-03], + [ 2.0676e-03, 5.7316e-04, 2.9039e-04, ..., 3.4690e-04, + 1.9932e-03, 4.8184e-04], + ..., + [ 2.0504e-03, 1.7338e-03, 4.3702e-04, ..., 1.9483e-06, + 2.4662e-03, 6.0272e-04], + [-1.3342e-03, 3.6297e-03, 2.8973e-03, ..., 5.0011e-03, + 5.6534e-03, -7.2575e-04], + [-1.0025e-02, -6.2027e-03, -1.5593e-03, ..., 4.5188e-06, + -9.7351e-03, -8.6975e-04]], device='cuda:0') +Epoch 192, bias, value: tensor([ 0.0131, -0.0054, -0.0005, -0.0189, 0.0117, -0.0067, -0.0151, -0.0230, + 0.0114, -0.0041], device='cuda:0'), grad: tensor([ 0.0248, 0.0020, 0.0007, 0.0018, -0.0112, 0.0326, 0.0169, 0.0178, + 0.0022, -0.0877], device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 216.53, cls_loss 0.5395 cls_loss_mapping 0.0110 cls_loss_causal 0.5086 re_mapping 0.0086 re_causal 0.0215 /// teacc 98.55 lr 0.00010000 +Epoch 193, weight, value: tensor([[ 0.0231, 0.0833, -0.1472, ..., -0.0843, 0.0621, 0.0039], + [-0.0761, -0.0848, 0.0780, ..., -0.0312, -0.0887, 0.0490], + [-0.0604, -0.0486, -0.0943, ..., -0.1106, -0.0258, 0.0436], + ..., + [-0.0158, -0.0886, -0.0994, ..., -0.1024, -0.0198, 0.0592], + [-0.0967, -0.0843, -0.0864, ..., -0.0559, -0.0538, -0.0779], + [ 0.0123, -0.0177, -0.0591, ..., -0.0597, 0.0278, -0.0176]], + device='cuda:0'), grad: tensor([[-7.2250e-03, -9.1457e-04, 5.9366e-05, ..., -8.5602e-03, + -3.6812e-03, -4.2605e-04], + [-3.8433e-03, 1.5414e-04, -2.9640e-03, ..., 2.8253e-04, + -3.8934e-04, -4.8256e-03], + [-4.1084e-03, -1.8167e-03, 5.1975e-04, ..., -3.2063e-03, + -2.0771e-03, 1.2894e-03], + ..., + [-5.0402e-04, 5.6839e-04, -7.0333e-05, ..., 4.2033e-04, + -3.3879e-04, -2.0008e-03], + [ 4.5738e-03, 2.8496e-03, 5.7220e-04, ..., 8.1491e-04, + 3.0022e-03, 1.9484e-03], + [ 3.9787e-03, 1.8728e-04, 4.4537e-04, ..., 5.8317e-04, + 1.8187e-03, 1.1482e-03]], device='cuda:0') +Epoch 193, bias, value: tensor([ 0.0123, -0.0052, -0.0004, -0.0197, 0.0126, -0.0066, -0.0148, -0.0219, + 0.0111, -0.0051], device='cuda:0'), grad: tensor([-0.0361, -0.0452, 0.0195, 0.0136, 0.0420, 0.0199, -0.0229, -0.0381, + 0.0246, 0.0228], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 192---------------------------------------------------- +epoch 192, time 216.77, cls_loss 0.5076 cls_loss_mapping 0.0099 cls_loss_causal 0.4810 re_mapping 0.0089 re_causal 0.0221 /// teacc 98.89 lr 0.00010000 +Epoch 194, weight, value: tensor([[ 0.0244, 0.0839, -0.1466, ..., -0.0837, 0.0627, 0.0044], + [-0.0771, -0.0851, 0.0784, ..., -0.0315, -0.0895, 0.0493], + [-0.0596, -0.0485, -0.0942, ..., -0.1106, -0.0244, 0.0438], + ..., + [-0.0167, -0.0890, -0.1002, ..., -0.1030, -0.0210, 0.0587], + [-0.0988, -0.0849, -0.0857, ..., -0.0554, -0.0543, -0.0784], + [ 0.0110, -0.0165, -0.0600, ..., -0.0602, 0.0276, -0.0178]], + device='cuda:0'), grad: tensor([[ 2.2392e-03, 5.2378e-06, 3.1441e-06, ..., 3.7581e-05, + 1.0710e-03, 9.4771e-06], + [-1.9054e-03, 2.9802e-07, -4.7714e-05, ..., -2.9063e-04, + -9.6273e-04, -7.6175e-05], + [ 9.6846e-04, -5.4777e-05, 1.1347e-05, ..., 2.9147e-05, + 6.8426e-04, 6.8992e-06], + ..., + [ 1.7691e-03, 1.2927e-05, 8.6799e-06, ..., 3.4392e-05, + 7.8011e-04, 7.7188e-06], + [-4.3373e-03, 2.1458e-05, 6.3360e-05, ..., 3.1620e-05, + -1.5955e-03, 1.6410e-06], + [ 1.1320e-03, 1.5661e-05, 1.9237e-05, ..., 2.9311e-05, + 4.8923e-04, 1.0282e-05]], device='cuda:0') +Epoch 194, bias, value: tensor([ 0.0124, -0.0052, -0.0007, -0.0200, 0.0135, -0.0058, -0.0146, -0.0226, + 0.0106, -0.0051], device='cuda:0'), grad: tensor([ 0.0115, -0.0137, 0.0126, -0.0152, 0.0058, -0.0208, 0.0161, 0.0102, + -0.0167, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 216.17, cls_loss 0.4969 cls_loss_mapping 0.0077 cls_loss_causal 0.4689 re_mapping 0.0083 re_causal 0.0201 /// teacc 98.64 lr 0.00010000 +Epoch 195, weight, value: tensor([[ 0.0235, 0.0838, -0.1469, ..., -0.0834, 0.0622, 0.0048], + [-0.0763, -0.0858, 0.0796, ..., -0.0314, -0.0893, 0.0493], + [-0.0594, -0.0498, -0.0943, ..., -0.1097, -0.0250, 0.0438], + ..., + [-0.0164, -0.0898, -0.0994, ..., -0.1027, -0.0195, 0.0591], + [-0.0992, -0.0841, -0.0866, ..., -0.0548, -0.0538, -0.0784], + [ 0.0120, -0.0174, -0.0612, ..., -0.0608, 0.0278, -0.0188]], + device='cuda:0'), grad: tensor([[ 1.1835e-03, -1.0151e-06, 1.0931e-04, ..., 1.7369e-04, + 1.2445e-03, 4.4227e-04], + [ 1.5450e-03, 3.8669e-06, 2.7180e-04, ..., -5.3024e-04, + 9.0694e-04, -9.5272e-04], + [ 1.3323e-03, 1.0364e-05, 1.2171e-04, ..., 2.3830e-04, + 1.0347e-03, 8.7309e-04], + ..., + [ 9.2459e-04, 5.0552e-06, 1.2374e-04, ..., 6.6042e-05, + 5.1308e-04, -1.6665e-04], + [ 1.0614e-03, 1.0240e-04, 7.1764e-04, ..., 8.5688e-04, + 9.3269e-04, 4.3678e-04], + [ 2.3861e-03, 1.4372e-05, 9.2566e-05, ..., 4.2558e-05, + 1.6642e-03, 1.8549e-04]], device='cuda:0') +Epoch 195, bias, value: tensor([ 0.0126, -0.0041, -0.0008, -0.0197, 0.0123, -0.0062, -0.0152, -0.0223, + 0.0097, -0.0039], device='cuda:0'), grad: tensor([-0.0152, 0.0046, 0.0154, -0.0117, -0.0606, 0.0069, 0.0184, 0.0087, + 0.0187, 0.0148], device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 216.32, cls_loss 0.5397 cls_loss_mapping 0.0075 cls_loss_causal 0.5202 re_mapping 0.0082 re_causal 0.0207 /// teacc 98.71 lr 0.00010000 +Epoch 196, weight, value: tensor([[ 0.0239, 0.0844, -0.1470, ..., -0.0841, 0.0616, 0.0044], + [-0.0766, -0.0863, 0.0804, ..., -0.0312, -0.0894, 0.0502], + [-0.0596, -0.0492, -0.0944, ..., -0.1104, -0.0244, 0.0433], + ..., + [-0.0163, -0.0894, -0.1001, ..., -0.1033, -0.0200, 0.0579], + [-0.0999, -0.0852, -0.0865, ..., -0.0550, -0.0539, -0.0784], + [ 0.0125, -0.0182, -0.0624, ..., -0.0597, 0.0279, -0.0168]], + device='cuda:0'), grad: tensor([[-8.8425e-03, -5.2825e-06, 9.2685e-06, ..., 6.2656e-04, + -5.5199e-03, 4.3440e-04], + [-5.3673e-03, -6.3218e-06, -1.0306e-04, ..., -4.7050e-06, + -1.5574e-03, -1.7138e-03], + [-1.0419e-04, 1.9744e-06, 1.0230e-05, ..., 2.9016e-04, + 5.0926e-04, -9.5510e-04], + ..., + [ 2.7828e-03, 1.8049e-06, 1.5512e-05, ..., 4.3058e-04, + 1.0796e-03, 3.7622e-04], + [ 2.6321e-04, 1.4283e-05, 3.9071e-05, ..., -2.9125e-03, + 3.3331e-04, 4.6158e-04], + [ 2.6970e-03, 6.0387e-06, 2.2426e-05, ..., 3.9124e-04, + 1.1501e-03, 6.1846e-04]], device='cuda:0') +Epoch 196, bias, value: tensor([ 0.0128, -0.0041, -0.0006, -0.0198, 0.0128, -0.0064, -0.0154, -0.0219, + 0.0094, -0.0043], device='cuda:0'), grad: tensor([-0.0085, -0.0422, -0.0129, 0.0200, -0.0107, 0.0146, 0.0174, 0.0204, + -0.0168, 0.0186], device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 216.54, cls_loss 0.5428 cls_loss_mapping 0.0063 cls_loss_causal 0.5180 re_mapping 0.0077 re_causal 0.0206 /// teacc 98.71 lr 0.00010000 +Epoch 197, weight, value: tensor([[ 0.0247, 0.0835, -0.1480, ..., -0.0846, 0.0623, 0.0026], + [-0.0775, -0.0863, 0.0811, ..., -0.0307, -0.0899, 0.0495], + [-0.0595, -0.0475, -0.0933, ..., -0.1098, -0.0242, 0.0430], + ..., + [-0.0162, -0.0904, -0.1010, ..., -0.1034, -0.0202, 0.0589], + [-0.0982, -0.0857, -0.0875, ..., -0.0551, -0.0528, -0.0784], + [ 0.0119, -0.0203, -0.0631, ..., -0.0602, 0.0265, -0.0165]], + device='cuda:0'), grad: tensor([[ 0.0029, 0.0003, 0.0004, ..., 0.0003, 0.0039, 0.0007], + [ 0.0010, 0.0003, 0.0004, ..., 0.0001, 0.0008, 0.0001], + [-0.0155, -0.0043, -0.0043, ..., -0.0018, -0.0243, -0.0046], + ..., + [-0.0036, -0.0013, -0.0029, ..., -0.0013, -0.0026, -0.0023], + [ 0.0018, 0.0005, 0.0006, ..., 0.0005, 0.0014, 0.0010], + [-0.0015, 0.0001, 0.0003, ..., 0.0002, 0.0034, 0.0005]], + device='cuda:0') +Epoch 197, bias, value: tensor([ 0.0119, -0.0053, 0.0004, -0.0193, 0.0127, -0.0061, -0.0150, -0.0223, + 0.0102, -0.0049], device='cuda:0'), grad: tensor([ 0.0144, -0.0260, -0.0858, 0.0381, 0.0292, 0.0165, 0.0130, -0.0202, + 0.0187, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 216.44, cls_loss 0.5159 cls_loss_mapping 0.0072 cls_loss_causal 0.4893 re_mapping 0.0084 re_causal 0.0220 /// teacc 98.78 lr 0.00010000 +Epoch 198, weight, value: tensor([[ 0.0243, 0.0826, -0.1489, ..., -0.0838, 0.0623, 0.0034], + [-0.0781, -0.0859, 0.0817, ..., -0.0317, -0.0909, 0.0495], + [-0.0595, -0.0478, -0.0938, ..., -0.1107, -0.0241, 0.0419], + ..., + [-0.0164, -0.0921, -0.1023, ..., -0.1031, -0.0209, 0.0603], + [-0.0984, -0.0855, -0.0871, ..., -0.0556, -0.0526, -0.0776], + [ 0.0128, -0.0189, -0.0622, ..., -0.0599, 0.0264, -0.0169]], + device='cuda:0'), grad: tensor([[-0.0012, -0.0035, 0.0003, ..., 0.0005, 0.0019, 0.0005], + [ 0.0010, 0.0002, 0.0004, ..., 0.0006, 0.0022, 0.0016], + [ 0.0015, 0.0011, 0.0003, ..., 0.0005, 0.0008, 0.0008], + ..., + [-0.0020, 0.0012, 0.0004, ..., -0.0011, -0.0002, -0.0038], + [ 0.0026, 0.0016, -0.0020, ..., 0.0010, 0.0010, 0.0006], + [ 0.0031, 0.0050, 0.0022, ..., -0.0008, -0.0005, -0.0013]], + device='cuda:0') +Epoch 198, bias, value: tensor([ 0.0114, -0.0054, -0.0003, -0.0195, 0.0131, -0.0066, -0.0140, -0.0223, + 0.0101, -0.0044], device='cuda:0'), grad: tensor([ 0.0214, 0.0271, 0.0140, -0.0228, -0.0140, -0.0126, 0.0025, -0.0123, + 0.0119, -0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 216.11, cls_loss 0.5633 cls_loss_mapping 0.0071 cls_loss_causal 0.5337 re_mapping 0.0079 re_causal 0.0214 /// teacc 98.83 lr 0.00010000 +Epoch 199, weight, value: tensor([[ 0.0241, 0.0835, -0.1491, ..., -0.0837, 0.0623, 0.0030], + [-0.0773, -0.0849, 0.0813, ..., -0.0304, -0.0898, 0.0499], + [-0.0594, -0.0482, -0.0937, ..., -0.1117, -0.0241, 0.0422], + ..., + [-0.0178, -0.0929, -0.1021, ..., -0.1034, -0.0224, 0.0602], + [-0.0986, -0.0865, -0.0880, ..., -0.0555, -0.0537, -0.0774], + [ 0.0131, -0.0198, -0.0614, ..., -0.0603, 0.0257, -0.0185]], + device='cuda:0'), grad: tensor([[ 1.0185e-03, -2.8368e-06, 7.3552e-05, ..., 3.4642e-04, + 4.1676e-04, 6.0558e-04], + [ 1.2951e-03, -4.4894e-04, -1.1911e-03, ..., 8.9169e-04, + 6.4135e-04, -1.1971e-02], + [-2.5139e-03, 2.5332e-05, 1.3757e-04, ..., -5.4502e-04, + 7.0858e-04, -9.0218e-04], + ..., + [ 3.3150e-03, 2.1064e-04, 5.2404e-04, ..., 2.0134e-04, + 2.1973e-03, 8.8425e-03], + [ 1.7538e-03, -1.0195e-03, 1.4699e-04, ..., 4.5586e-04, + 8.9693e-04, 1.8444e-03], + [-6.8426e-04, 4.6897e-04, 2.4393e-05, ..., 2.8417e-05, + -5.3978e-04, 1.5411e-03]], device='cuda:0') +Epoch 199, bias, value: tensor([ 0.0105, -0.0041, -0.0008, -0.0189, 0.0129, -0.0055, -0.0145, -0.0229, + 0.0095, -0.0040], device='cuda:0'), grad: tensor([ 0.0114, -0.0518, -0.0096, 0.0013, 0.0460, -0.0416, 0.0044, 0.0202, + 0.0259, -0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 216.13, cls_loss 0.5154 cls_loss_mapping 0.0085 cls_loss_causal 0.4984 re_mapping 0.0081 re_causal 0.0208 /// teacc 98.77 lr 0.00010000 +Epoch 200, weight, value: tensor([[ 0.0234, 0.0832, -0.1498, ..., -0.0848, 0.0622, 0.0038], + [-0.0778, -0.0840, 0.0816, ..., -0.0312, -0.0907, 0.0498], + [-0.0593, -0.0470, -0.0939, ..., -0.1125, -0.0225, 0.0434], + ..., + [-0.0179, -0.0932, -0.1011, ..., -0.1030, -0.0221, 0.0598], + [-0.0977, -0.0862, -0.0887, ..., -0.0571, -0.0526, -0.0777], + [ 0.0133, -0.0201, -0.0630, ..., -0.0602, 0.0240, -0.0183]], + device='cuda:0'), grad: tensor([[-2.6150e-03, -1.8282e-03, 5.7667e-05, ..., 2.6107e-05, + -1.7595e-03, 3.4213e-04], + [-1.5163e-03, -6.4945e-04, -3.7694e-04, ..., -2.2817e-04, + -1.1616e-03, 8.7738e-05], + [ 3.4714e-04, 3.5439e-03, 2.0866e-03, ..., 3.1620e-05, + 1.7481e-03, 6.0797e-04], + ..., + [-2.6360e-03, 3.7217e-04, 2.9787e-05, ..., 1.5482e-05, + -4.4556e-03, 6.2084e-04], + [ 2.1362e-03, 1.7328e-03, 2.2256e-04, ..., 2.5794e-05, + 1.2331e-03, 7.3385e-04], + [ 7.6675e-03, 2.2926e-03, 2.9102e-05, ..., 1.8314e-05, + 5.8479e-03, 1.1654e-03]], device='cuda:0') +Epoch 200, bias, value: tensor([ 0.0112, -0.0039, -0.0006, -0.0190, 0.0125, -0.0049, -0.0148, -0.0236, + 0.0096, -0.0042], device='cuda:0'), grad: tensor([ 0.0084, -0.0345, 0.0044, 0.0322, 0.0044, 0.0066, -0.0393, -0.0019, + 0.0078, 0.0119], device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 216.25, cls_loss 0.5354 cls_loss_mapping 0.0090 cls_loss_causal 0.5104 re_mapping 0.0077 re_causal 0.0199 /// teacc 98.73 lr 0.00010000 +Epoch 201, weight, value: tensor([[ 0.0236, 0.0842, -0.1496, ..., -0.0841, 0.0621, 0.0060], + [-0.0776, -0.0860, 0.0807, ..., -0.0319, -0.0910, 0.0498], + [-0.0592, -0.0472, -0.0944, ..., -0.1130, -0.0224, 0.0433], + ..., + [-0.0184, -0.0936, -0.1019, ..., -0.1030, -0.0229, 0.0598], + [-0.0969, -0.0878, -0.0895, ..., -0.0578, -0.0520, -0.0783], + [ 0.0120, -0.0201, -0.0639, ..., -0.0610, 0.0233, -0.0186]], + device='cuda:0'), grad: tensor([[ 6.7902e-04, -5.3406e-04, -3.3545e-04, ..., -3.0398e-04, + 3.4475e-04, 1.5700e-04], + [ 1.3580e-03, 8.6689e-04, 2.3735e-04, ..., 1.7476e-04, + 2.0351e-03, -2.2137e-04], + [ 1.8358e-03, 1.0157e-03, 3.9959e-04, ..., 3.2234e-04, + 1.5659e-03, 2.4343e-04], + ..., + [-1.1444e-03, -1.1854e-03, 7.1573e-04, ..., 2.3425e-04, + -4.7951e-03, 1.7560e-04], + [ 9.3460e-03, 3.5820e-03, 1.6155e-03, ..., -5.2065e-05, + 1.9760e-03, 1.2529e-04], + [-9.5749e-03, 3.7193e-03, 4.2458e-03, ..., 5.5122e-04, + -2.0027e-03, -1.0548e-03]], device='cuda:0') +Epoch 201, bias, value: tensor([ 0.0118, -0.0040, -0.0009, -0.0185, 0.0133, -0.0047, -0.0150, -0.0235, + 0.0089, -0.0051], device='cuda:0'), grad: tensor([-0.0032, 0.0212, 0.0235, -0.0199, -0.0924, 0.0413, 0.0353, -0.0390, + 0.0491, -0.0161], device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 215.72, cls_loss 0.5316 cls_loss_mapping 0.0077 cls_loss_causal 0.5055 re_mapping 0.0080 re_causal 0.0199 /// teacc 98.78 lr 0.00010000 +Epoch 202, weight, value: tensor([[ 0.0239, 0.0842, -0.1509, ..., -0.0845, 0.0620, 0.0052], + [-0.0782, -0.0859, 0.0810, ..., -0.0325, -0.0916, 0.0482], + [-0.0598, -0.0480, -0.0946, ..., -0.1124, -0.0229, 0.0434], + ..., + [-0.0192, -0.0939, -0.1023, ..., -0.1027, -0.0228, 0.0604], + [-0.0972, -0.0881, -0.0906, ..., -0.0585, -0.0529, -0.0772], + [ 0.0135, -0.0211, -0.0655, ..., -0.0618, 0.0238, -0.0190]], + device='cuda:0'), grad: tensor([[ 2.9278e-03, 2.1088e-04, 1.1349e-04, ..., 5.5075e-04, + 2.3308e-03, 4.6325e-04], + [ 1.4830e-03, 6.2585e-05, 8.1122e-05, ..., 2.8595e-05, + 1.0033e-03, 5.3406e-04], + [-4.4785e-03, -4.1342e-04, 4.9412e-05, ..., -8.4996e-05, + -3.8795e-03, -2.4929e-03], + ..., + [-4.3774e-04, 8.9943e-05, 3.8087e-05, ..., 1.9908e-05, + -5.7745e-04, 1.4675e-04], + [ 2.1534e-03, 2.1076e-04, 9.7752e-05, ..., 4.0412e-04, + 1.3027e-03, 5.4502e-04], + [ 1.3008e-03, 3.0112e-04, 2.7156e-04, ..., 8.0049e-05, + 1.4067e-03, 5.1737e-04]], device='cuda:0') +Epoch 202, bias, value: tensor([ 0.0110, -0.0051, -0.0016, -0.0185, 0.0136, -0.0043, -0.0147, -0.0233, + 0.0092, -0.0039], device='cuda:0'), grad: tensor([ 0.0196, 0.0150, -0.0443, -0.0169, 0.0116, 0.0159, -0.0220, -0.0139, + 0.0191, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 216.37, cls_loss 0.5119 cls_loss_mapping 0.0059 cls_loss_causal 0.4828 re_mapping 0.0086 re_causal 0.0223 /// teacc 98.56 lr 0.00010000 +Epoch 203, weight, value: tensor([[ 0.0230, 0.0839, -0.1508, ..., -0.0859, 0.0619, 0.0041], + [-0.0770, -0.0855, 0.0816, ..., -0.0327, -0.0907, 0.0488], + [-0.0595, -0.0486, -0.0953, ..., -0.1126, -0.0229, 0.0436], + ..., + [-0.0198, -0.0949, -0.1036, ..., -0.1032, -0.0241, 0.0610], + [-0.0979, -0.0885, -0.0892, ..., -0.0576, -0.0530, -0.0779], + [ 0.0150, -0.0215, -0.0655, ..., -0.0619, 0.0252, -0.0205]], + device='cuda:0'), grad: tensor([[-0.0030, -0.0082, -0.0020, ..., -0.0032, -0.0127, 0.0004], + [-0.0037, 0.0003, 0.0007, ..., 0.0003, 0.0028, 0.0015], + [ 0.0004, 0.0006, 0.0008, ..., 0.0003, 0.0002, 0.0003], + ..., + [ 0.0045, 0.0008, 0.0023, ..., 0.0001, 0.0018, 0.0023], + [ 0.0027, 0.0005, 0.0006, ..., 0.0003, 0.0020, 0.0006], + [ 0.0080, 0.0014, -0.0014, ..., 0.0006, 0.0019, -0.0020]], + device='cuda:0') +Epoch 203, bias, value: tensor([ 0.0104, -0.0040, -0.0016, -0.0172, 0.0132, -0.0055, -0.0147, -0.0240, + 0.0094, -0.0039], device='cuda:0'), grad: tensor([-0.0363, 0.0109, -0.0050, -0.0884, 0.0036, 0.0229, 0.0032, 0.0406, + 0.0273, 0.0213], device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 216.20, cls_loss 0.5326 cls_loss_mapping 0.0080 cls_loss_causal 0.5074 re_mapping 0.0080 re_causal 0.0215 /// teacc 98.67 lr 0.00010000 +Epoch 204, weight, value: tensor([[ 0.0238, 0.0853, -0.1510, ..., -0.0858, 0.0625, 0.0037], + [-0.0765, -0.0871, 0.0816, ..., -0.0324, -0.0906, 0.0491], + [-0.0597, -0.0474, -0.0935, ..., -0.1124, -0.0214, 0.0447], + ..., + [-0.0186, -0.0951, -0.1059, ..., -0.1037, -0.0240, 0.0606], + [-0.0992, -0.0893, -0.0888, ..., -0.0582, -0.0534, -0.0789], + [ 0.0138, -0.0214, -0.0650, ..., -0.0619, 0.0243, -0.0200]], + device='cuda:0'), grad: tensor([[-3.6407e-04, -5.5790e-04, 3.2615e-06, ..., -2.6951e-03, + 3.7432e-04, -1.5207e-05], + [-1.5507e-03, 1.0971e-06, -2.6792e-05, ..., 2.2984e-04, + -5.0879e-04, -6.5994e-03], + [ 5.4312e-04, 9.2834e-06, 1.2398e-05, ..., 2.7800e-04, + 7.0333e-04, 3.3436e-03], + ..., + [-3.3665e-03, 1.7688e-05, 2.3842e-05, ..., -1.6842e-03, + -1.5030e-03, -6.5994e-04], + [ 5.5456e-04, 1.0654e-05, 1.8209e-05, ..., 1.6904e-04, + 5.9319e-04, -2.2011e-03], + [ 1.8339e-03, 8.4937e-06, 1.1049e-05, ..., 2.1875e-04, + 1.0834e-03, 2.1534e-03]], device='cuda:0') +Epoch 204, bias, value: tensor([ 0.0103, -0.0036, -0.0008, -0.0179, 0.0118, -0.0056, -0.0143, -0.0228, + 0.0088, -0.0038], device='cuda:0'), grad: tensor([ 0.0081, -0.0510, 0.0271, 0.0271, 0.0013, 0.0203, -0.0059, -0.0301, + -0.0215, 0.0245], device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 216.19, cls_loss 0.5100 cls_loss_mapping 0.0067 cls_loss_causal 0.4799 re_mapping 0.0080 re_causal 0.0212 /// teacc 98.68 lr 0.00010000 +Epoch 205, weight, value: tensor([[ 0.0241, 0.0858, -0.1514, ..., -0.0861, 0.0627, 0.0041], + [-0.0761, -0.0866, 0.0824, ..., -0.0328, -0.0910, 0.0496], + [-0.0593, -0.0461, -0.0935, ..., -0.1124, -0.0220, 0.0440], + ..., + [-0.0185, -0.0969, -0.1069, ..., -0.1051, -0.0243, 0.0611], + [-0.0996, -0.0905, -0.0888, ..., -0.0584, -0.0545, -0.0793], + [ 0.0138, -0.0215, -0.0649, ..., -0.0619, 0.0253, -0.0190]], + device='cuda:0'), grad: tensor([[-1.4591e-03, -3.2101e-03, 5.5122e-04, ..., -9.1028e-04, + -2.4357e-03, 1.2035e-03], + [ 1.8463e-03, 1.0949e-04, 1.1492e-03, ..., 5.4359e-04, + 1.2093e-03, 2.4204e-03], + [-1.3962e-03, 5.7068e-03, 6.1951e-03, ..., 1.4582e-03, + -7.5960e-04, 1.2417e-03], + ..., + [ 7.5626e-04, 9.4831e-05, 3.7360e-04, ..., 2.2900e-04, + 5.4979e-04, 1.0042e-03], + [ 1.3580e-03, 6.0129e-04, 1.1272e-03, ..., 5.2309e-04, + 9.7322e-04, 1.5059e-03], + [ 9.1648e-04, 4.6301e-04, 3.3045e-04, ..., 3.8147e-04, + 8.1778e-04, 8.1205e-04]], device='cuda:0') +Epoch 205, bias, value: tensor([ 0.0114, -0.0033, -0.0005, -0.0188, 0.0126, -0.0067, -0.0145, -0.0236, + 0.0089, -0.0034], device='cuda:0'), grad: tensor([ 0.0036, 0.0500, -0.0030, -0.0295, -0.0490, -0.0210, 0.0002, 0.0146, + 0.0210, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 216.13, cls_loss 0.5076 cls_loss_mapping 0.0093 cls_loss_causal 0.4834 re_mapping 0.0077 re_causal 0.0199 /// teacc 98.56 lr 0.00010000 +Epoch 206, weight, value: tensor([[ 0.0237, 0.0861, -0.1520, ..., -0.0850, 0.0626, 0.0036], + [-0.0764, -0.0857, 0.0829, ..., -0.0332, -0.0901, 0.0498], + [-0.0596, -0.0473, -0.0943, ..., -0.1135, -0.0221, 0.0434], + ..., + [-0.0185, -0.0973, -0.1082, ..., -0.1050, -0.0250, 0.0615], + [-0.1001, -0.0912, -0.0890, ..., -0.0584, -0.0555, -0.0785], + [ 0.0132, -0.0196, -0.0634, ..., -0.0624, 0.0252, -0.0192]], + device='cuda:0'), grad: tensor([[-9.3079e-04, -3.1223e-03, -4.8566e-04, ..., -1.3046e-03, + -1.7624e-03, 5.5742e-04], + [ 1.4906e-03, 1.6943e-05, 3.6678e-03, ..., 1.3161e-04, + 1.5011e-03, 4.8561e-03], + [ 6.0940e-04, 1.1253e-04, 1.2058e-04, ..., 5.1945e-05, + 2.8300e-04, 1.0052e-03], + ..., + [ 9.2745e-04, 1.2346e-05, 3.5614e-05, ..., 6.3181e-06, + 2.3437e-04, 1.3876e-03], + [-1.5488e-03, 3.1590e-04, 5.5885e-04, ..., 1.4579e-04, + 2.5725e-04, -3.3264e-03], + [ 6.5470e-04, 1.6320e-04, 2.7919e-04, ..., 7.4923e-05, + 1.6749e-04, 8.8739e-04]], device='cuda:0') +Epoch 206, bias, value: tensor([ 0.0111, -0.0050, 0.0009, -0.0181, 0.0122, -0.0057, -0.0149, -0.0233, + 0.0086, -0.0036], device='cuda:0'), grad: tensor([ 0.0015, 0.0101, 0.0136, 0.0147, -0.0139, -0.0238, 0.0083, 0.0191, + -0.0418, 0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 216.06, cls_loss 0.5170 cls_loss_mapping 0.0062 cls_loss_causal 0.4891 re_mapping 0.0078 re_causal 0.0203 /// teacc 98.82 lr 0.00010000 +Epoch 207, weight, value: tensor([[ 0.0234, 0.0870, -0.1517, ..., -0.0850, 0.0621, 0.0032], + [-0.0770, -0.0862, 0.0822, ..., -0.0339, -0.0906, 0.0491], + [-0.0594, -0.0475, -0.0944, ..., -0.1129, -0.0223, 0.0435], + ..., + [-0.0188, -0.0963, -0.1091, ..., -0.1064, -0.0261, 0.0604], + [-0.0990, -0.0908, -0.0897, ..., -0.0583, -0.0555, -0.0774], + [ 0.0139, -0.0206, -0.0629, ..., -0.0624, 0.0254, -0.0174]], + device='cuda:0'), grad: tensor([[-6.4325e-04, 1.1530e-03, 7.1859e-04, ..., 1.3493e-05, + -4.5586e-03, 6.6042e-04], + [ 1.1730e-03, 2.1651e-05, 7.9274e-06, ..., 3.1471e-05, + 1.1053e-03, 7.1192e-04], + [ 4.2038e-03, 2.0809e-03, 1.1959e-03, ..., 4.2915e-06, + 4.5891e-03, 1.1797e-03], + ..., + [-1.2035e-03, 1.3363e-04, 8.0407e-05, ..., 8.7544e-08, + -5.0831e-04, -1.7433e-03], + [-5.2338e-03, 1.1997e-03, 5.6314e-04, ..., 1.3363e-04, + -5.4016e-03, -1.6050e-03], + [-1.0437e-02, -1.0185e-02, -6.3324e-03, ..., 6.9663e-06, + -5.5275e-03, -2.1896e-03]], device='cuda:0') +Epoch 207, bias, value: tensor([ 0.0114, -0.0054, 0.0006, -0.0185, 0.0122, -0.0051, -0.0145, -0.0254, + 0.0096, -0.0026], device='cuda:0'), grad: tensor([-0.0140, 0.0122, 0.0311, 0.0428, 0.0194, -0.0269, 0.0431, -0.0205, + -0.0461, -0.0410], device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 216.11, cls_loss 0.4985 cls_loss_mapping 0.0086 cls_loss_causal 0.4712 re_mapping 0.0077 re_causal 0.0191 /// teacc 98.73 lr 0.00010000 +Epoch 208, weight, value: tensor([[ 0.0234, 0.0862, -0.1515, ..., -0.0852, 0.0622, 0.0030], + [-0.0766, -0.0861, 0.0821, ..., -0.0336, -0.0901, 0.0486], + [-0.0595, -0.0488, -0.0959, ..., -0.1134, -0.0223, 0.0448], + ..., + [-0.0191, -0.0961, -0.1092, ..., -0.1076, -0.0260, 0.0607], + [-0.0994, -0.0916, -0.0898, ..., -0.0594, -0.0559, -0.0776], + [ 0.0139, -0.0211, -0.0636, ..., -0.0635, 0.0243, -0.0172]], + device='cuda:0'), grad: tensor([[-4.0627e-03, -5.5962e-03, -1.7319e-03, ..., -4.0627e-03, + -4.8332e-03, 1.2150e-03], + [ 8.5878e-04, 5.7125e-04, -9.6226e-04, ..., 4.0340e-04, + 1.2980e-03, 7.3290e-04], + [-4.8943e-03, 9.4700e-04, 1.0757e-03, ..., 1.2054e-03, + -5.3177e-03, -3.6907e-04], + ..., + [ 8.8787e-04, 2.9206e-04, 7.0667e-04, ..., -8.0615e-06, + 1.0681e-03, -2.6226e-03], + [-2.1763e-03, -6.2361e-06, -5.5742e-04, ..., -8.5068e-04, + -4.6425e-03, -7.8821e-04], + [ 1.9073e-03, 7.2813e-04, 1.0977e-03, ..., 4.8137e-04, + 2.8400e-03, 1.5345e-03]], device='cuda:0') +Epoch 208, bias, value: tensor([ 0.0119, -0.0048, 0.0001, -0.0187, 0.0123, -0.0051, -0.0155, -0.0254, + 0.0098, -0.0024], device='cuda:0'), grad: tensor([-0.0267, 0.0070, -0.0107, -0.0355, 0.0232, 0.0287, 0.0237, -0.0181, + -0.0115, 0.0199], device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 216.01, cls_loss 0.5293 cls_loss_mapping 0.0074 cls_loss_causal 0.5054 re_mapping 0.0079 re_causal 0.0213 /// teacc 98.81 lr 0.00010000 +Epoch 209, weight, value: tensor([[ 0.0230, 0.0855, -0.1519, ..., -0.0850, 0.0625, 0.0014], + [-0.0767, -0.0866, 0.0824, ..., -0.0322, -0.0907, 0.0489], + [-0.0591, -0.0491, -0.0963, ..., -0.1136, -0.0218, 0.0444], + ..., + [-0.0196, -0.0959, -0.1098, ..., -0.1061, -0.0276, 0.0607], + [-0.1000, -0.0912, -0.0902, ..., -0.0577, -0.0560, -0.0780], + [ 0.0131, -0.0211, -0.0644, ..., -0.0638, 0.0242, -0.0155]], + device='cuda:0'), grad: tensor([[-1.2989e-03, -9.9763e-06, 1.7953e-04, ..., 1.0246e-04, + -8.5545e-04, 4.3583e-04], + [-1.8778e-03, 1.8496e-06, 1.4710e-04, ..., 1.8805e-05, + 7.6532e-05, -4.3640e-03], + [ 1.5192e-03, 1.2897e-05, 1.1551e-04, ..., 2.2054e-05, + 9.2649e-04, 6.2323e-04], + ..., + [-1.9331e-03, -5.3024e-04, 1.2219e-04, ..., 2.4475e-06, + -2.6665e-03, -6.2866e-03], + [ 2.4223e-03, 2.3887e-05, 2.2912e-04, ..., 1.4138e-04, + 1.3838e-03, 7.4577e-04], + [ 2.4452e-03, 4.4346e-04, 1.6785e-04, ..., 1.5810e-05, + 2.2793e-03, 5.2071e-03]], device='cuda:0') +Epoch 209, bias, value: tensor([ 0.0114, -0.0053, 0.0003, -0.0186, 0.0124, -0.0043, -0.0155, -0.0260, + 0.0099, -0.0019], device='cuda:0'), grad: tensor([-0.0449, -0.0033, 0.0215, 0.0312, -0.0064, 0.0303, -0.0546, -0.0173, + 0.0131, 0.0305], device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 216.16, cls_loss 0.5119 cls_loss_mapping 0.0058 cls_loss_causal 0.4840 re_mapping 0.0077 re_causal 0.0198 /// teacc 98.67 lr 0.00010000 +Epoch 210, weight, value: tensor([[ 0.0230, 0.0859, -0.1524, ..., -0.0853, 0.0622, 0.0028], + [-0.0765, -0.0850, 0.0834, ..., -0.0324, -0.0901, 0.0490], + [-0.0582, -0.0489, -0.0960, ..., -0.1148, -0.0209, 0.0439], + ..., + [-0.0195, -0.0969, -0.1092, ..., -0.1060, -0.0270, 0.0607], + [-0.1004, -0.0916, -0.0912, ..., -0.0573, -0.0571, -0.0777], + [ 0.0133, -0.0217, -0.0639, ..., -0.0632, 0.0244, -0.0148]], + device='cuda:0'), grad: tensor([[0.0025, 0.0037, 0.0011, ..., 0.0073, 0.0073, 0.0035], + [0.0021, 0.0001, 0.0006, ..., 0.0003, 0.0017, 0.0022], + [0.0034, 0.0006, 0.0002, ..., 0.0002, 0.0005, 0.0012], + ..., + [0.0027, 0.0002, 0.0012, ..., 0.0005, 0.0019, 0.0011], + [0.0021, 0.0002, 0.0010, ..., 0.0008, 0.0016, 0.0020], + [0.0029, 0.0003, 0.0016, ..., 0.0006, 0.0021, 0.0020]], + device='cuda:0') +Epoch 210, bias, value: tensor([ 0.0108, -0.0051, 0.0008, -0.0178, 0.0127, -0.0059, -0.0156, -0.0253, + 0.0100, -0.0025], device='cuda:0'), grad: tensor([ 0.0406, 0.0224, 0.0255, -0.0403, -0.0738, -0.0118, -0.0209, 0.0182, + 0.0175, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 216.30, cls_loss 0.5527 cls_loss_mapping 0.0078 cls_loss_causal 0.5233 re_mapping 0.0075 re_causal 0.0196 /// teacc 98.72 lr 0.00010000 +Epoch 211, weight, value: tensor([[ 0.0234, 0.0859, -0.1537, ..., -0.0861, 0.0628, 0.0034], + [-0.0771, -0.0853, 0.0842, ..., -0.0326, -0.0910, 0.0496], + [-0.0596, -0.0486, -0.0967, ..., -0.1160, -0.0216, 0.0432], + ..., + [-0.0179, -0.0965, -0.1102, ..., -0.1063, -0.0266, 0.0617], + [-0.1009, -0.0912, -0.0913, ..., -0.0579, -0.0560, -0.0780], + [ 0.0121, -0.0216, -0.0629, ..., -0.0622, 0.0229, -0.0159]], + device='cuda:0'), grad: tensor([[ 2.2278e-03, 1.4615e-04, 1.0538e-03, ..., 3.2997e-04, + 7.5111e-03, 1.8606e-03], + [ 3.3569e-04, 1.2428e-05, 1.4467e-03, ..., 1.3530e-04, + 1.2436e-03, 7.2975e-03], + [ 1.3895e-03, 5.4449e-05, 6.4659e-04, ..., 1.8561e-04, + -1.5867e-04, 6.4392e-03], + ..., + [ 3.2387e-03, 3.5554e-05, 4.4918e-04, ..., 3.2902e-04, + 3.0918e-03, 2.7542e-03], + [ 2.0466e-03, 1.5306e-04, -1.9159e-03, ..., 2.1720e-04, + 4.5180e-04, -8.3237e-03], + [-4.7340e-03, -2.4021e-04, -1.4582e-03, ..., -2.4605e-03, + -6.9237e-04, -6.5470e-04]], device='cuda:0') +Epoch 211, bias, value: tensor([ 0.0109, -0.0046, 0.0008, -0.0184, 0.0126, -0.0065, -0.0144, -0.0247, + 0.0099, -0.0034], device='cuda:0'), grad: tensor([ 0.0637, 0.0412, 0.0231, -0.0022, -0.0695, 0.0294, -0.0823, 0.0402, + -0.0168, -0.0270], device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 216.22, cls_loss 0.5125 cls_loss_mapping 0.0072 cls_loss_causal 0.4806 re_mapping 0.0075 re_causal 0.0186 /// teacc 98.75 lr 0.00010000 +Epoch 212, weight, value: tensor([[ 0.0240, 0.0864, -0.1523, ..., -0.0850, 0.0633, 0.0045], + [-0.0774, -0.0858, 0.0830, ..., -0.0334, -0.0916, 0.0497], + [-0.0592, -0.0494, -0.0980, ..., -0.1160, -0.0215, 0.0435], + ..., + [-0.0190, -0.0970, -0.1084, ..., -0.1078, -0.0273, 0.0624], + [-0.0996, -0.0904, -0.0916, ..., -0.0581, -0.0542, -0.0774], + [ 0.0129, -0.0208, -0.0604, ..., -0.0622, 0.0227, -0.0176]], + device='cuda:0'), grad: tensor([[-6.9189e-04, 6.6876e-05, -1.0556e-04, ..., 4.7827e-04, + -6.0320e-04, -6.3944e-04], + [ 8.7595e-04, 2.5436e-05, -1.3304e-04, ..., 1.1879e-04, + 4.5633e-04, 2.4414e-03], + [ 1.0767e-03, 7.7307e-05, 2.6493e-03, ..., 3.7694e-04, + 2.4261e-03, -2.7866e-03], + ..., + [-3.6411e-03, -6.5446e-05, -7.3385e-04, ..., -3.3321e-03, + -3.2539e-03, -1.5771e-04], + [ 2.0847e-03, 5.8317e-04, -2.4872e-03, ..., 5.6171e-04, + -1.2064e-03, 8.8930e-04], + [-2.7504e-03, 1.0401e-04, -1.0705e-04, ..., 1.9526e-04, + -1.4210e-03, -7.3004e-04]], device='cuda:0') +Epoch 212, bias, value: tensor([ 0.0110, -0.0053, 0.0014, -0.0184, 0.0128, -0.0062, -0.0148, -0.0245, + 0.0103, -0.0040], device='cuda:0'), grad: tensor([-0.0168, 0.0082, 0.0208, -0.0133, 0.0179, -0.0035, 0.0140, -0.0126, + 0.0082, -0.0227], device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 216.56, cls_loss 0.4944 cls_loss_mapping 0.0076 cls_loss_causal 0.4680 re_mapping 0.0076 re_causal 0.0184 /// teacc 98.75 lr 0.00010000 +Epoch 213, weight, value: tensor([[ 0.0231, 0.0866, -0.1532, ..., -0.0853, 0.0624, 0.0038], + [-0.0776, -0.0859, 0.0838, ..., -0.0333, -0.0919, 0.0503], + [-0.0597, -0.0498, -0.0986, ..., -0.1158, -0.0209, 0.0447], + ..., + [-0.0198, -0.0979, -0.1087, ..., -0.1081, -0.0279, 0.0616], + [-0.0995, -0.0904, -0.0921, ..., -0.0572, -0.0544, -0.0782], + [ 0.0136, -0.0201, -0.0606, ..., -0.0625, 0.0242, -0.0186]], + device='cuda:0'), grad: tensor([[ 1.3971e-03, 3.3045e-04, 1.0300e-03, ..., 1.6427e-04, + 7.6199e-04, 8.1158e-04], + [-1.3387e-04, 6.4433e-05, 9.7215e-05, ..., 3.9101e-05, + 1.1653e-04, -7.5161e-05], + [ 3.4094e-04, 2.1708e-04, -6.0320e-04, ..., 6.0946e-05, + 3.4571e-04, -1.8942e-04], + ..., + [ 1.3952e-03, 1.0866e-04, 5.5552e-04, ..., 7.1704e-05, + 8.0156e-04, -5.2118e-04], + [ 3.8071e-03, 2.9907e-03, 3.0231e-03, ..., 2.3270e-03, + 2.5291e-03, 5.1022e-04], + [ 5.1842e-03, 6.6042e-04, 1.2007e-03, ..., 3.8266e-04, + 4.5967e-03, 7.5912e-04]], device='cuda:0') +Epoch 213, bias, value: tensor([ 0.0097, -0.0052, 0.0014, -0.0184, 0.0142, -0.0056, -0.0150, -0.0252, + 0.0099, -0.0037], device='cuda:0'), grad: tensor([ 0.0229, -0.0071, -0.0075, -0.0374, -0.0493, -0.0280, 0.0267, 0.0133, + 0.0352, 0.0312], device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 216.43, cls_loss 0.5266 cls_loss_mapping 0.0088 cls_loss_causal 0.4923 re_mapping 0.0079 re_causal 0.0214 /// teacc 98.70 lr 0.00010000 +Epoch 214, weight, value: tensor([[ 0.0233, 0.0865, -0.1531, ..., -0.0851, 0.0625, 0.0035], + [-0.0775, -0.0863, 0.0835, ..., -0.0338, -0.0927, 0.0496], + [-0.0605, -0.0502, -0.0987, ..., -0.1171, -0.0203, 0.0456], + ..., + [-0.0200, -0.0978, -0.1091, ..., -0.1091, -0.0290, 0.0619], + [-0.0986, -0.0899, -0.0910, ..., -0.0569, -0.0526, -0.0786], + [ 0.0140, -0.0206, -0.0607, ..., -0.0629, 0.0239, -0.0190]], + device='cuda:0'), grad: tensor([[ 3.8548e-03, 5.7459e-05, 4.1461e-04, ..., 2.2984e-03, + 5.2757e-03, 2.3675e-04], + [ 9.9945e-04, 6.6102e-05, 3.2115e-04, ..., 2.9281e-06, + 7.4387e-04, 4.7398e-04], + [-1.8587e-03, 1.8692e-04, -6.0892e-04, ..., 2.8033e-06, + -2.4281e-03, 3.9196e-04], + ..., + [ 5.0783e-05, 1.2851e-04, 7.5102e-04, ..., 6.3330e-08, + 1.0662e-03, -1.8654e-03], + [ 2.2621e-03, 3.7313e-04, 6.5613e-04, ..., 5.6058e-05, + 1.0014e-03, 1.0414e-03], + [ 1.6470e-03, 7.1764e-05, -9.0227e-06, ..., 6.0769e-07, + 1.1568e-03, -7.9489e-04]], device='cuda:0') +Epoch 214, bias, value: tensor([ 0.0098, -0.0058, 0.0017, -0.0184, 0.0138, -0.0057, -0.0152, -0.0252, + 0.0107, -0.0035], device='cuda:0'), grad: tensor([ 0.0248, -0.0127, -0.0398, -0.0211, -0.0172, 0.0154, -0.0313, 0.0016, + 0.0672, 0.0130], device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 216.64, cls_loss 0.5171 cls_loss_mapping 0.0057 cls_loss_causal 0.4911 re_mapping 0.0077 re_causal 0.0217 /// teacc 98.67 lr 0.00010000 +Epoch 215, weight, value: tensor([[ 0.0223, 0.0877, -0.1524, ..., -0.0862, 0.0624, 0.0032], + [-0.0776, -0.0863, 0.0843, ..., -0.0338, -0.0925, 0.0486], + [-0.0611, -0.0505, -0.0989, ..., -0.1172, -0.0207, 0.0466], + ..., + [-0.0192, -0.0987, -0.1106, ..., -0.1086, -0.0282, 0.0614], + [-0.0986, -0.0899, -0.0915, ..., -0.0576, -0.0528, -0.0793], + [ 0.0143, -0.0200, -0.0613, ..., -0.0633, 0.0241, -0.0183]], + device='cuda:0'), grad: tensor([[-4.2496e-03, -2.4624e-03, -2.3403e-03, ..., -7.1001e-04, + -3.0785e-03, -1.1797e-03], + [-1.4925e-03, 8.6427e-05, -7.1678e-03, ..., -6.2675e-03, + -3.1662e-03, -2.6340e-03], + [-3.4294e-03, 5.7077e-04, 2.1172e-04, ..., 1.1802e-05, + -4.6182e-04, -1.0445e-02], + ..., + [ 1.0557e-03, 2.9817e-05, -5.2643e-04, ..., 9.6738e-05, + 8.2195e-05, 4.7035e-03], + [ 2.1534e-03, 4.8375e-04, 7.6485e-04, ..., 4.6849e-04, + 1.2617e-03, 1.7118e-03], + [-1.2917e-02, 1.8883e-04, 5.5981e-04, ..., 1.0800e-04, + -4.9171e-03, 1.7090e-03]], device='cuda:0') +Epoch 215, bias, value: tensor([ 0.0094, -0.0059, 0.0013, -0.0184, 0.0136, -0.0063, -0.0142, -0.0246, + 0.0107, -0.0034], device='cuda:0'), grad: tensor([-0.0104, -0.0162, -0.0596, 0.0100, 0.0208, 0.0100, 0.0272, 0.0213, + 0.0226, -0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 216.51, cls_loss 0.5045 cls_loss_mapping 0.0069 cls_loss_causal 0.4832 re_mapping 0.0074 re_causal 0.0204 /// teacc 98.71 lr 0.00010000 +Epoch 216, weight, value: tensor([[ 0.0229, 0.0887, -0.1512, ..., -0.0860, 0.0621, 0.0041], + [-0.0767, -0.0871, 0.0837, ..., -0.0338, -0.0924, 0.0478], + [-0.0626, -0.0510, -0.0998, ..., -0.1171, -0.0211, 0.0463], + ..., + [-0.0185, -0.0987, -0.1099, ..., -0.1095, -0.0272, 0.0624], + [-0.1000, -0.0895, -0.0915, ..., -0.0581, -0.0538, -0.0804], + [ 0.0150, -0.0203, -0.0612, ..., -0.0628, 0.0241, -0.0177]], + device='cuda:0'), grad: tensor([[-9.7046e-03, 3.2902e-04, 3.7599e-04, ..., 2.4486e-04, + -1.2497e-02, 1.3626e-04], + [ 2.1706e-03, 4.6253e-05, 8.3685e-04, ..., 1.3649e-04, + 8.4639e-04, 1.3657e-03], + [-3.8223e-03, 1.9836e-04, -8.8358e-04, ..., 1.3924e-04, + -2.3651e-03, -1.1148e-03], + ..., + [ 1.4801e-03, 2.3499e-05, 7.6628e-04, ..., 9.7096e-05, + 9.2316e-04, 1.0290e-03], + [-5.4016e-03, 1.3685e-04, -2.7657e-03, ..., -3.4499e-04, + -1.4734e-03, -5.1498e-03], + [ 3.7632e-03, 1.7190e-04, 9.1410e-04, ..., 1.4484e-04, + 3.0918e-03, 1.2531e-03]], device='cuda:0') +Epoch 216, bias, value: tensor([ 0.0102, -0.0060, -0.0006, -0.0182, 0.0133, -0.0056, -0.0155, -0.0229, + 0.0099, -0.0025], device='cuda:0'), grad: tensor([-0.0362, 0.0098, -0.0355, 0.0212, 0.0126, -0.0028, 0.0177, 0.0084, + -0.0194, 0.0242], device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 216.43, cls_loss 0.4855 cls_loss_mapping 0.0079 cls_loss_causal 0.4611 re_mapping 0.0076 re_causal 0.0193 /// teacc 98.80 lr 0.00010000 +Epoch 217, weight, value: tensor([[ 0.0239, 0.0891, -0.1518, ..., -0.0858, 0.0634, 0.0048], + [-0.0765, -0.0882, 0.0818, ..., -0.0345, -0.0920, 0.0483], + [-0.0628, -0.0505, -0.0998, ..., -0.1178, -0.0213, 0.0459], + ..., + [-0.0183, -0.0988, -0.1103, ..., -0.1104, -0.0274, 0.0621], + [-0.0999, -0.0903, -0.0915, ..., -0.0577, -0.0542, -0.0806], + [ 0.0147, -0.0216, -0.0627, ..., -0.0628, 0.0230, -0.0168]], + device='cuda:0'), grad: tensor([[ 3.9024e-03, 1.7252e-03, 1.2312e-03, ..., 8.0633e-04, + 2.3308e-03, 1.2579e-03], + [ 1.5392e-03, 1.8358e-04, 1.4877e-03, ..., 1.6463e-04, + 2.0618e-03, 3.9041e-05], + [-4.2868e-04, 7.2098e-04, 1.0309e-03, ..., 2.8372e-04, + 2.4929e-03, 2.4853e-03], + ..., + [ 5.6267e-03, 1.0366e-03, 7.7152e-04, ..., 5.9366e-05, + 2.9068e-03, 8.7976e-04], + [-4.0197e-04, -1.6098e-03, 3.0470e-04, ..., -8.7070e-04, + -2.1946e-04, -4.3225e-04], + [-7.5226e-03, -1.0258e-04, -2.1420e-03, ..., 1.0616e-04, + -5.1079e-03, -1.9350e-03]], device='cuda:0') +Epoch 217, bias, value: tensor([ 0.0108, -0.0058, -0.0011, -0.0179, 0.0129, -0.0064, -0.0157, -0.0232, + 0.0102, -0.0019], device='cuda:0'), grad: tensor([ 0.0268, 0.0160, 0.0370, -0.0011, 0.0222, -0.0526, -0.0145, 0.0338, + 0.0138, -0.0814], device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 216.38, cls_loss 0.5328 cls_loss_mapping 0.0058 cls_loss_causal 0.4991 re_mapping 0.0076 re_causal 0.0197 /// teacc 98.82 lr 0.00010000 +Epoch 218, weight, value: tensor([[ 0.0251, 0.0896, -0.1507, ..., -0.0869, 0.0637, 0.0062], + [-0.0754, -0.0884, 0.0818, ..., -0.0337, -0.0904, 0.0483], + [-0.0634, -0.0512, -0.1003, ..., -0.1185, -0.0222, 0.0460], + ..., + [-0.0173, -0.0989, -0.1105, ..., -0.1106, -0.0273, 0.0635], + [-0.1001, -0.0909, -0.0925, ..., -0.0577, -0.0543, -0.0807], + [ 0.0130, -0.0220, -0.0624, ..., -0.0638, 0.0219, -0.0185]], + device='cuda:0'), grad: tensor([[ 1.2579e-03, -2.0158e-04, 8.8811e-05, ..., 1.0180e-04, + 3.9887e-04, 1.1168e-03], + [ 1.3866e-03, 1.7256e-05, -2.3627e-04, ..., 1.3483e-04, + 3.5691e-04, -3.3379e-03], + [-1.6654e-04, 2.3603e-05, 8.3923e-05, ..., 3.6895e-05, + 1.4913e-04, 8.6021e-04], + ..., + [-8.7128e-03, -7.4804e-05, 5.4741e-04, ..., -8.6737e-04, + -3.1414e-03, 2.7351e-03], + [ 5.9166e-03, 2.8534e-03, 4.7035e-03, ..., 2.0065e-03, + 7.0572e-04, 1.8101e-03], + [ 4.1656e-03, 1.0902e-04, 1.9426e-03, ..., 1.2863e-04, + 1.2026e-03, -5.3549e-04]], device='cuda:0') +Epoch 218, bias, value: tensor([ 0.0107, -0.0050, -0.0013, -0.0185, 0.0126, -0.0064, -0.0151, -0.0229, + 0.0103, -0.0024], device='cuda:0'), grad: tensor([ 0.0189, 0.0146, 0.0047, 0.0195, -0.0200, -0.0088, -0.0147, -0.0546, + 0.0244, 0.0161], device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 216.23, cls_loss 0.4859 cls_loss_mapping 0.0068 cls_loss_causal 0.4560 re_mapping 0.0079 re_causal 0.0200 /// teacc 98.77 lr 0.00010000 +Epoch 219, weight, value: tensor([[ 0.0254, 0.0900, -0.1502, ..., -0.0878, 0.0640, 0.0065], + [-0.0755, -0.0894, 0.0824, ..., -0.0337, -0.0907, 0.0492], + [-0.0643, -0.0524, -0.1027, ..., -0.1191, -0.0216, 0.0467], + ..., + [-0.0173, -0.0984, -0.1098, ..., -0.1093, -0.0274, 0.0627], + [-0.1008, -0.0932, -0.0929, ..., -0.0589, -0.0541, -0.0807], + [ 0.0138, -0.0219, -0.0631, ..., -0.0665, 0.0217, -0.0198]], + device='cuda:0'), grad: tensor([[-1.4591e-03, -2.3162e-04, -5.1022e-04, ..., -1.3614e-04, + -1.0052e-03, -9.4604e-04], + [ 1.8625e-03, 1.3053e-04, 1.2035e-03, ..., 2.8777e-04, + 6.4135e-04, 1.2913e-03], + [ 2.2583e-03, 1.7834e-04, -5.7518e-05, ..., 1.3494e-04, + 3.5858e-03, -3.2127e-05], + ..., + [-4.2648e-03, -5.5170e-04, -2.1954e-03, ..., -1.1368e-03, + -1.2062e-05, -5.2719e-03], + [-2.1458e-03, 1.0824e-04, -2.1420e-03, ..., 9.3579e-06, + -1.0500e-03, -8.2111e-04], + [ 9.9182e-05, 1.2410e-04, -5.8621e-05, ..., 8.0884e-05, + -6.4802e-04, 1.1711e-03]], device='cuda:0') +Epoch 219, bias, value: tensor([ 0.0101, -0.0042, -0.0015, -0.0178, 0.0127, -0.0060, -0.0157, -0.0234, + 0.0102, -0.0024], device='cuda:0'), grad: tensor([-0.0193, 0.0227, -0.0068, 0.0206, 0.0119, 0.0232, 0.0179, -0.0229, + -0.0419, -0.0056], device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 216.56, cls_loss 0.5188 cls_loss_mapping 0.0075 cls_loss_causal 0.4883 re_mapping 0.0080 re_causal 0.0213 /// teacc 98.71 lr 0.00010000 +Epoch 220, weight, value: tensor([[ 0.0264, 0.0915, -0.1489, ..., -0.0867, 0.0649, 0.0069], + [-0.0765, -0.0897, 0.0824, ..., -0.0336, -0.0910, 0.0507], + [-0.0635, -0.0520, -0.1029, ..., -0.1189, -0.0212, 0.0465], + ..., + [-0.0161, -0.0969, -0.1100, ..., -0.1092, -0.0273, 0.0629], + [-0.1008, -0.0947, -0.0933, ..., -0.0598, -0.0545, -0.0821], + [ 0.0132, -0.0228, -0.0645, ..., -0.0665, 0.0211, -0.0199]], + device='cuda:0'), grad: tensor([[ 1.0166e-03, -5.5008e-03, -7.5483e-04, ..., 1.0408e-05, + -2.2411e-03, 1.0090e-03], + [ 1.3247e-03, 1.5557e-05, -1.9288e-04, ..., -1.3316e-04, + 8.5688e-04, 1.0300e-03], + [ 8.3637e-04, 2.3925e-04, 1.1355e-04, ..., 6.1333e-05, + 9.5320e-04, -1.0729e-03], + ..., + [-2.2583e-03, 5.7109e-06, 2.1502e-05, ..., 2.7955e-05, + -7.7963e-04, -3.1281e-03], + [-7.8058e-04, 4.5681e-04, 9.2864e-05, ..., 2.5392e-04, + -8.5449e-04, 8.1539e-04], + [-8.2541e-04, 9.4831e-05, 1.7345e-05, ..., 8.6725e-06, + -1.0834e-03, -3.3426e-04]], device='cuda:0') +Epoch 220, bias, value: tensor([ 0.0102, -0.0037, -0.0011, -0.0181, 0.0130, -0.0067, -0.0159, -0.0231, + 0.0100, -0.0026], device='cuda:0'), grad: tensor([ 0.0077, 0.0214, 0.0053, -0.0298, -0.0088, 0.0189, 0.0153, -0.0311, + 0.0130, -0.0120], device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 216.35, cls_loss 0.4985 cls_loss_mapping 0.0087 cls_loss_causal 0.4778 re_mapping 0.0076 re_causal 0.0203 /// teacc 98.65 lr 0.00010000 +Epoch 221, weight, value: tensor([[ 0.0262, 0.0916, -0.1487, ..., -0.0877, 0.0652, 0.0079], + [-0.0767, -0.0907, 0.0826, ..., -0.0348, -0.0913, 0.0501], + [-0.0636, -0.0520, -0.1025, ..., -0.1181, -0.0204, 0.0475], + ..., + [-0.0154, -0.0959, -0.1087, ..., -0.1062, -0.0273, 0.0619], + [-0.1006, -0.0939, -0.0932, ..., -0.0599, -0.0538, -0.0817], + [ 0.0131, -0.0225, -0.0648, ..., -0.0674, 0.0207, -0.0203]], + device='cuda:0'), grad: tensor([[-5.9462e-04, 9.1270e-06, 1.0967e-05, ..., 1.6856e-04, + -6.1035e-04, 3.6430e-04], + [-2.0313e-03, 4.9584e-06, 1.1355e-05, ..., -1.8215e-03, + -1.2102e-03, -1.1902e-02], + [ 4.9137e-06, 2.1386e-04, 1.1569e-04, ..., 1.3328e-04, + 6.9475e-04, 6.7863e-03], + ..., + [-3.9148e-04, 4.1425e-05, 1.1146e-05, ..., 5.9336e-05, + -2.0349e-04, -7.0810e-05], + [ 1.4944e-03, 9.3102e-05, 7.2718e-05, ..., 2.5749e-04, + 8.7500e-04, 1.4381e-03], + [-4.5371e-04, 1.5974e-05, 2.1935e-05, ..., 1.5724e-04, + -5.6076e-04, 7.6151e-04]], device='cuda:0') +Epoch 221, bias, value: tensor([ 0.0111, -0.0039, -0.0006, -0.0177, 0.0121, -0.0058, -0.0161, -0.0244, + 0.0100, -0.0026], device='cuda:0'), grad: tensor([-0.0099, -0.0656, 0.0289, 0.0128, -0.0047, -0.0086, 0.0267, -0.0021, + 0.0284, -0.0060], device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 216.32, cls_loss 0.5077 cls_loss_mapping 0.0070 cls_loss_causal 0.4848 re_mapping 0.0072 re_causal 0.0198 /// teacc 98.71 lr 0.00010000 +Epoch 222, weight, value: tensor([[ 0.0266, 0.0931, -0.1479, ..., -0.0880, 0.0654, 0.0073], + [-0.0767, -0.0922, 0.0823, ..., -0.0341, -0.0906, 0.0511], + [-0.0638, -0.0522, -0.1031, ..., -0.1186, -0.0197, 0.0473], + ..., + [-0.0159, -0.0966, -0.1100, ..., -0.1056, -0.0283, 0.0617], + [-0.1008, -0.0927, -0.0936, ..., -0.0602, -0.0538, -0.0815], + [ 0.0136, -0.0220, -0.0631, ..., -0.0694, 0.0210, -0.0208]], + device='cuda:0'), grad: tensor([[ 2.9564e-03, 3.4809e-03, 2.7046e-03, ..., 1.6098e-03, + 1.9855e-03, 1.8816e-03], + [ 1.0023e-03, 6.0856e-05, -1.7433e-03, ..., 2.0278e-04, + 3.7122e-04, 3.3989e-03], + [-8.4734e-04, -5.5733e-03, -3.9368e-03, ..., -2.6436e-03, + 7.6485e-04, -2.0256e-03], + ..., + [ 1.7061e-03, 6.6161e-05, 4.0507e-04, ..., 6.0648e-05, + 1.1482e-03, -2.4929e-03], + [-7.5035e-03, -3.8838e-04, -1.1168e-03, ..., 1.6060e-03, + -7.1487e-03, -4.3335e-03], + [ 4.2839e-03, 1.2999e-03, 4.8256e-03, ..., 2.2745e-04, + 4.1161e-03, 2.3384e-03]], device='cuda:0') +Epoch 222, bias, value: tensor([ 0.0119, -0.0037, -0.0003, -0.0177, 0.0113, -0.0057, -0.0167, -0.0246, + 0.0096, -0.0019], device='cuda:0'), grad: tensor([ 0.0344, 0.0248, -0.0497, 0.0043, 0.0106, -0.0165, -0.0214, 0.0110, + -0.0234, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 216.28, cls_loss 0.5177 cls_loss_mapping 0.0078 cls_loss_causal 0.4926 re_mapping 0.0076 re_causal 0.0195 /// teacc 98.78 lr 0.00010000 +Epoch 223, weight, value: tensor([[ 0.0264, 0.0940, -0.1489, ..., -0.0884, 0.0658, 0.0067], + [-0.0777, -0.0931, 0.0831, ..., -0.0352, -0.0913, 0.0503], + [-0.0644, -0.0523, -0.1028, ..., -0.1180, -0.0201, 0.0481], + ..., + [-0.0157, -0.0982, -0.1101, ..., -0.1040, -0.0278, 0.0623], + [-0.1012, -0.0929, -0.0927, ..., -0.0602, -0.0536, -0.0807], + [ 0.0136, -0.0233, -0.0641, ..., -0.0719, 0.0207, -0.0208]], + device='cuda:0'), grad: tensor([[ 0.0008, -0.0002, 0.0005, ..., 0.0001, 0.0001, 0.0003], + [ 0.0011, 0.0004, -0.0010, ..., -0.0004, 0.0008, -0.0010], + [ 0.0015, 0.0007, 0.0012, ..., 0.0002, 0.0015, 0.0005], + ..., + [ 0.0003, 0.0004, 0.0008, ..., 0.0002, 0.0008, -0.0006], + [ 0.0015, 0.0004, 0.0006, ..., 0.0002, 0.0010, 0.0005], + [ 0.0033, 0.0013, 0.0019, ..., 0.0002, 0.0020, 0.0015]], + device='cuda:0') +Epoch 223, bias, value: tensor([ 0.0118, -0.0038, -0.0008, -0.0177, 0.0106, -0.0053, -0.0161, -0.0247, + 0.0094, -0.0013], device='cuda:0'), grad: tensor([ 0.0123, -0.0003, 0.0157, -0.0107, -0.0344, -0.0394, 0.0112, 0.0203, + 0.0161, 0.0091], device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 216.06, cls_loss 0.4947 cls_loss_mapping 0.0052 cls_loss_causal 0.4686 re_mapping 0.0070 re_causal 0.0189 /// teacc 98.76 lr 0.00010000 +Epoch 224, weight, value: tensor([[ 0.0262, 0.0937, -0.1487, ..., -0.0894, 0.0663, 0.0065], + [-0.0772, -0.0941, 0.0827, ..., -0.0351, -0.0916, 0.0499], + [-0.0652, -0.0517, -0.1024, ..., -0.1175, -0.0207, 0.0480], + ..., + [-0.0156, -0.0988, -0.1099, ..., -0.1046, -0.0279, 0.0628], + [-0.1014, -0.0907, -0.0921, ..., -0.0610, -0.0526, -0.0817], + [ 0.0131, -0.0233, -0.0645, ..., -0.0714, 0.0204, -0.0210]], + device='cuda:0'), grad: tensor([[ 1.3170e-03, 3.1033e-03, -4.8995e-05, ..., 3.8099e-04, + 2.3880e-03, 2.2144e-03], + [ 2.1229e-03, 2.0087e-04, -3.7718e-04, ..., 8.1658e-05, + 1.5268e-03, 1.1139e-03], + [-3.9411e-04, -4.4060e-03, 1.0097e-04, ..., 8.4281e-05, + -6.6948e-04, 6.7472e-04], + ..., + [-5.9280e-03, 1.8632e-04, -2.0817e-05, ..., -4.9925e-04, + -2.7637e-03, -4.6310e-03], + [ 2.0390e-03, 4.0746e-04, 3.2663e-04, ..., 2.3818e-04, + 1.6785e-03, 1.7824e-03], + [ 6.3896e-04, 1.6296e-04, 3.0780e-04, ..., -3.5465e-05, + -1.7900e-03, -4.7569e-03]], device='cuda:0') +Epoch 224, bias, value: tensor([ 0.0109, -0.0039, -0.0009, -0.0182, 0.0112, -0.0048, -0.0154, -0.0252, + 0.0096, -0.0012], device='cuda:0'), grad: tensor([ 0.0084, 0.0253, 0.0066, 0.0267, 0.0567, 0.0015, -0.0076, -0.0673, + 0.0042, -0.0544], device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 215.90, cls_loss 0.4991 cls_loss_mapping 0.0039 cls_loss_causal 0.4692 re_mapping 0.0076 re_causal 0.0200 /// teacc 98.84 lr 0.00010000 +Epoch 225, weight, value: tensor([[ 0.0254, 0.0937, -0.1487, ..., -0.0890, 0.0662, 0.0058], + [-0.0784, -0.0949, 0.0827, ..., -0.0355, -0.0926, 0.0506], + [-0.0655, -0.0515, -0.1027, ..., -0.1178, -0.0212, 0.0477], + ..., + [-0.0148, -0.0992, -0.1086, ..., -0.1049, -0.0274, 0.0627], + [-0.1014, -0.0886, -0.0921, ..., -0.0604, -0.0528, -0.0821], + [ 0.0131, -0.0227, -0.0630, ..., -0.0719, 0.0207, -0.0200]], + device='cuda:0'), grad: tensor([[-5.5981e-04, -3.1304e-04, 7.8455e-06, ..., 0.0000e+00, + -4.9210e-04, -1.5860e-03], + [ 6.8808e-04, 9.3877e-06, 6.4163e-03, ..., 0.0000e+00, + 3.1042e-04, 2.7885e-03], + [ 4.9496e-04, 1.5378e-05, 3.8862e-05, ..., 0.0000e+00, + 1.9526e-04, 1.5802e-03], + ..., + [ 2.0161e-03, 7.7105e-04, 1.0614e-03, ..., 0.0000e+00, + 1.3981e-03, 3.2330e-03], + [ 3.7909e-04, 1.4150e-04, 1.4591e-04, ..., 0.0000e+00, + 2.6989e-04, 1.5593e-03], + [-1.7872e-03, -5.7173e-04, -7.9346e-03, ..., 0.0000e+00, + -1.1539e-03, -8.0414e-03]], device='cuda:0') +Epoch 225, bias, value: tensor([ 0.0100, -0.0039, -0.0009, -0.0179, 0.0106, -0.0047, -0.0144, -0.0244, + 0.0090, -0.0014], device='cuda:0'), grad: tensor([-0.0195, 0.0015, 0.0168, -0.0167, -0.0066, -0.0196, 0.0171, 0.0240, + 0.0177, -0.0147], device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 215.97, cls_loss 0.5177 cls_loss_mapping 0.0046 cls_loss_causal 0.4918 re_mapping 0.0075 re_causal 0.0205 /// teacc 98.80 lr 0.00010000 +Epoch 226, weight, value: tensor([[ 0.0255, 0.0940, -0.1494, ..., -0.0888, 0.0670, 0.0062], + [-0.0791, -0.0946, 0.0833, ..., -0.0345, -0.0933, 0.0501], + [-0.0650, -0.0518, -0.1035, ..., -0.1172, -0.0204, 0.0480], + ..., + [-0.0151, -0.1000, -0.1080, ..., -0.1053, -0.0270, 0.0628], + [-0.1015, -0.0887, -0.0926, ..., -0.0604, -0.0527, -0.0817], + [ 0.0127, -0.0230, -0.0627, ..., -0.0717, 0.0203, -0.0193]], + device='cuda:0'), grad: tensor([[ 2.4109e-03, -2.6107e-04, 5.0850e-06, ..., 1.0471e-03, + 1.6999e-04, 1.5020e-03], + [-1.4896e-03, 5.0992e-05, 1.0437e-04, ..., 2.7108e-04, + -8.3208e-04, -6.5374e-04], + [ 1.2817e-03, 9.0897e-05, 9.1374e-05, ..., 1.1826e-04, + 3.7193e-04, 1.4524e-03], + ..., + [ 3.0231e-04, 1.5602e-03, 7.4234e-03, ..., 1.0806e-04, + 4.9639e-04, 2.7752e-03], + [ 2.6321e-03, 7.4208e-05, 4.8012e-05, ..., -8.8549e-04, + -1.3609e-03, 2.8305e-03], + [-7.9956e-03, 4.6015e-05, 9.7036e-05, ..., 1.1057e-04, + -3.3593e-04, -1.7767e-03]], device='cuda:0') +Epoch 226, bias, value: tensor([ 0.0113, -0.0046, -0.0006, -0.0183, 0.0101, -0.0052, -0.0145, -0.0231, + 0.0082, -0.0012], device='cuda:0'), grad: tensor([ 0.0319, -0.0032, 0.0211, 0.0036, 0.0240, -0.0408, 0.0007, -0.0364, + 0.0252, -0.0259], device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 215.96, cls_loss 0.5074 cls_loss_mapping 0.0083 cls_loss_causal 0.4821 re_mapping 0.0076 re_causal 0.0201 /// teacc 98.66 lr 0.00010000 +Epoch 227, weight, value: tensor([[ 0.0267, 0.0940, -0.1490, ..., -0.0891, 0.0672, 0.0046], + [-0.0794, -0.0942, 0.0840, ..., -0.0346, -0.0936, 0.0509], + [-0.0655, -0.0519, -0.1047, ..., -0.1179, -0.0212, 0.0484], + ..., + [-0.0157, -0.1012, -0.1089, ..., -0.1057, -0.0276, 0.0615], + [-0.1024, -0.0883, -0.0920, ..., -0.0602, -0.0523, -0.0823], + [ 0.0138, -0.0221, -0.0626, ..., -0.0719, 0.0203, -0.0194]], + device='cuda:0'), grad: tensor([[-8.3017e-04, 2.8163e-05, 1.1444e-04, ..., 1.5700e-04, + -6.0797e-04, 2.0683e-04], + [ 7.9870e-04, 2.8923e-05, 1.4222e-04, ..., 1.8930e-04, + 6.4135e-04, -1.5569e-04], + [ 7.6962e-04, 3.7611e-05, 4.0460e-04, ..., 7.3099e-04, + 6.8378e-04, 2.1648e-03], + ..., + [ 6.2084e-04, 1.0625e-05, 4.7088e-05, ..., 6.2525e-05, + 4.9877e-04, -4.8370e-03], + [-2.1172e-03, 9.9719e-05, 5.9366e-04, ..., 4.4394e-04, + -1.5402e-03, 6.8331e-04], + [ 7.1573e-04, 4.8250e-05, 2.1315e-04, ..., 3.0470e-04, + 6.0225e-04, 5.9223e-04]], device='cuda:0') +Epoch 227, bias, value: tensor([ 0.0118, -0.0039, -0.0014, -0.0179, 0.0111, -0.0065, -0.0145, -0.0239, + 0.0072, -0.0002], device='cuda:0'), grad: tensor([ 0.0015, -0.0127, 0.0249, 0.0085, 0.0201, 0.0287, -0.0416, -0.0100, + -0.0402, 0.0208], device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 215.96, cls_loss 0.5324 cls_loss_mapping 0.0056 cls_loss_causal 0.5036 re_mapping 0.0073 re_causal 0.0195 /// teacc 98.56 lr 0.00010000 +Epoch 228, weight, value: tensor([[ 0.0267, 0.0948, -0.1488, ..., -0.0892, 0.0671, 0.0048], + [-0.0795, -0.0948, 0.0833, ..., -0.0348, -0.0929, 0.0525], + [-0.0651, -0.0519, -0.1053, ..., -0.1182, -0.0216, 0.0483], + ..., + [-0.0165, -0.1021, -0.1082, ..., -0.1054, -0.0294, 0.0620], + [-0.1026, -0.0891, -0.0935, ..., -0.0606, -0.0523, -0.0832], + [ 0.0152, -0.0206, -0.0606, ..., -0.0731, 0.0221, -0.0210]], + device='cuda:0'), grad: tensor([[ 1.0433e-03, -1.6937e-03, -7.0000e-04, ..., -5.0974e-04, + 8.9073e-04, 2.2304e-04], + [ 1.5345e-03, 3.3230e-05, -1.8105e-05, ..., 4.5598e-06, + 1.0262e-03, 1.4770e-04], + [-2.9125e-03, 4.2987e-04, 1.1885e-04, ..., 9.0778e-05, + -5.9120e-06, -6.7806e-04], + ..., + [ 1.7767e-03, 1.3256e-04, 2.4453e-05, ..., 1.2659e-05, + 1.1702e-03, 7.0381e-04], + [ 4.6768e-03, 8.4839e-03, 6.7596e-03, ..., 1.8358e-04, + 3.2692e-03, 1.4663e-05], + [-2.5768e-03, -7.3586e-03, -6.7940e-03, ..., 5.9545e-05, + -3.3054e-03, 6.0558e-04]], device='cuda:0') +Epoch 228, bias, value: tensor([ 0.0116, -0.0042, -0.0013, -0.0175, 0.0107, -0.0060, -0.0136, -0.0241, + 0.0068, -0.0004], device='cuda:0'), grad: tensor([ 0.0186, 0.0210, -0.0421, 0.0278, -0.0381, 0.0145, -0.0278, 0.0242, + -0.0043, 0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 215.93, cls_loss 0.5036 cls_loss_mapping 0.0053 cls_loss_causal 0.4698 re_mapping 0.0075 re_causal 0.0203 /// teacc 98.86 lr 0.00010000 +Epoch 229, weight, value: tensor([[ 0.0268, 0.0952, -0.1488, ..., -0.0888, 0.0670, 0.0042], + [-0.0804, -0.0959, 0.0834, ..., -0.0342, -0.0943, 0.0525], + [-0.0649, -0.0510, -0.1052, ..., -0.1181, -0.0207, 0.0476], + ..., + [-0.0160, -0.1033, -0.1097, ..., -0.1063, -0.0292, 0.0627], + [-0.1027, -0.0901, -0.0936, ..., -0.0611, -0.0524, -0.0832], + [ 0.0145, -0.0198, -0.0603, ..., -0.0731, 0.0222, -0.0206]], + device='cuda:0'), grad: tensor([[ 1.0328e-03, 1.3494e-04, 7.2670e-04, ..., 1.3351e-04, + 1.2722e-03, 8.0442e-04], + [-4.8561e-03, 2.7514e-04, 9.1934e-03, ..., 4.9442e-05, + -4.4212e-03, 3.1872e-03], + [-3.2310e-03, -4.4365e-03, -6.0043e-03, ..., -1.5841e-03, + -1.1238e-02, 6.8092e-04], + ..., + [ 4.3449e-03, 4.5687e-05, 4.3035e-04, ..., 6.1750e-05, + 2.0752e-03, 2.1381e-03], + [-9.7275e-04, 4.0665e-03, 2.6760e-03, ..., 2.6264e-03, + 1.0538e-03, -2.6855e-03], + [-1.1854e-03, 5.4896e-05, 3.4451e-04, ..., 4.2677e-05, + 1.0328e-03, -8.5068e-04]], device='cuda:0') +Epoch 229, bias, value: tensor([ 0.0113, -0.0045, -0.0019, -0.0170, 0.0099, -0.0076, -0.0114, -0.0233, + 0.0068, -0.0004], device='cuda:0'), grad: tensor([ 0.0126, -0.0182, -0.0345, 0.0103, 0.0149, 0.0022, -0.0139, 0.0260, + 0.0112, -0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 216.34, cls_loss 0.5076 cls_loss_mapping 0.0047 cls_loss_causal 0.4779 re_mapping 0.0076 re_causal 0.0191 /// teacc 98.77 lr 0.00010000 +Epoch 230, weight, value: tensor([[ 0.0279, 0.0956, -0.1487, ..., -0.0884, 0.0670, 0.0035], + [-0.0801, -0.0957, 0.0833, ..., -0.0341, -0.0933, 0.0524], + [-0.0658, -0.0514, -0.1055, ..., -0.1185, -0.0222, 0.0478], + ..., + [-0.0148, -0.1035, -0.1102, ..., -0.1061, -0.0307, 0.0629], + [-0.1023, -0.0910, -0.0943, ..., -0.0607, -0.0524, -0.0826], + [ 0.0132, -0.0191, -0.0609, ..., -0.0720, 0.0229, -0.0199]], + device='cuda:0'), grad: tensor([[ 6.7174e-05, -1.5819e-04, 5.4054e-06, ..., 5.9139e-07, + -9.3222e-04, 5.6362e-04], + [-2.5711e-03, 4.8466e-06, -2.0218e-04, ..., -3.2634e-05, + 2.2221e-04, -2.6340e-03], + [ 4.8518e-04, 2.1309e-05, 3.4064e-05, ..., 3.8929e-06, + -5.9509e-04, -1.9398e-03], + ..., + [ 6.1035e-04, 3.1982e-06, 4.7207e-05, ..., 7.2122e-06, + 4.0269e-04, -4.2224e-04], + [ 7.1287e-04, 1.3545e-05, 8.0228e-05, ..., 1.1899e-05, + 4.2248e-04, 1.2369e-03], + [-1.4153e-03, 4.3243e-05, 1.7926e-05, ..., 2.0582e-06, + -1.4315e-03, 9.7847e-04]], device='cuda:0') +Epoch 230, bias, value: tensor([ 0.0116, -0.0038, -0.0016, -0.0179, 0.0110, -0.0080, -0.0128, -0.0238, + 0.0076, -0.0005], device='cuda:0'), grad: tensor([ 0.0038, -0.0181, -0.0098, 0.0085, 0.0079, 0.0051, 0.0080, 0.0054, + -0.0134, 0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 216.05, cls_loss 0.5130 cls_loss_mapping 0.0058 cls_loss_causal 0.4862 re_mapping 0.0075 re_causal 0.0192 /// teacc 98.81 lr 0.00010000 +Epoch 231, weight, value: tensor([[ 0.0278, 0.0952, -0.1504, ..., -0.0889, 0.0660, 0.0033], + [-0.0800, -0.0954, 0.0845, ..., -0.0330, -0.0939, 0.0534], + [-0.0658, -0.0525, -0.1065, ..., -0.1195, -0.0213, 0.0482], + ..., + [-0.0150, -0.1039, -0.1095, ..., -0.1065, -0.0312, 0.0623], + [-0.1028, -0.0917, -0.0961, ..., -0.0608, -0.0512, -0.0823], + [ 0.0134, -0.0204, -0.0618, ..., -0.0725, 0.0227, -0.0193]], + device='cuda:0'), grad: tensor([[-2.1191e-03, 1.3880e-05, 1.7390e-05, ..., 3.1907e-06, + -1.2894e-03, 1.6123e-05], + [ 8.2684e-04, 5.9396e-05, -1.2703e-05, ..., 1.1697e-06, + 2.9635e-04, 4.1199e-04], + [-2.0428e-03, 1.9297e-05, -4.0025e-05, ..., -1.2122e-05, + -1.2999e-03, 6.2037e-04], + ..., + [ 2.2850e-03, 9.0301e-05, 2.3142e-05, ..., 1.5348e-06, + 2.8849e-04, -3.7594e-03], + [ 5.2691e-04, 1.6344e-04, 4.4256e-05, ..., 3.5651e-06, + 2.8324e-04, 1.5974e-05], + [ 1.0178e-02, 6.1393e-05, 5.4061e-05, ..., 8.0932e-07, + 3.3522e-04, 2.4509e-03]], device='cuda:0') +Epoch 231, bias, value: tensor([ 0.0105, -0.0042, -0.0003, -0.0185, 0.0120, -0.0072, -0.0131, -0.0251, + 0.0079, -0.0001], device='cuda:0'), grad: tensor([-0.0205, 0.0163, -0.0051, 0.0259, -0.0218, 0.0026, -0.0159, 0.0079, + 0.0004, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 216.12, cls_loss 0.5067 cls_loss_mapping 0.0071 cls_loss_causal 0.4789 re_mapping 0.0071 re_causal 0.0180 /// teacc 98.77 lr 0.00010000 +Epoch 232, weight, value: tensor([[ 0.0286, 0.0959, -0.1500, ..., -0.0888, 0.0657, 0.0035], + [-0.0799, -0.0947, 0.0850, ..., -0.0335, -0.0936, 0.0535], + [-0.0658, -0.0534, -0.1068, ..., -0.1204, -0.0213, 0.0480], + ..., + [-0.0140, -0.1049, -0.1098, ..., -0.1063, -0.0304, 0.0626], + [-0.1034, -0.0920, -0.0949, ..., -0.0604, -0.0512, -0.0827], + [ 0.0122, -0.0213, -0.0630, ..., -0.0735, 0.0228, -0.0197]], + device='cuda:0'), grad: tensor([[ 7.0190e-04, 3.0231e-04, 1.1331e-04, ..., 3.7813e-04, + 4.2081e-04, 9.5510e-04], + [ 1.1692e-03, 1.5008e-04, -7.2736e-07, ..., 3.8773e-05, + 1.4949e-04, -1.9951e-03], + [-1.8616e-03, -2.8362e-03, -8.4266e-06, ..., 2.4152e-04, + -4.6802e-04, -1.6699e-03], + ..., + [ 3.5515e-03, 7.9536e-04, 1.2457e-05, ..., 1.1516e-04, + 1.0538e-03, 3.1319e-03], + [-2.4662e-03, 3.9291e-04, 5.9932e-05, ..., 2.9445e-04, + -3.4161e-03, 8.7881e-04], + [-1.5087e-03, -4.4799e-04, 9.8869e-06, ..., 4.8780e-04, + -1.3256e-03, 1.6184e-03]], device='cuda:0') +Epoch 232, bias, value: tensor([ 0.0110, -0.0036, 0.0004, -0.0189, 0.0117, -0.0072, -0.0129, -0.0250, + 0.0071, -0.0006], device='cuda:0'), grad: tensor([-0.0090, 0.0107, -0.0370, 0.0421, -0.0262, -0.0067, 0.0209, 0.0480, + -0.0250, -0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 216.14, cls_loss 0.5150 cls_loss_mapping 0.0103 cls_loss_causal 0.4829 re_mapping 0.0074 re_causal 0.0184 /// teacc 98.75 lr 0.00010000 +Epoch 233, weight, value: tensor([[ 0.0284, 0.0970, -0.1507, ..., -0.0889, 0.0658, 0.0032], + [-0.0793, -0.0943, 0.0852, ..., -0.0355, -0.0935, 0.0544], + [-0.0670, -0.0538, -0.1072, ..., -0.1215, -0.0220, 0.0473], + ..., + [-0.0146, -0.1053, -0.1100, ..., -0.1060, -0.0305, 0.0632], + [-0.1030, -0.0927, -0.0950, ..., -0.0610, -0.0507, -0.0824], + [ 0.0125, -0.0216, -0.0634, ..., -0.0730, 0.0227, -0.0199]], + device='cuda:0'), grad: tensor([[-2.5253e-03, -3.5739e-04, 1.4317e-04, ..., 1.5342e-04, + -5.8889e-04, -5.9929e-03], + [ 1.8473e-03, 1.8227e-04, 1.3018e-04, ..., 1.5259e-04, + 1.2608e-03, 1.7138e-03], + [ 1.1902e-03, 2.1064e-04, 7.2956e-05, ..., -1.7869e-04, + 6.5708e-04, 3.6354e-03], + ..., + [ 1.9550e-03, 2.2924e-04, 6.5506e-05, ..., 1.0073e-04, + 1.2293e-03, 1.1168e-03], + [ 1.7796e-03, 2.1732e-04, 3.0637e-05, ..., 8.1003e-05, + 1.2569e-03, 1.0138e-03], + [ 2.4567e-03, 5.9128e-04, 1.4973e-04, ..., 1.4055e-04, + 1.9836e-03, 9.6846e-04]], device='cuda:0') +Epoch 233, bias, value: tensor([ 0.0102, -0.0038, 0.0006, -0.0194, 0.0124, -0.0071, -0.0133, -0.0250, + 0.0072, 0.0002], device='cuda:0'), grad: tensor([-0.0698, 0.0299, 0.0391, -0.0391, -0.0092, -0.0137, -0.0119, 0.0259, + 0.0234, 0.0254], device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 216.34, cls_loss 0.5130 cls_loss_mapping 0.0070 cls_loss_causal 0.4858 re_mapping 0.0072 re_causal 0.0181 /// teacc 98.79 lr 0.00010000 +Epoch 234, weight, value: tensor([[ 0.0289, 0.0972, -0.1511, ..., -0.0905, 0.0655, 0.0029], + [-0.0803, -0.0951, 0.0842, ..., -0.0340, -0.0935, 0.0568], + [-0.0660, -0.0517, -0.1071, ..., -0.1199, -0.0213, 0.0479], + ..., + [-0.0153, -0.1067, -0.1099, ..., -0.1063, -0.0304, 0.0620], + [-0.1036, -0.0927, -0.0939, ..., -0.0616, -0.0501, -0.0832], + [ 0.0133, -0.0216, -0.0632, ..., -0.0723, 0.0221, -0.0205]], + device='cuda:0'), grad: tensor([[ 1.5461e-04, -6.6981e-06, 3.9907e-07, ..., 2.0146e-04, + 1.5461e-04, 5.7173e-04], + [ 3.5048e-04, 6.0722e-07, 7.0874e-07, ..., 4.0865e-04, + 1.7083e-04, 1.6747e-03], + [-4.6110e-04, 5.3644e-06, 2.6748e-06, ..., 3.7599e-04, + 3.2139e-04, -2.6722e-03], + ..., + [ 1.0433e-03, 9.7096e-05, 1.9874e-06, ..., 1.8096e-04, + -1.6575e-03, 7.2670e-04], + [ 3.3808e-04, 9.3728e-06, 5.5917e-06, ..., 4.4894e-04, + 7.8630e-04, -3.2520e-04], + [-5.1308e-04, -9.1553e-05, 2.8126e-06, ..., 2.9659e-04, + -1.6141e-04, 5.8556e-04]], device='cuda:0') +Epoch 234, bias, value: tensor([ 0.0103, -0.0038, 0.0021, -0.0190, 0.0114, -0.0066, -0.0141, -0.0257, + 0.0070, 0.0003], device='cuda:0'), grad: tensor([ 0.0050, 0.0102, 0.0008, 0.0186, -0.0545, 0.0062, 0.0133, -0.0097, + 0.0031, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 216.15, cls_loss 0.5240 cls_loss_mapping 0.0062 cls_loss_causal 0.4966 re_mapping 0.0075 re_causal 0.0194 /// teacc 98.83 lr 0.00010000 +Epoch 235, weight, value: tensor([[ 0.0286, 0.0971, -0.1512, ..., -0.0904, 0.0654, 0.0022], + [-0.0808, -0.0962, 0.0849, ..., -0.0338, -0.0945, 0.0580], + [-0.0662, -0.0506, -0.1076, ..., -0.1198, -0.0224, 0.0475], + ..., + [-0.0144, -0.1077, -0.1101, ..., -0.1067, -0.0299, 0.0621], + [-0.1040, -0.0932, -0.0934, ..., -0.0615, -0.0504, -0.0833], + [ 0.0130, -0.0219, -0.0628, ..., -0.0737, 0.0214, -0.0202]], + device='cuda:0'), grad: tensor([[-6.3837e-05, -1.1024e-03, 6.2823e-05, ..., 1.0061e-03, + -2.0924e-03, 6.0463e-04], + [ 1.1797e-03, 3.6955e-05, 1.3103e-03, ..., 2.4929e-03, + 1.0735e-04, 7.5722e-03], + [ 2.7466e-04, 7.8678e-05, 1.4102e-04, ..., 2.4676e-04, + 9.3341e-05, -5.7602e-03], + ..., + [ 7.7200e-04, 2.0111e-04, 2.3401e-04, ..., 3.7742e-04, + 2.7204e-04, 1.7958e-03], + [ 2.0580e-03, 1.8444e-03, 6.8998e-04, ..., 9.2363e-04, + 1.5717e-03, 2.0561e-03], + [ 2.3880e-03, 1.3046e-03, 6.1035e-04, ..., 8.1825e-04, + 6.8617e-04, 1.8358e-03]], device='cuda:0') +Epoch 235, bias, value: tensor([ 0.0106, -0.0034, 0.0020, -0.0189, 0.0110, -0.0062, -0.0150, -0.0258, + 0.0073, 0.0003], device='cuda:0'), grad: tensor([ 0.0107, 0.0292, -0.0189, 0.0073, -0.0019, -0.0573, -0.0089, 0.0012, + 0.0193, 0.0195], device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 216.21, cls_loss 0.4892 cls_loss_mapping 0.0074 cls_loss_causal 0.4640 re_mapping 0.0078 re_causal 0.0195 /// teacc 98.71 lr 0.00010000 +Epoch 236, weight, value: tensor([[ 0.0290, 0.0970, -0.1520, ..., -0.0910, 0.0655, 0.0016], + [-0.0811, -0.0961, 0.0850, ..., -0.0341, -0.0953, 0.0571], + [-0.0673, -0.0512, -0.1076, ..., -0.1204, -0.0229, 0.0481], + ..., + [-0.0144, -0.1089, -0.1093, ..., -0.1064, -0.0295, 0.0621], + [-0.1041, -0.0934, -0.0937, ..., -0.0594, -0.0501, -0.0823], + [ 0.0135, -0.0230, -0.0622, ..., -0.0742, 0.0219, -0.0210]], + device='cuda:0'), grad: tensor([[ 5.6601e-04, 1.7571e-04, 1.2082e-04, ..., 1.3888e-04, + 2.5183e-05, 6.1655e-04], + [-1.1024e-03, -3.9220e-04, -2.2209e-04, ..., 4.9546e-07, + -9.6917e-05, -9.2773e-03], + [ 3.5381e-04, 1.6779e-05, 1.3746e-05, ..., -1.0014e-05, + 1.2740e-05, 9.0313e-04], + ..., + [ 5.0116e-04, 6.4194e-05, 2.1070e-05, ..., 4.0568e-06, + 1.7810e-04, 1.3037e-03], + [ 3.5286e-04, 6.9320e-05, 5.2005e-05, ..., 1.9580e-05, + 1.4462e-05, 8.0681e-04], + [ 3.6263e-04, 4.9442e-05, 8.0541e-06, ..., 2.4978e-06, + 5.9396e-05, 2.3689e-03]], device='cuda:0') +Epoch 236, bias, value: tensor([ 0.0104, -0.0034, 0.0020, -0.0187, 0.0106, -0.0062, -0.0144, -0.0267, + 0.0075, 0.0009], device='cuda:0'), grad: tensor([ 0.0153, -0.0511, 0.0116, 0.0137, -0.0196, 0.0086, 0.0191, -0.0329, + 0.0157, 0.0196], device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 216.48, cls_loss 0.4994 cls_loss_mapping 0.0056 cls_loss_causal 0.4726 re_mapping 0.0066 re_causal 0.0170 /// teacc 98.77 lr 0.00010000 +Epoch 237, weight, value: tensor([[ 0.0287, 0.0973, -0.1523, ..., -0.0912, 0.0650, 0.0022], + [-0.0820, -0.0960, 0.0851, ..., -0.0337, -0.0966, 0.0567], + [-0.0678, -0.0519, -0.1083, ..., -0.1211, -0.0230, 0.0476], + ..., + [-0.0155, -0.1100, -0.1096, ..., -0.1069, -0.0297, 0.0628], + [-0.1027, -0.0927, -0.0929, ..., -0.0587, -0.0500, -0.0811], + [ 0.0127, -0.0237, -0.0623, ..., -0.0740, 0.0214, -0.0221]], + device='cuda:0'), grad: tensor([[-3.2711e-03, -1.5259e-03, 3.0613e-04, ..., 1.7893e-04, + 2.3758e-04, 1.5247e-04], + [ 1.2159e-03, 4.2415e-04, 1.7428e-04, ..., 9.1314e-05, + 4.2939e-04, 3.2640e-04], + [ 9.2745e-04, 4.1056e-04, 1.6332e-04, ..., 7.8499e-05, + 3.4046e-04, -1.5450e-02], + ..., + [ 1.2941e-03, -1.0052e-03, 1.4782e-04, ..., 7.4029e-05, + 3.8028e-04, 1.4442e-02], + [-3.0637e-04, 1.0462e-03, 2.8610e-04, ..., 1.0109e-03, + -8.9121e-04, -2.7037e-04], + [-3.9077e-04, 6.6090e-04, 1.2577e-04, ..., 1.8477e-04, + -8.3733e-04, 1.2934e-04]], device='cuda:0') +Epoch 237, bias, value: tensor([ 0.0098, -0.0039, 0.0012, -0.0178, 0.0118, -0.0065, -0.0153, -0.0263, + 0.0083, 0.0004], device='cuda:0'), grad: tensor([-0.0155, 0.0138, -0.0210, 0.0250, 0.0120, 0.0129, -0.0393, 0.0286, + -0.0080, -0.0086], device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 216.24, cls_loss 0.5001 cls_loss_mapping 0.0052 cls_loss_causal 0.4810 re_mapping 0.0069 re_causal 0.0184 /// teacc 98.74 lr 0.00010000 +Epoch 238, weight, value: tensor([[ 0.0288, 0.0983, -0.1531, ..., -0.0904, 0.0649, 0.0037], + [-0.0820, -0.0957, 0.0856, ..., -0.0333, -0.0968, 0.0559], + [-0.0670, -0.0507, -0.1078, ..., -0.1209, -0.0224, 0.0498], + ..., + [-0.0160, -0.1120, -0.1099, ..., -0.1069, -0.0308, 0.0620], + [-0.1031, -0.0944, -0.0941, ..., -0.0588, -0.0497, -0.0810], + [ 0.0122, -0.0244, -0.0632, ..., -0.0738, 0.0209, -0.0232]], + device='cuda:0'), grad: tensor([[ 2.0962e-03, 3.5267e-03, 5.0306e-05, ..., 5.9319e-04, + 8.0156e-04, 2.3785e-03], + [-8.1420e-05, -2.9135e-04, 2.5287e-05, ..., -2.4109e-03, + -5.6505e-04, -4.2648e-03], + [-1.2032e-02, -6.2895e-04, -1.9956e-04, ..., 5.5265e-04, + -7.4768e-03, -5.5237e-03], + ..., + [-1.2312e-03, 3.3760e-04, 5.1975e-05, ..., 4.4537e-04, + -1.1892e-03, 2.9259e-03], + [ 3.4447e-03, -4.8294e-03, 7.3493e-05, ..., 1.3714e-03, + 1.2093e-03, 2.4796e-03], + [-2.3305e-05, -3.2425e-03, 4.4644e-05, ..., 4.7970e-04, + 1.2560e-03, -3.0065e-04]], device='cuda:0') +Epoch 238, bias, value: tensor([ 0.0096, -0.0043, 0.0024, -0.0177, 0.0119, -0.0066, -0.0154, -0.0263, + 0.0078, 0.0003], device='cuda:0'), grad: tensor([ 0.0455, -0.0099, -0.0353, -0.0177, 0.0028, 0.0500, -0.0154, 0.0038, + 0.0006, -0.0244], device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 216.54, cls_loss 0.5416 cls_loss_mapping 0.0059 cls_loss_causal 0.5175 re_mapping 0.0070 re_causal 0.0189 /// teacc 98.75 lr 0.00010000 +Epoch 239, weight, value: tensor([[ 0.0281, 0.0990, -0.1530, ..., -0.0906, 0.0654, 0.0051], + [-0.0812, -0.0944, 0.0860, ..., -0.0329, -0.0963, 0.0559], + [-0.0670, -0.0510, -0.1082, ..., -0.1213, -0.0225, 0.0492], + ..., + [-0.0158, -0.1124, -0.1103, ..., -0.1076, -0.0316, 0.0613], + [-0.1043, -0.0948, -0.0945, ..., -0.0594, -0.0514, -0.0828], + [ 0.0125, -0.0239, -0.0630, ..., -0.0726, 0.0214, -0.0221]], + device='cuda:0'), grad: tensor([[-9.2208e-05, 8.7261e-04, 9.5926e-08, ..., -1.2207e-03, + 2.2721e-04, -2.9678e-03], + [ 3.8886e-04, 2.3162e-04, -4.2841e-07, ..., 8.5640e-04, + 1.6832e-04, -4.8351e-04], + [ 1.7452e-03, 2.9016e-04, 6.6590e-07, ..., 9.6941e-04, + 1.5771e-04, 3.8109e-03], + ..., + [-2.9888e-03, 1.3268e-04, 1.8999e-07, ..., -8.5294e-05, + 1.1015e-04, -5.7602e-03], + [ 1.6270e-03, 3.2401e-04, 2.2631e-07, ..., 1.9341e-03, + 1.9169e-04, 3.0403e-03], + [ 1.9436e-03, 1.9002e-04, 1.2014e-07, ..., 2.5773e-04, + 1.5223e-04, 2.2640e-03]], device='cuda:0') +Epoch 239, bias, value: tensor([ 0.0100, -0.0040, 0.0024, -0.0179, 0.0113, -0.0064, -0.0150, -0.0274, + 0.0080, 0.0007], device='cuda:0'), grad: tensor([-0.0340, -0.0035, 0.0279, -0.0063, -0.0367, 0.0026, -0.0081, -0.0532, + 0.0482, 0.0630], device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 216.31, cls_loss 0.4966 cls_loss_mapping 0.0052 cls_loss_causal 0.4575 re_mapping 0.0070 re_causal 0.0180 /// teacc 98.80 lr 0.00010000 +Epoch 240, weight, value: tensor([[ 0.0283, 0.0993, -0.1537, ..., -0.0909, 0.0659, 0.0050], + [-0.0815, -0.0943, 0.0863, ..., -0.0329, -0.0963, 0.0548], + [-0.0674, -0.0531, -0.1082, ..., -0.1231, -0.0236, 0.0495], + ..., + [-0.0155, -0.1125, -0.1114, ..., -0.1084, -0.0319, 0.0628], + [-0.1055, -0.0949, -0.0946, ..., -0.0601, -0.0523, -0.0837], + [ 0.0130, -0.0228, -0.0623, ..., -0.0733, 0.0220, -0.0220]], + device='cuda:0'), grad: tensor([[7.2527e-04, 8.9569e-03, 2.9621e-03, ..., 7.9498e-03, 1.9522e-03, + 1.0347e-03], + [1.1988e-03, 1.1033e-04, 6.5982e-05, ..., 1.7214e-04, 6.0272e-04, + 1.2665e-03], + [1.9093e-03, 9.1374e-05, 9.6187e-06, ..., 7.2575e-04, 1.2379e-03, + 1.0366e-03], + ..., + [5.2185e-03, 1.1396e-03, 6.3801e-04, ..., 1.4186e-04, 2.5101e-03, + 2.5711e-03], + [1.9512e-03, 5.7449e-03, 2.5043e-03, ..., 4.5700e-03, 1.4153e-03, + 1.0014e-03], + [2.0676e-03, 5.8651e-04, 2.2888e-04, ..., 3.0398e-04, 1.2455e-03, + 1.5612e-03]], device='cuda:0') +Epoch 240, bias, value: tensor([ 0.0104, -0.0039, 0.0023, -0.0182, 0.0116, -0.0070, -0.0151, -0.0260, + 0.0070, 0.0006], device='cuda:0'), grad: tensor([ 0.0033, 0.0177, -0.0052, 0.0281, -0.0341, -0.0249, -0.0629, 0.0339, + 0.0281, 0.0160], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 239---------------------------------------------------- +epoch 239, time 217.11, cls_loss 0.4661 cls_loss_mapping 0.0053 cls_loss_causal 0.4462 re_mapping 0.0071 re_causal 0.0185 /// teacc 98.94 lr 0.00010000 +Epoch 241, weight, value: tensor([[ 0.0295, 0.0995, -0.1543, ..., -0.0903, 0.0666, 0.0063], + [-0.0810, -0.0944, 0.0863, ..., -0.0334, -0.0954, 0.0536], + [-0.0668, -0.0525, -0.1063, ..., -0.1225, -0.0228, 0.0499], + ..., + [-0.0167, -0.1125, -0.1120, ..., -0.1067, -0.0325, 0.0626], + [-0.1037, -0.0955, -0.0955, ..., -0.0617, -0.0524, -0.0841], + [ 0.0146, -0.0230, -0.0620, ..., -0.0723, 0.0218, -0.0218]], + device='cuda:0'), grad: tensor([[ 1.5297e-03, 3.6120e-04, 2.9659e-04, ..., 9.5844e-04, + 5.6744e-05, 1.7319e-03], + [ 9.6226e-04, 7.8559e-05, 2.5332e-05, ..., 2.4843e-04, + 6.7353e-05, 1.2341e-03], + [-2.1152e-03, 1.0653e-03, 4.0579e-04, ..., 1.7881e-05, + 2.0063e-04, -2.0752e-03], + ..., + [ 1.9245e-03, 4.0293e-04, 1.3733e-04, ..., 6.4421e-04, + 2.3413e-04, 1.2522e-03], + [ 1.5392e-03, 4.5013e-04, 1.6046e-04, ..., 6.2895e-04, + 2.3842e-04, 6.2513e-04], + [ 3.5305e-03, 1.2865e-03, 4.5562e-04, ..., 1.4153e-03, + 5.6648e-04, -8.3542e-04]], device='cuda:0') +Epoch 241, bias, value: tensor([ 0.0114, -0.0041, 0.0020, -0.0191, 0.0112, -0.0062, -0.0152, -0.0255, + 0.0068, 0.0005], device='cuda:0'), grad: tensor([ 0.0131, 0.0104, -0.0023, 0.0168, -0.0223, 0.0139, -0.0468, 0.0131, + -0.0059, 0.0101], device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 216.20, cls_loss 0.5225 cls_loss_mapping 0.0056 cls_loss_causal 0.4957 re_mapping 0.0074 re_causal 0.0203 /// teacc 98.70 lr 0.00010000 +Epoch 242, weight, value: tensor([[ 0.0291, 0.1003, -0.1542, ..., -0.0910, 0.0655, 0.0064], + [-0.0806, -0.0954, 0.0871, ..., -0.0346, -0.0954, 0.0529], + [-0.0668, -0.0531, -0.1079, ..., -0.1227, -0.0224, 0.0498], + ..., + [-0.0168, -0.1117, -0.1124, ..., -0.1060, -0.0317, 0.0625], + [-0.1036, -0.0956, -0.0959, ..., -0.0602, -0.0526, -0.0844], + [ 0.0138, -0.0230, -0.0629, ..., -0.0734, 0.0229, -0.0229]], + device='cuda:0'), grad: tensor([[ 1.7538e-03, 3.3998e-04, 1.4102e-04, ..., 6.7377e-04, + 6.4802e-04, 2.5463e-03], + [ 1.6985e-03, 1.1837e-04, 2.1309e-05, ..., 5.1451e-04, + 5.5313e-04, 2.2717e-03], + [-1.3924e-04, -1.2789e-03, 1.0747e-04, ..., 3.8171e-04, + 5.3465e-05, 1.0841e-02], + ..., + [ 6.0921e-03, 1.0386e-03, 1.0645e-04, ..., 1.2326e-04, + 1.6336e-03, 8.0872e-03], + [ 1.2913e-03, 3.2949e-04, 2.1517e-04, ..., 4.3869e-04, + 5.8174e-04, 2.9526e-03], + [-9.3231e-03, 1.7824e-03, 1.4515e-03, ..., 1.2646e-03, + -1.6403e-03, -1.1505e-02]], device='cuda:0') +Epoch 242, bias, value: tensor([ 0.0114, -0.0034, 0.0013, -0.0189, 0.0120, -0.0060, -0.0142, -0.0262, + 0.0065, -0.0005], device='cuda:0'), grad: tensor([ 0.0220, 0.0219, 0.0434, -0.0174, -0.0242, -0.0430, 0.0191, 0.0222, + 0.0200, -0.0641], device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 216.42, cls_loss 0.5266 cls_loss_mapping 0.0064 cls_loss_causal 0.4990 re_mapping 0.0070 re_causal 0.0191 /// teacc 98.70 lr 0.00010000 +Epoch 243, weight, value: tensor([[ 0.0285, 0.0992, -0.1557, ..., -0.0918, 0.0653, 0.0064], + [-0.0809, -0.0964, 0.0869, ..., -0.0340, -0.0958, 0.0532], + [-0.0664, -0.0532, -0.1088, ..., -0.1222, -0.0219, 0.0498], + ..., + [-0.0163, -0.1127, -0.1120, ..., -0.1053, -0.0323, 0.0624], + [-0.1043, -0.0952, -0.0953, ..., -0.0617, -0.0524, -0.0851], + [ 0.0147, -0.0225, -0.0634, ..., -0.0736, 0.0230, -0.0211]], + device='cuda:0'), grad: tensor([[ 0.0014, 0.0002, 0.0001, ..., 0.0005, 0.0002, 0.0015], + [-0.0104, -0.0020, -0.0008, ..., -0.0086, -0.0015, -0.0143], + [ 0.0028, 0.0007, 0.0005, ..., 0.0019, 0.0004, 0.0032], + ..., + [ 0.0053, 0.0004, 0.0002, ..., 0.0018, 0.0029, 0.0184], + [ 0.0027, 0.0008, 0.0004, ..., 0.0014, 0.0005, 0.0025], + [-0.0068, 0.0003, 0.0002, ..., 0.0015, -0.0033, -0.0151]], + device='cuda:0') +Epoch 243, bias, value: tensor([ 0.0104, -0.0021, 0.0011, -0.0194, 0.0110, -0.0060, -0.0142, -0.0255, + 0.0062, 0.0004], device='cuda:0'), grad: tensor([ 0.0165, -0.0969, 0.0259, 0.0168, -0.0047, 0.0207, 0.0181, 0.0495, + 0.0203, -0.0662], device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 216.43, cls_loss 0.5100 cls_loss_mapping 0.0066 cls_loss_causal 0.4834 re_mapping 0.0069 re_causal 0.0176 /// teacc 98.64 lr 0.00010000 +Epoch 244, weight, value: tensor([[ 0.0294, 0.0998, -0.1558, ..., -0.0912, 0.0655, 0.0072], + [-0.0808, -0.0970, 0.0862, ..., -0.0324, -0.0962, 0.0537], + [-0.0664, -0.0526, -0.1081, ..., -0.1226, -0.0194, 0.0498], + ..., + [-0.0176, -0.1129, -0.1119, ..., -0.1062, -0.0327, 0.0617], + [-0.1041, -0.0967, -0.0960, ..., -0.0622, -0.0519, -0.0866], + [ 0.0151, -0.0221, -0.0630, ..., -0.0745, 0.0235, -0.0205]], + device='cuda:0'), grad: tensor([[-2.0733e-03, -1.8430e-04, 3.2473e-04, ..., 3.7718e-04, + -1.6251e-03, 2.8496e-03], + [ 6.4421e-04, 1.0020e-04, 1.7810e-04, ..., -2.2202e-03, + 2.9951e-05, 6.9284e-04], + [-3.9482e-03, 8.2433e-05, 2.3985e-04, ..., -1.6079e-03, + 2.1911e-04, -2.0767e-02], + ..., + [-4.0131e-03, 7.2539e-05, 3.5238e-04, ..., 4.9210e-04, + 6.8426e-05, -7.2441e-03], + [ 2.2850e-03, 1.1927e-04, 5.3835e-04, ..., 2.4738e-03, + 3.0422e-04, 2.7618e-03], + [ 2.1696e-04, 2.0170e-04, -2.2278e-03, ..., 3.5119e-04, + -2.6646e-03, 2.2125e-03]], device='cuda:0') +Epoch 244, bias, value: tensor([ 0.0105, -0.0018, 0.0011, -0.0200, 0.0116, -0.0060, -0.0144, -0.0262, + 0.0066, 0.0006], device='cuda:0'), grad: tensor([ 0.0177, 0.0151, -0.0612, -0.0254, 0.0211, 0.0070, -0.0155, -0.0501, + 0.0676, 0.0238], device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 216.60, cls_loss 0.5265 cls_loss_mapping 0.0058 cls_loss_causal 0.5058 re_mapping 0.0068 re_causal 0.0184 /// teacc 98.76 lr 0.00010000 +Epoch 245, weight, value: tensor([[ 0.0307, 0.1007, -0.1563, ..., -0.0912, 0.0654, 0.0077], + [-0.0801, -0.0967, 0.0865, ..., -0.0321, -0.0961, 0.0533], + [-0.0681, -0.0529, -0.1081, ..., -0.1232, -0.0203, 0.0492], + ..., + [-0.0175, -0.1127, -0.1128, ..., -0.1058, -0.0328, 0.0623], + [-0.1046, -0.0979, -0.0963, ..., -0.0622, -0.0518, -0.0871], + [ 0.0147, -0.0232, -0.0634, ..., -0.0756, 0.0232, -0.0208]], + device='cuda:0'), grad: tensor([[ 2.2526e-03, 1.6344e-04, 5.8651e-05, ..., 6.3944e-04, + 2.7180e-04, 2.2278e-03], + [-2.0618e-03, 9.2030e-05, -1.1120e-03, ..., 6.8307e-05, + 1.4675e-04, -2.5463e-03], + [ 9.0218e-04, 8.2433e-05, 3.3319e-05, ..., 1.2290e-04, + -9.5844e-05, -3.9291e-03], + ..., + [ 5.6744e-04, 1.0794e-04, 5.3453e-04, ..., 5.2571e-05, + 1.8668e-04, 4.1161e-03], + [ 4.0412e-04, 3.8934e-04, 1.1909e-04, ..., 2.8825e-04, + 1.1492e-04, 8.7547e-04], + [ 1.5707e-03, 1.0061e-03, 2.3532e-04, ..., 4.4417e-04, + 1.8048e-04, -1.6766e-03]], device='cuda:0') +Epoch 245, bias, value: tensor([ 0.0108, -0.0014, 0.0005, -0.0207, 0.0113, -0.0063, -0.0145, -0.0245, + 0.0060, 0.0007], device='cuda:0'), grad: tensor([ 0.0309, -0.0144, -0.0076, -0.0635, 0.0048, 0.0295, -0.0177, 0.0146, + 0.0144, 0.0090], device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 216.25, cls_loss 0.4879 cls_loss_mapping 0.0052 cls_loss_causal 0.4630 re_mapping 0.0068 re_causal 0.0170 /// teacc 98.64 lr 0.00010000 +Epoch 246, weight, value: tensor([[ 0.0311, 0.1008, -0.1568, ..., -0.0915, 0.0655, 0.0070], + [-0.0811, -0.0969, 0.0877, ..., -0.0312, -0.0969, 0.0534], + [-0.0680, -0.0540, -0.1091, ..., -0.1235, -0.0196, 0.0491], + ..., + [-0.0163, -0.1119, -0.1134, ..., -0.1056, -0.0317, 0.0624], + [-0.1047, -0.0983, -0.0967, ..., -0.0625, -0.0525, -0.0875], + [ 0.0152, -0.0238, -0.0628, ..., -0.0764, 0.0233, -0.0207]], + device='cuda:0'), grad: tensor([[ 3.6383e-04, 7.9036e-05, 2.7269e-05, ..., 1.2897e-05, + 1.0395e-04, 3.8695e-04], + [ 6.5327e-04, 1.7345e-04, 2.1502e-05, ..., 1.0515e-06, + 1.4091e-04, -5.4970e-03], + [ 3.7909e-04, 1.7560e-04, 6.2406e-05, ..., 1.5363e-05, + 9.5963e-05, 5.1651e-03], + ..., + [-5.1575e-03, -5.0049e-03, -5.3024e-03, ..., -7.1335e-04, + 2.4676e-04, -2.3308e-03], + [ 1.3275e-03, 7.1573e-04, 5.8937e-04, ..., 7.1466e-05, + 1.8048e-04, 1.0033e-03], + [-2.3518e-03, 5.7554e-04, 3.1799e-05, ..., 7.3574e-06, + 4.0603e-04, -3.1071e-03]], device='cuda:0') +Epoch 246, bias, value: tensor([ 0.0097, -0.0015, 0.0013, -0.0199, 0.0102, -0.0063, -0.0152, -0.0237, + 0.0055, 0.0015], device='cuda:0'), grad: tensor([ 0.0066, -0.0128, 0.0231, -0.0022, -0.0145, 0.0167, 0.0044, -0.0464, + 0.0329, -0.0078], device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 216.40, cls_loss 0.5211 cls_loss_mapping 0.0079 cls_loss_causal 0.4950 re_mapping 0.0071 re_causal 0.0185 /// teacc 98.77 lr 0.00010000 +Epoch 247, weight, value: tensor([[ 0.0311, 0.1006, -0.1582, ..., -0.0918, 0.0656, 0.0069], + [-0.0811, -0.0972, 0.0886, ..., -0.0302, -0.0975, 0.0539], + [-0.0684, -0.0546, -0.1079, ..., -0.1238, -0.0198, 0.0491], + ..., + [-0.0166, -0.1120, -0.1140, ..., -0.1064, -0.0323, 0.0619], + [-0.1048, -0.0984, -0.0970, ..., -0.0625, -0.0521, -0.0879], + [ 0.0156, -0.0239, -0.0617, ..., -0.0761, 0.0232, -0.0201]], + device='cuda:0'), grad: tensor([[ 1.8473e-03, 5.9319e-04, 1.8673e-06, ..., 1.4794e-04, + 8.9121e-04, 6.9237e-04], + [-5.5552e-04, -4.2844e-04, 1.3318e-07, ..., -9.6977e-05, + 8.9407e-05, -1.6642e-03], + [ 9.6369e-04, 8.7404e-04, 1.0930e-05, ..., 8.8215e-05, + 2.5702e-04, 5.4979e-04], + ..., + [-4.3373e-03, -1.5812e-03, -1.5879e-06, ..., 9.1136e-05, + -3.7575e-03, -1.6928e-03], + [ 9.0885e-04, 4.0817e-04, 5.0575e-05, ..., 1.6344e-04, + 1.9884e-04, 4.0102e-04], + [-7.0477e-04, -6.2895e-04, 1.3806e-05, ..., 1.7428e-04, + 7.7248e-04, -6.8331e-04]], device='cuda:0') +Epoch 247, bias, value: tensor([ 0.0103, -0.0014, 0.0006, -0.0193, 0.0106, -0.0052, -0.0148, -0.0250, + 0.0048, 0.0014], device='cuda:0'), grad: tensor([ 0.0263, -0.0143, 0.0246, 0.0370, 0.0251, -0.0351, -0.0047, -0.0666, + 0.0187, -0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 216.61, cls_loss 0.4804 cls_loss_mapping 0.0043 cls_loss_causal 0.4577 re_mapping 0.0074 re_causal 0.0198 /// teacc 98.85 lr 0.00010000 +Epoch 248, weight, value: tensor([[ 0.0312, 0.1007, -0.1566, ..., -0.0901, 0.0657, 0.0066], + [-0.0812, -0.0980, 0.0878, ..., -0.0299, -0.0971, 0.0538], + [-0.0692, -0.0556, -0.1080, ..., -0.1235, -0.0208, 0.0486], + ..., + [-0.0152, -0.1119, -0.1133, ..., -0.1062, -0.0320, 0.0628], + [-0.1052, -0.0971, -0.0960, ..., -0.0622, -0.0519, -0.0883], + [ 0.0158, -0.0245, -0.0631, ..., -0.0763, 0.0221, -0.0205]], + device='cuda:0'), grad: tensor([[-6.4373e-05, -6.4278e-04, 2.9400e-05, ..., -7.7844e-05, + -5.0783e-04, 3.1066e-04], + [-8.8644e-04, 4.0025e-05, 1.1516e-04, ..., 1.9193e-04, + -6.5386e-05, -4.3249e-04], + [-2.8133e-03, -7.6294e-03, -1.3649e-02, ..., -8.8120e-03, + -3.0537e-03, 5.7030e-04], + ..., + [-8.8406e-04, 5.7906e-05, 1.2450e-05, ..., 8.0615e-06, + -6.8367e-05, -4.0722e-04], + [ 3.3951e-04, -2.3687e-04, -3.2568e-04, ..., -2.9087e-04, + 6.7174e-05, 5.2500e-04], + [-1.3809e-03, 8.9407e-05, 1.1899e-05, ..., 1.1213e-05, + -9.1434e-05, -2.6264e-03]], device='cuda:0') +Epoch 248, bias, value: tensor([ 0.0094, -0.0018, 0.0006, -0.0187, 0.0096, -0.0055, -0.0137, -0.0243, + 0.0049, 0.0014], device='cuda:0'), grad: tensor([ 0.0068, -0.0201, -0.0187, 0.0388, 0.0095, 0.0144, 0.0098, -0.0203, + 0.0019, -0.0220], device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 216.49, cls_loss 0.4957 cls_loss_mapping 0.0062 cls_loss_causal 0.4740 re_mapping 0.0067 re_causal 0.0170 /// teacc 98.71 lr 0.00010000 +Epoch 249, weight, value: tensor([[ 0.0307, 0.1008, -0.1575, ..., -0.0913, 0.0654, 0.0066], + [-0.0820, -0.0965, 0.0893, ..., -0.0303, -0.0980, 0.0531], + [-0.0693, -0.0549, -0.1070, ..., -0.1233, -0.0200, 0.0489], + ..., + [-0.0158, -0.1117, -0.1128, ..., -0.1075, -0.0321, 0.0631], + [-0.1051, -0.0971, -0.0960, ..., -0.0613, -0.0521, -0.0871], + [ 0.0159, -0.0261, -0.0630, ..., -0.0764, 0.0215, -0.0217]], + device='cuda:0'), grad: tensor([[ 1.2660e-04, -1.1921e-05, 1.0524e-06, ..., 3.5465e-05, + 1.2107e-05, 1.1663e-03], + [-2.4819e-04, 5.4203e-06, -9.5069e-06, ..., 5.6744e-05, + 8.7082e-05, 2.7142e-03], + [ 2.4223e-04, 6.3896e-05, 3.7197e-06, ..., 1.3614e-04, + 1.1140e-04, 2.0447e-03], + ..., + [-1.7033e-03, -4.9496e-04, 2.8744e-05, ..., 5.5224e-05, + 2.6250e-04, -4.8103e-03], + [ 5.2273e-05, 1.3344e-05, 6.6534e-06, ..., 5.8293e-05, + 3.6895e-05, 9.0933e-04], + [ 2.8402e-05, -8.6613e-08, -2.5600e-05, ..., 3.5226e-05, + -2.4997e-06, 2.0256e-03]], device='cuda:0') +Epoch 249, bias, value: tensor([ 0.0082, -0.0022, 0.0006, -0.0187, 0.0092, -0.0058, -0.0132, -0.0238, + 0.0054, 0.0019], device='cuda:0'), grad: tensor([ 0.0125, -0.0061, 0.0231, -0.0167, -0.0024, -0.0489, -0.0087, 0.0109, + 0.0184, 0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 216.03, cls_loss 0.5060 cls_loss_mapping 0.0065 cls_loss_causal 0.4831 re_mapping 0.0069 re_causal 0.0183 /// teacc 98.75 lr 0.00010000 +Epoch 250, weight, value: tensor([[ 0.0311, 0.1000, -0.1571, ..., -0.0920, 0.0654, 0.0066], + [-0.0820, -0.0966, 0.0894, ..., -0.0303, -0.0978, 0.0520], + [-0.0699, -0.0547, -0.1081, ..., -0.1222, -0.0199, 0.0489], + ..., + [-0.0150, -0.1103, -0.1115, ..., -0.1078, -0.0315, 0.0644], + [-0.1051, -0.0967, -0.0963, ..., -0.0614, -0.0526, -0.0871], + [ 0.0153, -0.0262, -0.0636, ..., -0.0781, 0.0218, -0.0212]], + device='cuda:0'), grad: tensor([[ 1.4563e-03, 1.2720e-04, 1.5593e-04, ..., 7.3957e-04, + 4.5562e-04, 2.2545e-03], + [-1.5850e-03, -4.9973e-04, -4.6778e-04, ..., -7.4482e-04, + -2.7370e-04, 2.1439e-03], + [-2.9850e-04, -4.5395e-04, -6.9380e-04, ..., -3.0136e-03, + -3.8683e-05, -4.7417e-03], + ..., + [ 1.4620e-03, 1.2034e-04, 1.4579e-04, ..., 8.9502e-04, + 6.1274e-04, 2.0874e-02], + [-7.4434e-04, 1.2076e-04, 1.3459e-04, ..., 7.0286e-04, + -1.8730e-03, 3.2157e-05], + [ 1.5497e-03, 8.5294e-05, 9.6321e-05, ..., 4.9973e-04, + 6.2370e-04, 3.4466e-03]], device='cuda:0') +Epoch 250, bias, value: tensor([ 0.0085, -0.0024, -0.0001, -0.0188, 0.0084, -0.0051, -0.0127, -0.0236, + 0.0056, 0.0018], device='cuda:0'), grad: tensor([ 0.0239, -0.0201, -0.0245, 0.0241, -0.0974, -0.0135, 0.0385, 0.0226, + 0.0217, 0.0247], device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 216.34, cls_loss 0.5300 cls_loss_mapping 0.0089 cls_loss_causal 0.5002 re_mapping 0.0070 re_causal 0.0174 /// teacc 98.78 lr 0.00010000 +Epoch 251, weight, value: tensor([[ 0.0308, 0.1005, -0.1578, ..., -0.0930, 0.0656, 0.0067], + [-0.0819, -0.0964, 0.0891, ..., -0.0306, -0.0978, 0.0526], + [-0.0703, -0.0560, -0.1089, ..., -0.1223, -0.0207, 0.0490], + ..., + [-0.0152, -0.1133, -0.1127, ..., -0.1093, -0.0317, 0.0637], + [-0.1051, -0.0968, -0.0962, ..., -0.0609, -0.0519, -0.0888], + [ 0.0151, -0.0268, -0.0628, ..., -0.0785, 0.0225, -0.0205]], + device='cuda:0'), grad: tensor([[-2.0943e-03, -3.5262e-04, 1.9833e-05, ..., -1.3371e-03, + -2.8305e-03, -3.8576e-04], + [-3.2830e-04, 1.0014e-04, 1.1486e-04, ..., -1.2703e-03, + -1.1883e-03, -2.0561e-03], + [ 7.8964e-04, 2.3675e-04, -1.8921e-03, ..., 6.3372e-04, + 1.0910e-03, -6.0234e-03], + ..., + [ 1.1339e-03, 1.9252e-04, 2.7966e-04, ..., 4.5919e-04, + 5.4932e-04, 1.0967e-03], + [ 1.3285e-03, 3.3498e-05, 7.7295e-04, ..., 8.4305e-04, + 1.0843e-03, -1.5039e-03], + [ 5.3444e-03, 1.6165e-04, 5.8383e-05, ..., 3.8981e-04, + 1.5955e-03, 8.3399e-04]], device='cuda:0') +Epoch 251, bias, value: tensor([ 0.0084, -0.0028, -0.0006, -0.0179, 0.0084, -0.0049, -0.0123, -0.0232, + 0.0052, 0.0013], device='cuda:0'), grad: tensor([-0.0249, -0.0309, -0.0203, 0.0319, -0.0031, 0.0065, -0.0006, 0.0195, + -0.0007, 0.0226], device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 216.24, cls_loss 0.5143 cls_loss_mapping 0.0053 cls_loss_causal 0.4894 re_mapping 0.0074 re_causal 0.0193 /// teacc 98.85 lr 0.00010000 +Epoch 252, weight, value: tensor([[ 0.0301, 0.1005, -0.1603, ..., -0.0937, 0.0650, 0.0065], + [-0.0818, -0.0972, 0.0905, ..., -0.0309, -0.0980, 0.0526], + [-0.0711, -0.0550, -0.1081, ..., -0.1241, -0.0206, 0.0493], + ..., + [-0.0145, -0.1129, -0.1146, ..., -0.1084, -0.0325, 0.0645], + [-0.1048, -0.0965, -0.0963, ..., -0.0608, -0.0517, -0.0888], + [ 0.0144, -0.0275, -0.0627, ..., -0.0798, 0.0222, -0.0210]], + device='cuda:0'), grad: tensor([[-0.0042, -0.0042, 0.0004, ..., -0.0023, -0.0017, 0.0009], + [-0.0031, 0.0003, -0.0020, ..., -0.0026, -0.0019, -0.0046], + [ 0.0002, 0.0004, 0.0003, ..., -0.0004, 0.0002, 0.0035], + ..., + [ 0.0034, 0.0051, 0.0004, ..., 0.0017, 0.0012, 0.0093], + [ 0.0011, 0.0004, 0.0005, ..., 0.0006, 0.0005, -0.0089], + [ 0.0009, 0.0065, 0.0007, ..., 0.0004, 0.0003, 0.0029]], + device='cuda:0') +Epoch 252, bias, value: tensor([ 0.0087, -0.0029, -0.0011, -0.0180, 0.0089, -0.0045, -0.0130, -0.0231, + 0.0056, 0.0011], device='cuda:0'), grad: tensor([ 0.0231, -0.0593, -0.0188, -0.0084, 0.0162, -0.0720, 0.0283, 0.0665, + -0.0058, 0.0302], device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 215.96, cls_loss 0.5084 cls_loss_mapping 0.0063 cls_loss_causal 0.4858 re_mapping 0.0065 re_causal 0.0164 /// teacc 98.84 lr 0.00010000 +Epoch 253, weight, value: tensor([[ 0.0320, 0.1012, -0.1597, ..., -0.0933, 0.0656, 0.0067], + [-0.0809, -0.0955, 0.0915, ..., -0.0296, -0.0975, 0.0529], + [-0.0717, -0.0553, -0.1083, ..., -0.1237, -0.0204, 0.0516], + ..., + [-0.0149, -0.1132, -0.1149, ..., -0.1105, -0.0329, 0.0633], + [-0.1056, -0.0970, -0.0959, ..., -0.0591, -0.0531, -0.0890], + [ 0.0146, -0.0272, -0.0626, ..., -0.0796, 0.0209, -0.0211]], + device='cuda:0'), grad: tensor([[-4.9305e-04, -9.4366e-04, 2.7323e-04, ..., -1.0290e-03, + -1.0777e-03, 4.4751e-04], + [-3.0231e-04, 3.0026e-06, -3.4046e-04, ..., 5.1260e-05, + 2.0508e-06, -1.0967e-03], + [ 2.3711e-04, 8.2776e-06, 2.0981e-04, ..., 5.1916e-05, + 8.6352e-06, 7.5960e-04], + ..., + [ 3.0303e-04, 6.5446e-05, 2.0385e-04, ..., 5.9873e-05, + 5.8770e-05, 1.2197e-03], + [ 3.5429e-04, -3.5577e-07, 1.9503e-04, ..., 1.2767e-04, + 1.9237e-05, 1.3170e-03], + [ 5.9046e-06, -2.3866e-04, 1.9431e-04, ..., 1.2122e-05, + -1.8692e-04, 7.1144e-04]], device='cuda:0') +Epoch 253, bias, value: tensor([ 0.0089, -0.0020, -0.0009, -0.0185, 0.0087, -0.0051, -0.0136, -0.0233, + 0.0054, 0.0018], device='cuda:0'), grad: tensor([ 0.0070, -0.0261, 0.0075, -0.0217, 0.0125, 0.0069, -0.0184, 0.0113, + 0.0131, 0.0079], device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 216.01, cls_loss 0.5008 cls_loss_mapping 0.0064 cls_loss_causal 0.4804 re_mapping 0.0072 re_causal 0.0185 /// teacc 98.76 lr 0.00010000 +Epoch 254, weight, value: tensor([[ 0.0300, 0.1007, -0.1602, ..., -0.0924, 0.0643, 0.0063], + [-0.0808, -0.0952, 0.0919, ..., -0.0295, -0.0963, 0.0524], + [-0.0722, -0.0568, -0.1091, ..., -0.1246, -0.0212, 0.0511], + ..., + [-0.0150, -0.1133, -0.1156, ..., -0.1113, -0.0332, 0.0642], + [-0.1061, -0.0962, -0.0963, ..., -0.0595, -0.0520, -0.0883], + [ 0.0152, -0.0273, -0.0631, ..., -0.0794, 0.0218, -0.0214]], + device='cuda:0'), grad: tensor([[-1.0881e-03, -2.5253e-03, -6.0892e-04, ..., -1.8253e-03, + -7.2098e-04, -2.4586e-03], + [ 4.4537e-04, 9.8050e-05, 1.4019e-04, ..., 2.9922e-04, + 1.6403e-04, 1.0891e-03], + [-3.2663e-04, -8.8120e-04, -4.9305e-04, ..., 2.0993e-04, + 6.2323e-04, -1.6769e-02], + ..., + [-5.0843e-05, -6.2943e-04, 5.7411e-04, ..., 6.1941e-04, + 5.4455e-04, 1.6830e-02], + [ 7.6294e-04, 3.8028e-04, 3.4690e-04, ..., 4.1127e-04, + 3.0708e-04, 8.4686e-04], + [ 6.0415e-04, 5.1689e-04, 4.7803e-04, ..., 5.3453e-04, + 4.2391e-04, 1.7891e-03]], device='cuda:0') +Epoch 254, bias, value: tensor([ 0.0088, -0.0024, -0.0014, -0.0189, 0.0086, -0.0049, -0.0129, -0.0226, + 0.0057, 0.0016], device='cuda:0'), grad: tensor([-0.0381, 0.0089, -0.0053, 0.0394, 0.0147, -0.0355, -0.0197, 0.0157, + 0.0073, 0.0124], device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 216.09, cls_loss 0.5033 cls_loss_mapping 0.0041 cls_loss_causal 0.4765 re_mapping 0.0074 re_causal 0.0192 /// teacc 98.72 lr 0.00010000 +Epoch 255, weight, value: tensor([[ 0.0292, 0.0992, -0.1616, ..., -0.0927, 0.0642, 0.0069], + [-0.0814, -0.0955, 0.0917, ..., -0.0304, -0.0965, 0.0514], + [-0.0701, -0.0550, -0.1069, ..., -0.1234, -0.0211, 0.0522], + ..., + [-0.0141, -0.1132, -0.1162, ..., -0.1098, -0.0332, 0.0638], + [-0.1075, -0.0964, -0.0958, ..., -0.0590, -0.0522, -0.0881], + [ 0.0155, -0.0269, -0.0627, ..., -0.0783, 0.0216, -0.0210]], + device='cuda:0'), grad: tensor([[ 1.0765e-02, 8.8196e-03, 4.2170e-05, ..., 3.3302e-03, + 5.3711e-03, 5.7831e-03], + [-9.4795e-04, -3.3069e-04, -1.1845e-03, ..., -1.3618e-03, + 1.7965e-04, -2.5749e-03], + [ 2.4204e-03, 4.6372e-04, 3.8147e-04, ..., 4.6825e-04, + 1.0481e-03, 3.6755e-03], + ..., + [ 2.1458e-03, 6.3753e-04, 6.3121e-05, ..., 3.4189e-04, + 7.3910e-04, -8.3971e-04], + [ 8.6164e-04, 4.8971e-04, 7.0906e-04, ..., 4.6492e-04, + 6.1131e-04, 3.4161e-03], + [-1.1040e-02, -1.1467e-02, 3.9458e-05, ..., -3.4561e-03, + -4.9706e-03, -5.5466e-03]], device='cuda:0') +Epoch 255, bias, value: tensor([ 0.0089, -0.0018, -0.0009, -0.0180, 0.0079, -0.0055, -0.0127, -0.0224, + 0.0044, 0.0015], device='cuda:0'), grad: tensor([ 0.0466, -0.0165, 0.0283, 0.0013, 0.0184, -0.0017, -0.0125, 0.0116, + -0.0084, -0.0671], device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 216.21, cls_loss 0.5176 cls_loss_mapping 0.0037 cls_loss_causal 0.4914 re_mapping 0.0068 re_causal 0.0188 /// teacc 98.72 lr 0.00010000 +Epoch 256, weight, value: tensor([[ 0.0299, 0.1000, -0.1613, ..., -0.0940, 0.0650, 0.0071], + [-0.0812, -0.0958, 0.0912, ..., -0.0309, -0.0966, 0.0517], + [-0.0692, -0.0549, -0.1069, ..., -0.1247, -0.0208, 0.0518], + ..., + [-0.0148, -0.1145, -0.1153, ..., -0.1101, -0.0335, 0.0635], + [-0.1076, -0.0968, -0.0965, ..., -0.0588, -0.0530, -0.0865], + [ 0.0151, -0.0259, -0.0626, ..., -0.0786, 0.0211, -0.0218]], + device='cuda:0'), grad: tensor([[-1.8167e-03, -2.7065e-03, 1.4496e-04, ..., 5.1767e-05, + 1.5891e-04, 1.6737e-03], + [-9.4700e-04, 5.8711e-05, -4.0855e-03, ..., -8.0943e-05, + 1.9670e-04, -1.4439e-03], + [-2.6107e-04, 2.1589e-04, 4.2725e-04, ..., 2.0266e-05, + 3.0470e-04, -4.7455e-03], + ..., + [-6.3553e-03, 3.1781e-04, 2.9206e-04, ..., 1.2517e-04, + -2.3365e-03, -1.1726e-02], + [ 9.0170e-04, 2.3854e-04, 3.4332e-04, ..., -6.0797e-05, + 1.2887e-04, 6.8169e-03], + [-4.5252e-04, 3.0875e-04, 2.8753e-04, ..., -6.9976e-05, + 6.4194e-05, -7.3147e-04]], device='cuda:0') +Epoch 256, bias, value: tensor([ 0.0087, -0.0016, -0.0007, -0.0179, 0.0084, -0.0055, -0.0130, -0.0225, + 0.0042, 0.0014], device='cuda:0'), grad: tensor([-0.0039, -0.0615, -0.0225, 0.0251, 0.0541, 0.0352, -0.0517, -0.0034, + 0.0367, -0.0081], device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 216.65, cls_loss 0.4882 cls_loss_mapping 0.0038 cls_loss_causal 0.4638 re_mapping 0.0072 re_causal 0.0193 /// teacc 98.85 lr 0.00010000 +Epoch 257, weight, value: tensor([[ 0.0299, 0.1010, -0.1599, ..., -0.0954, 0.0655, 0.0063], + [-0.0817, -0.0961, 0.0910, ..., -0.0322, -0.0975, 0.0525], + [-0.0702, -0.0561, -0.1061, ..., -0.1233, -0.0218, 0.0518], + ..., + [-0.0140, -0.1148, -0.1160, ..., -0.1107, -0.0337, 0.0638], + [-0.1068, -0.0951, -0.0957, ..., -0.0578, -0.0518, -0.0856], + [ 0.0149, -0.0262, -0.0630, ..., -0.0797, 0.0219, -0.0227]], + device='cuda:0'), grad: tensor([[ 2.7161e-03, 2.7714e-03, 2.3556e-04, ..., 2.6360e-03, + -6.7024e-03, 1.2035e-03], + [-7.8440e-04, 1.8442e-04, 1.1280e-05, ..., 1.0099e-03, + 3.5949e-06, -2.6054e-03], + [ 3.2091e-04, 1.9622e-04, 4.0263e-05, ..., 9.4295e-05, + 4.1306e-05, 2.2030e-03], + ..., + [-1.3626e-04, 1.9050e-04, 3.1805e-04, ..., 1.8036e-04, + 1.6069e-04, -8.5449e-03], + [ 3.0746e-03, -4.9448e-04, -3.7122e-04, ..., -3.2864e-03, + 1.3704e-03, 3.6354e-03], + [ 8.3065e-04, 5.7125e-04, 4.7404e-07, ..., 6.8855e-04, + 1.0086e-02, 1.4210e-03]], device='cuda:0') +Epoch 257, bias, value: tensor([ 0.0087, -0.0011, -0.0014, -0.0181, 0.0081, -0.0049, -0.0137, -0.0220, + 0.0053, 0.0006], device='cuda:0'), grad: tensor([ 0.0066, -0.0358, 0.0191, 0.0193, -0.0431, -0.0057, -0.0030, -0.0306, + 0.0140, 0.0593], device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 216.68, cls_loss 0.5186 cls_loss_mapping 0.0056 cls_loss_causal 0.4975 re_mapping 0.0072 re_causal 0.0187 /// teacc 98.77 lr 0.00010000 +Epoch 258, weight, value: tensor([[ 0.0310, 0.1023, -0.1595, ..., -0.0949, 0.0661, 0.0061], + [-0.0816, -0.0963, 0.0914, ..., -0.0316, -0.0970, 0.0515], + [-0.0707, -0.0562, -0.1068, ..., -0.1242, -0.0231, 0.0519], + ..., + [-0.0145, -0.1144, -0.1161, ..., -0.1108, -0.0337, 0.0642], + [-0.1061, -0.0956, -0.0962, ..., -0.0579, -0.0500, -0.0858], + [ 0.0152, -0.0259, -0.0616, ..., -0.0792, 0.0206, -0.0236]], + device='cuda:0'), grad: tensor([[ 2.2087e-03, 8.6546e-04, 3.3140e-04, ..., 3.7265e-04, + 1.1206e-03, -1.1606e-03], + [-1.8005e-03, -1.0979e-02, -8.3847e-03, ..., -3.5572e-04, + -3.8648e-04, -3.5648e-03], + [ 1.8966e-04, 8.7118e-04, 6.7186e-04, ..., 5.9903e-05, + 6.8486e-05, -8.4400e-04], + ..., + [ 5.0783e-04, 4.7684e-04, 6.8998e-04, ..., 8.9645e-05, + 4.1544e-05, 6.0987e-04], + [ 2.8205e-04, 1.8203e-04, 2.5177e-04, ..., 5.8979e-05, + 5.3912e-05, 1.3590e-03], + [-1.5998e-04, 8.1015e-04, 6.4182e-04, ..., 1.3828e-04, + 1.3983e-04, 6.5422e-04]], device='cuda:0') +Epoch 258, bias, value: tensor([ 0.0090, -0.0005, -0.0010, -0.0178, 0.0077, -0.0050, -0.0147, -0.0223, + 0.0055, 0.0005], device='cuda:0'), grad: tensor([ 0.0055, -0.0774, 0.0031, 0.0234, 0.0032, 0.0085, 0.0069, 0.0100, + 0.0103, 0.0065], device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 216.42, cls_loss 0.5060 cls_loss_mapping 0.0056 cls_loss_causal 0.4775 re_mapping 0.0067 re_causal 0.0167 /// teacc 98.76 lr 0.00010000 +Epoch 259, weight, value: tensor([[ 0.0310, 0.1015, -0.1607, ..., -0.0940, 0.0647, 0.0058], + [-0.0822, -0.0952, 0.0915, ..., -0.0308, -0.0967, 0.0508], + [-0.0702, -0.0571, -0.1073, ..., -0.1239, -0.0236, 0.0521], + ..., + [-0.0148, -0.1139, -0.1152, ..., -0.1107, -0.0341, 0.0650], + [-0.1068, -0.0968, -0.0977, ..., -0.0588, -0.0503, -0.0859], + [ 0.0156, -0.0260, -0.0616, ..., -0.0799, 0.0209, -0.0241]], + device='cuda:0'), grad: tensor([[ 4.6849e-04, 6.6876e-05, 4.5300e-05, ..., 1.4670e-05, + 1.0633e-04, 6.5517e-04], + [ 1.1539e-03, 4.2176e-04, 2.6703e-04, ..., 2.5034e-04, + 5.8603e-04, 2.6321e-03], + [ 4.0936e-04, -1.4362e-03, 5.9247e-05, ..., -4.9171e-03, + 1.5247e-04, -6.6910e-03], + ..., + [ 8.8215e-04, 2.5773e-04, 2.0981e-04, ..., 2.9206e-05, + 2.2244e-04, 1.5736e-03], + [ 2.9397e-04, 5.4884e-04, 8.8871e-05, ..., 5.4032e-05, + 1.1933e-04, 1.5259e-03], + [-4.9400e-03, 1.4037e-05, -1.2531e-03, ..., -9.0063e-05, + -1.8606e-03, -7.0839e-03]], device='cuda:0') +Epoch 259, bias, value: tensor([ 0.0081, -0.0008, -0.0013, -0.0186, 0.0073, -0.0055, -0.0130, -0.0214, + 0.0061, 0.0004], device='cuda:0'), grad: tensor([ 0.0153, 0.0317, -0.0302, 0.0175, 0.0056, -0.0443, 0.0161, 0.0254, + 0.0470, -0.0842], device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 216.69, cls_loss 0.4969 cls_loss_mapping 0.0054 cls_loss_causal 0.4700 re_mapping 0.0070 re_causal 0.0180 /// teacc 98.93 lr 0.00010000 +Epoch 260, weight, value: tensor([[ 0.0316, 0.1010, -0.1614, ..., -0.0942, 0.0653, 0.0054], + [-0.0818, -0.0948, 0.0914, ..., -0.0294, -0.0954, 0.0508], + [-0.0720, -0.0570, -0.1069, ..., -0.1228, -0.0251, 0.0520], + ..., + [-0.0161, -0.1153, -0.1155, ..., -0.1126, -0.0357, 0.0656], + [-0.1075, -0.0974, -0.0979, ..., -0.0583, -0.0518, -0.0871], + [ 0.0149, -0.0274, -0.0625, ..., -0.0801, 0.0204, -0.0240]], + device='cuda:0'), grad: tensor([[ 4.6119e-06, -1.4200e-03, -2.5120e-03, ..., -3.7174e-03, + 1.0872e-03, 3.2926e-04], + [ 4.6945e-04, -9.3689e-03, -2.5482e-03, ..., -2.1038e-03, + 6.1750e-04, 8.5926e-04], + [-1.5774e-03, -9.5844e-04, -9.7752e-04, ..., 3.8409e-04, + -9.5546e-05, -1.8015e-03], + ..., + [ 1.4496e-04, 1.8024e-04, 1.5867e-04, ..., 1.8144e-04, + 9.0981e-04, -2.4247e-04], + [ 4.2152e-04, 1.7233e-03, 3.6144e-04, ..., 1.5202e-03, + 7.2479e-04, 4.3988e-04], + [ 2.0146e-04, 7.7820e-04, 2.7752e-04, ..., 8.8644e-04, + 1.0803e-05, 8.1539e-04]], device='cuda:0') +Epoch 260, bias, value: tensor([ 0.0075, -0.0004, -0.0022, -0.0179, 0.0076, -0.0056, -0.0128, -0.0211, + 0.0059, 0.0002], device='cuda:0'), grad: tensor([ 0.0013, -0.0093, -0.0595, 0.0271, 0.0164, 0.0201, -0.0182, -0.0156, + 0.0235, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 216.57, cls_loss 0.5192 cls_loss_mapping 0.0056 cls_loss_causal 0.4958 re_mapping 0.0068 re_causal 0.0180 /// teacc 98.78 lr 0.00010000 +Epoch 261, weight, value: tensor([[ 0.0316, 0.1004, -0.1622, ..., -0.0945, 0.0646, 0.0054], + [-0.0819, -0.0950, 0.0899, ..., -0.0303, -0.0956, 0.0508], + [-0.0712, -0.0564, -0.1059, ..., -0.1219, -0.0242, 0.0532], + ..., + [-0.0152, -0.1146, -0.1162, ..., -0.1117, -0.0354, 0.0663], + [-0.1081, -0.0983, -0.0988, ..., -0.0594, -0.0519, -0.0865], + [ 0.0148, -0.0274, -0.0619, ..., -0.0807, 0.0196, -0.0242]], + device='cuda:0'), grad: tensor([[-1.4143e-03, 7.9250e-04, 5.2184e-05, ..., -1.6699e-03, + 9.3281e-05, -5.6381e-03], + [-1.0214e-03, -3.2753e-05, -1.7347e-03, ..., 7.7903e-05, + 6.9942e-07, -1.9054e-03], + [-3.2158e-03, -7.3967e-03, 1.5879e-04, ..., 1.7226e-04, + -1.0328e-03, -6.3667e-03], + ..., + [-1.1692e-03, 7.8344e-04, 3.2830e-04, ..., 1.9681e-04, + 5.8785e-06, 8.2970e-04], + [ 1.8253e-03, 1.5392e-03, 5.4884e-04, ..., 4.5466e-04, + 1.5497e-04, 4.2496e-03], + [ 1.4286e-03, 3.3951e-04, 1.2553e-04, ..., 1.1295e-04, + 2.6539e-05, 1.2178e-03]], device='cuda:0') +Epoch 261, bias, value: tensor([ 0.0066, -0.0005, -0.0003, -0.0185, 0.0070, -0.0055, -0.0138, -0.0203, + 0.0066, -0.0004], device='cuda:0'), grad: tensor([-0.0240, -0.0245, -0.0562, 0.0559, -0.0065, -0.0148, 0.0350, 0.0246, + 0.0123, -0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 216.72, cls_loss 0.4784 cls_loss_mapping 0.0055 cls_loss_causal 0.4523 re_mapping 0.0068 re_causal 0.0181 /// teacc 98.81 lr 0.00010000 +Epoch 262, weight, value: tensor([[ 0.0310, 0.1010, -0.1627, ..., -0.0944, 0.0650, 0.0043], + [-0.0812, -0.0948, 0.0911, ..., -0.0303, -0.0946, 0.0497], + [-0.0716, -0.0542, -0.1051, ..., -0.1205, -0.0236, 0.0526], + ..., + [-0.0154, -0.1157, -0.1175, ..., -0.1112, -0.0349, 0.0664], + [-0.1073, -0.0981, -0.0989, ..., -0.0591, -0.0513, -0.0870], + [ 0.0130, -0.0299, -0.0627, ..., -0.0818, 0.0177, -0.0236]], + device='cuda:0'), grad: tensor([[ 1.7369e-04, -2.9588e-04, -9.5308e-05, ..., -3.8266e-05, + -8.1435e-06, 6.0272e-04], + [-1.7557e-03, 5.6326e-05, -6.3442e-06, ..., 1.1418e-06, + -1.3943e-03, -1.8663e-03], + [ 4.5252e-04, 1.6475e-04, 6.5155e-06, ..., 5.6736e-06, + 3.0684e-04, 7.4196e-04], + ..., + [-8.1253e-04, 5.1647e-05, 1.0291e-06, ..., 1.6205e-06, + -4.4584e-04, -3.7403e-03], + [ 7.1859e-04, 4.3941e-04, 1.7524e-05, ..., 4.1395e-05, + 3.6216e-04, 1.9350e-03], + [ 4.6253e-05, 3.2878e-04, 5.4926e-05, ..., 4.8339e-05, + -2.9802e-05, -2.9659e-04]], device='cuda:0') +Epoch 262, bias, value: tensor([ 6.3323e-03, -2.4431e-04, -9.3571e-05, -1.8305e-02, 7.5203e-03, + -5.4886e-03, -1.4011e-02, -2.0989e-02, 6.8823e-03, -5.3389e-04], + device='cuda:0'), grad: tensor([ 0.0115, -0.0443, 0.0146, 0.0217, -0.0140, -0.0074, 0.0124, -0.0230, + 0.0221, 0.0063], device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 216.37, cls_loss 0.4738 cls_loss_mapping 0.0052 cls_loss_causal 0.4530 re_mapping 0.0069 re_causal 0.0175 /// teacc 98.77 lr 0.00010000 +Epoch 263, weight, value: tensor([[ 0.0305, 0.1020, -0.1631, ..., -0.0952, 0.0646, 0.0054], + [-0.0810, -0.0943, 0.0922, ..., -0.0311, -0.0946, 0.0495], + [-0.0723, -0.0540, -0.1057, ..., -0.1203, -0.0243, 0.0521], + ..., + [-0.0144, -0.1161, -0.1186, ..., -0.1107, -0.0342, 0.0667], + [-0.1072, -0.0993, -0.0999, ..., -0.0596, -0.0515, -0.0864], + [ 0.0122, -0.0289, -0.0621, ..., -0.0823, 0.0174, -0.0240]], + device='cuda:0'), grad: tensor([[ 4.6706e-04, -7.0524e-04, 3.8314e-04, ..., 6.4960e-07, + 5.7125e-04, 7.7152e-04], + [ 2.1687e-03, 1.2808e-03, 1.5459e-03, ..., 2.8871e-07, + 2.3136e-03, 2.7142e-03], + [ 2.8253e-04, 1.4143e-03, 2.8419e-04, ..., 8.4611e-07, + 3.1910e-03, 7.0333e-04], + ..., + [ 3.2353e-04, 3.7932e-04, 3.0971e-04, ..., 1.0803e-06, + -8.1406e-03, -6.2370e-04], + [ 4.2868e-04, 9.0075e-04, 6.6090e-04, ..., 5.4808e-07, + 5.1403e-04, 1.1892e-03], + [ 2.5387e-03, 2.0170e-04, 8.2922e-04, ..., 2.0992e-06, + 3.5248e-03, 7.6962e-04]], device='cuda:0') +Epoch 263, bias, value: tensor([ 0.0063, 0.0005, -0.0002, -0.0191, 0.0086, -0.0056, -0.0142, -0.0213, + 0.0065, -0.0002], device='cuda:0'), grad: tensor([ 0.0165, 0.0335, -0.0061, -0.0401, -0.0182, 0.0165, -0.0108, -0.0051, + 0.0205, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 216.43, cls_loss 0.5023 cls_loss_mapping 0.0054 cls_loss_causal 0.4763 re_mapping 0.0069 re_causal 0.0183 /// teacc 98.78 lr 0.00010000 +Epoch 264, weight, value: tensor([[ 0.0297, 0.1020, -0.1623, ..., -0.0958, 0.0656, 0.0046], + [-0.0823, -0.0925, 0.0930, ..., -0.0308, -0.0959, 0.0489], + [-0.0723, -0.0548, -0.1061, ..., -0.1206, -0.0250, 0.0520], + ..., + [-0.0142, -0.1167, -0.1181, ..., -0.1110, -0.0346, 0.0662], + [-0.1081, -0.1002, -0.0997, ..., -0.0585, -0.0533, -0.0864], + [ 0.0132, -0.0287, -0.0625, ..., -0.0821, 0.0189, -0.0221]], + device='cuda:0'), grad: tensor([[ 6.2895e-04, 2.8944e-04, 2.6894e-04, ..., 2.8539e-04, + 2.7871e-04, 1.0452e-03], + [ 1.3447e-03, 2.1064e-04, 7.4565e-05, ..., 5.5265e-04, + 1.1015e-03, 1.8549e-03], + [-2.5082e-03, -6.0921e-03, -3.1643e-03, ..., 9.0241e-05, + 2.1720e-04, -4.9553e-03], + ..., + [ 1.3704e-03, 3.6740e-04, 2.2161e-04, ..., 2.2590e-05, + 2.7394e-04, 2.7561e-03], + [ 1.3515e-05, 9.5224e-04, 6.1226e-04, ..., 2.0933e-04, + 3.4499e-04, -1.1784e-04], + [-1.7822e-04, 2.3341e-04, 7.0691e-05, ..., 5.6922e-05, + 4.5133e-04, -1.4076e-03]], device='cuda:0') +Epoch 264, bias, value: tensor([ 6.0938e-03, 3.5793e-04, -6.4326e-05, -1.8146e-02, 7.9776e-03, + -5.7667e-03, -1.3606e-02, -2.1665e-02, 5.4614e-03, 4.8054e-04], + device='cuda:0'), grad: tensor([ 0.0129, 0.0263, -0.0381, -0.0125, -0.0228, 0.0230, 0.0059, 0.0248, + -0.0105, -0.0089], device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 216.48, cls_loss 0.5033 cls_loss_mapping 0.0056 cls_loss_causal 0.4741 re_mapping 0.0065 re_causal 0.0170 /// teacc 98.78 lr 0.00010000 +Epoch 265, weight, value: tensor([[ 0.0300, 0.1024, -0.1614, ..., -0.0967, 0.0653, 0.0051], + [-0.0833, -0.0925, 0.0925, ..., -0.0320, -0.0972, 0.0487], + [-0.0724, -0.0551, -0.1060, ..., -0.1205, -0.0247, 0.0527], + ..., + [-0.0154, -0.1166, -0.1177, ..., -0.1091, -0.0341, 0.0661], + [-0.1089, -0.1005, -0.0994, ..., -0.0588, -0.0529, -0.0874], + [ 0.0144, -0.0296, -0.0650, ..., -0.0822, 0.0190, -0.0213]], + device='cuda:0'), grad: tensor([[ 2.2736e-03, 3.2310e-03, 4.5216e-07, ..., 1.8597e-05, + 7.0152e-03, 3.1543e-04], + [-1.0538e-03, 1.3545e-05, 2.7660e-07, ..., -2.3210e-04, + -8.9455e-04, -7.6218e-03], + [-1.6308e-04, 2.0540e-04, 8.8103e-07, ..., -9.9838e-05, + 2.5344e-04, -3.7718e-04], + ..., + [ 5.9319e-04, 3.2753e-05, 3.9395e-07, ..., 1.8522e-05, + 3.0446e-04, 4.2081e-04], + [-1.7414e-03, -3.7918e-03, 9.8050e-06, ..., 1.5497e-05, + -5.9204e-03, -9.8324e-04], + [ 2.8205e-04, 9.5510e-04, 1.4342e-06, ..., 1.1325e-05, + -5.9128e-03, 5.8699e-04]], device='cuda:0') +Epoch 265, bias, value: tensor([ 0.0065, -0.0013, -0.0004, -0.0186, 0.0087, -0.0052, -0.0137, -0.0216, + 0.0060, 0.0009], device='cuda:0'), grad: tensor([ 0.0359, -0.0233, 0.0010, 0.0052, 0.0267, 0.0033, 0.0059, 0.0058, + -0.0581, -0.0024], device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 216.57, cls_loss 0.5170 cls_loss_mapping 0.0049 cls_loss_causal 0.4955 re_mapping 0.0065 re_causal 0.0177 /// teacc 98.78 lr 0.00010000 +Epoch 266, weight, value: tensor([[ 0.0315, 0.1019, -0.1617, ..., -0.0970, 0.0664, 0.0045], + [-0.0838, -0.0923, 0.0924, ..., -0.0324, -0.0981, 0.0489], + [-0.0711, -0.0547, -0.1060, ..., -0.1213, -0.0242, 0.0528], + ..., + [-0.0154, -0.1157, -0.1169, ..., -0.1094, -0.0350, 0.0660], + [-0.1087, -0.1012, -0.1007, ..., -0.0582, -0.0518, -0.0876], + [ 0.0155, -0.0296, -0.0649, ..., -0.0824, 0.0195, -0.0213]], + device='cuda:0'), grad: tensor([[ 7.0047e-04, 8.7023e-05, 2.4259e-04, ..., 1.1283e-04, + 3.7575e-04, 4.5133e-04], + [ 5.0497e-04, -7.0524e-04, -8.4877e-04, ..., 6.8784e-05, + -8.7321e-06, -3.7880e-03], + [ 2.0516e-04, 1.7560e-04, 9.8288e-05, ..., 1.6189e-04, + -3.5310e-04, 1.1520e-03], + ..., + [ 5.6839e-04, 9.6679e-05, 7.5221e-05, ..., 4.4167e-05, + 3.4571e-04, 1.2197e-03], + [-1.6050e-03, 2.1231e-04, 2.5058e-04, ..., 2.3139e-04, + -3.9196e-04, -8.3303e-04], + [-2.5043e-03, -8.1825e-04, -1.1265e-05, ..., -5.5742e-04, + -2.5845e-03, -6.8045e-04]], device='cuda:0') +Epoch 266, bias, value: tensor([ 0.0053, -0.0006, 0.0007, -0.0189, 0.0076, -0.0047, -0.0128, -0.0225, + 0.0058, 0.0014], device='cuda:0'), grad: tensor([ 0.0247, -0.0016, -0.0011, -0.0078, -0.0338, 0.0143, 0.0274, 0.0235, + -0.0291, -0.0162], device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 216.83, cls_loss 0.5138 cls_loss_mapping 0.0042 cls_loss_causal 0.4849 re_mapping 0.0069 re_causal 0.0196 /// teacc 98.87 lr 0.00010000 +Epoch 267, weight, value: tensor([[ 0.0319, 0.1026, -0.1620, ..., -0.0971, 0.0667, 0.0049], + [-0.0841, -0.0926, 0.0935, ..., -0.0319, -0.0979, 0.0490], + [-0.0722, -0.0560, -0.1084, ..., -0.1214, -0.0249, 0.0532], + ..., + [-0.0153, -0.1152, -0.1172, ..., -0.1091, -0.0350, 0.0663], + [-0.1090, -0.1000, -0.1007, ..., -0.0586, -0.0508, -0.0877], + [ 0.0155, -0.0304, -0.0652, ..., -0.0814, 0.0201, -0.0209]], + device='cuda:0'), grad: tensor([[ 3.3855e-04, -2.7542e-03, 2.7943e-04, ..., -1.5450e-03, + 8.1837e-05, 4.6587e-04], + [-1.4610e-03, 8.9109e-05, 6.1893e-04, ..., 1.3962e-03, + 2.3055e-04, -1.7700e-03], + [ 9.1553e-04, 1.7166e-03, -4.6182e-04, ..., -1.5366e-04, + 4.6873e-04, 1.1148e-03], + ..., + [-3.7842e-03, 5.3549e-04, -3.7169e-04, ..., 4.1223e-04, + -5.5218e-04, -1.1978e-03], + [ 1.6003e-03, 1.0681e-03, -9.1743e-04, ..., -6.9523e-04, + 4.1699e-04, -1.6222e-03], + [-1.9350e-03, -5.4216e-04, 1.0347e-03, ..., 5.8746e-04, + 4.4537e-04, -2.0561e-03]], device='cuda:0') +Epoch 267, bias, value: tensor([ 0.0061, -0.0006, 0.0012, -0.0192, 0.0075, -0.0052, -0.0140, -0.0218, + 0.0052, 0.0021], device='cuda:0'), grad: tensor([-0.0039, 0.0021, 0.0275, 0.0216, 0.0246, 0.0210, 0.0203, 0.0089, + -0.0694, -0.0526], device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 216.64, cls_loss 0.5085 cls_loss_mapping 0.0055 cls_loss_causal 0.4861 re_mapping 0.0071 re_causal 0.0181 /// teacc 98.62 lr 0.00010000 +Epoch 268, weight, value: tensor([[ 0.0327, 0.1036, -0.1615, ..., -0.0977, 0.0676, 0.0050], + [-0.0830, -0.0932, 0.0939, ..., -0.0328, -0.0980, 0.0494], + [-0.0726, -0.0559, -0.1076, ..., -0.1217, -0.0253, 0.0538], + ..., + [-0.0165, -0.1162, -0.1187, ..., -0.1098, -0.0361, 0.0657], + [-0.1076, -0.0986, -0.1004, ..., -0.0584, -0.0504, -0.0864], + [ 0.0160, -0.0304, -0.0640, ..., -0.0809, 0.0214, -0.0203]], + device='cuda:0'), grad: tensor([[ 1.2946e-04, 4.8894e-08, 6.9384e-08, ..., 6.7689e-06, + 1.0036e-05, 2.7704e-04], + [ 1.6224e-04, 7.1712e-08, -1.6624e-06, ..., 3.4273e-07, + 1.3813e-05, 4.7994e-04], + [-1.5755e-03, 1.6717e-07, -7.0222e-07, ..., 3.6834e-07, + 2.8208e-05, 5.2404e-04], + ..., + [ 1.0455e-04, 4.3884e-06, 5.6857e-07, ..., 1.2759e-07, + -2.2780e-06, -2.0332e-03], + [ 2.0278e-04, 3.1032e-06, 9.9000e-07, ..., 2.4319e-05, + -4.0710e-05, -8.3494e-04], + [ 1.2708e-04, -1.9148e-05, 1.4249e-07, ..., 1.1520e-06, + -8.5458e-06, 3.4189e-04]], device='cuda:0') +Epoch 268, bias, value: tensor([ 0.0061, -0.0008, 0.0016, -0.0195, 0.0075, -0.0062, -0.0139, -0.0225, + 0.0067, 0.0019], device='cuda:0'), grad: tensor([ 0.0168, 0.0223, -0.0075, 0.0233, -0.0429, 0.0232, 0.0190, -0.0140, + -0.0341, -0.0060], device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 216.31, cls_loss 0.4931 cls_loss_mapping 0.0052 cls_loss_causal 0.4658 re_mapping 0.0066 re_causal 0.0171 /// teacc 98.68 lr 0.00010000 +Epoch 269, weight, value: tensor([[ 0.0339, 0.1041, -0.1625, ..., -0.0981, 0.0690, 0.0048], + [-0.0830, -0.0936, 0.0941, ..., -0.0331, -0.0989, 0.0498], + [-0.0714, -0.0555, -0.1067, ..., -0.1204, -0.0247, 0.0535], + ..., + [-0.0166, -0.1167, -0.1188, ..., -0.1100, -0.0359, 0.0662], + [-0.1082, -0.1006, -0.1016, ..., -0.0592, -0.0499, -0.0866], + [ 0.0156, -0.0299, -0.0637, ..., -0.0814, 0.0210, -0.0200]], + device='cuda:0'), grad: tensor([[ 3.0547e-05, 2.0161e-03, -3.4332e-05, ..., 2.1625e-06, + 1.2469e-04, 2.6435e-05], + [ 4.5151e-05, 1.5527e-05, -2.7329e-05, ..., -1.7703e-05, + 1.8394e-04, 1.1779e-05], + [ 3.2157e-05, 5.4026e-04, 1.6198e-05, ..., 3.3788e-06, + 1.3065e-04, -5.1212e-04], + ..., + [ 4.7743e-05, 6.1452e-05, 1.8016e-05, ..., 4.6268e-06, + 1.9431e-04, 3.7122e-04], + [ 3.1382e-05, 1.4675e-04, 4.5925e-05, ..., 1.1843e-04, + 1.2755e-04, 4.3482e-05], + [-1.4043e-04, 4.5925e-05, 2.4229e-05, ..., 1.0580e-06, + -5.7983e-04, -6.9082e-05]], device='cuda:0') +Epoch 269, bias, value: tensor([ 0.0061, -0.0008, 0.0015, -0.0198, 0.0086, -0.0064, -0.0137, -0.0218, + 0.0056, 0.0018], device='cuda:0'), grad: tensor([ 0.0240, 0.0118, 0.0141, -0.0189, -0.0177, -0.0159, -0.0202, 0.0228, + 0.0148, -0.0148], device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 216.25, cls_loss 0.4846 cls_loss_mapping 0.0039 cls_loss_causal 0.4570 re_mapping 0.0067 re_causal 0.0182 /// teacc 98.85 lr 0.00010000 +Epoch 270, weight, value: tensor([[ 0.0342, 0.1047, -0.1634, ..., -0.0978, 0.0690, 0.0062], + [-0.0839, -0.0949, 0.0936, ..., -0.0336, -0.1010, 0.0503], + [-0.0707, -0.0568, -0.1072, ..., -0.1216, -0.0246, 0.0533], + ..., + [-0.0172, -0.1152, -0.1181, ..., -0.1102, -0.0353, 0.0652], + [-0.1078, -0.1008, -0.1024, ..., -0.0594, -0.0479, -0.0863], + [ 0.0160, -0.0306, -0.0639, ..., -0.0822, 0.0204, -0.0200]], + device='cuda:0'), grad: tensor([[ 1.3857e-03, 7.4244e-04, 3.9339e-04, ..., 8.0538e-04, + 1.7185e-03, 1.1816e-03], + [-3.5286e-03, -1.2495e-05, 1.0329e-04, ..., 1.1902e-03, + -8.4257e-04, -3.3295e-02], + [ 1.2398e-03, 8.5068e-04, 5.1832e-04, ..., 5.0259e-04, + 1.3628e-03, -3.7718e-04], + ..., + [ 3.4866e-03, 5.0402e-04, 2.2030e-04, ..., 3.1781e-04, + 1.2608e-03, 2.5681e-02], + [ 9.2125e-04, 7.3576e-04, 4.1246e-04, ..., 1.0328e-03, + 1.1196e-03, 1.0624e-03], + [-5.3024e-03, -4.0665e-03, -3.8929e-03, ..., -1.8377e-03, + -5.8517e-03, 3.8242e-03]], device='cuda:0') +Epoch 270, bias, value: tensor([ 6.8533e-03, -8.2987e-04, 1.5447e-03, -1.8811e-02, 9.1519e-03, + -6.6691e-03, -1.4126e-02, -2.1675e-02, 5.7795e-03, -5.0555e-05], + device='cuda:0'), grad: tensor([ 0.0312, -0.0573, -0.0021, -0.0170, -0.0026, -0.0184, 0.0160, 0.0515, + 0.0274, -0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 216.53, cls_loss 0.4885 cls_loss_mapping 0.0048 cls_loss_causal 0.4603 re_mapping 0.0068 re_causal 0.0181 /// teacc 98.67 lr 0.00010000 +Epoch 271, weight, value: tensor([[ 0.0344, 0.1047, -0.1644, ..., -0.0979, 0.0680, 0.0060], + [-0.0823, -0.0956, 0.0931, ..., -0.0345, -0.0989, 0.0513], + [-0.0703, -0.0560, -0.1069, ..., -0.1215, -0.0253, 0.0531], + ..., + [-0.0175, -0.1146, -0.1175, ..., -0.1094, -0.0334, 0.0659], + [-0.1086, -0.1005, -0.1017, ..., -0.0599, -0.0485, -0.0854], + [ 0.0152, -0.0282, -0.0656, ..., -0.0834, 0.0204, -0.0217]], + device='cuda:0'), grad: tensor([[ 9.7961e-03, 7.7782e-03, 8.4937e-05, ..., 3.6068e-03, + 7.3547e-03, 3.7594e-03], + [ 2.0370e-03, 3.4839e-05, 8.0061e-04, ..., 3.0780e-04, + 1.4315e-03, 4.3335e-03], + [ 6.2084e-04, 3.7766e-04, 1.0836e-04, ..., 9.1732e-05, + 5.1212e-04, -7.5531e-04], + ..., + [-4.3335e-03, 1.9693e-04, -1.5612e-03, ..., -5.3883e-04, + -3.0060e-03, -4.1885e-03], + [-7.7629e-04, -1.4067e-04, 5.6736e-06, ..., 7.7963e-05, + -3.2520e-04, -3.3226e-03], + [ 1.1511e-03, 1.2341e-03, 1.6797e-04, ..., 2.2149e-04, + 8.6117e-04, -3.2673e-03]], device='cuda:0') +Epoch 271, bias, value: tensor([ 0.0069, -0.0001, 0.0010, -0.0187, 0.0087, -0.0073, -0.0134, -0.0208, + 0.0061, -0.0014], device='cuda:0'), grad: tensor([-0.0206, 0.0266, -0.0014, 0.0227, -0.0104, 0.0093, 0.0116, -0.0273, + -0.0012, -0.0093], device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 216.31, cls_loss 0.5048 cls_loss_mapping 0.0040 cls_loss_causal 0.4807 re_mapping 0.0068 re_causal 0.0181 /// teacc 98.79 lr 0.00010000 +Epoch 272, weight, value: tensor([[ 0.0349, 0.1053, -0.1633, ..., -0.0985, 0.0688, 0.0049], + [-0.0824, -0.0953, 0.0939, ..., -0.0335, -0.0986, 0.0519], + [-0.0705, -0.0565, -0.1075, ..., -0.1220, -0.0262, 0.0535], + ..., + [-0.0179, -0.1166, -0.1183, ..., -0.1093, -0.0341, 0.0653], + [-0.1090, -0.1010, -0.1022, ..., -0.0613, -0.0488, -0.0848], + [ 0.0155, -0.0281, -0.0660, ..., -0.0830, 0.0208, -0.0207]], + device='cuda:0'), grad: tensor([[-4.4036e-04, 8.4698e-05, 4.3958e-05, ..., 4.2844e-04, + -6.4564e-04, 1.0885e-05], + [ 8.5449e-04, 1.7488e-04, 1.7214e-04, ..., 5.1308e-04, + 1.4687e-03, 4.6802e-04], + [ 5.2929e-04, 1.9515e-04, 1.8775e-04, ..., 4.2081e-04, + 8.6594e-04, 3.2496e-04], + ..., + [-1.6289e-03, -7.3671e-04, 2.3079e-04, ..., 1.6010e-04, + -2.4261e-03, -1.0633e-03], + [ 8.1968e-04, 7.4863e-04, 8.1015e-04, ..., -1.0031e-04, + 1.4944e-03, 6.1321e-04], + [ 8.2254e-04, 9.7179e-04, 2.1529e-04, ..., 9.3365e-04, + 1.2865e-03, 9.7322e-04]], device='cuda:0') +Epoch 272, bias, value: tensor([ 0.0082, -0.0006, 0.0007, -0.0191, 0.0079, -0.0076, -0.0128, -0.0223, + 0.0070, -0.0002], device='cuda:0'), grad: tensor([-0.0121, 0.0179, 0.0153, -0.0192, 0.0202, -0.0108, 0.0097, -0.0490, + 0.0006, 0.0273], device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 216.35, cls_loss 0.5079 cls_loss_mapping 0.0083 cls_loss_causal 0.4844 re_mapping 0.0065 re_causal 0.0168 /// teacc 98.64 lr 0.00010000 +Epoch 273, weight, value: tensor([[ 0.0337, 0.1045, -0.1636, ..., -0.0982, 0.0677, 0.0055], + [-0.0819, -0.0952, 0.0942, ..., -0.0339, -0.0976, 0.0515], + [-0.0707, -0.0567, -0.1076, ..., -0.1241, -0.0275, 0.0526], + ..., + [-0.0185, -0.1176, -0.1181, ..., -0.1101, -0.0334, 0.0660], + [-0.1092, -0.1026, -0.1026, ..., -0.0613, -0.0504, -0.0843], + [ 0.0156, -0.0266, -0.0653, ..., -0.0843, 0.0223, -0.0202]], + device='cuda:0'), grad: tensor([[-5.8365e-03, -5.9853e-03, -3.0398e-04, ..., 2.9564e-04, + -3.2177e-03, -4.6086e-04], + [ 5.2780e-05, 8.6546e-04, 7.1812e-04, ..., 9.0122e-04, + 6.2609e-04, 1.5755e-03], + [ 1.2741e-03, 1.4191e-03, 2.6202e-04, ..., 2.3377e-04, + 1.7853e-03, 1.0319e-03], + ..., + [ 1.3435e-04, -2.7199e-03, -1.6365e-03, ..., -1.9369e-03, + -1.3180e-03, -1.9722e-03], + [ 2.5845e-04, -7.7171e-03, 2.5749e-04, ..., 4.8447e-04, + 1.4400e-03, 6.3896e-04], + [ 1.7157e-03, 2.2678e-03, 4.5872e-04, ..., 8.0967e-04, + 2.9984e-03, -2.2736e-03]], device='cuda:0') +Epoch 273, bias, value: tensor([ 7.5501e-03, -4.4537e-04, 6.2948e-05, -1.8081e-02, 7.0223e-03, + -7.9334e-03, -1.2432e-02, -2.1873e-02, 5.4944e-03, 1.7605e-03], + device='cuda:0'), grad: tensor([-0.0114, 0.0251, 0.0173, 0.0021, -0.0184, 0.0433, 0.0160, -0.0529, + -0.0204, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 216.85, cls_loss 0.5097 cls_loss_mapping 0.0035 cls_loss_causal 0.4861 re_mapping 0.0063 re_causal 0.0171 /// teacc 98.84 lr 0.00010000 +Epoch 274, weight, value: tensor([[ 0.0342, 0.1038, -0.1642, ..., -0.0996, 0.0673, 0.0041], + [-0.0824, -0.0956, 0.0955, ..., -0.0329, -0.0989, 0.0524], + [-0.0710, -0.0563, -0.1084, ..., -0.1234, -0.0275, 0.0526], + ..., + [-0.0185, -0.1185, -0.1193, ..., -0.1111, -0.0324, 0.0664], + [-0.1098, -0.1024, -0.1024, ..., -0.0613, -0.0504, -0.0851], + [ 0.0161, -0.0266, -0.0654, ..., -0.0840, 0.0221, -0.0210]], + device='cuda:0'), grad: tensor([[ 2.5892e-04, 1.0562e-04, 1.7488e-04, ..., 2.5177e-04, + 6.6710e-04, 7.4291e-04], + [ 4.6492e-04, 2.5725e-04, 3.6120e-04, ..., 5.4741e-04, + 1.0490e-03, 1.7996e-03], + [ 3.1304e-04, 1.7774e-04, 2.7800e-04, ..., 3.4451e-04, + 1.5669e-03, 5.0659e-03], + ..., + [ 9.6798e-04, 2.9778e-04, 4.4465e-04, ..., 5.0211e-04, + -5.4407e-04, -8.7509e-03], + [ 2.7704e-04, 2.1827e-04, 3.8743e-04, ..., 4.8518e-04, + 6.6566e-04, 7.7105e-04], + [-3.3379e-04, 2.0111e-04, 3.0303e-04, ..., 3.8934e-04, + -9.2387e-05, -2.5139e-03]], device='cuda:0') +Epoch 274, bias, value: tensor([ 0.0068, -0.0013, 0.0004, -0.0174, 0.0074, -0.0082, -0.0120, -0.0212, + 0.0054, 0.0012], device='cuda:0'), grad: tensor([ 0.0130, 0.0221, -0.0028, -0.0462, -0.0096, -0.0032, 0.0102, -0.0017, + 0.0126, 0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 216.67, cls_loss 0.5089 cls_loss_mapping 0.0048 cls_loss_causal 0.4822 re_mapping 0.0061 re_causal 0.0165 /// teacc 98.84 lr 0.00010000 +Epoch 275, weight, value: tensor([[ 0.0341, 0.1036, -0.1648, ..., -0.0996, 0.0663, 0.0048], + [-0.0831, -0.0963, 0.0957, ..., -0.0324, -0.0977, 0.0532], + [-0.0710, -0.0575, -0.1082, ..., -0.1222, -0.0271, 0.0525], + ..., + [-0.0175, -0.1192, -0.1202, ..., -0.1112, -0.0331, 0.0662], + [-0.1101, -0.1033, -0.1029, ..., -0.0622, -0.0489, -0.0860], + [ 0.0164, -0.0268, -0.0649, ..., -0.0826, 0.0215, -0.0205]], + device='cuda:0'), grad: tensor([[ 2.8000e-03, -1.9111e-06, 6.6042e-05, ..., 1.2803e-04, + 2.6894e-03, 6.0987e-04], + [ 4.2200e-04, 2.3735e-04, 9.5367e-04, ..., 7.5102e-04, + 8.5306e-04, 1.8358e-05], + [ 9.5415e-04, 1.2994e-04, 3.2616e-04, ..., 3.8052e-04, + 1.4324e-03, -1.1454e-03], + ..., + [-1.9350e-03, 3.6097e-04, 1.1557e-04, ..., 1.8382e-04, + 1.5917e-03, -3.3760e-03], + [ 4.8518e-04, 1.4248e-03, 7.3719e-04, ..., 8.0729e-04, + -2.0421e-04, 1.0014e-03], + [-2.9640e-03, -1.7586e-03, 5.9664e-05, ..., 1.6785e-04, + -8.8501e-03, 4.9858e-03]], device='cuda:0') +Epoch 275, bias, value: tensor([ 0.0066, -0.0006, 0.0006, -0.0174, 0.0075, -0.0088, -0.0118, -0.0221, + 0.0058, 0.0013], device='cuda:0'), grad: tensor([-0.0296, -0.0077, -0.0318, 0.0285, -0.0026, 0.0261, -0.0092, 0.0173, + 0.0452, -0.0362], device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 216.70, cls_loss 0.4847 cls_loss_mapping 0.0057 cls_loss_causal 0.4477 re_mapping 0.0060 re_causal 0.0156 /// teacc 98.84 lr 0.00010000 +Epoch 276, weight, value: tensor([[ 0.0344, 0.1033, -0.1650, ..., -0.1001, 0.0675, 0.0046], + [-0.0835, -0.0966, 0.0964, ..., -0.0330, -0.0974, 0.0530], + [-0.0713, -0.0570, -0.1087, ..., -0.1216, -0.0274, 0.0530], + ..., + [-0.0160, -0.1191, -0.1210, ..., -0.1112, -0.0329, 0.0670], + [-0.1102, -0.1047, -0.1026, ..., -0.0624, -0.0485, -0.0867], + [ 0.0163, -0.0255, -0.0651, ..., -0.0833, 0.0238, -0.0210]], + device='cuda:0'), grad: tensor([[ 3.2282e-04, -3.0460e-03, 2.3171e-06, ..., -8.1444e-04, + 2.8658e-04, 3.2501e-03], + [ 2.3770e-04, 4.7237e-05, 1.9515e-04, ..., 2.4056e-04, + 2.5511e-04, -3.0308e-03], + [ 3.0780e-04, 2.5959e-03, -3.4928e-05, ..., 7.4506e-05, + -6.1035e-04, -1.7252e-03], + ..., + [ 7.3099e-04, 4.9919e-05, 7.6890e-06, ..., 1.3244e-04, + 5.4789e-04, 5.5923e-03], + [-3.1223e-03, -2.2430e-03, -2.0516e-04, ..., -6.2704e-04, + -2.0943e-03, -1.1520e-02], + [ 5.7125e-04, 2.2960e-04, 1.0477e-06, ..., 1.6999e-04, + 5.2500e-04, -2.5010e-04]], device='cuda:0') +Epoch 276, bias, value: tensor([ 0.0066, 0.0001, 0.0015, -0.0188, 0.0068, -0.0092, -0.0112, -0.0225, + 0.0054, 0.0022], device='cuda:0'), grad: tensor([ 0.0305, 0.0033, -0.0341, 0.0188, 0.0249, 0.0218, 0.0022, 0.0367, + -0.0891, -0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 216.51, cls_loss 0.4727 cls_loss_mapping 0.0049 cls_loss_causal 0.4464 re_mapping 0.0065 re_causal 0.0170 /// teacc 98.89 lr 0.00010000 +Epoch 277, weight, value: tensor([[ 0.0358, 0.1039, -0.1652, ..., -0.0991, 0.0683, 0.0040], + [-0.0846, -0.0967, 0.0970, ..., -0.0346, -0.0976, 0.0534], + [-0.0718, -0.0577, -0.1092, ..., -0.1218, -0.0275, 0.0526], + ..., + [-0.0165, -0.1195, -0.1221, ..., -0.1124, -0.0337, 0.0672], + [-0.1104, -0.1027, -0.1020, ..., -0.0625, -0.0484, -0.0880], + [ 0.0158, -0.0256, -0.0647, ..., -0.0836, 0.0238, -0.0211]], + device='cuda:0'), grad: tensor([[-7.2632e-03, -3.5114e-03, -2.4643e-03, ..., -2.7466e-03, + -6.4964e-03, -5.1079e-03], + [ 5.5981e-04, 2.4140e-04, 3.2783e-04, ..., 2.9230e-04, + 4.6206e-04, 6.6280e-04], + [ 2.3861e-03, 1.8339e-03, 1.4172e-03, ..., 1.3371e-03, + 1.3819e-03, 7.8125e-03], + ..., + [ 1.3065e-03, 7.8559e-05, 6.0707e-05, ..., 1.6093e-04, + 7.6866e-04, -4.3488e-03], + [ 1.9064e-03, 3.5439e-03, 4.9210e-03, ..., 3.7804e-03, + 2.8725e-03, 3.5596e-04], + [ 1.5221e-03, 3.7384e-04, 3.4571e-04, ..., 3.5787e-04, + 1.0824e-03, 2.8591e-03]], device='cuda:0') +Epoch 277, bias, value: tensor([ 0.0063, -0.0002, 0.0011, -0.0195, 0.0075, -0.0077, -0.0118, -0.0213, + 0.0052, 0.0015], device='cuda:0'), grad: tensor([-0.0523, 0.0088, 0.0028, -0.0320, 0.0157, -0.0042, 0.0136, -0.0045, + 0.0212, 0.0309], device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 216.45, cls_loss 0.4761 cls_loss_mapping 0.0057 cls_loss_causal 0.4518 re_mapping 0.0069 re_causal 0.0187 /// teacc 98.78 lr 0.00010000 +Epoch 278, weight, value: tensor([[ 0.0358, 0.1037, -0.1658, ..., -0.0990, 0.0680, 0.0039], + [-0.0857, -0.0966, 0.0969, ..., -0.0326, -0.0981, 0.0538], + [-0.0721, -0.0575, -0.1095, ..., -0.1231, -0.0274, 0.0529], + ..., + [-0.0165, -0.1199, -0.1222, ..., -0.1118, -0.0326, 0.0673], + [-0.1099, -0.1021, -0.1019, ..., -0.0613, -0.0471, -0.0881], + [ 0.0160, -0.0260, -0.0653, ..., -0.0846, 0.0247, -0.0223]], + device='cuda:0'), grad: tensor([[ 0.0015, 0.0037, 0.0021, ..., 0.0016, 0.0018, 0.0017], + [ 0.0007, 0.0003, -0.0022, ..., 0.0002, 0.0002, -0.0025], + [-0.0023, -0.0010, -0.0001, ..., -0.0005, -0.0017, -0.0035], + ..., + [-0.0008, -0.0008, -0.0004, ..., -0.0010, -0.0003, 0.0003], + [ 0.0007, 0.0009, 0.0026, ..., 0.0006, 0.0006, 0.0031], + [ 0.0009, 0.0006, 0.0003, ..., 0.0003, 0.0008, 0.0009]], + device='cuda:0') +Epoch 278, bias, value: tensor([ 0.0063, 0.0005, 0.0007, -0.0189, 0.0075, -0.0079, -0.0123, -0.0220, + 0.0058, 0.0014], device='cuda:0'), grad: tensor([ 0.0214, -0.0026, -0.0202, 0.0045, -0.0034, -0.0113, 0.0072, -0.0214, + 0.0162, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 216.82, cls_loss 0.4755 cls_loss_mapping 0.0048 cls_loss_causal 0.4467 re_mapping 0.0064 re_causal 0.0167 /// teacc 98.86 lr 0.00010000 +Epoch 279, weight, value: tensor([[ 0.0352, 0.1022, -0.1665, ..., -0.1006, 0.0673, 0.0036], + [-0.0868, -0.0961, 0.0963, ..., -0.0328, -0.0985, 0.0539], + [-0.0730, -0.0579, -0.1098, ..., -0.1241, -0.0275, 0.0528], + ..., + [-0.0156, -0.1209, -0.1239, ..., -0.1112, -0.0329, 0.0672], + [-0.1099, -0.1014, -0.1025, ..., -0.0601, -0.0470, -0.0884], + [ 0.0161, -0.0254, -0.0647, ..., -0.0847, 0.0251, -0.0223]], + device='cuda:0'), grad: tensor([[-2.9774e-03, -3.0632e-03, -3.7241e-04, ..., 2.3469e-05, + -1.4734e-03, -4.4594e-03], + [ 1.0431e-04, 4.3440e-04, 1.3268e-04, ..., 4.8923e-04, + 2.6852e-05, 3.6550e-04], + [ 5.3549e-04, 4.4322e-04, 1.6785e-04, ..., 1.7786e-04, + 2.0742e-04, 6.2513e-04], + ..., + [ 1.8311e-03, 1.3943e-03, 2.8014e-04, ..., 2.3043e-04, + 6.1798e-04, 2.0103e-03], + [ 1.2169e-03, 2.7790e-03, 2.0466e-03, ..., 2.2144e-03, + 1.3196e-04, 1.0557e-03], + [-4.2664e-02, 8.7786e-04, 3.4189e-04, ..., 3.7718e-04, + 2.0862e-04, 7.7724e-04]], device='cuda:0') +Epoch 279, bias, value: tensor([ 0.0067, -0.0002, 0.0005, -0.0175, 0.0082, -0.0078, -0.0124, -0.0222, + 0.0054, 0.0002], device='cuda:0'), grad: tensor([-0.0155, 0.0042, 0.0237, -0.0636, 0.0277, 0.0078, 0.0053, 0.0106, + 0.0126, -0.0127], device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 216.50, cls_loss 0.4637 cls_loss_mapping 0.0038 cls_loss_causal 0.4357 re_mapping 0.0066 re_causal 0.0171 /// teacc 98.78 lr 0.00010000 +Epoch 280, weight, value: tensor([[ 0.0353, 0.1027, -0.1661, ..., -0.1010, 0.0671, 0.0050], + [-0.0870, -0.0958, 0.0968, ..., -0.0318, -0.0988, 0.0544], + [-0.0739, -0.0576, -0.1095, ..., -0.1237, -0.0279, 0.0527], + ..., + [-0.0160, -0.1190, -0.1243, ..., -0.1105, -0.0320, 0.0673], + [-0.1095, -0.1019, -0.1027, ..., -0.0599, -0.0470, -0.0891], + [ 0.0179, -0.0251, -0.0637, ..., -0.0848, 0.0254, -0.0215]], + device='cuda:0'), grad: tensor([[ 2.7537e-04, -2.8777e-04, -3.2753e-05, ..., 8.7500e-04, + -1.6537e-03, -9.4748e-04], + [ 5.2023e-04, 1.1051e-04, 2.4913e-08, ..., -1.0977e-03, + 4.4537e-04, -1.6384e-03], + [ 4.0555e-04, 6.4254e-05, 1.8496e-06, ..., 4.0507e-04, + 4.7565e-04, 1.8711e-03], + ..., + [ 7.6103e-04, 1.9342e-05, 2.0862e-07, ..., 2.4772e-04, + 1.9443e-04, -1.8311e-03], + [ 7.3004e-04, 2.8920e-04, 2.2054e-05, ..., 4.1366e-04, + 5.0402e-04, 7.7963e-04], + [ 3.9902e-03, 1.4293e-04, 5.6066e-06, ..., 4.6921e-04, + 1.4381e-03, 2.1591e-03]], device='cuda:0') +Epoch 280, bias, value: tensor([ 0.0077, -0.0003, 0.0020, -0.0177, 0.0073, -0.0076, -0.0134, -0.0226, + 0.0046, 0.0011], device='cuda:0'), grad: tensor([-0.0398, -0.0038, 0.0254, 0.0212, -0.0054, -0.0108, 0.0111, -0.0672, + 0.0249, 0.0443], device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 216.66, cls_loss 0.4831 cls_loss_mapping 0.0045 cls_loss_causal 0.4620 re_mapping 0.0062 re_causal 0.0160 /// teacc 98.86 lr 0.00010000 +Epoch 281, weight, value: tensor([[ 0.0358, 0.1040, -0.1636, ..., -0.1010, 0.0680, 0.0059], + [-0.0872, -0.0971, 0.0975, ..., -0.0308, -0.0991, 0.0543], + [-0.0751, -0.0588, -0.1101, ..., -0.1241, -0.0290, 0.0532], + ..., + [-0.0159, -0.1190, -0.1241, ..., -0.1104, -0.0324, 0.0665], + [-0.1091, -0.1007, -0.1031, ..., -0.0608, -0.0463, -0.0885], + [ 0.0182, -0.0258, -0.0639, ..., -0.0854, 0.0253, -0.0219]], + device='cuda:0'), grad: tensor([[-1.7548e-04, 1.5748e-04, 3.3617e-05, ..., 2.1100e-04, + -4.2963e-04, -5.1022e-04], + [ 1.3423e-04, -5.1689e-04, 1.9953e-05, ..., -2.3603e-04, + 1.5891e-04, -5.7793e-04], + [-5.5838e-04, -3.7885e-04, -4.2081e-05, ..., -8.4877e-04, + -5.9891e-04, -1.3933e-03], + ..., + [ 1.8227e-04, 1.1522e-04, 3.9726e-05, ..., 9.1851e-05, + 2.1958e-04, -1.1959e-03], + [ 2.2554e-04, 3.3593e-04, 6.4194e-05, ..., 2.2459e-04, + 2.4104e-04, 1.2865e-03], + [ 3.1137e-04, 7.5638e-05, 1.7241e-05, ..., 9.6202e-05, + 3.5405e-04, 4.0436e-04]], device='cuda:0') +Epoch 281, bias, value: tensor([ 0.0090, -0.0005, 0.0006, -0.0178, 0.0084, -0.0081, -0.0144, -0.0224, + 0.0050, 0.0012], device='cuda:0'), grad: tensor([-0.0132, 0.0054, -0.0472, 0.0201, -0.0166, -0.0078, 0.0105, 0.0089, + 0.0213, 0.0187], device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 216.41, cls_loss 0.5208 cls_loss_mapping 0.0052 cls_loss_causal 0.5056 re_mapping 0.0067 re_causal 0.0183 /// teacc 98.81 lr 0.00010000 +Epoch 282, weight, value: tensor([[ 0.0345, 0.1040, -0.1640, ..., -0.1012, 0.0679, 0.0057], + [-0.0866, -0.0978, 0.0977, ..., -0.0332, -0.1001, 0.0539], + [-0.0759, -0.0580, -0.1107, ..., -0.1252, -0.0297, 0.0530], + ..., + [-0.0168, -0.1193, -0.1233, ..., -0.1105, -0.0331, 0.0678], + [-0.1074, -0.1015, -0.1027, ..., -0.0596, -0.0452, -0.0896], + [ 0.0173, -0.0264, -0.0667, ..., -0.0875, 0.0248, -0.0216]], + device='cuda:0'), grad: tensor([[ 2.8086e-04, 3.0661e-04, 3.0041e-04, ..., 6.6566e-04, + 1.6129e-04, 3.2997e-04], + [ 1.9681e-04, 1.9240e-04, 1.5843e-04, ..., 4.2129e-04, + 1.3232e-04, 2.3210e-04], + [ 2.8443e-04, 5.3787e-04, 4.1676e-04, ..., 6.6757e-04, + 2.5558e-04, 4.1533e-04], + ..., + [ 3.7622e-04, 4.7731e-04, 2.9302e-04, ..., 2.9612e-04, + 2.8658e-04, 5.6314e-04], + [-3.8576e-04, -1.0309e-03, -9.1171e-04, ..., -1.0042e-03, + -9.0659e-05, -2.8491e-04], + [-1.3437e-03, -6.8331e-04, -4.2367e-04, ..., -2.4757e-03, + -1.0481e-03, -2.4738e-03]], device='cuda:0') +Epoch 282, bias, value: tensor([ 0.0092, -0.0009, 0.0003, -0.0180, 0.0093, -0.0080, -0.0148, -0.0224, + 0.0060, 0.0006], device='cuda:0'), grad: tensor([ 0.0051, 0.0034, 0.0061, 0.0005, 0.0044, 0.0074, 0.0046, 0.0056, + -0.0056, -0.0315], device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 216.29, cls_loss 0.5183 cls_loss_mapping 0.0053 cls_loss_causal 0.4890 re_mapping 0.0066 re_causal 0.0180 /// teacc 98.61 lr 0.00010000 +Epoch 283, weight, value: tensor([[ 0.0344, 0.1032, -0.1636, ..., -0.1007, 0.0683, 0.0076], + [-0.0872, -0.0988, 0.0970, ..., -0.0329, -0.0998, 0.0532], + [-0.0752, -0.0567, -0.1108, ..., -0.1252, -0.0292, 0.0527], + ..., + [-0.0173, -0.1194, -0.1235, ..., -0.1102, -0.0335, 0.0687], + [-0.1072, -0.1014, -0.1022, ..., -0.0608, -0.0458, -0.0910], + [ 0.0176, -0.0251, -0.0644, ..., -0.0886, 0.0250, -0.0204]], + device='cuda:0'), grad: tensor([[ 7.5340e-04, 4.5753e-04, 9.7990e-05, ..., 3.6621e-04, + 9.8896e-04, 2.7704e-04], + [ 1.0681e-03, -2.8670e-05, -7.2575e-04, ..., -2.5692e-03, + -1.8728e-04, -7.8201e-04], + [ 5.9748e-04, 7.2622e-04, 1.3328e-04, ..., 2.4140e-04, + 6.6662e-04, -1.0843e-03], + ..., + [ 9.2745e-04, 8.4448e-04, 6.6280e-05, ..., 2.3174e-04, + 1.0071e-03, 2.6155e-04], + [-3.5577e-06, 5.2547e-04, 3.9554e-04, ..., 1.1663e-03, + -9.5415e-04, -2.8634e-04], + [-1.1454e-03, 2.2659e-03, 2.8825e-04, ..., 8.1778e-04, + 2.3317e-04, 1.9193e-04]], device='cuda:0') +Epoch 283, bias, value: tensor([ 0.0092, -0.0006, 0.0002, -0.0186, 0.0089, -0.0083, -0.0136, -0.0226, + 0.0050, 0.0015], device='cuda:0'), grad: tensor([ 0.0306, -0.0872, 0.0169, 0.0178, -0.0034, 0.0310, -0.0052, 0.0281, + 0.0089, -0.0375], device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 216.61, cls_loss 0.4729 cls_loss_mapping 0.0041 cls_loss_causal 0.4445 re_mapping 0.0068 re_causal 0.0176 /// teacc 98.72 lr 0.00010000 +Epoch 284, weight, value: tensor([[ 0.0342, 0.1036, -0.1634, ..., -0.1006, 0.0687, 0.0073], + [-0.0876, -0.0981, 0.0966, ..., -0.0331, -0.0996, 0.0537], + [-0.0759, -0.0566, -0.1107, ..., -0.1248, -0.0287, 0.0531], + ..., + [-0.0167, -0.1203, -0.1238, ..., -0.1117, -0.0336, 0.0689], + [-0.1076, -0.1011, -0.1013, ..., -0.0605, -0.0458, -0.0922], + [ 0.0175, -0.0248, -0.0646, ..., -0.0896, 0.0255, -0.0200]], + device='cuda:0'), grad: tensor([[ 1.1490e-02, 1.6794e-03, 4.1634e-05, ..., 3.2120e-03, + 6.2752e-03, 2.3518e-03], + [ 4.1151e-04, -1.0347e-03, 9.3126e-04, ..., -5.5456e-04, + -1.8387e-03, -2.0313e-03], + [-2.0790e-03, -4.3178e-04, -1.3695e-03, ..., -1.8158e-03, + -1.6136e-03, -1.6661e-03], + ..., + [ 3.2425e-04, 8.1587e-04, 8.6367e-05, ..., -3.5405e-05, + -3.8314e-04, 3.9649e-04], + [ 4.5586e-04, 9.2649e-04, 9.6083e-05, ..., 1.3332e-03, + 1.0657e-04, 1.0366e-03], + [-8.8263e-04, -1.5650e-03, 1.2136e-04, ..., -3.6287e-04, + -1.3103e-03, -4.7350e-04]], device='cuda:0') +Epoch 284, bias, value: tensor([ 0.0093, -0.0004, 0.0002, -0.0183, 0.0092, -0.0090, -0.0140, -0.0224, + 0.0058, 0.0008], device='cuda:0'), grad: tensor([ 0.0504, 0.0098, -0.0343, 0.0225, 0.0219, -0.0099, -0.0139, 0.0010, + -0.0056, -0.0420], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 283---------------------------------------------------- +epoch 283, time 216.71, cls_loss 0.4548 cls_loss_mapping 0.0029 cls_loss_causal 0.4374 re_mapping 0.0067 re_causal 0.0172 /// teacc 98.95 lr 0.00010000 +Epoch 285, weight, value: tensor([[ 0.0337, 0.1040, -0.1636, ..., -0.1004, 0.0688, 0.0071], + [-0.0873, -0.0978, 0.0970, ..., -0.0324, -0.0981, 0.0544], + [-0.0764, -0.0567, -0.1106, ..., -0.1245, -0.0295, 0.0524], + ..., + [-0.0174, -0.1201, -0.1238, ..., -0.1126, -0.0345, 0.0688], + [-0.1088, -0.1005, -0.1009, ..., -0.0606, -0.0465, -0.0927], + [ 0.0183, -0.0249, -0.0650, ..., -0.0897, 0.0260, -0.0204]], + device='cuda:0'), grad: tensor([[ 1.9467e-04, -1.2226e-03, 6.4215e-07, ..., -2.2483e-04, + 1.4524e-03, 1.7858e-04], + [ 1.4353e-04, 2.6774e-04, 4.2841e-08, ..., 1.6232e-03, + -6.3801e-04, -3.6144e-03], + [ 5.9271e-04, 4.0131e-03, 1.1511e-06, ..., 8.7619e-05, + 1.6689e-03, 1.2617e-03], + ..., + [ 6.5851e-04, 3.2473e-04, 1.7844e-06, ..., 8.0884e-05, + 1.4811e-03, 2.0695e-03], + [ 2.0349e-04, -5.2834e-03, -1.5058e-05, ..., 8.2314e-05, + 6.7043e-04, 3.4332e-04], + [ 7.4959e-03, 4.1509e-04, 9.7528e-06, ..., 6.8426e-05, + 5.4283e-03, 4.2439e-04]], device='cuda:0') +Epoch 285, bias, value: tensor([ 0.0090, 0.0007, -0.0007, -0.0189, 0.0098, -0.0089, -0.0133, -0.0223, + 0.0051, 0.0007], device='cuda:0'), grad: tensor([ 0.0052, 0.0103, 0.0218, -0.0051, -0.0653, 0.0206, -0.0285, 0.0243, + -0.0093, 0.0259], device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 216.56, cls_loss 0.4817 cls_loss_mapping 0.0030 cls_loss_causal 0.4632 re_mapping 0.0061 re_causal 0.0162 /// teacc 98.70 lr 0.00010000 +Epoch 286, weight, value: tensor([[ 0.0341, 0.1032, -0.1642, ..., -0.0995, 0.0690, 0.0076], + [-0.0874, -0.0978, 0.0979, ..., -0.0348, -0.0979, 0.0533], + [-0.0764, -0.0570, -0.1103, ..., -0.1254, -0.0285, 0.0519], + ..., + [-0.0184, -0.1202, -0.1245, ..., -0.1125, -0.0354, 0.0679], + [-0.1090, -0.1016, -0.1020, ..., -0.0598, -0.0474, -0.0930], + [ 0.0180, -0.0248, -0.0657, ..., -0.0888, 0.0257, -0.0189]], + device='cuda:0'), grad: tensor([[ 0.0014, 0.0003, 0.0002, ..., 0.0014, 0.0017, 0.0005], + [-0.0013, -0.0008, -0.0010, ..., -0.0036, -0.0020, 0.0008], + [ 0.0020, 0.0007, 0.0003, ..., 0.0013, 0.0019, 0.0023], + ..., + [ 0.0016, 0.0020, 0.0006, ..., 0.0024, 0.0011, 0.0037], + [ 0.0010, -0.0049, -0.0035, ..., -0.0060, -0.0026, -0.0055], + [-0.0051, 0.0009, 0.0003, ..., 0.0017, 0.0018, 0.0010]], + device='cuda:0') +Epoch 286, bias, value: tensor([ 0.0088, -0.0002, -0.0001, -0.0184, 0.0088, -0.0075, -0.0132, -0.0223, + 0.0039, 0.0014], device='cuda:0'), grad: tensor([ 0.0224, -0.0035, 0.0255, 0.0398, -0.0343, -0.0109, 0.0382, -0.0224, + -0.0319, -0.0230], device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 216.67, cls_loss 0.5072 cls_loss_mapping 0.0057 cls_loss_causal 0.4854 re_mapping 0.0061 re_causal 0.0154 /// teacc 98.80 lr 0.00010000 +Epoch 287, weight, value: tensor([[ 0.0336, 0.1045, -0.1641, ..., -0.1002, 0.0690, 0.0066], + [-0.0874, -0.0970, 0.0977, ..., -0.0340, -0.0969, 0.0544], + [-0.0774, -0.0573, -0.1087, ..., -0.1270, -0.0287, 0.0520], + ..., + [-0.0192, -0.1206, -0.1260, ..., -0.1127, -0.0348, 0.0681], + [-0.1081, -0.1021, -0.1018, ..., -0.0611, -0.0473, -0.0929], + [ 0.0176, -0.0260, -0.0659, ..., -0.0898, 0.0247, -0.0198]], + device='cuda:0'), grad: tensor([[ 1.0115e-04, 1.9300e-04, 1.1975e-04, ..., 1.1358e-03, + 3.8528e-04, 3.7313e-04], + [ 6.8605e-05, 2.2721e-04, 3.2473e-04, ..., 5.1546e-04, + 5.2786e-04, 6.0892e-04], + [ 7.5042e-05, 1.3041e-04, 1.1939e-04, ..., 8.1825e-04, + 2.9039e-04, 3.7479e-04], + ..., + [ 1.7881e-05, 4.7088e-05, 1.0949e-04, ..., 2.4581e-04, + 2.0146e-04, -1.9913e-03], + [ 1.1402e-04, 2.1648e-04, 1.3220e-04, ..., 9.2697e-04, + 3.9673e-04, 3.7694e-04], + [ 2.8044e-05, 7.8678e-05, 6.7472e-05, ..., 4.6325e-04, + -6.4802e-04, 3.6645e-04]], device='cuda:0') +Epoch 287, bias, value: tensor([ 9.0360e-03, -1.4786e-05, 1.9306e-04, -1.8110e-02, 7.9707e-03, + -8.3505e-03, -1.2123e-02, -2.2014e-02, 4.5183e-03, 1.0887e-04], + device='cuda:0'), grad: tensor([ 0.0167, 0.0211, 0.0172, -0.0397, 0.0144, 0.0127, -0.0111, -0.0021, + 0.0184, -0.0476], device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 216.48, cls_loss 0.4953 cls_loss_mapping 0.0047 cls_loss_causal 0.4645 re_mapping 0.0069 re_causal 0.0172 /// teacc 98.87 lr 0.00010000 +Epoch 288, weight, value: tensor([[ 0.0335, 0.1045, -0.1636, ..., -0.0989, 0.0691, 0.0061], + [-0.0870, -0.0950, 0.0979, ..., -0.0353, -0.0946, 0.0544], + [-0.0766, -0.0556, -0.1096, ..., -0.1265, -0.0288, 0.0520], + ..., + [-0.0193, -0.1221, -0.1250, ..., -0.1129, -0.0347, 0.0678], + [-0.1090, -0.1023, -0.1012, ..., -0.0606, -0.0477, -0.0922], + [ 0.0180, -0.0264, -0.0653, ..., -0.0903, 0.0247, -0.0190]], + device='cuda:0'), grad: tensor([[ 6.3705e-03, 1.1724e-04, 2.0921e-05, ..., 6.8009e-05, + 7.4921e-03, -7.9203e-04], + [-7.2289e-04, -1.1110e-03, -3.4034e-05, ..., 2.6047e-05, + -2.6264e-03, 2.4092e-04], + [ 1.1313e-04, 9.6679e-05, 3.6508e-05, ..., -2.1264e-05, + 1.4858e-03, -6.5422e-04], + ..., + [ 1.5402e-04, 1.3399e-04, 6.6161e-05, ..., 6.9551e-06, + -3.5167e-04, -6.4898e-04], + [ 2.6274e-04, 1.0496e-04, 2.8992e-04, ..., 1.3077e-04, + -2.2469e-03, 1.0401e-04], + [ 7.6437e-04, 1.2922e-04, -1.3828e-03, ..., 1.9193e-05, + 2.6665e-03, 7.6580e-04]], device='cuda:0') +Epoch 288, bias, value: tensor([ 0.0085, 0.0006, -0.0004, -0.0182, 0.0090, -0.0086, -0.0129, -0.0224, + 0.0042, 0.0013], device='cuda:0'), grad: tensor([-0.0270, -0.0400, -0.0264, 0.0306, 0.0049, 0.0140, 0.0427, -0.0148, + -0.0049, 0.0208], device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 216.46, cls_loss 0.5000 cls_loss_mapping 0.0057 cls_loss_causal 0.4739 re_mapping 0.0060 re_causal 0.0152 /// teacc 98.83 lr 0.00010000 +Epoch 289, weight, value: tensor([[ 0.0331, 0.1051, -0.1639, ..., -0.1000, 0.0684, 0.0050], + [-0.0860, -0.0936, 0.0995, ..., -0.0355, -0.0925, 0.0539], + [-0.0763, -0.0566, -0.1096, ..., -0.1253, -0.0292, 0.0527], + ..., + [-0.0190, -0.1213, -0.1246, ..., -0.1120, -0.0341, 0.0682], + [-0.1095, -0.1026, -0.1019, ..., -0.0604, -0.0491, -0.0923], + [ 0.0178, -0.0260, -0.0651, ..., -0.0910, 0.0233, -0.0191]], + device='cuda:0'), grad: tensor([[ 8.2254e-05, 5.0545e-04, 3.3998e-04, ..., 7.7486e-06, + 1.4317e-04, 6.5136e-04], + [ 1.5545e-04, 6.8855e-04, 5.0974e-04, ..., 7.7039e-06, + 3.2043e-04, 6.7568e-04], + [ 2.3043e-04, 1.4219e-03, 1.0080e-03, ..., 2.4036e-05, + 4.5395e-04, 1.0462e-03], + ..., + [ 1.9038e-04, 5.5981e-04, -7.6962e-04, ..., -1.1601e-05, + 5.0783e-04, 2.0733e-03], + [-8.7833e-04, -5.1403e-04, 4.3082e-04, ..., -3.0234e-05, + 4.1747e-04, -3.8300e-03], + [ 7.2432e-04, 1.3804e-04, 1.2836e-03, ..., 2.2098e-05, + -5.6648e-04, 7.1287e-04]], device='cuda:0') +Epoch 289, bias, value: tensor([ 0.0072, 0.0010, 0.0007, -0.0181, 0.0090, -0.0092, -0.0131, -0.0223, + 0.0048, 0.0013], device='cuda:0'), grad: tensor([-0.0160, 0.0205, -0.0043, 0.0244, 0.0218, -0.0310, -0.0113, 0.0327, + -0.0503, 0.0134], device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 216.30, cls_loss 0.4711 cls_loss_mapping 0.0026 cls_loss_causal 0.4507 re_mapping 0.0064 re_causal 0.0170 /// teacc 98.84 lr 0.00010000 +Epoch 290, weight, value: tensor([[ 0.0339, 0.1066, -0.1630, ..., -0.0997, 0.0695, 0.0049], + [-0.0855, -0.0942, 0.0986, ..., -0.0368, -0.0923, 0.0531], + [-0.0769, -0.0567, -0.1086, ..., -0.1249, -0.0296, 0.0530], + ..., + [-0.0177, -0.1205, -0.1249, ..., -0.1106, -0.0346, 0.0685], + [-0.1110, -0.1036, -0.1017, ..., -0.0597, -0.0499, -0.0923], + [ 0.0172, -0.0255, -0.0647, ..., -0.0919, 0.0234, -0.0210]], + device='cuda:0'), grad: tensor([[-1.7357e-03, -8.3237e-03, -4.1924e-03, ..., 1.2815e-04, + -2.9964e-03, -1.0796e-03], + [ 1.3936e-04, 7.9632e-04, 3.2640e-04, ..., -6.2108e-05, + 6.0701e-04, 7.8559e-05], + [ 9.7942e-04, 4.3678e-03, 2.0752e-03, ..., 1.9550e-04, + 1.4572e-03, 8.3017e-04], + ..., + [ 5.5254e-05, 3.7479e-04, 7.9513e-05, ..., 5.6177e-05, + -1.2094e-04, 1.7738e-04], + [-2.8276e-04, 1.0645e-04, -1.1277e-04, ..., -3.8600e-04, + 5.8842e-04, -1.7710e-03], + [ 5.0575e-05, -7.5111e-03, 2.2233e-04, ..., 6.3956e-05, + 4.9782e-04, 1.9836e-04]], device='cuda:0') +Epoch 290, bias, value: tensor([ 0.0080, 0.0006, 0.0004, -0.0179, 0.0093, -0.0083, -0.0135, -0.0223, + 0.0043, 0.0004], device='cuda:0'), grad: tensor([-0.0177, 0.0216, -0.0275, 0.0347, -0.0085, 0.0443, -0.0126, -0.0123, + -0.0105, -0.0117], device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 216.65, cls_loss 0.4797 cls_loss_mapping 0.0035 cls_loss_causal 0.4543 re_mapping 0.0063 re_causal 0.0173 /// teacc 98.79 lr 0.00010000 +Epoch 291, weight, value: tensor([[ 0.0348, 0.1075, -0.1618, ..., -0.0989, 0.0705, 0.0055], + [-0.0853, -0.0944, 0.0995, ..., -0.0364, -0.0931, 0.0539], + [-0.0786, -0.0559, -0.1085, ..., -0.1254, -0.0301, 0.0512], + ..., + [-0.0178, -0.1211, -0.1259, ..., -0.1113, -0.0354, 0.0698], + [-0.1121, -0.1031, -0.1013, ..., -0.0608, -0.0506, -0.0934], + [ 0.0170, -0.0256, -0.0651, ..., -0.0921, 0.0223, -0.0212]], + device='cuda:0'), grad: tensor([[ 7.4434e-04, 7.7343e-04, 3.3170e-05, ..., 5.5414e-07, + 2.1152e-03, 1.6193e-03], + [ 4.5681e-04, 4.5800e-04, -1.0462e-03, ..., -4.5151e-05, + 1.8358e-03, -1.9112e-03], + [ 1.7118e-03, -1.1314e-02, -1.7519e-03, ..., 4.6380e-07, + 3.9673e-03, 2.1000e-03], + ..., + [ 1.2386e-04, 6.1417e-04, 2.9230e-04, ..., 7.4180e-07, + -2.5177e-03, -3.4637e-03], + [-1.7118e-04, -2.0351e-03, 7.4100e-04, ..., 1.4208e-05, + -1.2665e-03, 1.0777e-03], + [ 1.0700e-03, 9.2697e-04, 3.5912e-05, ..., 6.1840e-07, + 1.1911e-03, 1.4286e-03]], device='cuda:0') +Epoch 291, bias, value: tensor([ 0.0083, 0.0004, -0.0001, -0.0175, 0.0103, -0.0089, -0.0133, -0.0221, + 0.0035, 0.0005], device='cuda:0'), grad: tensor([ 3.0457e-02, 1.0857e-02, 1.2566e-02, -4.2938e-02, 3.6499e-02, + 2.0493e-02, 2.4796e-02, -5.6091e-02, -3.6713e-02, 6.0439e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 216.53, cls_loss 0.4472 cls_loss_mapping 0.0034 cls_loss_causal 0.4183 re_mapping 0.0063 re_causal 0.0170 /// teacc 98.62 lr 0.00010000 +Epoch 292, weight, value: tensor([[ 0.0357, 0.1075, -0.1611, ..., -0.0987, 0.0711, 0.0051], + [-0.0861, -0.0955, 0.1003, ..., -0.0372, -0.0937, 0.0538], + [-0.0777, -0.0547, -0.1079, ..., -0.1251, -0.0303, 0.0521], + ..., + [-0.0177, -0.1225, -0.1264, ..., -0.1122, -0.0365, 0.0692], + [-0.1130, -0.1027, -0.1027, ..., -0.0608, -0.0514, -0.0934], + [ 0.0163, -0.0250, -0.0651, ..., -0.0914, 0.0223, -0.0208]], + device='cuda:0'), grad: tensor([[-1.2910e-04, -6.2752e-04, 3.8457e-04, ..., 1.7986e-05, + -6.9523e-04, 2.2089e-04], + [ 4.3392e-04, -7.4387e-04, -2.8801e-03, ..., -5.7220e-04, + -7.9751e-05, -1.8349e-03], + [ 4.7255e-04, 1.8206e-03, 1.5707e-03, ..., 7.1430e-04, + 1.5669e-03, 2.1057e-03], + ..., + [-1.0460e-02, 6.2037e-04, 3.2961e-05, ..., 1.1313e-04, + -1.2581e-02, 2.0564e-04], + [-1.8816e-03, -3.4380e-04, 4.1175e-04, ..., -3.0828e-04, + -2.3422e-03, -1.5268e-03], + [ 2.3193e-03, -2.5177e-03, 3.3617e-04, ..., 1.5724e-04, + 1.1292e-03, 4.3058e-04]], device='cuda:0') +Epoch 292, bias, value: tensor([ 0.0079, 0.0006, -0.0007, -0.0180, 0.0107, -0.0082, -0.0130, -0.0222, + 0.0031, 0.0009], device='cuda:0'), grad: tensor([ 0.0128, 0.0054, 0.0204, 0.0107, 0.0378, 0.0439, -0.0473, -0.0197, + -0.0499, -0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 216.46, cls_loss 0.5091 cls_loss_mapping 0.0041 cls_loss_causal 0.4799 re_mapping 0.0062 re_causal 0.0170 /// teacc 98.85 lr 0.00010000 +Epoch 293, weight, value: tensor([[ 0.0350, 0.1082, -0.1602, ..., -0.0995, 0.0706, 0.0040], + [-0.0872, -0.0955, 0.1008, ..., -0.0367, -0.0948, 0.0535], + [-0.0776, -0.0565, -0.1093, ..., -0.1256, -0.0306, 0.0519], + ..., + [-0.0167, -0.1222, -0.1257, ..., -0.1116, -0.0350, 0.0692], + [-0.1144, -0.1028, -0.1029, ..., -0.0605, -0.0513, -0.0933], + [ 0.0154, -0.0256, -0.0645, ..., -0.0927, 0.0216, -0.0213]], + device='cuda:0'), grad: tensor([[-6.1569e-03, -4.8065e-03, -4.1695e-03, ..., -5.8174e-03, + -3.9253e-03, 5.1141e-05], + [-1.5659e-03, -7.5936e-05, 2.8300e-04, ..., 2.8062e-04, + -5.0449e-04, -1.3275e-03], + [-1.9188e-03, 2.9755e-04, -1.9932e-03, ..., -3.5210e-03, + -4.7073e-03, 7.2777e-05], + ..., + [-1.4448e-03, -2.2945e-03, 9.6262e-05, ..., -5.1641e-04, + 3.1281e-04, 6.7139e-04], + [ 1.3151e-03, 7.8583e-04, 9.6607e-04, ..., 1.4782e-03, + 1.4238e-03, 6.8247e-05], + [ 8.0872e-04, 5.0974e-04, 2.5153e-04, ..., 5.0068e-04, + 1.2600e-04, 4.7773e-05]], device='cuda:0') +Epoch 293, bias, value: tensor([ 0.0066, 0.0006, -0.0006, -0.0182, 0.0101, -0.0076, -0.0121, -0.0213, + 0.0035, 0.0002], device='cuda:0'), grad: tensor([-0.0404, -0.0135, -0.0454, 0.0340, -0.0097, 0.0676, -0.0013, -0.0059, + 0.0264, -0.0118], device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 216.46, cls_loss 0.4704 cls_loss_mapping 0.0043 cls_loss_causal 0.4476 re_mapping 0.0060 re_causal 0.0157 /// teacc 98.71 lr 0.00010000 +Epoch 294, weight, value: tensor([[ 0.0358, 0.1079, -0.1601, ..., -0.0995, 0.0711, 0.0037], + [-0.0868, -0.0964, 0.1013, ..., -0.0355, -0.0945, 0.0519], + [-0.0792, -0.0571, -0.1093, ..., -0.1267, -0.0312, 0.0517], + ..., + [-0.0178, -0.1223, -0.1260, ..., -0.1118, -0.0360, 0.0706], + [-0.1144, -0.1034, -0.1026, ..., -0.0618, -0.0503, -0.0937], + [ 0.0149, -0.0259, -0.0648, ..., -0.0933, 0.0213, -0.0217]], + device='cuda:0'), grad: tensor([[-4.0627e-03, -3.3951e-03, -4.5323e-04, ..., -3.4237e-03, + -1.9798e-03, 9.8884e-05], + [-9.9277e-04, 5.9605e-04, 1.9913e-03, ..., -3.7122e-04, + 1.7548e-04, 1.3971e-04], + [ 4.2605e-04, -1.1177e-03, -5.3024e-03, ..., 4.2105e-04, + 2.0742e-04, 1.0574e-04], + ..., + [ 4.9543e-04, 1.4365e-04, 2.5153e-04, ..., 3.2306e-04, + 2.1017e-04, 1.2302e-04], + [ 8.8978e-04, 5.4073e-04, 4.0817e-04, ..., 8.2302e-04, + 4.5061e-04, 1.4019e-04], + [-1.5587e-02, -2.3232e-03, -8.5831e-04, ..., 4.5156e-04, + -8.6060e-03, -1.1387e-03]], device='cuda:0') +Epoch 294, bias, value: tensor([ 0.0072, 0.0014, -0.0012, -0.0183, 0.0103, -0.0065, -0.0140, -0.0217, + 0.0037, 0.0002], device='cuda:0'), grad: tensor([-0.0143, -0.0331, -0.0037, 0.0258, 0.0599, -0.0012, 0.0279, 0.0145, + 0.0014, -0.0773], device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 216.47, cls_loss 0.4734 cls_loss_mapping 0.0039 cls_loss_causal 0.4487 re_mapping 0.0060 re_causal 0.0161 /// teacc 98.86 lr 0.00010000 +Epoch 295, weight, value: tensor([[ 0.0356, 0.1079, -0.1620, ..., -0.0991, 0.0711, 0.0040], + [-0.0860, -0.0961, 0.1010, ..., -0.0358, -0.0942, 0.0517], + [-0.0812, -0.0574, -0.1096, ..., -0.1274, -0.0318, 0.0524], + ..., + [-0.0183, -0.1232, -0.1261, ..., -0.1096, -0.0357, 0.0702], + [-0.1160, -0.1038, -0.1018, ..., -0.0598, -0.0507, -0.0946], + [ 0.0155, -0.0246, -0.0646, ..., -0.0915, 0.0221, -0.0215]], + device='cuda:0'), grad: tensor([[ 9.5129e-04, 3.3188e-04, 2.1935e-04, ..., 8.8930e-04, + 1.2934e-04, 5.5408e-04], + [-6.5804e-04, -4.3321e-04, -1.3304e-03, ..., -1.4811e-03, + -7.7343e-04, -1.6441e-03], + [-8.1778e-04, 8.8096e-05, 1.1092e-04, ..., 2.0552e-04, + 1.2600e-04, 7.1049e-04], + ..., + [-2.5215e-03, 9.9421e-05, 1.9014e-04, ..., 2.6345e-04, + 2.3556e-04, -1.1425e-03], + [ 6.5994e-04, 2.8610e-04, 3.1829e-04, ..., 6.9284e-04, + 2.5940e-04, 8.4829e-04], + [ 1.5535e-03, 1.4806e-04, 3.4142e-04, ..., 4.8447e-04, + 6.1035e-04, 9.5463e-04]], device='cuda:0') +Epoch 295, bias, value: tensor([ 0.0074, 0.0022, -0.0011, -0.0194, 0.0099, -0.0078, -0.0133, -0.0209, + 0.0029, 0.0012], device='cuda:0'), grad: tensor([ 0.0170, -0.0388, -0.0052, 0.0254, -0.0207, 0.0111, -0.0051, -0.0167, + 0.0149, 0.0181], device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 216.59, cls_loss 0.5082 cls_loss_mapping 0.0055 cls_loss_causal 0.4839 re_mapping 0.0062 re_causal 0.0165 /// teacc 98.70 lr 0.00010000 +Epoch 296, weight, value: tensor([[ 0.0349, 0.1074, -0.1626, ..., -0.0988, 0.0708, 0.0042], + [-0.0873, -0.0969, 0.1014, ..., -0.0357, -0.0955, 0.0511], + [-0.0810, -0.0576, -0.1104, ..., -0.1288, -0.0315, 0.0524], + ..., + [-0.0193, -0.1238, -0.1269, ..., -0.1106, -0.0357, 0.0705], + [-0.1151, -0.1040, -0.1013, ..., -0.0601, -0.0505, -0.0946], + [ 0.0150, -0.0251, -0.0649, ..., -0.0920, 0.0223, -0.0210]], + device='cuda:0'), grad: tensor([[-0.0032, 0.0001, 0.0001, ..., 0.0003, -0.0030, -0.0012], + [-0.0018, -0.0024, -0.0030, ..., -0.0040, -0.0012, -0.0015], + [-0.0003, 0.0002, 0.0002, ..., -0.0004, 0.0009, -0.0025], + ..., + [ 0.0009, 0.0003, 0.0003, ..., 0.0005, 0.0005, 0.0015], + [ 0.0005, 0.0003, 0.0003, ..., 0.0006, 0.0004, 0.0004], + [ 0.0001, 0.0052, 0.0008, ..., 0.0013, 0.0006, 0.0006]], + device='cuda:0') +Epoch 296, bias, value: tensor([ 0.0058, 0.0017, -0.0016, -0.0196, 0.0113, -0.0084, -0.0131, -0.0208, + 0.0037, 0.0018], device='cuda:0'), grad: tensor([-0.0503, -0.0457, 0.0047, -0.0237, -0.0096, 0.0239, 0.0201, 0.0227, + 0.0192, 0.0389], device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 216.39, cls_loss 0.4883 cls_loss_mapping 0.0039 cls_loss_causal 0.4688 re_mapping 0.0061 re_causal 0.0164 /// teacc 98.76 lr 0.00010000 +Epoch 297, weight, value: tensor([[ 0.0364, 0.1076, -0.1636, ..., -0.0996, 0.0716, 0.0042], + [-0.0872, -0.0967, 0.1009, ..., -0.0363, -0.0953, 0.0507], + [-0.0808, -0.0577, -0.1106, ..., -0.1291, -0.0312, 0.0523], + ..., + [-0.0201, -0.1239, -0.1276, ..., -0.1115, -0.0376, 0.0703], + [-0.1154, -0.1037, -0.1013, ..., -0.0603, -0.0495, -0.0948], + [ 0.0155, -0.0251, -0.0645, ..., -0.0919, 0.0230, -0.0199]], + device='cuda:0'), grad: tensor([[-5.9795e-04, 1.6794e-03, 3.7044e-05, ..., -1.1711e-03, + -1.9302e-03, -1.0633e-03], + [ 1.8704e-04, 1.8373e-05, -9.6500e-05, ..., 5.5933e-04, + 7.1096e-04, 1.1295e-04], + [ 1.1700e-04, 1.0806e-04, 2.3663e-05, ..., 3.2592e-04, + -7.4339e-04, 1.0896e-04], + ..., + [ 9.2745e-05, 1.1438e-04, 3.6836e-05, ..., 1.2350e-04, + 2.2054e-04, 1.5402e-04], + [ 1.3256e-04, 1.1396e-03, 6.8724e-05, ..., 4.2629e-04, + 4.3941e-04, 1.8179e-04], + [ 1.4639e-04, -5.0049e-03, -3.2401e-04, ..., 1.8752e-04, + 4.4537e-04, -1.2302e-04]], device='cuda:0') +Epoch 297, bias, value: tensor([ 0.0075, 0.0022, -0.0016, -0.0203, 0.0114, -0.0095, -0.0138, -0.0203, + 0.0032, 0.0019], device='cuda:0'), grad: tensor([ 0.0026, 0.0274, -0.0144, 0.0182, 0.0211, -0.0032, -0.0578, -0.0094, + 0.0235, -0.0080], device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 216.35, cls_loss 0.4710 cls_loss_mapping 0.0042 cls_loss_causal 0.4460 re_mapping 0.0055 re_causal 0.0143 /// teacc 98.68 lr 0.00010000 +Epoch 298, weight, value: tensor([[ 0.0362, 0.1073, -0.1647, ..., -0.1003, 0.0719, 0.0043], + [-0.0876, -0.0970, 0.1017, ..., -0.0357, -0.0958, 0.0515], + [-0.0811, -0.0585, -0.1112, ..., -0.1288, -0.0316, 0.0544], + ..., + [-0.0193, -0.1240, -0.1271, ..., -0.1108, -0.0380, 0.0683], + [-0.1150, -0.1040, -0.1016, ..., -0.0595, -0.0494, -0.0955], + [ 0.0151, -0.0245, -0.0629, ..., -0.0926, 0.0225, -0.0194]], + device='cuda:0'), grad: tensor([[-1.4031e-04, -5.6028e-04, -2.4825e-05, ..., -1.0163e-04, + -2.5249e-04, -3.0785e-03], + [ 1.6379e-04, -1.8120e-03, 7.6666e-06, ..., 4.4179e-04, + 8.4996e-05, -6.8617e-04], + [ 5.3406e-04, 4.8709e-04, 5.9962e-05, ..., 8.7643e-04, + 1.6260e-04, 1.2341e-03], + ..., + [-9.1248e-03, -3.0899e-04, 1.1221e-05, ..., -1.9131e-03, + -1.4687e-03, -5.6763e-03], + [ 1.0090e-03, 7.8630e-04, 3.0488e-05, ..., -5.1117e-04, + 7.3016e-05, -9.4557e-04], + [ 3.8967e-03, 3.8314e-04, 1.0259e-05, ..., 3.0065e-04, + 6.6757e-04, 2.2621e-03]], device='cuda:0') +Epoch 298, bias, value: tensor([ 0.0076, 0.0023, -0.0022, -0.0198, 0.0105, -0.0094, -0.0129, -0.0204, + 0.0031, 0.0020], device='cuda:0'), grad: tensor([-0.0074, -0.0215, 0.0185, -0.0055, -0.0108, 0.0258, 0.0497, -0.0350, + -0.0362, 0.0223], device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 216.56, cls_loss 0.5155 cls_loss_mapping 0.0043 cls_loss_causal 0.4946 re_mapping 0.0063 re_causal 0.0174 /// teacc 98.81 lr 0.00010000 +Epoch 299, weight, value: tensor([[ 0.0356, 0.1075, -0.1640, ..., -0.0987, 0.0716, 0.0056], + [-0.0872, -0.0975, 0.1014, ..., -0.0363, -0.0964, 0.0508], + [-0.0795, -0.0580, -0.1105, ..., -0.1286, -0.0312, 0.0556], + ..., + [-0.0183, -0.1240, -0.1300, ..., -0.1112, -0.0384, 0.0683], + [-0.1158, -0.1044, -0.1029, ..., -0.0588, -0.0497, -0.0958], + [ 0.0156, -0.0242, -0.0598, ..., -0.0937, 0.0228, -0.0182]], + device='cuda:0'), grad: tensor([[-1.7077e-05, -1.6403e-03, 2.7585e-04, ..., -1.3380e-03, + -1.4572e-03, -3.9597e-03], + [ 2.1327e-04, 3.4094e-04, 2.2814e-05, ..., 2.2435e-04, + 3.8671e-04, 1.2779e-03], + [ 4.7779e-04, 6.6900e-04, 2.0170e-04, ..., 1.4913e-04, + 4.1485e-04, 1.2465e-03], + ..., + [ 2.3687e-04, 5.3120e-04, 4.1902e-05, ..., 1.0866e-04, + 1.7858e-04, -6.4898e-04], + [ 6.3753e-04, 2.1191e-03, 4.9114e-04, ..., 1.9658e-04, + 4.4560e-04, 2.5578e-03], + [ 8.9312e-04, 1.3075e-03, 4.3654e-04, ..., 1.3804e-04, + 6.0940e-04, 2.3880e-03]], device='cuda:0') +Epoch 299, bias, value: tensor([ 0.0084, 0.0021, -0.0014, -0.0204, 0.0114, -0.0105, -0.0145, -0.0202, + 0.0031, 0.0027], device='cuda:0'), grad: tensor([-0.0291, -0.0023, 0.0243, -0.0095, 0.0032, -0.0114, -0.0066, -0.0446, + 0.0377, 0.0383], device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 216.85, cls_loss 0.4694 cls_loss_mapping 0.0026 cls_loss_causal 0.4387 re_mapping 0.0065 re_causal 0.0176 /// teacc 98.59 lr 0.00010000 +Epoch 300, weight, value: tensor([[ 0.0359, 0.1073, -0.1650, ..., -0.0982, 0.0724, 0.0054], + [-0.0887, -0.0987, 0.1005, ..., -0.0380, -0.0974, 0.0502], + [-0.0785, -0.0577, -0.1102, ..., -0.1292, -0.0314, 0.0550], + ..., + [-0.0184, -0.1234, -0.1281, ..., -0.1098, -0.0378, 0.0700], + [-0.1168, -0.1042, -0.1026, ..., -0.0585, -0.0501, -0.0947], + [ 0.0157, -0.0249, -0.0596, ..., -0.0943, 0.0236, -0.0197]], + device='cuda:0'), grad: tensor([[ 4.8727e-05, 3.2163e-04, 5.2929e-04, ..., 2.0885e-04, + 1.1790e-04, 2.5773e-04], + [ 9.6202e-05, 6.5756e-04, 7.7105e-04, ..., 3.4666e-04, + 2.7466e-04, 3.1710e-04], + [ 8.4639e-05, 1.7667e-04, 2.6608e-04, ..., 1.1200e-04, + 7.0632e-05, 5.1409e-05], + ..., + [ 1.5104e-04, 8.7261e-05, 1.3351e-04, ..., 4.7952e-05, + 4.9442e-05, 4.7994e-04], + [ 1.2255e-04, 5.2214e-04, 1.0090e-03, ..., 4.1223e-04, + 1.6785e-04, 3.9315e-04], + [-6.5899e-04, 1.2863e-04, 1.8382e-04, ..., 7.7367e-05, + 5.6356e-05, -4.0561e-05]], device='cuda:0') +Epoch 300, bias, value: tensor([ 0.0084, 0.0023, -0.0006, -0.0202, 0.0104, -0.0103, -0.0145, -0.0192, + 0.0027, 0.0019], device='cuda:0'), grad: tensor([ 0.0114, 0.0188, 0.0096, 0.0131, -0.0187, -0.0157, 0.0087, 0.0183, + 0.0211, -0.0667], device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 216.24, cls_loss 0.4952 cls_loss_mapping 0.0037 cls_loss_causal 0.4695 re_mapping 0.0060 re_causal 0.0157 /// teacc 98.84 lr 0.00010000 +Epoch 301, weight, value: tensor([[ 0.0359, 0.1073, -0.1652, ..., -0.0989, 0.0722, 0.0049], + [-0.0887, -0.0998, 0.1015, ..., -0.0381, -0.0977, 0.0492], + [-0.0801, -0.0588, -0.1103, ..., -0.1293, -0.0324, 0.0543], + ..., + [-0.0180, -0.1237, -0.1268, ..., -0.1101, -0.0389, 0.0713], + [-0.1160, -0.1029, -0.1037, ..., -0.0583, -0.0497, -0.0943], + [ 0.0164, -0.0252, -0.0601, ..., -0.0956, 0.0251, -0.0196]], + device='cuda:0'), grad: tensor([[ 2.6774e-04, -1.2846e-03, 8.5974e-04, ..., 2.1636e-05, + -2.5806e-03, 7.7868e-04], + [ 4.4346e-04, 8.8334e-05, -9.4604e-04, ..., -1.0133e-04, + 5.3501e-04, -2.6798e-04], + [ 5.0735e-04, -4.4518e-03, -4.5052e-03, ..., 1.0207e-05, + -1.9093e-03, -2.7637e-03], + ..., + [-2.5711e-03, 3.1543e-04, 3.0756e-04, ..., 7.5102e-06, + -4.7231e-04, -4.3945e-03], + [-2.6360e-03, 1.3638e-03, 1.2875e-03, ..., 1.5646e-05, + -1.1549e-03, 1.9588e-03], + [ 1.9855e-03, 5.5647e-04, 3.3188e-04, ..., 7.2755e-06, + 1.9350e-03, 4.9543e-04]], device='cuda:0') +Epoch 301, bias, value: tensor([ 0.0068, 0.0022, -0.0015, -0.0192, 0.0107, -0.0098, -0.0143, -0.0191, + 0.0026, 0.0021], device='cuda:0'), grad: tensor([-0.0196, 0.0056, -0.0079, 0.0144, 0.0122, 0.0206, 0.0188, -0.0694, + 0.0009, 0.0245], device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 216.80, cls_loss 0.4744 cls_loss_mapping 0.0036 cls_loss_causal 0.4468 re_mapping 0.0059 re_causal 0.0159 /// teacc 98.86 lr 0.00010000 +Epoch 302, weight, value: tensor([[ 0.0354, 0.1080, -0.1643, ..., -0.0984, 0.0719, 0.0044], + [-0.0889, -0.1007, 0.1016, ..., -0.0384, -0.0968, 0.0488], + [-0.0810, -0.0591, -0.1100, ..., -0.1293, -0.0327, 0.0551], + ..., + [-0.0167, -0.1236, -0.1271, ..., -0.1097, -0.0387, 0.0725], + [-0.1162, -0.1039, -0.1049, ..., -0.0596, -0.0488, -0.0940], + [ 0.0157, -0.0261, -0.0606, ..., -0.0964, 0.0231, -0.0198]], + device='cuda:0'), grad: tensor([[-3.0880e-03, -1.0700e-03, 4.1097e-05, ..., -1.3323e-03, + -4.6120e-03, -1.0475e-02], + [ 8.3637e-04, 3.2640e-04, 1.5929e-05, ..., 3.7432e-04, + 7.7629e-04, 1.6956e-03], + [ 3.7074e-04, 1.4853e-04, 3.4988e-05, ..., 1.7750e-04, + 2.7490e-04, 6.1893e-04], + ..., + [ 6.3133e-04, 1.0037e-04, 5.8077e-06, ..., 1.3506e-04, + 1.7900e-03, 5.8174e-03], + [ 1.2436e-03, 7.3528e-04, 1.4949e-04, ..., 3.2020e-04, + 2.8086e-04, 1.7281e-03], + [ 3.7313e-04, 1.3983e-04, 1.6928e-05, ..., 1.6713e-04, + 3.1018e-04, 7.9775e-04]], device='cuda:0') +Epoch 302, bias, value: tensor([ 0.0064, 0.0029, -0.0010, -0.0195, 0.0110, -0.0096, -0.0147, -0.0192, + 0.0035, 0.0007], device='cuda:0'), grad: tensor([-0.0638, 0.0197, 0.0097, -0.0104, 0.0140, -0.0511, 0.0146, 0.0232, + 0.0346, 0.0095], device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 216.21, cls_loss 0.4909 cls_loss_mapping 0.0031 cls_loss_causal 0.4636 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.68 lr 0.00010000 +Epoch 303, weight, value: tensor([[ 0.0355, 0.1076, -0.1641, ..., -0.0971, 0.0730, 0.0052], + [-0.0891, -0.1006, 0.1013, ..., -0.0385, -0.0962, 0.0499], + [-0.0808, -0.0597, -0.1093, ..., -0.1295, -0.0326, 0.0567], + ..., + [-0.0170, -0.1235, -0.1275, ..., -0.1092, -0.0389, 0.0707], + [-0.1179, -0.1041, -0.1054, ..., -0.0613, -0.0503, -0.0961], + [ 0.0165, -0.0252, -0.0604, ..., -0.0959, 0.0236, -0.0206]], + device='cuda:0'), grad: tensor([[ 0.0008, 0.0007, 0.0005, ..., 0.0008, 0.0011, 0.0012], + [ 0.0003, 0.0003, 0.0002, ..., 0.0003, 0.0005, 0.0008], + [-0.0003, 0.0003, 0.0004, ..., 0.0004, 0.0004, -0.0018], + ..., + [-0.0002, 0.0002, 0.0004, ..., 0.0001, -0.0008, -0.0003], + [ 0.0004, -0.0109, 0.0007, ..., 0.0007, 0.0006, 0.0008], + [ 0.0003, 0.0002, -0.0016, ..., 0.0003, 0.0005, -0.0012]], + device='cuda:0') +Epoch 303, bias, value: tensor([ 0.0063, 0.0030, -0.0005, -0.0190, 0.0105, -0.0089, -0.0140, -0.0189, + 0.0020, 0.0002], device='cuda:0'), grad: tensor([ 0.0241, 0.0178, -0.0087, 0.0670, 0.0154, -0.0101, -0.0382, -0.0112, + -0.0132, -0.0427], device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 216.47, cls_loss 0.5195 cls_loss_mapping 0.0046 cls_loss_causal 0.4992 re_mapping 0.0057 re_causal 0.0148 /// teacc 98.59 lr 0.00010000 +Epoch 304, weight, value: tensor([[ 0.0355, 0.1074, -0.1635, ..., -0.0971, 0.0728, 0.0046], + [-0.0904, -0.0993, 0.1033, ..., -0.0381, -0.0964, 0.0495], + [-0.0801, -0.0610, -0.1110, ..., -0.1291, -0.0321, 0.0575], + ..., + [-0.0171, -0.1241, -0.1284, ..., -0.1100, -0.0392, 0.0698], + [-0.1157, -0.1034, -0.1064, ..., -0.0611, -0.0504, -0.0940], + [ 0.0177, -0.0254, -0.0602, ..., -0.0959, 0.0247, -0.0211]], + device='cuda:0'), grad: tensor([[ 5.0879e-04, 3.4839e-05, 7.0274e-05, ..., 2.5388e-06, + 5.9462e-04, 7.3576e-04], + [-6.1846e-04, -2.9349e-04, -7.5960e-04, ..., 6.7055e-08, + 1.3721e-04, -2.3842e-03], + [-2.5654e-03, -3.8171e-04, 1.1817e-05, ..., 1.8720e-07, + -1.1606e-03, -5.3024e-04], + ..., + [ 1.2712e-03, 1.9014e-04, 2.6059e-04, ..., 1.3039e-08, + 9.2554e-04, 3.5782e-03], + [ 5.0068e-04, -1.8203e-04, -3.0422e-04, ..., 9.0897e-06, + 2.0778e-04, 9.0361e-04], + [ 5.3585e-05, 1.0145e-04, 2.2352e-04, ..., 9.9652e-08, + -6.8784e-05, -2.1553e-03]], device='cuda:0') +Epoch 304, bias, value: tensor([ 0.0059, 0.0018, -0.0006, -0.0190, 0.0107, -0.0096, -0.0136, -0.0190, + 0.0023, 0.0017], device='cuda:0'), grad: tensor([ 0.0219, -0.0318, -0.0058, 0.0268, -0.0045, -0.0113, -0.0419, 0.0394, + 0.0195, -0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 216.16, cls_loss 0.4753 cls_loss_mapping 0.0045 cls_loss_causal 0.4486 re_mapping 0.0060 re_causal 0.0155 /// teacc 98.68 lr 0.00010000 +Epoch 305, weight, value: tensor([[ 0.0352, 0.1081, -0.1638, ..., -0.0976, 0.0730, 0.0042], + [-0.0904, -0.0989, 0.1051, ..., -0.0385, -0.0963, 0.0500], + [-0.0807, -0.0611, -0.1111, ..., -0.1293, -0.0322, 0.0567], + ..., + [-0.0178, -0.1259, -0.1296, ..., -0.1092, -0.0400, 0.0688], + [-0.1171, -0.1043, -0.1074, ..., -0.0617, -0.0519, -0.0949], + [ 0.0179, -0.0262, -0.0614, ..., -0.0960, 0.0236, -0.0197]], + device='cuda:0'), grad: tensor([[ 4.8327e-04, 2.0623e-04, 1.8224e-05, ..., 9.7334e-05, + 1.1568e-03, 4.7708e-04], + [ 3.5375e-05, 8.0705e-05, -2.9540e-04, ..., 2.1803e-04, + 1.5326e-03, -1.4462e-05], + [-3.5534e-03, 1.1587e-03, 1.9804e-05, ..., 1.7452e-04, + -3.1223e-03, 5.8985e-04], + ..., + [ 8.1825e-04, 3.3188e-04, 1.0151e-04, ..., 1.5545e-04, + 1.2722e-03, 1.3514e-03], + [ 4.2081e-04, 1.8263e-04, 3.3915e-05, ..., 8.0705e-05, + 9.6035e-04, 3.1590e-04], + [-7.1526e-04, -7.9060e-04, -3.9905e-05, ..., -1.3027e-03, + -9.2745e-04, -3.5133e-03]], device='cuda:0') +Epoch 305, bias, value: tensor([ 0.0062, 0.0013, -0.0011, -0.0190, 0.0106, -0.0097, -0.0127, -0.0190, + 0.0018, 0.0022], device='cuda:0'), grad: tensor([ 0.0212, -0.0294, -0.0008, -0.0145, 0.0024, -0.0126, 0.0264, 0.0323, + 0.0186, -0.0439], device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 216.56, cls_loss 0.4867 cls_loss_mapping 0.0029 cls_loss_causal 0.4611 re_mapping 0.0057 re_causal 0.0157 /// teacc 98.83 lr 0.00010000 +Epoch 306, weight, value: tensor([[ 0.0348, 0.1075, -0.1640, ..., -0.0986, 0.0727, 0.0031], + [-0.0900, -0.0984, 0.1059, ..., -0.0389, -0.0966, 0.0503], + [-0.0807, -0.0618, -0.1109, ..., -0.1284, -0.0321, 0.0577], + ..., + [-0.0167, -0.1265, -0.1300, ..., -0.1088, -0.0400, 0.0688], + [-0.1177, -0.1044, -0.1072, ..., -0.0617, -0.0530, -0.0948], + [ 0.0181, -0.0264, -0.0606, ..., -0.0967, 0.0242, -0.0198]], + device='cuda:0'), grad: tensor([[ 4.0472e-05, 1.5152e-04, 5.5820e-05, ..., 1.9813e-04, + 2.0826e-04, 3.6883e-04], + [ 6.8128e-05, -3.9116e-08, -1.2188e-03, ..., 2.1350e-04, + 2.6917e-04, -4.5371e-04], + [ 1.2350e-04, 2.3079e-04, 3.2163e-04, ..., 3.1471e-04, + 4.5395e-04, 1.4105e-03], + ..., + [ 5.7459e-05, 6.8855e-04, 6.0272e-04, ..., 8.3745e-05, + 1.8275e-04, 1.0824e-03], + [ 7.2539e-05, 1.0824e-03, 5.5522e-05, ..., 1.1140e-04, + 2.5797e-04, 8.7693e-06], + [ 5.6356e-05, -2.9316e-03, 3.2663e-04, ..., 1.3268e-04, + 7.7009e-05, 6.1798e-04]], device='cuda:0') +Epoch 306, bias, value: tensor([ 0.0058, 0.0017, -0.0013, -0.0182, 0.0099, -0.0098, -0.0124, -0.0188, + 0.0017, 0.0019], device='cuda:0'), grad: tensor([ 0.0074, 0.0037, 0.0167, 0.0114, 0.0138, 0.0075, -0.0374, 0.0117, + 0.0030, -0.0378], device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 216.47, cls_loss 0.4821 cls_loss_mapping 0.0030 cls_loss_causal 0.4561 re_mapping 0.0059 re_causal 0.0160 /// teacc 98.71 lr 0.00010000 +Epoch 307, weight, value: tensor([[ 0.0337, 0.1078, -0.1654, ..., -0.0993, 0.0726, 0.0025], + [-0.0903, -0.0982, 0.1065, ..., -0.0380, -0.0968, 0.0515], + [-0.0795, -0.0626, -0.1112, ..., -0.1283, -0.0316, 0.0573], + ..., + [-0.0171, -0.1269, -0.1295, ..., -0.1090, -0.0411, 0.0691], + [-0.1181, -0.1047, -0.1070, ..., -0.0613, -0.0533, -0.0955], + [ 0.0182, -0.0271, -0.0611, ..., -0.0965, 0.0238, -0.0190]], + device='cuda:0'), grad: tensor([[ 1.7121e-05, 7.3051e-03, 1.8539e-03, ..., 6.4313e-05, + 7.9453e-05, 2.5558e-04], + [ 1.8895e-05, 9.6321e-04, 7.0190e-03, ..., 2.2659e-03, + 8.5950e-05, -4.8971e-04], + [ 1.3210e-05, -1.5430e-03, -9.1476e-03, ..., -2.6817e-03, + 6.8367e-05, 4.6968e-04], + ..., + [-1.6582e-04, -5.6553e-04, -3.4189e-04, ..., 6.1002e-07, + -6.7616e-04, 3.3379e-03], + [ 1.7136e-05, 1.7710e-03, 8.2552e-05, ..., 3.0589e-04, + 9.3430e-06, 5.4121e-04], + [ 2.5839e-05, 4.7898e-04, 4.2343e-04, ..., 5.4389e-06, + 1.0443e-04, 3.4022e-04]], device='cuda:0') +Epoch 307, bias, value: tensor([ 0.0059, 0.0020, -0.0019, -0.0180, 0.0092, -0.0100, -0.0123, -0.0184, + 0.0023, 0.0017], device='cuda:0'), grad: tensor([ 0.0370, -0.0010, 0.0002, -0.0691, 0.0143, 0.0334, 0.0022, -0.0225, + -0.0146, 0.0202], device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 216.92, cls_loss 0.4783 cls_loss_mapping 0.0048 cls_loss_causal 0.4528 re_mapping 0.0065 re_causal 0.0171 /// teacc 98.74 lr 0.00010000 +Epoch 308, weight, value: tensor([[ 0.0351, 0.1083, -0.1647, ..., -0.0984, 0.0738, 0.0046], + [-0.0910, -0.0997, 0.1049, ..., -0.0388, -0.0980, 0.0507], + [-0.0798, -0.0623, -0.1109, ..., -0.1271, -0.0315, 0.0565], + ..., + [-0.0171, -0.1263, -0.1289, ..., -0.1105, -0.0399, 0.0698], + [-0.1190, -0.1053, -0.1073, ..., -0.0620, -0.0533, -0.0961], + [ 0.0180, -0.0276, -0.0620, ..., -0.0965, 0.0242, -0.0185]], + device='cuda:0'), grad: tensor([[ 1.7059e-04, 5.3272e-07, 2.3782e-04, ..., 2.5388e-06, + 2.2784e-05, 4.2987e-04], + [ 1.8799e-04, 3.8236e-05, -3.6073e-04, ..., -1.6633e-06, + 2.5049e-05, -6.6833e-03], + [ 2.2316e-04, 4.9949e-05, 3.3617e-05, ..., 1.6568e-06, + 3.2544e-05, 1.8463e-03], + ..., + [ 5.6791e-04, 1.9753e-04, 5.2989e-05, ..., 1.2293e-07, + 6.7353e-05, 1.4944e-03], + [-5.4836e-04, -8.2350e-04, 2.2185e-04, ..., 3.5018e-05, + 4.0919e-05, 1.2207e-03], + [ 1.7672e-03, 3.4690e-04, 5.8842e-04, ..., 6.2957e-07, + 6.7294e-05, 7.5769e-04]], device='cuda:0') +Epoch 308, bias, value: tensor([ 0.0067, 0.0013, -0.0018, -0.0181, 0.0094, -0.0105, -0.0128, -0.0179, + 0.0025, 0.0017], device='cuda:0'), grad: tensor([ 0.0045, -0.0333, 0.0139, 0.0063, 0.0142, 0.0034, -0.0270, -0.0151, + 0.0151, 0.0180], device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 216.52, cls_loss 0.4809 cls_loss_mapping 0.0043 cls_loss_causal 0.4498 re_mapping 0.0059 re_causal 0.0153 /// teacc 98.69 lr 0.00010000 +Epoch 309, weight, value: tensor([[ 0.0349, 0.1092, -0.1651, ..., -0.0976, 0.0741, 0.0042], + [-0.0917, -0.0999, 0.1041, ..., -0.0379, -0.0975, 0.0514], + [-0.0804, -0.0619, -0.1100, ..., -0.1276, -0.0313, 0.0566], + ..., + [-0.0163, -0.1262, -0.1285, ..., -0.1113, -0.0391, 0.0693], + [-0.1185, -0.1067, -0.1079, ..., -0.0626, -0.0526, -0.0954], + [ 0.0181, -0.0300, -0.0618, ..., -0.0974, 0.0218, -0.0178]], + device='cuda:0'), grad: tensor([[-1.3041e-04, -2.2812e-03, 7.3373e-05, ..., -1.0452e-03, + -4.7636e-04, 4.6945e-04], + [ 1.6975e-04, 1.9932e-04, 4.1962e-05, ..., 1.1271e-04, + -6.6566e-04, -1.2980e-03], + [ 6.9761e-04, 1.1930e-03, 2.2006e-04, ..., 4.4703e-04, + -3.0303e-04, 7.7724e-04], + ..., + [-1.6680e-03, -1.1940e-03, 1.5914e-04, ..., 2.4307e-04, + -1.6060e-03, -1.2789e-03], + [ 3.1304e-04, 6.4039e-04, 5.4646e-04, ..., 2.1324e-03, + 4.6897e-04, 4.5395e-04], + [-1.2102e-03, -2.0161e-03, -4.6492e-04, ..., -2.1801e-03, + -4.1795e-04, -1.2197e-03]], device='cuda:0') +Epoch 309, bias, value: tensor([ 0.0073, 0.0008, -0.0020, -0.0183, 0.0095, -0.0108, -0.0129, -0.0181, + 0.0040, 0.0010], device='cuda:0'), grad: tensor([ 6.5613e-03, -4.7974e-02, -1.1223e-02, 2.1042e-02, 1.5495e-02, + 9.2924e-05, 1.5450e-02, -1.5457e-02, 1.7059e-02, -1.0653e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 216.73, cls_loss 0.4808 cls_loss_mapping 0.0028 cls_loss_causal 0.4504 re_mapping 0.0062 re_causal 0.0161 /// teacc 98.69 lr 0.00010000 +Epoch 310, weight, value: tensor([[ 0.0351, 0.1093, -0.1663, ..., -0.0985, 0.0747, 0.0040], + [-0.0917, -0.0989, 0.1060, ..., -0.0374, -0.0978, 0.0526], + [-0.0816, -0.0617, -0.1105, ..., -0.1279, -0.0300, 0.0570], + ..., + [-0.0161, -0.1265, -0.1292, ..., -0.1121, -0.0401, 0.0699], + [-0.1182, -0.1059, -0.1082, ..., -0.0619, -0.0540, -0.0965], + [ 0.0190, -0.0298, -0.0625, ..., -0.0955, 0.0218, -0.0181]], + device='cuda:0'), grad: tensor([[ 1.4076e-03, 7.5626e-04, 4.0650e-05, ..., 9.2602e-04, + 1.3523e-03, 9.5224e-04], + [-1.2045e-03, -1.5259e-03, 1.4043e-04, ..., -4.8828e-04, + -3.6740e-04, -8.6365e-03], + [-6.4898e-04, 8.5020e-04, 1.1694e-04, ..., -2.4281e-03, + 4.1223e-04, 3.0041e-03], + ..., + [-2.8191e-03, -2.0866e-03, -4.9621e-05, ..., 7.8058e-04, + -3.0823e-03, -2.7618e-03], + [-2.4986e-03, -2.2030e-03, -2.6512e-04, ..., -5.7507e-04, + -1.2770e-03, 5.8603e-04], + [ 2.3098e-03, 6.2895e-04, 1.1122e-04, ..., 8.1348e-04, + 1.8549e-03, 3.4676e-03]], device='cuda:0') +Epoch 310, bias, value: tensor([ 0.0069, 0.0017, -0.0018, -0.0183, 0.0095, -0.0105, -0.0133, -0.0184, + 0.0034, 0.0013], device='cuda:0'), grad: tensor([ 0.0395, -0.0666, -0.0352, 0.0205, 0.0048, -0.0063, 0.0259, -0.0067, + -0.0057, 0.0297], device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 216.39, cls_loss 0.4766 cls_loss_mapping 0.0037 cls_loss_causal 0.4496 re_mapping 0.0063 re_causal 0.0162 /// teacc 98.68 lr 0.00010000 +Epoch 311, weight, value: tensor([[ 0.0346, 0.1098, -0.1675, ..., -0.0994, 0.0728, 0.0035], + [-0.0917, -0.0984, 0.1070, ..., -0.0369, -0.0969, 0.0527], + [-0.0818, -0.0620, -0.1105, ..., -0.1288, -0.0303, 0.0575], + ..., + [-0.0157, -0.1287, -0.1296, ..., -0.1130, -0.0399, 0.0695], + [-0.1181, -0.1050, -0.1072, ..., -0.0616, -0.0534, -0.0949], + [ 0.0193, -0.0297, -0.0624, ..., -0.0961, 0.0210, -0.0192]], + device='cuda:0'), grad: tensor([[-2.9888e-03, -5.0697e-03, -9.6941e-04, ..., -1.6909e-03, + -2.0695e-03, -6.0940e-04], + [-2.0385e-05, 1.6958e-05, -4.5085e-04, ..., 6.1989e-04, + -8.1301e-04, -1.7834e-03], + [ 3.6192e-04, -1.1721e-03, 3.9291e-04, ..., 1.8132e-04, + 7.8201e-04, -1.5974e-03], + ..., + [-3.1281e-04, 1.1182e-04, 3.3236e-04, ..., 7.0453e-05, + -3.2008e-05, -2.9068e-03], + [ 3.3140e-04, 4.4274e-04, 5.8603e-04, ..., 2.7537e-04, + 3.6693e-04, 5.7459e-04], + [ 7.5150e-04, 5.7459e-05, 1.7428e-04, ..., 2.1529e-04, + 4.6945e-04, 3.6373e-03]], device='cuda:0') +Epoch 311, bias, value: tensor([ 0.0065, 0.0021, -0.0022, -0.0188, 0.0100, -0.0118, -0.0119, -0.0180, + 0.0035, 0.0008], device='cuda:0'), grad: tensor([-7.8506e-03, -1.0040e-02, 2.6822e-04, 2.4216e-02, -6.5923e-05, + -4.3335e-02, 1.6571e-02, -9.5825e-03, 8.3160e-03, 2.1530e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 216.34, cls_loss 0.4816 cls_loss_mapping 0.0058 cls_loss_causal 0.4585 re_mapping 0.0061 re_causal 0.0161 /// teacc 98.76 lr 0.00010000 +Epoch 312, weight, value: tensor([[ 0.0340, 0.1112, -0.1680, ..., -0.0999, 0.0715, 0.0047], + [-0.0919, -0.1006, 0.1072, ..., -0.0377, -0.0979, 0.0537], + [-0.0824, -0.0617, -0.1108, ..., -0.1295, -0.0304, 0.0569], + ..., + [-0.0151, -0.1290, -0.1290, ..., -0.1128, -0.0388, 0.0703], + [-0.1182, -0.1051, -0.1075, ..., -0.0624, -0.0522, -0.0954], + [ 0.0188, -0.0291, -0.0636, ..., -0.0968, 0.0203, -0.0210]], + device='cuda:0'), grad: tensor([[ 2.0237e-03, 1.3107e-02, 2.4652e-04, ..., 4.4465e-04, + 1.7700e-02, 1.7953e-04], + [ 1.8269e-05, -6.5851e-04, -2.7103e-03, ..., -9.5367e-04, + -1.5221e-03, -1.0767e-03], + [-1.8406e-03, -1.3298e-02, 2.0826e-04, ..., 1.9670e-04, + -1.7548e-02, -1.0815e-03], + ..., + [ 1.3106e-05, 6.6757e-05, 4.3941e-04, ..., 9.7275e-05, + 2.4509e-04, -1.7619e-04], + [ 6.9380e-05, 4.2653e-04, 3.9840e-04, ..., 1.2565e-04, + 5.6267e-04, 2.6870e-04], + [ 7.2598e-05, 7.8142e-05, 2.0087e-04, ..., 8.0228e-05, + 2.7871e-04, 3.2187e-04]], device='cuda:0') +Epoch 312, bias, value: tensor([ 0.0073, 0.0016, -0.0023, -0.0183, 0.0091, -0.0112, -0.0117, -0.0176, + 0.0033, 0.0003], device='cuda:0'), grad: tensor([ 0.0408, -0.0458, -0.0126, -0.0091, 0.0172, 0.0088, -0.0205, -0.0030, + 0.0119, 0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 216.22, cls_loss 0.4874 cls_loss_mapping 0.0042 cls_loss_causal 0.4643 re_mapping 0.0061 re_causal 0.0163 /// teacc 98.86 lr 0.00010000 +Epoch 313, weight, value: tensor([[ 0.0352, 0.1115, -0.1690, ..., -0.0992, 0.0712, 0.0048], + [-0.0932, -0.1019, 0.1058, ..., -0.0387, -0.0973, 0.0543], + [-0.0820, -0.0620, -0.1111, ..., -0.1288, -0.0286, 0.0570], + ..., + [-0.0156, -0.1276, -0.1275, ..., -0.1120, -0.0396, 0.0702], + [-0.1185, -0.1043, -0.1088, ..., -0.0630, -0.0521, -0.0966], + [ 0.0181, -0.0293, -0.0631, ..., -0.0970, 0.0194, -0.0210]], + device='cuda:0'), grad: tensor([[ 6.8188e-04, 3.3832e-04, 1.8036e-04, ..., 4.0196e-06, + 5.1832e-04, 4.6134e-05], + [ 1.3673e-04, 7.9775e-04, 4.2820e-04, ..., 1.9688e-06, + 1.7476e-04, 5.4181e-05], + [-5.0774e-03, -2.0504e-03, -1.5545e-03, ..., 5.3549e-04, + -3.1815e-03, 6.2466e-04], + ..., + [ 3.3617e-04, 2.8801e-04, 1.4579e-04, ..., 6.8918e-06, + 2.8396e-04, 2.4065e-05], + [ 9.2793e-04, 3.4302e-05, 5.4789e-04, ..., -6.3133e-04, + 9.1600e-04, -1.4973e-03], + [ 3.4857e-04, 1.9205e-04, 1.2052e-04, ..., 3.8855e-06, + 2.7680e-04, 5.7369e-05]], device='cuda:0') +Epoch 313, bias, value: tensor([ 0.0068, 0.0023, -0.0022, -0.0184, 0.0095, -0.0117, -0.0110, -0.0184, + 0.0033, 0.0003], device='cuda:0'), grad: tensor([-0.0251, 0.0439, -0.0529, -0.0174, 0.0113, 0.0108, 0.0138, -0.0183, + 0.0265, 0.0074], device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 216.72, cls_loss 0.4838 cls_loss_mapping 0.0040 cls_loss_causal 0.4609 re_mapping 0.0067 re_causal 0.0170 /// teacc 98.79 lr 0.00010000 +Epoch 314, weight, value: tensor([[ 0.0358, 0.1116, -0.1692, ..., -0.0977, 0.0711, 0.0049], + [-0.0941, -0.1012, 0.1066, ..., -0.0391, -0.0981, 0.0541], + [-0.0824, -0.0634, -0.1116, ..., -0.1292, -0.0290, 0.0565], + ..., + [-0.0158, -0.1281, -0.1280, ..., -0.1119, -0.0388, 0.0701], + [-0.1192, -0.1039, -0.1108, ..., -0.0643, -0.0522, -0.0965], + [ 0.0183, -0.0287, -0.0622, ..., -0.0972, 0.0205, -0.0218]], + device='cuda:0'), grad: tensor([[ 2.7108e-04, 1.1653e-05, 1.6260e-04, ..., -2.8872e-04, + 4.8423e-04, 7.3910e-04], + [-4.2152e-04, 3.1292e-05, -7.7486e-06, ..., 1.0744e-05, + -3.9667e-05, 2.5501e-03], + [ 2.3425e-04, 1.3280e-04, 5.8353e-05, ..., 1.9521e-05, + 3.2949e-04, 1.0033e-03], + ..., + [ 3.8338e-04, -3.6788e-04, -4.0412e-04, ..., -1.4946e-05, + 8.0633e-04, 2.2526e-03], + [ 2.0134e-04, 1.7822e-04, 2.2724e-05, ..., 1.5058e-05, + 3.8981e-04, 9.2125e-04], + [-1.3332e-03, 4.0674e-04, 3.3522e-04, ..., 3.1948e-05, + -2.0924e-03, -1.0195e-03]], device='cuda:0') +Epoch 314, bias, value: tensor([ 0.0081, 0.0029, -0.0038, -0.0191, 0.0090, -0.0113, -0.0118, -0.0173, + 0.0034, 0.0002], device='cuda:0'), grad: tensor([ 0.0214, -0.0239, 0.0190, -0.0407, 0.0243, -0.0028, -0.0106, 0.0442, + 0.0268, -0.0578], device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 216.34, cls_loss 0.4937 cls_loss_mapping 0.0036 cls_loss_causal 0.4774 re_mapping 0.0062 re_causal 0.0163 /// teacc 98.84 lr 0.00010000 +Epoch 315, weight, value: tensor([[ 0.0360, 0.1114, -0.1696, ..., -0.0981, 0.0711, 0.0048], + [-0.0939, -0.1017, 0.1073, ..., -0.0395, -0.0987, 0.0535], + [-0.0825, -0.0639, -0.1109, ..., -0.1290, -0.0286, 0.0571], + ..., + [-0.0172, -0.1279, -0.1292, ..., -0.1106, -0.0391, 0.0711], + [-0.1192, -0.1035, -0.1109, ..., -0.0634, -0.0518, -0.0966], + [ 0.0203, -0.0287, -0.0611, ..., -0.0967, 0.0209, -0.0222]], + device='cuda:0'), grad: tensor([[-1.5535e-03, -1.5860e-03, -1.6088e-03, ..., -1.4820e-03, + -1.1377e-03, 6.4182e-04], + [ 1.0433e-03, 3.2640e-04, 6.2323e-04, ..., 4.2605e-04, + 8.5211e-04, 7.5006e-04], + [ 1.6594e-03, 3.4356e-04, 6.4993e-04, ..., 4.4680e-04, + 9.0742e-04, 1.2627e-03], + ..., + [-4.8184e-04, 6.2525e-05, -2.0170e-04, ..., -3.7622e-04, + 1.9860e-04, -1.4582e-03], + [-5.0087e-03, 8.3819e-06, -1.7509e-03, ..., -3.9577e-04, + -9.0218e-04, -3.6259e-03], + [ 1.6069e-03, 5.1737e-04, 8.1778e-04, ..., 6.3276e-04, + 1.2274e-03, 9.4557e-04]], device='cuda:0') +Epoch 315, bias, value: tensor([ 0.0079, 0.0029, -0.0031, -0.0192, 0.0082, -0.0117, -0.0113, -0.0168, + 0.0025, 0.0009], device='cuda:0'), grad: tensor([-0.0171, 0.0283, 0.0056, 0.0290, -0.0127, 0.0212, -0.0147, -0.0102, + -0.0554, 0.0259], device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 216.41, cls_loss 0.4990 cls_loss_mapping 0.0028 cls_loss_causal 0.4803 re_mapping 0.0067 re_causal 0.0191 /// teacc 98.85 lr 0.00010000 +Epoch 316, weight, value: tensor([[ 0.0359, 0.1114, -0.1678, ..., -0.0982, 0.0709, 0.0049], + [-0.0938, -0.1004, 0.1065, ..., -0.0401, -0.0980, 0.0536], + [-0.0843, -0.0642, -0.1117, ..., -0.1285, -0.0291, 0.0565], + ..., + [-0.0172, -0.1282, -0.1300, ..., -0.1107, -0.0387, 0.0708], + [-0.1181, -0.1036, -0.1112, ..., -0.0634, -0.0517, -0.0960], + [ 0.0197, -0.0288, -0.0608, ..., -0.0978, 0.0198, -0.0229]], + device='cuda:0'), grad: tensor([[ 1.1206e-04, 2.3210e-04, 1.6287e-05, ..., 4.8542e-04, + 1.2946e-04, 2.6631e-04], + [-9.4223e-04, 2.8059e-05, -7.9572e-05, ..., 9.1553e-04, + -1.1139e-03, 6.0701e-04], + [ 9.4533e-05, 2.2459e-04, 4.3482e-05, ..., 3.5882e-04, + 1.0586e-04, 2.0576e-04], + ..., + [-3.9506e-04, 4.3929e-05, 4.0010e-06, ..., 6.8307e-05, + 1.9443e-04, -3.5334e-04], + [ 2.1780e-04, 9.8765e-05, -6.6662e-04, ..., 4.2111e-05, + 1.4305e-04, 2.1851e-04], + [-3.4237e-04, 1.0729e-03, 3.9995e-05, ..., 1.2076e-04, + -1.7929e-04, -4.3988e-04]], device='cuda:0') +Epoch 316, bias, value: tensor([ 0.0079, 0.0039, -0.0031, -0.0185, 0.0077, -0.0116, -0.0127, -0.0173, + 0.0033, 0.0008], device='cuda:0'), grad: tensor([ 0.0065, -0.0270, 0.0047, -0.0052, 0.0074, 0.0060, -0.0043, 0.0023, + 0.0051, 0.0045], device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 216.16, cls_loss 0.4725 cls_loss_mapping 0.0046 cls_loss_causal 0.4535 re_mapping 0.0062 re_causal 0.0164 /// teacc 98.88 lr 0.00010000 +Epoch 317, weight, value: tensor([[ 0.0354, 0.1111, -0.1664, ..., -0.0981, 0.0708, 0.0050], + [-0.0945, -0.1012, 0.1060, ..., -0.0417, -0.0987, 0.0523], + [-0.0843, -0.0634, -0.1117, ..., -0.1285, -0.0291, 0.0565], + ..., + [-0.0168, -0.1285, -0.1288, ..., -0.1090, -0.0386, 0.0713], + [-0.1198, -0.1046, -0.1118, ..., -0.0637, -0.0524, -0.0965], + [ 0.0202, -0.0303, -0.0612, ..., -0.0959, 0.0206, -0.0226]], + device='cuda:0'), grad: tensor([[-1.1597e-03, 4.4554e-05, 5.7602e-07, ..., 1.7297e-04, + 2.7776e-04, 3.3006e-06], + [-2.2030e-03, 4.3735e-06, 5.7789e-07, ..., 5.5224e-05, + -2.1744e-03, -1.4290e-02], + [ 3.8586e-03, 2.4140e-05, -1.8524e-06, ..., 2.0142e-03, + 4.9114e-04, -7.3910e-04], + ..., + [-4.4823e-03, 1.5132e-05, 1.3243e-06, ..., 3.4831e-06, + -1.0319e-03, 1.4580e-02], + [ 1.2207e-03, 1.7214e-03, 1.5535e-03, ..., 5.7793e-03, + 3.1209e-04, 3.5262e-04], + [-3.7975e-03, 8.0109e-05, 1.6298e-06, ..., 3.0518e-05, + 4.5347e-04, 1.5192e-05]], device='cuda:0') +Epoch 317, bias, value: tensor([ 0.0094, 0.0024, -0.0021, -0.0189, 0.0082, -0.0119, -0.0125, -0.0174, + 0.0026, 0.0007], device='cuda:0'), grad: tensor([-0.0128, -0.1083, 0.0092, 0.0187, 0.0248, 0.0087, 0.0045, 0.0392, + 0.0255, -0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 216.76, cls_loss 0.4841 cls_loss_mapping 0.0048 cls_loss_causal 0.4584 re_mapping 0.0057 re_causal 0.0143 /// teacc 98.78 lr 0.00010000 +Epoch 318, weight, value: tensor([[ 0.0347, 0.1113, -0.1672, ..., -0.0981, 0.0691, 0.0038], + [-0.0943, -0.1020, 0.1061, ..., -0.0419, -0.0972, 0.0529], + [-0.0845, -0.0628, -0.1114, ..., -0.1275, -0.0291, 0.0574], + ..., + [-0.0151, -0.1287, -0.1295, ..., -0.1095, -0.0404, 0.0704], + [-0.1216, -0.1047, -0.1117, ..., -0.0655, -0.0517, -0.0967], + [ 0.0212, -0.0300, -0.0600, ..., -0.0958, 0.0210, -0.0224]], + device='cuda:0'), grad: tensor([[ 1.0805e-03, 9.1696e-04, 1.9026e-04, ..., 7.2575e-04, + 8.7643e-04, 3.2568e-04], + [-8.9836e-04, 2.9707e-04, 1.9252e-04, ..., 4.6635e-04, + 8.5735e-04, 1.5020e-03], + [ 1.2856e-03, 6.7472e-04, 3.1066e-04, ..., 3.2473e-04, + 9.1887e-04, -3.3550e-03], + ..., + [-3.6430e-03, -3.7384e-03, -3.1986e-03, ..., -2.9159e-04, + -4.4746e-03, -1.6189e-04], + [-9.7942e-04, -5.9038e-05, 8.2970e-05, ..., 2.4402e-04, + -1.8063e-03, -9.6130e-04], + [ 1.4954e-03, 1.0509e-03, 5.1689e-04, ..., 4.5705e-04, + 1.3084e-03, 7.2193e-04]], device='cuda:0') +Epoch 318, bias, value: tensor([ 0.0092, 0.0031, -0.0014, -0.0201, 0.0080, -0.0118, -0.0126, -0.0186, + 0.0031, 0.0015], device='cuda:0'), grad: tensor([ 0.0342, 0.0076, -0.0511, 0.0102, 0.0254, -0.0119, -0.0063, 0.0030, + -0.0464, 0.0354], device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 216.30, cls_loss 0.4657 cls_loss_mapping 0.0045 cls_loss_causal 0.4407 re_mapping 0.0063 re_causal 0.0165 /// teacc 98.87 lr 0.00010000 +Epoch 319, weight, value: tensor([[ 0.0372, 0.1108, -0.1672, ..., -0.0979, 0.0694, 0.0040], + [-0.0954, -0.1025, 0.1061, ..., -0.0427, -0.0977, 0.0533], + [-0.0859, -0.0624, -0.1128, ..., -0.1272, -0.0298, 0.0576], + ..., + [-0.0165, -0.1293, -0.1288, ..., -0.1093, -0.0407, 0.0694], + [-0.1213, -0.1047, -0.1110, ..., -0.0654, -0.0520, -0.0964], + [ 0.0208, -0.0297, -0.0594, ..., -0.0964, 0.0211, -0.0213]], + device='cuda:0'), grad: tensor([[ 4.2200e-04, 5.4806e-05, 1.6898e-05, ..., 1.5926e-06, + 5.2834e-04, 9.9838e-05], + [-1.7204e-03, 4.6492e-05, 1.1720e-05, ..., 1.5600e-07, + -1.6232e-03, -6.8903e-04], + [ 5.3263e-04, 2.2161e-04, 8.7917e-05, ..., 9.8348e-07, + 9.2888e-04, 1.2153e-04], + ..., + [ 6.6614e-04, -3.6931e-04, 1.4596e-05, ..., 1.8720e-07, + 1.1091e-03, -4.7684e-04], + [ 5.5456e-04, 1.3041e-04, 4.1872e-05, ..., 4.9397e-06, + 6.4421e-04, 1.8501e-04], + [-5.5933e-04, 3.9244e-04, 4.8727e-05, ..., 1.7639e-06, + -8.0585e-04, 8.4102e-05]], device='cuda:0') +Epoch 319, bias, value: tensor([ 0.0092, 0.0025, -0.0016, -0.0196, 0.0070, -0.0108, -0.0133, -0.0183, + 0.0036, 0.0019], device='cuda:0'), grad: tensor([ 0.0102, -0.0199, 0.0125, -0.0023, 0.0127, 0.0102, -0.0210, 0.0033, + 0.0120, -0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 216.46, cls_loss 0.4810 cls_loss_mapping 0.0040 cls_loss_causal 0.4595 re_mapping 0.0059 re_causal 0.0156 /// teacc 98.72 lr 0.00010000 +Epoch 320, weight, value: tensor([[ 0.0380, 0.1115, -0.1663, ..., -0.0980, 0.0699, 0.0041], + [-0.0950, -0.1028, 0.1060, ..., -0.0425, -0.0969, 0.0535], + [-0.0876, -0.0623, -0.1131, ..., -0.1269, -0.0306, 0.0573], + ..., + [-0.0152, -0.1291, -0.1300, ..., -0.1085, -0.0401, 0.0697], + [-0.1209, -0.1065, -0.1101, ..., -0.0654, -0.0516, -0.0959], + [ 0.0199, -0.0297, -0.0596, ..., -0.0969, 0.0202, -0.0209]], + device='cuda:0'), grad: tensor([[ 2.7370e-04, 2.1681e-05, 3.0428e-05, ..., 8.9169e-05, + -1.2331e-03, 2.8205e-04], + [ 2.4581e-04, 9.9540e-05, 8.0347e-05, ..., 7.6103e-04, + 2.9039e-04, -1.2612e-04], + [ 3.5596e-04, 2.9182e-04, -3.5286e-04, ..., 9.4533e-05, + 2.8276e-04, 1.8072e-04], + ..., + [ 4.2033e-04, 9.5963e-05, 6.4969e-05, ..., 3.6657e-05, + 3.2091e-04, 2.7823e-04], + [-7.3493e-05, -3.7422e-03, 5.7411e-04, ..., -4.6158e-03, + 2.3329e-04, 2.5177e-04], + [-4.5776e-04, -4.9973e-04, -2.3842e-04, ..., 9.3341e-05, + -1.3196e-04, -2.7037e-04]], device='cuda:0') +Epoch 320, bias, value: tensor([ 0.0086, 0.0036, -0.0021, -0.0203, 0.0067, -0.0103, -0.0137, -0.0181, + 0.0035, 0.0025], device='cuda:0'), grad: tensor([-0.0180, 0.0171, 0.0054, 0.0186, -0.0312, 0.0138, -0.0002, 0.0143, + -0.0285, 0.0085], device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 216.86, cls_loss 0.4961 cls_loss_mapping 0.0031 cls_loss_causal 0.4672 re_mapping 0.0061 re_causal 0.0167 /// teacc 98.84 lr 0.00010000 +Epoch 321, weight, value: tensor([[ 0.0381, 0.1123, -0.1667, ..., -0.0978, 0.0704, 0.0036], + [-0.0949, -0.1028, 0.1058, ..., -0.0440, -0.0972, 0.0520], + [-0.0898, -0.0629, -0.1130, ..., -0.1286, -0.0307, 0.0581], + ..., + [-0.0150, -0.1272, -0.1294, ..., -0.1086, -0.0400, 0.0692], + [-0.1197, -0.1044, -0.1107, ..., -0.0640, -0.0509, -0.0952], + [ 0.0196, -0.0297, -0.0602, ..., -0.0969, 0.0196, -0.0202]], + device='cuda:0'), grad: tensor([[ 1.6508e-03, 5.6811e-06, 2.2314e-06, ..., 6.8307e-05, + 1.0347e-03, 1.6146e-03], + [-7.8773e-04, 2.6133e-06, 1.7062e-06, ..., -1.9383e-04, + -1.9968e-04, -2.9678e-03], + [-4.6844e-03, 1.9878e-05, 1.2808e-05, ..., 2.6271e-05, + -2.7466e-03, -1.4668e-03], + ..., + [ 4.5943e-04, 5.2974e-06, 3.4142e-06, ..., 1.6558e-04, + 1.3518e-04, -1.9426e-03], + [ 5.2881e-04, 1.9014e-05, 1.2293e-05, ..., 2.8126e-06, + 2.5916e-04, 9.6703e-04], + [ 4.1747e-04, 1.8939e-05, 1.2323e-05, ..., 1.3607e-06, + 1.5819e-04, 7.3051e-04]], device='cuda:0') +Epoch 321, bias, value: tensor([ 0.0086, 0.0032, -0.0027, -0.0198, 0.0078, -0.0110, -0.0143, -0.0174, + 0.0044, 0.0015], device='cuda:0'), grad: tensor([ 0.0270, -0.0465, -0.0306, 0.0186, 0.0232, 0.0144, -0.0150, -0.0056, + 0.0199, -0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 216.44, cls_loss 0.4849 cls_loss_mapping 0.0027 cls_loss_causal 0.4565 re_mapping 0.0063 re_causal 0.0165 /// teacc 98.82 lr 0.00010000 +Epoch 322, weight, value: tensor([[ 0.0375, 0.1114, -0.1680, ..., -0.0982, 0.0705, 0.0042], + [-0.0938, -0.1023, 0.1058, ..., -0.0427, -0.0963, 0.0517], + [-0.0906, -0.0626, -0.1137, ..., -0.1290, -0.0303, 0.0580], + ..., + [-0.0151, -0.1274, -0.1297, ..., -0.1096, -0.0396, 0.0701], + [-0.1212, -0.1049, -0.1096, ..., -0.0635, -0.0512, -0.0959], + [ 0.0199, -0.0303, -0.0601, ..., -0.0977, 0.0191, -0.0205]], + device='cuda:0'), grad: tensor([[-2.0313e-03, 1.7792e-05, 1.3210e-05, ..., 2.2674e-04, + 2.0046e-03, -8.0109e-04], + [ 1.7080e-03, 5.5414e-08, -2.5630e-05, ..., 1.2375e-05, + 1.2693e-03, 2.5368e-04], + [ 7.7200e-04, 8.2701e-06, 6.6906e-06, ..., -2.0730e-04, + 6.9761e-04, -1.1415e-03], + ..., + [ 8.7309e-04, 6.1374e-07, 1.3366e-05, ..., 2.7791e-05, + 4.6778e-04, 2.2054e-04], + [ 9.3508e-04, 2.2367e-05, 1.7777e-05, ..., 7.2122e-05, + 6.2943e-04, 4.0054e-04], + [-6.1035e-04, 8.0913e-06, 5.8934e-06, ..., 1.8954e-05, + -8.3923e-03, 2.7347e-04]], device='cuda:0') +Epoch 322, bias, value: tensor([ 0.0081, 0.0030, -0.0027, -0.0198, 0.0091, -0.0106, -0.0149, -0.0179, + 0.0049, 0.0009], device='cuda:0'), grad: tensor([-0.0055, 0.0302, 0.0026, -0.0093, -0.0142, 0.0184, 0.0199, -0.0348, + -0.0071, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 216.78, cls_loss 0.4706 cls_loss_mapping 0.0035 cls_loss_causal 0.4374 re_mapping 0.0062 re_causal 0.0164 /// teacc 98.93 lr 0.00010000 +Epoch 323, weight, value: tensor([[ 0.0370, 0.1115, -0.1698, ..., -0.0986, 0.0703, 0.0030], + [-0.0937, -0.1023, 0.1061, ..., -0.0427, -0.0969, 0.0518], + [-0.0905, -0.0627, -0.1147, ..., -0.1298, -0.0293, 0.0588], + ..., + [-0.0142, -0.1259, -0.1270, ..., -0.1081, -0.0396, 0.0703], + [-0.1211, -0.1051, -0.1094, ..., -0.0640, -0.0509, -0.0955], + [ 0.0204, -0.0292, -0.0606, ..., -0.0981, 0.0203, -0.0205]], + device='cuda:0'), grad: tensor([[ 9.8228e-04, 2.1502e-05, 3.7581e-05, ..., 4.3213e-06, + 1.9825e-04, 2.5558e-03], + [ 2.2659e-03, 9.9987e-06, 2.0429e-05, ..., 1.8924e-06, + 2.7561e-04, 6.8932e-03], + [-3.5286e-03, -1.6479e-03, -3.5496e-03, ..., -2.9445e-04, + -7.4655e-06, -1.2466e-02], + ..., + [-1.4277e-03, 6.2883e-05, 1.1468e-04, ..., 2.0728e-05, + -7.6771e-04, -1.6918e-03], + [ 3.2997e-04, 4.4012e-04, 1.0958e-03, ..., 9.1970e-05, + 1.8299e-04, 1.1225e-03], + [ 9.3317e-04, 5.9414e-04, 1.1816e-03, ..., 8.6606e-05, + 2.5296e-04, 1.1015e-03]], device='cuda:0') +Epoch 323, bias, value: tensor([ 0.0083, 0.0034, -0.0019, -0.0193, 0.0082, -0.0109, -0.0149, -0.0186, + 0.0044, 0.0016], device='cuda:0'), grad: tensor([ 0.0212, -0.0174, -0.0783, -0.0090, 0.0198, 0.0138, 0.0152, -0.0113, + 0.0133, 0.0327], device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 216.48, cls_loss 0.4771 cls_loss_mapping 0.0041 cls_loss_causal 0.4497 re_mapping 0.0056 re_causal 0.0151 /// teacc 98.73 lr 0.00010000 +Epoch 324, weight, value: tensor([[ 0.0381, 0.1123, -0.1703, ..., -0.0988, 0.0702, 0.0035], + [-0.0949, -0.1027, 0.1056, ..., -0.0431, -0.0974, 0.0519], + [-0.0904, -0.0637, -0.1146, ..., -0.1299, -0.0286, 0.0599], + ..., + [-0.0143, -0.1262, -0.1269, ..., -0.1083, -0.0403, 0.0702], + [-0.1206, -0.1068, -0.1099, ..., -0.0640, -0.0520, -0.0953], + [ 0.0210, -0.0299, -0.0607, ..., -0.0975, 0.0205, -0.0207]], + device='cuda:0'), grad: tensor([[ 3.9148e-04, 2.3785e-03, 1.0628e-04, ..., 1.3381e-05, + 3.2401e-04, 1.4412e-04], + [ 2.3675e-04, 2.7165e-05, -2.5816e-06, ..., 4.0345e-06, + 1.6177e-04, -4.0770e-04], + [-2.8944e-04, 2.5845e-04, 1.2362e-04, ..., 1.8075e-05, + 1.6165e-04, 7.8344e-04], + ..., + [ 4.3535e-04, 1.2934e-04, 5.9931e-07, ..., 8.8692e-05, + 4.7159e-04, 2.0199e-03], + [ 2.0802e-04, 8.3208e-05, 1.0604e-04, ..., 2.0432e-04, + 1.6403e-04, -1.3237e-03], + [ 3.4857e-04, -2.7791e-05, 2.2501e-06, ..., 3.2258e-04, + 1.6677e-04, -1.6441e-03]], device='cuda:0') +Epoch 324, bias, value: tensor([ 0.0082, 0.0031, -0.0015, -0.0193, 0.0092, -0.0108, -0.0156, -0.0192, + 0.0041, 0.0020], device='cuda:0'), grad: tensor([ 0.0227, 0.0036, -0.0176, 0.0070, -0.0143, 0.0017, 0.0111, 0.0303, + -0.0180, -0.0265], device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 216.53, cls_loss 0.4802 cls_loss_mapping 0.0039 cls_loss_causal 0.4575 re_mapping 0.0060 re_causal 0.0163 /// teacc 98.76 lr 0.00010000 +Epoch 325, weight, value: tensor([[ 0.0387, 0.1131, -0.1697, ..., -0.0972, 0.0713, 0.0037], + [-0.0955, -0.1043, 0.1048, ..., -0.0441, -0.0979, 0.0519], + [-0.0908, -0.0628, -0.1146, ..., -0.1297, -0.0282, 0.0600], + ..., + [-0.0142, -0.1252, -0.1262, ..., -0.1077, -0.0402, 0.0698], + [-0.1214, -0.1080, -0.1106, ..., -0.0648, -0.0518, -0.0948], + [ 0.0215, -0.0287, -0.0600, ..., -0.0978, 0.0209, -0.0205]], + device='cuda:0'), grad: tensor([[-1.0204e-03, -6.1655e-04, 5.6475e-05, ..., 3.4046e-04, + -5.5027e-04, 2.1875e-04], + [ 3.0780e-04, -7.5483e-04, -2.5387e-03, ..., -4.2763e-03, + 1.3053e-04, 1.9836e-04], + [ 5.1498e-04, 1.7762e-04, 4.4435e-05, ..., 2.7037e-04, + 1.3328e-04, 8.2550e-03], + ..., + [ 2.7609e-04, 9.9361e-05, 3.6210e-05, ..., 2.4462e-04, + 1.1313e-04, -7.7171e-03], + [ 5.3644e-04, 5.3167e-04, 2.5768e-03, ..., 6.5279e-04, + 1.9395e-04, 7.2241e-04], + [-1.6940e-04, 2.8419e-04, 1.0186e-04, ..., 4.7326e-04, + 2.6560e-04, -3.0303e-04]], device='cuda:0') +Epoch 325, bias, value: tensor([ 0.0085, 0.0026, -0.0017, -0.0180, 0.0080, -0.0110, -0.0142, -0.0189, + 0.0036, 0.0014], device='cuda:0'), grad: tensor([-0.0184, -0.0106, 0.0327, 0.0072, -0.0465, 0.0076, -0.0109, 0.0035, + 0.0237, 0.0117], device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 216.25, cls_loss 0.4560 cls_loss_mapping 0.0031 cls_loss_causal 0.4389 re_mapping 0.0060 re_causal 0.0166 /// teacc 98.68 lr 0.00010000 +Epoch 326, weight, value: tensor([[ 0.0391, 0.1135, -0.1687, ..., -0.0966, 0.0713, 0.0029], + [-0.0957, -0.1033, 0.1058, ..., -0.0422, -0.0994, 0.0522], + [-0.0915, -0.0636, -0.1149, ..., -0.1309, -0.0281, 0.0606], + ..., + [-0.0158, -0.1259, -0.1268, ..., -0.1084, -0.0410, 0.0694], + [-0.1218, -0.1069, -0.1112, ..., -0.0644, -0.0514, -0.0948], + [ 0.0208, -0.0294, -0.0599, ..., -0.0989, 0.0202, -0.0208]], + device='cuda:0'), grad: tensor([[-3.9506e-04, 1.5235e-04, 2.5892e-04, ..., 5.0163e-04, + 1.9595e-05, -3.0136e-03], + [ 7.1955e-04, 1.0617e-05, 8.0347e-05, ..., 5.0592e-04, + 8.3804e-05, 3.3512e-03], + [ 4.7207e-04, 3.3200e-05, 6.2108e-05, ..., 3.1304e-04, + 6.3777e-05, 1.4763e-03], + ..., + [ 3.3951e-04, 5.0478e-06, -8.3089e-05, ..., -1.3328e-04, + 7.8201e-05, -1.3702e-02], + [ 8.7976e-04, 5.0688e-04, 4.8256e-04, ..., 1.5106e-03, + 6.6161e-05, 1.4191e-03], + [ 3.2997e-04, 4.4763e-05, 2.2262e-05, ..., 1.2779e-04, + 8.5831e-05, 7.6218e-03]], device='cuda:0') +Epoch 326, bias, value: tensor([ 0.0082, 0.0029, -0.0021, -0.0172, 0.0084, -0.0098, -0.0147, -0.0191, + 0.0032, 0.0004], device='cuda:0'), grad: tensor([-0.0154, 0.0378, -0.0005, -0.0100, -0.0189, 0.0289, -0.0424, -0.0311, + 0.0263, 0.0253], device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 216.54, cls_loss 0.4745 cls_loss_mapping 0.0032 cls_loss_causal 0.4500 re_mapping 0.0056 re_causal 0.0148 /// teacc 98.85 lr 0.00010000 +Epoch 327, weight, value: tensor([[ 0.0385, 0.1131, -0.1685, ..., -0.0969, 0.0703, 0.0031], + [-0.0954, -0.1036, 0.1056, ..., -0.0432, -0.0987, 0.0511], + [-0.0889, -0.0626, -0.1149, ..., -0.1306, -0.0277, 0.0615], + ..., + [-0.0167, -0.1271, -0.1277, ..., -0.1083, -0.0406, 0.0691], + [-0.1232, -0.1070, -0.1123, ..., -0.0642, -0.0521, -0.0960], + [ 0.0224, -0.0274, -0.0585, ..., -0.0988, 0.0212, -0.0215]], + device='cuda:0'), grad: tensor([[-4.5624e-03, 4.4060e-04, -3.9577e-04, ..., 1.3983e-04, + -9.5069e-05, 3.1471e-03], + [ 3.1400e-04, -1.0386e-05, 8.7142e-05, ..., 1.0446e-05, + 8.6240e-07, 3.2845e-03], + [-2.0042e-05, 5.6744e-05, 3.4571e-05, ..., 4.3440e-04, + 5.0694e-05, -2.0924e-03], + ..., + [-7.6115e-05, 8.5688e-04, 4.0233e-05, ..., 7.4208e-06, + 7.1339e-07, 9.9335e-03], + [ 3.0632e-03, 1.5521e-04, 4.8548e-05, ..., 4.0084e-05, + 6.1505e-06, -4.4060e-03], + [ 1.6088e-03, -1.4534e-03, 3.2902e-05, ..., 2.3097e-05, + 3.2987e-06, -1.2138e-02]], device='cuda:0') +Epoch 327, bias, value: tensor([ 0.0082, 0.0026, -0.0017, -0.0176, 0.0078, -0.0090, -0.0146, -0.0193, + 0.0031, 0.0009], device='cuda:0'), grad: tensor([-0.0620, 0.0385, -0.0103, -0.0082, -0.0137, 0.0158, 0.0165, 0.0072, + 0.0221, -0.0061], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 326---------------------------------------------------- +epoch 326, time 217.26, cls_loss 0.4954 cls_loss_mapping 0.0032 cls_loss_causal 0.4688 re_mapping 0.0060 re_causal 0.0166 /// teacc 99.00 lr 0.00010000 +Epoch 328, weight, value: tensor([[ 0.0387, 0.1128, -0.1689, ..., -0.0967, 0.0705, 0.0025], + [-0.0956, -0.1037, 0.1060, ..., -0.0436, -0.0984, 0.0510], + [-0.0912, -0.0631, -0.1157, ..., -0.1307, -0.0288, 0.0614], + ..., + [-0.0153, -0.1277, -0.1276, ..., -0.1086, -0.0409, 0.0683], + [-0.1245, -0.1067, -0.1119, ..., -0.0645, -0.0517, -0.0956], + [ 0.0206, -0.0274, -0.0596, ..., -0.0993, 0.0216, -0.0202]], + device='cuda:0'), grad: tensor([[ 9.1457e-04, 3.5316e-05, 5.6922e-05, ..., 1.0006e-05, + 1.2612e-04, 5.6982e-05], + [ 1.8244e-03, 4.5784e-06, 5.6438e-06, ..., 7.4366e-07, + 2.9182e-04, 1.0139e-04], + [ 9.4080e-04, -2.2640e-03, -1.3018e-03, ..., -1.1069e-04, + 1.2970e-04, -3.6955e-05], + ..., + [ 1.8358e-03, 6.8173e-06, 4.5784e-06, ..., 3.8370e-07, + 3.4928e-04, -2.5392e-04], + [ 9.7322e-04, 8.0884e-05, 2.2924e-04, ..., 3.8475e-05, + 2.8825e-04, 5.0873e-05], + [-1.0330e-02, 5.8971e-06, 5.3868e-06, ..., 2.3004e-07, + -2.8095e-03, 4.4727e-04]], device='cuda:0') +Epoch 328, bias, value: tensor([ 0.0074, 0.0023, -0.0030, -0.0169, 0.0084, -0.0096, -0.0140, -0.0177, + 0.0018, 0.0014], device='cuda:0'), grad: tensor([ 0.0144, 0.0188, 0.0070, 0.0194, -0.0480, -0.0174, 0.0151, 0.0190, + 0.0131, -0.0414], device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 216.67, cls_loss 0.4826 cls_loss_mapping 0.0028 cls_loss_causal 0.4539 re_mapping 0.0060 re_causal 0.0170 /// teacc 98.91 lr 0.00010000 +Epoch 329, weight, value: tensor([[ 0.0386, 0.1121, -0.1693, ..., -0.0970, 0.0709, 0.0039], + [-0.0957, -0.1041, 0.1062, ..., -0.0432, -0.0990, 0.0509], + [-0.0911, -0.0638, -0.1167, ..., -0.1315, -0.0294, 0.0609], + ..., + [-0.0158, -0.1280, -0.1276, ..., -0.1092, -0.0417, 0.0689], + [-0.1242, -0.1069, -0.1114, ..., -0.0647, -0.0523, -0.0952], + [ 0.0207, -0.0269, -0.0597, ..., -0.0998, 0.0222, -0.0214]], + device='cuda:0'), grad: tensor([[-1.1368e-03, -3.4118e-04, 1.1645e-05, ..., 3.3760e-04, + -1.3240e-05, 1.1647e-04], + [ 6.7997e-04, 2.8443e-04, -3.9756e-05, ..., 6.0767e-05, + 1.2141e-04, -3.7742e-04], + [ 6.7949e-04, 4.2796e-04, 1.4700e-05, ..., 3.4642e-04, + 1.5771e-04, 8.4460e-05], + ..., + [ 5.5790e-04, 1.6344e-04, 1.4596e-05, ..., 7.2122e-05, + 8.0347e-05, -1.1069e-04], + [ 7.0953e-04, 7.6246e-04, 5.2881e-04, ..., 8.2064e-04, + 1.4436e-04, 1.5926e-04], + [ 1.3494e-03, 4.0364e-04, -7.6771e-05, ..., 2.9397e-04, + 1.6785e-04, -4.7415e-05]], device='cuda:0') +Epoch 329, bias, value: tensor([ 0.0078, 0.0022, -0.0029, -0.0168, 0.0089, -0.0098, -0.0150, -0.0175, + 0.0024, 0.0009], device='cuda:0'), grad: tensor([-0.0177, 0.0101, 0.0117, 0.0150, -0.0504, 0.0154, 0.0087, 0.0133, + -0.0190, 0.0129], device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 216.48, cls_loss 0.4700 cls_loss_mapping 0.0027 cls_loss_causal 0.4528 re_mapping 0.0060 re_causal 0.0157 /// teacc 98.85 lr 0.00010000 +Epoch 330, weight, value: tensor([[ 0.0375, 0.1119, -0.1693, ..., -0.0966, 0.0711, 0.0021], + [-0.0955, -0.1047, 0.1067, ..., -0.0445, -0.0986, 0.0514], + [-0.0903, -0.0643, -0.1166, ..., -0.1301, -0.0299, 0.0608], + ..., + [-0.0166, -0.1265, -0.1274, ..., -0.1096, -0.0413, 0.0690], + [-0.1250, -0.1064, -0.1109, ..., -0.0656, -0.0530, -0.0955], + [ 0.0224, -0.0279, -0.0610, ..., -0.1000, 0.0217, -0.0204]], + device='cuda:0'), grad: tensor([[ 1.0567e-03, 2.3782e-04, 2.0242e-04, ..., 4.8935e-05, + 4.9973e-04, 7.2956e-04], + [-1.0824e-03, -4.4403e-03, -3.2215e-03, ..., -7.9966e-04, + -2.8944e-04, -6.0234e-03], + [-1.3485e-03, 1.6260e-03, 1.1797e-03, ..., 2.9516e-04, + -5.6410e-04, 9.5797e-04], + ..., + [-1.0366e-03, 6.0129e-04, 4.2391e-04, ..., 1.0473e-04, + 1.3304e-04, 1.2159e-03], + [ 9.6846e-04, 5.2404e-04, 6.0225e-04, ..., 8.1122e-05, + 4.9448e-04, -4.9353e-04], + [ 1.1463e-03, 3.7408e-04, 3.0041e-04, ..., 7.1168e-05, + 4.6045e-05, 4.5156e-04]], device='cuda:0') +Epoch 330, bias, value: tensor([ 0.0078, 0.0026, -0.0029, -0.0164, 0.0085, -0.0109, -0.0143, -0.0174, + 0.0018, 0.0013], device='cuda:0'), grad: tensor([ 0.0166, -0.0311, -0.0046, 0.0314, -0.0139, 0.0088, -0.0154, 0.0066, + -0.0131, 0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 216.30, cls_loss 0.4690 cls_loss_mapping 0.0037 cls_loss_causal 0.4476 re_mapping 0.0063 re_causal 0.0166 /// teacc 98.78 lr 0.00010000 +Epoch 331, weight, value: tensor([[ 0.0364, 0.1125, -0.1700, ..., -0.0970, 0.0710, 0.0016], + [-0.0945, -0.1046, 0.1065, ..., -0.0451, -0.0989, 0.0516], + [-0.0886, -0.0653, -0.1164, ..., -0.1303, -0.0303, 0.0616], + ..., + [-0.0166, -0.1271, -0.1281, ..., -0.1102, -0.0404, 0.0686], + [-0.1256, -0.1075, -0.1112, ..., -0.0652, -0.0530, -0.0961], + [ 0.0224, -0.0280, -0.0611, ..., -0.0989, 0.0226, -0.0201]], + device='cuda:0'), grad: tensor([[-1.5182e-03, 5.1498e-03, 9.4995e-08, ..., -1.0751e-05, + -5.3120e-04, -8.5878e-04], + [ 1.5039e-03, 1.1575e-04, 1.6764e-08, ..., 3.1712e-07, + 4.8661e-04, -3.9244e-04], + [ 8.1062e-04, 1.1820e-04, 2.5313e-06, ..., 4.0047e-06, + 2.7323e-04, 1.3599e-03], + ..., + [-1.0443e-03, 2.0373e-04, 3.6880e-07, ..., 2.0675e-07, + -3.3426e-04, -1.2360e-03], + [ 6.7377e-04, 7.4244e-04, 4.8205e-06, ..., 5.6028e-06, + 2.4772e-04, 5.6934e-04], + [-4.9095e-03, -5.5408e-04, 5.4250e-07, ..., 2.1979e-06, + -3.5515e-03, 8.1873e-04]], device='cuda:0') +Epoch 331, bias, value: tensor([ 0.0078, 0.0039, -0.0034, -0.0156, 0.0086, -0.0122, -0.0142, -0.0177, + 0.0013, 0.0016], device='cuda:0'), grad: tensor([-0.0092, -0.0031, 0.0147, 0.0147, 0.0335, -0.0054, -0.0479, -0.0116, + 0.0124, 0.0018], device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 216.35, cls_loss 0.4734 cls_loss_mapping 0.0037 cls_loss_causal 0.4413 re_mapping 0.0059 re_causal 0.0164 /// teacc 98.88 lr 0.00010000 +Epoch 332, weight, value: tensor([[ 0.0367, 0.1113, -0.1698, ..., -0.0974, 0.0712, -0.0004], + [-0.0959, -0.1044, 0.1069, ..., -0.0452, -0.0994, 0.0522], + [-0.0882, -0.0662, -0.1177, ..., -0.1307, -0.0302, 0.0603], + ..., + [-0.0166, -0.1275, -0.1282, ..., -0.1103, -0.0397, 0.0687], + [-0.1255, -0.1065, -0.1106, ..., -0.0645, -0.0528, -0.0953], + [ 0.0236, -0.0276, -0.0615, ..., -0.0990, 0.0230, -0.0192]], + device='cuda:0'), grad: tensor([[-7.9041e-03, -5.9738e-03, -2.5921e-03, ..., -1.2913e-03, + -2.9163e-03, 9.7156e-05], + [-4.3607e-04, -1.0729e-03, -8.0442e-04, ..., 6.5088e-05, + -8.7833e-04, -1.8234e-03], + [ 5.0783e-04, 7.4577e-04, 5.3883e-04, ..., 4.9543e-04, + 1.2052e-04, -6.3956e-05], + ..., + [-8.5449e-04, -2.5463e-04, 1.8334e-04, ..., -3.1376e-03, + 7.1108e-05, -6.3705e-04], + [ 2.3804e-03, 1.2712e-03, 6.3610e-04, ..., 7.1955e-04, + 9.1219e-04, 2.3651e-04], + [ 1.5907e-03, 2.0790e-03, 1.2722e-03, ..., 7.9679e-04, + 6.8378e-04, 1.4000e-03]], device='cuda:0') +Epoch 332, bias, value: tensor([ 0.0072, 0.0042, -0.0032, -0.0158, 0.0076, -0.0128, -0.0131, -0.0180, + 0.0023, 0.0014], device='cuda:0'), grad: tensor([-0.0397, 0.0112, 0.0165, 0.0187, -0.0160, 0.0177, 0.0037, -0.0227, + 0.0213, -0.0107], device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 216.67, cls_loss 0.4901 cls_loss_mapping 0.0034 cls_loss_causal 0.4647 re_mapping 0.0057 re_causal 0.0158 /// teacc 98.85 lr 0.00010000 +Epoch 333, weight, value: tensor([[ 0.0382, 0.1118, -0.1702, ..., -0.0969, 0.0725, 0.0015], + [-0.0969, -0.1037, 0.1060, ..., -0.0460, -0.0997, 0.0530], + [-0.0873, -0.0667, -0.1179, ..., -0.1286, -0.0291, 0.0604], + ..., + [-0.0160, -0.1282, -0.1284, ..., -0.1107, -0.0404, 0.0674], + [-0.1261, -0.1069, -0.1096, ..., -0.0650, -0.0529, -0.0951], + [ 0.0221, -0.0285, -0.0621, ..., -0.0998, 0.0216, -0.0187]], + device='cuda:0'), grad: tensor([[-3.6907e-04, 9.7789e-09, 2.3260e-07, ..., -1.5497e-03, + 4.0627e-04, 1.1083e-06], + [ 1.5867e-04, 1.7928e-08, -1.4203e-06, ..., 4.2409e-05, + 1.0198e-04, 2.0638e-06], + [ 2.2435e-04, 4.3539e-07, 9.2201e-07, ..., 5.3078e-05, + 6.9499e-05, -2.1115e-05], + ..., + [-1.1444e-03, 4.2096e-07, 4.0838e-07, ..., 8.0645e-05, + -1.3418e-03, 6.5006e-06], + [-1.3514e-03, 3.9078e-06, 8.2627e-06, ..., 2.7585e-04, + -2.2674e-04, 2.4885e-06], + [ 1.7672e-03, 5.3830e-07, 2.3004e-06, ..., 1.6618e-04, + 6.0654e-04, -1.5395e-06]], device='cuda:0') +Epoch 333, bias, value: tensor([ 0.0085, 0.0039, -0.0018, -0.0165, 0.0067, -0.0122, -0.0136, -0.0176, + 0.0011, 0.0015], device='cuda:0'), grad: tensor([-0.0038, -0.0136, -0.0142, 0.0236, -0.0135, 0.0181, 0.0222, -0.0364, + -0.0162, 0.0338], device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 216.47, cls_loss 0.4830 cls_loss_mapping 0.0041 cls_loss_causal 0.4625 re_mapping 0.0053 re_causal 0.0147 /// teacc 98.89 lr 0.00010000 +Epoch 334, weight, value: tensor([[ 0.0381, 0.1127, -0.1718, ..., -0.0977, 0.0722, 0.0015], + [-0.0977, -0.1030, 0.1065, ..., -0.0461, -0.0998, 0.0541], + [-0.0876, -0.0652, -0.1167, ..., -0.1290, -0.0286, 0.0598], + ..., + [-0.0162, -0.1271, -0.1294, ..., -0.1115, -0.0410, 0.0677], + [-0.1246, -0.1087, -0.1103, ..., -0.0649, -0.0534, -0.0950], + [ 0.0224, -0.0276, -0.0604, ..., -0.0999, 0.0217, -0.0192]], + device='cuda:0'), grad: tensor([[-7.0724e-03, 1.1606e-03, 3.4833e-04, ..., 5.5170e-04, + 3.1638e-04, 7.3814e-04], + [ 4.7445e-04, 3.5316e-05, 2.0906e-05, ..., 2.1055e-05, + 1.1466e-05, 5.0402e-04], + [ 3.4380e-04, 1.6618e-04, 1.2672e-04, ..., 1.0920e-04, + 5.6833e-05, 3.1304e-04], + ..., + [ 4.8804e-04, 1.4067e-04, 5.1141e-05, ..., 7.3791e-05, + 4.5747e-05, 3.5262e-04], + [-3.1424e-04, 4.6611e-04, 1.5831e-04, ..., 2.5344e-04, + 1.7011e-04, -1.1072e-03], + [ 1.6766e-03, 1.4324e-03, 4.1318e-04, ..., 7.2193e-04, + 4.6825e-04, 8.2779e-04]], device='cuda:0') +Epoch 334, bias, value: tensor([ 0.0076, 0.0049, -0.0024, -0.0172, 0.0073, -0.0129, -0.0124, -0.0175, + 0.0006, 0.0022], device='cuda:0'), grad: tensor([ 0.0100, 0.0212, 0.0159, -0.0450, -0.0798, 0.0258, 0.0397, 0.0238, + -0.0414, 0.0298], device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 216.74, cls_loss 0.4742 cls_loss_mapping 0.0033 cls_loss_causal 0.4586 re_mapping 0.0056 re_causal 0.0149 /// teacc 98.91 lr 0.00010000 +Epoch 335, weight, value: tensor([[ 0.0388, 0.1131, -0.1722, ..., -0.0976, 0.0726, 0.0018], + [-0.1000, -0.1040, 0.1070, ..., -0.0458, -0.1008, 0.0535], + [-0.0885, -0.0660, -0.1179, ..., -0.1290, -0.0281, 0.0608], + ..., + [-0.0151, -0.1279, -0.1292, ..., -0.1128, -0.0419, 0.0679], + [-0.1233, -0.1088, -0.1095, ..., -0.0655, -0.0537, -0.0955], + [ 0.0217, -0.0278, -0.0603, ..., -0.1008, 0.0214, -0.0204]], + device='cuda:0'), grad: tensor([[ 1.1940e-03, 7.8604e-06, 7.1712e-08, ..., 4.4674e-05, + 3.5667e-04, -7.7486e-04], + [-9.7609e-04, -1.0576e-03, 4.3772e-08, ..., 4.0680e-06, + -2.3956e-03, -4.9896e-03], + [-2.1072e-02, 2.0862e-05, 3.4831e-07, ..., -8.4543e-04, + -1.0424e-03, -2.8687e-03], + ..., + [ 1.9188e-03, 8.1825e-04, 4.2934e-07, ..., 7.5400e-06, + 1.7071e-03, 3.5572e-03], + [ 2.0695e-03, 3.1561e-05, 1.8448e-05, ..., 2.6196e-05, + 3.4761e-04, 1.5078e-03], + [ 1.2569e-03, 6.2168e-05, -4.7125e-06, ..., 3.3796e-05, + 2.3830e-04, 9.8038e-04]], device='cuda:0') +Epoch 335, bias, value: tensor([ 0.0087, 0.0044, -0.0020, -0.0173, 0.0074, -0.0125, -0.0135, -0.0184, + 0.0013, 0.0020], device='cuda:0'), grad: tensor([-0.0123, -0.0770, -0.0456, 0.0165, 0.0144, 0.0093, 0.0251, 0.0381, + 0.0178, 0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 216.33, cls_loss 0.4964 cls_loss_mapping 0.0050 cls_loss_causal 0.4772 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.75 lr 0.00010000 +Epoch 336, weight, value: tensor([[ 0.0377, 0.1123, -0.1745, ..., -0.0986, 0.0711, 0.0014], + [-0.0998, -0.1036, 0.1071, ..., -0.0459, -0.1015, 0.0525], + [-0.0883, -0.0655, -0.1176, ..., -0.1289, -0.0272, 0.0612], + ..., + [-0.0149, -0.1289, -0.1288, ..., -0.1140, -0.0423, 0.0671], + [-0.1233, -0.1085, -0.1087, ..., -0.0645, -0.0536, -0.0940], + [ 0.0223, -0.0261, -0.0599, ..., -0.1003, 0.0224, -0.0209]], + device='cuda:0'), grad: tensor([[ 7.4530e-04, 1.0449e-04, 2.2173e-05, ..., 2.2292e-04, + 6.5982e-05, 7.4720e-04], + [ 9.2936e-04, 1.9813e-04, -1.2197e-05, ..., 4.8339e-05, + 1.1981e-04, 3.4256e-03], + [-3.0899e-03, -1.0180e-04, -1.8811e-04, ..., 5.2065e-05, + -6.2418e-04, 3.9053e-04], + ..., + [-1.5795e-04, -1.4949e-04, 2.6417e-04, ..., -1.0991e-04, + -7.0453e-05, -6.7101e-03], + [-3.3188e-03, 2.6554e-05, 3.4243e-05, ..., -7.9334e-05, + -3.4404e-04, -1.9464e-03], + [-1.1027e-04, -1.2369e-03, -5.7602e-04, ..., 4.1425e-05, + 1.2410e-04, 1.0509e-03]], device='cuda:0') +Epoch 336, bias, value: tensor([ 0.0076, 0.0051, -0.0032, -0.0170, 0.0074, -0.0117, -0.0131, -0.0174, + 0.0008, 0.0016], device='cuda:0'), grad: tensor([ 0.0114, 0.0268, -0.0120, 0.0311, 0.0097, 0.0130, 0.0100, -0.0421, + -0.0494, 0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 216.32, cls_loss 0.4917 cls_loss_mapping 0.0047 cls_loss_causal 0.4706 re_mapping 0.0062 re_causal 0.0165 /// teacc 98.92 lr 0.00010000 +Epoch 337, weight, value: tensor([[ 0.0384, 0.1131, -0.1750, ..., -0.0989, 0.0715, -0.0004], + [-0.0995, -0.1034, 0.1079, ..., -0.0448, -0.1013, 0.0541], + [-0.0890, -0.0671, -0.1179, ..., -0.1277, -0.0272, 0.0611], + ..., + [-0.0127, -0.1272, -0.1291, ..., -0.1140, -0.0425, 0.0673], + [-0.1244, -0.1081, -0.1085, ..., -0.0649, -0.0541, -0.0948], + [ 0.0214, -0.0275, -0.0607, ..., -0.0997, 0.0217, -0.0211]], + device='cuda:0'), grad: tensor([[ 5.8031e-04, 1.1530e-03, 1.6950e-06, ..., 2.1413e-05, + 9.4399e-06, 3.1900e-04], + [ 1.0842e-04, 3.9846e-05, -1.9893e-05, ..., 5.3644e-07, + 6.6496e-07, 5.1308e-04], + [ 4.4250e-04, 8.6427e-05, 4.4331e-06, ..., 3.6538e-05, + 3.5971e-05, 1.1740e-03], + ..., + [ 2.7657e-04, 4.3303e-05, 2.0396e-06, ..., 2.2957e-07, + 2.0415e-05, -4.1466e-03], + [-1.5707e-03, -3.4714e-03, 1.8716e-05, ..., 3.8594e-06, + 3.9972e-06, 3.7289e-04], + [ 1.5287e-03, 6.4421e-04, 1.1185e-06, ..., 3.5036e-06, + -1.2614e-05, 3.5691e-04]], device='cuda:0') +Epoch 337, bias, value: tensor([ 0.0069, 0.0053, -0.0034, -0.0159, 0.0062, -0.0122, -0.0122, -0.0170, + 0.0012, 0.0010], device='cuda:0'), grad: tensor([ 0.0203, 0.0215, 0.0175, -0.0179, 0.0107, -0.0019, -0.0158, -0.0390, + -0.0178, 0.0224], device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 216.46, cls_loss 0.4875 cls_loss_mapping 0.0039 cls_loss_causal 0.4708 re_mapping 0.0059 re_causal 0.0148 /// teacc 98.85 lr 0.00010000 +Epoch 338, weight, value: tensor([[ 0.0380, 0.1134, -0.1756, ..., -0.0983, 0.0713, 0.0003], + [-0.1000, -0.1022, 0.1076, ..., -0.0453, -0.1015, 0.0534], + [-0.0888, -0.0671, -0.1183, ..., -0.1295, -0.0274, 0.0615], + ..., + [-0.0129, -0.1287, -0.1294, ..., -0.1152, -0.0419, 0.0668], + [-0.1231, -0.1080, -0.1084, ..., -0.0639, -0.0539, -0.0960], + [ 0.0206, -0.0277, -0.0617, ..., -0.0997, 0.0220, -0.0200]], + device='cuda:0'), grad: tensor([[ 7.2861e-04, 5.1498e-05, 1.3888e-05, ..., 4.0364e-04, + 4.9844e-06, 5.7602e-04], + [ 1.2398e-03, 3.0518e-05, 2.4140e-05, ..., 6.9761e-04, + 1.0557e-05, 1.2341e-03], + [ 6.8712e-04, 3.6031e-05, 1.5318e-05, ..., 1.4763e-03, + 4.8466e-06, 2.8629e-03], + ..., + [ 1.4818e-04, 6.7875e-06, 1.1988e-05, ..., 2.2926e-03, + 3.4988e-05, 4.1618e-03], + [ 3.9172e-04, -2.3422e-03, -1.0929e-03, ..., 6.0558e-04, + 1.5482e-05, 2.7347e-04], + [-1.9016e-03, 3.3289e-05, -6.6757e-05, ..., -3.4370e-03, + -1.1796e-04, -8.6670e-03]], device='cuda:0') +Epoch 338, bias, value: tensor([ 0.0073, 0.0050, -0.0041, -0.0155, 0.0059, -0.0116, -0.0124, -0.0167, + 0.0009, 0.0012], device='cuda:0'), grad: tensor([ 1.8555e-02, -2.3518e-03, 2.7847e-02, 2.9716e-03, 1.5556e-02, + -2.2049e-02, 7.1883e-05, 1.2611e-02, -1.8478e-02, -3.4698e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 216.63, cls_loss 0.4849 cls_loss_mapping 0.0041 cls_loss_causal 0.4539 re_mapping 0.0063 re_causal 0.0161 /// teacc 98.81 lr 0.00010000 +Epoch 339, weight, value: tensor([[ 0.0384, 0.1159, -0.1741, ..., -0.0978, 0.0713, 0.0022], + [-0.1005, -0.1022, 0.1072, ..., -0.0447, -0.1023, 0.0526], + [-0.0890, -0.0664, -0.1185, ..., -0.1294, -0.0277, 0.0617], + ..., + [-0.0129, -0.1284, -0.1297, ..., -0.1147, -0.0417, 0.0663], + [-0.1243, -0.1095, -0.1096, ..., -0.0649, -0.0549, -0.0960], + [ 0.0219, -0.0278, -0.0629, ..., -0.0993, 0.0220, -0.0197]], + device='cuda:0'), grad: tensor([[ 2.4242e-03, -5.9366e-05, 7.8058e-04, ..., 4.9591e-04, + -2.5719e-05, -4.6229e-04], + [-9.0647e-04, 4.1455e-05, 8.0109e-04, ..., 4.3154e-04, + -1.7166e-05, 3.4404e-04], + [-1.8597e-04, 2.2024e-05, 1.1438e-04, ..., 1.4353e-04, + -1.7166e-05, -1.0271e-03], + ..., + [-2.2869e-03, 1.2092e-05, 2.2620e-05, ..., 1.2301e-05, + -1.4174e-04, -1.0862e-03], + [ 2.5463e-03, 3.1173e-05, 7.3624e-04, ..., 1.0943e-04, + 5.3227e-05, 7.8058e-04], + [-3.9520e-03, 3.6061e-05, 1.2541e-04, ..., 7.1704e-05, + 4.5568e-05, 8.1921e-04]], device='cuda:0') +Epoch 339, bias, value: tensor([ 8.0922e-03, 5.2635e-03, -3.6541e-03, -1.6177e-02, 5.7349e-03, + -1.1812e-02, -1.2846e-02, -1.8082e-02, 8.0180e-05, 3.3521e-03], + device='cuda:0'), grad: tensor([-0.0147, -0.0471, -0.0083, -0.0311, 0.0443, -0.0061, 0.0349, -0.0298, + 0.0325, 0.0253], device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 216.46, cls_loss 0.4720 cls_loss_mapping 0.0030 cls_loss_causal 0.4482 re_mapping 0.0061 re_causal 0.0158 /// teacc 98.80 lr 0.00010000 +Epoch 340, weight, value: tensor([[ 0.0382, 0.1170, -0.1742, ..., -0.0989, 0.0716, 0.0025], + [-0.1000, -0.1020, 0.1074, ..., -0.0457, -0.1026, 0.0515], + [-0.0889, -0.0676, -0.1188, ..., -0.1291, -0.0275, 0.0621], + ..., + [-0.0130, -0.1300, -0.1307, ..., -0.1147, -0.0419, 0.0668], + [-0.1240, -0.1105, -0.1090, ..., -0.0633, -0.0547, -0.0956], + [ 0.0199, -0.0287, -0.0621, ..., -0.1000, 0.0206, -0.0204]], + device='cuda:0'), grad: tensor([[ 9.2983e-04, -2.1207e-04, 1.7229e-08, ..., -8.0094e-06, + 1.3542e-04, 3.2163e-04], + [ 1.7157e-03, 1.0848e-04, 0.0000e+00, ..., 4.6074e-05, + 5.2166e-04, 1.5640e-03], + [ 1.6022e-03, 9.2089e-05, 4.1910e-09, ..., 8.6352e-06, + 2.4676e-04, 6.8617e-04], + ..., + [ 1.1024e-03, 7.4029e-05, 0.0000e+00, ..., 1.5028e-05, + 3.0828e-04, -7.4720e-04], + [-2.4586e-03, -4.4584e-04, 2.6403e-07, ..., 3.9577e-05, + -1.3123e-03, 1.0204e-04], + [ 8.6451e-04, 8.3148e-05, 9.3132e-10, ..., 1.9893e-05, + 6.8855e-04, 7.0381e-04]], device='cuda:0') +Epoch 340, bias, value: tensor([ 0.0077, 0.0056, -0.0033, -0.0174, 0.0064, -0.0121, -0.0123, -0.0186, + 0.0004, 0.0035], device='cuda:0'), grad: tensor([ 0.0207, 0.0239, 0.0283, -0.0464, -0.0615, 0.0114, 0.0226, 0.0205, + -0.0095, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 216.40, cls_loss 0.4965 cls_loss_mapping 0.0045 cls_loss_causal 0.4721 re_mapping 0.0063 re_causal 0.0157 /// teacc 98.82 lr 0.00010000 +Epoch 341, weight, value: tensor([[ 0.0373, 0.1170, -0.1758, ..., -0.0991, 0.0697, 0.0030], + [-0.1005, -0.1003, 0.1082, ..., -0.0458, -0.1017, 0.0511], + [-0.0890, -0.0677, -0.1194, ..., -0.1290, -0.0269, 0.0624], + ..., + [-0.0127, -0.1285, -0.1312, ..., -0.1151, -0.0414, 0.0671], + [-0.1238, -0.1104, -0.1097, ..., -0.0647, -0.0542, -0.0964], + [ 0.0202, -0.0284, -0.0629, ..., -0.0981, 0.0227, -0.0205]], + device='cuda:0'), grad: tensor([[ 5.6648e-04, 2.2089e-04, 5.7578e-05, ..., 8.4890e-07, + 2.1422e-04, 2.5439e-04], + [ 1.0424e-03, 2.1744e-03, 2.1095e-03, ..., 7.4530e-04, + -3.0375e-04, 1.1425e-03], + [ 1.2083e-03, 1.0805e-03, 7.8058e-04, ..., 5.1439e-05, + 2.4557e-04, 6.3181e-04], + ..., + [-1.5364e-03, 1.1045e-04, 1.4625e-05, ..., 6.3097e-07, + -9.4296e-07, -1.5650e-03], + [ 5.1165e-04, 4.1866e-04, 2.6011e-04, ..., 3.5839e-03, + 1.8978e-04, 1.8656e-04], + [ 1.6842e-03, 3.7336e-04, 1.6391e-04, ..., 2.2631e-07, + 5.7411e-04, -8.2874e-04]], device='cuda:0') +Epoch 341, bias, value: tensor([ 0.0079, 0.0056, -0.0025, -0.0160, 0.0060, -0.0128, -0.0127, -0.0191, + 0.0003, 0.0031], device='cuda:0'), grad: tensor([-0.0191, -0.0029, 0.0208, -0.0002, 0.0190, 0.0033, -0.0333, -0.0175, + 0.0294, 0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 216.41, cls_loss 0.4892 cls_loss_mapping 0.0030 cls_loss_causal 0.4639 re_mapping 0.0058 re_causal 0.0159 /// teacc 98.96 lr 0.00010000 +Epoch 342, weight, value: tensor([[ 0.0379, 0.1166, -0.1766, ..., -0.0974, 0.0698, 0.0038], + [-0.0998, -0.1006, 0.1088, ..., -0.0456, -0.1009, 0.0511], + [-0.0905, -0.0682, -0.1200, ..., -0.1298, -0.0273, 0.0625], + ..., + [-0.0129, -0.1287, -0.1309, ..., -0.1149, -0.0418, 0.0668], + [-0.1242, -0.1115, -0.1092, ..., -0.0652, -0.0564, -0.0967], + [ 0.0215, -0.0274, -0.0639, ..., -0.0960, 0.0242, -0.0200]], + device='cuda:0'), grad: tensor([[ 2.3854e-04, 1.3737e-06, 1.6391e-07, ..., 3.2596e-08, + 5.6893e-05, 9.9652e-07], + [ 2.5690e-05, 3.3051e-05, -3.2205e-06, ..., 1.0245e-08, + 6.4182e-04, 2.4390e-04], + [ 3.8338e-04, 1.4770e-04, 6.8219e-07, ..., 6.4261e-08, + 1.6403e-04, 4.0588e-03], + ..., + [ 1.0371e-04, 2.8133e-05, -1.0473e-04, ..., 2.0862e-07, + 4.0621e-05, -4.4746e-03], + [ 6.0225e-04, 5.7332e-06, 3.0287e-06, ..., 4.1770e-07, + 1.3196e-04, 1.3709e-04], + [-3.0060e-03, -6.4634e-06, 5.2541e-05, ..., 6.0536e-09, + -6.0415e-04, 8.2076e-05]], device='cuda:0') +Epoch 342, bias, value: tensor([ 0.0073, 0.0063, -0.0028, -0.0158, 0.0065, -0.0118, -0.0134, -0.0192, + -0.0005, 0.0032], device='cuda:0'), grad: tensor([ 0.0126, 0.0358, 0.0096, 0.0191, -0.0637, 0.0143, 0.0147, -0.0236, + -0.0148, -0.0040], device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 216.75, cls_loss 0.4662 cls_loss_mapping 0.0054 cls_loss_causal 0.4397 re_mapping 0.0060 re_causal 0.0158 /// teacc 98.88 lr 0.00010000 +Epoch 343, weight, value: tensor([[ 0.0376, 0.1167, -0.1762, ..., -0.0963, 0.0700, 0.0036], + [-0.1000, -0.1014, 0.1097, ..., -0.0447, -0.1013, 0.0501], + [-0.0907, -0.0682, -0.1196, ..., -0.1296, -0.0274, 0.0619], + ..., + [-0.0123, -0.1280, -0.1322, ..., -0.1148, -0.0417, 0.0685], + [-0.1245, -0.1113, -0.1095, ..., -0.0653, -0.0572, -0.0974], + [ 0.0215, -0.0273, -0.0646, ..., -0.0968, 0.0239, -0.0204]], + device='cuda:0'), grad: tensor([[ 3.0327e-04, -1.1183e-05, 1.5080e-05, ..., 3.4332e-05, + -3.8218e-04, 1.1754e-04], + [ 3.5334e-04, 1.1856e-06, 9.7379e-06, ..., 2.4855e-05, + 2.9895e-06, 1.3614e-04], + [ 3.2187e-04, 8.1882e-06, 1.5274e-05, ..., 2.1115e-05, + 1.1645e-05, 1.2741e-03], + ..., + [ 5.2547e-04, 1.0006e-05, 9.4697e-06, ..., 1.1526e-05, + 8.2329e-06, -9.9373e-04], + [ 5.1689e-04, 1.3031e-05, 4.2528e-05, ..., 4.5091e-05, + 7.3053e-06, 2.0945e-04], + [ 9.1124e-04, -1.2890e-05, -1.7121e-05, ..., 2.3872e-05, + 8.0615e-06, 3.5214e-04]], device='cuda:0') +Epoch 343, bias, value: tensor([ 0.0075, 0.0062, -0.0024, -0.0154, 0.0063, -0.0131, -0.0127, -0.0181, + -0.0008, 0.0024], device='cuda:0'), grad: tensor([-0.0268, 0.0053, 0.0055, 0.0046, 0.0068, -0.0265, 0.0087, 0.0047, + 0.0080, 0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 216.72, cls_loss 0.5028 cls_loss_mapping 0.0038 cls_loss_causal 0.4771 re_mapping 0.0059 re_causal 0.0154 /// teacc 98.88 lr 0.00010000 +Epoch 344, weight, value: tensor([[ 0.0366, 0.1175, -0.1750, ..., -0.0969, 0.0706, 0.0027], + [-0.0995, -0.1019, 0.1109, ..., -0.0438, -0.1010, 0.0500], + [-0.0910, -0.0671, -0.1207, ..., -0.1287, -0.0268, 0.0616], + ..., + [-0.0128, -0.1279, -0.1334, ..., -0.1151, -0.0419, 0.0686], + [-0.1249, -0.1125, -0.1100, ..., -0.0654, -0.0573, -0.0963], + [ 0.0219, -0.0274, -0.0635, ..., -0.0984, 0.0229, -0.0208]], + device='cuda:0'), grad: tensor([[ 9.8050e-05, -1.0383e-04, 1.0574e-04, ..., -1.2153e-04, + -1.1548e-06, 3.2425e-05], + [ 3.9434e-04, 1.1164e-04, 1.7679e-04, ..., 5.5432e-05, + 1.8626e-08, 1.6177e-04], + [ 4.2295e-04, 7.1883e-05, -3.1042e-04, ..., 2.9016e-04, + 9.6206e-07, 1.2398e-04], + ..., + [-2.5444e-03, -8.9312e-04, 8.7619e-05, ..., 2.3767e-05, + -1.0710e-06, -1.3914e-03], + [ 3.0422e-04, 7.3254e-05, 2.3234e-04, ..., 1.9538e-04, + 2.9728e-06, 7.7188e-05], + [ 1.9627e-03, 1.5867e-04, 8.8155e-05, ..., 4.1425e-05, + 2.5660e-05, 2.4986e-04]], device='cuda:0') +Epoch 344, bias, value: tensor([ 0.0079, 0.0059, -0.0032, -0.0156, 0.0062, -0.0117, -0.0132, -0.0188, + -0.0005, 0.0031], device='cuda:0'), grad: tensor([ 0.0098, 0.0003, -0.0144, 0.0363, 0.0120, -0.0149, -0.0552, -0.0097, + 0.0180, 0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 216.49, cls_loss 0.4699 cls_loss_mapping 0.0029 cls_loss_causal 0.4491 re_mapping 0.0059 re_causal 0.0144 /// teacc 98.90 lr 0.00010000 +Epoch 345, weight, value: tensor([[ 0.0372, 0.1185, -0.1752, ..., -0.0967, 0.0714, 0.0020], + [-0.0993, -0.1035, 0.1118, ..., -0.0435, -0.1014, 0.0511], + [-0.0919, -0.0668, -0.1209, ..., -0.1290, -0.0271, 0.0616], + ..., + [-0.0125, -0.1284, -0.1344, ..., -0.1145, -0.0420, 0.0690], + [-0.1258, -0.1134, -0.1100, ..., -0.0662, -0.0577, -0.0981], + [ 0.0223, -0.0265, -0.0636, ..., -0.0987, 0.0231, -0.0210]], + device='cuda:0'), grad: tensor([[ 1.4601e-03, -1.3232e-04, 8.9586e-05, ..., 5.7220e-06, + 1.3411e-04, 6.2323e-04], + [ 1.4744e-03, 5.4315e-06, 3.0413e-05, ..., 6.7987e-08, + 2.2578e-04, 9.0361e-04], + [ 1.5202e-03, 3.1400e-04, 5.7793e-04, ..., 5.5656e-06, + 3.1400e-04, 1.0481e-03], + ..., + [-4.7989e-03, 1.0662e-05, -1.9848e-04, ..., 2.8545e-07, + -1.5297e-03, -6.4812e-03], + [ 9.9277e-04, -9.7609e-04, -1.7424e-03, ..., -3.6567e-05, + 1.2708e-04, 5.0831e-04], + [ 9.5320e-04, 1.7822e-04, 1.8549e-04, ..., 4.1164e-07, + 1.1730e-04, 5.6887e-04]], device='cuda:0') +Epoch 345, bias, value: tensor([ 0.0076, 0.0058, -0.0028, -0.0166, 0.0069, -0.0124, -0.0122, -0.0189, + -0.0013, 0.0039], device='cuda:0'), grad: tensor([-0.0176, 0.0196, 0.0222, -0.0381, 0.0150, -0.0480, 0.0212, -0.0224, + 0.0315, 0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 216.64, cls_loss 0.4596 cls_loss_mapping 0.0030 cls_loss_causal 0.4373 re_mapping 0.0060 re_causal 0.0158 /// teacc 98.73 lr 0.00010000 +Epoch 346, weight, value: tensor([[ 0.0369, 0.1190, -0.1760, ..., -0.0971, 0.0709, 0.0018], + [-0.0989, -0.1021, 0.1139, ..., -0.0444, -0.1011, 0.0502], + [-0.0933, -0.0662, -0.1211, ..., -0.1292, -0.0280, 0.0619], + ..., + [-0.0117, -0.1282, -0.1339, ..., -0.1148, -0.0419, 0.0689], + [-0.1266, -0.1130, -0.1108, ..., -0.0662, -0.0585, -0.0974], + [ 0.0226, -0.0271, -0.0635, ..., -0.0991, 0.0223, -0.0206]], + device='cuda:0'), grad: tensor([[ 3.4428e-03, 1.0192e-04, 3.3522e-04, ..., 5.4576e-06, + 5.2929e-04, 1.2410e-04], + [ 9.1219e-04, 4.0770e-04, 1.1110e-03, ..., 2.3276e-05, + 2.8348e-04, 4.3750e-04], + [-1.9093e-03, -1.2484e-03, -1.5707e-03, ..., 1.2241e-05, + -4.9019e-04, -1.6956e-03], + ..., + [ 6.1417e-03, 3.4881e-04, 9.3842e-04, ..., 9.6112e-06, + 2.3794e-04, 7.2956e-04], + [-1.4758e-04, 2.5511e-04, 8.8596e-04, ..., 8.6725e-06, + 2.0301e-04, 3.9601e-04], + [-5.9052e-03, 4.0889e-04, -1.5402e-03, ..., 1.3679e-05, + 2.8372e-04, -3.6669e-04]], device='cuda:0') +Epoch 346, bias, value: tensor([ 0.0074, 0.0063, -0.0020, -0.0168, 0.0063, -0.0121, -0.0133, -0.0193, + -0.0013, 0.0047], device='cuda:0'), grad: tensor([ 0.0207, 0.0163, -0.0148, -0.0059, -0.0084, -0.0159, 0.0270, 0.0300, + -0.0190, -0.0301], device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 216.78, cls_loss 0.5084 cls_loss_mapping 0.0036 cls_loss_causal 0.4790 re_mapping 0.0059 re_causal 0.0157 /// teacc 98.89 lr 0.00010000 +Epoch 347, weight, value: tensor([[ 0.0386, 0.1174, -0.1765, ..., -0.0978, 0.0712, 0.0021], + [-0.0979, -0.1022, 0.1135, ..., -0.0457, -0.1017, 0.0502], + [-0.0937, -0.0662, -0.1227, ..., -0.1311, -0.0287, 0.0612], + ..., + [-0.0123, -0.1275, -0.1313, ..., -0.1115, -0.0411, 0.0692], + [-0.1269, -0.1123, -0.1111, ..., -0.0647, -0.0594, -0.0966], + [ 0.0221, -0.0278, -0.0646, ..., -0.1008, 0.0226, -0.0217]], + device='cuda:0'), grad: tensor([[-4.4556e-03, 3.5251e-07, -6.4898e-04, ..., -1.0860e-04, + -9.3365e-04, -2.4776e-03], + [ 9.3412e-04, 2.3469e-07, 2.9397e-04, ..., 3.6299e-05, + 9.0718e-05, 6.7186e-04], + [ 6.7329e-04, 7.8836e-07, -2.0397e-04, ..., -1.2048e-05, + 7.3195e-05, 3.2806e-04], + ..., + [-1.0357e-03, 8.0187e-07, 1.7118e-04, ..., 2.1830e-05, + 8.2552e-05, -1.6470e-03], + [ 1.5802e-03, 7.4040e-07, 1.2529e-04, ..., 1.7107e-05, + 7.5340e-05, 4.5180e-04], + [ 2.0180e-03, -3.9674e-06, -2.2733e-04, ..., 7.2829e-06, + 8.4937e-05, 8.5402e-04]], device='cuda:0') +Epoch 347, bias, value: tensor([ 0.0083, 0.0065, -0.0030, -0.0164, 0.0079, -0.0124, -0.0135, -0.0191, + -0.0014, 0.0032], device='cuda:0'), grad: tensor([-0.0419, 0.0231, -0.0180, 0.0210, 0.0041, 0.0084, -0.0107, -0.0060, + 0.0237, -0.0037], device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 216.74, cls_loss 0.4779 cls_loss_mapping 0.0032 cls_loss_causal 0.4525 re_mapping 0.0056 re_causal 0.0148 /// teacc 98.70 lr 0.00010000 +Epoch 348, weight, value: tensor([[ 0.0380, 0.1169, -0.1788, ..., -0.0980, 0.0720, 0.0014], + [-0.0981, -0.0998, 0.1142, ..., -0.0456, -0.1011, 0.0491], + [-0.0930, -0.0672, -0.1232, ..., -0.1319, -0.0283, 0.0622], + ..., + [-0.0125, -0.1273, -0.1318, ..., -0.1090, -0.0415, 0.0692], + [-0.1272, -0.1127, -0.1126, ..., -0.0656, -0.0615, -0.0971], + [ 0.0210, -0.0286, -0.0628, ..., -0.1006, 0.0218, -0.0209]], + device='cuda:0'), grad: tensor([[ 1.1343e-04, -1.8489e-04, 9.8720e-07, ..., 8.8150e-07, + -2.3580e-04, 2.8801e-04], + [ 3.0208e-04, 2.2333e-06, -9.2089e-06, ..., 3.1665e-08, + 7.4990e-06, 8.0013e-04], + [ 4.6945e-04, 5.3942e-06, 8.5160e-06, ..., 8.6334e-07, + 2.2352e-05, 5.2309e-04], + ..., + [ 1.2922e-03, 3.7607e-06, 4.5449e-06, ..., 1.8626e-08, + 1.9893e-05, 9.5606e-04], + [ 4.7183e-04, 4.5970e-06, 6.1318e-06, ..., -2.5369e-06, + 3.5167e-05, 7.2050e-04], + [-1.0166e-03, 7.3463e-06, 2.0582e-06, ..., 1.1735e-07, + 1.4476e-05, -4.1122e-03]], device='cuda:0') +Epoch 348, bias, value: tensor([ 0.0079, 0.0065, -0.0028, -0.0160, 0.0078, -0.0125, -0.0133, -0.0194, + -0.0019, 0.0035], device='cuda:0'), grad: tensor([ 0.0048, 0.0133, 0.0088, 0.0074, -0.0007, 0.0075, 0.0087, -0.0111, + 0.0114, -0.0501], device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 216.80, cls_loss 0.4579 cls_loss_mapping 0.0026 cls_loss_causal 0.4312 re_mapping 0.0057 re_causal 0.0155 /// teacc 98.77 lr 0.00010000 +Epoch 349, weight, value: tensor([[ 0.0372, 0.1157, -0.1794, ..., -0.0979, 0.0711, 0.0024], + [-0.0980, -0.0986, 0.1152, ..., -0.0458, -0.1014, 0.0489], + [-0.0929, -0.0675, -0.1225, ..., -0.1318, -0.0272, 0.0626], + ..., + [-0.0127, -0.1274, -0.1329, ..., -0.1091, -0.0424, 0.0693], + [-0.1273, -0.1129, -0.1128, ..., -0.0669, -0.0612, -0.0978], + [ 0.0224, -0.0267, -0.0613, ..., -0.1003, 0.0238, -0.0201]], + device='cuda:0'), grad: tensor([[ 2.7823e-04, 3.9861e-07, 2.2399e-04, ..., 3.4750e-05, + 4.3899e-05, 1.7703e-04], + [ 4.4870e-04, -7.5847e-06, -1.8616e-03, ..., -3.0112e-04, + 1.1414e-04, -4.7922e-04], + [-7.4482e-04, 5.1484e-06, 2.5892e-04, ..., 4.6521e-05, + 8.3268e-05, -2.3758e-04], + ..., + [-3.3607e-03, 3.8929e-06, 5.0926e-04, ..., 4.1962e-05, + 9.0301e-05, -9.7809e-03], + [ 8.3923e-04, -1.1519e-05, 6.2275e-04, ..., 2.9850e-04, + 8.8811e-05, 8.4829e-04], + [ 4.2458e-03, 7.4245e-06, 4.2009e-04, ..., 4.5985e-05, + 9.9599e-05, 9.8801e-03]], device='cuda:0') +Epoch 349, bias, value: tensor([ 0.0091, 0.0064, -0.0033, -0.0160, 0.0080, -0.0118, -0.0145, -0.0199, + -0.0019, 0.0040], device='cuda:0'), grad: tensor([ 0.0135, -0.0348, -0.0062, 0.0172, 0.0157, -0.0175, -0.0168, -0.0385, + 0.0211, 0.0464], device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 216.82, cls_loss 0.4902 cls_loss_mapping 0.0045 cls_loss_causal 0.4576 re_mapping 0.0055 re_causal 0.0149 /// teacc 98.75 lr 0.00010000 +Epoch 350, weight, value: tensor([[ 0.0371, 0.1162, -0.1796, ..., -0.0983, 0.0714, 0.0023], + [-0.0987, -0.0992, 0.1143, ..., -0.0471, -0.1029, 0.0493], + [-0.0925, -0.0683, -0.1228, ..., -0.1331, -0.0272, 0.0626], + ..., + [-0.0133, -0.1270, -0.1333, ..., -0.1088, -0.0433, 0.0701], + [-0.1266, -0.1136, -0.1131, ..., -0.0667, -0.0606, -0.0978], + [ 0.0227, -0.0267, -0.0606, ..., -0.1009, 0.0239, -0.0218]], + device='cuda:0'), grad: tensor([[-7.3528e-04, -1.7891e-03, 2.7463e-05, ..., 1.0088e-05, + -4.5753e-04, 4.2200e-04], + [ 1.0681e-04, 3.2663e-04, 7.3242e-04, ..., 4.6939e-05, + 1.2387e-07, 5.5218e-04], + [ 5.0497e-04, 7.8619e-05, 8.2314e-05, ..., 7.1563e-06, + 5.6885e-06, 1.4150e-04], + ..., + [ 5.2595e-04, 3.4142e-04, 2.2399e-04, ..., 5.0478e-06, + 2.9549e-05, 3.9053e-04], + [ 2.0611e-04, -3.1433e-03, -2.3956e-03, ..., 6.4969e-05, + 6.2108e-05, 3.7384e-04], + [-4.8232e-04, 1.9562e-04, 8.4758e-05, ..., 2.8223e-05, + -1.4022e-05, 1.5461e-04]], device='cuda:0') +Epoch 350, bias, value: tensor([ 0.0093, 0.0050, -0.0040, -0.0148, 0.0089, -0.0114, -0.0147, -0.0189, + -0.0023, 0.0030], device='cuda:0'), grad: tensor([ 0.0072, 0.0120, 0.0113, 0.0383, 0.0040, -0.0088, -0.0218, -0.0164, + -0.0029, -0.0231], device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 216.82, cls_loss 0.4664 cls_loss_mapping 0.0028 cls_loss_causal 0.4387 re_mapping 0.0059 re_causal 0.0161 /// teacc 98.78 lr 0.00010000 +Epoch 351, weight, value: tensor([[ 0.0379, 0.1164, -0.1802, ..., -0.0985, 0.0724, 0.0022], + [-0.0992, -0.0982, 0.1152, ..., -0.0465, -0.1025, 0.0495], + [-0.0932, -0.0681, -0.1226, ..., -0.1336, -0.0276, 0.0645], + ..., + [-0.0136, -0.1277, -0.1334, ..., -0.1077, -0.0450, 0.0697], + [-0.1253, -0.1139, -0.1132, ..., -0.0674, -0.0609, -0.0981], + [ 0.0222, -0.0262, -0.0611, ..., -0.1011, 0.0238, -0.0226]], + device='cuda:0'), grad: tensor([[ 4.1199e-04, 3.4928e-04, 2.4581e-04, ..., 2.4939e-04, + 2.5296e-04, 3.0398e-04], + [ 2.5964e-04, 3.0184e-04, 4.5347e-04, ..., 2.1899e-04, + 6.5863e-05, 3.3355e-04], + [ 6.8521e-04, 7.6532e-04, 4.1676e-04, ..., 2.7347e-04, + 2.2304e-04, 4.9973e-04], + ..., + [-2.9030e-03, -1.1587e-04, 1.3494e-03, ..., 9.3162e-05, + -1.0319e-03, -1.4591e-03], + [-1.2369e-03, -6.9523e-04, -9.0981e-04, ..., 1.9181e-04, + -6.3610e-04, -3.7646e-04], + [ 5.4598e-04, 6.9201e-05, -1.9312e-03, ..., 5.7220e-05, + 1.9670e-04, 3.0351e-04]], device='cuda:0') +Epoch 351, bias, value: tensor([ 0.0090, 0.0049, -0.0030, -0.0156, 0.0099, -0.0118, -0.0143, -0.0200, + -0.0015, 0.0026], device='cuda:0'), grad: tensor([ 0.0154, 0.0208, 0.0186, 0.0161, -0.0119, -0.0059, -0.0275, -0.0256, + 0.0107, -0.0106], device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 216.94, cls_loss 0.4713 cls_loss_mapping 0.0039 cls_loss_causal 0.4451 re_mapping 0.0055 re_causal 0.0147 /// teacc 98.75 lr 0.00010000 +Epoch 352, weight, value: tensor([[ 0.0397, 0.1171, -0.1800, ..., -0.0987, 0.0734, 0.0029], + [-0.0991, -0.0987, 0.1153, ..., -0.0459, -0.1029, 0.0494], + [-0.0931, -0.0672, -0.1218, ..., -0.1339, -0.0269, 0.0642], + ..., + [-0.0137, -0.1276, -0.1343, ..., -0.1086, -0.0456, 0.0707], + [-0.1270, -0.1152, -0.1141, ..., -0.0681, -0.0620, -0.0981], + [ 0.0220, -0.0267, -0.0620, ..., -0.1012, 0.0239, -0.0226]], + device='cuda:0'), grad: tensor([[ 9.5963e-05, 1.5843e-04, 2.8759e-05, ..., 6.9082e-05, + 9.4116e-05, 7.1049e-05], + [ 8.5056e-05, 1.5819e-04, 8.6010e-05, ..., 1.0061e-04, + 5.0068e-05, -1.2779e-03], + [ 6.6578e-05, -1.3018e-03, 4.4703e-05, ..., 2.1294e-05, + 3.5644e-05, 6.6996e-05], + ..., + [-3.6597e-04, 2.6608e-04, 8.7917e-05, ..., 2.3190e-06, + -2.2423e-04, 5.2691e-04], + [ 6.7174e-05, -2.6741e-03, -3.2501e-03, ..., -4.3030e-03, + 4.1604e-05, -8.1491e-04], + [-3.2514e-05, 1.2302e-04, -2.6917e-04, ..., 4.0159e-06, + -3.9823e-06, 4.8447e-04]], device='cuda:0') +Epoch 352, bias, value: tensor([ 0.0097, 0.0043, -0.0027, -0.0151, 0.0089, -0.0118, -0.0149, -0.0190, + -0.0018, 0.0028], device='cuda:0'), grad: tensor([ 0.0060, 0.0020, -0.0141, 0.0174, 0.0044, 0.0118, 0.0022, -0.0240, + -0.0110, 0.0054], device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 216.68, cls_loss 0.4680 cls_loss_mapping 0.0030 cls_loss_causal 0.4445 re_mapping 0.0056 re_causal 0.0145 /// teacc 98.78 lr 0.00010000 +Epoch 353, weight, value: tensor([[ 0.0400, 0.1174, -0.1789, ..., -0.0984, 0.0733, 0.0036], + [-0.1001, -0.0996, 0.1142, ..., -0.0461, -0.1035, 0.0496], + [-0.0917, -0.0673, -0.1211, ..., -0.1353, -0.0262, 0.0637], + ..., + [-0.0145, -0.1272, -0.1349, ..., -0.1100, -0.0466, 0.0718], + [-0.1270, -0.1150, -0.1135, ..., -0.0680, -0.0622, -0.0996], + [ 0.0230, -0.0262, -0.0610, ..., -0.0989, 0.0254, -0.0230]], + device='cuda:0'), grad: tensor([[ 8.5068e-04, 3.8695e-04, 3.3045e-04, ..., 3.2961e-05, + 2.7633e-04, 1.5867e-04], + [-1.5888e-03, 7.1824e-05, -2.7008e-03, ..., 1.4567e-04, + 5.2482e-05, 6.7902e-04], + [ 1.5128e-04, 1.7047e-04, 2.2268e-04, ..., -4.7398e-04, + 1.1867e-04, -2.3746e-03], + ..., + [-3.4161e-03, 1.0109e-04, -5.0163e-04, ..., 3.4332e-05, + 1.0031e-04, 3.0613e-04], + [ 9.0599e-04, 1.3554e-04, 4.2152e-04, ..., 4.3631e-05, + 9.4533e-05, 2.2030e-04], + [ 1.5778e-02, 1.1790e-04, 5.3835e-04, ..., 3.1769e-05, + 1.4400e-04, 1.7083e-04]], device='cuda:0') +Epoch 353, bias, value: tensor([ 0.0105, 0.0044, -0.0019, -0.0159, 0.0080, -0.0124, -0.0146, -0.0202, + -0.0008, 0.0032], device='cuda:0'), grad: tensor([-0.0083, -0.0380, -0.0138, -0.0105, 0.0257, -0.0081, -0.0097, -0.0074, + 0.0201, 0.0500], device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 216.59, cls_loss 0.5068 cls_loss_mapping 0.0036 cls_loss_causal 0.4771 re_mapping 0.0054 re_causal 0.0144 /// teacc 98.71 lr 0.00010000 +Epoch 354, weight, value: tensor([[ 0.0394, 0.1180, -0.1802, ..., -0.0987, 0.0736, 0.0024], + [-0.0996, -0.1001, 0.1135, ..., -0.0470, -0.1032, 0.0502], + [-0.0914, -0.0682, -0.1212, ..., -0.1355, -0.0250, 0.0640], + ..., + [-0.0136, -0.1273, -0.1346, ..., -0.1083, -0.0465, 0.0724], + [-0.1270, -0.1155, -0.1133, ..., -0.0670, -0.0627, -0.0992], + [ 0.0227, -0.0266, -0.0597, ..., -0.0991, 0.0245, -0.0234]], + device='cuda:0'), grad: tensor([[-3.6259e-03, 3.3379e-04, 2.0210e-07, ..., -2.6152e-05, + -2.9125e-03, 2.5730e-03], + [-1.1272e-03, 3.4928e-05, 2.4121e-07, ..., -3.4351e-03, + 1.6531e-06, -6.7329e-03], + [ 1.0020e-04, 5.1975e-04, 3.7439e-07, ..., 2.1667e-03, + 1.0962e-06, 3.5133e-03], + ..., + [-1.8682e-03, 2.8014e-04, 5.6885e-06, ..., 1.4806e-04, + 1.9893e-06, 2.2869e-03], + [-2.6536e-04, -1.5841e-03, 6.6198e-06, ..., -1.0157e-03, + 2.4661e-06, -2.2602e-03], + [ 2.0638e-03, 3.3045e-04, 2.8890e-06, ..., 3.5691e-04, + 4.4778e-06, -3.8099e-04]], device='cuda:0') +Epoch 354, bias, value: tensor([ 0.0109, 0.0041, -0.0010, -0.0166, 0.0088, -0.0129, -0.0153, -0.0208, + -0.0002, 0.0032], device='cuda:0'), grad: tensor([ 0.0241, -0.0683, 0.0163, -0.0064, -0.0128, 0.0019, 0.0289, 0.0214, + -0.0034, -0.0018], device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 216.30, cls_loss 0.5023 cls_loss_mapping 0.0028 cls_loss_causal 0.4766 re_mapping 0.0053 re_causal 0.0143 /// teacc 98.78 lr 0.00010000 +Epoch 355, weight, value: tensor([[ 0.0398, 0.1189, -0.1804, ..., -0.0980, 0.0733, 0.0022], + [-0.0971, -0.0999, 0.1145, ..., -0.0470, -0.1035, 0.0513], + [-0.0932, -0.0688, -0.1213, ..., -0.1354, -0.0259, 0.0640], + ..., + [-0.0121, -0.1273, -0.1367, ..., -0.1095, -0.0444, 0.0726], + [-0.1288, -0.1168, -0.1143, ..., -0.0684, -0.0639, -0.1005], + [ 0.0220, -0.0286, -0.0603, ..., -0.0996, 0.0239, -0.0219]], + device='cuda:0'), grad: tensor([[ 1.6972e-05, -3.2291e-03, 8.1956e-05, ..., 7.0095e-05, + 3.7104e-05, 3.2276e-05], + [ 2.0730e-04, 2.1368e-05, 7.2718e-04, ..., 2.7132e-04, + 4.5657e-05, 8.3065e-04], + [ 1.4901e-04, 6.6340e-05, -1.7757e-03, ..., -8.7070e-04, + 3.8058e-05, -2.1667e-03], + ..., + [ 6.0892e-04, 3.0785e-03, 6.7139e-04, ..., 3.3236e-04, + 3.7670e-05, -4.8637e-04], + [ 2.2328e-04, -3.5119e-04, -6.5155e-03, ..., -3.0537e-03, + -7.6473e-05, 7.3242e-04], + [-3.6049e-03, 6.4731e-05, 3.5572e-04, ..., 2.3842e-04, + 4.3660e-05, -2.6073e-03]], device='cuda:0') +Epoch 355, bias, value: tensor([ 0.0119, 0.0045, -0.0023, -0.0158, 0.0084, -0.0132, -0.0151, -0.0206, + -0.0017, 0.0041], device='cuda:0'), grad: tensor([ 0.0272, 0.0100, -0.0059, 0.0179, -0.0123, 0.0132, 0.0076, 0.0098, + -0.0034, -0.0642], device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 216.63, cls_loss 0.4816 cls_loss_mapping 0.0048 cls_loss_causal 0.4639 re_mapping 0.0053 re_causal 0.0141 /// teacc 98.54 lr 0.00010000 +Epoch 356, weight, value: tensor([[ 0.0397, 0.1198, -0.1794, ..., -0.0976, 0.0728, 0.0009], + [-0.0970, -0.0996, 0.1162, ..., -0.0471, -0.1039, 0.0517], + [-0.0931, -0.0697, -0.1220, ..., -0.1353, -0.0266, 0.0633], + ..., + [-0.0122, -0.1275, -0.1354, ..., -0.1077, -0.0451, 0.0728], + [-0.1277, -0.1170, -0.1139, ..., -0.0661, -0.0627, -0.0985], + [ 0.0214, -0.0288, -0.0606, ..., -0.0984, 0.0239, -0.0228]], + device='cuda:0'), grad: tensor([[ 6.2370e-03, 1.2159e-03, 8.4829e-04, ..., 1.1387e-03, + 3.9411e-04, 2.1496e-03], + [ 1.6499e-03, 1.7631e-04, 4.6325e-04, ..., 3.6001e-04, + 7.5936e-05, -1.0881e-03], + [-1.1749e-03, 1.1533e-04, -1.3723e-03, ..., 3.4904e-04, + 6.4254e-05, -1.7309e-04], + ..., + [ 1.3189e-03, 1.4925e-04, 4.4465e-04, ..., 1.2827e-04, + 2.3469e-05, 4.6659e-04], + [ 2.1992e-03, 3.7670e-04, 4.8971e-04, ..., 1.6804e-03, + 8.2195e-05, 7.7105e-04], + [ 1.3113e-03, 1.6963e-04, 4.5347e-04, ..., 1.6391e-04, + 3.0369e-05, 4.6587e-04]], device='cuda:0') +Epoch 356, bias, value: tensor([ 0.0105, 0.0045, -0.0033, -0.0154, 0.0090, -0.0139, -0.0146, -0.0202, + -0.0002, 0.0039], device='cuda:0'), grad: tensor([ 0.0540, 0.0023, -0.0354, -0.0010, -0.0003, -0.0672, -0.0397, 0.0275, + 0.0351, 0.0247], device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 216.23, cls_loss 0.5086 cls_loss_mapping 0.0039 cls_loss_causal 0.4801 re_mapping 0.0056 re_causal 0.0141 /// teacc 98.62 lr 0.00010000 +Epoch 357, weight, value: tensor([[ 0.0401, 0.1205, -0.1796, ..., -0.0980, 0.0731, 0.0017], + [-0.0982, -0.1006, 0.1172, ..., -0.0482, -0.1051, 0.0517], + [-0.0933, -0.0698, -0.1217, ..., -0.1352, -0.0264, 0.0637], + ..., + [-0.0126, -0.1285, -0.1357, ..., -0.1071, -0.0448, 0.0726], + [-0.1273, -0.1160, -0.1143, ..., -0.0663, -0.0628, -0.0981], + [ 0.0229, -0.0284, -0.0606, ..., -0.0972, 0.0241, -0.0235]], + device='cuda:0'), grad: tensor([[ 3.1433e-03, 4.3511e-06, 2.4006e-05, ..., 2.9540e-04, + 5.3835e-04, 7.8058e-04], + [-2.0313e-03, 5.0163e-04, 3.7956e-03, ..., 1.0757e-03, + -5.0783e-04, -7.7095e-03], + [ 1.0653e-03, 2.4050e-05, 2.5964e-04, ..., 1.0490e-04, + 3.2187e-04, 1.3714e-03], + ..., + [-1.1654e-03, 8.5384e-06, -1.3947e-04, ..., 4.2826e-05, + 3.4499e-04, 5.4207e-03], + [ 2.5826e-03, -5.7554e-04, -4.3602e-03, ..., -1.5945e-03, + 3.7813e-04, -2.7657e-03], + [ 1.2474e-03, 1.1683e-05, 2.0099e-04, ..., 1.1069e-04, + 6.1607e-04, 1.1473e-03]], device='cuda:0') +Epoch 357, bias, value: tensor([ 0.0092, 0.0055, -0.0028, -0.0159, 0.0095, -0.0138, -0.0139, -0.0205, + -0.0008, 0.0033], device='cuda:0'), grad: tensor([-0.0060, -0.0054, -0.0046, 0.0296, -0.0092, 0.0220, -0.0618, 0.0183, + -0.0019, 0.0189], device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 216.60, cls_loss 0.4696 cls_loss_mapping 0.0035 cls_loss_causal 0.4475 re_mapping 0.0062 re_causal 0.0160 /// teacc 98.78 lr 0.00010000 +Epoch 358, weight, value: tensor([[ 0.0395, 0.1213, -0.1798, ..., -0.0984, 0.0734, 0.0022], + [-0.0972, -0.1003, 0.1168, ..., -0.0488, -0.1059, 0.0515], + [-0.0929, -0.0691, -0.1207, ..., -0.1343, -0.0258, 0.0634], + ..., + [-0.0123, -0.1274, -0.1341, ..., -0.1050, -0.0444, 0.0732], + [-0.1272, -0.1165, -0.1156, ..., -0.0674, -0.0618, -0.0981], + [ 0.0237, -0.0291, -0.0612, ..., -0.0982, 0.0244, -0.0234]], + device='cuda:0'), grad: tensor([[ 1.3220e-04, 2.2745e-04, 2.7156e-04, ..., 3.0935e-05, + 2.2143e-05, -4.9055e-05], + [ 1.8680e-04, 2.2137e-04, 2.1565e-04, ..., 3.3200e-05, + 8.8394e-05, 3.1972e-04], + [ 4.0913e-04, 5.5122e-04, 1.4615e-04, ..., 1.1057e-05, + 8.3983e-05, 3.8695e-04], + ..., + [ 1.2169e-03, 1.6842e-03, 1.5450e-04, ..., 4.6998e-05, + 9.1910e-05, 9.7809e-03], + [ 3.3355e-04, 4.3631e-04, 1.9896e-04, ..., 2.6792e-05, + 6.6400e-05, 4.7421e-04], + [-7.8058e-04, -1.0052e-03, -2.0275e-03, ..., -2.7037e-04, + -6.4325e-04, -1.0118e-03]], device='cuda:0') +Epoch 358, bias, value: tensor([ 0.0087, 0.0056, -0.0022, -0.0163, 0.0093, -0.0125, -0.0141, -0.0210, + -0.0016, 0.0042], device='cuda:0'), grad: tensor([-0.0106, 0.0233, -0.0083, -0.0409, 0.0226, 0.0006, 0.0248, 0.0407, + -0.0088, -0.0433], device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 216.05, cls_loss 0.4695 cls_loss_mapping 0.0062 cls_loss_causal 0.4480 re_mapping 0.0060 re_causal 0.0150 /// teacc 98.58 lr 0.00010000 +Epoch 359, weight, value: tensor([[ 0.0388, 0.1219, -0.1802, ..., -0.0976, 0.0737, 0.0021], + [-0.0971, -0.0989, 0.1165, ..., -0.0466, -0.1060, 0.0506], + [-0.0937, -0.0685, -0.1196, ..., -0.1345, -0.0249, 0.0636], + ..., + [-0.0123, -0.1279, -0.1337, ..., -0.1047, -0.0438, 0.0729], + [-0.1267, -0.1172, -0.1170, ..., -0.0688, -0.0621, -0.0979], + [ 0.0232, -0.0284, -0.0603, ..., -0.0991, 0.0247, -0.0230]], + device='cuda:0'), grad: tensor([[ 2.1577e-04, 5.8270e-04, -1.0693e-04, ..., 2.7299e-04, + -4.1164e-06, 1.1663e-03], + [ 6.8712e-04, -5.1546e-04, 2.4676e-04, ..., 5.6744e-04, + 1.1874e-06, -1.0281e-03], + [-4.5280e-03, 2.0294e-03, 2.6665e-03, ..., -3.3264e-03, + 1.8343e-05, -2.4891e-03], + ..., + [-2.2144e-03, 7.2670e-04, 2.8038e-04, ..., 5.2309e-04, + -2.5228e-05, 8.6689e-04], + [ 4.8757e-04, 4.1199e-04, 2.0766e-04, ..., 3.4785e-04, + 6.5081e-06, 7.2432e-04], + [ 1.1740e-03, -7.4291e-04, -1.0163e-04, ..., -1.3578e-04, + 7.6473e-05, -3.1033e-03]], device='cuda:0') +Epoch 359, bias, value: tensor([ 0.0085, 0.0050, -0.0032, -0.0154, 0.0088, -0.0119, -0.0128, -0.0206, + -0.0024, 0.0042], device='cuda:0'), grad: tensor([-0.0397, -0.0002, -0.0082, 0.0048, 0.0001, 0.0267, 0.0228, 0.0083, + 0.0186, -0.0333], device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 216.29, cls_loss 0.4948 cls_loss_mapping 0.0037 cls_loss_causal 0.4691 re_mapping 0.0057 re_causal 0.0145 /// teacc 98.79 lr 0.00010000 +Epoch 360, weight, value: tensor([[ 0.0395, 0.1225, -0.1807, ..., -0.0993, 0.0739, -0.0011], + [-0.0959, -0.0974, 0.1170, ..., -0.0458, -0.1045, 0.0519], + [-0.0925, -0.0701, -0.1189, ..., -0.1337, -0.0254, 0.0628], + ..., + [-0.0135, -0.1292, -0.1332, ..., -0.1059, -0.0455, 0.0740], + [-0.1264, -0.1163, -0.1162, ..., -0.0674, -0.0626, -0.0976], + [ 0.0233, -0.0286, -0.0606, ..., -0.0993, 0.0249, -0.0233]], + device='cuda:0'), grad: tensor([[-1.2760e-03, -1.7204e-03, 2.9397e-04, ..., 1.0312e-05, + -1.1053e-03, 5.0735e-04], + [ 3.0112e-04, 1.2898e-04, -2.7695e-03, ..., -8.3160e-04, + 8.3029e-05, -7.9775e-04], + [ 2.8181e-04, 1.6367e-04, 3.4738e-04, ..., 3.1013e-06, + 1.0806e-04, 5.9509e-03], + ..., + [ 4.3106e-04, 1.6665e-04, 4.1270e-04, ..., 6.7204e-06, + 1.1849e-04, -8.0032e-03], + [-1.5345e-03, 1.7607e-04, -8.6308e-04, ..., 7.7820e-04, + 1.3041e-04, 1.2655e-03], + [ 9.4128e-04, 2.5582e-04, 6.6948e-04, ..., -1.4555e-04, + -3.4833e-04, 1.3456e-03]], device='cuda:0') +Epoch 360, bias, value: tensor([ 0.0077, 0.0054, -0.0032, -0.0153, 0.0089, -0.0117, -0.0134, -0.0215, + -0.0016, 0.0046], device='cuda:0'), grad: tensor([-1.0178e-02, -6.9946e-02, 3.2562e-02, -4.8180e-03, 2.8015e-02, + 2.5864e-02, 1.9958e-02, -2.1820e-02, 6.0737e-05, 2.4462e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 216.12, cls_loss 0.4828 cls_loss_mapping 0.0023 cls_loss_causal 0.4605 re_mapping 0.0057 re_causal 0.0152 /// teacc 98.79 lr 0.00010000 +Epoch 361, weight, value: tensor([[ 0.0395, 0.1231, -0.1809, ..., -0.0980, 0.0734, 0.0004], + [-0.0940, -0.0975, 0.1166, ..., -0.0455, -0.1046, 0.0512], + [-0.0930, -0.0705, -0.1188, ..., -0.1339, -0.0258, 0.0631], + ..., + [-0.0127, -0.1285, -0.1332, ..., -0.1047, -0.0444, 0.0739], + [-0.1264, -0.1176, -0.1171, ..., -0.0679, -0.0627, -0.0979], + [ 0.0229, -0.0275, -0.0612, ..., -0.0997, 0.0241, -0.0235]], + device='cuda:0'), grad: tensor([[-3.3617e-04, 4.2648e-03, -2.4261e-03, ..., 2.0826e-04, + -5.3215e-04, 1.1835e-03], + [ 4.4799e-04, 5.8651e-04, 9.9373e-04, ..., 3.4046e-04, + 6.0707e-05, 3.6764e-04], + [ 2.9087e-04, 3.3355e-04, -1.8191e-04, ..., 5.6863e-05, + 6.1929e-05, 4.4870e-04], + ..., + [-1.5678e-03, 4.1872e-05, 6.6137e-04, ..., 5.4747e-05, + 5.4210e-05, -3.0594e-03], + [ 2.6436e-03, 3.7956e-03, 2.2049e-03, ..., 2.9812e-03, + 7.0691e-05, 1.8568e-03], + [-6.2294e-03, -5.3520e-03, -7.6294e-03, ..., -8.4076e-03, + 6.3956e-05, -3.2444e-03]], device='cuda:0') +Epoch 361, bias, value: tensor([ 0.0086, 0.0053, -0.0043, -0.0159, 0.0083, -0.0122, -0.0125, -0.0206, + -0.0017, 0.0049], device='cuda:0'), grad: tensor([ 0.0054, 0.0318, 0.0006, -0.0484, 0.0237, 0.0416, -0.0241, 0.0141, + 0.0186, -0.0634], device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 216.64, cls_loss 0.4538 cls_loss_mapping 0.0022 cls_loss_causal 0.4345 re_mapping 0.0054 re_causal 0.0139 /// teacc 98.91 lr 0.00010000 +Epoch 362, weight, value: tensor([[ 0.0388, 0.1227, -0.1815, ..., -0.0981, 0.0728, 0.0007], + [-0.0941, -0.0980, 0.1176, ..., -0.0455, -0.1062, 0.0504], + [-0.0937, -0.0698, -0.1196, ..., -0.1338, -0.0265, 0.0633], + ..., + [-0.0130, -0.1292, -0.1337, ..., -0.1049, -0.0470, 0.0738], + [-0.1259, -0.1193, -0.1175, ..., -0.0681, -0.0605, -0.0981], + [ 0.0226, -0.0270, -0.0605, ..., -0.0997, 0.0235, -0.0228]], + device='cuda:0'), grad: tensor([[ 6.2346e-05, -1.1809e-05, 1.2350e-04, ..., 8.3819e-09, + 7.1430e-04, -4.6768e-03], + [ 1.4007e-04, 1.6853e-05, 1.6570e-04, ..., -3.8650e-08, + 9.5558e-04, 2.0787e-05], + [ 8.3923e-05, 1.0097e-04, 1.0711e-04, ..., 7.9162e-09, + 7.8964e-04, 2.5673e-03], + ..., + [-2.7485e-03, -4.4537e-04, -1.4248e-03, ..., 1.4901e-08, + -1.8501e-03, -8.5783e-04], + [ 1.0347e-04, 2.0862e-04, 1.5509e-04, ..., 3.2596e-07, + 6.3086e-04, 9.5487e-05], + [ 1.0109e-03, 6.4230e-04, 7.3242e-04, ..., 1.3504e-08, + 2.1801e-03, 1.6804e-03]], device='cuda:0') +Epoch 362, bias, value: tensor([ 0.0092, 0.0055, -0.0048, -0.0162, 0.0088, -0.0124, -0.0128, -0.0207, + -0.0017, 0.0051], device='cuda:0'), grad: tensor([-0.0048, 0.0064, 0.0103, 0.0126, -0.0184, 0.0067, 0.0053, -0.0388, + 0.0051, 0.0156], device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 216.28, cls_loss 0.4502 cls_loss_mapping 0.0019 cls_loss_causal 0.4279 re_mapping 0.0054 re_causal 0.0146 /// teacc 98.88 lr 0.00010000 +Epoch 363, weight, value: tensor([[ 0.0397, 0.1228, -0.1826, ..., -0.0991, 0.0728, 0.0015], + [-0.0951, -0.0979, 0.1184, ..., -0.0465, -0.1081, 0.0500], + [-0.0943, -0.0701, -0.1202, ..., -0.1326, -0.0253, 0.0645], + ..., + [-0.0128, -0.1273, -0.1333, ..., -0.1041, -0.0468, 0.0733], + [-0.1254, -0.1200, -0.1179, ..., -0.0693, -0.0595, -0.0987], + [ 0.0230, -0.0262, -0.0600, ..., -0.0997, 0.0239, -0.0237]], + device='cuda:0'), grad: tensor([[ 5.3853e-05, 4.1664e-05, -2.2411e-03, ..., 4.5985e-05, + 8.7142e-05, -4.6234e-03], + [ 1.3530e-04, -8.1718e-05, -1.1623e-04, ..., 1.4760e-05, + 3.6567e-05, 1.7471e-03], + [ 2.7657e-04, 7.4267e-05, 6.5613e-04, ..., 3.6836e-05, + 8.0407e-05, 1.4391e-03], + ..., + [-2.6340e-03, 3.9816e-05, 3.1829e-04, ..., 4.2766e-05, + 6.8486e-05, -3.1033e-03], + [ 5.6982e-04, 4.5687e-05, 3.5787e-04, ..., 9.0003e-05, + 1.0282e-04, 6.0606e-04], + [ 2.9602e-03, 1.2386e-04, 2.0802e-04, ..., 1.1963e-04, + 3.0923e-04, 2.7885e-03]], device='cuda:0') +Epoch 363, bias, value: tensor([ 0.0089, 0.0055, -0.0037, -0.0167, 0.0090, -0.0132, -0.0128, -0.0203, + -0.0015, 0.0046], device='cuda:0'), grad: tensor([-5.0323e-02, 1.7700e-02, 1.2047e-02, 8.1406e-03, -3.9032e-02, + 9.5749e-03, 1.4397e-02, -6.3419e-05, 1.0712e-02, 1.6846e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 216.12, cls_loss 0.4965 cls_loss_mapping 0.0031 cls_loss_causal 0.4708 re_mapping 0.0054 re_causal 0.0154 /// teacc 98.78 lr 0.00010000 +Epoch 364, weight, value: tensor([[ 0.0386, 0.1221, -0.1819, ..., -0.0992, 0.0720, 0.0013], + [-0.0960, -0.0983, 0.1169, ..., -0.0475, -0.1100, 0.0499], + [-0.0935, -0.0705, -0.1196, ..., -0.1332, -0.0248, 0.0635], + ..., + [-0.0122, -0.1264, -0.1328, ..., -0.1044, -0.0473, 0.0748], + [-0.1244, -0.1196, -0.1186, ..., -0.0694, -0.0602, -0.0997], + [ 0.0220, -0.0262, -0.0581, ..., -0.0996, 0.0241, -0.0236]], + device='cuda:0'), grad: tensor([[ 7.3862e-04, 5.3501e-04, 4.4060e-04, ..., 1.0341e-04, + 2.1815e-04, 9.2268e-04], + [ 6.3610e-04, 1.6391e-05, -1.7986e-03, ..., 1.2994e-04, + 1.3971e-04, 8.2350e-04], + [ 4.7159e-04, -3.4809e-03, 3.6907e-04, ..., 7.4387e-05, + 9.2924e-05, -5.8136e-03], + ..., + [-9.0957e-05, -2.8992e-04, 4.7755e-04, ..., 8.1956e-05, + 1.1367e-04, -1.2010e-04], + [-2.5368e-03, 6.0177e-04, 2.6226e-04, ..., 7.5877e-05, + -1.0891e-03, 2.0294e-03], + [-6.8521e-04, 7.4744e-05, -9.6989e-04, ..., -3.7789e-04, + 1.5736e-04, -2.9016e-04]], device='cuda:0') +Epoch 364, bias, value: tensor([ 0.0083, 0.0051, -0.0036, -0.0174, 0.0095, -0.0129, -0.0133, -0.0196, + -0.0020, 0.0056], device='cuda:0'), grad: tensor([ 0.0248, 0.0053, -0.0873, -0.0011, -0.0083, 0.0265, 0.0267, 0.0202, + -0.0001, -0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 216.61, cls_loss 0.5035 cls_loss_mapping 0.0027 cls_loss_causal 0.4776 re_mapping 0.0056 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 +Epoch 365, weight, value: tensor([[ 0.0394, 0.1230, -0.1831, ..., -0.0993, 0.0723, 0.0014], + [-0.0966, -0.0974, 0.1175, ..., -0.0478, -0.1105, 0.0498], + [-0.0939, -0.0704, -0.1203, ..., -0.1344, -0.0241, 0.0629], + ..., + [-0.0126, -0.1274, -0.1336, ..., -0.1035, -0.0475, 0.0754], + [-0.1237, -0.1173, -0.1176, ..., -0.0695, -0.0604, -0.0991], + [ 0.0214, -0.0274, -0.0598, ..., -0.0994, 0.0235, -0.0240]], + device='cuda:0'), grad: tensor([[-1.3905e-03, -6.4000e-06, 2.8044e-05, ..., -1.8533e-06, + -1.3399e-03, -7.3481e-04], + [ 7.6175e-05, 3.6340e-06, 9.9093e-06, ..., 2.9802e-07, + 2.9579e-05, 5.3376e-05], + [-3.0975e-03, -2.7823e-04, -1.5192e-03, ..., 2.3078e-06, + 7.1168e-05, -4.0588e-03], + ..., + [ 6.0177e-04, 5.1320e-05, 2.5845e-04, ..., 1.5972e-07, + 3.2693e-05, 5.5981e-04], + [-7.4806e-03, -2.2621e-03, -3.1700e-03, ..., 2.0061e-06, + -4.0855e-03, -8.6927e-04], + [ 7.9041e-03, 2.3155e-03, 3.4866e-03, ..., 1.2079e-06, + 4.1466e-03, 1.8492e-03]], device='cuda:0') +Epoch 365, bias, value: tensor([ 0.0084, 0.0054, -0.0037, -0.0184, 0.0096, -0.0114, -0.0143, -0.0198, + -0.0020, 0.0058], device='cuda:0'), grad: tensor([-0.0052, 0.0043, -0.0196, 0.0062, 0.0046, 0.0154, 0.0101, 0.0058, + -0.0527, 0.0312], device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 216.41, cls_loss 0.4919 cls_loss_mapping 0.0025 cls_loss_causal 0.4682 re_mapping 0.0054 re_causal 0.0142 /// teacc 98.87 lr 0.00010000 +Epoch 366, weight, value: tensor([[ 0.0397, 0.1240, -0.1827, ..., -0.1002, 0.0725, 0.0016], + [-0.0957, -0.0979, 0.1175, ..., -0.0478, -0.1112, 0.0509], + [-0.0942, -0.0707, -0.1204, ..., -0.1342, -0.0242, 0.0631], + ..., + [-0.0110, -0.1281, -0.1350, ..., -0.1049, -0.0475, 0.0754], + [-0.1237, -0.1169, -0.1167, ..., -0.0702, -0.0599, -0.0993], + [ 0.0209, -0.0278, -0.0607, ..., -0.1001, 0.0230, -0.0243]], + device='cuda:0'), grad: tensor([[ 6.8998e-04, 1.2267e-04, 1.4234e-04, ..., 4.0364e-04, + 7.2598e-05, 7.2098e-04], + [ 2.5606e-04, 2.4542e-05, -3.2449e-04, ..., 1.4603e-04, + 5.4836e-05, -3.7422e-03], + [ 2.0278e-04, 3.7909e-05, 1.0228e-04, ..., 3.6657e-05, + 4.6223e-05, 4.6778e-04], + ..., + [ 3.7146e-04, 6.7234e-05, 1.5497e-04, ..., 1.2732e-04, + 9.4414e-05, 1.5182e-03], + [ 1.3027e-03, 1.7488e-04, 3.2473e-04, ..., 6.3848e-04, + 2.6894e-04, 1.8778e-03], + [-3.0565e-04, -7.9107e-04, -7.1287e-04, ..., 2.4390e-04, + -4.1795e-04, 1.1110e-03]], device='cuda:0') +Epoch 366, bias, value: tensor([ 0.0087, 0.0059, -0.0031, -0.0182, 0.0088, -0.0114, -0.0147, -0.0200, + -0.0021, 0.0058], device='cuda:0'), grad: tensor([ 0.0207, -0.0754, 0.0202, 0.0216, -0.0119, -0.0354, 0.0070, 0.0191, + 0.0184, 0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 216.30, cls_loss 0.4500 cls_loss_mapping 0.0037 cls_loss_causal 0.4248 re_mapping 0.0057 re_causal 0.0151 /// teacc 98.75 lr 0.00010000 +Epoch 367, weight, value: tensor([[ 0.0395, 0.1235, -0.1832, ..., -0.1006, 0.0725, 0.0019], + [-0.0969, -0.0977, 0.1173, ..., -0.0468, -0.1115, 0.0536], + [-0.0946, -0.0718, -0.1208, ..., -0.1340, -0.0247, 0.0627], + ..., + [-0.0101, -0.1288, -0.1347, ..., -0.1056, -0.0480, 0.0741], + [-0.1239, -0.1172, -0.1155, ..., -0.0695, -0.0588, -0.0991], + [ 0.0207, -0.0287, -0.0604, ..., -0.1013, 0.0229, -0.0244]], + device='cuda:0'), grad: tensor([[ 1.3435e-04, -7.1144e-04, 1.2204e-05, ..., 3.5530e-07, + -9.7137e-07, 2.5558e-04], + [ 4.1604e-04, -2.3227e-06, -2.5839e-05, ..., 2.3134e-06, + 8.2422e-08, 6.3419e-04], + [ 1.7023e-04, 2.1684e-04, 4.2081e-04, ..., 4.0187e-07, + 4.5169e-08, 1.1673e-03], + ..., + [ 6.0892e-04, 2.0385e-05, 3.9898e-06, ..., 3.4599e-07, + 7.7067e-07, 5.9694e-05], + [ 3.4475e-04, 5.6565e-05, 1.5087e-05, ..., -8.5821e-07, + 8.7079e-08, 4.8804e-04], + [ 3.7527e-04, 5.9605e-04, 1.2897e-05, ..., 1.1642e-06, + 4.2394e-06, 1.1101e-03]], device='cuda:0') +Epoch 367, bias, value: tensor([ 0.0074, 0.0057, -0.0038, -0.0179, 0.0093, -0.0107, -0.0143, -0.0197, + -0.0028, 0.0064], device='cuda:0'), grad: tensor([ 0.0047, 0.0124, -0.0200, 0.0123, -0.0138, -0.0216, 0.0075, 0.0170, + 0.0149, -0.0134], device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 216.23, cls_loss 0.4724 cls_loss_mapping 0.0031 cls_loss_causal 0.4577 re_mapping 0.0054 re_causal 0.0145 /// teacc 98.70 lr 0.00010000 +Epoch 368, weight, value: tensor([[ 0.0397, 0.1237, -0.1832, ..., -0.0996, 0.0723, 0.0052], + [-0.0971, -0.0970, 0.1165, ..., -0.0464, -0.1122, 0.0522], + [-0.0949, -0.0712, -0.1205, ..., -0.1334, -0.0254, 0.0636], + ..., + [-0.0108, -0.1293, -0.1356, ..., -0.1066, -0.0483, 0.0741], + [-0.1250, -0.1167, -0.1152, ..., -0.0689, -0.0576, -0.0985], + [ 0.0218, -0.0282, -0.0589, ..., -0.0999, 0.0228, -0.0246]], + device='cuda:0'), grad: tensor([[-6.3467e-04, -5.6791e-04, 8.2922e-04, ..., 9.7561e-04, + 9.6679e-05, -4.8518e-04], + [ 5.9700e-04, 3.6097e-04, -2.8763e-03, ..., -1.7366e-03, + 4.1425e-06, 3.8128e-03], + [-3.0193e-03, -4.0092e-03, -4.1504e-03, ..., 1.9515e-04, + 1.9193e-04, -1.0033e-02], + ..., + [ 4.3755e-03, 7.5340e-04, 4.5800e-04, ..., 1.5843e-04, + 4.4479e-03, 1.1658e-02], + [-1.4772e-03, 9.1493e-05, -4.1437e-04, ..., 9.4652e-05, + 1.1557e-04, -1.2932e-03], + [-2.9278e-03, 4.8518e-04, 6.9523e-04, ..., 3.1948e-04, + -5.0659e-03, -4.8180e-03]], device='cuda:0') +Epoch 368, bias, value: tensor([ 0.0076, 0.0054, -0.0031, -0.0184, 0.0095, -0.0116, -0.0140, -0.0197, + -0.0022, 0.0061], device='cuda:0'), grad: tensor([ 0.0221, 0.0067, -0.0758, 0.0250, 0.0142, -0.0037, 0.0015, 0.0484, + -0.0329, -0.0054], device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 216.30, cls_loss 0.4851 cls_loss_mapping 0.0029 cls_loss_causal 0.4617 re_mapping 0.0056 re_causal 0.0153 /// teacc 98.86 lr 0.00010000 +Epoch 369, weight, value: tensor([[ 0.0392, 0.1231, -0.1845, ..., -0.1003, 0.0729, 0.0045], + [-0.0972, -0.0979, 0.1173, ..., -0.0456, -0.1131, 0.0518], + [-0.0949, -0.0704, -0.1211, ..., -0.1341, -0.0246, 0.0639], + ..., + [-0.0113, -0.1300, -0.1367, ..., -0.1073, -0.0500, 0.0743], + [-0.1248, -0.1166, -0.1146, ..., -0.0677, -0.0564, -0.0998], + [ 0.0218, -0.0297, -0.0581, ..., -0.1005, 0.0230, -0.0251]], + device='cuda:0'), grad: tensor([[-2.3232e-03, -4.5568e-05, 7.1621e-04, ..., -1.0777e-03, + 2.5511e-04, -2.6550e-03], + [ 5.2977e-04, 1.0467e-04, -8.9547e-07, ..., 1.1897e-04, + -4.5180e-04, 4.3774e-04], + [ 2.8801e-04, 1.0300e-04, 6.0225e-04, ..., -1.8740e-04, + 2.3389e-04, -2.0921e-04], + ..., + [ 1.5440e-03, 1.1796e-04, 1.0996e-03, ..., 3.4499e-04, + 4.0793e-04, 1.1635e-03], + [ 6.6900e-04, 1.2219e-04, 1.1034e-03, ..., 3.3927e-04, + 2.7275e-04, 7.1430e-04], + [-6.4964e-03, 2.3341e-04, -9.1858e-03, ..., 3.4809e-04, + -7.5221e-05, -1.1749e-03]], device='cuda:0') +Epoch 369, bias, value: tensor([ 0.0076, 0.0058, -0.0038, -0.0185, 0.0102, -0.0115, -0.0140, -0.0208, + -0.0026, 0.0072], device='cuda:0'), grad: tensor([ 0.0013, -0.0367, 0.0137, -0.0210, 0.0528, 0.0303, -0.0380, 0.0367, + -0.0369, -0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 216.52, cls_loss 0.4508 cls_loss_mapping 0.0031 cls_loss_causal 0.4309 re_mapping 0.0056 re_causal 0.0155 /// teacc 98.80 lr 0.00010000 +Epoch 370, weight, value: tensor([[ 0.0398, 0.1246, -0.1824, ..., -0.0987, 0.0733, 0.0052], + [-0.0969, -0.0989, 0.1161, ..., -0.0459, -0.1130, 0.0524], + [-0.0945, -0.0701, -0.1203, ..., -0.1352, -0.0243, 0.0627], + ..., + [-0.0126, -0.1299, -0.1371, ..., -0.1092, -0.0499, 0.0749], + [-0.1253, -0.1167, -0.1149, ..., -0.0673, -0.0567, -0.0995], + [ 0.0217, -0.0308, -0.0583, ..., -0.1017, 0.0224, -0.0266]], + device='cuda:0'), grad: tensor([[-7.9918e-04, -5.7030e-03, 4.4823e-05, ..., 2.0489e-06, + -2.6283e-03, -1.1244e-03], + [ 3.7074e-05, 2.9296e-05, 6.1035e-05, ..., 2.7474e-07, + 5.2989e-05, 1.9894e-03], + [ 2.0599e-04, 2.1935e-03, 2.1458e-04, ..., 2.6785e-06, + 1.0996e-03, 1.3247e-03], + ..., + [-4.4298e-04, 6.3419e-05, -1.1158e-03, ..., 2.4214e-08, + 5.3346e-05, -5.0163e-03], + [ 1.1140e-04, 9.9182e-05, 3.6359e-05, ..., 7.2969e-07, + 5.5492e-05, 1.5430e-03], + [ 3.9387e-04, 4.8804e-04, 1.1724e-04, ..., 5.0291e-08, + 1.5807e-04, 1.3733e-03]], device='cuda:0') +Epoch 370, bias, value: tensor([ 0.0080, 0.0061, -0.0033, -0.0189, 0.0107, -0.0107, -0.0148, -0.0209, + -0.0029, 0.0063], device='cuda:0'), grad: tensor([-0.0096, -0.0016, 0.0028, -0.0060, -0.0144, 0.0199, -0.0157, -0.0136, + 0.0177, 0.0206], device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 216.60, cls_loss 0.4643 cls_loss_mapping 0.0042 cls_loss_causal 0.4395 re_mapping 0.0053 re_causal 0.0135 /// teacc 99.00 lr 0.00010000 +Epoch 371, weight, value: tensor([[ 0.0396, 0.1243, -0.1819, ..., -0.0983, 0.0722, 0.0051], + [-0.0973, -0.0986, 0.1159, ..., -0.0462, -0.1147, 0.0515], + [-0.0952, -0.0702, -0.1192, ..., -0.1342, -0.0243, 0.0617], + ..., + [-0.0132, -0.1305, -0.1366, ..., -0.1093, -0.0501, 0.0762], + [-0.1254, -0.1172, -0.1154, ..., -0.0682, -0.0572, -0.0996], + [ 0.0224, -0.0295, -0.0600, ..., -0.1019, 0.0216, -0.0271]], + device='cuda:0'), grad: tensor([[-1.2672e-04, -3.1423e-06, 2.0072e-05, ..., 2.0564e-06, + -2.4586e-03, -4.9782e-04], + [ 2.7943e-04, 7.5221e-05, 9.9182e-04, ..., 1.9944e-04, + 2.3818e-04, 1.9131e-03], + [ 3.7813e-04, 6.2227e-05, -1.2140e-03, ..., -2.8849e-04, + 6.5660e-04, -2.5921e-03], + ..., + [ 1.9526e-04, 1.4886e-05, 1.4806e-04, ..., 2.8566e-05, + 2.2674e-04, 3.5024e-04], + [ 4.1246e-04, 1.5998e-04, 2.6560e-04, ..., 1.1936e-05, + 2.4605e-04, 3.5286e-04], + [ 4.4322e-04, 1.2201e-04, 1.9062e-04, ..., 6.4336e-06, + 2.6059e-04, 1.6105e-04]], device='cuda:0') +Epoch 371, bias, value: tensor([ 0.0082, 0.0048, -0.0042, -0.0189, 0.0112, -0.0098, -0.0142, -0.0210, + -0.0025, 0.0062], device='cuda:0'), grad: tensor([-0.0165, -0.0063, 0.0086, 0.0129, -0.0600, -0.0400, 0.0172, 0.0152, + 0.0362, 0.0328], device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 216.22, cls_loss 0.4856 cls_loss_mapping 0.0028 cls_loss_causal 0.4631 re_mapping 0.0061 re_causal 0.0156 /// teacc 98.85 lr 0.00010000 +Epoch 372, weight, value: tensor([[ 0.0396, 0.1261, -0.1823, ..., -0.0984, 0.0736, 0.0032], + [-0.0982, -0.0988, 0.1158, ..., -0.0468, -0.1155, 0.0528], + [-0.0963, -0.0708, -0.1198, ..., -0.1343, -0.0233, 0.0631], + ..., + [-0.0136, -0.1311, -0.1371, ..., -0.1099, -0.0499, 0.0752], + [-0.1253, -0.1170, -0.1152, ..., -0.0677, -0.0565, -0.1003], + [ 0.0233, -0.0296, -0.0598, ..., -0.1017, 0.0207, -0.0261]], + device='cuda:0'), grad: tensor([[ 2.3632e-03, 7.2813e-04, 1.5044e-04, ..., 9.8228e-04, + 6.1846e-04, 1.7242e-03], + [ 6.0272e-04, 5.4300e-05, 9.8526e-05, ..., 4.4107e-04, + 2.1911e-04, -1.7910e-03], + [-1.9312e-03, 2.6035e-04, 2.1183e-04, ..., -2.8210e-03, + -5.0926e-04, -2.6855e-03], + ..., + [ 8.0395e-04, 1.1623e-04, 1.6236e-04, ..., 2.2912e-04, + 2.2554e-04, 1.1950e-03], + [-4.0779e-03, -2.5787e-03, -2.8992e-04, ..., 8.5020e-04, + -8.5402e-04, -1.5691e-05], + [ 9.3689e-03, 6.2799e-04, 6.8188e-04, ..., 3.0565e-04, + 1.3075e-03, 3.9215e-03]], device='cuda:0') +Epoch 372, bias, value: tensor([ 0.0065, 0.0058, -0.0044, -0.0180, 0.0110, -0.0103, -0.0131, -0.0207, + -0.0026, 0.0054], device='cuda:0'), grad: tensor([ 0.0344, -0.0254, -0.0117, -0.0059, -0.0031, -0.0040, -0.0352, 0.0320, + -0.0026, 0.0215], device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 216.39, cls_loss 0.4758 cls_loss_mapping 0.0027 cls_loss_causal 0.4500 re_mapping 0.0058 re_causal 0.0157 /// teacc 98.80 lr 0.00010000 +Epoch 373, weight, value: tensor([[ 0.0400, 0.1273, -0.1818, ..., -0.0992, 0.0740, 0.0037], + [-0.0998, -0.0974, 0.1168, ..., -0.0460, -0.1145, 0.0530], + [-0.0949, -0.0699, -0.1192, ..., -0.1347, -0.0229, 0.0642], + ..., + [-0.0131, -0.1318, -0.1375, ..., -0.1106, -0.0488, 0.0751], + [-0.1254, -0.1172, -0.1169, ..., -0.0682, -0.0573, -0.0995], + [ 0.0233, -0.0309, -0.0606, ..., -0.1016, 0.0200, -0.0274]], + device='cuda:0'), grad: tensor([[ 1.6851e-03, 1.5306e-03, 2.6631e-04, ..., 7.8249e-04, + 1.0376e-03, 2.4433e-03], + [ 2.2590e-04, 2.1076e-04, 2.7728e-04, ..., 1.1247e-04, + 4.9114e-05, -1.6718e-03], + [ 2.9373e-04, 2.8396e-04, -3.6061e-05, ..., 5.3024e-04, + 1.0532e-04, 4.2796e-04], + ..., + [-4.4975e-03, -1.8940e-03, -2.0657e-03, ..., -2.4462e-04, + -1.4143e-03, -5.0774e-03], + [ 3.1519e-04, 2.6155e-04, 3.0518e-04, ..., 2.0714e-03, + 1.3006e-04, 2.4376e-03], + [ 1.5173e-03, -1.5049e-03, 2.6655e-04, ..., -6.4182e-04, + 4.5598e-05, 2.1782e-03]], device='cuda:0') +Epoch 373, bias, value: tensor([ 0.0074, 0.0056, -0.0053, -0.0183, 0.0109, -0.0105, -0.0131, -0.0192, + -0.0024, 0.0045], device='cuda:0'), grad: tensor([ 0.0159, 0.0120, -0.0147, -0.0164, -0.0408, 0.0211, 0.0135, -0.0435, + 0.0234, 0.0295], device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 216.30, cls_loss 0.4650 cls_loss_mapping 0.0030 cls_loss_causal 0.4409 re_mapping 0.0055 re_causal 0.0145 /// teacc 98.86 lr 0.00010000 +Epoch 374, weight, value: tensor([[ 0.0413, 0.1273, -0.1817, ..., -0.0994, 0.0761, 0.0045], + [-0.1006, -0.0974, 0.1174, ..., -0.0457, -0.1146, 0.0537], + [-0.0958, -0.0709, -0.1188, ..., -0.1357, -0.0217, 0.0630], + ..., + [-0.0131, -0.1330, -0.1377, ..., -0.1120, -0.0481, 0.0744], + [-0.1241, -0.1165, -0.1177, ..., -0.0675, -0.0565, -0.0995], + [ 0.0213, -0.0315, -0.0607, ..., -0.1020, 0.0193, -0.0268]], + device='cuda:0'), grad: tensor([[-8.4937e-05, -4.8804e-04, 9.0361e-05, ..., 8.2374e-05, + 1.2316e-05, 2.8324e-04], + [ 7.5758e-05, 3.2902e-05, 7.9155e-05, ..., 3.8505e-05, + 1.3828e-05, 2.5535e-04], + [-3.6508e-05, 1.8489e-04, 1.0157e-04, ..., 1.2922e-04, + 6.8188e-05, -9.4318e-04], + ..., + [-1.4293e-04, 6.5625e-05, -6.3479e-05, ..., 5.4270e-05, + 3.4660e-05, -3.4976e-04], + [ 3.5906e-04, 6.8855e-04, 5.9843e-04, ..., 5.2691e-04, + 2.8753e-04, 4.9973e-04], + [ 3.1033e-03, 9.7036e-05, 1.8761e-05, ..., 5.5134e-05, + 2.5883e-05, 1.7583e-04]], device='cuda:0') +Epoch 374, bias, value: tensor([ 0.0079, 0.0057, -0.0052, -0.0186, 0.0112, -0.0101, -0.0139, -0.0193, + -0.0015, 0.0036], device='cuda:0'), grad: tensor([ 0.0148, 0.0239, -0.0220, -0.0242, 0.0121, 0.0169, 0.0185, -0.0470, + 0.0193, -0.0125], device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 216.67, cls_loss 0.5010 cls_loss_mapping 0.0026 cls_loss_causal 0.4808 re_mapping 0.0054 re_causal 0.0149 /// teacc 98.70 lr 0.00010000 +Epoch 375, weight, value: tensor([[ 0.0417, 0.1277, -0.1815, ..., -0.0996, 0.0752, 0.0043], + [-0.1004, -0.0979, 0.1179, ..., -0.0459, -0.1156, 0.0533], + [-0.0961, -0.0722, -0.1190, ..., -0.1362, -0.0228, 0.0630], + ..., + [-0.0130, -0.1324, -0.1378, ..., -0.1116, -0.0477, 0.0744], + [-0.1238, -0.1151, -0.1178, ..., -0.0680, -0.0557, -0.0994], + [ 0.0213, -0.0319, -0.0608, ..., -0.1026, 0.0198, -0.0274]], + device='cuda:0'), grad: tensor([[ 1.4887e-03, 5.2834e-04, 3.0780e-04, ..., 1.4553e-03, + 1.3625e-06, 1.1578e-03], + [ 2.5606e-04, 3.3545e-04, 3.6073e-04, ..., 8.8692e-05, + 5.6904e-07, 1.0004e-03], + [ 6.2037e-04, 1.3614e-04, 2.6679e-04, ..., 8.7619e-05, + 1.3979e-06, 8.1110e-04], + ..., + [ 4.5371e-04, 3.1161e-04, 2.5678e-04, ..., 1.0365e-04, + 1.1218e-04, 1.1549e-03], + [ 3.1781e-04, -1.5726e-03, -1.3876e-03, ..., 5.9223e-04, + 5.6714e-05, 8.0299e-04], + [-8.0585e-04, 2.9206e-04, 2.1958e-04, ..., 4.6062e-04, + -7.0763e-04, 6.7663e-04]], device='cuda:0') +Epoch 375, bias, value: tensor([ 0.0076, 0.0057, -0.0059, -0.0185, 0.0109, -0.0096, -0.0135, -0.0186, + -0.0028, 0.0045], device='cuda:0'), grad: tensor([ 0.0335, 0.0248, 0.0202, -0.0307, -0.0401, -0.0034, 0.0009, 0.0364, + -0.0328, -0.0088], device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 216.55, cls_loss 0.4839 cls_loss_mapping 0.0025 cls_loss_causal 0.4612 re_mapping 0.0057 re_causal 0.0148 /// teacc 98.72 lr 0.00010000 +Epoch 376, weight, value: tensor([[ 0.0408, 0.1285, -0.1808, ..., -0.0993, 0.0755, 0.0044], + [-0.1006, -0.0982, 0.1190, ..., -0.0453, -0.1161, 0.0528], + [-0.0951, -0.0736, -0.1197, ..., -0.1363, -0.0228, 0.0634], + ..., + [-0.0132, -0.1302, -0.1375, ..., -0.1107, -0.0487, 0.0735], + [-0.1237, -0.1159, -0.1185, ..., -0.0678, -0.0545, -0.0993], + [ 0.0229, -0.0332, -0.0601, ..., -0.1026, 0.0197, -0.0264]], + device='cuda:0'), grad: tensor([[ 1.0405e-03, 2.2852e-04, 7.3528e-04, ..., 1.0529e-03, + 1.1790e-04, 5.0354e-04], + [-7.8392e-04, 7.5459e-05, -3.5152e-03, ..., 4.1695e-03, + 9.0241e-05, 2.9697e-03], + [-7.8125e-03, 9.9182e-04, 8.3208e-04, ..., 6.2180e-04, + 1.4806e-04, -4.5586e-03], + ..., + [ 3.5553e-03, 4.9496e-04, 2.7084e-03, ..., 6.9380e-05, + 2.0730e-04, 1.2264e-03], + [ 2.0638e-03, 9.8801e-04, 1.4124e-03, ..., 5.3978e-04, + 8.3447e-05, 3.4499e-04], + [-1.5244e-02, 2.7485e-03, 1.9007e-03, ..., 1.6534e-04, + 2.0778e-04, 9.4318e-04]], device='cuda:0') +Epoch 376, bias, value: tensor([ 0.0074, 0.0060, -0.0056, -0.0181, 0.0106, -0.0110, -0.0141, -0.0188, + -0.0018, 0.0051], device='cuda:0'), grad: tensor([ 0.0302, -0.0064, -0.0334, -0.0108, 0.0331, -0.0023, -0.0740, 0.0388, + 0.0090, 0.0158], device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 216.73, cls_loss 0.4539 cls_loss_mapping 0.0024 cls_loss_causal 0.4304 re_mapping 0.0055 re_causal 0.0141 /// teacc 98.86 lr 0.00010000 +Epoch 377, weight, value: tensor([[ 0.0414, 0.1297, -0.1810, ..., -0.0996, 0.0758, 0.0037], + [-0.1013, -0.0989, 0.1189, ..., -0.0451, -0.1162, 0.0519], + [-0.0954, -0.0738, -0.1200, ..., -0.1366, -0.0231, 0.0633], + ..., + [-0.0132, -0.1303, -0.1375, ..., -0.1105, -0.0485, 0.0740], + [-0.1237, -0.1157, -0.1172, ..., -0.0677, -0.0533, -0.1002], + [ 0.0235, -0.0328, -0.0602, ..., -0.1017, 0.0192, -0.0253]], + device='cuda:0'), grad: tensor([[ 1.5944e-05, -7.3109e-08, 2.8777e-04, ..., 4.6194e-07, + -3.1106e-07, 1.6069e-04], + [ 3.7730e-05, 2.8405e-08, 4.2987e-04, ..., -8.9854e-06, + 5.5879e-09, 9.1851e-05], + [ 8.5056e-05, 1.0245e-06, 2.9659e-04, ..., 9.0385e-07, + 2.1560e-07, 3.5286e-04], + ..., + [-9.7418e-04, 1.8450e-06, -3.2310e-03, ..., 1.8338e-06, + 5.5879e-09, -2.9049e-03], + [-8.0032e-03, -1.9501e-02, -1.2672e-02, ..., 3.7625e-06, + 1.3430e-06, 3.3617e-04], + [ 9.4604e-03, 1.7834e-04, 4.8733e-04, ..., 7.4366e-07, + 7.6368e-08, 8.4734e-04]], device='cuda:0') +Epoch 377, bias, value: tensor([ 0.0076, 0.0068, -0.0046, -0.0179, 0.0108, -0.0119, -0.0143, -0.0195, + -0.0022, 0.0048], device='cuda:0'), grad: tensor([ 0.0105, -0.0207, 0.0120, -0.0103, 0.0108, 0.0292, 0.0108, -0.0431, + -0.0353, 0.0360], device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 216.57, cls_loss 0.4337 cls_loss_mapping 0.0031 cls_loss_causal 0.4177 re_mapping 0.0054 re_causal 0.0141 /// teacc 98.74 lr 0.00010000 +Epoch 378, weight, value: tensor([[ 0.0407, 0.1294, -0.1816, ..., -0.1001, 0.0750, 0.0037], + [-0.1018, -0.0993, 0.1177, ..., -0.0444, -0.1165, 0.0524], + [-0.0952, -0.0721, -0.1197, ..., -0.1365, -0.0218, 0.0632], + ..., + [-0.0134, -0.1309, -0.1370, ..., -0.1108, -0.0491, 0.0743], + [-0.1252, -0.1161, -0.1160, ..., -0.0690, -0.0541, -0.1014], + [ 0.0246, -0.0314, -0.0599, ..., -0.1014, 0.0182, -0.0247]], + device='cuda:0'), grad: tensor([[ 1.6794e-03, 7.7963e-04, 2.7132e-04, ..., 9.9480e-05, + 3.8600e-04, 9.4557e-04], + [ 6.9809e-04, 2.0814e-04, 2.8777e-04, ..., 2.0218e-04, + 2.7180e-04, 1.6460e-03], + [ 7.7486e-04, 1.8358e-04, 1.8597e-04, ..., -6.8188e-04, + 3.3879e-04, -9.1324e-03], + ..., + [ 8.1253e-04, 2.8849e-04, -5.1832e-04, ..., 9.5904e-05, + 2.1756e-04, 2.1591e-03], + [-1.0460e-02, -1.0056e-02, -8.6899e-03, ..., -3.4142e-03, + -8.7662e-03, -2.0504e-04], + [ 2.5311e-03, 1.3876e-03, 2.2149e-04, ..., 8.9467e-05, + 3.0684e-04, 8.5115e-04]], device='cuda:0') +Epoch 378, bias, value: tensor([ 0.0074, 0.0063, -0.0050, -0.0178, 0.0112, -0.0119, -0.0144, -0.0185, + -0.0030, 0.0050], device='cuda:0'), grad: tensor([ 0.0198, 0.0296, -0.0244, 0.0467, -0.0129, -0.0137, 0.0131, -0.0079, + -0.0782, 0.0279], device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 216.76, cls_loss 0.4893 cls_loss_mapping 0.0022 cls_loss_causal 0.4637 re_mapping 0.0053 re_causal 0.0142 /// teacc 98.98 lr 0.00010000 +Epoch 379, weight, value: tensor([[ 0.0398, 0.1288, -0.1819, ..., -0.1013, 0.0750, 0.0034], + [-0.1035, -0.0995, 0.1175, ..., -0.0442, -0.1176, 0.0533], + [-0.0955, -0.0727, -0.1210, ..., -0.1367, -0.0218, 0.0644], + ..., + [-0.0131, -0.1308, -0.1381, ..., -0.1114, -0.0502, 0.0729], + [-0.1257, -0.1153, -0.1150, ..., -0.0681, -0.0534, -0.1018], + [ 0.0255, -0.0311, -0.0588, ..., -0.1008, 0.0178, -0.0230]], + device='cuda:0'), grad: tensor([[ 3.1638e-04, 5.4777e-05, 1.2171e-04, ..., 1.4508e-04, + 2.0528e-04, 1.3518e-04], + [ 2.2516e-05, 4.9293e-05, 1.9455e-04, ..., 3.4183e-05, + 1.5542e-05, 2.8181e-04], + [ 2.6986e-05, 1.5035e-05, 1.1945e-04, ..., 2.6941e-05, + 2.2575e-05, 5.6696e-04], + ..., + [ 1.4611e-05, 1.7315e-05, 1.2779e-04, ..., 7.7337e-06, + 5.1633e-06, -3.2544e-04], + [ 6.6161e-05, 2.1553e-04, 1.7703e-04, ..., 1.8179e-04, + 1.0766e-05, 1.5771e-04], + [ 9.7811e-05, 6.1929e-05, -1.0290e-03, ..., 3.0607e-05, + 9.2447e-05, -1.5030e-03]], device='cuda:0') +Epoch 379, bias, value: tensor([ 0.0072, 0.0056, -0.0031, -0.0189, 0.0106, -0.0118, -0.0143, -0.0196, + -0.0022, 0.0059], device='cuda:0'), grad: tensor([ 0.0072, 0.0008, 0.0108, -0.0209, 0.0137, 0.0009, 0.0117, 0.0042, + -0.0032, -0.0251], device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 216.19, cls_loss 0.4998 cls_loss_mapping 0.0031 cls_loss_causal 0.4747 re_mapping 0.0050 re_causal 0.0135 /// teacc 98.97 lr 0.00010000 +Epoch 380, weight, value: tensor([[ 0.0407, 0.1292, -0.1816, ..., -0.1021, 0.0749, 0.0039], + [-0.1054, -0.1001, 0.1181, ..., -0.0447, -0.1199, 0.0543], + [-0.0955, -0.0738, -0.1195, ..., -0.1369, -0.0211, 0.0641], + ..., + [-0.0119, -0.1308, -0.1390, ..., -0.1115, -0.0509, 0.0730], + [-0.1261, -0.1158, -0.1154, ..., -0.0679, -0.0525, -0.1012], + [ 0.0251, -0.0309, -0.0589, ..., -0.1004, 0.0183, -0.0232]], + device='cuda:0'), grad: tensor([[-3.3550e-03, 4.3571e-05, 1.1575e-04, ..., 2.2620e-05, + 1.3828e-04, 7.0393e-05], + [-3.4630e-05, 2.1551e-06, 3.5577e-06, ..., 5.9046e-07, + 1.0263e-06, 1.2386e-04], + [ 4.8256e-04, -4.1509e-04, -1.0234e-04, ..., -2.2590e-04, + 1.0073e-05, -1.0605e-03], + ..., + [ 4.3344e-04, 4.6849e-05, 1.5929e-05, ..., 2.4855e-05, + 6.1728e-06, 1.8132e-04], + [ 5.5647e-04, -1.7233e-03, -4.6616e-03, ..., 1.6198e-05, + 1.3657e-05, 1.4532e-04], + [-2.9540e-04, 2.6878e-06, 5.2825e-06, ..., 2.7455e-06, + 2.2963e-05, -7.3528e-04]], device='cuda:0') +Epoch 380, bias, value: tensor([ 0.0079, 0.0054, -0.0030, -0.0187, 0.0096, -0.0122, -0.0154, -0.0190, + -0.0014, 0.0062], device='cuda:0'), grad: tensor([ 0.0018, -0.0211, 0.0065, 0.0116, 0.0130, 0.0254, 0.0167, -0.0181, + 0.0055, -0.0413], device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 216.15, cls_loss 0.4545 cls_loss_mapping 0.0031 cls_loss_causal 0.4284 re_mapping 0.0055 re_causal 0.0149 /// teacc 98.87 lr 0.00010000 +Epoch 381, weight, value: tensor([[ 0.0405, 0.1298, -0.1803, ..., -0.1019, 0.0744, 0.0042], + [-0.1038, -0.0979, 0.1193, ..., -0.0451, -0.1188, 0.0536], + [-0.0952, -0.0732, -0.1205, ..., -0.1364, -0.0208, 0.0644], + ..., + [-0.0116, -0.1307, -0.1374, ..., -0.1103, -0.0501, 0.0733], + [-0.1272, -0.1157, -0.1161, ..., -0.0683, -0.0529, -0.1023], + [ 0.0241, -0.0317, -0.0592, ..., -0.1016, 0.0166, -0.0227]], + device='cuda:0'), grad: tensor([[-1.4486e-03, 6.2704e-05, 3.9673e-04, ..., -1.5955e-03, + -2.8896e-04, -6.5756e-04], + [ 8.0645e-05, 1.9491e-04, -2.5425e-03, ..., 3.1781e-04, + 3.0017e-04, -9.8896e-04], + [ 5.1451e-04, 8.8835e-04, 8.4209e-04, ..., 4.2629e-04, + 1.0262e-03, 5.9662e-03], + ..., + [ 1.6320e-04, 2.9349e-04, 5.4312e-04, ..., 2.1100e-04, + 3.7456e-04, -2.9411e-03], + [ 8.9467e-05, 1.8466e-04, 4.2367e-04, ..., 2.2244e-04, + 2.5749e-04, 1.1883e-03], + [ 8.0943e-05, 2.1303e-04, -1.0939e-03, ..., 1.1301e-04, + 2.6655e-04, -2.0065e-03]], device='cuda:0') +Epoch 381, bias, value: tensor([ 0.0076, 0.0055, -0.0029, -0.0182, 0.0108, -0.0133, -0.0152, -0.0194, + -0.0015, 0.0062], device='cuda:0'), grad: tensor([-0.0254, 0.0073, 0.0459, -0.0081, 0.0172, 0.0219, -0.0057, 0.0088, + -0.0119, -0.0499], device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 216.21, cls_loss 0.4811 cls_loss_mapping 0.0031 cls_loss_causal 0.4626 re_mapping 0.0052 re_causal 0.0138 /// teacc 98.77 lr 0.00010000 +Epoch 382, weight, value: tensor([[ 0.0414, 0.1296, -0.1817, ..., -0.1023, 0.0742, 0.0039], + [-0.1050, -0.0989, 0.1195, ..., -0.0447, -0.1195, 0.0539], + [-0.0963, -0.0734, -0.1206, ..., -0.1375, -0.0202, 0.0629], + ..., + [-0.0115, -0.1309, -0.1370, ..., -0.1104, -0.0498, 0.0742], + [-0.1262, -0.1153, -0.1167, ..., -0.0679, -0.0487, -0.1018], + [ 0.0236, -0.0319, -0.0597, ..., -0.1020, 0.0149, -0.0232]], + device='cuda:0'), grad: tensor([[ 2.3603e-05, 7.4482e-04, 5.7745e-04, ..., 1.8108e-04, + 2.1362e-04, 1.1492e-03], + [ 6.6614e-04, 8.4102e-05, -2.8539e-04, ..., -3.2878e-04, + 3.8385e-04, 9.5224e-04], + [-5.7077e-04, 1.6222e-03, -5.4741e-04, ..., 3.5834e-04, + -1.4992e-03, -8.5754e-03], + ..., + [ 2.0084e-03, 7.0858e-04, 9.3937e-04, ..., 2.2554e-04, + 5.0497e-04, 2.8801e-03], + [ 2.2259e-03, 2.1534e-03, 1.0452e-03, ..., 5.2691e-04, + 2.0909e-04, 1.2589e-03], + [-3.2959e-03, -1.4000e-03, -4.0841e-04, ..., 1.0520e-04, + -2.5511e-04, -7.0801e-03]], device='cuda:0') +Epoch 382, bias, value: tensor([ 0.0080, 0.0056, -0.0044, -0.0182, 0.0110, -0.0135, -0.0143, -0.0197, + -0.0014, 0.0062], device='cuda:0'), grad: tensor([ 0.0082, 0.0095, -0.0396, -0.0049, 0.0045, 0.0237, 0.0087, 0.0096, + 0.0213, -0.0411], device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 216.43, cls_loss 0.4700 cls_loss_mapping 0.0023 cls_loss_causal 0.4422 re_mapping 0.0055 re_causal 0.0150 /// teacc 98.79 lr 0.00010000 +Epoch 383, weight, value: tensor([[ 0.0433, 0.1310, -0.1814, ..., -0.1039, 0.0754, 0.0033], + [-0.1054, -0.0995, 0.1187, ..., -0.0455, -0.1212, 0.0530], + [-0.0952, -0.0727, -0.1199, ..., -0.1380, -0.0187, 0.0627], + ..., + [-0.0108, -0.1310, -0.1366, ..., -0.1106, -0.0497, 0.0739], + [-0.1258, -0.1165, -0.1157, ..., -0.0679, -0.0487, -0.1011], + [ 0.0244, -0.0329, -0.0603, ..., -0.1010, 0.0171, -0.0231]], + device='cuda:0'), grad: tensor([[-5.3444e-03, 1.1015e-04, 4.0859e-05, ..., 1.0759e-04, + -2.0809e-03, -4.2953e-03], + [ 3.5954e-03, -1.3103e-03, 1.3202e-05, ..., -1.3838e-03, + 1.9753e-04, -1.7452e-03], + [ 1.8253e-03, 2.5320e-04, 1.8513e-04, ..., 1.1855e-04, + 1.5192e-03, 2.1458e-03], + ..., + [ 3.6335e-03, -1.3208e-04, -1.0496e-04, ..., 3.4153e-05, + 3.0589e-04, 6.0310e-03], + [ 4.6196e-03, 4.7469e-04, 5.1975e-05, ..., 5.1641e-04, + 1.4381e-03, 3.3722e-03], + [-8.4925e-04, -1.2505e-04, 3.7879e-05, ..., 2.3261e-05, + -2.8496e-03, -6.5765e-03]], device='cuda:0') +Epoch 383, bias, value: tensor([ 0.0078, 0.0045, -0.0032, -0.0185, 0.0105, -0.0141, -0.0132, -0.0192, + -0.0007, 0.0056], device='cuda:0'), grad: tensor([-0.0150, 0.0048, -0.0142, -0.0138, 0.0412, -0.0353, -0.0309, 0.0457, + 0.0427, -0.0253], device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 216.39, cls_loss 0.5184 cls_loss_mapping 0.0045 cls_loss_causal 0.4966 re_mapping 0.0053 re_causal 0.0146 /// teacc 98.89 lr 0.00010000 +Epoch 384, weight, value: tensor([[ 0.0428, 0.1306, -0.1799, ..., -0.1026, 0.0766, 0.0045], + [-0.1047, -0.0986, 0.1181, ..., -0.0463, -0.1185, 0.0521], + [-0.0950, -0.0720, -0.1191, ..., -0.1376, -0.0187, 0.0624], + ..., + [-0.0104, -0.1316, -0.1363, ..., -0.1113, -0.0493, 0.0733], + [-0.1262, -0.1162, -0.1152, ..., -0.0670, -0.0489, -0.1011], + [ 0.0241, -0.0347, -0.0612, ..., -0.1013, 0.0152, -0.0220]], + device='cuda:0'), grad: tensor([[ 2.0905e-03, -2.2089e-04, 1.1277e-04, ..., 1.2153e-04, + -1.2398e-03, 1.1988e-03], + [-9.3689e-03, -1.1721e-03, -2.1801e-03, ..., -3.1891e-03, + -6.9046e-04, -6.0997e-03], + [ 1.7433e-03, 8.0585e-05, 1.6522e-04, ..., 1.1301e-04, + 4.0627e-04, 2.8343e-03], + ..., + [ 2.2221e-03, 2.0957e-04, 4.4823e-03, ..., 5.2166e-04, + 1.0777e-03, 1.2383e-02], + [ 9.7179e-04, 8.3303e-04, 1.0300e-03, ..., 2.0218e-03, + 3.5954e-04, 4.0894e-03], + [-4.5013e-03, 1.4126e-04, -4.0779e-03, ..., 8.6188e-05, + -1.4153e-03, -2.3315e-02]], device='cuda:0') +Epoch 384, bias, value: tensor([ 0.0084, 0.0036, -0.0034, -0.0182, 0.0119, -0.0139, -0.0137, -0.0197, + -0.0014, 0.0059], device='cuda:0'), grad: tensor([-0.0449, -0.0163, 0.0387, 0.0019, 0.0273, 0.0170, 0.0212, -0.0154, + 0.0273, -0.0569], device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 216.18, cls_loss 0.4722 cls_loss_mapping 0.0018 cls_loss_causal 0.4473 re_mapping 0.0056 re_causal 0.0155 /// teacc 98.88 lr 0.00010000 +Epoch 385, weight, value: tensor([[ 0.0422, 0.1307, -0.1800, ..., -0.1026, 0.0765, 0.0048], + [-0.1026, -0.0993, 0.1172, ..., -0.0466, -0.1155, 0.0531], + [-0.0943, -0.0720, -0.1181, ..., -0.1375, -0.0173, 0.0622], + ..., + [-0.0105, -0.1322, -0.1365, ..., -0.1127, -0.0499, 0.0727], + [-0.1267, -0.1160, -0.1152, ..., -0.0681, -0.0501, -0.1012], + [ 0.0249, -0.0361, -0.0622, ..., -0.1003, 0.0149, -0.0213]], + device='cuda:0'), grad: tensor([[ 5.4407e-04, 3.0494e-04, 8.7166e-04, ..., 5.2806e-07, + 2.5249e-04, -4.4346e-04], + [-3.5343e-03, 6.5923e-05, -1.7262e-03, ..., 1.0571e-07, + 1.0335e-04, -5.1069e-04], + [ 6.6662e-04, 1.4019e-04, 1.0052e-03, ..., 4.9826e-08, + 1.5318e-04, -6.3229e-04], + ..., + [ 1.0414e-03, 1.1215e-02, -4.1199e-03, ..., 3.7719e-08, + 1.0210e-04, 6.3858e-03], + [ 7.9298e-04, 1.6630e-04, 1.0691e-03, ..., 1.7360e-06, + 3.1781e-04, 1.4114e-03], + [ 7.0155e-05, 3.8099e-04, 8.8835e-04, ..., 7.4506e-09, + 4.8351e-04, 1.1545e-04]], device='cuda:0') +Epoch 385, bias, value: tensor([ 0.0087, 0.0052, -0.0033, -0.0180, 0.0113, -0.0150, -0.0139, -0.0197, + -0.0018, 0.0059], device='cuda:0'), grad: tensor([-0.0008, -0.0683, 0.0050, 0.0028, -0.0156, 0.0173, -0.0219, 0.0339, + 0.0367, 0.0107], device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 216.39, cls_loss 0.4829 cls_loss_mapping 0.0024 cls_loss_causal 0.4546 re_mapping 0.0056 re_causal 0.0148 /// teacc 98.95 lr 0.00010000 +Epoch 386, weight, value: tensor([[ 0.0433, 0.1313, -0.1814, ..., -0.1034, 0.0763, 0.0045], + [-0.1034, -0.0995, 0.1177, ..., -0.0474, -0.1157, 0.0526], + [-0.0947, -0.0716, -0.1185, ..., -0.1379, -0.0167, 0.0613], + ..., + [-0.0099, -0.1333, -0.1358, ..., -0.1131, -0.0489, 0.0745], + [-0.1271, -0.1161, -0.1143, ..., -0.0675, -0.0503, -0.1009], + [ 0.0248, -0.0354, -0.0624, ..., -0.1009, 0.0163, -0.0226]], + device='cuda:0'), grad: tensor([[-3.3550e-03, 2.5120e-03, 4.2391e-04, ..., 9.0933e-04, + -2.7065e-03, -7.2670e-04], + [ 1.7536e-04, 2.0075e-04, 4.5538e-05, ..., 6.6698e-05, + 1.4031e-04, 1.7710e-03], + [ 5.2023e-04, 1.0729e-03, 2.2018e-04, ..., 1.9109e-04, + 3.3259e-04, -1.6613e-03], + ..., + [ 1.1024e-03, 2.8086e-04, 5.4389e-05, ..., 1.3518e-04, + 1.5342e-04, -1.1902e-03], + [ 7.1716e-04, 4.0913e-04, -6.8724e-05, ..., 7.6413e-05, + 2.2185e-04, 1.1625e-03], + [-1.0471e-03, 6.6042e-04, 1.3554e-04, ..., 2.1887e-04, + 2.0349e-04, -1.3418e-03]], device='cuda:0') +Epoch 386, bias, value: tensor([ 0.0088, 0.0049, -0.0042, -0.0192, 0.0114, -0.0137, -0.0128, -0.0191, + -0.0028, 0.0061], device='cuda:0'), grad: tensor([-0.0138, -0.0001, -0.0114, -0.0021, -0.0084, 0.0246, 0.0193, -0.0253, + 0.0200, -0.0028], device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 216.58, cls_loss 0.4776 cls_loss_mapping 0.0036 cls_loss_causal 0.4498 re_mapping 0.0054 re_causal 0.0145 /// teacc 98.85 lr 0.00010000 +Epoch 387, weight, value: tensor([[ 0.0417, 0.1313, -0.1820, ..., -0.1035, 0.0747, 0.0038], + [-0.1036, -0.0999, 0.1176, ..., -0.0474, -0.1141, 0.0536], + [-0.0953, -0.0718, -0.1191, ..., -0.1380, -0.0174, 0.0611], + ..., + [-0.0088, -0.1340, -0.1363, ..., -0.1129, -0.0500, 0.0748], + [-0.1265, -0.1162, -0.1148, ..., -0.0671, -0.0499, -0.0992], + [ 0.0238, -0.0356, -0.0605, ..., -0.1003, 0.0165, -0.0224]], + device='cuda:0'), grad: tensor([[-5.7697e-04, -3.1424e-04, 6.7139e-04, ..., 1.9908e-05, + -5.2261e-04, 4.6659e-04], + [ 3.9482e-04, 8.5711e-05, -1.0042e-03, ..., 4.1157e-05, + -5.9080e-04, -2.7599e-03], + [-2.1782e-03, -8.2064e-04, -1.4677e-03, ..., -4.2796e-04, + -1.5259e-04, 4.9305e-04], + ..., + [ 2.6684e-03, 7.0393e-05, 3.4733e-03, ..., 1.4625e-05, + 1.9646e-03, -2.2564e-03], + [ 4.4012e-04, 7.9989e-05, 1.7834e-03, ..., 3.9071e-05, + 2.6441e-04, 1.6022e-03], + [ 2.9373e-03, 3.8075e-04, 3.2825e-03, ..., 8.0109e-05, + 2.0161e-03, 4.7898e-04]], device='cuda:0') +Epoch 387, bias, value: tensor([ 0.0086, 0.0050, -0.0038, -0.0200, 0.0104, -0.0127, -0.0132, -0.0189, + -0.0024, 0.0064], device='cuda:0'), grad: tensor([-0.0110, -0.0177, -0.0037, 0.0267, -0.0323, 0.0226, -0.0319, 0.0131, + 0.0271, 0.0070], device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 216.76, cls_loss 0.4830 cls_loss_mapping 0.0030 cls_loss_causal 0.4644 re_mapping 0.0054 re_causal 0.0143 /// teacc 98.87 lr 0.00010000 +Epoch 388, weight, value: tensor([[ 0.0415, 0.1307, -0.1815, ..., -0.1041, 0.0752, 0.0023], + [-0.1039, -0.1006, 0.1166, ..., -0.0474, -0.1144, 0.0539], + [-0.0962, -0.0686, -0.1170, ..., -0.1385, -0.0160, 0.0607], + ..., + [-0.0088, -0.1332, -0.1365, ..., -0.1117, -0.0505, 0.0745], + [-0.1268, -0.1163, -0.1129, ..., -0.0672, -0.0504, -0.0986], + [ 0.0236, -0.0364, -0.0609, ..., -0.0999, 0.0157, -0.0212]], + device='cuda:0'), grad: tensor([[ 3.1173e-05, 2.2337e-05, 2.1434e-04, ..., 2.7940e-09, + -1.4856e-05, 4.7255e-04], + [ 2.8038e-04, 8.5056e-05, -4.5657e-04, ..., 0.0000e+00, + 4.3958e-07, -1.0147e-03], + [ 2.2864e-04, 2.6941e-04, 5.7411e-04, ..., 0.0000e+00, + 1.3292e-04, 3.5248e-03], + ..., + [ 3.5733e-05, 3.0339e-05, 2.0194e-04, ..., 0.0000e+00, + 1.6168e-06, 7.8917e-04], + [ 2.4438e-04, 6.1369e-04, 1.1396e-03, ..., 7.9162e-09, + 7.5176e-06, -3.5310e-04], + [ 2.9469e-04, 1.2410e-04, 3.2377e-04, ..., 0.0000e+00, + 8.0615e-06, 5.4836e-04]], device='cuda:0') +Epoch 388, bias, value: tensor([ 0.0076, 0.0056, -0.0033, -0.0195, 0.0086, -0.0116, -0.0142, -0.0192, + -0.0021, 0.0075], device='cuda:0'), grad: tensor([ 0.0142, -0.0553, 0.0237, -0.0459, 0.0092, -0.0106, 0.0075, 0.0181, + 0.0140, 0.0251], device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 216.08, cls_loss 0.4399 cls_loss_mapping 0.0030 cls_loss_causal 0.4173 re_mapping 0.0056 re_causal 0.0153 /// teacc 98.73 lr 0.00010000 +Epoch 389, weight, value: tensor([[ 0.0413, 0.1315, -0.1805, ..., -0.1041, 0.0753, 0.0038], + [-0.1035, -0.1006, 0.1160, ..., -0.0471, -0.1143, 0.0540], + [-0.0955, -0.0693, -0.1183, ..., -0.1380, -0.0158, 0.0597], + ..., + [-0.0090, -0.1336, -0.1354, ..., -0.1125, -0.0503, 0.0759], + [-0.1260, -0.1170, -0.1141, ..., -0.0682, -0.0500, -0.0975], + [ 0.0239, -0.0347, -0.0608, ..., -0.1004, 0.0152, -0.0225]], + device='cuda:0'), grad: tensor([[ 4.5824e-04, 3.4094e-05, 7.5531e-04, ..., 1.4162e-04, + 1.5128e-04, 1.1892e-03], + [ 2.6298e-04, 1.4938e-05, 2.6941e-04, ..., 4.8876e-05, + 1.3900e-04, 1.6241e-03], + [ 4.5061e-04, 1.2912e-05, 1.8752e-04, ..., 3.1292e-05, + 1.7095e-04, 1.1292e-03], + ..., + [ 7.2432e-04, 1.2182e-05, 2.3782e-04, ..., 4.1038e-05, + 1.6630e-04, -3.9520e-03], + [ 1.0586e-03, 2.7835e-05, 4.6706e-04, ..., 8.5771e-05, + 1.0020e-04, 1.3380e-03], + [-7.1831e-03, 2.6315e-05, 4.0793e-04, ..., 7.3910e-05, + -1.2836e-03, -4.0588e-03]], device='cuda:0') +Epoch 389, bias, value: tensor([ 0.0085, 0.0050, -0.0035, -0.0185, 0.0088, -0.0118, -0.0144, -0.0184, + -0.0023, 0.0062], device='cuda:0'), grad: tensor([ 0.0191, 0.0183, 0.0176, 0.0225, -0.0081, 0.0269, -0.0367, -0.0012, + 0.0194, -0.0778], device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 216.35, cls_loss 0.4548 cls_loss_mapping 0.0028 cls_loss_causal 0.4342 re_mapping 0.0056 re_causal 0.0150 /// teacc 98.77 lr 0.00010000 +Epoch 390, weight, value: tensor([[ 0.0413, 0.1314, -0.1808, ..., -0.1054, 0.0764, 0.0030], + [-0.1035, -0.1013, 0.1161, ..., -0.0465, -0.1141, 0.0550], + [-0.0961, -0.0697, -0.1186, ..., -0.1372, -0.0172, 0.0598], + ..., + [-0.0093, -0.1337, -0.1361, ..., -0.1131, -0.0513, 0.0752], + [-0.1262, -0.1187, -0.1146, ..., -0.0677, -0.0499, -0.0976], + [ 0.0230, -0.0357, -0.0617, ..., -0.1017, 0.0145, -0.0237]], + device='cuda:0'), grad: tensor([[-1.4849e-03, -1.6298e-03, 1.7351e-06, ..., 4.0308e-06, + -1.5316e-03, 1.2136e-04], + [ 5.9128e-04, 2.6077e-05, -4.3213e-05, ..., -3.9395e-07, + 1.4555e-04, 1.4191e-03], + [ 3.6240e-04, 8.4817e-05, 1.4743e-06, ..., 2.5034e-06, + 3.9749e-03, 7.2365e-03], + ..., + [ 5.0926e-04, 7.3075e-05, 2.9728e-06, ..., 3.6228e-07, + -3.7403e-03, -5.6725e-03], + [ 5.1355e-04, 1.2362e-04, 3.2187e-05, ..., 7.1563e-06, + 2.0432e-04, 8.5735e-04], + [ 8.8596e-04, 6.6185e-04, 1.9781e-06, ..., 3.1479e-07, + 7.2193e-04, 1.0290e-03]], device='cuda:0') +Epoch 390, bias, value: tensor([ 0.0084, 0.0055, -0.0037, -0.0177, 0.0096, -0.0121, -0.0147, -0.0189, + -0.0025, 0.0054], device='cuda:0'), grad: tensor([ 0.0004, 0.0096, 0.0215, -0.0490, 0.0205, 0.0083, 0.0136, -0.0221, + 0.0088, -0.0117], device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 216.57, cls_loss 0.4836 cls_loss_mapping 0.0038 cls_loss_causal 0.4508 re_mapping 0.0053 re_causal 0.0141 /// teacc 98.86 lr 0.00010000 +Epoch 391, weight, value: tensor([[ 0.0401, 0.1305, -0.1815, ..., -0.1056, 0.0767, 0.0028], + [-0.1042, -0.1014, 0.1169, ..., -0.0473, -0.1150, 0.0533], + [-0.0963, -0.0693, -0.1179, ..., -0.1375, -0.0178, 0.0594], + ..., + [-0.0090, -0.1337, -0.1369, ..., -0.1132, -0.0502, 0.0757], + [-0.1262, -0.1182, -0.1148, ..., -0.0675, -0.0495, -0.0961], + [ 0.0229, -0.0348, -0.0613, ..., -0.1013, 0.0139, -0.0234]], + device='cuda:0'), grad: tensor([[-3.7074e-04, -6.3992e-04, 2.9802e-04, ..., 1.3001e-06, + -9.3639e-05, -7.8773e-04], + [ 1.1849e-04, -2.7847e-07, -1.2922e-03, ..., -4.4256e-05, + -7.8440e-05, -1.1885e-04], + [ 1.5860e-03, 1.1605e-04, 4.6206e-04, ..., 1.8880e-05, + 7.8678e-04, 9.1696e-04], + ..., + [ 4.1771e-04, -6.3133e-04, -1.1492e-03, ..., 2.0526e-06, + -2.0874e-04, -2.7637e-03], + [-4.3983e-03, 5.1498e-05, 3.8671e-04, ..., 1.3024e-05, + -2.2011e-03, 1.0080e-03], + [-8.0109e-04, 7.1168e-05, -1.2989e-03, ..., 1.0328e-06, + 6.3419e-05, -3.3875e-03]], device='cuda:0') +Epoch 391, bias, value: tensor([ 0.0087, 0.0050, -0.0036, -0.0178, 0.0087, -0.0125, -0.0141, -0.0194, + -0.0025, 0.0069], device='cuda:0'), grad: tensor([-0.0250, -0.0173, 0.0254, 0.0420, 0.0138, 0.0143, 0.0118, -0.0077, + -0.0134, -0.0439], device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 219.03, cls_loss 0.4995 cls_loss_mapping 0.0037 cls_loss_causal 0.4718 re_mapping 0.0047 re_causal 0.0122 /// teacc 98.87 lr 0.00010000 +Epoch 392, weight, value: tensor([[ 0.0395, 0.1303, -0.1818, ..., -0.1060, 0.0751, 0.0041], + [-0.1047, -0.1025, 0.1161, ..., -0.0470, -0.1145, 0.0528], + [-0.0966, -0.0693, -0.1176, ..., -0.1379, -0.0183, 0.0591], + ..., + [-0.0082, -0.1337, -0.1385, ..., -0.1130, -0.0505, 0.0765], + [-0.1263, -0.1176, -0.1144, ..., -0.0676, -0.0490, -0.0981], + [ 0.0223, -0.0351, -0.0605, ..., -0.1004, 0.0144, -0.0234]], + device='cuda:0'), grad: tensor([[-6.8903e-04, 1.1152e-04, -1.0548e-03, ..., 2.3410e-05, + -4.2772e-04, -1.4458e-03], + [-4.6396e-04, -2.0349e-04, -7.9966e-04, ..., -2.6703e-04, + 3.8259e-06, -8.9722e-03], + [ 2.9993e-04, 2.2268e-04, 1.0548e-03, ..., 3.7402e-05, + 3.5435e-05, -1.2064e-04], + ..., + [-6.3553e-03, 3.5977e-04, -2.9488e-03, ..., 4.3184e-05, + 3.7462e-05, -2.5501e-03], + [ 1.0099e-03, 5.5170e-04, 1.3599e-03, ..., 4.8816e-05, + 1.6379e-04, 1.9608e-03], + [ 6.6452e-03, 7.6723e-04, 1.2388e-03, ..., 2.3544e-05, + 1.9681e-04, 4.5052e-03]], device='cuda:0') +Epoch 392, bias, value: tensor([ 0.0101, 0.0050, -0.0038, -0.0166, 0.0089, -0.0129, -0.0150, -0.0199, + -0.0043, 0.0080], device='cuda:0'), grad: tensor([-0.0128, -0.0226, 0.0286, -0.0517, 0.0206, 0.0131, 0.0093, -0.0540, + 0.0276, 0.0419], device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 216.52, cls_loss 0.4736 cls_loss_mapping 0.0020 cls_loss_causal 0.4513 re_mapping 0.0050 re_causal 0.0139 /// teacc 98.90 lr 0.00010000 +Epoch 393, weight, value: tensor([[ 0.0398, 0.1299, -0.1827, ..., -0.1054, 0.0749, 0.0041], + [-0.1051, -0.1027, 0.1160, ..., -0.0471, -0.1138, 0.0531], + [-0.0953, -0.0695, -0.1176, ..., -0.1375, -0.0183, 0.0589], + ..., + [-0.0093, -0.1328, -0.1383, ..., -0.1121, -0.0497, 0.0769], + [-0.1278, -0.1175, -0.1151, ..., -0.0679, -0.0497, -0.0980], + [ 0.0235, -0.0331, -0.0588, ..., -0.0986, 0.0139, -0.0228]], + device='cuda:0'), grad: tensor([[ 7.5996e-05, 4.9114e-04, 4.9025e-05, ..., -3.0160e-04, + 1.4193e-06, 7.9107e-04], + [ 6.7055e-05, 1.1122e-04, 1.5080e-04, ..., 7.4646e-07, + 3.6985e-05, 4.3201e-04], + [ 1.2951e-03, 8.5306e-04, 1.2612e-04, ..., 4.0606e-07, + 1.3165e-05, 4.4746e-03], + ..., + [-1.0347e-04, 8.0013e-04, 6.4790e-05, ..., 2.2259e-07, + 5.3681e-06, -6.1073e-03], + [ 5.4932e-04, 1.2827e-03, -2.1706e-03, ..., -5.2834e-03, + 8.5011e-06, 6.5279e-04], + [-2.5606e-04, 2.0294e-03, 1.0967e-04, ..., 4.6343e-06, + 1.0289e-05, 2.1896e-03]], device='cuda:0') +Epoch 393, bias, value: tensor([ 0.0095, 0.0049, -0.0031, -0.0172, 0.0085, -0.0150, -0.0139, -0.0181, + -0.0047, 0.0084], device='cuda:0'), grad: tensor([ 0.0074, 0.0068, 0.0255, -0.0542, -0.0230, 0.0368, 0.0424, -0.0292, + -0.0284, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 216.43, cls_loss 0.4626 cls_loss_mapping 0.0019 cls_loss_causal 0.4292 re_mapping 0.0055 re_causal 0.0157 /// teacc 98.91 lr 0.00010000 +Epoch 394, weight, value: tensor([[ 0.0411, 0.1300, -0.1834, ..., -0.1060, 0.0749, 0.0024], + [-0.1041, -0.1031, 0.1159, ..., -0.0480, -0.1134, 0.0552], + [-0.0961, -0.0703, -0.1174, ..., -0.1361, -0.0158, 0.0585], + ..., + [-0.0107, -0.1327, -0.1383, ..., -0.1116, -0.0501, 0.0775], + [-0.1298, -0.1181, -0.1159, ..., -0.0678, -0.0504, -0.0995], + [ 0.0235, -0.0337, -0.0583, ..., -0.0984, 0.0129, -0.0210]], + device='cuda:0'), grad: tensor([[-4.0627e-03, -3.4523e-04, 2.0719e-04, ..., 3.0845e-05, + -4.9782e-03, -4.8828e-03], + [ 5.4502e-04, 5.8025e-05, -3.8004e-04, ..., 4.7922e-05, + 7.4244e-04, 2.4307e-04], + [ 1.1444e-03, 1.8346e-04, -3.1090e-04, ..., 1.6022e-04, + 1.0624e-03, 6.6280e-04], + ..., + [ 4.8590e-04, 8.9288e-05, 3.7599e-04, ..., 6.1572e-05, + 2.3091e-04, 6.3562e-04], + [ 6.8235e-04, 1.1843e-04, 3.3569e-04, ..., 9.9123e-05, + 3.2616e-04, 7.9393e-04], + [ 1.5008e-04, 2.9159e-04, 4.0674e-04, ..., 2.3603e-05, + 3.0065e-04, 7.5674e-04]], device='cuda:0') +Epoch 394, bias, value: tensor([ 0.0090, 0.0066, -0.0029, -0.0181, 0.0081, -0.0148, -0.0142, -0.0181, + -0.0050, 0.0088], device='cuda:0'), grad: tensor([-0.0088, -0.0159, -0.0114, -0.0405, 0.0126, 0.0168, 0.0061, 0.0145, + 0.0143, 0.0123], device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 216.11, cls_loss 0.4690 cls_loss_mapping 0.0030 cls_loss_causal 0.4529 re_mapping 0.0058 re_causal 0.0155 /// teacc 98.67 lr 0.00010000 +Epoch 395, weight, value: tensor([[ 0.0399, 0.1300, -0.1836, ..., -0.1063, 0.0753, 0.0026], + [-0.1064, -0.1036, 0.1150, ..., -0.0483, -0.1139, 0.0550], + [-0.0965, -0.0701, -0.1166, ..., -0.1358, -0.0159, 0.0594], + ..., + [-0.0091, -0.1328, -0.1401, ..., -0.1124, -0.0513, 0.0776], + [-0.1299, -0.1187, -0.1166, ..., -0.0674, -0.0494, -0.1000], + [ 0.0220, -0.0347, -0.0579, ..., -0.0979, 0.0120, -0.0220]], + device='cuda:0'), grad: tensor([[ 4.0550e-03, 3.0443e-05, 1.4424e-04, ..., 1.5140e-04, + 3.1114e-04, -7.5579e-04], + [-7.5645e-03, 1.1273e-05, 9.3365e-04, ..., -4.6349e-03, + 4.4644e-05, -8.1100e-03], + [ 1.5144e-03, 2.4185e-05, 2.5010e-04, ..., 3.3945e-05, + 1.0037e-04, 1.5945e-03], + ..., + [ 1.1253e-03, 8.5688e-04, 2.8648e-03, ..., 1.8105e-05, + 3.3826e-05, 1.9035e-03], + [ 1.5125e-03, 1.2553e-04, 1.8539e-03, ..., 1.2457e-04, + 1.4651e-04, 1.9131e-03], + [ 1.5419e-02, -9.3842e-04, -1.0735e-02, ..., 2.3365e-05, + 2.9445e-05, -3.1509e-03]], device='cuda:0') +Epoch 395, bias, value: tensor([ 0.0087, 0.0062, -0.0023, -0.0173, 0.0085, -0.0150, -0.0150, -0.0178, + -0.0046, 0.0080], device='cuda:0'), grad: tensor([-0.0019, -0.0087, 0.0153, -0.0089, -0.0023, 0.0173, -0.0049, 0.0243, + 0.0203, -0.0504], device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 216.78, cls_loss 0.4854 cls_loss_mapping 0.0029 cls_loss_causal 0.4674 re_mapping 0.0054 re_causal 0.0146 /// teacc 98.86 lr 0.00010000 +Epoch 396, weight, value: tensor([[ 0.0390, 0.1300, -0.1825, ..., -0.1065, 0.0753, 0.0021], + [-0.1077, -0.1040, 0.1171, ..., -0.0479, -0.1148, 0.0562], + [-0.0976, -0.0711, -0.1170, ..., -0.1357, -0.0163, 0.0597], + ..., + [-0.0078, -0.1336, -0.1416, ..., -0.1130, -0.0514, 0.0772], + [-0.1314, -0.1190, -0.1178, ..., -0.0678, -0.0496, -0.1011], + [ 0.0216, -0.0345, -0.0576, ..., -0.0983, 0.0115, -0.0214]], + device='cuda:0'), grad: tensor([[ 8.1301e-05, -2.6554e-05, 2.8014e-04, ..., 7.0512e-05, + -2.6310e-07, 2.5630e-04], + [ 1.2720e-04, 6.7754e-07, 4.3201e-04, ..., 1.0186e-04, + 1.0310e-06, 3.9077e-04], + [ 8.4102e-05, 2.7437e-06, 2.8706e-04, ..., 1.9217e-03, + 1.4203e-06, 7.6628e-04], + ..., + [ 4.3344e-04, 5.4687e-06, 3.4714e-04, ..., 1.3456e-05, + 3.0026e-05, 2.9049e-03], + [ 8.8871e-05, 2.5466e-05, 3.0565e-04, ..., 2.2614e-04, + 2.8517e-06, 5.4598e-04], + [-7.5645e-03, -8.8882e-04, -1.1358e-03, ..., 2.7359e-05, + -5.1956e-03, 3.5501e-04]], device='cuda:0') +Epoch 396, bias, value: tensor([ 0.0085, 0.0065, -0.0031, -0.0182, 0.0080, -0.0147, -0.0143, -0.0178, + -0.0041, 0.0084], device='cuda:0'), grad: tensor([ 0.0077, 0.0099, 0.0149, 0.0096, -0.0375, 0.0073, -0.0066, -0.0136, + 0.0134, -0.0051], device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 216.17, cls_loss 0.4741 cls_loss_mapping 0.0024 cls_loss_causal 0.4581 re_mapping 0.0051 re_causal 0.0145 /// teacc 98.70 lr 0.00010000 +Epoch 397, weight, value: tensor([[ 0.0383, 0.1306, -0.1819, ..., -0.1062, 0.0758, 0.0033], + [-0.1077, -0.1043, 0.1165, ..., -0.0480, -0.1150, 0.0563], + [-0.0970, -0.0715, -0.1174, ..., -0.1356, -0.0161, 0.0590], + ..., + [-0.0081, -0.1337, -0.1411, ..., -0.1140, -0.0495, 0.0768], + [-0.1325, -0.1175, -0.1181, ..., -0.0681, -0.0501, -0.1020], + [ 0.0221, -0.0337, -0.0580, ..., -0.0998, 0.0123, -0.0226]], + device='cuda:0'), grad: tensor([[ 6.6996e-04, -4.4763e-05, 4.1652e-04, ..., 7.5102e-05, + 4.3929e-05, 2.2185e-04], + [ 8.2302e-04, 1.0699e-04, 4.8161e-04, ..., 5.5343e-05, + 4.8518e-05, 2.4052e-03], + [-1.5392e-03, 9.0659e-05, 3.7241e-04, ..., 1.2785e-05, + -4.4346e-04, -2.2068e-03], + ..., + [-2.7504e-03, 1.3605e-05, 6.7425e-04, ..., 5.0999e-06, + 5.2452e-05, 9.7132e-04], + [ 8.5258e-04, 1.9188e-03, 1.8291e-03, ..., 7.0095e-04, + 4.2230e-05, 1.0091e-04], + [ 9.6512e-03, 1.5450e-04, 6.5041e-04, ..., 3.6776e-05, + 6.1989e-05, 6.1226e-04]], device='cuda:0') +Epoch 397, bias, value: tensor([ 0.0085, 0.0064, -0.0028, -0.0191, 0.0083, -0.0141, -0.0152, -0.0171, + -0.0041, 0.0082], device='cuda:0'), grad: tensor([ 0.0119, 0.0219, -0.0228, -0.1191, 0.0017, 0.0197, 0.0093, 0.0060, + 0.0246, 0.0468], device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 216.25, cls_loss 0.4627 cls_loss_mapping 0.0032 cls_loss_causal 0.4436 re_mapping 0.0052 re_causal 0.0143 /// teacc 98.87 lr 0.00010000 +Epoch 398, weight, value: tensor([[ 0.0387, 0.1304, -0.1814, ..., -0.1066, 0.0753, 0.0041], + [-0.1089, -0.1035, 0.1178, ..., -0.0479, -0.1136, 0.0569], + [-0.0949, -0.0713, -0.1182, ..., -0.1371, -0.0165, 0.0583], + ..., + [-0.0094, -0.1341, -0.1410, ..., -0.1128, -0.0506, 0.0774], + [-0.1313, -0.1158, -0.1186, ..., -0.0672, -0.0483, -0.1031], + [ 0.0215, -0.0335, -0.0575, ..., -0.1007, 0.0122, -0.0236]], + device='cuda:0'), grad: tensor([[ 1.6937e-03, 6.2037e-04, 1.2035e-03, ..., 7.6914e-04, + 7.3385e-04, 9.7275e-04], + [-9.0122e-04, 1.4126e-04, -2.1248e-03, ..., 9.4116e-05, + -4.0948e-05, -2.5330e-03], + [ 2.0580e-03, 4.1533e-04, 1.1721e-03, ..., 2.3198e-04, + 2.2354e-03, 1.0323e-02], + ..., + [ 1.6632e-03, 1.6356e-04, 7.5626e-04, ..., 8.5771e-05, + -7.7152e-04, -7.1983e-03], + [-8.7357e-04, 1.1390e-04, -3.0923e-04, ..., -1.2599e-05, + -9.5940e-04, 1.3819e-03], + [ 7.7724e-04, -3.0518e-04, -1.1206e-05, ..., 7.4387e-05, + 5.9795e-04, 4.4870e-04]], device='cuda:0') +Epoch 398, bias, value: tensor([ 0.0090, 0.0064, -0.0029, -0.0185, 0.0095, -0.0155, -0.0152, -0.0169, + -0.0043, 0.0075], device='cuda:0'), grad: tensor([ 0.0229, -0.0272, 0.0400, 0.0034, 0.0030, -0.0426, -0.0082, 0.0108, + 0.0093, -0.0114], device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 216.34, cls_loss 0.4601 cls_loss_mapping 0.0026 cls_loss_causal 0.4396 re_mapping 0.0051 re_causal 0.0141 /// teacc 98.98 lr 0.00010000 +Epoch 399, weight, value: tensor([[ 0.0384, 0.1309, -0.1821, ..., -0.1068, 0.0753, 0.0040], + [-0.1079, -0.1050, 0.1171, ..., -0.0476, -0.1147, 0.0579], + [-0.0938, -0.0724, -0.1184, ..., -0.1371, -0.0165, 0.0569], + ..., + [-0.0099, -0.1344, -0.1405, ..., -0.1130, -0.0510, 0.0772], + [-0.1312, -0.1168, -0.1198, ..., -0.0674, -0.0481, -0.1015], + [ 0.0221, -0.0329, -0.0567, ..., -0.1020, 0.0122, -0.0242]], + device='cuda:0'), grad: tensor([[ 1.9989e-03, 8.4972e-04, 9.0408e-04, ..., 1.4830e-03, + -8.0913e-06, 4.1428e-03], + [-2.8648e-03, -1.1034e-03, -1.6041e-03, ..., -1.7118e-03, + 3.0454e-07, -6.8474e-03], + [ 1.7643e-04, 9.9838e-06, 3.7742e-04, ..., 1.4372e-05, + 1.0189e-06, 7.6962e-04], + ..., + [-3.7313e-04, 3.4999e-06, -5.8126e-04, ..., 4.7013e-06, + 6.8359e-07, 1.3342e-03], + [ 5.1308e-04, 1.4174e-04, 8.3971e-04, ..., 2.2018e-04, + 7.3528e-07, 1.5373e-03], + [ 1.9944e-04, 1.0401e-05, -7.9298e-04, ..., 3.3975e-05, + 3.1199e-06, -3.8605e-03]], device='cuda:0') +Epoch 399, bias, value: tensor([ 0.0089, 0.0074, -0.0038, -0.0194, 0.0086, -0.0137, -0.0146, -0.0171, + -0.0051, 0.0080], device='cuda:0'), grad: tensor([-0.0060, -0.0012, 0.0126, 0.0148, 0.0172, 0.0140, -0.0182, -0.0124, + 0.0176, -0.0385], device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 216.67, cls_loss 0.4950 cls_loss_mapping 0.0035 cls_loss_causal 0.4777 re_mapping 0.0050 re_causal 0.0132 /// teacc 98.86 lr 0.00010000 +Epoch 400, weight, value: tensor([[ 0.0386, 0.1307, -0.1822, ..., -0.1067, 0.0756, 0.0038], + [-0.1080, -0.1062, 0.1175, ..., -0.0479, -0.1156, 0.0575], + [-0.0958, -0.0721, -0.1188, ..., -0.1370, -0.0169, 0.0567], + ..., + [-0.0098, -0.1347, -0.1415, ..., -0.1138, -0.0505, 0.0774], + [-0.1324, -0.1170, -0.1198, ..., -0.0674, -0.0480, -0.1021], + [ 0.0230, -0.0326, -0.0564, ..., -0.1027, 0.0120, -0.0245]], + device='cuda:0'), grad: tensor([[ 2.1286e-03, 4.4560e-04, -6.5727e-03, ..., 2.4343e-04, + 2.6250e-04, 2.1591e-03], + [ 7.7391e-04, 1.4019e-04, 1.0967e-03, ..., 1.1486e-04, + 1.1140e-04, 2.6321e-03], + [ 9.0551e-04, 1.9026e-04, 7.7534e-04, ..., 1.3196e-04, + 1.1700e-04, -4.7798e-03], + ..., + [ 8.8453e-04, 1.5175e-04, 9.2697e-04, ..., 4.1515e-05, + 1.0818e-04, 1.5583e-03], + [ 3.2878e-04, 2.3150e-04, -1.8466e-04, ..., -9.6655e-04, + 1.3292e-04, -1.2980e-03], + [-1.0824e-03, -7.8678e-04, 9.7036e-04, ..., 3.9846e-05, + 9.4354e-05, -1.1034e-03]], device='cuda:0') +Epoch 400, bias, value: tensor([ 0.0088, 0.0073, -0.0040, -0.0186, 0.0078, -0.0123, -0.0140, -0.0183, + -0.0052, 0.0080], device='cuda:0'), grad: tensor([ 0.0083, 0.0368, -0.0193, -0.0897, -0.0030, 0.0383, 0.0162, 0.0363, + 0.0091, -0.0331], device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 216.56, cls_loss 0.4681 cls_loss_mapping 0.0023 cls_loss_causal 0.4455 re_mapping 0.0051 re_causal 0.0136 /// teacc 98.85 lr 0.00001000 +Epoch 401, weight, value: tensor([[ 0.0381, 0.1307, -0.1819, ..., -0.1074, 0.0762, 0.0037], + [-0.1084, -0.1055, 0.1167, ..., -0.0470, -0.1144, 0.0568], + [-0.0971, -0.0722, -0.1185, ..., -0.1366, -0.0172, 0.0560], + ..., + [-0.0095, -0.1343, -0.1411, ..., -0.1143, -0.0509, 0.0772], + [-0.1329, -0.1158, -0.1184, ..., -0.0665, -0.0475, -0.1010], + [ 0.0221, -0.0323, -0.0568, ..., -0.1033, 0.0116, -0.0234]], + device='cuda:0'), grad: tensor([[-1.1848e-02, -8.2321e-03, -4.2992e-03, ..., -7.9930e-05, + -5.5580e-03, 4.5821e-06], + [-1.3103e-03, 1.9789e-05, 1.3494e-04, ..., 1.8487e-06, + 2.7943e-04, -2.6054e-03], + [ 6.3562e-04, 3.3903e-04, 3.4642e-04, ..., 5.2638e-06, + 5.6839e-04, -1.3375e-04], + ..., + [ 2.6188e-03, 7.0035e-05, 2.0301e-04, ..., 8.6473e-07, + 6.1893e-04, 2.5826e-03], + [ 1.0433e-03, 2.0516e-04, 2.7180e-04, ..., 8.9779e-06, + 5.7793e-04, 3.4809e-05], + [-2.5883e-03, 2.4021e-04, -2.6779e-03, ..., 3.1702e-06, + 7.1383e-04, -2.5916e-04]], device='cuda:0') +Epoch 401, bias, value: tensor([ 0.0085, 0.0063, -0.0046, -0.0173, 0.0077, -0.0133, -0.0137, -0.0178, + -0.0054, 0.0089], device='cuda:0'), grad: tensor([-0.0344, 0.0029, -0.0034, 0.0239, 0.0217, 0.0171, -0.0218, 0.0129, + 0.0139, -0.0328], device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 216.45, cls_loss 0.4730 cls_loss_mapping 0.0015 cls_loss_causal 0.4433 re_mapping 0.0051 re_causal 0.0144 /// teacc 98.87 lr 0.00001000 +Epoch 402, weight, value: tensor([[ 0.0383, 0.1309, -0.1819, ..., -0.1074, 0.0764, 0.0038], + [-0.1085, -0.1055, 0.1168, ..., -0.0470, -0.1145, 0.0568], + [-0.0969, -0.0722, -0.1185, ..., -0.1366, -0.0172, 0.0561], + ..., + [-0.0095, -0.1342, -0.1411, ..., -0.1142, -0.0510, 0.0772], + [-0.1329, -0.1160, -0.1185, ..., -0.0666, -0.0476, -0.1011], + [ 0.0220, -0.0323, -0.0569, ..., -0.1033, 0.0114, -0.0235]], + device='cuda:0'), grad: tensor([[-1.1225e-03, -2.7390e-03, -2.3975e-03, ..., -1.8530e-03, + -1.3018e-03, -5.9938e-04], + [ 1.0653e-03, 1.5831e-04, 2.4676e-04, ..., 1.3709e-04, + 7.6473e-05, 7.1478e-04], + [-1.3208e-03, 6.6519e-05, -4.1924e-03, ..., -2.7800e-04, + 3.7849e-05, -1.4982e-03], + ..., + [-5.9700e-03, 3.9220e-05, 2.1279e-04, ..., 3.3259e-05, + 1.9178e-05, -7.5483e-04], + [ 7.7391e-04, 2.5630e-04, 4.1604e-04, ..., 2.1195e-04, + 1.2219e-04, -2.3174e-03], + [ 1.4186e-04, 1.9968e-04, 4.3726e-04, ..., 1.7023e-04, + 1.0115e-04, -1.6165e-04]], device='cuda:0') +Epoch 402, bias, value: tensor([ 0.0085, 0.0063, -0.0045, -0.0171, 0.0076, -0.0134, -0.0137, -0.0178, + -0.0054, 0.0088], device='cuda:0'), grad: tensor([ 0.0051, 0.0061, -0.0455, 0.0299, 0.0161, 0.0241, 0.0202, -0.0430, + -0.0189, 0.0058], device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 216.40, cls_loss 0.4249 cls_loss_mapping 0.0014 cls_loss_causal 0.4040 re_mapping 0.0049 re_causal 0.0136 /// teacc 98.89 lr 0.00001000 +Epoch 403, weight, value: tensor([[ 0.0384, 0.1309, -0.1817, ..., -0.1074, 0.0765, 0.0038], + [-0.1086, -0.1055, 0.1169, ..., -0.0470, -0.1147, 0.0568], + [-0.0969, -0.0721, -0.1186, ..., -0.1365, -0.0170, 0.0562], + ..., + [-0.0094, -0.1342, -0.1411, ..., -0.1142, -0.0508, 0.0771], + [-0.1329, -0.1161, -0.1186, ..., -0.0666, -0.0475, -0.1010], + [ 0.0219, -0.0322, -0.0569, ..., -0.1034, 0.0113, -0.0237]], + device='cuda:0'), grad: tensor([[ 3.4165e-04, -1.1545e-04, 5.3495e-05, ..., 2.2948e-05, + -1.1399e-06, -1.1361e-04], + [ 8.5115e-04, 1.0192e-05, 3.4153e-05, ..., 8.6352e-06, + 7.8380e-05, 1.0929e-03], + [ 8.2684e-04, -3.8296e-05, -2.5630e-04, ..., -6.4731e-05, + 3.7003e-04, 5.0068e-04], + ..., + [-1.7052e-03, 3.9823e-06, 1.2532e-05, ..., 3.0156e-06, + 6.1035e-05, -4.1199e-03], + [ 3.6073e-04, 3.5107e-05, 9.2685e-05, ..., 2.9951e-05, + 1.1468e-04, 5.1689e-04], + [ 2.5196e-03, 8.8155e-05, 6.3516e-06, ..., 1.8505e-06, + 1.9050e-04, 4.7531e-03]], device='cuda:0') +Epoch 403, bias, value: tensor([ 0.0086, 0.0063, -0.0044, -0.0171, 0.0075, -0.0135, -0.0137, -0.0178, + -0.0055, 0.0087], device='cuda:0'), grad: tensor([-0.0198, 0.0050, 0.0110, 0.0092, -0.0186, -0.0433, 0.0397, -0.0079, + 0.0161, 0.0086], device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 216.43, cls_loss 0.4861 cls_loss_mapping 0.0012 cls_loss_causal 0.4616 re_mapping 0.0048 re_causal 0.0137 /// teacc 98.93 lr 0.00001000 +Epoch 404, weight, value: tensor([[ 0.0385, 0.1309, -0.1817, ..., -0.1076, 0.0767, 0.0039], + [-0.1085, -0.1056, 0.1167, ..., -0.0470, -0.1148, 0.0568], + [-0.0969, -0.0721, -0.1186, ..., -0.1364, -0.0172, 0.0560], + ..., + [-0.0093, -0.1343, -0.1412, ..., -0.1142, -0.0509, 0.0773], + [-0.1329, -0.1161, -0.1184, ..., -0.0667, -0.0474, -0.1011], + [ 0.0217, -0.0320, -0.0567, ..., -0.1034, 0.0112, -0.0239]], + device='cuda:0'), grad: tensor([[ 5.7125e-04, 3.8207e-05, 3.1853e-04, ..., 5.9128e-05, + 4.0859e-05, 1.1930e-03], + [ 5.3978e-04, 1.3083e-05, -1.2732e-04, ..., -2.3448e-04, + 1.0794e-04, 7.8201e-04], + [ 5.0783e-04, 3.2306e-05, -5.7983e-04, ..., 4.2915e-05, + 4.7922e-05, 5.4073e-04], + ..., + [-2.4681e-03, 7.8797e-05, 5.8317e-04, ..., 4.0501e-05, + 2.0385e-04, -4.4670e-03], + [ 8.3256e-04, 1.0595e-03, 5.8317e-04, ..., 1.2624e-04, + 1.4710e-04, 9.1505e-04], + [-4.5729e-04, -6.5088e-05, 6.9737e-05, ..., 6.3121e-05, + -3.1543e-04, 8.6641e-04]], device='cuda:0') +Epoch 404, bias, value: tensor([ 0.0087, 0.0063, -0.0046, -0.0172, 0.0075, -0.0135, -0.0135, -0.0178, + -0.0055, 0.0087], device='cuda:0'), grad: tensor([ 0.0129, 0.0144, -0.0078, 0.0012, -0.0154, -0.0048, -0.0139, -0.0165, + 0.0167, 0.0132], device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 216.29, cls_loss 0.4843 cls_loss_mapping 0.0014 cls_loss_causal 0.4596 re_mapping 0.0046 re_causal 0.0135 /// teacc 98.91 lr 0.00001000 +Epoch 405, weight, value: tensor([[ 0.0386, 0.1308, -0.1815, ..., -0.1076, 0.0768, 0.0040], + [-0.1086, -0.1057, 0.1167, ..., -0.0469, -0.1149, 0.0568], + [-0.0969, -0.0721, -0.1186, ..., -0.1364, -0.0173, 0.0560], + ..., + [-0.0092, -0.1344, -0.1412, ..., -0.1141, -0.0510, 0.0773], + [-0.1329, -0.1162, -0.1184, ..., -0.0667, -0.0474, -0.1012], + [ 0.0216, -0.0320, -0.0568, ..., -0.1034, 0.0112, -0.0238]], + device='cuda:0'), grad: tensor([[-4.7296e-05, 1.1642e-06, 3.7766e-04, ..., 2.1741e-05, + 3.0875e-05, 7.3719e-04], + [-2.7657e-03, 1.9707e-06, 5.4789e-04, ..., -1.3793e-04, + 3.6567e-05, -5.7755e-03], + [ 4.4465e-04, 3.4541e-05, 5.7077e-04, ..., 4.8503e-06, + 4.6432e-05, 8.9884e-04], + ..., + [ 5.3024e-04, 1.2564e-06, -1.0960e-05, ..., 1.3269e-05, + 4.2230e-05, 5.4932e-04], + [ 6.3562e-04, 6.2510e-06, 4.7851e-04, ..., 2.5615e-05, + 3.0845e-05, 1.1959e-03], + [-1.0750e-02, 1.2442e-06, 8.8167e-04, ..., 1.2130e-05, + -7.4692e-03, 2.5043e-03]], device='cuda:0') +Epoch 405, bias, value: tensor([ 0.0089, 0.0063, -0.0047, -0.0171, 0.0076, -0.0136, -0.0135, -0.0178, + -0.0056, 0.0087], device='cuda:0'), grad: tensor([-0.0390, -0.0228, 0.0145, 0.0156, 0.0018, 0.0302, -0.0087, -0.0117, + 0.0150, 0.0053], device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 216.04, cls_loss 0.4775 cls_loss_mapping 0.0013 cls_loss_causal 0.4538 re_mapping 0.0044 re_causal 0.0128 /// teacc 98.96 lr 0.00001000 +Epoch 406, weight, value: tensor([[ 0.0387, 0.1308, -0.1813, ..., -0.1075, 0.0770, 0.0040], + [-0.1088, -0.1059, 0.1168, ..., -0.0470, -0.1149, 0.0566], + [-0.0969, -0.0721, -0.1185, ..., -0.1363, -0.0174, 0.0561], + ..., + [-0.0093, -0.1345, -0.1413, ..., -0.1142, -0.0510, 0.0774], + [-0.1329, -0.1161, -0.1183, ..., -0.0668, -0.0475, -0.1012], + [ 0.0217, -0.0319, -0.0568, ..., -0.1036, 0.0113, -0.0238]], + device='cuda:0'), grad: tensor([[ 8.6927e-04, -9.5320e-04, 3.6740e-04, ..., 3.7819e-05, + 6.1607e-04, 8.4448e-04], + [-2.3139e-04, 2.8610e-04, 2.6345e-04, ..., 2.9686e-07, + 1.1164e-04, -1.6317e-03], + [ 7.3528e-04, 2.1207e-04, 2.9731e-04, ..., 5.3644e-07, + 4.1175e-04, 5.9509e-04], + ..., + [ 7.6447e-03, 3.7074e-04, -1.8873e-03, ..., 9.4529e-08, + 1.1263e-03, 3.4218e-03], + [-6.8855e-04, -8.8406e-04, -4.6897e-04, ..., 4.1686e-06, + -2.0790e-03, -9.2983e-04], + [ 3.1624e-03, 3.4404e-04, 2.9635e-04, ..., 7.5903e-07, + -8.0919e-04, 2.4331e-04]], device='cuda:0') +Epoch 406, bias, value: tensor([ 0.0090, 0.0062, -0.0046, -0.0171, 0.0075, -0.0137, -0.0135, -0.0178, + -0.0055, 0.0087], device='cuda:0'), grad: tensor([ 0.0131, -0.0184, 0.0100, -0.0128, -0.0155, 0.0141, 0.0030, 0.0123, + -0.0199, 0.0140], device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 216.63, cls_loss 0.4276 cls_loss_mapping 0.0010 cls_loss_causal 0.4047 re_mapping 0.0044 re_causal 0.0122 /// teacc 99.00 lr 0.00001000 +Epoch 407, weight, value: tensor([[ 0.0388, 0.1309, -0.1813, ..., -0.1075, 0.0771, 0.0042], + [-0.1085, -0.1059, 0.1169, ..., -0.0469, -0.1151, 0.0566], + [-0.0970, -0.0721, -0.1185, ..., -0.1363, -0.0174, 0.0560], + ..., + [-0.0090, -0.1345, -0.1413, ..., -0.1143, -0.0512, 0.0773], + [-0.1329, -0.1161, -0.1181, ..., -0.0667, -0.0475, -0.1011], + [ 0.0217, -0.0319, -0.0568, ..., -0.1037, 0.0113, -0.0237]], + device='cuda:0'), grad: tensor([[ 1.6470e-03, 9.2506e-04, -8.0490e-04, ..., 0.0000e+00, + 4.7851e-04, 5.7507e-04], + [-1.0490e-03, 4.0889e-05, 4.7188e-03, ..., 0.0000e+00, + 2.2843e-05, 3.9215e-03], + [-4.2610e-03, 1.1462e-04, -4.6234e-03, ..., 0.0000e+00, + -3.7594e-03, -7.5569e-03], + ..., + [-7.4806e-03, -6.9504e-03, 5.6887e-04, ..., 0.0000e+00, + -5.4932e-04, 1.1301e-03], + [ 1.8072e-03, 7.8249e-04, 6.7616e-04, ..., 0.0000e+00, + 8.2791e-05, 6.6137e-04], + [ 1.5182e-03, 3.9983e-04, 6.3896e-04, ..., 0.0000e+00, + 7.8630e-04, -3.6068e-03]], device='cuda:0') +Epoch 407, bias, value: tensor([ 0.0092, 0.0064, -0.0047, -0.0171, 0.0073, -0.0137, -0.0137, -0.0179, + -0.0053, 0.0088], device='cuda:0'), grad: tensor([-0.0084, 0.0049, -0.0601, 0.0052, 0.0041, 0.0389, -0.0045, 0.0033, + 0.0025, 0.0141], device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 216.28, cls_loss 0.4390 cls_loss_mapping 0.0010 cls_loss_causal 0.4145 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.00 lr 0.00001000 +Epoch 408, weight, value: tensor([[ 0.0387, 0.1310, -0.1814, ..., -0.1075, 0.0770, 0.0042], + [-0.1085, -0.1058, 0.1170, ..., -0.0470, -0.1150, 0.0567], + [-0.0970, -0.0721, -0.1185, ..., -0.1363, -0.0174, 0.0560], + ..., + [-0.0091, -0.1344, -0.1413, ..., -0.1145, -0.0513, 0.0773], + [-0.1329, -0.1163, -0.1182, ..., -0.0669, -0.0476, -0.1011], + [ 0.0217, -0.0320, -0.0569, ..., -0.1039, 0.0113, -0.0236]], + device='cuda:0'), grad: tensor([[-1.1759e-03, -5.1446e-06, -2.2659e-03, ..., 1.1820e-04, + 2.7108e-04, -1.1368e-03], + [-7.6246e-04, -1.8207e-06, -4.9896e-03, ..., -2.6779e-03, + 1.5426e-04, -1.0727e-02], + [-8.4543e-04, 3.1404e-06, 1.5697e-03, ..., 9.6464e-04, + -2.1935e-03, -1.8053e-03], + ..., + [ 6.4659e-04, -6.7391e-06, 1.1454e-03, ..., 1.4460e-04, + 4.8661e-04, 1.0958e-03], + [ 5.9557e-04, 2.8759e-05, 8.8644e-04, ..., -6.6519e-05, + 4.3750e-04, 1.5373e-03], + [ 8.3208e-04, 2.8923e-05, 1.0490e-03, ..., 9.8586e-05, + 6.0225e-04, 2.6588e-03]], device='cuda:0') +Epoch 408, bias, value: tensor([ 0.0092, 0.0064, -0.0046, -0.0172, 0.0074, -0.0138, -0.0138, -0.0180, + -0.0052, 0.0089], device='cuda:0'), grad: tensor([-0.0041, -0.1091, 0.0025, 0.0266, -0.0282, -0.0079, 0.0332, 0.0277, + 0.0253, 0.0341], device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 216.38, cls_loss 0.4641 cls_loss_mapping 0.0011 cls_loss_causal 0.4369 re_mapping 0.0044 re_causal 0.0129 /// teacc 98.97 lr 0.00001000 +Epoch 409, weight, value: tensor([[ 0.0388, 0.1309, -0.1813, ..., -0.1077, 0.0770, 0.0042], + [-0.1084, -0.1057, 0.1169, ..., -0.0471, -0.1151, 0.0567], + [-0.0969, -0.0720, -0.1184, ..., -0.1362, -0.0172, 0.0560], + ..., + [-0.0091, -0.1343, -0.1414, ..., -0.1144, -0.0514, 0.0774], + [-0.1330, -0.1164, -0.1182, ..., -0.0670, -0.0475, -0.1010], + [ 0.0217, -0.0319, -0.0570, ..., -0.1041, 0.0113, -0.0238]], + device='cuda:0'), grad: tensor([[ 1.2026e-03, -7.7933e-06, 5.2929e-05, ..., -5.8785e-06, + -3.3863e-06, 4.9651e-05], + [-1.7662e-03, 1.5869e-03, 1.2192e-02, ..., 7.2643e-08, + 1.2200e-07, 9.6436e-03], + [-9.6464e-04, -1.6613e-03, -1.2802e-02, ..., 2.3935e-07, + 9.2667e-07, -1.0071e-02], + ..., + [ 8.0681e-04, 3.6031e-05, 1.1194e-04, ..., 1.8300e-07, + 7.2867e-06, -9.5904e-05], + [ 7.1478e-04, -4.7386e-05, -7.5638e-05, ..., 1.2834e-06, + 4.9993e-06, -2.5177e-04], + [ 1.5335e-03, -7.8499e-05, 4.1652e-04, ..., -8.2701e-06, + -4.0817e-04, 4.8637e-04]], device='cuda:0') +Epoch 409, bias, value: tensor([ 0.0093, 0.0064, -0.0047, -0.0172, 0.0073, -0.0138, -0.0137, -0.0180, + -0.0052, 0.0088], device='cuda:0'), grad: tensor([ 0.0061, 0.0121, -0.0205, -0.0266, 0.0091, 0.0058, 0.0066, 0.0050, + -0.0154, 0.0178], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 408---------------------------------------------------- +epoch 408, time 216.88, cls_loss 0.4550 cls_loss_mapping 0.0010 cls_loss_causal 0.4238 re_mapping 0.0044 re_causal 0.0130 /// teacc 99.04 lr 0.00001000 +Epoch 410, weight, value: tensor([[ 0.0387, 0.1310, -0.1814, ..., -0.1077, 0.0770, 0.0042], + [-0.1084, -0.1057, 0.1169, ..., -0.0472, -0.1152, 0.0566], + [-0.0966, -0.0718, -0.1183, ..., -0.1361, -0.0170, 0.0561], + ..., + [-0.0091, -0.1343, -0.1414, ..., -0.1144, -0.0513, 0.0775], + [-0.1330, -0.1165, -0.1182, ..., -0.0669, -0.0475, -0.1009], + [ 0.0216, -0.0320, -0.0570, ..., -0.1041, 0.0110, -0.0240]], + device='cuda:0'), grad: tensor([[ 1.2708e-04, -7.6592e-05, 4.1056e-04, ..., 1.8021e-07, + -8.2195e-05, 9.9945e-04], + [ 3.9172e-04, 1.0878e-06, 6.5088e-04, ..., 1.8626e-09, + 4.9025e-06, 1.9503e-03], + [ 4.2772e-04, 1.3269e-05, 5.8270e-04, ..., -4.7088e-06, + 1.5192e-05, 1.6842e-03], + ..., + [ 1.0872e-03, 1.6410e-06, 5.7316e-04, ..., 3.2596e-08, + 8.2329e-06, 4.6616e-03], + [-2.0332e-03, 2.7508e-05, -1.1511e-03, ..., 4.2617e-06, + 7.6033e-06, -3.1757e-03], + [ 2.0230e-04, 2.0757e-05, 9.6846e-04, ..., 6.5193e-09, + 1.9327e-05, 1.3363e-04]], device='cuda:0') +Epoch 410, bias, value: tensor([ 0.0093, 0.0064, -0.0046, -0.0173, 0.0074, -0.0137, -0.0138, -0.0180, + -0.0051, 0.0087], device='cuda:0'), grad: tensor([ 0.0120, 0.0196, 0.0156, 0.0167, -0.0783, -0.0196, 0.0120, 0.0103, + -0.0145, 0.0262], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 409---------------------------------------------------- +epoch 409, time 216.89, cls_loss 0.4536 cls_loss_mapping 0.0008 cls_loss_causal 0.4221 re_mapping 0.0042 re_causal 0.0131 /// teacc 99.05 lr 0.00001000 +Epoch 411, weight, value: tensor([[ 0.0387, 0.1311, -0.1813, ..., -0.1079, 0.0771, 0.0042], + [-0.1083, -0.1058, 0.1168, ..., -0.0473, -0.1152, 0.0566], + [-0.0966, -0.0718, -0.1183, ..., -0.1362, -0.0170, 0.0560], + ..., + [-0.0090, -0.1344, -0.1415, ..., -0.1145, -0.0513, 0.0776], + [-0.1330, -0.1165, -0.1182, ..., -0.0669, -0.0476, -0.1011], + [ 0.0215, -0.0319, -0.0570, ..., -0.1042, 0.0108, -0.0239]], + device='cuda:0'), grad: tensor([[-4.3640e-03, -1.9018e-06, -1.0862e-03, ..., -1.4277e-03, + -1.4114e-04, -1.2283e-03], + [ 3.6573e-04, 8.7824e-07, -1.6022e-03, ..., 2.4343e-04, + 2.6286e-05, 6.2084e-04], + [ 2.9635e-04, 4.1537e-06, 8.6784e-04, ..., 4.0960e-04, + 1.8880e-05, 1.5802e-03], + ..., + [ 2.9707e-04, 1.1446e-06, 4.9162e-04, ..., 5.0932e-05, + 2.4736e-05, 3.0684e-04], + [ 1.7500e-03, 9.4855e-07, 1.9360e-03, ..., 3.4714e-04, + 1.7166e-05, 2.2449e-03], + [-1.5965e-03, 4.8019e-06, -2.0504e-03, ..., 1.4293e-04, + 3.2157e-05, -8.2970e-04]], device='cuda:0') +Epoch 411, bias, value: tensor([ 0.0092, 0.0065, -0.0047, -0.0173, 0.0074, -0.0137, -0.0138, -0.0179, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([-0.0441, -0.0050, 0.0201, 0.0124, 0.0263, 0.0087, -0.0279, 0.0118, + 0.0026, -0.0049], device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 216.76, cls_loss 0.4344 cls_loss_mapping 0.0010 cls_loss_causal 0.4081 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.04 lr 0.00001000 +Epoch 412, weight, value: tensor([[ 0.0387, 0.1310, -0.1811, ..., -0.1077, 0.0771, 0.0044], + [-0.1085, -0.1058, 0.1167, ..., -0.0473, -0.1153, 0.0565], + [-0.0966, -0.0718, -0.1184, ..., -0.1363, -0.0172, 0.0559], + ..., + [-0.0089, -0.1344, -0.1415, ..., -0.1144, -0.0513, 0.0776], + [-0.1330, -0.1166, -0.1182, ..., -0.0668, -0.0476, -0.1010], + [ 0.0215, -0.0318, -0.0569, ..., -0.1043, 0.0108, -0.0240]], + device='cuda:0'), grad: tensor([[-4.8866e-03, 2.0757e-05, 3.2139e-04, ..., 5.7817e-05, + -1.3084e-03, -6.1893e-04], + [ 1.5640e-04, 1.1533e-04, -1.1845e-03, ..., 2.0278e-04, + 1.3757e-04, 9.9063e-05], + [ 4.7569e-03, 7.5626e-04, 1.1349e-03, ..., 1.0508e-04, + 4.7541e-04, -7.1526e-03], + ..., + [ 4.9639e-04, 1.1571e-05, 4.7874e-04, ..., 3.0607e-05, + 1.0121e-04, 6.6986e-03], + [ 1.1024e-03, 2.1899e-04, 5.4359e-04, ..., 4.0978e-05, + 2.4438e-04, 1.0376e-03], + [ 4.4560e-04, 3.7700e-05, 3.6955e-04, ..., 1.4412e-04, + 1.1182e-04, 1.0843e-03]], device='cuda:0') +Epoch 412, bias, value: tensor([ 0.0094, 0.0064, -0.0048, -0.0173, 0.0074, -0.0137, -0.0137, -0.0179, + -0.0052, 0.0088], device='cuda:0'), grad: tensor([-0.0138, -0.0165, 0.0026, 0.0197, 0.0147, -0.0750, 0.0077, 0.0327, + 0.0144, 0.0135], device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 216.08, cls_loss 0.4225 cls_loss_mapping 0.0008 cls_loss_causal 0.3948 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.02 lr 0.00001000 +Epoch 413, weight, value: tensor([[ 0.0387, 0.1311, -0.1811, ..., -0.1077, 0.0773, 0.0046], + [-0.1086, -0.1058, 0.1168, ..., -0.0471, -0.1154, 0.0565], + [-0.0967, -0.0718, -0.1183, ..., -0.1363, -0.0173, 0.0560], + ..., + [-0.0089, -0.1344, -0.1414, ..., -0.1144, -0.0514, 0.0776], + [-0.1330, -0.1166, -0.1182, ..., -0.0667, -0.0477, -0.1010], + [ 0.0215, -0.0317, -0.0569, ..., -0.1044, 0.0107, -0.0240]], + device='cuda:0'), grad: tensor([[ 1.2522e-03, 6.7890e-05, 4.1723e-05, ..., 5.7602e-04, + 7.7009e-05, 1.4801e-03], + [ 1.3876e-03, 6.6906e-06, 7.1466e-05, ..., 1.3471e-04, + 1.7285e-04, 1.4982e-03], + [-1.0386e-03, 7.5698e-06, 7.0453e-05, ..., 1.3769e-04, + -8.1968e-04, -7.9193e-03], + ..., + [ 5.0735e-03, 4.1202e-06, 1.1188e-04, ..., 2.0814e-04, + 1.1377e-03, 1.0353e-02], + [ 1.5907e-03, 4.4167e-05, 1.2708e-04, ..., 1.4580e-02, + 2.1386e-04, 1.7300e-03], + [-4.3678e-03, 1.3128e-05, -5.6744e-04, ..., -1.1091e-03, + -1.4896e-03, -5.7678e-03]], device='cuda:0') +Epoch 413, bias, value: tensor([ 0.0094, 0.0063, -0.0049, -0.0174, 0.0074, -0.0136, -0.0138, -0.0177, + -0.0051, 0.0087], device='cuda:0'), grad: tensor([ 0.0140, 0.0159, -0.0149, 0.0142, -0.0144, -0.0439, -0.0338, 0.0425, + 0.0349, -0.0144], device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 216.74, cls_loss 0.4668 cls_loss_mapping 0.0011 cls_loss_causal 0.4392 re_mapping 0.0043 re_causal 0.0127 /// teacc 99.02 lr 0.00001000 +Epoch 414, weight, value: tensor([[ 0.0387, 0.1311, -0.1811, ..., -0.1077, 0.0773, 0.0047], + [-0.1085, -0.1057, 0.1168, ..., -0.0472, -0.1154, 0.0565], + [-0.0966, -0.0718, -0.1182, ..., -0.1363, -0.0172, 0.0561], + ..., + [-0.0090, -0.1344, -0.1413, ..., -0.1144, -0.0515, 0.0775], + [-0.1331, -0.1167, -0.1183, ..., -0.0667, -0.0478, -0.1011], + [ 0.0214, -0.0317, -0.0571, ..., -0.1044, 0.0109, -0.0240]], + device='cuda:0'), grad: tensor([[ 4.1008e-04, -1.6779e-05, 1.9789e-04, ..., 1.6212e-05, + 5.6118e-05, 7.0286e-04], + [ 5.2786e-04, 3.0641e-06, 2.4533e-04, ..., 1.5408e-05, + 5.9456e-05, 9.3079e-04], + [ 5.3787e-04, 2.3305e-05, 2.4962e-04, ..., 3.5057e-03, + 6.0529e-05, 9.9659e-04], + ..., + [ 4.3726e-04, 3.8110e-06, 2.8944e-04, ..., 1.2610e-06, + 6.8963e-05, 5.6124e-04], + [ 1.2989e-03, 6.5613e-04, 1.9264e-04, ..., 2.3484e-04, + 5.7667e-05, 5.3072e-04], + [-2.2182e-03, -7.6771e-04, 2.1577e-04, ..., 1.4286e-06, + -3.2783e-04, -1.2302e-03]], device='cuda:0') +Epoch 414, bias, value: tensor([ 0.0094, 0.0063, -0.0046, -0.0174, 0.0075, -0.0135, -0.0140, -0.0178, + -0.0052, 0.0086], device='cuda:0'), grad: tensor([ 0.0090, 0.0107, 0.0282, -0.0236, -0.0152, 0.0136, -0.0098, 0.0095, + 0.0141, -0.0365], device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 216.91, cls_loss 0.4279 cls_loss_mapping 0.0009 cls_loss_causal 0.4006 re_mapping 0.0043 re_causal 0.0126 /// teacc 99.03 lr 0.00001000 +Epoch 415, weight, value: tensor([[ 0.0386, 0.1312, -0.1811, ..., -0.1076, 0.0773, 0.0047], + [-0.1085, -0.1058, 0.1167, ..., -0.0472, -0.1153, 0.0565], + [-0.0966, -0.0717, -0.1182, ..., -0.1364, -0.0174, 0.0561], + ..., + [-0.0090, -0.1346, -0.1413, ..., -0.1144, -0.0515, 0.0775], + [-0.1330, -0.1167, -0.1182, ..., -0.0667, -0.0477, -0.1009], + [ 0.0212, -0.0317, -0.0571, ..., -0.1043, 0.0109, -0.0240]], + device='cuda:0'), grad: tensor([[-4.8876e-04, -4.1199e-04, -3.4657e-03, ..., 0.0000e+00, + -2.1210e-03, 6.6996e-05], + [-1.9789e-04, 1.3545e-05, -1.2360e-03, ..., 0.0000e+00, + 2.0969e-04, -9.5510e-04], + [ 1.2088e-04, -7.2531e-06, 7.4863e-04, ..., 0.0000e+00, + 3.9959e-04, 1.3769e-04], + ..., + [ 3.7402e-05, 1.3903e-05, 3.7456e-04, ..., 0.0000e+00, + 8.8274e-05, 4.5687e-05], + [ 7.1883e-05, 5.0753e-05, 3.8004e-04, ..., 0.0000e+00, + 1.4794e-04, 1.0955e-04], + [-1.0088e-05, 4.9353e-05, 3.9124e-04, ..., 0.0000e+00, + 1.4019e-04, 5.5522e-05]], device='cuda:0') +Epoch 415, bias, value: tensor([ 0.0094, 0.0064, -0.0046, -0.0175, 0.0076, -0.0137, -0.0142, -0.0178, + -0.0049, 0.0084], device='cuda:0'), grad: tensor([-0.0287, -0.0248, -0.0194, 0.0162, 0.0092, 0.0161, 0.0119, 0.0060, + 0.0098, 0.0038], device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 217.20, cls_loss 0.4381 cls_loss_mapping 0.0009 cls_loss_causal 0.4169 re_mapping 0.0042 re_causal 0.0124 /// teacc 99.03 lr 0.00001000 +Epoch 416, weight, value: tensor([[ 0.0386, 0.1313, -0.1810, ..., -0.1077, 0.0772, 0.0047], + [-0.1085, -0.1059, 0.1166, ..., -0.0472, -0.1153, 0.0563], + [-0.0966, -0.0718, -0.1182, ..., -0.1365, -0.0174, 0.0561], + ..., + [-0.0090, -0.1348, -0.1415, ..., -0.1145, -0.0515, 0.0775], + [-0.1331, -0.1167, -0.1181, ..., -0.0666, -0.0477, -0.1008], + [ 0.0213, -0.0318, -0.0572, ..., -0.1044, 0.0109, -0.0238]], + device='cuda:0'), grad: tensor([[ 1.0133e-04, -2.9163e-03, 2.7008e-08, ..., 1.4439e-05, + -9.2850e-03, -5.2986e-03], + [ 1.3697e-04, 1.4476e-05, -1.8254e-07, ..., 3.1237e-06, + 3.1328e-04, -4.5013e-03], + [ 1.4675e-04, 1.7204e-03, 1.8943e-06, ..., 2.7195e-06, + 6.3934e-03, 6.1073e-03], + ..., + [ 1.7357e-04, 5.7407e-06, 1.8952e-06, ..., 1.7742e-07, + 3.0661e-04, 4.3344e-04], + [-1.3361e-03, 3.8981e-05, 1.7742e-07, ..., 5.7101e-05, + -2.5139e-03, 6.8378e-04], + [ 2.7585e-04, 1.7583e-05, -2.5164e-06, ..., 5.2154e-07, + 4.7255e-04, 4.9543e-04]], device='cuda:0') +Epoch 416, bias, value: tensor([ 0.0094, 0.0063, -0.0047, -0.0174, 0.0076, -0.0138, -0.0141, -0.0179, + -0.0048, 0.0086], device='cuda:0'), grad: tensor([-0.0425, -0.0230, 0.0489, 0.0135, 0.0069, -0.0257, 0.0332, 0.0058, + -0.0245, 0.0075], device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 216.41, cls_loss 0.4615 cls_loss_mapping 0.0010 cls_loss_causal 0.4381 re_mapping 0.0040 re_causal 0.0125 /// teacc 99.00 lr 0.00001000 +Epoch 417, weight, value: tensor([[ 0.0385, 0.1314, -0.1810, ..., -0.1077, 0.0772, 0.0046], + [-0.1085, -0.1060, 0.1168, ..., -0.0473, -0.1153, 0.0564], + [-0.0966, -0.0717, -0.1183, ..., -0.1364, -0.0175, 0.0560], + ..., + [-0.0088, -0.1349, -0.1416, ..., -0.1144, -0.0515, 0.0775], + [-0.1331, -0.1168, -0.1181, ..., -0.0668, -0.0475, -0.1009], + [ 0.0214, -0.0318, -0.0572, ..., -0.1043, 0.0110, -0.0239]], + device='cuda:0'), grad: tensor([[ 1.7548e-04, -6.5684e-05, 6.6376e-04, ..., 1.4293e-04, + 1.3101e-04, 8.4066e-04], + [-1.8799e-04, -1.0061e-04, -1.0033e-03, ..., -1.6665e-04, + 4.7743e-05, -5.0068e-04], + [ 2.3019e-04, -4.2033e-04, 5.8699e-04, ..., 5.9992e-05, + -1.8349e-03, 7.6580e-04], + ..., + [ 3.2485e-05, 3.9268e-04, -2.3918e-03, ..., -5.3692e-04, + 1.7004e-03, 1.5199e-04], + [ 2.0730e-04, 9.3579e-05, 7.3195e-04, ..., 1.3804e-04, + 2.3234e-04, -3.5343e-03], + [ 9.3043e-05, 5.4598e-05, 6.6805e-04, ..., 5.8889e-05, + 2.9302e-04, 8.0824e-04]], device='cuda:0') +Epoch 417, bias, value: tensor([ 0.0095, 0.0064, -0.0047, -0.0172, 0.0075, -0.0138, -0.0141, -0.0180, + -0.0049, 0.0086], device='cuda:0'), grad: tensor([ 0.0137, -0.0128, -0.0021, 0.0156, -0.0142, 0.0121, -0.0083, -0.0023, + -0.0165, 0.0146], device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 216.01, cls_loss 0.4444 cls_loss_mapping 0.0010 cls_loss_causal 0.4153 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.00 lr 0.00001000 +Epoch 418, weight, value: tensor([[ 0.0386, 0.1314, -0.1810, ..., -0.1077, 0.0772, 0.0046], + [-0.1085, -0.1059, 0.1167, ..., -0.0475, -0.1154, 0.0564], + [-0.0966, -0.0718, -0.1184, ..., -0.1363, -0.0174, 0.0560], + ..., + [-0.0088, -0.1350, -0.1416, ..., -0.1143, -0.0516, 0.0774], + [-0.1331, -0.1168, -0.1181, ..., -0.0667, -0.0476, -0.1010], + [ 0.0213, -0.0318, -0.0573, ..., -0.1043, 0.0109, -0.0239]], + device='cuda:0'), grad: tensor([[-1.7385e-03, -2.0063e-04, 2.7680e-04, ..., 1.8942e-04, + -3.3170e-05, -4.4136e-03], + [-7.6473e-05, -5.5361e-04, 1.4496e-04, ..., -7.2527e-04, + 9.5963e-06, -2.4700e-03], + [ 7.2145e-04, 8.4400e-05, 2.3174e-04, ..., 2.2486e-05, + -1.0042e-03, -4.5204e-03], + ..., + [ 1.8034e-03, 3.7283e-05, 2.9612e-04, ..., 3.7551e-05, + 6.5470e-04, 3.9711e-03], + [ 4.8494e-04, 7.4565e-05, 2.7585e-04, ..., 4.3958e-05, + 2.8819e-05, 1.4248e-03], + [ 6.2599e-03, 8.2493e-05, 3.3522e-04, ..., 4.0889e-05, + 3.8815e-04, 5.2338e-03]], device='cuda:0') +Epoch 418, bias, value: tensor([ 0.0095, 0.0064, -0.0048, -0.0172, 0.0076, -0.0137, -0.0142, -0.0179, + -0.0049, 0.0085], device='cuda:0'), grad: tensor([-0.0181, -0.0203, -0.0390, 0.0006, -0.0073, 0.0294, -0.0154, 0.0225, + 0.0133, 0.0342], device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 216.44, cls_loss 0.4529 cls_loss_mapping 0.0011 cls_loss_causal 0.4286 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.04 lr 0.00001000 +Epoch 419, weight, value: tensor([[ 0.0387, 0.1316, -0.1811, ..., -0.1077, 0.0773, 0.0048], + [-0.1084, -0.1059, 0.1167, ..., -0.0475, -0.1152, 0.0562], + [-0.0966, -0.0716, -0.1183, ..., -0.1364, -0.0175, 0.0561], + ..., + [-0.0089, -0.1350, -0.1417, ..., -0.1144, -0.0516, 0.0775], + [-0.1333, -0.1167, -0.1181, ..., -0.0667, -0.0476, -0.1011], + [ 0.0213, -0.0318, -0.0574, ..., -0.1044, 0.0110, -0.0239]], + device='cuda:0'), grad: tensor([[ 4.0531e-04, 1.2326e-04, 3.7956e-04, ..., 2.0802e-04, + 3.7402e-05, 2.9612e-04], + [ 4.1842e-04, 7.9274e-05, 4.7374e-04, ..., 6.0380e-05, + 1.7807e-05, 1.8129e-03], + [-1.9073e-05, 4.3213e-05, 3.0828e-04, ..., 5.0992e-05, + 1.2048e-05, -3.3283e-03], + ..., + [ 1.1311e-03, 1.0586e-04, 4.1509e-04, ..., 9.4712e-05, + 3.9130e-05, 5.9967e-03], + [ 8.3208e-04, 3.2306e-04, -1.3180e-03, ..., 3.3808e-04, + 3.4189e-04, 1.7881e-03], + [-4.8218e-03, -2.0008e-03, -2.5654e-03, ..., -1.7776e-03, + -7.4625e-04, -1.0368e-02]], device='cuda:0') +Epoch 419, bias, value: tensor([ 0.0096, 0.0063, -0.0046, -0.0173, 0.0076, -0.0138, -0.0141, -0.0180, + -0.0050, 0.0085], device='cuda:0'), grad: tensor([-0.0149, 0.0253, 0.0071, 0.0190, 0.0167, 0.0166, -0.0126, 0.0262, + -0.0048, -0.0786], device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 216.24, cls_loss 0.4422 cls_loss_mapping 0.0009 cls_loss_causal 0.4173 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.02 lr 0.00001000 +Epoch 420, weight, value: tensor([[ 0.0389, 0.1314, -0.1811, ..., -0.1077, 0.0774, 0.0047], + [-0.1085, -0.1058, 0.1168, ..., -0.0474, -0.1150, 0.0563], + [-0.0965, -0.0716, -0.1183, ..., -0.1363, -0.0176, 0.0562], + ..., + [-0.0089, -0.1351, -0.1417, ..., -0.1143, -0.0516, 0.0775], + [-0.1334, -0.1167, -0.1180, ..., -0.0668, -0.0476, -0.1012], + [ 0.0214, -0.0319, -0.0573, ..., -0.1044, 0.0110, -0.0238]], + device='cuda:0'), grad: tensor([[-2.9316e-03, -2.5196e-03, -1.8530e-03, ..., -3.1071e-03, + -5.2124e-05, -8.4000e-03], + [ 7.1669e-04, 2.8181e-04, 2.6250e-04, ..., 8.2731e-05, + 2.6628e-05, 2.0313e-03], + [ 1.0633e-03, 3.9840e-04, 3.2282e-04, ..., 1.7536e-04, + 8.6665e-05, 2.5043e-03], + ..., + [ 1.7710e-03, 7.7820e-04, 7.8249e-04, ..., 7.3433e-04, + 6.2287e-05, 3.6793e-03], + [ 1.6193e-03, 5.2834e-04, 4.1533e-04, ..., 9.6738e-05, + 1.9097e-04, 1.8902e-03], + [-1.0788e-02, -3.8695e-04, -2.6226e-04, ..., 1.3435e-04, + -3.4356e-04, 1.3647e-03]], device='cuda:0') +Epoch 420, bias, value: tensor([ 0.0096, 0.0064, -0.0046, -0.0173, 0.0076, -0.0138, -0.0142, -0.0180, + -0.0051, 0.0086], device='cuda:0'), grad: tensor([-0.0453, 0.0231, 0.0233, -0.0358, 0.0246, -0.0134, -0.0393, 0.0287, + 0.0282, 0.0061], device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 216.71, cls_loss 0.4287 cls_loss_mapping 0.0009 cls_loss_causal 0.4016 re_mapping 0.0040 re_causal 0.0118 /// teacc 99.03 lr 0.00001000 +Epoch 421, weight, value: tensor([[ 0.0389, 0.1314, -0.1812, ..., -0.1076, 0.0774, 0.0048], + [-0.1086, -0.1058, 0.1169, ..., -0.0474, -0.1150, 0.0564], + [-0.0965, -0.0716, -0.1184, ..., -0.1363, -0.0176, 0.0562], + ..., + [-0.0089, -0.1352, -0.1416, ..., -0.1144, -0.0516, 0.0774], + [-0.1335, -0.1165, -0.1181, ..., -0.0668, -0.0477, -0.1012], + [ 0.0213, -0.0319, -0.0574, ..., -0.1043, 0.0109, -0.0238]], + device='cuda:0'), grad: tensor([[ 4.9858e-03, 6.2108e-05, -1.4410e-03, ..., 4.0717e-06, + 1.1671e-04, -3.4833e-04], + [ 2.1935e-04, 5.7742e-06, 7.9422e-03, ..., 3.5428e-06, + 2.2218e-05, 1.2980e-03], + [ 3.7789e-04, 2.5615e-05, 1.3075e-03, ..., 3.8594e-06, + 4.4316e-05, 5.2738e-04], + ..., + [ 4.9973e-04, 2.6405e-05, -7.0572e-03, ..., 1.9014e-04, + 1.3673e-04, -7.8154e-04], + [ 9.9754e-04, 7.0155e-05, 6.7234e-04, ..., 3.2812e-05, + 1.1104e-04, 3.4189e-04], + [ 2.2459e-04, -1.9455e-04, 7.7915e-04, ..., 5.1117e-04, + -1.0389e-04, 1.0395e-03]], device='cuda:0') +Epoch 421, bias, value: tensor([ 0.0097, 0.0064, -0.0046, -0.0174, 0.0077, -0.0136, -0.0143, -0.0181, + -0.0051, 0.0085], device='cuda:0'), grad: tensor([-0.0128, 0.0220, 0.0107, 0.0107, -0.0553, 0.0092, -0.0029, -0.0015, + 0.0102, 0.0097], device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 216.44, cls_loss 0.4675 cls_loss_mapping 0.0010 cls_loss_causal 0.4397 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.05 lr 0.00001000 +Epoch 422, weight, value: tensor([[ 0.0388, 0.1314, -0.1812, ..., -0.1076, 0.0775, 0.0047], + [-0.1087, -0.1058, 0.1168, ..., -0.0475, -0.1150, 0.0564], + [-0.0966, -0.0716, -0.1185, ..., -0.1363, -0.0176, 0.0562], + ..., + [-0.0089, -0.1352, -0.1416, ..., -0.1142, -0.0515, 0.0773], + [-0.1334, -0.1165, -0.1179, ..., -0.0667, -0.0477, -0.1010], + [ 0.0213, -0.0320, -0.0574, ..., -0.1043, 0.0108, -0.0238]], + device='cuda:0'), grad: tensor([[-1.3256e-03, -1.3232e-04, 2.9421e-04, ..., 1.7810e-04, + -2.5773e-04, -6.8932e-03], + [ 3.6860e-04, -1.2732e-04, -3.4027e-03, ..., 1.2100e-05, + 1.3888e-05, -2.4357e-03], + [ 2.3022e-03, -1.1892e-03, 3.3975e-04, ..., 2.6420e-05, + 3.4690e-05, 4.1313e-03], + ..., + [ 3.7479e-04, 2.7323e-04, 2.6894e-03, ..., 5.9493e-06, + 3.3647e-05, 4.3526e-03], + [ 2.3866e-04, 1.5330e-04, 3.2783e-04, ..., 4.2272e-04, + 1.2350e-04, 7.8869e-04], + [ 1.7121e-05, 1.4997e-04, 8.6594e-04, ..., 2.6971e-05, + 1.3132e-07, 1.3742e-03]], device='cuda:0') +Epoch 422, bias, value: tensor([ 0.0096, 0.0064, -0.0046, -0.0173, 0.0077, -0.0136, -0.0143, -0.0182, + -0.0050, 0.0085], device='cuda:0'), grad: tensor([-0.0532, -0.0020, 0.0105, -0.0420, 0.0143, 0.0115, 0.0053, 0.0302, + 0.0123, 0.0131], device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 216.29, cls_loss 0.4585 cls_loss_mapping 0.0010 cls_loss_causal 0.4351 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.00 lr 0.00001000 +Epoch 423, weight, value: tensor([[ 0.0390, 0.1314, -0.1814, ..., -0.1075, 0.0776, 0.0048], + [-0.1089, -0.1058, 0.1166, ..., -0.0476, -0.1149, 0.0562], + [-0.0965, -0.0715, -0.1183, ..., -0.1363, -0.0175, 0.0563], + ..., + [-0.0088, -0.1352, -0.1415, ..., -0.1140, -0.0515, 0.0773], + [-0.1333, -0.1165, -0.1178, ..., -0.0668, -0.0477, -0.1008], + [ 0.0214, -0.0319, -0.0574, ..., -0.1044, 0.0108, -0.0238]], + device='cuda:0'), grad: tensor([[ 3.7789e-04, -1.8984e-05, 8.1825e-04, ..., 6.6720e-06, + 3.2663e-04, 9.3746e-04], + [ 6.6376e-04, 1.0887e-06, 1.1272e-03, ..., 1.2191e-06, + 5.9366e-04, 1.5202e-03], + [ 4.4537e-04, 9.2387e-07, 8.8644e-04, ..., 1.5898e-06, + 4.9257e-04, 7.0524e-04], + ..., + [-3.8052e-03, 2.9542e-06, 9.2173e-04, ..., 1.6764e-07, + -3.3798e-03, -7.7400e-03], + [ 2.6155e-04, 1.1998e-04, -1.4200e-03, ..., 2.8573e-06, + 2.4235e-04, 3.8481e-04], + [ 4.9448e-04, 9.6560e-06, 8.9836e-04, ..., 9.6485e-07, + 4.8590e-04, 9.8038e-04]], device='cuda:0') +Epoch 423, bias, value: tensor([ 0.0095, 0.0063, -0.0045, -0.0172, 0.0077, -0.0136, -0.0143, -0.0181, + -0.0049, 0.0085], device='cuda:0'), grad: tensor([ 0.0185, 0.0234, -0.0158, -0.0133, 0.0183, -0.0152, -0.0100, -0.0133, + -0.0110, 0.0184], device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 216.23, cls_loss 0.4687 cls_loss_mapping 0.0010 cls_loss_causal 0.4447 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.01 lr 0.00001000 +Epoch 424, weight, value: tensor([[ 0.0390, 0.1315, -0.1814, ..., -0.1074, 0.0777, 0.0049], + [-0.1089, -0.1058, 0.1165, ..., -0.0476, -0.1150, 0.0562], + [-0.0967, -0.0717, -0.1183, ..., -0.1362, -0.0176, 0.0562], + ..., + [-0.0087, -0.1350, -0.1414, ..., -0.1140, -0.0515, 0.0775], + [-0.1334, -0.1166, -0.1178, ..., -0.0669, -0.0477, -0.1007], + [ 0.0215, -0.0320, -0.0575, ..., -0.1045, 0.0109, -0.0238]], + device='cuda:0'), grad: tensor([[-3.0918e-03, -1.2052e-04, 5.9175e-04, ..., -7.4506e-05, + -2.2507e-04, -3.6755e-03], + [ 5.0449e-04, 4.5914e-07, -1.4200e-03, ..., 3.1618e-07, + 3.9749e-06, -3.1128e-03], + [ 1.0138e-03, 4.6007e-06, 6.5660e-04, ..., 6.6264e-07, + 3.5673e-05, 5.3310e-04], + ..., + [-2.8458e-03, 1.4734e-06, 6.8521e-04, ..., 2.5611e-08, + -3.1757e-04, 5.8794e-04], + [-1.1871e-02, 6.8769e-06, 7.0429e-04, ..., 5.4166e-06, + 9.4175e-05, 1.0061e-03], + [ 4.7035e-03, 3.4094e-05, 6.6853e-04, ..., 6.7847e-07, + 1.8358e-04, 5.9509e-04]], device='cuda:0') +Epoch 424, bias, value: tensor([ 0.0095, 0.0062, -0.0045, -0.0171, 0.0077, -0.0136, -0.0143, -0.0181, + -0.0050, 0.0085], device='cuda:0'), grad: tensor([-0.0107, -0.0302, 0.0130, 0.0374, -0.0378, -0.0110, 0.0267, -0.0030, + -0.0096, 0.0252], device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 216.18, cls_loss 0.4485 cls_loss_mapping 0.0008 cls_loss_causal 0.4248 re_mapping 0.0040 re_causal 0.0119 /// teacc 99.00 lr 0.00001000 +Epoch 425, weight, value: tensor([[ 0.0389, 0.1315, -0.1815, ..., -0.1075, 0.0777, 0.0049], + [-0.1090, -0.1059, 0.1164, ..., -0.0478, -0.1152, 0.0562], + [-0.0968, -0.0716, -0.1182, ..., -0.1363, -0.0176, 0.0562], + ..., + [-0.0088, -0.1352, -0.1414, ..., -0.1141, -0.0518, 0.0773], + [-0.1335, -0.1166, -0.1177, ..., -0.0668, -0.0476, -0.1008], + [ 0.0215, -0.0320, -0.0576, ..., -0.1045, 0.0109, -0.0238]], + device='cuda:0'), grad: tensor([[ 2.7919e-04, -5.9217e-05, 3.8457e-04, ..., 4.2289e-05, + -8.5950e-05, 6.1512e-04], + [ 4.3583e-04, 8.2105e-06, 5.4979e-04, ..., 3.3289e-05, + 2.6032e-05, 3.2921e-03], + [ 1.7366e-03, 6.4313e-05, 4.2105e-04, ..., 2.9802e-04, + 1.1122e-04, -1.0662e-03], + ..., + [ 9.2506e-04, 3.0085e-05, 5.0259e-04, ..., 7.1883e-05, + 4.0054e-05, 9.7847e-04], + [-2.4338e-03, 1.7095e-04, 6.0081e-04, ..., -6.0892e-04, + -1.9383e-04, -4.2419e-03], + [ 3.8981e-04, 3.1799e-05, 4.5943e-04, ..., 5.4032e-05, + 4.5776e-05, 7.1907e-04]], device='cuda:0') +Epoch 425, bias, value: tensor([ 0.0094, 0.0062, -0.0045, -0.0172, 0.0078, -0.0138, -0.0143, -0.0180, + -0.0050, 0.0086], device='cuda:0'), grad: tensor([-0.0010, 0.0232, -0.0153, -0.0321, -0.0083, -0.0199, 0.0320, 0.0144, + -0.0111, 0.0182], device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 216.60, cls_loss 0.4461 cls_loss_mapping 0.0009 cls_loss_causal 0.4192 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.03 lr 0.00001000 +Epoch 426, weight, value: tensor([[ 0.0390, 0.1317, -0.1815, ..., -0.1075, 0.0778, 0.0050], + [-0.1088, -0.1059, 0.1163, ..., -0.0477, -0.1151, 0.0562], + [-0.0967, -0.0716, -0.1182, ..., -0.1362, -0.0177, 0.0564], + ..., + [-0.0087, -0.1352, -0.1415, ..., -0.1142, -0.0516, 0.0773], + [-0.1335, -0.1167, -0.1177, ..., -0.0666, -0.0477, -0.1007], + [ 0.0214, -0.0318, -0.0572, ..., -0.1044, 0.0109, -0.0238]], + device='cuda:0'), grad: tensor([[-2.1801e-03, -3.9506e-04, 3.9935e-04, ..., 2.8634e-04, + -1.8797e-03, -1.8873e-03], + [-5.3444e-03, 7.7486e-05, 2.9278e-04, ..., 1.2445e-04, + 2.0885e-04, -4.1351e-03], + [ 1.1024e-03, 1.0633e-04, 3.1734e-04, ..., 1.6379e-04, + 2.6011e-04, 1.2627e-03], + ..., + [ 9.8724e-03, 2.2471e-04, 2.4235e-04, ..., 4.6432e-05, + 9.2268e-04, 2.0737e-02], + [ 1.5659e-03, 3.6716e-04, 7.9203e-04, ..., 7.8773e-04, + 4.1986e-04, 1.8730e-03], + [-6.8283e-03, 1.3864e-04, 3.2401e-04, ..., 1.5700e-04, + 3.5572e-04, -1.7151e-02]], device='cuda:0') +Epoch 426, bias, value: tensor([ 0.0095, 0.0063, -0.0044, -0.0173, 0.0077, -0.0138, -0.0144, -0.0181, + -0.0050, 0.0086], device='cuda:0'), grad: tensor([-0.0024, -0.0448, 0.0188, 0.0197, 0.0149, 0.0249, -0.0715, 0.0524, + -0.0054, -0.0066], device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 216.15, cls_loss 0.4585 cls_loss_mapping 0.0010 cls_loss_causal 0.4340 re_mapping 0.0039 re_causal 0.0117 /// teacc 99.03 lr 0.00001000 +Epoch 427, weight, value: tensor([[ 0.0390, 0.1318, -0.1816, ..., -0.1075, 0.0778, 0.0050], + [-0.1088, -0.1059, 0.1162, ..., -0.0478, -0.1151, 0.0563], + [-0.0968, -0.0714, -0.1180, ..., -0.1362, -0.0177, 0.0563], + ..., + [-0.0087, -0.1352, -0.1416, ..., -0.1143, -0.0516, 0.0772], + [-0.1334, -0.1168, -0.1178, ..., -0.0668, -0.0477, -0.1008], + [ 0.0214, -0.0320, -0.0574, ..., -0.1045, 0.0107, -0.0239]], + device='cuda:0'), grad: tensor([[ 1.1187e-03, 5.2929e-05, 2.6846e-04, ..., 2.4307e-04, + 7.3624e-04, 8.9693e-04], + [ 5.1117e-04, 1.1438e-04, 3.4666e-04, ..., 2.8992e-04, + 6.0225e-04, 1.5297e-03], + [-3.2997e-04, 9.9182e-05, -1.6136e-03, ..., -2.8954e-03, + -6.0310e-03, -1.9932e-03], + ..., + [ 2.9683e-04, 1.2219e-04, 3.4618e-04, ..., 1.9276e-04, + 6.0892e-04, 9.2125e-04], + [-1.3399e-03, 9.6977e-05, 3.7742e-04, ..., 1.0948e-03, + 2.8954e-03, 1.2407e-03], + [ 4.4107e-04, 8.8036e-05, 2.9731e-04, ..., 1.4520e-04, + 4.2677e-04, 1.3657e-03]], device='cuda:0') +Epoch 427, bias, value: tensor([ 0.0095, 0.0063, -0.0043, -0.0172, 0.0078, -0.0138, -0.0144, -0.0181, + -0.0051, 0.0085], device='cuda:0'), grad: tensor([ 0.0118, 0.0126, -0.0485, -0.0411, 0.0166, -0.0110, 0.0184, 0.0120, + 0.0163, 0.0129], device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 216.55, cls_loss 0.4375 cls_loss_mapping 0.0009 cls_loss_causal 0.4130 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.00 lr 0.00001000 +Epoch 428, weight, value: tensor([[ 0.0389, 0.1319, -0.1815, ..., -0.1076, 0.0779, 0.0048], + [-0.1088, -0.1060, 0.1164, ..., -0.0477, -0.1150, 0.0564], + [-0.0969, -0.0716, -0.1181, ..., -0.1362, -0.0176, 0.0563], + ..., + [-0.0088, -0.1353, -0.1416, ..., -0.1142, -0.0516, 0.0772], + [-0.1334, -0.1169, -0.1180, ..., -0.0668, -0.0477, -0.1008], + [ 0.0215, -0.0320, -0.0573, ..., -0.1046, 0.0107, -0.0237]], + device='cuda:0'), grad: tensor([[-1.7214e-04, -4.2677e-04, 6.6185e-04, ..., 1.1516e-04, + -2.0194e-04, 2.4170e-05], + [ 7.4720e-04, 7.3552e-05, 1.0138e-03, ..., 8.2791e-05, + 2.8229e-04, 5.4026e-04], + [ 4.5848e-04, 4.3869e-05, 6.3753e-04, ..., 6.2227e-05, + 1.8215e-04, 1.0918e-02], + ..., + [ 3.0098e-03, 1.8573e-04, 4.5919e-04, ..., 2.4319e-05, + 6.9284e-04, -9.1782e-03], + [ 7.6056e-04, 1.0872e-03, 7.2670e-04, ..., 7.2658e-05, + 2.8658e-04, 2.0370e-03], + [-1.8466e-04, 1.6856e-04, -1.4992e-03, ..., 3.9011e-05, + 1.3936e-04, 2.6189e-06]], device='cuda:0') +Epoch 428, bias, value: tensor([ 0.0095, 0.0064, -0.0045, -0.0170, 0.0077, -0.0139, -0.0142, -0.0182, + -0.0051, 0.0085], device='cuda:0'), grad: tensor([ 0.0115, 0.0150, -0.0158, -0.0438, 0.0158, 0.0091, -0.0197, 0.0055, + 0.0325, -0.0102], device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 216.34, cls_loss 0.4381 cls_loss_mapping 0.0007 cls_loss_causal 0.4150 re_mapping 0.0039 re_causal 0.0119 /// teacc 98.95 lr 0.00001000 +Epoch 429, weight, value: tensor([[ 0.0389, 0.1318, -0.1815, ..., -0.1076, 0.0779, 0.0048], + [-0.1089, -0.1059, 0.1165, ..., -0.0477, -0.1150, 0.0563], + [-0.0969, -0.0715, -0.1182, ..., -0.1362, -0.0177, 0.0562], + ..., + [-0.0087, -0.1354, -0.1416, ..., -0.1144, -0.0514, 0.0771], + [-0.1333, -0.1169, -0.1180, ..., -0.0667, -0.0479, -0.1008], + [ 0.0214, -0.0321, -0.0576, ..., -0.1048, 0.0107, -0.0238]], + device='cuda:0'), grad: tensor([[ 1.0605e-03, 1.3075e-03, 2.7180e-04, ..., 5.6171e-04, + 4.4608e-04, 7.5960e-04], + [ 5.8079e-04, 4.5991e-04, 3.0255e-04, ..., 6.1417e-04, + 4.3535e-04, 9.5129e-04], + [ 6.1703e-04, 1.4486e-03, 6.2227e-05, ..., -4.5371e-04, + 3.0828e-04, -1.7605e-03], + ..., + [ 5.2547e-04, 7.2765e-04, 2.9922e-04, ..., 4.6229e-04, + 3.3927e-04, 7.8011e-04], + [ 1.0672e-03, 1.1234e-03, 2.1172e-04, ..., 5.0735e-04, + 2.8658e-04, 4.5037e-04], + [-2.4509e-03, 8.9169e-04, 3.0875e-04, ..., -2.7485e-03, + -2.7275e-03, -5.6648e-03]], device='cuda:0') +Epoch 429, bias, value: tensor([ 0.0095, 0.0065, -0.0045, -0.0170, 0.0076, -0.0139, -0.0142, -0.0181, + -0.0049, 0.0084], device='cuda:0'), grad: tensor([ 0.0171, 0.0119, 0.0086, 0.0102, 0.0013, -0.0309, -0.0117, -0.0140, + 0.0158, -0.0082], device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 216.57, cls_loss 0.4412 cls_loss_mapping 0.0009 cls_loss_causal 0.4205 re_mapping 0.0039 re_causal 0.0121 /// teacc 98.99 lr 0.00001000 +Epoch 430, weight, value: tensor([[ 0.0390, 0.1319, -0.1816, ..., -0.1077, 0.0779, 0.0048], + [-0.1090, -0.1060, 0.1165, ..., -0.0478, -0.1150, 0.0563], + [-0.0969, -0.0715, -0.1183, ..., -0.1362, -0.0177, 0.0563], + ..., + [-0.0087, -0.1355, -0.1415, ..., -0.1143, -0.0515, 0.0772], + [-0.1333, -0.1170, -0.1180, ..., -0.0666, -0.0480, -0.1007], + [ 0.0214, -0.0319, -0.0575, ..., -0.1048, 0.0106, -0.0239]], + device='cuda:0'), grad: tensor([[-1.4200e-03, -3.0575e-03, -4.3631e-04, ..., -1.4439e-03, + -1.6436e-05, -1.2159e-03], + [ 3.1686e-04, 2.4343e-04, 4.8161e-04, ..., 1.4949e-04, + 3.2503e-07, 1.3456e-03], + [ 1.6272e-04, 1.0920e-04, 2.9612e-04, ..., 7.6294e-05, + 1.3914e-06, 9.1887e-04], + ..., + [-2.5883e-05, 1.3244e-04, 3.5000e-04, ..., 7.8321e-05, + 5.6298e-07, 1.3053e-04], + [ 3.1304e-04, 3.6669e-04, 4.6659e-04, ..., 1.9479e-04, + 1.1716e-06, 1.0033e-03], + [-6.0654e-04, 3.6573e-04, -2.3937e-03, ..., 1.6439e-04, + 9.5963e-06, -1.7462e-03]], device='cuda:0') +Epoch 430, bias, value: tensor([ 0.0094, 0.0064, -0.0045, -0.0171, 0.0075, -0.0138, -0.0141, -0.0181, + -0.0048, 0.0083], device='cuda:0'), grad: tensor([-0.0154, 0.0127, 0.0081, -0.0242, 0.0058, 0.0107, 0.0101, -0.0024, + 0.0094, -0.0147], device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 216.45, cls_loss 0.4559 cls_loss_mapping 0.0009 cls_loss_causal 0.4278 re_mapping 0.0038 re_causal 0.0121 /// teacc 98.99 lr 0.00001000 +Epoch 431, weight, value: tensor([[ 0.0390, 0.1319, -0.1818, ..., -0.1077, 0.0778, 0.0048], + [-0.1091, -0.1060, 0.1165, ..., -0.0478, -0.1151, 0.0562], + [-0.0970, -0.0716, -0.1184, ..., -0.1363, -0.0178, 0.0562], + ..., + [-0.0086, -0.1356, -0.1416, ..., -0.1142, -0.0516, 0.0772], + [-0.1332, -0.1171, -0.1180, ..., -0.0667, -0.0478, -0.1008], + [ 0.0215, -0.0320, -0.0575, ..., -0.1048, 0.0106, -0.0239]], + device='cuda:0'), grad: tensor([[ 2.9144e-03, 1.4286e-03, 4.6897e-04, ..., 1.5199e-04, + 1.0757e-03, 1.1396e-04], + [-1.8368e-03, -1.0633e-03, 1.4758e-04, ..., -2.5177e-04, + 2.9135e-04, 9.1612e-05], + [-3.6850e-03, 3.1400e-04, 1.6224e-04, ..., 1.5736e-05, + -4.2114e-03, 1.4710e-04], + ..., + [ 2.4662e-03, 1.1259e-04, 5.8711e-05, ..., 7.4841e-06, + 1.1520e-03, 1.6594e-04], + [ 1.0338e-02, 4.1485e-04, 3.8576e-04, ..., -1.9088e-05, + 9.3508e-04, 1.2088e-04], + [-1.5205e-02, 2.3556e-04, 2.9635e-04, ..., 1.5900e-05, + -2.4471e-03, 1.7798e-04]], device='cuda:0') +Epoch 431, bias, value: tensor([ 0.0094, 0.0063, -0.0045, -0.0170, 0.0076, -0.0139, -0.0141, -0.0181, + -0.0049, 0.0085], device='cuda:0'), grad: tensor([ 0.0259, -0.0243, -0.0218, -0.0089, 0.0186, -0.0060, 0.0163, 0.0124, + 0.0236, -0.0357], device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 216.73, cls_loss 0.4403 cls_loss_mapping 0.0009 cls_loss_causal 0.4102 re_mapping 0.0037 re_causal 0.0113 /// teacc 98.95 lr 0.00001000 +Epoch 432, weight, value: tensor([[ 0.0389, 0.1319, -0.1819, ..., -0.1078, 0.0778, 0.0048], + [-0.1090, -0.1059, 0.1166, ..., -0.0476, -0.1151, 0.0564], + [-0.0967, -0.0716, -0.1184, ..., -0.1364, -0.0178, 0.0561], + ..., + [-0.0087, -0.1357, -0.1415, ..., -0.1144, -0.0516, 0.0773], + [-0.1334, -0.1173, -0.1182, ..., -0.0667, -0.0477, -0.1009], + [ 0.0216, -0.0320, -0.0576, ..., -0.1048, 0.0106, -0.0238]], + device='cuda:0'), grad: tensor([[ 7.4022e-06, 3.6340e-06, 3.3259e-04, ..., 4.5728e-07, + -5.0711e-07, 6.0987e-04], + [ 2.5034e-05, -3.4720e-05, -9.2983e-04, ..., 6.0536e-08, + 2.2557e-06, -1.1177e-03], + [ 1.0407e-04, -8.3912e-07, 3.6287e-04, ..., 1.0384e-07, + 7.1758e-07, -3.3741e-03], + ..., + [ 8.9931e-04, 3.3602e-06, 4.0126e-04, ..., 8.5216e-08, + 2.1923e-04, -2.5964e-04], + [ 4.5747e-05, 1.8880e-05, 4.7874e-04, ..., 3.1702e-06, + 1.7256e-05, 7.3814e-04], + [-1.9627e-03, 4.5486e-06, 3.4380e-04, ..., 3.5018e-07, + -6.6423e-04, 7.0763e-04]], device='cuda:0') +Epoch 432, bias, value: tensor([ 0.0094, 0.0064, -0.0045, -0.0170, 0.0076, -0.0139, -0.0142, -0.0180, + -0.0051, 0.0085], device='cuda:0'), grad: tensor([ 0.0069, -0.0096, -0.0224, -0.0218, 0.0105, 0.0062, 0.0101, 0.0078, + 0.0076, 0.0047], device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 216.16, cls_loss 0.4551 cls_loss_mapping 0.0009 cls_loss_causal 0.4245 re_mapping 0.0038 re_causal 0.0114 /// teacc 98.97 lr 0.00001000 +Epoch 433, weight, value: tensor([[ 0.0389, 0.1320, -0.1819, ..., -0.1076, 0.0779, 0.0047], + [-0.1089, -0.1058, 0.1164, ..., -0.0476, -0.1152, 0.0564], + [-0.0966, -0.0714, -0.1184, ..., -0.1363, -0.0176, 0.0563], + ..., + [-0.0088, -0.1358, -0.1415, ..., -0.1146, -0.0516, 0.0772], + [-0.1334, -0.1174, -0.1182, ..., -0.0666, -0.0478, -0.1010], + [ 0.0217, -0.0321, -0.0577, ..., -0.1049, 0.0107, -0.0238]], + device='cuda:0'), grad: tensor([[ 7.2861e-04, 1.6999e-04, 2.3186e-04, ..., 4.9658e-06, + 2.5868e-05, 1.1663e-03], + [ 5.0068e-04, 3.3170e-05, 8.9645e-05, ..., 1.2806e-06, + 7.8380e-05, 2.0599e-03], + [-9.9869e-03, -1.4706e-03, 1.8084e-04, ..., 2.6245e-06, + 6.7532e-05, -6.4774e-03], + ..., + [ 1.0157e-03, 1.3864e-04, 3.3498e-04, ..., 7.3668e-07, + -4.1556e-04, 1.4610e-03], + [ 2.2182e-03, 3.1996e-04, 8.0395e-04, ..., 1.9774e-05, + 4.0919e-05, 2.7847e-03], + [ 4.2892e-04, 1.5867e-04, 4.7636e-04, ..., 1.3476e-06, + 8.7976e-05, -3.7360e-04]], device='cuda:0') +Epoch 433, bias, value: tensor([ 0.0094, 0.0064, -0.0044, -0.0172, 0.0075, -0.0137, -0.0142, -0.0180, + -0.0050, 0.0086], device='cuda:0'), grad: tensor([ 0.0100, 0.0135, -0.0515, -0.0267, 0.0137, 0.0326, -0.0199, -0.0022, + 0.0223, 0.0082], device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 216.60, cls_loss 0.4457 cls_loss_mapping 0.0010 cls_loss_causal 0.4217 re_mapping 0.0037 re_causal 0.0111 /// teacc 98.98 lr 0.00001000 +Epoch 434, weight, value: tensor([[ 0.0391, 0.1320, -0.1819, ..., -0.1076, 0.0779, 0.0046], + [-0.1089, -0.1057, 0.1165, ..., -0.0476, -0.1151, 0.0565], + [-0.0965, -0.0714, -0.1184, ..., -0.1363, -0.0174, 0.0564], + ..., + [-0.0088, -0.1359, -0.1414, ..., -0.1147, -0.0517, 0.0771], + [-0.1335, -0.1174, -0.1182, ..., -0.0666, -0.0478, -0.1011], + [ 0.0218, -0.0322, -0.0577, ..., -0.1047, 0.0107, -0.0237]], + device='cuda:0'), grad: tensor([[ 1.9894e-03, -8.7118e-04, 1.3001e-05, ..., 1.2636e-04, + 2.2888e-03, -6.9656e-03], + [ 2.6989e-04, 6.8128e-05, 1.8299e-05, ..., 1.3804e-04, + 3.4165e-04, 2.3232e-03], + [ 2.8920e-04, 3.5405e-04, 3.0294e-05, ..., 7.1287e-04, + 2.6512e-04, 4.6844e-03], + ..., + [ 3.2997e-04, 9.3043e-05, 5.0843e-05, ..., 9.2745e-05, + 3.1209e-04, 2.0466e-03], + [ 3.0255e-04, 9.5785e-05, -8.5211e-04, ..., 2.6321e-04, + 5.2834e-04, 1.3227e-03], + [-2.5520e-03, 2.9355e-05, 4.4584e-04, ..., 1.3614e-04, + -3.0251e-03, 1.4839e-03]], device='cuda:0') +Epoch 434, bias, value: tensor([ 0.0094, 0.0065, -0.0043, -0.0174, 0.0076, -0.0137, -0.0144, -0.0181, + -0.0050, 0.0086], device='cuda:0'), grad: tensor([ 0.0044, 0.0172, 0.0261, 0.0143, -0.0099, -0.0231, -0.0091, 0.0147, + 0.0072, -0.0418], device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 216.34, cls_loss 0.4694 cls_loss_mapping 0.0009 cls_loss_causal 0.4392 re_mapping 0.0035 re_causal 0.0112 /// teacc 98.96 lr 0.00001000 +Epoch 435, weight, value: tensor([[ 0.0388, 0.1320, -0.1819, ..., -0.1077, 0.0776, 0.0046], + [-0.1088, -0.1056, 0.1164, ..., -0.0477, -0.1150, 0.0564], + [-0.0964, -0.0713, -0.1182, ..., -0.1363, -0.0172, 0.0564], + ..., + [-0.0088, -0.1357, -0.1412, ..., -0.1145, -0.0517, 0.0772], + [-0.1335, -0.1175, -0.1182, ..., -0.0667, -0.0478, -0.1011], + [ 0.0217, -0.0322, -0.0578, ..., -0.1048, 0.0107, -0.0236]], + device='cuda:0'), grad: tensor([[ 3.3545e-04, -3.0128e-07, 3.1263e-05, ..., 1.2442e-06, + 7.1898e-06, 6.8378e-04], + [ 4.1461e-04, 4.1910e-09, -3.8218e-04, ..., -6.2525e-05, + 1.5078e-06, 6.3848e-04], + [-2.9392e-03, 1.2387e-07, 4.0442e-05, ..., 6.0126e-06, + -9.4026e-06, -4.6692e-03], + ..., + [ 8.2159e-04, 3.7253e-09, 4.8113e-04, ..., 1.7732e-05, + 1.7989e-04, 3.2978e-03], + [ 2.5654e-04, 2.2464e-06, 1.3721e-04, ..., 1.7136e-05, + 2.4691e-05, -8.4534e-03], + [-4.6015e-04, 6.6590e-08, -6.9284e-04, ..., 5.3868e-06, + -3.3689e-04, 5.0354e-03]], device='cuda:0') +Epoch 435, bias, value: tensor([ 0.0094, 0.0064, -0.0042, -0.0175, 0.0074, -0.0137, -0.0142, -0.0180, + -0.0050, 0.0086], device='cuda:0'), grad: tensor([ 0.0062, -0.0211, -0.0263, 0.0094, 0.0073, 0.0057, 0.0056, 0.0181, + -0.0216, 0.0166], device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 216.29, cls_loss 0.4581 cls_loss_mapping 0.0009 cls_loss_causal 0.4301 re_mapping 0.0037 re_causal 0.0116 /// teacc 98.97 lr 0.00001000 +Epoch 436, weight, value: tensor([[ 0.0388, 0.1321, -0.1817, ..., -0.1077, 0.0777, 0.0046], + [-0.1088, -0.1056, 0.1164, ..., -0.0477, -0.1151, 0.0565], + [-0.0964, -0.0713, -0.1183, ..., -0.1364, -0.0171, 0.0564], + ..., + [-0.0089, -0.1356, -0.1413, ..., -0.1145, -0.0518, 0.0772], + [-0.1335, -0.1176, -0.1183, ..., -0.0669, -0.0479, -0.1011], + [ 0.0215, -0.0321, -0.0579, ..., -0.1047, 0.0108, -0.0237]], + device='cuda:0'), grad: tensor([[ 3.5787e-04, 2.4900e-05, 6.3324e-04, ..., 5.7332e-06, + 5.1439e-05, 1.4296e-03], + [ 4.2748e-04, 7.9513e-05, 1.3332e-03, ..., 2.0519e-05, + 2.7966e-04, 2.9850e-03], + [ 4.2033e-04, 1.0252e-04, 1.5211e-03, ..., 4.1872e-05, + 4.0174e-04, 3.1643e-03], + ..., + [-2.3055e-04, -2.8920e-04, 5.6171e-04, ..., -3.2043e-04, + 5.9992e-05, -5.8212e-03], + [-3.8242e-03, -4.0412e-04, 5.5408e-04, ..., 2.4647e-05, + 1.2004e-04, -2.1896e-03], + [ 1.9150e-03, 3.8218e-04, -1.2379e-03, ..., 1.5008e-04, + 2.0266e-04, 1.7653e-03]], device='cuda:0') +Epoch 436, bias, value: tensor([ 0.0094, 0.0064, -0.0042, -0.0175, 0.0075, -0.0137, -0.0143, -0.0181, + -0.0050, 0.0087], device='cuda:0'), grad: tensor([ 0.0135, 0.0217, 0.0203, -0.0090, -0.0178, 0.0190, 0.0134, -0.0135, + -0.0402, -0.0074], device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 216.56, cls_loss 0.4667 cls_loss_mapping 0.0008 cls_loss_causal 0.4430 re_mapping 0.0037 re_causal 0.0118 /// teacc 98.97 lr 0.00001000 +Epoch 437, weight, value: tensor([[ 0.0388, 0.1321, -0.1818, ..., -0.1077, 0.0778, 0.0045], + [-0.1088, -0.1056, 0.1165, ..., -0.0477, -0.1152, 0.0565], + [-0.0964, -0.0714, -0.1184, ..., -0.1365, -0.0171, 0.0563], + ..., + [-0.0089, -0.1357, -0.1415, ..., -0.1144, -0.0520, 0.0772], + [-0.1333, -0.1176, -0.1182, ..., -0.0668, -0.0478, -0.1008], + [ 0.0216, -0.0321, -0.0579, ..., -0.1045, 0.0108, -0.0237]], + device='cuda:0'), grad: tensor([[ 1.5175e-04, 1.3018e-04, 4.7660e-04, ..., 3.2640e-04, + 1.5342e-04, -5.5885e-03], + [ 2.8443e-04, 1.6379e-04, 5.8603e-04, ..., -3.1590e-04, + 1.5914e-04, -3.2482e-03], + [ 2.5439e-04, 1.2338e-04, 4.5919e-04, ..., 1.2732e-04, + 1.5020e-04, 1.5717e-03], + ..., + [ 2.9755e-04, 2.5034e-04, 6.6710e-04, ..., 1.3828e-04, + 1.6344e-04, 1.3151e-03], + [ 3.1471e-04, 3.5191e-04, 7.7534e-04, ..., 1.3220e-04, + 1.7083e-04, 1.7929e-03], + [-2.4402e-04, -2.3460e-03, -3.3054e-03, ..., 8.5056e-05, + 1.9193e-04, 1.6527e-03]], device='cuda:0') +Epoch 437, bias, value: tensor([ 0.0094, 0.0064, -0.0043, -0.0174, 0.0075, -0.0137, -0.0143, -0.0182, + -0.0048, 0.0085], device='cuda:0'), grad: tensor([-0.0168, 0.0039, 0.0129, 0.0168, 0.0190, -0.0142, -0.0117, 0.0150, + -0.0146, -0.0102], device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 217.01, cls_loss 0.4267 cls_loss_mapping 0.0008 cls_loss_causal 0.3958 re_mapping 0.0037 re_causal 0.0113 /// teacc 98.98 lr 0.00001000 +Epoch 438, weight, value: tensor([[ 0.0388, 0.1322, -0.1818, ..., -0.1076, 0.0779, 0.0047], + [-0.1090, -0.1057, 0.1163, ..., -0.0476, -0.1152, 0.0565], + [-0.0965, -0.0714, -0.1184, ..., -0.1365, -0.0171, 0.0564], + ..., + [-0.0088, -0.1357, -0.1415, ..., -0.1144, -0.0521, 0.0773], + [-0.1332, -0.1177, -0.1183, ..., -0.0669, -0.0480, -0.1009], + [ 0.0215, -0.0319, -0.0579, ..., -0.1046, 0.0110, -0.0238]], + device='cuda:0'), grad: tensor([[ 3.1185e-04, -1.3418e-03, -4.0483e-04, ..., 2.4962e-04, + -2.6369e-04, -1.2169e-03], + [ 6.9189e-04, 6.0511e-04, -3.1948e-03, ..., -1.5240e-03, + 2.1970e-04, -1.5402e-03], + [ 2.0485e-03, -1.3027e-03, 1.6344e-04, ..., 2.0659e-04, + 8.6164e-04, -1.0748e-03], + ..., + [ 1.3046e-03, 5.7697e-04, -1.2550e-03, ..., 8.1122e-05, + 3.6335e-04, 7.9632e-04], + [ 1.8368e-03, 2.6951e-03, 4.5547e-03, ..., 1.6346e-03, + 6.5041e-04, 5.8174e-03], + [ 1.2674e-03, 1.6747e-03, 9.4938e-04, ..., 1.7679e-04, + 9.4795e-04, 1.6022e-03]], device='cuda:0') +Epoch 438, bias, value: tensor([ 0.0095, 0.0062, -0.0043, -0.0175, 0.0077, -0.0138, -0.0142, -0.0181, + -0.0048, 0.0084], device='cuda:0'), grad: tensor([-0.0135, 0.0105, -0.0288, 0.0410, -0.0164, -0.0361, -0.0150, -0.0124, + 0.0515, 0.0191], device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 215.94, cls_loss 0.4639 cls_loss_mapping 0.0008 cls_loss_causal 0.4317 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.03 lr 0.00001000 +Epoch 439, weight, value: tensor([[ 0.0389, 0.1320, -0.1818, ..., -0.1078, 0.0778, 0.0046], + [-0.1091, -0.1058, 0.1164, ..., -0.0476, -0.1152, 0.0566], + [-0.0965, -0.0712, -0.1186, ..., -0.1365, -0.0171, 0.0564], + ..., + [-0.0087, -0.1356, -0.1413, ..., -0.1143, -0.0518, 0.0774], + [-0.1333, -0.1178, -0.1184, ..., -0.0670, -0.0481, -0.1011], + [ 0.0215, -0.0318, -0.0579, ..., -0.1046, 0.0109, -0.0239]], + device='cuda:0'), grad: tensor([[-2.3827e-05, 1.0710e-06, -1.4744e-03, ..., 1.6857e-06, + 8.1733e-06, -7.4327e-05], + [-1.4505e-03, 1.2349e-06, 1.8752e-04, ..., -5.7369e-07, + 1.6466e-05, 2.5959e-03], + [ 3.6567e-05, 4.7493e-04, 1.8418e-04, ..., 8.9221e-07, + -1.2660e-04, -8.2626e-03], + ..., + [ 1.9610e-04, 9.6858e-06, -7.4482e-04, ..., 2.6077e-08, + 1.8433e-05, 1.0815e-03], + [ 1.4699e-04, 5.1297e-06, 7.8964e-04, ..., 4.3288e-06, + 6.4410e-06, 7.1383e-04], + [ 3.2926e-04, 4.9882e-06, 2.0695e-04, ..., 1.7695e-07, + 3.2689e-06, 4.6396e-04]], device='cuda:0') +Epoch 439, bias, value: tensor([ 0.0094, 0.0063, -0.0042, -0.0175, 0.0077, -0.0139, -0.0141, -0.0180, + -0.0049, 0.0084], device='cuda:0'), grad: tensor([-0.0227, -0.0123, -0.0198, 0.0174, -0.0162, 0.0081, 0.0108, -0.0122, + 0.0311, 0.0158], device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 216.07, cls_loss 0.4544 cls_loss_mapping 0.0009 cls_loss_causal 0.4243 re_mapping 0.0036 re_causal 0.0115 /// teacc 98.99 lr 0.00001000 +Epoch 440, weight, value: tensor([[ 0.0390, 0.1319, -0.1818, ..., -0.1078, 0.0778, 0.0045], + [-0.1092, -0.1058, 0.1164, ..., -0.0475, -0.1153, 0.0566], + [-0.0966, -0.0712, -0.1187, ..., -0.1364, -0.0171, 0.0563], + ..., + [-0.0087, -0.1357, -0.1414, ..., -0.1142, -0.0519, 0.0775], + [-0.1335, -0.1178, -0.1185, ..., -0.0671, -0.0483, -0.1013], + [ 0.0216, -0.0317, -0.0579, ..., -0.1047, 0.0110, -0.0238]], + device='cuda:0'), grad: tensor([[-2.7065e-03, -4.4746e-03, -1.0204e-03, ..., 5.2899e-07, + -3.7155e-03, 4.8804e-04], + [ 5.1308e-04, -2.4567e-03, -3.9043e-03, ..., 4.0978e-08, + -1.5008e-04, -1.0843e-03], + [ 3.0145e-05, 2.2678e-03, 6.9714e-04, ..., -9.1456e-07, + 3.9148e-04, 3.7169e-04], + ..., + [ 1.3332e-03, 1.0681e-03, 1.6279e-03, ..., 3.8929e-07, + 5.1737e-04, 2.0199e-03], + [ 1.3905e-03, 2.0046e-03, 9.2649e-04, ..., 1.0245e-07, + 1.2560e-03, 1.2665e-03], + [-1.2312e-03, 3.8600e-04, -8.0490e-04, ..., 3.7253e-09, + 1.7911e-05, -9.2649e-04]], device='cuda:0') +Epoch 440, bias, value: tensor([ 0.0094, 0.0062, -0.0043, -0.0174, 0.0077, -0.0139, -0.0141, -0.0179, + -0.0051, 0.0085], device='cuda:0'), grad: tensor([-0.0137, -0.0046, -0.0373, -0.0094, 0.0090, 0.0097, 0.0075, 0.0202, + 0.0203, -0.0018], device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 216.12, cls_loss 0.4665 cls_loss_mapping 0.0009 cls_loss_causal 0.4289 re_mapping 0.0038 re_causal 0.0121 /// teacc 98.94 lr 0.00001000 +Epoch 441, weight, value: tensor([[ 0.0389, 0.1319, -0.1818, ..., -0.1080, 0.0778, 0.0045], + [-0.1094, -0.1059, 0.1165, ..., -0.0476, -0.1154, 0.0566], + [-0.0966, -0.0712, -0.1187, ..., -0.1362, -0.0171, 0.0565], + ..., + [-0.0087, -0.1356, -0.1414, ..., -0.1142, -0.0521, 0.0773], + [-0.1334, -0.1178, -0.1186, ..., -0.0670, -0.0483, -0.1013], + [ 0.0217, -0.0315, -0.0578, ..., -0.1045, 0.0112, -0.0238]], + device='cuda:0'), grad: tensor([[ 4.4703e-04, -7.9274e-06, 3.8594e-05, ..., 4.7609e-06, + 1.9002e-04, 7.9966e-04], + [ 2.4259e-04, 5.9038e-05, 3.1888e-05, ..., 7.7337e-06, + 1.0633e-04, -6.7711e-04], + [-1.7456e-02, -2.6703e-04, 1.5147e-05, ..., 1.2387e-06, + 2.6822e-04, -1.3596e-02], + ..., + [ 1.8036e-02, 6.4850e-05, 6.0290e-05, ..., 4.6082e-06, + 1.8847e-04, 1.3336e-02], + [ 2.0523e-03, 1.8859e-04, 8.0109e-03, ..., 6.3438e-03, + 1.1339e-03, 2.0962e-03], + [-1.0071e-02, -1.0405e-03, -4.1428e-03, ..., -3.1114e-04, + -5.9891e-03, -5.7869e-03]], device='cuda:0') +Epoch 441, bias, value: tensor([ 0.0093, 0.0062, -0.0042, -0.0173, 0.0077, -0.0139, -0.0141, -0.0180, + -0.0050, 0.0085], device='cuda:0'), grad: tensor([ 0.0122, 0.0024, -0.0225, -0.0132, 0.0350, -0.0129, 0.0220, 0.0430, + 0.0092, -0.0752], device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 216.28, cls_loss 0.4406 cls_loss_mapping 0.0008 cls_loss_causal 0.4109 re_mapping 0.0040 re_causal 0.0123 /// teacc 98.95 lr 0.00001000 +Epoch 442, weight, value: tensor([[ 0.0390, 0.1320, -0.1816, ..., -0.1078, 0.0777, 0.0046], + [-0.1095, -0.1060, 0.1165, ..., -0.0476, -0.1153, 0.0565], + [-0.0965, -0.0711, -0.1187, ..., -0.1360, -0.0171, 0.0566], + ..., + [-0.0087, -0.1357, -0.1416, ..., -0.1142, -0.0520, 0.0772], + [-0.1334, -0.1177, -0.1184, ..., -0.0669, -0.0484, -0.1013], + [ 0.0216, -0.0315, -0.0578, ..., -0.1044, 0.0112, -0.0237]], + device='cuda:0'), grad: tensor([[ 0.0017, 0.0003, 0.0006, ..., 0.0006, 0.0006, 0.0027], + [ 0.0010, 0.0003, 0.0008, ..., 0.0005, 0.0001, 0.0021], + [ 0.0017, 0.0002, 0.0005, ..., 0.0008, 0.0005, -0.0036], + ..., + [-0.0059, -0.0014, -0.0031, ..., -0.0003, -0.0022, -0.0067], + [ 0.0019, 0.0004, 0.0007, ..., 0.0002, 0.0002, 0.0053], + [ 0.0013, 0.0002, 0.0006, ..., 0.0003, 0.0004, 0.0025]], + device='cuda:0') +Epoch 442, bias, value: tensor([ 0.0095, 0.0062, -0.0041, -0.0173, 0.0076, -0.0138, -0.0143, -0.0181, + -0.0051, 0.0086], device='cuda:0'), grad: tensor([ 0.0169, 0.0228, -0.0061, 0.0056, -0.0419, 0.0002, -0.0063, -0.0539, + 0.0419, 0.0207], device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 216.24, cls_loss 0.4404 cls_loss_mapping 0.0008 cls_loss_causal 0.4126 re_mapping 0.0037 re_causal 0.0115 /// teacc 98.97 lr 0.00001000 +Epoch 443, weight, value: tensor([[ 0.0389, 0.1319, -0.1815, ..., -0.1078, 0.0777, 0.0046], + [-0.1095, -0.1062, 0.1163, ..., -0.0478, -0.1154, 0.0564], + [-0.0966, -0.0712, -0.1186, ..., -0.1361, -0.0171, 0.0567], + ..., + [-0.0085, -0.1359, -0.1416, ..., -0.1144, -0.0521, 0.0772], + [-0.1335, -0.1177, -0.1183, ..., -0.0669, -0.0485, -0.1012], + [ 0.0216, -0.0315, -0.0578, ..., -0.1044, 0.0113, -0.0236]], + device='cuda:0'), grad: tensor([[ 9.5248e-05, -5.8079e-04, 3.5390e-07, ..., -7.7903e-05, + -2.2936e-04, -3.7384e-04], + [ 4.5121e-05, 4.8839e-06, -8.8476e-07, ..., 9.7416e-07, + 2.2858e-05, 2.8539e-04], + [ 3.8266e-04, 1.0824e-04, 2.6971e-06, ..., 1.6809e-05, + 2.2197e-04, 3.0613e-04], + ..., + [ 1.3618e-03, 3.0130e-05, 2.0802e-05, ..., 4.9584e-06, + 6.0797e-05, 7.4089e-05], + [ 1.5473e-04, 2.0552e-04, 2.5392e-04, ..., 8.0287e-05, + 1.0628e-04, 3.0231e-04], + [ 1.8060e-04, 1.4114e-04, -5.9754e-05, ..., 2.0862e-05, + 4.4346e-04, 4.1580e-04]], device='cuda:0') +Epoch 443, bias, value: tensor([ 0.0093, 0.0063, -0.0042, -0.0171, 0.0077, -0.0139, -0.0143, -0.0182, + -0.0051, 0.0086], device='cuda:0'), grad: tensor([ 0.0025, 0.0042, 0.0039, -0.0282, 0.0086, 0.0038, -0.0070, 0.0048, + 0.0040, 0.0035], device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 216.09, cls_loss 0.4564 cls_loss_mapping 0.0009 cls_loss_causal 0.4272 re_mapping 0.0037 re_causal 0.0115 /// teacc 98.97 lr 0.00001000 +Epoch 444, weight, value: tensor([[ 0.0390, 0.1319, -0.1815, ..., -0.1078, 0.0777, 0.0046], + [-0.1096, -0.1062, 0.1164, ..., -0.0478, -0.1155, 0.0564], + [-0.0966, -0.0712, -0.1187, ..., -0.1360, -0.0171, 0.0566], + ..., + [-0.0085, -0.1359, -0.1415, ..., -0.1144, -0.0520, 0.0772], + [-0.1334, -0.1177, -0.1183, ..., -0.0668, -0.0486, -0.1011], + [ 0.0215, -0.0316, -0.0579, ..., -0.1044, 0.0113, -0.0236]], + device='cuda:0'), grad: tensor([[ 4.3392e-04, 4.0978e-08, -3.4847e-03, ..., 1.3113e-06, + 2.1625e-04, 9.6750e-04], + [ 2.9016e-04, 5.0105e-07, 1.5860e-03, ..., 6.2175e-06, + 2.9922e-04, 1.1930e-03], + [-9.6893e-04, 8.3894e-06, 1.0557e-03, ..., 8.0988e-06, + 2.0528e-04, -3.8300e-03], + ..., + [ 5.0449e-04, 4.4890e-07, -1.0281e-03, ..., 5.5879e-09, + -1.9474e-03, -5.0507e-03], + [ 2.3770e-04, 9.0376e-06, -1.1826e-03, ..., 1.6719e-05, + 1.8179e-04, 6.7425e-04], + [ 3.1471e-04, -2.0415e-06, -1.1044e-03, ..., 8.9407e-08, + 2.6226e-04, 2.7523e-03]], device='cuda:0') +Epoch 444, bias, value: tensor([ 0.0093, 0.0064, -0.0043, -0.0171, 0.0077, -0.0140, -0.0143, -0.0181, + -0.0050, 0.0084], device='cuda:0'), grad: tensor([-0.0375, 0.0327, 0.0030, 0.0249, 0.0243, 0.0175, -0.0085, -0.0462, + -0.0099, -0.0002], device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 216.33, cls_loss 0.4451 cls_loss_mapping 0.0009 cls_loss_causal 0.4204 re_mapping 0.0038 re_causal 0.0116 /// teacc 98.99 lr 0.00001000 +Epoch 445, weight, value: tensor([[ 0.0390, 0.1319, -0.1814, ..., -0.1078, 0.0778, 0.0047], + [-0.1097, -0.1062, 0.1163, ..., -0.0478, -0.1155, 0.0564], + [-0.0965, -0.0710, -0.1186, ..., -0.1360, -0.0171, 0.0567], + ..., + [-0.0085, -0.1360, -0.1416, ..., -0.1145, -0.0522, 0.0771], + [-0.1334, -0.1177, -0.1183, ..., -0.0668, -0.0487, -0.1012], + [ 0.0216, -0.0316, -0.0578, ..., -0.1043, 0.0113, -0.0236]], + device='cuda:0'), grad: tensor([[-7.8154e-04, -3.3970e-03, 6.2132e-04, ..., 3.3528e-07, + -3.7632e-03, -1.2188e-03], + [ 4.0054e-04, 3.5763e-04, -1.2255e-03, ..., 9.3132e-09, + 6.0368e-04, -1.5950e-04], + [ 3.7551e-04, 4.6277e-04, -3.0022e-03, ..., 5.4762e-07, + 6.8426e-04, -9.9182e-03], + ..., + [ 3.4118e-04, 2.6584e-04, 3.5610e-03, ..., 1.4901e-08, + 5.9652e-04, 8.4839e-03], + [ 4.7016e-04, 4.0555e-04, 6.7759e-04, ..., 2.1718e-06, + 6.9857e-04, 6.3753e-04], + [ 5.8413e-04, 4.1938e-04, 1.8585e-04, ..., 5.4017e-08, + 7.2479e-04, 1.0386e-03]], device='cuda:0') +Epoch 445, bias, value: tensor([ 0.0094, 0.0064, -0.0041, -0.0172, 0.0077, -0.0139, -0.0142, -0.0181, + -0.0052, 0.0084], device='cuda:0'), grad: tensor([-0.0347, -0.0107, -0.0100, -0.0409, 0.0201, 0.0170, -0.0159, 0.0380, + 0.0164, 0.0206], device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 216.96, cls_loss 0.4404 cls_loss_mapping 0.0008 cls_loss_causal 0.4156 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.04 lr 0.00001000 +Epoch 446, weight, value: tensor([[ 0.0390, 0.1321, -0.1815, ..., -0.1080, 0.0779, 0.0047], + [-0.1096, -0.1064, 0.1164, ..., -0.0477, -0.1156, 0.0566], + [-0.0965, -0.0711, -0.1188, ..., -0.1361, -0.0171, 0.0565], + ..., + [-0.0086, -0.1362, -0.1417, ..., -0.1146, -0.0522, 0.0771], + [-0.1336, -0.1178, -0.1182, ..., -0.0668, -0.0488, -0.1011], + [ 0.0215, -0.0316, -0.0578, ..., -0.1042, 0.0113, -0.0238]], + device='cuda:0'), grad: tensor([[ 6.5470e-04, 4.8310e-05, 6.9237e-04, ..., 1.0610e-04, + 3.2425e-04, -2.5302e-05], + [-4.4212e-03, 9.4473e-05, 9.8877e-03, ..., 1.1253e-02, + 4.4727e-04, 1.4938e-02], + [ 9.1743e-04, -3.4547e-04, -3.9139e-03, ..., -4.2105e-04, + -6.9559e-05, -1.4210e-03], + ..., + [ 5.2309e-04, 4.3690e-05, 9.1219e-04, ..., 4.0531e-04, + -7.1144e-04, -7.1335e-03], + [ 1.3256e-03, 5.0485e-05, -1.1663e-03, ..., -3.1166e-03, + 3.1996e-04, -3.4676e-03], + [ 2.2755e-03, 2.9042e-05, 6.3467e-04, ..., 2.2471e-04, + 2.4056e-04, 6.8436e-03]], device='cuda:0') +Epoch 446, bias, value: tensor([ 0.0093, 0.0064, -0.0043, -0.0172, 0.0077, -0.0136, -0.0142, -0.0181, + -0.0051, 0.0082], device='cuda:0'), grad: tensor([ 0.0005, 0.0253, -0.0118, -0.0337, 0.0318, -0.0103, -0.0031, -0.0162, + -0.0165, 0.0341], device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 216.37, cls_loss 0.4455 cls_loss_mapping 0.0007 cls_loss_causal 0.4165 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.03 lr 0.00001000 +Epoch 447, weight, value: tensor([[ 0.0388, 0.1320, -0.1814, ..., -0.1080, 0.0779, 0.0047], + [-0.1097, -0.1065, 0.1162, ..., -0.0479, -0.1158, 0.0565], + [-0.0965, -0.0709, -0.1188, ..., -0.1363, -0.0170, 0.0566], + ..., + [-0.0084, -0.1362, -0.1416, ..., -0.1146, -0.0521, 0.0771], + [-0.1336, -0.1176, -0.1181, ..., -0.0667, -0.0489, -0.1011], + [ 0.0214, -0.0316, -0.0577, ..., -0.1042, 0.0112, -0.0239]], + device='cuda:0'), grad: tensor([[ 4.5252e-04, -1.5628e-04, 2.0909e-04, ..., 4.6007e-06, + 1.8263e-04, 4.9257e-04], + [-1.7614e-03, 1.0721e-05, 3.1257e-04, ..., 1.6205e-07, + -1.2884e-03, -2.3689e-03], + [-1.3863e-02, 6.5088e-05, 3.0065e-04, ..., 3.1799e-05, + 1.1265e-04, -1.6088e-03], + ..., + [ 9.0933e-04, 8.8990e-05, 2.8658e-04, ..., 3.4645e-07, + 1.5450e-04, 3.9768e-04], + [ 3.5715e-04, 8.2850e-05, 4.2987e-04, ..., 1.2016e-04, + 1.2696e-04, 3.8242e-04], + [ 9.7942e-04, 5.9366e-05, -2.1019e-03, ..., 4.2468e-07, + 1.7917e-04, 3.1781e-04]], device='cuda:0') +Epoch 447, bias, value: tensor([ 0.0092, 0.0061, -0.0043, -0.0172, 0.0078, -0.0135, -0.0142, -0.0180, + -0.0050, 0.0082], device='cuda:0'), grad: tensor([ 0.0131, -0.0191, -0.0505, -0.0194, 0.0342, 0.0139, -0.0050, 0.0387, + 0.0229, -0.0289], device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 216.48, cls_loss 0.4649 cls_loss_mapping 0.0008 cls_loss_causal 0.4375 re_mapping 0.0037 re_causal 0.0119 /// teacc 99.01 lr 0.00001000 +Epoch 448, weight, value: tensor([[ 0.0388, 0.1320, -0.1815, ..., -0.1079, 0.0779, 0.0047], + [-0.1095, -0.1065, 0.1161, ..., -0.0478, -0.1156, 0.0566], + [-0.0963, -0.0708, -0.1188, ..., -0.1362, -0.0170, 0.0567], + ..., + [-0.0084, -0.1363, -0.1417, ..., -0.1148, -0.0523, 0.0770], + [-0.1335, -0.1177, -0.1181, ..., -0.0667, -0.0488, -0.1011], + [ 0.0213, -0.0315, -0.0576, ..., -0.1042, 0.0112, -0.0241]], + device='cuda:0'), grad: tensor([[ 5.3674e-05, 8.8692e-05, 4.2629e-04, ..., 1.8626e-07, + -1.0341e-05, 1.6224e-04], + [ 3.6508e-05, 1.7807e-05, 5.4359e-04, ..., 2.2724e-07, + 4.1537e-07, 1.8287e-04], + [-1.0055e-04, 1.0401e-04, -3.7270e-03, ..., -5.0329e-06, + 2.0508e-06, -1.0986e-03], + ..., + [-1.5497e-03, 4.8250e-05, 4.3511e-04, ..., 3.0249e-06, + -5.5879e-05, -1.6947e-03], + [-5.3883e-04, -1.5821e-03, -2.8491e-04, ..., 1.0151e-06, + 1.3877e-06, -4.4465e-04], + [ 1.4019e-03, 1.2851e-04, 4.9305e-04, ..., 1.4901e-08, + 5.8144e-05, 1.6956e-03]], device='cuda:0') +Epoch 448, bias, value: tensor([ 0.0092, 0.0061, -0.0042, -0.0171, 0.0080, -0.0137, -0.0141, -0.0182, + -0.0051, 0.0082], device='cuda:0'), grad: tensor([ 0.0078, 0.0109, -0.0543, 0.0100, 0.0094, 0.0139, 0.0064, 0.0029, + -0.0325, 0.0254], device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 216.24, cls_loss 0.4522 cls_loss_mapping 0.0008 cls_loss_causal 0.4238 re_mapping 0.0038 re_causal 0.0121 /// teacc 99.03 lr 0.00001000 +Epoch 449, weight, value: tensor([[ 0.0388, 0.1319, -0.1816, ..., -0.1079, 0.0781, 0.0048], + [-0.1095, -0.1066, 0.1161, ..., -0.0477, -0.1157, 0.0565], + [-0.0965, -0.0707, -0.1187, ..., -0.1363, -0.0171, 0.0567], + ..., + [-0.0084, -0.1362, -0.1417, ..., -0.1147, -0.0523, 0.0769], + [-0.1336, -0.1177, -0.1181, ..., -0.0668, -0.0488, -0.1011], + [ 0.0214, -0.0315, -0.0576, ..., -0.1043, 0.0114, -0.0241]], + device='cuda:0'), grad: tensor([[-4.2458e-03, -2.7733e-03, 6.9523e-04, ..., 2.3320e-05, + -3.5992e-03, -1.9970e-03], + [-2.9159e-04, 2.9698e-05, -1.9741e-03, ..., 4.2260e-05, + 2.4605e-04, -1.0424e-03], + [ 8.5068e-04, 6.4516e-04, 7.6485e-04, ..., -3.0804e-04, + 8.4209e-04, 5.5456e-04], + ..., + [ 5.8079e-04, 2.3186e-04, -8.5831e-04, ..., 6.3442e-06, + 3.4022e-04, 6.1178e-04], + [ 1.2121e-03, 2.4681e-03, 6.7854e-04, ..., -3.8177e-05, + 2.5821e-04, -4.0936e-04], + [ 2.3575e-03, 1.0595e-03, 8.5878e-04, ..., 5.0440e-06, + 1.5421e-03, 8.8692e-04]], device='cuda:0') +Epoch 449, bias, value: tensor([ 0.0092, 0.0061, -0.0041, -0.0172, 0.0079, -0.0137, -0.0141, -0.0182, + -0.0050, 0.0082], device='cuda:0'), grad: tensor([-0.0032, -0.0142, 0.0108, -0.0425, 0.0221, -0.0205, 0.0107, -0.0151, + 0.0303, 0.0216], device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 216.23, cls_loss 0.4189 cls_loss_mapping 0.0007 cls_loss_causal 0.3922 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.03 lr 0.00001000 +Epoch 450, weight, value: tensor([[ 0.0389, 0.1320, -0.1816, ..., -0.1079, 0.0783, 0.0047], + [-0.1096, -0.1066, 0.1161, ..., -0.0478, -0.1157, 0.0565], + [-0.0963, -0.0707, -0.1187, ..., -0.1362, -0.0170, 0.0568], + ..., + [-0.0085, -0.1362, -0.1416, ..., -0.1147, -0.0524, 0.0769], + [-0.1337, -0.1179, -0.1181, ..., -0.0668, -0.0488, -0.1011], + [ 0.0213, -0.0316, -0.0575, ..., -0.1044, 0.0112, -0.0241]], + device='cuda:0'), grad: tensor([[-2.9683e-04, -6.0081e-04, -1.8234e-03, ..., -7.1001e-04, + 3.5524e-05, 8.8632e-05], + [ 8.1015e-04, 7.7486e-05, 5.0879e-04, ..., 2.1324e-05, + 4.0126e-04, 1.9798e-03], + [ 1.0300e-03, 2.1160e-04, 6.9809e-04, ..., 6.4254e-05, + 2.5535e-04, 1.6804e-03], + ..., + [-2.4092e-04, 8.0526e-05, 5.5170e-04, ..., 5.6699e-06, + 4.2248e-04, 3.7718e-04], + [-2.4929e-03, -5.1727e-03, 4.8637e-04, ..., 1.3578e-04, + 1.7571e-04, 5.8889e-04], + [ 5.1451e-04, 1.8334e-04, 5.7364e-04, ..., 7.2777e-05, + 2.8324e-04, 1.2417e-03]], device='cuda:0') +Epoch 450, bias, value: tensor([ 0.0091, 0.0061, -0.0041, -0.0173, 0.0079, -0.0136, -0.0141, -0.0181, + -0.0050, 0.0083], device='cuda:0'), grad: tensor([-0.0270, 0.0144, 0.0112, -0.0527, 0.0103, 0.0305, 0.0098, 0.0113, + -0.0177, 0.0099], device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 216.48, cls_loss 0.4713 cls_loss_mapping 0.0007 cls_loss_causal 0.4432 re_mapping 0.0039 re_causal 0.0124 /// teacc 99.05 lr 0.00001000 +Epoch 451, weight, value: tensor([[ 0.0388, 0.1320, -0.1816, ..., -0.1079, 0.0782, 0.0048], + [-0.1092, -0.1065, 0.1163, ..., -0.0479, -0.1156, 0.0566], + [-0.0963, -0.0708, -0.1188, ..., -0.1361, -0.0171, 0.0568], + ..., + [-0.0085, -0.1363, -0.1418, ..., -0.1148, -0.0524, 0.0770], + [-0.1337, -0.1179, -0.1182, ..., -0.0668, -0.0489, -0.1013], + [ 0.0214, -0.0316, -0.0576, ..., -0.1043, 0.0113, -0.0241]], + device='cuda:0'), grad: tensor([[ 8.9931e-04, 1.7881e-07, 5.4026e-04, ..., 1.5460e-07, + 6.4516e-04, 2.2972e-04], + [-1.9825e-04, -4.0978e-08, -2.4300e-03, ..., 9.3132e-09, + -1.5402e-04, 4.2367e-04], + [ 8.8406e-04, -8.8848e-07, 6.2752e-04, ..., 2.9802e-08, + 6.3705e-04, 2.2781e-04], + ..., + [ 6.2370e-04, 2.1234e-07, 4.0770e-04, ..., 0.0000e+00, + 9.2387e-05, 2.0587e-04], + [ 5.9319e-04, 1.7136e-07, 1.0996e-03, ..., 7.2643e-08, + 2.4009e-04, -2.1172e-03], + [ 1.0939e-03, 5.4017e-08, 4.2057e-04, ..., 5.5879e-09, + 1.1820e-04, 2.0206e-04]], device='cuda:0') +Epoch 451, bias, value: tensor([ 0.0091, 0.0062, -0.0040, -0.0172, 0.0079, -0.0136, -0.0141, -0.0182, + -0.0052, 0.0082], device='cuda:0'), grad: tensor([ 0.0135, -0.0050, 0.0139, -0.0486, -0.0179, 0.0108, 0.0144, 0.0078, + -0.0005, 0.0116], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 450---------------------------------------------------- +epoch 450, time 216.86, cls_loss 0.4308 cls_loss_mapping 0.0007 cls_loss_causal 0.4035 re_mapping 0.0039 re_causal 0.0119 /// teacc 99.06 lr 0.00001000 +Epoch 452, weight, value: tensor([[ 0.0389, 0.1319, -0.1818, ..., -0.1080, 0.0783, 0.0047], + [-0.1091, -0.1064, 0.1162, ..., -0.0479, -0.1157, 0.0568], + [-0.0964, -0.0708, -0.1190, ..., -0.1363, -0.0170, 0.0567], + ..., + [-0.0085, -0.1361, -0.1417, ..., -0.1147, -0.0523, 0.0770], + [-0.1336, -0.1179, -0.1182, ..., -0.0669, -0.0490, -0.1013], + [ 0.0213, -0.0317, -0.0576, ..., -0.1042, 0.0114, -0.0242]], + device='cuda:0'), grad: tensor([[ 4.1389e-03, 8.2731e-04, 2.4915e-04, ..., 1.3103e-03, + 1.6441e-03, 7.2622e-04], + [ 1.7083e-04, 2.3037e-05, 2.7418e-04, ..., 1.6659e-05, + 2.1160e-05, 1.5116e-03], + [-5.1451e-04, -6.8903e-05, 1.5867e-04, ..., 2.3901e-05, + -5.0604e-05, -1.9112e-03], + ..., + [ 1.9133e-04, 8.5607e-06, 2.8753e-04, ..., 2.8517e-06, + 5.7034e-06, 2.4185e-03], + [ 2.5105e-04, 7.4089e-05, -1.9503e-03, ..., 1.3864e-04, + 1.7512e-04, -1.2732e-03], + [ 2.2638e-04, 2.2322e-05, 2.5463e-04, ..., 1.1347e-05, + 1.9073e-05, 1.0052e-03]], device='cuda:0') +Epoch 452, bias, value: tensor([ 0.0090, 0.0063, -0.0041, -0.0173, 0.0078, -0.0135, -0.0140, -0.0181, + -0.0050, 0.0081], device='cuda:0'), grad: tensor([ 0.0259, 0.0127, -0.0294, 0.0158, -0.0502, 0.0126, 0.0101, 0.0330, + -0.0184, -0.0122], device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 216.38, cls_loss 0.4292 cls_loss_mapping 0.0007 cls_loss_causal 0.4056 re_mapping 0.0038 re_causal 0.0122 /// teacc 99.04 lr 0.00001000 +Epoch 453, weight, value: tensor([[ 0.0388, 0.1319, -0.1817, ..., -0.1079, 0.0783, 0.0046], + [-0.1091, -0.1064, 0.1163, ..., -0.0479, -0.1158, 0.0568], + [-0.0965, -0.0708, -0.1190, ..., -0.1363, -0.0170, 0.0567], + ..., + [-0.0086, -0.1361, -0.1419, ..., -0.1148, -0.0523, 0.0769], + [-0.1338, -0.1178, -0.1183, ..., -0.0669, -0.0491, -0.1015], + [ 0.0214, -0.0317, -0.0577, ..., -0.1041, 0.0114, -0.0241]], + device='cuda:0'), grad: tensor([[ 7.0477e-04, -2.7919e-04, 2.0611e-04, ..., 1.6093e-04, + 1.3077e-04, 1.8740e-04], + [ 5.3883e-04, 6.7465e-06, -7.8297e-04, ..., 1.0841e-05, + -7.2002e-05, -3.0804e-04], + [ 4.3893e-04, 1.4234e-04, 7.8726e-04, ..., 2.5898e-05, + 1.0669e-05, 3.5882e-04], + ..., + [ 1.6327e-02, 9.1940e-06, 3.9124e-04, ..., 3.6322e-06, + 4.5374e-06, 2.9421e-04], + [ 9.5272e-04, -5.6887e-04, -1.8339e-03, ..., 6.7174e-05, + 7.0214e-05, 2.7728e-04], + [ 1.3695e-02, 8.0585e-05, 3.8552e-04, ..., 4.6730e-05, + 5.5492e-05, 5.5361e-04]], device='cuda:0') +Epoch 453, bias, value: tensor([ 0.0090, 0.0062, -0.0041, -0.0172, 0.0078, -0.0135, -0.0139, -0.0182, + -0.0053, 0.0082], device='cuda:0'), grad: tensor([ 0.0048, 0.0019, 0.0148, -0.0387, -0.0178, 0.0166, -0.0031, 0.0177, + -0.0121, 0.0160], device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 216.48, cls_loss 0.4239 cls_loss_mapping 0.0007 cls_loss_causal 0.4031 re_mapping 0.0038 re_causal 0.0119 /// teacc 99.06 lr 0.00001000 +Epoch 454, weight, value: tensor([[ 0.0388, 0.1321, -0.1816, ..., -0.1079, 0.0783, 0.0046], + [-0.1092, -0.1066, 0.1163, ..., -0.0480, -0.1159, 0.0566], + [-0.0965, -0.0708, -0.1190, ..., -0.1364, -0.0169, 0.0568], + ..., + [-0.0087, -0.1361, -0.1418, ..., -0.1148, -0.0522, 0.0769], + [-0.1337, -0.1178, -0.1182, ..., -0.0667, -0.0492, -0.1014], + [ 0.0213, -0.0318, -0.0576, ..., -0.1041, 0.0113, -0.0241]], + device='cuda:0'), grad: tensor([[ 3.6120e-04, -9.1046e-06, 1.8227e-04, ..., 3.8326e-05, + 1.1008e-06, 3.7646e-04], + [ 4.9305e-04, -7.2382e-06, 2.1791e-04, ..., 8.6737e-04, + 8.3819e-07, 1.4377e-04], + [ 4.6110e-04, 2.3767e-06, 1.9825e-04, ..., 3.8594e-05, + 1.1027e-06, -9.5892e-04], + ..., + [ 1.8225e-03, 2.9057e-07, 2.2960e-04, ..., 2.0802e-05, + 3.1888e-06, 1.1673e-03], + [ 1.2226e-03, 6.1810e-05, 5.2643e-04, ..., 8.4257e-04, + 1.0058e-06, 5.4550e-04], + [-7.0724e-03, 6.4634e-06, 2.0540e-04, ..., 3.5018e-05, + 6.7912e-06, 4.4894e-04]], device='cuda:0') +Epoch 454, bias, value: tensor([ 0.0091, 0.0061, -0.0040, -0.0173, 0.0078, -0.0135, -0.0139, -0.0180, + -0.0053, 0.0082], device='cuda:0'), grad: tensor([ 0.0075, 0.0136, 0.0046, 0.0110, -0.0229, -0.0112, -0.0153, 0.0151, + 0.0163, -0.0186], device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 216.44, cls_loss 0.4152 cls_loss_mapping 0.0008 cls_loss_causal 0.3887 re_mapping 0.0038 re_causal 0.0114 /// teacc 99.02 lr 0.00001000 +Epoch 455, weight, value: tensor([[ 0.0388, 0.1322, -0.1815, ..., -0.1078, 0.0783, 0.0046], + [-0.1093, -0.1067, 0.1163, ..., -0.0481, -0.1161, 0.0565], + [-0.0965, -0.0709, -0.1191, ..., -0.1365, -0.0169, 0.0568], + ..., + [-0.0085, -0.1361, -0.1418, ..., -0.1149, -0.0522, 0.0770], + [-0.1337, -0.1177, -0.1183, ..., -0.0668, -0.0491, -0.1014], + [ 0.0213, -0.0318, -0.0576, ..., -0.1042, 0.0113, -0.0242]], + device='cuda:0'), grad: tensor([[ 3.6454e-04, 2.9244e-07, 4.4775e-04, ..., 1.5032e-06, + 1.2743e-04, 6.9857e-04], + [ 8.5211e-04, 3.6620e-06, 5.4264e-04, ..., 6.1616e-06, + 2.3293e-04, 1.1749e-03], + [ 5.7077e-04, 1.2434e-04, 5.3930e-04, ..., 4.9882e-06, + 1.7595e-04, 9.3651e-04], + ..., + [ 1.5459e-03, 7.5549e-06, 6.3896e-04, ..., 6.2957e-07, + 1.5497e-04, 1.9417e-03], + [ 6.1083e-04, 5.1446e-06, 3.4847e-03, ..., 7.7295e-04, + 1.1688e-04, 8.8120e-04], + [-1.7204e-03, -9.3162e-05, -3.0041e-03, ..., 2.4494e-06, + -3.1471e-03, -6.7711e-04]], device='cuda:0') +Epoch 455, bias, value: tensor([ 0.0090, 0.0060, -0.0040, -0.0171, 0.0079, -0.0136, -0.0140, -0.0180, + -0.0052, 0.0082], device='cuda:0'), grad: tensor([ 0.0104, 0.0194, 0.0128, -0.1027, 0.0247, 0.0288, 0.0117, -0.0063, + 0.0271, -0.0260], device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 216.64, cls_loss 0.4217 cls_loss_mapping 0.0009 cls_loss_causal 0.3947 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.04 lr 0.00001000 +Epoch 456, weight, value: tensor([[ 0.0387, 0.1321, -0.1815, ..., -0.1079, 0.0781, 0.0046], + [-0.1093, -0.1066, 0.1162, ..., -0.0480, -0.1161, 0.0566], + [-0.0964, -0.0709, -0.1190, ..., -0.1366, -0.0169, 0.0570], + ..., + [-0.0087, -0.1361, -0.1416, ..., -0.1150, -0.0523, 0.0769], + [-0.1339, -0.1176, -0.1185, ..., -0.0670, -0.0490, -0.1017], + [ 0.0214, -0.0317, -0.0575, ..., -0.1040, 0.0115, -0.0241]], + device='cuda:0'), grad: tensor([[-1.6394e-03, -3.9983e-04, 3.6502e-04, ..., -7.2598e-05, + -1.1539e-03, 2.5582e-04], + [ 6.5708e-04, 1.8086e-06, 4.4656e-04, ..., 3.7160e-06, + 2.4867e-04, -5.5790e-05], + [ 4.4799e-04, 6.4075e-06, 3.8743e-04, ..., 8.0653e-07, + 1.7047e-04, 6.1226e-04], + ..., + [-6.2084e-04, 3.9674e-06, -1.0233e-03, ..., 1.0617e-07, + -2.1064e-04, -1.7681e-03], + [ 9.4938e-04, 7.7307e-05, 3.3259e-04, ..., 3.5554e-05, + 4.5252e-04, 6.5088e-04], + [ 2.9793e-03, 6.1318e-06, -1.8673e-03, ..., 2.5369e-06, + 4.3344e-04, 2.3499e-05]], device='cuda:0') +Epoch 456, bias, value: tensor([ 0.0090, 0.0061, -0.0040, -0.0171, 0.0079, -0.0135, -0.0140, -0.0181, + -0.0054, 0.0084], device='cuda:0'), grad: tensor([-0.0220, 0.0107, 0.0030, -0.0062, -0.0191, 0.0132, 0.0181, -0.0140, + 0.0142, 0.0021], device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 216.56, cls_loss 0.4425 cls_loss_mapping 0.0009 cls_loss_causal 0.4137 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.05 lr 0.00001000 +Epoch 457, weight, value: tensor([[ 0.0387, 0.1320, -0.1816, ..., -0.1079, 0.0780, 0.0045], + [-0.1093, -0.1066, 0.1163, ..., -0.0480, -0.1162, 0.0566], + [-0.0965, -0.0708, -0.1190, ..., -0.1368, -0.0169, 0.0569], + ..., + [-0.0086, -0.1363, -0.1416, ..., -0.1151, -0.0523, 0.0770], + [-0.1338, -0.1175, -0.1184, ..., -0.0669, -0.0487, -0.1014], + [ 0.0216, -0.0317, -0.0574, ..., -0.1039, 0.0116, -0.0241]], + device='cuda:0'), grad: tensor([[ 3.7146e-04, -2.7437e-06, 4.5681e-04, ..., 1.7917e-04, + 3.3528e-07, 1.2207e-03], + [ 4.8923e-04, 1.4633e-05, 1.1549e-03, ..., 3.1281e-03, + 2.9564e-05, 8.4610e-03], + [-4.3297e-03, 4.8801e-06, 5.2309e-04, ..., 2.2852e-04, + 8.6576e-06, -3.9902e-03], + ..., + [-4.6120e-03, -1.4105e-03, -6.5804e-03, ..., -6.0768e-03, + -2.8801e-03, -1.8402e-02], + [ 8.6737e-04, 1.1909e-04, 1.0080e-03, ..., 3.4714e-04, + 2.4152e-04, 2.0275e-03], + [ 3.6907e-03, 9.9659e-04, 2.2888e-03, ..., 8.3637e-04, + 1.7271e-03, 1.9932e-03]], device='cuda:0') +Epoch 457, bias, value: tensor([ 0.0090, 0.0060, -0.0041, -0.0172, 0.0078, -0.0134, -0.0141, -0.0181, + -0.0053, 0.0085], device='cuda:0'), grad: tensor([ 0.0129, 0.0363, -0.0144, 0.0169, 0.0248, 0.0157, -0.0095, -0.1047, + 0.0177, 0.0042], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 456---------------------------------------------------- +epoch 456, time 217.08, cls_loss 0.4383 cls_loss_mapping 0.0008 cls_loss_causal 0.4097 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.07 lr 0.00001000 +Epoch 458, weight, value: tensor([[ 0.0388, 0.1320, -0.1816, ..., -0.1079, 0.0780, 0.0045], + [-0.1093, -0.1065, 0.1162, ..., -0.0481, -0.1162, 0.0565], + [-0.0966, -0.0709, -0.1191, ..., -0.1368, -0.0169, 0.0570], + ..., + [-0.0085, -0.1363, -0.1416, ..., -0.1149, -0.0523, 0.0770], + [-0.1338, -0.1176, -0.1184, ..., -0.0669, -0.0487, -0.1014], + [ 0.0217, -0.0316, -0.0573, ..., -0.1039, 0.0116, -0.0239]], + device='cuda:0'), grad: tensor([[ 1.1391e-02, -3.8147e-05, 7.2384e-04, ..., 5.7131e-05, + -1.9312e-04, 1.6499e-03], + [ 8.6021e-04, 7.1168e-05, 9.4843e-04, ..., 3.6389e-05, + 3.0875e-05, 3.1700e-03], + [ 2.8774e-05, 2.5928e-05, -2.1381e-03, ..., 8.8811e-05, + 6.0983e-06, -1.1940e-02], + ..., + [ 2.4052e-03, 4.2200e-05, 1.3266e-03, ..., 2.4065e-05, + 1.8284e-05, 2.5101e-03], + [-2.8443e-04, -8.0013e-04, -8.5950e-05, ..., -7.2622e-04, + -1.6165e-04, 7.8917e-04], + [ 3.1738e-03, 6.8665e-05, 2.5387e-03, ..., 4.7117e-05, + 4.2319e-05, -1.3142e-03]], device='cuda:0') +Epoch 458, bias, value: tensor([ 0.0089, 0.0061, -0.0041, -0.0173, 0.0078, -0.0134, -0.0142, -0.0181, + -0.0052, 0.0086], device='cuda:0'), grad: tensor([ 0.0230, 0.0276, -0.0463, -0.0443, 0.0063, 0.0257, 0.0006, 0.0229, + -0.0075, -0.0079], device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 217.38, cls_loss 0.4646 cls_loss_mapping 0.0007 cls_loss_causal 0.4348 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.07 lr 0.00001000 +Epoch 459, weight, value: tensor([[ 0.0387, 0.1321, -0.1817, ..., -0.1080, 0.0780, 0.0043], + [-0.1092, -0.1066, 0.1162, ..., -0.0481, -0.1162, 0.0566], + [-0.0965, -0.0709, -0.1190, ..., -0.1369, -0.0170, 0.0571], + ..., + [-0.0085, -0.1362, -0.1414, ..., -0.1149, -0.0522, 0.0771], + [-0.1337, -0.1176, -0.1184, ..., -0.0666, -0.0487, -0.1014], + [ 0.0216, -0.0318, -0.0574, ..., -0.1040, 0.0115, -0.0239]], + device='cuda:0'), grad: tensor([[ 1.2712e-03, 5.4359e-04, 7.3957e-04, ..., 1.6365e-03, + 1.5068e-04, 4.6425e-03], + [-4.2534e-03, -2.7370e-04, 5.8365e-04, ..., -5.4979e-04, + -6.8188e-04, -4.1351e-03], + [ 8.3733e-04, 2.5320e-04, 8.5592e-04, ..., 4.5919e-04, + 2.0695e-04, -5.1842e-03], + ..., + [-5.5075e-04, 8.1182e-05, 9.1219e-04, ..., 9.5427e-05, + 2.2292e-04, 2.9421e-04], + [ 5.1641e-04, -2.2754e-05, 3.8052e-04, ..., -1.9193e-04, + 1.9515e-04, 2.3842e-03], + [ 5.9557e-04, 2.0337e-04, -1.6632e-03, ..., 2.1923e-04, + 4.1819e-04, -4.5776e-03]], device='cuda:0') +Epoch 459, bias, value: tensor([ 0.0088, 0.0063, -0.0040, -0.0172, 0.0077, -0.0134, -0.0143, -0.0180, + -0.0052, 0.0085], device='cuda:0'), grad: tensor([ 0.0280, -0.0182, -0.0079, 0.0061, -0.0085, 0.0349, -0.0182, -0.0013, + 0.0180, -0.0329], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 458---------------------------------------------------- +epoch 458, time 217.27, cls_loss 0.3964 cls_loss_mapping 0.0008 cls_loss_causal 0.3695 re_mapping 0.0036 re_causal 0.0109 /// teacc 99.08 lr 0.00001000 +Epoch 460, weight, value: tensor([[ 0.0387, 0.1321, -0.1817, ..., -0.1080, 0.0779, 0.0042], + [-0.1090, -0.1066, 0.1161, ..., -0.0481, -0.1162, 0.0566], + [-0.0965, -0.0709, -0.1192, ..., -0.1369, -0.0169, 0.0570], + ..., + [-0.0087, -0.1361, -0.1415, ..., -0.1150, -0.0522, 0.0770], + [-0.1338, -0.1176, -0.1185, ..., -0.0667, -0.0488, -0.1015], + [ 0.0217, -0.0319, -0.0573, ..., -0.1038, 0.0115, -0.0236]], + device='cuda:0'), grad: tensor([[ 1.1082e-03, 8.2180e-06, 4.7517e-04, ..., 1.7679e-04, + 8.0526e-05, 8.8549e-04], + [-4.4465e-05, 8.2031e-06, -1.6470e-03, ..., 2.1741e-05, + 9.3102e-05, -7.8278e-03], + [ 2.5153e-04, -8.1599e-05, 2.5654e-04, ..., 8.7637e-07, + -3.8952e-05, 7.7963e-04], + ..., + [ 5.9271e-04, 6.1154e-05, 6.9332e-04, ..., 1.3877e-07, + 2.4724e-04, 2.4166e-03], + [ 1.2445e-03, 1.5140e-05, 5.2547e-04, ..., 1.9825e-04, + 8.9347e-05, 6.7520e-03], + [-1.1778e-04, -8.0347e-05, 2.4629e-04, ..., 3.8482e-06, + -4.3809e-05, -2.8305e-03]], device='cuda:0') +Epoch 460, bias, value: tensor([ 0.0087, 0.0063, -0.0042, -0.0173, 0.0078, -0.0135, -0.0141, -0.0181, + -0.0053, 0.0088], device='cuda:0'), grad: tensor([ 0.0156, -0.0356, 0.0073, 0.0174, -0.0178, -0.0184, -0.0037, 0.0187, + 0.0311, -0.0146], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 459---------------------------------------------------- +epoch 459, time 216.83, cls_loss 0.4218 cls_loss_mapping 0.0008 cls_loss_causal 0.3924 re_mapping 0.0038 re_causal 0.0117 /// teacc 99.14 lr 0.00001000 +Epoch 461, weight, value: tensor([[ 0.0388, 0.1319, -0.1818, ..., -0.1083, 0.0779, 0.0042], + [-0.1090, -0.1069, 0.1160, ..., -0.0482, -0.1162, 0.0565], + [-0.0967, -0.0709, -0.1193, ..., -0.1369, -0.0169, 0.0571], + ..., + [-0.0087, -0.1362, -0.1415, ..., -0.1150, -0.0522, 0.0770], + [-0.1337, -0.1176, -0.1184, ..., -0.0667, -0.0488, -0.1015], + [ 0.0217, -0.0319, -0.0574, ..., -0.1039, 0.0115, -0.0236]], + device='cuda:0'), grad: tensor([[ 3.5453e-04, -2.5049e-05, 7.6532e-04, ..., 1.3816e-04, + -4.2105e-04, 9.2793e-04], + [ 5.2023e-04, 3.1982e-06, 7.9012e-04, ..., 1.6379e-04, + 5.4741e-04, 1.1663e-03], + [-1.9062e-04, -5.3108e-05, -3.5267e-03, ..., 1.5569e-04, + -3.1900e-04, -9.6655e-04], + ..., + [ 5.0163e-04, 5.8860e-06, 7.7200e-04, ..., 2.1291e-04, + 2.6774e-04, 1.2531e-03], + [-2.0719e-04, -6.8283e-04, -2.9926e-03, ..., -8.9216e-04, + 3.3307e-04, -1.2083e-03], + [-5.7316e-04, 3.8415e-05, 9.0981e-04, ..., 1.2141e-04, + -1.9932e-03, -7.5006e-04]], device='cuda:0') +Epoch 461, bias, value: tensor([ 0.0087, 0.0063, -0.0043, -0.0172, 0.0078, -0.0135, -0.0139, -0.0182, + -0.0053, 0.0087], device='cuda:0'), grad: tensor([ 0.0143, 0.0199, -0.0271, 0.0217, 0.0187, 0.0155, -0.0151, 0.0163, + -0.0491, -0.0151], device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 217.00, cls_loss 0.4161 cls_loss_mapping 0.0007 cls_loss_causal 0.3969 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.09 lr 0.00001000 +Epoch 462, weight, value: tensor([[ 0.0388, 0.1319, -0.1819, ..., -0.1083, 0.0780, 0.0041], + [-0.1091, -0.1070, 0.1161, ..., -0.0482, -0.1163, 0.0565], + [-0.0966, -0.0709, -0.1194, ..., -0.1369, -0.0170, 0.0571], + ..., + [-0.0086, -0.1362, -0.1415, ..., -0.1149, -0.0521, 0.0769], + [-0.1338, -0.1176, -0.1185, ..., -0.0666, -0.0488, -0.1014], + [ 0.0216, -0.0321, -0.0575, ..., -0.1041, 0.0114, -0.0237]], + device='cuda:0'), grad: tensor([[-1.1024e-03, -2.0400e-05, -2.1801e-03, ..., 6.1467e-07, + -9.9391e-06, -1.4868e-03], + [ 9.7942e-04, 2.2072e-06, 1.2474e-03, ..., 3.5390e-08, + 7.3295e-07, 1.6737e-03], + [ 2.5768e-03, 1.1128e-04, 7.8583e-04, ..., 7.9945e-06, + 3.6061e-05, 2.8667e-03], + ..., + [ 6.8283e-04, 4.8548e-05, 8.8263e-04, ..., 2.0489e-08, + 6.7651e-06, 1.0633e-03], + [ 7.8106e-04, 5.2118e-04, 7.3624e-04, ..., 1.3327e-06, + 2.5611e-06, 1.1339e-03], + [-1.8816e-03, -1.1282e-03, -2.0599e-03, ..., 2.2165e-07, + 1.3702e-05, -5.5656e-03]], device='cuda:0') +Epoch 462, bias, value: tensor([ 0.0087, 0.0063, -0.0043, -0.0171, 0.0078, -0.0135, -0.0140, -0.0182, + -0.0053, 0.0087], device='cuda:0'), grad: tensor([-0.0467, 0.0226, 0.0276, 0.0191, -0.0093, 0.0230, -0.0089, 0.0159, + 0.0264, -0.0698], device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 216.50, cls_loss 0.4249 cls_loss_mapping 0.0007 cls_loss_causal 0.4000 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.08 lr 0.00001000 +Epoch 463, weight, value: tensor([[ 0.0390, 0.1320, -0.1817, ..., -0.1082, 0.0780, 0.0043], + [-0.1091, -0.1070, 0.1159, ..., -0.0482, -0.1163, 0.0565], + [-0.0966, -0.0710, -0.1193, ..., -0.1369, -0.0171, 0.0573], + ..., + [-0.0086, -0.1362, -0.1416, ..., -0.1150, -0.0522, 0.0769], + [-0.1340, -0.1177, -0.1185, ..., -0.0667, -0.0487, -0.1015], + [ 0.0217, -0.0321, -0.0576, ..., -0.1041, 0.0113, -0.0236]], + device='cuda:0'), grad: tensor([[ 8.6308e-05, -9.7230e-06, 2.4033e-04, ..., -1.8626e-09, + -1.6764e-05, -3.2101e-03], + [-8.8692e-04, 7.6462e-07, -2.4014e-03, ..., 2.7940e-08, + -2.8253e-05, 2.1076e-04], + [-2.2352e-04, -6.4754e-04, 2.4283e-04, ..., 7.2643e-07, + 3.0547e-06, 7.2021e-03], + ..., + [ 2.9945e-04, 2.4550e-06, 3.4237e-04, ..., 3.5390e-08, + 3.9674e-06, -9.7656e-03], + [ 1.1787e-05, 4.3325e-06, 2.5153e-04, ..., 5.1782e-07, + 5.0105e-06, 1.1625e-03], + [ 1.0033e-03, 4.8429e-06, 1.4150e-04, ..., 6.1095e-07, + 7.4729e-06, 9.6273e-04]], device='cuda:0') +Epoch 463, bias, value: tensor([ 0.0088, 0.0062, -0.0042, -0.0171, 0.0077, -0.0136, -0.0139, -0.0182, + -0.0053, 0.0087], device='cuda:0'), grad: tensor([-0.0209, -0.0191, 0.0291, 0.0134, 0.0052, -0.0170, 0.0123, -0.0256, + 0.0146, 0.0081], device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 216.78, cls_loss 0.4353 cls_loss_mapping 0.0006 cls_loss_causal 0.4067 re_mapping 0.0037 re_causal 0.0119 /// teacc 99.09 lr 0.00001000 +Epoch 464, weight, value: tensor([[ 0.0388, 0.1319, -0.1818, ..., -0.1081, 0.0779, 0.0044], + [-0.1091, -0.1069, 0.1159, ..., -0.0481, -0.1163, 0.0565], + [-0.0967, -0.0711, -0.1194, ..., -0.1370, -0.0173, 0.0572], + ..., + [-0.0086, -0.1361, -0.1415, ..., -0.1150, -0.0522, 0.0770], + [-0.1339, -0.1177, -0.1185, ..., -0.0666, -0.0487, -0.1015], + [ 0.0218, -0.0322, -0.0576, ..., -0.1041, 0.0113, -0.0237]], + device='cuda:0'), grad: tensor([[ 0.0013, 0.0001, 0.0019, ..., 0.0001, 0.0003, 0.0012], + [-0.0014, -0.0008, -0.0066, ..., 0.0011, 0.0009, -0.0008], + [-0.0054, 0.0004, -0.0008, ..., 0.0003, 0.0007, 0.0006], + ..., + [-0.0012, 0.0005, -0.0019, ..., 0.0005, -0.0022, -0.0027], + [ 0.0014, 0.0011, 0.0027, ..., 0.0009, 0.0006, 0.0020], + [ 0.0029, -0.0029, -0.0021, ..., -0.0040, -0.0007, -0.0057]], + device='cuda:0') +Epoch 464, bias, value: tensor([ 0.0088, 0.0060, -0.0042, -0.0172, 0.0077, -0.0136, -0.0138, -0.0181, + -0.0052, 0.0086], device='cuda:0'), grad: tensor([ 2.0233e-02, -5.9784e-02, -1.7517e-02, 3.5919e-02, -6.9809e-03, + 2.0020e-02, 2.2003e-02, -3.8971e-02, 2.5101e-02, -7.4506e-05], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 216.80, cls_loss 0.4405 cls_loss_mapping 0.0008 cls_loss_causal 0.4130 re_mapping 0.0037 re_causal 0.0116 /// teacc 99.06 lr 0.00001000 +Epoch 465, weight, value: tensor([[ 0.0388, 0.1320, -0.1818, ..., -0.1081, 0.0779, 0.0043], + [-0.1090, -0.1069, 0.1160, ..., -0.0481, -0.1164, 0.0565], + [-0.0966, -0.0711, -0.1194, ..., -0.1370, -0.0174, 0.0573], + ..., + [-0.0085, -0.1359, -0.1415, ..., -0.1148, -0.0520, 0.0769], + [-0.1338, -0.1177, -0.1185, ..., -0.0665, -0.0486, -0.1014], + [ 0.0219, -0.0322, -0.0576, ..., -0.1041, 0.0113, -0.0238]], + device='cuda:0'), grad: tensor([[ 2.4319e-04, 1.3411e-04, 5.5122e-04, ..., 9.8169e-05, + 1.0110e-05, 9.0265e-04], + [-4.3068e-03, 8.3983e-05, -1.6842e-03, ..., 9.0122e-05, + 1.3426e-05, -1.2302e-03], + [ 4.2319e-04, 2.0480e-04, 4.2647e-05, ..., 2.0444e-04, + 3.0994e-05, -3.7408e-04], + ..., + [ 6.2895e-04, 2.2566e-04, 7.4959e-04, ..., 2.2602e-04, + 3.4064e-05, 1.7757e-03], + [ 4.4632e-04, -4.3869e-03, 5.5456e-04, ..., 1.9360e-04, + 2.9325e-05, 1.2770e-03], + [ 1.9331e-03, 1.7815e-03, 6.3944e-04, ..., -1.4124e-03, + -2.0981e-04, -2.5558e-03]], device='cuda:0') +Epoch 465, bias, value: tensor([ 0.0087, 0.0062, -0.0042, -0.0174, 0.0075, -0.0137, -0.0137, -0.0180, + -0.0051, 0.0086], device='cuda:0'), grad: tensor([ 0.0131, -0.0519, 0.0171, 0.0210, -0.0155, -0.0182, -0.0135, 0.0222, + 0.0089, 0.0169], device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 217.11, cls_loss 0.4338 cls_loss_mapping 0.0007 cls_loss_causal 0.4042 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.05 lr 0.00001000 +Epoch 466, weight, value: tensor([[ 0.0389, 0.1320, -0.1819, ..., -0.1082, 0.0780, 0.0043], + [-0.1090, -0.1068, 0.1160, ..., -0.0483, -0.1166, 0.0564], + [-0.0966, -0.0713, -0.1196, ..., -0.1371, -0.0175, 0.0573], + ..., + [-0.0085, -0.1360, -0.1415, ..., -0.1149, -0.0520, 0.0769], + [-0.1339, -0.1178, -0.1185, ..., -0.0665, -0.0487, -0.1015], + [ 0.0218, -0.0322, -0.0576, ..., -0.1040, 0.0113, -0.0237]], + device='cuda:0'), grad: tensor([[ 2.6536e-04, 3.4049e-06, 9.8133e-04, ..., 1.0878e-04, + 2.3210e-04, 9.9182e-04], + [ 9.9277e-04, 2.6822e-07, 2.8362e-03, ..., 3.6430e-04, + 1.0395e-03, 4.0512e-03], + [ 3.1662e-04, -1.2177e-04, 1.4830e-03, ..., 2.4700e-04, + 3.0184e-04, 1.5364e-03], + ..., + [-1.7157e-03, 1.5767e-06, 1.2865e-03, ..., 9.6738e-05, + -2.2526e-03, -4.1580e-03], + [-2.8872e-04, 1.8609e-04, -2.8839e-03, ..., 1.1176e-04, + 2.4128e-04, -7.0524e-04], + [-8.4591e-04, 8.8066e-06, -3.0003e-03, ..., -1.8015e-03, + -8.8787e-04, -3.4275e-03]], device='cuda:0') +Epoch 466, bias, value: tensor([ 0.0088, 0.0062, -0.0043, -0.0173, 0.0075, -0.0136, -0.0137, -0.0180, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([ 0.0128, 0.0388, 0.0162, -0.0043, 0.0188, -0.0135, -0.0191, -0.0097, + -0.0092, -0.0307], device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 217.29, cls_loss 0.4346 cls_loss_mapping 0.0007 cls_loss_causal 0.4133 re_mapping 0.0038 re_causal 0.0116 /// teacc 99.06 lr 0.00001000 +Epoch 467, weight, value: tensor([[ 0.0389, 0.1320, -0.1817, ..., -0.1082, 0.0779, 0.0043], + [-0.1090, -0.1068, 0.1161, ..., -0.0483, -0.1167, 0.0564], + [-0.0966, -0.0712, -0.1196, ..., -0.1372, -0.0175, 0.0572], + ..., + [-0.0085, -0.1360, -0.1415, ..., -0.1150, -0.0521, 0.0769], + [-0.1341, -0.1179, -0.1187, ..., -0.0666, -0.0488, -0.1015], + [ 0.0219, -0.0322, -0.0576, ..., -0.1040, 0.0114, -0.0235]], + device='cuda:0'), grad: tensor([[-5.9414e-04, -1.1520e-03, 9.3985e-04, ..., 6.0163e-07, + -5.3358e-04, 2.7204e-04], + [ 3.9101e-04, 5.6922e-06, 1.2856e-03, ..., -6.0610e-06, + 1.7226e-05, 6.0606e-04], + [ 6.6423e-04, 5.4985e-05, 1.0471e-03, ..., 4.2189e-07, + 1.4782e-04, 1.7891e-03], + ..., + [-4.2653e-04, 1.3554e-04, -4.5853e-03, ..., 1.8673e-06, + 1.4329e-04, -3.2272e-03], + [-1.1110e-03, 1.0788e-04, -1.8358e-03, ..., 1.7518e-06, + 1.3077e-04, -2.1820e-03], + [-1.6994e-03, 2.4676e-04, 1.2131e-03, ..., 3.9116e-07, + -5.0735e-04, 8.2827e-04]], device='cuda:0') +Epoch 467, bias, value: tensor([ 0.0089, 0.0062, -0.0043, -0.0173, 0.0074, -0.0135, -0.0137, -0.0182, + -0.0052, 0.0089], device='cuda:0'), grad: tensor([ 0.0112, 0.0193, 0.0170, 0.0170, 0.0194, 0.0149, -0.0146, -0.0448, + -0.0499, 0.0105], device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 216.48, cls_loss 0.4554 cls_loss_mapping 0.0008 cls_loss_causal 0.4184 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.05 lr 0.00001000 +Epoch 468, weight, value: tensor([[ 0.0389, 0.1320, -0.1819, ..., -0.1081, 0.0779, 0.0044], + [-0.1090, -0.1069, 0.1161, ..., -0.0482, -0.1167, 0.0563], + [-0.0966, -0.0712, -0.1196, ..., -0.1374, -0.0176, 0.0572], + ..., + [-0.0083, -0.1358, -0.1413, ..., -0.1150, -0.0519, 0.0769], + [-0.1341, -0.1179, -0.1187, ..., -0.0666, -0.0488, -0.1015], + [ 0.0219, -0.0320, -0.0576, ..., -0.1042, 0.0112, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.8063e-03, 2.1708e-04, 4.0746e-04, ..., -3.7402e-06, + -2.9787e-05, 5.0157e-05], + [ 4.1056e-04, 6.3293e-06, -4.3064e-05, ..., 4.0084e-06, + 1.8151e-06, -3.7670e-04], + [ 1.2326e-04, 1.2495e-05, 8.9824e-05, ..., 3.4213e-05, + 4.4778e-06, 3.7956e-04], + ..., + [-7.7114e-06, 8.9183e-06, 4.7594e-05, ..., 2.7254e-05, + -2.3901e-05, -8.1778e-05], + [ 6.4659e-03, -6.6996e-04, -1.1215e-03, ..., 4.5657e-05, + 1.0453e-05, 1.0651e-04], + [ 2.3735e-04, 9.0897e-05, 1.5175e-04, ..., 1.2340e-06, + 2.8983e-05, 1.2803e-04]], device='cuda:0') +Epoch 468, bias, value: tensor([ 0.0088, 0.0061, -0.0044, -0.0173, 0.0074, -0.0134, -0.0136, -0.0181, + -0.0053, 0.0088], device='cuda:0'), grad: tensor([ 0.0061, 0.0112, 0.0025, 0.0023, 0.0013, -0.0261, -0.0475, 0.0008, + 0.0471, 0.0023], device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 216.72, cls_loss 0.4069 cls_loss_mapping 0.0008 cls_loss_causal 0.3781 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.04 lr 0.00001000 +Epoch 469, weight, value: tensor([[ 0.0388, 0.1320, -0.1818, ..., -0.1082, 0.0779, 0.0042], + [-0.1089, -0.1068, 0.1162, ..., -0.0480, -0.1166, 0.0565], + [-0.0965, -0.0711, -0.1194, ..., -0.1372, -0.0175, 0.0572], + ..., + [-0.0083, -0.1359, -0.1414, ..., -0.1150, -0.0520, 0.0769], + [-0.1342, -0.1179, -0.1189, ..., -0.0667, -0.0488, -0.1015], + [ 0.0218, -0.0320, -0.0577, ..., -0.1042, 0.0111, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.6999e-04, 4.9081e-07, 4.8041e-05, ..., 1.0259e-05, + 6.9499e-05, 3.7527e-04], + [ 1.1998e-04, 1.3411e-07, -2.1803e-04, ..., -3.0342e-06, + 4.7535e-05, -9.2459e-04], + [ 3.5810e-04, 2.3216e-05, 2.7347e-04, ..., 2.3283e-07, + 1.3876e-04, 1.2999e-03], + ..., + [-2.0790e-03, 1.8347e-07, -4.9162e-04, ..., 8.6613e-08, + -8.1730e-04, -4.2725e-03], + [ 3.1924e-04, -6.4015e-05, 4.0531e-05, ..., -4.6492e-06, + 1.1241e-04, 4.2343e-04], + [ 2.3437e-04, 2.3935e-07, 8.3029e-05, ..., 1.3225e-07, + 5.3793e-06, 6.2323e-04]], device='cuda:0') +Epoch 469, bias, value: tensor([ 0.0088, 0.0062, -0.0042, -0.0173, 0.0075, -0.0134, -0.0137, -0.0182, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([ 0.0081, 0.0067, -0.0105, 0.0090, 0.0118, 0.0075, 0.0085, -0.0273, + -0.0229, 0.0092], device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 217.50, cls_loss 0.4381 cls_loss_mapping 0.0007 cls_loss_causal 0.4176 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.05 lr 0.00001000 +Epoch 470, weight, value: tensor([[ 0.0388, 0.1320, -0.1819, ..., -0.1082, 0.0778, 0.0040], + [-0.1088, -0.1067, 0.1163, ..., -0.0481, -0.1163, 0.0565], + [-0.0965, -0.0709, -0.1193, ..., -0.1373, -0.0173, 0.0573], + ..., + [-0.0083, -0.1358, -0.1415, ..., -0.1150, -0.0520, 0.0771], + [-0.1343, -0.1180, -0.1190, ..., -0.0668, -0.0488, -0.1015], + [ 0.0217, -0.0321, -0.0576, ..., -0.1044, 0.0110, -0.0236]], + device='cuda:0'), grad: tensor([[ 2.7895e-04, 7.1898e-06, 8.9407e-04, ..., 4.1842e-05, + 1.5721e-06, 7.9870e-04], + [-2.7466e-03, -1.4710e-04, -1.7204e-03, ..., -8.6880e-04, + 2.9039e-06, -2.8915e-03], + [-7.0381e-04, 7.9989e-05, -3.6087e-03, ..., 1.7181e-05, + 3.5483e-06, -3.4084e-03], + ..., + [ 3.0098e-03, 3.9563e-06, 1.0443e-03, ..., 1.0960e-05, + 3.0537e-03, 8.5163e-04], + [ 5.6124e-04, 1.9103e-05, 9.2173e-04, ..., 6.6578e-05, + 3.3956e-06, 9.9277e-04], + [ 5.3930e-04, 2.7686e-05, 1.0920e-03, ..., 3.7462e-05, + 2.6774e-04, 1.0252e-03]], device='cuda:0') +Epoch 470, bias, value: tensor([ 0.0087, 0.0062, -0.0043, -0.0171, 0.0076, -0.0134, -0.0138, -0.0182, + -0.0053, 0.0086], device='cuda:0'), grad: tensor([ 0.0166, -0.0312, -0.0797, -0.0308, 0.0046, 0.0095, 0.0230, 0.0434, + 0.0214, 0.0232], device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 216.63, cls_loss 0.4533 cls_loss_mapping 0.0007 cls_loss_causal 0.4187 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.11 lr 0.00001000 +Epoch 471, weight, value: tensor([[ 0.0388, 0.1320, -0.1818, ..., -0.1081, 0.0779, 0.0040], + [-0.1090, -0.1067, 0.1162, ..., -0.0482, -0.1164, 0.0565], + [-0.0965, -0.0709, -0.1193, ..., -0.1373, -0.0174, 0.0572], + ..., + [-0.0081, -0.1359, -0.1415, ..., -0.1149, -0.0521, 0.0771], + [-0.1343, -0.1180, -0.1188, ..., -0.0668, -0.0488, -0.1014], + [ 0.0216, -0.0320, -0.0577, ..., -0.1043, 0.0110, -0.0235]], + device='cuda:0'), grad: tensor([[ 1.4603e-04, 1.5521e-04, 3.9577e-04, ..., 4.0269e-04, + -2.4736e-05, 9.5940e-04], + [-1.5774e-03, 1.1601e-05, -1.2207e-03, ..., -1.7767e-03, + 1.5693e-06, 1.0231e-02], + [ 3.6550e-04, 4.3511e-05, 2.0564e-04, ..., 4.1413e-04, + 5.0515e-06, 2.5482e-03], + ..., + [ 1.4582e-03, 6.4564e-04, 1.0853e-03, ..., 1.3006e-04, + 5.5313e-04, -2.1606e-02], + [ 6.9714e-04, 3.0160e-04, 6.5041e-04, ..., 9.7132e-04, + 1.0900e-05, 2.8038e-03], + [-7.4806e-03, -6.0272e-04, -7.6866e-04, ..., 2.1398e-04, + -5.7316e-04, -3.2330e-03]], device='cuda:0') +Epoch 471, bias, value: tensor([ 0.0087, 0.0062, -0.0042, -0.0171, 0.0077, -0.0134, -0.0141, -0.0182, + -0.0053, 0.0086], device='cuda:0'), grad: tensor([ 0.0087, -0.0512, 0.0122, 0.0079, 0.0346, 0.0185, -0.0081, -0.0128, + 0.0180, -0.0276], device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 216.91, cls_loss 0.4147 cls_loss_mapping 0.0008 cls_loss_causal 0.3896 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.08 lr 0.00001000 +Epoch 472, weight, value: tensor([[ 0.0389, 0.1321, -0.1818, ..., -0.1081, 0.0779, 0.0041], + [-0.1090, -0.1068, 0.1160, ..., -0.0483, -0.1164, 0.0564], + [-0.0966, -0.0709, -0.1195, ..., -0.1373, -0.0173, 0.0572], + ..., + [-0.0082, -0.1359, -0.1414, ..., -0.1150, -0.0522, 0.0770], + [-0.1342, -0.1180, -0.1188, ..., -0.0668, -0.0488, -0.1013], + [ 0.0218, -0.0321, -0.0577, ..., -0.1043, 0.0109, -0.0234]], + device='cuda:0'), grad: tensor([[ 1.0490e-03, 2.1801e-03, 5.7125e-04, ..., 1.7146e-06, + 2.8825e-04, 2.9030e-03], + [ 4.3559e-04, -4.6417e-06, 9.6202e-05, ..., -1.8790e-05, + 2.1204e-05, 1.5011e-03], + [-3.9411e-04, -2.1744e-03, -6.8903e-05, ..., 2.1420e-06, + -2.4748e-04, -3.6068e-03], + ..., + [-2.4390e-04, 9.2864e-05, 5.0974e-04, ..., 9.7882e-07, + 7.1943e-05, 8.1205e-04], + [ 3.2496e-04, 7.3791e-05, 3.8886e-04, ..., 2.9821e-06, + 5.5850e-05, 1.0529e-03], + [ 1.4095e-03, 4.5121e-05, 3.5715e-04, ..., 7.3947e-07, + 4.2975e-05, 1.6880e-03]], device='cuda:0') +Epoch 472, bias, value: tensor([ 0.0089, 0.0061, -0.0044, -0.0171, 0.0077, -0.0134, -0.0141, -0.0182, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([ 0.0149, -0.0181, -0.0089, 0.0140, -0.0230, -0.0218, 0.0086, 0.0119, + 0.0101, 0.0123], device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 216.54, cls_loss 0.4295 cls_loss_mapping 0.0008 cls_loss_causal 0.4065 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.07 lr 0.00001000 +Epoch 473, weight, value: tensor([[ 0.0390, 0.1321, -0.1818, ..., -0.1079, 0.0780, 0.0041], + [-0.1090, -0.1068, 0.1160, ..., -0.0483, -0.1165, 0.0563], + [-0.0968, -0.0708, -0.1196, ..., -0.1373, -0.0174, 0.0573], + ..., + [-0.0083, -0.1360, -0.1413, ..., -0.1151, -0.0523, 0.0770], + [-0.1342, -0.1180, -0.1189, ..., -0.0668, -0.0488, -0.1011], + [ 0.0217, -0.0323, -0.0578, ..., -0.1044, 0.0109, -0.0235]], + device='cuda:0'), grad: tensor([[-1.6804e-03, -1.4238e-03, 8.2910e-05, ..., -1.2970e-03, + 2.9850e-04, 1.0767e-03], + [ 1.1539e-03, 1.9920e-04, 7.2670e-04, ..., 7.4446e-05, + 4.4799e-04, 1.4915e-03], + [ 3.5076e-03, 2.9411e-03, 2.6703e-03, ..., 9.9468e-04, + 7.3099e-04, 7.8354e-03], + ..., + [ 4.1294e-04, 4.6730e-04, -2.9526e-03, ..., 5.4985e-05, + 3.4094e-04, -3.3975e-04], + [ 8.5783e-04, 2.0063e-04, 5.7507e-04, ..., 1.0931e-04, + 3.0231e-04, 1.1673e-03], + [-3.3970e-03, 1.0155e-02, 6.0081e-03, ..., 4.8685e-04, + 2.5311e-03, 2.2030e-03]], device='cuda:0') +Epoch 473, bias, value: tensor([ 0.0089, 0.0060, -0.0045, -0.0170, 0.0078, -0.0133, -0.0141, -0.0182, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([-0.0017, 0.0191, 0.0165, -0.0641, 0.0029, 0.0215, 0.0252, -0.0098, + 0.0159, -0.0254], device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 216.51, cls_loss 0.4513 cls_loss_mapping 0.0008 cls_loss_causal 0.4208 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.09 lr 0.00001000 +Epoch 474, weight, value: tensor([[ 0.0390, 0.1322, -0.1818, ..., -0.1079, 0.0780, 0.0042], + [-0.1090, -0.1069, 0.1158, ..., -0.0485, -0.1166, 0.0563], + [-0.0970, -0.0711, -0.1196, ..., -0.1374, -0.0174, 0.0572], + ..., + [-0.0084, -0.1361, -0.1412, ..., -0.1151, -0.0523, 0.0769], + [-0.1343, -0.1181, -0.1191, ..., -0.0670, -0.0490, -0.1012], + [ 0.0216, -0.0323, -0.0579, ..., -0.1043, 0.0106, -0.0235]], + device='cuda:0'), grad: tensor([[ 2.5058e-04, 2.5883e-05, 9.5940e-04, ..., 1.6999e-04, + -4.0054e-04, -1.8826e-03], + [-1.3857e-03, 3.0130e-05, -5.5313e-03, ..., 1.5116e-04, + -1.1883e-03, -5.9509e-03], + [ 1.1349e-03, 1.6487e-04, 1.8559e-03, ..., 1.7524e-04, + -3.8528e-04, 2.4796e-03], + ..., + [ 1.3466e-02, 6.3062e-05, 2.0580e-03, ..., 7.1228e-05, + 1.1311e-03, 3.0060e-03], + [-3.6278e-03, 1.1665e-04, -6.7234e-04, ..., 2.3019e-04, + 3.1114e-04, -1.8921e-03], + [ 7.8201e-05, 3.6597e-05, -4.3821e-04, ..., 8.9467e-05, + 8.4209e-04, -5.9932e-05]], device='cuda:0') +Epoch 474, bias, value: tensor([ 0.0091, 0.0061, -0.0046, -0.0171, 0.0080, -0.0132, -0.0142, -0.0183, + -0.0053, 0.0085], device='cuda:0'), grad: tensor([-0.0207, -0.0368, 0.0269, -0.0150, 0.0161, 0.0084, -0.0019, 0.0698, + -0.0208, -0.0258], device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 216.36, cls_loss 0.4349 cls_loss_mapping 0.0007 cls_loss_causal 0.4055 re_mapping 0.0035 re_causal 0.0113 /// teacc 99.08 lr 0.00001000 +Epoch 475, weight, value: tensor([[ 0.0390, 0.1324, -0.1818, ..., -0.1080, 0.0781, 0.0041], + [-0.1090, -0.1069, 0.1158, ..., -0.0482, -0.1166, 0.0565], + [-0.0971, -0.0708, -0.1194, ..., -0.1375, -0.0175, 0.0572], + ..., + [-0.0084, -0.1360, -0.1413, ..., -0.1151, -0.0522, 0.0770], + [-0.1344, -0.1182, -0.1191, ..., -0.0671, -0.0489, -0.1013], + [ 0.0216, -0.0323, -0.0578, ..., -0.1044, 0.0106, -0.0236]], + device='cuda:0'), grad: tensor([[ 1.0519e-03, 1.9741e-04, 1.3542e-03, ..., 1.6785e-04, + 4.3774e-04, 1.2426e-03], + [ 8.8263e-04, 5.0776e-06, 1.4496e-03, ..., -2.7902e-06, + 2.0599e-04, -9.1314e-04], + [ 1.6582e-04, -8.9874e-03, -5.1193e-03, ..., 4.8995e-05, + -1.4486e-03, -4.3488e-03], + ..., + [-1.0424e-03, 1.7732e-05, -1.8826e-03, ..., 9.4995e-06, + 1.8001e-04, -6.9475e-04], + [-1.0033e-02, -1.1414e-02, -4.7493e-03, ..., -8.8654e-03, + -7.8659e-03, -6.8893e-03], + [ 1.9665e-03, 2.8178e-05, 1.1902e-03, ..., 3.0026e-05, + 3.3426e-04, 2.3003e-03]], device='cuda:0') +Epoch 475, bias, value: tensor([ 0.0090, 0.0062, -0.0045, -0.0171, 0.0079, -0.0132, -0.0142, -0.0182, + -0.0054, 0.0086], device='cuda:0'), grad: tensor([ 0.0206, 0.0185, -0.0117, 0.0762, 0.0193, -0.0457, -0.0147, -0.0105, + -0.0723, 0.0203], device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 216.60, cls_loss 0.4324 cls_loss_mapping 0.0007 cls_loss_causal 0.4064 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.10 lr 0.00001000 +Epoch 476, weight, value: tensor([[ 0.0389, 0.1323, -0.1820, ..., -0.1080, 0.0781, 0.0040], + [-0.1089, -0.1069, 0.1159, ..., -0.0482, -0.1166, 0.0565], + [-0.0971, -0.0708, -0.1195, ..., -0.1376, -0.0175, 0.0572], + ..., + [-0.0084, -0.1359, -0.1412, ..., -0.1152, -0.0523, 0.0769], + [-0.1344, -0.1181, -0.1191, ..., -0.0670, -0.0488, -0.1012], + [ 0.0215, -0.0324, -0.0580, ..., -0.1044, 0.0105, -0.0236]], + device='cuda:0'), grad: tensor([[ 1.4114e-04, 6.7335e-07, 4.5395e-04, ..., 1.1548e-07, + 1.4164e-05, 1.8692e-04], + [ 2.4438e-04, 7.2271e-06, 3.4523e-04, ..., 1.8626e-09, + 9.8422e-06, 4.7517e-04], + [ 1.6701e-04, -1.6975e-03, 4.3631e-04, ..., 1.2107e-08, + 9.0450e-06, 3.0231e-04], + ..., + [-1.1522e-04, 1.2711e-05, 4.9543e-04, ..., 9.3132e-10, + 1.7226e-04, -1.5202e-03], + [ 7.2765e-04, 1.5482e-05, 6.2799e-04, ..., -3.5390e-08, + 1.6600e-05, 3.3379e-04], + [-4.3583e-04, 8.8736e-06, -1.9293e-03, ..., 2.7940e-09, + 1.7986e-03, 2.7031e-05]], device='cuda:0') +Epoch 476, bias, value: tensor([ 0.0090, 0.0062, -0.0046, -0.0171, 0.0078, -0.0131, -0.0140, -0.0182, + -0.0053, 0.0085], device='cuda:0'), grad: tensor([ 0.0059, 0.0078, -0.0165, 0.0283, -0.0272, 0.0072, 0.0069, 0.0029, + 0.0080, -0.0233], device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 216.48, cls_loss 0.4356 cls_loss_mapping 0.0007 cls_loss_causal 0.4092 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.07 lr 0.00001000 +Epoch 477, weight, value: tensor([[ 0.0389, 0.1324, -0.1821, ..., -0.1081, 0.0781, 0.0040], + [-0.1090, -0.1068, 0.1160, ..., -0.0481, -0.1166, 0.0564], + [-0.0969, -0.0706, -0.1194, ..., -0.1377, -0.0174, 0.0572], + ..., + [-0.0083, -0.1359, -0.1411, ..., -0.1151, -0.0521, 0.0770], + [-0.1343, -0.1179, -0.1189, ..., -0.0669, -0.0488, -0.1012], + [ 0.0216, -0.0324, -0.0579, ..., -0.1043, 0.0104, -0.0236]], + device='cuda:0'), grad: tensor([[ 3.5739e-04, -3.0294e-05, 3.2282e-04, ..., 1.1735e-07, + 3.7868e-06, 1.1921e-03], + [-2.0142e-03, 1.0505e-06, 3.7527e-04, ..., 4.0047e-08, + -3.5834e-04, -2.4719e-03], + [-1.2684e-04, 1.3700e-06, 3.1137e-04, ..., 4.3772e-08, + 3.6180e-05, 8.9467e-05], + ..., + [ 4.1008e-04, 1.7108e-06, 3.5191e-04, ..., 9.3132e-10, + 5.4240e-05, 1.0271e-03], + [ 5.1355e-04, 5.3607e-06, 2.9874e-04, ..., 2.6450e-07, + 6.9380e-05, 1.6356e-03], + [-6.1846e-04, -7.9125e-06, -2.8706e-03, ..., 6.5193e-09, + 1.0349e-05, -4.4365e-03]], device='cuda:0') +Epoch 477, bias, value: tensor([ 0.0088, 0.0062, -0.0046, -0.0171, 0.0079, -0.0133, -0.0140, -0.0180, + -0.0053, 0.0086], device='cuda:0'), grad: tensor([ 0.0141, -0.0149, 0.0102, 0.0102, 0.0166, -0.0142, 0.0072, 0.0147, + 0.0175, -0.0614], device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 216.73, cls_loss 0.4296 cls_loss_mapping 0.0007 cls_loss_causal 0.4095 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.11 lr 0.00001000 +Epoch 478, weight, value: tensor([[ 0.0388, 0.1323, -0.1820, ..., -0.1082, 0.0780, 0.0039], + [-0.1089, -0.1068, 0.1161, ..., -0.0482, -0.1166, 0.0565], + [-0.0969, -0.0705, -0.1195, ..., -0.1377, -0.0173, 0.0572], + ..., + [-0.0082, -0.1358, -0.1411, ..., -0.1149, -0.0521, 0.0770], + [-0.1343, -0.1179, -0.1189, ..., -0.0669, -0.0488, -0.1011], + [ 0.0215, -0.0325, -0.0579, ..., -0.1045, 0.0104, -0.0237]], + device='cuda:0'), grad: tensor([[ 4.5514e-04, 6.5207e-05, 6.6519e-04, ..., 2.3997e-04, + -1.1444e-04, 5.0306e-04], + [-4.1199e-03, -4.9734e-04, 1.5628e-04, ..., -1.2121e-03, + 6.5148e-05, -2.7637e-03], + [ 3.9363e-04, 2.4509e-04, 3.0947e-04, ..., 1.2410e-04, + 1.3459e-04, -8.5068e-03], + ..., + [ 2.2469e-03, 7.2289e-04, 5.1785e-04, ..., 2.4033e-04, + 1.0319e-03, 4.1885e-03], + [ 8.6069e-04, 4.9973e-04, 5.5408e-04, ..., 2.9469e-04, + 2.7251e-04, 5.7030e-03], + [-2.5158e-03, -6.6757e-04, -2.2340e-04, ..., 2.9612e-04, + -2.1057e-03, -9.0361e-04]], device='cuda:0') +Epoch 478, bias, value: tensor([ 0.0088, 0.0062, -0.0046, -0.0172, 0.0078, -0.0133, -0.0140, -0.0179, + -0.0052, 0.0085], device='cuda:0'), grad: tensor([ 0.0057, -0.0264, -0.0147, 0.0087, 0.0093, -0.0170, 0.0089, 0.0201, + 0.0174, -0.0119], device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 216.54, cls_loss 0.4323 cls_loss_mapping 0.0007 cls_loss_causal 0.4079 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.11 lr 0.00001000 +Epoch 479, weight, value: tensor([[ 0.0390, 0.1324, -0.1820, ..., -0.1083, 0.0780, 0.0039], + [-0.1086, -0.1068, 0.1161, ..., -0.0482, -0.1165, 0.0565], + [-0.0970, -0.0705, -0.1194, ..., -0.1376, -0.0174, 0.0571], + ..., + [-0.0083, -0.1360, -0.1412, ..., -0.1151, -0.0522, 0.0771], + [-0.1344, -0.1179, -0.1188, ..., -0.0667, -0.0487, -0.1013], + [ 0.0215, -0.0325, -0.0579, ..., -0.1045, 0.0104, -0.0236]], + device='cuda:0'), grad: tensor([[-3.5419e-03, -7.1907e-04, 3.3665e-04, ..., 6.8378e-04, + -6.6423e-04, -1.4534e-03], + [ 3.7599e-04, 1.0973e-04, 3.8195e-04, ..., 3.3712e-04, + 1.7405e-05, -1.3542e-03], + [ 1.2197e-03, 3.2187e-04, 3.3545e-04, ..., 6.1131e-04, + 1.5783e-04, -4.3225e-04], + ..., + [ 4.4942e-04, -2.1839e-04, 3.3784e-04, ..., 2.6107e-04, + 3.1233e-05, 3.1033e-03], + [-2.4033e-03, -7.6675e-04, -1.1549e-03, ..., -4.1656e-03, + 4.5985e-05, -1.0395e-03], + [ 1.5411e-03, 3.8481e-04, -1.5488e-03, ..., 4.3559e-04, + 1.3065e-04, 1.3981e-03]], device='cuda:0') +Epoch 479, bias, value: tensor([ 0.0088, 0.0062, -0.0045, -0.0172, 0.0078, -0.0134, -0.0141, -0.0180, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([-0.0026, -0.0071, 0.0140, 0.0349, 0.0080, -0.0131, -0.0352, 0.0192, + -0.0065, -0.0117], device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 216.56, cls_loss 0.4406 cls_loss_mapping 0.0008 cls_loss_causal 0.4144 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.10 lr 0.00001000 +Epoch 480, weight, value: tensor([[ 0.0390, 0.1324, -0.1819, ..., -0.1082, 0.0782, 0.0039], + [-0.1085, -0.1069, 0.1162, ..., -0.0482, -0.1165, 0.0565], + [-0.0971, -0.0705, -0.1194, ..., -0.1376, -0.0175, 0.0572], + ..., + [-0.0082, -0.1360, -0.1413, ..., -0.1151, -0.0522, 0.0771], + [-0.1344, -0.1178, -0.1189, ..., -0.0666, -0.0489, -0.1014], + [ 0.0216, -0.0324, -0.0576, ..., -0.1045, 0.0105, -0.0235]], + device='cuda:0'), grad: tensor([[ 4.0054e-03, 1.3266e-03, 1.1292e-03, ..., 1.8673e-03, + 1.5459e-03, 2.8877e-03], + [ 9.8133e-04, 1.6779e-05, 8.9884e-04, ..., 1.0067e-04, + 2.7224e-05, 1.6851e-03], + [ 7.6485e-04, 4.7827e-04, -2.1782e-03, ..., 1.4424e-04, + 8.9121e-04, -1.3971e-03], + ..., + [ 1.8930e-03, 1.0681e-04, 6.6948e-04, ..., 1.5110e-05, + 2.4235e-04, 4.7112e-03], + [ 2.2240e-03, 1.0109e-03, 9.8228e-04, ..., 1.3742e-03, + 1.1902e-03, 2.4967e-03], + [ 2.0428e-03, 1.9634e-04, 7.7248e-04, ..., 6.1750e-05, + 5.5981e-04, 1.7290e-03]], device='cuda:0') +Epoch 480, bias, value: tensor([ 0.0088, 0.0063, -0.0045, -0.0174, 0.0076, -0.0132, -0.0142, -0.0180, + -0.0052, 0.0089], device='cuda:0'), grad: tensor([-0.0047, 0.0159, -0.0176, 0.0091, -0.0186, 0.0222, -0.0435, 0.0194, + -0.0215, 0.0391], device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 216.52, cls_loss 0.4379 cls_loss_mapping 0.0007 cls_loss_causal 0.4153 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.06 lr 0.00001000 +Epoch 481, weight, value: tensor([[ 0.0391, 0.1326, -0.1819, ..., -0.1081, 0.0781, 0.0040], + [-0.1086, -0.1068, 0.1161, ..., -0.0480, -0.1166, 0.0563], + [-0.0972, -0.0707, -0.1195, ..., -0.1378, -0.0176, 0.0569], + ..., + [-0.0082, -0.1361, -0.1412, ..., -0.1151, -0.0522, 0.0773], + [-0.1346, -0.1179, -0.1190, ..., -0.0667, -0.0489, -0.1015], + [ 0.0216, -0.0326, -0.0578, ..., -0.1044, 0.0106, -0.0234]], + device='cuda:0'), grad: tensor([[ 1.3723e-03, 4.4489e-04, 4.3106e-04, ..., 2.8682e-04, + 6.7092e-06, 9.6893e-04], + [ 1.5569e-04, 1.2830e-05, 6.1226e-04, ..., 4.0078e-04, + 2.4550e-06, -4.4403e-03], + [-5.7840e-04, 2.0587e-04, -3.2310e-03, ..., -4.1161e-03, + 1.1392e-05, -5.1041e-03], + ..., + [ 2.3198e-04, 3.7372e-05, -1.7433e-03, ..., 1.6737e-04, + 8.4698e-05, 1.5535e-03], + [ 2.2876e-04, 1.1183e-05, 2.8014e-04, ..., 2.2745e-04, + 6.5118e-06, 8.2111e-04], + [ 2.3613e-03, 4.7982e-05, 6.0892e-04, ..., 1.8919e-04, + 1.2589e-03, 1.3199e-03]], device='cuda:0') +Epoch 481, bias, value: tensor([ 0.0088, 0.0062, -0.0046, -0.0172, 0.0076, -0.0131, -0.0141, -0.0179, + -0.0053, 0.0088], device='cuda:0'), grad: tensor([ 0.0159, -0.0202, -0.0203, 0.0099, 0.0033, -0.0094, 0.0168, -0.0168, + 0.0075, 0.0133], device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 216.84, cls_loss 0.4202 cls_loss_mapping 0.0007 cls_loss_causal 0.3915 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.07 lr 0.00001000 +Epoch 482, weight, value: tensor([[ 0.0390, 0.1328, -0.1818, ..., -0.1081, 0.0782, 0.0040], + [-0.1087, -0.1068, 0.1163, ..., -0.0482, -0.1166, 0.0565], + [-0.0972, -0.0706, -0.1195, ..., -0.1377, -0.0175, 0.0570], + ..., + [-0.0080, -0.1363, -0.1413, ..., -0.1150, -0.0522, 0.0772], + [-0.1347, -0.1179, -0.1190, ..., -0.0666, -0.0490, -0.1015], + [ 0.0216, -0.0327, -0.0579, ..., -0.1045, 0.0106, -0.0236]], + device='cuda:0'), grad: tensor([[-4.5113e-06, -9.7752e-06, 7.4971e-08, ..., 4.1239e-06, + -9.1344e-06, 2.6894e-04], + [ 1.7449e-05, 8.9332e-06, 8.3819e-08, ..., 1.5087e-07, + 1.9697e-07, -2.4891e-03], + [ 9.0480e-05, 4.9829e-05, 2.3190e-06, ..., 1.0533e-06, + 1.7677e-06, 1.2951e-03], + ..., + [ 3.6042e-06, 3.2261e-06, 1.8515e-06, ..., 1.7695e-08, + 3.0035e-07, 7.3290e-04], + [ 1.7481e-06, 2.9206e-06, 5.9186e-07, ..., 1.5516e-06, + 1.9036e-06, -1.9531e-03], + [ 2.4617e-05, 1.4335e-05, -1.0077e-06, ..., 1.5460e-07, + 2.4643e-06, 3.4428e-04]], device='cuda:0') +Epoch 482, bias, value: tensor([ 0.0088, 0.0063, -0.0046, -0.0172, 0.0075, -0.0131, -0.0141, -0.0180, + -0.0053, 0.0088], device='cuda:0'), grad: tensor([ 0.0017, -0.0134, 0.0111, 0.0061, 0.0013, 0.0037, 0.0029, 0.0045, + -0.0206, 0.0027], device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 216.28, cls_loss 0.4127 cls_loss_mapping 0.0006 cls_loss_causal 0.3852 re_mapping 0.0036 re_causal 0.0114 /// teacc 99.07 lr 0.00001000 +Epoch 483, weight, value: tensor([[ 0.0389, 0.1328, -0.1820, ..., -0.1081, 0.0783, 0.0041], + [-0.1088, -0.1068, 0.1164, ..., -0.0481, -0.1167, 0.0564], + [-0.0970, -0.0705, -0.1195, ..., -0.1377, -0.0175, 0.0572], + ..., + [-0.0081, -0.1362, -0.1413, ..., -0.1149, -0.0522, 0.0771], + [-0.1347, -0.1179, -0.1188, ..., -0.0666, -0.0490, -0.1015], + [ 0.0216, -0.0328, -0.0579, ..., -0.1046, 0.0105, -0.0236]], + device='cuda:0'), grad: tensor([[-1.4896e-03, -6.1226e-04, -2.4109e-03, ..., 8.0764e-06, + 0.0000e+00, -7.1945e-03], + [ 7.7820e-04, 2.9778e-04, 1.2264e-03, ..., -9.1270e-06, + 0.0000e+00, 2.3460e-03], + [ 1.2174e-05, 6.8903e-05, -3.8719e-03, ..., 6.1579e-06, + 0.0000e+00, -9.7322e-04], + ..., + [ 1.1330e-03, 2.1589e-04, 8.6069e-04, ..., 3.0156e-06, + 1.0477e-07, 2.1248e-03], + [ 2.9635e-04, 2.5201e-04, 9.4032e-04, ..., 1.6749e-04, + 4.1910e-09, 9.5510e-04], + [-9.0313e-04, 2.0087e-04, 8.7357e-04, ..., 4.0159e-06, + -8.1435e-06, 3.2711e-04]], device='cuda:0') +Epoch 483, bias, value: tensor([ 0.0088, 0.0063, -0.0046, -0.0173, 0.0076, -0.0131, -0.0140, -0.0181, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([-0.0206, -0.0085, -0.0438, 0.0150, 0.0112, -0.0117, 0.0224, 0.0154, + 0.0113, 0.0093], device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 216.71, cls_loss 0.4375 cls_loss_mapping 0.0008 cls_loss_causal 0.4116 re_mapping 0.0034 re_causal 0.0110 /// teacc 99.08 lr 0.00001000 +Epoch 484, weight, value: tensor([[ 0.0391, 0.1329, -0.1819, ..., -0.1079, 0.0784, 0.0043], + [-0.1088, -0.1067, 0.1162, ..., -0.0481, -0.1167, 0.0564], + [-0.0970, -0.0707, -0.1195, ..., -0.1377, -0.0175, 0.0570], + ..., + [-0.0083, -0.1362, -0.1413, ..., -0.1150, -0.0521, 0.0771], + [-0.1348, -0.1178, -0.1186, ..., -0.0662, -0.0492, -0.1015], + [ 0.0215, -0.0329, -0.0579, ..., -0.1046, 0.0106, -0.0235]], + device='cuda:0'), grad: tensor([[ 3.9339e-04, -8.7202e-05, 1.3485e-03, ..., 2.7325e-06, + 4.3869e-04, 8.3256e-04], + [-2.8820e-03, 3.4198e-06, -3.8967e-03, ..., 5.8627e-07, + -3.7479e-03, -4.3526e-03], + [ 4.1962e-04, 1.4924e-05, 5.5933e-04, ..., 1.0669e-05, + 1.1759e-03, 2.8682e-04], + ..., + [ 8.3160e-03, 3.5204e-06, -1.9169e-03, ..., 9.8813e-07, + -1.9550e-05, 1.8950e-03], + [ 8.6486e-05, 1.7792e-05, 7.1621e-04, ..., 1.2226e-05, + 9.8419e-04, 3.5191e-04], + [ 3.0384e-03, 2.7239e-05, 2.1782e-03, ..., 3.1069e-06, + 5.2118e-04, 1.1282e-03]], device='cuda:0') +Epoch 484, bias, value: tensor([ 0.0088, 0.0063, -0.0047, -0.0173, 0.0076, -0.0130, -0.0141, -0.0179, + -0.0053, 0.0087], device='cuda:0'), grad: tensor([ 0.0218, -0.0640, -0.0016, -0.0320, 0.0297, 0.0249, 0.0220, 0.0052, + -0.0035, -0.0024], device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 216.61, cls_loss 0.4229 cls_loss_mapping 0.0009 cls_loss_causal 0.4016 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.05 lr 0.00001000 +Epoch 485, weight, value: tensor([[ 0.0392, 0.1329, -0.1819, ..., -0.1080, 0.0784, 0.0043], + [-0.1088, -0.1067, 0.1162, ..., -0.0482, -0.1165, 0.0563], + [-0.0971, -0.0707, -0.1196, ..., -0.1377, -0.0175, 0.0571], + ..., + [-0.0084, -0.1363, -0.1413, ..., -0.1150, -0.0522, 0.0770], + [-0.1348, -0.1178, -0.1186, ..., -0.0659, -0.0493, -0.1014], + [ 0.0215, -0.0330, -0.0579, ..., -0.1047, 0.0107, -0.0235]], + device='cuda:0'), grad: tensor([[-2.7008e-03, -3.5706e-03, -6.5851e-04, ..., 1.1049e-05, + -6.4230e-04, -1.8368e-03], + [ 3.7718e-04, 6.8069e-05, 5.6076e-04, ..., 1.3877e-06, + 6.1870e-05, -4.9782e-04], + [ 9.0551e-04, 1.5831e-04, 4.5252e-04, ..., 7.0706e-06, + 8.8215e-05, 1.8854e-03], + ..., + [-1.2436e-02, 8.6212e-04, -1.3437e-03, ..., 3.4366e-07, + -4.9515e-03, -1.0025e-02], + [ 2.8229e-03, 3.0327e-04, 5.3453e-04, ..., 2.2084e-05, + 5.3078e-05, -4.0398e-03], + [ 8.6517e-03, 7.9966e-04, 7.1287e-04, ..., 3.0808e-06, + 3.3798e-03, 6.2637e-03]], device='cuda:0') +Epoch 485, bias, value: tensor([ 0.0088, 0.0063, -0.0048, -0.0173, 0.0075, -0.0131, -0.0140, -0.0178, + -0.0052, 0.0087], device='cuda:0'), grad: tensor([-0.0096, -0.0063, 0.0188, 0.0254, 0.0291, -0.0398, 0.0152, -0.0311, + -0.0079, 0.0063], device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 216.37, cls_loss 0.4520 cls_loss_mapping 0.0010 cls_loss_causal 0.4270 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.09 lr 0.00001000 +Epoch 486, weight, value: tensor([[ 0.0392, 0.1330, -0.1820, ..., -0.1079, 0.0784, 0.0043], + [-0.1089, -0.1068, 0.1162, ..., -0.0484, -0.1164, 0.0562], + [-0.0969, -0.0706, -0.1195, ..., -0.1376, -0.0174, 0.0573], + ..., + [-0.0085, -0.1363, -0.1414, ..., -0.1151, -0.0522, 0.0770], + [-0.1347, -0.1178, -0.1186, ..., -0.0659, -0.0492, -0.1014], + [ 0.0215, -0.0331, -0.0580, ..., -0.1049, 0.0106, -0.0236]], + device='cuda:0'), grad: tensor([[ 9.6083e-04, -2.6941e-05, 4.8089e-04, ..., -8.2329e-06, + 4.3273e-04, 1.6012e-03], + [-4.2686e-03, -4.5449e-07, 7.5531e-04, ..., 3.2131e-07, + 4.0007e-04, 8.8024e-04], + [-3.3875e-03, 9.3058e-06, 5.5599e-04, ..., 4.6939e-07, + -3.0518e-03, 3.1376e-03], + ..., + [-1.7824e-03, 9.1270e-07, 5.6887e-04, ..., 5.5879e-09, + 6.3658e-04, -1.0490e-02], + [ 3.4771e-03, -2.2814e-05, 4.0030e-04, ..., 4.7311e-07, + 2.9612e-04, 1.5764e-03], + [ 3.9825e-03, 1.1496e-05, 5.9414e-04, ..., 3.2410e-07, + 4.5967e-04, 2.0943e-03]], device='cuda:0') +Epoch 486, bias, value: tensor([ 0.0088, 0.0063, -0.0047, -0.0172, 0.0075, -0.0132, -0.0140, -0.0179, + -0.0051, 0.0086], device='cuda:0'), grad: tensor([ 0.0162, -0.0082, -0.0198, -0.0364, 0.0282, 0.0191, -0.0190, -0.0333, + 0.0221, 0.0311], device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 216.13, cls_loss 0.4548 cls_loss_mapping 0.0010 cls_loss_causal 0.4319 re_mapping 0.0033 re_causal 0.0105 /// teacc 99.06 lr 0.00001000 +Epoch 487, weight, value: tensor([[ 0.0392, 0.1331, -0.1820, ..., -0.1079, 0.0784, 0.0042], + [-0.1087, -0.1069, 0.1163, ..., -0.0483, -0.1165, 0.0563], + [-0.0969, -0.0707, -0.1196, ..., -0.1376, -0.0173, 0.0573], + ..., + [-0.0084, -0.1363, -0.1413, ..., -0.1151, -0.0522, 0.0770], + [-0.1347, -0.1178, -0.1183, ..., -0.0657, -0.0490, -0.1013], + [ 0.0215, -0.0331, -0.0580, ..., -0.1049, 0.0105, -0.0235]], + device='cuda:0'), grad: tensor([[ 2.8896e-04, 9.5904e-05, 9.1887e-04, ..., 1.7555e-06, + 8.3804e-05, 7.7057e-04], + [ 3.9005e-04, 7.3433e-05, 1.5392e-03, ..., 5.1223e-09, + 6.4373e-05, 7.7915e-04], + [ 3.4428e-04, 1.7309e-04, 1.0490e-03, ..., 3.7756e-06, + 4.9442e-05, 9.2888e-04], + ..., + [-1.5106e-03, -1.2836e-03, -5.8556e-04, ..., 3.2596e-09, + 1.1349e-04, -2.0256e-03], + [-2.7447e-03, -1.2140e-03, -3.6669e-04, ..., 5.3823e-05, + -1.8492e-03, -3.1033e-03], + [ 5.7840e-04, 2.3675e-04, -2.5940e-03, ..., 5.8906e-07, + 1.7881e-04, -4.3893e-04]], device='cuda:0') +Epoch 487, bias, value: tensor([ 0.0088, 0.0064, -0.0048, -0.0173, 0.0074, -0.0131, -0.0141, -0.0179, + -0.0050, 0.0087], device='cuda:0'), grad: tensor([ 0.0105, 0.0141, 0.0113, 0.0395, -0.0194, -0.0282, 0.0165, -0.0175, + -0.0137, -0.0130], device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 216.25, cls_loss 0.4455 cls_loss_mapping 0.0009 cls_loss_causal 0.4187 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.06 lr 0.00001000 +Epoch 488, weight, value: tensor([[ 0.0393, 0.1331, -0.1820, ..., -0.1080, 0.0785, 0.0043], + [-0.1087, -0.1069, 0.1162, ..., -0.0484, -0.1164, 0.0562], + [-0.0970, -0.0707, -0.1196, ..., -0.1377, -0.0174, 0.0573], + ..., + [-0.0084, -0.1363, -0.1413, ..., -0.1151, -0.0522, 0.0770], + [-0.1347, -0.1180, -0.1184, ..., -0.0657, -0.0489, -0.1013], + [ 0.0216, -0.0330, -0.0579, ..., -0.1049, 0.0105, -0.0235]], + device='cuda:0'), grad: tensor([[ 8.4162e-05, 8.0168e-05, 4.3344e-04, ..., 1.9325e-07, + 1.4491e-05, -3.3722e-03], + [ 9.8407e-05, -8.3772e-07, 5.3358e-04, ..., -8.2096e-07, + 1.8906e-07, 8.5592e-04], + [-4.4966e-04, -9.5291e-03, -1.1833e-02, ..., 2.8405e-08, + -1.7271e-03, -2.8687e-03], + ..., + [ 2.9445e-04, 3.0044e-06, 5.5122e-04, ..., 1.0431e-07, + 4.9779e-07, 5.7268e-04], + [ 2.2709e-04, 1.9893e-05, 4.2319e-04, ..., 6.3237e-07, + 3.4831e-06, 6.6471e-04], + [ 1.3638e-04, 2.5019e-05, 4.8423e-04, ..., 3.7253e-08, + 4.4405e-06, 4.9448e-04]], device='cuda:0') +Epoch 488, bias, value: tensor([ 0.0087, 0.0063, -0.0048, -0.0175, 0.0073, -0.0130, -0.0141, -0.0179, + -0.0049, 0.0089], device='cuda:0'), grad: tensor([-0.0222, 0.0133, -0.0718, 0.0251, 0.0087, 0.0153, 0.0038, 0.0114, + 0.0049, 0.0116], device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 216.03, cls_loss 0.4407 cls_loss_mapping 0.0008 cls_loss_causal 0.4134 re_mapping 0.0035 re_causal 0.0110 /// teacc 99.08 lr 0.00001000 +Epoch 489, weight, value: tensor([[ 0.0392, 0.1332, -0.1820, ..., -0.1080, 0.0784, 0.0043], + [-0.1089, -0.1071, 0.1161, ..., -0.0484, -0.1164, 0.0561], + [-0.0970, -0.0706, -0.1196, ..., -0.1378, -0.0174, 0.0573], + ..., + [-0.0083, -0.1363, -0.1413, ..., -0.1152, -0.0523, 0.0770], + [-0.1348, -0.1182, -0.1184, ..., -0.0657, -0.0489, -0.1013], + [ 0.0217, -0.0330, -0.0580, ..., -0.1051, 0.0104, -0.0234]], + device='cuda:0'), grad: tensor([[ 7.5054e-04, -1.7881e-05, -1.8301e-03, ..., 4.9800e-05, + 4.7684e-04, 9.4223e-04], + [ 2.9182e-04, 1.7554e-05, 3.8338e-03, ..., 5.6148e-05, + 1.6522e-04, 1.4095e-03], + [ 1.4219e-03, 3.2306e-05, 8.7118e-04, ..., 2.7442e-04, + 8.6021e-04, 3.4599e-03], + ..., + [ 3.4380e-04, 1.8045e-05, 4.7922e-04, ..., 3.1203e-05, + 1.9455e-04, 3.4180e-03], + [-5.7716e-03, -4.0741e-03, -2.1713e-02, ..., -5.2185e-03, + 2.0742e-04, -1.2573e-02], + [ 4.4174e-03, 2.9392e-03, 1.2917e-02, ..., 3.8776e-03, + 1.2517e-04, 8.3923e-03]], device='cuda:0') +Epoch 489, bias, value: tensor([ 0.0087, 0.0062, -0.0049, -0.0175, 0.0072, -0.0128, -0.0140, -0.0179, + -0.0047, 0.0088], device='cuda:0'), grad: tensor([-0.0204, 0.0243, 0.0241, 0.0033, -0.0224, 0.0027, -0.0025, 0.0214, + -0.0311, 0.0006], device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 216.28, cls_loss 0.4252 cls_loss_mapping 0.0009 cls_loss_causal 0.3984 re_mapping 0.0035 re_causal 0.0109 /// teacc 99.08 lr 0.00001000 +Epoch 490, weight, value: tensor([[ 0.0392, 0.1333, -0.1819, ..., -0.1079, 0.0783, 0.0044], + [-0.1089, -0.1071, 0.1159, ..., -0.0485, -0.1164, 0.0560], + [-0.0971, -0.0705, -0.1195, ..., -0.1377, -0.0175, 0.0572], + ..., + [-0.0084, -0.1364, -0.1412, ..., -0.1152, -0.0523, 0.0771], + [-0.1348, -0.1183, -0.1184, ..., -0.0657, -0.0491, -0.1013], + [ 0.0218, -0.0330, -0.0580, ..., -0.1053, 0.0105, -0.0235]], + device='cuda:0'), grad: tensor([[ 3.8004e-04, -2.7523e-03, 5.5695e-04, ..., 1.7309e-04, + 7.4327e-05, 1.1044e-03], + [ 4.1676e-04, 1.2159e-04, 7.4387e-04, ..., 1.3828e-04, + 7.4565e-05, 1.0853e-03], + [-2.0742e-04, 3.4928e-04, 5.3883e-04, ..., -1.0624e-03, + -2.4438e-04, -4.2000e-03], + ..., + [ 4.1938e-04, 1.2767e-04, 6.9904e-04, ..., 7.7426e-05, + 6.7413e-05, 8.1348e-04], + [-2.7704e-04, 2.7227e-04, -1.4524e-03, ..., 1.2660e-04, + 9.8169e-05, -9.9838e-05], + [ 2.9469e-04, -2.4959e-05, 4.9496e-04, ..., 7.3731e-05, + -7.2241e-05, 8.9121e-04]], device='cuda:0') +Epoch 490, bias, value: tensor([ 0.0088, 0.0060, -0.0050, -0.0175, 0.0073, -0.0127, -0.0140, -0.0179, + -0.0049, 0.0088], device='cuda:0'), grad: tensor([-0.0177, 0.0148, -0.0132, 0.0297, -0.0500, 0.0116, 0.0160, 0.0131, + -0.0181, 0.0137], device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 216.34, cls_loss 0.4299 cls_loss_mapping 0.0010 cls_loss_causal 0.3973 re_mapping 0.0034 re_causal 0.0101 /// teacc 99.11 lr 0.00001000 +Epoch 491, weight, value: tensor([[ 0.0392, 0.1330, -0.1819, ..., -0.1080, 0.0784, 0.0044], + [-0.1090, -0.1070, 0.1158, ..., -0.0484, -0.1164, 0.0560], + [-0.0972, -0.0704, -0.1195, ..., -0.1377, -0.0175, 0.0572], + ..., + [-0.0085, -0.1364, -0.1412, ..., -0.1152, -0.0524, 0.0771], + [-0.1347, -0.1183, -0.1184, ..., -0.0657, -0.0490, -0.1012], + [ 0.0218, -0.0331, -0.0581, ..., -0.1053, 0.0104, -0.0235]], + device='cuda:0'), grad: tensor([[-1.5152e-02, -1.8005e-03, 4.7755e-04, ..., 4.0382e-06, + 6.6817e-05, -1.2465e-03], + [ 7.9346e-04, 1.5140e-04, -1.7958e-03, ..., 7.5065e-07, + 3.2878e-04, 4.1342e-04], + [ 2.2805e-04, 3.7456e-04, -4.4107e-04, ..., 7.9572e-06, + 3.2902e-04, -1.9178e-03], + ..., + [ 4.1122e-03, 1.7929e-04, 7.7105e-04, ..., 1.0245e-07, + 7.7009e-04, 9.2316e-04], + [ 3.8910e-03, -1.0347e-03, 8.2827e-04, ..., 3.1352e-05, + 3.2949e-04, 3.9597e-03], + [ 1.4734e-03, 1.1539e-03, -1.8330e-03, ..., 4.3493e-07, + 2.8539e-04, 1.8728e-04]], device='cuda:0') +Epoch 491, bias, value: tensor([ 0.0088, 0.0059, -0.0051, -0.0174, 0.0074, -0.0128, -0.0139, -0.0179, + -0.0048, 0.0088], device='cuda:0'), grad: tensor([-0.0406, -0.0135, -0.0070, -0.0033, 0.0163, 0.0121, 0.0058, 0.0049, + 0.0402, -0.0148], device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 216.04, cls_loss 0.4412 cls_loss_mapping 0.0009 cls_loss_causal 0.4107 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.09 lr 0.00001000 +Epoch 492, weight, value: tensor([[ 0.0392, 0.1331, -0.1819, ..., -0.1081, 0.0782, 0.0045], + [-0.1090, -0.1070, 0.1159, ..., -0.0483, -0.1165, 0.0561], + [-0.0972, -0.0704, -0.1194, ..., -0.1377, -0.0175, 0.0572], + ..., + [-0.0085, -0.1364, -0.1411, ..., -0.1152, -0.0524, 0.0770], + [-0.1347, -0.1184, -0.1185, ..., -0.0657, -0.0490, -0.1012], + [ 0.0218, -0.0331, -0.0581, ..., -0.1055, 0.0103, -0.0236]], + device='cuda:0'), grad: tensor([[-3.3188e-04, -1.5163e-03, 1.2526e-06, ..., 1.6922e-06, + -2.2662e-04, 1.2064e-03], + [ 2.5439e-04, 1.0830e-04, -1.8597e-05, ..., 9.1270e-08, + 1.3840e-04, -3.3798e-03], + [ 2.9993e-04, 3.3355e-04, 5.5507e-07, ..., 5.4948e-08, + 1.7059e-04, 2.8954e-03], + ..., + [ 2.8038e-04, 8.6844e-05, 7.4785e-07, ..., 1.1176e-08, + 1.5318e-04, 1.3733e-03], + [ 2.2399e-04, 4.0221e-04, 1.1545e-04, ..., 1.5587e-05, + 1.7536e-04, -1.2617e-03], + [ 1.8156e-04, 1.4687e-04, 1.2079e-06, ..., 1.0710e-07, + 8.1435e-06, 1.1082e-03]], device='cuda:0') +Epoch 492, bias, value: tensor([ 0.0089, 0.0060, -0.0051, -0.0173, 0.0073, -0.0128, -0.0140, -0.0178, + -0.0049, 0.0089], device='cuda:0'), grad: tensor([ 0.0035, -0.0196, 0.0180, 0.0109, 0.0106, 0.0052, 0.0062, 0.0113, + -0.0528, 0.0067], device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 216.35, cls_loss 0.4243 cls_loss_mapping 0.0007 cls_loss_causal 0.3935 re_mapping 0.0035 re_causal 0.0108 /// teacc 99.09 lr 0.00001000 +Epoch 493, weight, value: tensor([[ 0.0393, 0.1331, -0.1821, ..., -0.1081, 0.0784, 0.0044], + [-0.1089, -0.1070, 0.1158, ..., -0.0482, -0.1165, 0.0562], + [-0.0973, -0.0703, -0.1193, ..., -0.1378, -0.0175, 0.0572], + ..., + [-0.0086, -0.1363, -0.1411, ..., -0.1152, -0.0526, 0.0771], + [-0.1347, -0.1184, -0.1185, ..., -0.0656, -0.0490, -0.1012], + [ 0.0218, -0.0332, -0.0581, ..., -0.1055, 0.0102, -0.0235]], + device='cuda:0'), grad: tensor([[ 2.6569e-03, 1.2789e-03, 1.4448e-03, ..., 5.3316e-05, + 9.5034e-04, 1.6174e-03], + [ 1.0900e-03, 3.8338e-04, 1.3018e-03, ..., 8.4758e-05, + 3.8791e-04, -3.0441e-03], + [-4.1366e-04, 3.0684e-04, -2.1973e-03, ..., -1.4343e-03, + 5.2500e-04, -1.1444e-03], + ..., + [-6.3801e-04, 2.8324e-04, 7.3099e-04, ..., 1.6224e-04, + -1.4293e-04, -9.3307e-03], + [-8.3017e-04, 1.0478e-04, -3.0136e-03, ..., 3.4785e-04, + -6.2048e-05, 3.2940e-03], + [ 3.5725e-03, 1.1816e-03, 7.9250e-04, ..., 1.2851e-04, + 1.2693e-03, 5.4893e-03]], device='cuda:0') +Epoch 493, bias, value: tensor([ 0.0089, 0.0061, -0.0051, -0.0173, 0.0073, -0.0130, -0.0140, -0.0178, + -0.0049, 0.0089], device='cuda:0'), grad: tensor([ 0.0100, 0.0040, -0.0231, -0.0258, -0.0507, 0.0224, 0.0273, -0.0191, + 0.0124, 0.0425], device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 216.41, cls_loss 0.4282 cls_loss_mapping 0.0008 cls_loss_causal 0.4038 re_mapping 0.0036 re_causal 0.0109 /// teacc 99.05 lr 0.00001000 +Epoch 494, weight, value: tensor([[ 0.0394, 0.1332, -0.1820, ..., -0.1078, 0.0785, 0.0044], + [-0.1090, -0.1070, 0.1158, ..., -0.0481, -0.1165, 0.0562], + [-0.0974, -0.0703, -0.1193, ..., -0.1379, -0.0174, 0.0572], + ..., + [-0.0085, -0.1363, -0.1411, ..., -0.1153, -0.0526, 0.0770], + [-0.1347, -0.1184, -0.1185, ..., -0.0655, -0.0491, -0.1013], + [ 0.0219, -0.0332, -0.0580, ..., -0.1054, 0.0103, -0.0236]], + device='cuda:0'), grad: tensor([[-1.1005e-03, -1.7891e-03, -5.4502e-04, ..., -1.4658e-03, + -6.1178e-04, -5.5733e-03], + [-8.2064e-04, -2.3663e-04, -5.8031e-04, ..., 6.2525e-05, + -6.6900e-04, -7.6389e-04], + [ 4.5753e-04, 2.5320e-04, 1.1520e-03, ..., 1.5116e-04, + 1.9336e-04, -4.2381e-03], + ..., + [ 1.2684e-03, 7.4208e-05, -1.2255e-03, ..., 2.3708e-05, + 9.0480e-05, 1.0939e-03], + [ 3.1870e-06, 3.0851e-04, -1.2779e-03, ..., 1.9693e-04, + 1.5402e-04, 8.9502e-04], + [ 2.1042e-02, 2.2674e-04, 1.1797e-03, ..., 1.1981e-04, + 2.4021e-04, 7.8964e-03]], device='cuda:0') +Epoch 494, bias, value: tensor([ 0.0090, 0.0060, -0.0051, -0.0173, 0.0072, -0.0128, -0.0141, -0.0177, + -0.0051, 0.0089], device='cuda:0'), grad: tensor([-0.0087, -0.0086, -0.0117, -0.0081, -0.0096, 0.0193, 0.0042, -0.0079, + -0.0134, 0.0445], device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 216.22, cls_loss 0.4589 cls_loss_mapping 0.0007 cls_loss_causal 0.4332 re_mapping 0.0036 re_causal 0.0117 /// teacc 99.04 lr 0.00001000 +Epoch 495, weight, value: tensor([[ 0.0394, 0.1333, -0.1820, ..., -0.1078, 0.0786, 0.0044], + [-0.1089, -0.1070, 0.1157, ..., -0.0482, -0.1166, 0.0562], + [-0.0974, -0.0703, -0.1192, ..., -0.1379, -0.0175, 0.0572], + ..., + [-0.0085, -0.1363, -0.1411, ..., -0.1154, -0.0527, 0.0771], + [-0.1348, -0.1182, -0.1184, ..., -0.0654, -0.0492, -0.1014], + [ 0.0218, -0.0330, -0.0579, ..., -0.1054, 0.0105, -0.0238]], + device='cuda:0'), grad: tensor([[ 9.1851e-05, -1.7953e-04, 3.6287e-04, ..., 4.2617e-06, + -1.9598e-04, -1.9913e-03], + [ 4.4775e-04, 3.0661e-04, 1.3485e-03, ..., -3.1218e-06, + 4.5276e-04, 1.8616e-03], + [ 1.7881e-04, 1.7345e-04, 4.1580e-04, ..., 6.5863e-06, + 7.2777e-05, 1.2941e-03], + ..., + [-1.0628e-04, 1.7846e-04, 4.6539e-04, ..., 4.7088e-06, + -1.6356e-04, 1.5650e-03], + [-2.6178e-04, -2.3043e-04, -4.1580e-03, ..., -1.7905e-04, + 7.6592e-06, 1.0767e-03], + [ 1.3247e-03, 5.0402e-04, 6.2037e-04, ..., 3.1143e-05, + 3.4881e-04, -5.8632e-03]], device='cuda:0') +Epoch 495, bias, value: tensor([ 0.0089, 0.0060, -0.0050, -0.0173, 0.0074, -0.0129, -0.0140, -0.0178, + -0.0051, 0.0088], device='cuda:0'), grad: tensor([-0.0449, 0.0300, 0.0173, -0.0100, 0.0146, 0.0204, 0.0192, 0.0188, + -0.0425, -0.0229], device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 216.02, cls_loss 0.4428 cls_loss_mapping 0.0007 cls_loss_causal 0.4234 re_mapping 0.0036 re_causal 0.0117 /// teacc 99.07 lr 0.00001000 +Epoch 496, weight, value: tensor([[ 0.0394, 0.1332, -0.1820, ..., -0.1079, 0.0786, 0.0045], + [-0.1090, -0.1071, 0.1157, ..., -0.0482, -0.1167, 0.0561], + [-0.0974, -0.0703, -0.1192, ..., -0.1379, -0.0176, 0.0573], + ..., + [-0.0084, -0.1363, -0.1410, ..., -0.1152, -0.0527, 0.0772], + [-0.1349, -0.1182, -0.1184, ..., -0.0654, -0.0493, -0.1015], + [ 0.0218, -0.0331, -0.0581, ..., -0.1055, 0.0105, -0.0239]], + device='cuda:0'), grad: tensor([[ 1.3220e-04, -1.6624e-06, -1.7853e-03, ..., 3.4332e-05, + 1.3383e-06, -2.1350e-04], + [ 2.5344e-04, -7.2084e-07, 5.4550e-04, ..., 2.5053e-07, + 3.2503e-07, 6.5041e-04], + [ 1.8120e-04, 1.5153e-06, 4.3011e-04, ..., 1.3262e-06, + 2.2352e-06, 4.8876e-04], + ..., + [ 1.5068e-03, -1.3029e-06, 4.7612e-04, ..., 2.5146e-08, + 7.1859e-04, 1.3018e-03], + [ 2.1374e-04, 6.2771e-06, 3.9458e-04, ..., 1.4976e-06, + 4.3124e-05, 4.2748e-04], + [-3.5214e-04, 9.9987e-06, 4.2820e-04, ..., 9.2201e-08, + -1.5616e-04, 3.5691e-04]], device='cuda:0') +Epoch 496, bias, value: tensor([ 0.0089, 0.0059, -0.0049, -0.0171, 0.0074, -0.0130, -0.0139, -0.0177, + -0.0052, 0.0086], device='cuda:0'), grad: tensor([-0.0242, 0.0113, 0.0085, -0.0221, -0.0234, 0.0080, 0.0082, 0.0184, + 0.0078, 0.0074], device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 216.20, cls_loss 0.4360 cls_loss_mapping 0.0007 cls_loss_causal 0.4120 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.07 lr 0.00001000 +Epoch 497, weight, value: tensor([[ 0.0392, 0.1332, -0.1821, ..., -0.1078, 0.0786, 0.0044], + [-0.1091, -0.1072, 0.1157, ..., -0.0483, -0.1167, 0.0563], + [-0.0974, -0.0704, -0.1192, ..., -0.1379, -0.0176, 0.0572], + ..., + [-0.0083, -0.1364, -0.1410, ..., -0.1153, -0.0526, 0.0773], + [-0.1348, -0.1182, -0.1186, ..., -0.0654, -0.0493, -0.1017], + [ 0.0216, -0.0331, -0.0581, ..., -0.1056, 0.0104, -0.0239]], + device='cuda:0'), grad: tensor([[-4.2694e-02, 7.8430e-03, 3.0613e-03, ..., 2.0332e-03, + 8.3148e-05, -2.0847e-03], + [ 2.2697e-03, 3.8277e-07, 3.2687e-04, ..., -4.5542e-07, + 1.0514e-04, 4.4861e-03], + [ 1.6518e-03, 1.3255e-05, 1.6761e-04, ..., 5.2527e-07, + 1.6773e-04, 3.4237e-03], + ..., + [ 2.3210e-04, -3.7774e-06, 1.5998e-04, ..., 1.2480e-07, + -2.8563e-04, 1.3094e-03], + [ 1.1377e-03, 1.0997e-04, 2.0373e-04, ..., 3.1471e-05, + 9.5785e-05, 2.7504e-03], + [ 2.3537e-03, 2.2724e-06, 1.2255e-04, ..., 4.2841e-07, + 4.3893e-04, -3.5515e-03]], device='cuda:0') +Epoch 497, bias, value: tensor([ 0.0088, 0.0059, -0.0049, -0.0171, 0.0074, -0.0129, -0.0137, -0.0176, + -0.0053, 0.0085], device='cuda:0'), grad: tensor([-0.0262, 0.0276, 0.0195, -0.0172, -0.0245, 0.0334, -0.0452, 0.0123, + 0.0243, -0.0039], device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 216.22, cls_loss 0.4474 cls_loss_mapping 0.0008 cls_loss_causal 0.4140 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.06 lr 0.00001000 +Epoch 498, weight, value: tensor([[ 0.0394, 0.1332, -0.1822, ..., -0.1077, 0.0786, 0.0044], + [-0.1090, -0.1071, 0.1158, ..., -0.0482, -0.1166, 0.0564], + [-0.0975, -0.0704, -0.1192, ..., -0.1379, -0.0176, 0.0572], + ..., + [-0.0086, -0.1364, -0.1411, ..., -0.1154, -0.0527, 0.0771], + [-0.1348, -0.1182, -0.1187, ..., -0.0654, -0.0492, -0.1017], + [ 0.0218, -0.0333, -0.0582, ..., -0.1057, 0.0104, -0.0237]], + device='cuda:0'), grad: tensor([[ 4.0054e-04, 4.1842e-04, 1.2007e-03, ..., 7.2479e-05, + 4.5228e-04, 1.9569e-03], + [-2.8095e-03, -4.9496e-04, 2.3866e-04, ..., 9.6679e-05, + -3.1924e-04, 6.2275e-04], + [ 5.2595e-04, 3.1166e-03, 3.3379e-03, ..., 9.4593e-05, + 2.1191e-03, 4.9324e-03], + ..., + [ 8.2016e-04, 9.9301e-05, 1.2808e-03, ..., 7.4863e-05, + 3.2115e-04, 2.8076e-03], + [-4.6349e-04, -8.7786e-04, -7.2289e-04, ..., 7.4208e-05, + -4.4680e-04, -1.7662e-03], + [-1.2016e-03, 2.9993e-04, -1.1864e-03, ..., -6.9952e-04, + -6.4039e-04, -5.3635e-03]], device='cuda:0') +Epoch 498, bias, value: tensor([ 0.0088, 0.0061, -0.0051, -0.0170, 0.0073, -0.0131, -0.0136, -0.0178, + -0.0052, 0.0085], device='cuda:0'), grad: tensor([ 0.0237, 0.0240, 0.0531, -0.0697, 0.0040, 0.0313, -0.0005, 0.0324, + -0.0658, -0.0327], device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 216.39, cls_loss 0.4261 cls_loss_mapping 0.0008 cls_loss_causal 0.4029 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.06 lr 0.00001000 +Epoch 499, weight, value: tensor([[ 0.0394, 0.1332, -0.1823, ..., -0.1077, 0.0786, 0.0045], + [-0.1091, -0.1071, 0.1157, ..., -0.0483, -0.1166, 0.0563], + [-0.0975, -0.0705, -0.1192, ..., -0.1378, -0.0177, 0.0572], + ..., + [-0.0086, -0.1364, -0.1411, ..., -0.1155, -0.0526, 0.0771], + [-0.1346, -0.1183, -0.1187, ..., -0.0653, -0.0493, -0.1016], + [ 0.0218, -0.0333, -0.0581, ..., -0.1057, 0.0105, -0.0235]], + device='cuda:0'), grad: tensor([[ 2.2256e-04, -4.2953e-03, 4.2394e-06, ..., 5.6684e-05, + 1.5507e-03, 1.6415e-04], + [-1.4615e-04, 1.3435e-04, -7.6443e-06, ..., 1.1332e-05, + 3.0780e-04, -1.7900e-03], + [ 2.6751e-04, 7.1108e-05, 3.1203e-05, ..., 2.1845e-05, + 5.0926e-04, 2.9063e-04], + ..., + [-2.4271e-04, 1.7956e-05, 1.2629e-05, ..., 2.3860e-06, + 3.4118e-04, 8.0407e-05], + [-6.1464e-04, 3.1948e-03, -2.9039e-04, ..., 2.6703e-05, + -5.4054e-03, 2.4116e-04], + [ 1.4567e-04, 5.2452e-05, 8.2329e-06, ..., 6.8173e-06, + 4.8780e-04, 2.2328e-04]], device='cuda:0') +Epoch 499, bias, value: tensor([ 0.0087, 0.0060, -0.0050, -0.0171, 0.0074, -0.0131, -0.0137, -0.0178, + -0.0051, 0.0088], device='cuda:0'), grad: tensor([-0.0123, -0.0236, 0.0072, 0.0084, 0.0061, 0.0059, 0.0087, 0.0050, + -0.0113, 0.0060], device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 216.34, cls_loss 0.4480 cls_loss_mapping 0.0008 cls_loss_causal 0.4246 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.04 lr 0.00001000 +Epoch 500, weight, value: tensor([[ 0.0395, 0.1333, -0.1823, ..., -0.1075, 0.0787, 0.0044], + [-0.1090, -0.1072, 0.1157, ..., -0.0482, -0.1166, 0.0564], + [-0.0975, -0.0706, -0.1192, ..., -0.1379, -0.0178, 0.0572], + ..., + [-0.0085, -0.1365, -0.1412, ..., -0.1156, -0.0528, 0.0770], + [-0.1346, -0.1185, -0.1187, ..., -0.0654, -0.0493, -0.1015], + [ 0.0218, -0.0333, -0.0582, ..., -0.1056, 0.0105, -0.0235]], + device='cuda:0'), grad: tensor([[-4.7340e-03, -4.5109e-04, -8.2493e-04, ..., 3.1739e-05, + -2.3346e-03, -2.5711e-03], + [ 1.8442e-04, 3.6418e-05, -2.1493e-04, ..., 1.1481e-05, + 8.8692e-05, -1.1873e-03], + [ 6.1703e-04, 1.0586e-04, 1.5030e-03, ..., 2.1130e-05, + 1.9169e-04, 2.8419e-03], + ..., + [ 1.2989e-03, 3.2210e-04, -4.8027e-03, ..., 1.0692e-05, + 8.6641e-04, 1.2827e-03], + [ 1.4763e-03, 8.8501e-04, -9.9373e-04, ..., -2.0638e-05, + 3.1447e-04, 2.5616e-03], + [-1.0509e-03, -1.2531e-03, -9.2745e-04, ..., 5.6103e-06, + 3.0375e-04, -6.6223e-03]], device='cuda:0') +Epoch 500, bias, value: tensor([ 0.0087, 0.0060, -0.0050, -0.0171, 0.0074, -0.0129, -0.0137, -0.0179, + -0.0051, 0.0088], device='cuda:0'), grad: tensor([-3.2013e-02, -4.3303e-05, 3.5614e-02, 3.5767e-02, 8.9417e-03, + 2.5131e-02, 3.3264e-02, -4.9561e-02, 1.0506e-02, -6.7688e-02], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 216.48, cls_loss 0.4289 cls_loss_mapping 0.0008 cls_loss_causal 0.3987 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.05 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.779999 98.790001 ... 89.138016 76.584227 +ShearY 98.909996 98.919998 ... 89.138016 73.370252 +AutoContrast 98.900002 98.979996 ... 89.138016 66.252623 +Invert 98.970001 99.019997 ... 89.138016 72.543168 +Equalize 98.409996 98.439995 ... 89.138016 66.964914 +Solarize 98.430000 98.599998 ... 89.138016 68.342981 +SolarizeAdd 98.599998 98.619995 ... 89.138016 72.159985 +Posterize 98.959999 99.000000 ... 89.138016 76.318024 +Contrast 99.139999 99.169998 ... 89.138016 78.336514 +Color 99.019997 99.049995 ... 89.138016 67.886716 +Brightness 99.119995 99.199997 ... 89.138016 77.374899 +Sharpness 99.000000 99.029999 ... 89.138016 78.293999 +NoiseSalt 99.029999 99.049995 ... 89.138016 70.436442 +NoiseGaussian 98.979996 99.070000 ... 89.138016 65.005304 +w/o do (original x) 99.050000 0.000000 ... 0.000000 79.455813 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.01 70.432545 77.824686 79.692243 89.287494 79.309242 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last +randm: False +stride: 3 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.879997 98.919998 ... 89.486794 76.881126 +ShearY 98.879997 98.849998 ... 89.486794 73.554698 +AutoContrast 98.930000 99.019997 ... 89.486794 66.492076 +Invert 98.979996 99.000000 ... 89.486794 72.618074 +Equalize 98.419998 98.459999 ... 89.486794 67.132525 +Solarize 98.470001 98.580002 ... 89.486794 68.999302 +SolarizeAdd 98.619995 98.659996 ... 89.486794 72.501183 +Posterize 98.949997 99.019997 ... 89.486794 76.556273 +Contrast 99.169998 99.199997 ... 89.486794 78.399466 +Color 99.059998 99.029999 ... 89.486794 67.984040 +Brightness 99.159996 99.190002 ... 89.486794 77.668425 +Sharpness 99.010002 99.049995 ... 89.486794 78.452616 +NoiseSalt 99.049995 99.059998 ... 89.486794 70.660622 +NoiseGaussian 98.979996 99.040001 ... 89.486794 65.199389 +w/o do (original x) 99.030000 0.000000 ... 0.000000 79.652907 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.08 70.79748 77.613598 79.80739 89.78575 79.501055 diff --git a/Meta-causal/code-withStyleAttack/71591.error b/Meta-causal/code-withStyleAttack/71591.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/71591.log b/Meta-causal/code-withStyleAttack/71591.log new file mode 100644 index 0000000000000000000000000000000000000000..73d707496918f5bee214826849703f2b973f115a --- /dev/null +++ b/Meta-causal/code-withStyleAttack/71591.log @@ -0,0 +1,14129 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[ 0.0224, -0.0232, 0.0043, ..., -0.0119, -0.0122, 0.0253], + [-0.0080, 0.0261, -0.0234, ..., 0.0273, 0.0169, 0.0103], + [-0.0056, -0.0279, -0.0151, ..., 0.0159, -0.0221, -0.0221], + ..., + [-0.0188, 0.0174, 0.0250, ..., -0.0149, -0.0295, -0.0102], + [-0.0309, -0.0118, 0.0149, ..., 0.0134, -0.0106, 0.0078], + [-0.0157, -0.0052, -0.0018, ..., -0.0136, -0.0029, 0.0029]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0230, 0.0176, 0.0105, 0.0247, 0.0259, -0.0210, 0.0279, 0.0249, + -0.0094, 0.0097], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 257.47, cls_loss 1.2559 cls_loss_mapping 1.8083 cls_loss_causal 2.2263 re_mapping 0.1568 re_causal 0.1696 /// teacc 87.85 lr 0.00010000 +Epoch 2, weight, value: tensor([[ 0.0241, -0.0285, 0.0023, ..., -0.0155, -0.0122, 0.0194], + [-0.0152, 0.0298, -0.0252, ..., 0.0230, 0.0169, 0.0144], + [-0.0088, -0.0348, -0.0152, ..., 0.0214, -0.0221, -0.0263], + ..., + [-0.0186, 0.0241, 0.0235, ..., -0.0218, -0.0295, -0.0052], + [-0.0291, -0.0171, 0.0128, ..., 0.0112, -0.0106, 0.0029], + [-0.0181, 0.0007, -0.0035, ..., -0.0194, -0.0029, 0.0085]], + device='cuda:0'), grad: tensor([[-0.0174, -0.0008, 0.0000, ..., -0.0024, 0.0000, 0.0036], + [ 0.0025, -0.0122, 0.0000, ..., -0.0083, 0.0000, -0.0067], + [ 0.0043, 0.0056, 0.0000, ..., -0.0072, 0.0000, 0.0059], + ..., + [ 0.0145, 0.0328, 0.0000, ..., 0.0021, 0.0000, 0.0338], + [ 0.0212, 0.0201, 0.0000, ..., 0.0119, 0.0000, 0.0200], + [-0.0119, -0.0600, 0.0000, ..., 0.0052, 0.0000, -0.0406]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0235, 0.0177, 0.0102, 0.0258, 0.0249, -0.0200, 0.0273, 0.0242, + -0.0097, 0.0101], device='cuda:0'), grad: tensor([-0.0154, -0.0094, 0.0124, -0.0131, -0.0129, -0.0883, 0.0379, 0.0379, + 0.0498, 0.0011], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 257.97, cls_loss 0.3567 cls_loss_mapping 0.7536 cls_loss_causal 1.9254 re_mapping 0.2080 re_causal 0.2755 /// teacc 93.43 lr 0.00010000 +Epoch 3, weight, value: tensor([[ 0.0251, -0.0308, 0.0023, ..., -0.0185, -0.0056, 0.0171], + [-0.0185, 0.0324, -0.0253, ..., 0.0239, 0.0081, 0.0161], + [-0.0098, -0.0366, -0.0152, ..., 0.0242, -0.0270, -0.0277], + ..., + [-0.0193, 0.0275, 0.0235, ..., -0.0223, -0.0193, -0.0030], + [-0.0275, -0.0200, 0.0127, ..., 0.0103, -0.0203, -0.0002], + [-0.0162, 0.0022, -0.0035, ..., -0.0248, -0.0107, 0.0108]], + device='cuda:0'), grad: tensor([[ 7.3385e-04, 1.4381e-03, 0.0000e+00, ..., 2.1706e-03, + -1.8013e-04, 1.9550e-03], + [ 3.3340e-03, -1.5396e-02, 0.0000e+00, ..., -6.9237e-03, + -8.6129e-05, -1.2657e-02], + [ 1.0513e-02, 6.8817e-03, 0.0000e+00, ..., 2.0020e-02, + 8.8215e-05, 9.2850e-03], + ..., + [ 4.4800e-02, 8.1482e-02, 0.0000e+00, ..., 5.5389e-03, + 5.2071e-03, 5.6061e-02], + [-6.2866e-02, -7.4280e-02, 0.0000e+00, ..., -1.7746e-02, + -5.4321e-03, -4.8676e-02], + [ 4.3526e-03, -2.8648e-03, 0.0000e+00, ..., 5.0735e-03, + 1.4663e-04, -3.8643e-03]], device='cuda:0') +Epoch 3, bias, value: tensor([-0.0237, 0.0180, 0.0099, 0.0255, 0.0246, -0.0192, 0.0269, 0.0236, + -0.0095, 0.0107], device='cuda:0'), grad: tensor([ 0.0019, -0.0063, 0.0215, 0.0150, -0.0021, -0.0075, -0.0162, 0.0435, + -0.0566, 0.0069], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 257.13, cls_loss 0.2092 cls_loss_mapping 0.4211 cls_loss_causal 1.6320 re_mapping 0.1518 re_causal 0.2407 /// teacc 95.52 lr 0.00010000 +Epoch 4, weight, value: tensor([[ 0.0252, -0.0332, 0.0023, ..., -0.0194, 0.0012, 0.0146], + [-0.0201, 0.0342, -0.0253, ..., 0.0244, 0.0167, 0.0174], + [-0.0097, -0.0383, -0.0152, ..., 0.0260, -0.0281, -0.0274], + ..., + [-0.0186, 0.0297, 0.0235, ..., -0.0209, -0.0110, -0.0013], + [-0.0272, -0.0226, 0.0127, ..., 0.0101, -0.0299, -0.0029], + [-0.0158, 0.0032, -0.0035, ..., -0.0289, -0.0181, 0.0122]], + device='cuda:0'), grad: tensor([[ 1.1311e-03, 3.0899e-03, 0.0000e+00, ..., 1.6088e-03, + 7.3254e-05, 4.0855e-03], + [-2.0752e-03, -9.1248e-03, 0.0000e+00, ..., -6.0997e-03, + -4.6730e-03, -7.2746e-03], + [ 4.9210e-03, 3.9291e-03, 0.0000e+00, ..., 1.7109e-03, + 1.0643e-03, 3.6640e-03], + ..., + [-1.6556e-02, -4.4098e-02, 0.0000e+00, ..., -3.6964e-03, + -1.1543e-02, -3.7567e-02], + [ 3.7422e-03, 9.9335e-03, 0.0000e+00, ..., 6.0081e-03, + 5.1651e-03, 3.0136e-03], + [ 1.2169e-02, 2.4658e-02, 0.0000e+00, ..., 3.8376e-03, + 5.7220e-03, 2.5681e-02]], device='cuda:0') +Epoch 4, bias, value: tensor([-0.0234, 0.0180, 0.0100, 0.0254, 0.0243, -0.0195, 0.0266, 0.0239, + -0.0093, 0.0109], device='cuda:0'), grad: tensor([ 0.0035, -0.0077, 0.0067, -0.0054, 0.0124, -0.0022, -0.0071, -0.0330, + 0.0099, 0.0229], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 257.21, cls_loss 0.1541 cls_loss_mapping 0.2857 cls_loss_causal 1.4368 re_mapping 0.1179 re_causal 0.2076 /// teacc 96.04 lr 0.00010000 +Epoch 5, weight, value: tensor([[ 0.0250, -0.0348, 0.0022, ..., -0.0206, 0.0023, 0.0129], + [-0.0212, 0.0352, -0.0254, ..., 0.0248, 0.0245, 0.0181], + [-0.0096, -0.0405, -0.0154, ..., 0.0282, -0.0281, -0.0266], + ..., + [-0.0189, 0.0312, 0.0232, ..., -0.0210, -0.0071, -0.0004], + [-0.0267, -0.0239, 0.0127, ..., 0.0086, -0.0345, -0.0048], + [-0.0158, 0.0037, -0.0037, ..., -0.0324, -0.0208, 0.0130]], + device='cuda:0'), grad: tensor([[-1.0805e-03, 5.5122e-04, 0.0000e+00, ..., -5.0879e-04, + 3.4189e-04, -6.5899e-04], + [ 6.9389e-03, 3.6144e-03, 0.0000e+00, ..., 5.4588e-03, + -7.5102e-05, 7.0076e-03], + [ 1.1711e-03, 1.2236e-03, 0.0000e+00, ..., -3.9196e-04, + 3.0823e-03, 6.9427e-04], + ..., + [-1.1063e-03, -3.9177e-03, 0.0000e+00, ..., 2.8133e-03, + -2.2812e-03, -1.9217e-04], + [-8.8272e-03, -3.1734e-04, 0.0000e+00, ..., -2.7435e-02, + -1.0025e-02, -1.3275e-02], + [ 9.3639e-05, 4.0507e-04, 0.0000e+00, ..., 2.5482e-03, + 1.6775e-03, -9.0456e-04]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0235, 0.0180, 0.0101, 0.0257, 0.0240, -0.0194, 0.0264, 0.0236, + -0.0092, 0.0110], device='cuda:0'), grad: tensor([-0.0024, 0.0159, 0.0062, 0.0068, 0.0086, -0.0017, 0.0017, 0.0011, + -0.0389, 0.0028], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 256.43, cls_loss 0.1133 cls_loss_mapping 0.2108 cls_loss_causal 1.3546 re_mapping 0.0935 re_causal 0.1818 /// teacc 97.05 lr 0.00010000 +Epoch 6, weight, value: tensor([[ 0.0254, -0.0372, 0.0022, ..., -0.0218, 0.0021, 0.0104], + [-0.0217, 0.0357, -0.0254, ..., 0.0251, 0.0302, 0.0186], + [-0.0097, -0.0419, -0.0154, ..., 0.0299, -0.0279, -0.0255], + ..., + [-0.0187, 0.0330, 0.0232, ..., -0.0219, -0.0032, 0.0007], + [-0.0266, -0.0252, 0.0127, ..., 0.0087, -0.0391, -0.0061], + [-0.0156, 0.0038, -0.0037, ..., -0.0347, -0.0239, 0.0136]], + device='cuda:0'), grad: tensor([[-0.0034, -0.0009, 0.0000, ..., -0.0012, 0.0002, -0.0010], + [ 0.0022, -0.0045, 0.0000, ..., 0.0053, -0.0022, -0.0004], + [-0.0040, 0.0008, 0.0000, ..., -0.0127, -0.0033, -0.0066], + ..., + [ 0.0037, -0.0011, 0.0000, ..., 0.0026, 0.0004, 0.0012], + [ 0.0034, 0.0028, 0.0000, ..., 0.0003, 0.0016, 0.0026], + [ 0.0076, 0.0035, 0.0000, ..., 0.0009, 0.0014, 0.0002]], + device='cuda:0') +Epoch 6, bias, value: tensor([-0.0234, 0.0178, 0.0100, 0.0258, 0.0239, -0.0196, 0.0263, 0.0235, + -0.0089, 0.0111], device='cuda:0'), grad: tensor([-0.0069, 0.0013, -0.0064, -0.0048, -0.0015, -0.0074, 0.0073, 0.0049, + 0.0051, 0.0085], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 256.52, cls_loss 0.1039 cls_loss_mapping 0.1858 cls_loss_causal 1.2690 re_mapping 0.0754 re_causal 0.1534 /// teacc 97.53 lr 0.00010000 +Epoch 7, weight, value: tensor([[ 0.0261, -0.0387, 0.0022, ..., -0.0231, 0.0025, 0.0088], + [-0.0219, 0.0365, -0.0254, ..., 0.0250, 0.0356, 0.0191], + [-0.0095, -0.0440, -0.0154, ..., 0.0315, -0.0284, -0.0253], + ..., + [-0.0194, 0.0343, 0.0232, ..., -0.0225, -0.0011, 0.0018], + [-0.0265, -0.0263, 0.0127, ..., 0.0086, -0.0432, -0.0074], + [-0.0154, 0.0040, -0.0037, ..., -0.0374, -0.0260, 0.0140]], + device='cuda:0'), grad: tensor([[ 0.0026, 0.0049, 0.0000, ..., 0.0011, 0.0019, 0.0042], + [ 0.0060, 0.0029, 0.0000, ..., 0.0068, 0.0007, 0.0029], + [ 0.0066, 0.0054, 0.0000, ..., 0.0055, 0.0020, 0.0031], + ..., + [-0.0035, -0.0091, 0.0000, ..., -0.0084, -0.0077, -0.0009], + [ 0.0023, 0.0031, 0.0000, ..., 0.0002, 0.0010, -0.0013], + [-0.0034, -0.0161, 0.0000, ..., -0.0015, 0.0006, -0.0151]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0231, 0.0179, 0.0101, 0.0260, 0.0237, -0.0197, 0.0260, 0.0233, + -0.0088, 0.0111], device='cuda:0'), grad: tensor([ 0.0059, 0.0116, 0.0125, -0.0158, 0.0069, 0.0016, -0.0012, -0.0090, + 0.0012, -0.0138], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 256.53, cls_loss 0.0832 cls_loss_mapping 0.1463 cls_loss_causal 1.1815 re_mapping 0.0645 re_causal 0.1366 /// teacc 97.75 lr 0.00010000 +Epoch 8, weight, value: tensor([[ 0.0262, -0.0406, 0.0022, ..., -0.0244, 0.0016, 0.0064], + [-0.0219, 0.0372, -0.0254, ..., 0.0248, 0.0395, 0.0198], + [-0.0095, -0.0453, -0.0154, ..., 0.0330, -0.0280, -0.0244], + ..., + [-0.0196, 0.0356, 0.0232, ..., -0.0235, 0.0013, 0.0025], + [-0.0260, -0.0273, 0.0127, ..., 0.0084, -0.0471, -0.0087], + [-0.0152, 0.0042, -0.0037, ..., -0.0393, -0.0283, 0.0147]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0007, 0.0000, ..., 0.0003, 0.0004, 0.0007], + [ 0.0009, 0.0004, 0.0000, ..., 0.0010, 0.0005, 0.0008], + [-0.0027, -0.0005, 0.0000, ..., -0.0050, -0.0021, -0.0023], + ..., + [ 0.0056, 0.0032, 0.0000, ..., 0.0036, -0.0001, 0.0058], + [-0.0021, 0.0007, 0.0000, ..., -0.0005, 0.0002, 0.0011], + [-0.0049, -0.0074, 0.0000, ..., 0.0002, -0.0002, -0.0066]], + device='cuda:0') +Epoch 8, bias, value: tensor([-0.0232, 0.0178, 0.0104, 0.0260, 0.0237, -0.0198, 0.0256, 0.0232, + -0.0084, 0.0111], device='cuda:0'), grad: tensor([ 0.0014, 0.0015, -0.0024, -0.0007, -0.0011, 0.0021, 0.0005, 0.0070, + -0.0029, -0.0052], device='cuda:0') +100 +0.0001 +changing lr +epoch 7, time 255.84, cls_loss 0.0838 cls_loss_mapping 0.1494 cls_loss_causal 1.1530 re_mapping 0.0562 re_causal 0.1223 /// teacc 97.37 lr 0.00010000 +Epoch 9, weight, value: tensor([[ 0.0268, -0.0415, 0.0022, ..., -0.0243, 0.0009, 0.0048], + [-0.0223, 0.0374, -0.0254, ..., 0.0241, 0.0422, 0.0197], + [-0.0093, -0.0469, -0.0154, ..., 0.0342, -0.0286, -0.0233], + ..., + [-0.0197, 0.0368, 0.0232, ..., -0.0242, 0.0044, 0.0033], + [-0.0257, -0.0279, 0.0127, ..., 0.0083, -0.0497, -0.0097], + [-0.0152, 0.0043, -0.0037, ..., -0.0416, -0.0303, 0.0150]], + device='cuda:0'), grad: tensor([[ 3.8552e-04, 8.8978e-04, 0.0000e+00, ..., 6.4039e-04, + 3.0589e-04, 9.1887e-04], + [-8.5545e-04, -4.1753e-05, 0.0000e+00, ..., 1.1597e-03, + -1.4629e-03, 1.7667e-04], + [-1.0090e-03, 4.0016e-03, 0.0000e+00, ..., -1.2655e-03, + -2.0170e-04, 2.6474e-03], + ..., + [-5.9929e-03, -1.3519e-02, 0.0000e+00, ..., 5.9938e-04, + -4.8065e-03, -1.1536e-02], + [ 2.1687e-03, 3.3226e-03, 0.0000e+00, ..., 2.2755e-03, + 1.2569e-03, 3.6831e-03], + [ 5.6915e-03, 1.1658e-02, 0.0000e+00, ..., 2.2388e-04, + 4.2076e-03, 8.9645e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0230, 0.0173, 0.0105, 0.0263, 0.0236, -0.0201, 0.0254, 0.0234, + -0.0081, 0.0110], device='cuda:0'), grad: tensor([ 0.0017, 0.0012, 0.0036, -0.0080, -0.0111, 0.0070, -0.0001, -0.0131, + 0.0067, 0.0121], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 256.65, cls_loss 0.0806 cls_loss_mapping 0.1362 cls_loss_causal 1.0933 re_mapping 0.0513 re_causal 0.1157 /// teacc 97.81 lr 0.00010000 +Epoch 10, weight, value: tensor([[ 0.0272, -0.0432, 0.0022, ..., -0.0245, 0.0011, 0.0033], + [-0.0221, 0.0383, -0.0254, ..., 0.0239, 0.0453, 0.0201], + [-0.0093, -0.0484, -0.0154, ..., 0.0356, -0.0298, -0.0232], + ..., + [-0.0200, 0.0376, 0.0232, ..., -0.0249, 0.0057, 0.0038], + [-0.0253, -0.0282, 0.0127, ..., 0.0072, -0.0524, -0.0103], + [-0.0151, 0.0047, -0.0037, ..., -0.0430, -0.0311, 0.0155]], + device='cuda:0'), grad: tensor([[ 6.1131e-04, 7.0667e-04, 0.0000e+00, ..., 2.3117e-03, + 2.4295e-04, 3.7408e-04], + [ 2.1347e-02, 1.1879e-02, 0.0000e+00, ..., 7.2956e-05, + 1.4297e-02, 6.4049e-03], + [ 8.1348e-04, 5.2166e-04, 0.0000e+00, ..., 7.7665e-05, + 2.5916e-04, 3.1638e-04], + ..., + [-4.8780e-04, -1.4143e-03, 0.0000e+00, ..., 8.2970e-05, + -9.2506e-04, -1.0128e-03], + [-1.5316e-03, 1.6785e-04, 0.0000e+00, ..., 4.8423e-04, + 4.9353e-04, 2.5749e-04], + [-2.3239e-02, -1.4091e-02, 0.0000e+00, ..., 2.2769e-04, + -1.5732e-02, -7.3013e-03]], device='cuda:0') +Epoch 10, bias, value: tensor([-0.0227, 0.0173, 0.0106, 0.0264, 0.0232, -0.0200, 0.0248, 0.0232, + -0.0079, 0.0113], device='cuda:0'), grad: tensor([ 3.6716e-03, 2.8427e-02, 1.3494e-03, 2.8152e-03, 3.7938e-05, + 1.5383e-03, -4.9667e-03, -2.0123e-04, -1.0252e-03, -3.1647e-02], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 256.37, cls_loss 0.0589 cls_loss_mapping 0.1077 cls_loss_causal 1.0827 re_mapping 0.0458 re_causal 0.1084 /// teacc 97.82 lr 0.00010000 +Epoch 11, weight, value: tensor([[ 0.0273, -0.0443, 0.0022, ..., -0.0250, 0.0006, 0.0023], + [-0.0221, 0.0381, -0.0254, ..., 0.0238, 0.0477, 0.0205], + [-0.0092, -0.0495, -0.0154, ..., 0.0367, -0.0307, -0.0226], + ..., + [-0.0201, 0.0387, 0.0232, ..., -0.0257, 0.0076, 0.0042], + [-0.0252, -0.0282, 0.0127, ..., 0.0067, -0.0546, -0.0107], + [-0.0152, 0.0050, -0.0037, ..., -0.0441, -0.0311, 0.0159]], + device='cuda:0'), grad: tensor([[-4.4250e-03, -2.8801e-03, 0.0000e+00, ..., -1.6747e-03, + 1.7762e-04, -9.9361e-05], + [ 3.5000e-03, 1.6632e-03, 0.0000e+00, ..., 1.9360e-03, + 1.2016e-03, 3.2959e-03], + [-1.0040e-02, -6.9733e-03, 0.0000e+00, ..., -9.7656e-03, + -7.0801e-03, -1.4786e-02], + ..., + [ 5.2986e-03, 3.9597e-03, 0.0000e+00, ..., 6.0844e-03, + 3.9558e-03, 8.3008e-03], + [ 2.4586e-03, 1.4696e-03, 0.0000e+00, ..., 9.0456e-04, + 6.9857e-04, 1.7128e-03], + [-1.4267e-03, -2.2144e-03, 0.0000e+00, ..., 2.1112e-04, + 2.1362e-04, -1.9836e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0229, 0.0174, 0.0106, 0.0266, 0.0231, -0.0203, 0.0250, 0.0233, + -0.0077, 0.0111], device='cuda:0'), grad: tensor([-0.0159, 0.0048, -0.0163, 0.0236, 0.0017, -0.0213, 0.0110, 0.0104, + 0.0040, -0.0020], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 256.71, cls_loss 0.0515 cls_loss_mapping 0.0953 cls_loss_causal 1.0188 re_mapping 0.0421 re_causal 0.1035 /// teacc 98.27 lr 0.00010000 +Epoch 12, weight, value: tensor([[ 2.7578e-02, -4.5355e-02, 2.2385e-03, ..., -2.5915e-02, + 8.7393e-05, 9.3150e-04], + [-2.2334e-02, 3.8262e-02, -2.5420e-02, ..., 2.3188e-02, + 4.8786e-02, 2.0211e-02], + [-8.9505e-03, -5.0475e-02, -1.5361e-02, ..., 3.7974e-02, + -3.0312e-02, -2.1672e-02], + ..., + [-2.0410e-02, 3.9681e-02, 2.3202e-02, ..., -2.6385e-02, + 9.7325e-03, 4.7005e-03], + [-2.4811e-02, -2.8785e-02, 1.2677e-02, ..., 6.4490e-03, + -5.7005e-02, -1.1297e-02], + [-1.5696e-02, 4.7733e-03, -3.7345e-03, ..., -4.5391e-02, + -3.2535e-02, 1.5676e-02]], device='cuda:0'), grad: tensor([[ 0.0004, 0.0004, 0.0000, ..., 0.0008, 0.0003, 0.0009], + [-0.0003, 0.0005, 0.0000, ..., -0.0003, -0.0007, -0.0002], + [ 0.0018, 0.0015, 0.0000, ..., 0.0009, 0.0013, 0.0020], + ..., + [-0.0026, -0.0067, 0.0000, ..., -0.0041, -0.0047, -0.0078], + [ 0.0025, 0.0020, 0.0000, ..., 0.0023, 0.0019, 0.0036], + [ 0.0007, 0.0007, 0.0000, ..., 0.0010, 0.0005, 0.0016]], + device='cuda:0') +Epoch 12, bias, value: tensor([-0.0229, 0.0170, 0.0108, 0.0268, 0.0233, -0.0200, 0.0245, 0.0233, + -0.0075, 0.0107], device='cuda:0'), grad: tensor([ 0.0020, 0.0002, 0.0051, -0.0071, -0.0014, 0.0031, 0.0002, -0.0137, + 0.0077, 0.0038], device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 255.62, cls_loss 0.0565 cls_loss_mapping 0.1010 cls_loss_causal 1.0194 re_mapping 0.0380 re_causal 0.0976 /// teacc 98.25 lr 0.00010000 +Epoch 13, weight, value: tensor([[ 0.0278, -0.0463, 0.0022, ..., -0.0267, -0.0008, -0.0004], + [-0.0220, 0.0389, -0.0254, ..., 0.0234, 0.0518, 0.0208], + [-0.0091, -0.0513, -0.0154, ..., 0.0386, -0.0312, -0.0209], + ..., + [-0.0205, 0.0406, 0.0232, ..., -0.0274, 0.0111, 0.0050], + [-0.0248, -0.0293, 0.0127, ..., 0.0065, -0.0597, -0.0120], + [-0.0153, 0.0049, -0.0037, ..., -0.0466, -0.0336, 0.0160]], + device='cuda:0'), grad: tensor([[ 1.0830e-04, 8.8930e-05, 0.0000e+00, ..., 4.6659e-04, + 8.0526e-05, 1.7452e-04], + [ 1.3128e-05, -5.1832e-04, 0.0000e+00, ..., 2.1255e-04, + -1.0338e-03, -4.3917e-04], + [ 1.1134e-04, 2.0778e-04, 0.0000e+00, ..., -5.4598e-04, + 1.4520e-04, -2.8968e-04], + ..., + [ 1.4639e-04, -7.2718e-04, 0.0000e+00, ..., 1.1063e-04, + -2.5892e-04, -2.3866e-04], + [-8.9836e-04, 1.4162e-04, 0.0000e+00, ..., 1.6165e-04, + 2.3162e-04, 2.4486e-04], + [ 1.7977e-03, 3.9935e-04, 0.0000e+00, ..., 1.4138e-04, + 3.3236e-04, 6.7759e-04]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0230, 0.0171, 0.0108, 0.0267, 0.0232, -0.0201, 0.0246, 0.0232, + -0.0076, 0.0111], device='cuda:0'), grad: tensor([ 6.0940e-04, -5.4359e-05, 4.0984e-04, 1.0672e-03, -2.0593e-05, + -6.3362e-03, 7.7343e-04, 3.0422e-04, -1.7118e-03, 4.9629e-03], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 256.43, cls_loss 0.0439 cls_loss_mapping 0.0805 cls_loss_causal 0.9475 re_mapping 0.0352 re_causal 0.0912 /// teacc 98.38 lr 0.00010000 +Epoch 14, weight, value: tensor([[ 0.0280, -0.0473, 0.0022, ..., -0.0272, -0.0016, -0.0016], + [-0.0220, 0.0395, -0.0254, ..., 0.0231, 0.0536, 0.0210], + [-0.0088, -0.0523, -0.0154, ..., 0.0395, -0.0319, -0.0200], + ..., + [-0.0207, 0.0412, 0.0232, ..., -0.0280, 0.0123, 0.0052], + [-0.0246, -0.0302, 0.0127, ..., 0.0063, -0.0614, -0.0131], + [-0.0155, 0.0051, -0.0037, ..., -0.0478, -0.0344, 0.0164]], + device='cuda:0'), grad: tensor([[-2.1038e-03, 3.2306e-04, 0.0000e+00, ..., -1.3247e-03, + 2.3067e-05, -4.9174e-05], + [ 1.8454e-04, 4.7088e-04, 0.0000e+00, ..., -1.8811e-04, + -3.8195e-04, -9.7632e-05], + [ 8.5306e-04, 2.9683e-04, 0.0000e+00, ..., 4.3845e-04, + 4.5490e-04, 3.1257e-04], + ..., + [-3.2539e-03, -6.8665e-03, 0.0000e+00, ..., 8.1778e-05, + -4.6120e-03, -2.6035e-03], + [ 7.3338e-04, 5.5313e-04, 0.0000e+00, ..., 1.5202e-03, + 4.0722e-04, 4.6825e-04], + [ 2.4834e-03, 3.7098e-03, 0.0000e+00, ..., 2.1648e-04, + 3.0041e-03, 1.0653e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0229, 0.0168, 0.0109, 0.0269, 0.0233, -0.0202, 0.0243, 0.0232, + -0.0072, 0.0110], device='cuda:0'), grad: tensor([-0.0038, 0.0005, 0.0017, 0.0060, 0.0010, -0.0062, 0.0014, -0.0057, + 0.0011, 0.0041], device='cuda:0') +100 +0.0001 +changing lr +epoch 13, time 256.59, cls_loss 0.0544 cls_loss_mapping 0.0930 cls_loss_causal 0.9703 re_mapping 0.0329 re_causal 0.0886 /// teacc 98.19 lr 0.00010000 +Epoch 15, weight, value: tensor([[ 0.0282, -0.0485, 0.0022, ..., -0.0279, -0.0021, -0.0029], + [-0.0216, 0.0400, -0.0254, ..., 0.0226, 0.0554, 0.0217], + [-0.0088, -0.0532, -0.0154, ..., 0.0402, -0.0327, -0.0200], + ..., + [-0.0210, 0.0419, 0.0232, ..., -0.0280, 0.0140, 0.0056], + [-0.0242, -0.0305, 0.0127, ..., 0.0060, -0.0631, -0.0139], + [-0.0153, 0.0051, -0.0037, ..., -0.0484, -0.0357, 0.0166]], + device='cuda:0'), grad: tensor([[-6.6566e-04, 3.5977e-04, 0.0000e+00, ..., 1.2815e-04, + 3.0279e-04, 9.3162e-05], + [ 1.7653e-03, 1.6069e-03, 0.0000e+00, ..., 1.0567e-03, + 8.0919e-04, 2.1973e-03], + [-4.3221e-03, -3.0651e-03, 0.0000e+00, ..., -4.7493e-03, + -1.6642e-03, -6.3057e-03], + ..., + [-1.9050e-04, -8.3694e-03, 0.0000e+00, ..., 2.0294e-03, + -5.4665e-03, -4.8447e-03], + [ 1.5459e-03, 7.9679e-04, 0.0000e+00, ..., 3.9577e-04, + 5.4598e-04, 7.8487e-04], + [ 3.2406e-03, 7.9575e-03, 0.0000e+00, ..., 1.3590e-04, + 4.1008e-03, 9.2087e-03]], device='cuda:0') +Epoch 15, bias, value: tensor([-0.0232, 0.0167, 0.0108, 0.0269, 0.0229, -0.0201, 0.0245, 0.0232, + -0.0070, 0.0112], device='cuda:0'), grad: tensor([-0.0008, 0.0043, -0.0097, 0.0003, -0.0026, -0.0071, 0.0037, -0.0040, + 0.0045, 0.0114], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 258.53, cls_loss 0.0448 cls_loss_mapping 0.0834 cls_loss_causal 0.9786 re_mapping 0.0307 re_causal 0.0889 /// teacc 98.54 lr 0.00010000 +Epoch 16, weight, value: tensor([[ 0.0291, -0.0495, 0.0022, ..., -0.0285, -0.0030, -0.0042], + [-0.0212, 0.0400, -0.0254, ..., 0.0222, 0.0578, 0.0219], + [-0.0088, -0.0547, -0.0154, ..., 0.0413, -0.0339, -0.0196], + ..., + [-0.0211, 0.0431, 0.0232, ..., -0.0284, 0.0160, 0.0066], + [-0.0243, -0.0305, 0.0127, ..., 0.0054, -0.0640, -0.0143], + [-0.0154, 0.0051, -0.0037, ..., -0.0495, -0.0369, 0.0165]], + device='cuda:0'), grad: tensor([[ 3.4189e-04, 4.1795e-04, 0.0000e+00, ..., 5.8365e-04, + 2.8801e-04, 3.8671e-04], + [ 5.9414e-04, 5.8699e-04, 0.0000e+00, ..., 1.3599e-03, + 3.5715e-04, 8.5783e-04], + [ 1.9205e-04, 2.8133e-04, 0.0000e+00, ..., 9.7942e-04, + 7.5340e-05, 3.0327e-04], + ..., + [ 9.2316e-03, 9.9258e-03, 0.0000e+00, ..., 3.4833e-04, + 7.0381e-03, 4.8447e-03], + [-1.0620e-02, -1.3107e-02, 0.0000e+00, ..., 1.5163e-03, + -9.6817e-03, -6.5193e-03], + [ 1.5554e-03, 1.9484e-03, 0.0000e+00, ..., 9.1672e-05, + 1.1206e-03, 1.0328e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0228, 0.0170, 0.0109, 0.0266, 0.0228, -0.0200, 0.0243, 0.0236, + -0.0073, 0.0109], device='cuda:0'), grad: tensor([ 0.0011, 0.0021, 0.0015, -0.0030, -0.0050, 0.0023, -0.0005, 0.0152, + -0.0163, 0.0026], device='cuda:0') +100 +0.0001 +changing lr +epoch 15, time 257.80, cls_loss 0.0442 cls_loss_mapping 0.0810 cls_loss_causal 0.9234 re_mapping 0.0291 re_causal 0.0819 /// teacc 98.49 lr 0.00010000 +Epoch 17, weight, value: tensor([[ 0.0292, -0.0508, 0.0022, ..., -0.0288, -0.0038, -0.0053], + [-0.0215, 0.0405, -0.0254, ..., 0.0214, 0.0596, 0.0218], + [-0.0085, -0.0557, -0.0154, ..., 0.0423, -0.0343, -0.0187], + ..., + [-0.0212, 0.0436, 0.0232, ..., -0.0287, 0.0168, 0.0071], + [-0.0241, -0.0312, 0.0127, ..., 0.0049, -0.0656, -0.0153], + [-0.0152, 0.0053, -0.0037, ..., -0.0509, -0.0376, 0.0168]], + device='cuda:0'), grad: tensor([[-4.6730e-04, 2.9340e-05, 0.0000e+00, ..., -5.6362e-04, + 6.9380e-05, 7.7963e-05], + [ 1.7029e-02, 3.8738e-03, 0.0000e+00, ..., 2.5320e-04, + 1.3748e-02, 2.5678e-04], + [-9.4366e-04, -1.9181e-04, 0.0000e+00, ..., -7.1478e-04, + -5.2977e-04, -1.2417e-03], + ..., + [ 7.4720e-04, 4.3392e-05, 0.0000e+00, ..., 5.3644e-04, + 2.5582e-04, 7.1287e-04], + [-1.7761e-02, -3.9177e-03, 0.0000e+00, ..., 2.5511e-04, + -1.4633e-02, 4.7278e-04], + [ 2.1172e-04, 3.6740e-04, 0.0000e+00, ..., 1.2040e-04, + 1.3912e-04, 9.2888e-04]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0230, 0.0167, 0.0112, 0.0268, 0.0226, -0.0197, 0.0243, 0.0235, + -0.0073, 0.0108], device='cuda:0'), grad: tensor([-0.0017, 0.0276, -0.0015, 0.0003, -0.0017, 0.0008, 0.0009, 0.0015, + -0.0277, 0.0015], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 257.26, cls_loss 0.0372 cls_loss_mapping 0.0736 cls_loss_causal 0.8876 re_mapping 0.0287 re_causal 0.0815 /// teacc 98.67 lr 0.00010000 +Epoch 18, weight, value: tensor([[ 0.0293, -0.0515, 0.0022, ..., -0.0296, -0.0044, -0.0064], + [-0.0210, 0.0414, -0.0254, ..., 0.0208, 0.0618, 0.0223], + [-0.0085, -0.0565, -0.0154, ..., 0.0430, -0.0349, -0.0180], + ..., + [-0.0214, 0.0440, 0.0232, ..., -0.0291, 0.0177, 0.0073], + [-0.0238, -0.0317, 0.0127, ..., 0.0048, -0.0669, -0.0160], + [-0.0156, 0.0051, -0.0037, ..., -0.0517, -0.0391, 0.0168]], + device='cuda:0'), grad: tensor([[ 2.4152e-04, 6.7651e-05, 0.0000e+00, ..., 2.4343e-04, + 2.1458e-04, 3.0231e-04], + [ 7.0333e-04, 4.3344e-04, 0.0000e+00, ..., 1.4830e-04, + 3.9649e-04, 6.8426e-04], + [-3.1209e-04, 1.2231e-04, 0.0000e+00, ..., -8.8024e-04, + -6.2656e-04, -6.4182e-04], + ..., + [ 1.6756e-03, 6.7234e-04, 0.0000e+00, ..., 5.8556e-04, + -1.8954e-04, 1.5011e-03], + [-2.2659e-03, 9.8467e-05, 0.0000e+00, ..., 1.3793e-04, + -2.5320e-04, -1.1835e-03], + [-1.9705e-04, -1.8082e-03, 0.0000e+00, ..., 2.5451e-05, + 1.7548e-04, -1.5564e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0232, 0.0167, 0.0113, 0.0270, 0.0228, -0.0197, 0.0243, 0.0235, + -0.0069, 0.0102], device='cuda:0'), grad: tensor([ 0.0003, 0.0015, -0.0008, -0.0002, 0.0007, 0.0006, -0.0001, 0.0037, + -0.0036, -0.0021], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 256.63, cls_loss 0.0340 cls_loss_mapping 0.0710 cls_loss_causal 0.8941 re_mapping 0.0274 re_causal 0.0757 /// teacc 98.69 lr 0.00010000 +Epoch 19, weight, value: tensor([[ 0.0294, -0.0528, 0.0022, ..., -0.0301, -0.0056, -0.0075], + [-0.0211, 0.0412, -0.0254, ..., 0.0203, 0.0628, 0.0218], + [-0.0084, -0.0576, -0.0154, ..., 0.0442, -0.0363, -0.0179], + ..., + [-0.0217, 0.0447, 0.0232, ..., -0.0293, 0.0196, 0.0081], + [-0.0233, -0.0315, 0.0127, ..., 0.0041, -0.0676, -0.0164], + [-0.0155, 0.0051, -0.0037, ..., -0.0524, -0.0399, 0.0170]], + device='cuda:0'), grad: tensor([[ 1.0042e-03, 8.3864e-05, 0.0000e+00, ..., 3.1042e-04, + 1.7583e-04, 2.3258e-04], + [ 1.3428e-03, 3.6359e-04, 0.0000e+00, ..., 3.5501e-04, + 5.5218e-04, 6.7949e-04], + [-5.4073e-04, 2.6250e-04, 0.0000e+00, ..., -3.3321e-03, + -6.0320e-04, -1.8396e-03], + ..., + [-3.4183e-05, -2.5063e-03, 0.0000e+00, ..., 2.2316e-03, + -1.1864e-03, -4.0388e-04], + [ 1.7290e-03, 2.3890e-04, 0.0000e+00, ..., 4.2462e-04, + -5.7411e-04, -2.8300e-04], + [ 1.4906e-03, 1.0099e-03, 0.0000e+00, ..., 5.3614e-05, + 7.7724e-04, 8.5735e-04]], device='cuda:0') +Epoch 19, bias, value: tensor([-0.0234, 0.0163, 0.0112, 0.0275, 0.0227, -0.0200, 0.0240, 0.0236, + -0.0066, 0.0103], device='cuda:0'), grad: tensor([ 0.0019, 0.0030, -0.0023, -0.0119, 0.0009, 0.0023, -0.0011, 0.0005, + 0.0037, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 18, time 255.71, cls_loss 0.0272 cls_loss_mapping 0.0576 cls_loss_causal 0.8790 re_mapping 0.0258 re_causal 0.0755 /// teacc 98.60 lr 0.00010000 +Epoch 20, weight, value: tensor([[ 0.0296, -0.0531, 0.0022, ..., -0.0306, -0.0058, -0.0082], + [-0.0210, 0.0413, -0.0254, ..., 0.0196, 0.0641, 0.0216], + [-0.0086, -0.0585, -0.0154, ..., 0.0450, -0.0365, -0.0172], + ..., + [-0.0217, 0.0456, 0.0232, ..., -0.0302, 0.0214, 0.0087], + [-0.0230, -0.0320, 0.0127, ..., 0.0036, -0.0694, -0.0172], + [-0.0152, 0.0052, -0.0037, ..., -0.0536, -0.0405, 0.0171]], + device='cuda:0'), grad: tensor([[ 9.6464e-04, 4.0102e-04, 0.0000e+00, ..., 5.9366e-05, + 2.2233e-04, 5.7411e-04], + [-3.4833e-04, 1.2627e-03, 0.0000e+00, ..., -9.7847e-04, + -4.1151e-04, 2.1446e-04], + [ 3.7785e-03, 1.2751e-03, 0.0000e+00, ..., 1.7393e-04, + 9.5844e-04, 1.9894e-03], + ..., + [-9.7580e-03, -8.2550e-03, 0.0000e+00, ..., 1.0550e-04, + -5.3444e-03, -7.5150e-03], + [ 1.3733e-03, 6.0749e-04, 0.0000e+00, ..., 1.8370e-04, + 5.0020e-04, 6.6423e-04], + [ 6.9427e-03, 4.9095e-03, 0.0000e+00, ..., 1.0073e-04, + 2.5902e-03, 4.2076e-03]], device='cuda:0') +Epoch 20, bias, value: tensor([-0.0234, 0.0162, 0.0112, 0.0272, 0.0226, -0.0197, 0.0241, 0.0239, + -0.0068, 0.0104], device='cuda:0'), grad: tensor([ 0.0015, -0.0010, 0.0058, -0.0040, -0.0025, 0.0017, 0.0001, -0.0143, + 0.0023, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 256.13, cls_loss 0.0230 cls_loss_mapping 0.0512 cls_loss_causal 0.8657 re_mapping 0.0258 re_causal 0.0770 /// teacc 98.63 lr 0.00010000 +Epoch 21, weight, value: tensor([[ 0.0295, -0.0543, 0.0022, ..., -0.0314, -0.0069, -0.0096], + [-0.0205, 0.0416, -0.0254, ..., 0.0195, 0.0655, 0.0219], + [-0.0086, -0.0592, -0.0154, ..., 0.0458, -0.0367, -0.0166], + ..., + [-0.0220, 0.0462, 0.0232, ..., -0.0308, 0.0225, 0.0088], + [-0.0226, -0.0318, 0.0127, ..., 0.0030, -0.0700, -0.0175], + [-0.0151, 0.0053, -0.0037, ..., -0.0541, -0.0412, 0.0173]], + device='cuda:0'), grad: tensor([[ 2.6822e-05, 8.7738e-05, 0.0000e+00, ..., 2.9206e-05, + 7.8022e-05, 6.5923e-05], + [-4.8294e-03, -7.7844e-05, 0.0000e+00, ..., -3.3226e-03, + -5.4779e-03, -4.3945e-03], + [ 3.8700e-03, 3.6150e-05, 0.0000e+00, ..., 2.4376e-03, + 4.6310e-03, 3.5229e-03], + ..., + [ 1.3268e-04, -4.5538e-04, 0.0000e+00, ..., 1.8454e-04, + -2.7275e-04, -2.5213e-05], + [ 2.5272e-04, 8.9884e-05, 0.0000e+00, ..., 4.8804e-04, + 1.7679e-04, 3.9482e-04], + [-1.1253e-03, -3.7694e-04, 0.0000e+00, ..., 3.2574e-05, + 2.5129e-04, -1.3804e-04]], device='cuda:0') +Epoch 21, bias, value: tensor([-0.0239, 0.0164, 0.0114, 0.0271, 0.0225, -0.0197, 0.0241, 0.0239, + -0.0067, 0.0105], device='cuda:0'), grad: tensor([ 7.7128e-05, -8.2779e-03, 6.5651e-03, 1.2741e-03, 1.1754e-04, + 7.5674e-04, 2.4533e-04, 1.2910e-04, 6.6519e-04, -1.5545e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 20, time 255.97, cls_loss 0.0242 cls_loss_mapping 0.0533 cls_loss_causal 0.8174 re_mapping 0.0234 re_causal 0.0696 /// teacc 98.51 lr 0.00010000 +Epoch 22, weight, value: tensor([[ 0.0298, -0.0549, 0.0022, ..., -0.0316, -0.0075, -0.0102], + [-0.0203, 0.0418, -0.0254, ..., 0.0191, 0.0668, 0.0221], + [-0.0087, -0.0595, -0.0154, ..., 0.0465, -0.0381, -0.0159], + ..., + [-0.0219, 0.0468, 0.0232, ..., -0.0311, 0.0246, 0.0092], + [-0.0224, -0.0324, 0.0127, ..., 0.0026, -0.0714, -0.0183], + [-0.0152, 0.0052, -0.0037, ..., -0.0549, -0.0419, 0.0174]], + device='cuda:0'), grad: tensor([[-2.3580e-04, 2.4676e-05, 0.0000e+00, ..., 9.8228e-05, + 2.0936e-05, 9.3952e-06], + [ 9.0241e-05, -3.7134e-05, 0.0000e+00, ..., 3.2663e-04, + 4.8697e-05, 1.8060e-04], + [-2.7701e-05, 2.1011e-05, 0.0000e+00, ..., -5.9748e-04, + -2.2626e-04, -4.4155e-04], + ..., + [ 1.2898e-04, -4.0859e-05, 0.0000e+00, ..., 1.7858e-04, + -3.4850e-06, 1.5020e-04], + [-4.2886e-05, 1.1373e-04, 0.0000e+00, ..., 2.5511e-04, + 4.6372e-05, 1.4913e-04], + [-1.9813e-04, -2.8133e-04, 0.0000e+00, ..., 3.6061e-05, + 3.7402e-05, -3.1376e-04]], device='cuda:0') +Epoch 22, bias, value: tensor([-0.0235, 0.0164, 0.0112, 0.0270, 0.0225, -0.0199, 0.0241, 0.0242, + -0.0068, 0.0105], device='cuda:0'), grad: tensor([-0.0003, 0.0003, -0.0002, 0.0003, 0.0004, 0.0010, -0.0015, 0.0002, + 0.0003, -0.0005], device='cuda:0') +100 +0.0001 +changing lr +epoch 21, time 256.05, cls_loss 0.0272 cls_loss_mapping 0.0550 cls_loss_causal 0.8081 re_mapping 0.0229 re_causal 0.0669 /// teacc 98.59 lr 0.00010000 +Epoch 23, weight, value: tensor([[ 0.0298, -0.0561, 0.0022, ..., -0.0319, -0.0085, -0.0114], + [-0.0205, 0.0416, -0.0254, ..., 0.0189, 0.0676, 0.0219], + [-0.0089, -0.0600, -0.0154, ..., 0.0472, -0.0388, -0.0152], + ..., + [-0.0221, 0.0475, 0.0232, ..., -0.0316, 0.0262, 0.0095], + [-0.0221, -0.0328, 0.0127, ..., 0.0022, -0.0724, -0.0193], + [-0.0154, 0.0053, -0.0037, ..., -0.0553, -0.0426, 0.0180]], + device='cuda:0'), grad: tensor([[-6.3360e-05, 1.4091e-04, 0.0000e+00, ..., 1.0490e-04, + 1.4079e-04, 6.7890e-05], + [ 3.3975e-04, 7.2908e-04, 0.0000e+00, ..., 1.3804e-04, + 6.1655e-04, 8.2016e-04], + [ 2.2876e-04, 1.9217e-04, 0.0000e+00, ..., -3.0732e-04, + -3.2354e-06, -1.1235e-04], + ..., + [ 6.4583e-03, 1.0223e-02, 0.0000e+00, ..., 1.4138e-04, + 1.0345e-02, 7.7095e-03], + [-5.0316e-03, -1.0239e-02, 0.0000e+00, ..., 1.7023e-03, + -9.9792e-03, -5.3902e-03], + [-1.9417e-03, -2.7962e-03, 0.0000e+00, ..., 3.4541e-05, + -2.4414e-03, -4.2572e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0239, 0.0161, 0.0112, 0.0276, 0.0224, -0.0198, 0.0239, 0.0241, + -0.0067, 0.0107], device='cuda:0'), grad: tensor([ 3.9792e-04, 1.0967e-03, 3.8719e-04, 3.0479e-03, 1.5268e-03, + 4.5896e-05, -9.2697e-03, 1.7502e-02, -9.3994e-03, -5.3329e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 256.41, cls_loss 0.0243 cls_loss_mapping 0.0521 cls_loss_causal 0.8529 re_mapping 0.0224 re_causal 0.0674 /// teacc 98.61 lr 0.00010000 +Epoch 24, weight, value: tensor([[ 0.0299, -0.0566, 0.0022, ..., -0.0325, -0.0092, -0.0122], + [-0.0200, 0.0418, -0.0254, ..., 0.0182, 0.0687, 0.0216], + [-0.0091, -0.0608, -0.0154, ..., 0.0481, -0.0396, -0.0148], + ..., + [-0.0217, 0.0484, 0.0232, ..., -0.0320, 0.0281, 0.0103], + [-0.0218, -0.0330, 0.0127, ..., 0.0021, -0.0733, -0.0199], + [-0.0158, 0.0053, -0.0037, ..., -0.0561, -0.0434, 0.0182]], + device='cuda:0'), grad: tensor([[-8.8739e-04, 9.6500e-05, 0.0000e+00, ..., -2.4843e-04, + 7.9870e-05, 8.9169e-05], + [-1.2589e-03, -1.4925e-03, 0.0000e+00, ..., 1.1273e-05, + -2.8992e-03, -1.7023e-03], + [-1.5411e-03, 4.2272e-04, 0.0000e+00, ..., -1.2503e-03, + 5.2595e-04, -2.2125e-03], + ..., + [ 1.3180e-03, -6.6261e-03, 0.0000e+00, ..., 1.0443e-03, + -1.5354e-03, -3.1528e-03], + [ 6.8207e-03, 3.3116e-04, 0.0000e+00, ..., 8.9788e-04, + 3.4070e-04, 7.9870e-04], + [ 4.4107e-04, 7.9775e-04, 0.0000e+00, ..., 6.9141e-04, + 2.3782e-04, 1.7633e-03]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0242, 0.0160, 0.0113, 0.0274, 0.0223, -0.0198, 0.0241, 0.0244, + -0.0067, 0.0107], device='cuda:0'), grad: tensor([-0.0009, -0.0018, -0.0044, 0.0090, 0.0053, -0.0161, -0.0010, -0.0028, + 0.0098, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 256.70, cls_loss 0.0217 cls_loss_mapping 0.0457 cls_loss_causal 0.8185 re_mapping 0.0220 re_causal 0.0657 /// teacc 98.86 lr 0.00010000 +Epoch 25, weight, value: tensor([[ 0.0305, -0.0561, 0.0022, ..., -0.0323, -0.0087, -0.0112], + [-0.0200, 0.0419, -0.0254, ..., 0.0178, 0.0703, 0.0217], + [-0.0092, -0.0615, -0.0154, ..., 0.0487, -0.0405, -0.0145], + ..., + [-0.0219, 0.0491, 0.0232, ..., -0.0330, 0.0294, 0.0104], + [-0.0219, -0.0333, 0.0127, ..., 0.0017, -0.0747, -0.0204], + [-0.0157, 0.0052, -0.0037, ..., -0.0568, -0.0443, 0.0181]], + device='cuda:0'), grad: tensor([[-1.4365e-04, 3.2812e-05, 0.0000e+00, ..., 1.3083e-05, + 3.3319e-05, 4.2230e-05], + [-2.5019e-05, -1.1313e-04, 0.0000e+00, ..., 5.1737e-05, + -2.1267e-04, -5.1558e-05], + [ 1.8910e-05, 5.9366e-05, 0.0000e+00, ..., -5.3596e-04, + -1.3602e-04, -4.2057e-04], + ..., + [ 2.5010e-04, -1.6596e-06, 0.0000e+00, ..., 3.7646e-04, + 3.2037e-05, 2.0015e-04], + [ 1.2815e-04, 7.2181e-05, 0.0000e+00, ..., 4.2886e-05, + 5.3912e-05, 6.6161e-05], + [ 1.2827e-03, 1.2503e-03, 0.0000e+00, ..., 1.4767e-05, + 7.8201e-05, 2.6488e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0239, 0.0160, 0.0114, 0.0271, 0.0222, -0.0195, 0.0240, 0.0243, + -0.0070, 0.0108], device='cuda:0'), grad: tensor([-2.5654e-04, -5.5790e-05, -5.0735e-04, -2.3899e-03, 1.2577e-04, + -3.1233e-04, 9.2387e-05, 7.1383e-04, 2.7084e-04, 2.3174e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 255.24, cls_loss 0.0192 cls_loss_mapping 0.0400 cls_loss_causal 0.8269 re_mapping 0.0208 re_causal 0.0639 /// teacc 98.83 lr 0.00010000 +Epoch 26, weight, value: tensor([[ 0.0305, -0.0567, 0.0022, ..., -0.0332, -0.0094, -0.0120], + [-0.0198, 0.0422, -0.0254, ..., 0.0172, 0.0712, 0.0218], + [-0.0088, -0.0618, -0.0154, ..., 0.0496, -0.0404, -0.0140], + ..., + [-0.0222, 0.0493, 0.0232, ..., -0.0335, 0.0302, 0.0104], + [-0.0217, -0.0333, 0.0127, ..., 0.0016, -0.0753, -0.0206], + [-0.0156, 0.0054, -0.0037, ..., -0.0575, -0.0451, 0.0184]], + device='cuda:0'), grad: tensor([[-1.5240e-03, 5.3763e-05, 0.0000e+00, ..., -6.9857e-04, + 4.8429e-05, 1.2577e-04], + [ 1.3149e-04, 4.4048e-05, 0.0000e+00, ..., 8.4019e-04, + -2.0945e-04, 5.1260e-04], + [ 1.5414e-04, -3.1531e-05, 0.0000e+00, ..., -4.2295e-04, + -2.6965e-04, -9.0742e-04], + ..., + [ 5.2261e-04, 2.6536e-04, 0.0000e+00, ..., 1.0157e-03, + 3.7861e-04, 1.1711e-03], + [ 9.9277e-04, 9.2387e-05, 0.0000e+00, ..., 5.7077e-04, + 1.3113e-04, 2.2233e-04], + [ 1.2245e-03, 7.2813e-04, 0.0000e+00, ..., 8.8334e-05, + 3.8242e-04, 4.6253e-04]], device='cuda:0') +Epoch 26, bias, value: tensor([-0.0244, 0.0157, 0.0119, 0.0271, 0.0221, -0.0194, 0.0240, 0.0242, + -0.0070, 0.0110], device='cuda:0'), grad: tensor([-0.0021, 0.0011, -0.0003, -0.0048, -0.0020, 0.0014, 0.0008, 0.0019, + 0.0017, 0.0023], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 253.50, cls_loss 0.0167 cls_loss_mapping 0.0411 cls_loss_causal 0.8116 re_mapping 0.0202 re_causal 0.0641 /// teacc 98.69 lr 0.00010000 +Epoch 27, weight, value: tensor([[ 0.0309, -0.0572, 0.0022, ..., -0.0335, -0.0101, -0.0126], + [-0.0194, 0.0425, -0.0259, ..., 0.0169, 0.0726, 0.0222], + [-0.0091, -0.0627, -0.0154, ..., 0.0502, -0.0414, -0.0138], + ..., + [-0.0224, 0.0496, 0.0228, ..., -0.0340, 0.0306, 0.0106], + [-0.0213, -0.0336, 0.0127, ..., 0.0011, -0.0758, -0.0212], + [-0.0158, 0.0057, -0.0038, ..., -0.0587, -0.0451, 0.0186]], + device='cuda:0'), grad: tensor([[ 7.8976e-06, 6.5148e-05, 0.0000e+00, ..., 1.2875e-04, + 1.0145e-04, 6.4552e-05], + [-3.4213e-04, 7.0810e-05, 0.0000e+00, ..., 5.5820e-05, + -4.8923e-04, 9.1553e-05], + [ 5.2691e-04, 3.7575e-04, 0.0000e+00, ..., 1.2493e-04, + 4.4036e-04, 1.7440e-04], + ..., + [ 1.6851e-03, 6.8951e-04, 0.0000e+00, ..., 9.0361e-05, + -8.8453e-04, -4.9293e-05], + [ 2.2721e-04, 1.0306e-04, 0.0000e+00, ..., 2.3937e-04, + 2.6059e-04, 1.0759e-04], + [ 3.6597e-04, 2.0897e-04, 0.0000e+00, ..., 1.4949e-04, + 1.2684e-04, 1.8263e-04]], device='cuda:0') +Epoch 27, bias, value: tensor([-0.0243, 0.0159, 0.0114, 0.0272, 0.0223, -0.0195, 0.0243, 0.0240, + -0.0068, 0.0110], device='cuda:0'), grad: tensor([ 2.8563e-04, -2.6059e-04, 1.5039e-03, -5.3482e-03, -8.0109e-05, + 1.8740e-03, -2.3937e-03, 2.3079e-03, 1.0347e-03, 1.0815e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 249.77, cls_loss 0.0159 cls_loss_mapping 0.0370 cls_loss_causal 0.8277 re_mapping 0.0198 re_causal 0.0610 /// teacc 98.65 lr 0.00010000 +Epoch 28, weight, value: tensor([[ 0.0311, -0.0580, 0.0022, ..., -0.0338, -0.0109, -0.0136], + [-0.0193, 0.0425, -0.0260, ..., 0.0162, 0.0735, 0.0221], + [-0.0091, -0.0633, -0.0154, ..., 0.0509, -0.0418, -0.0134], + ..., + [-0.0224, 0.0505, 0.0226, ..., -0.0344, 0.0319, 0.0110], + [-0.0210, -0.0339, 0.0127, ..., 0.0008, -0.0765, -0.0216], + [-0.0158, 0.0056, -0.0038, ..., -0.0597, -0.0461, 0.0189]], + device='cuda:0'), grad: tensor([[ 5.5544e-06, 2.3291e-05, 0.0000e+00, ..., 4.7982e-05, + 2.5600e-05, 3.2008e-05], + [-6.9082e-05, -5.5492e-05, 0.0000e+00, ..., 1.6093e-05, + -1.9169e-04, -6.9380e-05], + [ 2.7657e-05, 8.4639e-05, 0.0000e+00, ..., -1.1367e-04, + 7.1704e-05, -5.1945e-05], + ..., + [-1.0443e-04, -6.8378e-04, 0.0000e+00, ..., 1.4395e-05, + -3.9792e-04, -4.7040e-04], + [-1.2553e-04, 7.4983e-05, 0.0000e+00, ..., 2.9430e-05, + 7.7128e-05, 1.4949e-04], + [ 1.6975e-04, 3.8600e-04, 0.0000e+00, ..., 9.6679e-05, + 2.6608e-04, 4.2772e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0243, 0.0157, 0.0115, 0.0270, 0.0224, -0.0195, 0.0244, 0.0242, + -0.0069, 0.0110], device='cuda:0'), grad: tensor([-3.1376e-04, -9.6858e-05, 4.8786e-05, -1.9073e-05, -1.1176e-04, + 1.4400e-04, 2.5797e-04, -4.2462e-04, -1.2612e-04, 6.4087e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 247.41, cls_loss 0.0154 cls_loss_mapping 0.0360 cls_loss_causal 0.7687 re_mapping 0.0196 re_causal 0.0587 /// teacc 98.82 lr 0.00010000 +Epoch 29, weight, value: tensor([[ 0.0312, -0.0583, 0.0014, ..., -0.0345, -0.0115, -0.0144], + [-0.0191, 0.0426, -0.0259, ..., 0.0160, 0.0744, 0.0220], + [-0.0091, -0.0641, -0.0168, ..., 0.0518, -0.0423, -0.0128], + ..., + [-0.0225, 0.0512, 0.0259, ..., -0.0352, 0.0331, 0.0113], + [-0.0211, -0.0343, 0.0121, ..., 0.0004, -0.0776, -0.0224], + [-0.0159, 0.0055, -0.0073, ..., -0.0604, -0.0471, 0.0189]], + device='cuda:0'), grad: tensor([[ 2.7466e-04, 1.3351e-04, 0.0000e+00, ..., 3.9101e-04, + 9.4712e-05, 1.1629e-04], + [ 1.6320e-04, 1.1557e-04, 0.0000e+00, ..., 1.5891e-04, + 1.4618e-05, 1.8525e-04], + [-3.2234e-03, 6.8486e-05, 0.0000e+00, ..., -3.7346e-03, + 3.4839e-05, -3.4881e-04], + ..., + [ 6.1941e-04, 3.9697e-04, 0.0000e+00, ..., 4.8137e-04, + -3.2806e-04, 1.2760e-03], + [ 3.6812e-04, 1.7393e-04, 0.0000e+00, ..., 3.4571e-04, + 3.3855e-05, 3.0231e-04], + [ 9.5606e-04, 3.2768e-03, 0.0000e+00, ..., 6.9022e-05, + 2.5845e-04, 5.4398e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0246, 0.0155, 0.0116, 0.0273, 0.0224, -0.0194, 0.0242, 0.0244, + -0.0070, 0.0108], device='cuda:0'), grad: tensor([ 6.7353e-05, 5.7173e-04, -6.8092e-03, 3.0766e-03, -1.0597e-02, + 1.0729e-03, 1.0090e-03, 2.4624e-03, 1.0729e-03, 8.0719e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 28, time 247.28, cls_loss 0.0168 cls_loss_mapping 0.0413 cls_loss_causal 0.7871 re_mapping 0.0185 re_causal 0.0558 /// teacc 98.78 lr 0.00010000 +Epoch 30, weight, value: tensor([[ 0.0315, -0.0589, -0.0098, ..., -0.0346, -0.0124, -0.0150], + [-0.0192, 0.0426, -0.0245, ..., 0.0157, 0.0746, 0.0217], + [-0.0092, -0.0646, -0.0175, ..., 0.0523, -0.0426, -0.0122], + ..., + [-0.0225, 0.0519, 0.0255, ..., -0.0357, 0.0347, 0.0116], + [-0.0209, -0.0346, 0.0044, ..., 0.0001, -0.0781, -0.0229], + [-0.0157, 0.0055, -0.0077, ..., -0.0606, -0.0477, 0.0191]], + device='cuda:0'), grad: tensor([[ 1.7462e-03, 3.8177e-05, 0.0000e+00, ..., 6.0196e-03, + 4.8310e-05, 5.2691e-05], + [-2.6441e-04, -2.2924e-04, 0.0000e+00, ..., -1.0085e-04, + -7.1478e-04, -4.5347e-04], + [ 3.2768e-03, 5.4216e-04, 0.0000e+00, ..., 5.7936e-04, + 3.3808e-04, 5.1928e-04], + ..., + [ 1.3723e-03, 1.5819e-04, 0.0000e+00, ..., 8.0407e-05, + -1.9252e-04, 3.8433e-04], + [ 5.7411e-04, 1.5211e-04, 0.0000e+00, ..., 1.2064e-04, + 6.5923e-05, 2.4986e-04], + [-5.5933e-04, -1.7011e-04, 0.0000e+00, ..., 9.8467e-05, + 9.8765e-05, -4.3941e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0245, 0.0151, 0.0117, 0.0271, 0.0223, -0.0192, 0.0239, 0.0246, + -0.0069, 0.0110], device='cuda:0'), grad: tensor([ 0.0138, -0.0006, 0.0065, -0.0109, 0.0006, 0.0014, -0.0140, 0.0028, + 0.0014, -0.0010], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 247.58, cls_loss 0.0139 cls_loss_mapping 0.0382 cls_loss_causal 0.7461 re_mapping 0.0186 re_causal 0.0561 /// teacc 98.79 lr 0.00010000 +Epoch 31, weight, value: tensor([[ 3.1546e-02, -5.9198e-02, -9.8350e-03, ..., -3.5676e-02, + -1.2900e-02, -1.5569e-02], + [-1.8851e-02, 4.2870e-02, -2.4519e-02, ..., 1.5045e-02, + 7.5752e-02, 2.1819e-02], + [-9.1102e-03, -6.5186e-02, -1.7468e-02, ..., 5.3203e-02, + -4.3015e-02, -1.2029e-02], + ..., + [-2.3044e-02, 5.2554e-02, 2.5503e-02, ..., -3.6110e-02, + 3.5453e-02, 1.1836e-02], + [-2.0862e-02, -3.4773e-02, 4.4109e-03, ..., -9.0113e-05, + -7.9172e-02, -2.3423e-02], + [-1.5367e-02, 5.5597e-03, -7.7215e-03, ..., -6.1657e-02, + -4.7825e-02, 1.9520e-02]], device='cuda:0'), grad: tensor([[ 3.0905e-05, 1.6510e-05, 0.0000e+00, ..., 2.8655e-05, + 2.9221e-05, 1.8746e-05], + [-7.7438e-04, -3.1042e-04, 0.0000e+00, ..., -1.6606e-04, + -1.2741e-03, -4.8828e-04], + [ 2.8157e-04, 2.4527e-05, 0.0000e+00, ..., 3.3170e-05, + 4.3964e-04, 1.7250e-04], + ..., + [ 2.7680e-04, -2.0355e-05, 0.0000e+00, ..., 5.1588e-05, + 5.7407e-06, 4.7833e-05], + [ 5.2500e-04, 9.3877e-05, 0.0000e+00, ..., 5.0485e-05, + 3.4380e-04, 1.2970e-04], + [ 1.7166e-04, 2.2631e-06, 0.0000e+00, ..., 8.7023e-06, + 1.1718e-04, -2.3201e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0247, 0.0151, 0.0118, 0.0273, 0.0223, -0.0194, 0.0237, 0.0245, + -0.0069, 0.0113], device='cuda:0'), grad: tensor([ 9.3043e-05, -1.4610e-03, 4.7636e-04, -7.2336e-04, 1.8036e-04, + -1.5192e-05, -5.0552e-06, 4.1366e-04, 8.4639e-04, 1.9526e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 247.78, cls_loss 0.0126 cls_loss_mapping 0.0319 cls_loss_causal 0.7504 re_mapping 0.0181 re_causal 0.0551 /// teacc 98.70 lr 0.00010000 +Epoch 32, weight, value: tensor([[ 3.1718e-02, -5.9576e-02, -9.8389e-03, ..., -3.6147e-02, + -1.3590e-02, -1.6067e-02], + [-1.8095e-02, 4.3704e-02, -2.4525e-02, ..., 1.4557e-02, + 7.7584e-02, 2.2167e-02], + [-9.1107e-03, -6.5505e-02, -1.7454e-02, ..., 5.3730e-02, + -4.3251e-02, -1.1491e-02], + ..., + [-2.3092e-02, 5.2848e-02, 2.5503e-02, ..., -3.6585e-02, + 3.5784e-02, 1.2140e-02], + [-2.0738e-02, -3.4930e-02, 4.4102e-03, ..., -8.4070e-05, + -8.0119e-02, -2.3934e-02], + [-1.5473e-02, 5.5312e-03, -7.7216e-03, ..., -6.2370e-02, + -4.8808e-02, 1.9445e-02]], device='cuda:0'), grad: tensor([[ 7.0512e-05, 1.4856e-05, 0.0000e+00, ..., 7.3493e-05, + 6.4597e-06, 7.1585e-05], + [ 3.0446e-04, -1.0192e-05, 0.0000e+00, ..., 8.5533e-05, + -6.5982e-05, 1.6046e-04], + [ 1.8388e-05, -1.6165e-04, 0.0000e+00, ..., -1.5712e-04, + 2.8625e-05, -3.7408e-04], + ..., + [ 2.0361e-04, -1.6764e-05, 0.0000e+00, ..., 9.4056e-05, + -6.9976e-05, 1.6570e-04], + [-5.3940e-03, 4.0174e-05, 0.0000e+00, ..., -1.1663e-03, + 2.0891e-05, -2.5253e-03], + [ 3.8223e-03, 2.2277e-05, 0.0000e+00, ..., 8.5306e-04, + 7.7784e-05, 1.8196e-03]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0247, 0.0154, 0.0118, 0.0273, 0.0222, -0.0195, 0.0237, 0.0245, + -0.0068, 0.0112], device='cuda:0'), grad: tensor([ 0.0002, 0.0006, -0.0005, 0.0001, 0.0007, 0.0005, 0.0002, 0.0006, + -0.0087, 0.0063], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 248.79, cls_loss 0.0152 cls_loss_mapping 0.0342 cls_loss_causal 0.7317 re_mapping 0.0168 re_causal 0.0511 /// teacc 98.70 lr 0.00010000 +Epoch 33, weight, value: tensor([[ 0.0318, -0.0598, -0.0098, ..., -0.0366, -0.0139, -0.0168], + [-0.0180, 0.0437, -0.0245, ..., 0.0143, 0.0783, 0.0221], + [-0.0092, -0.0662, -0.0174, ..., 0.0543, -0.0439, -0.0112], + ..., + [-0.0233, 0.0534, 0.0255, ..., -0.0369, 0.0367, 0.0125], + [-0.0206, -0.0348, 0.0044, ..., -0.0006, -0.0801, -0.0244], + [-0.0157, 0.0052, -0.0077, ..., -0.0627, -0.0499, 0.0194]], + device='cuda:0'), grad: tensor([[-5.0962e-05, 2.3022e-05, 0.0000e+00, ..., 2.8506e-05, + 3.2365e-05, 2.8893e-05], + [-1.7524e-04, -2.0099e-04, 0.0000e+00, ..., 8.3327e-05, + -4.9448e-04, -6.8069e-05], + [-2.5177e-04, 1.2541e-04, 0.0000e+00, ..., -6.1893e-04, + 1.2672e-04, -2.6560e-04], + ..., + [-6.8569e-04, -1.2312e-03, 0.0000e+00, ..., 4.6790e-05, + -1.4315e-03, -4.4346e-04], + [ 2.2173e-04, 9.3341e-05, 0.0000e+00, ..., 1.6421e-05, + 1.3304e-04, 1.2791e-04], + [-1.6913e-05, -1.1945e-04, 0.0000e+00, ..., 2.5094e-05, + 5.1290e-05, -2.4819e-04]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0249, 0.0153, 0.0118, 0.0278, 0.0223, -0.0197, 0.0241, 0.0244, + -0.0070, 0.0112], device='cuda:0'), grad: tensor([-0.0001, -0.0002, -0.0004, -0.0154, 0.0004, 0.0178, 0.0002, -0.0022, + 0.0003, -0.0004], device='cuda:0') +100 +0.0001 +changing lr +epoch 32, time 249.15, cls_loss 0.0112 cls_loss_mapping 0.0309 cls_loss_causal 0.7493 re_mapping 0.0163 re_causal 0.0515 /// teacc 98.74 lr 0.00010000 +Epoch 34, weight, value: tensor([[ 0.0321, -0.0602, -0.0098, ..., -0.0371, -0.0140, -0.0174], + [-0.0181, 0.0437, -0.0245, ..., 0.0138, 0.0789, 0.0219], + [-0.0092, -0.0668, -0.0174, ..., 0.0549, -0.0444, -0.0109], + ..., + [-0.0232, 0.0541, 0.0255, ..., -0.0372, 0.0380, 0.0129], + [-0.0202, -0.0351, 0.0044, ..., -0.0008, -0.0810, -0.0249], + [-0.0159, 0.0053, -0.0077, ..., -0.0634, -0.0506, 0.0196]], + device='cuda:0'), grad: tensor([[ 1.6916e-04, 4.4298e-04, 0.0000e+00, ..., 2.5943e-05, + 3.5954e-04, 3.0136e-04], + [-7.8440e-05, -3.1471e-05, 0.0000e+00, ..., 2.6897e-06, + -2.4271e-04, -6.3181e-05], + [ 2.3353e-04, 9.4533e-05, 0.0000e+00, ..., 9.2864e-05, + 1.1200e-04, 1.1164e-04], + ..., + [-2.9802e-04, -9.1219e-04, 0.0000e+00, ..., 9.4101e-06, + -8.3303e-04, -4.6611e-04], + [-3.3450e-04, 8.3447e-05, 0.0000e+00, ..., -1.1092e-04, + 1.0872e-04, -4.0233e-07], + [-1.3840e-04, -1.5378e-04, 0.0000e+00, ..., 3.4064e-05, + 1.5581e-04, -2.8014e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0251, 0.0149, 0.0118, 0.0278, 0.0222, -0.0198, 0.0240, 0.0247, + -0.0068, 0.0111], device='cuda:0'), grad: tensor([ 0.0008, -0.0001, 0.0005, 0.0002, 0.0002, 0.0008, -0.0001, -0.0015, + -0.0004, -0.0003], device='cuda:0') +100 +0.0001 +changing lr +epoch 33, time 248.87, cls_loss 0.0155 cls_loss_mapping 0.0394 cls_loss_causal 0.7420 re_mapping 0.0162 re_causal 0.0492 /// teacc 98.58 lr 0.00010000 +Epoch 35, weight, value: tensor([[ 0.0320, -0.0608, -0.0103, ..., -0.0375, -0.0152, -0.0187], + [-0.0172, 0.0442, -0.0256, ..., 0.0135, 0.0803, 0.0219], + [-0.0087, -0.0674, -0.0158, ..., 0.0555, -0.0446, -0.0102], + ..., + [-0.0238, 0.0546, 0.0255, ..., -0.0377, 0.0388, 0.0133], + [-0.0201, -0.0355, 0.0041, ..., -0.0010, -0.0824, -0.0255], + [-0.0159, 0.0054, -0.0077, ..., -0.0646, -0.0508, 0.0198]], + device='cuda:0'), grad: tensor([[-4.5002e-06, 7.5884e-06, 0.0000e+00, ..., 7.9095e-05, + 7.6219e-06, 1.9222e-05], + [ 1.4313e-05, 3.6091e-05, 0.0000e+00, ..., 2.7701e-05, + -1.2867e-05, 6.2823e-05], + [ 1.5545e-04, 1.3793e-04, 0.0000e+00, ..., 1.6034e-05, + 1.2481e-04, 1.6356e-04], + ..., + [-3.0899e-04, -6.0749e-04, 0.0000e+00, ..., 3.5137e-05, + -5.7268e-04, -5.5647e-04], + [-1.8203e-04, 4.7892e-05, 0.0000e+00, ..., 1.9372e-05, + 4.4167e-05, 9.4354e-05], + [ 1.3912e-04, 4.1962e-05, 0.0000e+00, ..., 2.4235e-04, + 6.1989e-05, 4.9305e-04]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0253, 0.0152, 0.0120, 0.0275, 0.0222, -0.0195, 0.0241, 0.0246, + -0.0069, 0.0111], device='cuda:0'), grad: tensor([ 1.0812e-04, 1.1319e-04, 3.5644e-04, 5.4479e-05, -8.7214e-04, + 9.4557e-04, -4.4370e-04, -8.2779e-04, -1.9372e-04, 7.5769e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 249.60, cls_loss 0.0097 cls_loss_mapping 0.0258 cls_loss_causal 0.7511 re_mapping 0.0161 re_causal 0.0504 /// teacc 99.02 lr 0.00010000 +Epoch 36, weight, value: tensor([[ 0.0322, -0.0612, -0.0166, ..., -0.0380, -0.0160, -0.0191], + [-0.0168, 0.0443, -0.0250, ..., 0.0132, 0.0814, 0.0219], + [-0.0088, -0.0682, -0.0147, ..., 0.0560, -0.0459, -0.0103], + ..., + [-0.0239, 0.0552, 0.0257, ..., -0.0379, 0.0402, 0.0138], + [-0.0202, -0.0357, 0.0015, ..., -0.0016, -0.0834, -0.0262], + [-0.0157, 0.0054, -0.0082, ..., -0.0652, -0.0513, 0.0199]], + device='cuda:0'), grad: tensor([[-1.0186e-04, -8.6054e-06, 0.0000e+00, ..., 1.5587e-05, + 4.7386e-06, 4.3124e-05], + [ 2.8476e-05, 2.2184e-06, 0.0000e+00, ..., 4.2588e-05, + -2.2799e-05, 2.2426e-05], + [ 7.5960e-04, 3.5524e-05, 0.0000e+00, ..., -6.7763e-06, + 2.0087e-05, -2.9874e-04], + ..., + [ 1.0204e-04, -1.3697e-04, 0.0000e+00, ..., 1.0097e-04, + -9.2626e-05, -8.8394e-05], + [ 3.5095e-04, 1.5631e-05, 0.0000e+00, ..., 1.7250e-04, + 1.2755e-05, 3.1680e-05], + [-4.1389e-04, 2.5526e-05, 0.0000e+00, ..., 9.7394e-05, + 4.4197e-05, 1.9088e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0254, 0.0153, 0.0119, 0.0277, 0.0223, -0.0200, 0.0243, 0.0251, + -0.0073, 0.0110], device='cuda:0'), grad: tensor([-0.0002, 0.0001, 0.0007, -0.0024, 0.0004, 0.0004, 0.0007, 0.0001, + 0.0009, -0.0008], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 248.80, cls_loss 0.0090 cls_loss_mapping 0.0245 cls_loss_causal 0.7085 re_mapping 0.0157 re_causal 0.0473 /// teacc 98.98 lr 0.00010000 +Epoch 37, weight, value: tensor([[ 0.0328, -0.0616, -0.0196, ..., -0.0384, -0.0166, -0.0195], + [-0.0165, 0.0444, -0.0247, ..., 0.0131, 0.0824, 0.0220], + [-0.0091, -0.0686, -0.0146, ..., 0.0564, -0.0468, -0.0102], + ..., + [-0.0238, 0.0557, 0.0256, ..., -0.0384, 0.0412, 0.0140], + [-0.0201, -0.0359, 0.0023, ..., -0.0016, -0.0843, -0.0265], + [-0.0157, 0.0053, -0.0077, ..., -0.0658, -0.0516, 0.0200]], + device='cuda:0'), grad: tensor([[-4.3124e-05, 6.4634e-06, 7.6776e-08, ..., 4.7088e-05, + 7.8157e-06, 4.3929e-05], + [ 4.2409e-05, 2.2370e-06, -3.5707e-06, ..., 1.0484e-04, + -2.9281e-05, 6.7949e-05], + [-2.3293e-04, 3.0667e-05, 4.7777e-07, ..., -7.9930e-05, + 6.8103e-08, -1.3828e-04], + ..., + [ 4.2349e-05, -1.1945e-04, 8.8569e-07, ..., 9.3937e-05, + -8.3208e-05, -3.5584e-05], + [ 5.0992e-05, 1.8433e-05, 8.0001e-07, ..., 1.4746e-04, + 2.1875e-05, 1.1462e-04], + [ 1.1897e-04, 7.0512e-05, 3.4156e-07, ..., 1.3971e-04, + 3.1799e-05, 2.6870e-04]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0248, 0.0154, 0.0115, 0.0279, 0.0225, -0.0201, 0.0241, 0.0250, + -0.0072, 0.0109], device='cuda:0'), grad: tensor([-8.9347e-05, 1.6475e-04, -3.8815e-04, 7.8559e-05, -4.1842e-04, + -1.8167e-04, -5.5790e-05, 5.0932e-05, 2.5868e-04, 5.8079e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 249.11, cls_loss 0.0085 cls_loss_mapping 0.0272 cls_loss_causal 0.7146 re_mapping 0.0147 re_causal 0.0463 /// teacc 98.91 lr 0.00010000 +Epoch 38, weight, value: tensor([[ 0.0329, -0.0620, -0.0215, ..., -0.0388, -0.0173, -0.0201], + [-0.0160, 0.0445, -0.0238, ..., 0.0129, 0.0835, 0.0222], + [-0.0094, -0.0690, -0.0140, ..., 0.0569, -0.0477, -0.0101], + ..., + [-0.0238, 0.0563, 0.0253, ..., -0.0387, 0.0424, 0.0144], + [-0.0199, -0.0362, 0.0020, ..., -0.0017, -0.0852, -0.0270], + [-0.0158, 0.0053, -0.0093, ..., -0.0665, -0.0523, 0.0201]], + device='cuda:0'), grad: tensor([[ 6.5379e-06, 1.7941e-05, 4.3237e-07, ..., 2.3767e-05, + 9.2015e-06, 3.5107e-05], + [-8.4341e-06, -3.2187e-06, -1.1260e-06, ..., 4.5270e-05, + -2.8536e-05, 5.1022e-05], + [ 8.6129e-06, 3.6597e-05, -4.2394e-06, ..., 2.0885e-04, + 1.4402e-05, 3.0899e-04], + ..., + [-5.1744e-06, -8.3685e-05, 1.3383e-06, ..., 6.6578e-05, + -7.1526e-05, 4.9233e-05], + [ 3.0428e-05, 2.5302e-05, 1.3048e-06, ..., 2.8983e-05, + 1.4842e-05, 5.2899e-05], + [-8.7500e-05, -3.9071e-05, 3.4133e-07, ..., 1.2672e-04, + 1.5661e-05, 3.3826e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0249, 0.0156, 0.0113, 0.0278, 0.0223, -0.0201, 0.0240, 0.0253, + -0.0071, 0.0109], device='cuda:0'), grad: tensor([ 6.2346e-05, 6.6400e-05, 4.9496e-04, 1.4293e-04, -1.1377e-03, + 2.8625e-05, 1.1355e-04, 9.8765e-05, 1.5557e-04, -2.5541e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 37, time 248.86, cls_loss 0.0106 cls_loss_mapping 0.0302 cls_loss_causal 0.7267 re_mapping 0.0149 re_causal 0.0465 /// teacc 98.89 lr 0.00010000 +Epoch 39, weight, value: tensor([[ 0.0330, -0.0625, -0.0239, ..., -0.0394, -0.0177, -0.0207], + [-0.0166, 0.0443, -0.0239, ..., 0.0122, 0.0835, 0.0218], + [-0.0087, -0.0693, -0.0129, ..., 0.0578, -0.0477, -0.0095], + ..., + [-0.0239, 0.0567, 0.0245, ..., -0.0389, 0.0437, 0.0146], + [-0.0198, -0.0364, 0.0021, ..., -0.0021, -0.0859, -0.0274], + [-0.0157, 0.0057, -0.0094, ..., -0.0670, -0.0528, 0.0205]], + device='cuda:0'), grad: tensor([[ 3.6657e-05, 3.8929e-06, 0.0000e+00, ..., 4.0889e-05, + 2.0623e-05, 1.8343e-05], + [-2.4092e-04, -1.2994e-04, 0.0000e+00, ..., 1.5393e-05, + -6.1893e-04, -1.5354e-04], + [-1.3602e-04, 5.4091e-06, 0.0000e+00, ..., -1.4424e-04, + 5.9485e-05, -6.8486e-05], + ..., + [ 6.2346e-05, 1.4924e-05, 0.0000e+00, ..., 1.2800e-05, + 6.2406e-05, 3.2216e-05], + [ 1.6153e-04, 2.6777e-05, 0.0000e+00, ..., 9.2745e-05, + 1.5235e-04, 7.4923e-05], + [ 1.0145e-04, 1.1995e-05, 0.0000e+00, ..., 1.3418e-05, + 5.0157e-05, 2.5779e-06]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0253, 0.0149, 0.0122, 0.0279, 0.0222, -0.0200, 0.0239, 0.0254, + -0.0071, 0.0109], device='cuda:0'), grad: tensor([ 0.0001, -0.0007, -0.0002, -0.0002, 0.0002, 0.0001, -0.0001, 0.0001, + 0.0004, 0.0002], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 248.74, cls_loss 0.0131 cls_loss_mapping 0.0343 cls_loss_causal 0.7485 re_mapping 0.0148 re_causal 0.0471 /// teacc 98.83 lr 0.00010000 +Epoch 40, weight, value: tensor([[ 0.0329, -0.0628, -0.0274, ..., -0.0399, -0.0180, -0.0222], + [-0.0163, 0.0446, -0.0242, ..., 0.0120, 0.0848, 0.0222], + [-0.0088, -0.0697, -0.0128, ..., 0.0584, -0.0482, -0.0092], + ..., + [-0.0243, 0.0572, 0.0269, ..., -0.0397, 0.0442, 0.0148], + [-0.0198, -0.0367, 0.0034, ..., -0.0022, -0.0864, -0.0281], + [-0.0153, 0.0056, -0.0114, ..., -0.0676, -0.0538, 0.0206]], + device='cuda:0'), grad: tensor([[-2.6561e-06, 1.4856e-05, 0.0000e+00, ..., 8.1778e-05, + -1.9178e-05, 4.0203e-05], + [ 2.1026e-05, 2.1085e-05, 0.0000e+00, ..., 2.2709e-05, + -3.4869e-05, 3.1739e-05], + [ 2.7990e-04, 2.5535e-04, 0.0000e+00, ..., -6.4611e-05, + 8.0585e-05, 1.9240e-04], + ..., + [-6.9523e-04, -8.5878e-04, 0.0000e+00, ..., 2.3976e-05, + -1.9169e-04, -8.1587e-04], + [ 1.8597e-05, 2.4050e-05, 0.0000e+00, ..., 4.2826e-05, + 1.3649e-05, 5.5760e-05], + [ 1.5461e-04, 9.6500e-05, 0.0000e+00, ..., 4.5240e-05, + 4.3571e-05, 2.6083e-04]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0254, 0.0150, 0.0122, 0.0280, 0.0224, -0.0198, 0.0236, 0.0252, + -0.0074, 0.0109], device='cuda:0'), grad: tensor([ 5.7250e-05, 8.2672e-05, 4.6897e-04, 1.3745e-04, 2.2268e-04, + 3.8266e-05, -5.6696e-04, -1.2217e-03, 1.2767e-04, 6.5279e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 248.90, cls_loss 0.0096 cls_loss_mapping 0.0274 cls_loss_causal 0.6816 re_mapping 0.0152 re_causal 0.0453 /// teacc 98.81 lr 0.00010000 +Epoch 41, weight, value: tensor([[ 0.0329, -0.0632, -0.0278, ..., -0.0403, -0.0186, -0.0231], + [-0.0159, 0.0449, -0.0243, ..., 0.0118, 0.0863, 0.0224], + [-0.0090, -0.0703, -0.0128, ..., 0.0587, -0.0490, -0.0095], + ..., + [-0.0241, 0.0578, 0.0269, ..., -0.0400, 0.0449, 0.0150], + [-0.0194, -0.0370, 0.0037, ..., -0.0023, -0.0872, -0.0280], + [-0.0155, 0.0057, -0.0109, ..., -0.0689, -0.0548, 0.0207]], + device='cuda:0'), grad: tensor([[-7.5817e-05, 5.1916e-05, 0.0000e+00, ..., -2.5302e-05, + 3.0577e-05, 7.3671e-05], + [ 4.8709e-04, 1.6346e-03, 0.0000e+00, ..., 3.1735e-07, + 2.2564e-03, 2.2373e-03], + [ 6.5744e-05, 5.5742e-04, 0.0000e+00, ..., 9.5814e-06, + 3.8099e-04, 6.9618e-04], + ..., + [-5.0402e-04, -1.1223e-02, 0.0000e+00, ..., 3.1777e-06, + -8.7662e-03, -1.4008e-02], + [ 4.7565e-05, 8.0764e-05, 0.0000e+00, ..., 4.5039e-06, + 5.1558e-05, 1.2803e-04], + [-5.2357e-04, -8.3256e-04, 0.0000e+00, ..., 4.8317e-06, + 1.3256e-04, -2.2945e-03]], device='cuda:0') +Epoch 41, bias, value: tensor([-0.0255, 0.0152, 0.0118, 0.0279, 0.0228, -0.0201, 0.0239, 0.0253, + -0.0070, 0.0107], device='cuda:0'), grad: tensor([-0.0001, 0.0023, 0.0008, -0.0002, 0.0132, 0.0002, 0.0001, -0.0132, + 0.0002, -0.0033], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 248.93, cls_loss 0.0080 cls_loss_mapping 0.0224 cls_loss_causal 0.6944 re_mapping 0.0145 re_causal 0.0449 /// teacc 98.99 lr 0.00010000 +Epoch 42, weight, value: tensor([[ 0.0331, -0.0637, -0.0280, ..., -0.0407, -0.0191, -0.0238], + [-0.0161, 0.0443, -0.0243, ..., 0.0115, 0.0866, 0.0217], + [-0.0090, -0.0710, -0.0128, ..., 0.0593, -0.0497, -0.0090], + ..., + [-0.0241, 0.0589, 0.0270, ..., -0.0405, 0.0466, 0.0160], + [-0.0193, -0.0373, 0.0037, ..., -0.0025, -0.0882, -0.0286], + [-0.0154, 0.0056, -0.0109, ..., -0.0696, -0.0555, 0.0208]], + device='cuda:0'), grad: tensor([[-2.2963e-05, 5.8562e-06, 2.4665e-09, ..., 2.8461e-05, + 7.6815e-06, 1.4208e-05], + [ 5.1111e-06, -2.1741e-05, 2.2337e-08, ..., 1.6809e-05, + -3.5226e-05, 3.6322e-06], + [-3.1918e-05, 2.0280e-05, -1.0943e-07, ..., -1.0192e-04, + -1.0841e-05, -6.5565e-05], + ..., + [ 6.1691e-06, -7.2718e-05, 3.2742e-08, ..., 4.3243e-05, + -5.5134e-05, -4.1246e-05], + [ 5.1111e-05, 1.6078e-05, 1.6458e-08, ..., 5.9664e-05, + 2.1666e-05, 3.0279e-05], + [ 2.6155e-04, -5.6416e-05, 2.1464e-10, ..., 7.9498e-06, + 2.3872e-05, -1.1361e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0254, 0.0146, 0.0116, 0.0280, 0.0227, -0.0201, 0.0241, 0.0261, + -0.0073, 0.0107], device='cuda:0'), grad: tensor([-4.4741e-06, 1.4700e-05, -8.5533e-05, -4.6164e-05, 1.0657e-04, + -7.2384e-04, -2.1815e-05, -6.2250e-06, 1.8454e-04, 5.8222e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 248.84, cls_loss 0.0085 cls_loss_mapping 0.0235 cls_loss_causal 0.6933 re_mapping 0.0136 re_causal 0.0421 /// teacc 98.84 lr 0.00010000 +Epoch 43, weight, value: tensor([[ 0.0333, -0.0640, -0.0285, ..., -0.0414, -0.0196, -0.0244], + [-0.0150, 0.0450, -0.0244, ..., 0.0112, 0.0882, 0.0224], + [-0.0086, -0.0716, -0.0127, ..., 0.0602, -0.0503, -0.0089], + ..., + [-0.0245, 0.0590, 0.0272, ..., -0.0409, 0.0468, 0.0160], + [-0.0195, -0.0374, 0.0040, ..., -0.0030, -0.0888, -0.0289], + [-0.0155, 0.0057, -0.0109, ..., -0.0700, -0.0563, 0.0213]], + device='cuda:0'), grad: tensor([[ 1.6659e-05, 1.0528e-05, 0.0000e+00, ..., 2.0415e-05, + 5.6624e-06, 2.5570e-05], + [ 1.9640e-05, -1.8716e-05, 0.0000e+00, ..., 5.0604e-05, + -3.2574e-05, 3.7402e-05], + [-8.9183e-06, 8.4192e-06, 0.0000e+00, ..., -2.3925e-04, + -1.7118e-04, -2.4104e-04], + ..., + [ 1.5938e-04, 6.6340e-05, 0.0000e+00, ..., 1.7440e-04, + 1.4663e-04, 4.0174e-04], + [ 7.1406e-05, 5.8264e-05, 0.0000e+00, ..., 4.2766e-05, + 7.8976e-06, 1.1730e-04], + [-2.8944e-04, -2.5249e-04, 0.0000e+00, ..., 4.0799e-05, + 2.2277e-06, -5.8317e-04]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0256, 0.0151, 0.0123, 0.0278, 0.0222, -0.0202, 0.0241, 0.0258, + -0.0076, 0.0108], device='cuda:0'), grad: tensor([ 7.3612e-05, 9.9659e-05, -2.2221e-04, -6.2904e-03, 2.3079e-04, + 6.1722e-03, -9.5367e-05, 6.7806e-04, 3.8648e-04, -1.0338e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 248.72, cls_loss 0.0066 cls_loss_mapping 0.0202 cls_loss_causal 0.6736 re_mapping 0.0132 re_causal 0.0412 /// teacc 98.89 lr 0.00010000 +Epoch 44, weight, value: tensor([[ 0.0333, -0.0643, -0.0290, ..., -0.0416, -0.0202, -0.0247], + [-0.0147, 0.0450, -0.0245, ..., 0.0108, 0.0890, 0.0225], + [-0.0085, -0.0719, -0.0127, ..., 0.0609, -0.0508, -0.0085], + ..., + [-0.0246, 0.0595, 0.0273, ..., -0.0417, 0.0475, 0.0161], + [-0.0192, -0.0376, 0.0044, ..., -0.0032, -0.0891, -0.0294], + [-0.0154, 0.0056, -0.0109, ..., -0.0705, -0.0570, 0.0212]], + device='cuda:0'), grad: tensor([[ 2.1189e-05, 1.4514e-05, 6.9791e-08, ..., 6.1989e-06, + 6.7316e-06, 3.0965e-05], + [-2.0897e-04, -9.9391e-06, 7.9599e-09, ..., -2.8038e-04, + -1.7452e-03, -1.1683e-03], + [ 1.9658e-04, 8.9481e-06, 1.2224e-08, ..., 2.5320e-04, + 1.5135e-03, 1.0443e-03], + ..., + [ 6.8128e-05, 8.0988e-06, 2.9150e-07, ..., 2.5272e-05, + 8.4460e-05, 1.2070e-04], + [-6.2846e-06, 1.1362e-05, 1.2247e-07, ..., -2.3935e-06, + 1.9118e-05, 1.6257e-05], + [-1.8394e-04, -1.3185e-04, -1.1837e-06, ..., 7.1704e-05, + 1.6689e-05, -9.2149e-05]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0257, 0.0149, 0.0124, 0.0278, 0.0226, -0.0204, 0.0242, 0.0257, + -0.0074, 0.0106], device='cuda:0'), grad: tensor([ 5.2273e-05, -1.5163e-03, 1.3809e-03, 1.7130e-04, -1.0169e-04, + -8.8871e-05, 1.2863e-04, 1.9264e-04, 8.7097e-06, -2.2709e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 249.97, cls_loss 0.0093 cls_loss_mapping 0.0249 cls_loss_causal 0.6863 re_mapping 0.0128 re_causal 0.0403 /// teacc 98.76 lr 0.00010000 +Epoch 45, weight, value: tensor([[ 0.0336, -0.0647, -0.0354, ..., -0.0422, -0.0205, -0.0253], + [-0.0145, 0.0447, -0.0245, ..., 0.0105, 0.0897, 0.0224], + [-0.0084, -0.0722, -0.0133, ..., 0.0613, -0.0513, -0.0082], + ..., + [-0.0246, 0.0604, 0.0306, ..., -0.0417, 0.0481, 0.0165], + [-0.0192, -0.0379, 0.0043, ..., -0.0031, -0.0900, -0.0300], + [-0.0152, 0.0054, -0.0115, ..., -0.0714, -0.0572, 0.0214]], + device='cuda:0'), grad: tensor([[-5.6553e-04, 9.9316e-06, 2.4796e-07, ..., -1.8139e-03, + 1.6168e-05, 2.1413e-05], + [-1.0949e-04, -3.7134e-05, 4.0955e-07, ..., 1.1109e-05, + -2.3782e-04, -6.5267e-05], + [ 3.0375e-04, 3.7163e-05, 2.1793e-07, ..., 3.7932e-04, + 9.5308e-05, 8.5771e-05], + ..., + [ 1.3280e-04, 5.1409e-05, 2.2035e-06, ..., 8.2910e-05, + -1.7643e-05, 1.0967e-04], + [-8.1730e-04, -7.7724e-05, 3.5018e-06, ..., -4.4554e-05, + 4.4942e-05, -2.2709e-04], + [ 3.6597e-04, -1.1134e-04, -1.3061e-05, ..., 6.5947e-04, + 2.2054e-05, -6.5625e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0257, 0.0147, 0.0126, 0.0276, 0.0225, -0.0199, 0.0238, 0.0261, + -0.0074, 0.0106], device='cuda:0'), grad: tensor([-0.0046, -0.0001, 0.0012, 0.0005, 0.0001, 0.0008, 0.0013, 0.0004, + -0.0014, 0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 249.81, cls_loss 0.0085 cls_loss_mapping 0.0218 cls_loss_causal 0.7006 re_mapping 0.0134 re_causal 0.0423 /// teacc 98.91 lr 0.00010000 +Epoch 46, weight, value: tensor([[ 0.0338, -0.0651, -0.0388, ..., -0.0426, -0.0213, -0.0260], + [-0.0140, 0.0446, -0.0244, ..., 0.0103, 0.0905, 0.0225], + [-0.0078, -0.0729, -0.0183, ..., 0.0623, -0.0522, -0.0075], + ..., + [-0.0248, 0.0610, 0.0308, ..., -0.0421, 0.0489, 0.0167], + [-0.0195, -0.0379, 0.0103, ..., -0.0036, -0.0908, -0.0308], + [-0.0155, 0.0055, -0.0104, ..., -0.0723, -0.0577, 0.0215]], + device='cuda:0'), grad: tensor([[ 2.5678e-04, 9.9614e-06, 0.0000e+00, ..., 2.8968e-04, + 1.2629e-05, 7.6741e-06], + [ 1.2003e-05, 1.8859e-04, 0.0000e+00, ..., 3.5949e-06, + 1.1635e-04, 6.9797e-05], + [ 3.0637e-05, 4.0740e-05, 0.0000e+00, ..., 5.2564e-06, + 4.8637e-05, 2.3484e-05], + ..., + [-1.3220e-04, -6.2609e-04, 0.0000e+00, ..., 6.0443e-07, + -5.8699e-04, -2.9230e-04], + [-7.4685e-05, 1.8165e-05, 0.0000e+00, ..., 6.4433e-05, + 2.3216e-05, 1.6034e-05], + [ 2.4900e-05, 2.2590e-05, 0.0000e+00, ..., 6.5193e-06, + 3.7163e-05, -2.9970e-06]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0257, 0.0149, 0.0133, 0.0276, 0.0221, -0.0196, 0.0236, 0.0258, + -0.0078, 0.0105], device='cuda:0'), grad: tensor([ 1.0948e-03, 1.3721e-04, 8.9526e-05, 3.4618e-04, 7.3314e-05, + 8.0872e-04, -2.0466e-03, -5.9986e-04, 6.4015e-05, 3.3468e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 249.62, cls_loss 0.0083 cls_loss_mapping 0.0227 cls_loss_causal 0.6700 re_mapping 0.0127 re_causal 0.0396 /// teacc 98.88 lr 0.00010000 +Epoch 47, weight, value: tensor([[ 0.0339, -0.0654, -0.0401, ..., -0.0430, -0.0217, -0.0265], + [-0.0142, 0.0445, -0.0246, ..., 0.0097, 0.0905, 0.0220], + [-0.0080, -0.0731, -0.0182, ..., 0.0628, -0.0530, -0.0074], + ..., + [-0.0248, 0.0615, 0.0311, ..., -0.0425, 0.0503, 0.0175], + [-0.0193, -0.0381, 0.0103, ..., -0.0036, -0.0913, -0.0312], + [-0.0151, 0.0056, -0.0105, ..., -0.0725, -0.0579, 0.0217]], + device='cuda:0'), grad: tensor([[ 4.2105e-04, 1.3135e-05, 5.4017e-08, ..., 3.2973e-04, + -3.2753e-05, 2.5797e-04], + [ 3.9250e-05, -1.1474e-06, 8.3965e-09, ..., 3.2067e-05, + -2.4766e-05, 5.3763e-05], + [-1.0929e-03, 1.0751e-05, -3.0937e-08, ..., -8.1205e-04, + 1.7822e-05, -9.1505e-04], + ..., + [ 1.3864e-04, 1.1927e-04, 3.6648e-07, ..., 7.6652e-05, + 2.9013e-05, 2.6798e-04], + [ 1.3185e-04, -3.5733e-05, 1.3225e-07, ..., 2.1601e-04, + -4.3839e-05, 1.1671e-04], + [ 1.4234e-04, -2.3162e-04, -1.3318e-06, ..., 8.7440e-05, + 2.5392e-05, -3.0026e-05]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0260, 0.0142, 0.0130, 0.0276, 0.0219, -0.0197, 0.0240, 0.0264, + -0.0078, 0.0109], device='cuda:0'), grad: tensor([ 1.6537e-03, 1.4317e-04, -3.6983e-03, 3.6502e-04, 1.5222e-05, + 5.7936e-04, -4.7803e-04, 6.2227e-04, 8.6975e-04, -7.7188e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 249.88, cls_loss 0.0074 cls_loss_mapping 0.0211 cls_loss_causal 0.6772 re_mapping 0.0121 re_causal 0.0385 /// teacc 98.90 lr 0.00010000 +Epoch 48, weight, value: tensor([[ 0.0340, -0.0657, -0.0467, ..., -0.0437, -0.0219, -0.0270], + [-0.0136, 0.0451, -0.0247, ..., 0.0094, 0.0917, 0.0224], + [-0.0082, -0.0737, -0.0196, ..., 0.0632, -0.0537, -0.0074], + ..., + [-0.0252, 0.0617, 0.0336, ..., -0.0429, 0.0506, 0.0174], + [-0.0189, -0.0383, 0.0096, ..., -0.0040, -0.0920, -0.0317], + [-0.0153, 0.0058, -0.0105, ..., -0.0728, -0.0583, 0.0225]], + device='cuda:0'), grad: tensor([[-3.8594e-06, 7.8231e-06, 1.5106e-06, ..., 1.0347e-04, + -8.1137e-06, 2.0966e-05], + [-1.4827e-05, -9.3937e-05, -7.6145e-06, ..., 2.0540e-04, + -2.2447e-04, 1.6391e-05], + [-4.1306e-05, 2.7820e-05, -1.0051e-05, ..., -2.0742e-04, + 3.6746e-05, -1.4353e-04], + ..., + [ 4.8101e-05, -7.5877e-05, 5.8673e-06, ..., 2.7582e-05, + -2.9132e-05, -7.5221e-05], + [ 7.0691e-05, 1.4044e-05, 2.7642e-06, ..., 9.9123e-05, + 3.5077e-05, 3.0786e-05], + [ 1.0109e-04, 5.2482e-05, 1.5050e-06, ..., 1.8030e-05, + 6.4433e-05, 5.8860e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0263, 0.0144, 0.0127, 0.0277, 0.0217, -0.0198, 0.0244, 0.0262, + -0.0076, 0.0110], device='cuda:0'), grad: tensor([ 1.8644e-04, 2.8086e-04, -1.3959e-04, -1.3000e-02, 2.4700e-04, + 1.2833e-02, -1.0290e-03, 6.0769e-07, 3.4070e-04, 2.7370e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 47, time 249.88, cls_loss 0.0073 cls_loss_mapping 0.0203 cls_loss_causal 0.7000 re_mapping 0.0121 re_causal 0.0375 /// teacc 98.93 lr 0.00010000 +Epoch 49, weight, value: tensor([[ 0.0342, -0.0661, -0.0525, ..., -0.0441, -0.0225, -0.0275], + [-0.0135, 0.0451, -0.0272, ..., 0.0085, 0.0922, 0.0220], + [-0.0083, -0.0742, -0.0176, ..., 0.0640, -0.0540, -0.0068], + ..., + [-0.0254, 0.0623, 0.0338, ..., -0.0434, 0.0515, 0.0177], + [-0.0185, -0.0384, 0.0095, ..., -0.0044, -0.0926, -0.0320], + [-0.0156, 0.0056, -0.0106, ..., -0.0736, -0.0591, 0.0225]], + device='cuda:0'), grad: tensor([[-1.1480e-04, 6.3442e-06, 9.7509e-07, ..., 9.0778e-05, + 5.9605e-06, 1.5080e-05], + [ 8.8632e-05, -2.6509e-05, -3.3360e-06, ..., 1.3582e-05, + -4.8280e-05, -1.4096e-05], + [ 1.1605e-04, 5.1111e-06, 8.2562e-07, ..., 2.1338e-05, + 1.3649e-05, -1.7077e-05], + ..., + [ 6.8545e-05, 2.2203e-06, 7.5251e-06, ..., 3.3617e-05, + -6.8657e-06, 5.4955e-05], + [-2.7657e-04, 2.6807e-05, 4.0419e-06, ..., 2.4259e-05, + -4.9382e-05, 2.9877e-05], + [ 8.6962e-08, -7.0572e-05, -2.0623e-05, ..., 1.4059e-05, + 1.2487e-05, -1.1039e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0263, 0.0141, 0.0127, 0.0280, 0.0217, -0.0195, 0.0242, 0.0264, + -0.0074, 0.0106], device='cuda:0'), grad: tensor([-7.5758e-05, 2.7442e-04, 2.9206e-04, -6.9141e-04, 6.9857e-05, + 5.3215e-04, 2.6196e-05, 2.2709e-04, -5.6505e-04, -8.9467e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 250.12, cls_loss 0.0063 cls_loss_mapping 0.0172 cls_loss_causal 0.6548 re_mapping 0.0121 re_causal 0.0386 /// teacc 98.92 lr 0.00010000 +Epoch 50, weight, value: tensor([[ 0.0344, -0.0664, -0.0557, ..., -0.0443, -0.0230, -0.0281], + [-0.0132, 0.0451, -0.0271, ..., 0.0082, 0.0927, 0.0219], + [-0.0086, -0.0754, -0.0179, ..., 0.0644, -0.0546, -0.0068], + ..., + [-0.0253, 0.0631, 0.0336, ..., -0.0434, 0.0520, 0.0183], + [-0.0183, -0.0384, 0.0101, ..., -0.0048, -0.0930, -0.0326], + [-0.0157, 0.0056, -0.0097, ..., -0.0741, -0.0592, 0.0225]], + device='cuda:0'), grad: tensor([[-1.7726e-04, 1.8170e-06, 0.0000e+00, ..., 8.5115e-05, + 3.3695e-06, 3.8028e-05], + [-3.1620e-05, -4.2707e-05, 0.0000e+00, ..., 1.2495e-05, + -1.0383e-04, -3.2902e-05], + [-2.5854e-05, 5.0887e-06, 0.0000e+00, ..., -1.5688e-03, + 9.4101e-06, -6.5899e-04], + ..., + [ 2.8074e-05, 7.5921e-06, 0.0000e+00, ..., 3.8475e-05, + 2.9907e-05, 2.4393e-05], + [ 3.0667e-05, 5.1670e-06, 0.0000e+00, ..., 4.8578e-05, + 1.1057e-05, 2.6897e-05], + [ 2.0415e-05, 5.8748e-06, 0.0000e+00, ..., 2.2814e-05, + 1.4268e-05, 7.1563e-06]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0263, 0.0138, 0.0124, 0.0282, 0.0221, -0.0197, 0.0242, 0.0268, + -0.0074, 0.0104], device='cuda:0'), grad: tensor([-4.0317e-04, -6.8426e-05, -1.7834e-03, 1.4389e-04, 1.1044e-03, + 1.0557e-05, 7.0953e-04, 1.0186e-04, 1.1367e-04, 7.1287e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 250.53, cls_loss 0.0066 cls_loss_mapping 0.0189 cls_loss_causal 0.6700 re_mapping 0.0123 re_causal 0.0375 /// teacc 98.89 lr 0.00010000 +Epoch 51, weight, value: tensor([[ 0.0344, -0.0666, -0.0588, ..., -0.0439, -0.0238, -0.0285], + [-0.0130, 0.0453, -0.0273, ..., 0.0079, 0.0933, 0.0219], + [-0.0087, -0.0760, -0.0186, ..., 0.0649, -0.0552, -0.0066], + ..., + [-0.0254, 0.0635, 0.0337, ..., -0.0435, 0.0528, 0.0187], + [-0.0179, -0.0386, 0.0104, ..., -0.0048, -0.0936, -0.0329], + [-0.0157, 0.0055, -0.0100, ..., -0.0751, -0.0599, 0.0224]], + device='cuda:0'), grad: tensor([[-9.2149e-05, 2.2009e-05, 1.2577e-05, ..., 2.1413e-05, + 2.1756e-05, 3.7134e-05], + [-3.3951e-04, -1.9884e-04, -3.0417e-06, ..., -6.0797e-05, + -5.5122e-04, -1.7107e-04], + [-2.1255e-04, -1.9169e-04, -2.5892e-04, ..., -4.9305e-04, + 1.2064e-04, -4.4799e-04], + ..., + [ 2.9325e-04, 1.4281e-04, 1.2624e-04, ..., 2.6703e-04, + 1.0401e-04, 2.8610e-04], + [ 2.7156e-04, 1.5140e-04, 9.3102e-05, ..., 1.9765e-04, + 1.3888e-04, 2.6155e-04], + [-5.2720e-05, -3.8147e-05, -1.2964e-05, ..., 9.3058e-06, + 1.6510e-05, -9.1910e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([-0.0260, 0.0137, 0.0121, 0.0279, 0.0221, -0.0196, 0.0239, 0.0271, + -0.0070, 0.0101], device='cuda:0'), grad: tensor([-2.3711e-04, -8.8596e-04, -7.6008e-04, 6.0380e-05, 8.3923e-05, + 5.4121e-05, 1.4687e-04, 8.4543e-04, 8.1444e-04, -1.2314e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 250.35, cls_loss 0.0067 cls_loss_mapping 0.0209 cls_loss_causal 0.6814 re_mapping 0.0115 re_causal 0.0359 /// teacc 98.94 lr 0.00010000 +Epoch 52, weight, value: tensor([[ 0.0347, -0.0669, -0.0623, ..., -0.0433, -0.0239, -0.0294], + [-0.0122, 0.0460, -0.0247, ..., 0.0076, 0.0944, 0.0221], + [-0.0087, -0.0764, -0.0188, ..., 0.0655, -0.0558, -0.0060], + ..., + [-0.0264, 0.0635, 0.0293, ..., -0.0446, 0.0530, 0.0185], + [-0.0178, -0.0391, 0.0116, ..., -0.0050, -0.0946, -0.0338], + [-0.0153, 0.0058, -0.0090, ..., -0.0759, -0.0605, 0.0229]], + device='cuda:0'), grad: tensor([[-2.7373e-05, -2.7761e-05, 4.5402e-07, ..., 4.6045e-05, + -2.1517e-05, 3.9071e-05], + [ 2.9549e-05, -2.4721e-05, -3.2270e-07, ..., 3.9458e-04, + -7.7903e-05, 2.4652e-04], + [-1.2052e-04, 6.4410e-06, -3.0175e-07, ..., -1.3256e-03, + -4.5188e-06, -9.5510e-04], + ..., + [ 4.9800e-05, 1.2435e-05, 3.1805e-07, ..., 3.9428e-05, + 3.9995e-05, 4.8578e-05], + [ 6.8843e-05, 7.6592e-06, 7.2503e-07, ..., 7.8857e-05, + 2.1413e-05, 7.1645e-05], + [ 4.1485e-05, 7.7561e-06, 2.7847e-07, ..., 4.0680e-05, + 1.4536e-05, 1.0133e-04]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0259, 0.0141, 0.0121, 0.0278, 0.0219, -0.0193, 0.0245, 0.0263, + -0.0074, 0.0106], device='cuda:0'), grad: tensor([-2.7150e-05, 5.6696e-04, -1.8187e-03, -8.1539e-04, -6.9916e-05, + 5.3215e-04, 9.4986e-04, 1.6403e-04, 2.3937e-04, 2.7966e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 250.12, cls_loss 0.0059 cls_loss_mapping 0.0170 cls_loss_causal 0.6565 re_mapping 0.0115 re_causal 0.0357 /// teacc 98.82 lr 0.00010000 +Epoch 53, weight, value: tensor([[ 0.0349, -0.0673, -0.0647, ..., -0.0434, -0.0240, -0.0303], + [-0.0120, 0.0466, -0.0248, ..., 0.0069, 0.0955, 0.0225], + [-0.0089, -0.0768, -0.0187, ..., 0.0663, -0.0563, -0.0054], + ..., + [-0.0266, 0.0639, 0.0292, ..., -0.0451, 0.0529, 0.0187], + [-0.0174, -0.0392, 0.0119, ..., -0.0053, -0.0946, -0.0342], + [-0.0154, 0.0056, -0.0092, ..., -0.0768, -0.0612, 0.0226]], + device='cuda:0'), grad: tensor([[-1.4082e-05, 1.3620e-05, 0.0000e+00, ..., 7.3537e-06, + 1.1064e-05, 1.5616e-05], + [-3.5197e-05, 3.8594e-05, 0.0000e+00, ..., -1.7747e-05, + -8.9049e-05, -1.7313e-06], + [ 7.9989e-05, 2.8595e-05, 0.0000e+00, ..., -2.2411e-04, + 2.9728e-05, -1.8191e-04], + ..., + [-1.2264e-05, -3.7336e-04, 0.0000e+00, ..., 2.4915e-04, + -2.4223e-04, -1.3304e-04], + [ 6.3360e-05, 5.8472e-05, 0.0000e+00, ..., 1.9997e-05, + 3.7462e-05, 7.1466e-05], + [-2.2209e-04, -8.5354e-05, 0.0000e+00, ..., -2.2084e-05, + 1.8597e-04, -1.4913e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0258, 0.0141, 0.0121, 0.0281, 0.0219, -0.0195, 0.0248, 0.0262, + -0.0072, 0.0101], device='cuda:0'), grad: tensor([-8.2329e-06, -5.6863e-05, -8.4460e-05, -2.6584e-04, 1.0195e-03, + 1.1593e-04, 6.3002e-05, -7.3254e-05, 2.5725e-04, -9.6607e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 249.97, cls_loss 0.0058 cls_loss_mapping 0.0189 cls_loss_causal 0.6409 re_mapping 0.0116 re_causal 0.0356 /// teacc 98.74 lr 0.00010000 +Epoch 54, weight, value: tensor([[ 0.0351, -0.0677, -0.0662, ..., -0.0440, -0.0247, -0.0308], + [-0.0118, 0.0461, -0.0242, ..., 0.0069, 0.0956, 0.0222], + [-0.0088, -0.0773, -0.0189, ..., 0.0669, -0.0566, -0.0049], + ..., + [-0.0268, 0.0645, 0.0291, ..., -0.0454, 0.0540, 0.0191], + [-0.0174, -0.0392, 0.0117, ..., -0.0057, -0.0949, -0.0348], + [-0.0152, 0.0056, -0.0079, ..., -0.0774, -0.0619, 0.0227]], + device='cuda:0'), grad: tensor([[-3.0249e-06, 1.9986e-06, 0.0000e+00, ..., 1.7822e-05, + 3.4701e-06, 4.5821e-06], + [-1.7062e-05, -6.9588e-06, 0.0000e+00, ..., 9.4026e-06, + -3.7521e-05, -1.1418e-06], + [ 3.7663e-06, 1.3910e-05, 0.0000e+00, ..., -5.1670e-06, + 1.8552e-05, 2.9095e-06], + ..., + [-1.5974e-05, -9.5546e-05, 0.0000e+00, ..., -1.4775e-05, + -9.5785e-05, -7.5817e-05], + [ 1.5482e-05, 2.8536e-05, 0.0000e+00, ..., 4.0531e-05, + 3.5942e-05, 2.9147e-05], + [ 8.4564e-06, 9.5814e-06, 0.0000e+00, ..., 4.4793e-05, + 1.6928e-05, 5.1558e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0257, 0.0140, 0.0122, 0.0284, 0.0219, -0.0195, 0.0243, 0.0265, + -0.0074, 0.0100], device='cuda:0'), grad: tensor([-1.0207e-05, -1.8105e-06, 3.8654e-05, -8.0943e-05, -3.6925e-05, + 5.5224e-05, -5.2840e-05, -1.4329e-04, 1.2118e-04, 1.1134e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 250.13, cls_loss 0.0056 cls_loss_mapping 0.0170 cls_loss_causal 0.6705 re_mapping 0.0112 re_causal 0.0342 /// teacc 98.94 lr 0.00010000 +Epoch 55, weight, value: tensor([[ 0.0351, -0.0679, -0.0668, ..., -0.0444, -0.0250, -0.0314], + [-0.0116, 0.0461, -0.0242, ..., 0.0068, 0.0964, 0.0223], + [-0.0088, -0.0777, -0.0188, ..., 0.0675, -0.0574, -0.0046], + ..., + [-0.0271, 0.0648, 0.0290, ..., -0.0459, 0.0543, 0.0193], + [-0.0172, -0.0393, 0.0116, ..., -0.0061, -0.0953, -0.0351], + [-0.0152, 0.0058, -0.0079, ..., -0.0784, -0.0623, 0.0229]], + device='cuda:0'), grad: tensor([[ 1.3374e-05, 1.7891e-06, 7.2701e-08, ..., 1.0890e-04, + 5.7481e-06, 1.0744e-05], + [ 4.6909e-05, -8.2850e-06, 9.9360e-08, ..., 6.3002e-05, + -2.8789e-05, 2.7105e-05], + [ 8.1360e-05, 5.6066e-06, 2.0070e-07, ..., -2.7275e-04, + 4.8935e-05, -1.5974e-04], + ..., + [ 6.2168e-05, -9.6485e-06, 2.0038e-08, ..., 7.8231e-06, + 1.7092e-05, -3.6526e-06], + [ 6.8307e-05, 2.0675e-06, -1.5302e-06, ..., 1.9360e-04, + 1.2100e-05, 1.1986e-04], + [ 9.2804e-05, -4.1276e-06, 2.1700e-07, ..., 3.2187e-05, + 4.9174e-06, 5.6485e-07]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0259, 0.0142, 0.0123, 0.0281, 0.0217, -0.0192, 0.0246, 0.0262, + -0.0076, 0.0103], device='cuda:0'), grad: tensor([ 3.6740e-04, 2.1613e-04, -1.1897e-04, -9.0332e-03, 1.9521e-06, + 7.4120e-03, 2.4176e-04, 1.3661e-04, 4.2582e-04, 3.5262e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 250.31, cls_loss 0.0060 cls_loss_mapping 0.0182 cls_loss_causal 0.6359 re_mapping 0.0112 re_causal 0.0346 /// teacc 98.92 lr 0.00010000 +Epoch 56, weight, value: tensor([[ 0.0350, -0.0685, -0.0678, ..., -0.0450, -0.0258, -0.0319], + [-0.0109, 0.0463, -0.0242, ..., 0.0056, 0.0975, 0.0226], + [-0.0088, -0.0781, -0.0185, ..., 0.0687, -0.0583, -0.0042], + ..., + [-0.0276, 0.0650, 0.0294, ..., -0.0463, 0.0545, 0.0191], + [-0.0168, -0.0392, 0.0118, ..., -0.0065, -0.0953, -0.0358], + [-0.0153, 0.0059, -0.0081, ..., -0.0792, -0.0628, 0.0230]], + device='cuda:0'), grad: tensor([[ 2.9311e-05, 4.3027e-06, 3.0035e-07, ..., 3.3975e-05, + 7.3314e-06, 2.4870e-05], + [-4.2796e-05, -4.6223e-05, 1.4361e-06, ..., 3.7581e-05, + -1.0759e-04, -6.4336e-06], + [-1.0926e-04, 9.8050e-06, -3.2894e-06, ..., -2.9731e-04, + -4.0948e-05, -2.8324e-04], + ..., + [ 4.3154e-05, -9.0152e-06, 4.7660e-07, ..., 1.3745e-04, + 5.8711e-05, 1.6105e-04], + [ 8.4996e-05, 1.6540e-05, 1.3579e-06, ..., 4.8727e-05, + 2.6807e-05, 5.2631e-05], + [ 1.6749e-05, -1.5721e-05, -2.6561e-06, ..., 5.4762e-06, + 1.3955e-05, -5.2184e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([-0.0262, 0.0141, 0.0127, 0.0279, 0.0222, -0.0189, 0.0243, 0.0259, + -0.0074, 0.0102], device='cuda:0'), grad: tensor([ 7.9751e-05, -5.6356e-05, -5.0163e-04, 5.4091e-05, 9.0897e-05, + -1.5974e-04, 8.4877e-05, 2.7204e-04, 2.0075e-04, -6.4909e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 250.39, cls_loss 0.0044 cls_loss_mapping 0.0142 cls_loss_causal 0.6370 re_mapping 0.0112 re_causal 0.0343 /// teacc 98.81 lr 0.00010000 +Epoch 57, weight, value: tensor([[ 0.0351, -0.0689, -0.0702, ..., -0.0454, -0.0264, -0.0324], + [-0.0104, 0.0463, -0.0238, ..., 0.0055, 0.0983, 0.0225], + [-0.0089, -0.0788, -0.0186, ..., 0.0693, -0.0594, -0.0040], + ..., + [-0.0275, 0.0657, 0.0300, ..., -0.0467, 0.0553, 0.0197], + [-0.0166, -0.0395, 0.0114, ..., -0.0068, -0.0964, -0.0363], + [-0.0153, 0.0058, -0.0091, ..., -0.0798, -0.0636, 0.0228]], + device='cuda:0'), grad: tensor([[ 2.6040e-06, 6.1169e-06, 0.0000e+00, ..., 4.1127e-06, + 9.2909e-06, 1.3962e-05], + [ 1.4037e-05, -1.3247e-05, 0.0000e+00, ..., 1.2182e-05, + -9.2626e-05, 2.3901e-05], + [ 6.0171e-05, 1.6972e-05, 0.0000e+00, ..., 3.8370e-07, + 3.6389e-05, 3.3379e-05], + ..., + [ 2.5421e-05, -5.9962e-05, 0.0000e+00, ..., 2.5406e-06, + -5.2303e-05, -5.3346e-05], + [-1.6916e-04, 8.8587e-06, 0.0000e+00, ..., -1.8209e-05, + 1.6734e-05, -5.6297e-05], + [-1.3602e-04, 7.2352e-08, 0.0000e+00, ..., 2.8964e-06, + 1.9729e-05, -6.7997e-04]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0263, 0.0143, 0.0126, 0.0276, 0.0226, -0.0190, 0.0240, 0.0263, + -0.0072, 0.0098], device='cuda:0'), grad: tensor([ 1.7643e-05, 1.1116e-04, 2.0015e-04, 2.0039e-04, 1.2836e-03, + -1.6701e-04, 1.3745e-04, -1.0006e-05, -5.1594e-04, -1.2589e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 250.09, cls_loss 0.0054 cls_loss_mapping 0.0170 cls_loss_causal 0.6696 re_mapping 0.0113 re_causal 0.0347 /// teacc 98.94 lr 0.00010000 +Epoch 58, weight, value: tensor([[ 0.0349, -0.0692, -0.0716, ..., -0.0458, -0.0269, -0.0332], + [-0.0104, 0.0461, -0.0241, ..., 0.0052, 0.0986, 0.0222], + [-0.0088, -0.0792, -0.0184, ..., 0.0700, -0.0599, -0.0033], + ..., + [-0.0276, 0.0664, 0.0306, ..., -0.0473, 0.0562, 0.0201], + [-0.0166, -0.0399, 0.0114, ..., -0.0072, -0.0972, -0.0370], + [-0.0151, 0.0059, -0.0094, ..., -0.0811, -0.0642, 0.0227]], + device='cuda:0'), grad: tensor([[ 3.7342e-05, 1.4372e-05, 2.6819e-08, ..., 4.0054e-05, + 1.5348e-05, 5.3160e-06], + [ 3.2097e-05, 4.3660e-06, -8.7358e-07, ..., 1.0170e-05, + -1.7464e-05, -4.5188e-06], + [ 3.8683e-05, 1.8954e-05, 1.1380e-07, ..., -1.3728e-06, + 2.3514e-05, -1.0114e-06], + ..., + [ 1.3435e-04, 1.0498e-05, 3.3318e-07, ..., 7.2978e-06, + 1.8850e-05, -5.1409e-06], + [ 1.4830e-04, 5.5909e-05, 1.6182e-07, ..., 1.3161e-04, + 5.3853e-05, 5.2750e-05], + [ 3.5018e-05, 2.4796e-05, 6.1875e-08, ..., 5.1875e-07, + 5.3406e-05, -1.5819e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0266, 0.0141, 0.0130, 0.0270, 0.0229, -0.0184, 0.0239, 0.0265, + -0.0076, 0.0099], device='cuda:0'), grad: tensor([ 1.7238e-04, 9.3997e-05, 1.0616e-04, -1.5888e-03, 2.0194e-04, + 4.8375e-04, -3.0279e-04, 3.0565e-04, 5.8413e-04, -5.6207e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 57, time 249.57, cls_loss 0.0062 cls_loss_mapping 0.0161 cls_loss_causal 0.6155 re_mapping 0.0109 re_causal 0.0328 /// teacc 98.87 lr 0.00010000 +Epoch 59, weight, value: tensor([[ 0.0351, -0.0697, -0.0756, ..., -0.0462, -0.0276, -0.0342], + [-0.0103, 0.0460, -0.0245, ..., 0.0049, 0.0991, 0.0221], + [-0.0090, -0.0798, -0.0183, ..., 0.0707, -0.0607, -0.0030], + ..., + [-0.0279, 0.0670, 0.0306, ..., -0.0478, 0.0566, 0.0200], + [-0.0164, -0.0402, 0.0127, ..., -0.0074, -0.0985, -0.0377], + [-0.0148, 0.0060, -0.0088, ..., -0.0819, -0.0635, 0.0231]], + device='cuda:0'), grad: tensor([[-2.7680e-04, 5.4128e-06, 0.0000e+00, ..., 1.4579e-04, + 6.3851e-06, -9.3132e-06], + [ 2.4080e-05, 2.0444e-04, 0.0000e+00, ..., 1.8358e-05, + 1.2201e-04, 1.9562e-04], + [-3.2168e-06, 5.6297e-05, 0.0000e+00, ..., -1.4198e-04, + 4.8488e-05, -6.8247e-05], + ..., + [-7.7710e-06, -4.5276e-04, 0.0000e+00, ..., 3.4451e-05, + -3.7694e-04, -4.1485e-04], + [ 4.6268e-06, 2.7448e-05, 0.0000e+00, ..., 5.3138e-05, + 2.4453e-05, 5.8889e-05], + [ 9.1672e-05, 5.0440e-06, 0.0000e+00, ..., 9.2536e-06, + 2.1890e-05, 1.6779e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0266, 0.0137, 0.0129, 0.0277, 0.0230, -0.0185, 0.0241, 0.0259, + -0.0077, 0.0101], device='cuda:0'), grad: tensor([-9.4461e-04, 3.6407e-04, 5.4449e-05, 2.1422e-04, 2.4533e-04, + 2.0659e-04, -8.9943e-05, -5.9891e-04, 1.4675e-04, 4.0197e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 249.17, cls_loss 0.0047 cls_loss_mapping 0.0157 cls_loss_causal 0.6201 re_mapping 0.0102 re_causal 0.0322 /// teacc 98.96 lr 0.00010000 +Epoch 60, weight, value: tensor([[ 0.0353, -0.0700, -0.0768, ..., -0.0470, -0.0282, -0.0347], + [-0.0102, 0.0462, -0.0245, ..., 0.0044, 0.0996, 0.0221], + [-0.0089, -0.0801, -0.0184, ..., 0.0715, -0.0608, -0.0024], + ..., + [-0.0281, 0.0675, 0.0308, ..., -0.0483, 0.0571, 0.0203], + [-0.0161, -0.0401, 0.0128, ..., -0.0077, -0.0986, -0.0383], + [-0.0148, 0.0057, -0.0090, ..., -0.0824, -0.0643, 0.0234]], + device='cuda:0'), grad: tensor([[-3.5703e-05, 5.2527e-06, 0.0000e+00, ..., 8.0606e-07, + 4.2692e-06, 6.6981e-06], + [-2.7139e-06, 6.3218e-06, 0.0000e+00, ..., 6.3404e-06, + -1.8552e-05, 5.3681e-06], + [-6.6385e-06, 1.8299e-05, 0.0000e+00, ..., -8.7976e-05, + 5.7556e-06, -6.1035e-05], + ..., + [ 1.7658e-05, -1.0377e-04, 0.0000e+00, ..., 8.6665e-05, + -5.7906e-05, -2.3037e-05], + [-3.8058e-05, 7.9796e-06, 0.0000e+00, ..., 1.1630e-05, + 9.3207e-06, 8.0690e-06], + [ 1.5423e-05, 2.3827e-05, 0.0000e+00, ..., 2.9244e-06, + 1.9073e-05, 1.9163e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([-0.0266, 0.0136, 0.0132, 0.0276, 0.0224, -0.0185, 0.0245, 0.0260, + -0.0077, 0.0102], device='cuda:0'), grad: tensor([-2.4867e-04, 3.4124e-05, -6.0946e-06, 2.7001e-05, 8.1360e-05, + 7.8499e-05, 1.0943e-04, -6.9261e-05, -8.1837e-05, 7.5281e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 249.10, cls_loss 0.0056 cls_loss_mapping 0.0164 cls_loss_causal 0.6528 re_mapping 0.0105 re_causal 0.0323 /// teacc 98.99 lr 0.00010000 +Epoch 61, weight, value: tensor([[ 0.0353, -0.0703, -0.0774, ..., -0.0477, -0.0289, -0.0354], + [-0.0098, 0.0461, -0.0247, ..., 0.0046, 0.0999, 0.0221], + [-0.0093, -0.0808, -0.0183, ..., 0.0719, -0.0616, -0.0022], + ..., + [-0.0279, 0.0683, 0.0311, ..., -0.0485, 0.0579, 0.0208], + [-0.0156, -0.0403, 0.0129, ..., -0.0084, -0.0982, -0.0391], + [-0.0146, 0.0055, -0.0082, ..., -0.0826, -0.0649, 0.0237]], + device='cuda:0'), grad: tensor([[-1.0751e-05, 1.2323e-05, 8.0763e-10, ..., 6.2704e-05, + 5.2489e-06, 1.2827e-04], + [-3.6985e-05, -6.3717e-05, 4.8967e-09, ..., 3.5316e-05, + -1.1188e-04, 3.4660e-05], + [-6.8307e-05, 8.8736e-06, 7.0213e-09, ..., -3.1662e-04, + 7.5735e-06, -5.4169e-04], + ..., + [ 7.2837e-05, 2.7013e-04, -3.1810e-08, ..., 9.7811e-05, + 5.2333e-05, 7.4053e-04], + [ 1.0796e-05, 2.4229e-05, 7.7125e-10, ..., 2.0787e-05, + 2.6375e-05, 5.0783e-05], + [-3.8862e-05, -4.2415e-04, 1.9209e-09, ..., 3.7253e-05, + -2.0564e-05, -8.8453e-04]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0270, 0.0136, 0.0127, 0.0278, 0.0222, -0.0190, 0.0243, 0.0262, + -0.0070, 0.0104], device='cuda:0'), grad: tensor([ 1.1456e-04, 5.3309e-06, -8.8882e-04, 1.3995e-04, 5.5218e-04, + 7.2837e-05, 5.1647e-05, 1.2665e-03, 1.0663e-04, -1.4210e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 249.04, cls_loss 0.0048 cls_loss_mapping 0.0152 cls_loss_causal 0.6577 re_mapping 0.0103 re_causal 0.0321 /// teacc 98.94 lr 0.00010000 +Epoch 62, weight, value: tensor([[ 0.0354, -0.0705, -0.0793, ..., -0.0481, -0.0293, -0.0360], + [-0.0096, 0.0460, -0.0247, ..., 0.0044, 0.1003, 0.0220], + [-0.0093, -0.0814, -0.0183, ..., 0.0724, -0.0622, -0.0019], + ..., + [-0.0282, 0.0688, 0.0311, ..., -0.0488, 0.0587, 0.0211], + [-0.0153, -0.0399, 0.0130, ..., -0.0089, -0.0985, -0.0393], + [-0.0148, 0.0053, -0.0071, ..., -0.0833, -0.0653, 0.0240]], + device='cuda:0'), grad: tensor([[ 1.2808e-05, 7.0259e-06, 2.9299e-06, ..., 1.1824e-05, + 2.3052e-05, 1.0192e-05], + [-3.7956e-04, -1.2326e-04, -9.0301e-05, ..., -1.1635e-04, + -6.6185e-04, -1.5521e-04], + [ 2.2995e-04, 7.1347e-05, 3.9488e-05, ..., 8.1062e-05, + 3.1114e-04, 8.8513e-05], + ..., + [ 5.7161e-05, -8.1897e-05, 1.3456e-05, ..., 1.9372e-05, + 1.0602e-05, -4.8906e-05], + [-5.0932e-05, 3.0786e-05, 9.4771e-06, ..., 1.5855e-05, + 1.2469e-04, 3.8892e-05], + [ 3.2455e-05, 4.8727e-05, 7.8902e-06, ..., 8.2515e-07, + 6.7770e-05, -1.7717e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0271, 0.0134, 0.0128, 0.0280, 0.0218, -0.0191, 0.0245, 0.0263, + -0.0068, 0.0104], device='cuda:0'), grad: tensor([ 3.9011e-05, -8.9121e-04, 6.0558e-04, 1.3053e-04, 2.0206e-04, + -1.5900e-05, -1.1139e-05, 3.8952e-05, -1.0478e-04, 6.0685e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 249.35, cls_loss 0.0048 cls_loss_mapping 0.0131 cls_loss_causal 0.6545 re_mapping 0.0104 re_causal 0.0328 /// teacc 98.93 lr 0.00010000 +Epoch 63, weight, value: tensor([[ 0.0354, -0.0707, -0.0803, ..., -0.0486, -0.0296, -0.0365], + [-0.0092, 0.0461, -0.0246, ..., 0.0043, 0.1008, 0.0222], + [-0.0092, -0.0818, -0.0179, ..., 0.0730, -0.0628, -0.0016], + ..., + [-0.0285, 0.0694, 0.0316, ..., -0.0494, 0.0594, 0.0214], + [-0.0152, -0.0402, 0.0128, ..., -0.0095, -0.0990, -0.0399], + [-0.0145, 0.0051, -0.0071, ..., -0.0831, -0.0658, 0.0233]], + device='cuda:0'), grad: tensor([[ 3.3267e-06, 7.3016e-06, 6.6240e-08, ..., 9.4473e-06, + 1.4842e-05, 1.7270e-05], + [-3.0160e-04, 3.8370e-06, -8.4564e-07, ..., -2.2972e-04, + -3.3808e-04, -9.5785e-05], + [ 2.2197e-04, 6.4552e-05, 6.3470e-07, ..., 2.0695e-04, + 3.4451e-04, 1.7583e-04], + ..., + [ 2.2084e-05, -1.1259e-04, 9.4587e-08, ..., -7.6115e-05, + -1.8728e-04, -1.3423e-04], + [ 2.2590e-05, 1.2696e-05, -1.2005e-06, ..., 2.5779e-05, + 4.3899e-05, 3.2753e-05], + [-3.9756e-05, -4.3452e-05, 7.1304e-08, ..., 1.7613e-05, + -2.5585e-05, -1.2505e-04]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0274, 0.0136, 0.0128, 0.0278, 0.0230, -0.0193, 0.0244, 0.0264, + -0.0069, 0.0099], device='cuda:0'), grad: tensor([ 3.5822e-05, -7.8154e-04, 7.7009e-04, 6.6519e-05, 1.6713e-04, + 3.3617e-05, 1.0866e-04, -2.1148e-04, 1.0222e-04, -2.9063e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 248.87, cls_loss 0.0056 cls_loss_mapping 0.0147 cls_loss_causal 0.6053 re_mapping 0.0096 re_causal 0.0306 /// teacc 98.94 lr 0.00010000 +Epoch 64, weight, value: tensor([[ 0.0354, -0.0709, -0.0807, ..., -0.0490, -0.0301, -0.0369], + [-0.0089, 0.0460, -0.0245, ..., 0.0040, 0.1014, 0.0222], + [-0.0094, -0.0826, -0.0178, ..., 0.0736, -0.0636, -0.0015], + ..., + [-0.0286, 0.0702, 0.0318, ..., -0.0495, 0.0602, 0.0218], + [-0.0152, -0.0404, 0.0127, ..., -0.0098, -0.0997, -0.0403], + [-0.0146, 0.0050, -0.0073, ..., -0.0835, -0.0666, 0.0234]], + device='cuda:0'), grad: tensor([[ 1.3605e-05, 5.5023e-06, 6.0769e-08, ..., 8.3968e-06, + 7.2308e-06, 8.4937e-06], + [-8.3074e-06, -1.3195e-05, -1.1744e-06, ..., 4.1798e-06, + -5.0902e-05, -7.0035e-06], + [ 2.3529e-05, -2.0754e-04, 4.0932e-07, ..., -2.2101e-04, + -2.5415e-04, -3.4547e-04], + ..., + [ 2.3946e-05, 1.6367e-04, 4.0140e-07, ..., 2.1219e-04, + 2.2507e-04, 2.8610e-04], + [ 8.6725e-05, 6.1505e-06, 1.7101e-07, ..., 3.7607e-06, + 9.7156e-06, 1.1250e-05], + [ 3.7961e-06, 1.8150e-05, -3.4226e-07, ..., 3.8855e-06, + 2.5988e-05, 5.9791e-06]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0274, 0.0136, 0.0126, 0.0282, 0.0228, -0.0195, 0.0245, 0.0268, + -0.0071, 0.0096], device='cuda:0'), grad: tensor([ 6.1154e-05, 2.1771e-05, -5.3453e-04, -2.5198e-05, 5.3495e-05, + -5.1785e-04, 3.1054e-05, 5.8889e-04, 3.0255e-04, 1.8507e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 63, time 248.83, cls_loss 0.0048 cls_loss_mapping 0.0159 cls_loss_causal 0.6186 re_mapping 0.0099 re_causal 0.0311 /// teacc 99.00 lr 0.00010000 +Epoch 65, weight, value: tensor([[ 0.0352, -0.0710, -0.0814, ..., -0.0497, -0.0304, -0.0392], + [-0.0089, 0.0460, -0.0245, ..., 0.0037, 0.1017, 0.0219], + [-0.0095, -0.0830, -0.0177, ..., 0.0741, -0.0643, -0.0014], + ..., + [-0.0288, 0.0707, 0.0320, ..., -0.0499, 0.0608, 0.0222], + [-0.0152, -0.0406, 0.0127, ..., -0.0101, -0.1000, -0.0407], + [-0.0141, 0.0049, -0.0074, ..., -0.0842, -0.0666, 0.0241]], + device='cuda:0'), grad: tensor([[-5.1165e-04, 1.5981e-06, 2.6630e-09, ..., -1.5378e-04, + 1.3011e-06, -1.5005e-05], + [ 7.0520e-06, 2.6528e-08, -9.9826e-08, ..., 3.3438e-05, + -2.3052e-05, 2.5272e-05], + [ 6.3300e-05, 8.7172e-06, 3.3499e-08, ..., -9.9540e-05, + 1.0341e-05, -9.3162e-05], + ..., + [ 3.8177e-05, -2.4855e-05, 2.3749e-08, ..., 5.5462e-05, + -2.9281e-05, 4.9740e-05], + [-1.0937e-05, 4.1462e-06, 1.4508e-08, ..., 1.4484e-05, + 4.5151e-06, 1.8179e-05], + [ 2.8038e-04, 2.3991e-05, 5.5516e-09, ..., 9.1136e-05, + 1.9565e-05, 6.0886e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0283, 0.0133, 0.0124, 0.0280, 0.0227, -0.0192, 0.0249, 0.0268, + -0.0074, 0.0106], device='cuda:0'), grad: tensor([-1.9913e-03, 1.1063e-04, 4.3839e-05, 8.8573e-05, -2.2483e-04, + 1.6475e-04, 4.1533e-04, 2.4772e-04, 2.4512e-05, 1.1215e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 64, time 248.98, cls_loss 0.0035 cls_loss_mapping 0.0124 cls_loss_causal 0.5869 re_mapping 0.0101 re_causal 0.0305 /// teacc 99.02 lr 0.00010000 +Epoch 66, weight, value: tensor([[ 0.0356, -0.0718, -0.0821, ..., -0.0498, -0.0323, -0.0395], + [-0.0085, 0.0460, -0.0245, ..., 0.0034, 0.1022, 0.0218], + [-0.0096, -0.0835, -0.0177, ..., 0.0746, -0.0646, -0.0009], + ..., + [-0.0289, 0.0713, 0.0318, ..., -0.0505, 0.0614, 0.0223], + [-0.0150, -0.0407, 0.0133, ..., -0.0103, -0.1005, -0.0413], + [-0.0142, 0.0049, -0.0073, ..., -0.0850, -0.0668, 0.0245]], + device='cuda:0'), grad: tensor([[-5.8442e-05, 2.1793e-06, 7.0198e-08, ..., -1.5765e-05, + 1.8561e-06, 6.1467e-06], + [-1.1019e-05, -1.6898e-05, 1.0961e-07, ..., 3.9376e-06, + -3.5942e-05, -1.0148e-05], + [ 3.0339e-05, 1.5840e-05, 6.5949e-08, ..., -2.3488e-06, + 1.5393e-05, 9.7156e-06], + ..., + [ 1.8373e-05, -2.4095e-05, 5.5559e-08, ..., 5.0403e-06, + -1.6838e-05, -5.9940e-06], + [ 7.8082e-06, 6.8881e-06, 6.9290e-07, ..., 8.3372e-06, + 8.8066e-06, 1.2822e-05], + [-2.6509e-05, -4.4629e-06, 2.0012e-07, ..., 6.1058e-06, + 7.5214e-06, -5.5611e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0281, 0.0134, 0.0122, 0.0279, 0.0225, -0.0191, 0.0250, 0.0268, + -0.0073, 0.0107], device='cuda:0'), grad: tensor([-4.4274e-04, -9.3207e-06, 1.4162e-04, 6.5267e-05, 6.4611e-05, + 2.4274e-05, 9.7275e-05, 2.5615e-05, 1.0586e-04, -7.2241e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 249.16, cls_loss 0.0046 cls_loss_mapping 0.0158 cls_loss_causal 0.6459 re_mapping 0.0092 re_causal 0.0291 /// teacc 98.96 lr 0.00010000 +Epoch 67, weight, value: tensor([[ 0.0359, -0.0723, -0.0837, ..., -0.0501, -0.0329, -0.0399], + [-0.0079, 0.0463, -0.0245, ..., 0.0033, 0.1031, 0.0221], + [-0.0097, -0.0842, -0.0169, ..., 0.0756, -0.0653, -0.0005], + ..., + [-0.0292, 0.0719, 0.0318, ..., -0.0516, 0.0619, 0.0224], + [-0.0150, -0.0407, 0.0119, ..., -0.0110, -0.1012, -0.0417], + [-0.0142, 0.0045, -0.0062, ..., -0.0863, -0.0677, 0.0243]], + device='cuda:0'), grad: tensor([[-9.6500e-05, 1.9167e-06, 0.0000e+00, ..., -3.1561e-05, + 2.5053e-06, 3.8259e-06], + [-4.8161e-05, -2.8789e-05, 0.0000e+00, ..., -1.3940e-05, + -1.3196e-04, -4.5061e-05], + [ 3.8803e-05, 6.1393e-06, 0.0000e+00, ..., 6.5491e-06, + 1.6540e-05, 2.0973e-06], + ..., + [ 8.4788e-06, -2.3797e-05, 0.0000e+00, ..., 5.7481e-06, + -1.6421e-05, -4.8764e-06], + [ 5.9009e-05, 6.3404e-06, 0.0000e+00, ..., 2.8625e-05, + 5.0038e-05, 2.3067e-05], + [-5.6475e-06, 1.2644e-05, 0.0000e+00, ..., 5.2042e-06, + 2.2665e-05, -8.6948e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0279, 0.0138, 0.0125, 0.0280, 0.0228, -0.0190, 0.0245, 0.0264, + -0.0074, 0.0103], device='cuda:0'), grad: tensor([-4.4465e-04, -1.3530e-04, 1.7023e-04, 5.0873e-05, 4.9472e-05, + 2.4959e-05, 3.0294e-05, 1.5438e-05, 2.4939e-04, -1.0990e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 249.08, cls_loss 0.0039 cls_loss_mapping 0.0129 cls_loss_causal 0.6127 re_mapping 0.0097 re_causal 0.0300 /// teacc 98.92 lr 0.00010000 +Epoch 68, weight, value: tensor([[ 0.0363, -0.0726, -0.0840, ..., -0.0500, -0.0333, -0.0403], + [-0.0075, 0.0472, -0.0239, ..., 0.0029, 0.1041, 0.0222], + [-0.0099, -0.0848, -0.0170, ..., 0.0763, -0.0658, 0.0001], + ..., + [-0.0293, 0.0722, 0.0309, ..., -0.0522, 0.0622, 0.0229], + [-0.0150, -0.0409, 0.0119, ..., -0.0114, -0.1016, -0.0423], + [-0.0145, 0.0040, -0.0061, ..., -0.0874, -0.0688, 0.0238]], + device='cuda:0'), grad: tensor([[-7.1339e-07, 3.6545e-06, 7.2690e-07, ..., 2.6654e-06, + 3.4329e-06, 5.9828e-06], + [ 3.3993e-06, 5.6215e-06, 1.3588e-06, ..., 2.9802e-06, + 3.2187e-06, 7.2643e-06], + [ 1.9327e-05, 1.4961e-05, 6.4895e-06, ..., -2.7604e-06, + 1.6928e-05, 1.1347e-05], + ..., + [-1.2256e-06, -4.8399e-05, 1.4831e-07, ..., 1.1683e-05, + -6.7055e-05, -8.1509e-06], + [-3.7951e-07, 1.4536e-05, -1.0461e-05, ..., 8.6203e-06, + 1.0319e-05, 2.8476e-05], + [-6.7651e-05, -6.0916e-05, 1.7451e-07, ..., -4.2580e-06, + 1.7464e-05, -1.3912e-04]], device='cuda:0') +Epoch 68, bias, value: tensor([-0.0274, 0.0140, 0.0125, 0.0283, 0.0228, -0.0190, 0.0244, 0.0266, + -0.0077, 0.0096], device='cuda:0'), grad: tensor([ 8.9258e-06, 2.6211e-05, 6.5386e-05, 4.1455e-05, 1.4496e-04, + -3.2276e-05, 3.0071e-05, -4.0144e-05, 4.8876e-05, -2.9302e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 249.40, cls_loss 0.0043 cls_loss_mapping 0.0130 cls_loss_causal 0.6333 re_mapping 0.0095 re_causal 0.0293 /// teacc 98.97 lr 0.00010000 +Epoch 69, weight, value: tensor([[ 0.0363, -0.0728, -0.0878, ..., -0.0507, -0.0340, -0.0406], + [-0.0078, 0.0471, -0.0256, ..., 0.0025, 0.1044, 0.0220], + [-0.0102, -0.0851, -0.0196, ..., 0.0770, -0.0661, 0.0010], + ..., + [-0.0296, 0.0727, 0.0315, ..., -0.0534, 0.0626, 0.0230], + [-0.0147, -0.0411, 0.0120, ..., -0.0121, -0.1025, -0.0429], + [-0.0136, 0.0040, -0.0031, ..., -0.0884, -0.0684, 0.0239]], + device='cuda:0'), grad: tensor([[ 5.4203e-06, 5.2862e-06, 2.5518e-07, ..., 1.7598e-05, + 4.1425e-06, 1.2852e-05], + [ 1.9595e-06, 3.0976e-06, 8.4867e-08, ..., 9.5516e-06, + -1.3180e-05, 5.8785e-06], + [ 8.6010e-05, 2.7016e-05, 4.7963e-08, ..., 7.7128e-05, + 2.6152e-05, 3.8505e-05], + ..., + [ 1.7077e-05, -5.5581e-05, 7.0455e-07, ..., 2.1309e-06, + -4.3362e-05, -5.1171e-05], + [-1.5330e-04, 2.4978e-06, 5.7975e-07, ..., -1.2743e-04, + 3.0827e-06, -1.4603e-05], + [ 1.5991e-06, 2.2292e-05, -3.8184e-06, ..., 3.3855e-05, + 9.6187e-06, 2.1592e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0274, 0.0135, 0.0124, 0.0281, 0.0231, -0.0189, 0.0249, 0.0263, + -0.0078, 0.0101], device='cuda:0'), grad: tensor([ 5.2005e-05, 2.8297e-05, 3.1734e-04, 2.7284e-05, -6.4492e-05, + -3.8207e-05, 1.7539e-05, -1.7181e-05, -4.4322e-04, 1.2022e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 248.97, cls_loss 0.0055 cls_loss_mapping 0.0158 cls_loss_causal 0.5879 re_mapping 0.0094 re_causal 0.0278 /// teacc 98.93 lr 0.00010000 +Epoch 70, weight, value: tensor([[ 0.0366, -0.0733, -0.0896, ..., -0.0509, -0.0350, -0.0410], + [-0.0074, 0.0472, -0.0256, ..., 0.0018, 0.1051, 0.0220], + [-0.0101, -0.0859, -0.0181, ..., 0.0777, -0.0667, 0.0010], + ..., + [-0.0295, 0.0734, 0.0313, ..., -0.0534, 0.0634, 0.0235], + [-0.0154, -0.0414, 0.0092, ..., -0.0122, -0.1037, -0.0439], + [-0.0134, 0.0038, -0.0028, ..., -0.0890, -0.0690, 0.0239]], + device='cuda:0'), grad: tensor([[ 1.1325e-06, 4.8615e-07, 1.1714e-08, ..., 1.5058e-05, + 1.5320e-06, 8.3372e-06], + [ 9.7156e-05, -1.8571e-06, 2.0082e-09, ..., 1.6260e-04, + -1.6131e-06, 8.7619e-05], + [-1.3673e-04, 4.1127e-06, 1.6444e-09, ..., -2.3437e-04, + -9.0227e-06, -1.2743e-04], + ..., + [ 2.4572e-05, -3.0085e-05, 6.2922e-08, ..., 2.6613e-05, + -2.4423e-05, -8.3372e-06], + [ 1.6177e-04, 1.5348e-06, 1.7986e-07, ..., 1.5333e-05, + 3.6210e-06, 1.2112e-04], + [-1.7595e-04, 4.1239e-06, -3.5390e-07, ..., 3.8110e-06, + 5.3644e-06, -1.1945e-04]], device='cuda:0') +Epoch 70, bias, value: tensor([-0.0273, 0.0135, 0.0123, 0.0289, 0.0228, -0.0191, 0.0246, 0.0268, + -0.0088, 0.0103], device='cuda:0'), grad: tensor([ 1.9252e-05, 3.4904e-04, -4.8327e-04, -2.6852e-05, 7.7665e-05, + 6.6817e-05, -3.9786e-05, 3.6746e-05, 3.9673e-04, -3.9697e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 69---------------------------------------------------- +epoch 69, time 249.99, cls_loss 0.0037 cls_loss_mapping 0.0120 cls_loss_causal 0.6155 re_mapping 0.0094 re_causal 0.0283 /// teacc 99.05 lr 0.00010000 +Epoch 71, weight, value: tensor([[ 0.0368, -0.0735, -0.0900, ..., -0.0517, -0.0353, -0.0415], + [-0.0075, 0.0470, -0.0257, ..., 0.0012, 0.1053, 0.0216], + [-0.0101, -0.0862, -0.0181, ..., 0.0785, -0.0670, 0.0016], + ..., + [-0.0297, 0.0741, 0.0312, ..., -0.0540, 0.0641, 0.0241], + [-0.0151, -0.0415, 0.0090, ..., -0.0123, -0.1040, -0.0444], + [-0.0135, 0.0037, -0.0028, ..., -0.0903, -0.0694, 0.0237]], + device='cuda:0'), grad: tensor([[-1.2390e-05, 7.5884e-06, 0.0000e+00, ..., 1.3225e-05, + 1.1146e-05, 1.6391e-05], + [ 2.0519e-05, 3.4183e-05, 0.0000e+00, ..., 3.2902e-05, + 4.2826e-05, 6.2585e-05], + [-3.0115e-05, 1.6797e-04, 0.0000e+00, ..., 2.8893e-05, + 2.3377e-04, 2.1267e-04], + ..., + [ 9.0227e-06, -3.0541e-04, 0.0000e+00, ..., -1.6689e-04, + -4.1032e-04, -4.4084e-04], + [ 3.6031e-05, 1.3515e-05, 0.0000e+00, ..., 3.9190e-05, + 2.1234e-05, 3.8028e-05], + [ 5.3160e-06, 1.9461e-05, 0.0000e+00, ..., 3.5912e-05, + 3.4541e-05, 7.1526e-05]], device='cuda:0') +Epoch 71, bias, value: tensor([-0.0276, 0.0130, 0.0126, 0.0290, 0.0230, -0.0190, 0.0249, 0.0271, + -0.0087, 0.0099], device='cuda:0'), grad: tensor([-7.7784e-05, 2.4009e-04, 5.8460e-04, -4.5270e-05, -5.1647e-05, + 1.3220e-04, 4.0412e-05, -1.1387e-03, 1.6522e-04, 1.5128e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 249.24, cls_loss 0.0036 cls_loss_mapping 0.0112 cls_loss_causal 0.6578 re_mapping 0.0086 re_causal 0.0283 /// teacc 98.95 lr 0.00010000 +Epoch 72, weight, value: tensor([[ 0.0371, -0.0741, -0.0902, ..., -0.0518, -0.0362, -0.0421], + [-0.0070, 0.0474, -0.0257, ..., 0.0011, 0.1063, 0.0218], + [-0.0101, -0.0864, -0.0182, ..., 0.0792, -0.0674, 0.0023], + ..., + [-0.0299, 0.0745, 0.0313, ..., -0.0546, 0.0645, 0.0243], + [-0.0149, -0.0417, 0.0090, ..., -0.0126, -0.1044, -0.0449], + [-0.0138, 0.0034, -0.0028, ..., -0.0911, -0.0703, 0.0235]], + device='cuda:0'), grad: tensor([[ 6.2212e-06, 5.7593e-06, 0.0000e+00, ..., 2.4259e-05, + 4.4778e-06, 2.9370e-05], + [ 1.2420e-05, 8.7917e-06, 0.0000e+00, ..., 3.2485e-05, + 8.4098e-07, 3.9309e-05], + [-4.3809e-05, 1.8477e-05, 0.0000e+00, ..., -2.0039e-04, + 1.7270e-05, -1.8895e-04], + ..., + [ 1.0207e-05, -2.3615e-04, 0.0000e+00, ..., 4.9114e-05, + -1.8966e-04, -1.3518e-04], + [ 7.9513e-05, 8.6486e-05, 0.0000e+00, ..., 5.1171e-05, + 5.2750e-05, 1.0860e-04], + [-1.7691e-04, 6.5744e-05, 0.0000e+00, ..., -4.6819e-05, + 7.6652e-05, -1.0270e-04]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0272, 0.0134, 0.0130, 0.0290, 0.0234, -0.0189, 0.0236, 0.0270, + -0.0085, 0.0093], device='cuda:0'), grad: tensor([ 7.0095e-05, 1.1063e-04, -5.2738e-04, 1.7989e-04, 3.1281e-04, + 8.0913e-06, -2.5675e-05, -1.7786e-04, 3.7837e-04, -3.3021e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 249.01, cls_loss 0.0042 cls_loss_mapping 0.0112 cls_loss_causal 0.6267 re_mapping 0.0088 re_causal 0.0273 /// teacc 99.03 lr 0.00010000 +Epoch 73, weight, value: tensor([[ 0.0373, -0.0745, -0.0907, ..., -0.0523, -0.0369, -0.0425], + [-0.0068, 0.0474, -0.0258, ..., 0.0008, 0.1066, 0.0214], + [-0.0104, -0.0871, -0.0183, ..., 0.0797, -0.0680, 0.0026], + ..., + [-0.0300, 0.0751, 0.0314, ..., -0.0551, 0.0652, 0.0244], + [-0.0146, -0.0416, 0.0090, ..., -0.0129, -0.1048, -0.0454], + [-0.0138, 0.0033, -0.0027, ..., -0.0903, -0.0706, 0.0241]], + device='cuda:0'), grad: tensor([[ 1.5479e-06, 1.2098e-06, 0.0000e+00, ..., 1.0118e-05, + 7.5670e-07, 1.1653e-05], + [ 2.0931e-07, -6.0201e-06, 0.0000e+00, ..., 1.4752e-05, + -1.6183e-05, 6.1579e-06], + [-1.6952e-04, -1.4171e-05, 0.0000e+00, ..., -3.8791e-04, + 5.3532e-06, -2.6393e-04], + ..., + [ 1.6403e-04, 1.4029e-05, 0.0000e+00, ..., 3.2640e-04, + 1.5628e-06, 2.3603e-04], + [ 8.5011e-06, 2.1104e-06, 0.0000e+00, ..., 1.7583e-05, + 1.8915e-06, 1.5914e-05], + [-5.5015e-05, -4.9472e-06, 0.0000e+00, ..., 9.8795e-06, + 1.3635e-06, -5.2989e-05]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0274, 0.0132, 0.0128, 0.0290, 0.0230, -0.0189, 0.0232, 0.0271, + -0.0084, 0.0101], device='cuda:0'), grad: tensor([ 1.4603e-05, 1.9073e-05, -7.0858e-04, 1.3141e-06, 1.9908e-05, + 8.0466e-05, 1.1079e-05, 6.5517e-04, 4.1932e-05, -1.3447e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 249.13, cls_loss 0.0039 cls_loss_mapping 0.0108 cls_loss_causal 0.6336 re_mapping 0.0086 re_causal 0.0277 /// teacc 98.90 lr 0.00010000 +Epoch 74, weight, value: tensor([[ 0.0379, -0.0748, -0.0914, ..., -0.0524, -0.0373, -0.0429], + [-0.0070, 0.0469, -0.0258, ..., 0.0005, 0.1067, 0.0211], + [-0.0106, -0.0876, -0.0185, ..., 0.0802, -0.0687, 0.0029], + ..., + [-0.0301, 0.0760, 0.0320, ..., -0.0557, 0.0663, 0.0249], + [-0.0139, -0.0418, 0.0090, ..., -0.0131, -0.1053, -0.0457], + [-0.0139, 0.0032, -0.0028, ..., -0.0909, -0.0712, 0.0242]], + device='cuda:0'), grad: tensor([[ 3.7879e-05, 4.7944e-06, 9.6043e-09, ..., 1.6376e-05, + 7.6368e-06, 1.5393e-05], + [-7.2837e-05, -1.1921e-04, 9.0222e-10, ..., 5.1856e-06, + -3.3331e-04, -1.1837e-04], + [-3.1620e-05, 3.9458e-05, 1.4843e-09, ..., -8.3327e-05, + 1.2338e-04, -2.8282e-05], + ..., + [ 5.8800e-05, 4.5896e-05, 2.9802e-08, ..., 1.3374e-05, + 1.3828e-04, 7.6771e-05], + [ 3.6031e-05, 9.1493e-06, 1.9136e-08, ..., 1.2763e-05, + 1.5482e-05, 1.9133e-05], + [ 2.7850e-05, 3.8184e-06, -1.2200e-07, ..., 1.8001e-05, + 1.1317e-05, 9.8646e-06]], device='cuda:0') +Epoch 74, bias, value: tensor([-0.0269, 0.0125, 0.0126, 0.0296, 0.0228, -0.0196, 0.0229, 0.0275, + -0.0079, 0.0099], device='cuda:0'), grad: tensor([ 9.5069e-05, -3.1614e-04, -3.8832e-05, 2.8539e-04, -7.6294e-05, + -3.7551e-04, -1.3120e-05, 2.3568e-04, 1.0538e-04, 9.7573e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 73, time 249.16, cls_loss 0.0035 cls_loss_mapping 0.0120 cls_loss_causal 0.6178 re_mapping 0.0093 re_causal 0.0280 /// teacc 99.03 lr 0.00010000 +Epoch 75, weight, value: tensor([[ 0.0383, -0.0752, -0.0917, ..., -0.0526, -0.0380, -0.0432], + [-0.0064, 0.0472, -0.0258, ..., 0.0002, 0.1081, 0.0213], + [-0.0104, -0.0877, -0.0186, ..., 0.0812, -0.0692, 0.0042], + ..., + [-0.0308, 0.0761, 0.0319, ..., -0.0566, 0.0660, 0.0247], + [-0.0138, -0.0417, 0.0090, ..., -0.0135, -0.1058, -0.0461], + [-0.0141, 0.0031, -0.0027, ..., -0.0914, -0.0716, 0.0243]], + device='cuda:0'), grad: tensor([[-1.3104e-06, 1.7639e-06, 1.8615e-07, ..., 3.0667e-05, + 2.9858e-06, 8.8140e-06], + [-1.7390e-05, -8.7842e-06, 9.2899e-07, ..., 4.3422e-05, + -5.0008e-05, 1.6272e-05], + [ 6.2883e-06, 9.0376e-06, -2.7046e-06, ..., -9.6619e-05, + 1.6600e-05, -4.6909e-05], + ..., + [ 7.0632e-06, -2.6301e-05, 2.5844e-07, ..., 1.0438e-05, + -2.0459e-05, -2.2590e-05], + [ 1.3508e-05, 7.2084e-06, 5.6485e-07, ..., 7.4208e-05, + 1.6302e-05, 2.5392e-05], + [-5.7846e-05, -4.4480e-06, 1.1042e-07, ..., 5.9009e-06, + 1.7568e-05, -8.9347e-05]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0264, 0.0131, 0.0134, 0.0295, 0.0226, -0.0196, 0.0227, 0.0269, + -0.0077, 0.0095], device='cuda:0'), grad: tensor([ 5.7369e-05, 4.1008e-05, -1.5450e-04, 2.5555e-05, 1.9896e-04, + 7.3791e-05, -2.6512e-04, -7.0445e-06, 1.8883e-04, -1.5938e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 74---------------------------------------------------- +epoch 74, time 249.56, cls_loss 0.0035 cls_loss_mapping 0.0113 cls_loss_causal 0.6124 re_mapping 0.0086 re_causal 0.0270 /// teacc 99.07 lr 0.00010000 +Epoch 76, weight, value: tensor([[ 3.8142e-02, -7.5874e-02, -9.2389e-02, ..., -5.2984e-02, + -3.8668e-02, -4.3888e-02], + [-5.9852e-03, 4.7375e-02, -2.5881e-02, ..., 5.5953e-06, + 1.0889e-01, 2.1456e-02], + [-1.0404e-02, -8.8269e-02, -1.7966e-02, ..., 8.2020e-02, + -6.9877e-02, 4.6408e-03], + ..., + [-3.1037e-02, 7.6573e-02, 3.1821e-02, ..., -5.6948e-02, + 6.6422e-02, 2.4920e-02], + [-1.3428e-02, -4.1898e-02, 8.6652e-03, ..., -1.4050e-02, + -1.0674e-01, -4.6816e-02], + [-1.4081e-02, 3.0303e-03, -2.4950e-03, ..., -9.1853e-02, + -7.2324e-02, 2.4360e-02]], device='cuda:0'), grad: tensor([[ 2.3060e-06, 2.5406e-06, 4.1234e-07, ..., 2.9989e-06, + 2.9225e-06, 3.4031e-06], + [-5.9456e-06, 6.0834e-06, -2.5705e-06, ..., -2.2541e-08, + -1.6600e-05, -2.5965e-06], + [ 2.6390e-05, 7.4022e-06, 2.1681e-06, ..., -3.5278e-06, + 1.7956e-05, 1.1235e-05], + ..., + [ 1.6838e-05, -9.2804e-05, 3.1218e-06, ..., 3.1553e-06, + -5.0694e-05, -5.9992e-05], + [-1.7494e-05, 6.3889e-06, 1.7453e-06, ..., 2.7958e-06, + 9.4324e-06, -3.0752e-06], + [ 1.1697e-05, 5.8800e-05, -1.7462e-10, ..., 2.5649e-06, + 5.0843e-05, 3.5882e-05]], device='cuda:0') +Epoch 76, bias, value: tensor([-0.0269, 0.0134, 0.0136, 0.0292, 0.0225, -0.0195, 0.0229, 0.0268, + -0.0075, 0.0094], device='cuda:0'), grad: tensor([ 1.7896e-05, 2.2277e-05, 8.8871e-05, -1.3673e-04, -2.9534e-05, + -2.3365e-05, 6.0461e-06, -5.4419e-05, -3.0294e-05, 1.3947e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 248.64, cls_loss 0.0030 cls_loss_mapping 0.0104 cls_loss_causal 0.6116 re_mapping 0.0091 re_causal 0.0272 /// teacc 99.07 lr 0.00010000 +Epoch 77, weight, value: tensor([[ 0.0390, -0.0762, -0.0934, ..., -0.0536, -0.0392, -0.0444], + [-0.0057, 0.0475, -0.0259, ..., -0.0005, 0.1094, 0.0215], + [-0.0105, -0.0891, -0.0176, ..., 0.0827, -0.0705, 0.0049], + ..., + [-0.0313, 0.0772, 0.0320, ..., -0.0575, 0.0670, 0.0252], + [-0.0137, -0.0422, 0.0084, ..., -0.0148, -0.1073, -0.0474], + [-0.0147, 0.0028, -0.0026, ..., -0.0923, -0.0733, 0.0242]], + device='cuda:0'), grad: tensor([[ 2.5705e-05, 1.3500e-05, 4.0600e-09, ..., 1.7202e-04, + 1.0833e-05, 1.3418e-05], + [ 2.3190e-06, 2.9892e-05, 1.2078e-09, ..., 4.8541e-06, + 2.2933e-05, 2.0310e-05], + [-8.1062e-06, 2.8312e-06, 4.2201e-10, ..., -2.6613e-05, + 3.9600e-06, -1.7926e-05], + ..., + [ 4.2468e-06, -6.2406e-05, 1.6007e-09, ..., 7.7114e-06, + -6.0201e-05, -2.5183e-05], + [-8.2329e-06, 4.0494e-06, 1.0230e-08, ..., 2.0444e-05, + 3.9712e-06, 9.5665e-06], + [ 9.5516e-06, 2.1001e-07, 4.0018e-09, ..., 7.9200e-06, + 7.4878e-06, -1.6451e-05]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0265, 0.0133, 0.0134, 0.0293, 0.0230, -0.0189, 0.0235, 0.0268, + -0.0081, 0.0087], device='cuda:0'), grad: tensor([ 3.6287e-04, 3.9995e-05, -4.5598e-05, -1.0654e-05, 1.8820e-05, + 3.2216e-05, -4.0507e-04, -3.9786e-05, 2.3305e-05, 2.4259e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 76, time 248.78, cls_loss 0.0039 cls_loss_mapping 0.0100 cls_loss_causal 0.6054 re_mapping 0.0087 re_causal 0.0261 /// teacc 98.98 lr 0.00010000 +Epoch 78, weight, value: tensor([[ 0.0391, -0.0768, -0.0957, ..., -0.0544, -0.0399, -0.0448], + [-0.0057, 0.0474, -0.0266, ..., -0.0008, 0.1098, 0.0213], + [-0.0103, -0.0894, -0.0176, ..., 0.0834, -0.0710, 0.0057], + ..., + [-0.0320, 0.0777, 0.0316, ..., -0.0579, 0.0674, 0.0251], + [-0.0136, -0.0424, 0.0081, ..., -0.0150, -0.1079, -0.0481], + [-0.0138, 0.0032, -0.0018, ..., -0.0927, -0.0733, 0.0247]], + device='cuda:0'), grad: tensor([[-1.4149e-05, 1.3355e-06, 3.2538e-08, ..., 9.7603e-06, + 1.2787e-06, 4.9174e-06], + [-7.5810e-06, -6.0303e-07, 4.3394e-08, ..., 3.5134e-07, + -1.1459e-05, -2.4047e-06], + [ 4.1872e-06, 1.2338e-05, -3.1013e-07, ..., 1.0146e-07, + 1.1280e-05, 8.3447e-06], + ..., + [ 2.9709e-06, -5.6654e-05, 4.3685e-08, ..., 1.9073e-06, + -3.8177e-05, -3.5524e-05], + [ 4.2990e-06, 1.6764e-05, 6.0303e-08, ..., 4.9584e-06, + 1.3612e-05, 1.3977e-05], + [ 5.5833e-07, 1.2860e-05, -8.5129e-09, ..., 7.5549e-06, + 1.1228e-05, 8.2254e-06]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0268, 0.0130, 0.0139, 0.0289, 0.0227, -0.0189, 0.0235, 0.0263, + -0.0078, 0.0093], device='cuda:0'), grad: tensor([-1.2696e-05, -4.9025e-06, 2.7835e-05, 1.4201e-05, -1.6296e-04, + 2.7448e-05, 1.0037e-04, -7.9155e-05, 4.8339e-05, 4.1544e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 249.15, cls_loss 0.0035 cls_loss_mapping 0.0095 cls_loss_causal 0.6262 re_mapping 0.0081 re_causal 0.0259 /// teacc 99.04 lr 0.00010000 +Epoch 79, weight, value: tensor([[ 0.0391, -0.0776, -0.0967, ..., -0.0548, -0.0405, -0.0454], + [-0.0054, 0.0476, -0.0268, ..., -0.0010, 0.1105, 0.0214], + [-0.0099, -0.0900, -0.0174, ..., 0.0838, -0.0716, 0.0059], + ..., + [-0.0325, 0.0782, 0.0321, ..., -0.0582, 0.0678, 0.0253], + [-0.0134, -0.0426, 0.0082, ..., -0.0154, -0.1084, -0.0485], + [-0.0137, 0.0030, -0.0018, ..., -0.0931, -0.0738, 0.0238]], + device='cuda:0'), grad: tensor([[-2.4941e-06, 4.4778e-06, 1.1598e-08, ..., 3.0771e-06, + 4.1686e-06, 7.7039e-06], + [-3.3323e-06, -1.7555e-06, 2.0184e-08, ..., 3.7160e-06, + -2.0564e-05, -6.4541e-07], + [ 1.1712e-05, 4.6864e-06, 2.4433e-08, ..., -1.8582e-05, + 7.9125e-06, -1.4208e-05], + ..., + [-8.1003e-05, -1.5163e-04, 3.7340e-08, ..., 5.4948e-06, + -1.0788e-04, -1.7023e-04], + [-7.0296e-06, 2.4140e-06, 4.4965e-08, ..., 3.3639e-06, + 3.9861e-06, 6.2138e-06], + [ 1.3396e-05, 1.2875e-05, -1.2561e-07, ..., 1.3635e-06, + 1.1317e-05, 1.9446e-05]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0270, 0.0130, 0.0146, 0.0289, 0.0239, -0.0193, 0.0238, 0.0263, + -0.0082, 0.0086], device='cuda:0'), grad: tensor([ 5.2713e-06, 1.3582e-05, 3.3170e-05, -1.1832e-04, 8.7097e-06, + 3.4857e-04, 7.3425e-06, -3.6216e-04, -1.4074e-05, 7.7367e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 248.95, cls_loss 0.0038 cls_loss_mapping 0.0118 cls_loss_causal 0.5950 re_mapping 0.0086 re_causal 0.0253 /// teacc 98.83 lr 0.00010000 +Epoch 80, weight, value: tensor([[ 0.0396, -0.0780, -0.0972, ..., -0.0550, -0.0410, -0.0457], + [-0.0047, 0.0473, -0.0269, ..., -0.0015, 0.1111, 0.0215], + [-0.0103, -0.0907, -0.0176, ..., 0.0844, -0.0724, 0.0060], + ..., + [-0.0329, 0.0787, 0.0320, ..., -0.0590, 0.0680, 0.0250], + [-0.0131, -0.0428, 0.0080, ..., -0.0143, -0.1090, -0.0491], + [-0.0138, 0.0030, -0.0017, ..., -0.0936, -0.0741, 0.0241]], + device='cuda:0'), grad: tensor([[ 7.6648e-07, 1.8356e-06, 0.0000e+00, ..., 8.8289e-06, + 1.6931e-06, 6.3069e-06], + [-4.3004e-07, 9.8571e-06, 0.0000e+00, ..., 3.2950e-06, + 8.7824e-07, 6.9402e-06], + [ 3.5428e-06, 4.9770e-06, 0.0000e+00, ..., -1.2052e-04, + 7.5772e-06, -4.8369e-05], + ..., + [ 1.0617e-05, -5.3525e-05, 0.0000e+00, ..., 4.0382e-05, + -5.7846e-05, -1.2033e-05], + [-1.4096e-05, -5.9092e-07, 0.0000e+00, ..., 1.0349e-05, + 2.1812e-06, 2.5824e-05], + [-7.0482e-06, -5.2117e-06, 0.0000e+00, ..., 2.3656e-06, + 9.2313e-06, -5.9992e-05]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0262, 0.0132, 0.0143, 0.0285, 0.0241, -0.0189, 0.0224, 0.0259, + -0.0073, 0.0086], device='cuda:0'), grad: tensor([ 2.6703e-05, 3.9399e-05, -1.9741e-04, -2.2605e-05, 6.4433e-05, + 1.5092e-04, -2.0996e-05, 4.3541e-05, -2.8196e-07, -8.4102e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 249.18, cls_loss 0.0028 cls_loss_mapping 0.0083 cls_loss_causal 0.5624 re_mapping 0.0084 re_causal 0.0251 /// teacc 99.06 lr 0.00010000 +Epoch 81, weight, value: tensor([[ 0.0398, -0.0784, -0.0974, ..., -0.0559, -0.0415, -0.0462], + [-0.0045, 0.0482, -0.0269, ..., -0.0020, 0.1122, 0.0219], + [-0.0104, -0.0911, -0.0176, ..., 0.0853, -0.0728, 0.0069], + ..., + [-0.0335, 0.0784, 0.0319, ..., -0.0596, 0.0676, 0.0247], + [-0.0132, -0.0430, 0.0080, ..., -0.0150, -0.1093, -0.0497], + [-0.0134, 0.0032, -0.0015, ..., -0.0943, -0.0744, 0.0244]], + device='cuda:0'), grad: tensor([[ 7.8827e-06, 3.1758e-06, 8.5216e-08, ..., 8.2031e-06, + 8.4490e-06, 8.1882e-06], + [-1.0264e-04, -3.4392e-05, 2.3330e-07, ..., -6.1020e-06, + -1.3876e-04, -6.7830e-05], + [ 1.4193e-05, 4.4070e-06, 1.2328e-07, ..., -3.1609e-06, + 2.1756e-05, 9.0003e-06], + ..., + [ 1.4745e-05, 6.6832e-06, 4.9733e-07, ..., 2.0582e-06, + 1.0900e-05, 1.3649e-05], + [ 6.3181e-05, 1.9461e-05, 3.3504e-07, ..., 5.6028e-06, + 6.1750e-05, 3.9965e-05], + [-2.2620e-05, -2.3264e-06, -8.4797e-07, ..., 1.7285e-06, + 1.2875e-05, -6.1333e-05]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0264, 0.0134, 0.0145, 0.0288, 0.0237, -0.0195, 0.0235, 0.0253, + -0.0076, 0.0090], device='cuda:0'), grad: tensor([ 2.8774e-05, -2.3544e-04, 3.8564e-05, -6.1356e-06, 4.1842e-05, + 1.7777e-05, 2.8580e-05, 5.4598e-05, 1.6439e-04, -1.3316e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 249.02, cls_loss 0.0031 cls_loss_mapping 0.0095 cls_loss_causal 0.5711 re_mapping 0.0081 re_causal 0.0250 /// teacc 98.98 lr 0.00010000 +Epoch 82, weight, value: tensor([[ 0.0400, -0.0786, -0.0978, ..., -0.0561, -0.0420, -0.0466], + [-0.0044, 0.0482, -0.0269, ..., -0.0022, 0.1127, 0.0220], + [-0.0108, -0.0916, -0.0177, ..., 0.0856, -0.0736, 0.0071], + ..., + [-0.0336, 0.0789, 0.0321, ..., -0.0602, 0.0681, 0.0248], + [-0.0129, -0.0430, 0.0080, ..., -0.0151, -0.1097, -0.0502], + [-0.0134, 0.0032, -0.0015, ..., -0.0949, -0.0747, 0.0250]], + device='cuda:0'), grad: tensor([[ 1.0645e-04, 2.0955e-06, 0.0000e+00, ..., 7.1563e-06, + 1.3717e-05, 4.4517e-06], + [ 1.1044e-03, 3.2056e-06, 0.0000e+00, ..., 1.5404e-06, + 1.0103e-04, 5.3197e-05], + [ 7.9453e-05, 3.5353e-06, 0.0000e+00, ..., -6.8173e-06, + 3.5614e-05, 6.0275e-06], + ..., + [ 6.5863e-06, -2.4348e-05, 0.0000e+00, ..., 1.2610e-06, + -9.3803e-06, -4.4107e-06], + [-1.5059e-03, 9.9745e-07, 0.0000e+00, ..., 1.8729e-06, + -1.4186e-04, 6.3814e-06], + [ 3.5048e-05, 5.4166e-06, 0.0000e+00, ..., 3.4533e-06, + 2.2471e-05, 2.1681e-05]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0264, 0.0132, 0.0139, 0.0289, 0.0236, -0.0194, 0.0237, 0.0253, + -0.0075, 0.0092], device='cuda:0'), grad: tensor([ 2.7251e-04, 2.7847e-03, 2.0945e-04, -6.2823e-05, -1.8847e-04, + 1.6427e-04, 2.8896e-04, 1.2271e-05, -3.6163e-03, 1.3661e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 81, time 249.12, cls_loss 0.0031 cls_loss_mapping 0.0089 cls_loss_causal 0.5856 re_mapping 0.0081 re_causal 0.0251 /// teacc 99.07 lr 0.00010000 +Epoch 83, weight, value: tensor([[ 0.0404, -0.0786, -0.0983, ..., -0.0568, -0.0425, -0.0470], + [-0.0042, 0.0483, -0.0270, ..., -0.0025, 0.1134, 0.0222], + [-0.0113, -0.0922, -0.0177, ..., 0.0860, -0.0744, 0.0069], + ..., + [-0.0336, 0.0800, 0.0320, ..., -0.0604, 0.0687, 0.0259], + [-0.0128, -0.0434, 0.0079, ..., -0.0154, -0.1105, -0.0522], + [-0.0131, 0.0025, -0.0015, ..., -0.0953, -0.0760, 0.0252]], + device='cuda:0'), grad: tensor([[ 6.7316e-06, 3.1199e-06, 2.7311e-07, ..., 5.3570e-06, + 1.5438e-05, 5.6252e-06], + [-1.3256e-04, 6.3553e-06, 1.3842e-07, ..., -3.6299e-05, + -2.1982e-04, -2.0251e-05], + [ 9.1374e-05, 8.1882e-06, 2.5816e-06, ..., 1.8761e-05, + 8.1003e-05, 3.1799e-05], + ..., + [ 9.8050e-06, -4.1693e-05, 2.5204e-08, ..., 2.6580e-06, + -2.7716e-05, -2.2680e-05], + [-1.9491e-05, -2.7865e-06, -4.9993e-06, ..., 1.4581e-05, + 9.2387e-05, -1.7524e-05], + [ 4.2319e-05, 6.5081e-06, 1.2489e-06, ..., 9.6336e-06, + 9.6709e-06, 1.3247e-05]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0264, 0.0131, 0.0129, 0.0292, 0.0231, -0.0192, 0.0235, 0.0260, + -0.0075, 0.0094], device='cuda:0'), grad: tensor([ 1.6555e-05, -4.1604e-04, 3.1257e-04, 3.2449e-04, -3.4988e-05, + -3.2401e-04, 6.9976e-05, -1.9476e-05, -6.1989e-05, 1.3340e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 248.91, cls_loss 0.0027 cls_loss_mapping 0.0089 cls_loss_causal 0.5771 re_mapping 0.0080 re_causal 0.0242 /// teacc 98.97 lr 0.00010000 +Epoch 84, weight, value: tensor([[ 0.0401, -0.0793, -0.1003, ..., -0.0578, -0.0433, -0.0475], + [-0.0040, 0.0484, -0.0274, ..., -0.0030, 0.1140, 0.0221], + [-0.0114, -0.0928, -0.0185, ..., 0.0871, -0.0748, 0.0073], + ..., + [-0.0337, 0.0805, 0.0323, ..., -0.0610, 0.0691, 0.0261], + [-0.0126, -0.0436, 0.0087, ..., -0.0160, -0.1114, -0.0528], + [-0.0129, 0.0024, -0.0011, ..., -0.0960, -0.0764, 0.0252]], + device='cuda:0'), grad: tensor([[ 8.0676e-08, 5.2303e-06, 1.1676e-07, ..., 3.2373e-06, + 2.0154e-06, 1.0222e-05], + [ 1.7332e-06, 1.0920e-04, 6.6683e-07, ..., 7.2904e-06, + 1.1253e-04, 9.9480e-05], + [-5.8971e-06, 2.1607e-05, -2.3711e-06, ..., -1.6972e-05, + 2.1487e-05, 3.7216e-06], + ..., + [ 1.1772e-05, -1.1408e-04, 3.0524e-07, ..., 2.5377e-05, + -1.5390e-04, -3.7044e-05], + [ 7.9302e-07, 1.2122e-05, 9.5461e-07, ..., 4.3027e-06, + 3.7290e-06, 3.1590e-05], + [-6.5148e-05, -1.5020e-04, 2.3734e-08, ..., -4.1753e-05, + 4.2170e-06, -3.5644e-04]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0272, 0.0130, 0.0129, 0.0293, 0.0233, -0.0194, 0.0239, 0.0261, + -0.0075, 0.0093], device='cuda:0'), grad: tensor([ 2.4185e-05, 1.7929e-04, 1.3791e-05, 5.3383e-06, 6.9714e-04, + 6.9141e-05, 4.7803e-05, 3.1628e-06, 3.2961e-05, -1.0729e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 249.16, cls_loss 0.0032 cls_loss_mapping 0.0113 cls_loss_causal 0.6205 re_mapping 0.0078 re_causal 0.0245 /// teacc 98.88 lr 0.00010000 +Epoch 85, weight, value: tensor([[ 0.0403, -0.0799, -0.1031, ..., -0.0583, -0.0440, -0.0479], + [-0.0040, 0.0483, -0.0275, ..., -0.0029, 0.1151, 0.0226], + [-0.0116, -0.0937, -0.0188, ..., 0.0877, -0.0768, 0.0068], + ..., + [-0.0341, 0.0809, 0.0321, ..., -0.0615, 0.0694, 0.0263], + [-0.0123, -0.0435, 0.0093, ..., -0.0165, -0.1117, -0.0532], + [-0.0130, 0.0024, -0.0012, ..., -0.0965, -0.0767, 0.0251]], + device='cuda:0'), grad: tensor([[ 2.5965e-06, 4.3982e-07, 1.6380e-07, ..., 6.6459e-06, + 1.8664e-06, 3.8054e-06], + [ 4.4494e-07, -1.4640e-06, 6.6182e-08, ..., 5.4091e-06, + -1.5330e-04, -8.8036e-05], + [ 7.7486e-06, 7.2969e-07, 1.1170e-07, ..., -8.6203e-06, + 6.0380e-05, 2.9564e-05], + ..., + [ 4.3251e-06, -2.6803e-06, 3.2317e-07, ..., 1.7695e-06, + 2.6062e-05, 1.9774e-05], + [-2.6017e-05, 1.7975e-06, 9.0664e-07, ..., -2.8405e-06, + 6.6012e-06, 1.4186e-05], + [-1.3590e-05, -2.7101e-06, -2.7362e-06, ..., 1.2713e-06, + 3.3993e-06, -2.2069e-05]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0271, 0.0131, 0.0124, 0.0289, 0.0237, -0.0191, 0.0243, 0.0261, + -0.0073, 0.0089], device='cuda:0'), grad: tensor([ 2.2158e-05, -1.6177e-04, 9.5785e-05, 3.7462e-05, 4.7892e-05, + 4.8459e-05, 2.6580e-06, 4.9621e-05, -7.6413e-05, -6.6161e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 249.06, cls_loss 0.0031 cls_loss_mapping 0.0100 cls_loss_causal 0.5811 re_mapping 0.0077 re_causal 0.0238 /// teacc 98.98 lr 0.00010000 +Epoch 86, weight, value: tensor([[ 0.0407, -0.0795, -0.1035, ..., -0.0584, -0.0448, -0.0483], + [-0.0037, 0.0483, -0.0275, ..., -0.0033, 0.1156, 0.0224], + [-0.0115, -0.0944, -0.0188, ..., 0.0887, -0.0774, 0.0078], + ..., + [-0.0344, 0.0817, 0.0322, ..., -0.0620, 0.0700, 0.0267], + [-0.0120, -0.0437, 0.0093, ..., -0.0168, -0.1121, -0.0536], + [-0.0129, 0.0025, -0.0011, ..., -0.0968, -0.0773, 0.0252]], + device='cuda:0'), grad: tensor([[-6.7726e-06, 7.8380e-06, 2.3504e-07, ..., 6.1870e-05, + 5.2713e-06, 2.7120e-05], + [-7.9930e-05, -4.3035e-05, 5.1558e-06, ..., 7.8306e-06, + -1.5724e-04, -1.9982e-05], + [-2.9102e-05, 2.4647e-05, 9.8627e-07, ..., -1.6558e-04, + 2.3082e-05, -2.7120e-05], + ..., + [ 5.9605e-05, 5.4866e-05, -1.7732e-05, ..., 1.2167e-05, + 2.7120e-05, 1.1849e-04], + [-1.0088e-05, 1.3426e-05, 2.5681e-07, ..., 2.8700e-05, + 1.7300e-05, 3.0324e-05], + [ 1.4612e-06, 3.2604e-05, 4.0047e-06, ..., 9.7379e-06, + 6.4552e-05, 4.2990e-06]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0262, 0.0128, 0.0130, 0.0284, 0.0235, -0.0191, 0.0242, 0.0264, + -0.0073, 0.0088], device='cuda:0'), grad: tensor([ 1.3363e-04, -3.7163e-05, -2.7728e-04, -1.5039e-03, 9.0778e-05, + 7.3385e-04, 8.6188e-05, 5.6839e-04, 8.3089e-05, 1.2290e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 249.01, cls_loss 0.0035 cls_loss_mapping 0.0112 cls_loss_causal 0.5999 re_mapping 0.0077 re_causal 0.0236 /// teacc 98.81 lr 0.00010000 +Epoch 87, weight, value: tensor([[ 0.0413, -0.0803, -0.1043, ..., -0.0591, -0.0455, -0.0491], + [-0.0034, 0.0483, -0.0276, ..., -0.0043, 0.1149, 0.0210], + [-0.0114, -0.0948, -0.0180, ..., 0.0889, -0.0801, 0.0062], + ..., + [-0.0345, 0.0821, 0.0320, ..., -0.0603, 0.0725, 0.0285], + [-0.0127, -0.0439, 0.0090, ..., -0.0180, -0.1128, -0.0540], + [-0.0125, 0.0030, -0.0010, ..., -0.0974, -0.0778, 0.0260]], + device='cuda:0'), grad: tensor([[ 2.0653e-05, 1.6089e-07, 1.8002e-06, ..., 3.3289e-05, + 2.4331e-07, 1.6123e-05], + [ 2.1353e-05, 6.7661e-07, 1.9360e-07, ..., 4.4435e-05, + 2.8461e-06, 1.3165e-05], + [-4.4942e-05, 1.1356e-07, -4.5784e-06, ..., -1.2338e-04, + -1.7077e-05, -1.1760e-04], + ..., + [ 8.2254e-06, -3.2000e-06, 2.8987e-07, ..., 4.7266e-05, + 1.0915e-05, 4.8906e-05], + [ 4.3437e-06, 7.9744e-08, 1.2256e-06, ..., 3.0115e-05, + 1.6147e-07, 1.7956e-05], + [ 1.6466e-05, 1.2238e-06, 1.8463e-07, ..., 1.0915e-05, + 1.4575e-06, 1.8790e-05]], device='cuda:0') +Epoch 87, bias, value: tensor([-0.0262, 0.0118, 0.0121, 0.0285, 0.0229, -0.0196, 0.0250, 0.0278, + -0.0081, 0.0098], device='cuda:0'), grad: tensor([ 9.8109e-05, 1.8489e-04, -2.3270e-04, 2.9743e-05, -5.6684e-05, + 1.0711e-04, -3.8528e-04, 8.3089e-05, 7.4923e-05, 9.7156e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 249.03, cls_loss 0.0025 cls_loss_mapping 0.0066 cls_loss_causal 0.5727 re_mapping 0.0075 re_causal 0.0231 /// teacc 98.96 lr 0.00010000 +Epoch 88, weight, value: tensor([[ 0.0416, -0.0809, -0.1053, ..., -0.0599, -0.0461, -0.0497], + [-0.0033, 0.0484, -0.0276, ..., -0.0046, 0.1153, 0.0210], + [-0.0110, -0.0959, -0.0178, ..., 0.0898, -0.0803, 0.0067], + ..., + [-0.0348, 0.0827, 0.0317, ..., -0.0606, 0.0727, 0.0288], + [-0.0127, -0.0440, 0.0090, ..., -0.0185, -0.1131, -0.0546], + [-0.0127, 0.0030, -0.0010, ..., -0.0987, -0.0782, 0.0259]], + device='cuda:0'), grad: tensor([[-4.0352e-05, 3.7206e-07, 3.0355e-08, ..., -8.2776e-06, + 5.8068e-07, 2.9802e-06], + [-2.4354e-07, -3.0287e-06, 5.3726e-08, ..., 1.5497e-06, + -8.2776e-06, -3.8208e-07], + [ 1.3560e-05, 4.3144e-07, 3.0175e-07, ..., 1.1269e-06, + 9.1922e-07, -3.6620e-06], + ..., + [ 6.7204e-06, 1.1651e-06, 2.8289e-08, ..., 4.6082e-06, + 2.4810e-06, 6.9849e-06], + [-1.9278e-06, 3.0026e-06, -8.0094e-07, ..., -1.7241e-05, + 1.6410e-06, 1.1988e-05], + [ 1.6525e-05, -3.9861e-06, 1.0029e-07, ..., 7.6368e-06, + 5.1502e-07, -2.9773e-05]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0262, 0.0117, 0.0122, 0.0282, 0.0234, -0.0192, 0.0250, 0.0280, + -0.0083, 0.0092], device='cuda:0'), grad: tensor([-1.0383e-04, 4.7833e-06, 4.3511e-05, -7.6652e-05, 3.1233e-05, + 1.3225e-05, 3.2187e-05, 3.6001e-05, 6.0529e-05, -4.1038e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 249.23, cls_loss 0.0028 cls_loss_mapping 0.0081 cls_loss_causal 0.5744 re_mapping 0.0069 re_causal 0.0221 /// teacc 98.92 lr 0.00010000 +Epoch 89, weight, value: tensor([[ 0.0416, -0.0819, -0.1057, ..., -0.0605, -0.0469, -0.0503], + [-0.0029, 0.0486, -0.0277, ..., -0.0048, 0.1160, 0.0211], + [-0.0109, -0.0962, -0.0177, ..., 0.0909, -0.0806, 0.0073], + ..., + [-0.0352, 0.0831, 0.0318, ..., -0.0613, 0.0728, 0.0287], + [-0.0125, -0.0442, 0.0091, ..., -0.0189, -0.1136, -0.0553], + [-0.0128, 0.0031, -0.0012, ..., -0.0997, -0.0786, 0.0259]], + device='cuda:0'), grad: tensor([[ 5.9530e-06, 2.7958e-06, 0.0000e+00, ..., 2.6807e-05, + 3.7439e-06, 7.8753e-06], + [-1.9521e-05, -2.1588e-06, 0.0000e+00, ..., 2.8685e-06, + -1.8024e-04, -9.1016e-05], + [-2.7508e-05, 2.4334e-05, 0.0000e+00, ..., -1.2957e-05, + 6.6221e-05, 2.1458e-05], + ..., + [ 1.9580e-05, -4.3124e-05, 0.0000e+00, ..., 1.1951e-05, + 5.5075e-05, 7.1116e-06], + [ 1.3493e-05, 4.8578e-06, 0.0000e+00, ..., 3.4243e-05, + 6.2883e-06, 1.6004e-05], + [-6.2585e-06, -3.6582e-06, 0.0000e+00, ..., 1.0535e-05, + 4.9286e-06, -3.5651e-06]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0267, 0.0119, 0.0128, 0.0283, 0.0237, -0.0194, 0.0252, 0.0276, + -0.0084, 0.0091], device='cuda:0'), grad: tensor([ 5.3525e-05, -1.8120e-04, 4.6223e-05, -2.5332e-05, 1.0866e-04, + 5.1856e-05, -2.1470e-04, 7.3493e-05, 8.3029e-05, 4.4517e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 248.72, cls_loss 0.0028 cls_loss_mapping 0.0098 cls_loss_causal 0.5497 re_mapping 0.0078 re_causal 0.0227 /// teacc 98.83 lr 0.00010000 +Epoch 90, weight, value: tensor([[ 0.0417, -0.0826, -0.1061, ..., -0.0611, -0.0481, -0.0509], + [-0.0024, 0.0486, -0.0278, ..., -0.0051, 0.1169, 0.0214], + [-0.0111, -0.0971, -0.0177, ..., 0.0914, -0.0812, 0.0074], + ..., + [-0.0356, 0.0838, 0.0317, ..., -0.0617, 0.0731, 0.0289], + [-0.0121, -0.0444, 0.0092, ..., -0.0196, -0.1141, -0.0558], + [-0.0126, 0.0030, -0.0012, ..., -0.1004, -0.0792, 0.0260]], + device='cuda:0'), grad: tensor([[ 2.2650e-06, 1.4585e-06, 6.2818e-07, ..., 2.8592e-06, + 1.8487e-06, 4.1313e-06], + [ 4.3847e-06, 1.4871e-05, 7.7263e-06, ..., 3.3770e-06, + 1.9446e-05, 2.0415e-05], + [ 2.7880e-05, 1.9252e-05, 1.1280e-05, ..., -1.8477e-05, + 2.7850e-05, -5.8562e-06], + ..., + [ 5.3309e-06, -6.1929e-05, -2.3633e-05, ..., 1.1981e-05, + -9.1255e-05, -5.3912e-05], + [ 1.5929e-05, 3.1050e-06, 2.3115e-06, ..., 4.9137e-06, + 4.5225e-06, 9.1940e-06], + [ 1.3381e-05, 1.4147e-06, 1.1455e-06, ..., 6.5705e-07, + 5.8040e-06, -1.0371e-05]], device='cuda:0') +Epoch 90, bias, value: tensor([-0.0270, 0.0121, 0.0125, 0.0280, 0.0239, -0.0191, 0.0251, 0.0278, + -0.0085, 0.0091], device='cuda:0'), grad: tensor([ 2.1145e-05, 6.7234e-05, 6.5923e-05, -4.1604e-04, 5.7101e-05, + 2.5153e-04, -3.7849e-05, -9.6619e-05, 7.6592e-05, 1.1273e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 249.14, cls_loss 0.0041 cls_loss_mapping 0.0133 cls_loss_causal 0.5777 re_mapping 0.0074 re_causal 0.0222 /// teacc 99.01 lr 0.00010000 +Epoch 91, weight, value: tensor([[ 0.0419, -0.0832, -0.1067, ..., -0.0616, -0.0490, -0.0519], + [-0.0022, 0.0494, -0.0280, ..., -0.0061, 0.1183, 0.0220], + [-0.0106, -0.0977, -0.0176, ..., 0.0933, -0.0816, 0.0085], + ..., + [-0.0359, 0.0834, 0.0319, ..., -0.0620, 0.0726, 0.0286], + [-0.0121, -0.0444, 0.0093, ..., -0.0201, -0.1149, -0.0564], + [-0.0125, 0.0028, -0.0011, ..., -0.1016, -0.0800, 0.0253]], + device='cuda:0'), grad: tensor([[ 3.3062e-06, 1.6820e-06, 1.7171e-08, ..., 3.9637e-06, + 2.1067e-06, 4.3996e-06], + [-3.5018e-05, -7.3493e-05, 7.5903e-08, ..., 4.1351e-06, + -1.1212e-04, -4.0084e-05], + [-8.7395e-06, 6.1467e-06, 1.6589e-08, ..., -2.8223e-05, + 1.1325e-05, -7.1377e-06], + ..., + [ 2.6301e-05, 2.7314e-05, 7.5204e-08, ..., 5.4911e-06, + 4.7922e-05, 2.0564e-05], + [ 2.7996e-06, 2.8908e-06, 6.8860e-08, ..., 2.0564e-05, + 4.5672e-06, 1.3046e-05], + [ 9.4175e-06, 4.8168e-06, 1.2596e-07, ..., 1.4186e-05, + 7.5325e-06, -4.5486e-06]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0273, 0.0125, 0.0140, 0.0282, 0.0243, -0.0192, 0.0251, 0.0272, + -0.0087, 0.0083], device='cuda:0'), grad: tensor([ 5.6662e-06, -1.0991e-04, -1.7345e-05, -4.5747e-05, -3.2157e-05, + -7.8559e-05, 6.1393e-05, 1.0675e-04, 2.6986e-05, 8.2970e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 248.82, cls_loss 0.0031 cls_loss_mapping 0.0092 cls_loss_causal 0.6076 re_mapping 0.0073 re_causal 0.0222 /// teacc 98.95 lr 0.00010000 +Epoch 92, weight, value: tensor([[ 0.0426, -0.0838, -0.1058, ..., -0.0619, -0.0500, -0.0528], + [-0.0009, 0.0498, -0.0285, ..., -0.0064, 0.1190, 0.0226], + [-0.0112, -0.0981, -0.0175, ..., 0.0937, -0.0821, 0.0085], + ..., + [-0.0369, 0.0835, 0.0331, ..., -0.0625, 0.0725, 0.0285], + [-0.0119, -0.0443, 0.0093, ..., -0.0206, -0.1155, -0.0569], + [-0.0123, 0.0023, -0.0013, ..., -0.1017, -0.0811, 0.0257]], + device='cuda:0'), grad: tensor([[ 3.1918e-05, 4.7907e-06, 9.7454e-06, ..., 2.8033e-06, + 2.6878e-06, 7.1228e-05], + [-1.9208e-05, -1.0647e-05, 4.0396e-07, ..., 2.3507e-06, + -5.1737e-05, -7.9572e-06], + [ 2.5228e-05, 7.8902e-06, 5.5395e-06, ..., -8.3223e-06, + 8.7917e-06, 2.9311e-05], + ..., + [ 1.2182e-05, -9.8161e-07, 1.2880e-06, ..., 5.0627e-06, + 1.4305e-06, 6.1058e-06], + [-1.0461e-04, -1.3459e-04, 2.4512e-06, ..., 2.9169e-06, + 2.3857e-05, 2.3514e-05], + [ 3.7737e-06, 8.6665e-05, -2.6897e-05, ..., 4.4145e-06, + 5.5321e-06, -1.9109e-04]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0269, 0.0130, 0.0133, 0.0285, 0.0240, -0.0195, 0.0249, 0.0270, + -0.0087, 0.0087], device='cuda:0'), grad: tensor([ 1.5187e-04, -1.7032e-05, 9.1553e-05, 4.1747e-04, -2.5749e-04, + -2.6941e-04, 3.2830e-04, 4.1127e-05, -4.4084e-04, -4.4972e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 248.99, cls_loss 0.0026 cls_loss_mapping 0.0092 cls_loss_causal 0.5583 re_mapping 0.0075 re_causal 0.0226 /// teacc 98.98 lr 0.00010000 +Epoch 93, weight, value: tensor([[ 0.0429, -0.0840, -0.1100, ..., -0.0628, -0.0504, -0.0532], + [-0.0007, 0.0497, -0.0286, ..., -0.0069, 0.1191, 0.0225], + [-0.0110, -0.0984, -0.0171, ..., 0.0942, -0.0823, 0.0088], + ..., + [-0.0372, 0.0838, 0.0330, ..., -0.0628, 0.0727, 0.0287], + [-0.0117, -0.0442, 0.0089, ..., -0.0208, -0.1159, -0.0572], + [-0.0124, 0.0021, -0.0011, ..., -0.1024, -0.0815, 0.0252]], + device='cuda:0'), grad: tensor([[ 1.5665e-06, 1.1958e-06, 4.3510e-09, ..., 8.7963e-07, + 8.6613e-07, 8.2888e-07], + [ 8.0420e-07, 2.0303e-06, 4.0600e-09, ..., 1.8300e-06, + -1.2098e-06, 1.6894e-06], + [ 3.1814e-06, 3.8184e-06, 1.5323e-08, ..., -2.3115e-06, + 3.0790e-06, -3.4142e-06], + ..., + [ 1.5333e-05, 7.3984e-06, 6.6939e-09, ..., 1.0282e-06, + 3.0212e-06, -1.1204e-06], + [ 3.6154e-06, 1.6578e-06, 2.3007e-08, ..., 1.6289e-06, + 1.5805e-06, 2.2519e-06], + [ 5.7854e-06, 1.6000e-06, 1.7637e-08, ..., 7.5763e-07, + 2.2501e-06, -1.4473e-06]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0272, 0.0128, 0.0133, 0.0283, 0.0247, -0.0191, 0.0253, 0.0270, + -0.0087, 0.0079], device='cuda:0'), grad: tensor([ 1.4506e-05, 1.9416e-05, 2.3142e-05, -1.7786e-04, 9.5740e-06, + -6.8188e-05, 8.1360e-06, 1.1933e-04, 2.3305e-05, 2.8536e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 92, time 248.77, cls_loss 0.0024 cls_loss_mapping 0.0077 cls_loss_causal 0.5595 re_mapping 0.0074 re_causal 0.0232 /// teacc 99.06 lr 0.00010000 +Epoch 94, weight, value: tensor([[ 0.0432, -0.0843, -0.1111, ..., -0.0638, -0.0511, -0.0536], + [-0.0003, 0.0497, -0.0287, ..., -0.0069, 0.1194, 0.0227], + [-0.0112, -0.0995, -0.0164, ..., 0.0950, -0.0834, 0.0084], + ..., + [-0.0374, 0.0842, 0.0328, ..., -0.0633, 0.0731, 0.0290], + [-0.0115, -0.0440, 0.0074, ..., -0.0212, -0.1164, -0.0577], + [-0.0123, 0.0021, -0.0013, ..., -0.1030, -0.0818, 0.0255]], + device='cuda:0'), grad: tensor([[ 1.8964e-07, 1.6326e-06, 1.9339e-08, ..., 1.7472e-06, + 1.7490e-06, 3.1926e-06], + [-1.1520e-06, 1.1124e-05, 5.8062e-09, ..., 3.9265e-06, + 1.0356e-05, 1.3523e-05], + [-1.7891e-06, 2.9244e-06, 3.6089e-09, ..., -1.3843e-05, + 2.6580e-06, -8.1360e-06], + ..., + [ 5.5274e-07, -3.4839e-05, 2.4447e-08, ..., 2.8871e-06, + -5.8502e-05, -1.9297e-05], + [-2.4657e-07, 5.2787e-06, 3.4197e-08, ..., 1.3057e-06, + 3.4664e-06, 8.0019e-06], + [ 2.5914e-07, -5.0291e-06, -2.0454e-07, ..., 7.7672e-07, + 2.2247e-05, -2.0415e-05]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0275, 0.0130, 0.0128, 0.0286, 0.0241, -0.0188, 0.0252, 0.0272, + -0.0085, 0.0078], device='cuda:0'), grad: tensor([ 6.0983e-06, 2.5496e-05, -1.5706e-05, 1.0848e-05, 2.8953e-05, + 2.0340e-06, 7.0855e-06, -3.8147e-05, 1.1228e-05, -3.7998e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 249.10, cls_loss 0.0030 cls_loss_mapping 0.0093 cls_loss_causal 0.6135 re_mapping 0.0067 re_causal 0.0219 /// teacc 98.98 lr 0.00010000 +Epoch 95, weight, value: tensor([[ 0.0426, -0.0851, -0.1120, ..., -0.0636, -0.0520, -0.0560], + [-0.0002, 0.0493, -0.0288, ..., -0.0071, 0.1193, 0.0224], + [-0.0112, -0.1001, -0.0160, ..., 0.0959, -0.0839, 0.0088], + ..., + [-0.0374, 0.0851, 0.0330, ..., -0.0635, 0.0738, 0.0296], + [-0.0113, -0.0444, 0.0076, ..., -0.0216, -0.1173, -0.0585], + [-0.0115, 0.0016, -0.0015, ..., -0.1044, -0.0830, 0.0258]], + device='cuda:0'), grad: tensor([[ 1.2910e-04, 3.2131e-06, 4.4401e-07, ..., 2.6464e-05, + 3.0566e-06, 2.9415e-05], + [ 1.0237e-05, 4.9770e-06, 5.4808e-07, ..., 3.6396e-06, + 4.1276e-06, 1.3031e-05], + [ 5.7369e-05, 2.7455e-06, 1.7993e-06, ..., -1.1861e-05, + 2.2370e-06, -7.7039e-06], + ..., + [ 2.8327e-05, 8.6129e-05, 1.3001e-06, ..., 1.4566e-05, + 1.0550e-04, 1.5223e-04], + [-5.4312e-04, 2.2948e-06, 1.5777e-06, ..., -1.1837e-04, + 1.7975e-06, -1.0729e-04], + [ 3.2806e-04, -1.2660e-04, 5.2527e-07, ..., 7.5519e-05, + -1.5283e-04, -3.9846e-05]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0284, 0.0127, 0.0130, 0.0282, 0.0242, -0.0189, 0.0251, 0.0278, + -0.0086, 0.0082], device='cuda:0'), grad: tensor([ 4.9782e-04, 6.2406e-05, 1.9562e-04, 3.0971e-04, -3.8087e-05, + -3.7432e-04, 5.6535e-05, 4.2200e-04, -2.0847e-03, 9.5415e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 248.93, cls_loss 0.0026 cls_loss_mapping 0.0089 cls_loss_causal 0.5930 re_mapping 0.0073 re_causal 0.0223 /// teacc 98.97 lr 0.00010000 +Epoch 96, weight, value: tensor([[ 0.0423, -0.0855, -0.1130, ..., -0.0638, -0.0524, -0.0566], + [ 0.0001, 0.0495, -0.0289, ..., -0.0073, 0.1196, 0.0225], + [-0.0114, -0.1005, -0.0159, ..., 0.0963, -0.0841, 0.0090], + ..., + [-0.0378, 0.0853, 0.0329, ..., -0.0639, 0.0739, 0.0296], + [-0.0110, -0.0447, 0.0074, ..., -0.0218, -0.1179, -0.0590], + [-0.0107, 0.0014, -0.0015, ..., -0.1044, -0.0835, 0.0258]], + device='cuda:0'), grad: tensor([[ 1.9725e-06, 4.1910e-06, 0.0000e+00, ..., 3.4869e-06, + 4.3400e-06, 6.5416e-06], + [-1.0684e-05, -1.7425e-06, 0.0000e+00, ..., 1.2005e-06, + -2.2128e-05, 2.6315e-05], + [-6.1393e-06, 1.7047e-05, 0.0000e+00, ..., -1.3985e-05, + 1.5453e-05, 2.2709e-05], + ..., + [ 4.0196e-06, -1.0127e-04, 0.0000e+00, ..., 3.5241e-06, + -7.6294e-05, -8.2655e-07], + [ 4.0643e-06, 8.1509e-06, 0.0000e+00, ..., 5.3905e-06, + 1.1012e-05, 3.3826e-05], + [ 1.2014e-06, 2.8715e-05, 0.0000e+00, ..., 2.0228e-06, + 2.0683e-05, 1.5688e-04]], device='cuda:0') +Epoch 96, bias, value: tensor([-0.0288, 0.0128, 0.0127, 0.0281, 0.0243, -0.0194, 0.0254, 0.0277, + -0.0083, 0.0086], device='cuda:0'), grad: tensor([ 2.3156e-05, 2.2471e-05, 4.5508e-05, 8.5831e-05, -4.3440e-04, + 3.8087e-05, 2.2650e-06, -1.2279e-04, 7.3433e-05, 2.6631e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 248.55, cls_loss 0.0026 cls_loss_mapping 0.0088 cls_loss_causal 0.5626 re_mapping 0.0069 re_causal 0.0212 /// teacc 98.95 lr 0.00010000 +Epoch 97, weight, value: tensor([[ 0.0422, -0.0859, -0.1130, ..., -0.0644, -0.0530, -0.0569], + [ 0.0005, 0.0497, -0.0289, ..., -0.0073, 0.1201, 0.0228], + [-0.0114, -0.1010, -0.0159, ..., 0.0969, -0.0845, 0.0093], + ..., + [-0.0383, 0.0856, 0.0330, ..., -0.0645, 0.0740, 0.0296], + [-0.0108, -0.0446, 0.0074, ..., -0.0222, -0.1185, -0.0596], + [-0.0111, 0.0013, -0.0016, ..., -0.1052, -0.0840, 0.0257]], + device='cuda:0'), grad: tensor([[-1.4910e-06, 1.9949e-06, 4.6974e-08, ..., 1.7121e-05, + 1.0151e-06, 1.8794e-06], + [ 4.5598e-06, 8.5056e-05, 1.2282e-07, ..., 5.7109e-06, + 1.1539e-04, 7.5519e-05], + [-2.5421e-05, 4.3474e-06, -1.1120e-06, ..., -4.7028e-05, + 3.4347e-06, -5.8085e-05], + ..., + [-5.0776e-06, -1.5485e-04, 2.2730e-08, ..., 1.2061e-06, + -1.9884e-04, -1.2970e-04], + [ 2.8908e-06, 7.7188e-06, 5.3365e-07, ..., 8.3521e-06, + 8.8438e-06, 9.1791e-06], + [ 1.4983e-05, 5.2042e-06, 8.4983e-09, ..., 5.2333e-05, + 1.0803e-05, 5.2065e-05]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0294, 0.0131, 0.0127, 0.0279, 0.0246, -0.0184, 0.0255, 0.0275, + -0.0083, 0.0080], device='cuda:0'), grad: tensor([ 1.8671e-05, 1.6391e-04, -8.9645e-05, 4.1962e-05, 2.3496e-04, + 2.7165e-05, -2.4891e-04, -2.7013e-04, 2.9758e-05, 9.2506e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 248.99, cls_loss 0.0027 cls_loss_mapping 0.0076 cls_loss_causal 0.5660 re_mapping 0.0067 re_causal 0.0207 /// teacc 99.01 lr 0.00010000 +Epoch 98, weight, value: tensor([[ 0.0427, -0.0864, -0.1131, ..., -0.0651, -0.0537, -0.0573], + [ 0.0003, 0.0494, -0.0289, ..., -0.0075, 0.1200, 0.0226], + [-0.0115, -0.1014, -0.0159, ..., 0.0975, -0.0848, 0.0097], + ..., + [-0.0385, 0.0863, 0.0333, ..., -0.0649, 0.0746, 0.0301], + [-0.0106, -0.0452, 0.0074, ..., -0.0225, -0.1195, -0.0605], + [-0.0108, 0.0011, -0.0016, ..., -0.1069, -0.0842, 0.0254]], + device='cuda:0'), grad: tensor([[ 1.5777e-06, 1.1623e-06, 6.6939e-10, ..., 5.5842e-06, + 1.2033e-06, 4.1276e-06], + [-2.7125e-07, -8.0676e-08, 1.4697e-09, ..., 2.8051e-06, + -9.3877e-06, 4.2166e-07], + [ 1.8463e-05, 1.1194e-06, 9.6188e-09, ..., -4.0792e-06, + 3.6508e-06, -6.4746e-06], + ..., + [ 4.0829e-06, 9.6142e-05, 1.7608e-09, ..., 2.6077e-06, + 4.9949e-05, 1.4770e-04], + [-4.3452e-05, 2.1681e-06, 2.9831e-09, ..., -1.4193e-06, + 5.4948e-06, 6.0685e-06], + [-2.7083e-06, -1.2201e-04, 1.3388e-09, ..., 4.1574e-06, + -6.0350e-05, -1.9848e-04]], device='cuda:0') +Epoch 98, bias, value: tensor([-0.0292, 0.0125, 0.0127, 0.0277, 0.0248, -0.0182, 0.0258, 0.0279, + -0.0085, 0.0078], device='cuda:0'), grad: tensor([ 4.5985e-05, 1.5497e-04, 8.1897e-05, -6.9678e-05, 1.4734e-04, + 1.0103e-04, 1.3530e-04, 3.7646e-04, -5.7030e-04, -4.0197e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 248.72, cls_loss 0.0024 cls_loss_mapping 0.0081 cls_loss_causal 0.5687 re_mapping 0.0070 re_causal 0.0212 /// teacc 98.99 lr 0.00010000 +Epoch 99, weight, value: tensor([[ 0.0428, -0.0870, -0.1132, ..., -0.0658, -0.0547, -0.0576], + [ 0.0006, 0.0493, -0.0290, ..., -0.0077, 0.1202, 0.0224], + [-0.0115, -0.1016, -0.0160, ..., 0.0980, -0.0850, 0.0101], + ..., + [-0.0389, 0.0868, 0.0335, ..., -0.0654, 0.0750, 0.0305], + [-0.0101, -0.0453, 0.0074, ..., -0.0228, -0.1205, -0.0611], + [-0.0107, 0.0006, -0.0017, ..., -0.1080, -0.0849, 0.0254]], + device='cuda:0'), grad: tensor([[ 1.1131e-05, 3.0939e-06, 0.0000e+00, ..., 1.5795e-05, + 4.3474e-06, 5.3644e-06], + [-4.4927e-06, -3.2154e-07, 0.0000e+00, ..., 4.9956e-06, + -9.8422e-06, 3.7327e-06], + [-7.3135e-05, -1.8895e-05, 0.0000e+00, ..., -1.6665e-04, + 1.6009e-06, -3.4064e-05], + ..., + [ 8.6203e-06, -9.4920e-06, 0.0000e+00, ..., 1.2763e-05, + -9.5665e-06, -1.3765e-06], + [ 1.9923e-05, 5.5321e-06, 0.0000e+00, ..., 1.8224e-05, + 8.5011e-06, 7.3947e-06], + [-6.5938e-06, -2.4959e-06, 0.0000e+00, ..., 4.3698e-06, + -4.7348e-06, -1.3031e-05]], device='cuda:0') +Epoch 99, bias, value: tensor([-0.0294, 0.0123, 0.0128, 0.0279, 0.0248, -0.0190, 0.0257, 0.0282, + -0.0077, 0.0076], device='cuda:0'), grad: tensor([ 5.4568e-05, 1.7896e-05, -3.1042e-04, 5.6922e-05, 1.1981e-05, + -4.0904e-06, 3.7432e-05, 3.6359e-05, 1.0055e-04, -1.8794e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 248.79, cls_loss 0.0024 cls_loss_mapping 0.0071 cls_loss_causal 0.5535 re_mapping 0.0071 re_causal 0.0213 /// teacc 98.90 lr 0.00010000 +Epoch 100, weight, value: tensor([[ 0.0434, -0.0876, -0.1133, ..., -0.0659, -0.0554, -0.0579], + [ 0.0007, 0.0492, -0.0290, ..., -0.0080, 0.1203, 0.0222], + [-0.0117, -0.1023, -0.0161, ..., 0.0983, -0.0853, 0.0101], + ..., + [-0.0391, 0.0875, 0.0335, ..., -0.0656, 0.0755, 0.0312], + [-0.0102, -0.0455, 0.0076, ..., -0.0235, -0.1214, -0.0621], + [-0.0110, 0.0002, -0.0017, ..., -0.1097, -0.0859, 0.0255]], + device='cuda:0'), grad: tensor([[-7.2159e-06, 4.5984e-07, 0.0000e+00, ..., 7.2084e-06, + 7.1665e-07, 3.8743e-06], + [-8.8057e-07, 1.9968e-06, 0.0000e+00, ..., 3.6769e-06, + -1.9819e-06, 3.9227e-06], + [ 7.0222e-06, 3.7346e-06, 0.0000e+00, ..., 4.5486e-06, + 5.1409e-06, 1.0394e-05], + ..., + [ 2.4196e-06, -1.0870e-05, 0.0000e+00, ..., 2.4512e-06, + -8.7991e-06, -3.9041e-06], + [ 1.6972e-05, 6.1747e-07, 0.0000e+00, ..., -4.8690e-06, + 8.2562e-07, 6.3241e-05], + [-1.4222e-04, 9.7323e-07, 0.0000e+00, ..., 1.1511e-05, + 9.9372e-07, -3.5167e-04]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0291, 0.0121, 0.0123, 0.0286, 0.0253, -0.0190, 0.0257, 0.0288, + -0.0084, 0.0071], device='cuda:0'), grad: tensor([-1.5587e-05, 1.6525e-05, 4.6462e-05, 1.2824e-06, 7.7486e-04, + 5.6058e-05, -1.9640e-05, 4.5635e-06, 1.3447e-04, -9.9945e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 99---------------------------------------------------- +epoch 99, time 249.36, cls_loss 0.0032 cls_loss_mapping 0.0095 cls_loss_causal 0.5624 re_mapping 0.0068 re_causal 0.0202 /// teacc 99.08 lr 0.00010000 +Epoch 101, weight, value: tensor([[ 0.0455, -0.0882, -0.1134, ..., -0.0651, -0.0561, -0.0590], + [ 0.0004, 0.0492, -0.0290, ..., -0.0083, 0.1205, 0.0219], + [-0.0115, -0.1026, -0.0162, ..., 0.0998, -0.0860, 0.0110], + ..., + [-0.0396, 0.0878, 0.0335, ..., -0.0664, 0.0757, 0.0313], + [-0.0098, -0.0457, 0.0076, ..., -0.0244, -0.1223, -0.0636], + [-0.0112, 0.0002, -0.0017, ..., -0.1116, -0.0854, 0.0258]], + device='cuda:0'), grad: tensor([[-3.6731e-06, 2.8918e-07, 0.0000e+00, ..., 2.5574e-06, + 1.6252e-07, 1.1688e-06], + [-3.2280e-06, -4.1947e-06, 0.0000e+00, ..., 2.7530e-06, + -9.4548e-06, -2.6394e-06], + [-2.0005e-06, -5.3411e-07, 0.0000e+00, ..., -5.2750e-06, + 1.0887e-06, -2.6859e-06], + ..., + [ 2.9746e-06, 1.6922e-06, 0.0000e+00, ..., 1.8068e-06, + 2.7474e-06, 5.1744e-06], + [ 3.6079e-06, 1.4221e-06, 0.0000e+00, ..., 4.6082e-06, + 1.7826e-06, 3.3993e-06], + [ 1.3933e-06, -1.1465e-06, 0.0000e+00, ..., 1.2331e-06, + 1.2666e-06, -1.2904e-05]], device='cuda:0') +Epoch 101, bias, value: tensor([-0.0273, 0.0115, 0.0130, 0.0281, 0.0255, -0.0195, 0.0250, 0.0286, + -0.0084, 0.0072], device='cuda:0'), grad: tensor([-9.2462e-06, -4.7572e-06, -4.6454e-06, -2.4691e-05, 1.1615e-05, + 1.7494e-05, 6.4843e-08, 1.7568e-05, 1.7956e-05, -2.1428e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 248.77, cls_loss 0.0025 cls_loss_mapping 0.0081 cls_loss_causal 0.5323 re_mapping 0.0071 re_causal 0.0207 /// teacc 98.98 lr 0.00010000 +Epoch 102, weight, value: tensor([[ 0.0458, -0.0885, -0.1134, ..., -0.0658, -0.0566, -0.0593], + [ 0.0006, 0.0491, -0.0290, ..., -0.0085, 0.1207, 0.0218], + [-0.0114, -0.1030, -0.0162, ..., 0.1005, -0.0866, 0.0114], + ..., + [-0.0400, 0.0882, 0.0335, ..., -0.0669, 0.0760, 0.0316], + [-0.0096, -0.0463, 0.0076, ..., -0.0245, -0.1231, -0.0653], + [-0.0107, 0.0003, -0.0017, ..., -0.1127, -0.0856, 0.0260]], + device='cuda:0'), grad: tensor([[ 6.7770e-05, 5.7183e-06, 0.0000e+00, ..., 2.3615e-04, + 1.9744e-06, 3.4422e-05], + [-2.1875e-05, -2.0966e-05, 0.0000e+00, ..., 1.1176e-05, + -4.9323e-05, -1.2174e-05], + [ 6.4261e-06, 1.2942e-05, 0.0000e+00, ..., -2.5094e-05, + 7.5176e-06, -3.7793e-06], + ..., + [ 6.1952e-06, -4.9382e-05, 0.0000e+00, ..., 3.1143e-05, + 1.1697e-05, -4.5806e-05], + [ 6.2659e-06, 1.7546e-06, 0.0000e+00, ..., 2.6524e-05, + 3.2037e-06, 9.3058e-06], + [ 2.4080e-05, 4.3362e-05, 0.0000e+00, ..., 1.5050e-05, + 1.6585e-05, 9.4354e-05]], device='cuda:0') +Epoch 102, bias, value: tensor([-0.0275, 0.0115, 0.0131, 0.0287, 0.0253, -0.0202, 0.0252, 0.0287, + -0.0089, 0.0078], device='cuda:0'), grad: tensor([ 6.2704e-04, -3.7640e-05, -2.7567e-06, 3.7968e-05, -4.2605e-04, + -3.9116e-06, -4.9257e-04, -3.6061e-05, 7.6473e-05, 2.5749e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 249.06, cls_loss 0.0022 cls_loss_mapping 0.0075 cls_loss_causal 0.5917 re_mapping 0.0067 re_causal 0.0217 /// teacc 98.96 lr 0.00010000 +Epoch 103, weight, value: tensor([[ 0.0461, -0.0891, -0.1134, ..., -0.0668, -0.0578, -0.0598], + [ 0.0012, 0.0493, -0.0290, ..., -0.0086, 0.1213, 0.0220], + [-0.0118, -0.1033, -0.0162, ..., 0.1013, -0.0871, 0.0120], + ..., + [-0.0404, 0.0886, 0.0335, ..., -0.0675, 0.0763, 0.0316], + [-0.0091, -0.0474, 0.0076, ..., -0.0250, -0.1252, -0.0664], + [-0.0107, 0.0003, -0.0017, ..., -0.1133, -0.0864, 0.0260]], + device='cuda:0'), grad: tensor([[ 2.7032e-07, 2.5611e-07, 0.0000e+00, ..., 7.5391e-07, + 2.9919e-07, 4.5821e-07], + [-1.2498e-06, 9.2248e-07, 0.0000e+00, ..., 9.2201e-07, + -1.2955e-06, 6.3656e-07], + [-3.0268e-06, 1.3262e-06, 0.0000e+00, ..., -6.0089e-06, + 1.8291e-06, 1.6829e-06], + ..., + [ 6.9942e-07, -4.5560e-06, 0.0000e+00, ..., 7.1572e-07, + -3.8259e-06, -3.5837e-06], + [ 3.8326e-05, 2.4354e-07, 0.0000e+00, ..., 5.7742e-06, + 5.0059e-07, 2.7701e-05], + [-6.1691e-05, 1.4040e-07, 0.0000e+00, ..., 2.6617e-06, + 8.1770e-07, -4.7028e-05]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0275, 0.0119, 0.0130, 0.0285, 0.0257, -0.0202, 0.0251, 0.0288, + -0.0092, 0.0074], device='cuda:0'), grad: tensor([ 1.6121e-06, 2.2389e-06, -3.6526e-06, -1.0654e-05, 5.0843e-05, + 7.7114e-06, 4.0121e-06, -4.6566e-06, 9.8884e-05, -1.4627e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 249.19, cls_loss 0.0021 cls_loss_mapping 0.0084 cls_loss_causal 0.5470 re_mapping 0.0068 re_causal 0.0207 /// teacc 99.05 lr 0.00010000 +Epoch 104, weight, value: tensor([[ 0.0461, -0.0896, -0.1134, ..., -0.0673, -0.0583, -0.0601], + [ 0.0016, 0.0493, -0.0290, ..., -0.0092, 0.1216, 0.0220], + [-0.0121, -0.1037, -0.0162, ..., 0.1015, -0.0875, 0.0123], + ..., + [-0.0409, 0.0887, 0.0335, ..., -0.0678, 0.0764, 0.0314], + [-0.0090, -0.0476, 0.0076, ..., -0.0253, -0.1255, -0.0668], + [-0.0105, 0.0007, -0.0017, ..., -0.1138, -0.0863, 0.0263]], + device='cuda:0'), grad: tensor([[-2.1793e-06, 1.7602e-06, 0.0000e+00, ..., 1.4096e-05, + 4.6901e-06, 2.0768e-06], + [-7.3373e-05, 2.6241e-05, 0.0000e+00, ..., 2.0042e-06, + -5.6148e-05, -1.9222e-05], + [ 1.0751e-05, 6.6221e-05, 0.0000e+00, ..., 2.8729e-05, + 4.2558e-05, 9.0778e-05], + ..., + [-1.3141e-06, -3.8195e-04, 0.0000e+00, ..., -3.1978e-05, + -3.9721e-04, -2.5868e-04], + [ 4.1336e-05, 1.3232e-05, 0.0000e+00, ..., 2.6431e-06, + 5.6744e-05, 2.3082e-05], + [ 2.9191e-05, 1.0598e-04, 0.0000e+00, ..., 1.4435e-06, + 1.4925e-04, 6.3956e-05]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0277, 0.0119, 0.0123, 0.0297, 0.0253, -0.0195, 0.0254, 0.0282, + -0.0093, 0.0077], device='cuda:0'), grad: tensor([ 3.3259e-05, -6.6400e-05, 1.7393e-04, 2.2018e-04, 7.2122e-05, + 2.7955e-05, -9.2149e-05, -7.6962e-04, 1.1814e-04, 2.8253e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 248.74, cls_loss 0.0020 cls_loss_mapping 0.0057 cls_loss_causal 0.5575 re_mapping 0.0064 re_causal 0.0198 /// teacc 99.00 lr 0.00010000 +Epoch 105, weight, value: tensor([[ 0.0460, -0.0894, -0.1134, ..., -0.0683, -0.0589, -0.0605], + [ 0.0019, 0.0492, -0.0290, ..., -0.0096, 0.1218, 0.0219], + [-0.0121, -0.1043, -0.0162, ..., 0.1019, -0.0877, 0.0127], + ..., + [-0.0411, 0.0892, 0.0335, ..., -0.0682, 0.0767, 0.0317], + [-0.0089, -0.0477, 0.0076, ..., -0.0256, -0.1261, -0.0673], + [-0.0108, 0.0006, -0.0017, ..., -0.1144, -0.0866, 0.0265]], + device='cuda:0'), grad: tensor([[-7.7859e-07, 5.3435e-08, 0.0000e+00, ..., 5.7230e-07, + 2.9779e-07, 2.9197e-07], + [-9.1121e-06, -5.7835e-07, 0.0000e+00, ..., -9.3430e-06, + -2.2873e-05, -1.1235e-05], + [ 9.9987e-06, 7.2271e-07, 0.0000e+00, ..., 7.6666e-06, + 1.6481e-05, 8.6650e-06], + ..., + [ 1.5631e-05, 7.8380e-06, 0.0000e+00, ..., 5.1316e-07, + 2.0098e-06, 2.2035e-06], + [ 4.7907e-06, 8.6147e-07, 0.0000e+00, ..., 3.1199e-07, + 9.4669e-07, 8.0606e-07], + [ 1.6587e-06, -4.3516e-07, 0.0000e+00, ..., 2.7893e-07, + 5.9139e-07, -5.0198e-07]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0281, 0.0118, 0.0121, 0.0300, 0.0252, -0.0194, 0.0256, 0.0284, + -0.0093, 0.0075], device='cuda:0'), grad: tensor([-8.4750e-07, -2.5973e-05, 3.3289e-05, -1.8048e-04, -1.8068e-06, + 8.8394e-05, 4.7162e-06, 5.8204e-05, 1.8299e-05, 6.0163e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 248.95, cls_loss 0.0026 cls_loss_mapping 0.0075 cls_loss_causal 0.5535 re_mapping 0.0064 re_causal 0.0193 /// teacc 99.00 lr 0.00010000 +Epoch 106, weight, value: tensor([[ 4.6018e-02, -9.0405e-02, -1.1338e-01, ..., -6.9156e-02, + -5.9732e-02, -6.0899e-02], + [ 2.6019e-03, 4.9380e-02, -2.8996e-02, ..., -9.1717e-03, + 1.2258e-01, 2.2521e-02], + [-1.1257e-02, -1.0465e-01, -1.6176e-02, ..., 1.0310e-01, + -8.9068e-02, 1.3090e-02], + ..., + [-4.1379e-02, 8.9569e-02, 3.3547e-02, ..., -6.8819e-02, + 7.6881e-02, 3.1890e-02], + [-9.0471e-03, -4.7770e-02, 7.6484e-03, ..., -2.6619e-02, + -1.2658e-01, -6.8568e-02], + [-1.1113e-02, 1.0463e-04, -1.7162e-03, ..., -1.1530e-01, + -8.7433e-02, 2.6085e-02]], device='cuda:0'), grad: tensor([[ 2.0321e-06, 2.4047e-06, 0.0000e+00, ..., 1.3761e-05, + 6.0024e-07, 4.7088e-06], + [-8.8960e-06, 1.2349e-06, 0.0000e+00, ..., 3.1665e-06, + -1.3299e-05, 1.9427e-06], + [ 2.0191e-06, 7.7859e-06, 0.0000e+00, ..., 2.3879e-06, + 2.1476e-06, 9.5218e-06], + ..., + [ 2.0657e-06, 1.3649e-05, 0.0000e+00, ..., 1.5348e-05, + -4.9584e-06, 3.0205e-05], + [ 2.8405e-06, 8.4639e-06, 0.0000e+00, ..., 1.4253e-05, + 2.9225e-06, 1.5780e-05], + [ 1.3895e-05, 1.4389e-04, 0.0000e+00, ..., 1.1510e-04, + 1.6894e-06, 2.6441e-04]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0284, 0.0125, 0.0129, 0.0300, 0.0261, -0.0200, 0.0254, 0.0285, + -0.0095, 0.0065], device='cuda:0'), grad: tensor([ 5.2631e-05, -2.9951e-06, 4.8548e-05, 9.5442e-06, -9.8228e-04, + -9.8050e-05, -1.2763e-05, 9.7394e-05, 7.4565e-05, 8.1301e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 248.68, cls_loss 0.0026 cls_loss_mapping 0.0085 cls_loss_causal 0.5336 re_mapping 0.0064 re_causal 0.0200 /// teacc 98.93 lr 0.00010000 +Epoch 107, weight, value: tensor([[ 0.0461, -0.0909, -0.1134, ..., -0.0697, -0.0606, -0.0616], + [ 0.0045, 0.0497, -0.0290, ..., -0.0110, 0.1239, 0.0233], + [-0.0113, -0.1057, -0.0162, ..., 0.1044, -0.0895, 0.0140], + ..., + [-0.0434, 0.0895, 0.0335, ..., -0.0698, 0.0761, 0.0311], + [-0.0089, -0.0480, 0.0077, ..., -0.0271, -0.1271, -0.0691], + [-0.0109, -0.0002, -0.0017, ..., -0.1159, -0.0883, 0.0265]], + device='cuda:0'), grad: tensor([[ 8.2515e-07, 6.5891e-08, 0.0000e+00, ..., 6.2631e-07, + 8.6240e-07, 6.6217e-07], + [-2.5287e-05, -1.3672e-06, 0.0000e+00, ..., -1.2979e-05, + -2.7791e-05, -1.3202e-05], + [ 1.7164e-06, 1.2375e-07, 0.0000e+00, ..., 4.5798e-07, + 1.1809e-06, 6.5845e-07], + ..., + [ 4.5868e-07, -1.2584e-07, 0.0000e+00, ..., 1.9604e-07, + -3.1898e-08, 3.0454e-07], + [-2.0955e-06, 6.4028e-08, 0.0000e+00, ..., 2.2650e-06, + 6.6729e-07, 3.5055e-06], + [ 8.4698e-05, -1.4470e-07, 0.0000e+00, ..., 1.7858e-04, + 2.5216e-07, 4.0579e-04]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0286, 0.0120, 0.0133, 0.0309, 0.0260, -0.0211, 0.0271, 0.0276, + -0.0098, 0.0066], device='cuda:0'), grad: tensor([ 2.5760e-06, -5.7042e-05, 5.4389e-06, 2.4989e-05, -1.4439e-03, + -1.8731e-05, 5.3972e-05, 1.8636e-06, 3.2485e-06, 1.4277e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 249.04, cls_loss 0.0025 cls_loss_mapping 0.0079 cls_loss_causal 0.5470 re_mapping 0.0065 re_causal 0.0200 /// teacc 98.75 lr 0.00010000 +Epoch 108, weight, value: tensor([[ 0.0462, -0.0915, -0.1134, ..., -0.0701, -0.0619, -0.0619], + [ 0.0045, 0.0488, -0.0290, ..., -0.0114, 0.1235, 0.0223], + [-0.0112, -0.1062, -0.0162, ..., 0.1052, -0.0903, 0.0144], + ..., + [-0.0426, 0.0915, 0.0335, ..., -0.0700, 0.0778, 0.0328], + [-0.0087, -0.0488, 0.0077, ..., -0.0276, -0.1277, -0.0698], + [-0.0113, -0.0014, -0.0017, ..., -0.1166, -0.0906, 0.0259]], + device='cuda:0'), grad: tensor([[-1.0841e-06, 1.3020e-06, 0.0000e+00, ..., 1.9129e-06, + 3.0007e-06, 1.5153e-06], + [-3.0175e-05, -5.0254e-06, 0.0000e+00, ..., -1.0310e-06, + -4.6045e-05, -1.3076e-05], + [ 1.4000e-05, 5.2042e-06, 0.0000e+00, ..., 2.6468e-06, + 2.2352e-05, 7.6592e-06], + ..., + [ 4.7348e-06, -2.4274e-05, 0.0000e+00, ..., 5.8627e-07, + -1.0677e-05, -2.8074e-05], + [ 1.9353e-06, 2.3898e-06, 0.0000e+00, ..., 3.7216e-06, + 3.7886e-06, 1.7853e-06], + [ 2.8796e-06, 3.2671e-06, 0.0000e+00, ..., 3.9325e-07, + 5.5358e-06, 3.7178e-06]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0286, 0.0113, 0.0135, 0.0312, 0.0261, -0.0217, 0.0270, 0.0294, + -0.0099, 0.0056], device='cuda:0'), grad: tensor([-6.0303e-07, -6.3598e-05, 4.2558e-05, 1.1154e-05, 3.6180e-05, + 9.3505e-06, -2.8327e-05, -3.1471e-05, 1.2740e-05, 1.2003e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 249.07, cls_loss 0.0024 cls_loss_mapping 0.0063 cls_loss_causal 0.5377 re_mapping 0.0064 re_causal 0.0197 /// teacc 99.02 lr 0.00010000 +Epoch 109, weight, value: tensor([[ 0.0463, -0.0921, -0.1134, ..., -0.0702, -0.0627, -0.0624], + [ 0.0050, 0.0488, -0.0290, ..., -0.0117, 0.1238, 0.0223], + [-0.0119, -0.1068, -0.0162, ..., 0.1056, -0.0906, 0.0150], + ..., + [-0.0429, 0.0917, 0.0335, ..., -0.0709, 0.0779, 0.0328], + [-0.0076, -0.0486, 0.0077, ..., -0.0274, -0.1286, -0.0703], + [-0.0117, -0.0014, -0.0017, ..., -0.1174, -0.0906, 0.0258]], + device='cuda:0'), grad: tensor([[ 3.8631e-06, 1.1418e-06, 0.0000e+00, ..., 6.6422e-06, + 1.0971e-06, 2.2482e-06], + [ 1.8507e-05, 6.4634e-06, 0.0000e+00, ..., 2.2560e-05, + 4.6976e-06, 3.8706e-06], + [-2.2240e-06, 3.4831e-06, 0.0000e+00, ..., -1.1601e-05, + 3.3043e-06, -1.7628e-05], + ..., + [ 6.0722e-06, -6.2957e-06, 0.0000e+00, ..., 1.0937e-05, + -7.0222e-06, -4.2208e-06], + [-8.3923e-05, -2.3946e-05, 0.0000e+00, ..., -4.9651e-05, + -1.1921e-05, 4.7646e-06], + [ 4.9770e-06, 3.0641e-06, 0.0000e+00, ..., 5.0329e-06, + 3.6974e-06, 8.6986e-07]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0285, 0.0114, 0.0127, 0.0297, 0.0261, -0.0203, 0.0272, 0.0291, + -0.0085, 0.0051], device='cuda:0'), grad: tensor([ 1.9118e-05, 1.1575e-04, 2.0280e-05, -1.8954e-04, 4.2051e-05, + 1.2004e-04, 6.2466e-05, 4.5776e-05, -2.6703e-04, 3.0756e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 249.06, cls_loss 0.0022 cls_loss_mapping 0.0065 cls_loss_causal 0.5326 re_mapping 0.0064 re_causal 0.0195 /// teacc 98.94 lr 0.00010000 +Epoch 110, weight, value: tensor([[ 0.0458, -0.0929, -0.1134, ..., -0.0708, -0.0640, -0.0628], + [ 0.0052, 0.0487, -0.0290, ..., -0.0121, 0.1239, 0.0221], + [-0.0122, -0.1072, -0.0162, ..., 0.1058, -0.0911, 0.0148], + ..., + [-0.0430, 0.0925, 0.0335, ..., -0.0709, 0.0786, 0.0337], + [-0.0072, -0.0487, 0.0077, ..., -0.0272, -0.1288, -0.0711], + [-0.0117, -0.0016, -0.0017, ..., -0.1179, -0.0911, 0.0255]], + device='cuda:0'), grad: tensor([[ 2.7451e-07, 4.6706e-07, 0.0000e+00, ..., 2.6301e-06, + 2.9989e-07, 3.0566e-06], + [-2.5816e-06, -5.7835e-07, 0.0000e+00, ..., 1.2042e-06, + -6.6385e-06, 2.4564e-07], + [ 5.8394e-07, 3.5856e-07, 0.0000e+00, ..., -7.7859e-06, + 1.2768e-06, -5.2154e-06], + ..., + [ 1.4137e-06, 3.3528e-06, 0.0000e+00, ..., 4.7423e-06, + 1.9260e-06, 1.1668e-05], + [ 4.8568e-07, 5.7882e-07, 0.0000e+00, ..., 1.3700e-06, + 1.3141e-06, 2.4233e-06], + [-1.6168e-06, -2.6505e-06, 0.0000e+00, ..., 7.5139e-06, + -1.0487e-06, 5.0291e-06]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0292, 0.0111, 0.0123, 0.0301, 0.0265, -0.0201, 0.0268, 0.0297, + -0.0083, 0.0047], device='cuda:0'), grad: tensor([ 9.2685e-06, -3.0380e-06, -9.4622e-06, 8.0839e-06, -6.8605e-05, + -2.8536e-05, 2.9728e-05, 3.2991e-05, 1.0535e-05, 1.8850e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 249.04, cls_loss 0.0020 cls_loss_mapping 0.0061 cls_loss_causal 0.5647 re_mapping 0.0065 re_causal 0.0197 /// teacc 98.90 lr 0.00010000 +Epoch 111, weight, value: tensor([[ 0.0452, -0.0937, -0.1134, ..., -0.0713, -0.0647, -0.0635], + [ 0.0054, 0.0487, -0.0290, ..., -0.0125, 0.1242, 0.0220], + [-0.0122, -0.1079, -0.0159, ..., 0.1066, -0.0914, 0.0154], + ..., + [-0.0433, 0.0929, 0.0335, ..., -0.0715, 0.0788, 0.0339], + [-0.0066, -0.0490, 0.0075, ..., -0.0274, -0.1295, -0.0717], + [-0.0118, -0.0016, -0.0017, ..., -0.1181, -0.0914, 0.0257]], + device='cuda:0'), grad: tensor([[-1.5541e-07, 4.0815e-07, 0.0000e+00, ..., 2.6114e-06, + 4.0396e-07, 2.8443e-06], + [-7.6508e-07, 9.0525e-07, 0.0000e+00, ..., 1.3765e-06, + -3.5972e-07, 1.7844e-06], + [ 4.1500e-06, -4.1234e-07, 0.0000e+00, ..., -1.8850e-05, + 1.2126e-06, -2.5451e-05], + ..., + [ 1.1694e-07, -4.2543e-06, 0.0000e+00, ..., 1.5467e-05, + -4.8131e-06, 1.4640e-05], + [-1.0459e-06, 7.4646e-07, 0.0000e+00, ..., 2.6226e-06, + 8.4424e-07, 3.6545e-06], + [ 2.8778e-07, 6.0489e-07, 0.0000e+00, ..., 7.8827e-06, + 5.8953e-07, 9.0599e-06]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0302, 0.0110, 0.0124, 0.0303, 0.0262, -0.0204, 0.0269, 0.0298, + -0.0074, 0.0046], device='cuda:0'), grad: tensor([ 5.2154e-06, 3.8147e-06, -3.1978e-05, -1.3538e-05, -3.0145e-05, + 4.0755e-06, 3.3751e-06, 2.9266e-05, 4.1053e-06, 2.5719e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 249.04, cls_loss 0.0024 cls_loss_mapping 0.0075 cls_loss_causal 0.5648 re_mapping 0.0067 re_causal 0.0195 /// teacc 98.83 lr 0.00010000 +Epoch 112, weight, value: tensor([[ 0.0460, -0.0939, -0.1134, ..., -0.0717, -0.0653, -0.0639], + [ 0.0060, 0.0484, -0.0290, ..., -0.0119, 0.1242, 0.0221], + [-0.0127, -0.1088, -0.0159, ..., 0.1074, -0.0922, 0.0156], + ..., + [-0.0435, 0.0938, 0.0335, ..., -0.0722, 0.0794, 0.0345], + [-0.0065, -0.0490, 0.0075, ..., -0.0280, -0.1299, -0.0724], + [-0.0121, -0.0021, -0.0017, ..., -0.1188, -0.0922, 0.0254]], + device='cuda:0'), grad: tensor([[ 1.3843e-05, 9.1037e-07, 1.6298e-08, ..., -1.2685e-06, + 1.5106e-06, -6.7707e-07], + [-2.8200e-06, -1.1832e-05, -7.0082e-08, ..., -9.9242e-06, + -2.9609e-05, -1.1228e-05], + [ 1.5154e-05, 1.6540e-06, 9.2987e-09, ..., 5.9465e-07, + 3.2149e-06, 6.4112e-06], + ..., + [ 2.3663e-05, 1.2167e-05, 1.5949e-07, ..., 3.7141e-06, + 8.3819e-06, 5.2780e-05], + [ 1.4335e-05, -6.9989e-07, 1.7506e-08, ..., 1.0850e-06, + 3.9116e-06, 4.7125e-06], + [ 4.5210e-05, 2.1651e-05, -3.8277e-07, ..., 6.5900e-06, + 1.0274e-05, 1.3447e-04]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0295, 0.0113, 0.0124, 0.0298, 0.0263, -0.0200, 0.0266, 0.0301, + -0.0075, 0.0040], device='cuda:0'), grad: tensor([ 8.3506e-05, 6.1452e-05, 7.8142e-05, 2.5673e-03, -2.7561e-04, + -3.4275e-03, 1.3375e-04, 1.9872e-04, 9.4473e-05, 4.8637e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 111, time 249.07, cls_loss 0.0028 cls_loss_mapping 0.0077 cls_loss_causal 0.5504 re_mapping 0.0064 re_causal 0.0190 /// teacc 99.07 lr 0.00010000 +Epoch 113, weight, value: tensor([[ 0.0465, -0.0945, -0.1134, ..., -0.0717, -0.0662, -0.0644], + [ 0.0069, 0.0493, -0.0289, ..., -0.0126, 0.1253, 0.0225], + [-0.0130, -0.1072, -0.0159, ..., 0.1094, -0.0920, 0.0180], + ..., + [-0.0448, 0.0925, 0.0333, ..., -0.0745, 0.0786, 0.0327], + [-0.0064, -0.0491, 0.0075, ..., -0.0283, -0.1305, -0.0734], + [-0.0118, -0.0016, -0.0017, ..., -0.1194, -0.0924, 0.0264]], + device='cuda:0'), grad: tensor([[-1.2189e-05, -7.5586e-06, 6.6211e-09, ..., 4.1979e-07, + 6.8033e-07, 4.1607e-07], + [-1.8328e-05, -2.7940e-05, 1.2660e-09, ..., 5.5460e-07, + -5.2214e-05, -1.8492e-05], + [-2.6803e-06, 1.1288e-06, 3.1287e-09, ..., -1.0125e-05, + 1.8934e-06, -1.2301e-05], + ..., + [ 1.5259e-05, 2.0519e-05, 9.8516e-09, ..., 9.8348e-07, + 3.1263e-05, 1.3225e-05], + [ 3.5507e-08, 6.3516e-06, 2.1697e-08, ..., -5.2969e-08, + 7.9498e-06, 3.2373e-06], + [ 3.4198e-06, 2.5742e-06, -1.1566e-07, ..., 1.4957e-06, + 3.3937e-06, 2.1905e-06]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0290, 0.0116, 0.0135, 0.0291, 0.0264, -0.0186, 0.0265, 0.0279, + -0.0078, 0.0049], device='cuda:0'), grad: tensor([-4.5329e-05, -5.9783e-05, -1.9804e-05, 2.9966e-05, 7.8529e-06, + 6.6534e-06, 1.9312e-05, 5.1528e-05, -2.4009e-06, 1.1861e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 248.55, cls_loss 0.0022 cls_loss_mapping 0.0057 cls_loss_causal 0.5496 re_mapping 0.0064 re_causal 0.0194 /// teacc 98.98 lr 0.00010000 +Epoch 114, weight, value: tensor([[ 0.0464, -0.0948, -0.1143, ..., -0.0721, -0.0667, -0.0651], + [ 0.0068, 0.0489, -0.0290, ..., -0.0136, 0.1248, 0.0214], + [-0.0127, -0.1064, -0.0167, ..., 0.1116, -0.0900, 0.0203], + ..., + [-0.0451, 0.0928, 0.0331, ..., -0.0765, 0.0786, 0.0324], + [-0.0064, -0.0495, 0.0076, ..., -0.0288, -0.1311, -0.0741], + [-0.0112, -0.0014, -0.0018, ..., -0.1199, -0.0924, 0.0271]], + device='cuda:0'), grad: tensor([[ 1.5339e-06, 7.1675e-06, 0.0000e+00, ..., 8.7777e-07, + 5.5991e-06, 5.1521e-06], + [-2.8864e-05, 3.9160e-05, 0.0000e+00, ..., 2.4983e-07, + -3.7551e-06, 2.1249e-05], + [ 4.4145e-06, 3.9816e-05, 0.0000e+00, ..., -9.4622e-07, + 1.3642e-05, 2.6181e-05], + ..., + [-4.5836e-05, -3.0255e-04, 0.0000e+00, ..., 3.5390e-07, + -2.1660e-04, -2.1923e-04], + [ 1.1005e-05, 1.1712e-05, 0.0000e+00, ..., 1.3765e-06, + 1.8120e-05, 1.1563e-05], + [ 4.2915e-05, 1.6022e-04, 0.0000e+00, ..., 4.2072e-07, + 1.4806e-04, 1.1790e-04]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0294, 0.0108, 0.0152, 0.0287, 0.0264, -0.0185, 0.0267, 0.0273, + -0.0080, 0.0056], device='cuda:0'), grad: tensor([ 1.4462e-05, 1.3389e-05, 8.2076e-05, 2.1443e-05, 4.3482e-05, + 1.8850e-05, -1.6671e-06, -5.8603e-04, 4.4316e-05, 3.4976e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 113, time 248.73, cls_loss 0.0024 cls_loss_mapping 0.0064 cls_loss_causal 0.5179 re_mapping 0.0063 re_causal 0.0188 /// teacc 98.95 lr 0.00010000 +Epoch 115, weight, value: tensor([[ 0.0469, -0.0955, -0.1145, ..., -0.0723, -0.0677, -0.0655], + [ 0.0065, 0.0478, -0.0290, ..., -0.0137, 0.1244, 0.0204], + [-0.0131, -0.1087, -0.0167, ..., 0.1112, -0.0915, 0.0192], + ..., + [-0.0443, 0.0952, 0.0331, ..., -0.0753, 0.0800, 0.0346], + [-0.0062, -0.0493, 0.0077, ..., -0.0295, -0.1313, -0.0747], + [-0.0112, -0.0028, -0.0019, ..., -0.1204, -0.0936, 0.0265]], + device='cuda:0'), grad: tensor([[-8.5607e-06, -9.4473e-06, 3.0734e-08, ..., 2.6636e-06, + -2.4941e-06, 6.6869e-07], + [ 2.2464e-06, -8.0559e-07, 7.9069e-07, ..., 2.3805e-06, + -2.8536e-06, 5.4277e-06], + [ 3.6880e-07, 2.7521e-07, 3.5216e-09, ..., -1.0926e-04, + -4.8965e-05, -1.7977e-04], + ..., + [ 7.5102e-06, 8.7544e-06, 6.8569e-08, ..., 1.0669e-04, + 5.2482e-05, 1.7619e-04], + [ 6.2631e-07, 3.8533e-07, 1.2413e-08, ..., 2.7865e-06, + 4.9034e-07, 6.7195e-07], + [-7.4692e-06, -1.9576e-06, -1.9167e-06, ..., 1.5870e-06, + -2.3767e-06, -9.1866e-06]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0290, 0.0101, 0.0138, 0.0285, 0.0265, -0.0186, 0.0267, 0.0295, + -0.0080, 0.0048], device='cuda:0'), grad: tensor([-3.4213e-05, 1.4417e-05, -2.4605e-04, 3.6713e-06, 8.1360e-06, + 7.4923e-05, -8.7380e-05, 2.7776e-04, 9.4250e-06, -2.0891e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 248.92, cls_loss 0.0021 cls_loss_mapping 0.0075 cls_loss_causal 0.5622 re_mapping 0.0063 re_causal 0.0197 /// teacc 98.90 lr 0.00010000 +Epoch 116, weight, value: tensor([[ 0.0472, -0.0962, -0.1146, ..., -0.0727, -0.0687, -0.0660], + [ 0.0072, 0.0476, -0.0290, ..., -0.0138, 0.1247, 0.0199], + [-0.0135, -0.1098, -0.0168, ..., 0.1114, -0.0926, 0.0190], + ..., + [-0.0446, 0.0960, 0.0331, ..., -0.0751, 0.0805, 0.0352], + [-0.0056, -0.0476, 0.0077, ..., -0.0298, -0.1301, -0.0731], + [-0.0112, -0.0038, -0.0019, ..., -0.1204, -0.0954, 0.0261]], + device='cuda:0'), grad: tensor([[ 1.0338e-06, 1.9092e-08, 0.0000e+00, ..., 1.2275e-06, + 7.8580e-08, 1.7614e-07], + [-1.5309e-07, -4.8522e-07, 0.0000e+00, ..., 7.7591e-08, + -1.6512e-06, -4.2329e-07], + [ 1.6885e-06, 5.2794e-08, 0.0000e+00, ..., 5.5210e-08, + 2.0768e-07, 1.0530e-07], + ..., + [ 4.0829e-06, 8.3062e-08, 0.0000e+00, ..., 5.3551e-08, + 6.7521e-07, 4.3702e-07], + [ 1.5171e-06, 8.0501e-08, 0.0000e+00, ..., 2.2363e-07, + 2.7614e-07, 2.5635e-07], + [ 1.0288e-08, 6.0303e-08, 0.0000e+00, ..., 5.9110e-08, + 2.3027e-07, -1.9092e-06]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0287, 0.0098, 0.0131, 0.0282, 0.0268, -0.0189, 0.0268, 0.0301, + -0.0065, 0.0035], device='cuda:0'), grad: tensor([ 7.1526e-06, 4.9882e-06, 8.3074e-06, -5.0694e-05, 2.6803e-06, + 4.3027e-06, -3.3267e-06, 1.8716e-05, 7.1749e-06, 7.0781e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 249.02, cls_loss 0.0018 cls_loss_mapping 0.0055 cls_loss_causal 0.5411 re_mapping 0.0061 re_causal 0.0188 /// teacc 98.98 lr 0.00010000 +Epoch 117, weight, value: tensor([[ 0.0471, -0.0968, -0.1146, ..., -0.0736, -0.0701, -0.0669], + [ 0.0076, 0.0475, -0.0291, ..., -0.0141, 0.1249, 0.0199], + [-0.0135, -0.1099, -0.0167, ..., 0.1119, -0.0927, 0.0193], + ..., + [-0.0448, 0.0962, 0.0331, ..., -0.0752, 0.0807, 0.0352], + [-0.0056, -0.0476, 0.0077, ..., -0.0302, -0.1304, -0.0736], + [-0.0110, -0.0038, -0.0019, ..., -0.1208, -0.0956, 0.0263]], + device='cuda:0'), grad: tensor([[ 2.1094e-07, 5.3970e-07, 0.0000e+00, ..., 3.2950e-06, + 2.4419e-06, 9.3598e-07], + [-1.3515e-05, -4.4405e-06, 0.0000e+00, ..., 1.3513e-06, + -3.9876e-05, -1.4350e-05], + [ 2.6617e-06, 1.0999e-06, 0.0000e+00, ..., -1.6838e-06, + 1.3143e-05, 4.1164e-06], + ..., + [ 4.3251e-06, -5.5972e-07, 0.0000e+00, ..., 5.0850e-07, + 1.0811e-05, 3.5316e-06], + [ 4.6082e-06, 1.2228e-06, 0.0000e+00, ..., 1.9968e-06, + 5.4613e-06, 2.2221e-06], + [ 2.3823e-06, 6.2026e-07, 0.0000e+00, ..., 7.1619e-07, + 2.4196e-06, 1.2899e-06]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0296, 0.0098, 0.0132, 0.0282, 0.0269, -0.0192, 0.0275, 0.0301, + -0.0067, 0.0036], device='cuda:0'), grad: tensor([ 8.3670e-06, -4.1634e-05, 1.1869e-05, 1.3478e-05, 2.8387e-05, + -2.1666e-05, -4.1634e-05, 1.3262e-05, 1.8910e-05, 1.0647e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 249.37, cls_loss 0.0024 cls_loss_mapping 0.0051 cls_loss_causal 0.5543 re_mapping 0.0059 re_causal 0.0175 /// teacc 98.99 lr 0.00010000 +Epoch 118, weight, value: tensor([[ 0.0476, -0.0974, -0.1148, ..., -0.0740, -0.0714, -0.0674], + [ 0.0080, 0.0476, -0.0291, ..., -0.0142, 0.1253, 0.0199], + [-0.0133, -0.1101, -0.0167, ..., 0.1127, -0.0929, 0.0198], + ..., + [-0.0455, 0.0967, 0.0331, ..., -0.0756, 0.0809, 0.0354], + [-0.0056, -0.0478, 0.0077, ..., -0.0310, -0.1308, -0.0742], + [-0.0107, -0.0039, -0.0019, ..., -0.1212, -0.0962, 0.0270]], + device='cuda:0'), grad: tensor([[-1.1548e-05, 5.0245e-07, 0.0000e+00, ..., 4.3400e-06, + 8.3540e-07, 9.1922e-07], + [ 2.7884e-06, 2.3488e-06, 0.0000e+00, ..., 3.3416e-06, + 2.6897e-06, 4.6529e-06], + [ 5.2787e-06, 4.2953e-06, 0.0000e+00, ..., -2.5615e-05, + -4.3446e-07, -1.6630e-05], + ..., + [ 7.4096e-06, -1.5616e-05, 0.0000e+00, ..., 6.6422e-06, + -1.7926e-05, -8.5160e-06], + [ 1.3746e-05, 8.2003e-07, 0.0000e+00, ..., 1.6630e-05, + 3.7160e-06, 1.3731e-05], + [ 2.8554e-06, 1.6410e-06, 0.0000e+00, ..., 1.9856e-06, + 2.0973e-06, -4.5039e-06]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0293, 0.0099, 0.0134, 0.0278, 0.0260, -0.0188, 0.0275, 0.0302, + -0.0070, 0.0041], device='cuda:0'), grad: tensor([-4.0270e-06, 2.7552e-05, -2.8834e-05, -1.4198e-04, 4.4346e-05, + 6.3837e-05, -5.7131e-05, 1.6257e-05, 8.7380e-05, -7.6815e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 248.93, cls_loss 0.0018 cls_loss_mapping 0.0063 cls_loss_causal 0.5506 re_mapping 0.0059 re_causal 0.0181 /// teacc 98.93 lr 0.00010000 +Epoch 119, weight, value: tensor([[ 0.0479, -0.0981, -0.1149, ..., -0.0746, -0.0729, -0.0679], + [ 0.0089, 0.0478, -0.0291, ..., -0.0146, 0.1257, 0.0198], + [-0.0132, -0.1104, -0.0167, ..., 0.1132, -0.0934, 0.0198], + ..., + [-0.0459, 0.0970, 0.0331, ..., -0.0757, 0.0814, 0.0358], + [-0.0059, -0.0481, 0.0077, ..., -0.0315, -0.1316, -0.0748], + [-0.0109, -0.0043, -0.0019, ..., -0.1217, -0.0972, 0.0265]], + device='cuda:0'), grad: tensor([[ 5.2387e-08, 2.3737e-07, 0.0000e+00, ..., 1.4445e-06, + 1.3865e-07, 2.2501e-06], + [-2.5872e-06, -2.7046e-06, 0.0000e+00, ..., 7.5856e-07, + -6.9030e-06, -1.7807e-06], + [ 1.1045e-06, 5.2154e-07, 0.0000e+00, ..., 4.5970e-06, + 1.2442e-06, 6.6720e-06], + ..., + [ 3.7085e-06, 4.1947e-06, 0.0000e+00, ..., 8.0047e-07, + 3.0845e-06, 6.9290e-06], + [ 3.5390e-07, 8.3167e-07, 0.0000e+00, ..., 9.4017e-07, + 1.1418e-06, 2.1644e-06], + [-3.6471e-06, -4.8690e-06, 0.0000e+00, ..., 5.6475e-06, + 1.5413e-07, 1.2852e-07]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0294, 0.0101, 0.0132, 0.0275, 0.0268, -0.0183, 0.0274, 0.0304, + -0.0075, 0.0035], device='cuda:0'), grad: tensor([ 4.8764e-06, -3.1646e-06, 1.7986e-05, -6.3963e-06, -1.0973e-04, + 1.5706e-05, 5.5432e-05, 2.0966e-05, 6.1542e-06, -1.8757e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 249.07, cls_loss 0.0018 cls_loss_mapping 0.0046 cls_loss_causal 0.5238 re_mapping 0.0058 re_causal 0.0176 /// teacc 98.89 lr 0.00010000 +Epoch 120, weight, value: tensor([[ 0.0489, -0.0981, -0.1149, ..., -0.0751, -0.0738, -0.0682], + [ 0.0092, 0.0478, -0.0290, ..., -0.0149, 0.1260, 0.0198], + [-0.0135, -0.1107, -0.0167, ..., 0.1137, -0.0938, 0.0198], + ..., + [-0.0462, 0.0973, 0.0331, ..., -0.0759, 0.0815, 0.0359], + [-0.0055, -0.0480, 0.0077, ..., -0.0317, -0.1319, -0.0752], + [-0.0113, -0.0045, -0.0019, ..., -0.1224, -0.0976, 0.0267]], + device='cuda:0'), grad: tensor([[-1.3322e-05, 3.7136e-08, 0.0000e+00, ..., 9.8604e-08, + 6.7637e-08, 2.0443e-07], + [ 5.3039e-07, 2.7716e-06, 0.0000e+00, ..., 3.2689e-07, + 1.9353e-06, 2.0433e-06], + [ 2.7101e-07, 2.8103e-07, 0.0000e+00, ..., -5.0850e-06, + 5.2154e-07, -1.5004e-06], + ..., + [ 2.5909e-06, -6.9514e-06, 0.0000e+00, ..., 1.4016e-06, + -6.4783e-06, -4.3549e-06], + [ 4.6566e-06, 3.1595e-07, 0.0000e+00, ..., 4.9686e-07, + 8.6427e-07, 5.8906e-07], + [ 1.1371e-06, 1.6112e-06, 0.0000e+00, ..., 1.6054e-07, + 1.8245e-06, 4.0047e-07]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0289, 0.0100, 0.0131, 0.0278, 0.0265, -0.0186, 0.0276, 0.0303, + -0.0072, 0.0035], device='cuda:0'), grad: tensor([-4.4376e-05, 7.7561e-06, -3.7998e-06, 4.6566e-06, 1.3774e-06, + 3.8743e-06, 6.3591e-06, 1.1642e-06, 1.7226e-05, 5.8189e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 119, time 249.05, cls_loss 0.0024 cls_loss_mapping 0.0069 cls_loss_causal 0.5222 re_mapping 0.0059 re_causal 0.0180 /// teacc 98.91 lr 0.00010000 +Epoch 121, weight, value: tensor([[ 0.0490, -0.0989, -0.1149, ..., -0.0760, -0.0750, -0.0689], + [ 0.0097, 0.0480, -0.0290, ..., -0.0152, 0.1265, 0.0199], + [-0.0138, -0.1109, -0.0167, ..., 0.1144, -0.0940, 0.0202], + ..., + [-0.0469, 0.0973, 0.0331, ..., -0.0762, 0.0815, 0.0354], + [-0.0060, -0.0490, 0.0077, ..., -0.0328, -0.1324, -0.0762], + [-0.0104, -0.0034, -0.0019, ..., -0.1230, -0.0976, 0.0275]], + device='cuda:0'), grad: tensor([[ 3.2037e-06, 9.6741e-08, 0.0000e+00, ..., 1.3467e-06, + 1.6775e-07, 1.8917e-07], + [ 1.2051e-06, -1.0766e-06, 0.0000e+00, ..., 3.0990e-07, + -4.0680e-06, 9.1689e-07], + [ 7.4226e-07, 6.4634e-07, 0.0000e+00, ..., -3.6675e-06, + 1.1977e-06, 6.9197e-07], + ..., + [ 1.6484e-06, -1.5106e-06, 0.0000e+00, ..., 4.5262e-07, + 5.7276e-08, -1.0477e-06], + [-2.0325e-05, 2.7195e-07, 0.0000e+00, ..., -1.0813e-06, + 7.0361e-07, 4.5984e-07], + [-3.0827e-06, 6.4215e-07, 0.0000e+00, ..., 4.1886e-07, + 5.9744e-07, -4.5374e-06]], device='cuda:0') +Epoch 121, bias, value: tensor([-0.0296, 0.0101, 0.0131, 0.0282, 0.0277, -0.0177, 0.0267, 0.0296, + -0.0085, 0.0044], device='cuda:0'), grad: tensor([ 1.5065e-05, 7.5698e-06, 3.7048e-06, 3.7044e-05, 4.7125e-06, + 1.4700e-05, 9.2685e-06, 1.6484e-06, -8.6248e-05, -7.5065e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 249.01, cls_loss 0.0021 cls_loss_mapping 0.0051 cls_loss_causal 0.5488 re_mapping 0.0058 re_causal 0.0180 /// teacc 99.04 lr 0.00010000 +Epoch 122, weight, value: tensor([[ 0.0492, -0.0993, -0.1150, ..., -0.0771, -0.0755, -0.0693], + [ 0.0091, 0.0479, -0.0290, ..., -0.0155, 0.1265, 0.0193], + [-0.0141, -0.1119, -0.0168, ..., 0.1152, -0.0947, 0.0198], + ..., + [-0.0473, 0.0980, 0.0331, ..., -0.0765, 0.0819, 0.0361], + [-0.0056, -0.0491, 0.0079, ..., -0.0330, -0.1326, -0.0766], + [-0.0089, -0.0030, -0.0019, ..., -0.1236, -0.0972, 0.0286]], + device='cuda:0'), grad: tensor([[ 1.4715e-06, 6.8545e-07, 0.0000e+00, ..., 1.6298e-06, + 1.0971e-06, 3.7439e-06], + [-6.0350e-06, -3.5875e-06, 0.0000e+00, ..., 5.4436e-07, + -1.5333e-05, -2.1085e-06], + [ 1.5274e-06, 1.1595e-06, 0.0000e+00, ..., -1.8207e-06, + 2.2147e-06, 1.3290e-06], + ..., + [ 2.2631e-06, -1.9372e-06, 0.0000e+00, ..., 7.4087e-07, + 2.6310e-07, 1.3714e-07], + [ 4.2245e-06, 1.6848e-06, 0.0000e+00, ..., 2.3711e-06, + 4.0792e-06, 8.8736e-06], + [-8.5235e-06, -3.0845e-06, 0.0000e+00, ..., 3.9139e-07, + 1.9222e-06, -2.1994e-05]], device='cuda:0') +Epoch 122, bias, value: tensor([-0.0300, 0.0094, 0.0125, 0.0299, 0.0269, -0.0192, 0.0269, 0.0301, + -0.0083, 0.0055], device='cuda:0'), grad: tensor([ 1.5303e-05, -1.5616e-05, 8.7544e-06, 9.9093e-06, 2.4095e-05, + 2.7400e-06, -6.4038e-06, 6.3516e-06, 3.5048e-05, -8.0168e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 248.83, cls_loss 0.0020 cls_loss_mapping 0.0059 cls_loss_causal 0.5524 re_mapping 0.0061 re_causal 0.0185 /// teacc 98.96 lr 0.00010000 +Epoch 123, weight, value: tensor([[ 0.0493, -0.0996, -0.1150, ..., -0.0786, -0.0768, -0.0702], + [ 0.0104, 0.0481, -0.0290, ..., -0.0158, 0.1274, 0.0196], + [-0.0134, -0.1120, -0.0168, ..., 0.1166, -0.0950, 0.0207], + ..., + [-0.0481, 0.0980, 0.0331, ..., -0.0770, 0.0818, 0.0361], + [-0.0061, -0.0492, 0.0078, ..., -0.0339, -0.1341, -0.0772], + [-0.0089, -0.0033, -0.0019, ..., -0.1240, -0.0977, 0.0287]], + device='cuda:0'), grad: tensor([[-1.1154e-05, 0.0000e+00, 0.0000e+00, ..., -1.7071e-06, + 3.8301e-08, 3.7748e-08], + [-2.5006e-07, 0.0000e+00, 0.0000e+00, ..., -2.5844e-07, + -1.5395e-06, -9.2201e-07], + [ 1.1355e-05, 0.0000e+00, 0.0000e+00, ..., 2.4997e-06, + 8.3866e-07, 3.6508e-07], + ..., + [ 7.1619e-07, 0.0000e+00, 0.0000e+00, ..., 1.5600e-07, + 5.1950e-08, 7.1945e-08], + [-4.4927e-06, 0.0000e+00, 0.0000e+00, ..., -7.7952e-07, + 8.4925e-08, -4.7090e-08], + [ 4.6939e-06, 0.0000e+00, 0.0000e+00, ..., 7.2550e-07, + 4.7672e-08, 9.4878e-08]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0306, 0.0103, 0.0135, 0.0297, 0.0263, -0.0191, 0.0269, 0.0296, + -0.0089, 0.0054], device='cuda:0'), grad: tensor([-3.8803e-05, 1.7164e-06, 4.5538e-05, -2.0921e-05, 2.8051e-06, + -2.3574e-05, 1.2882e-05, 4.5076e-06, -8.7172e-06, 2.4408e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 248.97, cls_loss 0.0025 cls_loss_mapping 0.0074 cls_loss_causal 0.5437 re_mapping 0.0060 re_causal 0.0179 /// teacc 98.97 lr 0.00010000 +Epoch 124, weight, value: tensor([[ 0.0494, -0.1000, -0.1150, ..., -0.0790, -0.0777, -0.0708], + [ 0.0094, 0.0453, -0.0290, ..., -0.0159, 0.1252, 0.0168], + [-0.0143, -0.1123, -0.0168, ..., 0.1168, -0.0956, 0.0207], + ..., + [-0.0470, 0.1010, 0.0331, ..., -0.0774, 0.0847, 0.0387], + [-0.0049, -0.0495, 0.0078, ..., -0.0334, -0.1344, -0.0775], + [-0.0090, -0.0032, -0.0019, ..., -0.1247, -0.0982, 0.0289]], + device='cuda:0'), grad: tensor([[-2.8420e-08, 4.7777e-07, 0.0000e+00, ..., 2.8647e-06, + 3.7765e-07, 1.6587e-06], + [-6.7018e-06, -1.2726e-05, 0.0000e+00, ..., 9.4157e-07, + -3.2037e-05, -1.4141e-05], + [ 2.7823e-07, -1.8450e-06, 0.0000e+00, ..., -1.8626e-05, + 8.8988e-07, -1.8194e-05], + ..., + [ 4.7572e-06, 8.9332e-06, 0.0000e+00, ..., 5.3532e-06, + 2.2948e-05, 1.6779e-05], + [ 1.0226e-06, 1.8682e-06, 0.0000e+00, ..., 6.2734e-06, + 3.2205e-06, 4.3064e-06], + [-3.0093e-08, 6.2073e-07, 0.0000e+00, ..., 3.2913e-06, + 9.8068e-07, 1.7965e-06]], device='cuda:0') +Epoch 124, bias, value: tensor([-0.0305, 0.0078, 0.0129, 0.0295, 0.0263, -0.0188, 0.0263, 0.0318, + -0.0080, 0.0054], device='cuda:0'), grad: tensor([ 8.7842e-06, -3.6329e-05, -5.8115e-05, 1.3024e-05, 1.8924e-06, + 9.3877e-06, -1.6302e-05, 4.7505e-05, 2.2292e-05, 7.8380e-06], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 123---------------------------------------------------- +epoch 123, time 249.64, cls_loss 0.0023 cls_loss_mapping 0.0079 cls_loss_causal 0.5751 re_mapping 0.0059 re_causal 0.0184 /// teacc 99.15 lr 0.00010000 +Epoch 125, weight, value: tensor([[ 0.0492, -0.1011, -0.1150, ..., -0.0802, -0.0793, -0.0717], + [ 0.0114, 0.0458, -0.0290, ..., -0.0162, 0.1261, 0.0175], + [-0.0144, -0.1132, -0.0167, ..., 0.1182, -0.0966, 0.0200], + ..., + [-0.0487, 0.1013, 0.0331, ..., -0.0773, 0.0845, 0.0389], + [-0.0048, -0.0497, 0.0078, ..., -0.0339, -0.1347, -0.0785], + [-0.0089, -0.0036, -0.0019, ..., -0.1254, -0.0992, 0.0287]], + device='cuda:0'), grad: tensor([[ 7.2003e-08, 1.2119e-07, 0.0000e+00, ..., 1.0775e-06, + 2.4898e-08, 6.5425e-07], + [ 3.0976e-06, 5.7220e-06, 0.0000e+00, ..., 4.4443e-06, + 3.5688e-06, 1.9982e-05], + [-2.5984e-07, 1.8743e-07, 0.0000e+00, ..., -1.0151e-06, + 1.1764e-07, -1.1213e-06], + ..., + [ 3.2205e-06, 5.3823e-05, 0.0000e+00, ..., 1.1418e-06, + -7.3051e-08, 2.1428e-05], + [ 1.1504e-05, -6.3360e-05, 0.0000e+00, ..., 1.0831e-06, + 2.1118e-07, -1.5691e-05], + [-2.5122e-07, 4.7833e-06, 0.0000e+00, ..., 3.7681e-06, + 1.8254e-06, 2.0061e-06]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0313, 0.0088, 0.0135, 0.0288, 0.0259, -0.0182, 0.0254, 0.0315, + -0.0084, 0.0054], device='cuda:0'), grad: tensor([ 1.3269e-05, 6.0797e-05, 1.4892e-06, 3.9130e-05, -2.1577e-04, + -4.9591e-05, 8.0585e-05, 2.0587e-04, -1.6022e-04, 2.4870e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 248.89, cls_loss 0.0022 cls_loss_mapping 0.0066 cls_loss_causal 0.5331 re_mapping 0.0056 re_causal 0.0167 /// teacc 99.00 lr 0.00010000 +Epoch 126, weight, value: tensor([[ 0.0500, -0.1013, -0.1150, ..., -0.0805, -0.0798, -0.0723], + [ 0.0125, 0.0465, -0.0290, ..., -0.0165, 0.1271, 0.0180], + [-0.0135, -0.1133, -0.0167, ..., 0.1190, -0.0968, 0.0207], + ..., + [-0.0500, 0.1009, 0.0331, ..., -0.0776, 0.0840, 0.0385], + [-0.0054, -0.0498, 0.0078, ..., -0.0354, -0.1350, -0.0794], + [-0.0087, -0.0036, -0.0019, ..., -0.1260, -0.1000, 0.0292]], + device='cuda:0'), grad: tensor([[-6.0070e-07, 7.2469e-08, 0.0000e+00, ..., 6.7335e-07, + 1.0064e-07, 2.4796e-07], + [ 2.3469e-06, 1.4226e-07, 0.0000e+00, ..., 4.1574e-06, + -6.8336e-08, 4.5225e-06], + [ 1.7341e-06, 5.6485e-07, 0.0000e+00, ..., -1.9409e-06, + -7.0664e-08, -3.4645e-06], + ..., + [ 9.6858e-07, -2.3935e-06, 0.0000e+00, ..., 3.2573e-07, + -2.1122e-06, -2.3879e-06], + [-1.7405e-05, 1.6508e-07, 0.0000e+00, ..., 1.5199e-06, + 3.5227e-07, -4.1090e-06], + [ 3.5278e-06, 7.7765e-07, 0.0000e+00, ..., 2.6845e-07, + 8.8010e-07, 1.3383e-06]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0304, 0.0096, 0.0139, 0.0285, 0.0248, -0.0164, 0.0238, 0.0306, + -0.0092, 0.0061], device='cuda:0'), grad: tensor([-2.7791e-06, 1.2547e-05, 1.0051e-05, 1.4596e-05, 5.3421e-06, + 1.9789e-05, -1.2755e-05, -3.6880e-07, -6.3002e-05, 1.6555e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 248.92, cls_loss 0.0016 cls_loss_mapping 0.0052 cls_loss_causal 0.5238 re_mapping 0.0058 re_causal 0.0174 /// teacc 98.98 lr 0.00010000 +Epoch 127, weight, value: tensor([[ 0.0503, -0.1016, -0.1150, ..., -0.0807, -0.0805, -0.0726], + [ 0.0127, 0.0465, -0.0287, ..., -0.0169, 0.1273, 0.0180], + [-0.0139, -0.1134, -0.0167, ..., 0.1193, -0.0972, 0.0208], + ..., + [-0.0503, 0.1011, 0.0328, ..., -0.0780, 0.0840, 0.0385], + [-0.0052, -0.0498, 0.0078, ..., -0.0360, -0.1354, -0.0796], + [-0.0089, -0.0040, -0.0020, ..., -0.1272, -0.1007, 0.0289]], + device='cuda:0'), grad: tensor([[ 6.9849e-08, 2.3714e-07, 7.0606e-08, ..., 4.2561e-07, + 7.6089e-07, 4.0373e-07], + [-8.3297e-06, -1.0401e-05, -3.8259e-06, ..., 2.8615e-07, + -3.7968e-05, -1.3158e-05], + [ 3.7532e-07, 1.4063e-06, 3.0524e-07, ..., -1.0803e-06, + 3.7793e-06, 5.6066e-07], + ..., + [ 2.4159e-06, 1.3895e-06, 1.0002e-06, ..., 2.4843e-07, + 8.1286e-06, 2.3395e-06], + [ 5.7928e-07, 1.0245e-06, 3.7323e-07, ..., 6.2259e-07, + 3.7663e-06, 1.5302e-06], + [ 6.8452e-07, 1.1576e-06, 2.5611e-07, ..., 6.3097e-07, + 3.0193e-06, 1.6978e-06]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0300, 0.0095, 0.0136, 0.0283, 0.0257, -0.0162, 0.0238, 0.0306, + -0.0090, 0.0055], device='cuda:0'), grad: tensor([ 1.4529e-06, -5.5045e-05, 5.8003e-06, 1.4268e-06, 1.5169e-05, + 7.6070e-06, -1.4552e-07, 1.2286e-05, 4.0606e-06, 7.3761e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 249.34, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5671 re_mapping 0.0055 re_causal 0.0180 /// teacc 98.94 lr 0.00010000 +Epoch 128, weight, value: tensor([[ 0.0497, -0.1017, -0.1151, ..., -0.0805, -0.0813, -0.0736], + [ 0.0129, 0.0465, -0.0280, ..., -0.0175, 0.1274, 0.0179], + [-0.0138, -0.1141, -0.0168, ..., 0.1193, -0.0985, 0.0198], + ..., + [-0.0507, 0.1016, 0.0325, ..., -0.0778, 0.0846, 0.0393], + [-0.0048, -0.0500, 0.0077, ..., -0.0366, -0.1356, -0.0803], + [-0.0084, -0.0045, -0.0020, ..., -0.1280, -0.1018, 0.0284]], + device='cuda:0'), grad: tensor([[-8.7311e-11, 1.4188e-08, 0.0000e+00, ..., 2.0710e-07, + 4.4820e-08, 1.5879e-07], + [-7.2597e-07, 1.1129e-07, 0.0000e+00, ..., 1.7590e-07, + -1.1241e-06, -4.6473e-07], + [ 1.9476e-07, 1.8044e-08, 0.0000e+00, ..., -1.2852e-07, + 2.5728e-07, -9.0746e-08], + ..., + [ 2.9290e-07, -4.4517e-07, 0.0000e+00, ..., 1.9569e-07, + -2.6077e-07, 3.1153e-07], + [-4.9779e-07, 2.6499e-08, 0.0000e+00, ..., 5.4389e-07, + 3.5274e-07, 6.4122e-07], + [-3.3132e-07, 1.3248e-07, 0.0000e+00, ..., 2.0731e-06, + 1.8091e-07, 2.7791e-06]], device='cuda:0') +Epoch 128, bias, value: tensor([-0.0305, 0.0092, 0.0125, 0.0283, 0.0264, -0.0164, 0.0240, 0.0312, + -0.0087, 0.0053], device='cuda:0'), grad: tensor([ 6.2585e-07, -1.0636e-06, 8.5402e-07, 2.0918e-06, 1.3858e-05, + 2.5202e-06, -2.4989e-05, 1.2536e-06, -2.1290e-06, 6.9961e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 249.19, cls_loss 0.0019 cls_loss_mapping 0.0056 cls_loss_causal 0.5435 re_mapping 0.0058 re_causal 0.0170 /// teacc 98.97 lr 0.00010000 +Epoch 129, weight, value: tensor([[ 0.0501, -0.1019, -0.1151, ..., -0.0809, -0.0819, -0.0739], + [ 0.0136, 0.0466, -0.0278, ..., -0.0177, 0.1279, 0.0181], + [-0.0126, -0.1146, -0.0168, ..., 0.1203, -0.0997, 0.0198], + ..., + [-0.0522, 0.1017, 0.0323, ..., -0.0787, 0.0845, 0.0392], + [-0.0051, -0.0502, 0.0077, ..., -0.0377, -0.1362, -0.0812], + [-0.0075, -0.0042, -0.0021, ..., -0.1280, -0.1011, 0.0289]], + device='cuda:0'), grad: tensor([[-2.6426e-07, 4.2521e-08, 4.2201e-10, ..., 7.7188e-06, + 4.7759e-08, 2.4843e-07], + [-6.9034e-08, -2.4738e-10, 1.6007e-10, ..., 9.1223e-07, + -5.1782e-07, 8.0618e-08], + [ 4.6962e-07, 1.5250e-07, 2.0373e-10, ..., 1.7267e-06, + 1.7742e-07, 2.1351e-07], + ..., + [ 4.5868e-07, -1.0943e-06, 2.1828e-10, ..., 4.6974e-08, + -7.3714e-07, -8.3353e-07], + [ 9.3924e-07, 9.8546e-08, 8.0036e-10, ..., 1.3992e-05, + 1.9441e-07, 1.4678e-06], + [-2.6915e-06, 5.0385e-07, 1.1059e-09, ..., 3.2899e-07, + 4.3120e-07, -2.9728e-06]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0302, 0.0095, 0.0126, 0.0273, 0.0266, -0.0160, 0.0241, 0.0309, + -0.0092, 0.0057], device='cuda:0'), grad: tensor([ 1.2852e-05, 2.9337e-06, 6.4038e-06, -4.8541e-06, 4.6194e-06, + 1.1042e-05, -5.7489e-05, 1.0701e-06, 3.6895e-05, -1.3493e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 248.74, cls_loss 0.0020 cls_loss_mapping 0.0053 cls_loss_causal 0.5204 re_mapping 0.0054 re_causal 0.0164 /// teacc 98.97 lr 0.00010000 +Epoch 130, weight, value: tensor([[ 0.0503, -0.1023, -0.1152, ..., -0.0818, -0.0824, -0.0743], + [ 0.0135, 0.0465, -0.0282, ..., -0.0207, 0.1277, 0.0176], + [-0.0113, -0.1147, -0.0164, ..., 0.1220, -0.0987, 0.0213], + ..., + [-0.0525, 0.1020, 0.0323, ..., -0.0791, 0.0847, 0.0393], + [-0.0051, -0.0503, 0.0077, ..., -0.0384, -0.1366, -0.0818], + [-0.0073, -0.0048, -0.0021, ..., -0.1287, -0.1019, 0.0291]], + device='cuda:0'), grad: tensor([[ 4.7078e-07, 4.2119e-07, 0.0000e+00, ..., 3.6112e-07, + 8.3959e-07, 6.9477e-07], + [-1.1429e-05, -1.0692e-05, 0.0000e+00, ..., 2.1944e-07, + -2.1040e-05, -8.3670e-06], + [ 2.1607e-07, 2.3786e-06, 0.0000e+00, ..., -3.7570e-06, + 4.4741e-06, -2.4829e-06], + ..., + [ 7.5586e-06, 5.2601e-06, 0.0000e+00, ..., 2.9914e-06, + 1.0252e-05, 7.7337e-06], + [ 7.7719e-07, 5.5414e-07, 0.0000e+00, ..., 3.6531e-07, + 1.3197e-06, 8.1584e-07], + [ 7.0361e-07, 7.4506e-07, 0.0000e+00, ..., 2.5448e-07, + 1.3886e-06, 5.3737e-07]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0302, 0.0089, 0.0134, 0.0269, 0.0269, -0.0158, 0.0238, 0.0310, + -0.0095, 0.0062], device='cuda:0'), grad: tensor([ 1.8869e-06, -3.6180e-05, -2.9453e-07, 1.3681e-06, 1.4435e-06, + 1.7295e-06, -5.2853e-07, 2.5302e-05, 2.9132e-06, 2.3283e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 248.70, cls_loss 0.0016 cls_loss_mapping 0.0048 cls_loss_causal 0.5330 re_mapping 0.0054 re_causal 0.0169 /// teacc 98.99 lr 0.00010000 +Epoch 131, weight, value: tensor([[ 0.0509, -0.1025, -0.1152, ..., -0.0836, -0.0833, -0.0745], + [ 0.0137, 0.0465, -0.0284, ..., -0.0212, 0.1278, 0.0175], + [-0.0113, -0.1149, -0.0162, ..., 0.1229, -0.0989, 0.0219], + ..., + [-0.0526, 0.1024, 0.0323, ..., -0.0798, 0.0849, 0.0395], + [-0.0053, -0.0505, 0.0077, ..., -0.0393, -0.1371, -0.0823], + [-0.0072, -0.0055, -0.0021, ..., -0.1293, -0.1026, 0.0290]], + device='cuda:0'), grad: tensor([[-7.1116e-06, 5.4250e-07, 0.0000e+00, ..., -3.4496e-06, + 8.1956e-07, 1.1846e-06], + [-2.4284e-07, 2.7716e-06, 0.0000e+00, ..., 1.3178e-06, + 1.1995e-06, 4.2021e-06], + [ 6.5286e-07, 4.4219e-06, 0.0000e+00, ..., 2.6748e-06, + 5.9232e-06, 7.4506e-06], + ..., + [ 1.2899e-06, -1.9789e-05, 0.0000e+00, ..., 1.7881e-06, + -2.6062e-05, -1.9327e-05], + [ 1.2210e-06, 3.2224e-06, 0.0000e+00, ..., 1.7369e-06, + 3.5353e-06, 4.8839e-06], + [-4.1188e-07, 4.1574e-06, 0.0000e+00, ..., 8.6725e-06, + 6.5379e-06, 1.1280e-05]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0304, 0.0088, 0.0136, 0.0267, 0.0265, -0.0159, 0.0243, 0.0312, + -0.0099, 0.0060], device='cuda:0'), grad: tensor([-3.2037e-05, 1.3158e-05, 2.4840e-05, 8.4713e-06, -1.5117e-05, + 1.0334e-05, -7.7412e-06, -4.6194e-05, 8.1435e-06, 3.6150e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 248.92, cls_loss 0.0019 cls_loss_mapping 0.0063 cls_loss_causal 0.5710 re_mapping 0.0054 re_causal 0.0172 /// teacc 98.88 lr 0.00010000 +Epoch 132, weight, value: tensor([[ 0.0511, -0.1030, -0.1153, ..., -0.0834, -0.0838, -0.0751], + [ 0.0143, 0.0468, -0.0278, ..., -0.0215, 0.1284, 0.0177], + [-0.0109, -0.1154, -0.0162, ..., 0.1234, -0.0995, 0.0223], + ..., + [-0.0531, 0.1027, 0.0319, ..., -0.0798, 0.0851, 0.0397], + [-0.0051, -0.0507, 0.0077, ..., -0.0406, -0.1377, -0.0828], + [-0.0074, -0.0065, -0.0021, ..., -0.1311, -0.1044, 0.0285]], + device='cuda:0'), grad: tensor([[-5.6103e-06, 9.6043e-08, 0.0000e+00, ..., -1.3467e-06, + 1.7078e-07, 5.3085e-07], + [-5.6103e-06, -4.0270e-06, 0.0000e+00, ..., -8.6054e-07, + -7.8976e-06, -1.1632e-06], + [ 4.5747e-06, 3.5227e-07, 0.0000e+00, ..., 1.0179e-06, + 6.6636e-07, -1.1697e-06], + ..., + [ 1.8878e-06, 4.3563e-07, 0.0000e+00, ..., 9.3644e-07, + 1.1837e-06, 4.5821e-06], + [ 4.2319e-06, 2.3302e-06, 0.0000e+00, ..., 1.1791e-06, + 4.4182e-06, 2.4904e-06], + [ 1.0328e-06, 1.3143e-07, 0.0000e+00, ..., 7.1293e-07, + 2.0023e-07, -7.1637e-06]], device='cuda:0') +Epoch 132, bias, value: tensor([-0.0302, 0.0091, 0.0135, 0.0268, 0.0261, -0.0162, 0.0248, 0.0315, + -0.0101, 0.0053], device='cuda:0'), grad: tensor([-2.0623e-05, -9.7528e-06, 1.5222e-05, -5.2303e-06, 3.3118e-06, + -2.0359e-06, 3.2503e-06, 1.3217e-05, 1.2577e-05, -9.9391e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 248.79, cls_loss 0.0017 cls_loss_mapping 0.0046 cls_loss_causal 0.4882 re_mapping 0.0051 re_causal 0.0159 /// teacc 98.97 lr 0.00010000 +Epoch 133, weight, value: tensor([[ 0.0511, -0.1035, -0.1154, ..., -0.0840, -0.0846, -0.0756], + [ 0.0150, 0.0476, -0.0277, ..., -0.0226, 0.1293, 0.0181], + [-0.0109, -0.1156, -0.0162, ..., 0.1241, -0.0997, 0.0225], + ..., + [-0.0541, 0.1022, 0.0318, ..., -0.0800, 0.0844, 0.0392], + [-0.0050, -0.0509, 0.0077, ..., -0.0410, -0.1379, -0.0833], + [-0.0073, -0.0064, -0.0021, ..., -0.1313, -0.1046, 0.0291]], + device='cuda:0'), grad: tensor([[-2.4587e-06, 2.8216e-08, 0.0000e+00, ..., -1.8382e-07, + 5.3813e-08, 1.0064e-07], + [-5.1456e-07, -2.2212e-07, 0.0000e+00, ..., 2.6240e-07, + -1.0580e-06, -6.9849e-08], + [ 3.2317e-07, 2.2061e-07, 0.0000e+00, ..., -1.7835e-06, + 2.2666e-07, -1.5954e-06], + ..., + [ 6.7335e-07, -4.5914e-07, 0.0000e+00, ..., 1.6550e-06, + -1.3225e-07, 1.1576e-06], + [ 5.1502e-07, 9.9011e-08, 0.0000e+00, ..., 8.5216e-07, + 2.9569e-07, 5.1968e-07], + [-5.9488e-08, 1.0408e-07, 0.0000e+00, ..., 3.8301e-07, + 1.3434e-07, -8.8988e-07]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0304, 0.0094, 0.0135, 0.0272, 0.0253, -0.0161, 0.0249, 0.0309, + -0.0104, 0.0061], device='cuda:0'), grad: tensor([-9.5218e-06, -3.8091e-07, -2.0443e-07, 3.4235e-06, 2.3246e-06, + 3.0529e-06, -4.5970e-06, 3.8669e-06, 3.6713e-06, -1.6764e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 248.90, cls_loss 0.0018 cls_loss_mapping 0.0050 cls_loss_causal 0.5402 re_mapping 0.0053 re_causal 0.0164 /// teacc 98.99 lr 0.00010000 +Epoch 134, weight, value: tensor([[ 0.0514, -0.1039, -0.1154, ..., -0.0840, -0.0859, -0.0759], + [ 0.0163, 0.0478, -0.0277, ..., -0.0218, 0.1302, 0.0186], + [-0.0118, -0.1158, -0.0162, ..., 0.1243, -0.1012, 0.0224], + ..., + [-0.0549, 0.1022, 0.0318, ..., -0.0807, 0.0842, 0.0391], + [-0.0048, -0.0512, 0.0077, ..., -0.0414, -0.1386, -0.0838], + [-0.0078, -0.0066, -0.0021, ..., -0.1319, -0.1052, 0.0291]], + device='cuda:0'), grad: tensor([[ 5.9896e-08, 6.4867e-07, 1.4552e-11, ..., 1.8142e-06, + 1.3923e-07, 3.6005e-06], + [-1.0873e-07, 3.9185e-07, -8.8767e-10, ..., 3.1441e-06, + -1.1707e-06, 6.3032e-06], + [ 1.5860e-06, 8.8383e-07, 1.1642e-10, ..., -7.2457e-07, + 2.7078e-07, 2.3786e-06], + ..., + [ 1.3821e-06, 1.8135e-05, 5.6752e-10, ..., 1.8567e-05, + 5.0329e-06, 7.9036e-05], + [ 2.0284e-06, 1.3569e-06, 5.8208e-11, ..., 5.4725e-06, + 3.0687e-07, 1.0908e-05], + [ 7.5391e-07, -1.4424e-05, 4.3656e-11, ..., 1.4246e-05, + -5.2415e-06, -2.2411e-05]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0302, 0.0101, 0.0131, 0.0270, 0.0253, -0.0162, 0.0252, 0.0305, + -0.0105, 0.0059], device='cuda:0'), grad: tensor([ 1.2942e-05, 2.4348e-05, 1.9461e-05, -3.7611e-05, -3.0327e-04, + 2.1636e-05, 2.4691e-05, 2.5606e-04, 4.3780e-05, -6.2227e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 248.49, cls_loss 0.0015 cls_loss_mapping 0.0046 cls_loss_causal 0.5176 re_mapping 0.0052 re_causal 0.0166 /// teacc 99.07 lr 0.00010000 +Epoch 135, weight, value: tensor([[ 0.0519, -0.1038, -0.1154, ..., -0.0835, -0.0862, -0.0761], + [ 0.0168, 0.0478, -0.0277, ..., -0.0220, 0.1305, 0.0186], + [-0.0122, -0.1160, -0.0162, ..., 0.1245, -0.1015, 0.0224], + ..., + [-0.0552, 0.1023, 0.0318, ..., -0.0809, 0.0842, 0.0390], + [-0.0046, -0.0513, 0.0077, ..., -0.0418, -0.1391, -0.0843], + [-0.0079, -0.0068, -0.0021, ..., -0.1326, -0.1058, 0.0292]], + device='cuda:0'), grad: tensor([[-3.7224e-08, 3.0338e-07, 0.0000e+00, ..., 4.8778e-08, + 2.8242e-07, 2.5355e-07], + [-7.2736e-07, 2.0284e-06, 0.0000e+00, ..., 5.8615e-08, + 4.8010e-07, 1.1278e-06], + [ 7.6834e-08, 1.3197e-06, 0.0000e+00, ..., -1.1898e-07, + 1.2303e-06, 8.2748e-07], + ..., + [ 9.4180e-08, -9.6560e-06, 0.0000e+00, ..., 1.1502e-07, + -8.0466e-06, -6.4783e-06], + [ 4.6380e-07, 8.7451e-07, 0.0000e+00, ..., 9.6683e-08, + 1.0896e-06, 8.1491e-07], + [ 3.8446e-08, 2.7623e-06, 0.0000e+00, ..., 2.6706e-07, + 2.4494e-06, 2.0340e-06]], device='cuda:0') +Epoch 135, bias, value: tensor([-0.0292, 0.0103, 0.0126, 0.0274, 0.0256, -0.0164, 0.0252, 0.0303, + -0.0104, 0.0057], device='cuda:0'), grad: tensor([ 8.2934e-07, 2.6617e-06, 2.9504e-06, 2.1588e-06, 8.9686e-07, + -4.5709e-06, 4.3884e-06, -1.9789e-05, 4.1164e-06, 6.3293e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 249.12, cls_loss 0.0019 cls_loss_mapping 0.0054 cls_loss_causal 0.5368 re_mapping 0.0052 re_causal 0.0159 /// teacc 99.05 lr 0.00010000 +Epoch 136, weight, value: tensor([[ 0.0510, -0.1042, -0.1154, ..., -0.0842, -0.0873, -0.0766], + [ 0.0173, 0.0476, -0.0277, ..., -0.0222, 0.1306, 0.0184], + [-0.0121, -0.1162, -0.0162, ..., 0.1252, -0.1019, 0.0227], + ..., + [-0.0554, 0.1028, 0.0318, ..., -0.0812, 0.0846, 0.0395], + [-0.0048, -0.0517, 0.0077, ..., -0.0429, -0.1398, -0.0850], + [-0.0079, -0.0072, -0.0021, ..., -0.1333, -0.1065, 0.0290]], + device='cuda:0'), grad: tensor([[ 2.4840e-05, 2.3618e-08, 0.0000e+00, ..., 1.9884e-07, + 7.0839e-08, 1.5821e-07], + [ 1.7218e-07, -4.7404e-07, 0.0000e+00, ..., 1.9709e-07, + -1.4268e-06, -2.1327e-07], + [ 2.9821e-06, 1.4249e-07, 0.0000e+00, ..., -5.9837e-07, + 4.3516e-07, -3.6019e-07], + ..., + [ 5.6205e-07, 1.5146e-07, 0.0000e+00, ..., 2.9313e-07, + 3.5460e-07, 4.6380e-07], + [-3.7014e-05, 3.9057e-08, 0.0000e+00, ..., -5.2666e-07, + 1.1362e-07, 1.9453e-07], + [ 5.0217e-06, 1.8961e-08, 0.0000e+00, ..., 2.5937e-07, + 4.1095e-08, -4.7521e-07]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0303, 0.0101, 0.0128, 0.0272, 0.0252, -0.0161, 0.0256, 0.0309, + -0.0112, 0.0053], device='cuda:0'), grad: tensor([ 6.0886e-05, 4.3167e-07, 1.0386e-05, 8.0466e-07, 4.2492e-07, + 1.2917e-06, 4.9807e-06, 3.0529e-06, -9.3520e-05, 1.1325e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 249.08, cls_loss 0.0018 cls_loss_mapping 0.0043 cls_loss_causal 0.5394 re_mapping 0.0053 re_causal 0.0163 /// teacc 99.10 lr 0.00010000 +Epoch 137, weight, value: tensor([[ 0.0509, -0.1043, -0.1154, ..., -0.0847, -0.0876, -0.0771], + [ 0.0176, 0.0476, -0.0277, ..., -0.0225, 0.1308, 0.0185], + [-0.0126, -0.1166, -0.0162, ..., 0.1257, -0.1025, 0.0224], + ..., + [-0.0556, 0.1032, 0.0318, ..., -0.0813, 0.0849, 0.0398], + [-0.0038, -0.0517, 0.0077, ..., -0.0431, -0.1400, -0.0853], + [-0.0080, -0.0073, -0.0021, ..., -0.1341, -0.1067, 0.0283]], + device='cuda:0'), grad: tensor([[-5.2713e-07, 1.1656e-08, 0.0000e+00, ..., 2.7358e-09, + 1.4377e-08, 1.1918e-08], + [ 5.3970e-07, 2.4284e-07, 0.0000e+00, ..., 3.3615e-09, + 5.8702e-08, 3.4983e-08], + [ 6.9384e-08, 2.2643e-08, 0.0000e+00, ..., 1.3970e-09, + 3.8097e-08, 1.5760e-08], + ..., + [ 3.7975e-07, 5.1892e-08, 0.0000e+00, ..., 4.0454e-09, + 3.1170e-08, -1.1933e-08], + [ 5.2014e-07, 2.1805e-07, 0.0000e+00, ..., -2.1362e-08, + 2.1292e-07, 6.7463e-08], + [ 4.9919e-07, 1.3341e-07, 0.0000e+00, ..., 3.8999e-09, + 1.3411e-07, -1.9174e-07]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0306, 0.0101, 0.0126, 0.0261, 0.0275, -0.0158, 0.0255, 0.0311, + -0.0105, 0.0037], device='cuda:0'), grad: tensor([-3.3602e-06, 4.7646e-06, 3.4738e-07, -8.1360e-06, 8.0978e-07, + -6.0014e-06, 2.9169e-06, 2.5798e-06, 3.5036e-06, 2.5388e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 248.84, cls_loss 0.0021 cls_loss_mapping 0.0060 cls_loss_causal 0.5336 re_mapping 0.0049 re_causal 0.0157 /// teacc 99.02 lr 0.00010000 +Epoch 138, weight, value: tensor([[ 0.0526, -0.1015, -0.1155, ..., -0.0844, -0.0881, -0.0775], + [ 0.0177, 0.0476, -0.0277, ..., -0.0229, 0.1310, 0.0184], + [-0.0110, -0.1172, -0.0160, ..., 0.1275, -0.1028, 0.0235], + ..., + [-0.0559, 0.1035, 0.0318, ..., -0.0818, 0.0850, 0.0397], + [-0.0048, -0.0519, 0.0077, ..., -0.0451, -0.1404, -0.0873], + [-0.0081, -0.0075, -0.0021, ..., -0.1350, -0.1070, 0.0282]], + device='cuda:0'), grad: tensor([[-5.4808e-07, 1.1368e-07, 0.0000e+00, ..., -4.4680e-07, + 3.9086e-08, 4.1607e-07], + [-1.1493e-06, 5.7295e-06, 0.0000e+00, ..., -1.1671e-07, + 1.7853e-06, 8.2478e-06], + [ 1.1446e-06, 9.2387e-07, 0.0000e+00, ..., 5.4995e-07, + 1.6885e-06, 1.1269e-06], + ..., + [ 3.2503e-07, -4.4964e-06, 0.0000e+00, ..., 1.8328e-06, + -5.1670e-06, -1.4296e-06], + [-1.2340e-06, 6.3051e-07, 0.0000e+00, ..., 4.1211e-07, + 1.0151e-06, 1.1083e-06], + [ 5.6252e-07, 3.1339e-07, 0.0000e+00, ..., 4.5728e-07, + 1.2550e-07, 1.5227e-06]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0289, 0.0099, 0.0141, 0.0259, 0.0282, -0.0161, 0.0255, 0.0309, + -0.0117, 0.0035], device='cuda:0'), grad: tensor([-2.5071e-06, 1.7270e-05, 1.7598e-05, -2.9519e-05, -3.5673e-05, + 1.5765e-05, 5.0217e-06, 5.3719e-06, -2.2619e-07, 6.8732e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 248.84, cls_loss 0.0021 cls_loss_mapping 0.0064 cls_loss_causal 0.5411 re_mapping 0.0052 re_causal 0.0158 /// teacc 98.93 lr 0.00010000 +Epoch 139, weight, value: tensor([[ 0.0525, -0.1012, -0.1155, ..., -0.0829, -0.0896, -0.0781], + [ 0.0181, 0.0476, -0.0277, ..., -0.0229, 0.1314, 0.0186], + [-0.0133, -0.1182, -0.0160, ..., 0.1252, -0.1038, 0.0235], + ..., + [-0.0562, 0.1037, 0.0318, ..., -0.0825, 0.0852, 0.0398], + [-0.0025, -0.0510, 0.0077, ..., -0.0419, -0.1408, -0.0875], + [-0.0079, -0.0073, -0.0021, ..., -0.1357, -0.1073, 0.0286]], + device='cuda:0'), grad: tensor([[ 6.2445e-07, 5.5367e-07, 0.0000e+00, ..., 8.0327e-07, + 1.2852e-06, 9.7975e-07], + [-1.7017e-05, -1.6421e-05, 0.0000e+00, ..., -4.2305e-07, + -3.6895e-05, -1.0200e-05], + [ 2.0750e-06, 9.1782e-07, 0.0000e+00, ..., -1.1683e-05, + 2.4922e-06, -8.6874e-06], + ..., + [ 5.3011e-06, 1.8865e-05, 0.0000e+00, ..., 6.6720e-06, + 5.1975e-05, 2.0623e-05], + [ 2.1011e-05, 1.4707e-05, 0.0000e+00, ..., 2.6524e-06, + 1.4074e-05, 1.6659e-05], + [-1.5184e-05, -8.1435e-06, 0.0000e+00, ..., 1.8324e-07, + 4.9062e-06, -1.1988e-05]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0291, 0.0101, 0.0118, 0.0260, 0.0280, -0.0164, 0.0249, 0.0306, + -0.0088, 0.0043], device='cuda:0'), grad: tensor([ 5.4166e-06, -4.6104e-05, -1.7568e-05, -2.4724e-04, 1.3009e-05, + 3.2801e-06, 9.8571e-06, 2.2161e-04, 7.5102e-05, -1.7673e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 248.94, cls_loss 0.0016 cls_loss_mapping 0.0044 cls_loss_causal 0.5235 re_mapping 0.0055 re_causal 0.0161 /// teacc 99.02 lr 0.00010000 +Epoch 140, weight, value: tensor([[ 0.0525, -0.1016, -0.1156, ..., -0.0833, -0.0907, -0.0786], + [ 0.0185, 0.0476, -0.0277, ..., -0.0230, 0.1317, 0.0186], + [-0.0132, -0.1185, -0.0160, ..., 0.1256, -0.1038, 0.0242], + ..., + [-0.0565, 0.1038, 0.0318, ..., -0.0832, 0.0852, 0.0396], + [-0.0026, -0.0514, 0.0077, ..., -0.0420, -0.1420, -0.0881], + [-0.0077, -0.0067, -0.0022, ..., -0.1364, -0.1073, 0.0290]], + device='cuda:0'), grad: tensor([[ 1.4175e-06, 4.9651e-08, 0.0000e+00, ..., 3.7742e-07, + 8.3295e-08, 1.1725e-06], + [ 1.0673e-06, -4.5868e-07, 0.0000e+00, ..., 2.6561e-06, + -2.0862e-06, 2.3227e-06], + [-1.7909e-06, 2.2806e-07, 0.0000e+00, ..., -8.8513e-06, + 4.4261e-07, -6.5640e-06], + ..., + [ 2.7455e-06, -5.8347e-07, 0.0000e+00, ..., 5.1735e-07, + 3.1409e-07, 8.7405e-07], + [ 3.4110e-07, 1.0827e-07, 0.0000e+00, ..., 3.4571e-06, + 2.0326e-07, 6.1095e-06], + [-2.7299e-05, 1.6124e-07, 0.0000e+00, ..., 2.3120e-07, + 2.1420e-07, -2.0534e-05]], device='cuda:0') +Epoch 140, bias, value: tensor([-0.0291, 0.0102, 0.0120, 0.0263, 0.0280, -0.0163, 0.0249, 0.0301, + -0.0091, 0.0046], device='cuda:0'), grad: tensor([ 9.8050e-06, 1.0394e-05, -1.3188e-05, -1.5221e-08, 1.2094e-04, + 1.5214e-05, 7.0408e-06, 1.5706e-05, 8.0047e-07, -1.6689e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 248.85, cls_loss 0.0018 cls_loss_mapping 0.0053 cls_loss_causal 0.5306 re_mapping 0.0049 re_causal 0.0154 /// teacc 99.05 lr 0.00010000 +Epoch 141, weight, value: tensor([[ 0.0537, -0.0984, -0.1156, ..., -0.0835, -0.0881, -0.0762], + [ 0.0189, 0.0476, -0.0277, ..., -0.0232, 0.1319, 0.0186], + [-0.0133, -0.1188, -0.0160, ..., 0.1259, -0.1047, 0.0243], + ..., + [-0.0568, 0.1039, 0.0318, ..., -0.0838, 0.0855, 0.0395], + [-0.0029, -0.0525, 0.0077, ..., -0.0422, -0.1437, -0.0891], + [-0.0083, -0.0070, -0.0022, ..., -0.1372, -0.1077, 0.0289]], + device='cuda:0'), grad: tensor([[ 4.8336e-07, 2.7649e-08, 0.0000e+00, ..., 1.0962e-06, + 8.8301e-08, 5.2387e-07], + [-7.4506e-07, -4.7451e-07, 0.0000e+00, ..., 5.4063e-07, + -1.9800e-06, -2.8114e-08], + [-3.0617e-07, 1.0198e-07, 0.0000e+00, ..., -1.0105e-06, + 3.8301e-07, -3.7416e-07], + ..., + [ 5.7556e-07, 1.3271e-07, 0.0000e+00, ..., 8.4331e-07, + 5.3784e-07, 1.6922e-06], + [-9.1083e-07, 9.2201e-08, 0.0000e+00, ..., 2.1094e-07, + 3.4343e-07, 6.1048e-07], + [ 1.7957e-08, 5.2445e-08, 0.0000e+00, ..., 3.0193e-06, + 2.1048e-07, 4.9062e-06]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0267, 0.0102, 0.0118, 0.0265, 0.0282, -0.0164, 0.0251, 0.0300, + -0.0097, 0.0041], device='cuda:0'), grad: tensor([ 3.7998e-06, -6.1095e-07, 3.3900e-07, 2.0303e-06, -2.8849e-05, + 5.3924e-07, 7.3314e-06, 4.6454e-06, -2.9523e-06, 1.3746e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 248.50, cls_loss 0.0018 cls_loss_mapping 0.0047 cls_loss_causal 0.5473 re_mapping 0.0050 re_causal 0.0160 /// teacc 98.94 lr 0.00010000 +Epoch 142, weight, value: tensor([[ 0.0532, -0.0990, -0.1158, ..., -0.0847, -0.0885, -0.0778], + [ 0.0192, 0.0476, -0.0276, ..., -0.0234, 0.1322, 0.0185], + [-0.0135, -0.1191, -0.0160, ..., 0.1261, -0.1053, 0.0244], + ..., + [-0.0572, 0.1041, 0.0317, ..., -0.0842, 0.0856, 0.0396], + [-0.0025, -0.0526, 0.0077, ..., -0.0423, -0.1440, -0.0894], + [-0.0076, -0.0068, -0.0022, ..., -0.1385, -0.1080, 0.0295]], + device='cuda:0'), grad: tensor([[-2.3952e-08, 1.2945e-07, 0.0000e+00, ..., 6.4028e-08, + 1.5274e-07, 1.8289e-07], + [-3.7719e-07, 4.4005e-07, 0.0000e+00, ..., 1.8510e-08, + -6.2305e-07, 7.3388e-07], + [ 6.2049e-08, 6.3330e-07, 0.0000e+00, ..., -1.0344e-07, + 6.8638e-07, 6.3237e-07], + ..., + [ 4.7428e-07, -2.1362e-07, 0.0000e+00, ..., 1.9558e-08, + 7.7346e-07, 2.5257e-06], + [-4.7404e-07, 1.6124e-07, 0.0000e+00, ..., 2.0053e-08, + 2.8126e-07, 2.4098e-07], + [-1.4377e-07, -1.9819e-06, 0.0000e+00, ..., 9.3423e-09, + -2.4661e-06, -5.4613e-06]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0278, 0.0102, 0.0116, 0.0262, 0.0288, -0.0163, 0.0250, 0.0297, + -0.0093, 0.0045], device='cuda:0'), grad: tensor([ 3.4040e-07, 1.9073e-06, 2.7381e-06, -1.0490e-05, 1.1576e-06, + 4.4107e-06, 1.0170e-06, 1.0662e-05, -1.6810e-07, -1.1563e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 248.74, cls_loss 0.0015 cls_loss_mapping 0.0045 cls_loss_causal 0.5424 re_mapping 0.0051 re_causal 0.0165 /// teacc 99.00 lr 0.00010000 +Epoch 143, weight, value: tensor([[ 0.0529, -0.0994, -0.1159, ..., -0.0857, -0.0891, -0.0781], + [ 0.0199, 0.0476, -0.0276, ..., -0.0232, 0.1325, 0.0184], + [-0.0133, -0.1191, -0.0160, ..., 0.1264, -0.1057, 0.0247], + ..., + [-0.0576, 0.1043, 0.0317, ..., -0.0846, 0.0857, 0.0396], + [-0.0029, -0.0528, 0.0077, ..., -0.0424, -0.1444, -0.0897], + [-0.0070, -0.0065, -0.0022, ..., -0.1389, -0.1083, 0.0302]], + device='cuda:0'), grad: tensor([[ 1.1205e-07, 3.1316e-08, 0.0000e+00, ..., 3.6974e-07, + 3.4546e-08, 5.7567e-08], + [ 1.7544e-07, 1.6007e-07, 0.0000e+00, ..., 3.8045e-07, + 1.2328e-07, 1.5297e-07], + [-2.7437e-06, 6.9034e-08, 0.0000e+00, ..., -4.5672e-06, + 8.4750e-08, -3.7043e-07], + ..., + [ 9.3540e-08, -6.6822e-07, 0.0000e+00, ..., 1.1222e-07, + -6.9477e-07, -4.6985e-07], + [ 3.8464e-07, 5.9779e-08, 0.0000e+00, ..., 6.8080e-07, + 7.5321e-08, 9.9128e-08], + [ 8.4401e-09, 2.1164e-07, 0.0000e+00, ..., 5.3609e-08, + 2.2387e-07, 1.3039e-07]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0285, 0.0103, 0.0118, 0.0276, 0.0284, -0.0170, 0.0249, 0.0296, + -0.0097, 0.0052], device='cuda:0'), grad: tensor([ 4.9314e-07, 9.8720e-07, -7.8380e-06, 4.8243e-06, 4.5495e-07, + 7.2177e-07, -6.4168e-07, -8.4378e-07, 1.4091e-06, 4.6636e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 249.19, cls_loss 0.0013 cls_loss_mapping 0.0039 cls_loss_causal 0.5077 re_mapping 0.0052 re_causal 0.0162 /// teacc 99.05 lr 0.00010000 +Epoch 144, weight, value: tensor([[ 0.0530, -0.0995, -0.1165, ..., -0.0870, -0.0893, -0.0785], + [ 0.0199, 0.0476, -0.0279, ..., -0.0244, 0.1325, 0.0181], + [-0.0133, -0.1193, -0.0158, ..., 0.1268, -0.1066, 0.0244], + ..., + [-0.0578, 0.1047, 0.0317, ..., -0.0845, 0.0863, 0.0404], + [-0.0025, -0.0529, 0.0078, ..., -0.0424, -0.1447, -0.0901], + [-0.0066, -0.0068, -0.0022, ..., -0.1393, -0.1089, 0.0301]], + device='cuda:0'), grad: tensor([[ 4.9447e-08, 6.1607e-07, 0.0000e+00, ..., 5.3048e-06, + 6.2305e-07, 5.8766e-07], + [-1.8962e-06, -5.4576e-07, 0.0000e+00, ..., 1.2890e-06, + -3.6974e-06, -5.2992e-07], + [ 7.4180e-07, 1.0831e-06, 0.0000e+00, ..., -5.8580e-07, + 1.6233e-06, -1.3737e-07], + ..., + [ 1.0068e-06, -9.4771e-06, 0.0000e+00, ..., 2.5029e-07, + -6.7577e-06, -7.2829e-06], + [-2.7120e-06, 3.7835e-07, 0.0000e+00, ..., 8.5868e-07, + 6.3702e-07, 5.5833e-07], + [ 1.4594e-06, 2.3302e-06, 0.0000e+00, ..., 3.9372e-07, + 2.6394e-06, 1.7127e-06]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0290, 0.0099, 0.0117, 0.0279, 0.0281, -0.0175, 0.0253, 0.0301, + -0.0095, 0.0052], device='cuda:0'), grad: tensor([ 9.9912e-06, -4.7265e-07, 4.8019e-06, 2.5779e-06, 1.4603e-05, + 1.0021e-05, -2.6703e-05, -1.4432e-05, -1.1794e-05, 1.1370e-05], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 143---------------------------------------------------- +epoch 143, time 249.40, cls_loss 0.0016 cls_loss_mapping 0.0038 cls_loss_causal 0.5052 re_mapping 0.0051 re_causal 0.0155 /// teacc 99.16 lr 0.00010000 +Epoch 145, weight, value: tensor([[ 0.0530, -0.0996, -0.1166, ..., -0.0881, -0.0897, -0.0790], + [ 0.0207, 0.0456, -0.0281, ..., -0.0245, 0.1313, 0.0166], + [-0.0134, -0.1195, -0.0156, ..., 0.1271, -0.1074, 0.0243], + ..., + [-0.0584, 0.1069, 0.0317, ..., -0.0847, 0.0882, 0.0421], + [-0.0024, -0.0533, 0.0078, ..., -0.0426, -0.1456, -0.0906], + [-0.0067, -0.0074, -0.0022, ..., -0.1401, -0.1101, 0.0298]], + device='cuda:0'), grad: tensor([[ 3.0687e-07, 1.3399e-07, 0.0000e+00, ..., 2.8382e-07, + 4.7265e-07, 3.3411e-07], + [ 9.7265e-08, 2.6048e-08, 0.0000e+00, ..., -5.6140e-06, + -1.9148e-05, -1.3635e-05], + [ 1.2573e-06, -1.4424e-05, 0.0000e+00, ..., -2.0444e-05, + -1.7136e-05, -1.8641e-05], + ..., + [ 1.4091e-06, 1.2815e-05, 0.0000e+00, ..., 2.5511e-05, + 3.7074e-05, 2.8923e-05], + [ 5.6177e-06, 2.9686e-07, 0.0000e+00, ..., 9.3924e-07, + 1.5786e-06, 5.8534e-07], + [ 6.2771e-07, 2.7963e-07, 0.0000e+00, ..., 2.3283e-07, + 9.1642e-07, 2.0640e-07]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0294, 0.0084, 0.0116, 0.0276, 0.0282, -0.0176, 0.0255, 0.0321, + -0.0096, 0.0049], device='cuda:0'), grad: tensor([ 3.5632e-06, -1.9982e-05, -6.1333e-05, -4.3035e-05, 3.3118e-06, + -3.6150e-05, 8.5533e-06, 1.0890e-04, 3.0726e-05, 5.3681e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 248.89, cls_loss 0.0015 cls_loss_mapping 0.0034 cls_loss_causal 0.5023 re_mapping 0.0049 re_causal 0.0151 /// teacc 99.02 lr 0.00010000 +Epoch 146, weight, value: tensor([[ 0.0533, -0.0997, -0.1169, ..., -0.0886, -0.0902, -0.0793], + [ 0.0212, 0.0456, -0.0279, ..., -0.0245, 0.1316, 0.0167], + [-0.0133, -0.1195, -0.0156, ..., 0.1272, -0.1076, 0.0246], + ..., + [-0.0588, 0.1072, 0.0314, ..., -0.0854, 0.0883, 0.0423], + [-0.0025, -0.0537, 0.0077, ..., -0.0429, -0.1464, -0.0914], + [-0.0069, -0.0082, -0.0023, ..., -0.1413, -0.1112, 0.0292]], + device='cuda:0'), grad: tensor([[ 1.1933e-07, 4.6042e-08, 0.0000e+00, ..., 2.8402e-05, + 5.0670e-08, 1.7113e-07], + [-7.6892e-08, -3.7230e-07, 0.0000e+00, ..., 2.7660e-06, + -7.6881e-07, -8.2073e-09], + [ 2.1013e-07, 7.5437e-08, 0.0000e+00, ..., 1.0729e-06, + 1.3190e-07, -1.0394e-06], + ..., + [ 2.8242e-07, 1.8487e-07, 0.0000e+00, ..., 5.5134e-07, + 2.5076e-07, 1.0505e-06], + [ 5.0105e-07, 6.2981e-08, 0.0000e+00, ..., 4.0568e-06, + 9.7963e-08, 3.0664e-07], + [-6.4867e-07, -3.7160e-07, 0.0000e+00, ..., 2.3931e-05, + -2.6193e-08, -2.9784e-06]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0293, 0.0085, 0.0114, 0.0280, 0.0282, -0.0175, 0.0256, 0.0321, + -0.0098, 0.0044], device='cuda:0'), grad: tensor([ 8.9109e-05, 8.4192e-06, 6.0685e-06, -8.4639e-06, 6.1452e-05, + 1.5087e-05, -2.5535e-04, 4.7237e-06, 1.6525e-05, 6.2168e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 248.74, cls_loss 0.0016 cls_loss_mapping 0.0045 cls_loss_causal 0.4954 re_mapping 0.0047 re_causal 0.0143 /// teacc 99.08 lr 0.00010000 +Epoch 147, weight, value: tensor([[ 0.0537, -0.0998, -0.1181, ..., -0.0894, -0.0907, -0.0798], + [ 0.0213, 0.0456, -0.0279, ..., -0.0250, 0.1319, 0.0167], + [-0.0132, -0.1198, -0.0156, ..., 0.1276, -0.1080, 0.0247], + ..., + [-0.0592, 0.1075, 0.0314, ..., -0.0857, 0.0884, 0.0423], + [-0.0037, -0.0542, 0.0073, ..., -0.0432, -0.1468, -0.0920], + [-0.0054, -0.0083, -0.0023, ..., -0.1427, -0.1118, 0.0294]], + device='cuda:0'), grad: tensor([[-1.7323e-07, 6.6531e-08, 0.0000e+00, ..., 4.6287e-07, + 1.7090e-07, 1.3597e-07], + [-2.3805e-06, -1.4612e-06, 0.0000e+00, ..., 2.7078e-07, + -5.0515e-06, -1.8040e-06], + [ 5.9232e-07, 1.3271e-07, 0.0000e+00, ..., 2.4855e-08, + 2.7707e-07, 3.6962e-08], + ..., + [ 1.1856e-06, 5.1269e-07, 0.0000e+00, ..., 3.8370e-07, + 1.9632e-06, 1.1344e-06], + [-3.9078e-06, 1.7695e-07, 0.0000e+00, ..., -3.9907e-07, + 5.6671e-07, 3.1665e-07], + [ 7.8510e-07, 1.7579e-07, 0.0000e+00, ..., 2.2678e-07, + 6.9756e-07, 2.8429e-07]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0294, 0.0084, 0.0116, 0.0281, 0.0284, -0.0176, 0.0262, 0.0320, + -0.0108, 0.0047], device='cuda:0'), grad: tensor([ 1.2135e-06, -5.6811e-06, 2.4550e-06, 1.3933e-06, -6.7949e-06, + 1.1355e-05, 2.8443e-06, 5.8413e-06, -1.6153e-05, 3.4943e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 248.74, cls_loss 0.0018 cls_loss_mapping 0.0052 cls_loss_causal 0.5374 re_mapping 0.0048 re_causal 0.0151 /// teacc 98.91 lr 0.00010000 +Epoch 148, weight, value: tensor([[ 0.0537, -0.1000, -0.1184, ..., -0.0906, -0.0909, -0.0804], + [ 0.0217, 0.0457, -0.0280, ..., -0.0256, 0.1322, 0.0168], + [-0.0131, -0.1198, -0.0157, ..., 0.1283, -0.1078, 0.0254], + ..., + [-0.0599, 0.1075, 0.0314, ..., -0.0870, 0.0882, 0.0421], + [-0.0028, -0.0538, 0.0076, ..., -0.0433, -0.1471, -0.0914], + [-0.0061, -0.0084, -0.0025, ..., -0.1439, -0.1125, 0.0293]], + device='cuda:0'), grad: tensor([[ 7.1898e-07, 7.8231e-07, 2.6193e-10, ..., 4.1835e-06, + 9.1083e-07, 9.5367e-07], + [-1.7390e-05, -1.9714e-05, -6.1409e-09, ..., 4.9453e-07, + -2.5064e-05, -7.2494e-06], + [ 2.3330e-07, 1.1381e-06, 4.0745e-10, ..., -3.5390e-07, + 1.4585e-06, 7.4331e-08], + ..., + [ 1.5780e-05, -7.2876e-07, 3.3178e-09, ..., 2.3621e-07, + 6.9663e-07, 2.4214e-06], + [ 1.7574e-06, 1.7593e-06, 4.3656e-10, ..., 5.3132e-07, + 1.9614e-06, 2.0284e-06], + [-4.9435e-06, -9.2573e-07, 4.9477e-10, ..., 1.8207e-07, + 4.3027e-07, -9.5591e-06]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0301, 0.0084, 0.0119, 0.0286, 0.0282, -0.0178, 0.0268, 0.0313, + -0.0098, 0.0041], device='cuda:0'), grad: tensor([ 1.5497e-05, -3.9250e-05, 3.2932e-06, 5.9962e-05, 7.1041e-06, + -1.1973e-05, -1.5199e-05, 8.8066e-06, 1.2010e-05, -4.0323e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 248.66, cls_loss 0.0018 cls_loss_mapping 0.0057 cls_loss_causal 0.5120 re_mapping 0.0053 re_causal 0.0158 /// teacc 99.12 lr 0.00010000 +Epoch 149, weight, value: tensor([[ 0.0536, -0.1002, -0.1190, ..., -0.0919, -0.0912, -0.0830], + [ 0.0222, 0.0459, -0.0282, ..., -0.0261, 0.1327, 0.0170], + [-0.0129, -0.1201, -0.0142, ..., 0.1288, -0.1081, 0.0257], + ..., + [-0.0606, 0.1094, 0.0313, ..., -0.0873, 0.0892, 0.0438], + [-0.0026, -0.0540, 0.0075, ..., -0.0434, -0.1474, -0.0919], + [-0.0056, -0.0112, -0.0025, ..., -0.1441, -0.1155, 0.0271]], + device='cuda:0'), grad: tensor([[ 5.1316e-07, 4.9971e-08, 9.6043e-10, ..., 2.4252e-06, + 7.7439e-07, 3.3039e-07], + [-8.9854e-06, -6.7651e-06, -3.1339e-07, ..., -1.9313e-07, + -2.2560e-05, -8.5011e-06], + [ 2.0545e-06, 6.8394e-08, 1.3679e-09, ..., -2.4633e-07, + 3.2727e-06, 1.2829e-07], + ..., + [ 3.7868e-06, 5.3570e-06, 2.4540e-07, ..., 1.9080e-07, + 1.0930e-05, 5.1856e-06], + [-4.4443e-06, 6.0012e-08, 9.3132e-10, ..., 9.5321e-07, + 2.4494e-06, 1.0449e-06], + [ 8.6753e-07, -4.1502e-08, 1.2398e-08, ..., 4.1388e-06, + 5.3085e-07, -6.0210e-07]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0306, 0.0086, 0.0122, 0.0279, 0.0298, -0.0175, 0.0266, 0.0328, + -0.0098, 0.0019], device='cuda:0'), grad: tensor([ 8.8662e-06, -3.1769e-05, 7.2941e-06, 6.2250e-06, 9.4101e-06, + 8.3596e-06, -2.2262e-05, 1.6436e-05, -1.7211e-05, 1.4536e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 248.98, cls_loss 0.0016 cls_loss_mapping 0.0042 cls_loss_causal 0.5237 re_mapping 0.0050 re_causal 0.0156 /// teacc 98.89 lr 0.00010000 +Epoch 150, weight, value: tensor([[ 0.0539, -0.1003, -0.1191, ..., -0.0924, -0.0915, -0.0832], + [ 0.0226, 0.0460, -0.0282, ..., -0.0263, 0.1332, 0.0171], + [-0.0098, -0.1208, -0.0126, ..., 0.1292, -0.1088, 0.0269], + ..., + [-0.0614, 0.1095, 0.0313, ..., -0.0877, 0.0890, 0.0436], + [-0.0028, -0.0541, 0.0063, ..., -0.0438, -0.1478, -0.0923], + [-0.0055, -0.0112, -0.0025, ..., -0.1446, -0.1155, 0.0273]], + device='cuda:0'), grad: tensor([[-9.5984e-08, 2.4331e-08, 0.0000e+00, ..., 1.7171e-08, + 4.7905e-08, 5.0961e-08], + [-3.6671e-07, -2.5611e-07, 0.0000e+00, ..., 1.0827e-08, + -7.9675e-07, -2.0675e-07], + [ 2.4796e-07, 8.5495e-07, 0.0000e+00, ..., -3.3760e-08, + 1.2349e-06, 1.5888e-06], + ..., + [ 2.5844e-07, -9.7323e-07, 0.0000e+00, ..., 2.8842e-08, + -1.0906e-06, -2.0005e-06], + [-9.6578e-07, 5.1688e-08, 0.0000e+00, ..., -8.3703e-08, + 1.2119e-07, 8.9698e-08], + [ 9.4483e-07, 1.3376e-07, 0.0000e+00, ..., 6.2573e-08, + 1.9546e-07, 1.5181e-07]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0305, 0.0087, 0.0152, 0.0252, 0.0301, -0.0177, 0.0269, 0.0325, + -0.0101, 0.0020], device='cuda:0'), grad: tensor([-1.8347e-07, -7.7114e-07, 4.0419e-06, -8.5589e-07, 5.3132e-07, + -2.2668e-06, 1.4342e-06, -2.7586e-06, -2.7604e-06, 3.5744e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 250.60, cls_loss 0.0022 cls_loss_mapping 0.0069 cls_loss_causal 0.5349 re_mapping 0.0051 re_causal 0.0160 /// teacc 98.98 lr 0.00010000 +Epoch 151, weight, value: tensor([[ 0.0541, -0.1005, -0.1198, ..., -0.0936, -0.0923, -0.0834], + [ 0.0260, 0.0476, -0.0282, ..., -0.0283, 0.1364, 0.0175], + [-0.0096, -0.1214, -0.0126, ..., 0.1300, -0.1080, 0.0276], + ..., + [-0.0637, 0.1090, 0.0312, ..., -0.0879, 0.0883, 0.0435], + [-0.0021, -0.0570, 0.0062, ..., -0.0443, -0.1500, -0.0927], + [-0.0066, -0.0114, -0.0026, ..., -0.1452, -0.1159, 0.0272]], + device='cuda:0'), grad: tensor([[ 4.2212e-07, 5.5967e-08, 0.0000e+00, ..., 3.9185e-07, + 7.9919e-08, 6.7288e-08], + [-4.0745e-08, -6.7195e-07, 0.0000e+00, ..., 1.2228e-06, + -1.5832e-06, -1.2096e-07], + [ 2.4959e-06, 1.2154e-07, 0.0000e+00, ..., 1.1566e-07, + 2.1257e-07, -1.0914e-08], + ..., + [ 8.4862e-06, 5.9651e-07, 0.0000e+00, ..., 1.1135e-07, + 5.3272e-07, 1.3644e-06], + [ 2.8545e-07, 1.8836e-07, 0.0000e+00, ..., 5.3225e-07, + 2.9081e-07, 1.7253e-07], + [ 1.1558e-06, 2.0349e-07, 0.0000e+00, ..., 1.0856e-07, + 1.1915e-07, 2.7753e-07]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0305, 0.0118, 0.0153, 0.0239, 0.0298, -0.0184, 0.0274, 0.0318, + -0.0094, 0.0015], device='cuda:0'), grad: tensor([ 2.6356e-06, 2.6803e-06, 1.0043e-05, -9.7036e-05, -3.8892e-06, + 4.3243e-05, -4.9435e-06, 3.3468e-05, 5.7444e-06, 7.9572e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 256.70, cls_loss 0.0017 cls_loss_mapping 0.0045 cls_loss_causal 0.5356 re_mapping 0.0048 re_causal 0.0151 /// teacc 99.00 lr 0.00010000 +Epoch 152, weight, value: tensor([[ 0.0541, -0.1005, -0.1230, ..., -0.0965, -0.0927, -0.0836], + [ 0.0265, 0.0476, -0.0283, ..., -0.0281, 0.1369, 0.0176], + [-0.0097, -0.1217, -0.0129, ..., 0.1300, -0.1091, 0.0272], + ..., + [-0.0641, 0.1092, 0.0312, ..., -0.0880, 0.0885, 0.0436], + [-0.0022, -0.0575, 0.0061, ..., -0.0446, -0.1505, -0.0931], + [-0.0074, -0.0114, -0.0026, ..., -0.1457, -0.1160, 0.0273]], + device='cuda:0'), grad: tensor([[-9.7963e-08, 1.1805e-07, 0.0000e+00, ..., 5.6869e-08, + 4.0559e-07, 4.7288e-07], + [-1.3165e-05, -8.2850e-06, 0.0000e+00, ..., -5.3868e-06, + -3.3110e-05, -1.2681e-05], + [ 5.6345e-07, 4.2142e-07, 0.0000e+00, ..., 1.8231e-07, + 1.4082e-06, 1.2117e-06], + ..., + [ 4.2911e-07, 1.1758e-07, 0.0000e+00, ..., 1.0803e-06, + 6.3749e-07, 6.3777e-06], + [ 6.0815e-07, 1.2817e-07, 0.0000e+00, ..., 3.3039e-07, + 3.8254e-07, 5.8394e-07], + [ 7.4564e-08, 5.3225e-07, 0.0000e+00, ..., 2.5034e-06, + 2.3760e-07, 1.4059e-05]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0317, 0.0122, 0.0152, 0.0244, 0.0294, -0.0184, 0.0275, 0.0317, + -0.0095, 0.0010], device='cuda:0'), grad: tensor([ 5.8813e-07, -5.3674e-05, 4.0531e-06, 1.1530e-06, -5.1022e-05, + -5.6718e-07, 5.3197e-05, 1.2808e-05, 3.6322e-06, 2.9802e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 256.51, cls_loss 0.0016 cls_loss_mapping 0.0038 cls_loss_causal 0.5111 re_mapping 0.0049 re_causal 0.0147 /// teacc 99.03 lr 0.00010000 +Epoch 153, weight, value: tensor([[ 0.0547, -0.1004, -0.1230, ..., -0.0966, -0.0934, -0.0839], + [ 0.0265, 0.0475, -0.0283, ..., -0.0286, 0.1370, 0.0174], + [-0.0095, -0.1224, -0.0127, ..., 0.1309, -0.1096, 0.0277], + ..., + [-0.0647, 0.1096, 0.0312, ..., -0.0891, 0.0887, 0.0438], + [-0.0022, -0.0577, 0.0060, ..., -0.0447, -0.1505, -0.0936], + [-0.0072, -0.0114, -0.0026, ..., -0.1461, -0.1161, 0.0273]], + device='cuda:0'), grad: tensor([[ 7.6077e-08, 2.2905e-08, 0.0000e+00, ..., 4.6974e-08, + 3.9843e-08, 1.3888e-07], + [-2.4564e-07, 8.0094e-08, 0.0000e+00, ..., 4.1968e-08, + -4.5216e-07, 1.2422e-07], + [ 1.6810e-07, 1.7683e-07, 0.0000e+00, ..., -3.0710e-07, + 2.6985e-07, -4.0367e-08], + ..., + [ 3.6252e-07, -1.2172e-06, 0.0000e+00, ..., 2.3597e-07, + -7.1805e-07, -1.9488e-07], + [-3.7369e-07, 6.2515e-08, 0.0000e+00, ..., -1.3341e-07, + 1.4110e-07, 2.4564e-07], + [-8.2003e-07, 3.7765e-07, 0.0000e+00, ..., 7.4040e-08, + 3.1828e-07, -3.6098e-06]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0315, 0.0120, 0.0152, 0.0243, 0.0295, -0.0181, 0.0274, 0.0317, + -0.0093, 0.0009], device='cuda:0'), grad: tensor([ 5.6904e-07, 1.9022e-07, 4.5914e-07, -2.3225e-08, 4.3772e-06, + 3.7160e-06, 1.4687e-06, 7.8464e-07, -9.2760e-07, -1.0625e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 251.33, cls_loss 0.0016 cls_loss_mapping 0.0040 cls_loss_causal 0.5334 re_mapping 0.0046 re_causal 0.0148 /// teacc 99.00 lr 0.00010000 +Epoch 154, weight, value: tensor([[ 0.0548, -0.1005, -0.1230, ..., -0.0968, -0.0939, -0.0842], + [ 0.0270, 0.0484, -0.0285, ..., -0.0289, 0.1380, 0.0187], + [-0.0095, -0.1231, -0.0124, ..., 0.1315, -0.1100, 0.0277], + ..., + [-0.0653, 0.1098, 0.0312, ..., -0.0894, 0.0889, 0.0438], + [-0.0022, -0.0579, 0.0060, ..., -0.0450, -0.1506, -0.0943], + [-0.0087, -0.0124, -0.0026, ..., -0.1469, -0.1177, 0.0268]], + device='cuda:0'), grad: tensor([[-2.0652e-07, 3.6962e-08, 0.0000e+00, ..., -8.4809e-08, + 8.0967e-08, 1.4016e-07], + [-1.2387e-07, -5.2201e-07, 0.0000e+00, ..., 2.4005e-07, + -1.3253e-06, 9.5286e-08], + [-3.1479e-07, -5.0198e-07, 0.0000e+00, ..., -1.6838e-05, + -1.1690e-05, -1.6078e-05], + ..., + [ 2.5332e-07, 3.9954e-07, 0.0000e+00, ..., 1.4260e-05, + 1.0915e-05, 1.5274e-05], + [ 4.0489e-07, 2.3865e-07, 0.0000e+00, ..., 1.1837e-06, + 7.1991e-07, 1.4110e-06], + [-3.9837e-07, -2.2294e-07, 0.0000e+00, ..., 7.8324e-07, + 3.8592e-08, 1.5339e-06]], device='cuda:0') +Epoch 154, bias, value: tensor([-3.1364e-02, 1.2959e-02, 1.5216e-02, 2.4715e-02, 2.8798e-02, + -1.8623e-02, 2.7313e-02, 3.1599e-02, -9.5108e-03, 7.8537e-05], + device='cuda:0'), grad: tensor([-4.1537e-07, 2.6543e-08, -5.4032e-05, 2.1115e-05, -1.1705e-05, + -2.4945e-05, 1.2480e-05, 4.7505e-05, 6.0201e-06, 3.8520e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 248.99, cls_loss 0.0015 cls_loss_mapping 0.0035 cls_loss_causal 0.5235 re_mapping 0.0046 re_causal 0.0146 /// teacc 99.05 lr 0.00010000 +Epoch 155, weight, value: tensor([[ 0.0549, -0.1007, -0.1229, ..., -0.0971, -0.0943, -0.0845], + [ 0.0270, 0.0481, -0.0286, ..., -0.0291, 0.1379, 0.0183], + [-0.0097, -0.1235, -0.0123, ..., 0.1317, -0.1108, 0.0275], + ..., + [-0.0658, 0.1098, 0.0311, ..., -0.0899, 0.0890, 0.0437], + [-0.0022, -0.0580, 0.0060, ..., -0.0453, -0.1506, -0.0949], + [-0.0076, -0.0116, -0.0026, ..., -0.1471, -0.1173, 0.0276]], + device='cuda:0'), grad: tensor([[ 3.3225e-07, 2.3370e-08, 0.0000e+00, ..., 2.1083e-07, + 1.7486e-07, 1.9011e-07], + [-1.7613e-05, 2.1933e-07, 0.0000e+00, ..., -4.8280e-06, + -1.2055e-05, -7.8902e-06], + [ 3.2820e-06, 4.3353e-07, 0.0000e+00, ..., -3.8967e-06, + 4.5039e-06, 1.5134e-06], + ..., + [ 4.4703e-07, -1.0226e-06, 0.0000e+00, ..., 1.5283e-06, + -5.4762e-07, -6.4867e-07], + [ 9.2238e-06, 6.9849e-08, 0.0000e+00, ..., 6.0312e-06, + 4.2245e-06, 4.2729e-06], + [ 2.1979e-06, 1.1537e-07, 0.0000e+00, ..., 7.8324e-07, + 1.5404e-06, 1.3812e-06]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0313, 0.0127, 0.0151, 0.0248, 0.0283, -0.0186, 0.0273, 0.0312, + -0.0095, 0.0012], device='cuda:0'), grad: tensor([ 1.6866e-06, -3.8326e-05, 3.6415e-06, 6.3442e-06, -1.4331e-07, + -1.5318e-05, 7.9572e-06, 7.7300e-07, 2.6360e-05, 7.0371e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 154, time 249.14, cls_loss 0.0012 cls_loss_mapping 0.0034 cls_loss_causal 0.5027 re_mapping 0.0051 re_causal 0.0151 /// teacc 99.04 lr 0.00010000 +Epoch 156, weight, value: tensor([[ 0.0553, -0.1009, -0.1227, ..., -0.0972, -0.0949, -0.0848], + [ 0.0270, 0.0479, -0.0286, ..., -0.0291, 0.1379, 0.0180], + [-0.0097, -0.1239, -0.0125, ..., 0.1320, -0.1113, 0.0276], + ..., + [-0.0660, 0.1101, 0.0311, ..., -0.0903, 0.0894, 0.0438], + [-0.0025, -0.0587, 0.0058, ..., -0.0458, -0.1507, -0.0958], + [-0.0075, -0.0114, -0.0027, ..., -0.1477, -0.1174, 0.0278]], + device='cuda:0'), grad: tensor([[ 1.7183e-06, 1.4697e-08, 0.0000e+00, ..., 2.3178e-07, + 7.2177e-08, 1.0217e-06], + [-6.5975e-06, -1.3551e-07, 0.0000e+00, ..., 3.2433e-07, + -2.1920e-05, -4.8093e-06], + [-7.6368e-07, 2.8027e-08, 0.0000e+00, ..., -6.2399e-07, + 2.0140e-07, -6.0443e-07], + ..., + [ 6.5453e-06, 3.9185e-07, 0.0000e+00, ..., 1.4761e-07, + 1.8716e-05, 5.9046e-06], + [ 3.4622e-07, 3.1461e-08, 0.0000e+00, ..., 1.4273e-07, + 2.1036e-07, 3.8743e-07], + [ 2.1458e-05, 1.3621e-07, 0.0000e+00, ..., 9.8571e-06, + 1.6647e-07, 4.6730e-05]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0311, 0.0126, 0.0151, 0.0250, 0.0282, -0.0183, 0.0275, 0.0313, + -0.0100, 0.0013], device='cuda:0'), grad: tensor([ 6.0350e-06, -3.0100e-05, -1.3513e-06, 8.1807e-06, -2.1911e-04, + -9.5740e-06, 1.3663e-06, 3.2037e-05, 1.8347e-06, 2.1100e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 249.43, cls_loss 0.0016 cls_loss_mapping 0.0044 cls_loss_causal 0.5173 re_mapping 0.0046 re_causal 0.0146 /// teacc 98.95 lr 0.00010000 +Epoch 157, weight, value: tensor([[ 0.0558, -0.1012, -0.1227, ..., -0.0978, -0.0958, -0.0852], + [ 0.0269, 0.0477, -0.0285, ..., -0.0299, 0.1379, 0.0178], + [-0.0095, -0.1242, -0.0123, ..., 0.1327, -0.1113, 0.0285], + ..., + [-0.0666, 0.1107, 0.0309, ..., -0.0910, 0.0900, 0.0442], + [-0.0026, -0.0590, 0.0058, ..., -0.0463, -0.1508, -0.0967], + [-0.0075, -0.0119, -0.0027, ..., -0.1485, -0.1178, 0.0276]], + device='cuda:0'), grad: tensor([[ 1.7637e-07, 3.0705e-08, 0.0000e+00, ..., 2.5164e-06, + 1.3215e-06, 4.0457e-06], + [ 1.1744e-06, 4.6589e-07, 0.0000e+00, ..., 4.2245e-06, + 2.4959e-06, 7.5549e-06], + [ 9.6916e-08, 1.1828e-07, 0.0000e+00, ..., -8.2254e-05, + -4.4137e-05, -1.3125e-04], + ..., + [ 3.7509e-07, -4.4797e-07, 0.0000e+00, ..., 6.6876e-05, + 3.5942e-05, 1.0717e-04], + [-1.1269e-06, 4.4791e-08, 0.0000e+00, ..., 3.1777e-06, + 1.7760e-06, 5.0664e-06], + [-1.5544e-06, -3.8766e-07, 0.0000e+00, ..., 4.7265e-07, + 9.6625e-09, -3.6694e-07]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0312, 0.0124, 0.0152, 0.0252, 0.0281, -0.0181, 0.0282, 0.0314, + -0.0102, 0.0008], device='cuda:0'), grad: tensor([ 1.0908e-05, 2.1085e-05, -3.4332e-04, 6.7391e-06, 4.7497e-06, + 2.6217e-07, 1.0870e-05, 2.8157e-04, 9.8944e-06, -2.2370e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 249.17, cls_loss 0.0013 cls_loss_mapping 0.0039 cls_loss_causal 0.4925 re_mapping 0.0048 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +Epoch 158, weight, value: tensor([[ 0.0562, -0.1013, -0.1227, ..., -0.0983, -0.0961, -0.0854], + [ 0.0268, 0.0475, -0.0289, ..., -0.0305, 0.1379, 0.0176], + [-0.0095, -0.1246, -0.0119, ..., 0.1332, -0.1115, 0.0288], + ..., + [-0.0669, 0.1110, 0.0309, ..., -0.0916, 0.0903, 0.0443], + [-0.0026, -0.0589, 0.0055, ..., -0.0465, -0.1508, -0.0975], + [-0.0071, -0.0118, -0.0028, ..., -0.1492, -0.1178, 0.0277]], + device='cuda:0'), grad: tensor([[ 5.9605e-07, 1.0962e-06, 0.0000e+00, ..., 3.6322e-07, + 4.7917e-07, 2.5146e-06], + [ 7.3400e-08, 4.1933e-07, 0.0000e+00, ..., 6.1467e-07, + -6.2445e-07, 1.5050e-06], + [ 1.6252e-07, -3.5256e-05, 0.0000e+00, ..., -5.4836e-05, + -1.0294e-04, -6.4433e-05], + ..., + [ 4.4890e-06, 4.2588e-05, 0.0000e+00, ..., 5.2750e-05, + 1.0294e-04, 7.9751e-05], + [-7.2271e-07, 3.5949e-07, 0.0000e+00, ..., -1.9209e-09, + 3.1642e-07, 7.8045e-07], + [-1.2480e-05, -2.3112e-05, 0.0000e+00, ..., 1.0099e-07, + -1.4203e-06, -5.1409e-05]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0313, 0.0121, 0.0152, 0.0253, 0.0279, -0.0181, 0.0284, 0.0315, + -0.0101, 0.0009], device='cuda:0'), grad: tensor([ 9.0152e-06, 7.3947e-06, -2.9063e-04, 4.5300e-06, 1.0866e-04, + -2.3142e-05, 2.6569e-05, 3.5000e-04, -1.4706e-06, -1.9085e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 249.32, cls_loss 0.0013 cls_loss_mapping 0.0041 cls_loss_causal 0.5203 re_mapping 0.0045 re_causal 0.0144 /// teacc 99.05 lr 0.00010000 +Epoch 159, weight, value: tensor([[ 0.0562, -0.1015, -0.1227, ..., -0.0985, -0.0964, -0.0857], + [ 0.0268, 0.0475, -0.0288, ..., -0.0310, 0.1379, 0.0174], + [-0.0094, -0.1249, -0.0120, ..., 0.1341, -0.1114, 0.0294], + ..., + [-0.0676, 0.1111, 0.0309, ..., -0.0926, 0.0903, 0.0441], + [-0.0025, -0.0591, 0.0055, ..., -0.0468, -0.1508, -0.0981], + [-0.0067, -0.0117, -0.0028, ..., -0.1499, -0.1178, 0.0278]], + device='cuda:0'), grad: tensor([[-3.7067e-06, 2.1618e-07, 0.0000e+00, ..., -3.2336e-06, + 8.9640e-08, 2.2934e-07], + [ 4.2812e-08, 1.0267e-05, 0.0000e+00, ..., 3.1316e-07, + 3.2298e-06, 2.6971e-06], + [-9.4995e-07, 3.2634e-06, 0.0000e+00, ..., -4.0680e-06, + 1.3821e-06, -5.6718e-07], + ..., + [ 1.0594e-07, -3.8832e-05, 0.0000e+00, ..., 4.5355e-07, + -1.2867e-05, -9.1121e-06], + [ 9.6299e-07, 1.8552e-05, 0.0000e+00, ..., 3.4440e-06, + 5.7332e-06, 5.5805e-06], + [ 6.7055e-08, 1.6456e-06, 0.0000e+00, ..., 1.4035e-06, + 5.9139e-07, 3.0864e-06]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0313, 0.0119, 0.0155, 0.0252, 0.0282, -0.0177, 0.0281, 0.0311, + -0.0101, 0.0010], device='cuda:0'), grad: tensor([-3.4034e-05, 1.8656e-05, -5.5656e-06, 6.2138e-06, -6.2957e-06, + 3.5465e-06, 3.0220e-05, -6.4433e-05, 4.3064e-05, 8.5607e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 249.29, cls_loss 0.0016 cls_loss_mapping 0.0046 cls_loss_causal 0.5209 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.04 lr 0.00010000 +Epoch 160, weight, value: tensor([[ 0.0535, -0.1018, -0.1227, ..., -0.0993, -0.0970, -0.0891], + [ 0.0268, 0.0472, -0.0288, ..., -0.0316, 0.1380, 0.0172], + [-0.0094, -0.1265, -0.0120, ..., 0.1342, -0.1126, 0.0288], + ..., + [-0.0707, 0.1106, 0.0308, ..., -0.0922, 0.0900, 0.0439], + [-0.0019, -0.0570, 0.0055, ..., -0.0471, -0.1507, -0.0987], + [-0.0033, -0.0108, -0.0028, ..., -0.1502, -0.1172, 0.0290]], + device='cuda:0'), grad: tensor([[ 5.3318e-07, 1.7229e-08, 0.0000e+00, ..., 1.4128e-06, + 1.7259e-08, 1.3458e-06], + [ 1.6149e-06, 3.2625e-08, 0.0000e+00, ..., 3.1404e-06, + -1.2107e-07, 2.7735e-06], + [-2.1849e-06, 1.1106e-07, 0.0000e+00, ..., -6.1803e-06, + 1.0227e-07, -5.8450e-06], + ..., + [ 1.4482e-06, -1.5339e-06, 0.0000e+00, ..., 3.4273e-06, + -1.1791e-06, 1.7257e-06], + [ 9.7789e-07, 4.1589e-08, 0.0000e+00, ..., 1.6596e-06, + 6.7637e-08, 1.5795e-06], + [ 6.7241e-06, 1.1409e-06, 0.0000e+00, ..., 1.1832e-05, + 9.3598e-07, 1.1377e-05]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0337, 0.0117, 0.0153, 0.0252, 0.0285, -0.0181, 0.0281, 0.0300, + -0.0095, 0.0030], device='cuda:0'), grad: tensor([ 3.5837e-06, 8.9779e-06, -1.5751e-05, -5.7044e-07, -4.7177e-05, + 1.5432e-06, 1.2331e-06, 6.4783e-06, 5.6215e-06, 3.6061e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 249.60, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.5231 re_mapping 0.0047 re_causal 0.0147 /// teacc 99.00 lr 0.00010000 +Epoch 161, weight, value: tensor([[ 0.0537, -0.1018, -0.1227, ..., -0.0996, -0.0972, -0.0901], + [ 0.0268, 0.0471, -0.0288, ..., -0.0319, 0.1380, 0.0171], + [-0.0092, -0.1271, -0.0116, ..., 0.1352, -0.1129, 0.0291], + ..., + [-0.0713, 0.1108, 0.0308, ..., -0.0931, 0.0903, 0.0439], + [-0.0019, -0.0568, 0.0054, ..., -0.0479, -0.1507, -0.0995], + [-0.0025, -0.0107, -0.0028, ..., -0.1511, -0.1173, 0.0293]], + device='cuda:0'), grad: tensor([[ 1.3167e-07, 5.7858e-08, 0.0000e+00, ..., 2.4273e-08, + 8.9989e-08, 2.8010e-07], + [-1.6317e-06, -9.1502e-07, 0.0000e+00, ..., 3.8155e-08, + -3.9227e-06, -9.2108e-07], + [ 1.2191e-06, 8.9547e-07, 0.0000e+00, ..., -2.7765e-08, + 2.5108e-06, 1.5097e-06], + ..., + [ 1.5143e-06, 8.0839e-07, 0.0000e+00, ..., 2.9802e-07, + -1.1764e-07, 2.7511e-06], + [ 7.7346e-07, 2.4377e-07, 0.0000e+00, ..., 1.3632e-07, + 5.1828e-07, 1.2936e-06], + [-3.6061e-06, -1.8366e-06, 0.0000e+00, ..., -8.6427e-07, + 7.6368e-08, -9.0823e-06]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0329, 0.0116, 0.0155, 0.0249, 0.0285, -0.0179, 0.0278, 0.0299, + -0.0098, 0.0033], device='cuda:0'), grad: tensor([ 8.3819e-07, -6.8247e-06, 6.4075e-06, 3.6117e-06, 6.9775e-06, + -3.4943e-06, 1.5106e-06, 7.7039e-06, 4.0457e-06, -2.0817e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 249.07, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.5420 re_mapping 0.0048 re_causal 0.0146 /// teacc 99.08 lr 0.00010000 +Epoch 162, weight, value: tensor([[ 0.0538, -0.1019, -0.1227, ..., -0.0996, -0.0978, -0.0901], + [ 0.0269, 0.0469, -0.0295, ..., -0.0318, 0.1381, 0.0170], + [-0.0092, -0.1279, -0.0112, ..., 0.1355, -0.1136, 0.0290], + ..., + [-0.0716, 0.1112, 0.0308, ..., -0.0934, 0.0906, 0.0441], + [-0.0019, -0.0570, 0.0053, ..., -0.0483, -0.1508, -0.1000], + [-0.0023, -0.0105, -0.0028, ..., -0.1519, -0.1174, 0.0294]], + device='cuda:0'), grad: tensor([[ 7.6042e-07, 2.9244e-07, 0.0000e+00, ..., 1.9558e-08, + 3.6741e-07, 2.6776e-07], + [-1.8990e-06, 1.7628e-05, 0.0000e+00, ..., -6.5484e-08, + 1.2442e-05, 1.4186e-05], + [ 9.9000e-07, 2.1383e-06, 0.0000e+00, ..., -4.8021e-09, + 2.5444e-06, 1.7593e-06], + ..., + [-9.1456e-07, -3.6210e-05, 0.0000e+00, ..., 7.7416e-08, + -3.3110e-05, -2.9817e-05], + [ 8.2562e-07, 4.1053e-06, 0.0000e+00, ..., 5.0844e-08, + 3.9637e-06, 3.2242e-06], + [ 8.7824e-07, 7.0184e-06, 0.0000e+00, ..., 9.8080e-09, + 6.7949e-06, 5.5991e-06]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0327, 0.0116, 0.0154, 0.0250, 0.0286, -0.0182, 0.0278, 0.0300, + -0.0099, 0.0034], device='cuda:0'), grad: tensor([ 4.7535e-06, 2.9206e-05, 8.7246e-06, 4.1127e-06, 9.1568e-06, + -2.2441e-05, 4.4145e-06, -6.6936e-05, 1.2591e-05, 1.6451e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 248.96, cls_loss 0.0012 cls_loss_mapping 0.0034 cls_loss_causal 0.5036 re_mapping 0.0045 re_causal 0.0139 /// teacc 99.04 lr 0.00010000 +Epoch 163, weight, value: tensor([[ 0.0539, -0.1021, -0.1227, ..., -0.0996, -0.0979, -0.0902], + [ 0.0270, 0.0468, -0.0290, ..., -0.0322, 0.1382, 0.0170], + [-0.0093, -0.1281, -0.0115, ..., 0.1359, -0.1144, 0.0290], + ..., + [-0.0720, 0.1114, 0.0307, ..., -0.0940, 0.0909, 0.0442], + [-0.0020, -0.0570, 0.0053, ..., -0.0485, -0.1508, -0.1006], + [-0.0020, -0.0105, -0.0028, ..., -0.1521, -0.1174, 0.0295]], + device='cuda:0'), grad: tensor([[-7.7263e-06, 1.3104e-06, 0.0000e+00, ..., 5.8353e-08, + 1.2089e-06, 9.5554e-07], + [-7.0967e-07, 3.5185e-06, 0.0000e+00, ..., 9.6217e-08, + 1.9968e-06, 2.2519e-06], + [ 4.6636e-07, 3.0082e-06, 0.0000e+00, ..., -6.8033e-07, + 2.7902e-06, 1.7062e-06], + ..., + [ 9.4669e-07, -3.3855e-05, 0.0000e+00, ..., 2.9034e-07, + -3.0324e-05, -2.3246e-05], + [ 7.1898e-07, 2.0638e-06, 0.0000e+00, ..., 1.4622e-07, + 2.0508e-06, 1.5488e-06], + [ 1.6876e-06, 4.8913e-06, 0.0000e+00, ..., 1.6228e-07, + 4.3847e-06, 3.2354e-06]], device='cuda:0') +Epoch 163, bias, value: tensor([-0.0326, 0.0115, 0.0153, 0.0250, 0.0284, -0.0179, 0.0276, 0.0300, + -0.0099, 0.0035], device='cuda:0'), grad: tensor([-4.7356e-05, 7.7412e-06, 8.7321e-06, 8.3819e-06, 3.2961e-05, + 1.8179e-05, 9.8273e-06, -6.9916e-05, 8.9258e-06, 2.2545e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 249.35, cls_loss 0.0013 cls_loss_mapping 0.0047 cls_loss_causal 0.5243 re_mapping 0.0049 re_causal 0.0148 /// teacc 99.10 lr 0.00010000 +Epoch 164, weight, value: tensor([[ 0.0545, -0.1022, -0.1227, ..., -0.0995, -0.0979, -0.0903], + [ 0.0269, 0.0467, -0.0292, ..., -0.0336, 0.1382, 0.0169], + [-0.0092, -0.1289, -0.0115, ..., 0.1367, -0.1149, 0.0289], + ..., + [-0.0721, 0.1118, 0.0306, ..., -0.0945, 0.0912, 0.0444], + [-0.0018, -0.0567, 0.0055, ..., -0.0487, -0.1508, -0.1009], + [-0.0015, -0.0106, -0.0029, ..., -0.1497, -0.1175, 0.0298]], + device='cuda:0'), grad: tensor([[-1.2014e-07, 1.1205e-08, 3.4925e-10, ..., 2.3562e-07, + 1.0536e-08, 6.4843e-08], + [ 3.2503e-07, 3.6741e-07, 1.8917e-09, ..., 3.9930e-07, + 2.5239e-07, 3.2084e-07], + [ 4.0159e-06, 2.6473e-07, 3.8708e-09, ..., 6.3181e-06, + 2.3551e-07, 3.5012e-08], + ..., + [ 2.3260e-07, -1.2135e-06, 3.8999e-09, ..., 1.4657e-07, + -1.0226e-06, -5.8906e-07], + [-5.6475e-06, 3.4022e-08, 4.9477e-10, ..., -1.5404e-06, + 2.1857e-08, 9.9943e-08], + [ 3.9861e-07, 6.1817e-08, 2.0082e-09, ..., 6.8918e-07, + 7.9453e-08, 1.3290e-06]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0320, 0.0112, 0.0154, 0.0249, 0.0269, -0.0183, 0.0282, 0.0302, + -0.0097, 0.0040], device='cuda:0'), grad: tensor([-2.8755e-07, 4.2729e-06, 3.3528e-05, -1.5810e-05, -6.5826e-06, + 6.4932e-06, -9.8720e-06, 1.5981e-06, -2.2203e-05, 8.8289e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 248.84, cls_loss 0.0015 cls_loss_mapping 0.0036 cls_loss_causal 0.5112 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.04 lr 0.00010000 +Epoch 165, weight, value: tensor([[ 0.0546, -0.1025, -0.1227, ..., -0.0998, -0.0990, -0.0905], + [ 0.0270, 0.0464, -0.0292, ..., -0.0351, 0.1385, 0.0165], + [-0.0093, -0.1291, -0.0115, ..., 0.1372, -0.1148, 0.0295], + ..., + [-0.0726, 0.1120, 0.0306, ..., -0.0957, 0.0911, 0.0443], + [-0.0018, -0.0571, 0.0055, ..., -0.0487, -0.1509, -0.1018], + [-0.0015, -0.0106, -0.0029, ..., -0.1505, -0.1177, 0.0297]], + device='cuda:0'), grad: tensor([[-2.2148e-08, 1.6449e-07, 0.0000e+00, ..., 3.1316e-07, + 3.5483e-07, 2.0489e-07], + [-3.2820e-06, -5.3179e-07, 0.0000e+00, ..., 5.9896e-08, + -7.3761e-06, -1.9297e-06], + [ 6.4261e-07, 2.2147e-06, 0.0000e+00, ..., 1.8277e-08, + 3.2429e-06, 2.6245e-06], + ..., + [ 4.4913e-07, -4.8317e-06, 0.0000e+00, ..., 2.9890e-08, + -4.6194e-06, -5.7742e-06], + [ 3.9744e-07, 1.0040e-06, 0.0000e+00, ..., 1.6845e-07, + 2.2221e-06, 1.4715e-06], + [-7.6892e-08, 2.7986e-07, 0.0000e+00, ..., 1.6263e-07, + 4.7451e-07, 2.8964e-07]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0320, 0.0110, 0.0153, 0.0250, 0.0287, -0.0184, 0.0275, 0.0298, + -0.0098, 0.0039], device='cuda:0'), grad: tensor([ 9.0431e-07, -1.0975e-05, 6.9067e-06, 5.7518e-06, 6.1728e-06, + 5.2005e-06, -4.8876e-06, -1.2599e-05, 4.9472e-06, -1.4454e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 249.04, cls_loss 0.0014 cls_loss_mapping 0.0042 cls_loss_causal 0.5089 re_mapping 0.0049 re_causal 0.0148 /// teacc 98.83 lr 0.00010000 +Epoch 166, weight, value: tensor([[ 0.0547, -0.1026, -0.1227, ..., -0.1001, -0.0995, -0.0906], + [ 0.0271, 0.0478, -0.0292, ..., -0.0325, 0.1403, 0.0182], + [-0.0093, -0.1298, -0.0115, ..., 0.1386, -0.1154, 0.0297], + ..., + [-0.0727, 0.1124, 0.0306, ..., -0.0962, 0.0914, 0.0446], + [-0.0018, -0.0573, 0.0055, ..., -0.0489, -0.1510, -0.1028], + [-0.0016, -0.0107, -0.0029, ..., -0.1523, -0.1179, 0.0288]], + device='cuda:0'), grad: tensor([[ 1.3632e-07, 2.8464e-08, 0.0000e+00, ..., 1.6380e-07, + 4.1415e-08, 2.1572e-07], + [ 3.5996e-07, 1.2363e-07, 0.0000e+00, ..., 6.2108e-08, + 2.5355e-07, 2.3982e-07], + [ 4.0010e-06, 6.9104e-07, 0.0000e+00, ..., 1.5786e-06, + 1.1111e-06, 3.9451e-06], + ..., + [ 2.0384e-07, -1.1707e-06, 0.0000e+00, ..., 3.1228e-08, + -1.3132e-06, -1.7025e-06], + [-3.2131e-06, 2.8842e-08, 0.0000e+00, ..., 8.4192e-07, + 1.1036e-07, 5.4762e-07], + [-4.6566e-06, 1.7602e-07, 0.0000e+00, ..., -1.2461e-06, + 2.2561e-07, -4.1276e-06]], device='cuda:0') +Epoch 166, bias, value: tensor([-0.0317, 0.0127, 0.0154, 0.0249, 0.0274, -0.0180, 0.0271, 0.0300, + -0.0098, 0.0027], device='cuda:0'), grad: tensor([ 1.8468e-06, 3.4161e-06, 3.5584e-05, 4.1872e-06, 3.2205e-06, + 5.3570e-06, -2.6710e-06, -1.6307e-06, -1.5706e-05, -3.3617e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 248.85, cls_loss 0.0015 cls_loss_mapping 0.0045 cls_loss_causal 0.5337 re_mapping 0.0045 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +Epoch 167, weight, value: tensor([[ 0.0547, -0.1028, -0.1227, ..., -0.1003, -0.1001, -0.0907], + [ 0.0270, 0.0473, -0.0292, ..., -0.0329, 0.1402, 0.0179], + [-0.0095, -0.1306, -0.0115, ..., 0.1390, -0.1160, 0.0297], + ..., + [-0.0730, 0.1131, 0.0306, ..., -0.0966, 0.0920, 0.0449], + [-0.0017, -0.0574, 0.0055, ..., -0.0492, -0.1510, -0.1034], + [-0.0017, -0.0108, -0.0029, ..., -0.1542, -0.1180, 0.0276]], + device='cuda:0'), grad: tensor([[ 1.8738e-06, 2.0047e-07, 0.0000e+00, ..., 8.1770e-07, + 1.6502e-08, 3.8155e-08], + [ 5.0552e-06, 4.2561e-07, 0.0000e+00, ..., 2.1756e-06, + -2.5029e-07, 2.9744e-08], + [ 7.5884e-06, 4.8196e-07, 0.0000e+00, ..., 1.2731e-06, + 4.5809e-08, -3.1129e-07], + ..., + [ 1.2312e-06, 1.6775e-07, 0.0000e+00, ..., 6.3609e-07, + 9.5053e-08, 1.7462e-07], + [-3.8028e-05, -3.3975e-06, 0.0000e+00, ..., -1.2957e-05, + 1.5105e-08, 2.2899e-07], + [ 3.5372e-06, 3.1525e-07, 0.0000e+00, ..., 2.6133e-06, + 9.9244e-09, 9.6858e-07]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0319, 0.0121, 0.0153, 0.0257, 0.0287, -0.0190, 0.0286, 0.0303, + -0.0099, 0.0013], device='cuda:0'), grad: tensor([ 1.1012e-05, 2.9951e-05, 4.3303e-05, 5.5224e-05, -6.7130e-06, + 2.5764e-05, 2.5690e-05, 7.8380e-06, -2.1863e-04, 2.6554e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 248.60, cls_loss 0.0022 cls_loss_mapping 0.0044 cls_loss_causal 0.4896 re_mapping 0.0048 re_causal 0.0139 /// teacc 99.06 lr 0.00010000 +Epoch 168, weight, value: tensor([[ 0.0548, -0.1033, -0.1228, ..., -0.1005, -0.1008, -0.0910], + [ 0.0263, 0.0494, -0.0292, ..., -0.0335, 0.1425, 0.0200], + [-0.0098, -0.1345, -0.0125, ..., 0.1393, -0.1193, 0.0277], + ..., + [-0.0726, 0.1113, 0.0305, ..., -0.0956, 0.0900, 0.0432], + [-0.0014, -0.0573, 0.0054, ..., -0.0495, -0.1511, -0.1051], + [-0.0021, -0.0111, -0.0030, ..., -0.1547, -0.1184, 0.0274]], + device='cuda:0'), grad: tensor([[ 3.7486e-08, 2.2410e-08, 0.0000e+00, ..., 1.4959e-08, + 4.5169e-08, 2.1828e-08], + [-3.1642e-07, -1.4051e-07, 0.0000e+00, ..., 2.3341e-08, + -6.7754e-07, -1.1059e-09], + [ 3.0675e-08, 4.4703e-08, 0.0000e+00, ..., -8.0967e-08, + 6.7870e-08, -2.7881e-08], + ..., + [ 9.0280e-08, -3.7393e-07, 0.0000e+00, ..., 2.3283e-08, + -1.6834e-07, -3.6345e-07], + [ 1.6112e-07, 1.2829e-07, 0.0000e+00, ..., 5.5588e-08, + 2.9337e-07, 1.5856e-07], + [ 3.0850e-07, 1.0314e-07, 0.0000e+00, ..., 3.0093e-08, + 1.2631e-07, 5.8324e-08]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0318, 0.0136, 0.0141, 0.0256, 0.0290, -0.0192, 0.0304, 0.0288, + -0.0096, 0.0009], device='cuda:0'), grad: tensor([ 4.2468e-07, 8.8185e-08, 6.6776e-07, -5.2191e-06, 9.3307e-08, + -1.3877e-06, 8.6334e-07, -1.5949e-07, 7.1432e-07, 3.9004e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 249.41, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.5031 re_mapping 0.0045 re_causal 0.0142 /// teacc 99.00 lr 0.00010000 +Epoch 169, weight, value: tensor([[ 0.0549, -0.1034, -0.1229, ..., -0.1005, -0.1010, -0.0911], + [ 0.0263, 0.0493, -0.0295, ..., -0.0337, 0.1425, 0.0199], + [-0.0099, -0.1348, -0.0137, ..., 0.1398, -0.1194, 0.0280], + ..., + [-0.0728, 0.1114, 0.0305, ..., -0.0958, 0.0901, 0.0432], + [-0.0012, -0.0571, 0.0054, ..., -0.0498, -0.1511, -0.1061], + [-0.0022, -0.0112, -0.0030, ..., -0.1548, -0.1185, 0.0274]], + device='cuda:0'), grad: tensor([[-1.8661e-07, 1.9616e-08, 0.0000e+00, ..., 1.4796e-07, + 6.7754e-08, 1.3737e-08], + [ 4.8466e-06, 5.8766e-07, 0.0000e+00, ..., 3.9116e-06, + 2.6952e-06, -9.8895e-08], + [ 1.4380e-05, 2.1569e-06, 0.0000e+00, ..., 1.1057e-05, + 8.9854e-06, 4.9244e-08], + ..., + [ 1.0412e-06, 2.7497e-07, 0.0000e+00, ..., 7.2783e-07, + 7.1106e-07, 8.8522e-07], + [ 1.6242e-06, 2.9732e-07, 0.0000e+00, ..., 1.4044e-06, + 1.1763e-06, 3.6648e-07], + [ 4.7870e-07, 9.3831e-08, 0.0000e+00, ..., 3.5809e-07, + 2.9500e-07, 1.7171e-08]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0317, 0.0134, 0.0141, 0.0257, 0.0290, -0.0191, 0.0301, 0.0289, + -0.0093, 0.0008], device='cuda:0'), grad: tensor([-6.5565e-07, 2.3738e-05, 6.8903e-05, -1.2815e-04, -2.2352e-06, + 2.0906e-05, -5.4669e-07, 6.8843e-06, 8.1882e-06, 2.7549e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 249.34, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.5159 re_mapping 0.0045 re_causal 0.0141 /// teacc 99.03 lr 0.00010000 +Epoch 170, weight, value: tensor([[ 0.0549, -0.1035, -0.1229, ..., -0.1007, -0.1014, -0.0913], + [ 0.0263, 0.0493, -0.0294, ..., -0.0339, 0.1424, 0.0198], + [-0.0099, -0.1353, -0.0137, ..., 0.1399, -0.1214, 0.0267], + ..., + [-0.0730, 0.1116, 0.0301, ..., -0.0949, 0.0904, 0.0437], + [-0.0012, -0.0572, 0.0055, ..., -0.0498, -0.1512, -0.1069], + [-0.0022, -0.0114, -0.0031, ..., -0.1550, -0.1188, 0.0273]], + device='cuda:0'), grad: tensor([[ 9.0105e-08, 1.0006e-07, 0.0000e+00, ..., 1.0122e-07, + 1.9302e-07, 1.3527e-07], + [-4.2245e-06, -1.8636e-06, 0.0000e+00, ..., 9.8604e-08, + -5.4985e-06, -9.2387e-07], + [-7.0082e-08, 2.0657e-06, 0.0000e+00, ..., -2.4177e-06, + 3.2540e-06, 3.1926e-06], + ..., + [ 1.7202e-06, -3.0231e-06, 0.0000e+00, ..., 6.5018e-08, + -3.2634e-06, -5.5991e-06], + [ 5.1130e-07, 5.0571e-07, 0.0000e+00, ..., 1.2359e-06, + 1.0850e-06, 6.0257e-07], + [ 1.2834e-06, 1.3821e-06, 0.0000e+00, ..., 1.5018e-08, + 2.7027e-06, 1.4445e-06]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0317, 0.0133, 0.0134, 0.0257, 0.0291, -0.0187, 0.0297, 0.0292, + -0.0093, 0.0006], device='cuda:0'), grad: tensor([ 6.4867e-07, -9.1195e-06, 2.8256e-06, 3.4608e-06, 6.1886e-07, + 9.6392e-07, 4.5681e-07, -8.7097e-06, 3.5912e-06, 5.2266e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 248.52, cls_loss 0.0014 cls_loss_mapping 0.0034 cls_loss_causal 0.5381 re_mapping 0.0042 re_causal 0.0138 /// teacc 98.92 lr 0.00010000 +Epoch 171, weight, value: tensor([[ 0.0549, -0.1037, -0.1229, ..., -0.1012, -0.1018, -0.0914], + [ 0.0263, 0.0493, -0.0291, ..., -0.0342, 0.1424, 0.0198], + [-0.0098, -0.1354, -0.0137, ..., 0.1410, -0.1219, 0.0268], + ..., + [-0.0734, 0.1116, 0.0296, ..., -0.0954, 0.0904, 0.0438], + [-0.0009, -0.0570, 0.0053, ..., -0.0500, -0.1512, -0.1080], + [-0.0016, -0.0113, -0.0031, ..., -0.1548, -0.1188, 0.0275]], + device='cuda:0'), grad: tensor([[ 1.0477e-08, 1.2631e-08, 5.8208e-10, ..., 1.2631e-07, + 2.6717e-08, 3.6089e-08], + [-3.7393e-07, -1.6612e-07, 1.4552e-09, ..., 7.6892e-08, + -9.4669e-07, -5.3027e-08], + [-1.2736e-07, 5.5414e-08, 5.2387e-10, ..., -1.0412e-06, + -8.3237e-09, -1.1697e-06], + ..., + [ 7.3481e-07, 1.3066e-06, 5.2387e-10, ..., 7.4832e-07, + 8.4424e-07, 2.7828e-06], + [-4.5809e-08, 1.5134e-07, -3.4866e-08, ..., 2.5867e-07, + 2.2911e-07, 3.9232e-07], + [-6.4913e-07, -2.1122e-06, 6.4028e-10, ..., 1.7986e-08, + -6.0350e-07, -3.1125e-06]], device='cuda:0') +Epoch 171, bias, value: tensor([-0.0319, 0.0132, 0.0136, 0.0257, 0.0288, -0.0187, 0.0291, 0.0292, + -0.0087, 0.0009], device='cuda:0'), grad: tensor([ 3.9488e-07, -7.0175e-07, -2.3749e-06, 3.6974e-07, 2.2333e-06, + 7.2364e-07, -5.3877e-07, 7.6517e-06, 6.5006e-07, -8.4043e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 248.86, cls_loss 0.0015 cls_loss_mapping 0.0041 cls_loss_causal 0.5224 re_mapping 0.0040 re_causal 0.0134 /// teacc 99.04 lr 0.00010000 +Epoch 172, weight, value: tensor([[ 0.0554, -0.1027, -0.1230, ..., -0.1019, -0.1019, -0.0915], + [ 0.0263, 0.0492, -0.0290, ..., -0.0345, 0.1424, 0.0197], + [-0.0098, -0.1356, -0.0137, ..., 0.1422, -0.1219, 0.0272], + ..., + [-0.0741, 0.1118, 0.0295, ..., -0.0962, 0.0905, 0.0439], + [-0.0006, -0.0567, 0.0053, ..., -0.0503, -0.1514, -0.1087], + [-0.0011, -0.0118, -0.0031, ..., -0.1547, -0.1192, 0.0276]], + device='cuda:0'), grad: tensor([[-7.5670e-09, 9.1386e-09, 0.0000e+00, ..., 4.4529e-08, + 1.3155e-08, 5.2503e-08], + [-6.5891e-08, -2.9861e-08, 0.0000e+00, ..., 5.4657e-08, + -1.7905e-07, 8.0501e-08], + [ 2.9686e-09, 7.2760e-09, 0.0000e+00, ..., -7.0455e-07, + 3.4226e-08, -7.6555e-07], + ..., + [ 1.0675e-07, 4.5344e-08, 0.0000e+00, ..., 3.5623e-07, + 3.8592e-08, 4.9965e-07], + [ 6.3272e-08, 2.0489e-08, 0.0000e+00, ..., 2.0419e-07, + 2.8871e-08, 2.6799e-07], + [-1.9115e-07, 6.4669e-08, 0.0000e+00, ..., 3.2014e-08, + 1.5891e-08, -6.1758e-08]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0317, 0.0131, 0.0138, 0.0249, 0.0285, -0.0174, 0.0290, 0.0292, + -0.0082, 0.0010], device='cuda:0'), grad: tensor([ 6.8161e-08, 1.4994e-07, -1.4957e-06, -6.3144e-07, -6.1048e-07, + 6.3609e-07, 6.9616e-08, 1.4892e-06, 7.8976e-07, -4.3958e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 248.68, cls_loss 0.0014 cls_loss_mapping 0.0031 cls_loss_causal 0.5235 re_mapping 0.0041 re_causal 0.0137 /// teacc 98.94 lr 0.00010000 +Epoch 173, weight, value: tensor([[ 0.0553, -0.1029, -0.1230, ..., -0.1024, -0.1026, -0.0916], + [ 0.0263, 0.0492, -0.0291, ..., -0.0348, 0.1425, 0.0196], + [-0.0097, -0.1358, -0.0136, ..., 0.1435, -0.1218, 0.0279], + ..., + [-0.0745, 0.1118, 0.0294, ..., -0.0977, 0.0906, 0.0439], + [-0.0007, -0.0572, 0.0052, ..., -0.0510, -0.1515, -0.1101], + [-0.0008, -0.0118, -0.0031, ..., -0.1548, -0.1195, 0.0277]], + device='cuda:0'), grad: tensor([[-3.9814e-07, 6.2282e-09, 0.0000e+00, ..., 2.7218e-07, + 7.6834e-09, 9.3889e-08], + [-7.2119e-08, -1.0029e-07, 0.0000e+00, ..., 2.4308e-07, + -2.2550e-07, 2.1711e-08], + [ 4.5635e-08, 1.9441e-08, 0.0000e+00, ..., -1.6745e-06, + 2.9395e-08, -8.8615e-07], + ..., + [ 9.0920e-08, -8.9128e-07, 0.0000e+00, ..., 2.6962e-07, + -5.5647e-07, -5.2806e-07], + [ 1.0070e-07, 2.4564e-08, 0.0000e+00, ..., 4.0862e-07, + 3.4168e-08, 1.8091e-07], + [-1.4040e-07, 9.1328e-08, 0.0000e+00, ..., 8.7079e-08, + 8.9000e-08, -8.2073e-08]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0318, 0.0130, 0.0126, 0.0267, 0.0285, -0.0174, 0.0289, 0.0292, + -0.0087, 0.0011], device='cuda:0'), grad: tensor([-9.3784e-07, 3.6252e-07, -3.3695e-06, 3.4682e-06, 1.0300e-06, + 1.2731e-06, -1.5516e-06, -7.3528e-07, 1.5171e-06, -1.0636e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 249.47, cls_loss 0.0014 cls_loss_mapping 0.0045 cls_loss_causal 0.5132 re_mapping 0.0045 re_causal 0.0138 /// teacc 98.87 lr 0.00010000 +Epoch 174, weight, value: tensor([[ 0.0562, -0.1030, -0.1233, ..., -0.1004, -0.1030, -0.0919], + [ 0.0263, 0.0491, -0.0296, ..., -0.0353, 0.1424, 0.0195], + [-0.0095, -0.1358, -0.0111, ..., 0.1466, -0.1218, 0.0294], + ..., + [-0.0746, 0.1120, 0.0284, ..., -0.0993, 0.0907, 0.0440], + [-0.0008, -0.0574, 0.0047, ..., -0.0517, -0.1516, -0.1120], + [-0.0007, -0.0121, -0.0033, ..., -0.1549, -0.1197, 0.0277]], + device='cuda:0'), grad: tensor([[-8.9698e-08, 3.1258e-08, 0.0000e+00, ..., 6.0536e-08, + 6.5425e-08, 1.2596e-07], + [-5.0757e-07, 5.5227e-07, 0.0000e+00, ..., 2.1397e-07, + -4.3958e-07, 1.2415e-06], + [ 1.1292e-07, 9.5286e-08, 0.0000e+00, ..., 8.5274e-08, + 1.8440e-07, 2.4238e-07], + ..., + [ 2.8755e-07, -1.3122e-06, 0.0000e+00, ..., 4.7637e-07, + -8.4657e-07, 1.2922e-08], + [-9.7440e-08, 9.8720e-08, 0.0000e+00, ..., 8.1782e-08, + 2.3132e-07, 4.0932e-07], + [ 7.2119e-08, 1.7043e-07, 0.0000e+00, ..., 5.3924e-07, + 2.4959e-07, 1.8235e-06]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0301, 0.0129, 0.0139, 0.0262, 0.0283, -0.0174, 0.0280, 0.0292, + -0.0093, 0.0011], device='cuda:0'), grad: tensor([ 1.0885e-08, 2.4699e-06, 1.5413e-06, 7.9954e-07, -1.9625e-05, + 1.1139e-06, 1.2768e-06, 3.9935e-06, 2.9383e-07, 8.0764e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 248.89, cls_loss 0.0015 cls_loss_mapping 0.0045 cls_loss_causal 0.5254 re_mapping 0.0044 re_causal 0.0138 /// teacc 98.96 lr 0.00010000 +Epoch 175, weight, value: tensor([[ 0.0562, -0.1034, -0.1233, ..., -0.1009, -0.1038, -0.0922], + [ 0.0263, 0.0491, -0.0311, ..., -0.0359, 0.1424, 0.0194], + [-0.0094, -0.1355, -0.0107, ..., 0.1487, -0.1219, 0.0308], + ..., + [-0.0748, 0.1120, 0.0284, ..., -0.0999, 0.0907, 0.0439], + [-0.0006, -0.0582, 0.0051, ..., -0.0528, -0.1517, -0.1151], + [-0.0006, -0.0122, -0.0033, ..., -0.1553, -0.1199, 0.0273]], + device='cuda:0'), grad: tensor([[-7.6834e-09, 2.8033e-07, 0.0000e+00, ..., 1.6589e-08, + 2.1467e-07, 1.1141e-07], + [-1.9022e-07, 7.9535e-07, 0.0000e+00, ..., 3.8941e-08, + 2.0594e-07, 6.1933e-07], + [ 5.6636e-08, 1.1735e-06, 0.0000e+00, ..., 4.4587e-08, + 8.8289e-07, 5.1130e-07], + ..., + [ 9.5170e-08, -1.2137e-05, 0.0000e+00, ..., 2.5495e-08, + -8.9779e-06, -6.5863e-06], + [ 1.9080e-07, 4.0280e-07, 0.0000e+00, ..., 2.5670e-08, + 4.4703e-07, 2.5728e-07], + [ 1.2398e-08, 6.3591e-06, 0.0000e+00, ..., 3.1060e-07, + 4.8615e-06, 3.9041e-06]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0303, 0.0126, 0.0151, 0.0260, 0.0294, -0.0175, 0.0274, 0.0291, + -0.0098, 0.0006], device='cuda:0'), grad: tensor([ 1.4370e-06, 2.1663e-06, 4.0196e-06, -2.4855e-08, -5.6170e-08, + 1.1306e-06, 8.1724e-07, -2.9460e-05, 2.3991e-06, 1.7494e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 248.90, cls_loss 0.0013 cls_loss_mapping 0.0033 cls_loss_causal 0.5121 re_mapping 0.0044 re_causal 0.0133 /// teacc 99.07 lr 0.00010000 +Epoch 176, weight, value: tensor([[ 0.0561, -0.1036, -0.1233, ..., -0.1013, -0.1045, -0.0924], + [ 0.0264, 0.0490, -0.0313, ..., -0.0361, 0.1424, 0.0191], + [-0.0095, -0.1361, -0.0107, ..., 0.1497, -0.1222, 0.0308], + ..., + [-0.0751, 0.1122, 0.0284, ..., -0.1003, 0.0908, 0.0439], + [-0.0004, -0.0585, 0.0051, ..., -0.0527, -0.1519, -0.1156], + [-0.0006, -0.0121, -0.0033, ..., -0.1554, -0.1192, 0.0285]], + device='cuda:0'), grad: tensor([[-3.0547e-07, 2.5204e-08, 0.0000e+00, ..., -3.2852e-07, + 2.3399e-08, 2.8289e-08], + [-3.1549e-07, -4.4610e-07, 0.0000e+00, ..., 1.1176e-08, + -1.1679e-06, 5.0000e-08], + [ 1.6217e-07, 1.3062e-07, 0.0000e+00, ..., 1.5134e-07, + 1.3527e-07, 8.9116e-08], + ..., + [ 1.4086e-07, -7.1479e-07, 0.0000e+00, ..., 7.8580e-09, + -2.7218e-07, -7.5763e-07], + [ 3.2759e-07, 3.4180e-07, 0.0000e+00, ..., -4.0163e-09, + 4.0000e-07, 2.0617e-07], + [-2.5565e-07, 2.8452e-07, 0.0000e+00, ..., 4.5635e-08, + 5.1269e-07, 1.1380e-07]], device='cuda:0') +Epoch 176, bias, value: tensor([-0.0302, 0.0123, 0.0151, 0.0257, 0.0292, -0.0171, 0.0271, 0.0290, + -0.0096, 0.0017], device='cuda:0'), grad: tensor([-2.1029e-06, -1.7639e-06, 1.2787e-06, 8.8243e-07, 4.6403e-07, + 1.4494e-08, 2.9779e-07, -9.1363e-07, 1.4836e-06, 3.5344e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 249.02, cls_loss 0.0012 cls_loss_mapping 0.0034 cls_loss_causal 0.5039 re_mapping 0.0045 re_causal 0.0133 /// teacc 99.03 lr 0.00010000 +Epoch 177, weight, value: tensor([[ 0.0562, -0.1039, -0.1236, ..., -0.1015, -0.1051, -0.0925], + [ 0.0265, 0.0489, -0.0313, ..., -0.0361, 0.1424, 0.0190], + [-0.0097, -0.1370, -0.0107, ..., 0.1497, -0.1228, 0.0306], + ..., + [-0.0753, 0.1125, 0.0283, ..., -0.1006, 0.0909, 0.0442], + [-0.0004, -0.0587, 0.0050, ..., -0.0529, -0.1520, -0.1160], + [-0.0005, -0.0125, -0.0033, ..., -0.1556, -0.1196, 0.0284]], + device='cuda:0'), grad: tensor([[-8.7358e-07, 8.7486e-08, 0.0000e+00, ..., 4.5728e-07, + 7.4273e-08, 8.7719e-08], + [-1.6147e-07, 6.5379e-07, 0.0000e+00, ..., 5.8266e-08, + 5.3842e-08, 6.6543e-07], + [ 1.0617e-07, 6.1467e-07, 0.0000e+00, ..., -1.6880e-08, + 4.8988e-07, 5.7416e-07], + ..., + [ 1.4412e-07, -8.4713e-06, 0.0000e+00, ..., 1.3970e-08, + -5.4725e-06, -6.9216e-06], + [ 8.0746e-07, 1.6333e-07, 0.0000e+00, ..., 3.8743e-07, + 2.1106e-07, 1.1586e-06], + [-6.5658e-07, 5.8636e-06, 0.0000e+00, ..., 4.9651e-08, + 4.5151e-06, 2.5872e-06]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0302, 0.0123, 0.0147, 0.0260, 0.0293, -0.0173, 0.0267, 0.0293, + -0.0098, 0.0018], device='cuda:0'), grad: tensor([-5.6326e-06, 1.3113e-06, 1.9968e-06, -7.7188e-06, 3.0715e-06, + 9.2760e-07, -1.5814e-06, -6.1989e-06, 7.9498e-06, 5.8301e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 249.03, cls_loss 0.0014 cls_loss_mapping 0.0044 cls_loss_causal 0.5068 re_mapping 0.0044 re_causal 0.0131 /// teacc 98.96 lr 0.00010000 +Epoch 178, weight, value: tensor([[ 0.0562, -0.1041, -0.1243, ..., -0.1018, -0.1055, -0.0927], + [ 0.0266, 0.0489, -0.0317, ..., -0.0364, 0.1424, 0.0190], + [-0.0097, -0.1376, -0.0108, ..., 0.1503, -0.1230, 0.0308], + ..., + [-0.0757, 0.1125, 0.0281, ..., -0.1018, 0.0910, 0.0441], + [-0.0003, -0.0589, 0.0063, ..., -0.0537, -0.1521, -0.1164], + [-0.0004, -0.0127, -0.0034, ..., -0.1558, -0.1198, 0.0285]], + device='cuda:0'), grad: tensor([[ 5.3421e-06, 5.7509e-08, 0.0000e+00, ..., 3.4750e-08, + 3.7777e-08, 9.5519e-08], + [ 1.8533e-07, 7.8091e-07, 0.0000e+00, ..., 1.4773e-07, + 4.9686e-07, 8.4657e-07], + [ 3.5064e-07, 2.5961e-07, 0.0000e+00, ..., -9.0920e-08, + 1.6717e-07, 1.8533e-07], + ..., + [ 2.8964e-07, -3.8967e-06, 0.0000e+00, ..., 1.1665e-07, + -2.4512e-06, -3.1292e-06], + [-1.3359e-05, 9.5111e-08, 0.0000e+00, ..., 4.5635e-08, + 6.2515e-08, 1.6578e-07], + [ 5.1036e-06, 2.1029e-06, 0.0000e+00, ..., 3.6079e-06, + 1.2545e-06, 5.2340e-06]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0304, 0.0122, 0.0147, 0.0261, 0.0297, -0.0169, 0.0267, 0.0291, + -0.0098, 0.0018], device='cuda:0'), grad: tensor([ 3.0324e-05, 3.1348e-06, 2.6878e-06, 8.6799e-06, -1.5013e-05, + 1.4426e-06, 3.6210e-06, -5.0552e-06, -7.6890e-05, 4.7147e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 248.88, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.4883 re_mapping 0.0044 re_causal 0.0138 /// teacc 98.97 lr 0.00010000 +Epoch 179, weight, value: tensor([[ 0.0562, -0.1043, -0.1243, ..., -0.1023, -0.1060, -0.0929], + [ 0.0260, 0.0489, -0.0318, ..., -0.0366, 0.1424, 0.0189], + [-0.0097, -0.1383, -0.0108, ..., 0.1508, -0.1232, 0.0307], + ..., + [-0.0760, 0.1127, 0.0281, ..., -0.1023, 0.0910, 0.0443], + [ 0.0002, -0.0582, 0.0063, ..., -0.0542, -0.1515, -0.1167], + [-0.0005, -0.0137, -0.0034, ..., -0.1563, -0.1203, 0.0280]], + device='cuda:0'), grad: tensor([[ 1.0547e-07, 2.6310e-07, 1.6752e-07, ..., 1.3513e-06, + 2.4331e-07, 2.3236e-07], + [ 1.6321e-07, 9.7882e-07, 7.4506e-09, ..., 1.8708e-07, + 8.4983e-07, 7.7393e-07], + [ 4.1444e-08, 2.7381e-07, 2.6193e-09, ..., -4.7218e-07, + 2.0990e-07, -1.1077e-07], + ..., + [ 9.7207e-09, 2.4527e-05, 1.1642e-10, ..., 1.0629e-07, + -2.1625e-06, 2.9862e-05], + [-4.5053e-07, 5.4901e-07, 1.5134e-08, ..., -4.4238e-09, + 1.3621e-07, 6.4680e-07], + [-2.3574e-08, -2.7612e-05, 7.0431e-09, ..., 9.2259e-08, + 1.7486e-07, -3.2485e-05]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0307, 0.0118, 0.0146, 0.0260, 0.0301, -0.0169, 0.0271, 0.0293, + -0.0093, 0.0013], device='cuda:0'), grad: tensor([ 4.1723e-06, 4.3362e-06, 2.9011e-07, 2.4475e-06, 1.0058e-06, + 6.7689e-06, -8.9779e-06, 7.2837e-05, -1.8068e-06, -8.1182e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 249.36, cls_loss 0.0014 cls_loss_mapping 0.0045 cls_loss_causal 0.5113 re_mapping 0.0041 re_causal 0.0131 /// teacc 98.82 lr 0.00010000 +Epoch 180, weight, value: tensor([[ 5.6317e-02, -1.0427e-01, -1.2434e-01, ..., -1.0288e-01, + -1.0662e-01, -9.3212e-02], + [ 2.6121e-02, 4.8859e-02, -3.1823e-02, ..., -3.6968e-02, + 1.4240e-01, 1.8902e-02], + [-9.7369e-03, -1.3891e-01, -1.0743e-02, ..., 1.5115e-01, + -1.2349e-01, 3.0629e-02], + ..., + [-7.6451e-02, 1.1277e-01, 2.8054e-02, ..., -1.0268e-01, + 9.1130e-02, 4.4415e-02], + [ 1.3948e-04, -5.8052e-02, 6.3755e-03, ..., -5.3928e-02, + -1.5143e-01, -1.1696e-01], + [ 2.8459e-04, -1.3948e-02, -3.4362e-03, ..., -1.5683e-01, + -1.2066e-01, 2.7726e-02]], device='cuda:0'), grad: tensor([[-2.0198e-08, -5.5297e-09, -7.5670e-10, ..., 1.4424e-07, + 1.6182e-08, 1.3062e-07], + [ 5.8953e-07, 4.6275e-08, 5.8208e-11, ..., 1.6131e-06, + 4.5402e-09, 1.3476e-06], + [-1.4147e-06, -1.3283e-07, 0.0000e+00, ..., -3.6452e-06, + -2.4796e-07, -2.8592e-06], + ..., + [ 3.0245e-07, 5.0990e-08, 5.8208e-11, ..., 5.5414e-07, + -2.1304e-08, 6.9151e-07], + [ 1.3423e-07, 5.3959e-08, 1.7462e-10, ..., 6.3283e-07, + 8.8534e-08, 4.2329e-07], + [-4.9185e-08, -1.2864e-07, 1.1642e-10, ..., 1.5879e-07, + 3.9756e-08, -5.7882e-07]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0320, 0.0117, 0.0145, 0.0261, 0.0308, -0.0170, 0.0279, 0.0294, + -0.0092, 0.0007], device='cuda:0'), grad: tensor([ 2.0140e-07, 4.1053e-06, -8.7172e-06, 1.9427e-06, 6.9290e-07, + 5.4389e-07, -7.0035e-07, 2.3860e-06, 7.9302e-07, -1.2480e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 249.31, cls_loss 0.0019 cls_loss_mapping 0.0050 cls_loss_causal 0.5283 re_mapping 0.0043 re_causal 0.0133 /// teacc 98.98 lr 0.00010000 +Epoch 181, weight, value: tensor([[ 0.0564, -0.1046, -0.1256, ..., -0.1038, -0.1075, -0.0937], + [ 0.0290, 0.0494, -0.0323, ..., -0.0372, 0.1430, 0.0188], + [-0.0098, -0.1419, -0.0102, ..., 0.1505, -0.1248, 0.0296], + ..., + [-0.0767, 0.1132, 0.0271, ..., -0.1004, 0.0914, 0.0449], + [-0.0027, -0.0620, 0.0034, ..., -0.0587, -0.1544, -0.1178], + [ 0.0019, -0.0146, -0.0035, ..., -0.1562, -0.1214, 0.0290]], + device='cuda:0'), grad: tensor([[-1.6415e-08, 9.4878e-09, 5.2387e-10, ..., 1.0594e-08, + 3.7893e-08, 1.3271e-08], + [ 3.6345e-07, -7.4320e-07, 2.9104e-10, ..., 2.1141e-07, + -4.5472e-07, -1.2410e-07], + [ 1.2782e-07, 4.0745e-08, -2.1246e-08, ..., -2.0489e-07, + 1.9174e-07, -2.2165e-07], + ..., + [ 1.7497e-07, 1.0245e-08, 3.4925e-09, ..., 5.9314e-08, + 2.4517e-07, 2.9686e-08], + [-2.0936e-06, 1.2491e-07, 2.2701e-09, ..., 2.9220e-08, + -2.5164e-06, 7.7649e-08], + [ 6.3702e-07, 4.0606e-07, 1.1642e-10, ..., 7.8231e-08, + 8.3633e-07, 2.3353e-07]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0327, 0.0132, 0.0137, 0.0268, 0.0287, -0.0175, 0.0316, 0.0301, + -0.0139, 0.0027], device='cuda:0'), grad: tensor([ 7.9512e-08, 1.2994e-05, 8.2888e-07, 9.4026e-06, 5.7779e-06, + -3.4031e-06, 1.7378e-06, 1.8347e-06, -3.3587e-05, 4.3325e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 249.04, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.5111 re_mapping 0.0043 re_causal 0.0137 /// teacc 99.05 lr 0.00010000 +Epoch 182, weight, value: tensor([[ 0.0565, -0.1045, -0.1260, ..., -0.1041, -0.1078, -0.0938], + [ 0.0291, 0.0492, -0.0323, ..., -0.0372, 0.1430, 0.0186], + [-0.0100, -0.1420, -0.0098, ..., 0.1509, -0.1251, 0.0297], + ..., + [-0.0771, 0.1134, 0.0267, ..., -0.1011, 0.0915, 0.0451], + [-0.0023, -0.0628, 0.0033, ..., -0.0584, -0.1544, -0.1179], + [ 0.0012, -0.0136, -0.0036, ..., -0.1564, -0.1217, 0.0291]], + device='cuda:0'), grad: tensor([[ 8.5100e-08, 4.0804e-08, 0.0000e+00, ..., 1.3225e-07, + 5.8091e-08, 1.1816e-07], + [ 1.0378e-07, 3.6089e-07, 0.0000e+00, ..., 6.9966e-08, + 2.1188e-08, 9.6299e-07], + [ 8.3703e-08, 2.5090e-06, 0.0000e+00, ..., -3.0664e-07, + 2.7716e-06, 2.9244e-06], + ..., + [ 2.6845e-07, -3.8594e-06, 0.0000e+00, ..., 1.6170e-07, + -4.0941e-06, -4.3958e-06], + [ 3.7230e-07, 2.2852e-07, 0.0000e+00, ..., 3.6438e-08, + 2.9244e-07, 4.3004e-07], + [-2.6897e-06, -6.0885e-08, 0.0000e+00, ..., 4.2026e-08, + 6.1525e-08, -3.3826e-06]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0327, 0.0130, 0.0136, 0.0258, 0.0287, -0.0160, 0.0313, 0.0302, + -0.0131, 0.0024], device='cuda:0'), grad: tensor([ 1.2638e-06, 2.5816e-06, 7.2569e-06, 7.6368e-06, 1.0349e-05, + -1.3717e-05, 1.8198e-06, -8.5905e-06, 4.2170e-06, -1.2815e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 249.25, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.4971 re_mapping 0.0041 re_causal 0.0137 /// teacc 99.01 lr 0.00010000 +Epoch 183, weight, value: tensor([[ 0.0568, -0.1045, -0.1262, ..., -0.1044, -0.1080, -0.0940], + [ 0.0292, 0.0492, -0.0323, ..., -0.0372, 0.1430, 0.0186], + [-0.0101, -0.1423, -0.0094, ..., 0.1517, -0.1257, 0.0297], + ..., + [-0.0773, 0.1135, 0.0257, ..., -0.1018, 0.0915, 0.0451], + [-0.0022, -0.0628, 0.0034, ..., -0.0582, -0.1544, -0.1181], + [ 0.0011, -0.0136, -0.0036, ..., -0.1567, -0.1218, 0.0291]], + device='cuda:0'), grad: tensor([[-4.8196e-07, 4.0221e-08, 0.0000e+00, ..., 2.7125e-07, + 6.4261e-08, 4.0571e-08], + [-4.1490e-07, 4.4256e-06, 0.0000e+00, ..., 7.6310e-08, + 3.5409e-06, 4.8019e-06], + [ 1.3364e-07, 2.2189e-07, 0.0000e+00, ..., 3.3760e-09, + 2.5146e-07, 1.9278e-07], + ..., + [ 2.3108e-07, -7.9423e-06, 0.0000e+00, ..., 3.6380e-08, + -7.7263e-06, -8.3447e-06], + [ 4.6007e-07, 1.7358e-07, 0.0000e+00, ..., 7.7998e-08, + 4.9453e-07, 1.8615e-07], + [-1.7593e-06, 8.7172e-07, 0.0000e+00, ..., 1.1676e-07, + 1.0114e-06, 9.3179e-07]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0325, 0.0130, 0.0137, 0.0259, 0.0288, -0.0166, 0.0307, 0.0302, + -0.0128, 0.0023], device='cuda:0'), grad: tensor([-6.0815e-07, 8.1733e-06, 1.7537e-06, 1.2189e-07, 7.2271e-06, + 5.2452e-06, -6.8592e-07, -1.3202e-05, 3.8296e-06, -1.1876e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 248.96, cls_loss 0.0010 cls_loss_mapping 0.0028 cls_loss_causal 0.4991 re_mapping 0.0042 re_causal 0.0135 /// teacc 99.07 lr 0.00010000 +Epoch 184, weight, value: tensor([[ 0.0566, -0.1058, -0.1262, ..., -0.1053, -0.1093, -0.0943], + [ 0.0292, 0.0492, -0.0323, ..., -0.0374, 0.1431, 0.0186], + [-0.0101, -0.1428, -0.0094, ..., 0.1523, -0.1259, 0.0298], + ..., + [-0.0776, 0.1137, 0.0257, ..., -0.1022, 0.0917, 0.0454], + [-0.0021, -0.0628, 0.0034, ..., -0.0583, -0.1545, -0.1183], + [ 0.0010, -0.0144, -0.0037, ..., -0.1570, -0.1228, 0.0288]], + device='cuda:0'), grad: tensor([[ 4.6915e-07, 8.9058e-09, 0.0000e+00, ..., 2.1176e-07, + 2.6484e-08, 2.1793e-07], + [-1.7730e-07, -1.3050e-07, 0.0000e+00, ..., 3.5681e-08, + -6.4680e-07, -6.8161e-08], + [ 1.5146e-07, 6.7637e-08, 0.0000e+00, ..., 1.7565e-06, + 1.5576e-07, 1.7893e-07], + ..., + [ 1.9139e-07, -9.3773e-08, 0.0000e+00, ..., 2.1595e-08, + 2.7940e-08, -4.7614e-08], + [ 6.1430e-06, 2.3923e-08, 0.0000e+00, ..., 5.6811e-08, + 8.8883e-08, 1.1257e-07], + [-5.2191e-06, 3.1025e-08, 0.0000e+00, ..., 2.2817e-08, + 8.7195e-08, -1.6661e-06]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0333, 0.0130, 0.0137, 0.0259, 0.0288, -0.0166, 0.0311, 0.0303, + -0.0127, 0.0021], device='cuda:0'), grad: tensor([ 3.3155e-06, -3.6554e-07, 1.7688e-05, -1.4812e-05, 5.5917e-06, + -9.8422e-06, 2.3842e-07, 8.1910e-07, 1.7658e-05, -2.0310e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 248.97, cls_loss 0.0012 cls_loss_mapping 0.0034 cls_loss_causal 0.5027 re_mapping 0.0040 re_causal 0.0129 /// teacc 99.15 lr 0.00010000 +Epoch 185, weight, value: tensor([[ 0.0569, -0.1058, -0.1262, ..., -0.1056, -0.1096, -0.0945], + [ 0.0294, 0.0492, -0.0324, ..., -0.0376, 0.1431, 0.0185], + [-0.0103, -0.1442, -0.0094, ..., 0.1533, -0.1263, 0.0294], + ..., + [-0.0780, 0.1140, 0.0256, ..., -0.1024, 0.0917, 0.0456], + [-0.0020, -0.0630, 0.0035, ..., -0.0584, -0.1546, -0.1186], + [ 0.0009, -0.0145, -0.0037, ..., -0.1572, -0.1230, 0.0289]], + device='cuda:0'), grad: tensor([[ 1.2377e-06, 2.5844e-08, 0.0000e+00, ..., 1.3015e-07, + 2.4505e-08, 8.6380e-08], + [ 4.5775e-07, 5.0699e-08, 0.0000e+00, ..., 1.3784e-07, + -2.1362e-07, 1.2224e-07], + [ 3.0268e-07, 5.3726e-08, 5.8208e-11, ..., -5.3458e-07, + 5.8091e-08, -2.4517e-07], + ..., + [ 4.5565e-07, -9.0979e-08, 0.0000e+00, ..., 1.2829e-07, + -1.7846e-07, -5.8208e-11], + [-5.6252e-06, -2.1548e-07, 5.8208e-11, ..., 1.1019e-07, + 7.3924e-08, 1.4785e-07], + [ 9.1875e-07, -2.1141e-07, 0.0000e+00, ..., 1.6810e-07, + 4.2550e-08, -5.1735e-07]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0332, 0.0130, 0.0134, 0.0256, 0.0287, -0.0167, 0.0309, 0.0306, + -0.0126, 0.0021], device='cuda:0'), grad: tensor([ 7.9125e-06, 3.7849e-06, 7.3295e-07, 3.8184e-06, -2.1048e-07, + 3.5446e-06, 4.6603e-06, 3.3733e-06, -3.2037e-05, 4.3325e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 249.38, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.5123 re_mapping 0.0042 re_causal 0.0132 /// teacc 99.13 lr 0.00010000 +Epoch 186, weight, value: tensor([[ 0.0571, -0.1061, -0.1262, ..., -0.1060, -0.1103, -0.0948], + [ 0.0294, 0.0492, -0.0322, ..., -0.0378, 0.1432, 0.0185], + [-0.0102, -0.1446, -0.0094, ..., 0.1540, -0.1267, 0.0293], + ..., + [-0.0782, 0.1141, 0.0254, ..., -0.1030, 0.0918, 0.0457], + [-0.0019, -0.0627, 0.0035, ..., -0.0587, -0.1546, -0.1185], + [ 0.0009, -0.0152, -0.0037, ..., -0.1574, -0.1238, 0.0287]], + device='cuda:0'), grad: tensor([[-1.0757e-06, 7.3342e-09, 3.4925e-09, ..., 6.8976e-08, + 2.3865e-09, 1.0186e-08], + [-2.5262e-08, 1.1059e-09, 1.7288e-08, ..., 1.0279e-07, + -1.8906e-07, 3.0210e-08], + [ 7.1421e-08, 2.3050e-08, 1.0652e-08, ..., 2.9278e-08, + 4.6100e-08, 1.9325e-08], + ..., + [ 8.1200e-08, -6.5484e-08, 2.9395e-08, ..., 1.0419e-08, + 3.1199e-08, -2.6310e-08], + [ 9.4704e-08, 4.5809e-08, 3.2014e-08, ..., 1.8114e-07, + 6.7521e-08, 1.6124e-08], + [ 8.9314e-07, 6.8452e-08, 2.5146e-08, ..., 2.8522e-07, + 7.5379e-08, 8.0210e-08]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0335, 0.0129, 0.0134, 0.0256, 0.0289, -0.0165, 0.0312, 0.0307, + -0.0125, 0.0019], device='cuda:0'), grad: tensor([-6.8806e-06, 1.3402e-06, 1.9632e-06, -8.7470e-06, -2.9933e-06, + 1.9092e-06, 1.1548e-06, 1.6373e-06, 2.7921e-06, 7.8678e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 249.26, cls_loss 0.0009 cls_loss_mapping 0.0027 cls_loss_causal 0.4940 re_mapping 0.0041 re_causal 0.0131 /// teacc 99.04 lr 0.00010000 +Epoch 187, weight, value: tensor([[ 0.0574, -0.1059, -0.1262, ..., -0.1062, -0.1107, -0.0949], + [ 0.0294, 0.0491, -0.0323, ..., -0.0382, 0.1432, 0.0185], + [-0.0103, -0.1449, -0.0094, ..., 0.1542, -0.1269, 0.0293], + ..., + [-0.0785, 0.1142, 0.0254, ..., -0.1034, 0.0919, 0.0458], + [-0.0019, -0.0628, 0.0035, ..., -0.0588, -0.1546, -0.1186], + [ 0.0009, -0.0152, -0.0038, ..., -0.1576, -0.1240, 0.0287]], + device='cuda:0'), grad: tensor([[-6.6299e-08, 4.5286e-08, 0.0000e+00, ..., 8.3062e-08, + 4.3132e-08, 8.0967e-08], + [-1.8976e-08, 3.8208e-07, 0.0000e+00, ..., 3.2713e-08, + 3.4925e-07, 4.8988e-07], + [ 8.9058e-09, 2.7148e-07, 0.0000e+00, ..., -6.0943e-08, + 2.6263e-07, 2.1560e-07], + ..., + [ 2.0664e-08, -2.0619e-06, 0.0000e+00, ..., 4.8720e-08, + -2.0824e-06, -2.5537e-06], + [ 1.8044e-08, 7.5845e-08, 0.0000e+00, ..., 2.6659e-08, + 6.9907e-08, 1.0687e-07], + [-1.2631e-07, 7.6275e-07, 0.0000e+00, ..., 1.3039e-08, + 8.5589e-07, 9.3644e-07]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0334, 0.0128, 0.0133, 0.0256, 0.0289, -0.0165, 0.0317, 0.0307, + -0.0125, 0.0019], device='cuda:0'), grad: tensor([-7.5379e-08, 1.4491e-06, 3.6024e-06, -1.8151e-06, 6.7661e-07, + 4.7823e-07, 7.1013e-09, -6.7353e-06, 4.9360e-07, 1.9055e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 249.38, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.5128 re_mapping 0.0039 re_causal 0.0127 /// teacc 98.95 lr 0.00010000 +Epoch 188, weight, value: tensor([[ 0.0575, -0.1061, -0.1262, ..., -0.1066, -0.1113, -0.0952], + [ 0.0296, 0.0491, -0.0323, ..., -0.0384, 0.1433, 0.0184], + [-0.0104, -0.1455, -0.0094, ..., 0.1553, -0.1273, 0.0295], + ..., + [-0.0789, 0.1144, 0.0254, ..., -0.1039, 0.0920, 0.0459], + [-0.0018, -0.0629, 0.0035, ..., -0.0590, -0.1547, -0.1189], + [ 0.0006, -0.0156, -0.0038, ..., -0.1582, -0.1246, 0.0285]], + device='cuda:0'), grad: tensor([[-1.0617e-07, 6.5425e-08, 0.0000e+00, ..., 1.3562e-08, + 9.7381e-08, 2.8056e-07], + [-3.3602e-06, 1.4284e-07, 0.0000e+00, ..., 3.9232e-08, + -5.2452e-06, -1.7928e-06], + [ 2.6706e-07, 2.6007e-07, 0.0000e+00, ..., -5.4919e-08, + 4.3330e-07, 3.3528e-07], + ..., + [ 1.0571e-06, 2.4438e-06, 0.0000e+00, ..., -3.6875e-08, + 1.3243e-06, 1.1630e-05], + [ 2.5891e-07, 3.6001e-08, 0.0000e+00, ..., 5.1310e-08, + 7.0361e-07, 3.9581e-07], + [ 2.1840e-07, -3.3714e-06, 0.0000e+00, ..., 2.8027e-08, + 8.9698e-08, -1.3188e-05]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0336, 0.0128, 0.0134, 0.0258, 0.0290, -0.0166, 0.0314, 0.0307, + -0.0124, 0.0016], device='cuda:0'), grad: tensor([-2.8522e-07, -1.0900e-05, 1.7062e-06, 5.8301e-07, 4.4368e-06, + 3.9227e-06, 1.0645e-06, 3.6448e-05, 6.0489e-07, -3.7581e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 249.14, cls_loss 0.0013 cls_loss_mapping 0.0028 cls_loss_causal 0.4948 re_mapping 0.0042 re_causal 0.0133 /// teacc 99.07 lr 0.00010000 +Epoch 189, weight, value: tensor([[ 0.0574, -0.1064, -0.1262, ..., -0.1068, -0.1121, -0.0957], + [ 0.0296, 0.0491, -0.0323, ..., -0.0387, 0.1433, 0.0184], + [-0.0105, -0.1458, -0.0094, ..., 0.1559, -0.1277, 0.0295], + ..., + [-0.0796, 0.1145, 0.0254, ..., -0.1042, 0.0921, 0.0460], + [-0.0020, -0.0630, 0.0035, ..., -0.0596, -0.1548, -0.1189], + [ 0.0014, -0.0157, -0.0038, ..., -0.1584, -0.1248, 0.0289]], + device='cuda:0'), grad: tensor([[-1.9395e-07, 2.3283e-09, 0.0000e+00, ..., 1.2427e-08, + 1.3330e-08, 1.8452e-08], + [-1.0955e-07, -1.1135e-07, 0.0000e+00, ..., 1.4756e-08, + -3.9930e-07, -9.0920e-08], + [ 1.0786e-07, 1.4610e-08, 0.0000e+00, ..., -1.5867e-07, + 5.1892e-08, -8.3004e-08], + ..., + [ 2.8568e-07, 3.3615e-08, 0.0000e+00, ..., 1.1729e-07, + 1.1048e-07, 1.2119e-07], + [ 1.3085e-07, 1.6327e-08, 0.0000e+00, ..., 1.3388e-08, + 5.3667e-08, 3.7544e-08], + [ 1.0256e-07, 2.4447e-09, 0.0000e+00, ..., 7.6834e-09, + 2.3370e-08, -9.1502e-08]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0337, 0.0127, 0.0133, 0.0252, 0.0289, -0.0165, 0.0325, 0.0307, + -0.0128, 0.0019], device='cuda:0'), grad: tensor([-1.0291e-06, 9.5461e-08, 6.9942e-07, -9.1270e-06, 2.2189e-07, + 2.2259e-06, 1.0198e-06, 2.9001e-06, 1.7453e-06, 1.2172e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 249.05, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.5275 re_mapping 0.0038 re_causal 0.0127 /// teacc 99.07 lr 0.00010000 +Epoch 190, weight, value: tensor([[ 0.0574, -0.1066, -0.1262, ..., -0.1070, -0.1127, -0.0963], + [ 0.0296, 0.0491, -0.0324, ..., -0.0398, 0.1433, 0.0182], + [-0.0108, -0.1462, -0.0091, ..., 0.1588, -0.1282, 0.0301], + ..., + [-0.0803, 0.1146, 0.0243, ..., -0.1047, 0.0922, 0.0461], + [-0.0018, -0.0630, 0.0035, ..., -0.0595, -0.1548, -0.1192], + [ 0.0014, -0.0157, -0.0038, ..., -0.1586, -0.1249, 0.0290]], + device='cuda:0'), grad: tensor([[ 3.4343e-09, 1.6880e-09, 0.0000e+00, ..., 6.8103e-09, + 4.2783e-09, 1.7084e-08], + [-1.0373e-07, -4.2957e-08, 0.0000e+00, ..., 2.3370e-08, + -3.7462e-07, 5.6985e-08], + [ 5.0495e-08, 2.0635e-08, 0.0000e+00, ..., -4.6741e-08, + 4.8167e-08, -3.3062e-08], + ..., + [ 1.9860e-07, -5.4861e-08, 0.0000e+00, ..., 3.2829e-08, + 4.7236e-08, 6.0070e-08], + [ 2.0897e-07, 3.7078e-08, 0.0000e+00, ..., 9.0222e-09, + 1.9651e-07, 8.0443e-08], + [-2.4065e-06, 1.5309e-08, 0.0000e+00, ..., 2.5990e-08, + 1.7695e-08, -5.1111e-06]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0339, 0.0124, 0.0137, 0.0263, 0.0290, -0.0173, 0.0320, 0.0308, + -0.0126, 0.0018], device='cuda:0'), grad: tensor([ 6.4843e-08, 3.1362e-07, 1.2724e-07, 1.8878e-06, 1.7703e-05, + -4.1723e-06, 7.5903e-07, 9.5740e-07, 3.2177e-07, -1.7941e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 249.19, cls_loss 0.0011 cls_loss_mapping 0.0029 cls_loss_causal 0.4992 re_mapping 0.0041 re_causal 0.0128 /// teacc 99.00 lr 0.00010000 +Epoch 191, weight, value: tensor([[ 0.0572, -0.1062, -0.1262, ..., -0.1064, -0.1132, -0.0964], + [ 0.0299, 0.0492, -0.0324, ..., -0.0399, 0.1435, 0.0183], + [-0.0109, -0.1467, -0.0090, ..., 0.1591, -0.1286, 0.0303], + ..., + [-0.0810, 0.1147, 0.0240, ..., -0.1055, 0.0922, 0.0462], + [-0.0020, -0.0633, 0.0035, ..., -0.0598, -0.1551, -0.1197], + [ 0.0015, -0.0159, -0.0039, ..., -0.1590, -0.1252, 0.0289]], + device='cuda:0'), grad: tensor([[ 9.6217e-08, 2.8842e-08, 0.0000e+00, ..., 1.0774e-07, + 1.2899e-07, 3.3993e-08], + [-2.5854e-05, -4.4182e-06, 0.0000e+00, ..., -2.1338e-05, + -3.0160e-05, -9.0292e-07], + [ 9.4809e-07, 3.2363e-07, 0.0000e+00, ..., 8.5589e-07, + 1.2163e-06, 3.1944e-07], + ..., + [ 8.8010e-07, -4.8568e-07, 0.0000e+00, ..., 7.0641e-07, + 2.6915e-07, -9.8906e-07], + [ 6.5975e-06, 1.4110e-06, 0.0000e+00, ..., 5.5507e-06, + 8.8662e-06, 4.4517e-07], + [ 4.4121e-08, 2.8918e-07, 0.0000e+00, ..., 2.5658e-07, + 4.5239e-07, -4.4587e-08]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0343, 0.0126, 0.0137, 0.0263, 0.0292, -0.0168, 0.0319, 0.0307, + -0.0129, 0.0017], device='cuda:0'), grad: tensor([ 6.6264e-07, -1.1969e-04, 5.4464e-06, -2.1774e-06, 2.6956e-05, + 7.3016e-06, 5.0426e-05, 3.5055e-06, 2.7582e-05, -8.7894e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 249.20, cls_loss 0.0012 cls_loss_mapping 0.0030 cls_loss_causal 0.4861 re_mapping 0.0041 re_causal 0.0124 /// teacc 98.98 lr 0.00010000 +Epoch 192, weight, value: tensor([[ 0.0574, -0.1063, -0.1263, ..., -0.1067, -0.1137, -0.0966], + [ 0.0300, 0.0492, -0.0332, ..., -0.0402, 0.1436, 0.0182], + [-0.0109, -0.1470, -0.0080, ..., 0.1594, -0.1287, 0.0308], + ..., + [-0.0815, 0.1149, 0.0229, ..., -0.1066, 0.0923, 0.0463], + [-0.0019, -0.0631, 0.0032, ..., -0.0598, -0.1553, -0.1200], + [ 0.0019, -0.0163, -0.0040, ..., -0.1593, -0.1257, 0.0287]], + device='cuda:0'), grad: tensor([[-9.0222e-10, 9.1095e-09, 1.1642e-09, ..., 5.5553e-07, + 1.2311e-08, 2.9686e-08], + [-1.0961e-07, 3.7922e-08, 8.1491e-10, ..., 1.1269e-07, + -1.5122e-07, 5.1502e-07], + [ 3.2567e-08, 3.7020e-08, 1.5716e-09, ..., 2.1304e-08, + 4.7236e-08, 1.3970e-09], + ..., + [ 8.8883e-08, -3.0780e-07, 1.7462e-10, ..., 2.1595e-08, + -2.0571e-07, -5.9837e-08], + [-9.7498e-08, 4.8341e-08, -1.0245e-08, ..., 3.6764e-07, + 9.2201e-08, 9.5111e-08], + [-2.5553e-08, 1.0506e-08, 5.3551e-09, ..., 4.6857e-08, + 2.7765e-08, 6.4401e-07]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0342, 0.0125, 0.0129, 0.0271, 0.0294, -0.0170, 0.0316, 0.0306, + -0.0129, 0.0017], device='cuda:0'), grad: tensor([ 1.2834e-06, 2.5369e-06, 2.8545e-07, -7.2271e-07, -7.1526e-06, + 2.3935e-06, -4.3064e-06, 4.1840e-07, 1.1707e-06, 4.0941e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 248.97, cls_loss 0.0015 cls_loss_mapping 0.0040 cls_loss_causal 0.5267 re_mapping 0.0041 re_causal 0.0127 /// teacc 98.97 lr 0.00010000 +Epoch 193, weight, value: tensor([[ 0.0575, -0.1076, -0.1263, ..., -0.1070, -0.1152, -0.0971], + [ 0.0296, 0.0492, -0.0332, ..., -0.0410, 0.1438, 0.0179], + [-0.0106, -0.1478, -0.0079, ..., 0.1607, -0.1293, 0.0315], + ..., + [-0.0822, 0.1151, 0.0229, ..., -0.1074, 0.0923, 0.0464], + [-0.0016, -0.0629, 0.0032, ..., -0.0599, -0.1554, -0.1204], + [ 0.0021, -0.0166, -0.0041, ..., -0.1601, -0.1258, 0.0285]], + device='cuda:0'), grad: tensor([[-2.9802e-07, -2.7241e-08, 0.0000e+00, ..., 2.5588e-07, + 1.1787e-08, 3.5157e-08], + [-6.1875e-08, 1.4866e-07, 0.0000e+00, ..., 5.1176e-07, + 1.5693e-07, 6.7987e-07], + [ 3.7544e-09, 1.7724e-08, 0.0000e+00, ..., -6.7987e-07, + -1.3586e-07, -8.3493e-07], + ..., + [ 7.3400e-08, -3.1218e-06, 0.0000e+00, ..., 2.4308e-07, + -3.6191e-06, -1.6112e-06], + [ 2.6455e-08, 6.6240e-08, 0.0000e+00, ..., 1.2876e-07, + 1.0623e-07, 1.5204e-07], + [ 7.2818e-08, 1.3399e-07, 0.0000e+00, ..., 2.5681e-07, + 1.6449e-07, 3.2480e-07]], device='cuda:0') +Epoch 193, bias, value: tensor([-0.0342, 0.0120, 0.0134, 0.0275, 0.0300, -0.0171, 0.0310, 0.0305, + -0.0124, 0.0013], device='cuda:0'), grad: tensor([-5.7416e-07, 2.2519e-06, -1.6782e-06, 6.6906e-06, -3.1348e-06, + 6.0629e-07, 6.2457e-08, -6.5640e-06, 6.6962e-07, 1.6605e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 249.00, cls_loss 0.0012 cls_loss_mapping 0.0027 cls_loss_causal 0.5036 re_mapping 0.0046 re_causal 0.0139 /// teacc 99.08 lr 0.00010000 +Epoch 194, weight, value: tensor([[ 0.0577, -0.1084, -0.1264, ..., -0.1073, -0.1162, -0.0977], + [ 0.0312, 0.0510, -0.0350, ..., -0.0404, 0.1452, 0.0196], + [-0.0124, -0.1511, -0.0069, ..., 0.1607, -0.1316, 0.0307], + ..., + [-0.0829, 0.1151, 0.0223, ..., -0.1094, 0.0923, 0.0462], + [-0.0016, -0.0630, 0.0032, ..., -0.0602, -0.1555, -0.1210], + [ 0.0019, -0.0169, -0.0041, ..., -0.1608, -0.1263, 0.0281]], + device='cuda:0'), grad: tensor([[ 4.1258e-07, -6.7812e-09, 0.0000e+00, ..., 1.0198e-06, + 3.8836e-07, 6.5193e-07], + [-4.9733e-06, 1.0002e-06, 0.0000e+00, ..., 2.3842e-07, + -5.6177e-06, 1.8543e-06], + [-7.5810e-07, 2.1624e-08, 0.0000e+00, ..., -6.6273e-06, + 5.9418e-07, -3.9078e-06], + ..., + [ 1.4352e-06, 3.1386e-07, 0.0000e+00, ..., 9.0804e-07, + 1.3001e-06, 9.8255e-07], + [ 2.7716e-06, 2.1042e-08, 0.0000e+00, ..., 1.3290e-06, + 1.9111e-06, 7.9069e-07], + [ 3.0492e-06, -1.6056e-06, 0.0000e+00, ..., 6.9523e-07, + -8.2655e-07, -2.7865e-06]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0341, 0.0143, 0.0116, 0.0274, 0.0310, -0.0172, 0.0296, 0.0301, + -0.0126, 0.0010], device='cuda:0'), grad: tensor([ 2.3916e-06, -7.7561e-06, -1.4946e-05, 2.4270e-06, 5.3793e-06, + -1.7390e-05, 5.8264e-06, 5.7146e-06, 1.0043e-05, 8.2999e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 248.64, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.5408 re_mapping 0.0039 re_causal 0.0130 /// teacc 99.05 lr 0.00010000 +Epoch 195, weight, value: tensor([[ 0.0577, -0.1082, -0.1267, ..., -0.1077, -0.1162, -0.0984], + [ 0.0312, 0.0510, -0.0346, ..., -0.0406, 0.1452, 0.0195], + [-0.0124, -0.1512, -0.0065, ..., 0.1615, -0.1317, 0.0309], + ..., + [-0.0836, 0.1154, 0.0205, ..., -0.1113, 0.0925, 0.0465], + [-0.0016, -0.0628, 0.0031, ..., -0.0607, -0.1556, -0.1214], + [ 0.0025, -0.0170, -0.0038, ..., -0.1610, -0.1265, 0.0283]], + device='cuda:0'), grad: tensor([[-5.6374e-08, 4.5955e-08, 1.9632e-06, ..., 6.3628e-06, + 3.6700e-08, 7.7009e-08], + [-2.7567e-07, 5.2154e-07, 8.2015e-08, ..., 3.3458e-07, + 7.0024e-08, 8.7591e-07], + [ 6.0012e-08, 3.5646e-07, 1.4883e-06, ..., 4.4219e-06, + 4.0117e-07, 2.1397e-07], + ..., + [ 2.3213e-07, -1.7919e-06, 1.6007e-08, ..., 7.2760e-08, + -2.1085e-06, -2.0694e-06], + [ 6.5612e-07, 3.4599e-07, 5.2201e-07, ..., 1.7332e-06, + 6.0955e-07, 5.5227e-07], + [-3.6438e-07, 5.8877e-08, 4.8574e-08, ..., 2.0722e-07, + 4.9314e-07, -4.4447e-07]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0342, 0.0142, 0.0117, 0.0272, 0.0306, -0.0172, 0.0297, 0.0302, + -0.0125, 0.0013], device='cuda:0'), grad: tensor([ 2.1592e-05, 1.8412e-06, 1.6972e-05, -4.2868e-04, 9.2946e-07, + 2.2147e-06, 3.7980e-04, -3.4068e-06, 9.3579e-06, -1.1781e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 249.27, cls_loss 0.0011 cls_loss_mapping 0.0024 cls_loss_causal 0.5213 re_mapping 0.0040 re_causal 0.0128 /// teacc 99.03 lr 0.00010000 +Epoch 196, weight, value: tensor([[ 0.0575, -0.1080, -0.1277, ..., -0.1080, -0.1162, -0.0990], + [ 0.0313, 0.0509, -0.0351, ..., -0.0408, 0.1453, 0.0195], + [-0.0124, -0.1513, -0.0045, ..., 0.1621, -0.1318, 0.0310], + ..., + [-0.0840, 0.1155, 0.0200, ..., -0.1123, 0.0926, 0.0465], + [-0.0015, -0.0628, 0.0025, ..., -0.0612, -0.1557, -0.1218], + [ 0.0027, -0.0169, -0.0042, ..., -0.1614, -0.1268, 0.0283]], + device='cuda:0'), grad: tensor([[ 9.0396e-08, 8.3237e-09, 0.0000e+00, ..., 5.9605e-08, + 6.6590e-08, 6.6881e-08], + [-3.4645e-07, -1.1781e-07, 2.9104e-11, ..., 2.1642e-07, + -1.0040e-06, 9.4820e-08], + [-1.1252e-07, 3.6845e-08, 5.8208e-11, ..., -1.0887e-06, + 4.2981e-07, -9.5833e-07], + ..., + [ 3.4529e-07, -1.7637e-08, 2.9104e-11, ..., 7.2131e-07, + 1.4715e-07, 8.8196e-07], + [ 2.7963e-07, 1.9354e-08, 8.7311e-11, ..., 1.9954e-07, + 7.3982e-08, 1.8172e-07], + [ 7.8697e-08, 2.2090e-08, 0.0000e+00, ..., 2.4326e-06, + 4.9011e-08, 1.5525e-06]], device='cuda:0') +Epoch 196, bias, value: tensor([-0.0346, 0.0142, 0.0118, 0.0277, 0.0307, -0.0173, 0.0294, 0.0301, + -0.0126, 0.0013], device='cuda:0'), grad: tensor([ 6.2631e-07, -4.5518e-07, -1.5907e-06, 2.1365e-06, -8.8960e-06, + -3.8333e-06, -4.4215e-07, 2.7269e-06, 1.7677e-06, 7.9498e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 249.13, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.4948 re_mapping 0.0039 re_causal 0.0123 /// teacc 99.09 lr 0.00010000 +Epoch 197, weight, value: tensor([[ 0.0576, -0.1083, -0.1279, ..., -0.1083, -0.1168, -0.0999], + [ 0.0313, 0.0509, -0.0369, ..., -0.0410, 0.1453, 0.0195], + [-0.0125, -0.1513, -0.0038, ..., 0.1628, -0.1319, 0.0311], + ..., + [-0.0845, 0.1153, 0.0198, ..., -0.1136, 0.0926, 0.0463], + [-0.0013, -0.0616, 0.0025, ..., -0.0601, -0.1556, -0.1212], + [ 0.0029, -0.0164, -0.0043, ..., -0.1619, -0.1268, 0.0287]], + device='cuda:0'), grad: tensor([[-2.2090e-08, 3.0338e-07, 0.0000e+00, ..., 2.5466e-08, + 4.6869e-07, 1.9954e-07], + [-8.5384e-06, -1.6853e-05, 0.0000e+00, ..., 4.3452e-08, + -2.2516e-05, -8.3521e-06], + [ 6.5076e-08, 3.9348e-07, 0.0000e+00, ..., -6.8801e-08, + 4.3330e-07, 5.4715e-07], + ..., + [ 4.0568e-06, 6.8769e-06, 0.0000e+00, ..., 2.3603e-08, + 9.5740e-06, 2.3972e-06], + [ 1.0856e-07, 1.0300e-06, 0.0000e+00, ..., 5.8062e-08, + 1.3756e-06, 6.3889e-07], + [ 2.9281e-06, 5.7369e-06, 0.0000e+00, ..., 1.3708e-08, + 7.2904e-06, 3.1702e-06]], device='cuda:0') +Epoch 197, bias, value: tensor([-0.0346, 0.0142, 0.0118, 0.0277, 0.0307, -0.0189, 0.0300, 0.0296, + -0.0117, 0.0014], device='cuda:0'), grad: tensor([ 4.4500e-08, -4.2230e-05, 1.4389e-06, 9.3644e-07, 5.5954e-06, + 5.3365e-07, 1.3126e-08, 1.6749e-05, 7.7533e-07, 1.6183e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 196, time 248.72, cls_loss 0.0014 cls_loss_mapping 0.0036 cls_loss_causal 0.5097 re_mapping 0.0038 re_causal 0.0122 /// teacc 98.95 lr 0.00010000 +Epoch 198, weight, value: tensor([[ 0.0574, -0.1090, -0.1280, ..., -0.1089, -0.1185, -0.1008], + [ 0.0311, 0.0505, -0.0372, ..., -0.0413, 0.1451, 0.0191], + [-0.0125, -0.1513, -0.0037, ..., 0.1632, -0.1320, 0.0311], + ..., + [-0.0830, 0.1164, 0.0197, ..., -0.1161, 0.0934, 0.0469], + [-0.0012, -0.0616, 0.0023, ..., -0.0603, -0.1558, -0.1217], + [ 0.0034, -0.0186, -0.0046, ..., -0.1640, -0.1289, 0.0270]], + device='cuda:0'), grad: tensor([[-3.8999e-09, 3.0734e-08, 0.0000e+00, ..., 1.2195e-08, + 3.9872e-08, 7.4389e-08], + [ 2.3458e-08, 1.4505e-07, 0.0000e+00, ..., 3.9086e-08, + 1.5588e-07, 2.7893e-07], + [ 1.0041e-08, 3.2084e-07, 0.0000e+00, ..., -2.6566e-07, + 4.2329e-07, 5.3691e-07], + ..., + [ 3.7689e-08, -7.3155e-07, 0.0000e+00, ..., 5.2707e-08, + -9.0804e-07, -1.3458e-06], + [-9.2259e-08, 6.4843e-08, 0.0000e+00, ..., 4.9971e-08, + 7.7242e-08, 1.5774e-07], + [-2.2672e-08, 7.0548e-08, 0.0000e+00, ..., 7.1537e-08, + 9.0513e-08, 1.8300e-07]], device='cuda:0') +Epoch 198, bias, value: tensor([-0.0350, 0.0139, 0.0118, 0.0272, 0.0319, -0.0191, 0.0303, 0.0303, + -0.0116, 0.0007], device='cuda:0'), grad: tensor([ 1.0128e-07, 6.9430e-07, 7.0687e-07, 8.6613e-07, -8.9465e-08, + -1.0561e-06, 7.7160e-07, -2.3898e-06, 1.3586e-07, 2.8033e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 249.06, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5217 re_mapping 0.0039 re_causal 0.0127 /// teacc 99.02 lr 0.00010000 +Epoch 199, weight, value: tensor([[ 0.0566, -0.1104, -0.1282, ..., -0.1066, -0.1193, -0.1036], + [ 0.0312, 0.0505, -0.0374, ..., -0.0414, 0.1452, 0.0190], + [-0.0129, -0.1514, -0.0038, ..., 0.1632, -0.1321, 0.0311], + ..., + [-0.0834, 0.1166, 0.0197, ..., -0.1164, 0.0935, 0.0470], + [-0.0008, -0.0617, 0.0031, ..., -0.0610, -0.1560, -0.1218], + [ 0.0039, -0.0187, -0.0063, ..., -0.1646, -0.1292, 0.0271]], + device='cuda:0'), grad: tensor([[-4.6529e-06, 2.2119e-09, 6.0536e-09, ..., 4.1036e-09, + 2.9395e-09, 4.1618e-09], + [ 1.0739e-07, -5.2387e-08, 2.3574e-07, ..., 7.5903e-08, + -1.6484e-07, -2.3778e-08], + [ 1.2445e-07, 4.8603e-09, 5.1176e-07, ..., 1.3318e-07, + 1.0565e-08, -3.7835e-10], + ..., + [ 1.8731e-07, -1.0041e-08, 2.2002e-08, ..., 1.1962e-08, + 2.8231e-09, 4.8894e-08], + [ 7.4273e-07, 4.3889e-08, -2.5388e-06, ..., -6.8359e-07, + 9.7905e-08, 4.0745e-08], + [ 7.4180e-07, -2.4971e-08, 1.2084e-07, ..., 4.2404e-08, + 1.4406e-08, -8.0036e-08]], device='cuda:0') +Epoch 199, bias, value: tensor([-0.0347, 0.0139, 0.0118, 0.0268, 0.0321, -0.0192, 0.0294, 0.0305, + -0.0112, 0.0010], device='cuda:0'), grad: tensor([-1.4678e-05, 1.2033e-06, 1.8803e-06, 1.4432e-05, 7.2643e-08, + -5.4874e-06, 3.4776e-06, 7.3668e-07, -4.3027e-06, 2.6710e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 248.92, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.4582 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.05 lr 0.00010000 +Epoch 200, weight, value: tensor([[ 0.0568, -0.1105, -0.1283, ..., -0.1065, -0.1196, -0.1039], + [ 0.0312, 0.0504, -0.0379, ..., -0.0416, 0.1452, 0.0190], + [-0.0129, -0.1514, -0.0040, ..., 0.1634, -0.1321, 0.0312], + ..., + [-0.0839, 0.1169, 0.0194, ..., -0.1164, 0.0935, 0.0471], + [-0.0006, -0.0623, 0.0040, ..., -0.0613, -0.1562, -0.1229], + [ 0.0047, -0.0186, -0.0072, ..., -0.1645, -0.1291, 0.0290]], + device='cuda:0'), grad: tensor([[-6.5775e-08, 1.0617e-07, 6.9849e-10, ..., -3.4668e-07, + 2.6100e-07, 6.1176e-08], + [-3.5763e-06, -4.2468e-06, 3.2014e-10, ..., 2.2433e-07, + -1.1273e-05, -1.9893e-06], + [ 1.1642e-07, 1.0681e-07, 7.8580e-10, ..., 8.4634e-08, + 2.7288e-07, 4.2317e-08], + ..., + [ 1.8319e-06, 2.0415e-06, 1.2806e-09, ..., 5.1514e-08, + 5.6252e-06, 1.0394e-06], + [ 1.1869e-07, 1.1636e-07, 1.1350e-09, ..., 9.8487e-08, + 3.1409e-07, 9.8313e-08], + [ 8.7079e-07, 1.0990e-06, -5.1223e-09, ..., 3.5320e-07, + 2.8238e-06, 7.3109e-07]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0345, 0.0139, 0.0117, 0.0266, 0.0307, -0.0193, 0.0295, 0.0305, + -0.0112, 0.0024], device='cuda:0'), grad: tensor([-1.2508e-06, -2.4691e-05, 1.0058e-06, 2.4401e-06, -4.2189e-07, + 7.7952e-07, 1.0990e-06, 1.2852e-05, 1.0207e-06, 7.1600e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 248.57, cls_loss 0.0011 cls_loss_mapping 0.0034 cls_loss_causal 0.5111 re_mapping 0.0038 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 201, weight, value: tensor([[ 0.0576, -0.1115, -0.1284, ..., -0.1066, -0.1202, -0.1043], + [ 0.0313, 0.0505, -0.0384, ..., -0.0421, 0.1453, 0.0189], + [-0.0131, -0.1514, -0.0033, ..., 0.1642, -0.1321, 0.0314], + ..., + [-0.0848, 0.1169, 0.0172, ..., -0.1172, 0.0935, 0.0469], + [-0.0006, -0.0625, 0.0038, ..., -0.0615, -0.1565, -0.1234], + [ 0.0049, -0.0188, -0.0076, ..., -0.1651, -0.1295, 0.0292]], + device='cuda:0'), grad: tensor([[ 8.0327e-07, 4.8894e-09, 7.2469e-09, ..., 9.4529e-07, + 8.4401e-09, 2.2969e-07], + [ 1.0962e-06, 1.6851e-08, 8.6438e-09, ..., 3.2736e-07, + -3.6624e-07, 2.9802e-07], + [ 1.4063e-07, 1.7870e-08, 1.2573e-08, ..., -1.6391e-06, + 2.8493e-08, -1.3392e-06], + ..., + [ 3.2526e-07, -2.0617e-07, 7.6834e-09, ..., 2.6263e-07, + -2.0838e-07, 6.4541e-07], + [-7.1079e-06, 4.4849e-08, 3.3033e-08, ..., 8.1584e-07, + 1.7032e-07, 7.3062e-07], + [-2.2200e-07, 4.4616e-08, 6.5309e-08, ..., 3.0897e-07, + 6.5018e-08, -2.0000e-07]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0344, 0.0139, 0.0118, 0.0267, 0.0309, -0.0193, 0.0295, 0.0302, + -0.0114, 0.0024], device='cuda:0'), grad: tensor([ 8.9779e-06, 1.0654e-05, -9.9279e-07, 1.6361e-05, -4.1835e-06, + 7.1041e-06, -9.1782e-07, 5.1744e-06, -4.8041e-05, 5.9158e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 248.95, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.5125 re_mapping 0.0038 re_causal 0.0124 /// teacc 99.11 lr 0.00010000 +Epoch 202, weight, value: tensor([[ 0.0583, -0.1118, -0.1317, ..., -0.1080, -0.1214, -0.1046], + [ 0.0306, 0.0504, -0.0393, ..., -0.0440, 0.1451, 0.0186], + [-0.0120, -0.1515, -0.0033, ..., 0.1665, -0.1319, 0.0318], + ..., + [-0.0838, 0.1173, 0.0181, ..., -0.1171, 0.0940, 0.0471], + [-0.0007, -0.0627, 0.0036, ..., -0.0621, -0.1568, -0.1239], + [ 0.0050, -0.0190, -0.0074, ..., -0.1657, -0.1299, 0.0293]], + device='cuda:0'), grad: tensor([[ 7.8522e-08, 3.9581e-08, 0.0000e+00, ..., 3.7486e-07, + 6.1409e-08, 1.0937e-07], + [-5.2532e-08, 2.3432e-06, 2.9104e-11, ..., 1.2747e-07, + 2.9411e-06, 3.0193e-06], + [-5.0990e-07, 9.8720e-07, 5.8208e-11, ..., -3.4813e-06, + 1.3625e-06, 2.2759e-07], + ..., + [ 5.0379e-08, -4.6529e-06, 2.9104e-11, ..., 6.8045e-08, + -6.2212e-06, -5.8636e-06], + [ 6.5938e-07, 1.3935e-07, 1.4552e-10, ..., 3.2112e-06, + 2.2480e-07, 1.2061e-06], + [-5.3085e-07, 1.0319e-06, 2.9104e-11, ..., -1.6676e-08, + 1.3867e-06, 7.2876e-07]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0364, 0.0134, 0.0123, 0.0263, 0.0308, -0.0195, 0.0311, 0.0308, + -0.0116, 0.0023], device='cuda:0'), grad: tensor([ 1.2573e-06, 6.5491e-06, -4.3958e-06, 1.1381e-06, 2.1756e-06, + -3.5968e-06, 1.6568e-06, -1.2405e-05, 8.1360e-06, -4.6613e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 249.08, cls_loss 0.0011 cls_loss_mapping 0.0027 cls_loss_causal 0.5059 re_mapping 0.0038 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 203, weight, value: tensor([[ 0.0584, -0.1121, -0.1317, ..., -0.1092, -0.1236, -0.1049], + [ 0.0309, 0.0504, -0.0396, ..., -0.0441, 0.1453, 0.0185], + [-0.0120, -0.1517, -0.0029, ..., 0.1665, -0.1321, 0.0317], + ..., + [-0.0842, 0.1176, 0.0182, ..., -0.1154, 0.0942, 0.0475], + [-0.0008, -0.0630, 0.0034, ..., -0.0624, -0.1573, -0.1245], + [ 0.0050, -0.0193, -0.0088, ..., -0.1662, -0.1304, 0.0294]], + device='cuda:0'), grad: tensor([[-7.2760e-10, 2.3574e-09, 2.3865e-09, ..., 1.7113e-07, + 1.6415e-08, 1.0565e-08], + [-6.3679e-08, -9.4005e-09, 2.0082e-09, ..., -6.8452e-07, + -1.2636e-05, -5.1521e-06], + [ 1.1118e-08, 1.5978e-08, 6.9849e-10, ..., 6.9663e-07, + 2.7660e-06, 1.1567e-06], + ..., + [ 2.1653e-08, -4.7701e-08, 3.2014e-10, ..., 6.5798e-07, + 8.9407e-06, 3.7253e-06], + [ 1.0681e-08, 1.0710e-08, 5.9081e-09, ..., 3.9372e-07, + 4.4762e-08, 1.1094e-07], + [ 3.2887e-09, 9.7207e-09, 1.5134e-09, ..., 2.0326e-07, + 4.1706e-08, 2.6985e-07]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0366, 0.0134, 0.0122, 0.0267, 0.0308, -0.0193, 0.0310, 0.0312, + -0.0118, 0.0023], device='cuda:0'), grad: tensor([ 8.2189e-07, -1.3977e-05, 4.3176e-06, 6.9477e-07, 1.9581e-07, + 2.0057e-05, -2.5421e-05, 1.0453e-05, 1.8971e-06, 9.7975e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 249.14, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.4957 re_mapping 0.0039 re_causal 0.0124 /// teacc 99.07 lr 0.00010000 +Epoch 204, weight, value: tensor([[ 0.0586, -0.1125, -0.1317, ..., -0.1094, -0.1248, -0.1053], + [ 0.0312, 0.0505, -0.0398, ..., -0.0442, 0.1456, 0.0185], + [-0.0121, -0.1518, -0.0025, ..., 0.1670, -0.1323, 0.0316], + ..., + [-0.0855, 0.1178, 0.0179, ..., -0.1161, 0.0942, 0.0476], + [-0.0011, -0.0632, 0.0034, ..., -0.0634, -0.1575, -0.1253], + [ 0.0050, -0.0196, -0.0087, ..., -0.1666, -0.1307, 0.0292]], + device='cuda:0'), grad: tensor([[ 8.4168e-08, 3.8155e-08, 8.4110e-09, ..., 8.6101e-07, + 1.3201e-07, 4.7102e-07], + [-4.8093e-06, -1.8068e-06, 3.7835e-09, ..., 1.2033e-06, + -8.4490e-06, -2.7176e-06], + [ 2.9523e-07, 1.7474e-07, 1.4261e-09, ..., 2.7120e-06, + 5.7183e-07, 1.4016e-06], + ..., + [ 1.4873e-06, 2.6729e-07, 5.0059e-09, ..., 1.1595e-06, + 2.1141e-06, 2.0321e-06], + [ 5.8394e-07, 2.9430e-07, 1.0128e-08, ..., 1.2349e-06, + 1.0533e-06, 1.2722e-06], + [ 1.2135e-06, 1.0226e-06, -1.6799e-07, ..., 1.2696e-05, + 2.7660e-06, 7.3537e-06]], device='cuda:0') +Epoch 204, bias, value: tensor([-0.0361, 0.0135, 0.0122, 0.0269, 0.0308, -0.0190, 0.0308, 0.0311, + -0.0123, 0.0021], device='cuda:0'), grad: tensor([ 2.8517e-06, -1.3418e-05, 8.7768e-06, 1.9008e-06, -6.7949e-05, + 2.9728e-06, 9.8199e-06, 9.2909e-06, 6.1281e-06, 3.9607e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 248.94, cls_loss 0.0010 cls_loss_mapping 0.0025 cls_loss_causal 0.5059 re_mapping 0.0037 re_causal 0.0119 /// teacc 99.00 lr 0.00010000 +Epoch 205, weight, value: tensor([[ 0.0582, -0.1127, -0.1318, ..., -0.1095, -0.1258, -0.1076], + [ 0.0316, 0.0506, -0.0368, ..., -0.0442, 0.1459, 0.0186], + [-0.0121, -0.1519, -0.0027, ..., 0.1673, -0.1325, 0.0316], + ..., + [-0.0865, 0.1179, 0.0148, ..., -0.1166, 0.0942, 0.0478], + [-0.0011, -0.0634, 0.0036, ..., -0.0638, -0.1579, -0.1259], + [ 0.0053, -0.0199, -0.0092, ..., -0.1670, -0.1312, 0.0294]], + device='cuda:0'), grad: tensor([[ 1.2701e-07, 1.0739e-08, 0.0000e+00, ..., 4.8475e-07, + 5.7626e-09, 1.9220e-07], + [ 6.9267e-09, 3.6904e-08, 0.0000e+00, ..., 4.7171e-07, + 1.4610e-08, 4.2468e-07], + [ 8.2946e-09, 5.0059e-08, 0.0000e+00, ..., 9.6625e-08, + 7.2760e-09, 1.0477e-07], + ..., + [ 1.2480e-07, 1.0169e-07, 0.0000e+00, ..., 5.1782e-07, + 2.3586e-07, 8.3493e-07], + [ 1.0506e-07, 1.4639e-08, 0.0000e+00, ..., 2.1840e-07, + 1.2282e-08, 2.2585e-07], + [-4.1886e-07, 6.9267e-09, 0.0000e+00, ..., 4.9826e-07, + -2.2887e-07, -3.7835e-07]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0368, 0.0136, 0.0121, 0.0271, 0.0308, -0.0192, 0.0306, 0.0311, + -0.0123, 0.0023], device='cuda:0'), grad: tensor([ 2.1253e-06, 1.5432e-06, 6.1514e-07, -1.8172e-07, -9.3728e-06, + -1.9465e-07, 3.1684e-06, 2.8014e-06, 1.0151e-06, -1.5134e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 249.01, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.4782 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.01 lr 0.00010000 +Epoch 206, weight, value: tensor([[ 0.0583, -0.1129, -0.1318, ..., -0.1096, -0.1265, -0.1080], + [ 0.0317, 0.0506, -0.0369, ..., -0.0445, 0.1461, 0.0185], + [-0.0122, -0.1520, -0.0026, ..., 0.1681, -0.1326, 0.0318], + ..., + [-0.0874, 0.1181, 0.0148, ..., -0.1177, 0.0942, 0.0478], + [-0.0010, -0.0636, 0.0036, ..., -0.0638, -0.1583, -0.1266], + [ 0.0063, -0.0202, -0.0094, ..., -0.1673, -0.1317, 0.0296]], + device='cuda:0'), grad: tensor([[ 8.3819e-09, 2.5902e-09, 0.0000e+00, ..., 9.1677e-08, + 3.9581e-09, 9.5461e-09], + [-6.3854e-08, -2.7649e-09, 2.9104e-11, ..., 7.8988e-08, + -1.0541e-07, 5.2387e-10], + [ 3.9581e-09, 6.5775e-09, 5.8208e-11, ..., 6.5251e-08, + 1.0303e-08, -3.1869e-08], + ..., + [ 1.4988e-08, -2.3656e-07, 2.9104e-11, ..., 2.9104e-09, + -1.8091e-07, -1.9022e-07], + [ 4.0687e-08, 2.3487e-08, 5.8208e-11, ..., 2.8545e-07, + 6.6706e-08, 1.9820e-08], + [ 9.1386e-09, 1.6019e-07, 2.9104e-11, ..., 2.1071e-08, + 1.4214e-07, 1.4668e-07]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0368, 0.0136, 0.0122, 0.0268, 0.0309, -0.0190, 0.0302, 0.0309, + -0.0121, 0.0025], device='cuda:0'), grad: tensor([ 2.2794e-07, 3.8039e-08, 1.9372e-07, 2.6333e-07, 2.5006e-07, + -6.5076e-08, -1.6652e-06, -3.6205e-07, 7.6788e-07, 3.7183e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 248.86, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.4926 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.07 lr 0.00010000 +Epoch 207, weight, value: tensor([[ 0.0587, -0.1119, -0.1319, ..., -0.1087, -0.1242, -0.1084], + [ 0.0316, 0.0506, -0.0372, ..., -0.0449, 0.1462, 0.0185], + [-0.0124, -0.1520, -0.0027, ..., 0.1681, -0.1327, 0.0318], + ..., + [-0.0883, 0.1182, 0.0144, ..., -0.1182, 0.0943, 0.0479], + [-0.0008, -0.0636, 0.0024, ..., -0.0639, -0.1584, -0.1269], + [ 0.0076, -0.0206, -0.0087, ..., -0.1679, -0.1323, 0.0294]], + device='cuda:0'), grad: tensor([[ 6.2981e-08, 1.9383e-08, 0.0000e+00, ..., 1.0547e-07, + 9.0804e-09, 4.0658e-08], + [ 2.2806e-07, 5.0815e-08, 0.0000e+00, ..., 3.5064e-07, + -2.3865e-07, 1.5402e-07], + [ 6.1356e-06, 1.3486e-06, 0.0000e+00, ..., 6.4038e-06, + 6.0245e-08, 2.7101e-06], + ..., + [ 1.4517e-07, -8.7079e-08, 0.0000e+00, ..., 1.1199e-07, + -6.8743e-08, -2.2323e-08], + [-7.0743e-06, -1.5432e-06, 0.0000e+00, ..., -7.4059e-06, + 8.0792e-08, -3.0696e-06], + [-1.7965e-06, 6.5891e-08, 0.0000e+00, ..., 6.2806e-08, + 4.9273e-08, -1.9744e-06]], device='cuda:0') +Epoch 207, bias, value: tensor([-0.0368, 0.0135, 0.0121, 0.0271, 0.0309, -0.0190, 0.0303, 0.0307, + -0.0119, 0.0028], device='cuda:0'), grad: tensor([ 3.8277e-07, 1.3541e-06, 3.0845e-05, 3.9162e-07, 8.1435e-06, + 4.9314e-07, 2.4820e-07, 4.7963e-07, -3.5107e-05, -7.2494e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 249.00, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.4882 re_mapping 0.0036 re_causal 0.0114 /// teacc 98.95 lr 0.00010000 +Epoch 208, weight, value: tensor([[ 5.9003e-02, -1.1210e-01, -1.3188e-01, ..., -1.0887e-01, + -1.2448e-01, -1.0856e-01], + [ 3.1620e-02, 5.0533e-02, -3.7071e-02, ..., -4.4759e-02, + 1.4633e-01, 1.8506e-02], + [-1.2100e-02, -1.5210e-01, -2.8116e-03, ..., 1.6900e-01, + -1.3290e-01, 3.1803e-02], + ..., + [-8.9069e-02, 1.1848e-01, 1.5012e-02, ..., -1.1863e-01, + 9.4497e-02, 4.8116e-02], + [-5.0419e-06, -6.2768e-02, 2.3477e-03, ..., -6.4211e-02, + -1.5863e-01, -1.2740e-01], + [ 7.9636e-03, -2.1183e-02, -9.0684e-03, ..., -1.6799e-01, + -1.3308e-01, 2.9193e-02]], device='cuda:0'), grad: tensor([[ 1.0536e-07, 1.5818e-08, 0.0000e+00, ..., 1.0812e-08, + 1.5352e-08, 2.2454e-08], + [ 1.4470e-07, 7.5321e-08, 0.0000e+00, ..., 3.0646e-08, + 1.2678e-07, 4.8289e-07], + [ 1.8184e-07, 9.9477e-08, 0.0000e+00, ..., 3.7107e-09, + 8.5565e-08, 4.1560e-08], + ..., + [ 1.4976e-06, 7.1479e-07, 0.0000e+00, ..., 1.4115e-09, + 7.6788e-07, 7.1153e-07], + [ 2.9453e-07, 1.8554e-08, 0.0000e+00, ..., -5.3842e-08, + 1.0704e-07, 1.0827e-07], + [ 1.1380e-07, 1.5844e-07, 0.0000e+00, ..., 3.3481e-07, + 2.0992e-06, 8.7321e-06]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0367, 0.0135, 0.0122, 0.0267, 0.0309, -0.0200, 0.0305, 0.0309, + -0.0113, 0.0028], device='cuda:0'), grad: tensor([ 7.7626e-07, 2.2482e-06, 1.3802e-06, -9.6634e-06, -2.5183e-05, + -1.1370e-05, 3.4012e-06, 1.2450e-05, 2.6394e-06, 2.3276e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 249.38, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4866 re_mapping 0.0037 re_causal 0.0123 /// teacc 99.03 lr 0.00010000 +Epoch 209, weight, value: tensor([[ 5.9846e-02, -1.1298e-01, -1.3191e-01, ..., -1.0919e-01, + -1.2504e-01, -1.0877e-01], + [ 3.1558e-02, 5.0302e-02, -3.7320e-02, ..., -4.4868e-02, + 1.4621e-01, 1.8317e-02], + [-1.2139e-02, -1.5213e-01, -2.6231e-03, ..., 1.6935e-01, + -1.3297e-01, 3.1841e-02], + ..., + [-8.8844e-02, 1.1892e-01, 1.5363e-02, ..., -1.1905e-01, + 9.4782e-02, 4.8461e-02], + [ 1.6534e-04, -6.2750e-02, 2.5992e-03, ..., -6.4606e-02, + -1.5874e-01, -1.2759e-01], + [ 7.6511e-03, -2.1637e-02, -9.3959e-03, ..., -1.6858e-01, + -1.3372e-01, 2.8332e-02]], device='cuda:0'), grad: tensor([[-1.3190e-07, -1.6531e-08, 0.0000e+00, ..., 1.8626e-09, + -1.1787e-08, 3.2160e-09], + [ 2.2163e-08, 5.4453e-08, 0.0000e+00, ..., 5.7917e-09, + 5.2270e-08, 5.7014e-08], + [ 2.9002e-08, 1.3446e-08, 0.0000e+00, ..., -6.1118e-09, + 1.1845e-08, -4.9477e-10], + ..., + [ 2.6674e-08, -3.0152e-07, 2.9104e-11, ..., 4.2637e-09, + -3.2154e-07, -3.1362e-07], + [ 7.9977e-08, 4.8749e-09, 0.0000e+00, ..., -9.7789e-09, + 4.6566e-09, 8.6875e-09], + [ 7.4855e-08, 1.9488e-07, -1.3097e-10, ..., 8.2364e-09, + 2.1979e-07, 1.9290e-07]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0363, 0.0134, 0.0123, 0.0262, 0.0317, -0.0199, 0.0305, 0.0312, + -0.0112, 0.0021], device='cuda:0'), grad: tensor([-6.1281e-07, 2.4145e-07, 1.5728e-07, 6.8871e-07, 4.7672e-08, + -1.4706e-06, 3.8487e-07, -4.6194e-07, 2.8964e-07, 7.5391e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 208, time 248.98, cls_loss 0.0013 cls_loss_mapping 0.0034 cls_loss_causal 0.5296 re_mapping 0.0036 re_causal 0.0123 /// teacc 99.12 lr 0.00010000 +Epoch 210, weight, value: tensor([[ 0.0596, -0.1142, -0.1319, ..., -0.1106, -0.1262, -0.1095], + [ 0.0319, 0.0502, -0.0375, ..., -0.0450, 0.1463, 0.0182], + [-0.0122, -0.1520, -0.0021, ..., 0.1711, -0.1329, 0.0324], + ..., + [-0.0897, 0.1185, 0.0154, ..., -0.1227, 0.0947, 0.0477], + [-0.0010, -0.0615, 0.0028, ..., -0.0659, -0.1588, -0.1270], + [ 0.0080, -0.0220, -0.0095, ..., -0.1694, -0.1341, 0.0282]], + device='cuda:0'), grad: tensor([[ 1.0384e-07, 2.6281e-08, 2.9104e-11, ..., 1.6275e-07, + 2.0358e-08, 5.5414e-08], + [ 1.1390e-04, 8.4519e-05, 1.3097e-10, ..., 5.0664e-07, + 6.5327e-05, 1.2934e-04], + [ 3.5875e-06, 3.8696e-07, 4.3656e-11, ..., 3.6694e-06, + 2.6287e-07, 1.0524e-07], + ..., + [ 1.8524e-06, 9.0990e-07, 2.6193e-10, ..., 8.9931e-08, + 7.5437e-07, 1.6382e-06], + [-8.7172e-06, 4.9872e-07, 1.6007e-10, ..., -9.7379e-06, + 3.8627e-07, 8.3819e-07], + [-1.2422e-04, -9.3400e-05, -7.7125e-10, ..., 1.0999e-06, + -7.2181e-05, -1.4281e-04]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0369, 0.0133, 0.0125, 0.0277, 0.0315, -0.0180, 0.0302, 0.0292, + -0.0125, 0.0020], device='cuda:0'), grad: tensor([ 8.5216e-07, 4.9353e-04, 3.0145e-05, 2.9892e-05, 4.1574e-05, + 1.9968e-06, 2.3730e-06, 8.0317e-06, -7.5459e-05, -5.3215e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 248.89, cls_loss 0.0015 cls_loss_mapping 0.0042 cls_loss_causal 0.5197 re_mapping 0.0036 re_causal 0.0120 /// teacc 99.14 lr 0.00010000 +Epoch 211, weight, value: tensor([[ 0.0605, -0.1169, -0.1319, ..., -0.1105, -0.1291, -0.1105], + [ 0.0321, 0.0501, -0.0375, ..., -0.0451, 0.1465, 0.0179], + [-0.0124, -0.1521, -0.0020, ..., 0.1735, -0.1323, 0.0332], + ..., + [-0.0902, 0.1184, 0.0158, ..., -0.1276, 0.0945, 0.0474], + [-0.0011, -0.0622, 0.0028, ..., -0.0661, -0.1599, -0.1278], + [ 0.0087, -0.0219, -0.0101, ..., -0.1700, -0.1342, 0.0267]], + device='cuda:0'), grad: tensor([[-6.8161e-08, 9.5315e-09, 0.0000e+00, ..., 1.0754e-08, + 3.9727e-09, 2.1522e-08], + [-5.8149e-08, -9.2550e-08, 0.0000e+00, ..., 6.5716e-08, + -4.0117e-07, 4.7497e-08], + [ 3.7689e-08, 2.1071e-08, 0.0000e+00, ..., -5.0850e-07, + -1.3242e-09, -3.4692e-07], + ..., + [ 7.6951e-08, 5.8120e-08, 0.0000e+00, ..., 1.4273e-07, + 1.1100e-07, 2.0675e-07], + [ 2.2422e-07, 4.1444e-08, 0.0000e+00, ..., 1.2806e-07, + 9.9884e-08, 3.0058e-07], + [-7.7346e-07, -1.5611e-07, 0.0000e+00, ..., -3.1112e-08, + 8.8476e-09, -1.0403e-06]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0368, 0.0132, 0.0133, 0.0294, 0.0344, -0.0181, 0.0299, 0.0268, + -0.0127, -0.0004], device='cuda:0'), grad: tensor([-2.7521e-07, -1.8324e-07, -8.5356e-07, 4.4866e-07, 2.9132e-06, + -6.0827e-08, 1.9756e-07, 6.8452e-07, 1.5143e-06, -4.3735e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 248.93, cls_loss 0.0011 cls_loss_mapping 0.0032 cls_loss_causal 0.5029 re_mapping 0.0040 re_causal 0.0120 /// teacc 98.94 lr 0.00010000 +Epoch 212, weight, value: tensor([[ 0.0602, -0.1171, -0.1319, ..., -0.1107, -0.1292, -0.1120], + [ 0.0322, 0.0500, -0.0377, ..., -0.0452, 0.1466, 0.0178], + [-0.0125, -0.1524, -0.0019, ..., 0.1737, -0.1325, 0.0331], + ..., + [-0.0907, 0.1190, 0.0164, ..., -0.1278, 0.0949, 0.0480], + [-0.0011, -0.0625, 0.0028, ..., -0.0668, -0.1603, -0.1292], + [ 0.0101, -0.0229, -0.0110, ..., -0.1707, -0.1349, 0.0264]], + device='cuda:0'), grad: tensor([[ 4.3027e-07, 1.4261e-08, 0.0000e+00, ..., 6.1846e-09, + 3.3557e-08, 1.5832e-08], + [-5.7556e-07, -6.4075e-07, 0.0000e+00, ..., -1.0151e-07, + -2.2613e-06, -6.7754e-07], + [ 1.3912e-07, 1.2340e-07, 0.0000e+00, ..., -2.5914e-07, + 3.3434e-07, -8.4168e-08], + ..., + [ 2.0047e-07, -7.3574e-08, 0.0000e+00, ..., 1.8929e-07, + 1.7427e-07, 7.9628e-08], + [ 1.1520e-06, 6.4448e-07, 0.0000e+00, ..., 3.1752e-08, + 4.9686e-07, 2.9150e-07], + [ 1.7928e-07, 5.8644e-08, 0.0000e+00, ..., 6.1846e-09, + 6.3039e-08, 2.7663e-08]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0370, 0.0131, 0.0132, 0.0297, 0.0349, -0.0181, 0.0297, 0.0272, + -0.0131, -0.0004], device='cuda:0'), grad: tensor([ 1.7118e-06, -3.0790e-06, 1.8661e-07, 4.6305e-06, 1.0952e-06, + -1.3202e-05, 2.5779e-06, 8.2329e-07, 4.4964e-06, 7.4226e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 248.62, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.5164 re_mapping 0.0036 re_causal 0.0116 /// teacc 99.13 lr 0.00010000 +Epoch 213, weight, value: tensor([[ 0.0604, -0.1175, -0.1320, ..., -0.1111, -0.1295, -0.1122], + [ 0.0326, 0.0501, -0.0377, ..., -0.0451, 0.1470, 0.0179], + [-0.0126, -0.1526, -0.0018, ..., 0.1738, -0.1331, 0.0330], + ..., + [-0.0921, 0.1191, 0.0164, ..., -0.1279, 0.0949, 0.0481], + [-0.0008, -0.0612, 0.0028, ..., -0.0668, -0.1594, -0.1290], + [ 0.0106, -0.0236, -0.0110, ..., -0.1710, -0.1362, 0.0284]], + device='cuda:0'), grad: tensor([[-2.1094e-07, 1.7681e-08, 0.0000e+00, ..., 4.3074e-09, + 1.5090e-08, 2.3225e-08], + [-4.9738e-08, 7.0548e-07, 0.0000e+00, ..., 4.9185e-09, + 6.8394e-08, 6.2864e-07], + [ 9.7934e-09, 4.4378e-07, 0.0000e+00, ..., -7.9570e-08, + 3.2037e-07, 3.8953e-07], + ..., + [ 3.0646e-08, -3.2168e-06, 0.0000e+00, ..., 2.4738e-09, + -2.0936e-06, -3.3919e-06], + [ 6.3214e-08, 1.1502e-07, 0.0000e+00, ..., 5.3609e-08, + 1.0838e-07, 1.9849e-07], + [ 2.2395e-08, 1.4082e-06, 0.0000e+00, ..., 1.4406e-09, + 1.0347e-06, 1.5413e-06]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0370, 0.0133, 0.0131, 0.0294, 0.0332, -0.0183, 0.0297, 0.0269, + -0.0121, 0.0009], device='cuda:0'), grad: tensor([-7.6368e-07, 9.6299e-07, 1.1651e-06, -4.6985e-07, 4.1653e-07, + 8.5495e-07, 1.7288e-07, -6.6385e-06, 1.0245e-06, 3.2783e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 249.07, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.5040 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +Epoch 214, weight, value: tensor([[ 0.0607, -0.1177, -0.1320, ..., -0.1114, -0.1297, -0.1124], + [ 0.0326, 0.0499, -0.0380, ..., -0.0454, 0.1470, 0.0178], + [-0.0125, -0.1530, -0.0015, ..., 0.1729, -0.1338, 0.0322], + ..., + [-0.0926, 0.1199, 0.0166, ..., -0.1260, 0.0956, 0.0495], + [-0.0006, -0.0612, 0.0028, ..., -0.0672, -0.1597, -0.1300], + [ 0.0105, -0.0243, -0.0111, ..., -0.1714, -0.1371, 0.0286]], + device='cuda:0'), grad: tensor([[ 6.4174e-09, 4.0745e-10, 0.0000e+00, ..., 4.2957e-08, + -5.1368e-09, 7.9453e-09], + [ 3.1723e-09, -4.8312e-09, 1.4552e-11, ..., 2.4098e-08, + -1.7011e-08, 8.4110e-09], + [ 2.4040e-08, 3.1578e-09, 0.0000e+00, ..., -5.9954e-08, + -3.4110e-08, -5.5705e-08], + ..., + [ 2.3108e-08, 8.7311e-10, 1.4552e-11, ..., 5.2503e-08, + 2.1755e-08, 6.4319e-08], + [ 4.1066e-08, 4.7003e-09, 2.9104e-11, ..., 2.0349e-07, + 1.3722e-08, 7.5321e-08], + [-2.9337e-07, -1.2093e-08, 0.0000e+00, ..., 5.0495e-09, + 7.2760e-09, -1.4773e-07]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0370, 0.0132, 0.0126, 0.0291, 0.0329, -0.0178, 0.0297, 0.0285, + -0.0123, 0.0008], device='cuda:0'), grad: tensor([ 8.8185e-08, 9.7440e-08, -3.0821e-08, 7.1479e-07, 3.8464e-07, + 1.1956e-07, -7.0781e-07, 2.2852e-07, 8.9128e-07, -1.7779e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 249.05, cls_loss 0.0008 cls_loss_mapping 0.0028 cls_loss_causal 0.4780 re_mapping 0.0039 re_causal 0.0121 /// teacc 99.12 lr 0.00010000 +Epoch 215, weight, value: tensor([[ 0.0610, -0.1177, -0.1320, ..., -0.1114, -0.1297, -0.1125], + [ 0.0325, 0.0498, -0.0363, ..., -0.0456, 0.1470, 0.0175], + [-0.0127, -0.1532, -0.0016, ..., 0.1728, -0.1340, 0.0321], + ..., + [-0.0931, 0.1208, 0.0149, ..., -0.1258, 0.0963, 0.0510], + [-0.0006, -0.0613, 0.0030, ..., -0.0674, -0.1601, -0.1306], + [ 0.0110, -0.0247, -0.0112, ..., -0.1719, -0.1374, 0.0285]], + device='cuda:0'), grad: tensor([[ 1.0396e-07, 7.1945e-08, 2.9104e-11, ..., 6.3388e-08, + 1.4435e-07, 2.9337e-08], + [-3.4533e-06, -2.3153e-06, 7.2760e-11, ..., 2.1304e-08, + -4.7758e-06, -9.0012e-07], + [ 6.6496e-07, 4.8103e-07, 4.3656e-11, ..., 1.3446e-08, + 9.3738e-07, 1.8207e-07], + ..., + [ 1.3621e-07, -4.9919e-07, 1.7753e-09, ..., 1.5425e-09, + -2.9616e-07, -2.9453e-07], + [ 1.4994e-06, 1.0012e-06, 1.1642e-10, ..., 1.1775e-07, + 2.0824e-06, 4.3144e-07], + [ 8.6322e-08, 5.5786e-07, -3.5652e-09, ..., 6.1118e-09, + 5.5833e-07, 1.3621e-07]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0368, 0.0130, 0.0124, 0.0290, 0.0324, -0.0178, 0.0299, 0.0296, + -0.0123, 0.0007], device='cuda:0'), grad: tensor([ 5.2992e-07, -1.1131e-05, 2.2668e-06, 6.3377e-07, 7.5391e-07, + 8.1444e-07, 5.5926e-07, -3.4762e-07, 5.1633e-06, 7.4040e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 248.82, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.5011 re_mapping 0.0036 re_causal 0.0119 /// teacc 99.05 lr 0.00010000 +Epoch 216, weight, value: tensor([[ 0.0615, -0.1178, -0.1320, ..., -0.1112, -0.1298, -0.1127], + [ 0.0326, 0.0493, -0.0362, ..., -0.0456, 0.1469, 0.0173], + [-0.0127, -0.1533, -0.0016, ..., 0.1730, -0.1344, 0.0321], + ..., + [-0.0930, 0.1216, 0.0151, ..., -0.1260, 0.0969, 0.0517], + [-0.0006, -0.0614, 0.0031, ..., -0.0676, -0.1604, -0.1310], + [ 0.0111, -0.0254, -0.0113, ..., -0.1729, -0.1382, 0.0282]], + device='cuda:0'), grad: tensor([[ 1.4655e-05, 1.2413e-08, 2.9104e-11, ..., 7.4651e-09, + 4.2171e-08, 2.8908e-05], + [-4.8801e-07, -5.4104e-08, 9.7498e-10, ..., 1.4756e-08, + -8.4424e-07, 6.1409e-08], + [ 1.2117e-06, 1.8219e-07, 1.1350e-09, ..., 4.6892e-07, + 2.8918e-07, 7.2597e-07], + ..., + [ 2.5262e-07, -6.0210e-07, -4.6130e-09, ..., 1.3591e-08, + -4.3563e-07, -3.0501e-07], + [-3.8301e-07, 7.3691e-08, 1.0186e-10, ..., -4.1979e-07, + 3.3248e-07, 1.8708e-07], + [-1.7509e-05, 4.4267e-08, 1.8917e-10, ..., 2.9337e-08, + 9.9593e-08, -3.1769e-05]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0366, 0.0130, 0.0123, 0.0288, 0.0324, -0.0178, 0.0299, 0.0302, + -0.0124, 0.0005], device='cuda:0'), grad: tensor([ 1.2028e-04, -9.5740e-07, 1.2651e-05, -1.1298e-07, 6.6087e-06, + 4.0382e-06, 1.1185e-06, -3.1607e-08, -7.8604e-06, -1.3554e-04], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 215---------------------------------------------------- +epoch 215, time 249.49, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.5144 re_mapping 0.0035 re_causal 0.0120 /// teacc 99.18 lr 0.00010000 +Epoch 217, weight, value: tensor([[ 0.0614, -0.1191, -0.1320, ..., -0.1114, -0.1300, -0.1144], + [ 0.0330, 0.0492, -0.0377, ..., -0.0458, 0.1470, 0.0173], + [-0.0128, -0.1534, -0.0016, ..., 0.1736, -0.1344, 0.0323], + ..., + [-0.0933, 0.1219, 0.0171, ..., -0.1263, 0.0971, 0.0517], + [-0.0008, -0.0617, 0.0033, ..., -0.0678, -0.1609, -0.1319], + [ 0.0116, -0.0252, -0.0115, ..., -0.1736, -0.1387, 0.0285]], + device='cuda:0'), grad: tensor([[-4.8312e-09, 4.3656e-10, 0.0000e+00, ..., 3.3615e-08, + 6.6939e-10, 3.5419e-08], + [ 2.9278e-08, 1.0477e-08, 0.0000e+00, ..., 5.0204e-08, + 6.8976e-09, 7.4448e-08], + [-2.5658e-07, 5.8790e-09, 0.0000e+00, ..., -6.3609e-07, + 6.4611e-09, -6.6264e-07], + ..., + [ 1.4581e-08, -3.8097e-08, 0.0000e+00, ..., 2.6659e-08, + -3.9494e-08, 1.1933e-09], + [ 3.8475e-08, 2.5320e-09, 0.0000e+00, ..., 1.0332e-07, + 3.7544e-09, 1.1618e-07], + [ 1.4249e-07, 9.2259e-09, 0.0000e+00, ..., 2.7847e-07, + 9.6334e-09, 3.3388e-07]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0370, 0.0130, 0.0124, 0.0282, 0.0323, -0.0176, 0.0298, 0.0303, + -0.0126, 0.0008], device='cuda:0'), grad: tensor([ 1.2078e-08, 3.1013e-07, -2.2464e-06, 1.5821e-07, -2.8359e-07, + 8.0559e-08, 1.5483e-07, 5.0437e-08, 3.9721e-07, 1.3821e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 248.70, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4533 re_mapping 0.0037 re_causal 0.0116 /// teacc 98.90 lr 0.00010000 +Epoch 218, weight, value: tensor([[ 0.0620, -0.1193, -0.1320, ..., -0.1114, -0.1302, -0.1147], + [ 0.0332, 0.0493, -0.0377, ..., -0.0460, 0.1472, 0.0172], + [-0.0129, -0.1537, -0.0011, ..., 0.1734, -0.1348, 0.0320], + ..., + [-0.0939, 0.1227, 0.0170, ..., -0.1256, 0.0975, 0.0535], + [-0.0006, -0.0617, 0.0034, ..., -0.0680, -0.1610, -0.1323], + [ 0.0110, -0.0253, -0.0115, ..., -0.1762, -0.1390, 0.0278]], + device='cuda:0'), grad: tensor([[-7.9744e-08, 3.6438e-08, 2.9104e-11, ..., 1.4133e-07, + 1.4156e-07, 8.0618e-08], + [-1.5236e-06, -3.8068e-07, -6.4028e-10, ..., 6.6881e-08, + -2.3283e-06, -1.3718e-06], + [ 6.8732e-07, 1.1706e-07, 8.7311e-11, ..., 6.4122e-07, + 1.0226e-06, 6.4354e-07], + ..., + [ 1.3539e-07, 7.2818e-08, 8.7311e-11, ..., 1.4203e-08, + 1.7893e-07, 1.1601e-07], + [ 2.1514e-07, 8.3819e-08, 2.9104e-10, ..., 1.2177e-07, + 2.0734e-07, 1.0518e-07], + [ 2.7183e-08, -2.5320e-09, 2.9104e-11, ..., 1.0361e-08, + 4.7876e-08, -1.9092e-08]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0367, 0.0130, 0.0123, 0.0286, 0.0326, -0.0181, 0.0297, 0.0315, + -0.0125, -0.0002], device='cuda:0'), grad: tensor([-2.1420e-08, -5.1931e-06, 4.2059e-06, 1.5914e-07, 2.7823e-07, + 2.9057e-07, -1.4231e-06, 5.5181e-07, 1.1241e-06, 4.5868e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 217, time 248.85, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4952 re_mapping 0.0035 re_causal 0.0114 /// teacc 99.13 lr 0.00010000 +Epoch 219, weight, value: tensor([[ 0.0626, -0.1193, -0.1320, ..., -0.1132, -0.1302, -0.1148], + [ 0.0334, 0.0493, -0.0377, ..., -0.0460, 0.1474, 0.0173], + [-0.0133, -0.1540, -0.0013, ..., 0.1723, -0.1355, 0.0316], + ..., + [-0.0943, 0.1231, 0.0170, ..., -0.1253, 0.0978, 0.0540], + [-0.0006, -0.0619, 0.0040, ..., -0.0683, -0.1612, -0.1328], + [ 0.0109, -0.0256, -0.0115, ..., -0.1768, -0.1393, 0.0275]], + device='cuda:0'), grad: tensor([[ 6.7055e-08, 1.1933e-09, 5.1630e-08, ..., 1.8190e-08, + 4.2230e-08, 2.5466e-08], + [-3.1927e-08, -7.2992e-08, 1.8132e-08, ..., 2.0576e-08, + -4.8161e-05, -2.5928e-05], + [ 7.9453e-08, 1.0565e-08, 3.8388e-08, ..., -4.8982e-08, + 1.6451e-05, 8.8215e-06], + ..., + [ 5.7073e-08, 4.2986e-08, 9.2550e-09, ..., 5.5006e-09, + 2.8819e-05, 1.5527e-05], + [-7.9023e-07, 7.8289e-09, -4.4703e-07, ..., 6.8627e-08, + 2.0745e-07, 1.1810e-07], + [ 1.1624e-07, 2.2992e-09, 6.3737e-08, ..., 8.5565e-09, + 1.2689e-07, 5.6491e-08]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0368, 0.0131, 0.0117, 0.0290, 0.0327, -0.0176, 0.0298, 0.0320, + -0.0127, -0.0004], device='cuda:0'), grad: tensor([ 4.7009e-07, -7.0691e-05, 2.4468e-05, 1.2796e-06, 2.4606e-06, + 2.4028e-06, -1.1869e-07, 4.2528e-05, -3.5297e-06, 7.6508e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 248.83, cls_loss 0.0010 cls_loss_mapping 0.0020 cls_loss_causal 0.5029 re_mapping 0.0036 re_causal 0.0118 /// teacc 99.01 lr 0.00010000 +Epoch 220, weight, value: tensor([[ 0.0630, -0.1191, -0.1320, ..., -0.1134, -0.1301, -0.1149], + [ 0.0352, 0.0493, -0.0384, ..., -0.0444, 0.1486, 0.0183], + [-0.0157, -0.1540, -0.0021, ..., 0.1708, -0.1373, 0.0306], + ..., + [-0.0951, 0.1238, 0.0184, ..., -0.1254, 0.0981, 0.0545], + [-0.0004, -0.0620, 0.0047, ..., -0.0675, -0.1613, -0.1333], + [ 0.0110, -0.0274, -0.0118, ..., -0.1771, -0.1415, 0.0270]], + device='cuda:0'), grad: tensor([[-5.0990e-08, 7.2177e-09, 1.7462e-10, ..., 3.6758e-08, + 9.0513e-09, 9.6741e-08], + [-4.7963e-08, 6.0129e-08, 1.1642e-10, ..., 1.2841e-07, + -9.2201e-08, 2.5332e-07], + [ 1.5803e-08, 6.3982e-07, 2.9104e-11, ..., -1.9802e-07, + 4.9639e-07, 4.6496e-07], + ..., + [ 8.4168e-08, -1.2843e-06, 3.4925e-10, ..., 2.8661e-07, + -9.2853e-07, -1.0021e-06], + [ 7.4622e-08, 1.4273e-07, 2.3283e-10, ..., 3.8446e-08, + 1.2899e-07, 3.8277e-07], + [-1.8964e-07, 1.2573e-07, -2.6484e-09, ..., 4.5460e-08, + 1.1490e-07, -4.2934e-07]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0367, 0.0139, 0.0107, 0.0287, 0.0327, -0.0175, 0.0297, 0.0324, + -0.0120, -0.0009], device='cuda:0'), grad: tensor([ 8.1549e-08, 6.3004e-07, 1.5423e-06, 1.1092e-06, -9.3412e-07, + 2.4517e-07, 2.4028e-07, -2.1104e-06, 1.3905e-06, -2.1942e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 249.27, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.4997 re_mapping 0.0037 re_causal 0.0119 /// teacc 99.14 lr 0.00010000 +Epoch 221, weight, value: tensor([[ 6.2953e-02, -1.1918e-01, -1.3206e-01, ..., -1.1359e-01, + -1.3028e-01, -1.1525e-01], + [ 3.5340e-02, 4.9309e-02, -3.8609e-02, ..., -4.4497e-02, + 1.4881e-01, 1.8249e-02], + [-1.6201e-02, -1.5408e-01, -4.4585e-03, ..., 1.7100e-01, + -1.3736e-01, 3.0833e-02], + ..., + [-9.6122e-02, 1.2366e-01, 1.7994e-02, ..., -1.2621e-01, + 9.8080e-02, 5.3989e-02], + [-9.9592e-05, -6.1682e-02, 3.4173e-03, ..., -6.8996e-02, + -1.6144e-01, -1.3383e-01], + [ 1.1115e-02, -2.7578e-02, -8.9636e-03, ..., -1.7728e-01, + -1.4202e-01, 2.6975e-02]], device='cuda:0'), grad: tensor([[ 4.6566e-08, 1.5541e-08, 2.7067e-09, ..., 4.5286e-07, + 1.7695e-08, 4.3749e-07], + [ 2.7474e-08, 1.3690e-07, 3.8417e-09, ..., 7.9535e-07, + -8.7195e-08, 8.7405e-07], + [-7.6881e-07, 4.8633e-08, -8.7311e-08, ..., -6.3740e-06, + 5.8237e-08, -5.0329e-06], + ..., + [ 2.0163e-07, -4.6045e-06, 8.4401e-10, ..., 8.7870e-07, + -3.8631e-06, -2.6077e-06], + [ 4.9127e-07, 3.7253e-07, 9.2841e-09, ..., 1.4752e-06, + 3.6042e-07, 1.8990e-06], + [ 6.0769e-08, 3.6992e-06, 5.8208e-11, ..., 1.7800e-07, + 3.1777e-06, 2.9467e-06]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0368, 0.0139, 0.0106, 0.0284, 0.0332, -0.0171, 0.0297, 0.0318, + -0.0116, -0.0010], device='cuda:0'), grad: tensor([ 1.5814e-06, 3.5930e-06, -2.2829e-05, 4.2282e-06, 3.9451e-06, + -2.6710e-06, 2.9672e-06, -8.6278e-06, 6.9328e-06, 1.0811e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 249.04, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4918 re_mapping 0.0035 re_causal 0.0119 /// teacc 99.14 lr 0.00010000 +Epoch 222, weight, value: tensor([[ 6.3227e-02, -1.1923e-01, -1.3210e-01, ..., -1.1330e-01, + -1.3037e-01, -1.1545e-01], + [ 3.5389e-02, 4.9147e-02, -3.7580e-02, ..., -4.4735e-02, + 1.4866e-01, 1.7873e-02], + [-1.6009e-02, -1.5407e-01, -4.6168e-03, ..., 1.7177e-01, + -1.3741e-01, 3.1302e-02], + ..., + [-9.6933e-02, 1.2398e-01, 1.7107e-02, ..., -1.2677e-01, + 9.8433e-02, 5.4197e-02], + [ 1.9101e-05, -6.1851e-02, 3.6579e-03, ..., -6.9823e-02, + -1.6167e-01, -1.3496e-01], + [ 9.8616e-03, -2.7742e-02, -8.8402e-03, ..., -1.7955e-01, + -1.4226e-01, 2.5853e-02]], device='cuda:0'), grad: tensor([[ 7.4878e-07, 5.4133e-09, 3.9814e-08, ..., 3.8912e-08, + 1.6065e-08, 1.8219e-07], + [-7.2352e-08, -1.0291e-07, 5.6170e-09, ..., 1.9632e-06, + -6.1095e-07, 2.4550e-06], + [ 4.3563e-07, 4.5198e-08, 1.9529e-08, ..., -2.4140e-06, + 1.2852e-07, -3.0138e-06], + ..., + [ 1.2561e-07, -1.1892e-07, 1.1933e-09, ..., 2.2736e-07, + -1.3330e-08, 2.0897e-07], + [-2.8163e-06, 5.5967e-08, -1.3481e-07, ..., 4.7468e-08, + 1.8114e-07, 2.2922e-07], + [ 8.3167e-07, 1.6851e-08, 5.2940e-08, ..., 5.1135e-08, + 3.0297e-08, -6.3889e-07]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0366, 0.0137, 0.0109, 0.0286, 0.0342, -0.0173, 0.0296, 0.0319, + -0.0117, -0.0022], device='cuda:0'), grad: tensor([ 3.1013e-06, 6.0610e-06, -5.5879e-06, 7.5810e-07, 1.2070e-06, + 1.6587e-06, 6.5984e-07, 1.0086e-06, -9.4175e-06, 5.2387e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 249.22, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4769 re_mapping 0.0036 re_causal 0.0117 /// teacc 99.08 lr 0.00010000 +Epoch 223, weight, value: tensor([[ 6.3371e-02, -1.1949e-01, -1.3214e-01, ..., -1.1336e-01, + -1.3062e-01, -1.1577e-01], + [ 3.5468e-02, 4.9103e-02, -3.7611e-02, ..., -4.5032e-02, + 1.4874e-01, 1.7622e-02], + [-1.5983e-02, -1.5420e-01, -4.7820e-03, ..., 1.7232e-01, + -1.3739e-01, 3.1829e-02], + ..., + [-9.7409e-02, 1.2427e-01, 1.7148e-02, ..., -1.2714e-01, + 9.8534e-02, 5.4147e-02], + [ 2.0850e-05, -6.1987e-02, 4.2047e-03, ..., -7.0147e-02, + -1.6188e-01, -1.3568e-01], + [ 9.9044e-03, -2.7946e-02, -9.0789e-03, ..., -1.8146e-01, + -1.4254e-01, 2.4613e-02]], device='cuda:0'), grad: tensor([[-1.1903e-07, 5.7335e-09, 0.0000e+00, ..., 3.3760e-09, + 1.2078e-08, 1.4988e-08], + [-9.8313e-08, -7.5786e-08, 0.0000e+00, ..., 6.5775e-09, + -3.2829e-07, -4.2113e-08], + [ 3.4139e-08, 5.7771e-08, 0.0000e+00, ..., -4.3632e-07, + 8.5856e-08, -3.0966e-07], + ..., + [ 4.8283e-08, -7.3854e-07, 0.0000e+00, ..., 3.5320e-07, + -4.4890e-07, -2.3702e-07], + [-6.8394e-09, 3.6875e-08, 0.0000e+00, ..., 8.9349e-09, + 8.4809e-08, 5.5879e-08], + [ 1.6094e-08, 6.0862e-07, 0.0000e+00, ..., 1.8044e-09, + 4.6589e-07, 1.6170e-07]], device='cuda:0') +Epoch 223, bias, value: tensor([-0.0366, 0.0136, 0.0111, 0.0289, 0.0351, -0.0174, 0.0295, 0.0319, + -0.0119, -0.0031], device='cuda:0'), grad: tensor([-5.6485e-07, -3.4925e-07, -5.7137e-07, 3.5064e-07, 5.1782e-07, + 3.7998e-07, 1.0995e-07, -1.0070e-07, 7.6252e-09, 2.4028e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 248.96, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.4934 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.10 lr 0.00010000 +Epoch 224, weight, value: tensor([[ 6.4454e-02, -1.1956e-01, -1.3215e-01, ..., -1.1289e-01, + -1.3077e-01, -1.1607e-01], + [ 3.5487e-02, 4.8988e-02, -3.7637e-02, ..., -4.5178e-02, + 1.4875e-01, 1.7444e-02], + [-1.6110e-02, -1.5433e-01, -4.8074e-03, ..., 1.7270e-01, + -1.3749e-01, 3.1925e-02], + ..., + [-9.7917e-02, 1.2466e-01, 1.7186e-02, ..., -1.2732e-01, + 9.8771e-02, 5.4432e-02], + [-1.1290e-05, -6.2154e-02, 4.3296e-03, ..., -7.0806e-02, + -1.6211e-01, -1.3667e-01], + [ 1.0062e-02, -2.8252e-02, -9.0739e-03, ..., -1.8192e-01, + -1.4285e-01, 2.4295e-02]], device='cuda:0'), grad: tensor([[ 3.6962e-08, 9.4878e-09, 0.0000e+00, ..., 1.3446e-08, + 3.8766e-08, 3.9581e-08], + [-1.1493e-06, -6.8313e-07, 0.0000e+00, ..., 6.8219e-08, + -3.8594e-06, -1.3588e-06], + [ 1.7893e-07, 5.1834e-08, 0.0000e+00, ..., -1.7358e-07, + 2.2107e-07, -2.8685e-07], + ..., + [ 4.1164e-07, 1.1077e-07, 0.0000e+00, ..., 2.6368e-08, + 1.0468e-06, 4.2398e-07], + [-8.5402e-07, 8.4925e-08, 0.0000e+00, ..., 1.6589e-09, + 3.8557e-07, 2.3574e-07], + [ 1.2026e-07, 3.0093e-08, 0.0000e+00, ..., 1.9412e-08, + 2.0652e-07, 6.4902e-09]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0361, 0.0135, 0.0111, 0.0288, 0.0356, -0.0174, 0.0294, 0.0321, + -0.0122, -0.0033], device='cuda:0'), grad: tensor([ 2.1618e-07, -6.8210e-06, 2.1874e-07, 2.0880e-06, 2.3600e-06, + 2.3134e-06, 1.0198e-06, 2.5705e-06, -4.4890e-06, 5.0478e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 248.84, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4970 re_mapping 0.0034 re_causal 0.0115 /// teacc 99.11 lr 0.00010000 +Epoch 225, weight, value: tensor([[ 6.4823e-02, -1.1968e-01, -1.3215e-01, ..., -1.1303e-01, + -1.3080e-01, -1.1654e-01], + [ 3.5599e-02, 4.8880e-02, -3.7527e-02, ..., -4.5334e-02, + 1.4911e-01, 1.7589e-02], + [-1.6120e-02, -1.5435e-01, -4.9097e-03, ..., 1.7323e-01, + -1.3803e-01, 3.1958e-02], + ..., + [-9.8410e-02, 1.2496e-01, 1.7140e-02, ..., -1.2784e-01, + 9.8953e-02, 5.4382e-02], + [-2.0537e-05, -6.2340e-02, 4.2659e-03, ..., -7.1306e-02, + -1.6239e-01, -1.3786e-01], + [ 1.0461e-02, -2.8407e-02, -6.5802e-03, ..., -1.8210e-01, + -1.4332e-01, 2.4651e-02]], device='cuda:0'), grad: tensor([[-1.8335e-09, 4.1386e-08, 0.0000e+00, ..., 1.5134e-08, + 5.2067e-08, 4.0716e-08], + [ 4.1036e-08, 4.3050e-07, 0.0000e+00, ..., 7.1304e-09, + 5.0385e-07, 3.1432e-07], + [ 1.6880e-08, 6.1351e-08, 0.0000e+00, ..., -8.1782e-09, + 7.3982e-08, 3.5070e-08], + ..., + [ 1.2718e-08, -1.1707e-06, 0.0000e+00, ..., 2.3283e-09, + -1.3690e-06, -8.1630e-07], + [-4.3213e-07, 3.5914e-08, 0.0000e+00, ..., 9.7207e-09, + 4.4907e-08, 3.2131e-08], + [ 1.0827e-07, 2.0408e-07, 0.0000e+00, ..., 8.7311e-10, + 2.4051e-07, 1.3039e-07]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0360, 0.0136, 0.0111, 0.0286, 0.0353, -0.0175, 0.0295, 0.0319, + -0.0124, -0.0030], device='cuda:0'), grad: tensor([ 1.5041e-07, 1.2591e-06, 2.1944e-07, 8.5589e-07, 2.6170e-07, + 7.3016e-07, 2.0303e-07, -2.7027e-06, -1.8375e-06, 8.7079e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 249.17, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5015 re_mapping 0.0033 re_causal 0.0111 /// teacc 99.01 lr 0.00010000 +Epoch 226, weight, value: tensor([[ 0.0646, -0.1200, -0.1322, ..., -0.1137, -0.1311, -0.1172], + [ 0.0359, 0.0488, -0.0376, ..., -0.0454, 0.1495, 0.0176], + [-0.0163, -0.1543, -0.0050, ..., 0.1740, -0.1380, 0.0328], + ..., + [-0.0993, 0.1253, 0.0173, ..., -0.1285, 0.0991, 0.0540], + [ 0.0002, -0.0619, 0.0045, ..., -0.0712, -0.1633, -0.1384], + [ 0.0107, -0.0293, -0.0061, ..., -0.1822, -0.1444, 0.0246]], + device='cuda:0'), grad: tensor([[ 1.9576e-06, 2.8522e-09, 0.0000e+00, ..., 7.3016e-07, + 3.6962e-09, 1.7841e-08], + [ 6.7502e-06, 7.6834e-08, 0.0000e+00, ..., 1.7062e-06, + 8.8650e-08, 1.9965e-07], + [ 1.1057e-04, 2.0693e-08, 0.0000e+00, ..., 3.8952e-05, + 2.6339e-08, 3.9639e-08], + ..., + [ 1.6550e-06, -2.5099e-07, 0.0000e+00, ..., 5.7090e-07, + -3.1665e-07, -6.2864e-09], + [-1.4687e-04, 1.1176e-08, 0.0000e+00, ..., -4.9680e-05, + 1.4727e-08, 3.5187e-08], + [ 2.2352e-06, 2.7387e-08, 0.0000e+00, ..., 8.5821e-07, + 3.1956e-08, 1.0310e-06]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0364, 0.0137, 0.0114, 0.0284, 0.0350, -0.0181, 0.0297, 0.0315, + -0.0119, -0.0030], device='cuda:0'), grad: tensor([ 1.5102e-05, 7.0035e-05, 7.8106e-04, 5.9962e-05, 3.6359e-06, + 6.1929e-05, 6.9499e-05, 1.2852e-05, -1.0958e-03, 2.2799e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 248.97, cls_loss 0.0015 cls_loss_mapping 0.0032 cls_loss_causal 0.5036 re_mapping 0.0036 re_causal 0.0120 /// teacc 99.17 lr 0.00010000 +Epoch 227, weight, value: tensor([[ 0.0646, -0.1200, -0.1322, ..., -0.1156, -0.1311, -0.1175], + [ 0.0324, 0.0488, -0.0376, ..., -0.0456, 0.1469, 0.0175], + [-0.0175, -0.1543, -0.0050, ..., 0.1751, -0.1383, 0.0341], + ..., + [-0.1012, 0.1257, 0.0173, ..., -0.1294, 0.0993, 0.0535], + [ 0.0051, -0.0612, 0.0046, ..., -0.0730, -0.1594, -0.1387], + [ 0.0110, -0.0303, -0.0060, ..., -0.1826, -0.1453, 0.0248]], + device='cuda:0'), grad: tensor([[-1.6799e-07, 4.6712e-08, 0.0000e+00, ..., 1.3388e-09, + -1.1467e-08, 2.2131e-07], + [-9.1968e-08, 3.4645e-07, 0.0000e+00, ..., 3.9290e-09, + -2.5448e-07, 5.8627e-07], + [ 4.8807e-08, 7.9069e-07, 0.0000e+00, ..., -1.1933e-08, + 2.2049e-07, 1.0245e-06], + ..., + [ 6.4319e-08, -3.2812e-05, 0.0000e+00, ..., 4.8021e-09, + -3.7607e-06, -4.5896e-05], + [ 1.0210e-07, 3.0897e-07, 0.0000e+00, ..., 7.0140e-09, + 1.4191e-07, 3.3248e-07], + [-1.1624e-07, 2.8595e-05, 0.0000e+00, ..., 2.2701e-09, + 3.2522e-06, 3.9846e-05]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0373, 0.0102, 0.0113, 0.0281, 0.0341, -0.0183, 0.0311, 0.0308, + -0.0064, -0.0029], device='cuda:0'), grad: tensor([-4.4471e-07, 1.5674e-06, 3.0808e-06, 8.6948e-06, 2.1737e-06, + 5.2061e-07, 9.3947e-08, -1.2851e-04, 1.4380e-06, 1.1128e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 248.83, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.5226 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.07 lr 0.00010000 +Epoch 228, weight, value: tensor([[ 0.0678, -0.1184, -0.1322, ..., -0.1140, -0.1295, -0.1180], + [ 0.0324, 0.0489, -0.0373, ..., -0.0457, 0.1470, 0.0176], + [-0.0177, -0.1544, -0.0050, ..., 0.1753, -0.1384, 0.0341], + ..., + [-0.1019, 0.1261, 0.0172, ..., -0.1295, 0.0993, 0.0539], + [ 0.0050, -0.0614, 0.0045, ..., -0.0735, -0.1594, -0.1395], + [ 0.0110, -0.0316, -0.0059, ..., -0.1828, -0.1459, 0.0243]], + device='cuda:0'), grad: tensor([[ 4.7148e-09, 5.2009e-08, 0.0000e+00, ..., 5.4046e-08, + 2.9133e-08, 7.1770e-08], + [-9.5170e-08, 1.1306e-06, 0.0000e+00, ..., 3.4180e-07, + 8.3912e-07, 1.3597e-06], + [ 1.4668e-08, 1.1217e-07, 0.0000e+00, ..., -5.8254e-07, + 1.0087e-07, -4.5216e-07], + ..., + [ 3.2218e-08, -2.2072e-06, 0.0000e+00, ..., 9.9244e-08, + -1.7798e-06, -1.8273e-06], + [ 5.0262e-08, 4.0443e-07, 0.0000e+00, ..., 1.1915e-07, + 4.0559e-07, 4.9593e-07], + [-2.1246e-08, 2.6845e-07, 0.0000e+00, ..., 1.8103e-07, + 2.1094e-07, 3.5786e-07]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0355, 0.0103, 0.0112, 0.0280, 0.0339, -0.0182, 0.0310, 0.0312, + -0.0064, -0.0033], device='cuda:0'), grad: tensor([ 2.8568e-07, 3.2280e-06, -1.2461e-06, 1.6717e-07, -8.7498e-07, + -5.3551e-09, 2.1188e-07, -4.1053e-06, 1.4035e-06, 9.4064e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 248.89, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4859 re_mapping 0.0034 re_causal 0.0111 /// teacc 99.12 lr 0.00010000 +Epoch 229, weight, value: tensor([[ 0.0678, -0.1186, -0.1322, ..., -0.1146, -0.1297, -0.1185], + [ 0.0324, 0.0492, -0.0373, ..., -0.0459, 0.1474, 0.0180], + [-0.0172, -0.1544, -0.0048, ..., 0.1763, -0.1389, 0.0342], + ..., + [-0.1035, 0.1260, 0.0171, ..., -0.1297, 0.0990, 0.0537], + [ 0.0050, -0.0616, 0.0041, ..., -0.0749, -0.1595, -0.1408], + [ 0.0111, -0.0317, -0.0055, ..., -0.1829, -0.1464, 0.0243]], + device='cuda:0'), grad: tensor([[ 2.8056e-07, 3.7882e-07, 1.5134e-09, ..., 1.3516e-07, + 8.8476e-09, 3.6461e-07], + [-2.0990e-07, -1.0856e-07, -4.5693e-09, ..., 1.6892e-07, + -5.2573e-07, 6.8278e-08], + [ 2.7707e-07, 9.6334e-09, 4.1036e-09, ..., -1.5246e-06, + -5.1979e-08, -1.7062e-06], + ..., + [ 1.3073e-07, 7.0257e-08, 3.6089e-09, ..., 4.1886e-07, + 1.8207e-07, 7.8650e-07], + [ 2.9784e-06, 5.7713e-08, 1.5163e-08, ..., 5.9092e-07, + 2.6333e-07, 1.5981e-06], + [-4.1053e-06, -5.2247e-07, 3.1840e-08, ..., 4.8109e-08, + 3.2538e-08, -1.7891e-06]], device='cuda:0') +Epoch 229, bias, value: tensor([-0.0358, 0.0103, 0.0115, 0.0281, 0.0340, -0.0185, 0.0311, 0.0308, + -0.0065, -0.0032], device='cuda:0'), grad: tensor([ 2.3358e-06, 6.7987e-08, -4.6752e-06, 1.6298e-06, 2.0172e-06, + 3.7439e-07, -9.0012e-07, 2.4177e-06, 1.5765e-05, -1.9073e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 248.76, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4824 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.11 lr 0.00010000 +Epoch 230, weight, value: tensor([[ 0.0678, -0.1188, -0.1322, ..., -0.1150, -0.1300, -0.1189], + [ 0.0325, 0.0492, -0.0374, ..., -0.0459, 0.1476, 0.0181], + [-0.0169, -0.1546, -0.0044, ..., 0.1768, -0.1393, 0.0342], + ..., + [-0.1040, 0.1265, 0.0171, ..., -0.1299, 0.0994, 0.0541], + [ 0.0050, -0.0618, 0.0037, ..., -0.0757, -0.1596, -0.1420], + [ 0.0116, -0.0321, -0.0057, ..., -0.1834, -0.1472, 0.0241]], + device='cuda:0'), grad: tensor([[-7.4797e-09, 1.1059e-09, 0.0000e+00, ..., 2.2555e-08, + 2.2119e-09, 6.5775e-09], + [-4.9185e-09, -1.0332e-08, 0.0000e+00, ..., 5.6520e-08, + -8.3237e-08, 3.7835e-08], + [ 3.2538e-08, 1.4756e-08, 2.9104e-11, ..., -6.5309e-08, + 2.3167e-08, 4.9477e-10], + ..., + [ 1.7491e-08, -4.4587e-08, 0.0000e+00, ..., 1.0623e-08, + -3.2713e-08, -3.2276e-08], + [-1.2957e-07, 1.4144e-08, 5.8208e-11, ..., 4.9477e-09, + 4.3277e-08, 2.6339e-08], + [-1.8987e-07, 6.5484e-09, 2.9104e-11, ..., 1.2427e-08, + 1.2486e-08, 1.6211e-08]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0359, 0.0103, 0.0116, 0.0280, 0.0339, -0.0186, 0.0312, 0.0312, + -0.0065, -0.0034], device='cuda:0'), grad: tensor([-4.7788e-08, 2.1933e-07, 9.6974e-08, 1.6543e-07, -1.9488e-07, + 1.2359e-06, -1.2957e-07, -2.5728e-08, -5.3039e-07, -7.8604e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 249.16, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4836 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 +Epoch 231, weight, value: tensor([[ 0.0691, -0.1189, -0.1322, ..., -0.1154, -0.1303, -0.1191], + [ 0.0325, 0.0492, -0.0374, ..., -0.0461, 0.1478, 0.0182], + [-0.0171, -0.1547, -0.0044, ..., 0.1773, -0.1395, 0.0342], + ..., + [-0.1047, 0.1267, 0.0175, ..., -0.1302, 0.0995, 0.0542], + [ 0.0050, -0.0620, 0.0036, ..., -0.0757, -0.1596, -0.1429], + [ 0.0118, -0.0324, -0.0066, ..., -0.1836, -0.1477, 0.0241]], + device='cuda:0'), grad: tensor([[-2.2887e-07, 1.3475e-08, 5.8208e-11, ..., 1.4930e-08, + 2.1217e-08, 1.9267e-08], + [-3.3295e-07, 6.4354e-07, 3.4925e-10, ..., 2.0664e-09, + 1.5565e-07, 9.4716e-07], + [ 4.1910e-08, 1.1065e-07, 6.6939e-10, ..., 3.0559e-09, + 1.5332e-07, 1.5623e-07], + ..., + [ 1.1467e-07, -2.3488e-06, -8.4401e-10, ..., 1.5134e-09, + -2.2110e-06, -3.4142e-06], + [ 2.7358e-07, 2.9500e-07, 1.9791e-09, ..., -2.6193e-10, + 4.8243e-07, 4.2864e-07], + [ 3.1840e-08, 9.0059e-07, 3.7835e-10, ..., 3.9290e-09, + 9.2061e-07, 1.3197e-06]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0355, 0.0104, 0.0115, 0.0280, 0.0339, -0.0189, 0.0310, 0.0312, + -0.0065, -0.0032], device='cuda:0'), grad: tensor([-8.9360e-07, 1.1809e-06, 4.6217e-07, 3.3434e-07, 7.4087e-07, + 1.6065e-07, 2.6339e-08, -6.6496e-06, 1.8207e-06, 2.8014e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 248.76, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4860 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.07 lr 0.00010000 +Epoch 232, weight, value: tensor([[ 0.0692, -0.1191, -0.1323, ..., -0.1158, -0.1306, -0.1215], + [ 0.0325, 0.0490, -0.0399, ..., -0.0462, 0.1478, 0.0181], + [-0.0172, -0.1547, -0.0044, ..., 0.1779, -0.1397, 0.0344], + ..., + [-0.1055, 0.1279, 0.0199, ..., -0.1306, 0.1005, 0.0552], + [ 0.0050, -0.0623, 0.0037, ..., -0.0761, -0.1597, -0.1442], + [ 0.0121, -0.0349, -0.0072, ..., -0.1838, -0.1510, 0.0229]], + device='cuda:0'), grad: tensor([[ 5.4715e-09, 4.8778e-08, 5.8208e-11, ..., 4.7788e-08, + 7.3924e-08, 1.3493e-07], + [-3.5600e-07, -1.1030e-07, 5.8208e-11, ..., 4.0024e-07, + -4.4378e-07, 7.4459e-07], + [-1.0047e-07, 1.8103e-07, 0.0000e+00, ..., -6.2957e-07, + 2.7101e-07, -5.0617e-07], + ..., + [ 4.6659e-07, -1.6596e-06, 1.1642e-10, ..., 1.6880e-07, + -2.2575e-06, -3.3751e-06], + [ 4.0000e-07, 2.5542e-07, 1.1642e-10, ..., 6.8126e-07, + 4.1374e-07, 1.1781e-06], + [-1.1211e-07, 5.3726e-08, -7.5670e-10, ..., 1.6745e-06, + 9.2085e-08, 8.9873e-07]], device='cuda:0') +Epoch 232, bias, value: tensor([-0.0361, 0.0103, 0.0116, 0.0279, 0.0343, -0.0182, 0.0308, 0.0318, + -0.0066, -0.0040], device='cuda:0'), grad: tensor([ 4.1490e-07, 2.0098e-06, -1.5944e-06, 6.2361e-06, -7.1526e-06, + -2.6776e-07, 4.6170e-07, -6.9775e-06, 5.1670e-06, 1.6689e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 249.35, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.4932 re_mapping 0.0034 re_causal 0.0111 /// teacc 98.97 lr 0.00010000 +Epoch 233, weight, value: tensor([[ 0.0703, -0.1189, -0.1324, ..., -0.1161, -0.1305, -0.1221], + [ 0.0325, 0.0493, -0.0399, ..., -0.0463, 0.1483, 0.0185], + [-0.0173, -0.1547, -0.0045, ..., 0.1784, -0.1399, 0.0345], + ..., + [-0.1071, 0.1279, 0.0199, ..., -0.1308, 0.1001, 0.0544], + [ 0.0050, -0.0625, 0.0041, ..., -0.0764, -0.1598, -0.1452], + [ 0.0116, -0.0358, -0.0073, ..., -0.1841, -0.1526, 0.0236]], + device='cuda:0'), grad: tensor([[-1.2550e-07, 2.7940e-09, 1.1642e-10, ..., 4.7206e-08, + 1.1874e-08, 4.8312e-09], + [-1.4587e-07, -6.7696e-08, 1.1642e-10, ..., 1.7870e-08, + -3.6694e-07, -6.3446e-08], + [ 1.2992e-07, 1.4028e-08, 1.1642e-10, ..., 1.7462e-09, + 5.7858e-08, 1.9267e-08], + ..., + [ 7.8930e-08, 7.9744e-09, 3.4925e-10, ..., 1.1350e-08, + 1.2061e-07, 3.6787e-08], + [-6.0303e-08, 2.2235e-08, 1.3970e-09, ..., 1.5018e-08, + 9.4064e-08, 4.3947e-08], + [-8.7894e-09, 3.7835e-09, -1.5716e-09, ..., 3.5740e-08, + 1.2107e-08, 1.4494e-08]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0357, 0.0105, 0.0116, 0.0285, 0.0341, -0.0183, 0.0310, 0.0306, + -0.0066, -0.0037], device='cuda:0'), grad: tensor([-7.0734e-07, -3.6880e-07, 6.5845e-07, 3.1036e-07, -2.5122e-07, + -1.9674e-07, 2.4401e-07, 3.6811e-07, -1.2724e-07, 7.6659e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 249.16, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.5061 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.11 lr 0.00010000 +Epoch 234, weight, value: tensor([[ 0.0704, -0.1192, -0.1324, ..., -0.1165, -0.1308, -0.1226], + [ 0.0326, 0.0495, -0.0371, ..., -0.0464, 0.1486, 0.0185], + [-0.0175, -0.1548, -0.0073, ..., 0.1786, -0.1402, 0.0344], + ..., + [-0.1099, 0.1281, 0.0185, ..., -0.1311, 0.1000, 0.0544], + [ 0.0050, -0.0627, 0.0062, ..., -0.0766, -0.1599, -0.1463], + [ 0.0146, -0.0361, -0.0063, ..., -0.1849, -0.1531, 0.0241]], + device='cuda:0'), grad: tensor([[ 5.1083e-07, 1.9209e-09, 0.0000e+00, ..., 1.6456e-06, + 4.1327e-09, 1.9334e-06], + [ 1.7113e-07, 6.6240e-08, 0.0000e+00, ..., 4.9500e-07, + 1.0332e-07, 6.8871e-07], + [-5.9865e-06, 4.9942e-08, 0.0000e+00, ..., -1.8969e-05, + 1.0582e-07, -2.2352e-05], + ..., + [ 1.0617e-06, -2.9826e-07, 0.0000e+00, ..., 3.2336e-06, + -5.7742e-07, 3.3788e-06], + [ 4.0270e-06, 2.0722e-08, 0.0000e+00, ..., 1.2912e-05, + 4.1677e-08, 1.5274e-05], + [ 3.1898e-08, 7.7882e-08, 0.0000e+00, ..., 1.3632e-07, + 1.5961e-07, 2.3609e-07]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0358, 0.0104, 0.0115, 0.0283, 0.0342, -0.0184, 0.0309, 0.0303, + -0.0066, -0.0026], device='cuda:0'), grad: tensor([ 7.9051e-06, 2.7474e-06, -9.1076e-05, 2.0247e-06, 6.1933e-07, + 9.0804e-07, -6.6729e-07, 1.4521e-05, 6.2168e-05, 7.5763e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 249.24, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4877 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.08 lr 0.00010000 +Epoch 235, weight, value: tensor([[ 0.0716, -0.1210, -0.1325, ..., -0.1168, -0.1308, -0.1236], + [ 0.0326, 0.0494, -0.0369, ..., -0.0469, 0.1487, 0.0184], + [-0.0173, -0.1549, -0.0074, ..., 0.1797, -0.1403, 0.0348], + ..., + [-0.1108, 0.1283, 0.0182, ..., -0.1320, 0.1001, 0.0543], + [ 0.0049, -0.0628, 0.0059, ..., -0.0779, -0.1599, -0.1482], + [ 0.0161, -0.0360, -0.0065, ..., -0.1853, -0.1533, 0.0240]], + device='cuda:0'), grad: tensor([[-3.0617e-08, 4.8894e-09, 1.1642e-10, ..., 9.7672e-08, + 1.9209e-09, 3.2247e-08], + [ 8.5100e-08, 1.2282e-08, 1.7462e-10, ..., 3.6391e-07, + 2.9046e-08, 3.8464e-07], + [ 6.2922e-08, 7.4506e-09, 0.0000e+00, ..., -6.0908e-07, + -9.3249e-08, -4.8382e-07], + ..., + [ 3.2526e-07, 7.6019e-08, 2.7940e-09, ..., 7.3283e-08, + 1.8103e-08, 2.0326e-07], + [ 5.9779e-08, -1.9325e-07, 4.0745e-10, ..., 7.2643e-08, + 2.0431e-08, 5.2573e-07], + [-5.4017e-07, 4.4063e-08, -6.9849e-09, ..., 1.2980e-08, + 3.3760e-09, -7.5484e-07]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0358, 0.0104, 0.0118, 0.0284, 0.0344, -0.0183, 0.0306, 0.0301, + -0.0067, -0.0023], device='cuda:0'), grad: tensor([ 3.3109e-07, 2.1011e-06, -8.4192e-07, -1.6578e-06, 8.1770e-07, + 5.0217e-06, -4.3400e-06, 3.1684e-06, -1.1316e-06, -3.4738e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 248.89, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.5029 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.02 lr 0.00010000 +Epoch 236, weight, value: tensor([[ 0.0720, -0.1212, -0.1325, ..., -0.1184, -0.1310, -0.1239], + [ 0.0326, 0.0494, -0.0370, ..., -0.0483, 0.1488, 0.0181], + [-0.0172, -0.1549, -0.0074, ..., 0.1806, -0.1405, 0.0350], + ..., + [-0.1114, 0.1294, 0.0183, ..., -0.1320, 0.1007, 0.0559], + [ 0.0049, -0.0621, 0.0028, ..., -0.0788, -0.1600, -0.1489], + [ 0.0162, -0.0366, -0.0073, ..., -0.1855, -0.1540, 0.0238]], + device='cuda:0'), grad: tensor([[ 8.7894e-09, 1.1816e-08, 4.6566e-10, ..., 2.6484e-08, + 1.2922e-08, 3.9116e-08], + [ 1.5590e-06, 2.0303e-06, 1.9209e-09, ..., 8.8592e-08, + 1.8654e-06, 1.0673e-06], + [ 3.0617e-08, 1.1415e-07, 9.3132e-10, ..., -1.9222e-06, + 1.2445e-07, -1.3877e-06], + ..., + [ 8.0792e-08, -2.6599e-06, 2.2002e-08, ..., 5.5740e-07, + -2.1048e-06, -1.1725e-06], + [-1.4585e-06, -2.1141e-07, 1.4552e-09, ..., 1.0128e-07, + -5.1595e-07, 3.6554e-07], + [-8.0210e-08, 1.9441e-07, 1.1525e-08, ..., 5.7684e-08, + 1.6426e-07, -2.4890e-07]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0364, 0.0103, 0.0120, 0.0283, 0.0328, -0.0183, 0.0317, 0.0314, + -0.0067, -0.0025], device='cuda:0'), grad: tensor([ 2.3097e-07, 8.3894e-06, -7.0520e-06, 5.3421e-06, 7.1386e-07, + -2.7362e-06, 1.3663e-06, -1.6196e-06, -2.7269e-06, -1.9185e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 248.81, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.4909 re_mapping 0.0034 re_causal 0.0115 /// teacc 99.04 lr 0.00010000 +Epoch 237, weight, value: tensor([[ 0.0724, -0.1214, -0.1326, ..., -0.1186, -0.1313, -0.1242], + [ 0.0326, 0.0494, -0.0365, ..., -0.0473, 0.1494, 0.0186], + [-0.0174, -0.1557, -0.0077, ..., 0.1801, -0.1425, 0.0344], + ..., + [-0.1125, 0.1313, 0.0183, ..., -0.1318, 0.1016, 0.0565], + [ 0.0049, -0.0623, 0.0026, ..., -0.0795, -0.1601, -0.1503], + [ 0.0161, -0.0371, -0.0075, ..., -0.1870, -0.1545, 0.0231]], + device='cuda:0'), grad: tensor([[ 1.9791e-09, 1.5076e-08, 0.0000e+00, ..., 1.6124e-08, + 2.1711e-08, 5.9896e-08], + [-4.6159e-08, 8.3703e-08, 0.0000e+00, ..., 6.2049e-08, + -4.3139e-06, -1.3588e-06], + [ 5.9372e-09, 5.2096e-08, 0.0000e+00, ..., -3.8277e-07, + 6.5984e-07, -2.1176e-07], + ..., + [ 1.5774e-08, 8.0392e-06, 0.0000e+00, ..., 1.3423e-07, + 1.0736e-05, 2.7493e-05], + [ 2.7998e-08, 3.3353e-08, 0.0000e+00, ..., 4.9360e-08, + 9.7789e-08, 1.2643e-07], + [-4.1211e-08, -9.1419e-06, 0.0000e+00, ..., 3.4634e-08, + -8.7693e-06, -2.9430e-05]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0364, 0.0104, 0.0116, 0.0266, 0.0340, -0.0187, 0.0321, 0.0326, + -0.0067, -0.0035], device='cuda:0'), grad: tensor([ 1.7695e-07, -5.1521e-06, -6.6543e-07, 1.3914e-06, 6.9849e-06, + 1.1863e-07, 1.0559e-07, 6.4790e-05, 5.3179e-07, -6.8247e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 248.79, cls_loss 0.0009 cls_loss_mapping 0.0030 cls_loss_causal 0.4784 re_mapping 0.0033 re_causal 0.0109 /// teacc 98.87 lr 0.00010000 +Epoch 238, weight, value: tensor([[ 0.0717, -0.1213, -0.1326, ..., -0.1185, -0.1313, -0.1268], + [ 0.0327, 0.0493, -0.0364, ..., -0.0474, 0.1496, 0.0186], + [-0.0174, -0.1560, -0.0078, ..., 0.1809, -0.1429, 0.0347], + ..., + [-0.1136, 0.1314, 0.0184, ..., -0.1322, 0.1017, 0.0565], + [ 0.0048, -0.0620, 0.0027, ..., -0.0806, -0.1601, -0.1514], + [ 0.0165, -0.0370, -0.0076, ..., -0.1871, -0.1544, 0.0237]], + device='cuda:0'), grad: tensor([[-4.2515e-07, -4.2608e-07, 0.0000e+00, ..., 1.4086e-07, + -3.6461e-07, 1.8114e-07], + [-6.0722e-07, -3.9628e-07, -3.5507e-09, ..., -9.8255e-07, + -4.5970e-06, -2.9635e-06], + [ 7.3400e-08, -2.3469e-07, 4.6566e-10, ..., -1.8375e-06, + 3.3528e-06, -1.9372e-06], + ..., + [ 4.2538e-07, 1.3420e-06, 1.4552e-09, ..., 1.7351e-06, + 8.8150e-07, 3.8818e-06], + [ 2.0745e-07, 2.4238e-07, 6.9849e-10, ..., 3.2783e-07, + 4.0024e-07, 7.0874e-07], + [ 4.8196e-08, -8.6799e-07, 1.1642e-10, ..., 3.8068e-08, + -7.5495e-08, -9.4576e-07]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0370, 0.0104, 0.0118, 0.0262, 0.0339, -0.0148, 0.0292, 0.0322, + -0.0068, -0.0030], device='cuda:0'), grad: tensor([-5.6960e-06, -9.4548e-06, -3.2280e-06, 1.4473e-06, 5.0059e-07, + 1.4789e-06, 8.8941e-07, 1.4976e-05, 2.1905e-06, -3.1460e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 249.10, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.5152 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.08 lr 0.00010000 +Epoch 239, weight, value: tensor([[ 0.0722, -0.1218, -0.1326, ..., -0.1192, -0.1316, -0.1273], + [ 0.0327, 0.0489, -0.0364, ..., -0.0477, 0.1497, 0.0182], + [-0.0172, -0.1561, -0.0080, ..., 0.1822, -0.1431, 0.0351], + ..., + [-0.1150, 0.1297, 0.0192, ..., -0.1327, 0.0999, 0.0547], + [ 0.0048, -0.0628, 0.0028, ..., -0.0816, -0.1603, -0.1532], + [ 0.0169, -0.0337, -0.0078, ..., -0.1876, -0.1514, 0.0263]], + device='cuda:0'), grad: tensor([[-1.8219e-08, 2.8522e-09, 0.0000e+00, ..., 8.7311e-10, + 4.0163e-09, 3.5507e-09], + [-3.9639e-08, 2.3516e-08, 0.0000e+00, ..., -3.0268e-09, + -6.7113e-08, 1.3213e-08], + [ 5.8499e-08, 2.7649e-08, 0.0000e+00, ..., 1.2515e-08, + 6.3563e-08, 4.7323e-08], + ..., + [ 1.4959e-08, -4.1956e-07, 0.0000e+00, ..., 9.9535e-09, + -4.3726e-07, -3.8254e-07], + [ 9.6334e-08, 2.0431e-08, 0.0000e+00, ..., 1.0303e-08, + 5.8964e-08, 3.6438e-08], + [ 7.3342e-09, 2.7101e-07, 0.0000e+00, ..., 1.1059e-08, + 2.8615e-07, 2.4890e-07]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0369, 0.0104, 0.0121, 0.0257, 0.0338, -0.0148, 0.0284, 0.0297, + -0.0069, -0.0002], device='cuda:0'), grad: tensor([-6.9616e-08, -7.1013e-09, 4.5309e-07, -1.1167e-06, 8.9058e-09, + 2.3539e-07, 3.5041e-08, -9.7975e-07, 6.6310e-07, 7.9256e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 248.84, cls_loss 0.0018 cls_loss_mapping 0.0042 cls_loss_causal 0.4759 re_mapping 0.0036 re_causal 0.0111 /// teacc 99.13 lr 0.00010000 +Epoch 240, weight, value: tensor([[ 0.0723, -0.1221, -0.1327, ..., -0.1215, -0.1322, -0.1276], + [ 0.0328, 0.0496, -0.0366, ..., -0.0478, 0.1504, 0.0186], + [-0.0171, -0.1563, -0.0080, ..., 0.1835, -0.1434, 0.0354], + ..., + [-0.1183, 0.1300, 0.0202, ..., -0.1342, 0.0998, 0.0546], + [ 0.0048, -0.0632, 0.0027, ..., -0.0829, -0.1605, -0.1546], + [ 0.0173, -0.0342, -0.0091, ..., -0.1897, -0.1519, 0.0291]], + device='cuda:0'), grad: tensor([[-1.1642e-09, 4.6566e-10, 0.0000e+00, ..., 1.2107e-08, + 7.5670e-10, 1.1991e-08], + [-2.8522e-09, 5.8208e-09, 0.0000e+00, ..., 7.6077e-08, + -1.5309e-08, 8.9698e-08], + [ 3.9057e-08, 2.8522e-09, 0.0000e+00, ..., 1.2224e-09, + 6.4611e-09, 1.1292e-08], + ..., + [ 9.0804e-09, -3.2189e-08, 0.0000e+00, ..., 4.0105e-08, + -1.7812e-08, 2.3982e-08], + [-1.8044e-09, 5.3551e-09, 0.0000e+00, ..., 1.3039e-08, + 1.0070e-08, 3.1549e-08], + [ 2.0082e-08, 1.7055e-08, 0.0000e+00, ..., 7.1637e-06, + 1.5309e-08, 8.9854e-06]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0377, 0.0106, 0.0123, 0.0260, 0.0310, -0.0149, 0.0287, 0.0295, + -0.0070, 0.0024], device='cuda:0'), grad: tensor([ 3.2771e-08, 3.6624e-07, 3.5414e-07, -1.3420e-06, -3.6061e-05, + 9.1083e-07, 2.3539e-07, 1.5320e-07, 2.2852e-07, 3.5167e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 250.56, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4765 re_mapping 0.0034 re_causal 0.0110 /// teacc 99.07 lr 0.00010000 +Epoch 241, weight, value: tensor([[ 0.0731, -0.1226, -0.1321, ..., -0.1216, -0.1331, -0.1279], + [ 0.0329, 0.0495, -0.0373, ..., -0.0479, 0.1506, 0.0186], + [-0.0176, -0.1565, -0.0076, ..., 0.1839, -0.1438, 0.0353], + ..., + [-0.1195, 0.1309, 0.0218, ..., -0.1344, 0.1006, 0.0554], + [ 0.0048, -0.0634, 0.0025, ..., -0.0830, -0.1606, -0.1554], + [ 0.0172, -0.0350, -0.0126, ..., -0.1903, -0.1527, 0.0290]], + device='cuda:0'), grad: tensor([[ 2.6659e-08, 6.1700e-09, 4.7148e-09, ..., 2.7823e-08, + 5.6461e-09, 3.3004e-08], + [-1.1816e-08, 2.3749e-08, 2.3283e-09, ..., 4.4180e-08, + -3.5507e-09, 8.1549e-08], + [ 5.0641e-09, 4.6333e-08, 2.3283e-09, ..., -3.4552e-07, + 5.0757e-08, -2.5029e-07], + ..., + [ 1.2747e-08, 1.0722e-07, 7.0373e-08, ..., 2.0466e-07, + -4.7556e-08, 5.6904e-07], + [ 3.6496e-08, 5.6112e-08, 7.7998e-09, ..., 6.5600e-08, + 5.4715e-08, 1.2061e-07], + [ 2.7753e-06, -2.8522e-07, -8.7719e-08, ..., 1.8895e-05, + -9.0164e-08, 2.8417e-05]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0373, 0.0106, 0.0121, 0.0263, 0.0310, -0.0153, 0.0286, 0.0301, + -0.0070, 0.0024], device='cuda:0'), grad: tensor([ 2.9616e-07, 1.8778e-07, -7.0455e-07, 3.7812e-07, -7.7307e-05, + -8.2888e-07, 1.1746e-07, 1.5078e-06, 5.2573e-07, 7.5817e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 250.41, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4998 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.10 lr 0.00010000 +Epoch 242, weight, value: tensor([[ 0.0732, -0.1230, -0.1322, ..., -0.1219, -0.1341, -0.1284], + [ 0.0329, 0.0494, -0.0373, ..., -0.0478, 0.1507, 0.0188], + [-0.0178, -0.1566, -0.0078, ..., 0.1841, -0.1443, 0.0352], + ..., + [-0.1218, 0.1311, 0.0211, ..., -0.1345, 0.1009, 0.0551], + [ 0.0048, -0.0635, 0.0024, ..., -0.0833, -0.1606, -0.1567], + [ 0.0183, -0.0350, -0.0106, ..., -0.1908, -0.1528, 0.0290]], + device='cuda:0'), grad: tensor([[-1.8319e-06, 3.4925e-10, 0.0000e+00, ..., -1.4082e-06, + 1.1059e-09, 5.9372e-09], + [ 3.4343e-09, -1.8626e-09, 0.0000e+00, ..., 6.4087e-08, + -5.2562e-08, 4.5227e-08], + [ 9.4250e-07, 7.7416e-09, 0.0000e+00, ..., 4.3027e-07, + 1.9209e-09, -2.8266e-07], + ..., + [ 1.7812e-08, -2.3574e-08, 0.0000e+00, ..., 1.5297e-07, + -1.5716e-09, 1.4959e-07], + [ 3.4878e-07, 9.1386e-09, 0.0000e+00, ..., 8.4750e-08, + 3.9698e-08, 2.2002e-08], + [ 1.4424e-07, 4.6566e-09, 0.0000e+00, ..., 2.0862e-07, + 6.5193e-09, 4.0163e-09]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0375, 0.0106, 0.0119, 0.0263, 0.0310, -0.0154, 0.0286, 0.0296, + -0.0070, 0.0024], device='cuda:0'), grad: tensor([-4.4703e-06, 2.8452e-07, 1.8645e-06, -1.1995e-06, 1.2154e-07, + -4.1281e-07, 9.4203e-07, 4.7078e-07, 2.6096e-06, -1.9209e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 250.43, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4946 re_mapping 0.0032 re_causal 0.0109 /// teacc 99.09 lr 0.00010000 +Epoch 243, weight, value: tensor([[ 0.0739, -0.1233, -0.1322, ..., -0.1229, -0.1347, -0.1288], + [ 0.0329, 0.0493, -0.0391, ..., -0.0498, 0.1507, 0.0182], + [-0.0174, -0.1569, -0.0061, ..., 0.1862, -0.1444, 0.0358], + ..., + [-0.1224, 0.1314, 0.0222, ..., -0.1347, 0.1012, 0.0553], + [ 0.0047, -0.0637, 0.0023, ..., -0.0844, -0.1607, -0.1583], + [ 0.0185, -0.0351, -0.0111, ..., -0.1911, -0.1528, 0.0290]], + device='cuda:0'), grad: tensor([[-4.1968e-08, 2.2643e-08, 5.8208e-11, ..., 4.1327e-09, + 2.6543e-08, 2.8929e-08], + [-1.3039e-07, 9.8837e-08, 5.8208e-11, ..., 1.3853e-08, + -9.8487e-08, 6.7113e-08], + [ 2.3749e-08, 1.2270e-07, 5.8208e-11, ..., -3.6031e-08, + 1.6147e-07, 1.1572e-07], + ..., + [ 4.7788e-08, -7.7952e-07, 0.0000e+00, ..., 7.3924e-09, + -7.7998e-07, -8.8010e-07], + [ 2.6193e-09, 2.2817e-08, -5.2387e-10, ..., 4.6974e-08, + 4.8138e-08, 5.8673e-08], + [ 3.6554e-08, 3.7951e-07, 1.1642e-10, ..., 2.4773e-07, + 4.1653e-07, 8.6194e-07]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0375, 0.0105, 0.0125, 0.0263, 0.0310, -0.0152, 0.0289, 0.0298, + -0.0070, 0.0024], device='cuda:0'), grad: tensor([-8.6962e-08, 3.8138e-07, 7.3155e-07, 2.8755e-07, -1.7108e-06, + 3.5437e-07, -2.1164e-07, -1.9670e-06, -9.7882e-07, 3.2112e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 250.61, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4708 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.12 lr 0.00010000 +Epoch 244, weight, value: tensor([[ 0.0745, -0.1235, -0.1323, ..., -0.1232, -0.1351, -0.1289], + [ 0.0329, 0.0507, -0.0391, ..., -0.0501, 0.1521, 0.0194], + [-0.0175, -0.1571, -0.0059, ..., 0.1864, -0.1447, 0.0358], + ..., + [-0.1238, 0.1304, 0.0229, ..., -0.1345, 0.1000, 0.0543], + [ 0.0048, -0.0635, 0.0022, ..., -0.0846, -0.1607, -0.1588], + [ 0.0183, -0.0352, -0.0127, ..., -0.1916, -0.1531, 0.0289]], + device='cuda:0'), grad: tensor([[ 1.9674e-08, 1.9034e-08, 0.0000e+00, ..., 3.1781e-08, + 5.1165e-08, 1.8859e-08], + [-4.3493e-07, -4.0466e-07, 0.0000e+00, ..., 1.1455e-07, + -1.4398e-06, -9.5519e-08], + [ 7.5204e-08, 6.6939e-08, -2.3283e-10, ..., -5.2480e-07, + 2.0687e-07, -6.0117e-07], + ..., + [ 6.5891e-08, -2.1467e-07, 0.0000e+00, ..., 2.0396e-07, + -1.2724e-07, 5.5647e-08], + [-8.6671e-08, 2.2736e-07, 1.1642e-10, ..., 1.8743e-07, + 6.7847e-07, 3.5297e-07], + [ 4.5344e-08, 9.4296e-08, 0.0000e+00, ..., 6.9267e-09, + 1.1246e-07, 8.5100e-08]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0361, 0.0109, 0.0124, 0.0265, 0.0312, -0.0154, 0.0271, 0.0288, + -0.0070, 0.0023], device='cuda:0'), grad: tensor([ 2.1630e-07, -1.7667e-06, -1.2843e-06, 9.6299e-07, 1.8231e-07, + 2.7241e-07, 4.7591e-07, 3.1502e-07, 1.0477e-07, 5.1921e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 250.45, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.5160 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.10 lr 0.00010000 +Epoch 245, weight, value: tensor([[ 0.0745, -0.1240, -0.1322, ..., -0.1238, -0.1364, -0.1293], + [ 0.0329, 0.0523, -0.0395, ..., -0.0503, 0.1536, 0.0208], + [-0.0177, -0.1574, -0.0059, ..., 0.1870, -0.1450, 0.0360], + ..., + [-0.1250, 0.1294, 0.0249, ..., -0.1347, 0.0989, 0.0537], + [ 0.0048, -0.0638, 0.0021, ..., -0.0852, -0.1609, -0.1603], + [ 0.0183, -0.0363, -0.0153, ..., -0.1918, -0.1540, 0.0288]], + device='cuda:0'), grad: tensor([[-4.1560e-08, 4.4238e-09, 0.0000e+00, ..., 3.1199e-08, + -4.1910e-08, 1.7462e-08], + [ 4.9709e-08, -1.3912e-08, 1.9209e-09, ..., 2.5611e-08, + -1.4459e-07, 1.8277e-08], + [ 6.3388e-08, 4.5029e-07, 2.0198e-08, ..., -1.2619e-07, + 5.0804e-07, 5.2247e-07], + ..., + [ 8.6671e-08, -4.7963e-07, -2.2643e-08, ..., -6.7172e-08, + -4.7334e-07, -6.3702e-07], + [ 4.1607e-07, 9.6625e-09, 5.8208e-11, ..., 1.0955e-07, + 3.2014e-08, 1.0757e-07], + [ 2.6124e-07, 7.7416e-09, 5.8208e-11, ..., 4.5402e-08, + 5.4482e-08, -5.5530e-08]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0363, 0.0114, 0.0125, 0.0266, 0.0312, -0.0161, 0.0280, 0.0280, + -0.0071, 0.0022], device='cuda:0'), grad: tensor([-5.0338e-07, 1.0524e-06, 1.1995e-06, -1.6555e-05, 8.8476e-09, + 6.9328e-06, 2.0303e-07, -2.3562e-07, 5.4352e-06, 2.4512e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 250.30, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4757 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.11 lr 0.00010000 +Epoch 246, weight, value: tensor([[ 0.0748, -0.1243, -0.1330, ..., -0.1245, -0.1369, -0.1297], + [ 0.0330, 0.0524, -0.0384, ..., -0.0508, 0.1536, 0.0204], + [-0.0177, -0.1575, -0.0059, ..., 0.1874, -0.1454, 0.0359], + ..., + [-0.1268, 0.1294, 0.0239, ..., -0.1351, 0.0991, 0.0542], + [ 0.0047, -0.0640, 0.0018, ..., -0.0862, -0.1609, -0.1611], + [ 0.0180, -0.0364, -0.0153, ..., -0.1922, -0.1542, 0.0288]], + device='cuda:0'), grad: tensor([[ 9.4878e-09, 5.2387e-10, 5.8208e-11, ..., 6.1118e-09, + 8.1491e-10, 6.4028e-10], + [ 3.2596e-09, 2.9104e-09, 0.0000e+00, ..., 6.1118e-09, + -6.4611e-09, 7.0431e-09], + [ 8.2073e-09, 5.5297e-09, 0.0000e+00, ..., -1.4203e-08, + 6.2282e-09, -5.1805e-09], + ..., + [ 1.3621e-08, -2.3807e-08, 0.0000e+00, ..., 4.3656e-09, + -1.6764e-08, -1.5949e-08], + [ 1.7602e-07, 4.9477e-09, 1.1642e-10, ..., 5.0059e-09, + 8.7894e-09, 6.0536e-09], + [ 2.8801e-07, 7.1013e-09, 0.0000e+00, ..., 7.6834e-09, + 5.8790e-09, 1.4959e-08]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0365, 0.0113, 0.0124, 0.0260, 0.0312, -0.0151, 0.0284, 0.0282, + -0.0072, 0.0022], device='cuda:0'), grad: tensor([ 7.5903e-08, 9.2317e-08, 1.0361e-07, 4.4564e-07, 1.5600e-08, + -3.9786e-06, 2.3074e-07, 1.1991e-07, 1.0300e-06, 1.8803e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 250.43, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5023 re_mapping 0.0032 re_causal 0.0105 /// teacc 99.12 lr 0.00010000 +Epoch 247, weight, value: tensor([[ 0.0753, -0.1244, -0.1337, ..., -0.1232, -0.1371, -0.1300], + [ 0.0330, 0.0525, -0.0386, ..., -0.0510, 0.1536, 0.0202], + [-0.0181, -0.1576, -0.0081, ..., 0.1871, -0.1458, 0.0358], + ..., + [-0.1276, 0.1294, 0.0233, ..., -0.1354, 0.0992, 0.0545], + [ 0.0049, -0.0643, 0.0060, ..., -0.0834, -0.1611, -0.1614], + [ 0.0178, -0.0366, -0.0154, ..., -0.1929, -0.1544, 0.0288]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.5134e-09, 8.7311e-10, ..., 1.4144e-08, + 1.9791e-09, 6.1118e-09], + [-2.9569e-08, -9.6625e-09, 9.3132e-10, ..., 2.5844e-08, + -7.4971e-08, 5.1223e-09], + [ 1.1118e-08, 8.2073e-09, 5.2387e-10, ..., -1.3586e-07, + 1.2747e-08, -9.8837e-08], + ..., + [ 2.4913e-08, -2.3749e-08, -9.8953e-10, ..., 4.1095e-08, + -8.3237e-09, 1.7870e-08], + [-2.5961e-08, 1.1176e-08, 8.1491e-10, ..., 6.1991e-08, + 3.0559e-08, 5.6811e-08], + [ 2.6193e-09, 1.1583e-08, -5.2969e-09, ..., 2.1304e-08, + 1.0536e-08, 8.3819e-09]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0358, 0.0112, 0.0120, 0.0256, 0.0312, -0.0158, 0.0279, 0.0283, + -0.0069, 0.0021], device='cuda:0'), grad: tensor([ 6.5076e-08, 5.4948e-08, -2.2468e-07, 1.7718e-07, -1.5146e-07, + -2.5239e-07, 3.9814e-08, 1.7567e-07, 6.1002e-08, 6.2049e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 250.34, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.4973 re_mapping 0.0033 re_causal 0.0105 /// teacc 99.13 lr 0.00010000 +Epoch 248, weight, value: tensor([[ 0.0759, -0.1249, -0.1340, ..., -0.1227, -0.1377, -0.1303], + [ 0.0330, 0.0519, -0.0387, ..., -0.0517, 0.1533, 0.0196], + [-0.0182, -0.1574, -0.0096, ..., 0.1890, -0.1453, 0.0370], + ..., + [-0.1271, 0.1301, 0.0232, ..., -0.1376, 0.0997, 0.0546], + [ 0.0048, -0.0649, 0.0046, ..., -0.0843, -0.1613, -0.1632], + [ 0.0176, -0.0366, -0.0154, ..., -0.1936, -0.1545, 0.0288]], + device='cuda:0'), grad: tensor([[-4.0745e-09, 7.8580e-09, 0.0000e+00, ..., 3.7136e-08, + 2.0314e-08, 4.6333e-08], + [-3.6997e-07, -3.3295e-07, 1.7462e-10, ..., 5.3085e-08, + -1.0096e-06, -1.8324e-07], + [ 3.5740e-08, 1.7288e-08, 5.8208e-11, ..., -5.4063e-07, + 5.3726e-08, -7.4785e-07], + ..., + [ 7.1654e-08, 2.7358e-08, -4.0745e-10, ..., 1.4692e-07, + 1.4587e-07, 2.3004e-07], + [ 2.3888e-07, 1.9674e-07, 0.0000e+00, ..., 5.9837e-08, + 5.5879e-07, 2.1874e-07], + [ 7.7533e-08, 3.7311e-08, 1.1642e-10, ..., 1.2456e-08, + 8.8417e-08, 4.3656e-08]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0357, 0.0111, 0.0128, 0.0287, 0.0311, -0.0164, 0.0287, 0.0282, + -0.0070, 0.0021], device='cuda:0'), grad: tensor([ 1.5891e-07, -1.3169e-06, -3.4291e-06, 2.3488e-06, 3.7020e-07, + -2.8834e-06, 1.4994e-06, 1.2815e-06, 1.4855e-06, 4.9593e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 250.32, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4935 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.10 lr 0.00010000 +Epoch 249, weight, value: tensor([[ 0.0743, -0.1252, -0.1341, ..., -0.1241, -0.1382, -0.1327], + [ 0.0331, 0.0514, -0.0388, ..., -0.0521, 0.1532, 0.0193], + [-0.0183, -0.1576, -0.0097, ..., 0.1901, -0.1455, 0.0376], + ..., + [-0.1270, 0.1307, 0.0230, ..., -0.1378, 0.1001, 0.0548], + [ 0.0048, -0.0658, 0.0047, ..., -0.0846, -0.1615, -0.1647], + [ 0.0178, -0.0366, -0.0147, ..., -0.1938, -0.1546, 0.0288]], + device='cuda:0'), grad: tensor([[ 1.7893e-07, 5.2387e-10, 6.4785e-08, ..., 9.8778e-08, + 8.9640e-09, 2.0373e-08], + [ 1.7369e-07, -6.8103e-09, 6.2108e-08, ..., 2.7730e-07, + -2.3516e-08, 1.3516e-07], + [ 1.6356e-07, 7.5670e-09, 5.3144e-08, ..., -1.0314e-07, + 1.8161e-08, -1.8231e-07], + ..., + [ 1.3504e-08, -2.4389e-08, 1.5134e-09, ..., 7.8406e-08, + 2.1781e-07, 5.4343e-07], + [-2.0675e-06, 1.1001e-08, -7.0687e-07, ..., -7.6042e-07, + 4.3306e-08, 5.2329e-08], + [ 6.9290e-07, 3.7253e-09, 2.4610e-07, ..., 3.1944e-07, + -8.2515e-07, -1.7211e-06]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0374, 0.0110, 0.0131, 0.0284, 0.0311, -0.0158, 0.0290, 0.0282, + -0.0070, 0.0021], device='cuda:0'), grad: tensor([ 8.4611e-07, 1.8263e-06, 3.5460e-07, 1.5344e-07, 4.0270e-06, + 2.2605e-05, -2.0653e-05, 1.8608e-06, -7.9423e-06, -3.1088e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 250.29, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4764 re_mapping 0.0034 re_causal 0.0111 /// teacc 99.14 lr 0.00010000 +Epoch 250, weight, value: tensor([[ 0.0748, -0.1255, -0.1344, ..., -0.1241, -0.1385, -0.1331], + [ 0.0330, 0.0512, -0.0384, ..., -0.0522, 0.1531, 0.0189], + [-0.0186, -0.1577, -0.0100, ..., 0.1906, -0.1457, 0.0377], + ..., + [-0.1271, 0.1309, 0.0230, ..., -0.1381, 0.1003, 0.0550], + [ 0.0048, -0.0660, 0.0045, ..., -0.0852, -0.1616, -0.1659], + [ 0.0205, -0.0366, -0.0128, ..., -0.1942, -0.1547, 0.0289]], + device='cuda:0'), grad: tensor([[-7.9628e-07, 3.4925e-10, 5.8208e-11, ..., -5.7975e-08, + 1.9791e-09, 4.0804e-08], + [-3.9290e-08, -5.1805e-08, 1.1642e-10, ..., 6.8569e-08, + -3.8021e-07, -1.5507e-07], + [ 1.0349e-07, 1.7346e-08, 0.0000e+00, ..., -2.0640e-07, + 5.3318e-08, -1.2841e-07], + ..., + [ 6.5949e-08, -5.0059e-09, 2.9104e-10, ..., 3.1258e-08, + 6.7230e-08, 5.7509e-08], + [ 9.3482e-08, 2.9569e-08, 5.8208e-11, ..., 4.5984e-08, + 9.0105e-08, 7.2410e-08], + [-7.1363e-08, 9.3132e-10, -3.6671e-09, ..., 1.2049e-08, + 3.6089e-09, -9.8895e-08]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0372, 0.0108, 0.0131, 0.0283, 0.0311, -0.0158, 0.0290, 0.0283, + -0.0071, 0.0022], device='cuda:0'), grad: tensor([-2.4643e-06, -3.4401e-08, 6.5775e-08, -1.7649e-06, 6.8033e-07, + 1.9893e-06, 9.1782e-07, 3.7672e-07, 5.9837e-07, -3.6252e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 250.23, cls_loss 0.0008 cls_loss_mapping 0.0013 cls_loss_causal 0.4758 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.11 lr 0.00010000 +Epoch 251, weight, value: tensor([[ 0.0766, -0.1258, -0.1345, ..., -0.1223, -0.1394, -0.1335], + [ 0.0330, 0.0512, -0.0388, ..., -0.0525, 0.1532, 0.0188], + [-0.0196, -0.1580, -0.0100, ..., 0.1895, -0.1460, 0.0375], + ..., + [-0.1275, 0.1309, 0.0235, ..., -0.1386, 0.1004, 0.0551], + [ 0.0048, -0.0662, 0.0044, ..., -0.0855, -0.1617, -0.1671], + [ 0.0206, -0.0366, -0.0125, ..., -0.1944, -0.1547, 0.0289]], + device='cuda:0'), grad: tensor([[-7.9744e-09, 1.7462e-10, 5.8208e-11, ..., 2.2037e-07, + 4.6566e-10, 1.2165e-07], + [-1.1467e-08, -6.4028e-09, 1.0477e-09, ..., 6.5949e-08, + -4.6042e-08, 2.5553e-08], + [ 4.2492e-09, 2.7940e-09, 1.1642e-10, ..., -9.5274e-07, + 9.1968e-09, -6.1747e-07], + ..., + [ 5.8208e-09, -1.0827e-08, -1.6298e-09, ..., 2.7474e-08, + -6.9849e-09, 7.5670e-09], + [ 5.7742e-08, 6.5775e-09, 1.1059e-09, ..., 2.9313e-07, + 2.8347e-08, 1.9372e-07], + [ 8.1491e-10, 2.5611e-09, -6.2282e-09, ..., 4.2492e-08, + 3.3178e-09, 1.6298e-09]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0358, 0.0107, 0.0123, 0.0284, 0.0311, -0.0160, 0.0287, 0.0283, + -0.0070, 0.0022], device='cuda:0'), grad: tensor([ 6.3237e-07, 1.6461e-07, -2.9951e-06, 5.0850e-07, 7.9069e-07, + -3.8603e-07, -2.0396e-07, 9.9011e-08, 1.3253e-06, 6.1933e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 250---------------------------------------------------- +epoch 250, time 251.05, cls_loss 0.0008 cls_loss_mapping 0.0018 cls_loss_causal 0.4867 re_mapping 0.0030 re_causal 0.0104 /// teacc 99.19 lr 0.00010000 +Epoch 252, weight, value: tensor([[ 0.0768, -0.1259, -0.1366, ..., -0.1217, -0.1397, -0.1336], + [ 0.0330, 0.0513, -0.0369, ..., -0.0529, 0.1534, 0.0189], + [-0.0195, -0.1584, -0.0082, ..., 0.1895, -0.1465, 0.0374], + ..., + [-0.1284, 0.1310, 0.0228, ..., -0.1387, 0.1004, 0.0551], + [ 0.0048, -0.0664, 0.0040, ..., -0.0862, -0.1619, -0.1683], + [ 0.0206, -0.0367, -0.0130, ..., -0.1947, -0.1548, 0.0289]], + device='cuda:0'), grad: tensor([[-4.1686e-06, 1.2515e-09, 4.9477e-10, ..., -4.2468e-06, + 1.7462e-09, 4.0454e-09], + [ 1.6706e-08, 2.0314e-07, 1.8626e-09, ..., 1.1793e-07, + 1.6566e-07, 3.7742e-07], + [ 7.5391e-07, 4.7963e-07, 3.4925e-10, ..., 7.8045e-07, + 5.5507e-07, 7.2876e-07], + ..., + [ 2.1042e-08, -7.6601e-07, 7.3051e-09, ..., 2.6805e-08, + -8.7917e-07, -1.1167e-06], + [ 9.7323e-08, 3.1228e-08, 2.2643e-08, ..., 7.8231e-08, + 5.3202e-08, 1.1042e-07], + [ 1.3020e-06, 3.3586e-08, -5.2474e-08, ..., 1.3877e-06, + 3.6118e-08, -1.0699e-07]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0355, 0.0108, 0.0122, 0.0287, 0.0311, -0.0160, 0.0287, 0.0283, + -0.0071, 0.0022], device='cuda:0'), grad: tensor([-1.2122e-05, 9.7509e-07, 3.7234e-06, 6.5612e-07, -1.4377e-07, + 5.3272e-07, 4.6007e-06, -2.2501e-06, 6.2957e-07, 3.3509e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 250.13, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4809 re_mapping 0.0032 re_causal 0.0102 /// teacc 99.11 lr 0.00010000 +Epoch 253, weight, value: tensor([[ 0.0766, -0.1264, -0.1367, ..., -0.1217, -0.1425, -0.1355], + [ 0.0330, 0.0509, -0.0371, ..., -0.0564, 0.1531, 0.0173], + [-0.0198, -0.1587, -0.0083, ..., 0.1898, -0.1472, 0.0372], + ..., + [-0.1299, 0.1312, 0.0240, ..., -0.1393, 0.1006, 0.0553], + [ 0.0049, -0.0666, 0.0041, ..., -0.0864, -0.1619, -0.1694], + [ 0.0206, -0.0369, -0.0137, ..., -0.1953, -0.1551, 0.0289]], + device='cuda:0'), grad: tensor([[-9.5170e-09, 1.1642e-10, 3.2014e-10, ..., 1.4843e-09, + 3.4925e-10, 2.0664e-09], + [-1.2806e-08, 5.0641e-09, 6.4028e-10, ..., 5.3842e-09, + -2.5349e-08, 6.7812e-09], + [ 6.4902e-09, 2.8056e-08, 3.4925e-10, ..., -1.6851e-08, + 2.1450e-08, 6.8976e-09], + ..., + [ 1.1205e-08, -4.6362e-08, 3.5798e-09, ..., 5.7626e-09, + -2.3516e-08, -9.2550e-09], + [ 7.2585e-08, 3.6962e-09, 3.4197e-08, ..., 3.7253e-09, + 1.8044e-08, 8.6613e-08], + [-7.9570e-08, 4.8021e-09, -4.9011e-08, ..., 1.2515e-08, + 4.1036e-09, -1.0390e-07]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0362, 0.0102, 0.0121, 0.0289, 0.0313, -0.0164, 0.0289, 0.0284, + -0.0070, 0.0022], device='cuda:0'), grad: tensor([-3.8417e-08, 2.4127e-08, 5.4861e-08, 1.9465e-07, 8.0909e-08, + -5.7090e-07, 1.8766e-07, 1.5803e-08, 6.4587e-07, -5.7835e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 250.22, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4702 re_mapping 0.0035 re_causal 0.0110 /// teacc 99.13 lr 0.00010000 +Epoch 254, weight, value: tensor([[ 0.0769, -0.1266, -0.1367, ..., -0.1217, -0.1427, -0.1358], + [ 0.0331, 0.0510, -0.0378, ..., -0.0566, 0.1533, 0.0174], + [-0.0199, -0.1590, -0.0078, ..., 0.1906, -0.1474, 0.0375], + ..., + [-0.1301, 0.1315, 0.0254, ..., -0.1395, 0.1007, 0.0556], + [ 0.0048, -0.0672, 0.0033, ..., -0.0875, -0.1622, -0.1719], + [ 0.0201, -0.0374, -0.0146, ..., -0.1956, -0.1555, 0.0283]], + device='cuda:0'), grad: tensor([[-2.3952e-08, 6.5775e-09, 0.0000e+00, ..., 2.6193e-10, + 1.3941e-08, 5.7917e-09], + [-4.4191e-07, -1.3772e-07, 2.9104e-10, ..., 5.8208e-10, + -8.8150e-07, 2.1129e-08], + [ 3.8737e-08, 1.2352e-07, 5.5297e-10, ..., 1.7462e-10, + 1.6880e-07, 1.0699e-07], + ..., + [ 3.4081e-08, -9.2480e-07, -5.5297e-10, ..., 1.2515e-09, + -7.4040e-07, -6.8499e-07], + [ 3.1432e-07, 3.5786e-07, -1.0768e-09, ..., 5.5297e-10, + 8.2329e-07, 1.8650e-07], + [ 3.2742e-08, 1.7823e-07, 3.7835e-10, ..., 2.6193e-09, + 1.6764e-07, 1.3493e-07]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0360, 0.0103, 0.0124, 0.0289, 0.0319, -0.0163, 0.0288, 0.0286, + -0.0072, 0.0016], device='cuda:0'), grad: tensor([-1.4203e-07, -1.6997e-06, 5.1595e-07, 4.8941e-07, 2.0210e-07, + -5.4203e-07, 4.0070e-07, -2.2277e-06, 2.1625e-06, 8.4424e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 250.41, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.4820 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.07 lr 0.00010000 +Epoch 255, weight, value: tensor([[ 0.0771, -0.1267, -0.1368, ..., -0.1217, -0.1429, -0.1361], + [ 0.0332, 0.0510, -0.0383, ..., -0.0573, 0.1533, 0.0171], + [-0.0198, -0.1590, -0.0065, ..., 0.1926, -0.1476, 0.0384], + ..., + [-0.1310, 0.1330, 0.0250, ..., -0.1412, 0.1012, 0.0577], + [ 0.0046, -0.0672, 0.0027, ..., -0.0889, -0.1623, -0.1745], + [ 0.0220, -0.0380, -0.0147, ..., -0.1968, -0.1557, 0.0276]], + device='cuda:0'), grad: tensor([[ 4.0745e-09, 1.2631e-08, 2.3283e-10, ..., 2.1246e-09, + 1.7491e-08, 1.0827e-08], + [-8.3994e-08, 5.0466e-08, 1.4843e-09, ..., 1.0186e-09, + -2.1816e-07, -3.9494e-08], + [ 2.6572e-08, 1.1275e-07, 1.6589e-09, ..., -4.1473e-08, + 1.8370e-07, 3.1723e-08], + ..., + [ 3.1199e-08, -7.8697e-07, -1.6153e-08, ..., 4.4238e-09, + -7.5623e-07, -5.6112e-07], + [ 2.9802e-08, 3.0326e-08, 6.6939e-10, ..., 6.2864e-09, + 9.4355e-08, 4.7585e-08], + [ 2.9482e-08, 2.6589e-07, 7.2760e-09, ..., 1.8626e-09, + 2.5705e-07, 2.3458e-07]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0359, 0.0102, 0.0137, 0.0287, 0.0324, -0.0163, 0.0285, 0.0303, + -0.0076, 0.0009], device='cuda:0'), grad: tensor([ 6.4960e-08, -2.0606e-07, -1.5774e-08, 1.0887e-06, 1.5239e-07, + -4.8615e-07, 8.6089e-08, -1.6810e-06, 2.0664e-07, 8.0187e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 250.20, cls_loss 0.0010 cls_loss_mapping 0.0023 cls_loss_causal 0.4630 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.08 lr 0.00010000 +Epoch 256, weight, value: tensor([[ 0.0761, -0.1271, -0.1368, ..., -0.1221, -0.1434, -0.1380], + [ 0.0332, 0.0509, -0.0383, ..., -0.0576, 0.1535, 0.0169], + [-0.0199, -0.1593, -0.0065, ..., 0.1936, -0.1479, 0.0390], + ..., + [-0.1315, 0.1330, 0.0244, ..., -0.1423, 0.1012, 0.0570], + [ 0.0045, -0.0674, 0.0024, ..., -0.0894, -0.1626, -0.1764], + [ 0.0230, -0.0371, -0.0131, ..., -0.1990, -0.1545, 0.0276]], + device='cuda:0'), grad: tensor([[ 2.3923e-08, 2.2992e-08, 0.0000e+00, ..., 2.2934e-08, + 7.2410e-08, 3.4226e-08], + [-1.8254e-06, -1.4231e-06, 0.0000e+00, ..., 1.3399e-07, + -6.5044e-06, -9.5228e-07], + [ 1.2096e-07, 2.9476e-07, 0.0000e+00, ..., -4.2084e-08, + 7.4226e-07, 3.9418e-07], + ..., + [ 3.2363e-07, -6.5006e-07, 0.0000e+00, ..., 4.5518e-08, + -1.8231e-07, -1.0459e-06], + [ 1.1744e-06, 1.0412e-06, 0.0000e+00, ..., 1.8242e-07, + 4.0717e-06, 1.0710e-06], + [ 1.3376e-07, 2.0640e-07, 0.0000e+00, ..., 9.5926e-08, + 6.2818e-07, 4.1653e-07]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0371, 0.0102, 0.0140, 0.0279, 0.0325, -0.0164, 0.0287, 0.0298, + -0.0078, 0.0010], device='cuda:0'), grad: tensor([ 2.5099e-07, -1.1988e-05, 1.5832e-06, 2.0359e-06, -1.1027e-06, + -6.0350e-06, 4.9435e-06, -7.5158e-07, 9.2089e-06, 1.7909e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 250.19, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4809 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 +Epoch 257, weight, value: tensor([[ 0.0748, -0.1273, -0.1385, ..., -0.1221, -0.1463, -0.1381], + [ 0.0333, 0.0507, -0.0388, ..., -0.0578, 0.1536, 0.0167], + [-0.0204, -0.1594, -0.0067, ..., 0.1943, -0.1477, 0.0393], + ..., + [-0.1321, 0.1333, 0.0253, ..., -0.1434, 0.1014, 0.0572], + [ 0.0046, -0.0678, 0.0035, ..., -0.0896, -0.1628, -0.1775], + [ 0.0227, -0.0373, -0.0136, ..., -0.1993, -0.1549, 0.0276]], + device='cuda:0'), grad: tensor([[ 5.2387e-10, 1.1642e-10, 1.1642e-10, ..., 9.3074e-08, + 1.3388e-09, 7.6834e-09], + [-3.3295e-08, -5.2387e-09, 2.3283e-10, ..., 6.6881e-08, + -1.0064e-07, -1.1933e-08], + [ 1.1467e-08, 3.5507e-09, 5.8208e-11, ..., 8.9873e-08, + 2.7416e-08, 1.5483e-08], + ..., + [ 5.7626e-09, -5.7044e-09, 2.0955e-09, ..., 4.4820e-09, + 1.0477e-08, 1.7986e-08], + [-1.5716e-09, 1.5716e-09, 5.8208e-11, ..., 2.0990e-07, + 1.7404e-08, 8.4983e-09], + [-1.2049e-08, 2.5029e-09, -5.3551e-09, ..., 2.5670e-08, + 3.4343e-09, -4.6799e-08]], device='cuda:0') +Epoch 257, bias, value: tensor([-0.0382, 0.0102, 0.0141, 0.0285, 0.0325, -0.0171, 0.0288, 0.0300, + -0.0076, 0.0010], device='cuda:0'), grad: tensor([ 3.1898e-07, 9.8953e-09, 3.6345e-07, 9.2143e-08, 4.9185e-08, + 6.9141e-06, -8.2403e-06, 7.2294e-08, 6.4727e-07, -2.3888e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 250.32, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.5113 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.09 lr 0.00010000 +Epoch 258, weight, value: tensor([[ 0.0750, -0.1274, -0.1386, ..., -0.1224, -0.1464, -0.1383], + [ 0.0333, 0.0508, -0.0388, ..., -0.0578, 0.1537, 0.0167], + [-0.0206, -0.1596, -0.0072, ..., 0.1948, -0.1480, 0.0395], + ..., + [-0.1327, 0.1334, 0.0253, ..., -0.1436, 0.1015, 0.0573], + [ 0.0054, -0.0680, 0.0023, ..., -0.0906, -0.1629, -0.1785], + [ 0.0225, -0.0374, -0.0138, ..., -0.1996, -0.1549, 0.0276]], + device='cuda:0'), grad: tensor([[-9.8313e-08, 8.7311e-10, 0.0000e+00, ..., 2.3632e-08, + 5.6461e-09, 1.3621e-08], + [-2.3458e-08, -2.8813e-08, 0.0000e+00, ..., 8.9873e-08, + -2.0966e-07, 8.2480e-08], + [ 2.6484e-08, 6.8685e-09, 0.0000e+00, ..., -2.4738e-08, + 4.6042e-08, 9.0222e-09], + ..., + [ 1.9441e-08, 3.2596e-09, 0.0000e+00, ..., 5.8790e-08, + 2.1188e-08, 2.2911e-07], + [ 5.9884e-07, 9.7789e-09, 0.0000e+00, ..., 8.6613e-08, + 6.9267e-08, 1.4296e-06], + [-2.5961e-07, 3.4925e-10, 0.0000e+00, ..., 3.7206e-07, + 2.1537e-09, -1.2387e-06]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0386, 0.0102, 0.0142, 0.0285, 0.0325, -0.0191, 0.0290, 0.0300, + -0.0063, 0.0010], device='cuda:0'), grad: tensor([-5.0897e-07, 4.7428e-07, 2.6124e-07, 2.1588e-06, -1.2908e-06, + -5.5879e-06, 9.0571e-07, 1.0785e-06, 1.1548e-05, -9.0301e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 249.81, cls_loss 0.0010 cls_loss_mapping 0.0019 cls_loss_causal 0.4754 re_mapping 0.0031 re_causal 0.0103 /// teacc 99.18 lr 0.00010000 +Epoch 259, weight, value: tensor([[ 0.0750, -0.1282, -0.1386, ..., -0.1225, -0.1472, -0.1389], + [ 0.0335, 0.0509, -0.0389, ..., -0.0577, 0.1541, 0.0169], + [-0.0210, -0.1602, -0.0072, ..., 0.1954, -0.1491, 0.0392], + ..., + [-0.1333, 0.1336, 0.0254, ..., -0.1434, 0.1017, 0.0571], + [ 0.0053, -0.0696, 0.0023, ..., -0.0918, -0.1635, -0.1834], + [ 0.0234, -0.0373, -0.0143, ..., -0.2001, -0.1551, 0.0278]], + device='cuda:0'), grad: tensor([[ 3.0210e-08, 1.3970e-09, 0.0000e+00, ..., -5.7044e-09, + 2.0955e-09, 4.6566e-09], + [ 1.8962e-06, -3.2596e-08, 1.7462e-10, ..., 3.4925e-09, + -1.9930e-07, -2.0780e-08], + [ 1.1030e-07, 3.6962e-08, -3.4925e-10, ..., -7.3342e-09, + 5.6461e-08, 8.1025e-08], + ..., + [ 2.9744e-08, -8.4052e-08, -1.1642e-10, ..., 1.2340e-08, + -5.6112e-08, -6.8452e-08], + [ 2.4531e-06, 1.3097e-08, 5.8208e-11, ..., 3.6671e-09, + 6.2399e-08, 1.4873e-06], + [-2.4177e-06, 1.6880e-08, 5.8208e-11, ..., 3.6089e-09, + 2.2468e-08, -2.2091e-06]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0390, 0.0103, 0.0140, 0.0283, 0.0324, -0.0191, 0.0288, 0.0296, + -0.0065, 0.0013], device='cuda:0'), grad: tensor([ 1.6880e-09, 4.3660e-06, 5.2946e-07, 2.3656e-06, 1.8720e-06, + -6.7092e-06, 5.1083e-07, -6.3155e-08, 1.1049e-05, -1.3970e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 250.09, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4834 re_mapping 0.0032 re_causal 0.0105 /// teacc 99.06 lr 0.00010000 +Epoch 260, weight, value: tensor([[ 0.0750, -0.1284, -0.1386, ..., -0.1226, -0.1473, -0.1391], + [ 0.0334, 0.0508, -0.0390, ..., -0.0580, 0.1545, 0.0164], + [-0.0213, -0.1616, -0.0072, ..., 0.1959, -0.1504, 0.0387], + ..., + [-0.1335, 0.1352, 0.0254, ..., -0.1436, 0.1029, 0.0586], + [ 0.0053, -0.0703, 0.0023, ..., -0.0923, -0.1636, -0.1854], + [ 0.0241, -0.0393, -0.0143, ..., -0.1999, -0.1571, 0.0277]], + device='cuda:0'), grad: tensor([[-1.4491e-06, 3.5507e-09, 2.0955e-09, ..., -4.2585e-07, + 8.6729e-09, 5.7626e-09], + [ 2.7148e-07, 5.0000e-08, 1.1118e-08, ..., 1.2759e-07, + -3.4133e-07, -4.9942e-08], + [ 6.7987e-07, 3.6147e-08, 5.8208e-09, ..., 2.5751e-07, + 9.2725e-08, 5.2212e-08], + ..., + [ 5.2096e-08, -5.2014e-07, 9.1596e-07, ..., 2.2701e-09, + -8.6846e-07, -6.1421e-07], + [ 1.0506e-07, 3.3004e-08, 5.6461e-09, ..., 1.7986e-08, + 1.2445e-07, 6.9849e-08], + [ 5.9954e-08, 2.5285e-07, 3.0792e-08, ..., 5.5297e-09, + 4.8988e-07, 3.0221e-07]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0392, 0.0100, 0.0136, 0.0283, 0.0323, -0.0191, 0.0287, 0.0314, + -0.0064, 0.0011], device='cuda:0'), grad: tensor([-4.7386e-06, 1.0356e-06, 2.5593e-06, -9.5218e-06, 8.3447e-07, + -5.4110e-07, 6.3656e-07, 7.3165e-06, 7.3155e-07, 1.7108e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 250.59, cls_loss 0.0017 cls_loss_mapping 0.0042 cls_loss_causal 0.5093 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.04 lr 0.00010000 +Epoch 261, weight, value: tensor([[ 0.0749, -0.1286, -0.1388, ..., -0.1227, -0.1479, -0.1394], + [ 0.0335, 0.0507, -0.0391, ..., -0.0584, 0.1546, 0.0161], + [-0.0219, -0.1598, -0.0073, ..., 0.1985, -0.1513, 0.0409], + ..., + [-0.1348, 0.1342, 0.0254, ..., -0.1471, 0.1038, 0.0571], + [ 0.0053, -0.0707, 0.0024, ..., -0.0931, -0.1638, -0.1874], + [ 0.0242, -0.0406, -0.0144, ..., -0.1985, -0.1585, 0.0290]], + device='cuda:0'), grad: tensor([[ 2.5029e-08, 7.1595e-09, 0.0000e+00, ..., 1.5716e-09, + 1.2806e-08, 1.1874e-08], + [-2.7299e-08, 1.0215e-07, 0.0000e+00, ..., 2.4273e-08, + 7.6834e-08, 1.6706e-07], + [ 4.4121e-08, 1.0161e-06, 0.0000e+00, ..., 1.4564e-07, + 1.5330e-06, 1.4678e-06], + ..., + [ 2.4564e-08, -1.4706e-06, 0.0000e+00, ..., -2.0245e-07, + -2.1812e-06, -2.1495e-06], + [-2.0850e-07, 7.2061e-08, 0.0000e+00, ..., 1.2224e-08, + 1.3493e-07, 1.1449e-07], + [ 5.3318e-08, 5.1688e-08, 0.0000e+00, ..., 1.2864e-07, + 7.9570e-08, 1.6589e-07]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0395, 0.0099, 0.0155, 0.0282, 0.0302, -0.0190, 0.0291, 0.0299, + -0.0065, 0.0031], device='cuda:0'), grad: tensor([ 1.6682e-07, 5.2107e-07, 4.7684e-06, 3.1898e-07, -1.1758e-08, + 3.4692e-07, 1.2177e-07, -6.4075e-06, -6.7428e-07, 8.4518e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 250.52, cls_loss 0.0011 cls_loss_mapping 0.0025 cls_loss_causal 0.4805 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.13 lr 0.00010000 +Epoch 262, weight, value: tensor([[ 0.0752, -0.1293, -0.1389, ..., -0.1228, -0.1482, -0.1399], + [ 0.0335, 0.0507, -0.0392, ..., -0.0586, 0.1550, 0.0161], + [-0.0207, -0.1597, -0.0063, ..., 0.1999, -0.1563, 0.0415], + ..., + [-0.1354, 0.1343, 0.0255, ..., -0.1481, 0.1049, 0.0567], + [ 0.0052, -0.0721, 0.0016, ..., -0.0955, -0.1643, -0.1912], + [ 0.0241, -0.0407, -0.0144, ..., -0.1986, -0.1586, 0.0289]], + device='cuda:0'), grad: tensor([[-2.2724e-07, 1.7462e-10, 1.5541e-08, ..., -6.2690e-08, + 6.9849e-10, 5.8208e-10], + [ 3.1490e-08, -2.2701e-09, 8.0909e-09, ..., 2.9453e-08, + -5.3784e-08, 1.8335e-08], + [ 6.1525e-08, 1.7462e-09, 7.4506e-09, ..., -8.1491e-09, + 9.9535e-09, -2.5902e-08], + ..., + [ 1.0186e-08, -9.0804e-09, 8.1491e-10, ..., 6.2282e-09, + -5.8208e-10, 6.4028e-10], + [-5.3039e-07, 2.8522e-09, -2.0443e-07, ..., -9.8429e-08, + 2.6834e-08, 4.7148e-09], + [ 1.9697e-07, 2.5611e-09, 5.7684e-08, ..., 3.7951e-08, + 2.9104e-09, 6.6357e-09]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0394, 0.0100, 0.0158, 0.0266, 0.0302, -0.0189, 0.0291, 0.0297, + -0.0068, 0.0031], device='cuda:0'), grad: tensor([-1.1679e-06, 2.9895e-07, 3.4273e-07, 8.8650e-08, 1.0832e-07, + 4.5449e-07, 1.8692e-06, 7.1654e-08, -2.9597e-06, 8.9686e-07], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 261---------------------------------------------------- +epoch 261, time 251.18, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4841 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.20 lr 0.00010000 +Epoch 263, weight, value: tensor([[ 0.0758, -0.1299, -0.1389, ..., -0.1229, -0.1489, -0.1408], + [ 0.0337, 0.0506, -0.0391, ..., -0.0583, 0.1554, 0.0161], + [-0.0209, -0.1598, -0.0066, ..., 0.1999, -0.1569, 0.0414], + ..., + [-0.1367, 0.1344, 0.0263, ..., -0.1481, 0.1052, 0.0568], + [ 0.0052, -0.0727, 0.0018, ..., -0.0961, -0.1649, -0.1933], + [ 0.0240, -0.0408, -0.0146, ..., -0.1986, -0.1587, 0.0290]], + device='cuda:0'), grad: tensor([[-1.6415e-08, 4.6566e-10, 0.0000e+00, ..., 4.7730e-09, + 2.5611e-09, 3.4925e-09], + [-1.8999e-07, -8.3237e-08, 6.1700e-09, ..., 9.8604e-08, + -5.1456e-07, -2.6776e-08], + [-3.8766e-08, 1.0594e-08, 1.0477e-09, ..., -8.5565e-08, + 3.2131e-08, -5.6229e-08], + ..., + [ 4.3306e-08, -2.6426e-08, -1.3271e-08, ..., 6.5076e-08, + 6.8219e-08, 3.4575e-08], + [ 1.7323e-07, 7.1363e-08, 3.4925e-10, ..., 1.2363e-07, + 3.4599e-07, 1.0780e-07], + [ 3.7253e-09, 5.9372e-09, 1.2806e-09, ..., 5.5763e-08, + 1.0943e-08, 3.5623e-08]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0391, 0.0102, 0.0158, 0.0263, 0.0301, -0.0190, 0.0284, 0.0298, + -0.0068, 0.0031], device='cuda:0'), grad: tensor([-5.9139e-08, -5.8115e-07, -1.4156e-07, 2.9104e-08, -1.0524e-06, + 1.4924e-07, 2.4983e-07, 2.8079e-07, 9.5740e-07, 1.8708e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 250.23, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4860 re_mapping 0.0033 re_causal 0.0107 /// teacc 99.04 lr 0.00010000 +Epoch 264, weight, value: tensor([[ 0.0759, -0.1303, -0.1389, ..., -0.1230, -0.1492, -0.1412], + [ 0.0338, 0.0494, -0.0392, ..., -0.0584, 0.1547, 0.0154], + [-0.0218, -0.1598, -0.0067, ..., 0.2000, -0.1572, 0.0414], + ..., + [-0.1376, 0.1350, 0.0267, ..., -0.1481, 0.1062, 0.0569], + [ 0.0052, -0.0737, 0.0018, ..., -0.0956, -0.1653, -0.1949], + [ 0.0239, -0.0409, -0.0147, ..., -0.1990, -0.1589, 0.0289]], + device='cuda:0'), grad: tensor([[-1.1642e-10, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 2.3283e-09, 1.0477e-09], + [-1.3039e-07, -1.2107e-08, 1.1642e-10, ..., -9.0804e-09, + -2.2387e-07, -6.5309e-08], + [ 4.4354e-08, 1.6065e-08, 5.8208e-10, ..., 6.8685e-09, + 6.2631e-08, 2.6659e-08], + ..., + [ 1.9907e-08, -2.7241e-08, -1.5134e-09, ..., 1.7462e-09, + 4.8894e-09, -1.5134e-08], + [ 2.1653e-08, 8.6147e-09, 1.1642e-10, ..., 1.0477e-09, + 1.1304e-07, 3.5740e-08], + [ 3.0734e-08, 8.8476e-09, 4.6566e-10, ..., 2.0373e-08, + 1.0710e-08, 1.3853e-08]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0391, 0.0098, 0.0158, 0.0266, 0.0301, -0.0190, 0.0272, 0.0299, + -0.0067, 0.0031], device='cuda:0'), grad: tensor([-6.8685e-09, -3.9791e-07, 2.4913e-07, -5.7276e-07, 3.6089e-09, + 5.4948e-07, 2.9686e-08, 7.4506e-09, -7.9628e-08, 2.2212e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 250.18, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4944 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.15 lr 0.00010000 +Epoch 265, weight, value: tensor([[ 0.0777, -0.1308, -0.1391, ..., -0.1228, -0.1497, -0.1413], + [ 0.0339, 0.0486, -0.0396, ..., -0.0583, 0.1543, 0.0150], + [-0.0224, -0.1598, -0.0072, ..., 0.1999, -0.1576, 0.0414], + ..., + [-0.1392, 0.1355, 0.0271, ..., -0.1481, 0.1068, 0.0570], + [ 0.0052, -0.0746, 0.0016, ..., -0.0956, -0.1656, -0.1964], + [ 0.0239, -0.0410, -0.0161, ..., -0.1997, -0.1590, 0.0287]], + device='cuda:0'), grad: tensor([[ 2.9569e-08, 1.3271e-08, 1.1642e-10, ..., 1.2806e-09, + 2.0140e-08, 2.2119e-08], + [-1.1111e-06, 3.1432e-08, -1.9209e-08, ..., 1.7928e-08, + -1.3448e-06, -7.7346e-07], + [ 9.6275e-08, 1.6927e-07, 2.0955e-09, ..., -1.2037e-07, + 2.9546e-07, 1.7043e-07], + ..., + [ 3.9604e-07, -3.2526e-07, 3.8417e-09, ..., 9.5228e-08, + 1.1816e-07, -2.9686e-08], + [ 3.1875e-07, 1.7462e-08, 4.4238e-09, ..., 1.6065e-08, + 3.5018e-07, 2.2619e-07], + [ 2.0093e-07, 4.7148e-08, 6.2864e-09, ..., 3.6322e-08, + 3.7579e-07, 2.3493e-07]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0379, 0.0096, 0.0158, 0.0273, 0.0302, -0.0190, 0.0270, 0.0300, + -0.0067, 0.0029], device='cuda:0'), grad: tensor([ 3.1432e-07, -3.8669e-06, 9.0292e-07, 2.3260e-07, 4.0047e-07, + -2.1979e-06, 1.4976e-06, 7.0594e-07, 1.3271e-06, 7.1106e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 250.62, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4887 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 +Epoch 266, weight, value: tensor([[ 0.0798, -0.1299, -0.1389, ..., -0.1219, -0.1500, -0.1388], + [ 0.0341, 0.0486, -0.0402, ..., -0.0583, 0.1547, 0.0151], + [-0.0231, -0.1599, -0.0091, ..., 0.1999, -0.1583, 0.0414], + ..., + [-0.1426, 0.1356, 0.0259, ..., -0.1481, 0.1068, 0.0570], + [ 0.0052, -0.0752, 0.0021, ..., -0.0951, -0.1661, -0.1977], + [ 0.0232, -0.0411, -0.0178, ..., -0.1999, -0.1591, 0.0287]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 2.3283e-10, 4.6566e-10, ..., 2.7940e-09, + 2.3283e-10, 1.9791e-09], + [ 2.3283e-10, 2.7940e-09, 2.3283e-10, ..., 2.7358e-08, + 2.5611e-09, 1.0594e-08], + [ 2.3283e-10, 6.2864e-09, 2.3283e-10, ..., -2.3283e-10, + 5.5879e-09, 3.2596e-09], + ..., + [ 8.3819e-09, -1.8044e-08, 6.7521e-09, ..., 1.1525e-08, + -1.6531e-08, 5.2387e-09], + [ 3.0268e-09, 9.3132e-10, 8.1491e-10, ..., 9.5926e-08, + 9.3132e-10, 2.7940e-09], + [ 1.2806e-09, 4.3074e-09, -4.8894e-09, ..., 3.6205e-08, + 3.8417e-09, 2.8405e-08]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0356, 0.0098, 0.0157, 0.0276, 0.0302, -0.0190, 0.0268, 0.0300, + -0.0067, 0.0029], device='cuda:0'), grad: tensor([ 2.2235e-08, 9.4762e-08, 4.4005e-08, -4.6939e-07, -2.0524e-07, + 5.3830e-07, -7.9256e-07, 3.8370e-07, 2.7381e-07, 1.2224e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 250.20, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4589 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.10 lr 0.00010000 +Epoch 267, weight, value: tensor([[ 0.0791, -0.1299, -0.1388, ..., -0.1216, -0.1501, -0.1408], + [ 0.0342, 0.0487, -0.0395, ..., -0.0583, 0.1551, 0.0153], + [-0.0228, -0.1599, -0.0092, ..., 0.2001, -0.1584, 0.0415], + ..., + [-0.1443, 0.1356, 0.0259, ..., -0.1481, 0.1068, 0.0571], + [ 0.0051, -0.0754, 0.0020, ..., -0.0974, -0.1664, -0.1999], + [ 0.0231, -0.0413, -0.0183, ..., -0.2005, -0.1592, 0.0285]], + device='cuda:0'), grad: tensor([[-1.3842e-07, 1.4552e-08, 0.0000e+00, ..., 4.8894e-09, + 2.0140e-08, 1.6647e-08], + [-2.2701e-08, 3.3295e-08, 0.0000e+00, ..., 3.7369e-08, + -2.3399e-08, 5.1688e-08], + [ 9.6625e-09, 9.7789e-09, 0.0000e+00, ..., -6.8569e-08, + 1.5832e-08, -4.4471e-08], + ..., + [ 2.0373e-08, -4.4773e-07, 0.0000e+00, ..., 1.2107e-08, + -5.5460e-07, -4.4261e-07], + [ 1.3039e-08, 1.6880e-08, 0.0000e+00, ..., 2.6310e-08, + 3.1316e-08, 3.2713e-08], + [ 7.1595e-08, 3.0966e-07, 0.0000e+00, ..., 3.1781e-08, + 4.1537e-07, 3.6322e-07]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0363, 0.0099, 0.0158, 0.0281, 0.0303, -0.0192, 0.0269, 0.0300, + -0.0068, 0.0028], device='cuda:0'), grad: tensor([-4.8988e-07, 1.0710e-07, -3.2946e-08, 2.3865e-08, 1.2480e-07, + 5.1456e-07, -6.2957e-07, -1.1250e-06, 1.7229e-07, 1.3495e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 250.29, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4748 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.03 lr 0.00010000 +Epoch 268, weight, value: tensor([[ 0.0793, -0.1300, -0.1388, ..., -0.1220, -0.1502, -0.1409], + [ 0.0343, 0.0486, -0.0396, ..., -0.0585, 0.1551, 0.0152], + [-0.0232, -0.1600, -0.0094, ..., 0.2000, -0.1587, 0.0415], + ..., + [-0.1447, 0.1357, 0.0258, ..., -0.1481, 0.1069, 0.0571], + [ 0.0052, -0.0755, 0.0023, ..., -0.0975, -0.1665, -0.2008], + [ 0.0230, -0.0414, -0.0186, ..., -0.2006, -0.1593, 0.0285]], + device='cuda:0'), grad: tensor([[-2.5495e-08, 1.8626e-09, 0.0000e+00, ..., 2.8382e-07, + 5.7044e-09, 3.2573e-07], + [-1.7951e-07, 9.0804e-09, 0.0000e+00, ..., 8.0676e-08, + -1.4005e-07, 9.1386e-08], + [ 2.9802e-08, 9.3132e-09, 0.0000e+00, ..., 1.1048e-07, + 3.2480e-08, 1.2980e-07], + ..., + [ 8.6147e-09, -9.8371e-08, 0.0000e+00, ..., 3.5414e-07, + -1.4575e-07, 2.7055e-07], + [ 5.2736e-08, 1.0012e-08, 1.1642e-10, ..., 4.1095e-08, + 4.6683e-08, 5.2154e-08], + [ 2.3749e-08, 2.6892e-08, 0.0000e+00, ..., 1.4715e-06, + 4.7730e-08, 1.6335e-06]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0364, 0.0099, 0.0157, 0.0288, 0.0304, -0.0193, 0.0272, 0.0300, + -0.0068, 0.0027], device='cuda:0'), grad: tensor([ 1.5609e-06, 7.0687e-07, 2.6878e-06, -4.8243e-06, -1.4193e-05, + -1.6093e-06, 4.5002e-06, 2.0955e-06, 8.2795e-07, 8.2552e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 249.37, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.4824 re_mapping 0.0030 re_causal 0.0102 /// teacc 99.07 lr 0.00010000 +Epoch 269, weight, value: tensor([[ 0.0775, -0.1304, -0.1388, ..., -0.1231, -0.1506, -0.1410], + [ 0.0345, 0.0487, -0.0396, ..., -0.0586, 0.1555, 0.0152], + [-0.0236, -0.1600, -0.0095, ..., 0.2000, -0.1589, 0.0415], + ..., + [-0.1463, 0.1359, 0.0259, ..., -0.1481, 0.1071, 0.0571], + [ 0.0051, -0.0760, 0.0025, ..., -0.0975, -0.1670, -0.2015], + [ 0.0227, -0.0418, -0.0187, ..., -0.2007, -0.1596, 0.0284]], + device='cuda:0'), grad: tensor([[-4.3423e-08, 8.1491e-10, 0.0000e+00, ..., -9.3132e-10, + -1.0128e-08, 1.0477e-09], + [ 1.2806e-08, -9.1968e-09, 1.6298e-09, ..., 7.7998e-09, + -3.5623e-08, 3.6089e-09], + [ 1.1758e-08, 1.3039e-08, 4.7730e-09, ..., 1.8626e-09, + 2.0838e-08, 1.6647e-08], + ..., + [ 1.1525e-08, -2.2119e-08, -9.1968e-09, ..., 7.1013e-09, + -8.6147e-09, -2.1537e-08], + [-7.6834e-09, 5.5879e-09, 6.9849e-10, ..., 3.7253e-09, + 1.4319e-08, 5.5879e-09], + [ 4.8894e-09, 4.8894e-09, 4.6566e-10, ..., 6.8685e-08, + 5.0059e-09, 7.9395e-08]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0384, 0.0100, 0.0157, 0.0297, 0.0304, -0.0192, 0.0273, 0.0301, + -0.0068, 0.0027], device='cuda:0'), grad: tensor([-2.1316e-07, 1.3085e-07, 1.1025e-07, 1.7462e-09, -3.4878e-07, + 6.5193e-09, 7.0781e-08, 1.9558e-08, -5.2387e-08, 2.9011e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 248.57, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4621 re_mapping 0.0031 re_causal 0.0097 /// teacc 99.08 lr 0.00010000 +Epoch 270, weight, value: tensor([[ 0.0773, -0.1302, -0.1390, ..., -0.1240, -0.1506, -0.1411], + [ 0.0345, 0.0488, -0.0381, ..., -0.0582, 0.1560, 0.0156], + [-0.0242, -0.1600, -0.0100, ..., 0.2000, -0.1595, 0.0414], + ..., + [-0.1476, 0.1359, 0.0258, ..., -0.1482, 0.1070, 0.0571], + [ 0.0051, -0.0764, 0.0024, ..., -0.0985, -0.1671, -0.2024], + [ 0.0224, -0.0418, -0.0188, ..., -0.2007, -0.1598, 0.0284]], + device='cuda:0'), grad: tensor([[ 1.0710e-08, 5.7044e-09, 0.0000e+00, ..., 1.6391e-07, + 2.1770e-08, 1.8626e-08], + [-6.3004e-07, -3.0571e-07, -1.1642e-09, ..., 7.8580e-08, + -1.2247e-06, 1.0594e-08], + [ 5.2038e-08, 1.2224e-08, 2.3283e-10, ..., 8.0909e-08, + 6.4727e-08, 6.7055e-08], + ..., + [ 1.2899e-07, 1.8743e-08, 1.1642e-10, ..., 6.7055e-08, + 1.0850e-07, 3.5809e-07], + [ 3.5809e-07, 2.0978e-07, 4.6566e-10, ..., 1.7532e-07, + 8.0327e-07, 5.0059e-08], + [-6.8103e-08, 2.6776e-08, 0.0000e+00, ..., 7.4273e-08, + 9.8487e-08, -4.2142e-07]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0388, 0.0102, 0.0156, 0.0294, 0.0304, -0.0189, 0.0280, 0.0300, + -0.0069, 0.0027], device='cuda:0'), grad: tensor([ 5.3784e-07, -2.9933e-06, 6.1467e-07, 3.6228e-07, 4.6194e-06, + 2.1197e-06, -7.7486e-06, 1.5423e-06, 2.2277e-06, -1.2759e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 250.11, cls_loss 0.0011 cls_loss_mapping 0.0028 cls_loss_causal 0.5107 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.04 lr 0.00010000 +Epoch 271, weight, value: tensor([[ 0.0772, -0.1312, -0.1391, ..., -0.1244, -0.1512, -0.1413], + [ 0.0377, 0.0518, -0.0381, ..., -0.0587, 0.1592, 0.0170], + [-0.0246, -0.1600, -0.0097, ..., 0.2000, -0.1599, 0.0414], + ..., + [-0.1482, 0.1363, 0.0259, ..., -0.1482, 0.1074, 0.0573], + [ 0.0021, -0.0797, 0.0026, ..., -0.0989, -0.1704, -0.2063], + [ 0.0208, -0.0422, -0.0190, ..., -0.2008, -0.1600, 0.0284]], + device='cuda:0'), grad: tensor([[ 5.1805e-08, -1.2922e-08, 0.0000e+00, ..., 6.2049e-08, + 2.3982e-08, 8.0327e-09], + [-2.3982e-08, -4.1444e-08, 0.0000e+00, ..., 1.4610e-07, + -1.9080e-07, -6.0769e-08], + [ 1.1455e-06, 1.7229e-08, 0.0000e+00, ..., 1.5674e-06, + 7.0198e-08, 2.0838e-08], + ..., + [ 5.7276e-08, 1.5134e-08, 9.3132e-10, ..., 4.3190e-08, + 3.8533e-08, 1.7695e-08], + [-1.4724e-06, 4.1910e-09, -1.0477e-09, ..., -2.1923e-06, + 2.8056e-08, 1.3271e-08], + [ 1.5344e-07, 1.0943e-08, 0.0000e+00, ..., 1.9744e-07, + 4.8894e-09, 3.1432e-09]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0391, 0.0133, 0.0156, 0.0289, 0.0304, -0.0190, 0.0281, 0.0302, + -0.0098, 0.0025], device='cuda:0'), grad: tensor([ 1.2096e-07, 5.6904e-07, 1.0751e-05, 2.4941e-06, 4.6217e-08, + -2.4959e-06, 8.8848e-07, 4.6613e-07, -1.4424e-05, 1.5795e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 250.50, cls_loss 0.0010 cls_loss_mapping 0.0022 cls_loss_causal 0.4751 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.12 lr 0.00010000 +Epoch 272, weight, value: tensor([[ 0.0777, -0.1316, -0.1392, ..., -0.1246, -0.1511, -0.1415], + [ 0.0378, 0.0508, -0.0382, ..., -0.0579, 0.1585, 0.0150], + [-0.0251, -0.1601, -0.0099, ..., 0.2001, -0.1610, 0.0414], + ..., + [-0.1467, 0.1380, 0.0261, ..., -0.1483, 0.1096, 0.0577], + [ 0.0020, -0.0798, 0.0025, ..., -0.1008, -0.1705, -0.2070], + [ 0.0210, -0.0422, -0.0191, ..., -0.2009, -0.1601, 0.0284]], + device='cuda:0'), grad: tensor([[ 9.1968e-09, 5.8208e-10, 0.0000e+00, ..., 1.3853e-08, + 4.8894e-09, 1.3388e-08], + [ 1.0058e-07, 3.0501e-08, 0.0000e+00, ..., 4.9546e-07, + -5.8208e-10, 4.9872e-07], + [-2.5472e-07, 7.9162e-09, 0.0000e+00, ..., -8.6520e-07, + -1.1933e-07, -7.8417e-07], + ..., + [ 6.7521e-08, -1.1188e-07, -0.0000e+00, ..., 1.4389e-07, + -5.7044e-08, -9.5461e-09], + [ 2.6589e-07, 9.4296e-09, 0.0000e+00, ..., 1.9907e-08, + 4.6683e-08, 2.8173e-08], + [ 4.8894e-09, 8.4983e-09, 0.0000e+00, ..., 1.8626e-09, + 1.1758e-08, 1.1176e-08]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0390, 0.0131, 0.0156, 0.0289, 0.0304, -0.0192, 0.0289, 0.0306, + -0.0099, 0.0025], device='cuda:0'), grad: tensor([ 4.9593e-08, 1.0831e-06, -2.0452e-06, 1.2096e-07, 5.3318e-07, + -1.1614e-06, 3.8929e-07, 1.8242e-07, 8.2096e-07, 4.0629e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 271---------------------------------------------------- +epoch 271, time 251.04, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4532 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.21 lr 0.00010000 +Epoch 273, weight, value: tensor([[ 0.0777, -0.1320, -0.1392, ..., -0.1248, -0.1516, -0.1416], + [ 0.0377, 0.0508, -0.0385, ..., -0.0577, 0.1586, 0.0152], + [-0.0255, -0.1602, -0.0103, ..., 0.2001, -0.1621, 0.0414], + ..., + [-0.1472, 0.1383, 0.0264, ..., -0.1483, 0.1099, 0.0578], + [ 0.0020, -0.0798, 0.0024, ..., -0.1021, -0.1705, -0.2072], + [ 0.0207, -0.0430, -0.0201, ..., -0.2009, -0.1605, 0.0283]], + device='cuda:0'), grad: tensor([[ 3.8650e-08, 0.0000e+00, 0.0000e+00, ..., 2.5611e-09, + 1.5134e-09, 3.9581e-09], + [-2.4447e-08, 2.3283e-10, 0.0000e+00, ..., 9.3132e-09, + -2.4214e-08, 5.7044e-09], + [ 2.3167e-08, 2.3283e-10, 0.0000e+00, ..., -1.4552e-08, + 4.7730e-09, -1.3970e-08], + ..., + [ 1.5483e-08, 0.0000e+00, 3.4925e-10, ..., 3.9581e-09, + 6.9849e-10, 7.7998e-09], + [ 1.8370e-07, 0.0000e+00, 0.0000e+00, ..., 7.3342e-09, + 4.7730e-09, 1.8394e-08], + [-2.8778e-07, 2.3283e-10, -8.1491e-10, ..., 3.4925e-10, + 1.2806e-09, -6.2631e-08]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0391, 0.0131, 0.0155, 0.0298, 0.0304, -0.0188, 0.0287, 0.0307, + -0.0099, 0.0024], device='cuda:0'), grad: tensor([ 2.5844e-07, -2.7474e-08, 9.2201e-08, 8.6054e-07, 4.2142e-07, + -1.8831e-06, 1.5448e-07, 8.0443e-08, 1.0617e-06, -1.0096e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 250.39, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4588 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.13 lr 0.00010000 +Epoch 274, weight, value: tensor([[ 0.0775, -0.1322, -0.1392, ..., -0.1249, -0.1518, -0.1419], + [ 0.0378, 0.0509, -0.0376, ..., -0.0561, 0.1592, 0.0166], + [-0.0259, -0.1602, -0.0104, ..., 0.2001, -0.1623, 0.0413], + ..., + [-0.1479, 0.1383, 0.0264, ..., -0.1483, 0.1098, 0.0578], + [ 0.0020, -0.0798, 0.0023, ..., -0.1022, -0.1705, -0.2073], + [ 0.0211, -0.0430, -0.0203, ..., -0.2010, -0.1605, 0.0283]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 2.3283e-09, + 1.6298e-09, 1.6298e-09], + [-5.2387e-09, -3.4925e-10, 1.2806e-09, ..., 1.3004e-07, + -4.8778e-08, 5.7160e-08], + [ 4.5402e-09, 3.8999e-08, -1.6298e-09, ..., -1.0186e-07, + 5.6229e-08, -1.8626e-08], + ..., + [ 1.6764e-08, -7.6601e-08, 1.1642e-10, ..., 1.9791e-08, + -7.7998e-08, -4.6217e-08], + [ 1.8510e-08, 1.2922e-08, 1.1642e-10, ..., 4.5169e-08, + 3.1665e-08, 2.8987e-08], + [ 1.7567e-07, 5.4715e-09, 0.0000e+00, ..., 6.9523e-07, + 8.2655e-09, 2.9732e-07]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0395, 0.0134, 0.0155, 0.0299, 0.0302, -0.0188, 0.0282, 0.0306, + -0.0099, 0.0024], device='cuda:0'), grad: tensor([ 1.3271e-08, 3.6787e-07, -6.4028e-08, -1.3621e-08, -3.4813e-06, + -6.2515e-08, 3.0734e-07, -8.9640e-08, 2.1956e-07, 2.8051e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 250.51, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4648 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.11 lr 0.00010000 +Epoch 275, weight, value: tensor([[ 0.0768, -0.1324, -0.1392, ..., -0.1250, -0.1519, -0.1440], + [ 0.0378, 0.0511, -0.0387, ..., -0.0547, 0.1597, 0.0178], + [-0.0261, -0.1603, -0.0106, ..., 0.2002, -0.1625, 0.0414], + ..., + [-0.1483, 0.1383, 0.0270, ..., -0.1484, 0.1098, 0.0577], + [ 0.0020, -0.0798, 0.0020, ..., -0.1030, -0.1705, -0.2075], + [ 0.0217, -0.0430, -0.0213, ..., -0.2012, -0.1606, 0.0283]], + device='cuda:0'), grad: tensor([[ 8.1491e-10, 1.1991e-08, 0.0000e+00, ..., 2.6426e-08, + 1.2456e-08, 1.2689e-08], + [-2.7707e-08, 6.5519e-07, 0.0000e+00, ..., 2.6077e-08, + 6.6496e-07, 6.2073e-07], + [ 7.4506e-09, 1.2142e-07, 0.0000e+00, ..., 2.4447e-08, + 1.4447e-07, 1.4342e-07], + ..., + [ 5.4715e-09, -1.3653e-06, 0.0000e+00, ..., -8.4634e-08, + -1.4110e-06, -1.4007e-06], + [ 8.9640e-09, 3.0501e-08, 0.0000e+00, ..., 5.5064e-08, + 4.3423e-08, 3.4226e-08], + [ 8.1491e-10, 3.8417e-07, 0.0000e+00, ..., 8.6147e-09, + 3.5902e-07, 3.9255e-07]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0405, 0.0136, 0.0155, 0.0298, 0.0300, -0.0183, 0.0274, 0.0306, + -0.0100, 0.0024], device='cuda:0'), grad: tensor([ 9.3365e-08, 1.5516e-06, 3.7020e-07, 1.6904e-07, 3.4319e-07, + 3.3993e-07, -6.3423e-07, -3.4813e-06, 2.4028e-07, 1.0030e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 250.66, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4860 re_mapping 0.0030 re_causal 0.0100 /// teacc 99.09 lr 0.00010000 +Epoch 276, weight, value: tensor([[ 0.0768, -0.1333, -0.1392, ..., -0.1251, -0.1527, -0.1444], + [ 0.0378, 0.0511, -0.0388, ..., -0.0547, 0.1597, 0.0177], + [-0.0265, -0.1603, -0.0108, ..., 0.2002, -0.1630, 0.0414], + ..., + [-0.1487, 0.1385, 0.0271, ..., -0.1484, 0.1100, 0.0578], + [ 0.0020, -0.0798, 0.0021, ..., -0.1036, -0.1705, -0.2077], + [ 0.0218, -0.0430, -0.0213, ..., -0.2014, -0.1607, 0.0283]], + device='cuda:0'), grad: tensor([[-1.5891e-07, 5.1223e-09, 0.0000e+00, ..., 4.6566e-10, + 8.4983e-09, 3.4925e-09], + [-9.3947e-08, -5.3551e-08, 1.1642e-10, ..., 4.4238e-09, + -2.2631e-07, -3.9116e-08], + [ 3.5740e-08, 2.4098e-08, 0.0000e+00, ..., 1.2806e-09, + 4.6100e-08, 1.3271e-08], + ..., + [ 4.4005e-08, 1.0477e-08, 8.1491e-10, ..., 5.3551e-09, + 3.8301e-08, 1.4901e-08], + [-1.4435e-08, -7.4389e-08, 1.1642e-10, ..., 2.3283e-09, + 6.5891e-08, 2.2002e-08], + [ 7.1013e-08, 2.7358e-08, -1.7462e-09, ..., 2.0140e-07, + 2.0722e-08, 3.2037e-07]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0406, 0.0136, 0.0155, 0.0298, 0.0300, -0.0184, 0.0278, 0.0306, + -0.0100, 0.0024], device='cuda:0'), grad: tensor([-6.4727e-07, -6.4960e-08, 2.5635e-07, 5.0385e-07, -9.4716e-07, + 1.8033e-07, 1.7427e-07, 3.1153e-07, -1.0412e-06, 1.2796e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 250.84, cls_loss 0.0009 cls_loss_mapping 0.0020 cls_loss_causal 0.4545 re_mapping 0.0029 re_causal 0.0091 /// teacc 99.12 lr 0.00010000 +Epoch 277, weight, value: tensor([[ 0.0769, -0.1339, -0.1392, ..., -0.1257, -0.1531, -0.1447], + [ 0.0378, 0.0510, -0.0390, ..., -0.0547, 0.1597, 0.0177], + [-0.0270, -0.1604, -0.0110, ..., 0.2002, -0.1635, 0.0413], + ..., + [-0.1492, 0.1387, 0.0273, ..., -0.1485, 0.1103, 0.0578], + [ 0.0020, -0.0798, 0.0020, ..., -0.1040, -0.1706, -0.2082], + [ 0.0218, -0.0432, -0.0216, ..., -0.2015, -0.1609, 0.0282]], + device='cuda:0'), grad: tensor([[-4.5169e-08, 1.9791e-09, 8.1491e-10, ..., 2.0955e-09, + 3.0268e-09, 7.3342e-09], + [-3.3528e-08, 3.7020e-07, 2.1828e-07, ..., 5.4715e-09, + 2.8266e-07, 3.2107e-07], + [ 1.9791e-08, 3.6787e-08, 1.5716e-08, ..., 3.0268e-09, + 5.7276e-08, 4.3772e-08], + ..., + [ 4.0047e-08, -1.5525e-06, -8.7079e-07, ..., 6.2864e-09, + -1.5581e-06, -1.3290e-06], + [ 3.2596e-09, 1.1851e-07, 6.3563e-08, ..., 1.1991e-08, + 1.4680e-07, 1.4692e-07], + [-5.6345e-08, 6.6264e-07, 3.7090e-07, ..., 1.5041e-07, + 6.8359e-07, 6.6822e-07]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0410, 0.0136, 0.0154, 0.0295, 0.0301, -0.0184, 0.0284, 0.0307, + -0.0100, 0.0023], device='cuda:0'), grad: tensor([-2.8266e-07, 1.2955e-06, 2.7195e-07, 6.7009e-07, -7.8650e-07, + 3.6275e-07, 1.9849e-07, -5.0664e-06, 6.1980e-07, 2.7325e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 250.62, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4684 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.11 lr 0.00010000 +Epoch 278, weight, value: tensor([[ 0.0776, -0.1341, -0.1392, ..., -0.1251, -0.1534, -0.1449], + [ 0.0378, 0.0511, -0.0392, ..., -0.0547, 0.1598, 0.0177], + [-0.0275, -0.1604, -0.0110, ..., 0.2002, -0.1638, 0.0413], + ..., + [-0.1498, 0.1387, 0.0274, ..., -0.1485, 0.1103, 0.0578], + [ 0.0020, -0.0798, 0.0020, ..., -0.1037, -0.1706, -0.2084], + [ 0.0219, -0.0432, -0.0220, ..., -0.2018, -0.1609, 0.0281]], + device='cuda:0'), grad: tensor([[-1.7462e-09, 4.6566e-10, 1.1642e-10, ..., 3.4925e-10, + 9.3132e-10, 5.8208e-10], + [-1.3155e-08, 4.1910e-08, 6.7521e-09, ..., 3.3760e-09, + 1.6298e-08, 3.4925e-08], + [ 1.7462e-09, 1.0896e-07, 1.6065e-08, ..., -4.7730e-09, + 1.0617e-07, 8.7311e-08], + ..., + [ 3.6089e-09, -1.8219e-07, -2.7241e-08, ..., 9.3132e-09, + -1.6799e-07, -1.3877e-07], + [ 5.4715e-09, 4.8894e-09, 4.6566e-10, ..., 1.8626e-09, + 1.4901e-08, 6.7521e-09], + [ 1.1642e-09, 7.5670e-09, 1.1642e-09, ..., 5.3551e-09, + 8.6147e-09, 1.3388e-08]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0399, 0.0136, 0.0154, 0.0293, 0.0302, -0.0183, 0.0274, 0.0307, + -0.0100, 0.0022], device='cuda:0'), grad: tensor([-1.8859e-08, 6.2748e-08, 5.0478e-07, -3.7975e-07, -3.0268e-08, + 1.1059e-08, 7.3342e-09, -2.0640e-07, 2.1653e-08, 3.6904e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 250.53, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4848 re_mapping 0.0029 re_causal 0.0099 /// teacc 99.01 lr 0.00010000 +Epoch 279, weight, value: tensor([[ 0.0772, -0.1342, -0.1392, ..., -0.1248, -0.1532, -0.1466], + [ 0.0378, 0.0511, -0.0364, ..., -0.0548, 0.1599, 0.0178], + [-0.0277, -0.1604, -0.0109, ..., 0.2003, -0.1640, 0.0414], + ..., + [-0.1516, 0.1387, 0.0260, ..., -0.1485, 0.1102, 0.0578], + [ 0.0020, -0.0798, 0.0016, ..., -0.1052, -0.1706, -0.2087], + [ 0.0227, -0.0432, -0.0221, ..., -0.2018, -0.1610, 0.0281]], + device='cuda:0'), grad: tensor([[-2.1188e-08, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, + 1.1642e-10, 6.9849e-10], + [ 1.3467e-06, 8.1491e-10, 0.0000e+00, ..., 2.4796e-08, + 4.1910e-09, 3.7719e-08], + [ 2.2701e-08, 2.9104e-09, 0.0000e+00, ..., 7.9162e-09, + 4.1910e-09, 2.5611e-09], + ..., + [ 2.5961e-08, -9.7789e-09, 0.0000e+00, ..., 5.9372e-09, + -1.0361e-08, 1.7812e-08], + [-1.4817e-06, 6.9849e-10, 0.0000e+00, ..., 5.1223e-09, + -2.3283e-09, 9.7789e-09], + [ 6.8219e-08, 1.1642e-09, 0.0000e+00, ..., 2.4447e-08, + 1.5134e-09, -5.3085e-08]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0403, 0.0136, 0.0154, 0.0290, 0.0303, -0.0181, 0.0270, 0.0306, + -0.0100, 0.0023], device='cuda:0'), grad: tensor([-6.5658e-08, 7.7635e-06, 4.6100e-07, -3.7486e-07, -8.9407e-08, + -2.7008e-07, 2.0361e-07, 3.1898e-07, -8.2180e-06, 2.6077e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 250.73, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4621 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.06 lr 0.00010000 +Epoch 280, weight, value: tensor([[ 0.0770, -0.1343, -0.1392, ..., -0.1250, -0.1530, -0.1476], + [ 0.0379, 0.0511, -0.0346, ..., -0.0549, 0.1600, 0.0182], + [-0.0280, -0.1604, -0.0112, ..., 0.2010, -0.1642, 0.0416], + ..., + [-0.1527, 0.1387, 0.0251, ..., -0.1486, 0.1098, 0.0577], + [ 0.0021, -0.0798, 0.0014, ..., -0.1057, -0.1706, -0.2088], + [ 0.0210, -0.0432, -0.0232, ..., -0.2029, -0.1612, 0.0279]], + device='cuda:0'), grad: tensor([[-2.1560e-07, 1.1642e-10, 0.0000e+00, ..., 8.1491e-10, + 1.1642e-09, 1.7462e-09], + [-2.0955e-08, 0.0000e+00, 0.0000e+00, ..., 2.9802e-08, + -5.6694e-08, 3.3760e-09], + [ 2.6193e-08, 4.1910e-09, 0.0000e+00, ..., -1.0966e-07, + -2.7707e-08, -6.3679e-08], + ..., + [ 2.3865e-08, -4.0745e-09, 0.0000e+00, ..., 3.0966e-08, + 1.9791e-08, 6.1234e-08], + [ 4.4354e-08, 1.1642e-09, 0.0000e+00, ..., 4.5402e-08, + 4.8545e-08, 5.8440e-08], + [ 2.3283e-08, -1.3970e-09, 0.0000e+00, ..., 3.4925e-10, + 3.6089e-09, -8.4052e-08]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0408, 0.0137, 0.0157, 0.0289, 0.0303, -0.0178, 0.0260, 0.0305, + -0.0098, 0.0020], device='cuda:0'), grad: tensor([-1.3364e-06, 4.4354e-08, -4.4121e-08, 3.6182e-07, 1.0745e-07, + 2.4214e-07, 1.0605e-07, 3.3085e-07, 4.0163e-07, -1.9523e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 250.41, cls_loss 0.0008 cls_loss_mapping 0.0016 cls_loss_causal 0.4770 re_mapping 0.0030 re_causal 0.0099 /// teacc 99.07 lr 0.00010000 +Epoch 281, weight, value: tensor([[ 0.0772, -0.1347, -0.1392, ..., -0.1250, -0.1531, -0.1478], + [ 0.0379, 0.0511, -0.0346, ..., -0.0550, 0.1601, 0.0182], + [-0.0284, -0.1604, -0.0112, ..., 0.2013, -0.1645, 0.0417], + ..., + [-0.1532, 0.1387, 0.0252, ..., -0.1489, 0.1099, 0.0576], + [ 0.0021, -0.0799, 0.0014, ..., -0.1061, -0.1706, -0.2092], + [ 0.0211, -0.0433, -0.0236, ..., -0.2040, -0.1613, 0.0275]], + device='cuda:0'), grad: tensor([[-2.8755e-08, 8.1491e-10, 0.0000e+00, ..., 1.6298e-08, + 3.2596e-09, 1.9441e-08], + [-5.0059e-08, -3.2363e-08, 1.6298e-09, ..., 1.1176e-07, + -1.4203e-07, 8.9174e-08], + [ 1.4901e-08, 3.2131e-08, 6.4028e-09, ..., -7.3155e-07, + 6.3330e-08, -4.4075e-07], + ..., + [ 1.9791e-08, -2.7590e-08, -9.1968e-09, ..., 8.6264e-08, + -5.4715e-09, 9.8161e-07], + [ 1.0477e-08, 9.8953e-09, 1.1642e-10, ..., 4.3586e-07, + 3.4692e-08, 2.9779e-07], + [ 1.2689e-08, 6.4028e-09, 2.3283e-10, ..., 5.1223e-08, + 1.3853e-08, -1.1949e-06]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0406, 0.0137, 0.0158, 0.0281, 0.0306, -0.0173, 0.0249, 0.0304, + -0.0099, 0.0017], device='cuda:0'), grad: tensor([-1.0419e-07, 1.5693e-07, -1.5385e-06, 1.2375e-07, 4.5728e-07, + -3.9302e-07, 5.4110e-07, 2.5891e-06, 1.0729e-06, -2.8964e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 250.45, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4893 re_mapping 0.0029 re_causal 0.0094 /// teacc 99.10 lr 0.00010000 +Epoch 282, weight, value: tensor([[ 0.0777, -0.1349, -0.1392, ..., -0.1249, -0.1532, -0.1479], + [ 0.0379, 0.0511, -0.0343, ..., -0.0553, 0.1602, 0.0179], + [-0.0291, -0.1604, -0.0115, ..., 0.2013, -0.1651, 0.0417], + ..., + [-0.1546, 0.1387, 0.0251, ..., -0.1489, 0.1099, 0.0576], + [ 0.0021, -0.0799, 0.0017, ..., -0.1065, -0.1707, -0.2095], + [ 0.0209, -0.0435, -0.0237, ..., -0.2044, -0.1614, 0.0273]], + device='cuda:0'), grad: tensor([[-1.6298e-09, 1.5018e-08, 3.2596e-09, ..., 1.3970e-09, + 2.5029e-08, 1.7113e-08], + [-8.3586e-08, 2.0186e-07, 7.0198e-08, ..., 2.6776e-09, + 2.5658e-07, 1.5378e-07], + [ 1.5716e-08, 5.0943e-07, 1.4447e-07, ..., -1.5134e-09, + 9.0851e-07, 4.9360e-07], + ..., + [ 2.6193e-08, -1.6596e-06, -5.6392e-07, ..., 1.0477e-09, + -3.3081e-06, -1.7686e-06], + [-1.1176e-08, 3.0850e-07, 9.7090e-08, ..., 5.4715e-09, + 5.9092e-07, 3.0803e-07], + [-1.2806e-09, 5.8254e-07, 1.3073e-07, ..., 2.0606e-08, + 8.8289e-07, 4.5076e-07]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0403, 0.0136, 0.0157, 0.0285, 0.0311, -0.0175, 0.0217, 0.0304, + -0.0098, 0.0016], device='cuda:0'), grad: tensor([ 5.4948e-08, 5.9512e-07, 1.9763e-06, -2.4568e-06, 6.4261e-07, + 8.0187e-07, -3.3178e-08, -5.0217e-06, 1.1437e-06, 2.2966e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 250.80, cls_loss 0.0009 cls_loss_mapping 0.0015 cls_loss_causal 0.4933 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.07 lr 0.00010000 +Epoch 283, weight, value: tensor([[ 0.0785, -0.1359, -0.1392, ..., -0.1247, -0.1528, -0.1480], + [ 0.0379, 0.0511, -0.0344, ..., -0.0554, 0.1602, 0.0178], + [-0.0289, -0.1605, -0.0090, ..., 0.2014, -0.1658, 0.0417], + ..., + [-0.1556, 0.1404, 0.0252, ..., -0.1489, 0.1113, 0.0586], + [ 0.0021, -0.0799, -0.0009, ..., -0.1093, -0.1707, -0.2099], + [ 0.0215, -0.0434, -0.0239, ..., -0.2045, -0.1615, 0.0274]], + device='cuda:0'), grad: tensor([[ 1.4994e-07, 1.1642e-10, 0.0000e+00, ..., 9.5647e-07, + 5.2387e-10, 2.2119e-09], + [-7.2177e-09, -6.9267e-09, 0.0000e+00, ..., 4.1502e-08, + -3.1374e-08, 1.8626e-09], + [ 7.9744e-09, 3.8999e-09, 0.0000e+00, ..., -4.3481e-08, + 7.5088e-09, -3.0675e-08], + ..., + [ 9.1386e-09, -6.8685e-09, 0.0000e+00, ..., 2.9511e-08, + 3.4925e-10, 2.3574e-08], + [-1.3004e-07, 3.3760e-09, 0.0000e+00, ..., 6.2049e-08, + 1.4668e-08, 1.4668e-08], + [ 1.1059e-09, 8.7311e-10, 0.0000e+00, ..., 1.4668e-07, + 1.9209e-09, 1.3073e-07]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0400, 0.0136, 0.0157, 0.0286, 0.0305, -0.0176, 0.0220, 0.0311, + -0.0099, 0.0016], device='cuda:0'), grad: tensor([ 2.9113e-06, 1.2899e-07, -3.7078e-08, 1.3225e-07, -7.4087e-07, + 2.0750e-06, -4.8615e-06, 1.3574e-07, -4.5239e-07, 7.1619e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 250.81, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4713 re_mapping 0.0028 re_causal 0.0095 /// teacc 99.12 lr 0.00010000 +Epoch 284, weight, value: tensor([[ 0.0788, -0.1361, -0.1392, ..., -0.1243, -0.1529, -0.1481], + [ 0.0379, 0.0511, -0.0344, ..., -0.0556, 0.1602, 0.0177], + [-0.0288, -0.1608, -0.0090, ..., 0.2016, -0.1664, 0.0415], + ..., + [-0.1565, 0.1406, 0.0252, ..., -0.1490, 0.1123, 0.0589], + [ 0.0021, -0.0799, -0.0009, ..., -0.1103, -0.1707, -0.2106], + [ 0.0216, -0.0434, -0.0239, ..., -0.2046, -0.1615, 0.0274]], + device='cuda:0'), grad: tensor([[-6.1351e-08, 4.7148e-09, 0.0000e+00, ..., -4.4878e-08, + 4.5984e-09, 7.4506e-09], + [-2.6845e-07, 6.5134e-08, 0.0000e+00, ..., 5.3202e-08, + -2.3341e-07, 6.2573e-08], + [ 5.2620e-08, 1.0978e-07, 0.0000e+00, ..., -6.9849e-09, + 1.4994e-07, 1.1327e-07], + ..., + [ 9.5577e-08, -5.6764e-07, 0.0000e+00, ..., 1.5425e-08, + -5.2527e-07, -6.6264e-07], + [ 1.1118e-07, 5.5355e-08, 0.0000e+00, ..., 2.8289e-08, + 1.6647e-07, 9.0338e-08], + [ 2.0547e-08, 1.7881e-07, 0.0000e+00, ..., 3.0664e-07, + 2.0675e-07, 3.0827e-07]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0396, 0.0136, 0.0154, 0.0282, 0.0305, -0.0176, 0.0221, 0.0315, + -0.0099, 0.0016], device='cuda:0'), grad: tensor([-2.9942e-07, -2.8871e-08, 5.8115e-07, 4.3353e-07, -1.4165e-06, + -2.6287e-07, 4.3493e-07, -1.6904e-06, 5.9279e-07, 1.6624e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 250.42, cls_loss 0.0010 cls_loss_mapping 0.0021 cls_loss_causal 0.4724 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.04 lr 0.00010000 +Epoch 285, weight, value: tensor([[ 0.0793, -0.1362, -0.1392, ..., -0.1243, -0.1530, -0.1481], + [ 0.0379, 0.0510, -0.0345, ..., -0.0560, 0.1602, 0.0175], + [-0.0289, -0.1609, -0.0091, ..., 0.2017, -0.1666, 0.0414], + ..., + [-0.1578, 0.1395, 0.0253, ..., -0.1505, 0.1114, 0.0579], + [ 0.0021, -0.0799, -0.0009, ..., -0.1117, -0.1708, -0.2110], + [ 0.0217, -0.0436, -0.0239, ..., -0.2050, -0.1616, 0.0273]], + device='cuda:0'), grad: tensor([[-2.2119e-09, 1.1642e-10, 0.0000e+00, ..., 1.6065e-08, + 1.3970e-09, 8.1491e-10], + [ 1.0245e-07, -1.6065e-08, 0.0000e+00, ..., 1.6240e-07, + -6.4611e-08, -2.7241e-08], + [ 1.2210e-06, 2.0955e-09, 0.0000e+00, ..., 1.4789e-06, + 7.3342e-09, 2.2119e-09], + ..., + [ 2.3632e-08, 4.4238e-09, 0.0000e+00, ..., 1.8044e-08, + 1.7113e-08, 8.2655e-09], + [-1.5795e-06, 3.9581e-09, 0.0000e+00, ..., -1.9353e-06, + 1.4668e-08, 6.8685e-09], + [ 1.1933e-07, 6.9849e-10, 0.0000e+00, ..., 1.3586e-07, + 4.4238e-09, 5.8208e-10]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0393, 0.0135, 0.0154, 0.0305, 0.0313, -0.0194, 0.0227, 0.0307, + -0.0099, 0.0015], device='cuda:0'), grad: tensor([ 2.5262e-08, 9.3179e-07, 9.5814e-06, 1.7544e-07, 9.1735e-08, + 4.9244e-08, 5.7090e-07, 1.5390e-07, -1.2450e-05, 9.1037e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 250.46, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4509 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +Epoch 286, weight, value: tensor([[ 0.0795, -0.1373, -0.1392, ..., -0.1244, -0.1537, -0.1483], + [ 0.0379, 0.0510, -0.0345, ..., -0.0561, 0.1602, 0.0174], + [-0.0293, -0.1610, -0.0091, ..., 0.2017, -0.1668, 0.0414], + ..., + [-0.1583, 0.1407, 0.0253, ..., -0.1507, 0.1121, 0.0593], + [ 0.0021, -0.0799, -0.0009, ..., -0.1119, -0.1708, -0.2112], + [ 0.0217, -0.0438, -0.0239, ..., -0.2050, -0.1618, 0.0273]], + device='cuda:0'), grad: tensor([[ 2.4447e-09, 2.3283e-10, 0.0000e+00, ..., 7.1013e-09, + 5.8208e-10, 4.7730e-09], + [-6.7521e-09, -1.8626e-09, 0.0000e+00, ..., 5.1339e-08, + -2.7474e-08, 2.5728e-08], + [ 5.4715e-09, 4.8894e-09, 0.0000e+00, ..., -1.3667e-07, + 8.8476e-09, -7.6718e-08], + ..., + [ 1.9674e-08, -7.5670e-09, 0.0000e+00, ..., 2.4447e-09, + 4.4238e-09, 3.1432e-09], + [-3.2247e-08, 2.6776e-09, 0.0000e+00, ..., 2.7707e-08, + 6.8685e-09, 2.2585e-08], + [ 5.8208e-09, -1.6298e-09, 0.0000e+00, ..., 4.7730e-09, + 1.1642e-09, -1.4552e-08]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0392, 0.0135, 0.0153, 0.0303, 0.0305, -0.0193, 0.0227, 0.0317, + -0.0099, 0.0015], device='cuda:0'), grad: tensor([ 2.4796e-08, 1.2107e-07, -2.8708e-07, 2.2538e-07, 5.6811e-08, + -1.4226e-07, 8.3819e-08, 8.5915e-08, -1.6124e-07, 6.9849e-10], + device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 250.46, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4766 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.09 lr 0.00010000 +Epoch 287, weight, value: tensor([[ 0.0797, -0.1380, -0.1392, ..., -0.1242, -0.1542, -0.1484], + [ 0.0379, 0.0510, -0.0346, ..., -0.0559, 0.1604, 0.0175], + [-0.0297, -0.1611, -0.0091, ..., 0.2017, -0.1671, 0.0414], + ..., + [-0.1595, 0.1408, 0.0254, ..., -0.1506, 0.1124, 0.0594], + [ 0.0021, -0.0799, -0.0010, ..., -0.1142, -0.1708, -0.2117], + [ 0.0218, -0.0440, -0.0240, ..., -0.2053, -0.1620, 0.0272]], + device='cuda:0'), grad: tensor([[-1.3970e-09, 1.6880e-08, 0.0000e+00, ..., 4.9244e-08, + 2.3632e-08, 1.6764e-08], + [ 2.7940e-09, 4.7381e-08, 0.0000e+00, ..., 4.6450e-08, + 3.8883e-08, 8.7777e-08], + [ 9.5461e-09, 1.7579e-08, 0.0000e+00, ..., 2.0955e-08, + 2.7707e-08, 2.0722e-08], + ..., + [ 5.4715e-09, -2.1490e-07, 0.0000e+00, ..., 1.0245e-08, + -2.9011e-07, -1.7777e-07], + [ 5.5879e-09, 7.6834e-09, 0.0000e+00, ..., 3.5507e-08, + 1.2689e-08, 1.0477e-08], + [ 5.5879e-09, 6.3097e-08, 0.0000e+00, ..., 1.4319e-08, + 8.6613e-08, 7.9628e-08]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0390, 0.0136, 0.0152, 0.0302, 0.0305, -0.0193, 0.0221, 0.0318, + -0.0100, 0.0015], device='cuda:0'), grad: tensor([ 2.1630e-07, 3.8464e-07, 2.5937e-07, -4.4610e-07, -1.4482e-07, + 1.6326e-06, -2.0396e-06, -5.1036e-07, 3.2596e-07, 3.3039e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 250.09, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4756 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.09 lr 0.00010000 +Epoch 288, weight, value: tensor([[ 0.0798, -0.1378, -0.1392, ..., -0.1243, -0.1541, -0.1486], + [ 0.0379, 0.0510, -0.0346, ..., -0.0561, 0.1605, 0.0176], + [-0.0299, -0.1611, -0.0091, ..., 0.2018, -0.1671, 0.0414], + ..., + [-0.1605, 0.1408, 0.0254, ..., -0.1506, 0.1124, 0.0594], + [ 0.0021, -0.0799, -0.0010, ..., -0.1149, -0.1708, -0.2120], + [ 0.0220, -0.0441, -0.0240, ..., -0.2054, -0.1620, 0.0272]], + device='cuda:0'), grad: tensor([[-1.4482e-07, 6.9849e-10, 0.0000e+00, ..., 2.6077e-08, + -5.5414e-08, 3.6322e-08], + [-2.0154e-06, -3.4925e-10, 0.0000e+00, ..., -1.0999e-06, + -1.5311e-06, -1.5460e-06], + [ 1.0775e-06, 2.0838e-08, 0.0000e+00, ..., 6.8336e-08, + 8.1956e-07, 5.8860e-07], + ..., + [ 3.0850e-08, -7.5204e-08, 0.0000e+00, ..., 2.6426e-08, + -4.9942e-08, -3.9581e-08], + [ 1.9139e-07, 8.1491e-09, 0.0000e+00, ..., 7.1595e-08, + 1.2433e-07, 1.1059e-07], + [ 4.7032e-08, 2.5611e-08, 0.0000e+00, ..., 1.7369e-07, + 4.4820e-08, 5.4063e-07]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0389, 0.0136, 0.0152, 0.0302, 0.0305, -0.0192, 0.0221, 0.0318, + -0.0100, 0.0015], device='cuda:0'), grad: tensor([-6.5658e-07, -6.2138e-06, 1.8347e-06, 7.4096e-06, -1.2619e-06, + -5.3011e-06, 1.1912e-06, 2.5472e-07, 8.5915e-07, 1.8999e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 250.38, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4723 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.14 lr 0.00010000 +Epoch 289, weight, value: tensor([[ 0.0807, -0.1368, -0.1393, ..., -0.1242, -0.1544, -0.1489], + [ 0.0379, 0.0510, -0.0346, ..., -0.0562, 0.1606, 0.0177], + [-0.0318, -0.1613, -0.0094, ..., 0.2018, -0.1676, 0.0414], + ..., + [-0.1619, 0.1410, 0.0254, ..., -0.1506, 0.1125, 0.0594], + [ 0.0021, -0.0800, -0.0012, ..., -0.1142, -0.1708, -0.2123], + [ 0.0223, -0.0446, -0.0240, ..., -0.2058, -0.1625, 0.0270]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.4925e-09, 0.0000e+00, ..., 1.1642e-10, + 6.7521e-09, 3.8417e-09], + [-1.0403e-06, -1.2983e-06, 0.0000e+00, ..., 8.1491e-10, + -2.4810e-06, -1.0785e-06], + [ 4.2026e-08, 3.8184e-08, 0.0000e+00, ..., 1.1642e-10, + 9.8371e-08, 4.2142e-08], + ..., + [ 8.8615e-07, 1.0896e-06, 1.1642e-10, ..., 9.3132e-10, + 2.0638e-06, 8.9640e-07], + [ 1.3213e-07, 8.4750e-08, 0.0000e+00, ..., 4.6566e-10, + 1.6578e-07, 7.5670e-08], + [ 2.0955e-09, 2.9104e-08, -3.4925e-10, ..., 1.7462e-09, + 5.0990e-08, 6.2864e-09]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0375, 0.0136, 0.0152, 0.0300, 0.0306, -0.0192, 0.0216, 0.0318, + -0.0099, 0.0013], device='cuda:0'), grad: tensor([ 1.9441e-08, -4.5411e-06, 1.9511e-07, 2.3365e-07, 2.0547e-07, + -5.2899e-07, 8.2422e-08, 3.8408e-06, 6.0722e-07, -1.2119e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 250.35, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4702 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.04 lr 0.00010000 +Epoch 290, weight, value: tensor([[ 0.0808, -0.1385, -0.1393, ..., -0.1244, -0.1555, -0.1491], + [ 0.0380, 0.0510, -0.0342, ..., -0.0562, 0.1608, 0.0178], + [-0.0328, -0.1614, -0.0094, ..., 0.2017, -0.1680, 0.0414], + ..., + [-0.1647, 0.1414, 0.0251, ..., -0.1506, 0.1135, 0.0596], + [ 0.0021, -0.0800, -0.0013, ..., -0.1147, -0.1709, -0.2127], + [ 0.0227, -0.0449, -0.0239, ..., -0.2059, -0.1629, 0.0270]], + device='cuda:0'), grad: tensor([[-5.5647e-08, 1.5134e-09, 0.0000e+00, ..., 8.1491e-10, + 3.0268e-09, 1.7462e-09], + [-7.2410e-08, -6.4145e-08, 0.0000e+00, ..., 2.7940e-09, + -2.1444e-07, -7.3807e-08], + [ 1.2456e-08, 1.6065e-08, 0.0000e+00, ..., -1.0827e-08, + 3.1083e-08, 5.8208e-09], + ..., + [ 3.2946e-08, -6.9849e-09, 0.0000e+00, ..., 2.0955e-09, + 6.0070e-08, 1.2806e-08], + [ 1.5600e-08, 1.4319e-08, 0.0000e+00, ..., 1.6298e-09, + 3.2014e-08, 1.3039e-08], + [ 1.6880e-08, 5.9372e-09, 0.0000e+00, ..., 3.3760e-09, + 1.1292e-08, 7.3342e-09]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0375, 0.0136, 0.0151, 0.0284, 0.0306, -0.0190, 0.0216, 0.0321, + -0.0099, 0.0013], device='cuda:0'), grad: tensor([-3.4738e-07, -3.1758e-07, 4.9826e-08, 1.1199e-07, 9.3016e-08, + 6.1933e-08, 5.0524e-08, 1.0419e-07, 8.7079e-08, 1.1583e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 250.71, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4604 re_mapping 0.0029 re_causal 0.0095 /// teacc 98.98 lr 0.00010000 +Epoch 291, weight, value: tensor([[ 0.0807, -0.1401, -0.1393, ..., -0.1246, -0.1564, -0.1492], + [ 0.0380, 0.0510, -0.0342, ..., -0.0564, 0.1609, 0.0179], + [-0.0334, -0.1615, -0.0094, ..., 0.2019, -0.1681, 0.0414], + ..., + [-0.1668, 0.1415, 0.0251, ..., -0.1506, 0.1136, 0.0595], + [ 0.0021, -0.0801, -0.0013, ..., -0.1177, -0.1709, -0.2135], + [ 0.0228, -0.0444, -0.0239, ..., -0.2060, -0.1626, 0.0271]], + device='cuda:0'), grad: tensor([[ 3.9581e-09, 3.8417e-09, 0.0000e+00, ..., 1.5832e-08, + 8.7311e-09, 3.9581e-09], + [-2.4680e-08, -2.3516e-08, 0.0000e+00, ..., 1.5250e-08, + -7.2177e-08, -9.5461e-09], + [ 8.1491e-09, 9.4296e-09, 0.0000e+00, ..., -1.1991e-08, + 1.8743e-08, -8.0327e-09], + ..., + [ 1.5716e-08, -4.4238e-09, 0.0000e+00, ..., 2.2002e-08, + 1.8394e-08, 2.3050e-08], + [ 4.5402e-09, 5.5879e-09, 0.0000e+00, ..., 2.1071e-08, + 1.0245e-08, 6.8685e-09], + [ 2.4447e-09, 5.5879e-09, 0.0000e+00, ..., 1.0943e-08, + 5.3551e-09, 1.2806e-08]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0377, 0.0137, 0.0151, 0.0283, 0.0307, -0.0190, 0.0227, 0.0320, + -0.0101, 0.0014], device='cuda:0'), grad: tensor([ 7.6368e-08, -2.9569e-08, 3.0734e-08, -7.2061e-08, -1.0489e-07, + 7.2345e-06, -7.5139e-06, 1.3865e-07, 1.4016e-07, 7.8115e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 250.60, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.5107 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.03 lr 0.00010000 +Epoch 292, weight, value: tensor([[ 0.0807, -0.1416, -0.1393, ..., -0.1252, -0.1571, -0.1495], + [ 0.0380, 0.0510, -0.0342, ..., -0.0571, 0.1609, 0.0176], + [-0.0335, -0.1615, -0.0094, ..., 0.2024, -0.1677, 0.0415], + ..., + [-0.1691, 0.1416, 0.0251, ..., -0.1509, 0.1137, 0.0594], + [ 0.0021, -0.0801, -0.0013, ..., -0.1182, -0.1710, -0.2141], + [ 0.0228, -0.0438, -0.0237, ..., -0.2065, -0.1630, 0.0272]], + device='cuda:0'), grad: tensor([[-3.3062e-08, -6.9849e-09, 0.0000e+00, ..., 2.8173e-08, + 5.3551e-09, 2.1304e-08], + [-5.2154e-08, -5.9372e-08, 0.0000e+00, ..., 8.8126e-08, + -1.3877e-07, 4.6333e-08], + [ 1.9674e-08, 5.1223e-09, 0.0000e+00, ..., -1.6019e-07, + -1.3493e-07, -1.5949e-07], + ..., + [ 3.0966e-08, 2.9453e-08, 0.0000e+00, ..., 8.8243e-08, + 1.2363e-07, 1.5658e-07], + [ 3.5390e-08, 2.2235e-08, 0.0000e+00, ..., 3.6787e-08, + 7.4506e-08, 4.1211e-08], + [ 4.7730e-09, 1.9791e-09, 0.0000e+00, ..., 2.7660e-07, + 7.1013e-09, 9.5041e-07]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0380, 0.0136, 0.0152, 0.0280, 0.0307, -0.0193, 0.0238, 0.0318, + -0.0101, 0.0015], device='cuda:0'), grad: tensor([-2.8173e-08, 2.1840e-07, -7.0361e-07, 3.4226e-07, -4.0643e-06, + -1.8615e-07, 2.2585e-08, 6.8545e-07, 3.2224e-07, 3.3807e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 250.49, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4610 re_mapping 0.0028 re_causal 0.0095 /// teacc 99.05 lr 0.00010000 +Epoch 293, weight, value: tensor([[ 0.0812, -0.1424, -0.1393, ..., -0.1254, -0.1572, -0.1497], + [ 0.0381, 0.0510, -0.0342, ..., -0.0575, 0.1610, 0.0177], + [-0.0338, -0.1637, -0.0094, ..., 0.2012, -0.1702, 0.0405], + ..., + [-0.1721, 0.1427, 0.0251, ..., -0.1499, 0.1150, 0.0600], + [ 0.0022, -0.0802, -0.0013, ..., -0.1182, -0.1710, -0.2144], + [ 0.0227, -0.0439, -0.0237, ..., -0.2068, -0.1632, 0.0271]], + device='cuda:0'), grad: tensor([[ 4.2026e-08, 3.4925e-10, 0.0000e+00, ..., 3.4925e-10, + 5.8208e-10, 1.0477e-09], + [-2.0023e-08, -3.8417e-08, 0.0000e+00, ..., 2.3283e-09, + -9.5228e-08, -4.3889e-08], + [ 2.3516e-08, 2.7940e-09, 0.0000e+00, ..., -6.5193e-09, + 5.8208e-09, -4.6566e-10], + ..., + [ 1.1828e-07, 2.6659e-08, 0.0000e+00, ..., 1.9791e-09, + 6.4261e-08, 3.3993e-08], + [ 4.7032e-08, 6.5193e-09, 0.0000e+00, ..., 9.3132e-10, + 1.3970e-08, 7.7998e-09], + [ 1.3970e-08, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, + 1.6298e-09, -3.1432e-09]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0376, 0.0136, 0.0143, 0.0280, 0.0308, -0.0194, 0.0237, 0.0325, + -0.0100, 0.0015], device='cuda:0'), grad: tensor([ 2.4680e-07, -1.0803e-07, 1.2689e-07, 1.1232e-06, 6.7404e-08, + -2.5742e-06, 7.3458e-08, 6.9663e-07, 2.8685e-07, 7.9977e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 250.34, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4660 re_mapping 0.0028 re_causal 0.0095 /// teacc 99.06 lr 0.00010000 +Epoch 294, weight, value: tensor([[ 0.0814, -0.1427, -0.1393, ..., -0.1255, -0.1574, -0.1499], + [ 0.0381, 0.0510, -0.0342, ..., -0.0576, 0.1611, 0.0177], + [-0.0345, -0.1637, -0.0095, ..., 0.2015, -0.1704, 0.0405], + ..., + [-0.1735, 0.1426, 0.0251, ..., -0.1499, 0.1149, 0.0600], + [ 0.0022, -0.0799, -0.0011, ..., -0.1183, -0.1709, -0.2147], + [ 0.0230, -0.0441, -0.0240, ..., -0.2070, -0.1634, 0.0271]], + device='cuda:0'), grad: tensor([[ 6.5193e-09, 2.3283e-10, 0.0000e+00, ..., 8.1491e-10, + 3.4925e-10, 6.9849e-10], + [-7.6834e-09, 2.4447e-09, 0.0000e+00, ..., 1.0477e-09, + -1.1642e-08, 2.3283e-09], + [ 5.3551e-09, 3.0268e-09, 0.0000e+00, ..., 3.4925e-10, + 4.6566e-09, 3.4925e-09], + ..., + [ 1.4203e-08, -1.3737e-08, 0.0000e+00, ..., 1.3970e-09, + -9.3132e-09, -5.4715e-09], + [ 3.6671e-08, 3.7253e-09, 0.0000e+00, ..., 1.5134e-09, + 9.1968e-09, 1.0827e-08], + [-1.4668e-08, 2.3283e-09, -4.6566e-10, ..., 3.8417e-09, + 2.4447e-09, -2.0606e-08]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0376, 0.0136, 0.0144, 0.0278, 0.0308, -0.0194, 0.0236, 0.0323, + -0.0099, 0.0014], device='cuda:0'), grad: tensor([ 2.7590e-08, 7.6834e-09, 2.7008e-08, 1.8114e-07, 6.5891e-08, + -8.2236e-07, 4.4168e-07, 2.9453e-08, 1.5460e-07, -9.3831e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 250.12, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4765 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.05 lr 0.00010000 +Epoch 295, weight, value: tensor([[ 0.0816, -0.1429, -0.1393, ..., -0.1256, -0.1576, -0.1500], + [ 0.0381, 0.0510, -0.0344, ..., -0.0579, 0.1612, 0.0178], + [-0.0347, -0.1638, -0.0095, ..., 0.2016, -0.1704, 0.0406], + ..., + [-0.1767, 0.1430, 0.0253, ..., -0.1500, 0.1151, 0.0601], + [ 0.0022, -0.0804, -0.0011, ..., -0.1185, -0.1712, -0.2171], + [ 0.0234, -0.0440, -0.0240, ..., -0.2072, -0.1635, 0.0270]], + device='cuda:0'), grad: tensor([[-2.3632e-08, -1.2573e-08, 0.0000e+00, ..., 8.1491e-10, + -3.2014e-08, -6.8685e-09], + [-5.4482e-08, 4.5868e-08, 0.0000e+00, ..., 2.0955e-09, + -4.3190e-08, 7.4506e-09], + [ 1.5018e-08, 4.3423e-08, 0.0000e+00, ..., 1.6298e-09, + 7.6601e-08, 4.2375e-08], + ..., + [ 3.0501e-08, -2.2887e-07, 0.0000e+00, ..., 1.5134e-09, + -2.6496e-07, -1.8312e-07], + [ 2.9337e-08, 5.4599e-08, 0.0000e+00, ..., 1.3970e-09, + 1.1432e-07, 5.7276e-08], + [ 4.7730e-09, 3.7253e-08, 0.0000e+00, ..., 4.1910e-09, + 5.2154e-08, 3.7020e-08]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0374, 0.0136, 0.0144, 0.0277, 0.0309, -0.0194, 0.0237, 0.0324, + -0.0101, 0.0013], device='cuda:0'), grad: tensor([-1.7649e-07, 4.7614e-08, 2.2049e-07, -1.6065e-07, 7.9395e-08, + 1.0745e-07, 9.3132e-10, -5.0664e-07, 2.6217e-07, 1.4796e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 250.35, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4529 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.11 lr 0.00010000 +Epoch 296, weight, value: tensor([[ 0.0816, -0.1430, -0.1393, ..., -0.1265, -0.1588, -0.1502], + [ 0.0382, 0.0511, -0.0344, ..., -0.0579, 0.1615, 0.0182], + [-0.0350, -0.1638, -0.0095, ..., 0.2017, -0.1704, 0.0406], + ..., + [-0.1792, 0.1429, 0.0253, ..., -0.1500, 0.1150, 0.0600], + [ 0.0022, -0.0804, -0.0011, ..., -0.1186, -0.1713, -0.2174], + [ 0.0245, -0.0443, -0.0239, ..., -0.2072, -0.1638, 0.0270]], + device='cuda:0'), grad: tensor([[-1.1642e-09, 6.9849e-10, 0.0000e+00, ..., 1.9791e-09, + 3.8417e-09, 2.6776e-09], + [-1.2771e-07, -8.3819e-09, 0.0000e+00, ..., 4.6566e-09, + -3.4715e-07, -1.0082e-07], + [ 9.5461e-09, 7.7998e-09, 0.0000e+00, ..., -2.6193e-08, + 2.1537e-08, -4.4238e-09], + ..., + [ 2.5029e-08, -1.0827e-08, 0.0000e+00, ..., 1.1292e-08, + 3.4459e-08, 1.9791e-08], + [ 3.6787e-08, 5.5879e-09, 0.0000e+00, ..., 3.4925e-09, + 5.9721e-08, 2.4913e-08], + [-9.7905e-08, 6.9849e-10, 0.0000e+00, ..., 1.2806e-09, + 2.7125e-08, -6.2515e-08]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0378, 0.0137, 0.0144, 0.0279, 0.0309, -0.0195, 0.0238, 0.0323, + -0.0100, 0.0013], device='cuda:0'), grad: tensor([ 3.1432e-09, -7.5111e-07, 2.5262e-08, 1.6345e-07, 7.7859e-07, + 1.3725e-07, 6.4145e-08, 1.4610e-07, 2.2398e-07, -7.6834e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 250.14, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4724 re_mapping 0.0027 re_causal 0.0094 /// teacc 99.01 lr 0.00010000 +Epoch 297, weight, value: tensor([[ 0.0819, -0.1433, -0.1394, ..., -0.1266, -0.1590, -0.1503], + [ 0.0382, 0.0511, -0.0343, ..., -0.0595, 0.1613, 0.0179], + [-0.0339, -0.1638, -0.0096, ..., 0.2021, -0.1699, 0.0407], + ..., + [-0.1809, 0.1435, 0.0253, ..., -0.1501, 0.1156, 0.0603], + [ 0.0022, -0.0805, -0.0011, ..., -0.1195, -0.1713, -0.2177], + [ 0.0246, -0.0444, -0.0240, ..., -0.2074, -0.1639, 0.0270]], + device='cuda:0'), grad: tensor([[ 2.2119e-08, 1.4203e-08, 0.0000e+00, ..., 4.4121e-08, + 3.6089e-08, 2.4913e-08], + [-5.4576e-07, -4.9314e-07, 0.0000e+00, ..., 7.7416e-08, + -1.2312e-06, -3.1060e-07], + [-2.8987e-08, 1.7160e-07, 0.0000e+00, ..., -6.4587e-07, + 4.2422e-07, -6.0350e-07], + ..., + [ 1.7311e-07, 7.9046e-08, 0.0000e+00, ..., 1.6647e-07, + 2.1944e-07, 2.4913e-07], + [ 1.3807e-07, 1.5378e-07, 0.0000e+00, ..., 1.2072e-07, + 3.2876e-07, 1.5728e-07], + [ 4.7148e-08, 1.2806e-08, 0.0000e+00, ..., 2.5146e-08, + 5.8440e-08, 4.4238e-09]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0377, 0.0135, 0.0145, 0.0255, 0.0309, -0.0189, 0.0244, 0.0327, + -0.0101, 0.0013], device='cuda:0'), grad: tensor([ 2.3597e-07, -2.0992e-06, -2.2836e-06, 1.8384e-06, 1.0207e-06, + 2.1083e-07, -1.3122e-06, 1.3970e-06, 5.5693e-07, 4.1304e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 250.42, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4833 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.05 lr 0.00010000 +Epoch 298, weight, value: tensor([[ 0.0825, -0.1437, -0.1394, ..., -0.1271, -0.1592, -0.1504], + [ 0.0381, 0.0511, -0.0343, ..., -0.0599, 0.1614, 0.0176], + [-0.0314, -0.1638, -0.0096, ..., 0.2024, -0.1700, 0.0408], + ..., + [-0.1836, 0.1436, 0.0253, ..., -0.1502, 0.1156, 0.0602], + [ 0.0022, -0.0805, -0.0011, ..., -0.1203, -0.1713, -0.2183], + [ 0.0251, -0.0442, -0.0239, ..., -0.2079, -0.1640, 0.0270]], + device='cuda:0'), grad: tensor([[-2.4796e-08, 1.0128e-08, 0.0000e+00, ..., -1.2689e-08, + 2.0606e-08, 1.0361e-08], + [-2.9267e-07, -6.2049e-08, 0.0000e+00, ..., -1.8650e-07, + -4.3819e-07, -1.1665e-07], + [ 1.9127e-07, 1.8184e-07, 0.0000e+00, ..., 1.2584e-07, + 2.5495e-07, 3.4086e-07], + ..., + [ 3.6117e-06, 1.6198e-05, 0.0000e+00, ..., 1.1525e-08, + 6.4522e-06, 2.6837e-05], + [ 1.3085e-07, 2.8568e-07, 0.0000e+00, ..., 3.2247e-08, + 2.0594e-07, 4.6450e-07], + [-3.7756e-06, -1.7077e-05, 0.0000e+00, ..., 2.6193e-08, + -6.7614e-06, -2.8297e-05]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0375, 0.0134, 0.0146, 0.0263, 0.0310, -0.0198, 0.0246, 0.0326, + -0.0101, 0.0014], device='cuda:0'), grad: tensor([-7.6718e-08, -6.6869e-07, 1.4091e-06, 1.2794e-07, 2.0415e-06, + 3.5530e-07, 9.4413e-08, 9.6142e-05, 1.7788e-06, -1.0133e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 250.30, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4646 re_mapping 0.0030 re_causal 0.0099 /// teacc 99.05 lr 0.00010000 +Epoch 299, weight, value: tensor([[ 0.0859, -0.1440, -0.1394, ..., -0.1266, -0.1580, -0.1506], + [ 0.0381, 0.0510, -0.0341, ..., -0.0601, 0.1614, 0.0165], + [-0.0330, -0.1639, -0.0096, ..., 0.2026, -0.1701, 0.0408], + ..., + [-0.1865, 0.1424, 0.0251, ..., -0.1503, 0.1154, 0.0587], + [ 0.0022, -0.0805, -0.0012, ..., -0.1214, -0.1713, -0.2189], + [ 0.0264, -0.0404, -0.0239, ..., -0.2083, -0.1624, 0.0295]], + device='cuda:0'), grad: tensor([[-5.5414e-08, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, + -0.0000e+00, 1.9791e-09], + [-1.4319e-07, -3.9930e-08, 0.0000e+00, ..., 1.2806e-09, + -1.8731e-07, -9.7323e-08], + [ 4.8662e-08, 2.6659e-08, 0.0000e+00, ..., 7.7998e-09, + 7.1595e-08, 4.6915e-08], + ..., + [ 2.5379e-08, -1.2806e-08, 0.0000e+00, ..., 2.3283e-10, + 7.9162e-09, -3.7253e-09], + [ 4.9127e-08, 1.3504e-08, 0.0000e+00, ..., 3.4925e-10, + 5.9023e-08, 3.2829e-08], + [-1.0361e-08, 6.2864e-09, 0.0000e+00, ..., 1.9791e-09, + 2.1304e-08, -4.0047e-08]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0352, 0.0132, 0.0146, 0.0264, 0.0310, -0.0201, 0.0246, 0.0305, + -0.0101, 0.0035], device='cuda:0'), grad: tensor([-3.1060e-07, -4.2492e-07, 3.8370e-07, -3.0361e-07, 2.5705e-07, + 2.9104e-08, 2.9337e-08, 1.9558e-07, 1.8999e-07, -3.2247e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 250.62, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4927 re_mapping 0.0029 re_causal 0.0097 /// teacc 99.16 lr 0.00010000 +Epoch 300, weight, value: tensor([[ 0.0878, -0.1448, -0.1394, ..., -0.1267, -0.1589, -0.1507], + [ 0.0381, 0.0505, -0.0341, ..., -0.0603, 0.1611, 0.0155], + [-0.0343, -0.1640, -0.0099, ..., 0.2026, -0.1703, 0.0407], + ..., + [-0.1885, 0.1428, 0.0251, ..., -0.1503, 0.1162, 0.0589], + [ 0.0022, -0.0805, -0.0021, ..., -0.1216, -0.1714, -0.2193], + [ 0.0255, -0.0405, -0.0240, ..., -0.2093, -0.1627, 0.0293]], + device='cuda:0'), grad: tensor([[-1.2922e-08, 2.3283e-10, 0.0000e+00, ..., 9.8138e-08, + 9.3132e-10, 5.8208e-10], + [-3.5740e-08, -1.5018e-08, 0.0000e+00, ..., 7.9628e-08, + -7.7998e-08, -2.8755e-08], + [ 8.3819e-09, 4.0745e-09, 0.0000e+00, ..., 2.8708e-07, + 9.8953e-09, 5.8208e-09], + ..., + [ 2.2002e-08, 4.0745e-09, 0.0000e+00, ..., 6.7521e-09, + 3.3178e-08, 1.1991e-08], + [-7.1013e-09, 3.3760e-09, 0.0000e+00, ..., 2.0536e-07, + 1.6764e-08, 7.1013e-09], + [ 1.5600e-08, 1.0477e-09, 0.0000e+00, ..., 2.8755e-08, + 5.2387e-09, 6.2864e-09]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0340, 0.0130, 0.0144, 0.0260, 0.0311, -0.0199, 0.0249, 0.0307, + -0.0101, 0.0033], device='cuda:0'), grad: tensor([ 1.6904e-07, 8.1025e-08, 6.1002e-07, 4.4587e-08, 1.6228e-07, + 3.9898e-06, -5.6587e-06, 9.7556e-08, 3.6880e-07, 1.5507e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 250.64, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4894 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.08 lr 0.00010000 +Epoch 301, weight, value: tensor([[ 0.0878, -0.1452, -0.1394, ..., -0.1269, -0.1595, -0.1509], + [ 0.0382, 0.0504, -0.0342, ..., -0.0604, 0.1614, 0.0158], + [-0.0341, -0.1641, -0.0099, ..., 0.2031, -0.1708, 0.0407], + ..., + [-0.1906, 0.1430, 0.0252, ..., -0.1504, 0.1162, 0.0589], + [ 0.0022, -0.0805, -0.0022, ..., -0.1229, -0.1714, -0.2196], + [ 0.0247, -0.0407, -0.0241, ..., -0.2100, -0.1633, 0.0291]], + device='cuda:0'), grad: tensor([[ 8.0327e-09, 1.1642e-09, 0.0000e+00, ..., 1.1874e-08, + 2.9104e-09, 1.8161e-08], + [-1.7253e-07, -1.2852e-07, 0.0000e+00, ..., 1.0640e-07, + -5.4296e-07, -1.2002e-07], + [ 2.1304e-08, 9.8953e-09, 0.0000e+00, ..., -1.9162e-07, + 3.5274e-08, -2.2817e-08], + ..., + [ 1.8568e-07, 8.8825e-08, 0.0000e+00, ..., 9.5577e-08, + 3.6415e-07, 3.8138e-07], + [ 7.2992e-08, 9.5461e-09, 0.0000e+00, ..., 8.0327e-08, + 2.6193e-08, 1.3853e-07], + [-4.8103e-07, -5.5297e-08, 0.0000e+00, ..., 1.9837e-06, + 1.7928e-08, 2.5909e-06]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0341, 0.0131, 0.0144, 0.0260, 0.0313, -0.0199, 0.0247, 0.0307, + -0.0101, 0.0031], device='cuda:0'), grad: tensor([ 6.7987e-08, -5.3970e-07, -1.9360e-07, 1.3574e-07, -8.3074e-06, + -3.3434e-07, 7.3994e-07, 1.3150e-06, 5.4389e-07, 6.5789e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 250.59, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4537 re_mapping 0.0029 re_causal 0.0094 /// teacc 99.09 lr 0.00010000 +Epoch 302, weight, value: tensor([[ 0.0879, -0.1466, -0.1394, ..., -0.1301, -0.1612, -0.1511], + [ 0.0383, 0.0505, -0.0345, ..., -0.0605, 0.1616, 0.0161], + [-0.0347, -0.1642, -0.0100, ..., 0.2033, -0.1710, 0.0407], + ..., + [-0.1925, 0.1430, 0.0256, ..., -0.1504, 0.1163, 0.0589], + [ 0.0022, -0.0806, -0.0022, ..., -0.1233, -0.1714, -0.2200], + [ 0.0244, -0.0408, -0.0245, ..., -0.2103, -0.1636, 0.0290]], + device='cuda:0'), grad: tensor([[ 3.8301e-08, 6.9849e-10, 0.0000e+00, ..., 7.5204e-08, + 2.4447e-09, 1.0477e-09], + [-1.3388e-08, -2.3283e-09, 0.0000e+00, ..., 6.5193e-09, + -3.7835e-08, -1.0361e-08], + [ 6.8918e-08, 7.5670e-09, 0.0000e+00, ..., 3.4925e-10, + 1.6880e-08, 1.0361e-08], + ..., + [ 4.2375e-08, 3.7486e-08, 0.0000e+00, ..., 1.0477e-09, + -3.6089e-09, 3.2480e-08], + [-1.7951e-07, 3.0268e-09, 0.0000e+00, ..., 7.3342e-09, + 1.1409e-08, 5.4715e-09], + [ 2.4796e-08, -6.1351e-08, 0.0000e+00, ..., 1.1642e-09, + 2.7940e-09, -5.5879e-08]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0370, 0.0132, 0.0144, 0.0237, 0.0313, -0.0176, 0.0272, 0.0307, + -0.0100, 0.0030], device='cuda:0'), grad: tensor([ 3.7812e-07, -1.5134e-09, 2.5472e-07, 3.1898e-08, 2.6496e-07, + -2.4214e-08, -4.7078e-07, 2.6613e-07, -5.7416e-07, -1.0792e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 250.52, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4707 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.00 lr 0.00010000 +Epoch 303, weight, value: tensor([[ 0.0880, -0.1472, -0.1394, ..., -0.1302, -0.1618, -0.1513], + [ 0.0384, 0.0506, -0.0345, ..., -0.0603, 0.1620, 0.0167], + [-0.0351, -0.1642, -0.0099, ..., 0.2035, -0.1714, 0.0407], + ..., + [-0.1939, 0.1430, 0.0256, ..., -0.1505, 0.1162, 0.0589], + [ 0.0022, -0.0806, -0.0022, ..., -0.1225, -0.1715, -0.2205], + [ 0.0236, -0.0409, -0.0245, ..., -0.2107, -0.1643, 0.0289]], + device='cuda:0'), grad: tensor([[-4.1910e-09, 1.1642e-10, 0.0000e+00, ..., 3.3760e-09, + 2.3283e-10, 4.6566e-09], + [-8.0327e-09, -2.3283e-10, 0.0000e+00, ..., 3.7067e-07, + -1.3039e-08, 5.4110e-07], + [ 1.3970e-09, 6.1700e-09, 0.0000e+00, ..., 5.4715e-08, + 6.9849e-09, 9.4064e-08], + ..., + [ 3.2596e-09, -1.0477e-08, 0.0000e+00, ..., 2.9407e-07, + -9.3132e-09, 4.2189e-07], + [ 1.5716e-08, 2.0955e-09, 0.0000e+00, ..., 4.0163e-08, + 1.0245e-08, 5.8673e-08], + [-9.6625e-09, 1.2806e-09, 0.0000e+00, ..., 1.6578e-06, + 1.3970e-09, 2.4345e-06]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0369, 0.0133, 0.0144, 0.0238, 0.0314, -0.0178, 0.0268, 0.0306, + -0.0099, 0.0029], device='cuda:0'), grad: tensor([-2.3283e-10, 1.5497e-06, 2.7521e-07, 9.5228e-08, -1.0297e-05, + -1.8685e-07, 9.7440e-08, 1.2266e-06, 2.6543e-07, 6.9775e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 250.61, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4813 re_mapping 0.0028 re_causal 0.0091 /// teacc 99.11 lr 0.00010000 +Epoch 304, weight, value: tensor([[ 0.0881, -0.1481, -0.1394, ..., -0.1303, -0.1627, -0.1520], + [ 0.0384, 0.0507, -0.0346, ..., -0.0606, 0.1623, 0.0168], + [-0.0352, -0.1643, -0.0100, ..., 0.2036, -0.1715, 0.0407], + ..., + [-0.1951, 0.1431, 0.0258, ..., -0.1505, 0.1163, 0.0589], + [ 0.0022, -0.0807, -0.0022, ..., -0.1227, -0.1716, -0.2223], + [ 0.0251, -0.0408, -0.0248, ..., -0.2112, -0.1644, 0.0289]], + device='cuda:0'), grad: tensor([[-1.1758e-07, 5.8208e-10, 0.0000e+00, ..., -8.8476e-09, + 1.1642e-09, 2.5611e-09], + [-3.7486e-08, 2.1770e-08, 0.0000e+00, ..., 3.7253e-09, + 8.4983e-09, 9.3482e-08], + [ 4.1910e-09, 7.3924e-08, 0.0000e+00, ..., 5.8208e-10, + 1.2282e-07, 1.2747e-07], + ..., + [ 3.0850e-08, -3.0152e-07, 0.0000e+00, ..., 8.1491e-09, + -4.8568e-07, -4.9127e-07], + [ 6.9500e-08, 1.2224e-08, 0.0000e+00, ..., 4.5402e-09, + 2.5379e-08, 1.7264e-07], + [-2.2235e-08, 1.5856e-07, 0.0000e+00, ..., 1.1711e-07, + 2.6566e-07, 3.7579e-07]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0370, 0.0134, 0.0144, 0.0238, 0.0315, -0.0178, 0.0268, 0.0306, + -0.0101, 0.0029], device='cuda:0'), grad: tensor([-7.7020e-07, 2.8592e-07, 3.6974e-07, 1.1094e-07, 2.2408e-06, + -1.1455e-07, 6.8452e-08, -1.1865e-06, 2.1253e-06, -3.1069e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 250.46, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4279 re_mapping 0.0029 re_causal 0.0091 /// teacc 98.99 lr 0.00010000 +Epoch 305, weight, value: tensor([[ 0.0880, -0.1495, -0.1394, ..., -0.1304, -0.1638, -0.1525], + [ 0.0385, 0.0507, -0.0346, ..., -0.0620, 0.1625, 0.0167], + [-0.0353, -0.1643, -0.0100, ..., 0.2050, -0.1717, 0.0409], + ..., + [-0.1968, 0.1434, 0.0258, ..., -0.1506, 0.1166, 0.0590], + [ 0.0022, -0.0808, -0.0022, ..., -0.1232, -0.1718, -0.2235], + [ 0.0254, -0.0411, -0.0248, ..., -0.2118, -0.1651, 0.0287]], + device='cuda:0'), grad: tensor([[ 1.0373e-07, 5.3551e-09, 0.0000e+00, ..., 1.6764e-08, + 1.3167e-07, 1.1572e-07], + [ 1.7369e-07, 1.8366e-06, 0.0000e+00, ..., 7.9162e-09, + -3.4757e-06, -3.0063e-06], + [ 2.9523e-07, 5.9954e-09, 0.0000e+00, ..., -6.9849e-10, + 3.9116e-07, 3.4273e-07], + ..., + [ 3.4389e-07, 7.6252e-09, 0.0000e+00, ..., 1.3213e-08, + 4.3958e-07, 4.0745e-07], + [-2.9840e-06, -2.1588e-06, 0.0000e+00, ..., 3.2596e-09, + 3.5274e-07, 2.9197e-07], + [ 3.9255e-07, 3.0443e-08, 0.0000e+00, ..., 2.7008e-08, + 4.7404e-07, 4.6613e-07]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0371, 0.0134, 0.0146, 0.0238, 0.0315, -0.0178, 0.0266, 0.0307, + -0.0102, 0.0028], device='cuda:0'), grad: tensor([ 6.6776e-07, -6.9551e-06, 1.8077e-06, 6.3740e-06, 1.5693e-07, + 1.5078e-06, 7.0687e-07, 2.1234e-06, -8.8289e-06, 2.4382e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 250.18, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4577 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.13 lr 0.00010000 +Epoch 306, weight, value: tensor([[ 0.0879, -0.1511, -0.1394, ..., -0.1304, -0.1646, -0.1528], + [ 0.0386, 0.0513, -0.0346, ..., -0.0638, 0.1632, 0.0171], + [-0.0360, -0.1644, -0.0100, ..., 0.2062, -0.1718, 0.0411], + ..., + [-0.2016, 0.1432, 0.0258, ..., -0.1508, 0.1160, 0.0587], + [ 0.0022, -0.0809, -0.0022, ..., -0.1235, -0.1719, -0.2250], + [ 0.0256, -0.0412, -0.0248, ..., -0.2125, -0.1656, 0.0286]], + device='cuda:0'), grad: tensor([[ 3.6089e-09, 1.2806e-09, 1.1642e-10, ..., 1.7637e-08, + 3.7835e-09, 2.1886e-08], + [-3.4051e-08, 6.8394e-08, 1.1642e-10, ..., -3.1781e-07, + -2.3097e-06, -1.8729e-06], + [-4.3772e-08, 1.0303e-08, 1.6298e-09, ..., -2.1874e-07, + 9.3132e-07, 3.7393e-07], + ..., + [ 4.1677e-08, -1.8976e-07, 5.8208e-10, ..., 3.6391e-07, + 1.0328e-06, 1.0561e-06], + [ 1.7288e-08, 3.8126e-08, 4.0745e-10, ..., 1.0850e-07, + 9.8429e-08, 1.9546e-07], + [-5.7044e-09, 2.0373e-08, 1.7462e-10, ..., 2.6892e-08, + 3.3120e-08, 4.8196e-08]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0371, 0.0136, 0.0147, 0.0237, 0.0317, -0.0177, 0.0265, 0.0304, + -0.0102, 0.0028], device='cuda:0'), grad: tensor([ 6.2166e-08, -4.5262e-06, 1.0561e-06, -6.4843e-08, 3.2224e-07, + 1.1869e-07, -2.4156e-08, 2.4773e-06, 4.5286e-07, 1.3015e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 250.76, cls_loss 0.0008 cls_loss_mapping 0.0012 cls_loss_causal 0.4869 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.03 lr 0.00010000 +Epoch 307, weight, value: tensor([[ 0.0878, -0.1521, -0.1394, ..., -0.1305, -0.1655, -0.1532], + [ 0.0383, 0.0517, -0.0344, ..., -0.0640, 0.1636, 0.0168], + [-0.0376, -0.1645, -0.0101, ..., 0.2065, -0.1720, 0.0412], + ..., + [-0.2049, 0.1431, 0.0258, ..., -0.1508, 0.1158, 0.0586], + [ 0.0023, -0.0810, -0.0023, ..., -0.1226, -0.1721, -0.2262], + [ 0.0292, -0.0412, -0.0248, ..., -0.2129, -0.1652, 0.0289]], + device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.6193e-09, 0.0000e+00, ..., 2.3865e-09, + 2.9686e-09, 4.1327e-09], + [ 1.0303e-08, 1.1578e-07, 0.0000e+00, ..., 2.1479e-08, + 9.8546e-08, 8.6729e-08], + [ 1.4948e-07, 9.8662e-08, 0.0000e+00, ..., 9.1910e-08, + 1.0541e-07, 6.8860e-08], + ..., + [ 1.0070e-08, -3.4645e-07, 0.0000e+00, ..., 1.4377e-08, + -3.4110e-07, -2.1921e-07], + [-2.4098e-07, 2.3399e-08, 0.0000e+00, ..., -1.5099e-07, + 2.9628e-08, 2.5437e-08], + [ 3.3993e-08, 3.1665e-08, 0.0000e+00, ..., 8.3004e-08, + 5.2562e-08, 1.4342e-07]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0372, 0.0131, 0.0146, 0.0237, 0.0317, -0.0177, 0.0260, 0.0302, + -0.0100, 0.0034], device='cuda:0'), grad: tensor([ 2.8871e-08, 4.5635e-07, 9.4809e-07, 1.4203e-08, -3.4925e-07, + 6.8860e-08, 8.4168e-08, -8.0327e-07, -1.0896e-06, 6.6124e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 250.67, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4359 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.00 lr 0.00010000 +Epoch 308, weight, value: tensor([[ 0.0877, -0.1526, -0.1394, ..., -0.1305, -0.1658, -0.1534], + [ 0.0383, 0.0516, -0.0344, ..., -0.0642, 0.1636, 0.0166], + [-0.0377, -0.1645, -0.0101, ..., 0.2067, -0.1721, 0.0412], + ..., + [-0.2055, 0.1432, 0.0258, ..., -0.1509, 0.1159, 0.0586], + [ 0.0023, -0.0810, -0.0023, ..., -0.1226, -0.1721, -0.2267], + [ 0.0292, -0.0413, -0.0248, ..., -0.2135, -0.1655, 0.0288]], + device='cuda:0'), grad: tensor([[ 2.8755e-08, 1.9209e-09, 0.0000e+00, ..., 7.5670e-09, + 5.8208e-10, 7.5670e-10], + [ 1.0675e-07, 4.8545e-08, 0.0000e+00, ..., 1.0303e-08, + 2.9453e-08, 3.3178e-08], + [ 2.5088e-08, 1.0186e-08, 0.0000e+00, ..., 1.1642e-09, + 1.1292e-08, 7.3342e-09], + ..., + [ 1.7521e-08, -1.3143e-07, 0.0000e+00, ..., 7.1595e-09, + -1.4959e-07, -8.0559e-08], + [ 6.4773e-07, 2.4040e-08, 0.0000e+00, ..., -5.0641e-09, + 9.3714e-09, -1.9209e-09], + [ 1.7055e-07, 7.9221e-08, 0.0000e+00, ..., 1.6531e-08, + 7.8056e-08, 6.9092e-08]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0373, 0.0131, 0.0146, 0.0237, 0.0318, -0.0177, 0.0261, 0.0302, + -0.0100, 0.0032], device='cuda:0'), grad: tensor([ 1.2608e-07, 5.1688e-07, 1.2992e-07, 3.5074e-06, -5.1921e-08, + -8.6054e-06, 1.2945e-06, -3.4517e-08, 2.2277e-06, 9.0431e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 250.66, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4611 re_mapping 0.0030 re_causal 0.0097 /// teacc 98.98 lr 0.00010000 +Epoch 309, weight, value: tensor([[ 0.0877, -0.1534, -0.1394, ..., -0.1306, -0.1660, -0.1536], + [ 0.0384, 0.0518, -0.0345, ..., -0.0644, 0.1642, 0.0173], + [-0.0378, -0.1646, -0.0101, ..., 0.2070, -0.1723, 0.0412], + ..., + [-0.2079, 0.1431, 0.0258, ..., -0.1510, 0.1154, 0.0585], + [ 0.0023, -0.0810, -0.0023, ..., -0.1229, -0.1722, -0.2273], + [ 0.0303, -0.0413, -0.0249, ..., -0.2139, -0.1656, 0.0288]], + device='cuda:0'), grad: tensor([[ 5.8208e-10, 4.6566e-10, 0.0000e+00, ..., 2.6776e-09, + 1.1059e-09, 1.1059e-09], + [-9.6916e-08, -9.4238e-08, 0.0000e+00, ..., -3.4925e-09, + -2.1560e-07, -6.0536e-08], + [ 1.4668e-08, 1.7055e-08, 0.0000e+00, ..., -2.0897e-08, + 3.2247e-08, 9.3714e-09], + ..., + [ 3.0617e-08, 1.7055e-08, 0.0000e+00, ..., 1.2980e-08, + 3.5332e-08, 5.5123e-08], + [ 6.1875e-08, 5.5996e-08, 0.0000e+00, ..., 1.5600e-08, + 1.2270e-07, 4.2724e-08], + [-2.2119e-08, -5.8208e-09, 0.0000e+00, ..., 2.5786e-08, + 1.9209e-09, -1.7986e-08]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0373, 0.0132, 0.0146, 0.0237, 0.0319, -0.0178, 0.0256, 0.0300, + -0.0099, 0.0034], device='cuda:0'), grad: tensor([ 9.0804e-09, -3.7835e-07, 1.4494e-08, -2.9686e-09, -9.3947e-08, + 2.1188e-08, 1.0885e-08, 2.3888e-07, 2.6263e-07, -7.2410e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 250.47, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4960 re_mapping 0.0029 re_causal 0.0098 /// teacc 99.06 lr 0.00010000 +Epoch 310, weight, value: tensor([[ 0.0867, -0.1542, -0.1394, ..., -0.1309, -0.1664, -0.1538], + [ 0.0384, 0.0518, -0.0346, ..., -0.0647, 0.1645, 0.0173], + [-0.0384, -0.1646, -0.0101, ..., 0.2071, -0.1724, 0.0412], + ..., + [-0.2087, 0.1432, 0.0259, ..., -0.1510, 0.1153, 0.0584], + [ 0.0025, -0.0810, -0.0023, ..., -0.1216, -0.1722, -0.2278], + [ 0.0301, -0.0414, -0.0249, ..., -0.2145, -0.1657, 0.0287]], + device='cuda:0'), grad: tensor([[ 1.9791e-09, 9.8953e-10, 0.0000e+00, ..., 6.4028e-10, + 2.7940e-09, 3.8417e-09], + [-1.9046e-07, -9.7265e-08, 0.0000e+00, ..., -4.5402e-09, + -3.0268e-07, -1.1612e-07], + [ 4.3306e-08, 2.9395e-08, 0.0000e+00, ..., -1.6997e-08, + 8.0385e-08, 2.1129e-08], + ..., + [ 3.3237e-08, -5.2387e-10, 0.0000e+00, ..., 3.4925e-09, + 2.8289e-08, 1.1816e-08], + [ 8.6206e-08, 4.4412e-08, 0.0000e+00, ..., 4.0745e-08, + 1.3039e-07, 5.8732e-08], + [-2.2701e-09, 9.3714e-09, 0.0000e+00, ..., 4.6566e-09, + 2.5495e-08, 2.9104e-10]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0378, 0.0133, 0.0146, 0.0237, 0.0320, -0.0178, 0.0252, 0.0300, + -0.0097, 0.0033], device='cuda:0'), grad: tensor([ 6.2864e-09, -6.5798e-07, 1.3411e-07, -1.3271e-08, 8.6147e-08, + 1.3504e-07, -1.6787e-07, 8.6147e-08, 4.3446e-07, -9.5461e-09], + device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 250.52, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4607 re_mapping 0.0026 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +Epoch 311, weight, value: tensor([[ 0.0864, -0.1568, -0.1394, ..., -0.1309, -0.1677, -0.1541], + [ 0.0385, 0.0516, -0.0351, ..., -0.0651, 0.1644, 0.0169], + [-0.0387, -0.1647, -0.0102, ..., 0.2073, -0.1726, 0.0413], + ..., + [-0.2092, 0.1435, 0.0264, ..., -0.1511, 0.1158, 0.0586], + [ 0.0025, -0.0811, -0.0023, ..., -0.1219, -0.1723, -0.2286], + [ 0.0303, -0.0415, -0.0252, ..., -0.2151, -0.1660, 0.0287]], + device='cuda:0'), grad: tensor([[ 2.9104e-10, 4.0163e-09, 0.0000e+00, ..., 2.2817e-08, + 3.2014e-09, 1.1583e-08], + [ 1.1292e-08, 3.3469e-08, 0.0000e+00, ..., 4.9942e-08, + 2.9046e-08, 4.6974e-08], + [ 4.3074e-09, 1.9406e-07, 0.0000e+00, ..., -4.9639e-07, + 1.7206e-07, -3.0617e-08], + ..., + [ 6.9849e-10, -3.2852e-07, 0.0000e+00, ..., 4.0047e-08, + -2.9104e-07, -2.4587e-07], + [-1.7462e-08, 1.6007e-08, 0.0000e+00, ..., 2.8813e-08, + 1.2049e-08, 2.9802e-08], + [ 3.8999e-09, 3.3178e-08, 0.0000e+00, ..., 2.9453e-08, + 2.7008e-08, 5.4017e-08]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0380, 0.0131, 0.0146, 0.0237, 0.0320, -0.0178, 0.0254, 0.0301, + -0.0097, 0.0033], device='cuda:0'), grad: tensor([ 6.5891e-08, 2.6985e-07, -5.0291e-07, 7.7765e-07, 9.8371e-09, + 4.2201e-08, 1.6764e-08, -8.8057e-07, 9.1386e-09, 2.1688e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 250.75, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4687 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.09 lr 0.00010000 +Epoch 312, weight, value: tensor([[ 0.0864, -0.1575, -0.1394, ..., -0.1310, -0.1682, -0.1543], + [ 0.0385, 0.0515, -0.0352, ..., -0.0652, 0.1643, 0.0166], + [-0.0376, -0.1649, -0.0102, ..., 0.2081, -0.1728, 0.0416], + ..., + [-0.2094, 0.1438, 0.0265, ..., -0.1511, 0.1163, 0.0587], + [ 0.0025, -0.0812, -0.0023, ..., -0.1219, -0.1724, -0.2290], + [ 0.0301, -0.0416, -0.0253, ..., -0.2171, -0.1662, 0.0284]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 4.5984e-09, 0.0000e+00, ..., 7.5670e-10, + 3.4051e-08, 1.3912e-08], + [-4.0466e-07, -2.1537e-09, 0.0000e+00, ..., -1.8859e-08, + -6.0769e-07, -1.1956e-07], + [ 8.8359e-08, 8.0443e-08, 0.0000e+00, ..., 6.5775e-09, + 1.8917e-07, 1.2596e-07], + ..., + [ 3.3120e-08, -5.8673e-07, 0.0000e+00, ..., 9.1968e-09, + -3.6764e-07, -6.2771e-07], + [ 1.9965e-07, 7.0140e-08, 0.0000e+00, ..., 2.1013e-08, + 2.1292e-07, 1.0524e-07], + [-2.0233e-07, 1.1700e-07, 0.0000e+00, ..., 1.1548e-07, + 9.3016e-08, 3.2200e-07]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0381, 0.0130, 0.0150, 0.0238, 0.0321, -0.0179, 0.0253, 0.0302, + -0.0097, 0.0029], device='cuda:0'), grad: tensor([-6.1525e-08, -1.2191e-06, 5.6392e-07, 5.3260e-08, 4.0652e-07, + 3.4389e-07, 3.2433e-07, -1.5749e-06, 7.9721e-07, 3.5088e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 250.78, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4935 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.12 lr 0.00010000 +Epoch 313, weight, value: tensor([[ 0.0892, -0.1587, -0.1394, ..., -0.1306, -0.1690, -0.1546], + [ 0.0386, 0.0517, -0.0352, ..., -0.0652, 0.1649, 0.0171], + [-0.0385, -0.1652, -0.0102, ..., 0.2081, -0.1734, 0.0414], + ..., + [-0.2103, 0.1440, 0.0265, ..., -0.1511, 0.1165, 0.0587], + [ 0.0025, -0.0813, -0.0023, ..., -0.1231, -0.1726, -0.2305], + [ 0.0315, -0.0417, -0.0253, ..., -0.2167, -0.1665, 0.0287]], + device='cuda:0'), grad: tensor([[-5.1223e-08, -3.8999e-09, 0.0000e+00, ..., 2.2585e-08, + -6.1700e-09, 6.9849e-10], + [ 7.4564e-08, -6.4028e-09, 0.0000e+00, ..., 1.0250e-07, + -5.0873e-08, -1.5367e-08], + [ 1.7986e-08, 1.0943e-08, 0.0000e+00, ..., 2.7940e-08, + 1.1816e-08, 1.1816e-08], + ..., + [ 1.4959e-08, -1.5192e-08, 0.0000e+00, ..., 6.7521e-09, + -1.3970e-09, -1.0536e-08], + [-5.3272e-07, 3.5507e-09, 0.0000e+00, ..., -1.5076e-07, + 1.3388e-08, 5.0641e-09], + [ 2.1653e-08, 3.7253e-09, 0.0000e+00, ..., 1.7171e-08, + 5.5297e-09, -6.9849e-10]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0364, 0.0132, 0.0147, 0.0238, 0.0316, -0.0180, 0.0253, 0.0303, + -0.0099, 0.0034], device='cuda:0'), grad: tensor([-2.1863e-07, 5.1688e-07, 1.9383e-07, -1.8545e-07, 2.2491e-07, + 4.1202e-06, -2.8294e-06, 5.9430e-08, -2.0098e-06, 1.2061e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 250.48, cls_loss 0.0009 cls_loss_mapping 0.0021 cls_loss_causal 0.4632 re_mapping 0.0028 re_causal 0.0091 /// teacc 99.06 lr 0.00010000 +Epoch 314, weight, value: tensor([[ 0.0892, -0.1600, -0.1394, ..., -0.1307, -0.1703, -0.1550], + [ 0.0417, 0.0549, -0.0352, ..., -0.0666, 0.1683, 0.0205], + [-0.0370, -0.1653, -0.0102, ..., 0.2089, -0.1733, 0.0416], + ..., + [-0.2134, 0.1438, 0.0265, ..., -0.1512, 0.1156, 0.0583], + [ 0.0024, -0.0814, -0.0023, ..., -0.1256, -0.1729, -0.2324], + [ 0.0308, -0.0423, -0.0253, ..., -0.2174, -0.1687, 0.0280]], + device='cuda:0'), grad: tensor([[ 2.5611e-09, 0.0000e+00, 0.0000e+00, ..., 4.4005e-08, + 5.8208e-10, 1.6880e-09], + [ 2.7590e-08, -1.7462e-09, 0.0000e+00, ..., 1.2165e-08, + -6.6357e-09, 2.8056e-08], + [-6.9849e-10, 1.2806e-09, 0.0000e+00, ..., -7.2236e-08, + 6.9849e-10, -4.3656e-08], + ..., + [ 6.3446e-09, 3.4925e-10, 0.0000e+00, ..., 5.8557e-08, + 1.2224e-09, 5.9779e-08], + [ 1.8626e-09, 1.9791e-09, 0.0000e+00, ..., 1.0419e-08, + 6.0536e-09, 1.6938e-08], + [-4.8778e-08, 3.4925e-10, 0.0000e+00, ..., 8.4110e-08, + 4.0745e-10, 1.8557e-07]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0364, 0.0165, 0.0149, 0.0238, 0.0285, -0.0181, 0.0253, 0.0299, + -0.0102, 0.0029], device='cuda:0'), grad: tensor([ 5.9430e-08, 1.6601e-07, -6.8103e-08, -1.0309e-07, -7.6368e-07, + 5.3959e-08, -1.4529e-07, 2.1153e-07, 3.8825e-08, 5.3830e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 313, time 250.56, cls_loss 0.0009 cls_loss_mapping 0.0017 cls_loss_causal 0.4680 re_mapping 0.0027 re_causal 0.0089 /// teacc 98.96 lr 0.00010000 +Epoch 315, weight, value: tensor([[ 0.0889, -0.1615, -0.1394, ..., -0.1313, -0.1714, -0.1557], + [ 0.0417, 0.0549, -0.0351, ..., -0.0678, 0.1683, 0.0204], + [-0.0386, -0.1654, -0.0102, ..., 0.2088, -0.1743, 0.0415], + ..., + [-0.2137, 0.1439, 0.0265, ..., -0.1513, 0.1157, 0.0583], + [ 0.0026, -0.0814, -0.0023, ..., -0.1247, -0.1728, -0.2333], + [ 0.0310, -0.0425, -0.0253, ..., -0.2184, -0.1690, 0.0276]], + device='cuda:0'), grad: tensor([[ 4.8475e-07, 4.6566e-10, 0.0000e+00, ..., 1.2806e-09, + 6.9849e-10, 1.0477e-09], + [-1.5600e-08, -2.4564e-08, 0.0000e+00, ..., 1.6065e-08, + -1.2538e-07, 7.4739e-08], + [ 8.9407e-08, 5.1223e-09, 0.0000e+00, ..., -6.0652e-08, + 6.8685e-09, -3.8650e-08], + ..., + [ 2.2934e-08, 6.8569e-08, 0.0000e+00, ..., 2.3283e-08, + 3.3295e-08, 2.1106e-07], + [-7.9907e-07, 4.7497e-08, 0.0000e+00, ..., 5.9954e-08, + 9.5810e-08, 8.7777e-08], + [ 1.8510e-07, 5.4366e-08, 0.0000e+00, ..., 2.1886e-08, + 2.5379e-08, 1.5472e-07]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0371, 0.0164, 0.0144, 0.0239, 0.0286, -0.0182, 0.0257, 0.0298, + -0.0098, 0.0027], device='cuda:0'), grad: tensor([ 1.4137e-06, 1.1362e-07, 1.5018e-07, 1.8976e-08, -9.1782e-07, + 3.1199e-08, 2.7590e-08, 4.6240e-07, -2.1253e-06, 8.4611e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 250.58, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4766 re_mapping 0.0027 re_causal 0.0088 /// teacc 99.08 lr 0.00010000 +Epoch 316, weight, value: tensor([[ 0.0883, -0.1627, -0.1395, ..., -0.1314, -0.1696, -0.1581], + [ 0.0417, 0.0547, -0.0351, ..., -0.0711, 0.1682, 0.0200], + [-0.0392, -0.1658, -0.0102, ..., 0.2091, -0.1749, 0.0414], + ..., + [-0.2141, 0.1444, 0.0265, ..., -0.1514, 0.1163, 0.0584], + [ 0.0027, -0.0815, -0.0021, ..., -0.1253, -0.1728, -0.2340], + [ 0.0320, -0.0428, -0.0252, ..., -0.2195, -0.1695, 0.0282]], + device='cuda:0'), grad: tensor([[-5.1223e-09, 3.4925e-10, 0.0000e+00, ..., 4.0745e-09, + 3.4925e-10, 2.3982e-08], + [-1.9791e-09, 1.4901e-08, 0.0000e+00, ..., 1.3155e-08, + 6.5193e-09, 2.6426e-08], + [ 4.8894e-09, 8.1491e-09, 0.0000e+00, ..., -1.6473e-07, + 6.0536e-09, -8.9291e-08], + ..., + [ 9.0804e-09, -4.4936e-08, 0.0000e+00, ..., 1.3155e-08, + -4.3772e-08, -4.0745e-09], + [ 6.7521e-09, 4.4238e-09, 0.0000e+00, ..., 1.1118e-07, + 7.5670e-09, 8.2538e-08], + [-9.6625e-09, 4.7730e-09, 0.0000e+00, ..., 4.1910e-09, + 4.6566e-09, 3.4110e-08]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0382, 0.0159, 0.0143, 0.0238, 0.0291, -0.0182, 0.0256, 0.0299, + -0.0097, 0.0033], device='cuda:0'), grad: tensor([ 5.3318e-08, 1.2980e-07, -3.2969e-07, -1.0943e-07, -2.3935e-07, + 9.3132e-09, 1.2689e-08, 5.0059e-08, 3.1060e-07, 1.2526e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 250.29, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4912 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.06 lr 0.00010000 +Epoch 317, weight, value: tensor([[ 0.0885, -0.1632, -0.1395, ..., -0.1314, -0.1694, -0.1587], + [ 0.0417, 0.0547, -0.0352, ..., -0.0712, 0.1683, 0.0200], + [-0.0400, -0.1659, -0.0102, ..., 0.2093, -0.1754, 0.0414], + ..., + [-0.2145, 0.1445, 0.0266, ..., -0.1515, 0.1165, 0.0584], + [ 0.0026, -0.0815, -0.0021, ..., -0.1267, -0.1729, -0.2347], + [ 0.0321, -0.0429, -0.0251, ..., -0.2220, -0.1702, 0.0268]], + device='cuda:0'), grad: tensor([[ 7.3342e-09, 8.1491e-10, 0.0000e+00, ..., 8.9640e-09, + 5.2387e-09, 2.1653e-08], + [-3.3528e-08, -7.5321e-08, 0.0000e+00, ..., 1.6298e-07, + -1.7264e-07, 4.8848e-07], + [ 2.1770e-08, 3.8417e-09, 0.0000e+00, ..., -6.7288e-08, + 1.6415e-08, -3.9581e-09], + ..., + [ 4.4936e-08, 3.2247e-08, 0.0000e+00, ..., 6.9500e-08, + 1.0501e-07, 2.9081e-07], + [-1.4564e-07, 3.0035e-08, 0.0000e+00, ..., -8.4518e-08, + 1.0571e-07, 6.2399e-08], + [ 2.9076e-06, 4.6566e-09, 0.0000e+00, ..., 1.1623e-05, + 4.6007e-06, 4.5955e-05]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0382, 0.0158, 0.0142, 0.0238, 0.0294, -0.0182, 0.0243, 0.0299, + -0.0098, 0.0018], device='cuda:0'), grad: tensor([ 1.1420e-07, 2.0675e-06, 1.8743e-08, 3.1339e-07, -1.9264e-04, + 2.0920e-07, 2.4610e-07, 1.2051e-06, -4.1118e-07, 1.8883e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 250.49, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4607 re_mapping 0.0028 re_causal 0.0096 /// teacc 98.98 lr 0.00010000 +Epoch 318, weight, value: tensor([[ 0.0889, -0.1611, -0.1395, ..., -0.1315, -0.1696, -0.1587], + [ 0.0417, 0.0547, -0.0353, ..., -0.0712, 0.1683, 0.0200], + [-0.0400, -0.1659, -0.0102, ..., 0.2096, -0.1754, 0.0415], + ..., + [-0.2148, 0.1447, 0.0267, ..., -0.1516, 0.1166, 0.0584], + [ 0.0026, -0.0816, -0.0021, ..., -0.1274, -0.1731, -0.2357], + [ 0.0319, -0.0430, -0.0251, ..., -0.2229, -0.1707, 0.0263]], + device='cuda:0'), grad: tensor([[ 2.5611e-09, 2.3283e-10, 0.0000e+00, ..., 2.7823e-08, + 6.9849e-10, 5.8208e-10], + [-2.2119e-08, -1.5483e-08, 0.0000e+00, ..., 5.9372e-09, + -5.8906e-08, -3.2480e-08], + [ 6.1700e-09, 6.1700e-09, 0.0000e+00, ..., -5.4715e-09, + 1.7928e-08, 6.2864e-09], + ..., + [ 3.8417e-09, 2.7940e-09, 0.0000e+00, ..., 1.3970e-09, + 9.7789e-09, 6.8685e-09], + [ 5.7044e-09, 4.1910e-09, 0.0000e+00, ..., 2.0955e-09, + 1.4319e-08, 8.6147e-09], + [ 9.3132e-10, 1.9791e-09, 0.0000e+00, ..., 1.0477e-09, + 3.9581e-09, 2.5611e-09]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0379, 0.0158, 0.0142, 0.0244, 0.0295, -0.0188, 0.0244, 0.0299, + -0.0098, 0.0013], device='cuda:0'), grad: tensor([ 4.9244e-08, -1.0210e-07, 2.5029e-08, 6.0536e-09, 7.4855e-08, + 3.2480e-08, -1.3842e-07, 2.6659e-08, 3.4925e-08, 1.2224e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 250.34, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4601 re_mapping 0.0027 re_causal 0.0090 /// teacc 98.98 lr 0.00010000 +Epoch 319, weight, value: tensor([[ 0.0885, -0.1609, -0.1395, ..., -0.1318, -0.1700, -0.1604], + [ 0.0417, 0.0547, -0.0353, ..., -0.0713, 0.1683, 0.0200], + [-0.0405, -0.1660, -0.0103, ..., 0.2098, -0.1757, 0.0415], + ..., + [-0.2152, 0.1449, 0.0267, ..., -0.1517, 0.1170, 0.0585], + [ 0.0027, -0.0818, -0.0021, ..., -0.1271, -0.1734, -0.2372], + [ 0.0323, -0.0431, -0.0251, ..., -0.2238, -0.1709, 0.0262]], + device='cuda:0'), grad: tensor([[-3.0547e-07, 1.0012e-08, 0.0000e+00, ..., 4.8894e-09, + 1.8161e-08, 1.3621e-08], + [-1.3865e-07, 4.1956e-07, 0.0000e+00, ..., 3.2596e-09, + 3.1665e-07, 6.3051e-07], + [ 2.7125e-08, 2.3039e-07, 0.0000e+00, ..., 6.0536e-09, + 3.0478e-07, 3.1269e-07], + ..., + [ 1.8161e-08, -9.4697e-06, 0.0000e+00, ..., 2.2119e-09, + -1.2614e-05, -1.3284e-05], + [ 1.1129e-07, 8.6129e-06, 0.0000e+00, ..., 6.7521e-09, + 1.1705e-05, 1.2062e-05], + [ 2.5635e-07, 7.9744e-08, 0.0000e+00, ..., 1.2806e-09, + 9.2317e-08, 9.3132e-08]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0395, 0.0158, 0.0141, 0.0244, 0.0295, -0.0188, 0.0241, 0.0299, + -0.0094, 0.0013], device='cuda:0'), grad: tensor([-1.1092e-06, 1.6615e-06, 1.1176e-06, 1.2352e-07, 5.0478e-07, + 6.6939e-08, -3.3411e-08, -4.4763e-05, 4.1217e-05, 1.2824e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 250.50, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4555 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.08 lr 0.00010000 +Epoch 320, weight, value: tensor([[ 0.0886, -0.1612, -0.1395, ..., -0.1318, -0.1705, -0.1606], + [ 0.0417, 0.0544, -0.0353, ..., -0.0714, 0.1681, 0.0198], + [-0.0400, -0.1661, -0.0103, ..., 0.2105, -0.1754, 0.0417], + ..., + [-0.2146, 0.1470, 0.0267, ..., -0.1518, 0.1201, 0.0598], + [ 0.0026, -0.0820, -0.0021, ..., -0.1290, -0.1739, -0.2415], + [ 0.0315, -0.0434, -0.0251, ..., -0.2258, -0.1716, 0.0248]], + device='cuda:0'), grad: tensor([[ 7.6834e-09, 1.3853e-08, 0.0000e+00, ..., 7.3225e-08, + 1.4051e-07, 5.1921e-08], + [-1.2713e-07, 1.8149e-07, 0.0000e+00, ..., 1.1008e-06, + 2.6599e-06, 9.1037e-07], + [ 8.2655e-09, 5.9954e-08, 0.0000e+00, ..., 3.3760e-08, + 1.0896e-07, 9.4762e-08], + ..., + [ 4.3190e-08, -2.3209e-06, 0.0000e+00, ..., 5.3202e-08, + -1.6950e-06, -3.2075e-06], + [-6.3214e-08, 1.1583e-07, 0.0000e+00, ..., 3.6485e-07, + 8.5030e-07, 2.7614e-07], + [ 1.8394e-08, 2.3050e-08, 0.0000e+00, ..., 8.8057e-07, + 1.0070e-07, 5.7463e-07]], device='cuda:0') +Epoch 320, bias, value: tensor([-3.9511e-02, 1.5646e-02, 1.4300e-02, 2.4448e-02, 2.9719e-02, + -1.8779e-02, 2.4179e-02, 3.1213e-02, -9.8676e-03, 4.8556e-05], + device='cuda:0'), grad: tensor([ 1.2405e-06, 2.3559e-05, 7.1898e-07, 8.0327e-07, 4.2245e-06, + 5.6289e-06, -3.9518e-05, -5.4836e-06, 5.4538e-06, 3.3639e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 250.67, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4731 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.03 lr 0.00010000 +Epoch 321, weight, value: tensor([[ 0.0884, -0.1617, -0.1395, ..., -0.1320, -0.1710, -0.1614], + [ 0.0417, 0.0544, -0.0357, ..., -0.0715, 0.1681, 0.0198], + [-0.0404, -0.1664, -0.0104, ..., 0.2107, -0.1757, 0.0416], + ..., + [-0.2152, 0.1471, 0.0270, ..., -0.1519, 0.1202, 0.0598], + [ 0.0026, -0.0821, -0.0022, ..., -0.1292, -0.1740, -0.2428], + [ 0.0309, -0.0435, -0.0254, ..., -0.2275, -0.1720, 0.0236]], + device='cuda:0'), grad: tensor([[-9.5111e-08, 5.8208e-10, 0.0000e+00, ..., 1.2666e-07, + 1.2806e-09, 2.1106e-07], + [ 1.9441e-08, 3.9465e-07, 0.0000e+00, ..., 3.1991e-07, + 3.1781e-07, 9.4110e-07], + [ 2.1420e-08, 4.2142e-08, 0.0000e+00, ..., -7.4226e-07, + 3.8301e-08, -1.2806e-06], + ..., + [ 1.6298e-08, -2.9393e-06, 0.0000e+00, ..., 6.0536e-08, + -2.5332e-06, -2.7474e-06], + [-7.9046e-08, 1.8382e-07, 0.0000e+00, ..., 8.5216e-08, + 1.7113e-07, 3.1129e-07], + [ 7.9395e-08, 2.3004e-06, 0.0000e+00, ..., 4.9709e-08, + 1.9893e-06, 2.3060e-06]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0400, 0.0156, 0.0142, 0.0245, 0.0300, -0.0188, 0.0243, 0.0311, + -0.0099, -0.0010], device='cuda:0'), grad: tensor([ 4.5658e-07, 3.0994e-06, -4.5896e-06, 1.0803e-07, 5.2806e-07, + 2.6310e-07, 7.4622e-08, -6.1244e-06, 4.4703e-07, 5.7444e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 250.21, cls_loss 0.0009 cls_loss_mapping 0.0022 cls_loss_causal 0.4985 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.13 lr 0.00010000 +Epoch 322, weight, value: tensor([[ 0.0886, -0.1626, -0.1395, ..., -0.1322, -0.1718, -0.1618], + [ 0.0407, 0.0518, -0.0362, ..., -0.0715, 0.1658, 0.0184], + [-0.0408, -0.1666, -0.0103, ..., 0.2111, -0.1760, 0.0416], + ..., + [-0.2124, 0.1506, 0.0274, ..., -0.1521, 0.1235, 0.0630], + [ 0.0027, -0.0824, -0.0022, ..., -0.1295, -0.1744, -0.2448], + [ 0.0293, -0.0436, -0.0251, ..., -0.2290, -0.1732, 0.0236]], + device='cuda:0'), grad: tensor([[ 1.6415e-08, 9.8953e-09, 0.0000e+00, ..., 3.1083e-08, + 2.3167e-08, 1.8743e-08], + [-2.4540e-07, -1.6915e-07, 0.0000e+00, ..., 8.7311e-08, + -4.6799e-07, 1.1432e-07], + [-2.4796e-08, 9.8953e-09, 0.0000e+00, ..., -3.0058e-07, + 2.3749e-08, -1.8021e-07], + ..., + [ 8.3353e-08, 4.8429e-08, 0.0000e+00, ..., 1.1921e-07, + 1.4156e-07, 1.1467e-07], + [ 1.4273e-07, 8.2655e-08, 0.0000e+00, ..., 4.9593e-08, + 2.5495e-07, 5.4133e-08], + [ 2.2119e-09, 3.6089e-09, 0.0000e+00, ..., 1.2107e-08, + 6.9849e-09, 2.2468e-08]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0401, 0.0138, 0.0142, 0.0245, 0.0301, -0.0187, 0.0240, 0.0342, + -0.0099, -0.0013], device='cuda:0'), grad: tensor([ 1.1385e-07, -1.9267e-07, -7.5158e-07, 1.9325e-07, -3.9767e-07, + 1.9209e-08, -3.3644e-08, 5.3411e-07, 4.5821e-07, 6.9849e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 250.79, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4681 re_mapping 0.0031 re_causal 0.0098 /// teacc 98.87 lr 0.00010000 +Epoch 323, weight, value: tensor([[ 0.0886, -0.1629, -0.1395, ..., -0.1322, -0.1729, -0.1621], + [ 0.0407, 0.0518, -0.0361, ..., -0.0716, 0.1658, 0.0184], + [-0.0406, -0.1662, -0.0103, ..., 0.2123, -0.1761, 0.0420], + ..., + [-0.2124, 0.1506, 0.0273, ..., -0.1526, 0.1235, 0.0629], + [ 0.0030, -0.0825, -0.0022, ..., -0.1296, -0.1745, -0.2457], + [ 0.0295, -0.0436, -0.0245, ..., -0.2291, -0.1733, 0.0236]], + device='cuda:0'), grad: tensor([[-8.4983e-09, -2.3283e-09, 0.0000e+00, ..., 1.0594e-08, + -3.6089e-09, 5.7044e-09], + [-2.3283e-10, -2.6776e-09, 0.0000e+00, ..., 1.0943e-08, + -1.3039e-08, 6.9849e-10], + [-4.1560e-08, 4.6566e-10, 0.0000e+00, ..., -3.7439e-07, + -2.6077e-08, -3.5320e-07], + ..., + [ 3.1665e-08, 1.7462e-09, 0.0000e+00, ..., 2.5635e-07, + 2.4214e-08, 2.4447e-07], + [ 7.7998e-09, 1.8626e-09, 0.0000e+00, ..., 2.0023e-08, + 8.2655e-09, 2.0140e-08], + [ 1.2806e-09, 3.4925e-10, 0.0000e+00, ..., 2.5611e-09, + 9.3132e-10, 2.0955e-09]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0401, 0.0138, 0.0146, 0.0244, 0.0301, -0.0188, 0.0239, 0.0341, + -0.0097, -0.0013], device='cuda:0'), grad: tensor([-3.6089e-09, 2.6776e-08, -1.2759e-06, 2.8126e-07, 2.0210e-07, + -4.6566e-08, -1.9430e-07, 8.8755e-07, 1.1455e-07, 1.6065e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 250.49, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4510 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.05 lr 0.00010000 +Epoch 324, weight, value: tensor([[ 0.0893, -0.1632, -0.1395, ..., -0.1324, -0.1707, -0.1623], + [ 0.0406, 0.0518, -0.0361, ..., -0.0716, 0.1659, 0.0184], + [-0.0407, -0.1663, -0.0103, ..., 0.2126, -0.1763, 0.0421], + ..., + [-0.2124, 0.1506, 0.0273, ..., -0.1528, 0.1235, 0.0629], + [ 0.0031, -0.0825, -0.0022, ..., -0.1300, -0.1746, -0.2466], + [ 0.0308, -0.0437, -0.0244, ..., -0.2294, -0.1736, 0.0235]], + device='cuda:0'), grad: tensor([[-5.3691e-07, 2.0955e-09, 0.0000e+00, ..., 1.5018e-08, + 3.4925e-10, 8.1491e-10], + [-1.3039e-08, -5.1223e-09, 0.0000e+00, ..., 2.8173e-08, + -5.0408e-08, -2.3167e-08], + [ 1.5949e-08, 7.7998e-09, 0.0000e+00, ..., 5.9721e-08, + 3.4925e-09, 6.5193e-09], + ..., + [ 1.4203e-08, -6.0536e-09, 0.0000e+00, ..., 7.9162e-09, + 3.1432e-09, 1.3853e-08], + [ 4.2957e-08, 4.9127e-08, 0.0000e+00, ..., 5.4389e-07, + 3.5740e-08, 2.3516e-08], + [ 4.3772e-08, 1.5134e-09, 0.0000e+00, ..., 2.9220e-08, + 1.6298e-09, 5.3784e-08]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0399, 0.0137, 0.0146, 0.0244, 0.0301, -0.0190, 0.0254, 0.0341, + -0.0096, -0.0010], device='cuda:0'), grad: tensor([-2.4997e-06, 8.3703e-08, 3.2084e-07, 3.7556e-07, -3.0687e-07, + 1.8347e-06, -2.5071e-06, 1.4005e-07, 2.0973e-06, 4.5705e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 250.68, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.4778 re_mapping 0.0026 re_causal 0.0087 /// teacc 98.87 lr 0.00010000 +Epoch 325, weight, value: tensor([[ 0.0894, -0.1633, -0.1395, ..., -0.1323, -0.1709, -0.1647], + [ 0.0406, 0.0533, -0.0362, ..., -0.0717, 0.1672, 0.0193], + [-0.0412, -0.1664, -0.0108, ..., 0.2130, -0.1764, 0.0422], + ..., + [-0.2125, 0.1489, 0.0274, ..., -0.1529, 0.1219, 0.0612], + [ 0.0031, -0.0826, -0.0021, ..., -0.1302, -0.1749, -0.2478], + [ 0.0311, -0.0438, -0.0245, ..., -0.2295, -0.1740, 0.0236]], + device='cuda:0'), grad: tensor([[-1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 5.3551e-08, + 1.0477e-09, 1.9791e-09], + [-3.0035e-08, -2.1188e-08, 0.0000e+00, ..., 1.9674e-08, + -7.5088e-08, -3.3644e-08], + [ 1.6298e-09, 3.8417e-09, 0.0000e+00, ..., -1.0477e-08, + 2.3283e-09, -1.3504e-08], + ..., + [ 1.5134e-08, 8.1491e-10, 1.1642e-10, ..., 1.0827e-08, + 2.2468e-08, 1.8510e-08], + [ 7.9162e-09, 9.7789e-09, 0.0000e+00, ..., 3.6205e-08, + 3.1199e-08, 1.9907e-08], + [-1.0361e-08, 3.3760e-09, -6.9849e-10, ..., 8.7311e-09, + 9.0804e-09, -1.5134e-09]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0406, 0.0146, 0.0147, 0.0244, 0.0301, -0.0190, 0.0253, 0.0327, + -0.0096, -0.0008], device='cuda:0'), grad: tensor([ 9.0688e-08, -1.0617e-07, 1.9791e-09, 9.4413e-08, 1.2736e-07, + 8.3470e-08, -4.4145e-07, 8.9291e-08, 1.1479e-07, -3.8301e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 250.66, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4380 re_mapping 0.0027 re_causal 0.0086 /// teacc 98.99 lr 0.00010000 +Epoch 326, weight, value: tensor([[ 0.0896, -0.1636, -0.1396, ..., -0.1324, -0.1705, -0.1649], + [ 0.0406, 0.0533, -0.0344, ..., -0.0718, 0.1672, 0.0193], + [-0.0418, -0.1665, -0.0110, ..., 0.2131, -0.1767, 0.0422], + ..., + [-0.2125, 0.1489, 0.0264, ..., -0.1530, 0.1219, 0.0611], + [ 0.0031, -0.0828, -0.0024, ..., -0.1302, -0.1751, -0.2489], + [ 0.0314, -0.0436, -0.0247, ..., -0.2298, -0.1742, 0.0235]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.4925e-10, 0.0000e+00, ..., 5.5879e-09, + 1.9791e-09, 8.0327e-09], + [-2.1653e-08, -1.2689e-08, 0.0000e+00, ..., 3.5274e-08, + -5.0641e-08, 3.4692e-08], + [-1.6298e-09, 1.1642e-09, 0.0000e+00, ..., -8.4168e-08, + -1.9791e-09, -7.6019e-08], + ..., + [ 7.1013e-09, 3.7253e-09, 0.0000e+00, ..., 8.7777e-08, + 1.7229e-08, 1.0629e-07], + [ 1.1758e-08, 5.0059e-09, 0.0000e+00, ..., 8.0909e-08, + 2.5961e-08, 9.5693e-08], + [-4.9360e-08, 8.1491e-10, 0.0000e+00, ..., 1.7090e-06, + -1.4040e-07, 3.0566e-06]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0406, 0.0146, 0.0146, 0.0243, 0.0301, -0.0188, 0.0253, 0.0326, + -0.0097, -0.0008], device='cuda:0'), grad: tensor([ 5.3900e-08, 1.2887e-07, 5.1456e-08, -9.6299e-07, -1.2994e-05, + 3.3760e-07, 9.9838e-07, 5.2433e-07, 3.7858e-07, 1.1466e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 250.22, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4466 re_mapping 0.0028 re_causal 0.0096 /// teacc 98.99 lr 0.00010000 +Epoch 327, weight, value: tensor([[ 0.0896, -0.1638, -0.1396, ..., -0.1327, -0.1703, -0.1650], + [ 0.0406, 0.0533, -0.0342, ..., -0.0718, 0.1672, 0.0193], + [-0.0421, -0.1667, -0.0119, ..., 0.2132, -0.1769, 0.0421], + ..., + [-0.2126, 0.1489, 0.0266, ..., -0.1530, 0.1219, 0.0611], + [ 0.0031, -0.0828, -0.0025, ..., -0.1304, -0.1752, -0.2495], + [ 0.0316, -0.0436, -0.0248, ..., -0.2300, -0.1744, 0.0235]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 2.2119e-09, 0.0000e+00, ..., 1.1642e-10, + 3.2596e-09, 2.6776e-09], + [-1.7579e-08, 6.0885e-08, 0.0000e+00, ..., 5.1223e-09, + 5.4948e-08, 7.0082e-08], + [ 1.9791e-09, 4.5984e-08, 0.0000e+00, ..., 2.4447e-09, + 5.9721e-08, 5.4366e-08], + ..., + [ 6.7521e-09, -2.8010e-07, 0.0000e+00, ..., 1.2689e-08, + -3.4808e-07, -3.0524e-07], + [ 1.7346e-08, 3.3178e-08, 0.0000e+00, ..., 4.6566e-10, + 5.0757e-08, 3.9116e-08], + [ 4.3074e-09, 1.0524e-07, 0.0000e+00, ..., 1.9092e-08, + 1.3446e-07, 1.3481e-07]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0408, 0.0146, 0.0145, 0.0244, 0.0301, -0.0189, 0.0251, 0.0326, + -0.0096, -0.0007], device='cuda:0'), grad: tensor([ 1.0012e-08, 1.7055e-07, 1.4948e-07, 7.4506e-08, -3.9814e-08, + -1.3458e-07, 3.8999e-08, -8.0187e-07, 1.6671e-07, 3.8021e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 250.48, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4857 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.05 lr 0.00010000 +Epoch 328, weight, value: tensor([[ 0.0898, -0.1642, -0.1396, ..., -0.1328, -0.1705, -0.1653], + [ 0.0407, 0.0533, -0.0336, ..., -0.0719, 0.1673, 0.0193], + [-0.0438, -0.1670, -0.0120, ..., 0.2139, -0.1777, 0.0422], + ..., + [-0.2127, 0.1489, 0.0264, ..., -0.1532, 0.1219, 0.0611], + [ 0.0031, -0.0831, -0.0026, ..., -0.1312, -0.1757, -0.2513], + [ 0.0323, -0.0435, -0.0259, ..., -0.2303, -0.1744, 0.0235]], + device='cuda:0'), grad: tensor([[ 2.3749e-08, 2.3283e-08, 0.0000e+00, ..., 8.8476e-09, + 4.8545e-08, 3.5041e-08], + [-4.2655e-07, -6.7754e-08, 0.0000e+00, ..., 8.2189e-08, + -5.2201e-07, -1.5332e-07], + [ 4.1327e-08, 6.0303e-08, 0.0000e+00, ..., -2.3376e-07, + 1.1199e-07, -1.7649e-07], + ..., + [ 7.5088e-08, -5.0943e-07, 0.0000e+00, ..., 5.6694e-08, + -5.4715e-07, -2.5402e-07], + [ 1.4342e-07, 4.9826e-08, 0.0000e+00, ..., 1.8626e-08, + 2.0082e-07, 1.1572e-07], + [ 5.9372e-08, 3.2829e-07, 0.0000e+00, ..., 7.7533e-08, + 4.4541e-07, 6.0676e-07]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0407, 0.0147, 0.0144, 0.0246, 0.0301, -0.0190, 0.0252, 0.0326, + -0.0097, -0.0007], device='cuda:0'), grad: tensor([ 2.0175e-07, -1.1269e-06, -4.1584e-07, 9.1270e-07, -4.8755e-07, + -9.4390e-07, 2.8638e-07, -1.3057e-06, 7.7859e-07, 2.1067e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 250.36, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4762 re_mapping 0.0027 re_causal 0.0089 /// teacc 99.14 lr 0.00010000 +Epoch 329, weight, value: tensor([[ 0.0902, -0.1646, -0.1396, ..., -0.1330, -0.1713, -0.1660], + [ 0.0408, 0.0534, -0.0333, ..., -0.0721, 0.1673, 0.0194], + [-0.0467, -0.1674, -0.0122, ..., 0.2153, -0.1783, 0.0425], + ..., + [-0.2127, 0.1489, 0.0266, ..., -0.1535, 0.1219, 0.0611], + [ 0.0030, -0.0835, -0.0027, ..., -0.1316, -0.1762, -0.2531], + [ 0.0326, -0.0437, -0.0261, ..., -0.2308, -0.1750, 0.0232]], + device='cuda:0'), grad: tensor([[-1.5134e-09, 3.3760e-09, 0.0000e+00, ..., 5.5507e-07, + 6.0536e-09, 7.2177e-09], + [-6.1118e-08, 3.5623e-08, 0.0000e+00, ..., 5.2154e-08, + -3.3877e-08, 4.4936e-08], + [ 1.0012e-08, 2.4796e-08, 0.0000e+00, ..., 1.2340e-08, + 4.8196e-08, 3.6438e-08], + ..., + [ 1.9441e-08, -2.5565e-07, 0.0000e+00, ..., 4.9709e-08, + -3.9046e-07, -1.4761e-07], + [ 5.0757e-08, 1.8859e-08, 0.0000e+00, ..., 4.1677e-08, + 8.3586e-08, 5.8906e-08], + [ 2.4564e-08, 1.5099e-07, 0.0000e+00, ..., 1.1979e-07, + 2.4377e-07, 2.8359e-07]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0407, 0.0147, 0.0146, 0.0245, 0.0302, -0.0189, 0.0251, 0.0326, + -0.0099, -0.0008], device='cuda:0'), grad: tensor([ 2.0582e-06, 2.2096e-07, 1.7963e-07, -1.6298e-09, -8.4704e-07, + -8.8336e-07, -2.2985e-06, -3.1153e-07, 4.8429e-07, 1.4044e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 250.88, cls_loss 0.0012 cls_loss_mapping 0.0026 cls_loss_causal 0.4595 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.20 lr 0.00010000 +Epoch 330, weight, value: tensor([[ 0.0905, -0.1650, -0.1401, ..., -0.1354, -0.1711, -0.1665], + [ 0.0408, 0.0533, -0.0336, ..., -0.0722, 0.1673, 0.0193], + [-0.0474, -0.1677, -0.0135, ..., 0.2151, -0.1789, 0.0424], + ..., + [-0.2128, 0.1490, 0.0275, ..., -0.1537, 0.1219, 0.0611], + [ 0.0030, -0.0836, -0.0050, ..., -0.1324, -0.1764, -0.2543], + [ 0.0323, -0.0438, -0.0271, ..., -0.2318, -0.1754, 0.0212]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 6.9849e-10], + [ 1.4785e-08, 1.2340e-08, 0.0000e+00, ..., 6.2981e-08, + 1.6298e-09, 4.2724e-08], + [ 6.9849e-10, 1.2806e-09, 0.0000e+00, ..., -1.7043e-07, + 1.5134e-09, -1.4179e-07], + ..., + [ 4.5402e-09, 3.2596e-09, 0.0000e+00, ..., 8.1374e-08, + 8.4983e-09, 7.3109e-08], + [-2.0373e-08, -2.0373e-08, 0.0000e+00, ..., 2.0606e-08, + -1.6415e-08, 1.9907e-08], + [-2.7940e-09, 9.3132e-10, 0.0000e+00, ..., 5.8208e-10, + 1.0477e-09, -4.0745e-09]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0409, 0.0146, 0.0141, 0.0245, 0.0300, -0.0189, 0.0282, 0.0326, + -0.0101, -0.0013], device='cuda:0'), grad: tensor([ 2.2119e-09, 2.0757e-07, -4.2305e-07, 1.0245e-08, 2.1420e-08, + 5.5879e-09, 4.3074e-09, 2.2410e-07, -2.2468e-08, -1.8044e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 250.52, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4569 re_mapping 0.0027 re_causal 0.0092 /// teacc 99.18 lr 0.00010000 +Epoch 331, weight, value: tensor([[ 0.0905, -0.1655, -0.1402, ..., -0.1355, -0.1690, -0.1665], + [ 0.0408, 0.0533, -0.0337, ..., -0.0722, 0.1673, 0.0193], + [-0.0477, -0.1679, -0.0137, ..., 0.2152, -0.1795, 0.0423], + ..., + [-0.2128, 0.1490, 0.0277, ..., -0.1537, 0.1220, 0.0612], + [ 0.0031, -0.0837, -0.0046, ..., -0.1322, -0.1765, -0.2549], + [ 0.0323, -0.0440, -0.0272, ..., -0.2320, -0.1760, 0.0212]], + device='cuda:0'), grad: tensor([[ 1.8394e-08, 1.1642e-10, 0.0000e+00, ..., 1.8626e-08, + 2.3283e-10, 1.0477e-09], + [ 1.8044e-08, 1.5134e-08, 0.0000e+00, ..., 2.4913e-08, + 1.0594e-08, 2.2934e-08], + [ 2.5495e-08, 4.9942e-08, 0.0000e+00, ..., 5.8208e-08, + 5.5763e-08, 7.9744e-08], + ..., + [ 4.5402e-09, -2.4913e-07, 0.0000e+00, ..., -1.8405e-07, + -2.7288e-07, -3.6485e-07], + [-7.4622e-08, 3.0268e-09, 0.0000e+00, ..., -9.5344e-08, + 7.4506e-09, 9.0804e-09], + [-9.6275e-08, 1.7346e-08, 0.0000e+00, ..., 1.4552e-08, + 1.9674e-08, 2.5029e-08]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0409, 0.0146, 0.0138, 0.0245, 0.0300, -0.0189, 0.0280, 0.0326, + -0.0100, -0.0013], device='cuda:0'), grad: tensor([ 9.9302e-08, 1.9581e-07, 3.7393e-07, -8.6753e-07, 7.0315e-07, + 4.1490e-07, 2.4610e-07, -5.3179e-07, -3.9465e-07, -2.2817e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 250.30, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4719 re_mapping 0.0027 re_causal 0.0093 /// teacc 99.11 lr 0.00010000 +Epoch 332, weight, value: tensor([[ 0.0905, -0.1654, -0.1402, ..., -0.1356, -0.1686, -0.1665], + [ 0.0408, 0.0533, -0.0337, ..., -0.0717, 0.1674, 0.0194], + [-0.0477, -0.1681, -0.0137, ..., 0.2138, -0.1824, 0.0413], + ..., + [-0.2129, 0.1490, 0.0277, ..., -0.1540, 0.1220, 0.0612], + [ 0.0032, -0.0837, -0.0046, ..., -0.1326, -0.1766, -0.2554], + [ 0.0323, -0.0443, -0.0273, ..., -0.2324, -0.1765, 0.0212]], + device='cuda:0'), grad: tensor([[-6.1700e-09, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, + 2.3283e-10, 5.8208e-10], + [-8.6147e-09, -3.7253e-09, 0.0000e+00, ..., 5.5879e-09, + -1.8277e-08, -5.3551e-09], + [ 4.6566e-10, 2.2119e-09, 0.0000e+00, ..., -1.9791e-08, + 2.7940e-09, -1.3388e-08], + ..., + [ 5.7044e-09, -1.3504e-08, 0.0000e+00, ..., 6.5193e-09, + -8.9640e-09, -5.8208e-10], + [ 3.3760e-09, 5.7044e-09, 0.0000e+00, ..., 6.5193e-09, + 1.2107e-08, 1.1874e-08], + [-3.4925e-09, 4.0745e-09, 0.0000e+00, ..., 1.3970e-09, + 4.8894e-09, -1.9092e-08]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0409, 0.0147, 0.0128, 0.0245, 0.0300, -0.0189, 0.0281, 0.0326, + -0.0099, -0.0013], device='cuda:0'), grad: tensor([-2.3399e-08, -7.7998e-09, -3.7602e-08, -4.9127e-08, 1.0955e-07, + 3.6554e-08, -5.3435e-08, 3.2131e-08, 4.0396e-08, -4.5286e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 250.55, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4931 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.08 lr 0.00010000 +Epoch 333, weight, value: tensor([[ 0.0905, -0.1660, -0.1403, ..., -0.1357, -0.1688, -0.1665], + [ 0.0408, 0.0533, -0.0337, ..., -0.0718, 0.1675, 0.0194], + [-0.0481, -0.1693, -0.0151, ..., 0.2136, -0.1835, 0.0407], + ..., + [-0.2129, 0.1491, 0.0282, ..., -0.1539, 0.1220, 0.0613], + [ 0.0032, -0.0839, -0.0045, ..., -0.1330, -0.1768, -0.2564], + [ 0.0323, -0.0444, -0.0277, ..., -0.2340, -0.1769, 0.0209]], + device='cuda:0'), grad: tensor([[-1.2841e-07, 1.6298e-09, 0.0000e+00, ..., -2.6659e-08, + 3.6089e-09, 3.3760e-09], + [-8.5682e-08, 1.4552e-08, 0.0000e+00, ..., 2.0955e-09, + -2.3609e-07, -1.3073e-07], + [ 1.8510e-08, 1.7462e-08, 0.0000e+00, ..., 2.2119e-09, + 3.7253e-08, 3.5390e-08], + ..., + [ 4.1793e-08, -2.3236e-07, 0.0000e+00, ..., 1.8626e-09, + -1.9034e-07, -2.8289e-07], + [-9.8953e-09, 2.0955e-09, 0.0000e+00, ..., -1.1292e-08, + 5.9139e-08, 3.5274e-08], + [ 5.3202e-08, 1.3865e-07, 0.0000e+00, ..., 3.0268e-09, + 1.8335e-07, 2.1153e-07]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0409, 0.0147, 0.0119, 0.0248, 0.0304, -0.0191, 0.0281, 0.0327, + -0.0099, -0.0014], device='cuda:0'), grad: tensor([-8.0373e-07, -3.7812e-07, 1.6962e-07, -9.6043e-08, 3.1758e-07, + 1.0175e-07, 1.6182e-07, -3.5134e-07, -2.6193e-08, 9.2480e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 249.93, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4858 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.07 lr 0.00010000 +Epoch 334, weight, value: tensor([[ 0.0905, -0.1668, -0.1403, ..., -0.1357, -0.1691, -0.1665], + [ 0.0408, 0.0531, -0.0337, ..., -0.0724, 0.1673, 0.0192], + [-0.0481, -0.1708, -0.0151, ..., 0.2136, -0.1848, 0.0400], + ..., + [-0.2129, 0.1495, 0.0282, ..., -0.1518, 0.1224, 0.0618], + [ 0.0031, -0.0842, -0.0048, ..., -0.1337, -0.1772, -0.2581], + [ 0.0323, -0.0448, -0.0278, ..., -0.2342, -0.1777, 0.0209]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 5.8208e-10, + 0.0000e+00, 3.4925e-10], + [ 8.6147e-09, -1.6298e-09, 0.0000e+00, ..., 1.5134e-09, + -6.7521e-09, -2.4447e-09], + [ 1.6298e-09, 1.1642e-10, -1.1642e-10, ..., -1.9907e-08, + 2.3283e-10, -2.0023e-08], + ..., + [ 7.1013e-09, 6.9849e-10, 1.1642e-10, ..., 1.6531e-08, + 2.7940e-09, 1.9558e-08], + [-4.1560e-07, 1.1642e-09, 0.0000e+00, ..., -3.4925e-10, + 5.0059e-09, 1.9791e-09], + [ 3.7253e-08, 1.1642e-10, 0.0000e+00, ..., 1.0477e-09, + 2.3283e-10, -3.4925e-09]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0409, 0.0145, 0.0111, 0.0247, 0.0304, -0.0190, 0.0280, 0.0331, + -0.0102, -0.0015], device='cuda:0'), grad: tensor([ 2.5611e-09, 4.3074e-08, -4.2142e-08, 5.4715e-09, 1.6880e-08, + 1.3616e-06, 6.1817e-08, 7.9046e-08, -1.7304e-06, 2.2200e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 250.40, cls_loss 0.0007 cls_loss_mapping 0.0014 cls_loss_causal 0.4885 re_mapping 0.0026 re_causal 0.0091 /// teacc 99.00 lr 0.00010000 +Epoch 335, weight, value: tensor([[ 0.0905, -0.1676, -0.1403, ..., -0.1359, -0.1694, -0.1665], + [ 0.0409, 0.0531, -0.0336, ..., -0.0728, 0.1673, 0.0192], + [-0.0485, -0.1709, -0.0152, ..., 0.2139, -0.1850, 0.0400], + ..., + [-0.2130, 0.1496, 0.0282, ..., -0.1524, 0.1224, 0.0618], + [ 0.0030, -0.0846, -0.0049, ..., -0.1342, -0.1777, -0.2603], + [ 0.0323, -0.0456, -0.0278, ..., -0.2345, -0.1801, 0.0209]], + device='cuda:0'), grad: tensor([[ 5.2387e-09, 1.3388e-09, 0.0000e+00, ..., 7.2003e-08, + 2.4447e-09, 5.0699e-08], + [-4.1036e-08, -2.5437e-08, 0.0000e+00, ..., 1.3260e-07, + -1.4389e-07, 3.7369e-07], + [ 7.9744e-09, 1.7753e-08, 0.0000e+00, ..., -8.2888e-08, + 2.2934e-08, -2.3283e-10], + ..., + [ 8.1374e-08, -3.5332e-08, 0.0000e+00, ..., 5.3493e-08, + -2.7067e-08, 6.5891e-07], + [ 9.7323e-08, 3.5681e-08, 0.0000e+00, ..., 6.5193e-09, + 9.1328e-08, 2.1281e-07], + [-5.9977e-07, -2.8813e-08, 0.0000e+00, ..., 6.5612e-07, + 1.0186e-08, 7.4552e-07]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0409, 0.0145, 0.0111, 0.0246, 0.0305, -0.0190, 0.0286, 0.0331, + -0.0105, -0.0015], device='cuda:0'), grad: tensor([ 2.5844e-07, 1.2498e-06, -1.1123e-07, 1.2503e-07, -4.6380e-06, + 1.6857e-07, 2.0978e-07, 2.7791e-06, 1.0114e-06, -1.0580e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 250.36, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4456 re_mapping 0.0027 re_causal 0.0086 /// teacc 99.14 lr 0.00010000 +Epoch 336, weight, value: tensor([[ 0.0905, -0.1686, -0.1403, ..., -0.1359, -0.1701, -0.1665], + [ 0.0411, 0.0531, -0.0336, ..., -0.0730, 0.1674, 0.0192], + [-0.0491, -0.1711, -0.0152, ..., 0.2150, -0.1850, 0.0403], + ..., + [-0.2130, 0.1496, 0.0281, ..., -0.1526, 0.1225, 0.0618], + [ 0.0025, -0.0865, -0.0049, ..., -0.1350, -0.1800, -0.2639], + [ 0.0323, -0.0458, -0.0278, ..., -0.2349, -0.1805, 0.0209]], + device='cuda:0'), grad: tensor([[ 2.0373e-09, 1.1642e-10, 0.0000e+00, ..., 1.8044e-09, + 4.0745e-10, 1.9791e-09], + [-2.7241e-08, -2.3283e-08, 0.0000e+00, ..., 6.1700e-08, + -6.6124e-08, 2.8522e-08], + [ 1.9209e-09, 1.2806e-09, 0.0000e+00, ..., 1.1642e-10, + 1.5134e-09, 5.8790e-09], + ..., + [ 1.8452e-08, 1.1176e-08, 0.0000e+00, ..., 3.5914e-08, + 3.2887e-08, 6.3097e-08], + [ 2.1770e-08, 4.0163e-09, 0.0000e+00, ..., 3.0443e-08, + 1.2107e-08, 3.5972e-08], + [-2.9104e-10, 1.8626e-09, 0.0000e+00, ..., 8.5216e-08, + 4.4820e-09, 8.7370e-08]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0409, 0.0146, 0.0113, 0.0246, 0.0305, -0.0190, 0.0286, 0.0331, + -0.0120, -0.0015], device='cuda:0'), grad: tensor([ 3.5856e-08, 1.9628e-07, 1.0896e-07, 1.4193e-06, -1.7360e-06, + -1.9576e-06, 9.2248e-07, 2.8475e-07, 3.5204e-07, 3.7323e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 250.32, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4772 re_mapping 0.0026 re_causal 0.0089 /// teacc 99.12 lr 0.00010000 +Epoch 337, weight, value: tensor([[ 0.0905, -0.1687, -0.1404, ..., -0.1361, -0.1703, -0.1665], + [ 0.0413, 0.0532, -0.0337, ..., -0.0746, 0.1675, 0.0193], + [-0.0468, -0.1715, -0.0150, ..., 0.2156, -0.1828, 0.0394], + ..., + [-0.2131, 0.1496, 0.0263, ..., -0.1530, 0.1224, 0.0617], + [ 0.0025, -0.0864, -0.0050, ..., -0.1365, -0.1801, -0.2643], + [ 0.0321, -0.0473, -0.0248, ..., -0.2358, -0.1833, 0.0207]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 1.6298e-09, + 6.9849e-10, 2.6193e-09], + [-2.3574e-08, -1.3853e-08, 0.0000e+00, ..., 1.3562e-08, + -4.7614e-08, -8.0909e-09], + [ 2.0373e-09, 1.1059e-09, 0.0000e+00, ..., -2.2375e-07, + -4.3074e-09, -3.4529e-07], + ..., + [ 1.0186e-08, 5.7044e-09, 1.1642e-10, ..., 1.4890e-07, + 2.6310e-08, 2.4145e-07], + [ 5.8208e-09, 3.5507e-09, 0.0000e+00, ..., 3.4343e-09, + 1.1991e-08, 1.1234e-08], + [ 2.3283e-10, 8.7311e-10, -2.3283e-10, ..., 3.2480e-08, + 2.3283e-09, 4.5926e-08]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0408, 0.0146, 0.0106, 0.0271, 0.0306, -0.0191, 0.0289, 0.0330, + -0.0121, -0.0016], device='cuda:0'), grad: tensor([ 5.5879e-09, -5.8033e-08, -5.4482e-07, 4.6974e-08, 2.7881e-08, + 2.1188e-08, 1.0128e-08, 4.1071e-07, 3.1781e-08, 7.2003e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 250.40, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4715 re_mapping 0.0026 re_causal 0.0091 /// teacc 99.16 lr 0.00010000 +Epoch 338, weight, value: tensor([[ 0.0903, -0.1690, -0.1404, ..., -0.1363, -0.1704, -0.1667], + [ 0.0410, 0.0531, -0.0338, ..., -0.0746, 0.1674, 0.0193], + [-0.0469, -0.1719, -0.0150, ..., 0.2151, -0.1830, 0.0389], + ..., + [-0.2132, 0.1496, 0.0265, ..., -0.1533, 0.1224, 0.0616], + [ 0.0039, -0.0843, -0.0050, ..., -0.1369, -0.1786, -0.2646], + [ 0.0326, -0.0469, -0.0249, ..., -0.2389, -0.1835, 0.0209]], + device='cuda:0'), grad: tensor([[-4.7614e-08, 5.8208e-10, 0.0000e+00, ..., -6.8685e-09, + 1.2806e-09, 8.7311e-10], + [-6.3737e-08, -5.2678e-08, 0.0000e+00, ..., -1.4668e-08, + -1.2899e-07, -8.0850e-08], + [ 1.1234e-08, 1.0477e-08, 0.0000e+00, ..., -1.1059e-09, + 1.9092e-08, 1.0768e-08], + ..., + [ 1.0186e-08, 2.5029e-09, 0.0000e+00, ..., 7.5670e-10, + 1.2224e-08, 6.0536e-09], + [ 5.3726e-08, 2.0140e-08, 0.0000e+00, ..., 4.9477e-09, + 4.9185e-08, 3.0559e-08], + [ 2.8289e-08, 1.5716e-09, 0.0000e+00, ..., 3.0850e-09, + 3.0850e-09, 2.2701e-09]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0411, 0.0144, 0.0101, 0.0276, 0.0304, -0.0192, 0.0288, 0.0328, + -0.0101, -0.0012], device='cuda:0'), grad: tensor([-3.4948e-07, -3.0454e-07, 5.7975e-08, 5.8790e-08, 5.4832e-08, + -1.8068e-07, 2.0559e-07, 4.4936e-08, 2.5751e-07, 1.8044e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 250.42, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4554 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.05 lr 0.00010000 +Epoch 339, weight, value: tensor([[ 0.0903, -0.1701, -0.1404, ..., -0.1362, -0.1706, -0.1667], + [ 0.0411, 0.0531, -0.0340, ..., -0.0749, 0.1675, 0.0193], + [-0.0470, -0.1722, -0.0150, ..., 0.2152, -0.1831, 0.0389], + ..., + [-0.2133, 0.1496, 0.0266, ..., -0.1534, 0.1224, 0.0616], + [ 0.0038, -0.0844, -0.0050, ..., -0.1372, -0.1789, -0.2649], + [ 0.0326, -0.0469, -0.0248, ..., -0.2392, -0.1835, 0.0209]], + device='cuda:0'), grad: tensor([[ 6.1700e-09, 7.0431e-09, 0.0000e+00, ..., 9.3132e-10, + 1.8044e-08, 1.1409e-08], + [-1.5739e-07, -9.5693e-08, 0.0000e+00, ..., 1.7462e-09, + -2.9267e-07, -1.5763e-07], + [ 3.0326e-08, 3.4866e-08, 0.0000e+00, ..., -3.1432e-09, + 7.4913e-08, 4.6624e-08], + ..., + [ 5.6229e-08, -2.2515e-07, 0.0000e+00, ..., -5.1223e-09, + -1.9360e-07, -2.0687e-07], + [ 3.3702e-08, 3.5157e-08, 0.0000e+00, ..., 1.4552e-09, + 7.8289e-08, 4.9244e-08], + [ 3.3178e-09, 1.6729e-07, 0.0000e+00, ..., 3.8999e-09, + 2.0664e-07, 1.7160e-07]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0411, 0.0145, 0.0101, 0.0275, 0.0304, -0.0191, 0.0288, 0.0328, + -0.0102, -0.0012], device='cuda:0'), grad: tensor([ 3.4866e-08, -6.3563e-07, 1.7975e-07, 1.0728e-07, 1.6647e-07, + 1.6065e-08, 2.3108e-08, -6.9803e-07, 1.8789e-07, 6.2957e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 250.66, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4674 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 +Epoch 340, weight, value: tensor([[ 0.0903, -0.1706, -0.1404, ..., -0.1370, -0.1705, -0.1667], + [ 0.0411, 0.0531, -0.0340, ..., -0.0731, 0.1678, 0.0196], + [-0.0472, -0.1723, -0.0150, ..., 0.2173, -0.1856, 0.0412], + ..., + [-0.2133, 0.1496, 0.0266, ..., -0.1543, 0.1224, 0.0616], + [ 0.0040, -0.0843, -0.0050, ..., -0.1389, -0.1790, -0.2653], + [ 0.0326, -0.0472, -0.0247, ..., -0.2395, -0.1837, 0.0209]], + device='cuda:0'), grad: tensor([[-5.5297e-09, 4.0745e-10, 0.0000e+00, ..., 1.7462e-10, + 5.8208e-10, 6.4028e-10], + [-4.0221e-08, -3.2131e-08, 0.0000e+00, ..., 9.4296e-09, + -1.7951e-07, -1.1775e-07], + [ 0.0000e+00, 1.7229e-08, 0.0000e+00, ..., -1.0536e-08, + 6.4319e-08, 4.4238e-08], + ..., + [ 2.1770e-08, -1.7928e-08, 0.0000e+00, ..., 5.6461e-09, + 3.6962e-08, 2.7241e-08], + [ 3.3760e-09, 1.2456e-08, 0.0000e+00, ..., 1.0885e-08, + 2.6601e-08, 2.4505e-08], + [ 5.2387e-09, 4.2492e-09, 0.0000e+00, ..., 4.5577e-08, + 8.4401e-09, 5.1747e-08]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0411, 0.0148, 0.0123, 0.0251, 0.0304, -0.0196, 0.0290, 0.0328, + -0.0099, -0.0012], device='cuda:0'), grad: tensor([-3.6613e-08, -3.1688e-07, 1.1391e-07, 4.0513e-08, -1.4296e-07, + -1.2282e-08, 6.5251e-08, 8.3761e-08, 2.3982e-08, 1.8324e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 250.47, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4382 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.18 lr 0.00010000 +Epoch 341, weight, value: tensor([[ 0.0904, -0.1714, -0.1405, ..., -0.1371, -0.1705, -0.1667], + [ 0.0411, 0.0531, -0.0340, ..., -0.0728, 0.1678, 0.0197], + [-0.0472, -0.1732, -0.0148, ..., 0.2173, -0.1861, 0.0411], + ..., + [-0.2134, 0.1497, 0.0266, ..., -0.1546, 0.1225, 0.0616], + [ 0.0040, -0.0844, -0.0051, ..., -0.1395, -0.1792, -0.2658], + [ 0.0326, -0.0473, -0.0243, ..., -0.2397, -0.1838, 0.0209]], + device='cuda:0'), grad: tensor([[-2.7940e-09, 6.2282e-09, 0.0000e+00, ..., 8.1491e-10, + 9.4878e-09, 6.4611e-09], + [-3.7788e-07, -1.7253e-07, 0.0000e+00, ..., 3.4925e-10, + -4.3795e-07, -1.1566e-07], + [ 3.9581e-08, 6.0257e-07, 5.8208e-11, ..., 2.3283e-10, + 5.8301e-07, 6.0862e-07], + ..., + [ 2.5204e-08, -9.4669e-07, 0.0000e+00, ..., 5.8208e-11, + -7.6927e-07, -8.8895e-07], + [ 2.7986e-07, 2.5099e-07, 0.0000e+00, ..., 1.4552e-09, + 3.4575e-07, 1.5355e-07], + [ 6.3446e-09, 1.0629e-07, 0.0000e+00, ..., 2.1537e-09, + 9.5228e-08, 9.3307e-08]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0411, 0.0148, 0.0122, 0.0252, 0.0304, -0.0197, 0.0287, 0.0328, + -0.0100, -0.0012], device='cuda:0'), grad: tensor([-1.5774e-08, -1.0245e-06, 1.8841e-06, 2.2934e-07, 2.2561e-07, + 2.8755e-08, 3.9407e-08, -2.5835e-06, 9.0525e-07, 3.2596e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 250.65, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4639 re_mapping 0.0024 re_causal 0.0086 /// teacc 99.06 lr 0.00010000 +Epoch 342, weight, value: tensor([[ 0.0904, -0.1726, -0.1405, ..., -0.1371, -0.1712, -0.1667], + [ 0.0412, 0.0535, -0.0340, ..., -0.0729, 0.1681, 0.0198], + [-0.0482, -0.1738, -0.0147, ..., 0.2173, -0.1864, 0.0411], + ..., + [-0.2136, 0.1494, 0.0267, ..., -0.1548, 0.1222, 0.0614], + [ 0.0037, -0.0846, -0.0052, ..., -0.1407, -0.1798, -0.2669], + [ 0.0326, -0.0474, -0.0242, ..., -0.2399, -0.1839, 0.0209]], + device='cuda:0'), grad: tensor([[ 6.9849e-10, 3.4925e-10, 0.0000e+00, ..., 6.1700e-08, + 1.1642e-10, 5.3435e-08], + [-1.0594e-08, -6.6357e-09, 0.0000e+00, ..., 1.1818e-06, + -2.1653e-08, 1.0449e-06], + [ 2.4447e-09, 8.1491e-10, 0.0000e+00, ..., 1.0692e-06, + 4.7730e-09, 9.4669e-07], + ..., + [ 1.0943e-08, 7.4506e-09, 0.0000e+00, ..., 6.4820e-07, + 2.3283e-09, 6.1560e-07], + [ 1.5949e-08, 3.4925e-09, 0.0000e+00, ..., 4.4238e-08, + 7.6834e-09, 4.6799e-08], + [-1.8626e-08, -1.4435e-08, 0.0000e+00, ..., 3.2857e-06, + 1.0477e-09, 2.8275e-06]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0411, 0.0151, 0.0121, 0.0252, 0.0304, -0.0198, 0.0289, 0.0324, + -0.0106, -0.0012], device='cuda:0'), grad: tensor([ 3.0664e-07, 3.3248e-06, 3.1620e-05, -3.0667e-05, -1.7986e-05, + 1.2887e-07, 5.0012e-07, 3.4571e-06, 3.1386e-07, 9.0152e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 250.36, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4158 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.04 lr 0.00010000 +Epoch 343, weight, value: tensor([[ 0.0904, -0.1728, -0.1405, ..., -0.1372, -0.1713, -0.1667], + [ 0.0413, 0.0535, -0.0338, ..., -0.0729, 0.1682, 0.0198], + [-0.0484, -0.1741, -0.0145, ..., 0.2173, -0.1867, 0.0411], + ..., + [-0.2136, 0.1494, 0.0264, ..., -0.1552, 0.1221, 0.0614], + [ 0.0035, -0.0842, -0.0052, ..., -0.1414, -0.1797, -0.2672], + [ 0.0327, -0.0476, -0.0241, ..., -0.2403, -0.1840, 0.0210]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 3.4925e-10, 0.0000e+00, ..., 2.6776e-09, + 1.0477e-09, 3.1432e-09], + [-5.2270e-08, -3.1176e-07, 0.0000e+00, ..., 1.0699e-07, + -4.4517e-07, -1.1153e-07], + [ 9.3132e-10, 1.5134e-09, 0.0000e+00, ..., -1.2864e-07, + -2.1653e-08, -1.1479e-07], + ..., + [ 3.4575e-08, 2.2096e-07, 0.0000e+00, ..., 3.5041e-08, + 3.3155e-07, 1.9697e-07], + [ 4.1910e-09, 1.0361e-08, 0.0000e+00, ..., 8.0327e-09, + 1.8510e-08, 1.5367e-08], + [ 8.8476e-09, 6.6939e-08, 0.0000e+00, ..., 1.2736e-07, + 9.6625e-08, 8.2701e-07]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0411, 0.0151, 0.0121, 0.0252, 0.0303, -0.0196, 0.0289, 0.0323, + -0.0106, -0.0012], device='cuda:0'), grad: tensor([ 1.0710e-08, -7.2410e-07, -3.0966e-07, 4.0745e-09, -2.9411e-06, + 8.7311e-09, 4.1910e-08, 8.7824e-07, 6.3563e-08, 2.9579e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 250.52, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4611 re_mapping 0.0025 re_causal 0.0086 /// teacc 99.06 lr 0.00010000 +Epoch 344, weight, value: tensor([[ 0.0904, -0.1731, -0.1405, ..., -0.1374, -0.1709, -0.1667], + [ 0.0413, 0.0535, -0.0337, ..., -0.0729, 0.1682, 0.0198], + [-0.0486, -0.1744, -0.0144, ..., 0.2173, -0.1867, 0.0411], + ..., + [-0.2137, 0.1494, 0.0263, ..., -0.1554, 0.1221, 0.0614], + [ 0.0033, -0.0841, -0.0052, ..., -0.1418, -0.1797, -0.2674], + [ 0.0327, -0.0482, -0.0241, ..., -0.2417, -0.1842, 0.0208]], + device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 8.9640e-09, + 8.1491e-10, 7.3342e-09], + [ 3.8417e-09, 2.5611e-09, 0.0000e+00, ..., 1.4913e-07, + 4.7381e-08, 2.1502e-07], + [ 1.0477e-09, 1.0477e-09, 0.0000e+00, ..., -1.2352e-07, + -8.9640e-09, -8.3703e-08], + ..., + [ 5.8208e-09, 1.3039e-08, 0.0000e+00, ..., 5.1223e-08, + 9.9302e-08, 7.0897e-08], + [ 7.5554e-08, 9.3132e-10, 0.0000e+00, ..., 2.2701e-08, + 5.0059e-09, 2.0373e-08], + [-1.6298e-09, 1.2806e-09, 0.0000e+00, ..., 2.7474e-08, + 1.0361e-08, 1.4319e-08]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0411, 0.0151, 0.0121, 0.0253, 0.0305, -0.0197, 0.0291, 0.0323, + -0.0108, -0.0012], device='cuda:0'), grad: tensor([ 4.0280e-08, 6.4820e-07, -2.1956e-07, -3.8669e-06, -6.6124e-07, + -7.5763e-07, 2.8173e-07, 3.8482e-06, 5.8953e-07, 1.2247e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 249.94, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.5014 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.13 lr 0.00010000 +Epoch 345, weight, value: tensor([[ 0.0904, -0.1733, -0.1406, ..., -0.1393, -0.1684, -0.1667], + [ 0.0412, 0.0535, -0.0338, ..., -0.0731, 0.1682, 0.0198], + [-0.0496, -0.1746, -0.0145, ..., 0.2173, -0.1868, 0.0411], + ..., + [-0.2138, 0.1494, 0.0262, ..., -0.1556, 0.1222, 0.0615], + [ 0.0038, -0.0841, -0.0052, ..., -0.1404, -0.1797, -0.2674], + [ 0.0327, -0.0486, -0.0240, ..., -0.2419, -0.1845, 0.0208]], + device='cuda:0'), grad: tensor([[ 2.5611e-09, 1.0477e-09, 0.0000e+00, ..., 5.7044e-09, + 2.0955e-09, 6.5193e-09], + [-2.2352e-08, -7.3574e-08, 0.0000e+00, ..., 1.1991e-08, + -2.9616e-07, -1.6694e-07], + [ 8.7311e-09, 4.7730e-09, 0.0000e+00, ..., -5.5414e-08, + 1.9209e-08, -3.0966e-08], + ..., + [ 5.9372e-08, 1.3271e-08, 0.0000e+00, ..., 2.1886e-08, + 1.1607e-07, 8.0559e-08], + [-1.0652e-07, 2.0955e-08, 0.0000e+00, ..., 1.2107e-08, + 4.6333e-08, 4.8778e-08], + [ 1.2340e-08, 1.2573e-08, 0.0000e+00, ..., 9.7789e-09, + 3.7136e-08, 2.6426e-08]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0411, 0.0151, 0.0120, 0.0253, 0.0305, -0.0200, 0.0307, 0.0323, + -0.0095, -0.0013], device='cuda:0'), grad: tensor([ 3.2713e-08, -1.3749e-07, -1.2224e-07, 5.1223e-08, 8.4983e-08, + 6.6124e-08, 1.9209e-08, 3.6042e-07, -4.4145e-07, 1.1269e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 250.54, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4911 re_mapping 0.0026 re_causal 0.0089 /// teacc 99.08 lr 0.00010000 +Epoch 346, weight, value: tensor([[ 0.0903, -0.1741, -0.1406, ..., -0.1393, -0.1686, -0.1667], + [ 0.0412, 0.0535, -0.0337, ..., -0.0731, 0.1682, 0.0198], + [-0.0493, -0.1753, -0.0146, ..., 0.2174, -0.1870, 0.0411], + ..., + [-0.2140, 0.1494, 0.0262, ..., -0.1562, 0.1222, 0.0615], + [ 0.0047, -0.0838, -0.0052, ..., -0.1415, -0.1796, -0.2676], + [ 0.0327, -0.0489, -0.0235, ..., -0.2433, -0.1848, 0.0207]], + device='cuda:0'), grad: tensor([[ 1.9441e-08, 1.1642e-10, 0.0000e+00, ..., 1.2689e-08, + 4.6566e-10, 5.0059e-09], + [ 7.4506e-09, -2.5728e-08, 0.0000e+00, ..., 2.7474e-08, + -7.1945e-08, -2.2585e-08], + [ 1.8277e-08, 1.6298e-09, 0.0000e+00, ..., -1.7346e-08, + -3.0268e-09, -3.3760e-09], + ..., + [ 2.1653e-08, 1.3853e-08, 1.1642e-10, ..., 8.7311e-09, + 3.9930e-08, 2.8056e-08], + [-2.0023e-08, 4.8894e-09, 0.0000e+00, ..., -4.7730e-09, + 1.4435e-08, 1.2224e-08], + [-2.8801e-07, 1.0477e-09, -5.8208e-10, ..., 3.9348e-08, + 2.7940e-09, -8.9640e-09]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0411, 0.0150, 0.0120, 0.0253, 0.0306, -0.0198, 0.0304, 0.0322, + -0.0077, -0.0013], device='cuda:0'), grad: tensor([ 1.4761e-07, 7.8231e-08, 7.4739e-08, 1.0664e-06, 2.1746e-07, + 9.8138e-08, -5.1456e-08, 1.5285e-07, -2.0838e-08, -1.7509e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 250.86, cls_loss 0.0007 cls_loss_mapping 0.0010 cls_loss_causal 0.4966 re_mapping 0.0026 re_causal 0.0089 /// teacc 99.17 lr 0.00010000 +Epoch 347, weight, value: tensor([[ 0.0903, -0.1768, -0.1406, ..., -0.1394, -0.1694, -0.1667], + [ 0.0412, 0.0535, -0.0337, ..., -0.0732, 0.1682, 0.0198], + [-0.0496, -0.1787, -0.0146, ..., 0.2174, -0.1896, 0.0406], + ..., + [-0.2140, 0.1499, 0.0262, ..., -0.1552, 0.1228, 0.0622], + [ 0.0050, -0.0839, -0.0052, ..., -0.1418, -0.1799, -0.2681], + [ 0.0327, -0.0504, -0.0235, ..., -0.2429, -0.1854, 0.0208]], + device='cuda:0'), grad: tensor([[-8.7311e-09, 1.5134e-09, 0.0000e+00, ..., 5.8208e-10, + 1.3970e-09, 2.2119e-09], + [-1.0943e-08, 8.1607e-08, 0.0000e+00, ..., 1.5134e-09, + 5.1572e-08, 6.7055e-08], + [ 9.3132e-10, 4.1211e-08, 0.0000e+00, ..., -5.8208e-09, + 3.8533e-08, 3.1432e-08], + ..., + [ 5.2387e-09, -2.0431e-07, 0.0000e+00, ..., 1.2806e-09, + -1.8103e-07, -1.7672e-07], + [ 1.8626e-09, 6.9849e-09, 0.0000e+00, ..., 2.3283e-10, + 1.3039e-08, 9.1968e-09], + [ 1.1642e-10, 2.7940e-08, 0.0000e+00, ..., 8.6147e-09, + 2.4913e-08, 1.5716e-08]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0411, 0.0150, 0.0115, 0.0252, 0.0304, -0.0198, 0.0304, 0.0331, + -0.0072, -0.0013], device='cuda:0'), grad: tensor([-3.7486e-08, 1.5658e-07, 8.1258e-08, 2.0140e-08, 1.2841e-07, + 2.2119e-09, 3.4692e-08, -4.2794e-07, 2.1886e-08, 3.9465e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 251.34, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4419 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +Epoch 348, weight, value: tensor([[ 0.0903, -0.1774, -0.1407, ..., -0.1396, -0.1695, -0.1667], + [ 0.0412, 0.0534, -0.0337, ..., -0.0732, 0.1682, 0.0197], + [-0.0497, -0.1805, -0.0146, ..., 0.2173, -0.1904, 0.0404], + ..., + [-0.2141, 0.1503, 0.0262, ..., -0.1539, 0.1230, 0.0627], + [ 0.0049, -0.0840, -0.0052, ..., -0.1434, -0.1800, -0.2690], + [ 0.0327, -0.0513, -0.0235, ..., -0.2431, -0.1859, 0.0208]], + device='cuda:0'), grad: tensor([[ 1.2806e-09, 1.1292e-08, 0.0000e+00, ..., 2.3283e-10, + 1.0943e-08, 1.3039e-08], + [-2.3283e-09, 7.0781e-08, 0.0000e+00, ..., 7.7998e-09, + 6.4494e-08, 8.8592e-08], + [ 2.3283e-10, 6.2864e-08, 0.0000e+00, ..., 2.9104e-09, + 6.0652e-08, 6.6473e-08], + ..., + [ 3.3760e-09, -2.3295e-07, 0.0000e+00, ..., 5.4715e-09, + -2.2200e-07, -2.4866e-07], + [ 1.2224e-08, 6.7521e-09, 0.0000e+00, ..., 2.3283e-09, + 8.3819e-09, 9.5461e-09], + [ 9.3132e-10, 7.1013e-09, 0.0000e+00, ..., 6.2864e-08, + 7.1013e-09, 9.9419e-08]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0411, 0.0149, 0.0113, 0.0253, 0.0303, -0.0199, 0.0306, 0.0335, + -0.0074, -0.0013], device='cuda:0'), grad: tensor([ 4.3306e-08, 2.5565e-07, 2.0256e-07, 1.4668e-08, -1.7066e-07, + -1.1420e-07, 4.8662e-08, -6.9942e-07, 7.5554e-08, 3.5483e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 250.64, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4665 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.21 lr 0.00010000 +Epoch 349, weight, value: tensor([[ 0.0903, -0.1781, -0.1407, ..., -0.1396, -0.1696, -0.1667], + [ 0.0413, 0.0535, -0.0337, ..., -0.0733, 0.1683, 0.0198], + [-0.0499, -0.1816, -0.0146, ..., 0.2172, -0.1909, 0.0402], + ..., + [-0.2143, 0.1504, 0.0262, ..., -0.1532, 0.1230, 0.0628], + [ 0.0051, -0.0841, -0.0052, ..., -0.1440, -0.1802, -0.2697], + [ 0.0327, -0.0515, -0.0235, ..., -0.2428, -0.1860, 0.0209]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 3.4925e-10], + [-4.4005e-08, -2.8173e-08, 0.0000e+00, ..., 3.3760e-09, + -8.7428e-08, -2.5379e-08], + [ 6.9849e-10, 5.3551e-09, 0.0000e+00, ..., 1.3737e-08, + 4.3074e-09, 2.6776e-09], + ..., + [ 1.1292e-08, -1.8626e-09, 0.0000e+00, ..., 2.6776e-09, + 1.5600e-08, 1.7462e-09], + [ 2.6543e-08, 1.7229e-08, 0.0000e+00, ..., 1.1642e-09, + 4.8662e-08, 1.6182e-08], + [ 3.0268e-09, 2.9104e-09, 0.0000e+00, ..., 2.3283e-09, + 5.9372e-09, 3.2596e-09]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0411, 0.0150, 0.0111, 0.0253, 0.0303, -0.0197, 0.0304, 0.0335, + -0.0072, -0.0012], device='cuda:0'), grad: tensor([ 1.2224e-08, -1.4564e-07, 2.7660e-07, -5.4715e-07, 1.5949e-08, + 7.2760e-08, 1.4203e-08, 1.4051e-07, 1.3097e-07, 4.7730e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 250.53, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4548 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.14 lr 0.00010000 +Epoch 350, weight, value: tensor([[ 0.0903, -0.1790, -0.1422, ..., -0.1402, -0.1700, -0.1667], + [ 0.0414, 0.0535, -0.0332, ..., -0.0734, 0.1684, 0.0198], + [-0.0496, -0.1817, -0.0133, ..., 0.2173, -0.1910, 0.0403], + ..., + [-0.2144, 0.1505, 0.0261, ..., -0.1536, 0.1230, 0.0627], + [ 0.0050, -0.0843, -0.0054, ..., -0.1455, -0.1806, -0.2705], + [ 0.0327, -0.0531, -0.0235, ..., -0.2430, -0.1874, 0.0208]], + device='cuda:0'), grad: tensor([[ 2.6776e-09, 2.4447e-09, 0.0000e+00, ..., 5.8208e-10, + 2.6776e-09, 3.4925e-09], + [ 6.7521e-09, 6.8103e-08, 0.0000e+00, ..., 1.5134e-09, + 7.0431e-08, 8.3819e-08], + [ 8.1491e-10, 1.7090e-07, 0.0000e+00, ..., 5.7044e-09, + 1.6531e-07, 1.8976e-07], + ..., + [ 4.5402e-09, -2.9360e-07, 0.0000e+00, ..., -5.3551e-09, + -2.9965e-07, -3.2340e-07], + [-5.2154e-08, 1.0245e-08, 0.0000e+00, ..., 1.3970e-09, + 1.2806e-08, 1.2922e-08], + [-1.4901e-08, 1.5367e-08, 0.0000e+00, ..., 2.2119e-09, + 1.7113e-08, -1.6415e-08]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0411, 0.0151, 0.0112, 0.0254, 0.0304, -0.0202, 0.0304, 0.0334, + -0.0074, -0.0013], device='cuda:0'), grad: tensor([ 2.7241e-08, 3.1060e-07, 5.5647e-07, 1.5053e-07, 1.3201e-07, + 1.7788e-07, 2.4447e-09, -9.4203e-07, -3.1362e-07, -8.3004e-08], + device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 349---------------------------------------------------- +epoch 349, time 251.26, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4559 re_mapping 0.0025 re_causal 0.0082 /// teacc 99.24 lr 0.00010000 +Epoch 351, weight, value: tensor([[ 0.0903, -0.1803, -0.1429, ..., -0.1416, -0.1704, -0.1667], + [ 0.0415, 0.0535, -0.0333, ..., -0.0734, 0.1684, 0.0198], + [-0.0496, -0.1818, -0.0127, ..., 0.2173, -0.1911, 0.0403], + ..., + [-0.2145, 0.1519, 0.0262, ..., -0.1538, 0.1243, 0.0646], + [ 0.0049, -0.0846, -0.0055, ..., -0.1459, -0.1812, -0.2716], + [ 0.0327, -0.0536, -0.0235, ..., -0.2431, -0.1877, 0.0208]], + device='cuda:0'), grad: tensor([[-2.9922e-05, -2.3283e-10, 0.0000e+00, ..., -4.3392e-05, + 0.0000e+00, 2.3283e-10], + [ 1.2340e-08, -1.4435e-08, 0.0000e+00, ..., 3.4808e-08, + -2.8522e-08, -1.5134e-08], + [ 8.2655e-09, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, + 1.2806e-09, -2.7940e-09], + ..., + [ 1.8510e-08, 1.1059e-08, 0.0000e+00, ..., 5.7044e-09, + 2.1188e-08, 1.5832e-08], + [ 2.6776e-08, 1.1642e-09, 0.0000e+00, ..., 2.9220e-08, + 2.2119e-09, 1.7462e-09], + [ 1.5832e-08, 9.3132e-10, 0.0000e+00, ..., 6.9849e-09, + 1.3970e-09, -3.2596e-09]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0411, 0.0150, 0.0112, 0.0255, 0.0289, -0.0209, 0.0315, 0.0352, + -0.0078, -0.0013], device='cuda:0'), grad: tensor([-1.9050e-04, 1.0454e-07, 3.9930e-08, 5.8208e-08, 2.9569e-08, + 1.8487e-07, 1.8954e-04, 9.8604e-08, 1.6042e-07, 7.9977e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 251.11, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4508 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.04 lr 0.00010000 +Epoch 352, weight, value: tensor([[ 0.0904, -0.1799, -0.1431, ..., -0.1394, -0.1705, -0.1668], + [ 0.0415, 0.0535, -0.0332, ..., -0.0735, 0.1685, 0.0198], + [-0.0498, -0.1819, -0.0127, ..., 0.2174, -0.1911, 0.0403], + ..., + [-0.2146, 0.1519, 0.0262, ..., -0.1540, 0.1243, 0.0645], + [ 0.0051, -0.0847, -0.0055, ..., -0.1469, -0.1814, -0.2723], + [ 0.0327, -0.0537, -0.0234, ..., -0.2432, -0.1878, 0.0207]], + device='cuda:0'), grad: tensor([[ 8.1491e-10, 2.3283e-09, 0.0000e+00, ..., 7.4506e-09, + 2.7940e-09, 3.6089e-09], + [-2.6985e-07, -7.1805e-07, 0.0000e+00, ..., 5.0175e-08, + -9.7789e-07, -3.6275e-07], + [ 2.4447e-09, 5.6112e-08, 0.0000e+00, ..., -5.2736e-08, + 6.1817e-08, 3.2713e-08], + ..., + [ 2.5495e-07, 5.1595e-07, 0.0000e+00, ..., 2.8522e-08, + 7.4226e-07, 2.9127e-07], + [-4.1211e-08, 3.0501e-08, 0.0000e+00, ..., 3.3062e-08, + 4.1444e-08, 2.9104e-08], + [ 3.9232e-08, 7.0781e-08, 0.0000e+00, ..., 6.2399e-07, + 8.0676e-08, 1.5963e-06]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0411, 0.0150, 0.0112, 0.0255, 0.0290, -0.0209, 0.0299, 0.0352, + -0.0077, -0.0013], device='cuda:0'), grad: tensor([ 3.1083e-08, -2.0154e-06, 4.0978e-08, 2.6659e-08, -4.3623e-06, + -3.9581e-09, -8.1258e-08, 1.8785e-06, 1.9791e-09, 4.4778e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 250.83, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4379 re_mapping 0.0025 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +Epoch 353, weight, value: tensor([[ 0.0904, -0.1806, -0.1434, ..., -0.1395, -0.1709, -0.1668], + [ 0.0415, 0.0535, -0.0339, ..., -0.0736, 0.1685, 0.0197], + [-0.0496, -0.1820, -0.0128, ..., 0.2175, -0.1911, 0.0403], + ..., + [-0.2147, 0.1521, 0.0283, ..., -0.1543, 0.1244, 0.0646], + [ 0.0052, -0.0847, -0.0056, ..., -0.1474, -0.1815, -0.2729], + [ 0.0327, -0.0557, -0.0256, ..., -0.2439, -0.1890, 0.0206]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1758e-08, 0.0000e+00, ..., 3.8417e-09, + 1.1409e-08, 9.4296e-09], + [-2.0955e-09, 1.5562e-06, 0.0000e+00, ..., 4.3074e-09, + 1.4855e-06, 1.2238e-06], + [ 1.1642e-10, 6.5775e-08, 0.0000e+00, ..., -2.0955e-09, + 6.3097e-08, 4.8080e-08], + ..., + [ 5.8208e-10, -8.5831e-06, 0.0000e+00, ..., 4.6566e-10, + -8.2180e-06, -6.7391e-06], + [ 2.2119e-09, 1.8789e-07, 0.0000e+00, ..., 2.4447e-09, + 1.8138e-07, 1.4750e-07], + [ 2.3283e-10, 6.4522e-06, 0.0000e+00, ..., 7.2177e-09, + 6.1803e-06, 5.0850e-06]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0411, 0.0149, 0.0112, 0.0254, 0.0291, -0.0205, 0.0293, 0.0352, + -0.0067, -0.0014], device='cuda:0'), grad: tensor([ 4.6799e-08, 4.3511e-06, 1.7358e-07, 3.9581e-07, 8.4843e-07, + 4.8196e-08, -5.5879e-07, -2.3976e-05, 5.4156e-07, 1.8105e-05], + device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 250.67, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4449 re_mapping 0.0027 re_causal 0.0089 /// teacc 99.15 lr 0.00010000 +Epoch 354, weight, value: tensor([[ 0.0904, -0.1806, -0.1434, ..., -0.1396, -0.1710, -0.1668], + [ 0.0416, 0.0535, -0.0340, ..., -0.0736, 0.1685, 0.0197], + [-0.0500, -0.1823, -0.0128, ..., 0.2175, -0.1911, 0.0403], + ..., + [-0.2148, 0.1521, 0.0283, ..., -0.1545, 0.1245, 0.0647], + [ 0.0054, -0.0849, -0.0056, ..., -0.1479, -0.1815, -0.2734], + [ 0.0327, -0.0570, -0.0255, ..., -0.2440, -0.1899, 0.0205]], + device='cuda:0'), grad: tensor([[ 8.1491e-10, 2.5611e-09, 0.0000e+00, ..., 1.2806e-09, + 2.7940e-09, 2.3283e-09], + [-7.0897e-08, -2.2119e-08, 0.0000e+00, ..., 2.5611e-09, + -1.0349e-07, -3.7136e-08], + [ 3.3760e-09, 1.1525e-08, 0.0000e+00, ..., -5.5879e-09, + 1.3155e-08, 3.9581e-09], + ..., + [ 2.2934e-08, -8.0792e-08, 0.0000e+00, ..., 3.1432e-09, + -2.5379e-08, -3.3993e-08], + [ 3.1316e-08, 1.8161e-08, 0.0000e+00, ..., 4.6566e-10, + 3.6554e-08, 2.5262e-08], + [-1.9441e-08, 3.2131e-08, 0.0000e+00, ..., 2.7940e-09, + 2.7241e-08, -6.8685e-09]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0411, 0.0149, 0.0112, 0.0255, 0.0291, -0.0203, 0.0290, 0.0352, + -0.0065, -0.0014], device='cuda:0'), grad: tensor([ 2.3632e-08, 2.5053e-07, 6.3446e-08, 2.0675e-07, 2.3819e-07, + 5.4599e-08, 3.8184e-08, -1.0768e-07, -7.7067e-07, 2.3167e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 250.45, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4576 re_mapping 0.0025 re_causal 0.0084 /// teacc 98.99 lr 0.00010000 +Epoch 355, weight, value: tensor([[ 0.0904, -0.1812, -0.1434, ..., -0.1397, -0.1713, -0.1668], + [ 0.0416, 0.0535, -0.0344, ..., -0.0737, 0.1685, 0.0197], + [-0.0501, -0.1825, -0.0125, ..., 0.2175, -0.1913, 0.0403], + ..., + [-0.2149, 0.1522, 0.0284, ..., -0.1547, 0.1245, 0.0647], + [ 0.0053, -0.0850, -0.0055, ..., -0.1486, -0.1817, -0.2742], + [ 0.0327, -0.0573, -0.0256, ..., -0.2443, -0.1900, 0.0205]], + device='cuda:0'), grad: tensor([[ 1.2922e-08, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + -2.3283e-10, 1.3970e-09], + [-1.5716e-08, 5.6694e-08, -4.7730e-09, ..., -4.4471e-08, + -1.1176e-08, 2.1071e-08], + [ 8.3819e-09, 2.0955e-09, 1.1642e-09, ..., 1.0827e-08, + 1.7928e-08, 7.3342e-09], + ..., + [ 8.0327e-09, -6.1817e-08, 2.3283e-10, ..., 8.1491e-10, + -6.0070e-08, -3.6904e-08], + [ 2.3982e-08, 8.1491e-10, 2.4447e-09, ..., 2.2701e-08, + 3.4808e-08, 1.5018e-08], + [-1.8976e-08, 1.1642e-09, -1.5134e-09, ..., 2.5611e-09, + 2.4447e-09, -3.8883e-08]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0411, 0.0149, 0.0112, 0.0255, 0.0292, -0.0203, 0.0292, 0.0352, + -0.0068, -0.0014], device='cuda:0'), grad: tensor([ 8.3703e-08, 1.1642e-08, 8.6962e-08, 4.3539e-07, 4.2049e-07, + -1.0738e-06, 4.4494e-07, 9.3132e-10, 3.0780e-07, -7.0827e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 250.76, cls_loss 0.0008 cls_loss_mapping 0.0015 cls_loss_causal 0.4671 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.08 lr 0.00010000 +Epoch 356, weight, value: tensor([[ 0.0904, -0.1820, -0.1435, ..., -0.1398, -0.1715, -0.1668], + [ 0.0416, 0.0528, -0.0346, ..., -0.0739, 0.1681, 0.0193], + [-0.0502, -0.1828, -0.0124, ..., 0.2176, -0.1915, 0.0403], + ..., + [-0.2150, 0.1530, 0.0285, ..., -0.1555, 0.1251, 0.0652], + [ 0.0053, -0.0851, -0.0048, ..., -0.1496, -0.1820, -0.2764], + [ 0.0327, -0.0574, -0.0256, ..., -0.2450, -0.1905, 0.0205]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, + 0.0000e+00, 1.1642e-10], + [-2.8871e-08, -3.0734e-08, 0.0000e+00, ..., 8.1491e-10, + -4.2259e-08, -4.5053e-08], + [ 6.9849e-10, 3.4925e-10, 0.0000e+00, ..., 1.1642e-10, + 5.8208e-10, 5.8208e-10], + ..., + [ 2.6193e-08, 2.5844e-08, 0.0000e+00, ..., 5.8208e-10, + 3.6205e-08, 3.9465e-08], + [ 1.7462e-09, 2.9104e-09, 0.0000e+00, ..., 2.3283e-10, + 4.3074e-09, 4.4238e-09], + [-1.7462e-09, 1.1642e-09, 0.0000e+00, ..., 6.9849e-10, + 1.6298e-09, -7.3342e-09]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0411, 0.0144, 0.0112, 0.0254, 0.0295, -0.0199, 0.0288, 0.0358, + -0.0071, -0.0014], device='cuda:0'), grad: tensor([ 1.8626e-09, -1.1874e-07, 6.9849e-09, -2.0256e-08, 2.8173e-08, + 1.6997e-08, 2.3283e-09, 1.1746e-07, 8.2655e-09, -1.5600e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 251.62, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4964 re_mapping 0.0026 re_causal 0.0087 /// teacc 99.06 lr 0.00010000 +Epoch 357, weight, value: tensor([[ 0.0904, -0.1824, -0.1436, ..., -0.1397, -0.1709, -0.1668], + [ 0.0417, 0.0528, -0.0348, ..., -0.0740, 0.1681, 0.0192], + [-0.0503, -0.1829, -0.0127, ..., 0.2176, -0.1917, 0.0403], + ..., + [-0.2151, 0.1531, 0.0284, ..., -0.1558, 0.1252, 0.0654], + [ 0.0045, -0.0852, -0.0034, ..., -0.1510, -0.1822, -0.2776], + [ 0.0327, -0.0589, -0.0255, ..., -0.2460, -0.1914, 0.0203]], + device='cuda:0'), grad: tensor([[ 1.1642e-09, 1.1642e-10, 1.1642e-10, ..., 1.2806e-09, + 9.3132e-10, 1.7462e-09], + [-3.8976e-07, -1.8626e-09, 4.6566e-10, ..., -3.7346e-07, + -1.1660e-06, -2.7847e-07], + [ 3.2596e-08, 1.3271e-08, 1.1642e-10, ..., 2.0722e-08, + 8.6962e-08, 2.6892e-08], + ..., + [ 2.2235e-08, -2.0256e-08, 8.1491e-10, ..., 5.2154e-08, + 2.9453e-08, 5.5996e-08], + [ 3.1502e-07, 2.0955e-09, 6.9849e-10, ..., 3.1199e-07, + 9.1782e-07, 2.4331e-07], + [-1.0710e-08, 3.3760e-09, -6.4028e-09, ..., 1.1525e-08, + 5.8208e-09, -2.5379e-08]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0411, 0.0143, 0.0112, 0.0256, 0.0297, -0.0197, 0.0289, 0.0358, + -0.0088, -0.0015], device='cuda:0'), grad: tensor([ 2.7125e-08, -2.1141e-06, 2.7497e-07, 4.7544e-07, -1.1502e-07, + -1.0533e-06, 1.4156e-07, 3.8091e-07, 2.0191e-06, -3.1781e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 251.57, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4433 re_mapping 0.0027 re_causal 0.0087 /// teacc 99.17 lr 0.00010000 +Epoch 358, weight, value: tensor([[ 0.0905, -0.1796, -0.1437, ..., -0.1397, -0.1709, -0.1668], + [ 0.0417, 0.0528, -0.0348, ..., -0.0741, 0.1681, 0.0191], + [-0.0508, -0.1833, -0.0127, ..., 0.2176, -0.1919, 0.0403], + ..., + [-0.2152, 0.1531, 0.0282, ..., -0.1560, 0.1252, 0.0653], + [ 0.0044, -0.0855, -0.0041, ..., -0.1526, -0.1827, -0.2791], + [ 0.0327, -0.0598, -0.0252, ..., -0.2478, -0.1917, 0.0199]], + device='cuda:0'), grad: tensor([[-3.7369e-08, 3.4925e-10, 0.0000e+00, ..., 1.2340e-08, + -1.3970e-09, 2.3283e-10], + [-1.3388e-08, -2.0140e-08, 0.0000e+00, ..., 5.0059e-09, + -3.7136e-08, -4.0745e-09], + [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.3074e-09, + 9.3132e-10, -5.8208e-10], + ..., + [ 6.4028e-09, 8.3819e-09, 0.0000e+00, ..., 1.3970e-09, + 1.5134e-08, 7.4506e-09], + [ 5.9372e-09, 6.7521e-09, 0.0000e+00, ..., 2.7940e-09, + 1.3504e-08, 5.3551e-09], + [ 2.8289e-08, 6.9849e-10, 0.0000e+00, ..., 4.6566e-09, + 2.7940e-09, 1.1758e-08]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0410, 0.0142, 0.0111, 0.0257, 0.0305, -0.0197, 0.0288, 0.0357, + -0.0092, -0.0017], device='cuda:0'), grad: tensor([-2.4401e-07, -4.3656e-08, 1.7928e-08, 1.1525e-08, 7.9279e-08, + 1.2154e-07, -2.7521e-07, 4.6217e-08, 4.4587e-08, 2.5611e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 250.89, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4419 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.21 lr 0.00010000 +Epoch 359, weight, value: tensor([[ 0.0905, -0.1796, -0.1437, ..., -0.1398, -0.1708, -0.1668], + [ 0.0418, 0.0527, -0.0348, ..., -0.0743, 0.1681, 0.0191], + [-0.0510, -0.1833, -0.0127, ..., 0.2177, -0.1919, 0.0403], + ..., + [-0.2153, 0.1533, 0.0281, ..., -0.1561, 0.1253, 0.0654], + [ 0.0044, -0.0857, -0.0038, ..., -0.1537, -0.1830, -0.2799], + [ 0.0327, -0.0601, -0.0252, ..., -0.2480, -0.1920, 0.0199]], + device='cuda:0'), grad: tensor([[-3.5274e-08, 1.7462e-09, 0.0000e+00, ..., -4.7730e-09, + 3.0268e-09, 2.0955e-09], + [-2.4075e-07, -4.3423e-07, 0.0000e+00, ..., 1.0477e-09, + -7.6648e-07, -2.9034e-07], + [ 7.7998e-09, 1.0710e-08, -9.3132e-10, ..., -4.6566e-09, + 2.3399e-08, 4.0745e-09], + ..., + [ 1.6345e-07, 2.8778e-07, 1.1642e-10, ..., 1.6298e-09, + 5.0804e-07, 2.0035e-07], + [ 4.2492e-08, 7.9162e-08, 8.1491e-10, ..., 1.3504e-08, + 1.3772e-07, 5.8673e-08], + [-5.3551e-09, 1.7695e-08, 0.0000e+00, ..., 5.8208e-10, + 3.0268e-08, -3.7253e-08]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0410, 0.0142, 0.0111, 0.0257, 0.0306, -0.0199, 0.0289, 0.0357, + -0.0093, -0.0018], device='cuda:0'), grad: tensor([-1.8207e-07, -1.7611e-06, 4.6683e-08, 1.4203e-08, 2.9523e-07, + 6.2864e-09, 1.5169e-07, 1.2312e-06, 3.5577e-07, -1.4168e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 251.45, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4871 re_mapping 0.0024 re_causal 0.0087 /// teacc 99.14 lr 0.00010000 +Epoch 360, weight, value: tensor([[ 0.0905, -0.1799, -0.1437, ..., -0.1397, -0.1713, -0.1668], + [ 0.0420, 0.0529, -0.0348, ..., -0.0742, 0.1685, 0.0192], + [-0.0515, -0.1834, -0.0127, ..., 0.2177, -0.1920, 0.0403], + ..., + [-0.2156, 0.1532, 0.0278, ..., -0.1569, 0.1251, 0.0652], + [ 0.0041, -0.0865, -0.0036, ..., -0.1549, -0.1845, -0.2822], + [ 0.0327, -0.0601, -0.0249, ..., -0.2483, -0.1922, 0.0199]], + device='cuda:0'), grad: tensor([[ 4.0745e-09, 8.1491e-10, 0.0000e+00, ..., 1.1642e-09, + 3.4925e-10, 4.3074e-09], + [ 1.3039e-08, 2.0559e-07, 0.0000e+00, ..., 1.1059e-08, + 6.9966e-08, 1.2608e-07], + [ 2.6776e-09, 2.6426e-08, 0.0000e+00, ..., -3.2596e-09, + 1.0245e-08, 1.3271e-08], + ..., + [ 1.4319e-08, -4.2212e-07, 0.0000e+00, ..., 1.6415e-08, + -1.4459e-07, -1.7870e-07], + [ 9.6625e-09, 5.7975e-08, 0.0000e+00, ..., 1.5134e-09, + 2.0140e-08, 4.0629e-08], + [-2.4494e-07, 2.1188e-08, 0.0000e+00, ..., 9.2201e-08, + 7.7998e-09, -2.5844e-08]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0410, 0.0145, 0.0111, 0.0257, 0.0306, -0.0197, 0.0288, 0.0354, + -0.0103, -0.0017], device='cuda:0'), grad: tensor([ 2.6659e-08, 5.1269e-07, 5.7626e-08, 1.2643e-07, -2.0652e-07, + 2.9197e-07, 4.6100e-08, -5.7882e-07, 1.3644e-07, -3.9861e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 251.30, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4583 re_mapping 0.0025 re_causal 0.0085 /// teacc 99.16 lr 0.00010000 +Epoch 361, weight, value: tensor([[ 0.0905, -0.1800, -0.1437, ..., -0.1398, -0.1711, -0.1668], + [ 0.0420, 0.0528, -0.0348, ..., -0.0744, 0.1685, 0.0191], + [-0.0519, -0.1835, -0.0127, ..., 0.2178, -0.1921, 0.0403], + ..., + [-0.2158, 0.1532, 0.0280, ..., -0.1574, 0.1251, 0.0651], + [ 0.0043, -0.0863, -0.0035, ..., -0.1555, -0.1845, -0.2828], + [ 0.0327, -0.0598, -0.0249, ..., -0.2485, -0.1922, 0.0200]], + device='cuda:0'), grad: tensor([[-2.0955e-09, 1.1642e-10, 0.0000e+00, ..., 4.6566e-10, + 4.6566e-10, 3.4925e-10], + [-1.5949e-08, -1.7812e-08, 0.0000e+00, ..., 5.4715e-09, + -3.9465e-08, 1.8626e-09], + [ 9.3132e-10, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, + 9.3132e-10, 1.0477e-09], + ..., + [ 4.1910e-09, 2.7940e-09, 0.0000e+00, ..., 5.9372e-09, + 1.0012e-08, 1.7928e-08], + [ 1.1991e-08, 1.4086e-08, 0.0000e+00, ..., 1.3970e-09, + 2.9919e-08, 1.0012e-08], + [ 2.3283e-10, 1.5134e-09, 0.0000e+00, ..., 1.1642e-09, + 2.4447e-09, -3.8417e-09]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0410, 0.0144, 0.0111, 0.0256, 0.0307, -0.0194, 0.0290, 0.0353, + -0.0101, -0.0017], device='cuda:0'), grad: tensor([-1.1525e-08, -2.4680e-08, 1.0477e-08, 1.1642e-09, -8.5100e-08, + 4.5402e-09, -6.7521e-09, 7.5321e-08, 7.1130e-08, -1.0361e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 251.55, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4631 re_mapping 0.0025 re_causal 0.0088 /// teacc 99.14 lr 0.00010000 +Epoch 362, weight, value: tensor([[ 0.0905, -0.1799, -0.1437, ..., -0.1398, -0.1704, -0.1668], + [ 0.0422, 0.0530, -0.0349, ..., -0.0742, 0.1687, 0.0193], + [-0.0521, -0.1836, -0.0127, ..., 0.2177, -0.1929, 0.0402], + ..., + [-0.2159, 0.1532, 0.0279, ..., -0.1576, 0.1251, 0.0650], + [ 0.0037, -0.0876, -0.0034, ..., -0.1559, -0.1857, -0.2842], + [ 0.0328, -0.0592, -0.0248, ..., -0.2495, -0.1920, 0.0200]], + device='cuda:0'), grad: tensor([[-1.0477e-09, 5.8208e-11, 0.0000e+00, ..., 1.1642e-10, + -6.4028e-10, 3.4925e-10], + [-7.4506e-09, -1.2340e-08, 0.0000e+00, ..., 6.8103e-09, + -2.2526e-08, 9.8953e-10], + [ 1.3388e-09, 1.2224e-09, 0.0000e+00, ..., -1.2224e-09, + 2.0373e-09, 6.9849e-10], + ..., + [ 1.0245e-08, 6.6357e-09, 0.0000e+00, ..., 1.5716e-09, + 1.2107e-08, 1.8626e-08], + [ 4.0163e-09, 3.7835e-09, 0.0000e+00, ..., 1.6880e-09, + 7.2760e-09, 7.1595e-09], + [-2.0780e-08, 7.5670e-10, 0.0000e+00, ..., 6.9849e-10, + 1.3388e-09, -4.1211e-08]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0410, 0.0145, 0.0110, 0.0256, 0.0308, -0.0192, 0.0289, 0.0351, + -0.0108, -0.0017], device='cuda:0'), grad: tensor([-5.2387e-09, -4.4820e-09, 6.4611e-09, 5.4715e-09, 1.1135e-07, + 7.1013e-09, -1.2806e-09, 9.1502e-08, 3.3702e-08, -2.2154e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 252.01, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4448 re_mapping 0.0023 re_causal 0.0081 /// teacc 99.09 lr 0.00010000 +Epoch 363, weight, value: tensor([[ 0.0905, -0.1801, -0.1438, ..., -0.1399, -0.1706, -0.1668], + [ 0.0422, 0.0530, -0.0350, ..., -0.0744, 0.1688, 0.0192], + [-0.0516, -0.1838, -0.0127, ..., 0.2178, -0.1930, 0.0402], + ..., + [-0.2160, 0.1532, 0.0278, ..., -0.1578, 0.1251, 0.0650], + [ 0.0038, -0.0876, -0.0034, ..., -0.1558, -0.1858, -0.2844], + [ 0.0328, -0.0591, -0.0244, ..., -0.2497, -0.1921, 0.0200]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 4.4820e-09, + 1.1642e-10, 3.3760e-09], + [-2.2701e-09, -2.0955e-09, 0.0000e+00, ..., 7.8580e-09, + -4.8312e-09, 6.9267e-09], + [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., -4.3190e-08, + 4.6566e-10, -3.9174e-08], + ..., + [ 1.4552e-09, 5.2387e-09, 0.0000e+00, ..., 1.5774e-08, + 5.9954e-09, 3.3004e-08], + [ 2.9104e-10, 9.3132e-10, 0.0000e+00, ..., 1.0477e-08, + 1.8626e-09, 4.7730e-09], + [-5.2387e-10, -6.4611e-09, 0.0000e+00, ..., 2.6193e-09, + -5.4133e-09, -2.0838e-08]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0410, 0.0145, 0.0110, 0.0256, 0.0308, -0.0193, 0.0288, 0.0351, + -0.0106, -0.0017], device='cuda:0'), grad: tensor([ 2.0547e-08, 2.9453e-08, -1.8999e-07, 6.4319e-08, 1.8510e-08, + 1.1589e-07, -1.5181e-07, 1.7206e-07, 4.7905e-08, -1.1519e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 251.84, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4832 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.10 lr 0.00010000 +Epoch 364, weight, value: tensor([[ 0.0905, -0.1804, -0.1439, ..., -0.1399, -0.1709, -0.1668], + [ 0.0423, 0.0530, -0.0350, ..., -0.0746, 0.1688, 0.0192], + [-0.0519, -0.1839, -0.0126, ..., 0.2178, -0.1931, 0.0402], + ..., + [-0.2162, 0.1533, 0.0277, ..., -0.1580, 0.1251, 0.0651], + [ 0.0040, -0.0876, -0.0037, ..., -0.1566, -0.1860, -0.2848], + [ 0.0328, -0.0594, -0.0235, ..., -0.2502, -0.1920, 0.0201]], + device='cuda:0'), grad: tensor([[-2.7649e-08, 1.6880e-09, 0.0000e+00, ..., 3.4925e-10, + 1.8044e-09, 5.3551e-09], + [ 7.1013e-09, 9.4937e-08, 0.0000e+00, ..., 9.8953e-10, + 9.0047e-08, 1.4494e-07], + [ 4.3074e-09, 3.6368e-07, 0.0000e+00, ..., 5.8208e-11, + 3.8696e-07, 5.9698e-07], + ..., + [ 2.9220e-08, -5.1875e-07, 0.0000e+00, ..., 1.2806e-09, + -5.4203e-07, -7.6089e-07], + [ 2.4505e-08, 1.4086e-08, 0.0000e+00, ..., 1.7462e-10, + 1.6531e-08, 2.8347e-08], + [-8.3062e-08, 8.9640e-09, 0.0000e+00, ..., 1.0186e-08, + 8.7894e-09, -7.2236e-08]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0410, 0.0145, 0.0110, 0.0256, 0.0308, -0.0192, 0.0292, 0.0351, + -0.0104, -0.0017], device='cuda:0'), grad: tensor([-1.0792e-07, 3.9465e-07, 1.5199e-06, 6.7579e-08, -8.9698e-08, + -1.5483e-08, 2.9453e-08, -1.8524e-06, 1.8207e-07, -1.2480e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 251.53, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4656 re_mapping 0.0024 re_causal 0.0085 /// teacc 99.02 lr 0.00010000 +Epoch 365, weight, value: tensor([[ 0.0905, -0.1806, -0.1441, ..., -0.1400, -0.1711, -0.1668], + [ 0.0423, 0.0530, -0.0352, ..., -0.0749, 0.1689, 0.0192], + [-0.0524, -0.1842, -0.0127, ..., 0.2178, -0.1933, 0.0401], + ..., + [-0.2164, 0.1533, 0.0269, ..., -0.1583, 0.1251, 0.0650], + [ 0.0043, -0.0878, -0.0040, ..., -0.1567, -0.1861, -0.2850], + [ 0.0328, -0.0595, -0.0221, ..., -0.2514, -0.1922, 0.0200]], + device='cuda:0'), grad: tensor([[ 1.6298e-09, 3.8068e-08, 0.0000e+00, ..., 5.8208e-10, + 4.0862e-08, 4.5286e-08], + [-2.8405e-08, 1.8077e-06, 0.0000e+00, ..., 1.9092e-08, + 1.8692e-06, 2.1867e-06], + [-2.0140e-08, 6.1095e-06, 0.0000e+00, ..., 2.4098e-08, + 6.4522e-06, 7.2978e-06], + ..., + [ 7.1013e-09, -9.6411e-06, 0.0000e+00, ..., -8.8126e-08, + -1.0096e-05, -1.1571e-05], + [ 3.2014e-08, 1.4144e-07, 0.0000e+00, ..., 3.7951e-08, + 1.5425e-07, 1.9849e-07], + [ 3.6089e-09, 4.0396e-07, 0.0000e+00, ..., 1.8626e-09, + 4.1211e-07, 4.8289e-07]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0410, 0.0144, 0.0109, 0.0256, 0.0311, -0.0192, 0.0292, 0.0349, + -0.0102, -0.0017], device='cuda:0'), grad: tensor([ 1.3993e-07, 6.5155e-06, 2.2009e-05, 7.6648e-07, 3.2131e-06, + 6.6007e-08, 3.0617e-08, -3.4809e-05, 6.3702e-07, 1.4594e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 251.45, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4325 re_mapping 0.0024 re_causal 0.0078 /// teacc 99.05 lr 0.00010000 +Epoch 366, weight, value: tensor([[ 0.0906, -0.1809, -0.1441, ..., -0.1400, -0.1714, -0.1668], + [ 0.0425, 0.0531, -0.0353, ..., -0.0750, 0.1690, 0.0192], + [-0.0512, -0.1870, -0.0123, ..., 0.2178, -0.1947, 0.0391], + ..., + [-0.2167, 0.1542, 0.0269, ..., -0.1563, 0.1255, 0.0664], + [ 0.0040, -0.0879, -0.0041, ..., -0.1600, -0.1865, -0.2870], + [ 0.0328, -0.0632, -0.0220, ..., -0.2543, -0.1944, 0.0193]], + device='cuda:0'), grad: tensor([[-2.0489e-08, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, + 3.4925e-10, 2.4447e-09], + [ 1.7462e-09, -2.0955e-09, 0.0000e+00, ..., 1.4319e-08, + -1.3970e-08, 4.6217e-08], + [ 3.2596e-09, 8.3819e-09, 0.0000e+00, ..., -1.3970e-09, + 8.1491e-09, 1.4086e-08], + ..., + [ 1.6764e-08, -2.9104e-09, 0.0000e+00, ..., 7.9162e-09, + -1.1642e-09, 2.6310e-08], + [ 1.0128e-08, 3.8417e-09, 0.0000e+00, ..., 2.5611e-09, + 4.7730e-09, 8.1491e-09], + [-1.6065e-08, -7.6834e-09, -2.3283e-10, ..., 1.2538e-07, + 2.5611e-09, 2.3481e-07]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0410, 0.0145, 0.0101, 0.0256, 0.0317, -0.0194, 0.0289, 0.0363, + -0.0111, -0.0020], device='cuda:0'), grad: tensor([-1.4191e-07, 1.3970e-07, 6.4727e-08, 7.5670e-08, -1.0263e-06, + -8.9523e-08, 1.6706e-07, 1.0827e-07, 7.4971e-08, 6.4541e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 251.08, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4772 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.08 lr 0.00010000 +Epoch 367, weight, value: tensor([[ 0.0906, -0.1811, -0.1441, ..., -0.1401, -0.1718, -0.1668], + [ 0.0427, 0.0532, -0.0354, ..., -0.0752, 0.1692, 0.0192], + [-0.0512, -0.1871, -0.0115, ..., 0.2180, -0.1947, 0.0392], + ..., + [-0.2170, 0.1543, 0.0269, ..., -0.1570, 0.1255, 0.0664], + [ 0.0037, -0.0885, -0.0042, ..., -0.1606, -0.1875, -0.2888], + [ 0.0328, -0.0639, -0.0220, ..., -0.2558, -0.1950, 0.0189]], + device='cuda:0'), grad: tensor([[-6.7847e-07, -2.9663e-07, 0.0000e+00, ..., -8.6729e-08, + -5.5600e-07, 2.3283e-10], + [ 5.9698e-07, 2.1968e-07, 0.0000e+00, ..., 8.5798e-08, + 4.3097e-07, -3.5274e-08], + [ 5.3551e-09, 3.4925e-09, 0.0000e+00, ..., 9.3132e-10, + 5.9372e-09, 1.3970e-09], + ..., + [ 2.2934e-08, 3.8184e-08, 0.0000e+00, ..., 3.4925e-10, + 5.6229e-08, 2.5379e-08], + [ 1.9907e-08, 1.9325e-08, 0.0000e+00, ..., 3.7253e-09, + 3.4110e-08, 1.4552e-08], + [ 2.6776e-09, 1.9791e-09, 0.0000e+00, ..., 7.9162e-09, + 3.3760e-09, 3.9232e-08]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0410, 0.0145, 0.0102, 0.0256, 0.0320, -0.0194, 0.0285, 0.0363, + -0.0117, -0.0022], device='cuda:0'), grad: tensor([-3.0268e-06, 2.6952e-06, 3.5390e-08, -1.6415e-08, -1.5309e-07, + 1.1874e-07, -4.5286e-08, 1.1828e-07, 1.1735e-07, 1.5681e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 251.53, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4274 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.14 lr 0.00010000 +Epoch 368, weight, value: tensor([[ 0.0906, -0.1836, -0.1442, ..., -0.1428, -0.1741, -0.1669], + [ 0.0427, 0.0532, -0.0355, ..., -0.0784, 0.1684, 0.0181], + [-0.0513, -0.1872, -0.0114, ..., 0.2199, -0.1921, 0.0406], + ..., + [-0.2172, 0.1543, 0.0269, ..., -0.1574, 0.1255, 0.0664], + [ 0.0032, -0.0887, -0.0041, ..., -0.1611, -0.1878, -0.2909], + [ 0.0329, -0.0627, -0.0220, ..., -0.2565, -0.1952, 0.0190]], + device='cuda:0'), grad: tensor([[-2.4447e-09, 5.8208e-10, 0.0000e+00, ..., 4.6566e-10, + -3.4925e-10, 5.8208e-10], + [-1.0114e-06, -4.2021e-06, 0.0000e+00, ..., 2.3283e-09, + -5.0180e-06, -2.1756e-06], + [ 1.6298e-09, 2.3399e-08, 0.0000e+00, ..., -8.1491e-09, + 1.9907e-08, 1.2806e-09], + ..., + [ 9.6206e-07, 3.9935e-06, 0.0000e+00, ..., 1.7462e-09, + 4.7684e-06, 2.0694e-06], + [ 2.3865e-08, 9.4762e-08, 0.0000e+00, ..., 1.5134e-09, + 1.1316e-07, 5.1572e-08], + [ 4.0745e-09, 7.9162e-09, 0.0000e+00, ..., 1.1642e-10, + 9.4296e-09, 2.9104e-09]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0412, 0.0128, 0.0119, 0.0255, 0.0323, -0.0194, 0.0312, 0.0362, + -0.0124, -0.0021], device='cuda:0'), grad: tensor([-7.9162e-09, -1.0513e-05, 2.3190e-07, -8.4168e-08, 2.6310e-08, + 1.2573e-08, 1.5134e-09, 1.0036e-05, 2.4750e-07, 2.5146e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 251.70, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4696 re_mapping 0.0026 re_causal 0.0086 /// teacc 99.06 lr 0.00010000 +Epoch 369, weight, value: tensor([[ 0.0907, -0.1833, -0.1442, ..., -0.1423, -0.1735, -0.1669], + [ 0.0429, 0.0533, -0.0356, ..., -0.0784, 0.1687, 0.0182], + [-0.0526, -0.1875, -0.0115, ..., 0.2199, -0.1923, 0.0406], + ..., + [-0.2174, 0.1544, 0.0269, ..., -0.1579, 0.1255, 0.0664], + [ 0.0032, -0.0897, -0.0034, ..., -0.1620, -0.1898, -0.2931], + [ 0.0329, -0.0627, -0.0220, ..., -0.2569, -0.1952, 0.0190]], + device='cuda:0'), grad: tensor([[-3.2596e-09, -1.7462e-09, 0.0000e+00, ..., 1.8743e-08, + 3.6089e-09, 5.8557e-08], + [-4.3190e-08, -6.1234e-08, 0.0000e+00, ..., 1.9697e-07, + -4.1910e-08, 5.5786e-07], + [ 1.6298e-09, 2.4447e-09, 0.0000e+00, ..., -2.7660e-06, + -2.4550e-06, -3.1181e-06], + ..., + [ 1.9209e-08, 2.3865e-08, 0.0000e+00, ..., 1.7453e-06, + 1.1194e-06, 3.6601e-06], + [ 4.3074e-09, 3.3760e-09, 0.0000e+00, ..., 3.7369e-08, + 2.3632e-08, 1.3132e-07], + [-2.9104e-09, 2.3283e-09, 0.0000e+00, ..., 7.5065e-06, + 2.0489e-08, 3.1918e-05]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0411, 0.0129, 0.0118, 0.0255, 0.0323, -0.0195, 0.0308, 0.0362, + -0.0128, -0.0021], device='cuda:0'), grad: tensor([ 1.7742e-07, 1.8366e-06, -1.3217e-05, 6.7018e-06, -1.1432e-04, + 6.0420e-08, 1.6869e-07, 1.3448e-05, 4.4610e-07, 1.0467e-04], + device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 251.77, cls_loss 0.0008 cls_loss_mapping 0.0017 cls_loss_causal 0.4679 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.16 lr 0.00010000 +Epoch 370, weight, value: tensor([[ 0.0879, -0.1871, -0.1443, ..., -0.1422, -0.1751, -0.1695], + [ 0.0427, 0.0533, -0.0358, ..., -0.0784, 0.1687, 0.0182], + [-0.0533, -0.1875, -0.0115, ..., 0.2201, -0.1923, 0.0407], + ..., + [-0.2176, 0.1542, 0.0239, ..., -0.1587, 0.1254, 0.0652], + [ 0.0043, -0.0892, -0.0034, ..., -0.1623, -0.1901, -0.2937], + [ 0.0357, -0.0582, -0.0184, ..., -0.2576, -0.1920, 0.0220]], + device='cuda:0'), grad: tensor([[-3.9581e-09, -3.0268e-09, 0.0000e+00, ..., 0.0000e+00, + -2.3283e-09, 0.0000e+00], + [-9.7789e-09, 5.0059e-09, 0.0000e+00, ..., -6.9849e-09, + -2.6077e-08, -1.0477e-08], + [ 3.1432e-09, 1.7462e-09, 0.0000e+00, ..., 4.6566e-10, + 2.7940e-09, 2.5611e-09], + ..., + [ 2.9104e-09, -1.7928e-08, 0.0000e+00, ..., 1.6298e-09, + -7.1013e-09, -1.0245e-08], + [-4.1910e-09, 1.1642e-10, 0.0000e+00, ..., 3.4925e-10, + 2.3283e-09, -1.5134e-09], + [ 9.5461e-09, 6.6357e-09, 0.0000e+00, ..., 8.1491e-10, + 4.6566e-09, 6.4028e-09]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0438, 0.0128, 0.0119, 0.0254, 0.0322, -0.0191, 0.0297, 0.0351, + -0.0118, 0.0007], device='cuda:0'), grad: tensor([-3.3295e-08, -7.1013e-09, 3.4808e-08, 4.7032e-08, 4.4471e-08, + -1.2910e-07, 5.0641e-08, -2.3399e-08, -1.1211e-07, 1.3434e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 251.58, cls_loss 0.0007 cls_loss_mapping 0.0016 cls_loss_causal 0.4755 re_mapping 0.0023 re_causal 0.0080 /// teacc 98.99 lr 0.00010000 +Epoch 371, weight, value: tensor([[ 0.0879, -0.1872, -0.1443, ..., -0.1423, -0.1752, -0.1696], + [ 0.0426, 0.0533, -0.0359, ..., -0.0784, 0.1688, 0.0181], + [-0.0537, -0.1877, -0.0115, ..., 0.2201, -0.1925, 0.0407], + ..., + [-0.2178, 0.1544, 0.0237, ..., -0.1591, 0.1256, 0.0652], + [ 0.0047, -0.0891, -0.0033, ..., -0.1624, -0.1903, -0.2940], + [ 0.0358, -0.0581, -0.0183, ..., -0.2579, -0.1920, 0.0221]], + device='cuda:0'), grad: tensor([[-9.8953e-09, 1.1642e-09, 0.0000e+00, ..., 2.0256e-08, + -1.2806e-09, 1.0943e-08], + [-9.4646e-08, -1.4727e-07, 0.0000e+00, ..., 5.9372e-09, + -3.0431e-07, -8.6613e-08], + [ 3.4925e-09, 2.4447e-09, 0.0000e+00, ..., -4.4936e-08, + 4.3074e-09, -2.1188e-08], + ..., + [ 7.7416e-08, 8.9058e-08, 0.0000e+00, ..., 2.0955e-09, + 1.8359e-07, 6.0070e-08], + [-5.7160e-08, 1.0710e-08, 0.0000e+00, ..., 9.0804e-09, + 3.7486e-08, 2.3516e-08], + [ 3.6554e-08, 1.4086e-08, 0.0000e+00, ..., 2.4331e-08, + 2.0838e-08, 6.6473e-08]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0439, 0.0127, 0.0119, 0.0253, 0.0321, -0.0192, 0.0301, 0.0352, + -0.0114, 0.0008], device='cuda:0'), grad: tensor([ 2.9686e-08, -4.2352e-07, -1.8964e-07, 7.8464e-08, -2.2235e-07, + 3.3993e-08, 3.2131e-08, 4.2608e-07, -2.3248e-07, 4.7218e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 251.17, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4777 re_mapping 0.0025 re_causal 0.0089 /// teacc 99.17 lr 0.00010000 +Epoch 372, weight, value: tensor([[ 0.0879, -0.1872, -0.1446, ..., -0.1423, -0.1751, -0.1696], + [ 0.0431, 0.0539, -0.0330, ..., -0.0784, 0.1695, 0.0184], + [-0.0541, -0.1878, -0.0115, ..., 0.2202, -0.1926, 0.0407], + ..., + [-0.2197, 0.1538, 0.0233, ..., -0.1597, 0.1249, 0.0647], + [ 0.0048, -0.0894, -0.0036, ..., -0.1625, -0.1909, -0.2948], + [ 0.0358, -0.0581, -0.0183, ..., -0.2585, -0.1920, 0.0223]], + device='cuda:0'), grad: tensor([[-1.2689e-08, -0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + -1.2806e-09, 5.8208e-10], + [-9.3132e-10, 9.6625e-08, 2.3283e-09, ..., 2.5611e-09, + 8.2422e-08, 8.0443e-08], + [ 1.3970e-09, 8.3819e-09, 0.0000e+00, ..., 3.4925e-10, + 9.0804e-09, 2.0489e-08], + ..., + [ 1.3970e-09, -1.2945e-07, -2.6776e-09, ..., 8.6147e-09, + -1.1548e-07, -2.9686e-08], + [ 5.5879e-09, 4.7730e-09, 1.1642e-10, ..., 6.9849e-10, + 6.8685e-09, 7.7998e-09], + [ 5.0059e-09, 5.4715e-09, 0.0000e+00, ..., 6.8685e-09, + 5.3551e-09, 5.2154e-08]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0439, 0.0131, 0.0119, 0.0253, 0.0318, -0.0192, 0.0306, 0.0344, + -0.0113, 0.0009], device='cuda:0'), grad: tensor([-8.3237e-08, 2.5984e-07, 6.6683e-07, -6.9477e-07, -4.3586e-07, + 8.2306e-08, 2.2817e-08, -5.9255e-08, 5.6578e-08, 2.0664e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 251.52, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4266 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.15 lr 0.00010000 +Epoch 373, weight, value: tensor([[ 0.0879, -0.1872, -0.1446, ..., -0.1423, -0.1751, -0.1696], + [ 0.0432, 0.0539, -0.0330, ..., -0.0785, 0.1696, 0.0184], + [-0.0543, -0.1879, -0.0115, ..., 0.2202, -0.1927, 0.0407], + ..., + [-0.2198, 0.1538, 0.0233, ..., -0.1599, 0.1249, 0.0647], + [ 0.0047, -0.0896, -0.0035, ..., -0.1626, -0.1914, -0.2955], + [ 0.0358, -0.0581, -0.0183, ..., -0.2587, -0.1920, 0.0223]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 3.0268e-09, + 1.1642e-10, 1.5134e-09], + [-1.3737e-08, -1.8626e-08, 0.0000e+00, ..., 1.7707e-07, + -3.7369e-08, 1.0559e-07], + [ 6.9849e-10, 1.9791e-09, 0.0000e+00, ..., 3.7253e-09, + 2.0955e-09, 3.4925e-10], + ..., + [ 9.4296e-09, 9.7789e-09, 0.0000e+00, ..., 4.2259e-08, + 2.2235e-08, 3.2480e-08], + [ 7.7998e-09, 3.6089e-09, 0.0000e+00, ..., 1.9209e-08, + 7.3342e-09, 9.3132e-09], + [-1.2806e-08, 6.9849e-10, -3.4925e-10, ..., 4.9407e-07, + 1.5134e-09, 2.8196e-07]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0439, 0.0131, 0.0119, 0.0253, 0.0318, -0.0194, 0.0308, 0.0343, + -0.0115, 0.0009], device='cuda:0'), grad: tensor([ 1.2573e-08, 6.2259e-07, 2.4098e-08, 2.5611e-08, -3.1702e-06, + 3.4226e-08, 3.3621e-07, 2.0175e-07, 1.0245e-07, 1.8179e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 251.61, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4684 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.16 lr 0.00010000 +Epoch 374, weight, value: tensor([[ 0.0879, -0.1872, -0.1446, ..., -0.1423, -0.1751, -0.1696], + [ 0.0433, 0.0539, -0.0330, ..., -0.0786, 0.1696, 0.0184], + [-0.0544, -0.1879, -0.0115, ..., 0.2204, -0.1927, 0.0408], + ..., + [-0.2200, 0.1537, 0.0233, ..., -0.1606, 0.1249, 0.0644], + [ 0.0046, -0.0898, -0.0035, ..., -0.1626, -0.1916, -0.2960], + [ 0.0358, -0.0580, -0.0183, ..., -0.2597, -0.1921, 0.0223]], + device='cuda:0'), grad: tensor([[ 5.8208e-10, 1.1642e-09, 0.0000e+00, ..., 3.8417e-09, + 3.6089e-09, 1.6298e-09], + [-6.5193e-08, 5.6159e-07, 0.0000e+00, ..., 1.9209e-08, + 4.3632e-07, 5.0291e-07], + [ 3.0268e-09, 3.7486e-08, 0.0000e+00, ..., -3.9581e-09, + 5.6345e-08, 2.8522e-08], + ..., + [ 2.2468e-08, -6.7567e-07, 0.0000e+00, ..., -1.0245e-08, + -6.1514e-07, -5.9186e-07], + [ 2.1537e-08, 2.2235e-08, 0.0000e+00, ..., 1.7579e-08, + 4.7847e-08, 2.1188e-08], + [-1.2689e-08, 1.9674e-08, 0.0000e+00, ..., 1.1642e-09, + 2.6193e-08, -3.9465e-08]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0439, 0.0130, 0.0119, 0.0253, 0.0323, -0.0194, 0.0303, 0.0341, + -0.0117, 0.0009], device='cuda:0'), grad: tensor([ 1.8510e-08, 1.0841e-06, 1.0151e-07, 1.8044e-08, 2.7195e-07, + 3.3178e-08, -1.7451e-07, -1.3923e-06, 2.0198e-07, -1.5635e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 251.58, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4854 re_mapping 0.0022 re_causal 0.0083 /// teacc 99.16 lr 0.00010000 +Epoch 375, weight, value: tensor([[ 0.0879, -0.1872, -0.1446, ..., -0.1424, -0.1751, -0.1696], + [ 0.0434, 0.0536, -0.0330, ..., -0.0786, 0.1695, 0.0182], + [-0.0546, -0.1881, -0.0116, ..., 0.2204, -0.1928, 0.0408], + ..., + [-0.2201, 0.1550, 0.0233, ..., -0.1607, 0.1254, 0.0660], + [ 0.0048, -0.0900, -0.0032, ..., -0.1626, -0.1919, -0.2964], + [ 0.0358, -0.0584, -0.0183, ..., -0.2600, -0.1924, 0.0217]], + device='cuda:0'), grad: tensor([[-1.5250e-08, -2.4447e-09, 0.0000e+00, ..., 2.7940e-09, + -4.3074e-09, 3.4925e-10], + [ 1.3504e-08, -1.8626e-09, 0.0000e+00, ..., 1.6298e-09, + -1.0012e-08, 1.2806e-09], + [ 2.8289e-08, 2.5611e-09, 0.0000e+00, ..., -5.4715e-09, + 1.6298e-09, -2.5611e-09], + ..., + [ 2.9919e-08, -7.4506e-09, 0.0000e+00, ..., 2.6776e-09, + -6.9849e-10, -3.6089e-09], + [-2.4214e-06, 2.5611e-09, 0.0000e+00, ..., 9.3132e-10, + 5.0059e-09, 1.8626e-09], + [ 8.0676e-08, 2.9104e-09, 0.0000e+00, ..., 9.3132e-10, + 3.7253e-09, 1.3970e-09]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0439, 0.0128, 0.0119, 0.0251, 0.0321, -0.0176, 0.0301, 0.0353, + -0.0115, 0.0007], device='cuda:0'), grad: tensor([-1.6426e-07, 2.1618e-07, 2.9337e-07, 8.8010e-08, 8.9640e-08, + 2.3872e-05, 6.7987e-08, 2.8824e-07, -2.5585e-05, 8.6660e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 251.65, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4652 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.19 lr 0.00010000 +Epoch 376, weight, value: tensor([[ 0.0879, -0.1872, -0.1447, ..., -0.1424, -0.1751, -0.1696], + [ 0.0436, 0.0537, -0.0330, ..., -0.0786, 0.1696, 0.0183], + [-0.0545, -0.1881, -0.0112, ..., 0.2205, -0.1928, 0.0409], + ..., + [-0.2202, 0.1551, 0.0233, ..., -0.1611, 0.1254, 0.0662], + [ 0.0052, -0.0902, -0.0032, ..., -0.1634, -0.1922, -0.2971], + [ 0.0358, -0.0585, -0.0183, ..., -0.2602, -0.1925, 0.0216]], + device='cuda:0'), grad: tensor([[-2.8638e-08, 0.0000e+00, 1.1642e-10, ..., 6.9849e-10, + -6.9849e-10, 9.3132e-10], + [-1.6298e-09, -1.0477e-09, 3.4925e-10, ..., 1.5134e-09, + -6.0536e-09, 9.3132e-10], + [ 1.8626e-09, 6.9849e-10, 0.0000e+00, ..., -9.3132e-09, + 1.0477e-09, -1.0361e-08], + ..., + [ 3.3760e-09, 1.3970e-09, 5.8208e-10, ..., 5.2387e-09, + 3.4925e-09, 9.7789e-09], + [-1.9791e-09, 1.1642e-09, 1.1642e-10, ..., 1.9791e-09, + 3.4925e-09, 3.1432e-09], + [ 6.9849e-10, 3.4925e-10, -1.1292e-08, ..., 0.0000e+00, + 1.5134e-09, -3.7835e-08]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0438, 0.0129, 0.0120, 0.0251, 0.0320, -0.0182, 0.0302, 0.0353, + -0.0109, 0.0007], device='cuda:0'), grad: tensor([-1.6636e-07, 1.7579e-08, 8.1491e-10, -4.6520e-07, 1.1770e-07, + 4.0513e-07, 1.1583e-07, 5.1456e-08, 6.9849e-09, -6.3796e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 251.32, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4380 re_mapping 0.0024 re_causal 0.0084 /// teacc 99.14 lr 0.00010000 +Epoch 377, weight, value: tensor([[ 0.0880, -0.1872, -0.1449, ..., -0.1425, -0.1751, -0.1696], + [ 0.0436, 0.0537, -0.0330, ..., -0.0787, 0.1697, 0.0182], + [-0.0547, -0.1882, -0.0109, ..., 0.2206, -0.1929, 0.0409], + ..., + [-0.2203, 0.1552, 0.0233, ..., -0.1613, 0.1254, 0.0662], + [ 0.0052, -0.0903, -0.0033, ..., -0.1636, -0.1923, -0.2976], + [ 0.0358, -0.0585, -0.0182, ..., -0.2605, -0.1925, 0.0217]], + device='cuda:0'), grad: tensor([[ 9.1968e-09, 3.4925e-10, 5.1223e-09, ..., 1.2806e-09, + 4.6566e-10, 8.1491e-10], + [-6.0536e-09, -1.8277e-08, 1.1642e-09, ..., 4.5402e-09, + -2.8173e-08, -5.4715e-09], + [ 3.2596e-09, 1.3970e-09, 6.9849e-10, ..., 6.9849e-10, + 1.6298e-09, 1.0477e-09], + ..., + [ 7.7998e-09, 5.9372e-09, 1.1642e-10, ..., 3.1432e-09, + 8.2655e-09, 1.3621e-08], + [-2.9453e-08, 9.5461e-09, -9.8953e-09, ..., -8.9640e-09, + 1.4785e-08, 6.5193e-09], + [-1.8161e-08, -4.5402e-09, 1.6298e-09, ..., 5.5879e-09, + 8.1491e-10, -4.5169e-08]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0438, 0.0129, 0.0120, 0.0258, 0.0320, -0.0206, 0.0298, 0.0353, + -0.0110, 0.0007], device='cuda:0'), grad: tensor([ 5.2736e-08, 7.6834e-09, 2.0838e-08, 7.6834e-09, 1.8138e-07, + 3.0152e-08, 2.4447e-08, 7.1479e-08, -1.7951e-07, -2.0443e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 251.63, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4327 re_mapping 0.0023 re_causal 0.0082 /// teacc 99.24 lr 0.00010000 +Epoch 378, weight, value: tensor([[ 0.0880, -0.1872, -0.1454, ..., -0.1425, -0.1751, -0.1696], + [ 0.0441, 0.0546, -0.0331, ..., -0.0788, 0.1705, 0.0186], + [-0.0547, -0.1882, -0.0113, ..., 0.2207, -0.1929, 0.0409], + ..., + [-0.2215, 0.1544, 0.0233, ..., -0.1615, 0.1244, 0.0656], + [ 0.0053, -0.0903, -0.0048, ..., -0.1638, -0.1924, -0.2979], + [ 0.0358, -0.0585, -0.0182, ..., -0.2606, -0.1925, 0.0217]], + device='cuda:0'), grad: tensor([[-3.6089e-09, -6.9849e-10, 0.0000e+00, ..., 5.5879e-09, + -0.0000e+00, 2.4447e-09], + [-5.5297e-08, -3.8068e-08, 0.0000e+00, ..., 5.2387e-09, + -9.7323e-08, -1.8277e-08], + [ 4.6566e-10, 2.9686e-08, 0.0000e+00, ..., -1.9907e-08, + 2.1071e-08, 2.6193e-08], + ..., + [ 5.5879e-09, -4.8545e-08, 0.0000e+00, ..., -8.6147e-09, + -2.6659e-08, 3.2829e-08], + [ 4.0745e-09, 3.1432e-09, 0.0000e+00, ..., 3.2596e-09, + 6.4028e-09, 5.0059e-09], + [ 6.5193e-09, 4.5402e-09, 0.0000e+00, ..., 1.3970e-09, + 7.3342e-09, 5.7044e-09]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0438, 0.0133, 0.0120, 0.0258, 0.0320, -0.0207, 0.0290, 0.0346, + -0.0111, 0.0007], device='cuda:0'), grad: tensor([ 4.7730e-09, -1.0629e-07, 3.2736e-07, -7.3016e-06, 2.4959e-07, + 3.1851e-06, 2.6426e-08, 3.0082e-06, 2.1269e-07, 4.0256e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 251.77, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4632 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.10 lr 0.00010000 +Epoch 379, weight, value: tensor([[ 0.0880, -0.1872, -0.1455, ..., -0.1425, -0.1751, -0.1696], + [ 0.0452, 0.0550, -0.0331, ..., -0.0788, 0.1710, 0.0187], + [-0.0551, -0.1879, -0.0113, ..., 0.2209, -0.1926, 0.0412], + ..., + [-0.2218, 0.1544, 0.0233, ..., -0.1637, 0.1244, 0.0655], + [ 0.0052, -0.0904, -0.0048, ..., -0.1642, -0.1927, -0.2985], + [ 0.0357, -0.0589, -0.0182, ..., -0.2607, -0.1939, 0.0218]], + device='cuda:0'), grad: tensor([[-1.3970e-09, 1.1642e-10, 0.0000e+00, ..., 1.2340e-08, + 2.6776e-09, 9.0804e-09], + [-1.8394e-08, -2.2585e-08, 0.0000e+00, ..., 1.7683e-07, + -8.1491e-09, 1.2096e-07], + [-9.3132e-10, 1.5134e-09, 0.0000e+00, ..., -3.7393e-07, + -6.9384e-08, -2.7660e-07], + ..., + [ 1.2340e-08, 9.4296e-09, 0.0000e+00, ..., 3.4459e-08, + 2.6659e-08, 2.8987e-08], + [ 8.8476e-09, 4.5402e-09, 0.0000e+00, ..., 2.2468e-08, + 1.2340e-08, 1.5250e-08], + [-8.3819e-09, 1.1642e-09, 0.0000e+00, ..., 4.5402e-09, + 2.0955e-09, -8.1491e-09]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0438, 0.0137, 0.0122, 0.0259, 0.0316, -0.0213, 0.0293, 0.0345, + -0.0113, 0.0007], device='cuda:0'), grad: tensor([ 4.8894e-08, 7.3295e-07, -1.6801e-06, -9.1386e-08, 4.6310e-07, + 2.0117e-07, 3.9348e-08, 2.0349e-07, 1.2107e-07, -2.3516e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 252.08, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4658 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.20 lr 0.00010000 +Epoch 380, weight, value: tensor([[ 0.0880, -0.1872, -0.1458, ..., -0.1436, -0.1751, -0.1696], + [ 0.0454, 0.0551, -0.0331, ..., -0.0788, 0.1720, 0.0192], + [-0.0560, -0.1881, -0.0096, ..., 0.2211, -0.1938, 0.0410], + ..., + [-0.2223, 0.1546, 0.0233, ..., -0.1642, 0.1238, 0.0651], + [ 0.0047, -0.0912, -0.0050, ..., -0.1653, -0.1937, -0.3004], + [ 0.0358, -0.0589, -0.0182, ..., -0.2627, -0.1939, 0.0218]], + device='cuda:0'), grad: tensor([[ 1.2806e-09, 1.1642e-10, 0.0000e+00, ..., 6.9849e-10, + -1.6298e-09, 4.6566e-10], + [ 3.0268e-09, 8.2655e-09, 0.0000e+00, ..., 9.7789e-09, + 7.5670e-09, 2.1537e-08], + [ 9.3132e-10, 7.1013e-09, 0.0000e+00, ..., 3.4925e-10, + 7.1013e-09, 8.2655e-09], + ..., + [ 3.3760e-09, -1.6845e-07, 0.0000e+00, ..., 1.0594e-08, + -2.0291e-07, -2.3341e-07], + [ 4.5402e-08, 2.3283e-09, 0.0000e+00, ..., 2.3283e-10, + 3.6089e-09, 2.2119e-09], + [ 8.8476e-09, 2.0955e-09, 0.0000e+00, ..., 5.8557e-08, + 3.3760e-09, 7.0431e-08]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0438, 0.0143, 0.0120, 0.0260, 0.0314, -0.0212, 0.0299, 0.0340, + -0.0123, 0.0007], device='cuda:0'), grad: tensor([ 1.9791e-09, 9.1619e-08, 3.7951e-08, 7.5437e-07, 3.5274e-08, + -1.0971e-06, 6.1584e-08, -4.1071e-07, 2.0850e-07, 3.3621e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 251.66, cls_loss 0.0006 cls_loss_mapping 0.0007 cls_loss_causal 0.4685 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.08 lr 0.00010000 +Epoch 381, weight, value: tensor([[ 0.0880, -0.1872, -0.1458, ..., -0.1436, -0.1751, -0.1696], + [ 0.0454, 0.0550, -0.0331, ..., -0.0790, 0.1720, 0.0191], + [-0.0552, -0.1882, -0.0096, ..., 0.2212, -0.1938, 0.0411], + ..., + [-0.2225, 0.1547, 0.0233, ..., -0.1649, 0.1239, 0.0651], + [ 0.0044, -0.0913, -0.0050, ..., -0.1657, -0.1938, -0.3010], + [ 0.0358, -0.0589, -0.0182, ..., -0.2632, -0.1940, 0.0218]], + device='cuda:0'), grad: tensor([[ 2.6776e-09, 2.0955e-09, 3.4925e-10, ..., 2.3283e-10, + 4.6566e-10, 1.7462e-09], + [-8.3819e-09, -6.4028e-09, 1.1642e-10, ..., 2.5611e-09, + -1.9558e-08, 1.9791e-09], + [ 3.0268e-09, 2.7940e-09, 0.0000e+00, ..., -7.6834e-08, + -1.7579e-08, -8.6031e-08], + ..., + [ 1.4203e-08, 1.9558e-08, 1.0477e-09, ..., 7.7533e-08, + 3.1549e-08, 1.2142e-07], + [ 6.9849e-10, 1.2806e-09, 1.1642e-10, ..., 9.3132e-10, + 1.9791e-09, 1.5134e-09], + [-1.1758e-08, -2.5611e-08, -2.2119e-09, ..., 1.3388e-08, + -3.4925e-10, -1.3621e-08]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0438, 0.0142, 0.0120, 0.0263, 0.0315, -0.0214, 0.0299, 0.0339, + -0.0133, 0.0007], device='cuda:0'), grad: tensor([ 3.0501e-08, 9.0804e-09, -1.9255e-07, -1.6741e-07, -9.9302e-08, + 9.4529e-08, 6.5193e-09, 4.6729e-07, 1.4319e-08, -1.3784e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 251.59, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4301 re_mapping 0.0023 re_causal 0.0078 /// teacc 99.21 lr 0.00010000 +Epoch 382, weight, value: tensor([[ 0.0879, -0.1872, -0.1458, ..., -0.1440, -0.1752, -0.1697], + [ 0.0454, 0.0550, -0.0331, ..., -0.0791, 0.1720, 0.0190], + [-0.0557, -0.1883, -0.0096, ..., 0.2213, -0.1939, 0.0412], + ..., + [-0.2227, 0.1549, 0.0233, ..., -0.1657, 0.1240, 0.0651], + [ 0.0045, -0.0912, -0.0050, ..., -0.1664, -0.1939, -0.3015], + [ 0.0359, -0.0589, -0.0182, ..., -0.2634, -0.1940, 0.0219]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 0.0000e+00, 1.1642e-10], + [-3.2596e-09, -8.1491e-10, 0.0000e+00, ..., -2.4447e-09, + -1.1292e-08, -8.1491e-10], + [ 1.1642e-09, 8.1491e-10, 0.0000e+00, ..., -1.6182e-08, + 1.1642e-10, -1.0710e-08], + ..., + [ 9.3132e-10, -4.1910e-09, 0.0000e+00, ..., 1.9209e-08, + 2.4447e-09, 1.1758e-08], + [ 4.3074e-09, 4.6566e-10, 0.0000e+00, ..., 2.4447e-09, + 3.2596e-09, 2.3283e-09], + [-5.8208e-10, 1.9791e-09, -1.1642e-10, ..., 1.0477e-09, + 1.5134e-09, -3.9581e-09]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0439, 0.0140, 0.0121, 0.0263, 0.0313, -0.0212, 0.0299, 0.0339, + -0.0131, 0.0008], device='cuda:0'), grad: tensor([ 1.9791e-09, -1.0245e-08, -2.4680e-08, 2.5495e-08, 9.1968e-09, + -2.2980e-07, 1.7439e-07, 4.0862e-08, 3.5390e-08, -1.9092e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 251.66, cls_loss 0.0008 cls_loss_mapping 0.0021 cls_loss_causal 0.4812 re_mapping 0.0024 re_causal 0.0079 /// teacc 99.11 lr 0.00010000 +Epoch 383, weight, value: tensor([[ 0.0881, -0.1872, -0.1458, ..., -0.1440, -0.1752, -0.1697], + [ 0.0456, 0.0550, -0.0331, ..., -0.0792, 0.1721, 0.0190], + [-0.0582, -0.1885, -0.0096, ..., 0.2228, -0.1927, 0.0421], + ..., + [-0.2234, 0.1551, 0.0233, ..., -0.1673, 0.1241, 0.0651], + [ 0.0049, -0.0915, -0.0050, ..., -0.1695, -0.1968, -0.3048], + [ 0.0358, -0.0589, -0.0182, ..., -0.2637, -0.1941, 0.0221]], + device='cuda:0'), grad: tensor([[ 2.3283e-09, 5.8208e-10, 0.0000e+00, ..., 3.0268e-09, + 2.2119e-09, 4.3074e-09], + [-9.0804e-09, 2.4214e-08, 0.0000e+00, ..., 8.4983e-09, + -2.4447e-09, 2.6077e-08], + [-7.1013e-09, 4.0745e-08, 0.0000e+00, ..., -1.8498e-07, + -9.9535e-08, -2.1327e-07], + ..., + [ 1.5600e-08, -1.1735e-07, 0.0000e+00, ..., 1.3749e-07, + 3.0850e-08, 9.1386e-08], + [ 2.0722e-08, 4.8894e-09, 0.0000e+00, ..., 4.4238e-08, + 3.5740e-08, 5.9954e-08], + [-1.0012e-08, 2.7125e-08, 0.0000e+00, ..., 1.2806e-09, + 1.7113e-08, -1.3271e-08]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0436, 0.0139, 0.0131, 0.0261, 0.0299, -0.0213, 0.0329, 0.0336, + -0.0157, 0.0007], device='cuda:0'), grad: tensor([ 2.4913e-08, 7.0664e-08, -6.6916e-07, -2.9686e-08, 1.2177e-07, + -1.3970e-07, 2.8522e-08, 3.3900e-07, 2.8289e-07, -2.7474e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 251.59, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4600 re_mapping 0.0024 re_causal 0.0077 /// teacc 99.10 lr 0.00010000 +Epoch 384, weight, value: tensor([[ 0.0881, -0.1872, -0.1458, ..., -0.1451, -0.1752, -0.1697], + [ 0.0456, 0.0548, -0.0331, ..., -0.0800, 0.1720, 0.0186], + [-0.0574, -0.1889, -0.0096, ..., 0.2232, -0.1928, 0.0424], + ..., + [-0.2238, 0.1556, 0.0233, ..., -0.1678, 0.1245, 0.0654], + [ 0.0048, -0.0917, -0.0050, ..., -0.1695, -0.1970, -0.3050], + [ 0.0359, -0.0590, -0.0182, ..., -0.2644, -0.1943, 0.0221]], + device='cuda:0'), grad: tensor([[-7.1013e-09, 1.1642e-10, 0.0000e+00, ..., 1.1642e-09, + -9.3132e-10, 8.1491e-10], + [ 4.3656e-08, 1.5716e-08, 0.0000e+00, ..., 1.3295e-07, + 8.5216e-08, 1.8987e-07], + [ 1.9674e-08, 5.8208e-09, 0.0000e+00, ..., 1.0361e-08, + 4.7730e-09, 1.3388e-08], + ..., + [ 1.2573e-08, 3.6089e-09, 0.0000e+00, ..., 6.7172e-08, + 4.8080e-08, 9.9419e-08], + [-1.2538e-07, -3.3411e-08, 0.0000e+00, ..., -3.2363e-08, + 1.5134e-09, -4.2608e-08], + [-4.5402e-09, 4.6566e-10, 0.0000e+00, ..., 3.6089e-09, + 1.6298e-09, -5.9372e-09]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0437, 0.0134, 0.0134, 0.0259, 0.0300, -0.0212, 0.0334, 0.0339, + -0.0159, 0.0007], device='cuda:0'), grad: tensor([-3.7020e-08, 6.5286e-07, 1.0384e-07, -2.6776e-09, -6.9523e-07, + 2.2457e-07, -1.5716e-08, 3.0594e-07, -5.0338e-07, -1.3504e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 251.58, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4546 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.09 lr 0.00010000 +Epoch 385, weight, value: tensor([[ 0.0881, -0.1872, -0.1458, ..., -0.1452, -0.1752, -0.1697], + [ 0.0458, 0.0548, -0.0331, ..., -0.0802, 0.1721, 0.0185], + [-0.0588, -0.1895, -0.0096, ..., 0.2233, -0.1931, 0.0424], + ..., + [-0.2242, 0.1557, 0.0233, ..., -0.1683, 0.1245, 0.0654], + [ 0.0045, -0.0918, -0.0050, ..., -0.1696, -0.1971, -0.3051], + [ 0.0359, -0.0590, -0.0182, ..., -0.2655, -0.1944, 0.0221]], + device='cuda:0'), grad: tensor([[-7.5088e-09, 2.9104e-10, 0.0000e+00, ..., 2.3283e-10, + 1.1059e-09, 1.0477e-09], + [-1.0227e-07, -1.1321e-07, 0.0000e+00, ..., 1.7462e-09, + -2.5891e-07, -1.6415e-07], + [ 4.8196e-08, 3.6147e-08, 0.0000e+00, ..., -1.8044e-09, + 9.1386e-08, 6.5193e-08], + ..., + [ 8.5915e-08, 4.2550e-08, 0.0000e+00, ..., 1.8044e-09, + 1.2596e-07, 1.0605e-07], + [-5.5507e-07, 5.0059e-09, 0.0000e+00, ..., 4.5984e-09, + -4.2049e-07, -5.8021e-07], + [ 5.5740e-07, 1.7171e-08, 0.0000e+00, ..., 1.2806e-09, + 4.1141e-07, 5.3272e-07]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0437, 0.0133, 0.0133, 0.0260, 0.0302, -0.0209, 0.0340, 0.0338, + -0.0162, 0.0006], device='cuda:0'), grad: tensor([-3.3062e-08, -5.5833e-07, 2.8894e-07, 3.8650e-08, 1.7881e-07, + -1.7397e-06, 1.1604e-06, 5.3411e-07, -3.4068e-06, 3.5651e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 251.43, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4484 re_mapping 0.0024 re_causal 0.0080 /// teacc 99.08 lr 0.00010000 +Epoch 386, weight, value: tensor([[ 0.0891, -0.1872, -0.1458, ..., -0.1421, -0.1752, -0.1697], + [ 0.0460, 0.0548, -0.0331, ..., -0.0802, 0.1721, 0.0184], + [-0.0596, -0.1902, -0.0096, ..., 0.2234, -0.1938, 0.0422], + ..., + [-0.2244, 0.1560, 0.0233, ..., -0.1687, 0.1249, 0.0658], + [ 0.0045, -0.0921, -0.0050, ..., -0.1697, -0.1972, -0.3052], + [ 0.0359, -0.0590, -0.0182, ..., -0.2661, -0.1945, 0.0219]], + device='cuda:0'), grad: tensor([[-8.0327e-09, 1.7462e-10, 0.0000e+00, ..., 1.0477e-09, + 0.0000e+00, 1.0477e-09], + [-1.1642e-10, 8.1491e-10, 0.0000e+00, ..., 1.3970e-08, + -1.5716e-09, 1.2689e-08], + [-9.8953e-10, -1.2456e-07, 0.0000e+00, ..., -6.5891e-07, + 3.4925e-10, -5.3924e-07], + ..., + [ 5.8208e-09, 1.2433e-07, 0.0000e+00, ..., 6.5193e-07, + 4.3656e-09, 5.5088e-07], + [ 1.2806e-09, 8.1491e-10, 0.0000e+00, ..., 2.8522e-09, + 7.5670e-10, 3.3178e-09], + [-1.3388e-09, 6.4028e-10, 5.8208e-11, ..., 5.7044e-09, + -2.9104e-10, 8.7311e-10]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0428, 0.0133, 0.0131, 0.0260, 0.0305, -0.0210, 0.0309, 0.0343, + -0.0163, 0.0006], device='cuda:0'), grad: tensor([-3.7486e-08, 4.4121e-08, -1.6624e-06, 2.7998e-08, -9.0455e-08, + 5.0641e-09, 5.0059e-09, 1.7006e-06, 1.4319e-08, 1.1583e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 251.70, cls_loss 0.0007 cls_loss_mapping 0.0013 cls_loss_causal 0.4485 re_mapping 0.0024 re_causal 0.0081 /// teacc 99.12 lr 0.00010000 +Epoch 387, weight, value: tensor([[ 0.0891, -0.1872, -0.1458, ..., -0.1421, -0.1752, -0.1697], + [ 0.0460, 0.0548, -0.0331, ..., -0.0803, 0.1722, 0.0183], + [-0.0604, -0.1902, -0.0096, ..., 0.2236, -0.1941, 0.0422], + ..., + [-0.2250, 0.1556, 0.0233, ..., -0.1697, 0.1249, 0.0649], + [ 0.0059, -0.0918, -0.0050, ..., -0.1696, -0.1973, -0.3053], + [ 0.0359, -0.0585, -0.0182, ..., -0.2665, -0.1940, 0.0225]], + device='cuda:0'), grad: tensor([[ 5.8208e-10, 1.1642e-10, 0.0000e+00, ..., 1.0477e-09, + 1.7462e-10, 1.7462e-10], + [-1.3446e-08, -2.3574e-08, 0.0000e+00, ..., 2.9104e-10, + -5.0117e-08, -1.8277e-08], + [ 7.7416e-09, 5.0641e-09, 0.0000e+00, ..., 2.2119e-09, + 1.1700e-08, 6.5193e-09], + ..., + [ 1.2515e-08, 1.2282e-08, 0.0000e+00, ..., 1.2224e-09, + 2.6368e-08, 1.0245e-08], + [-1.3225e-07, 1.3970e-09, 0.0000e+00, ..., -8.7311e-10, + 2.7940e-09, 1.1642e-09], + [ 1.6880e-09, 3.4925e-10, 0.0000e+00, ..., 2.0373e-09, + 6.4028e-10, -3.4925e-09]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0428, 0.0131, 0.0130, 0.0263, 0.0319, -0.0206, 0.0294, 0.0332, + -0.0155, 0.0008], device='cuda:0'), grad: tensor([ 6.8103e-09, -5.9139e-08, 4.1095e-08, 2.3236e-07, 1.8568e-08, + 5.9977e-07, -2.6869e-07, 6.5775e-08, -6.5053e-07, 1.3853e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 251.38, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4626 re_mapping 0.0026 re_causal 0.0085 /// teacc 99.10 lr 0.00010000 +Epoch 388, weight, value: tensor([[ 0.0891, -0.1872, -0.1458, ..., -0.1421, -0.1752, -0.1697], + [ 0.0471, 0.0542, -0.0331, ..., -0.0804, 0.1721, 0.0181], + [-0.0638, -0.1905, -0.0096, ..., 0.2236, -0.1945, 0.0420], + ..., + [-0.2271, 0.1565, 0.0233, ..., -0.1699, 0.1252, 0.0653], + [ 0.0066, -0.0923, -0.0050, ..., -0.1696, -0.1976, -0.3054], + [ 0.0359, -0.0589, -0.0182, ..., -0.2711, -0.1941, 0.0218]], + device='cuda:0'), grad: tensor([[ 2.9686e-09, 3.4925e-10, 0.0000e+00, ..., 1.5134e-09, + 1.5716e-09, 8.4401e-09], + [-1.2806e-08, -3.0268e-08, 0.0000e+00, ..., 3.8999e-08, + -3.6322e-08, 5.0408e-08], + [ 1.0303e-08, 2.9104e-09, 0.0000e+00, ..., -3.0035e-08, + 6.1118e-09, -1.5658e-08], + ..., + [ 5.4832e-08, 1.8685e-08, 0.0000e+00, ..., 9.3132e-09, + 9.7090e-08, 1.3376e-07], + [ 3.5041e-08, 2.0955e-09, 0.0000e+00, ..., 1.9209e-09, + 5.1805e-09, 5.1572e-08], + [-1.6589e-07, 1.0477e-09, 0.0000e+00, ..., 3.5507e-08, + -1.3527e-07, -3.6275e-07]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0428, 0.0131, 0.0128, 0.0263, 0.0330, -0.0213, 0.0293, 0.0332, + -0.0152, 0.0006], device='cuda:0'), grad: tensor([ 3.0675e-08, 1.8731e-07, -2.9337e-08, 1.1735e-07, 2.1933e-07, + 6.2049e-08, -3.5332e-08, 5.2992e-07, 2.1351e-07, -1.2917e-06], + device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 250.85, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4774 re_mapping 0.0024 re_causal 0.0079 /// teacc 98.99 lr 0.00010000 +Epoch 389, weight, value: tensor([[ 0.0891, -0.1872, -0.1458, ..., -0.1422, -0.1752, -0.1697], + [ 0.0503, 0.0572, -0.0331, ..., -0.0804, 0.1748, 0.0189], + [-0.0641, -0.1908, -0.0096, ..., 0.2238, -0.1947, 0.0422], + ..., + [-0.2308, 0.1532, 0.0233, ..., -0.1708, 0.1226, 0.0639], + [ 0.0041, -0.0932, -0.0050, ..., -0.1697, -0.1997, -0.3060], + [ 0.0360, -0.0588, -0.0182, ..., -0.2719, -0.1940, 0.0221]], + device='cuda:0'), grad: tensor([[-7.9744e-09, 1.1642e-10, 0.0000e+00, ..., 6.4028e-10, + -2.9104e-09, 5.2387e-10], + [-5.7626e-09, -1.4319e-08, 0.0000e+00, ..., 1.6298e-09, + -3.0675e-08, -6.9849e-09], + [ 1.4552e-09, 9.8953e-10, 0.0000e+00, ..., -1.6880e-09, + 2.1537e-09, 4.0745e-10], + ..., + [ 9.4878e-09, 7.4506e-09, 0.0000e+00, ..., 9.8953e-10, + 1.5250e-08, 9.6625e-09], + [-1.0128e-08, 2.3283e-10, 0.0000e+00, ..., 3.4925e-10, + 5.0059e-09, 2.0373e-09], + [-2.0373e-09, 7.5670e-10, 0.0000e+00, ..., 1.7462e-10, + 1.8626e-09, -5.2212e-08]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0428, 0.0159, 0.0129, 0.0261, 0.0330, -0.0211, 0.0292, 0.0295, + -0.0157, 0.0007], device='cuda:0'), grad: tensor([-2.2643e-08, -1.2515e-08, 1.2980e-08, 5.4250e-08, 1.6415e-07, + -8.0676e-08, 4.4820e-09, 6.4611e-08, -5.7858e-08, -1.2154e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 252.61, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4691 re_mapping 0.0023 re_causal 0.0082 /// teacc 99.23 lr 0.00010000 +Epoch 390, weight, value: tensor([[ 0.0891, -0.1872, -0.1458, ..., -0.1423, -0.1752, -0.1697], + [ 0.0505, 0.0572, -0.0331, ..., -0.0807, 0.1749, 0.0188], + [-0.0640, -0.1909, -0.0096, ..., 0.2240, -0.1948, 0.0423], + ..., + [-0.2309, 0.1532, 0.0233, ..., -0.1713, 0.1226, 0.0637], + [ 0.0045, -0.0931, -0.0050, ..., -0.1698, -0.1998, -0.3060], + [ 0.0359, -0.0588, -0.0182, ..., -0.2722, -0.1939, 0.0222]], + device='cuda:0'), grad: tensor([[-9.7789e-09, 1.7462e-10, 0.0000e+00, ..., 1.7462e-09, + 2.3283e-10, 3.4925e-10], + [-5.8790e-09, 1.9209e-09, 0.0000e+00, ..., 1.9209e-09, + -6.8685e-09, 1.1642e-10], + [ 3.8999e-09, 5.3551e-09, 0.0000e+00, ..., -7.5670e-09, + 7.9744e-09, -2.3283e-10], + ..., + [ 2.1537e-09, -1.4028e-08, 0.0000e+00, ..., 1.4552e-09, + -1.0710e-08, -9.6043e-09], + [ 2.5029e-09, 9.8953e-10, 0.0000e+00, ..., 1.1642e-09, + 2.6776e-09, 1.7462e-09], + [ 2.4447e-09, 2.3865e-09, 0.0000e+00, ..., 4.6566e-10, + 2.7358e-09, -1.5716e-09]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0428, 0.0158, 0.0130, 0.0263, 0.0330, -0.0212, 0.0293, 0.0293, + -0.0154, 0.0007], device='cuda:0'), grad: tensor([-3.5565e-08, 1.0012e-08, 8.3877e-08, -1.7299e-07, 4.4762e-08, + 5.2794e-08, -1.0070e-08, -4.7148e-09, 2.3458e-08, 1.0477e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 252.15, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4574 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.16 lr 0.00010000 +Epoch 391, weight, value: tensor([[ 0.0892, -0.1872, -0.1458, ..., -0.1424, -0.1752, -0.1697], + [ 0.0505, 0.0572, -0.0332, ..., -0.0808, 0.1749, 0.0188], + [-0.0644, -0.1911, -0.0092, ..., 0.2242, -0.1949, 0.0424], + ..., + [-0.2309, 0.1532, 0.0233, ..., -0.1716, 0.1226, 0.0637], + [ 0.0057, -0.0931, -0.0050, ..., -0.1699, -0.1999, -0.3061], + [ 0.0359, -0.0588, -0.0182, ..., -0.2724, -0.1939, 0.0223]], + device='cuda:0'), grad: tensor([[-1.1059e-08, -8.8476e-09, 0.0000e+00, ..., 1.7462e-10, + -6.4611e-09, 2.3865e-09], + [ 3.7253e-09, 6.5193e-09, 0.0000e+00, ..., 1.8044e-09, + 5.1805e-09, 1.3912e-08], + [ 1.9791e-09, 2.0373e-09, 0.0000e+00, ..., 2.9104e-10, + 1.6880e-09, 4.9477e-09], + ..., + [ 3.4343e-09, -3.6089e-09, 0.0000e+00, ..., 1.5716e-09, + -6.1118e-09, 1.4435e-08], + [ 4.2492e-09, 9.3132e-10, 0.0000e+00, ..., 5.8208e-11, + 6.9849e-10, 2.2119e-09], + [-2.3225e-08, -2.3574e-08, 0.0000e+00, ..., 6.7521e-09, + 3.4343e-09, -1.7590e-07]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0428, 0.0158, 0.0131, 0.0263, 0.0329, -0.0217, 0.0292, 0.0293, + -0.0147, 0.0008], device='cuda:0'), grad: tensor([-6.0827e-08, 6.5891e-08, 3.6554e-08, -1.2747e-08, 5.5460e-07, + -6.0536e-09, 1.8394e-08, 6.9966e-08, 2.6484e-08, -6.8173e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 251.15, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4737 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.23 lr 0.00010000 +Epoch 392, weight, value: tensor([[ 0.0892, -0.1872, -0.1458, ..., -0.1424, -0.1752, -0.1697], + [ 0.0505, 0.0571, -0.0332, ..., -0.0810, 0.1749, 0.0187], + [-0.0641, -0.1913, -0.0092, ..., 0.2244, -0.1950, 0.0425], + ..., + [-0.2310, 0.1533, 0.0233, ..., -0.1720, 0.1227, 0.0638], + [ 0.0026, -0.0932, -0.0050, ..., -0.1701, -0.1999, -0.3064], + [ 0.0359, -0.0588, -0.0182, ..., -0.2726, -0.1940, 0.0224]], + device='cuda:0'), grad: tensor([[ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 2.3865e-09, + 3.4925e-10, 1.4552e-09], + [ 1.9092e-08, -4.0745e-10, 0.0000e+00, ..., 6.5775e-09, + 5.8208e-11, 1.2515e-08], + [ 1.4086e-08, 2.3865e-09, 0.0000e+00, ..., -3.2422e-08, + 9.9535e-09, -1.1059e-09], + ..., + [ 1.8976e-08, -1.3504e-08, 0.0000e+00, ..., 8.9058e-09, + -8.1491e-10, 8.6729e-09], + [-7.7591e-08, 5.0059e-09, 0.0000e+00, ..., 2.3865e-08, + -2.6717e-08, -2.5379e-08], + [ 6.1118e-09, -4.3656e-09, 0.0000e+00, ..., 2.4214e-08, + 4.3074e-09, 1.4435e-08]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0428, 0.0157, 0.0132, 0.0287, 0.0327, -0.0217, 0.0292, 0.0293, + -0.0175, 0.0008], device='cuda:0'), grad: tensor([ 1.3155e-08, 1.6461e-07, 5.5647e-08, 9.8778e-08, -1.2410e-07, + 3.5565e-08, -2.2701e-09, 1.1246e-07, -4.6915e-07, 1.1717e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 252.37, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4322 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.12 lr 0.00010000 +Epoch 393, weight, value: tensor([[ 0.0892, -0.1872, -0.1458, ..., -0.1425, -0.1752, -0.1697], + [ 0.0506, 0.0571, -0.0332, ..., -0.0813, 0.1749, 0.0185], + [-0.0642, -0.1920, -0.0092, ..., 0.2245, -0.1952, 0.0425], + ..., + [-0.2311, 0.1535, 0.0233, ..., -0.1720, 0.1228, 0.0639], + [ 0.0026, -0.0932, -0.0050, ..., -0.1701, -0.2000, -0.3064], + [ 0.0359, -0.0588, -0.0182, ..., -0.2740, -0.1941, 0.0224]], + device='cuda:0'), grad: tensor([[ 6.4028e-10, 3.4925e-10, 0.0000e+00, ..., 9.3132e-10, + 1.0477e-09, 4.0745e-10], + [-4.9477e-08, -3.2654e-08, 0.0000e+00, ..., 1.1642e-09, + -1.0006e-07, -2.8289e-08], + [ 9.3714e-09, 4.5984e-09, 0.0000e+00, ..., -1.8044e-09, + 1.6589e-08, 4.8312e-09], + ..., + [ 1.5309e-08, 1.1816e-08, 0.0000e+00, ..., 2.9104e-10, + 3.2946e-08, 9.9535e-09], + [-1.0594e-08, 4.1910e-09, 0.0000e+00, ..., 5.2387e-10, + 1.1409e-08, 3.3178e-09], + [ 2.2002e-08, 1.9209e-09, 0.0000e+00, ..., 5.8208e-10, + 5.8790e-09, 1.4552e-09]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0428, 0.0156, 0.0132, 0.0286, 0.0327, -0.0211, 0.0292, 0.0294, + -0.0175, 0.0008], device='cuda:0'), grad: tensor([ 5.6461e-09, -1.8300e-07, 3.1141e-08, 1.2456e-08, 1.7975e-07, + -2.3982e-08, -1.2317e-07, 6.7870e-08, -6.0012e-08, 1.1286e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 252.25, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4876 re_mapping 0.0023 re_causal 0.0083 /// teacc 99.01 lr 0.00010000 +Epoch 394, weight, value: tensor([[ 0.0892, -0.1872, -0.1458, ..., -0.1425, -0.1752, -0.1697], + [ 0.0506, 0.0571, -0.0333, ..., -0.0805, 0.1752, 0.0189], + [-0.0642, -0.1923, -0.0092, ..., 0.2242, -0.1962, 0.0420], + ..., + [-0.2311, 0.1536, 0.0233, ..., -0.1726, 0.1230, 0.0641], + [ 0.0025, -0.0932, -0.0048, ..., -0.1702, -0.2001, -0.3066], + [ 0.0359, -0.0589, -0.0182, ..., -0.2747, -0.1942, 0.0223]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 5.8208e-11, 0.0000e+00, ..., 5.2387e-10, + 5.8208e-11, 3.4925e-10], + [ 1.1642e-10, 1.1642e-09, 0.0000e+00, ..., 3.6089e-09, + 9.8953e-10, 5.3551e-09], + [-2.9104e-10, 6.4028e-10, 0.0000e+00, ..., -2.8522e-09, + 6.9849e-10, -1.6298e-09], + ..., + [ 6.4028e-10, -1.3970e-09, 0.0000e+00, ..., 2.3283e-09, + -6.9849e-10, 1.6298e-09], + [ 9.8953e-10, 2.9104e-10, 0.0000e+00, ..., 8.1491e-10, + 5.8208e-10, 8.7311e-10], + [ 2.9104e-10, 1.2224e-09, 0.0000e+00, ..., 5.0641e-09, + 1.1059e-09, 8.2073e-09]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0428, 0.0159, 0.0126, 0.0286, 0.0327, -0.0205, 0.0292, 0.0294, + -0.0176, 0.0008], device='cuda:0'), grad: tensor([ 3.0850e-09, 1.5658e-08, 5.2387e-10, -2.9802e-08, -3.2713e-08, + 1.4843e-08, 2.3865e-09, 7.7998e-09, 1.0710e-08, 2.2002e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 252.24, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4396 re_mapping 0.0023 re_causal 0.0078 /// teacc 99.12 lr 0.00010000 +Epoch 395, weight, value: tensor([[ 0.0889, -0.1873, -0.1458, ..., -0.1425, -0.1752, -0.1697], + [ 0.0507, 0.0571, -0.0333, ..., -0.0806, 0.1753, 0.0188], + [-0.0641, -0.1924, -0.0091, ..., 0.2243, -0.1963, 0.0421], + ..., + [-0.2313, 0.1537, 0.0232, ..., -0.1734, 0.1231, 0.0642], + [ 0.0025, -0.0932, -0.0048, ..., -0.1702, -0.2001, -0.3067], + [ 0.0359, -0.0590, -0.0182, ..., -0.2768, -0.1945, 0.0219]], + device='cuda:0'), grad: tensor([[-9.8953e-10, 1.1642e-10, 0.0000e+00, ..., 2.9104e-10, + 1.7462e-10, 1.1642e-10], + [-7.1013e-09, -5.4715e-09, 0.0000e+00, ..., 1.3970e-09, + -1.2282e-08, -1.4552e-09], + [ 1.8626e-09, 1.1059e-09, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-09, 6.4028e-10], + ..., + [ 3.4343e-09, 2.4447e-09, 0.0000e+00, ..., 6.4028e-10, + 5.4133e-09, 2.0373e-09], + [ 1.7462e-10, 9.3132e-10, 0.0000e+00, ..., 6.9849e-10, + 1.9209e-09, 8.1491e-10], + [ 1.3970e-09, 1.7462e-10, 0.0000e+00, ..., 1.6298e-09, + 4.0745e-10, 1.8626e-09]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0429, 0.0158, 0.0127, 0.0285, 0.0329, -0.0186, 0.0292, 0.0294, + -0.0177, 0.0008], device='cuda:0'), grad: tensor([-1.5716e-09, -1.5018e-08, 8.5565e-09, 8.2655e-09, -1.0710e-08, + 3.4925e-10, -1.0477e-09, 1.5367e-08, -3.7253e-09, 1.6065e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 252.67, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4509 re_mapping 0.0022 re_causal 0.0076 /// teacc 99.17 lr 0.00010000 +Epoch 396, weight, value: tensor([[ 0.0889, -0.1873, -0.1458, ..., -0.1426, -0.1753, -0.1697], + [ 0.0508, 0.0571, -0.0333, ..., -0.0807, 0.1752, 0.0187], + [-0.0660, -0.1927, -0.0091, ..., 0.2244, -0.1970, 0.0418], + ..., + [-0.2314, 0.1539, 0.0232, ..., -0.1736, 0.1234, 0.0648], + [ 0.0026, -0.0932, -0.0048, ..., -0.1702, -0.2002, -0.3068], + [ 0.0359, -0.0591, -0.0182, ..., -0.2771, -0.1948, 0.0218]], + device='cuda:0'), grad: tensor([[ 1.4610e-08, 9.3132e-10, 0.0000e+00, ..., 2.9104e-09, + 8.1491e-10, 3.7253e-09], + [-5.2969e-08, -4.9011e-08, 0.0000e+00, ..., -9.3132e-09, + -1.2107e-07, -3.4634e-08], + [ 3.7893e-08, 1.6298e-08, 0.0000e+00, ..., -2.4855e-08, + 7.1188e-08, -1.6764e-08], + ..., + [ 1.2224e-08, 1.4435e-08, 0.0000e+00, ..., 2.2701e-08, + 1.9965e-08, 3.4226e-08], + [-2.0606e-08, 8.0327e-09, 0.0000e+00, ..., 6.9849e-09, + 1.2689e-08, 1.1001e-08], + [-1.0128e-08, 3.0268e-09, 0.0000e+00, ..., 5.4133e-09, + 4.9477e-09, -1.4727e-08]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0429, 0.0157, 0.0126, 0.0285, 0.0328, -0.0187, 0.0292, 0.0296, + -0.0175, 0.0007], device='cuda:0'), grad: tensor([ 8.5915e-08, -1.6880e-07, 7.3924e-09, -4.8720e-08, 4.7672e-08, + 4.9884e-08, 1.2980e-08, 1.6880e-07, -1.0739e-07, -1.7870e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 252.83, cls_loss 0.0004 cls_loss_mapping 0.0008 cls_loss_causal 0.4610 re_mapping 0.0022 re_causal 0.0078 /// teacc 99.21 lr 0.00010000 +Epoch 397, weight, value: tensor([[ 0.0889, -0.1873, -0.1458, ..., -0.1427, -0.1753, -0.1698], + [ 0.0504, 0.0571, -0.0333, ..., -0.0808, 0.1752, 0.0183], + [-0.0662, -0.1928, -0.0091, ..., 0.2245, -0.1972, 0.0418], + ..., + [-0.2314, 0.1540, 0.0232, ..., -0.1739, 0.1236, 0.0650], + [ 0.0027, -0.0932, -0.0048, ..., -0.1703, -0.2002, -0.3069], + [ 0.0361, -0.0592, -0.0182, ..., -0.2774, -0.1949, 0.0219]], + device='cuda:0'), grad: tensor([[-1.7462e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, + -1.1642e-10, 1.3970e-09], + [-7.1013e-09, -2.7358e-09, 0.0000e+00, ..., 1.0477e-09, + -1.5891e-08, -4.0163e-09], + [ 3.3178e-09, 3.7835e-09, 0.0000e+00, ..., -7.5670e-09, + 6.6357e-09, -1.9791e-09], + ..., + [ 4.3074e-09, 1.1001e-08, 0.0000e+00, ..., 4.1327e-09, + 3.1432e-09, 2.7241e-08], + [-5.3551e-09, -1.9791e-09, 0.0000e+00, ..., 2.9104e-10, + 2.6776e-09, 1.9209e-09], + [ 3.3178e-09, -2.3458e-08, 0.0000e+00, ..., 2.9104e-10, + 3.2014e-09, -4.9302e-08]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0429, 0.0155, 0.0126, 0.0285, 0.0329, -0.0188, 0.0292, 0.0297, + -0.0175, 0.0008], device='cuda:0'), grad: tensor([ 8.1491e-09, 3.8999e-09, 1.1409e-08, 1.4552e-09, 1.1490e-07, + 9.4878e-09, 1.4959e-08, 1.2782e-07, -6.8219e-08, -2.0268e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 253.06, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4301 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.15 lr 0.00010000 +Epoch 398, weight, value: tensor([[ 0.0889, -0.1873, -0.1459, ..., -0.1427, -0.1753, -0.1698], + [ 0.0505, 0.0571, -0.0333, ..., -0.0810, 0.1750, 0.0177], + [-0.0663, -0.1930, -0.0091, ..., 0.2247, -0.1972, 0.0420], + ..., + [-0.2316, 0.1540, 0.0232, ..., -0.1742, 0.1238, 0.0656], + [ 0.0026, -0.0930, -0.0049, ..., -0.1703, -0.2003, -0.3069], + [ 0.0361, -0.0592, -0.0182, ..., -0.2777, -0.1951, 0.0219]], + device='cuda:0'), grad: tensor([[ 6.4028e-10, 6.4028e-10, 0.0000e+00, ..., 1.1642e-10, + 6.9849e-10, 4.0745e-10], + [-1.9500e-08, -1.7346e-08, 0.0000e+00, ..., 2.9104e-10, + -3.2422e-08, -5.8790e-09], + [ 7.8580e-09, 9.0804e-09, 0.0000e+00, ..., 6.9849e-10, + 8.2655e-09, 5.9954e-09], + ..., + [ 9.0222e-09, -2.1013e-08, 0.0000e+00, ..., 1.1642e-10, + -7.7998e-09, -1.7462e-08], + [ 1.0128e-08, 1.0361e-08, 0.0000e+00, ..., 5.8208e-11, + 1.3795e-08, 5.0641e-09], + [ 1.8626e-09, 8.6729e-09, 0.0000e+00, ..., 2.9104e-10, + 7.9162e-09, 5.9954e-09]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0429, 0.0152, 0.0127, 0.0286, 0.0332, -0.0183, 0.0292, 0.0299, + -0.0176, 0.0007], device='cuda:0'), grad: tensor([ 7.6834e-09, -6.1118e-08, 1.0809e-07, -1.9209e-07, 2.1304e-08, + 5.8324e-08, 2.7358e-09, -3.1432e-08, 6.4203e-08, 3.0210e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 252.54, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4135 re_mapping 0.0023 re_causal 0.0073 /// teacc 99.16 lr 0.00010000 +Epoch 399, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1428, -0.1752, -0.1698], + [ 0.0510, 0.0575, -0.0333, ..., -0.0811, 0.1745, 0.0169], + [-0.0663, -0.1932, -0.0091, ..., 0.2257, -0.1974, 0.0427], + ..., + [-0.2321, 0.1537, 0.0232, ..., -0.1754, 0.1230, 0.0649], + [ 0.0027, -0.0932, -0.0049, ..., -0.1709, -0.2012, -0.3079], + [ 0.0361, -0.0593, -0.0182, ..., -0.2789, -0.1959, 0.0216]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 1.1642e-10, 0.0000e+00, ..., 5.8208e-10, + 2.9104e-10, 2.3283e-10], + [-2.1362e-08, -1.4261e-08, 0.0000e+00, ..., 2.5029e-09, + -4.2375e-08, -1.0245e-08], + [ 6.7521e-09, 5.1805e-09, 0.0000e+00, ..., 8.1491e-10, + 1.3853e-08, 4.7730e-09], + ..., + [ 9.1968e-09, 6.2864e-09, 0.0000e+00, ..., 1.5716e-09, + 1.7579e-08, 6.8685e-09], + [ 2.7358e-09, 1.4552e-09, 0.0000e+00, ..., 4.6566e-10, + 5.1223e-09, 2.7358e-09], + [-9.3132e-10, 8.7311e-10, 0.0000e+00, ..., 2.5553e-08, + 2.2701e-09, 1.3446e-08]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0429, 0.0142, 0.0135, 0.0287, 0.0356, -0.0193, 0.0292, 0.0289, + -0.0178, 0.0006], device='cuda:0'), grad: tensor([ 5.1223e-09, -6.8278e-08, 6.3039e-08, -7.7940e-08, -7.5321e-08, + 2.3225e-08, -6.5775e-09, 7.8056e-08, 1.7753e-08, 5.5414e-08], + device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 252.23, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4625 re_mapping 0.0022 re_causal 0.0078 /// teacc 99.10 lr 0.00010000 +Epoch 400, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1427, -0.1750, -0.1698], + [ 0.0511, 0.0575, -0.0333, ..., -0.0813, 0.1747, 0.0168], + [-0.0666, -0.1939, -0.0092, ..., 0.2259, -0.1977, 0.0427], + ..., + [-0.2323, 0.1539, 0.0232, ..., -0.1768, 0.1231, 0.0650], + [ 0.0030, -0.0933, -0.0050, ..., -0.1708, -0.2017, -0.3082], + [ 0.0361, -0.0594, -0.0182, ..., -0.2795, -0.1960, 0.0216]], + device='cuda:0'), grad: tensor([[-5.6461e-09, 8.7311e-10, 0.0000e+00, ..., 5.2387e-10, + 6.4028e-10, 8.1491e-10], + [ 3.6671e-09, 8.1607e-08, 0.0000e+00, ..., 6.2864e-09, + 8.1549e-08, 7.5146e-08], + [ 3.6671e-09, 2.2585e-08, 0.0000e+00, ..., -0.0000e+00, + 2.4622e-08, 1.8976e-08], + ..., + [ 1.2806e-09, -1.6775e-07, 0.0000e+00, ..., 2.3283e-09, + -1.5856e-07, -1.4005e-07], + [-3.0443e-08, 2.1537e-09, 0.0000e+00, ..., -1.0477e-08, + 1.8044e-09, 2.0373e-09], + [ 4.8312e-09, 3.2713e-08, 0.0000e+00, ..., 1.0710e-08, + 2.7590e-08, 3.4575e-08]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0429, 0.0142, 0.0135, 0.0289, 0.0356, -0.0195, 0.0287, 0.0289, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([-5.7509e-08, 3.0454e-07, 9.3598e-08, 8.9698e-08, -1.4028e-08, + 1.0600e-07, 5.6520e-08, -5.1269e-07, -2.3120e-07, 1.8114e-07], + device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 252.23, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4619 re_mapping 0.0022 re_causal 0.0075 /// teacc 99.15 lr 0.00001000 +Epoch 401, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1428, -0.1750, -0.1698], + [ 0.0517, 0.0575, -0.0336, ..., -0.0815, 0.1750, 0.0168], + [-0.0693, -0.1949, -0.0082, ..., 0.2261, -0.1986, 0.0426], + ..., + [-0.2327, 0.1541, 0.0232, ..., -0.1771, 0.1232, 0.0653], + [ 0.0031, -0.0934, -0.0050, ..., -0.1707, -0.2020, -0.3083], + [ 0.0361, -0.0594, -0.0182, ..., -0.2803, -0.1962, 0.0216]], + device='cuda:0'), grad: tensor([[ 8.1491e-10, 1.1642e-09, 0.0000e+00, ..., 1.3970e-09, + 2.6776e-09, 1.2806e-09], + [-3.9348e-08, -3.2189e-08, 0.0000e+00, ..., 8.6729e-09, + -7.6601e-08, -7.1013e-09], + [ 5.0059e-09, 4.0745e-09, 0.0000e+00, ..., -2.4005e-07, + 8.2073e-09, -2.6776e-07], + ..., + [ 1.6938e-08, 8.3819e-09, 0.0000e+00, ..., 2.4587e-07, + 2.6659e-08, 2.7521e-07], + [ 2.6776e-09, 9.0222e-09, 0.0000e+00, ..., 1.2806e-09, + 2.1129e-08, 4.8312e-09], + [ 2.4447e-09, 1.9791e-09, 0.0000e+00, ..., 7.7009e-08, + 3.5507e-09, 5.8790e-08]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0429, 0.0143, 0.0132, 0.0287, 0.0356, -0.0185, 0.0283, 0.0290, + -0.0169, 0.0006], device='cuda:0'), grad: tensor([ 7.7416e-09, -1.0600e-07, -6.5612e-07, 8.3062e-08, -4.2235e-07, + -2.5961e-08, 4.7905e-08, 7.5670e-07, 5.4715e-09, 3.1292e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 252.65, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4899 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.18 lr 0.00001000 +Epoch 402, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1428, -0.1750, -0.1698], + [ 0.0517, 0.0575, -0.0336, ..., -0.0815, 0.1750, 0.0168], + [-0.0694, -0.1950, -0.0082, ..., 0.2262, -0.1987, 0.0426], + ..., + [-0.2327, 0.1541, 0.0232, ..., -0.1773, 0.1232, 0.0653], + [ 0.0031, -0.0934, -0.0050, ..., -0.1707, -0.2021, -0.3084], + [ 0.0361, -0.0594, -0.0182, ..., -0.2804, -0.1962, 0.0215]], + device='cuda:0'), grad: tensor([[ 1.7521e-08, 1.1642e-10, 0.0000e+00, ..., 5.9954e-09, + 1.7462e-10, 3.7835e-09], + [-3.0850e-09, -5.2387e-10, 0.0000e+00, ..., 8.7311e-10, + -6.9849e-09, 3.8999e-09], + [ 8.7311e-10, 7.1595e-09, 0.0000e+00, ..., 1.0477e-09, + 7.7998e-09, 1.1525e-08], + ..., + [ 1.9209e-09, -1.7462e-08, 0.0000e+00, ..., 2.3283e-10, + -1.7986e-08, -1.2806e-08], + [-1.8044e-08, 4.6566e-10, 0.0000e+00, ..., -2.0955e-09, + 6.9849e-10, 2.1537e-09], + [-7.6834e-09, 1.9791e-09, 0.0000e+00, ..., 3.4925e-10, + 2.2119e-09, -5.9488e-08]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0429, 0.0143, 0.0133, 0.0286, 0.0356, -0.0185, 0.0283, 0.0290, + -0.0169, 0.0006], device='cuda:0'), grad: tensor([ 1.1711e-07, 7.1479e-08, 2.7753e-07, -3.1642e-07, 3.5623e-07, + -4.5495e-07, 2.1118e-07, 1.8161e-08, 1.3737e-08, -2.5821e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 252.30, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4538 re_mapping 0.0020 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 +Epoch 403, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1428, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0815, 0.1750, 0.0168], + [-0.0694, -0.1950, -0.0082, ..., 0.2262, -0.1987, 0.0426], + ..., + [-0.2328, 0.1541, 0.0232, ..., -0.1774, 0.1232, 0.0653], + [ 0.0031, -0.0934, -0.0050, ..., -0.1708, -0.2021, -0.3084], + [ 0.0361, -0.0594, -0.0182, ..., -0.2805, -0.1963, 0.0215]], + device='cuda:0'), grad: tensor([[ 1.1306e-06, 4.2864e-07, 0.0000e+00, ..., 6.9849e-10, + -1.6880e-09, 2.4564e-07], + [-1.2980e-08, -3.9581e-09, 0.0000e+00, ..., 3.6089e-09, + -2.1304e-08, 2.2119e-09], + [ 4.4820e-09, 5.2969e-09, 0.0000e+00, ..., -1.4086e-08, + 8.3819e-09, -3.4925e-09], + ..., + [ 9.1386e-09, -8.2073e-09, 0.0000e+00, ..., 2.5611e-09, + -1.5134e-09, -1.0245e-08], + [ 5.0059e-09, 3.0850e-09, 0.0000e+00, ..., 1.9791e-09, + 7.1595e-09, 3.4925e-09], + [-1.2079e-06, -4.5542e-07, 0.0000e+00, ..., 2.7358e-09, + 4.0163e-09, -2.5728e-07]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0429, 0.0143, 0.0133, 0.0286, 0.0356, -0.0185, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 5.8673e-06, -8.7311e-10, -3.0443e-08, 3.8941e-08, 3.4808e-07, + 1.7637e-08, 1.8568e-08, 5.1805e-09, 2.6077e-08, -6.2510e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 251.65, cls_loss 0.0005 cls_loss_mapping 0.0007 cls_loss_causal 0.4585 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.19 lr 0.00001000 +Epoch 404, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0815, 0.1750, 0.0168], + [-0.0695, -0.1950, -0.0082, ..., 0.2262, -0.1987, 0.0426], + ..., + [-0.2328, 0.1541, 0.0232, ..., -0.1775, 0.1232, 0.0652], + [ 0.0031, -0.0934, -0.0050, ..., -0.1708, -0.2021, -0.3084], + [ 0.0361, -0.0594, -0.0182, ..., -0.2805, -0.1963, 0.0215]], + device='cuda:0'), grad: tensor([[ 8.6729e-09, 3.2014e-09, 0.0000e+00, ..., 6.4028e-10, + 2.1537e-09, 1.9732e-08], + [ 8.7486e-08, -5.0059e-09, 0.0000e+00, ..., 1.0710e-08, + -4.8196e-08, 1.0813e-06], + [ 1.1525e-08, 7.1595e-09, 0.0000e+00, ..., -2.4098e-08, + 1.2165e-08, 3.1258e-08], + ..., + [ 5.2503e-08, -7.5670e-09, 0.0000e+00, ..., 1.0710e-08, + 9.1968e-09, 3.0966e-07], + [ 1.5367e-08, 2.3283e-09, 0.0000e+00, ..., 2.0373e-09, + 7.6834e-09, 3.6031e-08], + [-2.2980e-07, -8.7311e-09, 0.0000e+00, ..., 8.0909e-09, + -2.5029e-09, -1.7742e-06]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0429, 0.0143, 0.0133, 0.0286, 0.0356, -0.0184, 0.0283, 0.0290, + -0.0169, 0.0006], device='cuda:0'), grad: tensor([ 1.5576e-07, 4.8503e-06, 9.5367e-07, -1.8291e-06, 1.2117e-06, + 4.2981e-07, 3.8301e-08, 1.6158e-06, 3.7695e-07, -7.8082e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 251.60, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4512 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.16 lr 0.00001000 +Epoch 405, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0815, 0.1751, 0.0167], + [-0.0695, -0.1950, -0.0082, ..., 0.2262, -0.1987, 0.0426], + ..., + [-0.2328, 0.1541, 0.0232, ..., -0.1775, 0.1232, 0.0652], + [ 0.0031, -0.0934, -0.0050, ..., -0.1708, -0.2022, -0.3084], + [ 0.0361, -0.0594, -0.0182, ..., -0.2806, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[-3.2887e-08, 1.1642e-10, 0.0000e+00, ..., 1.1933e-09, + 2.0373e-10, 1.3679e-09], + [-1.0186e-09, -5.9954e-09, 0.0000e+00, ..., 4.9477e-10, + -1.9965e-08, 5.2387e-09], + [ 8.9931e-09, 2.0955e-09, 0.0000e+00, ..., -6.3446e-09, + 2.7940e-09, -2.8813e-09], + ..., + [ 1.0186e-08, 1.0768e-08, 0.0000e+00, ..., 1.2806e-09, + 5.7626e-09, 1.0070e-08], + [ 8.3237e-09, 1.3679e-09, 0.0000e+00, ..., 6.4028e-10, + 2.9686e-09, 6.1991e-09], + [-4.8603e-08, 1.7171e-09, 0.0000e+00, ..., 2.0373e-10, + 1.2515e-09, -1.0018e-07]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0429, 0.0143, 0.0133, 0.0286, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([-2.7614e-07, 7.2177e-08, 7.1304e-08, -1.5495e-07, 3.5111e-07, + 6.9849e-08, 1.6327e-08, 2.4750e-07, 6.5716e-08, -4.5565e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 251.81, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4368 re_mapping 0.0020 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 +Epoch 406, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0816, 0.1751, 0.0167], + [-0.0696, -0.1950, -0.0082, ..., 0.2262, -0.1987, 0.0426], + ..., + [-0.2328, 0.1542, 0.0232, ..., -0.1776, 0.1233, 0.0652], + [ 0.0031, -0.0934, -0.0050, ..., -0.1708, -0.2022, -0.3084], + [ 0.0362, -0.0594, -0.0182, ..., -0.2806, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[-3.5157e-08, -1.6298e-09, 0.0000e+00, ..., 8.1491e-10, + -1.1671e-08, 3.1432e-09], + [-1.0990e-07, -1.0146e-07, 0.0000e+00, ..., 4.6857e-09, + -2.0280e-07, -5.0088e-08], + [ 2.2206e-08, 1.8481e-08, 0.0000e+00, ..., -4.2171e-08, + 3.4837e-08, -1.7928e-08], + ..., + [ 2.7154e-08, 1.3184e-08, 0.0000e+00, ..., 2.8056e-08, + 3.5128e-08, 2.8143e-08], + [ 2.4971e-08, 1.8539e-08, 0.0000e+00, ..., 3.1723e-09, + 3.7864e-08, 1.3068e-08], + [ 3.1432e-09, 1.2602e-08, 0.0000e+00, ..., 9.0513e-09, + 1.3446e-08, -3.5798e-09]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0429, 0.0142, 0.0133, 0.0286, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([-1.5530e-07, -4.4075e-07, -2.0780e-08, 5.1979e-08, 1.6985e-07, + 2.5146e-08, 7.3225e-08, 1.8964e-07, 1.2200e-07, 2.9977e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 251.61, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4034 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.22 lr 0.00001000 +Epoch 407, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0816, 0.1751, 0.0167], + [-0.0696, -0.1950, -0.0082, ..., 0.2263, -0.1987, 0.0426], + ..., + [-0.2328, 0.1542, 0.0232, ..., -0.1777, 0.1233, 0.0652], + [ 0.0031, -0.0934, -0.0050, ..., -0.1708, -0.2022, -0.3085], + [ 0.0362, -0.0595, -0.0182, ..., -0.2806, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 1.7462e-10, 0.0000e+00, ..., 4.3656e-10, + 2.6193e-10, 4.3656e-10], + [-6.0245e-09, -2.6484e-09, 0.0000e+00, ..., 1.9791e-09, + -8.7603e-09, 1.0186e-09], + [ 1.0477e-09, 3.8708e-09, 0.0000e+00, ..., -9.1677e-09, + 3.7835e-09, -6.1991e-09], + ..., + [ 2.1828e-09, -4.0454e-09, 0.0000e+00, ..., 5.3260e-09, + -1.1933e-09, 9.3132e-10], + [ 2.9977e-09, 2.0373e-09, 0.0000e+00, ..., 1.2224e-09, + 3.9290e-09, 2.0955e-09], + [ 4.3656e-10, 1.1933e-09, 0.0000e+00, ..., 5.2387e-10, + 1.2224e-09, 1.2224e-09]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0429, 0.0142, 0.0133, 0.0286, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 3.4634e-09, -6.2573e-09, -2.8056e-08, -6.9849e-10, 6.2864e-09, + 7.4215e-09, 2.2119e-09, 1.5396e-08, 1.3388e-08, 5.5588e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 252.28, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4286 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.19 lr 0.00001000 +Epoch 408, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0816, 0.1751, 0.0167], + [-0.0696, -0.1951, -0.0082, ..., 0.2263, -0.1987, 0.0427], + ..., + [-0.2328, 0.1542, 0.0232, ..., -0.1777, 0.1233, 0.0652], + [ 0.0031, -0.0935, -0.0050, ..., -0.1708, -0.2022, -0.3085], + [ 0.0362, -0.0595, -0.0182, ..., -0.2806, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.4843e-09, 1.0477e-09, 0.0000e+00, ..., 7.2760e-10, + 2.3283e-09, 1.0186e-09], + [-6.4319e-08, -4.6624e-08, 0.0000e+00, ..., -6.1991e-09, + -1.0425e-07, -3.4750e-08], + [ 1.2544e-08, 1.0768e-08, 0.0000e+00, ..., 2.7649e-09, + 2.0897e-08, 8.9349e-09], + ..., + [ 5.9954e-09, -3.3469e-09, 0.0000e+00, ..., 1.0477e-09, + 4.0454e-09, -1.8335e-09], + [ 4.9477e-09, 4.0163e-09, 0.0000e+00, ..., 5.8208e-10, + 8.9349e-09, 3.0559e-09], + [ 1.6880e-09, 1.7753e-09, 0.0000e+00, ..., 3.5798e-09, + 2.3574e-09, 2.5611e-09]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0429, 0.0142, 0.0133, 0.0286, 0.0354, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 8.4401e-09, -2.7381e-07, 6.8743e-08, -6.3796e-08, 1.6624e-07, + 5.3900e-08, -1.8917e-09, 1.0943e-08, 2.2410e-08, 2.1741e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 251.86, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4307 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.18 lr 0.00001000 +Epoch 409, weight, value: tensor([[ 0.0889, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0816, 0.1751, 0.0167], + [-0.0696, -0.1951, -0.0082, ..., 0.2263, -0.1987, 0.0427], + ..., + [-0.2328, 0.1542, 0.0232, ..., -0.1778, 0.1233, 0.0652], + [ 0.0031, -0.0935, -0.0050, ..., -0.1708, -0.2022, -0.3085], + [ 0.0362, -0.0595, -0.0182, ..., -0.2807, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.5862e-08, 1.1642e-10, 0.0000e+00, ..., 6.1118e-10, + 2.0373e-10, 2.1537e-09], + [ 1.9791e-09, 7.7998e-09, 0.0000e+00, ..., 5.8208e-10, + 3.8126e-09, 1.2718e-08], + [ 2.5320e-09, 2.1537e-09, 0.0000e+00, ..., -1.1059e-09, + 2.5611e-09, 1.0477e-09], + ..., + [ 6.4611e-09, -2.0780e-08, 0.0000e+00, ..., 9.3132e-10, + -2.4680e-08, -2.1246e-08], + [ 7.8289e-09, 3.5507e-09, 0.0000e+00, ..., 3.4925e-10, + 6.7521e-09, 3.2596e-09], + [-5.2445e-08, 2.6193e-09, 0.0000e+00, ..., 4.9477e-10, + 3.3760e-09, -6.0827e-09]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 1.0675e-07, 9.7149e-08, 2.3720e-08, 6.0420e-08, 1.0780e-07, + -2.2847e-08, -4.5693e-09, -5.2183e-08, 4.8167e-08, -3.5041e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 251.88, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4589 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.12 lr 0.00001000 +Epoch 410, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0335, ..., -0.0816, 0.1751, 0.0167], + [-0.0696, -0.1951, -0.0082, ..., 0.2264, -0.1987, 0.0427], + ..., + [-0.2328, 0.1542, 0.0232, ..., -0.1778, 0.1233, 0.0652], + [ 0.0031, -0.0935, -0.0050, ..., -0.1708, -0.2023, -0.3085], + [ 0.0362, -0.0595, -0.0182, ..., -0.2807, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 6.1118e-10, 2.6193e-10, 0.0000e+00, ..., 1.6589e-09, + 6.9849e-10, 2.5902e-09], + [-2.2672e-08, -1.1380e-08, 0.0000e+00, ..., 4.4791e-08, + -3.5157e-08, 5.3638e-08], + [ 2.9104e-09, 1.9791e-09, 0.0000e+00, ..., 8.4401e-10, + 4.4820e-09, 4.7730e-09], + ..., + [ 7.3633e-09, 5.0350e-09, 0.0000e+00, ..., 3.2131e-08, + 1.3359e-08, 5.8470e-08], + [ 5.2096e-09, 3.4343e-09, 0.0000e+00, ..., 2.6776e-09, + 9.0804e-09, 7.1886e-09], + [-1.2224e-09, 6.8976e-09, 0.0000e+00, ..., 4.3481e-08, + 6.3446e-09, 4.4383e-08]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 1.3213e-08, 1.9441e-07, 2.3254e-08, 1.1001e-08, -7.1805e-07, + -1.4872e-08, 3.5419e-08, 2.5425e-07, 3.7544e-08, 1.8789e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 251.81, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4501 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.14 lr 0.00001000 +Epoch 411, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0816, 0.1751, 0.0167], + [-0.0697, -0.1951, -0.0081, ..., 0.2264, -0.1987, 0.0427], + ..., + [-0.2328, 0.1542, 0.0232, ..., -0.1779, 0.1233, 0.0652], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2023, -0.3085], + [ 0.0362, -0.0595, -0.0182, ..., -0.2808, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 4.1910e-09, 2.0664e-09, 0.0000e+00, ..., 6.3155e-09, + 2.9686e-09, 6.2864e-09], + [-6.8103e-09, 6.6939e-09, 0.0000e+00, ..., 2.1566e-08, + -3.3964e-08, 2.4156e-08], + [-3.5856e-08, -2.3108e-08, 0.0000e+00, ..., -7.7533e-08, + 3.7835e-09, -7.1130e-08], + ..., + [ 1.0303e-08, -7.8871e-09, 0.0000e+00, ..., 9.7207e-09, + 4.8021e-09, -1.4552e-10], + [ 1.3562e-08, 9.4587e-09, 0.0000e+00, ..., 2.6135e-08, + 6.1991e-09, 2.6077e-08], + [ 9.3132e-10, 2.5902e-09, 0.0000e+00, ..., 4.1036e-09, + 2.0373e-09, 1.7753e-09]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 3.9669e-08, 6.7172e-08, -3.6787e-07, -2.3603e-08, 5.8528e-08, + 2.4505e-08, 1.2602e-08, 4.8400e-08, 1.4016e-07, 1.0594e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 252.04, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4340 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.11 lr 0.00001000 +Epoch 412, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0817, 0.1751, 0.0167], + [-0.0697, -0.1951, -0.0081, ..., 0.2264, -0.1988, 0.0427], + ..., + [-0.2329, 0.1542, 0.0232, ..., -0.1780, 0.1233, 0.0652], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2023, -0.3085], + [ 0.0362, -0.0595, -0.0182, ..., -0.2808, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 1.1642e-10, 0.0000e+00, ..., 4.9477e-10, + 2.9104e-10, 2.0373e-10], + [-5.4424e-09, 3.1712e-07, 0.0000e+00, ..., 3.3178e-09, + 2.9337e-07, 2.1036e-07], + [ 9.3132e-10, 1.2020e-08, 0.0000e+00, ..., -2.7358e-09, + 1.1991e-08, 8.2073e-09], + ..., + [ 3.9581e-09, -3.5530e-07, 0.0000e+00, ..., 4.6566e-09, + -3.2946e-07, -2.3248e-07], + [-9.0222e-10, 4.0745e-09, 0.0000e+00, ..., 5.5297e-10, + 5.9372e-09, 2.7358e-09], + [ 8.1491e-10, 1.2980e-08, 0.0000e+00, ..., 2.1828e-09, + 1.2573e-08, 1.1845e-08]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 2.3283e-09, 7.0501e-07, 2.8260e-08, 1.9354e-08, -3.9872e-09, + 9.9535e-09, -1.7550e-08, -7.6694e-07, -8.7311e-10, 4.2754e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 252.26, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4192 re_mapping 0.0018 re_causal 0.0070 /// teacc 99.14 lr 0.00001000 +Epoch 413, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0336, ..., -0.0817, 0.1751, 0.0167], + [-0.0697, -0.1951, -0.0081, ..., 0.2264, -0.1988, 0.0427], + ..., + [-0.2329, 0.1542, 0.0232, ..., -0.1780, 0.1233, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2023, -0.3086], + [ 0.0362, -0.0595, -0.0182, ..., -0.2809, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 3.4925e-10, 0.0000e+00, ..., 2.6776e-09, + 5.2387e-10, 9.3132e-10], + [-8.8476e-09, 1.6880e-09, 0.0000e+00, ..., 4.8312e-09, + -6.5775e-09, 4.7148e-09], + [-3.0850e-09, 6.5775e-09, 0.0000e+00, ..., -3.7835e-08, + 7.5670e-09, -1.7055e-08], + ..., + [ 2.6776e-09, -3.1665e-08, 0.0000e+00, ..., 1.8626e-09, + -3.6554e-08, -1.4203e-08], + [ 7.9744e-09, 4.7148e-09, 0.0000e+00, ..., 3.1898e-08, + 9.5461e-09, 2.0780e-08], + [-1.5716e-09, 6.9849e-09, 0.0000e+00, ..., 1.7346e-08, + 8.0909e-09, 2.1071e-08]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 1.0303e-08, 2.2061e-08, -9.9011e-08, 2.9861e-08, -9.7323e-08, + 1.2456e-08, -1.6764e-08, -8.6497e-08, 1.2410e-07, 1.1432e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 252.16, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4252 re_mapping 0.0018 re_causal 0.0070 /// teacc 99.15 lr 0.00001000 +Epoch 414, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0335, ..., -0.0817, 0.1751, 0.0166], + [-0.0697, -0.1952, -0.0081, ..., 0.2264, -0.1988, 0.0427], + ..., + [-0.2329, 0.1542, 0.0232, ..., -0.1781, 0.1233, 0.0652], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2023, -0.3086], + [ 0.0362, -0.0595, -0.0182, ..., -0.2809, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.7462e-10, 2.3283e-10, 0.0000e+00, ..., 1.3970e-09, + 2.3283e-10, 4.6566e-10], + [-1.2864e-08, 6.4611e-09, 0.0000e+00, ..., 4.2492e-09, + -1.5076e-08, 8.6147e-09], + [ 1.6880e-09, 1.3679e-08, 0.0000e+00, ..., -9.4296e-09, + 8.6147e-09, 7.5670e-10], + ..., + [ 1.2689e-08, -3.1316e-08, 0.0000e+00, ..., 3.6089e-09, + -4.1327e-09, -1.8685e-08], + [-1.0128e-08, 2.1537e-09, 0.0000e+00, ..., -0.0000e+00, + 3.6089e-09, 3.4925e-10], + [ 8.0327e-09, 3.4343e-09, 0.0000e+00, ..., 5.8208e-10, + 2.9104e-09, 3.8417e-09]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.2456e-08, 6.9849e-09, 1.4494e-08, 2.2468e-08, + -1.4959e-08, -1.2806e-08, -3.4634e-08, -3.6904e-08, 3.6729e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 252.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4366 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.12 lr 0.00001000 +Epoch 415, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0335, ..., -0.0817, 0.1751, 0.0166], + [-0.0698, -0.1952, -0.0081, ..., 0.2265, -0.1988, 0.0427], + ..., + [-0.2329, 0.1542, 0.0232, ..., -0.1782, 0.1233, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2023, -0.3086], + [ 0.0362, -0.0595, -0.0182, ..., -0.2810, -0.1963, 0.0216]], + device='cuda:0'), grad: tensor([[-1.0477e-09, 5.8208e-11, 0.0000e+00, ..., 2.3283e-10, + -1.1642e-10, 2.3283e-10], + [-1.6298e-09, -1.3970e-09, 0.0000e+00, ..., 8.6147e-09, + -3.7253e-09, 1.0885e-08], + [ 6.9849e-10, 4.0745e-10, 0.0000e+00, ..., -3.3760e-09, + 9.8953e-10, -2.5029e-09], + ..., + [ 1.5134e-09, 1.0477e-09, 0.0000e+00, ..., 1.6415e-08, + 2.5611e-09, 2.1304e-08], + [-5.8208e-11, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, + 5.8208e-10, 4.0745e-10], + [ 8.1491e-10, 1.1642e-10, 0.0000e+00, ..., 6.7521e-09, + 3.4925e-10, 7.8580e-09]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([-2.5029e-09, 3.1025e-08, -5.5297e-09, 1.7462e-09, -1.0291e-07, + 2.0955e-09, 4.3656e-09, 6.5775e-08, -1.1642e-09, 2.7008e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 251.87, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4730 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.12 lr 0.00001000 +Epoch 416, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0335, ..., -0.0817, 0.1751, 0.0166], + [-0.0698, -0.1952, -0.0081, ..., 0.2265, -0.1988, 0.0428], + ..., + [-0.2329, 0.1543, 0.0232, ..., -0.1782, 0.1233, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2024, -0.3086], + [ 0.0362, -0.0595, -0.0182, ..., -0.2810, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[-9.8953e-10, -9.8953e-10, 0.0000e+00, ..., 4.6566e-10, + -1.3970e-09, 6.4028e-10], + [-3.1432e-09, -2.3283e-09, 0.0000e+00, ..., 1.3912e-08, + -6.4028e-09, 1.5309e-08], + [ 6.4028e-10, 3.4343e-09, 0.0000e+00, ..., -5.4133e-09, + 3.1432e-09, -0.0000e+00], + ..., + [ 2.9686e-09, -1.0477e-09, 0.0000e+00, ..., 6.6939e-09, + 3.5507e-09, 7.2760e-09], + [ 1.0477e-09, 8.7311e-10, 0.0000e+00, ..., 9.3132e-10, + 1.5716e-09, 1.5716e-09], + [-8.1491e-10, 1.3388e-09, 0.0000e+00, ..., 2.9628e-08, + 2.2701e-09, 2.1129e-08]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0429, 0.0142, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([-4.7148e-09, 5.6578e-08, -4.3074e-09, 8.5565e-09, -1.8429e-07, + 2.1537e-09, 1.6007e-08, 3.4459e-08, 7.9744e-09, 9.2434e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 251.78, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4300 re_mapping 0.0017 re_causal 0.0069 /// teacc 99.16 lr 0.00001000 +Epoch 417, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0335, ..., -0.0818, 0.1751, 0.0166], + [-0.0699, -0.1953, -0.0081, ..., 0.2265, -0.1988, 0.0428], + ..., + [-0.2329, 0.1543, 0.0232, ..., -0.1783, 0.1233, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2024, -0.3086], + [ 0.0362, -0.0595, -0.0182, ..., -0.2811, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[-6.7521e-09, 3.4925e-10, 0.0000e+00, ..., 1.0477e-08, + -1.9791e-09, 1.1991e-08], + [-3.0675e-08, -6.6939e-09, 0.0000e+00, ..., 1.4435e-08, + -3.2072e-08, 1.5716e-09], + [ 8.6147e-09, 2.0955e-09, 0.0000e+00, ..., -5.8208e-11, + 8.1491e-09, 7.9162e-09], + ..., + [ 3.3760e-09, 1.2806e-09, 0.0000e+00, ..., 2.1595e-08, + 3.0268e-09, 2.5495e-08], + [ 1.7928e-08, 3.7835e-09, 0.0000e+00, ..., 8.1491e-10, + 1.7346e-08, 8.7311e-09], + [ 2.0373e-09, 1.1642e-09, 0.0000e+00, ..., 1.3236e-07, + 2.4447e-09, 1.4820e-07]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0429, 0.0141, 0.0134, 0.0285, 0.0355, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([ 1.8685e-08, -2.7649e-08, 5.0000e-08, -5.0059e-09, -9.5041e-07, + 3.8417e-09, 1.9395e-07, 1.1543e-07, 6.1700e-08, 5.5367e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 252.23, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4168 re_mapping 0.0017 re_causal 0.0068 /// teacc 99.15 lr 0.00001000 +Epoch 418, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0518, 0.0575, -0.0335, ..., -0.0818, 0.1751, 0.0166], + [-0.0699, -0.1953, -0.0081, ..., 0.2265, -0.1988, 0.0428], + ..., + [-0.2329, 0.1543, 0.0232, ..., -0.1784, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2024, -0.3087], + [ 0.0362, -0.0595, -0.0182, ..., -0.2812, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[ 0.0000e+00, 5.8208e-11, 0.0000e+00, ..., 1.3388e-09, + 1.1642e-10, 4.0745e-10], + [ 5.8208e-11, 4.6566e-09, 0.0000e+00, ..., 2.1828e-08, + 4.1327e-09, 2.2876e-08], + [ 5.8208e-10, 2.6776e-09, 0.0000e+00, ..., -5.4715e-09, + 2.6193e-09, -8.0909e-09], + ..., + [ 7.5670e-10, -4.8312e-09, 0.0000e+00, ..., 2.2526e-08, + -3.2014e-09, 1.9441e-08], + [ 2.9104e-10, 3.4925e-10, 0.0000e+00, ..., 1.5716e-09, + 5.8208e-10, 9.3132e-10], + [ 5.8208e-11, 1.7462e-09, 0.0000e+00, ..., 1.0885e-08, + 8.1491e-10, 1.2515e-08]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0429, 0.0141, 0.0134, 0.0285, 0.0356, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([ 5.0059e-09, 7.9977e-08, 2.0955e-09, -8.5565e-09, -1.3132e-07, + 1.3795e-08, -5.0291e-08, 6.6531e-08, 7.9744e-09, 3.9872e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 252.10, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4586 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.15 lr 0.00001000 +Epoch 419, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0575, -0.0335, ..., -0.0818, 0.1751, 0.0165], + [-0.0700, -0.1954, -0.0081, ..., 0.2265, -0.1989, 0.0428], + ..., + [-0.2329, 0.1543, 0.0232, ..., -0.1784, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1709, -0.2024, -0.3087], + [ 0.0362, -0.0595, -0.0182, ..., -0.2812, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[ 3.4925e-10, 1.1642e-10, 0.0000e+00, ..., 8.1491e-10, + 1.1642e-10, 6.4028e-10], + [ 1.6298e-09, 1.2165e-08, 0.0000e+00, ..., 4.2492e-09, + 9.6043e-09, 1.3271e-08], + [-0.0000e+00, 5.5879e-09, 0.0000e+00, ..., -1.9907e-08, + 4.3074e-09, -7.9744e-09], + ..., + [ 7.5670e-10, -2.2526e-08, 0.0000e+00, ..., 3.1432e-09, + -1.5658e-08, -1.1583e-08], + [ 8.7311e-10, 5.8208e-10, 0.0000e+00, ..., 4.2492e-09, + 5.2387e-10, 3.0850e-09], + [ 1.9209e-09, 3.9581e-09, 0.0000e+00, ..., 2.3865e-09, + 3.2596e-09, 5.7044e-09]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0429, 0.0141, 0.0134, 0.0285, 0.0356, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([ 7.4506e-09, 5.6927e-08, 1.5018e-08, -4.9768e-08, -1.4552e-08, + -3.8184e-08, 8.2073e-09, -3.0093e-08, 2.5728e-08, 3.5448e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 252.34, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4377 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 +Epoch 420, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0575, -0.0335, ..., -0.0818, 0.1751, 0.0165], + [-0.0700, -0.1954, -0.0081, ..., 0.2265, -0.1989, 0.0428], + ..., + [-0.2330, 0.1543, 0.0232, ..., -0.1785, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1710, -0.2024, -0.3087], + [ 0.0362, -0.0595, -0.0182, ..., -0.2813, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.3388e-09, 0.0000e+00, 0.0000e+00, ..., 5.8790e-09, + 5.8208e-11, 1.8626e-09], + [ 1.9791e-09, 5.8208e-11, 0.0000e+00, ..., 2.5728e-08, + -8.6729e-09, 9.1386e-09], + [-1.7637e-08, 8.7311e-10, 0.0000e+00, ..., -8.1898e-08, + 9.3132e-10, -2.4680e-08], + ..., + [ 4.9477e-09, -7.3342e-09, 0.0000e+00, ..., 3.4925e-09, + 4.5402e-09, -2.6193e-09], + [ 8.1491e-09, 1.7462e-09, 0.0000e+00, ..., 2.8347e-08, + 2.1537e-09, 1.0012e-08], + [-1.7462e-10, 6.9849e-10, 0.0000e+00, ..., 7.5670e-09, + 4.6566e-10, 4.0745e-09]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0429, 0.0141, 0.0134, 0.0285, 0.0356, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([ 2.0431e-08, 8.6380e-08, -2.6636e-07, 4.5984e-09, -3.0850e-09, + 1.5192e-08, 1.2398e-08, 1.0768e-08, 1.0751e-07, 1.6822e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 419, time 252.40, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4081 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.18 lr 0.00001000 +Epoch 421, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0575, -0.0335, ..., -0.0819, 0.1751, 0.0165], + [-0.0701, -0.1954, -0.0081, ..., 0.2266, -0.1989, 0.0428], + ..., + [-0.2330, 0.1543, 0.0232, ..., -0.1786, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1710, -0.2024, -0.3087], + [ 0.0362, -0.0595, -0.0182, ..., -0.2814, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[-1.9616e-08, 3.3178e-09, 0.0000e+00, ..., -1.2282e-08, + 1.1642e-10, 6.9849e-09], + [-2.5029e-09, -1.9209e-09, 0.0000e+00, ..., 1.3388e-09, + -4.9477e-09, 1.0477e-09], + [ 7.1013e-09, 1.1642e-09, 0.0000e+00, ..., 2.2701e-09, + 1.0477e-09, -4.0745e-10], + ..., + [ 4.3656e-09, 2.0780e-08, 0.0000e+00, ..., 8.7311e-10, + 2.5611e-09, 4.2026e-08], + [ 1.7462e-09, 9.8953e-10, 0.0000e+00, ..., 1.0477e-09, + 1.6880e-09, 1.1642e-09], + [ 7.7416e-09, -2.5495e-08, 0.0000e+00, ..., 8.9640e-09, + 4.0745e-10, -5.1572e-08]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0429, 0.0141, 0.0134, 0.0285, 0.0356, -0.0184, 0.0283, 0.0289, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([-6.0594e-08, 4.6566e-10, 2.6368e-08, 1.0186e-08, 8.5565e-09, + -5.7044e-09, 1.3388e-09, 1.3772e-07, 1.0070e-08, -1.1642e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 252.40, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4127 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.17 lr 0.00001000 +Epoch 422, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0575, -0.0335, ..., -0.0819, 0.1751, 0.0165], + [-0.0701, -0.1955, -0.0081, ..., 0.2266, -0.1989, 0.0428], + ..., + [-0.2330, 0.1544, 0.0232, ..., -0.1787, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1710, -0.2024, -0.3087], + [ 0.0362, -0.0595, -0.0182, ..., -0.2814, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[-1.2049e-08, 1.1642e-10, 0.0000e+00, ..., 1.4552e-09, + -3.4925e-10, 1.7462e-10], + [-1.3970e-09, 6.1118e-09, 0.0000e+00, ..., 1.5134e-09, + 1.9791e-09, 7.6252e-09], + [ 1.2806e-09, 3.5274e-08, 0.0000e+00, ..., 1.2806e-09, + 2.4913e-08, 3.4750e-08], + ..., + [ 9.3132e-10, -5.1979e-08, 0.0000e+00, ..., 1.6880e-09, + -3.5507e-08, -4.9011e-08], + [-4.0745e-10, 8.7311e-10, 0.0000e+00, ..., 1.0477e-09, + 1.6298e-09, 6.4028e-10], + [ 1.1001e-08, 5.0641e-09, 0.0000e+00, ..., 1.2224e-09, + 4.3656e-09, 5.2387e-09]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0429, 0.0141, 0.0134, 0.0285, 0.0356, -0.0184, 0.0283, 0.0289, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([-4.8429e-08, 2.8173e-08, 1.1921e-07, 2.3108e-08, 1.3155e-08, + -6.5775e-09, -2.2701e-08, -1.4692e-07, 9.8953e-10, 6.5076e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 252.39, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4400 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 +Epoch 423, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0819, 0.1751, 0.0165], + [-0.0702, -0.1955, -0.0081, ..., 0.2266, -0.1989, 0.0428], + ..., + [-0.2330, 0.1544, 0.0232, ..., -0.1787, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1710, -0.2025, -0.3087], + [ 0.0362, -0.0595, -0.0182, ..., -0.2815, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[-1.0186e-08, 7.6252e-09, 0.0000e+00, ..., 4.0745e-10, + 4.6566e-09, 3.5507e-09], + [ 2.0722e-08, 7.9977e-08, 0.0000e+00, ..., 5.4715e-09, + 4.5809e-08, 4.5809e-08], + [ 8.6729e-09, 9.1211e-08, 0.0000e+00, ..., 3.4925e-10, + 5.9546e-08, 6.6822e-08], + ..., + [ 3.0210e-08, -2.9407e-07, 0.0000e+00, ..., 2.8522e-09, + -1.8068e-07, -1.6007e-07], + [-9.8895e-08, 4.9477e-09, 0.0000e+00, ..., 4.6566e-10, + 5.6461e-09, 3.4343e-09], + [ 4.2259e-08, 7.3749e-08, 0.0000e+00, ..., 1.3912e-08, + 4.5227e-08, 5.2678e-08]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0429, 0.0140, 0.0134, 0.0285, 0.0356, -0.0185, 0.0283, 0.0289, + -0.0170, 0.0007], device='cuda:0'), grad: tensor([-2.9104e-08, 4.1956e-07, 3.3551e-07, 5.9197e-08, -3.3295e-08, + 2.7299e-08, 1.0885e-08, -7.3435e-07, -5.8580e-07, 5.4389e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 252.26, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4444 re_mapping 0.0017 re_causal 0.0069 /// teacc 99.20 lr 0.00001000 +Epoch 424, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0820, 0.1751, 0.0164], + [-0.0703, -0.1956, -0.0081, ..., 0.2266, -0.1990, 0.0428], + ..., + [-0.2330, 0.1544, 0.0232, ..., -0.1788, 0.1234, 0.0653], + [ 0.0031, -0.0935, -0.0050, ..., -0.1710, -0.2025, -0.3088], + [ 0.0362, -0.0595, -0.0182, ..., -0.2816, -0.1964, 0.0216]], + device='cuda:0'), grad: tensor([[-7.6019e-08, 5.8208e-11, 0.0000e+00, ..., 3.8999e-09, + 1.7462e-10, 3.4343e-09], + [-1.3388e-09, -1.2224e-09, 0.0000e+00, ..., 2.0955e-08, + -4.4820e-09, 2.8696e-08], + [ 8.7311e-10, 1.9209e-09, 0.0000e+00, ..., -1.3679e-08, + 2.3283e-09, -5.9954e-09], + ..., + [ 3.3178e-09, -2.6193e-09, 0.0000e+00, ..., 2.1246e-08, + 2.0955e-09, 2.4273e-08], + [ 1.6706e-08, 9.3132e-10, 0.0000e+00, ..., 3.1432e-09, + 1.7462e-09, 4.4238e-09], + [ 9.2550e-09, 3.4925e-10, 0.0000e+00, ..., 2.7358e-08, + 3.4925e-09, 2.0838e-08]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0429, 0.0140, 0.0134, 0.0285, 0.0357, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([-3.5181e-07, 1.2247e-07, -2.1420e-08, 1.1764e-07, -3.6485e-07, + -8.5565e-08, 2.0384e-07, 1.2561e-07, 9.2725e-08, 1.6950e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 252.35, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4151 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.20 lr 0.00001000 +Epoch 425, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0820, 0.1751, 0.0164], + [-0.0703, -0.1957, -0.0081, ..., 0.2267, -0.1990, 0.0428], + ..., + [-0.2330, 0.1544, 0.0232, ..., -0.1788, 0.1235, 0.0654], + [ 0.0031, -0.0935, -0.0050, ..., -0.1710, -0.2025, -0.3088], + [ 0.0362, -0.0595, -0.0182, ..., -0.2817, -0.1965, 0.0216]], + device='cuda:0'), grad: tensor([[ 5.2387e-10, 5.8208e-11, 0.0000e+00, ..., 3.1432e-09, + 1.7462e-10, 1.8626e-09], + [-3.6438e-08, -3.1840e-08, 0.0000e+00, ..., 7.5670e-10, + -9.6683e-08, -2.9686e-08], + [ 8.1491e-10, 1.6298e-09, 0.0000e+00, ..., -2.6950e-08, + 4.2492e-09, -2.1246e-08], + ..., + [ 3.2480e-08, 2.5437e-08, 0.0000e+00, ..., 5.2387e-09, + 7.9977e-08, 2.9569e-08], + [-1.3388e-09, 4.1327e-09, 0.0000e+00, ..., 6.4028e-10, + 7.6834e-09, 3.4925e-09], + [ 2.7940e-09, 1.3388e-09, 0.0000e+00, ..., 1.2689e-08, + 2.2119e-09, 1.3562e-08]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0429, 0.0140, 0.0134, 0.0285, 0.0357, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 9.5461e-09, -1.3388e-07, -8.4634e-08, 2.5437e-08, 2.2119e-09, + 9.8953e-10, -2.2119e-09, 1.3621e-07, 3.6089e-09, 5.8906e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 252.29, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4320 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 +Epoch 426, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0820, 0.1751, 0.0164], + [-0.0704, -0.1958, -0.0081, ..., 0.2267, -0.1990, 0.0428], + ..., + [-0.2330, 0.1545, 0.0232, ..., -0.1789, 0.1235, 0.0654], + [ 0.0031, -0.0936, -0.0050, ..., -0.1710, -0.2025, -0.3088], + [ 0.0362, -0.0595, -0.0182, ..., -0.2817, -0.1965, 0.0216]], + device='cuda:0'), grad: tensor([[ 4.0745e-10, 2.3283e-10, 0.0000e+00, ..., 1.7462e-10, + 4.6566e-10, 4.6566e-10], + [-2.2992e-08, -2.1886e-08, 0.0000e+00, ..., 1.4319e-08, + -4.2550e-08, 1.8743e-08], + [ 3.4925e-09, 4.5984e-09, 0.0000e+00, ..., 1.2340e-08, + 8.1491e-09, 3.2596e-08], + ..., + [ 8.2073e-09, 7.6834e-09, 0.0000e+00, ..., 7.6834e-09, + 1.5076e-08, 2.1246e-08], + [ 1.0885e-08, 1.0012e-08, 0.0000e+00, ..., 9.3132e-10, + 1.9034e-08, 5.9372e-09], + [ 7.5670e-10, 8.7311e-10, 0.0000e+00, ..., 1.2154e-07, + 1.4552e-09, 2.6613e-07]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0429, 0.0140, 0.0135, 0.0285, 0.0357, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 3.4343e-09, -2.9046e-08, 8.1607e-08, -5.0815e-08, -6.9477e-07, + 3.6671e-08, -1.9209e-09, 7.3807e-08, 4.9593e-08, 5.4436e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 252.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4212 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.18 lr 0.00001000 +Epoch 427, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1429, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0821, 0.1751, 0.0164], + [-0.0704, -0.1959, -0.0081, ..., 0.2267, -0.1990, 0.0428], + ..., + [-0.2330, 0.1545, 0.0232, ..., -0.1790, 0.1235, 0.0654], + [ 0.0031, -0.0936, -0.0050, ..., -0.1710, -0.2026, -0.3088], + [ 0.0362, -0.0595, -0.0182, ..., -0.2818, -0.1965, 0.0216]], + device='cuda:0'), grad: tensor([[ 2.6776e-08, 5.8208e-11, 0.0000e+00, ..., 2.9104e-10, + 5.8208e-11, 1.7462e-09], + [ 1.4086e-08, 2.1537e-09, 0.0000e+00, ..., 1.4086e-08, + -3.4925e-09, 1.6880e-08], + [ 8.3237e-09, 2.3516e-08, 0.0000e+00, ..., -3.7078e-08, + 1.4785e-08, -6.6357e-09], + ..., + [ 8.7894e-09, -2.8871e-08, 0.0000e+00, ..., 1.8394e-08, + -1.5134e-08, -1.4727e-08], + [ 6.9849e-10, 5.8208e-10, 0.0000e+00, ..., 3.2596e-09, + 1.2224e-09, 3.0268e-09], + [-2.0373e-07, 1.4552e-09, 0.0000e+00, ..., 2.9104e-10, + 1.1642e-09, -1.0594e-08]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0429, 0.0140, 0.0134, 0.0285, 0.0357, -0.0184, 0.0283, 0.0290, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 1.6461e-07, 1.4435e-07, 2.9802e-08, 4.5076e-07, 1.2666e-07, + 2.7195e-07, -4.0163e-09, 1.4552e-08, 3.3120e-08, -1.2247e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 252.03, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4251 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 +Epoch 428, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0821, 0.1751, 0.0163], + [-0.0705, -0.1959, -0.0081, ..., 0.2268, -0.1991, 0.0428], + ..., + [-0.2331, 0.1546, 0.0232, ..., -0.1791, 0.1236, 0.0655], + [ 0.0031, -0.0936, -0.0050, ..., -0.1711, -0.2026, -0.3089], + [ 0.0362, -0.0595, -0.0182, ..., -0.2819, -0.1965, 0.0215]], + device='cuda:0'), grad: tensor([[-1.8859e-08, 1.1642e-09, 0.0000e+00, ..., 1.8044e-09, + 6.9849e-10, 5.6461e-09], + [ 6.9849e-09, 7.0431e-09, 0.0000e+00, ..., 2.5146e-08, + 4.8312e-09, 4.4005e-08], + [ 3.3760e-09, 2.5379e-08, 0.0000e+00, ..., -2.4505e-08, + 1.5367e-08, -6.9849e-10], + ..., + [ 1.3388e-08, -9.0804e-08, 0.0000e+00, ..., 1.4261e-08, + -5.0582e-08, -1.2224e-08], + [ 4.2492e-09, 4.4238e-09, 0.0000e+00, ..., 1.2224e-08, + 2.1537e-09, 1.8044e-08], + [-2.1013e-08, 1.3912e-08, 0.0000e+00, ..., 7.0839e-08, + 8.7894e-09, 5.5472e-08]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0357, -0.0184, 0.0283, 0.0291, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([-1.0693e-07, 2.9313e-07, 6.2515e-08, 1.9046e-07, -8.1630e-07, + 2.8522e-09, 8.7486e-08, -1.2829e-07, 7.7998e-08, 3.7346e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 252.13, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4317 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 +Epoch 429, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0519, 0.0574, -0.0335, ..., -0.0821, 0.1751, 0.0163], + [-0.0706, -0.1960, -0.0081, ..., 0.2268, -0.1991, 0.0428], + ..., + [-0.2331, 0.1546, 0.0232, ..., -0.1792, 0.1236, 0.0655], + [ 0.0031, -0.0936, -0.0050, ..., -0.1711, -0.2026, -0.3089], + [ 0.0363, -0.0595, -0.0182, ..., -0.2820, -0.1966, 0.0215]], + device='cuda:0'), grad: tensor([[-6.4028e-10, 3.4925e-10, 0.0000e+00, ..., 2.9104e-10, + 2.3283e-10, 2.9104e-10], + [ 2.5611e-09, 2.2841e-07, 0.0000e+00, ..., 3.8999e-09, + 1.9872e-07, 2.2340e-07], + [ 1.3388e-09, 2.5728e-08, 0.0000e+00, ..., 2.6193e-09, + 2.0780e-08, 2.2992e-08], + ..., + [ 1.8044e-09, -2.7381e-07, 0.0000e+00, ..., -3.6671e-09, + -2.3469e-07, -2.6124e-07], + [-1.0012e-08, 1.2224e-09, 0.0000e+00, ..., -2.1537e-09, + 2.3283e-09, 2.0373e-09], + [ 7.5670e-10, 6.0536e-09, 0.0000e+00, ..., 8.1491e-10, + 4.7730e-09, 5.0641e-09]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0284, 0.0357, -0.0184, 0.0283, 0.0291, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([-1.5716e-09, 6.1421e-07, 6.6997e-08, 6.1700e-09, 3.4750e-08, + 1.7870e-08, 2.1537e-09, -7.0501e-07, -2.9220e-08, 1.9558e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 251.70, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3760 re_mapping 0.0016 re_causal 0.0066 /// teacc 99.18 lr 0.00001000 +Epoch 430, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0519, 0.0573, -0.0335, ..., -0.0821, 0.1751, 0.0163], + [-0.0707, -0.1961, -0.0081, ..., 0.2268, -0.1991, 0.0428], + ..., + [-0.2331, 0.1546, 0.0232, ..., -0.1792, 0.1236, 0.0655], + [ 0.0031, -0.0936, -0.0050, ..., -0.1711, -0.2027, -0.3089], + [ 0.0363, -0.0596, -0.0182, ..., -0.2820, -0.1966, 0.0215]], + device='cuda:0'), grad: tensor([[ 2.9104e-10, 5.8208e-11, 0.0000e+00, ..., 1.1642e-10, + 5.8208e-11, 1.1642e-10], + [-9.8953e-10, -1.7462e-09, 0.0000e+00, ..., 1.1642e-10, + -3.3178e-09, -6.4028e-10], + [ 1.5716e-09, 3.4925e-10, 0.0000e+00, ..., 5.8208e-11, + 4.0745e-10, 2.9104e-10], + ..., + [ 1.3388e-09, 8.7311e-10, 0.0000e+00, ..., 0.0000e+00, + 1.5716e-09, 6.9849e-10], + [-6.9849e-10, 6.4028e-10, 0.0000e+00, ..., 1.1642e-10, + 1.0477e-09, 4.6566e-10], + [-3.3178e-09, 2.9104e-10, 0.0000e+00, ..., 1.1642e-10, + 4.6566e-10, -1.6880e-09]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0357, -0.0184, 0.0283, 0.0291, + -0.0170, 0.0006], device='cuda:0'), grad: tensor([ 2.6776e-09, 4.7730e-09, 1.4727e-08, 1.8277e-08, 1.8161e-08, + 5.7044e-09, -7.3342e-09, 7.4506e-09, -1.5483e-08, -3.3644e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 251.97, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4268 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 431, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0520, 0.0573, -0.0335, ..., -0.0822, 0.1751, 0.0162], + [-0.0707, -0.1962, -0.0081, ..., 0.2268, -0.1992, 0.0428], + ..., + [-0.2331, 0.1547, 0.0232, ..., -0.1793, 0.1237, 0.0656], + [ 0.0031, -0.0936, -0.0050, ..., -0.1711, -0.2027, -0.3090], + [ 0.0363, -0.0596, -0.0182, ..., -0.2821, -0.1966, 0.0215]], + device='cuda:0'), grad: tensor([[ 1.1642e-10, 2.9104e-10, 0.0000e+00, ..., 4.0745e-10, + 2.3283e-10, 5.8208e-10], + [ 2.3283e-10, 8.5565e-09, 0.0000e+00, ..., 9.8953e-09, + 6.8685e-09, 2.1595e-08], + [ 2.9104e-10, 5.7044e-09, 0.0000e+00, ..., -4.3714e-08, + 4.8312e-09, -6.2399e-08], + ..., + [ 5.8208e-10, -2.6310e-08, 0.0000e+00, ..., 8.7311e-10, + -2.0431e-08, -1.8626e-08], + [ 4.0745e-10, 5.8208e-10, 0.0000e+00, ..., 4.6566e-10, + 6.9849e-10, 1.1059e-09], + [-2.3283e-10, 6.5193e-09, 0.0000e+00, ..., 3.2247e-08, + 5.2387e-09, 5.2503e-08]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0357, -0.0184, 0.0283, 0.0291, + -0.0171, 0.0006], device='cuda:0'), grad: tensor([ 3.2014e-09, 7.2992e-08, -1.8603e-07, 1.4552e-08, 1.6822e-08, + -3.6671e-09, -4.0745e-10, -7.1246e-08, 6.1700e-09, 1.6415e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 252.18, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4400 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 +Epoch 432, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0520, 0.0573, -0.0335, ..., -0.0822, 0.1751, 0.0162], + [-0.0709, -0.1963, -0.0081, ..., 0.2269, -0.1992, 0.0428], + ..., + [-0.2331, 0.1547, 0.0232, ..., -0.1794, 0.1237, 0.0656], + [ 0.0031, -0.0937, -0.0050, ..., -0.1711, -0.2028, -0.3090], + [ 0.0363, -0.0596, -0.0182, ..., -0.2822, -0.1966, 0.0215]], + device='cuda:0'), grad: tensor([[ 2.3283e-10, 2.3283e-10, 0.0000e+00, ..., 1.1642e-10, + 2.3283e-10, 1.7462e-10], + [-2.5611e-09, 1.8044e-09, 0.0000e+00, ..., 4.6566e-10, + -3.2596e-09, 2.5029e-09], + [ 8.1491e-10, 5.8208e-09, 0.0000e+00, ..., 5.8208e-11, + 5.0059e-09, 4.0745e-09], + ..., + [ 1.8044e-09, -1.6240e-08, 0.0000e+00, ..., 2.9104e-10, + -1.1234e-08, -1.1118e-08], + [-5.0641e-09, 1.5134e-09, 0.0000e+00, ..., -0.0000e+00, + 2.2701e-09, 9.3132e-10], + [ 1.2224e-09, 3.2014e-09, 0.0000e+00, ..., 3.4925e-10, + 2.7358e-09, 2.5029e-09]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0358, -0.0185, 0.0283, 0.0291, + -0.0171, 0.0006], device='cuda:0'), grad: tensor([ 1.8044e-09, 6.6357e-09, 1.8859e-08, 2.5611e-09, 1.1933e-08, + 1.3621e-08, 3.1432e-09, -3.1374e-08, -1.6007e-08, 1.4901e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 251.95, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4114 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.17 lr 0.00001000 +Epoch 433, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0521, 0.0573, -0.0335, ..., -0.0822, 0.1751, 0.0162], + [-0.0710, -0.1964, -0.0081, ..., 0.2269, -0.1993, 0.0428], + ..., + [-0.2332, 0.1548, 0.0232, ..., -0.1794, 0.1237, 0.0657], + [ 0.0031, -0.0937, -0.0050, ..., -0.1711, -0.2029, -0.3091], + [ 0.0363, -0.0596, -0.0182, ..., -0.2822, -0.1967, 0.0215]], + device='cuda:0'), grad: tensor([[-1.4575e-07, 7.5670e-10, 0.0000e+00, ..., 1.1059e-09, + 1.2224e-09, 1.4552e-09], + [-1.8481e-08, -9.3132e-09, 0.0000e+00, ..., -1.5658e-08, + -6.5833e-08, -1.0914e-08], + [ 4.0396e-08, 5.0641e-09, 0.0000e+00, ..., -2.5320e-09, + 1.6124e-08, 2.9104e-11], + ..., + [ 1.7113e-08, -8.0036e-09, 0.0000e+00, ..., 1.3708e-08, + 3.1752e-08, 4.1327e-09], + [ 1.4494e-08, 4.2492e-09, 0.0000e+00, ..., 4.3656e-10, + 8.7311e-09, 3.4051e-09], + [ 5.4133e-08, 4.2201e-09, 0.0000e+00, ..., 5.5297e-10, + 3.7253e-09, -6.8394e-09]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0358, -0.0185, 0.0284, 0.0292, + -0.0171, 0.0006], device='cuda:0'), grad: tensor([-5.9232e-07, -4.6130e-08, 1.5437e-07, 3.7748e-08, 5.5326e-08, + 2.3661e-08, 8.2771e-08, 5.1368e-08, 4.9826e-08, 1.9919e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 252.29, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4171 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.16 lr 0.00001000 +Epoch 434, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0522, 0.0573, -0.0335, ..., -0.0822, 0.1751, 0.0162], + [-0.0711, -0.1966, -0.0081, ..., 0.2269, -0.1993, 0.0428], + ..., + [-0.2332, 0.1548, 0.0232, ..., -0.1795, 0.1237, 0.0657], + [ 0.0030, -0.0938, -0.0050, ..., -0.1711, -0.2031, -0.3091], + [ 0.0363, -0.0596, -0.0182, ..., -0.2823, -0.1967, 0.0215]], + device='cuda:0'), grad: tensor([[-2.0140e-08, 2.3283e-09, 0.0000e+00, ..., -2.2992e-09, + 1.0477e-09, 3.5507e-09], + [-2.1973e-08, -1.6589e-08, 0.0000e+00, ..., -4.4529e-09, + -3.1840e-08, -3.3760e-09], + [ 1.3446e-08, 3.7835e-09, 0.0000e+00, ..., 2.1828e-09, + 5.1223e-09, 2.5320e-09], + ..., + [ 1.0623e-08, 3.2538e-08, 0.0000e+00, ..., 2.9686e-09, + 1.4727e-08, 4.7585e-08], + [ 6.9558e-09, 4.3947e-09, 0.0000e+00, ..., 1.6880e-09, + 7.3924e-09, 2.3283e-09], + [ 1.0332e-08, -3.9494e-08, 0.0000e+00, ..., 3.2305e-09, + 2.0664e-09, -7.3924e-08]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0358, -0.0185, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([-5.4832e-08, -6.6822e-08, 5.2736e-08, 1.2689e-08, 8.0618e-08, + -8.4983e-09, 7.0431e-09, 2.0454e-07, 2.9366e-08, -2.3609e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 252.17, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4030 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.17 lr 0.00001000 +Epoch 435, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0522, 0.0573, -0.0335, ..., -0.0823, 0.1751, 0.0162], + [-0.0712, -0.1966, -0.0081, ..., 0.2269, -0.1994, 0.0428], + ..., + [-0.2332, 0.1549, 0.0232, ..., -0.1796, 0.1238, 0.0658], + [ 0.0030, -0.0938, -0.0050, ..., -0.1712, -0.2031, -0.3092], + [ 0.0363, -0.0596, -0.0182, ..., -0.2823, -0.1967, 0.0215]], + device='cuda:0'), grad: tensor([[ 8.7311e-11, 5.8208e-11, 0.0000e+00, ..., 6.6939e-10, + 1.1642e-10, 5.2387e-10], + [-3.1141e-09, -1.7462e-10, 0.0000e+00, ..., 2.3807e-08, + -2.7358e-09, 1.5745e-08], + [ 6.9849e-10, 2.9104e-09, 0.0000e+00, ..., -4.4121e-08, + 3.8999e-09, -2.3778e-08], + ..., + [ 1.8335e-09, -5.6752e-09, 0.0000e+00, ..., 2.0373e-09, + -4.6275e-09, -4.0454e-09], + [ 9.6043e-10, 8.1491e-10, 0.0000e+00, ..., 1.5978e-08, + 1.7171e-09, 1.0215e-08], + [ 3.4925e-10, 1.1933e-09, 0.0000e+00, ..., 2.9977e-09, + 1.4843e-09, 7.2760e-09]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0358, -0.0185, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([ 2.5611e-09, 6.3679e-08, -1.1298e-07, -1.7462e-10, -8.9349e-09, + 1.9500e-09, 3.5798e-09, -4.4238e-09, 4.8894e-08, 2.0198e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 252.26, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4279 re_mapping 0.0016 re_causal 0.0070 /// teacc 99.17 lr 0.00001000 +Epoch 436, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0522, 0.0573, -0.0335, ..., -0.0823, 0.1752, 0.0161], + [-0.0713, -0.1967, -0.0081, ..., 0.2270, -0.1994, 0.0428], + ..., + [-0.2332, 0.1549, 0.0232, ..., -0.1797, 0.1238, 0.0658], + [ 0.0030, -0.0938, -0.0050, ..., -0.1712, -0.2032, -0.3092], + [ 0.0363, -0.0596, -0.0182, ..., -0.2824, -0.1967, 0.0215]], + device='cuda:0'), grad: tensor([[ 2.7067e-09, 2.6484e-09, 0.0000e+00, ..., 2.1246e-09, + 5.0932e-09, 1.4843e-09], + [-8.9989e-08, -8.3353e-08, 0.0000e+00, ..., 2.5611e-09, + -1.8196e-07, -2.2410e-08], + [ 6.6648e-09, 2.3516e-08, 0.0000e+00, ..., -1.7550e-08, + 2.9424e-08, 3.7544e-09], + ..., + [ 1.7724e-08, -3.2654e-08, 0.0000e+00, ..., 3.9581e-09, + -2.2963e-08, -2.7823e-08], + [ 2.8551e-08, 5.3114e-08, 0.0000e+00, ..., 6.0827e-09, + 1.0966e-07, 2.1129e-08], + [ 1.0565e-08, 7.0431e-09, 0.0000e+00, ..., 4.6566e-10, + 8.5565e-09, 5.1223e-09]], device='cuda:0') +Epoch 436, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0358, -0.0185, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([ 2.0344e-08, -3.8440e-07, 4.0745e-08, 1.7812e-08, 4.9506e-08, + 2.1624e-08, 6.3912e-08, -2.8434e-08, 1.2526e-07, 8.0443e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 252.19, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3931 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.15 lr 0.00001000 +Epoch 437, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1430, -0.1751, -0.1698], + [ 0.0523, 0.0573, -0.0335, ..., -0.0823, 0.1752, 0.0161], + [-0.0714, -0.1968, -0.0081, ..., 0.2270, -0.1995, 0.0428], + ..., + [-0.2333, 0.1549, 0.0232, ..., -0.1798, 0.1238, 0.0658], + [ 0.0030, -0.0938, -0.0050, ..., -0.1712, -0.2032, -0.3092], + [ 0.0363, -0.0596, -0.0182, ..., -0.2824, -0.1968, 0.0215]], + device='cuda:0'), grad: tensor([[-6.4028e-10, 5.8208e-10, 0.0000e+00, ..., 5.6170e-09, + 1.7171e-09, 2.7940e-09], + [-3.2596e-09, 9.4296e-09, 0.0000e+00, ..., 1.2957e-07, + 4.4791e-08, 6.0361e-08], + [ 6.1118e-10, 2.9919e-08, 0.0000e+00, ..., -2.4075e-07, + -4.4238e-08, -8.6147e-08], + ..., + [ 1.9209e-09, -6.3214e-08, 0.0000e+00, ..., 2.0606e-08, + -5.0728e-08, -2.7940e-08], + [-1.3679e-09, 2.0082e-09, 0.0000e+00, ..., 1.6269e-08, + 7.1595e-09, 8.3237e-09], + [ 4.0745e-10, 2.9104e-09, 0.0000e+00, ..., 3.1432e-09, + 2.6484e-09, 1.8626e-09]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0429, 0.0139, 0.0134, 0.0285, 0.0358, -0.0186, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([ 1.2573e-08, 2.9663e-07, -4.5914e-07, 9.3831e-08, 9.8196e-08, + 8.9058e-09, 1.5541e-08, -8.8941e-08, 2.9511e-08, 1.2777e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 252.01, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4138 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.14 lr 0.00001000 +Epoch 438, weight, value: tensor([[ 0.0890, -0.1872, -0.1459, ..., -0.1431, -0.1751, -0.1699], + [ 0.0523, 0.0573, -0.0335, ..., -0.0823, 0.1752, 0.0161], + [-0.0714, -0.1969, -0.0081, ..., 0.2270, -0.1995, 0.0429], + ..., + [-0.2333, 0.1549, 0.0232, ..., -0.1799, 0.1238, 0.0658], + [ 0.0030, -0.0939, -0.0050, ..., -0.1712, -0.2032, -0.3093], + [ 0.0363, -0.0596, -0.0182, ..., -0.2825, -0.1968, 0.0215]], + device='cuda:0'), grad: tensor([[-4.9477e-10, 8.7311e-10, 0.0000e+00, ..., 4.6566e-10, + 1.3097e-09, 8.7311e-10], + [-3.8621e-08, -3.8854e-08, 0.0000e+00, ..., 5.5006e-09, + -6.6299e-08, -2.1420e-08], + [ 3.1141e-09, 5.5297e-09, 0.0000e+00, ..., -8.1200e-09, + 7.0140e-09, -3.3178e-09], + ..., + [ 9.9535e-09, 9.8953e-10, 0.0000e+00, ..., 1.8044e-09, + 1.0157e-08, 2.4738e-09], + [ 2.0955e-09, 2.7649e-09, 0.0000e+00, ..., 7.5670e-10, + 4.8312e-09, 2.5611e-09], + [ 4.8312e-09, 6.7521e-09, 0.0000e+00, ..., 7.8580e-10, + 8.3819e-09, 5.1514e-09]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0429, 0.0138, 0.0134, 0.0285, 0.0358, -0.0187, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([-2.2992e-09, -1.4948e-07, -2.6484e-09, 7.9162e-09, 7.9512e-08, + 3.6671e-09, 8.8185e-09, 2.7270e-08, 1.1903e-08, 3.3324e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 252.00, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4499 re_mapping 0.0016 re_causal 0.0071 /// teacc 99.18 lr 0.00001000 +Epoch 439, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1431, -0.1751, -0.1699], + [ 0.0523, 0.0573, -0.0335, ..., -0.0824, 0.1752, 0.0161], + [-0.0715, -0.1970, -0.0081, ..., 0.2271, -0.1995, 0.0429], + ..., + [-0.2334, 0.1550, 0.0232, ..., -0.1800, 0.1238, 0.0659], + [ 0.0030, -0.0939, -0.0050, ..., -0.1713, -0.2033, -0.3093], + [ 0.0363, -0.0596, -0.0182, ..., -0.2826, -0.1968, 0.0215]], + device='cuda:0'), grad: tensor([[ 4.6275e-09, 2.3865e-09, 0.0000e+00, ..., 8.7311e-11, + 3.6089e-09, 1.2806e-09], + [-4.6013e-08, -2.0809e-08, 0.0000e+00, ..., -4.3656e-10, + -7.1537e-08, -1.0099e-08], + [ 9.0804e-09, 4.5809e-08, 0.0000e+00, ..., 4.0745e-10, + 4.4820e-08, 3.8126e-08], + ..., + [ 9.4296e-09, -5.8295e-08, 0.0000e+00, ..., 1.8917e-09, + -2.6979e-08, -4.3685e-08], + [ 7.1886e-09, 1.7928e-08, 0.0000e+00, ..., 1.4552e-10, + 2.9628e-08, 8.9931e-09], + [ 8.0618e-09, 4.9185e-09, 0.0000e+00, ..., 3.6089e-09, + 6.1409e-09, 1.1030e-08]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0429, 0.0138, 0.0134, 0.0285, 0.0358, -0.0187, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([ 2.1100e-08, -1.0803e-07, 1.4401e-07, 1.9529e-08, -7.3633e-09, + 4.4529e-09, 9.1968e-09, -1.1630e-07, -1.0070e-08, 6.4494e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 251.90, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3947 re_mapping 0.0016 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 440, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1431, -0.1751, -0.1699], + [ 0.0523, 0.0573, -0.0335, ..., -0.0824, 0.1752, 0.0161], + [-0.0715, -0.1971, -0.0081, ..., 0.2272, -0.1996, 0.0429], + ..., + [-0.2334, 0.1550, 0.0232, ..., -0.1801, 0.1239, 0.0659], + [ 0.0030, -0.0939, -0.0050, ..., -0.1713, -0.2033, -0.3094], + [ 0.0363, -0.0596, -0.0182, ..., -0.2827, -0.1969, 0.0215]], + device='cuda:0'), grad: tensor([[-4.0163e-09, 2.1537e-09, 0.0000e+00, ..., 6.4028e-10, + 1.4552e-10, 4.4529e-09], + [ 2.9104e-10, 4.1939e-08, 0.0000e+00, ..., 1.9878e-08, + 3.0152e-08, 6.6531e-08], + [ 6.6939e-10, 2.6223e-08, 0.0000e+00, ..., 4.3656e-09, + 1.9558e-08, 2.9017e-08], + ..., + [ 4.0745e-10, -5.4424e-08, 0.0000e+00, ..., 1.2759e-07, + -6.2690e-08, 1.7986e-07], + [ 3.9290e-09, 1.3388e-09, 0.0000e+00, ..., 1.3097e-09, + 9.6043e-10, 2.4447e-09], + [ 2.7940e-09, -5.1514e-09, 0.0000e+00, ..., 1.8114e-07, + 9.7789e-09, 2.5774e-07]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0429, 0.0138, 0.0134, 0.0285, 0.0358, -0.0187, 0.0284, 0.0292, + -0.0172, 0.0006], device='cuda:0'), grad: tensor([-1.9791e-09, 2.1467e-07, 1.0349e-07, 1.5163e-08, -1.3886e-06, + -2.3720e-08, 1.8306e-08, 4.1956e-07, 2.3778e-08, 6.3051e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 252.35, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4184 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.22 lr 0.00001000 +Epoch 441, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1431, -0.1751, -0.1699], + [ 0.0523, 0.0573, -0.0335, ..., -0.0825, 0.1752, 0.0160], + [-0.0716, -0.1973, -0.0080, ..., 0.2272, -0.1996, 0.0429], + ..., + [-0.2334, 0.1551, 0.0232, ..., -0.1803, 0.1239, 0.0660], + [ 0.0030, -0.0939, -0.0050, ..., -0.1713, -0.2033, -0.3094], + [ 0.0363, -0.0597, -0.0182, ..., -0.2828, -0.1969, 0.0215]], + device='cuda:0'), grad: tensor([[-2.9133e-08, 1.4552e-10, 0.0000e+00, ..., 1.3679e-09, + -6.6648e-09, 1.3097e-09], + [-1.4843e-09, 4.0454e-09, 0.0000e+00, ..., 2.7328e-08, + -2.3283e-10, 2.5728e-08], + [ 1.4261e-09, 6.0245e-09, 0.0000e+00, ..., -1.2049e-08, + 5.0059e-09, -4.6566e-09], + ..., + [ 2.0955e-09, -1.8888e-08, 0.0000e+00, ..., 7.2760e-09, + -1.3009e-08, -6.6939e-09], + [ 1.8044e-09, 5.5297e-10, 0.0000e+00, ..., 8.4401e-10, + 1.1933e-09, 1.1642e-09], + [ 9.1095e-09, 1.1350e-09, 0.0000e+00, ..., 5.5588e-09, + 4.4238e-09, 1.5716e-09]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0429, 0.0137, 0.0135, 0.0285, 0.0358, -0.0187, 0.0285, 0.0293, + -0.0173, 0.0006], device='cuda:0'), grad: tensor([-2.2841e-07, 1.0803e-07, 1.5320e-07, -1.8871e-07, -7.4913e-08, + 7.3458e-08, 5.9139e-08, -2.8522e-09, 2.0955e-08, 9.2434e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 251.84, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4031 re_mapping 0.0016 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +Epoch 442, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1431, -0.1751, -0.1699], + [ 0.0524, 0.0572, -0.0335, ..., -0.0826, 0.1752, 0.0159], + [-0.0717, -0.1974, -0.0080, ..., 0.2273, -0.1997, 0.0429], + ..., + [-0.2335, 0.1552, 0.0232, ..., -0.1804, 0.1240, 0.0660], + [ 0.0030, -0.0939, -0.0050, ..., -0.1714, -0.2033, -0.3095], + [ 0.0363, -0.0597, -0.0182, ..., -0.2829, -0.1969, 0.0215]], + device='cuda:0'), grad: tensor([[-3.9086e-08, 1.1642e-10, 0.0000e+00, ..., 9.5170e-09, + 1.4552e-10, 2.4447e-09], + [-5.3551e-09, 1.8044e-09, 0.0000e+00, ..., 1.3300e-08, + -8.0909e-09, 5.5297e-09], + [-1.7200e-08, 1.8947e-08, 0.0000e+00, ..., -6.2981e-08, + 1.2689e-08, -2.1420e-08], + ..., + [ 3.7544e-09, -2.9628e-08, 0.0000e+00, ..., 2.1828e-09, + -1.5309e-08, -3.1636e-08], + [ 1.6735e-08, 1.1642e-09, 0.0000e+00, ..., 5.9023e-08, + 2.1537e-09, 3.7631e-08], + [ 3.5594e-08, 4.2492e-09, 0.0000e+00, ..., 6.6939e-10, + 3.3178e-09, 2.7067e-09]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0429, 0.0137, 0.0135, 0.0285, 0.0358, -0.0188, 0.0285, 0.0293, + -0.0173, 0.0006], device='cuda:0'), grad: tensor([-1.4785e-07, 3.9581e-08, -1.1467e-07, 1.4319e-08, 4.2870e-08, + 8.5565e-09, -6.9616e-08, -7.7300e-08, 1.6170e-07, 1.5856e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 251.70, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4098 re_mapping 0.0016 re_causal 0.0066 /// teacc 99.21 lr 0.00001000 +Epoch 443, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0524, 0.0572, -0.0335, ..., -0.0826, 0.1752, 0.0158], + [-0.0718, -0.1975, -0.0080, ..., 0.2273, -0.1997, 0.0429], + ..., + [-0.2335, 0.1553, 0.0232, ..., -0.1805, 0.1241, 0.0661], + [ 0.0030, -0.0939, -0.0050, ..., -0.1714, -0.2034, -0.3095], + [ 0.0363, -0.0597, -0.0182, ..., -0.2829, -0.1969, 0.0215]], + device='cuda:0'), grad: tensor([[ 2.8813e-09, 5.8208e-10, 0.0000e+00, ..., 5.5297e-10, + 4.0745e-10, 1.4552e-10], + [-4.3452e-08, -5.9285e-08, 0.0000e+00, ..., 2.0373e-10, + -7.0781e-08, -2.7678e-08], + [ 3.8999e-09, 4.3365e-09, 0.0000e+00, ..., 8.7311e-11, + 5.3551e-09, 2.2119e-09], + ..., + [ 2.4040e-08, 1.7957e-08, 0.0000e+00, ..., 1.1642e-10, + 2.1275e-08, 7.6834e-09], + [ 6.6939e-10, 3.2800e-08, 0.0000e+00, ..., 1.1642e-10, + 4.1939e-08, 1.6764e-08], + [ 9.0513e-09, 2.1537e-09, 0.0000e+00, ..., 3.2014e-10, + 1.7462e-09, 9.6043e-10]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0429, 0.0137, 0.0135, 0.0285, 0.0358, -0.0189, 0.0285, 0.0294, + -0.0173, 0.0006], device='cuda:0'), grad: tensor([ 2.8987e-08, -1.0722e-07, 2.4942e-08, -2.0955e-09, 6.4611e-09, + 2.9773e-08, 5.2096e-09, 1.1828e-07, -1.5728e-07, 7.7940e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 442, time 252.34, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4127 re_mapping 0.0016 re_causal 0.0069 /// teacc 99.18 lr 0.00001000 +Epoch 444, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0524, 0.0572, -0.0335, ..., -0.0826, 0.1752, 0.0158], + [-0.0718, -0.1977, -0.0080, ..., 0.2274, -0.1998, 0.0430], + ..., + [-0.2335, 0.1553, 0.0232, ..., -0.1808, 0.1241, 0.0662], + [ 0.0030, -0.0940, -0.0050, ..., -0.1714, -0.2034, -0.3096], + [ 0.0364, -0.0597, -0.0182, ..., -0.2830, -0.1970, 0.0215]], + device='cuda:0'), grad: tensor([[-5.8790e-09, 1.1642e-10, 0.0000e+00, ..., 6.4028e-10, + 1.1642e-10, 4.9477e-10], + [ 4.9477e-10, 5.1514e-08, 0.0000e+00, ..., 2.0664e-09, + 5.1805e-08, 5.0321e-08], + [ 6.1118e-10, 1.2689e-08, 0.0000e+00, ..., -1.0768e-08, + 9.2259e-09, -8.7311e-11], + ..., + [ 1.0768e-09, -7.2294e-08, 0.0000e+00, ..., 3.2887e-09, + -6.6939e-08, -6.3446e-08], + [-4.3656e-10, 8.7311e-10, 0.0000e+00, ..., 1.2224e-09, + 1.1642e-09, 1.7753e-09], + [ 2.5611e-09, 2.3574e-09, 0.0000e+00, ..., 3.7835e-10, + 1.9791e-09, 2.4447e-09]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0429, 0.0136, 0.0135, 0.0285, 0.0357, -0.0189, 0.0285, 0.0294, + -0.0173, 0.0007], device='cuda:0'), grad: tensor([-2.5786e-08, 1.5064e-07, 1.7142e-08, -2.9511e-08, 1.7200e-08, + 2.6892e-08, 4.2201e-09, -1.7881e-07, 7.3342e-09, 2.1420e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 252.35, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4045 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.15 lr 0.00001000 +Epoch 445, weight, value: tensor([[ 0.0891, -0.1872, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0524, 0.0572, -0.0335, ..., -0.0827, 0.1752, 0.0158], + [-0.0719, -0.1978, -0.0080, ..., 0.2275, -0.1999, 0.0430], + ..., + [-0.2336, 0.1554, 0.0232, ..., -0.1809, 0.1241, 0.0662], + [ 0.0030, -0.0940, -0.0050, ..., -0.1715, -0.2035, -0.3097], + [ 0.0363, -0.0597, -0.0182, ..., -0.2831, -0.1970, 0.0215]], + device='cuda:0'), grad: tensor([[-3.6764e-07, -2.1129e-08, 0.0000e+00, ..., -2.1013e-07, + -2.6484e-08, 5.2387e-10], + [ 1.6869e-07, 7.8580e-10, 0.0000e+00, ..., 1.0582e-07, + -1.0536e-08, -4.1327e-09], + [ 2.6484e-08, 8.4692e-09, 0.0000e+00, ..., -8.4692e-09, + 1.2864e-08, -9.5461e-09], + ..., + [ 4.3365e-09, -3.5594e-08, 0.0000e+00, ..., 2.9104e-09, + -2.4942e-08, -3.0617e-08], + [-5.0350e-09, 2.0082e-09, 0.0000e+00, ..., 3.3469e-09, + 4.3074e-09, 1.7171e-09], + [ 9.4296e-09, 3.8417e-09, 0.0000e+00, ..., 2.9104e-09, + 3.3469e-09, 2.5611e-09]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0429, 0.0136, 0.0135, 0.0285, 0.0357, -0.0190, 0.0285, 0.0294, + -0.0173, 0.0006], device='cuda:0'), grad: tensor([-1.4082e-06, 6.6031e-07, 4.5839e-08, 8.9814e-08, 9.2899e-08, + -2.2672e-08, 5.7789e-07, -6.0420e-08, -6.1991e-09, 4.2695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 252.22, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3901 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.20 lr 0.00001000 +Epoch 446, weight, value: tensor([[ 0.0891, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0525, 0.0571, -0.0335, ..., -0.0828, 0.1752, 0.0157], + [-0.0720, -0.1980, -0.0080, ..., 0.2276, -0.1999, 0.0430], + ..., + [-0.2336, 0.1555, 0.0232, ..., -0.1810, 0.1242, 0.0663], + [ 0.0030, -0.0940, -0.0050, ..., -0.1715, -0.2035, -0.3097], + [ 0.0364, -0.0597, -0.0182, ..., -0.2832, -0.1970, 0.0215]], + device='cuda:0'), grad: tensor([[ 5.2387e-10, 5.8208e-10, 0.0000e+00, ..., 6.6939e-10, + 1.6007e-09, 9.8953e-10], + [-7.4331e-08, -8.1200e-09, 0.0000e+00, ..., 4.6566e-10, + -8.4983e-08, -3.7719e-08], + [ 1.2777e-08, 4.7672e-08, 0.0000e+00, ..., -2.6776e-09, + 4.6712e-08, 3.9086e-08], + ..., + [ 2.8260e-08, -6.5134e-08, 0.0000e+00, ..., 1.7753e-09, + -1.6036e-08, -3.6147e-08], + [ 6.0827e-09, 2.7649e-09, 0.0000e+00, ..., 7.5670e-10, + 9.2550e-09, 4.1327e-09], + [ 7.6543e-09, 1.0419e-08, 0.0000e+00, ..., 1.1642e-10, + 1.4115e-08, 1.0768e-08]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0429, 0.0136, 0.0135, 0.0285, 0.0357, -0.0191, 0.0285, 0.0294, + -0.0173, 0.0006], device='cuda:0'), grad: tensor([ 7.5670e-09, -2.5565e-07, 1.9302e-07, -7.7765e-08, 4.1531e-08, + 8.7311e-08, 4.4820e-09, -8.9989e-08, 3.7369e-08, 6.1933e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 252.40, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4066 re_mapping 0.0016 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 447, weight, value: tensor([[ 0.0892, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0526, 0.0571, -0.0333, ..., -0.0828, 0.1752, 0.0157], + [-0.0721, -0.1982, -0.0080, ..., 0.2276, -0.2000, 0.0430], + ..., + [-0.2337, 0.1556, 0.0231, ..., -0.1811, 0.1242, 0.0664], + [ 0.0030, -0.0940, -0.0050, ..., -0.1715, -0.2036, -0.3098], + [ 0.0364, -0.0597, -0.0182, ..., -0.2834, -0.1971, 0.0215]], + device='cuda:0'), grad: tensor([[ 1.4552e-10, 2.0373e-10, 0.0000e+00, ..., 6.9849e-10, + 5.2387e-10, 7.5670e-10], + [-3.4343e-09, -1.7608e-08, 0.0000e+00, ..., -4.1327e-09, + -1.0419e-07, -3.4430e-08], + [ 3.0850e-09, 1.2835e-08, 0.0000e+00, ..., -1.3330e-08, + 3.7864e-08, 4.9185e-09], + ..., + [ 2.3865e-09, -7.2760e-09, 0.0000e+00, ..., 5.5297e-09, + 3.0675e-08, 5.6170e-09], + [-4.7730e-09, 1.5134e-09, 0.0000e+00, ..., 3.2014e-10, + 4.1327e-09, 3.2014e-09], + [ 9.0222e-10, 3.0559e-09, 0.0000e+00, ..., 5.8208e-10, + 4.1327e-09, 3.0559e-09]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0428, 0.0136, 0.0135, 0.0285, 0.0357, -0.0192, 0.0285, 0.0294, + -0.0174, 0.0006], device='cuda:0'), grad: tensor([ 4.4529e-09, -1.9942e-07, 5.1659e-08, 6.1991e-09, 6.4261e-08, + 2.0693e-08, 5.7626e-09, 6.8161e-08, -1.9063e-08, 1.7695e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 252.77, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4311 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.17 lr 0.00001000 +Epoch 448, weight, value: tensor([[ 0.0892, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0526, 0.0571, -0.0333, ..., -0.0829, 0.1753, 0.0156], + [-0.0723, -0.1984, -0.0080, ..., 0.2277, -0.2001, 0.0430], + ..., + [-0.2338, 0.1556, 0.0231, ..., -0.1812, 0.1243, 0.0664], + [ 0.0030, -0.0941, -0.0050, ..., -0.1715, -0.2037, -0.3098], + [ 0.0364, -0.0597, -0.0182, ..., -0.2835, -0.1972, 0.0215]], + device='cuda:0'), grad: tensor([[ 6.4028e-10, 6.1118e-10, 0.0000e+00, ..., 3.4925e-10, + 8.7311e-10, 8.4401e-10], + [-8.1374e-08, -1.1449e-07, 0.0000e+00, ..., -2.3545e-08, + -2.0803e-07, -1.6158e-07], + [ 2.3196e-08, 3.1112e-08, 0.0000e+00, ..., 3.5507e-09, + 5.0612e-08, 3.8097e-08], + ..., + [ 3.8155e-08, 4.0105e-08, 0.0000e+00, ..., 1.1467e-08, + 7.6252e-08, 6.0711e-08], + [-2.7998e-08, -4.1618e-09, 0.0000e+00, ..., 4.9477e-10, + 2.0955e-09, 1.6298e-09], + [ 1.7171e-09, 6.6648e-09, 0.0000e+00, ..., 3.4925e-09, + 1.0477e-08, 6.4028e-09]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0428, 0.0135, 0.0135, 0.0286, 0.0358, -0.0192, 0.0285, 0.0294, + -0.0174, 0.0006], device='cuda:0'), grad: tensor([ 4.5402e-09, -4.1793e-07, 1.2922e-07, 8.1200e-09, 1.6263e-07, + 4.8691e-08, 6.1991e-09, 1.8964e-07, -1.3725e-07, 2.1304e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 252.21, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3865 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 449, weight, value: tensor([[ 0.0892, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0527, 0.0571, -0.0333, ..., -0.0830, 0.1753, 0.0156], + [-0.0724, -0.1985, -0.0080, ..., 0.2277, -0.2002, 0.0430], + ..., + [-0.2338, 0.1557, 0.0231, ..., -0.1814, 0.1243, 0.0664], + [ 0.0030, -0.0941, -0.0050, ..., -0.1716, -0.2037, -0.3099], + [ 0.0364, -0.0598, -0.0182, ..., -0.2837, -0.1972, 0.0215]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 8.7311e-11, 0.0000e+00, ..., 5.8208e-11, + 1.1642e-10, 8.7311e-11], + [-2.1246e-09, 7.8580e-10, 0.0000e+00, ..., 1.0186e-09, + -3.2596e-09, 1.2806e-09], + [ 6.4028e-10, 2.5320e-09, 0.0000e+00, ..., -1.2806e-09, + 2.4738e-09, 1.2806e-09], + ..., + [ 1.7462e-09, -6.5484e-09, 0.0000e+00, ..., 6.9849e-10, + -2.4447e-09, -4.7439e-09], + [ 1.1642e-10, 6.9849e-10, 0.0000e+00, ..., 1.4552e-10, + 1.2806e-09, 6.6939e-10], + [ 4.0745e-10, 7.5670e-10, 0.0000e+00, ..., 1.7753e-09, + 8.1491e-10, 2.1537e-09]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0428, 0.0135, 0.0135, 0.0286, 0.0358, -0.0193, 0.0285, 0.0294, + -0.0174, 0.0006], device='cuda:0'), grad: tensor([ 6.4028e-10, 3.5798e-09, 5.8208e-09, -6.9849e-10, -1.9209e-09, + 4.7439e-09, 2.0373e-09, -9.0513e-09, 2.6193e-10, 1.0012e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 252.22, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4047 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.21 lr 0.00001000 +Epoch 450, weight, value: tensor([[ 0.0892, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0527, 0.0571, -0.0333, ..., -0.0830, 0.1753, 0.0156], + [-0.0725, -0.1987, -0.0080, ..., 0.2277, -0.2002, 0.0430], + ..., + [-0.2339, 0.1557, 0.0231, ..., -0.1814, 0.1244, 0.0665], + [ 0.0030, -0.0941, -0.0050, ..., -0.1716, -0.2038, -0.3100], + [ 0.0364, -0.0598, -0.0182, ..., -0.2837, -0.1972, 0.0215]], + device='cuda:0'), grad: tensor([[ 3.2014e-10, 5.8208e-11, 0.0000e+00, ..., 8.7311e-11, + 5.8208e-11, 1.7462e-10], + [-1.4552e-10, -1.7462e-10, 0.0000e+00, ..., 1.2515e-09, + -2.5902e-09, 2.2410e-09], + [ 1.0186e-09, 1.7171e-09, 0.0000e+00, ..., 8.7311e-11, + 1.3679e-09, 1.6589e-09], + ..., + [ 1.8626e-09, -1.9791e-09, 0.0000e+00, ..., 1.0186e-09, + -3.2014e-10, -3.7835e-10], + [-1.4901e-08, 5.8208e-10, 0.0000e+00, ..., 2.9104e-11, + 1.0186e-09, 9.6043e-10], + [ 6.0245e-09, 4.6566e-10, 0.0000e+00, ..., 2.8522e-09, + 4.3656e-10, -5.8208e-10]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0428, 0.0135, 0.0134, 0.0286, 0.0358, -0.0194, 0.0286, 0.0295, + -0.0174, 0.0006], device='cuda:0'), grad: tensor([ 2.1828e-09, 1.4319e-08, 9.9826e-09, 1.1583e-08, -1.6356e-08, + 1.8161e-08, 4.6857e-09, 8.7894e-09, -7.9395e-08, 4.8283e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 252.64, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4217 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 451, weight, value: tensor([[ 0.0892, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0527, 0.0571, -0.0333, ..., -0.0830, 0.1753, 0.0156], + [-0.0727, -0.1989, -0.0080, ..., 0.2278, -0.2003, 0.0430], + ..., + [-0.2339, 0.1558, 0.0231, ..., -0.1815, 0.1244, 0.0665], + [ 0.0030, -0.0942, -0.0050, ..., -0.1716, -0.2038, -0.3100], + [ 0.0364, -0.0598, -0.0182, ..., -0.2838, -0.1973, 0.0216]], + device='cuda:0'), grad: tensor([[ 5.8208e-11, 2.9104e-10, 0.0000e+00, ..., 4.9477e-10, + 5.2387e-10, 1.1205e-08], + [-3.1723e-09, -5.8208e-09, 0.0000e+00, ..., 1.3388e-09, + -1.1700e-08, 2.4884e-08], + [ 3.8999e-09, 7.5670e-10, 0.0000e+00, ..., 4.3656e-10, + 1.4552e-09, 8.6729e-09], + ..., + [ 2.7940e-09, 2.6484e-09, 0.0000e+00, ..., 7.2760e-10, + 5.1805e-09, 6.5600e-08], + [-5.0641e-09, 1.9209e-09, 0.0000e+00, ..., 6.6939e-10, + 3.6962e-09, 7.3051e-09], + [ 1.8626e-09, -1.0768e-09, 0.0000e+00, ..., 1.2835e-08, + -1.9791e-09, -2.9034e-07]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0428, 0.0135, 0.0134, 0.0287, 0.0357, -0.0194, 0.0285, 0.0295, + -0.0174, 0.0007], device='cuda:0'), grad: tensor([ 4.5198e-08, 1.1001e-07, 5.7888e-08, 1.3533e-08, 5.8394e-07, + -6.8976e-09, 6.5251e-08, 2.7800e-07, -5.8208e-10, -1.1297e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 252.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4249 re_mapping 0.0016 re_causal 0.0068 /// teacc 99.15 lr 0.00001000 +Epoch 452, weight, value: tensor([[ 0.0893, -0.1873, -0.1459, ..., -0.1432, -0.1751, -0.1699], + [ 0.0528, 0.0571, -0.0333, ..., -0.0831, 0.1753, 0.0155], + [-0.0728, -0.1992, -0.0080, ..., 0.2278, -0.2004, 0.0430], + ..., + [-0.2340, 0.1559, 0.0231, ..., -0.1816, 0.1245, 0.0666], + [ 0.0030, -0.0942, -0.0050, ..., -0.1717, -0.2040, -0.3101], + [ 0.0364, -0.0598, -0.0182, ..., -0.2839, -0.1973, 0.0216]], + device='cuda:0'), grad: tensor([[ 3.2014e-10, 1.7462e-10, 0.0000e+00, ..., 3.2014e-10, + 2.3283e-10, 4.6566e-10], + [ 2.0082e-09, 1.4552e-10, 0.0000e+00, ..., 5.3522e-08, + 6.9267e-09, 2.7503e-08], + [ 8.1491e-10, 3.1432e-09, 0.0000e+00, ..., -5.8790e-08, + -8.3528e-09, -2.4273e-08], + ..., + [ 2.1537e-09, -4.8603e-09, 0.0000e+00, ..., 3.6962e-09, + -1.3970e-09, 1.7462e-10], + [-6.4902e-09, 2.3283e-10, 0.0000e+00, ..., 8.4401e-10, + 1.7753e-09, 1.1642e-09], + [-1.5454e-08, 1.7462e-09, 0.0000e+00, ..., 5.8208e-10, + 1.2224e-09, -2.2759e-08]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0428, 0.0134, 0.0134, 0.0287, 0.0357, -0.0195, 0.0285, 0.0295, + -0.0175, 0.0007], device='cuda:0'), grad: tensor([ 2.2410e-09, 9.7090e-08, -8.2888e-08, 1.2951e-08, 6.7055e-08, + 5.8790e-09, 2.7940e-09, 3.3178e-09, -2.1624e-08, -6.5193e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 251.79, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4209 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.16 lr 0.00001000 +Epoch 453, weight, value: tensor([[ 0.0893, -0.1873, -0.1459, ..., -0.1432, -0.1750, -0.1699], + [ 0.0529, 0.0571, -0.0334, ..., -0.0832, 0.1753, 0.0154], + [-0.0730, -0.1994, -0.0080, ..., 0.2279, -0.2005, 0.0431], + ..., + [-0.2340, 0.1560, 0.0231, ..., -0.1818, 0.1245, 0.0667], + [ 0.0030, -0.0943, -0.0050, ..., -0.1717, -0.2040, -0.3102], + [ 0.0364, -0.0598, -0.0182, ..., -0.2840, -0.1974, 0.0216]], + device='cuda:0'), grad: tensor([[-9.2987e-09, -1.6298e-09, 0.0000e+00, ..., 5.0932e-10, + -2.5320e-09, 2.1828e-10], + [-5.5006e-09, 6.9267e-09, 0.0000e+00, ..., 5.2241e-09, + -3.9581e-09, 8.7748e-09], + [ 1.2515e-09, 8.4838e-09, 0.0000e+00, ..., 2.4738e-10, + 6.5193e-09, 5.5006e-09], + ..., + [ 2.7649e-09, -2.1508e-08, 0.0000e+00, ..., 1.8335e-09, + -1.2238e-08, -1.1671e-08], + [ 2.8376e-09, 1.4843e-09, 0.0000e+00, ..., 1.7462e-10, + 2.4593e-09, 1.1933e-09], + [ 4.9768e-09, 1.0128e-08, 0.0000e+00, ..., 8.1782e-09, + 5.8935e-09, 1.3039e-08]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0428, 0.0134, 0.0134, 0.0287, 0.0357, -0.0195, 0.0285, 0.0296, + -0.0175, 0.0007], device='cuda:0'), grad: tensor([-6.9325e-08, 2.6892e-08, 2.9017e-08, 2.1726e-08, -5.5821e-08, + -3.2189e-08, 2.1479e-08, -3.6904e-08, 1.8073e-08, 8.4750e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 252.05, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4303 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.14 lr 0.00001000 +Epoch 454, weight, value: tensor([[ 0.0893, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1699], + [ 0.0529, 0.0570, -0.0333, ..., -0.0833, 0.1754, 0.0154], + [-0.0731, -0.1995, -0.0080, ..., 0.2280, -0.2005, 0.0431], + ..., + [-0.2341, 0.1561, 0.0231, ..., -0.1821, 0.1246, 0.0667], + [ 0.0030, -0.0943, -0.0050, ..., -0.1717, -0.2041, -0.3103], + [ 0.0364, -0.0598, -0.0182, ..., -0.2842, -0.1974, 0.0216]], + device='cuda:0'), grad: tensor([[-1.4261e-09, 1.0186e-10, 0.0000e+00, ..., 6.3446e-09, + 1.1642e-10, 2.0373e-10], + [ 7.4215e-10, 4.6421e-09, 0.0000e+00, ..., 6.6939e-10, + 4.8603e-09, 4.6712e-09], + [ 1.3533e-09, 4.7003e-09, 0.0000e+00, ..., 1.6007e-10, + 4.8312e-09, 3.7107e-09], + ..., + [ 3.4925e-10, -1.5280e-08, 0.0000e+00, ..., 5.0932e-10, + -1.6313e-08, -1.0943e-08], + [ 3.3469e-10, 2.6193e-10, 0.0000e+00, ..., 8.0036e-10, + 2.4738e-10, 3.3469e-10], + [-2.9104e-10, 1.6880e-09, 0.0000e+00, ..., 9.4587e-10, + 1.8917e-09, -1.3388e-09]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0428, 0.0133, 0.0134, 0.0287, 0.0357, -0.0195, 0.0285, 0.0296, + -0.0175, 0.0007], device='cuda:0'), grad: tensor([ 3.1665e-08, 2.6426e-08, 5.2969e-08, -1.0617e-07, 9.5897e-09, + 4.6857e-08, -4.7177e-08, -3.4808e-08, 2.3225e-08, 4.3656e-11], + device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 251.81, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4198 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 455, weight, value: tensor([[ 0.0893, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1699], + [ 0.0529, 0.0570, -0.0333, ..., -0.0835, 0.1753, 0.0152], + [-0.0732, -0.1997, -0.0080, ..., 0.2282, -0.2006, 0.0432], + ..., + [-0.2342, 0.1562, 0.0231, ..., -0.1821, 0.1247, 0.0669], + [ 0.0030, -0.0943, -0.0050, ..., -0.1718, -0.2041, -0.3104], + [ 0.0364, -0.0598, -0.0182, ..., -0.2843, -0.1975, 0.0216]], + device='cuda:0'), grad: tensor([[-1.8626e-09, -2.9104e-11, 0.0000e+00, ..., 9.0222e-10, + -1.4552e-11, 1.1205e-09], + [ 1.0332e-09, 5.0641e-09, 0.0000e+00, ..., 1.6007e-10, + 1.1933e-09, 5.8935e-09], + [ 1.6444e-09, 1.8917e-09, 0.0000e+00, ..., 3.1578e-09, + 3.7253e-09, 1.7142e-08], + ..., + [ 1.6589e-09, -9.5024e-09, 0.0000e+00, ..., 3.3033e-09, + -5.4424e-09, 4.5111e-10], + [ 3.7980e-09, 1.1350e-09, 0.0000e+00, ..., 1.3242e-09, + 1.3679e-09, 4.5693e-09], + [-1.6007e-08, 1.8335e-09, 0.0000e+00, ..., 3.0414e-09, + 1.3824e-09, -8.5856e-09]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0427, 0.0132, 0.0135, 0.0287, 0.0358, -0.0196, 0.0285, 0.0296, + -0.0176, 0.0006], device='cuda:0'), grad: tensor([ 1.6589e-09, 3.6671e-08, 2.8592e-07, -4.1095e-07, 6.9209e-08, + 3.2858e-08, -1.0215e-08, 4.0105e-08, 3.1490e-08, -5.4599e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 251.68, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4208 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.18 lr 0.00001000 +Epoch 456, weight, value: tensor([[ 0.0894, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1699], + [ 0.0530, 0.0569, -0.0333, ..., -0.0835, 0.1753, 0.0151], + [-0.0733, -0.1999, -0.0080, ..., 0.2282, -0.2007, 0.0432], + ..., + [-0.2342, 0.1564, 0.0230, ..., -0.1822, 0.1248, 0.0670], + [ 0.0030, -0.0943, -0.0050, ..., -0.1718, -0.2042, -0.3105], + [ 0.0364, -0.0599, -0.0182, ..., -0.2844, -0.1976, 0.0216]], + device='cuda:0'), grad: tensor([[-4.6857e-09, 1.7462e-10, 0.0000e+00, ..., 5.8208e-10, + -3.3469e-10, 1.8044e-09], + [-7.5815e-09, -6.4902e-09, 0.0000e+00, ..., 1.8044e-09, + -1.3388e-08, -5.9663e-10], + [ 1.3242e-09, 2.1537e-09, 0.0000e+00, ..., 6.5484e-10, + 2.9977e-09, 2.5320e-09], + ..., + [ 3.8126e-09, -5.4715e-09, 0.0000e+00, ..., 1.2515e-09, + -4.0745e-09, 1.4115e-09], + [ 2.3574e-09, 1.1496e-09, 0.0000e+00, ..., 8.0036e-10, + 2.2410e-09, 2.4302e-09], + [-3.2160e-09, 1.5862e-09, 0.0000e+00, ..., 1.0055e-08, + 2.2555e-09, -7.8435e-09]], device='cuda:0') +Epoch 456, bias, value: tensor([-0.0427, 0.0132, 0.0135, 0.0287, 0.0357, -0.0197, 0.0286, 0.0297, + -0.0176, 0.0006], device='cuda:0'), grad: tensor([-2.4535e-08, -3.3469e-09, 1.5832e-08, 1.9500e-08, -4.1036e-09, + -1.3155e-08, 1.2151e-08, 2.1057e-08, 2.1188e-08, -3.8301e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 251.93, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4174 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.16 lr 0.00001000 +Epoch 457, weight, value: tensor([[ 0.0894, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1699], + [ 0.0530, 0.0569, -0.0333, ..., -0.0836, 0.1753, 0.0151], + [-0.0735, -0.2001, -0.0080, ..., 0.2283, -0.2008, 0.0432], + ..., + [-0.2343, 0.1565, 0.0230, ..., -0.1823, 0.1249, 0.0671], + [ 0.0031, -0.0944, -0.0050, ..., -0.1719, -0.2042, -0.3106], + [ 0.0364, -0.0599, -0.0182, ..., -0.2846, -0.1977, 0.0216]], + device='cuda:0'), grad: tensor([[-7.5670e-10, 2.9104e-10, 0.0000e+00, ..., 5.9663e-10, + -1.3097e-10, 7.2760e-10], + [-1.1816e-08, -7.3342e-09, 0.0000e+00, ..., 5.1368e-09, + -2.6019e-08, 4.5547e-09], + [ 1.6735e-09, 9.1241e-09, 0.0000e+00, ..., 1.0041e-09, + 7.2760e-09, 7.1450e-09], + ..., + [ 6.6502e-09, -1.9005e-08, 0.0000e+00, ..., 2.0955e-09, + -2.9686e-09, -6.9558e-09], + [-2.4156e-09, 2.2701e-09, 0.0000e+00, ..., 2.3283e-10, + 4.0163e-09, 1.9791e-09], + [ 1.9063e-09, 5.6316e-09, 0.0000e+00, ..., 7.0286e-09, + 5.7044e-09, 4.0542e-08]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0427, 0.0131, 0.0135, 0.0287, 0.0358, -0.0197, 0.0286, 0.0298, + -0.0176, 0.0006], device='cuda:0'), grad: tensor([ 8.0036e-10, 2.2701e-09, 3.1432e-08, 8.4256e-09, -1.5844e-07, + 6.2864e-09, 1.1860e-08, -7.9017e-09, -1.1220e-08, 1.2922e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 251.80, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4281 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.19 lr 0.00001000 +Epoch 458, weight, value: tensor([[ 0.0895, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1700], + [ 0.0531, 0.0568, -0.0333, ..., -0.0837, 0.1753, 0.0150], + [-0.0737, -0.2003, -0.0080, ..., 0.2284, -0.2009, 0.0432], + ..., + [-0.2344, 0.1566, 0.0230, ..., -0.1824, 0.1250, 0.0672], + [ 0.0031, -0.0944, -0.0050, ..., -0.1719, -0.2043, -0.3106], + [ 0.0364, -0.0599, -0.0182, ..., -0.2847, -0.1978, 0.0216]], + device='cuda:0'), grad: tensor([[-2.6193e-10, 1.3097e-10, 0.0000e+00, ..., 6.1118e-10, + 8.7311e-11, 8.8767e-10], + [ 2.3574e-09, 1.5352e-08, 0.0000e+00, ..., 2.8987e-08, + 1.2064e-08, 3.6205e-08], + [ 2.1391e-09, 3.2014e-09, 0.0000e+00, ..., -4.1327e-09, + 2.7212e-09, 1.3533e-09], + ..., + [ 3.7398e-09, -2.7023e-08, 0.0000e+00, ..., 5.1368e-09, + -1.9805e-08, -6.7521e-09], + [ 1.4843e-08, 6.8394e-10, 0.0000e+00, ..., 1.1933e-09, + 1.1350e-09, 8.8912e-09], + [-2.3239e-08, 1.1365e-08, 0.0000e+00, ..., 5.8353e-09, + 9.2696e-09, -1.8263e-08]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0427, 0.0131, 0.0134, 0.0286, 0.0358, -0.0197, 0.0286, 0.0298, + -0.0176, 0.0006], device='cuda:0'), grad: tensor([ 3.8854e-09, 1.5437e-07, 2.5757e-08, -7.5554e-08, -3.7195e-08, + -3.5361e-09, 8.1054e-09, -2.2148e-08, 1.0111e-07, -1.5111e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 251.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4067 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.23 lr 0.00001000 +Epoch 459, weight, value: tensor([[ 0.0895, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1700], + [ 0.0531, 0.0568, -0.0333, ..., -0.0838, 0.1753, 0.0149], + [-0.0738, -0.2006, -0.0080, ..., 0.2285, -0.2010, 0.0432], + ..., + [-0.2344, 0.1567, 0.0230, ..., -0.1826, 0.1250, 0.0673], + [ 0.0031, -0.0944, -0.0050, ..., -0.1719, -0.2043, -0.3107], + [ 0.0364, -0.0599, -0.0182, ..., -0.2850, -0.1978, 0.0215]], + device='cuda:0'), grad: tensor([[-8.5856e-10, 1.4552e-11, 0.0000e+00, ..., 6.5484e-10, + 4.3656e-11, 6.9849e-10], + [-6.1118e-10, 2.9104e-11, 0.0000e+00, ..., 2.1551e-08, + -1.4697e-09, 2.2177e-08], + [ 7.7125e-10, 5.6752e-10, 0.0000e+00, ..., -5.8208e-11, + 6.5484e-10, 2.3283e-10], + ..., + [ 7.7125e-10, -1.6007e-09, 0.0000e+00, ..., 6.8831e-09, + -3.9290e-10, 6.8540e-09], + [ 8.1491e-10, 1.4552e-10, 0.0000e+00, ..., 6.8394e-10, + 3.4925e-10, 8.2946e-10], + [ 6.4028e-10, 3.6380e-10, 0.0000e+00, ..., 5.5705e-08, + 3.2014e-10, 5.5879e-08]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0427, 0.0130, 0.0135, 0.0286, 0.0358, -0.0197, 0.0285, 0.0299, + -0.0176, 0.0006], device='cuda:0'), grad: tensor([ 2.3283e-10, 5.9663e-08, 3.5507e-09, -5.4424e-08, -2.3097e-07, + 5.6112e-08, 2.5684e-08, 2.2468e-08, 1.1059e-08, 1.3947e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 252.30, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4364 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +Epoch 460, weight, value: tensor([[ 0.0895, -0.1873, -0.1459, ..., -0.1433, -0.1750, -0.1700], + [ 0.0532, 0.0568, -0.0333, ..., -0.0839, 0.1754, 0.0149], + [-0.0740, -0.2008, -0.0080, ..., 0.2285, -0.2011, 0.0432], + ..., + [-0.2345, 0.1568, 0.0230, ..., -0.1827, 0.1251, 0.0674], + [ 0.0030, -0.0945, -0.0050, ..., -0.1720, -0.2046, -0.3109], + [ 0.0364, -0.0600, -0.0182, ..., -0.2851, -0.1979, 0.0215]], + device='cuda:0'), grad: tensor([[-3.2160e-09, 2.4593e-09, 0.0000e+00, ..., 8.8767e-10, + 2.0082e-09, 2.7212e-09], + [-3.2596e-09, 5.1397e-08, 0.0000e+00, ..., 2.7503e-09, + 2.7489e-08, 3.5681e-08], + [ 9.0222e-10, 9.2201e-08, 0.0000e+00, ..., -1.4639e-08, + 6.3912e-08, 5.0117e-08], + ..., + [ 2.3138e-09, -3.3900e-07, 0.0000e+00, ..., 3.7980e-09, + -2.2526e-07, -2.2398e-07], + [-3.4343e-09, 3.1258e-08, 0.0000e+00, ..., 6.5338e-09, + 2.2628e-08, 2.7198e-08], + [ 1.8917e-09, 4.2433e-08, 0.0000e+00, ..., 1.6298e-09, + 2.8871e-08, 3.0122e-08]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0427, 0.0130, 0.0134, 0.0285, 0.0358, -0.0196, 0.0285, 0.0299, + -0.0177, 0.0006], device='cuda:0'), grad: tensor([-1.6735e-09, 1.8370e-07, 2.9267e-07, 1.3178e-07, 2.6729e-07, + 4.4733e-08, 5.5152e-09, -1.1902e-06, 1.1607e-07, 1.6321e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 252.50, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4287 re_mapping 0.0015 re_causal 0.0068 /// teacc 99.12 lr 0.00001000 +Epoch 461, weight, value: tensor([[ 0.0895, -0.1873, -0.1459, ..., -0.1434, -0.1750, -0.1700], + [ 0.0533, 0.0568, -0.0333, ..., -0.0839, 0.1754, 0.0148], + [-0.0741, -0.2009, -0.0080, ..., 0.2286, -0.2012, 0.0432], + ..., + [-0.2346, 0.1569, 0.0230, ..., -0.1828, 0.1252, 0.0675], + [ 0.0030, -0.0946, -0.0050, ..., -0.1720, -0.2047, -0.3110], + [ 0.0364, -0.0600, -0.0182, ..., -0.2853, -0.1980, 0.0215]], + device='cuda:0'), grad: tensor([[-1.3562e-07, 1.7462e-10, 0.0000e+00, ..., 1.1642e-10, + 2.9104e-10, 7.5670e-10], + [-1.3679e-08, -7.3342e-09, 0.0000e+00, ..., 3.7398e-09, + -2.1799e-08, 5.2387e-10], + [ 1.3286e-08, 1.9645e-09, 0.0000e+00, ..., -1.3533e-09, + 2.6339e-09, 9.3132e-10], + ..., + [ 9.9972e-09, -2.8813e-09, 0.0000e+00, ..., 2.0664e-09, + 5.4133e-09, 5.6170e-09], + [ 1.2369e-09, 2.4593e-09, 0.0000e+00, ..., 1.3097e-10, + 4.7148e-09, 2.8522e-09], + [ 2.4418e-08, -8.7311e-11, 0.0000e+00, ..., 1.4261e-09, + 1.7753e-09, -4.0396e-08]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0426, 0.0130, 0.0134, 0.0285, 0.0358, -0.0195, 0.0285, 0.0300, + -0.0177, 0.0006], device='cuda:0'), grad: tensor([-6.1421e-07, -2.9249e-09, 6.1118e-08, 2.6263e-07, 1.2864e-07, + 1.4115e-08, 9.6741e-08, 3.6089e-08, -2.7067e-08, 4.6217e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 252.01, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4097 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.15 lr 0.00001000 +Epoch 462, weight, value: tensor([[ 0.0896, -0.1873, -0.1459, ..., -0.1434, -0.1750, -0.1700], + [ 0.0535, 0.0568, -0.0333, ..., -0.0840, 0.1755, 0.0148], + [-0.0744, -0.2011, -0.0080, ..., 0.2286, -0.2013, 0.0432], + ..., + [-0.2346, 0.1571, 0.0230, ..., -0.1829, 0.1253, 0.0677], + [ 0.0029, -0.0948, -0.0050, ..., -0.1721, -0.2049, -0.3112], + [ 0.0364, -0.0600, -0.0182, ..., -0.2855, -0.1981, 0.0215]], + device='cuda:0'), grad: tensor([[-4.4383e-09, 3.0559e-10, 0.0000e+00, ..., 4.3656e-11, + 5.5297e-10, 2.1828e-10], + [-1.3635e-08, -1.0972e-08, 0.0000e+00, ..., 2.9104e-10, + -2.3501e-08, -7.5524e-09], + [ 2.2701e-09, 1.1933e-09, 0.0000e+00, ..., 9.7498e-10, + 2.0955e-09, 1.6735e-09], + ..., + [ 5.1659e-09, 3.5507e-09, 0.0000e+00, ..., 2.3283e-10, + 7.1450e-09, 2.9540e-09], + [-1.3679e-09, 2.2847e-09, 0.0000e+00, ..., 2.9104e-11, + 6.5484e-09, 2.1682e-09], + [ 3.8999e-09, 8.7311e-10, 0.0000e+00, ..., 7.2760e-10, + 1.4552e-09, 1.6153e-09]], device='cuda:0') +Epoch 462, bias, value: tensor([-0.0426, 0.0130, 0.0134, 0.0285, 0.0359, -0.0195, 0.0285, 0.0301, + -0.0179, 0.0006], device='cuda:0'), grad: tensor([-2.4462e-08, -3.6118e-08, 1.5221e-08, 9.5461e-09, -2.1100e-09, + 1.0914e-08, 1.6618e-08, 2.1813e-08, -2.3298e-08, 2.5088e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 252.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3975 re_mapping 0.0015 re_causal 0.0063 /// teacc 99.18 lr 0.00001000 +Epoch 463, weight, value: tensor([[ 0.0896, -0.1873, -0.1459, ..., -0.1434, -0.1750, -0.1700], + [ 0.0536, 0.0568, -0.0332, ..., -0.0841, 0.1756, 0.0147], + [-0.0745, -0.2013, -0.0080, ..., 0.2287, -0.2015, 0.0432], + ..., + [-0.2348, 0.1572, 0.0230, ..., -0.1830, 0.1254, 0.0677], + [ 0.0029, -0.0948, -0.0050, ..., -0.1721, -0.2050, -0.3113], + [ 0.0364, -0.0601, -0.0182, ..., -0.2858, -0.1982, 0.0214]], + device='cuda:0'), grad: tensor([[ 2.1537e-09, 1.3097e-10, 0.0000e+00, ..., 3.2043e-08, + 2.4738e-10, 3.9290e-10], + [-9.9390e-09, -9.1823e-09, 0.0000e+00, ..., 3.0414e-09, + -2.1100e-08, -5.7480e-09], + [ 1.1350e-09, 1.7171e-09, 0.0000e+00, ..., -2.0664e-09, + 2.3720e-09, -1.5716e-09], + ..., + [ 5.8790e-09, 2.8667e-09, 0.0000e+00, ..., 6.8394e-10, + 8.6438e-09, 3.0559e-09], + [ 4.1036e-09, 1.7462e-09, 0.0000e+00, ..., 4.4820e-09, + 3.6962e-09, 2.1537e-09], + [-5.9226e-09, 1.0186e-09, 0.0000e+00, ..., 2.3720e-09, + 1.4988e-09, -2.5320e-09]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0426, 0.0130, 0.0134, 0.0284, 0.0359, -0.0195, 0.0285, 0.0301, + -0.0179, 0.0005], device='cuda:0'), grad: tensor([ 1.4820e-07, -2.3108e-08, 2.0227e-09, 1.4115e-08, 1.0419e-07, + 3.0617e-08, -2.9523e-07, 2.2017e-08, 3.5681e-08, -1.4537e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 251.94, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4135 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.21 lr 0.00001000 +Epoch 464, weight, value: tensor([[ 0.0897, -0.1873, -0.1459, ..., -0.1434, -0.1751, -0.1700], + [ 0.0537, 0.0568, -0.0332, ..., -0.0842, 0.1757, 0.0147], + [-0.0746, -0.2015, -0.0080, ..., 0.2288, -0.2016, 0.0432], + ..., + [-0.2349, 0.1573, 0.0230, ..., -0.1832, 0.1254, 0.0678], + [ 0.0029, -0.0949, -0.0050, ..., -0.1722, -0.2051, -0.3114], + [ 0.0364, -0.0601, -0.0182, ..., -0.2861, -0.1983, 0.0214]], + device='cuda:0'), grad: tensor([[-1.0186e-10, 1.1933e-09, 0.0000e+00, ..., 1.0041e-09, + 7.4215e-10, 1.6589e-09], + [ 4.8894e-09, 2.5640e-08, 0.0000e+00, ..., 2.5160e-08, + 1.1743e-08, 3.9814e-08], + [-3.2742e-09, 1.5163e-08, 0.0000e+00, ..., -4.8429e-08, + 9.7498e-09, -2.7765e-08], + ..., + [ 2.1100e-09, -8.1025e-08, 0.0000e+00, ..., 2.2264e-09, + -4.7294e-08, -6.2340e-08], + [-2.0664e-09, -1.7317e-09, 0.0000e+00, ..., 8.9349e-09, + 1.7026e-09, 8.0327e-09], + [ 2.2992e-09, 1.5454e-08, 0.0000e+00, ..., 2.5757e-09, + 9.4296e-09, 1.3111e-08]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0426, 0.0130, 0.0134, 0.0285, 0.0360, -0.0195, 0.0285, 0.0301, + -0.0180, 0.0005], device='cuda:0'), grad: tensor([ 9.7207e-09, 1.6263e-07, -8.8941e-08, 7.6019e-08, 5.9314e-08, + -1.0268e-07, 2.1391e-08, -2.2224e-07, 2.2075e-08, 7.0664e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 252.28, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4186 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.20 lr 0.00001000 +Epoch 465, weight, value: tensor([[ 0.0897, -0.1873, -0.1459, ..., -0.1435, -0.1751, -0.1700], + [ 0.0539, 0.0568, -0.0332, ..., -0.0842, 0.1758, 0.0146], + [-0.0749, -0.2017, -0.0080, ..., 0.2288, -0.2017, 0.0432], + ..., + [-0.2350, 0.1574, 0.0230, ..., -0.1832, 0.1255, 0.0679], + [ 0.0029, -0.0950, -0.0050, ..., -0.1722, -0.2053, -0.3115], + [ 0.0364, -0.0602, -0.0182, ..., -0.2863, -0.1984, 0.0213]], + device='cuda:0'), grad: tensor([[-5.5297e-10, 4.3656e-11, 0.0000e+00, ..., 7.7125e-10, + 7.2760e-11, 4.6566e-10], + [-2.9395e-09, 1.0768e-09, 0.0000e+00, ..., 4.7003e-09, + -3.1432e-09, 4.5693e-09], + [ 5.3842e-10, 1.4843e-09, 0.0000e+00, ..., -4.8021e-10, + 1.7317e-09, 3.2014e-10], + ..., + [ 1.7608e-09, -5.2096e-09, 0.0000e+00, ..., 2.0227e-09, + -2.4738e-09, -3.6234e-09], + [ 2.4593e-09, 7.5670e-10, 0.0000e+00, ..., 1.4261e-09, + 9.4587e-10, 1.9645e-09], + [ 2.0518e-09, 5.8208e-10, 0.0000e+00, ..., 7.2905e-09, + 5.6752e-10, 3.5943e-09]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0426, 0.0130, 0.0133, 0.0286, 0.0360, -0.0195, 0.0285, 0.0301, + -0.0180, 0.0004], device='cuda:0'), grad: tensor([ 2.1537e-09, 1.8714e-08, 6.3737e-08, -5.5152e-08, -5.7276e-08, + -7.0373e-08, 3.2363e-08, 1.0579e-08, 2.6281e-08, 4.0105e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 252.27, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4082 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.19 lr 0.00001000 +Epoch 466, weight, value: tensor([[ 0.0897, -0.1874, -0.1459, ..., -0.1435, -0.1751, -0.1700], + [ 0.0541, 0.0569, -0.0332, ..., -0.0843, 0.1759, 0.0147], + [-0.0752, -0.2020, -0.0080, ..., 0.2289, -0.2019, 0.0432], + ..., + [-0.2351, 0.1575, 0.0230, ..., -0.1833, 0.1255, 0.0679], + [ 0.0028, -0.0951, -0.0050, ..., -0.1722, -0.2055, -0.3117], + [ 0.0364, -0.0602, -0.0182, ..., -0.2866, -0.1985, 0.0213]], + device='cuda:0'), grad: tensor([[-1.0768e-09, 2.6193e-10, 0.0000e+00, ..., 1.4988e-09, + 3.7835e-10, 1.1059e-09], + [-7.5961e-09, 8.1491e-09, 0.0000e+00, ..., 3.2363e-08, + -8.8767e-09, 2.5742e-08], + [ 1.2078e-09, 4.9768e-09, 0.0000e+00, ..., -3.4750e-08, + 4.4674e-09, -1.9558e-08], + ..., + [ 4.2492e-09, -1.5832e-08, 0.0000e+00, ..., 1.6633e-08, + -1.8335e-09, 8.0909e-09], + [ 1.4115e-09, 1.3679e-09, 0.0000e+00, ..., 5.5297e-10, + 2.5757e-09, 1.6444e-09], + [ 1.0186e-09, 3.5361e-09, 0.0000e+00, ..., 6.6939e-09, + 2.1973e-09, 8.7748e-09]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0425, 0.0131, 0.0132, 0.0286, 0.0360, -0.0195, 0.0285, 0.0301, + -0.0182, 0.0004], device='cuda:0'), grad: tensor([-2.2847e-09, 8.7661e-08, -8.3644e-08, 2.3108e-08, -7.5263e-08, + 2.7794e-09, 7.3342e-09, 3.1781e-08, 5.1077e-09, 3.2771e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 252.29, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4235 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.18 lr 0.00001000 +Epoch 467, weight, value: tensor([[ 0.0898, -0.1874, -0.1459, ..., -0.1435, -0.1751, -0.1700], + [ 0.0543, 0.0569, -0.0332, ..., -0.0844, 0.1761, 0.0147], + [-0.0755, -0.2022, -0.0080, ..., 0.2290, -0.2021, 0.0431], + ..., + [-0.2352, 0.1575, 0.0230, ..., -0.1834, 0.1255, 0.0680], + [ 0.0027, -0.0953, -0.0050, ..., -0.1723, -0.2057, -0.3119], + [ 0.0364, -0.0602, -0.0182, ..., -0.2867, -0.1986, 0.0213]], + device='cuda:0'), grad: tensor([[ 4.0891e-09, 4.0309e-09, 0.0000e+00, ..., 6.2573e-10, + 8.0181e-09, 4.1910e-09], + [-1.2515e-07, 2.6193e-10, 0.0000e+00, ..., -2.7663e-08, + -1.6787e-07, 1.3999e-08], + [ 2.5568e-08, 3.3324e-08, 0.0000e+00, ..., 8.7020e-09, + 6.7404e-08, 2.6892e-08], + ..., + [ 2.4753e-08, -1.2876e-07, 0.0000e+00, ..., 8.1200e-09, + -7.7824e-08, -1.1479e-07], + [ 1.8568e-08, 1.6502e-08, 0.0000e+00, ..., 2.0664e-09, + 3.5827e-08, 1.3402e-08], + [ 6.8540e-09, 7.1159e-09, 0.0000e+00, ..., 2.6339e-09, + 2.6863e-08, -2.6979e-08]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0425, 0.0132, 0.0131, 0.0285, 0.0361, -0.0194, 0.0285, 0.0301, + -0.0183, 0.0004], device='cuda:0'), grad: tensor([ 2.8434e-08, -4.7288e-07, 2.1176e-07, 7.2294e-08, 2.9220e-07, + -2.3283e-10, 7.1828e-08, -2.1479e-07, 1.0472e-07, -8.3179e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 251.83, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4173 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.23 lr 0.00001000 +Epoch 468, weight, value: tensor([[ 0.0898, -0.1874, -0.1459, ..., -0.1435, -0.1751, -0.1700], + [ 0.0545, 0.0569, -0.0332, ..., -0.0844, 0.1762, 0.0146], + [-0.0758, -0.2024, -0.0080, ..., 0.2290, -0.2024, 0.0431], + ..., + [-0.2353, 0.1577, 0.0230, ..., -0.1836, 0.1257, 0.0681], + [ 0.0026, -0.0954, -0.0050, ..., -0.1723, -0.2058, -0.3120], + [ 0.0364, -0.0603, -0.0182, ..., -0.2870, -0.1987, 0.0213]], + device='cuda:0'), grad: tensor([[ 1.6007e-10, 1.0186e-10, 0.0000e+00, ..., 3.9290e-10, + 1.8917e-10, 1.8917e-10], + [-5.7626e-09, -5.1950e-09, 0.0000e+00, ..., 2.0955e-09, + -1.2034e-08, -2.1246e-09], + [ 1.1205e-09, 2.2847e-09, 0.0000e+00, ..., 9.5315e-09, + 2.4447e-09, 7.0577e-09], + ..., + [ 4.1473e-09, 4.0745e-10, 0.0000e+00, ..., 1.7462e-09, + 5.5879e-09, 1.0623e-09], + [-1.4261e-09, 7.8580e-10, 0.0000e+00, ..., 2.9104e-10, + 1.4988e-09, 6.1118e-10], + [ 9.4587e-10, 1.0186e-09, 0.0000e+00, ..., 6.3446e-09, + 1.2806e-09, 4.2637e-09]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0425, 0.0132, 0.0129, 0.0285, 0.0362, -0.0193, 0.0284, 0.0302, + -0.0184, 0.0004], device='cuda:0'), grad: tensor([ 1.9936e-09, -1.5396e-08, 3.4488e-08, 1.1350e-08, -5.1514e-08, + -1.9791e-09, 1.1743e-08, 1.6284e-08, -8.4838e-09, 2.4069e-08], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 467---------------------------------------------------- +epoch 467, time 252.55, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4151 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.26 lr 0.00001000 +Epoch 469, weight, value: tensor([[ 0.0899, -0.1874, -0.1459, ..., -0.1435, -0.1751, -0.1700], + [ 0.0546, 0.0569, -0.0332, ..., -0.0845, 0.1762, 0.0145], + [-0.0760, -0.2026, -0.0080, ..., 0.2291, -0.2025, 0.0431], + ..., + [-0.2354, 0.1578, 0.0230, ..., -0.1838, 0.1258, 0.0682], + [ 0.0026, -0.0954, -0.0050, ..., -0.1724, -0.2059, -0.3121], + [ 0.0364, -0.0603, -0.0182, ..., -0.2872, -0.1988, 0.0213]], + device='cuda:0'), grad: tensor([[ 1.3097e-10, 1.3097e-09, 0.0000e+00, ..., 4.8545e-08, + 7.2760e-11, 2.8376e-09], + [-5.3842e-10, 1.1496e-09, 0.0000e+00, ..., 3.9290e-09, + -1.9209e-09, 2.6339e-09], + [ 4.8021e-10, 3.1432e-09, 0.0000e+00, ..., 1.8626e-09, + 1.9645e-09, 3.2451e-09], + ..., + [ 2.2410e-09, 3.5943e-09, 0.0000e+00, ..., 4.2201e-10, + -3.1287e-09, 1.7390e-08], + [-1.0332e-09, 5.3842e-10, 0.0000e+00, ..., 4.7730e-09, + 4.2201e-10, 1.2369e-09], + [-4.8458e-09, -1.5862e-08, 0.0000e+00, ..., 3.6816e-09, + 1.3970e-09, -4.5227e-08]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0425, 0.0131, 0.0129, 0.0285, 0.0362, -0.0193, 0.0284, 0.0302, + -0.0185, 0.0004], device='cuda:0'), grad: tensor([ 2.3702e-07, 3.4634e-08, 2.5218e-08, -1.5469e-08, 2.3958e-07, + 4.3918e-08, -4.7032e-07, 9.2608e-08, 4.7876e-09, -1.7928e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 252.03, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4050 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.20 lr 0.00001000 +Epoch 470, weight, value: tensor([[ 0.0899, -0.1874, -0.1459, ..., -0.1435, -0.1751, -0.1700], + [ 0.0547, 0.0568, -0.0332, ..., -0.0846, 0.1763, 0.0145], + [-0.0761, -0.2028, -0.0080, ..., 0.2292, -0.2026, 0.0431], + ..., + [-0.2355, 0.1580, 0.0230, ..., -0.1840, 0.1259, 0.0683], + [ 0.0026, -0.0955, -0.0050, ..., -0.1724, -0.2059, -0.3122], + [ 0.0364, -0.0603, -0.0182, ..., -0.2874, -0.1989, 0.0213]], + device='cuda:0'), grad: tensor([[ 2.6193e-10, 5.8208e-11, 0.0000e+00, ..., 3.1199e-08, + 1.4552e-10, 1.0186e-09], + [ 3.4343e-09, -1.7608e-09, 0.0000e+00, ..., 2.2847e-09, + -4.8894e-09, 1.1190e-08], + [ 9.1677e-10, 5.5297e-10, 0.0000e+00, ..., 1.7171e-09, + 1.1205e-09, 6.7957e-09], + ..., + [ 1.8481e-09, 8.8767e-10, 0.0000e+00, ..., 2.4884e-09, + 2.0955e-09, 6.4028e-09], + [ 1.0914e-09, 6.2573e-10, 0.0000e+00, ..., 4.5111e-10, + 1.5134e-09, 1.8335e-09], + [-1.9500e-08, 3.7835e-10, 0.0000e+00, ..., 2.8667e-09, + 7.1304e-10, -1.7972e-08]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0425, 0.0131, 0.0129, 0.0285, 0.0362, -0.0193, 0.0284, 0.0303, + -0.0185, 0.0004], device='cuda:0'), grad: tensor([ 1.4855e-07, 4.7003e-08, 4.5227e-08, -5.8557e-08, 3.2247e-08, + -2.7067e-09, -1.4983e-07, 2.8682e-08, 8.1636e-09, -7.8289e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 251.68, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3964 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.25 lr 0.00001000 +Epoch 471, weight, value: tensor([[ 0.0899, -0.1874, -0.1459, ..., -0.1436, -0.1751, -0.1701], + [ 0.0547, 0.0568, -0.0332, ..., -0.0847, 0.1763, 0.0144], + [-0.0763, -0.2031, -0.0080, ..., 0.2293, -0.2028, 0.0431], + ..., + [-0.2356, 0.1582, 0.0230, ..., -0.1841, 0.1260, 0.0685], + [ 0.0026, -0.0955, -0.0050, ..., -0.1725, -0.2060, -0.3123], + [ 0.0364, -0.0604, -0.0182, ..., -0.2875, -0.1990, 0.0213]], + device='cuda:0'), grad: tensor([[-6.4902e-09, 1.3097e-10, 0.0000e+00, ..., 2.1828e-10, + 2.0373e-10, 1.4552e-10], + [-6.6066e-09, -6.2137e-09, 0.0000e+00, ..., 1.3824e-09, + -1.1903e-08, -1.9063e-09], + [ 1.2369e-09, 4.0745e-09, 0.0000e+00, ..., 4.0745e-10, + 3.9581e-09, 3.5070e-09], + ..., + [ 2.8376e-09, -2.3865e-09, 0.0000e+00, ..., 7.4215e-10, + 8.7311e-10, -1.7171e-09], + [ 2.4302e-09, 1.5280e-09, 0.0000e+00, ..., 2.3283e-10, + 2.5902e-09, 1.0186e-09], + [ 3.9145e-09, 1.2806e-09, 0.0000e+00, ..., 6.9849e-10, + 1.5280e-09, -1.2515e-09]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0424, 0.0131, 0.0128, 0.0285, 0.0362, -0.0194, 0.0284, 0.0303, + -0.0185, 0.0004], device='cuda:0'), grad: tensor([-4.4005e-08, -5.0641e-09, 3.2975e-08, -1.2678e-07, 7.3342e-09, + 2.6426e-08, 4.1182e-09, 1.5396e-08, 4.8487e-08, 5.5239e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 251.98, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4146 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +Epoch 472, weight, value: tensor([[ 0.0900, -0.1874, -0.1459, ..., -0.1436, -0.1751, -0.1701], + [ 0.0548, 0.0568, -0.0332, ..., -0.0848, 0.1764, 0.0144], + [-0.0764, -0.2032, -0.0080, ..., 0.2294, -0.2029, 0.0431], + ..., + [-0.2357, 0.1582, 0.0229, ..., -0.1842, 0.1260, 0.0685], + [ 0.0026, -0.0955, -0.0050, ..., -0.1725, -0.2060, -0.3124], + [ 0.0364, -0.0604, -0.0182, ..., -0.2877, -0.1991, 0.0213]], + device='cuda:0'), grad: tensor([[-9.8953e-10, -5.8208e-11, 0.0000e+00, ..., 3.3469e-10, + 2.0373e-10, 2.4738e-10], + [-3.0850e-09, 3.0996e-09, 0.0000e+00, ..., 7.5670e-10, + -8.8767e-10, 4.0163e-09], + [ 4.6566e-10, 1.0870e-08, 0.0000e+00, ..., 1.7462e-10, + 9.1677e-09, 9.8225e-09], + ..., + [ 2.0227e-09, -1.6284e-08, 0.0000e+00, ..., 3.2014e-10, + -1.1321e-08, -1.4974e-08], + [ 1.0623e-09, 9.6043e-10, 0.0000e+00, ..., 6.6939e-10, + 1.5571e-09, 8.8767e-10], + [ 2.6193e-10, 1.0332e-09, 0.0000e+00, ..., 5.8208e-11, + 9.1677e-10, -8.0036e-10]], device='cuda:0') +Epoch 472, bias, value: tensor([-0.0424, 0.0130, 0.0128, 0.0285, 0.0362, -0.0194, 0.0285, 0.0303, + -0.0185, 0.0003], device='cuda:0'), grad: tensor([-5.3260e-09, 7.8144e-09, 3.3615e-08, -2.1246e-09, 1.3926e-08, + 5.0204e-09, -1.7608e-09, -3.9407e-08, 7.1886e-09, 3.0122e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 252.29, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4186 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.20 lr 0.00001000 +Epoch 473, weight, value: tensor([[ 0.0900, -0.1874, -0.1459, ..., -0.1436, -0.1751, -0.1701], + [ 0.0549, 0.0567, -0.0332, ..., -0.0849, 0.1764, 0.0143], + [-0.0765, -0.2034, -0.0080, ..., 0.2294, -0.2030, 0.0431], + ..., + [-0.2358, 0.1583, 0.0229, ..., -0.1842, 0.1261, 0.0686], + [ 0.0026, -0.0956, -0.0050, ..., -0.1726, -0.2060, -0.3124], + [ 0.0365, -0.0604, -0.0182, ..., -0.2879, -0.1992, 0.0213]], + device='cuda:0'), grad: tensor([[ 5.0932e-10, 1.0186e-10, 0.0000e+00, ..., 1.6298e-09, + 2.2555e-10, 1.5134e-09], + [-1.0747e-08, -1.2660e-08, 0.0000e+00, ..., 8.4110e-09, + -2.0576e-08, 7.4215e-10], + [ 8.3674e-10, 6.7666e-10, 0.0000e+00, ..., -3.0268e-08, + 4.3656e-11, -1.7288e-08], + ..., + [ 1.1496e-08, 8.4692e-09, 0.0000e+00, ..., 2.8740e-09, + 1.4101e-08, 1.2304e-08], + [ 7.4942e-09, 3.2232e-09, 0.0000e+00, ..., 8.7093e-09, + 5.6098e-09, 1.0834e-08], + [-1.4050e-08, 5.0932e-10, 0.0000e+00, ..., 1.2587e-09, + 7.8580e-10, -2.0576e-08]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0424, 0.0130, 0.0127, 0.0286, 0.0362, -0.0196, 0.0285, 0.0303, + -0.0186, 0.0003], device='cuda:0'), grad: tensor([ 9.5388e-09, -6.6429e-09, -1.0879e-07, 2.1450e-08, 6.4145e-08, + -4.6930e-09, 4.9840e-09, 6.5251e-08, 7.3342e-08, -9.5461e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 251.76, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4137 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.22 lr 0.00001000 +Epoch 474, weight, value: tensor([[ 0.0901, -0.1874, -0.1459, ..., -0.1437, -0.1751, -0.1701], + [ 0.0550, 0.0567, -0.0332, ..., -0.0850, 0.1764, 0.0142], + [-0.0767, -0.2036, -0.0080, ..., 0.2295, -0.2032, 0.0431], + ..., + [-0.2359, 0.1585, 0.0229, ..., -0.1844, 0.1262, 0.0687], + [ 0.0026, -0.0956, -0.0050, ..., -0.1727, -0.2061, -0.3125], + [ 0.0365, -0.0605, -0.0182, ..., -0.2880, -0.1993, 0.0214]], + device='cuda:0'), grad: tensor([[ 1.2478e-08, 1.3111e-08, 0.0000e+00, ..., 4.1473e-10, + 2.8158e-08, 1.6080e-09], + [-1.2410e-07, -1.2410e-07, 0.0000e+00, ..., 1.8117e-09, + -2.6636e-07, -1.3933e-08], + [ 5.5443e-09, 5.4933e-09, 0.0000e+00, ..., -9.4587e-10, + 1.1525e-08, 2.3283e-10], + ..., + [ 6.3083e-09, 5.9299e-09, 0.0000e+00, ..., 1.8554e-09, + 1.2806e-08, 3.1578e-09], + [ 7.5088e-08, 7.3982e-08, 0.0000e+00, ..., 3.2742e-10, + 1.5832e-07, 8.9203e-09], + [ 3.7908e-09, 2.9468e-09, 0.0000e+00, ..., 2.1828e-09, + 6.2355e-09, 1.7099e-09]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0424, 0.0130, 0.0127, 0.0286, 0.0362, -0.0196, 0.0286, 0.0304, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([ 5.6607e-08, -5.4808e-07, 2.3196e-08, 4.5140e-08, 1.6676e-08, + -3.0093e-08, 6.4145e-08, 3.3906e-08, 3.3551e-07, 2.1653e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 251.96, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4054 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.19 lr 0.00001000 +Epoch 475, weight, value: tensor([[ 0.0901, -0.1874, -0.1459, ..., -0.1438, -0.1751, -0.1701], + [ 0.0551, 0.0568, -0.0332, ..., -0.0851, 0.1765, 0.0142], + [-0.0768, -0.2038, -0.0080, ..., 0.2296, -0.2033, 0.0432], + ..., + [-0.2361, 0.1585, 0.0229, ..., -0.1845, 0.1262, 0.0688], + [ 0.0026, -0.0957, -0.0050, ..., -0.1727, -0.2062, -0.3126], + [ 0.0366, -0.0605, -0.0182, ..., -0.2882, -0.1994, 0.0214]], + device='cuda:0'), grad: tensor([[ 8.0036e-11, 8.7311e-11, 0.0000e+00, ..., 3.8563e-10, + 1.2369e-10, 2.0373e-10], + [-2.5175e-09, 1.0325e-08, 0.0000e+00, ..., 9.0222e-10, + 3.3688e-09, 1.0288e-08], + [ 5.5297e-10, 3.9872e-09, 0.0000e+00, ..., 3.0559e-10, + 3.4051e-09, 3.5216e-09], + ..., + [ 1.5280e-09, -2.4127e-08, 0.0000e+00, ..., 1.0186e-09, + -1.5789e-08, -1.8073e-08], + [-6.5484e-10, 6.8394e-10, 0.0000e+00, ..., 1.9645e-10, + 1.3242e-09, 6.4028e-10], + [ 8.6584e-10, 9.6770e-10, 0.0000e+00, ..., 1.7753e-09, + 8.3674e-10, 1.1205e-09]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0424, 0.0129, 0.0127, 0.0286, 0.0362, -0.0197, 0.0286, 0.0303, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([ 2.4520e-09, 2.4782e-08, 1.2580e-08, 2.4229e-09, 8.0268e-08, + 4.5329e-09, -6.9151e-08, -4.0513e-08, -3.4779e-09, 7.6616e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 251.99, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3975 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.22 lr 0.00001000 +Epoch 476, weight, value: tensor([[ 0.0901, -0.1874, -0.1459, ..., -0.1438, -0.1751, -0.1701], + [ 0.0552, 0.0568, -0.0331, ..., -0.0852, 0.1767, 0.0142], + [-0.0769, -0.2040, -0.0080, ..., 0.2297, -0.2034, 0.0432], + ..., + [-0.2363, 0.1586, 0.0229, ..., -0.1846, 0.1262, 0.0688], + [ 0.0027, -0.0958, -0.0050, ..., -0.1727, -0.2063, -0.3127], + [ 0.0366, -0.0605, -0.0182, ..., -0.2884, -0.1995, 0.0214]], + device='cuda:0'), grad: tensor([[-2.8376e-10, 2.6193e-10, 0.0000e+00, ..., 7.1304e-10, + 4.7294e-10, 8.1491e-10], + [-8.9349e-09, -7.2760e-10, 0.0000e+00, ..., 1.1161e-08, + -7.3633e-09, 2.1115e-08], + [ 1.7972e-09, 3.7107e-09, 0.0000e+00, ..., -8.2000e-09, + 6.3737e-09, 9.0440e-09], + ..., + [ 3.3615e-09, -3.3440e-08, 0.0000e+00, ..., 1.4443e-08, + -4.5635e-08, -1.8961e-08], + [ 3.7107e-09, 1.8626e-09, 0.0000e+00, ..., 1.3242e-09, + 3.9654e-09, 2.0300e-09], + [ 1.0186e-09, 1.7826e-08, 0.0000e+00, ..., 3.4663e-08, + 2.5728e-08, 9.1561e-08]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0424, 0.0129, 0.0127, 0.0285, 0.0361, -0.0198, 0.0287, 0.0303, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([ 1.5280e-10, 2.9395e-08, 1.3031e-08, 1.5163e-08, -2.1176e-07, + -7.6616e-09, 6.5993e-09, -5.7218e-08, 2.1551e-08, 2.0582e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 252.20, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4001 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.24 lr 0.00001000 +Epoch 477, weight, value: tensor([[ 0.0902, -0.1874, -0.1459, ..., -0.1439, -0.1751, -0.1701], + [ 0.0553, 0.0568, -0.0331, ..., -0.0853, 0.1767, 0.0141], + [-0.0770, -0.2042, -0.0080, ..., 0.2298, -0.2036, 0.0432], + ..., + [-0.2363, 0.1587, 0.0229, ..., -0.1847, 0.1263, 0.0689], + [ 0.0027, -0.0958, -0.0050, ..., -0.1728, -0.2063, -0.3128], + [ 0.0366, -0.0606, -0.0182, ..., -0.2885, -0.1996, 0.0215]], + device='cuda:0'), grad: tensor([[ 3.0559e-10, 1.0914e-10, 0.0000e+00, ..., 2.2555e-10, + 1.7462e-10, 3.9290e-10], + [-5.5079e-09, -3.6307e-09, 0.0000e+00, ..., 7.2032e-10, + -1.0994e-08, -1.2660e-09], + [ 1.4552e-10, 1.8699e-09, 0.0000e+00, ..., -4.5839e-09, + 2.0154e-09, -9.0222e-10], + ..., + [ 3.1141e-09, -1.7535e-09, 0.0000e+00, ..., 8.4401e-10, + 1.8481e-09, -1.0914e-10], + [ 9.5315e-10, 1.1860e-09, 0.0000e+00, ..., 2.9468e-09, + 2.3501e-09, 2.2774e-09], + [-7.6761e-09, 1.3315e-09, 0.0000e+00, ..., 2.3283e-10, + 1.3752e-09, -1.3752e-08]], device='cuda:0') +Epoch 477, bias, value: tensor([-0.0423, 0.0129, 0.0126, 0.0285, 0.0360, -0.0198, 0.0287, 0.0303, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([ 3.9581e-09, -7.3924e-09, 3.3106e-09, -4.8894e-08, 5.9372e-08, + 1.9354e-08, 2.9977e-09, 7.0213e-09, 1.2937e-08, -3.3324e-08], + device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 476---------------------------------------------------- +epoch 476, time 252.88, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4301 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.27 lr 0.00001000 +Epoch 478, weight, value: tensor([[ 0.0902, -0.1874, -0.1459, ..., -0.1439, -0.1751, -0.1701], + [ 0.0555, 0.0569, -0.0331, ..., -0.0854, 0.1769, 0.0142], + [-0.0772, -0.2043, -0.0080, ..., 0.2299, -0.2037, 0.0432], + ..., + [-0.2365, 0.1588, 0.0229, ..., -0.1849, 0.1263, 0.0688], + [ 0.0026, -0.0959, -0.0050, ..., -0.1728, -0.2064, -0.3129], + [ 0.0366, -0.0606, -0.0182, ..., -0.2886, -0.1997, 0.0216]], + device='cuda:0'), grad: tensor([[-1.1714e-09, 4.2201e-10, 0.0000e+00, ..., 1.3824e-10, + 4.7294e-10, 5.6752e-10], + [-1.4974e-08, -1.3482e-08, 0.0000e+00, ..., 1.5716e-09, + -3.9290e-08, -7.5306e-09], + [ 1.2005e-09, 3.5070e-09, 0.0000e+00, ..., -1.2224e-09, + 3.7762e-09, 1.8481e-09], + ..., + [ 1.0870e-08, -2.4243e-08, 0.0000e+00, ..., 1.0768e-09, + -5.3187e-09, -1.3060e-08], + [-1.1933e-08, 3.7544e-09, 0.0000e+00, ..., 2.3574e-09, + 7.5161e-09, 3.3760e-09], + [ 1.2296e-09, 1.7040e-08, 0.0000e+00, ..., 5.2605e-09, + 1.5032e-08, 8.1709e-09]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0423, 0.0130, 0.0126, 0.0284, 0.0360, -0.0198, 0.0287, 0.0302, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([-4.5693e-09, -4.6828e-08, 1.0368e-08, 2.5815e-08, 8.4474e-09, + 1.7724e-08, -8.0981e-09, -3.5361e-08, -2.3341e-08, 5.0757e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 252.79, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4196 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.25 lr 0.00001000 +Epoch 479, weight, value: tensor([[ 0.0903, -0.1874, -0.1459, ..., -0.1439, -0.1751, -0.1701], + [ 0.0556, 0.0569, -0.0331, ..., -0.0854, 0.1770, 0.0142], + [-0.0774, -0.2044, -0.0080, ..., 0.2300, -0.2038, 0.0432], + ..., + [-0.2367, 0.1589, 0.0229, ..., -0.1851, 0.1263, 0.0688], + [ 0.0027, -0.0959, -0.0050, ..., -0.1729, -0.2065, -0.3130], + [ 0.0367, -0.0607, -0.0182, ..., -0.2889, -0.1999, 0.0216]], + device='cuda:0'), grad: tensor([[-4.8167e-08, 5.0932e-11, 0.0000e+00, ..., -4.7294e-08, + 1.0914e-10, 4.1473e-10], + [-1.1350e-09, 2.8740e-09, 0.0000e+00, ..., 1.5236e-08, + 5.8208e-10, 1.7419e-08], + [ 4.2171e-08, 8.1491e-10, 0.0000e+00, ..., 4.2608e-08, + 1.0259e-09, 1.4043e-09], + ..., + [ 1.8335e-09, -2.0955e-09, 0.0000e+00, ..., 7.2905e-09, + 7.2032e-10, 6.4247e-09], + [ 1.7171e-09, 5.9663e-10, 0.0000e+00, ..., 1.0259e-09, + 1.2296e-09, 6.8394e-10], + [ 7.2032e-10, 8.7311e-10, 0.0000e+00, ..., 7.7562e-09, + 1.2733e-09, 7.8289e-09]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0422, 0.0130, 0.0125, 0.0284, 0.0360, -0.0198, 0.0286, 0.0302, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([-1.8417e-07, 6.1176e-08, 1.6787e-07, -1.9762e-08, -1.1403e-07, + 7.0431e-09, 2.2410e-08, 3.0268e-08, 2.0707e-08, 3.1810e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 252.31, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3954 re_mapping 0.0015 re_causal 0.0065 /// teacc 99.20 lr 0.00001000 +Epoch 480, weight, value: tensor([[ 0.0904, -0.1874, -0.1459, ..., -0.1439, -0.1751, -0.1702], + [ 0.0557, 0.0569, -0.0331, ..., -0.0856, 0.1771, 0.0141], + [-0.0776, -0.2046, -0.0080, ..., 0.2302, -0.2039, 0.0433], + ..., + [-0.2369, 0.1590, 0.0228, ..., -0.1853, 0.1263, 0.0688], + [ 0.0027, -0.0960, -0.0050, ..., -0.1729, -0.2065, -0.3131], + [ 0.0366, -0.0607, -0.0182, ..., -0.2890, -0.2000, 0.0217]], + device='cuda:0'), grad: tensor([[ 1.3097e-10, 2.6193e-10, 0.0000e+00, ..., 6.1118e-10, + 1.5280e-10, 1.6225e-09], + [-1.4392e-08, 2.1246e-09, 0.0000e+00, ..., 1.8044e-09, + -2.3851e-08, 1.0157e-08], + [ 3.2014e-10, 2.8740e-09, 0.0000e+00, ..., -1.2486e-08, + 3.7107e-09, -2.5975e-09], + ..., + [ 6.8030e-09, 1.9267e-08, 0.0000e+00, ..., 6.0536e-09, + -1.0848e-08, 2.2980e-07], + [ 1.0914e-09, 1.9863e-09, 0.0000e+00, ..., 1.7899e-09, + 2.1028e-09, 8.5274e-09], + [ 3.2014e-10, -5.4395e-08, 0.0000e+00, ..., 5.1659e-10, + 1.0077e-08, -3.7998e-07]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0422, 0.0129, 0.0125, 0.0284, 0.0359, -0.0197, 0.0286, 0.0301, + -0.0186, 0.0004], device='cuda:0'), grad: tensor([ 6.6066e-09, -1.3235e-08, -1.6502e-08, 4.0862e-08, 4.2631e-07, + -1.0739e-08, 3.4852e-09, 7.9582e-07, 3.3702e-08, -1.2498e-06], + device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 252.47, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3945 re_mapping 0.0015 re_causal 0.0064 /// teacc 99.18 lr 0.00001000 +Epoch 481, weight, value: tensor([[ 0.0905, -0.1874, -0.1459, ..., -0.1439, -0.1751, -0.1702], + [ 0.0558, 0.0569, -0.0331, ..., -0.0857, 0.1772, 0.0140], + [-0.0777, -0.2047, -0.0080, ..., 0.2304, -0.2040, 0.0434], + ..., + [-0.2371, 0.1591, 0.0228, ..., -0.1855, 0.1264, 0.0689], + [ 0.0028, -0.0960, -0.0050, ..., -0.1730, -0.2066, -0.3132], + [ 0.0366, -0.0608, -0.0182, ..., -0.2893, -0.2002, 0.0216]], + device='cuda:0'), grad: tensor([[-8.1083e-08, 1.7608e-09, 0.0000e+00, ..., 3.6380e-11, + 6.1118e-10, 1.2296e-09], + [-3.2713e-08, -3.1607e-08, 0.0000e+00, ..., 7.2760e-12, + -5.0990e-08, -2.1129e-08], + [ 7.5379e-09, 1.6080e-09, 0.0000e+00, ..., 7.2760e-12, + 2.9759e-09, 1.0914e-09], + ..., + [ 6.9049e-09, 4.2419e-09, 0.0000e+00, ..., 7.2760e-12, + 7.2687e-09, 3.0341e-09], + [ 1.6851e-08, 7.6761e-09, 0.0000e+00, ..., 1.4552e-11, + 1.2886e-08, 5.1223e-09], + [ 6.0536e-08, 4.3001e-09, 0.0000e+00, ..., 0.0000e+00, + 7.1741e-09, 2.9104e-09]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0421, 0.0129, 0.0126, 0.0283, 0.0360, -0.0197, 0.0286, 0.0301, + -0.0185, 0.0004], device='cuda:0'), grad: tensor([-5.2806e-07, -1.1572e-07, 4.6712e-08, 3.8068e-08, 5.3871e-08, + 1.0041e-09, 3.4604e-08, 3.2451e-08, 8.6206e-08, 3.7486e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 252.42, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4204 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.18 lr 0.00001000 +Epoch 482, weight, value: tensor([[ 0.0906, -0.1874, -0.1459, ..., -0.1440, -0.1751, -0.1702], + [ 0.0559, 0.0569, -0.0331, ..., -0.0859, 0.1773, 0.0140], + [-0.0778, -0.2049, -0.0079, ..., 0.2308, -0.2042, 0.0436], + ..., + [-0.2372, 0.1592, 0.0228, ..., -0.1857, 0.1264, 0.0689], + [ 0.0029, -0.0960, -0.0050, ..., -0.1731, -0.2066, -0.3133], + [ 0.0366, -0.0609, -0.0182, ..., -0.2895, -0.2004, 0.0216]], + device='cuda:0'), grad: tensor([[-3.9698e-08, 1.6735e-10, 0.0000e+00, ..., 7.5670e-10, + 2.7649e-10, 2.6193e-10], + [-4.6493e-09, -4.1837e-09, 0.0000e+00, ..., 2.0736e-09, + -9.2623e-09, -1.2005e-09], + [ 3.9436e-09, 1.9718e-09, 0.0000e+00, ..., -2.1915e-08, + 2.6484e-09, -1.2878e-08], + ..., + [ 3.4997e-09, -2.6048e-09, 0.0000e+00, ..., 1.5047e-08, + 0.0000e+00, 8.1709e-09], + [ 4.9331e-09, 1.4552e-09, 0.0000e+00, ..., 6.9849e-10, + 2.6193e-09, 1.3315e-09], + [ 2.1741e-08, 1.7026e-09, 0.0000e+00, ..., 3.5652e-10, + 2.2119e-09, 1.4843e-09]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0421, 0.0128, 0.0128, 0.0281, 0.0360, -0.0198, 0.0285, 0.0300, + -0.0184, 0.0003], device='cuda:0'), grad: tensor([-2.0128e-07, -7.8362e-09, -2.9220e-08, 1.1601e-07, 1.9354e-08, + -1.9080e-07, 1.1030e-07, 3.9843e-08, 3.2276e-08, 1.2992e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 252.43, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4201 re_mapping 0.0015 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +Epoch 483, weight, value: tensor([[ 0.0906, -0.1874, -0.1459, ..., -0.1440, -0.1751, -0.1702], + [ 0.0561, 0.0570, -0.0331, ..., -0.0860, 0.1775, 0.0140], + [-0.0780, -0.2051, -0.0079, ..., 0.2309, -0.2044, 0.0436], + ..., + [-0.2374, 0.1593, 0.0228, ..., -0.1859, 0.1264, 0.0689], + [ 0.0029, -0.0962, -0.0050, ..., -0.1732, -0.2068, -0.3135], + [ 0.0366, -0.0610, -0.0182, ..., -0.2897, -0.2005, 0.0216]], + device='cuda:0'), grad: tensor([[ 1.0186e-10, 1.4552e-11, 0.0000e+00, ..., 4.2201e-10, + 2.9104e-11, 3.7835e-10], + [-1.2806e-09, -2.9104e-09, 0.0000e+00, ..., 1.6284e-08, + -6.1700e-09, 1.1802e-08], + [ 1.3097e-10, 7.8580e-10, 0.0000e+00, ..., 2.2555e-09, + 5.2387e-10, 2.4156e-09], + ..., + [ 1.3679e-09, 2.5029e-09, 0.0000e+00, ..., 7.3924e-09, + 4.6712e-09, 8.0472e-09], + [ 1.9063e-09, 2.9104e-10, 0.0000e+00, ..., 8.1491e-10, + 5.0932e-10, 8.5856e-10], + [ 5.2387e-10, 5.0932e-10, 0.0000e+00, ..., 1.2558e-08, + 4.0745e-10, 1.0186e-08]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0420, 0.0129, 0.0128, 0.0281, 0.0359, -0.0198, 0.0285, 0.0300, + -0.0184, 0.0003], device='cuda:0'), grad: tensor([ 2.2410e-09, 5.3173e-08, 1.0754e-08, 1.5891e-08, -1.4959e-07, + -2.7008e-08, 8.1345e-09, 3.7195e-08, 1.5076e-08, 5.0291e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 252.53, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3964 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.20 lr 0.00001000 +Epoch 484, weight, value: tensor([[ 0.0907, -0.1874, -0.1459, ..., -0.1441, -0.1751, -0.1702], + [ 0.0563, 0.0571, -0.0331, ..., -0.0861, 0.1776, 0.0141], + [-0.0781, -0.2052, -0.0079, ..., 0.2313, -0.2045, 0.0438], + ..., + [-0.2376, 0.1594, 0.0228, ..., -0.1862, 0.1264, 0.0689], + [ 0.0029, -0.0962, -0.0050, ..., -0.1733, -0.2069, -0.3136], + [ 0.0366, -0.0611, -0.0182, ..., -0.2899, -0.2007, 0.0215]], + device='cuda:0'), grad: tensor([[ 8.8767e-10, 5.5297e-10, 0.0000e+00, ..., 2.7649e-10, + 7.2760e-10, 7.5670e-10], + [-5.7771e-09, -5.9081e-09, 0.0000e+00, ..., 1.6880e-09, + -1.8350e-08, -3.3906e-09], + [ 5.1659e-09, 2.9540e-09, 0.0000e+00, ..., -1.3388e-09, + 1.5716e-09, -9.8953e-10], + ..., + [ 5.0495e-09, 3.5507e-09, 0.0000e+00, ..., 5.6752e-10, + 6.3592e-09, 3.6380e-09], + [-9.6479e-09, -3.8126e-09, 0.0000e+00, ..., -1.7608e-09, + 9.0949e-09, 3.3760e-09], + [ 5.0932e-10, 1.6589e-09, 0.0000e+00, ..., 1.2806e-09, + 2.1828e-10, -1.5745e-08]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0420, 0.0130, 0.0130, 0.0279, 0.0359, -0.0198, 0.0285, 0.0299, + -0.0185, 0.0002], device='cuda:0'), grad: tensor([ 6.9995e-09, 1.1278e-08, 2.9089e-08, -1.1220e-08, 5.1077e-08, + 1.2689e-08, 1.0390e-08, 2.9351e-08, -7.8406e-08, -5.3318e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 252.73, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3967 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.18 lr 0.00001000 +Epoch 485, weight, value: tensor([[ 0.0908, -0.1874, -0.1459, ..., -0.1441, -0.1751, -0.1702], + [ 0.0563, 0.0571, -0.0331, ..., -0.0862, 0.1777, 0.0140], + [-0.0782, -0.2054, -0.0079, ..., 0.2315, -0.2046, 0.0439], + ..., + [-0.2377, 0.1595, 0.0228, ..., -0.1863, 0.1265, 0.0690], + [ 0.0029, -0.0963, -0.0050, ..., -0.1734, -0.2069, -0.3137], + [ 0.0366, -0.0611, -0.0182, ..., -0.2901, -0.2009, 0.0215]], + device='cuda:0'), grad: tensor([[-5.7189e-09, -1.0332e-09, 0.0000e+00, ..., 4.2201e-10, + -1.4697e-09, 5.9663e-10], + [-8.1491e-10, 8.6293e-09, 0.0000e+00, ..., 1.0768e-09, + 2.9831e-09, 1.1729e-08], + [ 1.6444e-09, 1.1496e-09, 0.0000e+00, ..., -1.0681e-08, + 1.3242e-09, -5.4133e-09], + ..., + [ 3.8854e-09, -1.5687e-08, 0.0000e+00, ..., 3.3760e-09, + -9.1532e-09, -1.3490e-08], + [-3.8126e-09, 7.2760e-10, 0.0000e+00, ..., 4.9477e-10, + 1.1642e-09, 1.3242e-09], + [-2.7503e-09, 3.2742e-09, 0.0000e+00, ..., 2.3283e-10, + 2.8085e-09, -1.5469e-08]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0419, 0.0129, 0.0131, 0.0278, 0.0359, -0.0198, 0.0285, 0.0300, + -0.0185, 0.0002], device='cuda:0'), grad: tensor([-3.1694e-08, 3.2072e-08, -1.3897e-08, 2.6994e-08, 6.6182e-08, + 2.8522e-09, 5.8644e-09, -9.8516e-09, -2.5146e-08, -5.1368e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 252.69, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4001 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.24 lr 0.00001000 +Epoch 486, weight, value: tensor([[ 0.0908, -0.1874, -0.1459, ..., -0.1442, -0.1751, -0.1703], + [ 0.0564, 0.0571, -0.0331, ..., -0.0863, 0.1778, 0.0139], + [-0.0783, -0.2055, -0.0079, ..., 0.2317, -0.2047, 0.0440], + ..., + [-0.2378, 0.1596, 0.0228, ..., -0.1865, 0.1265, 0.0690], + [ 0.0030, -0.0963, -0.0050, ..., -0.1734, -0.2070, -0.3138], + [ 0.0367, -0.0612, -0.0182, ..., -0.2903, -0.2010, 0.0215]], + device='cuda:0'), grad: tensor([[ 6.1555e-09, 1.4552e-11, 0.0000e+00, ..., 7.2032e-09, + 1.4552e-11, 4.3656e-10], + [ 1.1059e-09, 2.4156e-09, 0.0000e+00, ..., 8.2509e-09, + 1.9791e-09, 4.2637e-09], + [ 7.1304e-10, 1.4406e-09, 0.0000e+00, ..., -7.4069e-09, + 1.2951e-09, -5.4424e-09], + ..., + [ 4.0745e-10, -1.9063e-09, 0.0000e+00, ..., 4.5547e-09, + -1.2806e-09, 1.7899e-09], + [-1.1030e-08, 1.1642e-10, 0.0000e+00, ..., 4.9477e-10, + 1.8917e-10, 4.6566e-10], + [ 2.2555e-09, 7.4215e-10, 0.0000e+00, ..., 7.7562e-09, + 6.8394e-10, 8.6584e-09]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0419, 0.0129, 0.0132, 0.0277, 0.0359, -0.0197, 0.0285, 0.0299, + -0.0184, 0.0002], device='cuda:0'), grad: tensor([ 4.3481e-08, 3.5798e-08, -1.1845e-08, 1.1729e-08, -6.1409e-09, + 4.8382e-07, -5.4855e-07, 1.3097e-08, -4.7818e-08, 4.1677e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 252.96, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4063 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.19 lr 0.00001000 +Epoch 487, weight, value: tensor([[ 0.0909, -0.1874, -0.1459, ..., -0.1443, -0.1751, -0.1703], + [ 0.0567, 0.0572, -0.0330, ..., -0.0864, 0.1780, 0.0140], + [-0.0785, -0.2058, -0.0079, ..., 0.2318, -0.2049, 0.0441], + ..., + [-0.2381, 0.1596, 0.0228, ..., -0.1867, 0.1264, 0.0690], + [ 0.0030, -0.0964, -0.0050, ..., -0.1735, -0.2071, -0.3140], + [ 0.0367, -0.0612, -0.0182, ..., -0.2905, -0.2012, 0.0215]], + device='cuda:0'), grad: tensor([[-1.7317e-09, 3.2014e-10, 0.0000e+00, ..., 3.0559e-10, + 2.7649e-10, 5.5297e-10], + [-1.8626e-09, 1.7404e-08, 0.0000e+00, ..., 2.5757e-09, + 1.1452e-08, 1.7651e-08], + [ 4.3656e-10, 6.2282e-09, 0.0000e+00, ..., -4.6566e-09, + 5.3697e-09, 2.1973e-09], + ..., + [ 1.1205e-09, -4.3015e-08, 0.0000e+00, ..., 1.5425e-09, + -3.3353e-08, -3.6438e-08], + [ 1.7899e-09, 2.1828e-09, 0.0000e+00, ..., 2.4156e-09, + 3.1287e-09, 3.2160e-09], + [-5.2387e-10, 9.4442e-09, 0.0000e+00, ..., 7.5670e-10, + 7.5961e-09, 5.0641e-09]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0419, 0.0130, 0.0131, 0.0278, 0.0358, -0.0198, 0.0285, 0.0298, + -0.0185, 0.0002], device='cuda:0'), grad: tensor([-8.4256e-09, 5.0757e-08, 9.4005e-09, 1.1816e-08, 2.0576e-08, + 3.6380e-09, -1.4115e-09, -1.0553e-07, 1.6516e-08, 1.9834e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 252.77, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4207 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.20 lr 0.00001000 +Epoch 488, weight, value: tensor([[ 0.0910, -0.1875, -0.1459, ..., -0.1443, -0.1751, -0.1703], + [ 0.0570, 0.0573, -0.0330, ..., -0.0865, 0.1783, 0.0140], + [-0.0788, -0.2060, -0.0079, ..., 0.2320, -0.2052, 0.0441], + ..., + [-0.2383, 0.1598, 0.0228, ..., -0.1869, 0.1265, 0.0691], + [ 0.0029, -0.0966, -0.0050, ..., -0.1736, -0.2073, -0.3142], + [ 0.0367, -0.0613, -0.0182, ..., -0.2906, -0.2015, 0.0216]], + device='cuda:0'), grad: tensor([[-9.8953e-10, 1.1642e-10, 0.0000e+00, ..., 1.3097e-10, + 2.9104e-10, 7.5670e-10], + [ 2.8813e-09, -5.2096e-09, 0.0000e+00, ..., 7.5670e-10, + -1.3693e-08, -2.2847e-09], + [ 1.8757e-08, 7.7125e-10, 0.0000e+00, ..., 2.7649e-10, + 2.4302e-09, 1.8626e-09], + ..., + [ 6.3737e-09, 1.8335e-09, 0.0000e+00, ..., 8.8767e-10, + 4.0891e-09, 5.2532e-09], + [-2.8638e-08, 1.1350e-09, 0.0000e+00, ..., 3.4925e-10, + 3.1141e-09, 5.1223e-09], + [-9.5170e-09, 4.6566e-10, 0.0000e+00, ..., 2.4738e-09, + 1.6153e-09, -1.7957e-08]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0418, 0.0130, 0.0131, 0.0277, 0.0358, -0.0199, 0.0285, 0.0298, + -0.0186, 0.0002], device='cuda:0'), grad: tensor([-5.2823e-09, 3.1170e-08, 8.1898e-08, 6.9034e-08, 6.6706e-08, + -1.2526e-07, 3.8359e-08, 3.4575e-08, -1.1217e-07, -6.0885e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 252.70, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.15 lr 0.00001000 +Epoch 489, weight, value: tensor([[ 0.0911, -0.1875, -0.1459, ..., -0.1443, -0.1751, -0.1703], + [ 0.0572, 0.0574, -0.0330, ..., -0.0866, 0.1785, 0.0140], + [-0.0790, -0.2064, -0.0079, ..., 0.2322, -0.2055, 0.0441], + ..., + [-0.2385, 0.1600, 0.0228, ..., -0.1871, 0.1267, 0.0693], + [ 0.0028, -0.0968, -0.0050, ..., -0.1737, -0.2075, -0.3145], + [ 0.0367, -0.0615, -0.0182, ..., -0.2908, -0.2018, 0.0217]], + device='cuda:0'), grad: tensor([[-7.2760e-11, 1.6007e-10, 0.0000e+00, ..., 2.3283e-10, + 1.7462e-10, 1.2660e-09], + [-3.7980e-09, 5.5588e-09, 0.0000e+00, ..., 5.3842e-09, + 1.0623e-09, 1.0565e-08], + [ 7.1304e-10, 1.1161e-08, 0.0000e+00, ..., -5.5734e-09, + 1.1307e-08, 8.0181e-09], + ..., + [ 3.2596e-09, -6.5600e-08, 0.0000e+00, ..., 1.6153e-09, + -6.1525e-08, -6.4203e-08], + [ 1.7753e-09, 3.5943e-09, 0.0000e+00, ..., 4.2201e-10, + 4.6566e-09, 3.6089e-09], + [-4.1327e-09, 2.1013e-08, 0.0000e+00, ..., 4.9622e-09, + 2.1319e-08, 1.5498e-08]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0417, 0.0131, 0.0130, 0.0276, 0.0357, -0.0200, 0.0284, 0.0299, + -0.0187, 0.0002], device='cuda:0'), grad: tensor([ 2.1828e-09, 3.0268e-08, 2.5437e-08, 1.8685e-08, 4.1648e-08, + 4.5111e-09, 6.1118e-09, -1.7392e-07, 1.5309e-08, 4.6624e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 252.50, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4075 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.26 lr 0.00001000 +Epoch 490, weight, value: tensor([[ 0.0912, -0.1875, -0.1459, ..., -0.1443, -0.1750, -0.1703], + [ 0.0572, 0.0573, -0.0330, ..., -0.0868, 0.1785, 0.0139], + [-0.0792, -0.2067, -0.0079, ..., 0.2323, -0.2058, 0.0442], + ..., + [-0.2386, 0.1603, 0.0228, ..., -0.1872, 0.1269, 0.0695], + [ 0.0028, -0.0969, -0.0050, ..., -0.1738, -0.2076, -0.3146], + [ 0.0367, -0.0617, -0.0182, ..., -0.2910, -0.2020, 0.0217]], + device='cuda:0'), grad: tensor([[ 8.7311e-11, 7.5670e-10, 0.0000e+00, ..., 1.6007e-10, + 1.1496e-09, 8.0036e-10], + [-1.3941e-08, -6.7666e-09, 0.0000e+00, ..., 6.1118e-10, + -1.7419e-08, -3.6234e-09], + [ 2.5757e-09, 3.5070e-09, 0.0000e+00, ..., 2.7358e-09, + 4.5839e-09, 6.1991e-09], + ..., + [ 4.0309e-09, -9.1532e-09, 0.0000e+00, ..., 5.4715e-09, + -4.9185e-09, 1.0768e-09], + [ 2.1828e-09, 2.8231e-09, 0.0000e+00, ..., 5.6752e-10, + 4.5839e-09, 3.0268e-09], + [-6.5629e-09, 1.0768e-09, 0.0000e+00, ..., 1.6735e-09, + 4.0891e-09, -1.5862e-08]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0417, 0.0130, 0.0130, 0.0277, 0.0356, -0.0201, 0.0284, 0.0300, + -0.0188, 0.0002], device='cuda:0'), grad: tensor([ 2.5611e-09, -3.6467e-08, 9.4529e-08, -1.2910e-07, 5.7189e-08, + 2.0897e-08, 4.6130e-09, 1.6968e-08, 2.6019e-08, -4.7032e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 252.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4272 re_mapping 0.0015 re_causal 0.0067 /// teacc 99.25 lr 0.00001000 +Epoch 491, weight, value: tensor([[ 0.0912, -0.1875, -0.1459, ..., -0.1444, -0.1751, -0.1704], + [ 0.0574, 0.0574, -0.0330, ..., -0.0869, 0.1787, 0.0139], + [-0.0793, -0.2069, -0.0079, ..., 0.2324, -0.2059, 0.0442], + ..., + [-0.2388, 0.1603, 0.0228, ..., -0.1873, 0.1268, 0.0695], + [ 0.0029, -0.0969, -0.0050, ..., -0.1739, -0.2076, -0.3147], + [ 0.0367, -0.0617, -0.0182, ..., -0.2912, -0.2021, 0.0216]], + device='cuda:0'), grad: tensor([[ 6.5484e-10, 1.0186e-09, 0.0000e+00, ..., 4.6566e-09, + 1.2078e-09, 7.1304e-10], + [-4.0093e-07, -6.5798e-07, 0.0000e+00, ..., 8.5856e-09, + -7.4226e-07, -2.7893e-07], + [ 6.1555e-09, 1.1656e-08, 0.0000e+00, ..., 7.2760e-11, + 1.3417e-08, 6.2719e-09], + ..., + [ 3.5949e-07, 5.8487e-07, 0.0000e+00, ..., 2.2410e-09, + 6.5891e-07, 2.5239e-07], + [ 2.1435e-08, 3.4895e-08, 0.0000e+00, ..., 4.8603e-09, + 3.9727e-08, 1.5672e-08], + [ 9.1241e-09, 2.0140e-08, 0.0000e+00, ..., 2.6048e-08, + 1.9572e-08, 3.1607e-08]], device='cuda:0') +Epoch 491, bias, value: tensor([-4.1684e-02, 1.3043e-02, 1.2947e-02, 2.7699e-02, 3.5755e-02, + -2.0227e-02, 2.8542e-02, 2.9953e-02, -1.8824e-02, 9.2140e-05], + device='cuda:0'), grad: tensor([ 3.0297e-08, -1.3458e-06, 3.5856e-08, 1.1671e-08, -9.7498e-08, + 7.9861e-08, -1.7823e-07, 1.2293e-06, 1.0541e-07, 1.4238e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 490, time 252.45, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4141 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.24 lr 0.00001000 +Epoch 492, weight, value: tensor([[ 0.0914, -0.1875, -0.1459, ..., -0.1444, -0.1751, -0.1704], + [ 0.0577, 0.0576, -0.0330, ..., -0.0871, 0.1790, 0.0141], + [-0.0794, -0.2070, -0.0079, ..., 0.2326, -0.2061, 0.0443], + ..., + [-0.2392, 0.1603, 0.0228, ..., -0.1875, 0.1266, 0.0694], + [ 0.0028, -0.0970, -0.0050, ..., -0.1740, -0.2077, -0.3149], + [ 0.0367, -0.0618, -0.0182, ..., -0.2914, -0.2022, 0.0215]], + device='cuda:0'), grad: tensor([[-1.5309e-08, -2.9104e-11, 0.0000e+00, ..., 4.2201e-10, + -1.4552e-11, 2.6193e-10], + [ 1.4115e-09, 5.3842e-10, 0.0000e+00, ..., 6.8394e-10, + 3.2014e-10, 1.5862e-09], + [-1.8481e-09, 1.0186e-10, 0.0000e+00, ..., -9.0658e-09, + 4.8021e-10, -6.3446e-09], + ..., + [ 1.4115e-09, 5.0932e-10, 0.0000e+00, ..., 1.2806e-09, + 5.9663e-10, 1.3388e-09], + [ 7.1304e-09, 3.4925e-10, 0.0000e+00, ..., 4.7876e-09, + 5.6752e-10, 3.4925e-09], + [ 2.4156e-09, 2.3283e-10, 0.0000e+00, ..., 3.7835e-10, + 2.3283e-10, -4.6712e-09]], device='cuda:0') +Epoch 492, bias, value: tensor([-4.1613e-02, 1.3152e-02, 1.2967e-02, 2.7681e-02, 3.5836e-02, + -2.0348e-02, 2.8601e-02, 2.9800e-02, -1.8924e-02, 1.0279e-05], + device='cuda:0'), grad: tensor([-8.3819e-08, 9.9244e-09, -1.9660e-08, 1.3722e-08, 2.4142e-08, + 3.2014e-09, 1.9063e-09, 9.1532e-09, 4.3132e-08, 1.0390e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 252.89, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4054 re_mapping 0.0014 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +Epoch 493, weight, value: tensor([[ 0.0915, -0.1875, -0.1459, ..., -0.1444, -0.1750, -0.1704], + [ 0.0578, 0.0577, -0.0330, ..., -0.0872, 0.1790, 0.0140], + [-0.0796, -0.2072, -0.0079, ..., 0.2328, -0.2062, 0.0443], + ..., + [-0.2393, 0.1603, 0.0228, ..., -0.1876, 0.1267, 0.0694], + [ 0.0028, -0.0971, -0.0050, ..., -0.1741, -0.2078, -0.3150], + [ 0.0368, -0.0618, -0.0182, ..., -0.2915, -0.2023, 0.0216]], + device='cuda:0'), grad: tensor([[ 7.2760e-11, 5.8208e-11, 0.0000e+00, ..., 7.7125e-10, + 8.7311e-11, 4.3656e-10], + [-1.6589e-09, 3.9290e-10, 0.0000e+00, ..., 3.3615e-09, + -2.1537e-09, 2.5611e-09], + [ 4.2201e-10, 4.2346e-09, 0.0000e+00, ..., -9.8516e-09, + 3.0559e-09, -1.0623e-09], + ..., + [ 1.3679e-09, -4.9622e-09, 0.0000e+00, ..., 3.1869e-09, + -2.0518e-09, -3.0996e-09], + [ 1.7899e-09, 5.9663e-10, 0.0000e+00, ..., 1.0768e-09, + 1.3388e-09, 1.0914e-09], + [ 3.6380e-10, 1.1059e-09, 0.0000e+00, ..., 9.0222e-10, + 9.7498e-10, 2.0082e-09]], device='cuda:0') +Epoch 493, bias, value: tensor([-4.1548e-02, 1.3092e-02, 1.3009e-02, 2.7685e-02, 3.5794e-02, + -2.0430e-02, 2.8643e-02, 2.9746e-02, -1.9004e-02, 3.3065e-05], + device='cuda:0'), grad: tensor([ 2.2410e-09, 8.0327e-09, -1.2078e-08, 5.5006e-09, 2.9831e-09, + -5.0932e-09, 2.0955e-09, -8.0036e-10, 9.5461e-09, 9.4005e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 252.78, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4005 re_mapping 0.0014 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +Epoch 494, weight, value: tensor([[ 0.0915, -0.1875, -0.1459, ..., -0.1444, -0.1750, -0.1704], + [ 0.0579, 0.0577, -0.0330, ..., -0.0874, 0.1791, 0.0139], + [-0.0797, -0.2075, -0.0079, ..., 0.2330, -0.2064, 0.0444], + ..., + [-0.2394, 0.1605, 0.0228, ..., -0.1878, 0.1268, 0.0696], + [ 0.0027, -0.0972, -0.0050, ..., -0.1742, -0.2079, -0.3151], + [ 0.0368, -0.0618, -0.0182, ..., -0.2917, -0.2024, 0.0217]], + device='cuda:0'), grad: tensor([[-6.1846e-09, -1.0987e-08, 0.0000e+00, ..., 1.6007e-10, + -1.3955e-08, 2.1391e-09], + [-3.2014e-09, 1.8175e-08, 0.0000e+00, ..., 1.8481e-09, + 1.1554e-08, 1.7200e-08], + [ 3.5652e-09, 3.9872e-09, 0.0000e+00, ..., -2.8085e-09, + 5.3697e-09, 2.3865e-09], + ..., + [ 5.3842e-09, -2.2585e-08, 0.0000e+00, ..., 2.3865e-09, + -1.7724e-08, -7.8435e-09], + [ 7.6689e-09, 2.8085e-09, 0.0000e+00, ..., 3.7835e-10, + 4.5984e-09, 1.7171e-09], + [-1.2020e-08, 2.8667e-09, 0.0000e+00, ..., 1.8044e-09, + 3.5507e-09, -5.1135e-08]], device='cuda:0') +Epoch 494, bias, value: tensor([-4.1517e-02, 1.3071e-02, 1.3037e-02, 2.7620e-02, 3.5682e-02, + -2.0490e-02, 2.8700e-02, 2.9791e-02, -1.9121e-02, 7.5768e-05], + device='cuda:0'), grad: tensor([-9.5344e-08, 6.4028e-08, 3.1490e-08, 5.3173e-08, 7.8930e-08, + -4.7788e-08, 1.9601e-08, -1.1496e-09, 4.7061e-08, -1.3283e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 252.72, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4117 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.21 lr 0.00001000 +Epoch 495, weight, value: tensor([[ 0.0916, -0.1875, -0.1459, ..., -0.1445, -0.1750, -0.1705], + [ 0.0580, 0.0576, -0.0330, ..., -0.0876, 0.1791, 0.0138], + [-0.0799, -0.2077, -0.0079, ..., 0.2332, -0.2065, 0.0445], + ..., + [-0.2395, 0.1608, 0.0228, ..., -0.1880, 0.1270, 0.0697], + [ 0.0027, -0.0973, -0.0050, ..., -0.1743, -0.2080, -0.3152], + [ 0.0369, -0.0619, -0.0182, ..., -0.2918, -0.2026, 0.0218]], + device='cuda:0'), grad: tensor([[ 1.4552e-10, 4.5111e-10, 0.0000e+00, ..., 1.1642e-09, + 7.2760e-11, 4.3656e-10], + [ 4.6566e-10, 3.9872e-09, 0.0000e+00, ..., 2.5175e-09, + 2.5029e-09, 3.0122e-09], + [ 3.3469e-10, 1.6007e-09, 0.0000e+00, ..., 9.6043e-10, + 1.3679e-09, 2.1391e-09], + ..., + [ 1.1205e-09, -2.6776e-09, 0.0000e+00, ..., 7.2760e-11, + -1.6007e-09, -1.5571e-09], + [ 1.3533e-09, 2.7067e-09, 0.0000e+00, ..., 5.6461e-09, + 8.8767e-10, 1.1059e-09], + [-2.8813e-09, 1.5280e-09, 0.0000e+00, ..., 2.0518e-09, + 1.3097e-09, 5.0786e-09]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0415, 0.0129, 0.0131, 0.0276, 0.0355, -0.0204, 0.0287, 0.0299, + -0.0192, 0.0001], device='cuda:0'), grad: tensor([ 6.0827e-09, 2.0154e-08, 1.1423e-08, 2.4593e-09, -1.4130e-08, + 6.7230e-09, -4.6246e-08, -1.9936e-09, 2.9191e-08, 1.2500e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 253.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4096 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.22 lr 0.00001000 +Epoch 496, weight, value: tensor([[ 0.0917, -0.1876, -0.1459, ..., -0.1446, -0.1750, -0.1705], + [ 0.0581, 0.0575, -0.0330, ..., -0.0877, 0.1791, 0.0136], + [-0.0801, -0.2079, -0.0079, ..., 0.2334, -0.2068, 0.0445], + ..., + [-0.2396, 0.1611, 0.0228, ..., -0.1881, 0.1272, 0.0700], + [ 0.0028, -0.0973, -0.0050, ..., -0.1744, -0.2080, -0.3154], + [ 0.0369, -0.0620, -0.0182, ..., -0.2920, -0.2028, 0.0217]], + device='cuda:0'), grad: tensor([[ 5.5297e-10, 2.9104e-10, 0.0000e+00, ..., 2.9104e-10, + 2.3283e-10, 4.9477e-10], + [ 4.3656e-09, 7.3051e-09, 0.0000e+00, ..., 6.5018e-08, + 4.5402e-09, 4.5286e-08], + [ 3.9290e-10, 6.0973e-09, 0.0000e+00, ..., -1.0768e-09, + 4.4529e-09, 4.3947e-09], + ..., + [ 7.4215e-10, -2.0576e-08, 0.0000e+00, ..., 4.7730e-09, + -1.4450e-08, -1.3810e-08], + [-1.2689e-08, -2.2410e-09, 0.0000e+00, ..., 1.3533e-09, + 7.4215e-10, 1.4552e-09], + [ 6.2137e-09, 4.3801e-09, 0.0000e+00, ..., 6.0681e-09, + 1.8772e-09, 5.4424e-09]], device='cuda:0') +Epoch 496, bias, value: tensor([-4.1436e-02, 1.2797e-02, 1.3061e-02, 2.7518e-02, 3.5524e-02, + -2.0480e-02, 2.8675e-02, 3.0086e-02, -1.9137e-02, 7.6603e-05], + device='cuda:0'), grad: tensor([ 4.9185e-09, 2.2701e-07, 2.2119e-08, -1.8976e-08, -2.2771e-07, + 1.8248e-08, 1.7651e-08, -3.7951e-08, -3.4110e-08, 4.4616e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 253.14, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4122 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.19 lr 0.00001000 +Epoch 497, weight, value: tensor([[ 0.0918, -0.1876, -0.1459, ..., -0.1446, -0.1750, -0.1705], + [ 0.0582, 0.0574, -0.0330, ..., -0.0878, 0.1792, 0.0135], + [-0.0802, -0.2082, -0.0079, ..., 0.2336, -0.2071, 0.0446], + ..., + [-0.2397, 0.1614, 0.0228, ..., -0.1882, 0.1274, 0.0702], + [ 0.0028, -0.0974, -0.0050, ..., -0.1745, -0.2081, -0.3155], + [ 0.0369, -0.0621, -0.0182, ..., -0.2921, -0.2029, 0.0218]], + device='cuda:0'), grad: tensor([[-1.4770e-08, -1.5425e-09, 0.0000e+00, ..., 2.4738e-10, + -2.0664e-09, 8.1491e-10], + [ 1.1787e-09, 1.3242e-09, 0.0000e+00, ..., 4.3510e-09, + 8.7311e-10, 5.1659e-09], + [ 9.6043e-10, 1.0041e-09, 0.0000e+00, ..., -1.0768e-09, + 8.2946e-10, 5.0932e-10], + ..., + [ 8.8767e-10, -8.4401e-10, 0.0000e+00, ..., 3.7835e-09, + -1.0914e-09, 3.0559e-09], + [ 1.8917e-09, 3.4925e-10, 0.0000e+00, ..., 4.3656e-10, + 4.6566e-10, 4.2201e-10], + [ 3.8563e-09, -1.2224e-09, 0.0000e+00, ..., 2.6776e-09, + 1.3679e-09, -6.6648e-09]], device='cuda:0') +Epoch 497, bias, value: tensor([-4.1369e-02, 1.2703e-02, 1.3064e-02, 2.7511e-02, 3.5482e-02, + -2.0791e-02, 2.8669e-02, 3.0251e-02, -1.9157e-02, 5.2418e-05], + device='cuda:0'), grad: tensor([-8.8010e-08, 2.4826e-08, 1.0579e-08, 3.1723e-09, 2.0664e-09, + 5.7480e-09, 1.9878e-08, 1.8132e-08, 9.6479e-09, 7.2032e-09], + device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 252.49, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3844 re_mapping 0.0014 re_causal 0.0062 /// teacc 99.19 lr 0.00001000 +Epoch 498, weight, value: tensor([[ 0.0919, -0.1876, -0.1459, ..., -0.1447, -0.1750, -0.1705], + [ 0.0582, 0.0573, -0.0330, ..., -0.0880, 0.1792, 0.0133], + [-0.0804, -0.2085, -0.0079, ..., 0.2337, -0.2073, 0.0446], + ..., + [-0.2398, 0.1617, 0.0228, ..., -0.1883, 0.1277, 0.0705], + [ 0.0029, -0.0975, -0.0050, ..., -0.1746, -0.2082, -0.3157], + [ 0.0369, -0.0623, -0.0182, ..., -0.2924, -0.2031, 0.0217]], + device='cuda:0'), grad: tensor([[-2.3138e-09, 5.8208e-11, 0.0000e+00, ..., 1.3097e-10, + 7.2760e-11, 3.2014e-10], + [ 9.0076e-09, 2.9249e-09, 0.0000e+00, ..., 7.2760e-10, + 2.8085e-09, 2.1973e-09], + [ 9.7498e-10, 8.1491e-10, 0.0000e+00, ..., -9.3132e-10, + 8.0036e-10, -7.2760e-11], + ..., + [ 1.8481e-09, 4.6566e-10, 0.0000e+00, ..., 9.0222e-10, + 9.0222e-10, 2.6630e-09], + [-9.8080e-09, -2.5902e-09, 0.0000e+00, ..., 1.8917e-10, + -2.4156e-09, 2.4884e-09], + [-6.0245e-09, 6.9849e-10, 0.0000e+00, ..., 1.6269e-08, + 6.5484e-10, 4.3074e-09]], device='cuda:0') +Epoch 498, bias, value: tensor([-4.1340e-02, 1.2524e-02, 1.3024e-02, 2.7515e-02, 3.5555e-02, + -2.1214e-02, 2.8892e-02, 3.0467e-02, -1.9095e-02, -4.3499e-05], + device='cuda:0'), grad: tensor([-4.8749e-09, 5.0873e-08, 5.7771e-09, -1.0114e-08, -2.8667e-08, + 1.3941e-08, 6.8248e-09, 1.3868e-08, -4.8080e-08, 2.2570e-08], + device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 252.59, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4077 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.21 lr 0.00001000 +Epoch 499, weight, value: tensor([[ 0.0920, -0.1876, -0.1459, ..., -0.1448, -0.1750, -0.1706], + [ 0.0583, 0.0572, -0.0330, ..., -0.0881, 0.1792, 0.0132], + [-0.0805, -0.2087, -0.0079, ..., 0.2338, -0.2076, 0.0446], + ..., + [-0.2399, 0.1619, 0.0228, ..., -0.1884, 0.1278, 0.0707], + [ 0.0031, -0.0975, -0.0050, ..., -0.1747, -0.2082, -0.3158], + [ 0.0369, -0.0623, -0.0182, ..., -0.2926, -0.2033, 0.0216]], + device='cuda:0'), grad: tensor([[-5.3842e-09, -1.1642e-10, 0.0000e+00, ..., 4.9477e-10, + 0.0000e+00, 6.9849e-10], + [ 1.4552e-09, 1.2806e-09, 0.0000e+00, ..., 4.4529e-09, + 2.9104e-10, 8.0909e-09], + [ 8.7311e-10, 3.7835e-10, 0.0000e+00, ..., -2.5902e-09, + 2.6193e-10, -3.5216e-09], + ..., + [ 2.1537e-09, 2.9104e-10, 0.0000e+00, ..., 1.7171e-09, + 0.0000e+00, 5.7335e-09], + [ 1.2515e-09, 2.9104e-11, 0.0000e+00, ..., 9.0222e-10, + 0.0000e+00, 1.2806e-09], + [-4.9185e-08, 1.4843e-09, 0.0000e+00, ..., 6.3737e-09, + 2.0373e-10, -7.6368e-08]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0413, 0.0124, 0.0130, 0.0275, 0.0356, -0.0213, 0.0289, 0.0306, + -0.0190, -0.0001], device='cuda:0'), grad: tensor([-2.5728e-08, 3.1665e-08, -4.1910e-09, 7.1013e-09, 2.4680e-07, + 4.5402e-09, 8.1782e-09, 2.3720e-08, 9.0222e-09, -2.7963e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 252.83, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4173 re_mapping 0.0014 re_causal 0.0065 /// teacc 99.20 lr 0.00001000 +Epoch 500, weight, value: tensor([[ 0.0921, -0.1876, -0.1459, ..., -0.1448, -0.1750, -0.1706], + [ 0.0584, 0.0571, -0.0330, ..., -0.0882, 0.1793, 0.0131], + [-0.0807, -0.2090, -0.0079, ..., 0.2339, -0.2078, 0.0446], + ..., + [-0.2400, 0.1623, 0.0228, ..., -0.1885, 0.1281, 0.0710], + [ 0.0031, -0.0977, -0.0050, ..., -0.1747, -0.2084, -0.3160], + [ 0.0369, -0.0626, -0.0182, ..., -0.2927, -0.2036, 0.0216]], + device='cuda:0'), grad: tensor([[-3.7544e-09, 2.9104e-11, 0.0000e+00, ..., 9.3132e-10, + 0.0000e+00, 9.3132e-10], + [ 2.2119e-09, 1.4552e-10, 0.0000e+00, ..., 5.9372e-09, + 8.7311e-11, 6.6066e-09], + [ 2.8522e-09, 2.6193e-10, 0.0000e+00, ..., -2.1362e-08, + 1.1642e-10, -1.7113e-08], + ..., + [ 4.9477e-10, 2.0082e-09, 0.0000e+00, ..., 1.4465e-08, + 6.4028e-10, 1.4930e-08], + [-7.5961e-09, 4.0745e-10, 0.0000e+00, ..., 4.3074e-09, + 2.3283e-10, 2.6484e-09], + [ 3.1723e-09, 1.1642e-10, 0.0000e+00, ..., 3.5798e-08, + 8.7311e-11, 5.0233e-08]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0412, 0.0124, 0.0129, 0.0274, 0.0355, -0.0215, 0.0290, 0.0308, + -0.0190, -0.0002], device='cuda:0'), grad: tensor([-1.4785e-08, 3.1258e-08, -4.0280e-08, -3.5914e-08, -2.0606e-07, + 1.4348e-08, 1.2573e-08, 8.1025e-08, -1.1030e-08, 1.7695e-07], + device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 252.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4244 re_mapping 0.0014 re_causal 0.0064 /// teacc 99.17 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.129997 99.059998 ... 88.938713 72.058903 +ShearY 98.979996 98.959999 ... 88.938713 69.911438 +AutoContrast 99.180000 99.180000 ... 88.938713 62.315996 +Invert 99.059998 97.419998 ... 88.938713 61.798100 +Equalize 98.509995 98.479996 ... 88.938713 72.956973 +Solarize 98.489998 98.540001 ... 88.938713 63.519738 +SolarizeAdd 98.799995 98.599998 ... 88.938713 70.380315 +Posterize 99.119995 99.150002 ... 88.938713 74.407317 +Contrast 99.099998 99.290001 ... 88.938713 69.076321 +Color 98.900002 99.269997 ... 88.938713 56.291151 +Brightness 98.979996 99.220001 ... 88.938713 70.326309 +Sharpness 99.029999 99.169998 ... 88.938713 72.508256 +NoiseSalt 99.010002 99.180000 ... 88.938713 60.808785 +NoiseGaussian 98.930000 99.269997 ... 88.938713 54.688457 +w/o do (original x) 99.270000 0.000000 ... 0.000000 76.284641 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.27 67.739705 78.069103 76.520465 87.643249 77.49313 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps5_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last +randm: False +stride: 3 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.129997 99.139999 ... 89.387146 72.153507 +ShearY 98.959999 99.010002 ... 89.387146 70.025443 +AutoContrast 99.110001 99.110001 ... 89.387146 62.317904 +Invert 99.059998 97.360001 ... 89.387146 63.034497 +Equalize 98.540001 98.479996 ... 89.387146 73.167739 +Solarize 98.500000 98.580002 ... 89.387146 64.381314 +SolarizeAdd 98.820000 98.680000 ... 89.387146 70.767879 +Posterize 99.110001 99.129997 ... 89.387146 74.491499 +Contrast 99.070000 99.229996 ... 89.387146 68.858720 +Color 98.940002 99.169998 ... 89.387146 55.757759 +Brightness 99.000000 99.169998 ... 89.387146 69.960325 +Sharpness 99.089996 99.169998 ... 89.387146 72.177784 +NoiseSalt 98.970001 99.150002 ... 89.387146 60.591439 +NoiseGaussian 99.019997 99.190002 ... 89.387146 54.453465 +w/o do (original x) 99.170000 0.000000 ... 0.000000 76.429195 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.24 67.628304 78.34685 76.719355 87.892377 77.646721 diff --git a/Meta-causal/code-withStyleAttack/73080.error b/Meta-causal/code-withStyleAttack/73080.error new file mode 100644 index 0000000000000000000000000000000000000000..d6efac5bd6b3adc5163adf70bcde543937a889b5 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73080.error @@ -0,0 +1,3 @@ +Downloading: "https://download.pytorch.org/models/resnet18-5c106cde.pth" to /home/yuqian_fu/.cache/torch/hub/checkpoints/resnet18-5c106cde.pth + 0%| | 0.00/44.7M [00:00=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +Downloading: "https://download.pytorch.org/models/resnet18-5c106cde.pth" to /home/yuqian_fu/.cache/torch/hub/checkpoints/resnet18-5c106cde.pth + 0%| | 0.00/44.7M [00:00=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.1) +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.8-py3-none-any.whl.metadata (53 kB) +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.65.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.4-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (71.0.4) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.4.0) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.19.0) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.24.5-py3-none-any.whl.metadata (13 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.13.0) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.7.4) +Requirement already satisfied: mpmath<1.4,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 213.3 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 23.0 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 102.8 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 220.5 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) +Downloading timm-1.0.8-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 219.4 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) +Downloading grpcio-1.65.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.7 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.7/5.7 MB 228.9 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) +Downloading protobuf-4.25.4-cp37-abi3-manylinux2014_x86_64.whl (294 kB) +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 113.3 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.24.5-py3-none-any.whl (417 kB) +Downloading packaging-24.1-py3-none-any.whl (53 kB) +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 166.2 MB/s eta 0:00:00 +Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) +Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94506 sha256=9e9dbea2962f09081d9d0e5d8b966d27885bea2398cb21aa7f3377b19780459b + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.65.2 h5py-3.11.0 huggingface_hub-0.24.5 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.4 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.8 tqdm-4.66.4 werkzeug-3.0.3 +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps5_RA_Adam', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[-0.0184, -0.0119, 0.0022, ..., 0.0157, 0.0061, -0.0129], + [ 0.0170, 0.0004, 0.0204, ..., 0.0014, 0.0035, -0.0166], + [ 0.0130, -0.0087, -0.0109, ..., 0.0161, -0.0087, -0.0111], + ..., + [-0.0124, 0.0201, 0.0069, ..., -0.0073, -0.0180, -0.0210], + [-0.0180, -0.0212, -0.0180, ..., 0.0162, -0.0175, 0.0149], + [-0.0038, -0.0085, 0.0052, ..., -0.0174, 0.0157, 0.0038]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0148, 0.0139, 0.0076, 0.0038, 0.0162, -0.0021, -0.0016], + device='cuda:0'), grad: None +588 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 768.94, cls_loss 3.5739 cls_loss_mapping 110.2172 cls_loss_causal 105.5066 re_mapping 2525.5655 re_causal 2193.0438 /// teacc 51.51 lr 0.00999497 +Epoch 2, weight, value: tensor([[ 0.0223, 0.0250, 0.0774, ..., -0.0066, 0.0195, 0.0088], + [ 0.0179, -0.0201, 0.0476, ..., 0.0630, 0.0328, 0.0235], + [-0.0528, -0.0651, -0.1461, ..., -0.0156, -0.0286, -0.0407], + ..., + [-0.0769, -0.0025, -0.0956, ..., 0.0682, 0.0606, 0.0698], + [ 0.0183, 0.0183, 0.0580, ..., -0.0346, -0.1123, -0.0328], + [-0.0052, -0.0021, 0.0394, ..., -0.0235, 0.0480, 0.0352]], + device='cuda:0'), grad: tensor([[ 3.2983e-01, 3.3276e-01, 2.2681e-01, ..., 2.0007e-01, + 2.1716e-01, 2.6611e-01], + [-1.3916e-01, -1.6992e-01, -1.0114e-01, ..., -2.1130e-01, + -2.1997e-01, -2.3523e-01], + [ 5.2989e-05, 1.1700e-04, 6.1274e-05, ..., 4.0174e-04, + 4.3678e-04, 3.7694e-04], + ..., + [ 5.1147e-02, 6.2561e-02, 3.7598e-02, ..., 7.5562e-02, + 7.5134e-02, 8.5510e-02], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 3.1796e-03, 2.9716e-03, 2.1496e-03, ..., 8.7070e-04, + 9.7942e-04, 1.5593e-03]], device='cuda:0') +Epoch 2, bias, value: tensor([ 1.2916e-02, 8.2871e-04, 2.0194e-02, 2.0127e-02, 5.4337e-02, + -4.9469e-02, 6.9660e-05], device='cuda:0'), grad: tensor([ 3.7085e-01, -3.4546e-01, 3.6287e-04, -1.6492e-01, 1.3696e-01, + 0.0000e+00, 2.1744e-03], device='cuda:0') +588 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 773.01, cls_loss 1.8538 cls_loss_mapping 1.8926 cls_loss_causal 1.8840 re_mapping 0.3082 re_causal 0.3068 /// teacc 63.07 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 0.0374, 0.0464, 0.0185, ..., 0.0325, 0.0157, 0.0311], + [ 0.0620, 0.0456, 0.1125, ..., 0.0401, 0.0043, -0.0202], + [ 0.0399, 0.0045, -0.1133, ..., -0.0064, -0.0302, -0.0488], + ..., + [-0.1396, -0.0521, -0.0657, ..., -0.0593, 0.0603, 0.0217], + [ 0.0314, 0.0201, 0.1118, ..., 0.0228, -0.0943, 0.0076], + [-0.1154, -0.1236, -0.0613, ..., 0.0711, 0.1170, 0.0888]], + device='cuda:0'), grad: tensor([[ 6.7215e-03, -2.1534e-03, -2.9343e-02, ..., 3.3539e-02, + 1.4696e-03, 4.2725e-03], + [-3.1738e-01, -2.6782e-01, -1.8347e-01, ..., -2.3206e-01, + -2.1057e-01, -2.1838e-01], + [ 1.9910e-01, 1.7639e-01, 1.4795e-01, ..., 7.6416e-02, + 7.6172e-02, 8.8013e-02], + ..., + [ 2.6665e-03, 2.3746e-03, 2.0657e-03, ..., 2.2869e-03, + 2.9621e-03, 2.7847e-03], + [ 5.3644e-07, 4.1723e-07, 3.5763e-07, ..., 5.9605e-08, + 5.9605e-08, 5.9605e-08], + [ 2.5781e-01, 2.2180e-01, 1.6968e-01, ..., 1.4087e-01, + 1.5295e-01, 1.4722e-01]], device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0091, 0.0036, 0.0441, 0.0197, 0.0410, -0.0548, -0.0045], + device='cuda:0'), grad: tensor([-8.9844e-02, -3.4863e-01, 3.0054e-01, -1.5210e-01, 5.1880e-03, + 5.3644e-07, 2.8491e-01], device='cuda:0') +588 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 780.58, cls_loss 1.5391 cls_loss_mapping 1.8986 cls_loss_causal 1.8912 re_mapping 0.1304 re_causal 0.1297 /// teacc 69.35 lr 0.00995475 +Epoch 4, weight, value: tensor([[ 0.0820, 0.0893, 0.0341, ..., 0.0367, 0.0471, 0.0508], + [ 0.0364, 0.0183, 0.0674, ..., 0.0311, -0.0154, -0.0208], + [ 0.0234, -0.0017, -0.0791, ..., 0.0038, -0.0132, -0.0277], + ..., + [-0.1511, -0.0522, -0.0584, ..., -0.0349, 0.0741, 0.0339], + [ 0.0934, 0.0830, 0.1424, ..., 0.0781, -0.0802, 0.0071], + [-0.1357, -0.1573, -0.1040, ..., 0.0300, 0.0307, 0.0358]], + device='cuda:0'), grad: tensor([[-1.0254e-01, -5.2948e-02, -2.8168e-02, ..., -7.8857e-02, + -5.2704e-02, -4.3488e-02], + [ 2.4255e-01, 1.6577e-01, 8.6914e-02, ..., 1.6101e-01, + 1.2683e-01, 1.1298e-01], + [ 2.8244e-02, 2.0615e-02, 9.7733e-03, ..., 1.6678e-02, + 1.3931e-02, 1.2566e-02], + ..., + [-4.4281e-02, -4.2908e-02, -2.6855e-02, ..., -2.6245e-02, + -2.7908e-02, -2.8290e-02], + [ 1.1921e-07, 5.9605e-08, 5.9605e-08, ..., 1.1921e-07, + 5.9605e-08, 1.1921e-07], + [ 1.0550e-05, 1.0490e-05, 6.1393e-06, ..., 5.9605e-06, + 6.3777e-06, 6.2585e-06]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0076, 0.0268, 0.0495, 0.0059, 0.0494, -0.0534, -0.0556], + device='cuda:0'), grad: tensor([-7.1899e-02, 3.4863e-01, 4.0131e-02, -1.6663e-01, -1.5027e-01, + 4.7684e-07, 3.2544e-05], device='cuda:0') +588 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 785.32, cls_loss 1.3587 cls_loss_mapping 2.1901 cls_loss_causal 2.0801 re_mapping 0.1097 re_causal 0.0748 /// teacc 73.87 lr 0.00991965 +Epoch 5, weight, value: tensor([[ 0.0695, 0.0671, 0.0358, ..., 0.0113, 0.0492, 0.0534], + [ 0.0274, 0.0196, 0.0236, ..., 0.0316, -0.0380, -0.0459], + [ 0.0144, -0.0048, -0.0799, ..., -0.0018, -0.0316, -0.0349], + ..., + [-0.1295, -0.0424, -0.0462, ..., -0.0368, 0.0855, 0.0309], + [ 0.1435, 0.1211, 0.1868, ..., 0.0738, -0.0791, 0.0190], + [-0.1357, -0.1652, -0.0843, ..., 0.0743, 0.1043, 0.0921]], + device='cuda:0'), grad: tensor([[ 3.9558e-03, 2.2221e-03, 1.2808e-03, ..., 1.0490e-02, + 8.7967e-03, 1.0811e-02], + [ 1.6518e-03, 9.1314e-04, 4.8804e-04, ..., 7.5102e-04, + 4.5705e-04, 3.6025e-04], + [ 1.1032e-02, 5.4398e-03, 2.7637e-03, ..., 5.5923e-03, + 3.9101e-03, 3.0422e-03], + ..., + [-1.8921e-02, -9.5291e-03, -5.0392e-03, ..., -2.0050e-02, + -1.5656e-02, -1.7059e-02], + [ 1.9765e-04, 1.0020e-04, 5.4598e-05, ..., 1.8251e-04, + 1.3185e-04, 1.4544e-04], + [ 3.9864e-03, 2.0237e-03, 1.1015e-03, ..., 3.6697e-03, + 2.6512e-03, 2.9221e-03]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0211, 0.0262, 0.0561, -0.0229, 0.0575, -0.0753, -0.0556], + device='cuda:0'), grad: tensor([ 0.0378, 0.0030, 0.0197, -0.0035, -0.0715, 0.0007, 0.0138], + device='cuda:0') +588 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 781.29, cls_loss 1.2020 cls_loss_mapping 1.9142 cls_loss_causal 1.8996 re_mapping 0.0660 re_causal 0.0658 /// teacc 85.93 lr 0.00987464 +Epoch 6, weight, value: tensor([[ 0.0266, 0.0229, -0.0079, ..., -0.0146, 0.0206, 0.0161], + [ 0.0501, 0.0481, 0.0592, ..., 0.0700, 0.0044, 0.0104], + [ 0.0558, 0.0519, -0.0151, ..., -0.0215, -0.0604, -0.0649], + ..., + [-0.1457, -0.0573, -0.0560, ..., -0.0435, 0.0920, 0.0356], + [ 0.1095, 0.0892, 0.1326, ..., 0.0458, -0.1377, -0.0370], + [-0.1541, -0.2007, -0.1094, ..., 0.1120, 0.1443, 0.1341]], + device='cuda:0'), grad: tensor([[-1.1389e-01, -5.4779e-02, -8.9111e-03, ..., -5.0507e-02, + -1.9196e-02, -1.9135e-02], + [ 3.9330e-03, 2.7237e-03, 2.1992e-03, ..., 1.3399e-03, + 1.1625e-03, 9.1076e-04], + [ 2.7657e-03, 1.5955e-03, 1.6918e-03, ..., -3.4785e-04, + -5.6267e-05, -3.9482e-04], + ..., + [ 9.0271e-02, 3.8300e-02, -4.5815e-03, ..., 4.3091e-02, + 1.2718e-02, 1.4366e-02], + [ 2.6202e-04, 1.8251e-04, 1.4889e-04, ..., 8.5175e-05, + 7.4387e-05, 5.6177e-05], + [ 3.6240e-04, 2.5272e-04, 2.0730e-04, ..., 1.0860e-04, + 9.5725e-05, 6.8247e-05]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0448, 0.0248, 0.0328, 0.0130, 0.0623, -0.0987, -0.1037], + device='cuda:0'), grad: tensor([-0.1582, 0.0141, 0.0086, 0.0615, 0.0718, 0.0010, 0.0014], + device='cuda:0') +588 +0.009874639560909117 +changing lr +epoch 5, time 782.25, cls_loss 1.0410 cls_loss_mapping 1.9548 cls_loss_causal 1.8761 re_mapping 0.0672 re_causal 0.0645 /// teacc 71.86 lr 0.00981981 +Epoch 7, weight, value: tensor([[ 0.0316, 0.0397, 0.0112, ..., 0.0122, 0.0542, 0.0416], + [ 0.0262, 0.0136, 0.0114, ..., 0.0690, -0.0098, 0.0035], + [ 0.0512, 0.0590, -0.0043, ..., -0.0567, -0.0998, -0.1060], + ..., + [-0.1470, -0.0585, -0.0542, ..., -0.0344, 0.0967, 0.0371], + [ 0.1206, 0.0975, 0.1345, ..., 0.0267, -0.1654, -0.0568], + [-0.1286, -0.1820, -0.0886, ..., 0.1344, 0.1690, 0.1676]], + device='cuda:0'), grad: tensor([[ 6.0272e-02, 2.7435e-02, 9.3155e-03, ..., 3.4790e-02, + 2.2552e-02, 2.3010e-02], + [ 7.8674e-02, 3.1830e-02, 7.7705e-03, ..., 4.8492e-02, + 3.1525e-02, 3.0685e-02], + [-1.4258e-01, -6.0852e-02, -1.5106e-02, ..., -8.3557e-02, + -5.3650e-02, -5.3711e-02], + ..., + [ 1.0071e-03, 5.9032e-04, 3.5238e-04, ..., 5.1308e-04, + 3.4189e-04, 3.8791e-04], + [ 5.9128e-03, 3.4428e-03, 1.8206e-03, ..., 2.8954e-03, + 1.8768e-03, 2.1820e-03], + [ 1.1683e-05, 6.8247e-06, 3.6061e-06, ..., 5.7220e-06, + 3.7253e-06, 4.3213e-06]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0657, 0.0294, 0.0322, -0.0343, 0.0530, -0.0953, -0.0699], + device='cuda:0'), grad: tensor([ 9.8755e-02, 7.3120e-02, -1.6650e-01, -2.9617e-02, 3.7117e-03, + 2.0508e-02, 4.0621e-05], device='cuda:0') +588 +0.009819814303479266 +changing lr +epoch 6, time 785.00, cls_loss 1.1321 cls_loss_mapping 2.0379 cls_loss_causal 1.9503 re_mapping 0.0874 re_causal 0.0683 /// teacc 67.09 lr 0.00975528 +Epoch 8, weight, value: tensor([[ 0.0497, 0.0576, 0.0631, ..., -0.0440, -0.0326, -0.0438], + [ 0.0461, 0.0368, 0.0215, ..., 0.1022, 0.0300, 0.0572], + [ 0.0155, 0.0182, -0.0703, ..., -0.0451, -0.0587, -0.0731], + ..., + [-0.1224, -0.0320, -0.0220, ..., -0.0342, 0.0880, 0.0214], + [ 0.0736, 0.0444, 0.0680, ..., -0.0154, -0.2294, -0.1180], + [-0.1203, -0.1739, -0.1006, ..., 0.1403, 0.1648, 0.1680]], + device='cuda:0'), grad: tensor([[-1.4307e-01, -6.6956e-02, -4.9164e-02, ..., -4.6051e-02, + -3.3051e-02, -3.1891e-02], + [ 1.3928e-01, 6.4148e-02, 4.3549e-02, ..., 3.6499e-02, + 2.2797e-02, 2.3361e-02], + [ 5.2185e-03, 3.7098e-03, 6.9733e-03, ..., 1.3771e-02, + 1.4526e-02, 1.2390e-02], + ..., + [ 8.8930e-05, 4.2528e-05, 4.7088e-05, ..., 1.0747e-04, + 1.0806e-04, 9.9599e-05], + [ 2.0757e-05, 8.7470e-06, 7.1675e-06, ..., 1.5885e-05, + 1.5512e-05, 1.5080e-05], + [ 8.2850e-05, 3.9488e-05, 4.2707e-05, ..., 1.0383e-04, + 1.0383e-04, 9.6321e-05]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0465, 0.0106, 0.0850, -0.0369, 0.0885, -0.1035, -0.1627], + device='cuda:0'), grad: tensor([-1.6052e-01, 1.7181e-02, 1.7004e-01, -2.8488e-02, 9.3079e-04, + 1.2159e-04, 8.1444e-04], device='cuda:0') +588 +0.009755282581475767 +changing lr +epoch 7, time 785.21, cls_loss 0.9270 cls_loss_mapping 1.9301 cls_loss_causal 1.7580 re_mapping 0.0610 re_causal 0.0614 /// teacc 73.62 lr 0.00968117 +Epoch 9, weight, value: tensor([[ 0.0706, 0.0821, 0.1112, ..., -0.0248, 0.0023, -0.0161], + [ 0.0381, 0.0519, 0.0004, ..., 0.1056, 0.0288, 0.0387], + [-0.0074, -0.0221, -0.1374, ..., -0.0416, -0.0661, -0.0858], + ..., + [-0.1207, -0.0353, -0.0097, ..., -0.0640, 0.0612, 0.0087], + [ 0.0864, 0.0459, 0.0736, ..., -0.0896, -0.3091, -0.1860], + [-0.1414, -0.1997, -0.1052, ..., 0.1783, 0.2116, 0.2316]], + device='cuda:0'), grad: tensor([[ 8.8196e-02, 3.5431e-02, 3.8544e-02, ..., 2.7359e-02, + 2.7740e-02, 2.8519e-02], + [ 1.1986e-02, 3.1319e-03, 2.1114e-03, ..., 6.5918e-03, + 5.9891e-03, 6.0616e-03], + [-1.8244e-03, -3.7837e-04, -2.4283e-04, ..., -1.5717e-03, + -1.7834e-03, -1.7385e-03], + ..., + [-9.8389e-02, -3.8177e-02, -4.0436e-02, ..., -3.2379e-02, + -3.1952e-02, -3.2867e-02], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 7.8976e-07, 2.3842e-07, 2.0862e-07, ..., 4.6194e-07, + 4.9174e-07, 4.7684e-07]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0800, 0.0006, 0.0876, -0.0736, 0.0668, -0.0791, -0.1341], + device='cuda:0'), grad: tensor([ 1.7090e-01, 2.6810e-02, -3.9062e-03, 1.1027e-06, -1.9385e-01, + 0.0000e+00, 1.6540e-06], device='cuda:0') +588 +0.009681174353198686 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 781.29, cls_loss 0.8739 cls_loss_mapping 1.9264 cls_loss_causal 1.7031 re_mapping 0.0548 re_causal 0.0548 /// teacc 89.20 lr 0.00959764 +Epoch 10, weight, value: tensor([[ 0.0749, 0.0902, 0.1229, ..., -0.0243, 0.0085, -0.0069], + [ 0.0096, 0.0120, -0.0589, ..., 0.0984, 0.0110, 0.0391], + [ 0.0221, 0.0364, -0.0500, ..., -0.0063, -0.0399, -0.0708], + ..., + [-0.1264, -0.0622, -0.0581, ..., -0.0483, 0.0814, 0.0171], + [ 0.0781, 0.0707, 0.1235, ..., -0.1009, -0.3176, -0.1900], + [-0.1512, -0.2267, -0.1250, ..., 0.1614, 0.1977, 0.2150]], + device='cuda:0'), grad: tensor([[-9.1858e-02, -3.2288e-02, -3.4973e-02, ..., -2.9587e-02, + -3.0762e-02, -2.7374e-02], + [ 1.5152e-02, 9.3231e-03, 1.0956e-02, ..., 5.9128e-03, + 6.6681e-03, 6.8779e-03], + [ 6.3324e-03, 3.0494e-04, -3.1328e-04, ..., -1.6379e-04, + -5.2738e-04, -2.1248e-03], + ..., + [ 1.5274e-02, 4.3793e-03, 4.5471e-03, ..., 4.7340e-03, + 4.8065e-03, 4.1237e-03], + [ 7.4506e-05, 5.6177e-05, 6.7770e-05, ..., 3.5584e-05, + 4.1127e-05, 4.6372e-05], + [ 6.9733e-03, 5.2795e-03, 6.3057e-03, ..., 2.8381e-03, + 3.3112e-03, 3.5248e-03]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0810, -0.0156, 0.0911, -0.0452, 0.0564, -0.0767, -0.1248], + device='cuda:0'), grad: tensor([-0.2581, 0.0810, -0.0205, 0.1172, 0.0347, 0.0005, 0.0452], + device='cuda:0') +588 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 785.85, cls_loss 0.8995 cls_loss_mapping 1.9225 cls_loss_causal 1.6496 re_mapping 0.0522 re_causal 0.0520 /// teacc 89.45 lr 0.00950484 +Epoch 11, weight, value: tensor([[ 0.0599, 0.0623, 0.1251, ..., -0.0106, 0.0308, -0.0036], + [ 0.0313, 0.0261, -0.0223, ..., 0.1425, 0.0620, 0.1000], + [ 0.0330, 0.0484, -0.0699, ..., -0.0325, -0.0637, -0.0841], + ..., + [-0.1210, -0.0126, -0.0284, ..., -0.0981, 0.0041, -0.0339], + [ 0.0780, 0.0251, 0.1002, ..., -0.1323, -0.3406, -0.2067], + [-0.1619, -0.2235, -0.1608, ..., 0.1217, 0.1507, 0.1442]], + device='cuda:0'), grad: tensor([[-8.3008e-02, -6.0997e-03, -5.0011e-03, ..., -2.9266e-02, + -3.5095e-02, -4.2633e-02], + [ 3.8452e-02, 9.3155e-03, 9.8801e-03, ..., 1.7490e-03, + 3.8033e-03, 6.6147e-03], + [ 1.0052e-01, 1.1597e-02, 1.2558e-02, ..., 3.0136e-02, + 3.6621e-02, 4.5044e-02], + ..., + [-5.1300e-02, -1.1230e-02, -1.0216e-02, ..., -2.0962e-03, + -5.0774e-03, -9.2163e-03], + [ 8.5163e-04, 2.4652e-04, 3.1686e-04, ..., 4.3809e-05, + 8.0884e-05, 1.3030e-04], + [-1.8234e-02, -7.0877e-03, -1.1215e-02, ..., -1.1625e-03, + -1.5802e-03, -2.0561e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0930, -0.0307, 0.0866, 0.0041, 0.0191, -0.0644, -0.0959], + device='cuda:0'), grad: tensor([-0.1167, 0.1121, 0.2109, 0.0428, -0.1073, 0.0039, -0.1458], + device='cuda:0') +588 +0.009504844339512096 +changing lr +epoch 10, time 788.52, cls_loss 0.8072 cls_loss_mapping 1.9321 cls_loss_causal 1.5963 re_mapping 0.0528 re_causal 0.0515 /// teacc 85.43 lr 0.00940298 +Epoch 12, weight, value: tensor([[ 0.0781, 0.0914, 0.1681, ..., -0.0547, -0.0107, -0.0282], + [ 0.0217, 0.0047, -0.0365, ..., 0.1635, 0.0752, 0.1021], + [ 0.0120, 0.0353, -0.1016, ..., -0.0444, -0.0656, -0.0869], + ..., + [-0.1132, 0.0120, 0.0207, ..., -0.0414, 0.0796, 0.0249], + [ 0.0775, -0.0169, 0.0332, ..., -0.1575, -0.3792, -0.2468], + [-0.1794, -0.2264, -0.1943, ..., 0.1030, 0.1175, 0.1258]], + device='cuda:0'), grad: tensor([[-2.8732e-02, -7.7209e-03, -1.1177e-02, ..., -4.9744e-03, + -1.1841e-02, -1.4404e-02], + [-4.5837e-02, -7.5684e-03, -8.1406e-03, ..., -1.1093e-02, + -1.5976e-02, -2.0782e-02], + [-1.7075e-02, -1.6022e-03, 1.0376e-03, ..., -6.7711e-03, + -4.5586e-03, -2.1820e-03], + ..., + [ 4.6539e-02, 1.0094e-02, 1.2726e-02, ..., 1.0178e-02, + 1.7502e-02, 2.0767e-02], + [ 3.0365e-03, 4.2748e-04, 2.7204e-04, ..., 9.1887e-04, + 9.6607e-04, 9.7513e-04], + [ 8.8692e-05, 1.2189e-05, 6.8024e-06, ..., 2.7642e-05, + 2.7806e-05, 2.6613e-05]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0643, -0.0029, 0.1133, 0.0198, 0.0366, -0.1691, -0.1317], + device='cuda:0'), grad: tensor([-0.1500, -0.1061, -0.0862, 0.1324, 0.1986, 0.0109, 0.0003], + device='cuda:0') +588 +0.009402977659283692 +changing lr +epoch 11, time 780.78, cls_loss 0.8625 cls_loss_mapping 1.9354 cls_loss_causal 1.5560 re_mapping 0.0516 re_causal 0.0509 /// teacc 86.18 lr 0.00929224 +Epoch 13, weight, value: tensor([[ 0.0994, 0.1175, 0.2072, ..., 0.0095, 0.0632, 0.0378], + [ 0.0239, 0.0202, -0.0117, ..., 0.1219, 0.0594, 0.0686], + [-0.0032, 0.0054, -0.1578, ..., -0.0145, -0.0509, -0.0981], + ..., + [-0.1119, 0.0189, 0.0309, ..., -0.1342, -0.0211, -0.0492], + [ 0.0584, -0.0392, -0.0244, ..., -0.2124, -0.4568, -0.3141], + [-0.1354, -0.1843, -0.1637, ..., 0.1503, 0.1429, 0.1685]], + device='cuda:0'), grad: tensor([[-6.9702e-02, -1.9058e-02, -1.6525e-02, ..., -6.1913e-03, + -1.2161e-02, -1.7960e-02], + [ 1.2039e-02, 3.2258e-04, 7.5626e-04, ..., 1.8477e-04, + 4.9877e-04, 1.1559e-03], + [-3.6194e-02, -1.0138e-03, -2.3384e-03, ..., -1.0548e-03, + -2.0084e-03, -4.0398e-03], + ..., + [ 7.4646e-02, 1.9211e-02, 1.6861e-02, ..., 6.4240e-03, + 1.2527e-02, 1.8616e-02], + [ 8.4114e-04, 2.3559e-05, 5.4330e-05, ..., 2.4512e-05, + 4.6670e-05, 9.3937e-05], + [ 1.7960e-02, 5.1308e-04, 1.1721e-03, ..., 6.0177e-04, + 1.0777e-03, 2.0943e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0716, -0.0209, 0.1285, -0.0644, 0.0349, -0.1278, -0.0401], + device='cuda:0'), grad: tensor([-0.1522, 0.0534, -0.1591, 0.0020, 0.1736, 0.0037, 0.0787], + device='cuda:0') +588 +0.009292243968009333 +changing lr +epoch 12, time 788.18, cls_loss 0.8157 cls_loss_mapping 1.9419 cls_loss_causal 1.5347 re_mapping 0.0501 re_causal 0.0484 /// teacc 84.17 lr 0.00917287 +Epoch 14, weight, value: tensor([[ 0.0897, 0.1369, 0.2297, ..., -0.0123, 0.0430, 0.0184], + [ 0.0054, -0.0383, -0.0697, ..., 0.1412, 0.0559, 0.0741], + [ 0.0216, 0.0205, -0.1270, ..., -0.0577, -0.0855, -0.1333], + ..., + [-0.1028, 0.0523, 0.0464, ..., -0.0920, 0.0234, -0.0142], + [ 0.0635, 0.0260, 0.0361, ..., -0.1465, -0.3877, -0.2240], + [-0.1648, -0.2473, -0.2080, ..., 0.1545, 0.1637, 0.2083]], + device='cuda:0'), grad: tensor([[ 4.7531e-03, 8.1158e-04, 1.0948e-03, ..., 1.3628e-03, + 1.6851e-03, 2.3518e-03], + [ 1.2238e-01, 9.2697e-03, 1.4427e-02, ..., 1.2444e-02, + 1.8356e-02, 3.8300e-02], + [ 3.0701e-02, 2.5482e-03, 3.8834e-03, ..., 3.5610e-03, + 5.0926e-03, 1.0033e-02], + ..., + [-8.8501e-02, -9.2545e-03, -1.3451e-02, ..., -1.4000e-02, + -1.8799e-02, -3.2593e-02], + [ 5.1335e-06, 3.7998e-07, 5.9605e-07, ..., 4.9919e-07, + 7.4506e-07, 1.5870e-06], + [-6.9519e-02, -3.3817e-03, -5.9700e-03, ..., -3.3894e-03, + -6.3667e-03, -1.8143e-02]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0903, -0.0141, 0.1083, -0.0382, 0.0304, -0.1286, -0.0785], + device='cuda:0'), grad: tensor([ 1.3031e-02, 2.8564e-01, 7.2632e-02, 3.7384e-04, -2.1790e-01, + 1.1936e-05, -1.5405e-01], device='cuda:0') +588 +0.009172866268606516 +changing lr +epoch 13, time 781.62, cls_loss 0.8214 cls_loss_mapping 1.9875 cls_loss_causal 1.5469 re_mapping 0.0599 re_causal 0.0463 /// teacc 83.17 lr 0.00904508 +Epoch 15, weight, value: tensor([[ 0.1048, 0.1656, 0.2340, ..., -0.1068, -0.0505, -0.0660], + [-0.0049, -0.0581, -0.0859, ..., 0.1688, 0.0811, 0.0820], + [ 0.0133, 0.0728, -0.0662, ..., -0.0319, -0.0422, -0.0959], + ..., + [-0.1108, 0.0209, 0.0174, ..., -0.0368, 0.0661, 0.0214], + [ 0.0604, -0.0521, -0.0192, ..., -0.2233, -0.4485, -0.2269], + [-0.1568, -0.2658, -0.1842, ..., 0.1709, 0.1814, 0.2454]], + device='cuda:0'), grad: tensor([[-7.9422e-03, -4.0603e-04, -6.6757e-04, ..., -1.2197e-03, + -1.7109e-03, -3.2368e-03], + [ 2.4104e-04, 1.5482e-05, 2.3097e-05, ..., 3.8922e-05, + 5.3585e-05, 1.0073e-04], + [ 3.5763e-05, 1.8440e-06, 3.0212e-06, ..., 5.5060e-06, + 7.7114e-06, 1.4596e-05], + ..., + [ 7.6370e-03, 3.8719e-04, 6.3896e-04, ..., 1.1721e-03, + 1.6432e-03, 3.1109e-03], + [ 3.2037e-07, 1.4901e-08, 2.6077e-08, ..., 4.8429e-08, + 7.0781e-08, 1.3039e-07], + [ 1.5259e-05, 8.6427e-07, 1.3597e-06, ..., 2.3954e-06, + 3.3304e-06, 6.2883e-06]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.1039, -0.0150, 0.1055, 0.0046, 0.0436, -0.1953, -0.1409], + device='cuda:0'), grad: tensor([-3.8330e-02, 1.1873e-03, 1.7273e-04, 3.8803e-05, 3.6835e-02, + 1.5497e-06, 7.4267e-05], device='cuda:0') +588 +0.00904508497187474 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 785.22, cls_loss 0.7987 cls_loss_mapping 1.9176 cls_loss_causal 1.5042 re_mapping 0.0453 re_causal 0.0440 /// teacc 90.70 lr 0.00890916 +Epoch 16, weight, value: tensor([[ 0.0932, 0.1151, 0.1646, ..., -0.1365, -0.0741, -0.0827], + [-0.0090, -0.0474, -0.0663, ..., 0.1953, 0.0896, 0.0805], + [ 0.0254, 0.0497, -0.0991, ..., -0.0568, -0.0479, -0.1008], + ..., + [-0.0952, 0.0957, 0.0867, ..., -0.0252, 0.0657, 0.0264], + [ 0.0606, -0.1301, -0.0576, ..., -0.1886, -0.3981, -0.2093], + [-0.1773, -0.2405, -0.1711, ..., 0.2728, 0.2728, 0.2919]], + device='cuda:0'), grad: tensor([[ 4.7989e-03, 6.4230e-04, 7.7724e-04, ..., 2.5082e-04, + 3.6550e-04, 1.6956e-03], + [-1.5533e-02, -2.6722e-03, -3.0994e-03, ..., 6.4373e-04, + 4.2081e-05, -5.1842e-03], + [ 5.4121e-04, 6.9022e-05, 8.3804e-05, ..., 3.2514e-05, + 4.5180e-05, 1.9038e-04], + ..., + [ 5.7373e-03, 1.3742e-03, 1.5306e-03, ..., -1.1511e-03, + -7.8726e-04, 1.7395e-03], + [ 1.4198e-04, 1.8656e-05, 2.2531e-05, ..., 7.3016e-06, + 1.0803e-05, 4.9710e-05], + [ 2.2373e-03, 2.9469e-04, 3.5572e-04, ..., 1.1384e-04, + 1.6928e-04, 7.8344e-04]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.1158, 0.0231, 0.1040, -0.0379, 0.0579, -0.3031, -0.1504], + device='cuda:0'), grad: tensor([ 0.0304, -0.1083, 0.0034, 0.0131, 0.0466, 0.0009, 0.0140], + device='cuda:0') +588 +0.008909157412340152 +changing lr +epoch 15, time 785.61, cls_loss 0.7490 cls_loss_mapping 1.9150 cls_loss_causal 1.4867 re_mapping 0.0391 re_causal 0.0375 /// teacc 87.19 lr 0.00876536 +Epoch 17, weight, value: tensor([[ 0.0881, 0.1401, 0.1780, ..., -0.0813, -0.0197, -0.0337], + [-0.0289, -0.1781, -0.1736, ..., 0.1876, 0.0876, 0.0780], + [ 0.0400, 0.0977, -0.0207, ..., -0.0433, -0.0435, -0.0837], + ..., + [-0.1047, 0.0791, 0.0697, ..., -0.0430, 0.0530, -0.0055], + [ 0.0515, -0.1268, -0.0618, ..., -0.3298, -0.5177, -0.2736], + [-0.1459, -0.1469, -0.1048, ..., 0.2251, 0.2375, 0.2757]], + device='cuda:0'), grad: tensor([[-5.1727e-03, -2.7084e-04, -2.8849e-04, ..., -9.2554e-04, + -1.3981e-03, -2.6894e-03], + [ 2.1383e-06, 5.8487e-07, 5.9232e-07, ..., 5.0664e-07, + 5.3272e-07, 8.2329e-07], + [-5.3272e-07, 7.4506e-09, 2.2352e-08, ..., -4.8801e-07, + -4.3958e-07, -4.9919e-07], + ..., + [ 5.1689e-03, 2.7013e-04, 2.8753e-04, ..., 9.2506e-04, + 1.3962e-03, 2.6855e-03], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 5.9605e-08, 2.9802e-08, 2.9802e-08, ..., 3.7253e-09, + 0.0000e+00, 7.4506e-09]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0820, 0.0419, 0.0772, -0.0038, 0.0782, -0.2720, -0.1550], + device='cuda:0'), grad: tensor([-1.9409e-02, 6.4485e-06, -3.6508e-07, 1.1176e-08, 1.9394e-02, + 0.0000e+00, 1.6019e-07], device='cuda:0') +588 +0.00876535733001806 +changing lr +epoch 16, time 787.07, cls_loss 1.0272 cls_loss_mapping 2.4374 cls_loss_causal 2.3161 re_mapping 0.0529 re_causal 0.1017 /// teacc 82.66 lr 0.00861397 +Epoch 18, weight, value: tensor([[ 0.1099, 0.1502, 0.1563, ..., 0.0062, 0.0320, 0.0096], + [-0.0287, -0.1242, -0.1041, ..., 0.1810, 0.1111, 0.1118], + [ 0.0531, 0.0996, 0.0422, ..., -0.1603, -0.1208, -0.1710], + ..., + [-0.1320, -0.0151, -0.0713, ..., -0.0658, 0.0121, 0.0082], + [ 0.0321, -0.2659, -0.2384, ..., -0.3475, -0.5984, -0.3172], + [-0.1466, -0.1135, -0.0220, ..., 0.2731, 0.3049, 0.2587]], + device='cuda:0'), grad: tensor([[-2.6978e-02, -1.6556e-03, -1.8330e-03, ..., -4.2496e-03, + -3.5191e-03, -3.1338e-03], + [-5.7831e-02, -4.4250e-03, -3.0384e-03, ..., -1.9836e-02, + -1.8967e-02, -2.4017e-02], + [ 2.0233e-02, 1.8778e-03, 1.3866e-03, ..., 7.6485e-03, + 7.3547e-03, 8.7128e-03], + ..., + [ 6.7566e-02, 4.3640e-03, 3.9825e-03, ..., 1.4503e-02, + 1.2993e-02, 1.6006e-02], + [ 1.2293e-07, 9.6858e-08, 5.9605e-08, ..., 4.0606e-07, + 4.2096e-07, 4.7311e-07], + [ 7.5722e-04, 5.4979e-04, 3.3545e-04, ..., 2.3022e-03, + 2.3746e-03, 2.6875e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.1178, 0.0349, 0.0845, -0.0011, 0.0849, -0.3069, -0.2269], + device='cuda:0'), grad: tensor([-3.9398e-02, -3.4058e-01, 1.2360e-01, -2.4281e-03, 2.1936e-01, + 6.9439e-06, 3.9337e-02], device='cuda:0') +588 +0.008613974319136962 +changing lr +epoch 17, time 786.48, cls_loss 0.7901 cls_loss_mapping 2.0370 cls_loss_causal 1.8258 re_mapping 0.0511 re_causal 0.0509 /// teacc 89.70 lr 0.00845531 +Epoch 19, weight, value: tensor([[ 9.9670e-02, 1.7887e-01, 1.8741e-01, ..., 6.4120e-02, + 9.3857e-02, 4.8437e-02], + [-1.4598e-02, -1.5046e-01, -1.2635e-01, ..., 1.1182e-01, + 5.6881e-02, 6.4223e-02], + [ 5.7039e-02, 1.4452e-01, 6.9422e-02, ..., -1.4164e-01, + -1.3186e-01, -1.5214e-01], + ..., + [-1.2695e-01, -4.2238e-02, -9.9209e-02, ..., -9.0369e-02, + -2.1096e-02, -4.8919e-03], + [ 2.6778e-02, -3.1781e-01, -2.8282e-01, ..., -3.0728e-01, + -5.5136e-01, -2.9990e-01], + [-1.5649e-01, -9.9608e-02, 7.9323e-05, ..., 3.1000e-01, + 3.4786e-01, 2.8849e-01]], device='cuda:0'), grad: tensor([[-4.1931e-02, -4.8943e-03, -5.2834e-03, ..., -2.4738e-03, + -3.5858e-03, -8.3237e-03], + [ 3.6652e-02, 4.8752e-03, 5.2414e-03, ..., 2.1725e-03, + 3.0804e-03, 6.9504e-03], + [ 2.6989e-04, 3.4086e-07, 1.7136e-06, ..., 1.3523e-05, + 2.3633e-05, 6.7115e-05], + ..., + [ 4.9896e-03, 1.9729e-05, 4.4107e-05, ..., 2.8801e-04, + 4.8280e-04, 1.3037e-03], + [ 3.8743e-07, 0.0000e+00, 1.8626e-09, ..., 1.8626e-08, + 3.3528e-08, 9.6858e-08], + [ 1.1176e-07, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, + 9.3132e-09, 2.7940e-08]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0687, 0.0727, 0.0997, 0.0147, 0.1020, -0.3036, -0.2705], + device='cuda:0'), grad: tensor([-1.0883e-01, 8.8562e-02, 1.0290e-03, 1.9334e-06, 1.9287e-02, + 1.4771e-06, 4.2655e-07], device='cuda:0') +588 +0.008455313244934327 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 792.48, cls_loss 0.6640 cls_loss_mapping 1.9030 cls_loss_causal 1.7181 re_mapping 0.0393 re_causal 0.0394 /// teacc 91.71 lr 0.00828969 +Epoch 20, weight, value: tensor([[ 0.0927, 0.1293, 0.1443, ..., -0.0596, -0.0156, -0.0514], + [-0.0186, -0.0840, -0.0982, ..., 0.1475, 0.0574, 0.0569], + [ 0.0534, 0.1153, 0.0411, ..., -0.1358, -0.1174, -0.1334], + ..., + [-0.1253, -0.0480, -0.0766, ..., -0.0348, 0.0274, 0.0258], + [ 0.0300, -0.2536, -0.2211, ..., -0.3071, -0.5326, -0.2698], + [-0.1354, -0.1153, -0.0204, ..., 0.3638, 0.4074, 0.3599]], + device='cuda:0'), grad: tensor([[ 5.9223e-04, 8.8632e-05, 9.8467e-05, ..., 1.4770e-04, + 1.5235e-04, 2.2733e-04], + [-5.8031e-04, -8.8930e-05, -9.8705e-05, ..., -1.3971e-04, + -1.4472e-04, -2.1744e-04], + [ 2.9989e-07, 3.9116e-08, 4.4703e-08, ..., 8.9407e-08, + 8.9407e-08, 1.3039e-07], + ..., + [-1.3396e-05, 8.0094e-08, -6.1467e-08, ..., -8.4266e-06, + -8.0243e-06, -1.0595e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 4.2841e-08, 7.4506e-09, 7.4506e-09, ..., 1.1176e-08, + 1.1176e-08, 1.6764e-08]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0435, 0.0453, 0.1091, 0.0806, 0.1088, -0.3200, -0.2686], + device='cuda:0'), grad: tensor([ 1.8625e-03, -1.7653e-03, 1.1176e-06, 2.0042e-06, -1.0282e-04, + 0.0000e+00, 1.3039e-07], device='cuda:0') +588 +0.008289693629698565 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 788.43, cls_loss 0.7175 cls_loss_mapping 1.9173 cls_loss_causal 1.6670 re_mapping 0.0330 re_causal 0.0330 /// teacc 92.96 lr 0.00811745 +Epoch 21, weight, value: tensor([[ 9.1891e-02, 8.1340e-02, 8.2969e-02, ..., -6.4563e-02, + -1.6301e-02, -5.2537e-02], + [-2.6142e-02, -1.9188e-01, -1.6663e-01, ..., 1.5026e-01, + 6.6145e-02, 8.6999e-02], + [ 6.0192e-02, 1.7453e-01, 1.2074e-01, ..., 5.0772e-03, + -3.2672e-03, -7.5871e-02], + ..., + [-1.2180e-01, 6.8406e-05, -5.3561e-02, ..., -1.1375e-01, + -3.1294e-02, -1.4242e-02], + [ 3.8203e-02, -1.2938e-01, -1.2141e-01, ..., -2.3461e-01, + -4.6549e-01, -2.5152e-01], + [-1.3058e-01, -4.1205e-02, 6.0092e-02, ..., 1.9548e-01, + 2.5214e-01, 2.9548e-01]], device='cuda:0'), grad: tensor([[-0.0468, -0.0046, -0.0053, ..., -0.0109, -0.0111, -0.0158], + [-0.0105, -0.0012, -0.0008, ..., -0.0048, -0.0046, -0.0102], + [ 0.0035, 0.0001, 0.0001, ..., 0.0002, 0.0003, 0.0012], + ..., + [ 0.0696, 0.0055, 0.0063, ..., 0.0131, 0.0139, 0.0247], + [-0.0305, -0.0009, -0.0011, ..., -0.0014, -0.0023, -0.0095], + [ 0.0133, 0.0010, 0.0008, ..., 0.0037, 0.0037, 0.0091]], + device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0989, 0.0398, 0.0508, 0.0519, 0.1190, -0.3271, -0.2437], + device='cuda:0'), grad: tensor([-0.0945, -0.0936, 0.0182, 0.0071, 0.2197, -0.1527, 0.0956], + device='cuda:0') +588 +0.00811744900929367 +changing lr +epoch 20, time 785.92, cls_loss 0.6661 cls_loss_mapping 1.9219 cls_loss_causal 1.5924 re_mapping 0.0329 re_causal 0.0327 /// teacc 91.96 lr 0.00793893 +Epoch 22, weight, value: tensor([[ 0.0629, 0.0440, 0.0395, ..., -0.0731, -0.0183, -0.0545], + [-0.0205, -0.1703, -0.1429, ..., 0.1584, 0.0753, 0.0772], + [ 0.0643, 0.1886, 0.1447, ..., -0.0214, -0.0353, -0.0622], + ..., + [-0.1204, 0.0411, -0.0259, ..., -0.0771, 0.0035, 0.0036], + [ 0.0665, -0.0236, -0.0147, ..., -0.3274, -0.5222, -0.3354], + [-0.1134, -0.0866, 0.0271, ..., 0.1807, 0.2289, 0.2838]], + device='cuda:0'), grad: tensor([[ 7.0572e-04, 2.7552e-05, 8.2236e-07, ..., 6.1393e-05, + 4.5925e-05, 2.7442e-04], + [ 4.3640e-03, 1.7095e-04, 5.9679e-06, ..., 3.7503e-04, + 2.7990e-04, 1.6928e-03], + [ 4.7035e-03, 1.8489e-04, 6.9290e-06, ..., 4.0793e-04, + 3.0518e-04, 1.8263e-03], + ..., + [-1.5823e-02, -6.1989e-04, -2.1398e-05, ..., -1.3638e-03, + -1.0176e-03, -6.1417e-03], + [ 1.4853e-04, 5.8003e-06, 1.8813e-07, ..., 1.2733e-05, + 9.4920e-06, 5.7638e-05], + [ 2.8057e-03, 1.0961e-04, 3.5595e-06, ..., 2.4056e-04, + 1.7941e-04, 1.0891e-03]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0517, -0.0274, 0.1397, 0.0850, 0.1357, -0.3868, -0.2248], + device='cuda:0'), grad: tensor([ 0.0060, 0.0372, 0.0401, 0.0264, -0.1350, 0.0013, 0.0239], + device='cuda:0') +588 +0.007938926261462368 +changing lr +epoch 21, time 789.20, cls_loss 0.7044 cls_loss_mapping 1.9627 cls_loss_causal 1.5618 re_mapping 0.0338 re_causal 0.0317 /// teacc 85.43 lr 0.00775448 +Epoch 23, weight, value: tensor([[ 0.0814, 0.0740, 0.0388, ..., -0.0022, 0.0538, -0.0081], + [-0.0296, -0.1631, -0.1394, ..., 0.1643, 0.0467, 0.0704], + [ 0.0663, 0.1293, 0.0185, ..., -0.0327, -0.0695, -0.0847], + ..., + [-0.1219, 0.0087, 0.0325, ..., -0.1030, 0.0113, -0.0175], + [ 0.0309, -0.0047, -0.0515, ..., -0.2170, -0.4692, -0.3170], + [-0.1410, -0.1190, 0.0104, ..., 0.1578, 0.2296, 0.2565]], + device='cuda:0'), grad: tensor([[-9.2316e-03, -2.4997e-06, -2.7232e-06, ..., -2.8920e-04, + -2.9588e-04, -4.0092e-03], + [ 8.7118e-04, 3.7253e-09, 5.5879e-09, ..., 2.5928e-05, + 2.4945e-05, 3.8552e-04], + [ 2.5501e-03, 7.6368e-08, 8.2888e-08, ..., 7.6234e-05, + 7.3671e-05, 1.1272e-03], + ..., + [ 2.2068e-03, 2.5406e-06, 2.7083e-06, ..., 8.0943e-05, + 9.5069e-05, 8.9645e-04], + [ 1.1528e-04, -1.2387e-07, -7.8231e-08, ..., 2.2985e-06, + 2.4829e-06, 5.5432e-05], + [ 3.6478e-04, 9.3132e-10, 9.3132e-10, ..., 1.0841e-05, + 1.0408e-05, 1.6153e-04]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.1000, -0.0150, 0.1328, 0.1201, 0.1050, -0.4625, -0.2734], + device='cuda:0'), grad: tensor([-0.1293, 0.0125, 0.0365, 0.0449, 0.0281, 0.0019, 0.0052], + device='cuda:0') +588 +0.007754484907260515 +changing lr +epoch 22, time 781.63, cls_loss 0.6161 cls_loss_mapping 1.9454 cls_loss_causal 1.4837 re_mapping 0.0307 re_causal 0.0308 /// teacc 92.71 lr 0.00756450 +Epoch 24, weight, value: tensor([[ 0.0800, 0.0338, -0.0104, ..., -0.0265, 0.0049, -0.0452], + [-0.0125, -0.0722, -0.0557, ..., 0.0961, -0.0114, 0.0456], + [ 0.0486, 0.1229, -0.0028, ..., 0.0315, -0.0277, -0.0674], + ..., + [-0.1049, 0.0251, 0.0667, ..., -0.0633, 0.0767, 0.0181], + [ 0.0445, -0.0205, -0.1205, ..., -0.2981, -0.5527, -0.2694], + [-0.1630, -0.1408, -0.0178, ..., 0.1374, 0.2132, 0.2691]], + device='cuda:0'), grad: tensor([[-2.1698e-02, -1.6994e-03, -1.8892e-03, ..., -3.2234e-03, + -3.4847e-03, -6.6299e-03], + [ 6.0806e-03, 1.1034e-03, 9.0456e-04, ..., 5.0402e-04, + 5.3978e-04, 1.4162e-03], + [ 4.6492e-04, 1.9938e-05, 2.9132e-05, ..., 7.8738e-05, + 8.5354e-05, 1.5247e-04], + ..., + [ 1.5152e-02, 5.7650e-04, 9.5654e-04, ..., 2.6398e-03, + 2.8572e-03, 5.0621e-03], + [ 3.9302e-06, 3.3528e-07, 3.3528e-07, ..., 6.7987e-07, + 7.1898e-07, 1.2740e-06], + [ 6.8359e-07, 2.4214e-08, 5.7742e-08, ..., 1.2852e-07, + 1.3784e-07, 2.4214e-07]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0437, 0.0075, 0.1359, 0.0993, 0.1586, -0.4476, -0.2938], + device='cuda:0'), grad: tensor([-8.7830e-02, 2.2202e-02, 1.8454e-03, 1.7583e-05, 6.3782e-02, + 1.9386e-05, 3.9339e-06], device='cuda:0') +588 +0.007564496387029534 +changing lr +epoch 23, time 787.72, cls_loss 0.6427 cls_loss_mapping 1.9558 cls_loss_causal 1.4656 re_mapping 0.0319 re_causal 0.0319 /// teacc 90.20 lr 0.00736934 +Epoch 25, weight, value: tensor([[ 8.8193e-02, 1.4741e-01, 9.0724e-02, ..., -2.2174e-02, + 3.3019e-02, -2.3089e-02], + [-1.1034e-02, -1.3567e-01, -1.3019e-01, ..., 1.1658e-01, + 6.8575e-05, 5.1640e-02], + [ 3.0338e-02, 1.3420e-01, -3.2888e-03, ..., 6.7560e-02, + 4.6044e-03, -5.2422e-02], + ..., + [-1.0876e-01, -5.3759e-02, 3.3776e-02, ..., -6.1071e-02, + 6.4407e-02, 1.8074e-02], + [ 5.2106e-02, -4.3553e-02, -1.6549e-01, ..., -2.9827e-01, + -5.7158e-01, -3.1876e-01], + [-1.3325e-01, -1.0131e-01, 1.2952e-02, ..., 2.7104e-02, + 1.2206e-01, 1.7748e-01]], device='cuda:0'), grad: tensor([[ 7.9834e-02, 3.9253e-03, 1.5764e-03, ..., 5.5695e-03, + 3.1586e-03, 1.2199e-02], + [ 6.2294e-03, 3.2115e-04, 2.3818e-04, ..., 5.8126e-04, + 2.8348e-04, 3.3875e-03], + [ 1.8967e-02, 1.2369e-03, 5.0497e-04, ..., 1.8320e-03, + 7.5245e-04, 4.6806e-03], + ..., + [-1.0699e-01, -5.6076e-03, -2.4185e-03, ..., -8.2397e-03, + -4.3335e-03, -2.1759e-02], + [ 1.0139e-04, 6.9030e-06, 5.7854e-06, ..., 1.5438e-05, + 8.5831e-06, 8.9586e-05], + [ 4.5633e-04, 1.9133e-05, 1.2949e-05, ..., 2.7701e-05, + 1.1049e-05, 1.6868e-04]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.1200, -0.0290, 0.1008, 0.1543, 0.1530, -0.4763, -0.3745], + device='cuda:0'), grad: tensor([ 0.1738, 0.0605, 0.0690, 0.0209, -0.3291, 0.0015, 0.0032], + device='cuda:0') +588 +0.007369343312364995 +changing lr +epoch 24, time 788.55, cls_loss 1.1315 cls_loss_mapping 3.4355 cls_loss_causal 2.2370 re_mapping 0.3131 re_causal 0.0427 /// teacc 89.20 lr 0.00716942 +Epoch 26, weight, value: tensor([[ 0.0878, 0.1991, 0.1638, ..., -0.0773, -0.0426, -0.0493], + [-0.0240, -0.1001, -0.0669, ..., 0.0547, -0.0221, 0.0230], + [ 0.0336, 0.0375, -0.1346, ..., 0.0546, 0.0035, -0.0137], + ..., + [-0.1058, -0.1054, -0.0226, ..., -0.0700, 0.0644, 0.0029], + [ 0.0710, 0.0174, -0.0849, ..., -0.2391, -0.5537, -0.2359], + [-0.1200, -0.0744, -0.0111, ..., 0.0516, 0.1376, 0.1838]], + device='cuda:0'), grad: tensor([[-2.4857e-02, -2.3329e-04, 1.6689e-04, ..., -1.2379e-03, + -4.9925e-04, -5.0964e-03], + [ 3.0079e-03, 8.9586e-05, 4.6998e-05, ..., 2.2125e-04, + 1.1927e-04, 8.1253e-04], + [-2.0187e-02, -1.2960e-03, -1.0767e-03, ..., -2.2945e-03, + -1.4668e-03, -7.6714e-03], + ..., + [ 3.5126e-02, 1.1816e-03, 6.9714e-04, ..., 2.7428e-03, + 1.5230e-03, 9.9258e-03], + [ 1.9932e-03, 7.1108e-05, 4.3988e-05, ..., 1.6034e-04, + 9.0301e-05, 5.7602e-04], + [ 4.5128e-03, 1.7214e-04, 1.1188e-04, ..., 3.7599e-04, + 2.1517e-04, 1.3399e-03]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.1609, -0.0482, 0.1062, 0.0832, 0.1208, -0.3899, -0.3219], + device='cuda:0'), grad: tensor([-0.1237, 0.0164, -0.1273, 0.0023, 0.1954, 0.0112, 0.0256], + device='cuda:0') +588 +0.0071694186955877925 +changing lr +epoch 25, time 785.12, cls_loss 0.5741 cls_loss_mapping 1.9114 cls_loss_causal 1.6204 re_mapping 0.0319 re_causal 0.0319 /// teacc 91.96 lr 0.00696513 +Epoch 27, weight, value: tensor([[ 0.0982, 0.1648, 0.1783, ..., -0.1124, -0.0876, -0.0807], + [-0.0085, 0.0185, 0.0316, ..., 0.0832, 0.0404, 0.0526], + [ 0.0371, -0.0723, -0.2495, ..., 0.0228, -0.0261, -0.0346], + ..., + [-0.1209, -0.0259, 0.0448, ..., -0.0154, 0.1043, 0.0033], + [ 0.0403, -0.0464, -0.1138, ..., -0.2935, -0.5336, -0.2503], + [-0.1358, -0.0385, -0.0302, ..., -0.0237, 0.0271, 0.1576]], + device='cuda:0'), grad: tensor([[ 8.6899e-03, 1.0281e-03, 7.6437e-04, ..., 1.4591e-03, + 7.8630e-04, 3.8605e-03], + [ 1.4791e-03, 1.9038e-04, 1.5974e-04, ..., 2.3341e-04, + 1.0222e-04, 6.7043e-04], + [-3.6316e-02, -3.6621e-03, -1.9283e-03, ..., -6.8626e-03, + -4.6654e-03, -1.5869e-02], + ..., + [ 2.5116e-02, 2.3098e-03, 8.9169e-04, ..., 5.0087e-03, + 3.7041e-03, 1.0880e-02], + [ 9.7573e-05, 1.2517e-05, 1.0505e-05, ..., 1.5259e-05, + 6.7465e-06, 4.3809e-05], + [ 2.2125e-04, 2.8417e-05, 2.3842e-05, ..., 3.4600e-05, + 1.5274e-05, 9.9421e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.1745, -0.0293, 0.0722, 0.1363, 0.1182, -0.4579, -0.3441], + device='cuda:0'), grad: tensor([ 0.0941, 0.0185, -0.2969, 0.0090, 0.1716, 0.0012, 0.0027], + device='cuda:0') +588 +0.0069651251582696205 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 792.61, cls_loss 0.5976 cls_loss_mapping 1.9169 cls_loss_causal 1.5012 re_mapping 0.0282 re_causal 0.0280 /// teacc 93.47 lr 0.00675687 +Epoch 28, weight, value: tensor([[ 0.0849, 0.2158, 0.1909, ..., -0.0670, -0.0224, -0.0653], + [-0.0050, -0.0781, 0.0187, ..., 0.0143, -0.0333, 0.0300], + [ 0.0321, -0.0056, -0.1753, ..., 0.1102, 0.0497, -0.0252], + ..., + [-0.1299, -0.1128, -0.0702, ..., -0.0092, 0.1033, 0.0364], + [ 0.0531, -0.0508, -0.1309, ..., -0.4602, -0.7650, -0.3344], + [-0.1232, 0.0577, 0.0606, ..., -0.1415, -0.1076, 0.1019]], + device='cuda:0'), grad: tensor([[ 1.3847e-02, 1.0290e-03, 7.7295e-04, ..., 2.3003e-03, + 2.2545e-03, 3.3627e-03], + [ 3.4165e-04, 1.9178e-05, 1.6481e-05, ..., 8.1480e-05, + 7.8142e-05, 1.5199e-04], + [ 1.6439e-04, 9.2834e-06, 7.9498e-06, ..., 3.8981e-05, + 3.7402e-05, 7.2539e-05], + ..., + [-1.4359e-02, -1.0576e-03, -7.9727e-04, ..., -2.4204e-03, + -2.3708e-03, -3.5896e-03], + [ 7.5903e-07, 4.2841e-08, 3.6322e-08, ..., 1.8161e-07, + 1.7416e-07, 3.3993e-07], + [ 2.0023e-07, 1.1176e-08, 9.3132e-09, ..., 4.7497e-08, + 4.5635e-08, 8.9407e-08]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.1484, 0.0285, 0.0899, 0.1157, 0.1162, -0.5097, -0.3546], + device='cuda:0'), grad: tensor([ 4.8004e-02, 1.9236e-03, 9.1887e-04, 1.1250e-05, -5.0842e-02, + 4.2990e-06, 1.1325e-06], device='cuda:0') +588 +0.006756874120406716 +changing lr +epoch 27, time 781.38, cls_loss 0.5419 cls_loss_mapping 1.9509 cls_loss_causal 1.3682 re_mapping 0.0263 re_causal 0.0265 /// teacc 90.95 lr 0.00654508 +Epoch 29, weight, value: tensor([[ 0.0795, 0.1775, 0.1434, ..., -0.0355, 0.0117, -0.0423], + [-0.0027, -0.1031, -0.0182, ..., -0.0129, -0.0968, -0.0158], + [ 0.0158, -0.0338, -0.2434, ..., 0.1364, 0.0689, 0.0018], + ..., + [-0.1121, -0.0934, -0.0026, ..., -0.0365, 0.1038, 0.0124], + [ 0.0433, 0.0887, -0.0647, ..., -0.4377, -0.7782, -0.3282], + [-0.1288, 0.1385, 0.1747, ..., -0.1319, -0.0666, 0.1398]], + device='cuda:0'), grad: tensor([[-3.0937e-03, -1.1024e-03, -2.4188e-04, ..., -2.1133e-03, + -3.8552e-04, -3.2291e-03], + [-7.0740e-02, -4.3640e-03, -1.5545e-03, ..., -1.3870e-02, + -1.0757e-02, -2.3834e-02], + [ 3.1395e-03, 1.0872e-03, 2.0838e-04, ..., 2.7447e-03, + 1.3008e-03, 4.4403e-03], + ..., + [ 6.3599e-02, 2.0905e-03, 1.1339e-03, ..., 7.3051e-03, + 6.9160e-03, 1.3115e-02], + [ 1.3781e-04, 5.1200e-05, 9.2462e-06, ..., 1.2922e-04, + 6.1989e-05, 2.1267e-04], + [ 4.6730e-03, 1.5717e-03, 3.0494e-04, ..., 4.0703e-03, + 2.0103e-03, 6.5575e-03]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.1206, 0.0119, 0.1236, 0.1193, 0.1013, -0.4441, -0.3368], + device='cuda:0'), grad: tensor([-0.1005, -0.4434, 0.0994, 0.0629, 0.2317, 0.0046, 0.1451], + device='cuda:0') +588 +0.00654508497187474 +changing lr +epoch 28, time 788.92, cls_loss 0.5397 cls_loss_mapping 1.9572 cls_loss_causal 1.3194 re_mapping 0.0269 re_causal 0.0270 /// teacc 93.47 lr 0.00633018 +Epoch 30, weight, value: tensor([[ 0.0970, 0.1313, 0.1570, ..., -0.0301, 0.0310, -0.0383], + [-0.0013, -0.0586, -0.0991, ..., 0.0811, -0.0720, 0.0178], + [ 0.0013, -0.0700, -0.1946, ..., 0.1307, 0.1063, 0.0015], + ..., + [-0.1193, -0.0050, -0.0050, ..., -0.0915, 0.0110, -0.0163], + [ 0.0203, 0.0178, -0.0341, ..., -0.3808, -0.6659, -0.3106], + [-0.1217, 0.0976, 0.1366, ..., -0.1656, -0.0705, 0.1269]], + device='cuda:0'), grad: tensor([[ 1.2390e-02, 4.3106e-04, 4.7064e-04, ..., 2.0142e-03, + 1.6708e-03, 2.8572e-03], + [-7.1373e-03, -1.6105e-04, -4.0865e-04, ..., -1.4248e-03, + -1.5392e-03, -2.0161e-03], + [-3.9825e-02, -1.9464e-03, -4.4966e-04, ..., -4.2038e-03, + -9.2793e-04, -6.1264e-03], + ..., + [ 1.8425e-03, 1.1265e-04, 2.6941e-05, ..., 2.5678e-04, + 6.4135e-05, 3.4761e-04], + [ 5.1707e-06, 2.8964e-07, 6.7987e-08, ..., 6.4541e-07, + 1.5274e-07, 8.9779e-07], + [ 6.6643e-03, 3.2306e-04, 7.4565e-05, ..., 6.9666e-04, + 1.5306e-04, 1.0176e-03]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0586, 0.1363, 0.1159, 0.1093, 0.1186, -0.4779, -0.4230], + device='cuda:0'), grad: tensor([ 3.7048e-02, 1.9970e-03, -2.8271e-01, 1.8079e-01, 1.5976e-02, + 4.1485e-05, 4.6997e-02], device='cuda:0') +588 +0.006330184227833378 +changing lr +epoch 29, time 788.04, cls_loss 0.5340 cls_loss_mapping 1.9680 cls_loss_causal 1.2786 re_mapping 0.0260 re_causal 0.0260 /// teacc 91.71 lr 0.00611260 +Epoch 31, weight, value: tensor([[ 0.0813, 0.1521, 0.2056, ..., -0.1037, -0.0121, -0.0705], + [-0.0157, -0.1477, -0.0928, ..., 0.0127, -0.1391, -0.0113], + [ 0.0296, 0.0112, -0.1848, ..., 0.1852, 0.1251, 0.0714], + ..., + [-0.1151, -0.0392, -0.0127, ..., -0.0562, 0.0719, -0.0291], + [ 0.0647, 0.0228, 0.0058, ..., -0.3576, -0.5895, -0.2920], + [-0.1091, 0.0916, 0.1145, ..., -0.1134, -0.0486, 0.1463]], + device='cuda:0'), grad: tensor([[ 7.5340e-04, 1.0788e-04, 2.8402e-05, ..., 3.4952e-04, + 9.8467e-05, 4.8780e-04], + [ 5.1260e-04, 7.3433e-05, 1.8924e-05, ..., 2.3699e-04, + 6.5148e-05, 3.3212e-04], + [-2.8610e-03, -4.9025e-05, -1.8671e-05, ..., -1.0020e-04, + -2.3615e-04, -1.5295e-04], + ..., + [ 1.2331e-03, 1.4114e-04, 3.7640e-05, ..., 4.5133e-04, + 1.4722e-04, 6.3133e-04], + [-3.5000e-03, -5.8031e-04, -1.5092e-04, ..., -1.8902e-03, + -4.8470e-04, -2.6379e-03], + [ 2.4815e-03, 1.8239e-04, 5.0724e-05, ..., 5.6219e-04, + 2.5821e-04, 7.9155e-04]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0227, 0.1434, 0.2009, 0.0940, 0.0847, -0.4690, -0.4327], + device='cuda:0'), grad: tensor([ 0.0267, 0.0183, 0.0109, 0.0261, 0.0327, -0.1489, 0.0341], + device='cuda:0') +588 +0.006112604669781575 +changing lr +epoch 30, time 788.91, cls_loss 0.4965 cls_loss_mapping 1.9696 cls_loss_causal 1.3076 re_mapping 0.0263 re_causal 0.0259 /// teacc 84.17 lr 0.00589278 +Epoch 32, weight, value: tensor([[ 0.0793, 0.1248, 0.1654, ..., -0.0718, -0.0028, -0.0436], + [-0.0039, -0.1375, -0.0999, ..., -0.0187, -0.1632, -0.0308], + [ 0.0468, -0.0828, -0.1934, ..., 0.1889, 0.1445, 0.0811], + ..., + [-0.1174, 0.0410, 0.0460, ..., -0.0418, 0.0861, -0.0411], + [ 0.0302, -0.0699, -0.2443, ..., -0.4236, -0.6916, -0.3525], + [-0.1332, 0.1385, 0.1457, ..., -0.1190, -0.0400, 0.1374]], + device='cuda:0'), grad: tensor([[ 2.1820e-03, 5.9557e-04, 5.0735e-04, ..., 2.0466e-03, + 1.4563e-03, 2.0466e-03], + [ 1.7691e-03, 4.8542e-04, 4.1270e-04, ..., 1.6298e-03, + 1.1492e-03, 1.6193e-03], + [ 2.6188e-03, 7.2193e-04, 6.1226e-04, ..., 2.3689e-03, + 1.6546e-03, 2.3384e-03], + ..., + [-4.1771e-03, -9.4938e-04, -8.8930e-04, ..., -6.1531e-03, + -5.1956e-03, -6.9199e-03], + [ 3.7408e-04, 1.1033e-04, 9.0599e-05, ..., 2.5511e-04, + 1.4651e-04, 2.2233e-04], + [ 1.0471e-03, 2.9111e-04, 2.4605e-04, ..., 9.1887e-04, + 6.3086e-04, 8.9741e-04]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0680, 0.0919, 0.1895, 0.0895, 0.0840, -0.4686, -0.4053], + device='cuda:0'), grad: tensor([ 0.0666, 0.0542, 0.0806, -0.1348, -0.1113, 0.0121, 0.0324], + device='cuda:0') +588 +0.005892784473993186 +changing lr +epoch 31, time 785.58, cls_loss 0.4924 cls_loss_mapping 1.9799 cls_loss_causal 1.2812 re_mapping 0.0287 re_causal 0.0285 /// teacc 90.95 lr 0.00567117 +Epoch 33, weight, value: tensor([[ 0.0898, 0.1106, 0.1689, ..., -0.0609, -0.0084, -0.0501], + [ 0.0060, -0.1027, -0.1031, ..., -0.0119, -0.1542, -0.0428], + [ 0.0192, -0.1245, -0.3801, ..., 0.2343, 0.1420, 0.0927], + ..., + [-0.1043, 0.1256, 0.1974, ..., -0.0666, 0.0506, -0.0522], + [ 0.0356, -0.1164, -0.2787, ..., -0.3279, -0.5156, -0.2662], + [-0.1434, 0.0185, 0.1391, ..., -0.1208, 0.0278, 0.1511]], + device='cuda:0'), grad: tensor([[-4.6921e-04, 3.9339e-05, -1.3277e-05, ..., -5.7332e-06, + -2.6608e-04, -4.8786e-05], + [ 2.9011e-03, 2.3961e-04, 1.2946e-04, ..., 7.8297e-04, + 8.8334e-05, 7.7534e-04], + [-5.1460e-03, -5.4550e-04, -2.3949e-04, ..., -1.6346e-03, + -6.1452e-05, -1.6384e-03], + ..., + [ 1.8759e-03, 1.7905e-04, 8.4639e-05, ..., 5.9032e-04, + 2.0921e-04, 6.3896e-04], + [ 2.0885e-04, 2.1845e-05, 9.6709e-06, ..., 6.6757e-05, + 7.5661e-06, 6.8247e-05], + [ 1.1289e-04, 1.1832e-05, 5.2303e-06, ..., 3.6031e-05, + 3.7141e-06, 3.6776e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0389, 0.1217, 0.2248, 0.1206, 0.0889, -0.5410, -0.4683], + device='cuda:0'), grad: tensor([ 0.0138, 0.0526, -0.1074, 0.0098, 0.0250, 0.0040, 0.0022], + device='cuda:0') +588 +0.00567116632908828 +changing lr +epoch 32, time 787.12, cls_loss 0.5030 cls_loss_mapping 2.0219 cls_loss_causal 1.2584 re_mapping 0.0316 re_causal 0.0280 /// teacc 92.96 lr 0.00544820 +Epoch 34, weight, value: tensor([[ 0.0866, 0.0697, 0.0957, ..., -0.0776, -0.0634, -0.0689], + [-0.0064, -0.0470, -0.0392, ..., 0.0678, -0.0588, 0.0327], + [ 0.0253, -0.1556, -0.3172, ..., 0.1847, 0.1656, 0.0918], + ..., + [-0.1022, 0.1191, 0.1856, ..., -0.0881, 0.0141, -0.0924], + [ 0.0263, -0.1992, -0.3390, ..., -0.3390, -0.5622, -0.2738], + [-0.1282, 0.1198, 0.1879, ..., -0.0373, 0.1293, 0.2007]], + device='cuda:0'), grad: tensor([[-1.9852e-02, -2.6722e-03, -1.2140e-03, ..., -2.3937e-03, + -2.1324e-03, -5.3711e-03], + [ 3.6061e-05, 4.8392e-06, 2.1979e-06, ..., 4.3474e-06, + 3.8706e-06, 9.7379e-06], + [ 1.9806e-02, 2.6569e-03, 1.2074e-03, ..., 2.3880e-03, + 2.1267e-03, 5.3482e-03], + ..., + [ 1.4916e-05, 8.9556e-06, 4.2692e-06, ..., 2.6170e-07, + 1.8673e-06, 1.1876e-05], + [ 1.4035e-06, 1.8836e-07, 8.5682e-08, ..., 1.6927e-07, + 1.5064e-07, 3.7928e-07], + [ 1.3597e-06, 1.8254e-07, 8.2888e-08, ..., 1.6391e-07, + 1.4598e-07, 3.6741e-07]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0526, 0.1499, 0.2116, 0.1250, 0.0584, -0.6006, -0.4359], + device='cuda:0'), grad: tensor([-1.5857e-01, 2.8729e-04, 1.5771e-01, 2.5004e-05, 5.0068e-04, + 1.1183e-05, 1.0833e-05], device='cuda:0') +588 +0.00544819654451717 +changing lr +epoch 33, time 786.99, cls_loss 0.4777 cls_loss_mapping 1.9825 cls_loss_causal 1.2119 re_mapping 0.0252 re_causal 0.0246 /// teacc 93.47 lr 0.00522432 +Epoch 35, weight, value: tensor([[ 9.6564e-02, 8.6402e-02, 1.2515e-01, ..., -5.9213e-02, + -1.7374e-02, -7.3853e-02], + [-8.0869e-03, -1.5768e-01, -1.4085e-01, ..., 2.2443e-02, + -9.8792e-02, 4.2673e-03], + [ 2.9069e-02, -9.4800e-02, -1.9037e-01, ..., 1.5574e-01, + 1.3214e-01, 4.1642e-02], + ..., + [-1.1820e-01, 1.0329e-01, 1.1748e-01, ..., -7.6169e-02, + -1.8759e-04, -6.7465e-02], + [-3.4037e-02, -2.1797e-01, -3.6182e-01, ..., -2.6841e-01, + -5.0084e-01, -2.3240e-01], + [-9.9478e-02, 1.6780e-01, 2.2658e-01, ..., 2.1547e-03, + 1.7345e-01, 2.7692e-01]], device='cuda:0'), grad: tensor([[ 1.1641e-04, 8.7097e-06, -1.0347e-06, ..., 3.1918e-05, + 2.6524e-05, 4.3392e-05], + [ 1.6347e-05, 3.0138e-06, 1.6927e-07, ..., 1.4566e-05, + 1.0543e-05, 1.8179e-05], + [ 6.7444e-03, 1.2379e-03, 7.1824e-05, ..., 5.9738e-03, + 4.3259e-03, 7.4615e-03], + ..., + [-6.8855e-03, -1.2512e-03, -7.0989e-05, ..., -6.0272e-03, + -4.3678e-03, -7.5302e-03], + [ 1.3597e-07, 2.5146e-08, 1.3970e-09, ..., 1.2107e-07, + 8.7777e-08, 1.5134e-07], + [ 1.5199e-06, 2.8033e-07, 1.5832e-08, ..., 1.3551e-06, + 9.8068e-07, 1.6913e-06]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0102, 0.1092, 0.2120, 0.1399, 0.0852, -0.5872, -0.3527], + device='cuda:0'), grad: tensor([ 7.7581e-04, 4.0340e-04, 1.6553e-01, 5.3763e-05, -1.6687e-01, + 3.3565e-06, 3.7521e-05], device='cuda:0') +588 +0.005224324151752577 +changing lr +epoch 34, time 789.07, cls_loss 0.4744 cls_loss_mapping 1.9965 cls_loss_causal 1.1882 re_mapping 0.0266 re_causal 0.0262 /// teacc 93.47 lr 0.00500000 +Epoch 36, weight, value: tensor([[ 0.0967, 0.1221, 0.1184, ..., -0.0430, -0.0615, -0.0757], + [-0.0043, -0.1667, -0.1240, ..., 0.0556, -0.0596, 0.0331], + [ 0.0120, -0.0279, -0.1300, ..., 0.1821, 0.2069, 0.0506], + ..., + [-0.0990, 0.0124, 0.0623, ..., -0.1354, -0.0033, -0.0557], + [ 0.0066, -0.1723, -0.2741, ..., -0.2797, -0.4600, -0.2167], + [-0.1336, 0.1297, 0.1504, ..., -0.0877, 0.0066, 0.1904]], + device='cuda:0'), grad: tensor([[ 4.1428e-03, -8.2970e-04, -6.4659e-04, ..., -8.5592e-04, + 1.7866e-05, -6.2180e-04], + [-9.0790e-03, 2.1160e-04, 1.9991e-04, ..., -8.9359e-04, + -9.9277e-04, -1.2178e-03], + [ 9.3842e-04, 1.3697e-04, 9.9242e-05, ..., 3.8648e-04, + 2.1183e-04, 4.0460e-04], + ..., + [ 2.7332e-03, 3.2091e-04, 2.3162e-04, ..., 9.1791e-04, + 5.1689e-04, 9.6798e-04], + [ 4.9591e-04, 6.2704e-05, 4.5419e-05, ..., 1.7476e-04, + 9.6738e-05, 1.8334e-04], + [ 5.3793e-05, 6.8285e-06, 4.9360e-06, ..., 1.8999e-05, + 1.0498e-05, 1.9938e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0993, 0.0918, 0.1681, 0.0937, 0.1258, -0.6833, -0.3940], + device='cuda:0'), grad: tensor([-0.0761, 0.0035, 0.0164, 0.0104, 0.0376, 0.0073, 0.0008], + device='cuda:0') +588 +0.005000000000000003 +changing lr +epoch 35, time 792.03, cls_loss 0.4560 cls_loss_mapping 1.9909 cls_loss_causal 1.1629 re_mapping 0.0249 re_causal 0.0246 /// teacc 88.69 lr 0.00477568 +Epoch 37, weight, value: tensor([[ 0.1013, 0.1890, 0.1736, ..., -0.0483, -0.0621, -0.0633], + [-0.0103, -0.1278, -0.1001, ..., 0.0404, -0.1113, 0.0351], + [ 0.0370, -0.0673, -0.1721, ..., 0.1677, 0.1644, 0.0353], + ..., + [-0.1176, -0.0319, -0.0022, ..., -0.1089, 0.0363, -0.0545], + [-0.0070, -0.0855, -0.1436, ..., -0.3884, -0.5403, -0.2805], + [-0.1227, 0.0629, 0.0988, ..., -0.0030, 0.1669, 0.2495]], + device='cuda:0'), grad: tensor([[ 4.1515e-05, 1.1928e-05, 5.3421e-06, ..., 1.5631e-05, + 2.5257e-06, 2.4036e-05], + [ 3.3062e-06, 9.8906e-07, 4.3958e-07, ..., 1.2908e-06, + 1.9628e-07, 1.9856e-06], + [ 9.0480e-05, 4.7386e-06, 3.6545e-06, ..., 6.7763e-06, + 5.5656e-06, 1.1995e-05], + ..., + [-9.1314e-05, -4.3660e-06, -3.5316e-06, ..., -6.3032e-06, + -5.6215e-06, -1.1317e-05], + [ 1.9395e-07, 5.8906e-08, 2.6077e-08, ..., 7.7067e-08, + 1.1874e-08, 1.1805e-07], + [-4.4197e-05, -1.3396e-05, -5.9493e-06, ..., -1.7524e-05, + -2.6841e-06, -2.6911e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.1112, 0.1319, 0.1627, 0.0850, 0.1299, -0.6363, -0.5115], + device='cuda:0'), grad: tensor([ 1.3371e-03, 1.1128e-04, 3.1662e-04, 6.9439e-06, -2.6846e-04, + 6.6236e-06, -1.5087e-03], device='cuda:0') +588 +0.004775675848247429 +changing lr +epoch 36, time 796.83, cls_loss 0.4659 cls_loss_mapping 1.9866 cls_loss_causal 1.2009 re_mapping 0.0260 re_causal 0.0257 /// teacc 91.46 lr 0.00455180 +Epoch 38, weight, value: tensor([[ 0.0813, 0.1262, 0.0609, ..., -0.0690, -0.0698, -0.0691], + [-0.0043, -0.1431, -0.1101, ..., 0.0090, -0.1583, 0.0231], + [ 0.0284, -0.0451, -0.1768, ..., 0.2120, 0.2247, 0.0699], + ..., + [-0.1171, 0.0164, 0.0747, ..., -0.0878, 0.0124, -0.0669], + [ 0.0369, -0.0323, -0.0879, ..., -0.3959, -0.4334, -0.2563], + [-0.1201, 0.0114, 0.0562, ..., 0.0023, 0.1179, 0.2207]], + device='cuda:0'), grad: tensor([[ 1.6708e-03, 2.5201e-04, 3.8184e-06, ..., 4.5013e-04, + 5.3227e-05, 7.4673e-04], + [-7.2289e-03, -1.0662e-03, -1.5929e-05, ..., -1.9140e-03, + -2.2995e-04, -3.2253e-03], + [ 2.1782e-03, 3.4761e-04, 5.4501e-06, ..., 6.1274e-04, + 6.9618e-05, 9.7752e-04], + ..., + [ 1.0767e-03, 1.4424e-04, 2.0191e-06, ..., 2.6560e-04, + 3.4064e-05, 4.7731e-04], + [ 2.9540e-04, 4.0531e-05, 5.7695e-07, ..., 7.4089e-05, + 9.3505e-06, 1.3113e-04], + [ 6.6996e-04, 9.0063e-05, 1.2629e-06, ..., 1.6558e-04, + 2.1189e-05, 2.9683e-04]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0797, 0.1661, 0.1733, 0.1131, 0.0780, -0.6217, -0.4580], + device='cuda:0'), grad: tensor([ 0.0606, -0.2644, 0.0772, 0.0495, 0.0407, 0.0111, 0.0253], + device='cuda:0') +588 +0.004551803455482836 +changing lr +epoch 37, time 793.56, cls_loss 0.4088 cls_loss_mapping 1.9945 cls_loss_causal 1.1390 re_mapping 0.0272 re_causal 0.0267 /// teacc 90.45 lr 0.00432883 +Epoch 39, weight, value: tensor([[ 0.0832, 0.0943, 0.0779, ..., -0.1074, -0.1026, -0.0777], + [ 0.0134, -0.0943, -0.1052, ..., 0.0421, -0.1147, 0.0317], + [ 0.0184, -0.1020, -0.2025, ..., 0.1754, 0.2036, 0.0666], + ..., + [-0.1180, -0.0023, 0.0671, ..., -0.0080, 0.0718, -0.0207], + [ 0.0664, 0.0084, -0.0064, ..., -0.4826, -0.5101, -0.3272], + [-0.1370, 0.0443, 0.0230, ..., -0.0018, 0.1435, 0.1990]], + device='cuda:0'), grad: tensor([[ 1.1313e-04, 4.7946e-04, 1.7059e-04, ..., 7.0810e-04, + 3.7551e-04, 6.2132e-04], + [-1.2312e-03, -5.7936e-04, -3.7885e-04, ..., -2.7731e-05, + -1.4746e-04, -2.9731e-04], + [ 1.2197e-03, 4.2462e-04, 1.5688e-04, ..., 8.1348e-04, + 4.8971e-04, 7.3385e-04], + ..., + [-1.9798e-03, -9.7084e-04, -1.7929e-04, ..., -2.7885e-03, + -1.4896e-03, -2.2106e-03], + [ 1.3804e-04, 4.9263e-05, 1.9655e-05, ..., 8.4996e-05, + 5.2363e-05, 7.9095e-05], + [ 2.9349e-04, 1.0264e-04, 3.8415e-05, ..., 1.9336e-04, + 1.1677e-04, 1.7524e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0425, 0.1473, 0.2778, 0.1131, 0.0742, -0.6858, -0.4781], + device='cuda:0'), grad: tensor([ 0.0753, -0.1073, 0.0614, 0.0701, -0.1216, 0.0073, 0.0149], + device='cuda:0') +588 +0.004328833670911726 +changing lr +epoch 38, time 790.75, cls_loss 0.4256 cls_loss_mapping 1.9899 cls_loss_causal 1.1630 re_mapping 0.0241 re_causal 0.0242 /// teacc 93.22 lr 0.00410722 +Epoch 40, weight, value: tensor([[ 8.6891e-02, 1.3303e-01, 6.1671e-02, ..., -9.0542e-02, + -7.1304e-02, -4.9730e-02], + [ 2.7732e-04, -6.9901e-02, -3.5240e-02, ..., -3.8250e-02, + -2.0263e-01, -5.6416e-02], + [ 6.6434e-03, -1.1944e-01, -2.1008e-01, ..., 1.9276e-01, + 2.7706e-01, 9.6234e-02], + ..., + [-9.2859e-02, -3.6517e-02, 2.1713e-02, ..., -5.2031e-03, + 6.0510e-02, -1.6178e-02], + [ 2.0870e-02, 4.6258e-02, 2.4939e-02, ..., -4.9472e-01, + -5.3191e-01, -2.9663e-01], + [-1.3070e-01, 2.9015e-02, 4.0670e-02, ..., 4.8675e-04, + 1.3944e-01, 1.7128e-01]], device='cuda:0'), grad: tensor([[ 5.8222e-04, 1.1253e-04, 7.4208e-05, ..., 3.7336e-04, + 1.4448e-04, 8.3542e-04], + [-5.7638e-05, 6.7353e-05, 4.9293e-05, ..., 2.5773e-04, + 1.7136e-05, 6.0368e-04], + [ 3.4451e-04, 7.1108e-05, 4.7863e-05, ..., 2.4331e-04, + 8.1003e-05, 5.4121e-04], + ..., + [ 7.5817e-04, 1.1164e-04, 7.3910e-05, ..., 3.7551e-04, + 1.5259e-04, 8.1205e-04], + [ 5.5224e-05, 1.1846e-05, 7.9870e-06, ..., 4.0680e-05, + 1.3225e-05, 9.0718e-05], + [ 4.4316e-05, 8.5086e-06, 5.6922e-06, ..., 2.8893e-05, + 1.0200e-05, 6.4075e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0364, 0.1547, 0.1998, 0.1858, 0.1211, -0.7020, -0.5193], + device='cuda:0'), grad: tensor([ 0.0376, 0.0268, 0.0241, -0.1310, 0.0356, 0.0040, 0.0028], + device='cuda:0') +588 +0.0041072155260068206 +changing lr +epoch 39, time 797.91, cls_loss 0.4333 cls_loss_mapping 2.0248 cls_loss_causal 1.1616 re_mapping 0.0283 re_causal 0.0232 /// teacc 87.44 lr 0.00388740 +Epoch 41, weight, value: tensor([[ 0.0911, 0.0764, 0.0081, ..., -0.1578, -0.1109, -0.1089], + [ 0.0021, -0.1234, -0.0031, ..., -0.0790, -0.1233, -0.0430], + [ 0.0287, -0.1029, -0.2422, ..., 0.2096, 0.2480, 0.1162], + ..., + [-0.1239, 0.0178, 0.0969, ..., 0.0444, 0.0611, 0.0067], + [ 0.0503, 0.0292, -0.0074, ..., -0.5563, -0.6339, -0.3520], + [-0.1007, 0.0255, 0.0523, ..., 0.0263, 0.2143, 0.2079]], + device='cuda:0'), grad: tensor([[ 3.7217e-04, 1.0151e-04, 1.0008e-04, ..., 6.9761e-04, + 6.1178e-04, 8.0538e-04], + [ 1.4484e-04, 4.7237e-05, 4.7177e-05, ..., 3.2640e-04, + 2.8729e-04, 3.7861e-04], + [-6.5029e-05, 5.9575e-05, 6.4790e-05, ..., 4.2558e-04, + 3.8528e-04, 5.0926e-04], + ..., + [ 1.6356e-04, 4.9561e-05, 4.9263e-05, ..., 3.4165e-04, + 3.0041e-04, 3.9577e-04], + [ 2.4170e-05, 8.0094e-06, 8.0094e-06, ..., 5.5373e-05, + 4.8786e-05, 6.4254e-05], + [-9.3555e-04, -3.2663e-04, -3.2759e-04, ..., -2.2602e-03, + -1.9932e-03, -2.6264e-03]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0730, 0.1973, 0.1965, 0.2019, 0.0446, -0.7820, -0.4795], + device='cuda:0'), grad: tensor([ 0.0490, 0.0229, 0.0301, 0.0290, 0.0240, 0.0039, -0.1588], + device='cuda:0') +588 +0.0038873953302184317 +changing lr +epoch 40, time 793.18, cls_loss 0.4262 cls_loss_mapping 2.0002 cls_loss_causal 1.1095 re_mapping 0.0238 re_causal 0.0233 /// teacc 93.22 lr 0.00366982 +Epoch 42, weight, value: tensor([[ 0.0840, 0.1488, 0.1224, ..., -0.1270, -0.0851, -0.0858], + [-0.0175, -0.1495, -0.0575, ..., -0.0498, -0.0857, 0.0024], + [ 0.0192, -0.1242, -0.2258, ..., 0.1480, 0.1890, 0.0565], + ..., + [-0.0936, 0.0161, 0.0502, ..., 0.0650, 0.0849, 0.0097], + [ 0.0668, 0.0112, -0.0315, ..., -0.5849, -0.7258, -0.3710], + [-0.1213, -0.0046, -0.0204, ..., 0.0504, 0.2552, 0.2326]], + device='cuda:0'), grad: tensor([[-6.9008e-03, -2.5196e-03, -2.0008e-03, ..., -2.9793e-03, + -2.1839e-03, -3.8757e-03], + [-4.0359e-03, -5.6028e-04, -4.0054e-04, ..., -1.4534e-03, + -1.2255e-03, -1.7681e-03], + [ 1.1845e-03, 4.1723e-04, 3.3045e-04, ..., 5.0640e-04, + 3.7384e-04, 6.5660e-04], + ..., + [ 8.8730e-03, 2.4624e-03, 1.9178e-03, ..., 3.5820e-03, + 2.7637e-03, 4.5586e-03], + [ 6.4559e-06, 2.2762e-06, 1.8030e-06, ..., 2.7604e-06, + 2.0377e-06, 3.5800e-06], + [ 7.7534e-04, 1.6177e-04, 1.2255e-04, ..., 2.9659e-04, + 2.3890e-04, 3.6979e-04]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0461, 0.2063, 0.1867, 0.1634, 0.1129, -0.7826, -0.5055], + device='cuda:0'), grad: tensor([-1.2646e-01, -3.0930e-02, 2.0981e-02, 1.9131e-03, 1.2598e-01, + 1.1444e-04, 8.4991e-03], device='cuda:0') +588 +0.003669815772166629 +changing lr +epoch 41, time 793.89, cls_loss 0.3935 cls_loss_mapping 2.0710 cls_loss_causal 1.1350 re_mapping 0.0310 re_causal 0.0246 /// teacc 92.71 lr 0.00345492 +Epoch 43, weight, value: tensor([[ 0.0643, 0.1088, 0.0626, ..., -0.0929, -0.0535, -0.0638], + [ 0.0028, -0.0901, -0.0426, ..., -0.0234, -0.1150, 0.0112], + [ 0.0149, -0.1513, -0.2476, ..., 0.1876, 0.2205, 0.0927], + ..., + [-0.0908, 0.0468, 0.1297, ..., 0.0240, 0.0925, -0.0197], + [ 0.0966, 0.0282, -0.0233, ..., -0.6239, -0.7280, -0.3785], + [-0.1093, -0.0436, -0.0647, ..., -0.0154, 0.1854, 0.1869]], + device='cuda:0'), grad: tensor([[ 1.0422e-02, 1.5628e-04, 7.0393e-05, ..., 2.0962e-03, + 2.0924e-03, 2.6207e-03], + [-1.1168e-03, -6.2883e-05, -3.4243e-05, ..., -5.1880e-04, + -1.6975e-03, -1.2026e-03], + [-6.5857e-02, -7.9918e-04, -3.2258e-04, ..., -1.2230e-02, + -8.1863e-03, -1.3405e-02], + ..., + [ 5.4901e-02, 6.2561e-04, 2.3735e-04, ..., 1.0056e-02, + 6.1264e-03, 1.0735e-02], + [ 5.6744e-04, 2.6718e-05, 1.6287e-05, ..., 2.0099e-04, + 5.5265e-04, 4.1747e-04], + [ 2.2662e-04, 1.0826e-05, 6.6124e-06, ..., 8.1003e-05, + 2.2435e-04, 1.6904e-04]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.1039, 0.1597, 0.2045, 0.1776, 0.0795, -0.7515, -0.5459], + device='cuda:0'), grad: tensor([ 0.0842, -0.0832, -0.2791, 0.0431, 0.1971, 0.0268, 0.0109], + device='cuda:0') +588 +0.0034549150281252667 +changing lr +epoch 42, time 793.69, cls_loss 0.3802 cls_loss_mapping 1.9892 cls_loss_causal 1.0924 re_mapping 0.0229 re_causal 0.0225 /// teacc 93.22 lr 0.00324313 +Epoch 44, weight, value: tensor([[ 0.0834, 0.1332, 0.0910, ..., -0.1525, -0.0840, -0.1161], + [-0.0021, -0.0873, -0.0451, ..., -0.0073, -0.1078, 0.0156], + [ 0.0146, -0.1524, -0.2777, ..., 0.1898, 0.2598, 0.1004], + ..., + [-0.1010, 0.0378, 0.1309, ..., 0.0671, 0.0940, 0.0226], + [ 0.0986, -0.0018, -0.0487, ..., -0.7223, -0.7836, -0.4384], + [-0.1270, 0.0059, 0.0115, ..., 0.0070, 0.2349, 0.2318]], + device='cuda:0'), grad: tensor([[ 7.7020e-07, 7.6601e-08, 4.8662e-08, ..., 1.6997e-07, + 1.7486e-07, 1.7299e-07], + [ 5.0430e-03, 5.3501e-04, 3.3927e-04, ..., 1.5249e-03, + 1.5764e-03, 1.4753e-03], + [-5.0659e-03, -5.3740e-04, -3.4094e-04, ..., -1.5326e-03, + -1.5841e-03, -1.4820e-03], + ..., + [ 2.6450e-05, 2.8070e-06, 1.7807e-06, ..., 8.0019e-06, + 8.2701e-06, 7.7412e-06], + [ 1.4133e-07, 1.4901e-08, 9.3132e-09, ..., 4.0280e-08, + 4.1677e-08, 3.9348e-08], + [ 1.0245e-08, 1.1642e-09, 6.9849e-10, ..., 3.0268e-09, + 3.2596e-09, 3.0268e-09]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0907, 0.1949, 0.2062, 0.1744, 0.0891, -0.8095, -0.5587], + device='cuda:0'), grad: tensor([ 4.3847e-06, 3.2471e-02, -3.2623e-02, -2.1327e-07, 1.7047e-04, + 8.8802e-07, 6.6124e-08], device='cuda:0') +588 +0.0032431258795932905 +changing lr +epoch 43, time 789.57, cls_loss 0.3522 cls_loss_mapping 1.9937 cls_loss_causal 1.0828 re_mapping 0.0234 re_causal 0.0229 /// teacc 91.96 lr 0.00303487 +Epoch 45, weight, value: tensor([[ 0.0955, 0.1317, 0.1385, ..., -0.1394, -0.0978, -0.1083], + [ 0.0087, -0.0818, -0.0973, ..., 0.0101, -0.0519, 0.0091], + [ 0.0183, -0.1682, -0.2719, ..., 0.1735, 0.2487, 0.1136], + ..., + [-0.1164, 0.0541, 0.1336, ..., 0.0493, 0.0661, 0.0058], + [ 0.0464, -0.0092, -0.0714, ..., -0.7157, -0.7955, -0.4830], + [-0.1022, -0.0298, -0.0185, ..., -0.0183, 0.2220, 0.2155]], + device='cuda:0'), grad: tensor([[-3.1910e-03, -8.5545e-04, -8.6164e-04, ..., -8.4400e-04, + -1.3742e-03, -1.5039e-03], + [ 1.7672e-03, 4.7231e-04, 4.7612e-04, ..., 4.6682e-04, + 7.5912e-04, 8.3160e-04], + [ 2.9039e-04, 7.3493e-05, 7.4208e-05, ..., 7.4744e-05, + 1.1790e-04, 1.3268e-04], + ..., + [ 7.6914e-04, 2.1195e-04, 2.1338e-04, ..., 2.0635e-04, + 3.4070e-04, 3.6836e-04], + [ 9.1136e-05, 2.4363e-05, 2.4557e-05, ..., 2.4080e-05, + 3.9160e-05, 4.2886e-05], + [ 1.2159e-04, 3.2514e-05, 3.2783e-05, ..., 3.2157e-05, + 5.2243e-05, 5.7250e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.1140, 0.1826, 0.1994, 0.1708, 0.1024, -0.8266, -0.5850], + device='cuda:0'), grad: tensor([-0.1249, 0.0690, 0.0109, 0.0059, 0.0308, 0.0036, 0.0048], + device='cuda:0') +588 +0.0030348748417303863 +changing lr +---------------------saving model at epoch 44---------------------------------------------------- +epoch 44, time 795.83, cls_loss 0.3978 cls_loss_mapping 1.9878 cls_loss_causal 1.1038 re_mapping 0.0225 re_causal 0.0223 /// teacc 93.97 lr 0.00283058 +Epoch 46, weight, value: tensor([[ 0.0839, 0.0965, 0.1240, ..., -0.1082, -0.0749, -0.0755], + [ 0.0055, -0.0306, -0.0517, ..., -0.0313, -0.1022, -0.0195], + [ 0.0100, -0.1071, -0.1644, ..., 0.1763, 0.2366, 0.0998], + ..., + [-0.0981, 0.0232, 0.0561, ..., 0.0557, 0.0843, 0.0088], + [ 0.0672, 0.0190, -0.0069, ..., -0.7138, -0.7913, -0.4696], + [-0.1312, -0.1181, -0.1156, ..., -0.0245, 0.2260, 0.1983]], + device='cuda:0'), grad: tensor([[ 1.1587e-03, 5.3078e-05, 1.8433e-05, ..., 4.1866e-04, + 3.5286e-04, 7.6532e-04], + [-1.5497e-04, -3.2276e-05, -1.1854e-05, ..., 1.2422e-04, + 1.0598e-04, 3.6764e-04], + [-1.7033e-03, -1.0192e-04, -1.9267e-05, ..., 2.3887e-05, + 6.8843e-05, 2.2292e-04], + ..., + [ 2.7733e-03, 1.2243e-04, 2.6718e-05, ..., 5.9557e-04, + 4.5776e-04, 1.0719e-03], + [ 4.2558e-05, 8.6892e-07, 2.9127e-07, ..., 2.3574e-05, + 1.9982e-05, 4.9174e-05], + [-2.4204e-03, -4.9174e-05, -1.6510e-05, ..., -1.3437e-03, + -1.1387e-03, -2.8038e-03]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.1065, 0.2135, 0.1948, 0.1774, 0.1072, -0.8170, -0.6395], + device='cuda:0'), grad: tensor([ 0.0378, 0.0275, 0.0229, 0.0188, 0.0518, 0.0028, -0.1616], + device='cuda:0') +588 +0.0028305813044122124 +changing lr +---------------------saving model at epoch 45---------------------------------------------------- +epoch 45, time 793.49, cls_loss 0.3623 cls_loss_mapping 1.9942 cls_loss_causal 1.0389 re_mapping 0.0226 re_causal 0.0221 /// teacc 95.73 lr 0.00263066 +Epoch 47, weight, value: tensor([[ 0.0841, 0.1589, 0.1415, ..., -0.0505, -0.0102, -0.0414], + [-0.0044, -0.0669, 0.0280, ..., -0.0547, -0.1378, -0.0437], + [ 0.0146, -0.1602, -0.2107, ..., 0.1659, 0.2378, 0.0657], + ..., + [-0.0972, 0.0043, 0.0218, ..., 0.0352, 0.0535, 0.0162], + [ 0.0674, 0.0272, 0.0462, ..., -0.6853, -0.7837, -0.4699], + [-0.1275, -0.1053, -0.0656, ..., -0.0593, 0.2231, 0.1930]], + device='cuda:0'), grad: tensor([[ 1.0528e-05, 9.8825e-05, 4.8339e-05, ..., -1.7762e-04, + 2.6751e-04, -1.5097e-03], + [-1.2131e-03, 5.4312e-04, 1.5461e-04, ..., -1.0080e-03, + -1.0281e-03, -9.6178e-04], + [ 6.4945e-04, 5.8937e-04, 2.1136e-04, ..., 2.4772e-04, + 1.8764e-04, 7.2575e-04], + ..., + [-1.2379e-03, -2.8477e-03, -9.9373e-04, ..., 2.5320e-04, + 4.7743e-05, -2.3520e-04], + [ 1.8501e-04, 1.6427e-04, 5.8860e-05, ..., 7.3135e-05, + 5.2392e-05, 2.1923e-04], + [ 3.4595e-04, 3.3283e-04, 1.1873e-04, ..., 1.2720e-04, + 8.9347e-05, 4.0579e-04]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.1119, 0.1700, 0.1375, 0.2191, 0.1577, -0.8327, -0.6262], + device='cuda:0'), grad: tensor([-0.0715, -0.0638, 0.0537, 0.1011, -0.0654, 0.0159, 0.0297], + device='cuda:0') +588 +0.0026306566876350096 +changing lr +epoch 46, time 786.80, cls_loss 0.3553 cls_loss_mapping 1.9916 cls_loss_causal 1.0388 re_mapping 0.0231 re_causal 0.0224 /// teacc 92.46 lr 0.00243550 +Epoch 48, weight, value: tensor([[ 0.0956, 0.1558, 0.0879, ..., -0.0899, -0.0491, -0.0674], + [-0.0115, -0.0421, 0.0354, ..., -0.0749, -0.1362, -0.0593], + [ 0.0077, -0.1381, -0.2124, ..., 0.2029, 0.2413, 0.1030], + ..., + [-0.0945, -0.0059, 0.0601, ..., 0.0411, 0.0617, 0.0120], + [ 0.0528, 0.0421, 0.0290, ..., -0.6839, -0.7994, -0.4708], + [-0.1295, -0.0696, -0.0456, ..., -0.0112, 0.2695, 0.2299]], + device='cuda:0'), grad: tensor([[ 9.5034e-04, 7.5281e-05, 7.4506e-05, ..., 2.2638e-04, + 1.2189e-04, 2.8396e-04], + [ 3.6478e-04, 3.2634e-05, 3.2872e-05, ..., 8.5950e-05, + 4.1932e-05, 1.1307e-04], + [ 5.7268e-04, 5.1200e-05, 5.1588e-05, ..., 1.3494e-04, + 6.5863e-05, 1.7750e-04], + ..., + [ 4.9162e-04, 5.3734e-05, 5.5492e-05, ..., 1.1331e-04, + 4.3839e-05, 1.6296e-04], + [-2.9774e-03, -2.6631e-04, -2.6822e-04, ..., -7.0143e-04, + -3.4213e-04, -9.2268e-04], + [ 1.2887e-04, 1.1526e-05, 1.1608e-05, ..., 3.0354e-05, + 1.4812e-05, 3.9935e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.1229, 0.1631, 0.1843, 0.2061, 0.1127, -0.8222, -0.6167], + device='cuda:0'), grad: tensor([ 0.0432, 0.0200, 0.0313, 0.0257, 0.0357, -0.1630, 0.0071], + device='cuda:0') +588 +0.0024355036129704724 +changing lr +epoch 47, time 795.90, cls_loss 0.3662 cls_loss_mapping 1.9847 cls_loss_causal 1.0493 re_mapping 0.0217 re_causal 0.0213 /// teacc 92.46 lr 0.00224552 +Epoch 49, weight, value: tensor([[ 0.0922, 0.1120, 0.0930, ..., -0.1250, -0.0653, -0.0949], + [-0.0051, -0.0318, -0.0315, ..., -0.0817, -0.1453, -0.0640], + [ 0.0037, -0.1050, -0.2166, ..., 0.1755, 0.2101, 0.0731], + ..., + [-0.1005, -0.0186, 0.1052, ..., 0.0833, 0.1078, 0.0529], + [ 0.0616, 0.0550, 0.0163, ..., -0.6891, -0.8328, -0.4716], + [-0.1069, -0.0613, -0.0591, ..., -0.0168, 0.2633, 0.2115]], + device='cuda:0'), grad: tensor([[ 9.0361e-04, 2.4486e-04, 1.2577e-04, ..., 5.3930e-04, + 4.0340e-04, 1.2808e-03], + [-3.0640e-02, -4.7135e-04, -3.2711e-04, ..., -5.8136e-03, + -5.9357e-03, -8.0338e-03], + [ 2.3234e-04, 6.7770e-05, 3.4958e-05, ..., 1.4591e-04, + 1.0866e-04, 3.5119e-04], + ..., + [-1.6909e-03, -4.7898e-04, -2.4652e-04, ..., -1.0414e-03, + -7.7629e-04, -2.4929e-03], + [ 3.3110e-05, 8.1658e-06, 4.2245e-06, ..., 1.8403e-05, + 1.3992e-05, 4.3273e-05], + [ 4.1902e-05, 1.1764e-05, 6.0722e-06, ..., 2.5570e-05, + 1.9133e-05, 6.1274e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0660, 0.2048, 0.1961, 0.2368, 0.1202, -0.8832, -0.6157], + device='cuda:0'), grad: tensor([ 0.0501, -0.1084, 0.0139, 0.1385, -0.0981, 0.0017, 0.0024], + device='cuda:0') +588 +0.00224551509273949 +changing lr +epoch 48, time 791.15, cls_loss 0.3587 cls_loss_mapping 1.9853 cls_loss_causal 1.0358 re_mapping 0.0222 re_causal 0.0221 /// teacc 94.97 lr 0.00206107 +Epoch 50, weight, value: tensor([[ 0.0846, 0.1816, 0.1254, ..., -0.0750, 0.0012, -0.0450], + [-0.0013, -0.0766, -0.0773, ..., -0.1077, -0.1881, -0.0801], + [ 0.0212, -0.1647, -0.2473, ..., 0.1684, 0.2135, 0.0857], + ..., + [-0.0958, -0.0384, 0.0915, ..., 0.0667, 0.0922, 0.0290], + [ 0.0626, 0.0544, 0.0271, ..., -0.6782, -0.8441, -0.4648], + [-0.1303, 0.0366, 0.0180, ..., -0.0128, 0.2398, 0.2009]], + device='cuda:0'), grad: tensor([[-1.5373e-03, -4.3774e-04, -4.2224e-04, ..., -1.9836e-04, + -6.0892e-04, -1.6680e-03], + [ 3.6526e-04, 1.0401e-04, 1.0031e-04, ..., 4.7117e-05, + 1.4460e-04, 3.9625e-04], + [ 3.9077e-04, 1.1337e-04, 1.0937e-04, ..., 5.0902e-05, + 1.5759e-04, 4.3225e-04], + ..., + [ 4.2677e-04, 1.1957e-04, 1.1528e-04, ..., 5.4568e-05, + 1.6630e-04, 4.5514e-04], + [ 5.6386e-05, 1.6049e-05, 1.5482e-05, ..., 7.2718e-06, + 2.2322e-05, 6.1154e-05], + [ 5.2214e-05, 1.4871e-05, 1.4335e-05, ..., 6.7353e-06, + 2.0683e-05, 5.6654e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.1598, 0.1734, 0.1801, 0.1802, 0.1135, -0.8684, -0.6332], + device='cuda:0'), grad: tensor([-0.1240, 0.0295, 0.0322, 0.0198, 0.0338, 0.0045, 0.0042], + device='cuda:0') +588 +0.002061073738537637 +changing lr +epoch 49, time 791.66, cls_loss 0.3642 cls_loss_mapping 1.9799 cls_loss_causal 1.0172 re_mapping 0.0228 re_causal 0.0216 /// teacc 91.96 lr 0.00188255 +Epoch 51, weight, value: tensor([[ 8.0050e-02, 1.1906e-01, 1.0821e-01, ..., -1.1237e-01, + -1.8257e-02, -8.6082e-02], + [-2.8441e-04, -4.6112e-02, -6.0809e-02, ..., -9.2129e-02, + -1.4189e-01, -6.3123e-02], + [ 3.1039e-02, -1.1264e-01, -1.9799e-01, ..., 1.6092e-01, + 1.9514e-01, 5.9242e-02], + ..., + [-1.0457e-01, -6.0067e-02, 2.4459e-02, ..., 7.8760e-02, + 8.4602e-02, 5.1598e-02], + [ 7.7575e-02, 3.1680e-02, 4.2495e-03, ..., -7.1626e-01, + -8.7211e-01, -4.7275e-01], + [-1.1534e-01, 5.9901e-02, 3.1899e-02, ..., 9.2074e-03, + 2.5169e-01, 2.3730e-01]], device='cuda:0'), grad: tensor([[-5.6298e-07, -1.2806e-09, 2.3283e-10, ..., -7.5612e-08, + -5.9546e-08, -9.7963e-08], + [-2.6776e-09, -5.8208e-11, -1.1642e-10, ..., 3.4925e-10, + -5.8208e-10, 0.0000e+00], + [-1.5378e-05, -1.0012e-06, -9.3947e-08, ..., -2.9653e-06, + -1.3877e-06, -2.8573e-06], + ..., + [ 1.5989e-05, 1.0058e-06, 9.5868e-08, ..., 3.0417e-06, + 1.4501e-06, 2.9579e-06], + [ 2.3283e-09, 1.7462e-10, 1.1642e-10, ..., 1.1642e-10, + 1.1642e-10, 1.7462e-10], + [-3.7893e-08, -2.7358e-09, -2.3283e-09, ..., -1.4552e-09, + -1.6298e-09, -2.6776e-09]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.1052, 0.2196, 0.1519, 0.1560, 0.1559, -0.8919, -0.5913], + device='cuda:0'), grad: tensor([-2.1085e-06, -5.0059e-09, -9.9659e-05, 2.3865e-09, 1.0186e-04, + 6.6357e-09, -1.0635e-07], device='cuda:0') +588 +0.0018825509907063344 +changing lr +epoch 50, time 792.19, cls_loss 0.3518 cls_loss_mapping 1.9837 cls_loss_causal 0.9953 re_mapping 0.0223 re_causal 0.0216 /// teacc 93.97 lr 0.00171031 +Epoch 52, weight, value: tensor([[ 0.0866, 0.0985, 0.1082, ..., -0.1027, -0.0159, -0.0661], + [ 0.0080, -0.0585, -0.0623, ..., -0.0809, -0.1541, -0.0534], + [ 0.0161, -0.1148, -0.1782, ..., 0.1115, 0.1755, 0.0307], + ..., + [-0.1067, -0.0302, 0.0260, ..., 0.0835, 0.0624, 0.0352], + [ 0.0713, 0.0242, -0.0059, ..., -0.6828, -0.8655, -0.4524], + [-0.1252, 0.0592, 0.0131, ..., 0.0226, 0.2841, 0.2409]], + device='cuda:0'), grad: tensor([[8.2970e-04, 1.3388e-09, 3.4925e-09, ..., 4.7952e-05, 4.6587e-04, + 3.2496e-04], + [1.5850e-03, 3.8766e-08, 1.0332e-07, ..., 9.2506e-05, 8.9121e-04, + 6.2180e-04], + [8.0967e-04, 1.9209e-09, 5.1805e-09, ..., 4.6819e-05, 4.5466e-04, + 3.1710e-04], + ..., + [8.0633e-04, 4.0745e-10, 8.7311e-10, ..., 4.6581e-05, 4.5276e-04, + 3.1567e-04], + [1.7512e-04, 5.8208e-11, 2.3283e-10, ..., 1.0118e-05, 9.8348e-05, + 6.8605e-05], + [1.8668e-04, 1.7462e-10, 4.0745e-10, ..., 1.0788e-05, 1.0484e-04, + 7.3135e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0871, 0.2250, 0.1671, 0.1756, 0.1291, -0.8618, -0.5868], + device='cuda:0'), grad: tensor([ 0.0266, 0.0507, 0.0259, -0.1405, 0.0258, 0.0056, 0.0060], + device='cuda:0') +588 +0.0017103063703014388 +changing lr +epoch 51, time 788.44, cls_loss 0.3190 cls_loss_mapping 1.9850 cls_loss_causal 0.9625 re_mapping 0.0214 re_causal 0.0201 /// teacc 95.73 lr 0.00154469 +Epoch 53, weight, value: tensor([[ 0.0876, 0.0600, 0.1045, ..., -0.0811, 0.0089, -0.0534], + [ 0.0014, -0.0317, -0.0467, ..., -0.0808, -0.1815, -0.0571], + [ 0.0183, -0.0910, -0.1617, ..., 0.1261, 0.1863, 0.0311], + ..., + [-0.1071, -0.0345, 0.0117, ..., 0.0714, 0.0735, 0.0359], + [ 0.0532, 0.0471, 0.0300, ..., -0.6734, -0.8537, -0.4532], + [-0.1118, 0.0073, -0.0313, ..., -0.0092, 0.2564, 0.2223]], + device='cuda:0'), grad: tensor([[-2.4796e-03, -1.2941e-03, -1.0529e-03, ..., -4.8027e-03, + -4.9057e-03, -5.3940e-03], + [ 1.5469e-03, 1.0319e-03, 8.9645e-04, ..., 4.0779e-03, + 2.5291e-03, 3.4771e-03], + [ 2.1863e-04, 4.2468e-05, 1.6421e-05, ..., 7.7665e-05, + 6.0129e-04, 4.3964e-04], + ..., + [ 4.7302e-04, 1.7166e-04, 1.2064e-04, ..., 5.5265e-04, + 1.1129e-03, 9.9087e-04], + [ 3.4988e-05, 1.2383e-05, 8.5831e-06, ..., 3.9369e-05, + 8.3029e-05, 7.3135e-05], + [ 5.2601e-05, 9.7007e-06, 3.4031e-06, ..., 1.6198e-05, + 1.4591e-04, 1.0550e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.1020, 0.2097, 0.1749, 0.1619, 0.1433, -0.8940, -0.5930], + device='cuda:0'), grad: tensor([-0.0985, -0.0142, 0.0328, 0.0244, 0.0441, 0.0034, 0.0081], + device='cuda:0') +588 +0.0015446867550656784 +changing lr +epoch 52, time 789.46, cls_loss 0.3353 cls_loss_mapping 1.9789 cls_loss_causal 0.9720 re_mapping 0.0202 re_causal 0.0195 /// teacc 95.48 lr 0.00138603 +Epoch 54, weight, value: tensor([[ 0.0870, 0.1018, 0.1322, ..., -0.0850, -0.0085, -0.0579], + [-0.0097, -0.0700, -0.0837, ..., -0.0925, -0.1853, -0.0620], + [ 0.0211, -0.0879, -0.1712, ..., 0.1150, 0.1763, 0.0230], + ..., + [-0.1036, -0.0421, 0.0291, ..., 0.0902, 0.0958, 0.0477], + [ 0.0499, 0.0694, 0.0337, ..., -0.6543, -0.8369, -0.4341], + [-0.1149, -0.0044, -0.0291, ..., 0.0015, 0.2656, 0.2268]], + device='cuda:0'), grad: tensor([[-5.0974e-04, -1.8522e-05, -2.5466e-05, ..., -8.1420e-05, + -8.4937e-05, -1.4079e-04], + [-3.1147e-03, -7.0408e-06, 1.8835e-05, ..., -1.3113e-04, + -7.6473e-05, -1.3566e-04], + [ 7.2131e-07, -2.0256e-07, -1.3458e-07, ..., -4.1537e-07, + -3.8161e-07, -2.7660e-07], + ..., + [ 3.6221e-03, 2.5690e-05, 6.6943e-06, ..., 2.1267e-04, + 1.6165e-04, 2.7657e-04], + [ 1.3539e-07, 4.6566e-09, 6.4028e-09, ..., 2.0489e-08, + 2.1420e-08, 3.5507e-08], + [ 5.9512e-07, 2.3167e-08, 2.9453e-08, ..., 9.3598e-08, + 9.7323e-08, 1.5879e-07]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0913, 0.2123, 0.1681, 0.1768, 0.1330, -0.8870, -0.5721], + device='cuda:0'), grad: tensor([-2.8477e-03, -1.0956e-02, -6.1877e-06, 6.0722e-06, 1.3802e-02, + 7.4832e-07, 3.3192e-06], device='cuda:0') +588 +0.001386025680863044 +changing lr +epoch 53, time 789.24, cls_loss 0.3444 cls_loss_mapping 1.9801 cls_loss_causal 0.9729 re_mapping 0.0201 re_causal 0.0194 /// teacc 95.23 lr 0.00123464 +Epoch 55, weight, value: tensor([[ 0.0899, 0.0994, 0.1185, ..., -0.0883, -0.0154, -0.0628], + [-0.0138, -0.0552, -0.0646, ..., -0.0786, -0.1759, -0.0448], + [ 0.0190, -0.0833, -0.1638, ..., 0.1143, 0.1703, 0.0314], + ..., + [-0.0976, -0.0140, 0.0644, ..., 0.0951, 0.1176, 0.0502], + [ 0.0533, 0.0488, 0.0212, ..., -0.6563, -0.8324, -0.4305], + [-0.1198, -0.0496, -0.0600, ..., -0.0246, 0.2251, 0.2029]], + device='cuda:0'), grad: tensor([[ 1.9855e-03, 8.1599e-05, 1.0729e-05, ..., 5.2899e-05, + 8.5890e-05, 1.5986e-04], + [-6.4308e-07, -3.5856e-08, -3.5041e-08, ..., -1.5623e-07, + -1.6892e-07, -1.9337e-07], + [-3.0777e-02, -1.2655e-03, -1.6606e-04, ..., -8.1873e-04, + -1.3304e-03, -2.4776e-03], + ..., + [ 2.8809e-02, 1.1845e-03, 1.5545e-04, ..., 7.6628e-04, + 1.2455e-03, 2.3193e-03], + [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [-2.6193e-08, -3.4925e-09, -2.9104e-09, ..., -3.3760e-09, + -4.4238e-09, -5.0059e-09]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.1008, 0.1947, 0.1942, 0.1807, 0.1339, -0.8819, -0.6171], + device='cuda:0'), grad: tensor([ 8.9722e-03, -2.7604e-06, -1.3916e-01, 1.6764e-08, 1.3025e-01, + 2.0955e-09, -6.2049e-08], device='cuda:0') +588 +0.0012346426699819469 +changing lr +epoch 54, time 787.86, cls_loss 0.3206 cls_loss_mapping 1.9793 cls_loss_causal 0.9498 re_mapping 0.0204 re_causal 0.0194 /// teacc 95.48 lr 0.00109084 +Epoch 56, weight, value: tensor([[ 0.0925, 0.0880, 0.1235, ..., -0.1024, -0.0247, -0.0701], + [-0.0047, -0.0246, -0.0448, ..., -0.0640, -0.1698, -0.0438], + [ 0.0114, -0.0813, -0.1900, ..., 0.1038, 0.1607, 0.0077], + ..., + [-0.1076, -0.0237, 0.0582, ..., 0.1092, 0.1341, 0.0602], + [ 0.0442, 0.0242, 0.0075, ..., -0.6532, -0.8245, -0.4242], + [-0.1167, -0.0489, -0.0508, ..., -0.0102, 0.2154, 0.2315]], + device='cuda:0'), grad: tensor([[7.3135e-05, 1.2144e-06, 4.8131e-06, ..., 2.8417e-05, 3.7432e-05, + 5.5701e-05], + [1.2434e-04, 2.0862e-06, 8.1658e-06, ..., 4.8101e-05, 6.3360e-05, + 9.4295e-05], + [1.6632e-03, 2.7791e-05, 1.0926e-04, ..., 6.4373e-04, 8.4877e-04, + 1.2627e-03], + ..., + [6.2585e-05, 1.0468e-06, 4.1127e-06, ..., 2.4229e-05, 3.1918e-05, + 4.7505e-05], + [7.0482e-06, 1.1781e-07, 4.6310e-07, ..., 2.7288e-06, 3.5968e-06, + 5.3495e-06], + [1.9205e-04, 3.2093e-06, 1.2614e-05, ..., 7.4327e-05, 9.7990e-05, + 1.4579e-04]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0872, 0.1928, 0.1729, 0.1769, 0.1607, -0.8810, -0.5969], + device='cuda:0'), grad: tensor([ 0.0018, 0.0031, 0.0416, -0.0531, 0.0016, 0.0002, 0.0048], + device='cuda:0') +588 +0.0010908425876598518 +changing lr +epoch 55, time 786.43, cls_loss 0.3282 cls_loss_mapping 1.9821 cls_loss_causal 0.9340 re_mapping 0.0193 re_causal 0.0177 /// teacc 95.73 lr 0.00095492 +Epoch 57, weight, value: tensor([[ 0.0849, 0.0755, 0.0935, ..., -0.1036, -0.0197, -0.0655], + [-0.0095, -0.0116, -0.0134, ..., -0.0643, -0.1662, -0.0527], + [ 0.0208, -0.0986, -0.2220, ..., 0.0728, 0.1324, -0.0059], + ..., + [-0.0978, 0.0050, 0.1041, ..., 0.1330, 0.1540, 0.0799], + [ 0.0418, 0.0328, 0.0083, ..., -0.6448, -0.8223, -0.4147], + [-0.1264, -0.0757, -0.0820, ..., 0.0061, 0.2072, 0.2298]], + device='cuda:0'), grad: tensor([[ 1.7536e-04, 9.3132e-10, 1.1642e-09, ..., 4.3988e-05, + 4.4763e-05, 1.3244e-04], + [ 1.6344e-04, -6.9849e-10, -9.3132e-10, ..., 4.1008e-05, + 4.1693e-05, 1.2350e-04], + [-7.1812e-04, 1.1642e-10, 2.3283e-10, ..., -1.8024e-04, + -1.8334e-04, -5.4264e-04], + ..., + [ 1.6940e-04, 1.1642e-10, 1.1642e-10, ..., 4.2498e-05, + 4.3243e-05, 1.2803e-04], + [ 2.8685e-05, 0.0000e+00, 0.0000e+00, ..., 7.1973e-06, + 7.3202e-06, 2.1666e-05], + [ 5.6595e-05, 0.0000e+00, 0.0000e+00, ..., 1.4201e-05, + 1.4447e-05, 4.2766e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0884, 0.1921, 0.1628, 0.1668, 0.1655, -0.8695, -0.5853], + device='cuda:0'), grad: tensor([ 0.0331, 0.0309, -0.1357, 0.0236, 0.0320, 0.0054, 0.0107], + device='cuda:0') +588 +0.000954915028125264 +changing lr +epoch 56, time 794.91, cls_loss 0.3137 cls_loss_mapping 1.9781 cls_loss_causal 0.9403 re_mapping 0.0187 re_causal 0.0169 /// teacc 95.48 lr 0.00082713 +Epoch 58, weight, value: tensor([[ 0.0945, 0.0867, 0.1059, ..., -0.1034, -0.0176, -0.0633], + [-0.0110, -0.0268, -0.0210, ..., -0.0691, -0.1674, -0.0602], + [ 0.0115, -0.1071, -0.2236, ..., 0.0860, 0.1386, 0.0063], + ..., + [-0.1007, 0.0193, 0.1096, ..., 0.1217, 0.1406, 0.0734], + [ 0.0280, 0.0210, -0.0052, ..., -0.6541, -0.8266, -0.4252], + [-0.1237, -0.0900, -0.1109, ..., -0.0034, 0.2090, 0.2245]], + device='cuda:0'), grad: tensor([[ 5.8508e-04, 6.0380e-05, 8.1122e-05, ..., 9.2030e-05, + 1.1188e-04, 9.8765e-05], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, + 1.1642e-10, 1.1642e-10], + [ 1.0477e-09, 0.0000e+00, 1.1642e-10, ..., 1.1642e-10, + 1.1642e-10, 1.1642e-10], + ..., + [-5.8508e-04, -6.0380e-05, -8.1122e-05, ..., -9.2030e-05, + -1.1188e-04, -9.8765e-05], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 1.1642e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0861, 0.1934, 0.1672, 0.1621, 0.1814, -0.8975, -0.5961], + device='cuda:0'), grad: tensor([ 1.3752e-03, 1.5134e-09, 2.2119e-09, 0.0000e+00, -1.3752e-03, + 0.0000e+00, 2.3283e-10], device='cuda:0') +588 +0.0008271337313934874 +changing lr +epoch 57, time 793.63, cls_loss 0.3182 cls_loss_mapping 1.9770 cls_loss_causal 0.9363 re_mapping 0.0179 re_causal 0.0162 /// teacc 94.22 lr 0.00070776 +Epoch 59, weight, value: tensor([[ 0.0888, 0.0869, 0.1046, ..., -0.0862, -0.0070, -0.0598], + [ 0.0009, -0.0291, -0.0251, ..., -0.0691, -0.1569, -0.0566], + [ 0.0186, -0.1012, -0.2123, ..., 0.1033, 0.1411, 0.0275], + ..., + [-0.1039, 0.0033, 0.0946, ..., 0.0985, 0.1191, 0.0543], + [ 0.0455, 0.0372, 0.0154, ..., -0.6556, -0.8375, -0.4216], + [-0.1209, -0.0725, -0.0921, ..., 0.0013, 0.2220, 0.2357]], + device='cuda:0'), grad: tensor([[ 1.0811e-02, 2.8998e-05, 1.1642e-10, ..., 3.8071e-03, + 3.9043e-03, 6.3400e-03], + [-6.7282e-04, -1.2779e-04, 1.1642e-10, ..., 2.0191e-06, + -1.4477e-03, -1.7529e-03], + [ 2.9755e-04, 3.4481e-05, 6.9849e-10, ..., 4.0889e-05, + 4.2963e-04, 5.3787e-04], + ..., + [-1.0628e-02, 2.7165e-05, 1.1642e-10, ..., -3.8490e-03, + -3.3054e-03, -5.6343e-03], + [ 2.8417e-05, 5.3495e-06, 0.0000e+00, ..., 8.6147e-09, + 6.0678e-05, 7.3493e-05], + [ 5.5879e-05, 1.0297e-05, 0.0000e+00, ..., 4.3353e-07, + 1.1718e-04, 1.4210e-04]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0663, 0.2097, 0.1770, 0.1631, 0.1550, -0.8730, -0.5753], + device='cuda:0'), grad: tensor([ 0.1864, -0.1350, 0.0381, 0.0228, -0.1288, 0.0057, 0.0109], + device='cuda:0') +588 +0.00070775603199067 +changing lr +epoch 58, time 791.25, cls_loss 0.2901 cls_loss_mapping 1.9804 cls_loss_causal 0.9192 re_mapping 0.0170 re_causal 0.0153 /// teacc 94.97 lr 0.00059702 +Epoch 60, weight, value: tensor([[ 9.1027e-02, 8.0252e-02, 1.0966e-01, ..., -7.8625e-02, + 3.5921e-03, -5.6498e-02], + [-1.4066e-03, -1.2845e-02, -1.5504e-02, ..., -6.0565e-02, + -1.5019e-01, -4.6106e-02], + [ 1.6564e-02, -1.1053e-01, -2.2458e-01, ..., 9.8393e-02, + 1.4435e-01, 2.9368e-02], + ..., + [-1.0433e-01, -7.6022e-04, 8.2667e-02, ..., 9.0935e-02, + 1.0773e-01, 4.9190e-02], + [ 3.7246e-02, 2.3048e-02, -2.6357e-04, ..., -6.7142e-01, + -8.5446e-01, -4.3935e-01], + [-1.1422e-01, -6.6161e-02, -8.7990e-02, ..., -1.8907e-04, + 2.1051e-01, 2.2841e-01]], device='cuda:0'), grad: tensor([[-1.5676e-04, 5.4955e-05, -1.0324e-04, ..., 5.1320e-05, + 2.7180e-04, 3.9124e-04], + [-9.4891e-04, -6.8188e-04, -3.0577e-05, ..., 3.3025e-06, + -9.8801e-04, -1.9236e-03], + [ 4.5896e-04, 2.1553e-04, 4.5747e-05, ..., 2.2784e-05, + 2.6202e-04, 5.4216e-04], + ..., + [ 3.3188e-04, 2.0230e-04, 7.3135e-05, ..., -7.9691e-05, + 1.5903e-04, 4.1056e-04], + [ 2.3142e-05, 1.6019e-05, 9.1642e-07, ..., 4.1618e-08, + 2.2918e-05, 4.4793e-05], + [ 9.5248e-05, 5.5879e-05, 6.5751e-06, ..., 2.1756e-06, + 7.5221e-05, 1.5008e-04]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0636, 0.2131, 0.1707, 0.1703, 0.1683, -0.8938, -0.5851], + device='cuda:0'), grad: tensor([ 0.0217, -0.1340, 0.0378, 0.0268, 0.0341, 0.0031, 0.0105], + device='cuda:0') +588 +0.0005970223407163104 +changing lr +---------------------saving model at epoch 59---------------------------------------------------- +epoch 59, time 792.35, cls_loss 0.3278 cls_loss_mapping 1.9809 cls_loss_causal 0.9372 re_mapping 0.0166 re_causal 0.0147 /// teacc 96.73 lr 0.00049516 +Epoch 61, weight, value: tensor([[ 8.5440e-02, 9.8089e-02, 1.2226e-01, ..., -7.7280e-02, + 1.0226e-02, -4.9343e-02], + [ 5.9321e-04, -1.5743e-02, -2.9336e-02, ..., -7.3289e-02, + -1.6599e-01, -5.3425e-02], + [ 1.6552e-02, -1.1387e-01, -2.2323e-01, ..., 9.5401e-02, + 1.4125e-01, 2.3359e-02], + ..., + [-1.0525e-01, -1.8519e-03, 8.4089e-02, ..., 9.0946e-02, + 1.1003e-01, 4.7462e-02], + [ 4.9495e-02, 3.3376e-02, 1.0728e-02, ..., -6.6826e-01, + -8.4971e-01, -4.3507e-01], + [-1.1555e-01, -8.8010e-02, -1.0304e-01, ..., 3.3772e-03, + 2.1152e-01, 2.2746e-01]], device='cuda:0'), grad: tensor([[ 1.8173e-02, 1.9369e-03, 1.6441e-03, ..., 3.3092e-03, + 2.0313e-03, 2.5291e-03], + [ 3.9315e-04, 7.5698e-05, -1.3113e-05, ..., -9.0361e-04, + -5.3501e-04, -8.6689e-04], + [ 6.2561e-04, 5.8770e-05, 6.7949e-05, ..., 3.4142e-04, + 2.0504e-04, 3.0208e-04], + ..., + [-1.9699e-02, -2.1191e-03, -1.7538e-03, ..., -3.0384e-03, + -1.8749e-03, -2.2240e-03], + [ 3.4273e-05, 2.9095e-06, 4.1723e-06, ..., 2.7701e-05, + 1.6570e-05, 2.5049e-05], + [ 2.8467e-04, 2.8133e-05, 2.8908e-05, ..., 1.1533e-04, + 6.9499e-05, 9.9659e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0783, 0.2048, 0.1796, 0.1541, 0.1735, -0.8931, -0.6030], + device='cuda:0'), grad: tensor([ 0.1573, -0.1240, 0.0352, 0.0166, -0.0989, 0.0031, 0.0108], + device='cuda:0') +588 +0.0004951556604879052 +changing lr +epoch 60, time 791.05, cls_loss 0.3158 cls_loss_mapping 1.9791 cls_loss_causal 0.9121 re_mapping 0.0157 re_causal 0.0140 /// teacc 94.72 lr 0.00040236 +Epoch 62, weight, value: tensor([[ 0.0857, 0.0924, 0.1144, ..., -0.0729, 0.0183, -0.0492], + [-0.0042, -0.0068, -0.0229, ..., -0.0765, -0.1683, -0.0542], + [ 0.0245, -0.0980, -0.2087, ..., 0.1062, 0.1521, 0.0308], + ..., + [-0.1061, -0.0096, 0.0811, ..., 0.0871, 0.1008, 0.0451], + [ 0.0382, 0.0153, -0.0056, ..., -0.6643, -0.8481, -0.4378], + [-0.1126, -0.1011, -0.1112, ..., -0.0118, 0.1943, 0.2124]], + device='cuda:0'), grad: tensor([[-1.1024e-02, -4.4060e-04, -3.2377e-04, ..., -1.8454e-03, + -7.7057e-04, -6.5470e-04], + [ 6.2370e-03, 5.1594e-04, 3.2806e-04, ..., 9.4843e-04, + 1.0052e-03, 1.1072e-03], + [ 2.1667e-03, 3.2043e-04, 1.9121e-04, ..., 2.7537e-04, + 6.4468e-04, 7.6580e-04], + ..., + [ 2.1629e-03, -5.9748e-04, -3.0828e-04, ..., 6.0987e-04, + -1.3008e-03, -1.7509e-03], + [ 3.3498e-05, 1.6540e-05, 9.2387e-06, ..., 1.6764e-07, + 3.4720e-05, 4.3929e-05], + [ 1.4818e-04, 4.6730e-05, 2.6509e-05, ..., 1.0215e-05, + 9.7275e-05, 1.2141e-04]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0728, 0.2098, 0.1846, 0.1634, 0.1688, -0.8905, -0.6168], + device='cuda:0'), grad: tensor([-0.0482, 0.0734, 0.0497, 0.0231, -0.1085, 0.0028, 0.0077], + device='cuda:0') +588 +0.00040236113724274745 +changing lr +epoch 61, time 787.53, cls_loss 0.3060 cls_loss_mapping 1.9756 cls_loss_causal 0.9226 re_mapping 0.0150 re_causal 0.0132 /// teacc 95.98 lr 0.00031883 +Epoch 63, weight, value: tensor([[ 0.0928, 0.0778, 0.1075, ..., -0.0762, 0.0147, -0.0527], + [-0.0077, -0.0070, -0.0268, ..., -0.0797, -0.1680, -0.0487], + [ 0.0146, -0.1010, -0.2117, ..., 0.1067, 0.1558, 0.0293], + ..., + [-0.1087, 0.0071, 0.0897, ..., 0.0919, 0.0955, 0.0425], + [ 0.0441, 0.0034, -0.0133, ..., -0.6673, -0.8508, -0.4418], + [-0.1197, -0.1039, -0.1095, ..., -0.0056, 0.1977, 0.2147]], + device='cuda:0'), grad: tensor([[ 2.9874e-04, 1.3243e-06, 3.0585e-06, ..., 1.0304e-05, + 4.1962e-04, 4.2820e-04], + [ 2.6894e-04, 1.1176e-08, 2.6776e-08, ..., 2.7642e-06, + 4.0984e-04, 4.1819e-04], + [-1.1330e-03, 8.9407e-07, 2.1253e-06, ..., -6.2659e-06, + -1.7529e-03, -1.7881e-03], + ..., + [ 2.0456e-04, -2.2389e-06, -5.2303e-06, ..., -1.0468e-05, + 3.7336e-04, 3.8123e-04], + [ 3.2306e-05, 0.0000e+00, 0.0000e+00, ..., 3.2457e-07, + 4.9293e-05, 5.0306e-05], + [ 7.0691e-05, 1.1642e-09, 2.0955e-09, ..., 7.1479e-07, + 1.0782e-04, 1.1003e-04]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0705, 0.2137, 0.1821, 0.1657, 0.1671, -0.8872, -0.6179], + device='cuda:0'), grad: tensor([ 0.0317, 0.0315, -0.1349, 0.0301, 0.0295, 0.0038, 0.0083], + device='cuda:0') +588 +0.00031882564680131423 +changing lr +epoch 62, time 783.02, cls_loss 0.3126 cls_loss_mapping 1.9805 cls_loss_causal 0.9083 re_mapping 0.0139 re_causal 0.0121 /// teacc 94.47 lr 0.00024472 +Epoch 64, weight, value: tensor([[ 9.4025e-02, 7.5913e-02, 1.0598e-01, ..., -7.7749e-02, + 1.3209e-02, -4.9579e-02], + [-2.3682e-03, -1.4028e-02, -3.3347e-02, ..., -7.4437e-02, + -1.6476e-01, -4.5317e-02], + [ 1.1081e-02, -9.7927e-02, -2.0655e-01, ..., 1.0106e-01, + 1.4903e-01, 2.4322e-02], + ..., + [-1.0632e-01, 8.5577e-03, 9.1031e-02, ..., 9.7435e-02, + 1.0349e-01, 4.5756e-02], + [ 4.3566e-02, -8.3931e-04, -1.6542e-02, ..., -6.6553e-01, + -8.4983e-01, -4.4415e-01], + [-1.2004e-01, -1.0169e-01, -1.0813e-01, ..., -1.3541e-02, + 1.9016e-01, 2.0893e-01]], device='cuda:0'), grad: tensor([[-1.8127e-02, -6.4373e-04, -2.5439e-04, ..., -2.0351e-03, + -1.2655e-03, -2.8305e-03], + [ 1.5135e-03, 4.8733e-04, 1.4710e-04, ..., 1.3208e-04, + 3.8218e-04, 5.6648e-04], + [ 9.4414e-04, 4.8018e-04, 1.4353e-04, ..., 6.8009e-05, + 3.5191e-04, 4.8876e-04], + ..., + [ 1.7059e-02, 1.7729e-03, 5.7697e-04, ..., 1.8110e-03, + 1.9331e-03, 3.5496e-03], + [ 5.3525e-05, 3.9935e-05, 1.1787e-05, ..., 2.6841e-06, + 2.7969e-05, 3.7313e-05], + [ 2.5749e-04, 1.0943e-04, 3.2723e-05, ..., 2.0131e-05, + 8.1897e-05, 1.1659e-04]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0722, 0.2086, 0.1834, 0.1726, 0.1670, -0.8905, -0.6231], + device='cuda:0'), grad: tensor([-0.0863, 0.0341, 0.0322, -0.1433, 0.1533, 0.0026, 0.0074], + device='cuda:0') +588 +0.0002447174185242325 +changing lr +epoch 63, time 791.10, cls_loss 0.2896 cls_loss_mapping 1.9798 cls_loss_causal 0.9002 re_mapping 0.0130 re_causal 0.0113 /// teacc 96.73 lr 0.00018019 +Epoch 65, weight, value: tensor([[ 0.0897, 0.0794, 0.1075, ..., -0.0752, 0.0138, -0.0476], + [-0.0028, -0.0199, -0.0362, ..., -0.0751, -0.1664, -0.0470], + [ 0.0160, -0.0956, -0.2014, ..., 0.1081, 0.1495, 0.0265], + ..., + [-0.1055, 0.0042, 0.0905, ..., 0.0933, 0.1036, 0.0431], + [ 0.0456, -0.0022, -0.0198, ..., -0.6700, -0.8502, -0.4428], + [-0.1198, -0.0995, -0.1094, ..., -0.0149, 0.1900, 0.2070]], + device='cuda:0'), grad: tensor([[ 3.8719e-04, 6.5088e-04, 3.5584e-05, ..., 1.7494e-05, + 9.7096e-05, 4.5800e-04], + [ 1.9753e-04, 6.5279e-04, 2.5883e-05, ..., -3.0156e-06, + 8.0347e-05, 4.3654e-04], + [-5.5885e-03, 3.3259e-04, -2.6321e-04, ..., -3.3379e-04, + -3.0375e-04, -3.0208e-04], + ..., + [ 5.8899e-03, 9.5892e-04, 3.1066e-04, ..., 3.3188e-04, + 4.6277e-04, 1.1625e-03], + [ 2.3320e-05, 6.8247e-05, 2.8629e-06, ..., 3.3434e-07, + 8.8438e-06, 4.6134e-05], + [ 6.8665e-05, 1.9968e-04, 8.4043e-06, ..., 1.0394e-06, + 2.5943e-05, 1.3518e-04]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0750, 0.2072, 0.1763, 0.1767, 0.1697, -0.8926, -0.6247], + device='cuda:0'), grad: tensor([ 0.0307, 0.0306, 0.0069, -0.1343, 0.0534, 0.0032, 0.0094], + device='cuda:0') +588 +0.0001801856965207339 +changing lr +epoch 64, time 790.68, cls_loss 0.3230 cls_loss_mapping 1.9763 cls_loss_causal 0.9211 re_mapping 0.0121 re_causal 0.0105 /// teacc 96.23 lr 0.00012536 +Epoch 66, weight, value: tensor([[ 0.0896, 0.0815, 0.1091, ..., -0.0733, 0.0158, -0.0460], + [-0.0033, -0.0255, -0.0421, ..., -0.0750, -0.1651, -0.0474], + [ 0.0194, -0.0961, -0.1969, ..., 0.1095, 0.1503, 0.0283], + ..., + [-0.1068, 0.0056, 0.0886, ..., 0.0900, 0.1001, 0.0383], + [ 0.0486, -0.0059, -0.0212, ..., -0.6668, -0.8473, -0.4375], + [-0.1204, -0.0982, -0.1078, ..., -0.0144, 0.1904, 0.2093]], + device='cuda:0'), grad: tensor([[ 3.3188e-04, 7.6234e-05, 7.5996e-05, ..., 1.4627e-04, + 4.2081e-05, 3.3665e-04], + [-1.5087e-03, -3.2306e-04, -3.2306e-04, ..., -6.4230e-04, + -2.0349e-04, -1.4515e-03], + [ 3.7408e-04, 7.9930e-05, 7.9930e-05, ..., 1.5903e-04, + 5.0604e-05, 3.5906e-04], + ..., + [ 3.6812e-04, 7.3791e-05, 7.4029e-05, ..., 1.5211e-04, + 5.2303e-05, 3.3760e-04], + [ 5.4836e-05, 1.1742e-05, 1.1742e-05, ..., 2.3350e-05, + 7.4022e-06, 5.2780e-05], + [ 9.3997e-05, 2.0131e-05, 2.0131e-05, ..., 4.0054e-05, + 1.2688e-05, 9.0480e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0744, 0.2027, 0.1767, 0.1782, 0.1681, -0.8861, -0.6207], + device='cuda:0'), grad: tensor([ 0.0316, -0.1344, 0.0332, 0.0255, 0.0308, 0.0049, 0.0084], + device='cuda:0') +588 +0.000125360439090882 +changing lr +epoch 65, time 792.38, cls_loss 0.3172 cls_loss_mapping 1.9716 cls_loss_causal 0.9230 re_mapping 0.0114 re_causal 0.0098 /// teacc 95.98 lr 0.00008035 +Epoch 67, weight, value: tensor([[ 0.0879, 0.0803, 0.1082, ..., -0.0725, 0.0165, -0.0452], + [-0.0021, -0.0270, -0.0425, ..., -0.0759, -0.1655, -0.0470], + [ 0.0177, -0.0954, -0.1975, ..., 0.1083, 0.1491, 0.0283], + ..., + [-0.1052, 0.0073, 0.0912, ..., 0.0872, 0.0981, 0.0358], + [ 0.0486, -0.0031, -0.0203, ..., -0.6690, -0.8496, -0.4403], + [-0.1216, -0.0972, -0.1065, ..., -0.0113, 0.1937, 0.2112]], + device='cuda:0'), grad: tensor([[ 6.2631e-07, 1.6764e-08, 5.8208e-10, ..., 7.2177e-09, + 9.7789e-09, 5.5647e-08], + [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 1.1642e-10], + [ 3.7509e-07, -8.7311e-09, -2.4564e-08, ..., -1.3621e-08, + -2.2119e-09, 4.3423e-08], + ..., + [-1.0012e-06, -7.9162e-09, 2.4098e-08, ..., 6.5193e-09, + -7.4506e-09, -9.9069e-08], + [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00], + [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, + 0.0000e+00, 0.0000e+00]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0739, 0.2011, 0.1782, 0.1772, 0.1696, -0.8900, -0.6189], + device='cuda:0'), grad: tensor([ 2.7716e-06, 3.3760e-09, 1.3104e-06, 9.3132e-10, -4.0829e-06, + 0.0000e+00, 1.0477e-09], device='cuda:0') +588 +8.03520570068517e-05 +changing lr +epoch 66, time 792.09, cls_loss 0.3132 cls_loss_mapping 1.9714 cls_loss_causal 0.9056 re_mapping 0.0109 re_causal 0.0092 /// teacc 95.73 lr 0.00004525 +Epoch 68, weight, value: tensor([[ 0.0903, 0.0820, 0.1100, ..., -0.0714, 0.0175, -0.0440], + [-0.0027, -0.0259, -0.0430, ..., -0.0751, -0.1636, -0.0455], + [ 0.0169, -0.0956, -0.1967, ..., 0.1085, 0.1495, 0.0269], + ..., + [-0.1063, 0.0068, 0.0903, ..., 0.0852, 0.0970, 0.0348], + [ 0.0504, -0.0033, -0.0198, ..., -0.6671, -0.8467, -0.4384], + [-0.1218, -0.0956, -0.1050, ..., -0.0105, 0.1951, 0.2123]], + device='cuda:0'), grad: tensor([[ 6.4969e-05, 1.0461e-05, 1.0937e-05, ..., 3.2902e-05, + 2.4676e-05, 4.6372e-05], + [-2.8104e-05, 1.4314e-06, 1.6550e-06, ..., -5.5879e-06, + -7.2382e-06, -2.6189e-06], + [ 7.3016e-05, 3.6597e-05, 3.8296e-05, ..., 7.2181e-05, + 4.5568e-05, 1.2672e-04], + ..., + [-1.3077e-04, -5.0217e-05, -5.2750e-05, ..., -1.0782e-04, + -6.9261e-05, -1.8013e-04], + [ 1.4659e-06, 3.4645e-07, 3.6298e-07, ..., 8.9966e-07, + 6.3097e-07, 1.3746e-06], + [ 4.8913e-06, 7.1246e-07, 7.5856e-07, ..., 2.3860e-06, + 1.7183e-06, 3.2131e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0743, 0.2025, 0.1760, 0.1748, 0.1704, -0.8874, -0.6184], + device='cuda:0'), grad: tensor([ 1.5821e-03, 3.6407e-04, 2.5311e-03, 6.9046e-04, -5.4092e-03, + 4.3780e-05, 2.0087e-04], device='cuda:0') +588 +4.5251191160326525e-05 +changing lr +epoch 67, time 792.08, cls_loss 0.3078 cls_loss_mapping 1.9745 cls_loss_causal 0.8983 re_mapping 0.0105 re_causal 0.0088 /// teacc 96.73 lr 0.00002013 +Epoch 69, weight, value: tensor([[ 0.0900, 0.0832, 0.1111, ..., -0.0700, 0.0194, -0.0424], + [-0.0029, -0.0280, -0.0443, ..., -0.0757, -0.1646, -0.0467], + [ 0.0167, -0.0953, -0.1967, ..., 0.1070, 0.1488, 0.0259], + ..., + [-0.1058, 0.0064, 0.0900, ..., 0.0859, 0.0970, 0.0354], + [ 0.0498, -0.0023, -0.0195, ..., -0.6663, -0.8453, -0.4373], + [-0.1214, -0.0954, -0.1051, ..., -0.0114, 0.1945, 0.2114]], + device='cuda:0'), grad: tensor([[ 4.9019e-04, -1.2049e-08, 7.5698e-06, ..., 4.6992e-04, + 3.4881e-04, 5.8270e-04], + [-2.1420e-03, 1.0425e-07, -3.2783e-05, ..., -2.0409e-03, + -1.5144e-03, -2.5311e-03], + [ 5.1308e-04, -3.5646e-07, 7.5921e-06, ..., 4.9019e-04, + 3.6383e-04, 6.0749e-04], + ..., + [ 4.8018e-04, 2.6287e-07, 7.5325e-06, ..., 4.5419e-04, + 3.3689e-04, 5.6314e-04], + [ 5.7459e-05, 0.0000e+00, 8.8010e-07, ..., 5.4657e-05, + 4.0561e-05, 6.7770e-05], + [ 2.0742e-04, 9.8953e-10, 3.1758e-06, ..., 1.9717e-04, + 1.4627e-04, 2.4438e-04]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0754, 0.2015, 0.1754, 0.1740, 0.1711, -0.8861, -0.6191], + device='cuda:0'), grad: tensor([ 0.0315, -0.1367, 0.0328, 0.0252, 0.0304, 0.0037, 0.0132], + device='cuda:0') +588 +2.0128530023804673e-05 +changing lr +epoch 68, time 787.10, cls_loss 0.3233 cls_loss_mapping 1.9749 cls_loss_causal 0.9070 re_mapping 0.0102 re_causal 0.0086 /// teacc 95.48 lr 0.00000503 +Epoch 70, weight, value: tensor([[ 0.0898, 0.0830, 0.1111, ..., -0.0705, 0.0195, -0.0426], + [-0.0026, -0.0271, -0.0435, ..., -0.0750, -0.1640, -0.0459], + [ 0.0172, -0.0963, -0.1971, ..., 0.1068, 0.1485, 0.0255], + ..., + [-0.1057, 0.0069, 0.0900, ..., 0.0860, 0.0970, 0.0355], + [ 0.0496, -0.0023, -0.0197, ..., -0.6663, -0.8452, -0.4372], + [-0.1219, -0.0956, -0.1055, ..., -0.0122, 0.1939, 0.2107]], + device='cuda:0'), grad: tensor([[-1.0443e-03, -3.0518e-05, -3.2634e-05, ..., -1.6773e-04, + -1.6141e-04, -2.8014e-04], + [ 8.9312e-04, 2.6107e-05, 2.7910e-05, ..., 1.4341e-04, + 1.3804e-04, 2.3973e-04], + [ 9.2536e-06, 2.7055e-07, 2.8918e-07, ..., 1.4864e-06, + 1.4296e-06, 2.4829e-06], + ..., + [ 1.4174e-04, 4.1462e-06, 4.4294e-06, ..., 2.2769e-05, + 2.1920e-05, 3.8058e-05], + [ 3.7253e-09, 1.1642e-10, 1.1642e-10, ..., 5.8208e-10, + 5.8208e-10, 9.8953e-10], + [ 2.3330e-07, 6.8103e-09, 7.2760e-09, ..., 3.7486e-08, + 3.6031e-08, 6.2631e-08]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0752, 0.2024, 0.1749, 0.1738, 0.1712, -0.8861, -0.6196], + device='cuda:0'), grad: tensor([-4.4060e-03, 3.7689e-03, 3.9071e-05, 7.1304e-08, 5.9843e-04, + 1.5774e-08, 9.8441e-07], device='cuda:0') +588 +5.034667293427056e-06 +changing lr +epoch 69, time 794.32, cls_loss 0.3137 cls_loss_mapping 1.9734 cls_loss_causal 0.9050 re_mapping 0.0102 re_causal 0.0085 /// teacc 95.48 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps5_RA_Adam', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps5_RA_Adam/sketch_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.770934 50.585938 60.452218 53.652695 54.89695 + sketch art_painting cartoon photo Avg +do 19.979639 18.603516 16.638225 11.556886 15.599542 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps5_RA_Adam', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps5_RA_Adam/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.541868 46.777344 64.846416 47.005988 52.876583 + sketch art_painting cartoon photo Avg +do 24.840926 20.751953 21.587031 27.125749 23.154911 diff --git a/Meta-causal/code-withStyleAttack/73090.error b/Meta-causal/code-withStyleAttack/73090.error new file mode 100644 index 0000000000000000000000000000000000000000..9c6218c0a16de0906055735a2be983a2e8bb5681 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73090.error @@ -0,0 +1,19 @@ +Solving dependencies +Downloading: "https://download.pytorch.org/models/resnet18-5c106cde.pth" to /home/yuqian_fu/.cache/torch/hub/checkpoints/resnet18-5c106cde.pth + 0%| | 0.00/44.7M [00:00=3.6)) +Reason for being yanked: deprecated, use 4.5.5.64 +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 672, in + experiment() + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 72, in experiment + os.makedirs(svroot) + File "", line 225, in makedirs +FileExistsError: [Errno 17] File exists: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2' +Downloading: "https://download.pytorch.org/models/resnet18-5c106cde.pth" to /home/yuqian_fu/.cache/torch/hub/checkpoints/resnet18-5c106cde.pth + 0%| | 0.00/44.7M [00:00 + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py", line 30, in main + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py", line 44, in evaluate_pacs + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/serialization.py", line 1065, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/serialization.py", line 468, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/serialization.py", line 449, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/best_cls_net.pkl' +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:47: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py", line 140, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py", line 30, in main + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py", line 47, in evaluate_pacs + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/serialization.py", line 1065, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/serialization.py", line 468, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/serialization.py", line 449, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/last_cls_net.pkl' +srun: error: gcpl4-eu-0: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-withStyleAttack/73725.log b/Meta-causal/code-withStyleAttack/73725.log new file mode 100644 index 0000000000000000000000000000000000000000..911c3968acc1b120967e5f36d0af8cdae3ab396b --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73725.log @@ -0,0 +1,105 @@ +Collecting h5py>=2.9.0 + Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) +Collecting ml-collections + Downloading ml_collections-0.1.1.tar.gz (77 kB) + Preparing metadata (setup.py): started + Preparing metadata (setup.py): finished with status 'done' +Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.1) +Collecting opencv-python==4.5.5.62 + Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) +Collecting scipy>=1.3.2 + Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) +Collecting tensorboard + Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) +Collecting tensorboardX>=1.4 + Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) +Collecting timm + Downloading timm-1.0.8-py3-none-any.whl.metadata (53 kB) +Collecting absl-py (from ml-collections) + Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) +Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) +Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) +Collecting contextlib2 (from ml-collections) + Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) +Collecting grpcio>=1.48.2 (from tensorboard) + Downloading grpcio-1.65.4-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) +Collecting markdown>=2.6.8 (from tensorboard) + Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) +Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) + Downloading protobuf-4.25.4-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) +Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (72.1.0) +Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) + Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) +Collecting werkzeug>=1.0.1 (from tensorboard) + Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) +Collecting packaging (from tensorboardX>=1.4) + Using cached packaging-24.1-py3-none-any.whl.metadata (3.2 kB) +Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.4.0) +Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.19.0) +Collecting huggingface_hub (from timm) + Downloading huggingface_hub-0.24.5-py3-none-any.whl.metadata (13 kB) +Collecting safetensors (from timm) + Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) +Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) +Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) +Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) + Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) +Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) +Collecting tqdm>=4.42.1 (from huggingface_hub->timm) + Using cached tqdm-4.66.5-py3-none-any.whl.metadata (57 kB) +Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) +Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.13.0) +Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) +Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) +Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) +Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) +Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) +Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) +Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.7.4) +Requirement already satisfied: mpmath<1.4,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) +Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 279.0 MB/s eta 0:00:00 +Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 333.2 MB/s eta 0:00:00 +Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 347.3 MB/s eta 0:00:00 +Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 184.5 MB/s eta 0:00:00 +Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) +Downloading timm-1.0.8-py3-none-any.whl (2.3 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 275.2 MB/s eta 0:00:00 +Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) +Downloading grpcio-1.65.4-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.7 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.7/5.7 MB 119.7 MB/s eta 0:00:00 +Downloading Markdown-3.6-py3-none-any.whl (105 kB) +Downloading protobuf-4.25.4-cp37-abi3-manylinux2014_x86_64.whl (294 kB) +Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 278.1 MB/s eta 0:00:00 +Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) +Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) +Downloading huggingface_hub-0.24.5-py3-none-any.whl (417 kB) +Using cached packaging-24.1-py3-none-any.whl (53 kB) +Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) + ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 167.9 MB/s eta 0:00:00 +Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) +Using cached tqdm-4.66.5-py3-none-any.whl (78 kB) +Building wheels for collected packages: ml-collections + Building wheel for ml-collections (setup.py): started + Building wheel for ml-collections (setup.py): finished with status 'done' + Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94507 sha256=f8f3fbc001d31b15b1dd062d26dcf5af8a419e69223933f1f3b206ea12eae5d6 + Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 +Successfully built ml-collections +Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm +Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.65.4 h5py-3.11.0 huggingface_hub-0.24.5 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.4 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.8 tqdm-4.66.5 werkzeug-3.0.3 +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/sketch_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last diff --git a/Meta-causal/code-withStyleAttack/73726.error b/Meta-causal/code-withStyleAttack/73726.error new file mode 100644 index 0000000000000000000000000000000000000000..d4cae9c14f1b02ce78b5efc4ecc9cba1f06923fb --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73726.error @@ -0,0 +1,18 @@ +Solving dependencies +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +run_my_joint_v13_test.sh: line 34: andm: command not found +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:44: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:58: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:68: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:47: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:61: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:71: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) diff --git a/Meta-causal/code-withStyleAttack/73726.log b/Meta-causal/code-withStyleAttack/73726.log new file mode 100644 index 0000000000000000000000000000000000000000..38d6b08519e47a44ca0812228fb7b0df2d2a2682 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73726.log @@ -0,0 +1,2030 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[ 5.0321e-03, -2.0984e-02, -9.3549e-03, ..., 6.2974e-03, + -1.1026e-02, 1.6907e-02], + [ 1.8178e-03, -9.4542e-04, 9.6098e-03, ..., -3.6409e-03, + 1.6892e-02, 1.7561e-02], + [ 4.8632e-03, 1.5569e-02, -9.7156e-03, ..., -6.5302e-03, + -7.2668e-04, -2.3526e-03], + ..., + [-2.8515e-03, -2.0712e-02, 1.6087e-02, ..., 1.3537e-02, + -8.7155e-03, -1.1101e-02], + [-1.6359e-02, 1.5327e-03, 4.4491e-03, ..., 6.7593e-04, + 1.7600e-02, -2.0811e-02], + [-9.0949e-03, -4.9629e-03, 1.9741e-02, ..., 1.5146e-02, + -4.0215e-03, 6.5127e-05]], device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0169, -0.0141, 0.0051, 0.0054, 0.0115, -0.0021, 0.0115], + device='cuda:0'), grad: None +588 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 828.31, cls_loss 5.2711 cls_loss_mapping 1.1924 cls_loss_causal 1.4418 re_mapping 0.4544 re_causal 0.4542 /// teacc 83.17 lr 0.00999497 +Epoch 2, weight, value: tensor([[ 0.1217, 0.1173, 0.0514, ..., -0.0046, -0.0003, 0.0027], + [ 0.1174, 0.0145, 0.1393, ..., -0.0966, -0.0285, -0.0377], + [-0.0571, 0.0007, -0.0756, ..., 0.0668, 0.0251, 0.0395], + ..., + [-0.1860, -0.1862, -0.1100, ..., 0.0255, -0.0027, -0.0029], + [ 0.0276, 0.0367, 0.0287, ..., 0.0273, 0.0208, -0.0063], + [ 0.0405, 0.0606, 0.0413, ..., -0.0429, -0.0609, -0.0464]], + device='cuda:0'), grad: tensor([[ 2.3999e-01, 1.5540e-01, 1.7395e-01, ..., 6.3599e-02, + 3.1830e-02, 1.4832e-02], + [-2.9297e-01, -1.7139e-01, -1.9385e-01, ..., -9.8938e-02, + -5.2307e-02, -2.4780e-02], + [-7.4272e-03, -2.2011e-03, -2.3670e-03, ..., -7.6981e-03, + -4.7684e-03, -3.5572e-03], + ..., + [ 4.8889e-02, 1.5366e-02, 1.8539e-02, ..., 3.4393e-02, + 2.0172e-02, 1.0971e-02], + [ 1.1436e-02, 2.8381e-03, 3.6564e-03, ..., 8.5602e-03, + 5.0163e-03, 2.5101e-03], + [ 1.0669e-04, 3.4630e-05, 4.2349e-05, ..., 7.0214e-05, + 4.0710e-05, 2.0385e-05]], device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0438, -0.0197, -0.0433, -0.1126, 0.0461, 0.0814, 0.0377], + device='cuda:0'), grad: tensor([ 2.2461e-01, -3.2373e-01, -1.5762e-02, 2.9569e-08, 9.1003e-02, + 2.3636e-02, 1.9741e-04], device='cuda:0') +588 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 799.62, cls_loss 0.6358 cls_loss_mapping 0.4896 cls_loss_causal 0.9883 re_mapping 0.1652 re_causal 0.1647 /// teacc 84.42 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 0.2203, 0.2033, 0.1431, ..., 0.0037, 0.0061, 0.0012], + [ 0.0460, -0.0347, 0.0835, ..., -0.1054, -0.0311, -0.0345], + [-0.0530, -0.0073, -0.0789, ..., 0.0554, 0.0163, 0.0260], + ..., + [-0.2380, -0.2207, -0.1522, ..., 0.0301, -0.0020, 0.0036], + [ 0.0159, 0.0271, 0.0189, ..., 0.0258, 0.0218, -0.0054], + [ 0.0311, 0.0490, 0.0286, ..., -0.0492, -0.0671, -0.0513]], + device='cuda:0'), grad: tensor([[ 1.8129e-03, 1.2245e-03, 1.4124e-03, ..., 3.6287e-04, + 1.1533e-04, 1.2660e-04], + [-3.8757e-02, -1.6647e-02, -2.4887e-02, ..., -2.3804e-02, + -1.8066e-02, -1.4252e-02], + [ 6.5880e-03, 4.1084e-03, 4.5586e-03, ..., 2.2697e-03, + 1.5936e-03, 1.1473e-03], + ..., + [ 1.9318e-02, 7.3471e-03, 1.2062e-02, ..., 1.3222e-02, + 1.0208e-02, 8.0872e-03], + [ 1.4770e-04, 8.6725e-05, 1.0210e-04, ..., 5.6893e-05, + 4.0770e-05, 2.8580e-05], + [ 6.7830e-05, 4.5627e-05, 4.9263e-05, ..., 1.7539e-05, + 1.1235e-05, 6.4932e-06]], device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0477, -0.0237, -0.0222, -0.0822, 0.0136, 0.0640, 0.0352], + device='cuda:0'), grad: tensor([ 2.1400e-03, -9.0088e-02, 8.7280e-03, 2.9373e-02, 4.9530e-02, + 2.3103e-04, 8.0168e-05], device='cuda:0') +588 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 801.14, cls_loss 0.2952 cls_loss_mapping 0.2818 cls_loss_causal 0.8366 re_mapping 0.1379 re_causal 0.1375 /// teacc 89.70 lr 0.00995475 +Epoch 4, weight, value: tensor([[ 0.2456, 0.2246, 0.1671, ..., 0.0053, 0.0070, 0.0030], + [ 0.0538, -0.0369, 0.0839, ..., -0.0998, -0.0256, -0.0324], + [-0.0791, -0.0294, -0.1034, ..., 0.0457, 0.0108, 0.0191], + ..., + [-0.2597, -0.2334, -0.1673, ..., 0.0324, -0.0027, 0.0040], + [ 0.0087, 0.0224, 0.0140, ..., 0.0238, 0.0202, -0.0057], + [ 0.0509, 0.0667, 0.0467, ..., -0.0484, -0.0651, -0.0492]], + device='cuda:0'), grad: tensor([[-1.7810e-01, -8.4961e-02, -9.7595e-02, ..., -4.4067e-02, + -2.4261e-02, -1.2955e-02], + [ 1.4136e-01, 7.3181e-02, 8.9783e-02, ..., 3.0365e-02, + 1.6113e-02, 8.3160e-03], + [ 2.0935e-02, 6.7139e-03, 3.6945e-03, ..., 7.2441e-03, + 4.4250e-03, 2.4929e-03], + ..., + [ 1.4091e-02, 4.5280e-03, 3.6755e-03, ..., 5.7526e-03, + 3.3150e-03, 1.9131e-03], + [ 2.3770e-04, 7.3910e-05, 6.9916e-05, ..., 1.0741e-04, + 6.1035e-05, 3.5971e-05], + [ 1.6940e-04, 6.6280e-05, 6.6817e-05, ..., 6.1005e-05, + 3.3110e-05, 1.8448e-05]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0681, -0.0062, -0.0313, -0.0872, 0.0010, 0.0518, 0.0354], + device='cuda:0'), grad: tensor([-0.2710, 0.1898, 0.0462, 0.0029, 0.0312, 0.0005, 0.0003], + device='cuda:0') +588 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 797.88, cls_loss 0.1408 cls_loss_mapping 0.1878 cls_loss_causal 0.7363 re_mapping 0.1118 re_causal 0.1116 /// teacc 90.45 lr 0.00991965 +Epoch 5, weight, value: tensor([[ 0.2791, 0.2501, 0.2026, ..., 0.0055, 0.0088, 0.0030], + [ 0.0312, -0.0494, 0.0644, ..., -0.1022, -0.0300, -0.0349], + [-0.0713, -0.0226, -0.0975, ..., 0.0506, 0.0151, 0.0221], + ..., + [-0.2745, -0.2473, -0.1853, ..., 0.0290, -0.0043, 0.0031], + [ 0.0101, 0.0234, 0.0154, ..., 0.0239, 0.0207, -0.0047], + [ 0.0501, 0.0629, 0.0435, ..., -0.0468, -0.0634, -0.0481]], + device='cuda:0'), grad: tensor([[ 1.3316e-04, 4.7386e-05, 6.3956e-05, ..., 7.7188e-05, + 5.0247e-05, 4.2111e-05], + [-5.4911e-06, -2.1532e-06, -3.4533e-06, ..., -1.8850e-06, + -1.5842e-06, -8.4657e-07], + [-1.4305e-04, -4.8727e-05, -6.7532e-05, ..., -9.1493e-05, + -6.0022e-05, -5.2571e-05], + ..., + [ 2.5645e-05, 9.9912e-06, 1.3128e-05, ..., 1.3083e-05, + 8.5607e-06, 6.8210e-06], + [-7.0930e-05, -3.4750e-05, -3.9726e-05, ..., -1.9625e-05, + -1.1273e-05, -5.3793e-06], + [ 3.4571e-05, 1.6645e-05, 1.9327e-05, ..., 1.0930e-05, + 6.5230e-06, 3.8370e-06]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0838, -0.0224, -0.0294, -0.0923, 0.0014, 0.0493, 0.0404], + device='cuda:0'), grad: tensor([ 3.7646e-04, -7.3053e-06, -4.4131e-04, 6.0856e-05, 6.4552e-05, + -1.1587e-04, 6.1989e-05], device='cuda:0') +588 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 805.53, cls_loss 0.0873 cls_loss_mapping 0.1471 cls_loss_causal 0.6787 re_mapping 0.1024 re_causal 0.1030 /// teacc 92.96 lr 0.00987464 +Epoch 6, weight, value: tensor([[ 0.2880, 0.2631, 0.2133, ..., 0.0096, 0.0128, 0.0064], + [ 0.0292, -0.0533, 0.0617, ..., -0.0958, -0.0255, -0.0311], + [-0.0630, -0.0201, -0.0924, ..., 0.0469, 0.0108, 0.0176], + ..., + [-0.2901, -0.2574, -0.1985, ..., 0.0203, -0.0107, -0.0024], + [ 0.0088, 0.0226, 0.0147, ..., 0.0234, 0.0208, -0.0040], + [ 0.0459, 0.0593, 0.0399, ..., -0.0480, -0.0637, -0.0485]], + device='cuda:0'), grad: tensor([[-1.7767e-03, -1.2865e-03, -1.3285e-03, ..., -2.8992e-04, + -1.9503e-04, -1.2255e-04], + [ 4.7207e-04, 3.9721e-04, 3.8815e-04, ..., 6.9976e-05, + 5.1200e-05, 4.0233e-05], + [ 3.7193e-04, 2.0504e-04, 2.3699e-04, ..., 1.0020e-04, + 6.5267e-05, 4.1932e-05], + ..., + [ 1.6391e-06, 1.1921e-06, 1.2303e-06, ..., 2.1886e-07, + 1.4389e-07, 7.5554e-08], + [ 5.0634e-05, 3.7193e-05, 3.8207e-05, ..., 6.4634e-06, + 4.2431e-06, 2.1756e-06], + [ 2.2482e-06, 1.6494e-06, 1.6959e-06, ..., 3.4180e-07, + 2.3004e-07, 1.4040e-07]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0861, -0.0168, -0.0168, -0.0818, -0.0185, 0.0443, 0.0335], + device='cuda:0'), grad: tensor([-1.8644e-03, 5.5695e-04, 4.9257e-04, 7.6818e-04, 1.4622e-06, + 4.4137e-05, 2.2613e-06], device='cuda:0') +588 +0.009874639560909117 +changing lr +epoch 5, time 795.78, cls_loss 0.0215 cls_loss_mapping 0.1020 cls_loss_causal 0.6322 re_mapping 0.0907 re_causal 0.0919 /// teacc 91.46 lr 0.00981981 +Epoch 7, weight, value: tensor([[ 0.2936, 0.2675, 0.2189, ..., 0.0081, 0.0102, 0.0045], + [ 0.0293, -0.0530, 0.0608, ..., -0.0915, -0.0228, -0.0291], + [-0.0717, -0.0265, -0.0971, ..., 0.0444, 0.0096, 0.0166], + ..., + [-0.2808, -0.2497, -0.1938, ..., 0.0218, -0.0081, -0.0004], + [ 0.0066, 0.0212, 0.0130, ..., 0.0227, 0.0201, -0.0039], + [ 0.0466, 0.0576, 0.0395, ..., -0.0465, -0.0620, -0.0472]], + device='cuda:0'), grad: tensor([[ 5.6992e-03, 2.0063e-04, 8.2016e-04, ..., 2.8534e-03, + 2.1381e-03, 1.2836e-03], + [ 9.4762e-07, -1.4377e-08, 6.7230e-08, ..., 5.1595e-07, + 3.5716e-07, 1.7753e-07], + [ 1.1247e-04, 4.3064e-06, 1.6466e-05, ..., 5.6356e-05, + 4.2289e-05, 2.5496e-05], + ..., + [-5.8174e-03, -2.0444e-04, -8.3685e-04, ..., -2.9144e-03, + -2.1820e-03, -1.3103e-03], + [ 3.6247e-06, 1.7602e-07, 5.7789e-07, ..., 1.8282e-06, + 1.3737e-06, 8.4285e-07], + [ 2.4717e-06, 1.7020e-07, 4.6194e-07, ..., 1.2461e-06, + 9.4157e-07, 5.9605e-07]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0911, -0.0088, -0.0323, -0.0890, -0.0097, 0.0401, 0.0377], + device='cuda:0'), grad: tensor([ 1.8372e-02, 3.5465e-06, 3.6073e-04, 1.2424e-06, -1.8753e-02, + 1.1586e-05, 7.7859e-06], device='cuda:0') +588 +0.009819814303479266 +changing lr +epoch 6, time 800.31, cls_loss 0.0243 cls_loss_mapping 0.0918 cls_loss_causal 0.6256 re_mapping 0.0766 re_causal 0.0788 /// teacc 91.46 lr 0.00975528 +Epoch 8, weight, value: tensor([[ 0.2980, 0.2746, 0.2264, ..., 0.0057, 0.0083, 0.0022], + [ 0.0237, -0.0570, 0.0542, ..., -0.0866, -0.0195, -0.0258], + [-0.0704, -0.0298, -0.0975, ..., 0.0443, 0.0106, 0.0166], + ..., + [-0.2812, -0.2497, -0.1957, ..., 0.0196, -0.0108, -0.0023], + [ 0.0065, 0.0209, 0.0131, ..., 0.0221, 0.0197, -0.0036], + [ 0.0429, 0.0549, 0.0364, ..., -0.0467, -0.0615, -0.0468]], + device='cuda:0'), grad: tensor([[ 4.1890e-04, -1.3962e-05, 2.1800e-05, ..., 3.5572e-04, + 2.9254e-04, 2.3329e-04], + [ 2.1112e-04, 2.6956e-05, 4.5240e-05, ..., 1.0735e-04, + 7.2122e-05, 5.5432e-05], + [ 1.2074e-03, 4.4554e-05, 1.4460e-04, ..., 9.3460e-04, + 7.7057e-04, 6.1226e-04], + ..., + [-2.0294e-03, -7.0989e-05, -2.4092e-04, ..., -1.5392e-03, + -1.2522e-03, -9.9373e-04], + [ 6.6102e-05, 4.1723e-06, 9.6187e-06, ..., 4.8727e-05, + 3.9905e-05, 3.1650e-05], + [ 8.1509e-06, 9.3179e-07, 1.5674e-06, ..., 5.8152e-06, + 4.8541e-06, 3.8333e-06]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0899, -0.0095, -0.0254, -0.0850, -0.0107, 0.0374, 0.0315], + device='cuda:0'), grad: tensor([ 1.3647e-03, 5.8794e-04, 3.6697e-03, 3.5238e-04, -6.1874e-03, + 1.9550e-04, 2.2992e-05], device='cuda:0') +588 +0.009755282581475767 +changing lr +epoch 7, time 805.18, cls_loss 0.0157 cls_loss_mapping 0.0757 cls_loss_causal 0.6106 re_mapping 0.0691 re_causal 0.0726 /// teacc 91.46 lr 0.00968117 +Epoch 9, weight, value: tensor([[ 0.3025, 0.2786, 0.2318, ..., 0.0029, 0.0058, -0.0004], + [ 0.0247, -0.0542, 0.0536, ..., -0.0836, -0.0183, -0.0246], + [-0.0734, -0.0333, -0.0990, ..., 0.0455, 0.0127, 0.0187], + ..., + [-0.2799, -0.2501, -0.1977, ..., 0.0195, -0.0109, -0.0026], + [ 0.0050, 0.0200, 0.0123, ..., 0.0215, 0.0194, -0.0033], + [ 0.0415, 0.0535, 0.0354, ..., -0.0462, -0.0604, -0.0460]], + device='cuda:0'), grad: tensor([[ 3.4999e-06, 1.0198e-06, 1.2470e-06, ..., 1.5171e-06, + 1.0785e-06, 9.0664e-07], + [ 7.9256e-07, -1.2777e-08, -7.0315e-08, ..., 8.2096e-07, + 6.3796e-07, 5.7789e-07], + [ 1.0952e-05, -3.9872e-08, 6.2166e-07, ..., 1.2264e-05, + 9.8646e-06, 9.0227e-06], + ..., + [-1.9744e-05, -1.2182e-06, -2.2389e-06, ..., -1.6436e-05, + -1.2703e-05, -1.1384e-05], + [-1.6108e-05, -9.7379e-06, -1.0706e-05, ..., -4.1425e-06, + -3.0976e-06, -2.4643e-06], + [ 1.4357e-05, 8.5533e-06, 9.4250e-06, ..., 3.7495e-06, + 2.7921e-06, 2.2240e-06]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0908, -0.0060, -0.0282, -0.0862, -0.0051, 0.0331, 0.0290], + device='cuda:0'), grad: tensor([ 8.4564e-06, 2.9206e-06, 4.2200e-05, 1.5944e-05, -6.7711e-05, + -2.1055e-05, 1.9222e-05], device='cuda:0') +588 +0.009681174353198686 +changing lr +epoch 8, time 802.22, cls_loss 0.0149 cls_loss_mapping 0.0665 cls_loss_causal 0.5558 re_mapping 0.0621 re_causal 0.0664 /// teacc 91.46 lr 0.00959764 +Epoch 10, weight, value: tensor([[ 3.0159e-01, 2.7923e-01, 2.3433e-01, ..., -2.8317e-03, + 1.3968e-04, -5.7595e-03], + [ 1.8288e-02, -5.7675e-02, 4.7190e-02, ..., -8.2277e-02, + -1.8411e-02, -2.4713e-02], + [-7.7096e-02, -3.6908e-02, -1.0139e-01, ..., 4.4417e-02, + 1.2294e-02, 1.8147e-02], + ..., + [-2.7502e-01, -2.4551e-01, -1.9469e-01, ..., 2.1172e-02, + -8.4590e-03, -3.5385e-04], + [ 7.3740e-03, 1.9798e-02, 1.2663e-02, ..., 2.2504e-02, + 2.0140e-02, -2.0309e-03], + [ 4.3695e-02, 5.3812e-02, 3.5940e-02, ..., -4.4764e-02, + -5.8660e-02, -4.4696e-02]], device='cuda:0'), grad: tensor([[-7.9956e-03, -4.4746e-03, -4.3983e-03, ..., -1.5936e-03, + -9.3937e-04, -7.4768e-04], + [ 1.0624e-03, 5.6171e-04, 5.6887e-04, ..., 2.3031e-04, + 1.4746e-04, 1.1587e-04], + [ 1.4887e-03, 8.1444e-04, 8.0538e-04, ..., 3.1209e-04, + 1.9157e-04, 1.5295e-04], + ..., + [ 3.4485e-03, 2.0180e-03, 1.9407e-03, ..., 6.3705e-04, + 3.4356e-04, 2.7776e-04], + [ 4.9543e-04, 2.6703e-04, 2.6989e-04, ..., 1.0324e-04, + 6.4194e-05, 5.0187e-05], + [ 1.2350e-03, 6.6280e-04, 6.6805e-04, ..., 2.6107e-04, + 1.6379e-04, 1.2863e-04]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0872, -0.0109, -0.0321, -0.0853, -0.0045, 0.0384, 0.0337], + device='cuda:0'), grad: tensor([-0.0128, 0.0018, 0.0024, 0.0004, 0.0054, 0.0008, 0.0020], + device='cuda:0') +588 +0.009597638862757255 +changing lr +epoch 9, time 803.61, cls_loss 0.0121 cls_loss_mapping 0.0578 cls_loss_causal 0.5632 re_mapping 0.0554 re_causal 0.0609 /// teacc 90.95 lr 0.00950484 +Epoch 11, weight, value: tensor([[ 0.2987, 0.2761, 0.2324, ..., -0.0008, 0.0016, -0.0040], + [ 0.0157, -0.0577, 0.0448, ..., -0.0808, -0.0185, -0.0249], + [-0.0757, -0.0366, -0.1007, ..., 0.0421, 0.0109, 0.0166], + ..., + [-0.2711, -0.2433, -0.1930, ..., 0.0202, -0.0087, -0.0009], + [ 0.0046, 0.0185, 0.0115, ..., 0.0212, 0.0192, -0.0023], + [ 0.0445, 0.0547, 0.0372, ..., -0.0443, -0.0575, -0.0439]], + device='cuda:0'), grad: tensor([[ 2.2084e-05, 5.6289e-06, 7.4245e-06, ..., 9.7901e-06, + 5.7407e-06, 5.5172e-06], + [ 2.4259e-05, 7.4990e-06, 9.7901e-06, ..., 8.8587e-06, + 5.1893e-06, 4.9546e-06], + [-1.1665e-04, -3.3945e-05, -4.4584e-05, ..., -4.6194e-05, + -2.7493e-05, -2.6301e-05], + ..., + [ 3.1054e-05, 9.4697e-06, 1.2361e-05, ..., 1.1608e-05, + 6.7763e-06, 6.5193e-06], + [ 5.4277e-06, 1.6382e-06, 2.1402e-06, ..., 2.0564e-06, + 1.2172e-06, 1.1604e-06], + [ 7.7337e-06, 2.5276e-06, 3.2857e-06, ..., 2.6897e-06, + 1.6196e-06, 1.5302e-06]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0942, -0.0130, -0.0315, -0.0807, -0.0044, 0.0305, 0.0308], + device='cuda:0'), grad: tensor([ 5.7161e-05, 5.7817e-05, -2.8586e-04, 6.5506e-05, 7.4863e-05, + 1.3068e-05, 1.7837e-05], device='cuda:0') +588 +0.009504844339512096 +changing lr +epoch 10, time 800.50, cls_loss 0.0193 cls_loss_mapping 0.0536 cls_loss_causal 0.5720 re_mapping 0.0534 re_causal 0.0604 /// teacc 91.71 lr 0.00940298 +Epoch 12, weight, value: tensor([[ 0.2950, 0.2705, 0.2280, ..., -0.0022, -0.0017, -0.0069], + [ 0.0223, -0.0496, 0.0510, ..., -0.0767, -0.0150, -0.0217], + [-0.0756, -0.0387, -0.1018, ..., 0.0406, 0.0101, 0.0156], + ..., + [-0.2702, -0.2412, -0.1921, ..., 0.0179, -0.0095, -0.0019], + [ 0.0051, 0.0183, 0.0114, ..., 0.0211, 0.0190, -0.0019], + [ 0.0425, 0.0536, 0.0363, ..., -0.0440, -0.0565, -0.0432]], + device='cuda:0'), grad: tensor([[ 1.8215e-04, 3.8624e-05, 4.7177e-05, ..., 1.3149e-04, + 1.0204e-04, 9.9838e-05], + [ 5.9814e-03, 2.3289e-03, 2.5291e-03, ..., 1.1272e-03, + 5.2547e-04, 5.1832e-04], + [-5.6267e-03, -2.2545e-03, -2.4681e-03, ..., -9.4128e-04, + -4.0197e-04, -3.9339e-04], + ..., + [-5.6887e-04, -1.2016e-04, -1.1533e-04, ..., -3.3188e-04, + -2.3425e-04, -2.3437e-04], + [-3.4153e-05, -1.9282e-05, -2.2233e-05, ..., -6.6087e-06, + -6.7391e-06, -4.1500e-06], + [ 4.2230e-05, 1.8030e-05, 1.9923e-05, ..., 1.1779e-05, + 8.8066e-06, 7.3500e-06]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.1016, -0.0127, -0.0305, -0.0851, -0.0045, 0.0304, 0.0259], + device='cuda:0'), grad: tensor([ 6.6280e-04, 1.2558e-02, -1.1635e-02, 5.4300e-05, -1.6785e-03, + -3.4899e-05, 7.4983e-05], device='cuda:0') +588 +0.009402977659283692 +changing lr +epoch 11, time 798.45, cls_loss 0.0046 cls_loss_mapping 0.0423 cls_loss_causal 0.5215 re_mapping 0.0447 re_causal 0.0523 /// teacc 92.21 lr 0.00929224 +Epoch 13, weight, value: tensor([[ 0.3001, 0.2751, 0.2348, ..., -0.0022, -0.0019, -0.0069], + [ 0.0182, -0.0511, 0.0465, ..., -0.0744, -0.0143, -0.0207], + [-0.0784, -0.0421, -0.1037, ..., 0.0389, 0.0092, 0.0144], + ..., + [-0.2661, -0.2381, -0.1907, ..., 0.0176, -0.0092, -0.0018], + [ 0.0048, 0.0180, 0.0111, ..., 0.0200, 0.0178, -0.0024], + [ 0.0395, 0.0506, 0.0335, ..., -0.0430, -0.0551, -0.0421]], + device='cuda:0'), grad: tensor([[4.9502e-05, 4.2245e-06, 1.2688e-05, ..., 4.2140e-05, 3.3408e-05, + 3.4004e-05], + [3.4690e-05, 1.9267e-05, 2.3529e-05, ..., 1.5870e-05, 1.2502e-05, + 1.3538e-05], + [3.8922e-05, 2.1085e-05, 2.4259e-05, ..., 1.8880e-05, 1.6168e-05, + 1.5691e-05], + ..., + [1.9991e-04, 8.3208e-05, 9.4414e-05, ..., 6.9439e-05, 5.0664e-05, + 4.9055e-05], + [6.3658e-05, 2.6479e-05, 3.1114e-05, ..., 2.2322e-05, 1.6332e-05, + 1.6034e-05], + [1.2767e-04, 5.2661e-05, 6.1005e-05, ..., 4.1693e-05, 3.0011e-05, + 2.8968e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.1029, -0.0135, -0.0320, -0.0831, -0.0018, 0.0279, 0.0242], + device='cuda:0'), grad: tensor([ 1.4329e-04, 2.8968e-05, 6.3419e-05, -1.0004e-03, 3.8838e-04, + 1.2207e-04, 2.5368e-04], device='cuda:0') +588 +0.009292243968009333 +changing lr +epoch 12, time 798.32, cls_loss 0.0049 cls_loss_mapping 0.0433 cls_loss_causal 0.5335 re_mapping 0.0382 re_causal 0.0468 /// teacc 91.71 lr 0.00917287 +Epoch 14, weight, value: tensor([[ 0.3032, 0.2785, 0.2400, ..., -0.0022, -0.0019, -0.0067], + [ 0.0162, -0.0523, 0.0430, ..., -0.0712, -0.0129, -0.0192], + [-0.0783, -0.0429, -0.1035, ..., 0.0361, 0.0071, 0.0123], + ..., + [-0.2636, -0.2361, -0.1904, ..., 0.0174, -0.0085, -0.0016], + [ 0.0028, 0.0167, 0.0099, ..., 0.0196, 0.0176, -0.0021], + [ 0.0369, 0.0479, 0.0314, ..., -0.0427, -0.0544, -0.0417]], + device='cuda:0'), grad: tensor([[-1.3769e-04, -6.1214e-05, -6.0767e-05, ..., -1.7315e-05, + -6.6720e-06, -5.9307e-06], + [-1.9763e-06, -1.5730e-06, -1.7546e-06, ..., -2.4680e-07, + -2.6729e-07, -2.4703e-07], + [ 4.6760e-05, 2.0847e-05, 2.0757e-05, ..., 5.7667e-06, + 2.2184e-06, 2.0135e-06], + ..., + [ 1.9595e-05, 9.2313e-06, 9.1717e-06, ..., 2.1346e-06, + 7.0641e-07, 5.6112e-07], + [ 1.0327e-05, 4.5635e-06, 4.5486e-06, ..., 1.3914e-06, + 5.8440e-07, 5.2527e-07], + [ 5.9336e-05, 2.6643e-05, 2.6569e-05, ..., 7.6443e-06, + 3.1237e-06, 2.7958e-06]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.1022, -0.0082, -0.0316, -0.0808, -0.0023, 0.0228, 0.0218], + device='cuda:0'), grad: tensor([-2.4629e-04, -1.5059e-06, 8.2552e-05, 7.3090e-06, 3.3557e-05, + 1.8746e-05, 1.0586e-04], device='cuda:0') +588 +0.009172866268606516 +changing lr +epoch 13, time 806.83, cls_loss 0.0036 cls_loss_mapping 0.0387 cls_loss_causal 0.5195 re_mapping 0.0331 re_causal 0.0424 /// teacc 91.71 lr 0.00904508 +Epoch 15, weight, value: tensor([[ 0.3061, 0.2818, 0.2440, ..., -0.0028, -0.0028, -0.0074], + [ 0.0131, -0.0530, 0.0402, ..., -0.0699, -0.0130, -0.0191], + [-0.0803, -0.0457, -0.1050, ..., 0.0343, 0.0060, 0.0111], + ..., + [-0.2625, -0.2347, -0.1900, ..., 0.0170, -0.0078, -0.0012], + [ 0.0047, 0.0168, 0.0103, ..., 0.0201, 0.0180, -0.0013], + [ 0.0353, 0.0459, 0.0296, ..., -0.0416, -0.0529, -0.0404]], + device='cuda:0'), grad: tensor([[ 1.4208e-05, -2.1197e-06, -2.3581e-06, ..., 9.9689e-06, + 6.9700e-06, 6.8434e-06], + [ 3.2187e-05, 5.5358e-06, 4.6119e-06, ..., 1.8895e-05, + 1.3009e-05, 1.3582e-05], + [-1.7479e-05, -7.4096e-06, -7.0855e-06, ..., 5.6066e-06, + 6.5379e-06, 7.4357e-06], + ..., + [-1.4894e-05, 6.0685e-06, 7.9274e-06, ..., -2.1115e-05, + -1.3746e-05, -1.5616e-05], + [-3.7223e-05, -8.8513e-06, -9.4548e-06, ..., -2.1309e-05, + -1.8582e-05, -1.7688e-05], + [ 3.4779e-05, 1.2688e-05, 1.2919e-05, ..., 1.4134e-05, + 1.1563e-05, 1.1049e-05]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0998, -0.0093, -0.0319, -0.0797, -0.0044, 0.0277, 0.0211], + device='cuda:0'), grad: tensor([ 6.1989e-05, 1.0294e-04, -3.8952e-05, -1.7032e-05, -8.8334e-05, + -9.9003e-05, 7.8380e-05], device='cuda:0') +588 +0.00904508497187474 +changing lr +epoch 14, time 797.25, cls_loss 0.0050 cls_loss_mapping 0.0371 cls_loss_causal 0.5289 re_mapping 0.0291 re_causal 0.0393 /// teacc 91.71 lr 0.00890916 +Epoch 16, weight, value: tensor([[ 0.3069, 0.2842, 0.2474, ..., -0.0036, -0.0034, -0.0078], + [ 0.0107, -0.0545, 0.0363, ..., -0.0687, -0.0135, -0.0194], + [-0.0799, -0.0465, -0.1049, ..., 0.0330, 0.0051, 0.0101], + ..., + [-0.2605, -0.2335, -0.1903, ..., 0.0171, -0.0070, -0.0008], + [ 0.0024, 0.0154, 0.0090, ..., 0.0195, 0.0176, -0.0012], + [ 0.0349, 0.0444, 0.0292, ..., -0.0407, -0.0516, -0.0394]], + device='cuda:0'), grad: tensor([[-5.1022e-04, -3.0041e-04, -2.7776e-04, ..., -7.4565e-05, + -4.5955e-05, -4.3511e-05], + [-1.5073e-05, 2.7359e-05, 2.2188e-05, ..., -1.0446e-05, + -5.6513e-06, -7.4841e-06], + [ 6.1035e-05, 3.2037e-05, 2.9325e-05, ..., 1.3839e-06, + -2.5630e-06, -9.0757e-07], + ..., + [ 1.2231e-04, 5.1975e-05, 5.0098e-05, ..., 2.5004e-05, + 1.5914e-05, 1.5207e-05], + [ 2.9877e-05, 1.5073e-05, 1.4149e-05, ..., 5.1111e-06, + 3.1367e-06, 2.9914e-06], + [ 1.4663e-04, 7.7069e-05, 7.1466e-05, ..., 2.3887e-05, + 1.4424e-05, 1.3933e-05]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0935, -0.0055, -0.0301, -0.0787, -0.0017, 0.0221, 0.0232], + device='cuda:0'), grad: tensor([-7.3004e-04, -1.7083e-04, 8.3864e-05, 2.6584e-04, 2.5558e-04, + 5.2065e-05, 2.4271e-04], device='cuda:0') +588 +0.008909157412340152 +changing lr +epoch 15, time 804.54, cls_loss 0.0026 cls_loss_mapping 0.0329 cls_loss_causal 0.4981 re_mapping 0.0250 re_causal 0.0353 /// teacc 92.46 lr 0.00876536 +Epoch 17, weight, value: tensor([[ 3.0877e-01, 2.8484e-01, 2.4943e-01, ..., -3.7709e-03, + -3.5700e-03, -7.9942e-03], + [ 9.6614e-03, -5.4450e-02, 3.4373e-02, ..., -6.6495e-02, + -1.2707e-02, -1.8452e-02], + [-8.0009e-02, -4.7167e-02, -1.0457e-01, ..., 3.1683e-02, + 4.3488e-03, 9.3263e-03], + ..., + [-2.5681e-01, -2.3070e-01, -1.8887e-01, ..., 1.7559e-02, + -5.9791e-03, 4.6386e-05], + [ 3.3051e-03, 1.5180e-02, 9.0167e-03, ..., 1.9487e-02, + 1.7549e-02, -8.4258e-04], + [ 2.9731e-02, 4.1046e-02, 2.6216e-02, ..., -4.0756e-02, + -5.1178e-02, -3.9193e-02]], device='cuda:0'), grad: tensor([[-9.5129e-04, -6.0892e-04, -5.7459e-04, ..., -7.3254e-05, + -4.3511e-05, -3.4660e-05], + [ 2.2995e-04, 1.4925e-04, 1.3995e-04, ..., 1.7419e-05, + 1.0207e-05, 7.9051e-06], + [ 4.1771e-04, 2.0003e-04, 1.8895e-04, ..., 8.5592e-05, + 6.5327e-05, 5.9038e-05], + ..., + [-4.3422e-05, 4.4525e-05, 3.9905e-05, ..., -6.4969e-05, + -5.6446e-05, -5.4479e-05], + [ 1.8954e-05, 1.4342e-05, 1.2636e-05, ..., -3.4012e-06, + -3.5074e-06, -3.6582e-06], + [ 7.2479e-05, 4.3869e-05, 4.1634e-05, ..., 8.1062e-06, + 5.5060e-06, 4.7162e-06]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0977, -0.0022, -0.0301, -0.0810, -0.0012, 0.0240, 0.0148], + device='cuda:0'), grad: tensor([-1.2197e-03, 2.8682e-04, 8.0919e-04, 3.6120e-04, -3.5715e-04, + 1.6138e-05, 1.0234e-04], device='cuda:0') +588 +0.00876535733001806 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 801.25, cls_loss 0.0035 cls_loss_mapping 0.0301 cls_loss_causal 0.4996 re_mapping 0.0233 re_causal 0.0345 /// teacc 93.22 lr 0.00861397 +Epoch 18, weight, value: tensor([[ 0.3084, 0.2859, 0.2511, ..., -0.0048, -0.0045, -0.0088], + [ 0.0077, -0.0551, 0.0318, ..., -0.0646, -0.0121, -0.0177], + [-0.0786, -0.0478, -0.1040, ..., 0.0308, 0.0038, 0.0087], + ..., + [-0.2536, -0.2282, -0.1873, ..., 0.0177, -0.0052, 0.0006], + [ 0.0015, 0.0141, 0.0080, ..., 0.0189, 0.0171, -0.0008], + [ 0.0288, 0.0393, 0.0249, ..., -0.0401, -0.0502, -0.0385]], + device='cuda:0'), grad: tensor([[-8.0681e-04, -1.2195e-04, -1.2898e-04, ..., -1.3053e-04, + -4.7460e-06, -1.6645e-05], + [ 1.6248e-04, 2.6643e-05, 2.7582e-05, ..., 3.3528e-05, + 9.1642e-06, 1.1556e-05], + [ 5.5361e-04, 8.6367e-05, 8.6010e-05, ..., 1.6189e-04, + 8.4102e-05, 9.1195e-05], + ..., + [-1.3089e-04, -2.9251e-05, -2.3514e-05, ..., -1.1784e-04, + -1.0931e-04, -1.1015e-04], + [ 8.4877e-05, 1.6317e-05, 1.6421e-05, ..., 2.2918e-05, + 1.0960e-05, 1.2271e-05], + [ 9.1136e-05, 1.4931e-05, 1.5408e-05, ..., 1.9282e-05, + 5.6885e-06, 7.0222e-06]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0932, -0.0010, -0.0250, -0.0807, -0.0009, 0.0199, 0.0159], + device='cuda:0'), grad: tensor([-0.0024, 0.0005, 0.0017, 0.0001, -0.0004, 0.0002, 0.0003], + device='cuda:0') +588 +0.008613974319136962 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 797.72, cls_loss 0.0028 cls_loss_mapping 0.0307 cls_loss_causal 0.5040 re_mapping 0.0213 re_causal 0.0336 /// teacc 93.47 lr 0.00845531 +Epoch 19, weight, value: tensor([[ 0.3081, 0.2848, 0.2510, ..., -0.0048, -0.0047, -0.0089], + [ 0.0072, -0.0541, 0.0308, ..., -0.0628, -0.0116, -0.0170], + [-0.0802, -0.0488, -0.1041, ..., 0.0286, 0.0023, 0.0071], + ..., + [-0.2502, -0.2258, -0.1860, ..., 0.0181, -0.0042, 0.0013], + [ 0.0010, 0.0135, 0.0076, ..., 0.0184, 0.0167, -0.0008], + [ 0.0275, 0.0378, 0.0237, ..., -0.0392, -0.0491, -0.0376]], + device='cuda:0'), grad: tensor([[ 2.8849e-04, 4.6074e-05, 5.1081e-05, ..., 1.8489e-04, + 1.5497e-04, 1.4234e-04], + [ 1.0481e-03, 1.0043e-04, 1.2106e-04, ..., 8.1396e-04, + 6.8808e-04, 6.3181e-04], + [-3.1054e-05, -1.2353e-05, -9.3281e-06, ..., 3.5018e-06, + 4.2990e-06, 3.9078e-06], + ..., + [-1.4515e-03, -1.3566e-04, -1.6534e-04, ..., -1.1339e-03, + -9.5844e-04, -8.8024e-04], + [ 1.5306e-04, 2.1562e-05, 2.5362e-05, ..., 1.0747e-04, + 9.0837e-05, 8.3327e-05], + [-5.7817e-05, -3.6657e-05, -4.1723e-05, ..., 4.6641e-06, + 3.9190e-06, 3.8408e-06]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0960, -0.0008, -0.0296, -0.0797, 0.0015, 0.0181, 0.0155], + device='cuda:0'), grad: tensor([ 8.9502e-04, 3.5439e-03, -6.1989e-05, 1.1826e-04, -4.9210e-03, + 4.8828e-04, -6.4075e-05], device='cuda:0') +588 +0.008455313244934327 +changing lr +epoch 18, time 798.76, cls_loss 0.0039 cls_loss_mapping 0.0328 cls_loss_causal 0.4978 re_mapping 0.0189 re_causal 0.0313 /// teacc 92.71 lr 0.00828969 +Epoch 20, weight, value: tensor([[ 3.0665e-01, 2.8359e-01, 2.5095e-01, ..., -6.0339e-03, + -5.9578e-03, -1.0030e-02], + [ 6.9393e-03, -5.2780e-02, 3.0491e-02, ..., -6.1266e-02, + -1.1229e-02, -1.6480e-02], + [-7.9515e-02, -4.9647e-02, -1.0421e-01, ..., 2.7716e-02, + 1.8759e-03, 6.4953e-03], + ..., + [-2.4765e-01, -2.2352e-01, -1.8469e-01, ..., 1.8843e-02, + -2.9624e-03, 2.3516e-03], + [ 6.2641e-05, 1.2833e-02, 7.0505e-03, ..., 1.8054e-02, + 1.6392e-02, -6.4744e-04], + [ 2.6739e-02, 3.6766e-02, 2.2924e-02, ..., -3.8863e-02, + -4.8415e-02, -3.7202e-02]], device='cuda:0'), grad: tensor([[ 5.4300e-05, 2.3454e-05, 2.3410e-05, ..., 1.1541e-05, + 8.7619e-06, 8.7172e-06], + [-4.7326e-05, -1.6063e-05, -2.0429e-05, ..., -4.0978e-06, + -1.6568e-06, -2.0191e-06], + [ 2.6673e-05, 1.2994e-05, 1.3642e-05, ..., 8.0541e-06, + 7.0632e-06, 7.0035e-06], + ..., + [-1.8075e-05, -9.0972e-06, -6.3851e-06, ..., -6.0201e-06, + -4.8541e-06, -4.6007e-06], + [ 2.5127e-06, 9.0385e-07, 1.0878e-06, ..., 3.3341e-07, + 2.0722e-07, 2.1921e-07], + [ 1.1466e-05, 3.9227e-06, 4.9397e-06, ..., 1.1260e-06, + 5.3737e-07, 6.1840e-07]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0950, -0.0014, -0.0262, -0.0786, 0.0015, 0.0157, 0.0145], + device='cuda:0'), grad: tensor([ 1.1820e-04, -1.0955e-04, 4.5121e-05, -4.2945e-05, -4.3094e-05, + 5.6624e-06, 2.6494e-05], device='cuda:0') +588 +0.008289693629698565 +changing lr +epoch 19, time 792.11, cls_loss 0.0039 cls_loss_mapping 0.0237 cls_loss_causal 0.4932 re_mapping 0.0180 re_causal 0.0315 /// teacc 91.46 lr 0.00811745 +Epoch 21, weight, value: tensor([[ 3.0447e-01, 2.8187e-01, 2.4979e-01, ..., -6.7076e-03, + -6.7756e-03, -1.0729e-02], + [ 4.7002e-03, -5.3421e-02, 2.7895e-02, ..., -5.9239e-02, + -1.0466e-02, -1.5559e-02], + [-8.0416e-02, -5.0230e-02, -1.0362e-01, ..., 2.6126e-02, + 9.0521e-04, 5.4391e-03], + ..., + [-2.4427e-01, -2.2061e-01, -1.8261e-01, ..., 1.8997e-02, + -2.1985e-03, 2.8948e-03], + [ 9.3206e-05, 1.2356e-02, 6.7242e-03, ..., 1.8128e-02, + 1.6502e-02, -1.2376e-04], + [ 2.8731e-02, 3.7008e-02, 2.3603e-02, ..., -3.8469e-02, + -4.7741e-02, -3.6795e-02]], device='cuda:0'), grad: tensor([[-6.7532e-05, -3.3796e-05, -3.6597e-05, ..., -1.3702e-05, + -1.1690e-05, -1.1109e-05], + [ 1.8811e-04, 4.0323e-05, 4.5031e-05, ..., 5.3257e-05, + 4.5836e-05, 4.2945e-05], + [-9.0361e-04, -1.1033e-04, -1.1712e-04, ..., -3.1137e-04, + -2.6441e-04, -2.4796e-04], + ..., + [ 6.2656e-04, 9.2924e-05, 9.9301e-05, ..., 2.0146e-04, + 1.6916e-04, 1.5903e-04], + [ 3.8415e-05, 6.9588e-06, 8.0541e-06, ..., 1.2241e-05, + 1.0744e-05, 1.0028e-05], + [-6.0856e-05, -3.1680e-05, -3.9607e-05, ..., 9.1866e-06, + 8.8736e-06, 8.0541e-06]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0943, -0.0015, -0.0295, -0.0791, 0.0009, 0.0159, 0.0189], + device='cuda:0'), grad: tensor([-1.0812e-04, 4.7445e-04, -2.5234e-03, 4.6134e-04, 1.6918e-03, + 1.0103e-04, -9.5367e-05], device='cuda:0') +588 +0.00811744900929367 +changing lr +epoch 20, time 801.93, cls_loss 0.0025 cls_loss_mapping 0.0274 cls_loss_causal 0.4824 re_mapping 0.0165 re_causal 0.0294 /// teacc 92.21 lr 0.00793893 +Epoch 22, weight, value: tensor([[ 3.0332e-01, 2.8126e-01, 2.4988e-01, ..., -7.2212e-03, + -7.1288e-03, -1.1022e-02], + [ 2.2383e-03, -5.3393e-02, 2.6200e-02, ..., -5.8581e-02, + -1.0962e-02, -1.5828e-02], + [-7.8736e-02, -5.1609e-02, -1.0401e-01, ..., 2.6550e-02, + 1.5312e-03, 5.9552e-03], + ..., + [-2.4178e-01, -2.1774e-01, -1.8075e-01, ..., 1.8485e-02, + -2.1285e-03, 2.7723e-03], + [-2.9574e-04, 1.1796e-02, 6.3332e-03, ..., 1.7844e-02, + 1.6253e-02, 4.7915e-06], + [ 2.7793e-02, 3.5903e-02, 2.2670e-02, ..., -3.8213e-02, + -4.7233e-02, -3.6489e-02]], device='cuda:0'), grad: tensor([[-3.6836e-04, -2.5177e-04, -2.4652e-04, ..., -4.3809e-05, + -3.1322e-05, -2.8759e-05], + [-2.4527e-05, 1.1601e-05, 1.1146e-05, ..., -2.8938e-05, + -2.6405e-05, -2.6599e-05], + [ 1.9491e-04, 1.1951e-04, 1.1772e-04, ..., 4.6253e-05, + 3.8534e-05, 3.7432e-05], + ..., + [ 1.2636e-04, 7.4267e-05, 7.3791e-05, ..., 3.0041e-05, + 2.5019e-05, 2.4006e-05], + [ 1.8492e-05, 1.2353e-05, 1.2174e-05, ..., 2.6375e-06, + 2.0005e-06, 1.8878e-06], + [ 8.9884e-05, 4.4286e-05, 4.3482e-05, ..., 2.6584e-05, + 2.2545e-05, 2.2218e-05]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0933, -0.0056, -0.0201, -0.0788, -0.0019, 0.0151, 0.0176], + device='cuda:0'), grad: tensor([-4.7088e-04, -1.3137e-04, 2.8849e-04, -8.9347e-05, 2.0564e-04, + 2.3976e-05, 1.7357e-04], device='cuda:0') +588 +0.007938926261462368 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 804.93, cls_loss 0.0017 cls_loss_mapping 0.0210 cls_loss_causal 0.4604 re_mapping 0.0152 re_causal 0.0278 /// teacc 93.97 lr 0.00775448 +Epoch 23, weight, value: tensor([[ 3.0241e-01, 2.8070e-01, 2.5009e-01, ..., -7.2762e-03, + -7.3297e-03, -1.1042e-02], + [ 3.3405e-03, -5.2328e-02, 2.5679e-02, ..., -5.6922e-02, + -1.0260e-02, -1.5070e-02], + [-8.0902e-02, -5.3023e-02, -1.0454e-01, ..., 2.4919e-02, + 4.5360e-04, 4.8131e-03], + ..., + [-2.3769e-01, -2.1498e-01, -1.7900e-01, ..., 1.9241e-02, + -9.9494e-04, 3.6478e-03], + [-6.8180e-04, 1.1238e-02, 5.9412e-03, ..., 1.7508e-02, + 1.5926e-02, 6.2798e-05], + [ 2.5671e-02, 3.4222e-02, 2.1278e-02, ..., -3.8128e-02, + -4.6865e-02, -3.6323e-02]], device='cuda:0'), grad: tensor([[-9.3460e-05, -9.2089e-05, -1.0115e-04, ..., 2.8029e-05, + 2.2590e-05, 2.4304e-05], + [-5.7906e-05, -3.4869e-05, -3.4958e-05, ..., 1.2800e-05, + 1.2353e-05, 1.2778e-05], + [ 3.9995e-05, 5.4002e-05, 5.8174e-05, ..., -7.7263e-06, + -3.3919e-06, -3.3993e-06], + ..., + [ 2.7910e-05, 1.4126e-05, 1.5378e-05, ..., 7.2680e-06, + 6.9104e-06, 6.5789e-06], + [ 3.9577e-05, 2.2858e-05, 2.4661e-05, ..., 3.2112e-06, + 3.1590e-06, 2.6021e-06], + [ 9.1255e-05, 5.1975e-05, 5.5701e-05, ..., 6.2399e-06, + 6.0759e-06, 4.8578e-06]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0929, -0.0009, -0.0245, -0.0785, 0.0014, 0.0141, 0.0145], + device='cuda:0'), grad: tensor([-2.0582e-07, -9.3699e-05, -4.3452e-05, -1.0294e-04, 4.5955e-05, + 5.7280e-05, 1.3685e-04], device='cuda:0') +588 +0.007754484907260515 +changing lr +epoch 22, time 804.92, cls_loss 0.0012 cls_loss_mapping 0.0169 cls_loss_causal 0.4926 re_mapping 0.0135 re_causal 0.0270 /// teacc 92.46 lr 0.00756450 +Epoch 24, weight, value: tensor([[ 3.0341e-01, 2.8073e-01, 2.5079e-01, ..., -7.3566e-03, + -7.5387e-03, -1.1151e-02], + [ 2.2362e-03, -5.2175e-02, 2.4299e-02, ..., -5.5901e-02, + -1.0195e-02, -1.4882e-02], + [-8.1328e-02, -5.4016e-02, -1.0463e-01, ..., 2.3827e-02, + -9.1272e-05, 4.1168e-03], + ..., + [-2.3558e-01, -2.1269e-01, -1.7767e-01, ..., 1.9353e-02, + -4.5102e-04, 4.0575e-03], + [-9.7764e-04, 1.0551e-02, 5.4267e-03, ..., 1.7267e-02, + 1.5704e-02, 1.6631e-04], + [ 2.3303e-02, 3.2839e-02, 2.0170e-02, ..., -3.8064e-02, + -4.6516e-02, -3.6156e-02]], device='cuda:0'), grad: tensor([[-3.5238e-04, -2.3055e-04, -2.3615e-04, ..., -2.4348e-05, + -2.0236e-05, -1.9029e-05], + [ 1.1915e-04, 6.5088e-05, 6.7651e-05, ..., 1.9163e-05, + 1.7196e-05, 1.4812e-05], + [ 2.9311e-05, 3.6269e-05, 3.9190e-05, ..., 4.6603e-06, + 7.0035e-06, 6.3032e-06], + ..., + [-1.1533e-05, 2.5406e-05, 2.4766e-05, ..., -3.7104e-05, + -3.6001e-05, -2.9743e-05], + [ 5.5015e-05, 2.6435e-05, 2.7090e-05, ..., 1.2539e-05, + 1.1414e-05, 9.7975e-06], + [ 5.5641e-05, 2.6867e-05, 2.7537e-05, ..., 1.1757e-05, + 1.0565e-05, 9.0525e-06]], device='cuda:0') +Epoch 24, bias, value: tensor([ 9.7089e-02, -1.6464e-03, -2.4646e-02, -7.6325e-02, -9.4057e-05, + 1.4250e-02, 9.9926e-03], device='cuda:0'), grad: tensor([-4.3774e-04, 1.9503e-04, -5.9791e-06, 1.9085e-04, -1.5306e-04, + 1.0592e-04, 1.0550e-04], device='cuda:0') +588 +0.007564496387029534 +changing lr +epoch 23, time 800.65, cls_loss 0.0010 cls_loss_mapping 0.0178 cls_loss_causal 0.4827 re_mapping 0.0127 re_causal 0.0264 /// teacc 92.96 lr 0.00736934 +Epoch 25, weight, value: tensor([[ 2.9767e-01, 2.7746e-01, 2.4832e-01, ..., -8.7068e-03, + -8.7573e-03, -1.2214e-02], + [ 3.5369e-03, -5.1034e-02, 2.4146e-02, ..., -5.4302e-02, + -9.6105e-03, -1.4181e-02], + [-8.0467e-02, -5.4146e-02, -1.0408e-01, ..., 2.3166e-02, + -4.0458e-04, 3.7015e-03], + ..., + [-2.3154e-01, -2.0966e-01, -1.7562e-01, ..., 1.9981e-02, + 6.3002e-04, 4.9106e-03], + [-1.7386e-03, 9.8716e-03, 4.8619e-03, ..., 1.6948e-02, + 1.5422e-02, 2.2180e-04], + [ 2.3662e-02, 3.2522e-02, 2.0136e-02, ..., -3.7595e-02, + -4.5840e-02, -3.5701e-02]], device='cuda:0'), grad: tensor([[ 2.1793e-06, -9.1791e-06, -1.0736e-05, ..., 6.3106e-06, + 3.8967e-06, 4.2617e-06], + [-6.9022e-05, -2.4393e-05, -2.2501e-05, ..., -1.6659e-05, + -1.2137e-05, -1.2122e-05], + [ 2.4751e-05, 1.1161e-05, 1.1109e-05, ..., 5.2005e-06, + 4.2506e-06, 4.1351e-06], + ..., + [ 1.8418e-05, 1.0662e-05, 1.0833e-05, ..., 2.2370e-06, + 1.9036e-06, 1.7798e-06], + [ 1.1772e-05, 6.7949e-06, 6.8247e-06, ..., 1.3644e-06, + 1.1018e-06, 1.0049e-06], + [ 1.3508e-05, 5.8524e-06, 5.6326e-06, ..., 2.6282e-06, + 1.9539e-06, 1.9129e-06]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0893, 0.0024, -0.0232, -0.0765, 0.0024, 0.0128, 0.0112], + device='cuda:0'), grad: tensor([ 4.2707e-05, -1.6236e-04, 4.9114e-05, -1.7397e-06, 2.7105e-05, + 1.7464e-05, 2.7597e-05], device='cuda:0') +588 +0.007369343312364995 +changing lr +epoch 24, time 797.55, cls_loss 0.0014 cls_loss_mapping 0.0192 cls_loss_causal 0.4524 re_mapping 0.0122 re_causal 0.0246 /// teacc 93.47 lr 0.00716942 +Epoch 26, weight, value: tensor([[ 0.3000, 0.2791, 0.2508, ..., -0.0088, -0.0089, -0.0123], + [ 0.0006, -0.0522, 0.0216, ..., -0.0535, -0.0096, -0.0140], + [-0.0822, -0.0562, -0.1054, ..., 0.0223, -0.0009, 0.0030], + ..., + [-0.2290, -0.2075, -0.1744, ..., 0.0201, 0.0011, 0.0053], + [-0.0018, 0.0097, 0.0049, ..., 0.0168, 0.0153, 0.0004], + [ 0.0232, 0.0319, 0.0197, ..., -0.0371, -0.0452, -0.0352]], + device='cuda:0'), grad: tensor([[ 6.5088e-04, 1.9658e-04, 2.0254e-04, ..., 2.2066e-04, + 1.8895e-04, 1.9455e-04], + [-1.8561e-04, -3.1143e-05, -4.0710e-05, ..., -2.4721e-05, + -1.6004e-05, -1.5855e-05], + [ 4.2081e-04, 1.2130e-04, 1.2314e-04, ..., 1.5497e-04, + 1.3554e-04, 1.3888e-04], + ..., + [-9.8705e-04, -2.8467e-04, -2.8110e-04, ..., -3.6788e-04, + -3.1900e-04, -3.2997e-04], + [ 1.2672e-04, 3.7640e-05, 3.7223e-05, ..., 4.6343e-05, + 3.9935e-05, 4.1425e-05], + [ 7.8797e-05, 1.8910e-05, 2.0757e-05, ..., 2.0325e-05, + 1.6555e-05, 1.6943e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0930, -0.0002, -0.0241, -0.0761, 0.0029, 0.0117, 0.0107], + device='cuda:0'), grad: tensor([ 0.0016, -0.0005, 0.0011, -0.0001, -0.0025, 0.0003, 0.0002], + device='cuda:0') +588 +0.0071694186955877925 +changing lr +epoch 25, time 798.32, cls_loss 0.0014 cls_loss_mapping 0.0173 cls_loss_causal 0.4765 re_mapping 0.0114 re_causal 0.0249 /// teacc 92.46 lr 0.00696513 +Epoch 27, weight, value: tensor([[ 0.2992, 0.2786, 0.2510, ..., -0.0096, -0.0097, -0.0130], + [ 0.0007, -0.0513, 0.0211, ..., -0.0522, -0.0092, -0.0135], + [-0.0827, -0.0573, -0.1057, ..., 0.0216, -0.0013, 0.0026], + ..., + [-0.2271, -0.2057, -0.1735, ..., 0.0200, 0.0014, 0.0054], + [-0.0018, 0.0097, 0.0051, ..., 0.0165, 0.0152, 0.0005], + [ 0.0222, 0.0305, 0.0185, ..., -0.0367, -0.0445, -0.0347]], + device='cuda:0'), grad: tensor([[-3.2234e-04, -2.1136e-04, -2.5582e-04, ..., -3.6538e-05, + -3.8832e-05, -3.0115e-05], + [ 6.3753e-04, 1.5867e-04, 2.0230e-04, ..., 1.2553e-04, + 1.0616e-04, 9.0301e-05], + [-1.1482e-03, -7.7665e-05, -1.2553e-04, ..., -2.6131e-04, + -2.0933e-04, -1.8072e-04], + ..., + [ 4.3869e-04, 4.6045e-05, 7.3969e-05, ..., 8.6069e-05, + 7.0870e-05, 5.7817e-05], + [ 6.4552e-05, 1.4663e-05, 1.8716e-05, ..., 1.3188e-05, + 1.1049e-05, 9.4995e-06], + [ 1.0896e-04, 1.7837e-05, 2.3589e-05, ..., 2.3484e-05, + 1.9148e-05, 1.6630e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0918, 0.0010, -0.0236, -0.0753, 0.0025, 0.0101, 0.0109], + device='cuda:0'), grad: tensor([-0.0004, 0.0016, -0.0036, 0.0006, 0.0013, 0.0002, 0.0003], + device='cuda:0') +588 +0.0069651251582696205 +changing lr +epoch 26, time 797.17, cls_loss 0.0013 cls_loss_mapping 0.0172 cls_loss_causal 0.4458 re_mapping 0.0104 re_causal 0.0235 /// teacc 93.47 lr 0.00675687 +Epoch 28, weight, value: tensor([[ 0.2990, 0.2783, 0.2510, ..., -0.0100, -0.0103, -0.0135], + [ 0.0011, -0.0507, 0.0206, ..., -0.0510, -0.0088, -0.0131], + [-0.0834, -0.0582, -0.1057, ..., 0.0205, -0.0020, 0.0019], + ..., + [-0.2254, -0.2040, -0.1726, ..., 0.0204, 0.0022, 0.0060], + [-0.0025, 0.0090, 0.0044, ..., 0.0163, 0.0150, 0.0006], + [ 0.0219, 0.0300, 0.0182, ..., -0.0365, -0.0440, -0.0344]], + device='cuda:0'), grad: tensor([[ 3.2735e-04, 1.5974e-04, 1.6665e-04, ..., 6.4611e-05, + 5.8174e-05, 5.8711e-05], + [ 2.7388e-05, 1.0744e-05, 1.0781e-05, ..., 6.4336e-06, + 5.6624e-06, 5.6103e-06], + [-1.7011e-04, -4.9680e-05, -4.5806e-05, ..., -1.3113e-05, + -1.2346e-05, -1.2852e-05], + ..., + [-3.4630e-05, -1.1139e-06, -1.0550e-08, ..., -2.4438e-05, + -2.1309e-05, -1.9833e-05], + [ 3.0413e-05, 9.4846e-06, 9.1419e-06, ..., 5.6848e-06, + 5.1446e-06, 5.0291e-06], + [ 2.1249e-05, 5.6885e-06, 5.2862e-06, ..., 5.7481e-06, + 5.0664e-06, 4.8839e-06]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0930, 0.0031, -0.0248, -0.0758, 0.0018, 0.0096, 0.0101], + device='cuda:0'), grad: tensor([ 5.7983e-04, 5.8293e-05, -4.2319e-04, -2.2304e-04, -1.2046e-04, + 7.3552e-05, 5.5075e-05], device='cuda:0') +588 +0.006756874120406716 +changing lr +epoch 27, time 794.87, cls_loss 0.0010 cls_loss_mapping 0.0156 cls_loss_causal 0.4354 re_mapping 0.0101 re_causal 0.0227 /// teacc 91.96 lr 0.00654508 +Epoch 29, weight, value: tensor([[ 0.2964, 0.2768, 0.2501, ..., -0.0106, -0.0110, -0.0140], + [-0.0003, -0.0508, 0.0192, ..., -0.0499, -0.0086, -0.0127], + [-0.0830, -0.0584, -0.1052, ..., 0.0195, -0.0025, 0.0013], + ..., + [-0.2227, -0.2019, -0.1714, ..., 0.0206, 0.0028, 0.0064], + [-0.0022, 0.0085, 0.0041, ..., 0.0162, 0.0149, 0.0008], + [ 0.0216, 0.0297, 0.0181, ..., -0.0362, -0.0435, -0.0341]], + device='cuda:0'), grad: tensor([[ 2.1100e-05, 2.8107e-06, 4.6752e-06, ..., 9.8497e-06, + 8.6352e-06, 7.9274e-06], + [ 2.0242e-04, 2.1428e-05, 2.7567e-05, ..., 7.5579e-05, + 6.1452e-05, 5.5641e-05], + [-6.2370e-04, -8.9765e-05, -1.2732e-04, ..., -2.3520e-04, + -1.9300e-04, -1.7798e-04], + ..., + [ 1.2058e-04, 6.2808e-06, 4.7833e-06, ..., 3.9488e-05, + 2.9057e-05, 2.6211e-05], + [ 1.7896e-05, 3.2187e-06, 4.3213e-06, ..., 6.6496e-06, + 5.6550e-06, 5.1633e-06], + [-1.3197e-06, -5.2676e-06, -4.2990e-06, ..., 3.9190e-06, + 3.0492e-06, 2.8163e-06]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0892, 0.0022, -0.0241, -0.0743, 0.0036, 0.0108, 0.0095], + device='cuda:0'), grad: tensor([ 7.8917e-05, 6.7806e-04, -2.1114e-03, 8.7309e-04, 4.0364e-04, + 5.7459e-05, 1.9580e-05], device='cuda:0') +588 +0.00654508497187474 +changing lr +epoch 28, time 805.55, cls_loss 0.0012 cls_loss_mapping 0.0149 cls_loss_causal 0.4483 re_mapping 0.0097 re_causal 0.0227 /// teacc 92.21 lr 0.00633018 +Epoch 30, weight, value: tensor([[ 0.2958, 0.2760, 0.2501, ..., -0.0108, -0.0112, -0.0142], + [-0.0015, -0.0508, 0.0179, ..., -0.0491, -0.0085, -0.0125], + [-0.0831, -0.0591, -0.1053, ..., 0.0189, -0.0029, 0.0009], + ..., + [-0.2210, -0.2002, -0.1704, ..., 0.0205, 0.0031, 0.0066], + [-0.0030, 0.0081, 0.0037, ..., 0.0159, 0.0146, 0.0007], + [ 0.0227, 0.0298, 0.0185, ..., -0.0358, -0.0429, -0.0337]], + device='cuda:0'), grad: tensor([[-1.0805e-03, -3.7551e-04, -4.2343e-04, ..., -1.4603e-04, + -8.7082e-05, -1.1414e-04], + [ 5.3835e-04, 1.9062e-04, 2.1446e-04, ..., 8.3685e-05, + 5.4508e-05, 6.8009e-05], + [ 3.0708e-04, 9.1732e-05, 9.2924e-05, ..., 7.2002e-05, + 5.4151e-05, 6.0499e-05], + ..., + [ 1.6141e-04, 7.1049e-05, 8.7857e-05, ..., 2.6733e-05, + 2.1249e-05, 2.4289e-05], + [ 5.6893e-05, 2.1517e-05, 2.3648e-05, ..., 1.0692e-05, + 8.1211e-06, 9.1568e-06], + [ 7.7963e-05, 2.8268e-05, 3.1024e-05, ..., 1.6436e-05, + 1.2979e-05, 1.4275e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0890, 0.0014, -0.0226, -0.0744, 0.0027, 0.0087, 0.0116], + device='cuda:0'), grad: tensor([-0.0024, 0.0012, 0.0008, -0.0001, 0.0003, 0.0001, 0.0002], + device='cuda:0') +588 +0.006330184227833378 +changing lr +epoch 29, time 800.78, cls_loss 0.0022 cls_loss_mapping 0.0162 cls_loss_causal 0.4417 re_mapping 0.0096 re_causal 0.0220 /// teacc 93.72 lr 0.00611260 +Epoch 31, weight, value: tensor([[ 2.9643e-01, 2.7640e-01, 2.5087e-01, ..., -1.1205e-02, + -1.1681e-02, -1.4603e-02], + [-2.0792e-03, -5.0847e-02, 1.6912e-02, ..., -4.7939e-02, + -8.1598e-03, -1.2038e-02], + [-8.3404e-02, -6.0154e-02, -1.0570e-01, ..., 1.8019e-02, + -3.3490e-03, 2.8479e-04], + ..., + [-2.1862e-01, -1.9835e-01, -1.6911e-01, ..., 2.0958e-02, + 3.8500e-03, 7.2092e-03], + [-4.0201e-03, 7.2901e-03, 2.9537e-03, ..., 1.5558e-02, + 1.4312e-02, 7.1897e-04], + [ 2.1437e-02, 2.9160e-02, 1.7969e-02, ..., -3.5824e-02, + -4.2671e-02, -3.3568e-02]], device='cuda:0'), grad: tensor([[-9.1457e-04, -5.7173e-04, -5.7650e-04, ..., -8.1003e-05, + -7.4625e-05, -7.1347e-05], + [ 6.6853e-04, 3.9935e-04, 4.0102e-04, ..., 6.8128e-05, + 6.2287e-05, 5.8621e-05], + [ 9.3758e-05, 7.7426e-05, 7.6532e-05, ..., 2.7195e-06, + 4.0010e-06, 4.9435e-06], + ..., + [ 6.8128e-05, 2.6152e-05, 2.8029e-05, ..., 2.3112e-05, + 2.1592e-05, 1.9968e-05], + [-5.3465e-05, -2.6207e-06, -4.3362e-06, ..., -3.8505e-05, + -3.7253e-05, -3.4034e-05], + [ 9.2506e-05, 4.5151e-05, 4.7624e-05, ..., 2.2009e-05, + 2.0757e-05, 1.9178e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0902, 0.0026, -0.0210, -0.0748, 0.0033, 0.0070, 0.0088], + device='cuda:0'), grad: tensor([-1.2283e-03, 9.7322e-04, 5.2720e-05, 6.4433e-05, 1.4603e-04, + -1.7130e-04, 1.6260e-04], device='cuda:0') +588 +0.006112604669781575 +changing lr +epoch 30, time 803.32, cls_loss 0.0010 cls_loss_mapping 0.0153 cls_loss_causal 0.4200 re_mapping 0.0093 re_causal 0.0218 /// teacc 93.72 lr 0.00589278 +Epoch 32, weight, value: tensor([[ 0.2955, 0.2757, 0.2508, ..., -0.0117, -0.0121, -0.0150], + [-0.0038, -0.0513, 0.0154, ..., -0.0471, -0.0080, -0.0118], + [-0.0839, -0.0608, -0.1056, ..., 0.0171, -0.0039, -0.0003], + ..., + [-0.2166, -0.1966, -0.1682, ..., 0.0209, 0.0041, 0.0074], + [-0.0041, 0.0070, 0.0027, ..., 0.0154, 0.0142, 0.0008], + [ 0.0220, 0.0289, 0.0178, ..., -0.0354, -0.0421, -0.0331]], + device='cuda:0'), grad: tensor([[ 1.1473e-03, 4.5085e-04, 5.0259e-04, ..., 6.2656e-04, + 6.0129e-04, 5.9652e-04], + [ 2.4128e-04, 9.1970e-05, 1.0252e-04, ..., 1.2875e-04, + 1.2362e-04, 1.2255e-04], + [ 9.9838e-06, 1.7092e-05, 1.9282e-05, ..., 1.9982e-05, + 1.8999e-05, 1.9163e-05], + ..., + [ 9.0408e-04, 3.5238e-04, 3.9291e-04, ..., 4.9210e-04, + 4.7183e-04, 4.6825e-04], + [-8.0824e-05, -3.1620e-05, -2.5362e-05, ..., 2.4382e-06, + 1.2638e-06, 1.6280e-06], + [ 1.1528e-04, 4.4107e-05, 4.3482e-05, ..., 3.4809e-05, + 3.4034e-05, 3.3557e-05]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0894, 0.0011, -0.0218, -0.0742, 0.0038, 0.0067, 0.0108], + device='cuda:0'), grad: tensor([ 2.1744e-03, 4.6849e-04, -3.5048e-05, -4.3869e-03, 1.7233e-03, + -1.9062e-04, 2.4414e-04], device='cuda:0') +588 +0.005892784473993186 +changing lr +epoch 31, time 807.94, cls_loss 0.0011 cls_loss_mapping 0.0141 cls_loss_causal 0.4438 re_mapping 0.0094 re_causal 0.0225 /// teacc 93.47 lr 0.00567117 +Epoch 33, weight, value: tensor([[ 0.2931, 0.2734, 0.2494, ..., -0.0120, -0.0124, -0.0153], + [-0.0035, -0.0507, 0.0149, ..., -0.0460, -0.0076, -0.0113], + [-0.0824, -0.0604, -0.1047, ..., 0.0165, -0.0042, -0.0007], + ..., + [-0.2136, -0.1940, -0.1665, ..., 0.0211, 0.0045, 0.0077], + [-0.0045, 0.0066, 0.0024, ..., 0.0151, 0.0140, 0.0008], + [ 0.0203, 0.0279, 0.0169, ..., -0.0353, -0.0418, -0.0330]], + device='cuda:0'), grad: tensor([[ 1.2314e-04, 5.6297e-05, 6.4850e-05, ..., 8.8885e-06, + 1.5147e-05, 1.2144e-05], + [-2.0421e-04, -6.9261e-05, -1.1009e-04, ..., -4.1753e-05, + -3.6001e-05, -3.6359e-05], + [-1.9407e-04, -1.0091e-04, -1.0794e-04, ..., -1.3933e-05, + -2.2918e-05, -1.8120e-05], + ..., + [ 2.7269e-05, 1.2420e-05, 1.5542e-05, ..., 4.0792e-06, + 3.9861e-06, 3.7514e-06], + [ 5.0187e-05, 1.8343e-05, 2.7403e-05, ..., 9.7379e-06, + 8.6203e-06, 8.5980e-06], + [ 9.4652e-05, 4.1217e-05, 5.5104e-05, ..., 1.5408e-05, + 1.3940e-05, 1.3582e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0883, 0.0033, -0.0191, -0.0753, 0.0045, 0.0059, 0.0079], + device='cuda:0'), grad: tensor([ 2.3878e-04, -3.5334e-04, -3.5167e-04, 1.8370e-04, 4.4793e-05, + 8.6069e-05, 1.5199e-04], device='cuda:0') +588 +0.00567116632908828 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 807.15, cls_loss 0.0010 cls_loss_mapping 0.0137 cls_loss_causal 0.4254 re_mapping 0.0090 re_causal 0.0209 /// teacc 94.47 lr 0.00544820 +Epoch 34, weight, value: tensor([[ 0.2921, 0.2724, 0.2489, ..., -0.0123, -0.0128, -0.0156], + [-0.0043, -0.0503, 0.0143, ..., -0.0453, -0.0075, -0.0111], + [-0.0827, -0.0608, -0.1045, ..., 0.0159, -0.0044, -0.0010], + ..., + [-0.2125, -0.1929, -0.1659, ..., 0.0206, 0.0043, 0.0074], + [-0.0045, 0.0063, 0.0022, ..., 0.0151, 0.0139, 0.0010], + [ 0.0197, 0.0272, 0.0164, ..., -0.0350, -0.0414, -0.0327]], + device='cuda:0'), grad: tensor([[ 8.4925e-04, 2.9707e-04, 3.1471e-04, ..., 2.5940e-04, + 2.5916e-04, 2.5082e-04], + [-7.5569e-03, -2.6493e-03, -2.8076e-03, ..., -2.3994e-03, + -2.3823e-03, -2.3232e-03], + [ 3.5629e-03, 1.2598e-03, 1.3294e-03, ..., 1.1501e-03, + 1.1454e-03, 1.1187e-03], + ..., + [ 5.9414e-04, 2.0432e-04, 2.1875e-04, ..., 1.9705e-04, + 1.9121e-04, 1.8919e-04], + [ 1.5664e-04, 5.3883e-05, 5.7608e-05, ..., 4.7475e-05, + 4.6819e-05, 4.5508e-05], + [ 7.3814e-04, 2.4331e-04, 2.6560e-04, ..., 2.0218e-04, + 1.9646e-04, 1.8907e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0883, 0.0018, -0.0194, -0.0726, 0.0036, 0.0059, 0.0076], + device='cuda:0'), grad: tensor([ 0.0019, -0.0166, 0.0077, 0.0035, 0.0013, 0.0004, 0.0017], + device='cuda:0') +588 +0.00544819654451717 +changing lr +epoch 33, time 801.32, cls_loss 0.0009 cls_loss_mapping 0.0120 cls_loss_causal 0.4182 re_mapping 0.0085 re_causal 0.0208 /// teacc 92.21 lr 0.00522432 +Epoch 35, weight, value: tensor([[ 0.2903, 0.2713, 0.2482, ..., -0.0129, -0.0133, -0.0160], + [-0.0034, -0.0494, 0.0143, ..., -0.0442, -0.0070, -0.0106], + [-0.0834, -0.0615, -0.1046, ..., 0.0150, -0.0050, -0.0017], + ..., + [-0.2097, -0.1909, -0.1645, ..., 0.0209, 0.0048, 0.0079], + [-0.0047, 0.0058, 0.0018, ..., 0.0149, 0.0138, 0.0011], + [ 0.0188, 0.0265, 0.0158, ..., -0.0349, -0.0410, -0.0325]], + device='cuda:0'), grad: tensor([[-1.6623e-03, -9.4938e-04, -1.0691e-03, ..., -2.4772e-04, + -2.3317e-04, -2.4343e-04], + [ 2.8062e-04, 1.6332e-04, 1.5497e-04, ..., 2.0042e-05, + 1.9893e-05, 2.1189e-05], + [ 1.2183e-04, 6.8367e-05, 7.3433e-05, ..., 1.7747e-05, + 1.6943e-05, 1.6987e-05], + ..., + [ 6.3324e-04, 3.6430e-04, 4.3082e-04, ..., 1.0502e-04, + 9.7692e-05, 1.0312e-04], + [ 1.0037e-04, 5.7071e-05, 6.6102e-05, ..., 1.6525e-05, + 1.5527e-05, 1.6168e-05], + [ 2.9516e-04, 1.6642e-04, 1.9372e-04, ..., 4.9919e-05, + 4.6909e-05, 4.8518e-05]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0854, 0.0042, -0.0203, -0.0732, 0.0063, 0.0067, 0.0059], + device='cuda:0'), grad: tensor([-0.0023, 0.0004, 0.0002, 0.0003, 0.0009, 0.0001, 0.0004], + device='cuda:0') +588 +0.005224324151752577 +changing lr +epoch 34, time 795.20, cls_loss 0.0010 cls_loss_mapping 0.0118 cls_loss_causal 0.4274 re_mapping 0.0082 re_causal 0.0200 /// teacc 91.71 lr 0.00500000 +Epoch 36, weight, value: tensor([[ 0.2902, 0.2709, 0.2482, ..., -0.0132, -0.0136, -0.0163], + [-0.0040, -0.0492, 0.0136, ..., -0.0437, -0.0070, -0.0105], + [-0.0830, -0.0617, -0.1043, ..., 0.0146, -0.0053, -0.0020], + ..., + [-0.2083, -0.1896, -0.1639, ..., 0.0210, 0.0052, 0.0081], + [-0.0055, 0.0054, 0.0014, ..., 0.0146, 0.0135, 0.0009], + [ 0.0182, 0.0259, 0.0154, ..., -0.0346, -0.0407, -0.0322]], + device='cuda:0'), grad: tensor([[-8.0287e-05, -5.7131e-05, -5.0098e-05, ..., -1.9763e-06, + 5.5617e-08, -1.5134e-07], + [ 3.2574e-05, 1.4395e-05, 1.4730e-05, ..., 1.5602e-05, + 1.4566e-05, 1.4544e-05], + [ 8.7798e-05, 3.5137e-05, 3.6985e-05, ..., 5.2035e-05, + 4.9829e-05, 4.9561e-05], + ..., + [ 1.1677e-04, 6.7174e-05, 6.2585e-05, ..., 2.9519e-05, + 2.5943e-05, 2.6077e-05], + [ 1.1258e-05, 5.0142e-06, 5.1297e-06, ..., 5.4352e-06, + 5.0813e-06, 5.0738e-06], + [ 1.5706e-05, 7.3649e-06, 7.4171e-06, ..., 6.7689e-06, + 6.3069e-06, 6.2957e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0863, 0.0040, -0.0189, -0.0729, 0.0064, 0.0047, 0.0053], + device='cuda:0'), grad: tensor([-1.0407e-04, 6.8367e-05, 1.9825e-04, -4.1604e-04, 1.9872e-04, + 2.3410e-05, 3.1650e-05], device='cuda:0') +588 +0.005000000000000003 +changing lr +epoch 35, time 792.13, cls_loss 0.0008 cls_loss_mapping 0.0095 cls_loss_causal 0.4105 re_mapping 0.0080 re_causal 0.0194 /// teacc 93.47 lr 0.00477568 +Epoch 37, weight, value: tensor([[ 0.2894, 0.2700, 0.2477, ..., -0.0135, -0.0139, -0.0165], + [-0.0049, -0.0490, 0.0131, ..., -0.0433, -0.0071, -0.0105], + [-0.0823, -0.0618, -0.1039, ..., 0.0142, -0.0054, -0.0022], + ..., + [-0.2075, -0.1884, -0.1631, ..., 0.0210, 0.0053, 0.0082], + [-0.0042, 0.0054, 0.0015, ..., 0.0149, 0.0138, 0.0014], + [ 0.0179, 0.0256, 0.0152, ..., -0.0344, -0.0404, -0.0321]], + device='cuda:0'), grad: tensor([[-1.3101e-04, -1.0580e-04, -1.0169e-04, ..., 1.4357e-05, + 1.5572e-05, 1.5512e-05], + [-3.6263e-04, -3.8356e-05, -4.4197e-05, ..., -1.2255e-04, + -1.2350e-04, -1.1641e-04], + [ 2.5916e-04, 1.0717e-04, 1.1051e-04, ..., 6.8307e-05, + 6.6459e-05, 6.4075e-05], + ..., + [ 2.1553e-04, 4.8310e-05, 5.2541e-05, ..., 7.7128e-05, + 7.5579e-05, 7.2181e-05], + [ 6.5923e-05, 2.0012e-05, 2.0698e-05, ..., 1.7375e-05, + 1.7196e-05, 1.6212e-05], + [ 8.0645e-05, 2.7642e-05, 2.8148e-05, ..., 1.8388e-05, + 1.8135e-05, 1.7047e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0864, 0.0019, -0.0167, -0.0745, 0.0044, 0.0081, 0.0049], + device='cuda:0'), grad: tensor([-8.1778e-05, -1.1969e-03, 5.4407e-04, -2.3472e-04, 6.0987e-04, + 1.6737e-04, 1.9264e-04], device='cuda:0') +588 +0.004775675848247429 +changing lr +epoch 36, time 798.79, cls_loss 0.0009 cls_loss_mapping 0.0122 cls_loss_causal 0.4150 re_mapping 0.0079 re_causal 0.0202 /// teacc 93.22 lr 0.00455180 +Epoch 38, weight, value: tensor([[ 0.2885, 0.2691, 0.2472, ..., -0.0137, -0.0140, -0.0167], + [-0.0042, -0.0482, 0.0131, ..., -0.0424, -0.0068, -0.0101], + [-0.0824, -0.0621, -0.1037, ..., 0.0136, -0.0057, -0.0026], + ..., + [-0.2068, -0.1874, -0.1625, ..., 0.0208, 0.0054, 0.0083], + [-0.0045, 0.0052, 0.0014, ..., 0.0148, 0.0137, 0.0015], + [ 0.0180, 0.0252, 0.0150, ..., -0.0341, -0.0400, -0.0318]], + device='cuda:0'), grad: tensor([[-1.1810e-02, -5.7907e-03, -6.4354e-03, ..., -1.9150e-03, + -1.8396e-03, -1.5230e-03], + [ 2.3289e-03, 1.0719e-03, 1.2245e-03, ..., 4.2105e-04, + 4.0293e-04, 3.3545e-04], + [ 2.2926e-03, 1.2970e-03, 1.3771e-03, ..., 2.3854e-04, + 2.2101e-04, 1.8394e-04], + ..., + [ 5.0240e-03, 2.3766e-03, 2.6627e-03, ..., 8.9931e-04, + 8.7404e-04, 7.1955e-04], + [ 3.4261e-04, 1.6701e-04, 1.8680e-04, ..., 5.4419e-05, + 5.1767e-05, 4.3064e-05], + [ 3.2520e-04, 1.6987e-04, 1.8561e-04, ..., 4.1962e-05, + 3.9637e-05, 3.3170e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0866, 0.0040, -0.0170, -0.0750, 0.0025, 0.0075, 0.0058], + device='cuda:0'), grad: tensor([-0.0224, 0.0047, 0.0035, 0.0030, 0.0101, 0.0006, 0.0005], + device='cuda:0') +588 +0.004551803455482836 +changing lr +epoch 37, time 800.57, cls_loss 0.0010 cls_loss_mapping 0.0106 cls_loss_causal 0.4152 re_mapping 0.0077 re_causal 0.0193 /// teacc 92.96 lr 0.00432883 +Epoch 39, weight, value: tensor([[ 0.2868, 0.2679, 0.2464, ..., -0.0142, -0.0145, -0.0171], + [-0.0044, -0.0479, 0.0127, ..., -0.0419, -0.0067, -0.0100], + [-0.0824, -0.0624, -0.1035, ..., 0.0131, -0.0060, -0.0029], + ..., + [-0.2046, -0.1860, -0.1616, ..., 0.0213, 0.0060, 0.0088], + [-0.0057, 0.0046, 0.0008, ..., 0.0144, 0.0133, 0.0013], + [ 0.0185, 0.0251, 0.0150, ..., -0.0338, -0.0395, -0.0314]], + device='cuda:0'), grad: tensor([[-3.4738e-04, -2.0671e-04, -2.4319e-04, ..., -1.1101e-05, + -1.1504e-05, -6.8583e-06], + [ 9.3997e-05, 4.2319e-05, 4.9889e-05, ..., 1.4775e-05, + 1.5303e-05, 1.3471e-05], + [-2.8521e-05, 1.8358e-05, 2.7299e-05, ..., -2.5079e-05, + -2.0444e-05, -2.2128e-05], + ..., + [ 5.8673e-06, 1.6987e-05, 1.8612e-05, ..., -1.2435e-05, + -1.3791e-05, -1.2688e-05], + [ 1.1146e-04, 6.3837e-05, 7.5161e-05, ..., 5.3756e-06, + 5.3532e-06, 3.8929e-06], + [ 8.1778e-05, 3.8028e-05, 4.3213e-05, ..., 1.0148e-05, + 9.0450e-06, 8.4341e-06]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0844, 0.0040, -0.0171, -0.0745, 0.0052, 0.0047, 0.0075], + device='cuda:0'), grad: tensor([-3.8004e-04, 1.6165e-04, -1.6963e-04, 1.7810e-04, -5.3316e-05, + 1.3161e-04, 1.3173e-04], device='cuda:0') +588 +0.004328833670911726 +changing lr +epoch 38, time 796.82, cls_loss 0.0008 cls_loss_mapping 0.0106 cls_loss_causal 0.4009 re_mapping 0.0075 re_causal 0.0187 /// teacc 92.21 lr 0.00410722 +Epoch 40, weight, value: tensor([[ 0.2865, 0.2674, 0.2463, ..., -0.0143, -0.0146, -0.0172], + [-0.0047, -0.0478, 0.0121, ..., -0.0415, -0.0067, -0.0099], + [-0.0831, -0.0629, -0.1036, ..., 0.0125, -0.0064, -0.0033], + ..., + [-0.2026, -0.1846, -0.1607, ..., 0.0213, 0.0062, 0.0089], + [-0.0061, 0.0043, 0.0006, ..., 0.0142, 0.0132, 0.0012], + [ 0.0169, 0.0243, 0.0142, ..., -0.0338, -0.0394, -0.0314]], + device='cuda:0'), grad: tensor([[-1.0471e-03, -6.4516e-04, -6.2609e-04, ..., -9.7811e-05, + -1.0026e-04, -8.8274e-05], + [ 1.0246e-04, 6.1035e-05, 5.9575e-05, ..., 1.1928e-05, + 1.2927e-05, 1.1586e-05], + [ 9.3520e-05, 5.3167e-05, 5.2869e-05, ..., 2.1726e-05, + 2.2560e-05, 2.1011e-05], + ..., + [ 4.9877e-04, 2.8777e-04, 2.8229e-04, ..., 5.8442e-05, + 5.5343e-05, 4.8816e-05], + [ 8.1003e-05, 4.9174e-05, 4.7952e-05, ..., 8.1807e-06, + 8.8364e-06, 7.8455e-06], + [ 1.1645e-05, 1.2942e-05, 1.1452e-05, ..., 1.0058e-06, + 5.8254e-07, 7.1852e-07]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0849, 0.0044, -0.0183, -0.0719, 0.0068, 0.0039, 0.0043], + device='cuda:0'), grad: tensor([-1.3752e-03, 1.4281e-04, 1.3685e-04, 2.7895e-04, 7.0906e-04, + 1.0878e-04, -1.1288e-06], device='cuda:0') +588 +0.0041072155260068206 +changing lr +epoch 39, time 799.50, cls_loss 0.0009 cls_loss_mapping 0.0103 cls_loss_causal 0.3938 re_mapping 0.0073 re_causal 0.0185 /// teacc 92.21 lr 0.00388740 +Epoch 41, weight, value: tensor([[ 0.2860, 0.2672, 0.2464, ..., -0.0146, -0.0149, -0.0174], + [-0.0048, -0.0475, 0.0118, ..., -0.0409, -0.0065, -0.0097], + [-0.0834, -0.0634, -0.1036, ..., 0.0123, -0.0064, -0.0034], + ..., + [-0.2027, -0.1840, -0.1605, ..., 0.0210, 0.0061, 0.0088], + [-0.0049, 0.0046, 0.0009, ..., 0.0143, 0.0132, 0.0014], + [ 0.0171, 0.0240, 0.0140, ..., -0.0335, -0.0390, -0.0311]], + device='cuda:0'), grad: tensor([[ 1.1873e-04, 2.8163e-05, 3.3200e-05, ..., 3.4928e-05, + 3.4213e-05, 3.3349e-05], + [-1.1492e-04, -4.1366e-05, -5.5224e-05, ..., -3.8594e-05, + -3.7491e-05, -3.2783e-05], + [ 4.0561e-05, 1.4029e-05, 1.6689e-05, ..., 1.1601e-05, + 1.1168e-05, 1.0289e-05], + ..., + [-5.4598e-05, 5.8040e-06, 1.1645e-05, ..., -7.5698e-06, + -4.4480e-06, -8.8960e-06], + [ 2.8580e-05, 8.0168e-06, 9.9689e-06, ..., 8.9481e-06, + 8.7991e-06, 8.2776e-06], + [ 4.1574e-05, 1.2174e-05, 1.4141e-05, ..., 1.1735e-05, + 1.1489e-05, 1.0811e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0841, 0.0050, -0.0180, -0.0731, 0.0038, 0.0064, 0.0057], + device='cuda:0'), grad: tensor([ 2.9898e-04, -2.4199e-04, 8.8632e-05, -1.0020e-04, -2.1112e-04, + 6.7532e-05, 9.8109e-05], device='cuda:0') +588 +0.0038873953302184317 +changing lr +epoch 40, time 795.42, cls_loss 0.0008 cls_loss_mapping 0.0091 cls_loss_causal 0.4129 re_mapping 0.0072 re_causal 0.0185 /// teacc 94.22 lr 0.00366982 +Epoch 42, weight, value: tensor([[ 0.2847, 0.2663, 0.2458, ..., -0.0150, -0.0152, -0.0178], + [-0.0043, -0.0469, 0.0117, ..., -0.0404, -0.0064, -0.0096], + [-0.0830, -0.0635, -0.1034, ..., 0.0120, -0.0065, -0.0035], + ..., + [-0.2009, -0.1828, -0.1596, ..., 0.0213, 0.0065, 0.0092], + [-0.0055, 0.0042, 0.0005, ..., 0.0141, 0.0131, 0.0014], + [ 0.0165, 0.0234, 0.0135, ..., -0.0334, -0.0388, -0.0309]], + device='cuda:0'), grad: tensor([[ 5.4538e-05, 2.4676e-05, 2.5213e-05, ..., 7.8529e-06, + 8.7097e-06, 8.2329e-06], + [ 1.2949e-05, 1.4193e-06, 2.9383e-07, ..., 3.0547e-06, + 2.4810e-06, 3.0231e-06], + [-1.2410e-04, -5.5641e-05, -5.7817e-05, ..., -4.5933e-06, + -8.3447e-06, -7.0632e-06], + ..., + [ 2.1502e-05, 1.6764e-05, 1.8194e-05, ..., -5.8934e-06, + -4.7050e-06, -5.0850e-06], + [ 1.2375e-05, 5.1744e-06, 5.4352e-06, ..., 1.1707e-06, + 1.5898e-06, 1.3737e-06], + [ 1.5512e-05, 6.5938e-06, 6.9365e-06, ..., 1.7397e-06, + 2.2594e-06, 1.9930e-06]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0822, 0.0061, -0.0170, -0.0737, 0.0054, 0.0054, 0.0053], + device='cuda:0'), grad: tensor([ 9.6619e-05, 2.9564e-05, -2.1553e-04, 2.1651e-05, 1.3605e-05, + 2.3916e-05, 3.0071e-05], device='cuda:0') +588 +0.003669815772166629 +changing lr +epoch 41, time 801.63, cls_loss 0.0008 cls_loss_mapping 0.0096 cls_loss_causal 0.3998 re_mapping 0.0070 re_causal 0.0182 /// teacc 92.96 lr 0.00345492 +Epoch 43, weight, value: tensor([[ 0.2846, 0.2662, 0.2459, ..., -0.0152, -0.0155, -0.0180], + [-0.0056, -0.0472, 0.0108, ..., -0.0403, -0.0067, -0.0098], + [-0.0829, -0.0637, -0.1033, ..., 0.0119, -0.0065, -0.0035], + ..., + [-0.1998, -0.1820, -0.1591, ..., 0.0215, 0.0069, 0.0095], + [-0.0059, 0.0039, 0.0003, ..., 0.0139, 0.0129, 0.0013], + [ 0.0164, 0.0230, 0.0133, ..., -0.0332, -0.0386, -0.0308]], + device='cuda:0'), grad: tensor([[-4.0722e-04, -2.7061e-04, -2.7323e-04, ..., -2.7493e-05, + -2.5436e-05, -2.3738e-05], + [ 1.4566e-05, 1.2442e-05, 1.2346e-05, ..., -1.4491e-06, + -1.7099e-06, -1.7192e-06], + [ 1.2808e-05, 2.1994e-05, 2.2396e-05, ..., -6.4671e-06, + -5.9083e-06, -6.0797e-06], + ..., + [ 6.3896e-05, 3.8564e-05, 3.9160e-05, ..., 6.4485e-06, + 6.1393e-06, 5.8524e-06], + [ 4.0621e-05, 2.6256e-05, 2.6584e-05, ..., 3.2447e-06, + 3.0492e-06, 2.8592e-06], + [ 4.6164e-05, 2.8789e-05, 2.9176e-05, ..., 4.4964e-06, + 4.2766e-06, 4.0680e-06]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0823, 0.0040, -0.0161, -0.0725, 0.0060, 0.0045, 0.0053], + device='cuda:0'), grad: tensor([-4.8351e-04, 7.0594e-06, -2.9013e-05, 3.0375e-04, 8.9645e-05, + 5.0664e-05, 6.1691e-05], device='cuda:0') +588 +0.0034549150281252667 +changing lr +epoch 42, time 796.69, cls_loss 0.0008 cls_loss_mapping 0.0101 cls_loss_causal 0.3981 re_mapping 0.0069 re_causal 0.0177 /// teacc 92.96 lr 0.00324313 +Epoch 44, weight, value: tensor([[ 2.8448e-01, 2.6600e-01, 2.4602e-01, ..., -1.5436e-02, + -1.5655e-02, -1.8136e-02], + [-6.1491e-03, -4.7257e-02, 1.0294e-02, ..., -3.9985e-02, + -6.6576e-03, -9.7128e-03], + [-8.2745e-02, -6.3955e-02, -1.0317e-01, ..., 1.1570e-02, + -6.6499e-03, -3.7478e-03], + ..., + [-1.9834e-01, -1.8096e-01, -1.5846e-01, ..., 2.1870e-02, + 7.3295e-03, 9.8903e-03], + [-6.0654e-03, 3.7579e-03, 1.6885e-04, ..., 1.3831e-02, + 1.2828e-02, 1.3772e-03], + [ 1.5728e-02, 2.2594e-02, 1.2900e-02, ..., -3.3166e-02, + -3.8398e-02, -3.0691e-02]], device='cuda:0'), grad: tensor([[ 7.6950e-05, 3.3706e-05, 3.8385e-05, ..., 2.7925e-05, + 2.6584e-05, 2.6196e-05], + [-8.2791e-05, -2.1219e-05, -2.8834e-05, ..., -1.4342e-05, + -1.2286e-05, -1.1615e-05], + [ 2.6226e-05, 1.0192e-05, 1.1869e-05, ..., 8.0168e-06, + 7.5251e-06, 7.3686e-06], + ..., + [ 5.4121e-05, 1.8701e-05, 2.2903e-05, ..., 1.1973e-05, + 1.0692e-05, 1.0453e-05], + [ 2.2337e-05, 8.5011e-06, 9.4399e-06, ..., 4.0531e-06, + 3.9600e-06, 3.8557e-06], + [-1.3709e-05, -5.4650e-06, -5.3830e-06, ..., 1.2210e-06, + 8.2562e-07, 8.4797e-07]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0827, 0.0037, -0.0154, -0.0732, 0.0073, 0.0040, 0.0044], + device='cuda:0'), grad: tensor([ 1.4389e-04, -2.1696e-04, 5.4330e-05, -1.2207e-04, 1.2040e-04, + 4.5657e-05, -2.5272e-05], device='cuda:0') +588 +0.0032431258795932905 +changing lr +epoch 43, time 802.38, cls_loss 0.0008 cls_loss_mapping 0.0091 cls_loss_causal 0.4083 re_mapping 0.0068 re_causal 0.0177 /// teacc 93.22 lr 0.00303487 +Epoch 45, weight, value: tensor([[ 2.8393e-01, 2.6552e-01, 2.4583e-01, ..., -1.5520e-02, + -1.5750e-02, -1.8209e-02], + [-5.5705e-03, -4.6848e-02, 1.0230e-02, ..., -3.9526e-02, + -6.4789e-03, -9.5158e-03], + [-8.3250e-02, -6.4266e-02, -1.0321e-01, ..., 1.1034e-02, + -7.0113e-03, -4.1401e-03], + ..., + [-1.9825e-01, -1.8043e-01, -1.5828e-01, ..., 2.1683e-02, + 7.2651e-03, 9.8060e-03], + [-6.1646e-03, 3.5860e-03, 3.1987e-05, ..., 1.3814e-02, + 1.2821e-02, 1.4692e-03], + [ 1.5339e-02, 2.2269e-02, 1.2642e-02, ..., -3.3106e-02, + -3.8273e-02, -3.0628e-02]], device='cuda:0'), grad: tensor([[ 3.4302e-05, 8.4788e-06, 9.6858e-06, ..., 9.8124e-06, + 9.0227e-06, 8.6725e-06], + [-2.8417e-05, -1.6123e-05, -1.9059e-05, ..., -6.7912e-06, + -7.6666e-06, -7.0557e-06], + [ 2.9299e-06, 1.5199e-06, 2.6412e-06, ..., 2.3805e-06, + 3.1367e-06, 2.9635e-06], + ..., + [-6.4790e-05, -4.4033e-06, -4.4666e-06, ..., -1.8165e-05, + -1.6659e-05, -1.6019e-05], + [ 2.3350e-05, 3.3956e-06, 3.2838e-06, ..., 4.9174e-06, + 5.0031e-06, 4.6790e-06], + [ 2.0295e-05, 5.9754e-06, 6.7651e-06, ..., 5.3309e-06, + 5.0589e-06, 4.7907e-06]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0823, 0.0057, -0.0166, -0.0713, 0.0055, 0.0038, 0.0038], + device='cuda:0'), grad: tensor([ 8.9884e-05, -5.4538e-05, 1.0334e-05, 3.8356e-05, -2.0170e-04, + 6.6519e-05, 5.1379e-05], device='cuda:0') +588 +0.0030348748417303863 +changing lr +epoch 44, time 799.44, cls_loss 0.0007 cls_loss_mapping 0.0072 cls_loss_causal 0.3894 re_mapping 0.0068 re_causal 0.0173 /// teacc 92.96 lr 0.00283058 +Epoch 46, weight, value: tensor([[ 2.8382e-01, 2.6527e-01, 2.4582e-01, ..., -1.5541e-02, + -1.5773e-02, -1.8226e-02], + [-5.8964e-03, -4.6766e-02, 9.8641e-03, ..., -3.9271e-02, + -6.4859e-03, -9.4906e-03], + [-8.2545e-02, -6.4172e-02, -1.0280e-01, ..., 1.0980e-02, + -6.9455e-03, -4.1089e-03], + ..., + [-1.9680e-01, -1.7954e-01, -1.5767e-01, ..., 2.1922e-02, + 7.6230e-03, 1.0123e-02], + [-6.4769e-03, 3.3920e-03, -1.4152e-04, ..., 1.3695e-02, + 1.2709e-02, 1.4647e-03], + [ 1.4935e-02, 2.1985e-02, 1.2435e-02, ..., -3.3013e-02, + -3.8111e-02, -3.0528e-02]], device='cuda:0'), grad: tensor([[ 1.3173e-05, 1.7118e-06, 1.0105e-06, ..., 5.3868e-06, + 4.4331e-06, 4.7274e-06], + [ 1.0185e-05, 1.5367e-07, -1.1753e-06, ..., 5.7220e-06, + 4.2543e-06, 4.0792e-06], + [-1.3161e-04, -1.0498e-05, -4.6603e-06, ..., -6.7949e-05, + -5.2184e-05, -5.5194e-05], + ..., + [ 2.9318e-06, -1.4808e-06, -1.1141e-07, ..., 3.6526e-06, + 2.5872e-06, 2.7567e-06], + [ 7.6108e-06, 1.0431e-06, 7.7253e-07, ..., 3.2224e-06, + 2.5909e-06, 2.7474e-06], + [ 1.3076e-05, 2.0694e-06, 1.6941e-06, ..., 5.5432e-06, + 4.4182e-06, 4.7535e-06]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0828, 0.0054, -0.0150, -0.0734, 0.0072, 0.0031, 0.0030], + device='cuda:0'), grad: tensor([ 4.0889e-05, 3.4660e-05, -4.1389e-04, 2.6488e-04, 1.2204e-05, + 2.3052e-05, 3.8326e-05], device='cuda:0') +588 +0.0028305813044122124 +changing lr +epoch 45, time 801.33, cls_loss 0.0006 cls_loss_mapping 0.0064 cls_loss_causal 0.3921 re_mapping 0.0067 re_causal 0.0170 /// teacc 92.71 lr 0.00263066 +Epoch 47, weight, value: tensor([[ 0.2831, 0.2648, 0.2456, ..., -0.0157, -0.0159, -0.0183], + [-0.0057, -0.0465, 0.0097, ..., -0.0389, -0.0064, -0.0094], + [-0.0824, -0.0644, -0.1027, ..., 0.0108, -0.0070, -0.0042], + ..., + [-0.1963, -0.1789, -0.1573, ..., 0.0219, 0.0077, 0.0101], + [-0.0069, 0.0031, -0.0004, ..., 0.0136, 0.0126, 0.0014], + [ 0.0151, 0.0218, 0.0124, ..., -0.0328, -0.0379, -0.0304]], + device='cuda:0'), grad: tensor([[-3.5644e-04, -2.7180e-04, -2.5177e-04, ..., -1.9185e-06, + 1.2331e-06, -3.2806e-07], + [ 8.8036e-05, 5.2333e-05, 4.8786e-05, ..., 4.4405e-06, + 3.6415e-06, 3.0380e-06], + [ 1.5706e-05, 6.1154e-05, 5.3763e-05, ..., -1.1146e-05, + -1.1571e-05, -9.0748e-06], + ..., + [ 3.6299e-05, 2.7597e-05, 2.6762e-05, ..., -5.6289e-06, + -6.2473e-06, -4.9658e-06], + [ 3.3677e-05, 1.6645e-05, 1.5840e-05, ..., 3.6806e-06, + 3.5185e-06, 3.0026e-06], + [ 8.1122e-05, 4.6939e-05, 4.4197e-05, ..., 6.7726e-06, + 6.3442e-06, 5.5581e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0819, 0.0062, -0.0144, -0.0731, 0.0064, 0.0021, 0.0039], + device='cuda:0'), grad: tensor([-2.9564e-04, 1.2982e-04, -1.6510e-04, 1.2386e-04, 1.8924e-05, + 6.2943e-05, 1.2505e-04], device='cuda:0') +588 +0.0026306566876350096 +changing lr +epoch 46, time 798.60, cls_loss 0.0007 cls_loss_mapping 0.0080 cls_loss_causal 0.3906 re_mapping 0.0066 re_causal 0.0169 /// teacc 93.47 lr 0.00243550 +Epoch 48, weight, value: tensor([[ 0.2834, 0.2647, 0.2457, ..., -0.0157, -0.0160, -0.0184], + [-0.0062, -0.0465, 0.0093, ..., -0.0386, -0.0063, -0.0093], + [-0.0827, -0.0646, -0.1027, ..., 0.0106, -0.0071, -0.0044], + ..., + [-0.1956, -0.1782, -0.1569, ..., 0.0219, 0.0078, 0.0103], + [-0.0069, 0.0030, -0.0005, ..., 0.0135, 0.0126, 0.0015], + [ 0.0150, 0.0216, 0.0123, ..., -0.0328, -0.0378, -0.0303]], + device='cuda:0'), grad: tensor([[-2.7746e-05, -1.9118e-05, -2.0504e-05, ..., -3.5646e-07, + -4.0345e-06, -3.7979e-06], + [-3.4124e-05, -7.2643e-06, -1.1057e-05, ..., -3.6228e-06, + -1.4268e-06, -2.8801e-07], + [-3.8333e-06, 2.4959e-06, 3.3453e-06, ..., -5.1335e-06, + -4.5225e-06, -4.2729e-06], + ..., + [ 4.4078e-05, 1.7509e-05, 2.0623e-05, ..., 6.7875e-06, + 7.3239e-06, 6.3516e-06], + [ 7.0073e-06, 2.4196e-06, 2.9020e-06, ..., 1.3020e-06, + 1.2852e-06, 1.1139e-06], + [ 1.2197e-05, 4.6082e-06, 5.2340e-06, ..., 1.8831e-06, + 1.9781e-06, 1.6866e-06]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0831, 0.0055, -0.0148, -0.0733, 0.0063, 0.0024, 0.0038], + device='cuda:0'), grad: tensor([-3.6061e-05, -7.7128e-05, -2.3782e-05, 1.1355e-05, 8.5950e-05, + 1.4737e-05, 2.4915e-05], device='cuda:0') +588 +0.0024355036129704724 +changing lr +epoch 47, time 799.38, cls_loss 0.0006 cls_loss_mapping 0.0071 cls_loss_causal 0.3829 re_mapping 0.0065 re_causal 0.0164 /// teacc 92.46 lr 0.00224552 +Epoch 49, weight, value: tensor([[ 0.2825, 0.2642, 0.2453, ..., -0.0160, -0.0162, -0.0186], + [-0.0069, -0.0466, 0.0088, ..., -0.0384, -0.0063, -0.0092], + [-0.0828, -0.0647, -0.1026, ..., 0.0102, -0.0074, -0.0046], + ..., + [-0.1945, -0.1775, -0.1563, ..., 0.0220, 0.0080, 0.0104], + [-0.0068, 0.0028, -0.0006, ..., 0.0135, 0.0126, 0.0016], + [ 0.0148, 0.0215, 0.0121, ..., -0.0328, -0.0377, -0.0303]], + device='cuda:0'), grad: tensor([[-2.8825e-04, -2.1040e-04, -2.0421e-04, ..., -1.5795e-05, + -1.5706e-05, -1.4976e-05], + [ 2.1815e-04, 1.1706e-04, 1.1510e-04, ..., 2.1115e-05, + 2.1413e-05, 1.8880e-05], + [-4.8733e-04, -3.6329e-05, -4.1038e-05, ..., -3.7730e-05, + -3.0607e-05, -2.3693e-05], + ..., + [-4.5472e-07, 1.8477e-05, 1.4365e-05, ..., -2.0981e-05, + -2.7463e-05, -2.2218e-05], + [ 3.7432e-05, 1.5169e-05, 1.5691e-05, ..., 6.3702e-06, + 7.2643e-06, 6.0685e-06], + [ 5.8204e-05, 2.2158e-05, 2.2754e-05, ..., 8.0839e-06, + 8.8960e-06, 7.4431e-06]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0819, 0.0044, -0.0151, -0.0720, 0.0073, 0.0029, 0.0034], + device='cuda:0'), grad: tensor([-2.9588e-04, 3.6693e-04, -1.6050e-03, 1.3800e-03, -5.4926e-05, + 7.8201e-05, 1.2743e-04], device='cuda:0') +588 +0.00224551509273949 +changing lr +epoch 48, time 786.01, cls_loss 0.0007 cls_loss_mapping 0.0063 cls_loss_causal 0.3663 re_mapping 0.0065 re_causal 0.0160 /// teacc 93.97 lr 0.00206107 +Epoch 50, weight, value: tensor([[ 0.2825, 0.2641, 0.2453, ..., -0.0160, -0.0163, -0.0186], + [-0.0069, -0.0464, 0.0087, ..., -0.0381, -0.0063, -0.0092], + [-0.0829, -0.0649, -0.1026, ..., 0.0100, -0.0075, -0.0047], + ..., + [-0.1937, -0.1771, -0.1560, ..., 0.0221, 0.0082, 0.0106], + [-0.0071, 0.0026, -0.0008, ..., 0.0134, 0.0125, 0.0016], + [ 0.0147, 0.0214, 0.0121, ..., -0.0327, -0.0376, -0.0302]], + device='cuda:0'), grad: tensor([[ 1.6713e-04, 7.1347e-05, 7.0751e-05, ..., 2.7284e-05, + 3.0518e-05, 2.8655e-05], + [-1.8609e-04, -6.7532e-05, -7.6532e-05, ..., -4.3631e-05, + -4.6760e-05, -4.6492e-05], + [-3.3192e-06, 4.4480e-06, 4.6082e-06, ..., -5.1633e-06, + -4.9211e-06, -4.6492e-06], + ..., + [ 6.2525e-05, 1.8641e-05, 2.6613e-05, ..., 1.5542e-05, + 1.5467e-05, 1.6257e-05], + [ 2.8655e-05, 1.3143e-05, 1.4052e-05, ..., 2.2389e-06, + 2.3972e-06, 2.2817e-06], + [-9.3400e-05, -4.9412e-05, -5.0008e-05, ..., 7.0222e-07, + 1.8487e-07, 9.0851e-07]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0826, 0.0046, -0.0151, -0.0728, 0.0080, 0.0023, 0.0032], + device='cuda:0'), grad: tensor([ 3.5238e-04, -4.3797e-04, -2.6524e-05, 5.2154e-05, 1.5450e-04, + 5.3644e-05, -1.4865e-04], device='cuda:0') +588 +0.002061073738537637 +changing lr +epoch 49, time 797.69, cls_loss 0.0005 cls_loss_mapping 0.0065 cls_loss_causal 0.3759 re_mapping 0.0063 re_causal 0.0155 /// teacc 94.22 lr 0.00188255 +Epoch 51, weight, value: tensor([[ 0.2822, 0.2638, 0.2452, ..., -0.0161, -0.0164, -0.0187], + [-0.0069, -0.0463, 0.0085, ..., -0.0378, -0.0062, -0.0091], + [-0.0827, -0.0649, -0.1024, ..., 0.0099, -0.0075, -0.0048], + ..., + [-0.1938, -0.1768, -0.1559, ..., 0.0220, 0.0082, 0.0105], + [-0.0074, 0.0025, -0.0010, ..., 0.0133, 0.0124, 0.0015], + [ 0.0151, 0.0214, 0.0122, ..., -0.0326, -0.0374, -0.0300]], + device='cuda:0'), grad: tensor([[ 2.1413e-05, 9.0152e-06, 9.7007e-06, ..., 4.5635e-06, + 4.5709e-06, 4.1835e-06], + [-1.8728e-04, -9.0241e-05, -9.6738e-05, ..., -2.8402e-05, + -2.8253e-05, -2.5168e-05], + [ 3.2693e-05, 2.1756e-05, 2.2322e-05, ..., 4.7423e-06, + 4.4219e-06, 4.1835e-06], + ..., + [ 1.9789e-05, 3.2187e-06, 5.0440e-06, ..., 2.2762e-06, + 2.1271e-06, 1.6373e-06], + [ 1.0066e-05, 2.8573e-06, 3.2932e-06, ..., 2.0619e-06, + 2.1830e-06, 1.9092e-06], + [ 9.8228e-05, 5.1469e-05, 5.4270e-05, ..., 1.5110e-05, + 1.5125e-05, 1.3597e-05]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0823, 0.0050, -0.0144, -0.0727, 0.0064, 0.0019, 0.0042], + device='cuda:0'), grad: tensor([ 4.7833e-05, -3.8409e-04, 4.9025e-05, 1.2919e-05, 5.9575e-05, + 2.6226e-05, 1.8859e-04], device='cuda:0') +588 +0.0018825509907063344 +changing lr +epoch 50, time 796.52, cls_loss 0.0007 cls_loss_mapping 0.0074 cls_loss_causal 0.3886 re_mapping 0.0062 re_causal 0.0157 /// teacc 93.22 lr 0.00171031 +Epoch 52, weight, value: tensor([[ 0.2819, 0.2636, 0.2451, ..., -0.0163, -0.0165, -0.0188], + [-0.0068, -0.0461, 0.0085, ..., -0.0376, -0.0061, -0.0089], + [-0.0821, -0.0649, -0.1023, ..., 0.0098, -0.0075, -0.0048], + ..., + [-0.1935, -0.1765, -0.1557, ..., 0.0220, 0.0083, 0.0106], + [-0.0075, 0.0024, -0.0010, ..., 0.0133, 0.0123, 0.0015], + [ 0.0146, 0.0211, 0.0119, ..., -0.0325, -0.0373, -0.0300]], + device='cuda:0'), grad: tensor([[-1.2130e-04, -7.1228e-05, -8.4400e-05, ..., -5.5805e-06, + -5.2676e-06, -4.9733e-06], + [ 1.9139e-07, 6.3889e-06, 8.0541e-06, ..., -2.7455e-06, + -3.0957e-06, -2.9020e-06], + [ 1.7852e-05, 1.5929e-05, 1.8746e-05, ..., -4.7013e-06, + -4.3735e-06, -4.2170e-06], + ..., + [ 2.4468e-05, 1.1340e-05, 1.3232e-05, ..., 3.0175e-06, + 2.8796e-06, 2.7996e-06], + [ 1.4387e-05, 7.3798e-06, 8.6948e-06, ..., 1.2759e-06, + 1.2778e-06, 1.1921e-06], + [ 2.1636e-05, 1.0438e-05, 1.2234e-05, ..., 2.1234e-06, + 2.1681e-06, 2.0266e-06]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0819, 0.0055, -0.0127, -0.0730, 0.0061, 0.0016, 0.0033], + device='cuda:0'), grad: tensor([-1.4007e-04, -2.4676e-05, -2.0321e-06, 7.0810e-05, 4.0650e-05, + 2.1055e-05, 3.4153e-05], device='cuda:0') +588 +0.0017103063703014388 +changing lr +epoch 51, time 797.46, cls_loss 0.0010 cls_loss_mapping 0.0071 cls_loss_causal 0.3851 re_mapping 0.0060 re_causal 0.0155 /// teacc 93.47 lr 0.00154469 +Epoch 53, weight, value: tensor([[ 0.2814, 0.2633, 0.2449, ..., -0.0165, -0.0167, -0.0190], + [-0.0062, -0.0459, 0.0085, ..., -0.0373, -0.0059, -0.0087], + [-0.0825, -0.0651, -0.1023, ..., 0.0096, -0.0077, -0.0050], + ..., + [-0.1931, -0.1761, -0.1555, ..., 0.0222, 0.0085, 0.0108], + [-0.0076, 0.0023, -0.0011, ..., 0.0132, 0.0123, 0.0015], + [ 0.0144, 0.0209, 0.0118, ..., -0.0325, -0.0372, -0.0300]], + device='cuda:0'), grad: tensor([[ 5.0306e-04, 1.1176e-04, 1.3840e-04, ..., 1.1438e-04, + 1.2243e-04, 1.0753e-04], + [-1.2469e-04, -2.7537e-05, -3.8087e-05, ..., -1.2085e-05, + -9.9987e-06, -8.7470e-06], + [-5.5742e-04, -1.1384e-04, -1.6129e-04, ..., -6.0469e-05, + -8.7380e-05, -6.1452e-05], + ..., + [-1.3426e-05, -8.6799e-06, 1.1496e-05, ..., -7.4804e-05, + -6.1810e-05, -6.7234e-05], + [ 3.3736e-05, 8.1658e-06, 9.3505e-06, ..., 7.0818e-06, + 7.5251e-06, 6.6124e-06], + [ 7.1108e-05, 1.7479e-05, 2.0579e-05, ..., 1.3396e-05, + 1.3880e-05, 1.2286e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0812, 0.0071, -0.0138, -0.0725, 0.0063, 0.0013, 0.0031], + device='cuda:0'), grad: tensor([ 1.3027e-03, -3.2806e-04, -1.4381e-03, 2.5582e-04, -5.6207e-05, + 8.5473e-05, 1.7917e-04], device='cuda:0') +588 +0.0015446867550656784 +changing lr +epoch 52, time 799.47, cls_loss 0.0006 cls_loss_mapping 0.0062 cls_loss_causal 0.3679 re_mapping 0.0061 re_causal 0.0153 /// teacc 92.46 lr 0.00138603 +Epoch 54, weight, value: tensor([[ 0.2812, 0.2631, 0.2448, ..., -0.0166, -0.0168, -0.0191], + [-0.0068, -0.0460, 0.0081, ..., -0.0372, -0.0060, -0.0088], + [-0.0824, -0.0652, -0.1022, ..., 0.0095, -0.0077, -0.0050], + ..., + [-0.1928, -0.1758, -0.1553, ..., 0.0222, 0.0085, 0.0108], + [-0.0075, 0.0022, -0.0011, ..., 0.0132, 0.0123, 0.0016], + [ 0.0146, 0.0209, 0.0118, ..., -0.0324, -0.0371, -0.0299]], + device='cuda:0'), grad: tensor([[ 2.4110e-05, 6.9663e-06, 7.5586e-06, ..., 1.0364e-05, + 1.0893e-05, 9.9912e-06], + [ 4.3609e-07, -5.4240e-06, -5.9977e-06, ..., 4.2245e-06, + 2.6245e-06, 2.1569e-06], + [-4.9204e-05, 4.0187e-07, 1.6494e-06, ..., -1.9670e-05, + -1.4894e-05, -1.3962e-05], + ..., + [ 3.5733e-05, 8.7023e-06, 1.0386e-05, ..., 1.7047e-05, + 1.8060e-05, 1.5751e-05], + [ 1.2994e-05, 3.1758e-06, 3.1926e-06, ..., 5.6587e-06, + 5.4687e-06, 5.0776e-06], + [ 2.2680e-05, 7.6294e-06, 8.2329e-06, ..., 1.0602e-05, + 1.1273e-05, 1.0207e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0810, 0.0060, -0.0134, -0.0723, 0.0059, 0.0018, 0.0036], + device='cuda:0'), grad: tensor([ 5.9038e-05, 2.4468e-05, -1.8048e-04, -8.2850e-05, 9.2089e-05, + 3.5256e-05, 5.2601e-05], device='cuda:0') +588 +0.001386025680863044 +changing lr +epoch 53, time 798.75, cls_loss 0.0006 cls_loss_mapping 0.0049 cls_loss_causal 0.3745 re_mapping 0.0061 re_causal 0.0150 /// teacc 92.96 lr 0.00123464 +Epoch 55, weight, value: tensor([[ 0.2806, 0.2628, 0.2445, ..., -0.0167, -0.0169, -0.0192], + [-0.0070, -0.0460, 0.0080, ..., -0.0371, -0.0060, -0.0088], + [-0.0821, -0.0652, -0.1021, ..., 0.0095, -0.0077, -0.0051], + ..., + [-0.1922, -0.1753, -0.1549, ..., 0.0222, 0.0086, 0.0109], + [-0.0076, 0.0021, -0.0012, ..., 0.0131, 0.0122, 0.0016], + [ 0.0144, 0.0208, 0.0117, ..., -0.0324, -0.0370, -0.0298]], + device='cuda:0'), grad: tensor([[-4.9561e-05, -3.4869e-05, -3.9577e-05, ..., -1.2685e-06, + -2.1383e-06, -1.1781e-06], + [-6.4909e-05, -5.1456e-07, -2.5555e-06, ..., -3.8207e-05, + -4.4256e-05, -4.2200e-05], + [-3.7760e-05, -9.9838e-07, -7.8464e-08, ..., -1.3851e-05, + -1.0043e-05, -1.0252e-05], + ..., + [ 7.9215e-05, 1.3858e-05, 1.6168e-05, ..., 3.2604e-05, + 3.4302e-05, 3.2842e-05], + [ 1.0483e-05, 3.6396e-06, 4.1686e-06, ..., 2.8908e-06, + 3.1721e-06, 2.9635e-06], + [ 2.3022e-05, 6.9551e-06, 8.1956e-06, ..., 7.1935e-06, + 7.9051e-06, 7.3984e-06]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0803, 0.0059, -0.0126, -0.0723, 0.0064, 0.0015, 0.0033], + device='cuda:0'), grad: tensor([-4.4674e-05, -2.2376e-04, -1.3554e-04, 9.5725e-05, 2.2900e-04, + 2.3529e-05, 5.5224e-05], device='cuda:0') +588 +0.0012346426699819469 +changing lr +epoch 54, time 794.37, cls_loss 0.0005 cls_loss_mapping 0.0049 cls_loss_causal 0.3706 re_mapping 0.0060 re_causal 0.0148 /// teacc 93.47 lr 0.00109084 +Epoch 56, weight, value: tensor([[ 0.2811, 0.2629, 0.2447, ..., -0.0166, -0.0168, -0.0191], + [-0.0071, -0.0460, 0.0078, ..., -0.0370, -0.0060, -0.0088], + [-0.0823, -0.0653, -0.1021, ..., 0.0093, -0.0078, -0.0051], + ..., + [-0.1918, -0.1750, -0.1548, ..., 0.0222, 0.0086, 0.0109], + [-0.0078, 0.0021, -0.0013, ..., 0.0131, 0.0122, 0.0015], + [ 0.0141, 0.0206, 0.0115, ..., -0.0324, -0.0370, -0.0298]], + device='cuda:0'), grad: tensor([[ 5.1051e-05, 1.3635e-05, 1.3933e-05, ..., 1.2957e-05, + 1.3240e-05, 1.3150e-05], + [-1.3411e-04, -4.5836e-05, -5.3883e-05, ..., -3.7611e-05, + -3.7819e-05, -3.2663e-05], + [ 2.5988e-05, 7.5363e-06, 8.6576e-06, ..., 9.8124e-06, + 1.0036e-05, 9.9838e-06], + ..., + [ 1.7047e-04, 4.1544e-05, 5.0306e-05, ..., 4.8578e-05, + 4.9680e-05, 4.7624e-05], + [ 4.1395e-05, 7.3984e-06, 8.5235e-06, ..., 1.0625e-05, + 1.1064e-05, 1.1601e-05], + [-1.5807e-04, -2.1905e-05, -2.5377e-05, ..., -3.7372e-05, + -3.9428e-05, -4.3184e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0815, 0.0058, -0.0127, -0.0724, 0.0065, 0.0011, 0.0027], + device='cuda:0'), grad: tensor([ 1.3351e-04, -3.0065e-04, 6.3956e-05, 2.3320e-05, 4.5133e-04, + 1.2201e-04, -4.9305e-04], device='cuda:0') +588 +0.0010908425876598518 +changing lr +epoch 55, time 827.55, cls_loss 0.0005 cls_loss_mapping 0.0045 cls_loss_causal 0.3826 re_mapping 0.0060 re_causal 0.0148 /// teacc 92.46 lr 0.00095492 +Epoch 57, weight, value: tensor([[ 0.2808, 0.2627, 0.2446, ..., -0.0167, -0.0169, -0.0192], + [-0.0070, -0.0458, 0.0078, ..., -0.0368, -0.0059, -0.0087], + [-0.0823, -0.0654, -0.1021, ..., 0.0092, -0.0078, -0.0052], + ..., + [-0.1916, -0.1748, -0.1546, ..., 0.0222, 0.0087, 0.0109], + [-0.0078, 0.0020, -0.0014, ..., 0.0131, 0.0122, 0.0016], + [ 0.0141, 0.0206, 0.0115, ..., -0.0323, -0.0369, -0.0298]], + device='cuda:0'), grad: tensor([[-4.7827e-04, -1.5104e-04, -2.0325e-04, ..., -1.0800e-04, + -1.0729e-04, -9.7275e-05], + [ 4.2200e-05, 1.0833e-05, 1.6347e-05, ..., 1.2152e-05, + 1.1913e-05, 1.0714e-05], + [ 3.6031e-05, 1.1928e-05, 1.6063e-05, ..., 8.0392e-06, + 7.8678e-06, 7.2978e-06], + ..., + [ 3.1161e-04, 9.9123e-05, 1.3292e-04, ..., 6.9380e-05, + 6.9082e-05, 6.2525e-05], + [ 2.9162e-05, 9.4175e-06, 1.2636e-05, ..., 6.4299e-06, + 6.3479e-06, 5.8264e-06], + [ 3.0756e-05, 1.0729e-05, 1.3247e-05, ..., 5.9083e-06, + 6.0424e-06, 5.4277e-06]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0812, 0.0062, -0.0128, -0.0724, 0.0061, 0.0012, 0.0029], + device='cuda:0'), grad: tensor([-9.7942e-04, 8.6308e-05, 7.3612e-05, 5.9694e-05, 6.3610e-04, + 5.9843e-05, 6.4433e-05], device='cuda:0') +588 +0.000954915028125264 +changing lr +epoch 56, time 844.10, cls_loss 0.0005 cls_loss_mapping 0.0051 cls_loss_causal 0.3976 re_mapping 0.0060 re_causal 0.0146 /// teacc 93.47 lr 0.00082713 +Epoch 58, weight, value: tensor([[ 0.2804, 0.2624, 0.2444, ..., -0.0168, -0.0170, -0.0193], + [-0.0070, -0.0458, 0.0077, ..., -0.0367, -0.0059, -0.0086], + [-0.0823, -0.0654, -0.1020, ..., 0.0092, -0.0079, -0.0052], + ..., + [-0.1914, -0.1745, -0.1544, ..., 0.0222, 0.0087, 0.0109], + [-0.0079, 0.0019, -0.0014, ..., 0.0130, 0.0121, 0.0016], + [ 0.0143, 0.0206, 0.0116, ..., -0.0323, -0.0369, -0.0297]], + device='cuda:0'), grad: tensor([[-4.5568e-05, -7.2658e-05, -7.1228e-05, ..., 1.8388e-05, + 1.6898e-05, 1.6958e-05], + [ 5.1498e-05, 3.2544e-05, 3.2544e-05, ..., 2.2873e-05, + 1.7777e-05, 1.9237e-05], + [-3.5372e-06, 2.7835e-05, 3.1590e-05, ..., 1.2621e-05, + 8.9929e-06, 8.4043e-06], + ..., + [ 2.6226e-04, 9.2864e-05, 9.4056e-05, ..., 6.2764e-05, + 5.7697e-05, 5.8204e-05], + [-7.9155e-04, -2.4867e-04, -2.5606e-04, ..., -2.5034e-04, + -2.2256e-04, -2.2459e-04], + [ 2.7633e-04, 9.3520e-05, 9.5129e-05, ..., 7.5698e-05, + 6.8307e-05, 6.8665e-05]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0807, 0.0064, -0.0127, -0.0724, 0.0060, 0.0010, 0.0034], + device='cuda:0'), grad: tensor([ 8.4877e-05, 3.3945e-05, -1.0407e-04, 5.8937e-04, 5.8365e-04, + -1.8044e-03, 6.1512e-04], device='cuda:0') +588 +0.0008271337313934874 +changing lr +epoch 57, time 831.80, cls_loss 0.0005 cls_loss_mapping 0.0055 cls_loss_causal 0.3892 re_mapping 0.0060 re_causal 0.0147 /// teacc 92.46 lr 0.00070776 +Epoch 59, weight, value: tensor([[ 0.2802, 0.2623, 0.2443, ..., -0.0168, -0.0170, -0.0193], + [-0.0070, -0.0457, 0.0076, ..., -0.0366, -0.0058, -0.0086], + [-0.0823, -0.0655, -0.1021, ..., 0.0091, -0.0079, -0.0053], + ..., + [-0.1911, -0.1743, -0.1543, ..., 0.0222, 0.0087, 0.0110], + [-0.0079, 0.0019, -0.0014, ..., 0.0130, 0.0121, 0.0016], + [ 0.0142, 0.0205, 0.0115, ..., -0.0323, -0.0368, -0.0297]], + device='cuda:0'), grad: tensor([[-1.4458e-03, -8.4400e-04, -8.3685e-04, ..., -2.1994e-04, + -2.1100e-04, -1.9038e-04], + [ 7.4434e-04, 4.0364e-04, 4.0793e-04, ..., 1.2231e-04, + 1.1736e-04, 1.0604e-04], + [-1.0234e-04, 3.4750e-05, 3.5793e-05, ..., -3.9726e-05, + -3.6955e-05, -3.5733e-05], + ..., + [ 2.8872e-04, 1.7309e-04, 1.6582e-04, ..., 4.2737e-05, + 4.0442e-05, 3.6955e-05], + [ 1.0991e-04, 5.7966e-05, 5.8204e-05, ..., 1.8507e-05, + 1.7703e-05, 1.6078e-05], + [ 1.6439e-04, 8.2612e-05, 8.2433e-05, ..., 2.8223e-05, + 2.7031e-05, 2.4632e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0805, 0.0063, -0.0128, -0.0723, 0.0063, 0.0009, 0.0034], + device='cuda:0'), grad: tensor([-0.0023, 0.0012, -0.0004, 0.0005, 0.0005, 0.0002, 0.0003], + device='cuda:0') +588 +0.00070775603199067 +changing lr +epoch 58, time 819.14, cls_loss 0.0005 cls_loss_mapping 0.0055 cls_loss_causal 0.3894 re_mapping 0.0059 re_causal 0.0143 /// teacc 91.46 lr 0.00059702 +Epoch 60, weight, value: tensor([[ 0.2803, 0.2623, 0.2443, ..., -0.0168, -0.0170, -0.0193], + [-0.0072, -0.0457, 0.0075, ..., -0.0365, -0.0058, -0.0086], + [-0.0823, -0.0655, -0.1020, ..., 0.0090, -0.0079, -0.0053], + ..., + [-0.1909, -0.1741, -0.1542, ..., 0.0222, 0.0087, 0.0110], + [-0.0080, 0.0018, -0.0015, ..., 0.0130, 0.0121, 0.0016], + [ 0.0142, 0.0204, 0.0115, ..., -0.0322, -0.0368, -0.0297]], + device='cuda:0'), grad: tensor([[-6.7787e-03, -3.0537e-03, -2.9202e-03, ..., -9.1457e-04, + -1.1253e-03, -9.9182e-04], + [ 1.0204e-03, 4.7708e-04, 4.9448e-04, ..., 1.5736e-04, + 1.8883e-04, 1.7345e-04], + [ 1.6956e-03, 6.5470e-04, 6.0511e-04, ..., 2.5892e-04, + 3.0756e-04, 2.7180e-04], + ..., + [ 1.0824e-03, 5.5647e-04, 5.0831e-04, ..., 8.0526e-05, + 1.2136e-04, 9.6023e-05], + [ 4.1556e-04, 1.7333e-04, 1.5962e-04, ..., 5.4002e-05, + 6.7174e-05, 5.8085e-05], + [ 8.7690e-04, 3.6573e-04, 3.3498e-04, ..., 1.1265e-04, + 1.4067e-04, 1.2106e-04]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0807, 0.0062, -0.0126, -0.0724, 0.0061, 0.0008, 0.0035], + device='cuda:0'), grad: tensor([-0.0133, 0.0019, 0.0037, 0.0031, 0.0019, 0.0009, 0.0018], + device='cuda:0') +588 +0.0005970223407163104 +changing lr +epoch 59, time 814.50, cls_loss 0.0007 cls_loss_mapping 0.0059 cls_loss_causal 0.3381 re_mapping 0.0059 re_causal 0.0138 /// teacc 92.71 lr 0.00049516 +Epoch 61, weight, value: tensor([[ 0.2802, 0.2622, 0.2443, ..., -0.0169, -0.0171, -0.0194], + [-0.0074, -0.0458, 0.0074, ..., -0.0365, -0.0059, -0.0086], + [-0.0823, -0.0655, -0.1020, ..., 0.0090, -0.0080, -0.0053], + ..., + [-0.1907, -0.1740, -0.1541, ..., 0.0222, 0.0088, 0.0110], + [-0.0080, 0.0018, -0.0015, ..., 0.0130, 0.0121, 0.0016], + [ 0.0144, 0.0204, 0.0115, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[-1.6832e-03, -1.0757e-03, -8.9216e-04, ..., -3.7283e-05, + -6.5982e-05, -5.0902e-05], + [ 3.9601e-04, 1.5604e-04, 1.4305e-04, ..., 1.2177e-04, + 1.1653e-04, 1.0395e-04], + [-1.2188e-03, -9.0659e-05, -9.7632e-05, ..., -8.6641e-04, + -8.2254e-04, -7.1621e-04], + ..., + [ 1.4887e-03, 6.0654e-04, 5.1260e-04, ..., 5.6934e-04, + 5.6267e-04, 4.7302e-04], + [ 1.3816e-04, 7.1645e-05, 6.0081e-05, ..., 1.7062e-05, + 1.7494e-05, 1.5557e-05], + [ 4.3225e-04, 1.9133e-04, 1.5461e-04, ..., 5.9456e-05, + 5.9456e-05, 5.4836e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0807, 0.0056, -0.0127, -0.0725, 0.0064, 0.0007, 0.0040], + device='cuda:0'), grad: tensor([-0.0024, 0.0009, -0.0043, 0.0011, 0.0036, 0.0003, 0.0009], + device='cuda:0') +588 +0.0004951556604879052 +changing lr +epoch 60, time 811.20, cls_loss 0.0007 cls_loss_mapping 0.0055 cls_loss_causal 0.3530 re_mapping 0.0060 re_causal 0.0139 /// teacc 91.71 lr 0.00040236 +Epoch 62, weight, value: tensor([[ 0.2804, 0.2622, 0.2444, ..., -0.0169, -0.0171, -0.0193], + [-0.0076, -0.0458, 0.0073, ..., -0.0364, -0.0059, -0.0086], + [-0.0824, -0.0656, -0.1020, ..., 0.0089, -0.0080, -0.0054], + ..., + [-0.1906, -0.1739, -0.1540, ..., 0.0222, 0.0088, 0.0110], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0143, 0.0204, 0.0114, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[-1.9569e-03, -8.9359e-04, -9.8228e-04, ..., -5.1689e-04, + -4.9973e-04, -4.8327e-04], + [ 1.1933e-04, 5.8621e-05, 6.2406e-05, ..., 3.1203e-05, + 3.0220e-05, 2.9534e-05], + [ 6.9666e-04, 3.0160e-04, 3.2949e-04, ..., 1.8990e-04, + 1.8394e-04, 1.7786e-04], + ..., + [ 2.5058e-04, 1.2970e-04, 1.4579e-04, ..., 4.7624e-05, + 4.6819e-05, 4.4048e-05], + [ 1.2529e-04, 5.3614e-05, 5.8681e-05, ..., 3.2544e-05, + 3.1888e-05, 3.0667e-05], + [ 4.9889e-05, 3.1292e-05, 3.4332e-05, ..., 2.0638e-05, + 1.8150e-05, 1.8433e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0810, 0.0054, -0.0129, -0.0724, 0.0061, 0.0014, 0.0038], + device='cuda:0'), grad: tensor([-3.7651e-03, 2.2376e-04, 1.3990e-03, 1.4057e-03, 4.2057e-04, + 2.5535e-04, 6.1870e-05], device='cuda:0') +588 +0.00040236113724274745 +changing lr +epoch 61, time 807.28, cls_loss 0.0004 cls_loss_mapping 0.0041 cls_loss_causal 0.3915 re_mapping 0.0059 re_causal 0.0142 /// teacc 93.72 lr 0.00031883 +Epoch 63, weight, value: tensor([[ 0.2802, 0.2622, 0.2443, ..., -0.0169, -0.0171, -0.0194], + [-0.0075, -0.0458, 0.0073, ..., -0.0364, -0.0059, -0.0086], + [-0.0823, -0.0656, -0.1020, ..., 0.0089, -0.0080, -0.0054], + ..., + [-0.1905, -0.1738, -0.1539, ..., 0.0222, 0.0088, 0.0110], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0142, 0.0203, 0.0114, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[ 4.9710e-05, 1.4678e-05, 1.5251e-05, ..., 1.3709e-05, + 1.3046e-05, 1.2688e-05], + [ 2.0742e-05, -6.0797e-06, -5.1297e-06, ..., 1.1548e-05, + 1.1817e-05, 1.0729e-05], + [-1.8489e-04, -3.1650e-05, -3.4392e-05, ..., -7.3910e-05, + -7.2956e-05, -6.9559e-05], + ..., + [ 3.3647e-05, 7.4916e-06, 8.0690e-06, ..., 1.1712e-05, + 1.1392e-05, 1.0930e-05], + [-6.0588e-05, -1.1645e-05, -1.4849e-05, ..., -8.4117e-06, + -7.7188e-06, -6.9439e-06], + [ 5.7191e-05, 1.1474e-05, 1.3858e-05, ..., 1.1086e-05, + 1.0386e-05, 9.6411e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0808, 0.0055, -0.0127, -0.0725, 0.0062, 0.0014, 0.0036], + device='cuda:0'), grad: tensor([ 1.2708e-04, 1.1134e-04, -5.8222e-04, 2.5630e-04, 9.6560e-05, + -1.7190e-04, 1.6367e-04], device='cuda:0') +588 +0.00031882564680131423 +changing lr +epoch 62, time 807.95, cls_loss 0.0005 cls_loss_mapping 0.0044 cls_loss_causal 0.3894 re_mapping 0.0058 re_causal 0.0142 /// teacc 93.47 lr 0.00024472 +Epoch 64, weight, value: tensor([[ 0.2801, 0.2621, 0.2443, ..., -0.0169, -0.0171, -0.0194], + [-0.0075, -0.0457, 0.0073, ..., -0.0364, -0.0059, -0.0086], + [-0.0823, -0.0656, -0.1020, ..., 0.0089, -0.0080, -0.0054], + ..., + [-0.1903, -0.1737, -0.1539, ..., 0.0222, 0.0088, 0.0111], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0203, 0.0114, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[ 1.5706e-05, -2.1428e-05, -1.8954e-05, ..., 1.9580e-05, + 2.1145e-05, 1.9655e-05], + [ 5.5850e-05, 2.8703e-06, 4.4145e-06, ..., 2.7135e-05, + 3.0041e-05, 2.7075e-05], + [ 1.0788e-05, 9.5218e-06, 9.8944e-06, ..., 8.3670e-06, + 7.8157e-06, 8.6278e-06], + ..., + [-1.0562e-04, -5.1782e-07, -1.4892e-06, ..., -5.6863e-05, + -6.1333e-05, -5.6207e-05], + [ 4.4182e-06, -1.6177e-06, -1.5702e-06, ..., 1.6177e-06, + 1.7490e-06, 1.6317e-06], + [ 6.3241e-05, 8.4937e-06, 1.1571e-05, ..., 2.5243e-05, + 2.7627e-05, 2.5675e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0806, 0.0057, -0.0127, -0.0726, 0.0063, 0.0014, 0.0035], + device='cuda:0'), grad: tensor([ 1.4091e-04, 1.9705e-04, 2.5444e-06, -1.6570e-04, -3.9673e-04, + 2.3782e-05, 1.9848e-04], device='cuda:0') +588 +0.0002447174185242325 +changing lr +epoch 63, time 800.79, cls_loss 0.0005 cls_loss_mapping 0.0043 cls_loss_causal 0.3645 re_mapping 0.0058 re_causal 0.0137 /// teacc 92.96 lr 0.00018019 +Epoch 65, weight, value: tensor([[ 0.2800, 0.2620, 0.2442, ..., -0.0170, -0.0172, -0.0194], + [-0.0075, -0.0457, 0.0073, ..., -0.0363, -0.0058, -0.0086], + [-0.0823, -0.0656, -0.1019, ..., 0.0089, -0.0080, -0.0054], + ..., + [-0.1903, -0.1736, -0.1538, ..., 0.0222, 0.0089, 0.0111], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0203, 0.0114, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[ 2.3663e-04, 6.5327e-05, 7.1287e-05, ..., 3.8981e-05, + 3.8862e-05, 3.2604e-05], + [-3.3331e-04, -6.7234e-05, -7.2896e-05, ..., -5.2780e-05, + -4.8369e-05, -3.5584e-05], + [ 1.0020e-04, 2.4155e-05, 2.6375e-05, ..., 1.6272e-05, + 1.6153e-05, 1.2092e-05], + ..., + [ 3.8803e-05, 1.9759e-05, 2.1338e-05, ..., 1.2852e-05, + 9.9316e-06, 9.9391e-06], + [ 3.4213e-05, 9.4250e-06, 1.0446e-05, ..., 6.3889e-06, + 6.6273e-06, 5.7556e-06], + [-5.7310e-05, -3.6985e-05, -3.8534e-05, ..., 2.3283e-06, + 1.1045e-06, 2.2212e-07]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0804, 0.0058, -0.0127, -0.0726, 0.0064, 0.0015, 0.0035], + device='cuda:0'), grad: tensor([ 5.6124e-04, -8.7214e-04, 2.4772e-04, -8.1211e-06, 5.9217e-05, + 8.1003e-05, -6.9082e-05], device='cuda:0') +588 +0.0001801856965207339 +changing lr +epoch 64, time 806.57, cls_loss 0.0005 cls_loss_mapping 0.0039 cls_loss_causal 0.3887 re_mapping 0.0059 re_causal 0.0144 /// teacc 92.21 lr 0.00012536 +Epoch 66, weight, value: tensor([[ 0.2799, 0.2619, 0.2442, ..., -0.0170, -0.0172, -0.0194], + [-0.0074, -0.0457, 0.0073, ..., -0.0363, -0.0058, -0.0086], + [-0.0823, -0.0656, -0.1019, ..., 0.0089, -0.0080, -0.0054], + ..., + [-0.1902, -0.1736, -0.1538, ..., 0.0223, 0.0089, 0.0111], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0202, 0.0113, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[-7.4208e-05, -4.9382e-05, -4.5568e-05, ..., -2.7120e-06, + -2.4829e-06, -2.1067e-06], + [ 7.6234e-05, 3.8773e-05, 4.0531e-05, ..., 2.2858e-05, + 2.4304e-05, 2.1905e-05], + [ 9.8720e-06, 6.0499e-06, 6.1169e-06, ..., 2.6654e-06, + 2.5295e-06, 2.2203e-06], + ..., + [ 3.4630e-05, 1.5914e-05, 1.6615e-05, ..., 1.1235e-05, + 1.2189e-05, 1.0826e-05], + [ 6.4336e-06, 3.5278e-06, 3.4962e-06, ..., 1.3625e-06, + 1.4864e-06, 1.3262e-06], + [ 1.7092e-05, 9.9391e-06, 9.7305e-06, ..., 2.8815e-06, + 3.1013e-06, 2.7642e-06]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0803, 0.0058, -0.0127, -0.0726, 0.0065, 0.0015, 0.0035], + device='cuda:0'), grad: tensor([-9.6381e-05, 1.2600e-04, 9.8869e-06, -1.4079e-04, 6.4850e-05, + 1.0490e-05, 2.5824e-05], device='cuda:0') +588 +0.000125360439090882 +changing lr +epoch 65, time 808.78, cls_loss 0.0005 cls_loss_mapping 0.0045 cls_loss_causal 0.3667 re_mapping 0.0059 re_causal 0.0134 /// teacc 93.22 lr 0.00008035 +Epoch 67, weight, value: tensor([[ 0.2798, 0.2619, 0.2441, ..., -0.0170, -0.0172, -0.0195], + [-0.0074, -0.0456, 0.0073, ..., -0.0363, -0.0058, -0.0086], + [-0.0823, -0.0656, -0.1019, ..., 0.0088, -0.0080, -0.0054], + ..., + [-0.1901, -0.1735, -0.1538, ..., 0.0223, 0.0089, 0.0111], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0202, 0.0113, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[ 4.2648e-03, 2.3117e-03, 2.3460e-03, ..., 2.6727e-04, + 3.2973e-04, 3.4213e-04], + [ 4.3821e-04, 1.2589e-04, 1.3447e-04, ..., 6.1989e-05, + 6.6161e-05, 6.2048e-05], + [-3.0971e-04, -1.3430e-06, -2.4870e-05, ..., -1.4102e-04, + -1.4400e-04, -1.3316e-04], + ..., + [ 3.3736e-04, 1.5569e-04, 1.6546e-04, ..., 3.4511e-05, + 3.9518e-05, 3.8922e-05], + [ 8.9049e-05, 2.9191e-05, 3.1352e-05, ..., 1.4775e-05, + 1.5661e-05, 1.4886e-05], + [-4.9210e-03, -2.6112e-03, -2.6436e-03, ..., -2.6274e-04, + -3.3402e-04, -3.4785e-04]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0802, 0.0059, -0.0127, -0.0725, 0.0065, 0.0014, 0.0035], + device='cuda:0'), grad: tensor([ 0.0064, 0.0010, -0.0012, 0.0004, 0.0006, 0.0002, -0.0075], + device='cuda:0') +588 +8.03520570068517e-05 +changing lr +epoch 66, time 805.21, cls_loss 0.0005 cls_loss_mapping 0.0043 cls_loss_causal 0.3723 re_mapping 0.0059 re_causal 0.0138 /// teacc 93.47 lr 0.00004525 +Epoch 68, weight, value: tensor([[ 0.2798, 0.2619, 0.2441, ..., -0.0170, -0.0172, -0.0195], + [-0.0074, -0.0456, 0.0073, ..., -0.0363, -0.0058, -0.0086], + [-0.0823, -0.0656, -0.1019, ..., 0.0088, -0.0080, -0.0054], + ..., + [-0.1901, -0.1735, -0.1537, ..., 0.0223, 0.0089, 0.0111], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0202, 0.0113, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[-2.5034e-04, -1.6713e-04, -1.6880e-04, ..., -1.0543e-05, + -1.1466e-05, -1.0572e-05], + [-5.5671e-05, -2.0284e-06, -1.7649e-06, ..., -1.5542e-05, + -1.5691e-05, -1.5885e-05], + [ 5.1588e-05, 5.4836e-05, 5.5194e-05, ..., -1.1787e-05, + -1.4707e-05, -1.1392e-05], + ..., + [ 1.1039e-04, 4.1306e-05, 4.2021e-05, ..., 2.2307e-05, + 2.4095e-05, 2.1860e-05], + [ 3.7134e-05, 2.1636e-05, 2.1875e-05, ..., 3.4645e-06, + 3.7570e-06, 3.4906e-06], + [ 4.5300e-05, 2.1130e-05, 2.1294e-05, ..., 6.8657e-06, + 7.4469e-06, 7.0930e-06]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0802, 0.0058, -0.0127, -0.0725, 0.0065, 0.0014, 0.0035], + device='cuda:0'), grad: tensor([-2.8110e-04, -2.2256e-04, -3.5822e-05, 1.2255e-04, 2.6083e-04, + 5.7101e-05, 9.9182e-05], device='cuda:0') +588 +4.5251191160326525e-05 +changing lr +epoch 67, time 806.20, cls_loss 0.0005 cls_loss_mapping 0.0048 cls_loss_causal 0.3926 re_mapping 0.0058 re_causal 0.0139 /// teacc 93.22 lr 0.00002013 +Epoch 69, weight, value: tensor([[ 0.2798, 0.2619, 0.2441, ..., -0.0170, -0.0172, -0.0195], + [-0.0074, -0.0456, 0.0073, ..., -0.0363, -0.0058, -0.0086], + [-0.0823, -0.0656, -0.1019, ..., 0.0088, -0.0080, -0.0054], + ..., + [-0.1901, -0.1735, -0.1537, ..., 0.0223, 0.0089, 0.0111], + [-0.0078, 0.0018, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0202, 0.0113, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[-1.7926e-05, -4.2558e-05, -4.1962e-05, ..., 2.8610e-05, + 2.9489e-05, 2.8580e-05], + [ 9.9689e-06, 5.4874e-06, 4.2394e-06, ..., 9.0972e-06, + 1.0937e-05, 9.6411e-06], + [ 7.0333e-05, 3.0875e-05, 3.1590e-05, ..., 2.4319e-05, + 2.4244e-05, 2.3410e-05], + ..., + [-9.7990e-05, -1.0200e-05, -7.7561e-06, ..., -5.3912e-05, + -5.5939e-05, -5.3674e-05], + [ 2.0862e-05, 9.3058e-06, 9.2313e-06, ..., 5.4464e-06, + 5.5470e-06, 5.2899e-06], + [ 3.7640e-05, 1.5102e-05, 1.5378e-05, ..., 6.6869e-06, + 5.9530e-06, 6.1281e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0802, 0.0058, -0.0127, -0.0725, 0.0066, 0.0014, 0.0034], + device='cuda:0'), grad: tensor([ 9.5785e-05, 1.4924e-05, 1.3804e-04, -4.7803e-05, -3.2473e-04, + 4.1425e-05, 8.2016e-05], device='cuda:0') +588 +2.0128530023804673e-05 +changing lr +epoch 68, time 806.78, cls_loss 0.0005 cls_loss_mapping 0.0051 cls_loss_causal 0.3597 re_mapping 0.0059 re_causal 0.0134 /// teacc 93.22 lr 0.00000503 +Epoch 70, weight, value: tensor([[ 0.2798, 0.2619, 0.2441, ..., -0.0170, -0.0172, -0.0195], + [-0.0074, -0.0456, 0.0073, ..., -0.0363, -0.0058, -0.0086], + [-0.0823, -0.0656, -0.1019, ..., 0.0088, -0.0080, -0.0054], + ..., + [-0.1901, -0.1735, -0.1537, ..., 0.0223, 0.0089, 0.0111], + [-0.0078, 0.0017, -0.0015, ..., 0.0130, 0.0122, 0.0017], + [ 0.0141, 0.0202, 0.0113, ..., -0.0322, -0.0367, -0.0296]], + device='cuda:0'), grad: tensor([[-6.9809e-04, -4.3750e-04, -4.5300e-04, ..., -6.7949e-05, + -6.5744e-05, -5.2512e-05], + [ 3.1948e-04, 1.9705e-04, 2.0468e-04, ..., 3.2932e-05, + 3.0845e-05, 2.4393e-05], + [ 1.3494e-04, 8.6308e-05, 9.0063e-05, ..., 1.5706e-05, + 1.4812e-05, 1.2159e-05], + ..., + [ 1.2732e-04, 7.5877e-05, 7.6950e-05, ..., 1.5974e-05, + 1.6287e-05, 1.4447e-05], + [ 4.7207e-05, 2.7984e-05, 2.8849e-05, ..., 6.9700e-06, + 6.7130e-06, 5.8748e-06], + [ 8.1778e-05, 4.8131e-05, 4.9859e-05, ..., 1.2770e-05, + 1.2040e-05, 1.0483e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0802, 0.0058, -0.0127, -0.0725, 0.0066, 0.0014, 0.0034], + device='cuda:0'), grad: tensor([-9.3508e-04, 4.4107e-04, 1.7381e-04, -6.0767e-05, 1.8752e-04, + 6.9916e-05, 1.2267e-04], device='cuda:0') +588 +5.034667293427056e-06 +changing lr +epoch 69, time 805.22, cls_loss 0.0004 cls_loss_mapping 0.0041 cls_loss_causal 0.3673 re_mapping 0.0059 re_causal 0.0137 /// teacc 92.21 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/sketch_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.516416 55.810547 64.974403 56.886228 59.223726 + sketch art_painting cartoon photo Avg +do 99.414609 48.144531 61.476109 53.353293 54.324645 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.363706 57.519531 66.552901 60.239521 61.437318 + sketch art_painting cartoon photo Avg +do 99.440061 56.103516 65.443686 59.281437 60.276213 diff --git a/Meta-causal/code-withStyleAttack/73728.error b/Meta-causal/code-withStyleAttack/73728.error new file mode 100644 index 0000000000000000000000000000000000000000..d4cae9c14f1b02ce78b5efc4ecc9cba1f06923fb --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73728.error @@ -0,0 +1,18 @@ +Solving dependencies +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +run_my_joint_v13_test.sh: line 34: andm: command not found +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:44: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:58: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:68: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:47: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:61: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:71: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) diff --git a/Meta-causal/code-withStyleAttack/73728.log b/Meta-causal/code-withStyleAttack/73728.log new file mode 100644 index 0000000000000000000000000000000000000000..b8bf2a30c4348037db6af835714e9465cd246d8c --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73728.log @@ -0,0 +1,1974 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[ 1.8066e-02, -1.4616e-02, -5.2350e-03, ..., -1.3573e-02, + -5.3800e-03, 8.4514e-03], + [-2.9903e-03, -3.7094e-03, 6.7388e-04, ..., 1.3490e-02, + 1.7055e-02, -9.0759e-03], + [ 6.4058e-03, -1.5238e-02, 2.4516e-04, ..., -1.4837e-02, + 1.9578e-03, -1.6192e-02], + ..., + [-2.0085e-02, -1.5276e-02, -1.6557e-02, ..., -9.0591e-03, + -3.4042e-03, 6.4908e-03], + [-1.9666e-02, -9.1805e-04, 1.4554e-02, ..., -4.7549e-03, + -8.1075e-05, -1.6379e-02], + [-2.2419e-03, -1.7907e-02, 1.3829e-02, ..., 1.4124e-03, + -7.4572e-03, 1.0296e-02]], device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0050, -0.0021, -0.0018, -0.0059, 0.0006, -0.0049, 0.0084], + device='cuda:0'), grad: None +588 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 777.99, cls_loss 4.3509 cls_loss_mapping 1.1279 cls_loss_causal 1.4185 re_mapping 0.4384 re_causal 0.4385 /// teacc 84.92 lr 0.00999497 +Epoch 2, weight, value: tensor([[-0.0641, -0.0505, -0.1411, ..., 0.0175, 0.0704, 0.0477], + [ 0.2240, 0.1512, 0.2202, ..., 0.0486, 0.0596, 0.0267], + [-0.0330, -0.0519, -0.0591, ..., -0.0819, -0.1133, -0.1124], + ..., + [-0.2529, -0.2231, -0.1790, ..., 0.0475, 0.0563, 0.0592], + [ 0.1038, 0.1071, 0.1337, ..., -0.1160, -0.1088, -0.1196], + [ 0.0398, 0.0373, 0.0445, ..., -0.0132, -0.0174, -0.0009]], + device='cuda:0'), grad: tensor([[ 4.2236e-02, 8.0948e-03, -1.1368e-02, ..., 6.8970e-02, + 6.2683e-02, 3.6743e-02], + [ 9.6375e-02, 6.5552e-02, 3.6011e-02, ..., 3.9551e-02, + 3.5309e-02, 1.4694e-02], + [-7.8613e-02, -4.2023e-02, -6.4964e-03, ..., -6.7871e-02, + -5.6824e-02, -3.2562e-02], + ..., + [-6.1005e-02, -3.2318e-02, -1.8494e-02, ..., -4.1046e-02, + -4.1504e-02, -1.8982e-02], + [ 5.3316e-05, 3.5554e-05, 2.1845e-05, ..., 1.8999e-05, + 1.8522e-05, 6.3777e-06], + [ 3.9130e-05, 2.6524e-05, 1.4313e-05, ..., 1.3396e-05, + 1.3120e-05, 4.1462e-06]], device='cuda:0') +Epoch 2, bias, value: tensor([ 3.1027e-02, -1.2553e-02, -3.6311e-02, -5.0901e-02, 1.2749e-02, + -8.4003e-05, 4.5286e-02], device='cuda:0'), grad: tensor([ 1.3892e-01, 1.2219e-01, -1.4893e-01, 1.3962e-03, -1.1365e-01, + 7.8976e-05, 6.0230e-05], device='cuda:0') +588 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 771.45, cls_loss 0.5868 cls_loss_mapping 0.4660 cls_loss_causal 1.0081 re_mapping 0.1610 re_causal 0.1604 /// teacc 91.96 lr 0.00997987 +Epoch 3, weight, value: tensor([[-0.0646, -0.0589, -0.1475, ..., 0.0246, 0.0791, 0.0600], + [ 0.2646, 0.1994, 0.2629, ..., 0.0578, 0.0666, 0.0365], + [-0.0793, -0.0915, -0.0975, ..., -0.0850, -0.1109, -0.1113], + ..., + [-0.2961, -0.2595, -0.2078, ..., 0.0233, 0.0273, 0.0278], + [ 0.1148, 0.1168, 0.1435, ..., -0.1130, -0.1042, -0.1159], + [ 0.0684, 0.0600, 0.0581, ..., -0.0140, -0.0198, -0.0034]], + device='cuda:0'), grad: tensor([[ 3.9246e-02, 1.4778e-02, 6.1378e-03, ..., 1.3214e-02, + 1.5762e-02, 6.4125e-03], + [-5.2460e-02, -2.0218e-02, -8.4991e-03, ..., -1.7319e-02, + -2.0676e-02, -8.2321e-03], + [ 7.7963e-04, 3.2425e-04, 1.3626e-04, ..., 2.9707e-04, + 3.2449e-04, 1.5128e-04], + ..., + [ 8.3389e-03, 3.5267e-03, 1.5554e-03, ..., 2.5043e-03, + 3.0041e-03, 1.0738e-03], + [ 3.8471e-03, 1.4982e-03, 6.3276e-04, ..., 1.2541e-03, + 1.4982e-03, 5.8889e-04], + [ 3.0518e-04, 1.1790e-04, 4.9353e-05, ..., 1.0031e-04, + 1.1981e-04, 4.7386e-05]], device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0393, -0.0130, -0.0661, -0.0372, 0.0051, 0.0004, 0.0611], + device='cuda:0'), grad: tensor([ 8.4473e-02, -1.1200e-01, 1.5860e-03, 1.8403e-05, 1.7105e-02, + 8.1711e-03, 6.5136e-04], device='cuda:0') +588 +0.009979871469976196 +changing lr +epoch 2, time 778.16, cls_loss 0.2506 cls_loss_mapping 0.2656 cls_loss_causal 0.8203 re_mapping 0.1303 re_causal 0.1296 /// teacc 86.68 lr 0.00995475 +Epoch 4, weight, value: tensor([[-0.0672, -0.0718, -0.1605, ..., 0.0241, 0.0755, 0.0530], + [ 0.2772, 0.2214, 0.2872, ..., 0.0624, 0.0698, 0.0431], + [-0.0532, -0.0682, -0.0817, ..., -0.0827, -0.1068, -0.1117], + ..., + [-0.3240, -0.2869, -0.2331, ..., 0.0108, 0.0187, 0.0220], + [ 0.1108, 0.1148, 0.1421, ..., -0.1111, -0.1019, -0.1128], + [ 0.0640, 0.0548, 0.0526, ..., -0.0097, -0.0166, -0.0010]], + device='cuda:0'), grad: tensor([[ 5.6601e-04, 3.8314e-04, 3.0184e-04, ..., 2.3842e-04, + 2.6822e-04, 1.9336e-04], + [-1.0973e-04, -8.6367e-05, -7.6056e-05, ..., -5.5619e-06, + -4.4368e-06, 3.6992e-06], + [ 2.3746e-03, 1.5621e-03, 1.2035e-03, ..., 1.1530e-03, + 1.3189e-03, 9.7084e-04], + ..., + [ 3.1412e-05, 1.7956e-05, 1.1876e-05, ..., 2.3767e-05, + 2.5392e-05, 2.0787e-05], + [-6.7838e-06, -5.1297e-06, -3.8780e-06, ..., -3.8091e-07, + -4.7777e-07, 6.4261e-08], + [ 1.0364e-05, 6.9141e-06, 5.0589e-06, ..., 4.0941e-06, + 4.5300e-06, 3.2634e-06]], device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0697, -0.0210, -0.0571, -0.0513, -0.0050, -0.0074, 0.0618], + device='cuda:0'), grad: tensor([ 7.2241e-04, -1.1706e-04, 3.0746e-03, -3.7327e-03, 5.1528e-05, + -6.3702e-06, 1.3195e-05], device='cuda:0') +588 +0.009954748808839675 +changing lr +epoch 3, time 778.29, cls_loss 0.1132 cls_loss_mapping 0.1794 cls_loss_causal 0.7350 re_mapping 0.1117 re_causal 0.1117 /// teacc 89.45 lr 0.00991965 +Epoch 5, weight, value: tensor([[-0.0509, -0.0596, -0.1472, ..., 0.0416, 0.0907, 0.0689], + [ 0.2854, 0.2319, 0.2971, ..., 0.0610, 0.0693, 0.0421], + [-0.0703, -0.0823, -0.0959, ..., -0.0906, -0.1146, -0.1192], + ..., + [-0.3450, -0.3093, -0.2570, ..., -0.0020, 0.0058, 0.0091], + [ 0.1094, 0.1143, 0.1412, ..., -0.1094, -0.1004, -0.1109], + [ 0.0732, 0.0634, 0.0606, ..., -0.0076, -0.0145, 0.0011]], + device='cuda:0'), grad: tensor([[-2.2495e-04, -7.9691e-05, -3.3528e-05, ..., -8.6010e-05, + -1.0067e-04, -6.0201e-05], + [ 1.0490e-05, -4.1164e-06, -8.4937e-06, ..., 6.0946e-06, + 8.0168e-06, 4.5225e-06], + [-3.0883e-06, -1.2498e-06, 4.3213e-07, ..., -7.8091e-07, + 1.1278e-06, 1.1064e-06], + ..., + [ 1.4758e-04, 5.5283e-05, 2.5570e-05, ..., 5.6654e-05, + 6.3896e-05, 3.8058e-05], + [ 9.3132e-06, 3.5204e-06, 1.6363e-06, ..., 3.4254e-06, + 4.0121e-06, 2.4028e-06], + [ 5.6297e-05, 2.4781e-05, 1.3664e-05, ..., 1.8552e-05, + 2.1458e-05, 1.2845e-05]], device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0699, -0.0210, -0.0638, -0.0557, 0.0103, -0.0102, 0.0605], + device='cuda:0'), grad: tensor([-5.7936e-04, 5.4985e-05, -5.3085e-06, 1.2010e-05, 3.7050e-04, + 2.3007e-05, 1.2362e-04], device='cuda:0') +588 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 787.89, cls_loss 0.0795 cls_loss_mapping 0.1361 cls_loss_causal 0.6984 re_mapping 0.0997 re_causal 0.1004 /// teacc 92.21 lr 0.00987464 +Epoch 6, weight, value: tensor([[-0.0650, -0.0745, -0.1608, ..., 0.0434, 0.0951, 0.0720], + [ 0.3094, 0.2547, 0.3183, ..., 0.0604, 0.0678, 0.0418], + [-0.0780, -0.0862, -0.0994, ..., -0.0866, -0.1113, -0.1157], + ..., + [-0.3432, -0.3108, -0.2608, ..., -0.0084, -0.0020, 0.0019], + [ 0.1100, 0.1147, 0.1405, ..., -0.1070, -0.0981, -0.1083], + [ 0.0684, 0.0621, 0.0612, ..., -0.0102, -0.0181, -0.0020]], + device='cuda:0'), grad: tensor([[-8.9539e-02, -3.6102e-02, -1.8234e-02, ..., -2.8183e-02, + -3.8666e-02, -3.2867e-02], + [ 3.6438e-02, 1.4847e-02, 7.5455e-03, ..., 1.1536e-02, + 1.5747e-02, 1.3451e-02], + [ 5.9204e-03, 2.2964e-03, 1.1244e-03, ..., 1.8950e-03, + 2.5864e-03, 2.1667e-03], + ..., + [ 4.5685e-02, 1.8448e-02, 9.3689e-03, ..., 1.4359e-02, + 1.9745e-02, 1.6754e-02], + [ 3.8528e-04, 1.5247e-04, 8.0585e-05, ..., 1.2076e-04, + 1.6952e-04, 1.4007e-04], + [ 4.9305e-04, 1.9550e-04, 9.8288e-05, ..., 1.5604e-04, + 2.1434e-04, 1.8048e-04]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.0739, -0.0085, -0.0726, -0.0523, 0.0149, -0.0101, 0.0451], + device='cuda:0'), grad: tensor([-0.1722, 0.0696, 0.0117, 0.0014, 0.0879, 0.0008, 0.0010], + device='cuda:0') +588 +0.009874639560909117 +changing lr +epoch 5, time 778.76, cls_loss 0.0530 cls_loss_mapping 0.1098 cls_loss_causal 0.6454 re_mapping 0.0894 re_causal 0.0910 /// teacc 91.96 lr 0.00981981 +Epoch 7, weight, value: tensor([[-0.0598, -0.0710, -0.1576, ..., 0.0420, 0.0938, 0.0694], + [ 0.3013, 0.2502, 0.3140, ..., 0.0608, 0.0663, 0.0417], + [-0.0732, -0.0849, -0.0995, ..., -0.0864, -0.1098, -0.1144], + ..., + [-0.3411, -0.3072, -0.2568, ..., -0.0054, 0.0004, 0.0052], + [ 0.1056, 0.1117, 0.1376, ..., -0.1047, -0.0965, -0.1060], + [ 0.0718, 0.0641, 0.0619, ..., -0.0105, -0.0178, -0.0023]], + device='cuda:0'), grad: tensor([[ 2.3975e-03, 6.8235e-04, 1.3387e-04, ..., 5.4264e-04, + 8.4114e-04, 4.9353e-04], + [ 2.0084e-03, 6.0892e-04, 1.5950e-04, ..., 4.7421e-04, + 7.0953e-04, 4.2844e-04], + [ 5.0813e-05, 1.9267e-05, 8.9779e-06, ..., 1.3664e-05, + 1.8343e-05, 1.2241e-05], + ..., + [-4.6844e-03, -1.3247e-03, -2.5201e-04, ..., -1.0643e-03, + -1.6470e-03, -9.6607e-04], + [ 7.9572e-05, 2.2620e-05, 4.4331e-06, ..., 1.8343e-05, + 2.8133e-05, 1.6630e-05], + [ 2.9922e-04, 8.6010e-05, 1.7822e-05, ..., 6.9499e-05, + 1.0592e-04, 6.2883e-05]], device='cuda:0') +Epoch 7, bias, value: tensor([ 0.0672, -0.0121, -0.0528, -0.0537, 0.0089, -0.0162, 0.0493], + device='cuda:0'), grad: tensor([ 0.0065, 0.0052, 0.0001, -0.0002, -0.0126, 0.0002, 0.0008], + device='cuda:0') +588 +0.009819814303479266 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 786.05, cls_loss 0.0297 cls_loss_mapping 0.0896 cls_loss_causal 0.6016 re_mapping 0.0792 re_causal 0.0812 /// teacc 92.71 lr 0.00975528 +Epoch 8, weight, value: tensor([[-0.0676, -0.0763, -0.1606, ..., 0.0425, 0.0935, 0.0699], + [ 0.3098, 0.2586, 0.3212, ..., 0.0589, 0.0645, 0.0406], + [-0.0757, -0.0879, -0.1031, ..., -0.0840, -0.1073, -0.1118], + ..., + [-0.3368, -0.3049, -0.2562, ..., -0.0082, -0.0026, 0.0019], + [ 0.1032, 0.1099, 0.1352, ..., -0.1024, -0.0945, -0.1036], + [ 0.0685, 0.0621, 0.0606, ..., -0.0112, -0.0184, -0.0031]], + device='cuda:0'), grad: tensor([[-7.1716e-04, -3.8743e-04, -2.8014e-04, ..., -2.1017e-04, + -3.3545e-04, -2.5296e-04], + [ 7.5054e-04, 3.9124e-04, 2.6035e-04, ..., 2.0492e-04, + 2.7323e-04, 2.1517e-04], + [-4.8232e-04, -2.2614e-04, -1.2803e-04, ..., -1.2010e-04, + -1.2362e-04, -9.9778e-05], + ..., + [ 7.5936e-05, 2.4676e-05, 9.2313e-06, ..., 1.8805e-05, + 2.4348e-05, 1.4916e-05], + [ 9.5218e-06, 4.8392e-06, 3.1833e-06, ..., 2.5872e-06, + 3.5297e-06, 2.7195e-06], + [ 3.4070e-04, 1.8251e-04, 1.2958e-04, ..., 9.8169e-05, + 1.5116e-04, 1.1492e-04]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.0598, -0.0080, -0.0461, -0.0518, 0.0121, -0.0188, 0.0437], + device='cuda:0'), grad: tensor([-1.3371e-03, 1.3237e-03, -8.6784e-04, 4.3958e-05, 1.9133e-04, + 1.7464e-05, 6.2799e-04], device='cuda:0') +588 +0.009755282581475767 +changing lr +epoch 7, time 772.92, cls_loss 0.0159 cls_loss_mapping 0.0684 cls_loss_causal 0.5923 re_mapping 0.0701 re_causal 0.0730 /// teacc 92.46 lr 0.00968117 +Epoch 9, weight, value: tensor([[-7.0294e-02, -8.2486e-02, -1.6514e-01, ..., 4.2083e-02, + 9.2127e-02, 6.8614e-02], + [ 3.0445e-01, 2.5849e-01, 3.2050e-01, ..., 5.8001e-02, + 6.3234e-02, 4.0235e-02], + [-7.0183e-02, -8.2932e-02, -9.8085e-02, ..., -8.1957e-02, + -1.0465e-01, -1.0967e-01], + ..., + [-3.3327e-01, -3.0170e-01, -2.5488e-01, ..., -1.0581e-02, + -5.5196e-03, -1.6071e-04], + [ 1.0367e-01, 1.0979e-01, 1.3427e-01, ..., -9.9582e-02, + -9.1773e-02, -1.0072e-01], + [ 6.6764e-02, 6.0905e-02, 5.9512e-02, ..., -1.0945e-02, + -1.7924e-02, -3.0606e-03]], device='cuda:0'), grad: tensor([[-1.6987e-04, -7.6294e-05, -6.2823e-05, ..., -2.9922e-05, + -5.8085e-05, -4.1664e-05], + [-3.2127e-05, -4.3064e-05, -3.6836e-05, ..., -4.4964e-06, + 1.2316e-05, 4.6194e-06], + [ 1.6284e-04, 9.9838e-05, 8.1062e-05, ..., 2.8148e-05, + 3.2336e-05, 2.7448e-05], + ..., + [ 3.5524e-05, 1.8373e-05, 1.5914e-05, ..., 4.7088e-06, + 1.0245e-05, 7.1898e-06], + [ 9.3281e-06, 5.2415e-06, 4.2692e-06, ..., 1.5702e-06, + 2.2147e-06, 1.7481e-06], + [-4.5240e-05, -2.9370e-05, -2.1487e-05, ..., -6.5789e-06, + -5.7928e-06, -5.2638e-06]], device='cuda:0') +Epoch 9, bias, value: tensor([ 0.0720, -0.0199, -0.0425, -0.0519, 0.0113, -0.0174, 0.0395], + device='cuda:0'), grad: tensor([-3.8147e-04, 6.1512e-05, 2.3568e-04, 5.1528e-05, 7.0751e-05, + 1.5721e-05, -5.4061e-05], device='cuda:0') +588 +0.009681174353198686 +changing lr +epoch 8, time 777.37, cls_loss 0.0148 cls_loss_mapping 0.0577 cls_loss_causal 0.5646 re_mapping 0.0630 re_causal 0.0673 /// teacc 91.71 lr 0.00959764 +Epoch 10, weight, value: tensor([[-0.0781, -0.0882, -0.1678, ..., 0.0405, 0.0900, 0.0669], + [ 0.3057, 0.2600, 0.3207, ..., 0.0574, 0.0631, 0.0404], + [-0.0690, -0.0810, -0.0974, ..., -0.0808, -0.1042, -0.1088], + ..., + [-0.3274, -0.2982, -0.2523, ..., -0.0112, -0.0062, -0.0009], + [ 0.1028, 0.1089, 0.1328, ..., -0.0970, -0.0894, -0.0981], + [ 0.0657, 0.0598, 0.0583, ..., -0.0110, -0.0178, -0.0034]], + device='cuda:0'), grad: tensor([[-2.7714e-03, -8.9121e-04, -4.0722e-04, ..., -9.0313e-04, + -8.7309e-04, -7.9441e-04], + [ 1.8206e-03, 5.9080e-04, 2.7418e-04, ..., 5.9128e-04, + 5.7220e-04, 5.2023e-04], + [ 5.6171e-04, 1.8156e-04, 8.4460e-05, ..., 1.8311e-04, + 1.7738e-04, 1.6129e-04], + ..., + [ 2.6321e-04, 8.7440e-05, 4.0233e-05, ..., 8.4162e-05, + 8.0884e-05, 7.3850e-05], + [ 3.3706e-05, 1.0908e-05, 5.2154e-06, ..., 1.1005e-05, + 1.0721e-05, 9.7305e-06], + [ 1.5631e-05, -4.7199e-06, -8.6278e-06, ..., 9.6932e-06, + 8.7395e-06, 7.8455e-06]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.0645, -0.0142, -0.0434, -0.0515, 0.0164, -0.0180, 0.0376], + device='cuda:0'), grad: tensor([-6.3286e-03, 4.1428e-03, 1.2827e-03, 1.7333e-04, 5.8937e-04, + 7.6950e-05, 6.7055e-05], device='cuda:0') +588 +0.009597638862757255 +changing lr +epoch 9, time 778.03, cls_loss 0.0051 cls_loss_mapping 0.0493 cls_loss_causal 0.5442 re_mapping 0.0534 re_causal 0.0586 /// teacc 91.46 lr 0.00950484 +Epoch 11, weight, value: tensor([[-0.0841, -0.0941, -0.1715, ..., 0.0380, 0.0867, 0.0641], + [ 0.3054, 0.2613, 0.3211, ..., 0.0561, 0.0617, 0.0394], + [-0.0692, -0.0806, -0.0967, ..., -0.0798, -0.1030, -0.1076], + ..., + [-0.3233, -0.2945, -0.2503, ..., -0.0112, -0.0066, -0.0011], + [ 0.1038, 0.1091, 0.1323, ..., -0.0944, -0.0866, -0.0951], + [ 0.0658, 0.0599, 0.0580, ..., -0.0108, -0.0175, -0.0035]], + device='cuda:0'), grad: tensor([[-3.3894e-03, -2.1896e-03, -1.7900e-03, ..., -3.7837e-04, + -5.4598e-04, -4.4060e-04], + [ 1.5087e-03, 9.5987e-04, 7.8297e-04, ..., 1.8430e-04, + 2.6417e-04, 2.1267e-04], + [ 1.5628e-04, 8.4937e-05, 6.8069e-05, ..., 3.9160e-05, + 5.3555e-05, 4.3809e-05], + ..., + [ 2.8342e-05, 7.5459e-05, 6.7115e-05, ..., -7.2420e-05, + -9.0361e-05, -7.4565e-05], + [ 6.9320e-05, 3.8296e-05, 3.0667e-05, ..., 1.6332e-05, + 2.1875e-05, 1.7911e-05], + [ 1.5001e-03, 9.5367e-04, 7.7868e-04, ..., 1.9145e-04, + 2.7084e-04, 2.1970e-04]], device='cuda:0') +Epoch 11, bias, value: tensor([ 0.0619, -0.0127, -0.0436, -0.0510, 0.0152, -0.0142, 0.0360], + device='cuda:0'), grad: tensor([-0.0044, 0.0020, 0.0003, 0.0002, -0.0002, 0.0001, 0.0020], + device='cuda:0') +588 +0.009504844339512096 +changing lr +epoch 10, time 774.60, cls_loss 0.0047 cls_loss_mapping 0.0510 cls_loss_causal 0.5466 re_mapping 0.0465 re_causal 0.0531 /// teacc 92.71 lr 0.00940298 +Epoch 12, weight, value: tensor([[-0.0839, -0.0960, -0.1720, ..., 0.0390, 0.0870, 0.0646], + [ 0.3010, 0.2583, 0.3176, ..., 0.0545, 0.0600, 0.0383], + [-0.0696, -0.0805, -0.0965, ..., -0.0786, -0.1018, -0.1064], + ..., + [-0.3213, -0.2923, -0.2497, ..., -0.0127, -0.0084, -0.0027], + [ 0.1061, 0.1114, 0.1336, ..., -0.0919, -0.0843, -0.0926], + [ 0.0650, 0.0597, 0.0578, ..., -0.0108, -0.0174, -0.0038]], + device='cuda:0'), grad: tensor([[ 3.0918e-03, 5.6839e-04, 2.7800e-04, ..., 1.4496e-03, + 1.6880e-03, 1.5783e-03], + [ 6.1095e-05, 2.4781e-05, 1.9237e-05, ..., 1.6943e-05, + 1.9178e-05, 1.7986e-05], + [ 8.3372e-06, -1.6585e-05, -1.8209e-05, ..., 2.3082e-05, + 2.9430e-05, 2.4959e-05], + ..., + [-3.1319e-03, -5.3263e-04, -2.3746e-04, ..., -1.5078e-03, + -1.7595e-03, -1.6422e-03], + [ 4.3333e-05, 1.2241e-05, 8.2031e-06, ..., 1.6436e-05, + 1.9014e-05, 1.7807e-05], + [-8.8274e-05, -6.1035e-05, -5.3346e-05, ..., -3.3136e-06, + -2.3209e-06, -1.9576e-06]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0681, -0.0108, -0.0434, -0.0517, 0.0130, -0.0156, 0.0323], + device='cuda:0'), grad: tensor([ 9.1553e-03, 1.2910e-04, 9.7752e-05, 3.8117e-05, -9.4452e-03, + 1.1194e-04, -9.1195e-05], device='cuda:0') +588 +0.009402977659283692 +changing lr +epoch 11, time 778.37, cls_loss 0.0064 cls_loss_mapping 0.0432 cls_loss_causal 0.5521 re_mapping 0.0414 re_causal 0.0490 /// teacc 92.46 lr 0.00929224 +Epoch 13, weight, value: tensor([[-0.0895, -0.1000, -0.1741, ..., 0.0388, 0.0858, 0.0638], + [ 0.3020, 0.2592, 0.3167, ..., 0.0531, 0.0586, 0.0375], + [-0.0682, -0.0785, -0.0950, ..., -0.0778, -0.1009, -0.1055], + ..., + [-0.3176, -0.2893, -0.2477, ..., -0.0138, -0.0098, -0.0040], + [ 0.1063, 0.1117, 0.1336, ..., -0.0899, -0.0825, -0.0905], + [ 0.0652, 0.0579, 0.0562, ..., -0.0098, -0.0157, -0.0025]], + device='cuda:0'), grad: tensor([[ 8.2433e-05, 1.9953e-05, 1.0490e-05, ..., 1.7807e-05, + 2.6301e-05, 1.9848e-05], + [ 6.4909e-05, 1.0408e-05, 1.3374e-06, ..., 1.1384e-05, + 1.7986e-05, 1.4320e-05], + [-3.5262e-04, -1.8263e-04, -1.2827e-04, ..., -1.3399e-04, + -1.5056e-04, -1.1611e-04], + ..., + [-2.6727e-04, -4.0770e-05, -4.0755e-06, ..., -4.4644e-05, + -7.2002e-05, -5.7787e-05], + [ 3.7223e-05, 7.6741e-06, 2.3544e-06, ..., 7.3239e-06, + 1.0781e-05, 8.6427e-06], + [ 1.0252e-04, 4.0174e-05, 2.3350e-05, ..., 3.1441e-05, + 3.7342e-05, 2.9862e-05]], device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0610, -0.0050, -0.0432, -0.0564, 0.0138, -0.0165, 0.0385], + device='cuda:0'), grad: tensor([ 0.0002, 0.0002, -0.0009, 0.0008, -0.0008, 0.0001, 0.0003], + device='cuda:0') +588 +0.009292243968009333 +changing lr +epoch 12, time 778.05, cls_loss 0.0034 cls_loss_mapping 0.0397 cls_loss_causal 0.4946 re_mapping 0.0357 re_causal 0.0443 /// teacc 92.46 lr 0.00917287 +Epoch 14, weight, value: tensor([[-0.0858, -0.0990, -0.1720, ..., 0.0390, 0.0854, 0.0636], + [ 0.2971, 0.2573, 0.3136, ..., 0.0509, 0.0561, 0.0357], + [-0.0689, -0.0782, -0.0943, ..., -0.0765, -0.0994, -0.1038], + ..., + [-0.3110, -0.2854, -0.2451, ..., -0.0137, -0.0101, -0.0043], + [ 0.1056, 0.1112, 0.1326, ..., -0.0876, -0.0801, -0.0880], + [ 0.0592, 0.0545, 0.0535, ..., -0.0107, -0.0167, -0.0036]], + device='cuda:0'), grad: tensor([[ 5.0020e-04, 1.7703e-04, 1.3399e-04, ..., 9.3341e-05, + 1.1951e-04, 8.8930e-05], + [-2.8658e-04, -1.8084e-04, -1.6677e-04, ..., -5.6177e-05, + -6.2406e-05, -5.9783e-05], + [ 2.1195e-04, 6.1929e-05, 4.0680e-05, ..., 3.9279e-05, + 5.1975e-05, 3.6150e-05], + ..., + [-5.5981e-04, -8.6486e-05, -2.0877e-05, ..., -1.0252e-04, + -1.4460e-04, -8.9288e-05], + [ 4.6194e-05, 9.7901e-06, 4.4219e-06, ..., 9.2089e-06, + 1.2517e-05, 8.6576e-06], + [ 4.4495e-05, 9.9018e-06, 4.8466e-06, ..., 8.2478e-06, + 1.1235e-05, 7.2941e-06]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0710, -0.0108, -0.0448, -0.0553, 0.0208, -0.0171, 0.0285], + device='cuda:0'), grad: tensor([ 0.0012, -0.0004, 0.0006, 0.0001, -0.0018, 0.0001, 0.0001], + device='cuda:0') +588 +0.009172866268606516 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 778.21, cls_loss 0.0027 cls_loss_mapping 0.0345 cls_loss_causal 0.5152 re_mapping 0.0299 re_causal 0.0393 /// teacc 93.97 lr 0.00904508 +Epoch 15, weight, value: tensor([[-0.0885, -0.1006, -0.1715, ..., 0.0375, 0.0830, 0.0615], + [ 0.2988, 0.2582, 0.3130, ..., 0.0516, 0.0570, 0.0370], + [-0.0679, -0.0788, -0.0953, ..., -0.0748, -0.0975, -0.1017], + ..., + [-0.3096, -0.2833, -0.2440, ..., -0.0152, -0.0120, -0.0062], + [ 0.1030, 0.1099, 0.1312, ..., -0.0860, -0.0787, -0.0864], + [ 0.0582, 0.0537, 0.0525, ..., -0.0109, -0.0167, -0.0039]], + device='cuda:0'), grad: tensor([[ 7.9393e-04, 2.1148e-04, 1.3363e-04, ..., 1.8585e-04, + 2.4772e-04, 2.1100e-04], + [ 6.7282e-04, 3.1376e-04, 2.6083e-04, ..., 1.1039e-04, + 1.4329e-04, 1.2684e-04], + [ 2.5883e-03, 5.7745e-04, 2.4629e-04, ..., 7.5293e-04, + 9.1219e-04, 8.1730e-04], + ..., + [-4.2000e-03, -9.4509e-04, -4.4274e-04, ..., -1.1654e-03, + -1.4477e-03, -1.2789e-03], + [-1.4439e-03, -9.9754e-04, -9.2840e-04, ..., -1.0061e-04, + -1.4162e-04, -1.3125e-04], + [ 8.4448e-04, 4.9305e-04, 4.4131e-04, ..., 9.5725e-05, + 1.2875e-04, 1.1522e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0666, -0.0021, -0.0379, -0.0540, 0.0156, -0.0225, 0.0269], + device='cuda:0'), grad: tensor([ 0.0021, 0.0012, 0.0068, 0.0013, -0.0112, -0.0013, 0.0011], + device='cuda:0') +588 +0.00904508497187474 +changing lr +epoch 14, time 776.19, cls_loss 0.0041 cls_loss_mapping 0.0315 cls_loss_causal 0.4928 re_mapping 0.0264 re_causal 0.0371 /// teacc 93.97 lr 0.00890916 +Epoch 16, weight, value: tensor([[-0.0897, -0.1021, -0.1711, ..., 0.0378, 0.0827, 0.0614], + [ 0.2948, 0.2564, 0.3104, ..., 0.0497, 0.0553, 0.0358], + [-0.0707, -0.0802, -0.0964, ..., -0.0744, -0.0968, -0.1007], + ..., + [-0.3038, -0.2791, -0.2413, ..., -0.0143, -0.0115, -0.0058], + [ 0.1054, 0.1116, 0.1325, ..., -0.0837, -0.0765, -0.0840], + [ 0.0578, 0.0526, 0.0509, ..., -0.0114, -0.0172, -0.0046]], + device='cuda:0'), grad: tensor([[ 9.8705e-05, 4.2081e-05, 3.3855e-05, ..., 2.5094e-05, + 2.2784e-05, 1.9744e-05], + [-4.3154e-05, -3.7581e-05, -3.4839e-05, ..., -1.2957e-05, + -1.2979e-05, -1.2062e-05], + [-7.9572e-05, -3.3170e-05, -2.7269e-05, ..., -1.7211e-05, + -1.2763e-05, -1.1444e-05], + ..., + [-5.4657e-05, -8.2552e-06, -2.6654e-06, ..., -1.4283e-05, + -1.3880e-05, -1.1198e-05], + [ 1.0870e-05, 4.1053e-06, 3.2391e-06, ..., 3.1311e-06, + 2.9299e-06, 2.5500e-06], + [ 2.3231e-05, 1.1705e-05, 1.0073e-05, ..., 6.2846e-06, + 5.3830e-06, 4.8503e-06]], device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0685, -0.0054, -0.0399, -0.0563, 0.0179, -0.0206, 0.0285], + device='cuda:0'), grad: tensor([ 1.9526e-04, -2.2843e-05, -1.5879e-04, 8.1658e-05, -1.5867e-04, + 2.3648e-05, 4.0025e-05], device='cuda:0') +588 +0.008909157412340152 +changing lr +epoch 15, time 778.58, cls_loss 0.0033 cls_loss_mapping 0.0375 cls_loss_causal 0.5059 re_mapping 0.0235 re_causal 0.0350 /// teacc 93.97 lr 0.00876536 +Epoch 17, weight, value: tensor([[-0.0916, -0.1036, -0.1711, ..., 0.0374, 0.0813, 0.0603], + [ 0.2933, 0.2555, 0.3084, ..., 0.0484, 0.0540, 0.0350], + [-0.0697, -0.0791, -0.0953, ..., -0.0731, -0.0955, -0.0993], + ..., + [-0.3017, -0.2775, -0.2407, ..., -0.0151, -0.0124, -0.0067], + [ 0.1059, 0.1123, 0.1329, ..., -0.0822, -0.0750, -0.0823], + [ 0.0565, 0.0515, 0.0498, ..., -0.0115, -0.0172, -0.0049]], + device='cuda:0'), grad: tensor([[-2.2972e-04, -1.1915e-04, -1.0139e-04, ..., -1.0985e-04, + -1.1837e-04, -1.1688e-04], + [ 2.2840e-04, 1.0729e-04, 8.7261e-05, ..., 1.0186e-04, + 1.1033e-04, 1.0931e-04], + [ 7.7412e-06, 1.4137e-06, 9.4529e-07, ..., 7.4431e-06, + 8.5086e-06, 8.9929e-06], + ..., + [-7.7128e-05, -1.6451e-05, -5.0515e-06, ..., -2.5421e-05, + -2.8759e-05, -2.9579e-05], + [ 2.9624e-05, 1.0096e-05, 6.1989e-06, ..., 1.0081e-05, + 1.1168e-05, 1.1221e-05], + [ 2.8998e-05, 1.2301e-05, 9.0301e-06, ..., 1.1556e-05, + 1.2405e-05, 1.2219e-05]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0674, -0.0017, -0.0403, -0.0556, 0.0183, -0.0222, 0.0269], + device='cuda:0'), grad: tensor([-3.7766e-04, 4.1556e-04, 2.2411e-05, 2.5913e-05, -2.1303e-04, + 6.8545e-05, 5.7757e-05], device='cuda:0') +588 +0.00876535733001806 +changing lr +epoch 16, time 787.75, cls_loss 0.0028 cls_loss_mapping 0.0351 cls_loss_causal 0.4896 re_mapping 0.0216 re_causal 0.0333 /// teacc 93.22 lr 0.00861397 +Epoch 18, weight, value: tensor([[-0.0939, -0.1067, -0.1731, ..., 0.0376, 0.0803, 0.0597], + [ 0.2905, 0.2543, 0.3057, ..., 0.0468, 0.0523, 0.0339], + [-0.0695, -0.0787, -0.0947, ..., -0.0723, -0.0942, -0.0979], + ..., + [-0.2982, -0.2739, -0.2381, ..., -0.0155, -0.0129, -0.0073], + [ 0.1080, 0.1135, 0.1342, ..., -0.0806, -0.0731, -0.0804], + [ 0.0557, 0.0509, 0.0493, ..., -0.0120, -0.0178, -0.0057]], + device='cuda:0'), grad: tensor([[ 2.2471e-04, 8.0824e-05, 5.8323e-05, ..., 4.1693e-05, + 4.3541e-05, 4.1455e-05], + [ 3.4332e-04, 1.7226e-04, 1.4496e-04, ..., 1.3685e-04, + 1.4579e-04, 1.4079e-04], + [-8.1711e-03, -2.6722e-03, -1.6327e-03, ..., -1.6499e-03, + -1.9569e-03, -1.7881e-03], + ..., + [ 7.4081e-03, 2.4147e-03, 1.4582e-03, ..., 1.5192e-03, + 1.8148e-03, 1.6556e-03], + [ 1.7858e-04, 6.1631e-05, 4.3243e-05, ..., 2.8327e-05, + 2.9683e-05, 2.8059e-05], + [ 9.5844e-05, 3.7581e-05, 2.8685e-05, ..., 2.2396e-05, + 2.3127e-05, 2.2218e-05]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0695, -0.0026, -0.0400, -0.0562, 0.0168, -0.0193, 0.0248], + device='cuda:0'), grad: tensor([ 4.9210e-04, 5.8699e-04, -1.9394e-02, 6.0976e-05, 1.7654e-02, + 3.9959e-04, 2.0039e-04], device='cuda:0') +588 +0.008613974319136962 +changing lr +epoch 17, time 783.23, cls_loss 0.0016 cls_loss_mapping 0.0270 cls_loss_causal 0.4554 re_mapping 0.0189 re_causal 0.0304 /// teacc 93.97 lr 0.00845531 +Epoch 19, weight, value: tensor([[-0.0932, -0.1060, -0.1708, ..., 0.0371, 0.0790, 0.0587], + [ 0.2853, 0.2510, 0.3015, ..., 0.0449, 0.0502, 0.0322], + [-0.0712, -0.0805, -0.0969, ..., -0.0709, -0.0929, -0.0964], + ..., + [-0.2919, -0.2689, -0.2341, ..., -0.0153, -0.0126, -0.0070], + [ 0.1086, 0.1144, 0.1350, ..., -0.0795, -0.0721, -0.0792], + [ 0.0551, 0.0500, 0.0483, ..., -0.0118, -0.0175, -0.0057]], + device='cuda:0'), grad: tensor([[-6.6340e-05, -2.0504e-05, -1.5438e-05, ..., -1.8775e-05, + -2.3022e-05, -1.7628e-05], + [ 9.1046e-06, 2.7250e-06, 1.8775e-06, ..., 2.8554e-06, + 3.0827e-06, 2.4457e-06], + [ 3.9130e-05, 1.0230e-05, 5.1521e-06, ..., 1.1802e-05, + 1.3918e-05, 1.1459e-05], + ..., + [-1.9446e-05, -3.7737e-06, 2.7474e-08, ..., -6.3181e-06, + -6.9141e-06, -6.2250e-06], + [ 6.3255e-06, 1.8487e-06, 1.2461e-06, ..., 1.9111e-06, + 2.1514e-06, 1.7406e-06], + [ 2.5466e-05, 7.8455e-06, 5.9605e-06, ..., 7.0967e-06, + 8.9630e-06, 6.8583e-06]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0694, -0.0055, -0.0364, -0.0572, 0.0192, -0.0211, 0.0248], + device='cuda:0'), grad: tensor([-1.4925e-04, 2.1026e-05, 9.8705e-05, 1.3396e-05, -5.6416e-05, + 1.4968e-05, 5.7429e-05], device='cuda:0') +588 +0.008455313244934327 +changing lr +epoch 18, time 784.05, cls_loss 0.0034 cls_loss_mapping 0.0282 cls_loss_causal 0.4894 re_mapping 0.0180 re_causal 0.0308 /// teacc 93.97 lr 0.00828969 +Epoch 20, weight, value: tensor([[-0.0906, -0.1053, -0.1686, ..., 0.0369, 0.0781, 0.0583], + [ 0.2827, 0.2493, 0.2987, ..., 0.0441, 0.0491, 0.0316], + [-0.0706, -0.0802, -0.0965, ..., -0.0704, -0.0921, -0.0955], + ..., + [-0.2892, -0.2655, -0.2317, ..., -0.0156, -0.0129, -0.0076], + [ 0.1067, 0.1132, 0.1336, ..., -0.0780, -0.0709, -0.0778], + [ 0.0528, 0.0484, 0.0469, ..., -0.0122, -0.0178, -0.0063]], + device='cuda:0'), grad: tensor([[-5.7280e-05, 1.3441e-05, 2.3514e-05, ..., -6.1333e-05, + -7.1764e-05, -6.0350e-05], + [ 3.2949e-04, 2.2340e-04, 2.0254e-04, ..., 2.7671e-05, + 3.7760e-05, 2.9266e-05], + [ 5.2840e-05, 2.3499e-05, 1.8448e-05, ..., 1.8016e-05, + 2.1964e-05, 1.8463e-05], + ..., + [ 1.2321e-03, 8.5735e-04, 7.7248e-04, ..., 1.0759e-04, + 1.6880e-04, 1.3793e-04], + [-2.7237e-03, -1.9026e-03, -1.7233e-03, ..., -2.0909e-04, + -3.2377e-04, -2.5988e-04], + [ 4.9543e-04, 3.3903e-04, 3.0684e-04, ..., 4.2081e-05, + 6.0827e-05, 4.8190e-05]], device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0739, -0.0047, -0.0337, -0.0556, 0.0157, -0.0240, 0.0219], + device='cuda:0'), grad: tensor([-0.0003, 0.0004, 0.0001, 0.0008, 0.0012, -0.0027, 0.0005], + device='cuda:0') +588 +0.008289693629698565 +changing lr +epoch 19, time 784.24, cls_loss 0.0016 cls_loss_mapping 0.0202 cls_loss_causal 0.4631 re_mapping 0.0153 re_causal 0.0277 /// teacc 92.21 lr 0.00811745 +Epoch 21, weight, value: tensor([[-0.0907, -0.1054, -0.1672, ..., 0.0355, 0.0760, 0.0565], + [ 0.2816, 0.2481, 0.2962, ..., 0.0442, 0.0493, 0.0321], + [-0.0731, -0.0815, -0.0975, ..., -0.0708, -0.0925, -0.0955], + ..., + [-0.2854, -0.2619, -0.2291, ..., -0.0157, -0.0134, -0.0080], + [ 0.1064, 0.1128, 0.1328, ..., -0.0764, -0.0694, -0.0761], + [ 0.0508, 0.0468, 0.0455, ..., -0.0126, -0.0182, -0.0068]], + device='cuda:0'), grad: tensor([[-5.0402e-04, -1.3399e-04, -6.7711e-05, ..., -1.0961e-04, + -1.3840e-04, -1.1963e-04], + [ 2.2531e-04, 7.9691e-05, 5.1290e-05, ..., 6.3121e-05, + 7.5698e-05, 6.9797e-05], + [-4.8375e-04, -2.2137e-04, -1.7023e-04, ..., -1.6260e-04, + -1.9073e-04, -1.8334e-04], + ..., + [ 4.2701e-04, 1.5950e-04, 1.0848e-04, ..., 1.1283e-04, + 1.3638e-04, 1.2875e-04], + [ 4.9263e-05, 1.5363e-05, 1.0908e-05, ..., 1.3426e-05, + 1.7136e-05, 1.4856e-05], + [ 9.1434e-05, 2.9832e-05, 1.6481e-05, ..., 2.1935e-05, + 2.6092e-05, 2.2903e-05]], device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0742, -0.0014, -0.0364, -0.0531, 0.0146, -0.0242, 0.0199], + device='cuda:0'), grad: tensor([-0.0013, 0.0005, -0.0008, 0.0004, 0.0009, 0.0001, 0.0002], + device='cuda:0') +588 +0.00811744900929367 +changing lr +epoch 20, time 781.25, cls_loss 0.0016 cls_loss_mapping 0.0238 cls_loss_causal 0.4656 re_mapping 0.0142 re_causal 0.0270 /// teacc 93.72 lr 0.00793893 +Epoch 22, weight, value: tensor([[-0.0905, -0.1050, -0.1653, ..., 0.0352, 0.0754, 0.0560], + [ 0.2779, 0.2455, 0.2927, ..., 0.0435, 0.0485, 0.0317], + [-0.0696, -0.0799, -0.0961, ..., -0.0688, -0.0903, -0.0933], + ..., + [-0.2810, -0.2585, -0.2269, ..., -0.0164, -0.0143, -0.0088], + [ 0.1060, 0.1125, 0.1321, ..., -0.0749, -0.0680, -0.0746], + [ 0.0503, 0.0465, 0.0453, ..., -0.0128, -0.0184, -0.0072]], + device='cuda:0'), grad: tensor([[ 1.1024e-03, 1.7571e-04, 2.6654e-06, ..., 1.3304e-04, + 1.4126e-04, 9.6440e-05], + [ 4.5240e-05, 1.0446e-05, 6.2324e-06, ..., 9.6932e-06, + 9.9763e-06, 7.6108e-06], + [ 1.4460e-04, 2.7388e-05, 6.2995e-06, ..., 2.1040e-05, + 2.2888e-05, 1.6823e-05], + ..., + [-1.3838e-03, -2.3115e-04, -2.0191e-05, ..., -1.7822e-04, + -1.8954e-04, -1.3220e-04], + [ 3.8713e-05, 7.9274e-06, 3.2261e-06, ..., 6.7167e-06, + 7.0482e-06, 5.2378e-06], + [ 4.1366e-05, 7.4916e-06, 1.3122e-06, ..., 5.8711e-06, + 6.3404e-06, 4.6156e-06]], device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0737, -0.0015, -0.0287, -0.0590, 0.0157, -0.0245, 0.0181], + device='cuda:0'), grad: tensor([ 3.0155e-03, 1.3125e-04, 3.9840e-04, 3.5793e-05, -3.8013e-03, + 1.0926e-04, 1.1390e-04], device='cuda:0') +588 +0.007938926261462368 +changing lr +epoch 21, time 782.38, cls_loss 0.0015 cls_loss_mapping 0.0242 cls_loss_causal 0.4788 re_mapping 0.0139 re_causal 0.0263 /// teacc 93.47 lr 0.00775448 +Epoch 23, weight, value: tensor([[-0.0903, -0.1048, -0.1638, ..., 0.0347, 0.0743, 0.0550], + [ 0.2745, 0.2426, 0.2889, ..., 0.0427, 0.0479, 0.0315], + [-0.0698, -0.0793, -0.0952, ..., -0.0680, -0.0893, -0.0923], + ..., + [-0.2765, -0.2548, -0.2240, ..., -0.0165, -0.0147, -0.0092], + [ 0.1057, 0.1121, 0.1314, ..., -0.0736, -0.0669, -0.0733], + [ 0.0496, 0.0458, 0.0445, ..., -0.0126, -0.0181, -0.0071]], + device='cuda:0'), grad: tensor([[-3.3379e-04, -5.0217e-05, 6.4960e-07, ..., -7.9572e-05, + -8.9288e-05, -7.8738e-05], + [ 6.7651e-05, 1.6302e-05, 5.7444e-06, ..., 1.6719e-05, + 1.8090e-05, 1.6406e-05], + [ 2.9877e-05, 2.6990e-06, -2.5518e-06, ..., 7.1749e-06, + 8.9183e-06, 7.8976e-06], + ..., + [ 1.4472e-04, 1.4268e-05, -5.7854e-06, ..., 3.6836e-05, + 4.1395e-05, 3.6150e-05], + [ 2.5228e-05, 4.8280e-06, 9.1037e-07, ..., 5.8971e-06, + 6.5528e-06, 5.8413e-06], + [ 4.7833e-05, 9.7379e-06, 1.9241e-06, ..., 1.0654e-05, + 1.1787e-05, 1.0490e-05]], device='cuda:0') +Epoch 23, bias, value: tensor([ 0.0737, -0.0002, -0.0307, -0.0591, 0.0165, -0.0243, 0.0180], + device='cuda:0'), grad: tensor([-9.7656e-04, 1.7679e-04, 9.5427e-05, 5.5522e-05, 4.4632e-04, + 7.0810e-05, 1.3161e-04], device='cuda:0') +588 +0.007754484907260515 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 785.57, cls_loss 0.0022 cls_loss_mapping 0.0237 cls_loss_causal 0.4663 re_mapping 0.0132 re_causal 0.0264 /// teacc 94.22 lr 0.00756450 +Epoch 24, weight, value: tensor([[-0.0925, -0.1062, -0.1638, ..., 0.0340, 0.0733, 0.0542], + [ 0.2716, 0.2406, 0.2860, ..., 0.0418, 0.0469, 0.0308], + [-0.0699, -0.0791, -0.0948, ..., -0.0674, -0.0885, -0.0913], + ..., + [-0.2725, -0.2517, -0.2217, ..., -0.0163, -0.0150, -0.0094], + [ 0.1059, 0.1125, 0.1314, ..., -0.0726, -0.0660, -0.0722], + [ 0.0495, 0.0454, 0.0441, ..., -0.0128, -0.0183, -0.0074]], + device='cuda:0'), grad: tensor([[-5.5552e-05, -1.0900e-05, -6.4932e-06, ..., -3.2425e-05, + -3.6955e-05, -3.3021e-05], + [ 1.9073e-05, 4.8093e-06, 2.3395e-06, ..., 9.2834e-06, + 9.8944e-06, 9.1344e-06], + [ 9.9659e-05, 1.9118e-05, 5.3756e-06, ..., 2.0966e-05, + 2.7210e-05, 2.2233e-05], + ..., + [-1.1027e-04, -1.6630e-05, 3.1432e-09, ..., -1.2614e-05, + -1.9699e-05, -1.4253e-05], + [ 1.2137e-05, 2.8312e-06, 1.2452e-06, ..., 3.2634e-06, + 3.9898e-06, 3.3993e-06], + [ 5.4151e-05, 1.5765e-05, 1.0066e-05, ..., 2.6867e-05, + 2.9981e-05, 2.7239e-05]], device='cuda:0') +Epoch 24, bias, value: tensor([ 0.0723, -0.0003, -0.0308, -0.0574, 0.0175, -0.0252, 0.0180], + device='cuda:0'), grad: tensor([-1.6594e-04, 5.0187e-05, 2.9349e-04, -5.8934e-06, -3.4261e-04, + 3.3677e-05, 1.3709e-04], device='cuda:0') +588 +0.007564496387029534 +changing lr +epoch 23, time 777.39, cls_loss 0.0027 cls_loss_mapping 0.0201 cls_loss_causal 0.4614 re_mapping 0.0129 re_causal 0.0259 /// teacc 93.47 lr 0.00736934 +Epoch 25, weight, value: tensor([[-0.0925, -0.1067, -0.1631, ..., 0.0337, 0.0723, 0.0535], + [ 0.2693, 0.2389, 0.2835, ..., 0.0412, 0.0462, 0.0305], + [-0.0700, -0.0792, -0.0949, ..., -0.0671, -0.0880, -0.0907], + ..., + [-0.2701, -0.2491, -0.2198, ..., -0.0171, -0.0159, -0.0104], + [ 0.1053, 0.1119, 0.1304, ..., -0.0713, -0.0648, -0.0709], + [ 0.0499, 0.0458, 0.0444, ..., -0.0126, -0.0178, -0.0072]], + device='cuda:0'), grad: tensor([[ 1.6289e-03, 8.6451e-04, 7.7534e-04, ..., 1.5330e-04, + 1.7023e-04, 1.8120e-04], + [-1.7748e-03, -9.5940e-04, -8.6403e-04, ..., -1.7333e-04, + -1.8895e-04, -1.9956e-04], + [ 1.9670e-04, 8.6963e-05, 6.9618e-05, ..., 3.5405e-05, + 4.6581e-05, 4.5359e-05], + ..., + [-7.8082e-06, 2.1253e-06, 4.8913e-06, ..., -2.8498e-06, + -3.9488e-06, -3.4459e-06], + [ 2.2471e-05, 1.1474e-05, 9.8199e-06, ..., 3.4980e-06, + 4.1015e-06, 4.0047e-06], + [-2.1005e-04, -7.8499e-05, -5.6773e-05, ..., -4.1306e-05, + -5.8591e-05, -5.7042e-05]], device='cuda:0') +Epoch 25, bias, value: tensor([ 0.0732, 0.0002, -0.0302, -0.0569, 0.0154, -0.0257, 0.0180], + device='cuda:0'), grad: tensor([ 2.4300e-03, -2.6035e-03, 3.6573e-04, 2.3913e-04, -3.2187e-05, + 3.6359e-05, -4.3821e-04], device='cuda:0') +588 +0.007369343312364995 +changing lr +---------------------saving model at epoch 24---------------------------------------------------- +epoch 24, time 784.29, cls_loss 0.0012 cls_loss_mapping 0.0202 cls_loss_causal 0.4193 re_mapping 0.0118 re_causal 0.0243 /// teacc 94.47 lr 0.00716942 +Epoch 26, weight, value: tensor([[-0.0932, -0.1077, -0.1631, ..., 0.0334, 0.0712, 0.0527], + [ 0.2674, 0.2375, 0.2814, ..., 0.0404, 0.0455, 0.0301], + [-0.0683, -0.0774, -0.0932, ..., -0.0662, -0.0869, -0.0895], + ..., + [-0.2676, -0.2463, -0.2176, ..., -0.0174, -0.0161, -0.0108], + [ 0.1044, 0.1111, 0.1293, ..., -0.0700, -0.0637, -0.0696], + [ 0.0503, 0.0456, 0.0441, ..., -0.0123, -0.0174, -0.0070]], + device='cuda:0'), grad: tensor([[ 6.6757e-04, 3.2020e-04, 2.6894e-04, ..., 2.1827e-04, + 2.4021e-04, 2.2185e-04], + [-1.0004e-03, -5.3501e-04, -4.6778e-04, ..., -3.1519e-04, + -3.4690e-04, -3.2616e-04], + [-1.9267e-05, 1.4506e-05, 2.4095e-05, ..., -2.6450e-05, + -2.3142e-05, -1.8403e-05], + ..., + [-4.3333e-05, 1.7703e-05, 2.6360e-05, ..., -1.3717e-05, + -1.7226e-05, -1.2234e-05], + [ 6.3777e-05, 2.9311e-05, 2.4229e-05, ..., 1.6570e-05, + 1.8924e-05, 1.7494e-05], + [ 7.4208e-05, 3.5852e-05, 2.9847e-05, ..., 2.6062e-05, + 2.8253e-05, 2.6092e-05]], device='cuda:0') +Epoch 26, bias, value: tensor([ 0.0744, 0.0013, -0.0292, -0.0586, 0.0133, -0.0262, 0.0192], + device='cuda:0'), grad: tensor([ 0.0012, -0.0016, -0.0001, 0.0005, -0.0002, 0.0001, 0.0001], + device='cuda:0') +588 +0.0071694186955877925 +changing lr +epoch 25, time 783.09, cls_loss 0.0015 cls_loss_mapping 0.0195 cls_loss_causal 0.4439 re_mapping 0.0110 re_causal 0.0235 /// teacc 93.47 lr 0.00696513 +Epoch 27, weight, value: tensor([[-0.0931, -0.1077, -0.1622, ..., 0.0328, 0.0699, 0.0516], + [ 0.2649, 0.2359, 0.2791, ..., 0.0397, 0.0447, 0.0297], + [-0.0692, -0.0773, -0.0928, ..., -0.0659, -0.0865, -0.0889], + ..., + [-0.2623, -0.2428, -0.2149, ..., -0.0172, -0.0158, -0.0106], + [ 0.1036, 0.1106, 0.1287, ..., -0.0688, -0.0627, -0.0684], + [ 0.0495, 0.0446, 0.0431, ..., -0.0125, -0.0172, -0.0071]], + device='cuda:0'), grad: tensor([[ 3.4404e-04, 1.6963e-04, 1.5175e-04, ..., 1.0163e-04, + 1.0586e-04, 1.0747e-04], + [-5.7697e-04, -3.1424e-04, -2.8729e-04, ..., -1.5736e-04, + -1.6725e-04, -1.6594e-04], + [ 3.3212e-04, 1.7452e-04, 1.7071e-04, ..., 6.0886e-05, + 7.8261e-05, 7.0035e-05], + ..., + [ 1.9908e-04, 1.4424e-04, 1.3816e-04, ..., 9.4175e-06, + 1.3933e-05, 1.1228e-05], + [-3.2783e-04, -2.2471e-04, -2.1291e-04, ..., -1.0632e-05, + -2.0429e-05, -1.5661e-05], + [ 2.5725e-04, 1.6308e-04, 1.5402e-04, ..., 3.9935e-05, + 4.8101e-05, 4.3929e-05]], device='cuda:0') +Epoch 27, bias, value: tensor([ 0.0737, 0.0005, -0.0319, -0.0594, 0.0191, -0.0274, 0.0198], + device='cuda:0'), grad: tensor([ 0.0006, -0.0010, 0.0005, -0.0004, 0.0002, -0.0004, 0.0003], + device='cuda:0') +588 +0.0069651251582696205 +changing lr +epoch 26, time 784.48, cls_loss 0.0016 cls_loss_mapping 0.0172 cls_loss_causal 0.4328 re_mapping 0.0104 re_causal 0.0226 /// teacc 93.97 lr 0.00675687 +Epoch 28, weight, value: tensor([[-0.0910, -0.1079, -0.1617, ..., 0.0331, 0.0699, 0.0515], + [ 0.2612, 0.2333, 0.2757, ..., 0.0386, 0.0434, 0.0288], + [-0.0675, -0.0760, -0.0914, ..., -0.0651, -0.0857, -0.0879], + ..., + [-0.2631, -0.2416, -0.2140, ..., -0.0181, -0.0169, -0.0116], + [ 0.1059, 0.1122, 0.1298, ..., -0.0674, -0.0613, -0.0670], + [ 0.0461, 0.0427, 0.0413, ..., -0.0131, -0.0179, -0.0078]], + device='cuda:0'), grad: tensor([[-3.8624e-05, -1.5020e-05, -1.2144e-05, ..., -9.7975e-06, + -1.0334e-05, -9.5740e-06], + [-5.6289e-06, -5.3123e-06, -4.6603e-06, ..., -2.4913e-07, + -9.1316e-07, -1.0745e-07], + [ 2.4755e-06, 1.6652e-06, 1.5339e-06, ..., 1.0310e-06, + 1.1958e-06, 1.1399e-06], + ..., + [ 1.1645e-05, 7.1600e-06, 6.2920e-06, ..., 3.0696e-06, + 3.2298e-06, 2.7381e-06], + [ 1.0073e-05, 4.6119e-06, 3.8557e-06, ..., 2.5518e-06, + 2.8424e-06, 2.4885e-06], + [ 1.9416e-05, 8.2329e-06, 6.7055e-06, ..., 4.7274e-06, + 5.1372e-06, 4.5113e-06]], device='cuda:0') +Epoch 28, bias, value: tensor([ 0.0811, -0.0006, -0.0296, -0.0574, 0.0113, -0.0252, 0.0148], + device='cuda:0'), grad: tensor([-7.1585e-05, -6.8471e-06, 1.1623e-06, 6.6198e-06, 1.5810e-05, + 1.7777e-05, 3.6836e-05], device='cuda:0') +588 +0.006756874120406716 +changing lr +epoch 27, time 787.67, cls_loss 0.0018 cls_loss_mapping 0.0170 cls_loss_causal 0.4260 re_mapping 0.0104 re_causal 0.0233 /// teacc 93.47 lr 0.00654508 +Epoch 29, weight, value: tensor([[-0.0942, -0.1096, -0.1622, ..., 0.0321, 0.0683, 0.0503], + [ 0.2606, 0.2329, 0.2746, ..., 0.0384, 0.0432, 0.0288], + [-0.0668, -0.0750, -0.0904, ..., -0.0651, -0.0855, -0.0877], + ..., + [-0.2580, -0.2385, -0.2116, ..., -0.0175, -0.0162, -0.0111], + [ 0.1045, 0.1113, 0.1287, ..., -0.0666, -0.0606, -0.0661], + [ 0.0453, 0.0417, 0.0403, ..., -0.0128, -0.0175, -0.0077]], + device='cuda:0'), grad: tensor([[ 8.3017e-04, 4.9400e-04, 4.5729e-04, ..., 2.7943e-04, + 2.8706e-04, 2.5630e-04], + [-1.5125e-03, -9.4414e-04, -8.8310e-04, ..., -5.1022e-04, + -5.1260e-04, -4.6444e-04], + [ 3.5524e-04, 2.2745e-04, 2.1183e-04, ..., 1.2589e-04, + 1.2755e-04, 1.1522e-04], + ..., + [-3.2544e-05, 1.2241e-05, 1.9267e-05, ..., -1.0952e-05, + -2.1741e-05, -1.3456e-05], + [ 5.8740e-05, 3.1263e-05, 2.8178e-05, ..., 1.9714e-05, + 2.1145e-05, 1.8328e-05], + [ 9.6500e-05, 5.5939e-05, 5.1767e-05, ..., 3.1322e-05, + 3.2336e-05, 2.8819e-05]], device='cuda:0') +Epoch 29, bias, value: tensor([ 0.0755, 0.0014, -0.0294, -0.0574, 0.0169, -0.0274, 0.0149], + device='cuda:0'), grad: tensor([ 1.1759e-03, -1.9722e-03, 4.3583e-04, 2.8968e-04, -1.6963e-04, + 9.8050e-05, 1.4305e-04], device='cuda:0') +588 +0.00654508497187474 +changing lr +epoch 28, time 782.41, cls_loss 0.0013 cls_loss_mapping 0.0135 cls_loss_causal 0.4478 re_mapping 0.0099 re_causal 0.0232 /// teacc 91.21 lr 0.00633018 +Epoch 30, weight, value: tensor([[-0.0950, -0.1099, -0.1616, ..., 0.0315, 0.0671, 0.0494], + [ 0.2588, 0.2313, 0.2725, ..., 0.0379, 0.0427, 0.0286], + [-0.0667, -0.0745, -0.0898, ..., -0.0645, -0.0848, -0.0869], + ..., + [-0.2561, -0.2372, -0.2107, ..., -0.0176, -0.0163, -0.0113], + [ 0.1060, 0.1123, 0.1292, ..., -0.0651, -0.0592, -0.0646], + [ 0.0453, 0.0416, 0.0400, ..., -0.0128, -0.0174, -0.0078]], + device='cuda:0'), grad: tensor([[ 9.0420e-05, 5.4091e-05, 5.1469e-05, ..., 3.5822e-05, + 3.7998e-05, 3.6001e-05], + [ 1.0986e-03, 5.4789e-04, 5.0545e-04, ..., 6.4230e-04, + 6.7425e-04, 6.4564e-04], + [ 1.1692e-03, 6.7568e-04, 6.7043e-04, ..., 3.3116e-04, + 3.7360e-04, 3.4857e-04], + ..., + [-1.8664e-06, -8.3596e-06, -1.1399e-05, ..., 1.2106e-04, + 1.2374e-04, 1.2410e-04], + [ 1.3912e-04, 6.4492e-05, 6.0350e-05, ..., 4.7445e-05, + 5.1826e-05, 4.7833e-05], + [ 1.3411e-04, 6.7055e-05, 6.4194e-05, ..., 5.7012e-05, + 6.1929e-05, 5.8502e-05]], device='cuda:0') +Epoch 30, bias, value: tensor([ 0.0736, 0.0026, -0.0302, -0.0588, 0.0181, -0.0259, 0.0152], + device='cuda:0'), grad: tensor([ 1.0484e-04, 1.8797e-03, 1.5717e-03, -4.0665e-03, 2.4527e-05, + 2.5821e-04, 2.2674e-04], device='cuda:0') +588 +0.006330184227833378 +changing lr +epoch 29, time 787.30, cls_loss 0.0014 cls_loss_mapping 0.0155 cls_loss_causal 0.4136 re_mapping 0.0095 re_causal 0.0221 /// teacc 93.97 lr 0.00611260 +Epoch 31, weight, value: tensor([[-0.0945, -0.1099, -0.1608, ..., 0.0311, 0.0662, 0.0486], + [ 0.2560, 0.2293, 0.2698, ..., 0.0376, 0.0423, 0.0285], + [-0.0661, -0.0739, -0.0889, ..., -0.0641, -0.0843, -0.0863], + ..., + [-0.2531, -0.2351, -0.2093, ..., -0.0175, -0.0162, -0.0112], + [ 0.1062, 0.1128, 0.1296, ..., -0.0643, -0.0583, -0.0638], + [ 0.0436, 0.0405, 0.0390, ..., -0.0130, -0.0176, -0.0081]], + device='cuda:0'), grad: tensor([[ 2.4152e-04, 9.8050e-05, 9.0122e-05, ..., 5.9277e-05, + 5.1796e-05, 5.5343e-05], + [ 3.0547e-05, -5.6960e-06, -9.4920e-06, ..., 3.0510e-06, + 5.8264e-06, 7.0110e-06], + [-1.3351e-04, -4.4465e-05, -3.5226e-05, ..., -7.9349e-06, + -2.6718e-05, -2.6569e-05], + ..., + [-1.4198e-04, 2.9847e-05, 4.0114e-05, ..., -1.0359e-04, + -7.9453e-05, -8.3447e-05], + [-1.0366e-03, -7.8630e-04, -7.5960e-04, ..., -1.0937e-05, + -3.3736e-05, -2.6807e-05], + [ 6.4993e-04, 4.7135e-04, 4.5252e-04, ..., 2.2724e-05, + 3.6240e-05, 3.1263e-05]], device='cuda:0') +Epoch 31, bias, value: tensor([ 0.0748, 0.0023, -0.0298, -0.0585, 0.0201, -0.0268, 0.0126], + device='cuda:0'), grad: tensor([ 0.0005, 0.0001, -0.0003, 0.0005, -0.0006, -0.0008, 0.0006], + device='cuda:0') +588 +0.006112604669781575 +changing lr +epoch 30, time 777.64, cls_loss 0.0009 cls_loss_mapping 0.0130 cls_loss_causal 0.4182 re_mapping 0.0090 re_causal 0.0209 /// teacc 93.47 lr 0.00589278 +Epoch 32, weight, value: tensor([[-0.0958, -0.1103, -0.1603, ..., 0.0302, 0.0649, 0.0475], + [ 0.2544, 0.2283, 0.2682, ..., 0.0371, 0.0417, 0.0282], + [-0.0650, -0.0733, -0.0882, ..., -0.0639, -0.0838, -0.0857], + ..., + [-0.2512, -0.2333, -0.2080, ..., -0.0174, -0.0163, -0.0112], + [ 0.1061, 0.1126, 0.1292, ..., -0.0635, -0.0577, -0.0630], + [ 0.0430, 0.0400, 0.0383, ..., -0.0130, -0.0174, -0.0082]], + device='cuda:0'), grad: tensor([[-1.7881e-04, -3.1769e-05, -8.3894e-06, ..., -7.2002e-05, + -8.5115e-05, -8.2433e-05], + [ 1.2219e-05, -2.5169e-07, -2.0284e-06, ..., 3.5614e-06, + 4.0829e-06, 4.0829e-06], + [ 1.1176e-05, 1.0207e-06, -4.6985e-07, ..., 6.5155e-06, + 9.6485e-06, 8.7246e-06], + ..., + [ 6.4254e-05, 1.0289e-05, 5.6461e-09, ..., 2.7999e-05, + 3.2693e-05, 3.2157e-05], + [ 1.6838e-05, 4.7274e-06, 3.0529e-06, ..., 5.6140e-06, + 6.5491e-06, 6.3069e-06], + [-3.6824e-06, -6.1244e-06, -6.4969e-06, ..., 1.8412e-06, + 2.0638e-06, 1.9595e-06]], device='cuda:0') +Epoch 32, bias, value: tensor([ 0.0717, 0.0021, -0.0272, -0.0570, 0.0196, -0.0265, 0.0121], + device='cuda:0'), grad: tensor([-5.0974e-04, 4.3571e-05, 3.6329e-05, 1.9956e-04, 1.7917e-04, + 4.3243e-05, 7.9870e-06], device='cuda:0') +588 +0.005892784473993186 +changing lr +epoch 31, time 786.99, cls_loss 0.0010 cls_loss_mapping 0.0126 cls_loss_causal 0.4029 re_mapping 0.0085 re_causal 0.0205 /// teacc 93.22 lr 0.00567117 +Epoch 33, weight, value: tensor([[-0.0957, -0.1104, -0.1598, ..., 0.0300, 0.0642, 0.0470], + [ 0.2545, 0.2280, 0.2673, ..., 0.0368, 0.0416, 0.0282], + [-0.0660, -0.0741, -0.0889, ..., -0.0638, -0.0835, -0.0854], + ..., + [-0.2500, -0.2317, -0.2068, ..., -0.0176, -0.0166, -0.0116], + [ 0.1066, 0.1130, 0.1294, ..., -0.0625, -0.0568, -0.0620], + [ 0.0430, 0.0399, 0.0383, ..., -0.0131, -0.0175, -0.0083]], + device='cuda:0'), grad: tensor([[-3.7479e-04, -1.0264e-04, -5.4061e-05, ..., -1.2934e-04, + -1.5962e-04, -1.3351e-04], + [ 1.0926e-04, 3.0875e-05, 1.6913e-05, ..., 3.5852e-05, + 4.4197e-05, 3.7044e-05], + [ 2.6718e-05, 7.8008e-06, 4.3102e-06, ..., 1.0461e-05, + 1.2524e-05, 1.0684e-05], + ..., + [ 1.2136e-04, 3.4034e-05, 1.8865e-05, ..., 4.2945e-05, + 5.3167e-05, 4.4525e-05], + [ 3.9786e-05, 1.0602e-05, 5.4576e-06, ..., 1.2681e-05, + 1.5855e-05, 1.3150e-05], + [ 4.3809e-05, 1.2331e-05, 6.6198e-06, ..., 1.6227e-05, + 1.9699e-05, 1.6645e-05]], device='cuda:0') +Epoch 33, bias, value: tensor([ 0.0725, 0.0054, -0.0270, -0.0588, 0.0172, -0.0261, 0.0118], + device='cuda:0'), grad: tensor([-9.7752e-04, 2.8324e-04, 6.5923e-05, 9.7632e-05, 3.1328e-04, + 1.0639e-04, 1.1134e-04], device='cuda:0') +588 +0.00567116632908828 +changing lr +epoch 32, time 781.76, cls_loss 0.0009 cls_loss_mapping 0.0111 cls_loss_causal 0.3864 re_mapping 0.0085 re_causal 0.0203 /// teacc 93.22 lr 0.00544820 +Epoch 34, weight, value: tensor([[-0.0953, -0.1098, -0.1585, ..., 0.0296, 0.0633, 0.0463], + [ 0.2515, 0.2261, 0.2648, ..., 0.0359, 0.0406, 0.0274], + [-0.0654, -0.0735, -0.0881, ..., -0.0632, -0.0828, -0.0845], + ..., + [-0.2463, -0.2292, -0.2048, ..., -0.0172, -0.0161, -0.0112], + [ 0.1047, 0.1117, 0.1278, ..., -0.0619, -0.0563, -0.0614], + [ 0.0430, 0.0396, 0.0380, ..., -0.0131, -0.0174, -0.0084]], + device='cuda:0'), grad: tensor([[-3.1257e-04, -1.0788e-04, -1.1796e-04, ..., -5.4955e-05, + -6.5863e-05, -4.9770e-05], + [ 1.6057e-04, 4.4435e-05, 3.5286e-05, ..., 4.7147e-05, + 5.1558e-05, 4.8280e-05], + [ 1.0710e-03, 1.9407e-04, 1.1021e-04, ..., 3.2425e-04, + 3.3426e-04, 3.2640e-04], + ..., + [-1.0185e-03, -1.4710e-04, -4.0054e-05, ..., -3.3808e-04, + -3.4451e-04, -3.4618e-04], + [ 3.4362e-05, 8.6948e-06, 7.3947e-06, ..., 8.5086e-06, + 9.5740e-06, 8.5533e-06], + [ 4.0114e-05, 1.0327e-05, 8.5458e-06, ..., 9.8795e-06, + 1.1548e-05, 1.0200e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([ 0.0718, 0.0034, -0.0265, -0.0587, 0.0206, -0.0282, 0.0127], + device='cuda:0'), grad: tensor([-6.4993e-04, 4.1318e-04, 3.2349e-03, 1.0186e-04, -3.2940e-03, + 8.9824e-05, 1.0437e-04], device='cuda:0') +588 +0.00544819654451717 +changing lr +epoch 33, time 786.81, cls_loss 0.0016 cls_loss_mapping 0.0138 cls_loss_causal 0.4062 re_mapping 0.0083 re_causal 0.0200 /// teacc 93.72 lr 0.00522432 +Epoch 35, weight, value: tensor([[-0.0950, -0.1096, -0.1576, ..., 0.0291, 0.0624, 0.0456], + [ 0.2496, 0.2247, 0.2629, ..., 0.0357, 0.0403, 0.0273], + [-0.0657, -0.0734, -0.0880, ..., -0.0630, -0.0825, -0.0841], + ..., + [-0.2449, -0.2281, -0.2041, ..., -0.0171, -0.0160, -0.0112], + [ 0.1066, 0.1127, 0.1286, ..., -0.0610, -0.0554, -0.0604], + [ 0.0418, 0.0390, 0.0373, ..., -0.0132, -0.0174, -0.0085]], + device='cuda:0'), grad: tensor([[ 7.8678e-05, 2.4125e-05, 1.5855e-05, ..., 1.8612e-05, + 1.6034e-05, 1.8910e-05], + [ 5.3018e-05, 1.6555e-05, 1.0923e-05, ..., 1.3947e-05, + 1.1861e-05, 1.3560e-05], + [ 3.6776e-05, 1.0327e-05, 6.2548e-06, ..., 1.0982e-05, + 9.8944e-06, 1.1176e-05], + ..., + [-4.6134e-04, -1.4389e-04, -9.5129e-05, ..., -1.2058e-04, + -1.0270e-04, -1.1784e-04], + [ 1.1903e-04, 3.7104e-05, 2.4602e-05, ..., 3.1441e-05, + 2.6762e-05, 3.0667e-05], + [ 3.1680e-05, 9.9242e-06, 6.5863e-06, ..., 8.6203e-06, + 7.3425e-06, 8.3521e-06]], device='cuda:0') +Epoch 35, bias, value: tensor([ 0.0719, 0.0029, -0.0268, -0.0584, 0.0200, -0.0249, 0.0105], + device='cuda:0'), grad: tensor([ 1.8919e-04, 1.2517e-04, 9.4056e-05, 3.2926e-04, -1.0948e-03, + 2.8300e-04, 7.4983e-05], device='cuda:0') +588 +0.005224324151752577 +changing lr +epoch 34, time 788.36, cls_loss 0.0010 cls_loss_mapping 0.0114 cls_loss_causal 0.4279 re_mapping 0.0082 re_causal 0.0208 /// teacc 92.21 lr 0.00500000 +Epoch 36, weight, value: tensor([[-0.0944, -0.1094, -0.1567, ..., 0.0292, 0.0622, 0.0455], + [ 0.2489, 0.2240, 0.2618, ..., 0.0355, 0.0402, 0.0274], + [-0.0642, -0.0729, -0.0875, ..., -0.0626, -0.0819, -0.0835], + ..., + [-0.2443, -0.2269, -0.2032, ..., -0.0173, -0.0163, -0.0115], + [ 0.1058, 0.1124, 0.1281, ..., -0.0607, -0.0553, -0.0602], + [ 0.0411, 0.0385, 0.0370, ..., -0.0133, -0.0174, -0.0086]], + device='cuda:0'), grad: tensor([[ 3.0369e-05, 4.1686e-06, 1.3700e-06, ..., 1.1645e-05, + 1.3180e-05, 1.2532e-05], + [ 4.6372e-04, 2.4199e-04, 2.2376e-04, ..., 1.7905e-04, + 1.9515e-04, 1.8990e-04], + [ 6.3944e-04, 3.5572e-04, 3.3736e-04, ..., 2.2244e-04, + 2.4486e-04, 2.3806e-04], + ..., + [-2.5678e-04, -2.8476e-05, -6.0163e-06, ..., -8.7380e-05, + -1.0180e-04, -9.4712e-05], + [-5.4538e-05, -4.6074e-05, -4.0442e-05, ..., -1.2489e-06, + -1.8086e-06, -2.1346e-06], + [ 5.3406e-05, 3.3915e-05, 2.9311e-05, ..., 9.1642e-06, + 1.0468e-05, 1.0230e-05]], device='cuda:0') +Epoch 36, bias, value: tensor([ 0.0732, 0.0045, -0.0225, -0.0599, 0.0171, -0.0265, 0.0094], + device='cuda:0'), grad: tensor([ 9.6619e-05, 7.7295e-04, 9.8133e-04, -1.0500e-03, -8.4162e-04, + -2.8640e-05, 6.9499e-05], device='cuda:0') +588 +0.005000000000000003 +changing lr +epoch 35, time 784.15, cls_loss 0.0008 cls_loss_mapping 0.0103 cls_loss_causal 0.4000 re_mapping 0.0078 re_causal 0.0196 /// teacc 94.47 lr 0.00477568 +Epoch 37, weight, value: tensor([[-0.0943, -0.1092, -0.1560, ..., 0.0286, 0.0611, 0.0447], + [ 0.2470, 0.2227, 0.2600, ..., 0.0350, 0.0397, 0.0270], + [-0.0640, -0.0725, -0.0870, ..., -0.0623, -0.0815, -0.0830], + ..., + [-0.2424, -0.2254, -0.2022, ..., -0.0172, -0.0162, -0.0114], + [ 0.1049, 0.1117, 0.1273, ..., -0.0603, -0.0550, -0.0598], + [ 0.0405, 0.0380, 0.0365, ..., -0.0135, -0.0176, -0.0089]], + device='cuda:0'), grad: tensor([[ 2.5749e-04, 6.3002e-05, 3.3140e-05, ..., 9.3997e-05, + 8.0764e-05, 8.1837e-05], + [-4.2468e-05, -3.8445e-05, -3.6210e-05, ..., -1.3217e-05, + -1.4812e-05, -1.3486e-05], + [-2.6727e-04, -8.1241e-05, -4.2796e-05, ..., -1.4842e-04, + -1.1343e-04, -1.1891e-04], + ..., + [-2.5702e-04, -3.8713e-05, -1.3314e-05, ..., -4.9621e-05, + -5.5194e-05, -5.4419e-05], + [ 7.0095e-05, 1.1258e-05, 3.5018e-06, ..., 1.5974e-05, + 1.7226e-05, 1.7881e-05], + [ 4.5627e-05, 1.3471e-05, 8.6352e-06, ..., 1.5035e-05, + 1.4067e-05, 1.4104e-05]], device='cuda:0') +Epoch 37, bias, value: tensor([ 0.0725, 0.0038, -0.0224, -0.0573, 0.0177, -0.0276, 0.0085], + device='cuda:0'), grad: tensor([ 6.5470e-04, -1.8090e-05, -6.0844e-04, 4.1199e-04, -7.5769e-04, + 2.0742e-04, 1.1051e-04], device='cuda:0') +588 +0.004775675848247429 +changing lr +epoch 36, time 779.70, cls_loss 0.0008 cls_loss_mapping 0.0094 cls_loss_causal 0.4166 re_mapping 0.0075 re_causal 0.0197 /// teacc 93.22 lr 0.00455180 +Epoch 38, weight, value: tensor([[-0.0950, -0.1094, -0.1556, ..., 0.0280, 0.0601, 0.0438], + [ 0.2470, 0.2225, 0.2594, ..., 0.0350, 0.0397, 0.0272], + [-0.0644, -0.0727, -0.0869, ..., -0.0623, -0.0813, -0.0827], + ..., + [-0.2408, -0.2241, -0.2012, ..., -0.0172, -0.0164, -0.0115], + [ 0.1047, 0.1116, 0.1270, ..., -0.0596, -0.0544, -0.0591], + [ 0.0405, 0.0378, 0.0363, ..., -0.0135, -0.0174, -0.0088]], + device='cuda:0'), grad: tensor([[ 2.8133e-04, 1.6737e-04, 1.5581e-04, ..., 8.6308e-05, + 8.5950e-05, 8.3745e-05], + [-1.0338e-03, -6.2656e-04, -5.8079e-04, ..., -3.2234e-04, + -3.1805e-04, -3.0947e-04], + [ 7.2050e-04, 3.7026e-04, 3.2830e-04, ..., 2.6965e-04, + 2.7084e-04, 2.6679e-04], + ..., + [ 2.1029e-04, 9.8705e-05, 8.5533e-05, ..., 9.5487e-05, + 9.6500e-05, 9.5129e-05], + [ 1.0371e-04, 5.8979e-05, 5.4002e-05, ..., 3.2783e-05, + 3.2812e-05, 3.2067e-05], + [ 1.5438e-04, 8.7798e-05, 7.9989e-05, ..., 5.2989e-05, + 5.2631e-05, 5.1528e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([ 0.0709, 0.0059, -0.0227, -0.0581, 0.0179, -0.0277, 0.0091], + device='cuda:0'), grad: tensor([ 0.0004, -0.0016, 0.0013, -0.0009, 0.0004, 0.0002, 0.0003], + device='cuda:0') +588 +0.004551803455482836 +changing lr +epoch 37, time 775.94, cls_loss 0.0007 cls_loss_mapping 0.0081 cls_loss_causal 0.4098 re_mapping 0.0072 re_causal 0.0189 /// teacc 93.47 lr 0.00432883 +Epoch 39, weight, value: tensor([[-0.0943, -0.1091, -0.1548, ..., 0.0278, 0.0596, 0.0435], + [ 0.2452, 0.2213, 0.2577, ..., 0.0346, 0.0393, 0.0269], + [-0.0641, -0.0724, -0.0865, ..., -0.0621, -0.0809, -0.0823], + ..., + [-0.2387, -0.2225, -0.2000, ..., -0.0173, -0.0165, -0.0117], + [ 0.1044, 0.1114, 0.1267, ..., -0.0591, -0.0539, -0.0585], + [ 0.0400, 0.0374, 0.0358, ..., -0.0136, -0.0175, -0.0090]], + device='cuda:0'), grad: tensor([[ 1.2234e-05, -2.5481e-06, -1.2654e-07, ..., 7.1228e-06, + 9.0897e-06, 7.6517e-06], + [ 2.0653e-05, 3.8557e-06, 1.0664e-06, ..., 5.8673e-06, + 7.1339e-06, 5.8860e-06], + [-9.2983e-05, -4.9472e-05, -3.7372e-05, ..., -9.7230e-06, + -7.4841e-06, -7.2457e-06], + ..., + [-1.0264e-04, -8.2627e-06, 2.5867e-07, ..., -4.2975e-05, + -5.5522e-05, -4.7594e-05], + [ 2.5004e-05, 5.2787e-06, 2.5723e-06, ..., 8.2478e-06, + 1.0408e-05, 8.9854e-06], + [ 3.2425e-05, 1.0610e-05, 6.4336e-06, ..., 7.8827e-06, + 9.5293e-06, 8.3074e-06]], device='cuda:0') +Epoch 39, bias, value: tensor([ 0.0719, 0.0052, -0.0220, -0.0590, 0.0191, -0.0281, 0.0084], + device='cuda:0'), grad: tensor([ 5.7787e-05, 5.9575e-05, -1.3900e-04, 2.2423e-04, -3.5095e-04, + 7.1526e-05, 7.6413e-05], device='cuda:0') +588 +0.004328833670911726 +changing lr +epoch 38, time 784.49, cls_loss 0.0009 cls_loss_mapping 0.0107 cls_loss_causal 0.3768 re_mapping 0.0071 re_causal 0.0182 /// teacc 93.47 lr 0.00410722 +Epoch 40, weight, value: tensor([[-0.0941, -0.1091, -0.1543, ..., 0.0276, 0.0591, 0.0430], + [ 0.2440, 0.2204, 0.2564, ..., 0.0344, 0.0391, 0.0268], + [-0.0645, -0.0725, -0.0865, ..., -0.0621, -0.0809, -0.0822], + ..., + [-0.2372, -0.2211, -0.1989, ..., -0.0171, -0.0164, -0.0116], + [ 0.1047, 0.1114, 0.1266, ..., -0.0585, -0.0533, -0.0579], + [ 0.0392, 0.0369, 0.0353, ..., -0.0137, -0.0175, -0.0091]], + device='cuda:0'), grad: tensor([[ 9.7096e-05, 9.8646e-05, 1.0026e-04, ..., 1.2271e-05, + 1.5959e-05, 1.0937e-05], + [-2.9489e-05, 5.1498e-05, 4.5955e-05, ..., -1.8001e-04, + -1.9634e-04, -1.8787e-04], + [ 5.7745e-04, 2.9302e-04, 2.8849e-04, ..., 2.5678e-04, + 2.7299e-04, 2.6202e-04], + ..., + [ 5.5933e-04, 3.7813e-04, 3.7217e-04, ..., 9.0182e-05, + 8.9288e-05, 8.9407e-05], + [-1.3695e-03, -9.8991e-04, -9.7179e-04, ..., -9.5427e-05, + -8.1360e-05, -8.7738e-05], + [ 1.2732e-04, 9.3222e-05, 9.1851e-05, ..., 1.2018e-05, + 1.1064e-05, 1.1325e-05]], device='cuda:0') +Epoch 40, bias, value: tensor([ 0.0726, 0.0053, -0.0227, -0.0585, 0.0188, -0.0274, 0.0074], + device='cuda:0'), grad: tensor([-4.2170e-05, -3.4165e-04, 1.0271e-03, -1.2577e-04, 6.2037e-04, + -1.2465e-03, 1.0931e-04], device='cuda:0') +588 +0.0041072155260068206 +changing lr +epoch 39, time 782.01, cls_loss 0.0009 cls_loss_mapping 0.0088 cls_loss_causal 0.4045 re_mapping 0.0069 re_causal 0.0178 /// teacc 93.97 lr 0.00388740 +Epoch 41, weight, value: tensor([[-0.0943, -0.1093, -0.1541, ..., 0.0272, 0.0584, 0.0424], + [ 0.2439, 0.2202, 0.2557, ..., 0.0342, 0.0388, 0.0267], + [-0.0636, -0.0721, -0.0860, ..., -0.0618, -0.0804, -0.0817], + ..., + [-0.2365, -0.2203, -0.1983, ..., -0.0171, -0.0165, -0.0117], + [ 0.1047, 0.1115, 0.1265, ..., -0.0580, -0.0529, -0.0574], + [ 0.0383, 0.0363, 0.0347, ..., -0.0137, -0.0175, -0.0092]], + device='cuda:0'), grad: tensor([[ 2.4867e-04, 1.1575e-04, 7.6890e-05, ..., 8.9705e-05, + 7.6592e-05, 8.2135e-05], + [ 2.8443e-04, 9.6500e-05, 4.0561e-05, ..., 1.0222e-04, + 9.0897e-05, 9.7454e-05], + [-7.4863e-04, -3.1734e-04, -1.9884e-04, ..., -2.7394e-04, + -2.1338e-04, -2.3150e-04], + ..., + [-3.0756e-04, -1.2153e-04, -8.8871e-05, ..., -6.4611e-05, + -9.6679e-05, -9.0182e-05], + [ 1.2326e-04, 5.1200e-05, 3.6359e-05, ..., 3.2485e-05, + 3.6359e-05, 3.5733e-05], + [ 1.5604e-04, 6.8247e-05, 4.7863e-05, ..., 4.7207e-05, + 4.4048e-05, 4.5091e-05]], device='cuda:0') +Epoch 41, bias, value: tensor([ 0.0724, 0.0070, -0.0205, -0.0598, 0.0176, -0.0278, 0.0067], + device='cuda:0'), grad: tensor([ 0.0005, 0.0006, -0.0014, 0.0005, -0.0007, 0.0003, 0.0003], + device='cuda:0') +588 +0.0038873953302184317 +changing lr +epoch 40, time 784.45, cls_loss 0.0007 cls_loss_mapping 0.0081 cls_loss_causal 0.3746 re_mapping 0.0068 re_causal 0.0167 /// teacc 93.72 lr 0.00366982 +Epoch 42, weight, value: tensor([[-0.0939, -0.1090, -0.1533, ..., 0.0269, 0.0578, 0.0420], + [ 0.2415, 0.2186, 0.2539, ..., 0.0337, 0.0383, 0.0263], + [-0.0639, -0.0721, -0.0860, ..., -0.0617, -0.0803, -0.0816], + ..., + [-0.2352, -0.2192, -0.1975, ..., -0.0172, -0.0167, -0.0118], + [ 0.1048, 0.1115, 0.1263, ..., -0.0575, -0.0524, -0.0569], + [ 0.0390, 0.0366, 0.0349, ..., -0.0136, -0.0172, -0.0090]], + device='cuda:0'), grad: tensor([[-7.7009e-05, -4.2282e-06, 2.4810e-06, ..., -4.7326e-05, + -5.2571e-05, -4.8190e-05], + [-9.6500e-05, -9.1612e-05, -8.8453e-05, ..., 1.2748e-05, + 9.9018e-06, 9.6783e-06], + [ 3.0492e-06, -1.2167e-05, -1.5259e-05, ..., 7.4804e-06, + 8.8289e-06, 8.7321e-06], + ..., + [ 6.4194e-05, 5.1707e-05, 4.8608e-05, ..., 4.0047e-06, + 5.0478e-06, 3.5167e-06], + [ 4.5717e-05, 2.8208e-05, 2.7463e-05, ..., 7.7859e-06, + 9.8199e-06, 8.8662e-06], + [-4.8764e-06, -4.0755e-06, -3.5353e-06, ..., -4.3809e-06, + -3.9600e-06, -4.5821e-06]], device='cuda:0') +Epoch 42, bias, value: tensor([ 0.0725, 0.0055, -0.0207, -0.0594, 0.0175, -0.0276, 0.0076], + device='cuda:0'), grad: tensor([-2.5010e-04, -2.9832e-05, 5.4538e-05, 1.1677e-04, 4.9055e-05, + 6.1452e-05, -2.0377e-06], device='cuda:0') +588 +0.003669815772166629 +changing lr +epoch 41, time 785.42, cls_loss 0.0008 cls_loss_mapping 0.0086 cls_loss_causal 0.3795 re_mapping 0.0068 re_causal 0.0172 /// teacc 93.47 lr 0.00345492 +Epoch 43, weight, value: tensor([[-0.0945, -0.1091, -0.1530, ..., 0.0265, 0.0571, 0.0414], + [ 0.2417, 0.2184, 0.2533, ..., 0.0337, 0.0384, 0.0265], + [-0.0639, -0.0720, -0.0858, ..., -0.0617, -0.0801, -0.0814], + ..., + [-0.2336, -0.2181, -0.1966, ..., -0.0170, -0.0166, -0.0117], + [ 0.1039, 0.1109, 0.1257, ..., -0.0571, -0.0521, -0.0565], + [ 0.0391, 0.0366, 0.0348, ..., -0.0135, -0.0171, -0.0090]], + device='cuda:0'), grad: tensor([[-2.8759e-05, -3.9190e-06, -2.4792e-06, ..., -1.1273e-05, + -1.5885e-05, -1.3143e-05], + [-2.1607e-05, -2.6926e-05, -2.3410e-05, ..., -3.2067e-05, + -3.2812e-05, -3.1412e-05], + [-3.6907e-04, -1.4079e-04, -1.2422e-04, ..., -7.2420e-05, + -9.6321e-05, -9.0182e-05], + ..., + [ 1.5509e-04, 7.0274e-05, 6.0767e-05, ..., 4.3333e-05, + 5.4598e-05, 5.1677e-05], + [ 4.4584e-05, 2.0862e-05, 1.7837e-05, ..., 1.0572e-05, + 1.3895e-05, 1.3337e-05], + [-9.7394e-05, -5.5313e-05, -4.6998e-05, ..., -1.6689e-05, + -2.3067e-05, -2.4334e-05]], device='cuda:0') +Epoch 43, bias, value: tensor([ 0.0706, 0.0078, -0.0209, -0.0599, 0.0187, -0.0288, 0.0080], + device='cuda:0'), grad: tensor([-9.0897e-05, 2.2352e-06, -6.7520e-04, 5.5695e-04, 2.7084e-04, + 7.7784e-05, -1.4234e-04], device='cuda:0') +588 +0.0034549150281252667 +changing lr +epoch 42, time 781.37, cls_loss 0.0008 cls_loss_mapping 0.0091 cls_loss_causal 0.3906 re_mapping 0.0066 re_causal 0.0175 /// teacc 93.47 lr 0.00324313 +Epoch 44, weight, value: tensor([[-0.0937, -0.1090, -0.1526, ..., 0.0266, 0.0569, 0.0413], + [ 0.2405, 0.2179, 0.2524, ..., 0.0333, 0.0380, 0.0262], + [-0.0640, -0.0720, -0.0857, ..., -0.0615, -0.0798, -0.0811], + ..., + [-0.2324, -0.2171, -0.1957, ..., -0.0170, -0.0166, -0.0118], + [ 0.1045, 0.1112, 0.1259, ..., -0.0567, -0.0517, -0.0561], + [ 0.0380, 0.0358, 0.0342, ..., -0.0137, -0.0172, -0.0091]], + device='cuda:0'), grad: tensor([[ 4.3178e-04, 1.4257e-04, 1.1826e-04, ..., 1.0210e-04, + 1.0818e-04, 9.1910e-05], + [ 3.0422e-04, 1.4377e-04, 1.3006e-04, ..., 5.9456e-05, + 6.4552e-05, 5.4389e-05], + [-9.0218e-04, -2.2757e-04, -1.5807e-04, ..., -3.4285e-04, + -3.5477e-04, -3.3140e-04], + ..., + [ 3.6502e-04, 1.1307e-04, 9.7275e-05, ..., 1.0073e-04, + 1.0383e-04, 9.0361e-05], + [ 2.5129e-04, 1.1909e-04, 1.0931e-04, ..., 4.1991e-05, + 4.6015e-05, 3.6687e-05], + [-1.3657e-03, -6.9523e-04, -6.5088e-04, ..., -1.8764e-04, + -2.1398e-04, -1.5986e-04]], device='cuda:0') +Epoch 44, bias, value: tensor([ 0.0728, 0.0069, -0.0209, -0.0608, 0.0186, -0.0277, 0.0067], + device='cuda:0'), grad: tensor([ 0.0010, 0.0005, -0.0023, 0.0017, 0.0008, 0.0004, -0.0022], + device='cuda:0') +588 +0.0032431258795932905 +changing lr +---------------------saving model at epoch 43---------------------------------------------------- +epoch 43, time 780.99, cls_loss 0.0007 cls_loss_mapping 0.0082 cls_loss_causal 0.3753 re_mapping 0.0066 re_causal 0.0167 /// teacc 95.23 lr 0.00303487 +Epoch 45, weight, value: tensor([[-0.0933, -0.1088, -0.1521, ..., 0.0264, 0.0566, 0.0410], + [ 0.2397, 0.2173, 0.2516, ..., 0.0330, 0.0377, 0.0260], + [-0.0641, -0.0719, -0.0855, ..., -0.0614, -0.0796, -0.0808], + ..., + [-0.2312, -0.2160, -0.1950, ..., -0.0170, -0.0167, -0.0118], + [ 0.1041, 0.1109, 0.1255, ..., -0.0563, -0.0514, -0.0557], + [ 0.0376, 0.0355, 0.0338, ..., -0.0136, -0.0171, -0.0091]], + device='cuda:0'), grad: tensor([[ 9.8526e-05, 5.9068e-05, 5.9336e-05, ..., 8.7097e-06, + 7.2382e-06, 8.2776e-06], + [ 1.6838e-06, -8.2180e-06, -8.1658e-06, ..., -2.8566e-05, + -3.1203e-05, -3.2872e-05], + [-4.6396e-04, -1.2732e-04, -8.7261e-05, ..., -1.0407e-04, + -8.4102e-05, -8.1182e-05], + ..., + [ 2.9778e-04, 1.4830e-04, 1.3936e-04, ..., 5.0753e-05, + 5.2691e-05, 5.0783e-05], + [-6.5565e-04, -4.4060e-04, -4.5657e-04, ..., -3.6985e-05, + -5.6416e-05, -5.3614e-05], + [ 3.3355e-04, 1.9693e-04, 1.9622e-04, ..., 3.7581e-05, + 4.3303e-05, 4.1753e-05]], device='cuda:0') +Epoch 45, bias, value: tensor([ 0.0736, 0.0071, -0.0210, -0.0604, 0.0186, -0.0283, 0.0062], + device='cuda:0'), grad: tensor([ 1.1885e-04, 7.0035e-06, -1.1644e-03, 7.2956e-04, 4.9019e-04, + -5.9938e-04, 4.1986e-04], device='cuda:0') +588 +0.0030348748417303863 +changing lr +epoch 44, time 779.23, cls_loss 0.0006 cls_loss_mapping 0.0056 cls_loss_causal 0.3810 re_mapping 0.0065 re_causal 0.0167 /// teacc 92.96 lr 0.00283058 +Epoch 46, weight, value: tensor([[-0.0944, -0.1093, -0.1522, ..., 0.0260, 0.0559, 0.0404], + [ 0.2392, 0.2169, 0.2509, ..., 0.0330, 0.0376, 0.0261], + [-0.0635, -0.0717, -0.0852, ..., -0.0610, -0.0792, -0.0804], + ..., + [-0.2298, -0.2150, -0.1942, ..., -0.0170, -0.0166, -0.0118], + [ 0.1042, 0.1109, 0.1254, ..., -0.0559, -0.0511, -0.0553], + [ 0.0374, 0.0354, 0.0337, ..., -0.0136, -0.0170, -0.0091]], + device='cuda:0'), grad: tensor([[-3.5381e-04, -6.0469e-05, -4.0948e-05, ..., -8.4877e-05, + -1.1462e-04, -1.0496e-04], + [ 5.4419e-05, -6.3516e-06, -8.9481e-06, ..., 3.1859e-05, + 3.7342e-05, 3.5703e-05], + [-5.6326e-05, -2.5675e-05, -1.9148e-05, ..., -1.3374e-06, + 1.3635e-06, 1.7937e-06], + ..., + [ 2.6846e-04, 8.0526e-05, 6.6280e-05, ..., 7.5161e-05, + 9.1076e-05, 8.5056e-05], + [ 5.8591e-05, 1.6183e-05, 1.2800e-05, ..., 1.4663e-05, + 1.8165e-05, 1.6823e-05], + [ 9.3877e-05, 3.3319e-05, 2.6554e-05, ..., 1.7792e-05, + 2.0459e-05, 1.8761e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([ 0.0715, 0.0075, -0.0197, -0.0609, 0.0193, -0.0281, 0.0061], + device='cuda:0'), grad: tensor([-1.0271e-03, 2.0909e-04, -1.0324e-04, -9.7573e-05, 6.5947e-04, + 1.4853e-04, 2.1064e-04], device='cuda:0') +588 +0.0028305813044122124 +changing lr +epoch 45, time 782.16, cls_loss 0.0006 cls_loss_mapping 0.0065 cls_loss_causal 0.4017 re_mapping 0.0063 re_causal 0.0168 /// teacc 94.47 lr 0.00263066 +Epoch 47, weight, value: tensor([[-0.0943, -0.1093, -0.1519, ..., 0.0259, 0.0556, 0.0402], + [ 0.2384, 0.2164, 0.2502, ..., 0.0327, 0.0374, 0.0259], + [-0.0635, -0.0715, -0.0850, ..., -0.0609, -0.0790, -0.0801], + ..., + [-0.2290, -0.2144, -0.1938, ..., -0.0170, -0.0166, -0.0118], + [ 0.1040, 0.1109, 0.1253, ..., -0.0556, -0.0508, -0.0550], + [ 0.0371, 0.0351, 0.0335, ..., -0.0136, -0.0170, -0.0092]], + device='cuda:0'), grad: tensor([[ 1.0943e-04, 9.1255e-05, 9.0539e-05, ..., 3.6418e-05, + 4.6253e-05, 3.9160e-05], + [-2.0409e-04, -1.3947e-04, -1.3614e-04, ..., -6.1810e-05, + -7.4685e-05, -6.5446e-05], + [ 1.5408e-05, 9.5367e-06, 9.0674e-06, ..., 3.7346e-06, + 4.6231e-06, 3.9823e-06], + ..., + [ 1.1688e-04, 8.0884e-05, 7.8976e-05, ..., 6.3926e-06, + 8.2701e-06, 8.3894e-06], + [-1.4687e-04, -1.0443e-04, -1.0210e-04, ..., -9.9912e-06, + -1.4298e-05, -1.2837e-05], + [ 4.1425e-05, 2.1636e-05, 2.0519e-05, ..., 7.1786e-06, + 8.2031e-06, 7.6443e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([ 0.0718, 0.0074, -0.0197, -0.0606, 0.0195, -0.0284, 0.0057], + device='cuda:0'), grad: tensor([ 6.6936e-05, -2.3532e-04, 2.1771e-05, 9.8288e-05, 1.0723e-04, + -1.2827e-04, 6.9022e-05], device='cuda:0') +588 +0.0026306566876350096 +changing lr +epoch 46, time 780.02, cls_loss 0.0009 cls_loss_mapping 0.0073 cls_loss_causal 0.3808 re_mapping 0.0062 re_causal 0.0164 /// teacc 94.22 lr 0.00243550 +Epoch 48, weight, value: tensor([[-0.0945, -0.1096, -0.1519, ..., 0.0256, 0.0551, 0.0398], + [ 0.2374, 0.2158, 0.2494, ..., 0.0326, 0.0371, 0.0258], + [-0.0631, -0.0713, -0.0848, ..., -0.0607, -0.0787, -0.0798], + ..., + [-0.2289, -0.2139, -0.1934, ..., -0.0172, -0.0169, -0.0121], + [ 0.1045, 0.1112, 0.1255, ..., -0.0553, -0.0505, -0.0547], + [ 0.0374, 0.0351, 0.0334, ..., -0.0133, -0.0167, -0.0089]], + device='cuda:0'), grad: tensor([[ 4.6939e-05, 2.9400e-05, 3.1143e-05, ..., 2.5705e-06, + 4.1761e-06, 3.1665e-06], + [-1.9699e-05, -2.8163e-05, -2.9176e-05, ..., -2.3488e-06, + -3.1944e-06, -4.9844e-06], + [-1.8775e-04, -5.6952e-05, -5.0545e-05, ..., -3.3647e-05, + -4.2319e-05, -3.2812e-05], + ..., + [ 3.1292e-05, 1.6630e-05, 1.5453e-05, ..., 9.0823e-06, + 9.8944e-06, 9.2238e-06], + [ 2.7731e-05, 8.2776e-06, 6.8024e-06, ..., 7.1861e-06, + 8.4415e-06, 7.1935e-06], + [ 7.0155e-05, 2.0653e-05, 1.6779e-05, ..., 1.8433e-05, + 2.2486e-05, 1.8969e-05]], device='cuda:0') +Epoch 48, bias, value: tensor([ 0.0715, 0.0070, -0.0184, -0.0608, 0.0178, -0.0278, 0.0066], + device='cuda:0'), grad: tensor([ 5.7459e-05, 2.7984e-05, -4.3344e-04, 7.0810e-05, 4.8548e-05, + 6.5327e-05, 1.6344e-04], device='cuda:0') +588 +0.0024355036129704724 +changing lr +epoch 47, time 777.64, cls_loss 0.0007 cls_loss_mapping 0.0077 cls_loss_causal 0.3881 re_mapping 0.0063 re_causal 0.0164 /// teacc 93.97 lr 0.00224552 +Epoch 49, weight, value: tensor([[-0.0944, -0.1097, -0.1519, ..., 0.0254, 0.0548, 0.0395], + [ 0.2377, 0.2158, 0.2492, ..., 0.0327, 0.0372, 0.0260], + [-0.0629, -0.0712, -0.0846, ..., -0.0606, -0.0785, -0.0797], + ..., + [-0.2282, -0.2132, -0.1928, ..., -0.0173, -0.0169, -0.0121], + [ 0.1047, 0.1113, 0.1255, ..., -0.0550, -0.0502, -0.0544], + [ 0.0365, 0.0345, 0.0329, ..., -0.0135, -0.0168, -0.0090]], + device='cuda:0'), grad: tensor([[ 8.9931e-04, 2.2614e-04, 1.8775e-04, ..., 3.4213e-04, + 2.8157e-04, 2.9850e-04], + [ 2.0866e-03, 6.0844e-04, 4.9543e-04, ..., 8.4305e-04, + 6.9046e-04, 7.1907e-04], + [-8.8272e-03, -2.4700e-03, -2.0409e-03, ..., -3.5534e-03, + -2.8877e-03, -3.0384e-03], + ..., + [ 2.6169e-03, 6.5708e-04, 5.7602e-04, ..., 1.0681e-03, + 8.5640e-04, 9.1982e-04], + [ 3.8743e-04, 1.1140e-04, 9.3937e-05, ..., 1.4520e-04, + 1.1879e-04, 1.2517e-04], + [ 3.3474e-04, 9.4175e-05, 6.2168e-05, ..., 1.6975e-04, + 1.3697e-04, 1.4067e-04]], device='cuda:0') +Epoch 49, bias, value: tensor([ 0.0722, 0.0084, -0.0180, -0.0616, 0.0171, -0.0276, 0.0054], + device='cuda:0'), grad: tensor([ 0.0023, 0.0051, -0.0218, 0.0059, 0.0067, 0.0009, 0.0008], + device='cuda:0') +588 +0.00224551509273949 +changing lr +epoch 48, time 781.02, cls_loss 0.0008 cls_loss_mapping 0.0071 cls_loss_causal 0.3980 re_mapping 0.0061 re_causal 0.0164 /// teacc 94.47 lr 0.00206107 +Epoch 50, weight, value: tensor([[-0.0949, -0.1100, -0.1519, ..., 0.0253, 0.0545, 0.0393], + [ 0.2367, 0.2151, 0.2484, ..., 0.0324, 0.0370, 0.0258], + [-0.0635, -0.0714, -0.0848, ..., -0.0607, -0.0787, -0.0798], + ..., + [-0.2271, -0.2124, -0.1922, ..., -0.0171, -0.0167, -0.0119], + [ 0.1045, 0.1112, 0.1253, ..., -0.0548, -0.0501, -0.0542], + [ 0.0373, 0.0349, 0.0332, ..., -0.0133, -0.0167, -0.0090]], + device='cuda:0'), grad: tensor([[-7.2360e-05, 4.4256e-06, 1.0900e-05, ..., -6.4969e-06, + -1.2942e-05, -2.9765e-06], + [ 1.5318e-04, 6.4552e-05, 5.5045e-05, ..., 4.1813e-05, + 4.8548e-05, 4.5240e-05], + [-3.1233e-04, -1.2100e-04, -1.0043e-04, ..., -7.5161e-05, + -9.3043e-05, -9.1016e-05], + ..., + [ 1.1617e-04, 2.2396e-05, 1.2688e-05, ..., 1.7375e-05, + 2.9966e-05, 2.3380e-05], + [ 4.3690e-05, 1.3724e-05, 1.0848e-05, ..., 1.0937e-05, + 1.2949e-05, 1.1988e-05], + [ 4.4823e-05, 1.2606e-05, 9.6634e-06, ..., 1.1720e-05, + 1.3515e-05, 1.2405e-05]], device='cuda:0') +Epoch 50, bias, value: tensor([ 0.0712, 0.0082, -0.0193, -0.0609, 0.0177, -0.0280, 0.0069], + device='cuda:0'), grad: tensor([-2.9421e-04, 2.9659e-04, -6.4230e-04, 8.2076e-05, 3.4451e-04, + 1.0306e-04, 1.1146e-04], device='cuda:0') +588 +0.002061073738537637 +changing lr +epoch 49, time 775.85, cls_loss 0.0007 cls_loss_mapping 0.0059 cls_loss_causal 0.3879 re_mapping 0.0061 re_causal 0.0162 /// teacc 93.72 lr 0.00188255 +Epoch 51, weight, value: tensor([[-0.0950, -0.1101, -0.1517, ..., 0.0251, 0.0542, 0.0390], + [ 0.2366, 0.2150, 0.2481, ..., 0.0324, 0.0370, 0.0258], + [-0.0631, -0.0713, -0.0846, ..., -0.0605, -0.0785, -0.0796], + ..., + [-0.2264, -0.2118, -0.1918, ..., -0.0171, -0.0167, -0.0120], + [ 0.1043, 0.1110, 0.1252, ..., -0.0546, -0.0499, -0.0539], + [ 0.0366, 0.0345, 0.0329, ..., -0.0134, -0.0167, -0.0091]], + device='cuda:0'), grad: tensor([[-2.4974e-05, -9.4399e-06, -9.8273e-06, ..., -2.8126e-07, + -4.0084e-06, -2.3097e-06], + [ 1.2159e-05, 4.7684e-06, 4.3511e-06, ..., 4.4964e-06, + 5.4128e-06, 4.9211e-06], + [-6.6543e-07, -5.6531e-07, -4.2899e-08, ..., 4.2608e-07, + 1.3607e-06, 1.2927e-06], + ..., + [-2.8894e-07, 3.3490e-06, 3.9898e-06, ..., 7.8813e-08, + 4.0256e-07, -2.2433e-07], + [ 8.4564e-06, 2.5351e-06, 2.1197e-06, ..., 2.7735e-06, + 3.3230e-06, 3.1330e-06], + [ 1.3173e-05, 4.8317e-06, 4.3996e-06, ..., 3.3658e-06, + 4.4480e-06, 3.8967e-06]], device='cuda:0') +Epoch 51, bias, value: tensor([ 0.0709, 0.0088, -0.0183, -0.0611, 0.0180, -0.0283, 0.0059], + device='cuda:0'), grad: tensor([-5.8919e-05, 2.6375e-05, 5.8254e-07, -6.9402e-06, -1.1913e-05, + 2.0921e-05, 3.0026e-05], device='cuda:0') +588 +0.0018825509907063344 +changing lr +epoch 50, time 778.58, cls_loss 0.0006 cls_loss_mapping 0.0055 cls_loss_causal 0.3620 re_mapping 0.0062 re_causal 0.0157 /// teacc 93.22 lr 0.00171031 +Epoch 52, weight, value: tensor([[-0.0944, -0.1098, -0.1513, ..., 0.0251, 0.0540, 0.0389], + [ 0.2356, 0.2144, 0.2473, ..., 0.0322, 0.0368, 0.0257], + [-0.0636, -0.0714, -0.0846, ..., -0.0605, -0.0785, -0.0795], + ..., + [-0.2261, -0.2114, -0.1915, ..., -0.0171, -0.0168, -0.0120], + [ 0.1048, 0.1113, 0.1254, ..., -0.0543, -0.0495, -0.0536], + [ 0.0365, 0.0344, 0.0327, ..., -0.0134, -0.0167, -0.0091]], + device='cuda:0'), grad: tensor([[ 7.1585e-05, 2.3738e-05, 2.2829e-05, ..., 2.7627e-05, + 3.2187e-05, 2.9474e-05], + [-7.3388e-06, -1.5572e-05, -1.5676e-05, ..., 1.5974e-05, + 1.7300e-05, 1.5572e-05], + [ 2.3872e-05, 6.3367e-06, 6.0759e-06, ..., 1.5751e-05, + 1.7658e-05, 1.6004e-05], + ..., + [-1.5843e-04, -4.1500e-06, 2.1309e-06, ..., -1.1569e-04, + -1.3137e-04, -1.2040e-04], + [-1.1539e-04, -9.3699e-05, -9.7096e-05, ..., 2.1607e-06, + 1.8592e-07, 1.5637e-06], + [ 6.2466e-05, 3.2395e-05, 3.2336e-05, ..., 1.4745e-05, + 1.7583e-05, 1.5691e-05]], device='cuda:0') +Epoch 52, bias, value: tensor([ 0.0720, 0.0081, -0.0192, -0.0609, 0.0174, -0.0273, 0.0058], + device='cuda:0'), grad: tensor([ 1.7464e-04, 4.0978e-05, 6.9141e-05, 2.5940e-04, -5.9509e-04, + -5.3793e-05, 1.0496e-04], device='cuda:0') +588 +0.0017103063703014388 +changing lr +epoch 51, time 777.04, cls_loss 0.0006 cls_loss_mapping 0.0064 cls_loss_causal 0.3832 re_mapping 0.0060 re_causal 0.0157 /// teacc 93.22 lr 0.00154469 +Epoch 53, weight, value: tensor([[-0.0944, -0.1098, -0.1512, ..., 0.0250, 0.0538, 0.0388], + [ 0.2352, 0.2141, 0.2469, ..., 0.0321, 0.0367, 0.0256], + [-0.0636, -0.0714, -0.0846, ..., -0.0604, -0.0783, -0.0793], + ..., + [-0.2253, -0.2108, -0.1910, ..., -0.0170, -0.0167, -0.0120], + [ 0.1046, 0.1112, 0.1252, ..., -0.0541, -0.0494, -0.0534], + [ 0.0368, 0.0345, 0.0328, ..., -0.0134, -0.0167, -0.0091]], + device='cuda:0'), grad: tensor([[-1.1519e-05, -1.4165e-06, 1.2703e-06, ..., 1.1958e-05, + 9.8720e-06, 1.2934e-05], + [ 5.2691e-05, 2.5686e-06, -4.5896e-06, ..., 2.7329e-05, + 2.8938e-05, 2.8655e-05], + [ 1.1331e-04, 2.3946e-05, 7.9796e-06, ..., 4.3511e-05, + 4.8459e-05, 4.8310e-05], + ..., + [-3.5477e-04, -7.0393e-05, -2.2277e-05, ..., -1.5438e-04, + -1.6773e-04, -1.6880e-04], + [ 7.0035e-05, 1.5274e-05, 4.5523e-06, ..., 2.5392e-05, + 2.8521e-05, 2.8431e-05], + [ 4.6134e-05, 1.1250e-05, 5.2303e-06, ..., 1.6332e-05, + 1.8328e-05, 1.7807e-05]], device='cuda:0') +Epoch 53, bias, value: tensor([ 0.0719, 0.0082, -0.0191, -0.0616, 0.0177, -0.0275, 0.0063], + device='cuda:0'), grad: tensor([-2.3514e-05, 1.7953e-04, 3.2234e-04, 2.3305e-04, -1.0347e-03, + 1.9968e-04, 1.2434e-04], device='cuda:0') +588 +0.0015446867550656784 +changing lr +epoch 52, time 782.09, cls_loss 0.0006 cls_loss_mapping 0.0067 cls_loss_causal 0.3884 re_mapping 0.0060 re_causal 0.0158 /// teacc 92.71 lr 0.00138603 +Epoch 54, weight, value: tensor([[-0.0945, -0.1098, -0.1511, ..., 0.0248, 0.0536, 0.0386], + [ 0.2345, 0.2137, 0.2463, ..., 0.0319, 0.0365, 0.0255], + [-0.0630, -0.0711, -0.0842, ..., -0.0602, -0.0780, -0.0791], + ..., + [-0.2252, -0.2105, -0.1909, ..., -0.0172, -0.0169, -0.0121], + [ 0.1048, 0.1114, 0.1254, ..., -0.0539, -0.0492, -0.0532], + [ 0.0366, 0.0343, 0.0326, ..., -0.0134, -0.0166, -0.0090]], + device='cuda:0'), grad: tensor([[-7.3338e-04, -2.7394e-04, -2.6226e-04, ..., -2.8658e-04, + -3.0828e-04, -2.6822e-04], + [ 2.9230e-04, 1.0693e-04, 9.9599e-05, ..., 9.7156e-05, + 1.0562e-04, 9.0897e-05], + [ 1.4770e-04, 4.1157e-05, 3.7640e-05, ..., 7.1108e-05, + 7.1049e-05, 6.4492e-05], + ..., + [-2.3529e-05, 1.5661e-05, 2.3380e-05, ..., 1.3188e-05, + 1.9163e-05, 1.5557e-05], + [ 7.1764e-05, 2.5257e-05, 2.3022e-05, ..., 1.9610e-05, + 2.1726e-05, 1.8224e-05], + [ 1.2422e-04, 4.5151e-05, 4.1783e-05, ..., 3.5584e-05, + 3.9577e-05, 3.3230e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([ 0.0715, 0.0078, -0.0179, -0.0614, 0.0171, -0.0274, 0.0062], + device='cuda:0'), grad: tensor([-0.0017, 0.0007, 0.0004, 0.0003, -0.0001, 0.0002, 0.0003], + device='cuda:0') +588 +0.001386025680863044 +changing lr +epoch 53, time 774.24, cls_loss 0.0005 cls_loss_mapping 0.0052 cls_loss_causal 0.3412 re_mapping 0.0059 re_causal 0.0145 /// teacc 92.21 lr 0.00123464 +Epoch 55, weight, value: tensor([[-0.0946, -0.1099, -0.1510, ..., 0.0247, 0.0534, 0.0384], + [ 0.2346, 0.2136, 0.2462, ..., 0.0320, 0.0365, 0.0256], + [-0.0629, -0.0711, -0.0842, ..., -0.0601, -0.0780, -0.0790], + ..., + [-0.2248, -0.2101, -0.1905, ..., -0.0172, -0.0170, -0.0122], + [ 0.1044, 0.1110, 0.1250, ..., -0.0538, -0.0491, -0.0531], + [ 0.0365, 0.0342, 0.0325, ..., -0.0134, -0.0166, -0.0091]], + device='cuda:0'), grad: tensor([[ 2.2089e-04, 1.0550e-04, 1.0586e-04, ..., 1.3101e-04, + 1.3411e-04, 1.3399e-04], + [-4.0889e-04, -2.0576e-04, -2.0719e-04, ..., -2.4796e-04, + -2.4986e-04, -2.5177e-04], + [ 9.8050e-05, 4.5896e-05, 4.5389e-05, ..., 6.5923e-05, + 6.7651e-05, 6.7174e-05], + ..., + [ 4.8250e-05, 3.1799e-05, 3.1918e-05, ..., 3.7879e-05, + 3.5912e-05, 3.7074e-05], + [-7.3984e-06, -1.1280e-05, -1.1750e-05, ..., 1.0759e-05, + 1.0639e-05, 1.0796e-05], + [ 4.4703e-05, 2.5362e-05, 2.5615e-05, ..., 1.9953e-05, + 2.0534e-05, 2.0444e-05]], device='cuda:0') +Epoch 55, bias, value: tensor([ 0.0714, 0.0085, -0.0177, -0.0612, 0.0167, -0.0280, 0.0062], + device='cuda:0'), grad: tensor([ 4.9305e-04, -8.8120e-04, 2.1744e-04, -5.1521e-06, 7.4267e-05, + 2.2992e-05, 7.9215e-05], device='cuda:0') +588 +0.0012346426699819469 +changing lr +epoch 54, time 782.77, cls_loss 0.0006 cls_loss_mapping 0.0059 cls_loss_causal 0.3807 re_mapping 0.0058 re_causal 0.0151 /// teacc 92.21 lr 0.00109084 +Epoch 56, weight, value: tensor([[-0.0947, -0.1100, -0.1509, ..., 0.0246, 0.0531, 0.0382], + [ 0.2345, 0.2135, 0.2460, ..., 0.0320, 0.0365, 0.0256], + [-0.0628, -0.0709, -0.0841, ..., -0.0601, -0.0778, -0.0788], + ..., + [-0.2242, -0.2097, -0.1902, ..., -0.0172, -0.0169, -0.0122], + [ 0.1043, 0.1110, 0.1249, ..., -0.0537, -0.0490, -0.0530], + [ 0.0362, 0.0340, 0.0323, ..., -0.0135, -0.0166, -0.0091]], + device='cuda:0'), grad: tensor([[ 2.7627e-05, 2.1264e-05, 1.8045e-05, ..., 2.0955e-06, + 3.7104e-06, 3.7644e-06], + [ 1.9491e-05, -3.7819e-05, -4.5240e-05, ..., 1.8194e-05, + 1.9312e-05, 1.8984e-05], + [-5.8532e-05, -1.4134e-05, -8.1658e-06, ..., -9.3356e-06, + -8.9183e-06, -8.4117e-06], + ..., + [-1.6868e-04, -4.2349e-05, -2.1830e-05, ..., -6.6578e-05, + -7.4983e-05, -7.2956e-05], + [ 3.6687e-05, 1.1131e-05, 7.0967e-06, ..., 1.2882e-05, + 1.4156e-05, 1.3798e-05], + [ 4.1872e-05, 1.5736e-05, 1.2420e-05, ..., 1.2293e-05, + 1.3627e-05, 1.2979e-05]], device='cuda:0') +Epoch 56, bias, value: tensor([ 0.0712, 0.0088, -0.0173, -0.0614, 0.0170, -0.0281, 0.0057], + device='cuda:0'), grad: tensor([ 1.0133e-06, 2.3293e-04, -1.2696e-04, 1.8799e-04, -4.7779e-04, + 9.3281e-05, 8.9526e-05], device='cuda:0') +588 +0.0010908425876598518 +changing lr +epoch 55, time 782.23, cls_loss 0.0006 cls_loss_mapping 0.0059 cls_loss_causal 0.3434 re_mapping 0.0058 re_causal 0.0145 /// teacc 93.72 lr 0.00095492 +Epoch 57, weight, value: tensor([[-0.0947, -0.1100, -0.1508, ..., 0.0245, 0.0530, 0.0381], + [ 0.2342, 0.2133, 0.2457, ..., 0.0319, 0.0364, 0.0256], + [-0.0631, -0.0710, -0.0841, ..., -0.0601, -0.0778, -0.0788], + ..., + [-0.2236, -0.2093, -0.1898, ..., -0.0171, -0.0169, -0.0121], + [ 0.1040, 0.1108, 0.1247, ..., -0.0536, -0.0490, -0.0529], + [ 0.0363, 0.0339, 0.0323, ..., -0.0134, -0.0166, -0.0091]], + device='cuda:0'), grad: tensor([[-3.4332e-04, -1.1641e-04, -8.1122e-05, ..., -6.3300e-05, + -6.0499e-05, -4.4078e-05], + [ 3.5620e-04, 1.1104e-04, 8.8036e-05, ..., 9.5963e-05, + 1.0800e-04, 9.4950e-05], + [ 1.6856e-04, 3.9130e-05, 2.7493e-05, ..., 7.2122e-05, + 8.4817e-05, 8.2672e-05], + ..., + [-4.0865e-04, -5.9426e-05, -3.9786e-05, ..., -1.6725e-04, + -2.0897e-04, -2.0361e-04], + [ 8.7440e-05, 2.0519e-05, 1.4551e-05, ..., 2.9668e-05, + 3.5048e-05, 3.2812e-05], + [ 1.0663e-04, 3.4690e-05, 2.8372e-05, ..., 3.4213e-05, + 3.9250e-05, 3.5703e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([ 0.0711, 0.0088, -0.0179, -0.0613, 0.0175, -0.0285, 0.0063], + device='cuda:0'), grad: tensor([-0.0007, 0.0008, 0.0005, 0.0002, -0.0013, 0.0002, 0.0002], + device='cuda:0') +588 +0.000954915028125264 +changing lr +epoch 56, time 778.58, cls_loss 0.0006 cls_loss_mapping 0.0045 cls_loss_causal 0.3808 re_mapping 0.0057 re_causal 0.0148 /// teacc 93.72 lr 0.00082713 +Epoch 58, weight, value: tensor([[-0.0945, -0.1099, -0.1507, ..., 0.0246, 0.0530, 0.0381], + [ 0.2335, 0.2129, 0.2452, ..., 0.0317, 0.0362, 0.0254], + [-0.0632, -0.0711, -0.0841, ..., -0.0601, -0.0778, -0.0788], + ..., + [-0.2232, -0.2089, -0.1896, ..., -0.0172, -0.0169, -0.0122], + [ 0.1043, 0.1110, 0.1249, ..., -0.0535, -0.0489, -0.0528], + [ 0.0361, 0.0338, 0.0322, ..., -0.0135, -0.0166, -0.0091]], + device='cuda:0'), grad: tensor([[-1.6630e-04, 6.9011e-07, 1.9506e-05, ..., -3.3617e-05, + -3.9756e-05, -3.4124e-05], + [-3.8166e-03, -2.1992e-03, -2.1572e-03, ..., -1.2531e-03, + -1.4200e-03, -1.2770e-03], + [ 8.0824e-04, 4.3297e-04, 4.1485e-04, ..., 3.1972e-04, + 3.5095e-04, 3.1400e-04], + ..., + [ 1.2236e-03, 6.4898e-04, 6.3467e-04, ..., 2.1338e-04, + 2.7490e-04, 2.5058e-04], + [ 2.1064e-04, 1.1522e-04, 1.1188e-04, ..., 6.6459e-05, + 7.5638e-05, 6.7890e-05], + [ 6.1464e-04, 3.5548e-04, 3.4618e-04, ..., 2.4509e-04, + 2.6989e-04, 2.4176e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([ 0.0715, 0.0081, -0.0182, -0.0611, 0.0178, -0.0282, 0.0060], + device='cuda:0'), grad: tensor([-0.0006, -0.0057, 0.0014, 0.0017, 0.0019, 0.0003, 0.0009], + device='cuda:0') +588 +0.0008271337313934874 +changing lr +epoch 57, time 809.48, cls_loss 0.0005 cls_loss_mapping 0.0042 cls_loss_causal 0.3936 re_mapping 0.0057 re_causal 0.0150 /// teacc 93.22 lr 0.00070776 +Epoch 59, weight, value: tensor([[-0.0946, -0.1099, -0.1506, ..., 0.0245, 0.0529, 0.0380], + [ 0.2334, 0.2128, 0.2451, ..., 0.0317, 0.0362, 0.0254], + [-0.0631, -0.0711, -0.0841, ..., -0.0600, -0.0777, -0.0787], + ..., + [-0.2230, -0.2087, -0.1894, ..., -0.0172, -0.0169, -0.0122], + [ 0.1041, 0.1108, 0.1247, ..., -0.0534, -0.0488, -0.0527], + [ 0.0362, 0.0339, 0.0322, ..., -0.0134, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[ 4.9144e-05, 3.4124e-05, 3.2842e-05, ..., 3.8035e-06, + 5.2154e-06, 4.7907e-06], + [-2.6989e-04, -1.5306e-04, -1.4567e-04, ..., -1.8135e-05, + -2.6599e-05, -1.7509e-05], + [ 3.6359e-05, 2.5690e-05, 2.6360e-05, ..., 4.1015e-06, + 6.3218e-06, 4.4480e-06], + ..., + [ 1.4770e-04, 7.4089e-05, 6.6221e-05, ..., 3.2574e-05, + 3.4869e-05, 3.1054e-05], + [ 6.5386e-05, 3.4660e-05, 3.2157e-05, ..., 1.0155e-05, + 1.1824e-05, 9.5814e-06], + [ 4.6521e-05, 2.3782e-05, 2.1100e-05, ..., 1.4767e-05, + 1.5408e-05, 1.4156e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([ 0.0714, 0.0082, -0.0178, -0.0612, 0.0175, -0.0284, 0.0063], + device='cuda:0'), grad: tensor([ 4.7773e-05, -4.1842e-04, 4.0114e-05, -1.2136e-04, 2.6059e-04, + 1.1051e-04, 8.0645e-05], device='cuda:0') +588 +0.00070775603199067 +changing lr +epoch 58, time 805.04, cls_loss 0.0006 cls_loss_mapping 0.0054 cls_loss_causal 0.3910 re_mapping 0.0058 re_causal 0.0149 /// teacc 93.72 lr 0.00059702 +Epoch 60, weight, value: tensor([[-0.0945, -0.1099, -0.1505, ..., 0.0245, 0.0528, 0.0380], + [ 0.2331, 0.2126, 0.2448, ..., 0.0316, 0.0361, 0.0253], + [-0.0631, -0.0711, -0.0841, ..., -0.0599, -0.0777, -0.0786], + ..., + [-0.2227, -0.2084, -0.1892, ..., -0.0172, -0.0169, -0.0122], + [ 0.1040, 0.1108, 0.1246, ..., -0.0533, -0.0487, -0.0527], + [ 0.0360, 0.0337, 0.0321, ..., -0.0135, -0.0166, -0.0091]], + device='cuda:0'), grad: tensor([[ 3.2961e-05, 4.0025e-05, 3.5167e-05, ..., -2.2024e-05, + -2.5541e-05, -2.6152e-05], + [ 3.3092e-04, 1.3506e-04, 9.7990e-05, ..., 7.9155e-05, + 6.9618e-05, 6.2406e-05], + [-1.0567e-03, -4.5300e-04, -3.4881e-04, ..., -2.2972e-04, + -2.1887e-04, -2.0504e-04], + ..., + [ 1.2070e-04, 5.7369e-05, 5.0813e-05, ..., 2.4587e-05, + 2.8566e-05, 3.1531e-05], + [ 8.1837e-05, 3.0264e-05, 2.2873e-05, ..., 2.2545e-05, + 2.2694e-05, 2.2009e-05], + [ 1.5068e-04, 5.3704e-05, 3.9071e-05, ..., 3.9101e-05, + 3.8505e-05, 3.6687e-05]], device='cuda:0') +Epoch 60, bias, value: tensor([ 0.0716, 0.0080, -0.0178, -0.0609, 0.0176, -0.0284, 0.0060], + device='cuda:0'), grad: tensor([-2.1264e-05, 6.4468e-04, -1.9836e-03, 6.7139e-04, 1.8978e-04, + 1.7500e-04, 3.2091e-04], device='cuda:0') +588 +0.0005970223407163104 +changing lr +epoch 59, time 807.15, cls_loss 0.0006 cls_loss_mapping 0.0047 cls_loss_causal 0.3793 re_mapping 0.0057 re_causal 0.0145 /// teacc 92.46 lr 0.00049516 +Epoch 61, weight, value: tensor([[-0.0946, -0.1099, -0.1505, ..., 0.0245, 0.0527, 0.0379], + [ 0.2328, 0.2124, 0.2446, ..., 0.0315, 0.0361, 0.0253], + [-0.0635, -0.0713, -0.0843, ..., -0.0600, -0.0777, -0.0786], + ..., + [-0.2225, -0.2083, -0.1891, ..., -0.0172, -0.0170, -0.0123], + [ 0.1044, 0.1110, 0.1248, ..., -0.0532, -0.0486, -0.0525], + [ 0.0360, 0.0337, 0.0320, ..., -0.0135, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[ 1.4031e-04, 1.0115e-04, 9.8348e-05, ..., 2.1204e-05, + 2.3738e-05, 2.2426e-05], + [-5.2404e-04, -3.5501e-04, -3.4118e-04, ..., -1.0306e-04, + -1.1355e-04, -1.0812e-04], + [ 4.5031e-05, 3.0726e-05, 2.8789e-05, ..., 1.2629e-05, + 1.4290e-05, 1.3478e-05], + ..., + [ 6.5982e-05, 5.2780e-05, 5.0902e-05, ..., 1.4029e-05, + 1.4067e-05, 1.5333e-05], + [ 4.8369e-05, 2.8491e-05, 2.7150e-05, ..., 1.1295e-05, + 1.2696e-05, 1.1601e-05], + [ 1.0043e-04, 6.2644e-05, 5.9962e-05, ..., 2.2292e-05, + 2.4587e-05, 2.3022e-05]], device='cuda:0') +Epoch 61, bias, value: tensor([ 0.0714, 0.0079, -0.0181, -0.0608, 0.0176, -0.0280, 0.0061], + device='cuda:0'), grad: tensor([ 1.3781e-04, -5.8556e-04, 4.9412e-05, 1.5855e-04, 3.7253e-05, + 7.0691e-05, 1.3208e-04], device='cuda:0') +588 +0.0004951556604879052 +changing lr +epoch 60, time 792.83, cls_loss 0.0006 cls_loss_mapping 0.0047 cls_loss_causal 0.3636 re_mapping 0.0057 re_causal 0.0141 /// teacc 92.71 lr 0.00040236 +Epoch 62, weight, value: tensor([[-0.0947, -0.1100, -0.1505, ..., 0.0244, 0.0526, 0.0378], + [ 0.2327, 0.2124, 0.2445, ..., 0.0315, 0.0360, 0.0253], + [-0.0635, -0.0713, -0.0843, ..., -0.0599, -0.0777, -0.0786], + ..., + [-0.2223, -0.2081, -0.1890, ..., -0.0172, -0.0169, -0.0122], + [ 0.1044, 0.1110, 0.1248, ..., -0.0532, -0.0486, -0.0525], + [ 0.0359, 0.0337, 0.0320, ..., -0.0135, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[1.5759e-04, 1.2422e-04, 1.1712e-04, ..., 6.3181e-05, 6.2168e-05, + 6.3777e-05], + [3.9101e-04, 1.7965e-04, 1.6463e-04, ..., 1.6248e-04, 1.6677e-04, + 1.5986e-04], + [6.8760e-04, 3.3975e-04, 3.1376e-04, ..., 2.7561e-04, 2.8062e-04, + 2.7156e-04], + ..., + [3.8803e-05, 2.3484e-05, 2.3335e-05, ..., 3.9011e-05, 3.3319e-05, + 3.4481e-05], + [1.5879e-04, 7.5519e-05, 6.9559e-05, ..., 5.1141e-05, 5.3763e-05, + 5.1528e-05], + [2.6274e-04, 1.2243e-04, 1.1265e-04, ..., 8.4460e-05, 8.9109e-05, + 8.5115e-05]], device='cuda:0') +Epoch 62, bias, value: tensor([ 0.0712, 0.0080, -0.0181, -0.0608, 0.0178, -0.0281, 0.0059], + device='cuda:0'), grad: tensor([ 7.4744e-05, 7.1096e-04, 1.1415e-03, -2.6989e-03, 3.1173e-05, + 2.7537e-04, 4.6659e-04], device='cuda:0') +588 +0.00040236113724274745 +changing lr +epoch 61, time 803.46, cls_loss 0.0005 cls_loss_mapping 0.0046 cls_loss_causal 0.3672 re_mapping 0.0057 re_causal 0.0143 /// teacc 93.47 lr 0.00031883 +Epoch 63, weight, value: tensor([[-0.0945, -0.1099, -0.1504, ..., 0.0244, 0.0526, 0.0379], + [ 0.2326, 0.2123, 0.2444, ..., 0.0315, 0.0360, 0.0253], + [-0.0635, -0.0713, -0.0843, ..., -0.0599, -0.0776, -0.0785], + ..., + [-0.2221, -0.2080, -0.1889, ..., -0.0172, -0.0170, -0.0123], + [ 0.1043, 0.1109, 0.1247, ..., -0.0531, -0.0486, -0.0525], + [ 0.0358, 0.0336, 0.0319, ..., -0.0135, -0.0166, -0.0092]], + device='cuda:0'), grad: tensor([[ 1.0508e-04, 1.4856e-05, 2.4930e-05, ..., 2.8849e-05, + 3.3826e-05, 3.2455e-05], + [ 2.8682e-04, 8.7380e-05, 1.0943e-04, ..., 7.9691e-05, + 1.0055e-04, 8.9288e-05], + [ 2.3678e-05, 2.8145e-06, 7.0296e-06, ..., 6.0871e-06, + 9.9540e-06, 8.2627e-06], + ..., + [-7.6818e-04, -1.8561e-04, -2.5368e-04, ..., -1.9324e-04, + -2.5415e-04, -2.2399e-04], + [ 1.2767e-04, 3.2902e-05, 4.3064e-05, ..., 3.1650e-05, + 4.2170e-05, 3.6776e-05], + [ 6.0111e-05, 1.6093e-05, 2.0161e-05, ..., 1.4834e-05, + 1.9684e-05, 1.7151e-05]], device='cuda:0') +Epoch 63, bias, value: tensor([ 0.0716, 0.0080, -0.0180, -0.0608, 0.0178, -0.0282, 0.0058], + device='cuda:0'), grad: tensor([ 2.8872e-04, 6.4993e-04, 6.9916e-05, 4.3464e-04, -1.8988e-03, + 3.1042e-04, 1.4484e-04], device='cuda:0') +588 +0.00031882564680131423 +changing lr +epoch 62, time 783.15, cls_loss 0.0005 cls_loss_mapping 0.0043 cls_loss_causal 0.3643 re_mapping 0.0057 re_causal 0.0142 /// teacc 92.71 lr 0.00024472 +Epoch 64, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0526, 0.0378], + [ 0.2326, 0.2123, 0.2443, ..., 0.0315, 0.0360, 0.0253], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0776, -0.0785], + ..., + [-0.2220, -0.2079, -0.1888, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1108, 0.1246, ..., -0.0531, -0.0485, -0.0524], + [ 0.0357, 0.0336, 0.0319, ..., -0.0135, -0.0166, -0.0092]], + device='cuda:0'), grad: tensor([[-2.6107e-04, -9.1016e-05, -6.5625e-05, ..., -1.0306e-04, + -1.1551e-04, -9.4712e-05], + [ 2.9159e-04, 9.7215e-05, 8.0347e-05, ..., 1.5521e-04, + 1.7440e-04, 1.5688e-04], + [ 5.9396e-05, 2.0087e-05, 1.6272e-05, ..., 3.4511e-05, + 3.7819e-05, 3.5465e-05], + ..., + [-5.8830e-05, 1.1109e-05, 1.3985e-05, ..., 1.7032e-05, + 1.1884e-05, 1.0535e-05], + [ 5.7578e-05, 1.2890e-05, 8.0094e-06, ..., 1.6004e-05, + 1.8656e-05, 1.7390e-05], + [ 7.3493e-05, 2.4989e-05, 1.5810e-05, ..., 2.9519e-05, + 3.0696e-05, 2.8580e-05]], device='cuda:0') +Epoch 64, bias, value: tensor([ 0.0715, 0.0081, -0.0179, -0.0607, 0.0178, -0.0284, 0.0057], + device='cuda:0'), grad: tensor([-0.0006, 0.0006, 0.0001, -0.0003, -0.0003, 0.0002, 0.0002], + device='cuda:0') +588 +0.0002447174185242325 +changing lr +epoch 63, time 783.23, cls_loss 0.0005 cls_loss_mapping 0.0040 cls_loss_causal 0.3704 re_mapping 0.0057 re_causal 0.0142 /// teacc 92.71 lr 0.00018019 +Epoch 65, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0525, 0.0378], + [ 0.2325, 0.2122, 0.2442, ..., 0.0315, 0.0360, 0.0252], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0775, -0.0785], + ..., + [-0.2219, -0.2078, -0.1887, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1108, 0.1246, ..., -0.0531, -0.0485, -0.0524], + [ 0.0357, 0.0335, 0.0319, ..., -0.0135, -0.0166, -0.0092]], + device='cuda:0'), grad: tensor([[ 1.2007e-03, 1.5318e-04, 3.4958e-05, ..., 1.9395e-04, + 2.8396e-04, 2.0969e-04], + [ 7.3254e-05, 1.9327e-05, 1.4015e-05, ..., 2.5243e-05, + 2.9102e-05, 2.5526e-05], + [ 9.2268e-05, 1.4566e-05, 8.6427e-06, ..., 2.4602e-05, + 3.1501e-05, 2.4706e-05], + ..., + [-1.4887e-03, -1.8537e-04, -4.3571e-05, ..., -2.4867e-04, + -3.6168e-04, -2.6584e-04], + [ 9.1374e-05, 1.2487e-05, 5.3495e-06, ..., 2.0772e-05, + 2.7791e-05, 2.0996e-05], + [ 5.6684e-05, 8.5309e-06, 3.6918e-06, ..., 1.2532e-05, + 1.6615e-05, 1.2919e-05]], device='cuda:0') +Epoch 65, bias, value: tensor([ 0.0714, 0.0081, -0.0179, -0.0607, 0.0178, -0.0284, 0.0057], + device='cuda:0'), grad: tensor([ 3.6163e-03, 1.8573e-04, 2.8801e-04, 6.7502e-06, -4.5547e-03, + 2.8849e-04, 1.7118e-04], device='cuda:0') +588 +0.0001801856965207339 +changing lr +epoch 64, time 784.41, cls_loss 0.0005 cls_loss_mapping 0.0039 cls_loss_causal 0.3865 re_mapping 0.0057 re_causal 0.0146 /// teacc 93.22 lr 0.00012536 +Epoch 66, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0525, 0.0378], + [ 0.2325, 0.2122, 0.2442, ..., 0.0314, 0.0360, 0.0252], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0775, -0.0785], + ..., + [-0.2218, -0.2078, -0.1887, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1109, 0.1246, ..., -0.0530, -0.0485, -0.0524], + [ 0.0357, 0.0335, 0.0319, ..., -0.0135, -0.0165, -0.0092]], + device='cuda:0'), grad: tensor([[ 1.3554e-04, 1.8835e-05, 8.1435e-06, ..., 2.0534e-05, + 3.6955e-05, 2.6882e-05], + [ 3.9041e-05, 1.3135e-05, 8.7395e-06, ..., 1.1697e-05, + 1.3746e-05, 1.2234e-05], + [-4.7863e-05, -3.6389e-05, -3.1203e-05, ..., -1.2845e-05, + -1.3970e-05, -1.5751e-05], + ..., + [-3.0518e-04, -3.8534e-05, -1.0140e-05, ..., -7.1526e-05, + -1.0008e-04, -7.6473e-05], + [ 4.9859e-05, 1.0267e-05, 5.2229e-06, ..., 1.3337e-05, + 1.7017e-05, 1.3866e-05], + [ 4.3541e-05, 9.4622e-06, 5.2005e-06, ..., 1.3575e-05, + 1.6257e-05, 1.3515e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([ 0.0715, 0.0081, -0.0179, -0.0607, 0.0178, -0.0284, 0.0057], + device='cuda:0'), grad: tensor([ 3.6740e-04, 8.8513e-05, -4.8727e-05, 2.0921e-04, -8.6117e-04, + 1.3030e-04, 1.1593e-04], device='cuda:0') +588 +0.000125360439090882 +changing lr +epoch 65, time 788.68, cls_loss 0.0006 cls_loss_mapping 0.0044 cls_loss_causal 0.4013 re_mapping 0.0057 re_causal 0.0149 /// teacc 93.22 lr 0.00008035 +Epoch 67, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0525, 0.0378], + [ 0.2325, 0.2122, 0.2442, ..., 0.0314, 0.0359, 0.0252], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0775, -0.0785], + ..., + [-0.2218, -0.2077, -0.1886, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1108, 0.1246, ..., -0.0530, -0.0485, -0.0524], + [ 0.0358, 0.0335, 0.0319, ..., -0.0135, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[-4.7535e-05, -1.0945e-05, -5.4426e-06, ..., -9.0525e-06, + -1.0304e-05, -8.7395e-06], + [ 1.0920e-04, 3.1292e-05, 2.2084e-05, ..., 4.4525e-05, + 4.4137e-05, 4.3720e-05], + [-1.2636e-05, -2.3730e-06, -3.1833e-06, ..., 1.3057e-06, + -4.8429e-07, 9.8348e-07], + ..., + [-1.2589e-04, -3.4958e-05, -2.3052e-05, ..., -5.9754e-05, + -5.6773e-05, -5.8293e-05], + [ 3.6716e-05, 9.8273e-06, 6.2212e-06, ..., 1.5318e-05, + 1.5132e-05, 1.4961e-05], + [ 2.5898e-05, 6.4559e-06, 3.9339e-06, ..., 8.5160e-06, + 8.9854e-06, 8.3372e-06]], device='cuda:0') +Epoch 67, bias, value: tensor([ 0.0714, 0.0081, -0.0179, -0.0608, 0.0178, -0.0284, 0.0058], + device='cuda:0'), grad: tensor([-1.1188e-04, 3.1734e-04, -2.0221e-05, 5.3823e-05, -4.1795e-04, + 1.1033e-04, 6.9141e-05], device='cuda:0') +588 +8.03520570068517e-05 +changing lr +epoch 66, time 789.18, cls_loss 0.0005 cls_loss_mapping 0.0035 cls_loss_causal 0.3567 re_mapping 0.0056 re_causal 0.0140 /// teacc 92.71 lr 0.00004525 +Epoch 68, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0525, 0.0377], + [ 0.2325, 0.2122, 0.2442, ..., 0.0314, 0.0359, 0.0252], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0775, -0.0785], + ..., + [-0.2218, -0.2077, -0.1886, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1108, 0.1246, ..., -0.0530, -0.0485, -0.0524], + [ 0.0357, 0.0335, 0.0319, ..., -0.0135, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[ 9.5904e-05, 6.1035e-05, 6.4611e-05, ..., 2.4572e-05, + 2.9638e-05, 2.5541e-05], + [-1.1568e-03, -6.7329e-04, -7.0858e-04, ..., -1.6499e-04, + -2.1207e-04, -1.7130e-04], + [ 1.8251e-04, 8.7857e-05, 9.2208e-05, ..., 5.7310e-05, + 6.7353e-05, 5.9038e-05], + ..., + [ 5.7602e-04, 3.3116e-04, 3.3569e-04, ..., 1.0914e-04, + 1.3900e-04, 1.1832e-04], + [ 1.6415e-04, 9.1374e-05, 1.0031e-04, ..., 2.2128e-05, + 2.6956e-05, 2.1100e-05], + [ 2.6441e-04, 1.5140e-04, 1.6344e-04, ..., 3.5971e-05, + 4.3869e-05, 3.4839e-05]], device='cuda:0') +Epoch 68, bias, value: tensor([ 0.0715, 0.0081, -0.0179, -0.0608, 0.0178, -0.0285, 0.0058], + device='cuda:0'), grad: tensor([ 0.0001, -0.0015, 0.0003, -0.0002, 0.0008, 0.0002, 0.0003], + device='cuda:0') +588 +4.5251191160326525e-05 +changing lr +epoch 67, time 790.92, cls_loss 0.0005 cls_loss_mapping 0.0037 cls_loss_causal 0.3522 re_mapping 0.0057 re_causal 0.0140 /// teacc 93.97 lr 0.00002013 +Epoch 69, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0525, 0.0377], + [ 0.2325, 0.2122, 0.2442, ..., 0.0314, 0.0359, 0.0252], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0775, -0.0784], + ..., + [-0.2218, -0.2077, -0.1886, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1108, 0.1246, ..., -0.0530, -0.0485, -0.0524], + [ 0.0357, 0.0335, 0.0318, ..., -0.0135, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[ 1.9145e-04, 5.7161e-05, 3.0041e-05, ..., 3.8981e-05, + 3.7760e-05, 3.4571e-05], + [ 1.0885e-05, 1.1064e-06, -1.4063e-06, ..., 1.0841e-06, + 6.1514e-07, 5.4296e-07], + [ 3.4332e-05, 8.8140e-06, 4.2804e-06, ..., 5.4985e-06, + 5.7667e-06, 4.8243e-06], + ..., + [-3.8934e-04, -1.1307e-04, -5.8293e-05, ..., -7.6592e-05, + -7.5161e-05, -6.7890e-05], + [ 4.4286e-05, 1.2510e-05, 6.5118e-06, ..., 8.3819e-06, + 8.5384e-06, 7.5027e-06], + [ 4.1634e-05, 1.2763e-05, 7.1228e-06, ..., 8.5682e-06, + 8.4639e-06, 7.7039e-06]], device='cuda:0') +Epoch 69, bias, value: tensor([ 0.0715, 0.0081, -0.0179, -0.0608, 0.0178, -0.0285, 0.0058], + device='cuda:0'), grad: tensor([ 4.7398e-04, 3.5584e-05, 9.0063e-05, 1.6248e-04, -9.7513e-04, + 1.1194e-04, 1.0157e-04], device='cuda:0') +588 +2.0128530023804673e-05 +changing lr +epoch 68, time 782.87, cls_loss 0.0006 cls_loss_mapping 0.0041 cls_loss_causal 0.3614 re_mapping 0.0057 re_causal 0.0139 /// teacc 94.72 lr 0.00000503 +Epoch 70, weight, value: tensor([[-0.0945, -0.1099, -0.1503, ..., 0.0244, 0.0525, 0.0377], + [ 0.2325, 0.2121, 0.2441, ..., 0.0314, 0.0359, 0.0252], + [-0.0634, -0.0712, -0.0842, ..., -0.0599, -0.0775, -0.0784], + ..., + [-0.2218, -0.2077, -0.1886, ..., -0.0172, -0.0170, -0.0123], + [ 0.1041, 0.1108, 0.1246, ..., -0.0530, -0.0485, -0.0524], + [ 0.0357, 0.0335, 0.0318, ..., -0.0135, -0.0165, -0.0091]], + device='cuda:0'), grad: tensor([[ 7.4625e-04, 3.4142e-04, 2.9564e-04, ..., 1.3793e-04, + 1.3316e-04, 1.3506e-04], + [-5.5695e-04, -3.1757e-04, -2.9588e-04, ..., -2.0981e-05, + -2.5749e-05, -2.4766e-05], + [-5.8651e-04, -1.5652e-04, -9.7096e-05, ..., -1.9526e-04, + -1.6868e-04, -1.7869e-04], + ..., + [ 3.4714e-04, 1.2934e-04, 1.0431e-04, ..., 1.0133e-04, + 9.6619e-05, 9.6619e-05], + [ 6.6698e-05, 2.9922e-05, 2.6062e-05, ..., 1.5393e-05, + 1.5378e-05, 1.5073e-05], + [ 1.0866e-04, 5.4061e-05, 4.9651e-05, ..., 2.6464e-05, + 2.9162e-05, 2.7150e-05]], device='cuda:0') +Epoch 70, bias, value: tensor([ 0.0715, 0.0081, -0.0179, -0.0608, 0.0178, -0.0285, 0.0058], + device='cuda:0'), grad: tensor([ 0.0015, -0.0009, -0.0016, -0.0001, 0.0008, 0.0001, 0.0002], + device='cuda:0') +588 +5.034667293427056e-06 +changing lr +epoch 69, time 787.19, cls_loss 0.0005 cls_loss_mapping 0.0051 cls_loss_causal 0.3718 re_mapping 0.0057 re_causal 0.0140 /// teacc 93.47 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/sketch_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.516416 55.810547 64.974403 56.886228 59.223726 + sketch art_painting cartoon photo Avg +do 99.414609 48.144531 61.390785 53.353293 54.296203 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps2/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.363706 57.519531 66.552901 60.239521 61.437318 + sketch art_painting cartoon photo Avg +do 99.440061 56.054688 65.486348 59.341317 60.294118 diff --git a/Meta-causal/code-withStyleAttack/73764.error b/Meta-causal/code-withStyleAttack/73764.error new file mode 100644 index 0000000000000000000000000000000000000000..66663d356c0c65db0c8baefb03adbc6cb56c3698 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73764.error @@ -0,0 +1,33 @@ +/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 672, in + experiment() + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-mbc2l5uijcy3/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 273, in experiment + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_RA, y, epsilon_list) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 187, in adversarial_attack_Incre + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 19, in calc_mean_std + assert (len(size) == 4) + ^^^^^^^^^^^^^^ +AssertionError +slurmstepd: error: *** JOB 73764 ON gcpl4-eu-0 CANCELLED AT 2024-08-05T08:57:29 *** +slurmstepd: error: *** STEP 73764.0 ON gcpl4-eu-0 CANCELLED AT 2024-08-05T08:57:29 *** diff --git a/Meta-causal/code-withStyleAttack/73764.log b/Meta-causal/code-withStyleAttack/73764.log new file mode 100644 index 0000000000000000000000000000000000000000..8a5d3b91fa031df5c88460a657d70db4e63c5281 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73764.log @@ -0,0 +1,22 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0010, 0.0102, -0.0287, ..., 0.0063, -0.0032, 0.0239], + [-0.0166, -0.0058, 0.0193, ..., -0.0011, 0.0189, -0.0066], + [ 0.0068, 0.0064, -0.0134, ..., 0.0104, -0.0278, 0.0060], + ..., + [ 0.0182, -0.0116, 0.0244, ..., -0.0167, 0.0049, -0.0256], + [-0.0230, -0.0194, -0.0141, ..., 0.0098, -0.0289, -0.0030], + [ 0.0196, 0.0216, 0.0258, ..., -0.0108, 0.0208, 0.0110]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0302, 0.0098, -0.0187, -0.0011, 0.0113, 0.0204, 0.0101, 0.0234, + -0.0148, -0.0095], device='cuda:0'), grad: None +100 diff --git a/Meta-causal/code-withStyleAttack/73765.error b/Meta-causal/code-withStyleAttack/73765.error new file mode 100644 index 0000000000000000000000000000000000000000..a8a3ce7e52bea242daa4b938eb68723d707d5e81 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73765.error @@ -0,0 +1,96 @@ +/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/storage.py:414: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + return torch.load(io.BytesIO(b)) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 672, in + experiment() + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 273, in experiment + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_RA, y, epsilon_list) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 187, in adversarial_attack_Incre + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 19, in calc_mean_std + assert (len(size) == 4) + ^^^^^^^^^^^^^^ +AssertionError +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:45: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 145, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 29, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 45, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/serialization.py", line 1065, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/serialization.py", line 468, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/serialization.py", line 449, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/best_cls_net.pkl' +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py:48: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 145, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 29, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 48, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/serialization.py", line 1065, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/serialization.py", line 468, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-72vhe6zjf3ib/lib/python3.11/site-packages/torch/serialization.py", line 449, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/last_cls_net.pkl' +srun: error: gcpl4-eu-4: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-withStyleAttack/73765.log b/Meta-causal/code-withStyleAttack/73765.log new file mode 100644 index 0000000000000000000000000000000000000000..adb7c53661786f6b081c3b44a4c065060a4c19d5 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73765.log @@ -0,0 +1,28 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0128, 0.0060, -0.0054, ..., -0.0203, 0.0303, 0.0084], + [-0.0285, -0.0046, 0.0098, ..., -0.0264, -0.0134, -0.0136], + [ 0.0266, -0.0277, -0.0084, ..., -0.0112, -0.0297, 0.0311], + ..., + [ 0.0247, 0.0260, -0.0072, ..., -0.0026, 0.0066, -0.0214], + [ 0.0018, 0.0162, 0.0054, ..., -0.0162, 0.0112, -0.0263], + [ 0.0310, 0.0115, -0.0126, ..., -0.0082, -0.0026, -0.0038]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0159, 0.0270, 0.0303, -0.0006, -0.0084, -0.0255, -0.0144, 0.0308, + -0.0104, 0.0292], device='cuda:0'), grad: None +100 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last diff --git a/Meta-causal/code-withStyleAttack/73766.error b/Meta-causal/code-withStyleAttack/73766.error new file mode 100644 index 0000000000000000000000000000000000000000..1a4d12840feeec2f0f264b245f7905f571c9b3d3 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73766.error @@ -0,0 +1,16 @@ +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:225: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. + scaler = GradScaler() +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py:247: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. + with autocast(): +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:44: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:58: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:68: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:47: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:61: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) +/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py:71: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature. + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) diff --git a/Meta-causal/code-withStyleAttack/73766.log b/Meta-causal/code-withStyleAttack/73766.log new file mode 100644 index 0000000000000000000000000000000000000000..f95a64a361303ad44abcd6c94aa00a7a699544a6 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73766.log @@ -0,0 +1,1870 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'photo', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps1_RA_repeat', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_train.hdf5 torch.Size([1499, 3, 227, 227]) torch.Size([1499]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_val.hdf5 torch.Size([171, 3, 227, 227]) torch.Size([171]) +-------------------------------------loading pretrain weights---------------------------------- +Epoch 1, weight, value: tensor([[ 0.0041, 0.0207, -0.0077, ..., 0.0090, 0.0170, 0.0028], + [ 0.0097, -0.0137, -0.0030, ..., 0.0091, 0.0208, -0.0191], + [-0.0200, -0.0138, -0.0094, ..., 0.0213, 0.0134, -0.0101], + ..., + [-0.0031, 0.0079, 0.0148, ..., 0.0083, -0.0022, -0.0086], + [ 0.0023, -0.0039, 0.0085, ..., -0.0169, -0.0207, -0.0105], + [-0.0204, -0.0101, 0.0135, ..., -0.0202, 0.0021, 0.0039]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0168, -0.0204, 0.0171, -0.0100, 0.0122, 0.0034, -0.0083], + device='cuda:0'), grad: None +249 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 328.04, cls_loss 12.3285 cls_loss_mapping 1.5743 cls_loss_causal 1.7093 re_mapping 0.5029 re_causal 0.5021 /// teacc 42.69 lr 0.00999497 +Epoch 2, weight, value: tensor([[ 0.0806, 0.0979, 0.0322, ..., 0.0049, 0.0013, 0.0016], + [-0.0891, -0.0909, -0.0997, ..., -0.0512, 0.0041, -0.0834], + [-0.0414, -0.0397, -0.0141, ..., 0.0265, -0.0240, -0.0375], + ..., + [-0.0395, -0.0549, -0.0132, ..., 0.0733, 0.0263, 0.0311], + [-0.0512, -0.0253, -0.0534, ..., -0.0343, -0.0698, -0.0448], + [ 0.1202, 0.1275, 0.1626, ..., -0.0923, -0.0109, -0.0019]], + device='cuda:0'), grad: tensor([[ 0.0258, 0.0135, 0.0039, ..., 0.0085, 0.0035, 0.0035], + [-0.0181, -0.0107, -0.0024, ..., -0.0072, -0.0023, -0.0009], + [ 0.0141, 0.0076, 0.0043, ..., 0.0067, 0.0030, 0.0018], + ..., + [ 0.0320, 0.0172, 0.0096, ..., 0.0152, 0.0067, 0.0040], + [-0.1096, -0.0575, -0.0323, ..., -0.0508, -0.0232, -0.0150], + [ 0.0326, 0.0175, 0.0098, ..., 0.0160, 0.0070, 0.0039]], + device='cuda:0') +Epoch 2, bias, value: tensor([ 0.0665, 0.0081, 0.0509, -0.0436, 0.0429, -0.0208, -0.0932], + device='cuda:0'), grad: tensor([ 0.0765, -0.0126, 0.0306, 0.0511, 0.0686, -0.2800, 0.0657], + device='cuda:0') +249 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 328.92, cls_loss 2.2266 cls_loss_mapping 1.1671 cls_loss_causal 1.3959 re_mapping 0.1256 re_causal 0.1245 /// teacc 54.39 lr 0.00997987 +Epoch 3, weight, value: tensor([[ 0.0755, 0.0854, 0.0223, ..., 0.0188, 0.0139, 0.0173], + [-0.1050, -0.1059, -0.1274, ..., -0.0483, 0.0053, -0.0817], + [-0.0278, -0.0293, -0.0017, ..., 0.0300, -0.0264, -0.0385], + ..., + [-0.0520, -0.0710, -0.0232, ..., 0.0739, 0.0279, 0.0344], + [-0.0462, -0.0186, -0.0451, ..., -0.0560, -0.0890, -0.0671], + [ 0.1301, 0.1493, 0.1794, ..., -0.1049, -0.0189, -0.0107]], + device='cuda:0'), grad: tensor([[ 0.0210, 0.0086, 0.0037, ..., 0.0146, 0.0075, 0.0057], + [ 0.0216, 0.0091, 0.0044, ..., 0.0155, 0.0087, 0.0070], + [ 0.0156, 0.0065, 0.0030, ..., 0.0102, 0.0051, 0.0038], + ..., + [ 0.0411, 0.0174, 0.0083, ..., 0.0268, 0.0137, 0.0106], + [-0.2141, -0.0903, -0.0414, ..., -0.1223, -0.0476, -0.0331], + [-0.0315, -0.0168, -0.0130, ..., -0.0124, -0.0091, -0.0092]], + device='cuda:0') +Epoch 3, bias, value: tensor([ 0.0612, 0.0128, 0.0711, -0.0212, 0.0568, -0.0341, -0.1361], + device='cuda:0'), grad: tensor([ 0.0480, 0.0515, 0.0365, 0.3518, 0.0977, -0.4834, -0.1021], + device='cuda:0') +249 +0.009979871469976196 +changing lr +epoch 2, time 328.07, cls_loss 1.6508 cls_loss_mapping 0.8533 cls_loss_causal 1.1405 re_mapping 0.1107 re_causal 0.1092 /// teacc 43.27 lr 0.00995475 +Epoch 4, weight, value: tensor([[ 0.0929, 0.1021, 0.0452, ..., 0.0205, 0.0201, 0.0275], + [-0.1317, -0.1291, -0.1553, ..., -0.0556, -0.0021, -0.0900], + [-0.0005, -0.0066, 0.0176, ..., 0.0492, -0.0091, -0.0232], + ..., + [-0.0565, -0.0722, -0.0220, ..., 0.0674, 0.0222, 0.0287], + [-0.0426, -0.0129, -0.0439, ..., -0.0612, -0.0975, -0.0749], + [ 0.1251, 0.1393, 0.1731, ..., -0.1051, -0.0189, -0.0132]], + device='cuda:0'), grad: tensor([[ 0.0323, 0.0141, 0.0141, ..., 0.0124, 0.0062, 0.0055], + [ 0.0334, 0.0115, 0.0130, ..., 0.0159, 0.0115, 0.0080], + [-0.0419, -0.0097, -0.0117, ..., -0.0201, -0.0086, -0.0077], + ..., + [-0.1156, -0.0612, -0.0610, ..., -0.0418, -0.0280, -0.0229], + [ 0.0024, 0.0010, 0.0011, ..., 0.0009, 0.0004, 0.0004], + [ 0.0709, 0.0370, 0.0364, ..., 0.0247, 0.0138, 0.0126]], + device='cuda:0') +Epoch 4, bias, value: tensor([ 0.0733, -0.0484, 0.0864, -0.0268, 0.0908, -0.0282, -0.1370], + device='cuda:0'), grad: tensor([ 0.0681, 0.1196, -0.1030, 0.0478, -0.2786, 0.0052, 0.1409], + device='cuda:0') +249 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 330.54, cls_loss 1.4102 cls_loss_mapping 0.6470 cls_loss_causal 0.9715 re_mapping 0.0946 re_causal 0.0925 /// teacc 77.19 lr 0.00991965 +Epoch 5, weight, value: tensor([[ 0.1021, 0.1082, 0.0558, ..., 0.0227, 0.0215, 0.0301], + [-0.1205, -0.1211, -0.1480, ..., -0.0445, 0.0067, -0.0803], + [ 0.0108, 0.0058, 0.0233, ..., 0.0449, -0.0193, -0.0309], + ..., + [-0.0646, -0.0826, -0.0366, ..., 0.0580, 0.0143, 0.0228], + [-0.0472, -0.0178, -0.0483, ..., -0.0654, -0.0994, -0.0794], + [ 0.1141, 0.1368, 0.1698, ..., -0.1095, -0.0189, -0.0151]], + device='cuda:0'), grad: tensor([[-0.0292, -0.0067, -0.0099, ..., -0.0085, -0.0042, -0.0030], + [ 0.0619, 0.0148, 0.0165, ..., 0.0259, 0.0137, 0.0124], + [-0.0691, -0.0168, -0.0143, ..., -0.0348, -0.0188, -0.0182], + ..., + [-0.0536, -0.0107, -0.0084, ..., -0.0421, -0.0257, -0.0264], + [ 0.0040, 0.0009, 0.0007, ..., 0.0025, 0.0015, 0.0015], + [ 0.0048, 0.0011, 0.0009, ..., 0.0026, 0.0014, 0.0014]], + device='cuda:0') +Epoch 5, bias, value: tensor([ 0.0704, -0.0303, 0.0656, 0.0026, 0.0692, -0.0089, -0.1585], + device='cuda:0'), grad: tensor([-0.1077, 0.1888, -0.1777, 0.1881, -0.1130, 0.0094, 0.0123], + device='cuda:0') +249 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 334.03, cls_loss 1.3158 cls_loss_mapping 0.5101 cls_loss_causal 0.8892 re_mapping 0.0855 re_causal 0.0834 /// teacc 84.21 lr 0.00987464 +Epoch 6, weight, value: tensor([[ 0.0793, 0.0906, 0.0408, ..., 0.0199, 0.0203, 0.0290], + [-0.1072, -0.1078, -0.1340, ..., -0.0465, 0.0058, -0.0799], + [ 0.0284, 0.0186, 0.0323, ..., 0.0362, -0.0274, -0.0397], + ..., + [-0.0744, -0.0941, -0.0491, ..., 0.0560, 0.0099, 0.0183], + [-0.0568, -0.0259, -0.0561, ..., -0.0676, -0.1014, -0.0813], + [ 0.1281, 0.1506, 0.1832, ..., -0.1055, -0.0169, -0.0125]], + device='cuda:0'), grad: tensor([[ 2.1164e-02, 8.8272e-03, 9.5825e-03, ..., 4.4441e-04, + -2.2602e-03, -3.6850e-03], + [ 2.5543e-02, 9.6207e-03, 1.1208e-02, ..., 5.3291e-03, + 3.2196e-03, 1.1883e-03], + [ 2.1011e-02, 7.0343e-03, 7.5188e-03, ..., 5.7373e-03, + 2.9163e-03, 1.3037e-03], + ..., + [ 2.6596e-02, 1.9318e-02, 2.2980e-02, ..., 2.6855e-03, + 4.1428e-03, 3.3798e-03], + [ 6.1150e-03, 2.1019e-03, 2.1915e-03, ..., 1.2999e-03, + 4.8089e-04, 2.6956e-05], + [-1.1462e-01, -5.1758e-02, -5.8624e-02, ..., -1.9928e-02, + -1.1055e-02, -3.7251e-03]], device='cuda:0') +Epoch 6, bias, value: tensor([ 0.1190, -0.0475, 0.0738, -0.0121, 0.0560, -0.0130, -0.1660], + device='cuda:0'), grad: tensor([ 0.0582, 0.0546, 0.0554, 0.0359, -0.0010, 0.0160, -0.2190], + device='cuda:0') +249 +0.009874639560909117 +changing lr +epoch 5, time 329.85, cls_loss 1.1547 cls_loss_mapping 0.4385 cls_loss_causal 0.8479 re_mapping 0.0812 re_causal 0.0792 /// teacc 79.53 lr 0.00981981 +Epoch 7, weight, value: tensor([[ 0.0883, 0.0981, 0.0492, ..., 0.0134, 0.0104, 0.0195], + [-0.1033, -0.1092, -0.1363, ..., -0.0510, 0.0040, -0.0806], + [ 0.0255, 0.0188, 0.0320, ..., 0.0357, -0.0222, -0.0383], + ..., + [-0.0910, -0.1046, -0.0603, ..., 0.0653, 0.0178, 0.0298], + [-0.0530, -0.0238, -0.0521, ..., -0.0684, -0.1016, -0.0838], + [ 0.1283, 0.1518, 0.1848, ..., -0.1020, -0.0158, -0.0117]], + device='cuda:0'), grad: tensor([[ 0.0604, 0.0214, 0.0201, ..., 0.0237, 0.0145, 0.0152], + [-0.0547, -0.0210, -0.0169, ..., -0.0184, -0.0130, -0.0121], + [-0.0213, -0.0029, -0.0016, ..., -0.0107, -0.0048, -0.0050], + ..., + [ 0.0221, 0.0061, 0.0028, ..., 0.0053, 0.0017, 0.0014], + [ 0.0040, 0.0015, 0.0015, ..., 0.0013, 0.0007, 0.0008], + [-0.0275, -0.0114, -0.0120, ..., -0.0073, -0.0026, -0.0043]], + device='cuda:0') +Epoch 7, bias, value: tensor([ 0.1099, -0.0230, 0.0583, 0.0044, 0.0656, -0.0406, -0.1646], + device='cuda:0'), grad: tensor([ 0.1075, -0.1266, -0.0558, 0.0421, 0.0897, 0.0098, -0.0667], + device='cuda:0') +249 +0.009819814303479266 +changing lr +epoch 6, time 326.67, cls_loss 1.1271 cls_loss_mapping 0.3584 cls_loss_causal 0.7879 re_mapping 0.0775 re_causal 0.0756 /// teacc 76.02 lr 0.00975528 +Epoch 8, weight, value: tensor([[ 0.0711, 0.0809, 0.0296, ..., 0.0088, 0.0043, 0.0145], + [-0.1045, -0.1063, -0.1341, ..., -0.0397, 0.0157, -0.0668], + [ 0.0211, 0.0105, 0.0239, ..., 0.0413, -0.0161, -0.0322], + ..., + [-0.0833, -0.0990, -0.0501, ..., 0.0567, 0.0107, 0.0225], + [-0.0458, -0.0236, -0.0506, ..., -0.0682, -0.1008, -0.0849], + [ 0.1406, 0.1675, 0.1989, ..., -0.1046, -0.0191, -0.0158]], + device='cuda:0'), grad: tensor([[ 1.6815e-02, 4.2381e-03, 4.5357e-03, ..., 4.2076e-03, + 2.6627e-03, 1.6079e-03], + [-1.6083e-02, -5.0449e-04, -1.7424e-03, ..., -3.0537e-03, + -1.0414e-03, -3.4392e-05], + [ 1.3191e-02, 3.0155e-03, 2.9793e-03, ..., 3.5172e-03, + 2.2888e-03, 1.3504e-03], + ..., + [ 3.3661e-02, 6.4659e-03, 6.8054e-03, ..., 8.5526e-03, + 5.1994e-03, 2.8610e-03], + [ 1.0887e-02, 3.1624e-03, 3.0499e-03, ..., 3.0060e-03, + 2.0752e-03, 1.3409e-03], + [-5.9784e-02, -1.6724e-02, -1.5961e-02, ..., -1.6602e-02, + -1.1436e-02, -7.2823e-03]], device='cuda:0') +Epoch 8, bias, value: tensor([ 0.1165, 0.0103, 0.0518, -0.0153, 0.0377, -0.0192, -0.1721], + device='cuda:0'), grad: tensor([ 0.0467, -0.0516, 0.0388, 0.0040, 0.1010, 0.0303, -0.1691], + device='cuda:0') +249 +0.009755282581475767 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 332.33, cls_loss 1.0205 cls_loss_mapping 0.3281 cls_loss_causal 0.7437 re_mapping 0.0703 re_causal 0.0688 /// teacc 94.74 lr 0.00968117 +Epoch 9, weight, value: tensor([[ 0.0655, 0.0778, 0.0290, ..., 0.0109, 0.0081, 0.0178], + [-0.0935, -0.1044, -0.1313, ..., -0.0413, 0.0119, -0.0695], + [ 0.0122, 0.0072, 0.0218, ..., 0.0457, -0.0096, -0.0250], + ..., + [-0.0895, -0.1031, -0.0576, ..., 0.0494, 0.0036, 0.0173], + [-0.0527, -0.0276, -0.0559, ..., -0.0737, -0.1073, -0.0920], + [ 0.1482, 0.1761, 0.2075, ..., -0.1117, -0.0270, -0.0244]], + device='cuda:0'), grad: tensor([[-0.0704, -0.0111, -0.0158, ..., -0.0218, -0.0181, -0.0171], + [ 0.0493, 0.0078, 0.0108, ..., 0.0125, 0.0097, 0.0085], + [ 0.0060, 0.0010, 0.0014, ..., 0.0014, 0.0011, 0.0009], + ..., + [ 0.0272, 0.0043, 0.0060, ..., 0.0072, 0.0057, 0.0050], + [-0.0365, -0.0060, -0.0084, ..., -0.0059, -0.0040, -0.0023], + [ 0.0010, 0.0002, 0.0002, ..., 0.0002, 0.0002, 0.0001]], + device='cuda:0') +Epoch 9, bias, value: tensor([ 0.1084, 0.0215, 0.0581, 0.0034, 0.0180, -0.0383, -0.1613], + device='cuda:0'), grad: tensor([-0.2130, 0.1293, 0.0164, 0.0725, 0.0736, -0.0812, 0.0026], + device='cuda:0') +249 +0.009681174353198686 +changing lr +epoch 8, time 331.92, cls_loss 0.9519 cls_loss_mapping 0.2813 cls_loss_causal 0.6873 re_mapping 0.0690 re_causal 0.0674 /// teacc 74.27 lr 0.00959764 +Epoch 10, weight, value: tensor([[ 0.0632, 0.0775, 0.0291, ..., 0.0156, 0.0120, 0.0230], + [-0.1114, -0.1079, -0.1396, ..., -0.0458, 0.0084, -0.0705], + [ 0.0107, 0.0053, 0.0206, ..., 0.0454, -0.0096, -0.0244], + ..., + [-0.0776, -0.0953, -0.0465, ..., 0.0499, 0.0077, 0.0182], + [-0.0492, -0.0308, -0.0586, ..., -0.0804, -0.1166, -0.1012], + [ 0.1539, 0.1826, 0.2116, ..., -0.1141, -0.0304, -0.0279]], + device='cuda:0'), grad: tensor([[-3.4149e-02, -1.7075e-02, -1.7670e-02, ..., -7.3738e-03, + -6.0005e-03, -6.1493e-03], + [-3.9093e-02, -1.2817e-02, -1.5022e-02, ..., -5.9776e-03, + -9.7351e-03, -5.1575e-03], + [ 4.4922e-02, 1.9516e-02, 2.0401e-02, ..., 9.8114e-03, + 9.0637e-03, 8.3160e-03], + ..., + [ 4.0375e-02, 1.1116e-02, 1.2596e-02, ..., 8.0872e-03, + 1.0765e-02, 7.0953e-03], + [-1.4709e-02, -1.2846e-03, -8.9836e-04, ..., -5.2071e-03, + -4.8714e-03, -4.6921e-03], + [ 1.8346e-04, -1.4395e-05, -1.1884e-06, ..., 2.6703e-05, + 5.5432e-05, 2.5854e-05]], device='cuda:0') +Epoch 10, bias, value: tensor([ 0.1135, 0.0183, 0.0807, -0.0265, 0.0165, -0.0341, -0.1587], + device='cuda:0'), grad: tensor([-0.0498, -0.1252, 0.0838, 0.0073, 0.1248, -0.0418, 0.0009], + device='cuda:0') +249 +0.009597638862757255 +changing lr +epoch 9, time 329.09, cls_loss 0.9671 cls_loss_mapping 0.2545 cls_loss_causal 0.6727 re_mapping 0.0667 re_causal 0.0652 /// teacc 88.30 lr 0.00950484 +Epoch 11, weight, value: tensor([[ 0.0512, 0.0741, 0.0275, ..., 0.0174, 0.0153, 0.0271], + [-0.0926, -0.0992, -0.1311, ..., -0.0423, 0.0122, -0.0663], + [ 0.0180, 0.0045, 0.0183, ..., 0.0390, -0.0136, -0.0289], + ..., + [-0.0884, -0.1046, -0.0546, ..., 0.0459, 0.0022, 0.0134], + [-0.0528, -0.0319, -0.0592, ..., -0.0815, -0.1193, -0.1015], + [ 0.1608, 0.1935, 0.2216, ..., -0.1105, -0.0272, -0.0268]], + device='cuda:0'), grad: tensor([[ 0.0069, 0.0024, 0.0031, ..., 0.0024, 0.0027, 0.0022], + [-0.0634, -0.0111, -0.0202, ..., -0.0131, -0.0175, -0.0111], + [ 0.0605, 0.0114, 0.0197, ..., 0.0150, 0.0192, 0.0132], + ..., + [ 0.0064, 0.0026, 0.0031, ..., 0.0019, 0.0022, 0.0017], + [ 0.0036, 0.0012, 0.0015, ..., 0.0021, 0.0023, 0.0020], + [-0.0086, -0.0053, -0.0056, ..., -0.0005, -0.0005, -0.0003]], + device='cuda:0') +Epoch 11, bias, value: tensor([ 0.1092, 0.0245, 0.0954, -0.0343, -0.0178, -0.0177, -0.1501], + device='cuda:0'), grad: tensor([ 0.0146, -0.1552, 0.1486, -0.0186, 0.0125, 0.0086, -0.0105], + device='cuda:0') +249 +0.009504844339512096 +changing lr +epoch 10, time 330.61, cls_loss 0.8677 cls_loss_mapping 0.2115 cls_loss_causal 0.6044 re_mapping 0.0645 re_causal 0.0630 /// teacc 88.89 lr 0.00940298 +Epoch 12, weight, value: tensor([[ 0.0533, 0.0726, 0.0248, ..., 0.0079, 0.0041, 0.0166], + [-0.0923, -0.0968, -0.1302, ..., -0.0392, 0.0156, -0.0649], + [ 0.0133, 0.0006, 0.0156, ..., 0.0298, -0.0240, -0.0363], + ..., + [-0.0960, -0.1126, -0.0620, ..., 0.0487, 0.0110, 0.0201], + [-0.0550, -0.0307, -0.0566, ..., -0.0801, -0.1174, -0.0991], + [ 0.1767, 0.2036, 0.2309, ..., -0.1105, -0.0298, -0.0289]], + device='cuda:0'), grad: tensor([[ 6.2103e-03, 1.2302e-03, 1.9684e-03, ..., 1.1110e-03, + 1.4257e-03, 1.4706e-03], + [ 1.1263e-03, 2.5225e-04, 3.6550e-04, ..., 2.1648e-04, + 2.7752e-04, 2.8324e-04], + [ 4.2076e-03, -1.3840e-04, 1.0347e-03, ..., 2.9159e-04, + 3.8481e-04, 4.8208e-04], + ..., + [-1.1940e-02, -1.4353e-03, -3.4981e-03, ..., -1.6928e-03, + -2.1839e-03, -2.3327e-03], + [ 1.8167e-04, 3.9846e-05, 6.1452e-05, ..., 3.0190e-05, + 3.8952e-05, 4.0352e-05], + [ 1.3196e-04, 2.9266e-05, 4.1127e-05, ..., 2.6554e-05, + 3.4243e-05, 3.4899e-05]], device='cuda:0') +Epoch 12, bias, value: tensor([ 0.0990, 0.0170, 0.0597, -0.0340, -0.0004, -0.0047, -0.1275], + device='cuda:0'), grad: tensor([ 0.0247, 0.0043, 0.0240, 0.0003, -0.0545, 0.0007, 0.0005], + device='cuda:0') +249 +0.009402977659283692 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 331.19, cls_loss 0.8594 cls_loss_mapping 0.1861 cls_loss_causal 0.5998 re_mapping 0.0623 re_causal 0.0610 /// teacc 96.49 lr 0.00929224 +Epoch 13, weight, value: tensor([[ 6.2814e-02, 7.3441e-02, 2.2405e-02, ..., -4.8237e-03, + -6.0718e-03, 5.2522e-03], + [-9.2762e-02, -9.2045e-02, -1.2446e-01, ..., -2.4398e-02, + 3.2369e-02, -4.9289e-02], + [ 1.6479e-02, -1.7004e-04, 1.2857e-02, ..., 3.3231e-02, + -2.0912e-02, -3.3961e-02], + ..., + [-1.0067e-01, -1.1331e-01, -6.0309e-02, ..., 5.0409e-02, + 1.1058e-02, 2.2917e-02], + [-5.6428e-02, -3.3847e-02, -6.0100e-02, ..., -8.7616e-02, + -1.2580e-01, -1.0642e-01], + [ 1.7163e-01, 2.0295e-01, 2.3100e-01, ..., -1.1376e-01, + -3.3123e-02, -3.3123e-02]], device='cuda:0'), grad: tensor([[ 0.0477, 0.0087, 0.0102, ..., 0.0077, 0.0133, 0.0070], + [ 0.0009, 0.0002, 0.0002, ..., 0.0005, 0.0005, 0.0005], + [ 0.0049, 0.0010, 0.0010, ..., 0.0020, 0.0022, 0.0022], + ..., + [-0.0584, -0.0110, -0.0124, ..., -0.0121, -0.0181, -0.0119], + [ 0.0008, 0.0002, 0.0002, ..., 0.0003, 0.0004, 0.0004], + [ 0.0008, 0.0002, 0.0002, ..., 0.0003, 0.0004, 0.0004]], + device='cuda:0') +Epoch 13, bias, value: tensor([ 0.0886, -0.0081, 0.0858, -0.0343, 0.0154, 0.0007, -0.1392], + device='cuda:0'), grad: tensor([ 0.1378, 0.0026, 0.0148, 0.0097, -0.1698, 0.0025, 0.0025], + device='cuda:0') +249 +0.009292243968009333 +changing lr +epoch 12, time 328.66, cls_loss 0.7884 cls_loss_mapping 0.1570 cls_loss_causal 0.5338 re_mapping 0.0606 re_causal 0.0592 /// teacc 85.96 lr 0.00917287 +Epoch 14, weight, value: tensor([[ 0.0615, 0.0769, 0.0281, ..., -0.0086, -0.0078, 0.0025], + [-0.1012, -0.0984, -0.1294, ..., -0.0246, 0.0306, -0.0496], + [ 0.0153, -0.0004, 0.0110, ..., 0.0283, -0.0253, -0.0378], + ..., + [-0.0983, -0.1116, -0.0615, ..., 0.0510, 0.0112, 0.0251], + [-0.0436, -0.0272, -0.0524, ..., -0.0877, -0.1262, -0.1079], + [ 0.1741, 0.2049, 0.2322, ..., -0.1117, -0.0330, -0.0334]], + device='cuda:0'), grad: tensor([[-1.2360e-02, -1.6689e-03, -2.1782e-03, ..., -1.3695e-03, + -1.8444e-03, -1.2569e-03], + [-2.6188e-03, -1.7560e-04, 4.6194e-05, ..., -1.2903e-03, + -1.0080e-03, -1.2169e-03], + [ 1.1909e-02, 1.7262e-03, 1.8282e-03, ..., 1.4267e-03, + 1.8158e-03, 1.7233e-03], + ..., + [ 2.5406e-02, 3.6488e-03, 3.6907e-03, ..., 3.9711e-03, + 4.4861e-03, 4.4441e-03], + [-3.0823e-02, -4.4365e-03, -4.2458e-03, ..., -3.6411e-03, + -4.6196e-03, -4.9553e-03], + [ 4.1847e-03, 2.7370e-04, 2.4772e-04, ..., 3.6621e-04, + 5.0259e-04, 5.5885e-04]], device='cuda:0') +Epoch 14, bias, value: tensor([ 0.0458, -0.0057, 0.0768, -0.0453, 0.0524, 0.0168, -0.1319], + device='cuda:0'), grad: tensor([-0.0580, -0.0057, 0.0432, 0.0136, 0.0892, -0.0966, 0.0143], + device='cuda:0') +249 +0.009172866268606516 +changing lr +epoch 13, time 330.53, cls_loss 0.7788 cls_loss_mapping 0.1604 cls_loss_causal 0.5419 re_mapping 0.0587 re_causal 0.0576 /// teacc 89.47 lr 0.00904508 +Epoch 15, weight, value: tensor([[ 0.0543, 0.0690, 0.0245, ..., -0.0101, -0.0094, 0.0020], + [-0.0891, -0.0909, -0.1231, ..., -0.0165, 0.0380, -0.0410], + [ 0.0096, -0.0032, 0.0056, ..., 0.0338, -0.0202, -0.0313], + ..., + [-0.1022, -0.1135, -0.0643, ..., 0.0398, 0.0018, 0.0146], + [-0.0513, -0.0273, -0.0520, ..., -0.0899, -0.1276, -0.1114], + [ 0.1794, 0.2118, 0.2379, ..., -0.1066, -0.0298, -0.0298]], + device='cuda:0'), grad: tensor([[-3.8727e-02, -5.3978e-03, -6.6414e-03, ..., -4.4403e-03, + -6.5880e-03, -5.6458e-03], + [ 2.5787e-02, 2.8992e-03, 3.8128e-03, ..., 3.4657e-03, + 4.7989e-03, 4.3488e-03], + [ 3.3234e-02, 1.5211e-03, 3.5324e-03, ..., 7.4463e-03, + 8.7738e-03, 8.9951e-03], + ..., + [-2.9022e-02, 3.4499e-04, -1.8139e-03, ..., -8.2016e-03, + -9.1019e-03, -9.8038e-03], + [ 8.5068e-04, 6.1452e-05, 1.0502e-04, ..., 1.6069e-04, + 1.9574e-04, 1.9622e-04], + [ 5.4836e-04, 5.0932e-05, 7.6413e-05, ..., 9.1612e-05, + 1.1808e-04, 1.1295e-04]], device='cuda:0') +Epoch 15, bias, value: tensor([ 0.0235, 0.0189, 0.0600, -0.0188, 0.0581, 0.0033, -0.1361], + device='cuda:0'), grad: tensor([-0.1204, 0.0803, 0.0999, 0.0220, -0.0862, 0.0026, 0.0017], + device='cuda:0') +249 +0.00904508497187474 +changing lr +epoch 14, time 332.05, cls_loss 0.7896 cls_loss_mapping 0.1331 cls_loss_causal 0.5382 re_mapping 0.0563 re_causal 0.0555 /// teacc 88.89 lr 0.00890916 +Epoch 16, weight, value: tensor([[ 0.0507, 0.0644, 0.0202, ..., -0.0115, -0.0100, 0.0012], + [-0.0902, -0.0853, -0.1148, ..., -0.0046, 0.0471, -0.0318], + [ 0.0073, -0.0051, 0.0031, ..., 0.0341, -0.0184, -0.0285], + ..., + [-0.0909, -0.1086, -0.0602, ..., 0.0317, -0.0028, 0.0092], + [-0.0494, -0.0279, -0.0526, ..., -0.0870, -0.1246, -0.1088], + [ 0.1756, 0.2082, 0.2327, ..., -0.1124, -0.0380, -0.0372]], + device='cuda:0'), grad: tensor([[-0.0344, -0.0060, -0.0095, ..., -0.0063, -0.0076, -0.0072], + [ 0.0205, 0.0037, 0.0054, ..., 0.0044, 0.0049, 0.0048], + [-0.0551, -0.0103, -0.0130, ..., -0.0145, -0.0147, -0.0149], + ..., + [ 0.0391, 0.0071, 0.0099, ..., 0.0091, 0.0097, 0.0096], + [ 0.0194, 0.0036, 0.0046, ..., 0.0051, 0.0051, 0.0052], + [ 0.0020, 0.0004, 0.0005, ..., 0.0005, 0.0005, 0.0005]], + device='cuda:0') +Epoch 16, bias, value: tensor([ 0.0318, 0.0176, 0.0590, -0.0301, 0.0591, -0.0022, -0.1263], + device='cuda:0'), grad: tensor([-0.1025, 0.0603, -0.1576, 0.0245, 0.1137, 0.0556, 0.0058], + device='cuda:0') +249 +0.008909157412340152 +changing lr +epoch 15, time 330.68, cls_loss 0.7538 cls_loss_mapping 0.1214 cls_loss_causal 0.5324 re_mapping 0.0535 re_causal 0.0527 /// teacc 89.47 lr 0.00876536 +Epoch 17, weight, value: tensor([[ 0.0495, 0.0606, 0.0159, ..., -0.0167, -0.0161, -0.0039], + [-0.0854, -0.0851, -0.1131, ..., -0.0048, 0.0466, -0.0312], + [ 0.0095, -0.0045, 0.0021, ..., 0.0302, -0.0208, -0.0313], + ..., + [-0.0978, -0.1112, -0.0638, ..., 0.0395, 0.0065, 0.0183], + [-0.0519, -0.0292, -0.0548, ..., -0.0917, -0.1277, -0.1138], + [ 0.1782, 0.2135, 0.2395, ..., -0.1085, -0.0353, -0.0343]], + device='cuda:0'), grad: tensor([[ 4.2748e-04, 1.1969e-04, 1.2207e-04, ..., 1.3006e-04, + 1.2010e-04, 1.0848e-04], + [ 1.0085e-04, 3.1233e-05, 3.2514e-05, ..., 3.7760e-05, + 3.5286e-05, 3.2365e-05], + [-7.5006e-04, -8.9109e-05, -1.0699e-04, ..., -2.6321e-04, + -2.5272e-04, -2.2566e-04], + ..., + [ 2.4629e-04, 6.0737e-05, 6.2585e-05, ..., 7.6532e-05, + 7.1466e-05, 6.4731e-05], + [ 1.7762e-04, 2.6643e-05, 2.8387e-05, ..., 5.7787e-05, + 5.5104e-05, 5.0515e-05], + [-3.4380e-04, -1.7953e-04, -1.7190e-04, ..., -9.0003e-05, + -7.7963e-05, -7.4148e-05]], device='cuda:0') +Epoch 17, bias, value: tensor([ 0.0431, 0.0347, 0.0788, -0.0381, 0.0351, -0.0139, -0.1310], + device='cuda:0'), grad: tensor([ 0.0012, 0.0003, -0.0028, 0.0005, 0.0007, 0.0006, -0.0005], + device='cuda:0') +249 +0.00876535733001806 +changing lr +epoch 16, time 330.33, cls_loss 0.7158 cls_loss_mapping 0.0959 cls_loss_causal 0.5369 re_mapping 0.0526 re_causal 0.0519 /// teacc 94.15 lr 0.00861397 +Epoch 18, weight, value: tensor([[ 0.0449, 0.0565, 0.0137, ..., -0.0169, -0.0154, -0.0049], + [-0.0928, -0.0910, -0.1204, ..., -0.0028, 0.0457, -0.0285], + [ 0.0077, -0.0064, 0.0016, ..., 0.0295, -0.0225, -0.0319], + ..., + [-0.0945, -0.1056, -0.0579, ..., 0.0404, 0.0117, 0.0230], + [-0.0543, -0.0308, -0.0577, ..., -0.0970, -0.1333, -0.1192], + [ 0.1882, 0.2196, 0.2448, ..., -0.1046, -0.0314, -0.0324]], + device='cuda:0'), grad: tensor([[ 2.9640e-03, 1.5087e-05, -1.2487e-05, ..., 6.4421e-04, + 9.0408e-04, 8.3351e-04], + [-3.9124e-02, -5.6763e-03, -5.1651e-03, ..., -8.2016e-03, + -1.0010e-02, -9.6970e-03], + [ 5.2719e-03, 8.2636e-04, 7.5102e-04, ..., 1.0967e-03, + 1.3180e-03, 1.2846e-03], + ..., + [ 7.3853e-03, 1.2684e-03, 1.1692e-03, ..., 1.5392e-03, + 1.8225e-03, 1.7834e-03], + [ 4.0970e-03, 5.9843e-04, 5.4550e-04, ..., 8.5926e-04, + 1.0481e-03, 1.0147e-03], + [ 3.9024e-03, 6.8951e-04, 6.3562e-04, ..., 8.1110e-04, + 9.5463e-04, 9.3699e-04]], device='cuda:0') +Epoch 18, bias, value: tensor([ 0.0450, 0.0189, 0.0744, -0.0380, 0.0326, -0.0089, -0.1155], + device='cuda:0'), grad: tensor([ 0.0156, -0.1628, 0.0215, 0.0645, 0.0291, 0.0170, 0.0153], + device='cuda:0') +249 +0.008613974319136962 +changing lr +epoch 17, time 332.15, cls_loss 0.6988 cls_loss_mapping 0.1077 cls_loss_causal 0.4824 re_mapping 0.0497 re_causal 0.0490 /// teacc 93.57 lr 0.00845531 +Epoch 19, weight, value: tensor([[ 0.0339, 0.0438, 0.0010, ..., -0.0201, -0.0196, -0.0092], + [-0.0835, -0.0845, -0.1132, ..., -0.0004, 0.0460, -0.0263], + [ 0.0090, -0.0067, 0.0012, ..., 0.0233, -0.0282, -0.0372], + ..., + [-0.0906, -0.1018, -0.0536, ..., 0.0414, 0.0144, 0.0254], + [-0.0498, -0.0311, -0.0574, ..., -0.1009, -0.1372, -0.1226], + [ 0.1849, 0.2209, 0.2455, ..., -0.1042, -0.0302, -0.0325]], + device='cuda:0'), grad: tensor([[ 1.2932e-03, 1.5438e-04, 1.7822e-04, ..., 5.6267e-04, + 5.2834e-04, 5.4789e-04], + [-1.9791e-02, -3.1681e-03, -3.7670e-03, ..., -7.5111e-03, + -7.0915e-03, -7.2899e-03], + [ 4.6349e-03, 7.2432e-04, 8.5878e-04, ..., 1.7843e-03, + 1.6842e-03, 1.7319e-03], + ..., + [ 1.2718e-02, 2.1267e-03, 2.5406e-03, ..., 4.6883e-03, + 4.4327e-03, 4.5471e-03], + [ 1.4830e-04, 2.1055e-05, 2.5973e-05, ..., 6.6638e-05, + 6.2168e-05, 6.6042e-05], + [ 2.1327e-04, 3.4451e-05, 4.0352e-05, ..., 8.3327e-05, + 7.8619e-05, 8.0943e-05]], device='cuda:0') +Epoch 19, bias, value: tensor([ 0.0250, 0.0310, 0.0902, -0.0317, 0.0387, -0.0252, -0.1194], + device='cuda:0'), grad: tensor([ 0.0054, -0.0831, 0.0194, 0.0032, 0.0535, 0.0007, 0.0009], + device='cuda:0') +249 +0.008455313244934327 +changing lr +epoch 18, time 332.32, cls_loss 0.7309 cls_loss_mapping 0.0882 cls_loss_causal 0.5184 re_mapping 0.0476 re_causal 0.0468 /// teacc 95.91 lr 0.00828969 +Epoch 20, weight, value: tensor([[ 0.0337, 0.0446, 0.0015, ..., -0.0211, -0.0204, -0.0103], + [-0.0897, -0.0866, -0.1140, ..., 0.0029, 0.0484, -0.0236], + [ 0.0073, -0.0084, -0.0016, ..., 0.0239, -0.0258, -0.0362], + ..., + [-0.0875, -0.0985, -0.0495, ..., 0.0371, 0.0106, 0.0229], + [-0.0529, -0.0351, -0.0617, ..., -0.1022, -0.1382, -0.1231], + [ 0.1910, 0.2242, 0.2488, ..., -0.1034, -0.0304, -0.0334]], + device='cuda:0'), grad: tensor([[ 0.0155, 0.0025, 0.0032, ..., 0.0038, 0.0036, 0.0040], + [-0.0185, -0.0019, -0.0018, ..., -0.0065, -0.0066, -0.0069], + [ 0.0057, 0.0012, 0.0016, ..., 0.0009, 0.0008, 0.0010], + ..., + [-0.0304, -0.0078, -0.0112, ..., -0.0026, -0.0014, -0.0026], + [ 0.0060, 0.0012, 0.0016, ..., 0.0011, 0.0010, 0.0012], + [ 0.0158, 0.0035, 0.0048, ..., 0.0023, 0.0018, 0.0024]], + device='cuda:0') +Epoch 20, bias, value: tensor([ 0.0229, 0.0047, 0.0509, -0.0327, 0.0590, 0.0081, -0.1043], + device='cuda:0'), grad: tensor([ 0.0580, -0.0692, 0.0212, 0.0220, -0.1139, 0.0226, 0.0591], + device='cuda:0') +249 +0.008289693629698565 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 331.35, cls_loss 0.6405 cls_loss_mapping 0.0975 cls_loss_causal 0.4552 re_mapping 0.0467 re_causal 0.0465 /// teacc 97.66 lr 0.00811745 +Epoch 21, weight, value: tensor([[ 0.0233, 0.0409, -0.0019, ..., -0.0184, -0.0186, -0.0071], + [-0.0808, -0.0887, -0.1161, ..., -0.0016, 0.0459, -0.0268], + [ 0.0061, -0.0060, 0.0007, ..., 0.0297, -0.0216, -0.0308], + ..., + [-0.0845, -0.0976, -0.0496, ..., 0.0321, 0.0055, 0.0183], + [-0.0483, -0.0342, -0.0608, ..., -0.1027, -0.1387, -0.1238], + [ 0.1853, 0.2265, 0.2511, ..., -0.1003, -0.0282, -0.0320]], + device='cuda:0'), grad: tensor([[ 0.0071, 0.0009, 0.0009, ..., 0.0016, 0.0014, 0.0014], + [-0.0539, -0.0104, -0.0101, ..., -0.0222, -0.0219, -0.0188], + [ 0.0016, 0.0002, 0.0002, ..., -0.0003, -0.0004, -0.0004], + ..., + [ 0.0222, 0.0062, 0.0053, ..., 0.0171, 0.0176, 0.0145], + [ 0.0126, 0.0017, 0.0020, ..., 0.0020, 0.0018, 0.0018], + [ 0.0014, 0.0002, 0.0002, ..., 0.0002, 0.0002, 0.0002]], + device='cuda:0') +Epoch 21, bias, value: tensor([ 0.0173, 0.0031, 0.0273, -0.0330, 0.0884, 0.0304, -0.1249], + device='cuda:0'), grad: tensor([ 0.0259, -0.1476, 0.0072, 0.0380, 0.0166, 0.0540, 0.0058], + device='cuda:0') +249 +0.00811744900929367 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 333.43, cls_loss 0.6067 cls_loss_mapping 0.0854 cls_loss_causal 0.4607 re_mapping 0.0461 re_causal 0.0455 /// teacc 98.25 lr 0.00793893 +Epoch 22, weight, value: tensor([[ 0.0405, 0.0514, 0.0106, ..., -0.0137, -0.0141, -0.0019], + [-0.0799, -0.0906, -0.1176, ..., 0.0028, 0.0508, -0.0216], + [ 0.0059, -0.0075, -0.0031, ..., 0.0258, -0.0251, -0.0343], + ..., + [-0.0937, -0.1027, -0.0542, ..., 0.0320, 0.0055, 0.0172], + [-0.0520, -0.0341, -0.0624, ..., -0.1042, -0.1404, -0.1254], + [ 0.1820, 0.2247, 0.2515, ..., -0.1011, -0.0297, -0.0345]], + device='cuda:0'), grad: tensor([[ 0.0032, 0.0005, -0.0003, ..., 0.0037, 0.0050, 0.0054], + [-0.0380, -0.0085, -0.0084, ..., -0.0027, -0.0043, -0.0056], + [ 0.0186, 0.0038, 0.0040, ..., 0.0095, 0.0096, 0.0096], + ..., + [ 0.0214, 0.0046, 0.0049, ..., 0.0069, 0.0071, 0.0073], + [ 0.0202, 0.0040, 0.0038, ..., 0.0141, 0.0146, 0.0146], + [ 0.0059, 0.0013, 0.0013, ..., 0.0014, 0.0016, 0.0018]], + device='cuda:0') +Epoch 22, bias, value: tensor([ 0.0392, 0.0122, 0.0013, -0.0322, 0.0914, 0.0076, -0.1111], + device='cuda:0'), grad: tensor([ 0.0113, -0.1250, 0.0560, -0.0873, 0.0663, 0.0597, 0.0190], + device='cuda:0') +249 +0.007938926261462368 +changing lr +epoch 21, time 327.98, cls_loss 0.6550 cls_loss_mapping 0.0824 cls_loss_causal 0.4821 re_mapping 0.0436 re_causal 0.0438 /// teacc 94.15 lr 0.00775448 +Epoch 23, weight, value: tensor([[ 3.4201e-02, 5.0662e-02, 8.9740e-03, ..., -2.1122e-04, + -2.7622e-03, 1.0790e-02], + [-7.5561e-02, -8.4475e-02, -1.1104e-01, ..., -4.7518e-04, + 5.0359e-02, -2.2617e-02], + [ 7.8655e-03, -8.7703e-03, -3.6687e-03, ..., 1.9358e-02, + -3.0439e-02, -4.0037e-02], + ..., + [-9.6102e-02, -1.0140e-01, -5.3479e-02, ..., 2.6631e-02, + -4.5946e-05, 1.2200e-02], + [-5.1474e-02, -3.3746e-02, -6.1860e-02, ..., -1.0549e-01, + -1.4164e-01, -1.2638e-01], + [ 1.8815e-01, 2.2104e-01, 2.4732e-01, ..., -9.9035e-02, + -3.0846e-02, -3.4547e-02]], device='cuda:0'), grad: tensor([[ 0.0206, 0.0029, 0.0030, ..., 0.0016, 0.0024, 0.0024], + [ 0.0091, 0.0012, 0.0016, ..., 0.0006, 0.0010, 0.0011], + [-0.0322, -0.0047, -0.0033, ..., -0.0027, -0.0037, -0.0035], + ..., + [ 0.0049, 0.0007, 0.0007, ..., 0.0003, 0.0005, 0.0006], + [ 0.0188, 0.0025, 0.0033, ..., 0.0012, 0.0021, 0.0022], + [-0.0283, -0.0036, -0.0068, ..., -0.0015, -0.0031, -0.0036]], + device='cuda:0') +Epoch 23, bias, value: tensor([-0.0102, 0.0090, 0.0143, -0.0455, 0.0791, 0.0312, -0.0695], + device='cuda:0'), grad: tensor([ 0.0809, 0.0351, -0.1302, 0.0270, 0.0193, 0.0726, -0.1045], + device='cuda:0') +249 +0.007754484907260515 +changing lr +epoch 22, time 330.59, cls_loss 0.6359 cls_loss_mapping 0.0691 cls_loss_causal 0.4715 re_mapping 0.0417 re_causal 0.0414 /// teacc 88.89 lr 0.00756450 +Epoch 24, weight, value: tensor([[ 0.0297, 0.0495, 0.0066, ..., 0.0025, -0.0029, 0.0126], + [-0.0726, -0.0838, -0.1098, ..., 0.0009, 0.0526, -0.0210], + [ 0.0102, -0.0069, -0.0015, ..., 0.0206, -0.0283, -0.0378], + ..., + [-0.0925, -0.1007, -0.0532, ..., 0.0323, 0.0066, 0.0201], + [-0.0484, -0.0315, -0.0589, ..., -0.1099, -0.1459, -0.1307], + [ 0.1805, 0.2173, 0.2427, ..., -0.1020, -0.0350, -0.0389]], + device='cuda:0'), grad: tensor([[ 0.0157, 0.0022, 0.0028, ..., 0.0115, 0.0073, 0.0072], + [ 0.0063, 0.0010, 0.0014, ..., 0.0021, 0.0018, 0.0019], + [ 0.0055, 0.0010, 0.0011, ..., 0.0009, 0.0011, 0.0012], + ..., + [-0.0447, -0.0077, -0.0091, ..., -0.0168, -0.0125, -0.0138], + [ 0.0102, 0.0021, 0.0024, ..., 0.0016, 0.0019, 0.0023], + [ 0.0037, 0.0002, 0.0003, ..., 0.0004, 0.0005, 0.0008]], + device='cuda:0') +Epoch 24, bias, value: tensor([-0.0262, 0.0202, 0.0096, -0.0479, 0.0986, 0.0423, -0.0883], + device='cuda:0'), grad: tensor([ 0.0757, 0.0298, 0.0225, 0.0187, -0.2087, 0.0410, 0.0209], + device='cuda:0') +249 +0.007564496387029534 +changing lr +epoch 23, time 328.95, cls_loss 0.6473 cls_loss_mapping 0.0632 cls_loss_causal 0.4710 re_mapping 0.0390 re_causal 0.0383 /// teacc 90.06 lr 0.00736934 +Epoch 25, weight, value: tensor([[ 0.0354, 0.0505, 0.0064, ..., -0.0128, -0.0179, -0.0011], + [-0.0726, -0.0786, -0.1029, ..., 0.0073, 0.0563, -0.0160], + [ 0.0039, -0.0110, -0.0055, ..., 0.0250, -0.0243, -0.0333], + ..., + [-0.0857, -0.0961, -0.0482, ..., 0.0355, 0.0116, 0.0246], + [-0.0518, -0.0322, -0.0606, ..., -0.1137, -0.1492, -0.1351], + [ 0.1791, 0.2133, 0.2393, ..., -0.1019, -0.0347, -0.0390]], + device='cuda:0'), grad: tensor([[-0.0074, -0.0009, -0.0006, ..., -0.0032, -0.0046, -0.0038], + [ 0.0115, 0.0015, 0.0015, ..., 0.0036, 0.0047, 0.0043], + [ 0.0023, 0.0003, 0.0004, ..., 0.0005, 0.0005, 0.0006], + ..., + [-0.0173, -0.0023, -0.0032, ..., -0.0030, -0.0032, -0.0037], + [ 0.0021, 0.0003, 0.0004, ..., 0.0003, 0.0004, 0.0004], + [ 0.0032, 0.0004, 0.0005, ..., 0.0007, 0.0007, 0.0008]], + device='cuda:0') +Epoch 25, bias, value: tensor([-0.0229, 0.0201, 0.0102, -0.0626, 0.1099, 0.0253, -0.0719], + device='cuda:0'), grad: tensor([-0.0216, 0.0528, 0.0142, 0.0318, -0.1116, 0.0145, 0.0199], + device='cuda:0') +249 +0.007369343312364995 +changing lr +epoch 24, time 332.00, cls_loss 0.6398 cls_loss_mapping 0.0677 cls_loss_causal 0.4755 re_mapping 0.0390 re_causal 0.0388 /// teacc 89.47 lr 0.00716942 +Epoch 26, weight, value: tensor([[ 0.0382, 0.0500, 0.0073, ..., -0.0081, -0.0154, 0.0036], + [-0.0729, -0.0787, -0.1029, ..., 0.0119, 0.0599, -0.0118], + [-0.0049, -0.0143, -0.0077, ..., 0.0240, -0.0250, -0.0342], + ..., + [-0.0900, -0.0957, -0.0487, ..., 0.0295, 0.0070, 0.0202], + [-0.0457, -0.0301, -0.0592, ..., -0.1130, -0.1475, -0.1339], + [ 0.1854, 0.2161, 0.2411, ..., -0.1009, -0.0339, -0.0398]], + device='cuda:0'), grad: tensor([[ 0.0233, 0.0034, 0.0050, ..., 0.0070, 0.0073, 0.0074], + [ 0.0038, 0.0006, 0.0008, ..., 0.0008, 0.0008, 0.0008], + [ 0.0019, 0.0003, 0.0004, ..., 0.0004, 0.0004, 0.0004], + ..., + [ 0.0039, 0.0006, 0.0007, ..., 0.0009, 0.0010, 0.0010], + [-0.0287, -0.0046, -0.0056, ..., -0.0058, -0.0062, -0.0063], + [ 0.0032, 0.0005, 0.0006, ..., 0.0007, 0.0007, 0.0007]], + device='cuda:0') +Epoch 26, bias, value: tensor([-0.0096, 0.0024, -0.0115, -0.0645, 0.1321, 0.0331, -0.0738], + device='cuda:0'), grad: tensor([ 0.0940, 0.0174, 0.0085, -0.0218, 0.0169, -0.1296, 0.0145], + device='cuda:0') +249 +0.0071694186955877925 +changing lr +epoch 25, time 330.45, cls_loss 0.6167 cls_loss_mapping 0.0631 cls_loss_causal 0.4583 re_mapping 0.0373 re_causal 0.0370 /// teacc 86.55 lr 0.00696513 +Epoch 27, weight, value: tensor([[ 0.0291, 0.0450, 0.0038, ..., -0.0099, -0.0150, 0.0033], + [-0.0689, -0.0753, -0.1012, ..., 0.0162, 0.0639, -0.0065], + [-0.0016, -0.0147, -0.0085, ..., 0.0246, -0.0256, -0.0337], + ..., + [-0.0891, -0.0949, -0.0479, ..., 0.0311, 0.0090, 0.0208], + [-0.0484, -0.0300, -0.0604, ..., -0.1132, -0.1478, -0.1338], + [ 0.1865, 0.2173, 0.2434, ..., -0.1022, -0.0355, -0.0417]], + device='cuda:0'), grad: tensor([[-0.0355, -0.0106, -0.0125, ..., -0.0094, -0.0115, -0.0123], + [ 0.0024, 0.0023, 0.0016, ..., 0.0023, 0.0015, 0.0001], + [ 0.0042, 0.0009, 0.0009, ..., 0.0004, 0.0007, 0.0008], + ..., + [ 0.0084, 0.0021, 0.0023, ..., 0.0014, 0.0019, 0.0022], + [ 0.0219, 0.0051, 0.0055, ..., 0.0032, 0.0045, 0.0053], + [ 0.0339, 0.0082, 0.0092, ..., 0.0056, 0.0077, 0.0090]], + device='cuda:0') +Epoch 27, bias, value: tensor([-0.0398, 0.0162, 0.0080, -0.0598, 0.1309, 0.0240, -0.0714], + device='cuda:0'), grad: tensor([-0.0991, -0.0373, 0.0158, -0.1063, 0.0296, 0.0779, 0.1193], + device='cuda:0') +249 +0.0069651251582696205 +changing lr +epoch 26, time 329.67, cls_loss 0.5797 cls_loss_mapping 0.0547 cls_loss_causal 0.4521 re_mapping 0.0359 re_causal 0.0356 /// teacc 94.74 lr 0.00675687 +Epoch 28, weight, value: tensor([[ 0.0296, 0.0479, 0.0056, ..., -0.0083, -0.0129, 0.0055], + [-0.0635, -0.0771, -0.1022, ..., 0.0173, 0.0646, -0.0031], + [ 0.0014, -0.0160, -0.0100, ..., 0.0202, -0.0299, -0.0372], + ..., + [-0.0881, -0.0948, -0.0474, ..., 0.0348, 0.0136, 0.0242], + [-0.0515, -0.0288, -0.0590, ..., -0.1128, -0.1473, -0.1337], + [ 0.1851, 0.2156, 0.2419, ..., -0.1035, -0.0383, -0.0448]], + device='cuda:0'), grad: tensor([[-3.2379e-02, -1.0920e-03, -2.8629e-03, ..., -4.0131e-03, + -5.5962e-03, -5.5771e-03], + [ 2.0889e-02, 2.8038e-03, 3.5915e-03, ..., 3.9444e-03, + 4.8485e-03, 4.9324e-03], + [ 6.1455e-03, 1.2851e-04, 7.3433e-04, ..., -6.7186e-04, + -2.5487e-04, -3.5954e-04], + ..., + [ 2.0447e-03, -2.9111e-04, -1.8477e-04, ..., 4.2272e-04, + 3.9196e-04, 4.4465e-04], + [ 2.1057e-03, 9.6560e-05, 2.0063e-04, ..., 3.1281e-04, + 4.1771e-04, 4.1795e-04], + [-1.3342e-03, -1.7681e-03, -1.7300e-03, ..., -3.4809e-04, + -2.8825e-04, -3.3569e-04]], device='cuda:0') +Epoch 28, bias, value: tensor([-0.0779, 0.0406, 0.0447, -0.0915, 0.1372, 0.0079, -0.0530], + device='cuda:0'), grad: tensor([-0.1472, 0.0815, 0.0300, 0.0113, 0.0111, 0.0094, 0.0041], + device='cuda:0') +249 +0.006756874120406716 +changing lr +epoch 27, time 331.20, cls_loss 0.5574 cls_loss_mapping 0.0509 cls_loss_causal 0.4335 re_mapping 0.0341 re_causal 0.0339 /// teacc 95.32 lr 0.00654508 +Epoch 29, weight, value: tensor([[ 0.0304, 0.0503, 0.0085, ..., -0.0058, -0.0120, 0.0072], + [-0.0768, -0.0828, -0.1073, ..., 0.0146, 0.0624, -0.0061], + [-0.0012, -0.0181, -0.0123, ..., 0.0214, -0.0270, -0.0360], + ..., + [-0.0904, -0.0955, -0.0495, ..., 0.0372, 0.0162, 0.0272], + [-0.0497, -0.0261, -0.0575, ..., -0.1099, -0.1450, -0.1310], + [ 0.1914, 0.2156, 0.2432, ..., -0.1031, -0.0392, -0.0447]], + device='cuda:0'), grad: tensor([[-1.3367e-02, -4.8141e-03, -4.5052e-03, ..., -1.5497e-03, + -1.6203e-03, -2.0981e-03], + [ 1.2231e-04, 5.2422e-05, 4.9323e-05, ..., 8.0764e-05, + 7.7248e-05, 1.1367e-04], + [ 1.6391e-05, 5.1928e-04, 6.5613e-04, ..., -4.4060e-04, + -3.6407e-04, -4.4990e-04], + ..., + [ 8.2550e-03, 2.9011e-03, 2.6970e-03, ..., 1.0481e-03, + 1.0977e-03, 1.3857e-03], + [ 1.6584e-03, 4.9019e-04, 4.2510e-04, ..., 3.0446e-04, + 3.0375e-04, 3.7599e-04], + [ 8.2350e-04, 2.5725e-04, 2.3580e-04, ..., 1.6701e-04, + 1.8442e-04, 2.0158e-04]], device='cuda:0') +Epoch 29, bias, value: tensor([-0.0906, 0.0002, 0.0289, -0.0507, 0.1348, 0.0144, -0.0292], + device='cuda:0'), grad: tensor([-0.0324, 0.0003, -0.0017, 0.0071, 0.0202, 0.0044, 0.0021], + device='cuda:0') +249 +0.00654508497187474 +changing lr +epoch 28, time 330.13, cls_loss 0.5201 cls_loss_mapping 0.0471 cls_loss_causal 0.3959 re_mapping 0.0325 re_causal 0.0319 /// teacc 98.25 lr 0.00633018 +Epoch 30, weight, value: tensor([[ 0.0278, 0.0433, 0.0012, ..., -0.0062, -0.0116, 0.0061], + [-0.0718, -0.0800, -0.1032, ..., 0.0200, 0.0680, -0.0004], + [-0.0048, -0.0208, -0.0152, ..., 0.0243, -0.0240, -0.0325], + ..., + [-0.0927, -0.0954, -0.0493, ..., 0.0318, 0.0101, 0.0223], + [-0.0482, -0.0270, -0.0577, ..., -0.1149, -0.1492, -0.1351], + [ 0.1935, 0.2213, 0.2486, ..., -0.1004, -0.0374, -0.0427]], + device='cuda:0'), grad: tensor([[ 3.3539e-02, 2.4433e-03, 4.3488e-03, ..., 9.6436e-03, + 9.1782e-03, 9.7198e-03], + [ 3.8635e-02, 4.2038e-03, 5.7144e-03, ..., 1.1887e-02, + 1.2558e-02, 1.4526e-02], + [ 1.7042e-03, 1.3375e-04, 2.2352e-04, ..., 4.9543e-04, + 4.7588e-04, 5.0545e-04], + ..., + [-3.3905e-02, -4.1199e-03, -5.1804e-03, ..., -1.0689e-02, + -1.1513e-02, -1.3420e-02], + [-4.2877e-02, -2.8706e-03, -5.4817e-03, ..., -1.2169e-02, + -1.1497e-02, -1.2177e-02], + [ 6.4850e-04, 4.6849e-05, 8.4221e-05, ..., 1.8620e-04, + 1.7869e-04, 1.9145e-04]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0856, 0.0118, 0.0120, -0.0648, 0.1293, 0.0469, -0.0419], + device='cuda:0'), grad: tensor([ 0.1093, 0.1780, 0.0057, 0.0076, -0.1642, -0.1384, 0.0022], + device='cuda:0') +249 +0.006330184227833378 +changing lr +epoch 29, time 328.61, cls_loss 0.5769 cls_loss_mapping 0.0458 cls_loss_causal 0.4448 re_mapping 0.0318 re_causal 0.0321 /// teacc 91.23 lr 0.00611260 +Epoch 31, weight, value: tensor([[ 0.0227, 0.0393, -0.0026, ..., 0.0007, -0.0055, 0.0133], + [-0.0735, -0.0797, -0.1031, ..., 0.0189, 0.0667, -0.0018], + [-0.0030, -0.0226, -0.0156, ..., 0.0212, -0.0272, -0.0356], + ..., + [-0.0894, -0.0942, -0.0486, ..., 0.0219, 0.0020, 0.0137], + [-0.0473, -0.0221, -0.0541, ..., -0.1089, -0.1438, -0.1290], + [ 0.1950, 0.2227, 0.2500, ..., -0.1019, -0.0392, -0.0445]], + device='cuda:0'), grad: tensor([[ 2.0752e-02, 3.4256e-03, 3.6106e-03, ..., 5.5504e-03, + 5.8594e-03, 6.2523e-03], + [ 9.7427e-03, 1.3885e-03, 1.6747e-03, ..., 1.0185e-03, + 1.2217e-03, 1.5631e-03], + [-6.3965e-02, -9.3765e-03, -1.0880e-02, ..., -9.0942e-03, + -1.0384e-02, -1.2352e-02], + ..., + [ 3.4294e-03, 3.5167e-04, 4.5967e-04, ..., -5.0575e-05, + 8.6367e-05, 2.1911e-04], + [ 7.1716e-03, 1.0052e-03, 1.2264e-03, ..., 6.1607e-04, + 7.6723e-04, 1.0319e-03], + [ 1.6373e-02, 2.2926e-03, 2.7981e-03, ..., 1.3905e-03, + 1.7357e-03, 2.3403e-03]], device='cuda:0') +Epoch 31, bias, value: tensor([-0.0840, 0.0259, 0.0214, -0.0854, 0.1202, 0.0166, -0.0070], + device='cuda:0'), grad: tensor([ 0.0748, 0.0286, -0.1986, 0.0188, 0.0090, 0.0206, 0.0470], + device='cuda:0') +249 +0.006112604669781575 +changing lr +epoch 30, time 331.79, cls_loss 0.5156 cls_loss_mapping 0.0503 cls_loss_causal 0.4165 re_mapping 0.0308 re_causal 0.0313 /// teacc 92.98 lr 0.00589278 +Epoch 32, weight, value: tensor([[ 0.0300, 0.0474, 0.0041, ..., -0.0004, -0.0067, 0.0131], + [-0.0771, -0.0815, -0.1040, ..., 0.0176, 0.0646, -0.0019], + [-0.0086, -0.0254, -0.0183, ..., 0.0227, -0.0271, -0.0362], + ..., + [-0.0879, -0.0921, -0.0471, ..., 0.0200, 0.0009, 0.0115], + [-0.0438, -0.0222, -0.0538, ..., -0.1073, -0.1405, -0.1267], + [ 0.1896, 0.2165, 0.2438, ..., -0.1061, -0.0442, -0.0494]], + device='cuda:0'), grad: tensor([[ 0.0073, 0.0009, 0.0012, ..., -0.0005, 0.0002, 0.0004], + [ 0.0295, 0.0042, 0.0056, ..., 0.0088, 0.0124, 0.0111], + [ 0.0060, 0.0009, 0.0011, ..., 0.0005, 0.0009, 0.0011], + ..., + [-0.0171, -0.0021, -0.0033, ..., -0.0075, -0.0106, -0.0087], + [ 0.0152, 0.0023, 0.0027, ..., 0.0009, 0.0019, 0.0024], + [-0.0491, -0.0074, -0.0089, ..., -0.0028, -0.0059, -0.0076]], + device='cuda:0') +Epoch 32, bias, value: tensor([-0.0674, 0.0170, -0.0022, -0.0813, 0.1231, 0.0158, 0.0026], + device='cuda:0'), grad: tensor([ 0.0229, 0.1025, 0.0175, 0.0244, -0.0687, 0.0443, -0.1429], + device='cuda:0') +249 +0.005892784473993186 +changing lr +epoch 31, time 330.29, cls_loss 0.5033 cls_loss_mapping 0.0417 cls_loss_causal 0.3811 re_mapping 0.0303 re_causal 0.0305 /// teacc 96.49 lr 0.00567117 +Epoch 33, weight, value: tensor([[ 2.1752e-02, 4.0650e-02, -1.1649e-03, ..., -1.9465e-04, + -6.8365e-03, 1.3906e-02], + [-7.6714e-02, -7.9450e-02, -1.0255e-01, ..., 1.9779e-02, + 6.5701e-02, 6.6563e-04], + [-2.2845e-03, -2.4987e-02, -1.7244e-02, ..., 2.6580e-02, + -2.3315e-02, -3.2678e-02], + ..., + [-8.3019e-02, -8.7031e-02, -4.2755e-02, ..., 1.7453e-02, + -1.3261e-04, 9.4224e-03], + [-4.5152e-02, -2.4074e-02, -5.6061e-02, ..., -1.0920e-01, + -1.4236e-01, -1.2901e-01], + [ 1.8911e-01, 2.1565e-01, 2.4291e-01, ..., -1.0694e-01, + -4.5432e-02, -5.0759e-02]], device='cuda:0'), grad: tensor([[ 0.0145, 0.0023, 0.0023, ..., 0.0044, 0.0057, 0.0057], + [ 0.0072, 0.0024, 0.0026, ..., 0.0016, 0.0022, 0.0022], + [-0.0324, -0.0049, -0.0055, ..., -0.0098, -0.0128, -0.0128], + ..., + [ 0.0130, 0.0023, 0.0026, ..., 0.0038, 0.0050, 0.0050], + [ 0.0008, 0.0002, 0.0002, ..., 0.0002, 0.0003, 0.0003], + [-0.0041, -0.0024, -0.0024, ..., -0.0005, -0.0007, -0.0008]], + device='cuda:0') +Epoch 33, bias, value: tensor([-0.0675, 0.0047, 0.0232, -0.0976, 0.1352, 0.0036, 0.0059], + device='cuda:0'), grad: tensor([ 0.0587, 0.0216, -0.1288, 0.0037, 0.0499, 0.0028, -0.0079], + device='cuda:0') +249 +0.00567116632908828 +changing lr +epoch 32, time 333.62, cls_loss 0.5276 cls_loss_mapping 0.0446 cls_loss_causal 0.4236 re_mapping 0.0290 re_causal 0.0293 /// teacc 89.47 lr 0.00544820 +Epoch 34, weight, value: tensor([[ 0.0221, 0.0405, -0.0008, ..., 0.0050, -0.0014, 0.0214], + [-0.0757, -0.0799, -0.1037, ..., 0.0182, 0.0634, -0.0005], + [-0.0067, -0.0259, -0.0180, ..., 0.0251, -0.0245, -0.0348], + ..., + [-0.0869, -0.0876, -0.0439, ..., 0.0177, 0.0007, 0.0110], + [-0.0463, -0.0271, -0.0585, ..., -0.1112, -0.1442, -0.1317], + [ 0.1905, 0.2169, 0.2438, ..., -0.1069, -0.0458, -0.0530]], + device='cuda:0'), grad: tensor([[ 7.9269e-03, 8.4066e-04, 1.2541e-03, ..., 1.1244e-03, + 1.1339e-03, 1.0099e-03], + [ 7.5073e-03, 7.8869e-04, 1.1806e-03, ..., 1.0662e-03, + 1.0748e-03, 9.5749e-04], + [-2.7054e-02, -2.7981e-03, -4.2114e-03, ..., -3.8090e-03, + -3.8433e-03, -3.4180e-03], + ..., + [ 4.0207e-03, 4.3082e-04, 6.3992e-04, ..., 5.8174e-04, + 5.8556e-04, 5.2357e-04], + [ 2.4376e-03, 2.4140e-04, 3.7503e-04, ..., 2.5368e-04, + 2.6608e-04, 2.2280e-04], + [ 3.8815e-04, -6.7055e-05, -4.5568e-05, ..., 2.9266e-05, + 2.8685e-05, 2.1294e-05]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0257, 0.0073, -0.0020, -0.0803, 0.1131, 0.0054, -0.0103], + device='cuda:0'), grad: tensor([ 0.0328, 0.0311, -0.1124, 0.0193, 0.0166, 0.0103, 0.0022], + device='cuda:0') +249 +0.00544819654451717 +changing lr +epoch 33, time 329.71, cls_loss 0.4845 cls_loss_mapping 0.0441 cls_loss_causal 0.3794 re_mapping 0.0287 re_causal 0.0289 /// teacc 92.40 lr 0.00522432 +Epoch 35, weight, value: tensor([[ 0.0168, 0.0360, -0.0060, ..., 0.0067, 0.0009, 0.0228], + [-0.0761, -0.0776, -0.1011, ..., 0.0209, 0.0654, 0.0024], + [-0.0038, -0.0265, -0.0190, ..., 0.0260, -0.0229, -0.0342], + ..., + [-0.0823, -0.0832, -0.0386, ..., 0.0122, -0.0051, 0.0061], + [-0.0506, -0.0288, -0.0606, ..., -0.1109, -0.1443, -0.1310], + [ 0.1946, 0.2181, 0.2456, ..., -0.1053, -0.0442, -0.0519]], + device='cuda:0'), grad: tensor([[-0.0533, -0.0167, -0.0165, ..., -0.0188, -0.0193, -0.0189], + [ 0.0446, 0.0144, 0.0143, ..., 0.0155, 0.0158, 0.0155], + [ 0.0016, 0.0005, 0.0005, ..., 0.0007, 0.0007, 0.0007], + ..., + [ 0.0037, 0.0012, 0.0011, ..., 0.0013, 0.0014, 0.0013], + [-0.0082, -0.0019, -0.0016, ..., -0.0042, -0.0047, -0.0044], + [-0.0067, -0.0030, -0.0032, ..., -0.0017, -0.0014, -0.0014]], + device='cuda:0') +Epoch 35, bias, value: tensor([-0.0323, -0.0002, 0.0048, -0.0758, 0.1144, 0.0019, -0.0051], + device='cuda:0'), grad: tensor([-0.1429, 0.1179, 0.0044, 0.0507, 0.0101, -0.0270, -0.0131], + device='cuda:0') +249 +0.005224324151752577 +changing lr +epoch 34, time 330.52, cls_loss 0.4793 cls_loss_mapping 0.0372 cls_loss_causal 0.3881 re_mapping 0.0276 re_causal 0.0279 /// teacc 98.25 lr 0.00500000 +Epoch 36, weight, value: tensor([[ 0.0177, 0.0385, -0.0037, ..., 0.0109, 0.0039, 0.0261], + [-0.0737, -0.0744, -0.0980, ..., 0.0228, 0.0668, 0.0034], + [-0.0021, -0.0254, -0.0185, ..., 0.0254, -0.0230, -0.0334], + ..., + [-0.0790, -0.0835, -0.0385, ..., 0.0075, -0.0074, 0.0026], + [-0.0455, -0.0271, -0.0583, ..., -0.1087, -0.1419, -0.1277], + [ 0.1882, 0.2120, 0.2393, ..., -0.1064, -0.0460, -0.0534]], + device='cuda:0'), grad: tensor([[ 2.0466e-03, 2.3782e-04, 4.1747e-04, ..., 9.5963e-05, + 2.3568e-04, 2.4724e-04], + [ 6.3777e-05, 9.9838e-06, 1.4998e-05, ..., 7.1898e-06, + 1.0766e-05, 1.1154e-05], + [ 4.1991e-05, 5.5321e-06, 8.9332e-06, ..., 5.7667e-06, + 8.1062e-06, 8.2999e-06], + ..., + [-2.1267e-03, -2.4056e-04, -4.2892e-04, ..., -8.3625e-05, + -2.3139e-04, -2.4307e-04], + [ 9.8571e-06, 2.1067e-06, 2.8666e-06, ..., 4.2766e-06, + 3.8184e-06, 3.8184e-06], + [-1.4260e-05, -1.0580e-05, -1.0520e-05, ..., -3.3900e-06, + -2.9430e-06, -2.9858e-06]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0506, -0.0003, 0.0084, -0.1080, 0.1157, 0.0425, -0.0002], + device='cuda:0'), grad: tensor([ 8.5678e-03, 2.5105e-04, 1.7083e-04, -5.4926e-05, -8.9493e-03, + 3.3110e-05, -1.0870e-05], device='cuda:0') +249 +0.005000000000000003 +changing lr +epoch 35, time 329.50, cls_loss 0.4577 cls_loss_mapping 0.0368 cls_loss_causal 0.3710 re_mapping 0.0269 re_causal 0.0275 /// teacc 88.89 lr 0.00477568 +Epoch 37, weight, value: tensor([[ 0.0243, 0.0433, 0.0029, ..., 0.0153, 0.0096, 0.0301], + [-0.0743, -0.0760, -0.0994, ..., 0.0199, 0.0636, 0.0008], + [-0.0072, -0.0256, -0.0193, ..., 0.0236, -0.0258, -0.0361], + ..., + [-0.0845, -0.0856, -0.0421, ..., 0.0069, -0.0088, 0.0023], + [-0.0473, -0.0269, -0.0583, ..., -0.1085, -0.1416, -0.1278], + [ 0.1903, 0.2097, 0.2371, ..., -0.1051, -0.0446, -0.0516]], + device='cuda:0'), grad: tensor([[-0.0939, -0.0243, -0.0273, ..., -0.0372, -0.0416, -0.0429], + [-0.0333, -0.0091, -0.0090, ..., -0.0050, -0.0061, -0.0099], + [ 0.0591, 0.0145, 0.0168, ..., 0.0236, 0.0245, 0.0262], + ..., + [ 0.0277, 0.0069, 0.0073, ..., 0.0098, 0.0125, 0.0130], + [ 0.0064, 0.0018, 0.0018, ..., 0.0015, 0.0018, 0.0023], + [ 0.0249, 0.0079, 0.0079, ..., 0.0051, 0.0060, 0.0079]], + device='cuda:0') +Epoch 37, bias, value: tensor([-0.0554, 0.0193, -0.0239, -0.0958, 0.1055, 0.0275, 0.0302], + device='cuda:0'), grad: tensor([-0.2903, -0.1429, 0.1881, 0.0349, 0.0937, 0.0249, 0.0914], + device='cuda:0') +249 +0.004775675848247429 +changing lr +epoch 36, time 332.31, cls_loss 0.4722 cls_loss_mapping 0.0348 cls_loss_causal 0.3787 re_mapping 0.0263 re_causal 0.0266 /// teacc 95.32 lr 0.00455180 +Epoch 38, weight, value: tensor([[ 0.0165, 0.0405, -0.0005, ..., 0.0109, 0.0055, 0.0266], + [-0.0711, -0.0751, -0.0988, ..., 0.0172, 0.0605, -0.0024], + [-0.0049, -0.0245, -0.0187, ..., 0.0232, -0.0269, -0.0370], + ..., + [-0.0820, -0.0839, -0.0394, ..., 0.0061, -0.0089, 0.0024], + [-0.0473, -0.0280, -0.0586, ..., -0.1047, -0.1381, -0.1243], + [ 0.1912, 0.2107, 0.2379, ..., -0.1023, -0.0421, -0.0494]], + device='cuda:0'), grad: tensor([[ 6.7663e-04, 4.4465e-05, 7.9989e-05, ..., 2.5034e-05, + 4.6313e-05, 5.1856e-05], + [ 3.5828e-02, 2.1362e-03, 4.0321e-03, ..., 1.2283e-03, + 2.3632e-03, 2.6569e-03], + [-4.2236e-02, -2.5177e-03, -4.7531e-03, ..., -1.4486e-03, + -2.7866e-03, -3.1319e-03], + ..., + [ 3.7422e-03, 2.2364e-04, 4.2176e-04, ..., 1.2887e-04, + 2.4748e-04, 2.7800e-04], + [ 7.0095e-04, 4.2766e-05, 7.9811e-05, ..., 2.4512e-05, + 4.6700e-05, 5.2422e-05], + [ 6.7997e-04, 3.4124e-05, 7.0632e-05, ..., 2.0579e-05, + 4.2439e-05, 4.8012e-05]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0852, 0.0306, -0.0319, -0.0947, 0.1212, 0.0330, 0.0343], + device='cuda:0'), grad: tensor([ 0.0026, 0.1387, -0.1635, 0.0023, 0.0145, 0.0027, 0.0027], + device='cuda:0') +249 +0.004551803455482836 +changing lr +epoch 37, time 331.27, cls_loss 0.4758 cls_loss_mapping 0.0314 cls_loss_causal 0.3870 re_mapping 0.0256 re_causal 0.0268 /// teacc 95.32 lr 0.00432883 +Epoch 39, weight, value: tensor([[ 0.0204, 0.0393, -0.0012, ..., 0.0124, 0.0081, 0.0289], + [-0.0737, -0.0760, -0.0994, ..., 0.0153, 0.0584, -0.0044], + [-0.0019, -0.0237, -0.0170, ..., 0.0240, -0.0261, -0.0358], + ..., + [-0.0790, -0.0820, -0.0383, ..., 0.0030, -0.0121, -0.0011], + [-0.0506, -0.0307, -0.0612, ..., -0.1041, -0.1372, -0.1233], + [ 0.1883, 0.2107, 0.2377, ..., -0.1039, -0.0440, -0.0514]], + device='cuda:0'), grad: tensor([[ 0.0248, 0.0056, 0.0056, ..., 0.0067, 0.0061, 0.0070], + [-0.0345, -0.0099, -0.0098, ..., -0.0067, -0.0051, -0.0066], + [ 0.0201, 0.0054, 0.0055, ..., 0.0059, 0.0054, 0.0061], + ..., + [ 0.0186, 0.0044, 0.0045, ..., 0.0054, 0.0049, 0.0056], + [-0.0507, -0.0112, -0.0118, ..., -0.0191, -0.0187, -0.0202], + [ 0.0048, 0.0012, 0.0012, ..., 0.0012, 0.0011, 0.0013]], + device='cuda:0') +Epoch 39, bias, value: tensor([-0.0647, 0.0238, -0.0303, -0.1071, 0.1277, 0.0334, 0.0246], + device='cuda:0'), grad: tensor([ 0.0852, -0.0952, 0.0595, 0.0500, 0.0615, -0.1760, 0.0149], + device='cuda:0') +249 +0.004328833670911726 +changing lr +epoch 38, time 329.95, cls_loss 0.4622 cls_loss_mapping 0.0261 cls_loss_causal 0.3888 re_mapping 0.0249 re_causal 0.0258 /// teacc 98.25 lr 0.00410722 +Epoch 40, weight, value: tensor([[ 0.0149, 0.0370, -0.0035, ..., 0.0090, 0.0052, 0.0259], + [-0.0703, -0.0735, -0.0964, ..., 0.0148, 0.0582, -0.0040], + [-0.0012, -0.0236, -0.0168, ..., 0.0260, -0.0249, -0.0345], + ..., + [-0.0772, -0.0814, -0.0387, ..., 0.0027, -0.0121, -0.0015], + [-0.0504, -0.0318, -0.0618, ..., -0.1026, -0.1350, -0.1216], + [ 0.1901, 0.2120, 0.2392, ..., -0.1025, -0.0427, -0.0498]], + device='cuda:0'), grad: tensor([[ 8.7280e-03, 1.4296e-03, 1.9016e-03, ..., 1.9369e-03, + 1.8005e-03, 2.0885e-03], + [ 1.9894e-03, 1.4043e-04, -1.2755e-04, ..., -4.7827e-04, + 1.3947e-04, 2.4438e-04], + [ 8.3771e-03, 1.2226e-03, 1.4896e-03, ..., 6.0511e-04, + 8.0347e-04, 1.1272e-03], + ..., + [ 4.4785e-03, 7.2813e-04, 8.6689e-04, ..., -3.9554e-04, + -2.9302e-04, -9.9182e-05], + [-3.6255e-02, -5.4550e-03, -6.4697e-03, ..., -2.7485e-03, + -3.8147e-03, -5.2071e-03], + [ 9.4299e-03, 1.4324e-03, 1.7195e-03, ..., 7.7343e-04, + 1.0099e-03, 1.3704e-03]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0847, 0.0377, -0.0310, -0.1113, 0.1245, 0.0389, 0.0333], + device='cuda:0'), grad: tensor([ 0.0291, 0.0120, 0.0299, 0.0112, 0.0142, -0.1298, 0.0334], + device='cuda:0') +249 +0.0041072155260068206 +changing lr +epoch 39, time 331.15, cls_loss 0.4891 cls_loss_mapping 0.0286 cls_loss_causal 0.3927 re_mapping 0.0243 re_causal 0.0260 /// teacc 96.49 lr 0.00388740 +Epoch 41, weight, value: tensor([[ 0.0134, 0.0357, -0.0046, ..., 0.0067, 0.0024, 0.0236], + [-0.0727, -0.0737, -0.0978, ..., 0.0178, 0.0615, -0.0006], + [-0.0003, -0.0209, -0.0139, ..., 0.0245, -0.0252, -0.0349], + ..., + [-0.0832, -0.0844, -0.0419, ..., 0.0015, -0.0137, -0.0031], + [-0.0488, -0.0334, -0.0633, ..., -0.1002, -0.1335, -0.1208], + [ 0.1972, 0.2152, 0.2422, ..., -0.1017, -0.0426, -0.0496]], + device='cuda:0'), grad: tensor([[-0.0276, -0.0074, -0.0107, ..., -0.0043, -0.0054, -0.0093], + [ 0.0037, 0.0011, 0.0015, ..., 0.0006, 0.0008, 0.0013], + [-0.0006, -0.0001, 0.0002, ..., -0.0007, -0.0007, -0.0003], + ..., + [ 0.0030, 0.0008, 0.0011, ..., 0.0006, 0.0007, 0.0011], + [ 0.0099, 0.0026, 0.0034, ..., 0.0024, 0.0031, 0.0038], + [ 0.0093, 0.0024, 0.0035, ..., 0.0015, 0.0019, 0.0032]], + device='cuda:0') +Epoch 41, bias, value: tensor([-0.0931, 0.0289, -0.0288, -0.0874, 0.1113, 0.0396, 0.0367], + device='cuda:0'), grad: tensor([-0.1532, 0.0199, 0.0037, 0.0134, 0.0159, 0.0478, 0.0524], + device='cuda:0') +249 +0.0038873953302184317 +changing lr +epoch 40, time 330.06, cls_loss 0.4254 cls_loss_mapping 0.0236 cls_loss_causal 0.3561 re_mapping 0.0231 re_causal 0.0244 /// teacc 92.98 lr 0.00366982 +Epoch 42, weight, value: tensor([[ 0.0150, 0.0362, -0.0038, ..., 0.0082, 0.0046, 0.0268], + [-0.0688, -0.0728, -0.0960, ..., 0.0217, 0.0657, 0.0036], + [-0.0004, -0.0203, -0.0136, ..., 0.0231, -0.0259, -0.0358], + ..., + [-0.0803, -0.0846, -0.0429, ..., -0.0012, -0.0176, -0.0070], + [-0.0511, -0.0327, -0.0628, ..., -0.0992, -0.1329, -0.1200], + [ 0.1931, 0.2138, 0.2410, ..., -0.1026, -0.0443, -0.0515]], + device='cuda:0'), grad: tensor([[ 0.0127, 0.0027, 0.0024, ..., 0.0022, 0.0020, 0.0021], + [ 0.0206, 0.0048, 0.0042, ..., 0.0034, 0.0036, 0.0039], + [ 0.0106, 0.0020, 0.0019, ..., 0.0019, 0.0016, 0.0017], + ..., + [ 0.0131, 0.0030, 0.0026, ..., 0.0022, 0.0022, 0.0024], + [-0.0783, -0.0177, -0.0156, ..., -0.0131, -0.0128, -0.0140], + [ 0.0177, 0.0043, 0.0037, ..., 0.0028, 0.0029, 0.0032]], + device='cuda:0') +Epoch 42, bias, value: tensor([-0.0796, 0.0295, -0.0175, -0.0914, 0.1212, 0.0197, 0.0252], + device='cuda:0'), grad: tensor([ 0.0528, 0.0803, 0.0445, 0.0146, 0.0527, -0.3176, 0.0727], + device='cuda:0') +249 +0.003669815772166629 +changing lr +epoch 41, time 328.71, cls_loss 0.4690 cls_loss_mapping 0.0286 cls_loss_causal 0.3983 re_mapping 0.0226 re_causal 0.0241 /// teacc 94.74 lr 0.00345492 +Epoch 43, weight, value: tensor([[ 0.0134, 0.0368, -0.0032, ..., 0.0068, 0.0038, 0.0265], + [-0.0706, -0.0740, -0.0973, ..., 0.0228, 0.0667, 0.0047], + [ 0.0005, -0.0201, -0.0136, ..., 0.0235, -0.0249, -0.0351], + ..., + [-0.0788, -0.0842, -0.0421, ..., 0.0018, -0.0162, -0.0059], + [-0.0478, -0.0321, -0.0624, ..., -0.1000, -0.1331, -0.1203], + [ 0.1924, 0.2145, 0.2422, ..., -0.1034, -0.0456, -0.0525]], + device='cuda:0'), grad: tensor([[-0.0735, -0.0339, -0.0370, ..., -0.0244, -0.0208, -0.0199], + [ 0.0218, 0.0130, 0.0155, ..., 0.0057, 0.0050, 0.0032], + [ 0.0021, 0.0008, 0.0008, ..., 0.0007, 0.0006, 0.0006], + ..., + [ 0.0274, 0.0106, 0.0107, ..., 0.0102, 0.0087, 0.0094], + [ 0.0026, 0.0010, 0.0011, ..., 0.0010, 0.0008, 0.0009], + [ 0.0171, 0.0075, 0.0080, ..., 0.0059, 0.0050, 0.0050]], + device='cuda:0') +Epoch 43, bias, value: tensor([-0.0770, 0.0188, -0.0099, -0.0947, 0.0983, 0.0380, 0.0337], + device='cuda:0'), grad: tensor([-0.1666, 0.0365, 0.0053, 0.0061, 0.0714, 0.0068, 0.0406], + device='cuda:0') +249 +0.0034549150281252667 +changing lr +epoch 42, time 329.62, cls_loss 0.4459 cls_loss_mapping 0.0224 cls_loss_causal 0.3648 re_mapping 0.0226 re_causal 0.0240 /// teacc 97.66 lr 0.00324313 +Epoch 44, weight, value: tensor([[ 0.0138, 0.0355, -0.0043, ..., 0.0053, 0.0032, 0.0258], + [-0.0675, -0.0713, -0.0948, ..., 0.0198, 0.0633, 0.0018], + [-0.0014, -0.0207, -0.0147, ..., 0.0218, -0.0262, -0.0359], + ..., + [-0.0824, -0.0850, -0.0429, ..., 0.0046, -0.0129, -0.0025], + [-0.0487, -0.0311, -0.0620, ..., -0.0980, -0.1314, -0.1188], + [ 0.1923, 0.2136, 0.2416, ..., -0.1026, -0.0453, -0.0526]], + device='cuda:0'), grad: tensor([[ 0.0080, 0.0016, 0.0021, ..., 0.0042, 0.0038, 0.0043], + [ 0.0070, 0.0014, 0.0017, ..., 0.0009, 0.0012, 0.0013], + [ 0.0114, 0.0022, 0.0029, ..., 0.0010, 0.0015, 0.0018], + ..., + [ 0.0037, 0.0008, 0.0010, ..., 0.0002, 0.0004, 0.0004], + [ 0.0077, 0.0015, 0.0020, ..., 0.0007, 0.0011, 0.0012], + [-0.0420, -0.0083, -0.0107, ..., -0.0075, -0.0086, -0.0098]], + device='cuda:0') +Epoch 44, bias, value: tensor([-0.0628, 0.0187, -0.0107, -0.0785, 0.0837, 0.0302, 0.0266], + device='cuda:0'), grad: tensor([-0.0007, 0.0299, 0.0517, 0.0183, 0.0171, 0.0347, -0.1510], + device='cuda:0') +249 +0.0032431258795932905 +changing lr +epoch 43, time 331.60, cls_loss 0.4038 cls_loss_mapping 0.0240 cls_loss_causal 0.3298 re_mapping 0.0213 re_causal 0.0221 /// teacc 96.49 lr 0.00303487 +Epoch 45, weight, value: tensor([[ 0.0169, 0.0380, -0.0020, ..., 0.0042, 0.0024, 0.0249], + [-0.0663, -0.0706, -0.0934, ..., 0.0192, 0.0633, 0.0023], + [-0.0003, -0.0197, -0.0144, ..., 0.0244, -0.0244, -0.0343], + ..., + [-0.0832, -0.0875, -0.0458, ..., 0.0054, -0.0117, -0.0012], + [-0.0486, -0.0313, -0.0620, ..., -0.0991, -0.1322, -0.1199], + [ 0.1938, 0.2147, 0.2430, ..., -0.1019, -0.0450, -0.0524]], + device='cuda:0'), grad: tensor([[ 1.5078e-03, 2.7275e-04, 2.8729e-04, ..., 6.9809e-04, + 7.5102e-04, 6.9666e-04], + [-9.6369e-04, -4.1127e-04, -3.8123e-04, ..., 3.4904e-04, + 2.3067e-04, 2.7132e-04], + [ 5.9557e-04, 9.5010e-05, 1.0246e-04, ..., 3.1090e-04, + 3.2711e-04, 3.0613e-04], + ..., + [ 5.4245e-03, 9.3651e-04, 9.9277e-04, ..., 2.5997e-03, + 2.7733e-03, 2.5806e-03], + [ 5.6267e-03, 8.1110e-04, 8.9502e-04, ..., 3.2158e-03, + 3.3360e-03, 3.1414e-03], + [ 2.8586e-04, 5.3197e-05, 5.4866e-05, ..., 1.1313e-04, + 1.2589e-04, 1.1539e-04]], device='cuda:0') +Epoch 45, bias, value: tensor([-0.0503, 0.0255, -0.0244, -0.0944, 0.0951, 0.0332, 0.0224], + device='cuda:0'), grad: tensor([ 0.0054, -0.0022, 0.0022, -0.0475, 0.0197, 0.0213, 0.0010], + device='cuda:0') +249 +0.0030348748417303863 +changing lr +epoch 44, time 329.83, cls_loss 0.4152 cls_loss_mapping 0.0238 cls_loss_causal 0.3450 re_mapping 0.0213 re_causal 0.0223 /// teacc 84.80 lr 0.00283058 +Epoch 46, weight, value: tensor([[ 0.0155, 0.0365, -0.0030, ..., 0.0026, 0.0018, 0.0244], + [-0.0683, -0.0706, -0.0937, ..., 0.0217, 0.0646, 0.0038], + [ 0.0038, -0.0191, -0.0134, ..., 0.0244, -0.0242, -0.0344], + ..., + [-0.0862, -0.0878, -0.0461, ..., 0.0034, -0.0137, -0.0035], + [-0.0493, -0.0316, -0.0624, ..., -0.1001, -0.1328, -0.1208], + [ 0.1910, 0.2135, 0.2419, ..., -0.1018, -0.0455, -0.0523]], + device='cuda:0'), grad: tensor([[ 6.7673e-03, 1.6346e-03, 1.1425e-03, ..., 2.4128e-03, + 2.2144e-03, 2.3632e-03], + [-4.7379e-03, -1.2341e-03, -6.3896e-04, ..., -6.4945e-04, + -3.0136e-04, -4.6825e-04], + [ 4.0007e-04, 8.6546e-05, 8.7738e-05, ..., 2.6608e-04, + 2.7943e-04, 2.8062e-04], + ..., + [-3.1090e-03, -6.3133e-04, -7.3338e-04, ..., -2.4700e-03, + -2.6531e-03, -2.6379e-03], + [-9.3132e-06, -6.1691e-06, -1.0207e-05, ..., -1.0028e-05, + -1.3426e-05, -1.2413e-05], + [ 3.9518e-05, 5.3868e-06, 1.2144e-06, ..., 1.3329e-05, + 1.2040e-05, 1.2964e-05]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0518, 0.0027, -0.0073, -0.0779, 0.0876, 0.0297, 0.0241], + device='cuda:0'), grad: tensor([ 2.3590e-02, -1.6129e-02, 1.4343e-03, 2.2850e-03, -1.1353e-02, + -9.8813e-07, 1.5974e-04], device='cuda:0') +249 +0.0028305813044122124 +changing lr +epoch 45, time 328.07, cls_loss 0.4047 cls_loss_mapping 0.0210 cls_loss_causal 0.3507 re_mapping 0.0208 re_causal 0.0218 /// teacc 97.66 lr 0.00263066 +Epoch 47, weight, value: tensor([[ 1.6543e-02, 3.7707e-02, -1.7139e-03, ..., -2.0231e-04, + -7.0720e-06, 2.2496e-02], + [-6.9729e-02, -7.1758e-02, -9.4664e-02, ..., 2.2367e-02, + 6.4598e-02, 4.2657e-03], + [ 1.1171e-03, -1.8672e-02, -1.3626e-02, ..., 2.3659e-02, + -2.5242e-02, -3.5041e-02], + ..., + [-8.2318e-02, -8.6204e-02, -4.4770e-02, ..., 6.1175e-03, + -1.1213e-02, -7.6069e-04], + [-4.9991e-02, -3.2823e-02, -6.3484e-02, ..., -9.9780e-02, + -1.3256e-01, -1.2100e-01], + [ 1.9374e-01, 2.1411e-01, 2.4282e-01, ..., -1.0082e-01, + -4.4396e-02, -5.1597e-02]], device='cuda:0'), grad: tensor([[ 9.4557e-04, 1.0705e-04, 1.7452e-04, ..., 4.2152e-04, + 3.1972e-04, 3.0303e-04], + [ 7.7295e-04, 8.3983e-05, 1.4055e-04, ..., 3.4189e-04, + 2.5654e-04, 2.4199e-04], + [-5.2452e-03, -4.2748e-04, -8.2350e-04, ..., -2.3937e-03, + -1.7929e-03, -1.7033e-03], + ..., + [ 3.0727e-03, 2.5177e-04, 4.8304e-04, ..., 1.3981e-03, + 1.0462e-03, 9.9277e-04], + [ 3.0994e-04, 2.3350e-05, 4.9412e-05, ..., 1.4484e-04, + 1.0562e-04, 1.0049e-04], + [-5.4896e-05, -6.5506e-05, -6.4015e-05, ..., 2.0601e-06, + -1.8012e-06, 2.0079e-06]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0488, -0.0029, -0.0215, -0.0863, 0.1010, 0.0324, 0.0333], + device='cuda:0'), grad: tensor([ 0.0042, 0.0035, -0.0245, 0.0009, 0.0143, 0.0015, 0.0001], + device='cuda:0') +249 +0.0026306566876350096 +changing lr +---------------------saving model at epoch 46---------------------------------------------------- +epoch 46, time 331.01, cls_loss 0.3899 cls_loss_mapping 0.0215 cls_loss_causal 0.3216 re_mapping 0.0202 re_causal 0.0212 /// teacc 99.42 lr 0.00243550 +Epoch 48, weight, value: tensor([[ 0.0213, 0.0411, 0.0021, ..., 0.0023, 0.0021, 0.0245], + [-0.0673, -0.0714, -0.0943, ..., 0.0232, 0.0658, 0.0055], + [-0.0024, -0.0201, -0.0155, ..., 0.0226, -0.0265, -0.0361], + ..., + [-0.0821, -0.0858, -0.0449, ..., 0.0043, -0.0129, -0.0025], + [-0.0498, -0.0328, -0.0636, ..., -0.0994, -0.1319, -0.1205], + [ 0.1895, 0.2118, 0.2408, ..., -0.1011, -0.0448, -0.0520]], + device='cuda:0'), grad: tensor([[ 0.0203, 0.0044, 0.0062, ..., 0.0067, 0.0066, 0.0069], + [-0.0407, -0.0050, -0.0075, ..., -0.0256, -0.0225, -0.0193], + [ 0.0055, 0.0016, 0.0021, ..., 0.0009, 0.0011, 0.0015], + ..., + [ 0.0432, 0.0081, 0.0111, ..., 0.0199, 0.0184, 0.0175], + [-0.0187, -0.0058, -0.0075, ..., -0.0020, -0.0029, -0.0046], + [ 0.0110, 0.0033, 0.0043, ..., 0.0015, 0.0020, 0.0029]], + device='cuda:0') +Epoch 48, bias, value: tensor([-4.2536e-02, 5.2242e-05, -3.5888e-02, -8.0597e-02, 1.0369e-01, + 3.2327e-02, 3.0015e-02], device='cuda:0'), grad: tensor([ 0.0827, -0.1013, 0.0276, -0.1138, 0.1481, -0.1000, 0.0569], + device='cuda:0') +249 +0.0024355036129704724 +changing lr +epoch 47, time 330.12, cls_loss 0.4134 cls_loss_mapping 0.0203 cls_loss_causal 0.3577 re_mapping 0.0194 re_causal 0.0206 /// teacc 98.25 lr 0.00224552 +Epoch 49, weight, value: tensor([[ 1.9520e-02, 4.0544e-02, 1.2135e-03, ..., 2.9020e-03, + 2.6058e-03, 2.5107e-02], + [-6.9452e-02, -7.1449e-02, -9.4182e-02, ..., 2.1665e-02, + 6.3694e-02, 4.0164e-03], + [-1.2825e-04, -2.0393e-02, -1.5923e-02, ..., 2.3945e-02, + -2.5404e-02, -3.4911e-02], + ..., + [-8.3666e-02, -8.5399e-02, -4.4409e-02, ..., 4.9752e-03, + -1.1643e-02, -1.6987e-03], + [-4.6602e-02, -3.2437e-02, -6.2693e-02, ..., -9.9265e-02, + -1.3136e-01, -1.1988e-01], + [ 1.9226e-01, 2.1261e-01, 2.4166e-01, ..., -1.0052e-01, + -4.4378e-02, -5.1521e-02]], device='cuda:0'), grad: tensor([[ 2.3823e-03, 4.2057e-04, 5.5218e-04, ..., 6.0415e-04, + 5.7507e-04, 5.6982e-04], + [ 1.8587e-03, 3.1400e-04, 4.1485e-04, ..., 3.9101e-04, + 3.6907e-04, 3.5381e-04], + [ 5.3062e-03, 8.8596e-04, 1.1721e-03, ..., 1.0624e-03, + 1.0033e-03, 9.4891e-04], + ..., + [-1.7975e-02, -2.9354e-03, -3.8948e-03, ..., -3.1948e-03, + -2.9869e-03, -2.7561e-03], + [ 9.0942e-03, 1.5259e-03, 2.0161e-03, ..., 1.8597e-03, + 1.7653e-03, 1.6737e-03], + [ 1.9431e-04, 3.4064e-05, 4.4614e-05, ..., 5.5224e-05, + 5.4121e-05, 5.3406e-05]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0478, -0.0049, -0.0309, -0.0930, 0.0918, 0.0527, 0.0392], + device='cuda:0'), grad: tensor([ 0.0107, 0.0084, 0.0242, -0.0032, -0.0823, 0.0413, 0.0009], + device='cuda:0') +249 +0.00224551509273949 +changing lr +epoch 48, time 330.06, cls_loss 0.4012 cls_loss_mapping 0.0186 cls_loss_causal 0.3383 re_mapping 0.0193 re_causal 0.0201 /// teacc 93.57 lr 0.00206107 +Epoch 50, weight, value: tensor([[ 2.0446e-02, 4.0472e-02, 1.2142e-03, ..., 1.8591e-03, + 1.7368e-03, 2.4057e-02], + [-6.7066e-02, -7.0794e-02, -9.3470e-02, ..., 2.2911e-02, + 6.4539e-02, 5.5077e-03], + [ 6.3006e-04, -2.0119e-02, -1.5844e-02, ..., 2.2897e-02, + -2.6317e-02, -3.5647e-02], + ..., + [-8.5042e-02, -8.5942e-02, -4.4809e-02, ..., 6.5366e-03, + -9.5915e-03, -1.1390e-05], + [-5.0727e-02, -3.3386e-02, -6.4097e-02, ..., -9.9113e-02, + -1.3153e-01, -1.2037e-01], + [ 1.9127e-01, 2.1286e-01, 2.4207e-01, ..., -1.0039e-01, + -4.4504e-02, -5.1503e-02]], device='cuda:0'), grad: tensor([[ 6.0730e-03, 1.2159e-03, 1.1368e-03, ..., 9.3985e-04, + 1.0500e-03, 1.3018e-03], + [-2.1698e-02, -3.1109e-03, -2.0142e-03, ..., -1.3132e-03, + -2.5463e-03, -2.7905e-03], + [-2.6245e-02, -4.0779e-03, -6.5346e-03, ..., -9.6359e-03, + -8.8882e-03, -8.6517e-03], + ..., + [ 2.4704e-02, 3.5038e-03, 3.9787e-03, ..., 5.1460e-03, + 5.5923e-03, 5.4703e-03], + [ 1.1536e-02, 1.7271e-03, 2.3155e-03, ..., 3.1910e-03, + 3.1624e-03, 3.1013e-03], + [ 8.8024e-04, -4.0084e-05, 2.1085e-05, ..., 1.3852e-04, + 1.4365e-04, 1.4412e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0437, -0.0013, -0.0297, -0.0772, 0.0900, 0.0326, 0.0362], + device='cuda:0'), grad: tensor([ 0.0250, -0.0861, -0.1057, 0.0184, 0.0978, 0.0461, 0.0045], + device='cuda:0') +249 +0.002061073738537637 +changing lr +epoch 49, time 330.05, cls_loss 0.4108 cls_loss_mapping 0.0204 cls_loss_causal 0.3672 re_mapping 0.0191 re_causal 0.0203 /// teacc 97.08 lr 0.00188255 +Epoch 51, weight, value: tensor([[ 1.7320e-02, 3.9216e-02, 6.2498e-07, ..., 1.9222e-03, + 1.7273e-03, 2.4075e-02], + [-6.7380e-02, -7.0292e-02, -9.3030e-02, ..., 2.2198e-02, + 6.3751e-02, 5.0907e-03], + [-1.6921e-04, -2.0815e-02, -1.6985e-02, ..., 2.2693e-02, + -2.6474e-02, -3.5978e-02], + ..., + [-8.1570e-02, -8.4228e-02, -4.3248e-02, ..., 7.7393e-03, + -8.4003e-03, 1.0179e-03], + [-5.2555e-02, -3.4182e-02, -6.4818e-02, ..., -9.9578e-02, + -1.3199e-01, -1.2057e-01], + [ 1.9409e-01, 2.1346e-01, 2.4298e-01, ..., -9.9893e-02, + -4.3940e-02, -5.1111e-02]], device='cuda:0'), grad: tensor([[-0.0468, -0.0091, -0.0094, ..., -0.0149, -0.0140, -0.0150], + [ 0.0134, 0.0031, 0.0034, ..., 0.0035, 0.0035, 0.0039], + [-0.0219, -0.0061, -0.0081, ..., -0.0014, -0.0028, -0.0044], + ..., + [ 0.0136, 0.0032, 0.0036, ..., 0.0036, 0.0036, 0.0040], + [ 0.0090, 0.0022, 0.0025, ..., 0.0015, 0.0016, 0.0021], + [ 0.0209, 0.0034, 0.0044, ..., 0.0052, 0.0056, 0.0064]], + device='cuda:0') +Epoch 51, bias, value: tensor([-0.0596, -0.0054, -0.0307, -0.0723, 0.0958, 0.0308, 0.0483], + device='cuda:0'), grad: tensor([-0.1361, 0.0511, -0.1448, 0.0470, 0.0534, 0.0397, 0.0896], + device='cuda:0') +249 +0.0018825509907063344 +changing lr +epoch 50, time 331.20, cls_loss 0.4085 cls_loss_mapping 0.0208 cls_loss_causal 0.3579 re_mapping 0.0189 re_causal 0.0201 /// teacc 97.08 lr 0.00171031 +Epoch 52, weight, value: tensor([[ 1.6594e-02, 3.8365e-02, -9.4104e-04, ..., 4.6807e-04, + 5.0040e-05, 2.2813e-02], + [-6.7123e-02, -7.0090e-02, -9.2892e-02, ..., 2.2839e-02, + 6.4205e-02, 5.7850e-03], + [-9.1552e-04, -2.0781e-02, -1.6831e-02, ..., 2.2784e-02, + -2.6147e-02, -3.5580e-02], + ..., + [-8.1133e-02, -8.4615e-02, -4.3569e-02, ..., 7.6624e-03, + -8.3209e-03, 5.4378e-04], + [-5.0682e-02, -3.3415e-02, -6.3757e-02, ..., -9.9805e-02, + -1.3180e-01, -1.2028e-01], + [ 1.9335e-01, 2.1379e-01, 2.4317e-01, ..., -9.9732e-02, + -4.3972e-02, -5.1258e-02]], device='cuda:0'), grad: tensor([[-0.0623, -0.0100, -0.0099, ..., -0.0070, -0.0105, -0.0144], + [ 0.0210, 0.0043, 0.0046, ..., 0.0054, 0.0062, 0.0073], + [ 0.0055, 0.0012, 0.0015, ..., 0.0006, 0.0008, 0.0012], + ..., + [ 0.0811, 0.0198, 0.0192, ..., 0.0397, 0.0405, 0.0426], + [ 0.0169, 0.0037, 0.0047, ..., 0.0022, 0.0030, 0.0042], + [-0.0541, -0.0158, -0.0138, ..., -0.0434, -0.0418, -0.0414]], + device='cuda:0') +Epoch 52, bias, value: tensor([-0.0616, -0.0062, -0.0324, -0.0720, 0.0968, 0.0415, 0.0409], + device='cuda:0'), grad: tensor([-0.2673, 0.0804, 0.0253, -0.0621, 0.2345, 0.0765, -0.0873], + device='cuda:0') +249 +0.0017103063703014388 +changing lr +epoch 51, time 332.18, cls_loss 0.3786 cls_loss_mapping 0.0164 cls_loss_causal 0.3241 re_mapping 0.0190 re_causal 0.0202 /// teacc 98.25 lr 0.00154469 +Epoch 53, weight, value: tensor([[ 0.0161, 0.0378, -0.0018, ..., 0.0009, 0.0011, 0.0239], + [-0.0643, -0.0686, -0.0911, ..., 0.0227, 0.0638, 0.0054], + [ 0.0010, -0.0205, -0.0164, ..., 0.0215, -0.0275, -0.0367], + ..., + [-0.0835, -0.0851, -0.0440, ..., 0.0069, -0.0092, -0.0004], + [-0.0520, -0.0340, -0.0643, ..., -0.0991, -0.1311, -0.1197], + [ 0.1919, 0.2130, 0.2424, ..., -0.0994, -0.0437, -0.0507]], + device='cuda:0'), grad: tensor([[ 0.0029, 0.0031, 0.0009, ..., -0.0008, 0.0015, 0.0004], + [-0.0264, -0.0075, -0.0098, ..., -0.0020, -0.0038, -0.0039], + [ 0.0136, 0.0032, 0.0042, ..., 0.0049, 0.0044, 0.0043], + ..., + [-0.0147, -0.0058, -0.0035, ..., -0.0055, -0.0070, -0.0054], + [ 0.0054, 0.0016, 0.0019, ..., 0.0002, 0.0006, 0.0005], + [ 0.0085, 0.0024, 0.0030, ..., 0.0006, 0.0012, 0.0011]], + device='cuda:0') +Epoch 53, bias, value: tensor([-6.3325e-02, 8.9842e-05, -1.9721e-02, -7.2095e-02, 8.4363e-02, + 3.6477e-02, 4.1203e-02], device='cuda:0'), grad: tensor([ 0.0168, -0.1346, 0.0454, 0.0424, -0.0390, 0.0275, 0.0414], + device='cuda:0') +249 +0.0015446867550656784 +changing lr +epoch 52, time 329.41, cls_loss 0.3857 cls_loss_mapping 0.0160 cls_loss_causal 0.3440 re_mapping 0.0185 re_causal 0.0199 /// teacc 92.98 lr 0.00138603 +Epoch 54, weight, value: tensor([[ 0.0178, 0.0388, -0.0011, ..., 0.0013, 0.0015, 0.0242], + [-0.0657, -0.0688, -0.0913, ..., 0.0215, 0.0624, 0.0042], + [ 0.0008, -0.0209, -0.0167, ..., 0.0207, -0.0282, -0.0375], + ..., + [-0.0844, -0.0850, -0.0442, ..., 0.0077, -0.0087, 0.0003], + [-0.0516, -0.0337, -0.0640, ..., -0.0985, -0.1303, -0.1193], + [ 0.1927, 0.2125, 0.2420, ..., -0.0995, -0.0438, -0.0505]], + device='cuda:0'), grad: tensor([[ 2.5902e-03, 4.3750e-04, 3.8147e-04, ..., 6.9046e-04, + 8.2159e-04, 9.0170e-04], + [ 6.4278e-04, 1.3268e-04, 9.0897e-05, ..., 1.6940e-04, + 1.5903e-04, 2.2840e-04], + [ 8.8730e-03, 1.6870e-03, 1.1358e-03, ..., 2.0676e-03, + 1.8272e-03, 2.6093e-03], + ..., + [-1.4648e-02, -2.7256e-03, -1.9217e-03, ..., -3.5095e-03, + -3.3245e-03, -4.4708e-03], + [ 1.5039e-03, 2.9707e-04, 2.1136e-04, ..., 3.5834e-04, + 3.2043e-04, 4.3845e-04], + [ 3.5554e-05, -6.6161e-05, -8.5950e-05, ..., -5.1498e-05, + -5.7280e-05, -3.1382e-05]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0653, -0.0009, -0.0222, -0.0690, 0.0776, 0.0348, 0.0520], + device='cuda:0'), grad: tensor([ 0.0105, 0.0027, 0.0405, 0.0042, -0.0654, 0.0068, 0.0007], + device='cuda:0') +249 +0.001386025680863044 +changing lr +epoch 53, time 330.86, cls_loss 0.3514 cls_loss_mapping 0.0183 cls_loss_causal 0.3007 re_mapping 0.0182 re_causal 0.0194 /// teacc 94.74 lr 0.00123464 +Epoch 55, weight, value: tensor([[ 0.0183, 0.0384, -0.0015, ..., 0.0009, 0.0010, 0.0238], + [-0.0660, -0.0690, -0.0910, ..., 0.0212, 0.0622, 0.0039], + [ 0.0013, -0.0212, -0.0171, ..., 0.0212, -0.0273, -0.0367], + ..., + [-0.0838, -0.0847, -0.0441, ..., 0.0079, -0.0085, 0.0006], + [-0.0518, -0.0337, -0.0639, ..., -0.0987, -0.1306, -0.1196], + [ 0.1940, 0.2136, 0.2431, ..., -0.0987, -0.0431, -0.0496]], + device='cuda:0'), grad: tensor([[-0.0344, -0.0039, -0.0057, ..., -0.0044, -0.0057, -0.0051], + [ 0.0092, 0.0016, 0.0022, ..., 0.0008, 0.0013, 0.0016], + [ 0.0147, 0.0026, 0.0034, ..., 0.0017, 0.0024, 0.0029], + ..., + [ 0.0098, 0.0017, 0.0022, ..., 0.0014, 0.0018, 0.0021], + [ 0.0111, 0.0020, 0.0026, ..., 0.0013, 0.0018, 0.0023], + [ 0.0208, 0.0039, 0.0050, ..., 0.0023, 0.0033, 0.0043]], + device='cuda:0') +Epoch 55, bias, value: tensor([-0.0610, 0.0027, -0.0191, -0.0790, 0.0788, 0.0319, 0.0527], + device='cuda:0'), grad: tensor([-0.1146, 0.0354, 0.0564, -0.1403, 0.0373, 0.0436, 0.0823], + device='cuda:0') +249 +0.0012346426699819469 +changing lr +epoch 54, time 331.05, cls_loss 0.3828 cls_loss_mapping 0.0129 cls_loss_causal 0.3438 re_mapping 0.0180 re_causal 0.0191 /// teacc 98.83 lr 0.00109084 +Epoch 56, weight, value: tensor([[ 0.0160, 0.0379, -0.0021, ..., 0.0004, 0.0005, 0.0234], + [-0.0657, -0.0690, -0.0911, ..., 0.0213, 0.0622, 0.0040], + [ 0.0015, -0.0211, -0.0169, ..., 0.0209, -0.0274, -0.0368], + ..., + [-0.0807, -0.0841, -0.0436, ..., 0.0085, -0.0079, 0.0012], + [-0.0518, -0.0332, -0.0635, ..., -0.0981, -0.1300, -0.1190], + [ 0.1926, 0.2129, 0.2426, ..., -0.0986, -0.0432, -0.0499]], + device='cuda:0'), grad: tensor([[ 0.0143, 0.0027, 0.0028, ..., 0.0048, 0.0043, 0.0044], + [-0.0716, -0.0192, -0.0203, ..., -0.0194, -0.0167, -0.0164], + [ 0.0094, 0.0017, 0.0018, ..., 0.0032, 0.0029, 0.0030], + ..., + [ 0.0139, 0.0026, 0.0026, ..., 0.0047, 0.0043, 0.0043], + [ 0.0093, 0.0034, 0.0037, ..., 0.0018, 0.0014, 0.0012], + [ 0.0234, 0.0084, 0.0091, ..., 0.0046, 0.0035, 0.0032]], + device='cuda:0') +Epoch 56, bias, value: tensor([-0.0685, 0.0034, -0.0140, -0.0769, 0.0874, 0.0293, 0.0463], + device='cuda:0'), grad: tensor([ 0.0470, -0.2128, 0.0312, 0.0043, 0.0464, 0.0238, 0.0600], + device='cuda:0') +249 +0.0010908425876598518 +changing lr +epoch 55, time 323.62, cls_loss 0.3661 cls_loss_mapping 0.0160 cls_loss_causal 0.3215 re_mapping 0.0177 re_causal 0.0189 /// teacc 98.25 lr 0.00095492 +Epoch 57, weight, value: tensor([[ 1.5005e-02, 3.7605e-02, -2.3149e-03, ..., -2.7426e-07, + 2.0216e-04, 2.3002e-02], + [-6.4370e-02, -6.8672e-02, -9.1084e-02, ..., 2.1848e-02, + 6.2258e-02, 4.4904e-03], + [ 1.1558e-03, -2.0930e-02, -1.6830e-02, ..., 2.1629e-02, + -2.6595e-02, -3.6126e-02], + ..., + [-8.3195e-02, -8.4801e-02, -4.4234e-02, ..., 7.2385e-03, + -8.7855e-03, 1.1951e-04], + [-5.1073e-02, -3.3311e-02, -6.3440e-02, ..., -9.8062e-02, + -1.2996e-01, -1.1895e-01], + [ 1.9303e-01, 2.1301e-01, 2.4275e-01, ..., -9.8213e-02, + -4.2823e-02, -4.9595e-02]], device='cuda:0'), grad: tensor([[-4.6349e-03, -1.4801e-03, -2.1553e-03, ..., -1.2388e-03, + -9.0551e-04, -1.3304e-03], + [ 3.6755e-03, 1.0653e-03, 1.4753e-03, ..., 9.8324e-04, + 7.9060e-04, 1.0071e-03], + [ 1.1816e-03, 3.3689e-04, 4.6349e-04, ..., 3.2043e-04, + 2.6178e-04, 3.2544e-04], + ..., + [ 1.2760e-03, 3.2973e-04, 4.2701e-04, ..., 3.4094e-04, + 2.9993e-04, 3.3116e-04], + [-1.9550e-03, -3.6120e-04, -3.4738e-04, ..., -5.1260e-04, + -5.4312e-04, -4.3273e-04], + [ 1.8120e-04, 4.5985e-05, 5.9009e-05, ..., 4.9561e-05, + 4.4197e-05, 4.8012e-05]], device='cuda:0') +Epoch 57, bias, value: tensor([-0.0722, 0.0066, -0.0188, -0.0720, 0.0825, 0.0322, 0.0486], + device='cuda:0'), grad: tensor([-0.0132, 0.0104, 0.0033, 0.0007, 0.0036, -0.0054, 0.0005], + device='cuda:0') +249 +0.000954915028125264 +changing lr +epoch 56, time 331.48, cls_loss 0.3338 cls_loss_mapping 0.0158 cls_loss_causal 0.2944 re_mapping 0.0172 re_causal 0.0179 /// teacc 97.66 lr 0.00082713 +Epoch 58, weight, value: tensor([[ 1.4829e-02, 3.7431e-02, -2.3595e-03, ..., -2.7541e-04, + 2.0141e-04, 2.2959e-02], + [-6.4693e-02, -6.8549e-02, -9.0969e-02, ..., 2.0967e-02, + 6.1273e-02, 3.5630e-03], + [ 6.0739e-04, -2.1134e-02, -1.7188e-02, ..., 2.2493e-02, + -2.5617e-02, -3.5140e-02], + ..., + [-8.0295e-02, -8.3671e-02, -4.3064e-02, ..., 7.5777e-03, + -8.6172e-03, 4.3596e-04], + [-5.1215e-02, -3.3439e-02, -6.3594e-02, ..., -9.7778e-02, + -1.2974e-01, -1.1868e-01], + [ 1.9182e-01, 2.1246e-01, 2.4225e-01, ..., -9.8180e-02, + -4.2844e-02, -4.9746e-02]], device='cuda:0'), grad: tensor([[ 2.6302e-03, 3.8552e-04, 4.5037e-04, ..., 8.6260e-04, + 4.9591e-04, 8.1873e-04], + [ 2.3556e-03, 4.9305e-04, 5.5075e-04, ..., 1.3342e-03, + 1.1988e-03, 1.2360e-03], + [ 8.0538e-04, 7.5459e-05, 8.9288e-05, ..., 2.7370e-04, + 1.3626e-04, 2.5654e-04], + ..., + [-7.7400e-03, -7.4768e-04, -9.3222e-04, ..., -1.2827e-03, + 2.6774e-04, -1.2789e-03], + [ 4.6539e-03, 7.1239e-04, 8.2493e-04, ..., 1.6880e-03, + 1.0872e-03, 1.5926e-03], + [ 7.3767e-04, 1.5116e-04, 1.6904e-04, ..., 4.0865e-04, + 3.6311e-04, 3.7909e-04]], device='cuda:0') +Epoch 58, bias, value: tensor([-0.0706, 0.0075, -0.0220, -0.0744, 0.0909, 0.0314, 0.0442], + device='cuda:0'), grad: tensor([ 0.0093, 0.0078, 0.0029, -0.0101, -0.0286, 0.0163, 0.0024], + device='cuda:0') +249 +0.0008271337313934874 +changing lr +epoch 57, time 329.85, cls_loss 0.3746 cls_loss_mapping 0.0148 cls_loss_causal 0.3414 re_mapping 0.0171 re_causal 0.0178 /// teacc 98.83 lr 0.00070776 +Epoch 59, weight, value: tensor([[ 1.5693e-02, 3.7125e-02, -2.5885e-03, ..., -5.5509e-04, + -2.1890e-04, 2.2631e-02], + [-6.3554e-02, -6.8061e-02, -9.0497e-02, ..., 2.1130e-02, + 6.1466e-02, 3.8598e-03], + [ 2.8481e-04, -2.1117e-02, -1.7230e-02, ..., 2.2447e-02, + -2.5683e-02, -3.5191e-02], + ..., + [-8.1024e-02, -8.3837e-02, -4.3133e-02, ..., 7.2621e-03, + -8.7246e-03, 2.5914e-04], + [-5.1622e-02, -3.3392e-02, -6.3493e-02, ..., -9.7656e-02, + -1.2955e-01, -1.1856e-01], + [ 1.9065e-01, 2.1233e-01, 2.4202e-01, ..., -9.8071e-02, + -4.2879e-02, -4.9871e-02]], device='cuda:0'), grad: tensor([[-4.0680e-02, -3.4428e-03, -5.0812e-03, ..., -4.4632e-03, + -3.8223e-03, -3.9101e-03], + [ 2.0187e-02, 1.7681e-03, 2.5806e-03, ..., 2.2449e-03, + 1.9264e-03, 1.9703e-03], + [ 1.2360e-02, 1.0509e-03, 1.5488e-03, ..., 1.3590e-03, + 1.1635e-03, 1.1902e-03], + ..., + [ 3.7937e-03, 3.2425e-04, 4.7684e-04, ..., 4.1795e-04, + 3.5810e-04, 3.6621e-04], + [ 3.7308e-03, 3.6693e-04, 5.1689e-04, ..., 4.2915e-04, + 3.6931e-04, 3.7837e-04], + [-1.5664e-04, -1.8132e-04, -1.8728e-04, ..., -8.1003e-05, + -7.5459e-05, -7.9095e-05]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0684, 0.0132, -0.0251, -0.0723, 0.0904, 0.0310, 0.0381], + device='cuda:0'), grad: tensor([-1.5039e-01, 7.4341e-02, 4.5654e-02, 2.5940e-03, 1.4000e-02, + 1.3565e-02, 1.3137e-04], device='cuda:0') +249 +0.00070775603199067 +changing lr +epoch 58, time 326.82, cls_loss 0.4041 cls_loss_mapping 0.0154 cls_loss_causal 0.3807 re_mapping 0.0166 re_causal 0.0179 /// teacc 97.66 lr 0.00059702 +Epoch 60, weight, value: tensor([[ 1.6642e-02, 3.7745e-02, -1.9779e-03, ..., 3.0557e-05, + 3.0679e-04, 2.3253e-02], + [-6.2860e-02, -6.7697e-02, -9.0057e-02, ..., 2.1445e-02, + 6.1701e-02, 4.3185e-03], + [ 4.1718e-04, -2.1119e-02, -1.7099e-02, ..., 2.2253e-02, + -2.5673e-02, -3.5164e-02], + ..., + [-8.2178e-02, -8.4207e-02, -4.3647e-02, ..., 6.9399e-03, + -9.0221e-03, -1.9886e-04], + [-5.1376e-02, -3.3487e-02, -6.3582e-02, ..., -9.7748e-02, + -1.2963e-01, -1.1882e-01], + [ 1.9097e-01, 2.1212e-01, 2.4184e-01, ..., -9.7635e-02, + -4.2626e-02, -4.9554e-02]], device='cuda:0'), grad: tensor([[0.0049, 0.0014, 0.0016, ..., 0.0007, 0.0006, 0.0009], + [0.0021, 0.0010, 0.0012, ..., 0.0002, 0.0002, 0.0005], + [0.0033, 0.0010, 0.0011, ..., 0.0005, 0.0004, 0.0006], + ..., + [0.0067, 0.0016, 0.0018, ..., 0.0010, 0.0009, 0.0012], + [0.0056, 0.0016, 0.0019, ..., 0.0008, 0.0007, 0.0010], + [0.0084, 0.0024, 0.0028, ..., 0.0012, 0.0010, 0.0015]], + device='cuda:0') +Epoch 60, bias, value: tensor([-0.0656, 0.0152, -0.0203, -0.0775, 0.0834, 0.0307, 0.0411], + device='cuda:0'), grad: tensor([ 0.0234, 0.0136, 0.0158, -0.1503, 0.0294, 0.0274, 0.0406], + device='cuda:0') +249 +0.0005970223407163104 +changing lr +epoch 59, time 329.57, cls_loss 0.3804 cls_loss_mapping 0.0173 cls_loss_causal 0.3254 re_mapping 0.0165 re_causal 0.0176 /// teacc 95.32 lr 0.00049516 +Epoch 61, weight, value: tensor([[ 1.5601e-02, 3.7483e-02, -2.2291e-03, ..., -4.6495e-04, + -2.6343e-04, 2.2696e-02], + [-6.4556e-02, -6.8083e-02, -9.0482e-02, ..., 2.1067e-02, + 6.1298e-02, 3.9165e-03], + [ 1.5231e-03, -2.0817e-02, -1.6741e-02, ..., 2.2012e-02, + -2.5944e-02, -3.5229e-02], + ..., + [-8.2038e-02, -8.4064e-02, -4.3485e-02, ..., 6.9604e-03, + -8.8938e-03, -1.1123e-04], + [-5.1393e-02, -3.3565e-02, -6.3658e-02, ..., -9.7557e-02, + -1.2941e-01, -1.1864e-01], + [ 1.9107e-01, 2.1190e-01, 2.4157e-01, ..., -9.7761e-02, + -4.2792e-02, -4.9808e-02]], device='cuda:0'), grad: tensor([[-4.9782e-04, -9.0778e-05, -6.9320e-05, ..., -1.3649e-04, + -1.5163e-04, -1.5879e-04], + [ 3.0351e-04, 5.7369e-05, 4.5091e-05, ..., 7.9036e-05, + 8.2791e-05, 8.6308e-05], + [ 1.0669e-04, 1.4327e-05, 9.5814e-06, ..., 2.2203e-05, + 1.9684e-05, 2.1145e-05], + ..., + [ 6.8307e-05, 2.4498e-05, 2.2128e-05, ..., 3.4571e-05, + 4.8399e-05, 4.9651e-05], + [ 2.5138e-05, 3.6769e-06, 2.5071e-06, ..., 4.6752e-06, + 4.3623e-06, 4.3921e-06], + [-1.4976e-05, -1.1064e-05, -1.1660e-05, ..., -6.4708e-06, + -5.9865e-06, -5.2452e-06]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0696, 0.0104, -0.0169, -0.0737, 0.0840, 0.0302, 0.0424], + device='cuda:0'), grad: tensor([-1.8530e-03, 1.0500e-03, 3.4451e-04, 2.8238e-05, 3.5954e-04, + 7.8559e-05, -8.9034e-06], device='cuda:0') +249 +0.0004951556604879052 +changing lr +epoch 60, time 331.34, cls_loss 0.3563 cls_loss_mapping 0.0130 cls_loss_causal 0.3074 re_mapping 0.0166 re_causal 0.0178 /// teacc 97.08 lr 0.00040236 +Epoch 62, weight, value: tensor([[ 1.5320e-02, 3.7093e-02, -2.5707e-03, ..., -2.8073e-04, + -6.4013e-05, 2.2859e-02], + [-6.4080e-02, -6.8139e-02, -9.0520e-02, ..., 2.1187e-02, + 6.1283e-02, 3.9405e-03], + [ 9.9581e-04, -2.0593e-02, -1.6530e-02, ..., 2.1803e-02, + -2.6072e-02, -3.5299e-02], + ..., + [-8.1551e-02, -8.3702e-02, -4.3287e-02, ..., 7.0809e-03, + -8.7540e-03, 2.4438e-06], + [-5.1627e-02, -3.3640e-02, -6.3646e-02, ..., -9.7765e-02, + -1.2953e-01, -1.1882e-01], + [ 1.9169e-01, 2.1191e-01, 2.4162e-01, ..., -9.7639e-02, + -4.2731e-02, -4.9642e-02]], device='cuda:0'), grad: tensor([[ 0.0177, 0.0029, 0.0033, ..., 0.0051, 0.0045, 0.0051], + [-0.0066, -0.0028, -0.0045, ..., 0.0027, 0.0024, 0.0018], + [-0.0643, -0.0097, -0.0098, ..., -0.0200, -0.0175, -0.0195], + ..., + [ 0.0273, 0.0044, 0.0049, ..., 0.0083, 0.0072, 0.0082], + [ 0.0079, 0.0016, 0.0018, ..., 0.0013, 0.0011, 0.0014], + [ 0.0092, 0.0019, 0.0023, ..., 0.0010, 0.0009, 0.0013]], + device='cuda:0') +Epoch 62, bias, value: tensor([-0.0700, 0.0114, -0.0195, -0.0764, 0.0855, 0.0296, 0.0464], + device='cuda:0'), grad: tensor([ 0.0704, -0.0613, -0.2361, 0.0386, 0.1059, 0.0364, 0.0459], + device='cuda:0') +249 +0.00040236113724274745 +changing lr +epoch 61, time 332.05, cls_loss 0.3396 cls_loss_mapping 0.0144 cls_loss_causal 0.3085 re_mapping 0.0164 re_causal 0.0173 /// teacc 98.25 lr 0.00031883 +Epoch 63, weight, value: tensor([[ 1.4954e-02, 3.6922e-02, -2.8140e-03, ..., -2.3422e-04, + -5.7432e-05, 2.2867e-02], + [-6.3762e-02, -6.8041e-02, -9.0369e-02, ..., 2.1279e-02, + 6.1407e-02, 4.1065e-03], + [ 8.8317e-04, -2.0730e-02, -1.6648e-02, ..., 2.1510e-02, + -2.6352e-02, -3.5580e-02], + ..., + [-8.1370e-02, -8.3663e-02, -4.3208e-02, ..., 7.1921e-03, + -8.6710e-03, 9.9769e-05], + [-5.1957e-02, -3.3730e-02, -6.3738e-02, ..., -9.7821e-02, + -1.2954e-01, -1.1883e-01], + [ 1.9213e-01, 2.1229e-01, 2.4199e-01, ..., -9.7457e-02, + -4.2612e-02, -4.9565e-02]], device='cuda:0'), grad: tensor([[-3.5797e-02, -5.6915e-03, -5.5962e-03, ..., -7.2517e-03, + -6.8703e-03, -6.6795e-03], + [ 7.2336e-04, 1.2314e-04, 1.2267e-04, ..., 1.5068e-04, + 1.4389e-04, 1.3912e-04], + [ 8.9788e-04, 1.4281e-04, 1.4067e-04, ..., 1.8299e-04, + 1.7345e-04, 1.6868e-04], + ..., + [ 3.4149e-02, 5.4245e-03, 5.3368e-03, ..., 6.9237e-03, + 6.5613e-03, 6.3782e-03], + [ 4.0859e-05, 8.4862e-06, 8.9407e-06, ..., 1.6809e-05, + 1.7121e-05, 1.6600e-05], + [-8.7023e-06, -9.3803e-06, -1.0960e-05, ..., -2.3209e-06, + -3.0641e-06, -1.6522e-06]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0718, 0.0127, -0.0193, -0.0762, 0.0862, 0.0291, 0.0462], + device='cuda:0'), grad: tensor([-1.5234e-01, 3.0193e-03, 3.8185e-03, 3.4392e-05, 1.4526e-01, + 1.4937e-04, 2.0429e-05], device='cuda:0') +249 +0.00031882564680131423 +changing lr +epoch 62, time 330.00, cls_loss 0.3464 cls_loss_mapping 0.0136 cls_loss_causal 0.3091 re_mapping 0.0164 re_causal 0.0172 /// teacc 98.25 lr 0.00024472 +Epoch 64, weight, value: tensor([[ 1.5206e-02, 3.6896e-02, -2.7770e-03, ..., -1.9652e-05, + 2.2984e-04, 2.3190e-02], + [-6.4100e-02, -6.8116e-02, -9.0477e-02, ..., 2.0861e-02, + 6.0934e-02, 3.6454e-03], + [ 7.6458e-04, -2.0777e-02, -1.6668e-02, ..., 2.1358e-02, + -2.6479e-02, -3.5712e-02], + ..., + [-8.1783e-02, -8.3904e-02, -4.3463e-02, ..., 7.2126e-03, + -8.6849e-03, 9.1893e-05], + [-5.1559e-02, -3.3591e-02, -6.3580e-02, ..., -9.7622e-02, + -1.2931e-01, -1.1860e-01], + [ 1.9201e-01, 2.1231e-01, 2.4198e-01, ..., -9.7476e-02, + -4.2681e-02, -4.9656e-02]], device='cuda:0'), grad: tensor([[-0.0392, -0.0079, -0.0093, ..., -0.0054, -0.0068, -0.0106], + [ 0.0072, 0.0015, 0.0017, ..., 0.0012, 0.0015, 0.0021], + [ 0.0056, 0.0011, 0.0013, ..., 0.0008, 0.0010, 0.0015], + ..., + [ 0.0041, 0.0009, 0.0010, ..., 0.0005, 0.0007, 0.0011], + [ 0.0041, 0.0009, 0.0011, ..., -0.0006, -0.0002, 0.0002], + [ 0.0106, 0.0021, 0.0025, ..., 0.0016, 0.0020, 0.0030]], + device='cuda:0') +Epoch 64, bias, value: tensor([-0.0705, 0.0115, -0.0199, -0.0755, 0.0861, 0.0309, 0.0443], + device='cuda:0'), grad: tensor([-0.1573, 0.0282, 0.0224, 0.0291, 0.0163, 0.0191, 0.0422], + device='cuda:0') +249 +0.0002447174185242325 +changing lr +epoch 63, time 329.55, cls_loss 0.3577 cls_loss_mapping 0.0146 cls_loss_causal 0.3315 re_mapping 0.0162 re_causal 0.0171 /// teacc 97.66 lr 0.00018019 +Epoch 65, weight, value: tensor([[ 1.5464e-02, 3.6975e-02, -2.6881e-03, ..., 2.0285e-04, + 4.9334e-04, 2.3442e-02], + [-6.4448e-02, -6.8152e-02, -9.0541e-02, ..., 2.0774e-02, + 6.0798e-02, 3.4957e-03], + [ 1.2724e-03, -2.0635e-02, -1.6506e-02, ..., 2.1437e-02, + -2.6383e-02, -3.5585e-02], + ..., + [-8.1584e-02, -8.3761e-02, -4.3345e-02, ..., 7.3348e-03, + -8.5704e-03, 2.2214e-04], + [-5.1986e-02, -3.3723e-02, -6.3708e-02, ..., -9.7758e-02, + -1.2946e-01, -1.1873e-01], + [ 1.9178e-01, 2.1214e-01, 2.4180e-01, ..., -9.7505e-02, + -4.2729e-02, -4.9686e-02]], device='cuda:0'), grad: tensor([[ 0.0439, 0.0069, 0.0067, ..., 0.0146, 0.0159, 0.0169], + [-0.0056, -0.0017, -0.0010, ..., 0.0003, 0.0004, 0.0002], + [ 0.0234, 0.0066, 0.0063, ..., 0.0034, 0.0037, 0.0051], + ..., + [-0.0472, -0.0066, -0.0065, ..., -0.0162, -0.0178, -0.0185], + [ 0.0065, 0.0022, 0.0023, ..., 0.0010, 0.0011, 0.0016], + [-0.0265, -0.0092, -0.0097, ..., -0.0038, -0.0042, -0.0068]], + device='cuda:0') +Epoch 65, bias, value: tensor([-0.0696, 0.0099, -0.0176, -0.0752, 0.0855, 0.0300, 0.0439], + device='cuda:0'), grad: tensor([ 0.1364, -0.0251, 0.0841, 0.0211, -0.1368, 0.0251, -0.1047], + device='cuda:0') +249 +0.0001801856965207339 +changing lr +epoch 64, time 329.98, cls_loss 0.3297 cls_loss_mapping 0.0141 cls_loss_causal 0.2899 re_mapping 0.0161 re_causal 0.0168 /// teacc 98.25 lr 0.00012536 +Epoch 66, weight, value: tensor([[ 1.5510e-02, 3.6970e-02, -2.6831e-03, ..., 3.1927e-04, + 5.6486e-04, 2.3529e-02], + [-6.4049e-02, -6.8082e-02, -9.0448e-02, ..., 2.0773e-02, + 6.0787e-02, 3.5026e-03], + [ 1.2444e-03, -2.0612e-02, -1.6482e-02, ..., 2.1480e-02, + -2.6332e-02, -3.5529e-02], + ..., + [-8.1536e-02, -8.3659e-02, -4.3253e-02, ..., 7.2265e-03, + -8.6300e-03, 1.4584e-04], + [-5.2100e-02, -3.3740e-02, -6.3728e-02, ..., -9.7790e-02, + -1.2948e-01, -1.1877e-01], + [ 1.9161e-01, 2.1207e-01, 2.4172e-01, ..., -9.7513e-02, + -4.2750e-02, -4.9713e-02]], device='cuda:0'), grad: tensor([[ 3.5048e-04, 7.7426e-05, 7.2837e-05, ..., 1.2398e-04, + 1.1045e-04, 1.1355e-04], + [ 3.4857e-04, 7.4923e-05, 6.8188e-05, ..., 1.1712e-04, + 9.9897e-05, 1.0514e-04], + [ 2.6965e-04, 4.6492e-05, 3.9548e-05, ..., 8.7678e-05, + 7.2956e-05, 7.8261e-05], + ..., + [ 3.9220e-04, 6.5684e-05, 5.0783e-05, ..., 1.1128e-04, + 8.0824e-05, 9.3877e-05], + [-1.7481e-03, -2.7728e-04, -2.2542e-04, ..., -5.5456e-04, + -4.5109e-04, -4.9114e-04], + [-1.3852e-04, -9.9540e-05, -1.0753e-04, ..., -5.6326e-05, + -5.5879e-05, -5.1141e-05]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0699, 0.0114, -0.0175, -0.0763, 0.0858, 0.0301, 0.0433], + device='cuda:0'), grad: tensor([ 0.0010, 0.0010, 0.0008, 0.0015, 0.0012, -0.0054, -0.0002], + device='cuda:0') +249 +0.000125360439090882 +changing lr +epoch 65, time 331.68, cls_loss 0.3528 cls_loss_mapping 0.0135 cls_loss_causal 0.3265 re_mapping 0.0159 re_causal 0.0170 /// teacc 98.25 lr 0.00008035 +Epoch 67, weight, value: tensor([[ 1.5643e-02, 3.6990e-02, -2.6502e-03, ..., 3.3949e-04, + 5.7431e-04, 2.3557e-02], + [-6.4198e-02, -6.8172e-02, -9.0535e-02, ..., 2.0703e-02, + 6.0710e-02, 3.4334e-03], + [ 1.1406e-03, -2.0663e-02, -1.6544e-02, ..., 2.1427e-02, + -2.6369e-02, -3.5590e-02], + ..., + [-8.1601e-02, -8.3584e-02, -4.3196e-02, ..., 7.3042e-03, + -8.5377e-03, 2.2703e-04], + [-5.1952e-02, -3.3717e-02, -6.3702e-02, ..., -9.7715e-02, + -1.2940e-01, -1.1869e-01], + [ 1.9175e-01, 2.1213e-01, 2.4178e-01, ..., -9.7403e-02, + -4.2665e-02, -4.9612e-02]], device='cuda:0'), grad: tensor([[ 0.0066, 0.0014, 0.0016, ..., 0.0016, 0.0019, 0.0018], + [ 0.0041, 0.0009, 0.0009, ..., 0.0008, 0.0012, 0.0010], + [ 0.0066, 0.0014, 0.0015, ..., 0.0015, 0.0019, 0.0018], + ..., + [ 0.0052, 0.0011, 0.0012, ..., 0.0009, 0.0013, 0.0012], + [-0.0372, -0.0080, -0.0086, ..., -0.0080, -0.0104, -0.0097], + [ 0.0080, 0.0017, 0.0019, ..., 0.0017, 0.0022, 0.0021]], + device='cuda:0') +Epoch 67, bias, value: tensor([-0.0692, 0.0109, -0.0178, -0.0765, 0.0852, 0.0306, 0.0438], + device='cuda:0'), grad: tensor([ 0.0256, 0.0171, 0.0263, 0.0271, 0.0211, -0.1495, 0.0322], + device='cuda:0') +249 +8.03520570068517e-05 +changing lr +epoch 66, time 330.34, cls_loss 0.3384 cls_loss_mapping 0.0129 cls_loss_causal 0.2931 re_mapping 0.0159 re_causal 0.0167 /// teacc 97.66 lr 0.00004525 +Epoch 68, weight, value: tensor([[ 0.0156, 0.0370, -0.0027, ..., 0.0003, 0.0005, 0.0235], + [-0.0641, -0.0681, -0.0905, ..., 0.0207, 0.0607, 0.0034], + [ 0.0011, -0.0207, -0.0165, ..., 0.0214, -0.0263, -0.0356], + ..., + [-0.0815, -0.0836, -0.0432, ..., 0.0073, -0.0085, 0.0003], + [-0.0520, -0.0337, -0.0637, ..., -0.0977, -0.1294, -0.1187], + [ 0.1917, 0.2121, 0.2417, ..., -0.0974, -0.0427, -0.0496]], + device='cuda:0'), grad: tensor([[ 0.0050, 0.0011, 0.0012, ..., 0.0006, 0.0008, 0.0011], + [ 0.0047, 0.0010, 0.0011, ..., 0.0005, 0.0008, 0.0010], + [-0.0364, -0.0079, -0.0087, ..., -0.0044, -0.0060, -0.0082], + ..., + [ 0.0043, 0.0009, 0.0010, ..., 0.0005, 0.0007, 0.0010], + [ 0.0075, 0.0016, 0.0018, ..., 0.0009, 0.0012, 0.0017], + [ 0.0101, 0.0022, 0.0024, ..., 0.0012, 0.0017, 0.0023]], + device='cuda:0') +Epoch 68, bias, value: tensor([-0.0695, 0.0113, -0.0178, -0.0766, 0.0859, 0.0300, 0.0436], + device='cuda:0'), grad: tensor([ 0.0188, 0.0186, -0.1398, 0.0184, 0.0165, 0.0287, 0.0387], + device='cuda:0') +249 +4.5251191160326525e-05 +changing lr +epoch 67, time 326.76, cls_loss 0.3621 cls_loss_mapping 0.0137 cls_loss_causal 0.3337 re_mapping 0.0158 re_causal 0.0167 /// teacc 98.25 lr 0.00002013 +Epoch 69, weight, value: tensor([[ 0.0155, 0.0370, -0.0027, ..., 0.0002, 0.0005, 0.0235], + [-0.0640, -0.0681, -0.0904, ..., 0.0207, 0.0607, 0.0034], + [ 0.0011, -0.0207, -0.0166, ..., 0.0214, -0.0263, -0.0356], + ..., + [-0.0815, -0.0836, -0.0432, ..., 0.0074, -0.0085, 0.0003], + [-0.0520, -0.0337, -0.0637, ..., -0.0977, -0.1294, -0.1187], + [ 0.1916, 0.2121, 0.2417, ..., -0.0975, -0.0427, -0.0497]], + device='cuda:0'), grad: tensor([[-5.5962e-03, -1.2188e-03, -1.9083e-03, ..., -1.5488e-03, + -2.3575e-03, -2.5043e-03], + [-2.8439e-03, -7.0095e-04, -6.6280e-04, ..., -5.6648e-04, + -3.6955e-04, -4.1246e-04], + [ 2.1219e-04, 4.7594e-05, 6.6340e-05, ..., 5.1439e-05, + 7.2539e-05, 7.7248e-05], + ..., + [ 5.6305e-03, 1.2331e-03, 1.8368e-03, ..., 1.5287e-03, + 2.1935e-03, 2.3365e-03], + [ 2.1744e-03, 5.4455e-04, 5.4502e-04, ..., 4.3416e-04, + 3.3426e-04, 3.6693e-04], + [ 9.1195e-05, 1.5035e-05, 2.2843e-05, ..., 1.9804e-05, + 2.7820e-05, 3.0816e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0697, 0.0114, -0.0179, -0.0767, 0.0859, 0.0302, 0.0437], + device='cuda:0'), grad: tensor([-0.0353, -0.0065, 0.0011, 0.0015, 0.0328, 0.0059, 0.0005], + device='cuda:0') +249 +2.0128530023804673e-05 +changing lr +epoch 68, time 329.02, cls_loss 0.3571 cls_loss_mapping 0.0154 cls_loss_causal 0.3297 re_mapping 0.0158 re_causal 0.0166 /// teacc 95.32 lr 0.00000503 +Epoch 70, weight, value: tensor([[ 1.5356e-02, 3.6919e-02, -2.7168e-03, ..., 2.1096e-04, + 4.6857e-04, 2.3457e-02], + [-6.3965e-02, -6.8048e-02, -9.0401e-02, ..., 2.0716e-02, + 6.0715e-02, 3.4511e-03], + [ 1.0391e-03, -2.0695e-02, -1.6568e-02, ..., 2.1415e-02, + -2.6356e-02, -3.5591e-02], + ..., + [-8.1340e-02, -8.3535e-02, -4.3149e-02, ..., 7.4071e-03, + -8.4435e-03, 3.2380e-04], + [-5.1926e-02, -3.3719e-02, -6.3704e-02, ..., -9.7643e-02, + -1.2933e-01, -1.1863e-01], + [ 1.9161e-01, 2.1204e-01, 2.4169e-01, ..., -9.7465e-02, + -4.2731e-02, -4.9671e-02]], device='cuda:0'), grad: tensor([[-0.0250, -0.0076, -0.0089, ..., -0.0044, -0.0027, -0.0015], + [ 0.0036, 0.0011, 0.0011, ..., 0.0010, 0.0010, 0.0010], + [-0.0169, -0.0047, -0.0050, ..., -0.0040, -0.0049, -0.0048], + ..., + [ 0.0385, 0.0114, 0.0129, ..., 0.0078, 0.0070, 0.0057], + [-0.0054, -0.0020, -0.0019, ..., -0.0043, -0.0042, -0.0042], + [ 0.0004, 0.0001, 0.0001, ..., 0.0003, 0.0003, 0.0003]], + device='cuda:0') +Epoch 70, bias, value: tensor([-0.0701, 0.0115, -0.0179, -0.0767, 0.0862, 0.0304, 0.0435], + device='cuda:0'), grad: tensor([-0.0362, 0.0102, -0.0623, 0.0113, 0.0884, -0.0124, 0.0011], + device='cuda:0') +249 +5.034667293427056e-06 +changing lr +epoch 69, time 331.03, cls_loss 0.3636 cls_loss_mapping 0.0142 cls_loss_causal 0.3308 re_mapping 0.0159 re_causal 0.0171 /// teacc 97.08 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps1_RA_repeat', 'source_domain': 'photo', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps1_RA_repeat/photo_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of best +randm: False +stride: 5 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +columns: ['photo', 'art_painting', 'cartoon', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + photo art_painting cartoon sketch Avg +w/o do (original x) 99.94012 67.236328 42.576792 57.241028 55.684716 + photo art_painting cartoon sketch Avg +do 99.820359 68.994141 52.047782 69.076101 63.372674 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps1_RA_repeat', 'source_domain': 'photo', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1_eps1_RA_repeat/photo_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['photo', 'art_painting', 'cartoon', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + photo art_painting cartoon sketch Avg +w/o do (original x) 99.700599 66.601562 50.554608 64.545686 60.567285 + photo art_painting cartoon sketch Avg +do 99.640719 65.478516 54.692833 67.981675 62.717674 diff --git a/Meta-causal/code-withStyleAttack/73771.error b/Meta-causal/code-withStyleAttack/73771.error new file mode 100644 index 0000000000000000000000000000000000000000..cec843779d7c967ab2b29f38358285fe08cafe5e --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73771.error @@ -0,0 +1,86 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 672, in + experiment() + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py", line 273, in experiment + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_RA, y, epsilon_list) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 187, in adversarial_attack_Incre + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/tool_func.py", line 19, in calc_mean_std + assert (len(size) == 4) + ^^^^^^^^^^^^^^ +AssertionError +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 145, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 29, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 45, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/torch/serialization.py", line 997, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/torch/serialization.py", line 444, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/torch/serialization.py", line 425, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/best_cls_net.pkl' +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 145, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 29, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code-withStyleAttack/main_test_digit_v13.py", line 48, in evaluate_digit + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/torch/serialization.py", line 997, in load + with _open_file_like(f, 'rb') as opened_file: + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/torch/serialization.py", line 444, in _open_file_like + return _open_file(name_or_buffer, mode) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-3wy7kgr4zaqz/lib/python3.11/site-packages/torch/serialization.py", line 425, in __init__ + super().__init__(open(name, mode)) + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/last_cls_net.pkl' +srun: error: gcpl4-eu-1: task 0: Exited with exit code 1 diff --git a/Meta-causal/code-withStyleAttack/73771.log b/Meta-causal/code-withStyleAttack/73771.log new file mode 100644 index 0000000000000000000000000000000000000000..62a5832c7da6e31856b7941323e1f1d2affebcad --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73771.log @@ -0,0 +1,28 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0261, 0.0158, 0.0133, ..., 0.0026, -0.0106, 0.0144], + [ 0.0192, -0.0037, 0.0040, ..., 0.0299, -0.0211, -0.0292], + [-0.0290, -0.0065, 0.0231, ..., 0.0248, 0.0046, 0.0166], + ..., + [ 0.0263, 0.0206, 0.0228, ..., -0.0254, -0.0240, 0.0079], + [ 0.0304, -0.0059, 0.0271, ..., 0.0253, 0.0035, -0.0263], + [ 0.0075, -0.0214, -0.0031, ..., -0.0144, 0.0155, -0.0123]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([ 0.0129, 0.0222, 0.0276, 0.0118, 0.0310, -0.0219, 0.0043, 0.0032, + 0.0284, -0.0059], device='cuda:0'), grad: None +100 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last diff --git a/Meta-causal/code-withStyleAttack/73772.error b/Meta-causal/code-withStyleAttack/73772.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code-withStyleAttack/73772.log b/Meta-causal/code-withStyleAttack/73772.log new file mode 100644 index 0000000000000000000000000000000000000000..d25c5da5742b398db00877d8519b854f548c4c0c --- /dev/null +++ b/Meta-causal/code-withStyleAttack/73772.log @@ -0,0 +1,13405 @@ +/home/yuqian_fu +here1 +here2 +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +Epoch 1, weight, value: tensor([[-0.0257, -0.0217, -0.0145, ..., 0.0098, -0.0079, -0.0043], + [ 0.0085, 0.0093, 0.0299, ..., -0.0185, -0.0250, 0.0172], + [ 0.0309, 0.0209, -0.0241, ..., 0.0148, -0.0250, 0.0080], + ..., + [-0.0115, -0.0008, 0.0226, ..., 0.0186, -0.0206, -0.0125], + [-0.0296, 0.0005, 0.0156, ..., 0.0194, -0.0230, 0.0121], + [ 0.0034, 0.0047, 0.0175, ..., -0.0118, -0.0244, -0.0308]], + device='cuda:0'), grad: None +Epoch 1, bias, value: tensor([-0.0295, -0.0087, -0.0069, -0.0114, -0.0144, 0.0117, -0.0009, 0.0181, + 0.0012, 0.0265], device='cuda:0'), grad: None +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 215.27, cls_loss 2.2952 cls_loss_mapping 2.2932 cls_loss_causal 2.3004 re_mapping 0.0021 re_causal 0.0022 /// teacc 40.22 lr 0.00010000 +Epoch 2, weight, value: tensor([[-0.0266, -0.0224, -0.0154, ..., 0.0093, -0.0077, -0.0068], + [ 0.0100, 0.0086, 0.0291, ..., -0.0149, -0.0229, 0.0181], + [ 0.0300, 0.0203, -0.0232, ..., 0.0159, -0.0255, 0.0077], + ..., + [-0.0105, -0.0002, 0.0203, ..., 0.0181, -0.0215, -0.0113], + [-0.0279, -0.0001, 0.0145, ..., 0.0186, -0.0240, 0.0117], + [ 0.0027, 0.0041, 0.0171, ..., -0.0145, -0.0252, -0.0304]], + device='cuda:0'), grad: tensor([[ 0.0033, 0.0000, 0.0000, ..., 0.0003, 0.0000, 0.0014], + [-0.0039, 0.0000, 0.0000, ..., 0.0001, 0.0000, 0.0010], + [-0.0009, 0.0000, 0.0000, ..., -0.0005, 0.0000, -0.0013], + ..., + [-0.0009, 0.0000, 0.0000, ..., -0.0003, 0.0000, 0.0012], + [ 0.0036, 0.0000, 0.0000, ..., 0.0003, 0.0000, -0.0022], + [-0.0013, 0.0000, 0.0000, ..., -0.0004, 0.0000, -0.0039]], + device='cuda:0') +Epoch 2, bias, value: tensor([-0.0305, -0.0062, -0.0074, -0.0108, -0.0143, 0.0082, -0.0010, 0.0198, + 0.0007, 0.0259], device='cuda:0'), grad: tensor([ 0.0401, -0.0173, -0.0286, 0.0070, 0.0047, 0.0200, -0.0013, 0.0184, + 0.0051, -0.0482], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 214.59, cls_loss 2.2133 cls_loss_mapping 1.7574 cls_loss_causal 2.1724 re_mapping 0.0493 re_causal 0.0267 /// teacc 69.41 lr 0.00010000 +Epoch 3, weight, value: tensor([[-0.0284, -0.0224, -0.0154, ..., 0.0050, -0.0077, -0.0092], + [ 0.0109, 0.0086, 0.0291, ..., -0.0116, -0.0229, 0.0158], + [ 0.0297, 0.0203, -0.0232, ..., 0.0168, -0.0255, 0.0073], + ..., + [-0.0108, -0.0002, 0.0203, ..., 0.0154, -0.0215, -0.0101], + [-0.0283, -0.0001, 0.0145, ..., 0.0180, -0.0240, 0.0093], + [ 0.0004, 0.0041, 0.0171, ..., -0.0178, -0.0252, -0.0297]], + device='cuda:0'), grad: tensor([[ 0.0055, 0.0000, 0.0000, ..., 0.0079, 0.0000, 0.0097], + [-0.0021, 0.0000, 0.0000, ..., -0.0054, 0.0000, -0.0017], + [ 0.0049, 0.0000, 0.0000, ..., -0.0001, 0.0000, 0.0074], + ..., + [ 0.0016, 0.0000, 0.0000, ..., 0.0072, 0.0000, 0.0032], + [ 0.0063, 0.0000, 0.0000, ..., -0.0011, 0.0000, 0.0048], + [-0.0132, 0.0000, 0.0000, ..., -0.0145, 0.0000, -0.0124]], + device='cuda:0') +Epoch 3, bias, value: tensor([-0.0310, -0.0046, -0.0087, -0.0107, -0.0146, 0.0074, 0.0005, 0.0195, + -0.0003, 0.0264], device='cuda:0'), grad: tensor([ 0.0634, 0.0126, 0.0408, -0.0457, 0.0363, -0.0687, 0.0420, 0.0433, + -0.0366, -0.0872], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 214.49, cls_loss 2.0050 cls_loss_mapping 0.8717 cls_loss_causal 1.8206 re_mapping 0.1308 re_causal 0.1030 /// teacc 87.88 lr 0.00010000 +Epoch 4, weight, value: tensor([[-0.0312, -0.0310, -0.0154, ..., 0.0030, -0.0154, -0.0132], + [ 0.0102, -0.0029, 0.0291, ..., -0.0088, -0.0178, 0.0121], + [ 0.0287, 0.0067, -0.0232, ..., 0.0216, -0.0246, 0.0064], + ..., + [-0.0099, 0.0063, 0.0203, ..., 0.0103, -0.0188, -0.0105], + [-0.0260, -0.0076, 0.0145, ..., 0.0187, -0.0227, 0.0103], + [-0.0017, 0.0059, 0.0171, ..., -0.0242, -0.0287, -0.0286]], + device='cuda:0'), grad: tensor([[-0.0024, 0.0014, 0.0000, ..., -0.0173, 0.0029, 0.0001], + [-0.0019, 0.0074, 0.0000, ..., 0.0002, 0.0033, -0.0002], + [-0.0035, -0.0108, 0.0000, ..., 0.0044, -0.0054, -0.0046], + ..., + [ 0.0042, 0.0070, 0.0000, ..., 0.0040, 0.0027, 0.0044], + [ 0.0036, 0.0012, 0.0000, ..., 0.0017, 0.0013, 0.0032], + [ 0.0050, 0.0070, 0.0000, ..., 0.0027, 0.0031, 0.0038]], + device='cuda:0') +Epoch 4, bias, value: tensor([-0.0309, -0.0045, -0.0095, -0.0100, -0.0147, 0.0082, -0.0002, 0.0192, + 0.0002, 0.0262], device='cuda:0'), grad: tensor([-0.0816, 0.0069, -0.0895, 0.0048, -0.0013, 0.0634, -0.0400, 0.0790, + -0.0061, 0.0643], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 214.41, cls_loss 1.7866 cls_loss_mapping 0.5233 cls_loss_causal 1.5828 re_mapping 0.1264 re_causal 0.1410 /// teacc 91.49 lr 0.00010000 +Epoch 5, weight, value: tensor([[-3.4419e-02, -3.5779e-02, -1.5380e-02, ..., 2.4680e-03, + -2.9849e-02, -1.5549e-02], + [ 9.3580e-03, -9.9548e-03, 2.9077e-02, ..., -6.0945e-03, + -7.3603e-03, 1.0554e-02], + [ 2.6511e-02, -1.7537e-03, -2.3167e-02, ..., 2.3826e-02, + -2.5357e-02, 4.3792e-03], + ..., + [-1.0976e-02, 1.2861e-02, 2.0267e-02, ..., 1.0183e-02, + -2.3878e-02, -1.2458e-02], + [-2.5294e-02, -7.2615e-03, 1.4495e-02, ..., 2.2081e-02, + -1.6458e-02, 7.7974e-03], + [-1.3568e-05, 1.2129e-02, 1.7080e-02, ..., -2.8597e-02, + -3.2479e-02, -2.7199e-02]], device='cuda:0'), grad: tensor([[ 2.5082e-03, 8.8959e-03, 0.0000e+00, ..., -1.3561e-03, + 1.5177e-05, 1.3523e-03], + [ 4.1580e-04, -1.1429e-02, 0.0000e+00, ..., -1.6031e-03, + -2.8968e-05, 3.3569e-03], + [ 2.4586e-03, -7.6370e-03, 0.0000e+00, ..., -5.3368e-03, + -9.7370e-04, 5.2147e-03], + ..., + [ 5.3368e-03, 1.2512e-02, 0.0000e+00, ..., 5.9986e-04, + 7.0810e-04, -1.4362e-03], + [ 2.2339e-02, 1.7822e-02, 0.0000e+00, ..., 8.4381e-03, + 7.0524e-04, 1.5114e-02], + [-1.7975e-02, -2.5421e-02, 0.0000e+00, ..., 2.1973e-03, + 1.4448e-03, -1.4191e-02]], device='cuda:0') +Epoch 5, bias, value: tensor([-0.0301, -0.0041, -0.0095, -0.0093, -0.0138, 0.0081, -0.0017, 0.0185, + 0.0003, 0.0256], device='cuda:0'), grad: tensor([ 0.0030, -0.0207, -0.0003, 0.0251, -0.0262, 0.0673, -0.0639, 0.0069, + 0.1094, -0.1006], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 214.48, cls_loss 1.6165 cls_loss_mapping 0.3603 cls_loss_causal 1.3744 re_mapping 0.1120 re_causal 0.1466 /// teacc 91.67 lr 0.00010000 +Epoch 6, weight, value: tensor([[-0.0359, -0.0376, -0.0153, ..., 0.0025, -0.0343, -0.0175], + [ 0.0078, -0.0134, 0.0272, ..., -0.0037, -0.0035, 0.0078], + [ 0.0241, -0.0050, -0.0236, ..., 0.0251, -0.0211, 0.0021], + ..., + [-0.0109, 0.0158, 0.0186, ..., 0.0117, -0.0250, -0.0138], + [-0.0257, -0.0081, 0.0148, ..., 0.0243, -0.0145, 0.0051], + [ 0.0017, 0.0174, 0.0165, ..., -0.0326, -0.0343, -0.0257]], + device='cuda:0'), grad: tensor([[ 0.0118, 0.0029, 0.0000, ..., 0.0068, 0.0026, 0.0123], + [ 0.0054, -0.0118, 0.0000, ..., -0.0050, -0.0099, 0.0013], + [ 0.0062, -0.0006, 0.0000, ..., -0.0018, 0.0020, 0.0072], + ..., + [ 0.0040, 0.0102, 0.0000, ..., 0.0147, 0.0062, 0.0075], + [-0.0035, 0.0043, 0.0000, ..., -0.0032, 0.0029, 0.0081], + [-0.0033, -0.0107, 0.0000, ..., -0.0051, 0.0011, -0.0101]], + device='cuda:0') +Epoch 6, bias, value: tensor([-0.0291, -0.0042, -0.0085, -0.0095, -0.0132, 0.0091, -0.0029, 0.0191, + -0.0016, 0.0250], device='cuda:0'), grad: tensor([ 0.0630, 0.0033, -0.0320, -0.0046, 0.0212, -0.0768, -0.0086, 0.0711, + 0.0387, -0.0752], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 214.68, cls_loss 1.5231 cls_loss_mapping 0.2814 cls_loss_causal 1.2788 re_mapping 0.0952 re_causal 0.1394 /// teacc 92.94 lr 0.00010000 +Epoch 7, weight, value: tensor([[-0.0376, -0.0381, -0.0163, ..., 0.0033, -0.0381, -0.0196], + [ 0.0070, -0.0148, 0.0277, ..., -0.0013, -0.0025, 0.0050], + [ 0.0214, -0.0093, -0.0222, ..., 0.0274, -0.0170, 0.0003], + ..., + [-0.0107, 0.0175, 0.0132, ..., 0.0112, -0.0288, -0.0146], + [-0.0252, -0.0081, 0.0109, ..., 0.0264, -0.0110, 0.0033], + [ 0.0019, 0.0196, 0.0113, ..., -0.0359, -0.0366, -0.0252]], + device='cuda:0'), grad: tensor([[ 0.0018, 0.0028, 0.0000, ..., -0.0023, 0.0051, -0.0061], + [-0.0097, 0.0020, 0.0000, ..., -0.0202, 0.0030, -0.0169], + [ 0.0077, 0.0024, 0.0000, ..., 0.0001, -0.0076, 0.0049], + ..., + [-0.0036, 0.0121, 0.0000, ..., 0.0060, 0.0131, -0.0103], + [-0.0048, -0.0305, 0.0000, ..., -0.0168, -0.0431, -0.0006], + [ 0.0026, 0.0505, 0.0000, ..., 0.0087, 0.0238, 0.0083]], + device='cuda:0') +Epoch 7, bias, value: tensor([-0.0292, -0.0045, -0.0085, -0.0084, -0.0129, 0.0083, -0.0027, 0.0193, + -0.0021, 0.0246], device='cuda:0'), grad: tensor([ 0.0168, -0.1144, -0.0107, 0.0742, -0.0011, -0.0155, 0.0515, -0.0100, + -0.0684, 0.0776], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 214.51, cls_loss 1.4263 cls_loss_mapping 0.2359 cls_loss_causal 1.1985 re_mapping 0.0810 re_causal 0.1366 /// teacc 94.32 lr 0.00010000 +Epoch 8, weight, value: tensor([[-0.0399, -0.0381, -0.0166, ..., 0.0039, -0.0389, -0.0206], + [ 0.0068, -0.0187, 0.0274, ..., 0.0003, -0.0006, 0.0033], + [ 0.0198, -0.0117, -0.0223, ..., 0.0283, -0.0150, -0.0025], + ..., + [-0.0112, 0.0188, 0.0125, ..., 0.0114, -0.0294, -0.0167], + [-0.0251, -0.0076, 0.0106, ..., 0.0268, -0.0094, 0.0023], + [ 0.0021, 0.0215, 0.0106, ..., -0.0376, -0.0377, -0.0236]], + device='cuda:0'), grad: tensor([[-0.0009, 0.0044, 0.0000, ..., 0.0032, 0.0010, -0.0077], + [ 0.0024, 0.0029, 0.0000, ..., -0.0023, 0.0004, 0.0051], + [ 0.0071, 0.0080, 0.0000, ..., 0.0074, -0.0032, 0.0082], + ..., + [-0.0035, -0.0287, 0.0000, ..., -0.0062, 0.0003, 0.0019], + [-0.0014, -0.0081, 0.0000, ..., -0.0097, 0.0010, 0.0038], + [-0.0044, 0.0081, 0.0000, ..., 0.0007, 0.0002, -0.0134]], + device='cuda:0') +Epoch 8, bias, value: tensor([-0.0285, -0.0041, -0.0085, -0.0080, -0.0128, 0.0077, -0.0032, 0.0194, + -0.0024, 0.0242], device='cuda:0'), grad: tensor([-0.0442, 0.0173, 0.0562, -0.0463, -0.0025, 0.0402, 0.0295, -0.0161, + -0.0328, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 214.66, cls_loss 1.3349 cls_loss_mapping 0.2118 cls_loss_causal 1.1245 re_mapping 0.0704 re_causal 0.1275 /// teacc 95.25 lr 0.00010000 +Epoch 9, weight, value: tensor([[-0.0420, -0.0382, -0.0205, ..., 0.0037, -0.0418, -0.0217], + [ 0.0056, -0.0189, 0.0241, ..., 0.0020, 0.0007, 0.0010], + [ 0.0173, -0.0142, -0.0218, ..., 0.0290, -0.0144, -0.0035], + ..., + [-0.0117, 0.0192, 0.0109, ..., 0.0117, -0.0313, -0.0170], + [-0.0252, -0.0069, 0.0124, ..., 0.0261, -0.0080, 0.0003], + [ 0.0035, 0.0221, 0.0078, ..., -0.0395, -0.0385, -0.0226]], + device='cuda:0'), grad: tensor([[-3.9291e-03, -2.0325e-02, -1.5993e-03, ..., -8.2092e-03, + 1.2512e-03, -9.1476e-03], + [ 1.5507e-03, 3.0255e-04, 9.7394e-05, ..., -3.2166e-02, + -5.8556e-03, 4.0436e-03], + [ 2.1954e-03, 4.0321e-03, 1.8489e-04, ..., 9.0637e-03, + -8.9455e-04, 4.9591e-03], + ..., + [ 4.6992e-04, 2.9926e-03, 5.2881e-04, ..., 1.6270e-03, + 1.5116e-03, 6.2981e-03], + [ 5.3406e-03, 1.4908e-02, 6.1226e-04, ..., 7.1182e-03, + 2.7351e-03, 1.0880e-02], + [ 5.3692e-04, -6.4735e-03, -6.7711e-05, ..., 5.5351e-03, + 3.6359e-04, 1.0109e-03]], device='cuda:0') +Epoch 9, bias, value: tensor([-0.0288, -0.0039, -0.0086, -0.0078, -0.0133, 0.0078, -0.0030, 0.0195, + -0.0023, 0.0242], device='cuda:0'), grad: tensor([-0.1016, -0.0212, 0.0164, 0.0358, 0.0102, -0.0010, -0.0588, 0.0231, + 0.0649, 0.0322], device='cuda:0') +100 +0.0001 +changing lr +epoch 8, time 213.70, cls_loss 1.2971 cls_loss_mapping 0.1972 cls_loss_causal 1.0997 re_mapping 0.0653 re_causal 0.1244 /// teacc 94.86 lr 0.00010000 +Epoch 10, weight, value: tensor([[-4.3552e-02, -3.7830e-02, -2.1098e-02, ..., 3.3888e-03, + -4.3135e-02, -2.3272e-02], + [ 6.0748e-03, -1.8447e-02, 1.8358e-02, ..., 3.7643e-03, + 6.2216e-05, 1.1227e-03], + [ 1.6568e-02, -1.5606e-02, -2.7065e-02, ..., 2.9607e-02, + -1.3574e-02, -4.9672e-03], + ..., + [-1.2801e-02, 1.9409e-02, 1.3275e-02, ..., 1.2541e-02, + -2.9741e-02, -1.7815e-02], + [-2.5315e-02, -7.5025e-03, 9.3757e-03, ..., 2.6315e-02, + -7.5325e-03, -1.2798e-03], + [ 4.7083e-03, 2.3244e-02, 6.9806e-03, ..., -4.0716e-02, + -3.9597e-02, -2.1435e-02]], device='cuda:0'), grad: tensor([[ 0.0013, 0.0075, 0.0000, ..., 0.0133, 0.0038, 0.0066], + [ 0.0004, 0.0009, 0.0000, ..., -0.0127, -0.0105, -0.0049], + [ 0.0006, 0.0013, 0.0000, ..., 0.0024, 0.0008, -0.0020], + ..., + [ 0.0007, 0.0434, 0.0000, ..., 0.0038, 0.0186, 0.0069], + [-0.0045, 0.0124, 0.0000, ..., 0.0002, 0.0091, 0.0041], + [ 0.0191, 0.0310, 0.0000, ..., 0.0008, 0.0136, 0.0075]], + device='cuda:0') +Epoch 10, bias, value: tensor([-0.0290, -0.0037, -0.0084, -0.0078, -0.0133, 0.0078, -0.0033, 0.0197, + -0.0025, 0.0245], device='cuda:0'), grad: tensor([ 0.0701, -0.0559, -0.0114, -0.0258, 0.0330, -0.0676, 0.0064, 0.0322, + -0.0017, 0.0208], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 214.34, cls_loss 1.2984 cls_loss_mapping 0.1952 cls_loss_causal 1.1072 re_mapping 0.0583 re_causal 0.1136 /// teacc 95.89 lr 0.00010000 +Epoch 11, weight, value: tensor([[-0.0456, -0.0365, -0.0209, ..., 0.0039, -0.0440, -0.0239], + [ 0.0058, -0.0187, 0.0172, ..., 0.0057, 0.0003, 0.0001], + [ 0.0153, -0.0175, -0.0290, ..., 0.0310, -0.0128, -0.0053], + ..., + [-0.0120, 0.0194, 0.0153, ..., 0.0115, -0.0287, -0.0172], + [-0.0260, -0.0059, 0.0085, ..., 0.0268, -0.0068, -0.0039], + [ 0.0046, 0.0232, 0.0045, ..., -0.0435, -0.0413, -0.0203]], + device='cuda:0'), grad: tensor([[ 2.8858e-03, 3.7880e-03, 5.0932e-05, ..., 4.7379e-03, + 2.7561e-03, 3.7594e-03], + [ 2.2488e-03, 2.3926e-02, 3.9196e-04, ..., 1.8829e-02, + 5.8411e-02, 5.2338e-03], + [-3.6697e-03, 3.9406e-03, 1.0997e-04, ..., 2.6550e-03, + -3.8376e-03, 3.8967e-03], + ..., + [ 1.0633e-03, 1.5335e-02, 3.2115e-04, ..., 7.8812e-03, + 2.4841e-02, 6.0120e-03], + [ 3.6430e-03, -2.3346e-02, 3.4475e-04, ..., -1.7090e-02, + -5.3284e-02, -1.3666e-03], + [ 2.3708e-03, -7.0047e-04, 1.0395e-03, ..., 4.3106e-03, + 1.6747e-03, 4.6539e-03]], device='cuda:0') +Epoch 11, bias, value: tensor([-0.0291, -0.0034, -0.0077, -0.0078, -0.0129, 0.0077, -0.0036, 0.0200, + -0.0032, 0.0241], device='cuda:0'), grad: tensor([ 0.0412, 0.0699, 0.0022, -0.0941, -0.0242, -0.0572, 0.0221, 0.0344, + -0.0367, 0.0425], device='cuda:0') +100 +0.0001 +changing lr +epoch 10, time 214.03, cls_loss 1.2460 cls_loss_mapping 0.1566 cls_loss_causal 1.0507 re_mapping 0.0544 re_causal 0.1092 /// teacc 95.12 lr 0.00010000 +Epoch 12, weight, value: tensor([[-0.0470, -0.0365, -0.0199, ..., 0.0049, -0.0428, -0.0249], + [ 0.0053, -0.0196, 0.0171, ..., 0.0063, 0.0007, -0.0012], + [ 0.0142, -0.0180, -0.0329, ..., 0.0316, -0.0125, -0.0057], + ..., + [-0.0129, 0.0200, 0.0181, ..., 0.0113, -0.0301, -0.0175], + [-0.0269, -0.0060, 0.0083, ..., 0.0281, -0.0072, -0.0047], + [ 0.0059, 0.0234, 0.0041, ..., -0.0444, -0.0419, -0.0188]], + device='cuda:0'), grad: tensor([[ 0.0019, 0.0021, 0.0000, ..., -0.0088, -0.0006, -0.0013], + [ 0.0018, -0.0016, 0.0000, ..., -0.0290, -0.0007, -0.0173], + [ 0.0027, 0.0050, 0.0000, ..., 0.0138, 0.0022, 0.0123], + ..., + [ 0.0015, 0.0058, 0.0000, ..., 0.0018, 0.0029, -0.0048], + [-0.0011, 0.0007, 0.0000, ..., 0.0019, -0.0028, -0.0012], + [-0.0018, -0.0118, 0.0000, ..., 0.0121, -0.0015, 0.0003]], + device='cuda:0') +Epoch 12, bias, value: tensor([-0.0289, -0.0030, -0.0077, -0.0079, -0.0132, 0.0069, -0.0034, 0.0194, + -0.0031, 0.0247], device='cuda:0'), grad: tensor([-0.0415, -0.0632, 0.0842, -0.0616, -0.0169, 0.0559, 0.0486, 0.0276, + -0.0253, -0.0077], device='cuda:0') +100 +0.0001 +changing lr +epoch 11, time 213.93, cls_loss 1.2324 cls_loss_mapping 0.1540 cls_loss_causal 1.0474 re_mapping 0.0533 re_causal 0.1093 /// teacc 95.60 lr 0.00010000 +Epoch 13, weight, value: tensor([[-0.0483, -0.0362, -0.0198, ..., 0.0055, -0.0427, -0.0268], + [ 0.0051, -0.0214, 0.0160, ..., 0.0066, 0.0010, -0.0021], + [ 0.0132, -0.0188, -0.0329, ..., 0.0324, -0.0124, -0.0068], + ..., + [-0.0127, 0.0200, 0.0173, ..., 0.0119, -0.0309, -0.0170], + [-0.0277, -0.0059, 0.0082, ..., 0.0287, -0.0074, -0.0053], + [ 0.0057, 0.0242, 0.0040, ..., -0.0460, -0.0408, -0.0191]], + device='cuda:0'), grad: tensor([[ 2.5864e-03, 3.2623e-02, 9.7603e-06, ..., 4.2999e-02, + 5.2719e-03, 3.6163e-03], + [ 5.6791e-04, 2.5368e-03, 3.8445e-05, ..., 1.7872e-03, + 3.7880e-03, 9.8610e-04], + [ 1.5736e-03, 1.3323e-03, 1.0915e-05, ..., -8.0109e-03, + -1.4908e-02, 4.6959e-03], + ..., + [ 1.3142e-03, 1.1730e-03, 4.1097e-05, ..., 4.1962e-03, + 1.9598e-04, 8.8959e-03], + [ 5.4970e-03, -1.8478e-02, 4.1097e-05, ..., -2.0996e-02, + -1.9531e-03, -8.4000e-03], + [ 9.5654e-04, 2.4681e-03, 2.7061e-04, ..., 9.5654e-04, + 5.3253e-03, 8.0185e-03]], device='cuda:0') +Epoch 13, bias, value: tensor([-0.0289, -0.0031, -0.0077, -0.0073, -0.0130, 0.0068, -0.0038, 0.0197, + -0.0030, 0.0242], device='cuda:0'), grad: tensor([ 0.0989, -0.0048, -0.0379, -0.0505, -0.0331, 0.0029, 0.0288, 0.0159, + -0.0348, 0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 12, time 213.87, cls_loss 1.1705 cls_loss_mapping 0.1441 cls_loss_causal 0.9753 re_mapping 0.0510 re_causal 0.1034 /// teacc 95.55 lr 0.00010000 +Epoch 14, weight, value: tensor([[-0.0486, -0.0362, -0.0203, ..., 0.0049, -0.0427, -0.0268], + [ 0.0049, -0.0232, 0.0146, ..., 0.0073, 0.0029, -0.0031], + [ 0.0117, -0.0196, -0.0341, ..., 0.0318, -0.0126, -0.0082], + ..., + [-0.0126, 0.0199, 0.0179, ..., 0.0127, -0.0302, -0.0169], + [-0.0281, -0.0068, 0.0086, ..., 0.0295, -0.0081, -0.0061], + [ 0.0071, 0.0253, 0.0019, ..., -0.0473, -0.0427, -0.0188]], + device='cuda:0'), grad: tensor([[ 4.8423e-04, 1.7729e-03, 1.2124e-04, ..., -2.7981e-03, + -1.3800e-03, -4.7302e-03], + [ 3.2120e-03, 2.0103e-03, 8.0168e-05, ..., -6.1722e-03, + -1.1454e-03, -9.4452e-03], + [ 1.4086e-03, 3.1471e-03, 6.5982e-05, ..., -2.6760e-03, + 1.8911e-03, 2.0905e-03], + ..., + [ 9.6512e-04, -2.8610e-04, 4.6420e-04, ..., -2.8172e-03, + -2.7823e-04, 3.6335e-03], + [-1.8072e-03, 4.7951e-03, 2.5702e-04, ..., 1.3418e-03, + 1.5306e-03, 4.3144e-03], + [ 4.2343e-03, -1.1650e-02, 4.3869e-03, ..., 3.5686e-03, + 1.1635e-03, 6.0043e-03]], device='cuda:0') +Epoch 14, bias, value: tensor([-0.0289, -0.0031, -0.0076, -0.0073, -0.0126, 0.0068, -0.0040, 0.0198, + -0.0028, 0.0236], device='cuda:0'), grad: tensor([-0.0390, -0.0440, -0.0130, 0.0498, 0.0197, -0.0407, 0.0337, 0.0077, + -0.0064, 0.0323], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 214.36, cls_loss 1.1478 cls_loss_mapping 0.1399 cls_loss_causal 0.9650 re_mapping 0.0493 re_causal 0.1019 /// teacc 96.32 lr 0.00010000 +Epoch 15, weight, value: tensor([[-0.0501, -0.0359, -0.0193, ..., 0.0048, -0.0422, -0.0272], + [ 0.0041, -0.0235, 0.0101, ..., 0.0078, 0.0025, -0.0040], + [ 0.0110, -0.0202, -0.0305, ..., 0.0329, -0.0129, -0.0082], + ..., + [-0.0124, 0.0189, 0.0183, ..., 0.0125, -0.0300, -0.0176], + [-0.0288, -0.0071, 0.0079, ..., 0.0309, -0.0075, -0.0066], + [ 0.0064, 0.0273, 0.0016, ..., -0.0478, -0.0413, -0.0190]], + device='cuda:0'), grad: tensor([[ 0.0010, -0.0328, 0.0002, ..., 0.0006, 0.0044, 0.0013], + [ 0.0009, 0.0005, 0.0002, ..., 0.0006, 0.0007, 0.0008], + [ 0.0017, 0.0049, 0.0004, ..., 0.0089, 0.0141, 0.0046], + ..., + [-0.0076, 0.0029, -0.0029, ..., -0.0006, 0.0023, -0.0035], + [-0.0017, 0.0122, 0.0013, ..., 0.0124, 0.0198, 0.0027], + [ 0.0019, -0.0105, 0.0017, ..., -0.0141, -0.0169, 0.0001]], + device='cuda:0') +Epoch 15, bias, value: tensor([-0.0285, -0.0030, -0.0071, -0.0074, -0.0127, 0.0062, -0.0046, 0.0198, + -0.0021, 0.0234], device='cuda:0'), grad: tensor([-0.0182, 0.0141, 0.0750, 0.0158, 0.0573, -0.0111, -0.1108, -0.0547, + 0.0310, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +epoch 14, time 213.84, cls_loss 1.1324 cls_loss_mapping 0.1232 cls_loss_causal 0.9501 re_mapping 0.0482 re_causal 0.1021 /// teacc 96.28 lr 0.00010000 +Epoch 16, weight, value: tensor([[-0.0515, -0.0362, -0.0202, ..., 0.0039, -0.0432, -0.0280], + [ 0.0042, -0.0243, 0.0101, ..., 0.0081, 0.0029, -0.0047], + [ 0.0106, -0.0217, -0.0315, ..., 0.0328, -0.0134, -0.0089], + ..., + [-0.0126, 0.0195, 0.0203, ..., 0.0133, -0.0294, -0.0181], + [-0.0294, -0.0073, 0.0077, ..., 0.0311, -0.0082, -0.0069], + [ 0.0069, 0.0274, 0.0008, ..., -0.0473, -0.0413, -0.0178]], + device='cuda:0'), grad: tensor([[ 1.0004e-03, 4.0591e-05, 0.0000e+00, ..., -5.3864e-03, + -8.7738e-05, -3.0365e-03], + [ 1.0996e-03, 2.0123e-04, 0.0000e+00, ..., -4.6844e-03, + -2.4090e-03, -5.8594e-03], + [ 3.1042e-04, -9.4652e-05, 0.0000e+00, ..., 1.7080e-03, + -5.2500e-04, 2.1915e-03], + ..., + [ 1.2398e-03, 6.4316e-03, 0.0000e+00, ..., -2.6817e-03, + 7.6103e-04, 1.0033e-03], + [-5.6458e-03, -6.1989e-03, 0.0000e+00, ..., 1.3313e-03, + -7.8630e-04, -7.3624e-04], + [ 1.6260e-03, 2.3544e-02, 0.0000e+00, ..., 8.0795e-03, + 8.6136e-03, 4.4136e-03]], device='cuda:0') +Epoch 16, bias, value: tensor([-0.0290, -0.0034, -0.0073, -0.0072, -0.0126, 0.0063, -0.0042, 0.0200, + -0.0024, 0.0236], device='cuda:0'), grad: tensor([-0.0291, -0.0085, -0.0170, 0.0105, -0.0662, -0.0108, 0.0465, 0.0114, + 0.0035, 0.0597], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 214.41, cls_loss 1.1358 cls_loss_mapping 0.1230 cls_loss_causal 0.9681 re_mapping 0.0443 re_causal 0.0972 /// teacc 96.54 lr 0.00010000 +Epoch 17, weight, value: tensor([[-5.3411e-02, -3.6009e-02, -1.9137e-02, ..., 3.6422e-03, + -4.3475e-02, -2.8798e-02], + [ 3.5445e-03, -2.4572e-02, 8.2040e-03, ..., 8.1961e-03, + 2.0793e-03, -4.5984e-03], + [ 1.0166e-02, -2.3038e-02, -3.2474e-02, ..., 3.3691e-02, + -1.2874e-02, -9.1427e-03], + ..., + [-1.2993e-02, 1.9616e-02, 1.9004e-02, ..., 1.4400e-02, + -2.8677e-02, -1.8872e-02], + [-3.0115e-02, -6.2024e-03, 7.6136e-03, ..., 3.1983e-02, + -8.7944e-03, -7.5004e-03], + [ 7.8345e-03, 2.7041e-02, -7.4626e-05, ..., -4.8305e-02, + -4.0654e-02, -1.7473e-02]], device='cuda:0'), grad: tensor([[ 6.0797e-04, 1.1501e-03, 7.0453e-05, ..., 4.9820e-03, + 4.3411e-03, 2.4967e-03], + [ 3.3379e-04, 2.3937e-04, 6.2227e-05, ..., -7.5226e-03, + -7.6342e-04, -1.3170e-03], + [ 8.6927e-04, 5.2547e-04, 6.0648e-05, ..., -1.7147e-03, + -6.4201e-03, -5.7449e-03], + ..., + [ 2.2297e-03, 2.3117e-03, -1.0824e-03, ..., -1.6966e-03, + 2.0981e-05, 3.3226e-03], + [-7.7400e-03, 2.5692e-03, 7.2479e-05, ..., -3.0212e-03, + -4.3449e-03, 1.1663e-03], + [ 1.9730e-02, 2.5436e-02, 1.3590e-03, ..., -8.7357e-04, + -6.2828e-03, 1.4160e-02]], device='cuda:0') +Epoch 17, bias, value: tensor([-0.0289, -0.0033, -0.0069, -0.0074, -0.0123, 0.0069, -0.0048, 0.0202, + -0.0028, 0.0232], device='cuda:0'), grad: tensor([ 0.0191, -0.0114, -0.0294, -0.0221, -0.0402, 0.0311, 0.0584, -0.0183, + -0.0270, 0.0398], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 214.38, cls_loss 1.0765 cls_loss_mapping 0.1155 cls_loss_causal 0.9108 re_mapping 0.0435 re_causal 0.0941 /// teacc 96.73 lr 0.00010000 +Epoch 18, weight, value: tensor([[-0.0549, -0.0355, -0.0194, ..., 0.0023, -0.0434, -0.0301], + [ 0.0022, -0.0243, 0.0092, ..., 0.0097, 0.0036, -0.0057], + [ 0.0087, -0.0215, -0.0330, ..., 0.0354, -0.0121, -0.0099], + ..., + [-0.0133, 0.0190, 0.0185, ..., 0.0144, -0.0288, -0.0202], + [-0.0302, -0.0063, 0.0062, ..., 0.0320, -0.0093, -0.0086], + [ 0.0086, 0.0271, -0.0018, ..., -0.0492, -0.0410, -0.0157]], + device='cuda:0'), grad: tensor([[ 7.2956e-04, 3.2544e-05, 3.9387e-04, ..., 6.8207e-03, + 3.9520e-03, 4.0894e-03], + [ 1.3380e-03, 4.1723e-04, 2.6913e-03, ..., 2.8400e-03, + 1.6804e-03, -6.0768e-03], + [-3.6502e-04, 5.5522e-05, -8.2932e-03, ..., -2.0645e-02, + -9.4604e-03, 1.2070e-04], + ..., + [-9.3460e-05, 2.6941e-04, 1.3936e-04, ..., 9.9182e-05, + 4.0550e-03, 5.0583e-03], + [-1.3723e-03, 9.8944e-05, 1.2302e-03, ..., -1.3924e-03, + -6.9904e-04, 1.8196e-03], + [ 3.5858e-03, 2.1935e-03, 5.1785e-04, ..., 9.0637e-03, + 5.2299e-03, 9.6741e-03]], device='cuda:0') +Epoch 18, bias, value: tensor([-0.0295, -0.0030, -0.0066, -0.0070, -0.0118, 0.0057, -0.0046, 0.0202, + -0.0028, 0.0233], device='cuda:0'), grad: tensor([ 0.0323, -0.0053, -0.0860, 0.0056, -0.0185, 0.0309, -0.0240, 0.0267, + -0.0131, 0.0514], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 17---------------------------------------------------- +epoch 17, time 214.39, cls_loss 1.0657 cls_loss_mapping 0.0996 cls_loss_causal 0.9001 re_mapping 0.0429 re_causal 0.0966 /// teacc 96.77 lr 0.00010000 +Epoch 19, weight, value: tensor([[-0.0557, -0.0356, -0.0181, ..., 0.0021, -0.0430, -0.0306], + [ 0.0019, -0.0258, 0.0090, ..., 0.0101, 0.0032, -0.0065], + [ 0.0072, -0.0210, -0.0331, ..., 0.0364, -0.0114, -0.0113], + ..., + [-0.0119, 0.0189, 0.0194, ..., 0.0145, -0.0291, -0.0200], + [-0.0302, -0.0063, 0.0043, ..., 0.0329, -0.0099, -0.0097], + [ 0.0087, 0.0284, -0.0026, ..., -0.0505, -0.0398, -0.0157]], + device='cuda:0'), grad: tensor([[ 0.0008, -0.0026, 0.0005, ..., 0.0016, 0.0039, 0.0011], + [-0.0026, -0.0186, 0.0021, ..., -0.0257, -0.0202, 0.0002], + [ 0.0006, 0.0012, 0.0035, ..., -0.0010, 0.0076, -0.0006], + ..., + [-0.0005, -0.0091, -0.0194, ..., 0.0108, -0.0344, 0.0014], + [-0.0045, 0.0158, 0.0006, ..., 0.0066, 0.0014, 0.0038], + [-0.0010, 0.0050, 0.0022, ..., -0.0045, -0.0015, -0.0054]], + device='cuda:0') +Epoch 19, bias, value: tensor([-0.0291, -0.0033, -0.0063, -0.0068, -0.0115, 0.0056, -0.0048, 0.0202, + -0.0028, 0.0226], device='cuda:0'), grad: tensor([ 0.0177, -0.0679, 0.0006, 0.0368, 0.0659, 0.0206, -0.0080, -0.0128, + -0.0180, -0.0349], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 214.45, cls_loss 1.0378 cls_loss_mapping 0.1012 cls_loss_causal 0.8786 re_mapping 0.0422 re_causal 0.0929 /// teacc 96.85 lr 0.00010000 +Epoch 20, weight, value: tensor([[-0.0561, -0.0350, -0.0181, ..., 0.0022, -0.0435, -0.0317], + [ 0.0017, -0.0253, 0.0102, ..., 0.0091, 0.0025, -0.0072], + [ 0.0067, -0.0209, -0.0344, ..., 0.0373, -0.0118, -0.0118], + ..., + [-0.0124, 0.0174, 0.0208, ..., 0.0148, -0.0284, -0.0200], + [-0.0312, -0.0064, 0.0034, ..., 0.0330, -0.0105, -0.0115], + [ 0.0091, 0.0288, -0.0039, ..., -0.0522, -0.0403, -0.0152]], + device='cuda:0'), grad: tensor([[-0.0040, -0.0044, -0.0027, ..., -0.0190, -0.0028, -0.0084], + [ 0.0011, 0.0004, 0.0221, ..., 0.0233, 0.0210, 0.0067], + [ 0.0019, 0.0006, 0.0030, ..., 0.0045, 0.0022, 0.0004], + ..., + [-0.0031, 0.0004, -0.0008, ..., -0.0100, -0.0044, -0.0120], + [-0.0089, 0.0007, -0.0036, ..., -0.0125, -0.0033, 0.0020], + [ 0.0049, 0.0003, 0.0037, ..., 0.0089, 0.0030, 0.0038]], + device='cuda:0') +Epoch 20, bias, value: tensor([-0.0291, -0.0034, -0.0067, -0.0061, -0.0110, 0.0058, -0.0050, 0.0209, + -0.0036, 0.0222], device='cuda:0'), grad: tensor([-0.0720, 0.0888, 0.0301, 0.0028, -0.0207, 0.0293, -0.0072, -0.0741, + -0.0208, 0.0438], device='cuda:0') +100 +0.0001 +changing lr +epoch 19, time 213.87, cls_loss 1.0126 cls_loss_mapping 0.0928 cls_loss_causal 0.8729 re_mapping 0.0404 re_causal 0.0917 /// teacc 95.77 lr 0.00010000 +Epoch 21, weight, value: tensor([[-0.0573, -0.0353, -0.0182, ..., 0.0020, -0.0444, -0.0322], + [ 0.0016, -0.0269, 0.0088, ..., 0.0103, 0.0034, -0.0069], + [ 0.0062, -0.0217, -0.0364, ..., 0.0372, -0.0124, -0.0118], + ..., + [-0.0125, 0.0172, 0.0189, ..., 0.0147, -0.0284, -0.0210], + [-0.0321, -0.0062, 0.0077, ..., 0.0323, -0.0113, -0.0131], + [ 0.0100, 0.0295, -0.0037, ..., -0.0505, -0.0389, -0.0143]], + device='cuda:0'), grad: tensor([[-0.0057, 0.0022, -0.0111, ..., -0.0007, 0.0005, 0.0003], + [ 0.0014, 0.0006, 0.0009, ..., 0.0013, -0.0019, 0.0016], + [ 0.0027, 0.0014, 0.0018, ..., 0.0062, 0.0015, 0.0044], + ..., + [ 0.0009, -0.0133, 0.0002, ..., -0.0044, 0.0007, -0.0023], + [ 0.0040, -0.0034, 0.0031, ..., 0.0004, 0.0027, 0.0019], + [ 0.0035, 0.0067, 0.0043, ..., 0.0013, 0.0012, 0.0021]], + device='cuda:0') +Epoch 21, bias, value: tensor([-0.0293, -0.0034, -0.0062, -0.0067, -0.0113, 0.0055, -0.0049, 0.0208, + -0.0035, 0.0229], device='cuda:0'), grad: tensor([-0.0309, 0.0101, 0.0370, -0.0602, 0.0139, -0.0463, 0.0466, -0.0153, + 0.0187, 0.0264], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 20---------------------------------------------------- +epoch 20, time 214.55, cls_loss 1.0210 cls_loss_mapping 0.1005 cls_loss_causal 0.8468 re_mapping 0.0392 re_causal 0.0819 /// teacc 96.98 lr 0.00010000 +Epoch 22, weight, value: tensor([[-0.0594, -0.0349, -0.0154, ..., 0.0019, -0.0458, -0.0329], + [ 0.0017, -0.0274, 0.0075, ..., 0.0109, 0.0025, -0.0072], + [ 0.0056, -0.0222, -0.0367, ..., 0.0370, -0.0126, -0.0127], + ..., + [-0.0135, 0.0189, 0.0211, ..., 0.0148, -0.0273, -0.0220], + [-0.0322, -0.0065, 0.0089, ..., 0.0325, -0.0116, -0.0135], + [ 0.0099, 0.0289, -0.0046, ..., -0.0512, -0.0396, -0.0137]], + device='cuda:0'), grad: tensor([[ 0.0016, 0.0001, 0.0009, ..., 0.0040, 0.0019, 0.0027], + [ 0.0005, 0.0011, 0.0016, ..., 0.0055, 0.0040, 0.0014], + [ 0.0006, 0.0051, 0.0011, ..., 0.0114, 0.0087, 0.0042], + ..., + [-0.0021, -0.0065, 0.0020, ..., -0.0248, -0.0199, 0.0025], + [ 0.0025, 0.0019, -0.0051, ..., -0.0026, -0.0013, 0.0036], + [ 0.0057, -0.0054, -0.0053, ..., 0.0012, 0.0013, -0.0032]], + device='cuda:0') +Epoch 22, bias, value: tensor([-0.0295, -0.0037, -0.0058, -0.0063, -0.0111, 0.0057, -0.0056, 0.0212, + -0.0033, 0.0223], device='cuda:0'), grad: tensor([ 0.0220, 0.0302, 0.0352, 0.0206, 0.0146, -0.0574, 0.0176, -0.0756, + -0.0042, -0.0029], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 21---------------------------------------------------- +epoch 21, time 214.82, cls_loss 1.0064 cls_loss_mapping 0.0877 cls_loss_causal 0.8580 re_mapping 0.0375 re_causal 0.0837 /// teacc 97.28 lr 0.00010000 +Epoch 23, weight, value: tensor([[-0.0610, -0.0349, -0.0162, ..., 0.0016, -0.0459, -0.0338], + [ 0.0011, -0.0277, 0.0073, ..., 0.0110, 0.0023, -0.0074], + [ 0.0038, -0.0212, -0.0374, ..., 0.0361, -0.0131, -0.0133], + ..., + [-0.0142, 0.0191, 0.0212, ..., 0.0162, -0.0265, -0.0223], + [-0.0322, -0.0060, 0.0084, ..., 0.0327, -0.0121, -0.0143], + [ 0.0107, 0.0292, -0.0052, ..., -0.0523, -0.0395, -0.0135]], + device='cuda:0'), grad: tensor([[ 1.5011e-03, 5.3883e-04, 3.8266e-05, ..., 3.6964e-03, + 1.2836e-03, 1.1568e-03], + [ 1.0386e-03, 3.8090e-03, 3.8671e-04, ..., -2.8725e-03, + -3.6068e-03, -1.6432e-03], + [ 1.2274e-03, 4.7660e-04, 1.1277e-04, ..., -1.2276e-02, + -3.5915e-03, 1.6708e-03], + ..., + [-2.1324e-03, 2.4147e-03, 3.7527e-04, ..., 6.7863e-03, + 3.5553e-03, 1.7605e-03], + [ 7.5569e-03, 1.6332e-04, 1.5831e-04, ..., 3.7994e-03, + -4.1351e-03, 4.0665e-03], + [ 1.2608e-03, -1.9547e-02, 8.4972e-04, ..., -1.1452e-02, + -8.6594e-03, -3.2616e-03]], device='cuda:0') +Epoch 23, bias, value: tensor([-0.0295, -0.0033, -0.0063, -0.0070, -0.0105, 0.0062, -0.0057, 0.0219, + -0.0037, 0.0218], device='cuda:0'), grad: tensor([ 0.0134, -0.0265, -0.0247, 0.0339, 0.0248, -0.0520, 0.0249, 0.0190, + 0.0284, -0.0412], device='cuda:0') +100 +0.0001 +changing lr +epoch 22, time 214.66, cls_loss 0.9904 cls_loss_mapping 0.0934 cls_loss_causal 0.8568 re_mapping 0.0378 re_causal 0.0843 /// teacc 97.08 lr 0.00010000 +Epoch 24, weight, value: tensor([[-0.0613, -0.0353, -0.0162, ..., 0.0020, -0.0457, -0.0337], + [ 0.0009, -0.0279, 0.0061, ..., 0.0110, 0.0020, -0.0081], + [ 0.0023, -0.0217, -0.0384, ..., 0.0356, -0.0133, -0.0144], + ..., + [-0.0135, 0.0185, 0.0203, ..., 0.0167, -0.0274, -0.0223], + [-0.0324, -0.0065, 0.0085, ..., 0.0327, -0.0121, -0.0152], + [ 0.0105, 0.0301, -0.0041, ..., -0.0521, -0.0401, -0.0139]], + device='cuda:0'), grad: tensor([[ 2.1152e-03, 3.5453e-04, 4.0568e-06, ..., 6.2752e-03, + 5.2643e-03, 2.9888e-03], + [ 6.0892e-04, 1.5888e-03, 1.0327e-05, ..., 4.7836e-03, + 6.0005e-03, 1.5488e-03], + [ 1.2493e-03, 5.8889e-04, 1.2541e-04, ..., -1.4481e-02, + -1.6861e-02, -1.2207e-03], + ..., + [ 7.7868e-04, 4.5280e-03, 1.8895e-04, ..., 1.0292e-02, + 4.9820e-03, 3.5000e-03], + [ 3.9177e-03, -6.7139e-03, 1.5914e-04, ..., -1.4229e-02, + -4.4250e-03, -2.6588e-03], + [-5.7755e-03, -5.4626e-03, -5.7411e-04, ..., 3.4828e-03, + -2.8496e-03, -2.6569e-03]], device='cuda:0') +Epoch 24, bias, value: tensor([-0.0296, -0.0039, -0.0064, -0.0068, -0.0099, 0.0059, -0.0055, 0.0215, + -0.0036, 0.0221], device='cuda:0'), grad: tensor([ 0.0450, 0.0284, -0.0471, -0.0001, 0.0109, 0.0132, -0.0010, 0.0220, + -0.0507, -0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 23, time 214.27, cls_loss 0.9814 cls_loss_mapping 0.0806 cls_loss_causal 0.8296 re_mapping 0.0375 re_causal 0.0823 /// teacc 96.89 lr 0.00010000 +Epoch 25, weight, value: tensor([[-0.0630, -0.0352, -0.0164, ..., 0.0021, -0.0451, -0.0341], + [ 0.0011, -0.0289, 0.0063, ..., 0.0116, 0.0027, -0.0086], + [ 0.0020, -0.0216, -0.0394, ..., 0.0360, -0.0125, -0.0147], + ..., + [-0.0140, 0.0188, 0.0199, ..., 0.0168, -0.0267, -0.0225], + [-0.0327, -0.0059, 0.0081, ..., 0.0333, -0.0116, -0.0160], + [ 0.0117, 0.0299, -0.0054, ..., -0.0525, -0.0408, -0.0122]], + device='cuda:0'), grad: tensor([[ 1.4381e-03, -1.9882e-02, 1.4149e-05, ..., 9.1028e-04, + 2.8210e-03, -4.5151e-05], + [ 1.4267e-03, 1.7405e-03, 1.6809e-05, ..., 4.4403e-03, + 4.0741e-03, 5.0783e-04], + [-4.9305e-04, 5.6114e-03, 1.7405e-05, ..., 1.8219e-02, + 1.7349e-02, -2.1791e-04], + ..., + [ 7.2823e-03, 2.8824e-02, 4.9978e-05, ..., 1.4313e-02, + 1.5175e-02, 1.4191e-03], + [ 1.3046e-03, 2.6428e-02, 1.7673e-05, ..., 9.9487e-03, + 5.2185e-03, 7.3385e-04], + [-1.0967e-03, -6.1127e-02, 2.3472e-04, ..., -1.4391e-03, + -1.7517e-02, 1.2674e-03]], device='cuda:0') +Epoch 25, bias, value: tensor([-0.0299, -0.0037, -0.0053, -0.0066, -0.0100, 0.0059, -0.0061, 0.0213, + -0.0039, 0.0222], device='cuda:0'), grad: tensor([-0.0477, 0.0216, 0.0273, -0.0892, 0.0505, -0.0183, 0.0038, 0.0479, + -0.0053, 0.0094], device='cuda:0') +100 +0.0001 +changing lr +epoch 24, time 214.07, cls_loss 0.9405 cls_loss_mapping 0.0924 cls_loss_causal 0.7898 re_mapping 0.0369 re_causal 0.0782 /// teacc 96.41 lr 0.00010000 +Epoch 26, weight, value: tensor([[-0.0631, -0.0352, -0.0158, ..., 0.0022, -0.0446, -0.0345], + [ 0.0019, -0.0302, 0.0049, ..., 0.0122, 0.0020, -0.0097], + [ 0.0009, -0.0235, -0.0403, ..., 0.0362, -0.0136, -0.0148], + ..., + [-0.0141, 0.0189, 0.0194, ..., 0.0178, -0.0249, -0.0220], + [-0.0327, -0.0051, 0.0079, ..., 0.0332, -0.0120, -0.0183], + [ 0.0120, 0.0297, -0.0060, ..., -0.0534, -0.0401, -0.0118]], + device='cuda:0'), grad: tensor([[ 0.0034, 0.0005, 0.0000, ..., 0.0037, 0.0011, 0.0047], + [ 0.0017, 0.0103, 0.0000, ..., 0.0128, 0.0087, 0.0035], + [ 0.0011, 0.0003, 0.0000, ..., 0.0098, 0.0096, 0.0058], + ..., + [ 0.0003, 0.0014, 0.0000, ..., 0.0016, 0.0056, 0.0121], + [ 0.0034, -0.0098, 0.0000, ..., 0.0028, 0.0042, 0.0048], + [-0.0060, 0.0026, 0.0000, ..., 0.0044, 0.0021, -0.0127]], + device='cuda:0') +Epoch 26, bias, value: tensor([-0.0300, -0.0038, -0.0051, -0.0064, -0.0098, 0.0060, -0.0063, 0.0216, + -0.0044, 0.0222], device='cuda:0'), grad: tensor([ 0.0184, 0.0254, 0.0115, -0.0589, -0.0382, -0.0010, -0.0042, 0.0152, + 0.0219, 0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 25, time 214.27, cls_loss 0.9942 cls_loss_mapping 0.0905 cls_loss_causal 0.8462 re_mapping 0.0353 re_causal 0.0793 /// teacc 97.17 lr 0.00010000 +Epoch 27, weight, value: tensor([[-0.0647, -0.0345, -0.0152, ..., 0.0019, -0.0446, -0.0352], + [ 0.0017, -0.0307, 0.0068, ..., 0.0131, 0.0023, -0.0108], + [-0.0013, -0.0237, -0.0398, ..., 0.0361, -0.0140, -0.0150], + ..., + [-0.0138, 0.0189, 0.0207, ..., 0.0181, -0.0248, -0.0219], + [-0.0336, -0.0060, 0.0085, ..., 0.0335, -0.0123, -0.0177], + [ 0.0131, 0.0300, -0.0061, ..., -0.0543, -0.0401, -0.0104]], + device='cuda:0'), grad: tensor([[-0.0025, -0.0001, 0.0017, ..., 0.0045, 0.0026, 0.0004], + [ 0.0022, 0.0006, 0.0019, ..., -0.0079, -0.0086, 0.0006], + [-0.0139, 0.0013, 0.0007, ..., 0.0066, 0.0048, 0.0010], + ..., + [ 0.0027, 0.0140, 0.0008, ..., 0.0075, -0.0010, 0.0020], + [ 0.0005, 0.0008, -0.0145, ..., -0.0087, -0.0019, 0.0011], + [ 0.0084, -0.0170, -0.0009, ..., -0.0089, 0.0026, 0.0019]], + device='cuda:0') +Epoch 27, bias, value: tensor([-0.0297, -0.0034, -0.0048, -0.0067, -0.0097, 0.0057, -0.0068, 0.0218, + -0.0043, 0.0217], device='cuda:0'), grad: tensor([ 0.0179, -0.0205, -0.0186, 0.0268, -0.0227, -0.0061, 0.0250, 0.0132, + -0.0119, -0.0031], device='cuda:0') +100 +0.0001 +changing lr +epoch 26, time 214.59, cls_loss 0.9781 cls_loss_mapping 0.0834 cls_loss_causal 0.8259 re_mapping 0.0353 re_causal 0.0751 /// teacc 97.01 lr 0.00010000 +Epoch 28, weight, value: tensor([[-0.0678, -0.0343, -0.0145, ..., 0.0019, -0.0441, -0.0363], + [ 0.0023, -0.0309, 0.0070, ..., 0.0132, 0.0016, -0.0108], + [-0.0020, -0.0243, -0.0415, ..., 0.0358, -0.0137, -0.0159], + ..., + [-0.0145, 0.0188, 0.0215, ..., 0.0184, -0.0249, -0.0230], + [-0.0341, -0.0060, 0.0087, ..., 0.0353, -0.0116, -0.0182], + [ 0.0139, 0.0308, -0.0061, ..., -0.0550, -0.0414, -0.0094]], + device='cuda:0'), grad: tensor([[-0.0015, 0.0002, -0.0010, ..., -0.0072, -0.0031, 0.0004], + [-0.0287, 0.0023, -0.0047, ..., -0.0292, -0.0007, -0.0004], + [ 0.0029, 0.0004, 0.0011, ..., -0.0010, -0.0024, 0.0011], + ..., + [ 0.0039, 0.0075, -0.0024, ..., 0.0043, 0.0055, -0.0033], + [ 0.0021, 0.0004, 0.0013, ..., 0.0057, 0.0067, -0.0006], + [ 0.0083, 0.0070, 0.0056, ..., 0.0051, 0.0107, 0.0046]], + device='cuda:0') +Epoch 28, bias, value: tensor([-0.0303, -0.0035, -0.0052, -0.0064, -0.0093, 0.0058, -0.0063, 0.0215, + -0.0038, 0.0215], device='cuda:0'), grad: tensor([-0.0315, -0.0773, -0.0070, 0.0156, -0.0135, 0.0079, 0.0230, 0.0170, + 0.0231, 0.0427], device='cuda:0') +100 +0.0001 +changing lr +epoch 27, time 214.48, cls_loss 0.9082 cls_loss_mapping 0.0781 cls_loss_causal 0.7707 re_mapping 0.0350 re_causal 0.0752 /// teacc 97.19 lr 0.00010000 +Epoch 29, weight, value: tensor([[-0.0690, -0.0342, -0.0137, ..., 0.0024, -0.0437, -0.0371], + [ 0.0035, -0.0306, 0.0070, ..., 0.0136, 0.0021, -0.0114], + [-0.0030, -0.0250, -0.0411, ..., 0.0357, -0.0142, -0.0155], + ..., + [-0.0143, 0.0187, 0.0222, ..., 0.0182, -0.0251, -0.0236], + [-0.0344, -0.0060, 0.0091, ..., 0.0353, -0.0114, -0.0200], + [ 0.0144, 0.0307, -0.0065, ..., -0.0557, -0.0420, -0.0090]], + device='cuda:0'), grad: tensor([[-0.0005, 0.0017, 0.0002, ..., -0.0139, -0.0078, -0.0013], + [ 0.0003, 0.0050, 0.0003, ..., 0.0367, 0.0102, -0.0008], + [ 0.0006, 0.0018, 0.0002, ..., 0.0047, 0.0014, 0.0018], + ..., + [-0.0006, -0.0029, -0.0053, ..., -0.0171, 0.0023, -0.0008], + [ 0.0026, -0.0245, 0.0023, ..., 0.0003, -0.0129, -0.0455], + [ 0.0078, -0.0304, 0.0028, ..., -0.0125, -0.0190, 0.0100]], + device='cuda:0') +Epoch 29, bias, value: tensor([-0.0302, -0.0033, -0.0053, -0.0064, -0.0089, 0.0064, -0.0069, 0.0216, + -0.0039, 0.0209], device='cuda:0'), grad: tensor([-0.0455, 0.0565, 0.0171, -0.0245, 0.0282, 0.0278, 0.0252, -0.0366, + -0.0260, -0.0221], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 214.82, cls_loss 0.9152 cls_loss_mapping 0.0649 cls_loss_causal 0.7640 re_mapping 0.0362 re_causal 0.0785 /// teacc 97.57 lr 0.00010000 +Epoch 30, weight, value: tensor([[-0.0700, -0.0351, -0.0128, ..., 0.0017, -0.0443, -0.0379], + [ 0.0031, -0.0317, 0.0069, ..., 0.0153, 0.0029, -0.0120], + [-0.0043, -0.0257, -0.0420, ..., 0.0360, -0.0140, -0.0165], + ..., + [-0.0138, 0.0196, 0.0227, ..., 0.0179, -0.0254, -0.0236], + [-0.0335, -0.0061, 0.0107, ..., 0.0358, -0.0121, -0.0197], + [ 0.0136, 0.0310, -0.0074, ..., -0.0568, -0.0403, -0.0088]], + device='cuda:0'), grad: tensor([[ 6.5346e-03, 8.1182e-05, 1.5373e-03, ..., -4.8141e-03, + 8.7738e-03, 1.2169e-03], + [-6.6490e-03, 2.7943e-04, -1.3313e-03, ..., -1.1244e-03, + 3.2921e-03, 1.3599e-03], + [ 1.4153e-03, 9.0539e-05, 1.0376e-03, ..., -4.0359e-03, + 7.8011e-03, 1.9073e-06], + ..., + [-7.3624e-03, 2.0618e-03, 1.2236e-03, ..., -1.4381e-03, + 2.3079e-03, -5.2071e-04], + [ 5.5695e-03, 8.2684e-04, 2.7580e-03, ..., -2.8229e-03, + 1.4458e-03, 3.7217e-04], + [-6.6605e-03, -3.1242e-03, -5.7697e-04, ..., -1.8723e-02, + -7.5684e-03, -5.8289e-03]], device='cuda:0') +Epoch 30, bias, value: tensor([-0.0306, -0.0032, -0.0051, -0.0065, -0.0089, 0.0063, -0.0073, 0.0219, + -0.0036, 0.0208], device='cuda:0'), grad: tensor([ 0.0257, -0.0084, -0.0237, 0.0232, 0.0395, 0.0166, -0.0076, -0.0289, + 0.0058, -0.0423], device='cuda:0') +100 +0.0001 +changing lr +epoch 29, time 214.63, cls_loss 0.9028 cls_loss_mapping 0.0782 cls_loss_causal 0.7685 re_mapping 0.0351 re_causal 0.0719 /// teacc 97.49 lr 0.00010000 +Epoch 31, weight, value: tensor([[-0.0711, -0.0345, -0.0120, ..., 0.0014, -0.0452, -0.0382], + [ 0.0027, -0.0316, 0.0064, ..., 0.0155, 0.0030, -0.0134], + [-0.0049, -0.0275, -0.0414, ..., 0.0357, -0.0141, -0.0167], + ..., + [-0.0143, 0.0192, 0.0219, ..., 0.0185, -0.0255, -0.0245], + [-0.0331, -0.0056, 0.0107, ..., 0.0365, -0.0132, -0.0198], + [ 0.0139, 0.0309, -0.0092, ..., -0.0567, -0.0403, -0.0081]], + device='cuda:0'), grad: tensor([[ 0.0003, -0.0186, -0.0009, ..., -0.0149, -0.0093, -0.0021], + [-0.0002, 0.0059, 0.0010, ..., 0.0049, 0.0036, 0.0002], + [-0.0129, 0.0027, -0.0039, ..., 0.0053, 0.0023, 0.0003], + ..., + [-0.0002, -0.0145, 0.0005, ..., 0.0008, -0.0033, 0.0013], + [ 0.0030, -0.0093, -0.0061, ..., -0.0023, -0.0040, 0.0010], + [ 0.0007, 0.0024, 0.0034, ..., 0.0029, 0.0033, -0.0033]], + device='cuda:0') +Epoch 31, bias, value: tensor([-0.0307, -0.0033, -0.0056, -0.0061, -0.0091, 0.0058, -0.0070, 0.0225, + -0.0031, 0.0204], device='cuda:0'), grad: tensor([-0.0697, 0.0264, -0.0152, 0.0435, -0.0588, 0.0179, 0.0884, -0.0110, + -0.0430, 0.0216], device='cuda:0') +100 +0.0001 +changing lr +epoch 30, time 214.29, cls_loss 0.9081 cls_loss_mapping 0.0647 cls_loss_causal 0.7586 re_mapping 0.0349 re_causal 0.0778 /// teacc 97.36 lr 0.00010000 +Epoch 32, weight, value: tensor([[-0.0726, -0.0348, -0.0126, ..., 0.0014, -0.0453, -0.0382], + [ 0.0023, -0.0331, 0.0068, ..., 0.0160, 0.0033, -0.0140], + [-0.0045, -0.0283, -0.0411, ..., 0.0359, -0.0137, -0.0174], + ..., + [-0.0146, 0.0194, 0.0221, ..., 0.0184, -0.0265, -0.0250], + [-0.0339, -0.0055, 0.0097, ..., 0.0365, -0.0140, -0.0205], + [ 0.0134, 0.0313, -0.0092, ..., -0.0567, -0.0399, -0.0075]], + device='cuda:0'), grad: tensor([[ 3.4380e-04, -5.4474e-03, 1.7309e-03, ..., -3.8452e-03, + -1.6413e-03, 2.6798e-04], + [ 6.6910e-03, -5.0211e-04, 4.8332e-03, ..., 7.2403e-03, + -4.1931e-02, 4.4327e-03], + [ 2.1744e-04, -3.9520e-03, 1.5020e-03, ..., -3.0121e-02, + -4.1084e-03, 5.2881e-04], + ..., + [ 2.0504e-03, 5.2299e-03, 3.4218e-03, ..., 6.5308e-03, + 4.1885e-03, -9.5248e-05], + [-6.8626e-03, 1.6190e-02, 3.0556e-03, ..., 1.5213e-02, + 1.2856e-02, -3.2654e-03], + [-8.6498e-04, -2.2354e-02, -7.5951e-03, ..., -1.8024e-03, + 3.0351e-04, -4.0474e-03]], device='cuda:0') +Epoch 32, bias, value: tensor([-0.0307, -0.0032, -0.0051, -0.0063, -0.0090, 0.0059, -0.0072, 0.0219, + -0.0032, 0.0207], device='cuda:0'), grad: tensor([-0.0102, 0.0184, -0.0826, 0.0204, 0.0309, 0.0463, -0.0720, 0.0289, + 0.0504, -0.0307], device='cuda:0') +100 +0.0001 +changing lr +epoch 31, time 214.14, cls_loss 0.9015 cls_loss_mapping 0.0657 cls_loss_causal 0.7572 re_mapping 0.0329 re_causal 0.0747 /// teacc 97.40 lr 0.00010000 +Epoch 33, weight, value: tensor([[-0.0743, -0.0337, -0.0128, ..., 0.0012, -0.0453, -0.0395], + [ 0.0023, -0.0337, 0.0064, ..., 0.0161, 0.0032, -0.0149], + [-0.0051, -0.0302, -0.0412, ..., 0.0358, -0.0137, -0.0183], + ..., + [-0.0150, 0.0194, 0.0227, ..., 0.0185, -0.0271, -0.0262], + [-0.0346, -0.0058, 0.0088, ..., 0.0364, -0.0140, -0.0212], + [ 0.0140, 0.0313, -0.0088, ..., -0.0566, -0.0396, -0.0073]], + device='cuda:0'), grad: tensor([[ 2.1362e-03, 8.8215e-06, 7.1347e-05, ..., -1.7929e-02, + -5.6763e-03, 2.2488e-03], + [ 3.3593e-04, -2.1315e-04, -1.8024e-04, ..., 1.9312e-03, + 1.2989e-03, 3.4070e-04], + [-2.8553e-03, 3.2425e-05, 7.8559e-05, ..., 3.1891e-03, + 1.2123e-02, -6.7329e-04], + ..., + [ 7.3671e-04, -1.2875e-03, 2.2876e-04, ..., -4.6349e-03, + -9.2077e-04, 1.1759e-03], + [ 1.2188e-03, 1.1873e-04, 1.6558e-04, ..., 3.8738e-03, + 2.5291e-03, 2.3117e-03], + [-1.4679e-02, 1.2512e-03, -1.1520e-03, ..., 4.6806e-03, + 2.1057e-03, -1.2543e-02]], device='cuda:0') +Epoch 33, bias, value: tensor([-0.0309, -0.0035, -0.0049, -0.0060, -0.0088, 0.0057, -0.0078, 0.0220, + -0.0034, 0.0213], device='cuda:0'), grad: tensor([-0.0443, 0.0099, 0.0015, -0.0157, 0.0224, 0.0428, -0.0269, -0.0115, + 0.0375, -0.0157], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 214.91, cls_loss 0.8947 cls_loss_mapping 0.0650 cls_loss_causal 0.7649 re_mapping 0.0321 re_causal 0.0722 /// teacc 97.71 lr 0.00010000 +Epoch 34, weight, value: tensor([[-0.0758, -0.0329, -0.0119, ..., 0.0010, -0.0451, -0.0407], + [ 0.0016, -0.0341, 0.0066, ..., 0.0171, 0.0046, -0.0150], + [-0.0061, -0.0296, -0.0402, ..., 0.0349, -0.0153, -0.0195], + ..., + [-0.0141, 0.0191, 0.0219, ..., 0.0181, -0.0278, -0.0255], + [-0.0347, -0.0053, 0.0077, ..., 0.0359, -0.0154, -0.0216], + [ 0.0148, 0.0312, -0.0092, ..., -0.0570, -0.0399, -0.0069]], + device='cuda:0'), grad: tensor([[ 3.2401e-04, -1.2770e-03, -2.1362e-03, ..., -2.0142e-03, + 2.7618e-03, -9.6130e-04], + [ 2.1613e-04, -6.1035e-03, -1.1559e-03, ..., -9.1553e-05, + -5.5695e-03, -1.4620e-03], + [-1.5163e-03, 1.3218e-03, 1.2112e-03, ..., 8.9645e-03, + 6.2447e-03, 1.0653e-03], + ..., + [-2.1229e-03, 2.7905e-03, 1.4315e-03, ..., 1.1833e-02, + 9.2316e-03, 1.8406e-03], + [ 6.3858e-03, 4.0588e-03, 5.6992e-03, ..., 2.8076e-02, + 2.1225e-02, 3.6469e-03], + [ 2.0275e-03, 1.3733e-03, 3.1948e-03, ..., -1.8555e-02, + -1.1200e-02, 8.1635e-04]], device='cuda:0') +Epoch 34, bias, value: tensor([-0.0299, -0.0030, -0.0058, -0.0062, -0.0091, 0.0066, -0.0082, 0.0221, + -0.0036, 0.0209], device='cuda:0'), grad: tensor([-0.0191, -0.0017, 0.0203, -0.0338, 0.0164, -0.0311, -0.0267, 0.0186, + 0.0718, -0.0147], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 215.34, cls_loss 0.9087 cls_loss_mapping 0.0564 cls_loss_causal 0.7747 re_mapping 0.0319 re_causal 0.0707 /// teacc 97.92 lr 0.00010000 +Epoch 35, weight, value: tensor([[-0.0758, -0.0336, -0.0110, ..., -0.0001, -0.0449, -0.0412], + [ 0.0017, -0.0338, 0.0076, ..., 0.0171, 0.0031, -0.0142], + [-0.0065, -0.0286, -0.0398, ..., 0.0360, -0.0158, -0.0189], + ..., + [-0.0144, 0.0196, 0.0212, ..., 0.0181, -0.0268, -0.0262], + [-0.0352, -0.0058, 0.0081, ..., 0.0363, -0.0162, -0.0229], + [ 0.0152, 0.0312, -0.0104, ..., -0.0567, -0.0390, -0.0060]], + device='cuda:0'), grad: tensor([[ 5.1484e-06, 9.1362e-04, 5.5134e-06, ..., -1.6541e-02, + -1.4565e-02, 2.3985e-04], + [ 4.5970e-06, 1.8997e-03, -2.4116e-04, ..., 3.1357e-03, + -1.4055e-04, 1.9467e-04], + [ 1.7717e-05, 3.4447e-03, 2.2743e-06, ..., 7.4081e-03, + 6.9962e-03, 2.0885e-03], + ..., + [ 1.4246e-04, 1.0880e-02, 2.0757e-05, ..., 1.4526e-02, + 6.7978e-03, 2.9736e-03], + [-2.8461e-05, -1.5945e-03, 1.7190e-04, ..., -2.9335e-03, + 1.7433e-03, 9.7275e-04], + [-3.3617e-04, -4.3030e-02, 8.3566e-05, ..., -1.7319e-02, + -1.2375e-02, -3.2787e-03]], device='cuda:0') +Epoch 35, bias, value: tensor([-0.0303, -0.0032, -0.0050, -0.0066, -0.0094, 0.0067, -0.0083, 0.0224, + -0.0037, 0.0212], device='cuda:0'), grad: tensor([-0.0513, 0.0091, 0.0258, -0.0067, 0.0224, 0.0118, 0.0071, 0.0404, + 0.0016, -0.0602], device='cuda:0') +100 +0.0001 +changing lr +epoch 34, time 214.57, cls_loss 0.8723 cls_loss_mapping 0.0598 cls_loss_causal 0.7437 re_mapping 0.0315 re_causal 0.0679 /// teacc 97.69 lr 0.00010000 +Epoch 36, weight, value: tensor([[-0.0776, -0.0336, -0.0108, ..., -0.0005, -0.0440, -0.0423], + [ 0.0018, -0.0334, 0.0073, ..., 0.0167, 0.0029, -0.0141], + [-0.0066, -0.0298, -0.0401, ..., 0.0361, -0.0170, -0.0194], + ..., + [-0.0154, 0.0192, 0.0212, ..., 0.0186, -0.0261, -0.0274], + [-0.0359, -0.0059, 0.0072, ..., 0.0374, -0.0159, -0.0236], + [ 0.0153, 0.0322, -0.0104, ..., -0.0562, -0.0388, -0.0051]], + device='cuda:0'), grad: tensor([[ 9.1434e-05, 1.7872e-03, -5.6496e-03, ..., -2.8706e-04, + -7.8125e-03, 4.0054e-04], + [ 4.3899e-05, 1.8787e-03, 3.1710e-04, ..., 1.6174e-03, + 4.4518e-03, -5.8842e-04], + [ 3.2640e-04, 8.0395e-04, 1.3456e-03, ..., 1.0818e-02, + 7.9956e-03, 5.8937e-04], + ..., + [ 6.4087e-03, 5.4436e-03, 5.3453e-04, ..., 1.9852e-02, + 1.4778e-02, 9.0866e-03], + [-1.5640e-03, -1.4038e-03, 1.9951e-03, ..., 1.7920e-03, + 8.2779e-03, 1.1826e-03], + [-6.8893e-03, -2.6226e-03, -3.4118e-04, ..., -3.7720e-02, + -3.0457e-02, -1.1398e-02]], device='cuda:0') +Epoch 36, bias, value: tensor([-0.0303, -0.0037, -0.0051, -0.0061, -0.0089, 0.0065, -0.0088, 0.0228, + -0.0037, 0.0211], device='cuda:0'), grad: tensor([-0.0203, -0.0092, 0.0351, -0.0063, 0.0001, -0.0126, 0.0356, 0.0639, + 0.0111, -0.0974], device='cuda:0') +100 +0.0001 +changing lr +epoch 35, time 214.40, cls_loss 0.8496 cls_loss_mapping 0.0584 cls_loss_causal 0.7033 re_mapping 0.0310 re_causal 0.0678 /// teacc 97.75 lr 0.00010000 +Epoch 37, weight, value: tensor([[-0.0777, -0.0333, -0.0102, ..., -0.0012, -0.0443, -0.0432], + [ 0.0019, -0.0331, 0.0073, ..., 0.0174, 0.0038, -0.0148], + [-0.0082, -0.0288, -0.0397, ..., 0.0375, -0.0167, -0.0196], + ..., + [-0.0161, 0.0205, 0.0215, ..., 0.0186, -0.0244, -0.0281], + [-0.0362, -0.0064, 0.0064, ..., 0.0369, -0.0176, -0.0244], + [ 0.0154, 0.0308, -0.0111, ..., -0.0566, -0.0385, -0.0054]], + device='cuda:0'), grad: tensor([[-3.9220e-04, -9.6741e-03, -2.8458e-03, ..., -1.2421e-02, + -1.5583e-03, -5.6267e-03], + [ 3.8087e-05, 2.1191e-03, 1.2093e-03, ..., 4.4823e-03, + 4.4479e-03, 2.4014e-03], + [ 1.8239e-04, -9.0790e-04, 3.0804e-03, ..., -2.0615e-02, + -7.3242e-02, -1.3494e-04], + ..., + [ 4.4990e-04, -3.7537e-03, -2.1393e-02, ..., -1.8005e-02, + -1.9470e-02, 2.5034e-04], + [ 3.4142e-04, 2.2831e-03, -3.4924e-03, ..., 2.6131e-03, + -4.7150e-03, 9.4414e-04], + [-5.4550e-03, 4.8180e-03, -1.2589e-04, ..., 1.2085e-02, + 4.8523e-03, -2.2373e-03]], device='cuda:0') +Epoch 37, bias, value: tensor([-0.0305, -0.0035, -0.0043, -0.0057, -0.0093, 0.0063, -0.0092, 0.0228, + -0.0040, 0.0212], device='cuda:0'), grad: tensor([-0.0564, 0.0254, -0.0612, -0.0053, 0.0146, 0.0188, 0.0775, -0.0311, + -0.0041, 0.0218], device='cuda:0') +100 +0.0001 +changing lr +epoch 36, time 214.94, cls_loss 0.8506 cls_loss_mapping 0.0618 cls_loss_causal 0.7219 re_mapping 0.0302 re_causal 0.0670 /// teacc 97.88 lr 0.00010000 +Epoch 38, weight, value: tensor([[-0.0780, -0.0326, -0.0097, ..., -0.0003, -0.0442, -0.0441], + [ 0.0009, -0.0336, 0.0082, ..., 0.0171, 0.0036, -0.0152], + [-0.0087, -0.0300, -0.0415, ..., 0.0367, -0.0161, -0.0206], + ..., + [-0.0172, 0.0192, 0.0215, ..., 0.0182, -0.0250, -0.0284], + [-0.0358, -0.0054, 0.0084, ..., 0.0381, -0.0175, -0.0239], + [ 0.0157, 0.0314, -0.0105, ..., -0.0574, -0.0381, -0.0049]], + device='cuda:0'), grad: tensor([[ 4.7302e-04, 5.3835e-04, 4.5514e-04, ..., 7.5302e-03, + 1.0422e-02, 1.8044e-03], + [ 8.4543e-04, 8.6355e-04, 8.4686e-04, ..., 3.2616e-03, + -1.1864e-02, 4.6110e-04], + [-5.3177e-03, 1.1206e-03, -7.0114e-03, ..., -9.0103e-03, + -4.6883e-03, -4.7646e-03], + ..., + [ 2.8229e-03, -9.0103e-03, 3.8795e-03, ..., -4.0588e-03, + 5.9853e-03, 1.9569e-03], + [ 1.5278e-03, 3.4313e-03, 9.0599e-05, ..., -5.5161e-03, + -1.6251e-02, 2.3670e-03], + [-5.1422e-03, 2.5921e-03, -5.1079e-03, ..., 1.8082e-03, + -1.7075e-02, -3.2749e-03]], device='cuda:0') +Epoch 38, bias, value: tensor([-0.0304, -0.0036, -0.0045, -0.0055, -0.0086, 0.0067, -0.0099, 0.0227, + -0.0037, 0.0205], device='cuda:0'), grad: tensor([ 0.0356, -0.0082, -0.0075, 0.0122, 0.0246, 0.0562, -0.0639, 0.0067, + -0.0458, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 37---------------------------------------------------- +epoch 37, time 215.58, cls_loss 0.8937 cls_loss_mapping 0.0647 cls_loss_causal 0.7608 re_mapping 0.0307 re_causal 0.0665 /// teacc 98.03 lr 0.00010000 +Epoch 39, weight, value: tensor([[-0.0790, -0.0311, -0.0096, ..., -0.0007, -0.0441, -0.0437], + [ 0.0007, -0.0344, 0.0092, ..., 0.0163, 0.0033, -0.0164], + [-0.0078, -0.0310, -0.0417, ..., 0.0362, -0.0159, -0.0203], + ..., + [-0.0182, 0.0178, 0.0227, ..., 0.0191, -0.0251, -0.0284], + [-0.0365, -0.0059, 0.0087, ..., 0.0377, -0.0184, -0.0250], + [ 0.0164, 0.0319, -0.0109, ..., -0.0575, -0.0382, -0.0043]], + device='cuda:0'), grad: tensor([[-9.3889e-04, -2.0210e-07, 5.2881e-04, ..., -6.2408e-03, + -5.3635e-03, -4.1461e-04], + [ 9.4771e-05, 1.6298e-08, 7.5102e-04, ..., -5.2071e-03, + -9.9411e-03, -1.1806e-03], + [ 2.6298e-04, 1.4529e-07, 8.6308e-04, ..., 5.1079e-03, + 1.1833e-02, 2.3460e-03], + ..., + [ 1.0996e-03, 7.9162e-09, 5.1346e-03, ..., 6.6032e-03, + 1.6220e-02, 1.0090e-03], + [ 7.6294e-04, 3.2522e-06, -2.7448e-05, ..., -3.3722e-03, + -5.3139e-03, -5.4026e-04], + [-1.5440e-03, 2.7986e-07, 2.3251e-03, ..., 4.1161e-03, + 9.3842e-03, 9.9087e-04]], device='cuda:0') +Epoch 39, bias, value: tensor([-0.0305, -0.0045, -0.0048, -0.0057, -0.0085, 0.0070, -0.0087, 0.0233, + -0.0041, 0.0203], device='cuda:0'), grad: tensor([-0.0196, -0.0324, 0.0374, -0.0600, -0.0247, 0.0459, 0.0122, 0.0376, + -0.0180, 0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 38, time 214.80, cls_loss 0.8355 cls_loss_mapping 0.0570 cls_loss_causal 0.7181 re_mapping 0.0288 re_causal 0.0638 /// teacc 97.79 lr 0.00010000 +Epoch 40, weight, value: tensor([[-0.0796, -0.0316, -0.0091, ..., -0.0014, -0.0450, -0.0440], + [ 0.0002, -0.0346, 0.0098, ..., 0.0169, 0.0037, -0.0165], + [-0.0069, -0.0317, -0.0433, ..., 0.0358, -0.0165, -0.0194], + ..., + [-0.0181, 0.0194, 0.0219, ..., 0.0185, -0.0258, -0.0281], + [-0.0375, -0.0061, 0.0089, ..., 0.0389, -0.0173, -0.0265], + [ 0.0168, 0.0312, -0.0125, ..., -0.0587, -0.0394, -0.0038]], + device='cuda:0'), grad: tensor([[ 1.0133e-04, -1.5205e-02, 3.0670e-03, ..., -5.7268e-04, + -4.4861e-03, 3.3081e-05], + [ 4.6581e-05, 1.3876e-03, 6.2981e-03, ..., 1.8539e-03, + 4.0054e-03, 2.0400e-05], + [ 8.5413e-05, -6.2485e-03, 3.1209e-04, ..., -9.2621e-03, + -5.2299e-03, 8.0764e-05], + ..., + [ 3.8109e-03, 3.4008e-03, -1.2999e-03, ..., -3.8719e-04, + -1.3380e-03, 1.2636e-03], + [ 5.2309e-04, 1.8444e-03, -1.7029e-02, ..., -1.0239e-02, + -1.6159e-02, 2.4080e-04], + [-5.4359e-03, 3.3588e-03, -3.2120e-03, ..., -1.7605e-03, + 2.1858e-03, -2.3098e-03]], device='cuda:0') +Epoch 40, bias, value: tensor([-0.0309, -0.0046, -0.0041, -0.0059, -0.0078, 0.0064, -0.0090, 0.0230, + -0.0036, 0.0201], device='cuda:0'), grad: tensor([-0.0017, 0.0192, -0.0307, -0.0453, 0.0423, 0.0353, 0.0275, 0.0028, + -0.0344, -0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 39, time 214.72, cls_loss 0.8318 cls_loss_mapping 0.0513 cls_loss_causal 0.6879 re_mapping 0.0300 re_causal 0.0643 /// teacc 97.52 lr 0.00010000 +Epoch 41, weight, value: tensor([[-0.0807, -0.0309, -0.0085, ..., -0.0014, -0.0443, -0.0457], + [-0.0001, -0.0364, 0.0098, ..., 0.0176, 0.0037, -0.0171], + [-0.0077, -0.0336, -0.0435, ..., 0.0349, -0.0169, -0.0212], + ..., + [-0.0190, 0.0197, 0.0209, ..., 0.0182, -0.0265, -0.0291], + [-0.0388, -0.0062, 0.0104, ..., 0.0392, -0.0178, -0.0261], + [ 0.0165, 0.0319, -0.0118, ..., -0.0579, -0.0383, -0.0030]], + device='cuda:0'), grad: tensor([[-0.0019, -0.0070, -0.0051, ..., -0.0231, -0.0124, -0.0002], + [-0.0004, -0.0006, 0.0014, ..., 0.0052, 0.0188, -0.0003], + [-0.0006, 0.0024, -0.0001, ..., -0.0007, 0.0019, 0.0005], + ..., + [-0.0053, -0.0011, -0.0037, ..., -0.0154, -0.0003, -0.0058], + [ 0.0010, -0.0003, 0.0012, ..., 0.0020, -0.0012, 0.0006], + [ 0.0056, 0.0010, 0.0028, ..., 0.0141, 0.0066, 0.0053]], + device='cuda:0') +Epoch 41, bias, value: tensor([-0.0310, -0.0046, -0.0052, -0.0055, -0.0072, 0.0067, -0.0092, 0.0227, + -0.0035, 0.0205], device='cuda:0'), grad: tensor([-0.0572, 0.0244, -0.0007, 0.0166, 0.0295, -0.0233, -0.0094, -0.0342, + 0.0016, 0.0527], device='cuda:0') +100 +0.0001 +changing lr +epoch 40, time 214.58, cls_loss 0.8316 cls_loss_mapping 0.0541 cls_loss_causal 0.6942 re_mapping 0.0294 re_causal 0.0640 /// teacc 97.81 lr 0.00010000 +Epoch 42, weight, value: tensor([[-0.0801, -0.0304, -0.0075, ..., -0.0020, -0.0442, -0.0450], + [ 0.0001, -0.0359, 0.0103, ..., 0.0176, 0.0033, -0.0181], + [-0.0080, -0.0346, -0.0442, ..., 0.0342, -0.0173, -0.0214], + ..., + [-0.0200, 0.0197, 0.0213, ..., 0.0182, -0.0259, -0.0300], + [-0.0403, -0.0060, 0.0102, ..., 0.0405, -0.0177, -0.0266], + [ 0.0174, 0.0318, -0.0128, ..., -0.0593, -0.0382, -0.0022]], + device='cuda:0'), grad: tensor([[ 1.6460e-03, 4.1847e-03, 2.4300e-03, ..., 4.2267e-03, + 4.3068e-03, 1.2350e-03], + [ 2.3997e-04, 7.8726e-04, 1.1711e-03, ..., 1.0918e-02, + 6.6986e-03, 3.3879e-04], + [ 8.4925e-04, 8.6784e-04, -8.4610e-03, ..., 6.3782e-03, + 3.8719e-03, 6.9141e-04], + ..., + [ 2.2831e-03, -7.8917e-04, 3.5324e-03, ..., -4.3671e-02, + -3.1586e-02, 1.8291e-03], + [-2.8133e-03, 6.7234e-04, -6.3744e-03, ..., 3.3531e-03, + 4.6806e-03, -3.5820e-03], + [ 1.2684e-03, 5.3253e-03, 7.2784e-03, ..., 8.6060e-03, + 4.1580e-03, -2.7061e-05]], device='cuda:0') +Epoch 42, bias, value: tensor([-0.0313, -0.0049, -0.0057, -0.0051, -0.0070, 0.0058, -0.0082, 0.0228, + -0.0031, 0.0203], device='cuda:0'), grad: tensor([ 0.0105, 0.0211, -0.0177, -0.0286, -0.0188, 0.0217, -0.0155, -0.0057, + -0.0094, 0.0422], device='cuda:0') +100 +0.0001 +changing lr +epoch 41, time 214.23, cls_loss 0.8361 cls_loss_mapping 0.0553 cls_loss_causal 0.7031 re_mapping 0.0282 re_causal 0.0623 /// teacc 97.90 lr 0.00010000 +Epoch 43, weight, value: tensor([[-0.0806, -0.0312, -0.0076, ..., -0.0024, -0.0443, -0.0456], + [-0.0005, -0.0370, 0.0091, ..., 0.0185, 0.0036, -0.0189], + [-0.0079, -0.0369, -0.0450, ..., 0.0338, -0.0183, -0.0228], + ..., + [-0.0208, 0.0203, 0.0214, ..., 0.0186, -0.0261, -0.0313], + [-0.0395, -0.0047, 0.0099, ..., 0.0402, -0.0179, -0.0264], + [ 0.0179, 0.0321, -0.0127, ..., -0.0603, -0.0373, -0.0006]], + device='cuda:0'), grad: tensor([[ 3.7253e-05, 1.8625e-03, 2.7132e-04, ..., 3.9177e-03, + -1.0735e-02, 8.9121e-04], + [ 5.6058e-05, -6.1264e-03, 9.3365e-04, ..., -3.7498e-03, + -8.5640e-04, -4.7255e-04], + [ 2.6107e-05, -1.3618e-03, 5.9307e-05, ..., 8.3447e-06, + -1.9741e-03, 3.0899e-04], + ..., + [ 1.9431e-05, 5.1498e-03, 2.9802e-04, ..., 4.4289e-03, + 4.5929e-03, 1.0853e-03], + [ 1.3638e-03, -1.3025e-01, -5.6572e-03, ..., -3.4027e-02, + -1.4458e-02, -1.2115e-02], + [-2.5101e-03, 1.1047e-01, 2.1210e-03, ..., 2.6321e-02, + 1.6495e-02, 2.5005e-03]], device='cuda:0') +Epoch 43, bias, value: tensor([-0.0315, -0.0052, -0.0061, -0.0049, -0.0068, 0.0058, -0.0084, 0.0227, + -0.0027, 0.0206], device='cuda:0'), grad: tensor([-1.2527e-02, -1.8570e-02, -1.0848e-05, -9.0637e-03, -2.8305e-02, + 3.2501e-02, 3.5004e-02, 2.4765e-02, -8.9722e-02, 6.5918e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 42, time 214.31, cls_loss 0.8345 cls_loss_mapping 0.0591 cls_loss_causal 0.7259 re_mapping 0.0280 re_causal 0.0613 /// teacc 97.73 lr 0.00010000 +Epoch 44, weight, value: tensor([[-0.0817, -0.0319, -0.0069, ..., -0.0018, -0.0436, -0.0462], + [ 0.0003, -0.0367, 0.0094, ..., 0.0187, 0.0037, -0.0199], + [-0.0083, -0.0372, -0.0458, ..., 0.0337, -0.0177, -0.0240], + ..., + [-0.0211, 0.0200, 0.0219, ..., 0.0191, -0.0266, -0.0318], + [-0.0405, -0.0040, 0.0094, ..., 0.0409, -0.0182, -0.0277], + [ 0.0182, 0.0315, -0.0121, ..., -0.0614, -0.0379, -0.0001]], + device='cuda:0'), grad: tensor([[ 2.9278e-04, 1.6747e-03, 1.1826e-03, ..., 5.8250e-03, + 5.0354e-03, 1.2636e-03], + [ 1.4963e-03, 3.7346e-03, 2.4300e-03, ..., 6.2294e-03, + 3.8757e-03, 2.4815e-03], + [-1.3039e-02, 2.6226e-03, -1.2026e-03, ..., -1.0231e-02, + -1.3218e-03, 7.4267e-05], + ..., + [-3.3226e-03, 1.7258e-02, 1.4067e-03, ..., -3.3665e-03, + 5.8174e-05, -6.2847e-04], + [ 1.3199e-03, 1.3634e-02, 5.8403e-03, ..., 5.9662e-03, + 4.1161e-03, 3.7708e-03], + [ 1.9035e-03, -2.3865e-02, 3.8853e-03, ..., -1.7662e-03, + -6.1321e-04, 5.9605e-06]], device='cuda:0') +Epoch 44, bias, value: tensor([-0.0313, -0.0048, -0.0056, -0.0055, -0.0068, 0.0053, -0.0087, 0.0231, + -0.0026, 0.0204], device='cuda:0'), grad: tensor([ 0.0335, 0.0440, -0.0609, 0.0407, -0.0370, -0.0567, -0.0096, 0.0088, + 0.0566, -0.0194], device='cuda:0') +100 +0.0001 +changing lr +epoch 43, time 214.40, cls_loss 0.8506 cls_loss_mapping 0.0527 cls_loss_causal 0.7285 re_mapping 0.0275 re_causal 0.0620 /// teacc 97.81 lr 0.00010000 +Epoch 45, weight, value: tensor([[-0.0829, -0.0328, -0.0070, ..., -0.0024, -0.0447, -0.0468], + [ 0.0018, -0.0374, 0.0087, ..., 0.0190, 0.0030, -0.0202], + [-0.0093, -0.0382, -0.0455, ..., 0.0347, -0.0172, -0.0250], + ..., + [-0.0216, 0.0203, 0.0229, ..., 0.0191, -0.0269, -0.0313], + [-0.0411, -0.0046, 0.0094, ..., 0.0411, -0.0193, -0.0282], + [ 0.0193, 0.0327, -0.0130, ..., -0.0608, -0.0357, -0.0011]], + device='cuda:0'), grad: tensor([[ 0.0048, -0.0113, 0.0044, ..., -0.0242, -0.0290, -0.0030], + [ 0.0003, -0.0278, -0.0028, ..., -0.0200, -0.0219, -0.0010], + [ 0.0014, 0.0033, -0.0059, ..., 0.0179, 0.0231, -0.0123], + ..., + [ 0.0008, 0.0036, 0.0024, ..., 0.0058, 0.0041, 0.0028], + [-0.0007, 0.0051, -0.0021, ..., -0.0021, 0.0025, 0.0012], + [-0.0147, 0.0058, -0.0151, ..., -0.0090, 0.0013, -0.0121]], + device='cuda:0') +Epoch 45, bias, value: tensor([-0.0317, -0.0048, -0.0051, -0.0060, -0.0078, 0.0054, -0.0077, 0.0228, + -0.0029, 0.0214], device='cuda:0'), grad: tensor([ 0.0045, -0.0580, -0.0140, 0.0431, -0.0068, 0.0568, 0.0535, 0.0255, + -0.0154, -0.0891], device='cuda:0') +100 +0.0001 +changing lr +epoch 44, time 214.13, cls_loss 0.8352 cls_loss_mapping 0.0471 cls_loss_causal 0.7006 re_mapping 0.0282 re_causal 0.0616 /// teacc 97.86 lr 0.00010000 +Epoch 46, weight, value: tensor([[-0.0836, -0.0332, -0.0068, ..., -0.0021, -0.0451, -0.0466], + [ 0.0012, -0.0375, 0.0081, ..., 0.0189, 0.0033, -0.0210], + [-0.0098, -0.0378, -0.0437, ..., 0.0351, -0.0176, -0.0255], + ..., + [-0.0218, 0.0215, 0.0231, ..., 0.0201, -0.0261, -0.0324], + [-0.0428, -0.0046, 0.0089, ..., 0.0413, -0.0197, -0.0302], + [ 0.0215, 0.0320, -0.0122, ..., -0.0622, -0.0362, 0.0003]], + device='cuda:0'), grad: tensor([[-2.4986e-03, 9.3794e-04, 9.6083e-04, ..., 6.7101e-03, + 2.0256e-03, 1.7633e-03], + [ 9.6416e-04, 6.7139e-04, 1.0719e-03, ..., 7.7133e-03, + 4.6234e-03, -4.0169e-03], + [ 5.1842e-03, 3.2787e-03, 4.0779e-03, ..., 4.3182e-03, + 3.1433e-03, 1.5402e-03], + ..., + [ 7.8738e-05, 1.3943e-03, 1.0471e-03, ..., 4.0283e-03, + 1.5831e-03, -1.4985e-04], + [-7.4272e-03, -1.9896e-04, -2.4643e-02, ..., -5.2681e-03, + 3.1605e-03, 8.8215e-04], + [ 6.3515e-03, -4.7073e-03, 8.5144e-03, ..., -2.4261e-03, + -1.8873e-03, 1.6346e-03]], device='cuda:0') +Epoch 46, bias, value: tensor([-0.0312, -0.0054, -0.0043, -0.0059, -0.0079, 0.0059, -0.0083, 0.0229, + -0.0037, 0.0216], device='cuda:0'), grad: tensor([ 0.0244, 0.0043, 0.0356, 0.0140, -0.0122, -0.0553, 0.0098, -0.0173, + -0.0350, 0.0316], device='cuda:0') +100 +0.0001 +changing lr +epoch 45, time 214.12, cls_loss 0.8469 cls_loss_mapping 0.0404 cls_loss_causal 0.7034 re_mapping 0.0279 re_causal 0.0612 /// teacc 97.83 lr 0.00010000 +Epoch 47, weight, value: tensor([[-0.0834, -0.0345, -0.0062, ..., -0.0010, -0.0445, -0.0471], + [ 0.0008, -0.0362, 0.0089, ..., 0.0184, 0.0031, -0.0213], + [-0.0105, -0.0379, -0.0444, ..., 0.0355, -0.0179, -0.0271], + ..., + [-0.0222, 0.0216, 0.0228, ..., 0.0215, -0.0262, -0.0337], + [-0.0428, -0.0060, 0.0090, ..., 0.0421, -0.0197, -0.0311], + [ 0.0210, 0.0324, -0.0119, ..., -0.0631, -0.0358, 0.0010]], + device='cuda:0'), grad: tensor([[ 1.5557e-04, 8.3923e-04, 1.6260e-04, ..., -1.1932e-02, + -2.2163e-03, 3.6573e-04], + [ 6.7472e-04, 7.2289e-03, 5.5275e-03, ..., 3.9406e-03, + 1.1559e-03, 8.2111e-04], + [ 3.6860e-04, 6.2037e-04, 3.2449e-04, ..., -7.7133e-03, + -2.4223e-04, -2.2516e-05], + ..., + [-7.7209e-03, 3.6102e-02, 5.3596e-04, ..., 4.7445e-04, + 3.3436e-03, -2.0618e-03], + [-4.4975e-03, 3.5706e-03, -3.2959e-03, ..., 6.1035e-03, + 1.7891e-03, 5.3444e-03], + [ 1.4397e-02, -5.7526e-02, 4.5891e-03, ..., 2.4170e-02, + -2.5864e-03, 3.7804e-03]], device='cuda:0') +Epoch 47, bias, value: tensor([-0.0312, -0.0057, -0.0039, -0.0061, -0.0075, 0.0057, -0.0086, 0.0231, + -0.0036, 0.0215], device='cuda:0'), grad: tensor([-0.0063, 0.0173, -0.0329, -0.0409, -0.0179, -0.0327, 0.0309, -0.0089, + 0.0205, 0.0709], device='cuda:0') +100 +0.0001 +changing lr +epoch 46, time 214.46, cls_loss 0.8150 cls_loss_mapping 0.0541 cls_loss_causal 0.6965 re_mapping 0.0269 re_causal 0.0595 /// teacc 97.98 lr 0.00010000 +Epoch 48, weight, value: tensor([[-0.0847, -0.0332, -0.0052, ..., -0.0028, -0.0458, -0.0478], + [ 0.0003, -0.0371, 0.0109, ..., 0.0193, 0.0033, -0.0218], + [-0.0105, -0.0369, -0.0464, ..., 0.0348, -0.0186, -0.0280], + ..., + [-0.0234, 0.0208, 0.0224, ..., 0.0225, -0.0263, -0.0344], + [-0.0430, -0.0070, 0.0089, ..., 0.0421, -0.0200, -0.0312], + [ 0.0211, 0.0325, -0.0123, ..., -0.0627, -0.0350, 0.0016]], + device='cuda:0'), grad: tensor([[ 2.1982e-04, 1.8561e-04, -5.8479e-03, ..., -7.6523e-03, + -7.2441e-03, 2.5635e-03], + [-3.0098e-03, 1.4770e-04, 5.7793e-04, ..., -2.0905e-03, + 1.1539e-03, 1.5469e-03], + [ 3.4761e-04, -9.4700e-04, 1.1396e-03, ..., 5.6190e-03, + 1.8873e-03, 2.3918e-03], + ..., + [ 3.9029e-04, -3.6144e-03, 1.2560e-03, ..., -4.4746e-03, + 1.9197e-03, -2.8849e-04], + [ 2.8152e-03, 8.3876e-04, 5.4741e-04, ..., -6.0463e-03, + -3.6163e-03, 3.2330e-04], + [ 9.8109e-05, 1.5717e-03, 1.8730e-03, ..., -2.9640e-03, + -1.3018e-03, -2.1744e-03]], device='cuda:0') +Epoch 48, bias, value: tensor([-0.0320, -0.0053, -0.0049, -0.0058, -0.0070, 0.0049, -0.0082, 0.0229, + -0.0031, 0.0221], device='cuda:0'), grad: tensor([-0.0045, 0.0101, 0.0153, -0.0215, -0.0048, -0.0107, 0.0512, -0.0055, + -0.0188, -0.0108], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 47---------------------------------------------------- +epoch 47, time 214.59, cls_loss 0.8129 cls_loss_mapping 0.0470 cls_loss_causal 0.6799 re_mapping 0.0281 re_causal 0.0602 /// teacc 98.10 lr 0.00010000 +Epoch 49, weight, value: tensor([[-0.0860, -0.0326, -0.0047, ..., -0.0040, -0.0467, -0.0483], + [ 0.0004, -0.0376, 0.0111, ..., 0.0202, 0.0033, -0.0220], + [-0.0098, -0.0364, -0.0469, ..., 0.0334, -0.0194, -0.0288], + ..., + [-0.0245, 0.0208, 0.0234, ..., 0.0233, -0.0262, -0.0355], + [-0.0436, -0.0065, 0.0091, ..., 0.0422, -0.0207, -0.0320], + [ 0.0214, 0.0327, -0.0124, ..., -0.0631, -0.0345, 0.0018]], + device='cuda:0'), grad: tensor([[-4.8089e-04, -2.3019e-04, 6.3848e-04, ..., 5.1117e-03, + -1.3542e-03, 1.7004e-03], + [-1.0490e-03, -3.2043e-03, 4.1080e-04, ..., 8.3542e-04, + 3.7217e-04, 8.4352e-04], + [ 4.9496e-04, -1.3151e-03, 5.0163e-04, ..., 2.1725e-03, + -6.9022e-05, 8.9741e-04], + ..., + [ 4.6587e-04, 4.5586e-03, -5.0049e-03, ..., -1.1368e-02, + -2.6340e-03, -4.5128e-03], + [ 1.0595e-03, 4.1428e-03, 1.6508e-03, ..., 7.1640e-03, + 1.1892e-03, 2.4834e-03], + [ 1.1158e-04, -1.0597e-02, 1.4534e-03, ..., -8.6670e-03, + -3.8033e-03, -2.7618e-03]], device='cuda:0') +Epoch 49, bias, value: tensor([-0.0326, -0.0059, -0.0051, -0.0057, -0.0065, 0.0051, -0.0082, 0.0235, + -0.0030, 0.0221], device='cuda:0'), grad: tensor([-0.0047, 0.0017, -0.0173, 0.0603, -0.0050, 0.0110, 0.0112, -0.0362, + 0.0277, -0.0487], device='cuda:0') +100 +0.0001 +changing lr +epoch 48, time 214.43, cls_loss 0.8233 cls_loss_mapping 0.0492 cls_loss_causal 0.7011 re_mapping 0.0268 re_causal 0.0577 /// teacc 98.07 lr 0.00010000 +Epoch 50, weight, value: tensor([[-0.0875, -0.0327, -0.0037, ..., -0.0032, -0.0463, -0.0493], + [ 0.0002, -0.0368, 0.0112, ..., 0.0200, 0.0037, -0.0227], + [-0.0103, -0.0358, -0.0476, ..., 0.0331, -0.0199, -0.0292], + ..., + [-0.0244, 0.0204, 0.0229, ..., 0.0230, -0.0264, -0.0361], + [-0.0432, -0.0061, 0.0091, ..., 0.0415, -0.0219, -0.0328], + [ 0.0209, 0.0327, -0.0120, ..., -0.0632, -0.0341, 0.0017]], + device='cuda:0'), grad: tensor([[ 2.6798e-04, 2.7344e-02, 2.4338e-03, ..., 8.8882e-03, + -1.7273e-02, 3.6979e-04], + [ 2.4259e-05, 6.3705e-04, 3.0231e-04, ..., -4.6577e-03, + -5.3024e-03, 1.2010e-04], + [ 1.3542e-04, -2.5749e-03, 1.5087e-03, ..., -7.1640e-03, + 2.1763e-03, -5.9280e-03], + ..., + [ 4.1723e-04, 1.2970e-03, 3.7308e-03, ..., 2.7008e-03, + 1.9875e-03, 3.1090e-04], + [ 2.4283e-04, -1.5554e-03, 2.4643e-03, ..., -3.9215e-03, + 8.4076e-03, 1.0014e-03], + [-7.6962e-04, 1.3023e-02, -4.9973e-03, ..., -1.4448e-03, + 2.3308e-03, 9.7418e-04]], device='cuda:0') +Epoch 50, bias, value: tensor([-0.0325, -0.0064, -0.0049, -0.0055, -0.0069, 0.0054, -0.0086, 0.0239, + -0.0028, 0.0218], device='cuda:0'), grad: tensor([ 0.0400, -0.0173, -0.0263, -0.0039, -0.0309, 0.0224, -0.0040, 0.0287, + 0.0004, -0.0091], device='cuda:0') +100 +0.0001 +changing lr +epoch 49, time 214.38, cls_loss 0.8170 cls_loss_mapping 0.0397 cls_loss_causal 0.6892 re_mapping 0.0256 re_causal 0.0560 /// teacc 98.05 lr 0.00010000 +Epoch 51, weight, value: tensor([[-0.0880, -0.0334, -0.0031, ..., -0.0035, -0.0462, -0.0492], + [ 0.0007, -0.0373, 0.0106, ..., 0.0212, 0.0038, -0.0232], + [-0.0102, -0.0359, -0.0475, ..., 0.0334, -0.0203, -0.0296], + ..., + [-0.0250, 0.0208, 0.0222, ..., 0.0226, -0.0272, -0.0371], + [-0.0432, -0.0063, 0.0086, ..., 0.0417, -0.0218, -0.0332], + [ 0.0198, 0.0323, -0.0137, ..., -0.0634, -0.0333, 0.0020]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0093, 0.0001, ..., 0.0058, 0.0020, 0.0016], + [ 0.0005, 0.0009, 0.0003, ..., 0.0097, 0.0035, 0.0007], + [ 0.0025, 0.0020, 0.0007, ..., 0.0007, -0.0028, 0.0036], + ..., + [-0.0006, 0.0014, 0.0005, ..., 0.0028, 0.0012, 0.0019], + [-0.0022, -0.0153, -0.0009, ..., -0.0057, -0.0006, -0.0017], + [ 0.0029, -0.0047, 0.0022, ..., 0.0004, 0.0025, -0.0008]], + device='cuda:0') +Epoch 51, bias, value: tensor([-0.0325, -0.0057, -0.0047, -0.0059, -0.0067, 0.0055, -0.0086, 0.0237, + -0.0030, 0.0214], device='cuda:0'), grad: tensor([ 0.0234, 0.0185, 0.0013, -0.0060, -0.0055, 0.0224, -0.0353, 0.0118, + -0.0395, 0.0088], device='cuda:0') +100 +0.0001 +changing lr +epoch 50, time 214.57, cls_loss 0.7946 cls_loss_mapping 0.0390 cls_loss_causal 0.6636 re_mapping 0.0256 re_causal 0.0575 /// teacc 97.64 lr 0.00010000 +Epoch 52, weight, value: tensor([[-0.0888, -0.0334, -0.0033, ..., -0.0046, -0.0468, -0.0504], + [ 0.0004, -0.0382, 0.0123, ..., 0.0211, 0.0042, -0.0234], + [-0.0106, -0.0371, -0.0484, ..., 0.0329, -0.0215, -0.0304], + ..., + [-0.0247, 0.0212, 0.0226, ..., 0.0232, -0.0268, -0.0383], + [-0.0443, -0.0060, 0.0081, ..., 0.0414, -0.0215, -0.0335], + [ 0.0199, 0.0325, -0.0131, ..., -0.0630, -0.0324, 0.0030]], + device='cuda:0'), grad: tensor([[ 1.0080e-03, 2.2449e-03, 1.4668e-03, ..., -1.9779e-03, + -4.7989e-03, 2.7084e-04], + [ 1.7214e-03, -3.8544e-02, -3.6031e-05, ..., 4.8470e-04, + 4.6120e-03, 1.1647e-04], + [-4.2343e-03, -3.1490e-03, -1.7023e-04, ..., -7.0953e-03, + 2.4796e-05, 6.7568e-04], + ..., + [-4.1084e-03, -2.2964e-03, 6.0606e-04, ..., -1.6281e-02, + -2.3865e-02, -9.2010e-03], + [ 2.5005e-03, 7.8049e-03, 8.5545e-04, ..., -2.2171e-02, + -1.1429e-02, 1.6136e-03], + [ 2.0523e-03, 2.1103e-02, 6.0425e-03, ..., 1.0017e-02, + 1.0757e-02, 9.4681e-03]], device='cuda:0') +Epoch 52, bias, value: tensor([-0.0332, -0.0055, -0.0052, -0.0057, -0.0074, 0.0058, -0.0085, 0.0245, + -0.0029, 0.0214], device='cuda:0'), grad: tensor([-0.0018, -0.0228, -0.0305, 0.0493, 0.0329, -0.0140, -0.0288, -0.0645, + -0.0047, 0.0848], device='cuda:0') +100 +0.0001 +changing lr +epoch 51, time 214.43, cls_loss 0.8064 cls_loss_mapping 0.0418 cls_loss_causal 0.6848 re_mapping 0.0254 re_causal 0.0569 /// teacc 97.99 lr 0.00010000 +Epoch 53, weight, value: tensor([[-0.0890, -0.0340, -0.0027, ..., -0.0048, -0.0469, -0.0507], + [ 0.0005, -0.0381, 0.0129, ..., 0.0217, 0.0043, -0.0251], + [-0.0110, -0.0371, -0.0498, ..., 0.0333, -0.0221, -0.0315], + ..., + [-0.0258, 0.0215, 0.0230, ..., 0.0225, -0.0263, -0.0392], + [-0.0448, -0.0057, 0.0071, ..., 0.0416, -0.0217, -0.0350], + [ 0.0217, 0.0326, -0.0128, ..., -0.0634, -0.0319, 0.0047]], + device='cuda:0'), grad: tensor([[ 1.4591e-04, 3.6716e-04, 1.4009e-03, ..., 5.0964e-03, + -3.3661e-02, 1.3447e-04], + [ 1.9038e-04, -1.1349e-03, 2.2717e-03, ..., -9.0027e-03, + -2.1515e-03, 7.1168e-05], + [ 1.4889e-04, -7.9880e-03, 3.1352e-04, ..., 5.6686e-03, + 5.7564e-03, 6.9427e-04], + ..., + [-5.0604e-05, 5.8556e-04, 6.8092e-04, ..., 5.7106e-03, + 6.9847e-03, 9.5427e-05], + [ 8.3876e-04, -3.0689e-03, -6.1722e-03, ..., -2.6302e-03, + -6.5041e-04, -2.7809e-03], + [ 3.0947e-04, 5.9891e-04, 5.6419e-03, ..., 5.4512e-03, + 1.9882e-02, 7.1526e-04]], device='cuda:0') +Epoch 53, bias, value: tensor([-0.0334, -0.0050, -0.0053, -0.0059, -0.0077, 0.0056, -0.0081, 0.0243, + -0.0025, 0.0214], device='cuda:0'), grad: tensor([ 0.0094, -0.0452, 0.0005, 0.0062, -0.0186, 0.0094, -0.0168, 0.0328, + -0.0213, 0.0436], device='cuda:0') +100 +0.0001 +changing lr +epoch 52, time 214.53, cls_loss 0.7726 cls_loss_mapping 0.0404 cls_loss_causal 0.6513 re_mapping 0.0252 re_causal 0.0558 /// teacc 97.88 lr 0.00010000 +Epoch 54, weight, value: tensor([[-8.8468e-02, -3.3681e-02, -2.7028e-03, ..., -4.1885e-03, + -4.5839e-02, -4.9988e-02], + [ 6.9997e-05, -3.8439e-02, 1.1607e-02, ..., 2.1339e-02, + 3.5160e-03, -2.5470e-02], + [-1.0773e-02, -3.6684e-02, -4.9982e-02, ..., 3.3299e-02, + -2.2985e-02, -3.2814e-02], + ..., + [-2.5811e-02, 2.2059e-02, 2.1377e-02, ..., 2.2945e-02, + -2.7185e-02, -3.9848e-02], + [-4.4911e-02, -5.4660e-03, 7.2780e-03, ..., 4.1432e-02, + -2.2396e-02, -3.6457e-02], + [ 2.2130e-02, 3.2353e-02, -1.3004e-02, ..., -6.3769e-02, + -3.1147e-02, 5.7734e-03]], device='cuda:0'), grad: tensor([[ 2.2662e-04, -7.6437e-04, 3.0398e-04, ..., -8.5602e-03, + -9.6588e-03, 3.1590e-04], + [ 3.3569e-04, -1.5459e-03, -1.3771e-03, ..., -2.1210e-02, + -9.6970e-03, -1.2894e-03], + [ 2.1231e-04, 1.0712e-02, 1.0519e-03, ..., 1.9211e-02, + 1.1131e-02, 1.5962e-04], + ..., + [ 2.7442e-04, -6.8970e-03, 2.1610e-03, ..., 8.3113e-04, + -2.7084e-03, 1.9670e-04], + [ 2.4274e-05, 3.8433e-03, 1.2846e-03, ..., 8.9264e-03, + 6.7368e-03, 2.1820e-03], + [ 2.7905e-03, -1.1642e-02, -3.5267e-03, ..., -1.6251e-02, + 1.2980e-03, 3.6316e-03]], device='cuda:0') +Epoch 54, bias, value: tensor([-0.0328, -0.0055, -0.0052, -0.0064, -0.0066, 0.0057, -0.0080, 0.0239, + -0.0029, 0.0214], device='cuda:0'), grad: tensor([-0.0087, -0.0604, 0.0672, -0.0056, -0.0033, -0.0065, 0.0342, 0.0003, + 0.0103, -0.0274], device='cuda:0') +100 +0.0001 +changing lr +epoch 53, time 214.59, cls_loss 0.7603 cls_loss_mapping 0.0368 cls_loss_causal 0.6423 re_mapping 0.0259 re_causal 0.0546 /// teacc 97.96 lr 0.00010000 +Epoch 55, weight, value: tensor([[-0.0890, -0.0338, -0.0034, ..., -0.0045, -0.0458, -0.0500], + [ 0.0004, -0.0381, 0.0123, ..., 0.0204, 0.0036, -0.0256], + [-0.0108, -0.0370, -0.0496, ..., 0.0326, -0.0236, -0.0348], + ..., + [-0.0261, 0.0213, 0.0223, ..., 0.0229, -0.0276, -0.0396], + [-0.0444, -0.0055, 0.0074, ..., 0.0419, -0.0225, -0.0365], + [ 0.0223, 0.0335, -0.0135, ..., -0.0640, -0.0299, 0.0047]], + device='cuda:0'), grad: tensor([[ 3.0351e-04, 1.3673e-04, 6.2799e-04, ..., 7.6580e-04, + 1.5259e-03, 2.3842e-04], + [ 4.7946e-04, -1.2789e-03, 1.6155e-03, ..., 1.6332e-04, + 1.7128e-03, 3.4189e-04], + [ 5.6839e-04, -1.6241e-03, 3.7694e-04, ..., -4.9744e-03, + -2.8572e-03, -7.4005e-03], + ..., + [ 3.1109e-03, -5.4264e-04, 1.3790e-03, ..., 9.3699e-05, + 6.2180e-04, 2.4261e-03], + [ 7.1049e-04, 1.5182e-03, 7.1621e-04, ..., 1.5097e-03, + 3.3035e-03, 1.1358e-03], + [ 1.2274e-03, 1.1247e-04, 2.2335e-03, ..., 1.5907e-03, + 2.9545e-03, 1.2674e-03]], device='cuda:0') +Epoch 55, bias, value: tensor([-0.0331, -0.0054, -0.0057, -0.0062, -0.0066, 0.0053, -0.0074, 0.0240, + -0.0025, 0.0212], device='cuda:0'), grad: tensor([-0.0090, 0.0088, -0.0431, -0.0127, -0.0019, -0.0073, 0.0102, 0.0156, + 0.0202, 0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 54, time 214.47, cls_loss 0.7830 cls_loss_mapping 0.0383 cls_loss_causal 0.6575 re_mapping 0.0249 re_causal 0.0555 /// teacc 97.76 lr 0.00010000 +Epoch 56, weight, value: tensor([[-0.0900, -0.0346, -0.0026, ..., -0.0040, -0.0436, -0.0500], + [-0.0002, -0.0373, 0.0115, ..., 0.0203, 0.0039, -0.0254], + [-0.0107, -0.0376, -0.0505, ..., 0.0324, -0.0247, -0.0362], + ..., + [-0.0265, 0.0219, 0.0225, ..., 0.0232, -0.0278, -0.0407], + [-0.0461, -0.0060, 0.0083, ..., 0.0424, -0.0229, -0.0364], + [ 0.0227, 0.0337, -0.0141, ..., -0.0652, -0.0306, 0.0063]], + device='cuda:0'), grad: tensor([[ 0.0013, 0.0063, 0.0012, ..., 0.0022, 0.0030, 0.0020], + [ 0.0047, 0.0155, -0.0019, ..., -0.0110, -0.0011, 0.0072], + [ 0.0040, 0.0019, 0.0019, ..., 0.0044, 0.0030, 0.0011], + ..., + [ 0.0030, 0.0085, 0.0036, ..., 0.0020, 0.0077, 0.0015], + [-0.0136, 0.0023, -0.0013, ..., -0.0069, 0.0023, 0.0082], + [-0.0076, -0.0453, 0.0004, ..., 0.0037, 0.0043, -0.0057]], + device='cuda:0') +Epoch 56, bias, value: tensor([-0.0328, -0.0053, -0.0055, -0.0058, -0.0063, 0.0055, -0.0079, 0.0237, + -0.0025, 0.0205], device='cuda:0'), grad: tensor([ 0.0186, 0.0350, 0.0211, -0.0557, 0.0060, -0.0047, -0.0177, 0.0283, + 0.0041, -0.0351], device='cuda:0') +100 +0.0001 +changing lr +epoch 55, time 214.51, cls_loss 0.8051 cls_loss_mapping 0.0350 cls_loss_causal 0.6857 re_mapping 0.0249 re_causal 0.0568 /// teacc 97.78 lr 0.00010000 +Epoch 57, weight, value: tensor([[-0.0889, -0.0341, -0.0033, ..., -0.0044, -0.0437, -0.0502], + [-0.0010, -0.0378, 0.0123, ..., 0.0202, 0.0036, -0.0265], + [-0.0125, -0.0378, -0.0506, ..., 0.0337, -0.0242, -0.0365], + ..., + [-0.0273, 0.0214, 0.0235, ..., 0.0230, -0.0277, -0.0413], + [-0.0455, -0.0058, 0.0083, ..., 0.0428, -0.0230, -0.0384], + [ 0.0236, 0.0340, -0.0148, ..., -0.0653, -0.0310, 0.0077]], + device='cuda:0'), grad: tensor([[ 0.0052, -0.0012, 0.0006, ..., 0.0026, 0.0005, 0.0013], + [ 0.0064, -0.0004, 0.0022, ..., -0.0022, -0.0003, -0.0018], + [ 0.0024, 0.0008, 0.0013, ..., 0.0034, 0.0010, 0.0018], + ..., + [ 0.0002, -0.0118, -0.0066, ..., -0.0036, -0.0090, 0.0013], + [-0.0104, -0.0013, -0.0109, ..., 0.0014, 0.0012, -0.0062], + [ 0.0148, 0.0085, 0.0127, ..., 0.0094, 0.0020, 0.0072]], + device='cuda:0') +Epoch 57, bias, value: tensor([-0.0331, -0.0054, -0.0047, -0.0066, -0.0056, 0.0057, -0.0079, 0.0237, + -0.0031, 0.0207], device='cuda:0'), grad: tensor([-0.0099, 0.0100, -0.0044, -0.0481, 0.0069, -0.0243, 0.0095, 0.0034, + -0.0149, 0.0718], device='cuda:0') +100 +0.0001 +changing lr +epoch 56, time 214.45, cls_loss 0.7843 cls_loss_mapping 0.0334 cls_loss_causal 0.6632 re_mapping 0.0246 re_causal 0.0553 /// teacc 98.01 lr 0.00010000 +Epoch 58, weight, value: tensor([[-0.0904, -0.0335, -0.0039, ..., -0.0046, -0.0444, -0.0506], + [-0.0023, -0.0382, 0.0121, ..., 0.0207, 0.0038, -0.0270], + [-0.0118, -0.0386, -0.0509, ..., 0.0332, -0.0246, -0.0378], + ..., + [-0.0273, 0.0218, 0.0252, ..., 0.0219, -0.0278, -0.0419], + [-0.0459, -0.0058, 0.0093, ..., 0.0431, -0.0225, -0.0391], + [ 0.0243, 0.0338, -0.0160, ..., -0.0661, -0.0314, 0.0093]], + device='cuda:0'), grad: tensor([[ 0.0012, 0.0002, 0.0017, ..., 0.0042, 0.0014, 0.0004], + [ 0.0002, 0.0004, 0.0005, ..., 0.0057, 0.0013, 0.0002], + [ 0.0012, 0.0007, 0.0002, ..., 0.0009, 0.0013, 0.0014], + ..., + [ 0.0009, 0.0022, 0.0028, ..., -0.0050, -0.0011, 0.0003], + [ 0.0118, -0.0297, 0.0023, ..., -0.0331, -0.0026, 0.0037], + [-0.0138, 0.0106, -0.0105, ..., 0.0089, -0.0023, -0.0050]], + device='cuda:0') +Epoch 58, bias, value: tensor([-0.0333, -0.0048, -0.0050, -0.0069, -0.0053, 0.0062, -0.0078, 0.0237, + -0.0033, 0.0200], device='cuda:0'), grad: tensor([ 0.0278, 0.0327, 0.0107, -0.0801, 0.0414, 0.0250, -0.0066, 0.0126, + 0.0048, -0.0684], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 57---------------------------------------------------- +epoch 57, time 215.08, cls_loss 0.7821 cls_loss_mapping 0.0385 cls_loss_causal 0.6566 re_mapping 0.0242 re_causal 0.0518 /// teacc 98.22 lr 0.00010000 +Epoch 59, weight, value: tensor([[-0.0912, -0.0337, -0.0032, ..., -0.0048, -0.0447, -0.0521], + [-0.0020, -0.0392, 0.0136, ..., 0.0207, 0.0044, -0.0267], + [-0.0121, -0.0385, -0.0519, ..., 0.0339, -0.0242, -0.0384], + ..., + [-0.0273, 0.0226, 0.0253, ..., 0.0226, -0.0282, -0.0429], + [-0.0453, -0.0064, 0.0098, ..., 0.0435, -0.0221, -0.0397], + [ 0.0242, 0.0337, -0.0166, ..., -0.0677, -0.0314, 0.0091]], + device='cuda:0'), grad: tensor([[ 2.4772e-04, 2.3727e-03, 3.8087e-05, ..., 5.8746e-03, + 1.1749e-03, 7.3385e-04], + [ 1.2094e-04, -2.5520e-03, -1.4532e-04, ..., 8.9836e-04, + -3.8567e-03, -2.3329e-04], + [ 2.1768e-04, 2.6550e-03, 2.8238e-05, ..., 5.6953e-03, + -4.0531e-04, 1.0751e-05], + ..., + [ 2.6584e-04, -1.6754e-02, 2.0429e-05, ..., -1.5297e-02, + -5.5084e-03, -4.9362e-03], + [ 8.8501e-04, 2.3613e-03, 6.9201e-05, ..., -1.9119e-02, + 2.5597e-03, 1.5631e-03], + [ 2.6241e-05, 3.1986e-03, 1.8612e-05, ..., 3.1128e-03, + 1.2722e-03, 1.3361e-03]], device='cuda:0') +Epoch 59, bias, value: tensor([-0.0336, -0.0042, -0.0044, -0.0069, -0.0048, 0.0063, -0.0089, 0.0241, + -0.0039, 0.0200], device='cuda:0'), grad: tensor([ 0.0289, -0.0181, 0.0252, 0.0406, 0.0155, 0.0132, -0.0375, -0.0382, + -0.0257, -0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 58, time 214.23, cls_loss 0.7289 cls_loss_mapping 0.0306 cls_loss_causal 0.6184 re_mapping 0.0238 re_causal 0.0519 /// teacc 98.13 lr 0.00010000 +Epoch 60, weight, value: tensor([[-0.0912, -0.0333, -0.0038, ..., -0.0053, -0.0444, -0.0537], + [-0.0026, -0.0392, 0.0139, ..., 0.0210, 0.0041, -0.0278], + [-0.0131, -0.0384, -0.0526, ..., 0.0348, -0.0245, -0.0393], + ..., + [-0.0265, 0.0224, 0.0249, ..., 0.0222, -0.0286, -0.0415], + [-0.0452, -0.0068, 0.0103, ..., 0.0436, -0.0225, -0.0402], + [ 0.0246, 0.0341, -0.0165, ..., -0.0682, -0.0305, 0.0097]], + device='cuda:0'), grad: tensor([[ 0.0006, 0.0007, 0.0001, ..., 0.0092, 0.0007, 0.0004], + [ 0.0014, 0.0022, 0.0002, ..., 0.0090, 0.0009, 0.0008], + [ 0.0015, -0.0001, 0.0003, ..., 0.0062, -0.0016, 0.0002], + ..., + [ 0.0001, 0.0162, 0.0005, ..., 0.0106, 0.0017, 0.0029], + [ 0.0013, 0.0011, -0.0013, ..., 0.0037, 0.0011, -0.0022], + [-0.0021, 0.0180, -0.0003, ..., -0.0087, -0.0025, -0.0007]], + device='cuda:0') +Epoch 60, bias, value: tensor([-0.0344, -0.0043, -0.0041, -0.0066, -0.0053, 0.0063, -0.0092, 0.0244, + -0.0034, 0.0202], device='cuda:0'), grad: tensor([ 0.0220, 0.0327, 0.0075, -0.0403, -0.1123, 0.0340, 0.0195, 0.0390, + 0.0164, -0.0185], device='cuda:0') +100 +0.0001 +changing lr +epoch 59, time 214.30, cls_loss 0.7548 cls_loss_mapping 0.0394 cls_loss_causal 0.6445 re_mapping 0.0237 re_causal 0.0520 /// teacc 98.12 lr 0.00010000 +Epoch 61, weight, value: tensor([[-0.0911, -0.0331, -0.0042, ..., -0.0049, -0.0447, -0.0541], + [-0.0033, -0.0391, 0.0140, ..., 0.0216, 0.0044, -0.0284], + [-0.0134, -0.0387, -0.0532, ..., 0.0347, -0.0242, -0.0395], + ..., + [-0.0261, 0.0212, 0.0252, ..., 0.0224, -0.0285, -0.0417], + [-0.0459, -0.0064, 0.0114, ..., 0.0439, -0.0231, -0.0410], + [ 0.0248, 0.0345, -0.0165, ..., -0.0668, -0.0288, 0.0098]], + device='cuda:0'), grad: tensor([[-7.3242e-03, 4.5700e-03, -7.8278e-03, ..., 6.3858e-03, + 2.1038e-03, 2.8715e-05], + [-2.7828e-03, 8.2684e-04, -1.6956e-03, ..., -6.5308e-03, + 1.0658e-02, 7.5877e-05], + [ 4.3750e-04, 7.1411e-03, 1.2350e-04, ..., 8.6670e-03, + 7.4339e-04, 6.6757e-05], + ..., + [ 8.3876e-04, -1.8520e-03, 5.0831e-04, ..., -3.1319e-03, + 1.1644e-03, 1.4186e-04], + [ 1.9855e-03, -1.8204e-02, 1.1692e-03, ..., -7.6370e-03, + -1.4702e-02, 1.2732e-04], + [ 6.8903e-04, 1.3893e-02, 3.2845e-03, ..., 9.5749e-03, + 5.4693e-04, 7.5674e-04]], device='cuda:0') +Epoch 61, bias, value: tensor([-0.0341, -0.0037, -0.0039, -0.0071, -0.0064, 0.0058, -0.0092, 0.0244, + -0.0037, 0.0214], device='cuda:0'), grad: tensor([ 0.0159, -0.0295, 0.0248, 0.0034, -0.0136, -0.0361, 0.0132, -0.0096, + -0.0078, 0.0393], device='cuda:0') +100 +0.0001 +changing lr +epoch 60, time 220.81, cls_loss 0.7612 cls_loss_mapping 0.0345 cls_loss_causal 0.6546 re_mapping 0.0236 re_causal 0.0535 /// teacc 97.93 lr 0.00010000 +Epoch 62, weight, value: tensor([[-0.0918, -0.0323, -0.0044, ..., -0.0035, -0.0437, -0.0554], + [-0.0028, -0.0376, 0.0145, ..., 0.0213, 0.0043, -0.0293], + [-0.0147, -0.0395, -0.0534, ..., 0.0354, -0.0248, -0.0405], + ..., + [-0.0256, 0.0218, 0.0254, ..., 0.0228, -0.0288, -0.0419], + [-0.0463, -0.0066, 0.0118, ..., 0.0432, -0.0231, -0.0426], + [ 0.0241, 0.0350, -0.0176, ..., -0.0673, -0.0293, 0.0114]], + device='cuda:0'), grad: tensor([[ 4.5204e-04, 1.2374e-04, 6.2180e-04, ..., 4.7541e-04, + 1.1748e-04, 2.0170e-04], + [ 6.1798e-04, 3.7730e-05, 6.8998e-04, ..., -7.3204e-03, + -2.4796e-03, 3.7217e-04], + [ 2.3961e-04, 1.1802e-04, 3.3307e-04, ..., 7.1430e-04, + 1.2016e-04, 2.7966e-04], + ..., + [ 5.8222e-04, -2.0981e-03, 6.5517e-04, ..., -5.1403e-04, + 1.4937e-04, 4.1795e-04], + [-4.2038e-03, -4.2939e-04, -6.4392e-03, ..., -2.8114e-03, + 1.6165e-03, -5.9853e-03], + [ 1.0624e-03, 2.9583e-03, 3.3398e-03, ..., 2.7409e-03, + 1.2660e-04, 1.5764e-03]], device='cuda:0') +Epoch 62, bias, value: tensor([-0.0339, -0.0042, -0.0035, -0.0066, -0.0063, 0.0058, -0.0094, 0.0241, + -0.0039, 0.0214], device='cuda:0'), grad: tensor([ 0.0063, -0.0310, 0.0052, 0.0171, -0.0114, 0.0171, -0.0043, 0.0076, + -0.0253, 0.0186], device='cuda:0') +100 +0.0001 +changing lr +epoch 61, time 216.06, cls_loss 0.7601 cls_loss_mapping 0.0276 cls_loss_causal 0.6473 re_mapping 0.0227 re_causal 0.0501 /// teacc 98.08 lr 0.00010000 +Epoch 63, weight, value: tensor([[-0.0923, -0.0325, -0.0044, ..., -0.0037, -0.0433, -0.0553], + [-0.0021, -0.0368, 0.0149, ..., 0.0209, 0.0041, -0.0289], + [-0.0149, -0.0389, -0.0545, ..., 0.0349, -0.0247, -0.0393], + ..., + [-0.0265, 0.0222, 0.0253, ..., 0.0230, -0.0300, -0.0424], + [-0.0474, -0.0075, 0.0121, ..., 0.0436, -0.0222, -0.0437], + [ 0.0245, 0.0343, -0.0181, ..., -0.0665, -0.0275, 0.0109]], + device='cuda:0'), grad: tensor([[-2.2125e-03, 2.6150e-03, -2.5845e-03, ..., 2.6464e-05, + 2.0695e-03, 4.0317e-04], + [ 5.5981e-04, -1.2722e-03, 2.2278e-03, ..., 1.1606e-03, + 8.8978e-04, 7.6771e-04], + [-3.4409e-03, 9.5701e-04, -2.5520e-03, ..., 2.0468e-04, + -2.4319e-03, 2.0981e-04], + ..., + [ 2.0933e-04, 3.9062e-03, 1.4563e-03, ..., -1.7524e-05, + -4.0507e-04, 9.4795e-04], + [ 1.8673e-03, -1.8940e-03, 3.6278e-03, ..., -6.1417e-03, + -2.3537e-03, -2.4776e-03], + [ 9.4223e-04, -9.1629e-03, -1.2436e-02, ..., -3.5591e-03, + -7.3090e-03, -1.0042e-03]], device='cuda:0') +Epoch 63, bias, value: tensor([-0.0337, -0.0044, -0.0034, -0.0064, -0.0063, 0.0052, -0.0092, 0.0243, + -0.0039, 0.0213], device='cuda:0'), grad: tensor([-0.0202, 0.0049, -0.0081, 0.0502, -0.0038, 0.0300, -0.0081, 0.0055, + 0.0003, -0.0507], device='cuda:0') +100 +0.0001 +changing lr +epoch 62, time 214.53, cls_loss 0.7777 cls_loss_mapping 0.0336 cls_loss_causal 0.6719 re_mapping 0.0227 re_causal 0.0507 /// teacc 98.00 lr 0.00010000 +Epoch 64, weight, value: tensor([[-0.0929, -0.0326, -0.0045, ..., -0.0036, -0.0439, -0.0558], + [-0.0033, -0.0375, 0.0148, ..., 0.0212, 0.0034, -0.0298], + [-0.0139, -0.0389, -0.0559, ..., 0.0355, -0.0253, -0.0402], + ..., + [-0.0265, 0.0219, 0.0267, ..., 0.0226, -0.0293, -0.0422], + [-0.0486, -0.0073, 0.0127, ..., 0.0440, -0.0218, -0.0444], + [ 0.0238, 0.0348, -0.0195, ..., -0.0670, -0.0275, 0.0107]], + device='cuda:0'), grad: tensor([[ 1.2722e-03, -2.6840e-02, 2.8777e-04, ..., 4.5624e-03, + 8.7118e-04, -2.9507e-03], + [ 5.4741e-04, 8.3566e-05, 5.0735e-04, ..., 3.7708e-03, + 2.2297e-03, 4.2892e-04], + [ 1.0004e-03, 1.4362e-03, 1.9026e-03, ..., 1.8738e-02, + 1.3451e-02, 3.7861e-03], + ..., + [ 1.0557e-03, 2.9707e-04, 1.3840e-04, ..., -3.2837e-02, + -1.6464e-02, 3.7241e-04], + [-4.7188e-03, -4.8447e-04, -3.3054e-03, ..., 3.1395e-03, + 3.0746e-03, -3.8490e-03], + [-1.1263e-03, 8.4152e-03, 6.7139e-04, ..., 4.8180e-03, + 2.5291e-03, 1.5545e-03]], device='cuda:0') +Epoch 64, bias, value: tensor([-0.0338, -0.0044, -0.0025, -0.0065, -0.0061, 0.0053, -0.0096, 0.0245, + -0.0042, 0.0210], device='cuda:0'), grad: tensor([ 0.0228, 0.0227, -0.0147, 0.0184, -0.0229, -0.0001, -0.0078, -0.0287, + -0.0141, 0.0245], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 63---------------------------------------------------- +epoch 63, time 215.12, cls_loss 0.7441 cls_loss_mapping 0.0345 cls_loss_causal 0.6407 re_mapping 0.0233 re_causal 0.0516 /// teacc 98.23 lr 0.00010000 +Epoch 65, weight, value: tensor([[-0.0943, -0.0325, -0.0046, ..., -0.0039, -0.0433, -0.0567], + [-0.0023, -0.0380, 0.0150, ..., 0.0205, 0.0028, -0.0300], + [-0.0151, -0.0385, -0.0559, ..., 0.0366, -0.0253, -0.0413], + ..., + [-0.0273, 0.0216, 0.0261, ..., 0.0228, -0.0283, -0.0437], + [-0.0493, -0.0075, 0.0120, ..., 0.0433, -0.0221, -0.0452], + [ 0.0246, 0.0349, -0.0194, ..., -0.0682, -0.0290, 0.0124]], + device='cuda:0'), grad: tensor([[-1.1176e-06, -2.4948e-03, 1.6034e-04, ..., 2.2945e-03, + 5.4169e-04, 5.6362e-04], + [ 1.8919e-04, 9.5034e-04, 3.1137e-04, ..., 5.1231e-03, + 4.0579e-04, 3.6621e-04], + [ 5.4407e-04, 2.5597e-03, 5.3644e-04, ..., -5.4245e-03, + 6.9284e-04, 1.1177e-03], + ..., + [ 2.8496e-03, -4.1733e-03, 2.2564e-03, ..., 1.3866e-03, + 3.6335e-04, 1.4162e-03], + [ 2.0266e-05, 1.8530e-03, -5.6171e-04, ..., -7.3700e-03, + 5.6410e-04, -9.4604e-04], + [-6.0501e-03, 9.4080e-04, -5.0735e-03, ..., 3.1223e-03, + 6.9094e-04, -2.7828e-03]], device='cuda:0') +Epoch 65, bias, value: tensor([-0.0339, -0.0050, -0.0027, -0.0061, -0.0057, 0.0054, -0.0089, 0.0243, + -0.0044, 0.0205], device='cuda:0'), grad: tensor([ 0.0140, 0.0206, -0.0113, 0.0289, 0.0242, -0.0159, -0.0409, 0.0151, + -0.0316, -0.0031], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 64---------------------------------------------------- +epoch 64, time 215.34, cls_loss 0.7265 cls_loss_mapping 0.0313 cls_loss_causal 0.6048 re_mapping 0.0230 re_causal 0.0493 /// teacc 98.31 lr 0.00010000 +Epoch 66, weight, value: tensor([[-0.0955, -0.0322, -0.0043, ..., -0.0052, -0.0446, -0.0576], + [-0.0023, -0.0375, 0.0153, ..., 0.0203, 0.0021, -0.0296], + [-0.0159, -0.0389, -0.0564, ..., 0.0371, -0.0260, -0.0428], + ..., + [-0.0277, 0.0210, 0.0251, ..., 0.0229, -0.0287, -0.0451], + [-0.0490, -0.0076, 0.0122, ..., 0.0444, -0.0230, -0.0452], + [ 0.0256, 0.0347, -0.0200, ..., -0.0687, -0.0292, 0.0130]], + device='cuda:0'), grad: tensor([[ 5.5027e-04, 3.5632e-06, 5.5361e-04, ..., 1.9426e-03, + 4.4155e-04, -1.3132e-03], + [-5.7487e-03, -3.9905e-05, 5.3167e-04, ..., -2.9316e-03, + 1.1196e-03, 4.3660e-05], + [-2.0516e-04, 2.5654e-04, -3.0365e-03, ..., -5.8136e-03, + -4.5180e-04, 1.5187e-04], + ..., + [ 1.0052e-03, -1.0759e-05, 2.3956e-03, ..., -4.0627e-03, + 1.1319e-04, -2.2781e-04], + [ 5.7106e-03, -3.1114e-04, 1.6003e-03, ..., 4.4975e-03, + 6.3229e-04, 2.1877e-03], + [ 9.7561e-04, 1.0327e-05, 1.6356e-03, ..., 2.3594e-03, + 2.5272e-04, 3.8767e-04]], device='cuda:0') +Epoch 66, bias, value: tensor([-0.0347, -0.0049, -0.0024, -0.0067, -0.0054, 0.0057, -0.0092, 0.0242, + -0.0041, 0.0210], device='cuda:0'), grad: tensor([ 0.0030, -0.0244, -0.0194, 0.0130, 0.0152, -0.0308, -0.0040, -0.0077, + 0.0372, 0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 65, time 214.63, cls_loss 0.7399 cls_loss_mapping 0.0310 cls_loss_causal 0.6237 re_mapping 0.0236 re_causal 0.0521 /// teacc 98.10 lr 0.00010000 +Epoch 67, weight, value: tensor([[-0.0963, -0.0310, -0.0046, ..., -0.0055, -0.0462, -0.0578], + [-0.0026, -0.0383, 0.0151, ..., 0.0204, 0.0023, -0.0300], + [-0.0162, -0.0395, -0.0556, ..., 0.0374, -0.0243, -0.0432], + ..., + [-0.0281, 0.0207, 0.0253, ..., 0.0225, -0.0274, -0.0463], + [-0.0492, -0.0078, 0.0117, ..., 0.0448, -0.0231, -0.0464], + [ 0.0260, 0.0357, -0.0203, ..., -0.0693, -0.0304, 0.0133]], + device='cuda:0'), grad: tensor([[-6.1531e-03, 2.9898e-04, -9.3918e-03, ..., -2.9812e-03, + 3.8147e-04, -4.3945e-03], + [ 2.6817e-03, 7.9107e-04, 1.0729e-03, ..., -3.4122e-03, + 1.4963e-03, 6.1703e-04], + [-1.5068e-03, 1.0711e-04, 4.8280e-05, ..., -9.4681e-03, + -3.5648e-03, -4.2763e-03], + ..., + [-1.5839e-02, -3.7022e-03, 2.5558e-04, ..., 9.3341e-05, + 4.3154e-05, -3.6125e-03], + [ 3.7346e-03, 8.6069e-04, 3.3321e-03, ..., 4.4441e-03, + 1.7691e-03, 2.7523e-03], + [ 2.2217e-02, 2.5272e-03, 1.4849e-03, ..., 1.7271e-03, + 2.0847e-03, 1.0078e-02]], device='cuda:0') +Epoch 67, bias, value: tensor([-0.0349, -0.0051, -0.0020, -0.0066, -0.0057, 0.0055, -0.0095, 0.0243, + -0.0040, 0.0215], device='cuda:0'), grad: tensor([-0.0208, -0.0102, -0.0264, 0.0203, 0.0033, -0.0177, 0.0177, -0.0271, + 0.0281, 0.0327], device='cuda:0') +100 +0.0001 +changing lr +epoch 66, time 214.16, cls_loss 0.7268 cls_loss_mapping 0.0279 cls_loss_causal 0.6242 re_mapping 0.0230 re_causal 0.0521 /// teacc 97.99 lr 0.00010000 +Epoch 68, weight, value: tensor([[-0.0948, -0.0315, -0.0039, ..., -0.0046, -0.0465, -0.0570], + [-0.0025, -0.0384, 0.0149, ..., 0.0207, 0.0022, -0.0300], + [-0.0173, -0.0402, -0.0560, ..., 0.0366, -0.0246, -0.0430], + ..., + [-0.0281, 0.0206, 0.0241, ..., 0.0229, -0.0268, -0.0470], + [-0.0497, -0.0071, 0.0115, ..., 0.0455, -0.0231, -0.0481], + [ 0.0257, 0.0364, -0.0198, ..., -0.0696, -0.0296, 0.0140]], + device='cuda:0'), grad: tensor([[ 0.0015, 0.0028, 0.0004, ..., 0.0076, -0.0056, 0.0028], + [-0.0025, 0.0012, 0.0005, ..., -0.0071, 0.0014, 0.0004], + [-0.0034, -0.0231, 0.0002, ..., -0.0194, 0.0004, -0.0022], + ..., + [-0.0029, 0.0153, -0.0057, ..., 0.0134, 0.0031, -0.0038], + [ 0.0032, 0.0027, -0.0007, ..., 0.0065, 0.0018, 0.0007], + [-0.0007, -0.0096, -0.0043, ..., 0.0098, 0.0005, 0.0007]], + device='cuda:0') +Epoch 68, bias, value: tensor([-0.0337, -0.0046, -0.0026, -0.0068, -0.0059, 0.0049, -0.0089, 0.0243, + -0.0040, 0.0209], device='cuda:0'), grad: tensor([ 1.9089e-02, -5.8563e-02, -7.7209e-02, 9.9640e-03, 1.3161e-02, + 6.3106e-06, 5.3482e-03, 3.3020e-02, 3.4454e-02, 2.0676e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 67, time 214.37, cls_loss 0.7422 cls_loss_mapping 0.0268 cls_loss_causal 0.6226 re_mapping 0.0225 re_causal 0.0484 /// teacc 98.07 lr 0.00010000 +Epoch 69, weight, value: tensor([[-0.0950, -0.0321, -0.0024, ..., -0.0057, -0.0480, -0.0574], + [-0.0024, -0.0387, 0.0144, ..., 0.0208, 0.0009, -0.0309], + [-0.0167, -0.0401, -0.0573, ..., 0.0374, -0.0239, -0.0430], + ..., + [-0.0287, 0.0217, 0.0239, ..., 0.0238, -0.0271, -0.0477], + [-0.0518, -0.0065, 0.0119, ..., 0.0452, -0.0210, -0.0495], + [ 0.0267, 0.0360, -0.0199, ..., -0.0699, -0.0285, 0.0153]], + device='cuda:0'), grad: tensor([[-6.6986e-03, -4.3106e-03, -1.7242e-02, ..., -1.5991e-02, + -8.0948e-03, -2.5940e-03], + [ 1.1057e-04, 3.8795e-03, 3.5362e-03, ..., 7.9193e-03, + 5.0011e-03, 2.9474e-05], + [ 1.0509e-03, 2.3441e-03, 1.5955e-03, ..., 7.9422e-03, + 1.9050e-04, 1.7023e-04], + ..., + [ 3.6449e-03, 1.8112e-02, -3.2349e-03, ..., 1.2642e-02, + -6.8398e-03, 1.8091e-03], + [ 2.2850e-03, -3.5973e-03, -2.1343e-03, ..., -1.0773e-02, + -2.4548e-03, 3.5930e-04], + [ 1.0614e-03, -2.9877e-02, 6.4354e-03, ..., 4.1351e-03, + 6.0463e-03, -8.9407e-05]], device='cuda:0') +Epoch 69, bias, value: tensor([-0.0346, -0.0046, -0.0013, -0.0079, -0.0064, 0.0054, -0.0088, 0.0246, + -0.0042, 0.0213], device='cuda:0'), grad: tensor([-0.0509, 0.0246, 0.0118, 0.0177, 0.0176, -0.0733, -0.0080, 0.0390, + 0.0016, 0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 68, time 214.33, cls_loss 0.7552 cls_loss_mapping 0.0342 cls_loss_causal 0.6409 re_mapping 0.0225 re_causal 0.0480 /// teacc 98.23 lr 0.00010000 +Epoch 70, weight, value: tensor([[-0.0945, -0.0326, -0.0021, ..., -0.0059, -0.0484, -0.0582], + [-0.0029, -0.0382, 0.0138, ..., 0.0212, 0.0002, -0.0325], + [-0.0160, -0.0398, -0.0572, ..., 0.0376, -0.0245, -0.0440], + ..., + [-0.0291, 0.0216, 0.0257, ..., 0.0227, -0.0267, -0.0480], + [-0.0525, -0.0070, 0.0117, ..., 0.0451, -0.0191, -0.0509], + [ 0.0275, 0.0367, -0.0213, ..., -0.0705, -0.0295, 0.0175]], + device='cuda:0'), grad: tensor([[ 0.0033, 0.0007, 0.0023, ..., -0.0238, -0.0017, 0.0001], + [ 0.0005, 0.0004, 0.0014, ..., 0.0130, 0.0023, 0.0003], + [ 0.0061, 0.0006, 0.0050, ..., 0.0204, 0.0042, 0.0025], + ..., + [ 0.0163, 0.0225, 0.0013, ..., 0.0089, 0.0023, 0.0090], + [ 0.0028, -0.0190, -0.0099, ..., 0.0020, -0.0051, 0.0023], + [ 0.0018, -0.0030, -0.0037, ..., -0.0082, -0.0045, -0.0048]], + device='cuda:0') +Epoch 70, bias, value: tensor([-0.0350, -0.0047, -0.0012, -0.0070, -0.0066, 0.0055, -0.0086, 0.0249, + -0.0048, 0.0210], device='cuda:0'), grad: tensor([-0.0612, 0.0424, 0.0770, -0.0321, 0.0511, -0.0486, -0.0115, 0.0160, + -0.0135, -0.0196], device='cuda:0') +100 +0.0001 +changing lr +epoch 69, time 214.27, cls_loss 0.7782 cls_loss_mapping 0.0246 cls_loss_causal 0.6577 re_mapping 0.0219 re_causal 0.0499 /// teacc 98.21 lr 0.00010000 +Epoch 71, weight, value: tensor([[-0.0959, -0.0328, -0.0031, ..., -0.0057, -0.0492, -0.0599], + [-0.0031, -0.0384, 0.0124, ..., 0.0224, -0.0005, -0.0321], + [-0.0164, -0.0399, -0.0562, ..., 0.0376, -0.0228, -0.0447], + ..., + [-0.0295, 0.0215, 0.0260, ..., 0.0231, -0.0274, -0.0494], + [-0.0526, -0.0067, 0.0121, ..., 0.0445, -0.0196, -0.0514], + [ 0.0278, 0.0372, -0.0213, ..., -0.0707, -0.0292, 0.0181]], + device='cuda:0'), grad: tensor([[ 0.0190, 0.0006, 0.0004, ..., 0.0051, 0.0108, 0.0005], + [ 0.0002, 0.0003, -0.0029, ..., -0.0064, -0.0111, 0.0003], + [ 0.0008, 0.0006, 0.0006, ..., 0.0038, 0.0032, 0.0009], + ..., + [ 0.0022, 0.0007, 0.0028, ..., 0.0032, 0.0022, 0.0023], + [ 0.0021, 0.0005, 0.0010, ..., 0.0039, 0.0024, 0.0019], + [ 0.0012, 0.0012, -0.0003, ..., 0.0052, 0.0038, 0.0019]], + device='cuda:0') +Epoch 71, bias, value: tensor([-0.0348, -0.0042, -0.0016, -0.0080, -0.0059, 0.0054, -0.0096, 0.0250, + -0.0042, 0.0212], device='cuda:0'), grad: tensor([ 0.0621, -0.0423, 0.0208, -0.0164, -0.0797, 0.0499, -0.0732, 0.0231, + 0.0260, 0.0296], device='cuda:0') +100 +0.0001 +changing lr +epoch 70, time 214.82, cls_loss 0.7309 cls_loss_mapping 0.0263 cls_loss_causal 0.6219 re_mapping 0.0212 re_causal 0.0472 /// teacc 98.26 lr 0.00010000 +Epoch 72, weight, value: tensor([[-0.0967, -0.0332, -0.0025, ..., -0.0059, -0.0481, -0.0604], + [-0.0038, -0.0387, 0.0137, ..., 0.0224, -0.0008, -0.0315], + [-0.0175, -0.0403, -0.0571, ..., 0.0382, -0.0232, -0.0450], + ..., + [-0.0304, 0.0205, 0.0247, ..., 0.0232, -0.0279, -0.0498], + [-0.0521, -0.0076, 0.0124, ..., 0.0442, -0.0208, -0.0522], + [ 0.0269, 0.0373, -0.0206, ..., -0.0709, -0.0281, 0.0177]], + device='cuda:0'), grad: tensor([[ 3.7932e-04, 3.4027e-03, 4.4012e-04, ..., 5.0964e-03, + 8.9645e-03, 1.7440e-04], + [ 3.7551e-04, -8.9931e-04, -1.3202e-05, ..., 1.9207e-03, + -9.5673e-03, 1.0443e-03], + [ 4.4012e-04, 1.7586e-03, 9.7370e-04, ..., -5.0507e-03, + -3.0708e-03, 4.4060e-04], + ..., + [-1.8463e-03, -8.1940e-03, -5.7259e-03, ..., -1.8120e-03, + 3.4332e-03, 1.5867e-04], + [-1.1959e-03, -6.2447e-03, -1.5097e-03, ..., -5.0697e-03, + -5.9052e-03, -6.1722e-03], + [ 7.8201e-04, 7.4615e-03, 1.6556e-03, ..., 1.5736e-04, + 5.4455e-04, 1.2589e-03]], device='cuda:0') +Epoch 72, bias, value: tensor([-0.0344, -0.0044, -0.0015, -0.0082, -0.0051, 0.0048, -0.0097, 0.0250, + -0.0048, 0.0217], device='cuda:0'), grad: tensor([ 0.0256, -0.0211, 0.0154, 0.0353, -0.0058, -0.0250, 0.0318, -0.0124, + -0.0476, 0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 71, time 214.66, cls_loss 0.7178 cls_loss_mapping 0.0294 cls_loss_causal 0.6090 re_mapping 0.0209 re_causal 0.0459 /// teacc 98.13 lr 0.00010000 +Epoch 73, weight, value: tensor([[-0.0978, -0.0324, -0.0004, ..., -0.0071, -0.0471, -0.0613], + [-0.0035, -0.0395, 0.0127, ..., 0.0227, -0.0009, -0.0327], + [-0.0171, -0.0411, -0.0572, ..., 0.0378, -0.0247, -0.0450], + ..., + [-0.0313, 0.0207, 0.0245, ..., 0.0247, -0.0278, -0.0503], + [-0.0521, -0.0071, 0.0118, ..., 0.0455, -0.0212, -0.0535], + [ 0.0272, 0.0376, -0.0205, ..., -0.0735, -0.0288, 0.0183]], + device='cuda:0'), grad: tensor([[-1.3363e-04, 2.8286e-03, -9.9792e-03, ..., 4.1733e-03, + 3.3245e-03, 3.5667e-03], + [ 7.0715e-04, 8.6517e-03, 7.3433e-03, ..., 6.3705e-03, + 3.8929e-03, 9.3985e-04], + [-4.1389e-03, -3.3894e-03, -5.6601e-04, ..., -1.1040e-02, + -6.6452e-03, -5.9662e-03], + ..., + [-8.6606e-05, -3.4189e-04, 2.3293e-04, ..., -1.2192e-02, + -4.7646e-03, 6.6566e-04], + [-3.7823e-03, -9.3307e-03, -2.1458e-03, ..., -8.4686e-03, + -5.9700e-03, 3.0518e-04], + [ 2.9564e-03, 4.2191e-03, 3.4547e-04, ..., -5.8174e-04, + -1.8024e-03, 1.9703e-03]], device='cuda:0') +Epoch 73, bias, value: tensor([-0.0345, -0.0041, -0.0025, -0.0079, -0.0048, 0.0051, -0.0097, 0.0258, + -0.0048, 0.0207], device='cuda:0'), grad: tensor([ 0.0101, 0.0265, -0.0445, 0.0464, 0.0159, -0.0050, 0.0137, -0.0382, + -0.0293, 0.0042], device='cuda:0') +100 +0.0001 +changing lr +epoch 72, time 214.37, cls_loss 0.7506 cls_loss_mapping 0.0250 cls_loss_causal 0.6321 re_mapping 0.0206 re_causal 0.0461 /// teacc 98.18 lr 0.00010000 +Epoch 74, weight, value: tensor([[-0.0989, -0.0310, 0.0004, ..., -0.0067, -0.0481, -0.0616], + [-0.0045, -0.0386, 0.0117, ..., 0.0232, -0.0006, -0.0337], + [-0.0163, -0.0416, -0.0570, ..., 0.0378, -0.0257, -0.0449], + ..., + [-0.0314, 0.0209, 0.0249, ..., 0.0236, -0.0283, -0.0497], + [-0.0531, -0.0076, 0.0114, ..., 0.0461, -0.0207, -0.0545], + [ 0.0273, 0.0378, -0.0200, ..., -0.0732, -0.0278, 0.0186]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0014, 0.0002, ..., 0.0045, 0.0035, 0.0007], + [-0.0012, 0.0090, -0.0060, ..., 0.0085, 0.0019, 0.0012], + [ 0.0033, 0.0018, 0.0003, ..., -0.0056, -0.0053, 0.0016], + ..., + [ 0.0013, -0.0021, 0.0008, ..., -0.0074, -0.0035, -0.0004], + [ 0.0028, -0.0088, 0.0052, ..., 0.0015, -0.0055, 0.0008], + [-0.0013, 0.0021, -0.0014, ..., 0.0043, 0.0023, -0.0002]], + device='cuda:0') +Epoch 74, bias, value: tensor([-0.0344, -0.0040, -0.0026, -0.0078, -0.0052, 0.0061, -0.0101, 0.0257, + -0.0044, 0.0203], device='cuda:0'), grad: tensor([ 0.0271, 0.0170, -0.0256, -0.0158, -0.0079, -0.0439, 0.0432, -0.0161, + 0.0042, 0.0179], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 73---------------------------------------------------- +epoch 73, time 215.34, cls_loss 0.7539 cls_loss_mapping 0.0295 cls_loss_causal 0.6515 re_mapping 0.0204 re_causal 0.0464 /// teacc 98.39 lr 0.00010000 +Epoch 75, weight, value: tensor([[-1.0066e-01, -3.1539e-02, 1.1570e-03, ..., -7.1371e-03, + -4.8355e-02, -6.2627e-02], + [-5.5121e-03, -3.8157e-02, 1.2446e-02, ..., 2.2803e-02, + -8.8621e-05, -3.5304e-02], + [-1.7032e-02, -4.1763e-02, -5.7699e-02, ..., 3.8446e-02, + -2.5801e-02, -4.5856e-02], + ..., + [-3.0982e-02, 2.1652e-02, 2.3889e-02, ..., 2.3208e-02, + -2.8668e-02, -4.8909e-02], + [-5.2276e-02, -7.6586e-03, 1.1834e-02, ..., 4.5792e-02, + -2.1970e-02, -5.3894e-02], + [ 2.7572e-02, 3.7161e-02, -1.8581e-02, ..., -7.3759e-02, + -2.9613e-02, 1.8219e-02]], device='cuda:0'), grad: tensor([[ 1.0271e-03, 1.5335e-02, 3.1605e-03, ..., 9.1400e-03, + 6.0654e-03, 3.5524e-04], + [-2.5558e-03, -1.8902e-03, -7.2479e-04, ..., 1.2230e-02, + 1.8417e-02, -2.1152e-03], + [ 2.0103e-03, -3.1261e-03, 4.3917e-04, ..., -5.8632e-03, + -2.6764e-02, 6.2180e-04], + ..., + [-3.1900e-04, 7.0906e-04, 3.4004e-05, ..., 3.3379e-03, + -3.2825e-03, -1.0853e-03], + [ 2.3079e-03, 1.8511e-03, -5.4245e-03, ..., 6.8016e-03, + 5.1689e-03, 7.0858e-04], + [ 5.7697e-05, -1.8921e-02, 3.2997e-04, ..., -3.3661e-02, + -1.0353e-02, 2.2185e-04]], device='cuda:0') +Epoch 75, bias, value: tensor([-0.0345, -0.0045, -0.0026, -0.0078, -0.0057, 0.0067, -0.0094, 0.0257, + -0.0045, 0.0201], device='cuda:0'), grad: tensor([ 0.0653, 0.0041, 0.0087, -0.0352, -0.0017, 0.0273, 0.0218, -0.0020, + 0.0057, -0.0942], device='cuda:0') +100 +0.0001 +changing lr +epoch 74, time 216.54, cls_loss 0.7100 cls_loss_mapping 0.0275 cls_loss_causal 0.6002 re_mapping 0.0208 re_causal 0.0443 /// teacc 98.34 lr 0.00010000 +Epoch 76, weight, value: tensor([[-0.1004, -0.0318, 0.0007, ..., -0.0072, -0.0498, -0.0615], + [-0.0055, -0.0385, 0.0123, ..., 0.0224, -0.0009, -0.0350], + [-0.0177, -0.0419, -0.0579, ..., 0.0383, -0.0257, -0.0470], + ..., + [-0.0305, 0.0232, 0.0242, ..., 0.0232, -0.0292, -0.0490], + [-0.0533, -0.0068, 0.0111, ..., 0.0448, -0.0234, -0.0554], + [ 0.0280, 0.0364, -0.0185, ..., -0.0735, -0.0302, 0.0191]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0001, 0.0013, ..., 0.0006, 0.0013, 0.0002], + [ 0.0013, 0.0003, 0.0010, ..., 0.0016, 0.0026, 0.0003], + [ 0.0017, 0.0005, 0.0014, ..., 0.0020, 0.0021, 0.0008], + ..., + [ 0.0027, -0.0009, 0.0020, ..., -0.0004, 0.0006, 0.0016], + [-0.0217, 0.0017, -0.0093, ..., -0.0060, -0.0075, -0.0143], + [-0.0140, 0.0009, -0.0130, ..., -0.0038, -0.0038, -0.0099]], + device='cuda:0') +Epoch 76, bias, value: tensor([-0.0346, -0.0044, -0.0031, -0.0071, -0.0054, 0.0074, -0.0096, 0.0258, + -0.0052, 0.0198], device='cuda:0'), grad: tensor([ 0.0113, 0.0209, 0.0122, 0.0196, 0.0426, -0.0171, 0.0246, 0.0139, + -0.0527, -0.0753], device='cuda:0') +100 +0.0001 +changing lr +epoch 75, time 214.65, cls_loss 0.6792 cls_loss_mapping 0.0248 cls_loss_causal 0.5717 re_mapping 0.0204 re_causal 0.0449 /// teacc 98.27 lr 0.00010000 +Epoch 77, weight, value: tensor([[-0.1011, -0.0323, 0.0002, ..., -0.0067, -0.0508, -0.0619], + [-0.0046, -0.0392, 0.0121, ..., 0.0226, -0.0008, -0.0358], + [-0.0177, -0.0414, -0.0561, ..., 0.0386, -0.0257, -0.0464], + ..., + [-0.0315, 0.0243, 0.0239, ..., 0.0236, -0.0287, -0.0501], + [-0.0535, -0.0065, 0.0120, ..., 0.0448, -0.0232, -0.0565], + [ 0.0292, 0.0367, -0.0188, ..., -0.0741, -0.0299, 0.0198]], + device='cuda:0'), grad: tensor([[ 1.4019e-03, 8.4102e-05, 1.6436e-05, ..., 4.1389e-03, + 1.0986e-03, 8.3971e-04], + [ 1.4801e-03, 3.3131e-03, -1.2434e-04, ..., 1.3285e-03, + 2.4378e-04, 9.1910e-05], + [ 1.8005e-03, 4.6420e-04, 1.3642e-05, ..., 3.6621e-03, + 1.3170e-03, 1.8902e-03], + ..., + [-2.5291e-03, -3.0136e-03, 5.1439e-05, ..., -9.2697e-03, + 5.3120e-04, 7.0906e-04], + [ 1.8326e-02, 9.5320e-04, 1.3840e-04, ..., 4.5357e-03, + -4.9553e-03, 5.1403e-04], + [ 3.0041e-03, 4.2076e-03, 2.1305e-03, ..., 6.0730e-03, + 5.5933e-04, 3.3712e-04]], device='cuda:0') +Epoch 77, bias, value: tensor([-0.0346, -0.0038, -0.0023, -0.0078, -0.0053, 0.0075, -0.0102, 0.0252, + -0.0054, 0.0205], device='cuda:0'), grad: tensor([ 0.0382, 0.0103, 0.0301, -0.0345, -0.0151, -0.0786, 0.0263, -0.0605, + 0.0376, 0.0461], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 76---------------------------------------------------- +epoch 76, time 215.23, cls_loss 0.7240 cls_loss_mapping 0.0321 cls_loss_causal 0.6237 re_mapping 0.0201 re_causal 0.0438 /// teacc 98.42 lr 0.00010000 +Epoch 78, weight, value: tensor([[-1.0229e-01, -3.2048e-02, 1.8693e-04, ..., -6.6452e-03, + -5.1338e-02, -6.2215e-02], + [-4.2440e-03, -4.0016e-02, 1.2258e-02, ..., 2.2619e-02, + 5.9843e-05, -3.6112e-02], + [-1.7592e-02, -4.0706e-02, -5.6424e-02, ..., 3.8281e-02, + -2.6813e-02, -4.6330e-02], + ..., + [-3.1391e-02, 2.4409e-02, 2.3498e-02, ..., 2.3671e-02, + -2.9178e-02, -5.1592e-02], + [-5.4483e-02, -6.0722e-03, 1.2066e-02, ..., 4.5250e-02, + -2.3717e-02, -5.6993e-02], + [ 2.7999e-02, 3.6960e-02, -2.0505e-02, ..., -7.3778e-02, + -2.9040e-02, 2.0040e-02]], device='cuda:0'), grad: tensor([[ 5.3501e-04, -2.1553e-04, 4.6158e-04, ..., 1.2264e-03, + 1.7023e-03, 9.6858e-05], + [ 1.6203e-03, 3.2368e-03, 7.1096e-04, ..., 2.0504e-03, + 1.8444e-03, 3.8171e-04], + [ 6.9332e-04, 2.8849e-04, 1.4923e-02, ..., -4.0102e-04, + -2.1725e-03, -1.6479e-03], + ..., + [ 1.5807e-04, 1.2016e-03, 8.3113e-04, ..., -1.8091e-03, + 1.4486e-03, 1.5652e-04], + [ 2.7790e-03, 9.8896e-04, -3.6240e-03, ..., 3.1071e-03, + 1.5697e-03, 2.6627e-03], + [ 7.0763e-04, 3.8109e-03, 9.1124e-04, ..., 2.1248e-03, + 1.7941e-04, 4.9305e-04]], device='cuda:0') +Epoch 78, bias, value: tensor([-0.0345, -0.0039, -0.0023, -0.0078, -0.0049, 0.0067, -0.0099, 0.0255, + -0.0056, 0.0203], device='cuda:0'), grad: tensor([ 0.0126, 0.0255, 0.0155, 0.0160, -0.0009, -0.0434, -0.0516, 0.0069, + 0.0031, 0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 77, time 214.42, cls_loss 0.6871 cls_loss_mapping 0.0233 cls_loss_causal 0.5751 re_mapping 0.0201 re_causal 0.0436 /// teacc 98.31 lr 0.00010000 +Epoch 79, weight, value: tensor([[-1.0304e-01, -3.2197e-02, 2.0503e-05, ..., -6.9149e-03, + -5.2560e-02, -6.3131e-02], + [-4.3293e-03, -4.0805e-02, 1.1724e-02, ..., 2.2548e-02, + -5.5474e-04, -3.6920e-02], + [-1.7699e-02, -3.9200e-02, -5.6190e-02, ..., 3.8417e-02, + -2.6520e-02, -4.6914e-02], + ..., + [-3.1496e-02, 2.4293e-02, 2.2704e-02, ..., 2.3778e-02, + -2.9489e-02, -5.1945e-02], + [-5.3825e-02, -6.6946e-03, 1.2813e-02, ..., 4.5106e-02, + -2.2611e-02, -5.7671e-02], + [ 2.8089e-02, 3.6216e-02, -2.1271e-02, ..., -7.4582e-02, + -2.9713e-02, 2.0508e-02]], device='cuda:0'), grad: tensor([[-1.5001e-03, 8.3971e-04, 5.7042e-05, ..., 1.6098e-03, + 2.8515e-04, -1.7042e-03], + [ 1.9288e-04, 1.8539e-03, 2.0158e-04, ..., -9.9030e-03, + -1.6052e-02, 2.5177e-04], + [ 1.4172e-03, 6.5460e-03, 1.9932e-03, ..., 8.1863e-03, + 1.1253e-02, 3.3832e-04], + ..., + [ 1.3142e-03, 1.0178e-02, 3.0875e-04, ..., 1.8034e-03, + 2.7275e-03, -4.6301e-04], + [-5.5466e-03, 2.5215e-03, 1.2493e-04, ..., -8.3160e-03, + -2.0142e-03, -5.8699e-04], + [-2.4261e-03, 2.4551e-02, -1.7071e-03, ..., 3.4523e-03, + 9.0256e-03, 2.9624e-05]], device='cuda:0') +Epoch 79, bias, value: tensor([-0.0354, -0.0040, -0.0023, -0.0074, -0.0048, 0.0074, -0.0095, 0.0253, + -0.0054, 0.0197], device='cuda:0'), grad: tensor([ 0.0054, -0.0044, 0.0233, 0.0142, -0.0297, 0.0450, -0.0362, 0.0031, + -0.0627, 0.0419], device='cuda:0') +100 +0.0001 +changing lr +epoch 78, time 214.81, cls_loss 0.7108 cls_loss_mapping 0.0241 cls_loss_causal 0.5962 re_mapping 0.0212 re_causal 0.0455 /// teacc 98.10 lr 0.00010000 +Epoch 80, weight, value: tensor([[-0.1040, -0.0320, 0.0002, ..., -0.0070, -0.0514, -0.0640], + [-0.0053, -0.0416, 0.0122, ..., 0.0230, -0.0003, -0.0389], + [-0.0174, -0.0399, -0.0565, ..., 0.0379, -0.0276, -0.0472], + ..., + [-0.0323, 0.0243, 0.0227, ..., 0.0246, -0.0295, -0.0524], + [-0.0535, -0.0061, 0.0135, ..., 0.0450, -0.0236, -0.0600], + [ 0.0288, 0.0363, -0.0212, ..., -0.0744, -0.0285, 0.0210]], + device='cuda:0'), grad: tensor([[ 3.5095e-04, 1.0948e-03, 6.3848e-04, ..., 4.4785e-03, + 3.2120e-03, 2.7135e-05], + [ 1.4521e-05, -6.0654e-03, 1.0824e-04, ..., 2.2602e-04, + -5.4245e-03, 1.0461e-05], + [ 1.1688e-04, 2.6989e-03, 2.7514e-04, ..., -3.0842e-03, + 3.4103e-03, 2.8968e-05], + ..., + [ 2.1725e-03, 4.2267e-02, 5.3072e-04, ..., 1.9241e-02, + 1.1024e-02, 2.3746e-03], + [ 3.0383e-05, -8.0884e-05, -5.7340e-05, ..., 6.3705e-03, + 2.7370e-03, 1.8275e-04], + [-2.9182e-03, -4.2145e-02, -4.1461e-04, ..., -2.2079e-02, + -1.3687e-02, -3.2101e-03]], device='cuda:0') +Epoch 80, bias, value: tensor([-0.0351, -0.0044, -0.0028, -0.0072, -0.0046, 0.0070, -0.0097, 0.0254, + -0.0051, 0.0200], device='cuda:0'), grad: tensor([ 0.0226, -0.0123, 0.0004, -0.0072, -0.0281, -0.0041, 0.0191, 0.0404, + 0.0182, -0.0490], device='cuda:0') +100 +0.0001 +changing lr +epoch 79, time 214.69, cls_loss 0.7207 cls_loss_mapping 0.0216 cls_loss_causal 0.6112 re_mapping 0.0196 re_causal 0.0426 /// teacc 98.30 lr 0.00010000 +Epoch 81, weight, value: tensor([[-0.1045, -0.0314, 0.0007, ..., -0.0076, -0.0522, -0.0642], + [-0.0060, -0.0402, 0.0135, ..., 0.0242, 0.0011, -0.0398], + [-0.0181, -0.0410, -0.0558, ..., 0.0381, -0.0266, -0.0475], + ..., + [-0.0323, 0.0238, 0.0226, ..., 0.0236, -0.0310, -0.0522], + [-0.0544, -0.0042, 0.0125, ..., 0.0459, -0.0246, -0.0611], + [ 0.0289, 0.0352, -0.0219, ..., -0.0761, -0.0285, 0.0220]], + device='cuda:0'), grad: tensor([[ 4.0603e-04, 1.3638e-03, 4.1819e-04, ..., 9.9945e-04, + 6.4087e-04, -5.3823e-05], + [ 3.4070e-04, 1.6475e-04, 5.3215e-04, ..., 3.0556e-03, + 5.3368e-03, 1.3006e-04], + [-1.0223e-03, 2.7370e-03, 3.7050e-04, ..., -8.0872e-03, + -3.8414e-03, -4.7731e-04], + ..., + [ 1.1861e-04, 1.2436e-03, 3.9673e-03, ..., 1.4658e-03, + 2.0695e-03, 7.6890e-05], + [ 1.5850e-03, 2.0218e-02, 1.9627e-03, ..., 8.0948e-03, + 1.0628e-02, 8.4519e-05], + [ 4.6277e-04, 4.6463e-03, 4.0398e-03, ..., 6.8436e-03, + 6.9351e-03, 2.1327e-04]], device='cuda:0') +Epoch 81, bias, value: tensor([-0.0355, -0.0036, -0.0023, -0.0070, -0.0043, 0.0067, -0.0097, 0.0249, + -0.0055, 0.0198], device='cuda:0'), grad: tensor([-0.0043, 0.0107, -0.0789, 0.0039, 0.0285, -0.0781, 0.0257, 0.0060, + 0.0517, 0.0347], device='cuda:0') +100 +0.0001 +changing lr +epoch 80, time 214.76, cls_loss 0.7156 cls_loss_mapping 0.0236 cls_loss_causal 0.6183 re_mapping 0.0203 re_causal 0.0450 /// teacc 98.25 lr 0.00010000 +Epoch 82, weight, value: tensor([[-0.1042, -0.0312, 0.0010, ..., -0.0074, -0.0509, -0.0650], + [-0.0059, -0.0401, 0.0134, ..., 0.0241, 0.0014, -0.0411], + [-0.0176, -0.0403, -0.0559, ..., 0.0379, -0.0266, -0.0462], + ..., + [-0.0332, 0.0243, 0.0217, ..., 0.0240, -0.0305, -0.0537], + [-0.0550, -0.0041, 0.0118, ..., 0.0455, -0.0238, -0.0625], + [ 0.0296, 0.0349, -0.0226, ..., -0.0757, -0.0287, 0.0241]], + device='cuda:0'), grad: tensor([[ 6.6233e-04, 6.4516e-04, 2.3842e-03, ..., -4.9477e-03, + -1.0353e-02, 1.7047e-04], + [ 1.0020e-04, 2.5225e-04, 2.0409e-03, ..., 4.8370e-03, + 6.3629e-03, 3.9041e-06], + [ 2.0351e-03, 7.1335e-04, 1.6918e-03, ..., 2.2354e-03, + 3.4695e-03, 7.5459e-05], + ..., + [ 9.5320e-04, 1.1005e-03, 3.4199e-03, ..., 2.3327e-03, + 6.4316e-03, 1.3769e-04], + [ 8.7786e-04, 2.9278e-03, -4.5509e-03, ..., -1.0086e-02, + -1.2520e-02, 1.4591e-04], + [ 1.5230e-03, -9.8801e-03, 3.7804e-03, ..., 2.1152e-03, + 6.5880e-03, 1.9646e-04]], device='cuda:0') +Epoch 82, bias, value: tensor([-0.0350, -0.0038, -0.0024, -0.0083, -0.0040, 0.0075, -0.0104, 0.0251, + -0.0051, 0.0201], device='cuda:0'), grad: tensor([-0.0585, 0.0374, 0.0267, 0.0219, -0.0216, 0.0250, 0.0009, 0.0339, + -0.0751, 0.0095], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 81---------------------------------------------------- +epoch 81, time 216.48, cls_loss 0.7066 cls_loss_mapping 0.0192 cls_loss_causal 0.6117 re_mapping 0.0192 re_causal 0.0432 /// teacc 98.46 lr 0.00010000 +Epoch 83, weight, value: tensor([[-0.1043, -0.0309, 0.0003, ..., -0.0074, -0.0517, -0.0649], + [-0.0068, -0.0396, 0.0145, ..., 0.0236, 0.0008, -0.0410], + [-0.0170, -0.0410, -0.0564, ..., 0.0383, -0.0265, -0.0470], + ..., + [-0.0335, 0.0248, 0.0230, ..., 0.0245, -0.0314, -0.0547], + [-0.0547, -0.0047, 0.0120, ..., 0.0465, -0.0233, -0.0639], + [ 0.0303, 0.0355, -0.0230, ..., -0.0766, -0.0295, 0.0256]], + device='cuda:0'), grad: tensor([[ 2.5487e-04, -1.3672e-05, 7.3388e-06, ..., -3.2425e-03, + 3.0565e-04, 3.0518e-05], + [ 3.2234e-04, 1.3008e-03, 9.2015e-06, ..., 5.0659e-03, + 2.6245e-03, 2.0608e-05], + [ 2.2831e-03, 4.2629e-04, 4.4852e-05, ..., -5.9586e-03, + -5.7220e-03, 7.2539e-05], + ..., + [ 2.0103e-03, 2.5063e-03, 7.9036e-05, ..., 8.6823e-03, + 4.3259e-03, 1.5230e-03], + [ 3.2544e-04, 4.5967e-03, -2.8062e-04, ..., 5.3940e-03, + 4.9706e-03, 5.0592e-04], + [-7.7362e-03, -1.6052e-02, 2.5129e-04, ..., -2.2568e-02, + 1.0662e-03, -6.7291e-03]], device='cuda:0') +Epoch 83, bias, value: tensor([-0.0352, -0.0041, -0.0021, -0.0082, -0.0039, 0.0067, -0.0101, 0.0254, + -0.0046, 0.0198], device='cuda:0'), grad: tensor([-0.0170, 0.0233, -0.0385, 0.0162, 0.0313, -0.0533, 0.0289, 0.0316, + 0.0282, -0.0508], device='cuda:0') +100 +0.0001 +changing lr +epoch 82, time 215.97, cls_loss 0.7059 cls_loss_mapping 0.0222 cls_loss_causal 0.5968 re_mapping 0.0195 re_causal 0.0442 /// teacc 98.20 lr 0.00010000 +Epoch 84, weight, value: tensor([[-0.1045, -0.0310, 0.0010, ..., -0.0076, -0.0520, -0.0653], + [-0.0070, -0.0402, 0.0139, ..., 0.0239, 0.0014, -0.0415], + [-0.0169, -0.0412, -0.0565, ..., 0.0380, -0.0270, -0.0479], + ..., + [-0.0331, 0.0248, 0.0234, ..., 0.0243, -0.0323, -0.0555], + [-0.0549, -0.0056, 0.0122, ..., 0.0458, -0.0241, -0.0643], + [ 0.0296, 0.0357, -0.0237, ..., -0.0773, -0.0297, 0.0260]], + device='cuda:0'), grad: tensor([[ 3.3945e-05, -3.6812e-03, 4.1604e-04, ..., 3.8338e-03, + 3.1509e-03, 7.3051e-04], + [ 8.6486e-05, -2.0172e-02, 1.2264e-03, ..., -1.7967e-03, + -2.0874e-02, 1.3244e-04], + [ 3.1114e-04, 5.2309e-04, -4.7188e-03, ..., -2.8122e-02, + -2.8900e-02, 2.4223e-04], + ..., + [ 3.5226e-05, 2.3575e-02, 9.3937e-04, ..., 7.5912e-03, + 2.9785e-02, 3.3307e-04], + [ 5.4054e-03, -2.2793e-03, 2.1553e-03, ..., -2.1095e-03, + -9.6893e-03, 1.7776e-03], + [ 2.0218e-04, 8.4543e-04, 7.7534e-04, ..., 6.6490e-03, + 6.1569e-03, 4.6992e-04]], device='cuda:0') +Epoch 84, bias, value: tensor([-0.0352, -0.0034, -0.0026, -0.0080, -0.0030, 0.0072, -0.0105, 0.0249, + -0.0054, 0.0198], device='cuda:0'), grad: tensor([ 0.0226, -0.0151, -0.1359, 0.0225, 0.0454, -0.0020, 0.0097, 0.0495, + -0.0312, 0.0344], device='cuda:0') +100 +0.0001 +changing lr +epoch 83, time 216.18, cls_loss 0.6791 cls_loss_mapping 0.0250 cls_loss_causal 0.5678 re_mapping 0.0199 re_causal 0.0428 /// teacc 98.31 lr 0.00010000 +Epoch 85, weight, value: tensor([[-0.1051, -0.0302, 0.0007, ..., -0.0071, -0.0529, -0.0667], + [-0.0069, -0.0414, 0.0149, ..., 0.0243, 0.0015, -0.0413], + [-0.0174, -0.0421, -0.0550, ..., 0.0386, -0.0266, -0.0496], + ..., + [-0.0327, 0.0227, 0.0239, ..., 0.0237, -0.0331, -0.0547], + [-0.0547, -0.0054, 0.0115, ..., 0.0466, -0.0247, -0.0654], + [ 0.0298, 0.0377, -0.0248, ..., -0.0774, -0.0287, 0.0257]], + device='cuda:0'), grad: tensor([[ 9.1410e-04, 9.0456e-04, -1.4794e-04, ..., 3.7785e-03, + 1.4610e-03, 5.8556e-04], + [ 8.2135e-05, -2.7776e-04, 3.0565e-04, ..., -5.1727e-03, + -7.0534e-03, 5.2333e-05], + [ 4.6754e-04, 8.2245e-03, 1.0788e-04, ..., 6.9618e-03, + 1.0777e-03, 2.9874e-04], + ..., + [ 1.3647e-03, -1.6113e-02, -1.5955e-03, ..., -1.6235e-02, + -3.8261e-03, 1.4582e-03], + [-4.7646e-03, 3.0556e-03, 2.2709e-04, ..., 7.1478e-04, + 2.2659e-03, -3.1357e-03], + [-7.4625e-05, 1.4400e-03, 1.9336e-04, ..., 4.2305e-03, + 3.1204e-03, -2.6655e-04]], device='cuda:0') +Epoch 85, bias, value: tensor([-0.0352, -0.0037, -0.0022, -0.0078, -0.0035, 0.0069, -0.0100, 0.0245, + -0.0058, 0.0204], device='cuda:0'), grad: tensor([ 0.0159, -0.0430, 0.0224, 0.0161, 0.0169, -0.0205, 0.0307, -0.0475, + -0.0064, 0.0155], device='cuda:0') +100 +0.0001 +changing lr +epoch 84, time 216.30, cls_loss 0.6602 cls_loss_mapping 0.0219 cls_loss_causal 0.5622 re_mapping 0.0198 re_causal 0.0417 /// teacc 98.35 lr 0.00010000 +Epoch 86, weight, value: tensor([[-0.1070, -0.0300, 0.0004, ..., -0.0075, -0.0531, -0.0681], + [-0.0067, -0.0421, 0.0150, ..., 0.0251, 0.0010, -0.0418], + [-0.0180, -0.0422, -0.0555, ..., 0.0388, -0.0270, -0.0501], + ..., + [-0.0324, 0.0232, 0.0222, ..., 0.0237, -0.0325, -0.0546], + [-0.0552, -0.0057, 0.0122, ..., 0.0470, -0.0241, -0.0649], + [ 0.0294, 0.0371, -0.0244, ..., -0.0792, -0.0304, 0.0269]], + device='cuda:0'), grad: tensor([[ 7.4768e-04, 5.1832e-04, 7.4959e-04, ..., -6.1798e-03, + -5.9128e-03, 1.7023e-04], + [ 2.8849e-04, 9.8724e-03, 4.3464e-04, ..., 4.4632e-03, + 6.3477e-03, 5.5671e-05], + [ 1.1044e-03, -1.2054e-03, 9.8515e-04, ..., 2.4624e-03, + -2.5463e-03, 5.4550e-04], + ..., + [-4.1199e-03, -3.7781e-02, 4.8375e-04, ..., -1.4328e-02, + -2.9877e-02, 3.7265e-04], + [-4.2534e-03, -1.3916e-02, -1.9875e-03, ..., 2.5024e-03, + 6.8903e-04, -3.1834e-03], + [ 8.5678e-03, 1.8263e-03, 5.1117e-03, ..., 5.4398e-03, + 1.4420e-03, 1.1778e-03]], device='cuda:0') +Epoch 86, bias, value: tensor([-0.0355, -0.0035, -0.0024, -0.0081, -0.0030, 0.0082, -0.0097, 0.0247, + -0.0059, 0.0189], device='cuda:0'), grad: tensor([-0.0382, 0.0243, 0.0084, -0.0157, 0.0491, 0.0085, 0.0242, -0.0754, + -0.0211, 0.0359], device='cuda:0') +100 +0.0001 +changing lr +epoch 85, time 216.38, cls_loss 0.7146 cls_loss_mapping 0.0213 cls_loss_causal 0.6146 re_mapping 0.0196 re_causal 0.0447 /// teacc 98.33 lr 0.00010000 +Epoch 87, weight, value: tensor([[-1.0724e-01, -2.9468e-02, -6.7055e-07, ..., -7.7962e-03, + -5.2578e-02, -6.9750e-02], + [-7.1714e-03, -4.2247e-02, 1.6062e-02, ..., 2.6540e-02, + 1.7369e-03, -4.2300e-02], + [-1.9485e-02, -4.2606e-02, -5.5793e-02, ..., 3.6845e-02, + -2.7250e-02, -5.1601e-02], + ..., + [-3.2938e-02, 2.3249e-02, 2.2649e-02, ..., 2.4995e-02, + -3.2311e-02, -5.5354e-02], + [-5.5380e-02, -6.4965e-03, 1.2651e-02, ..., 4.8469e-02, + -2.4331e-02, -6.7280e-02], + [ 3.0679e-02, 3.7622e-02, -2.4840e-02, ..., -7.9100e-02, + -3.0917e-02, 3.0080e-02]], device='cuda:0'), grad: tensor([[ 0.0004, 0.0017, 0.0006, ..., -0.0003, 0.0023, 0.0004], + [ 0.0009, 0.0026, 0.0012, ..., 0.0048, -0.0004, 0.0011], + [ 0.0007, 0.0011, 0.0019, ..., 0.0055, 0.0034, 0.0007], + ..., + [-0.0006, -0.0113, -0.0103, ..., -0.0147, -0.0088, -0.0027], + [ 0.0011, 0.0029, 0.0016, ..., 0.0094, 0.0059, 0.0017], + [ 0.0002, 0.0073, 0.0017, ..., 0.0021, -0.0007, 0.0088]], + device='cuda:0') +Epoch 87, bias, value: tensor([-0.0355, -0.0035, -0.0032, -0.0077, -0.0037, 0.0079, -0.0095, 0.0251, + -0.0055, 0.0194], device='cuda:0'), grad: tensor([-0.0052, 0.0219, 0.0301, -0.0540, 0.0104, -0.0274, 0.0129, -0.0520, + 0.0390, 0.0244], device='cuda:0') +100 +0.0001 +changing lr +epoch 86, time 216.58, cls_loss 0.7185 cls_loss_mapping 0.0208 cls_loss_causal 0.6082 re_mapping 0.0191 re_causal 0.0424 /// teacc 98.34 lr 0.00010000 +Epoch 88, weight, value: tensor([[-0.1079, -0.0299, 0.0014, ..., -0.0071, -0.0517, -0.0690], + [-0.0074, -0.0425, 0.0158, ..., 0.0256, 0.0013, -0.0435], + [-0.0200, -0.0428, -0.0574, ..., 0.0358, -0.0277, -0.0545], + ..., + [-0.0327, 0.0237, 0.0224, ..., 0.0259, -0.0322, -0.0548], + [-0.0555, -0.0063, 0.0134, ..., 0.0490, -0.0247, -0.0688], + [ 0.0303, 0.0378, -0.0252, ..., -0.0801, -0.0306, 0.0298]], + device='cuda:0'), grad: tensor([[ 1.0276e-04, 5.9187e-05, 4.8351e-04, ..., -1.4381e-03, + 5.7793e-03, 3.2234e-04], + [ 9.8765e-05, 2.3127e-04, 2.9159e-04, ..., 2.5034e-04, + 1.4286e-03, 1.6153e-04], + [ 9.2840e-04, 6.2141e-03, -3.1261e-03, ..., -4.1771e-03, + -5.0621e-03, 1.7214e-03], + ..., + [ 4.0269e-04, 1.7138e-03, 7.3814e-04, ..., 7.6294e-05, + 9.4557e-04, 5.7125e-04], + [-8.9264e-03, -6.0768e-03, -3.6316e-03, ..., -1.7426e-02, + 2.3880e-03, -7.2594e-03], + [ 5.2414e-03, -2.3327e-03, 1.6212e-03, ..., 1.7807e-02, + 1.6756e-03, 4.2191e-03]], device='cuda:0') +Epoch 88, bias, value: tensor([-0.0341, -0.0041, -0.0036, -0.0075, -0.0033, 0.0069, -0.0098, 0.0259, + -0.0058, 0.0192], device='cuda:0'), grad: tensor([-0.0012, 0.0088, -0.0285, 0.0220, 0.0221, -0.0276, 0.0015, 0.0024, + -0.0331, 0.0335], device='cuda:0') +100 +0.0001 +changing lr +epoch 87, time 216.83, cls_loss 0.7007 cls_loss_mapping 0.0223 cls_loss_causal 0.5995 re_mapping 0.0182 re_causal 0.0394 /// teacc 98.28 lr 0.00010000 +Epoch 89, weight, value: tensor([[-0.1087, -0.0292, 0.0009, ..., -0.0072, -0.0519, -0.0703], + [-0.0071, -0.0433, 0.0156, ..., 0.0254, 0.0017, -0.0453], + [-0.0203, -0.0417, -0.0577, ..., 0.0354, -0.0277, -0.0555], + ..., + [-0.0336, 0.0232, 0.0233, ..., 0.0262, -0.0328, -0.0549], + [-0.0550, -0.0069, 0.0137, ..., 0.0493, -0.0249, -0.0716], + [ 0.0309, 0.0376, -0.0251, ..., -0.0805, -0.0306, 0.0300]], + device='cuda:0'), grad: tensor([[ 7.9632e-05, 2.0828e-03, 6.6376e-04, ..., 1.3062e-02, + 1.1816e-03, -7.6294e-04], + [ 4.7517e-04, -2.9713e-05, -6.8617e-04, ..., -5.4512e-03, + -1.6422e-03, -2.2340e-04], + [-3.7613e-03, -7.3624e-03, 6.3934e-03, ..., -1.4465e-02, + 2.5673e-03, 1.0258e-04], + ..., + [ 8.4639e-04, 8.0414e-03, 6.7139e-04, ..., 8.3008e-03, + 3.5381e-04, 1.9765e-04], + [ 1.3123e-03, 9.1019e-03, -8.5068e-03, ..., -3.6716e-03, + -1.8024e-04, 5.6505e-04], + [-2.6435e-05, -1.2703e-02, -2.6488e-04, ..., -1.7252e-03, + -1.2369e-03, -9.3317e-04]], device='cuda:0') +Epoch 89, bias, value: tensor([-0.0345, -0.0038, -0.0034, -0.0065, -0.0033, 0.0070, -0.0102, 0.0255, + -0.0062, 0.0190], device='cuda:0'), grad: tensor([ 0.0327, -0.0249, -0.0183, -0.0260, 0.0194, 0.0155, -0.0208, 0.0085, + 0.0097, 0.0042], device='cuda:0') +100 +0.0001 +changing lr +epoch 88, time 216.12, cls_loss 0.6970 cls_loss_mapping 0.0239 cls_loss_causal 0.5981 re_mapping 0.0190 re_causal 0.0407 /// teacc 98.19 lr 0.00010000 +Epoch 90, weight, value: tensor([[-0.1109, -0.0287, 0.0018, ..., -0.0072, -0.0524, -0.0699], + [-0.0078, -0.0433, 0.0146, ..., 0.0244, 0.0014, -0.0461], + [-0.0197, -0.0426, -0.0583, ..., 0.0346, -0.0274, -0.0537], + ..., + [-0.0344, 0.0231, 0.0216, ..., 0.0262, -0.0329, -0.0553], + [-0.0545, -0.0070, 0.0160, ..., 0.0510, -0.0245, -0.0718], + [ 0.0313, 0.0375, -0.0251, ..., -0.0807, -0.0296, 0.0296]], + device='cuda:0'), grad: tensor([[ 0.0007, 0.0018, -0.0032, ..., 0.0041, -0.0006, 0.0008], + [ 0.0003, -0.0043, 0.0013, ..., -0.0110, 0.0013, 0.0004], + [ 0.0003, 0.0006, -0.0004, ..., -0.0166, -0.0067, 0.0002], + ..., + [ 0.0007, 0.0065, 0.0019, ..., 0.0147, 0.0030, 0.0021], + [ 0.0005, 0.0018, -0.0069, ..., -0.0100, -0.0082, 0.0009], + [-0.0077, -0.0129, -0.0072, ..., -0.0067, -0.0005, -0.0084]], + device='cuda:0') +Epoch 90, bias, value: tensor([-0.0349, -0.0035, -0.0029, -0.0070, -0.0034, 0.0057, -0.0100, 0.0251, + -0.0049, 0.0192], device='cuda:0'), grad: tensor([-6.0177e-04, -1.9640e-05, -5.8594e-02, 1.6632e-02, 3.5522e-02, + 1.1505e-02, 3.2959e-02, 3.4851e-02, -4.5929e-02, -2.6398e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 89, time 216.08, cls_loss 0.6936 cls_loss_mapping 0.0205 cls_loss_causal 0.5965 re_mapping 0.0194 re_causal 0.0416 /// teacc 98.21 lr 0.00010000 +Epoch 91, weight, value: tensor([[-0.1110, -0.0291, 0.0011, ..., -0.0073, -0.0520, -0.0691], + [-0.0079, -0.0446, 0.0156, ..., 0.0257, 0.0028, -0.0460], + [-0.0210, -0.0434, -0.0582, ..., 0.0347, -0.0283, -0.0539], + ..., + [-0.0347, 0.0231, 0.0220, ..., 0.0264, -0.0342, -0.0549], + [-0.0549, -0.0079, 0.0158, ..., 0.0517, -0.0247, -0.0733], + [ 0.0319, 0.0386, -0.0254, ..., -0.0817, -0.0287, 0.0301]], + device='cuda:0'), grad: tensor([[ 2.4629e-04, 6.8130e-03, 3.6373e-03, ..., 1.8578e-03, + 1.8797e-03, 6.0558e-04], + [ 2.6450e-05, 4.2248e-04, 1.4420e-03, ..., 3.7956e-03, + 1.2951e-03, 4.1574e-05], + [ 1.7595e-03, 1.0509e-03, 1.4668e-03, ..., 8.6689e-04, + -2.1591e-03, 9.3952e-06], + ..., + [-4.7035e-03, -5.5885e-04, -6.1131e-04, ..., -5.7602e-03, + -1.0271e-03, 1.4436e-04], + [ 1.0014e-03, 3.5492e-02, 1.5144e-03, ..., 3.3131e-03, + 5.2986e-03, 7.4434e-04], + [ 4.3249e-04, -4.7546e-02, 1.2226e-03, ..., 3.6163e-03, + -5.4054e-03, 1.7691e-04]], device='cuda:0') +Epoch 91, bias, value: tensor([-0.0349, -0.0030, -0.0030, -0.0073, -0.0036, 0.0059, -0.0109, 0.0253, + -0.0044, 0.0194], device='cuda:0'), grad: tensor([-0.0017, 0.0242, 0.0138, 0.0158, -0.0118, 0.0013, -0.0057, -0.0821, + 0.0340, 0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 90, time 216.78, cls_loss 0.6707 cls_loss_mapping 0.0189 cls_loss_causal 0.5890 re_mapping 0.0188 re_causal 0.0412 /// teacc 98.26 lr 0.00010000 +Epoch 92, weight, value: tensor([[-0.1120, -0.0299, 0.0018, ..., -0.0074, -0.0522, -0.0677], + [-0.0082, -0.0441, 0.0145, ..., 0.0255, 0.0020, -0.0466], + [-0.0215, -0.0431, -0.0570, ..., 0.0347, -0.0292, -0.0546], + ..., + [-0.0343, 0.0231, 0.0213, ..., 0.0261, -0.0334, -0.0542], + [-0.0544, -0.0080, 0.0156, ..., 0.0515, -0.0249, -0.0738], + [ 0.0323, 0.0388, -0.0256, ..., -0.0814, -0.0290, 0.0307]], + device='cuda:0'), grad: tensor([[ 1.8196e-03, 5.0157e-05, 1.3635e-05, ..., 3.1281e-03, + 7.5912e-04, 4.1151e-04], + [ 1.4091e-04, -1.6537e-03, 1.8314e-05, ..., -2.0386e-02, + -1.1749e-02, 1.1408e-04], + [-2.8095e-03, 4.4942e-04, 8.3596e-06, ..., 2.1839e-03, + 4.5156e-04, 2.6393e-04], + ..., + [-1.3094e-03, -2.4548e-03, 3.3855e-05, ..., 1.6129e-02, + 4.6959e-03, -9.4271e-04], + [ 5.0240e-03, 6.1226e-04, -7.5340e-04, ..., 4.6043e-03, + 8.0872e-04, 4.5509e-03], + [-5.4665e-03, -3.8834e-03, 5.2214e-05, ..., 2.1133e-03, + 1.1711e-03, -8.6823e-03]], device='cuda:0') +Epoch 92, bias, value: tensor([-0.0351, -0.0034, -0.0031, -0.0081, -0.0032, 0.0076, -0.0111, 0.0251, + -0.0048, 0.0196], device='cuda:0'), grad: tensor([ 0.0267, -0.0774, 0.0039, -0.0255, -0.0297, 0.0484, 0.0230, -0.0045, + 0.0335, 0.0016], device='cuda:0') +100 +0.0001 +changing lr +epoch 91, time 214.85, cls_loss 0.6608 cls_loss_mapping 0.0207 cls_loss_causal 0.5662 re_mapping 0.0181 re_causal 0.0389 /// teacc 98.44 lr 0.00010000 +Epoch 93, weight, value: tensor([[-0.1129, -0.0294, 0.0020, ..., -0.0082, -0.0528, -0.0679], + [-0.0087, -0.0442, 0.0148, ..., 0.0253, 0.0026, -0.0466], + [-0.0219, -0.0440, -0.0566, ..., 0.0351, -0.0294, -0.0547], + ..., + [-0.0343, 0.0233, 0.0218, ..., 0.0254, -0.0333, -0.0549], + [-0.0547, -0.0076, 0.0156, ..., 0.0514, -0.0256, -0.0752], + [ 0.0329, 0.0389, -0.0261, ..., -0.0812, -0.0292, 0.0317]], + device='cuda:0'), grad: tensor([[ 1.5205e-02, 1.3120e-05, -3.8223e-03, ..., -1.8673e-03, + 2.1591e-03, -3.7909e-04], + [ 4.5091e-05, -2.5821e-04, 1.7667e-04, ..., 5.2595e-04, + 7.4387e-04, 1.7315e-05], + [ 6.6459e-05, 2.7075e-05, 1.4618e-05, ..., 5.8031e-04, + -8.3008e-03, 1.0663e-04], + ..., + [-4.3273e-04, 1.2971e-05, -2.7865e-05, ..., -3.4561e-03, + -7.6485e-03, -2.5177e-04], + [ 1.7366e-03, -3.0613e-04, 3.4199e-03, ..., 9.4748e-04, + 3.0460e-03, 7.8964e-04], + [ 1.2505e-04, 8.4221e-05, 8.7619e-05, ..., 8.8120e-04, + 1.1673e-03, 7.3850e-05]], device='cuda:0') +Epoch 93, bias, value: tensor([-0.0361, -0.0033, -0.0026, -0.0073, -0.0035, 0.0073, -0.0115, 0.0253, + -0.0052, 0.0203], device='cuda:0'), grad: tensor([ 0.0105, 0.0091, -0.0141, -0.0165, 0.0100, -0.0111, 0.0071, -0.0259, + 0.0177, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 92---------------------------------------------------- +epoch 92, time 217.47, cls_loss 0.7193 cls_loss_mapping 0.0216 cls_loss_causal 0.6199 re_mapping 0.0189 re_causal 0.0411 /// teacc 98.53 lr 0.00010000 +Epoch 94, weight, value: tensor([[-0.1136, -0.0292, 0.0037, ..., -0.0078, -0.0527, -0.0685], + [-0.0092, -0.0448, 0.0150, ..., 0.0252, 0.0023, -0.0462], + [-0.0214, -0.0446, -0.0578, ..., 0.0353, -0.0298, -0.0559], + ..., + [-0.0359, 0.0237, 0.0214, ..., 0.0256, -0.0331, -0.0558], + [-0.0548, -0.0071, 0.0140, ..., 0.0510, -0.0255, -0.0764], + [ 0.0329, 0.0387, -0.0259, ..., -0.0819, -0.0293, 0.0333]], + device='cuda:0'), grad: tensor([[-2.0752e-03, 1.5116e-04, -5.2299e-03, ..., -2.3785e-03, + 6.1750e-04, 5.1320e-05], + [ 3.5419e-03, 5.3358e-04, 7.3662e-03, ..., 3.5362e-03, + 1.4366e-02, 1.3582e-05], + [-1.1200e-04, 1.5764e-03, -5.7259e-03, ..., -4.0665e-03, + -8.9359e-04, 1.6046e-04], + ..., + [ 2.5988e-04, -3.4351e-03, 6.0797e-04, ..., -1.7757e-03, + -2.1591e-03, 1.1456e-04], + [ 2.1133e-03, 5.8413e-04, 2.2087e-03, ..., 1.7681e-03, + 6.7673e-03, 2.1827e-04], + [-2.9030e-03, 3.9983e-04, -2.0933e-04, ..., 9.9277e-04, + 1.7366e-03, -9.5272e-04]], device='cuda:0') +Epoch 94, bias, value: tensor([-0.0360, -0.0034, -0.0024, -0.0076, -0.0030, 0.0076, -0.0117, 0.0257, + -0.0050, 0.0193], device='cuda:0'), grad: tensor([ 0.0045, 0.0222, -0.0128, 0.0028, -0.0160, -0.0398, -0.0012, -0.0079, + 0.0396, 0.0086], device='cuda:0') +100 +0.0001 +changing lr +epoch 93, time 215.46, cls_loss 0.7183 cls_loss_mapping 0.0248 cls_loss_causal 0.6240 re_mapping 0.0182 re_causal 0.0400 /// teacc 98.28 lr 0.00010000 +Epoch 95, weight, value: tensor([[-0.1148, -0.0302, 0.0043, ..., -0.0092, -0.0539, -0.0707], + [-0.0076, -0.0453, 0.0142, ..., 0.0250, 0.0021, -0.0473], + [-0.0223, -0.0448, -0.0585, ..., 0.0360, -0.0295, -0.0555], + ..., + [-0.0362, 0.0239, 0.0205, ..., 0.0265, -0.0323, -0.0569], + [-0.0547, -0.0061, 0.0133, ..., 0.0507, -0.0258, -0.0753], + [ 0.0334, 0.0383, -0.0239, ..., -0.0826, -0.0298, 0.0332]], + device='cuda:0'), grad: tensor([[ 1.1295e-04, -2.4834e-03, 7.2718e-04, ..., 1.2922e-03, + 5.9471e-03, -2.2829e-04], + [ 1.3363e-04, 1.6861e-03, 7.6771e-04, ..., 1.9388e-03, + 3.2654e-03, 5.9545e-05], + [ 2.0428e-03, 1.3218e-03, -7.7677e-04, ..., -1.3878e-02, + -4.6326e-02, 1.5348e-05], + ..., + [ 3.9673e-03, -1.8530e-03, 8.9550e-04, ..., -1.1921e-04, + 1.8501e-03, 2.9793e-03], + [ 1.0347e-03, 3.1662e-03, 5.4240e-05, ..., 3.1948e-03, + -2.0218e-04, 5.4169e-04], + [-7.8278e-03, 1.9684e-03, 1.0490e-03, ..., 8.8739e-04, + 6.3057e-03, -5.4283e-03]], device='cuda:0') +Epoch 95, bias, value: tensor([-0.0378, -0.0041, -0.0016, -0.0072, -0.0013, 0.0074, -0.0114, 0.0255, + -0.0050, 0.0189], device='cuda:0'), grad: tensor([ 0.0228, 0.0264, -0.0709, 0.0294, -0.0582, 0.0433, 0.0167, 0.0117, + -0.0240, 0.0027], device='cuda:0') +100 +0.0001 +changing lr +epoch 94, time 216.72, cls_loss 0.6897 cls_loss_mapping 0.0193 cls_loss_causal 0.5862 re_mapping 0.0181 re_causal 0.0400 /// teacc 98.32 lr 0.00010000 +Epoch 96, weight, value: tensor([[-0.1147, -0.0303, 0.0041, ..., -0.0091, -0.0549, -0.0706], + [-0.0083, -0.0447, 0.0135, ..., 0.0253, 0.0020, -0.0487], + [-0.0234, -0.0451, -0.0602, ..., 0.0357, -0.0311, -0.0557], + ..., + [-0.0365, 0.0243, 0.0243, ..., 0.0260, -0.0319, -0.0577], + [-0.0540, -0.0059, 0.0133, ..., 0.0511, -0.0264, -0.0759], + [ 0.0332, 0.0385, -0.0241, ..., -0.0828, -0.0297, 0.0345]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0006, 0.0005, ..., -0.0026, 0.0008, 0.0009], + [ 0.0001, 0.0016, -0.0026, ..., 0.0001, 0.0016, 0.0004], + [ 0.0004, -0.0061, 0.0006, ..., 0.0031, -0.0019, 0.0013], + ..., + [ 0.0002, 0.0012, 0.0001, ..., 0.0028, 0.0008, -0.0045], + [ 0.0005, -0.0052, 0.0007, ..., -0.0016, -0.0029, 0.0019], + [ 0.0002, -0.0017, 0.0001, ..., -0.0032, -0.0037, 0.0024]], + device='cuda:0') +Epoch 96, bias, value: tensor([-0.0386, -0.0041, -0.0018, -0.0067, -0.0020, 0.0075, -0.0103, 0.0264, + -0.0050, 0.0178], device='cuda:0'), grad: tensor([-0.0141, 0.0150, -0.0113, -0.0471, 0.0323, 0.0197, 0.0198, -0.0030, + -0.0081, -0.0033], device='cuda:0') +100 +0.0001 +changing lr +epoch 95, time 214.85, cls_loss 0.6862 cls_loss_mapping 0.0173 cls_loss_causal 0.5946 re_mapping 0.0181 re_causal 0.0402 /// teacc 98.44 lr 0.00010000 +Epoch 97, weight, value: tensor([[-0.1157, -0.0302, 0.0037, ..., -0.0095, -0.0557, -0.0711], + [-0.0072, -0.0448, 0.0136, ..., 0.0252, 0.0021, -0.0488], + [-0.0252, -0.0447, -0.0610, ..., 0.0366, -0.0308, -0.0575], + ..., + [-0.0360, 0.0243, 0.0242, ..., 0.0265, -0.0329, -0.0574], + [-0.0542, -0.0069, 0.0143, ..., 0.0513, -0.0264, -0.0771], + [ 0.0332, 0.0387, -0.0242, ..., -0.0837, -0.0304, 0.0350]], + device='cuda:0'), grad: tensor([[-9.3794e-04, 9.1600e-04, 1.4553e-03, ..., 4.8904e-03, + 2.3384e-03, 3.1501e-05], + [ 2.0528e-04, -4.8370e-03, 1.2980e-03, ..., -1.5535e-03, + 2.5368e-03, 9.3460e-05], + [ 1.1168e-03, 1.2293e-03, 3.0155e-03, ..., 7.3767e-04, + 1.7700e-02, 5.8556e-04], + ..., + [-1.0115e-04, 4.1294e-04, 1.2474e-03, ..., 2.1172e-03, + 3.1033e-03, -6.2084e-04], + [-9.2888e-04, 1.1959e-03, 6.4898e-04, ..., 3.0022e-03, + 1.7767e-03, 3.8910e-04], + [-8.1873e-04, 1.6403e-03, 9.9850e-04, ..., 4.1466e-03, + 4.3182e-03, -7.1859e-04]], device='cuda:0') +Epoch 97, bias, value: tensor([-0.0383, -0.0039, -0.0020, -0.0065, -0.0019, 0.0074, -0.0103, 0.0262, + -0.0043, 0.0168], device='cuda:0'), grad: tensor([ 0.0164, 0.0009, -0.0091, 0.0259, -0.0249, -0.0083, -0.0512, 0.0195, + 0.0044, 0.0263], device='cuda:0') +100 +0.0001 +changing lr +epoch 96, time 217.11, cls_loss 0.7004 cls_loss_mapping 0.0160 cls_loss_causal 0.5914 re_mapping 0.0186 re_causal 0.0427 /// teacc 98.51 lr 0.00010000 +Epoch 98, weight, value: tensor([[-0.1151, -0.0296, 0.0048, ..., -0.0091, -0.0572, -0.0694], + [-0.0070, -0.0458, 0.0147, ..., 0.0251, 0.0022, -0.0507], + [-0.0254, -0.0445, -0.0617, ..., 0.0367, -0.0306, -0.0579], + ..., + [-0.0361, 0.0243, 0.0236, ..., 0.0267, -0.0350, -0.0574], + [-0.0546, -0.0069, 0.0139, ..., 0.0521, -0.0271, -0.0781], + [ 0.0337, 0.0391, -0.0243, ..., -0.0845, -0.0307, 0.0347]], + device='cuda:0'), grad: tensor([[ 0.0019, -0.0053, -0.0007, ..., -0.0010, -0.0011, -0.0033], + [-0.0003, 0.0002, -0.0005, ..., -0.0009, 0.0016, 0.0004], + [ 0.0006, 0.0018, 0.0007, ..., 0.0013, 0.0054, 0.0016], + ..., + [ 0.0041, -0.0016, 0.0038, ..., 0.0033, 0.0019, 0.0004], + [ 0.0016, 0.0017, 0.0015, ..., 0.0011, 0.0050, 0.0020], + [-0.0079, 0.0026, -0.0064, ..., -0.0027, 0.0026, -0.0035]], + device='cuda:0') +Epoch 98, bias, value: tensor([-0.0379, -0.0038, -0.0011, -0.0068, -0.0023, 0.0079, -0.0110, 0.0255, + -0.0048, 0.0175], device='cuda:0'), grad: tensor([-0.0295, 0.0069, 0.0254, -0.0107, 0.0289, 0.0106, -0.0622, 0.0207, + 0.0238, -0.0139], device='cuda:0') +100 +0.0001 +changing lr +epoch 97, time 214.99, cls_loss 0.7029 cls_loss_mapping 0.0180 cls_loss_causal 0.6155 re_mapping 0.0165 re_causal 0.0384 /// teacc 98.52 lr 0.00010000 +Epoch 99, weight, value: tensor([[-0.1148, -0.0302, 0.0035, ..., -0.0084, -0.0559, -0.0695], + [-0.0069, -0.0461, 0.0149, ..., 0.0251, 0.0024, -0.0510], + [-0.0256, -0.0446, -0.0616, ..., 0.0365, -0.0311, -0.0590], + ..., + [-0.0366, 0.0249, 0.0243, ..., 0.0271, -0.0363, -0.0583], + [-0.0554, -0.0069, 0.0136, ..., 0.0519, -0.0276, -0.0782], + [ 0.0337, 0.0390, -0.0249, ..., -0.0847, -0.0309, 0.0341]], + device='cuda:0'), grad: tensor([[ 0.0029, 0.0004, 0.0019, ..., 0.0033, 0.0029, 0.0009], + [ 0.0042, -0.0020, 0.0045, ..., 0.0014, -0.0048, 0.0015], + [ 0.0057, 0.0006, 0.0011, ..., 0.0054, 0.0017, 0.0011], + ..., + [-0.0042, -0.0034, -0.0053, ..., -0.0166, -0.0091, -0.0050], + [-0.0065, 0.0009, -0.0100, ..., -0.0083, 0.0021, -0.0013], + [-0.0317, 0.0017, 0.0009, ..., 0.0030, 0.0060, 0.0024]], + device='cuda:0') +Epoch 99, bias, value: tensor([-0.0377, -0.0037, -0.0023, -0.0068, -0.0015, 0.0074, -0.0109, 0.0259, + -0.0048, 0.0176], device='cuda:0'), grad: tensor([ 0.0343, -0.0289, 0.0315, -0.0082, 0.0022, 0.0417, 0.0212, -0.0671, + -0.0186, -0.0080], device='cuda:0') +100 +0.0001 +changing lr +epoch 98, time 214.75, cls_loss 0.6638 cls_loss_mapping 0.0136 cls_loss_causal 0.5603 re_mapping 0.0175 re_causal 0.0386 /// teacc 98.44 lr 0.00010000 +Epoch 100, weight, value: tensor([[-0.1150, -0.0307, 0.0039, ..., -0.0086, -0.0557, -0.0700], + [-0.0068, -0.0460, 0.0151, ..., 0.0245, 0.0024, -0.0498], + [-0.0257, -0.0442, -0.0610, ..., 0.0371, -0.0327, -0.0591], + ..., + [-0.0368, 0.0242, 0.0262, ..., 0.0269, -0.0362, -0.0579], + [-0.0554, -0.0065, 0.0142, ..., 0.0520, -0.0286, -0.0789], + [ 0.0343, 0.0393, -0.0251, ..., -0.0846, -0.0292, 0.0343]], + device='cuda:0'), grad: tensor([[ 2.7347e-04, 1.4870e-02, 7.5722e-04, ..., 3.0956e-03, + 3.4607e-02, 6.2847e-04], + [ 5.3167e-05, -6.8188e-04, 8.9598e-04, ..., 4.9248e-03, + 1.2093e-02, 2.2650e-04], + [ 7.8619e-05, 1.4858e-03, 3.9625e-04, ..., 3.4580e-03, + 3.8280e-03, 2.5582e-04], + ..., + [ 8.2433e-05, -1.6623e-03, 4.2081e-04, ..., -1.4849e-03, + 3.2425e-03, 9.6798e-04], + [ 3.0780e-04, -1.3527e-02, -2.2469e-03, ..., -6.6681e-03, + -4.5807e-02, 1.3065e-03], + [ 4.5002e-06, -1.4887e-03, 8.5402e-04, ..., 2.2640e-03, + 3.7613e-03, -3.6201e-03]], device='cuda:0') +Epoch 100, bias, value: tensor([-0.0379, -0.0037, -0.0018, -0.0072, -0.0024, 0.0071, -0.0099, 0.0269, + -0.0050, 0.0173], device='cuda:0'), grad: tensor([ 0.0532, 0.0428, 0.0182, 0.0108, -0.0126, -0.0379, -0.0109, -0.0133, + -0.0615, 0.0113], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 99---------------------------------------------------- +epoch 99, time 214.94, cls_loss 0.6294 cls_loss_mapping 0.0143 cls_loss_causal 0.5420 re_mapping 0.0177 re_causal 0.0384 /// teacc 98.55 lr 0.00010000 +Epoch 101, weight, value: tensor([[-0.1149, -0.0308, 0.0043, ..., -0.0086, -0.0564, -0.0707], + [-0.0075, -0.0459, 0.0172, ..., 0.0232, 0.0026, -0.0505], + [-0.0265, -0.0458, -0.0622, ..., 0.0374, -0.0345, -0.0601], + ..., + [-0.0357, 0.0244, 0.0260, ..., 0.0273, -0.0340, -0.0573], + [-0.0555, -0.0062, 0.0133, ..., 0.0517, -0.0283, -0.0794], + [ 0.0335, 0.0396, -0.0267, ..., -0.0852, -0.0290, 0.0342]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0004, 0.0002, ..., -0.0055, 0.0044, -0.0002], + [-0.0026, -0.0219, -0.0016, ..., -0.0181, -0.0042, 0.0001], + [ 0.0007, 0.0017, 0.0004, ..., 0.0048, 0.0016, 0.0002], + ..., + [ 0.0006, 0.0033, 0.0006, ..., 0.0058, 0.0010, 0.0008], + [ 0.0026, 0.0036, 0.0017, ..., 0.0065, 0.0026, 0.0002], + [-0.0009, 0.0034, -0.0018, ..., 0.0042, 0.0019, -0.0041]], + device='cuda:0') +Epoch 101, bias, value: tensor([-0.0385, -0.0036, -0.0023, -0.0065, -0.0024, 0.0060, -0.0093, 0.0275, + -0.0056, 0.0180], device='cuda:0'), grad: tensor([-0.0293, -0.0614, 0.0274, 0.0005, 0.0162, 0.0355, -0.0641, 0.0377, + 0.0125, 0.0251], device='cuda:0') +100 +0.0001 +changing lr +epoch 100, time 214.40, cls_loss 0.6868 cls_loss_mapping 0.0143 cls_loss_causal 0.5939 re_mapping 0.0174 re_causal 0.0396 /// teacc 98.21 lr 0.00010000 +Epoch 102, weight, value: tensor([[-0.1142, -0.0311, 0.0046, ..., -0.0082, -0.0563, -0.0710], + [-0.0070, -0.0457, 0.0176, ..., 0.0244, 0.0025, -0.0496], + [-0.0274, -0.0469, -0.0635, ..., 0.0379, -0.0338, -0.0625], + ..., + [-0.0359, 0.0240, 0.0263, ..., 0.0270, -0.0342, -0.0571], + [-0.0556, -0.0056, 0.0135, ..., 0.0520, -0.0279, -0.0805], + [ 0.0336, 0.0390, -0.0264, ..., -0.0857, -0.0303, 0.0341]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0003, 0.0021, ..., 0.0084, 0.0110, 0.0013], + [ 0.0016, -0.0013, 0.0017, ..., 0.0021, 0.0014, 0.0035], + [ 0.0004, 0.0012, 0.0008, ..., -0.0021, -0.0060, 0.0020], + ..., + [-0.0015, -0.0209, 0.0013, ..., -0.0040, 0.0009, -0.0136], + [ 0.0021, 0.0017, 0.0009, ..., 0.0051, 0.0056, -0.0039], + [-0.0034, 0.0140, 0.0007, ..., 0.0036, 0.0075, 0.0104]], + device='cuda:0') +Epoch 102, bias, value: tensor([-0.0385, -0.0029, -0.0018, -0.0065, -0.0030, 0.0055, -0.0101, 0.0270, + -0.0049, 0.0182], device='cuda:0'), grad: tensor([ 0.0229, -0.0009, -0.0113, 0.0204, -0.0734, 0.0299, -0.0028, -0.0177, + -0.0109, 0.0438], device='cuda:0') +100 +0.0001 +changing lr +epoch 101, time 214.93, cls_loss 0.7027 cls_loss_mapping 0.0173 cls_loss_causal 0.5938 re_mapping 0.0179 re_causal 0.0392 /// teacc 98.40 lr 0.00010000 +Epoch 103, weight, value: tensor([[-0.1137, -0.0303, 0.0054, ..., -0.0097, -0.0561, -0.0725], + [-0.0072, -0.0465, 0.0190, ..., 0.0248, 0.0032, -0.0491], + [-0.0282, -0.0479, -0.0642, ..., 0.0390, -0.0333, -0.0626], + ..., + [-0.0365, 0.0241, 0.0251, ..., 0.0266, -0.0342, -0.0576], + [-0.0541, -0.0052, 0.0136, ..., 0.0532, -0.0277, -0.0804], + [ 0.0328, 0.0398, -0.0267, ..., -0.0860, -0.0311, 0.0338]], + device='cuda:0'), grad: tensor([[-2.1801e-03, 1.8320e-03, 2.0313e-04, ..., 1.6083e-02, + 6.8665e-03, -2.1982e-04], + [ 4.1986e-04, 1.0475e-02, 1.3027e-03, ..., 7.2136e-03, + 2.3823e-03, 5.6362e-04], + [ 1.0090e-03, 4.5323e-04, -6.3419e-05, ..., 1.7033e-03, + 1.2388e-03, -1.1740e-03], + ..., + [ 1.3123e-03, 3.6182e-03, 3.3116e-04, ..., -1.3687e-02, + -1.2573e-02, 1.3561e-03], + [ 1.3666e-03, 2.8658e-04, -4.9667e-03, ..., -3.7575e-03, + -6.3438e-03, 1.1072e-03], + [-3.2101e-03, 4.0483e-04, 3.0303e-04, ..., -1.2413e-02, + 1.1358e-03, 1.4992e-03]], device='cuda:0') +Epoch 103, bias, value: tensor([-0.0399, -0.0019, -0.0015, -0.0061, -0.0033, 0.0057, -0.0100, 0.0266, + -0.0042, 0.0175], device='cuda:0'), grad: tensor([ 0.0384, 0.0323, 0.0011, -0.0006, -0.0346, 0.0349, 0.0252, -0.0361, + -0.0009, -0.0596], device='cuda:0') +100 +0.0001 +changing lr +epoch 102, time 215.87, cls_loss 0.6728 cls_loss_mapping 0.0186 cls_loss_causal 0.5734 re_mapping 0.0171 re_causal 0.0386 /// teacc 98.26 lr 0.00010000 +Epoch 104, weight, value: tensor([[-0.1141, -0.0295, 0.0059, ..., -0.0087, -0.0560, -0.0722], + [-0.0073, -0.0475, 0.0188, ..., 0.0245, 0.0027, -0.0484], + [-0.0288, -0.0474, -0.0657, ..., 0.0400, -0.0331, -0.0642], + ..., + [-0.0376, 0.0244, 0.0265, ..., 0.0260, -0.0345, -0.0583], + [-0.0538, -0.0052, 0.0138, ..., 0.0523, -0.0281, -0.0812], + [ 0.0332, 0.0397, -0.0268, ..., -0.0862, -0.0315, 0.0352]], + device='cuda:0'), grad: tensor([[ 3.4404e-04, 4.6581e-05, 1.0252e-03, ..., 1.1396e-03, + 3.5930e-04, 5.5838e-04], + [ 1.7476e-04, 6.9141e-05, 4.6134e-04, ..., 8.0395e-04, + 8.9407e-04, 2.9135e-04], + [ 2.5225e-04, 1.2290e-04, 2.0351e-03, ..., 1.2903e-03, + 2.9445e-04, 1.3885e-03], + ..., + [-4.5514e-04, 2.2385e-02, -8.0948e-03, ..., -4.1313e-03, + -1.9159e-03, -3.7537e-03], + [-8.3237e-03, 4.9680e-05, 8.9312e-04, ..., -6.2561e-03, + -9.5673e-03, 5.6696e-04], + [-3.6278e-03, -2.3453e-02, -5.9052e-03, ..., -2.8114e-03, + 5.4508e-05, -3.7231e-03]], device='cuda:0') +Epoch 104, bias, value: tensor([-0.0386, -0.0030, -0.0017, -0.0058, -0.0028, 0.0055, -0.0098, 0.0267, + -0.0047, 0.0174], device='cuda:0'), grad: tensor([ 0.0111, 0.0096, 0.0121, 0.0126, 0.0240, 0.0596, 0.0045, -0.0122, + -0.0787, -0.0427], device='cuda:0') +100 +0.0001 +changing lr +epoch 103, time 215.64, cls_loss 0.6844 cls_loss_mapping 0.0139 cls_loss_causal 0.6019 re_mapping 0.0177 re_causal 0.0397 /// teacc 98.30 lr 0.00010000 +Epoch 105, weight, value: tensor([[-0.1147, -0.0302, 0.0065, ..., -0.0095, -0.0568, -0.0729], + [-0.0074, -0.0489, 0.0190, ..., 0.0245, 0.0031, -0.0487], + [-0.0291, -0.0474, -0.0656, ..., 0.0401, -0.0334, -0.0646], + ..., + [-0.0372, 0.0251, 0.0272, ..., 0.0265, -0.0337, -0.0578], + [-0.0523, -0.0049, 0.0138, ..., 0.0531, -0.0285, -0.0814], + [ 0.0332, 0.0396, -0.0270, ..., -0.0862, -0.0320, 0.0341]], + device='cuda:0'), grad: tensor([[-1.4806e-04, 1.4484e-04, 6.0892e-04, ..., 4.1962e-03, + 2.9087e-03, -1.1808e-04], + [ 2.2650e-05, -1.9169e-03, 8.0585e-04, ..., 3.7518e-03, + 2.6608e-03, 3.5077e-05], + [ 8.8990e-05, 1.2627e-03, 4.7874e-03, ..., 1.3847e-02, + 3.2063e-03, 8.0729e-04], + ..., + [ 3.4924e-03, 1.5182e-03, 1.5701e-02, ..., 3.0589e-04, + 2.2106e-03, 2.8858e-03], + [ 1.1055e-02, 7.2670e-04, -7.9489e-04, ..., -4.9591e-03, + -2.2736e-03, 9.8114e-03], + [-4.1885e-03, -1.5345e-03, -1.9394e-02, ..., -1.3054e-02, + -1.8749e-03, -3.3951e-03]], device='cuda:0') +Epoch 105, bias, value: tensor([-0.0394, -0.0032, -0.0012, -0.0067, -0.0029, 0.0058, -0.0097, 0.0272, + -0.0049, 0.0181], device='cuda:0'), grad: tensor([ 0.0223, 0.0282, 0.0834, -0.0823, -0.0051, 0.0217, 0.0031, -0.0499, + 0.0105, -0.0320], device='cuda:0') +100 +0.0001 +changing lr +epoch 104, time 215.56, cls_loss 0.6399 cls_loss_mapping 0.0201 cls_loss_causal 0.5514 re_mapping 0.0173 re_causal 0.0372 /// teacc 98.32 lr 0.00010000 +Epoch 106, weight, value: tensor([[-0.1149, -0.0295, 0.0061, ..., -0.0099, -0.0569, -0.0736], + [-0.0089, -0.0493, 0.0202, ..., 0.0246, 0.0022, -0.0486], + [-0.0298, -0.0476, -0.0665, ..., 0.0399, -0.0331, -0.0646], + ..., + [-0.0372, 0.0251, 0.0273, ..., 0.0274, -0.0340, -0.0575], + [-0.0529, -0.0067, 0.0131, ..., 0.0532, -0.0299, -0.0826], + [ 0.0334, 0.0398, -0.0279, ..., -0.0863, -0.0315, 0.0339]], + device='cuda:0'), grad: tensor([[ 1.2028e-04, 1.1187e-03, 7.0381e-04, ..., 8.2636e-04, + 4.1175e-04, 4.5371e-04], + [ 1.9908e-05, -1.8051e-02, -2.8191e-03, ..., -1.5312e-02, + -5.1804e-03, 1.5414e-04], + [ 6.5756e-04, 4.8523e-03, 6.9952e-04, ..., 3.0270e-03, + 2.0123e-03, 2.2945e-03], + ..., + [ 6.6876e-05, 1.2398e-02, -4.1485e-04, ..., 2.2297e-03, + 5.0402e-04, 7.6218e-03], + [ 1.0198e-04, 3.8738e-03, 3.0875e-04, ..., 1.7214e-03, + 3.7241e-04, 3.3355e-04], + [ 4.2856e-05, -4.1466e-03, 9.9564e-04, ..., 1.4915e-03, + 3.2210e-04, -6.9695e-03]], device='cuda:0') +Epoch 106, bias, value: tensor([-0.0398, -0.0029, -0.0004, -0.0063, -0.0026, 0.0061, -0.0101, 0.0269, + -0.0060, 0.0182], device='cuda:0'), grad: tensor([ 0.0113, -0.0702, 0.0448, 0.0088, 0.0071, -0.0244, -0.0029, 0.0116, + 0.0111, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 105, time 216.10, cls_loss 0.6704 cls_loss_mapping 0.0185 cls_loss_causal 0.5777 re_mapping 0.0167 re_causal 0.0365 /// teacc 98.34 lr 0.00010000 +Epoch 107, weight, value: tensor([[-0.1155, -0.0303, 0.0061, ..., -0.0104, -0.0571, -0.0733], + [-0.0089, -0.0488, 0.0214, ..., 0.0256, 0.0022, -0.0479], + [-0.0303, -0.0476, -0.0662, ..., 0.0400, -0.0330, -0.0659], + ..., + [-0.0376, 0.0249, 0.0268, ..., 0.0270, -0.0344, -0.0575], + [-0.0527, -0.0057, 0.0133, ..., 0.0531, -0.0295, -0.0829], + [ 0.0338, 0.0398, -0.0285, ..., -0.0862, -0.0317, 0.0343]], + device='cuda:0'), grad: tensor([[ 7.3719e-04, 7.0953e-04, 7.6532e-04, ..., 1.3371e-03, + 1.2338e-04, 9.8515e-04], + [ 1.0023e-03, 1.1978e-03, 1.5659e-03, ..., 3.9268e-04, + 7.8529e-06, 5.7697e-04], + [-7.8964e-04, -1.2894e-02, -1.8463e-03, ..., -4.1924e-03, + 8.5533e-06, -4.0512e-03], + ..., + [-8.7814e-03, 6.9695e-03, -8.7204e-03, ..., -8.3828e-04, + -1.0262e-03, -6.8779e-03], + [ 1.3838e-03, 7.0457e-03, 1.2474e-03, ..., -6.3276e-04, + 1.8346e-04, 1.9426e-03], + [ 3.0041e-03, -2.2812e-03, 1.5745e-03, ..., 6.0225e-04, + 5.5170e-04, 6.7673e-03]], device='cuda:0') +Epoch 107, bias, value: tensor([-0.0392, -0.0024, -0.0009, -0.0072, -0.0028, 0.0060, -0.0103, 0.0268, + -0.0055, 0.0185], device='cuda:0'), grad: tensor([ 0.0314, 0.0065, -0.0688, 0.0398, -0.0086, 0.0085, -0.0159, -0.0161, + 0.0059, 0.0172], device='cuda:0') +100 +0.0001 +changing lr +epoch 106, time 215.91, cls_loss 0.6837 cls_loss_mapping 0.0138 cls_loss_causal 0.5957 re_mapping 0.0170 re_causal 0.0382 /// teacc 98.48 lr 0.00010000 +Epoch 108, weight, value: tensor([[-0.1148, -0.0300, 0.0062, ..., -0.0114, -0.0578, -0.0735], + [-0.0092, -0.0508, 0.0213, ..., 0.0258, 0.0012, -0.0486], + [-0.0306, -0.0475, -0.0656, ..., 0.0401, -0.0331, -0.0647], + ..., + [-0.0378, 0.0251, 0.0264, ..., 0.0266, -0.0345, -0.0571], + [-0.0526, -0.0051, 0.0129, ..., 0.0533, -0.0296, -0.0832], + [ 0.0336, 0.0402, -0.0283, ..., -0.0873, -0.0303, 0.0341]], + device='cuda:0'), grad: tensor([[ 4.8351e-04, 6.9275e-03, 5.0354e-04, ..., 6.0539e-03, + 4.2605e-04, 2.7676e-03], + [ 2.6059e-04, 6.0320e-04, 3.6526e-03, ..., -3.9597e-03, + -2.3041e-03, 1.6463e-04], + [ 2.9874e-04, 5.1003e-03, -2.8229e-03, ..., -1.0338e-02, + -1.4341e-04, 3.9673e-04], + ..., + [ 5.2223e-03, 6.8843e-05, 1.2856e-03, ..., 1.7242e-03, + -5.7793e-04, 3.7937e-03], + [ 1.3599e-03, -2.1866e-02, 9.0599e-04, ..., 4.5433e-03, + 8.5831e-04, 7.2527e-04], + [-9.0256e-03, 3.5076e-03, -5.5275e-03, ..., -1.4137e-02, + -1.1854e-03, -5.3291e-03]], device='cuda:0') +Epoch 108, bias, value: tensor([-0.0399, -0.0026, -0.0009, -0.0072, -0.0028, 0.0063, -0.0099, 0.0264, + -0.0052, 0.0188], device='cuda:0'), grad: tensor([ 0.0377, 0.0062, -0.0158, 0.0165, 0.0155, -0.0009, 0.0106, 0.0119, + 0.0032, -0.0848], device='cuda:0') +100 +0.0001 +changing lr +epoch 107, time 216.75, cls_loss 0.6623 cls_loss_mapping 0.0187 cls_loss_causal 0.5805 re_mapping 0.0170 re_causal 0.0376 /// teacc 98.26 lr 0.00010000 +Epoch 109, weight, value: tensor([[-0.1157, -0.0299, 0.0061, ..., -0.0121, -0.0584, -0.0756], + [-0.0094, -0.0504, 0.0211, ..., 0.0261, 0.0017, -0.0517], + [-0.0313, -0.0487, -0.0650, ..., 0.0396, -0.0331, -0.0652], + ..., + [-0.0373, 0.0252, 0.0268, ..., 0.0263, -0.0343, -0.0575], + [-0.0522, -0.0034, 0.0140, ..., 0.0539, -0.0301, -0.0837], + [ 0.0332, 0.0406, -0.0296, ..., -0.0885, -0.0303, 0.0351]], + device='cuda:0'), grad: tensor([[ 5.5730e-05, 5.4789e-04, 6.4850e-04, ..., 2.3327e-03, + 9.7322e-04, 2.1935e-04], + [-1.2505e-04, 3.0327e-04, 5.9462e-04, ..., 2.9469e-03, + 9.0170e-04, 3.6383e-04], + [ 8.8394e-05, 1.5414e-04, -4.3526e-03, ..., -3.8433e-03, + -4.9744e-03, 4.6730e-04], + ..., + [ 6.8307e-05, 1.6940e-04, 3.8218e-04, ..., 1.9913e-03, + 1.5664e-04, 2.5058e-04], + [ 7.2956e-04, -9.6273e-04, 6.4087e-04, ..., -3.4523e-03, + 5.9366e-04, 1.8530e-03], + [-2.4853e-03, 3.4332e-04, -1.2268e-02, ..., 2.2888e-03, + 3.9840e-04, -1.1238e-02]], device='cuda:0') +Epoch 109, bias, value: tensor([-0.0407, -0.0026, -0.0013, -0.0071, -0.0025, 0.0063, -0.0099, 0.0265, + -0.0045, 0.0186], device='cuda:0'), grad: tensor([ 0.0191, 0.0193, 0.0010, -0.0158, 0.0504, -0.0050, -0.0175, 0.0094, + -0.0148, -0.0460], device='cuda:0') +100 +0.0001 +changing lr +epoch 108, time 215.41, cls_loss 0.6649 cls_loss_mapping 0.0136 cls_loss_causal 0.5623 re_mapping 0.0167 re_causal 0.0368 /// teacc 98.45 lr 0.00010000 +Epoch 110, weight, value: tensor([[-0.1169, -0.0307, 0.0063, ..., -0.0120, -0.0590, -0.0772], + [-0.0096, -0.0502, 0.0211, ..., 0.0260, 0.0009, -0.0529], + [-0.0301, -0.0481, -0.0636, ..., 0.0397, -0.0334, -0.0640], + ..., + [-0.0371, 0.0251, 0.0269, ..., 0.0270, -0.0341, -0.0578], + [-0.0531, -0.0028, 0.0127, ..., 0.0526, -0.0302, -0.0838], + [ 0.0339, 0.0405, -0.0291, ..., -0.0889, -0.0300, 0.0358]], + device='cuda:0'), grad: tensor([[ 1.2608e-03, 5.7936e-04, 1.7185e-03, ..., 2.1477e-03, + 1.6594e-03, 8.7929e-04], + [ 2.6369e-04, -7.6914e-04, -8.5878e-04, ..., -1.6785e-03, + 1.1425e-03, 2.7156e-04], + [ 6.5804e-05, 3.9363e-04, 2.4090e-03, ..., -5.7983e-04, + -2.4948e-03, 2.5005e-03], + ..., + [ 1.0767e-03, 3.3379e-03, -2.8973e-03, ..., 1.7414e-03, + 2.0142e-03, 1.3342e-03], + [-3.7823e-03, 2.4261e-03, -3.2544e-04, ..., -2.8877e-03, + 1.7185e-03, -4.9210e-04], + [-8.2159e-04, -1.1719e-02, -3.4027e-03, ..., -4.1199e-03, + -2.7866e-03, -3.5057e-03]], device='cuda:0') +Epoch 110, bias, value: tensor([-0.0414, -0.0024, -0.0013, -0.0066, -0.0028, 0.0063, -0.0099, 0.0266, + -0.0050, 0.0191], device='cuda:0'), grad: tensor([ 0.0266, -0.0168, -0.0229, -0.0062, 0.0277, 0.0210, 0.0173, -0.0046, + -0.0095, -0.0327], device='cuda:0') +100 +0.0001 +changing lr +epoch 109, time 215.74, cls_loss 0.6645 cls_loss_mapping 0.0134 cls_loss_causal 0.5674 re_mapping 0.0165 re_causal 0.0371 /// teacc 98.40 lr 0.00010000 +Epoch 111, weight, value: tensor([[-0.1178, -0.0299, 0.0066, ..., -0.0117, -0.0578, -0.0773], + [-0.0093, -0.0505, 0.0205, ..., 0.0265, 0.0006, -0.0533], + [-0.0304, -0.0487, -0.0633, ..., 0.0392, -0.0349, -0.0646], + ..., + [-0.0376, 0.0259, 0.0270, ..., 0.0274, -0.0324, -0.0582], + [-0.0538, -0.0041, 0.0128, ..., 0.0525, -0.0316, -0.0843], + [ 0.0329, 0.0402, -0.0290, ..., -0.0894, -0.0292, 0.0353]], + device='cuda:0'), grad: tensor([[ 2.1148e-04, 4.1437e-04, -6.1803e-06, ..., 4.1428e-03, + 5.5275e-03, 1.0128e-03], + [ 1.7154e-04, 3.7918e-03, 9.7752e-05, ..., 1.1883e-03, + 2.1553e-03, 7.4291e-04], + [ 2.1601e-04, 1.2856e-03, 2.3437e-04, ..., 3.5610e-03, + 1.8673e-03, 1.4286e-03], + ..., + [ 7.8506e-03, -1.1032e-02, 8.2922e-04, ..., -5.9891e-04, + -6.5765e-03, 5.2338e-03], + [ 4.5562e-04, -1.3103e-03, 7.1526e-05, ..., 1.7872e-03, + 2.3079e-03, 1.1082e-03], + [ 1.1625e-03, 1.9226e-03, 7.5817e-04, ..., 2.6011e-04, + -1.3870e-02, 6.8045e-04]], device='cuda:0') +Epoch 111, bias, value: tensor([-0.0412, -0.0022, -0.0018, -0.0058, -0.0024, 0.0061, -0.0104, 0.0272, + -0.0052, 0.0185], device='cuda:0'), grad: tensor([ 0.0247, 0.0108, 0.0205, -0.0151, 0.0246, -0.0435, 0.0115, -0.0225, + 0.0172, -0.0283], device='cuda:0') +100 +0.0001 +changing lr +epoch 110, time 214.18, cls_loss 0.6494 cls_loss_mapping 0.0127 cls_loss_causal 0.5511 re_mapping 0.0164 re_causal 0.0377 /// teacc 98.49 lr 0.00010000 +Epoch 112, weight, value: tensor([[-0.1181, -0.0293, 0.0061, ..., -0.0113, -0.0577, -0.0771], + [-0.0099, -0.0504, 0.0207, ..., 0.0259, 0.0005, -0.0540], + [-0.0299, -0.0489, -0.0627, ..., 0.0398, -0.0342, -0.0627], + ..., + [-0.0383, 0.0264, 0.0257, ..., 0.0274, -0.0322, -0.0589], + [-0.0545, -0.0050, 0.0135, ..., 0.0522, -0.0322, -0.0857], + [ 0.0330, 0.0392, -0.0298, ..., -0.0897, -0.0289, 0.0354]], + device='cuda:0'), grad: tensor([[ 2.6494e-05, -2.3401e-04, -4.1313e-03, ..., -5.0621e-03, + 1.9479e-04, -1.1269e-02], + [ 2.3752e-05, 1.0672e-03, 8.3447e-04, ..., 3.0956e-03, + 3.0303e-04, 1.9798e-03], + [ 1.3077e-04, -4.7112e-03, -4.8599e-03, ..., -5.8937e-03, + -1.1215e-03, -3.0937e-03], + ..., + [ 1.5914e-04, 2.5635e-03, 6.9618e-04, ..., 1.1185e-02, + 5.2834e-03, 1.9169e-03], + [ 8.8120e-04, 6.1083e-04, 8.8358e-04, ..., 2.5272e-03, + 5.1880e-04, 2.0103e-03], + [ 3.2215e-03, 1.7548e-03, 5.2299e-03, ..., 5.5313e-03, + 4.5633e-04, 8.9722e-03]], device='cuda:0') +Epoch 112, bias, value: tensor([-0.0414, -0.0019, -0.0012, -0.0061, -0.0024, 0.0070, -0.0105, 0.0268, + -0.0061, 0.0184], device='cuda:0'), grad: tensor([-0.0478, 0.0136, -0.0179, -0.0033, -0.0613, 0.0178, 0.0201, 0.0326, + 0.0139, 0.0322], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 111---------------------------------------------------- +epoch 111, time 214.98, cls_loss 0.6636 cls_loss_mapping 0.0136 cls_loss_causal 0.5784 re_mapping 0.0164 re_causal 0.0372 /// teacc 98.56 lr 0.00010000 +Epoch 113, weight, value: tensor([[-0.1187, -0.0293, 0.0057, ..., -0.0108, -0.0571, -0.0777], + [-0.0110, -0.0512, 0.0207, ..., 0.0262, 0.0008, -0.0537], + [-0.0307, -0.0486, -0.0629, ..., 0.0398, -0.0344, -0.0653], + ..., + [-0.0383, 0.0271, 0.0260, ..., 0.0283, -0.0313, -0.0582], + [-0.0549, -0.0048, 0.0131, ..., 0.0529, -0.0325, -0.0859], + [ 0.0340, 0.0395, -0.0298, ..., -0.0902, -0.0296, 0.0356]], + device='cuda:0'), grad: tensor([[ 1.8135e-05, 3.5691e-04, 2.5616e-03, ..., 2.5158e-03, + 9.7692e-05, 1.6117e-03], + [ 2.0161e-05, -2.8667e-03, 4.1819e-04, ..., -6.6223e-03, + -8.9169e-04, -1.1426e-04], + [ 8.5607e-06, 3.7694e-04, 1.1177e-03, ..., 1.6994e-03, + 2.7537e-04, 6.7282e-04], + ..., + [ 1.8805e-05, 4.5204e-03, 1.0977e-03, ..., 5.5885e-03, + 1.8635e-03, 5.4359e-03], + [ 9.0972e-06, 3.8600e-04, 4.8218e-03, ..., 3.5706e-03, + 2.1207e-04, 3.0842e-03], + [ 1.9267e-05, 6.4926e-03, 2.1648e-03, ..., 3.8910e-03, + 1.0529e-03, 7.1716e-03]], device='cuda:0') +Epoch 113, bias, value: tensor([-0.0410, -0.0016, -0.0015, -0.0064, -0.0027, 0.0072, -0.0114, 0.0280, + -0.0061, 0.0183], device='cuda:0'), grad: tensor([ 0.0116, -0.0227, 0.0085, -0.0227, -0.0145, -0.0241, 0.0075, 0.0196, + 0.0168, 0.0201], device='cuda:0') +100 +0.0001 +changing lr +epoch 112, time 213.93, cls_loss 0.6511 cls_loss_mapping 0.0135 cls_loss_causal 0.5616 re_mapping 0.0167 re_causal 0.0376 /// teacc 98.54 lr 0.00010000 +Epoch 114, weight, value: tensor([[-0.1187, -0.0297, 0.0061, ..., -0.0120, -0.0573, -0.0790], + [-0.0110, -0.0516, 0.0205, ..., 0.0265, 0.0004, -0.0539], + [-0.0318, -0.0495, -0.0645, ..., 0.0401, -0.0355, -0.0665], + ..., + [-0.0368, 0.0277, 0.0258, ..., 0.0284, -0.0308, -0.0583], + [-0.0548, -0.0055, 0.0139, ..., 0.0514, -0.0319, -0.0862], + [ 0.0335, 0.0409, -0.0291, ..., -0.0907, -0.0298, 0.0369]], + device='cuda:0'), grad: tensor([[ 5.1022e-04, 1.3084e-03, 4.3988e-04, ..., 3.0842e-03, + 4.2282e-07, 3.1543e-04], + [ 4.7803e-04, 2.2259e-03, -3.8266e-05, ..., 4.1389e-03, + 2.0694e-06, 2.7999e-05], + [ 1.3745e-04, 6.3896e-04, 1.4048e-03, ..., 6.7596e-03, + -8.2776e-06, 1.1368e-03], + ..., + [ 4.1533e-04, -9.3651e-04, 2.3425e-04, ..., -2.3155e-03, + 1.7677e-06, 1.3828e-04], + [ 1.7920e-03, 1.2083e-03, 1.4391e-03, ..., -1.7414e-03, + 6.0257e-07, 1.3361e-03], + [ 1.5383e-03, 3.9577e-04, 1.7567e-03, ..., -4.6635e-04, + 3.6322e-07, 1.1311e-03]], device='cuda:0') +Epoch 114, bias, value: tensor([-0.0415, -0.0020, -0.0012, -0.0059, -0.0029, 0.0084, -0.0114, 0.0284, + -0.0068, 0.0177], device='cuda:0'), grad: tensor([ 0.0210, 0.0281, 0.0381, -0.0332, 0.0200, -0.0662, 0.0205, -0.0370, + -0.0003, 0.0089], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 113---------------------------------------------------- +epoch 113, time 214.99, cls_loss 0.6862 cls_loss_mapping 0.0153 cls_loss_causal 0.5965 re_mapping 0.0163 re_causal 0.0352 /// teacc 98.66 lr 0.00010000 +Epoch 115, weight, value: tensor([[-0.1189, -0.0289, 0.0074, ..., -0.0110, -0.0572, -0.0790], + [-0.0111, -0.0505, 0.0203, ..., 0.0265, 0.0004, -0.0538], + [-0.0325, -0.0502, -0.0633, ..., 0.0391, -0.0354, -0.0677], + ..., + [-0.0367, 0.0274, 0.0260, ..., 0.0288, -0.0310, -0.0591], + [-0.0561, -0.0043, 0.0130, ..., 0.0517, -0.0327, -0.0861], + [ 0.0345, 0.0400, -0.0282, ..., -0.0906, -0.0284, 0.0370]], + device='cuda:0'), grad: tensor([[ 1.2255e-04, 4.3488e-04, 8.8215e-05, ..., 3.7899e-03, + 2.6722e-03, 7.0572e-05], + [ 6.8307e-05, 1.2064e-03, -6.9022e-05, ..., -6.8321e-03, + -6.9695e-03, 7.5996e-05], + [ 2.0370e-05, 2.5482e-03, 3.3498e-05, ..., 5.3368e-03, + 3.5992e-03, 4.3213e-05], + ..., + [-1.8539e-03, -4.9934e-03, 8.1897e-05, ..., -5.8899e-03, + -1.7633e-03, 2.3711e-04], + [ 4.3362e-05, 8.9836e-04, 1.6880e-04, ..., 5.2948e-03, + 4.5052e-03, 2.3854e-04], + [ 1.3618e-03, 1.9388e-03, 9.1672e-05, ..., -6.7673e-03, + -9.3994e-03, 9.2936e-04]], device='cuda:0') +Epoch 115, bias, value: tensor([-0.0403, -0.0014, -0.0015, -0.0061, -0.0031, 0.0082, -0.0118, 0.0286, + -0.0075, 0.0179], device='cuda:0'), grad: tensor([ 0.0200, -0.0356, 0.0261, 0.0284, 0.0157, -0.0305, -0.0175, -0.0328, + 0.0301, -0.0040], device='cuda:0') +100 +0.0001 +changing lr +epoch 114, time 214.12, cls_loss 0.6663 cls_loss_mapping 0.0139 cls_loss_causal 0.5826 re_mapping 0.0157 re_causal 0.0364 /// teacc 98.57 lr 0.00010000 +Epoch 116, weight, value: tensor([[-0.1202, -0.0287, 0.0069, ..., -0.0110, -0.0580, -0.0801], + [-0.0122, -0.0505, 0.0201, ..., 0.0276, 0.0001, -0.0553], + [-0.0334, -0.0506, -0.0629, ..., 0.0386, -0.0349, -0.0678], + ..., + [-0.0387, 0.0268, 0.0259, ..., 0.0285, -0.0309, -0.0601], + [-0.0554, -0.0042, 0.0135, ..., 0.0525, -0.0326, -0.0868], + [ 0.0338, 0.0412, -0.0286, ..., -0.0897, -0.0282, 0.0373]], + device='cuda:0'), grad: tensor([[-2.4343e-04, 1.7996e-03, 4.1795e-04, ..., 3.4676e-03, + 1.6174e-03, -6.2847e-04], + [ 2.7657e-04, 3.9220e-04, -4.4556e-03, ..., -1.3466e-02, + -2.5826e-03, -5.7650e-04], + [-1.5516e-03, -1.3069e-02, -2.7962e-03, ..., -5.8899e-03, + -8.1863e-03, 1.5318e-04], + ..., + [ 8.2445e-04, 6.3133e-03, 8.6164e-04, ..., 6.0883e-03, + 2.5978e-03, 9.2506e-05], + [ 1.0052e-03, 2.1954e-03, 2.1648e-03, ..., -2.5921e-03, + 3.5820e-03, 3.2568e-04], + [ 5.7983e-04, 1.4982e-03, 9.6130e-04, ..., 6.2561e-03, + 2.1553e-03, 3.8862e-05]], device='cuda:0') +Epoch 116, bias, value: tensor([-0.0403, -0.0012, -0.0012, -0.0065, -0.0036, 0.0085, -0.0127, 0.0281, + -0.0060, 0.0177], device='cuda:0'), grad: tensor([ 0.0161, -0.0773, -0.0486, 0.0070, 0.0132, -0.0269, 0.0543, 0.0365, + -0.0106, 0.0361], device='cuda:0') +100 +0.0001 +changing lr +epoch 115, time 213.96, cls_loss 0.6459 cls_loss_mapping 0.0132 cls_loss_causal 0.5583 re_mapping 0.0162 re_causal 0.0361 /// teacc 98.37 lr 0.00010000 +Epoch 117, weight, value: tensor([[-0.1203, -0.0278, 0.0064, ..., -0.0115, -0.0567, -0.0814], + [-0.0130, -0.0503, 0.0203, ..., 0.0273, 0.0002, -0.0565], + [-0.0332, -0.0503, -0.0635, ..., 0.0392, -0.0340, -0.0682], + ..., + [-0.0387, 0.0272, 0.0261, ..., 0.0280, -0.0313, -0.0598], + [-0.0555, -0.0038, 0.0127, ..., 0.0518, -0.0335, -0.0868], + [ 0.0339, 0.0408, -0.0281, ..., -0.0893, -0.0285, 0.0387]], + device='cuda:0'), grad: tensor([[ 5.1069e-04, 3.3932e-03, 8.9526e-05, ..., 5.1613e-03, + 1.8096e-04, 1.7757e-03], + [ 1.0061e-04, 1.6165e-04, 1.5289e-05, ..., 1.2226e-03, + 5.0211e-04, 1.5831e-04], + [-6.8893e-03, -5.1346e-03, -4.1318e-04, ..., 3.9711e-03, + -4.8494e-04, -8.4915e-03], + ..., + [-1.0788e-05, -5.7602e-04, 1.2651e-05, ..., 3.6621e-04, + 3.0375e-04, 1.4818e-04], + [ 2.7061e-04, 8.8310e-04, 5.2243e-05, ..., -1.5762e-02, + -7.0858e-04, 1.1482e-03], + [ 1.8609e-04, -1.1589e-02, 1.1370e-05, ..., 4.8332e-03, + 2.2388e-04, 3.9554e-04]], device='cuda:0') +Epoch 117, bias, value: tensor([-0.0406, -0.0015, -0.0010, -0.0062, -0.0029, 0.0074, -0.0118, 0.0281, + -0.0067, 0.0180], device='cuda:0'), grad: tensor([ 0.0346, 0.0045, -0.0288, 0.0206, -0.0451, -0.0073, 0.0474, -0.0017, + -0.0356, 0.0115], device='cuda:0') +100 +0.0001 +changing lr +epoch 116, time 214.18, cls_loss 0.6857 cls_loss_mapping 0.0138 cls_loss_causal 0.5953 re_mapping 0.0163 re_causal 0.0364 /// teacc 98.47 lr 0.00010000 +Epoch 118, weight, value: tensor([[-0.1206, -0.0276, 0.0072, ..., -0.0124, -0.0575, -0.0818], + [-0.0140, -0.0508, 0.0205, ..., 0.0264, 0.0017, -0.0572], + [-0.0331, -0.0505, -0.0641, ..., 0.0411, -0.0331, -0.0687], + ..., + [-0.0394, 0.0274, 0.0262, ..., 0.0276, -0.0316, -0.0606], + [-0.0555, -0.0039, 0.0132, ..., 0.0528, -0.0346, -0.0871], + [ 0.0345, 0.0403, -0.0294, ..., -0.0911, -0.0294, 0.0387]], + device='cuda:0'), grad: tensor([[ 4.1890e-04, 1.4591e-04, 4.2915e-04, ..., -1.6088e-03, + 1.0128e-03, 7.5531e-04], + [-6.9332e-04, -3.5906e-04, 2.0897e-04, ..., -1.5593e-03, + -4.3640e-03, 6.5744e-05], + [ 3.4237e-04, 2.8300e-04, 8.9312e-04, ..., 4.7684e-06, + 6.4611e-04, 8.1587e-04], + ..., + [ 1.4472e-04, -3.4738e-04, 2.1923e-04, ..., 1.4715e-03, + 1.1415e-03, 4.1544e-05], + [ 1.9436e-03, -1.8187e-03, -1.3280e-04, ..., 2.8725e-03, + 3.3932e-03, 2.5558e-03], + [-3.3875e-03, 2.8872e-04, -8.0109e-04, ..., -1.3514e-03, + 1.6785e-03, -7.7972e-03]], device='cuda:0') +Epoch 118, bias, value: tensor([-0.0407, -0.0012, -0.0006, -0.0054, -0.0030, 0.0071, -0.0112, 0.0273, + -0.0064, 0.0169], device='cuda:0'), grad: tensor([-0.0062, -0.0285, 0.0024, 0.0385, -0.0118, -0.0166, -0.0007, 0.0190, + 0.0475, -0.0437], device='cuda:0') +100 +0.0001 +changing lr +epoch 117, time 214.03, cls_loss 0.6307 cls_loss_mapping 0.0141 cls_loss_causal 0.5465 re_mapping 0.0160 re_causal 0.0357 /// teacc 98.52 lr 0.00010000 +Epoch 119, weight, value: tensor([[-0.1205, -0.0281, 0.0075, ..., -0.0136, -0.0579, -0.0822], + [-0.0157, -0.0515, 0.0191, ..., 0.0270, 0.0016, -0.0578], + [-0.0339, -0.0522, -0.0640, ..., 0.0408, -0.0340, -0.0678], + ..., + [-0.0392, 0.0274, 0.0274, ..., 0.0278, -0.0308, -0.0616], + [-0.0550, -0.0030, 0.0131, ..., 0.0528, -0.0353, -0.0874], + [ 0.0342, 0.0399, -0.0295, ..., -0.0902, -0.0294, 0.0391]], + device='cuda:0'), grad: tensor([[ 5.6791e-04, 2.1970e-04, 7.8857e-05, ..., -3.9330e-03, + 1.0929e-03, 3.9130e-05], + [ 4.1628e-04, 2.5225e-04, -1.9729e-04, ..., -3.5534e-03, + -1.5726e-03, 8.1301e-05], + [ 1.8053e-03, -3.4962e-03, 3.2115e-04, ..., 2.7428e-03, + 1.0843e-03, 6.2323e-04], + ..., + [-1.0460e-02, -4.6182e-04, -2.2297e-03, ..., -1.5907e-03, + 1.6699e-03, -2.7447e-03], + [ 2.0485e-03, 6.2275e-04, 4.1056e-04, ..., 3.4466e-03, + 1.1559e-03, 3.0422e-04], + [ 6.0501e-03, -5.8317e-04, 1.2465e-03, ..., 3.0255e-04, + 1.4353e-03, 5.0545e-04]], device='cuda:0') +Epoch 119, bias, value: tensor([-0.0415, -0.0005, -0.0010, -0.0050, -0.0034, 0.0069, -0.0112, 0.0279, + -0.0064, 0.0170], device='cuda:0'), grad: tensor([-0.0119, -0.0029, 0.0124, -0.0075, -0.0403, 0.0163, 0.0163, -0.0079, + 0.0233, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 118, time 214.22, cls_loss 0.6379 cls_loss_mapping 0.0140 cls_loss_causal 0.5493 re_mapping 0.0156 re_causal 0.0342 /// teacc 98.47 lr 0.00010000 +Epoch 120, weight, value: tensor([[-0.1205, -0.0261, 0.0085, ..., -0.0125, -0.0577, -0.0814], + [-0.0161, -0.0517, 0.0205, ..., 0.0266, 0.0022, -0.0583], + [-0.0322, -0.0544, -0.0647, ..., 0.0399, -0.0338, -0.0679], + ..., + [-0.0395, 0.0277, 0.0266, ..., 0.0275, -0.0312, -0.0622], + [-0.0551, -0.0036, 0.0121, ..., 0.0537, -0.0353, -0.0870], + [ 0.0342, 0.0398, -0.0291, ..., -0.0906, -0.0300, 0.0388]], + device='cuda:0'), grad: tensor([[ 4.8494e-04, 7.7295e-04, 3.5793e-05, ..., -4.1389e-03, + -7.9441e-04, 2.3603e-04], + [ 1.9944e-04, 2.2392e-03, 1.4353e-04, ..., 1.2226e-03, + 8.8215e-05, 1.4572e-03], + [ 2.5201e-04, -9.7961e-03, 9.7733e-03, ..., 6.0425e-03, + 2.6474e-03, -1.3557e-02], + ..., + [ 1.6558e-04, 5.7945e-03, 3.6793e-03, ..., 3.7384e-03, + 9.2649e-04, 1.0309e-03], + [ 8.7500e-04, 3.9864e-03, -1.4709e-02, ..., -5.9013e-03, + -3.6736e-03, 1.9484e-03], + [ 2.3782e-04, -1.9333e-02, 1.8203e-04, ..., -3.9902e-03, + 2.0361e-04, 8.8739e-04]], device='cuda:0') +Epoch 120, bias, value: tensor([-0.0400, -0.0002, -0.0011, -0.0051, -0.0036, 0.0070, -0.0116, 0.0275, + -0.0063, 0.0164], device='cuda:0'), grad: tensor([-0.0268, 0.0136, 0.0054, 0.0131, 0.0149, -0.0110, -0.0077, 0.0223, + -0.0078, -0.0160], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 119---------------------------------------------------- +epoch 119, time 214.75, cls_loss 0.6473 cls_loss_mapping 0.0151 cls_loss_causal 0.5520 re_mapping 0.0155 re_causal 0.0359 /// teacc 98.71 lr 0.00010000 +Epoch 121, weight, value: tensor([[-0.1226, -0.0267, 0.0069, ..., -0.0125, -0.0586, -0.0828], + [-0.0167, -0.0520, 0.0202, ..., 0.0276, 0.0023, -0.0586], + [-0.0331, -0.0533, -0.0662, ..., 0.0400, -0.0347, -0.0676], + ..., + [-0.0385, 0.0281, 0.0244, ..., 0.0274, -0.0311, -0.0614], + [-0.0549, -0.0036, 0.0125, ..., 0.0534, -0.0361, -0.0888], + [ 0.0351, 0.0386, -0.0290, ..., -0.0894, -0.0301, 0.0387]], + device='cuda:0'), grad: tensor([[-1.6651e-03, 9.8109e-05, -2.3766e-03, ..., 7.9536e-04, + 5.1641e-04, -1.5697e-03], + [ 3.3021e-05, 5.0402e-04, 1.2684e-04, ..., -9.3460e-04, + 5.6922e-05, 3.0613e-04], + [ 2.7847e-04, -1.0010e-02, -1.6518e-03, ..., -7.6628e-04, + 6.9976e-05, -1.0910e-03], + ..., + [ 5.5254e-05, 1.1196e-03, -2.2888e-03, ..., -8.6427e-05, + 1.5378e-04, -4.4556e-03], + [ 3.9940e-03, 8.4839e-03, 2.8133e-03, ..., 9.0790e-04, + 6.9571e-04, 5.8441e-03], + [ 2.0361e-04, -6.6280e-04, 4.4084e-04, ..., 5.6553e-04, + 3.8886e-04, 1.1444e-03]], device='cuda:0') +Epoch 121, bias, value: tensor([-4.0708e-02, 5.9647e-05, -9.7154e-04, -5.0502e-03, -3.9269e-03, + 7.0832e-03, -1.1632e-02, 2.7602e-02, -6.4881e-03, 1.6793e-02], + device='cuda:0'), grad: tensor([ 0.0087, -0.0122, -0.0356, 0.0041, -0.0484, 0.0329, -0.0069, -0.0143, + 0.0565, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 120, time 214.21, cls_loss 0.6590 cls_loss_mapping 0.0152 cls_loss_causal 0.5722 re_mapping 0.0159 re_causal 0.0350 /// teacc 98.48 lr 0.00010000 +Epoch 122, weight, value: tensor([[-0.1232, -0.0275, 0.0074, ..., -0.0134, -0.0590, -0.0838], + [-0.0169, -0.0529, 0.0197, ..., 0.0279, 0.0021, -0.0589], + [-0.0327, -0.0531, -0.0663, ..., 0.0413, -0.0344, -0.0683], + ..., + [-0.0389, 0.0276, 0.0244, ..., 0.0284, -0.0306, -0.0617], + [-0.0548, -0.0045, 0.0120, ..., 0.0525, -0.0373, -0.0897], + [ 0.0348, 0.0389, -0.0285, ..., -0.0895, -0.0304, 0.0390]], + device='cuda:0'), grad: tensor([[-2.7008e-03, 1.2856e-03, -4.5204e-03, ..., 6.0120e-03, + -3.3360e-03, 1.1311e-03], + [-1.0548e-03, 2.2793e-03, 1.1258e-05, ..., -2.5768e-03, + 2.4014e-03, -2.7180e-03], + [ 2.5678e-04, 1.4124e-03, 5.2184e-05, ..., -1.4229e-03, + 2.2621e-03, 6.3229e-04], + ..., + [ 5.0116e-04, 4.3058e-04, 1.0467e-04, ..., 4.8637e-03, + 3.8719e-04, 8.2064e-04], + [-7.9155e-05, 6.5708e-04, 1.5211e-04, ..., 6.2714e-03, + 9.2888e-04, 9.2745e-04], + [-5.2309e-04, 2.1541e-04, -6.0618e-05, ..., -2.3003e-03, + 2.2328e-04, -5.2452e-04]], device='cuda:0') +Epoch 122, bias, value: tensor([-4.1007e-02, -5.9427e-04, -2.7417e-05, -5.2969e-03, -3.6806e-03, + 6.8738e-03, -1.1815e-02, 2.8935e-02, -8.0402e-03, 1.7384e-02], + device='cuda:0'), grad: tensor([ 0.0131, 0.0020, 0.0045, -0.0176, -0.0773, 0.0165, 0.0216, 0.0212, + 0.0274, -0.0114], device='cuda:0') +100 +0.0001 +changing lr +epoch 121, time 214.00, cls_loss 0.6822 cls_loss_mapping 0.0131 cls_loss_causal 0.5914 re_mapping 0.0155 re_causal 0.0345 /// teacc 98.50 lr 0.00010000 +Epoch 123, weight, value: tensor([[-0.1216, -0.0283, 0.0074, ..., -0.0139, -0.0586, -0.0835], + [-0.0172, -0.0543, 0.0202, ..., 0.0280, 0.0010, -0.0598], + [-0.0328, -0.0547, -0.0676, ..., 0.0421, -0.0343, -0.0670], + ..., + [-0.0401, 0.0273, 0.0258, ..., 0.0282, -0.0307, -0.0626], + [-0.0534, -0.0044, 0.0109, ..., 0.0531, -0.0358, -0.0891], + [ 0.0358, 0.0401, -0.0284, ..., -0.0901, -0.0303, 0.0380]], + device='cuda:0'), grad: tensor([[ 7.5150e-04, 2.4271e-04, 1.5659e-03, ..., -3.5534e-03, + 7.0524e-04, 1.5831e-04], + [ 4.4107e-04, 3.5000e-04, 1.3189e-03, ..., 1.8759e-03, + 1.0834e-03, 1.3041e-04], + [ 4.4632e-04, -1.1568e-03, 1.1702e-03, ..., -1.1765e-02, + -9.6817e-03, 2.4676e-04], + ..., + [ 2.6035e-04, -2.0266e-05, 9.6607e-04, ..., 1.8845e-03, + 1.0357e-03, 3.3665e-04], + [ 2.4819e-04, 9.3126e-04, 1.7853e-03, ..., 1.3685e-03, + 6.9046e-04, 9.1362e-04], + [-1.0509e-03, 9.1076e-04, 3.1738e-03, ..., 2.2449e-03, + 5.4979e-04, -9.6369e-04]], device='cuda:0') +Epoch 123, bias, value: tensor([-0.0410, -0.0006, 0.0006, -0.0051, -0.0033, 0.0059, -0.0122, 0.0286, + -0.0078, 0.0174], device='cuda:0'), grad: tensor([-0.0161, 0.0140, -0.0462, -0.0048, 0.0112, 0.0103, -0.0136, 0.0153, + 0.0132, 0.0166], device='cuda:0') +100 +0.0001 +changing lr +epoch 122, time 214.13, cls_loss 0.6422 cls_loss_mapping 0.0135 cls_loss_causal 0.5479 re_mapping 0.0158 re_causal 0.0349 /// teacc 98.40 lr 0.00010000 +Epoch 124, weight, value: tensor([[-0.1223, -0.0278, 0.0082, ..., -0.0130, -0.0585, -0.0823], + [-0.0181, -0.0537, 0.0202, ..., 0.0280, 0.0002, -0.0609], + [-0.0324, -0.0543, -0.0684, ..., 0.0414, -0.0347, -0.0666], + ..., + [-0.0387, 0.0274, 0.0267, ..., 0.0282, -0.0299, -0.0623], + [-0.0543, -0.0046, 0.0112, ..., 0.0539, -0.0360, -0.0902], + [ 0.0357, 0.0393, -0.0291, ..., -0.0903, -0.0293, 0.0384]], + device='cuda:0'), grad: tensor([[ 3.4752e-03, 2.4211e-04, 2.5177e-03, ..., -3.1757e-04, + 7.9632e-04, 1.2293e-03], + [ 1.8224e-05, -2.7828e-03, 1.2369e-03, ..., -6.5536e-03, + 1.0509e-03, 1.1545e-04], + [ 2.7251e-04, 9.9564e-04, 1.1139e-03, ..., 1.7605e-03, + 9.8801e-04, 2.9039e-04], + ..., + [ 5.9843e-04, -6.3171e-02, 1.1620e-02, ..., 8.3694e-03, + 4.4560e-04, 7.3671e-04], + [ 7.2718e-04, -3.1357e-03, -1.0246e-02, ..., -6.7177e-03, + 9.7692e-05, 9.7227e-04], + [-3.3021e-04, 6.4026e-02, 1.4057e-03, ..., 1.5602e-03, + 2.4915e-04, 1.5879e-04]], device='cuda:0') +Epoch 124, bias, value: tensor([-4.0826e-02, -6.9545e-05, 4.9023e-04, -5.6946e-03, -3.3404e-03, + 5.7964e-03, -1.2686e-02, 2.8223e-02, -6.9016e-03, 1.7681e-02], + device='cuda:0'), grad: tensor([-0.0051, -0.0048, 0.0347, 0.0444, 0.0113, -0.0956, -0.0025, 0.0186, + -0.0146, 0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 123, time 214.06, cls_loss 0.6517 cls_loss_mapping 0.0148 cls_loss_causal 0.5498 re_mapping 0.0148 re_causal 0.0327 /// teacc 98.32 lr 0.00010000 +Epoch 125, weight, value: tensor([[-1.2389e-01, -2.8001e-02, 8.6844e-03, ..., -1.3437e-02, + -5.8099e-02, -8.1284e-02], + [-1.7297e-02, -5.3578e-02, 1.9731e-02, ..., 2.7561e-02, + 3.6909e-05, -6.1791e-02], + [-3.3333e-02, -5.3676e-02, -6.9748e-02, ..., 4.2636e-02, + -3.4836e-02, -6.6161e-02], + ..., + [-3.7980e-02, 2.7927e-02, 2.7079e-02, ..., 2.6971e-02, + -3.1327e-02, -6.1369e-02], + [-5.4691e-02, -5.2084e-03, 1.1311e-02, ..., 5.5171e-02, + -3.6421e-02, -9.1268e-02], + [ 3.4489e-02, 3.9893e-02, -3.0131e-02, ..., -9.0470e-02, + -2.9973e-02, 3.7858e-02]], device='cuda:0'), grad: tensor([[ 7.5340e-03, 5.1079e-03, 9.5825e-03, ..., 9.4080e-04, + 3.6564e-03, 1.2369e-03], + [ 2.3353e-04, 1.9817e-03, 2.7313e-03, ..., -1.7023e-03, + -3.9177e-03, 9.3699e-04], + [ 7.1716e-04, 5.8937e-03, 8.2970e-04, ..., 1.0490e-03, + 2.2774e-03, 1.3237e-03], + ..., + [ 3.1929e-03, 3.4241e-02, -7.4272e-03, ..., 6.7711e-05, + -1.0605e-03, 1.0586e-03], + [ 5.5838e-04, -1.0017e-02, 1.0338e-03, ..., 8.5926e-04, + -6.9504e-03, 6.4468e-04], + [-7.8583e-03, -2.7313e-02, -8.8806e-03, ..., 2.0275e-03, + 1.4620e-03, 2.5578e-03]], device='cuda:0') +Epoch 125, bias, value: tensor([-0.0412, 0.0002, 0.0008, -0.0049, -0.0032, 0.0060, -0.0129, 0.0273, + -0.0072, 0.0177], device='cuda:0'), grad: tensor([ 0.0405, -0.0256, 0.0079, -0.0264, -0.0249, 0.0153, 0.0021, 0.0227, + 0.0051, -0.0167], device='cuda:0') +100 +0.0001 +changing lr +epoch 124, time 214.23, cls_loss 0.6420 cls_loss_mapping 0.0110 cls_loss_causal 0.5517 re_mapping 0.0158 re_causal 0.0349 /// teacc 98.24 lr 0.00010000 +Epoch 126, weight, value: tensor([[-0.1264, -0.0286, 0.0086, ..., -0.0139, -0.0570, -0.0816], + [-0.0181, -0.0513, 0.0193, ..., 0.0264, 0.0001, -0.0618], + [-0.0329, -0.0531, -0.0705, ..., 0.0438, -0.0351, -0.0664], + ..., + [-0.0407, 0.0273, 0.0262, ..., 0.0268, -0.0318, -0.0616], + [-0.0541, -0.0053, 0.0129, ..., 0.0555, -0.0365, -0.0906], + [ 0.0356, 0.0398, -0.0298, ..., -0.0896, -0.0315, 0.0380]], + device='cuda:0'), grad: tensor([[ 1.3840e-04, 8.9169e-04, -9.8646e-05, ..., 9.3126e-04, + 3.2037e-05, 1.0653e-03], + [-4.2648e-03, 1.1244e-03, -5.0507e-03, ..., -2.4757e-03, + 1.8150e-05, -3.8948e-03], + [ 2.4395e-03, 5.3749e-03, 1.8911e-03, ..., 1.7176e-03, + 1.6146e-03, 1.9398e-03], + ..., + [ 1.7443e-03, 2.8534e-03, 2.0142e-03, ..., -5.4693e-04, + -2.4509e-04, 2.4319e-03], + [ 6.3944e-04, 7.8430e-03, 2.6846e-04, ..., 1.8167e-03, + 1.2243e-04, 1.1158e-03], + [ 1.9293e-03, 1.0216e-02, 1.5440e-03, ..., 3.6011e-03, + 7.2956e-04, 2.4052e-03]], device='cuda:0') +Epoch 126, bias, value: tensor([-0.0417, 0.0003, 0.0013, -0.0050, -0.0035, 0.0068, -0.0134, 0.0271, + -0.0067, 0.0175], device='cuda:0'), grad: tensor([ 0.0081, -0.0060, 0.0204, -0.0558, -0.0321, 0.0023, 0.0077, 0.0035, + 0.0161, 0.0358], device='cuda:0') +100 +0.0001 +changing lr +epoch 125, time 214.15, cls_loss 0.6412 cls_loss_mapping 0.0147 cls_loss_causal 0.5561 re_mapping 0.0164 re_causal 0.0352 /// teacc 98.48 lr 0.00010000 +Epoch 127, weight, value: tensor([[-0.1273, -0.0289, 0.0084, ..., -0.0126, -0.0577, -0.0812], + [-0.0168, -0.0514, 0.0186, ..., 0.0259, -0.0014, -0.0607], + [-0.0315, -0.0537, -0.0711, ..., 0.0425, -0.0347, -0.0679], + ..., + [-0.0412, 0.0275, 0.0261, ..., 0.0274, -0.0332, -0.0622], + [-0.0540, -0.0058, 0.0124, ..., 0.0564, -0.0363, -0.0914], + [ 0.0358, 0.0398, -0.0290, ..., -0.0886, -0.0291, 0.0383]], + device='cuda:0'), grad: tensor([[ 3.5977e-04, 1.0505e-05, 2.5482e-03, ..., 4.2000e-03, + 3.8605e-03, 5.5838e-04], + [ 2.1422e-04, -8.2552e-05, 1.3056e-03, ..., 3.5114e-03, + 1.7118e-03, 3.0589e-04], + [ 9.7513e-05, 4.9211e-06, -1.1200e-02, ..., 2.6817e-03, + -2.8305e-02, 1.6332e-04], + ..., + [-1.0550e-05, 2.0340e-06, -1.7452e-03, ..., -8.3084e-03, + 2.1267e-03, -5.0402e-04], + [ 7.9870e-04, 4.0919e-05, -4.9171e-03, ..., -7.2823e-03, + -5.1994e-03, 1.1091e-03], + [-1.7852e-05, 1.0885e-05, 1.7653e-03, ..., 3.5019e-03, + 2.4776e-03, 1.2004e-04]], device='cuda:0') +Epoch 127, bias, value: tensor([-0.0408, -0.0003, 0.0003, -0.0052, -0.0033, 0.0066, -0.0139, 0.0283, + -0.0071, 0.0181], device='cuda:0'), grad: tensor([ 0.0191, 0.0118, -0.0179, 0.0109, -0.0192, 0.0178, 0.0168, -0.0343, + -0.0202, 0.0151], device='cuda:0') +100 +0.0001 +changing lr +epoch 126, time 214.78, cls_loss 0.6349 cls_loss_mapping 0.0113 cls_loss_causal 0.5562 re_mapping 0.0160 re_causal 0.0357 /// teacc 98.61 lr 0.00010000 +Epoch 128, weight, value: tensor([[-0.1267, -0.0291, 0.0084, ..., -0.0144, -0.0576, -0.0818], + [-0.0178, -0.0520, 0.0187, ..., 0.0270, -0.0013, -0.0610], + [-0.0304, -0.0536, -0.0706, ..., 0.0410, -0.0351, -0.0684], + ..., + [-0.0418, 0.0265, 0.0267, ..., 0.0267, -0.0337, -0.0627], + [-0.0546, -0.0062, 0.0129, ..., 0.0570, -0.0369, -0.0919], + [ 0.0376, 0.0403, -0.0299, ..., -0.0888, -0.0290, 0.0396]], + device='cuda:0'), grad: tensor([[ 1.0471e-03, -1.8616e-03, 1.1897e-04, ..., 2.7599e-03, + 2.3496e-04, 2.1279e-04], + [-3.3140e-05, 1.5453e-05, -7.9393e-05, ..., 1.1215e-03, + 2.6137e-05, 1.7583e-04], + [ 3.0956e-03, 4.3249e-04, 7.6389e-04, ..., 3.4866e-03, + 3.5238e-04, 1.9817e-03], + ..., + [ 2.2182e-03, -2.5854e-05, 1.4038e-03, ..., 4.9706e-03, + 3.1738e-03, 8.8501e-04], + [-7.4387e-04, 3.1424e-04, -4.6039e-04, ..., 1.7920e-03, + 1.3208e-04, 4.4703e-04], + [-1.0475e-02, -9.9754e-04, -2.8372e-04, ..., -5.4665e-03, + -1.6922e-02, -1.5430e-03]], device='cuda:0') +Epoch 128, bias, value: tensor([-4.1836e-02, 7.0876e-05, -8.6509e-06, -4.6336e-03, -2.7277e-03, + 5.9588e-03, -1.3718e-02, 2.8319e-02, -7.2488e-03, 1.8364e-02], + device='cuda:0'), grad: tensor([ 0.0231, 0.0040, 0.0083, -0.0025, 0.0264, 0.0090, -0.0790, 0.0149, + 0.0070, -0.0112], device='cuda:0') +100 +0.0001 +changing lr +epoch 127, time 218.18, cls_loss 0.6487 cls_loss_mapping 0.0123 cls_loss_causal 0.5666 re_mapping 0.0153 re_causal 0.0348 /// teacc 98.62 lr 0.00010000 +Epoch 129, weight, value: tensor([[-0.1275, -0.0292, 0.0091, ..., -0.0145, -0.0562, -0.0824], + [-0.0186, -0.0525, 0.0194, ..., 0.0269, -0.0017, -0.0612], + [-0.0311, -0.0547, -0.0700, ..., 0.0410, -0.0355, -0.0685], + ..., + [-0.0419, 0.0275, 0.0275, ..., 0.0272, -0.0327, -0.0625], + [-0.0541, -0.0061, 0.0154, ..., 0.0575, -0.0372, -0.0916], + [ 0.0375, 0.0399, -0.0318, ..., -0.0890, -0.0289, 0.0399]], + device='cuda:0'), grad: tensor([[ 1.2436e-03, 9.7871e-05, 3.9935e-04, ..., 3.8052e-03, + 1.2410e-04, 2.8968e-04], + [-3.5629e-03, 1.5593e-04, -9.1696e-04, ..., 2.9564e-03, + 3.5048e-04, -5.1384e-03], + [-1.0081e-05, -2.7733e-03, 3.8671e-04, ..., -1.3351e-03, + 3.3784e-04, -3.2730e-03], + ..., + [-4.1656e-03, 8.7261e-04, -1.4938e-02, ..., -1.7609e-02, + -2.8267e-03, 8.0872e-04], + [-2.9099e-02, -9.3937e-04, 2.4605e-03, ..., -9.6207e-03, + -1.8206e-03, -1.4877e-03], + [ 5.6877e-03, 7.0333e-04, 4.9820e-03, ..., 8.5983e-03, + 8.7547e-04, 4.5204e-03]], device='cuda:0') +Epoch 129, bias, value: tensor([-0.0403, 0.0002, 0.0003, -0.0050, -0.0039, 0.0058, -0.0133, 0.0288, + -0.0077, 0.0177], device='cuda:0'), grad: tensor([ 0.0201, 0.0046, -0.0037, -0.0010, 0.0548, -0.0488, 0.0453, -0.0505, + -0.0574, 0.0366], device='cuda:0') +100 +0.0001 +changing lr +epoch 128, time 216.18, cls_loss 0.6196 cls_loss_mapping 0.0127 cls_loss_causal 0.5326 re_mapping 0.0147 re_causal 0.0335 /// teacc 98.48 lr 0.00010000 +Epoch 130, weight, value: tensor([[-0.1271, -0.0290, 0.0087, ..., -0.0150, -0.0558, -0.0816], + [-0.0194, -0.0527, 0.0209, ..., 0.0269, -0.0021, -0.0613], + [-0.0317, -0.0540, -0.0699, ..., 0.0423, -0.0356, -0.0676], + ..., + [-0.0433, 0.0280, 0.0270, ..., 0.0282, -0.0318, -0.0641], + [-0.0546, -0.0061, 0.0143, ..., 0.0564, -0.0383, -0.0916], + [ 0.0379, 0.0397, -0.0326, ..., -0.0888, -0.0279, 0.0395]], + device='cuda:0'), grad: tensor([[ 9.1970e-05, -1.0467e-04, 1.1110e-03, ..., 4.2992e-03, + 1.2445e-03, 6.5744e-05], + [ 1.3256e-03, 6.5565e-06, 2.8534e-03, ..., 7.9422e-03, + 2.5845e-03, 8.8406e-04], + [ 1.1462e-04, -5.2185e-03, 9.8419e-04, ..., -1.1253e-02, + -4.4136e-03, -5.9223e-04], + ..., + [-3.2616e-03, 9.0551e-04, -7.3700e-03, ..., -1.4694e-02, + -8.4229e-03, -1.3599e-03], + [ 9.8038e-04, 1.0586e-03, -6.6614e-04, ..., -1.7023e-03, + 2.3327e-03, 7.5006e-04], + [ 1.7929e-03, -2.8000e-03, 3.7670e-03, ..., 4.4441e-03, + 1.5278e-03, 1.5049e-03]], device='cuda:0') +Epoch 130, bias, value: tensor([-0.0405, 0.0012, 0.0008, -0.0054, -0.0041, 0.0049, -0.0134, 0.0297, + -0.0080, 0.0175], device='cuda:0'), grad: tensor([ 0.0277, 0.0352, -0.0376, 0.0352, 0.0173, 0.0174, -0.0596, -0.0475, + -0.0081, 0.0201], device='cuda:0') +100 +0.0001 +changing lr +epoch 129, time 216.86, cls_loss 0.6438 cls_loss_mapping 0.0133 cls_loss_causal 0.5665 re_mapping 0.0155 re_causal 0.0358 /// teacc 98.40 lr 0.00010000 +Epoch 131, weight, value: tensor([[-0.1284, -0.0269, 0.0092, ..., -0.0162, -0.0556, -0.0809], + [-0.0180, -0.0540, 0.0202, ..., 0.0259, -0.0035, -0.0605], + [-0.0316, -0.0538, -0.0703, ..., 0.0413, -0.0364, -0.0672], + ..., + [-0.0428, 0.0282, 0.0267, ..., 0.0283, -0.0308, -0.0658], + [-0.0549, -0.0056, 0.0142, ..., 0.0580, -0.0376, -0.0919], + [ 0.0369, 0.0388, -0.0332, ..., -0.0890, -0.0274, 0.0405]], + device='cuda:0'), grad: tensor([[-2.1911e-04, -7.4921e-03, 4.1008e-03, ..., -4.4899e-03, + -3.2425e-03, 4.0913e-04], + [ 2.1249e-05, 4.0016e-03, 2.7627e-05, ..., 7.4387e-03, + 2.9144e-03, 6.1131e-04], + [ 3.9172e-04, 2.7866e-03, 6.1131e-04, ..., 5.3596e-03, + 1.8940e-03, 7.6437e-04], + ..., + [-7.7200e-04, -6.0577e-03, -1.2646e-03, ..., -7.3166e-03, + -7.7724e-05, -1.5712e-04], + [ 4.6158e-04, 2.6169e-03, 4.6134e-04, ..., 6.6681e-03, + 2.5024e-03, 1.4219e-03], + [-1.2102e-03, 2.7580e-03, 2.8229e-04, ..., -3.2616e-04, + 2.5826e-03, 7.0763e-04]], device='cuda:0') +Epoch 131, bias, value: tensor([-0.0402, -0.0002, 0.0002, -0.0061, -0.0035, 0.0063, -0.0137, 0.0292, + -0.0072, 0.0180], device='cuda:0'), grad: tensor([-0.0394, 0.0298, 0.0245, 0.0111, -0.0403, -0.0114, 0.0238, -0.0262, + 0.0320, -0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 130, time 215.02, cls_loss 0.6367 cls_loss_mapping 0.0124 cls_loss_causal 0.5427 re_mapping 0.0152 re_causal 0.0333 /// teacc 98.52 lr 0.00010000 +Epoch 132, weight, value: tensor([[-0.1309, -0.0287, 0.0097, ..., -0.0162, -0.0549, -0.0815], + [-0.0184, -0.0545, 0.0203, ..., 0.0264, -0.0035, -0.0606], + [-0.0321, -0.0546, -0.0706, ..., 0.0406, -0.0367, -0.0683], + ..., + [-0.0429, 0.0284, 0.0265, ..., 0.0282, -0.0306, -0.0664], + [-0.0543, -0.0054, 0.0148, ..., 0.0589, -0.0371, -0.0914], + [ 0.0371, 0.0384, -0.0338, ..., -0.0911, -0.0284, 0.0405]], + device='cuda:0'), grad: tensor([[-3.5197e-05, 4.0412e-05, 1.1927e-04, ..., -1.3418e-03, + 9.4461e-04, 1.6701e-04], + [ 1.0459e-06, 6.6869e-06, 5.6648e-04, ..., 2.0889e-02, + 5.5923e-03, 1.0478e-04], + [ 1.9707e-06, 1.0127e-04, -5.0974e-04, ..., -6.5002e-03, + -7.3090e-03, 4.1819e-04], + ..., + [ 1.5676e-05, 9.1136e-05, -5.3787e-04, ..., -8.2474e-03, + 3.9330e-03, -3.1137e-04], + [ 3.2991e-05, 5.1796e-05, 3.8218e-04, ..., -1.6602e-02, + -1.1604e-02, 3.1829e-04], + [-6.7830e-05, -1.3065e-04, -9.3174e-04, ..., -2.2297e-03, + -1.6766e-03, 8.1956e-05]], device='cuda:0') +Epoch 132, bias, value: tensor([-4.1679e-02, -2.1365e-05, 2.3002e-04, -6.5972e-03, -3.2754e-03, + 5.6250e-03, -1.1911e-02, 2.9683e-02, -5.9936e-03, 1.6487e-02], + device='cuda:0'), grad: tensor([-0.0020, 0.0539, -0.0091, -0.0058, -0.0087, 0.0167, 0.0015, -0.0314, + 0.0011, -0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 131, time 215.53, cls_loss 0.6404 cls_loss_mapping 0.0141 cls_loss_causal 0.5506 re_mapping 0.0149 re_causal 0.0333 /// teacc 98.58 lr 0.00010000 +Epoch 133, weight, value: tensor([[-0.1309, -0.0275, 0.0098, ..., -0.0167, -0.0558, -0.0804], + [-0.0191, -0.0540, 0.0199, ..., 0.0274, -0.0030, -0.0607], + [-0.0322, -0.0545, -0.0711, ..., 0.0390, -0.0374, -0.0680], + ..., + [-0.0425, 0.0282, 0.0258, ..., 0.0285, -0.0305, -0.0673], + [-0.0544, -0.0049, 0.0149, ..., 0.0583, -0.0372, -0.0918], + [ 0.0367, 0.0389, -0.0339, ..., -0.0907, -0.0280, 0.0412]], + device='cuda:0'), grad: tensor([[ 8.7738e-05, -2.3937e-03, -8.1177e-03, ..., 2.2774e-03, + -1.3748e-02, -3.8242e-03], + [ 6.4325e-04, 1.0500e-03, 5.4741e-04, ..., 6.9580e-03, + 2.9602e-03, 4.0269e-04], + [ 1.1539e-04, 9.6142e-05, 7.7391e-04, ..., 3.9043e-03, + 1.5259e-03, 4.2152e-04], + ..., + [-1.7441e-02, -1.1343e-04, -1.6184e-03, ..., 6.5994e-04, + -2.8362e-03, -1.0620e-02], + [-5.2147e-03, 6.0177e-04, -1.2350e-03, ..., -9.6741e-03, + 2.1019e-03, -2.6855e-03], + [ 1.6205e-02, 1.7347e-03, 2.9469e-03, ..., -6.0768e-03, + 5.7220e-03, 9.6741e-03]], device='cuda:0') +Epoch 133, bias, value: tensor([-0.0424, 0.0007, 0.0005, -0.0061, -0.0033, 0.0056, -0.0115, 0.0287, + -0.0071, 0.0174], device='cuda:0'), grad: tensor([-0.0356, 0.0328, 0.0199, -0.0215, 0.0339, 0.0436, -0.0228, -0.0097, + -0.0319, -0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 132, time 215.21, cls_loss 0.6433 cls_loss_mapping 0.0116 cls_loss_causal 0.5569 re_mapping 0.0149 re_causal 0.0341 /// teacc 98.43 lr 0.00010000 +Epoch 134, weight, value: tensor([[-0.1294, -0.0281, 0.0102, ..., -0.0168, -0.0552, -0.0816], + [-0.0199, -0.0537, 0.0215, ..., 0.0273, -0.0029, -0.0611], + [-0.0325, -0.0535, -0.0723, ..., 0.0396, -0.0371, -0.0679], + ..., + [-0.0431, 0.0286, 0.0265, ..., 0.0291, -0.0302, -0.0681], + [-0.0541, -0.0049, 0.0128, ..., 0.0580, -0.0371, -0.0920], + [ 0.0373, 0.0387, -0.0339, ..., -0.0914, -0.0287, 0.0413]], + device='cuda:0'), grad: tensor([[ 2.7752e-03, 2.3973e-04, 5.1737e-04, ..., 3.1471e-03, + 3.6011e-03, 2.8133e-03], + [ 9.5740e-06, 5.6496e-03, 9.1851e-05, ..., -7.7324e-03, + 6.9332e-04, 1.0930e-05], + [ 9.8038e-04, 4.8351e-04, 2.5201e-04, ..., 2.3384e-03, + 1.0653e-03, 9.8133e-04], + ..., + [ 8.4996e-05, 8.3923e-04, 7.9298e-04, ..., 5.0545e-03, + 3.2501e-03, 9.0837e-05], + [ 1.8585e-04, 7.8297e-04, 4.6277e-04, ..., -4.0016e-03, + -1.0666e-02, 4.6968e-04], + [ 2.1017e-04, -1.3538e-01, 2.5034e-04, ..., -2.5681e-02, + 3.6907e-03, 2.4772e-04]], device='cuda:0') +Epoch 134, bias, value: tensor([-0.0420, 0.0005, 0.0008, -0.0062, -0.0045, 0.0052, -0.0110, 0.0295, + -0.0073, 0.0174], device='cuda:0'), grad: tensor([ 0.0250, -0.0440, 0.0175, 0.0044, 0.0592, -0.0474, 0.0189, 0.0256, + 0.0034, -0.0626], device='cuda:0') +100 +0.0001 +changing lr +epoch 133, time 215.27, cls_loss 0.6479 cls_loss_mapping 0.0151 cls_loss_causal 0.5661 re_mapping 0.0147 re_causal 0.0320 /// teacc 98.54 lr 0.00010000 +Epoch 135, weight, value: tensor([[-0.1292, -0.0291, 0.0105, ..., -0.0170, -0.0557, -0.0816], + [-0.0202, -0.0538, 0.0203, ..., 0.0267, -0.0032, -0.0628], + [-0.0335, -0.0534, -0.0714, ..., 0.0405, -0.0375, -0.0683], + ..., + [-0.0427, 0.0278, 0.0259, ..., 0.0290, -0.0306, -0.0679], + [-0.0540, -0.0061, 0.0137, ..., 0.0591, -0.0356, -0.0909], + [ 0.0371, 0.0414, -0.0336, ..., -0.0906, -0.0285, 0.0403]], + device='cuda:0'), grad: tensor([[-2.8563e-04, 3.9554e-04, 1.0353e-04, ..., 4.1809e-03, + 3.3116e-04, 1.2267e-04], + [ 3.5185e-06, 3.1328e-04, 2.4748e-04, ..., 4.8866e-03, + 5.8603e-04, 4.4703e-06], + [ 8.4460e-05, 7.1192e-04, 2.9182e-04, ..., 6.5651e-03, + 1.3151e-03, 2.2817e-04], + ..., + [ 3.8475e-05, 4.5013e-04, -5.6000e-03, ..., -1.5078e-03, + 1.6298e-03, 3.3170e-05], + [ 3.0637e-04, -2.8019e-03, 2.7847e-04, ..., -1.6966e-03, + -1.0166e-03, -2.9945e-03], + [-3.2425e-05, 5.7554e-04, 9.9182e-04, ..., -7.5150e-04, + 9.4175e-04, 2.1338e-04]], device='cuda:0') +Epoch 135, bias, value: tensor([-4.2428e-02, 9.1617e-05, 1.0651e-03, -6.0409e-03, -5.5962e-03, + 4.9534e-03, -1.0143e-02, 2.9424e-02, -7.0573e-03, 1.8071e-02], + device='cuda:0'), grad: tensor([ 0.0190, 0.0245, 0.0275, 0.0044, -0.0591, 0.0181, 0.0101, -0.0190, + -0.0256, 0.0001], device='cuda:0') +100 +0.0001 +changing lr +epoch 134, time 215.15, cls_loss 0.6253 cls_loss_mapping 0.0102 cls_loss_causal 0.5371 re_mapping 0.0150 re_causal 0.0338 /// teacc 98.45 lr 0.00010000 +Epoch 136, weight, value: tensor([[-0.1305, -0.0291, 0.0095, ..., -0.0169, -0.0560, -0.0822], + [-0.0217, -0.0552, 0.0192, ..., 0.0265, -0.0033, -0.0637], + [-0.0325, -0.0522, -0.0709, ..., 0.0415, -0.0375, -0.0684], + ..., + [-0.0427, 0.0269, 0.0258, ..., 0.0288, -0.0318, -0.0680], + [-0.0545, -0.0051, 0.0143, ..., 0.0597, -0.0346, -0.0908], + [ 0.0364, 0.0413, -0.0344, ..., -0.0911, -0.0275, 0.0398]], + device='cuda:0'), grad: tensor([[ 4.2486e-04, 8.8024e-04, 8.3876e-04, ..., 2.4853e-03, + -4.3678e-04, 1.6785e-04], + [ 1.4842e-04, 9.0599e-04, 4.6492e-04, ..., -4.4289e-03, + 2.0657e-06, 3.4541e-05], + [ 3.6793e-03, 4.6873e-04, 3.4485e-03, ..., 6.7759e-04, + 1.5870e-05, 1.8346e-04], + ..., + [ 3.3398e-03, 8.8263e-04, 4.4975e-03, ..., 8.2932e-03, + 7.5772e-06, 1.7958e-03], + [ 4.0474e-03, 1.6129e-02, 1.1559e-03, ..., 3.0708e-03, + 1.2094e-04, 6.0654e-03], + [-1.4763e-02, 1.6556e-02, -1.0620e-02, ..., -4.2267e-03, + -1.8585e-04, -1.2505e-02]], device='cuda:0') +Epoch 136, bias, value: tensor([-0.0412, -0.0005, 0.0018, -0.0075, -0.0050, 0.0050, -0.0111, 0.0293, + -0.0064, 0.0180], device='cuda:0'), grad: tensor([ 0.0113, -0.0232, 0.0277, -0.0184, -0.0151, -0.0440, 0.0147, 0.0340, + 0.0365, -0.0234], device='cuda:0') +100 +0.0001 +changing lr +epoch 135, time 215.14, cls_loss 0.6660 cls_loss_mapping 0.0089 cls_loss_causal 0.5777 re_mapping 0.0144 re_causal 0.0341 /// teacc 98.49 lr 0.00010000 +Epoch 137, weight, value: tensor([[-0.1303, -0.0293, 0.0093, ..., -0.0174, -0.0556, -0.0833], + [-0.0216, -0.0538, 0.0201, ..., 0.0289, -0.0024, -0.0638], + [-0.0332, -0.0527, -0.0715, ..., 0.0398, -0.0376, -0.0669], + ..., + [-0.0424, 0.0271, 0.0255, ..., 0.0295, -0.0316, -0.0680], + [-0.0551, -0.0051, 0.0159, ..., 0.0586, -0.0348, -0.0916], + [ 0.0361, 0.0405, -0.0358, ..., -0.0912, -0.0279, 0.0398]], + device='cuda:0'), grad: tensor([[ 4.5538e-04, 1.1027e-04, 6.7806e-04, ..., -5.6801e-03, + 1.3914e-03, -3.6373e-03], + [ 5.7101e-05, 4.2224e-04, -1.5198e-02, ..., -1.8051e-02, + -9.3269e-04, -6.1989e-04], + [ 2.8419e-04, 2.1696e-04, 3.1891e-03, ..., 6.4926e-03, + 1.2636e-03, 3.9959e-04], + ..., + [ 1.4842e-05, -1.0443e-03, 1.4591e-03, ..., 5.4054e-03, + 6.9523e-04, 3.2496e-04], + [ 1.8668e-04, 4.5371e-04, 4.6234e-03, ..., 8.3466e-03, + 3.3321e-03, 9.1839e-04], + [ 2.7820e-05, -4.6463e-03, 8.4782e-04, ..., -1.3733e-03, + 6.3848e-04, -4.2510e-04]], device='cuda:0') +Epoch 137, bias, value: tensor([-0.0413, 0.0013, 0.0012, -0.0069, -0.0048, 0.0053, -0.0110, 0.0286, + -0.0075, 0.0176], device='cuda:0'), grad: tensor([-0.0289, -0.0669, 0.0300, 0.0124, 0.0056, 0.0404, -0.0326, 0.0211, + 0.0338, -0.0150], device='cuda:0') +100 +0.0001 +changing lr +epoch 136, time 215.29, cls_loss 0.6172 cls_loss_mapping 0.0102 cls_loss_causal 0.5282 re_mapping 0.0144 re_causal 0.0321 /// teacc 98.48 lr 0.00010000 +Epoch 138, weight, value: tensor([[-0.1299, -0.0298, 0.0104, ..., -0.0185, -0.0561, -0.0834], + [-0.0221, -0.0539, 0.0197, ..., 0.0273, -0.0026, -0.0640], + [-0.0330, -0.0530, -0.0712, ..., 0.0402, -0.0363, -0.0663], + ..., + [-0.0426, 0.0275, 0.0246, ..., 0.0301, -0.0318, -0.0685], + [-0.0548, -0.0049, 0.0155, ..., 0.0582, -0.0351, -0.0921], + [ 0.0361, 0.0397, -0.0347, ..., -0.0920, -0.0277, 0.0397]], + device='cuda:0'), grad: tensor([[ 2.0969e-04, 1.2487e-05, 2.0828e-03, ..., 2.0370e-03, + 3.3927e-04, -6.5374e-04], + [ 2.7227e-04, 9.8169e-05, -1.5221e-03, ..., 1.0033e-03, + 3.7813e-04, 8.0705e-05], + [-6.4697e-03, 5.7936e-05, -1.3990e-03, ..., 1.2932e-03, + 6.1560e-04, -4.8676e-03], + ..., + [ 1.2007e-03, 3.1090e-04, 1.1911e-03, ..., -2.2869e-03, + -1.9503e-03, 2.3735e-04], + [-2.4986e-04, 7.0000e-04, -6.3400e-03, ..., -4.0283e-03, + 3.7742e-04, 3.8862e-04], + [-2.3613e-03, -1.8511e-03, 3.1300e-03, ..., -1.0121e-04, + 4.8137e-04, 2.2545e-03]], device='cuda:0') +Epoch 138, bias, value: tensor([-0.0416, 0.0005, 0.0014, -0.0070, -0.0039, 0.0049, -0.0110, 0.0291, + -0.0070, 0.0171], device='cuda:0'), grad: tensor([ 0.0156, 0.0061, 0.0054, -0.0436, 0.0063, 0.0313, 0.0210, -0.0197, + -0.0328, 0.0104], device='cuda:0') +100 +0.0001 +changing lr +epoch 137, time 215.28, cls_loss 0.6337 cls_loss_mapping 0.0073 cls_loss_causal 0.5455 re_mapping 0.0142 re_causal 0.0337 /// teacc 98.65 lr 0.00010000 +Epoch 139, weight, value: tensor([[-0.1307, -0.0296, 0.0099, ..., -0.0191, -0.0560, -0.0843], + [-0.0219, -0.0547, 0.0209, ..., 0.0284, -0.0004, -0.0644], + [-0.0330, -0.0536, -0.0699, ..., 0.0391, -0.0366, -0.0666], + ..., + [-0.0427, 0.0266, 0.0245, ..., 0.0292, -0.0313, -0.0687], + [-0.0554, -0.0032, 0.0153, ..., 0.0594, -0.0348, -0.0930], + [ 0.0366, 0.0414, -0.0358, ..., -0.0913, -0.0283, 0.0400]], + device='cuda:0'), grad: tensor([[-1.3697e-04, 8.6880e-04, 2.3136e-03, ..., 4.3793e-03, + 1.0090e-03, 1.0042e-03], + [ 1.5602e-05, 3.1233e-04, 1.2989e-03, ..., 5.8060e-03, + 1.1730e-03, 2.2995e-04], + [ 1.8820e-05, 9.1410e-04, 2.6131e-03, ..., 5.1727e-03, + 1.1015e-03, 6.9160e-03], + ..., + [ 1.5271e-04, -1.9150e-03, -7.9203e-04, ..., -1.3733e-02, + 7.9918e-04, -1.4296e-03], + [ 2.0015e-04, 2.9492e-04, 1.9217e-03, ..., -6.9122e-03, + -3.5858e-03, 4.1103e-04], + [-5.3215e-04, 3.1967e-03, -5.6267e-03, ..., 2.6436e-03, + -1.8787e-03, 1.5621e-03]], device='cuda:0') +Epoch 139, bias, value: tensor([-0.0423, 0.0009, 0.0011, -0.0072, -0.0039, 0.0048, -0.0109, 0.0290, + -0.0067, 0.0175], device='cuda:0'), grad: tensor([ 0.0232, 0.0244, 0.0255, -0.0401, 0.0214, -0.0144, 0.0186, -0.0394, + -0.0238, 0.0045], device='cuda:0') +100 +0.0001 +changing lr +epoch 138, time 215.25, cls_loss 0.6245 cls_loss_mapping 0.0095 cls_loss_causal 0.5446 re_mapping 0.0147 re_causal 0.0336 /// teacc 98.58 lr 0.00010000 +Epoch 140, weight, value: tensor([[-0.1313, -0.0294, 0.0113, ..., -0.0184, -0.0563, -0.0851], + [-0.0209, -0.0559, 0.0214, ..., 0.0292, -0.0003, -0.0634], + [-0.0331, -0.0528, -0.0705, ..., 0.0389, -0.0378, -0.0673], + ..., + [-0.0433, 0.0279, 0.0232, ..., 0.0292, -0.0312, -0.0688], + [-0.0559, -0.0036, 0.0153, ..., 0.0606, -0.0333, -0.0940], + [ 0.0371, 0.0397, -0.0363, ..., -0.0914, -0.0277, 0.0406]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0004, 0.0010, ..., 0.0025, 0.0003, 0.0007], + [ 0.0039, 0.0047, 0.0043, ..., 0.0097, 0.0027, 0.0008], + [ 0.0007, -0.0043, 0.0025, ..., -0.0294, 0.0011, 0.0006], + ..., + [ 0.0006, 0.0002, 0.0026, ..., 0.0266, 0.0007, 0.0005], + [ 0.0003, 0.0030, -0.0038, ..., -0.0008, -0.0063, -0.0002], + [-0.0082, -0.0014, -0.0094, ..., -0.0009, -0.0008, -0.0017]], + device='cuda:0') +Epoch 140, bias, value: tensor([-0.0417, 0.0015, 0.0008, -0.0071, -0.0038, 0.0041, -0.0111, 0.0290, + -0.0061, 0.0167], device='cuda:0'), grad: tensor([ 0.0187, 0.0509, -0.0326, -0.0004, -0.0350, 0.0127, -0.0034, 0.0440, + -0.0149, -0.0400], device='cuda:0') +100 +0.0001 +changing lr +epoch 139, time 215.31, cls_loss 0.6407 cls_loss_mapping 0.0105 cls_loss_causal 0.5510 re_mapping 0.0143 re_causal 0.0338 /// teacc 98.54 lr 0.00010000 +Epoch 141, weight, value: tensor([[-1.3167e-01, -2.9014e-02, 1.1778e-02, ..., -1.9120e-02, + -5.7323e-02, -8.5191e-02], + [-2.2049e-02, -5.7292e-02, 2.1099e-02, ..., 2.9071e-02, + 7.8232e-06, -6.4065e-02], + [-3.3194e-02, -5.3109e-02, -7.0066e-02, ..., 4.0837e-02, + -3.8607e-02, -6.7493e-02], + ..., + [-4.3120e-02, 2.8189e-02, 2.4304e-02, ..., 2.9609e-02, + -3.1251e-02, -6.8027e-02], + [-5.5595e-02, -3.1558e-03, 1.5306e-02, ..., 6.0232e-02, + -3.3300e-02, -9.4613e-02], + [ 3.7355e-02, 3.9991e-02, -3.6958e-02, ..., -9.0309e-02, + -2.7752e-02, 4.0891e-02]], device='cuda:0'), grad: tensor([[ 3.2425e-05, 7.2765e-04, 1.1057e-04, ..., -5.0278e-03, + 6.5327e-04, 1.0473e-04], + [ 9.2566e-05, 2.6207e-03, 1.4043e-04, ..., 8.4152e-03, + 8.8406e-04, 3.5495e-05], + [ 4.4197e-05, 3.9387e-04, 1.1963e-04, ..., -1.0681e-03, + 4.2129e-04, 1.7196e-05], + ..., + [ 3.2306e-04, 2.6646e-03, 4.4346e-04, ..., -8.3113e-04, + 1.0986e-03, 1.3876e-04], + [ 3.1161e-04, 3.7823e-03, 2.9850e-04, ..., 5.3864e-03, + 9.5987e-04, 1.6367e-04], + [-2.1648e-03, -1.1703e-02, -9.6130e-04, ..., 3.0804e-04, + -7.2441e-03, -1.3685e-03]], device='cuda:0') +Epoch 141, bias, value: tensor([-0.0428, 0.0011, 0.0023, -0.0071, -0.0042, 0.0037, -0.0110, 0.0294, + -0.0068, 0.0176], device='cuda:0'), grad: tensor([-0.0131, 0.0323, -0.0150, 0.0236, 0.0007, -0.0154, -0.0106, 0.0036, + 0.0237, -0.0298], device='cuda:0') +100 +0.0001 +changing lr +epoch 140, time 215.22, cls_loss 0.6388 cls_loss_mapping 0.0123 cls_loss_causal 0.5481 re_mapping 0.0140 re_causal 0.0323 /// teacc 98.59 lr 0.00010000 +Epoch 142, weight, value: tensor([[-0.1332, -0.0279, 0.0118, ..., -0.0191, -0.0577, -0.0856], + [-0.0223, -0.0578, 0.0209, ..., 0.0281, -0.0008, -0.0650], + [-0.0349, -0.0547, -0.0698, ..., 0.0419, -0.0365, -0.0673], + ..., + [-0.0441, 0.0289, 0.0240, ..., 0.0289, -0.0319, -0.0681], + [-0.0541, -0.0040, 0.0148, ..., 0.0609, -0.0337, -0.0948], + [ 0.0386, 0.0408, -0.0364, ..., -0.0911, -0.0279, 0.0412]], + device='cuda:0'), grad: tensor([[ 1.1903e-04, -9.8813e-07, 5.9217e-05, ..., 1.0471e-03, + 5.5122e-03, 6.4313e-05], + [ 2.1160e-04, 1.4044e-05, -1.2338e-04, ..., 8.8215e-04, + 1.3404e-05, 2.5034e-05], + [ 8.9359e-04, 1.2960e-03, 1.8191e-04, ..., 5.3406e-04, + -1.8244e-03, 7.0989e-05], + ..., + [ 1.3475e-03, 1.3075e-03, -7.0572e-04, ..., -4.8714e-03, + 9.0742e-04, 1.2808e-05], + [-2.4395e-03, 1.1545e-04, -2.6112e-03, ..., -2.0123e-03, + -9.3460e-03, 5.6505e-04], + [ 1.3485e-03, 6.0141e-05, 2.5558e-03, ..., 3.9673e-03, + 1.0052e-03, 2.0191e-05]], device='cuda:0') +Epoch 142, bias, value: tensor([-0.0423, 0.0004, 0.0031, -0.0065, -0.0049, 0.0046, -0.0113, 0.0291, + -0.0073, 0.0173], device='cuda:0'), grad: tensor([ 2.4857e-02, 6.1264e-03, 2.0966e-05, 7.0877e-03, 9.6664e-03, + -2.3865e-02, 4.7150e-03, -2.3468e-02, -2.7344e-02, 2.2202e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 141, time 215.50, cls_loss 0.6288 cls_loss_mapping 0.0118 cls_loss_causal 0.5540 re_mapping 0.0135 re_causal 0.0305 /// teacc 98.43 lr 0.00010000 +Epoch 143, weight, value: tensor([[-0.1338, -0.0272, 0.0124, ..., -0.0190, -0.0581, -0.0854], + [-0.0215, -0.0568, 0.0214, ..., 0.0286, -0.0010, -0.0647], + [-0.0352, -0.0540, -0.0726, ..., 0.0413, -0.0360, -0.0694], + ..., + [-0.0439, 0.0277, 0.0244, ..., 0.0285, -0.0314, -0.0693], + [-0.0555, -0.0045, 0.0143, ..., 0.0615, -0.0335, -0.0947], + [ 0.0386, 0.0409, -0.0369, ..., -0.0911, -0.0283, 0.0422]], + device='cuda:0'), grad: tensor([[ 1.4567e-04, -1.4030e-02, 6.8760e-04, ..., 3.2463e-03, + 1.3771e-03, -1.3046e-02], + [ 2.7156e-04, 4.5270e-05, 4.4537e-04, ..., 5.6381e-03, + 2.0180e-03, 9.8109e-05], + [ 4.1413e-04, 4.0889e-04, 9.6416e-04, ..., -4.0588e-03, + 1.6766e-03, 4.1914e-04], + ..., + [ 3.2158e-03, 1.1826e-04, 2.5921e-03, ..., -3.1424e-04, + 7.4959e-04, 2.2817e-04], + [-1.4362e-03, 2.2924e-04, -3.8662e-03, ..., 3.4714e-03, + 9.1076e-04, -9.5510e-04], + [-1.5011e-03, 1.0643e-02, -3.9139e-03, ..., 4.5471e-03, + -7.0524e-04, 8.1406e-03]], device='cuda:0') +Epoch 143, bias, value: tensor([-0.0428, 0.0010, 0.0020, -0.0061, -0.0048, 0.0046, -0.0109, 0.0291, + -0.0070, 0.0172], device='cuda:0'), grad: tensor([-1.7532e-02, 2.5360e-02, -1.2230e-02, -1.1513e-02, -2.4475e-02, + 1.9043e-02, -2.2522e-02, 2.6627e-03, -3.7611e-05, 4.1290e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 142, time 215.49, cls_loss 0.6259 cls_loss_mapping 0.0099 cls_loss_causal 0.5321 re_mapping 0.0139 re_causal 0.0325 /// teacc 98.51 lr 0.00010000 +Epoch 144, weight, value: tensor([[-0.1335, -0.0267, 0.0122, ..., -0.0186, -0.0594, -0.0841], + [-0.0223, -0.0577, 0.0207, ..., 0.0288, -0.0021, -0.0660], + [-0.0356, -0.0543, -0.0736, ..., 0.0413, -0.0373, -0.0697], + ..., + [-0.0454, 0.0280, 0.0251, ..., 0.0282, -0.0321, -0.0696], + [-0.0564, -0.0059, 0.0143, ..., 0.0618, -0.0315, -0.0947], + [ 0.0399, 0.0423, -0.0364, ..., -0.0918, -0.0290, 0.0430]], + device='cuda:0'), grad: tensor([[ 6.6936e-05, -9.2447e-05, 1.5724e-04, ..., -2.1782e-03, + 6.6385e-06, -7.0155e-05], + [ 4.0817e-04, 1.5771e-04, 6.2180e-04, ..., -1.0052e-03, + 7.2643e-07, 3.6812e-04], + [ 9.2411e-04, 4.5985e-05, 1.1940e-03, ..., -5.6648e-03, + -5.1641e-04, 9.3794e-04], + ..., + [ 1.4114e-03, 5.7602e-04, 1.0777e-03, ..., -1.6136e-03, + 3.5405e-05, 1.4629e-03], + [-4.6158e-03, 1.2999e-03, -4.8256e-04, ..., 2.3403e-03, + 3.9577e-05, -1.3580e-03], + [ 1.0178e-02, 1.3252e-02, -2.5787e-03, ..., 2.8381e-03, + 3.1590e-06, 1.7807e-02]], device='cuda:0') +Epoch 144, bias, value: tensor([-0.0432, 0.0008, 0.0015, -0.0057, -0.0037, 0.0042, -0.0113, 0.0283, + -0.0055, 0.0166], device='cuda:0'), grad: tensor([-0.0386, -0.0052, -0.0270, -0.0272, -0.0103, 0.0176, 0.0287, -0.0141, + 0.0185, 0.0575], device='cuda:0') +100 +0.0001 +changing lr +epoch 143, time 215.41, cls_loss 0.6342 cls_loss_mapping 0.0114 cls_loss_causal 0.5453 re_mapping 0.0138 re_causal 0.0314 /// teacc 98.51 lr 0.00010000 +Epoch 145, weight, value: tensor([[-0.1338, -0.0281, 0.0127, ..., -0.0186, -0.0596, -0.0843], + [-0.0234, -0.0598, 0.0208, ..., 0.0299, -0.0030, -0.0670], + [-0.0371, -0.0556, -0.0737, ..., 0.0412, -0.0371, -0.0704], + ..., + [-0.0433, 0.0281, 0.0251, ..., 0.0272, -0.0330, -0.0691], + [-0.0570, -0.0059, 0.0136, ..., 0.0613, -0.0328, -0.0935], + [ 0.0399, 0.0433, -0.0366, ..., -0.0919, -0.0282, 0.0430]], + device='cuda:0'), grad: tensor([[-1.0994e-02, 1.1654e-03, -3.0193e-03, ..., -3.1395e-03, + 9.4700e-04, -3.0804e-03], + [-2.8515e-03, 3.8290e-04, -2.6188e-03, ..., -2.5177e-03, + 2.3804e-03, 1.5593e-04], + [ 1.7300e-03, 2.0012e-05, -2.1954e-03, ..., -2.2173e-04, + -7.4768e-03, 6.2227e-04], + ..., + [ 1.4296e-03, 3.3016e-03, 3.2997e-03, ..., 3.3417e-03, + 4.1962e-03, 3.0398e-04], + [-3.1204e-03, 8.5068e-03, 7.0047e-04, ..., 2.6016e-03, + 1.9608e-03, 6.6853e-04], + [ 4.0894e-03, -2.0645e-02, -5.9547e-03, ..., -6.2332e-03, + -3.6316e-03, -1.5497e-03]], device='cuda:0') +Epoch 145, bias, value: tensor([-0.0427, 0.0014, 0.0007, -0.0055, -0.0028, 0.0045, -0.0116, 0.0289, + -0.0067, 0.0160], device='cuda:0'), grad: tensor([-0.0215, -0.0125, -0.0018, 0.0149, 0.0434, -0.0516, 0.0203, 0.0373, + 0.0182, -0.0467], device='cuda:0') +100 +0.0001 +changing lr +epoch 144, time 215.61, cls_loss 0.6352 cls_loss_mapping 0.0133 cls_loss_causal 0.5488 re_mapping 0.0149 re_causal 0.0326 /// teacc 98.51 lr 0.00010000 +Epoch 146, weight, value: tensor([[-0.1336, -0.0276, 0.0124, ..., -0.0192, -0.0611, -0.0852], + [-0.0236, -0.0594, 0.0227, ..., 0.0301, -0.0003, -0.0670], + [-0.0380, -0.0542, -0.0746, ..., 0.0416, -0.0368, -0.0715], + ..., + [-0.0427, 0.0278, 0.0245, ..., 0.0277, -0.0345, -0.0694], + [-0.0573, -0.0068, 0.0138, ..., 0.0607, -0.0325, -0.0941], + [ 0.0408, 0.0433, -0.0373, ..., -0.0920, -0.0279, 0.0430]], + device='cuda:0'), grad: tensor([[ 7.4196e-04, 4.4327e-03, 8.5735e-04, ..., 1.7309e-03, + -5.8556e-03, 4.4918e-04], + [-5.2500e-04, -1.9455e-03, -3.7813e-04, ..., -1.4984e-02, + -2.1469e-02, 8.8930e-05], + [ 4.2653e-04, 2.0275e-03, 2.3508e-04, ..., 1.6190e-02, + 2.5421e-02, 1.6642e-04], + ..., + [-6.1455e-03, 1.0204e-03, -2.6722e-03, ..., 1.2388e-03, + 1.7328e-03, -5.7335e-03], + [ 7.0000e-04, -1.5259e-02, 2.3091e-04, ..., -2.2411e-03, + -2.4796e-05, -2.7314e-05], + [ 5.2605e-03, 6.0034e-04, 2.1763e-03, ..., 2.9297e-03, + 3.3131e-03, 3.7956e-03]], device='cuda:0') +Epoch 146, bias, value: tensor([-0.0436, 0.0024, 0.0008, -0.0056, -0.0028, 0.0050, -0.0120, 0.0282, + -0.0076, 0.0172], device='cuda:0'), grad: tensor([ 0.0152, -0.0454, 0.0471, -0.0110, -0.0331, 0.0053, 0.0230, -0.0014, + -0.0344, 0.0348], device='cuda:0') +100 +0.0001 +changing lr +epoch 145, time 215.52, cls_loss 0.6267 cls_loss_mapping 0.0094 cls_loss_causal 0.5424 re_mapping 0.0143 re_causal 0.0319 /// teacc 98.62 lr 0.00010000 +Epoch 147, weight, value: tensor([[-1.3435e-01, -2.7540e-02, 1.3251e-02, ..., -1.9509e-02, + -6.0111e-02, -8.5825e-02], + [-2.4170e-02, -5.9526e-02, 2.2393e-02, ..., 2.9982e-02, + -3.3975e-06, -6.7638e-02], + [-3.7395e-02, -5.5080e-02, -7.4315e-02, ..., 4.0965e-02, + -3.6712e-02, -7.1260e-02], + ..., + [-4.2642e-02, 2.7517e-02, 2.5700e-02, ..., 2.7541e-02, + -3.5340e-02, -6.8569e-02], + [-5.6710e-02, -6.8011e-03, 1.4177e-02, ..., 5.9994e-02, + -3.2395e-02, -9.4678e-02], + [ 3.9982e-02, 4.3059e-02, -3.8726e-02, ..., -9.1799e-02, + -2.8021e-02, 4.1716e-02]], device='cuda:0'), grad: tensor([[-2.5616e-03, 2.5201e-04, 4.1649e-06, ..., -3.4046e-03, + -4.1084e-03, 9.4324e-06], + [-7.7057e-04, -6.5994e-04, -1.9608e-03, ..., -3.2310e-03, + 8.7023e-04, 3.7365e-06], + [ 2.1210e-03, 7.9803e-03, 1.6470e-03, ..., 1.1377e-03, + 2.2430e-03, 3.0577e-05], + ..., + [-5.6696e-04, -1.7365e-02, -2.1381e-03, ..., -7.1001e-04, + -9.3699e-05, 1.4486e-03], + [-1.0717e-04, 1.8234e-03, 7.8297e-04, ..., 7.3862e-04, + 2.0838e-04, -3.9043e-03], + [ 2.0638e-03, 2.4586e-03, 1.4009e-03, ..., 1.0900e-03, + 3.9768e-04, 1.5497e-03]], device='cuda:0') +Epoch 147, bias, value: tensor([-0.0441, 0.0024, -0.0001, -0.0053, -0.0020, 0.0059, -0.0116, 0.0281, + -0.0078, 0.0166], device='cuda:0'), grad: tensor([-0.0847, -0.0361, 0.0667, 0.0454, -0.0032, 0.0197, 0.0290, -0.0825, + 0.0163, 0.0293], device='cuda:0') +100 +0.0001 +changing lr +epoch 146, time 215.34, cls_loss 0.6253 cls_loss_mapping 0.0091 cls_loss_causal 0.5497 re_mapping 0.0139 re_causal 0.0318 /// teacc 98.52 lr 0.00010000 +Epoch 148, weight, value: tensor([[-1.3464e-01, -2.8885e-02, 1.2889e-02, ..., -2.0254e-02, + -6.1833e-02, -8.6921e-02], + [-2.5922e-02, -5.9812e-02, 2.1840e-02, ..., 3.0222e-02, + 5.7060e-06, -6.8135e-02], + [-3.5620e-02, -5.5718e-02, -7.4854e-02, ..., 4.1619e-02, + -3.6157e-02, -7.1089e-02], + ..., + [-4.2622e-02, 2.8643e-02, 2.5181e-02, ..., 2.7809e-02, + -3.5973e-02, -6.9587e-02], + [-5.6488e-02, -6.3884e-03, 1.4457e-02, ..., 6.0712e-02, + -3.2094e-02, -9.4473e-02], + [ 4.0215e-02, 4.2762e-02, -3.9473e-02, ..., -9.2000e-02, + -2.7728e-02, 4.2401e-02]], device='cuda:0'), grad: tensor([[ 1.5469e-03, 2.8324e-04, 2.3766e-03, ..., 5.3835e-04, + 3.6407e-04, 1.3638e-03], + [-9.9564e-04, 3.0499e-03, -3.3798e-03, ..., -1.9264e-03, + 1.0982e-05, 4.9412e-05], + [ 1.8167e-04, -1.3474e-02, -1.8477e-04, ..., 5.5742e-04, + -8.9741e-04, 4.6444e-04], + ..., + [ 7.0000e-04, 1.2274e-03, 7.6771e-04, ..., 3.9387e-04, + 2.3261e-05, 5.5361e-04], + [ 1.0056e-02, 4.4785e-03, 8.8501e-03, ..., -2.6846e-04, + 1.9588e-03, 3.7785e-03], + [-7.9956e-03, -4.2847e-02, -1.2684e-03, ..., -1.8921e-03, + 3.2902e-05, -2.7710e-02]], device='cuda:0') +Epoch 148, bias, value: tensor([-0.0442, 0.0034, 0.0010, -0.0063, -0.0024, 0.0053, -0.0123, 0.0288, + -0.0077, 0.0164], device='cuda:0'), grad: tensor([ 0.0143, -0.0025, -0.0208, 0.0311, 0.0139, 0.0051, 0.0017, 0.0114, + 0.0158, -0.0701], device='cuda:0') +100 +0.0001 +changing lr +epoch 147, time 215.76, cls_loss 0.6163 cls_loss_mapping 0.0112 cls_loss_causal 0.5429 re_mapping 0.0142 re_causal 0.0328 /// teacc 98.49 lr 0.00010000 +Epoch 149, weight, value: tensor([[-0.1334, -0.0270, 0.0135, ..., -0.0194, -0.0626, -0.0857], + [-0.0259, -0.0607, 0.0216, ..., 0.0303, -0.0004, -0.0681], + [-0.0372, -0.0551, -0.0756, ..., 0.0417, -0.0361, -0.0718], + ..., + [-0.0424, 0.0291, 0.0255, ..., 0.0282, -0.0368, -0.0693], + [-0.0572, -0.0065, 0.0143, ..., 0.0606, -0.0313, -0.0948], + [ 0.0398, 0.0432, -0.0398, ..., -0.0932, -0.0266, 0.0430]], + device='cuda:0'), grad: tensor([[ 2.0838e-04, 8.6606e-05, 9.1672e-05, ..., 5.3368e-03, + 6.6109e-03, 9.8825e-05], + [-3.7193e-03, -4.0126e-04, -2.8496e-03, ..., -1.7471e-02, + -1.1978e-03, 1.5116e-04], + [ 1.4091e-04, -9.6798e-04, 2.7347e-04, ..., -8.9979e-04, + 1.6060e-03, 4.8542e-04], + ..., + [ 7.9536e-04, 1.0929e-03, -3.0098e-03, ..., -9.9640e-03, + -1.0941e-02, 1.1373e-04], + [ 1.7776e-03, 1.4832e-02, 1.7385e-03, ..., 8.0338e-03, + 1.9169e-03, 5.8746e-04], + [ 1.7967e-03, -2.5330e-02, 3.4466e-03, ..., 2.6855e-03, + 1.9083e-03, 9.6369e-04]], device='cuda:0') +Epoch 149, bias, value: tensor([-0.0430, 0.0030, 0.0012, -0.0061, -0.0032, 0.0061, -0.0123, 0.0288, + -0.0084, 0.0161], device='cuda:0'), grad: tensor([ 0.0382, -0.0611, -0.0066, 0.0114, 0.0453, -0.0090, -0.0234, -0.0347, + 0.0440, -0.0043], device='cuda:0') +100 +0.0001 +changing lr +epoch 148, time 215.48, cls_loss 0.6061 cls_loss_mapping 0.0120 cls_loss_causal 0.5259 re_mapping 0.0135 re_causal 0.0305 /// teacc 98.59 lr 0.00010000 +Epoch 150, weight, value: tensor([[-0.1335, -0.0264, 0.0139, ..., -0.0201, -0.0621, -0.0856], + [-0.0265, -0.0606, 0.0215, ..., 0.0300, -0.0002, -0.0674], + [-0.0376, -0.0565, -0.0753, ..., 0.0419, -0.0363, -0.0727], + ..., + [-0.0437, 0.0293, 0.0259, ..., 0.0288, -0.0360, -0.0702], + [-0.0575, -0.0076, 0.0152, ..., 0.0603, -0.0317, -0.0960], + [ 0.0397, 0.0441, -0.0394, ..., -0.0930, -0.0264, 0.0435]], + device='cuda:0'), grad: tensor([[ 5.7250e-05, -1.6880e-04, -6.7444e-03, ..., -1.1307e-02, + 0.0000e+00, -2.2507e-03], + [ 1.5602e-05, 9.4652e-04, 5.1451e-04, ..., -3.5686e-03, + 0.0000e+00, 4.0793e-04], + [ 3.2091e-04, 1.1721e-03, 2.1229e-03, ..., 3.7231e-03, + 0.0000e+00, 2.1782e-03], + ..., + [ 3.0975e-03, -3.7670e-04, 2.5959e-03, ..., -2.1286e-03, + 0.0000e+00, 3.5362e-03], + [-2.0719e-04, 1.1301e-03, 4.8294e-03, ..., 4.3335e-03, + 0.0000e+00, 1.9627e-03], + [ 2.1744e-03, -1.8387e-02, 2.5787e-03, ..., -5.1842e-03, + 0.0000e+00, -2.2766e-02]], device='cuda:0') +Epoch 150, bias, value: tensor([-0.0428, 0.0030, 0.0011, -0.0061, -0.0037, 0.0062, -0.0120, 0.0298, + -0.0089, 0.0158], device='cuda:0'), grad: tensor([-0.0752, -0.0153, 0.0357, 0.0494, 0.0474, -0.0357, -0.0360, -0.0251, + 0.0352, 0.0195], device='cuda:0') +100 +0.0001 +changing lr +epoch 149, time 215.57, cls_loss 0.6365 cls_loss_mapping 0.0101 cls_loss_causal 0.5542 re_mapping 0.0132 re_causal 0.0296 /// teacc 98.61 lr 0.00010000 +Epoch 151, weight, value: tensor([[-0.1344, -0.0273, 0.0145, ..., -0.0201, -0.0627, -0.0865], + [-0.0265, -0.0610, 0.0223, ..., 0.0302, -0.0008, -0.0670], + [-0.0375, -0.0569, -0.0754, ..., 0.0419, -0.0351, -0.0741], + ..., + [-0.0433, 0.0296, 0.0256, ..., 0.0283, -0.0369, -0.0693], + [-0.0580, -0.0086, 0.0147, ..., 0.0608, -0.0318, -0.0968], + [ 0.0400, 0.0434, -0.0390, ..., -0.0930, -0.0269, 0.0442]], + device='cuda:0'), grad: tensor([[ 4.6301e-04, 4.6539e-04, 3.5691e-04, ..., 2.0809e-03, + -3.1042e-04, 1.2875e-04], + [-1.7500e-03, -2.6703e-04, -8.7070e-04, ..., -9.6283e-03, + 3.3426e-04, -1.2199e-02], + [ 2.8229e-03, 2.7046e-03, 1.9312e-04, ..., 6.6757e-03, + 2.1422e-04, 2.0161e-03], + ..., + [-8.3685e-04, 6.3777e-05, 1.9705e-04, ..., -6.2523e-03, + -8.0049e-05, 9.9659e-05], + [-1.9245e-03, 4.6802e-04, 6.5660e-04, ..., -2.4452e-03, + -9.5510e-04, 5.8985e-04], + [ 1.0948e-03, -2.1982e-04, -2.4033e-03, ..., 2.4948e-03, + 1.1361e-04, 3.8475e-05]], device='cuda:0') +Epoch 151, bias, value: tensor([-0.0425, 0.0033, 0.0009, -0.0063, -0.0040, 0.0062, -0.0126, 0.0298, + -0.0077, 0.0153], device='cuda:0'), grad: tensor([ 0.0147, -0.0828, 0.0682, 0.0450, 0.0255, 0.0509, -0.0328, -0.0766, + 0.0133, -0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 150, time 215.59, cls_loss 0.5870 cls_loss_mapping 0.0083 cls_loss_causal 0.5038 re_mapping 0.0144 re_causal 0.0314 /// teacc 98.59 lr 0.00010000 +Epoch 152, weight, value: tensor([[-0.1359, -0.0267, 0.0144, ..., -0.0196, -0.0624, -0.0883], + [-0.0258, -0.0619, 0.0213, ..., 0.0298, -0.0019, -0.0664], + [-0.0389, -0.0579, -0.0760, ..., 0.0425, -0.0361, -0.0747], + ..., + [-0.0441, 0.0293, 0.0253, ..., 0.0285, -0.0367, -0.0707], + [-0.0576, -0.0090, 0.0156, ..., 0.0605, -0.0320, -0.0978], + [ 0.0397, 0.0438, -0.0392, ..., -0.0936, -0.0267, 0.0439]], + device='cuda:0'), grad: tensor([[ 4.8399e-04, 7.1049e-04, 1.7071e-04, ..., -6.5660e-04, + 7.4434e-04, 8.1062e-04], + [ 3.2902e-04, 9.5963e-05, 1.4210e-04, ..., 7.6675e-03, + 3.3646e-03, 1.4102e-04], + [ 3.1700e-03, 1.1377e-03, -3.4199e-03, ..., -4.2648e-03, + -3.4542e-03, -1.9436e-03], + ..., + [ 1.2035e-03, -3.3569e-03, 1.2982e-04, ..., -5.1041e-03, + 1.6079e-03, 1.1034e-03], + [ 1.1909e-04, 4.5359e-05, 1.3580e-03, ..., -2.5902e-03, + -2.4109e-03, -9.4843e-04], + [ 3.5405e-04, 5.5361e-04, 2.4211e-04, ..., 3.6602e-03, + 5.0116e-04, 6.7806e-04]], device='cuda:0') +Epoch 152, bias, value: tensor([-0.0426, 0.0039, 0.0010, -0.0060, -0.0047, 0.0060, -0.0117, 0.0290, + -0.0074, 0.0148], device='cuda:0'), grad: tensor([-0.0060, 0.0330, -0.0254, -0.0743, -0.0055, 0.0391, 0.0178, 0.0018, + -0.0021, 0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 151, time 215.73, cls_loss 0.6013 cls_loss_mapping 0.0082 cls_loss_causal 0.5187 re_mapping 0.0138 re_causal 0.0317 /// teacc 98.64 lr 0.00010000 +Epoch 153, weight, value: tensor([[-0.1347, -0.0272, 0.0143, ..., -0.0195, -0.0627, -0.0896], + [-0.0262, -0.0620, 0.0223, ..., 0.0300, -0.0008, -0.0660], + [-0.0379, -0.0583, -0.0759, ..., 0.0429, -0.0358, -0.0742], + ..., + [-0.0437, 0.0295, 0.0238, ..., 0.0281, -0.0370, -0.0698], + [-0.0587, -0.0087, 0.0168, ..., 0.0601, -0.0327, -0.0992], + [ 0.0402, 0.0435, -0.0392, ..., -0.0935, -0.0268, 0.0443]], + device='cuda:0'), grad: tensor([[-9.6440e-05, 1.0328e-03, 9.8038e-04, ..., 1.7815e-03, + 1.5535e-03, 1.5819e-04], + [ 8.2850e-05, -6.6467e-02, 4.1695e-03, ..., 2.3270e-03, + 1.5450e-03, 3.6693e-04], + [ 1.3232e-04, 2.8467e-04, 1.3466e-03, ..., 1.6298e-03, + 1.1711e-03, 3.7241e-04], + ..., + [ 3.1304e-04, 5.9387e-02, -2.7161e-03, ..., -2.6398e-03, + 1.0462e-03, -5.1594e-04], + [ 5.9032e-04, 4.2796e-04, 2.6474e-03, ..., -3.0289e-03, + -4.4441e-03, 1.2035e-03], + [ 6.7043e-04, 8.3389e-03, 3.1834e-03, ..., -5.0306e-04, + -6.7043e-04, 1.3781e-03]], device='cuda:0') +Epoch 153, bias, value: tensor([-0.0415, 0.0040, 0.0007, -0.0059, -0.0049, 0.0064, -0.0120, 0.0284, + -0.0082, 0.0153], device='cuda:0'), grad: tensor([ 0.0351, -0.0133, 0.0283, 0.0521, 0.0217, -0.0019, -0.0695, -0.0303, + -0.0005, -0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 152, time 215.45, cls_loss 0.6134 cls_loss_mapping 0.0106 cls_loss_causal 0.5369 re_mapping 0.0130 re_causal 0.0297 /// teacc 98.59 lr 0.00010000 +Epoch 154, weight, value: tensor([[-0.1351, -0.0265, 0.0147, ..., -0.0197, -0.0631, -0.0912], + [-0.0258, -0.0617, 0.0231, ..., 0.0303, -0.0009, -0.0664], + [-0.0394, -0.0586, -0.0767, ..., 0.0432, -0.0354, -0.0731], + ..., + [-0.0431, 0.0298, 0.0232, ..., 0.0278, -0.0383, -0.0691], + [-0.0595, -0.0084, 0.0164, ..., 0.0607, -0.0327, -0.0995], + [ 0.0403, 0.0430, -0.0389, ..., -0.0939, -0.0275, 0.0444]], + device='cuda:0'), grad: tensor([[-0.0026, -0.0032, 0.0009, ..., 0.0012, 0.0014, 0.0010], + [-0.0033, -0.0038, -0.0026, ..., -0.0013, 0.0015, -0.0048], + [ 0.0008, 0.0007, 0.0011, ..., 0.0016, 0.0011, 0.0011], + ..., + [ 0.0029, 0.0010, 0.0050, ..., 0.0043, 0.0023, 0.0010], + [ 0.0009, -0.0006, 0.0010, ..., -0.0007, 0.0011, 0.0009], + [-0.0039, -0.0032, -0.0071, ..., -0.0051, -0.0017, -0.0012]], + device='cuda:0') +Epoch 154, bias, value: tensor([-0.0428, 0.0041, 0.0009, -0.0067, -0.0050, 0.0068, -0.0105, 0.0280, + -0.0076, 0.0152], device='cuda:0'), grad: tensor([ 0.0220, -0.0293, 0.0267, 0.0061, -0.0437, 0.0032, 0.0014, 0.0371, + -0.0062, -0.0172], device='cuda:0') +100 +0.0001 +changing lr +epoch 153, time 215.63, cls_loss 0.6206 cls_loss_mapping 0.0100 cls_loss_causal 0.5507 re_mapping 0.0138 re_causal 0.0323 /// teacc 98.59 lr 0.00010000 +Epoch 155, weight, value: tensor([[-0.1365, -0.0267, 0.0150, ..., -0.0198, -0.0629, -0.0922], + [-0.0249, -0.0618, 0.0248, ..., 0.0300, -0.0011, -0.0651], + [-0.0398, -0.0592, -0.0780, ..., 0.0424, -0.0355, -0.0729], + ..., + [-0.0432, 0.0300, 0.0233, ..., 0.0283, -0.0387, -0.0672], + [-0.0590, -0.0077, 0.0167, ..., 0.0609, -0.0330, -0.1008], + [ 0.0397, 0.0428, -0.0391, ..., -0.0939, -0.0286, 0.0436]], + device='cuda:0'), grad: tensor([[ 1.1759e-03, -9.1374e-05, 1.0443e-03, ..., 1.1921e-03, + 3.2997e-04, 4.7731e-04], + [ 8.9455e-04, 2.6319e-06, 1.4648e-03, ..., 4.7040e-04, + 1.1187e-03, -5.0640e-04], + [-1.3847e-03, 4.4137e-05, -6.2103e-03, ..., -6.7978e-03, + -1.1015e-03, -2.5921e-03], + ..., + [ 1.6441e-03, 2.2277e-06, -2.1706e-03, ..., 5.7077e-04, + -2.0676e-03, 3.9792e-04], + [ 2.3556e-03, 8.0407e-05, 2.0771e-03, ..., 1.6079e-03, + 4.9448e-04, 1.1063e-03], + [ 2.9526e-03, 1.1861e-05, 3.1796e-03, ..., 2.8572e-03, + 7.7152e-04, 1.4019e-03]], device='cuda:0') +Epoch 155, bias, value: tensor([-0.0433, 0.0047, -0.0001, -0.0065, -0.0041, 0.0065, -0.0105, 0.0276, + -0.0069, 0.0150], device='cuda:0'), grad: tensor([ 0.0137, 0.0051, -0.0537, 0.0219, -0.0189, -0.0156, 0.0231, -0.0332, + 0.0212, 0.0364], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 154---------------------------------------------------- +epoch 154, time 216.28, cls_loss 0.6193 cls_loss_mapping 0.0062 cls_loss_causal 0.5336 re_mapping 0.0138 re_causal 0.0331 /// teacc 98.77 lr 0.00010000 +Epoch 156, weight, value: tensor([[-0.1357, -0.0263, 0.0149, ..., -0.0202, -0.0636, -0.0944], + [-0.0262, -0.0610, 0.0229, ..., 0.0297, -0.0010, -0.0637], + [-0.0401, -0.0596, -0.0781, ..., 0.0424, -0.0356, -0.0733], + ..., + [-0.0437, 0.0306, 0.0234, ..., 0.0290, -0.0379, -0.0676], + [-0.0599, -0.0086, 0.0163, ..., 0.0609, -0.0339, -0.1040], + [ 0.0403, 0.0427, -0.0366, ..., -0.0947, -0.0280, 0.0439]], + device='cuda:0'), grad: tensor([[ 2.7001e-05, -1.9409e-02, 2.9251e-05, ..., 2.8682e-04, + 8.3685e-05, 2.3663e-04], + [ 1.4476e-05, 1.4753e-03, 1.0818e-04, ..., -1.9121e-04, + 1.1200e-04, 1.4377e-04], + [-3.5226e-05, -1.7532e-02, 7.1645e-05, ..., 6.9094e-04, + -2.7218e-03, 7.2122e-05], + ..., + [ 4.5955e-05, 2.0950e-02, 2.3687e-04, ..., -3.2501e-03, + -2.8152e-03, 4.7994e-04], + [ 1.0929e-03, 2.1561e-02, 2.4390e-04, ..., 1.8787e-03, + 2.8362e-03, 4.0894e-03], + [-1.4579e-04, -1.2169e-02, 6.4135e-04, ..., -1.0386e-03, + 3.4213e-04, -4.9248e-03]], device='cuda:0') +Epoch 156, bias, value: tensor([-0.0433, 0.0050, -0.0008, -0.0065, -0.0037, 0.0064, -0.0107, 0.0276, + -0.0076, 0.0160], device='cuda:0'), grad: tensor([-0.0246, 0.0089, -0.0199, 0.0092, 0.0082, 0.0019, 0.0060, 0.0244, + 0.0532, -0.0672], device='cuda:0') +100 +0.0001 +changing lr +epoch 155, time 215.67, cls_loss 0.6105 cls_loss_mapping 0.0076 cls_loss_causal 0.5198 re_mapping 0.0139 re_causal 0.0319 /// teacc 98.61 lr 0.00010000 +Epoch 157, weight, value: tensor([[-0.1332, -0.0254, 0.0160, ..., -0.0203, -0.0648, -0.0934], + [-0.0258, -0.0624, 0.0241, ..., 0.0293, -0.0018, -0.0636], + [-0.0394, -0.0593, -0.0802, ..., 0.0424, -0.0362, -0.0744], + ..., + [-0.0430, 0.0310, 0.0225, ..., 0.0293, -0.0363, -0.0691], + [-0.0604, -0.0098, 0.0153, ..., 0.0601, -0.0348, -0.1036], + [ 0.0390, 0.0442, -0.0373, ..., -0.0944, -0.0281, 0.0452]], + device='cuda:0'), grad: tensor([[ 6.5386e-05, -2.4872e-03, 1.0014e-03, ..., 2.5654e-03, + 3.2067e-04, 2.4009e-04], + [ 4.1425e-05, 2.8893e-05, -2.8000e-03, ..., -5.3482e-03, + 2.3091e-04, 3.1948e-05], + [ 1.0669e-04, 3.3045e-04, 1.6069e-03, ..., -1.9569e-03, + 4.3893e-04, 1.2410e-04], + ..., + [-6.7902e-04, 7.2241e-05, 3.8795e-03, ..., 7.4768e-03, + 2.4247e-04, -7.0572e-04], + [ 2.9755e-04, 1.5223e-04, -6.4163e-03, ..., -1.7481e-03, + 9.3699e-05, 1.5175e-04], + [ 2.5201e-04, 5.5647e-04, -7.1049e-04, ..., 1.7633e-03, + -3.3188e-04, -3.3569e-04]], device='cuda:0') +Epoch 157, bias, value: tensor([-0.0431, 0.0044, -0.0019, -0.0056, -0.0040, 0.0064, -0.0100, 0.0287, + -0.0084, 0.0159], device='cuda:0'), grad: tensor([ 0.0036, -0.0152, -0.0032, 0.0120, -0.0186, -0.0493, 0.0396, 0.0298, + -0.0137, 0.0149], device='cuda:0') +100 +0.0001 +changing lr +epoch 156, time 215.53, cls_loss 0.6284 cls_loss_mapping 0.0101 cls_loss_causal 0.5417 re_mapping 0.0136 re_causal 0.0302 /// teacc 98.57 lr 0.00010000 +Epoch 158, weight, value: tensor([[-0.1339, -0.0270, 0.0172, ..., -0.0208, -0.0653, -0.0940], + [-0.0258, -0.0630, 0.0232, ..., 0.0292, -0.0023, -0.0631], + [-0.0406, -0.0588, -0.0809, ..., 0.0430, -0.0352, -0.0751], + ..., + [-0.0433, 0.0302, 0.0231, ..., 0.0289, -0.0369, -0.0698], + [-0.0614, -0.0092, 0.0158, ..., 0.0607, -0.0347, -0.1042], + [ 0.0403, 0.0454, -0.0379, ..., -0.0949, -0.0286, 0.0463]], + device='cuda:0'), grad: tensor([[ 4.1313e-06, 6.3121e-05, 2.4116e-04, ..., 1.6174e-03, + 7.1168e-05, 1.3940e-05], + [ 3.1495e-04, -2.8205e-04, 6.9141e-04, ..., 1.7891e-03, + 1.7416e-04, 2.6450e-06], + [ 2.4624e-06, 7.6294e-05, 1.6851e-03, ..., -2.4204e-03, + 6.1035e-05, 3.3140e-05], + ..., + [ 2.6941e-05, 7.5626e-04, 1.4477e-03, ..., 2.7313e-03, + 4.6253e-04, 2.7299e-05], + [-3.4690e-04, 2.2640e-03, 1.0109e-02, ..., 1.1406e-02, + 3.1624e-03, 3.6895e-05], + [-1.9968e-05, 1.0460e-02, -4.6005e-03, ..., -3.6068e-03, + -1.9407e-03, -1.0155e-05]], device='cuda:0') +Epoch 158, bias, value: tensor([-0.0437, 0.0042, -0.0008, -0.0052, -0.0038, 0.0060, -0.0102, 0.0282, + -0.0079, 0.0155], device='cuda:0'), grad: tensor([ 0.0090, 0.0097, -0.0244, 0.0098, 0.0067, -0.0181, -0.0201, 0.0164, + 0.0359, -0.0250], device='cuda:0') +100 +0.0001 +changing lr +epoch 157, time 215.73, cls_loss 0.6001 cls_loss_mapping 0.0084 cls_loss_causal 0.5104 re_mapping 0.0139 re_causal 0.0313 /// teacc 98.56 lr 0.00010000 +Epoch 159, weight, value: tensor([[-0.1328, -0.0274, 0.0182, ..., -0.0209, -0.0651, -0.0940], + [-0.0267, -0.0636, 0.0233, ..., 0.0294, -0.0043, -0.0642], + [-0.0406, -0.0603, -0.0800, ..., 0.0443, -0.0349, -0.0762], + ..., + [-0.0437, 0.0308, 0.0226, ..., 0.0286, -0.0351, -0.0697], + [-0.0613, -0.0086, 0.0156, ..., 0.0599, -0.0352, -0.1060], + [ 0.0401, 0.0453, -0.0374, ..., -0.0953, -0.0283, 0.0465]], + device='cuda:0'), grad: tensor([[ 2.4959e-05, 8.0109e-03, -1.4210e-04, ..., 1.8787e-03, + -9.9182e-05, -6.9761e-04], + [ 2.2575e-05, 1.4961e-04, 1.2362e-04, ..., 3.6755e-03, + 3.0017e-04, 8.9645e-05], + [ 5.4896e-05, 1.6766e-03, 6.7174e-05, ..., -1.6146e-03, + 2.4164e-04, 3.3975e-04], + ..., + [ 5.4948e-06, 2.9993e-04, -8.5592e-05, ..., 4.8141e-03, + 1.5962e-04, 7.0906e-04], + [ 6.2943e-05, 1.4772e-03, 1.5318e-04, ..., -1.0956e-02, + 2.8753e-04, -2.4624e-03], + [ 4.5389e-05, -1.0004e-03, 4.8804e-04, ..., 7.6904e-03, + 3.7336e-04, 2.0981e-03]], device='cuda:0') +Epoch 159, bias, value: tensor([-0.0435, 0.0040, 0.0004, -0.0061, -0.0029, 0.0059, -0.0107, 0.0275, + -0.0078, 0.0154], device='cuda:0'), grad: tensor([ 0.0278, 0.0287, 0.0004, -0.0069, 0.0185, -0.0489, -0.0186, 0.0270, + -0.0296, 0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 158, time 216.05, cls_loss 0.5965 cls_loss_mapping 0.0091 cls_loss_causal 0.4993 re_mapping 0.0136 re_causal 0.0311 /// teacc 98.53 lr 0.00010000 +Epoch 160, weight, value: tensor([[-0.1331, -0.0270, 0.0183, ..., -0.0212, -0.0654, -0.0958], + [-0.0270, -0.0630, 0.0226, ..., 0.0287, -0.0039, -0.0634], + [-0.0412, -0.0595, -0.0791, ..., 0.0443, -0.0361, -0.0755], + ..., + [-0.0439, 0.0306, 0.0238, ..., 0.0282, -0.0357, -0.0704], + [-0.0621, -0.0092, 0.0168, ..., 0.0613, -0.0350, -0.1078], + [ 0.0393, 0.0434, -0.0383, ..., -0.0946, -0.0289, 0.0458]], + device='cuda:0'), grad: tensor([[ 2.7671e-05, 2.9385e-05, 8.6784e-04, ..., 1.0500e-03, + -4.7226e-03, 1.8013e-04], + [ 1.7548e-04, 2.2268e-04, 7.2050e-04, ..., -8.1778e-04, + 2.6398e-03, 3.9887e-04], + [ 1.2046e-04, -1.0616e-04, 5.8222e-04, ..., -1.6842e-03, + 5.2959e-05, 2.8539e-04], + ..., + [ 1.6785e-04, 1.2512e-03, 2.2507e-03, ..., 1.6108e-03, + 6.1464e-04, 3.6240e-04], + [ 1.2183e-04, 1.5879e-04, -6.8321e-03, ..., 1.1802e-04, + -1.8120e-03, 7.4673e-04], + [ 1.0419e-04, -2.0351e-03, -6.0043e-03, ..., 1.2436e-03, + 2.4629e-04, -9.7418e-04]], device='cuda:0') +Epoch 160, bias, value: tensor([-0.0435, 0.0040, -0.0003, -0.0061, -0.0024, 0.0062, -0.0112, 0.0283, + -0.0073, 0.0146], device='cuda:0'), grad: tensor([ 0.0073, -0.0010, -0.0181, -0.0476, 0.0061, 0.0078, -0.0017, 0.0261, + 0.0078, 0.0133], device='cuda:0') +100 +0.0001 +changing lr +epoch 159, time 215.72, cls_loss 0.5835 cls_loss_mapping 0.0076 cls_loss_causal 0.5039 re_mapping 0.0134 re_causal 0.0309 /// teacc 98.42 lr 0.00010000 +Epoch 161, weight, value: tensor([[-0.1331, -0.0278, 0.0205, ..., -0.0204, -0.0659, -0.0961], + [-0.0275, -0.0643, 0.0236, ..., 0.0293, -0.0034, -0.0632], + [-0.0408, -0.0600, -0.0786, ..., 0.0432, -0.0368, -0.0745], + ..., + [-0.0443, 0.0304, 0.0236, ..., 0.0298, -0.0367, -0.0723], + [-0.0623, -0.0097, 0.0162, ..., 0.0609, -0.0348, -0.1093], + [ 0.0394, 0.0437, -0.0393, ..., -0.0959, -0.0283, 0.0470]], + device='cuda:0'), grad: tensor([[-6.9284e-04, 3.2806e-04, -8.6069e-04, ..., -8.6308e-04, + 7.1669e-04, -1.2589e-03], + [ 2.6554e-05, -4.0412e-05, 9.4295e-05, ..., 2.9507e-03, + 1.1164e-04, 4.0150e-04], + [ 8.5950e-05, 2.0161e-03, 3.9077e-04, ..., 2.4605e-03, + 4.9829e-04, 4.3201e-04], + ..., + [ 1.3149e-04, 2.8870e-02, 1.7941e-04, ..., 2.8439e-03, + 1.8930e-04, 3.2759e-04], + [ 1.3571e-03, -3.6678e-03, 1.6489e-03, ..., 3.3379e-03, + 1.7576e-03, -1.3657e-03], + [ 2.7008e-03, 4.5662e-03, 3.3207e-03, ..., 2.7981e-03, + 3.5419e-03, 1.9836e-04]], device='cuda:0') +Epoch 161, bias, value: tensor([-0.0432, 0.0046, 0.0003, -0.0058, -0.0036, 0.0055, -0.0108, 0.0289, + -0.0077, 0.0140], device='cuda:0'), grad: tensor([-0.0122, 0.0157, 0.0223, -0.0353, -0.0181, -0.0308, -0.0059, 0.0378, + 0.0040, 0.0225], device='cuda:0') +100 +0.0001 +changing lr +epoch 160, time 215.69, cls_loss 0.6010 cls_loss_mapping 0.0090 cls_loss_causal 0.5221 re_mapping 0.0132 re_causal 0.0293 /// teacc 98.48 lr 0.00010000 +Epoch 162, weight, value: tensor([[-0.1332, -0.0275, 0.0204, ..., -0.0219, -0.0663, -0.0955], + [-0.0279, -0.0653, 0.0248, ..., 0.0296, -0.0041, -0.0631], + [-0.0403, -0.0610, -0.0791, ..., 0.0431, -0.0373, -0.0746], + ..., + [-0.0447, 0.0312, 0.0238, ..., 0.0296, -0.0360, -0.0731], + [-0.0628, -0.0087, 0.0148, ..., 0.0613, -0.0342, -0.1112], + [ 0.0400, 0.0422, -0.0392, ..., -0.0964, -0.0297, 0.0481]], + device='cuda:0'), grad: tensor([[-9.5892e-04, -3.2522e-06, -7.6332e-03, ..., -1.0862e-03, + -4.1175e-04, -2.1219e-04], + [ 2.6369e-04, 1.2228e-06, 6.6519e-04, ..., -4.4250e-03, + 1.1152e-04, 6.0886e-05], + [ 9.7334e-05, 2.7791e-05, 7.1764e-04, ..., 1.9798e-03, + 2.1343e-03, 3.5435e-05], + ..., + [-3.4070e-04, -6.7890e-05, 2.3270e-04, ..., 4.0889e-04, + 2.1839e-04, 1.3697e-04], + [ 8.3160e-03, 1.5473e-04, 5.0240e-03, ..., -4.9515e-03, + 1.7862e-03, 6.8283e-04], + [-8.8196e-03, -2.0397e-04, -6.3896e-03, ..., 3.7422e-03, + -6.7997e-04, -8.3637e-04]], device='cuda:0') +Epoch 162, bias, value: tensor([-0.0439, 0.0043, 0.0003, -0.0058, -0.0024, 0.0056, -0.0107, 0.0287, + -0.0072, 0.0136], device='cuda:0'), grad: tensor([-0.0188, -0.0223, 0.0201, -0.0385, 0.0115, 0.0165, 0.0146, 0.0039, + 0.0208, -0.0079], device='cuda:0') +100 +0.0001 +changing lr +epoch 161, time 215.18, cls_loss 0.5935 cls_loss_mapping 0.0075 cls_loss_causal 0.5088 re_mapping 0.0134 re_causal 0.0300 /// teacc 98.59 lr 0.00010000 +Epoch 163, weight, value: tensor([[-0.1341, -0.0272, 0.0201, ..., -0.0216, -0.0657, -0.0964], + [-0.0278, -0.0662, 0.0245, ..., 0.0295, -0.0053, -0.0642], + [-0.0410, -0.0618, -0.0792, ..., 0.0433, -0.0388, -0.0740], + ..., + [-0.0450, 0.0310, 0.0233, ..., 0.0295, -0.0357, -0.0741], + [-0.0632, -0.0102, 0.0157, ..., 0.0610, -0.0336, -0.1117], + [ 0.0406, 0.0438, -0.0387, ..., -0.0965, -0.0302, 0.0489]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0011, 0.0024, ..., 0.0035, 0.0018, 0.0019], + [ 0.0002, 0.0003, 0.0018, ..., 0.0020, 0.0025, 0.0003], + [ 0.0018, 0.0007, 0.0020, ..., -0.0074, 0.0014, -0.0212], + ..., + [ 0.0008, 0.0013, 0.0015, ..., 0.0046, 0.0005, 0.0011], + [ 0.0012, 0.0005, 0.0012, ..., -0.0066, 0.0005, 0.0013], + [ 0.0009, -0.0045, 0.0034, ..., 0.0022, 0.0010, -0.0027]], + device='cuda:0') +Epoch 163, bias, value: tensor([-0.0437, 0.0046, 0.0004, -0.0057, -0.0032, 0.0056, -0.0100, 0.0283, + -0.0074, 0.0134], device='cuda:0'), grad: tensor([ 0.0274, 0.0245, -0.0332, 0.0287, -0.0530, -0.0002, 0.0189, 0.0289, + -0.0584, 0.0163], device='cuda:0') +100 +0.0001 +changing lr +epoch 162, time 214.87, cls_loss 0.5819 cls_loss_mapping 0.0088 cls_loss_causal 0.5082 re_mapping 0.0132 re_causal 0.0302 /// teacc 98.56 lr 0.00010000 +Epoch 164, weight, value: tensor([[-0.1348, -0.0264, 0.0189, ..., -0.0217, -0.0652, -0.0975], + [-0.0280, -0.0659, 0.0237, ..., 0.0290, -0.0066, -0.0643], + [-0.0410, -0.0643, -0.0801, ..., 0.0435, -0.0387, -0.0733], + ..., + [-0.0450, 0.0321, 0.0228, ..., 0.0294, -0.0364, -0.0747], + [-0.0636, -0.0083, 0.0159, ..., 0.0621, -0.0348, -0.1123], + [ 0.0412, 0.0432, -0.0386, ..., -0.0961, -0.0293, 0.0490]], + device='cuda:0'), grad: tensor([[ 1.1235e-04, 1.5259e-04, 8.5258e-04, ..., 1.2243e-04, + 4.2772e-04, 2.6777e-05], + [ 1.3196e-04, -2.4021e-05, 4.0293e-04, ..., 3.1650e-05, + -8.7559e-05, 1.0318e-04], + [ 2.7962e-03, 5.6410e-04, 4.0245e-03, ..., 1.1244e-03, + 2.7580e-03, 9.7632e-05], + ..., + [ 2.1935e-04, 3.1042e-04, -1.2169e-03, ..., -9.8765e-05, + 7.7295e-04, -1.2505e-04], + [ 1.3447e-03, 4.8780e-04, 1.2560e-03, ..., 3.6168e-04, + 1.0710e-03, 4.2009e-04], + [ 3.9959e-04, -1.7986e-03, 1.7853e-03, ..., 3.6502e-04, + -2.6245e-03, 4.9114e-04]], device='cuda:0') +Epoch 164, bias, value: tensor([-0.0445, 0.0038, 0.0008, -0.0063, -0.0036, 0.0057, -0.0099, 0.0283, + -0.0062, 0.0141], device='cuda:0'), grad: tensor([ 0.0082, -0.0242, 0.0333, 0.0130, -0.0263, -0.0165, 0.0069, 0.0017, + 0.0146, -0.0108], device='cuda:0') +100 +0.0001 +changing lr +epoch 163, time 214.72, cls_loss 0.6227 cls_loss_mapping 0.0086 cls_loss_causal 0.5427 re_mapping 0.0133 re_causal 0.0307 /// teacc 98.65 lr 0.00010000 +Epoch 165, weight, value: tensor([[-0.1352, -0.0274, 0.0196, ..., -0.0216, -0.0664, -0.0978], + [-0.0268, -0.0663, 0.0227, ..., 0.0291, -0.0064, -0.0643], + [-0.0411, -0.0645, -0.0806, ..., 0.0431, -0.0393, -0.0725], + ..., + [-0.0451, 0.0318, 0.0226, ..., 0.0305, -0.0361, -0.0751], + [-0.0639, -0.0080, 0.0156, ..., 0.0616, -0.0370, -0.1133], + [ 0.0398, 0.0445, -0.0380, ..., -0.0969, -0.0288, 0.0484]], + device='cuda:0'), grad: tensor([[ 1.7631e-04, -4.0507e-04, 2.7776e-04, ..., -1.4162e-03, + 5.7787e-05, 2.3341e-04], + [ 4.5681e-04, 2.3861e-03, -8.7662e-03, ..., -1.2283e-03, + -1.5287e-03, -2.3258e-04], + [ 5.7793e-04, 4.2820e-04, 3.9768e-04, ..., 6.8998e-04, + 1.5306e-04, 6.4087e-04], + ..., + [-4.0102e-04, 2.1946e-04, 3.7003e-03, ..., 4.7898e-04, + 1.5154e-03, -1.6747e-03], + [ 4.6802e-04, 2.0477e-02, 5.0926e-04, ..., 5.8651e-04, + 9.1019e-03, 5.9795e-04], + [ 5.8699e-04, 3.0494e-04, 3.3894e-03, ..., 1.0157e-03, + 8.5115e-04, 1.2455e-03]], device='cuda:0') +Epoch 165, bias, value: tensor([-0.0450, 0.0036, 0.0008, -0.0071, -0.0024, 0.0055, -0.0102, 0.0283, + -0.0068, 0.0153], device='cuda:0'), grad: tensor([-0.0459, -0.0346, 0.0234, -0.0100, -0.0099, 0.0187, -0.0044, 0.0224, + 0.0419, -0.0017], device='cuda:0') +100 +0.0001 +changing lr +epoch 164, time 214.75, cls_loss 0.6179 cls_loss_mapping 0.0087 cls_loss_causal 0.5323 re_mapping 0.0135 re_causal 0.0302 /// teacc 98.50 lr 0.00010000 +Epoch 166, weight, value: tensor([[-0.1358, -0.0273, 0.0196, ..., -0.0219, -0.0664, -0.0978], + [-0.0257, -0.0685, 0.0229, ..., 0.0296, -0.0072, -0.0646], + [-0.0415, -0.0632, -0.0808, ..., 0.0443, -0.0391, -0.0721], + ..., + [-0.0449, 0.0322, 0.0221, ..., 0.0296, -0.0379, -0.0748], + [-0.0641, -0.0092, 0.0162, ..., 0.0617, -0.0360, -0.1142], + [ 0.0397, 0.0438, -0.0376, ..., -0.0968, -0.0291, 0.0480]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0016, 0.0003, ..., -0.0002, 0.0009, 0.0003], + [ 0.0001, 0.0006, -0.0022, ..., 0.0009, -0.0025, 0.0002], + [ 0.0159, 0.0028, 0.0004, ..., -0.0010, 0.0008, 0.0004], + ..., + [ 0.0002, 0.0006, 0.0006, ..., -0.0004, 0.0009, 0.0003], + [ 0.0020, 0.0008, 0.0008, ..., 0.0013, 0.0009, 0.0005], + [ 0.0007, 0.0008, 0.0004, ..., 0.0012, 0.0012, 0.0007]], + device='cuda:0') +Epoch 166, bias, value: tensor([-0.0449, 0.0041, 0.0011, -0.0073, -0.0026, 0.0047, -0.0100, 0.0286, + -0.0068, 0.0153], device='cuda:0'), grad: tensor([-0.0100, -0.0114, 0.0258, 0.0054, -0.0002, 0.0145, -0.0840, 0.0127, + 0.0256, 0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 165, time 214.58, cls_loss 0.5959 cls_loss_mapping 0.0096 cls_loss_causal 0.5187 re_mapping 0.0132 re_causal 0.0299 /// teacc 98.56 lr 0.00010000 +Epoch 167, weight, value: tensor([[-0.1366, -0.0271, 0.0199, ..., -0.0216, -0.0662, -0.0984], + [-0.0263, -0.0680, 0.0236, ..., 0.0304, -0.0092, -0.0640], + [-0.0416, -0.0638, -0.0820, ..., 0.0445, -0.0382, -0.0725], + ..., + [-0.0455, 0.0326, 0.0204, ..., 0.0297, -0.0384, -0.0759], + [-0.0643, -0.0088, 0.0162, ..., 0.0618, -0.0350, -0.1148], + [ 0.0408, 0.0444, -0.0376, ..., -0.0973, -0.0297, 0.0496]], + device='cuda:0'), grad: tensor([[-7.5912e-04, 4.8304e-04, 6.4850e-04, ..., 1.8158e-03, + 1.2627e-03, 2.0117e-06], + [ 7.8201e-04, 1.3723e-03, 1.3371e-03, ..., 3.7155e-03, + 3.0670e-03, 2.5555e-06], + [ 3.6216e-04, -1.4496e-03, 2.8539e-04, ..., -1.9779e-03, + -1.1702e-03, 5.7966e-06], + ..., + [ 6.4039e-04, -5.2338e-03, 1.4706e-03, ..., 2.4509e-04, + 3.3646e-03, -2.4632e-05], + [ 1.1520e-03, 6.8378e-04, 2.2049e-03, ..., 1.7948e-03, + 1.1971e-02, 8.5831e-06], + [-1.8196e-03, 1.2579e-03, -2.8133e-05, ..., 1.6375e-03, + 1.8988e-03, 2.4915e-05]], device='cuda:0') +Epoch 167, bias, value: tensor([-0.0442, 0.0041, 0.0004, -0.0062, -0.0032, 0.0043, -0.0102, 0.0287, + -0.0065, 0.0148], device='cuda:0'), grad: tensor([ 0.0015, 0.0301, -0.0327, -0.0138, 0.0097, -0.0457, 0.0262, 0.0071, + 0.0002, 0.0174], device='cuda:0') +100 +0.0001 +changing lr +epoch 166, time 214.40, cls_loss 0.6369 cls_loss_mapping 0.0079 cls_loss_causal 0.5456 re_mapping 0.0130 re_causal 0.0305 /// teacc 98.44 lr 0.00010000 +Epoch 168, weight, value: tensor([[-0.1380, -0.0268, 0.0198, ..., -0.0214, -0.0651, -0.0977], + [-0.0257, -0.0666, 0.0233, ..., 0.0298, -0.0099, -0.0646], + [-0.0413, -0.0645, -0.0816, ..., 0.0446, -0.0382, -0.0718], + ..., + [-0.0448, 0.0309, 0.0212, ..., 0.0293, -0.0386, -0.0757], + [-0.0640, -0.0094, 0.0158, ..., 0.0604, -0.0366, -0.1151], + [ 0.0407, 0.0445, -0.0370, ..., -0.0960, -0.0283, 0.0476]], + device='cuda:0'), grad: tensor([[ 6.8583e-06, 2.7671e-05, 3.5763e-04, ..., 4.4670e-03, + 5.0850e-03, 1.5348e-06], + [ 5.4181e-05, 6.3658e-05, 5.1403e-04, ..., 3.3836e-03, + 2.4891e-03, 9.1270e-08], + [ 8.4266e-06, 3.1281e-02, 1.5345e-03, ..., 5.5237e-03, + -1.1971e-02, 2.9653e-06], + ..., + [ 5.1469e-05, -4.4084e-04, -5.5981e-04, ..., -1.1887e-02, + 6.6757e-06, 5.1260e-05], + [-1.8120e-04, 6.4898e-04, -4.1652e-04, ..., 1.3237e-03, + -6.2180e-03, 4.9561e-05], + [ 8.3923e-05, -1.0361e-02, 7.9870e-04, ..., -1.4658e-03, + 1.7090e-03, -2.1565e-04]], device='cuda:0') +Epoch 168, bias, value: tensor([-0.0436, 0.0040, 0.0015, -0.0073, -0.0032, 0.0049, -0.0107, 0.0288, + -0.0070, 0.0147], device='cuda:0'), grad: tensor([ 0.0359, 0.0230, 0.0326, -0.0222, 0.0172, -0.0464, 0.0249, -0.0651, + 0.0033, -0.0035], device='cuda:0') +100 +0.0001 +changing lr +epoch 167, time 214.76, cls_loss 0.6060 cls_loss_mapping 0.0077 cls_loss_causal 0.5203 re_mapping 0.0131 re_causal 0.0302 /// teacc 98.68 lr 0.00010000 +Epoch 169, weight, value: tensor([[-0.1389, -0.0282, 0.0197, ..., -0.0216, -0.0661, -0.0977], + [-0.0256, -0.0680, 0.0226, ..., 0.0297, -0.0114, -0.0655], + [-0.0422, -0.0647, -0.0819, ..., 0.0453, -0.0385, -0.0717], + ..., + [-0.0459, 0.0310, 0.0223, ..., 0.0288, -0.0376, -0.0772], + [-0.0648, -0.0081, 0.0165, ..., 0.0608, -0.0372, -0.1152], + [ 0.0414, 0.0461, -0.0367, ..., -0.0954, -0.0278, 0.0483]], + device='cuda:0'), grad: tensor([[-3.3081e-05, 2.0485e-03, 1.3983e-04, ..., 2.5883e-03, + 2.5539e-03, 1.1432e-04], + [ 7.8306e-06, 1.5430e-03, -2.3613e-03, ..., 7.5722e-03, + 2.1343e-03, 8.4400e-05], + [ 4.7505e-05, 3.1829e-04, 1.2505e-04, ..., -1.9440e-02, + 4.2992e-03, 1.0300e-04], + ..., + [-1.6284e-04, -3.2692e-03, 2.0182e-04, ..., 4.8752e-03, + 3.6411e-03, 1.7023e-04], + [ 2.5153e-04, -8.4610e-03, 5.9557e-04, ..., 2.9583e-03, + 2.0638e-03, 2.6250e-04], + [ 3.0923e-04, 4.5395e-03, 1.9062e-04, ..., -5.6534e-03, + -4.8447e-04, -7.3385e-04]], device='cuda:0') +Epoch 169, bias, value: tensor([-0.0436, 0.0042, 0.0012, -0.0061, -0.0032, 0.0040, -0.0118, 0.0284, + -0.0065, 0.0155], device='cuda:0'), grad: tensor([ 0.0191, 0.0230, -0.0452, 0.0238, -0.0053, 0.0281, -0.0435, 0.0228, + 0.0015, -0.0244], device='cuda:0') +100 +0.0001 +changing lr +epoch 168, time 214.44, cls_loss 0.6081 cls_loss_mapping 0.0085 cls_loss_causal 0.5204 re_mapping 0.0133 re_causal 0.0313 /// teacc 98.55 lr 0.00010000 +Epoch 170, weight, value: tensor([[-0.1403, -0.0279, 0.0195, ..., -0.0234, -0.0668, -0.0974], + [-0.0254, -0.0692, 0.0229, ..., 0.0296, -0.0108, -0.0655], + [-0.0423, -0.0646, -0.0836, ..., 0.0464, -0.0378, -0.0713], + ..., + [-0.0459, 0.0304, 0.0230, ..., 0.0290, -0.0364, -0.0769], + [-0.0647, -0.0078, 0.0161, ..., 0.0613, -0.0371, -0.1153], + [ 0.0407, 0.0472, -0.0372, ..., -0.0947, -0.0270, 0.0475]], + device='cuda:0'), grad: tensor([[ 2.5451e-05, 2.0528e-04, 3.0661e-04, ..., 5.3787e-03, + 1.3695e-03, 1.0952e-05], + [ 1.2958e-04, 1.1754e-04, 4.0627e-04, ..., 6.5765e-03, + 1.4887e-03, 2.2039e-05], + [ 8.4564e-06, 2.2143e-05, 1.1331e-04, ..., 6.1569e-03, + 1.6117e-03, 1.5521e-04], + ..., + [ 5.3912e-05, 2.0421e-04, 1.5354e-04, ..., 7.0419e-03, + 3.3092e-03, -1.4615e-04], + [-1.5364e-03, 1.4555e-04, -1.8272e-03, ..., -5.7907e-03, + 1.2913e-03, 1.2481e-04], + [ 8.3017e-04, 2.1160e-04, 1.8282e-03, ..., -2.1343e-03, + -8.3313e-03, 5.7608e-05]], device='cuda:0') +Epoch 170, bias, value: tensor([-0.0451, 0.0041, 0.0026, -0.0064, -0.0045, 0.0032, -0.0107, 0.0284, + -0.0057, 0.0158], device='cuda:0'), grad: tensor([ 0.0223, 0.0289, 0.0278, -0.0073, -0.0361, -0.0468, 0.0022, 0.0282, + -0.0207, 0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 169, time 214.66, cls_loss 0.6454 cls_loss_mapping 0.0096 cls_loss_causal 0.5670 re_mapping 0.0128 re_causal 0.0311 /// teacc 98.64 lr 0.00010000 +Epoch 171, weight, value: tensor([[-0.1404, -0.0296, 0.0191, ..., -0.0235, -0.0657, -0.0970], + [-0.0253, -0.0696, 0.0242, ..., 0.0303, -0.0116, -0.0659], + [-0.0432, -0.0648, -0.0833, ..., 0.0457, -0.0385, -0.0717], + ..., + [-0.0457, 0.0327, 0.0229, ..., 0.0286, -0.0364, -0.0780], + [-0.0658, -0.0078, 0.0151, ..., 0.0613, -0.0362, -0.1174], + [ 0.0418, 0.0460, -0.0381, ..., -0.0946, -0.0283, 0.0501]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0004, 0.0015, ..., 0.0017, 0.0009, 0.0002], + [ 0.0003, 0.0005, -0.0094, ..., -0.0029, -0.0086, 0.0001], + [ 0.0002, 0.0028, 0.0015, ..., 0.0013, 0.0006, 0.0001], + ..., + [ 0.0003, 0.0034, 0.0023, ..., 0.0023, 0.0013, 0.0001], + [ 0.0023, 0.0008, 0.0038, ..., 0.0026, 0.0012, 0.0010], + [-0.0060, 0.0011, -0.0012, ..., 0.0005, 0.0011, -0.0023]], + device='cuda:0') +Epoch 171, bias, value: tensor([-0.0461, 0.0040, 0.0024, -0.0069, -0.0031, 0.0041, -0.0106, 0.0278, + -0.0055, 0.0158], device='cuda:0'), grad: tensor([ 0.0126, -0.0174, 0.0186, -0.0046, 0.0070, 0.0149, -0.0811, 0.0263, + 0.0205, 0.0032], device='cuda:0') +100 +0.0001 +changing lr +epoch 170, time 214.64, cls_loss 0.5911 cls_loss_mapping 0.0072 cls_loss_causal 0.5104 re_mapping 0.0128 re_causal 0.0292 /// teacc 98.65 lr 0.00010000 +Epoch 172, weight, value: tensor([[-0.1408, -0.0273, 0.0191, ..., -0.0247, -0.0664, -0.0974], + [-0.0258, -0.0697, 0.0237, ..., 0.0318, -0.0103, -0.0669], + [-0.0409, -0.0644, -0.0835, ..., 0.0462, -0.0384, -0.0710], + ..., + [-0.0459, 0.0325, 0.0221, ..., 0.0271, -0.0371, -0.0778], + [-0.0664, -0.0078, 0.0170, ..., 0.0625, -0.0360, -0.1177], + [ 0.0415, 0.0461, -0.0381, ..., -0.0943, -0.0285, 0.0514]], + device='cuda:0'), grad: tensor([[ 1.0115e-04, 2.2888e-05, 3.3355e-04, ..., 2.0046e-03, + 3.5381e-03, 5.1641e-04], + [ 2.7919e-04, 6.4969e-06, 2.3708e-03, ..., 2.2926e-03, + 3.0079e-03, 7.5865e-04], + [ 1.0729e-04, -1.1772e-05, -2.2984e-03, ..., 1.4706e-03, + -6.8359e-03, 4.3154e-04], + ..., + [ 1.2636e-04, 1.9222e-05, 5.2595e-04, ..., 1.6212e-03, + 3.0398e-04, 8.2016e-04], + [ 2.7657e-04, 5.7667e-06, 8.8930e-04, ..., 1.8511e-03, + -5.6982e-04, 7.8535e-04], + [ 4.0817e-04, 2.9922e-04, 2.4915e-04, ..., 1.7567e-03, + 6.6280e-04, 1.4153e-03]], device='cuda:0') +Epoch 172, bias, value: tensor([-0.0460, 0.0049, 0.0027, -0.0071, -0.0036, 0.0044, -0.0112, 0.0272, + -0.0054, 0.0159], device='cuda:0'), grad: tensor([ 0.0290, 0.0537, -0.0074, -0.0247, -0.0401, -0.0185, -0.0278, 0.0168, + -0.0069, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 171, time 214.63, cls_loss 0.6013 cls_loss_mapping 0.0079 cls_loss_causal 0.5139 re_mapping 0.0123 re_causal 0.0284 /// teacc 98.69 lr 0.00010000 +Epoch 173, weight, value: tensor([[-0.1407, -0.0284, 0.0192, ..., -0.0240, -0.0663, -0.0974], + [-0.0257, -0.0690, 0.0237, ..., 0.0325, -0.0094, -0.0670], + [-0.0414, -0.0640, -0.0849, ..., 0.0465, -0.0386, -0.0715], + ..., + [-0.0452, 0.0322, 0.0224, ..., 0.0268, -0.0357, -0.0776], + [-0.0676, -0.0087, 0.0172, ..., 0.0627, -0.0347, -0.1191], + [ 0.0416, 0.0462, -0.0379, ..., -0.0958, -0.0298, 0.0517]], + device='cuda:0'), grad: tensor([[ 2.2203e-06, 2.4307e-04, 1.8415e-03, ..., 2.8248e-03, + 1.6022e-03, 8.0824e-05], + [ 4.0010e-06, 3.6740e-04, -1.3294e-03, ..., -3.8757e-03, + -5.7068e-03, 8.4281e-05], + [ 5.1707e-06, 7.6115e-05, 2.1324e-03, ..., 2.5654e-03, + 1.8024e-04, 7.3051e-04], + ..., + [ 1.7130e-04, 1.3638e-03, -4.7035e-03, ..., -4.2915e-04, + -2.4967e-03, 4.2582e-04], + [ 4.5121e-05, 3.6049e-04, 1.5898e-03, ..., -7.4768e-04, + 2.4052e-03, 2.8467e-04], + [-2.1410e-04, -9.9850e-04, 3.4904e-03, ..., 5.2872e-03, + 2.6741e-03, -2.2006e-04]], device='cuda:0') +Epoch 173, bias, value: tensor([-0.0453, 0.0058, 0.0032, -0.0069, -0.0036, 0.0048, -0.0121, 0.0271, + -0.0063, 0.0152], device='cuda:0'), grad: tensor([ 1.8951e-02, -9.3155e-03, 1.4587e-02, 1.5457e-02, -2.8152e-02, + -2.0203e-02, -1.2772e-02, 3.2455e-05, -8.0719e-03, 2.9449e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 172, time 214.58, cls_loss 0.5971 cls_loss_mapping 0.0087 cls_loss_causal 0.5110 re_mapping 0.0133 re_causal 0.0291 /// teacc 98.67 lr 0.00010000 +Epoch 174, weight, value: tensor([[-0.1421, -0.0287, 0.0185, ..., -0.0250, -0.0664, -0.0988], + [-0.0250, -0.0709, 0.0240, ..., 0.0319, -0.0094, -0.0660], + [-0.0420, -0.0641, -0.0851, ..., 0.0465, -0.0380, -0.0723], + ..., + [-0.0448, 0.0309, 0.0226, ..., 0.0272, -0.0366, -0.0785], + [-0.0681, -0.0084, 0.0172, ..., 0.0630, -0.0339, -0.1207], + [ 0.0406, 0.0474, -0.0398, ..., -0.0959, -0.0299, 0.0515]], + device='cuda:0'), grad: tensor([[-6.6948e-03, -1.0941e-02, -3.0746e-03, ..., -4.8828e-03, + 1.5411e-03, -5.3635e-03], + [ 3.8624e-05, 2.0838e-04, -4.3602e-03, ..., 7.6342e-04, + -1.5326e-03, 1.8668e-04], + [ 2.6464e-04, 7.4539e-03, 2.8157e-04, ..., 1.5593e-03, + 1.6050e-03, 3.2210e-04], + ..., + [ 3.1519e-04, -6.2904e-03, 2.8467e-04, ..., 1.5097e-03, + 8.8930e-04, 5.3692e-04], + [ 1.8144e-04, 1.3666e-03, 3.5143e-04, ..., 8.5545e-04, + 2.5234e-03, 2.1720e-04], + [ 4.5657e-04, -3.5439e-03, 7.0000e-04, ..., -1.9331e-03, + 1.2093e-03, 4.6778e-04]], device='cuda:0') +Epoch 174, bias, value: tensor([-0.0461, 0.0059, 0.0031, -0.0068, -0.0030, 0.0052, -0.0118, 0.0280, + -0.0067, 0.0141], device='cuda:0'), grad: tensor([-0.0271, -0.0075, 0.0237, 0.0491, 0.0351, -0.0592, -0.0225, 0.0062, + 0.0185, -0.0164], device='cuda:0') +100 +0.0001 +changing lr +epoch 173, time 214.69, cls_loss 0.6196 cls_loss_mapping 0.0095 cls_loss_causal 0.5402 re_mapping 0.0132 re_causal 0.0288 /// teacc 98.63 lr 0.00010000 +Epoch 175, weight, value: tensor([[-0.1425, -0.0271, 0.0185, ..., -0.0236, -0.0659, -0.0985], + [-0.0256, -0.0724, 0.0240, ..., 0.0312, -0.0093, -0.0654], + [-0.0428, -0.0649, -0.0852, ..., 0.0464, -0.0383, -0.0732], + ..., + [-0.0450, 0.0315, 0.0222, ..., 0.0272, -0.0373, -0.0799], + [-0.0675, -0.0081, 0.0166, ..., 0.0636, -0.0337, -0.1200], + [ 0.0407, 0.0470, -0.0398, ..., -0.0970, -0.0303, 0.0520]], + device='cuda:0'), grad: tensor([[-1.0319e-05, 9.8705e-04, -3.6097e-04, ..., 8.0919e-04, + -3.1590e-04, 1.2579e-03], + [ 9.3579e-06, 9.1732e-05, -1.0567e-03, ..., -2.5425e-03, + 4.4145e-06, 1.8954e-05], + [ 2.3529e-05, 9.4795e-04, -5.8651e-04, ..., 4.3321e-04, + 3.3307e-04, 5.1403e-04], + ..., + [ 1.2293e-05, 7.8297e-04, 6.7759e-04, ..., 4.5156e-04, + 5.8383e-05, 1.2267e-04], + [ 3.4958e-05, 4.3082e-04, 4.5943e-04, ..., 3.0661e-04, + 2.1625e-04, 1.8048e-04], + [ 1.1832e-05, -7.6175e-05, 3.6564e-03, ..., 3.0270e-03, + 2.5482e-03, 3.7441e-03]], device='cuda:0') +Epoch 175, bias, value: tensor([-0.0454, 0.0059, 0.0034, -0.0068, -0.0023, 0.0053, -0.0126, 0.0282, + -0.0066, 0.0130], device='cuda:0'), grad: tensor([ 0.0114, -0.0343, 0.0058, -0.0229, -0.0303, 0.0055, 0.0160, 0.0155, + 0.0052, 0.0281], device='cuda:0') +100 +0.0001 +changing lr +epoch 174, time 214.48, cls_loss 0.6325 cls_loss_mapping 0.0076 cls_loss_causal 0.5539 re_mapping 0.0122 re_causal 0.0289 /// teacc 98.70 lr 0.00010000 +Epoch 176, weight, value: tensor([[-0.1435, -0.0275, 0.0188, ..., -0.0236, -0.0669, -0.0992], + [-0.0267, -0.0736, 0.0239, ..., 0.0311, -0.0097, -0.0664], + [-0.0431, -0.0639, -0.0861, ..., 0.0456, -0.0386, -0.0731], + ..., + [-0.0451, 0.0321, 0.0211, ..., 0.0276, -0.0387, -0.0806], + [-0.0667, -0.0094, 0.0165, ..., 0.0648, -0.0328, -0.1208], + [ 0.0408, 0.0472, -0.0386, ..., -0.0980, -0.0296, 0.0515]], + device='cuda:0'), grad: tensor([[-0.0002, 0.0012, -0.0016, ..., 0.0007, 0.0003, 0.0004], + [ 0.0002, 0.0016, 0.0021, ..., 0.0030, 0.0035, 0.0010], + [ 0.0218, -0.0010, 0.0009, ..., -0.0018, -0.0015, -0.0002], + ..., + [ 0.0003, -0.0003, 0.0019, ..., 0.0006, 0.0016, 0.0009], + [-0.0026, -0.0016, -0.0069, ..., -0.0052, -0.0084, -0.0049], + [ 0.0006, 0.0009, 0.0026, ..., -0.0017, -0.0044, 0.0019]], + device='cuda:0') +Epoch 176, bias, value: tensor([-0.0461, 0.0053, 0.0030, -0.0063, -0.0025, 0.0055, -0.0127, 0.0284, + -0.0058, 0.0131], device='cuda:0'), grad: tensor([ 0.0259, 0.0524, -0.0018, 0.0114, -0.0337, 0.0329, 0.0078, 0.0132, + -0.0969, -0.0111], device='cuda:0') +100 +0.0001 +changing lr +epoch 175, time 214.66, cls_loss 0.6050 cls_loss_mapping 0.0108 cls_loss_causal 0.5031 re_mapping 0.0133 re_causal 0.0290 /// teacc 98.65 lr 0.00010000 +Epoch 177, weight, value: tensor([[-0.1448, -0.0277, 0.0187, ..., -0.0232, -0.0658, -0.0998], + [-0.0271, -0.0732, 0.0245, ..., 0.0323, -0.0092, -0.0671], + [-0.0451, -0.0642, -0.0865, ..., 0.0450, -0.0382, -0.0733], + ..., + [-0.0458, 0.0320, 0.0215, ..., 0.0284, -0.0393, -0.0809], + [-0.0654, -0.0093, 0.0175, ..., 0.0648, -0.0319, -0.1206], + [ 0.0426, 0.0475, -0.0408, ..., -0.0981, -0.0301, 0.0506]], + device='cuda:0'), grad: tensor([[-7.3433e-05, 1.4591e-04, -2.8276e-04, ..., 1.5793e-03, + 7.5626e-04, 1.1966e-05], + [ 1.6677e-04, 1.7989e-04, -8.1539e-04, ..., 3.4637e-03, + -7.7295e-04, 1.2755e-05], + [ 1.7095e-04, 1.0210e-04, 6.0606e-04, ..., 2.4834e-03, + 1.0582e-02, -1.5140e-04], + ..., + [-5.6419e-03, -8.7404e-04, -3.5992e-03, ..., -1.5945e-02, + -4.7188e-03, 3.9548e-05], + [ 1.3075e-03, -1.2851e-04, 3.1161e-04, ..., 2.8706e-04, + 1.2484e-03, 2.4605e-04], + [-2.5845e-04, 3.2473e-04, 1.6975e-03, ..., 4.9782e-03, + 9.8228e-04, -6.7139e-04]], device='cuda:0') +Epoch 177, bias, value: tensor([-0.0460, 0.0054, 0.0027, -0.0059, -0.0037, 0.0053, -0.0128, 0.0296, + -0.0061, 0.0132], device='cuda:0'), grad: tensor([ 0.0147, 0.0017, 0.0015, -0.0547, 0.0086, 0.0145, 0.0183, -0.0551, + 0.0189, 0.0315], device='cuda:0') +100 +0.0001 +changing lr +epoch 176, time 215.07, cls_loss 0.5660 cls_loss_mapping 0.0065 cls_loss_causal 0.4844 re_mapping 0.0132 re_causal 0.0291 /// teacc 98.50 lr 0.00010000 +Epoch 178, weight, value: tensor([[-0.1451, -0.0269, 0.0185, ..., -0.0240, -0.0666, -0.1005], + [-0.0273, -0.0733, 0.0251, ..., 0.0326, -0.0089, -0.0669], + [-0.0446, -0.0642, -0.0869, ..., 0.0433, -0.0379, -0.0724], + ..., + [-0.0475, 0.0316, 0.0210, ..., 0.0289, -0.0407, -0.0816], + [-0.0657, -0.0094, 0.0173, ..., 0.0653, -0.0307, -0.1215], + [ 0.0424, 0.0476, -0.0408, ..., -0.0978, -0.0302, 0.0508]], + device='cuda:0'), grad: tensor([[ 2.9874e-04, 2.6417e-03, 8.4257e-04, ..., 1.2302e-03, + 6.8307e-05, 1.1206e-03], + [ 1.9908e-04, -3.6163e-03, 1.7822e-04, ..., 1.5440e-03, + -4.2975e-05, 3.7670e-04], + [ 1.4429e-03, 1.2293e-03, 1.2207e-03, ..., -2.6436e-03, + 1.3649e-04, 1.4725e-03], + ..., + [ 1.0080e-03, -3.0017e-04, 2.4872e-03, ..., 1.5545e-03, + 6.3705e-04, 1.9007e-03], + [-5.7411e-03, 1.6966e-03, -7.9575e-03, ..., 3.3021e-04, + 3.2544e-05, -5.3062e-03], + [ 8.3303e-04, 4.2496e-03, 1.2846e-03, ..., 3.1528e-03, + 2.5725e-04, 1.5326e-03]], device='cuda:0') +Epoch 178, bias, value: tensor([-0.0465, 0.0053, 0.0028, -0.0065, -0.0036, 0.0058, -0.0122, 0.0294, + -0.0059, 0.0134], device='cuda:0'), grad: tensor([ 0.0021, -0.0010, -0.0102, -0.0236, 0.0326, 0.0237, -0.0716, 0.0277, + -0.0085, 0.0287], device='cuda:0') +100 +0.0001 +changing lr +epoch 177, time 224.77, cls_loss 0.6485 cls_loss_mapping 0.0074 cls_loss_causal 0.5647 re_mapping 0.0121 re_causal 0.0294 /// teacc 98.55 lr 0.00010000 +Epoch 179, weight, value: tensor([[-0.1462, -0.0269, 0.0192, ..., -0.0243, -0.0669, -0.1007], + [-0.0280, -0.0740, 0.0251, ..., 0.0338, -0.0082, -0.0677], + [-0.0454, -0.0633, -0.0879, ..., 0.0434, -0.0375, -0.0737], + ..., + [-0.0474, 0.0309, 0.0213, ..., 0.0278, -0.0411, -0.0812], + [-0.0650, -0.0103, 0.0190, ..., 0.0655, -0.0308, -0.1214], + [ 0.0433, 0.0486, -0.0409, ..., -0.0968, -0.0312, 0.0508]], + device='cuda:0'), grad: tensor([[ 3.9220e-04, 4.1366e-05, 2.0828e-02, ..., 6.2904e-03, + 2.2030e-03, 6.6936e-05], + [-1.8328e-05, 7.3202e-07, -2.5063e-03, ..., -1.0170e-02, + 2.1000e-03, 3.2872e-05], + [ 2.8777e-04, 9.9763e-06, 3.9787e-03, ..., 1.1139e-02, + 1.7653e-03, 4.3392e-05], + ..., + [ 2.1496e-03, 4.5449e-05, -2.4368e-02, ..., -5.2109e-03, + 1.6413e-03, 4.6015e-04], + [-1.2436e-02, -6.1913e-03, -1.5419e-02, ..., -1.0002e-02, + -2.3575e-03, -7.7286e-03], + [ 2.9869e-03, 5.9319e-03, 6.8359e-03, ..., 1.2047e-02, + 6.2561e-04, 6.7177e-03]], device='cuda:0') +Epoch 179, bias, value: tensor([-0.0460, 0.0058, 0.0027, -0.0068, -0.0034, 0.0052, -0.0114, 0.0286, + -0.0068, 0.0141], device='cuda:0'), grad: tensor([ 0.0373, -0.0156, 0.0370, 0.0172, 0.0206, -0.0303, -0.0145, -0.0308, + -0.0592, 0.0383], device='cuda:0') +100 +0.0001 +changing lr +epoch 178, time 217.90, cls_loss 0.5724 cls_loss_mapping 0.0070 cls_loss_causal 0.4834 re_mapping 0.0128 re_causal 0.0287 /// teacc 98.51 lr 0.00010000 +Epoch 180, weight, value: tensor([[-0.1454, -0.0260, 0.0183, ..., -0.0246, -0.0678, -0.1002], + [-0.0287, -0.0749, 0.0262, ..., 0.0338, -0.0096, -0.0677], + [-0.0455, -0.0637, -0.0883, ..., 0.0435, -0.0374, -0.0743], + ..., + [-0.0470, 0.0310, 0.0204, ..., 0.0271, -0.0421, -0.0811], + [-0.0650, -0.0105, 0.0201, ..., 0.0657, -0.0313, -0.1222], + [ 0.0430, 0.0494, -0.0391, ..., -0.0962, -0.0305, 0.0505]], + device='cuda:0'), grad: tensor([[ 5.4061e-05, 5.7564e-03, 2.8992e-04, ..., 5.4264e-04, + 1.3399e-03, 4.4703e-05], + [ 2.8431e-05, 7.7391e-04, -3.3408e-05, ..., 4.7755e-04, + 2.4719e-03, 4.6074e-05], + [ 1.2465e-03, 5.8508e-04, 2.7714e-03, ..., -1.4114e-03, + -1.0042e-03, 8.6129e-05], + ..., + [ 1.4246e-04, 4.2725e-03, 1.0490e-04, ..., 2.8825e-04, + 9.8038e-04, 6.9857e-04], + [ 1.2481e-04, 1.4715e-03, 3.3760e-04, ..., 1.3695e-03, + -1.9131e-03, 3.9601e-04], + [-1.0747e-04, -7.5188e-03, 2.0862e-04, ..., 4.3225e-04, + 1.0796e-03, -1.3895e-03]], device='cuda:0') +Epoch 180, bias, value: tensor([-0.0458, 0.0062, 0.0024, -0.0074, -0.0042, 0.0059, -0.0108, 0.0284, + -0.0071, 0.0144], device='cuda:0'), grad: tensor([ 0.0275, 0.0267, 0.0141, -0.0648, -0.0076, -0.0280, 0.0106, 0.0220, + -0.0086, 0.0083], device='cuda:0') +100 +0.0001 +changing lr +epoch 179, time 216.46, cls_loss 0.5969 cls_loss_mapping 0.0073 cls_loss_causal 0.5257 re_mapping 0.0123 re_causal 0.0293 /// teacc 98.69 lr 0.00010000 +Epoch 181, weight, value: tensor([[-0.1463, -0.0257, 0.0184, ..., -0.0243, -0.0681, -0.0992], + [-0.0284, -0.0750, 0.0258, ..., 0.0332, -0.0095, -0.0676], + [-0.0455, -0.0636, -0.0881, ..., 0.0430, -0.0366, -0.0743], + ..., + [-0.0470, 0.0313, 0.0201, ..., 0.0272, -0.0422, -0.0816], + [-0.0646, -0.0112, 0.0204, ..., 0.0656, -0.0322, -0.1226], + [ 0.0434, 0.0498, -0.0396, ..., -0.0971, -0.0300, 0.0499]], + device='cuda:0'), grad: tensor([[ 1.0556e-04, 2.8276e-04, -8.3017e-04, ..., 1.4133e-03, + -3.4943e-03, 1.3255e-05], + [-5.8031e-04, 9.0361e-04, 4.5991e-04, ..., -2.1973e-03, + -1.3466e-03, 2.6166e-05], + [ 1.4019e-03, 2.2984e-03, 1.1806e-03, ..., 2.5234e-03, + -2.0943e-03, 4.9412e-05], + ..., + [-3.1643e-03, -4.3640e-03, 1.2932e-03, ..., -7.7209e-03, + 2.2964e-03, 6.7294e-05], + [ 1.7130e-04, -4.5700e-03, 1.6699e-03, ..., 1.9627e-03, + -4.1542e-03, 1.5032e-04], + [ 6.2752e-03, 1.1139e-03, 7.7744e-03, ..., -2.7418e-05, + 2.7084e-03, 3.6068e-03]], device='cuda:0') +Epoch 181, bias, value: tensor([-0.0454, 0.0062, 0.0020, -0.0086, -0.0038, 0.0061, -0.0101, 0.0288, + -0.0076, 0.0145], device='cuda:0'), grad: tensor([-0.0019, -0.0447, 0.0103, -0.0184, -0.0048, 0.0282, 0.0329, -0.0276, + 0.0073, 0.0187], device='cuda:0') +100 +0.0001 +changing lr +epoch 180, time 224.09, cls_loss 0.6058 cls_loss_mapping 0.0088 cls_loss_causal 0.5177 re_mapping 0.0125 re_causal 0.0282 /// teacc 98.75 lr 0.00010000 +Epoch 182, weight, value: tensor([[-0.1467, -0.0263, 0.0175, ..., -0.0242, -0.0685, -0.0992], + [-0.0277, -0.0736, 0.0279, ..., 0.0331, -0.0083, -0.0699], + [-0.0465, -0.0647, -0.0885, ..., 0.0429, -0.0378, -0.0755], + ..., + [-0.0469, 0.0318, 0.0213, ..., 0.0283, -0.0421, -0.0819], + [-0.0638, -0.0099, 0.0200, ..., 0.0651, -0.0316, -0.1226], + [ 0.0429, 0.0500, -0.0398, ..., -0.0976, -0.0300, 0.0498]], + device='cuda:0'), grad: tensor([[ 1.3790e-03, 6.8069e-05, 1.1005e-03, ..., 9.3794e-04, + 1.0672e-03, 9.1887e-04], + [ 2.0351e-03, 1.9426e-03, 2.7733e-03, ..., 1.2026e-03, + 5.8975e-03, 2.5425e-03], + [-6.4316e-03, 3.9458e-04, -1.3641e-02, ..., -9.7322e-04, + -1.1215e-02, -1.9395e-04], + ..., + [ 2.5959e-03, 4.5061e-04, 4.5300e-06, ..., 1.1349e-03, + -1.2655e-03, -5.6410e-04], + [ 1.7204e-03, -4.2343e-03, 5.9204e-03, ..., 1.1616e-03, + 7.2432e-04, -6.1989e-04], + [-5.3978e-03, 1.3866e-03, -5.5122e-03, ..., -2.3994e-03, + -2.7161e-03, -4.4098e-03]], device='cuda:0') +Epoch 182, bias, value: tensor([-0.0452, 0.0072, 0.0014, -0.0088, -0.0041, 0.0053, -0.0096, 0.0296, + -0.0078, 0.0139], device='cuda:0'), grad: tensor([ 0.0176, 0.0248, -0.0799, 0.0117, 0.0380, -0.0331, 0.0307, 0.0081, + 0.0118, -0.0298], device='cuda:0') +100 +0.0001 +changing lr +epoch 181, time 223.91, cls_loss 0.5841 cls_loss_mapping 0.0054 cls_loss_causal 0.5073 re_mapping 0.0121 re_causal 0.0278 /// teacc 98.71 lr 0.00010000 +Epoch 183, weight, value: tensor([[-0.1458, -0.0262, 0.0176, ..., -0.0244, -0.0689, -0.0991], + [-0.0277, -0.0734, 0.0281, ..., 0.0327, -0.0081, -0.0719], + [-0.0458, -0.0647, -0.0882, ..., 0.0433, -0.0387, -0.0746], + ..., + [-0.0457, 0.0321, 0.0209, ..., 0.0284, -0.0429, -0.0798], + [-0.0649, -0.0103, 0.0193, ..., 0.0652, -0.0323, -0.1237], + [ 0.0426, 0.0501, -0.0397, ..., -0.0982, -0.0304, 0.0498]], + device='cuda:0'), grad: tensor([[ 2.7680e-04, 8.6689e-04, 1.4009e-03, ..., 2.5387e-03, + 5.9891e-04, -2.3842e-04], + [ 1.6165e-04, 2.5749e-03, 4.6539e-04, ..., 2.9755e-03, + 1.3142e-03, 6.0171e-05], + [ 9.8705e-04, 1.5287e-03, 5.1212e-04, ..., -8.3399e-04, + -2.7657e-03, 7.2908e-04], + ..., + [ 1.3809e-03, -5.7945e-03, 2.3689e-03, ..., -1.3931e-02, + -2.5311e-03, 1.4839e-03], + [ 1.4200e-03, 3.3016e-03, 5.0850e-03, ..., 3.6793e-03, + 8.1396e-04, 3.1972e-04], + [-1.2039e-02, 3.6221e-03, -3.2177e-03, ..., 5.4502e-04, + 8.4019e-04, -8.6365e-03]], device='cuda:0') +Epoch 183, bias, value: tensor([-0.0445, 0.0067, 0.0021, -0.0084, -0.0038, 0.0054, -0.0105, 0.0294, + -0.0079, 0.0137], device='cuda:0'), grad: tensor([ 0.0184, 0.0296, -0.0427, -0.0044, -0.0090, -0.0096, 0.0189, -0.0082, + 0.0370, -0.0300], device='cuda:0') +100 +0.0001 +changing lr +epoch 182, time 222.32, cls_loss 0.6121 cls_loss_mapping 0.0073 cls_loss_causal 0.5328 re_mapping 0.0120 re_causal 0.0278 /// teacc 98.67 lr 0.00010000 +Epoch 184, weight, value: tensor([[-0.1451, -0.0258, 0.0165, ..., -0.0256, -0.0708, -0.0994], + [-0.0279, -0.0727, 0.0271, ..., 0.0323, -0.0076, -0.0719], + [-0.0462, -0.0651, -0.0876, ..., 0.0428, -0.0385, -0.0745], + ..., + [-0.0464, 0.0324, 0.0218, ..., 0.0293, -0.0440, -0.0795], + [-0.0647, -0.0097, 0.0199, ..., 0.0651, -0.0329, -0.1248], + [ 0.0429, 0.0502, -0.0388, ..., -0.0991, -0.0301, 0.0501]], + device='cuda:0'), grad: tensor([[ 2.0707e-04, 1.8096e-04, 4.4632e-04, ..., 1.1730e-03, + 8.7643e-04, 7.0751e-05], + [ 3.2157e-05, 1.2016e-03, 4.2886e-05, ..., 1.4658e-03, + 1.9283e-03, 1.3661e-04], + [ 1.5342e-04, -5.6801e-03, -2.2182e-03, ..., -3.5024e-04, + -1.1078e-02, 2.7493e-05], + ..., + [ 1.5870e-05, 1.5106e-03, -6.8951e-04, ..., -7.6370e-03, + -6.2256e-03, 3.9160e-05], + [ 2.5043e-03, 1.4687e-03, 1.6832e-03, ..., 1.2608e-03, + -1.8477e-05, 1.9297e-05], + [ 7.9334e-05, 1.7204e-03, 2.9993e-04, ..., 2.7394e-04, + 7.0953e-04, 3.8505e-05]], device='cuda:0') +Epoch 184, bias, value: tensor([-0.0447, 0.0068, 0.0015, -0.0076, -0.0039, 0.0043, -0.0106, 0.0310, + -0.0079, 0.0131], device='cuda:0'), grad: tensor([ 0.0096, 0.0176, -0.0283, 0.0245, -0.0153, 0.0042, -0.0077, -0.0144, + -0.0044, 0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 183, time 229.27, cls_loss 0.5958 cls_loss_mapping 0.0051 cls_loss_causal 0.5169 re_mapping 0.0117 re_causal 0.0284 /// teacc 98.72 lr 0.00010000 +Epoch 185, weight, value: tensor([[-0.1453, -0.0254, 0.0167, ..., -0.0257, -0.0697, -0.0989], + [-0.0283, -0.0724, 0.0276, ..., 0.0328, -0.0075, -0.0699], + [-0.0466, -0.0653, -0.0878, ..., 0.0428, -0.0390, -0.0748], + ..., + [-0.0470, 0.0322, 0.0218, ..., 0.0303, -0.0434, -0.0811], + [-0.0641, -0.0100, 0.0190, ..., 0.0635, -0.0328, -0.1252], + [ 0.0430, 0.0506, -0.0392, ..., -0.0985, -0.0305, 0.0500]], + device='cuda:0'), grad: tensor([[-1.3571e-03, 1.8954e-04, 2.2256e-04, ..., 1.1845e-03, + 3.2654e-03, 7.9811e-05], + [ 6.0368e-04, -4.1275e-03, 1.0633e-03, ..., -2.2621e-03, + 6.4926e-03, 3.7456e-04], + [ 1.3990e-03, 4.0855e-03, 1.0118e-03, ..., 6.4850e-03, + 5.2261e-03, 2.3508e-04], + ..., + [-1.7303e-02, 3.5038e-03, 8.7738e-04, ..., -7.4244e-04, + 3.8395e-03, -1.1696e-02], + [ 7.3004e-04, 3.5501e-04, 9.3079e-04, ..., -2.0065e-03, + 4.3297e-03, 1.2022e-04], + [ 1.1009e-02, -4.9477e-03, -4.4250e-03, ..., -1.6594e-04, + 5.1003e-03, 7.3853e-03]], device='cuda:0') +Epoch 185, bias, value: tensor([-0.0448, 0.0070, 0.0016, -0.0072, -0.0037, 0.0046, -0.0116, 0.0306, + -0.0075, 0.0130], device='cuda:0'), grad: tensor([ 0.0005, -0.0130, 0.0467, 0.0143, 0.0314, -0.0422, -0.0214, -0.0062, + -0.0175, 0.0075], device='cuda:0') +100 +0.0001 +changing lr +epoch 184, time 229.30, cls_loss 0.6088 cls_loss_mapping 0.0075 cls_loss_causal 0.5312 re_mapping 0.0126 re_causal 0.0290 /// teacc 98.70 lr 0.00010000 +Epoch 186, weight, value: tensor([[-0.1459, -0.0242, 0.0152, ..., -0.0264, -0.0711, -0.0992], + [-0.0284, -0.0722, 0.0277, ..., 0.0322, -0.0095, -0.0710], + [-0.0463, -0.0660, -0.0867, ..., 0.0436, -0.0383, -0.0749], + ..., + [-0.0477, 0.0327, 0.0221, ..., 0.0296, -0.0434, -0.0812], + [-0.0640, -0.0097, 0.0191, ..., 0.0636, -0.0330, -0.1248], + [ 0.0445, 0.0499, -0.0379, ..., -0.0975, -0.0295, 0.0506]], + device='cuda:0'), grad: tensor([[-2.1133e-03, 2.7585e-04, -3.5334e-04, ..., -3.3550e-03, + 5.4091e-05, -1.1473e-03], + [ 2.8820e-03, 1.7691e-03, 5.6953e-03, ..., 6.9797e-05, + 1.5354e-04, 1.2481e-04], + [ 2.2240e-03, 5.7697e-04, 2.1992e-03, ..., 3.9597e-03, + 3.1292e-05, 1.4019e-03], + ..., + [ 5.3167e-04, 3.5496e-03, 1.4925e-03, ..., 3.6030e-03, + 1.1909e-04, 2.6560e-04], + [-8.8654e-03, 2.8732e-02, -6.9008e-03, ..., -5.1994e-03, + -2.0943e-03, -3.7403e-03], + [-3.2845e-03, -2.8610e-02, -6.3095e-03, ..., 1.2226e-03, + 3.7527e-04, -2.7809e-03]], device='cuda:0') +Epoch 186, bias, value: tensor([-0.0460, 0.0064, 0.0021, -0.0077, -0.0035, 0.0048, -0.0116, 0.0301, + -0.0067, 0.0140], device='cuda:0'), grad: tensor([-0.0258, 0.0094, 0.0216, 0.0086, -0.0106, 0.0203, -0.0185, 0.0305, + 0.0122, -0.0478], device='cuda:0') +100 +0.0001 +changing lr +epoch 185, time 217.65, cls_loss 0.5982 cls_loss_mapping 0.0085 cls_loss_causal 0.5123 re_mapping 0.0130 re_causal 0.0283 /// teacc 98.61 lr 0.00010000 +Epoch 187, weight, value: tensor([[-0.1458, -0.0242, 0.0155, ..., -0.0274, -0.0716, -0.0996], + [-0.0286, -0.0723, 0.0275, ..., 0.0324, -0.0089, -0.0706], + [-0.0467, -0.0648, -0.0872, ..., 0.0431, -0.0371, -0.0759], + ..., + [-0.0473, 0.0311, 0.0222, ..., 0.0308, -0.0440, -0.0829], + [-0.0620, -0.0102, 0.0211, ..., 0.0630, -0.0324, -0.1253], + [ 0.0426, 0.0509, -0.0389, ..., -0.0980, -0.0306, 0.0515]], + device='cuda:0'), grad: tensor([[ 1.2898e-04, 6.7294e-05, 2.7490e-04, ..., 2.5196e-03, + -7.6580e-04, 2.7919e-04], + [ 4.4882e-05, 3.3453e-06, 8.1301e-04, ..., 2.9964e-03, + 2.0638e-05, 4.4912e-05], + [ 3.5238e-04, 1.0026e-04, 8.3590e-04, ..., 2.7313e-03, + 7.1049e-05, 3.1781e-04], + ..., + [ 1.7166e-04, 1.2422e-04, 9.9564e-04, ..., 3.1471e-03, + 2.4512e-05, 3.7193e-04], + [ 4.6641e-05, -1.8967e-02, 6.1703e-04, ..., -3.3131e-03, + 5.8591e-05, 3.5572e-04], + [ 2.0826e-04, 1.7761e-02, 1.3294e-03, ..., -3.1033e-03, + 1.4611e-05, 3.7551e-04]], device='cuda:0') +Epoch 187, bias, value: tensor([-0.0462, 0.0067, 0.0028, -0.0077, -0.0035, 0.0044, -0.0113, 0.0312, + -0.0074, 0.0131], device='cuda:0'), grad: tensor([-0.0041, 0.0198, 0.0282, -0.0324, -0.0010, -0.0148, 0.0094, 0.0217, + -0.0289, 0.0021], device='cuda:0') +100 +0.0001 +changing lr +epoch 186, time 214.65, cls_loss 0.5929 cls_loss_mapping 0.0060 cls_loss_causal 0.5142 re_mapping 0.0123 re_causal 0.0284 /// teacc 98.76 lr 0.00010000 +Epoch 188, weight, value: tensor([[-0.1464, -0.0254, 0.0159, ..., -0.0260, -0.0711, -0.1003], + [-0.0292, -0.0730, 0.0274, ..., 0.0335, -0.0084, -0.0706], + [-0.0487, -0.0651, -0.0869, ..., 0.0436, -0.0373, -0.0760], + ..., + [-0.0464, 0.0305, 0.0217, ..., 0.0306, -0.0450, -0.0811], + [-0.0638, -0.0095, 0.0202, ..., 0.0622, -0.0327, -0.1269], + [ 0.0436, 0.0513, -0.0388, ..., -0.0989, -0.0302, 0.0513]], + device='cuda:0'), grad: tensor([[ 1.0185e-03, 8.4043e-06, -4.4899e-03, ..., -3.5210e-03, + -4.4289e-03, 2.4989e-05], + [ 4.5872e-04, 1.3995e-04, 4.8676e-03, ..., 5.9662e-03, + 5.5361e-04, 3.4380e-04], + [ 5.3406e-03, 1.0854e-04, -5.2834e-03, ..., 9.1028e-04, + 4.7340e-03, 6.1631e-05], + ..., + [ 1.1263e-03, 2.2411e-04, 5.4169e-04, ..., 3.6583e-03, + 1.1663e-03, 5.4550e-04], + [ 3.3588e-03, -1.8847e-04, 5.5504e-04, ..., -3.9444e-03, + 2.3842e-03, 5.4169e-04], + [ 3.7193e-03, 1.3256e-04, 1.0042e-03, ..., -2.7618e-03, + 2.0885e-03, 3.7498e-03]], device='cuda:0') +Epoch 188, bias, value: tensor([-0.0458, 0.0070, 0.0028, -0.0071, -0.0033, 0.0032, -0.0120, 0.0316, + -0.0071, 0.0125], device='cuda:0'), grad: tensor([-0.0589, 0.0403, 0.0141, -0.0138, -0.0079, 0.0221, -0.0096, 0.0248, + -0.0070, -0.0041], device='cuda:0') +100 +0.0001 +changing lr +epoch 187, time 216.81, cls_loss 0.6082 cls_loss_mapping 0.0062 cls_loss_causal 0.5328 re_mapping 0.0122 re_causal 0.0286 /// teacc 98.59 lr 0.00010000 +Epoch 189, weight, value: tensor([[-0.1464, -0.0250, 0.0155, ..., -0.0259, -0.0707, -0.1007], + [-0.0298, -0.0715, 0.0276, ..., 0.0330, -0.0078, -0.0709], + [-0.0498, -0.0654, -0.0876, ..., 0.0439, -0.0379, -0.0762], + ..., + [-0.0461, 0.0309, 0.0218, ..., 0.0310, -0.0464, -0.0811], + [-0.0646, -0.0096, 0.0207, ..., 0.0617, -0.0319, -0.1277], + [ 0.0445, 0.0497, -0.0387, ..., -0.0993, -0.0308, 0.0523]], + device='cuda:0'), grad: tensor([[ 5.5361e-04, 6.5327e-05, 1.9646e-04, ..., 2.8515e-03, + 1.5283e-04, 1.8501e-04], + [ 7.9632e-04, 2.0885e-03, 8.3923e-04, ..., 3.3607e-03, + 6.8903e-04, 2.4188e-04], + [ 1.2970e-03, 2.5535e-04, 1.3590e-03, ..., -2.2640e-03, + 1.4582e-03, 3.4857e-04], + ..., + [ 4.6790e-05, 1.1187e-03, 1.6565e-03, ..., 8.5297e-03, + 1.7357e-03, 2.6741e-03], + [ 4.4990e-04, 1.4277e-03, 3.1304e-04, ..., -1.1879e-02, + 2.3317e-04, -3.2921e-03], + [ 2.1133e-02, 4.7836e-03, -1.0475e-02, ..., -3.4447e-03, + 8.7738e-03, 1.9014e-04]], device='cuda:0') +Epoch 189, bias, value: tensor([-0.0460, 0.0069, 0.0020, -0.0064, -0.0045, 0.0035, -0.0107, 0.0321, + -0.0076, 0.0124], device='cuda:0'), grad: tensor([ 0.0183, 0.0247, -0.0033, -0.0291, 0.0403, -0.0539, 0.0025, 0.0260, + -0.0499, 0.0246], device='cuda:0') +100 +0.0001 +changing lr +epoch 188, time 218.47, cls_loss 0.5973 cls_loss_mapping 0.0083 cls_loss_causal 0.5144 re_mapping 0.0127 re_causal 0.0288 /// teacc 98.69 lr 0.00010000 +Epoch 190, weight, value: tensor([[-0.1473, -0.0248, 0.0162, ..., -0.0266, -0.0719, -0.1006], + [-0.0315, -0.0723, 0.0276, ..., 0.0323, -0.0081, -0.0714], + [-0.0499, -0.0652, -0.0882, ..., 0.0437, -0.0369, -0.0761], + ..., + [-0.0479, 0.0302, 0.0218, ..., 0.0316, -0.0466, -0.0823], + [-0.0644, -0.0078, 0.0207, ..., 0.0622, -0.0317, -0.1274], + [ 0.0449, 0.0492, -0.0390, ..., -0.0997, -0.0312, 0.0529]], + device='cuda:0'), grad: tensor([[ 3.0351e-04, 3.5167e-04, 8.6054e-07, ..., -3.8662e-03, + 1.6779e-05, 4.7803e-04], + [ 1.4234e-04, 3.1090e-04, -3.9399e-05, ..., 2.3518e-03, + -9.8050e-06, 3.4881e-04], + [ 1.3590e-03, 2.5845e-03, 1.0893e-05, ..., 1.9588e-03, + 8.7619e-05, 1.9550e-03], + ..., + [-3.0746e-03, 1.5421e-03, 4.8615e-07, ..., -4.4441e-04, + -1.1802e-05, 3.1233e-04], + [ 7.9918e-04, -2.2068e-03, 9.7528e-06, ..., -2.3613e-03, + 8.0705e-05, 8.7166e-04], + [ 2.7847e-03, -3.4904e-03, 4.4632e-04, ..., 6.9761e-04, + 5.0440e-06, -5.4932e-04]], device='cuda:0') +Epoch 190, bias, value: tensor([-0.0463, 0.0073, 0.0023, -0.0073, -0.0041, 0.0036, -0.0116, 0.0314, + -0.0068, 0.0134], device='cuda:0'), grad: tensor([-0.0164, 0.0242, 0.0374, -0.0232, 0.0149, -0.0172, 0.0124, -0.0341, + -0.0116, 0.0135], device='cuda:0') +100 +0.0001 +changing lr +epoch 189, time 215.51, cls_loss 0.5897 cls_loss_mapping 0.0059 cls_loss_causal 0.5058 re_mapping 0.0123 re_causal 0.0276 /// teacc 98.75 lr 0.00010000 +Epoch 191, weight, value: tensor([[-0.1462, -0.0257, 0.0161, ..., -0.0272, -0.0716, -0.1008], + [-0.0340, -0.0725, 0.0270, ..., 0.0330, -0.0104, -0.0724], + [-0.0520, -0.0646, -0.0884, ..., 0.0432, -0.0377, -0.0771], + ..., + [-0.0486, 0.0303, 0.0222, ..., 0.0304, -0.0457, -0.0833], + [-0.0628, -0.0079, 0.0207, ..., 0.0638, -0.0312, -0.1269], + [ 0.0450, 0.0490, -0.0396, ..., -0.0999, -0.0292, 0.0532]], + device='cuda:0'), grad: tensor([[ 7.9930e-05, 6.6042e-04, 2.2054e-04, ..., 2.7523e-03, + 4.2510e-04, 2.6441e-04], + [ 4.5705e-04, -7.6175e-05, 6.2704e-04, ..., -4.5633e-04, + -2.0325e-04, 3.9130e-05], + [ 8.6486e-05, 4.5319e-03, 3.3164e-04, ..., 2.9354e-03, + 9.8419e-04, 3.3593e-04], + ..., + [ 1.8167e-04, -1.8396e-03, 5.5647e-04, ..., 4.5815e-03, + -3.9124e-04, 1.1101e-03], + [ 1.5287e-03, -2.1133e-03, -1.2058e-04, ..., -4.4746e-03, + 3.1447e-04, 1.2589e-03], + [-2.5153e-04, 1.5726e-03, -3.0365e-03, ..., -4.9019e-04, + -6.4373e-04, -1.9970e-03]], device='cuda:0') +Epoch 191, bias, value: tensor([-0.0465, 0.0072, 0.0014, -0.0068, -0.0045, 0.0039, -0.0113, 0.0313, + -0.0064, 0.0136], device='cuda:0'), grad: tensor([ 0.0143, -0.0098, 0.0165, -0.0413, 0.0225, 0.0015, 0.0009, 0.0192, + -0.0128, -0.0110], device='cuda:0') +100 +0.0001 +changing lr +epoch 190, time 220.42, cls_loss 0.6250 cls_loss_mapping 0.0072 cls_loss_causal 0.5451 re_mapping 0.0126 re_causal 0.0280 /// teacc 98.71 lr 0.00010000 +Epoch 192, weight, value: tensor([[-0.1453, -0.0253, 0.0158, ..., -0.0270, -0.0697, -0.1010], + [-0.0340, -0.0726, 0.0268, ..., 0.0333, -0.0112, -0.0733], + [-0.0520, -0.0643, -0.0876, ..., 0.0446, -0.0378, -0.0781], + ..., + [-0.0484, 0.0305, 0.0230, ..., 0.0307, -0.0456, -0.0833], + [-0.0624, -0.0078, 0.0200, ..., 0.0624, -0.0308, -0.1272], + [ 0.0446, 0.0483, -0.0405, ..., -0.0991, -0.0299, 0.0545]], + device='cuda:0'), grad: tensor([[-1.4269e-04, -1.0863e-05, -1.0830e-04, ..., -4.9858e-03, + -1.1261e-02, -4.4227e-04], + [ 9.1046e-06, 8.1122e-05, -3.8654e-05, ..., 1.1339e-03, + 1.1578e-03, 7.3850e-05], + [ 1.7347e-03, 1.2726e-05, 1.4496e-03, ..., 2.2907e-03, + 3.6201e-03, 9.0301e-05], + ..., + [ 1.1917e-02, -8.6725e-05, -2.3544e-04, ..., 2.2388e-04, + 9.2926e-03, 3.5973e-03], + [-1.2627e-03, -1.2708e-04, -7.2575e-04, ..., -8.4209e-04, + 2.8191e-03, 2.7586e-06], + [-1.1810e-02, 2.6464e-05, 1.8001e-04, ..., 5.2929e-04, + -6.8398e-03, -3.5706e-03]], device='cuda:0') +Epoch 192, bias, value: tensor([-0.0453, 0.0078, 0.0016, -0.0068, -0.0056, 0.0036, -0.0120, 0.0315, + -0.0076, 0.0145], device='cuda:0'), grad: tensor([-0.0494, 0.0094, 0.0288, -0.0228, 0.0116, 0.0115, 0.0036, 0.0204, + -0.0060, -0.0069], device='cuda:0') +100 +0.0001 +changing lr +epoch 191, time 222.23, cls_loss 0.5892 cls_loss_mapping 0.0062 cls_loss_causal 0.5051 re_mapping 0.0126 re_causal 0.0287 /// teacc 98.54 lr 0.00010000 +Epoch 193, weight, value: tensor([[-0.1451, -0.0229, 0.0171, ..., -0.0281, -0.0687, -0.1027], + [-0.0313, -0.0725, 0.0275, ..., 0.0332, -0.0114, -0.0746], + [-0.0517, -0.0649, -0.0870, ..., 0.0452, -0.0377, -0.0776], + ..., + [-0.0486, 0.0309, 0.0216, ..., 0.0306, -0.0473, -0.0832], + [-0.0638, -0.0077, 0.0191, ..., 0.0622, -0.0322, -0.1293], + [ 0.0443, 0.0479, -0.0404, ..., -0.0989, -0.0297, 0.0549]], + device='cuda:0'), grad: tensor([[ 0.0015, 0.0033, 0.0011, ..., 0.0014, 0.0007, 0.0001], + [ 0.0021, 0.0006, 0.0006, ..., 0.0041, 0.0008, 0.0006], + [-0.0001, -0.0208, 0.0006, ..., -0.0031, 0.0011, 0.0002], + ..., + [-0.0038, -0.0003, -0.0037, ..., -0.0003, -0.0047, 0.0004], + [ 0.0009, 0.0023, -0.0009, ..., 0.0017, 0.0024, 0.0004], + [ 0.0005, 0.0013, 0.0009, ..., 0.0013, 0.0008, -0.0021]], + device='cuda:0') +Epoch 193, bias, value: tensor([-0.0451, 0.0083, 0.0025, -0.0071, -0.0053, 0.0049, -0.0126, 0.0310, + -0.0093, 0.0144], device='cuda:0'), grad: tensor([ 0.0360, 0.0390, -0.0479, 0.0265, -0.0303, 0.0007, -0.0066, -0.0395, + -0.0013, 0.0234], device='cuda:0') +100 +0.0001 +changing lr +epoch 192, time 214.67, cls_loss 0.6156 cls_loss_mapping 0.0073 cls_loss_causal 0.5321 re_mapping 0.0126 re_causal 0.0275 /// teacc 98.59 lr 0.00010000 +Epoch 194, weight, value: tensor([[-0.1462, -0.0230, 0.0164, ..., -0.0268, -0.0690, -0.1019], + [-0.0323, -0.0728, 0.0272, ..., 0.0324, -0.0121, -0.0725], + [-0.0508, -0.0635, -0.0870, ..., 0.0457, -0.0377, -0.0779], + ..., + [-0.0498, 0.0316, 0.0216, ..., 0.0312, -0.0453, -0.0836], + [-0.0650, -0.0075, 0.0190, ..., 0.0616, -0.0319, -0.1323], + [ 0.0452, 0.0471, -0.0409, ..., -0.1005, -0.0304, 0.0558]], + device='cuda:0'), grad: tensor([[ 2.5482e-03, 7.2622e-04, -2.9817e-05, ..., 2.2049e-03, + 3.7622e-04, 1.4420e-03], + [ 6.6638e-05, -4.4479e-03, 3.0816e-05, ..., 2.3384e-03, + 2.8563e-04, 6.2561e-04], + [-4.0169e-03, -1.0347e-03, -2.1114e-03, ..., -8.6670e-03, + -4.0221e-04, -6.1417e-03], + ..., + [ 8.1301e-04, 1.3323e-03, 4.1699e-04, ..., 4.4098e-03, + 4.1604e-04, 2.4147e-03], + [ 1.4508e-04, 7.7200e-04, 3.5733e-05, ..., -4.4785e-03, + 3.3593e-04, 7.8535e-04], + [ 7.3147e-04, 1.8005e-03, 1.2374e-04, ..., 3.6888e-03, + 5.2834e-04, 2.2850e-03]], device='cuda:0') +Epoch 194, bias, value: tensor([-0.0455, 0.0082, 0.0026, -0.0080, -0.0051, 0.0053, -0.0111, 0.0318, + -0.0095, 0.0131], device='cuda:0'), grad: tensor([ 0.0166, -0.0029, -0.0606, 0.0602, 0.0128, -0.0698, 0.0071, 0.0302, + -0.0153, 0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 193, time 214.61, cls_loss 0.5770 cls_loss_mapping 0.0077 cls_loss_causal 0.5056 re_mapping 0.0128 re_causal 0.0287 /// teacc 98.74 lr 0.00010000 +Epoch 195, weight, value: tensor([[-0.1464, -0.0236, 0.0165, ..., -0.0276, -0.0695, -0.1014], + [-0.0331, -0.0733, 0.0272, ..., 0.0330, -0.0126, -0.0732], + [-0.0497, -0.0642, -0.0872, ..., 0.0458, -0.0379, -0.0775], + ..., + [-0.0500, 0.0320, 0.0211, ..., 0.0314, -0.0454, -0.0835], + [-0.0655, -0.0068, 0.0189, ..., 0.0622, -0.0323, -0.1321], + [ 0.0458, 0.0473, -0.0406, ..., -0.1008, -0.0306, 0.0555]], + device='cuda:0'), grad: tensor([[ 3.9458e-04, -8.6069e-04, 2.5177e-04, ..., 3.3340e-03, + 6.9666e-04, 3.0398e-04], + [ 3.0565e-04, -4.9877e-04, 2.1183e-04, ..., -1.6336e-03, + 9.3818e-05, 1.2314e-04], + [ 4.0894e-03, 7.7915e-04, 2.7905e-03, ..., 2.9659e-04, + 2.2373e-03, 2.7885e-03], + ..., + [ 4.9353e-04, -1.3374e-02, 3.6597e-04, ..., -1.6724e-02, + -6.3820e-03, 4.3941e-04], + [ 1.7509e-03, 9.1982e-04, 1.0748e-03, ..., -6.2408e-03, + 9.6703e-04, 1.6394e-03], + [ 4.5204e-04, 5.0850e-03, 3.7193e-04, ..., 5.8784e-03, + 2.8992e-03, 3.1710e-04]], device='cuda:0') +Epoch 195, bias, value: tensor([-0.0459, 0.0080, 0.0032, -0.0081, -0.0040, 0.0048, -0.0112, 0.0317, + -0.0090, 0.0123], device='cuda:0'), grad: tensor([ 0.0166, -0.0096, 0.0158, -0.0207, 0.0268, 0.0331, -0.0068, -0.0489, + -0.0266, 0.0204], device='cuda:0') +100 +0.0001 +changing lr +epoch 194, time 214.84, cls_loss 0.6136 cls_loss_mapping 0.0070 cls_loss_causal 0.5446 re_mapping 0.0118 re_causal 0.0274 /// teacc 98.72 lr 0.00010000 +Epoch 196, weight, value: tensor([[-0.1463, -0.0235, 0.0160, ..., -0.0288, -0.0710, -0.1014], + [-0.0333, -0.0719, 0.0280, ..., 0.0332, -0.0127, -0.0723], + [-0.0499, -0.0641, -0.0868, ..., 0.0463, -0.0386, -0.0789], + ..., + [-0.0484, 0.0307, 0.0222, ..., 0.0308, -0.0456, -0.0834], + [-0.0641, -0.0075, 0.0184, ..., 0.0615, -0.0321, -0.1324], + [ 0.0461, 0.0482, -0.0413, ..., -0.1007, -0.0294, 0.0562]], + device='cuda:0'), grad: tensor([[ 0.0004, 0.0002, -0.0002, ..., 0.0018, 0.0002, 0.0006], + [ 0.0003, 0.0037, 0.0004, ..., -0.0092, 0.0015, 0.0014], + [ 0.0003, 0.0012, 0.0008, ..., 0.0096, -0.0051, 0.0003], + ..., + [ 0.0008, 0.0051, 0.0014, ..., 0.0059, 0.0045, 0.0021], + [ 0.0007, -0.0028, 0.0009, ..., -0.0008, -0.0047, 0.0005], + [-0.0013, -0.0089, -0.0004, ..., -0.0017, 0.0022, 0.0007]], + device='cuda:0') +Epoch 196, bias, value: tensor([-0.0465, 0.0087, 0.0033, -0.0077, -0.0044, 0.0034, -0.0106, 0.0316, + -0.0091, 0.0132], device='cuda:0'), grad: tensor([ 0.0188, -0.0190, 0.0195, -0.0015, -0.0596, -0.0318, 0.0213, 0.0405, + 0.0217, -0.0099], device='cuda:0') +100 +0.0001 +changing lr +epoch 195, time 214.42, cls_loss 0.5920 cls_loss_mapping 0.0068 cls_loss_causal 0.5129 re_mapping 0.0119 re_causal 0.0271 /// teacc 98.71 lr 0.00010000 +Epoch 197, weight, value: tensor([[-0.1462, -0.0242, 0.0165, ..., -0.0283, -0.0714, -0.1008], + [-0.0326, -0.0727, 0.0287, ..., 0.0338, -0.0145, -0.0722], + [-0.0479, -0.0636, -0.0867, ..., 0.0457, -0.0384, -0.0791], + ..., + [-0.0486, 0.0309, 0.0219, ..., 0.0312, -0.0432, -0.0838], + [-0.0653, -0.0068, 0.0179, ..., 0.0616, -0.0317, -0.1335], + [ 0.0459, 0.0479, -0.0417, ..., -0.1006, -0.0297, 0.0556]], + device='cuda:0'), grad: tensor([[ 0.0147, 0.0015, 0.0010, ..., 0.0014, 0.0062, 0.0004], + [ 0.0026, -0.0025, 0.0026, ..., 0.0032, 0.0011, 0.0019], + [ 0.0003, 0.0017, -0.0035, ..., -0.0065, 0.0014, -0.0026], + ..., + [ 0.0001, 0.0113, 0.0015, ..., 0.0085, 0.0059, 0.0022], + [ 0.0014, -0.0208, 0.0027, ..., 0.0028, 0.0035, 0.0014], + [ 0.0009, -0.0109, -0.0158, ..., -0.0115, -0.0155, -0.0026]], + device='cuda:0') +Epoch 197, bias, value: tensor([-0.0462, 0.0085, 0.0030, -0.0077, -0.0053, 0.0039, -0.0103, 0.0316, + -0.0088, 0.0131], device='cuda:0'), grad: tensor([ 0.0259, 0.0163, -0.0225, 0.0128, 0.0056, 0.0109, -0.0188, 0.0264, + -0.0015, -0.0551], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 196---------------------------------------------------- +epoch 196, time 215.46, cls_loss 0.5613 cls_loss_mapping 0.0063 cls_loss_causal 0.4777 re_mapping 0.0125 re_causal 0.0276 /// teacc 98.84 lr 0.00010000 +Epoch 198, weight, value: tensor([[-0.1482, -0.0246, 0.0172, ..., -0.0290, -0.0713, -0.1011], + [-0.0330, -0.0730, 0.0292, ..., 0.0351, -0.0144, -0.0729], + [-0.0484, -0.0635, -0.0873, ..., 0.0453, -0.0388, -0.0793], + ..., + [-0.0485, 0.0312, 0.0222, ..., 0.0320, -0.0442, -0.0838], + [-0.0666, -0.0066, 0.0184, ..., 0.0614, -0.0325, -0.1349], + [ 0.0462, 0.0476, -0.0422, ..., -0.1020, -0.0290, 0.0563]], + device='cuda:0'), grad: tensor([[ 0.0003, -0.0003, 0.0002, ..., 0.0024, -0.0003, 0.0001], + [ 0.0008, 0.0002, 0.0017, ..., -0.0012, 0.0007, 0.0002], + [ 0.0029, 0.0001, 0.0091, ..., -0.0085, 0.0006, 0.0020], + ..., + [ 0.0008, 0.0047, 0.0014, ..., 0.0203, 0.0004, 0.0073], + [ 0.0011, -0.0008, 0.0019, ..., 0.0044, 0.0004, 0.0010], + [-0.0036, -0.0056, -0.0097, ..., -0.0100, 0.0002, -0.0124]], + device='cuda:0') +Epoch 198, bias, value: tensor([-0.0464, 0.0092, 0.0023, -0.0073, -0.0046, 0.0037, -0.0107, 0.0322, + -0.0085, 0.0117], device='cuda:0'), grad: tensor([ 0.0104, -0.0057, 0.0138, -0.0215, -0.0135, -0.0085, -0.0111, 0.0569, + 0.0217, -0.0425], device='cuda:0') +100 +0.0001 +changing lr +epoch 197, time 214.58, cls_loss 0.5746 cls_loss_mapping 0.0062 cls_loss_causal 0.4995 re_mapping 0.0123 re_causal 0.0264 /// teacc 98.71 lr 0.00010000 +Epoch 199, weight, value: tensor([[-0.1489, -0.0249, 0.0179, ..., -0.0288, -0.0697, -0.1004], + [-0.0332, -0.0736, 0.0277, ..., 0.0350, -0.0141, -0.0719], + [-0.0472, -0.0645, -0.0890, ..., 0.0453, -0.0381, -0.0814], + ..., + [-0.0486, 0.0321, 0.0227, ..., 0.0308, -0.0464, -0.0824], + [-0.0681, -0.0060, 0.0192, ..., 0.0627, -0.0337, -0.1362], + [ 0.0450, 0.0472, -0.0429, ..., -0.1019, -0.0262, 0.0560]], + device='cuda:0'), grad: tensor([[ 0.0010, -0.0029, 0.0051, ..., 0.0053, 0.0005, 0.0009], + [-0.0027, 0.0006, -0.0020, ..., -0.0018, 0.0011, 0.0011], + [ 0.0007, -0.0042, -0.0071, ..., -0.0065, -0.0023, 0.0008], + ..., + [ 0.0028, 0.0022, 0.0019, ..., 0.0032, 0.0007, 0.0038], + [ 0.0064, 0.0009, 0.0059, ..., 0.0016, 0.0005, 0.0077], + [ 0.0037, 0.0024, 0.0017, ..., 0.0019, 0.0005, 0.0036]], + device='cuda:0') +Epoch 199, bias, value: tensor([-0.0460, 0.0093, 0.0022, -0.0083, -0.0037, 0.0041, -0.0109, 0.0323, + -0.0087, 0.0113], device='cuda:0'), grad: tensor([ 0.0427, -0.0163, -0.0739, -0.0352, -0.0041, -0.0145, 0.0109, 0.0333, + 0.0336, 0.0235], device='cuda:0') +100 +0.0001 +changing lr +epoch 198, time 214.64, cls_loss 0.5830 cls_loss_mapping 0.0055 cls_loss_causal 0.5139 re_mapping 0.0120 re_causal 0.0263 /// teacc 98.58 lr 0.00010000 +Epoch 200, weight, value: tensor([[-0.1496, -0.0254, 0.0187, ..., -0.0287, -0.0706, -0.1002], + [-0.0342, -0.0745, 0.0276, ..., 0.0354, -0.0143, -0.0726], + [-0.0475, -0.0635, -0.0896, ..., 0.0455, -0.0395, -0.0808], + ..., + [-0.0503, 0.0322, 0.0211, ..., 0.0304, -0.0447, -0.0849], + [-0.0684, -0.0060, 0.0204, ..., 0.0624, -0.0335, -0.1389], + [ 0.0453, 0.0469, -0.0429, ..., -0.1008, -0.0268, 0.0579]], + device='cuda:0'), grad: tensor([[ 3.4094e-04, -6.1281e-07, 8.9502e-04, ..., 1.0386e-03, + 2.8782e-03, 1.2045e-03], + [ 2.9251e-05, 1.2314e-04, -1.4582e-03, ..., 9.6655e-04, + -2.1534e-03, 5.1727e-03], + [ 2.9316e-03, 4.6194e-07, 1.8911e-03, ..., -1.8530e-03, + 2.5005e-03, 3.9041e-05], + ..., + [ 3.4118e-04, 5.0240e-03, -1.5869e-03, ..., -2.6913e-03, + 3.2711e-03, 5.6000e-03], + [ 8.0538e-04, 2.2918e-05, 2.9087e-03, ..., 1.8358e-03, + 2.8515e-03, 2.2392e-03], + [-4.3793e-03, 1.7195e-03, -2.5711e-03, ..., -2.9793e-03, + -1.4191e-02, -1.0490e-03]], device='cuda:0') +Epoch 200, bias, value: tensor([-0.0454, 0.0099, 0.0023, -0.0082, -0.0046, 0.0039, -0.0109, 0.0317, + -0.0091, 0.0123], device='cuda:0'), grad: tensor([ 0.0133, -0.0031, 0.0029, -0.0637, 0.0394, 0.0258, -0.0087, -0.0025, + 0.0220, -0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 199, time 214.76, cls_loss 0.5671 cls_loss_mapping 0.0049 cls_loss_causal 0.4924 re_mapping 0.0123 re_causal 0.0282 /// teacc 98.71 lr 0.00010000 +Epoch 201, weight, value: tensor([[-0.1514, -0.0264, 0.0192, ..., -0.0293, -0.0730, -0.1008], + [-0.0342, -0.0750, 0.0275, ..., 0.0358, -0.0150, -0.0733], + [-0.0475, -0.0642, -0.0863, ..., 0.0449, -0.0402, -0.0808], + ..., + [-0.0492, 0.0319, 0.0208, ..., 0.0311, -0.0434, -0.0850], + [-0.0679, -0.0056, 0.0207, ..., 0.0624, -0.0326, -0.1397], + [ 0.0453, 0.0483, -0.0428, ..., -0.0998, -0.0248, 0.0572]], + device='cuda:0'), grad: tensor([[ 2.4581e-04, 1.4172e-03, 1.3075e-03, ..., 2.5978e-03, + 1.8129e-03, 9.4986e-04], + [-3.8147e-03, 7.3612e-05, -8.0338e-03, ..., 1.5659e-03, + 2.4033e-03, 1.2245e-03], + [ 6.2408e-03, 8.1635e-04, 2.7390e-03, ..., -5.1422e-03, + -2.2995e-04, -4.6349e-04], + ..., + [ 1.9588e-03, 6.7810e-02, 1.9722e-03, ..., 1.6678e-02, + 2.6169e-03, 1.2217e-03], + [ 3.7479e-03, 2.8248e-03, 9.5367e-03, ..., 7.5912e-03, + -1.0605e-02, -2.1744e-04], + [-2.5082e-03, -6.0944e-02, -7.4310e-03, ..., -2.8702e-02, + -1.2636e-03, -6.4507e-03]], device='cuda:0') +Epoch 201, bias, value: tensor([-0.0463, 0.0106, 0.0024, -0.0088, -0.0052, 0.0037, -0.0110, 0.0321, + -0.0083, 0.0126], device='cuda:0'), grad: tensor([ 1.7014e-02, -8.0261e-03, -8.0872e-03, 3.9940e-03, -8.5068e-03, + 2.6627e-03, 4.0771e-02, -1.1063e-02, -9.5963e-05, -2.8656e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 200, time 215.08, cls_loss 0.5546 cls_loss_mapping 0.0062 cls_loss_causal 0.4836 re_mapping 0.0117 re_causal 0.0274 /// teacc 98.67 lr 0.00010000 +Epoch 202, weight, value: tensor([[-0.1526, -0.0263, 0.0184, ..., -0.0285, -0.0749, -0.1014], + [-0.0344, -0.0754, 0.0267, ..., 0.0336, -0.0136, -0.0731], + [-0.0473, -0.0643, -0.0864, ..., 0.0457, -0.0405, -0.0818], + ..., + [-0.0497, 0.0307, 0.0206, ..., 0.0313, -0.0443, -0.0857], + [-0.0688, -0.0058, 0.0207, ..., 0.0623, -0.0328, -0.1415], + [ 0.0454, 0.0494, -0.0438, ..., -0.1005, -0.0256, 0.0587]], + device='cuda:0'), grad: tensor([[-6.4552e-05, 7.8440e-04, -2.8682e-04, ..., 1.2856e-03, + 3.8700e-03, 1.3924e-04], + [ 1.2979e-05, 2.7905e-03, 5.2681e-03, ..., 4.4785e-03, + 7.2517e-03, 3.4046e-03], + [ 1.9163e-05, 1.2951e-03, 4.9591e-04, ..., -3.4618e-03, + -2.2873e-02, 1.1044e-03], + ..., + [-3.7774e-06, -7.8888e-03, -2.1946e-04, ..., -1.0040e-02, + 3.9530e-04, -3.5439e-03], + [-1.1182e-04, 4.1122e-03, 1.5068e-03, ..., 3.0727e-03, + 4.3983e-03, 2.0466e-03], + [ 7.4953e-06, 1.0700e-03, 5.1785e-04, ..., 2.4128e-03, + 3.9177e-03, 6.6185e-04]], device='cuda:0') +Epoch 202, bias, value: tensor([-0.0466, 0.0103, 0.0022, -0.0085, -0.0055, 0.0045, -0.0105, 0.0325, + -0.0086, 0.0119], device='cuda:0'), grad: tensor([ 0.0138, 0.0405, -0.0345, -0.0136, -0.0235, 0.0131, 0.0177, -0.0643, + 0.0309, 0.0201], device='cuda:0') +100 +0.0001 +changing lr +epoch 201, time 214.56, cls_loss 0.6166 cls_loss_mapping 0.0072 cls_loss_causal 0.5466 re_mapping 0.0120 re_causal 0.0265 /// teacc 98.78 lr 0.00010000 +Epoch 203, weight, value: tensor([[-0.1524, -0.0276, 0.0201, ..., -0.0279, -0.0739, -0.1018], + [-0.0356, -0.0753, 0.0271, ..., 0.0336, -0.0132, -0.0742], + [-0.0459, -0.0653, -0.0856, ..., 0.0457, -0.0389, -0.0815], + ..., + [-0.0502, 0.0317, 0.0201, ..., 0.0324, -0.0437, -0.0870], + [-0.0684, -0.0040, 0.0190, ..., 0.0614, -0.0338, -0.1390], + [ 0.0456, 0.0479, -0.0447, ..., -0.1017, -0.0239, 0.0607]], + device='cuda:0'), grad: tensor([[ 4.1294e-04, 9.5701e-04, 1.7748e-03, ..., 1.8358e-03, + 3.1834e-03, 1.0767e-03], + [ 1.3895e-03, 2.0966e-02, 3.4523e-03, ..., 8.3971e-04, + 4.2000e-03, 1.1721e-03], + [ 2.4199e-04, 5.3644e-04, 2.4910e-03, ..., 2.3136e-03, + 2.3117e-03, 1.2245e-03], + ..., + [ 4.6659e-04, -2.1469e-02, -2.0485e-03, ..., 4.1962e-04, + -5.4550e-03, 1.4553e-03], + [ 2.2531e-04, 8.6641e-04, 1.1415e-03, ..., 3.0136e-03, + 2.4147e-03, 4.5090e-03], + [-4.8250e-05, 5.2691e-04, 1.4591e-03, ..., 1.4477e-03, + 3.2406e-03, 3.7766e-04]], device='cuda:0') +Epoch 203, bias, value: tensor([-0.0460, 0.0101, 0.0027, -0.0084, -0.0055, 0.0052, -0.0114, 0.0333, + -0.0090, 0.0108], device='cuda:0'), grad: tensor([ 0.0254, 0.0273, 0.0240, -0.0748, 0.0101, -0.0209, -0.0064, -0.0370, + 0.0319, 0.0203], device='cuda:0') +100 +0.0001 +changing lr +epoch 202, time 214.66, cls_loss 0.5879 cls_loss_mapping 0.0062 cls_loss_causal 0.5211 re_mapping 0.0125 re_causal 0.0270 /// teacc 98.68 lr 0.00010000 +Epoch 204, weight, value: tensor([[-0.1537, -0.0275, 0.0196, ..., -0.0279, -0.0747, -0.1026], + [-0.0370, -0.0766, 0.0279, ..., 0.0342, -0.0144, -0.0741], + [-0.0469, -0.0647, -0.0868, ..., 0.0457, -0.0406, -0.0811], + ..., + [-0.0499, 0.0316, 0.0201, ..., 0.0324, -0.0448, -0.0873], + [-0.0674, -0.0044, 0.0190, ..., 0.0609, -0.0345, -0.1364], + [ 0.0454, 0.0485, -0.0429, ..., -0.1014, -0.0232, 0.0615]], + device='cuda:0'), grad: tensor([[ 0.0003, -0.0012, 0.0003, ..., 0.0010, 0.0012, -0.0004], + [ 0.0003, 0.0012, 0.0005, ..., -0.0031, -0.0002, -0.0006], + [ 0.0004, -0.0001, 0.0004, ..., -0.0007, 0.0004, -0.0002], + ..., + [ 0.0009, 0.0045, -0.0019, ..., 0.0015, -0.0033, -0.0005], + [-0.0003, -0.0002, -0.0062, ..., 0.0010, -0.0029, -0.0009], + [-0.0009, -0.0151, 0.0032, ..., -0.0020, 0.0041, -0.0007]], + device='cuda:0') +Epoch 204, bias, value: tensor([-0.0467, 0.0097, 0.0030, -0.0076, -0.0054, 0.0051, -0.0122, 0.0322, + -0.0089, 0.0124], device='cuda:0'), grad: tensor([ 0.0064, -0.0139, 0.0045, 0.0301, 0.0176, 0.0260, -0.0429, -0.0201, + -0.0032, -0.0045], device='cuda:0') +100 +0.0001 +changing lr +epoch 203, time 214.98, cls_loss 0.5753 cls_loss_mapping 0.0060 cls_loss_causal 0.5089 re_mapping 0.0123 re_causal 0.0277 /// teacc 98.74 lr 0.00010000 +Epoch 205, weight, value: tensor([[-0.1527, -0.0282, 0.0203, ..., -0.0283, -0.0753, -0.1036], + [-0.0378, -0.0757, 0.0282, ..., 0.0352, -0.0119, -0.0740], + [-0.0467, -0.0645, -0.0871, ..., 0.0455, -0.0402, -0.0821], + ..., + [-0.0491, 0.0311, 0.0194, ..., 0.0318, -0.0452, -0.0878], + [-0.0679, -0.0035, 0.0184, ..., 0.0609, -0.0362, -0.1357], + [ 0.0459, 0.0488, -0.0425, ..., -0.1011, -0.0237, 0.0613]], + device='cuda:0'), grad: tensor([[ 1.0109e-03, -2.6657e-02, 1.0653e-03, ..., -1.0956e-02, + 1.9722e-03, 8.8024e-04], + [ 4.8995e-05, 4.2305e-03, 1.7941e-04, ..., 4.0512e-03, + 4.9438e-03, 6.5756e-04], + [ 3.9220e-04, -4.1084e-03, 1.3828e-03, ..., -5.2834e-04, + -8.3237e-03, 7.5150e-04], + ..., + [ 1.5771e-04, -2.2621e-03, 3.3522e-04, ..., -2.9812e-03, + 2.5959e-03, -1.4706e-03], + [ 7.0047e-04, 6.2561e-03, 8.1587e-04, ..., 3.3073e-03, + 3.4657e-03, 1.2989e-03], + [-1.4858e-03, -2.1687e-03, 7.8082e-06, ..., 3.5286e-03, + 1.4277e-03, -8.1711e-03]], device='cuda:0') +Epoch 205, bias, value: tensor([-0.0470, 0.0106, 0.0032, -0.0084, -0.0053, 0.0055, -0.0121, 0.0322, + -0.0089, 0.0120], device='cuda:0'), grad: tensor([-0.0247, 0.0390, -0.0200, 0.0016, 0.0204, 0.0258, -0.0491, -0.0180, + 0.0286, -0.0035], device='cuda:0') +100 +0.0001 +changing lr +epoch 204, time 214.85, cls_loss 0.5804 cls_loss_mapping 0.0038 cls_loss_causal 0.5010 re_mapping 0.0124 re_causal 0.0285 /// teacc 98.71 lr 0.00010000 +Epoch 206, weight, value: tensor([[-0.1531, -0.0273, 0.0209, ..., -0.0273, -0.0750, -0.1021], + [-0.0374, -0.0754, 0.0281, ..., 0.0350, -0.0123, -0.0726], + [-0.0475, -0.0661, -0.0867, ..., 0.0457, -0.0407, -0.0829], + ..., + [-0.0491, 0.0313, 0.0190, ..., 0.0327, -0.0449, -0.0870], + [-0.0683, -0.0041, 0.0186, ..., 0.0606, -0.0363, -0.1360], + [ 0.0450, 0.0497, -0.0432, ..., -0.1015, -0.0241, 0.0605]], + device='cuda:0'), grad: tensor([[ 7.5027e-06, 1.7571e-04, 2.3097e-05, ..., -1.8358e-03, + 3.6383e-04, -4.2057e-04], + [ 1.4052e-05, -1.9455e-03, -2.7156e-04, ..., 8.2159e-04, + -3.7384e-04, -9.8324e-04], + [-1.0930e-05, -3.9816e-05, -1.3614e-04, ..., 6.1131e-04, + -2.8152e-03, -4.0007e-04], + ..., + [ 3.5429e-04, 6.1846e-04, 4.2987e-04, ..., 1.2121e-03, + 1.2693e-03, 8.3733e-04], + [-1.1833e-02, 3.8815e-04, 2.1541e-04, ..., 8.2827e-04, + 5.5599e-04, 5.0497e-04], + [-1.3914e-03, -4.5633e-04, 5.3310e-04, ..., -2.6913e-03, + -6.0177e-04, -9.1982e-04]], device='cuda:0') +Epoch 206, bias, value: tensor([-0.0458, 0.0112, 0.0026, -0.0082, -0.0056, 0.0048, -0.0122, 0.0325, + -0.0094, 0.0119], device='cuda:0'), grad: tensor([-0.0190, 0.0056, -0.0232, 0.0287, 0.0155, -0.0179, 0.0109, 0.0226, + -0.0024, -0.0208], device='cuda:0') +100 +0.0001 +changing lr +epoch 205, time 214.85, cls_loss 0.5985 cls_loss_mapping 0.0046 cls_loss_causal 0.5178 re_mapping 0.0121 re_causal 0.0275 /// teacc 98.73 lr 0.00010000 +Epoch 207, weight, value: tensor([[-0.1533, -0.0272, 0.0197, ..., -0.0276, -0.0743, -0.1028], + [-0.0374, -0.0758, 0.0283, ..., 0.0353, -0.0116, -0.0732], + [-0.0484, -0.0658, -0.0869, ..., 0.0461, -0.0405, -0.0827], + ..., + [-0.0481, 0.0306, 0.0202, ..., 0.0329, -0.0461, -0.0868], + [-0.0683, -0.0050, 0.0193, ..., 0.0611, -0.0355, -0.1366], + [ 0.0450, 0.0497, -0.0429, ..., -0.1016, -0.0244, 0.0604]], + device='cuda:0'), grad: tensor([[ 0.0003, -0.0014, 0.0003, ..., 0.0007, 0.0018, 0.0004], + [-0.0014, 0.0004, -0.0014, ..., -0.0056, -0.0039, 0.0011], + [ 0.0008, -0.0010, -0.0006, ..., 0.0005, -0.0019, -0.0041], + ..., + [ 0.0007, 0.0020, 0.0037, ..., -0.0003, 0.0002, 0.0025], + [ 0.0010, 0.0003, 0.0031, ..., 0.0021, 0.0035, 0.0007], + [ 0.0001, -0.0017, -0.0018, ..., 0.0001, -0.0014, -0.0046]], + device='cuda:0') +Epoch 207, bias, value: tensor([-0.0453, 0.0112, 0.0037, -0.0082, -0.0057, 0.0044, -0.0132, 0.0326, + -0.0095, 0.0117], device='cuda:0'), grad: tensor([ 0.0093, -0.0478, -0.0077, 0.0234, 0.0193, -0.0170, -0.0076, 0.0283, + 0.0208, -0.0210], device='cuda:0') +100 +0.0001 +changing lr +epoch 206, time 214.86, cls_loss 0.5707 cls_loss_mapping 0.0042 cls_loss_causal 0.4918 re_mapping 0.0120 re_causal 0.0270 /// teacc 98.73 lr 0.00010000 +Epoch 208, weight, value: tensor([[-0.1527, -0.0272, 0.0181, ..., -0.0285, -0.0759, -0.1022], + [-0.0384, -0.0754, 0.0283, ..., 0.0358, -0.0103, -0.0737], + [-0.0476, -0.0664, -0.0866, ..., 0.0475, -0.0407, -0.0826], + ..., + [-0.0496, 0.0299, 0.0218, ..., 0.0322, -0.0461, -0.0873], + [-0.0677, -0.0044, 0.0186, ..., 0.0597, -0.0363, -0.1376], + [ 0.0456, 0.0488, -0.0433, ..., -0.1016, -0.0232, 0.0610]], + device='cuda:0'), grad: tensor([[ 1.5450e-04, 0.0000e+00, 1.2550e-03, ..., 1.9064e-03, + 2.1725e-03, 2.1026e-05], + [ 1.5891e-04, 0.0000e+00, 1.0063e-02, ..., -2.3293e-04, + 3.8934e-04, 5.2065e-05], + [ 7.6866e-04, 0.0000e+00, -1.7273e-02, ..., -9.2621e-03, + -7.7133e-03, 9.0241e-05], + ..., + [ 4.8876e-04, 0.0000e+00, 1.4477e-03, ..., -3.0842e-03, + 2.7866e-03, -1.5392e-03], + [ 7.5722e-04, 0.0000e+00, 8.8978e-04, ..., 2.1935e-03, + 3.4943e-03, 5.3078e-05], + [ 4.7636e-04, 0.0000e+00, 1.9217e-03, ..., 3.4199e-03, + 3.1471e-03, 1.7719e-03]], device='cuda:0') +Epoch 208, bias, value: tensor([-0.0461, 0.0119, 0.0045, -0.0084, -0.0062, 0.0048, -0.0123, 0.0319, + -0.0104, 0.0121], device='cuda:0'), grad: tensor([ 0.0176, -0.0112, -0.0623, -0.0030, 0.0310, -0.0106, 0.0069, -0.0233, + 0.0206, 0.0344], device='cuda:0') +100 +0.0001 +changing lr +epoch 207, time 214.73, cls_loss 0.5736 cls_loss_mapping 0.0048 cls_loss_causal 0.5017 re_mapping 0.0119 re_causal 0.0259 /// teacc 98.84 lr 0.00010000 +Epoch 209, weight, value: tensor([[-0.1522, -0.0275, 0.0177, ..., -0.0288, -0.0756, -0.1026], + [-0.0379, -0.0755, 0.0282, ..., 0.0343, -0.0118, -0.0748], + [-0.0478, -0.0659, -0.0862, ..., 0.0470, -0.0404, -0.0836], + ..., + [-0.0516, 0.0293, 0.0221, ..., 0.0334, -0.0459, -0.0863], + [-0.0670, -0.0047, 0.0188, ..., 0.0591, -0.0356, -0.1382], + [ 0.0450, 0.0487, -0.0429, ..., -0.1020, -0.0234, 0.0608]], + device='cuda:0'), grad: tensor([[ 3.7421e-06, 8.5640e-04, 2.0452e-06, ..., -2.3975e-03, + 2.1210e-03, -9.4223e-04], + [ 3.8557e-07, 2.9635e-04, 2.5518e-07, ..., -2.5444e-03, + 1.6928e-03, 5.5599e-04], + [ 8.1584e-07, 2.7523e-03, 1.2163e-06, ..., -3.2020e-04, + 3.3531e-03, 2.3727e-03], + ..., + [ 1.3471e-04, 4.6873e-04, 2.8163e-05, ..., 2.9106e-03, + 3.0823e-03, 9.8896e-04], + [ 3.9756e-05, 4.9400e-04, 1.4022e-05, ..., 1.6108e-03, + 2.3594e-03, 5.7268e-04], + [-1.9419e-04, 3.4523e-04, -4.4525e-05, ..., -5.3704e-05, + -8.5068e-03, -1.1368e-03]], device='cuda:0') +Epoch 209, bias, value: tensor([-0.0469, 0.0110, 0.0038, -0.0067, -0.0065, 0.0036, -0.0111, 0.0321, + -0.0098, 0.0121], device='cuda:0'), grad: tensor([-0.0164, -0.0130, 0.0056, -0.0074, 0.0192, -0.0144, -0.0088, 0.0256, + 0.0174, -0.0078], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 208---------------------------------------------------- +epoch 208, time 215.63, cls_loss 0.5649 cls_loss_mapping 0.0048 cls_loss_causal 0.4995 re_mapping 0.0118 re_causal 0.0262 /// teacc 98.91 lr 0.00010000 +Epoch 210, weight, value: tensor([[-0.1530, -0.0276, 0.0187, ..., -0.0286, -0.0749, -0.1029], + [-0.0387, -0.0754, 0.0283, ..., 0.0339, -0.0115, -0.0748], + [-0.0468, -0.0657, -0.0867, ..., 0.0468, -0.0402, -0.0831], + ..., + [-0.0527, 0.0287, 0.0236, ..., 0.0337, -0.0467, -0.0875], + [-0.0668, -0.0053, 0.0183, ..., 0.0580, -0.0372, -0.1390], + [ 0.0465, 0.0490, -0.0427, ..., -0.1018, -0.0232, 0.0620]], + device='cuda:0'), grad: tensor([[-3.8109e-03, -4.6158e-03, -9.4604e-04, ..., -2.7447e-03, + -1.9493e-03, -1.1543e-02], + [ 1.9741e-04, 2.7885e-03, 4.1008e-04, ..., 5.8670e-03, + 1.4105e-03, 3.0088e-04], + [ 2.3222e-04, 7.8022e-05, 1.2541e-04, ..., 3.0212e-03, + 7.8011e-04, 4.3488e-04], + ..., + [ 8.6641e-04, -3.0003e-03, 4.8280e-04, ..., -4.5547e-03, + 1.0576e-03, -9.5081e-04], + [-1.0929e-03, 2.8496e-03, -2.4757e-03, ..., -1.6375e-03, + 1.0576e-03, 1.2817e-03], + [-8.9931e-04, 1.8263e-03, 6.1846e-04, ..., -1.1482e-03, + 1.0548e-03, 7.3528e-04]], device='cuda:0') +Epoch 210, bias, value: tensor([-0.0461, 0.0117, 0.0042, -0.0065, -0.0059, 0.0035, -0.0113, 0.0317, + -0.0115, 0.0119], device='cuda:0'), grad: tensor([-0.0728, 0.0577, 0.0212, 0.0174, 0.0098, 0.0090, -0.0292, -0.0114, + -0.0085, 0.0067], device='cuda:0') +100 +0.0001 +changing lr +epoch 209, time 214.71, cls_loss 0.5627 cls_loss_mapping 0.0078 cls_loss_causal 0.4899 re_mapping 0.0118 re_causal 0.0259 /// teacc 98.64 lr 0.00010000 +Epoch 211, weight, value: tensor([[-0.1530, -0.0279, 0.0185, ..., -0.0295, -0.0751, -0.1033], + [-0.0378, -0.0755, 0.0287, ..., 0.0343, -0.0118, -0.0753], + [-0.0453, -0.0667, -0.0865, ..., 0.0462, -0.0403, -0.0835], + ..., + [-0.0524, 0.0282, 0.0237, ..., 0.0333, -0.0469, -0.0869], + [-0.0669, -0.0053, 0.0186, ..., 0.0597, -0.0376, -0.1392], + [ 0.0460, 0.0496, -0.0428, ..., -0.1016, -0.0235, 0.0633]], + device='cuda:0'), grad: tensor([[ 6.4135e-04, 8.4788e-06, 1.5745e-03, ..., 1.2474e-03, + 1.4114e-03, 5.9128e-04], + [ 8.9109e-05, 1.6499e-04, 3.6669e-04, ..., 2.9831e-03, + 1.5163e-03, 1.1034e-03], + [ 4.3416e-04, 6.5446e-05, 4.0741e-03, ..., -2.0199e-03, + 7.5817e-04, -8.4019e-04], + ..., + [ 1.0973e-04, -6.6233e-04, 2.2101e-04, ..., -2.8934e-03, + 1.2426e-03, -2.4567e-03], + [ 1.5583e-03, 2.1636e-05, 3.2883e-03, ..., -4.1938e-04, + 1.3485e-03, 1.2665e-03], + [-1.7891e-03, -4.0472e-05, -8.2321e-03, ..., -2.7199e-03, + 1.1358e-03, 4.1556e-04]], device='cuda:0') +Epoch 211, bias, value: tensor([-0.0462, 0.0121, 0.0043, -0.0066, -0.0056, 0.0035, -0.0120, 0.0308, + -0.0110, 0.0124], device='cuda:0'), grad: tensor([ 0.0231, 0.0397, -0.0017, 0.0020, -0.0164, 0.0374, -0.0163, -0.0128, + -0.0292, -0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 210, time 214.56, cls_loss 0.5725 cls_loss_mapping 0.0045 cls_loss_causal 0.4975 re_mapping 0.0124 re_causal 0.0277 /// teacc 98.68 lr 0.00010000 +Epoch 212, weight, value: tensor([[-0.1542, -0.0280, 0.0192, ..., -0.0305, -0.0751, -0.1045], + [-0.0380, -0.0754, 0.0280, ..., 0.0346, -0.0121, -0.0761], + [-0.0465, -0.0667, -0.0876, ..., 0.0462, -0.0406, -0.0833], + ..., + [-0.0538, 0.0286, 0.0245, ..., 0.0334, -0.0465, -0.0881], + [-0.0671, -0.0051, 0.0188, ..., 0.0593, -0.0375, -0.1387], + [ 0.0462, 0.0497, -0.0427, ..., -0.1015, -0.0234, 0.0624]], + device='cuda:0'), grad: tensor([[ 1.2743e-04, 1.7250e-04, 8.1658e-05, ..., 1.6689e-03, + 4.2677e-04, 1.6737e-03], + [ 1.3137e-04, 2.7676e-03, 5.9605e-05, ..., 3.0994e-03, + 6.1369e-04, 6.0892e-04], + [ 2.7609e-04, -6.5384e-03, 2.0552e-04, ..., -1.1604e-02, + 6.0368e-04, -2.5692e-03], + ..., + [ 1.0937e-04, 3.3593e-04, 2.8044e-05, ..., -1.0180e-04, + 4.1413e-04, -6.6233e-04], + [-3.1395e-03, 6.6948e-04, 7.5483e-04, ..., -3.7556e-03, + -3.8052e-03, 3.4904e-03], + [ 1.4782e-04, 5.0813e-06, 5.6088e-05, ..., -8.1444e-04, + 2.6965e-04, 1.1730e-03]], device='cuda:0') +Epoch 212, bias, value: tensor([-0.0473, 0.0113, 0.0049, -0.0072, -0.0054, 0.0041, -0.0123, 0.0311, + -0.0106, 0.0129], device='cuda:0'), grad: tensor([ 0.0176, 0.0290, -0.0777, 0.0373, 0.0200, 0.0101, -0.0034, 0.0073, + -0.0249, -0.0154], device='cuda:0') +100 +0.0001 +changing lr +epoch 211, time 215.02, cls_loss 0.6072 cls_loss_mapping 0.0053 cls_loss_causal 0.5324 re_mapping 0.0115 re_causal 0.0264 /// teacc 98.69 lr 0.00010000 +Epoch 213, weight, value: tensor([[-0.1571, -0.0283, 0.0195, ..., -0.0295, -0.0742, -0.1050], + [-0.0371, -0.0752, 0.0283, ..., 0.0347, -0.0122, -0.0766], + [-0.0465, -0.0671, -0.0878, ..., 0.0458, -0.0413, -0.0835], + ..., + [-0.0526, 0.0284, 0.0241, ..., 0.0330, -0.0463, -0.0874], + [-0.0656, -0.0055, 0.0189, ..., 0.0605, -0.0387, -0.1388], + [ 0.0446, 0.0502, -0.0433, ..., -0.1010, -0.0237, 0.0613]], + device='cuda:0'), grad: tensor([[ 3.4881e-04, -2.5146e-07, 1.2279e-04, ..., -2.2392e-03, + 1.9817e-03, 1.0859e-06], + [ 8.4937e-05, 2.2352e-08, 1.4555e-04, ..., 5.3368e-03, + 2.7828e-03, 1.6298e-06], + [ 2.5892e-04, 1.2852e-06, 6.2943e-05, ..., -2.5597e-03, + -9.7036e-04, 7.9907e-07], + ..., + [ 2.4378e-04, 8.9169e-05, 9.6023e-05, ..., -1.8911e-03, + 1.4029e-03, 1.6868e-05], + [ 3.8052e-04, -3.2883e-03, -7.0143e-04, ..., -2.7943e-03, + -1.9484e-03, -8.0681e-04], + [-3.9673e-04, 1.3294e-03, -5.7268e-04, ..., -1.9569e-03, + 2.3384e-03, 2.1052e-04]], device='cuda:0') +Epoch 213, bias, value: tensor([-0.0472, 0.0123, 0.0042, -0.0076, -0.0055, 0.0046, -0.0131, 0.0316, + -0.0107, 0.0128], device='cuda:0'), grad: tensor([-0.0108, 0.0341, -0.0109, 0.0234, -0.0089, -0.0014, 0.0168, -0.0119, + -0.0167, -0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 212, time 214.84, cls_loss 0.5648 cls_loss_mapping 0.0069 cls_loss_causal 0.4905 re_mapping 0.0123 re_causal 0.0258 /// teacc 98.72 lr 0.00010000 +Epoch 214, weight, value: tensor([[-0.1572, -0.0273, 0.0186, ..., -0.0307, -0.0736, -0.1054], + [-0.0370, -0.0758, 0.0287, ..., 0.0354, -0.0110, -0.0774], + [-0.0462, -0.0668, -0.0861, ..., 0.0458, -0.0418, -0.0839], + ..., + [-0.0527, 0.0286, 0.0234, ..., 0.0341, -0.0463, -0.0880], + [-0.0669, -0.0046, 0.0185, ..., 0.0614, -0.0384, -0.1394], + [ 0.0446, 0.0508, -0.0438, ..., -0.1024, -0.0246, 0.0610]], + device='cuda:0'), grad: tensor([[ 6.0558e-05, 1.3518e-04, -4.1275e-03, ..., -3.0518e-03, + 2.1553e-04, 2.1234e-05], + [ 7.4357e-06, 8.7976e-04, 1.7080e-03, ..., 6.9542e-03, + 1.7920e-03, 3.4682e-06], + [ 5.4747e-05, -5.3167e-04, 6.3848e-04, ..., -6.2084e-04, + 3.6645e-04, 2.7984e-05], + ..., + [-1.1665e-04, -1.0124e-02, 1.8072e-03, ..., -2.4811e-02, + -4.5319e-02, 5.9515e-05], + [-1.8492e-03, -6.9962e-03, 8.5640e-04, ..., 1.8930e-03, + 4.6396e-04, 1.8585e-04], + [ 9.1362e-04, 7.2098e-03, -2.0236e-05, ..., 2.5024e-03, + 1.5187e-04, -8.1825e-04]], device='cuda:0') +Epoch 214, bias, value: tensor([-0.0471, 0.0126, 0.0042, -0.0076, -0.0064, 0.0044, -0.0125, 0.0325, + -0.0112, 0.0128], device='cuda:0'), grad: tensor([-0.0207, 0.0219, 0.0117, 0.0318, 0.0145, -0.0482, 0.0092, -0.0391, + -0.0169, 0.0357], device='cuda:0') +100 +0.0001 +changing lr +epoch 213, time 214.57, cls_loss 0.5691 cls_loss_mapping 0.0065 cls_loss_causal 0.4949 re_mapping 0.0120 re_causal 0.0257 /// teacc 98.83 lr 0.00010000 +Epoch 215, weight, value: tensor([[-0.1577, -0.0274, 0.0187, ..., -0.0320, -0.0727, -0.1051], + [-0.0360, -0.0756, 0.0275, ..., 0.0365, -0.0108, -0.0767], + [-0.0470, -0.0663, -0.0853, ..., 0.0466, -0.0414, -0.0828], + ..., + [-0.0534, 0.0298, 0.0227, ..., 0.0332, -0.0462, -0.0881], + [-0.0664, -0.0049, 0.0205, ..., 0.0612, -0.0381, -0.1393], + [ 0.0448, 0.0505, -0.0440, ..., -0.1023, -0.0253, 0.0615]], + device='cuda:0'), grad: tensor([[ 1.3018e-04, 3.7290e-06, 2.2995e-04, ..., 1.1330e-03, + 7.8082e-05, 1.3048e-06], + [ 1.6212e-04, 3.2902e-05, 4.2439e-04, ..., -2.9278e-03, + 9.4771e-05, 9.5665e-06], + [ 6.2799e-04, 7.3910e-06, 4.9925e-04, ..., 1.8902e-03, + -2.4819e-04, 8.8885e-06], + ..., + [ 3.2961e-05, -1.9759e-05, -3.4027e-03, ..., -1.8215e-03, + -2.9755e-04, 1.8165e-05], + [-3.0446e-04, 1.8090e-05, 3.5048e-04, ..., 1.5392e-03, + 7.2956e-05, 1.5527e-05], + [-5.1521e-06, 3.1143e-05, 5.7602e-04, ..., 1.3704e-03, + 1.3149e-04, -3.5554e-05]], device='cuda:0') +Epoch 215, bias, value: tensor([-0.0482, 0.0120, 0.0054, -0.0066, -0.0069, 0.0051, -0.0127, 0.0312, + -0.0103, 0.0126], device='cuda:0'), grad: tensor([ 0.0099, -0.0166, 0.0152, 0.0151, 0.0087, 0.0126, -0.0350, -0.0332, + 0.0103, 0.0132], device='cuda:0') +100 +0.0001 +changing lr +epoch 214, time 214.81, cls_loss 0.5720 cls_loss_mapping 0.0055 cls_loss_causal 0.4932 re_mapping 0.0116 re_causal 0.0252 /// teacc 98.80 lr 0.00010000 +Epoch 216, weight, value: tensor([[-0.1582, -0.0279, 0.0189, ..., -0.0330, -0.0713, -0.1052], + [-0.0362, -0.0758, 0.0259, ..., 0.0362, -0.0108, -0.0772], + [-0.0474, -0.0656, -0.0863, ..., 0.0460, -0.0417, -0.0835], + ..., + [-0.0536, 0.0291, 0.0235, ..., 0.0341, -0.0464, -0.0881], + [-0.0673, -0.0050, 0.0193, ..., 0.0621, -0.0387, -0.1388], + [ 0.0450, 0.0515, -0.0437, ..., -0.1021, -0.0247, 0.0624]], + device='cuda:0'), grad: tensor([[ 1.2508e-06, 1.2708e-04, 2.6846e-04, ..., 2.9964e-03, + 1.2433e-06, 2.4402e-04], + [ 1.2582e-06, 3.7909e-05, 5.5885e-04, ..., -7.6904e-03, + -1.5184e-05, 1.2577e-04], + [ 3.1680e-05, -1.0216e-04, -1.3838e-03, ..., 2.5578e-03, + -5.6997e-06, 1.1415e-03], + ..., + [ 8.9183e-06, 1.6367e-04, 3.4404e-04, ..., 4.1885e-03, + 1.3001e-06, 4.5180e-04], + [ 1.8731e-05, 6.0511e-04, -1.6937e-03, ..., -5.2547e-04, + 4.1984e-06, -3.1071e-03], + [ 4.4405e-05, 7.5769e-04, 1.9681e-04, ..., 5.0621e-03, + 3.3043e-06, 6.2370e-04]], device='cuda:0') +Epoch 216, bias, value: tensor([-0.0482, 0.0123, 0.0051, -0.0074, -0.0055, 0.0047, -0.0136, 0.0321, + -0.0109, 0.0129], device='cuda:0'), grad: tensor([ 0.0251, -0.0572, 0.0139, 0.0410, -0.0163, -0.0996, 0.0209, 0.0251, + 0.0054, 0.0417], device='cuda:0') +100 +0.0001 +changing lr +epoch 215, time 215.10, cls_loss 0.5409 cls_loss_mapping 0.0056 cls_loss_causal 0.4646 re_mapping 0.0115 re_causal 0.0264 /// teacc 98.80 lr 0.00010000 +Epoch 217, weight, value: tensor([[-0.1579, -0.0283, 0.0176, ..., -0.0316, -0.0718, -0.1054], + [-0.0371, -0.0760, 0.0271, ..., 0.0351, -0.0095, -0.0779], + [-0.0465, -0.0658, -0.0865, ..., 0.0456, -0.0411, -0.0824], + ..., + [-0.0540, 0.0296, 0.0236, ..., 0.0335, -0.0463, -0.0880], + [-0.0673, -0.0056, 0.0192, ..., 0.0632, -0.0382, -0.1385], + [ 0.0450, 0.0517, -0.0438, ..., -0.1014, -0.0248, 0.0627]], + device='cuda:0'), grad: tensor([[ 1.5652e-04, 1.0673e-06, 1.0033e-03, ..., -3.8834e-03, + 7.2050e-04, 2.3711e-04], + [ 3.6836e-04, 1.1124e-05, -3.4847e-03, ..., -1.0735e-02, + -1.6108e-03, -1.3256e-03], + [ 2.3544e-04, 1.4082e-05, -1.8301e-03, ..., -1.1539e-04, + -1.2627e-03, 3.4070e-04], + ..., + [ 3.1796e-03, -2.5702e-04, 2.4738e-03, ..., 5.8556e-03, + 6.6566e-04, 4.1580e-03], + [ 1.2789e-03, 4.7296e-05, 1.7166e-03, ..., 3.6697e-03, + -2.7580e-03, 1.3008e-03], + [-8.5144e-03, 1.0508e-04, -9.7351e-03, ..., -2.1133e-03, + 8.8692e-04, -8.8577e-03]], device='cuda:0') +Epoch 217, bias, value: tensor([-0.0486, 0.0132, 0.0048, -0.0077, -0.0050, 0.0053, -0.0139, 0.0320, + -0.0108, 0.0123], device='cuda:0'), grad: tensor([-0.0152, -0.0668, -0.0041, 0.0439, 0.0133, 0.0382, -0.0463, 0.0378, + 0.0040, -0.0049], device='cuda:0') +100 +0.0001 +changing lr +epoch 216, time 214.66, cls_loss 0.5667 cls_loss_mapping 0.0054 cls_loss_causal 0.4845 re_mapping 0.0114 re_causal 0.0255 /// teacc 98.75 lr 0.00010000 +Epoch 218, weight, value: tensor([[-0.1579, -0.0267, 0.0181, ..., -0.0312, -0.0717, -0.1059], + [-0.0375, -0.0758, 0.0267, ..., 0.0356, -0.0089, -0.0779], + [-0.0460, -0.0650, -0.0874, ..., 0.0435, -0.0435, -0.0830], + ..., + [-0.0542, 0.0288, 0.0237, ..., 0.0336, -0.0455, -0.0887], + [-0.0671, -0.0060, 0.0207, ..., 0.0637, -0.0377, -0.1387], + [ 0.0447, 0.0517, -0.0438, ..., -0.1008, -0.0251, 0.0632]], + device='cuda:0'), grad: tensor([[ 2.3842e-06, 4.0978e-08, 2.3341e-04, ..., 2.7905e-03, + 1.0735e-04, 4.5747e-05], + [ 1.0729e-06, 4.6566e-09, 2.6627e-03, ..., 6.4507e-03, + 4.2772e-04, 2.7847e-04], + [ 1.2094e-04, 9.5926e-08, 7.9060e-04, ..., -1.8654e-03, + 2.4068e-04, 2.7919e-04], + ..., + [ 1.5132e-05, -6.2119e-07, 1.0595e-03, ..., 3.9406e-03, + 6.6280e-05, 3.6925e-05], + [ 5.6381e-03, 8.8103e-07, 1.5354e-03, ..., -2.5806e-03, + -3.4761e-04, 6.8321e-03], + [-2.7165e-05, -2.1979e-06, -2.1591e-03, ..., -1.0357e-03, + 5.8174e-05, -3.0816e-05]], device='cuda:0') +Epoch 218, bias, value: tensor([-0.0476, 0.0125, 0.0035, -0.0076, -0.0037, 0.0047, -0.0146, 0.0320, + -0.0102, 0.0125], device='cuda:0'), grad: tensor([ 0.0167, 0.0313, -0.0080, -0.1056, 0.0165, 0.0182, -0.0125, 0.0215, + 0.0138, 0.0080], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 217---------------------------------------------------- +epoch 217, time 215.39, cls_loss 0.5742 cls_loss_mapping 0.0062 cls_loss_causal 0.4960 re_mapping 0.0120 re_causal 0.0268 /// teacc 98.99 lr 0.00010000 +Epoch 219, weight, value: tensor([[-0.1583, -0.0255, 0.0173, ..., -0.0323, -0.0710, -0.1067], + [-0.0370, -0.0767, 0.0264, ..., 0.0353, -0.0101, -0.0784], + [-0.0456, -0.0642, -0.0879, ..., 0.0445, -0.0426, -0.0817], + ..., + [-0.0550, 0.0285, 0.0242, ..., 0.0335, -0.0449, -0.0897], + [-0.0669, -0.0061, 0.0204, ..., 0.0651, -0.0383, -0.1395], + [ 0.0453, 0.0518, -0.0433, ..., -0.1021, -0.0267, 0.0635]], + device='cuda:0'), grad: tensor([[ 2.2161e-04, 1.5676e-04, 3.0351e-04, ..., 3.5973e-03, + 1.4811e-03, 1.0259e-05], + [ 8.9979e-04, 4.9057e-03, 2.3880e-03, ..., 2.0275e-03, + 3.0003e-03, 7.2867e-06], + [ 4.3654e-04, 1.3037e-03, 7.8106e-04, ..., -4.8409e-03, + -1.5884e-02, 7.5936e-05], + ..., + [-1.6883e-05, -1.4563e-03, 1.6928e-04, ..., 3.1338e-03, + 5.1422e-03, -3.3569e-04], + [ 5.3930e-04, 8.3971e-04, 1.0700e-03, ..., 1.8930e-04, + 3.2063e-03, -5.4836e-06], + [ 2.0099e-04, 3.1471e-04, 2.3961e-04, ..., 4.0016e-03, + 1.8921e-03, 3.0175e-05]], device='cuda:0') +Epoch 219, bias, value: tensor([-0.0473, 0.0123, 0.0039, -0.0077, -0.0037, 0.0043, -0.0146, 0.0321, + -0.0096, 0.0116], device='cuda:0'), grad: tensor([ 0.0185, -0.0016, -0.0346, 0.0227, -0.0138, -0.0354, -0.0078, 0.0219, + 0.0105, 0.0196], device='cuda:0') +100 +0.0001 +changing lr +epoch 218, time 214.74, cls_loss 0.5787 cls_loss_mapping 0.0046 cls_loss_causal 0.5089 re_mapping 0.0121 re_causal 0.0261 /// teacc 98.79 lr 0.00010000 +Epoch 220, weight, value: tensor([[-0.1592, -0.0272, 0.0192, ..., -0.0328, -0.0712, -0.1071], + [-0.0366, -0.0763, 0.0249, ..., 0.0360, -0.0101, -0.0793], + [-0.0455, -0.0627, -0.0878, ..., 0.0448, -0.0424, -0.0805], + ..., + [-0.0544, 0.0278, 0.0240, ..., 0.0343, -0.0438, -0.0889], + [-0.0679, -0.0057, 0.0210, ..., 0.0631, -0.0384, -0.1389], + [ 0.0460, 0.0514, -0.0440, ..., -0.1028, -0.0262, 0.0628]], + device='cuda:0'), grad: tensor([[ 3.0547e-07, 4.5563e-02, 1.5348e-06, ..., 2.7580e-03, + 1.3895e-06, 4.5262e-07], + [ 3.4086e-07, 1.1347e-05, 5.0552e-06, ..., 2.9125e-03, + 8.9630e-06, 1.8626e-07], + [ 2.1517e-05, -2.7823e-04, 5.1767e-05, ..., -1.6373e-02, + 7.6771e-05, 1.2532e-05], + ..., + [ 1.7453e-06, 4.7237e-06, 1.2286e-05, ..., 1.4830e-03, + 8.0988e-06, 2.5518e-07], + [-1.4067e-05, 2.5082e-04, -6.2764e-05, ..., 3.1242e-03, + -1.4770e-04, 1.2904e-05], + [-1.6868e-05, -4.5746e-02, 1.3316e-04, ..., 1.5640e-03, + 1.1690e-05, -8.7917e-06]], device='cuda:0') +Epoch 220, bias, value: tensor([-0.0479, 0.0121, 0.0048, -0.0081, -0.0027, 0.0036, -0.0139, 0.0325, + -0.0096, 0.0106], device='cuda:0'), grad: tensor([ 0.0382, 0.0175, -0.0778, 0.0083, 0.0268, -0.0191, 0.0527, 0.0116, + -0.0459, -0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 219, time 214.40, cls_loss 0.5739 cls_loss_mapping 0.0046 cls_loss_causal 0.5030 re_mapping 0.0123 re_causal 0.0282 /// teacc 98.84 lr 0.00010000 +Epoch 221, weight, value: tensor([[-0.1600, -0.0284, 0.0197, ..., -0.0330, -0.0715, -0.1071], + [-0.0376, -0.0760, 0.0257, ..., 0.0367, -0.0105, -0.0792], + [-0.0464, -0.0633, -0.0885, ..., 0.0449, -0.0439, -0.0815], + ..., + [-0.0546, 0.0282, 0.0228, ..., 0.0339, -0.0442, -0.0889], + [-0.0682, -0.0063, 0.0210, ..., 0.0619, -0.0386, -0.1390], + [ 0.0459, 0.0526, -0.0441, ..., -0.1021, -0.0260, 0.0629]], + device='cuda:0'), grad: tensor([[-6.6805e-04, -2.3518e-03, -1.4992e-03, ..., 1.8492e-03, + -4.7183e-04, -8.3351e-04], + [ 1.1444e-05, 3.0613e-04, 3.0547e-05, ..., 9.1219e-04, + -1.5306e-03, 1.4901e-05], + [-1.3441e-05, 2.5201e-04, 3.3140e-05, ..., -2.8591e-03, + 5.3263e-04, 3.1739e-05], + ..., + [ 1.2070e-05, 1.2600e-04, 2.8253e-05, ..., 2.4109e-03, + 4.3273e-04, 9.7305e-06], + [ 1.4186e-04, 3.6740e-04, 2.5439e-04, ..., -2.0103e-03, + 3.4165e-04, 1.7893e-04], + [ 1.8203e-04, 7.3624e-04, 3.7956e-04, ..., 1.9817e-03, + 1.2856e-03, 2.3282e-04]], device='cuda:0') +Epoch 221, bias, value: tensor([-0.0484, 0.0127, 0.0043, -0.0082, -0.0031, 0.0046, -0.0143, 0.0320, + -0.0101, 0.0120], device='cuda:0'), grad: tensor([ 0.0147, -0.0272, -0.0416, 0.0306, 0.0281, -0.0663, 0.0138, 0.0277, + -0.0085, 0.0288], device='cuda:0') +100 +0.0001 +changing lr +epoch 220, time 214.46, cls_loss 0.5942 cls_loss_mapping 0.0055 cls_loss_causal 0.5188 re_mapping 0.0117 re_causal 0.0267 /// teacc 98.74 lr 0.00010000 +Epoch 222, weight, value: tensor([[-0.1603, -0.0270, 0.0193, ..., -0.0332, -0.0711, -0.1071], + [-0.0375, -0.0773, 0.0260, ..., 0.0371, -0.0098, -0.0796], + [-0.0472, -0.0633, -0.0882, ..., 0.0451, -0.0428, -0.0825], + ..., + [-0.0554, 0.0280, 0.0237, ..., 0.0337, -0.0444, -0.0888], + [-0.0693, -0.0076, 0.0204, ..., 0.0622, -0.0400, -0.1408], + [ 0.0462, 0.0514, -0.0447, ..., -0.1023, -0.0263, 0.0632]], + device='cuda:0'), grad: tensor([[ 3.8314e-04, 4.1866e-04, 5.3197e-06, ..., 2.7409e-03, + 4.1771e-04, 9.3222e-04], + [ 1.1548e-05, -4.5853e-03, -6.2585e-05, ..., 6.1655e-04, + 1.9236e-03, 1.8924e-05], + [ 4.4286e-05, 1.7710e-03, 1.4722e-05, ..., 6.1188e-03, + 2.1744e-03, 9.1016e-05], + ..., + [ 3.7342e-05, 8.5678e-03, 1.0490e-04, ..., 2.4475e-02, + 7.4005e-03, 3.7044e-05], + [ 1.1373e-04, 1.3390e-03, 9.1866e-06, ..., -3.1708e-02, + -4.7188e-03, 3.9268e-04], + [ 5.6118e-05, 9.5272e-04, 4.3726e-04, ..., 4.0016e-03, + 9.4652e-04, -1.0312e-04]], device='cuda:0') +Epoch 222, bias, value: tensor([-0.0479, 0.0136, 0.0050, -0.0079, -0.0035, 0.0048, -0.0158, 0.0321, + -0.0102, 0.0113], device='cuda:0'), grad: tensor([ 0.0157, -0.0122, 0.0228, -0.0155, -0.0066, -0.0168, 0.0031, 0.0588, + -0.0661, 0.0168], device='cuda:0') +100 +0.0001 +changing lr +epoch 221, time 214.84, cls_loss 0.5708 cls_loss_mapping 0.0070 cls_loss_causal 0.4925 re_mapping 0.0118 re_causal 0.0255 /// teacc 98.85 lr 0.00010000 +Epoch 223, weight, value: tensor([[-0.1610, -0.0271, 0.0197, ..., -0.0324, -0.0714, -0.1085], + [-0.0379, -0.0751, 0.0261, ..., 0.0373, -0.0092, -0.0809], + [-0.0464, -0.0640, -0.0905, ..., 0.0438, -0.0426, -0.0819], + ..., + [-0.0565, 0.0275, 0.0237, ..., 0.0333, -0.0441, -0.0897], + [-0.0695, -0.0077, 0.0214, ..., 0.0631, -0.0392, -0.1410], + [ 0.0461, 0.0519, -0.0457, ..., -0.1027, -0.0269, 0.0631]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0009, 0.0013, ..., -0.0008, 0.0005, 0.0007], + [ 0.0005, -0.0013, -0.0006, ..., -0.0031, 0.0001, 0.0007], + [ 0.0003, -0.0055, 0.0018, ..., 0.0018, -0.0038, 0.0009], + ..., + [ 0.0008, -0.0052, 0.0021, ..., 0.0033, 0.0041, 0.0010], + [ 0.0022, 0.0059, -0.0020, ..., 0.0003, 0.0038, 0.0009], + [ 0.0085, 0.0068, 0.0003, ..., 0.0015, -0.0046, 0.0119]], + device='cuda:0') +Epoch 223, bias, value: tensor([-0.0482, 0.0145, 0.0040, -0.0073, -0.0040, 0.0047, -0.0157, 0.0322, + -0.0103, 0.0115], device='cuda:0'), grad: tensor([-0.0091, -0.0157, 0.0024, -0.0840, 0.0337, 0.0230, -0.0001, 0.0348, + 0.0297, -0.0146], device='cuda:0') +100 +0.0001 +changing lr +epoch 222, time 214.50, cls_loss 0.5731 cls_loss_mapping 0.0043 cls_loss_causal 0.4990 re_mapping 0.0111 re_causal 0.0251 /// teacc 98.81 lr 0.00010000 +Epoch 224, weight, value: tensor([[-0.1616, -0.0279, 0.0191, ..., -0.0328, -0.0722, -0.1090], + [-0.0385, -0.0752, 0.0253, ..., 0.0374, -0.0094, -0.0819], + [-0.0458, -0.0640, -0.0908, ..., 0.0445, -0.0432, -0.0819], + ..., + [-0.0562, 0.0280, 0.0233, ..., 0.0332, -0.0437, -0.0893], + [-0.0683, -0.0076, 0.0217, ..., 0.0634, -0.0402, -0.1411], + [ 0.0467, 0.0507, -0.0446, ..., -0.1030, -0.0262, 0.0625]], + device='cuda:0'), grad: tensor([[ 6.1691e-06, -2.6894e-04, 8.8263e-04, ..., 2.6779e-03, + 2.6751e-04, -9.7573e-05], + [ 2.3562e-06, 1.8597e-04, -3.4370e-03, ..., -8.2474e-03, + -1.6937e-03, 4.7803e-05], + [-2.3651e-04, 1.6081e-04, 8.1444e-04, ..., 2.2259e-03, + 2.5630e-04, 8.6844e-05], + ..., + [-7.5817e-05, -1.6336e-03, 5.0545e-04, ..., -4.6921e-03, + 1.5485e-04, -6.2466e-04], + [ 3.8981e-05, 1.8418e-04, 4.9543e-04, ..., 1.7185e-03, + 1.2130e-04, 1.0169e-04], + [ 7.5817e-05, 9.7752e-04, 6.6090e-04, ..., 2.7752e-03, + 2.0945e-04, 1.1563e-04]], device='cuda:0') +Epoch 224, bias, value: tensor([-0.0486, 0.0142, 0.0042, -0.0061, -0.0034, 0.0043, -0.0161, 0.0323, + -0.0114, 0.0121], device='cuda:0'), grad: tensor([ 0.0270, -0.0971, 0.0217, 0.0267, 0.0279, 0.0164, -0.0330, -0.0320, + 0.0174, 0.0250], device='cuda:0') +100 +0.0001 +changing lr +epoch 223, time 214.43, cls_loss 0.5577 cls_loss_mapping 0.0074 cls_loss_causal 0.4960 re_mapping 0.0113 re_causal 0.0258 /// teacc 98.70 lr 0.00010000 +Epoch 225, weight, value: tensor([[-0.1612, -0.0269, 0.0190, ..., -0.0326, -0.0708, -0.1087], + [-0.0380, -0.0755, 0.0276, ..., 0.0374, -0.0097, -0.0807], + [-0.0466, -0.0640, -0.0899, ..., 0.0453, -0.0434, -0.0820], + ..., + [-0.0547, 0.0279, 0.0219, ..., 0.0333, -0.0438, -0.0886], + [-0.0682, -0.0076, 0.0209, ..., 0.0636, -0.0400, -0.1412], + [ 0.0461, 0.0505, -0.0448, ..., -0.1035, -0.0257, 0.0617]], + device='cuda:0'), grad: tensor([[ 1.9054e-03, 4.4554e-06, 7.3004e-04, ..., 1.8654e-03, + 2.7958e-06, 1.9464e-03], + [ 9.3400e-05, 1.9951e-03, 6.8521e-04, ..., -1.2884e-03, + 2.2911e-07, -3.4833e-04], + [ 6.3717e-05, 4.6007e-06, 2.4402e-04, ..., 1.3990e-03, + -4.1366e-04, 1.7309e-04], + ..., + [ 1.9777e-04, 4.8615e-07, 4.2510e-04, ..., -5.3139e-03, + 2.1607e-06, 2.6512e-04], + [ 7.4768e-04, 1.1438e-04, 3.4952e-04, ..., 1.7176e-03, + 3.8713e-05, 8.4639e-04], + [ 2.6155e-04, 1.1455e-06, 5.0306e-04, ..., 1.9989e-03, + 7.2531e-06, 3.8362e-04]], device='cuda:0') +Epoch 225, bias, value: tensor([-0.0489, 0.0146, 0.0047, -0.0066, -0.0043, 0.0036, -0.0151, 0.0331, + -0.0110, 0.0112], device='cuda:0'), grad: tensor([ 0.0231, -0.0069, 0.0106, -0.0087, -0.0436, -0.0177, 0.0141, -0.0066, + 0.0165, 0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 224, time 214.78, cls_loss 0.5613 cls_loss_mapping 0.0048 cls_loss_causal 0.4904 re_mapping 0.0113 re_causal 0.0255 /// teacc 98.66 lr 0.00010000 +Epoch 226, weight, value: tensor([[-0.1605, -0.0274, 0.0189, ..., -0.0326, -0.0717, -0.1105], + [-0.0394, -0.0762, 0.0269, ..., 0.0363, -0.0091, -0.0796], + [-0.0477, -0.0641, -0.0902, ..., 0.0459, -0.0439, -0.0836], + ..., + [-0.0541, 0.0289, 0.0225, ..., 0.0339, -0.0438, -0.0889], + [-0.0682, -0.0079, 0.0210, ..., 0.0636, -0.0406, -0.1423], + [ 0.0464, 0.0514, -0.0454, ..., -0.1045, -0.0260, 0.0618]], + device='cuda:0'), grad: tensor([[ 4.8339e-05, -8.2779e-04, 1.6570e-04, ..., 4.0269e-04, + 2.5105e-04, 8.3685e-05], + [ 2.7493e-06, 2.7418e-05, 4.9543e-04, ..., 1.2388e-03, + 4.6968e-04, 3.5405e-05], + [ 2.7075e-05, 2.8685e-05, 2.5892e-04, ..., 5.8174e-04, + 2.9945e-04, 6.1035e-05], + ..., + [ 7.1144e-04, -6.0648e-05, 6.2704e-04, ..., 1.3084e-03, + 4.1127e-04, 7.9870e-04], + [ 1.2088e-04, 7.4100e-04, 2.5201e-04, ..., 2.8539e-04, + 2.5201e-04, 6.6185e-04], + [ 1.4853e-04, 2.7609e-04, -2.3289e-03, ..., -2.5272e-03, + -2.4757e-03, 5.7030e-04]], device='cuda:0') +Epoch 226, bias, value: tensor([-0.0480, 0.0137, 0.0045, -0.0068, -0.0046, 0.0035, -0.0147, 0.0336, + -0.0107, 0.0108], device='cuda:0'), grad: tensor([ 0.0012, 0.0090, 0.0041, 0.0150, -0.0169, -0.0109, 0.0037, 0.0125, + 0.0054, -0.0231], device='cuda:0') +100 +0.0001 +changing lr +epoch 225, time 214.65, cls_loss 0.5688 cls_loss_mapping 0.0068 cls_loss_causal 0.4970 re_mapping 0.0118 re_causal 0.0258 /// teacc 98.89 lr 0.00010000 +Epoch 227, weight, value: tensor([[-0.1618, -0.0272, 0.0207, ..., -0.0328, -0.0730, -0.1115], + [-0.0403, -0.0762, 0.0266, ..., 0.0364, -0.0091, -0.0810], + [-0.0495, -0.0643, -0.0899, ..., 0.0460, -0.0430, -0.0830], + ..., + [-0.0541, 0.0286, 0.0218, ..., 0.0344, -0.0442, -0.0880], + [-0.0687, -0.0071, 0.0200, ..., 0.0636, -0.0393, -0.1429], + [ 0.0468, 0.0512, -0.0460, ..., -0.1045, -0.0259, 0.0623]], + device='cuda:0'), grad: tensor([[ 4.8280e-04, 3.6024e-06, 8.1444e-04, ..., 9.6750e-04, + 1.3304e-04, 2.0847e-03], + [ 6.9761e-04, 1.2852e-07, 6.6471e-04, ..., 1.0157e-03, + 4.6182e-04, 6.5422e-04], + [ 1.6708e-03, 6.4261e-07, 7.3290e-04, ..., 2.8687e-03, + 8.0347e-04, 6.6042e-04], + ..., + [ 1.4009e-03, 1.8620e-04, 1.3208e-04, ..., -5.4512e-03, + 9.0003e-05, -1.2112e-03], + [-5.3253e-03, 2.8461e-06, -5.5237e-03, ..., -3.0823e-03, + -2.4891e-03, -3.2177e-03], + [-4.7531e-03, -3.4738e-04, -2.0103e-03, ..., 5.5075e-04, + 1.6594e-04, -6.1417e-04]], device='cuda:0') +Epoch 227, bias, value: tensor([-0.0480, 0.0137, 0.0047, -0.0067, -0.0049, 0.0038, -0.0147, 0.0337, + -0.0114, 0.0111], device='cuda:0'), grad: tensor([ 0.0156, 0.0179, 0.0288, 0.0378, 0.0114, 0.0235, -0.0176, -0.0322, + -0.0354, -0.0496], device='cuda:0') +100 +0.0001 +changing lr +epoch 226, time 214.72, cls_loss 0.5822 cls_loss_mapping 0.0050 cls_loss_causal 0.5103 re_mapping 0.0116 re_causal 0.0262 /// teacc 98.82 lr 0.00010000 +Epoch 228, weight, value: tensor([[-0.1621, -0.0272, 0.0201, ..., -0.0329, -0.0734, -0.1121], + [-0.0392, -0.0749, 0.0264, ..., 0.0364, -0.0095, -0.0783], + [-0.0500, -0.0639, -0.0891, ..., 0.0460, -0.0430, -0.0837], + ..., + [-0.0552, 0.0274, 0.0216, ..., 0.0337, -0.0450, -0.0896], + [-0.0688, -0.0074, 0.0205, ..., 0.0631, -0.0384, -0.1437], + [ 0.0481, 0.0516, -0.0465, ..., -0.1043, -0.0255, 0.0633]], + device='cuda:0'), grad: tensor([[ 6.1560e-04, 3.6097e-04, 2.7313e-03, ..., 1.2341e-03, + 5.7888e-04, 8.2731e-04], + [ 2.2364e-04, 2.5558e-04, 6.4659e-03, ..., 3.8223e-03, + 3.4928e-04, 6.3515e-04], + [-9.2566e-05, 6.2132e-04, -9.6741e-03, ..., -5.6801e-03, + 1.6952e-04, 3.7837e-04], + ..., + [-6.6805e-04, 6.8617e-04, -9.9411e-03, ..., -7.3090e-03, + 1.9073e-05, -4.8561e-03], + [ 6.7186e-04, 3.6812e-04, 3.2692e-03, ..., 1.8463e-03, + 1.1997e-03, 1.1930e-03], + [ 9.1600e-04, 1.3971e-03, 9.1858e-03, ..., 2.0485e-03, + 3.6983e-03, 4.4975e-03]], device='cuda:0') +Epoch 228, bias, value: tensor([-0.0479, 0.0137, 0.0049, -0.0070, -0.0035, 0.0035, -0.0147, 0.0328, + -0.0108, 0.0105], device='cuda:0'), grad: tensor([ 0.0130, 0.0237, -0.0176, 0.0214, 0.0010, 0.0204, -0.0567, -0.0668, + 0.0157, 0.0458], device='cuda:0') +100 +0.0001 +changing lr +epoch 227, time 214.53, cls_loss 0.5237 cls_loss_mapping 0.0051 cls_loss_causal 0.4603 re_mapping 0.0115 re_causal 0.0254 /// teacc 98.87 lr 0.00010000 +Epoch 229, weight, value: tensor([[-0.1625, -0.0275, 0.0205, ..., -0.0332, -0.0744, -0.1125], + [-0.0399, -0.0751, 0.0259, ..., 0.0371, -0.0092, -0.0777], + [-0.0501, -0.0628, -0.0892, ..., 0.0450, -0.0412, -0.0840], + ..., + [-0.0556, 0.0270, 0.0211, ..., 0.0336, -0.0460, -0.0902], + [-0.0688, -0.0083, 0.0206, ..., 0.0632, -0.0384, -0.1437], + [ 0.0472, 0.0517, -0.0470, ..., -0.1052, -0.0256, 0.0626]], + device='cuda:0'), grad: tensor([[ 0.0003, 0.0097, 0.0016, ..., 0.0027, 0.0002, 0.0001], + [ 0.0001, 0.0006, 0.0015, ..., 0.0032, 0.0002, 0.0002], + [ 0.0001, 0.0002, -0.0055, ..., -0.0029, 0.0001, 0.0002], + ..., + [-0.0020, -0.0029, 0.0013, ..., -0.0003, 0.0005, -0.0014], + [ 0.0001, 0.0019, 0.0017, ..., 0.0026, 0.0005, 0.0004], + [ 0.0004, -0.0026, -0.0020, ..., 0.0028, 0.0002, 0.0077]], + device='cuda:0') +Epoch 229, bias, value: tensor([-0.0484, 0.0144, 0.0046, -0.0063, -0.0038, 0.0034, -0.0141, 0.0327, + -0.0107, 0.0096], device='cuda:0'), grad: tensor([ 0.0525, 0.0288, -0.0134, 0.0288, -0.0561, -0.0427, 0.0121, -0.0143, + 0.0321, -0.0278], device='cuda:0') +100 +0.0001 +changing lr +epoch 228, time 214.50, cls_loss 0.5332 cls_loss_mapping 0.0047 cls_loss_causal 0.4594 re_mapping 0.0119 re_causal 0.0259 /// teacc 98.86 lr 0.00010000 +Epoch 230, weight, value: tensor([[-0.1640, -0.0272, 0.0202, ..., -0.0328, -0.0748, -0.1132], + [-0.0397, -0.0756, 0.0262, ..., 0.0368, -0.0103, -0.0783], + [-0.0512, -0.0634, -0.0901, ..., 0.0449, -0.0418, -0.0855], + ..., + [-0.0558, 0.0263, 0.0215, ..., 0.0336, -0.0457, -0.0902], + [-0.0691, -0.0084, 0.0193, ..., 0.0630, -0.0394, -0.1440], + [ 0.0487, 0.0531, -0.0472, ..., -0.1054, -0.0239, 0.0627]], + device='cuda:0'), grad: tensor([[ 4.0770e-04, 2.5320e-04, 1.2112e-03, ..., 1.0719e-03, + 1.4753e-03, -3.0470e-04], + [-6.6071e-03, 2.0549e-05, -8.2092e-03, ..., 1.7023e-03, + -3.4885e-03, -1.5297e-03], + [ 3.2282e-04, 1.9979e-04, 9.2268e-04, ..., 1.0481e-03, + 5.1384e-03, 8.2195e-05], + ..., + [ 2.6631e-04, 8.8334e-05, -3.6502e-04, ..., -3.0155e-03, + 5.3501e-04, 1.4389e-04], + [-5.7678e-03, 9.8646e-05, 1.5783e-03, ..., 1.3952e-03, + 1.8969e-03, 3.0208e-04], + [ 9.5987e-04, 4.1275e-03, 2.0542e-03, ..., 1.3084e-03, + 1.4849e-03, 3.4218e-03]], device='cuda:0') +Epoch 230, bias, value: tensor([-0.0476, 0.0143, 0.0041, -0.0066, -0.0047, 0.0042, -0.0142, 0.0332, + -0.0117, 0.0105], device='cuda:0'), grad: tensor([-0.0152, -0.0157, 0.0153, 0.0174, 0.0069, 0.0313, -0.0394, -0.0228, + -0.0037, 0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 229, time 214.47, cls_loss 0.5784 cls_loss_mapping 0.0048 cls_loss_causal 0.5047 re_mapping 0.0116 re_causal 0.0264 /// teacc 98.85 lr 0.00010000 +Epoch 231, weight, value: tensor([[-0.1652, -0.0271, 0.0200, ..., -0.0315, -0.0750, -0.1147], + [-0.0380, -0.0748, 0.0262, ..., 0.0365, -0.0080, -0.0778], + [-0.0516, -0.0646, -0.0898, ..., 0.0447, -0.0419, -0.0858], + ..., + [-0.0566, 0.0270, 0.0217, ..., 0.0341, -0.0465, -0.0894], + [-0.0687, -0.0086, 0.0193, ..., 0.0636, -0.0399, -0.1455], + [ 0.0487, 0.0519, -0.0466, ..., -0.1062, -0.0237, 0.0626]], + device='cuda:0'), grad: tensor([[ 5.6922e-06, 2.6792e-05, 5.0621e-03, ..., 2.0523e-03, + 3.5036e-06, 2.7239e-05], + [ 1.4991e-05, 3.6144e-04, 2.4974e-05, ..., -2.0142e-03, + 1.9856e-06, 2.4796e-04], + [-9.5940e-04, 8.2374e-05, 1.9714e-05, ..., -5.7907e-03, + -1.2522e-03, 1.1420e-04], + ..., + [ 8.7678e-05, 3.4094e-04, 3.6001e-04, ..., 1.8606e-03, + 3.5435e-05, 2.9445e-04], + [ 2.2972e-04, 1.7846e-04, 6.5029e-05, ..., 2.4967e-03, + 2.4378e-05, 5.7220e-04], + [-3.7885e-04, 9.0027e-04, -5.7678e-03, ..., 7.2098e-04, + 2.9411e-06, 8.9288e-05]], device='cuda:0') +Epoch 231, bias, value: tensor([-0.0479, 0.0152, 0.0044, -0.0068, -0.0049, 0.0029, -0.0138, 0.0336, + -0.0115, 0.0101], device='cuda:0'), grad: tensor([ 0.0369, -0.0155, -0.0477, 0.0146, 0.0311, 0.0039, -0.0443, 0.0174, + 0.0213, -0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 230, time 214.70, cls_loss 0.5964 cls_loss_mapping 0.0053 cls_loss_causal 0.5245 re_mapping 0.0114 re_causal 0.0261 /// teacc 98.79 lr 0.00010000 +Epoch 232, weight, value: tensor([[-0.1652, -0.0270, 0.0186, ..., -0.0307, -0.0732, -0.1140], + [-0.0366, -0.0757, 0.0257, ..., 0.0366, -0.0093, -0.0780], + [-0.0510, -0.0654, -0.0906, ..., 0.0446, -0.0412, -0.0863], + ..., + [-0.0565, 0.0280, 0.0216, ..., 0.0330, -0.0457, -0.0885], + [-0.0697, -0.0084, 0.0206, ..., 0.0640, -0.0402, -0.1464], + [ 0.0490, 0.0524, -0.0456, ..., -0.1065, -0.0244, 0.0625]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0002, 0.0005, ..., 0.0008, 0.0005, 0.0002], + [ 0.0030, 0.0007, 0.0037, ..., 0.0015, 0.0011, 0.0016], + [ 0.0005, 0.0010, 0.0008, ..., 0.0023, 0.0005, 0.0003], + ..., + [ 0.0005, -0.0005, -0.0031, ..., -0.0003, 0.0037, -0.0047], + [-0.0042, 0.0002, -0.0048, ..., -0.0008, 0.0009, -0.0014], + [-0.0031, 0.0005, 0.0006, ..., -0.0040, -0.0032, 0.0001]], + device='cuda:0') +Epoch 232, bias, value: tensor([-0.0472, 0.0149, 0.0045, -0.0068, -0.0047, 0.0031, -0.0145, 0.0330, + -0.0119, 0.0109], device='cuda:0'), grad: tensor([-0.0107, 0.0415, 0.0234, -0.0349, 0.0112, 0.0164, 0.0081, -0.0152, + -0.0113, -0.0285], device='cuda:0') +100 +0.0001 +changing lr +epoch 231, time 214.60, cls_loss 0.5673 cls_loss_mapping 0.0050 cls_loss_causal 0.4968 re_mapping 0.0117 re_causal 0.0268 /// teacc 98.69 lr 0.00010000 +Epoch 233, weight, value: tensor([[-0.1665, -0.0270, 0.0174, ..., -0.0316, -0.0730, -0.1145], + [-0.0380, -0.0754, 0.0258, ..., 0.0367, -0.0106, -0.0785], + [-0.0502, -0.0662, -0.0900, ..., 0.0448, -0.0406, -0.0868], + ..., + [-0.0557, 0.0279, 0.0215, ..., 0.0341, -0.0468, -0.0878], + [-0.0703, -0.0083, 0.0206, ..., 0.0645, -0.0408, -0.1454], + [ 0.0498, 0.0521, -0.0453, ..., -0.1076, -0.0216, 0.0624]], + device='cuda:0'), grad: tensor([[ 2.4939e-04, 3.4237e-04, 1.9522e-03, ..., 1.7281e-03, + 2.2411e-03, 1.3268e-04], + [ 1.1283e-04, 1.2612e-04, 1.1759e-03, ..., -2.8839e-03, + 2.6760e-03, 1.1295e-04], + [ 6.5565e-04, 3.7861e-03, 4.6444e-04, ..., 3.8490e-03, + 1.9350e-03, 1.0252e-03], + ..., + [-2.3142e-05, 1.5326e-03, -6.1226e-03, ..., -2.6321e-03, + 1.4563e-03, -3.8862e-04], + [-6.8521e-04, -1.1473e-03, -1.2131e-03, ..., -4.7684e-07, + -5.4817e-03, 2.5225e-04], + [ 8.9931e-04, -5.1537e-03, 1.3762e-03, ..., -6.6452e-03, + -3.5019e-03, 6.7711e-04]], device='cuda:0') +Epoch 233, bias, value: tensor([-0.0484, 0.0139, 0.0044, -0.0063, -0.0046, 0.0032, -0.0140, 0.0342, + -0.0120, 0.0109], device='cuda:0'), grad: tensor([ 0.0251, 0.0084, 0.0341, -0.0294, 0.0334, 0.0461, -0.0402, -0.0191, + -0.0362, -0.0223], device='cuda:0') +100 +0.0001 +changing lr +epoch 232, time 214.82, cls_loss 0.5509 cls_loss_mapping 0.0054 cls_loss_causal 0.4774 re_mapping 0.0113 re_causal 0.0246 /// teacc 98.64 lr 0.00010000 +Epoch 234, weight, value: tensor([[-0.1661, -0.0259, 0.0185, ..., -0.0316, -0.0730, -0.1130], + [-0.0384, -0.0766, 0.0260, ..., 0.0369, -0.0103, -0.0792], + [-0.0517, -0.0664, -0.0915, ..., 0.0448, -0.0429, -0.0872], + ..., + [-0.0563, 0.0282, 0.0212, ..., 0.0346, -0.0467, -0.0897], + [-0.0700, -0.0095, 0.0212, ..., 0.0639, -0.0414, -0.1452], + [ 0.0509, 0.0528, -0.0446, ..., -0.1057, -0.0202, 0.0634]], + device='cuda:0'), grad: tensor([[ 8.6069e-04, -1.8024e-03, 2.3508e-04, ..., 8.3208e-04, + 1.0338e-03, 1.1301e-04], + [ 6.2063e-06, 3.7014e-05, 1.6600e-05, ..., 1.0500e-03, + 1.2815e-04, 5.1409e-06], + [ 6.6900e-04, 1.8880e-05, 8.2314e-05, ..., 9.7513e-04, + -1.1742e-02, 4.5061e-04], + ..., + [ 1.5616e-05, -2.1152e-03, 5.0753e-05, ..., 1.1473e-03, + 3.7155e-03, 2.6321e-04], + [ 8.6665e-05, 1.1498e-04, -1.8597e-04, ..., -2.8286e-03, + 9.3460e-04, 6.3598e-05], + [-3.8862e-05, 1.7481e-03, 1.2207e-04, ..., 6.0654e-04, + 8.2159e-04, -3.2568e-04]], device='cuda:0') +Epoch 234, bias, value: tensor([-0.0481, 0.0140, 0.0034, -0.0064, -0.0052, 0.0036, -0.0149, 0.0345, + -0.0113, 0.0115], device='cuda:0'), grad: tensor([ 0.0057, 0.0116, -0.0058, -0.0168, -0.0196, 0.0094, 0.0078, 0.0159, + -0.0188, 0.0105], device='cuda:0') +100 +0.0001 +changing lr +epoch 233, time 214.95, cls_loss 0.5799 cls_loss_mapping 0.0043 cls_loss_causal 0.5040 re_mapping 0.0119 re_causal 0.0277 /// teacc 98.70 lr 0.00010000 +Epoch 235, weight, value: tensor([[-0.1662, -0.0267, 0.0198, ..., -0.0315, -0.0726, -0.1130], + [-0.0379, -0.0777, 0.0249, ..., 0.0364, -0.0103, -0.0800], + [-0.0526, -0.0655, -0.0918, ..., 0.0460, -0.0426, -0.0864], + ..., + [-0.0572, 0.0285, 0.0216, ..., 0.0338, -0.0470, -0.0897], + [-0.0705, -0.0092, 0.0221, ..., 0.0634, -0.0408, -0.1466], + [ 0.0505, 0.0536, -0.0445, ..., -0.1052, -0.0205, 0.0639]], + device='cuda:0'), grad: tensor([[ 6.6459e-05, 2.8491e-04, 1.5676e-04, ..., 4.2200e-05, + 6.2418e-04, -1.3149e-04], + [-8.7214e-04, -4.0665e-03, 2.5129e-04, ..., -7.2813e-04, + -3.9406e-03, -6.3038e-04], + [ 1.1826e-04, -8.5640e-04, -2.4986e-03, ..., 7.4744e-05, + -3.3131e-03, -4.3660e-05], + ..., + [ 8.2076e-05, 9.4461e-04, 4.2129e-04, ..., 6.7294e-05, + 1.3924e-03, 1.5581e-04], + [-1.5059e-03, 5.8079e-04, -3.3188e-04, ..., 7.2718e-05, + 8.9836e-04, -3.1924e-04], + [ 1.2529e-04, 3.9825e-03, 2.7370e-04, ..., 5.9277e-05, + 1.2703e-03, 9.4986e-04]], device='cuda:0') +Epoch 235, bias, value: tensor([-0.0494, 0.0141, 0.0037, -0.0068, -0.0054, 0.0035, -0.0148, 0.0348, + -0.0115, 0.0128], device='cuda:0'), grad: tensor([ 0.0149, -0.0574, -0.0481, 0.0269, -0.0102, 0.0312, -0.0105, 0.0354, + -0.0161, 0.0340], device='cuda:0') +100 +0.0001 +changing lr +epoch 234, time 214.64, cls_loss 0.5520 cls_loss_mapping 0.0044 cls_loss_causal 0.4751 re_mapping 0.0114 re_causal 0.0257 /// teacc 98.85 lr 0.00010000 +Epoch 236, weight, value: tensor([[-0.1666, -0.0257, 0.0207, ..., -0.0311, -0.0724, -0.1139], + [-0.0392, -0.0769, 0.0243, ..., 0.0365, -0.0097, -0.0804], + [-0.0536, -0.0657, -0.0922, ..., 0.0463, -0.0422, -0.0865], + ..., + [-0.0574, 0.0282, 0.0205, ..., 0.0336, -0.0471, -0.0883], + [-0.0718, -0.0093, 0.0218, ..., 0.0633, -0.0409, -0.1474], + [ 0.0510, 0.0535, -0.0452, ..., -0.1048, -0.0212, 0.0638]], + device='cuda:0'), grad: tensor([[ 6.4087e-04, 2.6870e-04, 7.1859e-04, ..., 0.0000e+00, + 3.6247e-06, 2.4271e-04], + [ 1.3977e-05, 2.5177e-04, 1.1183e-05, ..., 1.8626e-09, + 1.0189e-06, 7.3910e-06], + [ 9.9087e-04, -3.3360e-03, 1.1520e-03, ..., 1.8626e-09, + 4.1202e-06, 3.9601e-04], + ..., + [ 1.5945e-03, 4.4227e-04, 7.6437e-04, ..., -1.1362e-07, + 2.3276e-05, 6.8998e-04], + [ 7.3719e-04, 1.0786e-03, 7.8249e-04, ..., 0.0000e+00, + 2.2188e-05, 2.9469e-04], + [-1.4977e-02, 4.5586e-04, -1.6205e-02, ..., 9.8720e-08, + -1.1450e-04, -5.5084e-03]], device='cuda:0') +Epoch 236, bias, value: tensor([-0.0490, 0.0139, 0.0039, -0.0063, -0.0042, 0.0043, -0.0163, 0.0341, + -0.0123, 0.0130], device='cuda:0'), grad: tensor([ 0.0109, 0.0077, -0.0057, 0.0053, 0.0148, -0.0095, 0.0076, -0.0045, + -0.0211, -0.0054], device='cuda:0') +100 +0.0001 +changing lr +epoch 235, time 214.66, cls_loss 0.5689 cls_loss_mapping 0.0055 cls_loss_causal 0.5078 re_mapping 0.0111 re_causal 0.0258 /// teacc 98.66 lr 0.00010000 +Epoch 237, weight, value: tensor([[-0.1693, -0.0247, 0.0193, ..., -0.0316, -0.0724, -0.1141], + [-0.0388, -0.0757, 0.0243, ..., 0.0367, -0.0099, -0.0799], + [-0.0517, -0.0662, -0.0933, ..., 0.0464, -0.0415, -0.0872], + ..., + [-0.0578, 0.0281, 0.0208, ..., 0.0336, -0.0472, -0.0884], + [-0.0711, -0.0097, 0.0217, ..., 0.0632, -0.0413, -0.1473], + [ 0.0495, 0.0538, -0.0444, ..., -0.1051, -0.0211, 0.0623]], + device='cuda:0'), grad: tensor([[ 1.8921e-03, -2.8954e-03, -5.9700e-04, ..., -1.6050e-03, + -5.7936e-04, 1.0765e-04], + [ 3.0184e-04, 3.4790e-03, 8.0299e-04, ..., 2.7084e-04, + 2.7132e-04, -4.3303e-05], + [ 1.1187e-03, -1.2665e-03, -8.4534e-03, ..., -9.0837e-04, + -9.1219e-04, -2.9969e-04], + ..., + [ 5.9128e-04, 5.6038e-03, 1.4935e-03, ..., 3.5739e-04, + 1.2106e-04, 8.8215e-04], + [ 3.7155e-03, 3.4218e-03, 1.5316e-03, ..., 6.9284e-04, + 1.1051e-04, 1.6632e-03], + [-1.2474e-03, 3.5286e-03, -3.9721e-04, ..., 1.7560e-04, + 1.0449e-04, -3.0575e-03]], device='cuda:0') +Epoch 237, bias, value: tensor([-0.0489, 0.0143, 0.0037, -0.0059, -0.0040, 0.0042, -0.0172, 0.0342, + -0.0124, 0.0129], device='cuda:0'), grad: tensor([-0.0109, 0.0185, -0.0483, 0.0023, 0.0134, -0.0167, 0.0081, 0.0240, + -0.0044, 0.0140], device='cuda:0') +100 +0.0001 +changing lr +epoch 236, time 214.64, cls_loss 0.5700 cls_loss_mapping 0.0059 cls_loss_causal 0.5048 re_mapping 0.0112 re_causal 0.0253 /// teacc 98.73 lr 0.00010000 +Epoch 238, weight, value: tensor([[-0.1691, -0.0237, 0.0192, ..., -0.0317, -0.0726, -0.1137], + [-0.0387, -0.0734, 0.0251, ..., 0.0366, -0.0098, -0.0800], + [-0.0520, -0.0682, -0.0940, ..., 0.0465, -0.0414, -0.0874], + ..., + [-0.0581, 0.0289, 0.0207, ..., 0.0339, -0.0464, -0.0890], + [-0.0713, -0.0084, 0.0209, ..., 0.0630, -0.0418, -0.1465], + [ 0.0504, 0.0522, -0.0436, ..., -0.1050, -0.0207, 0.0624]], + device='cuda:0'), grad: tensor([[ 1.5414e-04, 3.6359e-04, 3.0708e-04, ..., 2.4939e-04, + 1.0118e-03, 2.5010e-04], + [-3.2520e-04, -7.9250e-04, -1.5163e-04, ..., -6.1369e-04, + 1.4286e-03, 1.7118e-06], + [ 4.3124e-05, 1.0017e-02, 4.0197e-04, ..., 1.5364e-03, + 4.5128e-03, 1.4611e-05], + ..., + [ 3.3045e-04, -8.4076e-03, 1.5192e-03, ..., 1.8048e-04, + 2.1782e-03, 3.7879e-05], + [ 6.5684e-05, 3.6526e-04, 7.1406e-05, ..., 1.8775e-04, + 2.7142e-03, 1.5259e-04], + [-3.1799e-05, 1.6332e-04, -3.8738e-03, ..., -1.3008e-03, + 3.5286e-04, -8.1897e-05]], device='cuda:0') +Epoch 238, bias, value: tensor([-0.0488, 0.0144, 0.0029, -0.0059, -0.0038, 0.0035, -0.0176, 0.0349, + -0.0121, 0.0134], device='cuda:0'), grad: tensor([ 0.0125, 0.0039, 0.0387, -0.0171, -0.0517, 0.0029, 0.0076, -0.0035, + 0.0131, -0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 237, time 214.49, cls_loss 0.5797 cls_loss_mapping 0.0050 cls_loss_causal 0.5136 re_mapping 0.0110 re_causal 0.0256 /// teacc 98.85 lr 0.00010000 +Epoch 239, weight, value: tensor([[-0.1690, -0.0245, 0.0193, ..., -0.0316, -0.0724, -0.1143], + [-0.0384, -0.0729, 0.0269, ..., 0.0365, -0.0105, -0.0808], + [-0.0520, -0.0683, -0.0953, ..., 0.0462, -0.0419, -0.0869], + ..., + [-0.0581, 0.0279, 0.0205, ..., 0.0343, -0.0470, -0.0887], + [-0.0717, -0.0089, 0.0217, ..., 0.0636, -0.0412, -0.1468], + [ 0.0504, 0.0518, -0.0438, ..., -0.1048, -0.0210, 0.0625]], + device='cuda:0'), grad: tensor([[ 5.7936e-04, -5.2691e-04, -2.1267e-03, ..., 5.5760e-05, + 6.8045e-04, -4.0474e-03], + [ 9.5272e-04, 4.9496e-04, 1.4696e-03, ..., 2.2352e-04, + 1.2875e-03, 2.8300e-04], + [-2.6584e-04, 2.5249e-04, -8.3447e-04, ..., 1.2755e-04, + -2.8076e-02, 7.7868e-04], + ..., + [-1.1311e-03, 5.7650e-04, -2.4319e-03, ..., -8.2016e-04, + -3.1776e-03, 3.8719e-04], + [-3.3021e-04, 1.0262e-03, -1.5860e-03, ..., 8.4698e-05, + 1.2314e-02, 8.5592e-04], + [-1.3714e-03, 1.2465e-03, 9.8705e-05, ..., 9.6917e-05, + 8.9025e-04, -1.4544e-04]], device='cuda:0') +Epoch 239, bias, value: tensor([-0.0481, 0.0141, 0.0041, -0.0042, -0.0043, 0.0022, -0.0165, 0.0343, + -0.0121, 0.0115], device='cuda:0'), grad: tensor([ 0.0011, -0.0107, -0.0057, 0.0435, 0.0024, 0.0112, -0.0329, -0.0310, + 0.0063, 0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 238, time 214.88, cls_loss 0.5803 cls_loss_mapping 0.0045 cls_loss_causal 0.5061 re_mapping 0.0116 re_causal 0.0261 /// teacc 98.76 lr 0.00010000 +Epoch 240, weight, value: tensor([[-0.1704, -0.0237, 0.0196, ..., -0.0312, -0.0723, -0.1139], + [-0.0373, -0.0736, 0.0269, ..., 0.0368, -0.0109, -0.0808], + [-0.0517, -0.0690, -0.0951, ..., 0.0474, -0.0424, -0.0861], + ..., + [-0.0580, 0.0288, 0.0194, ..., 0.0333, -0.0478, -0.0887], + [-0.0723, -0.0093, 0.0212, ..., 0.0639, -0.0417, -0.1480], + [ 0.0511, 0.0520, -0.0421, ..., -0.1051, -0.0206, 0.0631]], + device='cuda:0'), grad: tensor([[ 2.0456e-04, 3.4237e-04, 2.8682e-04, ..., -2.5916e-04, + 1.3905e-03, 3.9840e-04], + [ 9.6202e-05, -9.8610e-04, 2.1038e-03, ..., 3.4332e-04, + 7.5674e-04, 4.8828e-04], + [ 4.2677e-04, 9.7466e-04, 1.6413e-03, ..., 1.7238e-04, + 3.4428e-03, 1.3094e-03], + ..., + [ 5.3930e-04, -1.2903e-03, -9.1553e-03, ..., -1.9228e-04, + 3.0155e-03, 3.4690e-04], + [ 1.7416e-04, 2.3556e-04, 1.9372e-05, ..., -5.9307e-05, + -1.5503e-02, 4.0221e-04], + [-2.9588e-04, 7.1478e-04, 3.4943e-03, ..., 2.5892e-04, + 4.3297e-04, 4.3035e-05]], device='cuda:0') +Epoch 240, bias, value: tensor([-0.0482, 0.0142, 0.0045, -0.0054, -0.0035, 0.0032, -0.0178, 0.0343, + -0.0123, 0.0122], device='cuda:0'), grad: tensor([-0.0187, 0.0346, 0.0188, 0.0015, 0.0020, 0.0122, -0.0124, 0.0097, + -0.0684, 0.0207], device='cuda:0') +100 +0.0001 +changing lr +epoch 239, time 214.86, cls_loss 0.5589 cls_loss_mapping 0.0040 cls_loss_causal 0.4848 re_mapping 0.0114 re_causal 0.0267 /// teacc 98.76 lr 0.00010000 +Epoch 241, weight, value: tensor([[-0.1711, -0.0236, 0.0194, ..., -0.0314, -0.0716, -0.1146], + [-0.0380, -0.0736, 0.0264, ..., 0.0374, -0.0097, -0.0809], + [-0.0515, -0.0688, -0.0962, ..., 0.0471, -0.0432, -0.0861], + ..., + [-0.0558, 0.0281, 0.0193, ..., 0.0330, -0.0465, -0.0889], + [-0.0728, -0.0106, 0.0221, ..., 0.0637, -0.0420, -0.1491], + [ 0.0518, 0.0531, -0.0428, ..., -0.1050, -0.0208, 0.0645]], + device='cuda:0'), grad: tensor([[ 1.5533e-04, 3.2574e-05, 2.5988e-04, ..., -2.7657e-04, + 3.7041e-03, 1.3900e-04], + [-1.8156e-04, -9.9719e-05, 1.0290e-03, ..., -3.3259e-04, + 4.6730e-05, 3.6049e-04], + [ 4.9973e-04, 1.9491e-04, 3.1395e-03, ..., 1.9598e-04, + 2.2984e-03, 5.1022e-04], + ..., + [ 4.3058e-04, -6.2525e-05, 7.0572e-04, ..., 2.2495e-04, + 1.0860e-04, 6.6137e-04], + [ 4.4441e-04, 6.8009e-05, 1.1757e-02, ..., 2.3162e-04, + 2.3937e-03, 2.2233e-04], + [-1.2465e-03, 4.6641e-05, -2.9850e-03, ..., -1.2455e-03, + 1.2217e-03, -3.3779e-03]], device='cuda:0') +Epoch 241, bias, value: tensor([-0.0478, 0.0140, 0.0040, -0.0054, -0.0040, 0.0038, -0.0168, 0.0341, + -0.0133, 0.0125], device='cuda:0'), grad: tensor([-0.0124, 0.0179, 0.0214, -0.0582, -0.0230, 0.0231, -0.0335, 0.0159, + 0.0510, -0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 240, time 215.03, cls_loss 0.5523 cls_loss_mapping 0.0041 cls_loss_causal 0.4733 re_mapping 0.0108 re_causal 0.0239 /// teacc 98.81 lr 0.00010000 +Epoch 242, weight, value: tensor([[-0.1701, -0.0237, 0.0194, ..., -0.0318, -0.0718, -0.1136], + [-0.0374, -0.0744, 0.0270, ..., 0.0382, -0.0094, -0.0812], + [-0.0516, -0.0672, -0.0959, ..., 0.0468, -0.0439, -0.0868], + ..., + [-0.0565, 0.0280, 0.0199, ..., 0.0325, -0.0458, -0.0893], + [-0.0718, -0.0094, 0.0213, ..., 0.0633, -0.0417, -0.1487], + [ 0.0512, 0.0529, -0.0437, ..., -0.1049, -0.0214, 0.0654]], + device='cuda:0'), grad: tensor([[ 2.9731e-04, 8.0824e-05, 3.7479e-04, ..., 7.1859e-04, + 8.8024e-04, 3.6001e-04], + [ 3.9315e-04, 1.5664e-04, 1.2350e-03, ..., 4.9877e-04, + 3.5477e-04, 4.5514e-04], + [-4.8828e-03, -2.7943e-04, -1.6708e-03, ..., -2.8725e-03, + 5.6648e-04, -7.3128e-03], + ..., + [-3.2692e-03, 1.9193e-04, -1.9426e-03, ..., -2.2411e-03, + 8.9216e-04, -2.3499e-03], + [ 8.9121e-04, 9.9945e-04, 1.1520e-03, ..., -1.8156e-04, + 3.5076e-03, 1.3151e-03], + [ 1.7548e-03, -3.5223e-06, 2.2488e-03, ..., 2.1896e-03, + 7.2622e-04, 1.0309e-03]], device='cuda:0') +Epoch 242, bias, value: tensor([-0.0475, 0.0147, 0.0026, -0.0048, -0.0030, 0.0048, -0.0170, 0.0340, + -0.0140, 0.0115], device='cuda:0'), grad: tensor([ 0.0170, 0.0203, -0.0701, 0.0156, -0.0148, 0.0213, -0.0154, 0.0097, + -0.0074, 0.0237], device='cuda:0') +100 +0.0001 +changing lr +epoch 241, time 214.87, cls_loss 0.5649 cls_loss_mapping 0.0035 cls_loss_causal 0.4829 re_mapping 0.0116 re_causal 0.0261 /// teacc 98.78 lr 0.00010000 +Epoch 243, weight, value: tensor([[-0.1718, -0.0230, 0.0205, ..., -0.0317, -0.0717, -0.1130], + [-0.0378, -0.0748, 0.0266, ..., 0.0385, -0.0092, -0.0822], + [-0.0524, -0.0675, -0.0953, ..., 0.0471, -0.0440, -0.0858], + ..., + [-0.0575, 0.0285, 0.0194, ..., 0.0329, -0.0461, -0.0887], + [-0.0691, -0.0092, 0.0217, ..., 0.0642, -0.0420, -0.1490], + [ 0.0507, 0.0546, -0.0442, ..., -0.1056, -0.0210, 0.0648]], + device='cuda:0'), grad: tensor([[ 1.3924e-04, 8.3387e-05, 6.8378e-04, ..., -3.3684e-03, + 7.3385e-04, 2.8491e-05], + [ 1.4663e-04, 1.5104e-04, 3.7491e-05, ..., 4.0588e-03, + 4.7833e-05, 8.8513e-05], + [-4.0932e-03, 1.8358e-04, -1.9741e-03, ..., -4.1533e-04, + -1.7691e-03, -2.5597e-03], + ..., + [-1.1406e-02, -1.7014e-02, 2.4819e-04, ..., 2.6665e-03, + -6.2981e-03, -6.1722e-03], + [ 3.3360e-03, 1.5602e-03, 1.1241e-04, ..., -2.7008e-03, + 2.7442e-04, 1.9608e-03], + [ 1.8204e-02, 6.1150e-03, 6.0081e-04, ..., 3.1528e-03, + 6.4125e-03, 1.0208e-02]], device='cuda:0') +Epoch 243, bias, value: tensor([-0.0470, 0.0146, 0.0039, -0.0060, -0.0032, 0.0049, -0.0177, 0.0342, + -0.0137, 0.0113], device='cuda:0'), grad: tensor([-0.0107, 0.0239, -0.0002, -0.0005, -0.0027, 0.0181, -0.0254, -0.0407, + -0.0090, 0.0472], device='cuda:0') +100 +0.0001 +changing lr +epoch 242, time 214.68, cls_loss 0.5995 cls_loss_mapping 0.0041 cls_loss_causal 0.5168 re_mapping 0.0112 re_causal 0.0262 /// teacc 98.74 lr 0.00010000 +Epoch 244, weight, value: tensor([[-0.1718, -0.0227, 0.0206, ..., -0.0310, -0.0715, -0.1138], + [-0.0351, -0.0751, 0.0269, ..., 0.0370, -0.0093, -0.0823], + [-0.0521, -0.0661, -0.0963, ..., 0.0486, -0.0447, -0.0861], + ..., + [-0.0580, 0.0289, 0.0205, ..., 0.0321, -0.0464, -0.0877], + [-0.0682, -0.0093, 0.0219, ..., 0.0645, -0.0428, -0.1498], + [ 0.0514, 0.0550, -0.0437, ..., -0.1068, -0.0218, 0.0654]], + device='cuda:0'), grad: tensor([[ 1.2767e-04, 9.6416e-04, 1.5049e-03, ..., 2.1591e-03, + 1.5011e-03, 1.1027e-04], + [ 1.3900e-04, 8.1396e-04, 3.5071e-04, ..., -1.9875e-03, + 1.2379e-03, 1.4400e-04], + [ 1.4391e-03, 5.5552e-04, 1.9207e-03, ..., 3.1776e-03, + 2.2793e-03, 2.7728e-04], + ..., + [ 4.4227e-04, -1.3285e-03, 9.8991e-04, ..., 2.3575e-03, + 1.1997e-03, 5.6684e-05], + [ 1.3466e-03, -1.2608e-03, 4.0131e-03, ..., 1.8463e-03, + 1.9741e-03, 1.7405e-03], + [-5.3864e-03, 1.3332e-03, -7.4196e-03, ..., -1.9836e-03, + -3.6087e-03, 5.3215e-04]], device='cuda:0') +Epoch 244, bias, value: tensor([-0.0467, 0.0138, 0.0037, -0.0057, -0.0025, 0.0044, -0.0175, 0.0339, + -0.0137, 0.0117], device='cuda:0'), grad: tensor([ 0.0267, -0.0055, 0.0245, -0.0583, 0.0286, 0.0125, -0.0165, 0.0215, + -0.0077, -0.0258], device='cuda:0') +100 +0.0001 +changing lr +epoch 243, time 215.04, cls_loss 0.5673 cls_loss_mapping 0.0048 cls_loss_causal 0.4999 re_mapping 0.0112 re_causal 0.0263 /// teacc 98.74 lr 0.00010000 +Epoch 245, weight, value: tensor([[-0.1718, -0.0222, 0.0207, ..., -0.0314, -0.0719, -0.1126], + [-0.0352, -0.0756, 0.0261, ..., 0.0364, -0.0104, -0.0830], + [-0.0512, -0.0651, -0.0974, ..., 0.0486, -0.0447, -0.0869], + ..., + [-0.0583, 0.0308, 0.0201, ..., 0.0324, -0.0450, -0.0873], + [-0.0692, -0.0095, 0.0213, ..., 0.0650, -0.0431, -0.1498], + [ 0.0523, 0.0549, -0.0424, ..., -0.1068, -0.0215, 0.0661]], + device='cuda:0'), grad: tensor([[ 1.1712e-04, -7.2384e-04, 7.4208e-06, ..., -4.4847e-04, + 9.8765e-05, -1.3828e-04], + [ 7.1466e-05, -1.8075e-05, 9.0301e-06, ..., -1.0586e-03, + 1.0151e-04, 9.6440e-05], + [ 3.4237e-04, 3.6144e-04, 3.2067e-04, ..., 5.9748e-04, + 4.5633e-04, 4.2796e-04], + ..., + [-2.0294e-03, -5.0278e-03, -2.0981e-03, ..., -2.5806e-03, + -4.1351e-03, -2.3746e-03], + [ 8.9025e-04, 4.5085e-04, 7.7581e-04, ..., 1.1082e-03, + 1.0996e-03, 9.1314e-04], + [ 1.4794e-04, 4.5013e-03, 1.2803e-04, ..., 5.2357e-04, + 1.1212e-04, 1.5812e-03]], device='cuda:0') +Epoch 245, bias, value: tensor([-0.0477, 0.0142, 0.0035, -0.0063, -0.0025, 0.0043, -0.0178, 0.0346, + -0.0141, 0.0131], device='cuda:0'), grad: tensor([ 0.0095, -0.0132, 0.0162, -0.0180, 0.0140, 0.0236, 0.0110, -0.0257, + -0.0115, -0.0059], device='cuda:0') +100 +0.0001 +changing lr +epoch 244, time 214.76, cls_loss 0.5565 cls_loss_mapping 0.0050 cls_loss_causal 0.4938 re_mapping 0.0109 re_causal 0.0248 /// teacc 98.77 lr 0.00010000 +Epoch 246, weight, value: tensor([[-0.1724, -0.0230, 0.0217, ..., -0.0306, -0.0710, -0.1118], + [-0.0345, -0.0768, 0.0274, ..., 0.0367, -0.0106, -0.0837], + [-0.0491, -0.0635, -0.0979, ..., 0.0481, -0.0434, -0.0868], + ..., + [-0.0601, 0.0314, 0.0188, ..., 0.0320, -0.0454, -0.0871], + [-0.0711, -0.0101, 0.0211, ..., 0.0646, -0.0437, -0.1496], + [ 0.0532, 0.0546, -0.0426, ..., -0.1079, -0.0203, 0.0645]], + device='cuda:0'), grad: tensor([[ 1.4460e-04, 1.6427e-04, -8.2135e-05, ..., 9.7930e-05, + 7.0512e-05, 2.6798e-04], + [ 7.5519e-05, -1.4954e-03, 1.4210e-04, ..., 8.1241e-05, + 7.8499e-05, 7.5936e-05], + [ 4.8447e-04, 2.7966e-04, 6.9809e-04, ..., 1.6320e-04, + 2.3258e-04, 1.6441e-03], + ..., + [-3.4308e-04, -3.8123e-04, 1.2767e-04, ..., -4.3797e-04, + -4.6790e-05, -1.0023e-03], + [ 4.5433e-03, 1.3685e-03, -3.9220e-04, ..., 9.0957e-05, + 6.3610e-04, 1.0614e-03], + [-1.8661e-02, -3.0727e-03, -1.3580e-03, ..., 3.5715e-04, + 2.2745e-04, -6.3782e-03]], device='cuda:0') +Epoch 246, bias, value: tensor([-0.0477, 0.0150, 0.0033, -0.0067, -0.0022, 0.0044, -0.0177, 0.0338, + -0.0138, 0.0127], device='cuda:0'), grad: tensor([-0.0249, 0.0115, 0.0259, 0.0189, 0.0150, -0.0046, 0.0253, 0.0083, + -0.0116, -0.0638], device='cuda:0') +100 +0.0001 +changing lr +epoch 245, time 214.97, cls_loss 0.5620 cls_loss_mapping 0.0054 cls_loss_causal 0.4884 re_mapping 0.0113 re_causal 0.0253 /// teacc 98.73 lr 0.00010000 +Epoch 247, weight, value: tensor([[-0.1740, -0.0232, 0.0219, ..., -0.0303, -0.0717, -0.1125], + [-0.0339, -0.0765, 0.0274, ..., 0.0374, -0.0124, -0.0835], + [-0.0503, -0.0630, -0.0976, ..., 0.0489, -0.0422, -0.0882], + ..., + [-0.0591, 0.0306, 0.0189, ..., 0.0320, -0.0452, -0.0866], + [-0.0725, -0.0089, 0.0212, ..., 0.0645, -0.0448, -0.1495], + [ 0.0535, 0.0542, -0.0428, ..., -0.1086, -0.0206, 0.0646]], + device='cuda:0'), grad: tensor([[ 2.4717e-06, 4.3082e-04, 1.2016e-03, ..., 3.0994e-04, + 3.2592e-04, 3.2991e-05], + [ 2.8625e-05, 3.9488e-05, 8.6641e-04, ..., 4.2081e-04, + 7.8821e-04, 2.2382e-05], + [ 4.8615e-06, 5.1308e-04, -5.2490e-03, ..., -3.9458e-05, + -2.8858e-03, 2.4170e-05], + ..., + [ 1.5903e-04, 9.9945e-03, 1.0185e-03, ..., -3.2425e-04, + 2.2805e-04, -2.5606e-04], + [ 9.0241e-05, 1.9503e-04, -8.6260e-04, ..., -9.3555e-04, + 1.9026e-04, 2.1517e-05], + [ 1.3475e-03, -1.1826e-02, 2.1229e-03, ..., 5.6171e-04, + 1.0109e-03, 1.5354e-04]], device='cuda:0') +Epoch 247, bias, value: tensor([-0.0477, 0.0141, 0.0042, -0.0075, -0.0027, 0.0053, -0.0177, 0.0346, + -0.0141, 0.0128], device='cuda:0'), grad: tensor([ 0.0121, 0.0157, -0.0161, 0.0122, 0.0081, -0.0237, 0.0088, 0.0293, + -0.0404, -0.0059], device='cuda:0') +100 +0.0001 +changing lr +epoch 246, time 214.83, cls_loss 0.5446 cls_loss_mapping 0.0050 cls_loss_causal 0.4747 re_mapping 0.0107 re_causal 0.0258 /// teacc 98.77 lr 0.00010000 +Epoch 248, weight, value: tensor([[-0.1743, -0.0201, 0.0213, ..., -0.0305, -0.0732, -0.1124], + [-0.0350, -0.0766, 0.0277, ..., 0.0368, -0.0131, -0.0840], + [-0.0506, -0.0637, -0.0985, ..., 0.0497, -0.0428, -0.0879], + ..., + [-0.0602, 0.0304, 0.0185, ..., 0.0313, -0.0437, -0.0873], + [-0.0721, -0.0092, 0.0213, ..., 0.0648, -0.0443, -0.1494], + [ 0.0540, 0.0540, -0.0437, ..., -0.1081, -0.0203, 0.0646]], + device='cuda:0'), grad: tensor([[ 3.0065e-04, 1.1995e-05, -3.9711e-03, ..., -5.5580e-03, + 7.0371e-06, -6.5374e-04], + [ 1.0115e-04, -1.1861e-04, 3.1567e-04, ..., 9.5510e-04, + 4.6492e-06, 7.0453e-05], + [ 3.4451e-04, -3.0708e-04, 2.2614e-04, ..., -1.2350e-04, + 3.0756e-05, 1.7560e-04], + ..., + [-2.0142e-03, -7.8821e-04, 9.8586e-05, ..., 8.8024e-04, + 1.2927e-05, -6.3229e-04], + [ 8.7309e-04, 1.0738e-03, 5.0402e-04, ..., 1.5078e-03, + 1.1653e-04, 2.9182e-04], + [-8.5258e-04, 5.9795e-04, 1.2875e-04, ..., 9.5844e-05, + -2.2078e-04, 1.2934e-04]], device='cuda:0') +Epoch 248, bias, value: tensor([-0.0467, 0.0134, 0.0052, -0.0075, -0.0033, 0.0054, -0.0190, 0.0339, + -0.0123, 0.0122], device='cuda:0'), grad: tensor([-0.0705, 0.0160, -0.0155, 0.0162, 0.0189, -0.0194, 0.0200, -0.0091, + 0.0253, 0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 247, time 214.57, cls_loss 0.5555 cls_loss_mapping 0.0052 cls_loss_causal 0.4890 re_mapping 0.0109 re_causal 0.0243 /// teacc 98.79 lr 0.00010000 +Epoch 249, weight, value: tensor([[-0.1750, -0.0201, 0.0218, ..., -0.0306, -0.0736, -0.1128], + [-0.0364, -0.0770, 0.0281, ..., 0.0363, -0.0127, -0.0848], + [-0.0508, -0.0642, -0.0997, ..., 0.0488, -0.0432, -0.0877], + ..., + [-0.0599, 0.0293, 0.0191, ..., 0.0311, -0.0433, -0.0875], + [-0.0719, -0.0097, 0.0208, ..., 0.0654, -0.0442, -0.1497], + [ 0.0544, 0.0547, -0.0428, ..., -0.1071, -0.0195, 0.0652]], + device='cuda:0'), grad: tensor([[ 1.5235e-04, 2.8331e-06, 1.2617e-03, ..., 1.2321e-03, + 4.3964e-04, 6.6280e-05], + [ 4.2057e-04, 6.5640e-06, 2.6727e-04, ..., -7.9727e-04, + 1.2636e-04, 4.2588e-05], + [ 1.0700e-03, 3.8385e-05, 1.2980e-03, ..., 1.8063e-03, + 1.1196e-03, 2.4891e-04], + ..., + [-1.5610e-02, 4.9949e-05, -1.2741e-02, ..., -9.8495e-03, + -2.8286e-03, -2.2259e-03], + [ 1.4124e-03, 2.0027e-04, 1.5211e-03, ..., -1.3990e-03, + 9.6023e-05, 2.9159e-04], + [ 1.3672e-02, 1.1653e-04, 1.2512e-02, ..., 6.7940e-03, + 7.2861e-04, 1.5507e-03]], device='cuda:0') +Epoch 249, bias, value: tensor([-0.0465, 0.0141, 0.0047, -0.0071, -0.0037, 0.0047, -0.0190, 0.0346, + -0.0134, 0.0127], device='cuda:0'), grad: tensor([ 0.0152, -0.0214, 0.0201, -0.0156, 0.0142, -0.0318, 0.0147, -0.0322, + 0.0215, 0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 248, time 215.00, cls_loss 0.5550 cls_loss_mapping 0.0044 cls_loss_causal 0.4919 re_mapping 0.0107 re_causal 0.0246 /// teacc 98.79 lr 0.00010000 +Epoch 250, weight, value: tensor([[-0.1751, -0.0203, 0.0228, ..., -0.0301, -0.0733, -0.1134], + [-0.0371, -0.0776, 0.0285, ..., 0.0364, -0.0115, -0.0850], + [-0.0502, -0.0637, -0.1001, ..., 0.0484, -0.0440, -0.0864], + ..., + [-0.0602, 0.0290, 0.0173, ..., 0.0304, -0.0440, -0.0881], + [-0.0726, -0.0089, 0.0215, ..., 0.0653, -0.0433, -0.1496], + [ 0.0543, 0.0540, -0.0416, ..., -0.1067, -0.0198, 0.0649]], + device='cuda:0'), grad: tensor([[-3.1403e-02, 5.5820e-05, -1.8191e-04, ..., -4.4174e-03, + -7.0477e-04, 6.1631e-05], + [ 3.0255e-04, 5.3644e-05, -1.4901e-05, ..., 7.6103e-04, + 3.2280e-06, 3.7849e-05], + [ 7.6723e-04, 3.1972e-04, 2.9847e-05, ..., 4.5800e-04, + 1.6832e-04, 1.2106e-04], + ..., + [-1.5087e-03, -9.4414e-05, 1.8626e-05, ..., -4.4098e-03, + 7.2718e-06, 4.2462e-04], + [ 5.4169e-04, -7.3075e-05, 6.6459e-05, ..., 3.3402e-04, + 2.8572e-03, 3.3641e-04], + [-5.6601e-04, -8.3160e-04, 1.9717e-04, ..., 7.1907e-04, + 1.5354e-04, -1.7900e-03]], device='cuda:0') +Epoch 250, bias, value: tensor([-0.0473, 0.0146, 0.0051, -0.0064, -0.0040, 0.0043, -0.0180, 0.0333, + -0.0133, 0.0129], device='cuda:0'), grad: tensor([-0.0253, -0.0578, 0.0324, -0.0018, 0.0007, -0.0036, 0.0558, -0.0042, + 0.0560, -0.0524], device='cuda:0') +100 +0.0001 +changing lr +epoch 249, time 214.46, cls_loss 0.5453 cls_loss_mapping 0.0048 cls_loss_causal 0.4740 re_mapping 0.0110 re_causal 0.0248 /// teacc 98.77 lr 0.00010000 +Epoch 251, weight, value: tensor([[-0.1741, -0.0205, 0.0242, ..., -0.0304, -0.0731, -0.1135], + [-0.0386, -0.0781, 0.0275, ..., 0.0371, -0.0119, -0.0853], + [-0.0504, -0.0638, -0.1006, ..., 0.0468, -0.0436, -0.0871], + ..., + [-0.0605, 0.0297, 0.0178, ..., 0.0309, -0.0437, -0.0881], + [-0.0719, -0.0093, 0.0222, ..., 0.0666, -0.0426, -0.1497], + [ 0.0540, 0.0540, -0.0426, ..., -0.1079, -0.0202, 0.0643]], + device='cuda:0'), grad: tensor([[ 1.6332e-05, 2.0466e-03, 1.1873e-04, ..., -7.6771e-04, + -5.5552e-05, 3.2063e-03], + [ 2.2631e-06, -2.0504e-03, -4.2877e-03, ..., -1.7986e-03, + 1.3679e-05, -3.5458e-03], + [ 5.6028e-06, 1.7607e-04, 8.1837e-05, ..., 1.0319e-03, + 1.1420e-04, 1.5557e-04], + ..., + [ 1.3307e-05, 4.5371e-04, 3.0518e-03, ..., 1.0109e-03, + 1.4491e-05, 7.5102e-04], + [ 3.8357e-03, 9.2411e-04, 1.0328e-03, ..., 3.5114e-03, + 1.9574e-04, 2.2888e-03], + [-9.0866e-03, -1.0595e-03, -5.3673e-03, ..., 5.5981e-04, + 2.4939e-04, -2.8305e-03]], device='cuda:0') +Epoch 251, bias, value: tensor([-0.0473, 0.0157, 0.0040, -0.0071, -0.0030, 0.0030, -0.0182, 0.0340, + -0.0129, 0.0130], device='cuda:0'), grad: tensor([-0.0030, -0.0468, 0.0124, -0.0143, 0.0373, 0.0030, -0.0140, 0.0156, + 0.0217, -0.0119], device='cuda:0') +100 +0.0001 +changing lr +epoch 250, time 214.89, cls_loss 0.5617 cls_loss_mapping 0.0045 cls_loss_causal 0.4965 re_mapping 0.0105 re_causal 0.0231 /// teacc 98.76 lr 0.00010000 +Epoch 252, weight, value: tensor([[-0.1740, -0.0204, 0.0244, ..., -0.0304, -0.0745, -0.1135], + [-0.0392, -0.0771, 0.0281, ..., 0.0381, -0.0125, -0.0845], + [-0.0507, -0.0621, -0.1004, ..., 0.0482, -0.0433, -0.0873], + ..., + [-0.0598, 0.0291, 0.0176, ..., 0.0298, -0.0435, -0.0886], + [-0.0731, -0.0102, 0.0217, ..., 0.0652, -0.0419, -0.1507], + [ 0.0538, 0.0539, -0.0423, ..., -0.1090, -0.0208, 0.0650]], + device='cuda:0'), grad: tensor([[ 1.0371e-05, -8.1584e-07, 5.0592e-04, ..., 1.0052e-03, + 2.0647e-04, 5.9456e-06], + [ 1.4178e-05, 1.6410e-06, 7.7391e-04, ..., 1.6060e-03, + 1.9658e-04, 1.5020e-05], + [ 4.7612e-04, 3.0607e-05, 9.2840e-04, ..., 1.6193e-03, + 1.0061e-04, 5.4884e-04], + ..., + [ 2.1589e-04, 1.1897e-04, 4.9353e-04, ..., -1.1282e-03, + 1.9622e-04, 1.0377e-04], + [ 8.0824e-04, 1.6466e-05, -4.5242e-03, ..., -1.0056e-02, + -5.2404e-04, 3.0470e-04], + [-1.5802e-03, -2.3353e-04, -7.9679e-04, ..., 8.9693e-04, + -1.1158e-03, -2.0146e-04]], device='cuda:0') +Epoch 252, bias, value: tensor([-0.0476, 0.0163, 0.0042, -0.0079, -0.0036, 0.0040, -0.0183, 0.0343, + -0.0136, 0.0133], device='cuda:0'), grad: tensor([ 0.0143, 0.0213, 0.0150, 0.0108, -0.0066, 0.0193, 0.0164, -0.0180, + -0.0831, 0.0106], device='cuda:0') +100 +0.0001 +changing lr +epoch 251, time 214.71, cls_loss 0.6169 cls_loss_mapping 0.0054 cls_loss_causal 0.5419 re_mapping 0.0107 re_causal 0.0260 /// teacc 98.75 lr 0.00010000 +Epoch 253, weight, value: tensor([[-0.1755, -0.0201, 0.0239, ..., -0.0314, -0.0742, -0.1139], + [-0.0405, -0.0783, 0.0269, ..., 0.0376, -0.0130, -0.0856], + [-0.0514, -0.0607, -0.1023, ..., 0.0483, -0.0429, -0.0879], + ..., + [-0.0596, 0.0293, 0.0199, ..., 0.0298, -0.0435, -0.0886], + [-0.0724, -0.0105, 0.0223, ..., 0.0663, -0.0413, -0.1497], + [ 0.0537, 0.0545, -0.0420, ..., -0.1105, -0.0212, 0.0644]], + device='cuda:0'), grad: tensor([[ 4.7851e-04, 1.2749e-02, -7.1526e-06, ..., 7.3481e-04, + 1.9133e-05, 1.5423e-06], + [ 6.4659e-04, 2.4068e-04, 4.0866e-06, ..., 7.5006e-04, + -1.8347e-06, 4.4331e-07], + [ 1.8120e-04, 6.7282e-04, -3.7819e-05, ..., 1.6413e-03, + 1.9088e-05, 5.2601e-05], + ..., + [ 4.2648e-03, -5.5275e-03, 1.7090e-03, ..., 2.6202e-04, + -1.3375e-04, 1.2875e-05], + [-2.4963e-02, -1.1871e-02, 8.8662e-06, ..., -1.4038e-02, + 8.1882e-06, -1.3506e-04], + [ 2.5988e-04, 7.8487e-04, 1.9759e-05, ..., 5.2929e-04, + 1.0855e-05, 2.9936e-05]], device='cuda:0') +Epoch 253, bias, value: tensor([-0.0475, 0.0152, 0.0046, -0.0065, -0.0043, 0.0045, -0.0192, 0.0349, + -0.0135, 0.0129], device='cuda:0'), grad: tensor([ 0.0403, 0.0196, 0.0229, -0.0494, 0.0177, -0.0464, 0.0457, -0.0088, + -0.0632, 0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 252, time 214.89, cls_loss 0.5667 cls_loss_mapping 0.0048 cls_loss_causal 0.4989 re_mapping 0.0108 re_causal 0.0252 /// teacc 98.62 lr 0.00010000 +Epoch 254, weight, value: tensor([[-0.1760, -0.0203, 0.0225, ..., -0.0312, -0.0745, -0.1146], + [-0.0393, -0.0781, 0.0283, ..., 0.0372, -0.0129, -0.0860], + [-0.0509, -0.0612, -0.1019, ..., 0.0482, -0.0429, -0.0868], + ..., + [-0.0597, 0.0269, 0.0189, ..., 0.0296, -0.0450, -0.0880], + [-0.0725, -0.0093, 0.0228, ..., 0.0672, -0.0405, -0.1507], + [ 0.0528, 0.0563, -0.0434, ..., -0.1104, -0.0215, 0.0643]], + device='cuda:0'), grad: tensor([[-2.4357e-03, 2.2087e-03, -1.0826e-02, ..., 1.0747e-04, + -7.9727e-03, 6.2287e-06], + [ 1.7333e-04, 1.8692e-04, 2.1553e-03, ..., -1.6749e-04, + 1.3332e-03, 5.7846e-05], + [ 1.8013e-04, 2.0657e-03, 1.4267e-03, ..., 5.7220e-04, + 1.0872e-03, 4.9800e-05], + ..., + [-1.8940e-03, 4.4274e-04, -1.6647e-02, ..., -2.0237e-03, + -6.7139e-03, -4.0550e-03], + [ 1.2314e-04, -5.2338e-03, 1.4524e-03, ..., 4.9019e-04, + -1.1539e-03, 4.3899e-05], + [ 1.8520e-03, -8.5640e-04, 1.7502e-02, ..., 2.5234e-03, + 7.6256e-03, 3.6621e-03]], device='cuda:0') +Epoch 254, bias, value: tensor([-0.0486, 0.0147, 0.0051, -0.0065, -0.0026, 0.0057, -0.0189, 0.0333, + -0.0137, 0.0126], device='cuda:0'), grad: tensor([-0.0396, -0.0028, 0.0347, -0.0159, 0.0278, 0.0198, 0.0266, -0.0252, + -0.0425, 0.0171], device='cuda:0') +100 +0.0001 +changing lr +epoch 253, time 214.72, cls_loss 0.5592 cls_loss_mapping 0.0049 cls_loss_causal 0.4961 re_mapping 0.0103 re_causal 0.0241 /// teacc 98.68 lr 0.00010000 +Epoch 255, weight, value: tensor([[-0.1754, -0.0204, 0.0241, ..., -0.0315, -0.0747, -0.1151], + [-0.0398, -0.0783, 0.0280, ..., 0.0379, -0.0118, -0.0872], + [-0.0514, -0.0627, -0.1034, ..., 0.0483, -0.0432, -0.0861], + ..., + [-0.0593, 0.0268, 0.0192, ..., 0.0295, -0.0455, -0.0876], + [-0.0731, -0.0065, 0.0234, ..., 0.0672, -0.0409, -0.1511], + [ 0.0521, 0.0554, -0.0445, ..., -0.1106, -0.0210, 0.0632]], + device='cuda:0'), grad: tensor([[-4.6844e-03, -4.1695e-03, 9.9850e-04, ..., -1.0312e-05, + 5.0974e-04, -1.3990e-03], + [ 4.2081e-04, 1.7786e-04, 4.7731e-04, ..., -7.5388e-04, + 1.0147e-03, 4.3392e-04], + [-3.3627e-03, -1.1093e-02, 3.9053e-04, ..., -7.7095e-03, + -1.2045e-03, -1.1797e-03], + ..., + [ 2.6393e-04, 1.7843e-03, 4.7684e-04, ..., 1.1940e-03, + -3.3817e-03, 4.1151e-04], + [-2.0242e-04, 3.9253e-03, -3.4599e-03, ..., -2.0504e-03, + 1.9908e-04, -1.0424e-03], + [ 9.9564e-04, -5.2299e-03, 1.1835e-03, ..., 7.9250e-04, + 9.4080e-04, 9.6369e-04]], device='cuda:0') +Epoch 255, bias, value: tensor([-0.0482, 0.0154, 0.0051, -0.0074, -0.0028, 0.0062, -0.0193, 0.0326, + -0.0127, 0.0123], device='cuda:0'), grad: tensor([-0.0086, -0.0096, -0.0419, 0.0080, 0.0286, -0.0162, 0.0218, -0.0063, + 0.0083, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 254, time 214.75, cls_loss 0.5521 cls_loss_mapping 0.0074 cls_loss_causal 0.4778 re_mapping 0.0102 re_causal 0.0234 /// teacc 98.70 lr 0.00010000 +Epoch 256, weight, value: tensor([[-0.1758, -0.0199, 0.0249, ..., -0.0327, -0.0744, -0.1147], + [-0.0406, -0.0780, 0.0276, ..., 0.0382, -0.0110, -0.0871], + [-0.0515, -0.0629, -0.1032, ..., 0.0478, -0.0427, -0.0868], + ..., + [-0.0597, 0.0265, 0.0190, ..., 0.0295, -0.0465, -0.0895], + [-0.0734, -0.0064, 0.0230, ..., 0.0673, -0.0421, -0.1522], + [ 0.0530, 0.0548, -0.0434, ..., -0.1096, -0.0193, 0.0633]], + device='cuda:0'), grad: tensor([[ 3.6073e-04, 1.5993e-03, 1.2312e-03, ..., 1.1530e-03, + 3.7613e-03, 1.1749e-03], + [ 3.2812e-05, 8.0705e-05, 1.0765e-04, ..., 3.2401e-04, + 2.1780e-04, 1.1027e-05], + [-1.1605e-04, -1.5526e-03, 1.0747e-04, ..., -2.0161e-03, + 8.4448e-04, 3.0994e-05], + ..., + [ 3.2020e-04, 3.1471e-04, 3.0708e-04, ..., 1.0223e-03, + 4.9973e-04, 5.5885e-04], + [ 5.1260e-04, 6.7234e-04, 6.0034e-04, ..., 1.8988e-03, + 1.6203e-03, 6.4468e-04], + [-1.2455e-03, -5.7459e-04, -8.9884e-05, ..., -2.4259e-04, + 1.4257e-03, -3.1300e-03]], device='cuda:0') +Epoch 256, bias, value: tensor([-0.0476, 0.0152, 0.0051, -0.0075, -0.0042, 0.0071, -0.0198, 0.0329, + -0.0130, 0.0131], device='cuda:0'), grad: tensor([ 0.0369, 0.0133, -0.0111, -0.0128, -0.0185, -0.0161, -0.0462, 0.0188, + 0.0282, 0.0074], device='cuda:0') +100 +0.0001 +changing lr +epoch 255, time 214.67, cls_loss 0.5787 cls_loss_mapping 0.0049 cls_loss_causal 0.5133 re_mapping 0.0104 re_causal 0.0243 /// teacc 98.77 lr 0.00010000 +Epoch 257, weight, value: tensor([[-0.1769, -0.0205, 0.0264, ..., -0.0326, -0.0742, -0.1153], + [-0.0404, -0.0788, 0.0269, ..., 0.0373, -0.0117, -0.0874], + [-0.0515, -0.0631, -0.1047, ..., 0.0482, -0.0431, -0.0867], + ..., + [-0.0585, 0.0266, 0.0204, ..., 0.0297, -0.0458, -0.0889], + [-0.0729, -0.0072, 0.0233, ..., 0.0657, -0.0417, -0.1527], + [ 0.0530, 0.0558, -0.0433, ..., -0.1074, -0.0194, 0.0635]], + device='cuda:0'), grad: tensor([[ 0.0040, -0.0092, 0.0037, ..., 0.0002, 0.0001, 0.0018], + [-0.0003, 0.0030, 0.0002, ..., 0.0001, 0.0002, 0.0003], + [ 0.0093, 0.0021, 0.0007, ..., 0.0004, 0.0002, 0.0023], + ..., + [ 0.0004, -0.0047, -0.0049, ..., -0.0026, -0.0014, -0.0007], + [ 0.0009, 0.0020, 0.0053, ..., 0.0010, 0.0004, 0.0004], + [ 0.0025, 0.0020, 0.0012, ..., 0.0005, 0.0001, 0.0004]], + device='cuda:0') +Epoch 257, bias, value: tensor([-0.0487, 0.0146, 0.0057, -0.0075, -0.0029, 0.0062, -0.0190, 0.0330, + -0.0129, 0.0127], device='cuda:0'), grad: tensor([-0.0020, 0.0121, 0.0294, -0.0115, 0.0104, -0.0292, -0.0137, -0.0339, + 0.0213, 0.0171], device='cuda:0') +100 +0.0001 +changing lr +epoch 256, time 214.96, cls_loss 0.5677 cls_loss_mapping 0.0039 cls_loss_causal 0.5032 re_mapping 0.0106 re_causal 0.0244 /// teacc 98.87 lr 0.00010000 +Epoch 258, weight, value: tensor([[-0.1783, -0.0194, 0.0259, ..., -0.0326, -0.0738, -0.1160], + [-0.0406, -0.0790, 0.0261, ..., 0.0387, -0.0122, -0.0878], + [-0.0523, -0.0617, -0.1053, ..., 0.0471, -0.0437, -0.0868], + ..., + [-0.0587, 0.0259, 0.0235, ..., 0.0297, -0.0447, -0.0897], + [-0.0724, -0.0082, 0.0242, ..., 0.0652, -0.0390, -0.1536], + [ 0.0535, 0.0568, -0.0459, ..., -0.1070, -0.0193, 0.0640]], + device='cuda:0'), grad: tensor([[-1.4436e-04, 1.2636e-03, -1.8799e-04, ..., 3.9291e-03, + 1.1263e-03, -1.9896e-04], + [ 6.4492e-05, 5.5408e-04, -5.3501e-04, ..., -1.6861e-02, + -4.5547e-03, 2.4176e-04], + [ 2.1577e-04, 1.1063e-03, 3.3522e-04, ..., 1.9913e-03, + 4.9210e-04, -5.6915e-03], + ..., + [ 8.1182e-05, 1.8082e-03, 1.5950e-04, ..., 1.9073e-04, + 2.8515e-04, 2.8920e-04], + [ 2.5291e-03, 1.5182e-03, 1.8940e-03, ..., 5.0964e-03, + 1.3371e-03, 1.6527e-03], + [ 5.4789e-04, 3.4885e-03, 5.4264e-04, ..., 1.2417e-03, + 2.7037e-04, 3.1376e-04]], device='cuda:0') +Epoch 258, bias, value: tensor([-0.0486, 0.0152, 0.0051, -0.0078, -0.0035, 0.0057, -0.0189, 0.0332, + -0.0120, 0.0128], device='cuda:0'), grad: tensor([ 0.0131, -0.0465, 0.0015, -0.0191, -0.0140, 0.0161, 0.0149, -0.0141, + 0.0283, 0.0199], device='cuda:0') +100 +0.0001 +changing lr +epoch 257, time 214.66, cls_loss 0.5804 cls_loss_mapping 0.0055 cls_loss_causal 0.5000 re_mapping 0.0111 re_causal 0.0243 /// teacc 98.85 lr 0.00010000 +Epoch 259, weight, value: tensor([[-0.1796, -0.0199, 0.0259, ..., -0.0336, -0.0733, -0.1168], + [-0.0393, -0.0784, 0.0273, ..., 0.0406, -0.0110, -0.0880], + [-0.0514, -0.0637, -0.1052, ..., 0.0477, -0.0438, -0.0871], + ..., + [-0.0576, 0.0270, 0.0240, ..., 0.0296, -0.0441, -0.0892], + [-0.0726, -0.0062, 0.0247, ..., 0.0655, -0.0395, -0.1527], + [ 0.0540, 0.0563, -0.0467, ..., -0.1088, -0.0189, 0.0653]], + device='cuda:0'), grad: tensor([[ 2.5010e-04, -3.0041e-03, -2.2907e-03, ..., -2.1458e-03, + -1.6413e-03, -8.2850e-05], + [-1.4486e-03, 8.9788e-04, -2.9793e-03, ..., -2.4433e-03, + -1.5717e-03, -1.9226e-03], + [-4.5815e-03, -1.3514e-03, -2.6588e-03, ..., -2.3422e-03, + 1.0414e-03, -2.6264e-03], + ..., + [ 4.5776e-03, 1.3252e-02, 2.1534e-03, ..., 2.1687e-03, + 9.5463e-04, 4.6234e-03], + [ 2.4128e-03, 8.4496e-04, 2.5806e-03, ..., -4.9973e-03, + 1.8024e-03, 1.6956e-03], + [-5.0240e-03, -1.2863e-02, -1.8301e-03, ..., 1.2312e-03, + -2.1420e-03, -3.2196e-03]], device='cuda:0') +Epoch 259, bias, value: tensor([-0.0492, 0.0163, 0.0058, -0.0099, -0.0038, 0.0054, -0.0184, 0.0341, + -0.0123, 0.0130], device='cuda:0'), grad: tensor([-1.4038e-02, -9.3937e-05, -3.2959e-02, 2.1820e-02, 5.7983e-03, + -1.5650e-03, 1.1955e-02, 1.5205e-02, 2.8820e-03, -9.0103e-03], + device='cuda:0') +100 +0.0001 +changing lr +epoch 258, time 214.75, cls_loss 0.5697 cls_loss_mapping 0.0048 cls_loss_causal 0.4997 re_mapping 0.0109 re_causal 0.0238 /// teacc 98.91 lr 0.00010000 +Epoch 260, weight, value: tensor([[-0.1796, -0.0193, 0.0256, ..., -0.0330, -0.0731, -0.1170], + [-0.0394, -0.0796, 0.0262, ..., 0.0411, -0.0123, -0.0887], + [-0.0509, -0.0637, -0.1056, ..., 0.0473, -0.0449, -0.0880], + ..., + [-0.0583, 0.0278, 0.0249, ..., 0.0316, -0.0433, -0.0902], + [-0.0738, -0.0056, 0.0248, ..., 0.0650, -0.0406, -0.1540], + [ 0.0544, 0.0566, -0.0469, ..., -0.1087, -0.0196, 0.0660]], + device='cuda:0'), grad: tensor([[ 3.0661e-04, 4.6268e-06, 4.1342e-04, ..., 3.0637e-04, + -2.2068e-03, 2.1017e-04], + [ 6.5923e-05, 3.5129e-06, -6.2370e-04, ..., 5.3978e-04, + 8.0943e-05, 1.7703e-04], + [ 4.8327e-04, 1.2219e-05, 1.3411e-04, ..., -9.9468e-04, + 1.1158e-03, 2.9373e-04], + ..., + [-7.3395e-03, 6.4522e-06, 2.1744e-04, ..., 6.4230e-04, + 9.4795e-04, -4.2496e-03], + [ 2.6741e-03, -2.7552e-05, -3.1376e-03, ..., 5.4502e-04, + -1.6844e-04, 2.4261e-03], + [ 1.2726e-02, 2.4781e-05, 5.6381e-03, ..., 1.3580e-03, + 3.2253e-03, 6.2408e-03]], device='cuda:0') +Epoch 260, bias, value: tensor([-0.0484, 0.0158, 0.0053, -0.0095, -0.0043, 0.0063, -0.0184, 0.0355, + -0.0138, 0.0129], device='cuda:0'), grad: tensor([-0.0117, 0.0107, -0.0172, 0.0304, 0.0109, -0.0514, 0.0032, 0.0038, + -0.0194, 0.0408], device='cuda:0') +100 +0.0001 +changing lr +epoch 259, time 214.66, cls_loss 0.5418 cls_loss_mapping 0.0036 cls_loss_causal 0.4666 re_mapping 0.0108 re_causal 0.0244 /// teacc 98.88 lr 0.00010000 +Epoch 261, weight, value: tensor([[-0.1784, -0.0193, 0.0269, ..., -0.0313, -0.0729, -0.1164], + [-0.0409, -0.0789, 0.0251, ..., 0.0401, -0.0109, -0.0895], + [-0.0517, -0.0655, -0.1067, ..., 0.0465, -0.0459, -0.0890], + ..., + [-0.0583, 0.0279, 0.0249, ..., 0.0319, -0.0440, -0.0904], + [-0.0753, -0.0044, 0.0245, ..., 0.0648, -0.0408, -0.1541], + [ 0.0549, 0.0560, -0.0460, ..., -0.1091, -0.0203, 0.0669]], + device='cuda:0'), grad: tensor([[ 1.1959e-03, 3.2473e-04, 5.3793e-06, ..., -2.5635e-03, + -4.0460e-04, 3.3140e-04], + [ 9.7007e-06, 2.5702e-04, 3.7532e-06, ..., 1.0033e-03, + 1.1784e-04, 3.8564e-05], + [-3.0518e-03, -8.0204e-04, 1.3590e-05, ..., 2.3956e-03, + 1.4114e-03, 2.4366e-04], + ..., + [-3.2857e-06, -7.9727e-03, -5.2989e-05, ..., -1.5421e-03, + 5.2977e-04, -8.0261e-03], + [ 2.0552e-04, 1.0061e-03, -1.7226e-05, ..., -1.5612e-03, + 1.4114e-04, 9.5415e-04], + [ 1.8895e-04, 5.9891e-03, 1.1683e-04, ..., 7.3147e-04, + 1.1933e-04, 6.1417e-03]], device='cuda:0') +Epoch 261, bias, value: tensor([-0.0484, 0.0153, 0.0049, -0.0091, -0.0034, 0.0058, -0.0178, 0.0350, + -0.0137, 0.0126], device='cuda:0'), grad: tensor([-0.0063, 0.0141, 0.0026, 0.0208, -0.0199, -0.0183, 0.0291, -0.0388, + -0.0170, 0.0338], device='cuda:0') +100 +0.0001 +changing lr +epoch 260, time 214.66, cls_loss 0.5329 cls_loss_mapping 0.0043 cls_loss_causal 0.4700 re_mapping 0.0106 re_causal 0.0234 /// teacc 98.71 lr 0.00010000 +Epoch 262, weight, value: tensor([[-0.1776, -0.0191, 0.0280, ..., -0.0323, -0.0715, -0.1159], + [-0.0417, -0.0797, 0.0260, ..., 0.0397, -0.0099, -0.0907], + [-0.0522, -0.0680, -0.1079, ..., 0.0456, -0.0457, -0.0887], + ..., + [-0.0581, 0.0296, 0.0253, ..., 0.0322, -0.0444, -0.0895], + [-0.0757, -0.0046, 0.0245, ..., 0.0641, -0.0389, -0.1555], + [ 0.0540, 0.0561, -0.0468, ..., -0.1077, -0.0211, 0.0661]], + device='cuda:0'), grad: tensor([[ 8.0705e-05, -7.1526e-05, -6.7368e-03, ..., -3.8090e-03, + 2.8498e-06, -1.6098e-03], + [ 2.7442e-04, 6.5279e-04, 2.4338e-03, ..., 1.8368e-03, + 3.6471e-06, 4.3941e-04], + [ 5.9509e-04, 1.1387e-03, 2.1496e-03, ..., 2.7390e-03, + 4.3102e-06, 1.5202e-03], + ..., + [-4.0078e-04, 2.2812e-03, 1.2522e-03, ..., 3.2349e-03, + 1.8273e-06, 7.6532e-05], + [ 1.6367e-04, 2.6464e-04, 3.2234e-04, ..., 1.1740e-03, + -3.1531e-05, 7.0953e-04], + [-1.4639e-03, -4.6806e-03, 4.1580e-04, ..., -5.7259e-03, + 2.9970e-06, -2.1439e-03]], device='cuda:0') +Epoch 262, bias, value: tensor([-0.0477, 0.0165, 0.0047, -0.0096, -0.0033, 0.0060, -0.0183, 0.0344, + -0.0137, 0.0124], device='cuda:0'), grad: tensor([-0.0436, 0.0202, 0.0347, -0.0441, 0.0140, 0.0095, -0.0162, 0.0139, + 0.0191, -0.0075], device='cuda:0') +100 +0.0001 +changing lr +epoch 261, time 214.63, cls_loss 0.5545 cls_loss_mapping 0.0041 cls_loss_causal 0.4766 re_mapping 0.0110 re_causal 0.0246 /// teacc 98.80 lr 0.00010000 +Epoch 263, weight, value: tensor([[-0.1784, -0.0196, 0.0281, ..., -0.0324, -0.0713, -0.1170], + [-0.0416, -0.0799, 0.0271, ..., 0.0395, -0.0098, -0.0906], + [-0.0505, -0.0679, -0.1060, ..., 0.0463, -0.0450, -0.0878], + ..., + [-0.0594, 0.0295, 0.0251, ..., 0.0314, -0.0435, -0.0894], + [-0.0751, -0.0027, 0.0246, ..., 0.0641, -0.0387, -0.1555], + [ 0.0535, 0.0561, -0.0481, ..., -0.1083, -0.0215, 0.0659]], + device='cuda:0'), grad: tensor([[ 1.7238e-04, 1.6880e-04, 1.0425e-04, ..., 3.6597e-04, + 3.3450e-04, 2.2184e-06], + [ 4.9412e-05, 2.7250e-06, 6.4564e-04, ..., -7.7963e-04, + 2.0337e-04, 9.6019e-07], + [ 5.6648e-04, 1.9760e-03, 3.0279e-04, ..., 1.1139e-03, + 1.6851e-03, 7.1955e-04], + ..., + [ 5.2959e-05, 1.4830e-04, -2.1286e-03, ..., -1.5860e-03, + 4.1628e-04, 2.4468e-05], + [-1.5373e-03, 6.2408e-03, -4.2200e-04, ..., 4.5853e-03, + 6.5231e-03, 2.3270e-04], + [-1.2201e-04, 2.0924e-03, 2.4223e-04, ..., 9.2936e-04, + 1.7424e-03, -1.5283e-04]], device='cuda:0') +Epoch 263, bias, value: tensor([-0.0475, 0.0166, 0.0056, -0.0091, -0.0034, 0.0051, -0.0182, 0.0336, + -0.0136, 0.0121], device='cuda:0'), grad: tensor([ 0.0065, 0.0049, 0.0053, -0.0322, 0.0114, 0.0158, -0.0181, -0.0254, + 0.0215, 0.0103], device='cuda:0') +100 +0.0001 +changing lr +epoch 262, time 214.76, cls_loss 0.5482 cls_loss_mapping 0.0043 cls_loss_causal 0.4780 re_mapping 0.0105 re_causal 0.0233 /// teacc 98.80 lr 0.00010000 +Epoch 264, weight, value: tensor([[-0.1790, -0.0192, 0.0279, ..., -0.0318, -0.0719, -0.1174], + [-0.0410, -0.0807, 0.0282, ..., 0.0402, -0.0099, -0.0907], + [-0.0514, -0.0685, -0.1061, ..., 0.0450, -0.0454, -0.0895], + ..., + [-0.0611, 0.0304, 0.0246, ..., 0.0313, -0.0439, -0.0895], + [-0.0748, -0.0029, 0.0243, ..., 0.0646, -0.0385, -0.1558], + [ 0.0541, 0.0564, -0.0482, ..., -0.1084, -0.0213, 0.0666]], + device='cuda:0'), grad: tensor([[-3.1710e-05, 1.1784e-04, -8.3399e-04, ..., 3.5095e-04, + -1.2619e-06, 2.6412e-06], + [ 1.6525e-05, 3.0661e-04, 1.8668e-04, ..., 1.4277e-03, + 1.2200e-07, 3.6299e-05], + [ 2.1305e-03, -2.5864e-02, -2.7919e-04, ..., -1.3344e-02, + 2.8964e-07, 1.3128e-05], + ..., + [-5.2452e-03, -2.5444e-03, 3.3140e-04, ..., 6.6757e-04, + 1.5832e-08, -4.0131e-03], + [ 2.8396e-04, 1.9714e-02, 1.5378e-04, ..., 5.6458e-03, + -1.8626e-09, 2.2733e-04], + [ 1.1748e-04, 6.9046e-03, 3.7980e-04, ..., 3.4580e-03, + 9.6764e-07, 2.3556e-03]], device='cuda:0') +Epoch 264, bias, value: tensor([-0.0477, 0.0174, 0.0045, -0.0095, -0.0023, 0.0060, -0.0181, 0.0333, + -0.0135, 0.0112], device='cuda:0'), grad: tensor([-0.0169, 0.0309, -0.0068, -0.0089, -0.0398, 0.0266, -0.0123, -0.0015, + 0.0108, 0.0180], device='cuda:0') +100 +0.0001 +changing lr +epoch 263, time 214.55, cls_loss 0.5488 cls_loss_mapping 0.0038 cls_loss_causal 0.4832 re_mapping 0.0107 re_causal 0.0248 /// teacc 98.92 lr 0.00010000 +Epoch 265, weight, value: tensor([[-0.1800, -0.0191, 0.0277, ..., -0.0315, -0.0727, -0.1178], + [-0.0408, -0.0815, 0.0293, ..., 0.0402, -0.0099, -0.0928], + [-0.0516, -0.0679, -0.1060, ..., 0.0459, -0.0456, -0.0887], + ..., + [-0.0624, 0.0303, 0.0253, ..., 0.0299, -0.0450, -0.0905], + [-0.0757, -0.0032, 0.0236, ..., 0.0644, -0.0384, -0.1557], + [ 0.0557, 0.0571, -0.0481, ..., -0.1078, -0.0209, 0.0678]], + device='cuda:0'), grad: tensor([[ 2.2297e-03, 8.5878e-04, -3.3927e-04, ..., 3.1757e-03, + 1.7726e-04, 2.2659e-03], + [ 8.5756e-06, -2.5272e-04, 8.1015e-04, ..., -1.3790e-03, + 1.3340e-04, 4.6968e-05], + [-1.0719e-02, -1.1559e-02, 3.4065e-03, ..., -2.3895e-02, + -1.8644e-03, -1.3840e-02], + ..., + [ 1.2890e-05, 7.2479e-03, -3.3779e-03, ..., 5.6267e-04, + -3.7670e-04, 1.7405e-05], + [ 2.3289e-03, 4.3716e-03, 1.7052e-03, ..., 7.8430e-03, + 3.5048e-04, 5.5771e-03], + [ 3.3998e-04, 1.0815e-03, 2.5368e-03, ..., 1.5163e-03, + 1.4961e-04, 3.8981e-04]], device='cuda:0') +Epoch 265, bias, value: tensor([-0.0486, 0.0170, 0.0051, -0.0102, -0.0031, 0.0047, -0.0175, 0.0341, + -0.0130, 0.0125], device='cuda:0'), grad: tensor([ 0.0179, -0.0353, -0.0291, -0.0240, 0.0229, 0.0148, -0.0155, 0.0055, + 0.0173, 0.0255], device='cuda:0') +100 +0.0001 +changing lr +epoch 264, time 214.81, cls_loss 0.5787 cls_loss_mapping 0.0046 cls_loss_causal 0.5129 re_mapping 0.0104 re_causal 0.0245 /// teacc 98.83 lr 0.00010000 +Epoch 266, weight, value: tensor([[-0.1813, -0.0204, 0.0280, ..., -0.0309, -0.0735, -0.1187], + [-0.0404, -0.0815, 0.0276, ..., 0.0406, -0.0106, -0.0926], + [-0.0520, -0.0672, -0.1067, ..., 0.0468, -0.0455, -0.0900], + ..., + [-0.0610, 0.0294, 0.0241, ..., 0.0302, -0.0453, -0.0907], + [-0.0760, -0.0044, 0.0241, ..., 0.0626, -0.0385, -0.1556], + [ 0.0545, 0.0581, -0.0480, ..., -0.1081, -0.0193, 0.0677]], + device='cuda:0'), grad: tensor([[ 4.8369e-05, 2.0230e-04, 1.4887e-03, ..., 6.9284e-04, + 7.8011e-04, 1.7488e-04], + [ 7.5936e-05, 2.3613e-03, 3.0289e-03, ..., 6.7043e-04, + 3.6836e-04, 6.9046e-04], + [-4.1652e-04, 2.8491e-04, -1.3208e-03, ..., 1.7190e-04, + 2.4748e-04, 2.1327e-04], + ..., + [ 5.8472e-05, 5.9700e-04, 1.0080e-03, ..., 1.9765e-04, + 2.5892e-04, 2.0385e-04], + [ 1.3041e-04, 4.2224e-04, 2.5730e-03, ..., 1.3199e-03, + 1.5926e-03, -1.0538e-03], + [-5.6148e-05, -8.3542e-03, 1.4181e-03, ..., 4.8923e-04, + 4.2105e-04, 1.9586e-04]], device='cuda:0') +Epoch 266, bias, value: tensor([-0.0493, 0.0170, 0.0043, -0.0083, -0.0026, 0.0035, -0.0171, 0.0345, + -0.0132, 0.0122], device='cuda:0'), grad: tensor([ 0.0102, 0.0204, 0.0020, 0.0220, 0.0287, -0.0231, -0.0569, 0.0083, + 0.0034, -0.0152], device='cuda:0') +100 +0.0001 +changing lr +epoch 265, time 214.65, cls_loss 0.5440 cls_loss_mapping 0.0040 cls_loss_causal 0.4727 re_mapping 0.0104 re_causal 0.0236 /// teacc 98.79 lr 0.00010000 +Epoch 267, weight, value: tensor([[-0.1810, -0.0209, 0.0272, ..., -0.0308, -0.0755, -0.1191], + [-0.0408, -0.0819, 0.0272, ..., 0.0406, -0.0123, -0.0918], + [-0.0518, -0.0658, -0.1066, ..., 0.0470, -0.0464, -0.0901], + ..., + [-0.0608, 0.0297, 0.0234, ..., 0.0301, -0.0473, -0.0909], + [-0.0765, -0.0052, 0.0232, ..., 0.0616, -0.0379, -0.1558], + [ 0.0539, 0.0566, -0.0486, ..., -0.1074, -0.0187, 0.0676]], + device='cuda:0'), grad: tensor([[-4.8637e-04, 8.0061e-04, 4.1556e-04, ..., -9.1076e-05, + 5.6565e-05, -9.4175e-04], + [-7.7400e-03, 8.8692e-04, 7.8011e-04, ..., 4.6229e-04, + 5.9396e-05, 6.7759e-04], + [-4.0321e-03, 1.3027e-03, -3.8948e-03, ..., 6.2466e-04, + -1.8845e-03, -4.5280e-03], + ..., + [ 2.5272e-03, 1.7872e-03, 1.5898e-03, ..., 1.0939e-03, + 1.5366e-04, 2.0523e-03], + [ 1.0292e-02, 7.1478e-04, 4.8523e-03, ..., 8.8167e-04, + 1.2386e-04, 2.8381e-03], + [-8.7433e-03, -2.1038e-03, -1.3092e-02, ..., -2.9182e-03, + 2.1350e-04, -5.5008e-03]], device='cuda:0') +Epoch 267, bias, value: tensor([-0.0491, 0.0158, 0.0044, -0.0083, -0.0028, 0.0058, -0.0177, 0.0342, + -0.0136, 0.0125], device='cuda:0'), grad: tensor([ 0.0104, -0.0438, -0.0121, 0.0260, -0.0035, 0.0168, 0.0111, 0.0262, + 0.0099, -0.0410], device='cuda:0') +100 +0.0001 +changing lr +epoch 266, time 214.84, cls_loss 0.5483 cls_loss_mapping 0.0047 cls_loss_causal 0.4853 re_mapping 0.0106 re_causal 0.0247 /// teacc 98.59 lr 0.00010000 +Epoch 268, weight, value: tensor([[-0.1817, -0.0195, 0.0272, ..., -0.0299, -0.0763, -0.1192], + [-0.0382, -0.0832, 0.0275, ..., 0.0408, -0.0144, -0.0931], + [-0.0519, -0.0668, -0.1066, ..., 0.0467, -0.0445, -0.0906], + ..., + [-0.0608, 0.0308, 0.0238, ..., 0.0305, -0.0472, -0.0906], + [-0.0772, -0.0047, 0.0221, ..., 0.0608, -0.0368, -0.1563], + [ 0.0540, 0.0562, -0.0489, ..., -0.1074, -0.0190, 0.0675]], + device='cuda:0'), grad: tensor([[ 2.1011e-05, 6.3992e-04, 1.0891e-03, ..., 2.4629e-04, + 1.0424e-03, 1.2913e-03], + [-3.5465e-05, 3.5024e-04, -1.4400e-04, ..., -7.6182e-07, + 6.9952e-04, 5.1737e-04], + [-1.7190e-04, 3.5429e-04, 1.8301e-03, ..., 1.7524e-04, + 4.5681e-04, 2.7752e-04], + ..., + [ 1.1519e-05, -5.4207e-03, -3.5076e-03, ..., -2.4929e-03, + -4.6730e-03, -8.3237e-03], + [ 4.6587e-04, 2.0111e-04, 3.0499e-03, ..., 3.3045e-04, + 3.1888e-05, 6.3133e-04], + [ 1.3605e-05, 1.3466e-03, 8.5878e-04, ..., 5.4741e-04, + 1.9894e-03, 1.8892e-03]], device='cuda:0') +Epoch 268, bias, value: tensor([-0.0490, 0.0164, 0.0051, -0.0087, -0.0020, 0.0051, -0.0175, 0.0352, + -0.0147, 0.0114], device='cuda:0'), grad: tensor([-0.0058, 0.0310, 0.0173, -0.0295, 0.0097, -0.0150, -0.0086, -0.0543, + 0.0236, 0.0316], device='cuda:0') +100 +0.0001 +changing lr +epoch 267, time 214.96, cls_loss 0.5674 cls_loss_mapping 0.0044 cls_loss_causal 0.4951 re_mapping 0.0102 re_causal 0.0233 /// teacc 98.72 lr 0.00010000 +Epoch 269, weight, value: tensor([[-0.1816, -0.0189, 0.0283, ..., -0.0293, -0.0766, -0.1185], + [-0.0387, -0.0839, 0.0262, ..., 0.0394, -0.0152, -0.0943], + [-0.0528, -0.0658, -0.1072, ..., 0.0462, -0.0444, -0.0924], + ..., + [-0.0616, 0.0308, 0.0227, ..., 0.0307, -0.0484, -0.0910], + [-0.0773, -0.0050, 0.0224, ..., 0.0612, -0.0351, -0.1571], + [ 0.0562, 0.0564, -0.0488, ..., -0.1076, -0.0190, 0.0696]], + device='cuda:0'), grad: tensor([[ 2.0466e-03, 1.3809e-03, 8.0442e-04, ..., 1.3514e-03, + 6.7663e-04, 5.3482e-03], + [ 5.1928e-04, 5.4896e-05, -1.2034e-04, ..., 1.3006e-04, + 1.1406e-03, 1.7011e-04], + [ 8.2874e-04, 5.4836e-04, 2.3448e-04, ..., 4.0960e-04, + 8.7452e-04, 1.5192e-03], + ..., + [ 2.4295e-04, -8.5783e-04, 7.2241e-05, ..., 8.5890e-05, + 1.1816e-03, 4.4680e-04], + [ 1.0586e-03, 6.3610e-04, 3.5620e-04, ..., 5.5790e-04, + 7.4959e-04, 2.6722e-03], + [ 4.0770e-04, 6.7139e-03, 1.1998e-04, ..., 2.7609e-04, + 9.0075e-04, 1.0014e-03]], device='cuda:0') +Epoch 269, bias, value: tensor([-0.0483, 0.0158, 0.0050, -0.0093, -0.0022, 0.0060, -0.0178, 0.0343, + -0.0138, 0.0115], device='cuda:0'), grad: tensor([ 0.0242, 0.0086, 0.0111, 0.0128, -0.0299, -0.0499, -0.0110, 0.0076, + 0.0168, 0.0097], device='cuda:0') +100 +0.0001 +changing lr +epoch 268, time 223.08, cls_loss 0.5351 cls_loss_mapping 0.0054 cls_loss_causal 0.4558 re_mapping 0.0104 re_causal 0.0237 /// teacc 98.88 lr 0.00010000 +Epoch 270, weight, value: tensor([[-0.1819, -0.0194, 0.0279, ..., -0.0300, -0.0769, -0.1200], + [-0.0396, -0.0854, 0.0268, ..., 0.0395, -0.0157, -0.0929], + [-0.0537, -0.0657, -0.1073, ..., 0.0457, -0.0431, -0.0931], + ..., + [-0.0615, 0.0311, 0.0215, ..., 0.0307, -0.0479, -0.0913], + [-0.0762, -0.0044, 0.0227, ..., 0.0612, -0.0363, -0.1576], + [ 0.0568, 0.0568, -0.0482, ..., -0.1070, -0.0196, 0.0704]], + device='cuda:0'), grad: tensor([[ 4.7421e-04, -9.3765e-03, 4.5323e-04, ..., 1.3661e-04, + 2.4939e-04, -3.4103e-03], + [ 2.6512e-04, 4.3702e-04, 2.1887e-04, ..., 7.3791e-05, + 6.6710e-04, 1.2553e-04], + [ 2.7180e-03, 5.7945e-03, 2.3518e-03, ..., 6.8855e-04, + 6.4373e-04, 3.7880e-03], + ..., + [-8.6746e-03, 1.2321e-03, -1.0391e-02, ..., 5.6535e-05, + 4.0603e-04, -2.2049e-02], + [ 3.0308e-03, 1.9665e-03, 5.1165e-04, ..., 7.6342e-04, + 2.8968e-04, 1.8749e-03], + [-7.4577e-03, 4.7607e-03, 6.1836e-03, ..., -3.0594e-03, + 3.0279e-04, 1.1055e-02]], device='cuda:0') +Epoch 270, bias, value: tensor([-0.0490, 0.0166, 0.0055, -0.0085, -0.0028, 0.0054, -0.0178, 0.0341, + -0.0145, 0.0121], device='cuda:0'), grad: tensor([-0.0293, 0.0067, 0.0336, -0.0030, -0.0348, -0.0144, 0.0323, -0.0218, + 0.0188, 0.0119], device='cuda:0') +100 +0.0001 +changing lr +epoch 269, time 220.74, cls_loss 0.5670 cls_loss_mapping 0.0041 cls_loss_causal 0.5070 re_mapping 0.0097 re_causal 0.0236 /// teacc 98.80 lr 0.00010000 +Epoch 271, weight, value: tensor([[-0.1823, -0.0184, 0.0284, ..., -0.0298, -0.0766, -0.1206], + [-0.0380, -0.0854, 0.0268, ..., 0.0414, -0.0159, -0.0918], + [-0.0541, -0.0648, -0.1083, ..., 0.0446, -0.0436, -0.0943], + ..., + [-0.0614, 0.0307, 0.0221, ..., 0.0308, -0.0474, -0.0904], + [-0.0762, -0.0056, 0.0221, ..., 0.0610, -0.0365, -0.1581], + [ 0.0575, 0.0568, -0.0480, ..., -0.1068, -0.0190, 0.0703]], + device='cuda:0'), grad: tensor([[ 3.3951e-04, 2.0771e-03, 1.4257e-03, ..., 6.2132e-04, + 1.7822e-04, 1.7631e-04], + [ 1.9217e-04, -4.9782e-04, 6.5327e-04, ..., -5.9652e-04, + -1.0767e-03, 2.1625e-04], + [ 4.4584e-04, 3.6502e-04, 1.9064e-03, ..., 6.9714e-04, + 1.0753e-04, 2.1911e-04], + ..., + [ 1.4000e-03, 2.4948e-03, 2.2678e-03, ..., 8.2636e-04, + 1.1671e-04, 1.8272e-03], + [-1.5297e-03, -8.7814e-03, -9.5062e-03, ..., -2.8095e-03, + 4.9067e-04, 6.8855e-04], + [ 1.4992e-02, -5.4092e-03, 9.2621e-03, ..., 2.0351e-03, + -3.0175e-05, 3.7212e-03]], device='cuda:0') +Epoch 271, bias, value: tensor([-0.0490, 0.0171, 0.0046, -0.0092, -0.0033, 0.0063, -0.0174, 0.0344, + -0.0142, 0.0119], device='cuda:0'), grad: tensor([ 0.0197, 0.0165, 0.0153, -0.0762, 0.0194, 0.0224, 0.0166, 0.0257, + -0.0499, -0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 270, time 216.20, cls_loss 0.5306 cls_loss_mapping 0.0039 cls_loss_causal 0.4637 re_mapping 0.0102 re_causal 0.0229 /// teacc 98.97 lr 0.00010000 +Epoch 272, weight, value: tensor([[-0.1818, -0.0185, 0.0293, ..., -0.0290, -0.0744, -0.1201], + [-0.0382, -0.0867, 0.0269, ..., 0.0406, -0.0166, -0.0919], + [-0.0534, -0.0640, -0.1085, ..., 0.0441, -0.0454, -0.0939], + ..., + [-0.0619, 0.0300, 0.0220, ..., 0.0311, -0.0472, -0.0909], + [-0.0767, -0.0055, 0.0220, ..., 0.0612, -0.0367, -0.1579], + [ 0.0579, 0.0569, -0.0476, ..., -0.1059, -0.0182, 0.0706]], + device='cuda:0'), grad: tensor([[ 8.2180e-06, 5.0180e-06, 1.8334e-04, ..., 1.2827e-04, + 1.3268e-04, 2.1338e-05], + [ 7.4729e-06, 8.3745e-06, 2.8419e-04, ..., 1.9407e-04, + 1.1861e-04, 4.6119e-06], + [-1.8150e-05, 4.7863e-05, 1.4627e-04, ..., 1.2171e-04, + 1.2898e-04, 1.4949e-04], + ..., + [ 2.9993e-04, -4.7469e-04, -1.2341e-03, ..., -1.2884e-03, + -7.2432e-04, -2.3441e-03], + [ 8.4937e-05, -1.1605e-04, 4.3106e-04, ..., 2.7156e-04, + 4.3225e-04, 2.0199e-03], + [-6.5136e-04, 3.7503e-04, -1.7815e-03, ..., -3.7146e-04, + 1.5771e-04, -3.8624e-04]], device='cuda:0') +Epoch 272, bias, value: tensor([-0.0494, 0.0169, 0.0049, -0.0092, -0.0037, 0.0058, -0.0172, 0.0342, + -0.0142, 0.0130], device='cuda:0'), grad: tensor([ 0.0061, 0.0076, 0.0062, 0.0074, -0.0199, 0.0082, -0.0013, -0.0374, + 0.0177, 0.0055], device='cuda:0') +100 +0.0001 +changing lr +epoch 271, time 215.19, cls_loss 0.5463 cls_loss_mapping 0.0044 cls_loss_causal 0.4781 re_mapping 0.0105 re_causal 0.0242 /// teacc 98.68 lr 0.00010000 +Epoch 273, weight, value: tensor([[-0.1825, -0.0184, 0.0291, ..., -0.0282, -0.0748, -0.1215], + [-0.0380, -0.0872, 0.0269, ..., 0.0402, -0.0169, -0.0933], + [-0.0526, -0.0639, -0.1082, ..., 0.0438, -0.0456, -0.0943], + ..., + [-0.0610, 0.0293, 0.0225, ..., 0.0310, -0.0463, -0.0896], + [-0.0759, -0.0055, 0.0221, ..., 0.0618, -0.0369, -0.1550], + [ 0.0572, 0.0569, -0.0479, ..., -0.1054, -0.0189, 0.0708]], + device='cuda:0'), grad: tensor([[ 4.3027e-07, 7.7868e-04, 1.9813e-04, ..., 3.2663e-05, + 2.2769e-04, 1.1921e-06], + [ 6.9477e-07, 8.1968e-04, 5.5742e-04, ..., 2.1112e-04, + 6.2752e-04, 9.2946e-07], + [ 1.8645e-06, 1.1435e-03, 4.8780e-04, ..., 3.6210e-05, + 2.2697e-04, 6.8247e-06], + ..., + [ 2.8467e-04, 8.3237e-03, -4.3983e-03, ..., -2.2678e-03, + 8.7891e-03, 1.4439e-03], + [ 1.5974e-04, 1.0223e-02, 1.3485e-03, ..., 1.0014e-03, + 3.2902e-04, 8.8310e-04], + [-5.2452e-04, -2.0218e-02, -2.1534e-03, ..., -5.2154e-05, + -2.8534e-03, -2.5711e-03]], device='cuda:0') +Epoch 273, bias, value: tensor([-0.0494, 0.0158, 0.0050, -0.0096, -0.0036, 0.0051, -0.0167, 0.0352, + -0.0146, 0.0138], device='cuda:0'), grad: tensor([ 0.0126, 0.0128, 0.0108, 0.0134, -0.0059, -0.0499, 0.0232, 0.0021, + 0.0247, -0.0439], device='cuda:0') +100 +0.0001 +changing lr +epoch 272, time 215.34, cls_loss 0.5815 cls_loss_mapping 0.0055 cls_loss_causal 0.5145 re_mapping 0.0104 re_causal 0.0240 /// teacc 98.84 lr 0.00010000 +Epoch 274, weight, value: tensor([[-0.1830, -0.0184, 0.0282, ..., -0.0286, -0.0750, -0.1220], + [-0.0386, -0.0871, 0.0274, ..., 0.0408, -0.0142, -0.0938], + [-0.0531, -0.0642, -0.1078, ..., 0.0433, -0.0463, -0.0936], + ..., + [-0.0622, 0.0289, 0.0224, ..., 0.0301, -0.0470, -0.0913], + [-0.0749, -0.0070, 0.0205, ..., 0.0628, -0.0383, -0.1546], + [ 0.0577, 0.0581, -0.0478, ..., -0.1058, -0.0193, 0.0718]], + device='cuda:0'), grad: tensor([[ 5.6066e-06, -6.6566e-03, 4.7870e-07, ..., 5.7340e-05, + 5.3585e-05, -8.2350e-04], + [ 4.0680e-06, -4.1466e-03, 1.4640e-06, ..., -9.6359e-03, + 8.0645e-05, -6.6910e-03], + [ 5.8562e-06, 3.3092e-03, 1.3839e-06, ..., 2.7433e-05, + 7.8022e-05, 6.7520e-04], + ..., + [-1.2136e-04, 4.8599e-03, -4.9263e-05, ..., 6.8436e-03, + 4.8697e-05, 6.0883e-03], + [ 2.8992e-04, 2.5511e-04, 1.4380e-05, ..., 3.5524e-05, + 4.9740e-05, -8.1015e-04], + [ 5.4270e-05, 1.2770e-03, 5.7459e-05, ..., 1.8454e-03, + 5.2780e-05, 2.0714e-03]], device='cuda:0') +Epoch 274, bias, value: tensor([-0.0509, 0.0165, 0.0046, -0.0094, -0.0044, 0.0064, -0.0167, 0.0355, + -0.0143, 0.0136], device='cuda:0'), grad: tensor([ 0.0109, -0.0970, 0.0113, 0.0197, 0.0255, 0.0108, -0.0360, 0.0542, + 0.0025, -0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 273, time 214.77, cls_loss 0.5644 cls_loss_mapping 0.0050 cls_loss_causal 0.5002 re_mapping 0.0103 re_causal 0.0230 /// teacc 98.85 lr 0.00010000 +Epoch 275, weight, value: tensor([[-0.1836, -0.0187, 0.0282, ..., -0.0281, -0.0743, -0.1235], + [-0.0386, -0.0879, 0.0269, ..., 0.0422, -0.0145, -0.0953], + [-0.0528, -0.0641, -0.1074, ..., 0.0443, -0.0462, -0.0929], + ..., + [-0.0630, 0.0297, 0.0219, ..., 0.0303, -0.0471, -0.0905], + [-0.0752, -0.0078, 0.0209, ..., 0.0611, -0.0390, -0.1552], + [ 0.0569, 0.0581, -0.0476, ..., -0.1064, -0.0200, 0.0720]], + device='cuda:0'), grad: tensor([[-4.7541e-04, 1.8871e-04, -4.2582e-04, ..., -4.7421e-04, + 2.1994e-04, -1.5378e-05], + [ 1.1806e-03, 5.6601e-04, 5.5647e-04, ..., 3.0270e-03, + 2.2554e-04, 6.2957e-07], + [-1.5755e-03, -1.9989e-03, -7.3481e-04, ..., -5.3864e-03, + -3.8681e-03, 5.4240e-06], + ..., + [ 2.8515e-04, -1.8845e-02, 1.7393e-04, ..., 6.4802e-04, + 4.4823e-04, -2.9445e-04], + [-6.2943e-04, -3.5435e-05, -4.1795e-04, ..., -1.5106e-03, + 9.6130e-04, 1.2720e-04], + [ 2.4962e-04, 3.2067e-04, 2.1756e-04, ..., 8.1491e-04, + 2.9063e-04, 1.0133e-04]], device='cuda:0') +Epoch 275, bias, value: tensor([-0.0504, 0.0172, 0.0055, -0.0103, -0.0043, 0.0064, -0.0162, 0.0355, + -0.0147, 0.0122], device='cuda:0'), grad: tensor([-0.0520, 0.0468, -0.0092, 0.0264, 0.0286, -0.0092, -0.0059, -0.0607, + 0.0078, 0.0274], device='cuda:0') +100 +0.0001 +changing lr +epoch 274, time 214.88, cls_loss 0.5587 cls_loss_mapping 0.0035 cls_loss_causal 0.4922 re_mapping 0.0108 re_causal 0.0243 /// teacc 98.81 lr 0.00010000 +Epoch 276, weight, value: tensor([[-0.1842, -0.0187, 0.0265, ..., -0.0294, -0.0736, -0.1238], + [-0.0377, -0.0889, 0.0271, ..., 0.0417, -0.0149, -0.0947], + [-0.0530, -0.0626, -0.1066, ..., 0.0442, -0.0463, -0.0931], + ..., + [-0.0617, 0.0283, 0.0221, ..., 0.0305, -0.0459, -0.0894], + [-0.0750, -0.0083, 0.0211, ..., 0.0603, -0.0392, -0.1558], + [ 0.0567, 0.0588, -0.0472, ..., -0.1055, -0.0201, 0.0717]], + device='cuda:0'), grad: tensor([[-2.4736e-05, -1.6235e-02, -2.6951e-03, ..., -1.3573e-02, + -1.8215e-03, 3.5114e-03], + [ 1.3649e-05, 1.1911e-03, -1.5879e-03, ..., -7.8487e-04, + 1.8492e-03, -4.9710e-05], + [ 3.9458e-05, 1.7670e-02, 9.3555e-04, ..., 1.2024e-02, + 8.3685e-04, 4.2081e-04], + ..., + [ 1.6940e-04, 4.2801e-03, 4.1294e-04, ..., 1.1816e-03, + 5.1594e-04, 9.2506e-04], + [ 1.7691e-04, -3.6983e-03, 6.1607e-04, ..., -1.6623e-03, + 2.1477e-03, -9.1887e-04], + [ 2.2280e-04, -1.7911e-05, 3.5477e-04, ..., 9.0218e-04, + 5.8174e-04, -4.5052e-03]], device='cuda:0') +Epoch 276, bias, value: tensor([-0.0510, 0.0164, 0.0056, -0.0106, -0.0034, 0.0063, -0.0165, 0.0363, + -0.0140, 0.0118], device='cuda:0'), grad: tensor([-0.0550, 0.0124, 0.0225, -0.0016, -0.0036, -0.0213, 0.0309, 0.0003, + 0.0121, 0.0034], device='cuda:0') +100 +0.0001 +changing lr +epoch 275, time 215.15, cls_loss 0.5142 cls_loss_mapping 0.0035 cls_loss_causal 0.4546 re_mapping 0.0104 re_causal 0.0228 /// teacc 98.88 lr 0.00010000 +Epoch 277, weight, value: tensor([[-0.1854, -0.0177, 0.0262, ..., -0.0275, -0.0733, -0.1242], + [-0.0379, -0.0897, 0.0284, ..., 0.0411, -0.0154, -0.0947], + [-0.0534, -0.0639, -0.1064, ..., 0.0439, -0.0458, -0.0940], + ..., + [-0.0615, 0.0298, 0.0220, ..., 0.0305, -0.0457, -0.0894], + [-0.0747, -0.0083, 0.0214, ..., 0.0596, -0.0395, -0.1556], + [ 0.0563, 0.0589, -0.0472, ..., -0.1049, -0.0200, 0.0720]], + device='cuda:0'), grad: tensor([[ 2.3603e-05, 1.5780e-05, 6.9559e-05, ..., 8.3670e-06, + 0.0000e+00, 4.2498e-05], + [ 2.3216e-05, 1.6645e-05, 6.6996e-04, ..., -2.6608e-04, + 0.0000e+00, 8.5115e-05], + [ 1.8251e-04, 8.1778e-05, 4.4435e-05, ..., 3.1543e-04, + 0.0000e+00, 2.6870e-04], + ..., + [ 2.6748e-05, -3.8624e-03, 1.1711e-03, ..., -2.0310e-05, + 0.0000e+00, -3.7766e-04], + [ 3.3164e-04, 4.7302e-04, 7.8321e-05, ..., 1.0848e-04, + 0.0000e+00, 6.8760e-04], + [-9.6262e-05, 2.4185e-03, 1.8959e-03, ..., 9.9689e-06, + 0.0000e+00, 1.2016e-04]], device='cuda:0') +Epoch 277, bias, value: tensor([-0.0507, 0.0174, 0.0070, -0.0118, -0.0048, 0.0052, -0.0169, 0.0365, + -0.0131, 0.0118], device='cuda:0'), grad: tensor([-0.0814, 0.0201, 0.0131, -0.0014, -0.0396, 0.0190, 0.0178, 0.0134, + 0.0168, 0.0222], device='cuda:0') +100 +0.0001 +changing lr +epoch 276, time 215.16, cls_loss 0.5546 cls_loss_mapping 0.0043 cls_loss_causal 0.4966 re_mapping 0.0100 re_causal 0.0237 /// teacc 98.91 lr 0.00010000 +Epoch 278, weight, value: tensor([[-0.1856, -0.0186, 0.0258, ..., -0.0282, -0.0729, -0.1244], + [-0.0387, -0.0902, 0.0290, ..., 0.0407, -0.0157, -0.0945], + [-0.0540, -0.0631, -0.1065, ..., 0.0443, -0.0464, -0.0956], + ..., + [-0.0618, 0.0301, 0.0219, ..., 0.0292, -0.0460, -0.0901], + [-0.0747, -0.0075, 0.0212, ..., 0.0595, -0.0396, -0.1558], + [ 0.0560, 0.0581, -0.0471, ..., -0.1055, -0.0196, 0.0724]], + device='cuda:0'), grad: tensor([[ 1.2480e-06, 1.5378e-04, 9.2506e-04, ..., 1.1921e-04, + 3.7885e-04, 1.6894e-06], + [ 3.9674e-07, 2.6560e-04, 5.7697e-04, ..., 2.9349e-04, + 2.1541e-04, 8.3260e-07], + [ 3.9861e-06, 2.6250e-04, 7.8487e-04, ..., 3.1638e-04, + 3.0541e-04, 6.9365e-06], + ..., + [ 4.4033e-06, 5.9843e-04, 8.0824e-04, ..., 3.5739e-04, + 2.6703e-04, 1.8790e-05], + [ 4.6641e-05, -2.1667e-03, 1.3123e-03, ..., 2.8801e-04, + 8.4496e-04, 1.3294e-03], + [ 1.7080e-06, 1.5819e-04, 4.2319e-04, ..., 1.8108e-04, + 1.5843e-04, -1.4029e-05]], device='cuda:0') +Epoch 278, bias, value: tensor([-0.0505, 0.0172, 0.0065, -0.0101, -0.0047, 0.0049, -0.0168, 0.0366, + -0.0138, 0.0115], device='cuda:0'), grad: tensor([-0.0303, 0.0343, -0.0173, 0.0036, -0.0096, 0.0266, -0.0159, 0.0227, + -0.0298, 0.0158], device='cuda:0') +100 +0.0001 +changing lr +epoch 277, time 215.00, cls_loss 0.5619 cls_loss_mapping 0.0046 cls_loss_causal 0.4894 re_mapping 0.0104 re_causal 0.0241 /// teacc 98.89 lr 0.00010000 +Epoch 279, weight, value: tensor([[-0.1855, -0.0193, 0.0265, ..., -0.0284, -0.0726, -0.1253], + [-0.0390, -0.0908, 0.0285, ..., 0.0400, -0.0166, -0.0932], + [-0.0552, -0.0635, -0.1069, ..., 0.0449, -0.0462, -0.0966], + ..., + [-0.0605, 0.0301, 0.0230, ..., 0.0287, -0.0464, -0.0891], + [-0.0742, -0.0083, 0.0213, ..., 0.0599, -0.0394, -0.1551], + [ 0.0563, 0.0590, -0.0480, ..., -0.1055, -0.0183, 0.0725]], + device='cuda:0'), grad: tensor([[ 8.7321e-05, 9.5177e-04, -8.8930e-04, ..., -9.8419e-04, + 2.4986e-04, 5.2303e-05], + [ 2.1949e-05, -8.2254e-04, 1.8919e-04, ..., 4.2510e-04, + 2.8992e-04, 1.4973e-04], + [-6.1631e-05, 2.5201e-04, 1.8573e-04, ..., 8.3113e-04, + -1.2102e-03, 1.9729e-04], + ..., + [ 5.5647e-04, -9.1374e-05, 5.4312e-04, ..., 3.3617e-04, + 3.4642e-04, 6.5041e-04], + [ 3.0308e-03, -6.1989e-04, 2.8057e-03, ..., 6.6090e-04, + 3.6263e-04, 1.8139e-03], + [ 1.0910e-03, -1.7204e-03, 1.1969e-03, ..., -1.3809e-03, + 3.1185e-04, -1.6108e-03]], device='cuda:0') +Epoch 279, bias, value: tensor([-0.0491, 0.0174, 0.0064, -0.0108, -0.0060, 0.0063, -0.0180, 0.0369, + -0.0140, 0.0119], device='cuda:0'), grad: tensor([-0.0095, 0.0134, -0.0104, -0.0104, -0.0102, -0.0331, 0.0020, 0.0236, + 0.0448, -0.0103], device='cuda:0') +100 +0.0001 +changing lr +epoch 278, time 215.15, cls_loss 0.5376 cls_loss_mapping 0.0031 cls_loss_causal 0.4742 re_mapping 0.0108 re_causal 0.0246 /// teacc 98.81 lr 0.00010000 +Epoch 280, weight, value: tensor([[-0.1861, -0.0193, 0.0257, ..., -0.0278, -0.0719, -0.1263], + [-0.0388, -0.0925, 0.0290, ..., 0.0405, -0.0169, -0.0934], + [-0.0556, -0.0642, -0.1062, ..., 0.0441, -0.0471, -0.0986], + ..., + [-0.0599, 0.0303, 0.0227, ..., 0.0287, -0.0458, -0.0887], + [-0.0737, -0.0084, 0.0216, ..., 0.0600, -0.0388, -0.1535], + [ 0.0558, 0.0597, -0.0490, ..., -0.1061, -0.0180, 0.0726]], + device='cuda:0'), grad: tensor([[ 1.2424e-06, -1.8215e-04, 5.1737e-04, ..., -5.8794e-04, + -6.4611e-04, 1.2290e-04], + [ 2.1271e-06, 6.5207e-05, 6.7902e-04, ..., 8.2374e-05, + 8.6650e-06, 2.6360e-05], + [ 3.0287e-06, 6.5118e-06, -4.5242e-03, ..., 2.7332e-03, + 2.1191e-03, 7.8738e-05], + ..., + [ 5.5504e-04, 6.6376e-04, 3.9792e-04, ..., -1.5383e-03, + 1.2083e-03, 6.4087e-04], + [ 4.5002e-05, 5.2338e-03, 7.0238e-04, ..., 2.3365e-03, + 3.1090e-04, 1.7529e-03], + [-8.1062e-04, 2.5635e-03, 6.7472e-04, ..., 9.9754e-04, + -8.2541e-04, 2.1267e-04]], device='cuda:0') +Epoch 280, bias, value: tensor([-0.0492, 0.0174, 0.0064, -0.0103, -0.0062, 0.0065, -0.0182, 0.0362, + -0.0139, 0.0122], device='cuda:0'), grad: tensor([ 0.0067, -0.0171, -0.0130, -0.0153, -0.0200, 0.0116, 0.0102, -0.0209, + 0.0371, 0.0207], device='cuda:0') +100 +0.0001 +changing lr +epoch 279, time 214.82, cls_loss 0.5546 cls_loss_mapping 0.0043 cls_loss_causal 0.4912 re_mapping 0.0111 re_causal 0.0246 /// teacc 98.79 lr 0.00010000 +Epoch 281, weight, value: tensor([[-0.1861, -0.0190, 0.0249, ..., -0.0271, -0.0732, -0.1251], + [-0.0393, -0.0914, 0.0283, ..., 0.0413, -0.0153, -0.0936], + [-0.0563, -0.0650, -0.1071, ..., 0.0438, -0.0478, -0.0995], + ..., + [-0.0600, 0.0304, 0.0251, ..., 0.0277, -0.0467, -0.0894], + [-0.0745, -0.0065, 0.0220, ..., 0.0611, -0.0389, -0.1550], + [ 0.0556, 0.0596, -0.0499, ..., -0.1064, -0.0180, 0.0728]], + device='cuda:0'), grad: tensor([[ 7.1228e-06, -7.6714e-03, 2.4125e-05, ..., 3.4161e-06, + 1.6947e-03, 1.0177e-05], + [ 6.1095e-06, 4.3362e-06, 6.7115e-05, ..., 2.4647e-05, + 1.2331e-03, 1.9759e-05], + [ 2.9579e-05, 1.7929e-04, -3.4899e-05, ..., -9.9838e-05, + 1.4439e-03, -6.2585e-07], + ..., + [ 1.1390e-04, 1.0048e-02, 2.7046e-05, ..., 2.5635e-03, + 1.9417e-03, 2.4891e-03], + [-6.7329e-04, 5.2750e-05, 1.2922e-04, ..., 4.2886e-05, + -4.1885e-03, 2.7037e-04], + [ 1.6785e-04, 1.3638e-04, 7.3075e-05, ..., 9.4056e-05, + -6.1417e-04, -3.1233e-05]], device='cuda:0') +Epoch 281, bias, value: tensor([-0.0490, 0.0180, 0.0057, -0.0100, -0.0052, 0.0060, -0.0183, 0.0362, + -0.0136, 0.0111], device='cuda:0'), grad: tensor([-0.0655, 0.0290, -0.0058, -0.0075, -0.0009, -0.0353, 0.0175, 0.0713, + -0.0003, -0.0025], device='cuda:0') +100 +0.0001 +changing lr +epoch 280, time 215.53, cls_loss 0.5406 cls_loss_mapping 0.0048 cls_loss_causal 0.4737 re_mapping 0.0106 re_causal 0.0237 /// teacc 98.83 lr 0.00010000 +Epoch 282, weight, value: tensor([[-0.1850, -0.0193, 0.0258, ..., -0.0269, -0.0737, -0.1240], + [-0.0400, -0.0913, 0.0281, ..., 0.0406, -0.0153, -0.0947], + [-0.0570, -0.0653, -0.1066, ..., 0.0435, -0.0480, -0.0989], + ..., + [-0.0605, 0.0309, 0.0242, ..., 0.0278, -0.0469, -0.0897], + [-0.0745, -0.0068, 0.0219, ..., 0.0614, -0.0391, -0.1549], + [ 0.0552, 0.0605, -0.0501, ..., -0.1059, -0.0177, 0.0727]], + device='cuda:0'), grad: tensor([[ 1.5748e-04, 2.6405e-05, -2.9278e-04, ..., 7.5459e-05, + -1.1653e-04, 1.7977e-04], + [ 4.5717e-05, 1.9753e-04, 2.9236e-05, ..., 2.2531e-05, + 2.2911e-06, 4.8846e-05], + [ 1.1259e-04, 3.4189e-04, 3.8117e-05, ..., 1.0687e-04, + 6.7279e-06, 1.3840e-04], + ..., + [-4.6959e-03, 1.1549e-03, 5.6863e-05, ..., -2.2774e-03, + 3.0696e-06, -7.3395e-03], + [ 4.7898e-04, -6.3753e-04, 1.8847e-04, ..., -7.1049e-04, + 2.8431e-05, 2.5988e-04], + [ 3.2854e-04, 3.9558e-03, 8.8394e-05, ..., 2.7251e-04, + 4.1500e-06, 4.3797e-04]], device='cuda:0') +Epoch 282, bias, value: tensor([-0.0486, 0.0179, 0.0054, -0.0093, -0.0040, 0.0064, -0.0183, 0.0359, + -0.0153, 0.0110], device='cuda:0'), grad: tensor([ 0.0109, -0.0081, -0.0174, -0.0022, -0.0222, 0.0215, 0.0181, -0.0015, + 0.0100, -0.0091], device='cuda:0') +100 +0.0001 +changing lr +epoch 281, time 215.43, cls_loss 0.5485 cls_loss_mapping 0.0029 cls_loss_causal 0.4880 re_mapping 0.0108 re_causal 0.0240 /// teacc 98.90 lr 0.00010000 +Epoch 283, weight, value: tensor([[-0.1856, -0.0196, 0.0269, ..., -0.0254, -0.0737, -0.1240], + [-0.0405, -0.0921, 0.0285, ..., 0.0413, -0.0148, -0.0940], + [-0.0564, -0.0641, -0.1066, ..., 0.0441, -0.0478, -0.0994], + ..., + [-0.0609, 0.0307, 0.0233, ..., 0.0279, -0.0469, -0.0898], + [-0.0745, -0.0056, 0.0223, ..., 0.0595, -0.0398, -0.1529], + [ 0.0564, 0.0590, -0.0498, ..., -0.1058, -0.0175, 0.0736]], + device='cuda:0'), grad: tensor([[ 1.6838e-06, 2.5415e-04, 5.8975e-03, ..., 1.3306e-02, + 7.9956e-03, 1.3150e-06], + [ 3.7253e-07, 2.1210e-03, 2.7084e-03, ..., 1.3103e-03, + 2.3956e-03, 1.2070e-06], + [ 8.3260e-07, 3.8218e-04, 4.4990e-04, ..., -5.6934e-04, + -1.8730e-03, 2.1756e-06], + ..., + [ 2.2054e-05, 8.3494e-04, 2.5201e-04, ..., -2.6569e-03, + -1.0567e-03, 3.9458e-05], + [ 8.4698e-05, 5.6553e-04, 8.3685e-04, ..., 5.2834e-04, + 1.0633e-03, 1.1110e-04], + [-1.3322e-05, 1.5249e-03, -1.5091e-02, ..., 1.1654e-03, + -6.8235e-04, 1.5843e-04]], device='cuda:0') +Epoch 283, bias, value: tensor([-0.0490, 0.0193, 0.0055, -0.0095, -0.0050, 0.0065, -0.0182, 0.0357, + -0.0164, 0.0119], device='cuda:0'), grad: tensor([ 0.0511, 0.0485, -0.0213, -0.0182, 0.0036, -0.0198, 0.0185, -0.0329, + 0.0171, -0.0467], device='cuda:0') +100 +0.0001 +changing lr +epoch 282, time 214.72, cls_loss 0.5596 cls_loss_mapping 0.0037 cls_loss_causal 0.4937 re_mapping 0.0103 re_causal 0.0238 /// teacc 98.93 lr 0.00010000 +Epoch 284, weight, value: tensor([[-0.1862, -0.0193, 0.0266, ..., -0.0257, -0.0739, -0.1236], + [-0.0414, -0.0933, 0.0287, ..., 0.0412, -0.0147, -0.0942], + [-0.0573, -0.0654, -0.1076, ..., 0.0436, -0.0459, -0.0997], + ..., + [-0.0612, 0.0298, 0.0229, ..., 0.0284, -0.0468, -0.0900], + [-0.0747, -0.0060, 0.0225, ..., 0.0594, -0.0403, -0.1531], + [ 0.0576, 0.0597, -0.0505, ..., -0.1058, -0.0175, 0.0746]], + device='cuda:0'), grad: tensor([[ 4.6313e-05, 8.4281e-05, 4.3303e-05, ..., 1.7643e-04, + -9.2545e-03, 2.2161e-04], + [ 1.3307e-05, 3.4761e-04, -1.7192e-06, ..., 1.1104e-04, + -1.4820e-03, 3.0488e-05], + [ 2.4867e-04, 8.5878e-04, 2.2829e-05, ..., 1.2484e-03, + -4.6997e-03, 2.9683e-04], + ..., + [ 1.0788e-05, -4.1313e-03, 6.8396e-06, ..., -4.7340e-03, + 7.9966e-04, 3.0503e-05], + [ 3.4809e-04, 1.5421e-03, 1.0961e-04, ..., 1.6432e-03, + 2.0523e-03, 1.0796e-03], + [ 1.6212e-05, 5.4264e-04, 1.7077e-05, ..., 1.5163e-03, + 3.3436e-03, 2.1267e-04]], device='cuda:0') +Epoch 284, bias, value: tensor([-0.0495, 0.0187, 0.0052, -0.0086, -0.0061, 0.0070, -0.0170, 0.0360, + -0.0176, 0.0129], device='cuda:0'), grad: tensor([-0.0054, 0.0098, 0.0129, -0.0042, -0.0427, -0.0192, 0.0208, -0.0054, + 0.0271, 0.0063], device='cuda:0') +100 +0.0001 +changing lr +epoch 283, time 214.81, cls_loss 0.5475 cls_loss_mapping 0.0045 cls_loss_causal 0.4940 re_mapping 0.0104 re_causal 0.0236 /// teacc 98.83 lr 0.00010000 +Epoch 285, weight, value: tensor([[-0.1856, -0.0199, 0.0255, ..., -0.0270, -0.0727, -0.1239], + [-0.0413, -0.0937, 0.0285, ..., 0.0402, -0.0144, -0.0947], + [-0.0563, -0.0661, -0.1071, ..., 0.0436, -0.0441, -0.0987], + ..., + [-0.0611, 0.0302, 0.0242, ..., 0.0294, -0.0480, -0.0883], + [-0.0758, -0.0061, 0.0230, ..., 0.0604, -0.0396, -0.1534], + [ 0.0585, 0.0596, -0.0517, ..., -0.1055, -0.0176, 0.0729]], + device='cuda:0'), grad: tensor([[ 9.6858e-06, 6.3610e-04, -2.3346e-03, ..., 6.2037e-04, + -4.7989e-03, -2.6189e-06], + [ 3.4943e-06, 7.3004e-04, 1.8096e-04, ..., 6.0225e-04, + 3.2210e-04, 1.0423e-05], + [ 5.5015e-05, 6.8378e-04, 3.1137e-04, ..., 7.3099e-04, + 6.1560e-04, 9.3699e-05], + ..., + [ 2.6178e-04, 1.1620e-02, 4.0197e-04, ..., -3.7174e-03, + 5.8603e-04, 3.0208e-04], + [ 3.5048e-04, 9.0122e-04, 4.6501e-03, ..., 4.3526e-03, + 4.7493e-03, 4.9639e-04], + [-7.9441e-04, -1.2199e-02, -3.7823e-03, ..., -5.0020e-04, + -1.5583e-03, 4.0321e-03]], device='cuda:0') +Epoch 285, bias, value: tensor([-0.0500, 0.0187, 0.0053, -0.0093, -0.0069, 0.0074, -0.0162, 0.0354, + -0.0172, 0.0135], device='cuda:0'), grad: tensor([ 0.0037, 0.0108, 0.0125, 0.0366, -0.0531, 0.0107, 0.0167, 0.0031, + 0.0082, -0.0492], device='cuda:0') +100 +0.0001 +changing lr +epoch 284, time 214.69, cls_loss 0.5412 cls_loss_mapping 0.0059 cls_loss_causal 0.4730 re_mapping 0.0104 re_causal 0.0230 /// teacc 98.84 lr 0.00010000 +Epoch 286, weight, value: tensor([[-0.1860, -0.0187, 0.0252, ..., -0.0272, -0.0727, -0.1237], + [-0.0423, -0.0933, 0.0280, ..., 0.0408, -0.0146, -0.0964], + [-0.0576, -0.0660, -0.1080, ..., 0.0435, -0.0442, -0.0992], + ..., + [-0.0616, 0.0300, 0.0243, ..., 0.0284, -0.0479, -0.0886], + [-0.0755, -0.0062, 0.0232, ..., 0.0602, -0.0389, -0.1542], + [ 0.0591, 0.0593, -0.0510, ..., -0.1053, -0.0169, 0.0733]], + device='cuda:0'), grad: tensor([[ 2.8706e-04, 7.1287e-05, 4.9710e-05, ..., 2.0075e-04, + 1.6665e-04, 1.9789e-04], + [ 1.4000e-05, 1.3399e-03, 6.2361e-06, ..., 1.5268e-03, + 1.9014e-04, 3.6430e-04], + [ 1.1253e-04, 1.4079e-04, -1.1063e-04, ..., 3.4571e-04, + -3.7823e-03, -1.9407e-03], + ..., + [-1.0586e-04, -6.5470e-04, 2.1473e-05, ..., 1.7614e-03, + 4.6158e-04, -1.9833e-05], + [ 1.7965e-04, 6.1750e-05, 1.2374e-04, ..., 2.5153e-04, + 1.4582e-03, 2.2876e-04], + [ 1.2779e-04, 1.5802e-03, 6.0618e-05, ..., 2.5902e-03, + 4.4489e-04, 1.3094e-03]], device='cuda:0') +Epoch 286, bias, value: tensor([-0.0485, 0.0185, 0.0054, -0.0087, -0.0076, 0.0070, -0.0164, 0.0352, + -0.0175, 0.0136], device='cuda:0'), grad: tensor([ 0.0084, 0.0195, -0.0158, -0.0046, -0.0190, 0.0107, -0.0210, 0.0148, + 0.0145, -0.0077], device='cuda:0') +100 +0.0001 +changing lr +epoch 285, time 215.29, cls_loss 0.5383 cls_loss_mapping 0.0033 cls_loss_causal 0.4671 re_mapping 0.0106 re_causal 0.0240 /// teacc 98.88 lr 0.00010000 +Epoch 287, weight, value: tensor([[-0.1869, -0.0195, 0.0237, ..., -0.0262, -0.0723, -0.1257], + [-0.0431, -0.0943, 0.0285, ..., 0.0391, -0.0148, -0.0961], + [-0.0585, -0.0670, -0.1080, ..., 0.0445, -0.0437, -0.0992], + ..., + [-0.0622, 0.0309, 0.0257, ..., 0.0295, -0.0476, -0.0887], + [-0.0751, -0.0064, 0.0243, ..., 0.0582, -0.0410, -0.1542], + [ 0.0589, 0.0594, -0.0526, ..., -0.1079, -0.0162, 0.0729]], + device='cuda:0'), grad: tensor([[ 4.0084e-05, 1.9855e-03, 7.1955e-04, ..., 4.5204e-03, + 5.5361e-04, 1.0145e-04], + [ 6.3069e-06, 1.7107e-04, -1.7920e-03, ..., -1.3237e-03, + -6.0425e-03, 3.9339e-05], + [ 1.0520e-05, -2.2340e-04, 6.1369e-04, ..., -1.2147e-04, + 4.8733e-04, -4.8113e-04], + ..., + [ 5.2869e-05, 9.6655e-04, -3.1967e-03, ..., -2.8706e-03, + 6.0225e-04, 6.0380e-05], + [ 4.3392e-04, 5.1832e-04, 9.8801e-04, ..., 1.3561e-03, + 7.1573e-04, 3.7265e-04], + [ 5.0497e-04, 5.3978e-03, 4.9400e-04, ..., 4.2953e-03, + 7.9060e-04, 3.6860e-04]], device='cuda:0') +Epoch 287, bias, value: tensor([-0.0492, 0.0171, 0.0062, -0.0082, -0.0073, 0.0068, -0.0161, 0.0352, + -0.0166, 0.0131], device='cuda:0'), grad: tensor([ 0.0413, -0.0098, 0.0154, 0.0081, 0.0010, -0.0135, -0.0634, -0.0453, + 0.0277, 0.0386], device='cuda:0') +100 +0.0001 +changing lr +epoch 286, time 217.09, cls_loss 0.5132 cls_loss_mapping 0.0030 cls_loss_causal 0.4464 re_mapping 0.0101 re_causal 0.0228 /// teacc 98.88 lr 0.00010000 +Epoch 288, weight, value: tensor([[-0.1864, -0.0206, 0.0241, ..., -0.0280, -0.0727, -0.1259], + [-0.0431, -0.0945, 0.0293, ..., 0.0380, -0.0151, -0.0952], + [-0.0579, -0.0683, -0.1085, ..., 0.0437, -0.0446, -0.0998], + ..., + [-0.0632, 0.0313, 0.0259, ..., 0.0301, -0.0474, -0.0905], + [-0.0757, -0.0070, 0.0239, ..., 0.0597, -0.0406, -0.1548], + [ 0.0600, 0.0604, -0.0523, ..., -0.1080, -0.0158, 0.0736]], + device='cuda:0'), grad: tensor([[ 1.8701e-05, -7.3776e-03, 4.1500e-06, ..., -2.6894e-04, + -5.6763e-03, 2.0409e-04], + [ 7.1943e-05, 4.5323e-04, -2.3520e-04, ..., 1.3027e-03, + 6.2609e-04, 3.3045e-04], + [ 4.7016e-04, 2.5215e-03, 3.0518e-05, ..., 6.5374e-04, + 5.4626e-03, 3.7026e-04], + ..., + [ 4.7088e-05, 3.4332e-03, 2.9773e-05, ..., -6.0806e-03, + 5.2071e-03, 1.9658e-04], + [ 1.4520e-04, -4.2105e-04, 6.2823e-05, ..., 1.3742e-03, + 8.9407e-05, 4.0650e-04], + [ 1.1766e-04, 8.3494e-04, 5.6863e-05, ..., 1.2875e-03, + 1.1845e-03, 2.6917e-04]], device='cuda:0') +Epoch 288, bias, value: tensor([-0.0508, 0.0170, 0.0058, -0.0076, -0.0070, 0.0063, -0.0154, 0.0361, + -0.0172, 0.0136], device='cuda:0'), grad: tensor([-0.0153, 0.0091, 0.0148, -0.0139, -0.0260, 0.0088, 0.0077, -0.0117, + 0.0057, 0.0207], device='cuda:0') +100 +0.0001 +changing lr +epoch 287, time 223.59, cls_loss 0.5242 cls_loss_mapping 0.0030 cls_loss_causal 0.4572 re_mapping 0.0102 re_causal 0.0233 /// teacc 98.93 lr 0.00010000 +Epoch 289, weight, value: tensor([[-0.1872, -0.0196, 0.0236, ..., -0.0284, -0.0728, -0.1263], + [-0.0434, -0.0955, 0.0290, ..., 0.0384, -0.0147, -0.0937], + [-0.0579, -0.0694, -0.1092, ..., 0.0435, -0.0451, -0.1007], + ..., + [-0.0630, 0.0306, 0.0267, ..., 0.0308, -0.0484, -0.0902], + [-0.0764, -0.0058, 0.0244, ..., 0.0593, -0.0411, -0.1545], + [ 0.0593, 0.0606, -0.0521, ..., -0.1087, -0.0151, 0.0729]], + device='cuda:0'), grad: tensor([[ 6.8247e-06, 3.9935e-04, 3.6340e-06, ..., 1.8522e-05, + 6.4278e-04, 4.7624e-05], + [ 1.5342e-04, 1.3361e-03, 3.6508e-07, ..., 3.1233e-05, + 4.9591e-04, 1.4329e-04], + [ 1.9855e-03, 5.6887e-04, 3.9697e-05, ..., 4.8131e-06, + -4.6234e-03, 1.6317e-03], + ..., + [ 2.6369e-04, -1.7281e-03, 1.4317e-04, ..., 1.5581e-04, + -7.1144e-04, 3.4952e-04], + [ 9.8825e-05, -2.0676e-03, -7.3314e-05, ..., 7.8827e-06, + 4.2844e-04, 2.3377e-04], + [ 5.5504e-04, 3.9005e-03, 1.0958e-03, ..., 1.8787e-04, + 5.0879e-04, 1.3170e-03]], device='cuda:0') +Epoch 289, bias, value: tensor([-0.0508, 0.0184, 0.0058, -0.0083, -0.0072, 0.0067, -0.0160, 0.0366, + -0.0171, 0.0125], device='cuda:0'), grad: tensor([ 0.0198, -0.0068, 0.0039, -0.0138, -0.0248, 0.0222, -0.0367, -0.0017, + 0.0185, 0.0193], device='cuda:0') +100 +0.0001 +changing lr +epoch 288, time 222.80, cls_loss 0.5509 cls_loss_mapping 0.0038 cls_loss_causal 0.4724 re_mapping 0.0096 re_causal 0.0232 /// teacc 98.79 lr 0.00010000 +Epoch 290, weight, value: tensor([[-0.1875, -0.0194, 0.0234, ..., -0.0280, -0.0725, -0.1276], + [-0.0452, -0.0960, 0.0287, ..., 0.0386, -0.0143, -0.0920], + [-0.0574, -0.0693, -0.1097, ..., 0.0431, -0.0451, -0.1018], + ..., + [-0.0617, 0.0313, 0.0273, ..., 0.0305, -0.0500, -0.0896], + [-0.0771, -0.0056, 0.0245, ..., 0.0601, -0.0397, -0.1549], + [ 0.0592, 0.0607, -0.0533, ..., -0.1081, -0.0157, 0.0725]], + device='cuda:0'), grad: tensor([[ 8.4209e-04, 6.8918e-08, 8.5402e-04, ..., 8.4352e-04, + 5.1975e-04, 5.3978e-04], + [ 1.2243e-04, 1.9558e-07, -2.3422e-03, ..., -2.2430e-03, + -1.0586e-03, -2.1271e-06], + [ 6.4468e-03, 4.9621e-06, 6.0129e-04, ..., 8.0585e-04, + 3.5238e-04, 9.1457e-04], + ..., + [-1.6998e-02, 1.2964e-06, 5.6267e-04, ..., 7.8964e-04, + 2.4629e-04, -1.3359e-05], + [ 5.9843e-04, -2.0653e-05, 4.3488e-04, ..., -1.8978e-03, + 2.3675e-04, 3.6120e-04], + [ 6.6757e-03, 4.6641e-06, 7.8154e-04, ..., 9.9659e-04, + 6.0749e-04, 3.2425e-04]], device='cuda:0') +Epoch 290, bias, value: tensor([-0.0514, 0.0189, 0.0056, -0.0085, -0.0071, 0.0063, -0.0153, 0.0364, + -0.0167, 0.0125], device='cuda:0'), grad: tensor([ 0.0150, -0.0292, 0.0260, 0.0110, 0.0296, 0.0155, -0.0146, 0.0044, + -0.0200, -0.0378], device='cuda:0') +100 +0.0001 +changing lr +epoch 289, time 223.38, cls_loss 0.5369 cls_loss_mapping 0.0039 cls_loss_causal 0.4663 re_mapping 0.0103 re_causal 0.0238 /// teacc 98.80 lr 0.00010000 +Epoch 291, weight, value: tensor([[-0.1872, -0.0187, 0.0241, ..., -0.0275, -0.0727, -0.1273], + [-0.0458, -0.0974, 0.0294, ..., 0.0398, -0.0151, -0.0914], + [-0.0582, -0.0696, -0.1094, ..., 0.0425, -0.0439, -0.1014], + ..., + [-0.0602, 0.0313, 0.0272, ..., 0.0320, -0.0507, -0.0900], + [-0.0770, -0.0055, 0.0246, ..., 0.0594, -0.0383, -0.1549], + [ 0.0585, 0.0602, -0.0529, ..., -0.1094, -0.0168, 0.0720]], + device='cuda:0'), grad: tensor([[ 2.5105e-04, 5.9605e-05, -5.9557e-04, ..., -4.5228e-04, + 5.6696e-04, 1.2495e-05], + [ 8.8755e-07, 3.8028e-05, 5.7983e-04, ..., -2.6875e-03, + 1.0090e-03, 1.4789e-06], + [ 5.4091e-05, 7.9012e-04, 7.2241e-04, ..., 1.3847e-03, + 4.2057e-04, 3.0696e-05], + ..., + [ 3.0971e-04, 2.2030e-04, 4.5538e-04, ..., 1.5135e-03, + 1.0319e-03, 6.6471e-04], + [ 2.1338e-04, -1.2569e-03, 4.5681e-04, ..., -4.4479e-03, + -4.2191e-03, 3.0446e-04], + [-9.0265e-04, -6.0225e-04, 3.8505e-04, ..., 9.9373e-04, + 7.8583e-04, -1.4687e-03]], device='cuda:0') +Epoch 291, bias, value: tensor([-0.0523, 0.0189, 0.0068, -0.0082, -0.0071, 0.0057, -0.0160, 0.0364, + -0.0161, 0.0123], device='cuda:0'), grad: tensor([ 0.0091, 0.0043, -0.0147, 0.0158, 0.0223, 0.0167, -0.0141, -0.0413, + -0.0117, 0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 290, time 223.76, cls_loss 0.5483 cls_loss_mapping 0.0041 cls_loss_causal 0.4836 re_mapping 0.0105 re_causal 0.0236 /// teacc 98.78 lr 0.00010000 +Epoch 292, weight, value: tensor([[-0.1865, -0.0175, 0.0247, ..., -0.0274, -0.0722, -0.1278], + [-0.0463, -0.0974, 0.0291, ..., 0.0398, -0.0161, -0.0918], + [-0.0590, -0.0704, -0.1092, ..., 0.0432, -0.0434, -0.1012], + ..., + [-0.0606, 0.0306, 0.0273, ..., 0.0306, -0.0532, -0.0906], + [-0.0762, -0.0061, 0.0240, ..., 0.0599, -0.0394, -0.1552], + [ 0.0584, 0.0603, -0.0531, ..., -0.1096, -0.0154, 0.0726]], + device='cuda:0'), grad: tensor([[ 2.8118e-05, 1.2852e-07, 6.5923e-05, ..., 3.3402e-04, + 5.3978e-04, 2.2256e-04], + [ 1.5900e-05, 1.4035e-06, 1.1998e-04, ..., 5.2452e-04, + 8.6689e-04, 1.4329e-04], + [ 1.5408e-05, 6.1207e-06, 6.3181e-04, ..., -8.7500e-04, + -6.3419e-04, 3.7956e-04], + ..., + [ 7.6830e-05, 1.0179e-06, 1.4162e-04, ..., 1.2312e-03, + 1.8330e-03, 2.5415e-04], + [ 1.6069e-04, -7.1824e-06, -2.1229e-03, ..., -4.7135e-04, + -6.8626e-03, 2.2030e-04], + [-3.2163e-04, 8.3148e-06, 3.2854e-04, ..., 7.7724e-04, + 1.6975e-03, 1.4186e-04]], device='cuda:0') +Epoch 292, bias, value: tensor([-0.0518, 0.0182, 0.0069, -0.0082, -0.0071, 0.0048, -0.0152, 0.0357, + -0.0155, 0.0127], device='cuda:0'), grad: tensor([-0.0259, 0.0230, 0.0157, 0.0319, -0.0395, -0.0595, 0.0247, 0.0025, + -0.0017, 0.0289], device='cuda:0') +100 +0.0001 +changing lr +epoch 291, time 221.68, cls_loss 0.5284 cls_loss_mapping 0.0037 cls_loss_causal 0.4641 re_mapping 0.0102 re_causal 0.0241 /// teacc 98.80 lr 0.00010000 +Epoch 293, weight, value: tensor([[-0.1877, -0.0188, 0.0238, ..., -0.0285, -0.0720, -0.1278], + [-0.0468, -0.0965, 0.0291, ..., 0.0398, -0.0153, -0.0931], + [-0.0598, -0.0712, -0.1086, ..., 0.0426, -0.0427, -0.1009], + ..., + [-0.0607, 0.0307, 0.0284, ..., 0.0311, -0.0534, -0.0912], + [-0.0751, -0.0048, 0.0227, ..., 0.0600, -0.0406, -0.1550], + [ 0.0581, 0.0605, -0.0528, ..., -0.1090, -0.0160, 0.0725]], + device='cuda:0'), grad: tensor([[-2.6226e-06, 5.2154e-08, -2.1141e-07, ..., -5.6610e-03, + 9.3132e-08, 9.2462e-06], + [ 1.5087e-07, 2.3432e-06, -1.7546e-06, ..., 6.7472e-05, + 4.6082e-06, 3.0398e-06], + [ 6.6217e-07, 6.2957e-06, 9.4064e-08, ..., 2.8687e-03, + 1.2383e-05, 9.7975e-06], + ..., + [-2.2966e-06, -4.0799e-05, 5.6345e-07, ..., 4.6587e-04, + -7.8201e-05, 9.2909e-06], + [ 3.3993e-07, 5.0738e-06, 4.4610e-07, ..., -7.7581e-04, + 9.9838e-06, -1.2598e-03], + [ 2.9713e-05, 8.9407e-07, 9.1553e-05, ..., 1.1563e-04, + 5.4277e-06, -3.5167e-05]], device='cuda:0') +Epoch 293, bias, value: tensor([-0.0518, 0.0184, 0.0063, -0.0083, -0.0071, 0.0054, -0.0146, 0.0358, + -0.0161, 0.0126], device='cuda:0'), grad: tensor([-0.0182, 0.0102, 0.0247, -0.0395, 0.0097, 0.0230, 0.0176, -0.0182, + 0.0090, -0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 292, time 220.63, cls_loss 0.5578 cls_loss_mapping 0.0033 cls_loss_causal 0.4917 re_mapping 0.0099 re_causal 0.0239 /// teacc 98.81 lr 0.00010000 +Epoch 294, weight, value: tensor([[-0.1872, -0.0190, 0.0237, ..., -0.0290, -0.0733, -0.1267], + [-0.0492, -0.0969, 0.0287, ..., 0.0394, -0.0158, -0.0930], + [-0.0602, -0.0713, -0.1075, ..., 0.0433, -0.0432, -0.1001], + ..., + [-0.0611, 0.0310, 0.0290, ..., 0.0315, -0.0529, -0.0910], + [-0.0750, -0.0049, 0.0218, ..., 0.0600, -0.0394, -0.1555], + [ 0.0580, 0.0603, -0.0523, ..., -0.1096, -0.0173, 0.0722]], + device='cuda:0'), grad: tensor([[ 1.1504e-05, 1.4193e-06, 3.5852e-05, ..., 9.3937e-04, + 1.0281e-03, 5.6103e-06], + [ 3.6359e-04, 3.5204e-07, 1.3947e-04, ..., -8.6308e-04, + -4.1847e-03, 2.6152e-06], + [ 1.6987e-04, 1.3447e-04, 1.7858e-04, ..., 1.9331e-03, + 1.4744e-03, 1.7181e-05], + ..., + [ 2.3766e-03, 7.5158e-07, 9.8228e-04, ..., 3.4008e-03, + 4.0627e-03, 1.1474e-04], + [ 8.9169e-04, -2.7490e-04, 3.0422e-04, ..., -1.4391e-03, + 1.1463e-03, 7.9346e-04], + [-3.6564e-03, 1.3217e-05, 9.8765e-05, ..., 1.2159e-03, + 1.0700e-03, -3.4428e-04]], device='cuda:0') +Epoch 294, bias, value: tensor([-0.0522, 0.0176, 0.0069, -0.0079, -0.0061, 0.0048, -0.0151, 0.0358, + -0.0162, 0.0128], device='cuda:0'), grad: tensor([ 0.0255, -0.0120, 0.0202, -0.0052, -0.0292, 0.0147, -0.0333, 0.0470, + -0.0117, -0.0160], device='cuda:0') +100 +0.0001 +changing lr +epoch 293, time 219.96, cls_loss 0.5621 cls_loss_mapping 0.0055 cls_loss_causal 0.4946 re_mapping 0.0099 re_causal 0.0232 /// teacc 98.60 lr 0.00010000 +Epoch 295, weight, value: tensor([[-0.1874, -0.0191, 0.0233, ..., -0.0285, -0.0734, -0.1265], + [-0.0494, -0.0969, 0.0304, ..., 0.0389, -0.0152, -0.0935], + [-0.0596, -0.0720, -0.1068, ..., 0.0438, -0.0434, -0.0998], + ..., + [-0.0601, 0.0307, 0.0297, ..., 0.0313, -0.0528, -0.0915], + [-0.0751, -0.0043, 0.0221, ..., 0.0600, -0.0389, -0.1561], + [ 0.0580, 0.0607, -0.0526, ..., -0.1095, -0.0177, 0.0728]], + device='cuda:0'), grad: tensor([[ 5.3883e-05, 2.7940e-09, 9.4318e-04, ..., 8.6248e-05, + -3.3545e-04, 4.9099e-06], + [ 6.1393e-05, 9.3132e-09, 2.5296e-04, ..., 1.4877e-04, + 1.4229e-03, 1.1563e-05], + [-1.5707e-03, 7.3854e-07, -2.9850e-03, ..., -7.7629e-03, + 8.9407e-04, 1.7688e-05], + ..., + [-6.6986e-03, 1.2014e-07, -8.0795e-03, ..., 2.7161e-03, + -1.1749e-03, -5.7526e-03], + [ 1.4753e-03, -2.2650e-06, 3.6221e-03, ..., 1.9569e-03, + 4.5052e-03, 2.2733e-04], + [ 5.9624e-03, 9.4809e-07, 3.2654e-03, ..., 2.2793e-03, + -1.2131e-02, 5.3940e-03]], device='cuda:0') +Epoch 295, bias, value: tensor([-0.0501, 0.0184, 0.0058, -0.0094, -0.0064, 0.0050, -0.0159, 0.0366, + -0.0154, 0.0120], device='cuda:0'), grad: tensor([ 0.0263, 0.0090, -0.0314, 0.0144, 0.0205, -0.0127, -0.0139, -0.0413, + 0.0228, 0.0061], device='cuda:0') +100 +0.0001 +changing lr +epoch 294, time 222.79, cls_loss 0.5379 cls_loss_mapping 0.0036 cls_loss_causal 0.4739 re_mapping 0.0099 re_causal 0.0233 /// teacc 98.70 lr 0.00010000 +Epoch 296, weight, value: tensor([[-0.1875, -0.0190, 0.0224, ..., -0.0282, -0.0746, -0.1268], + [-0.0486, -0.0971, 0.0311, ..., 0.0390, -0.0147, -0.0933], + [-0.0587, -0.0719, -0.1073, ..., 0.0438, -0.0423, -0.1000], + ..., + [-0.0612, 0.0302, 0.0287, ..., 0.0300, -0.0535, -0.0919], + [-0.0754, -0.0046, 0.0227, ..., 0.0604, -0.0386, -0.1565], + [ 0.0594, 0.0615, -0.0518, ..., -0.1098, -0.0171, 0.0741]], + device='cuda:0'), grad: tensor([[ 1.2503e-03, 5.6833e-05, 4.5109e-04, ..., 1.1625e-03, + 2.4109e-03, 2.6170e-07], + [ 4.0740e-05, -1.0595e-03, 2.2149e-04, ..., 3.7117e-03, + -3.7646e-04, 4.4890e-07], + [ 3.5906e-04, 2.6298e-04, 9.3937e-04, ..., 1.8806e-03, + 2.0790e-03, 6.4261e-07], + ..., + [-4.2915e-04, 6.9904e-04, -1.0920e-03, ..., -3.5620e-04, + 3.2444e-03, 7.3433e-05], + [-7.9193e-03, -2.2531e-04, 8.0442e-04, ..., 1.7986e-03, + -2.6512e-04, -3.0472e-02], + [-1.1883e-03, 7.0691e-05, -1.4877e-03, ..., -2.7752e-03, + -5.2490e-03, -1.0467e-04]], device='cuda:0') +Epoch 296, bias, value: tensor([-0.0509, 0.0179, 0.0065, -0.0080, -0.0055, 0.0047, -0.0173, 0.0362, + -0.0144, 0.0112], device='cuda:0'), grad: tensor([ 0.0195, 0.0146, 0.0246, 0.0415, -0.0762, 0.0234, -0.0207, 0.0178, + -0.0212, -0.0234], device='cuda:0') +100 +0.0001 +changing lr +epoch 295, time 224.00, cls_loss 0.5361 cls_loss_mapping 0.0048 cls_loss_causal 0.4633 re_mapping 0.0105 re_causal 0.0229 /// teacc 98.72 lr 0.00010000 +Epoch 297, weight, value: tensor([[-0.1886, -0.0189, 0.0217, ..., -0.0284, -0.0751, -0.1269], + [-0.0482, -0.0973, 0.0336, ..., 0.0408, -0.0130, -0.0925], + [-0.0591, -0.0714, -0.1073, ..., 0.0437, -0.0426, -0.0996], + ..., + [-0.0617, 0.0289, 0.0279, ..., 0.0285, -0.0543, -0.0928], + [-0.0742, -0.0056, 0.0223, ..., 0.0617, -0.0390, -0.1551], + [ 0.0584, 0.0626, -0.0529, ..., -0.1100, -0.0176, 0.0739]], + device='cuda:0'), grad: tensor([[-7.5483e-04, -1.2760e-03, -6.1798e-03, ..., -6.7787e-03, + -2.6684e-03, -8.3542e-04], + [ 6.0320e-05, 1.4496e-04, -5.3940e-03, ..., -9.5224e-04, + -3.3360e-03, 6.8486e-05], + [ 1.7095e-04, 1.6785e-04, 1.3018e-03, ..., 9.4843e-04, + 6.4659e-04, 1.1903e-04], + ..., + [-5.1155e-03, 6.5148e-05, 1.0881e-03, ..., 5.9938e-04, + 5.9843e-04, -2.2621e-03], + [ 3.9220e-04, 9.2363e-04, 3.8242e-03, ..., 3.0155e-03, + 1.7862e-03, 6.2752e-04], + [ 5.2185e-03, 4.1175e-04, 1.1539e-03, ..., 7.7772e-04, + 5.7220e-04, 2.3079e-03]], device='cuda:0') +Epoch 297, bias, value: tensor([-0.0506, 0.0193, 0.0058, -0.0076, -0.0051, 0.0042, -0.0168, 0.0358, + -0.0146, 0.0101], device='cuda:0'), grad: tensor([-0.0464, -0.0315, -0.0163, 0.0345, 0.0091, 0.0046, 0.0070, 0.0052, + 0.0194, 0.0145], device='cuda:0') +100 +0.0001 +changing lr +epoch 296, time 223.50, cls_loss 0.5391 cls_loss_mapping 0.0040 cls_loss_causal 0.4681 re_mapping 0.0104 re_causal 0.0233 /// teacc 98.88 lr 0.00010000 +Epoch 298, weight, value: tensor([[-0.1887, -0.0196, 0.0228, ..., -0.0289, -0.0751, -0.1271], + [-0.0480, -0.0970, 0.0346, ..., 0.0405, -0.0126, -0.0929], + [-0.0602, -0.0717, -0.1075, ..., 0.0438, -0.0419, -0.1002], + ..., + [-0.0631, 0.0303, 0.0277, ..., 0.0295, -0.0537, -0.0939], + [-0.0721, -0.0044, 0.0227, ..., 0.0620, -0.0392, -0.1555], + [ 0.0602, 0.0622, -0.0528, ..., -0.1101, -0.0189, 0.0756]], + device='cuda:0'), grad: tensor([[6.1631e-05, 1.3329e-05, 7.9393e-05, ..., 2.3496e-04, 2.4498e-05, + 6.8784e-05], + [1.5959e-05, 2.4159e-06, 6.6310e-06, ..., 3.2463e-03, 2.8327e-05, + 1.7881e-05], + [1.1854e-03, 6.7174e-05, 1.1349e-03, ..., 1.4095e-03, 1.3638e-04, + 1.2417e-03], + ..., + [2.4605e-04, 2.6917e-04, 2.3389e-04, ..., 1.3189e-03, 3.2210e-04, + 3.6979e-04], + [3.4547e-04, 2.2918e-05, 3.2043e-04, ..., 9.0170e-04, 7.3254e-05, + 3.5644e-04], + [1.9819e-05, 1.2875e-04, 9.8884e-05, ..., 5.2786e-04, 1.4615e-04, + 1.0049e-04]], device='cuda:0') +Epoch 298, bias, value: tensor([-0.0522, 0.0200, 0.0068, -0.0076, -0.0061, 0.0040, -0.0162, 0.0366, + -0.0147, 0.0097], device='cuda:0'), grad: tensor([ 0.0146, 0.0244, -0.0305, 0.0321, 0.0033, -0.0323, 0.0021, 0.0280, + -0.0406, -0.0011], device='cuda:0') +100 +0.0001 +changing lr +epoch 297, time 217.27, cls_loss 0.5486 cls_loss_mapping 0.0042 cls_loss_causal 0.4831 re_mapping 0.0098 re_causal 0.0232 /// teacc 98.82 lr 0.00010000 +Epoch 299, weight, value: tensor([[-0.1881, -0.0193, 0.0231, ..., -0.0297, -0.0752, -0.1270], + [-0.0480, -0.0975, 0.0341, ..., 0.0405, -0.0134, -0.0932], + [-0.0614, -0.0718, -0.1074, ..., 0.0437, -0.0421, -0.1015], + ..., + [-0.0633, 0.0305, 0.0284, ..., 0.0286, -0.0534, -0.0926], + [-0.0724, -0.0050, 0.0223, ..., 0.0622, -0.0388, -0.1552], + [ 0.0607, 0.0627, -0.0529, ..., -0.1093, -0.0194, 0.0756]], + device='cuda:0'), grad: tensor([[ 3.5793e-05, 4.8447e-03, -1.3733e-04, ..., 2.9063e-04, + 1.5223e-04, 2.7046e-05], + [ 1.4104e-05, 2.6971e-06, 1.5175e-04, ..., 8.3160e-04, + 4.0340e-04, 4.5709e-06], + [-1.2994e-04, 1.4505e-03, 2.1362e-04, ..., 7.2289e-04, + 3.8242e-04, 6.0946e-05], + ..., + [ 4.0007e-04, 1.8752e-04, 1.0794e-04, ..., 4.5228e-04, + 2.5249e-04, 5.4121e-04], + [ 7.2479e-05, 3.3035e-03, 9.2566e-05, ..., 1.0099e-03, + 3.0255e-04, 1.7369e-04], + [-5.9366e-04, -9.6207e-03, 4.8876e-05, ..., -2.1992e-03, + -2.5988e-04, -1.0138e-03]], device='cuda:0') +Epoch 299, bias, value: tensor([-0.0513, 0.0206, 0.0056, -0.0085, -0.0046, 0.0051, -0.0166, 0.0358, + -0.0150, 0.0096], device='cuda:0'), grad: tensor([ 2.9510e-02, 2.2552e-02, 4.9683e-02, -4.4861e-02, -1.0498e-02, + 2.1114e-03, 2.4109e-02, 1.9302e-02, 6.0201e-05, -9.1980e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 298, time 215.13, cls_loss 0.5370 cls_loss_mapping 0.0041 cls_loss_causal 0.4698 re_mapping 0.0098 re_causal 0.0228 /// teacc 98.78 lr 0.00010000 +Epoch 300, weight, value: tensor([[-0.1889, -0.0195, 0.0231, ..., -0.0308, -0.0766, -0.1269], + [-0.0500, -0.0954, 0.0343, ..., 0.0414, -0.0132, -0.0931], + [-0.0613, -0.0718, -0.1073, ..., 0.0436, -0.0423, -0.1018], + ..., + [-0.0622, 0.0312, 0.0296, ..., 0.0293, -0.0530, -0.0934], + [-0.0740, -0.0048, 0.0210, ..., 0.0623, -0.0390, -0.1561], + [ 0.0610, 0.0621, -0.0538, ..., -0.1111, -0.0188, 0.0776]], + device='cuda:0'), grad: tensor([[ 1.9610e-05, -1.8978e-04, -2.9469e-03, ..., -8.8549e-04, + -1.4591e-03, -1.0643e-03], + [ 5.0735e-04, 4.3392e-05, -2.0523e-03, ..., 2.0523e-03, + -1.9855e-03, 1.9455e-04], + [-7.8917e-04, 1.7285e-04, -3.5954e-03, ..., -5.1079e-03, + -2.4128e-03, 3.2854e-04], + ..., + [ 4.1902e-05, 2.1565e-04, 1.6136e-03, ..., -2.0301e-04, + 8.2827e-04, 2.6941e-04], + [ 4.4733e-05, -4.2987e-04, 1.9894e-03, ..., 1.5354e-04, + 9.6846e-04, 5.1641e-04], + [ 2.0117e-05, 1.3925e-05, 1.0042e-03, ..., -3.8099e-04, + 6.7997e-04, -5.5084e-03]], device='cuda:0') +Epoch 300, bias, value: tensor([-0.0532, 0.0205, 0.0065, -0.0099, -0.0038, 0.0048, -0.0161, 0.0363, + -0.0160, 0.0114], device='cuda:0'), grad: tensor([-0.0105, 0.0018, -0.0087, 0.0152, -0.0170, 0.0116, 0.0274, -0.0151, + 0.0150, -0.0197], device='cuda:0') +100 +0.0001 +changing lr +epoch 299, time 214.57, cls_loss 0.5294 cls_loss_mapping 0.0036 cls_loss_causal 0.4671 re_mapping 0.0098 re_causal 0.0228 /// teacc 98.76 lr 0.00010000 +Epoch 301, weight, value: tensor([[-0.1894, -0.0198, 0.0232, ..., -0.0315, -0.0780, -0.1269], + [-0.0503, -0.0963, 0.0354, ..., 0.0404, -0.0133, -0.0932], + [-0.0615, -0.0739, -0.1064, ..., 0.0439, -0.0425, -0.1016], + ..., + [-0.0635, 0.0308, 0.0280, ..., 0.0291, -0.0542, -0.0929], + [-0.0732, -0.0039, 0.0223, ..., 0.0621, -0.0392, -0.1551], + [ 0.0612, 0.0619, -0.0524, ..., -0.1104, -0.0171, 0.0760]], + device='cuda:0'), grad: tensor([[ 5.7650e-04, 1.9884e-04, 2.4724e-04, ..., -5.9986e-04, + -6.7787e-03, 3.0965e-05], + [-1.7653e-03, 8.2493e-05, -4.6234e-03, ..., -2.3499e-03, + -7.8392e-04, -7.4148e-04], + [ 1.2094e-04, -1.3294e-03, 1.9288e-04, ..., -4.1199e-03, + -1.2085e-02, 3.0100e-05], + ..., + [ 1.9455e-04, 1.1797e-03, 1.3514e-03, ..., 2.0599e-03, + 5.8136e-03, 6.7139e-04], + [-8.6546e-05, 1.5199e-04, -2.6488e-04, ..., 3.5000e-04, + 1.5574e-03, 8.0764e-05], + [ 9.9850e-04, 3.2654e-02, 1.3218e-03, ..., 1.3494e-03, + 3.5477e-03, 1.7075e-02]], device='cuda:0') +Epoch 301, bias, value: tensor([-0.0523, 0.0195, 0.0052, -0.0100, -0.0030, 0.0050, -0.0161, 0.0361, + -0.0153, 0.0113], device='cuda:0'), grad: tensor([-0.0446, -0.0317, -0.0206, 0.0198, -0.0310, 0.0208, -0.0168, 0.0237, + 0.0133, 0.0670], device='cuda:0') +100 +0.0001 +changing lr +epoch 300, time 214.65, cls_loss 0.5254 cls_loss_mapping 0.0032 cls_loss_causal 0.4577 re_mapping 0.0101 re_causal 0.0224 /// teacc 98.85 lr 0.00010000 +Epoch 302, weight, value: tensor([[-0.1898, -0.0195, 0.0231, ..., -0.0311, -0.0794, -0.1261], + [-0.0512, -0.0963, 0.0335, ..., 0.0405, -0.0136, -0.0934], + [-0.0610, -0.0741, -0.1061, ..., 0.0443, -0.0422, -0.1020], + ..., + [-0.0642, 0.0312, 0.0280, ..., 0.0276, -0.0545, -0.0942], + [-0.0732, -0.0044, 0.0224, ..., 0.0628, -0.0381, -0.1557], + [ 0.0603, 0.0616, -0.0517, ..., -0.1100, -0.0169, 0.0757]], + device='cuda:0'), grad: tensor([[ 9.5181e-07, 1.0669e-04, 2.4891e-04, ..., 4.3273e-04, + 1.1301e-03, 2.9616e-07], + [ 4.7684e-07, 2.4354e-04, -1.8778e-03, ..., -1.7033e-03, + -3.6545e-03, 1.0990e-05], + [ 1.8299e-05, 6.2704e-05, 3.7503e-04, ..., 5.4169e-04, + 1.4734e-03, 1.5005e-05], + ..., + [ 1.0103e-04, 8.4400e-05, 1.4806e-04, ..., -3.4962e-03, + -3.7479e-03, 1.2374e-04], + [-2.5892e-04, 1.4038e-03, 2.2328e-04, ..., 5.1832e-04, + -2.0618e-03, 2.7084e-04], + [ 2.6468e-06, -2.3819e-02, 1.2577e-04, ..., 1.5602e-03, + 1.9398e-03, -5.2786e-04]], device='cuda:0') +Epoch 302, bias, value: tensor([-0.0532, 0.0192, 0.0057, -0.0099, -0.0038, 0.0057, -0.0158, 0.0351, + -0.0150, 0.0126], device='cuda:0'), grad: tensor([-0.0176, -0.0349, 0.0150, 0.0181, 0.0426, 0.0143, -0.0146, -0.0017, + -0.0112, -0.0099], device='cuda:0') +100 +0.0001 +changing lr +epoch 301, time 214.78, cls_loss 0.5361 cls_loss_mapping 0.0035 cls_loss_causal 0.4792 re_mapping 0.0098 re_causal 0.0227 /// teacc 98.90 lr 0.00010000 +Epoch 303, weight, value: tensor([[-0.1902, -0.0196, 0.0239, ..., -0.0309, -0.0795, -0.1263], + [-0.0524, -0.0968, 0.0331, ..., 0.0408, -0.0149, -0.0933], + [-0.0610, -0.0741, -0.1059, ..., 0.0443, -0.0401, -0.1026], + ..., + [-0.0654, 0.0306, 0.0275, ..., 0.0271, -0.0553, -0.0945], + [-0.0739, -0.0043, 0.0230, ..., 0.0629, -0.0385, -0.1567], + [ 0.0605, 0.0621, -0.0518, ..., -0.1087, -0.0180, 0.0770]], + device='cuda:0'), grad: tensor([[ 6.6876e-05, 6.3467e-04, 9.5320e-04, ..., 4.6349e-04, + -1.5154e-03, 7.5400e-06], + [ 1.1653e-04, 1.2436e-03, 1.2102e-03, ..., 1.9274e-03, + -5.5122e-03, 7.5176e-06], + [ 4.2737e-05, 2.3823e-03, -2.8915e-03, ..., 1.5707e-03, + -2.0771e-03, 9.2924e-05], + ..., + [-7.7934e-03, -4.7874e-03, -1.1339e-03, ..., -1.2253e-02, + -1.7691e-03, -1.6146e-03], + [ 2.5845e-04, 1.3847e-03, -5.2109e-03, ..., 1.2407e-03, + 1.8711e-03, 4.1783e-05], + [ 1.8415e-03, 7.3481e-04, 1.7643e-03, ..., 1.6174e-03, + 1.6890e-03, 3.6383e-04]], device='cuda:0') +Epoch 303, bias, value: tensor([-0.0529, 0.0192, 0.0069, -0.0107, -0.0038, 0.0050, -0.0148, 0.0348, + -0.0151, 0.0121], device='cuda:0'), grad: tensor([-0.0179, -0.0053, 0.0127, 0.0289, -0.0083, 0.0372, -0.0071, -0.0334, + -0.0050, -0.0019], device='cuda:0') +100 +0.0001 +changing lr +epoch 302, time 214.74, cls_loss 0.5180 cls_loss_mapping 0.0043 cls_loss_causal 0.4589 re_mapping 0.0098 re_causal 0.0222 /// teacc 98.72 lr 0.00010000 +Epoch 304, weight, value: tensor([[-0.1914, -0.0183, 0.0241, ..., -0.0314, -0.0788, -0.1276], + [-0.0530, -0.0973, 0.0341, ..., 0.0406, -0.0133, -0.0936], + [-0.0597, -0.0751, -0.1055, ..., 0.0446, -0.0397, -0.1030], + ..., + [-0.0645, 0.0302, 0.0285, ..., 0.0275, -0.0548, -0.0945], + [-0.0738, -0.0036, 0.0233, ..., 0.0632, -0.0384, -0.1571], + [ 0.0609, 0.0620, -0.0515, ..., -0.1088, -0.0182, 0.0778]], + device='cuda:0'), grad: tensor([[ 2.1780e-04, 5.2989e-05, 5.9605e-04, ..., 5.7030e-04, + 5.3883e-04, 7.5197e-04], + [ 2.0254e-04, 8.3160e-04, 1.1282e-03, ..., 1.3885e-03, + 2.7409e-03, 1.3697e-04], + [ 1.0353e-04, 2.0802e-04, 3.1447e-04, ..., -1.2074e-03, + 2.4557e-04, 5.9652e-04], + ..., + [ 8.8310e-04, 3.2692e-03, 3.9625e-04, ..., 1.6689e-04, + -4.0131e-03, 6.4325e-04], + [-1.6308e-03, 4.9744e-03, -4.7035e-03, ..., -4.5471e-03, + -2.6646e-03, -2.7866e-03], + [-4.4179e-04, -1.0712e-02, -3.2272e-03, ..., -4.6206e-04, + -5.8746e-04, -1.0521e-02]], device='cuda:0') +Epoch 304, bias, value: tensor([-0.0524, 0.0200, 0.0067, -0.0092, -0.0038, 0.0038, -0.0148, 0.0340, + -0.0146, 0.0110], device='cuda:0'), grad: tensor([-0.0206, 0.0358, -0.0207, 0.0426, 0.0085, 0.0170, 0.0146, -0.0395, + -0.0039, -0.0338], device='cuda:0') +100 +0.0001 +changing lr +epoch 303, time 214.50, cls_loss 0.5794 cls_loss_mapping 0.0046 cls_loss_causal 0.5229 re_mapping 0.0103 re_causal 0.0245 /// teacc 98.88 lr 0.00010000 +Epoch 305, weight, value: tensor([[-0.1913, -0.0184, 0.0239, ..., -0.0319, -0.0770, -0.1292], + [-0.0522, -0.0981, 0.0338, ..., 0.0407, -0.0153, -0.0937], + [-0.0605, -0.0754, -0.1052, ..., 0.0441, -0.0395, -0.1017], + ..., + [-0.0646, 0.0305, 0.0292, ..., 0.0287, -0.0540, -0.0947], + [-0.0732, -0.0050, 0.0220, ..., 0.0632, -0.0390, -0.1574], + [ 0.0605, 0.0616, -0.0511, ..., -0.1094, -0.0182, 0.0787]], + device='cuda:0'), grad: tensor([[ 3.3569e-04, 3.7253e-09, 1.0424e-03, ..., 3.1137e-04, + 1.9236e-03, 6.2275e-04], + [ 2.8849e-04, 5.5879e-09, 6.4945e-04, ..., 1.4830e-04, + 2.5711e-03, 3.4124e-05], + [ 2.7132e-04, 5.5879e-09, -1.5564e-03, ..., -1.5278e-03, + 4.5586e-03, 3.6669e-04], + ..., + [ 1.7142e-04, -1.2014e-06, -9.0332e-03, ..., 9.2804e-05, + -2.7512e-02, 1.3781e-04], + [ 1.7414e-03, -1.0841e-06, 4.4289e-03, ..., 2.7275e-04, + 1.2535e-02, 9.4748e-04], + [ 1.0240e-04, 7.1712e-07, -1.0872e-03, ..., 1.9407e-04, + 3.7880e-03, -8.7643e-04]], device='cuda:0') +Epoch 305, bias, value: tensor([-0.0523, 0.0199, 0.0066, -0.0100, -0.0037, 0.0041, -0.0141, 0.0345, + -0.0156, 0.0114], device='cuda:0'), grad: tensor([ 0.0171, 0.0141, 0.0056, -0.0455, -0.0305, 0.0328, -0.0052, -0.0154, + 0.0114, 0.0155], device='cuda:0') +100 +0.0001 +changing lr +epoch 304, time 214.64, cls_loss 0.5419 cls_loss_mapping 0.0031 cls_loss_causal 0.4713 re_mapping 0.0108 re_causal 0.0246 /// teacc 98.85 lr 0.00010000 +Epoch 306, weight, value: tensor([[-0.1907, -0.0187, 0.0246, ..., -0.0329, -0.0776, -0.1307], + [-0.0523, -0.0985, 0.0326, ..., 0.0423, -0.0157, -0.0948], + [-0.0602, -0.0763, -0.1055, ..., 0.0433, -0.0400, -0.1003], + ..., + [-0.0644, 0.0300, 0.0292, ..., 0.0291, -0.0534, -0.0945], + [-0.0736, -0.0038, 0.0211, ..., 0.0630, -0.0374, -0.1589], + [ 0.0615, 0.0617, -0.0518, ..., -0.1087, -0.0180, 0.0790]], + device='cuda:0'), grad: tensor([[ 2.6274e-04, 3.7909e-05, 1.1921e-04, ..., 2.2340e-04, + 3.6716e-04, 4.8518e-04], + [ 2.0266e-06, -9.8801e-04, 1.3299e-05, ..., 3.2091e-04, + 4.2200e-04, 2.6241e-05], + [ 2.4587e-05, 1.1759e-03, 8.4192e-06, ..., 1.5049e-03, + 3.0541e-04, 2.9778e-04], + ..., + [ 5.2601e-05, -3.1013e-03, 3.9458e-04, ..., -3.5801e-03, + 8.9645e-04, 4.2343e-04], + [ 1.6421e-05, 2.9039e-04, 4.9710e-05, ..., -3.9768e-04, + 3.1829e-04, 2.5225e-04], + [-1.2708e-04, 2.0075e-04, -9.6750e-04, ..., -4.3035e-04, + -3.7022e-03, -9.5129e-04]], device='cuda:0') +Epoch 306, bias, value: tensor([-0.0517, 0.0200, 0.0057, -0.0093, -0.0042, 0.0042, -0.0147, 0.0352, + -0.0163, 0.0119], device='cuda:0'), grad: tensor([-0.0078, -0.0175, -0.0126, 0.0184, 0.0260, -0.0190, -0.0081, 0.0221, + 0.0179, -0.0192], device='cuda:0') +100 +0.0001 +changing lr +epoch 305, time 214.97, cls_loss 0.5196 cls_loss_mapping 0.0028 cls_loss_causal 0.4634 re_mapping 0.0104 re_causal 0.0234 /// teacc 98.75 lr 0.00010000 +Epoch 307, weight, value: tensor([[-0.1910, -0.0178, 0.0262, ..., -0.0325, -0.0773, -0.1315], + [-0.0529, -0.0973, 0.0332, ..., 0.0438, -0.0155, -0.0948], + [-0.0586, -0.0765, -0.1060, ..., 0.0427, -0.0405, -0.1006], + ..., + [-0.0632, 0.0311, 0.0288, ..., 0.0293, -0.0529, -0.0944], + [-0.0729, -0.0047, 0.0206, ..., 0.0641, -0.0359, -0.1589], + [ 0.0611, 0.0617, -0.0520, ..., -0.1095, -0.0185, 0.0787]], + device='cuda:0'), grad: tensor([[ 2.2985e-06, 8.4639e-05, -1.2014e-06, ..., 1.0433e-03, + 6.1512e-05, 1.1760e-04], + [ 2.5555e-05, 1.3132e-03, 5.6535e-05, ..., 4.0588e-03, + 8.9407e-04, 1.3018e-03], + [ 8.3625e-05, 1.4868e-03, 2.0754e-04, ..., 3.8357e-03, + 1.1575e-04, 1.2131e-03], + ..., + [-1.3947e-04, -2.4128e-04, -3.2353e-04, ..., -8.5220e-03, + 1.8501e-04, 2.0671e-04], + [ 1.6463e-04, -8.7917e-05, 7.7486e-06, ..., 3.8929e-03, + 1.5869e-03, 1.1044e-03], + [ 2.0766e-04, -4.0016e-03, 1.0002e-04, ..., 1.7118e-04, + 1.9574e-04, -4.3182e-03]], device='cuda:0') +Epoch 307, bias, value: tensor([-0.0518, 0.0207, 0.0055, -0.0099, -0.0045, 0.0051, -0.0153, 0.0363, + -0.0163, 0.0110], device='cuda:0'), grad: tensor([-0.0058, 0.0475, 0.0369, -0.0324, -0.0359, -0.0097, -0.0003, -0.0168, + 0.0144, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 306, time 214.89, cls_loss 0.5246 cls_loss_mapping 0.0041 cls_loss_causal 0.4650 re_mapping 0.0096 re_causal 0.0210 /// teacc 98.79 lr 0.00010000 +Epoch 308, weight, value: tensor([[-0.1927, -0.0171, 0.0251, ..., -0.0326, -0.0781, -0.1314], + [-0.0524, -0.0983, 0.0324, ..., 0.0440, -0.0151, -0.0952], + [-0.0583, -0.0779, -0.1068, ..., 0.0419, -0.0418, -0.1007], + ..., + [-0.0627, 0.0305, 0.0305, ..., 0.0296, -0.0516, -0.0952], + [-0.0722, -0.0055, 0.0209, ..., 0.0653, -0.0361, -0.1591], + [ 0.0601, 0.0630, -0.0518, ..., -0.1102, -0.0180, 0.0797]], + device='cuda:0'), grad: tensor([[ 2.2873e-06, 2.0981e-04, 1.7242e-03, ..., 1.2779e-03, + 1.7319e-03, 1.2851e-04], + [ 9.3505e-07, 3.9721e-04, -1.9722e-03, ..., -1.5459e-03, + -6.6280e-04, 1.2338e-04], + [ 3.0503e-05, -5.1117e-03, -1.0246e-02, ..., -6.9466e-03, + -1.4336e-02, -1.9825e-04], + ..., + [ 1.2323e-05, 5.8174e-04, 1.4839e-03, ..., 9.1171e-04, + 2.2411e-03, 5.3644e-04], + [ 1.1277e-04, 1.6117e-03, 3.4580e-03, ..., 8.4305e-04, + 1.6146e-03, 2.1057e-03], + [-9.9301e-05, 3.2258e-04, 1.4820e-03, ..., 1.2150e-03, + 1.7653e-03, 1.1082e-03]], device='cuda:0') +Epoch 308, bias, value: tensor([-0.0519, 0.0201, 0.0041, -0.0109, -0.0035, 0.0050, -0.0150, 0.0364, + -0.0152, 0.0115], device='cuda:0'), grad: tensor([ 0.0115, -0.0084, -0.0622, 0.0216, 0.0213, -0.0311, 0.0156, 0.0145, + 0.0327, -0.0155], device='cuda:0') +100 +0.0001 +changing lr +epoch 307, time 214.68, cls_loss 0.5572 cls_loss_mapping 0.0029 cls_loss_causal 0.4933 re_mapping 0.0098 re_causal 0.0222 /// teacc 98.95 lr 0.00010000 +Epoch 309, weight, value: tensor([[-0.1922, -0.0176, 0.0246, ..., -0.0332, -0.0781, -0.1316], + [-0.0534, -0.0953, 0.0316, ..., 0.0458, -0.0157, -0.0948], + [-0.0582, -0.0796, -0.1063, ..., 0.0402, -0.0418, -0.1018], + ..., + [-0.0632, 0.0309, 0.0294, ..., 0.0282, -0.0531, -0.0947], + [-0.0726, -0.0073, 0.0211, ..., 0.0657, -0.0356, -0.1594], + [ 0.0606, 0.0634, -0.0516, ..., -0.1110, -0.0185, 0.0801]], + device='cuda:0'), grad: tensor([[ 3.7998e-05, 3.0065e-04, -4.5013e-03, ..., -1.0643e-03, + -7.3624e-04, 3.1829e-05], + [ 6.5708e-04, 1.5116e-03, -1.9855e-03, ..., -2.5043e-03, + -5.6267e-03, 4.6396e-04], + [ 5.7125e-04, -4.4289e-03, -1.5221e-03, ..., 4.9591e-03, + 9.3307e-03, 4.0221e-04], + ..., + [ 3.3455e-03, 1.7681e-03, 2.0313e-03, ..., 3.7537e-03, + 1.3237e-03, 2.2488e-03], + [-7.8506e-03, -2.7256e-03, 2.8062e-04, ..., -6.6032e-03, + -2.5582e-04, -5.3978e-03], + [ 1.3456e-03, 9.4700e-04, -1.1497e-02, ..., 2.7618e-03, + 1.0157e-03, 9.4032e-04]], device='cuda:0') +Epoch 309, bias, value: tensor([-0.0513, 0.0206, 0.0042, -0.0105, -0.0036, 0.0058, -0.0156, 0.0356, + -0.0150, 0.0105], device='cuda:0'), grad: tensor([-0.0383, -0.0249, 0.0196, -0.0012, 0.0372, 0.0150, 0.0211, -0.0113, + -0.0151, -0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 308, time 214.59, cls_loss 0.5436 cls_loss_mapping 0.0029 cls_loss_causal 0.4858 re_mapping 0.0095 re_causal 0.0223 /// teacc 98.86 lr 0.00010000 +Epoch 310, weight, value: tensor([[-0.1917, -0.0146, 0.0256, ..., -0.0322, -0.0784, -0.1316], + [-0.0530, -0.0942, 0.0307, ..., 0.0450, -0.0177, -0.0928], + [-0.0585, -0.0815, -0.1071, ..., 0.0409, -0.0406, -0.1039], + ..., + [-0.0641, 0.0294, 0.0307, ..., 0.0275, -0.0521, -0.0935], + [-0.0721, -0.0060, 0.0221, ..., 0.0666, -0.0359, -0.1591], + [ 0.0606, 0.0632, -0.0519, ..., -0.1111, -0.0193, 0.0816]], + device='cuda:0'), grad: tensor([[ 1.7381e-04, 4.6849e-04, 2.6178e-04, ..., 6.9094e-04, + 8.7404e-04, 1.0359e-04], + [-1.1757e-02, -1.9989e-03, 8.5258e-04, ..., -8.0185e-03, + 3.1352e-04, -3.9520e-03], + [ 2.2447e-04, 2.1982e-04, -8.0049e-05, ..., 8.1825e-04, + 1.7673e-05, 5.2166e-04], + ..., + [ 6.6090e-04, 2.2717e-03, 1.5917e-03, ..., 1.3361e-03, + 6.0654e-04, 3.0947e-04], + [ 1.4944e-03, -1.1452e-02, 8.9550e-04, ..., 9.2316e-04, + 8.0729e-04, 7.2098e-04], + [ 3.9253e-03, 1.8738e-02, -6.6643e-03, ..., 2.4872e-03, + -3.5019e-03, -6.9237e-04]], device='cuda:0') +Epoch 310, bias, value: tensor([-0.0512, 0.0196, 0.0050, -0.0102, -0.0031, 0.0054, -0.0157, 0.0351, + -0.0148, 0.0107], device='cuda:0'), grad: tensor([ 0.0203, -0.0122, 0.0198, -0.0333, 0.0223, 0.0050, 0.0254, 0.0039, + -0.0568, 0.0054], device='cuda:0') +100 +0.0001 +changing lr +epoch 309, time 214.67, cls_loss 0.5275 cls_loss_mapping 0.0039 cls_loss_causal 0.4649 re_mapping 0.0097 re_causal 0.0218 /// teacc 98.82 lr 0.00010000 +Epoch 311, weight, value: tensor([[-0.1923, -0.0150, 0.0261, ..., -0.0320, -0.0791, -0.1313], + [-0.0521, -0.0938, 0.0313, ..., 0.0458, -0.0172, -0.0935], + [-0.0587, -0.0823, -0.1088, ..., 0.0409, -0.0409, -0.1042], + ..., + [-0.0620, 0.0297, 0.0310, ..., 0.0271, -0.0524, -0.0922], + [-0.0725, -0.0066, 0.0240, ..., 0.0659, -0.0364, -0.1591], + [ 0.0594, 0.0617, -0.0516, ..., -0.1095, -0.0184, 0.0804]], + device='cuda:0'), grad: tensor([[ 1.1012e-05, 4.4131e-04, 1.1468e-04, ..., 3.2473e-04, + 2.5198e-05, 5.4789e-04], + [ 9.3639e-05, 2.0447e-03, 3.8266e-04, ..., 2.4319e-03, + 6.5267e-05, 3.2854e-04], + [ 3.6687e-05, 8.1778e-04, 3.4237e-04, ..., 1.9252e-05, + 9.0778e-05, 5.6887e-04], + ..., + [ 3.1013e-03, -6.6032e-03, 1.2917e-02, ..., 1.4908e-02, + 7.9536e-04, -1.7786e-03], + [-6.6012e-06, 1.7824e-03, -6.4049e-03, ..., -1.7052e-03, + 2.1565e-04, 7.5865e-04], + [-3.4447e-03, 2.4109e-03, -1.3084e-02, ..., -2.1423e-02, + -2.1229e-03, 3.9911e-04]], device='cuda:0') +Epoch 311, bias, value: tensor([-0.0509, 0.0196, 0.0053, -0.0101, -0.0028, 0.0044, -0.0154, 0.0343, + -0.0153, 0.0117], device='cuda:0'), grad: tensor([ 0.0103, 0.0170, -0.0278, 0.0123, 0.0060, 0.0098, 0.0166, 0.0035, + 0.0125, -0.0602], device='cuda:0') +100 +0.0001 +changing lr +epoch 310, time 214.67, cls_loss 0.5377 cls_loss_mapping 0.0035 cls_loss_causal 0.4754 re_mapping 0.0096 re_causal 0.0224 /// teacc 98.89 lr 0.00010000 +Epoch 312, weight, value: tensor([[-0.1929, -0.0151, 0.0257, ..., -0.0319, -0.0794, -0.1329], + [-0.0530, -0.0933, 0.0313, ..., 0.0463, -0.0155, -0.0922], + [-0.0591, -0.0841, -0.1075, ..., 0.0410, -0.0402, -0.1052], + ..., + [-0.0630, 0.0288, 0.0305, ..., 0.0265, -0.0529, -0.0924], + [-0.0734, -0.0039, 0.0244, ..., 0.0659, -0.0365, -0.1574], + [ 0.0610, 0.0607, -0.0529, ..., -0.1105, -0.0199, 0.0797]], + device='cuda:0'), grad: tensor([[ 1.2383e-05, -1.5721e-05, -2.5253e-03, ..., 7.2365e-03, + -6.5327e-04, 6.2704e-05], + [ 1.3504e-06, 9.3579e-06, 2.4033e-03, ..., -6.3095e-03, + 1.2159e-03, 6.3896e-05], + [ 4.9621e-06, 5.4419e-05, -3.6278e-03, ..., -2.6169e-03, + 1.1845e-03, 1.8435e-03], + ..., + [ 3.9190e-05, 1.4937e-04, 2.9926e-03, ..., 3.8929e-03, + 8.6164e-04, 3.3784e-04], + [-1.1225e-03, 3.1543e-04, -1.3313e-03, ..., -4.7874e-03, + 1.3046e-03, -4.9448e-04], + [ 1.0061e-03, -1.7786e-03, 2.4414e-04, ..., -3.5858e-03, + -1.8663e-03, -2.5024e-03]], device='cuda:0') +Epoch 312, bias, value: tensor([-0.0501, 0.0209, 0.0057, -0.0103, -0.0032, 0.0049, -0.0162, 0.0333, + -0.0157, 0.0115], device='cuda:0'), grad: tensor([ 0.0318, 0.0184, -0.0165, -0.0095, -0.0060, 0.0176, 0.0195, 0.0291, + -0.0648, -0.0197], device='cuda:0') +100 +0.0001 +changing lr +epoch 311, time 214.65, cls_loss 0.5650 cls_loss_mapping 0.0032 cls_loss_causal 0.5038 re_mapping 0.0094 re_causal 0.0221 /// teacc 98.87 lr 0.00010000 +Epoch 313, weight, value: tensor([[-0.1939, -0.0138, 0.0255, ..., -0.0314, -0.0792, -0.1336], + [-0.0523, -0.0936, 0.0331, ..., 0.0464, -0.0165, -0.0920], + [-0.0597, -0.0835, -0.1087, ..., 0.0401, -0.0412, -0.1048], + ..., + [-0.0636, 0.0296, 0.0315, ..., 0.0265, -0.0526, -0.0936], + [-0.0741, -0.0041, 0.0240, ..., 0.0662, -0.0372, -0.1590], + [ 0.0608, 0.0614, -0.0549, ..., -0.1103, -0.0201, 0.0812]], + device='cuda:0'), grad: tensor([[ 6.0536e-07, 5.3085e-06, 6.8283e-04, ..., 1.2112e-04, + 5.5456e-04, 1.8728e-04], + [ 4.6939e-07, 8.9169e-05, -1.7061e-03, ..., 1.8711e-03, + -2.9030e-03, 2.4853e-03], + [-1.6212e-05, -2.6917e-04, -4.6806e-03, ..., -2.6360e-03, + -2.9869e-03, -3.4447e-03], + ..., + [ 9.3877e-06, 4.3422e-05, 1.6747e-03, ..., 4.5037e-04, + 1.3361e-03, 6.2847e-04], + [ 1.6410e-06, 8.9854e-06, 8.1205e-04, ..., 1.0860e-04, + 6.7759e-04, 1.9741e-04], + [-5.0478e-07, 1.9655e-05, 1.0414e-03, ..., 2.1803e-04, + 8.4543e-04, 3.1543e-04]], device='cuda:0') +Epoch 313, bias, value: tensor([-0.0497, 0.0214, 0.0052, -0.0107, -0.0043, 0.0046, -0.0160, 0.0352, + -0.0162, 0.0113], device='cuda:0'), grad: tensor([ 0.0115, 0.0175, -0.0111, -0.0358, 0.0120, -0.0166, -0.0220, 0.0175, + 0.0143, 0.0128], device='cuda:0') +100 +0.0001 +changing lr +epoch 312, time 215.56, cls_loss 0.5570 cls_loss_mapping 0.0037 cls_loss_causal 0.5025 re_mapping 0.0094 re_causal 0.0224 /// teacc 98.95 lr 0.00010000 +Epoch 314, weight, value: tensor([[-0.1936, -0.0128, 0.0251, ..., -0.0312, -0.0797, -0.1326], + [-0.0510, -0.0938, 0.0326, ..., 0.0461, -0.0156, -0.0925], + [-0.0609, -0.0831, -0.1069, ..., 0.0412, -0.0400, -0.1041], + ..., + [-0.0643, 0.0311, 0.0316, ..., 0.0267, -0.0524, -0.0931], + [-0.0747, -0.0050, 0.0238, ..., 0.0662, -0.0376, -0.1594], + [ 0.0613, 0.0612, -0.0534, ..., -0.1105, -0.0208, 0.0805]], + device='cuda:0'), grad: tensor([[-1.1396e-03, 1.9860e-04, 8.4639e-04, ..., -3.0575e-03, + 1.0767e-03, 9.1270e-07], + [ 1.5771e-04, 3.2568e-04, 1.1969e-03, ..., 5.7983e-04, + 1.5879e-03, 1.3113e-06], + [ 2.1768e-04, 1.7643e-04, 1.0777e-03, ..., 5.1260e-04, + 1.3752e-03, 7.3463e-06], + ..., + [ 1.0008e-04, -2.7161e-03, 4.7350e-04, ..., -4.5729e-04, + 1.3628e-03, 1.5181e-06], + [ 2.2042e-04, 6.2943e-04, 1.0347e-03, ..., 6.5804e-04, + 1.2932e-03, 1.7121e-05], + [-3.0270e-03, 7.2670e-04, -4.3610e-02, ..., -8.2779e-04, + -2.7222e-02, 6.8098e-06]], device='cuda:0') +Epoch 314, bias, value: tensor([-0.0499, 0.0206, 0.0052, -0.0104, -0.0039, 0.0047, -0.0156, 0.0353, + -0.0166, 0.0112], device='cuda:0'), grad: tensor([-0.0041, 0.0336, 0.0293, -0.0581, -0.0046, -0.0096, 0.0020, 0.0123, + 0.0298, -0.0306], device='cuda:0') +100 +0.0001 +changing lr +---------------------saving model at epoch 313---------------------------------------------------- +epoch 313, time 216.20, cls_loss 0.5775 cls_loss_mapping 0.0033 cls_loss_causal 0.5115 re_mapping 0.0095 re_causal 0.0235 /// teacc 99.00 lr 0.00010000 +Epoch 315, weight, value: tensor([[-0.1936, -0.0128, 0.0246, ..., -0.0317, -0.0810, -0.1328], + [-0.0516, -0.0945, 0.0326, ..., 0.0471, -0.0158, -0.0930], + [-0.0612, -0.0837, -0.1051, ..., 0.0422, -0.0379, -0.1048], + ..., + [-0.0647, 0.0310, 0.0312, ..., 0.0261, -0.0537, -0.0944], + [-0.0736, -0.0042, 0.0237, ..., 0.0662, -0.0370, -0.1593], + [ 0.0615, 0.0618, -0.0527, ..., -0.1108, -0.0196, 0.0820]], + device='cuda:0'), grad: tensor([[ 1.9103e-05, 1.2326e-04, 2.1422e-04, ..., 1.7910e-03, + 7.3373e-05, 2.5898e-05], + [ 1.5590e-06, -1.1473e-03, 2.5725e-04, ..., -3.4237e-03, + -2.7580e-03, 1.4734e-06], + [ 5.9128e-05, 3.7909e-04, 3.3712e-04, ..., 2.8000e-03, + 1.5078e-03, 8.5175e-05], + ..., + [-1.2815e-04, -3.4165e-04, -2.1973e-03, ..., 2.7275e-04, + 3.5381e-04, -1.1997e-03], + [-1.0061e-04, 2.3103e-04, -1.8466e-04, ..., 2.4891e-04, + -9.0647e-04, 1.1390e-04], + [ 1.9431e-04, -7.3195e-04, 5.2357e-04, ..., 3.5548e-04, + 8.1873e-04, 1.1921e-03]], device='cuda:0') +Epoch 315, bias, value: tensor([-0.0508, 0.0205, 0.0060, -0.0112, -0.0038, 0.0057, -0.0153, 0.0349, + -0.0164, 0.0112], device='cuda:0'), grad: tensor([ 0.0142, -0.0084, 0.0146, 0.0083, -0.0214, 0.0123, -0.0194, -0.0198, + 0.0121, 0.0076], device='cuda:0') +100 +0.0001 +changing lr +epoch 314, time 215.86, cls_loss 0.5230 cls_loss_mapping 0.0031 cls_loss_causal 0.4563 re_mapping 0.0095 re_causal 0.0211 /// teacc 98.86 lr 0.00010000 +Epoch 316, weight, value: tensor([[-0.1939, -0.0114, 0.0244, ..., -0.0319, -0.0822, -0.1330], + [-0.0523, -0.0946, 0.0321, ..., 0.0462, -0.0177, -0.0933], + [-0.0608, -0.0843, -0.1062, ..., 0.0419, -0.0379, -0.1060], + ..., + [-0.0640, 0.0305, 0.0316, ..., 0.0265, -0.0521, -0.0937], + [-0.0738, -0.0028, 0.0231, ..., 0.0666, -0.0367, -0.1593], + [ 0.0604, 0.0611, -0.0536, ..., -0.1115, -0.0194, 0.0811]], + device='cuda:0'), grad: tensor([[ 4.7588e-04, 4.1771e-04, 9.6607e-04, ..., -7.2403e-03, + -1.3985e-02, 5.4693e-04], + [ 5.2363e-05, 6.2180e-04, 7.3195e-04, ..., 7.6246e-04, + 1.2302e-04, 4.3869e-04], + [ 9.3365e-04, 9.4938e-04, 2.7657e-03, ..., 3.5439e-03, + 8.8501e-04, 1.0843e-03], + ..., + [ 5.5885e-04, 1.9054e-03, -5.8508e-04, ..., 1.8711e-03, + 5.7030e-04, 1.5802e-03], + [ 7.5197e-04, 5.7459e-04, 1.4296e-03, ..., 4.2000e-03, + 1.1702e-03, 1.3123e-03], + [ 7.1049e-04, -5.5199e-03, 1.9398e-03, ..., 8.6784e-05, + -2.2158e-05, 1.4486e-03]], device='cuda:0') +Epoch 316, bias, value: tensor([-0.0519, 0.0205, 0.0054, -0.0103, -0.0032, 0.0049, -0.0150, 0.0357, + -0.0165, 0.0110], device='cuda:0'), grad: tensor([-0.0376, -0.0206, 0.0229, 0.0253, -0.0506, 0.0029, 0.0352, 0.0153, + 0.0228, -0.0156], device='cuda:0') +100 +0.0001 +changing lr +epoch 315, time 221.36, cls_loss 0.5577 cls_loss_mapping 0.0027 cls_loss_causal 0.4912 re_mapping 0.0096 re_causal 0.0225 /// teacc 98.67 lr 0.00010000 +Epoch 317, weight, value: tensor([[-0.1925, -0.0109, 0.0248, ..., -0.0309, -0.0815, -0.1315], + [-0.0529, -0.0951, 0.0328, ..., 0.0467, -0.0174, -0.0934], + [-0.0609, -0.0851, -0.1069, ..., 0.0416, -0.0380, -0.1066], + ..., + [-0.0635, 0.0306, 0.0302, ..., 0.0271, -0.0521, -0.0943], + [-0.0746, -0.0034, 0.0238, ..., 0.0658, -0.0369, -0.1619], + [ 0.0603, 0.0604, -0.0538, ..., -0.1117, -0.0216, 0.0816]], + device='cuda:0'), grad: tensor([[ 1.0672e-03, 5.5224e-05, 1.5011e-03, ..., 1.9491e-04, + 2.6913e-03, 1.4722e-05], + [ 1.6600e-05, 1.0216e-04, 2.5868e-04, ..., 3.3855e-04, + 1.6050e-03, 2.7165e-05], + [ 9.2804e-05, 1.1718e-04, 8.6546e-04, ..., 2.3341e-04, + 1.8177e-03, 2.0579e-05], + ..., + [ 1.0365e-04, 1.4818e-04, 1.1511e-03, ..., 7.0953e-04, + 3.2768e-03, 1.5473e-04], + [-3.1185e-04, -2.6393e-04, 6.2904e-03, ..., -1.3459e-04, + 2.6207e-03, 4.1038e-05], + [ 6.0940e-04, -4.0131e-03, -5.9586e-03, ..., -2.7962e-03, + -7.8888e-03, -8.8787e-04]], device='cuda:0') +Epoch 317, bias, value: tensor([-0.0510, 0.0210, 0.0048, -0.0117, -0.0039, 0.0061, -0.0149, 0.0360, + -0.0170, 0.0112], device='cuda:0'), grad: tensor([ 0.0322, 0.0214, -0.0134, 0.0189, -0.0102, -0.0156, -0.0003, -0.0043, + 0.0294, -0.0582], device='cuda:0') +100 +0.0001 +changing lr +epoch 316, time 226.55, cls_loss 0.5258 cls_loss_mapping 0.0033 cls_loss_causal 0.4616 re_mapping 0.0100 re_causal 0.0228 /// teacc 98.81 lr 0.00010000 +Epoch 318, weight, value: tensor([[-0.1934, -0.0109, 0.0249, ..., -0.0326, -0.0815, -0.1318], + [-0.0530, -0.0956, 0.0330, ..., 0.0464, -0.0181, -0.0926], + [-0.0606, -0.0833, -0.1063, ..., 0.0422, -0.0373, -0.1072], + ..., + [-0.0626, 0.0309, 0.0300, ..., 0.0276, -0.0526, -0.0931], + [-0.0754, -0.0038, 0.0244, ..., 0.0654, -0.0369, -0.1633], + [ 0.0593, 0.0605, -0.0541, ..., -0.1117, -0.0209, 0.0804]], + device='cuda:0'), grad: tensor([[-2.9299e-06, 2.4974e-05, 7.9966e-04, ..., 7.6473e-05, + 1.2169e-03, 6.9082e-05], + [ 2.0843e-06, 2.1887e-04, 3.1090e-03, ..., 6.1941e-04, + -2.6588e-03, 9.3269e-04], + [ 1.1541e-05, 1.4417e-05, 5.3930e-04, ..., 1.0204e-04, + 9.8228e-04, 4.3720e-05], + ..., + [ 5.7191e-05, 3.5071e-04, 1.1368e-03, ..., -6.4909e-05, + 1.5965e-03, 5.0306e-04], + [ 2.3276e-05, 2.6875e-03, 7.5674e-04, ..., 2.5034e-05, + 2.9325e-05, -2.1362e-03], + [-6.9141e-05, -3.6545e-03, 8.5354e-04, ..., 1.4424e-04, + 1.7281e-03, 4.5729e-04]], device='cuda:0') +Epoch 318, bias, value: tensor([-0.0522, 0.0198, 0.0055, -0.0113, -0.0034, 0.0070, -0.0155, 0.0365, + -0.0166, 0.0108], device='cuda:0'), grad: tensor([ 0.0119, -0.0299, 0.0155, 0.0104, -0.0131, -0.0159, 0.0182, -0.0163, + 0.0161, 0.0029], device='cuda:0') +100 +0.0001 +changing lr +epoch 317, time 226.15, cls_loss 0.5216 cls_loss_mapping 0.0032 cls_loss_causal 0.4675 re_mapping 0.0095 re_causal 0.0220 /// teacc 98.73 lr 0.00010000 +Epoch 319, weight, value: tensor([[-0.1938, -0.0112, 0.0261, ..., -0.0327, -0.0806, -0.1319], + [-0.0526, -0.0954, 0.0339, ..., 0.0462, -0.0188, -0.0931], + [-0.0600, -0.0838, -0.1071, ..., 0.0423, -0.0367, -0.1072], + ..., + [-0.0628, 0.0293, 0.0300, ..., 0.0283, -0.0525, -0.0927], + [-0.0759, -0.0041, 0.0248, ..., 0.0655, -0.0372, -0.1632], + [ 0.0591, 0.0614, -0.0552, ..., -0.1112, -0.0201, 0.0803]], + device='cuda:0'), grad: tensor([[ 4.2498e-05, 5.3704e-05, 4.7188e-03, ..., 2.2869e-03, + 1.3757e-04, 1.9913e-03], + [ 6.4850e-05, 4.3368e-04, -2.2602e-03, ..., -2.4967e-03, + -2.9755e-04, 1.3018e-04], + [ 7.7426e-05, 3.4499e-04, 2.9068e-03, ..., 1.8053e-03, + 4.6229e-04, 6.6233e-04], + ..., + [ 4.5228e-04, -2.2640e-03, -1.6336e-03, ..., 1.7052e-03, + 2.8877e-03, -1.8034e-03], + [ 7.1192e-04, 3.9816e-04, 3.8586e-03, ..., -3.8300e-03, + -3.2005e-03, 1.0910e-03], + [-3.9330e-03, 4.4465e-04, -3.6888e-03, ..., -4.5662e-03, + -2.5158e-03, -1.0551e-02]], device='cuda:0') +Epoch 319, bias, value: tensor([-0.0523, 0.0213, 0.0062, -0.0117, -0.0029, 0.0060, -0.0162, 0.0368, + -0.0164, 0.0100], device='cuda:0'), grad: tensor([ 0.0110, -0.0030, 0.0319, -0.0043, 0.0271, -0.0083, -0.0117, 0.0169, + -0.0294, -0.0302], device='cuda:0') +100 +0.0001 +changing lr +epoch 318, time 229.08, cls_loss 0.5169 cls_loss_mapping 0.0029 cls_loss_causal 0.4438 re_mapping 0.0092 re_causal 0.0219 /// teacc 98.73 lr 0.00010000 +Epoch 320, weight, value: tensor([[-0.1951, -0.0114, 0.0271, ..., -0.0316, -0.0808, -0.1330], + [-0.0519, -0.0970, 0.0349, ..., 0.0454, -0.0196, -0.0932], + [-0.0600, -0.0843, -0.1085, ..., 0.0424, -0.0348, -0.1069], + ..., + [-0.0645, 0.0295, 0.0298, ..., 0.0276, -0.0525, -0.0942], + [-0.0758, -0.0040, 0.0238, ..., 0.0665, -0.0381, -0.1635], + [ 0.0598, 0.0605, -0.0551, ..., -0.1118, -0.0196, 0.0805]], + device='cuda:0'), grad: tensor([[ 1.0765e-04, 6.7902e-04, 9.1314e-04, ..., 2.1958e-04, + 1.0604e-04, 5.3406e-04], + [ 3.0305e-06, 5.8746e-04, 1.8442e-04, ..., 2.1911e-04, + 6.3121e-05, 2.0564e-04], + [ 1.8501e-04, 8.2827e-04, 1.4400e-03, ..., 2.3055e-04, + 1.8167e-04, 5.6696e-04], + ..., + [ 6.5491e-06, 8.1635e-04, 8.0776e-04, ..., 2.6703e-04, + 2.5725e-04, 1.5211e-04], + [-9.0361e-05, 3.9339e-04, -5.3864e-03, ..., 1.2851e-04, + 1.1736e-04, -9.1970e-05], + [ 8.3894e-06, 7.4434e-04, 1.9350e-03, ..., 1.5032e-04, + -3.4451e-04, -3.8147e-03]], device='cuda:0') +Epoch 320, bias, value: tensor([-0.0512, 0.0200, 0.0059, -0.0111, -0.0036, 0.0059, -0.0158, 0.0365, + -0.0166, 0.0107], device='cuda:0'), grad: tensor([-0.0072, 0.0011, 0.0193, -0.0435, -0.0061, 0.0229, 0.0275, 0.0055, + -0.0251, 0.0057], device='cuda:0') +100 +0.0001 +changing lr +epoch 319, time 226.29, cls_loss 0.5202 cls_loss_mapping 0.0034 cls_loss_causal 0.4579 re_mapping 0.0091 re_causal 0.0225 /// teacc 98.82 lr 0.00010000 +Epoch 321, weight, value: tensor([[-0.1956, -0.0116, 0.0268, ..., -0.0308, -0.0810, -0.1339], + [-0.0525, -0.0965, 0.0353, ..., 0.0454, -0.0196, -0.0923], + [-0.0594, -0.0840, -0.1090, ..., 0.0412, -0.0358, -0.1066], + ..., + [-0.0661, 0.0308, 0.0299, ..., 0.0285, -0.0529, -0.0948], + [-0.0755, -0.0031, 0.0246, ..., 0.0661, -0.0390, -0.1636], + [ 0.0598, 0.0605, -0.0561, ..., -0.1122, -0.0199, 0.0804]], + device='cuda:0'), grad: tensor([[ 1.5485e-04, 8.4925e-04, 3.0303e-04, ..., 3.6073e-04, + 1.8225e-03, 8.1003e-05], + [ 1.0781e-05, 1.3626e-04, -7.5293e-04, ..., -1.1024e-03, + 1.4839e-03, 9.0823e-06], + [ 5.1826e-05, -6.6376e-03, 3.5071e-04, ..., -8.2636e-04, + -8.8596e-04, 4.2617e-05], + ..., + [ 5.5847e-03, 8.1444e-04, -7.6818e-04, ..., -2.5368e-04, + -8.3389e-03, 5.0392e-03], + [ 3.8362e-04, 4.1795e-04, 3.6073e-04, ..., 2.3365e-04, + 1.6060e-03, 2.2244e-04], + [-7.2250e-03, 1.9817e-03, -7.5674e-04, ..., 7.5293e-04, + 1.9321e-03, -6.0577e-03]], device='cuda:0') +Epoch 321, bias, value: tensor([-0.0513, 0.0201, 0.0066, -0.0106, -0.0044, 0.0061, -0.0163, 0.0364, + -0.0167, 0.0109], device='cuda:0'), grad: tensor([ 0.0271, -0.0085, -0.0384, 0.0231, 0.0322, -0.0410, -0.0081, -0.0150, + 0.0237, 0.0049], device='cuda:0') +100 +0.0001 +changing lr +epoch 320, time 227.21, cls_loss 0.5365 cls_loss_mapping 0.0036 cls_loss_causal 0.4627 re_mapping 0.0092 re_causal 0.0219 /// teacc 98.83 lr 0.00010000 +Epoch 322, weight, value: tensor([[-0.1961, -0.0117, 0.0259, ..., -0.0315, -0.0827, -0.1335], + [-0.0529, -0.0970, 0.0350, ..., 0.0453, -0.0211, -0.0922], + [-0.0588, -0.0840, -0.1089, ..., 0.0413, -0.0367, -0.1067], + ..., + [-0.0672, 0.0305, 0.0293, ..., 0.0283, -0.0531, -0.0954], + [-0.0754, -0.0033, 0.0247, ..., 0.0666, -0.0375, -0.1642], + [ 0.0611, 0.0603, -0.0544, ..., -0.1120, -0.0206, 0.0807]], + device='cuda:0'), grad: tensor([[ 2.5660e-05, 2.3270e-04, 7.2241e-05, ..., 4.0936e-04, + 2.0447e-02, 2.0409e-04], + [ 4.2915e-05, 5.8889e-04, 1.1820e-04, ..., 2.3448e-04, + 1.0109e-04, 4.1652e-04], + [ 1.5914e-04, -1.2560e-03, 4.4870e-04, ..., -4.0078e-04, + 2.9588e-04, -2.2888e-03], + ..., + [ 3.5715e-04, -1.5488e-03, 9.4175e-04, ..., 3.1924e-04, + 4.6968e-04, 2.0325e-04], + [ 2.3155e-03, 7.3767e-04, 6.4850e-03, ..., 1.9817e-03, + -1.8402e-02, 2.2852e-04], + [ 1.0653e-03, 5.7888e-04, 3.2482e-03, ..., 1.0128e-03, + 1.5545e-03, -6.0415e-04]], device='cuda:0') +Epoch 322, bias, value: tensor([-0.0518, 0.0206, 0.0053, -0.0103, -0.0030, 0.0060, -0.0153, 0.0357, + -0.0173, 0.0108], device='cuda:0'), grad: tensor([-0.0199, 0.0164, -0.0110, -0.0291, 0.0016, 0.0230, 0.0175, -0.0156, + -0.0027, 0.0198], device='cuda:0') +100 +0.0001 +changing lr +epoch 321, time 227.51, cls_loss 0.5585 cls_loss_mapping 0.0039 cls_loss_causal 0.4893 re_mapping 0.0094 re_causal 0.0224 /// teacc 98.84 lr 0.00010000 +Epoch 323, weight, value: tensor([[-0.1963, -0.0117, 0.0265, ..., -0.0317, -0.0833, -0.1331], + [-0.0521, -0.0973, 0.0346, ..., 0.0451, -0.0203, -0.0923], + [-0.0594, -0.0847, -0.1112, ..., 0.0405, -0.0374, -0.1065], + ..., + [-0.0673, 0.0310, 0.0289, ..., 0.0285, -0.0537, -0.0957], + [-0.0760, -0.0012, 0.0251, ..., 0.0664, -0.0375, -0.1636], + [ 0.0616, 0.0605, -0.0537, ..., -0.1117, -0.0215, 0.0813]], + device='cuda:0'), grad: tensor([[ 1.7583e-05, 2.1964e-05, 1.5278e-03, ..., 1.3590e-03, + 1.8778e-03, 1.0699e-05], + [ 3.9220e-04, 9.8348e-07, 3.9749e-03, ..., 4.9248e-03, + -6.9923e-03, 3.0816e-05], + [-5.7173e-04, 2.9150e-06, 1.8377e-03, ..., 1.8463e-03, + 2.3823e-03, -2.9683e-04], + ..., + [ 8.3387e-05, 6.5640e-06, -6.5880e-03, ..., -3.1781e-04, + -7.3471e-03, 5.0634e-05], + [-2.4891e-04, 6.5207e-05, 6.4898e-04, ..., -1.1301e-03, + 1.5669e-03, 4.8637e-04], + [ 1.5378e-05, 1.3142e-03, -1.3943e-03, ..., 2.1350e-04, + 2.0809e-03, -5.4073e-04]], device='cuda:0') +Epoch 323, bias, value: tensor([-0.0519, 0.0201, 0.0042, -0.0110, -0.0030, 0.0072, -0.0146, 0.0351, + -0.0177, 0.0124], device='cuda:0'), grad: tensor([ 0.0274, -0.0185, 0.0287, -0.0380, -0.0628, -0.0044, 0.0193, 0.0060, + 0.0199, 0.0224], device='cuda:0') +100 +0.0001 +changing lr +epoch 322, time 226.70, cls_loss 0.5326 cls_loss_mapping 0.0036 cls_loss_causal 0.4666 re_mapping 0.0101 re_causal 0.0232 /// teacc 98.90 lr 0.00010000 +Epoch 324, weight, value: tensor([[-0.1962, -0.0124, 0.0254, ..., -0.0312, -0.0840, -0.1323], + [-0.0516, -0.0978, 0.0352, ..., 0.0455, -0.0192, -0.0917], + [-0.0599, -0.0865, -0.1111, ..., 0.0403, -0.0375, -0.1063], + ..., + [-0.0674, 0.0318, 0.0298, ..., 0.0287, -0.0532, -0.0945], + [-0.0766, -0.0022, 0.0265, ..., 0.0668, -0.0366, -0.1636], + [ 0.0610, 0.0617, -0.0550, ..., -0.1118, -0.0229, 0.0816]], + device='cuda:0'), grad: tensor([[ 7.9162e-07, 1.3888e-04, 5.3978e-04, ..., 9.0218e-04, + 9.4032e-04, 4.3288e-06], + [ 4.0419e-07, 2.7442e-04, 7.0620e-04, ..., 8.2064e-04, + 1.3084e-03, 1.0785e-06], + [ 1.2890e-06, -1.7967e-03, 5.0020e-04, ..., 1.0796e-03, + -1.5984e-03, 8.4937e-06], + ..., + [ 1.2107e-07, 1.2856e-03, 3.6392e-03, ..., 2.7269e-05, + 1.6937e-03, 1.7095e-04], + [ 3.2708e-06, 4.2081e-04, 4.4155e-04, ..., -2.7728e-04, + 1.5354e-03, 3.2306e-05], + [-1.1325e-05, -6.5207e-05, 9.7322e-04, ..., 8.6498e-04, + -1.0559e-02, -4.3154e-04]], device='cuda:0') +Epoch 324, bias, value: tensor([-0.0526, 0.0199, 0.0041, -0.0120, -0.0036, 0.0079, -0.0142, 0.0365, + -0.0174, 0.0120], device='cuda:0'), grad: tensor([-0.0261, 0.0182, -0.0154, 0.0236, -0.0030, 0.0194, -0.0515, 0.0286, + 0.0174, -0.0111], device='cuda:0') +100 +0.0001 +changing lr +epoch 323, time 227.89, cls_loss 0.5395 cls_loss_mapping 0.0037 cls_loss_causal 0.4823 re_mapping 0.0096 re_causal 0.0223 /// teacc 98.90 lr 0.00010000 +Epoch 325, weight, value: tensor([[-0.1964, -0.0132, 0.0256, ..., -0.0298, -0.0823, -0.1333], + [-0.0514, -0.0983, 0.0354, ..., 0.0474, -0.0192, -0.0910], + [-0.0615, -0.0860, -0.1101, ..., 0.0393, -0.0371, -0.1072], + ..., + [-0.0678, 0.0301, 0.0311, ..., 0.0271, -0.0528, -0.0959], + [-0.0769, -0.0011, 0.0262, ..., 0.0684, -0.0379, -0.1647], + [ 0.0612, 0.0620, -0.0550, ..., -0.1117, -0.0233, 0.0822]], + device='cuda:0'), grad: tensor([[ 1.6063e-05, 5.5879e-07, 5.7518e-05, ..., 3.2604e-05, + 5.3830e-07, 7.6443e-06], + [ 1.0543e-05, 2.4214e-07, -3.8719e-04, ..., 1.6674e-05, + 3.9674e-07, 5.1484e-06], + [ 1.1110e-04, 5.2378e-06, 2.0933e-04, ..., 1.2188e-03, + 1.5712e-04, 3.6001e-05], + ..., + [-8.4579e-05, 2.5034e-06, 2.3308e-03, ..., 1.9302e-02, + 2.5616e-03, -1.5482e-05], + [ 2.6301e-05, 1.8746e-05, 4.1515e-05, ..., 2.7195e-05, + 2.0310e-05, 6.2644e-05], + [ 1.0240e-04, -1.7774e-04, 1.3027e-03, ..., 2.4235e-04, + -1.0818e-04, -2.8825e-04]], device='cuda:0') +Epoch 325, bias, value: tensor([-0.0508, 0.0194, 0.0052, -0.0117, -0.0029, 0.0078, -0.0151, 0.0352, + -0.0181, 0.0118], device='cuda:0'), grad: tensor([-0.0268, 0.0013, 0.0064, -0.0270, 0.0466, -0.0034, -0.0234, 0.0350, + -0.0189, 0.0102], device='cuda:0') +100 +0.0001 +changing lr +epoch 324, time 228.08, cls_loss 0.5154 cls_loss_mapping 0.0033 cls_loss_causal 0.4568 re_mapping 0.0102 re_causal 0.0231 /// teacc 98.87 lr 0.00010000 +Epoch 326, weight, value: tensor([[-0.1962, -0.0129, 0.0260, ..., -0.0288, -0.0823, -0.1343], + [-0.0523, -0.0985, 0.0345, ..., 0.0474, -0.0183, -0.0903], + [-0.0614, -0.0846, -0.1103, ..., 0.0386, -0.0364, -0.1073], + ..., + [-0.0668, 0.0293, 0.0295, ..., 0.0253, -0.0520, -0.0958], + [-0.0764, -0.0015, 0.0267, ..., 0.0683, -0.0387, -0.1657], + [ 0.0613, 0.0639, -0.0537, ..., -0.1095, -0.0220, 0.0821]], + device='cuda:0'), grad: tensor([[ 2.8778e-06, 7.4387e-05, 1.2755e-04, ..., -8.2874e-04, + 4.4394e-04, 2.5082e-04], + [ 3.6936e-06, -6.9714e-04, -2.2411e-03, ..., 6.4392e-03, + -2.0657e-03, 1.6344e-04], + [-6.3658e-05, 1.5521e-04, 6.0368e-04, ..., 2.1114e-03, + 9.5558e-04, 1.2541e-03], + ..., + [ 3.2306e-05, 9.8109e-05, 3.3998e-04, ..., -8.7128e-03, + 6.1035e-04, 2.5487e-04], + [ 4.0263e-05, 1.0216e-04, 5.1594e-04, ..., 1.2083e-03, + 9.8991e-04, 3.1304e-04], + [-4.5329e-05, 2.2297e-03, 4.4942e-04, ..., -2.1210e-03, + -2.9697e-03, 2.2850e-03]], device='cuda:0') +Epoch 326, bias, value: tensor([-0.0507, 0.0204, 0.0046, -0.0127, -0.0034, 0.0080, -0.0157, 0.0355, + -0.0181, 0.0127], device='cuda:0'), grad: tensor([-0.0166, -0.0242, 0.0448, 0.0044, -0.0153, 0.0138, -0.0089, -0.0132, + 0.0200, -0.0048], device='cuda:0') +100 +0.0001 +changing lr +epoch 325, time 226.57, cls_loss 0.5480 cls_loss_mapping 0.0032 cls_loss_causal 0.4781 re_mapping 0.0095 re_causal 0.0228 /// teacc 98.86 lr 0.00010000 +Epoch 327, weight, value: tensor([[-0.1978, -0.0130, 0.0260, ..., -0.0295, -0.0829, -0.1348], + [-0.0518, -0.0982, 0.0358, ..., 0.0469, -0.0179, -0.0900], + [-0.0615, -0.0846, -0.1109, ..., 0.0384, -0.0367, -0.1077], + ..., + [-0.0657, 0.0289, 0.0299, ..., 0.0252, -0.0508, -0.0953], + [-0.0771, -0.0011, 0.0257, ..., 0.0683, -0.0395, -0.1664], + [ 0.0612, 0.0638, -0.0531, ..., -0.1077, -0.0215, 0.0825]], + device='cuda:0'), grad: tensor([[ 4.9621e-06, -4.7874e-03, 1.0884e-04, ..., 6.8951e-04, + -1.5430e-03, -2.8362e-03], + [ 3.7104e-06, -4.2496e-03, 1.3018e-03, ..., 1.0014e-03, + -6.1302e-03, 1.0788e-05], + [ 6.6236e-06, 5.7840e-04, 1.5438e-04, ..., -1.9951e-03, + 5.0116e-04, 2.2793e-04], + ..., + [-1.1511e-05, 2.0397e-04, 1.0657e-04, ..., 1.1988e-03, + 4.1103e-04, 5.8115e-05], + [ 8.3074e-06, 2.8706e-03, 1.1768e-03, ..., 5.3215e-04, + 3.8090e-03, 1.4913e-04], + [-6.3404e-06, 8.9693e-04, 1.5903e-04, ..., -3.5248e-03, + 4.4775e-04, 4.3702e-04]], device='cuda:0') +Epoch 327, bias, value: tensor([-0.0510, 0.0208, 0.0053, -0.0129, -0.0039, 0.0075, -0.0170, 0.0356, + -0.0180, 0.0141], device='cuda:0'), grad: tensor([-0.0437, -0.0257, -0.0146, 0.0135, 0.0250, 0.0198, -0.0041, 0.0138, + 0.0377, -0.0217], device='cuda:0') +100 +0.0001 +changing lr +epoch 326, time 228.42, cls_loss 0.5229 cls_loss_mapping 0.0026 cls_loss_causal 0.4573 re_mapping 0.0096 re_causal 0.0218 /// teacc 98.77 lr 0.00010000 +Epoch 328, weight, value: tensor([[-0.1986, -0.0153, 0.0276, ..., -0.0287, -0.0830, -0.1349], + [-0.0514, -0.0967, 0.0356, ..., 0.0470, -0.0175, -0.0901], + [-0.0614, -0.0859, -0.1103, ..., 0.0385, -0.0369, -0.1081], + ..., + [-0.0643, 0.0290, 0.0310, ..., 0.0249, -0.0525, -0.0956], + [-0.0784, -0.0015, 0.0241, ..., 0.0687, -0.0383, -0.1677], + [ 0.0614, 0.0641, -0.0536, ..., -0.1076, -0.0221, 0.0827]], + device='cuda:0'), grad: tensor([[ 7.9498e-06, 3.3259e-04, 1.6320e-04, ..., 1.1846e-05, + 2.1815e-04, 2.9519e-05], + [ 4.8168e-06, 1.3888e-04, 3.0351e-04, ..., 4.5478e-05, + 4.9067e-04, 2.2560e-05], + [-1.0419e-04, 1.3554e-04, 1.9896e-04, ..., -4.3640e-03, + -1.9522e-03, -1.4771e-06], + ..., + [ 1.5557e-05, 8.9455e-04, 6.9857e-04, ..., 1.8568e-03, + 2.0294e-03, 8.7857e-05], + [ 1.0170e-05, 1.0767e-03, 9.8801e-04, ..., 1.9054e-03, + 2.5749e-03, 5.1439e-05], + [ 1.9446e-06, -9.4833e-03, -1.7509e-03, ..., 1.4937e-04, + -2.8458e-03, -5.5742e-04]], device='cuda:0') +Epoch 328, bias, value: tensor([-0.0524, 0.0217, 0.0056, -0.0128, -0.0042, 0.0075, -0.0158, 0.0357, + -0.0181, 0.0132], device='cuda:0'), grad: tensor([-0.0264, 0.0071, -0.0203, 0.0075, -0.0037, -0.0034, 0.0135, 0.0194, + 0.0182, -0.0119], device='cuda:0') +100 +0.0001 +changing lr +epoch 327, time 227.31, cls_loss 0.5292 cls_loss_mapping 0.0024 cls_loss_causal 0.4685 re_mapping 0.0092 re_causal 0.0225 /// teacc 98.69 lr 0.00010000 +Epoch 329, weight, value: tensor([[-0.1973, -0.0140, 0.0284, ..., -0.0279, -0.0833, -0.1343], + [-0.0524, -0.0971, 0.0364, ..., 0.0468, -0.0173, -0.0903], + [-0.0612, -0.0863, -0.1097, ..., 0.0379, -0.0371, -0.1076], + ..., + [-0.0652, 0.0294, 0.0312, ..., 0.0249, -0.0528, -0.0968], + [-0.0784, -0.0024, 0.0245, ..., 0.0693, -0.0373, -0.1690], + [ 0.0607, 0.0638, -0.0553, ..., -0.1073, -0.0220, 0.0828]], + device='cuda:0'), grad: tensor([[ 4.9782e-03, 1.0147e-03, 9.1314e-04, ..., 2.3723e-04, + 2.1887e-04, 1.2217e-03], + [ 7.4387e-05, 1.3828e-03, -4.0398e-03, ..., 5.2834e-04, + -3.7313e-04, 1.2720e-04], + [ 3.2306e-04, 1.0061e-03, -1.6775e-03, ..., -1.9908e-04, + -1.0405e-03, 3.1686e-04], + ..., + [ 7.0930e-05, -5.1636e-02, 3.7694e-04, ..., 2.3441e-03, + 2.8181e-04, -2.4078e-02], + [-1.3466e-03, 1.4048e-03, 1.9779e-03, ..., -9.4376e-03, + -3.0384e-03, 3.0375e-04], + [ 1.0004e-03, 5.3314e-02, 3.7003e-04, ..., 4.5776e-03, + 2.3136e-03, 2.4277e-02]], device='cuda:0') +Epoch 329, bias, value: tensor([-0.0513, 0.0204, 0.0061, -0.0129, -0.0039, 0.0076, -0.0159, 0.0349, + -0.0161, 0.0115], device='cuda:0'), grad: tensor([ 0.0370, -0.0168, -0.0235, -0.0119, -0.0021, -0.0145, 0.0241, 0.0068, + 0.0021, -0.0012], device='cuda:0') +100 +0.0001 +changing lr +epoch 328, time 227.53, cls_loss 0.5330 cls_loss_mapping 0.0021 cls_loss_causal 0.4732 re_mapping 0.0096 re_causal 0.0224 /// teacc 98.84 lr 0.00010000 +Epoch 330, weight, value: tensor([[-0.1980, -0.0139, 0.0284, ..., -0.0276, -0.0831, -0.1344], + [-0.0504, -0.0965, 0.0367, ..., 0.0485, -0.0177, -0.0915], + [-0.0608, -0.0860, -0.1088, ..., 0.0391, -0.0376, -0.1077], + ..., + [-0.0655, 0.0300, 0.0299, ..., 0.0234, -0.0549, -0.0956], + [-0.0792, -0.0016, 0.0251, ..., 0.0691, -0.0375, -0.1705], + [ 0.0606, 0.0630, -0.0548, ..., -0.1083, -0.0202, 0.0821]], + device='cuda:0'), grad: tensor([[ 4.5985e-05, 1.8311e-04, -3.8505e-04, ..., -4.1046e-03, + -3.8433e-03, 4.9561e-05], + [ 9.1136e-05, 2.7679e-06, 3.0661e-04, ..., 4.2748e-04, + 6.4945e-04, 9.0599e-05], + [ 5.6416e-05, 9.9778e-05, 2.2590e-04, ..., 6.8760e-04, + 2.4395e-03, 3.4547e-04], + ..., + [-2.8629e-03, -3.2377e-04, -2.8515e-03, ..., -1.3027e-03, + 2.1515e-03, -1.6155e-03], + [ 3.8195e-04, 1.6823e-05, 3.7789e-04, ..., 5.8556e-04, + 1.2131e-03, 3.1519e-04], + [ 2.1667e-03, 1.1846e-05, 2.3556e-03, ..., 1.3552e-03, + 1.3618e-03, 1.6794e-03]], device='cuda:0') +Epoch 330, bias, value: tensor([-0.0512, 0.0207, 0.0070, -0.0121, -0.0041, 0.0075, -0.0160, 0.0341, + -0.0167, 0.0114], device='cuda:0'), grad: tensor([-0.0488, 0.0148, 0.0307, 0.0096, 0.0113, -0.0070, -0.0205, -0.0358, + 0.0176, 0.0280], device='cuda:0') +100 +0.0001 +changing lr +epoch 329, time 226.85, cls_loss 0.5326 cls_loss_mapping 0.0021 cls_loss_causal 0.4698 re_mapping 0.0094 re_causal 0.0215 /// teacc 98.91 lr 0.00010000 +Epoch 331, weight, value: tensor([[-0.1988, -0.0143, 0.0283, ..., -0.0286, -0.0831, -0.1343], + [-0.0507, -0.0972, 0.0376, ..., 0.0479, -0.0182, -0.0921], + [-0.0609, -0.0856, -0.1101, ..., 0.0391, -0.0365, -0.1085], + ..., + [-0.0653, 0.0307, 0.0301, ..., 0.0228, -0.0561, -0.0942], + [-0.0796, -0.0013, 0.0250, ..., 0.0713, -0.0379, -0.1703], + [ 0.0605, 0.0624, -0.0546, ..., -0.1074, -0.0200, 0.0816]], + device='cuda:0'), grad: tensor([[ 1.7494e-05, 6.3956e-05, 7.4911e-04, ..., 1.1273e-05, + 7.0930e-05, 5.9783e-05], + [ 2.8759e-06, -2.7210e-05, 4.3511e-04, ..., 3.7216e-06, + 5.3495e-06, 8.7470e-06], + [ 1.2986e-05, -3.4761e-04, -1.0223e-02, ..., -5.3853e-05, + -3.2020e-04, -3.2306e-04], + ..., + [-2.0489e-07, 1.5879e-03, 2.1011e-05, ..., 9.2015e-07, + 1.5661e-05, 1.2627e-03], + [ 2.2340e-04, 1.6296e-04, 9.5320e-04, ..., -4.1336e-05, + 1.7273e-04, 1.4532e-04], + [ 1.1642e-06, -2.7542e-03, 1.8320e-03, ..., 1.0386e-05, + 1.7941e-05, -2.2087e-03]], device='cuda:0') +Epoch 331, bias, value: tensor([-0.0512, 0.0217, 0.0061, -0.0127, -0.0054, 0.0076, -0.0148, 0.0339, + -0.0168, 0.0122], device='cuda:0'), grad: tensor([-0.0222, 0.0130, -0.0264, -0.0158, 0.0150, 0.0217, 0.0012, -0.0186, + 0.0218, 0.0104], device='cuda:0') +100 +0.0001 +changing lr +epoch 330, time 226.87, cls_loss 0.5309 cls_loss_mapping 0.0028 cls_loss_causal 0.4749 re_mapping 0.0097 re_causal 0.0223 /// teacc 98.85 lr 0.00010000 +Epoch 332, weight, value: tensor([[-0.1983, -0.0146, 0.0285, ..., -0.0279, -0.0819, -0.1335], + [-0.0511, -0.0972, 0.0372, ..., 0.0470, -0.0191, -0.0930], + [-0.0612, -0.0848, -0.1101, ..., 0.0394, -0.0361, -0.1081], + ..., + [-0.0665, 0.0301, 0.0297, ..., 0.0230, -0.0569, -0.0949], + [-0.0803, -0.0013, 0.0245, ..., 0.0717, -0.0370, -0.1698], + [ 0.0619, 0.0634, -0.0553, ..., -0.1073, -0.0211, 0.0820]], + device='cuda:0'), grad: tensor([[-7.8440e-04, 2.9862e-05, 6.5613e-04, ..., 7.2479e-04, + -6.5708e-04, 1.4804e-05], + [ 1.6034e-05, 4.5240e-05, -1.1194e-04, ..., 2.4930e-05, + 4.3586e-06, 2.0057e-05], + [ 8.8990e-05, 2.5451e-05, 6.8521e-04, ..., 6.0797e-04, + 3.1799e-05, 4.6678e-06], + ..., + [ 1.6344e-04, -1.4269e-04, 5.1594e-04, ..., 1.7917e-04, + 3.2187e-05, -5.7817e-06], + [ 2.0370e-03, 2.7905e-03, 3.0727e-03, ..., -5.1785e-04, + 3.4660e-05, 1.3990e-03], + [-6.1264e-03, 2.1070e-05, -1.4465e-02, ..., -3.5152e-03, + 1.8692e-04, -3.4779e-05]], device='cuda:0') +Epoch 332, bias, value: tensor([-0.0508, 0.0213, 0.0067, -0.0134, -0.0054, 0.0076, -0.0136, 0.0333, + -0.0161, 0.0112], device='cuda:0'), grad: tensor([ 0.0187, -0.0199, 0.0201, -0.0221, 0.0288, -0.0447, 0.0192, 0.0125, + 0.0058, -0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 331, time 228.03, cls_loss 0.5243 cls_loss_mapping 0.0024 cls_loss_causal 0.4536 re_mapping 0.0095 re_causal 0.0224 /// teacc 98.92 lr 0.00010000 +Epoch 333, weight, value: tensor([[-0.1988, -0.0149, 0.0294, ..., -0.0268, -0.0819, -0.1338], + [-0.0516, -0.0991, 0.0377, ..., 0.0476, -0.0188, -0.0918], + [-0.0599, -0.0859, -0.1086, ..., 0.0405, -0.0349, -0.1081], + ..., + [-0.0663, 0.0280, 0.0294, ..., 0.0223, -0.0582, -0.0928], + [-0.0797, -0.0003, 0.0225, ..., 0.0704, -0.0366, -0.1714], + [ 0.0616, 0.0637, -0.0553, ..., -0.1070, -0.0203, 0.0827]], + device='cuda:0'), grad: tensor([[-5.0592e-04, 9.5987e-04, 1.3342e-03, ..., 3.6526e-04, + 1.4496e-03, 9.2745e-05], + [-6.8092e-04, -3.0270e-03, -1.0204e-03, ..., -2.0542e-03, + -2.7733e-03, -6.3820e-03], + [ 3.1471e-04, 3.0231e-04, -2.5711e-03, ..., 2.7156e-04, + -4.3607e-04, 3.6526e-04], + ..., + [ 4.4098e-03, -2.5864e-02, 9.6207e-03, ..., 4.6883e-03, + -2.0866e-03, 9.5062e-03], + [ 6.4707e-04, -3.4332e-03, 1.3809e-03, ..., 4.9114e-04, + 3.3283e-04, 6.4087e-04], + [ 7.0953e-04, 2.7420e-02, 1.3170e-03, ..., -1.1530e-03, + 1.6375e-03, 7.5293e-04]], device='cuda:0') +Epoch 333, bias, value: tensor([-0.0515, 0.0213, 0.0068, -0.0129, -0.0057, 0.0080, -0.0144, 0.0337, + -0.0160, 0.0112], device='cuda:0'), grad: tensor([-0.0117, -0.0225, -0.0178, -0.0375, -0.0095, 0.0172, 0.0089, 0.0417, + 0.0174, 0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 332, time 228.26, cls_loss 0.5360 cls_loss_mapping 0.0028 cls_loss_causal 0.4726 re_mapping 0.0096 re_causal 0.0215 /// teacc 98.80 lr 0.00010000 +Epoch 334, weight, value: tensor([[-0.1982, -0.0160, 0.0303, ..., -0.0266, -0.0806, -0.1347], + [-0.0530, -0.0993, 0.0373, ..., 0.0479, -0.0184, -0.0919], + [-0.0579, -0.0850, -0.1097, ..., 0.0402, -0.0353, -0.1069], + ..., + [-0.0681, 0.0302, 0.0284, ..., 0.0218, -0.0593, -0.0938], + [-0.0788, -0.0004, 0.0222, ..., 0.0704, -0.0373, -0.1717], + [ 0.0615, 0.0623, -0.0541, ..., -0.1077, -0.0199, 0.0835]], + device='cuda:0'), grad: tensor([[ 4.9084e-05, 1.2085e-05, 7.8773e-04, ..., 2.4870e-05, + 6.6900e-04, 2.0921e-05], + [ 2.6152e-05, 4.7028e-05, -1.1177e-03, ..., 1.4015e-05, + 8.5640e-04, 2.3574e-05], + [-7.8297e-04, 9.4833e-03, -2.3804e-03, ..., 7.3051e-03, + -5.0278e-03, -3.7253e-05], + ..., + [ 5.6982e-04, 8.2684e-04, 1.7309e-03, ..., 1.1396e-03, + 1.0042e-03, 1.9622e-04], + [ 1.1873e-04, 4.4435e-05, 6.6566e-04, ..., 2.1517e-05, + 4.6253e-04, 8.2493e-05], + [-3.6263e-04, 4.0550e-03, 7.9298e-04, ..., 9.5293e-06, + 4.7421e-04, -4.6968e-04]], device='cuda:0') +Epoch 334, bias, value: tensor([-0.0517, 0.0225, 0.0074, -0.0132, -0.0058, 0.0077, -0.0141, 0.0335, + -0.0176, 0.0120], device='cuda:0'), grad: tensor([ 0.0129, 0.0104, -0.0075, -0.0160, 0.0188, 0.0119, -0.0089, -0.0205, + -0.0202, 0.0191], device='cuda:0') +100 +0.0001 +changing lr +epoch 333, time 228.34, cls_loss 0.5173 cls_loss_mapping 0.0033 cls_loss_causal 0.4486 re_mapping 0.0095 re_causal 0.0210 /// teacc 98.83 lr 0.00010000 +Epoch 335, weight, value: tensor([[-0.1984, -0.0161, 0.0309, ..., -0.0268, -0.0800, -0.1346], + [-0.0539, -0.0993, 0.0375, ..., 0.0486, -0.0181, -0.0929], + [-0.0572, -0.0866, -0.1095, ..., 0.0405, -0.0356, -0.1065], + ..., + [-0.0691, 0.0305, 0.0282, ..., 0.0223, -0.0598, -0.0947], + [-0.0784, -0.0014, 0.0224, ..., 0.0696, -0.0376, -0.1730], + [ 0.0627, 0.0625, -0.0546, ..., -0.1075, -0.0205, 0.0845]], + device='cuda:0'), grad: tensor([[ 3.1665e-08, 5.6386e-05, 8.8215e-06, ..., 9.4250e-07, + 7.0238e-04, 2.2531e-05], + [ 4.5449e-07, 6.5279e-04, -6.3515e-03, ..., 2.3231e-05, + 2.1000e-03, 4.9397e-06], + [ 2.5518e-07, 1.2815e-04, 1.6654e-04, ..., 1.6999e-04, + 2.7962e-03, 1.1660e-05], + ..., + [ 1.2051e-06, -8.7833e-04, 1.1057e-04, ..., 1.6481e-05, + 1.5821e-03, 9.7632e-05], + [ 4.8578e-06, 3.6979e-04, -4.4674e-05, ..., -2.8539e-04, + -1.3901e-02, 1.3769e-04], + [ 1.1299e-02, -5.6887e-04, 4.8103e-03, ..., 5.8532e-05, + 7.7105e-04, 3.2735e-04]], device='cuda:0') +Epoch 335, bias, value: tensor([-0.0506, 0.0226, 0.0067, -0.0135, -0.0063, 0.0077, -0.0138, 0.0333, + -0.0179, 0.0123], device='cuda:0'), grad: tensor([-0.0263, -0.0352, 0.0159, 0.0083, -0.0044, 0.0073, 0.0256, -0.0028, + -0.0205, 0.0321], device='cuda:0') +100 +0.0001 +changing lr +epoch 334, time 225.77, cls_loss 0.5360 cls_loss_mapping 0.0035 cls_loss_causal 0.4720 re_mapping 0.0096 re_causal 0.0218 /// teacc 98.84 lr 0.00010000 +Epoch 336, weight, value: tensor([[-0.1982, -0.0168, 0.0312, ..., -0.0265, -0.0801, -0.1350], + [-0.0542, -0.0996, 0.0373, ..., 0.0487, -0.0192, -0.0936], + [-0.0577, -0.0850, -0.1097, ..., 0.0401, -0.0361, -0.1063], + ..., + [-0.0700, 0.0309, 0.0278, ..., 0.0216, -0.0590, -0.0958], + [-0.0788, -0.0019, 0.0225, ..., 0.0694, -0.0373, -0.1743], + [ 0.0636, 0.0624, -0.0548, ..., -0.1068, -0.0205, 0.0858]], + device='cuda:0'), grad: tensor([[ 8.5354e-04, 3.8934e-04, 1.2426e-03, ..., 1.0147e-03, + 7.0763e-04, 1.8568e-03], + [ 1.2434e-04, -6.4850e-04, -4.2572e-03, ..., -1.0366e-03, + 8.0204e-04, -5.4436e-03], + [ 6.8474e-04, 8.2636e-04, 1.2169e-03, ..., 1.5726e-03, + 8.3237e-03, 1.8873e-03], + ..., + [ 8.9407e-04, -9.8419e-04, -6.8545e-05, ..., 9.0313e-04, + 2.0370e-03, 1.8511e-03], + [-2.0428e-03, -4.8614e-04, 9.2328e-05, ..., -3.1090e-03, + 1.6575e-03, -1.5163e-03], + [ 1.8101e-03, 8.8358e-04, -4.8137e-04, ..., 3.5896e-03, + 6.9427e-04, 1.0574e-04]], device='cuda:0') +Epoch 336, bias, value: tensor([-0.0505, 0.0220, 0.0066, -0.0146, -0.0059, 0.0079, -0.0133, 0.0331, + -0.0175, 0.0128], device='cuda:0'), grad: tensor([ 0.0227, -0.0436, 0.0369, -0.0001, 0.0170, -0.0358, 0.0001, 0.0254, + -0.0349, 0.0122], device='cuda:0') +100 +0.0001 +changing lr +epoch 335, time 227.48, cls_loss 0.5316 cls_loss_mapping 0.0034 cls_loss_causal 0.4650 re_mapping 0.0096 re_causal 0.0226 /// teacc 98.85 lr 0.00010000 +Epoch 337, weight, value: tensor([[-0.1980, -0.0161, 0.0298, ..., -0.0262, -0.0789, -0.1335], + [-0.0544, -0.0986, 0.0364, ..., 0.0476, -0.0194, -0.0961], + [-0.0575, -0.0832, -0.1098, ..., 0.0391, -0.0372, -0.1068], + ..., + [-0.0717, 0.0303, 0.0282, ..., 0.0222, -0.0594, -0.0966], + [-0.0785, -0.0026, 0.0223, ..., 0.0694, -0.0374, -0.1735], + [ 0.0643, 0.0624, -0.0540, ..., -0.1068, -0.0216, 0.0846]], + device='cuda:0'), grad: tensor([[ 4.5824e-04, 1.1420e-04, 8.9705e-05, ..., 4.1991e-05, + 1.2174e-05, 6.8235e-04], + [-4.3449e-03, 6.0230e-05, 1.1778e-04, ..., 2.9588e-04, + 9.9754e-04, -3.2616e-03], + [ 6.5565e-04, 7.2598e-05, 1.2693e-03, ..., 9.6130e-04, + 3.0003e-03, 9.8705e-04], + ..., + [-8.5783e-04, 3.4094e-05, 2.8759e-05, ..., 1.4700e-05, + 3.5226e-05, -2.7218e-03], + [ 4.3988e-04, -1.1997e-03, 3.2854e-04, ..., -7.5847e-06, + -3.3426e-04, -5.8842e-04], + [ 7.2384e-04, 3.5930e-04, 2.0158e-04, ..., 2.3335e-05, + 4.0650e-05, 1.5593e-03]], device='cuda:0') +Epoch 337, bias, value: tensor([-0.0494, 0.0211, 0.0073, -0.0146, -0.0059, 0.0075, -0.0127, 0.0332, + -0.0180, 0.0123], device='cuda:0'), grad: tensor([ 0.0102, -0.0354, 0.0288, -0.0141, -0.0249, 0.0167, 0.0310, -0.0233, + 0.0010, 0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 336, time 227.31, cls_loss 0.5130 cls_loss_mapping 0.0029 cls_loss_causal 0.4581 re_mapping 0.0096 re_causal 0.0227 /// teacc 98.81 lr 0.00010000 +Epoch 338, weight, value: tensor([[-0.1989, -0.0163, 0.0292, ..., -0.0254, -0.0805, -0.1339], + [-0.0543, -0.0998, 0.0354, ..., 0.0471, -0.0196, -0.0956], + [-0.0571, -0.0844, -0.1094, ..., 0.0394, -0.0363, -0.1073], + ..., + [-0.0733, 0.0306, 0.0281, ..., 0.0216, -0.0594, -0.0960], + [-0.0789, -0.0019, 0.0231, ..., 0.0693, -0.0371, -0.1741], + [ 0.0655, 0.0632, -0.0547, ..., -0.1065, -0.0212, 0.0843]], + device='cuda:0'), grad: tensor([[-9.9480e-05, 5.7101e-05, 2.8539e-04, ..., 1.3244e-04, + 7.3528e-04, -2.0206e-04], + [-2.8825e-04, -1.6415e-04, 5.0449e-04, ..., 4.6444e-04, + 1.8892e-03, 3.0667e-05], + [ 1.4222e-04, 9.4414e-05, 1.6699e-03, ..., 9.9754e-04, + 2.3117e-03, 1.0419e-04], + ..., + [-1.2337e-02, -2.0885e-04, -3.6263e-04, ..., -1.2970e-04, + -7.4844e-03, -8.8272e-03], + [ 1.2040e-04, -1.6809e-05, 1.0071e-03, ..., 6.8378e-04, + 1.0433e-03, -7.1287e-05], + [ 1.1505e-02, -1.0471e-03, 1.3580e-03, ..., 1.5574e-03, + -3.2234e-04, 8.1787e-03]], device='cuda:0') +Epoch 338, bias, value: tensor([-0.0509, 0.0202, 0.0074, -0.0149, -0.0042, 0.0067, -0.0134, 0.0333, + -0.0165, 0.0128], device='cuda:0'), grad: tensor([ 0.0076, 0.0230, 0.0030, -0.0110, 0.0249, -0.0225, 0.0107, -0.0269, + 0.0008, -0.0098], device='cuda:0') +100 +0.0001 +changing lr +epoch 337, time 227.02, cls_loss 0.5071 cls_loss_mapping 0.0023 cls_loss_causal 0.4401 re_mapping 0.0093 re_causal 0.0217 /// teacc 98.96 lr 0.00010000 +Epoch 339, weight, value: tensor([[-0.1988, -0.0163, 0.0287, ..., -0.0249, -0.0817, -0.1342], + [-0.0545, -0.0999, 0.0367, ..., 0.0471, -0.0190, -0.0959], + [-0.0573, -0.0851, -0.1092, ..., 0.0383, -0.0376, -0.1079], + ..., + [-0.0720, 0.0304, 0.0273, ..., 0.0220, -0.0595, -0.0941], + [-0.0780, -0.0012, 0.0228, ..., 0.0693, -0.0368, -0.1753], + [ 0.0651, 0.0625, -0.0543, ..., -0.1061, -0.0192, 0.0836]], + device='cuda:0'), grad: tensor([[ 2.5004e-05, -6.9809e-03, 4.1910e-07, ..., 1.1884e-05, + 9.8133e-04, -3.1624e-03], + [ 1.4819e-05, -2.7294e-03, 7.6368e-07, ..., 1.0580e-05, + 2.2297e-03, 5.5170e-04], + [ 1.4210e-04, 6.7186e-04, 1.8120e-05, ..., 1.3620e-05, + 1.1330e-03, 3.3903e-04], + ..., + [ 1.6937e-03, 7.0496e-03, 6.0463e-04, ..., -3.1441e-05, + 2.5826e-03, 3.8376e-03], + [ 5.0306e-04, -1.0948e-03, 2.3581e-06, ..., 1.6779e-05, + 1.5974e-03, -1.0252e-03], + [-1.6356e-03, 6.2799e-04, -6.1703e-04, ..., 9.8497e-06, + -5.6114e-03, -1.0996e-03]], device='cuda:0') +Epoch 339, bias, value: tensor([-0.0500, 0.0209, 0.0060, -0.0148, -0.0041, 0.0076, -0.0134, 0.0327, + -0.0168, 0.0126], device='cuda:0'), grad: tensor([-0.0490, -0.0054, 0.0141, 0.0114, 0.0049, 0.0117, -0.0338, 0.0530, + 0.0074, -0.0142], device='cuda:0') +100 +0.0001 +changing lr +epoch 338, time 229.03, cls_loss 0.5403 cls_loss_mapping 0.0029 cls_loss_causal 0.4715 re_mapping 0.0088 re_causal 0.0214 /// teacc 98.94 lr 0.00010000 +Epoch 340, weight, value: tensor([[-0.1982, -0.0163, 0.0278, ..., -0.0250, -0.0812, -0.1344], + [-0.0549, -0.0987, 0.0370, ..., 0.0467, -0.0198, -0.0956], + [-0.0562, -0.0852, -0.1088, ..., 0.0395, -0.0375, -0.1085], + ..., + [-0.0718, 0.0296, 0.0273, ..., 0.0224, -0.0599, -0.0940], + [-0.0782, -0.0018, 0.0231, ..., 0.0692, -0.0368, -0.1744], + [ 0.0661, 0.0622, -0.0537, ..., -0.1062, -0.0195, 0.0841]], + device='cuda:0'), grad: tensor([[ 4.6611e-04, 2.4766e-05, 1.0891e-03, ..., 4.7755e-04, + 1.0939e-03, 4.0960e-04], + [ 6.9082e-05, 4.7708e-04, -1.0841e-02, ..., -1.3971e-03, + -5.6877e-03, 5.2333e-05], + [ 1.6427e-04, -1.4048e-03, 4.5156e-04, ..., -1.4887e-03, + -5.0426e-05, 1.4925e-04], + ..., + [ 1.0433e-03, 3.5191e-04, 4.4861e-03, ..., 6.1464e-04, + 2.8419e-03, 6.3324e-04], + [ 1.2312e-03, 4.1366e-04, 4.1962e-03, ..., 2.1019e-03, + -3.6449e-03, 1.2217e-03], + [-2.3727e-03, 9.3699e-05, -2.5673e-03, ..., -1.8139e-03, + 1.0586e-03, -1.5335e-03]], device='cuda:0') +Epoch 340, bias, value: tensor([-0.0500, 0.0208, 0.0056, -0.0143, -0.0038, 0.0075, -0.0144, 0.0329, + -0.0170, 0.0132], device='cuda:0'), grad: tensor([ 0.0095, -0.0129, -0.0033, -0.0085, 0.0187, -0.0056, 0.0134, -0.0053, + 0.0018, -0.0078], device='cuda:0') +100 +0.0001 +changing lr +epoch 339, time 231.10, cls_loss 0.5265 cls_loss_mapping 0.0035 cls_loss_causal 0.4578 re_mapping 0.0093 re_causal 0.0218 /// teacc 98.68 lr 0.00010000 +Epoch 341, weight, value: tensor([[-0.1980, -0.0166, 0.0273, ..., -0.0250, -0.0829, -0.1339], + [-0.0539, -0.0998, 0.0376, ..., 0.0466, -0.0192, -0.0945], + [-0.0571, -0.0854, -0.1079, ..., 0.0400, -0.0368, -0.1100], + ..., + [-0.0719, 0.0309, 0.0267, ..., 0.0223, -0.0603, -0.0946], + [-0.0772, -0.0017, 0.0222, ..., 0.0691, -0.0364, -0.1754], + [ 0.0643, 0.0616, -0.0547, ..., -0.1062, -0.0194, 0.0844]], + device='cuda:0'), grad: tensor([[ 3.2485e-05, 4.3772e-07, 2.7561e-04, ..., 1.5533e-04, + 8.0490e-04, 5.1916e-05], + [ 9.4235e-05, 8.4639e-06, 4.1723e-04, ..., -1.0818e-04, + 9.3985e-04, 1.3733e-04], + [-1.7033e-03, 5.8979e-05, 5.0640e-04, ..., 2.7680e-04, + 6.7282e-04, 6.1572e-05], + ..., + [ 5.5504e-04, -1.1003e-04, 3.7742e-04, ..., 1.1630e-05, + 1.1244e-03, 7.6056e-05], + [ 2.5821e-04, 2.0266e-06, -5.2404e-04, ..., -5.0640e-04, + 1.0548e-03, 3.9315e-04], + [ 4.8317e-06, 3.9011e-05, -2.7161e-03, ..., 8.6278e-06, + -8.3466e-03, 9.3877e-05]], device='cuda:0') +Epoch 341, bias, value: tensor([-0.0504, 0.0215, 0.0064, -0.0152, -0.0034, 0.0074, -0.0144, 0.0335, + -0.0182, 0.0133], device='cuda:0'), grad: tensor([ 0.0133, 0.0219, -0.0163, 0.0358, -0.0114, -0.0037, 0.0198, 0.0170, + -0.0312, -0.0452], device='cuda:0') +100 +0.0001 +changing lr +epoch 340, time 228.48, cls_loss 0.5410 cls_loss_mapping 0.0029 cls_loss_causal 0.4828 re_mapping 0.0095 re_causal 0.0226 /// teacc 98.78 lr 0.00010000 +Epoch 342, weight, value: tensor([[-0.1982, -0.0169, 0.0280, ..., -0.0257, -0.0818, -0.1349], + [-0.0530, -0.0997, 0.0383, ..., 0.0464, -0.0178, -0.0952], + [-0.0563, -0.0853, -0.1095, ..., 0.0397, -0.0358, -0.1095], + ..., + [-0.0714, 0.0315, 0.0264, ..., 0.0231, -0.0603, -0.0950], + [-0.0777, -0.0020, 0.0215, ..., 0.0688, -0.0369, -0.1757], + [ 0.0636, 0.0616, -0.0547, ..., -0.1061, -0.0182, 0.0847]], + device='cuda:0'), grad: tensor([[ 2.4751e-05, -1.4031e-04, -6.9237e-04, ..., 1.6630e-05, + -4.8561e-03, 2.4378e-05], + [-3.4809e-04, 1.9169e-04, 5.2261e-04, ..., -3.4714e-04, + 2.6512e-03, 5.5403e-05], + [ 1.3900e-04, 2.4772e-04, 1.8282e-03, ..., 1.1462e-04, + 1.1978e-03, 7.3969e-05], + ..., + [ 3.2158e-03, 1.4639e-03, 5.8060e-03, ..., 7.8058e-04, + 1.5421e-03, 2.7599e-03], + [ 1.0461e-04, 1.1606e-03, -3.4561e-03, ..., 6.4254e-05, + 1.1969e-03, 2.3115e-04], + [-9.3155e-03, -1.8127e-02, -3.8338e-03, ..., -8.0872e-04, + -4.9744e-03, -2.7733e-03]], device='cuda:0') +Epoch 342, bias, value: tensor([-0.0507, 0.0228, 0.0070, -0.0156, -0.0031, 0.0064, -0.0154, 0.0343, + -0.0183, 0.0130], device='cuda:0'), grad: tensor([-0.0235, 0.0207, 0.0197, -0.0204, 0.0101, -0.0154, 0.0148, 0.0487, + -0.0079, -0.0468], device='cuda:0') +100 +0.0001 +changing lr +epoch 341, time 228.03, cls_loss 0.5308 cls_loss_mapping 0.0040 cls_loss_causal 0.4690 re_mapping 0.0092 re_causal 0.0214 /// teacc 98.71 lr 0.00010000 +Epoch 343, weight, value: tensor([[-0.1967, -0.0167, 0.0288, ..., -0.0253, -0.0826, -0.1349], + [-0.0542, -0.1003, 0.0368, ..., 0.0462, -0.0173, -0.0947], + [-0.0592, -0.0859, -0.1078, ..., 0.0399, -0.0350, -0.1099], + ..., + [-0.0699, 0.0312, 0.0262, ..., 0.0230, -0.0593, -0.0964], + [-0.0771, -0.0022, 0.0223, ..., 0.0689, -0.0370, -0.1767], + [ 0.0638, 0.0621, -0.0556, ..., -0.1057, -0.0190, 0.0856]], + device='cuda:0'), grad: tensor([[ 1.5843e-04, 4.3571e-05, 7.0953e-04, ..., 5.1594e-04, + 4.8780e-04, 2.7561e-04], + [ 1.7321e-04, 3.9101e-05, 1.3571e-03, ..., 7.0333e-04, + 3.4976e-04, 4.6682e-04], + [ 4.6825e-04, 3.2097e-05, 2.4471e-03, ..., 1.7776e-03, + 1.1597e-03, 9.7036e-04], + ..., + [ 1.2884e-03, -2.4738e-03, 3.7270e-03, ..., 3.2024e-03, + 2.4109e-03, 1.6403e-03], + [-2.9945e-03, 1.5870e-05, -7.5951e-03, ..., -7.1564e-03, + -5.9662e-03, -3.5343e-03], + [ 9.1255e-05, 2.3918e-03, 1.1148e-03, ..., 6.3038e-04, + 9.0790e-04, 2.2352e-04]], device='cuda:0') +Epoch 343, bias, value: tensor([-0.0503, 0.0221, 0.0078, -0.0148, -0.0023, 0.0068, -0.0155, 0.0325, + -0.0188, 0.0130], device='cuda:0'), grad: tensor([-0.0047, -0.0123, 0.0205, -0.0309, -0.0128, -0.0176, 0.0132, 0.0233, + -0.0159, 0.0372], device='cuda:0') +100 +0.0001 +changing lr +epoch 342, time 228.04, cls_loss 0.5373 cls_loss_mapping 0.0031 cls_loss_causal 0.4776 re_mapping 0.0093 re_causal 0.0215 /// teacc 98.84 lr 0.00010000 +Epoch 344, weight, value: tensor([[-0.1977, -0.0173, 0.0284, ..., -0.0260, -0.0836, -0.1356], + [-0.0546, -0.1008, 0.0360, ..., 0.0457, -0.0193, -0.0949], + [-0.0594, -0.0852, -0.1078, ..., 0.0398, -0.0353, -0.1104], + ..., + [-0.0696, 0.0321, 0.0269, ..., 0.0224, -0.0593, -0.0955], + [-0.0768, -0.0014, 0.0219, ..., 0.0699, -0.0358, -0.1771], + [ 0.0644, 0.0612, -0.0545, ..., -0.1052, -0.0173, 0.0853]], + device='cuda:0'), grad: tensor([[ 2.2501e-05, 4.0627e-04, 1.2100e-04, ..., 0.0000e+00, + 1.0900e-03, 9.6262e-06], + [ 2.8992e-04, 1.0777e-03, 1.5116e-04, ..., 0.0000e+00, + 2.1152e-03, 2.0957e-04], + [-3.3140e-04, 8.8310e-04, 3.5822e-05, ..., 0.0000e+00, + 2.8706e-03, 9.6798e-05], + ..., + [ 3.6526e-04, -3.8567e-03, 1.1790e-04, ..., 0.0000e+00, + -1.4572e-02, 1.8024e-04], + [ 3.2806e-04, 5.4407e-04, 4.4918e-04, ..., 0.0000e+00, + 1.2093e-03, 1.6022e-04], + [ 1.4472e-04, 1.3908e-02, 7.8583e-04, ..., 0.0000e+00, + 1.9455e-03, 7.8201e-05]], device='cuda:0') +Epoch 344, bias, value: tensor([-0.0505, 0.0211, 0.0083, -0.0146, -0.0034, 0.0069, -0.0154, 0.0332, + -0.0191, 0.0142], device='cuda:0'), grad: tensor([ 0.0046, -0.0187, 0.0066, -0.0037, -0.0239, 0.0099, 0.0056, -0.0125, + 0.0084, 0.0236], device='cuda:0') +100 +0.0001 +changing lr +epoch 343, time 228.02, cls_loss 0.4975 cls_loss_mapping 0.0031 cls_loss_causal 0.4441 re_mapping 0.0095 re_causal 0.0211 /// teacc 98.96 lr 0.00010000 +Epoch 345, weight, value: tensor([[-0.1984, -0.0174, 0.0285, ..., -0.0266, -0.0824, -0.1374], + [-0.0526, -0.1015, 0.0361, ..., 0.0479, -0.0189, -0.0951], + [-0.0594, -0.0845, -0.1079, ..., 0.0405, -0.0369, -0.1106], + ..., + [-0.0696, 0.0325, 0.0279, ..., 0.0214, -0.0589, -0.0956], + [-0.0778, -0.0008, 0.0216, ..., 0.0685, -0.0352, -0.1772], + [ 0.0658, 0.0612, -0.0551, ..., -0.1049, -0.0178, 0.0851]], + device='cuda:0'), grad: tensor([[ 1.6034e-04, 1.3798e-05, 7.2300e-05, ..., 1.7345e-04, + 2.5892e-04, 3.4928e-04], + [ 1.1081e-04, 1.6809e-04, 2.2396e-05, ..., -4.8089e-04, + 5.0592e-04, -1.1482e-03], + [ 1.6327e-03, 2.4281e-03, -3.2902e-05, ..., 1.9703e-03, + 1.1559e-02, 1.2207e-04], + ..., + [ 2.0065e-03, 3.8834e-03, 5.4741e-04, ..., 2.0733e-03, + 6.1913e-03, 1.2693e-03], + [ 1.6890e-03, -4.6120e-03, 3.2425e-04, ..., 2.3155e-03, + -9.6436e-03, 5.1689e-04], + [ 1.0166e-03, -4.0398e-03, 3.1543e-04, ..., 4.9877e-04, + -6.2790e-03, 7.2193e-04]], device='cuda:0') +Epoch 345, bias, value: tensor([-0.0497, 0.0217, 0.0088, -0.0142, -0.0036, 0.0054, -0.0153, 0.0329, + -0.0186, 0.0131], device='cuda:0'), grad: tensor([ 0.0071, -0.0257, -0.0017, -0.0260, 0.0120, 0.0065, 0.0098, 0.0312, + 0.0122, -0.0252], device='cuda:0') +100 +0.0001 +changing lr +epoch 344, time 228.16, cls_loss 0.5412 cls_loss_mapping 0.0031 cls_loss_causal 0.4815 re_mapping 0.0097 re_causal 0.0232 /// teacc 98.88 lr 0.00010000 +Epoch 346, weight, value: tensor([[-0.1992, -0.0176, 0.0292, ..., -0.0268, -0.0811, -0.1390], + [-0.0538, -0.1013, 0.0359, ..., 0.0479, -0.0179, -0.0968], + [-0.0593, -0.0853, -0.1084, ..., 0.0406, -0.0385, -0.1108], + ..., + [-0.0702, 0.0321, 0.0284, ..., 0.0209, -0.0584, -0.0960], + [-0.0762, -0.0012, 0.0220, ..., 0.0686, -0.0364, -0.1764], + [ 0.0652, 0.0618, -0.0555, ..., -0.1050, -0.0168, 0.0862]], + device='cuda:0'), grad: tensor([[ 8.8215e-05, 6.2406e-05, 1.0033e-03, ..., 4.7421e-04, + 4.3130e-04, 1.9038e-04], + [ 6.1005e-05, 7.3075e-05, -6.8045e-04, ..., 1.1530e-03, + 1.5688e-03, 1.0985e-04], + [-2.8934e-03, 1.5962e-04, 7.7295e-04, ..., -8.9169e-04, + 2.0199e-03, -2.0027e-03], + ..., + [ 7.2420e-05, 8.2612e-05, 3.0589e-04, ..., 3.1805e-04, + 3.1519e-04, 8.2970e-05], + [ 1.5678e-03, 2.3699e-04, -4.0627e-03, ..., -6.8569e-04, + -9.4366e-04, 2.7943e-04], + [ 8.9526e-05, 6.4850e-04, 7.2432e-04, ..., 3.7432e-04, + 3.7956e-04, 1.0955e-04]], device='cuda:0') +Epoch 346, bias, value: tensor([-0.0493, 0.0206, 0.0082, -0.0137, -0.0043, 0.0051, -0.0143, 0.0324, + -0.0188, 0.0145], device='cuda:0'), grad: tensor([ 0.0156, 0.0163, -0.0169, 0.0243, 0.0468, -0.0222, -0.0387, 0.0111, + -0.0262, -0.0100], device='cuda:0') +100 +0.0001 +changing lr +epoch 345, time 228.53, cls_loss 0.5438 cls_loss_mapping 0.0033 cls_loss_causal 0.4906 re_mapping 0.0092 re_causal 0.0221 /// teacc 98.80 lr 0.00010000 +Epoch 347, weight, value: tensor([[-0.1985, -0.0171, 0.0294, ..., -0.0266, -0.0818, -0.1366], + [-0.0536, -0.1022, 0.0364, ..., 0.0474, -0.0174, -0.0980], + [-0.0608, -0.0855, -0.1084, ..., 0.0408, -0.0378, -0.1109], + ..., + [-0.0700, 0.0328, 0.0279, ..., 0.0206, -0.0574, -0.0964], + [-0.0774, -0.0014, 0.0218, ..., 0.0685, -0.0377, -0.1770], + [ 0.0646, 0.0618, -0.0554, ..., -0.1041, -0.0169, 0.0870]], + device='cuda:0'), grad: tensor([[ 6.9904e-04, 7.7903e-05, 2.7752e-04, ..., 4.0364e-04, + 9.5749e-04, 5.7650e-04], + [ 9.5367e-05, 5.8860e-05, 1.0765e-04, ..., 1.3006e-04, + 1.3523e-03, 7.7665e-05], + [-1.1702e-03, 7.6675e-04, -2.0828e-03, ..., -2.6646e-03, + 1.0443e-03, -4.6468e-04], + ..., + [ 4.2558e-04, 2.2316e-04, 5.5504e-04, ..., 6.4945e-04, + -3.1490e-03, 3.3164e-04], + [-1.9398e-03, -1.7258e-02, 4.1366e-04, ..., 3.7408e-04, + -1.0406e-02, -4.9114e-04], + [ 2.2717e-03, 8.4915e-03, 9.6846e-04, ..., 9.2745e-04, + 4.1428e-03, 1.7996e-03]], device='cuda:0') +Epoch 347, bias, value: tensor([-0.0500, 0.0209, 0.0084, -0.0130, -0.0045, 0.0060, -0.0150, 0.0326, + -0.0191, 0.0143], device='cuda:0'), grad: tensor([ 0.0051, 0.0191, 0.0123, -0.0244, 0.0260, 0.0012, 0.0222, -0.0076, + -0.0695, 0.0157], device='cuda:0') +100 +0.0001 +changing lr +epoch 346, time 228.27, cls_loss 0.5124 cls_loss_mapping 0.0027 cls_loss_causal 0.4477 re_mapping 0.0094 re_causal 0.0219 /// teacc 98.89 lr 0.00010000 +Epoch 348, weight, value: tensor([[-0.1984, -0.0169, 0.0292, ..., -0.0266, -0.0828, -0.1352], + [-0.0543, -0.1026, 0.0352, ..., 0.0479, -0.0177, -0.0990], + [-0.0625, -0.0868, -0.1090, ..., 0.0406, -0.0376, -0.1110], + ..., + [-0.0692, 0.0332, 0.0283, ..., 0.0208, -0.0566, -0.0966], + [-0.0781, -0.0008, 0.0234, ..., 0.0683, -0.0373, -0.1777], + [ 0.0640, 0.0618, -0.0549, ..., -0.1044, -0.0169, 0.0873]], + device='cuda:0'), grad: tensor([[ 1.1051e-04, 7.5400e-06, 1.3959e-04, ..., 1.0008e-04, + 9.4831e-05, 1.0276e-04], + [ 5.1498e-04, 6.9179e-06, 1.1330e-03, ..., 3.6299e-05, + 3.3855e-05, 5.4741e-04], + [ 1.3518e-04, 2.6703e-05, 1.7071e-04, ..., 1.2124e-04, + 1.1426e-04, 1.2553e-04], + ..., + [ 3.4022e-04, -3.0249e-06, 5.3215e-04, ..., 2.2662e-04, + 2.1422e-04, 3.3140e-04], + [ 3.7169e-04, -3.1352e-04, 7.3814e-04, ..., 3.2187e-04, + 3.1877e-04, 3.6883e-04], + [-3.3593e-04, 9.8944e-06, -1.4076e-03, ..., 3.6645e-04, + 3.4928e-04, -4.4513e-04]], device='cuda:0') +Epoch 348, bias, value: tensor([-0.0506, 0.0207, 0.0081, -0.0132, -0.0048, 0.0053, -0.0139, 0.0337, + -0.0191, 0.0142], device='cuda:0'), grad: tensor([-0.0211, 0.0226, 0.0131, 0.0214, -0.0480, 0.0227, 0.0194, -0.0148, + 0.0359, -0.0510], device='cuda:0') +100 +0.0001 +changing lr +epoch 347, time 229.13, cls_loss 0.5225 cls_loss_mapping 0.0035 cls_loss_causal 0.4688 re_mapping 0.0095 re_causal 0.0219 /// teacc 98.72 lr 0.00010000 +Epoch 349, weight, value: tensor([[-0.1997, -0.0171, 0.0303, ..., -0.0262, -0.0824, -0.1370], + [-0.0550, -0.1025, 0.0346, ..., 0.0481, -0.0171, -0.0991], + [-0.0614, -0.0869, -0.1086, ..., 0.0412, -0.0368, -0.1104], + ..., + [-0.0685, 0.0341, 0.0273, ..., 0.0207, -0.0578, -0.0964], + [-0.0777, -0.0010, 0.0220, ..., 0.0677, -0.0351, -0.1789], + [ 0.0630, 0.0619, -0.0551, ..., -0.1040, -0.0174, 0.0875]], + device='cuda:0'), grad: tensor([[ 1.3757e-04, 9.1612e-05, 1.7548e-04, ..., 2.4259e-04, + -6.1941e-04, 8.0109e-05], + [ 6.1933e-07, 3.4779e-05, 2.3913e-04, ..., 3.8767e-04, + 1.1432e-04, 2.9892e-05], + [ 1.7583e-05, 3.4642e-04, -1.4365e-04, ..., 5.8594e-03, + 5.8937e-04, 5.1260e-06], + ..., + [ 1.3605e-05, -1.3774e-06, 2.0117e-06, ..., -1.3123e-03, + 2.1279e-04, -2.0534e-05], + [ 4.7475e-05, -1.4687e-03, 7.9918e-04, ..., 2.7919e-04, + -3.6564e-03, -4.0460e-04], + [-2.4724e-04, 1.3101e-04, -1.4842e-04, ..., 2.2984e-04, + 5.8651e-04, 9.3281e-05]], device='cuda:0') +Epoch 349, bias, value: tensor([-0.0498, 0.0207, 0.0078, -0.0145, -0.0051, 0.0049, -0.0132, 0.0339, + -0.0187, 0.0144], device='cuda:0'), grad: tensor([-0.0061, 0.0085, 0.0115, 0.0306, 0.0202, 0.0157, -0.0207, -0.0220, + -0.0193, -0.0183], device='cuda:0') +100 +0.0001 +changing lr +epoch 348, time 229.26, cls_loss 0.5389 cls_loss_mapping 0.0043 cls_loss_causal 0.4746 re_mapping 0.0091 re_causal 0.0207 /// teacc 98.79 lr 0.00010000 +Epoch 350, weight, value: tensor([[-0.1997, -0.0168, 0.0298, ..., -0.0263, -0.0827, -0.1378], + [-0.0550, -0.1018, 0.0354, ..., 0.0470, -0.0162, -0.0995], + [-0.0623, -0.0871, -0.1085, ..., 0.0409, -0.0376, -0.1100], + ..., + [-0.0689, 0.0350, 0.0296, ..., 0.0215, -0.0581, -0.0961], + [-0.0779, -0.0022, 0.0218, ..., 0.0682, -0.0345, -0.1796], + [ 0.0639, 0.0627, -0.0558, ..., -0.1043, -0.0162, 0.0879]], + device='cuda:0'), grad: tensor([[ 8.1003e-05, -3.1528e-03, -1.7166e-03, ..., -1.3142e-03, + 6.9141e-04, 5.7518e-06], + [ 6.5342e-06, 5.5790e-04, -8.2321e-03, ..., -2.0676e-03, + 1.9097e-04, 7.5772e-06], + [ 1.5182e-03, 1.8406e-04, 3.3951e-04, ..., 1.0719e-03, + 3.3455e-03, 1.8016e-05], + ..., + [ 6.1929e-05, 2.3782e-04, 1.8764e-04, ..., 1.5593e-04, + 1.4257e-03, 1.0246e-04], + [-2.4204e-03, 2.0065e-03, 4.5109e-04, ..., -1.0595e-03, + -5.4665e-03, 1.0109e-04], + [ 4.5824e-04, -4.6158e-03, 2.1386e-04, ..., 1.2970e-04, + 1.4906e-03, -2.8896e-03]], device='cuda:0') +Epoch 350, bias, value: tensor([-0.0495, 0.0212, 0.0079, -0.0133, -0.0060, 0.0049, -0.0138, 0.0327, + -0.0185, 0.0149], device='cuda:0'), grad: tensor([-0.0228, 0.0133, -0.0099, -0.0160, 0.0040, 0.0114, 0.0399, 0.0129, + -0.0203, -0.0125], device='cuda:0') +100 +0.0001 +changing lr +epoch 349, time 228.07, cls_loss 0.5188 cls_loss_mapping 0.0022 cls_loss_causal 0.4577 re_mapping 0.0095 re_causal 0.0218 /// teacc 98.94 lr 0.00010000 +Epoch 351, weight, value: tensor([[-0.2001, -0.0167, 0.0297, ..., -0.0260, -0.0834, -0.1379], + [-0.0549, -0.1028, 0.0357, ..., 0.0477, -0.0157, -0.0987], + [-0.0620, -0.0877, -0.1085, ..., 0.0401, -0.0372, -0.1088], + ..., + [-0.0690, 0.0342, 0.0300, ..., 0.0221, -0.0587, -0.0966], + [-0.0786, -0.0027, 0.0215, ..., 0.0675, -0.0346, -0.1800], + [ 0.0636, 0.0640, -0.0547, ..., -0.1041, -0.0168, 0.0884]], + device='cuda:0'), grad: tensor([[-1.3363e-04, 3.3259e-04, 2.8825e-04, ..., -7.3254e-05, + 2.6631e-04, -4.7833e-05], + [ 1.5116e-04, 5.7983e-04, 7.9930e-05, ..., -6.1810e-05, + 5.4628e-05, 3.0324e-05], + [ 3.1853e-04, 6.5851e-04, 1.0157e-03, ..., 3.7193e-04, + 3.6478e-04, 2.0111e-04], + ..., + [ 3.9864e-04, 4.7798e-03, 1.8950e-03, ..., 2.2137e-04, + 1.4246e-04, 1.4153e-03], + [-7.5684e-03, 1.3390e-03, -9.3222e-04, ..., -2.3460e-03, + -5.9547e-03, -1.1721e-03], + [-1.2064e-03, 1.3447e-03, 8.6403e-04, ..., -1.3340e-04, + 1.5268e-03, 1.2183e-04]], device='cuda:0') +Epoch 351, bias, value: tensor([-0.0508, 0.0209, 0.0077, -0.0133, -0.0057, 0.0066, -0.0138, 0.0331, + -0.0189, 0.0146], device='cuda:0'), grad: tensor([ 5.9013e-03, -1.9958e-02, 1.3397e-02, 9.2545e-03, -3.6011e-02, + -9.3155e-03, 7.2420e-05, 3.0457e-02, -9.7275e-03, 1.5961e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 350, time 229.27, cls_loss 0.5169 cls_loss_mapping 0.0038 cls_loss_causal 0.4556 re_mapping 0.0091 re_causal 0.0205 /// teacc 98.87 lr 0.00010000 +Epoch 352, weight, value: tensor([[-0.1999, -0.0192, 0.0295, ..., -0.0273, -0.0847, -0.1389], + [-0.0556, -0.1006, 0.0359, ..., 0.0487, -0.0168, -0.0957], + [-0.0629, -0.0883, -0.1081, ..., 0.0393, -0.0376, -0.1092], + ..., + [-0.0692, 0.0346, 0.0296, ..., 0.0226, -0.0588, -0.0969], + [-0.0790, -0.0022, 0.0213, ..., 0.0675, -0.0343, -0.1814], + [ 0.0655, 0.0639, -0.0545, ..., -0.1062, -0.0161, 0.0887]], + device='cuda:0'), grad: tensor([[ 8.5402e-04, 5.9068e-05, 1.5745e-03, ..., 6.8521e-04, + 5.6219e-04, 1.7786e-04], + [ 7.9453e-05, 3.3408e-05, 1.2455e-03, ..., 6.2275e-04, + 3.4595e-04, 1.5354e-04], + [ 5.5599e-04, 2.7195e-05, 1.1082e-03, ..., 6.4707e-04, + 3.9673e-04, 4.4036e-04], + ..., + [ 6.7115e-05, 3.9078e-06, 6.1607e-04, ..., -8.0228e-05, + 4.3058e-04, 3.0935e-05], + [ 1.2140e-03, -7.8976e-05, 7.7546e-05, ..., -1.1647e-04, + 5.1165e-04, 4.4870e-04], + [ 2.3174e-04, 1.1742e-05, -5.0812e-03, ..., -1.9493e-03, + -2.3251e-03, 8.5473e-05]], device='cuda:0') +Epoch 352, bias, value: tensor([-0.0513, 0.0218, 0.0084, -0.0134, -0.0059, 0.0065, -0.0143, 0.0332, + -0.0185, 0.0139], device='cuda:0'), grad: tensor([ 0.0165, 0.0218, 0.0178, 0.0172, 0.0339, 0.0083, -0.0411, -0.0204, + -0.0020, -0.0520], device='cuda:0') +100 +0.0001 +changing lr +epoch 351, time 228.14, cls_loss 0.5024 cls_loss_mapping 0.0029 cls_loss_causal 0.4470 re_mapping 0.0091 re_causal 0.0208 /// teacc 98.93 lr 0.00010000 +Epoch 353, weight, value: tensor([[-0.1992, -0.0195, 0.0289, ..., -0.0278, -0.0857, -0.1387], + [-0.0565, -0.1018, 0.0360, ..., 0.0495, -0.0165, -0.0954], + [-0.0626, -0.0892, -0.1087, ..., 0.0380, -0.0375, -0.1097], + ..., + [-0.0700, 0.0350, 0.0299, ..., 0.0225, -0.0589, -0.0971], + [-0.0798, -0.0021, 0.0223, ..., 0.0670, -0.0352, -0.1824], + [ 0.0656, 0.0641, -0.0552, ..., -0.1047, -0.0156, 0.0889]], + device='cuda:0'), grad: tensor([[ 5.9865e-06, 1.5392e-03, 7.1704e-05, ..., 4.2057e-04, + 4.1795e-04, 2.4045e-04], + [ 3.5353e-06, -1.5850e-03, -5.2661e-05, ..., -1.1075e-04, + -3.4599e-03, 7.1526e-05], + [ 9.3699e-05, -5.2643e-04, -1.0061e-03, ..., -4.9858e-03, + 5.3346e-05, -2.5787e-03], + ..., + [-1.1432e-04, 1.2512e-03, 1.8597e-04, ..., 1.2293e-03, + 6.9761e-04, 3.7527e-04], + [ 1.5035e-05, 2.0447e-02, 1.8847e-04, ..., 1.0815e-03, + 1.0796e-03, 8.3923e-04], + [ 3.8464e-07, -2.1103e-02, 1.0622e-04, ..., 7.3433e-04, + 6.1691e-05, 2.9397e-04]], device='cuda:0') +Epoch 353, bias, value: tensor([-0.0517, 0.0220, 0.0069, -0.0145, -0.0061, 0.0078, -0.0135, 0.0334, + -0.0184, 0.0145], device='cuda:0'), grad: tensor([ 0.0134, -0.0137, -0.0337, -0.0142, -0.0334, 0.0277, 0.0165, 0.0174, + 0.0215, -0.0015], device='cuda:0') +100 +0.0001 +changing lr +epoch 352, time 228.49, cls_loss 0.5412 cls_loss_mapping 0.0031 cls_loss_causal 0.4821 re_mapping 0.0096 re_causal 0.0225 /// teacc 98.79 lr 0.00010000 +Epoch 354, weight, value: tensor([[-0.1970, -0.0192, 0.0300, ..., -0.0280, -0.0844, -0.1390], + [-0.0568, -0.1027, 0.0370, ..., 0.0503, -0.0171, -0.0955], + [-0.0624, -0.0889, -0.1096, ..., 0.0381, -0.0386, -0.1096], + ..., + [-0.0704, 0.0350, 0.0296, ..., 0.0205, -0.0592, -0.0973], + [-0.0810, -0.0028, 0.0220, ..., 0.0665, -0.0347, -0.1834], + [ 0.0654, 0.0632, -0.0537, ..., -0.1039, -0.0165, 0.0898]], + device='cuda:0'), grad: tensor([[ 1.9148e-06, 1.2624e-04, -1.6756e-03, ..., 7.6294e-04, + 6.8605e-05, -4.3464e-04], + [ 1.7703e-05, 1.8132e-04, -6.4492e-05, ..., 1.4591e-03, + 1.0949e-04, 2.5153e-04], + [ 1.0836e-04, 3.5620e-04, 5.4169e-04, ..., -4.9286e-03, + 2.9254e-04, 7.4720e-04], + ..., + [ 5.7161e-05, 9.9850e-04, 1.9526e-04, ..., 1.4126e-05, + 2.2125e-03, 7.4005e-04], + [ 2.2069e-05, 1.2293e-03, 3.9029e-04, ..., -5.2500e-04, + 9.8324e-04, 3.1590e-04], + [-5.2691e-05, -2.0771e-03, 3.5715e-04, ..., 6.3658e-05, + -5.2071e-03, -1.2197e-03]], device='cuda:0') +Epoch 354, bias, value: tensor([-0.0512, 0.0217, 0.0063, -0.0136, -0.0054, 0.0063, -0.0129, 0.0334, + -0.0186, 0.0145], device='cuda:0'), grad: tensor([-0.0132, 0.0298, -0.0419, 0.0189, 0.0238, -0.0069, -0.0016, -0.0039, + 0.0104, -0.0153], device='cuda:0') +100 +0.0001 +changing lr +epoch 353, time 227.51, cls_loss 0.5395 cls_loss_mapping 0.0025 cls_loss_causal 0.4715 re_mapping 0.0095 re_causal 0.0222 /// teacc 98.87 lr 0.00010000 +Epoch 355, weight, value: tensor([[-0.1968, -0.0187, 0.0296, ..., -0.0287, -0.0836, -0.1392], + [-0.0576, -0.1028, 0.0368, ..., 0.0487, -0.0167, -0.0962], + [-0.0625, -0.0895, -0.1090, ..., 0.0387, -0.0392, -0.1102], + ..., + [-0.0676, 0.0346, 0.0290, ..., 0.0203, -0.0586, -0.0963], + [-0.0820, -0.0031, 0.0215, ..., 0.0644, -0.0346, -0.1821], + [ 0.0654, 0.0632, -0.0547, ..., -0.1056, -0.0156, 0.0886]], + device='cuda:0'), grad: tensor([[-1.2566e-02, 2.0862e-06, -8.1100e-03, ..., 1.1861e-05, + -2.0935e-02, 7.8008e-06], + [ 1.0900e-05, 1.9953e-05, 1.3506e-04, ..., 2.3797e-05, + -1.1505e-02, 2.2277e-06], + [ 4.5866e-05, 1.8124e-06, 4.7588e-04, ..., 2.4211e-04, + 1.1526e-05, 2.8104e-05], + ..., + [ 2.4338e-03, 4.8339e-05, 2.2507e-03, ..., 2.0459e-05, + 3.8490e-03, 1.0884e-04], + [ 3.6907e-04, -1.3847e-03, -2.6932e-03, ..., -1.5869e-03, + -6.8009e-05, 3.5977e-04], + [ 1.0132e-02, -2.7954e-02, 7.4844e-03, ..., 8.9556e-06, + 1.7548e-02, -1.8871e-04]], device='cuda:0') +Epoch 355, bias, value: tensor([-0.0515, 0.0210, 0.0082, -0.0137, -0.0045, 0.0068, -0.0138, 0.0332, + -0.0196, 0.0144], device='cuda:0'), grad: tensor([-0.0506, -0.0172, 0.0141, 0.0212, 0.0362, -0.0204, 0.0114, 0.0213, + -0.0309, 0.0149], device='cuda:0') +100 +0.0001 +changing lr +epoch 354, time 223.56, cls_loss 0.5252 cls_loss_mapping 0.0023 cls_loss_causal 0.4646 re_mapping 0.0088 re_causal 0.0213 /// teacc 98.82 lr 0.00010000 +Epoch 356, weight, value: tensor([[-0.1961, -0.0171, 0.0301, ..., -0.0255, -0.0833, -0.1385], + [-0.0566, -0.1032, 0.0373, ..., 0.0486, -0.0163, -0.0967], + [-0.0627, -0.0885, -0.1092, ..., 0.0390, -0.0400, -0.1109], + ..., + [-0.0671, 0.0335, 0.0285, ..., 0.0194, -0.0575, -0.0985], + [-0.0827, -0.0041, 0.0221, ..., 0.0630, -0.0337, -0.1824], + [ 0.0658, 0.0646, -0.0552, ..., -0.1067, -0.0163, 0.0907]], + device='cuda:0'), grad: tensor([[ 1.3504e-03, -1.2123e-02, 5.9605e-08, ..., -4.4870e-04, + 4.4036e-04, 1.7643e-03], + [ 2.1636e-05, 6.9313e-03, 1.3165e-05, ..., -2.6345e-05, + 4.4815e-06, 2.4498e-05], + [ 2.2614e-04, 6.2752e-04, 2.0862e-05, ..., 9.5725e-05, + 2.9354e-03, 9.9468e-04], + ..., + [-1.9586e-04, 4.2796e-05, -1.7416e-04, ..., -1.4114e-04, + 1.2082e-04, 1.5962e-04], + [ 1.0556e-04, 6.7520e-03, 2.2620e-05, ..., 3.7432e-05, + 1.0696e-02, 8.7833e-04], + [-7.6599e-03, 3.9053e-04, -1.0574e-02, ..., -3.6354e-03, + 8.7118e-04, -1.6603e-03]], device='cuda:0') +Epoch 356, bias, value: tensor([-0.0513, 0.0220, 0.0083, -0.0133, -0.0059, 0.0061, -0.0137, 0.0341, + -0.0195, 0.0136], device='cuda:0'), grad: tensor([-0.0558, -0.0453, 0.0254, -0.0081, 0.0319, 0.0198, 0.0006, 0.0113, + 0.0334, -0.0131], device='cuda:0') +100 +0.0001 +changing lr +epoch 355, time 226.17, cls_loss 0.5131 cls_loss_mapping 0.0037 cls_loss_causal 0.4532 re_mapping 0.0089 re_causal 0.0203 /// teacc 98.99 lr 0.00010000 +Epoch 357, weight, value: tensor([[-0.1968, -0.0177, 0.0294, ..., -0.0260, -0.0847, -0.1381], + [-0.0556, -0.1036, 0.0367, ..., 0.0500, -0.0163, -0.0973], + [-0.0626, -0.0879, -0.1083, ..., 0.0397, -0.0397, -0.1116], + ..., + [-0.0667, 0.0336, 0.0283, ..., 0.0191, -0.0578, -0.0980], + [-0.0825, -0.0043, 0.0234, ..., 0.0615, -0.0339, -0.1815], + [ 0.0653, 0.0650, -0.0565, ..., -0.1074, -0.0162, 0.0904]], + device='cuda:0'), grad: tensor([[ 4.4227e-05, 2.8086e-04, 5.5265e-04, ..., 1.2279e-04, + 3.2902e-04, 2.7490e-04], + [ 1.1927e-04, 1.4663e-04, 1.0557e-03, ..., 3.1710e-05, + 2.7013e-04, 1.9193e-04], + [-1.8799e-04, 9.2173e-04, -1.4191e-03, ..., -5.5695e-04, + 2.8992e-04, -2.2292e-04], + ..., + [ 1.9908e-04, 5.9748e-04, 9.6130e-04, ..., 3.7241e-04, + 3.4142e-04, 8.2064e-04], + [ 3.0689e-03, 2.3785e-03, 5.3358e-04, ..., 1.6108e-03, + 7.4291e-04, 1.8778e-03], + [-1.6475e-04, -2.8954e-03, -2.1877e-03, ..., 3.7766e-04, + -2.8515e-03, -3.1815e-03]], device='cuda:0') +Epoch 357, bias, value: tensor([-0.0511, 0.0225, 0.0086, -0.0130, -0.0053, 0.0056, -0.0151, 0.0335, + -0.0192, 0.0137], device='cuda:0'), grad: tensor([ 0.0172, 0.0209, -0.0287, -0.0060, -0.0143, 0.0264, -0.0339, 0.0217, + 0.0363, -0.0395], device='cuda:0') +100 +0.0001 +changing lr +epoch 356, time 225.44, cls_loss 0.5103 cls_loss_mapping 0.0026 cls_loss_causal 0.4456 re_mapping 0.0091 re_causal 0.0217 /// teacc 98.87 lr 0.00010000 +Epoch 358, weight, value: tensor([[-0.1977, -0.0175, 0.0290, ..., -0.0269, -0.0842, -0.1400], + [-0.0549, -0.1048, 0.0372, ..., 0.0500, -0.0169, -0.0972], + [-0.0634, -0.0860, -0.1082, ..., 0.0400, -0.0394, -0.1126], + ..., + [-0.0680, 0.0338, 0.0275, ..., 0.0201, -0.0583, -0.0992], + [-0.0799, -0.0041, 0.0240, ..., 0.0623, -0.0343, -0.1823], + [ 0.0653, 0.0646, -0.0555, ..., -0.1088, -0.0139, 0.0908]], + device='cuda:0'), grad: tensor([[ 9.6738e-05, 9.6381e-05, 3.3913e-03, ..., 7.7486e-04, + 6.0177e-04, 8.5592e-04], + [ 1.4234e-04, 2.0936e-05, -1.9550e-03, ..., -4.3941e-04, + 7.3290e-04, 5.5599e-04], + [ 5.1403e-04, 4.2051e-05, 4.6959e-03, ..., 3.7556e-03, + 1.1988e-03, 1.6489e-03], + ..., + [-3.0384e-03, 2.5139e-03, -4.1924e-03, ..., -9.4271e-04, + 1.0490e-03, 3.4389e-03], + [ 6.3753e-04, 1.2245e-02, 3.8147e-03, ..., 1.1368e-02, + 5.2357e-04, 5.5504e-04], + [ 8.5068e-04, -4.0970e-03, 2.2354e-03, ..., 9.4366e-04, + -4.8790e-03, -4.1008e-03]], device='cuda:0') +Epoch 358, bias, value: tensor([-0.0505, 0.0220, 0.0090, -0.0133, -0.0057, 0.0052, -0.0139, 0.0333, + -0.0196, 0.0139], device='cuda:0'), grad: tensor([ 0.0296, -0.0429, 0.0478, -0.0381, 0.0194, -0.0284, -0.0091, 0.0177, + 0.0699, -0.0658], device='cuda:0') +100 +0.0001 +changing lr +epoch 357, time 224.21, cls_loss 0.5219 cls_loss_mapping 0.0030 cls_loss_causal 0.4586 re_mapping 0.0090 re_causal 0.0209 /// teacc 98.84 lr 0.00010000 +Epoch 359, weight, value: tensor([[-0.1981, -0.0174, 0.0305, ..., -0.0266, -0.0845, -0.1398], + [-0.0550, -0.1052, 0.0381, ..., 0.0503, -0.0171, -0.0961], + [-0.0636, -0.0865, -0.1098, ..., 0.0382, -0.0396, -0.1148], + ..., + [-0.0665, 0.0338, 0.0270, ..., 0.0210, -0.0587, -0.0988], + [-0.0802, -0.0046, 0.0251, ..., 0.0649, -0.0332, -0.1809], + [ 0.0642, 0.0648, -0.0557, ..., -0.1097, -0.0140, 0.0901]], + device='cuda:0'), grad: tensor([[ 1.7717e-05, 3.8669e-06, 2.0099e-04, ..., 1.0270e-04, + 1.2779e-03, 1.1414e-05], + [ 3.5316e-05, 2.0303e-07, 1.8001e-04, ..., 1.9753e-04, + 1.0576e-03, 2.8327e-05], + [-1.0514e-04, 3.7216e-06, 5.7966e-05, ..., 6.7174e-05, + 1.0977e-03, 1.5348e-05], + ..., + [ 1.4138e-04, 1.6510e-04, 4.4513e-04, ..., 1.5283e-04, + 2.0695e-03, 1.9038e-04], + [ 5.3287e-05, 2.1666e-05, 3.2067e-04, ..., -3.5596e-04, + -2.4323e-02, 7.7188e-05], + [ 4.7760e-03, -4.0030e-04, 8.6136e-03, ..., 4.8409e-03, + -2.0027e-03, 4.4518e-03]], device='cuda:0') +Epoch 359, bias, value: tensor([-0.0501, 0.0224, 0.0091, -0.0132, -0.0058, 0.0053, -0.0144, 0.0333, + -0.0191, 0.0128], device='cuda:0'), grad: tensor([ 0.0190, 0.0165, 0.0127, -0.0063, -0.0150, -0.0437, -0.0033, 0.0159, + -0.0052, 0.0094], device='cuda:0') +100 +0.0001 +changing lr +epoch 358, time 224.64, cls_loss 0.5438 cls_loss_mapping 0.0032 cls_loss_causal 0.4840 re_mapping 0.0087 re_causal 0.0213 /// teacc 98.79 lr 0.00010000 +Epoch 360, weight, value: tensor([[-0.1984, -0.0185, 0.0300, ..., -0.0274, -0.0844, -0.1406], + [-0.0560, -0.1055, 0.0376, ..., 0.0514, -0.0177, -0.0945], + [-0.0631, -0.0870, -0.1105, ..., 0.0398, -0.0393, -0.1144], + ..., + [-0.0665, 0.0336, 0.0279, ..., 0.0194, -0.0603, -0.0989], + [-0.0806, -0.0048, 0.0246, ..., 0.0650, -0.0328, -0.1813], + [ 0.0649, 0.0650, -0.0552, ..., -0.1100, -0.0128, 0.0907]], + device='cuda:0'), grad: tensor([[ 1.3225e-07, 2.0158e-04, 2.0714e-03, ..., 8.1301e-04, + 1.0624e-03, 1.7090e-07], + [ 4.5560e-06, 2.5565e-07, 4.1428e-03, ..., 6.8426e-04, + 1.8845e-03, 3.9227e-06], + [ 8.2552e-06, 6.4671e-06, 2.8782e-03, ..., 1.4048e-03, + 1.6842e-03, 1.4417e-05], + ..., + [ 1.1072e-05, 4.5542e-07, -3.0880e-03, ..., -2.3422e-03, + 8.0824e-04, -4.0555e-04], + [ 7.8455e-06, 1.3091e-05, 3.1185e-03, ..., 1.1206e-03, + 8.8739e-04, 7.3537e-06], + [ 4.9658e-06, 8.6501e-06, 2.2335e-03, ..., 7.8154e-04, + 1.5965e-03, 3.7956e-04]], device='cuda:0') +Epoch 360, bias, value: tensor([-0.0510, 0.0218, 0.0092, -0.0134, -0.0055, 0.0059, -0.0145, 0.0335, + -0.0191, 0.0134], device='cuda:0'), grad: tensor([ 0.0237, -0.0048, 0.0273, -0.0686, 0.0083, 0.0226, -0.0345, -0.0255, + 0.0294, 0.0220], device='cuda:0') +100 +0.0001 +changing lr +epoch 359, time 227.75, cls_loss 0.5203 cls_loss_mapping 0.0027 cls_loss_causal 0.4604 re_mapping 0.0091 re_causal 0.0212 /// teacc 98.90 lr 0.00010000 +Epoch 361, weight, value: tensor([[-0.1991, -0.0170, 0.0313, ..., -0.0266, -0.0859, -0.1413], + [-0.0565, -0.1056, 0.0376, ..., 0.0522, -0.0174, -0.0947], + [-0.0630, -0.0882, -0.1117, ..., 0.0396, -0.0395, -0.1148], + ..., + [-0.0660, 0.0343, 0.0272, ..., 0.0188, -0.0607, -0.0987], + [-0.0805, -0.0046, 0.0229, ..., 0.0645, -0.0328, -0.1822], + [ 0.0646, 0.0638, -0.0543, ..., -0.1093, -0.0119, 0.0908]], + device='cuda:0'), grad: tensor([[ 1.8740e-04, -1.8161e-07, 9.1457e-04, ..., 1.5821e-03, + 7.2718e-04, 3.6025e-04], + [ 7.8827e-06, 3.6554e-07, 4.7827e-04, ..., 9.4032e-04, + 4.0102e-04, 2.8044e-05], + [ 1.9503e-04, 5.2713e-07, 7.9918e-04, ..., 1.3752e-03, + 5.8270e-04, 4.1771e-04], + ..., + [ 8.6606e-05, -1.4920e-06, -9.6321e-05, ..., 5.3978e-04, + 1.7142e-04, -2.7823e-04], + [ 1.2493e-04, 1.2107e-08, 2.6274e-04, ..., 6.3944e-04, + 1.5950e-04, 2.3437e-04], + [-3.0446e-04, 5.0617e-07, 9.7513e-05, ..., 4.7064e-04, + 2.0218e-04, -3.4714e-04]], device='cuda:0') +Epoch 361, bias, value: tensor([-0.0516, 0.0227, 0.0085, -0.0137, -0.0056, 0.0069, -0.0153, 0.0332, + -0.0188, 0.0139], device='cuda:0'), grad: tensor([ 0.0103, 0.0056, 0.0095, -0.0066, -0.0226, -0.0120, 0.0052, 0.0016, + 0.0054, 0.0035], device='cuda:0') +100 +0.0001 +changing lr +epoch 360, time 228.11, cls_loss 0.5323 cls_loss_mapping 0.0028 cls_loss_causal 0.4745 re_mapping 0.0088 re_causal 0.0206 /// teacc 98.75 lr 0.00010000 +Epoch 362, weight, value: tensor([[-0.1994, -0.0155, 0.0313, ..., -0.0270, -0.0846, -0.1418], + [-0.0561, -0.1054, 0.0371, ..., 0.0534, -0.0188, -0.0942], + [-0.0630, -0.0890, -0.1114, ..., 0.0397, -0.0387, -0.1147], + ..., + [-0.0659, 0.0343, 0.0272, ..., 0.0188, -0.0605, -0.0987], + [-0.0801, -0.0040, 0.0230, ..., 0.0656, -0.0330, -0.1815], + [ 0.0644, 0.0635, -0.0545, ..., -0.1100, -0.0119, 0.0911]], + device='cuda:0'), grad: tensor([[ 4.6587e-04, 2.3603e-04, 3.1204e-03, ..., 2.2354e-03, + 0.0000e+00, 2.2240e-06], + [ 2.9057e-06, 2.5368e-04, 4.0932e-03, ..., -3.2024e-03, + 0.0000e+00, -2.4110e-05], + [ 2.7791e-05, 1.3006e-04, 1.1759e-03, ..., 6.3801e-04, + 0.0000e+00, 8.0606e-07], + ..., + [ 4.0084e-05, 1.2660e-04, 1.9026e-03, ..., 1.3571e-03, + 0.0000e+00, 4.2506e-06], + [-3.5000e-04, -2.4757e-03, 1.9703e-03, ..., 9.2983e-04, + 0.0000e+00, 1.9018e-06], + [-6.6280e-04, 7.5865e-04, 7.3195e-04, ..., -1.2982e-04, + 0.0000e+00, 5.0366e-06]], device='cuda:0') +Epoch 362, bias, value: tensor([-0.0526, 0.0220, 0.0089, -0.0129, -0.0054, 0.0074, -0.0161, 0.0335, + -0.0184, 0.0139], device='cuda:0'), grad: tensor([ 0.0358, -0.0022, 0.0086, -0.0051, 0.0322, -0.0199, -0.1020, 0.0208, + 0.0181, 0.0136], device='cuda:0') +100 +0.0001 +changing lr +epoch 361, time 227.14, cls_loss 0.5142 cls_loss_mapping 0.0024 cls_loss_causal 0.4564 re_mapping 0.0086 re_causal 0.0201 /// teacc 98.80 lr 0.00010000 +Epoch 363, weight, value: tensor([[-0.1995, -0.0161, 0.0318, ..., -0.0271, -0.0849, -0.1422], + [-0.0568, -0.1057, 0.0363, ..., 0.0492, -0.0193, -0.0944], + [-0.0627, -0.0909, -0.1111, ..., 0.0417, -0.0384, -0.1146], + ..., + [-0.0665, 0.0343, 0.0257, ..., 0.0199, -0.0607, -0.0982], + [-0.0800, -0.0041, 0.0235, ..., 0.0660, -0.0326, -0.1822], + [ 0.0651, 0.0639, -0.0536, ..., -0.1098, -0.0124, 0.0916]], + device='cuda:0'), grad: tensor([[ 4.9734e-04, 8.3745e-05, 2.7790e-03, ..., 1.5087e-03, + 1.8382e-04, 8.8835e-04], + [ 8.7798e-05, -8.2302e-04, 6.2656e-04, ..., 3.6430e-04, + 1.3399e-04, -4.4656e-04], + [ 2.1684e-04, 1.2505e-04, 2.9869e-03, ..., -2.2602e-03, + 2.0874e-04, 1.4389e-04], + ..., + [ 3.5095e-04, -5.3830e-06, 1.0118e-03, ..., 2.0180e-03, + 4.8339e-05, -2.0415e-05], + [ 1.6146e-03, 9.0885e-04, 3.5534e-03, ..., 5.4817e-03, + 1.3661e-04, 8.2254e-04], + [-4.3297e-03, -6.0558e-04, -5.8823e-03, ..., -4.1542e-03, + 6.0081e-05, -8.2111e-04]], device='cuda:0') +Epoch 363, bias, value: tensor([-0.0533, 0.0226, 0.0091, -0.0143, -0.0061, 0.0087, -0.0155, 0.0326, + -0.0186, 0.0149], device='cuda:0'), grad: tensor([ 2.7222e-02, 7.4120e-03, 3.6736e-03, -6.6956e-02, 1.5854e-02, + 2.7756e-02, -6.3419e-05, 1.4824e-02, 3.3173e-02, -6.2927e-02], + device='cuda:0') +100 +0.0001 +changing lr +epoch 362, time 225.94, cls_loss 0.5230 cls_loss_mapping 0.0033 cls_loss_causal 0.4688 re_mapping 0.0084 re_causal 0.0200 /// teacc 98.94 lr 0.00010000 +Epoch 364, weight, value: tensor([[-0.2009, -0.0164, 0.0320, ..., -0.0262, -0.0840, -0.1438], + [-0.0576, -0.1066, 0.0358, ..., 0.0496, -0.0197, -0.0948], + [-0.0634, -0.0907, -0.1123, ..., 0.0409, -0.0390, -0.1158], + ..., + [-0.0673, 0.0350, 0.0254, ..., 0.0189, -0.0610, -0.0982], + [-0.0800, -0.0032, 0.0237, ..., 0.0671, -0.0327, -0.1818], + [ 0.0662, 0.0634, -0.0532, ..., -0.1110, -0.0129, 0.0915]], + device='cuda:0'), grad: tensor([[-1.5885e-05, 3.0518e-05, 1.1063e-03, ..., 2.0962e-03, + -3.1781e-04, 3.7885e-04], + [ 4.4167e-05, -5.0068e-05, -1.2922e-03, ..., 1.3018e-03, + -2.7065e-03, 5.4598e-05], + [ 2.9415e-05, 3.7283e-05, -2.8458e-03, ..., -3.8395e-03, + -3.2349e-03, 3.1066e-04], + ..., + [ 4.7922e-04, 2.1172e-04, 2.1839e-03, ..., -2.5024e-03, + -1.9474e-03, -1.2770e-03], + [ 3.6001e-04, 5.6343e-03, 1.0281e-03, ..., 1.2207e-03, + 1.2331e-03, 2.5463e-04], + [-1.1892e-03, -7.3700e-03, -5.9090e-03, ..., -4.4212e-03, + -2.1706e-03, 6.6280e-04]], device='cuda:0') +Epoch 364, bias, value: tensor([-0.0521, 0.0218, 0.0085, -0.0145, -0.0052, 0.0081, -0.0162, 0.0329, + -0.0178, 0.0145], device='cuda:0'), grad: tensor([ 0.0177, -0.0023, -0.0380, -0.0006, 0.0298, 0.0025, 0.0397, -0.0333, + -0.0251, 0.0096], device='cuda:0') +100 +0.0001 +changing lr +epoch 363, time 225.05, cls_loss 0.5012 cls_loss_mapping 0.0024 cls_loss_causal 0.4391 re_mapping 0.0093 re_causal 0.0214 /// teacc 98.84 lr 0.00010000 +Epoch 365, weight, value: tensor([[-0.2019, -0.0173, 0.0321, ..., -0.0272, -0.0830, -0.1443], + [-0.0583, -0.1074, 0.0360, ..., 0.0506, -0.0208, -0.0950], + [-0.0616, -0.0921, -0.1126, ..., 0.0402, -0.0387, -0.1154], + ..., + [-0.0674, 0.0353, 0.0246, ..., 0.0190, -0.0618, -0.0966], + [-0.0805, -0.0034, 0.0233, ..., 0.0678, -0.0324, -0.1823], + [ 0.0654, 0.0646, -0.0532, ..., -0.1106, -0.0132, 0.0902]], + device='cuda:0'), grad: tensor([[ 5.8556e-04, 1.8489e-04, 1.4124e-03, ..., 1.7996e-03, + 9.4604e-04, 2.1875e-05], + [ 3.6216e-04, 4.8685e-04, 1.1377e-03, ..., 2.4071e-03, + -1.1536e-02, 1.1260e-06], + [-1.2791e-04, 4.1699e-04, -3.5038e-03, ..., -9.1982e-04, + -1.7166e-03, 3.3770e-06], + ..., + [-2.4414e-03, -2.3861e-03, -4.7188e-03, ..., -5.7411e-03, + 1.6749e-04, 1.7472e-06], + [ 6.9284e-04, 9.4175e-04, 1.3037e-03, ..., 3.4580e-03, + 7.5493e-03, -6.4075e-05], + [ 2.5487e-04, -1.0242e-03, 7.3004e-04, ..., 7.5607e-03, + 7.3357e-03, 6.4254e-05]], device='cuda:0') +Epoch 365, bias, value: tensor([-0.0522, 0.0217, 0.0081, -0.0139, -0.0050, 0.0083, -0.0158, 0.0320, + -0.0184, 0.0152], device='cuda:0'), grad: tensor([ 0.0164, -0.0209, -0.0126, -0.0157, -0.0002, 0.0376, -0.0442, -0.0198, + 0.0388, 0.0205], device='cuda:0') +100 +0.0001 +changing lr +epoch 364, time 227.22, cls_loss 0.5453 cls_loss_mapping 0.0043 cls_loss_causal 0.4812 re_mapping 0.0088 re_causal 0.0199 /// teacc 98.80 lr 0.00010000 +Epoch 366, weight, value: tensor([[-0.2015, -0.0174, 0.0306, ..., -0.0266, -0.0828, -0.1446], + [-0.0591, -0.1073, 0.0342, ..., 0.0522, -0.0208, -0.0953], + [-0.0616, -0.0933, -0.1126, ..., 0.0380, -0.0369, -0.1157], + ..., + [-0.0668, 0.0356, 0.0263, ..., 0.0191, -0.0619, -0.0968], + [-0.0816, -0.0035, 0.0227, ..., 0.0671, -0.0332, -0.1821], + [ 0.0647, 0.0646, -0.0519, ..., -0.1112, -0.0146, 0.0899]], + device='cuda:0'), grad: tensor([[ 4.8697e-05, 7.8008e-06, 5.0879e-04, ..., 1.9484e-03, + 5.4026e-04, 2.8089e-06], + [ 1.0329e-04, 2.4036e-05, 1.1454e-03, ..., 4.0169e-03, + 1.0757e-03, 7.2718e-06], + [ 7.0477e-04, 2.2507e-04, 4.8370e-03, ..., 7.2136e-03, + 3.8052e-03, 3.1203e-05], + ..., + [ 4.3273e-04, 1.6384e-03, -1.2827e-03, ..., -2.0847e-03, + -2.3613e-03, 1.1809e-05], + [ 2.8586e-04, 5.2118e-04, 1.9350e-03, ..., 3.8123e-04, + 1.5812e-03, 5.9679e-06], + [-1.8167e-03, -5.8861e-03, -1.0803e-02, ..., -1.3199e-02, + -7.1030e-03, -8.6948e-06]], device='cuda:0') +Epoch 366, bias, value: tensor([-0.0515, 0.0215, 0.0071, -0.0129, -0.0048, 0.0078, -0.0154, 0.0338, + -0.0196, 0.0142], device='cuda:0'), grad: tensor([ 0.0164, 0.0277, 0.0349, -0.0070, 0.0165, -0.0134, -0.0088, -0.0070, + -0.0077, -0.0516], device='cuda:0') +100 +0.0001 +changing lr +epoch 365, time 227.68, cls_loss 0.5029 cls_loss_mapping 0.0025 cls_loss_causal 0.4491 re_mapping 0.0091 re_causal 0.0218 /// teacc 98.80 lr 0.00010000 +Epoch 367, weight, value: tensor([[-0.2021, -0.0188, 0.0307, ..., -0.0258, -0.0823, -0.1452], + [-0.0595, -0.1068, 0.0343, ..., 0.0535, -0.0204, -0.0946], + [-0.0627, -0.0953, -0.1131, ..., 0.0369, -0.0380, -0.1166], + ..., + [-0.0676, 0.0360, 0.0264, ..., 0.0199, -0.0622, -0.0980], + [-0.0817, -0.0048, 0.0219, ..., 0.0655, -0.0313, -0.1830], + [ 0.0650, 0.0653, -0.0520, ..., -0.1105, -0.0153, 0.0909]], + device='cuda:0'), grad: tensor([[ 5.4693e-04, 3.7980e-04, 8.2195e-05, ..., -1.0590e-02, + -1.2848e-02, 1.4246e-04], + [ 2.1791e-04, 8.5413e-05, 6.3181e-05, ..., 2.6531e-03, + 2.0618e-03, 5.2363e-05], + [-5.3120e-04, -2.7061e-04, -4.3988e-04, ..., 1.2407e-03, + 1.0662e-03, -6.5756e-04], + ..., + [ 2.6059e-04, 1.6582e-04, 7.9632e-05, ..., 1.0328e-03, + 6.0034e-04, 1.2338e-04], + [ 1.5755e-03, 4.0030e-04, 2.3031e-04, ..., 3.1624e-03, + 2.2144e-03, 1.5426e-04], + [ 1.8520e-03, -8.4591e-04, -4.2176e-04, ..., 1.2131e-03, + 1.4038e-03, -7.8917e-04]], device='cuda:0') +Epoch 367, bias, value: tensor([-0.0514, 0.0219, 0.0063, -0.0123, -0.0046, 0.0071, -0.0144, 0.0343, + -0.0201, 0.0136], device='cuda:0'), grad: tensor([-0.0193, 0.0127, -0.0056, -0.0138, -0.0235, -0.0041, 0.0227, 0.0075, + 0.0171, 0.0062], device='cuda:0') +100 +0.0001 +changing lr +epoch 366, time 226.23, cls_loss 0.5290 cls_loss_mapping 0.0031 cls_loss_causal 0.4645 re_mapping 0.0093 re_causal 0.0223 /// teacc 98.91 lr 0.00010000 +Epoch 368, weight, value: tensor([[-0.2029, -0.0191, 0.0310, ..., -0.0270, -0.0826, -0.1460], + [-0.0588, -0.1067, 0.0348, ..., 0.0531, -0.0202, -0.0951], + [-0.0638, -0.0959, -0.1133, ..., 0.0359, -0.0381, -0.1171], + ..., + [-0.0657, 0.0366, 0.0266, ..., 0.0192, -0.0615, -0.0987], + [-0.0801, -0.0041, 0.0221, ..., 0.0680, -0.0306, -0.1820], + [ 0.0647, 0.0646, -0.0514, ..., -0.1102, -0.0146, 0.0925]], + device='cuda:0'), grad: tensor([[ 3.2987e-06, 6.0654e-04, -1.2589e-03, ..., -1.7900e-03, + 6.1929e-05, -6.2275e-04], + [ 4.6670e-05, 6.6710e-04, 8.0109e-05, ..., 9.6083e-04, + 1.1045e-04, 4.6730e-05], + [ 6.1207e-06, 1.1215e-03, 4.2391e-04, ..., 3.7365e-03, + 5.3978e-04, 3.7122e-04], + ..., + [ 7.5197e-04, 3.0460e-03, 9.7752e-04, ..., 2.2812e-03, + 4.9353e-04, 7.7200e-04], + [ 4.6283e-05, 4.9706e-03, 1.1337e-04, ..., 1.8435e-03, + 7.3814e-04, 9.1374e-05], + [-2.2850e-03, 9.1629e-03, -2.0599e-03, ..., 1.2960e-03, + 1.3065e-03, -1.0233e-03]], device='cuda:0') +Epoch 368, bias, value: tensor([-0.0529, 0.0210, 0.0056, -0.0110, -0.0046, 0.0075, -0.0134, 0.0332, + -0.0195, 0.0142], device='cuda:0'), grad: tensor([-0.0041, 0.0104, 0.0297, -0.0173, -0.0347, -0.0182, -0.0487, 0.0291, + 0.0256, 0.0281], device='cuda:0') +100 +0.0001 +changing lr +epoch 367, time 228.77, cls_loss 0.5181 cls_loss_mapping 0.0034 cls_loss_causal 0.4526 re_mapping 0.0089 re_causal 0.0201 /// teacc 98.75 lr 0.00010000 +Epoch 369, weight, value: tensor([[-0.2036, -0.0203, 0.0301, ..., -0.0275, -0.0839, -0.1455], + [-0.0581, -0.1081, 0.0361, ..., 0.0541, -0.0200, -0.0948], + [-0.0640, -0.0936, -0.1144, ..., 0.0368, -0.0380, -0.1170], + ..., + [-0.0652, 0.0339, 0.0274, ..., 0.0164, -0.0609, -0.0995], + [-0.0798, -0.0044, 0.0236, ..., 0.0686, -0.0319, -0.1817], + [ 0.0635, 0.0662, -0.0505, ..., -0.1099, -0.0135, 0.0928]], + device='cuda:0'), grad: tensor([[ 1.1432e-04, 2.0576e-04, 1.6603e-03, ..., 2.2793e-03, + 2.1706e-03, 7.8082e-05], + [ 2.2843e-05, 1.3113e-04, 5.2023e-04, ..., -2.3689e-03, + 9.4032e-04, 1.8090e-05], + [ 1.8644e-04, 1.2136e-04, 3.5515e-03, ..., 3.0403e-03, + 1.4553e-03, 1.6618e-04], + ..., + [ 3.1519e-04, 9.0718e-05, 3.0556e-03, ..., 1.9197e-03, + 8.1539e-04, 2.4676e-04], + [ 7.0763e-04, 1.7822e-04, 2.1858e-03, ..., 1.7910e-03, + 4.9877e-04, 4.9448e-04], + [ 2.0373e-04, -9.8944e-06, 1.9217e-03, ..., -2.4605e-03, + 5.6505e-04, 1.7726e-04]], device='cuda:0') +Epoch 369, bias, value: tensor([-0.0533, 0.0210, 0.0054, -0.0116, -0.0044, 0.0079, -0.0128, 0.0322, + -0.0191, 0.0148], device='cuda:0'), grad: tensor([ 0.0334, -0.0043, 0.0387, -0.0378, -0.0363, -0.0149, -0.0277, 0.0287, + 0.0250, -0.0050], device='cuda:0') +100 +0.0001 +changing lr +epoch 368, time 226.00, cls_loss 0.5376 cls_loss_mapping 0.0033 cls_loss_causal 0.4739 re_mapping 0.0092 re_causal 0.0217 /// teacc 98.84 lr 0.00010000 +Epoch 370, weight, value: tensor([[-0.2051, -0.0207, 0.0293, ..., -0.0300, -0.0844, -0.1460], + [-0.0587, -0.1082, 0.0362, ..., 0.0535, -0.0205, -0.0938], + [-0.0654, -0.0951, -0.1158, ..., 0.0364, -0.0394, -0.1181], + ..., + [-0.0656, 0.0343, 0.0285, ..., 0.0167, -0.0609, -0.0993], + [-0.0804, -0.0037, 0.0221, ..., 0.0669, -0.0323, -0.1837], + [ 0.0629, 0.0646, -0.0517, ..., -0.1099, -0.0118, 0.0931]], + device='cuda:0'), grad: tensor([[ 4.4078e-05, 2.7633e-04, 1.3256e-03, ..., 1.8244e-03, + 4.3464e-04, 5.2810e-05], + [ 4.8971e-04, 4.8131e-06, -4.0627e-03, ..., 4.4250e-04, + -3.9864e-03, 2.0897e-04], + [ 1.1659e-04, -5.5933e-04, 1.7738e-03, ..., 1.8320e-03, + 5.0688e-04, 9.8765e-05], + ..., + [ 4.3774e-04, 1.4722e-05, 2.5711e-03, ..., -1.4982e-03, + 4.7064e-04, 2.4557e-04], + [ 3.0851e-04, 2.8580e-05, 2.5101e-03, ..., -3.4714e-03, + 7.0381e-04, 1.8144e-04], + [-1.4763e-03, 5.3048e-05, -5.7106e-03, ..., -2.2774e-03, + -6.7139e-04, -2.1672e-04]], device='cuda:0') +Epoch 370, bias, value: tensor([-0.0544, 0.0210, 0.0041, -0.0104, -0.0037, 0.0080, -0.0123, 0.0326, + -0.0196, 0.0146], device='cuda:0'), grad: tensor([-0.0076, -0.0122, -0.0118, -0.0116, 0.0274, 0.0360, -0.0029, -0.0293, + -0.0005, 0.0125], device='cuda:0') +100 +0.0001 +changing lr +epoch 369, time 227.12, cls_loss 0.5292 cls_loss_mapping 0.0027 cls_loss_causal 0.4678 re_mapping 0.0093 re_causal 0.0214 /// teacc 98.88 lr 0.00010000 +Epoch 371, weight, value: tensor([[-0.2032, -0.0210, 0.0298, ..., -0.0288, -0.0843, -0.1458], + [-0.0589, -0.1082, 0.0372, ..., 0.0538, -0.0190, -0.0929], + [-0.0658, -0.0938, -0.1161, ..., 0.0373, -0.0394, -0.1184], + ..., + [-0.0650, 0.0341, 0.0289, ..., 0.0184, -0.0597, -0.0989], + [-0.0793, -0.0037, 0.0205, ..., 0.0650, -0.0323, -0.1843], + [ 0.0617, 0.0659, -0.0533, ..., -0.1110, -0.0126, 0.0927]], + device='cuda:0'), grad: tensor([[ 1.8396e-03, -7.8154e-04, 5.7831e-03, ..., 3.8185e-03, + 2.1774e-02, 4.4286e-05], + [ 1.0324e-04, 1.5793e-03, 3.0518e-03, ..., 3.1700e-03, + 1.2374e-04, 5.0426e-05], + [-3.4928e-04, -6.8951e-04, -1.5974e-03, ..., -4.0817e-03, + 1.8084e-04, -3.8013e-03], + ..., + [ 2.1839e-03, 7.0906e-04, -1.3496e-02, ..., 2.1515e-03, + 1.0271e-03, 3.5095e-03], + [ 3.7575e-04, 1.4222e-04, 8.1205e-04, ..., 2.5082e-03, + -2.6794e-02, 1.9228e-04], + [-5.3525e-05, 1.8299e-04, 2.0332e-03, ..., -1.2150e-03, + 3.0289e-03, 4.1628e-04]], device='cuda:0') +Epoch 371, bias, value: tensor([-0.0537, 0.0217, 0.0036, -0.0108, -0.0035, 0.0082, -0.0127, 0.0339, + -0.0201, 0.0137], device='cuda:0'), grad: tensor([ 0.0244, 0.0188, -0.0195, 0.0063, -0.0233, 0.0225, -0.0604, 0.0349, + -0.0048, 0.0009], device='cuda:0') +100 +0.0001 +changing lr +epoch 370, time 226.90, cls_loss 0.5453 cls_loss_mapping 0.0023 cls_loss_causal 0.4841 re_mapping 0.0094 re_causal 0.0223 /// teacc 98.92 lr 0.00010000 +Epoch 372, weight, value: tensor([[-0.2040, -0.0208, 0.0301, ..., -0.0282, -0.0847, -0.1464], + [-0.0591, -0.1084, 0.0378, ..., 0.0549, -0.0183, -0.0927], + [-0.0664, -0.0938, -0.1164, ..., 0.0382, -0.0391, -0.1170], + ..., + [-0.0638, 0.0331, 0.0286, ..., 0.0179, -0.0586, -0.0994], + [-0.0786, -0.0038, 0.0209, ..., 0.0650, -0.0312, -0.1851], + [ 0.0611, 0.0667, -0.0529, ..., -0.1115, -0.0129, 0.0936]], + device='cuda:0'), grad: tensor([[ 7.7426e-05, 9.6917e-05, 3.8052e-04, ..., 1.2517e-04, + 3.5334e-04, 4.6492e-06], + [ 2.1040e-05, 1.4365e-04, 4.4942e-04, ..., 1.9515e-04, + 4.5943e-04, 1.9483e-06], + [ 1.2398e-04, 7.9489e-04, 5.0402e-04, ..., 1.6630e-04, + 6.5136e-04, 1.9521e-05], + ..., + [-4.6844e-03, -2.0657e-03, 7.7772e-04, ..., 2.5558e-04, + 5.2309e-04, -1.8940e-03], + [ 2.1982e-04, 3.3665e-04, 6.9332e-04, ..., 2.1839e-04, + 5.3263e-04, 5.8442e-05], + [ 3.8395e-03, 1.0307e-02, -5.5122e-03, ..., -1.8082e-03, + -1.9989e-03, 1.9331e-03]], device='cuda:0') +Epoch 372, bias, value: tensor([-0.0540, 0.0226, 0.0042, -0.0105, -0.0029, 0.0087, -0.0139, 0.0339, + -0.0210, 0.0131], device='cuda:0'), grad: tensor([-0.0205, 0.0116, -0.0150, 0.0168, -0.0108, -0.0099, 0.0126, -0.0145, + 0.0121, 0.0176], device='cuda:0') +100 +0.0001 +changing lr +epoch 371, time 229.13, cls_loss 0.5329 cls_loss_mapping 0.0026 cls_loss_causal 0.4804 re_mapping 0.0089 re_causal 0.0210 /// teacc 98.82 lr 0.00010000 +Epoch 373, weight, value: tensor([[-0.2048, -0.0206, 0.0298, ..., -0.0277, -0.0851, -0.1462], + [-0.0576, -0.1077, 0.0365, ..., 0.0558, -0.0188, -0.0925], + [-0.0666, -0.0945, -0.1149, ..., 0.0381, -0.0385, -0.1148], + ..., + [-0.0634, 0.0336, 0.0273, ..., 0.0175, -0.0575, -0.0994], + [-0.0780, -0.0045, 0.0215, ..., 0.0642, -0.0310, -0.1854], + [ 0.0618, 0.0662, -0.0516, ..., -0.1115, -0.0135, 0.0933]], + device='cuda:0'), grad: tensor([[ 2.7753e-06, 2.3782e-05, 1.5755e-03, ..., 1.2531e-03, + 0.0000e+00, 8.3819e-08], + [ 1.0461e-04, 2.0161e-03, -1.1482e-03, ..., 1.4153e-03, + 0.0000e+00, 3.7439e-07], + [ 3.6583e-03, 3.1710e-04, 6.1569e-03, ..., 5.6419e-03, + 0.0000e+00, 8.9966e-07], + ..., + [-3.7384e-03, 2.0504e-03, -1.0902e-02, ..., -9.3231e-03, + 1.1176e-08, 3.2429e-06], + [ 7.0035e-05, 1.3447e-03, 1.8129e-03, ..., -7.3147e-04, + 0.0000e+00, 3.4161e-06], + [ 6.2180e-04, 2.7161e-02, 9.4147e-03, ..., 1.9817e-03, + 1.3649e-02, -4.1053e-06]], device='cuda:0') +Epoch 373, bias, value: tensor([-0.0552, 0.0237, 0.0047, -0.0102, -0.0025, 0.0076, -0.0133, 0.0341, + -0.0209, 0.0119], device='cuda:0'), grad: tensor([ 0.0131, 0.0193, 0.0621, 0.0131, -0.0415, 0.0092, -0.0118, -0.0376, + -0.0146, -0.0115], device='cuda:0') +100 +0.0001 +changing lr +epoch 372, time 226.62, cls_loss 0.5301 cls_loss_mapping 0.0027 cls_loss_causal 0.4804 re_mapping 0.0085 re_causal 0.0205 /// teacc 98.70 lr 0.00010000 +Epoch 374, weight, value: tensor([[-0.2067, -0.0204, 0.0291, ..., -0.0282, -0.0849, -0.1465], + [-0.0562, -0.1073, 0.0366, ..., 0.0555, -0.0187, -0.0920], + [-0.0662, -0.0948, -0.1149, ..., 0.0384, -0.0388, -0.1154], + ..., + [-0.0637, 0.0333, 0.0277, ..., 0.0175, -0.0596, -0.0989], + [-0.0780, -0.0040, 0.0216, ..., 0.0641, -0.0305, -0.1853], + [ 0.0616, 0.0670, -0.0523, ..., -0.1110, -0.0128, 0.0934]], + device='cuda:0'), grad: tensor([[-1.0647e-05, 9.2328e-05, 1.2302e-03, ..., 2.3460e-03, + 1.2178e-03, -3.2634e-05], + [ 1.1157e-06, 4.5128e-03, -8.3733e-04, ..., 7.8392e-04, + 7.5281e-05, 3.4243e-05], + [ 2.4345e-06, 1.9193e-04, -2.1286e-03, ..., -2.9640e-03, + 1.9932e-04, 2.7791e-05], + ..., + [ 3.7123e-06, -1.3481e-02, -1.9703e-03, ..., -2.6016e-03, + 6.0797e-04, -1.3981e-03], + [-1.3557e-02, 3.3661e-02, 1.2064e-03, ..., -1.0071e-03, + 1.2960e-03, -1.0788e-02], + [ 5.5581e-05, 4.8218e-03, 2.6531e-03, ..., 3.7498e-03, + -1.3885e-02, 8.9550e-04]], device='cuda:0') +Epoch 374, bias, value: tensor([-0.0550, 0.0237, 0.0047, -0.0103, -0.0024, 0.0077, -0.0139, 0.0338, + -0.0206, 0.0122], device='cuda:0'), grad: tensor([ 0.0356, 0.0510, -0.0115, -0.0307, 0.0242, 0.0020, -0.0205, -0.1013, + 0.0114, 0.0397], device='cuda:0') +100 +0.0001 +changing lr +epoch 373, time 226.28, cls_loss 0.5601 cls_loss_mapping 0.0041 cls_loss_causal 0.4943 re_mapping 0.0089 re_causal 0.0209 /// teacc 98.76 lr 0.00010000 +Epoch 375, weight, value: tensor([[-0.2063, -0.0212, 0.0299, ..., -0.0278, -0.0854, -0.1475], + [-0.0569, -0.1082, 0.0353, ..., 0.0556, -0.0206, -0.0924], + [-0.0655, -0.0953, -0.1157, ..., 0.0380, -0.0397, -0.1165], + ..., + [-0.0635, 0.0339, 0.0272, ..., 0.0175, -0.0599, -0.0985], + [-0.0778, -0.0051, 0.0228, ..., 0.0652, -0.0286, -0.1852], + [ 0.0616, 0.0674, -0.0519, ..., -0.1114, -0.0132, 0.0932]], + device='cuda:0'), grad: tensor([[ 5.8413e-04, 6.7241e-07, 2.8133e-04, ..., 1.2846e-03, + 3.9172e-04, 7.2360e-05], + [ 3.1900e-04, 3.8333e-06, -4.2796e-04, ..., 1.1120e-03, + 1.5926e-04, 1.8024e-04], + [-5.9166e-03, 2.0161e-05, 4.9257e-04, ..., -2.0294e-03, + -2.7962e-03, 1.7750e-04], + ..., + [ 1.2798e-03, -3.0011e-05, 1.7843e-03, ..., 1.3723e-03, + 1.3053e-04, 1.3876e-03], + [ 8.6069e-04, 2.2352e-07, 3.3116e-04, ..., -1.7424e-03, + 3.9172e-04, 3.5310e-04], + [ 2.2471e-04, 3.4459e-06, 2.1005e-04, ..., 1.7776e-03, + 4.2033e-04, -8.3065e-04]], device='cuda:0') +Epoch 375, bias, value: tensor([-0.0543, 0.0235, 0.0047, -0.0100, -0.0026, 0.0082, -0.0139, 0.0346, + -0.0208, 0.0106], device='cuda:0'), grad: tensor([ 0.0180, 0.0100, -0.0446, 0.0122, 0.0191, -0.0485, 0.0236, 0.0108, + -0.0143, 0.0137], device='cuda:0') +100 +0.0001 +changing lr +epoch 374, time 228.07, cls_loss 0.5247 cls_loss_mapping 0.0021 cls_loss_causal 0.4772 re_mapping 0.0091 re_causal 0.0214 /// teacc 98.73 lr 0.00010000 +Epoch 376, weight, value: tensor([[-0.2061, -0.0211, 0.0311, ..., -0.0279, -0.0817, -0.1471], + [-0.0566, -0.1080, 0.0352, ..., 0.0555, -0.0199, -0.0925], + [-0.0650, -0.0959, -0.1152, ..., 0.0377, -0.0402, -0.1165], + ..., + [-0.0643, 0.0337, 0.0253, ..., 0.0176, -0.0619, -0.0985], + [-0.0798, -0.0049, 0.0227, ..., 0.0651, -0.0294, -0.1854], + [ 0.0622, 0.0676, -0.0524, ..., -0.1104, -0.0116, 0.0934]], + device='cuda:0'), grad: tensor([[ 3.6001e-05, 1.5700e-04, 2.4834e-03, ..., 3.5477e-03, + 1.6832e-03, 2.5702e-04], + [ 1.5306e-04, 3.5119e-04, 3.0842e-03, ..., 3.7441e-03, + 2.5139e-03, 2.3615e-04], + [ 4.5598e-05, 1.0090e-03, -3.3016e-03, ..., -1.2589e-03, + -3.5515e-03, 2.5520e-03], + ..., + [-2.3308e-03, -2.1534e-03, -6.2828e-03, ..., -1.5116e-03, + -2.3193e-03, 2.8205e-04], + [ 1.6844e-04, 1.4663e-04, 1.8892e-03, ..., 3.0327e-03, + 1.5173e-03, 4.4274e-04], + [ 4.6992e-04, 8.2111e-04, 7.7152e-04, ..., -5.0812e-03, + -5.9967e-03, 1.2970e-03]], device='cuda:0') +Epoch 376, bias, value: tensor([-0.0532, 0.0226, 0.0055, -0.0098, -0.0025, 0.0083, -0.0144, 0.0341, + -0.0213, 0.0109], device='cuda:0'), grad: tensor([ 0.0225, 0.0278, -0.0338, -0.0338, 0.0565, 0.0130, -0.0230, -0.0212, + 0.0194, -0.0274], device='cuda:0') +100 +0.0001 +changing lr +epoch 375, time 226.14, cls_loss 0.5433 cls_loss_mapping 0.0024 cls_loss_causal 0.4766 re_mapping 0.0087 re_causal 0.0200 /// teacc 98.85 lr 0.00010000 +Epoch 377, weight, value: tensor([[-0.2062, -0.0212, 0.0320, ..., -0.0286, -0.0829, -0.1473], + [-0.0576, -0.1078, 0.0347, ..., 0.0551, -0.0208, -0.0934], + [-0.0654, -0.0959, -0.1154, ..., 0.0377, -0.0404, -0.1172], + ..., + [-0.0652, 0.0340, 0.0253, ..., 0.0162, -0.0632, -0.0991], + [-0.0794, -0.0042, 0.0218, ..., 0.0649, -0.0299, -0.1851], + [ 0.0617, 0.0668, -0.0525, ..., -0.1100, -0.0118, 0.0929]], + device='cuda:0'), grad: tensor([[ 1.2387e-06, 3.7408e-04, 4.0970e-03, ..., -2.7084e-03, + 1.5235e-04, 2.6340e-03], + [ 4.0121e-06, 1.4076e-03, 1.5557e-05, ..., -1.2226e-03, + 5.3406e-05, 6.4522e-06], + [ 2.6256e-05, 6.5231e-04, 4.6062e-04, ..., -1.4629e-03, + 1.9401e-05, 2.9635e-04], + ..., + [-1.0085e-04, 4.0579e-04, 6.8569e-04, ..., 1.3247e-03, + 1.7986e-05, 4.3416e-04], + [ 1.7151e-05, 1.8625e-03, 3.4180e-03, ..., 4.6616e-03, + 7.5912e-04, 5.6601e-04], + [ 1.8746e-05, -1.1292e-02, -6.6681e-03, ..., -3.8242e-04, + 6.9290e-06, -5.1842e-03]], device='cuda:0') +Epoch 377, bias, value: tensor([-0.0534, 0.0225, 0.0059, -0.0098, -0.0033, 0.0092, -0.0146, 0.0333, + -0.0212, 0.0114], device='cuda:0'), grad: tensor([-0.0073, -0.0212, -0.0199, 0.0156, 0.0264, -0.0102, 0.0107, 0.0135, + 0.0281, -0.0357], device='cuda:0') +100 +0.0001 +changing lr +epoch 376, time 233.10, cls_loss 0.5292 cls_loss_mapping 0.0031 cls_loss_causal 0.4543 re_mapping 0.0096 re_causal 0.0202 /// teacc 98.94 lr 0.00010000 +Epoch 378, weight, value: tensor([[-0.2057, -0.0216, 0.0318, ..., -0.0297, -0.0827, -0.1476], + [-0.0578, -0.1067, 0.0347, ..., 0.0554, -0.0221, -0.0933], + [-0.0676, -0.0949, -0.1157, ..., 0.0369, -0.0405, -0.1175], + ..., + [-0.0651, 0.0333, 0.0258, ..., 0.0166, -0.0619, -0.0996], + [-0.0802, -0.0058, 0.0222, ..., 0.0642, -0.0302, -0.1851], + [ 0.0623, 0.0670, -0.0532, ..., -0.1111, -0.0120, 0.0937]], + device='cuda:0'), grad: tensor([[ 2.2149e-04, 6.1512e-04, 9.4223e-04, ..., 1.0433e-03, + 3.1686e-04, 3.2377e-04], + [ 4.1798e-06, 6.9237e-04, -4.3068e-03, ..., -8.5449e-03, + -5.1788e-02, 6.0722e-06], + [ 2.0966e-05, 2.0008e-03, 5.0545e-04, ..., 2.2888e-03, + 1.9669e-02, 2.8729e-05], + ..., + [ 2.7239e-05, 1.2062e-02, -7.0839e-03, ..., -1.2482e-02, + 2.1011e-02, 3.8654e-05], + [ 2.7180e-04, 2.7332e-03, 4.8332e-03, ..., 9.8419e-03, + 1.4257e-03, 3.8362e-04], + [ 2.7403e-05, -2.5482e-02, 8.4639e-04, ..., 7.9918e-04, + 2.2554e-04, 3.9428e-05]], device='cuda:0') +Epoch 378, bias, value: tensor([-0.0531, 0.0226, 0.0052, -0.0101, -0.0034, 0.0084, -0.0140, 0.0348, + -0.0214, 0.0112], device='cuda:0'), grad: tensor([ 0.0192, -0.0635, 0.0152, 0.0037, 0.0174, 0.0458, 0.0300, -0.0753, + 0.0163, -0.0087], device='cuda:0') +100 +0.0001 +changing lr +epoch 377, time 228.12, cls_loss 0.5076 cls_loss_mapping 0.0028 cls_loss_causal 0.4523 re_mapping 0.0090 re_causal 0.0212 /// teacc 98.96 lr 0.00010000 +Epoch 379, weight, value: tensor([[-0.2057, -0.0211, 0.0309, ..., -0.0304, -0.0838, -0.1476], + [-0.0583, -0.1061, 0.0348, ..., 0.0549, -0.0201, -0.0926], + [-0.0698, -0.0943, -0.1170, ..., 0.0385, -0.0387, -0.1184], + ..., + [-0.0655, 0.0336, 0.0256, ..., 0.0170, -0.0625, -0.1007], + [-0.0808, -0.0053, 0.0230, ..., 0.0638, -0.0304, -0.1839], + [ 0.0631, 0.0670, -0.0534, ..., -0.1113, -0.0115, 0.0950]], + device='cuda:0'), grad: tensor([[ 4.3333e-05, 4.0442e-05, 3.4070e-04, ..., 2.4967e-03, + 3.7441e-03, 1.0443e-04], + [-4.9734e-04, -5.1928e-04, 4.7088e-04, ..., -1.6052e-02, + 3.9635e-03, -1.0328e-03], + [ 5.0455e-05, 4.5121e-05, 2.7657e-03, ..., 5.6801e-03, + 2.8496e-03, 1.2589e-04], + ..., + [ 1.1482e-03, 1.5717e-03, 2.5120e-03, ..., 6.5422e-03, + 1.7948e-03, 2.1152e-03], + [ 7.2837e-05, 9.4235e-05, 5.2023e-04, ..., 2.8229e-03, + -2.1988e-02, 2.0826e-04], + [-1.3428e-03, -3.1223e-03, -9.4175e-04, ..., -1.8282e-03, + 1.9627e-03, -3.2291e-03]], device='cuda:0') +Epoch 379, bias, value: tensor([-0.0528, 0.0221, 0.0065, -0.0108, -0.0034, 0.0091, -0.0146, 0.0340, + -0.0211, 0.0112], device='cuda:0'), grad: tensor([ 0.0195, -0.0419, 0.0349, -0.0360, 0.0280, -0.0401, 0.0168, 0.0244, + -0.0049, -0.0007], device='cuda:0') +100 +0.0001 +changing lr +epoch 378, time 228.79, cls_loss 0.5196 cls_loss_mapping 0.0028 cls_loss_causal 0.4497 re_mapping 0.0089 re_causal 0.0202 /// teacc 98.73 lr 0.00010000 +Epoch 380, weight, value: tensor([[-0.2060, -0.0204, 0.0309, ..., -0.0302, -0.0844, -0.1480], + [-0.0586, -0.1058, 0.0347, ..., 0.0546, -0.0209, -0.0933], + [-0.0690, -0.0951, -0.1171, ..., 0.0381, -0.0379, -0.1173], + ..., + [-0.0665, 0.0311, 0.0254, ..., 0.0164, -0.0644, -0.1000], + [-0.0825, -0.0054, 0.0229, ..., 0.0637, -0.0311, -0.1832], + [ 0.0635, 0.0678, -0.0539, ..., -0.1118, -0.0098, 0.0945]], + device='cuda:0'), grad: tensor([[-2.2659e-03, -1.8644e-03, -1.2865e-03, ..., -2.8858e-03, + 1.3306e-02, -5.4283e-03], + [ 4.3273e-05, 1.2457e-04, 1.1235e-04, ..., 8.8406e-04, + 6.2883e-05, 2.6011e-04], + [ 1.7512e-04, 2.4652e-04, 3.1877e-04, ..., -6.9523e-04, + 1.2177e-04, 5.9557e-04], + ..., + [ 1.6165e-04, 1.5891e-04, 3.1781e-04, ..., -3.7599e-04, + 3.1924e-04, 3.6502e-04], + [ 2.7370e-04, 1.3533e-03, -3.2635e-03, ..., -3.0060e-03, + -2.7122e-03, -1.3475e-03], + [ 4.0460e-04, 1.3742e-03, 4.2081e-04, ..., 1.8778e-03, + 9.0265e-04, 1.4238e-03]], device='cuda:0') +Epoch 380, bias, value: tensor([-0.0541, 0.0227, 0.0066, -0.0105, -0.0026, 0.0087, -0.0147, 0.0334, + -0.0211, 0.0118], device='cuda:0'), grad: tensor([-0.0005, -0.0103, -0.0062, 0.0122, 0.0060, 0.0338, -0.0616, -0.0051, + 0.0019, 0.0297], device='cuda:0') +100 +0.0001 +changing lr +epoch 379, time 227.42, cls_loss 0.5077 cls_loss_mapping 0.0030 cls_loss_causal 0.4474 re_mapping 0.0090 re_causal 0.0204 /// teacc 98.79 lr 0.00010000 +Epoch 381, weight, value: tensor([[-0.2053, -0.0208, 0.0322, ..., -0.0280, -0.0853, -0.1469], + [-0.0598, -0.1054, 0.0345, ..., 0.0550, -0.0211, -0.0944], + [-0.0689, -0.0963, -0.1177, ..., 0.0378, -0.0377, -0.1181], + ..., + [-0.0668, 0.0314, 0.0257, ..., 0.0166, -0.0642, -0.0999], + [-0.0836, -0.0058, 0.0229, ..., 0.0641, -0.0319, -0.1843], + [ 0.0646, 0.0671, -0.0541, ..., -0.1116, -0.0094, 0.0961]], + device='cuda:0'), grad: tensor([[ 1.4544e-03, 3.4839e-05, 2.5291e-03, ..., 1.8892e-03, + -5.5879e-09, 9.3079e-04], + [ 2.0474e-05, -1.1757e-05, 3.4124e-05, ..., -1.4305e-03, + 3.7253e-09, 1.3217e-05], + [ 3.6216e-04, 4.3839e-05, 6.3705e-04, ..., 9.5177e-04, + 0.0000e+00, 2.2626e-04], + ..., + [ 2.4378e-04, -8.0299e-04, 1.6046e-04, ..., 7.2432e-04, + 1.0803e-07, 2.7180e-04], + [ 2.2519e-04, 1.4811e-03, 7.3373e-05, ..., -2.7256e-03, + 0.0000e+00, 3.6430e-03], + [ 3.4118e-04, 6.5947e-04, 2.9335e-03, ..., 4.8876e-04, + 2.6636e-07, 2.4490e-03]], device='cuda:0') +Epoch 381, bias, value: tensor([-0.0536, 0.0235, 0.0063, -0.0102, -0.0041, 0.0080, -0.0138, 0.0330, + -0.0210, 0.0121], device='cuda:0'), grad: tensor([ 0.0180, -0.0717, 0.0150, 0.0104, -0.0453, 0.0236, 0.0216, 0.0133, + -0.0068, 0.0219], device='cuda:0') +100 +0.0001 +changing lr +epoch 380, time 227.92, cls_loss 0.5241 cls_loss_mapping 0.0027 cls_loss_causal 0.4742 re_mapping 0.0091 re_causal 0.0200 /// teacc 98.74 lr 0.00010000 +Epoch 382, weight, value: tensor([[-0.2064, -0.0203, 0.0326, ..., -0.0286, -0.0837, -0.1463], + [-0.0603, -0.1054, 0.0343, ..., 0.0559, -0.0207, -0.0960], + [-0.0686, -0.0946, -0.1183, ..., 0.0379, -0.0374, -0.1209], + ..., + [-0.0673, 0.0313, 0.0277, ..., 0.0160, -0.0639, -0.0993], + [-0.0842, -0.0067, 0.0224, ..., 0.0644, -0.0329, -0.1858], + [ 0.0656, 0.0670, -0.0532, ..., -0.1111, -0.0095, 0.0973]], + device='cuda:0'), grad: tensor([[ 1.1647e-04, 1.1510e-04, 7.9060e-04, ..., 6.5327e-04, + 7.6389e-04, 1.1081e-04], + [ 1.3161e-04, 1.5354e-03, 7.6914e-04, ..., 1.2703e-03, + 1.6575e-03, 1.9920e-04], + [ 5.7936e-04, -4.8180e-03, 1.6975e-04, ..., -4.2229e-03, + -7.8812e-03, 2.3019e-04], + ..., + [ 9.3699e-04, 9.7132e-04, 2.2526e-03, ..., 5.9557e-04, + 2.1400e-03, 2.4235e-04], + [ 5.9366e-04, 4.3249e-04, -5.3139e-03, ..., -1.8587e-03, + -1.6241e-03, 1.5736e-04], + [ 3.0845e-06, 1.6308e-04, 1.5574e-03, ..., 1.1244e-03, + 1.6003e-03, 2.7031e-05]], device='cuda:0') +Epoch 382, bias, value: tensor([-0.0533, 0.0228, 0.0060, -0.0117, -0.0043, 0.0091, -0.0142, 0.0343, + -0.0219, 0.0133], device='cuda:0'), grad: tensor([ 0.0145, 0.0493, -0.0130, 0.0114, -0.0779, -0.0211, 0.0261, -0.0010, + -0.0118, 0.0235], device='cuda:0') +100 +0.0001 +changing lr +epoch 381, time 227.80, cls_loss 0.5312 cls_loss_mapping 0.0029 cls_loss_causal 0.4680 re_mapping 0.0092 re_causal 0.0214 /// teacc 98.85 lr 0.00010000 +Epoch 383, weight, value: tensor([[-0.2064, -0.0201, 0.0326, ..., -0.0283, -0.0832, -0.1468], + [-0.0603, -0.1063, 0.0333, ..., 0.0558, -0.0216, -0.0960], + [-0.0699, -0.0946, -0.1191, ..., 0.0376, -0.0353, -0.1180], + ..., + [-0.0673, 0.0298, 0.0280, ..., 0.0153, -0.0659, -0.0984], + [-0.0837, -0.0068, 0.0220, ..., 0.0644, -0.0338, -0.1855], + [ 0.0666, 0.0684, -0.0526, ..., -0.1110, -0.0101, 0.0966]], + device='cuda:0'), grad: tensor([[ 6.4313e-05, 4.5449e-06, 1.8990e-04, ..., 1.7109e-03, + 5.5981e-04, 5.4762e-06], + [ 2.9013e-05, 8.5086e-06, 6.8307e-05, ..., -4.7340e-03, + 3.6860e-04, 5.2601e-06], + [ 4.1056e-04, 5.9307e-06, 5.9175e-04, ..., -4.8332e-03, + 1.1034e-03, 1.6296e-04], + ..., + [ 3.1143e-05, 7.5293e-04, 8.0228e-05, ..., 2.5330e-03, + 4.9543e-04, -4.1515e-05], + [-3.3236e-04, 2.0778e-04, -3.4459e-06, ..., 6.1941e-04, + 1.6289e-03, 1.4007e-04], + [ 1.2314e-04, -2.0294e-03, 2.2817e-04, ..., 1.9531e-03, + 8.2064e-04, 8.4341e-05]], device='cuda:0') +Epoch 383, bias, value: tensor([-0.0536, 0.0221, 0.0064, -0.0122, -0.0028, 0.0087, -0.0148, 0.0331, + -0.0212, 0.0144], device='cuda:0'), grad: tensor([ 0.0240, -0.0474, -0.0109, -0.0184, -0.0025, 0.0283, -0.0236, 0.0216, + 0.0131, 0.0159], device='cuda:0') +100 +0.0001 +changing lr +epoch 382, time 227.95, cls_loss 0.5066 cls_loss_mapping 0.0021 cls_loss_causal 0.4498 re_mapping 0.0087 re_causal 0.0206 /// teacc 98.84 lr 0.00010000 +Epoch 384, weight, value: tensor([[-0.2072, -0.0213, 0.0330, ..., -0.0275, -0.0829, -0.1480], + [-0.0589, -0.1057, 0.0331, ..., 0.0549, -0.0223, -0.0973], + [-0.0682, -0.0954, -0.1193, ..., 0.0374, -0.0357, -0.1190], + ..., + [-0.0674, 0.0314, 0.0282, ..., 0.0153, -0.0652, -0.0980], + [-0.0851, -0.0076, 0.0212, ..., 0.0641, -0.0340, -0.1858], + [ 0.0665, 0.0679, -0.0516, ..., -0.1127, -0.0114, 0.0982]], + device='cuda:0'), grad: tensor([[-2.2640e-03, 1.1116e-04, 1.6928e-04, ..., 1.0824e-03, + 7.2718e-04, 2.2256e-04], + [ 1.6153e-05, 1.4126e-04, 8.7786e-04, ..., 2.2926e-03, + 3.9077e-04, 6.8712e-04], + [ 3.9250e-05, 5.2452e-05, 5.3072e-04, ..., 3.1548e-03, + 6.5374e-04, 3.0708e-04], + ..., + [ 2.1309e-05, -4.3201e-04, 9.1505e-04, ..., -8.6546e-04, + 9.0742e-04, 6.8521e-04], + [ 4.6682e-04, -1.2150e-03, -4.3449e-03, ..., -3.3798e-03, + -1.7881e-04, -4.0398e-03], + [ 1.8382e-04, 5.3740e-04, 7.0238e-04, ..., -1.7061e-03, + 1.9875e-03, -2.2638e-04]], device='cuda:0') +Epoch 384, bias, value: tensor([-0.0531, 0.0224, 0.0059, -0.0109, -0.0020, 0.0085, -0.0155, 0.0336, + -0.0215, 0.0126], device='cuda:0'), grad: tensor([ 0.0052, -0.0105, 0.0149, 0.0228, 0.0161, 0.0114, -0.0421, -0.0129, + -0.0072, 0.0022], device='cuda:0') +100 +0.0001 +changing lr +epoch 383, time 227.61, cls_loss 0.5382 cls_loss_mapping 0.0017 cls_loss_causal 0.4692 re_mapping 0.0085 re_causal 0.0206 /// teacc 98.86 lr 0.00010000 +Epoch 385, weight, value: tensor([[-0.2086, -0.0223, 0.0330, ..., -0.0284, -0.0843, -0.1474], + [-0.0601, -0.1059, 0.0338, ..., 0.0552, -0.0233, -0.0979], + [-0.0665, -0.0965, -0.1187, ..., 0.0377, -0.0351, -0.1185], + ..., + [-0.0684, 0.0320, 0.0276, ..., 0.0171, -0.0628, -0.0987], + [-0.0868, -0.0075, 0.0206, ..., 0.0620, -0.0341, -0.1869], + [ 0.0660, 0.0682, -0.0523, ..., -0.1125, -0.0108, 0.1007]], + device='cuda:0'), grad: tensor([[ 1.4771e-06, 1.9908e-04, 1.3912e-04, ..., -5.0735e-04, + -1.1429e-02, 7.9095e-05], + [ 9.2834e-06, 5.2595e-04, 6.0177e-04, ..., -1.1148e-03, + 1.4038e-03, 1.9002e-04], + [ 2.7731e-05, -3.0594e-03, 1.8573e-04, ..., -9.2545e-03, + -3.8872e-03, 1.4329e-04], + ..., + [ 7.9796e-06, 1.0653e-03, 4.6754e-04, ..., 3.3779e-03, + 2.4338e-03, 1.8430e-04], + [ 1.2219e-05, 4.6325e-04, 5.4932e-04, ..., -1.3075e-03, + 3.0346e-03, 3.1400e-04], + [ 6.9737e-06, 2.8896e-04, 1.1005e-03, ..., 2.6588e-03, + 4.6883e-03, 5.8889e-04]], device='cuda:0') +Epoch 385, bias, value: tensor([-0.0534, 0.0235, 0.0061, -0.0116, -0.0031, 0.0088, -0.0140, 0.0337, + -0.0228, 0.0130], device='cuda:0'), grad: tensor([-0.0078, -0.0030, -0.0246, -0.0089, 0.0166, -0.0040, 0.0244, 0.0262, + -0.0110, -0.0079], device='cuda:0') +100 +0.0001 +changing lr +epoch 384, time 229.77, cls_loss 0.5345 cls_loss_mapping 0.0022 cls_loss_causal 0.4651 re_mapping 0.0086 re_causal 0.0203 /// teacc 98.89 lr 0.00010000 +Epoch 386, weight, value: tensor([[-0.2081, -0.0232, 0.0340, ..., -0.0293, -0.0837, -0.1455], + [-0.0612, -0.1064, 0.0329, ..., 0.0551, -0.0245, -0.0984], + [-0.0665, -0.0970, -0.1188, ..., 0.0377, -0.0338, -0.1192], + ..., + [-0.0684, 0.0330, 0.0269, ..., 0.0170, -0.0626, -0.0999], + [-0.0863, -0.0066, 0.0199, ..., 0.0630, -0.0343, -0.1890], + [ 0.0663, 0.0683, -0.0505, ..., -0.1125, -0.0119, 0.1020]], + device='cuda:0'), grad: tensor([[-3.2723e-05, -1.8967e-02, 4.3273e-04, ..., -1.6556e-03, + 4.5919e-04, 2.8872e-04], + [ 1.1611e-04, 4.6074e-05, 2.1625e-04, ..., 9.8038e-04, + -1.5345e-03, 4.8375e-04], + [-3.1605e-03, 1.5736e-04, 3.9840e-04, ..., 7.9870e-04, + 4.7088e-04, -1.4567e-04], + ..., + [ 9.8801e-04, -9.2387e-05, 1.9875e-03, ..., 2.1210e-03, + 6.6376e-04, 1.5631e-03], + [ 2.6178e-04, 9.0933e-04, 7.0095e-04, ..., 1.6327e-03, + 3.7193e-04, 7.2908e-04], + [-1.8110e-03, 1.8358e-04, -2.3422e-03, ..., -2.6093e-03, + 3.8576e-04, -3.3092e-03]], device='cuda:0') +Epoch 386, bias, value: tensor([-0.0539, 0.0234, 0.0061, -0.0122, -0.0034, 0.0092, -0.0139, 0.0346, + -0.0236, 0.0136], device='cuda:0'), grad: tensor([-0.0346, -0.0165, 0.0107, 0.0015, -0.0536, 0.0470, 0.0275, 0.0194, + 0.0221, -0.0236], device='cuda:0') +100 +0.0001 +changing lr +epoch 385, time 228.47, cls_loss 0.5097 cls_loss_mapping 0.0023 cls_loss_causal 0.4515 re_mapping 0.0081 re_causal 0.0196 /// teacc 98.86 lr 0.00010000 +Epoch 387, weight, value: tensor([[-0.2087, -0.0210, 0.0367, ..., -0.0285, -0.0842, -0.1459], + [-0.0607, -0.1073, 0.0333, ..., 0.0547, -0.0250, -0.0958], + [-0.0660, -0.0973, -0.1195, ..., 0.0376, -0.0347, -0.1199], + ..., + [-0.0686, 0.0328, 0.0267, ..., 0.0178, -0.0618, -0.1005], + [-0.0860, -0.0063, 0.0186, ..., 0.0627, -0.0344, -0.1885], + [ 0.0666, 0.0688, -0.0512, ..., -0.1126, -0.0124, 0.1018]], + device='cuda:0'), grad: tensor([[ 7.8045e-07, -3.7514e-06, -7.1526e-06, ..., 1.1759e-03, + 2.8682e-04, -3.4515e-06], + [ 4.0457e-06, 1.9938e-05, 4.7326e-05, ..., -1.8854e-03, + 1.2827e-04, 1.1008e-06], + [ 4.7982e-06, 1.9252e-05, 2.0161e-05, ..., 2.9984e-03, + 6.7234e-04, 1.3959e-04], + ..., + [ 1.4806e-04, 1.1849e-04, -1.5366e-04, ..., -1.0223e-03, + 2.6274e-04, 8.6844e-05], + [ 8.7619e-05, 2.3232e-03, 1.1615e-05, ..., 2.0447e-03, + 1.2341e-03, 1.8053e-03], + [-3.1805e-04, 4.2908e-02, 3.9726e-05, ..., 4.5598e-05, + 1.3387e-04, -1.1170e-04]], device='cuda:0') +Epoch 387, bias, value: tensor([-0.0534, 0.0230, 0.0061, -0.0131, -0.0039, 0.0084, -0.0128, 0.0355, + -0.0233, 0.0135], device='cuda:0'), grad: tensor([ 0.0203, -0.0031, 0.0493, -0.0435, -0.0616, -0.0164, 0.0241, -0.0050, + 0.0322, 0.0038], device='cuda:0') +100 +0.0001 +changing lr +epoch 386, time 226.60, cls_loss 0.5297 cls_loss_mapping 0.0031 cls_loss_causal 0.4680 re_mapping 0.0080 re_causal 0.0192 /// teacc 98.66 lr 0.00010000 +Epoch 388, weight, value: tensor([[-0.2106, -0.0204, 0.0361, ..., -0.0285, -0.0840, -0.1478], + [-0.0606, -0.1078, 0.0327, ..., 0.0551, -0.0260, -0.0926], + [-0.0661, -0.0977, -0.1196, ..., 0.0380, -0.0361, -0.1200], + ..., + [-0.0687, 0.0331, 0.0271, ..., 0.0173, -0.0600, -0.1009], + [-0.0864, -0.0068, 0.0179, ..., 0.0623, -0.0337, -0.1892], + [ 0.0663, 0.0687, -0.0514, ..., -0.1128, -0.0131, 0.1012]], + device='cuda:0'), grad: tensor([[ 7.2531e-06, 1.8433e-05, 3.9530e-04, ..., 2.8133e-05, + 6.4850e-04, 3.1412e-05], + [ 6.7279e-06, 8.1837e-05, 2.1207e-04, ..., 2.7299e-04, + 3.5310e-04, 1.2612e-04], + [ 1.4611e-05, 1.6999e-04, 6.5374e-04, ..., -3.6430e-03, + 1.0624e-03, 2.3878e-04], + ..., + [ 8.1435e-06, -8.0013e-04, 3.8028e-04, ..., -2.4796e-04, + 7.1144e-04, -1.1578e-03], + [-7.3433e-04, 7.6413e-05, 2.5063e-03, ..., 1.1390e-04, + 3.2864e-03, -3.2024e-03], + [ 3.2365e-05, 1.4722e-04, 5.2261e-04, ..., 6.6578e-05, + 7.8249e-04, 2.3007e-04]], device='cuda:0') +Epoch 388, bias, value: tensor([-0.0542, 0.0242, 0.0056, -0.0120, -0.0036, 0.0086, -0.0144, 0.0358, + -0.0232, 0.0130], device='cuda:0'), grad: tensor([ 0.0044, -0.0253, -0.0045, 0.0279, 0.0009, -0.0211, 0.0097, -0.0021, + 0.0040, 0.0061], device='cuda:0') +100 +0.0001 +changing lr +epoch 387, time 226.99, cls_loss 0.5191 cls_loss_mapping 0.0031 cls_loss_causal 0.4547 re_mapping 0.0085 re_causal 0.0208 /// teacc 98.87 lr 0.00010000 +Epoch 389, weight, value: tensor([[-0.2108, -0.0203, 0.0362, ..., -0.0280, -0.0839, -0.1476], + [-0.0611, -0.1083, 0.0322, ..., 0.0546, -0.0267, -0.0928], + [-0.0659, -0.0973, -0.1200, ..., 0.0379, -0.0356, -0.1198], + ..., + [-0.0688, 0.0330, 0.0275, ..., 0.0169, -0.0606, -0.1013], + [-0.0866, -0.0060, 0.0176, ..., 0.0621, -0.0332, -0.1890], + [ 0.0676, 0.0678, -0.0511, ..., -0.1127, -0.0128, 0.1012]], + device='cuda:0'), grad: tensor([[-5.9128e-04, -5.3692e-04, -1.1575e-04, ..., 3.8552e-04, + 8.3971e-04, -8.1444e-04], + [ 1.1194e-06, 1.0923e-05, 1.2264e-03, ..., 1.9817e-03, + 2.2411e-03, 1.3206e-06], + [ 3.0899e-04, 2.9802e-04, 1.1768e-03, ..., 2.2926e-03, + 7.4625e-04, 3.7551e-04], + ..., + [ 2.5202e-06, 1.6794e-05, -6.3515e-03, ..., -8.0338e-03, + -1.0307e-02, 6.5789e-06], + [ 2.4196e-06, -7.1831e-03, 1.0929e-03, ..., 1.3132e-03, + 1.6880e-03, 9.2864e-05], + [ 6.6981e-06, 3.2949e-04, 5.9175e-04, ..., 9.9850e-04, + 9.4318e-04, 8.0705e-05]], device='cuda:0') +Epoch 389, bias, value: tensor([-0.0546, 0.0248, 0.0066, -0.0115, -0.0037, 0.0072, -0.0134, 0.0350, + -0.0230, 0.0125], device='cuda:0'), grad: tensor([ 0.0029, 0.0124, 0.0202, -0.0457, 0.0196, -0.0086, 0.0198, -0.0177, + 0.0010, -0.0039], device='cuda:0') +100 +0.0001 +changing lr +epoch 388, time 227.28, cls_loss 0.5035 cls_loss_mapping 0.0020 cls_loss_causal 0.4490 re_mapping 0.0085 re_causal 0.0204 /// teacc 98.85 lr 0.00010000 +Epoch 390, weight, value: tensor([[-0.2123, -0.0207, 0.0362, ..., -0.0280, -0.0837, -0.1487], + [-0.0621, -0.1086, 0.0336, ..., 0.0548, -0.0277, -0.0935], + [-0.0656, -0.0974, -0.1203, ..., 0.0381, -0.0359, -0.1187], + ..., + [-0.0689, 0.0339, 0.0279, ..., 0.0181, -0.0613, -0.1004], + [-0.0884, -0.0062, 0.0175, ..., 0.0608, -0.0339, -0.1891], + [ 0.0681, 0.0676, -0.0511, ..., -0.1128, -0.0132, 0.1010]], + device='cuda:0'), grad: tensor([[ 1.1832e-04, 1.7130e-04, 5.8222e-04, ..., 9.2888e-04, + 1.4186e-04, 3.6550e-04], + [ 1.8537e-04, 1.0031e-04, -1.0498e-02, ..., -1.0986e-03, + 4.2319e-06, 2.8563e-04], + [ 1.7333e-04, -1.1377e-03, 9.4318e-04, ..., 9.2602e-04, + 1.0049e-04, 1.0929e-03], + ..., + [-8.7881e-04, 2.1625e-04, 6.3086e-04, ..., -1.5926e-03, + -1.3256e-04, 1.1358e-03], + [ 2.8634e-04, -1.3981e-03, 8.5735e-04, ..., 2.7442e-04, + 9.4414e-05, -3.5267e-03], + [ 4.8494e-04, 4.7588e-04, 9.7752e-04, ..., 2.2583e-03, + 5.8293e-05, 1.7977e-03]], device='cuda:0') +Epoch 390, bias, value: tensor([-0.0553, 0.0254, 0.0071, -0.0109, -0.0039, 0.0079, -0.0147, 0.0357, + -0.0235, 0.0121], device='cuda:0'), grad: tensor([ 0.0144, -0.0409, 0.0062, 0.0585, -0.0439, -0.0028, 0.0332, -0.0399, + -0.0078, 0.0231], device='cuda:0') +100 +0.0001 +changing lr +epoch 389, time 229.77, cls_loss 0.5240 cls_loss_mapping 0.0025 cls_loss_causal 0.4619 re_mapping 0.0079 re_causal 0.0192 /// teacc 98.89 lr 0.00010000 +Epoch 391, weight, value: tensor([[-0.2135, -0.0206, 0.0361, ..., -0.0282, -0.0840, -0.1486], + [-0.0631, -0.1088, 0.0329, ..., 0.0549, -0.0265, -0.0940], + [-0.0648, -0.0976, -0.1203, ..., 0.0376, -0.0358, -0.1204], + ..., + [-0.0693, 0.0354, 0.0280, ..., 0.0185, -0.0620, -0.1014], + [-0.0874, -0.0067, 0.0177, ..., 0.0608, -0.0342, -0.1886], + [ 0.0688, 0.0671, -0.0508, ..., -0.1141, -0.0133, 0.1020]], + device='cuda:0'), grad: tensor([[ 1.4022e-05, -7.7188e-06, 2.6774e-04, ..., 1.9321e-03, + 1.2760e-03, 9.7081e-06], + [ 3.9279e-05, 3.1665e-08, 3.6120e-05, ..., 2.5558e-03, + 4.8828e-04, 2.9624e-05], + [ 1.4079e-04, 4.5728e-07, 2.7490e-04, ..., 1.7033e-03, + 9.4032e-04, 1.0335e-04], + ..., + [ 3.3826e-05, 7.2643e-08, 3.0041e-04, ..., 2.1381e-03, + 5.8603e-04, 2.4766e-05], + [ 2.7442e-04, 6.2026e-07, 2.6417e-04, ..., -1.2455e-03, + -7.5684e-03, 2.3139e-04], + [ 2.3520e-04, 3.5837e-06, -1.2627e-03, ..., -5.0163e-03, + -3.0708e-03, 1.9360e-04]], device='cuda:0') +Epoch 391, bias, value: tensor([-0.0551, 0.0252, 0.0069, -0.0112, -0.0030, 0.0090, -0.0146, 0.0352, + -0.0237, 0.0112], device='cuda:0'), grad: tensor([ 0.0173, 0.0216, 0.0009, 0.0113, -0.0129, -0.0383, 0.0216, 0.0153, + -0.0258, -0.0109], device='cuda:0') +100 +0.0001 +changing lr +epoch 390, time 228.47, cls_loss 0.5305 cls_loss_mapping 0.0027 cls_loss_causal 0.4705 re_mapping 0.0084 re_causal 0.0196 /// teacc 98.93 lr 0.00010000 +Epoch 392, weight, value: tensor([[-0.2136, -0.0217, 0.0362, ..., -0.0282, -0.0829, -0.1473], + [-0.0614, -0.1081, 0.0335, ..., 0.0530, -0.0272, -0.0944], + [-0.0652, -0.0973, -0.1200, ..., 0.0396, -0.0354, -0.1190], + ..., + [-0.0690, 0.0356, 0.0275, ..., 0.0180, -0.0614, -0.1008], + [-0.0882, -0.0064, 0.0170, ..., 0.0604, -0.0343, -0.1886], + [ 0.0683, 0.0678, -0.0505, ..., -0.1131, -0.0146, 0.1023]], + device='cuda:0'), grad: tensor([[ 1.3864e-04, -3.5992e-03, -2.1660e-04, ..., 2.7275e-04, + -4.4376e-05, 7.1907e-04], + [ 1.7151e-05, 2.4033e-04, 1.6186e-06, ..., -6.3782e-03, + 3.1069e-06, -2.9926e-03], + [ 2.6131e-04, 2.7409e-03, 6.9737e-05, ..., 5.2185e-03, + 1.4389e-04, -1.1425e-03], + ..., + [ 5.9217e-05, 2.2564e-03, 9.5189e-05, ..., 6.4392e-03, + 6.2957e-06, 2.5616e-03], + [ 1.6952e-04, 3.8767e-04, 7.4089e-05, ..., 2.2926e-03, + 8.6844e-05, 6.4850e-04], + [-6.0987e-04, 1.4849e-03, 2.3711e-04, ..., -2.8114e-03, + 1.5840e-05, 2.0676e-03]], device='cuda:0') +Epoch 392, bias, value: tensor([-0.0546, 0.0240, 0.0071, -0.0109, -0.0026, 0.0101, -0.0154, 0.0350, + -0.0237, 0.0111], device='cuda:0'), grad: tensor([-0.0055, -0.0262, 0.0164, 0.0406, -0.0344, 0.0305, -0.0166, 0.0419, + -0.0152, -0.0316], device='cuda:0') +100 +0.0001 +changing lr +epoch 391, time 228.75, cls_loss 0.5034 cls_loss_mapping 0.0033 cls_loss_causal 0.4443 re_mapping 0.0088 re_causal 0.0203 /// teacc 98.89 lr 0.00010000 +Epoch 393, weight, value: tensor([[-0.2135, -0.0220, 0.0365, ..., -0.0282, -0.0820, -0.1466], + [-0.0605, -0.1070, 0.0338, ..., 0.0549, -0.0276, -0.0936], + [-0.0673, -0.0981, -0.1203, ..., 0.0397, -0.0357, -0.1173], + ..., + [-0.0688, 0.0349, 0.0268, ..., 0.0172, -0.0604, -0.1019], + [-0.0873, -0.0062, 0.0178, ..., 0.0607, -0.0341, -0.1894], + [ 0.0678, 0.0680, -0.0519, ..., -0.1143, -0.0152, 0.1021]], + device='cuda:0'), grad: tensor([[ 3.3230e-06, 3.8862e-05, 3.5667e-03, ..., 3.4542e-03, + 4.0345e-06, 1.9569e-03], + [ 7.4413e-07, 3.3593e-04, 9.7275e-05, ..., 8.0633e-04, + 3.4869e-05, 5.2452e-04], + [ 1.0012e-06, 1.3793e-04, 1.2369e-03, ..., 2.6131e-04, + 1.4313e-05, 4.2367e-04], + ..., + [ 8.4378e-07, 4.4417e-04, 1.6284e-04, ..., 8.7214e-04, + 4.6104e-05, 7.0143e-04], + [ 1.4223e-05, 4.0746e-04, -1.0109e-03, ..., 6.2037e-04, + 4.2289e-05, 9.0313e-04], + [-2.9113e-06, -2.0847e-03, -2.0790e-04, ..., -1.8997e-03, + -2.1625e-04, -3.1414e-03]], device='cuda:0') +Epoch 393, bias, value: tensor([-0.0540, 0.0248, 0.0069, -0.0111, -0.0023, 0.0086, -0.0151, 0.0341, + -0.0228, 0.0111], device='cuda:0'), grad: tensor([ 0.0272, 0.0034, 0.0085, -0.0073, 0.0014, 0.0027, -0.0305, 0.0043, + -0.0028, -0.0068], device='cuda:0') +100 +0.0001 +changing lr +epoch 392, time 229.15, cls_loss 0.5250 cls_loss_mapping 0.0035 cls_loss_causal 0.4707 re_mapping 0.0085 re_causal 0.0199 /// teacc 98.81 lr 0.00010000 +Epoch 394, weight, value: tensor([[-0.2158, -0.0216, 0.0366, ..., -0.0286, -0.0821, -0.1484], + [-0.0607, -0.1072, 0.0337, ..., 0.0542, -0.0284, -0.0929], + [-0.0668, -0.0982, -0.1207, ..., 0.0413, -0.0365, -0.1172], + ..., + [-0.0688, 0.0351, 0.0277, ..., 0.0174, -0.0595, -0.1018], + [-0.0876, -0.0062, 0.0178, ..., 0.0620, -0.0347, -0.1892], + [ 0.0672, 0.0682, -0.0506, ..., -0.1135, -0.0153, 0.1019]], + device='cuda:0'), grad: tensor([[ 1.9178e-05, 3.0255e-04, 2.6155e-04, ..., 3.3474e-04, + 2.0885e-04, 4.5085e-04], + [-7.2718e-04, 1.4663e-05, -7.5579e-04, ..., -1.0576e-03, + 1.8030e-05, -8.1182e-05], + [ 4.1127e-06, 5.8740e-05, 3.4571e-04, ..., 4.3535e-04, + 1.2326e-04, 1.1015e-04], + ..., + [-5.2065e-05, 4.0698e-04, -4.5815e-03, ..., -1.3828e-03, + 1.8463e-05, 6.1369e-04], + [ 1.7548e-04, 2.4581e-04, 3.4084e-03, ..., 7.6294e-04, + 9.5069e-06, 3.8695e-04], + [ 8.1658e-05, -1.6947e-03, -8.7166e-04, ..., 9.1600e-04, + -8.3983e-05, -3.0174e-03]], device='cuda:0') +Epoch 394, bias, value: tensor([-0.0535, 0.0232, 0.0080, -0.0121, -0.0024, 0.0089, -0.0152, 0.0346, + -0.0225, 0.0112], device='cuda:0'), grad: tensor([-0.0077, 0.0052, 0.0151, 0.0142, -0.0168, -0.0218, 0.0260, 0.0029, + -0.0125, -0.0044], device='cuda:0') +100 +0.0001 +changing lr +epoch 393, time 225.84, cls_loss 0.5388 cls_loss_mapping 0.0022 cls_loss_causal 0.4828 re_mapping 0.0081 re_causal 0.0196 /// teacc 98.88 lr 0.00010000 +Epoch 395, weight, value: tensor([[-0.2161, -0.0215, 0.0363, ..., -0.0298, -0.0820, -0.1485], + [-0.0604, -0.1073, 0.0340, ..., 0.0536, -0.0281, -0.0921], + [-0.0652, -0.0972, -0.1197, ..., 0.0424, -0.0354, -0.1169], + ..., + [-0.0685, 0.0341, 0.0277, ..., 0.0179, -0.0598, -0.1011], + [-0.0869, -0.0064, 0.0183, ..., 0.0610, -0.0356, -0.1896], + [ 0.0661, 0.0683, -0.0514, ..., -0.1136, -0.0156, 0.1011]], + device='cuda:0'), grad: tensor([[ 8.5384e-06, -1.2708e-04, 2.2078e-04, ..., 1.7223e-03, + 5.1212e-04, 1.6489e-03], + [ 5.8562e-06, -8.6844e-05, 2.1696e-05, ..., -2.8152e-03, + -2.8458e-03, -6.7520e-03], + [ 4.5039e-06, 5.5850e-05, 2.3976e-05, ..., 2.2411e-03, + 6.8712e-04, 7.7295e-04], + ..., + [ 4.0144e-05, -3.8743e-04, -4.0078e-04, ..., 8.1205e-04, + 8.3804e-05, -4.5180e-04], + [ 1.9580e-05, -7.6771e-04, 2.1324e-05, ..., -2.9907e-03, + 7.0477e-04, -1.3866e-03], + [ 1.6427e-04, 6.6471e-04, 6.7425e-04, ..., 1.6098e-03, + 8.0347e-04, 1.1444e-03]], device='cuda:0') +Epoch 395, bias, value: tensor([-0.0543, 0.0226, 0.0078, -0.0117, -0.0032, 0.0090, -0.0138, 0.0355, + -0.0236, 0.0117], device='cuda:0'), grad: tensor([ 0.0090, -0.0457, 0.0258, -0.0054, -0.0157, -0.0008, 0.0226, 0.0141, + -0.0212, 0.0174], device='cuda:0') +100 +0.0001 +changing lr +epoch 394, time 227.69, cls_loss 0.5102 cls_loss_mapping 0.0022 cls_loss_causal 0.4589 re_mapping 0.0081 re_causal 0.0199 /// teacc 98.86 lr 0.00010000 +Epoch 396, weight, value: tensor([[-0.2172, -0.0208, 0.0358, ..., -0.0301, -0.0827, -0.1471], + [-0.0613, -0.1084, 0.0342, ..., 0.0542, -0.0280, -0.0918], + [-0.0649, -0.0981, -0.1195, ..., 0.0433, -0.0351, -0.1158], + ..., + [-0.0689, 0.0338, 0.0269, ..., 0.0173, -0.0597, -0.1007], + [-0.0878, -0.0067, 0.0169, ..., 0.0610, -0.0356, -0.1906], + [ 0.0660, 0.0682, -0.0510, ..., -0.1147, -0.0158, 0.0999]], + device='cuda:0'), grad: tensor([[-2.5593e-06, 3.6210e-05, 7.6443e-06, ..., 3.8433e-04, + 2.1141e-07, 2.8834e-06], + [ 4.2543e-06, 4.3154e-05, -3.5167e-05, ..., 7.7057e-04, + 2.1793e-07, 1.0408e-05], + [ 7.6145e-06, 1.2798e-03, 6.6273e-06, ..., -3.4885e-03, + 2.9698e-05, 1.3277e-05], + ..., + [ 5.9903e-05, -7.3910e-05, 8.2180e-06, ..., 3.1662e-03, + 5.2415e-06, -4.8137e-04], + [ 2.2396e-05, -2.3117e-03, 9.1344e-06, ..., -1.0717e-04, + -5.8860e-05, 4.8101e-05], + [-1.6356e-04, -1.0836e-04, -4.1910e-06, ..., -4.5242e-03, + 1.0729e-06, 2.3329e-04]], device='cuda:0') +Epoch 396, bias, value: tensor([-0.0546, 0.0228, 0.0082, -0.0113, -0.0024, 0.0088, -0.0144, 0.0347, + -0.0227, 0.0108], device='cuda:0'), grad: tensor([-0.0029, 0.0042, 0.0020, 0.0113, 0.0048, 0.0041, 0.0036, 0.0070, + -0.0074, -0.0267], device='cuda:0') +100 +0.0001 +changing lr +epoch 395, time 226.25, cls_loss 0.4934 cls_loss_mapping 0.0024 cls_loss_causal 0.4303 re_mapping 0.0082 re_causal 0.0192 /// teacc 98.83 lr 0.00010000 +Epoch 397, weight, value: tensor([[-0.2171, -0.0219, 0.0374, ..., -0.0294, -0.0819, -0.1466], + [-0.0612, -0.1110, 0.0345, ..., 0.0550, -0.0273, -0.0916], + [-0.0654, -0.0989, -0.1200, ..., 0.0432, -0.0354, -0.1163], + ..., + [-0.0701, 0.0353, 0.0253, ..., 0.0165, -0.0600, -0.1005], + [-0.0873, -0.0060, 0.0162, ..., 0.0623, -0.0362, -0.1915], + [ 0.0650, 0.0686, -0.0505, ..., -0.1161, -0.0158, 0.1000]], + device='cuda:0'), grad: tensor([[ 8.3923e-04, 4.5633e-04, -2.9888e-03, ..., -1.6985e-03, + 8.7097e-06, 5.1767e-05], + [ 8.9025e-04, -1.6785e-03, 6.6662e-04, ..., 3.0174e-03, + 5.5879e-07, 1.2323e-05], + [ 1.1759e-03, 3.8695e-04, 9.1791e-04, ..., -1.0376e-03, + 7.8976e-06, 2.1175e-05], + ..., + [-6.4735e-03, 2.5578e-03, 5.1689e-04, ..., -2.3937e-04, + 9.3132e-08, 2.5177e-04], + [-6.8436e-03, 3.8395e-03, -2.2755e-03, ..., -1.0460e-02, + 1.5432e-06, 4.8339e-05], + [-1.4484e-04, -8.5754e-03, 1.8394e-04, ..., 1.9684e-03, + 4.7404e-07, -5.5599e-04]], device='cuda:0') +Epoch 397, bias, value: tensor([-0.0536, 0.0235, 0.0072, -0.0117, -0.0022, 0.0085, -0.0138, 0.0344, + -0.0223, 0.0102], device='cuda:0'), grad: tensor([-0.0014, 0.0189, -0.0060, 0.0363, -0.0409, 0.0042, -0.0037, 0.0128, + -0.0278, 0.0075], device='cuda:0') +100 +0.0001 +changing lr +epoch 396, time 227.29, cls_loss 0.5251 cls_loss_mapping 0.0031 cls_loss_causal 0.4661 re_mapping 0.0086 re_causal 0.0208 /// teacc 98.77 lr 0.00010000 +Epoch 398, weight, value: tensor([[-0.2181, -0.0244, 0.0378, ..., -0.0288, -0.0820, -0.1472], + [-0.0623, -0.1100, 0.0344, ..., 0.0548, -0.0272, -0.0921], + [-0.0648, -0.1013, -0.1207, ..., 0.0411, -0.0384, -0.1164], + ..., + [-0.0697, 0.0341, 0.0251, ..., 0.0170, -0.0595, -0.0998], + [-0.0868, -0.0050, 0.0168, ..., 0.0630, -0.0364, -0.1915], + [ 0.0654, 0.0695, -0.0491, ..., -0.1174, -0.0160, 0.1008]], + device='cuda:0'), grad: tensor([[ 9.6112e-07, 1.0176e-03, 1.1578e-05, ..., 6.5947e-04, + 3.3565e-06, 2.1420e-07], + [ 1.8552e-06, 1.1778e-03, -1.6451e-04, ..., 1.2970e-03, + 8.9049e-05, 4.8243e-07], + [ 1.9714e-05, 1.0853e-03, 2.5369e-06, ..., 7.6628e-04, + 4.9859e-05, 2.0973e-06], + ..., + [-1.1462e-04, -8.5175e-05, 8.1360e-06, ..., -4.7255e-04, + -4.3082e-04, 6.2063e-06], + [ 4.9025e-05, 1.3876e-03, 1.0890e-04, ..., -2.8114e-03, + 4.1991e-05, 3.7014e-05], + [-1.1206e-04, -1.5459e-03, -1.0490e-05, ..., 9.5272e-04, + 5.3197e-05, -8.4937e-05]], device='cuda:0') +Epoch 398, bias, value: tensor([-0.0542, 0.0227, 0.0066, -0.0105, -0.0028, 0.0092, -0.0141, 0.0347, + -0.0224, 0.0107], device='cuda:0'), grad: tensor([ 0.0094, 0.0174, 0.0005, 0.0080, -0.0098, -0.0255, 0.0117, 0.0018, + -0.0199, 0.0064], device='cuda:0') +100 +0.0001 +changing lr +epoch 397, time 225.53, cls_loss 0.5075 cls_loss_mapping 0.0028 cls_loss_causal 0.4473 re_mapping 0.0083 re_causal 0.0200 /// teacc 98.98 lr 0.00010000 +Epoch 399, weight, value: tensor([[-0.2161, -0.0240, 0.0383, ..., -0.0279, -0.0814, -0.1466], + [-0.0627, -0.1113, 0.0352, ..., 0.0561, -0.0276, -0.0927], + [-0.0653, -0.0996, -0.1183, ..., 0.0422, -0.0387, -0.1167], + ..., + [-0.0700, 0.0339, 0.0244, ..., 0.0163, -0.0589, -0.0993], + [-0.0865, -0.0059, 0.0167, ..., 0.0632, -0.0362, -0.1912], + [ 0.0642, 0.0716, -0.0510, ..., -0.1183, -0.0155, 0.1002]], + device='cuda:0'), grad: tensor([[-3.7774e-06, 7.7009e-04, -9.9945e-04, ..., -1.5545e-03, + -4.3030e-03, -8.2552e-06], + [ 1.7509e-06, -2.8763e-03, -9.1076e-04, ..., -1.4801e-02, + 3.3712e-04, 1.7434e-06], + [ 4.9844e-06, 1.3294e-03, 2.3508e-04, ..., 3.0499e-03, + 2.0051e-04, 3.8669e-06], + ..., + [-7.1339e-06, 8.0633e-04, 2.9147e-05, ..., 2.5806e-03, + 2.7633e-04, -3.5390e-06], + [-1.4007e-05, 2.8000e-03, 7.7188e-05, ..., 3.1776e-03, + 2.4939e-04, -5.1148e-06], + [-4.5586e-04, -1.0139e-02, 1.5140e-04, ..., -9.7351e-03, + -1.6642e-03, -3.8815e-04]], device='cuda:0') +Epoch 399, bias, value: tensor([-0.0540, 0.0245, 0.0078, -0.0114, -0.0032, 0.0093, -0.0143, 0.0337, + -0.0229, 0.0106], device='cuda:0'), grad: tensor([-0.0200, -0.0762, 0.0168, 0.0090, 0.0663, 0.0183, 0.0184, 0.0145, + 0.0198, -0.0668], device='cuda:0') +100 +0.0001 +changing lr +epoch 398, time 227.74, cls_loss 0.5076 cls_loss_mapping 0.0021 cls_loss_causal 0.4390 re_mapping 0.0089 re_causal 0.0216 /// teacc 98.98 lr 0.00010000 +Epoch 400, weight, value: tensor([[-0.2151, -0.0231, 0.0379, ..., -0.0286, -0.0811, -0.1475], + [-0.0639, -0.1130, 0.0356, ..., 0.0562, -0.0285, -0.0933], + [-0.0661, -0.0997, -0.1184, ..., 0.0422, -0.0402, -0.1174], + ..., + [-0.0704, 0.0326, 0.0247, ..., 0.0182, -0.0579, -0.0988], + [-0.0859, -0.0060, 0.0163, ..., 0.0628, -0.0348, -0.1912], + [ 0.0644, 0.0726, -0.0514, ..., -0.1175, -0.0149, 0.1001]], + device='cuda:0'), grad: tensor([[ 1.9312e-05, -2.1072e-02, 2.6107e-04, ..., 4.7569e-03, + 1.5759e-04, 1.2290e-04], + [ 2.8327e-05, 1.6222e-03, 2.1112e-04, ..., 6.9695e-03, + 4.8697e-05, 9.9778e-05], + [-5.7995e-05, 2.2755e-03, 7.6447e-03, ..., 8.5144e-03, + 2.7537e-05, 6.8617e-04], + ..., + [ 2.4819e-04, 1.6279e-03, 2.9397e-04, ..., 1.5841e-03, + 3.5197e-05, 3.4928e-04], + [ 2.0866e-03, 1.9779e-03, 8.6308e-04, ..., -2.4948e-03, + -8.9121e-04, 2.3537e-03], + [ 5.9748e-04, 5.6458e-03, 4.1699e-04, ..., 2.8725e-03, + 2.3866e-04, 1.0900e-03]], device='cuda:0') +Epoch 400, bias, value: tensor([-0.0538, 0.0241, 0.0073, -0.0110, -0.0041, 0.0072, -0.0143, 0.0352, + -0.0233, 0.0124], device='cuda:0'), grad: tensor([-0.0140, 0.0272, 0.0298, 0.0186, -0.0499, -0.0269, -0.0349, 0.0103, + 0.0220, 0.0178], device='cuda:0') +100 +0.0001 +changing lr +epoch 399, time 226.18, cls_loss 0.5355 cls_loss_mapping 0.0036 cls_loss_causal 0.4799 re_mapping 0.0086 re_causal 0.0211 /// teacc 98.87 lr 0.00001000 +Epoch 401, weight, value: tensor([[-0.2150, -0.0220, 0.0376, ..., -0.0298, -0.0817, -0.1475], + [-0.0646, -0.1131, 0.0352, ..., 0.0568, -0.0291, -0.0941], + [-0.0651, -0.0999, -0.1191, ..., 0.0445, -0.0389, -0.1161], + ..., + [-0.0705, 0.0325, 0.0252, ..., 0.0153, -0.0598, -0.0998], + [-0.0862, -0.0049, 0.0159, ..., 0.0625, -0.0352, -0.1908], + [ 0.0649, 0.0721, -0.0516, ..., -0.1159, -0.0150, 0.1005]], + device='cuda:0'), grad: tensor([[ 3.1382e-05, -6.2981e-03, 1.2579e-03, ..., -3.4733e-03, + 3.4720e-05, 8.8140e-06], + [ 3.6031e-05, 2.6941e-04, -7.6580e-04, ..., 1.0357e-03, + 3.3236e-04, 1.1794e-05], + [ 1.4031e-04, 1.7395e-02, 3.2783e-04, ..., 7.4005e-03, + 1.9538e-04, 6.8545e-05], + ..., + [ 5.7888e-04, -3.6438e-02, 2.5177e-04, ..., -4.9515e-03, + 1.4305e-04, 4.6515e-04], + [-1.6584e-03, 2.2293e-02, -2.1935e-03, ..., 7.5102e-04, + -7.7057e-04, -1.0195e-03], + [-1.3268e-02, -5.1994e-03, -3.6449e-03, ..., -7.6485e-03, + -1.3573e-02, -3.8087e-05]], device='cuda:0') +Epoch 401, bias, value: tensor([-0.0541, 0.0243, 0.0088, -0.0115, -0.0035, 0.0066, -0.0146, 0.0336, + -0.0226, 0.0127], device='cuda:0'), grad: tensor([-0.0093, -0.0026, 0.0372, 0.0072, 0.0577, 0.0081, -0.0200, -0.0490, + 0.0187, -0.0481], device='cuda:0') +100 +1e-05 +changing lr +epoch 400, time 226.15, cls_loss 0.5022 cls_loss_mapping 0.0033 cls_loss_causal 0.4427 re_mapping 0.0083 re_causal 0.0199 /// teacc 98.84 lr 0.00001000 +Epoch 402, weight, value: tensor([[-0.2150, -0.0220, 0.0375, ..., -0.0297, -0.0817, -0.1475], + [-0.0646, -0.1126, 0.0352, ..., 0.0570, -0.0289, -0.0941], + [-0.0652, -0.1001, -0.1193, ..., 0.0443, -0.0389, -0.1162], + ..., + [-0.0705, 0.0326, 0.0252, ..., 0.0153, -0.0597, -0.0997], + [-0.0863, -0.0052, 0.0159, ..., 0.0625, -0.0351, -0.1908], + [ 0.0649, 0.0716, -0.0517, ..., -0.1162, -0.0152, 0.1004]], + device='cuda:0'), grad: tensor([[ 3.4809e-04, 4.4918e-04, 4.3273e-04, ..., 2.0256e-03, + 1.8799e-04, 4.2963e-04], + [ 3.1561e-05, -1.0902e-04, 5.0259e-04, ..., -2.4281e-03, + 2.4581e-04, 2.6777e-05], + [ 9.0361e-04, 7.8726e-04, 5.9891e-04, ..., 2.7637e-03, + 2.4891e-04, 1.0347e-03], + ..., + [ 1.9133e-04, 1.0312e-04, 8.3494e-04, ..., 1.8892e-03, + 3.8576e-04, 9.0778e-05], + [ 3.3436e-03, -8.0776e-04, 5.3482e-03, ..., 3.6602e-03, + 2.5272e-03, 1.3008e-03], + [ 4.2939e-04, -1.1625e-03, 1.1683e-03, ..., 2.2907e-03, + 5.3978e-04, 2.3949e-04]], device='cuda:0') +Epoch 402, bias, value: tensor([-0.0541, 0.0245, 0.0087, -0.0113, -0.0035, 0.0068, -0.0145, 0.0337, + -0.0227, 0.0122], device='cuda:0'), grad: tensor([ 0.0110, -0.0202, 0.0131, -0.0359, -0.0226, 0.0041, 0.0117, 0.0111, + 0.0212, 0.0066], device='cuda:0') +100 +1e-05 +changing lr +epoch 401, time 226.33, cls_loss 0.4877 cls_loss_mapping 0.0014 cls_loss_causal 0.4314 re_mapping 0.0079 re_causal 0.0194 /// teacc 98.85 lr 0.00001000 +Epoch 403, weight, value: tensor([[-0.2151, -0.0220, 0.0375, ..., -0.0298, -0.0816, -0.1475], + [-0.0645, -0.1126, 0.0353, ..., 0.0571, -0.0289, -0.0940], + [-0.0652, -0.1002, -0.1193, ..., 0.0441, -0.0388, -0.1162], + ..., + [-0.0706, 0.0326, 0.0253, ..., 0.0155, -0.0597, -0.0997], + [-0.0863, -0.0053, 0.0159, ..., 0.0626, -0.0350, -0.1908], + [ 0.0648, 0.0715, -0.0519, ..., -0.1163, -0.0153, 0.1003]], + device='cuda:0'), grad: tensor([[ 8.0466e-07, -4.6432e-05, 4.8935e-05, ..., 4.8137e-04, + 1.7583e-04, -7.9628e-07], + [ 2.6792e-05, 2.0601e-06, -6.6853e-04, ..., -4.8041e-04, + 1.4806e-04, 2.5034e-05], + [ 1.1558e-06, 1.9982e-05, 2.3556e-04, ..., 1.5011e-03, + 1.1816e-03, 1.0030e-06], + ..., + [ 1.9774e-05, -3.2157e-05, 1.0663e-04, ..., 1.1759e-03, + 3.1400e-04, 2.3961e-05], + [ 3.8594e-05, -3.2991e-05, 5.0020e-04, ..., 1.1616e-03, + 1.7941e-04, 3.7193e-05], + [-2.4204e-03, 7.1383e-04, -1.6174e-03, ..., -8.4782e-04, + -2.1315e-04, -2.1248e-03]], device='cuda:0') +Epoch 403, bias, value: tensor([-0.0541, 0.0244, 0.0087, -0.0113, -0.0034, 0.0069, -0.0146, 0.0339, + -0.0227, 0.0121], device='cuda:0'), grad: tensor([ 0.0010, -0.0010, 0.0053, 0.0029, -0.0165, 0.0027, 0.0033, 0.0031, + 0.0028, -0.0037], device='cuda:0') +100 +1e-05 +changing lr +epoch 402, time 226.75, cls_loss 0.4965 cls_loss_mapping 0.0015 cls_loss_causal 0.4325 re_mapping 0.0077 re_causal 0.0196 /// teacc 98.92 lr 0.00001000 +Epoch 404, weight, value: tensor([[-0.2150, -0.0221, 0.0374, ..., -0.0300, -0.0816, -0.1474], + [-0.0645, -0.1127, 0.0352, ..., 0.0572, -0.0290, -0.0940], + [-0.0652, -0.1002, -0.1195, ..., 0.0439, -0.0390, -0.1162], + ..., + [-0.0707, 0.0326, 0.0253, ..., 0.0156, -0.0598, -0.0998], + [-0.0862, -0.0052, 0.0158, ..., 0.0625, -0.0349, -0.1908], + [ 0.0648, 0.0713, -0.0519, ..., -0.1163, -0.0153, 0.1002]], + device='cuda:0'), grad: tensor([[ 1.6177e-04, 2.3282e-04, 1.1158e-03, ..., -3.7155e-03, + 2.9874e-04, 2.2256e-04], + [ 1.4043e-04, -2.0635e-04, 1.9722e-03, ..., 5.1804e-03, + 1.3864e-04, 1.9774e-05], + [-1.2189e-04, 2.2137e-04, 5.3138e-05, ..., -1.5163e-03, + 1.9342e-05, 2.4557e-05], + ..., + [ 1.0590e-02, 1.6565e-03, 1.8139e-03, ..., 8.4839e-03, + 2.2554e-04, 5.3520e-03], + [ 8.9645e-04, 4.9686e-04, 8.8692e-05, ..., -2.1935e-03, + 5.3257e-05, 3.0088e-04], + [ 1.8108e-04, -3.5534e-03, -1.4725e-02, ..., -1.9409e-02, + -2.3346e-03, -3.9339e-04]], device='cuda:0') +Epoch 404, bias, value: tensor([-0.0541, 0.0244, 0.0085, -0.0112, -0.0033, 0.0068, -0.0144, 0.0340, + -0.0228, 0.0119], device='cuda:0'), grad: tensor([-0.0163, 0.0228, -0.0195, -0.0319, 0.0316, 0.0254, 0.0228, 0.0503, + -0.0177, -0.0675], device='cuda:0') +100 +1e-05 +changing lr +epoch 403, time 224.88, cls_loss 0.4845 cls_loss_mapping 0.0013 cls_loss_causal 0.4312 re_mapping 0.0076 re_causal 0.0188 /// teacc 98.94 lr 0.00001000 +Epoch 405, weight, value: tensor([[-0.2150, -0.0220, 0.0376, ..., -0.0298, -0.0816, -0.1474], + [-0.0645, -0.1126, 0.0353, ..., 0.0572, -0.0289, -0.0940], + [-0.0652, -0.1004, -0.1195, ..., 0.0438, -0.0390, -0.1163], + ..., + [-0.0706, 0.0326, 0.0254, ..., 0.0156, -0.0598, -0.0999], + [-0.0864, -0.0053, 0.0157, ..., 0.0624, -0.0350, -0.1910], + [ 0.0650, 0.0712, -0.0519, ..., -0.1163, -0.0153, 0.1002]], + device='cuda:0'), grad: tensor([[ 1.0294e-04, 3.0079e-03, 7.9956e-03, ..., 9.4376e-03, + 1.6699e-03, 1.0830e-04], + [ 1.4700e-05, 3.4752e-03, 1.2693e-03, ..., -3.0632e-03, + 4.7708e-04, 9.7528e-06], + [ 3.9315e-04, 5.1689e-04, 1.1339e-03, ..., -7.5684e-03, + -1.0290e-03, 4.2295e-04], + ..., + [ 5.0038e-05, 4.0579e-04, -4.6997e-03, ..., -3.9597e-03, + -2.8286e-03, 4.5717e-05], + [-7.4267e-05, 2.0790e-03, 3.7518e-03, ..., 5.3902e-03, + 9.0170e-04, 6.1035e-05], + [-5.6863e-05, 4.3726e-04, 1.2646e-03, ..., 3.3150e-03, + 5.8031e-04, -2.1875e-05]], device='cuda:0') +Epoch 405, bias, value: tensor([-0.0540, 0.0245, 0.0085, -0.0113, -0.0032, 0.0068, -0.0144, 0.0340, + -0.0229, 0.0119], device='cuda:0'), grad: tensor([ 0.0401, -0.0337, -0.0398, -0.0153, 0.0226, -0.0054, -0.0071, -0.0073, + 0.0294, 0.0165], device='cuda:0') +100 +1e-05 +changing lr +epoch 404, time 225.01, cls_loss 0.4885 cls_loss_mapping 0.0011 cls_loss_causal 0.4280 re_mapping 0.0075 re_causal 0.0199 /// teacc 98.92 lr 0.00001000 +Epoch 406, weight, value: tensor([[-0.2151, -0.0220, 0.0375, ..., -0.0299, -0.0817, -0.1474], + [-0.0644, -0.1126, 0.0353, ..., 0.0572, -0.0290, -0.0941], + [-0.0652, -0.1003, -0.1195, ..., 0.0439, -0.0391, -0.1163], + ..., + [-0.0703, 0.0327, 0.0255, ..., 0.0158, -0.0596, -0.0998], + [-0.0865, -0.0054, 0.0157, ..., 0.0621, -0.0352, -0.1911], + [ 0.0649, 0.0711, -0.0517, ..., -0.1162, -0.0153, 0.1002]], + device='cuda:0'), grad: tensor([[ 2.5463e-04, 3.3379e-05, 7.8297e-04, ..., 1.9484e-03, + 1.4877e-04, 1.9193e-05], + [ 6.2656e-04, 3.7163e-05, 1.6661e-03, ..., -2.6093e-03, + 3.6597e-04, 4.8071e-05], + [ 3.0327e-04, -3.6860e-04, 1.5182e-03, ..., -1.9484e-03, + 1.5616e-04, 2.0593e-05], + ..., + [-3.2921e-03, 2.4438e-05, -1.6527e-03, ..., -5.9280e-03, + -2.1610e-03, -2.8467e-04], + [-1.3161e-04, -1.0386e-03, -1.1597e-02, ..., -7.7248e-03, + 2.7990e-04, 3.6806e-05], + [ 7.4577e-04, 4.4554e-05, 2.9907e-03, ..., 3.0766e-03, + 4.0913e-04, 5.4181e-05]], device='cuda:0') +Epoch 406, bias, value: tensor([-0.0541, 0.0246, 0.0086, -0.0112, -0.0034, 0.0068, -0.0144, 0.0341, + -0.0232, 0.0119], device='cuda:0'), grad: tensor([ 0.0145, -0.0163, -0.0287, 0.0204, 0.0144, 0.0250, 0.0195, -0.0416, + -0.0246, 0.0176], device='cuda:0') +100 +1e-05 +changing lr +epoch 405, time 225.20, cls_loss 0.4988 cls_loss_mapping 0.0013 cls_loss_causal 0.4418 re_mapping 0.0073 re_causal 0.0188 /// teacc 98.93 lr 0.00001000 +Epoch 407, weight, value: tensor([[-0.2151, -0.0219, 0.0376, ..., -0.0298, -0.0816, -0.1475], + [-0.0643, -0.1126, 0.0352, ..., 0.0571, -0.0290, -0.0941], + [-0.0653, -0.1003, -0.1196, ..., 0.0437, -0.0391, -0.1164], + ..., + [-0.0703, 0.0327, 0.0255, ..., 0.0159, -0.0596, -0.0998], + [-0.0864, -0.0053, 0.0158, ..., 0.0622, -0.0352, -0.1910], + [ 0.0650, 0.0711, -0.0517, ..., -0.1161, -0.0152, 0.1002]], + device='cuda:0'), grad: tensor([[ 3.6907e-04, -9.1248e-03, -1.0605e-02, ..., -2.0599e-02, + -1.9257e-02, -6.2561e-04], + [-2.4967e-03, 4.4751e-04, -4.1237e-03, ..., -3.7708e-03, + 6.8855e-04, -1.9741e-03], + [ 1.0175e-04, 1.6747e-03, 3.3417e-03, ..., 7.6332e-03, + 4.7188e-03, 1.9920e-04], + ..., + [ 3.4779e-05, 4.2796e-04, 4.6897e-04, ..., 2.8706e-03, + 1.0605e-03, 5.5909e-05], + [ 5.1498e-04, 9.8801e-04, 6.0883e-03, ..., 5.1384e-03, + 6.9695e-03, 4.7183e-04], + [-3.5477e-03, 1.4615e-04, -6.3515e-03, ..., -5.7030e-03, + -3.2768e-03, -3.3684e-03]], device='cuda:0') +Epoch 407, bias, value: tensor([-0.0541, 0.0246, 0.0085, -0.0113, -0.0033, 0.0068, -0.0144, 0.0342, + -0.0232, 0.0120], device='cuda:0'), grad: tensor([-0.0671, -0.0060, 0.0301, 0.0319, -0.0076, -0.0135, 0.0418, 0.0133, + 0.0218, -0.0448], device='cuda:0') +100 +1e-05 +changing lr +epoch 406, time 224.77, cls_loss 0.5280 cls_loss_mapping 0.0011 cls_loss_causal 0.4628 re_mapping 0.0071 re_causal 0.0189 /// teacc 98.89 lr 0.00001000 +Epoch 408, weight, value: tensor([[-0.2152, -0.0220, 0.0376, ..., -0.0299, -0.0817, -0.1476], + [-0.0641, -0.1126, 0.0353, ..., 0.0572, -0.0291, -0.0941], + [-0.0654, -0.1004, -0.1197, ..., 0.0438, -0.0390, -0.1165], + ..., + [-0.0702, 0.0328, 0.0255, ..., 0.0160, -0.0595, -0.0997], + [-0.0865, -0.0053, 0.0159, ..., 0.0622, -0.0352, -0.1909], + [ 0.0650, 0.0711, -0.0517, ..., -0.1162, -0.0153, 0.1001]], + device='cuda:0'), grad: tensor([[ 2.1718e-06, 2.6684e-03, 2.0313e-03, ..., 4.5013e-03, + 8.5607e-06, 3.4189e-04], + [ 1.9986e-06, 5.1880e-03, 9.8441e-07, ..., 4.0474e-03, + 1.3316e-04, 5.8556e-04], + [ 1.3895e-06, 1.3342e-03, 7.6056e-04, ..., -2.3605e-02, + -2.1477e-03, -9.3002e-03], + ..., + [-3.8706e-06, 2.2125e-03, 3.3602e-06, ..., 2.2488e-03, + 5.4747e-05, 2.4354e-04], + [ 3.6135e-06, -4.1733e-03, 1.4246e-04, ..., -1.1683e-03, + 6.6087e-06, 6.8963e-05], + [-5.8949e-05, 4.6272e-03, 2.1473e-05, ..., 4.2763e-03, + 1.6704e-05, -1.3065e-04]], device='cuda:0') +Epoch 408, bias, value: tensor([-0.0541, 0.0246, 0.0085, -0.0112, -0.0034, 0.0067, -0.0144, 0.0344, + -0.0231, 0.0119], device='cuda:0'), grad: tensor([ 3.4515e-02, 3.7720e-02, -3.6835e-02, -9.2850e-03, -4.7302e-03, + -4.6082e-02, -1.2199e-02, -4.6110e-04, -4.6313e-05, 3.7384e-02], + device='cuda:0') +100 +1e-05 +changing lr +epoch 407, time 226.53, cls_loss 0.4742 cls_loss_mapping 0.0011 cls_loss_causal 0.4144 re_mapping 0.0070 re_causal 0.0179 /// teacc 98.89 lr 0.00001000 +Epoch 409, weight, value: tensor([[-0.2152, -0.0221, 0.0377, ..., -0.0298, -0.0818, -0.1475], + [-0.0639, -0.1126, 0.0354, ..., 0.0572, -0.0290, -0.0941], + [-0.0655, -0.1003, -0.1197, ..., 0.0438, -0.0390, -0.1164], + ..., + [-0.0702, 0.0327, 0.0256, ..., 0.0160, -0.0594, -0.0997], + [-0.0865, -0.0055, 0.0159, ..., 0.0622, -0.0351, -0.1910], + [ 0.0649, 0.0711, -0.0518, ..., -0.1161, -0.0155, 0.1002]], + device='cuda:0'), grad: tensor([[-6.3889e-06, 2.9993e-04, -2.2411e-05, ..., 1.2207e-03, + 8.5682e-08, 9.0063e-05], + [-1.7481e-06, 1.0986e-03, -3.0354e-05, ..., -5.1727e-03, + 5.4948e-08, 3.8594e-06], + [ 1.0723e-04, 1.6298e-03, 1.2171e-04, ..., 1.6975e-03, + -1.0906e-06, 2.3222e-04], + ..., + [ 9.9599e-05, 8.0919e-04, 5.6416e-05, ..., 2.8343e-03, + -4.6641e-06, 1.2767e-04], + [ 2.7359e-05, -9.4757e-03, 4.3601e-05, ..., -9.3918e-03, + 1.1455e-07, 9.0778e-05], + [-2.2709e-05, 2.8763e-03, 3.5018e-05, ..., 2.5444e-03, + 4.7684e-06, 5.9992e-05]], device='cuda:0') +Epoch 409, bias, value: tensor([-0.0542, 0.0245, 0.0085, -0.0113, -0.0033, 0.0067, -0.0143, 0.0344, + -0.0231, 0.0119], device='cuda:0'), grad: tensor([ 0.0091, -0.0420, 0.0126, 0.0090, 0.0134, 0.0116, 0.0106, 0.0196, + -0.0610, 0.0172], device='cuda:0') +100 +1e-05 +changing lr +epoch 408, time 224.12, cls_loss 0.5056 cls_loss_mapping 0.0010 cls_loss_causal 0.4461 re_mapping 0.0071 re_causal 0.0183 /// teacc 98.95 lr 0.00001000 +Epoch 410, weight, value: tensor([[-0.2151, -0.0220, 0.0378, ..., -0.0296, -0.0819, -0.1475], + [-0.0639, -0.1125, 0.0354, ..., 0.0572, -0.0289, -0.0942], + [-0.0654, -0.1004, -0.1196, ..., 0.0438, -0.0390, -0.1163], + ..., + [-0.0702, 0.0328, 0.0255, ..., 0.0158, -0.0593, -0.0997], + [-0.0864, -0.0056, 0.0159, ..., 0.0622, -0.0350, -0.1910], + [ 0.0647, 0.0710, -0.0520, ..., -0.1161, -0.0155, 0.1001]], + device='cuda:0'), grad: tensor([[ 3.1018e-04, 5.4693e-04, 9.7394e-05, ..., 6.5947e-04, + 5.1051e-05, 4.2289e-05], + [ 7.4059e-06, -7.9041e-03, 5.4693e-04, ..., -7.7477e-03, + -1.7757e-03, -4.5562e-04], + [ 7.8857e-05, 4.4212e-03, 2.6560e-04, ..., 1.0910e-03, + 9.7692e-05, 9.1910e-05], + ..., + [ 1.6183e-05, 1.6823e-03, 1.5593e-04, ..., 7.6723e-04, + 5.8770e-05, 5.4240e-05], + [ 1.7297e-04, 2.1229e-03, -1.9836e-03, ..., 8.8215e-04, + 3.0780e-04, 1.5581e-04], + [ 1.8597e-05, 4.2915e-03, 1.7023e-04, ..., 9.6035e-04, + 9.5904e-05, -4.8071e-05]], device='cuda:0') +Epoch 410, bias, value: tensor([-0.0541, 0.0245, 0.0085, -0.0113, -0.0031, 0.0067, -0.0144, 0.0343, + -0.0231, 0.0118], device='cuda:0'), grad: tensor([ 0.0065, -0.0787, 0.0116, 0.0208, -0.0056, 0.0108, 0.0076, 0.0075, + 0.0087, 0.0108], device='cuda:0') +100 +1e-05 +changing lr +epoch 409, time 226.37, cls_loss 0.4972 cls_loss_mapping 0.0011 cls_loss_causal 0.4349 re_mapping 0.0070 re_causal 0.0181 /// teacc 98.90 lr 0.00001000 +Epoch 411, weight, value: tensor([[-0.2150, -0.0220, 0.0379, ..., -0.0296, -0.0820, -0.1475], + [-0.0640, -0.1126, 0.0352, ..., 0.0572, -0.0289, -0.0943], + [-0.0655, -0.1005, -0.1196, ..., 0.0438, -0.0390, -0.1164], + ..., + [-0.0701, 0.0328, 0.0255, ..., 0.0159, -0.0592, -0.0997], + [-0.0862, -0.0056, 0.0159, ..., 0.0621, -0.0349, -0.1909], + [ 0.0644, 0.0712, -0.0521, ..., -0.1163, -0.0156, 0.1000]], + device='cuda:0'), grad: tensor([[ 1.8135e-05, 5.0634e-05, -3.2783e-05, ..., 6.5231e-04, + 5.1308e-04, 5.1069e-04], + [ 1.8919e-04, 8.2791e-05, 5.3458e-06, ..., 4.0722e-04, + 9.9018e-06, 6.0558e-05], + [ 1.1516e-04, -9.6989e-04, -1.6809e-05, ..., -1.8854e-03, + 3.1605e-03, 3.7994e-03], + ..., + [-1.4400e-03, 3.1996e-04, 7.2777e-05, ..., 7.5674e-04, + 1.7226e-04, 1.8907e-04], + [ 3.1829e-05, 1.4246e-04, 2.6509e-05, ..., 2.9850e-04, + 1.0592e-04, 1.1301e-04], + [ 1.1063e-03, 5.2691e-04, -8.0884e-05, ..., 6.5851e-04, + -5.6915e-03, -3.9558e-03]], device='cuda:0') +Epoch 411, bias, value: tensor([-0.0541, 0.0245, 0.0084, -0.0113, -0.0031, 0.0069, -0.0145, 0.0344, + -0.0232, 0.0116], device='cuda:0'), grad: tensor([ 0.0174, -0.0215, 0.0203, -0.0196, -0.0206, 0.0105, 0.0161, 0.0073, + 0.0075, -0.0174], device='cuda:0') +100 +1e-05 +changing lr +epoch 410, time 223.32, cls_loss 0.5118 cls_loss_mapping 0.0010 cls_loss_causal 0.4490 re_mapping 0.0069 re_causal 0.0180 /// teacc 98.91 lr 0.00001000 +Epoch 412, weight, value: tensor([[-0.2150, -0.0220, 0.0379, ..., -0.0295, -0.0820, -0.1474], + [-0.0640, -0.1127, 0.0353, ..., 0.0572, -0.0288, -0.0943], + [-0.0654, -0.1004, -0.1197, ..., 0.0438, -0.0391, -0.1165], + ..., + [-0.0701, 0.0328, 0.0254, ..., 0.0158, -0.0593, -0.0996], + [-0.0863, -0.0057, 0.0161, ..., 0.0621, -0.0349, -0.1910], + [ 0.0646, 0.0711, -0.0521, ..., -0.1162, -0.0154, 0.1001]], + device='cuda:0'), grad: tensor([[ 2.9731e-04, 8.2016e-05, -1.7602e-07, ..., -5.6725e-03, + 0.0000e+00, 1.2541e-04], + [ 2.2147e-06, 1.1283e-04, 7.4040e-08, ..., 1.6050e-03, + 0.0000e+00, 2.8229e-04], + [ 1.2398e-04, -8.3590e-04, 4.2794e-07, ..., 6.7472e-04, + 6.0163e-07, 8.1730e-04], + ..., + [ 5.3644e-06, 9.2108e-07, 3.2689e-07, ..., -3.3989e-03, + 8.8476e-09, -4.1580e-03], + [-2.5959e-03, 4.4912e-05, -1.2934e-05, ..., 5.8699e-04, + 3.8650e-07, 1.5724e-04], + [ 2.1696e-05, 1.5342e-04, 1.0155e-05, ..., 2.5806e-03, + 1.8207e-07, 1.9293e-03]], device='cuda:0') +Epoch 412, bias, value: tensor([-0.0541, 0.0245, 0.0082, -0.0113, -0.0029, 0.0070, -0.0145, 0.0344, + -0.0231, 0.0116], device='cuda:0'), grad: tensor([-0.0146, 0.0075, -0.0049, -0.0133, 0.0056, 0.0137, 0.0050, -0.0057, + -0.0021, 0.0089], device='cuda:0') +100 +1e-05 +changing lr +epoch 411, time 224.71, cls_loss 0.4856 cls_loss_mapping 0.0010 cls_loss_causal 0.4196 re_mapping 0.0071 re_causal 0.0181 /// teacc 98.94 lr 0.00001000 +Epoch 413, weight, value: tensor([[-0.2152, -0.0220, 0.0378, ..., -0.0295, -0.0819, -0.1475], + [-0.0641, -0.1126, 0.0351, ..., 0.0572, -0.0288, -0.0942], + [-0.0655, -0.1005, -0.1198, ..., 0.0437, -0.0391, -0.1166], + ..., + [-0.0701, 0.0328, 0.0255, ..., 0.0157, -0.0593, -0.0996], + [-0.0862, -0.0057, 0.0161, ..., 0.0623, -0.0349, -0.1911], + [ 0.0646, 0.0712, -0.0522, ..., -0.1162, -0.0154, 0.1001]], + device='cuda:0'), grad: tensor([[ 4.1164e-07, 6.9904e-04, 1.1444e-03, ..., 1.2951e-03, + 5.4436e-03, 4.7803e-05], + [ 6.7521e-07, -1.0424e-03, -2.3403e-03, ..., -5.2071e-03, + 9.9850e-04, 5.8800e-05], + [ 6.2399e-07, 3.6478e-04, 4.1604e-04, ..., 1.0567e-03, + 2.2640e-03, 4.6730e-05], + ..., + [ 5.4240e-06, 8.8596e-04, 2.4164e-04, ..., -2.9049e-03, + 8.6784e-04, 8.0919e-04], + [ 7.2047e-06, -1.2207e-03, -8.6367e-05, ..., -2.9964e-03, + 2.5158e-03, -9.3794e-04], + [-2.8998e-05, -8.9844e-02, 1.6975e-04, ..., 2.3727e-03, + 2.7657e-03, -2.6474e-02]], device='cuda:0') +Epoch 413, bias, value: tensor([-0.0540, 0.0245, 0.0082, -0.0114, -0.0028, 0.0070, -0.0146, 0.0344, + -0.0230, 0.0115], device='cuda:0'), grad: tensor([ 0.0153, -0.0199, 0.0088, 0.0099, 0.0151, -0.0214, 0.0186, -0.0016, + -0.0171, -0.0077], device='cuda:0') +100 +1e-05 +changing lr +epoch 412, time 225.56, cls_loss 0.4698 cls_loss_mapping 0.0008 cls_loss_causal 0.4051 re_mapping 0.0069 re_causal 0.0176 /// teacc 98.94 lr 0.00001000 +Epoch 414, weight, value: tensor([[-0.2152, -0.0218, 0.0379, ..., -0.0293, -0.0819, -0.1474], + [-0.0639, -0.1128, 0.0352, ..., 0.0571, -0.0288, -0.0941], + [-0.0656, -0.1005, -0.1198, ..., 0.0437, -0.0390, -0.1166], + ..., + [-0.0702, 0.0327, 0.0253, ..., 0.0156, -0.0592, -0.0997], + [-0.0864, -0.0058, 0.0161, ..., 0.0623, -0.0348, -0.1911], + [ 0.0648, 0.0712, -0.0522, ..., -0.1162, -0.0155, 0.1001]], + device='cuda:0'), grad: tensor([[ 1.1373e-04, -8.3113e-04, 3.0088e-04, ..., 1.5783e-03, + -7.6294e-04, -1.2135e-06], + [ 6.8426e-05, 8.1599e-05, 6.3515e-04, ..., -5.3139e-03, + 2.1076e-04, 2.2799e-05], + [-5.4502e-04, 1.0151e-04, 3.0494e-04, ..., -2.4757e-03, + 1.2636e-04, 1.9893e-05], + ..., + [ 2.0468e-04, -6.9962e-03, 1.0548e-03, ..., -5.7297e-03, + 3.7217e-04, -3.3455e-03], + [ 1.2279e-04, -1.9455e-03, 4.0197e-04, ..., 1.2608e-03, + 3.4451e-05, 1.7798e-04], + [ 3.1877e-04, 2.0180e-03, 1.3742e-03, ..., 4.3068e-03, + 9.3889e-04, 1.3649e-04]], device='cuda:0') +Epoch 414, bias, value: tensor([-0.0538, 0.0245, 0.0081, -0.0114, -0.0028, 0.0070, -0.0147, 0.0343, + -0.0230, 0.0116], device='cuda:0'), grad: tensor([ 0.0071, -0.0181, -0.0255, 0.0031, -0.0203, 0.0170, 0.0119, -0.0072, + 0.0114, 0.0204], device='cuda:0') +100 +1e-05 +changing lr +epoch 413, time 224.64, cls_loss 0.4811 cls_loss_mapping 0.0009 cls_loss_causal 0.4239 re_mapping 0.0070 re_causal 0.0181 /// teacc 98.98 lr 0.00001000 +Epoch 415, weight, value: tensor([[-0.2153, -0.0219, 0.0378, ..., -0.0294, -0.0816, -0.1475], + [-0.0639, -0.1127, 0.0352, ..., 0.0571, -0.0289, -0.0941], + [-0.0657, -0.1006, -0.1198, ..., 0.0437, -0.0388, -0.1166], + ..., + [-0.0703, 0.0328, 0.0254, ..., 0.0157, -0.0593, -0.0998], + [-0.0863, -0.0059, 0.0162, ..., 0.0622, -0.0350, -0.1912], + [ 0.0647, 0.0712, -0.0523, ..., -0.1161, -0.0155, 0.1002]], + device='cuda:0'), grad: tensor([[ 5.4911e-06, 1.7434e-05, -2.1291e-04, ..., -9.2316e-03, + -1.5823e-02, 9.6858e-06], + [ 1.0997e-05, 2.0698e-05, -5.4270e-05, ..., 2.1763e-03, + 1.3876e-04, 6.8434e-06], + [ 2.7657e-04, 2.2945e-03, 3.4714e-04, ..., 6.8893e-03, + 9.8419e-04, 3.6602e-03], + ..., + [ 1.8492e-05, 4.0770e-05, 8.0466e-05, ..., 2.9755e-03, + 6.5422e-04, 1.7788e-06], + [ 5.9357e-03, 9.4452e-03, 5.2299e-03, ..., 9.9869e-03, + 2.3804e-03, 6.1035e-05], + [ 1.6856e-04, 5.4789e-04, 7.1764e-04, ..., -1.1301e-04, + 8.0347e-04, 3.7663e-06]], device='cuda:0') +Epoch 415, bias, value: tensor([-0.0537, 0.0245, 0.0082, -0.0115, -0.0028, 0.0070, -0.0148, 0.0344, + -0.0231, 0.0116], device='cuda:0'), grad: tensor([-0.0751, 0.0160, 0.0384, 0.0028, -0.0116, -0.0381, 0.0185, 0.0188, + 0.0428, -0.0125], device='cuda:0') +100 +1e-05 +changing lr +epoch 414, time 225.58, cls_loss 0.4889 cls_loss_mapping 0.0009 cls_loss_causal 0.4330 re_mapping 0.0069 re_causal 0.0182 /// teacc 98.96 lr 0.00001000 +Epoch 416, weight, value: tensor([[-0.2153, -0.0219, 0.0379, ..., -0.0293, -0.0816, -0.1473], + [-0.0640, -0.1128, 0.0351, ..., 0.0571, -0.0289, -0.0942], + [-0.0657, -0.1007, -0.1199, ..., 0.0435, -0.0389, -0.1168], + ..., + [-0.0703, 0.0328, 0.0254, ..., 0.0159, -0.0593, -0.0998], + [-0.0863, -0.0059, 0.0163, ..., 0.0621, -0.0350, -0.1913], + [ 0.0648, 0.0714, -0.0523, ..., -0.1162, -0.0156, 0.1003]], + device='cuda:0'), grad: tensor([[-1.5192e-03, 5.7335e-03, -3.3970e-03, ..., -9.7961e-03, + -5.5850e-05, -3.3264e-03], + [ 1.1623e-06, 6.1607e-04, -1.0217e-06, ..., -2.8114e-03, + 3.2037e-05, 1.3185e-04], + [ 1.2793e-05, 2.4014e-03, 2.4512e-05, ..., 2.4891e-03, + 7.3910e-05, 8.6784e-05], + ..., + [ 3.4440e-06, 7.9203e-04, 1.0230e-05, ..., 1.8902e-03, + 6.6578e-05, 6.8665e-05], + [ 9.3102e-05, 2.2049e-03, 1.6427e-04, ..., 2.6474e-03, + 4.4632e-04, 5.7936e-04], + [ 5.9485e-05, 4.8518e-04, 1.4460e-04, ..., 1.1015e-03, + 1.0085e-04, 2.0123e-04]], device='cuda:0') +Epoch 416, bias, value: tensor([-0.0536, 0.0244, 0.0080, -0.0114, -0.0027, 0.0070, -0.0148, 0.0345, + -0.0231, 0.0115], device='cuda:0'), grad: tensor([-0.0414, -0.0211, 0.0151, -0.0047, -0.0079, -0.0026, 0.0228, 0.0120, + 0.0187, 0.0091], device='cuda:0') +100 +1e-05 +changing lr +epoch 415, time 225.34, cls_loss 0.4976 cls_loss_mapping 0.0010 cls_loss_causal 0.4237 re_mapping 0.0068 re_causal 0.0176 /// teacc 98.93 lr 0.00001000 +Epoch 417, weight, value: tensor([[-0.2155, -0.0217, 0.0377, ..., -0.0294, -0.0816, -0.1474], + [-0.0640, -0.1127, 0.0354, ..., 0.0573, -0.0289, -0.0942], + [-0.0656, -0.1008, -0.1199, ..., 0.0434, -0.0388, -0.1168], + ..., + [-0.0703, 0.0328, 0.0253, ..., 0.0159, -0.0594, -0.0997], + [-0.0865, -0.0058, 0.0163, ..., 0.0620, -0.0351, -0.1913], + [ 0.0648, 0.0711, -0.0523, ..., -0.1164, -0.0154, 0.1002]], + device='cuda:0'), grad: tensor([[ 3.3531e-03, 6.2656e-04, 2.8515e-03, ..., 6.8741e-03, + 7.1302e-06, 1.8206e-03], + [-4.2725e-03, 5.9357e-03, -4.0627e-03, ..., -4.7207e-04, + 1.1957e-04, 2.3746e-03], + [ 1.9014e-04, 3.9940e-03, 6.7091e-04, ..., -6.4468e-04, + 7.2122e-05, 1.7662e-03], + ..., + [ 2.3222e-04, -1.0979e-02, -3.5973e-03, ..., -4.7302e-03, + -2.4009e-04, -4.1046e-03], + [ 5.2643e-04, -1.4448e-03, 4.1270e-04, ..., -3.7937e-03, + 2.3425e-04, -1.0986e-03], + [ 2.5916e-04, 3.3951e-03, 4.7779e-04, ..., -4.1366e-04, + 2.7940e-05, 1.1187e-03]], device='cuda:0') +Epoch 417, bias, value: tensor([-0.0537, 0.0246, 0.0080, -0.0113, -0.0028, 0.0069, -0.0148, 0.0346, + -0.0231, 0.0114], device='cuda:0'), grad: tensor([ 0.0349, 0.0205, -0.0043, 0.0053, 0.0169, -0.0312, -0.0032, -0.0238, + -0.0098, -0.0054], device='cuda:0') +100 +1e-05 +changing lr +epoch 416, time 225.93, cls_loss 0.5019 cls_loss_mapping 0.0009 cls_loss_causal 0.4466 re_mapping 0.0068 re_causal 0.0180 /// teacc 98.95 lr 0.00001000 +Epoch 418, weight, value: tensor([[-0.2156, -0.0216, 0.0378, ..., -0.0296, -0.0817, -0.1473], + [-0.0640, -0.1126, 0.0354, ..., 0.0574, -0.0288, -0.0942], + [-0.0655, -0.1010, -0.1200, ..., 0.0434, -0.0388, -0.1169], + ..., + [-0.0703, 0.0330, 0.0254, ..., 0.0160, -0.0593, -0.0996], + [-0.0866, -0.0059, 0.0164, ..., 0.0620, -0.0352, -0.1913], + [ 0.0649, 0.0710, -0.0523, ..., -0.1163, -0.0154, 0.1003]], + device='cuda:0'), grad: tensor([[ 3.3736e-05, 4.9019e-04, 9.7036e-04, ..., 9.4318e-04, + 1.6391e-05, 2.1911e-04], + [ 6.1798e-04, -3.1662e-03, 1.5249e-03, ..., -1.8349e-03, + 1.0133e-05, 1.9360e-04], + [ 9.4533e-05, -3.8128e-03, 1.4820e-03, ..., -3.6449e-03, + -1.1492e-03, -4.2343e-03], + ..., + [-4.5896e-04, 5.3310e-04, 4.2486e-04, ..., 2.3060e-03, + 4.2534e-04, 5.8174e-04], + [ 1.0598e-04, 1.7571e-04, -4.9973e-03, ..., -5.8365e-03, + 5.4449e-05, 3.8171e-04], + [-7.9441e-04, 6.6423e-04, 1.2455e-03, ..., -2.6264e-03, + -2.0707e-04, -2.3890e-04]], device='cuda:0') +Epoch 418, bias, value: tensor([-0.0538, 0.0246, 0.0079, -0.0114, -0.0027, 0.0070, -0.0149, 0.0347, + -0.0231, 0.0114], device='cuda:0'), grad: tensor([-0.0181, -0.0071, -0.0089, 0.0071, 0.0059, 0.0177, 0.0153, 0.0117, + -0.0055, -0.0179], device='cuda:0') +100 +1e-05 +changing lr +epoch 417, time 226.81, cls_loss 0.5305 cls_loss_mapping 0.0009 cls_loss_causal 0.4673 re_mapping 0.0070 re_causal 0.0184 /// teacc 98.95 lr 0.00001000 +Epoch 419, weight, value: tensor([[-0.2156, -0.0217, 0.0378, ..., -0.0296, -0.0817, -0.1473], + [-0.0638, -0.1126, 0.0352, ..., 0.0573, -0.0289, -0.0942], + [-0.0657, -0.1010, -0.1202, ..., 0.0432, -0.0388, -0.1170], + ..., + [-0.0703, 0.0332, 0.0255, ..., 0.0160, -0.0593, -0.0996], + [-0.0866, -0.0060, 0.0165, ..., 0.0621, -0.0351, -0.1913], + [ 0.0649, 0.0708, -0.0522, ..., -0.1164, -0.0155, 0.1003]], + device='cuda:0'), grad: tensor([[ 6.4354e-07, -5.8599e-06, 1.5295e-04, ..., 1.6127e-03, + 0.0000e+00, -1.1243e-05], + [ 2.1458e-06, -1.2798e-03, 9.6321e-05, ..., 2.0905e-03, + 0.0000e+00, 4.0373e-07], + [ 6.0117e-07, 6.1607e-04, 2.7657e-04, ..., -5.5008e-03, + 0.0000e+00, 4.4927e-06], + ..., + [ 4.4703e-08, 6.4325e-04, 3.5834e-04, ..., 2.6207e-03, + 0.0000e+00, 7.4446e-05], + [-5.6535e-05, 3.4189e-04, 1.8299e-04, ..., 1.7481e-03, + 0.0000e+00, 2.2590e-05], + [ 1.0505e-06, 2.8381e-03, -1.6289e-03, ..., -5.1918e-03, + 0.0000e+00, -9.6560e-05]], device='cuda:0') +Epoch 419, bias, value: tensor([-0.0539, 0.0245, 0.0078, -0.0114, -0.0025, 0.0069, -0.0149, 0.0347, + -0.0229, 0.0113], device='cuda:0'), grad: tensor([ 0.0116, 0.0077, -0.0435, 0.0341, 0.0080, 0.0163, -0.0179, 0.0173, + -0.0178, -0.0159], device='cuda:0') +100 +1e-05 +changing lr +epoch 418, time 226.05, cls_loss 0.5017 cls_loss_mapping 0.0011 cls_loss_causal 0.4343 re_mapping 0.0067 re_causal 0.0174 /// teacc 98.97 lr 0.00001000 +Epoch 420, weight, value: tensor([[-0.2157, -0.0217, 0.0377, ..., -0.0296, -0.0817, -0.1473], + [-0.0638, -0.1125, 0.0350, ..., 0.0573, -0.0289, -0.0942], + [-0.0658, -0.1010, -0.1202, ..., 0.0432, -0.0389, -0.1170], + ..., + [-0.0702, 0.0332, 0.0255, ..., 0.0160, -0.0593, -0.0996], + [-0.0867, -0.0059, 0.0164, ..., 0.0621, -0.0351, -0.1913], + [ 0.0650, 0.0707, -0.0521, ..., -0.1163, -0.0155, 0.1004]], + device='cuda:0'), grad: tensor([[ 4.3273e-05, -8.7662e-03, 7.7903e-05, ..., -6.7091e-04, + 8.0156e-04, 3.4392e-05], + [ 1.2405e-06, 1.9097e-04, 6.4254e-05, ..., 1.1368e-03, + 9.5272e-04, 3.5949e-07], + [ 1.0237e-05, 3.2616e-04, 6.7413e-05, ..., -2.8324e-03, + 9.1600e-04, 2.8405e-06], + ..., + [ 1.6146e-03, 3.5801e-03, 1.1873e-04, ..., 2.0561e-03, + 1.6260e-03, 3.6240e-04], + [ 2.1782e-03, 7.4911e-04, 1.1140e-04, ..., 2.2659e-03, + 1.3914e-03, 5.0497e-04], + [ 6.5684e-05, 3.1147e-03, 2.8014e-04, ..., 2.0390e-03, + 3.7785e-03, 1.4007e-05]], device='cuda:0') +Epoch 420, bias, value: tensor([-0.0538, 0.0245, 0.0078, -0.0114, -0.0024, 0.0068, -0.0151, 0.0347, + -0.0229, 0.0114], device='cuda:0'), grad: tensor([-0.0375, -0.0072, -0.0131, -0.0081, -0.0395, -0.0126, 0.0204, 0.0357, + 0.0269, 0.0350], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 419---------------------------------------------------- +epoch 419, time 225.51, cls_loss 0.4652 cls_loss_mapping 0.0009 cls_loss_causal 0.4090 re_mapping 0.0069 re_causal 0.0177 /// teacc 99.02 lr 0.00001000 +Epoch 421, weight, value: tensor([[-0.2157, -0.0217, 0.0377, ..., -0.0295, -0.0818, -0.1473], + [-0.0638, -0.1125, 0.0351, ..., 0.0571, -0.0289, -0.0943], + [-0.0660, -0.1010, -0.1202, ..., 0.0433, -0.0388, -0.1171], + ..., + [-0.0703, 0.0332, 0.0254, ..., 0.0159, -0.0594, -0.0996], + [-0.0866, -0.0059, 0.0164, ..., 0.0621, -0.0351, -0.1914], + [ 0.0650, 0.0707, -0.0521, ..., -0.1164, -0.0155, 0.1005]], + device='cuda:0'), grad: tensor([[ 3.9101e-04, 1.3053e-04, 3.2902e-04, ..., 1.3657e-03, + 1.8686e-05, 2.6846e-04], + [ 1.5950e-04, 2.5094e-05, 1.0145e-04, ..., 1.2045e-03, + 5.1588e-05, 9.1672e-05], + [-4.1733e-03, -1.0757e-03, -3.6888e-03, ..., -1.2283e-03, + 2.9907e-05, 8.4400e-04], + ..., + [ 1.5039e-03, 2.4929e-03, 1.1511e-03, ..., -5.2223e-03, + 3.1948e-04, 1.5059e-03], + [ 1.1981e-04, 4.3130e-04, 8.2612e-05, ..., 1.8492e-03, + 1.5423e-05, 8.5735e-04], + [-1.5297e-03, 3.3140e-04, -9.3842e-04, ..., 4.9686e-04, + -6.2513e-04, -4.8399e-04]], device='cuda:0') +Epoch 421, bias, value: tensor([-0.0537, 0.0244, 0.0078, -0.0113, -0.0023, 0.0068, -0.0150, 0.0347, + -0.0229, 0.0114], device='cuda:0'), grad: tensor([ 0.0079, 0.0063, -0.0077, 0.0169, 0.0141, -0.0202, 0.0051, -0.0375, + 0.0099, 0.0052], device='cuda:0') +100 +1e-05 +changing lr +epoch 420, time 228.01, cls_loss 0.5011 cls_loss_mapping 0.0009 cls_loss_causal 0.4315 re_mapping 0.0066 re_causal 0.0179 /// teacc 99.02 lr 0.00001000 +Epoch 422, weight, value: tensor([[-0.2156, -0.0216, 0.0379, ..., -0.0294, -0.0818, -0.1474], + [-0.0638, -0.1126, 0.0350, ..., 0.0572, -0.0289, -0.0944], + [-0.0659, -0.1010, -0.1201, ..., 0.0433, -0.0386, -0.1169], + ..., + [-0.0703, 0.0334, 0.0255, ..., 0.0158, -0.0592, -0.0995], + [-0.0867, -0.0059, 0.0162, ..., 0.0620, -0.0351, -0.1916], + [ 0.0651, 0.0706, -0.0520, ..., -0.1163, -0.0154, 0.1005]], + device='cuda:0'), grad: tensor([[ 7.4916e-06, 4.5991e-04, 3.7737e-06, ..., 8.6451e-04, + 9.4831e-05, 2.9206e-06], + [ 1.4836e-06, -3.0136e-03, -2.0146e-05, ..., -8.1406e-03, + 8.7470e-06, -1.4133e-03], + [-1.5688e-04, 1.3161e-03, -6.4850e-05, ..., 6.4011e-03, + -3.1948e-04, 2.4354e-04], + ..., + [ 8.2552e-05, -2.1782e-03, 5.0128e-05, ..., -1.3618e-03, + 5.0545e-04, 5.3972e-05], + [ 5.0157e-05, 1.9222e-05, 5.9679e-06, ..., 1.4706e-03, + 2.1684e-04, 9.2089e-05], + [-3.5405e-05, -2.1458e-03, 3.9153e-06, ..., -3.2463e-03, + -6.2752e-04, 1.9145e-04]], device='cuda:0') +Epoch 422, bias, value: tensor([-0.0537, 0.0245, 0.0079, -0.0115, -0.0024, 0.0068, -0.0151, 0.0348, + -0.0230, 0.0114], device='cuda:0'), grad: tensor([ 0.0075, -0.0233, 0.0311, -0.0112, 0.0117, 0.0079, 0.0065, -0.0176, + -0.0172, 0.0048], device='cuda:0') +100 +1e-05 +changing lr +epoch 421, time 226.47, cls_loss 0.5107 cls_loss_mapping 0.0011 cls_loss_causal 0.4435 re_mapping 0.0067 re_causal 0.0176 /// teacc 98.96 lr 0.00001000 +Epoch 423, weight, value: tensor([[-0.2155, -0.0216, 0.0378, ..., -0.0293, -0.0819, -0.1472], + [-0.0638, -0.1125, 0.0348, ..., 0.0570, -0.0291, -0.0943], + [-0.0660, -0.1011, -0.1202, ..., 0.0432, -0.0387, -0.1170], + ..., + [-0.0702, 0.0335, 0.0256, ..., 0.0159, -0.0592, -0.0994], + [-0.0867, -0.0058, 0.0163, ..., 0.0621, -0.0351, -0.1916], + [ 0.0651, 0.0704, -0.0520, ..., -0.1162, -0.0153, 0.1003]], + device='cuda:0'), grad: tensor([[ 6.7174e-05, 1.0624e-03, 4.0674e-04, ..., 2.0485e-03, + 7.3850e-05, 2.3305e-04], + [-1.1873e-03, -2.5444e-03, -3.0088e-04, ..., -2.1362e-03, + 4.0698e-04, 5.1737e-05], + [ 1.0639e-04, 1.6966e-03, 1.1480e-04, ..., 2.1515e-03, + 2.3258e-04, 1.0794e-04], + ..., + [ 1.0842e-04, -3.2806e-03, -3.4599e-03, ..., -8.7814e-03, + 1.2159e-04, -2.1667e-03], + [ 1.4353e-04, -6.5384e-03, 8.2588e-04, ..., 2.0390e-03, + 8.0156e-04, 4.7922e-04], + [-2.5973e-05, -1.3504e-03, 8.8310e-04, ..., 3.9749e-03, + 3.3307e-04, 4.1842e-04]], device='cuda:0') +Epoch 423, bias, value: tensor([-0.0537, 0.0245, 0.0078, -0.0115, -0.0025, 0.0068, -0.0151, 0.0348, + -0.0229, 0.0114], device='cuda:0'), grad: tensor([ 0.0152, -0.0070, 0.0164, 0.0248, -0.0041, 0.0168, -0.0098, -0.0845, + 0.0120, 0.0202], device='cuda:0') +100 +1e-05 +changing lr +epoch 422, time 227.75, cls_loss 0.4815 cls_loss_mapping 0.0011 cls_loss_causal 0.4223 re_mapping 0.0064 re_causal 0.0169 /// teacc 98.94 lr 0.00001000 +Epoch 424, weight, value: tensor([[-0.2155, -0.0217, 0.0378, ..., -0.0293, -0.0820, -0.1472], + [-0.0638, -0.1124, 0.0349, ..., 0.0572, -0.0290, -0.0944], + [-0.0660, -0.1011, -0.1203, ..., 0.0432, -0.0388, -0.1170], + ..., + [-0.0700, 0.0337, 0.0256, ..., 0.0159, -0.0593, -0.0994], + [-0.0868, -0.0058, 0.0165, ..., 0.0620, -0.0352, -0.1916], + [ 0.0650, 0.0704, -0.0521, ..., -0.1161, -0.0153, 0.1004]], + device='cuda:0'), grad: tensor([[ 2.7227e-04, 2.6989e-04, 6.6519e-04, ..., -3.0079e-03, + -1.4515e-03, 5.2363e-05], + [-5.7667e-05, 5.4359e-04, 9.9564e-04, ..., 2.6245e-03, + 1.4973e-03, 5.9307e-06], + [ 1.9646e-04, 1.2946e-04, -5.9853e-03, ..., -5.2071e-03, + -4.2152e-03, 7.6890e-05], + ..., + [ 1.3113e-04, 5.0354e-04, 9.0122e-04, ..., 2.6703e-03, + -5.3787e-04, 2.4870e-05], + [ 3.0518e-03, 3.1352e-04, 1.9302e-03, ..., 2.8801e-03, + 1.1082e-03, 4.5547e-03], + [-1.0099e-03, 3.2425e-04, 4.0269e-04, ..., 1.5688e-03, + 7.2956e-04, -1.6963e-04]], device='cuda:0') +Epoch 424, bias, value: tensor([-0.0538, 0.0246, 0.0078, -0.0115, -0.0024, 0.0069, -0.0151, 0.0348, + -0.0229, 0.0114], device='cuda:0'), grad: tensor([-0.0163, 0.0251, -0.0429, 0.0021, -0.0038, 0.0223, -0.0271, -0.0052, + 0.0333, 0.0124], device='cuda:0') +100 +1e-05 +changing lr +epoch 423, time 229.97, cls_loss 0.4705 cls_loss_mapping 0.0009 cls_loss_causal 0.4149 re_mapping 0.0066 re_causal 0.0170 /// teacc 98.93 lr 0.00001000 +Epoch 425, weight, value: tensor([[-0.2155, -0.0215, 0.0378, ..., -0.0294, -0.0818, -0.1471], + [-0.0637, -0.1125, 0.0348, ..., 0.0572, -0.0290, -0.0944], + [-0.0660, -0.1010, -0.1203, ..., 0.0432, -0.0388, -0.1170], + ..., + [-0.0700, 0.0337, 0.0257, ..., 0.0161, -0.0593, -0.0994], + [-0.0868, -0.0058, 0.0167, ..., 0.0620, -0.0352, -0.1914], + [ 0.0651, 0.0703, -0.0522, ..., -0.1161, -0.0153, 0.1004]], + device='cuda:0'), grad: tensor([[ 7.1955e-04, 9.2793e-04, 4.0932e-03, ..., 5.3902e-03, + 4.4136e-03, 5.1290e-05], + [ 2.4223e-04, 2.0466e-03, 1.3800e-03, ..., 1.1368e-02, + 3.5439e-03, 5.1886e-05], + [-3.1090e-03, 7.0953e-04, 1.5068e-03, ..., -6.0940e-04, + -4.6310e-03, 1.0424e-03], + ..., + [-2.2519e-04, 9.6369e-04, 2.6779e-03, ..., 5.4703e-03, + 8.7128e-03, 4.5419e-05], + [ 1.9398e-03, 8.2922e-04, 3.0136e-03, ..., 4.7455e-03, + -1.2039e-02, 8.9884e-05], + [ 1.9073e-03, -8.4019e-04, 3.5934e-03, ..., 7.5073e-03, + 1.2253e-02, 1.0777e-04]], device='cuda:0') +Epoch 425, bias, value: tensor([-0.0537, 0.0246, 0.0079, -0.0116, -0.0024, 0.0069, -0.0152, 0.0347, + -0.0228, 0.0114], device='cuda:0'), grad: tensor([ 0.0252, 0.0309, 0.0039, 0.0282, -0.0210, -0.0311, -0.0446, 0.0038, + -0.0269, 0.0316], device='cuda:0') +100 +1e-05 +changing lr +epoch 424, time 226.96, cls_loss 0.4658 cls_loss_mapping 0.0008 cls_loss_causal 0.4074 re_mapping 0.0067 re_causal 0.0170 /// teacc 98.91 lr 0.00001000 +Epoch 426, weight, value: tensor([[-0.2155, -0.0215, 0.0378, ..., -0.0294, -0.0818, -0.1472], + [-0.0638, -0.1124, 0.0347, ..., 0.0572, -0.0292, -0.0944], + [-0.0661, -0.1009, -0.1205, ..., 0.0432, -0.0389, -0.1170], + ..., + [-0.0700, 0.0335, 0.0255, ..., 0.0160, -0.0593, -0.0994], + [-0.0869, -0.0059, 0.0166, ..., 0.0619, -0.0352, -0.1916], + [ 0.0653, 0.0704, -0.0519, ..., -0.1160, -0.0153, 0.1004]], + device='cuda:0'), grad: tensor([[ 1.0459e-06, 4.9800e-05, 1.1778e-04, ..., 1.3332e-03, + 5.8681e-05, 1.3597e-05], + [ 8.2478e-06, 1.2350e-04, 1.6689e-03, ..., 4.3678e-03, + 6.2704e-04, 2.5257e-05], + [ 2.2762e-06, 7.5006e-04, 1.6499e-03, ..., 5.3749e-03, + 7.4911e-04, 7.0632e-05], + ..., + [ 2.0072e-05, 1.1070e-02, 1.6747e-03, ..., 1.9501e-02, + 2.8954e-03, 9.8884e-05], + [ 1.0580e-05, 9.6703e-04, 5.0926e-04, ..., -5.4016e-03, + 3.1948e-04, 1.7178e-04], + [-8.3447e-05, 1.8253e-03, 3.4356e-04, ..., 5.9929e-03, + 4.2677e-04, 1.3936e-04]], device='cuda:0') +Epoch 426, bias, value: tensor([-0.0537, 0.0245, 0.0079, -0.0115, -0.0024, 0.0068, -0.0152, 0.0347, + -0.0229, 0.0115], device='cuda:0'), grad: tensor([ 0.0058, 0.0179, 0.0213, 0.0092, -0.0613, -0.0144, 0.0085, 0.0003, + -0.0006, 0.0134], device='cuda:0') +100 +1e-05 +changing lr +epoch 425, time 226.69, cls_loss 0.4713 cls_loss_mapping 0.0010 cls_loss_causal 0.4064 re_mapping 0.0064 re_causal 0.0167 /// teacc 98.93 lr 0.00001000 +Epoch 427, weight, value: tensor([[-0.2155, -0.0213, 0.0380, ..., -0.0292, -0.0816, -0.1473], + [-0.0639, -0.1123, 0.0348, ..., 0.0572, -0.0291, -0.0944], + [-0.0660, -0.1010, -0.1204, ..., 0.0431, -0.0390, -0.1171], + ..., + [-0.0699, 0.0335, 0.0256, ..., 0.0159, -0.0591, -0.0994], + [-0.0870, -0.0058, 0.0165, ..., 0.0621, -0.0353, -0.1917], + [ 0.0654, 0.0703, -0.0520, ..., -0.1161, -0.0154, 0.1004]], + device='cuda:0'), grad: tensor([[-1.4031e-04, 4.7296e-05, 2.9278e-03, ..., -8.4019e-04, + 3.8862e-04, -2.1130e-05], + [ 2.8521e-05, 3.9428e-05, -1.1452e-02, ..., -9.3994e-03, + -6.7425e-04, 4.3362e-06], + [-9.0885e-04, 1.4381e-03, -2.6122e-05, ..., -4.3144e-03, + 3.3355e-04, 6.5975e-06], + ..., + [ 1.7390e-05, 1.0471e-03, 5.3596e-03, ..., 5.3673e-03, + 5.0049e-03, 8.6948e-06], + [ 9.3889e-04, 6.0940e-04, -6.3744e-03, ..., -3.6869e-03, + -9.1400e-03, 3.6061e-05], + [ 7.3969e-05, 2.4843e-04, 1.7519e-03, ..., 2.3575e-03, + 7.8201e-04, 3.3174e-06]], device='cuda:0') +Epoch 427, bias, value: tensor([-0.0535, 0.0246, 0.0078, -0.0116, -0.0024, 0.0069, -0.0153, 0.0346, + -0.0228, 0.0114], device='cuda:0'), grad: tensor([-0.0134, -0.0782, -0.0222, 0.0057, 0.0294, 0.0185, 0.0163, 0.0276, + -0.0007, 0.0171], device='cuda:0') +100 +1e-05 +changing lr +epoch 426, time 226.18, cls_loss 0.5030 cls_loss_mapping 0.0009 cls_loss_causal 0.4444 re_mapping 0.0066 re_causal 0.0178 /// teacc 98.93 lr 0.00001000 +Epoch 428, weight, value: tensor([[-0.2151, -0.0215, 0.0381, ..., -0.0291, -0.0816, -0.1473], + [-0.0639, -0.1123, 0.0346, ..., 0.0570, -0.0292, -0.0944], + [-0.0661, -0.1012, -0.1203, ..., 0.0431, -0.0389, -0.1171], + ..., + [-0.0699, 0.0334, 0.0255, ..., 0.0158, -0.0593, -0.0995], + [-0.0871, -0.0058, 0.0166, ..., 0.0621, -0.0353, -0.1916], + [ 0.0652, 0.0704, -0.0521, ..., -0.1161, -0.0155, 0.1004]], + device='cuda:0'), grad: tensor([[ 5.2422e-05, 5.1689e-04, 2.9597e-06, ..., 2.8858e-03, + 6.4850e-05, 1.7043e-07], + [ 7.1883e-05, 1.1454e-03, 3.4329e-06, ..., 3.9291e-03, + 3.3528e-07, 1.6950e-07], + [ 1.4534e-03, -2.3117e-03, 9.1717e-06, ..., -6.1340e-03, + 6.8605e-05, 6.1654e-07], + ..., + [ 2.1502e-05, -2.8133e-04, -1.1581e-02, ..., 2.3899e-03, + 1.8448e-05, -2.4223e-04], + [ 2.1279e-05, 7.3242e-04, 1.0364e-05, ..., 2.9278e-03, + 3.6774e-02, 2.9393e-06], + [-2.7064e-06, -4.8920e-02, 7.0047e-04, ..., -7.3624e-03, + 1.3754e-05, 1.2126e-06]], device='cuda:0') +Epoch 428, bias, value: tensor([-0.0534, 0.0244, 0.0079, -0.0116, -0.0024, 0.0069, -0.0153, 0.0346, + -0.0227, 0.0114], device='cuda:0'), grad: tensor([ 0.0192, 0.0240, -0.0288, -0.0461, 0.0646, 0.0126, -0.0148, -0.0011, + 0.0482, -0.0778], device='cuda:0') +100 +1e-05 +changing lr +epoch 427, time 226.51, cls_loss 0.5136 cls_loss_mapping 0.0009 cls_loss_causal 0.4579 re_mapping 0.0067 re_causal 0.0176 /// teacc 98.96 lr 0.00001000 +Epoch 429, weight, value: tensor([[-0.2152, -0.0216, 0.0381, ..., -0.0293, -0.0817, -0.1472], + [-0.0637, -0.1123, 0.0347, ..., 0.0570, -0.0291, -0.0944], + [-0.0663, -0.1011, -0.1202, ..., 0.0431, -0.0388, -0.1170], + ..., + [-0.0700, 0.0335, 0.0256, ..., 0.0157, -0.0593, -0.0996], + [-0.0872, -0.0058, 0.0165, ..., 0.0620, -0.0354, -0.1917], + [ 0.0654, 0.0704, -0.0521, ..., -0.1160, -0.0155, 0.1006]], + device='cuda:0'), grad: tensor([[ 4.9360e-06, -2.7823e-04, 1.4267e-03, ..., 5.4512e-03, + 4.3368e-04, 7.3969e-05], + [ 9.1270e-08, 2.5425e-03, -2.7180e-04, ..., 4.7531e-03, + 5.3596e-04, 8.2627e-06], + [ 1.0114e-06, -4.6349e-03, -2.6531e-03, ..., -4.4632e-03, + 1.3481e-02, -7.0572e-03], + ..., + [ 2.9430e-06, 9.1219e-04, 1.5962e-04, ..., -4.6873e-04, + 8.5402e-04, 1.1340e-05], + [-4.1008e-05, 1.9836e-03, -3.5343e-03, ..., -4.2877e-03, + 1.1320e-03, 2.5272e-04], + [ 5.0873e-05, -2.0618e-03, 9.8765e-05, ..., -1.9951e-03, + -2.8973e-03, 1.1474e-05]], device='cuda:0') +Epoch 429, bias, value: tensor([-0.0536, 0.0245, 0.0079, -0.0116, -0.0024, 0.0069, -0.0152, 0.0346, + -0.0227, 0.0115], device='cuda:0'), grad: tensor([ 0.0289, 0.0377, -0.0338, -0.0086, 0.0198, -0.0068, -0.0009, -0.0049, + 0.0050, -0.0365], device='cuda:0') +100 +1e-05 +changing lr +epoch 428, time 226.49, cls_loss 0.4984 cls_loss_mapping 0.0010 cls_loss_causal 0.4396 re_mapping 0.0065 re_causal 0.0173 /// teacc 98.96 lr 0.00001000 +Epoch 430, weight, value: tensor([[-0.2151, -0.0217, 0.0381, ..., -0.0294, -0.0816, -0.1473], + [-0.0637, -0.1123, 0.0346, ..., 0.0571, -0.0291, -0.0945], + [-0.0665, -0.1011, -0.1202, ..., 0.0431, -0.0389, -0.1171], + ..., + [-0.0699, 0.0335, 0.0258, ..., 0.0157, -0.0592, -0.0996], + [-0.0873, -0.0057, 0.0165, ..., 0.0620, -0.0353, -0.1918], + [ 0.0653, 0.0704, -0.0520, ..., -0.1160, -0.0154, 0.1006]], + device='cuda:0'), grad: tensor([[ 4.1217e-05, 3.6740e-04, 2.0733e-03, ..., 3.1128e-03, + 5.5313e-04, 4.0047e-07], + [ 9.6977e-05, -2.8687e-03, -2.4204e-03, ..., -9.9106e-03, + -3.1033e-03, 2.6077e-07], + [ 2.1353e-05, -3.8338e-03, 2.6631e-04, ..., -1.0414e-03, + 3.3379e-04, 2.2876e-04], + ..., + [-2.8387e-06, 3.3073e-03, 4.8065e-04, ..., 8.6746e-03, + 4.6897e-04, 2.6658e-05], + [-1.0490e-03, 8.0061e-04, -5.1880e-04, ..., 2.3098e-03, + -2.4772e-04, 3.4428e-04], + [ 1.0353e-04, 1.3924e-04, 3.9029e-04, ..., 3.7460e-03, + 3.9268e-04, -1.8105e-05]], device='cuda:0') +Epoch 430, bias, value: tensor([-0.0537, 0.0244, 0.0079, -0.0115, -0.0026, 0.0069, -0.0152, 0.0348, + -0.0227, 0.0115], device='cuda:0'), grad: tensor([ 0.0430, -0.0686, 0.0054, -0.0171, -0.0168, -0.0190, 0.0156, 0.0301, + 0.0122, 0.0152], device='cuda:0') +100 +1e-05 +changing lr +epoch 429, time 228.51, cls_loss 0.5044 cls_loss_mapping 0.0009 cls_loss_causal 0.4502 re_mapping 0.0065 re_causal 0.0172 /// teacc 98.96 lr 0.00001000 +Epoch 431, weight, value: tensor([[-0.2151, -0.0218, 0.0380, ..., -0.0295, -0.0817, -0.1474], + [-0.0638, -0.1122, 0.0345, ..., 0.0570, -0.0292, -0.0946], + [-0.0665, -0.1012, -0.1201, ..., 0.0430, -0.0389, -0.1169], + ..., + [-0.0699, 0.0337, 0.0259, ..., 0.0157, -0.0593, -0.0995], + [-0.0874, -0.0058, 0.0165, ..., 0.0619, -0.0353, -0.1916], + [ 0.0651, 0.0704, -0.0520, ..., -0.1160, -0.0151, 0.1005]], + device='cuda:0'), grad: tensor([[ 4.6706e-04, -2.0275e-03, -1.5795e-04, ..., 3.6526e-04, + -8.2922e-04, 7.3290e-04], + [-7.3969e-05, 6.3801e-04, -1.1301e-04, ..., 4.2057e-04, + 2.0131e-05, 1.0036e-05], + [ 3.3140e-05, 1.1247e-04, 2.0146e-05, ..., -4.7836e-03, + -5.1117e-03, 6.2943e-05], + ..., + [-4.5508e-05, -1.2362e-04, -4.4703e-05, ..., -1.0509e-03, + 4.1842e-05, 1.4275e-05], + [ 1.4114e-04, -1.0526e-04, -1.5962e-04, ..., -1.9908e-04, + 9.3269e-04, 7.8559e-05], + [-2.0683e-05, 3.6168e-04, 5.6505e-05, ..., 5.6076e-04, + 1.9979e-04, -1.7524e-05]], device='cuda:0') +Epoch 431, bias, value: tensor([-0.0538, 0.0244, 0.0079, -0.0115, -0.0025, 0.0068, -0.0151, 0.0349, + -0.0227, 0.0115], device='cuda:0'), grad: tensor([-9.9564e-04, 7.3700e-03, -2.6550e-02, 1.3649e-02, 4.6501e-03, + -1.3275e-02, 1.4923e-02, -5.3596e-03, 3.1859e-05, 5.5618e-03], + device='cuda:0') +100 +1e-05 +changing lr +epoch 430, time 226.01, cls_loss 0.5025 cls_loss_mapping 0.0011 cls_loss_causal 0.4339 re_mapping 0.0064 re_causal 0.0167 /// teacc 99.00 lr 0.00001000 +Epoch 432, weight, value: tensor([[-0.2151, -0.0217, 0.0379, ..., -0.0294, -0.0817, -0.1473], + [-0.0638, -0.1121, 0.0345, ..., 0.0571, -0.0291, -0.0946], + [-0.0664, -0.1012, -0.1200, ..., 0.0432, -0.0390, -0.1169], + ..., + [-0.0699, 0.0337, 0.0258, ..., 0.0155, -0.0594, -0.0995], + [-0.0873, -0.0059, 0.0166, ..., 0.0620, -0.0353, -0.1916], + [ 0.0650, 0.0702, -0.0521, ..., -0.1161, -0.0150, 0.1004]], + device='cuda:0'), grad: tensor([[ 7.1049e-04, 2.2333e-06, 9.0981e-04, ..., 1.3638e-03, + 2.8968e-04, 1.2183e-04], + [ 2.0373e-04, 1.3661e-04, 1.4293e-04, ..., 1.5316e-03, + 9.3460e-05, 6.5684e-05], + [ 4.9639e-04, 1.3580e-02, 5.3978e-04, ..., 1.4477e-03, + 5.1727e-03, 1.6975e-04], + ..., + [-7.4310e-03, -1.6983e-02, -1.1063e-02, ..., -9.2316e-03, + -1.1063e-02, 6.2108e-05], + [ 1.2360e-03, 1.7338e-03, 1.3618e-03, ..., 2.1286e-03, + 1.1549e-03, 2.3866e-04], + [ 9.2125e-04, 1.3371e-03, 1.3275e-03, ..., -2.6379e-03, + 1.0290e-03, 2.7239e-05]], device='cuda:0') +Epoch 432, bias, value: tensor([-0.0537, 0.0245, 0.0080, -0.0117, -0.0026, 0.0069, -0.0151, 0.0347, + -0.0226, 0.0114], device='cuda:0'), grad: tensor([ 0.0010, 0.0091, 0.0240, 0.0124, 0.0127, 0.0134, -0.0287, -0.0393, + 0.0147, -0.0193], device='cuda:0') +100 +1e-05 +changing lr +epoch 431, time 229.02, cls_loss 0.5129 cls_loss_mapping 0.0010 cls_loss_causal 0.4475 re_mapping 0.0063 re_causal 0.0170 /// teacc 98.97 lr 0.00001000 +Epoch 433, weight, value: tensor([[-0.2151, -0.0216, 0.0378, ..., -0.0295, -0.0818, -0.1473], + [-0.0638, -0.1121, 0.0345, ..., 0.0569, -0.0290, -0.0946], + [-0.0663, -0.1013, -0.1201, ..., 0.0431, -0.0391, -0.1170], + ..., + [-0.0699, 0.0336, 0.0258, ..., 0.0155, -0.0591, -0.0994], + [-0.0874, -0.0060, 0.0165, ..., 0.0619, -0.0353, -0.1917], + [ 0.0650, 0.0703, -0.0522, ..., -0.1160, -0.0151, 0.1003]], + device='cuda:0'), grad: tensor([[ 3.8218e-04, 6.2585e-05, 5.1212e-04, ..., 2.1343e-03, + 4.7708e-04, 1.0902e-04], + [ 4.1771e-04, 1.9274e-03, 7.7724e-05, ..., -5.5885e-04, + 3.2425e-04, 1.5926e-04], + [-2.7752e-03, 3.2973e-04, -1.3008e-03, ..., 1.2369e-03, + 2.9016e-04, -1.6117e-03], + ..., + [ 1.2083e-03, 5.2719e-03, 1.6737e-03, ..., 4.6043e-03, + 2.3155e-03, 2.1768e-04], + [-5.2261e-03, 4.7836e-03, -2.4414e-03, ..., -4.7646e-03, + 2.2125e-03, -8.4352e-04], + [ 5.9967e-03, 3.7140e-02, 2.7008e-03, ..., 4.7760e-03, + 5.1422e-03, 1.1749e-03]], device='cuda:0') +Epoch 433, bias, value: tensor([-0.0538, 0.0244, 0.0079, -0.0115, -0.0025, 0.0071, -0.0150, 0.0346, + -0.0228, 0.0115], device='cuda:0'), grad: tensor([ 0.0109, -0.0014, 0.0014, -0.0205, -0.0303, -0.0106, 0.0182, 0.0231, + -0.0050, 0.0143], device='cuda:0') +100 +1e-05 +changing lr +epoch 432, time 227.29, cls_loss 0.5065 cls_loss_mapping 0.0009 cls_loss_causal 0.4424 re_mapping 0.0064 re_causal 0.0174 /// teacc 98.98 lr 0.00001000 +Epoch 434, weight, value: tensor([[-0.2151, -0.0216, 0.0378, ..., -0.0295, -0.0819, -0.1473], + [-0.0637, -0.1122, 0.0345, ..., 0.0568, -0.0289, -0.0944], + [-0.0664, -0.1013, -0.1202, ..., 0.0432, -0.0389, -0.1171], + ..., + [-0.0699, 0.0337, 0.0258, ..., 0.0156, -0.0591, -0.0995], + [-0.0874, -0.0060, 0.0166, ..., 0.0618, -0.0354, -0.1916], + [ 0.0650, 0.0702, -0.0521, ..., -0.1159, -0.0150, 0.1002]], + device='cuda:0'), grad: tensor([[ 5.9204e-03, 1.4639e-03, 9.0485e-03, ..., -1.1702e-03, + 1.8644e-04, 1.8626e-09], + [ 1.6856e-04, 7.1383e-04, 3.3998e-04, ..., 8.0109e-05, + 1.7440e-04, 0.0000e+00], + [ 1.3113e-04, 1.4591e-03, 3.0041e-04, ..., -1.0582e-02, + 2.2221e-04, 3.7253e-09], + ..., + [ 2.1477e-03, 9.0837e-04, 5.8937e-03, ..., 7.5836e-03, + 4.3983e-03, 3.3528e-07], + [-6.5536e-03, 3.7937e-03, -9.3842e-03, ..., -5.2834e-04, + 9.5224e-04, 5.6811e-06], + [-3.5458e-03, 9.2888e-04, -9.3842e-03, ..., -1.6975e-03, + -7.1831e-03, -6.7577e-06]], device='cuda:0') +Epoch 434, bias, value: tensor([-0.0539, 0.0244, 0.0079, -0.0116, -0.0026, 0.0071, -0.0151, 0.0346, + -0.0228, 0.0117], device='cuda:0'), grad: tensor([ 0.0117, -0.0116, -0.0061, -0.0110, 0.0214, -0.0286, 0.0206, 0.0354, + -0.0305, -0.0012], device='cuda:0') +100 +1e-05 +changing lr +epoch 433, time 228.15, cls_loss 0.4965 cls_loss_mapping 0.0008 cls_loss_causal 0.4369 re_mapping 0.0065 re_causal 0.0173 /// teacc 99.00 lr 0.00001000 +Epoch 435, weight, value: tensor([[-0.2154, -0.0217, 0.0379, ..., -0.0296, -0.0819, -0.1474], + [-0.0637, -0.1122, 0.0344, ..., 0.0568, -0.0286, -0.0943], + [-0.0664, -0.1014, -0.1200, ..., 0.0433, -0.0389, -0.1170], + ..., + [-0.0699, 0.0338, 0.0258, ..., 0.0156, -0.0591, -0.0995], + [-0.0873, -0.0061, 0.0166, ..., 0.0619, -0.0354, -0.1916], + [ 0.0650, 0.0701, -0.0522, ..., -0.1159, -0.0150, 0.1002]], + device='cuda:0'), grad: tensor([[ 2.3931e-05, 7.2746e-03, 6.2132e-04, ..., -1.2121e-03, + 1.7846e-04, 1.6391e-07], + [ 1.2789e-03, 2.0534e-05, 9.6083e-04, ..., 6.6233e-04, + 2.6035e-03, 2.1271e-06], + [ 2.2575e-05, 1.0738e-03, 2.3198e-04, ..., -2.6588e-03, + 2.3270e-04, 1.3858e-06], + ..., + [ 2.9492e-04, 1.3781e-03, 7.8869e-04, ..., -3.2067e-04, + -1.0204e-03, 1.4100e-06], + [ 1.0538e-04, -1.5762e-02, 2.9182e-04, ..., 1.3342e-03, + 2.0218e-04, 1.8343e-05], + [ 2.5749e-04, 1.2817e-03, 9.7275e-04, ..., 2.7962e-03, + 8.4591e-04, -4.7013e-06]], device='cuda:0') +Epoch 435, bias, value: tensor([-0.0540, 0.0244, 0.0079, -0.0117, -0.0026, 0.0073, -0.0152, 0.0347, + -0.0226, 0.0116], device='cuda:0'), grad: tensor([ 0.0055, -0.0018, -0.0139, -0.0081, -0.0067, 0.0182, 0.0203, -0.0160, + -0.0188, 0.0214], device='cuda:0') +100 +1e-05 +changing lr +epoch 434, time 227.53, cls_loss 0.5068 cls_loss_mapping 0.0010 cls_loss_causal 0.4519 re_mapping 0.0064 re_causal 0.0169 /// teacc 98.95 lr 0.00001000 +Epoch 436, weight, value: tensor([[-0.2153, -0.0216, 0.0380, ..., -0.0297, -0.0819, -0.1473], + [-0.0636, -0.1122, 0.0343, ..., 0.0568, -0.0285, -0.0944], + [-0.0664, -0.1013, -0.1199, ..., 0.0433, -0.0390, -0.1171], + ..., + [-0.0698, 0.0339, 0.0259, ..., 0.0155, -0.0593, -0.0995], + [-0.0874, -0.0063, 0.0164, ..., 0.0618, -0.0354, -0.1917], + [ 0.0649, 0.0700, -0.0522, ..., -0.1159, -0.0149, 0.1002]], + device='cuda:0'), grad: tensor([[ 0.0009, 0.0006, -0.0057, ..., 0.0024, 0.0002, 0.0004], + [ 0.0007, 0.0012, 0.0002, ..., 0.0003, 0.0011, 0.0002], + [ 0.0030, 0.0015, 0.0013, ..., 0.0005, 0.0001, 0.0014], + ..., + [ 0.0012, 0.0029, 0.0045, ..., 0.0063, 0.0024, 0.0006], + [-0.0007, -0.0067, 0.0076, ..., -0.0084, 0.0005, 0.0005], + [-0.0027, -0.0014, -0.0075, ..., -0.0065, -0.0053, -0.0016]], + device='cuda:0') +Epoch 436, bias, value: tensor([-0.0541, 0.0245, 0.0078, -0.0116, -0.0026, 0.0073, -0.0150, 0.0346, + -0.0228, 0.0116], device='cuda:0'), grad: tensor([ 0.0035, -0.0104, -0.0103, 0.0228, 0.0245, -0.0027, -0.0201, 0.0308, + -0.0324, -0.0059], device='cuda:0') +100 +1e-05 +changing lr +epoch 435, time 226.56, cls_loss 0.4781 cls_loss_mapping 0.0009 cls_loss_causal 0.4216 re_mapping 0.0063 re_causal 0.0165 /// teacc 98.97 lr 0.00001000 +Epoch 437, weight, value: tensor([[-0.2153, -0.0217, 0.0381, ..., -0.0296, -0.0820, -0.1473], + [-0.0637, -0.1123, 0.0341, ..., 0.0567, -0.0285, -0.0944], + [-0.0662, -0.1014, -0.1198, ..., 0.0434, -0.0390, -0.1171], + ..., + [-0.0698, 0.0340, 0.0259, ..., 0.0156, -0.0590, -0.0995], + [-0.0875, -0.0065, 0.0162, ..., 0.0617, -0.0353, -0.1917], + [ 0.0648, 0.0699, -0.0522, ..., -0.1159, -0.0149, 0.1001]], + device='cuda:0'), grad: tensor([[ 2.2829e-04, 2.4676e-04, 8.1396e-04, ..., 1.8950e-03, + 6.1607e-04, 3.7909e-05], + [ 2.8634e-04, 3.4976e-04, 1.3962e-03, ..., 2.2850e-03, + 9.8324e-04, 4.7714e-05], + [ 4.4060e-04, -1.8187e-03, 1.2474e-03, ..., -2.5482e-03, + -2.7962e-03, 7.3314e-05], + ..., + [ 7.3791e-05, 1.3208e-03, 7.3147e-04, ..., 3.0975e-03, + 4.1237e-03, 1.6704e-05], + [ 1.0753e-04, 2.3222e-04, 2.3861e-03, ..., -2.8858e-03, + -2.3174e-03, 2.1562e-05], + [ 8.7142e-05, 1.2422e-04, 6.3992e-04, ..., 2.4376e-03, + 3.1071e-03, -6.3702e-07]], device='cuda:0') +Epoch 437, bias, value: tensor([-0.0540, 0.0244, 0.0080, -0.0117, -0.0026, 0.0073, -0.0151, 0.0347, + -0.0229, 0.0115], device='cuda:0'), grad: tensor([ 0.0151, 0.0209, -0.0348, -0.0075, 0.0150, -0.0216, 0.0162, 0.0053, + -0.0310, 0.0225], device='cuda:0') +100 +1e-05 +changing lr +epoch 436, time 226.26, cls_loss 0.4742 cls_loss_mapping 0.0007 cls_loss_causal 0.4088 re_mapping 0.0067 re_causal 0.0176 /// teacc 99.00 lr 0.00001000 +Epoch 438, weight, value: tensor([[-0.2153, -0.0218, 0.0383, ..., -0.0295, -0.0820, -0.1473], + [-0.0638, -0.1125, 0.0341, ..., 0.0566, -0.0284, -0.0944], + [-0.0663, -0.1014, -0.1199, ..., 0.0434, -0.0389, -0.1171], + ..., + [-0.0698, 0.0341, 0.0259, ..., 0.0156, -0.0591, -0.0995], + [-0.0874, -0.0064, 0.0160, ..., 0.0618, -0.0353, -0.1918], + [ 0.0649, 0.0698, -0.0520, ..., -0.1158, -0.0149, 0.1001]], + device='cuda:0'), grad: tensor([[ 1.2457e-04, 8.8120e-04, 1.3380e-03, ..., 2.4090e-03, + 4.7798e-03, 1.3515e-05], + [ 1.0490e-04, 3.6764e-04, 1.0371e-04, ..., 1.6556e-03, + 1.2274e-03, 3.2961e-05], + [ 5.7554e-04, 9.9030e-03, 1.0633e-03, ..., 1.4175e-02, + 8.8310e-04, 2.2948e-05], + ..., + [-2.2793e-03, 8.3685e-04, 2.2620e-05, ..., 1.9798e-03, + 1.0691e-03, 1.6406e-05], + [ 1.7715e-04, 1.2779e-03, 6.5327e-04, ..., -1.2531e-03, + 1.5574e-03, 2.8461e-05], + [ 5.1651e-03, -2.0161e-03, 1.0628e-02, ..., -4.6539e-03, + -5.5199e-03, 5.1651e-03]], device='cuda:0') +Epoch 438, bias, value: tensor([-0.0540, 0.0244, 0.0080, -0.0117, -0.0025, 0.0073, -0.0151, 0.0347, + -0.0229, 0.0116], device='cuda:0'), grad: tensor([ 0.0347, -0.0129, 0.0485, -0.0422, -0.0371, 0.0260, -0.0145, 0.0045, + -0.0086, 0.0017], device='cuda:0') +100 +1e-05 +changing lr +epoch 437, time 227.87, cls_loss 0.4819 cls_loss_mapping 0.0007 cls_loss_causal 0.4234 re_mapping 0.0066 re_causal 0.0177 /// teacc 99.00 lr 0.00001000 +Epoch 439, weight, value: tensor([[-0.2154, -0.0218, 0.0384, ..., -0.0295, -0.0819, -0.1474], + [-0.0638, -0.1124, 0.0342, ..., 0.0566, -0.0283, -0.0945], + [-0.0663, -0.1013, -0.1200, ..., 0.0433, -0.0390, -0.1173], + ..., + [-0.0698, 0.0341, 0.0259, ..., 0.0156, -0.0592, -0.0995], + [-0.0874, -0.0063, 0.0158, ..., 0.0619, -0.0353, -0.1920], + [ 0.0648, 0.0698, -0.0521, ..., -0.1158, -0.0150, 0.1001]], + device='cuda:0'), grad: tensor([[ 8.9109e-05, 5.3644e-04, 1.2779e-03, ..., 5.8441e-03, + 5.5504e-03, 2.8324e-04], + [ 9.7394e-05, 1.9989e-03, -5.2757e-03, ..., 4.7684e-06, + 7.6103e-04, 5.9175e-04], + [-8.0317e-06, 1.3466e-03, 1.0500e-03, ..., 4.6463e-03, + 2.0218e-03, 2.7871e-04], + ..., + [ 2.4300e-03, -8.8882e-04, 1.0452e-03, ..., 3.5820e-03, + 2.7561e-03, -1.1063e-03], + [ 2.6083e-04, -2.5463e-03, 9.7275e-04, ..., -8.5373e-03, + -1.4870e-02, 1.1263e-03], + [-3.5343e-03, -2.3918e-03, -3.1414e-03, ..., -9.2010e-03, + -3.4809e-05, -2.7733e-03]], device='cuda:0') +Epoch 439, bias, value: tensor([-0.0541, 0.0245, 0.0079, -0.0118, -0.0025, 0.0074, -0.0152, 0.0347, + -0.0228, 0.0116], device='cuda:0'), grad: tensor([ 0.0379, 0.0251, 0.0277, 0.0044, -0.0013, 0.0106, 0.0262, 0.0095, + -0.0821, -0.0580], device='cuda:0') +100 +1e-05 +changing lr +epoch 438, time 225.30, cls_loss 0.4732 cls_loss_mapping 0.0007 cls_loss_causal 0.4063 re_mapping 0.0067 re_causal 0.0176 /// teacc 99.00 lr 0.00001000 +Epoch 440, weight, value: tensor([[-0.2154, -0.0219, 0.0383, ..., -0.0295, -0.0820, -0.1474], + [-0.0637, -0.1123, 0.0342, ..., 0.0564, -0.0283, -0.0946], + [-0.0662, -0.1013, -0.1199, ..., 0.0433, -0.0390, -0.1173], + ..., + [-0.0698, 0.0340, 0.0258, ..., 0.0155, -0.0593, -0.0996], + [-0.0875, -0.0062, 0.0158, ..., 0.0619, -0.0353, -0.1920], + [ 0.0648, 0.0699, -0.0520, ..., -0.1158, -0.0148, 0.1001]], + device='cuda:0'), grad: tensor([[ 4.7922e-05, 1.1826e-03, 1.0233e-03, ..., 1.8177e-03, + 1.8349e-03, 3.7342e-05], + [ 8.2731e-05, 1.2455e-03, 1.3390e-03, ..., 3.0956e-03, + 2.2697e-03, 2.7716e-05], + [ 8.7321e-05, -7.0457e-03, 5.1842e-03, ..., 9.7733e-03, + -3.1113e-02, 7.4983e-05], + ..., + [ 5.0366e-05, 1.1759e-03, 4.8661e-04, ..., 1.3313e-03, + 2.1225e-02, 8.5682e-06], + [ 3.0220e-05, 7.8583e-04, -5.0507e-03, ..., -8.4076e-03, + 2.2335e-03, 5.6863e-05], + [-6.1893e-04, 2.7390e-03, -5.0049e-03, ..., -1.1101e-02, + -5.4550e-03, -6.2704e-04]], device='cuda:0') +Epoch 440, bias, value: tensor([-0.0541, 0.0244, 0.0080, -0.0118, -0.0026, 0.0073, -0.0152, 0.0347, + -0.0227, 0.0116], device='cuda:0'), grad: tensor([ 0.0164, 0.0232, -0.0217, -0.0109, 0.0237, -0.0203, 0.0204, 0.0267, + -0.0137, -0.0439], device='cuda:0') +100 +1e-05 +changing lr +epoch 439, time 224.55, cls_loss 0.4895 cls_loss_mapping 0.0009 cls_loss_causal 0.4297 re_mapping 0.0064 re_causal 0.0170 /// teacc 99.02 lr 0.00001000 +Epoch 441, weight, value: tensor([[-0.2154, -0.0219, 0.0384, ..., -0.0294, -0.0820, -0.1474], + [-0.0637, -0.1123, 0.0344, ..., 0.0565, -0.0282, -0.0946], + [-0.0662, -0.1013, -0.1200, ..., 0.0432, -0.0390, -0.1174], + ..., + [-0.0697, 0.0339, 0.0258, ..., 0.0155, -0.0593, -0.0993], + [-0.0874, -0.0063, 0.0158, ..., 0.0619, -0.0354, -0.1921], + [ 0.0646, 0.0699, -0.0522, ..., -0.1160, -0.0149, 0.0998]], + device='cuda:0'), grad: tensor([[ 9.9468e-04, 3.1948e-04, 1.0757e-03, ..., 2.9163e-03, + 1.1139e-06, 1.3828e-03], + [ 2.2912e-04, 1.0242e-03, 2.8539e-04, ..., -5.2261e-03, + 4.7415e-05, 1.7941e-04], + [ 5.1498e-04, 2.6073e-03, 3.5381e-04, ..., 4.7340e-03, + 2.7001e-05, 9.9277e-04], + ..., + [-4.4179e-04, -2.3899e-03, -8.7690e-04, ..., -1.1520e-03, + -2.3246e-04, -2.3827e-05], + [ 7.2145e-04, -2.2774e-03, 3.1018e-04, ..., 1.0586e-03, + 3.2842e-05, -8.4162e-05], + [ 9.7990e-05, 1.3077e-04, -4.4060e-04, ..., 2.0523e-03, + 7.4744e-05, -1.7619e-04]], device='cuda:0') +Epoch 441, bias, value: tensor([-0.0541, 0.0246, 0.0081, -0.0117, -0.0027, 0.0075, -0.0153, 0.0347, + -0.0227, 0.0115], device='cuda:0'), grad: tensor([ 0.0175, -0.0300, 0.0252, -0.0585, 0.0141, -0.0042, 0.0205, -0.0051, + 0.0086, 0.0118], device='cuda:0') +100 +1e-05 +changing lr +epoch 440, time 226.38, cls_loss 0.4765 cls_loss_mapping 0.0010 cls_loss_causal 0.4129 re_mapping 0.0062 re_causal 0.0166 /// teacc 99.00 lr 0.00001000 +Epoch 442, weight, value: tensor([[-0.2154, -0.0220, 0.0384, ..., -0.0295, -0.0820, -0.1474], + [-0.0636, -0.1122, 0.0343, ..., 0.0566, -0.0282, -0.0946], + [-0.0664, -0.1013, -0.1200, ..., 0.0431, -0.0389, -0.1174], + ..., + [-0.0697, 0.0340, 0.0258, ..., 0.0155, -0.0594, -0.0993], + [-0.0876, -0.0063, 0.0157, ..., 0.0619, -0.0355, -0.1922], + [ 0.0647, 0.0699, -0.0521, ..., -0.1160, -0.0149, 0.0998]], + device='cuda:0'), grad: tensor([[ 2.8491e-04, 1.6193e-03, 5.4061e-05, ..., 2.4586e-03, + 1.8626e-08, 6.7425e-04], + [ 5.3453e-04, 8.7500e-04, 1.0014e-03, ..., 2.4319e-03, + 3.7253e-09, 1.0377e-04], + [ 3.7360e-04, 2.0084e-03, 1.5092e-04, ..., 1.6985e-03, + 4.4331e-07, 3.3522e-04], + ..., + [ 3.6488e-03, -3.5057e-03, 2.0065e-03, ..., 5.4512e-03, + 1.0684e-05, -1.3580e-03], + [-2.0218e-03, -3.4332e-03, 7.9095e-05, ..., -1.0712e-02, + 3.5949e-06, 1.7643e-04], + [-4.3182e-03, 2.4586e-03, -2.6913e-03, ..., -3.2597e-03, + -1.0006e-05, 3.5286e-04]], device='cuda:0') +Epoch 442, bias, value: tensor([-0.0542, 0.0246, 0.0080, -0.0115, -0.0027, 0.0074, -0.0154, 0.0347, + -0.0227, 0.0115], device='cuda:0'), grad: tensor([ 0.0213, 0.0196, 0.0212, -0.0191, 0.0020, 0.0032, 0.0220, 0.0114, + -0.0794, -0.0023], device='cuda:0') +100 +1e-05 +changing lr +epoch 441, time 224.72, cls_loss 0.4663 cls_loss_mapping 0.0009 cls_loss_causal 0.4083 re_mapping 0.0063 re_causal 0.0166 /// teacc 99.02 lr 0.00001000 +Epoch 443, weight, value: tensor([[-0.2156, -0.0220, 0.0383, ..., -0.0295, -0.0820, -0.1474], + [-0.0635, -0.1123, 0.0343, ..., 0.0566, -0.0283, -0.0950], + [-0.0665, -0.1013, -0.1199, ..., 0.0431, -0.0390, -0.1173], + ..., + [-0.0696, 0.0339, 0.0257, ..., 0.0154, -0.0595, -0.0993], + [-0.0876, -0.0064, 0.0157, ..., 0.0620, -0.0355, -0.1923], + [ 0.0647, 0.0698, -0.0521, ..., -0.1159, -0.0149, 0.0999]], + device='cuda:0'), grad: tensor([[ 1.3137e-04, 3.4422e-05, 2.1994e-04, ..., -1.0023e-03, + 1.1110e-04, 6.1631e-05], + [-5.1308e-04, -5.3167e-04, -1.8759e-03, ..., -4.0100e-02, + -1.0233e-03, 2.2680e-05], + [ 1.8668e-04, 1.2565e-04, 9.7811e-05, ..., 4.5776e-03, + 3.0041e-05, 6.5148e-05], + ..., + [ 1.4067e-04, 3.2673e-03, 1.4555e-04, ..., 4.5319e-03, + 4.8995e-05, 4.6939e-05], + [ 2.9159e-04, 2.9831e-03, 2.9898e-04, ..., 1.1421e-02, + 1.3459e-04, 1.5783e-04], + [ 1.9670e-04, -6.2065e-03, 2.2697e-04, ..., 1.6193e-03, + 7.3552e-05, 6.6578e-05]], device='cuda:0') +Epoch 443, bias, value: tensor([-0.0541, 0.0245, 0.0080, -0.0116, -0.0026, 0.0075, -0.0153, 0.0346, + -0.0228, 0.0116], device='cuda:0'), grad: tensor([-0.0228, -0.0045, -0.0195, -0.0248, 0.0098, 0.0122, 0.0123, 0.0165, + 0.0176, 0.0032], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 442---------------------------------------------------- +epoch 442, time 227.99, cls_loss 0.4877 cls_loss_mapping 0.0009 cls_loss_causal 0.4297 re_mapping 0.0063 re_causal 0.0167 /// teacc 99.04 lr 0.00001000 +Epoch 444, weight, value: tensor([[-0.2156, -0.0219, 0.0384, ..., -0.0295, -0.0818, -0.1475], + [-0.0634, -0.1125, 0.0343, ..., 0.0567, -0.0284, -0.0951], + [-0.0664, -0.1012, -0.1196, ..., 0.0432, -0.0389, -0.1173], + ..., + [-0.0697, 0.0340, 0.0257, ..., 0.0154, -0.0595, -0.0992], + [-0.0875, -0.0064, 0.0156, ..., 0.0620, -0.0355, -0.1925], + [ 0.0647, 0.0697, -0.0522, ..., -0.1159, -0.0149, 0.1001]], + device='cuda:0'), grad: tensor([[ 2.8349e-06, 7.8201e-05, 1.9836e-04, ..., 1.9855e-03, + 2.5606e-04, 1.0610e-04], + [ 4.5449e-05, 1.3590e-03, 1.1933e-04, ..., 1.9722e-03, + 1.5152e-04, 1.2946e-04], + [-1.3506e-04, -1.9627e-03, 1.1539e-04, ..., 1.0910e-03, + 1.3351e-04, 1.0753e-04], + ..., + [ 7.4744e-05, 5.1231e-03, 1.6499e-04, ..., 2.5368e-03, + 2.1112e-04, 8.7814e-03], + [ 1.3530e-05, 1.7762e-04, 1.0312e-04, ..., -3.3035e-03, + 3.4153e-05, 8.3208e-05], + [ 1.9968e-05, -4.6768e-03, 1.6010e-04, ..., 2.2602e-03, + 1.7178e-04, -8.4839e-03]], device='cuda:0') +Epoch 444, bias, value: tensor([-0.0540, 0.0245, 0.0080, -0.0116, -0.0026, 0.0076, -0.0155, 0.0347, + -0.0228, 0.0115], device='cuda:0'), grad: tensor([ 0.0253, 0.0015, 0.0198, -0.0323, 0.0103, -0.0078, -0.0375, 0.0337, + -0.0379, 0.0249], device='cuda:0') +100 +1e-05 +changing lr +epoch 443, time 226.62, cls_loss 0.4854 cls_loss_mapping 0.0010 cls_loss_causal 0.4186 re_mapping 0.0063 re_causal 0.0165 /// teacc 99.00 lr 0.00001000 +Epoch 445, weight, value: tensor([[-0.2156, -0.0219, 0.0384, ..., -0.0295, -0.0816, -0.1475], + [-0.0634, -0.1126, 0.0343, ..., 0.0567, -0.0283, -0.0950], + [-0.0663, -0.1012, -0.1194, ..., 0.0434, -0.0389, -0.1173], + ..., + [-0.0696, 0.0339, 0.0258, ..., 0.0155, -0.0595, -0.0994], + [-0.0876, -0.0062, 0.0156, ..., 0.0620, -0.0355, -0.1925], + [ 0.0647, 0.0696, -0.0522, ..., -0.1160, -0.0150, 0.1000]], + device='cuda:0'), grad: tensor([[ 1.0576e-03, 4.7255e-04, 5.3883e-04, ..., 4.8256e-03, + 9.9087e-04, 4.2725e-04], + [ 3.6907e-04, 2.4462e-04, 4.0674e-04, ..., 4.9286e-03, + 3.4771e-03, 9.1195e-05], + [-2.4128e-03, -1.2331e-03, -4.6043e-03, ..., -3.8574e-02, + -1.8295e-02, 5.2571e-05], + ..., + [ 5.1069e-04, 2.6131e-04, 2.3150e-04, ..., 4.0855e-03, + 1.4696e-03, 3.7074e-05], + [-1.1047e-02, -1.6708e-03, -5.9662e-03, ..., -4.9438e-03, + -5.7030e-03, 1.1292e-03], + [ 7.5188e-03, 2.0063e-04, 4.9400e-03, ..., 8.1787e-03, + 7.7820e-03, 5.3018e-05]], device='cuda:0') +Epoch 445, bias, value: tensor([-0.0540, 0.0245, 0.0081, -0.0116, -0.0026, 0.0074, -0.0155, 0.0349, + -0.0228, 0.0114], device='cuda:0'), grad: tensor([ 0.0188, -0.0123, -0.0779, 0.0042, 0.0158, 0.0136, -0.0012, 0.0229, + -0.0174, 0.0336], device='cuda:0') +100 +1e-05 +changing lr +epoch 444, time 224.10, cls_loss 0.4735 cls_loss_mapping 0.0008 cls_loss_causal 0.4224 re_mapping 0.0063 re_causal 0.0169 /// teacc 99.02 lr 0.00001000 +Epoch 446, weight, value: tensor([[-0.2157, -0.0220, 0.0384, ..., -0.0296, -0.0816, -0.1476], + [-0.0634, -0.1126, 0.0342, ..., 0.0565, -0.0284, -0.0951], + [-0.0664, -0.1012, -0.1195, ..., 0.0435, -0.0389, -0.1174], + ..., + [-0.0696, 0.0339, 0.0258, ..., 0.0155, -0.0595, -0.0993], + [-0.0874, -0.0062, 0.0156, ..., 0.0621, -0.0354, -0.1926], + [ 0.0648, 0.0696, -0.0522, ..., -0.1160, -0.0151, 0.1001]], + device='cuda:0'), grad: tensor([[ 2.5898e-05, 2.0444e-05, -3.8838e-04, ..., 3.1495e-04, + 6.7241e-06, 6.4671e-06], + [ 3.1479e-06, 1.9419e-04, -5.0545e-05, ..., 4.1199e-04, + 7.9349e-07, 9.7975e-07], + [ 3.7514e-06, 3.4213e-05, 2.8089e-05, ..., 3.4499e-04, + 1.0766e-06, 1.5378e-05], + ..., + [ 4.2230e-05, 3.1479e-06, 3.4451e-05, ..., 4.3082e-04, + 7.6741e-06, 5.6997e-06], + [ 1.2958e-04, -3.8362e-04, 4.6611e-05, ..., -3.2196e-03, + 5.6148e-05, 3.0547e-05], + [-2.1327e-04, 1.1861e-05, 4.6760e-05, ..., 3.4499e-04, + -4.0114e-05, -2.8729e-05]], device='cuda:0') +Epoch 446, bias, value: tensor([-0.0541, 0.0243, 0.0082, -0.0116, -0.0026, 0.0076, -0.0154, 0.0348, + -0.0228, 0.0113], device='cuda:0'), grad: tensor([ 0.0039, 0.0169, 0.0035, 0.0111, -0.0112, -0.0027, 0.0213, 0.0063, + -0.0560, 0.0069], device='cuda:0') +100 +1e-05 +changing lr +epoch 445, time 224.85, cls_loss 0.4895 cls_loss_mapping 0.0009 cls_loss_causal 0.4246 re_mapping 0.0061 re_causal 0.0166 /// teacc 98.99 lr 0.00001000 +Epoch 447, weight, value: tensor([[-0.2158, -0.0220, 0.0384, ..., -0.0295, -0.0816, -0.1476], + [-0.0633, -0.1125, 0.0343, ..., 0.0566, -0.0284, -0.0949], + [-0.0663, -0.1012, -0.1196, ..., 0.0435, -0.0389, -0.1174], + ..., + [-0.0695, 0.0341, 0.0260, ..., 0.0157, -0.0595, -0.0992], + [-0.0873, -0.0062, 0.0157, ..., 0.0622, -0.0353, -0.1926], + [ 0.0646, 0.0695, -0.0521, ..., -0.1161, -0.0153, 0.1000]], + device='cuda:0'), grad: tensor([[ 7.5221e-05, 5.5879e-08, 2.8872e-04, ..., 1.1187e-03, + 1.3771e-03, 6.4731e-05], + [-3.7575e-04, 1.3039e-07, -5.7554e-04, ..., -3.6259e-03, + 4.0102e-04, 5.6952e-05], + [ 4.4560e-04, 3.2485e-06, 1.9379e-03, ..., -2.7866e-03, + 1.6725e-04, 1.4937e-04], + ..., + [ 7.8082e-05, -4.6454e-06, -9.9945e-03, ..., 1.1549e-03, + 1.2541e-03, -7.9966e-04], + [ 4.5300e-04, 3.7253e-09, 9.5224e-04, ..., 1.4105e-03, + 4.9515e-03, 3.8624e-04], + [ 1.0997e-04, 8.5309e-07, 3.9520e-03, ..., 1.0557e-03, + -3.4485e-03, 4.2129e-04]], device='cuda:0') +Epoch 447, bias, value: tensor([-0.0541, 0.0243, 0.0081, -0.0116, -0.0027, 0.0076, -0.0153, 0.0349, + -0.0227, 0.0113], device='cuda:0'), grad: tensor([ 0.0116, -0.0259, -0.0126, -0.0236, 0.0160, -0.0041, 0.0138, -0.0077, + 0.0205, 0.0120], device='cuda:0') +100 +1e-05 +changing lr +epoch 446, time 225.72, cls_loss 0.4659 cls_loss_mapping 0.0010 cls_loss_causal 0.4032 re_mapping 0.0061 re_causal 0.0166 /// teacc 98.99 lr 0.00001000 +Epoch 448, weight, value: tensor([[-0.2155, -0.0220, 0.0383, ..., -0.0293, -0.0817, -0.1476], + [-0.0633, -0.1125, 0.0343, ..., 0.0566, -0.0283, -0.0949], + [-0.0664, -0.1013, -0.1195, ..., 0.0435, -0.0389, -0.1177], + ..., + [-0.0694, 0.0342, 0.0261, ..., 0.0158, -0.0594, -0.0992], + [-0.0875, -0.0062, 0.0157, ..., 0.0619, -0.0353, -0.1926], + [ 0.0647, 0.0694, -0.0520, ..., -0.1161, -0.0153, 0.1000]], + device='cuda:0'), grad: tensor([[ 6.4820e-06, 1.7290e-03, 1.1444e-03, ..., 4.1342e-04, + 1.2884e-03, 2.5928e-06], + [ 8.9407e-07, -1.4668e-03, -5.5351e-03, ..., -9.8267e-03, + -6.2904e-03, 3.6880e-07], + [ 1.0826e-05, 6.2408e-03, 8.0252e-04, ..., 2.0008e-03, + 3.8013e-03, 3.0957e-06], + ..., + [ 1.9640e-05, 1.3168e-02, 1.2455e-03, ..., 2.7618e-03, + 1.3380e-03, 1.1161e-05], + [ 5.5122e-04, 5.9204e-03, 5.2929e-04, ..., 1.5879e-03, + -1.1421e-02, 1.6987e-05], + [ 1.8179e-04, -3.1525e-02, 6.2847e-04, ..., 1.8892e-03, + 8.4877e-04, 7.0214e-05]], device='cuda:0') +Epoch 448, bias, value: tensor([-0.0540, 0.0244, 0.0081, -0.0116, -0.0028, 0.0075, -0.0153, 0.0351, + -0.0229, 0.0112], device='cuda:0'), grad: tensor([-0.0354, -0.0440, 0.0216, 0.0082, 0.0122, 0.0071, -0.0107, 0.0229, + 0.0128, 0.0054], device='cuda:0') +100 +1e-05 +changing lr +epoch 447, time 226.53, cls_loss 0.4864 cls_loss_mapping 0.0010 cls_loss_causal 0.4168 re_mapping 0.0063 re_causal 0.0168 /// teacc 99.01 lr 0.00001000 +Epoch 449, weight, value: tensor([[-0.2155, -0.0220, 0.0383, ..., -0.0294, -0.0817, -0.1477], + [-0.0633, -0.1126, 0.0343, ..., 0.0565, -0.0283, -0.0949], + [-0.0665, -0.1015, -0.1197, ..., 0.0435, -0.0389, -0.1177], + ..., + [-0.0694, 0.0344, 0.0261, ..., 0.0158, -0.0593, -0.0995], + [-0.0876, -0.0061, 0.0157, ..., 0.0620, -0.0352, -0.1928], + [ 0.0647, 0.0694, -0.0520, ..., -0.1161, -0.0154, 0.1002]], + device='cuda:0'), grad: tensor([[ 2.3544e-06, 2.3341e-04, 7.6485e-04, ..., -7.4615e-03, + 7.5817e-04, 9.0480e-05], + [ 9.5904e-05, 1.2267e-04, 9.1028e-04, ..., 4.5891e-03, + 1.0366e-03, 9.7573e-05], + [ 2.7418e-05, 9.0003e-05, -3.0670e-03, ..., -3.8452e-03, + -4.5624e-03, -1.5423e-05], + ..., + [ 2.3766e-03, 3.2043e-04, 8.0824e-04, ..., 1.3962e-03, + 4.8876e-04, 1.6613e-03], + [ 9.9659e-05, 1.8966e-04, 7.7057e-04, ..., 3.5133e-03, + 4.8542e-04, 1.6403e-04], + [ 6.2943e-05, -1.0424e-03, 4.0674e-04, ..., 2.8152e-03, + 4.8161e-04, -4.3011e-04]], device='cuda:0') +Epoch 449, bias, value: tensor([-0.0540, 0.0244, 0.0081, -0.0116, -0.0027, 0.0076, -0.0154, 0.0351, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([-0.0689, 0.0314, -0.0088, -0.0441, 0.0258, 0.0323, -0.0084, 0.0001, + 0.0228, 0.0177], device='cuda:0') +100 +1e-05 +changing lr +epoch 448, time 227.34, cls_loss 0.4779 cls_loss_mapping 0.0009 cls_loss_causal 0.4259 re_mapping 0.0063 re_causal 0.0170 /// teacc 98.99 lr 0.00001000 +Epoch 450, weight, value: tensor([[-0.2155, -0.0219, 0.0382, ..., -0.0294, -0.0817, -0.1475], + [-0.0632, -0.1126, 0.0343, ..., 0.0566, -0.0282, -0.0949], + [-0.0667, -0.1015, -0.1196, ..., 0.0437, -0.0389, -0.1177], + ..., + [-0.0694, 0.0346, 0.0261, ..., 0.0159, -0.0593, -0.0994], + [-0.0876, -0.0062, 0.0156, ..., 0.0619, -0.0353, -0.1929], + [ 0.0648, 0.0694, -0.0521, ..., -0.1161, -0.0154, 0.1001]], + device='cuda:0'), grad: tensor([[ 5.5969e-05, -8.1301e-04, 1.1883e-03, ..., -6.4545e-03, + -4.9744e-03, 2.2098e-05], + [ 1.7792e-05, 5.7983e-04, 5.2691e-04, ..., -4.9782e-03, + 1.6537e-03, 1.0125e-05], + [ 4.9829e-04, 2.0266e-04, 9.5558e-04, ..., 2.1267e-03, + 2.4719e-03, 1.3721e-04], + ..., + [ 7.1287e-04, 1.1711e-03, 7.6532e-04, ..., 2.2202e-03, + 2.7351e-03, 7.2956e-05], + [ 4.6234e-03, 1.2255e-03, 4.7255e-04, ..., 2.2087e-03, + 2.0561e-03, 1.4436e-04], + [-1.1040e-02, -1.1902e-03, 8.0633e-04, ..., 5.7983e-04, + -1.2379e-03, 5.3495e-05]], device='cuda:0') +Epoch 450, bias, value: tensor([-0.0539, 0.0245, 0.0081, -0.0117, -0.0030, 0.0076, -0.0154, 0.0352, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([-0.0200, -0.0133, 0.0145, -0.0394, -0.0324, 0.0489, 0.0311, 0.0176, + 0.0189, -0.0259], device='cuda:0') +100 +1e-05 +changing lr +epoch 449, time 227.63, cls_loss 0.4936 cls_loss_mapping 0.0010 cls_loss_causal 0.4342 re_mapping 0.0062 re_causal 0.0165 /// teacc 99.00 lr 0.00001000 +Epoch 451, weight, value: tensor([[-0.2156, -0.0220, 0.0382, ..., -0.0293, -0.0817, -0.1476], + [-0.0633, -0.1126, 0.0343, ..., 0.0566, -0.0284, -0.0950], + [-0.0667, -0.1015, -0.1196, ..., 0.0436, -0.0390, -0.1177], + ..., + [-0.0694, 0.0345, 0.0260, ..., 0.0160, -0.0592, -0.0994], + [-0.0877, -0.0061, 0.0156, ..., 0.0620, -0.0351, -0.1929], + [ 0.0647, 0.0694, -0.0521, ..., -0.1161, -0.0155, 0.1000]], + device='cuda:0'), grad: tensor([[ 3.0088e-04, 6.7139e-04, 1.0691e-03, ..., 3.0403e-03, + 3.6669e-04, 1.2326e-04], + [ 2.2519e-04, -1.1196e-03, 1.6830e-02, ..., 1.8494e-02, + 1.3423e-04, 1.2684e-04], + [-1.6618e-04, 2.7156e-04, 4.1199e-04, ..., -2.7370e-03, + 1.9407e-04, -4.0472e-05], + ..., + [ 4.7779e-04, 9.3918e-03, -1.0595e-03, ..., 8.7585e-03, + 4.8137e-04, 1.2827e-04], + [ 1.6193e-03, 6.5327e-04, -4.8447e-03, ..., -9.9335e-03, + 2.9011e-03, 2.2721e-04], + [ 1.0662e-03, -8.5907e-03, -1.2474e-02, ..., -1.6052e-02, + 4.2496e-03, -3.0899e-04]], device='cuda:0') +Epoch 451, bias, value: tensor([-0.0539, 0.0245, 0.0081, -0.0117, -0.0029, 0.0075, -0.0155, 0.0353, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0176, 0.0495, -0.0160, 0.0266, -0.0396, 0.0146, -0.0133, 0.0410, + -0.0384, -0.0418], device='cuda:0') +100 +1e-05 +changing lr +epoch 450, time 227.15, cls_loss 0.4922 cls_loss_mapping 0.0009 cls_loss_causal 0.4224 re_mapping 0.0061 re_causal 0.0169 /// teacc 99.04 lr 0.00001000 +Epoch 452, weight, value: tensor([[-0.2155, -0.0220, 0.0380, ..., -0.0294, -0.0819, -0.1476], + [-0.0632, -0.1125, 0.0341, ..., 0.0565, -0.0285, -0.0951], + [-0.0668, -0.1013, -0.1199, ..., 0.0435, -0.0390, -0.1177], + ..., + [-0.0693, 0.0346, 0.0262, ..., 0.0160, -0.0591, -0.0993], + [-0.0880, -0.0062, 0.0157, ..., 0.0619, -0.0352, -0.1929], + [ 0.0648, 0.0693, -0.0520, ..., -0.1161, -0.0155, 0.1001]], + device='cuda:0'), grad: tensor([[-7.0095e-05, 8.8453e-04, 3.4475e-04, ..., 1.1940e-03, + 6.6459e-05, -1.0902e-04], + [-7.7057e-04, 8.9407e-04, -2.2602e-04, ..., -3.1376e-03, + 8.0466e-05, -1.8721e-03], + [ 2.9612e-04, 2.4734e-02, 3.7527e-04, ..., -7.3719e-04, + 2.5964e-04, 5.2547e-04], + ..., + [ 3.7813e-04, -2.4231e-02, 8.2827e-04, ..., -1.0357e-03, + 6.3038e-04, 2.3985e-04], + [ 1.1883e-03, -6.6147e-03, 2.5616e-03, ..., 2.0218e-03, + 1.8778e-03, 8.3303e-04], + [ 3.3784e-04, 1.5144e-03, 7.3576e-04, ..., 2.6283e-03, + 5.4026e-04, 2.7752e-04]], device='cuda:0') +Epoch 452, bias, value: tensor([-0.0539, 0.0243, 0.0080, -0.0116, -0.0027, 0.0074, -0.0154, 0.0354, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0233, 0.0088, 0.0151, 0.0453, -0.0396, -0.0095, -0.0227, -0.0368, + -0.0057, 0.0220], device='cuda:0') +100 +1e-05 +changing lr +epoch 451, time 227.29, cls_loss 0.4930 cls_loss_mapping 0.0010 cls_loss_causal 0.4310 re_mapping 0.0062 re_causal 0.0164 /// teacc 99.00 lr 0.00001000 +Epoch 453, weight, value: tensor([[-0.2153, -0.0219, 0.0382, ..., -0.0292, -0.0819, -0.1476], + [-0.0632, -0.1125, 0.0342, ..., 0.0565, -0.0285, -0.0947], + [-0.0669, -0.1014, -0.1199, ..., 0.0435, -0.0391, -0.1178], + ..., + [-0.0693, 0.0345, 0.0263, ..., 0.0161, -0.0590, -0.0994], + [-0.0880, -0.0060, 0.0156, ..., 0.0618, -0.0352, -0.1931], + [ 0.0647, 0.0693, -0.0519, ..., -0.1161, -0.0156, 0.1001]], + device='cuda:0'), grad: tensor([[ 1.9753e-04, 3.7003e-04, 7.5912e-04, ..., 1.8234e-03, + 0.0000e+00, 2.5225e-04], + [ 2.8804e-05, 7.7724e-05, 7.0810e-04, ..., 2.1648e-03, + 0.0000e+00, 6.0081e-05], + [ 1.2362e-04, 2.4109e-02, 1.4057e-03, ..., -1.4229e-03, + 1.8626e-08, 5.3501e-04], + ..., + [ 1.9580e-05, 1.7426e-02, -1.0834e-03, ..., 4.7970e-04, + 3.7253e-09, 2.1305e-03], + [ 3.3522e-04, 8.5449e-04, 8.2922e-04, ..., 2.2049e-03, + 3.7253e-09, 5.7220e-04], + [ 8.5545e-04, -1.5495e-02, 1.4439e-03, ..., 3.2997e-03, + 0.0000e+00, -1.1473e-03]], device='cuda:0') +Epoch 453, bias, value: tensor([-0.0539, 0.0244, 0.0080, -0.0115, -0.0027, 0.0075, -0.0157, 0.0355, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0119, 0.0140, 0.0041, -0.0327, -0.0166, -0.0021, -0.0185, 0.0087, + 0.0144, 0.0168], device='cuda:0') +100 +1e-05 +changing lr +epoch 452, time 226.77, cls_loss 0.4833 cls_loss_mapping 0.0010 cls_loss_causal 0.4247 re_mapping 0.0060 re_causal 0.0160 /// teacc 99.00 lr 0.00001000 +Epoch 454, weight, value: tensor([[-0.2153, -0.0220, 0.0382, ..., -0.0291, -0.0819, -0.1474], + [-0.0633, -0.1124, 0.0342, ..., 0.0565, -0.0286, -0.0946], + [-0.0670, -0.1015, -0.1200, ..., 0.0436, -0.0391, -0.1178], + ..., + [-0.0694, 0.0343, 0.0262, ..., 0.0161, -0.0590, -0.0994], + [-0.0880, -0.0060, 0.0156, ..., 0.0618, -0.0352, -0.1933], + [ 0.0648, 0.0694, -0.0520, ..., -0.1162, -0.0157, 0.1001]], + device='cuda:0'), grad: tensor([[ 2.0623e-04, 3.1452e-03, 2.6822e-04, ..., 1.4610e-03, + -3.5629e-03, 7.9870e-04], + [-7.1526e-03, 2.8634e-04, -6.0501e-03, ..., -6.1531e-03, + 6.2752e-04, 2.0802e-04], + [ 3.5858e-04, 4.0936e-04, 2.9125e-03, ..., 4.2152e-03, + 6.7616e-04, 1.2808e-03], + ..., + [ 3.5048e-05, 3.2330e-04, 1.0979e-04, ..., 1.1444e-03, + 4.9639e-04, 2.7895e-04], + [ 3.6755e-03, 1.1034e-03, 2.7237e-03, ..., -9.9030e-03, + 2.0676e-03, 1.2274e-03], + [-9.8896e-04, 1.6613e-03, 2.0730e-04, ..., -3.6983e-03, + -1.5287e-03, 5.5027e-04]], device='cuda:0') +Epoch 454, bias, value: tensor([-0.0538, 0.0244, 0.0080, -0.0115, -0.0026, 0.0073, -0.0156, 0.0355, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([-0.0070, -0.0192, 0.0194, -0.0101, 0.0137, 0.0332, -0.0071, 0.0078, + -0.0059, -0.0249], device='cuda:0') +100 +1e-05 +changing lr +epoch 453, time 224.02, cls_loss 0.5118 cls_loss_mapping 0.0010 cls_loss_causal 0.4567 re_mapping 0.0059 re_causal 0.0168 /// teacc 99.01 lr 0.00001000 +Epoch 455, weight, value: tensor([[-0.2153, -0.0220, 0.0382, ..., -0.0293, -0.0819, -0.1475], + [-0.0634, -0.1123, 0.0342, ..., 0.0565, -0.0287, -0.0946], + [-0.0671, -0.1015, -0.1200, ..., 0.0434, -0.0391, -0.1179], + ..., + [-0.0694, 0.0342, 0.0264, ..., 0.0161, -0.0589, -0.0996], + [-0.0881, -0.0061, 0.0154, ..., 0.0618, -0.0351, -0.1934], + [ 0.0648, 0.0695, -0.0520, ..., -0.1161, -0.0157, 0.1001]], + device='cuda:0'), grad: tensor([[-2.1572e-03, -2.3767e-05, -3.9825e-03, ..., -2.9488e-03, + -8.9884e-04, -9.7084e-04], + [ 1.7142e-04, 2.4147e-03, 7.7629e-04, ..., 2.3441e-03, + 7.1406e-05, 8.9288e-05], + [ 4.1544e-05, 6.5446e-05, -1.5192e-03, ..., -6.1607e-03, + 1.6868e-05, 1.8501e-04], + ..., + [ 3.8171e-04, -1.1314e-02, 8.3780e-04, ..., 1.3180e-03, + 1.5748e-04, 1.2982e-04], + [ 6.1846e-04, -2.0237e-03, 1.3418e-03, ..., 5.3406e-04, + 2.5678e-04, 4.5657e-04], + [ 7.2002e-04, 2.5482e-03, 1.5669e-03, ..., 2.5864e-03, + 3.0613e-04, 4.0197e-04]], device='cuda:0') +Epoch 455, bias, value: tensor([-0.0539, 0.0244, 0.0078, -0.0114, -0.0026, 0.0074, -0.0156, 0.0355, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([-0.0080, 0.0180, -0.0531, 0.0090, 0.0122, 0.0134, 0.0033, -0.0131, + -0.0007, 0.0190], device='cuda:0') +100 +1e-05 +changing lr +epoch 454, time 225.98, cls_loss 0.4695 cls_loss_mapping 0.0009 cls_loss_causal 0.4018 re_mapping 0.0062 re_causal 0.0165 /// teacc 99.00 lr 0.00001000 +Epoch 456, weight, value: tensor([[-0.2153, -0.0221, 0.0381, ..., -0.0292, -0.0819, -0.1476], + [-0.0634, -0.1123, 0.0342, ..., 0.0566, -0.0287, -0.0946], + [-0.0671, -0.1014, -0.1201, ..., 0.0434, -0.0391, -0.1178], + ..., + [-0.0694, 0.0342, 0.0264, ..., 0.0160, -0.0590, -0.0997], + [-0.0881, -0.0062, 0.0156, ..., 0.0617, -0.0351, -0.1935], + [ 0.0647, 0.0695, -0.0520, ..., -0.1161, -0.0156, 0.1002]], + device='cuda:0'), grad: tensor([[ 0.0010, 0.0006, 0.0014, ..., 0.0025, 0.0002, 0.0003], + [ 0.0017, 0.0016, 0.0030, ..., 0.0052, 0.0005, 0.0014], + [ 0.0031, 0.0010, 0.0022, ..., 0.0044, 0.0009, 0.0010], + ..., + [-0.0054, 0.0007, -0.0040, ..., -0.0118, 0.0001, 0.0002], + [-0.0064, -0.0310, -0.0041, ..., -0.0075, -0.0035, -0.0034], + [ 0.0023, 0.0083, 0.0022, ..., 0.0024, 0.0003, 0.0005]], + device='cuda:0') +Epoch 456, bias, value: tensor([-0.0539, 0.0245, 0.0078, -0.0115, -0.0024, 0.0074, -0.0156, 0.0354, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([ 0.0112, 0.0299, 0.0190, 0.0035, 0.0292, -0.0147, 0.0192, -0.0498, + -0.0692, 0.0215], device='cuda:0') +100 +1e-05 +changing lr +epoch 455, time 225.54, cls_loss 0.4885 cls_loss_mapping 0.0008 cls_loss_causal 0.4299 re_mapping 0.0060 re_causal 0.0163 /// teacc 99.00 lr 0.00001000 +Epoch 457, weight, value: tensor([[-0.2151, -0.0220, 0.0381, ..., -0.0294, -0.0819, -0.1476], + [-0.0634, -0.1126, 0.0343, ..., 0.0568, -0.0286, -0.0947], + [-0.0671, -0.1015, -0.1201, ..., 0.0434, -0.0391, -0.1178], + ..., + [-0.0694, 0.0343, 0.0265, ..., 0.0160, -0.0587, -0.0997], + [-0.0880, -0.0063, 0.0156, ..., 0.0618, -0.0351, -0.1936], + [ 0.0648, 0.0697, -0.0520, ..., -0.1160, -0.0157, 0.1003]], + device='cuda:0'), grad: tensor([[ 1.6317e-06, 5.7507e-04, -1.7905e-04, ..., 3.1781e-04, + 9.4995e-08, 1.6857e-06], + [ 2.2445e-06, 8.1730e-04, 8.1062e-06, ..., 5.4979e-04, + 3.5018e-07, 3.0212e-06], + [ 2.4438e-06, 5.5981e-04, 2.0131e-05, ..., -4.8971e-04, + 1.0245e-07, 4.9435e-06], + ..., + [ 2.6464e-04, -3.0594e-03, 1.0079e-04, ..., 7.1573e-04, + 3.9525e-06, 2.0444e-04], + [ 4.3184e-05, -1.7190e-04, 5.5730e-05, ..., -7.8917e-04, + 1.8682e-06, -1.2264e-03], + [-5.7030e-04, 6.0749e-04, -2.3830e-04, ..., -4.1351e-03, + -1.9416e-05, -3.3069e-04]], device='cuda:0') +Epoch 457, bias, value: tensor([-0.0539, 0.0246, 0.0078, -0.0116, -0.0027, 0.0074, -0.0155, 0.0354, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0048, 0.0102, 0.0061, 0.0064, 0.0065, 0.0100, -0.0030, -0.0196, + 0.0054, -0.0269], device='cuda:0') +100 +1e-05 +changing lr +epoch 456, time 230.11, cls_loss 0.5059 cls_loss_mapping 0.0009 cls_loss_causal 0.4437 re_mapping 0.0062 re_causal 0.0171 /// teacc 99.01 lr 0.00001000 +Epoch 458, weight, value: tensor([[-0.2151, -0.0220, 0.0381, ..., -0.0295, -0.0819, -0.1476], + [-0.0634, -0.1126, 0.0342, ..., 0.0567, -0.0286, -0.0947], + [-0.0671, -0.1014, -0.1202, ..., 0.0434, -0.0391, -0.1178], + ..., + [-0.0694, 0.0341, 0.0264, ..., 0.0160, -0.0588, -0.0997], + [-0.0880, -0.0063, 0.0156, ..., 0.0618, -0.0352, -0.1937], + [ 0.0648, 0.0699, -0.0518, ..., -0.1160, -0.0157, 0.1003]], + device='cuda:0'), grad: tensor([[ 2.3994e-03, 1.1101e-06, 6.0387e-03, ..., 4.5395e-03, + 2.4014e-03, 9.0152e-06], + [ 2.0742e-05, 2.0337e-04, 8.2403e-06, ..., 3.6278e-03, + 6.8247e-06, 1.0568e-04], + [ 3.7537e-03, 2.1660e-04, 8.5115e-05, ..., 2.6932e-03, + 6.0126e-06, 1.7290e-03], + ..., + [ 1.8120e-04, -1.5450e-03, -2.2197e-04, ..., 1.0386e-03, + 1.3731e-05, -1.6880e-03], + [ 1.0365e-04, 1.1843e-04, 6.2752e-04, ..., -1.6037e-02, + 5.3525e-05, 1.0949e-04], + [-1.6975e-03, 5.1880e-04, 4.5538e-04, ..., 3.1548e-03, + -1.7738e-04, -3.5882e-04]], device='cuda:0') +Epoch 458, bias, value: tensor([-0.0540, 0.0246, 0.0079, -0.0115, -0.0026, 0.0076, -0.0156, 0.0353, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0450, 0.0167, -0.0097, -0.0269, 0.0192, 0.0156, -0.0648, 0.0010, + -0.0171, 0.0211], device='cuda:0') +100 +1e-05 +changing lr +epoch 457, time 225.77, cls_loss 0.5075 cls_loss_mapping 0.0009 cls_loss_causal 0.4515 re_mapping 0.0061 re_causal 0.0168 /// teacc 99.02 lr 0.00001000 +Epoch 459, weight, value: tensor([[-0.2152, -0.0218, 0.0380, ..., -0.0296, -0.0820, -0.1476], + [-0.0633, -0.1126, 0.0341, ..., 0.0567, -0.0287, -0.0947], + [-0.0672, -0.1013, -0.1203, ..., 0.0433, -0.0389, -0.1179], + ..., + [-0.0695, 0.0341, 0.0265, ..., 0.0160, -0.0589, -0.0998], + [-0.0879, -0.0063, 0.0156, ..., 0.0618, -0.0351, -0.1937], + [ 0.0649, 0.0698, -0.0518, ..., -0.1161, -0.0158, 0.1003]], + device='cuda:0'), grad: tensor([[ 8.5950e-05, 6.4325e-04, 3.5954e-04, ..., 2.4986e-03, + 5.2881e-04, 4.2841e-08], + [ 1.1712e-04, 6.9809e-04, -1.2934e-04, ..., -2.0742e-04, + 2.3365e-04, 1.5087e-07], + [ 1.8311e-04, 4.7350e-04, 1.4734e-04, ..., -4.9496e-04, + -1.5669e-03, 1.7256e-05], + ..., + [ 7.6818e-04, -8.3971e-04, 7.1335e-04, ..., 4.0398e-03, + 2.3794e-04, 1.7047e-05], + [ 1.4572e-03, 6.2895e-04, 5.4121e-04, ..., 1.0157e-03, + 6.7282e-04, 5.6364e-06], + [ 7.0286e-04, -1.2197e-03, -5.9891e-04, ..., -7.0763e-04, + 3.8552e-04, -2.9609e-05]], device='cuda:0') +Epoch 459, bias, value: tensor([-0.0540, 0.0247, 0.0079, -0.0116, -0.0024, 0.0075, -0.0157, 0.0353, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([ 0.0246, -0.0019, -0.0002, 0.0058, 0.0098, -0.0613, -0.0084, 0.0284, + -0.0006, 0.0037], device='cuda:0') +100 +1e-05 +changing lr +epoch 458, time 225.66, cls_loss 0.4997 cls_loss_mapping 0.0008 cls_loss_causal 0.4340 re_mapping 0.0061 re_causal 0.0170 /// teacc 98.98 lr 0.00001000 +Epoch 460, weight, value: tensor([[-0.2152, -0.0218, 0.0380, ..., -0.0295, -0.0821, -0.1475], + [-0.0633, -0.1126, 0.0341, ..., 0.0566, -0.0288, -0.0947], + [-0.0671, -0.1012, -0.1204, ..., 0.0432, -0.0388, -0.1180], + ..., + [-0.0693, 0.0341, 0.0266, ..., 0.0162, -0.0587, -0.0997], + [-0.0880, -0.0063, 0.0155, ..., 0.0618, -0.0351, -0.1939], + [ 0.0652, 0.0698, -0.0517, ..., -0.1160, -0.0159, 0.1005]], + device='cuda:0'), grad: tensor([[ 1.5211e-04, 8.8882e-04, 6.7949e-04, ..., 6.1941e-04, + 7.2539e-05, 1.5244e-05], + [ 1.1402e-04, 2.3975e-03, 7.0333e-04, ..., 1.3790e-03, + 5.6684e-05, 3.9488e-06], + [-1.2293e-03, -6.8617e-04, 6.3956e-05, ..., -4.7951e-03, + -7.1287e-04, 5.1707e-06], + ..., + [ 3.0804e-04, 1.2188e-03, 9.4831e-05, ..., 7.4911e-04, + 1.2553e-04, 9.5069e-05], + [-1.9501e-02, 2.5368e-03, 3.7909e-04, ..., -4.4785e-03, + 2.2948e-04, 3.3832e-04], + [ 2.3174e-03, -8.4152e-03, 1.1215e-03, ..., 7.5960e-04, + 6.8617e-04, 1.1158e-03]], device='cuda:0') +Epoch 460, bias, value: tensor([-0.0541, 0.0245, 0.0079, -0.0116, -0.0024, 0.0075, -0.0156, 0.0354, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0081, 0.0110, -0.0502, 0.0094, 0.0168, 0.0195, -0.0021, 0.0081, + -0.0164, -0.0041], device='cuda:0') +100 +1e-05 +changing lr +epoch 459, time 223.60, cls_loss 0.4977 cls_loss_mapping 0.0007 cls_loss_causal 0.4428 re_mapping 0.0064 re_causal 0.0175 /// teacc 98.94 lr 0.00001000 +Epoch 461, weight, value: tensor([[-0.2152, -0.0219, 0.0379, ..., -0.0296, -0.0820, -0.1476], + [-0.0633, -0.1125, 0.0341, ..., 0.0567, -0.0287, -0.0946], + [-0.0672, -0.1012, -0.1205, ..., 0.0432, -0.0389, -0.1181], + ..., + [-0.0693, 0.0342, 0.0266, ..., 0.0162, -0.0588, -0.0997], + [-0.0880, -0.0063, 0.0155, ..., 0.0618, -0.0350, -0.1939], + [ 0.0652, 0.0697, -0.0517, ..., -0.1160, -0.0159, 0.1005]], + device='cuda:0'), grad: tensor([[ 3.9153e-06, 3.1805e-04, 1.9036e-06, ..., 1.6079e-03, + 1.0282e-06, 9.0718e-05], + [-4.5225e-06, 1.6129e-04, -6.4790e-05, ..., -1.9522e-03, + 4.5300e-06, 1.0693e-04], + [ 4.3511e-06, -3.8849e-02, 2.1178e-06, ..., -2.6550e-03, + 4.1798e-06, 9.4235e-05], + ..., + [ 1.2442e-05, 3.3142e-02, 4.0412e-05, ..., 3.9520e-03, + -1.0900e-05, -9.6083e-04], + [ 5.3501e-04, 2.1994e-04, 5.7936e-04, ..., 1.7715e-04, + 1.1456e-04, 1.6034e-04], + [-1.4105e-03, 5.0392e-03, -1.6079e-03, ..., 2.1973e-03, + -3.2043e-04, 2.0817e-05]], device='cuda:0') +Epoch 461, bias, value: tensor([-0.0541, 0.0246, 0.0079, -0.0115, -0.0027, 0.0075, -0.0156, 0.0354, + -0.0229, 0.0111], device='cuda:0'), grad: tensor([ 0.0175, -0.0140, -0.0068, -0.0126, 0.0208, 0.0181, -0.0403, 0.0109, + -0.0118, 0.0182], device='cuda:0') +100 +1e-05 +changing lr +epoch 460, time 227.83, cls_loss 0.4764 cls_loss_mapping 0.0008 cls_loss_causal 0.4150 re_mapping 0.0062 re_causal 0.0166 /// teacc 98.96 lr 0.00001000 +Epoch 462, weight, value: tensor([[-0.2153, -0.0218, 0.0378, ..., -0.0297, -0.0822, -0.1477], + [-0.0632, -0.1127, 0.0341, ..., 0.0565, -0.0288, -0.0949], + [-0.0673, -0.1013, -0.1205, ..., 0.0431, -0.0388, -0.1183], + ..., + [-0.0693, 0.0343, 0.0266, ..., 0.0163, -0.0589, -0.0998], + [-0.0880, -0.0064, 0.0156, ..., 0.0620, -0.0350, -0.1939], + [ 0.0653, 0.0697, -0.0517, ..., -0.1161, -0.0158, 0.1005]], + device='cuda:0'), grad: tensor([[ 0.0002, 0.0004, -0.0035, ..., -0.0007, -0.0069, 0.0002], + [ 0.0013, 0.0004, 0.0032, ..., 0.0054, 0.0010, 0.0008], + [ 0.0012, 0.0006, 0.0024, ..., 0.0016, 0.0010, 0.0003], + ..., + [ 0.0002, 0.0011, 0.0056, ..., 0.0071, 0.0029, 0.0001], + [-0.0030, 0.0026, -0.0033, ..., -0.0023, 0.0016, 0.0002], + [ 0.0010, 0.0034, -0.0052, ..., -0.0032, 0.0029, 0.0005]], + device='cuda:0') +Epoch 462, bias, value: tensor([-0.0541, 0.0245, 0.0079, -0.0114, -0.0028, 0.0076, -0.0156, 0.0354, + -0.0228, 0.0110], device='cuda:0'), grad: tensor([-0.0072, 0.0332, -0.0026, -0.0309, -0.0059, 0.0022, -0.0423, 0.0412, + 0.0073, 0.0050], device='cuda:0') +100 +1e-05 +changing lr +epoch 461, time 228.37, cls_loss 0.5025 cls_loss_mapping 0.0009 cls_loss_causal 0.4403 re_mapping 0.0063 re_causal 0.0169 /// teacc 99.00 lr 0.00001000 +Epoch 463, weight, value: tensor([[-0.2153, -0.0218, 0.0378, ..., -0.0299, -0.0823, -0.1478], + [-0.0633, -0.1126, 0.0340, ..., 0.0565, -0.0288, -0.0950], + [-0.0672, -0.1014, -0.1205, ..., 0.0432, -0.0387, -0.1181], + ..., + [-0.0692, 0.0344, 0.0266, ..., 0.0164, -0.0589, -0.0998], + [-0.0879, -0.0065, 0.0156, ..., 0.0621, -0.0349, -0.1939], + [ 0.0653, 0.0696, -0.0517, ..., -0.1162, -0.0159, 0.1005]], + device='cuda:0'), grad: tensor([[ 4.0978e-07, 6.4468e-04, 8.8215e-04, ..., 3.1071e-03, + 4.3011e-04, 7.6532e-05], + [ 4.5002e-06, -5.4436e-03, -1.3170e-03, ..., -3.9520e-03, + 1.0414e-03, 3.9130e-05], + [ 2.0694e-06, 1.6403e-03, 2.1780e-04, ..., 4.3945e-03, + 2.4676e-04, 4.7415e-05], + ..., + [ 2.0957e-04, 1.9245e-03, 5.0640e-04, ..., 9.9411e-03, + 3.4618e-04, -5.9509e-04], + [ 1.0617e-05, -1.1883e-03, 4.9973e-04, ..., -1.8673e-03, + 2.7800e-04, -3.4285e-04], + [-1.1425e-03, 2.6360e-03, -5.6177e-05, ..., 7.6723e-04, + 1.4877e-04, 1.2092e-05]], device='cuda:0') +Epoch 463, bias, value: tensor([-0.0542, 0.0245, 0.0079, -0.0114, -0.0027, 0.0076, -0.0157, 0.0354, + -0.0228, 0.0110], device='cuda:0'), grad: tensor([ 0.0244, -0.0131, 0.0214, -0.0096, -0.0392, -0.0181, 0.0215, 0.0325, + -0.0132, -0.0067], device='cuda:0') +100 +1e-05 +changing lr +epoch 462, time 226.96, cls_loss 0.4776 cls_loss_mapping 0.0008 cls_loss_causal 0.4156 re_mapping 0.0061 re_causal 0.0163 /// teacc 98.97 lr 0.00001000 +Epoch 464, weight, value: tensor([[-0.2154, -0.0219, 0.0378, ..., -0.0299, -0.0823, -0.1478], + [-0.0633, -0.1127, 0.0341, ..., 0.0564, -0.0288, -0.0951], + [-0.0671, -0.1015, -0.1204, ..., 0.0432, -0.0387, -0.1181], + ..., + [-0.0691, 0.0345, 0.0266, ..., 0.0164, -0.0587, -0.0998], + [-0.0879, -0.0066, 0.0156, ..., 0.0620, -0.0349, -0.1940], + [ 0.0653, 0.0695, -0.0516, ..., -0.1162, -0.0157, 0.1005]], + device='cuda:0'), grad: tensor([[ 1.7285e-05, 1.2755e-04, 4.3583e-04, ..., 1.3695e-03, + 9.3132e-06, -1.2016e-04], + [ 2.6636e-07, 4.7421e-04, -1.0128e-03, ..., 1.6654e-04, + 2.4751e-05, 2.2268e-04], + [ 3.8017e-06, 7.3004e-04, 1.9503e-04, ..., 2.6474e-03, + 9.6321e-05, 4.4775e-04], + ..., + [ 6.9695e-03, -7.5960e-04, 2.0063e-04, ..., -4.0207e-03, + 1.5929e-05, 6.9797e-05], + [ 1.9848e-04, 3.4237e-04, -9.5654e-04, ..., 7.7391e-04, + -2.2137e-04, 1.3113e-04], + [-7.0038e-03, -3.3813e-02, 4.8399e-04, ..., 2.2736e-03, + 2.4334e-05, 4.6086e-04]], device='cuda:0') +Epoch 464, bias, value: tensor([-0.0542, 0.0245, 0.0081, -0.0114, -0.0027, 0.0076, -0.0157, 0.0355, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([ 0.0064, 0.0054, 0.0137, 0.0095, 0.0042, -0.0227, 0.0086, 0.0026, + 0.0050, -0.0327], device='cuda:0') +100 +1e-05 +changing lr +epoch 463, time 226.73, cls_loss 0.4860 cls_loss_mapping 0.0009 cls_loss_causal 0.4190 re_mapping 0.0061 re_causal 0.0162 /// teacc 99.03 lr 0.00001000 +Epoch 465, weight, value: tensor([[-0.2155, -0.0218, 0.0377, ..., -0.0298, -0.0823, -0.1479], + [-0.0633, -0.1127, 0.0340, ..., 0.0564, -0.0287, -0.0948], + [-0.0671, -0.1014, -0.1205, ..., 0.0431, -0.0387, -0.1181], + ..., + [-0.0692, 0.0344, 0.0265, ..., 0.0164, -0.0589, -0.0999], + [-0.0876, -0.0065, 0.0156, ..., 0.0620, -0.0349, -0.1941], + [ 0.0653, 0.0695, -0.0515, ..., -0.1160, -0.0155, 0.1004]], + device='cuda:0'), grad: tensor([[ 1.2755e-04, -1.0514e-04, 1.6630e-04, ..., 1.0242e-03, + -1.8477e-05, -2.0421e-04], + [ 1.3970e-06, -6.9082e-05, 7.2289e-04, ..., 2.3251e-03, + 3.2969e-07, 2.4308e-06], + [-4.0680e-06, -2.3353e-04, -2.4624e-03, ..., -1.0498e-02, + 4.0233e-06, 6.2227e-05], + ..., + [ 5.5581e-05, 2.2739e-05, 2.8744e-03, ..., 3.7098e-03, + 6.3851e-06, 5.3883e-05], + [ 4.7863e-05, 4.7798e-03, 1.3905e-03, ..., 2.0962e-03, + -5.6513e-06, 8.1837e-05], + [ 4.0621e-05, -5.6839e-03, 1.4296e-03, ..., 2.1992e-03, + 2.5749e-05, 1.1051e-04]], device='cuda:0') +Epoch 465, bias, value: tensor([-0.0541, 0.0245, 0.0079, -0.0115, -0.0027, 0.0076, -0.0157, 0.0354, + -0.0228, 0.0111], device='cuda:0'), grad: tensor([ 0.0075, 0.0121, -0.0507, 0.0119, -0.0163, 0.0113, -0.0208, 0.0193, + 0.0166, 0.0090], device='cuda:0') +100 +1e-05 +changing lr +epoch 464, time 228.71, cls_loss 0.4899 cls_loss_mapping 0.0009 cls_loss_causal 0.4159 re_mapping 0.0062 re_causal 0.0168 /// teacc 99.01 lr 0.00001000 +Epoch 466, weight, value: tensor([[-0.2154, -0.0220, 0.0378, ..., -0.0300, -0.0825, -0.1480], + [-0.0632, -0.1128, 0.0342, ..., 0.0565, -0.0286, -0.0949], + [-0.0671, -0.1015, -0.1205, ..., 0.0431, -0.0384, -0.1181], + ..., + [-0.0692, 0.0345, 0.0264, ..., 0.0164, -0.0589, -0.0999], + [-0.0877, -0.0065, 0.0155, ..., 0.0620, -0.0349, -0.1941], + [ 0.0653, 0.0696, -0.0514, ..., -0.1158, -0.0155, 0.1005]], + device='cuda:0'), grad: tensor([[ 1.2517e-04, -6.0806e-03, 2.0221e-05, ..., -1.3474e-02, + -1.1887e-02, 1.9804e-05], + [ 3.4733e-03, 3.7694e-04, 1.9956e-04, ..., 2.0981e-03, + 6.7520e-04, 1.9327e-05], + [ 9.6512e-04, 1.0864e-02, 7.0453e-05, ..., 4.5815e-03, + 2.0790e-03, 7.1812e-04], + ..., + [ 1.9245e-03, 4.0078e-04, 1.0967e-04, ..., 2.9507e-03, + 1.9274e-03, 3.6478e-05], + [ 4.5824e-04, 2.4529e-03, 4.6164e-05, ..., 1.6766e-03, + -1.2188e-03, 4.0382e-05], + [ 2.8019e-03, 2.6264e-03, 2.8419e-03, ..., 3.1528e-03, + 1.7071e-03, 1.2283e-03]], device='cuda:0') +Epoch 466, bias, value: tensor([-0.0543, 0.0245, 0.0080, -0.0114, -0.0028, 0.0075, -0.0158, 0.0356, + -0.0228, 0.0112], device='cuda:0'), grad: tensor([-0.0369, 0.0142, 0.0227, 0.0021, -0.0315, -0.0208, 0.0158, 0.0121, + 0.0032, 0.0190], device='cuda:0') +100 +1e-05 +changing lr +epoch 465, time 226.94, cls_loss 0.4604 cls_loss_mapping 0.0008 cls_loss_causal 0.4101 re_mapping 0.0063 re_causal 0.0167 /// teacc 99.00 lr 0.00001000 +Epoch 467, weight, value: tensor([[-0.2155, -0.0220, 0.0378, ..., -0.0299, -0.0825, -0.1478], + [-0.0632, -0.1128, 0.0342, ..., 0.0566, -0.0286, -0.0949], + [-0.0671, -0.1014, -0.1205, ..., 0.0430, -0.0384, -0.1182], + ..., + [-0.0693, 0.0344, 0.0263, ..., 0.0164, -0.0590, -0.0999], + [-0.0877, -0.0065, 0.0157, ..., 0.0620, -0.0348, -0.1939], + [ 0.0653, 0.0696, -0.0514, ..., -0.1158, -0.0154, 0.1006]], + device='cuda:0'), grad: tensor([[-1.1718e-04, 1.3329e-05, 1.3304e-04, ..., 7.9250e-04, + 3.9712e-06, 1.0263e-06], + [ 4.8012e-05, -3.8981e-04, 3.8362e-04, ..., 1.5991e-02, + 3.4034e-05, 1.7419e-05], + [ 4.2021e-05, 6.2406e-05, 2.9492e-04, ..., 1.1988e-03, + 7.3314e-06, 2.9560e-06], + ..., + [ 1.0884e-04, 3.9607e-05, -3.3016e-03, ..., -1.7197e-02, + 1.1659e-04, 7.9274e-05], + [ 1.6081e-04, 2.7800e-04, 4.9877e-04, ..., 2.1763e-03, + 2.7329e-05, 1.4579e-04], + [-5.4779e-03, -7.4863e-05, -8.5983e-03, ..., -5.3749e-03, + -3.3951e-03, -2.4796e-03]], device='cuda:0') +Epoch 467, bias, value: tensor([-0.0542, 0.0247, 0.0080, -0.0115, -0.0029, 0.0074, -0.0158, 0.0355, + -0.0228, 0.0111], device='cuda:0'), grad: tensor([ 0.0050, 0.0356, 0.0075, -0.0204, 0.0262, 0.0080, -0.0126, -0.0494, + 0.0131, -0.0130], device='cuda:0') +100 +1e-05 +changing lr +epoch 466, time 227.34, cls_loss 0.4873 cls_loss_mapping 0.0008 cls_loss_causal 0.4308 re_mapping 0.0062 re_causal 0.0170 /// teacc 99.02 lr 0.00001000 +Epoch 468, weight, value: tensor([[-0.2154, -0.0221, 0.0379, ..., -0.0297, -0.0825, -0.1478], + [-0.0633, -0.1126, 0.0340, ..., 0.0565, -0.0285, -0.0949], + [-0.0671, -0.1016, -0.1206, ..., 0.0430, -0.0386, -0.1182], + ..., + [-0.0693, 0.0343, 0.0263, ..., 0.0165, -0.0589, -0.1000], + [-0.0877, -0.0065, 0.0159, ..., 0.0621, -0.0347, -0.1940], + [ 0.0655, 0.0696, -0.0513, ..., -0.1158, -0.0155, 0.1007]], + device='cuda:0'), grad: tensor([[ 1.2919e-05, 8.2445e-04, 1.1772e-05, ..., -2.7332e-03, + 3.7432e-04, 7.9930e-05], + [ 2.2799e-05, -6.2904e-03, 4.6005e-03, ..., 1.6708e-03, + 5.2750e-05, 7.1600e-06], + [ 6.2525e-05, 7.5150e-04, 7.7561e-06, ..., 1.9875e-03, + 5.9032e-04, 7.1347e-05], + ..., + [ 6.4850e-04, -3.1815e-03, 2.2590e-04, ..., -6.4993e-04, + 2.3289e-03, 6.7616e-04], + [ 1.1406e-03, 6.4964e-03, 1.1641e-04, ..., 2.6054e-03, + 1.4315e-03, 3.0556e-03], + [ 6.1512e-04, 9.0637e-03, 3.2749e-03, ..., 3.0346e-03, + 1.1587e-03, 2.9488e-03]], device='cuda:0') +Epoch 468, bias, value: tensor([-0.0542, 0.0247, 0.0079, -0.0116, -0.0029, 0.0074, -0.0158, 0.0356, + -0.0228, 0.0111], device='cuda:0'), grad: tensor([-0.0378, -0.0207, 0.0238, -0.0049, -0.0223, -0.0082, 0.0204, -0.0026, + 0.0265, 0.0258], device='cuda:0') +100 +1e-05 +changing lr +epoch 467, time 226.06, cls_loss 0.4865 cls_loss_mapping 0.0009 cls_loss_causal 0.4338 re_mapping 0.0059 re_causal 0.0164 /// teacc 98.93 lr 0.00001000 +Epoch 469, weight, value: tensor([[-0.2155, -0.0220, 0.0379, ..., -0.0299, -0.0825, -0.1478], + [-0.0634, -0.1126, 0.0339, ..., 0.0564, -0.0285, -0.0949], + [-0.0673, -0.1017, -0.1207, ..., 0.0429, -0.0387, -0.1184], + ..., + [-0.0693, 0.0342, 0.0264, ..., 0.0165, -0.0589, -0.1000], + [-0.0875, -0.0065, 0.0157, ..., 0.0621, -0.0347, -0.1941], + [ 0.0654, 0.0696, -0.0513, ..., -0.1157, -0.0155, 0.1008]], + device='cuda:0'), grad: tensor([[ 6.3300e-05, 2.4557e-04, 6.5899e-04, ..., 1.9197e-03, + -1.5039e-03, 3.8075e-04], + [ 2.4529e-03, -6.0380e-05, 1.8206e-03, ..., 2.5673e-03, + 1.0616e-04, 8.1778e-05], + [ 1.0138e-03, 9.9301e-05, 1.5507e-03, ..., 6.6948e-03, + 7.0000e-04, 1.3471e-04], + ..., + [ 8.0919e-04, 2.8417e-05, 6.2370e-04, ..., -7.3662e-03, + 1.4722e-04, 4.1318e-04], + [-8.1711e-03, 1.6952e-04, -4.0207e-03, ..., -4.7073e-03, + -2.3499e-03, -3.9062e-03], + [ 2.0103e-03, 1.3523e-05, 1.1444e-03, ..., 5.6076e-03, + 1.7271e-03, 2.4147e-03]], device='cuda:0') +Epoch 469, bias, value: tensor([-0.0542, 0.0247, 0.0079, -0.0114, -0.0030, 0.0073, -0.0159, 0.0357, + -0.0228, 0.0112], device='cuda:0'), grad: tensor([ 0.0123, 0.0185, 0.0277, -0.0190, 0.0126, 0.0012, -0.0160, -0.0507, + -0.0079, 0.0214], device='cuda:0') +100 +1e-05 +changing lr +epoch 468, time 227.71, cls_loss 0.4784 cls_loss_mapping 0.0008 cls_loss_causal 0.4148 re_mapping 0.0062 re_causal 0.0168 /// teacc 98.94 lr 0.00001000 +Epoch 470, weight, value: tensor([[-0.2154, -0.0219, 0.0379, ..., -0.0299, -0.0825, -0.1478], + [-0.0635, -0.1126, 0.0339, ..., 0.0565, -0.0284, -0.0949], + [-0.0672, -0.1017, -0.1208, ..., 0.0428, -0.0387, -0.1184], + ..., + [-0.0692, 0.0343, 0.0263, ..., 0.0167, -0.0588, -0.1000], + [-0.0875, -0.0065, 0.0159, ..., 0.0621, -0.0347, -0.1940], + [ 0.0653, 0.0696, -0.0514, ..., -0.1159, -0.0154, 0.1007]], + device='cuda:0'), grad: tensor([[ 2.9802e-04, 1.4830e-03, 6.6328e-04, ..., 1.5001e-03, + 1.8382e-04, 2.6617e-06], + [ 3.5837e-06, 3.1948e-04, 1.1269e-02, ..., 6.3210e-03, + 4.6349e-04, 1.3858e-06], + [ 4.8615e-06, 4.2558e-05, -2.3973e-04, ..., -1.0519e-03, + -9.5139e-03, 1.6149e-06], + ..., + [-2.6211e-05, 1.5587e-05, 3.7003e-04, ..., 2.7776e-04, + 8.8406e-04, 2.4229e-05], + [-1.0926e-04, -3.7155e-03, 3.9792e-04, ..., 1.2684e-03, + 2.1267e-03, 1.1340e-05], + [ 1.1392e-05, 9.5010e-05, 4.9400e-04, ..., 3.9124e-04, + 1.3399e-03, -4.6343e-05]], device='cuda:0') +Epoch 470, bias, value: tensor([-0.0541, 0.0248, 0.0079, -0.0115, -0.0030, 0.0073, -0.0159, 0.0359, + -0.0229, 0.0109], device='cuda:0'), grad: tensor([ 0.0173, 0.0018, -0.0067, -0.0095, -0.0157, -0.0048, 0.0205, 0.0043, + -0.0144, 0.0072], device='cuda:0') +100 +1e-05 +changing lr +epoch 469, time 226.90, cls_loss 0.4926 cls_loss_mapping 0.0009 cls_loss_causal 0.4318 re_mapping 0.0061 re_causal 0.0163 /// teacc 98.95 lr 0.00001000 +Epoch 471, weight, value: tensor([[-0.2154, -0.0219, 0.0379, ..., -0.0299, -0.0826, -0.1478], + [-0.0634, -0.1128, 0.0338, ..., 0.0564, -0.0284, -0.0950], + [-0.0671, -0.1017, -0.1209, ..., 0.0430, -0.0388, -0.1184], + ..., + [-0.0692, 0.0345, 0.0261, ..., 0.0169, -0.0585, -0.0998], + [-0.0874, -0.0065, 0.0160, ..., 0.0620, -0.0348, -0.1941], + [ 0.0653, 0.0697, -0.0513, ..., -0.1158, -0.0154, 0.1006]], + device='cuda:0'), grad: tensor([[ 2.0653e-05, 7.6675e-04, 1.2398e-05, ..., 7.4196e-04, + 7.3731e-05, 4.3139e-06], + [ 8.8662e-06, 2.6150e-03, 5.8189e-06, ..., 1.1826e-03, + 2.8539e-04, 1.2480e-06], + [-1.8053e-03, 5.1880e-04, -3.0231e-03, ..., -1.2007e-03, + 4.5747e-05, -1.1902e-03], + ..., + [ 1.0023e-03, 4.7684e-04, 1.6602e-02, ..., 1.3561e-03, + 3.8743e-05, 2.4006e-05], + [-4.8828e-04, -6.1684e-03, -1.6678e-02, ..., -3.2196e-03, + -5.9509e-04, 1.2398e-05], + [-8.2922e-04, 7.1287e-04, 1.3971e-04, ..., -4.5806e-05, + 1.0037e-04, -4.9084e-05]], device='cuda:0') +Epoch 471, bias, value: tensor([-0.0541, 0.0248, 0.0081, -0.0115, -0.0031, 0.0072, -0.0160, 0.0360, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([ 0.0101, 0.0072, 0.0009, 0.0228, -0.0199, 0.0054, 0.0127, 0.0137, + -0.0322, -0.0208], device='cuda:0') +100 +1e-05 +changing lr +epoch 470, time 228.15, cls_loss 0.5116 cls_loss_mapping 0.0010 cls_loss_causal 0.4446 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.95 lr 0.00001000 +Epoch 472, weight, value: tensor([[-0.2154, -0.0220, 0.0380, ..., -0.0300, -0.0827, -0.1478], + [-0.0635, -0.1128, 0.0338, ..., 0.0563, -0.0284, -0.0950], + [-0.0671, -0.1018, -0.1210, ..., 0.0430, -0.0386, -0.1184], + ..., + [-0.0692, 0.0345, 0.0261, ..., 0.0169, -0.0585, -0.0999], + [-0.0875, -0.0064, 0.0161, ..., 0.0620, -0.0348, -0.1942], + [ 0.0654, 0.0697, -0.0513, ..., -0.1157, -0.0154, 0.1007]], + device='cuda:0'), grad: tensor([[ 0.0006, 0.0006, 0.0006, ..., 0.0015, 0.0003, 0.0003], + [ 0.0009, -0.0011, 0.0019, ..., -0.0008, -0.0002, 0.0004], + [ 0.0002, 0.0010, 0.0009, ..., -0.0005, 0.0004, 0.0003], + ..., + [ 0.0095, -0.0011, 0.0081, ..., 0.0100, 0.0004, 0.0040], + [ 0.0016, -0.0046, 0.0013, ..., 0.0022, 0.0003, -0.0038], + [-0.0089, 0.0011, 0.0024, ..., -0.0069, 0.0008, -0.0029]], + device='cuda:0') +Epoch 472, bias, value: tensor([-0.0541, 0.0247, 0.0080, -0.0115, -0.0030, 0.0072, -0.0159, 0.0361, + -0.0229, 0.0110], device='cuda:0'), grad: tensor([ 0.0169, -0.0081, 0.0175, 0.0155, -0.0348, -0.0106, 0.0267, 0.0412, + -0.0312, -0.0332], device='cuda:0') +100 +1e-05 +changing lr +epoch 471, time 225.49, cls_loss 0.5160 cls_loss_mapping 0.0010 cls_loss_causal 0.4458 re_mapping 0.0060 re_causal 0.0167 /// teacc 99.01 lr 0.00001000 +Epoch 473, weight, value: tensor([[-0.2153, -0.0220, 0.0381, ..., -0.0300, -0.0827, -0.1477], + [-0.0635, -0.1125, 0.0338, ..., 0.0563, -0.0283, -0.0950], + [-0.0670, -0.1017, -0.1210, ..., 0.0431, -0.0385, -0.1184], + ..., + [-0.0693, 0.0345, 0.0261, ..., 0.0169, -0.0586, -0.1000], + [-0.0876, -0.0065, 0.0160, ..., 0.0619, -0.0347, -0.1943], + [ 0.0655, 0.0696, -0.0513, ..., -0.1158, -0.0155, 0.1009]], + device='cuda:0'), grad: tensor([[ 2.5597e-03, 7.4053e-04, 4.4212e-03, ..., -5.8651e-05, + 3.8981e-04, 1.2426e-03], + [ 1.2279e-04, 6.5231e-04, 4.7588e-04, ..., 3.6201e-03, + 2.0579e-05, 6.6423e-04], + [-8.1024e-03, 3.7169e-04, -1.3885e-02, ..., -1.0109e-02, + -1.0347e-03, -3.6144e-03], + ..., + [-5.3930e-04, 5.3902e-03, -5.7888e-04, ..., -5.3062e-03, + 1.5423e-05, 2.4281e-03], + [ 3.6550e-04, 9.6273e-04, 5.5313e-04, ..., 1.8234e-03, + 5.2631e-05, 3.9077e-04], + [ 2.4471e-03, 8.2474e-03, 4.2343e-03, ..., 4.4518e-03, + 3.0589e-04, 6.5804e-03]], device='cuda:0') +Epoch 473, bias, value: tensor([-0.0541, 0.0248, 0.0081, -0.0116, -0.0030, 0.0073, -0.0159, 0.0361, + -0.0230, 0.0110], device='cuda:0'), grad: tensor([-0.0107, 0.0218, -0.0488, 0.0228, -0.0114, 0.0230, -0.0435, -0.0065, + 0.0126, 0.0406], device='cuda:0') +100 +1e-05 +changing lr +epoch 472, time 228.94, cls_loss 0.4884 cls_loss_mapping 0.0008 cls_loss_causal 0.4264 re_mapping 0.0059 re_causal 0.0160 /// teacc 98.99 lr 0.00001000 +Epoch 474, weight, value: tensor([[-0.2154, -0.0220, 0.0382, ..., -0.0300, -0.0826, -0.1479], + [-0.0636, -0.1125, 0.0336, ..., 0.0564, -0.0283, -0.0952], + [-0.0669, -0.1017, -0.1209, ..., 0.0432, -0.0386, -0.1183], + ..., + [-0.0694, 0.0345, 0.0260, ..., 0.0170, -0.0586, -0.1000], + [-0.0876, -0.0067, 0.0161, ..., 0.0617, -0.0347, -0.1942], + [ 0.0655, 0.0697, -0.0513, ..., -0.1158, -0.0156, 0.1007]], + device='cuda:0'), grad: tensor([[-2.8343e-03, 2.5344e-04, -2.0862e-04, ..., 2.5177e-03, + -4.4708e-03, 6.0558e-04], + [ 4.7356e-05, 3.4422e-05, 3.3212e-04, ..., -2.2125e-03, + 4.7541e-04, 1.0949e-04], + [ 5.7757e-05, 1.8346e-04, 7.4100e-04, ..., 2.4300e-03, + 2.8515e-04, 4.7135e-04], + ..., + [ 5.5850e-05, 5.9545e-05, 6.1691e-06, ..., 2.1286e-03, + 4.8041e-04, 8.6069e-04], + [-4.2582e-04, 1.0669e-04, -3.3875e-03, ..., -8.6308e-04, + 4.1246e-04, 1.1110e-04], + [ 8.3876e-04, 2.3258e-04, 1.1915e-04, ..., -5.9366e-04, + -1.7881e-03, -2.6584e-04]], device='cuda:0') +Epoch 474, bias, value: tensor([-0.0541, 0.0250, 0.0080, -0.0115, -0.0030, 0.0073, -0.0159, 0.0360, + -0.0231, 0.0109], device='cuda:0'), grad: tensor([-0.0074, -0.0127, -0.0072, -0.0103, -0.0070, -0.0065, 0.0412, 0.0178, + 0.0013, -0.0092], device='cuda:0') +100 +1e-05 +changing lr +epoch 473, time 227.47, cls_loss 0.4946 cls_loss_mapping 0.0008 cls_loss_causal 0.4411 re_mapping 0.0060 re_causal 0.0166 /// teacc 99.02 lr 0.00001000 +Epoch 475, weight, value: tensor([[-0.2154, -0.0219, 0.0381, ..., -0.0300, -0.0827, -0.1481], + [-0.0637, -0.1126, 0.0335, ..., 0.0564, -0.0284, -0.0952], + [-0.0669, -0.1016, -0.1210, ..., 0.0433, -0.0387, -0.1184], + ..., + [-0.0694, 0.0345, 0.0259, ..., 0.0170, -0.0585, -0.1000], + [-0.0876, -0.0068, 0.0161, ..., 0.0619, -0.0348, -0.1943], + [ 0.0656, 0.0699, -0.0512, ..., -0.1158, -0.0156, 0.1008]], + device='cuda:0'), grad: tensor([[-1.5460e-06, 1.9327e-05, 4.7624e-05, ..., -2.7943e-03, + 8.1956e-06, -3.5793e-05], + [ 9.1866e-06, 5.1117e-04, 6.1572e-05, ..., 2.4624e-03, + 4.5411e-06, 3.7938e-05], + [ 5.8636e-06, 1.5378e-04, 6.6698e-05, ..., 2.3766e-03, + 5.9791e-07, 2.6792e-05], + ..., + [ 9.6336e-06, -4.0932e-03, 7.6592e-05, ..., 2.2869e-03, + -5.0478e-07, -1.1367e-04], + [ 4.8399e-05, 2.3103e-04, 1.4043e-04, ..., 6.5041e-04, + 2.2605e-05, 7.7844e-05], + [ 2.0973e-06, 1.3523e-03, 1.2434e-04, ..., 2.1706e-03, + 6.6571e-06, 5.4002e-05]], device='cuda:0') +Epoch 475, bias, value: tensor([-0.0542, 0.0249, 0.0083, -0.0115, -0.0030, 0.0072, -0.0159, 0.0361, + -0.0230, 0.0108], device='cuda:0'), grad: tensor([-0.0096, 0.0261, 0.0223, -0.0419, -0.0031, -0.0432, 0.0213, 0.0145, + -0.0104, 0.0240], device='cuda:0') +100 +1e-05 +changing lr +epoch 474, time 227.98, cls_loss 0.4698 cls_loss_mapping 0.0008 cls_loss_causal 0.4120 re_mapping 0.0063 re_causal 0.0167 /// teacc 99.00 lr 0.00001000 +Epoch 476, weight, value: tensor([[-0.2153, -0.0219, 0.0381, ..., -0.0301, -0.0825, -0.1480], + [-0.0637, -0.1126, 0.0335, ..., 0.0564, -0.0285, -0.0952], + [-0.0668, -0.1016, -0.1210, ..., 0.0433, -0.0387, -0.1183], + ..., + [-0.0693, 0.0347, 0.0259, ..., 0.0170, -0.0584, -0.1000], + [-0.0876, -0.0068, 0.0162, ..., 0.0619, -0.0348, -0.1943], + [ 0.0655, 0.0698, -0.0513, ..., -0.1157, -0.0156, 0.1009]], + device='cuda:0'), grad: tensor([[ 1.3340e-04, 7.9095e-05, 3.9244e-04, ..., -2.2354e-03, + 1.5390e-04, 4.9233e-05], + [ 1.8284e-05, 5.3358e-04, 1.7548e-03, ..., 4.0131e-03, + 2.9945e-04, 5.8681e-05], + [ 7.0810e-05, 4.8518e-05, 1.6499e-03, ..., -1.4372e-03, + 5.4359e-04, 2.2936e-04], + ..., + [ 5.8794e-04, 1.1158e-03, 9.2697e-04, ..., -2.5883e-03, + 5.0402e-04, 1.8191e-04], + [ 6.8426e-04, 4.5598e-05, 8.0395e-04, ..., 2.1191e-03, + 2.3437e-04, 6.9678e-05], + [-2.3327e-03, 1.4191e-03, 1.3056e-03, ..., 2.9716e-03, + 8.9550e-04, 5.2005e-05]], device='cuda:0') +Epoch 476, bias, value: tensor([-0.0541, 0.0248, 0.0083, -0.0115, -0.0031, 0.0073, -0.0160, 0.0361, + -0.0230, 0.0108], device='cuda:0'), grad: tensor([-0.0168, 0.0285, -0.0084, -0.0020, -0.0268, -0.0115, 0.0178, -0.0375, + 0.0212, 0.0355], device='cuda:0') +100 +1e-05 +changing lr +epoch 475, time 226.88, cls_loss 0.4901 cls_loss_mapping 0.0007 cls_loss_causal 0.4339 re_mapping 0.0063 re_causal 0.0170 /// teacc 99.02 lr 0.00001000 +Epoch 477, weight, value: tensor([[-0.2153, -0.0219, 0.0383, ..., -0.0301, -0.0825, -0.1479], + [-0.0638, -0.1126, 0.0337, ..., 0.0565, -0.0284, -0.0952], + [-0.0667, -0.1014, -0.1210, ..., 0.0433, -0.0386, -0.1184], + ..., + [-0.0694, 0.0347, 0.0258, ..., 0.0170, -0.0584, -0.1000], + [-0.0876, -0.0068, 0.0162, ..., 0.0619, -0.0348, -0.1945], + [ 0.0656, 0.0698, -0.0514, ..., -0.1157, -0.0157, 0.1011]], + device='cuda:0'), grad: tensor([[ 0.0005, 0.0001, -0.0082, ..., -0.0002, 0.0014, 0.0003], + [ 0.0001, 0.0002, 0.0040, ..., 0.0051, 0.0008, 0.0002], + [-0.0019, 0.0002, -0.0016, ..., -0.0173, -0.0018, 0.0004], + ..., + [ 0.0007, 0.0002, 0.0013, ..., 0.0035, 0.0009, 0.0005], + [-0.0014, -0.0030, -0.0020, ..., -0.0039, 0.0003, -0.0037], + [ 0.0041, 0.0025, 0.0052, ..., 0.0089, 0.0024, 0.0037]], + device='cuda:0') +Epoch 477, bias, value: tensor([-0.0542, 0.0248, 0.0084, -0.0115, -0.0032, 0.0073, -0.0161, 0.0362, + -0.0230, 0.0109], device='cuda:0'), grad: tensor([-0.0055, 0.0312, -0.0581, 0.0152, 0.0123, 0.0021, -0.0227, 0.0213, + -0.0496, 0.0539], device='cuda:0') +100 +1e-05 +changing lr +epoch 476, time 227.97, cls_loss 0.4910 cls_loss_mapping 0.0009 cls_loss_causal 0.4372 re_mapping 0.0061 re_causal 0.0162 /// teacc 99.01 lr 0.00001000 +Epoch 478, weight, value: tensor([[-0.2152, -0.0219, 0.0383, ..., -0.0301, -0.0825, -0.1478], + [-0.0638, -0.1126, 0.0337, ..., 0.0565, -0.0284, -0.0953], + [-0.0667, -0.1013, -0.1210, ..., 0.0433, -0.0385, -0.1184], + ..., + [-0.0693, 0.0346, 0.0257, ..., 0.0169, -0.0584, -0.1000], + [-0.0876, -0.0066, 0.0162, ..., 0.0620, -0.0348, -0.1946], + [ 0.0657, 0.0697, -0.0514, ..., -0.1158, -0.0158, 0.1012]], + device='cuda:0'), grad: tensor([[ 2.6390e-05, 2.0909e-04, 1.5335e-03, ..., 2.8896e-03, + 1.9550e-03, 9.1363e-07], + [ 6.2346e-05, 9.5367e-04, 1.8244e-03, ..., 1.0605e-03, + 1.1969e-03, 1.3351e-05], + [ 1.6010e-04, 2.8858e-03, 1.4400e-03, ..., 3.4504e-03, + 8.8882e-04, 3.0661e-04], + ..., + [ 5.0277e-05, 8.9407e-04, 1.9779e-03, ..., 3.1261e-03, + 1.1921e-03, 2.9969e-04], + [ 2.8467e-04, 3.3207e-03, 3.0327e-03, ..., 5.6114e-03, + -8.1205e-04, 3.1888e-05], + [ 1.3232e-04, 7.0620e-04, -1.8015e-03, ..., -6.6414e-03, + 1.3237e-03, 1.9562e-04]], device='cuda:0') +Epoch 478, bias, value: tensor([-0.0541, 0.0248, 0.0084, -0.0115, -0.0032, 0.0072, -0.0161, 0.0362, + -0.0229, 0.0107], device='cuda:0'), grad: tensor([ 0.0285, 0.0249, 0.0323, -0.0254, -0.0306, -0.0063, -0.0341, 0.0302, + 0.0102, -0.0298], device='cuda:0') +100 +1e-05 +changing lr +epoch 477, time 226.00, cls_loss 0.5074 cls_loss_mapping 0.0009 cls_loss_causal 0.4451 re_mapping 0.0060 re_causal 0.0165 /// teacc 99.02 lr 0.00001000 +Epoch 479, weight, value: tensor([[-0.2152, -0.0220, 0.0382, ..., -0.0302, -0.0825, -0.1477], + [-0.0638, -0.1127, 0.0336, ..., 0.0564, -0.0285, -0.0952], + [-0.0668, -0.1014, -0.1210, ..., 0.0434, -0.0386, -0.1184], + ..., + [-0.0694, 0.0346, 0.0256, ..., 0.0169, -0.0582, -0.1002], + [-0.0877, -0.0067, 0.0165, ..., 0.0621, -0.0348, -0.1946], + [ 0.0657, 0.0698, -0.0513, ..., -0.1158, -0.0158, 0.1012]], + device='cuda:0'), grad: tensor([[ 2.9800e-02, -8.2970e-05, 5.9223e-04, ..., 1.8721e-03, + 1.8692e-03, 2.1124e-04], + [ 1.2827e-04, 8.5160e-06, -3.8376e-03, ..., -4.5013e-03, + -7.0076e-03, 5.7459e-05], + [ 2.3890e-04, 4.2379e-05, 7.7152e-04, ..., 2.3022e-03, + 1.9550e-03, 1.7810e-04], + ..., + [ 8.4639e-04, 9.1195e-05, 1.8539e-03, ..., 2.7504e-03, + 1.6832e-03, 6.9857e-04], + [ 4.8208e-04, 1.3232e-04, 7.9679e-04, ..., 2.7523e-03, + 2.0294e-03, 4.0388e-04], + [-2.3251e-03, -1.1696e-02, -2.3270e-03, ..., -6.2790e-03, + 1.5669e-03, -1.4515e-03]], device='cuda:0') +Epoch 479, bias, value: tensor([-0.0541, 0.0248, 0.0083, -0.0116, -0.0032, 0.0073, -0.0161, 0.0362, + -0.0227, 0.0107], device='cuda:0'), grad: tensor([ 0.0413, -0.0237, 0.0099, 0.0105, 0.0091, -0.0178, -0.0220, 0.0088, + 0.0120, -0.0282], device='cuda:0') +100 +1e-05 +changing lr +epoch 478, time 227.60, cls_loss 0.4372 cls_loss_mapping 0.0007 cls_loss_causal 0.3820 re_mapping 0.0061 re_causal 0.0160 /// teacc 99.01 lr 0.00001000 +Epoch 480, weight, value: tensor([[-0.2152, -0.0218, 0.0382, ..., -0.0299, -0.0824, -0.1474], + [-0.0638, -0.1128, 0.0335, ..., 0.0564, -0.0285, -0.0951], + [-0.0668, -0.1011, -0.1209, ..., 0.0434, -0.0386, -0.1184], + ..., + [-0.0695, 0.0346, 0.0256, ..., 0.0168, -0.0582, -0.1001], + [-0.0878, -0.0068, 0.0164, ..., 0.0620, -0.0348, -0.1946], + [ 0.0658, 0.0697, -0.0513, ..., -0.1158, -0.0158, 0.1012]], + device='cuda:0'), grad: tensor([[ 6.4969e-05, 3.2410e-06, -2.6369e-04, ..., 5.3501e-04, + -1.2527e-02, 4.9382e-05], + [ 4.6396e-04, 1.2672e-04, 8.0585e-04, ..., 2.5082e-03, + 6.4932e-06, 4.3321e-04], + [ 8.4972e-04, 6.2227e-05, 1.4744e-03, ..., 3.6564e-03, + 3.8052e-04, 6.9761e-04], + ..., + [ 7.8440e-04, 2.0158e-04, 1.3580e-03, ..., 4.4861e-03, + 2.0838e-04, 6.2704e-04], + [ 5.3263e-04, 4.5627e-05, 3.6097e-04, ..., 2.0714e-03, + 1.4734e-03, 6.7759e-04], + [-2.8872e-04, -1.3895e-03, -3.4475e-04, ..., -1.4839e-03, + -6.5041e-04, -1.4381e-03]], device='cuda:0') +Epoch 480, bias, value: tensor([-0.0539, 0.0248, 0.0084, -0.0116, -0.0032, 0.0074, -0.0162, 0.0361, + -0.0229, 0.0108], device='cuda:0'), grad: tensor([-0.0367, 0.0050, 0.0314, -0.0147, -0.0388, -0.0072, 0.0112, 0.0373, + 0.0249, -0.0123], device='cuda:0') +100 +1e-05 +changing lr +epoch 479, time 227.06, cls_loss 0.4997 cls_loss_mapping 0.0008 cls_loss_causal 0.4289 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.97 lr 0.00001000 +Epoch 481, weight, value: tensor([[-0.2152, -0.0219, 0.0382, ..., -0.0299, -0.0822, -0.1474], + [-0.0639, -0.1129, 0.0333, ..., 0.0563, -0.0286, -0.0953], + [-0.0669, -0.1011, -0.1209, ..., 0.0434, -0.0386, -0.1184], + ..., + [-0.0692, 0.0348, 0.0258, ..., 0.0168, -0.0582, -0.1002], + [-0.0878, -0.0068, 0.0165, ..., 0.0620, -0.0347, -0.1946], + [ 0.0655, 0.0696, -0.0514, ..., -0.1159, -0.0158, 0.1011]], + device='cuda:0'), grad: tensor([[ 7.2193e-04, 5.7793e-04, 1.5497e-03, ..., 3.0041e-03, + 1.3580e-03, 2.5809e-05], + [ 3.3474e-04, 3.8195e-04, 3.2234e-03, ..., 4.0512e-03, + 4.5586e-03, 7.4431e-06], + [-6.4194e-05, 8.3447e-04, 7.9441e-04, ..., 1.8883e-03, + 1.2884e-03, 2.8685e-05], + ..., + [ 3.8290e-04, -2.4323e-02, 1.5755e-03, ..., 2.6455e-03, + -6.1302e-03, 7.5102e-05], + [ 4.0078e-04, 8.0948e-03, 3.8713e-05, ..., -1.3161e-02, + 7.3528e-04, 1.1673e-03], + [-7.2861e-04, 1.3466e-03, -9.2239e-03, ..., -1.5364e-03, + -4.6806e-03, 3.3474e-04]], device='cuda:0') +Epoch 481, bias, value: tensor([-0.0540, 0.0247, 0.0082, -0.0115, -0.0031, 0.0073, -0.0161, 0.0362, + -0.0228, 0.0107], device='cuda:0'), grad: tensor([ 0.0238, 0.0376, 0.0190, -0.0381, 0.0193, 0.0396, -0.0048, -0.0374, + -0.0498, -0.0093], device='cuda:0') +100 +1e-05 +changing lr +epoch 480, time 226.81, cls_loss 0.4787 cls_loss_mapping 0.0009 cls_loss_causal 0.4195 re_mapping 0.0060 re_causal 0.0162 /// teacc 98.98 lr 0.00001000 +Epoch 482, weight, value: tensor([[-0.2153, -0.0219, 0.0381, ..., -0.0299, -0.0823, -0.1475], + [-0.0638, -0.1129, 0.0334, ..., 0.0564, -0.0286, -0.0952], + [-0.0670, -0.1012, -0.1208, ..., 0.0433, -0.0388, -0.1185], + ..., + [-0.0694, 0.0347, 0.0256, ..., 0.0167, -0.0582, -0.1002], + [-0.0877, -0.0069, 0.0165, ..., 0.0621, -0.0346, -0.1947], + [ 0.0656, 0.0697, -0.0513, ..., -0.1159, -0.0157, 0.1013]], + device='cuda:0'), grad: tensor([[-5.5403e-05, 9.6858e-05, 1.5664e-04, ..., -2.4490e-03, + 4.1485e-04, 1.7866e-05], + [ 1.3337e-05, 4.9925e-04, 5.4073e-04, ..., 2.9945e-03, + 7.4744e-05, 1.5751e-05], + [ 4.0025e-05, -1.3256e-03, -2.4509e-03, ..., -9.8190e-03, + -5.6982e-04, 5.6922e-05], + ..., + [-1.5393e-05, -5.3835e-04, -2.6250e-04, ..., -2.9469e-04, + 7.5996e-05, -9.3281e-05], + [ 3.9905e-05, 4.4608e-04, 8.3303e-04, ..., 3.7251e-03, + 7.2956e-05, 4.9621e-05], + [ 2.2009e-05, 3.0828e-04, 3.1877e-04, ..., 1.8253e-03, + 6.8605e-05, 5.0455e-05]], device='cuda:0') +Epoch 482, bias, value: tensor([-0.0541, 0.0247, 0.0082, -0.0115, -0.0031, 0.0073, -0.0162, 0.0361, + -0.0227, 0.0107], device='cuda:0'), grad: tensor([-0.0179, 0.0239, -0.0371, 0.0153, 0.0166, -0.0483, 0.0290, 0.0074, + -0.0051, 0.0162], device='cuda:0') +100 +1e-05 +changing lr +epoch 481, time 226.36, cls_loss 0.4845 cls_loss_mapping 0.0009 cls_loss_causal 0.4230 re_mapping 0.0058 re_causal 0.0154 /// teacc 99.02 lr 0.00001000 +Epoch 483, weight, value: tensor([[-0.2154, -0.0219, 0.0381, ..., -0.0300, -0.0823, -0.1476], + [-0.0638, -0.1127, 0.0335, ..., 0.0563, -0.0286, -0.0952], + [-0.0670, -0.1012, -0.1207, ..., 0.0432, -0.0389, -0.1185], + ..., + [-0.0692, 0.0346, 0.0256, ..., 0.0166, -0.0582, -0.1002], + [-0.0877, -0.0069, 0.0165, ..., 0.0622, -0.0345, -0.1947], + [ 0.0656, 0.0699, -0.0514, ..., -0.1159, -0.0157, 0.1013]], + device='cuda:0'), grad: tensor([[-6.5536e-03, -5.5656e-03, -5.9624e-03, ..., -8.8196e-03, + -3.0212e-03, -5.0634e-05], + [ 2.5368e-04, -1.2426e-03, 1.0920e-04, ..., 8.6355e-04, + 1.3542e-04, -1.4023e-02], + [ 4.3449e-03, 2.5082e-03, 3.9177e-03, ..., 6.0310e-03, + 6.2523e-03, -1.0586e-03], + ..., + [ 2.4357e-03, 9.5558e-04, 9.4366e-04, ..., 2.7390e-03, + -4.3488e-03, 3.7909e-04], + [ 3.6192e-04, 2.6894e-04, 1.4818e-04, ..., 7.5197e-04, + 2.3711e-04, 1.3232e-04], + [-2.8248e-03, -5.2757e-03, -9.4473e-06, ..., -5.6744e-04, + -2.2352e-05, -5.9652e-04]], device='cuda:0') +Epoch 483, bias, value: tensor([-0.0542, 0.0248, 0.0082, -0.0114, -0.0032, 0.0073, -0.0161, 0.0360, + -0.0226, 0.0107], device='cuda:0'), grad: tensor([-0.0252, -0.0240, 0.0242, 0.0161, 0.0103, 0.0068, -0.0212, 0.0072, + 0.0108, -0.0048], device='cuda:0') +100 +1e-05 +changing lr +epoch 482, time 225.15, cls_loss 0.5050 cls_loss_mapping 0.0010 cls_loss_causal 0.4473 re_mapping 0.0058 re_causal 0.0159 /// teacc 98.97 lr 0.00001000 +Epoch 484, weight, value: tensor([[-0.2154, -0.0218, 0.0381, ..., -0.0299, -0.0823, -0.1477], + [-0.0638, -0.1127, 0.0333, ..., 0.0562, -0.0287, -0.0951], + [-0.0669, -0.1013, -0.1208, ..., 0.0431, -0.0391, -0.1185], + ..., + [-0.0693, 0.0347, 0.0256, ..., 0.0166, -0.0581, -0.1000], + [-0.0877, -0.0068, 0.0165, ..., 0.0623, -0.0345, -0.1946], + [ 0.0657, 0.0698, -0.0513, ..., -0.1159, -0.0158, 0.1011]], + device='cuda:0'), grad: tensor([[ 4.8466e-06, 1.6499e-04, 1.3173e-04, ..., 1.8349e-03, + 0.0000e+00, 1.4877e-04], + [ 4.8637e-04, -3.0303e-04, 2.0943e-03, ..., -1.0040e-02, + 0.0000e+00, 1.7524e-04], + [ 2.3171e-05, 4.4078e-05, 1.1504e-04, ..., -4.4560e-04, + 0.0000e+00, 2.2411e-05], + ..., + [ 6.8069e-05, -9.9564e-04, -1.3952e-03, ..., -2.4166e-03, + 0.0000e+00, -1.3275e-03], + [ 9.4175e-04, 1.8263e-04, -3.7223e-05, ..., 2.0161e-03, + 0.0000e+00, 1.6403e-04], + [ 2.9489e-05, -3.8028e-04, 4.5943e-04, ..., 1.5221e-03, + 0.0000e+00, -3.5787e-04]], device='cuda:0') +Epoch 484, bias, value: tensor([-0.0542, 0.0247, 0.0081, -0.0116, -0.0033, 0.0075, -0.0160, 0.0361, + -0.0225, 0.0107], device='cuda:0'), grad: tensor([ 0.0163, -0.0174, -0.0490, 0.0123, 0.0046, -0.0159, 0.0299, -0.0180, + 0.0254, 0.0118], device='cuda:0') +100 +1e-05 +changing lr +epoch 483, time 226.06, cls_loss 0.4818 cls_loss_mapping 0.0010 cls_loss_causal 0.4255 re_mapping 0.0060 re_causal 0.0158 /// teacc 99.00 lr 0.00001000 +Epoch 485, weight, value: tensor([[-0.2155, -0.0217, 0.0381, ..., -0.0299, -0.0824, -0.1476], + [-0.0639, -0.1127, 0.0332, ..., 0.0561, -0.0287, -0.0951], + [-0.0668, -0.1013, -0.1208, ..., 0.0431, -0.0391, -0.1186], + ..., + [-0.0693, 0.0349, 0.0256, ..., 0.0166, -0.0582, -0.1002], + [-0.0878, -0.0069, 0.0166, ..., 0.0624, -0.0344, -0.1946], + [ 0.0657, 0.0696, -0.0514, ..., -0.1160, -0.0156, 0.1011]], + device='cuda:0'), grad: tensor([[-5.0515e-05, -5.0354e-04, -1.1659e-04, ..., 5.1451e-04, + 5.0449e-04, -4.1127e-06], + [ 3.1173e-05, 5.7526e-03, 1.9357e-05, ..., 2.2469e-03, + -4.1783e-05, 3.1084e-05], + [ 2.3082e-05, 2.4629e-04, 2.5010e-04, ..., -2.4033e-03, + -3.7441e-03, 1.9267e-05], + ..., + [-6.4135e-04, -7.5302e-03, 3.4906e-06, ..., -5.4455e-04, + 5.8699e-04, -2.5215e-03], + [ 2.4586e-03, 9.6607e-04, -3.0804e-04, ..., -3.0613e-03, + 8.7023e-04, 1.4601e-03], + [ 1.0319e-03, 1.0437e-04, 2.0079e-06, ..., 6.1560e-04, + 4.3106e-04, 6.5899e-04]], device='cuda:0') +Epoch 485, bias, value: tensor([-0.0542, 0.0247, 0.0081, -0.0115, -0.0032, 0.0074, -0.0159, 0.0360, + -0.0225, 0.0106], device='cuda:0'), grad: tensor([ 0.0080, 0.0083, -0.0233, 0.0016, 0.0066, 0.0446, -0.0200, -0.0214, + -0.0135, 0.0092], device='cuda:0') +100 +1e-05 +changing lr +epoch 484, time 228.50, cls_loss 0.4842 cls_loss_mapping 0.0009 cls_loss_causal 0.4235 re_mapping 0.0060 re_causal 0.0163 /// teacc 99.00 lr 0.00001000 +Epoch 486, weight, value: tensor([[-0.2157, -0.0218, 0.0382, ..., -0.0299, -0.0824, -0.1476], + [-0.0639, -0.1127, 0.0332, ..., 0.0561, -0.0287, -0.0951], + [-0.0668, -0.1013, -0.1208, ..., 0.0431, -0.0390, -0.1187], + ..., + [-0.0694, 0.0349, 0.0256, ..., 0.0165, -0.0581, -0.1002], + [-0.0878, -0.0069, 0.0166, ..., 0.0625, -0.0346, -0.1946], + [ 0.0659, 0.0696, -0.0513, ..., -0.1160, -0.0155, 0.1012]], + device='cuda:0'), grad: tensor([[ 1.5211e-04, 2.4486e-04, 2.7227e-04, ..., 1.7385e-03, + 5.2154e-08, 7.6771e-05], + [ 2.5749e-04, -1.6570e-04, -2.9011e-03, ..., -9.0103e-03, + 6.1467e-08, 2.2995e-04], + [ 1.2274e-03, 3.4785e-04, 5.5933e-04, ..., 3.0003e-03, + 1.7509e-05, 3.3951e-04], + ..., + [ 1.1593e-04, 7.8773e-04, 1.0246e-04, ..., 2.3460e-03, + 1.6037e-06, 7.4577e-04], + [-9.8991e-04, 3.5906e-04, 5.3596e-04, ..., 2.8267e-03, + 2.9039e-06, 2.4986e-04], + [ 9.5654e-04, 1.3456e-03, 3.6049e-04, ..., 3.8204e-03, + 1.1921e-06, 1.0614e-03]], device='cuda:0') +Epoch 486, bias, value: tensor([-0.0542, 0.0248, 0.0082, -0.0116, -0.0032, 0.0075, -0.0160, 0.0360, + -0.0226, 0.0106], device='cuda:0'), grad: tensor([ 0.0085, -0.0431, 0.0142, -0.0309, -0.0051, 0.0103, 0.0120, 0.0097, + 0.0104, 0.0139], device='cuda:0') +100 +1e-05 +changing lr +epoch 485, time 227.13, cls_loss 0.4446 cls_loss_mapping 0.0007 cls_loss_causal 0.3859 re_mapping 0.0060 re_causal 0.0159 /// teacc 99.02 lr 0.00001000 +Epoch 487, weight, value: tensor([[-0.2157, -0.0218, 0.0383, ..., -0.0298, -0.0823, -0.1476], + [-0.0639, -0.1126, 0.0333, ..., 0.0561, -0.0287, -0.0950], + [-0.0668, -0.1013, -0.1209, ..., 0.0434, -0.0391, -0.1188], + ..., + [-0.0695, 0.0350, 0.0254, ..., 0.0163, -0.0580, -0.1002], + [-0.0878, -0.0070, 0.0166, ..., 0.0624, -0.0346, -0.1947], + [ 0.0661, 0.0696, -0.0513, ..., -0.1159, -0.0154, 0.1013]], + device='cuda:0'), grad: tensor([[ 4.2021e-06, 1.4722e-04, 2.9993e-04, ..., 9.4032e-04, + 2.6226e-05, 1.0088e-05], + [ 5.4270e-05, 2.0847e-03, 1.5473e-04, ..., -3.2330e-04, + 1.0830e-04, 4.7088e-05], + [ 2.4274e-05, 2.7828e-03, 7.9679e-04, ..., 7.6599e-03, + 1.0967e-03, 2.5558e-04], + ..., + [ 4.3488e-04, -3.3813e-02, 6.8903e-04, ..., 1.1196e-03, + 4.7278e-04, 2.4021e-04], + [-1.2047e-02, 1.3447e-04, -1.3329e-02, ..., -8.5983e-03, + -1.1238e-02, 1.9968e-04], + [-1.1053e-03, 4.5967e-03, 2.3365e-05, ..., 5.1737e-04, + 7.8964e-04, -5.3167e-04]], device='cuda:0') +Epoch 487, bias, value: tensor([-0.0541, 0.0248, 0.0082, -0.0117, -0.0032, 0.0074, -0.0160, 0.0360, + -0.0227, 0.0108], device='cuda:0'), grad: tensor([ 0.0103, -0.0199, 0.0408, -0.0252, 0.0583, 0.0080, -0.0216, -0.0103, + -0.0507, 0.0103], device='cuda:0') +100 +1e-05 +changing lr +epoch 486, time 225.08, cls_loss 0.4685 cls_loss_mapping 0.0008 cls_loss_causal 0.4113 re_mapping 0.0059 re_causal 0.0155 /// teacc 99.04 lr 0.00001000 +Epoch 488, weight, value: tensor([[-0.2157, -0.0219, 0.0383, ..., -0.0298, -0.0824, -0.1476], + [-0.0638, -0.1126, 0.0335, ..., 0.0563, -0.0288, -0.0950], + [-0.0669, -0.1012, -0.1209, ..., 0.0434, -0.0390, -0.1189], + ..., + [-0.0694, 0.0349, 0.0253, ..., 0.0162, -0.0581, -0.1002], + [-0.0877, -0.0069, 0.0167, ..., 0.0624, -0.0345, -0.1947], + [ 0.0661, 0.0694, -0.0514, ..., -0.1159, -0.0154, 0.1014]], + device='cuda:0'), grad: tensor([[ 2.3618e-05, 1.7824e-03, -2.7218e-03, ..., 1.5044e-04, + -6.4182e-04, -3.4046e-04], + [ 1.4901e-05, 7.2908e-04, 4.3869e-04, ..., 7.4883e-03, + 4.4179e-04, 2.7299e-04], + [ 1.4448e-03, 9.9277e-04, 3.4285e-04, ..., 2.1553e-03, + 4.0007e-04, 5.6648e-04], + ..., + [-1.1988e-05, 4.4006e-02, -6.3598e-05, ..., 3.0212e-03, + 4.9543e-04, 1.9073e-03], + [ 7.2622e-04, -1.5144e-03, 4.7350e-04, ..., -8.9035e-03, + -5.4884e-04, -2.0409e-03], + [-2.3518e-03, 6.0730e-03, 2.9683e-04, ..., -2.2640e-03, + 3.8195e-04, -3.7060e-03]], device='cuda:0') +Epoch 488, bias, value: tensor([-0.0541, 0.0249, 0.0084, -0.0117, -0.0032, 0.0074, -0.0159, 0.0358, + -0.0227, 0.0107], device='cuda:0'), grad: tensor([-0.0253, 0.0146, 0.0121, 0.0101, -0.0132, -0.0008, 0.0093, 0.0272, + -0.0306, -0.0034], device='cuda:0') +100 +1e-05 +changing lr +epoch 487, time 225.52, cls_loss 0.4716 cls_loss_mapping 0.0008 cls_loss_causal 0.4157 re_mapping 0.0058 re_causal 0.0158 /// teacc 99.03 lr 0.00001000 +Epoch 489, weight, value: tensor([[-0.2158, -0.0220, 0.0382, ..., -0.0298, -0.0823, -0.1476], + [-0.0636, -0.1126, 0.0334, ..., 0.0563, -0.0288, -0.0949], + [-0.0670, -0.1013, -0.1209, ..., 0.0434, -0.0393, -0.1188], + ..., + [-0.0694, 0.0350, 0.0254, ..., 0.0162, -0.0578, -0.1002], + [-0.0876, -0.0068, 0.0167, ..., 0.0624, -0.0347, -0.1946], + [ 0.0661, 0.0695, -0.0513, ..., -0.1159, -0.0155, 0.1014]], + device='cuda:0'), grad: tensor([[ 2.5965e-06, 5.3406e-04, 5.0887e-06, ..., 7.8154e-04, + 4.1395e-05, 1.7321e-04], + [ 8.7172e-07, -5.6877e-03, 7.7200e-04, ..., 9.4843e-04, + 6.0387e-06, 2.3320e-05], + [ 3.5614e-06, 1.4210e-03, 1.7717e-05, ..., -1.6356e-03, + 2.7448e-05, 1.0401e-04], + ..., + [ 6.7377e-04, -1.6797e-04, 6.6102e-05, ..., 4.5848e-04, + 1.5333e-05, 5.1689e-04], + [ 1.1683e-04, -3.3607e-03, 2.3276e-05, ..., 5.0497e-04, + 1.7750e-04, 7.5436e-04], + [-1.1311e-03, 1.6441e-03, 6.5982e-05, ..., 1.3399e-03, + 2.1541e-04, 4.3690e-05]], device='cuda:0') +Epoch 489, bias, value: tensor([-0.0542, 0.0248, 0.0084, -0.0117, -0.0033, 0.0074, -0.0159, 0.0359, + -0.0226, 0.0107], device='cuda:0'), grad: tensor([ 0.0073, -0.0059, -0.0203, 0.0095, 0.0204, -0.0045, -0.0210, 0.0078, + -0.0048, 0.0116], device='cuda:0') +100 +1e-05 +changing lr +epoch 488, time 225.64, cls_loss 0.5085 cls_loss_mapping 0.0009 cls_loss_causal 0.4471 re_mapping 0.0061 re_causal 0.0167 /// teacc 99.02 lr 0.00001000 +Epoch 490, weight, value: tensor([[-0.2158, -0.0220, 0.0382, ..., -0.0297, -0.0823, -0.1475], + [-0.0636, -0.1128, 0.0333, ..., 0.0563, -0.0289, -0.0948], + [-0.0671, -0.1013, -0.1208, ..., 0.0435, -0.0393, -0.1187], + ..., + [-0.0693, 0.0350, 0.0256, ..., 0.0162, -0.0578, -0.1001], + [-0.0875, -0.0068, 0.0167, ..., 0.0624, -0.0347, -0.1948], + [ 0.0662, 0.0693, -0.0512, ..., -0.1160, -0.0157, 0.1013]], + device='cuda:0'), grad: tensor([[ 1.1012e-05, 4.3917e-04, 9.7275e-05, ..., 2.3575e-03, + 9.5591e-06, 7.9334e-05], + [ 2.6114e-06, 1.4782e-03, 4.2510e-04, ..., 4.5433e-03, + 2.8276e-04, 1.3578e-04], + [-6.3144e-06, -1.3933e-03, 1.6842e-03, ..., 2.7924e-03, + 4.1056e-04, 4.0078e-04], + ..., + [-1.2673e-05, 4.1428e-03, 5.4073e-04, ..., 6.8398e-03, + 1.0881e-03, -5.4207e-03], + [ 7.3016e-06, -7.6981e-03, 2.1982e-04, ..., -7.7019e-03, + -3.1528e-03, 1.7524e-04], + [ 1.0289e-05, 1.8489e-04, -6.2943e-03, ..., -2.1835e-02, + 6.0707e-05, -1.6856e-04]], device='cuda:0') +Epoch 490, bias, value: tensor([-0.0542, 0.0248, 0.0084, -0.0117, -0.0033, 0.0073, -0.0158, 0.0360, + -0.0226, 0.0107], device='cuda:0'), grad: tensor([ 0.0200, 0.0291, 0.0146, 0.0321, 0.0288, 0.0183, -0.0146, 0.0081, + -0.0341, -0.1022], device='cuda:0') +100 +1e-05 +changing lr +epoch 489, time 228.35, cls_loss 0.4906 cls_loss_mapping 0.0008 cls_loss_causal 0.4304 re_mapping 0.0061 re_causal 0.0163 /// teacc 99.02 lr 0.00001000 +Epoch 491, weight, value: tensor([[-0.2158, -0.0221, 0.0383, ..., -0.0297, -0.0823, -0.1476], + [-0.0635, -0.1129, 0.0332, ..., 0.0564, -0.0287, -0.0948], + [-0.0671, -0.1014, -0.1207, ..., 0.0434, -0.0392, -0.1187], + ..., + [-0.0694, 0.0350, 0.0256, ..., 0.0161, -0.0579, -0.1000], + [-0.0876, -0.0068, 0.0167, ..., 0.0623, -0.0347, -0.1947], + [ 0.0663, 0.0694, -0.0512, ..., -0.1158, -0.0157, 0.1013]], + device='cuda:0'), grad: tensor([[-1.2051e-06, 1.3054e-02, 2.4128e-04, ..., 9.2745e-04, + 4.2868e-04, 8.1211e-06], + [ 1.3806e-05, 3.8862e-04, 1.3323e-03, ..., 1.8892e-03, + 1.1148e-03, 2.4438e-06], + [-6.4135e-05, -3.1586e-02, 1.4997e-04, ..., -4.9400e-03, + 5.3024e-04, 2.6450e-06], + ..., + [ 1.1571e-05, 2.4509e-04, 3.8671e-04, ..., 1.5831e-03, + 6.4135e-04, 1.2942e-05], + [ 5.3234e-06, 1.4725e-02, 4.3559e-04, ..., 9.0218e-04, + 4.2725e-04, 4.0270e-06], + [ 8.5160e-06, 3.0565e-04, -2.3479e-03, ..., 1.2331e-03, + 6.4516e-04, 1.6296e-04]], device='cuda:0') +Epoch 491, bias, value: tensor([-0.0543, 0.0248, 0.0085, -0.0117, -0.0032, 0.0074, -0.0159, 0.0360, + -0.0228, 0.0110], device='cuda:0'), grad: tensor([ 0.0302, 0.0327, -0.0652, -0.0016, 0.0149, -0.0442, 0.0151, -0.0020, + 0.0308, -0.0108], device='cuda:0') +100 +1e-05 +changing lr +---------------------saving model at epoch 490---------------------------------------------------- +epoch 490, time 225.90, cls_loss 0.4583 cls_loss_mapping 0.0008 cls_loss_causal 0.3988 re_mapping 0.0062 re_causal 0.0161 /// teacc 99.06 lr 0.00001000 +Epoch 492, weight, value: tensor([[-0.2160, -0.0221, 0.0383, ..., -0.0297, -0.0824, -0.1474], + [-0.0636, -0.1128, 0.0332, ..., 0.0563, -0.0289, -0.0947], + [-0.0671, -0.1014, -0.1206, ..., 0.0435, -0.0393, -0.1188], + ..., + [-0.0693, 0.0350, 0.0255, ..., 0.0161, -0.0579, -0.1000], + [-0.0873, -0.0069, 0.0167, ..., 0.0622, -0.0347, -0.1948], + [ 0.0662, 0.0695, -0.0512, ..., -0.1159, -0.0157, 0.1013]], + device='cuda:0'), grad: tensor([[ 3.6538e-05, 1.3411e-04, 1.5497e-05, ..., 7.9870e-04, + 2.4021e-04, -1.9800e-06], + [ 7.6652e-05, -3.3722e-03, 1.3635e-05, ..., 1.0700e-03, + 3.7432e-04, 2.3663e-05], + [ 2.4348e-05, 2.7943e-04, 9.7603e-06, ..., 7.3338e-04, + 2.5916e-04, 6.2324e-06], + ..., + [ 3.0613e-04, 2.2888e-03, 2.7984e-05, ..., 2.2087e-03, + 1.2751e-03, 4.2439e-05], + [-1.4181e-03, 1.7242e-03, -2.2840e-04, ..., -1.9588e-03, + 3.0088e-04, 5.8025e-05], + [ 1.9722e-03, -1.5917e-03, -1.0163e-04, ..., -7.5836e-03, + -3.3169e-03, 3.2687e-04]], device='cuda:0') +Epoch 492, bias, value: tensor([-0.0543, 0.0248, 0.0084, -0.0117, -0.0031, 0.0074, -0.0160, 0.0360, + -0.0229, 0.0109], device='cuda:0'), grad: tensor([-0.0183, 0.0073, 0.0120, 0.0133, 0.0193, 0.0141, -0.0110, 0.0216, + -0.0215, -0.0367], device='cuda:0') +100 +1e-05 +changing lr +epoch 491, time 225.98, cls_loss 0.4780 cls_loss_mapping 0.0008 cls_loss_causal 0.4168 re_mapping 0.0061 re_causal 0.0162 /// teacc 99.02 lr 0.00001000 +Epoch 493, weight, value: tensor([[-0.2160, -0.0219, 0.0382, ..., -0.0298, -0.0824, -0.1475], + [-0.0636, -0.1129, 0.0332, ..., 0.0563, -0.0291, -0.0948], + [-0.0671, -0.1015, -0.1205, ..., 0.0434, -0.0394, -0.1189], + ..., + [-0.0695, 0.0352, 0.0254, ..., 0.0160, -0.0578, -0.1000], + [-0.0873, -0.0070, 0.0168, ..., 0.0622, -0.0345, -0.1949], + [ 0.0662, 0.0694, -0.0513, ..., -0.1158, -0.0157, 0.1013]], + device='cuda:0'), grad: tensor([[ 1.5132e-05, -1.0796e-03, 2.1896e-03, ..., 6.0177e-04, + 4.5635e-08, -4.9561e-05], + [ 7.7009e-05, 2.7370e-03, -2.1362e-03, ..., 1.2451e-02, + 3.5390e-08, 7.1144e-04], + [-3.6526e-04, -3.0651e-03, -9.2087e-03, ..., -1.8417e-02, + 4.7497e-08, 5.8681e-05], + ..., + [ 1.2957e-05, 1.2598e-03, 1.0052e-03, ..., 1.0014e-03, + 2.2445e-07, 6.8855e-04], + [ 1.5819e-04, 2.2945e-03, 1.0099e-03, ..., 1.3514e-03, + 1.7330e-05, 1.1435e-03], + [-1.7452e-04, -2.5311e-03, 1.4973e-03, ..., 2.0447e-03, + 1.1474e-06, -2.9602e-03]], device='cuda:0') +Epoch 493, bias, value: tensor([-0.0544, 0.0247, 0.0083, -0.0115, -0.0031, 0.0076, -0.0159, 0.0359, + -0.0228, 0.0108], device='cuda:0'), grad: tensor([ 0.0104, 0.0045, -0.0569, 0.0186, -0.0103, -0.0108, 0.0056, 0.0148, + 0.0179, 0.0062], device='cuda:0') +100 +1e-05 +changing lr +epoch 492, time 224.69, cls_loss 0.4789 cls_loss_mapping 0.0008 cls_loss_causal 0.4099 re_mapping 0.0060 re_causal 0.0158 /// teacc 99.03 lr 0.00001000 +Epoch 494, weight, value: tensor([[-0.2161, -0.0219, 0.0382, ..., -0.0296, -0.0822, -0.1475], + [-0.0635, -0.1130, 0.0333, ..., 0.0563, -0.0290, -0.0947], + [-0.0669, -0.1015, -0.1204, ..., 0.0435, -0.0394, -0.1189], + ..., + [-0.0693, 0.0351, 0.0256, ..., 0.0159, -0.0575, -0.0999], + [-0.0874, -0.0069, 0.0167, ..., 0.0622, -0.0344, -0.1949], + [ 0.0661, 0.0695, -0.0515, ..., -0.1157, -0.0159, 0.1012]], + device='cuda:0'), grad: tensor([[ 1.1420e-04, 1.0741e-04, 1.5497e-04, ..., 1.4677e-03, + -8.3637e-04, 4.3094e-05], + [ 9.7871e-05, 1.6570e-04, 1.7285e-04, ..., -1.8265e-02, + -2.8381e-03, 4.2826e-05], + [ 6.2466e-04, 4.8161e-04, 9.5415e-04, ..., 4.9973e-03, + 5.0688e-04, 2.9111e-04], + ..., + [-3.3169e-03, 7.1955e-04, -3.2463e-03, ..., -3.2139e-03, + -1.6761e-04, -2.0943e-03], + [ 2.8872e-04, 5.7745e-04, -1.5955e-03, ..., 8.3780e-04, + 5.1928e-04, 1.3673e-04], + [ 3.0479e-03, -8.3447e-05, 1.0700e-03, ..., 3.0727e-03, + 1.2434e-04, 1.5306e-03]], device='cuda:0') +Epoch 494, bias, value: tensor([-0.0542, 0.0247, 0.0083, -0.0116, -0.0032, 0.0074, -0.0157, 0.0358, + -0.0229, 0.0108], device='cuda:0'), grad: tensor([ 0.0119, -0.0706, 0.0251, -0.0159, -0.0159, 0.0144, 0.0223, 0.0001, + 0.0081, 0.0204], device='cuda:0') +100 +1e-05 +changing lr +epoch 493, time 228.03, cls_loss 0.4884 cls_loss_mapping 0.0009 cls_loss_causal 0.4246 re_mapping 0.0060 re_causal 0.0161 /// teacc 99.05 lr 0.00001000 +Epoch 495, weight, value: tensor([[-0.2162, -0.0218, 0.0382, ..., -0.0296, -0.0821, -0.1476], + [-0.0636, -0.1130, 0.0333, ..., 0.0564, -0.0291, -0.0945], + [-0.0669, -0.1015, -0.1205, ..., 0.0435, -0.0395, -0.1187], + ..., + [-0.0693, 0.0352, 0.0257, ..., 0.0160, -0.0575, -0.0999], + [-0.0875, -0.0068, 0.0168, ..., 0.0622, -0.0344, -0.1949], + [ 0.0662, 0.0694, -0.0515, ..., -0.1157, -0.0161, 0.1012]], + device='cuda:0'), grad: tensor([[-9.4147e-03, 3.1638e-04, 7.3719e-04, ..., 2.3880e-03, + 9.0742e-04, 1.5283e-04], + [ 1.5616e-04, 5.2303e-05, -1.1997e-03, ..., -2.5806e-03, + 2.0766e-04, 1.3888e-05], + [-5.2357e-04, -1.1339e-03, -1.4915e-03, ..., -3.8185e-03, + -1.8063e-03, 6.1321e-04], + ..., + [ 1.2624e-04, 1.5974e-05, 1.4555e-04, ..., -5.9395e-03, + -2.4471e-03, -6.6124e-07], + [ 2.4930e-05, 3.2258e-04, -3.8624e-04, ..., 2.3727e-03, + 1.2989e-03, 2.8300e-04], + [ 6.2637e-03, 5.6297e-05, 9.3281e-05, ..., 1.4601e-03, + 3.9983e-04, 4.0606e-06]], device='cuda:0') +Epoch 495, bias, value: tensor([-0.0541, 0.0248, 0.0082, -0.0116, -0.0032, 0.0074, -0.0157, 0.0359, + -0.0230, 0.0108], device='cuda:0'), grad: tensor([-0.0168, -0.0260, -0.0178, 0.0167, 0.0137, 0.0147, 0.0153, -0.0506, + 0.0190, 0.0319], device='cuda:0') +100 +1e-05 +changing lr +epoch 494, time 225.87, cls_loss 0.4873 cls_loss_mapping 0.0008 cls_loss_causal 0.4290 re_mapping 0.0061 re_causal 0.0166 /// teacc 99.02 lr 0.00001000 +Epoch 496, weight, value: tensor([[-0.2161, -0.0219, 0.0382, ..., -0.0297, -0.0821, -0.1475], + [-0.0634, -0.1127, 0.0333, ..., 0.0563, -0.0290, -0.0945], + [-0.0670, -0.1016, -0.1206, ..., 0.0434, -0.0393, -0.1188], + ..., + [-0.0692, 0.0351, 0.0257, ..., 0.0160, -0.0575, -0.1001], + [-0.0875, -0.0068, 0.0167, ..., 0.0621, -0.0345, -0.1949], + [ 0.0662, 0.0694, -0.0515, ..., -0.1158, -0.0161, 0.1013]], + device='cuda:0'), grad: tensor([[ 7.8022e-05, 1.6165e-04, 2.3155e-03, ..., 2.5501e-03, + 0.0000e+00, 5.0403e-06], + [ 8.9049e-05, 1.3676e-03, -1.6630e-05, ..., -1.8501e-03, + 0.0000e+00, 5.8562e-06], + [ 1.7405e-04, 4.3774e-04, 3.1531e-05, ..., 1.7529e-03, + 0.0000e+00, 5.6177e-06], + ..., + [-3.7670e-03, -6.2904e-03, 1.3001e-05, ..., -6.6948e-04, + 0.0000e+00, -2.4338e-03], + [-2.9063e-04, 2.1820e-03, 1.8537e-04, ..., 1.8320e-03, + 0.0000e+00, 1.7732e-05], + [ 1.2016e-04, 1.6689e-03, 9.0420e-05, ..., -1.2627e-03, + 0.0000e+00, -7.0906e-04]], device='cuda:0') +Epoch 496, bias, value: tensor([-0.0543, 0.0248, 0.0081, -0.0114, -0.0032, 0.0074, -0.0155, 0.0360, + -0.0231, 0.0107], device='cuda:0'), grad: tensor([-0.0108, -0.0139, 0.0137, 0.0175, 0.0179, 0.0211, -0.0031, -0.0191, + 0.0135, -0.0368], device='cuda:0') +100 +1e-05 +changing lr +epoch 495, time 229.03, cls_loss 0.5054 cls_loss_mapping 0.0009 cls_loss_causal 0.4367 re_mapping 0.0060 re_causal 0.0162 /// teacc 99.02 lr 0.00001000 +Epoch 497, weight, value: tensor([[-0.2162, -0.0217, 0.0381, ..., -0.0298, -0.0820, -0.1476], + [-0.0633, -0.1126, 0.0334, ..., 0.0563, -0.0290, -0.0945], + [-0.0671, -0.1016, -0.1205, ..., 0.0434, -0.0393, -0.1188], + ..., + [-0.0692, 0.0350, 0.0257, ..., 0.0160, -0.0574, -0.1000], + [-0.0874, -0.0068, 0.0169, ..., 0.0622, -0.0345, -0.1947], + [ 0.0661, 0.0695, -0.0516, ..., -0.1159, -0.0161, 0.1012]], + device='cuda:0'), grad: tensor([[ 5.5462e-05, 1.2054e-03, 2.3261e-05, ..., 7.0047e-04, + 3.3617e-04, 1.9932e-03], + [ 4.6992e-04, 6.0987e-04, 3.4642e-04, ..., 2.7924e-03, + 6.6423e-04, 1.3602e-04], + [ 8.8274e-05, 4.4727e-04, 5.4538e-05, ..., -1.4015e-02, + 9.8705e-05, 1.2672e-04], + ..., + [ 2.9087e-04, 3.0756e-04, 1.0767e-03, ..., 2.3022e-03, + 7.2327e-03, 1.9372e-04], + [ 4.8018e-04, 5.2109e-03, 2.5916e-04, ..., 1.9569e-03, + 4.9973e-04, 9.5272e-04], + [ 5.6791e-04, 2.4853e-03, 6.5756e-04, ..., 1.4591e-03, + 1.2732e-03, 4.8375e-04]], device='cuda:0') +Epoch 497, bias, value: tensor([-0.0543, 0.0248, 0.0082, -0.0114, -0.0032, 0.0073, -0.0155, 0.0360, + -0.0230, 0.0107], device='cuda:0'), grad: tensor([ 0.0016, 0.0225, -0.0469, -0.0292, 0.0048, -0.0065, -0.0036, 0.0338, + 0.0274, -0.0040], device='cuda:0') +100 +1e-05 +changing lr +epoch 496, time 226.48, cls_loss 0.4872 cls_loss_mapping 0.0009 cls_loss_causal 0.4209 re_mapping 0.0059 re_causal 0.0160 /// teacc 99.03 lr 0.00001000 +Epoch 498, weight, value: tensor([[-0.2162, -0.0217, 0.0380, ..., -0.0298, -0.0821, -0.1476], + [-0.0631, -0.1126, 0.0337, ..., 0.0564, -0.0290, -0.0943], + [-0.0672, -0.1014, -0.1206, ..., 0.0434, -0.0393, -0.1189], + ..., + [-0.0691, 0.0349, 0.0258, ..., 0.0160, -0.0574, -0.0999], + [-0.0875, -0.0070, 0.0169, ..., 0.0622, -0.0344, -0.1948], + [ 0.0659, 0.0694, -0.0517, ..., -0.1160, -0.0161, 0.1011]], + device='cuda:0'), grad: tensor([[ 1.4246e-04, -4.8027e-03, 1.6201e-04, ..., -5.1880e-04, + 1.8692e-04, -7.1764e-04], + [ 1.3784e-05, 7.4005e-04, 1.9445e-03, ..., 5.1384e-03, + 7.2575e-04, 7.6771e-05], + [ 2.9594e-05, 6.0120e-03, 1.6320e-04, ..., -8.0872e-03, + -1.7929e-03, 1.8120e-04], + ..., + [ 9.2089e-05, 2.1801e-03, -1.8148e-03, ..., -1.6670e-03, + 7.5531e-04, 3.6383e-04], + [ 2.2268e-04, -8.4496e-04, 2.3949e-04, ..., 1.8063e-03, + 1.7347e-03, 4.8542e-04], + [-7.7295e-04, 7.8249e-04, 4.3488e-04, ..., 2.1038e-03, + 4.6515e-04, -4.4644e-05]], device='cuda:0') +Epoch 498, bias, value: tensor([-0.0543, 0.0249, 0.0082, -0.0115, -0.0031, 0.0072, -0.0156, 0.0361, + -0.0230, 0.0107], device='cuda:0'), grad: tensor([-0.0281, 0.0275, -0.0361, -0.0083, 0.0170, -0.0309, 0.0054, 0.0110, + 0.0239, 0.0185], device='cuda:0') +100 +1e-05 +changing lr +epoch 497, time 225.86, cls_loss 0.4814 cls_loss_mapping 0.0008 cls_loss_causal 0.4181 re_mapping 0.0061 re_causal 0.0164 /// teacc 99.04 lr 0.00001000 +Epoch 499, weight, value: tensor([[-0.2163, -0.0217, 0.0380, ..., -0.0299, -0.0823, -0.1474], + [-0.0633, -0.1127, 0.0337, ..., 0.0563, -0.0290, -0.0945], + [-0.0673, -0.1014, -0.1206, ..., 0.0434, -0.0394, -0.1189], + ..., + [-0.0691, 0.0348, 0.0259, ..., 0.0159, -0.0574, -0.1000], + [-0.0876, -0.0070, 0.0169, ..., 0.0623, -0.0344, -0.1948], + [ 0.0661, 0.0695, -0.0515, ..., -0.1159, -0.0158, 0.1013]], + device='cuda:0'), grad: tensor([[ 1.9688e-06, 4.4048e-05, 1.2055e-05, ..., -2.6569e-03, + -2.4494e-07, 9.0778e-05], + [ 2.6561e-06, 1.5664e-04, -2.9778e-04, ..., 3.0804e-04, + 1.3970e-09, 1.4043e-04], + [ 3.2187e-06, -3.1686e-04, 9.5293e-06, ..., -2.1992e-03, + 4.0978e-08, -1.1663e-03], + ..., + [ 2.0117e-05, 2.6509e-05, -9.8441e-07, ..., 2.8640e-05, + 2.1001e-07, 5.2035e-05], + [-2.7657e-05, -1.1129e-03, -2.6330e-05, ..., 1.2083e-03, + 3.0734e-08, 1.5676e-04], + [-4.6015e-05, 5.5075e-05, 2.3931e-05, ..., 3.9530e-04, + -1.1316e-07, 6.0052e-05]], device='cuda:0') +Epoch 499, bias, value: tensor([-0.0543, 0.0247, 0.0082, -0.0114, -0.0033, 0.0073, -0.0156, 0.0359, + -0.0229, 0.0109], device='cuda:0'), grad: tensor([-0.0022, 0.0071, -0.0245, 0.0079, 0.0087, -0.0218, 0.0084, 0.0049, + 0.0057, 0.0059], device='cuda:0') +100 +1e-05 +changing lr +epoch 498, time 227.24, cls_loss 0.4588 cls_loss_mapping 0.0008 cls_loss_causal 0.4012 re_mapping 0.0059 re_causal 0.0158 /// teacc 99.06 lr 0.00001000 +Epoch 500, weight, value: tensor([[-0.2162, -0.0217, 0.0381, ..., -0.0299, -0.0823, -0.1473], + [-0.0632, -0.1126, 0.0337, ..., 0.0563, -0.0287, -0.0945], + [-0.0674, -0.1014, -0.1208, ..., 0.0433, -0.0395, -0.1190], + ..., + [-0.0690, 0.0349, 0.0258, ..., 0.0159, -0.0574, -0.0999], + [-0.0877, -0.0070, 0.0167, ..., 0.0622, -0.0345, -0.1949], + [ 0.0660, 0.0695, -0.0514, ..., -0.1158, -0.0159, 0.1014]], + device='cuda:0'), grad: tensor([[ 1.6510e-05, 7.9334e-05, 6.8843e-05, ..., 8.8787e-04, + 7.0238e-04, 1.6654e-04], + [ 9.3207e-06, 5.3167e-04, 2.8014e-04, ..., 2.8210e-03, + -7.8773e-04, 1.3971e-04], + [ 1.8209e-05, 1.2755e-04, 8.7559e-05, ..., 5.3406e-04, + 7.2622e-04, 1.7929e-04], + ..., + [ 5.4508e-05, -8.9455e-04, -7.6675e-04, ..., -8.8196e-03, + -5.5962e-03, 3.6049e-04], + [ 3.4541e-05, 8.4877e-05, 7.9334e-05, ..., 9.5987e-04, + 8.3447e-04, 1.7250e-04], + [-1.3542e-04, 2.3293e-04, 1.3006e-04, ..., 1.3494e-03, + 9.6369e-04, -3.1066e-04]], device='cuda:0') +Epoch 500, bias, value: tensor([-0.0543, 0.0247, 0.0081, -0.0114, -0.0033, 0.0074, -0.0157, 0.0360, + -0.0230, 0.0110], device='cuda:0'), grad: tensor([ 0.0110, -0.0037, -0.0200, -0.0128, 0.0090, 0.0153, 0.0120, -0.0386, + 0.0118, 0.0161], device='cuda:0') +100 +1e-05 +changing lr +epoch 499, time 225.78, cls_loss 0.4998 cls_loss_mapping 0.0008 cls_loss_causal 0.4380 re_mapping 0.0059 re_causal 0.0166 /// teacc 99.02 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 98.989998 98.879997 ... 84.90284 74.539656 +ShearY 98.869995 98.979996 ... 84.90284 68.235694 +AutoContrast 98.989998 99.000000 ... 84.90284 63.396949 +Invert 98.989998 98.919998 ... 84.90284 72.542549 +Equalize 98.250000 98.309998 ... 84.90284 70.364035 +Solarize 98.279999 98.089996 ... 84.90284 65.044853 +SolarizeAdd 98.470001 98.199997 ... 84.90284 72.076269 +Posterize 98.979996 98.940002 ... 84.90284 74.368635 +Contrast 99.040001 98.989998 ... 84.90284 73.096380 +Color 99.029999 99.059998 ... 84.90284 63.243871 +Brightness 98.930000 98.989998 ... 84.90284 73.377682 +Sharpness 99.040001 99.010002 ... 84.90284 74.417749 +NoiseSalt 99.099998 99.040001 ... 84.90284 67.804282 +NoiseGaussian 99.019997 99.070000 ... 84.90284 69.921497 +w/o do (original x) 99.060000 0.000000 ... 0.00000 78.214057 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.04 70.075292 77.858016 78.174395 85.201794 77.827374 +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1_eps1_RA/14factor_last.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'last', 'eval_mapping': True} +loading weight of last +randm: False +stride: 3 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat + mnist mnist_FA ... usps_FA Avg +ShearX 99.040001 98.959999 ... 84.554062 73.788023 +ShearY 98.790001 98.860001 ... 84.554062 67.782350 +AutoContrast 99.019997 99.059998 ... 84.554062 63.308972 +Invert 99.019997 98.940002 ... 84.554062 72.191246 +Equalize 98.199997 98.320000 ... 84.554062 71.175241 +Solarize 98.250000 98.129997 ... 84.554062 65.321118 +SolarizeAdd 98.470001 98.150002 ... 84.554062 72.141506 +Posterize 98.979996 98.919998 ... 84.554062 74.073261 +Contrast 99.070000 98.970001 ... 84.554062 72.475910 +Color 99.019997 99.019997 ... 84.554062 61.998447 +Brightness 99.000000 98.970001 ... 84.554062 72.502592 +Sharpness 98.989998 98.970001 ... 84.554062 74.041452 +NoiseSalt 98.989998 98.989998 ... 84.554062 67.042088 +NoiseGaussian 99.010002 99.010002 ... 84.554062 68.879292 +w/o do (original x) 99.020000 0.000000 ... 0.000000 78.022972 + +[15 rows x 11 columns] + mnist svhn mnist_m syndigit usps Avg +do 99.02 69.426091 78.091323 77.703339 84.10563 77.331596 diff --git a/Meta-causal/code-withStyleAttack/backbone_multiblock.py b/Meta-causal/code-withStyleAttack/backbone_multiblock.py new file mode 100644 index 0000000000000000000000000000000000000000..e4045f42709def4cff06ccbb144e0f21bf017c5d --- /dev/null +++ b/Meta-causal/code-withStyleAttack/backbone_multiblock.py @@ -0,0 +1,155 @@ +from torch import nn +from torch.utils import model_zoo +#from torchvision.models.resnet import BasicBlock, model_urls, Bottleneck +from torchvision.models.resnet import BasicBlock, Bottleneck + +import torch +import ssl +# from torch import nn as nn +# from utils.util import * + +ssl._create_default_https_context = ssl._create_unverified_context + +all = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101','resnet152'] + +model_urls = { +'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', +'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', +'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', +'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', +'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth', +} + + +class ResNetMultiBlock(nn.Module): + def __init__(self, block, layers,classes=7,c_dim=512): + self.inplanes = 64 + super(ResNetMultiBlock, self).__init__() + self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, + bias=False) + self.bn1 = nn.BatchNorm2d(64) + self.relu = nn.ReLU(inplace=True) + self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + self.layer1 = self._make_layer(block, 64, layers[0]) + self.layer2 = self._make_layer(block, 128, layers[1], stride=2) + self.layer3 = self._make_layer(block, 256, layers[2], stride=2) + self.layer4 = self._make_layer(block, 512, layers[3], stride=2) + self.avgpool = nn.AvgPool2d(7, stride=1) + self.class_classifier = nn.Linear(c_dim, classes) + + # for attacking + self.classifier = nn.Linear(c_dim, classes) + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + nn.init.constant_(m.weight, 1) + nn.init.constant_(m.bias, 0) + + def _make_layer(self, block, planes, blocks, stride=1): + downsample = None + if stride != 1 or self.inplanes != planes * block.expansion: + downsample = nn.Sequential( + nn.Conv2d(self.inplanes, planes * block.expansion, + kernel_size=1, stride=stride, bias=False), + nn.BatchNorm2d(planes * block.expansion), + ) + + layers = [] + layers.append(block(self.inplanes, planes, stride, downsample)) + self.inplanes = planes * block.expansion + for i in range(1, blocks): + layers.append(block(self.inplanes, planes)) + + return nn.Sequential(*layers) + + # def forward(self, x, mode='fc'): + # if mode == 'c': + # return self.class_classifier(x) + # else: + # x = self.conv1(x) + # x = self.bn1(x) + # x = self.relu(x) + # x = self.maxpool(x) + + # x = self.layer1(x) + # x = self.layer2(x) + # x = self.layer3(x) + # x = self.layer4(x) + # x = self.avgpool(x) + # x = x.view(x.size(0), -1) + # # print("x.shape:",x.shape) + # return self.class_classifier(x), x + + def forward_block1(self, x): + x = self.conv1(x) + x = self.bn1(x) + x = self.relu(x) + x = self.maxpool(x) + x = self.layer1(x) + return x + + def forward_block2(self, x): + x = self.layer2(x) + return x + + def forward_block3(self, x): + x = self.layer3(x) + return x + + def forward_block4(self, x): + x = self.layer4(x) + return x + + def forward_rest(self,x): + x = self.avgpool(x) + x = x.view(x.size(0), -1) + return self.class_classifier(x), x + + + def forward(self,x,mode='fc'): + if mode == 'c': + return self.class_classifier(x) + else: + layer1 = self.forward_block1(x) + layer2 = self.forward_block2(layer1) + layer3 = self.forward_block3(layer2) + layer4 = self.forward_block4(layer3) + p, f= self.forward_rest(layer4) + return layer1, layer2, layer3, layer4, p, f + + +def resnet18Multiblock(pretrained=True, **kwargs): + """Constructs a ResNet-18 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNetMultiBlock(BasicBlock, [2, 2, 2, 2], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet18']), strict=False) + return model + + +if __name__ =='__main__': + print('%'*100) + print('---test RN18 original--') + from network.resnet import resnet18 + cls_net = resnet18(classes=7,c_dim=2048).cuda() + print('cls_net:', cls_net) + x = torch.randn([16,3,227,227]).cuda() + p, f = cls_net(x) + print(p.shape, f.shape) + + + print('---test RN18 multiblock--') + cls_net = resnet18Multiblock(classes=7,c_dim=2048).cuda() + print('cls_net:', cls_net) + x = torch.randn([16,3,227,227]).cuda() + L1, L2, L3, L4, p, f = cls_net(x) + print(L1.shape, L2.shape, L3.shape, L4.shape, p.shape, f.shape) + + + + diff --git a/Meta-causal/code-withStyleAttack/data_loader_joint_v3.py b/Meta-causal/code-withStyleAttack/data_loader_joint_v3.py new file mode 100644 index 0000000000000000000000000000000000000000..dd2de2de850d89657507ff9a0e348c94c0e070d0 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/data_loader_joint_v3.py @@ -0,0 +1,743 @@ +''' Digit 实验 +''' +import torch +import torch.nn.functional as F +from torch.utils.data import Dataset, TensorDataset +from torchvision import transforms +from torchvision.datasets import MNIST, SVHN, CIFAR10, STL10, USPS + +import os +import pickle +import numpy as np +import h5py +#import cv2 +from scipy.io import loadmat +from PIL import Image + +from tools.autoaugment import SVHNPolicy, CIFAR10Policy +from tools.randaugment import RandAugment +from tools.causalaugment_v3 import RandAugment_incausal, FactualAugment_incausal, CounterfactualAugment_incausal, MultiCounterfactualAugment_incausal + +class myTensorDataset(Dataset): + def __init__(self, x, y, transform=None, transform2=None, transform3=None, twox=False): + self.x = x + self.y = y + self.transform = transform + self.transform2 = transform2 + self.transform3 = transform3 + self.twox = twox + def __len__(self): + return len(self.x) + def __getitem__(self, index): + x = self.x[index] + y = self.y[index] + c, h, w =x.shape + # print("x.shape:",x.shape) + if self.transform is not None: + x_RA = self.transform(x) + # print("x_RA.shape:",x_RA.shape) + if self.transform3 is not None: + x_CA = self.transform3(x_RA) + x_CA = x_CA.reshape(-1,c,h,w) + # print("x_CA.shape:",x_CA.shape) + if self.transform2 is not None: + x_FA = self.transform2(x) + # x_FA = x_FA.view(c,13,h,w) + x_FA = x_FA.reshape(-1,c,h,w) + # print("x_FA_in getitem.shape:",x_FA.shape) + # print("x_FA.shape:",x_FA.shape) + return (x, x_RA, x_FA, x_CA), y + else: + return (x, x_RA, x_CA), y + else: + if self.transform2 is not None: + x_FA = self.transform2(x) + x_FA = x_FA.reshape(-1,c,h,w) + return (x, x_RA, x_FA), y + else: + if self.twox: + return (x, x_RA), y + else: + return x_RA, y + +HOME = os.environ['HOME'] +print(HOME) +def resize_imgs(x, size): + ''' 目前只能处理单通道 + x [n, 28, 28] + size int + ''' + resize_x = np.zeros([x.shape[0], size, size]) + for i, im in enumerate(x): + im = Image.fromarray(im) + im = im.resize([size, size], Image.ANTIALIAS) + resize_x[i] = np.asarray(im) + return resize_x + +def load_mnist(split='train', translate=None, twox=False, ntr=None, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + ''' + autoaug == 'AA', AutoAugment + 'FastAA', Fast AutoAugment + 'RA', RandAugment + channels == 3 默认返回 rgb 3通道图像 + 1 返回单通道图像 + ''' + #path = f'data/mnist-{split}.pkl' + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/minst-{split}.pkl' + if not os.path.exists(path): + dataset = MNIST(f'{HOME}/.pytorch/MNIST', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + if split=='train': + x, y = x[0:10000], y[0:10000] + x = torch.tensor(resize_imgs(x.numpy(), 32)) + x = (x.float()/255.).unsqueeze(1).repeat(1,3,1,1) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + with open(path, 'rb') as f: + # print("reading!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!") + x, y = pickle.load(f) + if channels == 1: + x = x[:,0:1,:,:] + + if ntr is not None: + x, y = x[0:ntr], y[0:ntr] + + # 如果没有数据增强 + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + + # 数据增强管道 + transform = [transforms.ToPILImage()] + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def load_cifar10(split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + dataset = CIFAR10(f'{HOME}/.pytorch/CIFAR10', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y) + x = x.float()/255. + print(x.shape,y.shape) + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset +def load_IMG(task='S-U', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + # path = f'data/img2vid/{domain}/stanford40_12.npz' + if task == 'S-U': + path = f'data/img2vid/{task}/stanford40_12.npz' + elif task == 'E-H': + path = f'data/img2vid/{task}/EAD50_13.npz' + print(path) + dataset = np.load(path) + x, y = dataset['x'], dataset['y'] + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def load_VID(task='S-U',split='1'): + if task == 'S-U': + path = f'data/img2vid/{task}/ucf101_12_frame_sample8_{split}.npz' + elif task == 'E-H': + path = f'data/img2vid/{task}/hmdb51_13_frame_sample8_{split}.npz' + dataset = np.load(path) + print(path) + x, y = dataset['x'], dataset['y'] + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + dataset = TensorDataset(x, y) + return dataset + +def load_pacs(domain='photo', split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + #path = f'data/PACS/{domain}_{split}.hdf5' + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/{domain}_{split}.hdf5' + dataset = h5py.File(path, 'r') + x, y = dataset['images'], dataset['labels'] + #for i in range(20): + # cv2.imwrite('data/PACS/debug_images/img_cv2_'+domain+'_'+split+'_'+str(i)+'.png', x[i]) + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + y = y - 1 + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def read_dataset(domain, split): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/{domain}_{split}.hdf5' + dataset = h5py.File(path, 'r') + x_temp, y_temp = dataset['images'], dataset['labels'] + b, g, r = np.split(x_temp,3,axis=-1) + x_temp = np.concatenate((r,g,b),axis=-1) + x_temp = x_temp.transpose(0,3,1,2) + x_temp, y_temp = torch.tensor(x_temp), torch.tensor(y_temp, dtype=torch.long) + y_temp = y_temp - 1 + x_temp = x_temp.float()/255. + return x_temp, y_temp + +def load_pacs_multi(target_domain=['photo'], split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + domains = ['art_painting', 'cartoon', 'photo', 'sketch'] + source_domain = [i for i in domains if i != target_domain] + for i in range(len(source_domain)): + x_temp, y_temp = read_dataset(source_domain[i],split=split) + print(x_temp.shape,y_temp.shape) + if i == 0: + x = x_temp.clone() + y = y_temp.clone() + else: + x = torch.cat([x,x_temp],0) + y = torch.cat([y,y_temp],0) + print(x.shape,y.shape) + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + + +def load_cifar10_c_level1(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level1.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level1") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[0:10000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level1") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level2(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level2.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level2") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[10000:20000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level2") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level3(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level3.pkl' + if not os.path.exists(path): + print("generating cifar10_c_level3") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[20000:30000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level3") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level4(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level4.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level4") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[30000:40000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level4") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level5(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level5.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level5") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[40000:50000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level5") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c(dataroot): + y = np.load(os.path.join(dataroot, 'labels.npy')) + print("y.shape:",y.shape) + y_single = y[0:10000] + x1 = torch.zeros((190000,3,32,32)) + x2 = torch.zeros((190000,3,32,32)) + x3 = torch.zeros((190000,3,32,32)) + x4 = torch.zeros((190000,3,32,32)) + x5 = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y_total = y_single + else: + y_total = np.hstack((y_total,y_single)) + print("y_total.shape:",y_total.shape) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + x = np.load(os.path.join(dataroot,filename)) + x = x.transpose(0,3,1,2) + x = torch.tensor(x) + x = x.float()/255. + print(x.shape) + x1[index*10000:(index+1)*10000] = x[0:10000] + x2[index*10000:(index+1)*10000] = x[10000:20000] + x3[index*10000:(index+1)*10000] = x[20000:30000] + x4[index*10000:(index+1)*10000] = x[30000:40000] + x5[index*10000:(index+1)*10000] = x[40000:50000] + index = index + 1 + # x1, x2, x3, x4, x5, y_total = torch.tensor(x1), torch.tensor(x2), torch.tensor(x3),\ + # torch.tensor(x4),torch.tensor(x5),torch.tensor(y_total) + y_total = torch.tensor(y_total) + dataset1 = TensorDataset(x1, y_total) + dataset2 = TensorDataset(x2, y_total) + dataset3 = TensorDataset(x3, y_total) + dataset4 = TensorDataset(x4, y_total) + dataset5 = TensorDataset(x5, y_total) + return dataset1,dataset2,dataset3,dataset4,dataset5 + +def load_cifar10_c_class(dataroot,CORRUPTIONS): + y = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = y[0:10000] + y_single = torch.tensor(y_single) + print("y.shape:",y.shape) + x = np.load(os.path.join(dataroot,CORRUPTIONS+'.npy')) + print("loading data of",os.path.join(dataroot,CORRUPTIONS+'.npy')) + x = x.transpose(0,3,1,2) + x = torch.tensor(x) + x = x.float()/255. + dataset = [] + for i in range(5): + x_single = x[i*10000:(i+1)*10000] + dataset.append(TensorDataset(x_single, y_single)) + return dataset + +def load_usps(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/usps-{split}.pkl' + if not os.path.exists(path): + dataset = USPS(f'{HOME}/.pytorch/USPS', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + x = torch.tensor(resize_imgs(x, 32)) + x = (x.float()/255.).unsqueeze(1).repeat(1,3,1,1) + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + with open(path, 'rb') as f: + x, y = pickle.load(f) + if channels == 1: + x = x[:,0:1,:,:] + dataset = TensorDataset(x, y) + return dataset + +def load_svhn(split='train', channels=3): + dataset = SVHN(f'{HOME}/.pytorch/SVHN', split=split, download=True) + x, y = dataset.data, dataset.labels + x = x.astype('float32')/255. + x, y = torch.tensor(x), torch.tensor(y) + if channels == 1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + + +def load_syndigit(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/synth_{split}_32x32.mat' + data = loadmat(path) + x, y = data['X'], data['y'] + x = np.transpose(x, [3, 2, 0, 1]).astype('float32')/255. + y = y.squeeze() + x, y = torch.tensor(x), torch.tensor(y) + if channels == 1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + +def load_mnist_m(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/mnist_m-{split}.pkl' + with open(path, 'rb') as f: + x, y = pickle.load(f) + x, y = torch.tensor(x.astype('float32')/255.), torch.tensor(y) + if channels==1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + +if __name__=='__main__': + dataset = load_mnist(split='train') + print('mnist train', len(dataset)) + dataset = load_mnist('test') + print('mnist test', len(dataset)) + dataset = load_mnist_m('test') + print('mnsit_m test', len(dataset)) + dataset = load_svhn(split='test') + print('svhn', len(dataset)) + dataset = load_usps(split='test') + print('usps', len(dataset)) + dataset = load_syndigit(split='test') + print('syndigit', len(dataset)) + diff --git a/Meta-causal/code-withStyleAttack/env.yaml b/Meta-causal/code-withStyleAttack/env.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b0bd424fb7c5aa818f10a82173549eb0dd3199c7 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/env.yaml @@ -0,0 +1,119 @@ +name: Py3.7_torch1.8 +channels: + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ + - conda-forge + - bioconda + - defaults +dependencies: + - _libgcc_mutex=0.1=main + - asn1crypto=1.2.0=py37_0 + - blas=1.0=mkl + - bottleneck=1.3.2=py37heb32a55_1 + - bzip2=1.0.8=h7b6447c_0 + - ca-certificates=2021.10.8=ha878542_0 + - cairo=1.14.12=h8948797_3 + - certifi=2021.10.8=py37h89c1867_1 + - cffi=1.13.0=py37h2e261b9_0 + - chardet=3.0.4=py37_1003 + - click=8.0.3=pyhd3eb1b0_0 + - conda-package-handling=1.6.0=py37h7b6447c_0 + - cryptography=2.8=py37h1ba5d50_0 + - ffmpeg=4.0=hcdf2ecd_0 + - fontconfig=2.13.0=h9420a91_0 + - freeglut=3.0.0=hf484d3e_5 + - freetype=2.11.0=h70c0345_0 + - glib=2.63.1=h5a9c865_0 + - graphite2=1.3.14=h23475e2_0 + - h5py=2.8.0=py37h3010b51_1003 + - harfbuzz=1.8.8=hffaf4a1_0 + - hdf5=1.10.2=hba1933b_1 + - icu=58.2=he6710b0_3 + - idna=2.8=py37_0 + - intel-openmp=2021.3.0=h06a4308_3350 + - jasper=2.0.14=hd8c5072_2 + - jpeg=9d=h7f8727e_0 + - libedit=3.1.20181209=hc058e9b_0 + - libffi=3.2.1=hd88cf55_4 + - libgcc-ng=9.1.0=hdf63c60_0 + - libgfortran-ng=7.5.0=ha8ba4b0_17 + - libgfortran4=7.5.0=ha8ba4b0_17 + - libglu=9.0.0=hf484d3e_1 + - libopencv=3.4.2=hb342d67_1 + - libopus=1.3.1=h7b6447c_0 + - libpng=1.6.37=hbc83047_0 + - libprotobuf=3.17.2=h4ff587b_1 + - libstdcxx-ng=9.1.0=hdf63c60_0 + - libtiff=4.1.0=h2733197_0 + - libuuid=1.0.3=h7f8727e_2 + - libvpx=1.7.0=h439df22_0 + - libxcb=1.14=h7b6447c_0 + - libxml2=2.9.9=hea5a465_1 + - mkl=2021.3.0=h06a4308_520 + - mkl-service=2.4.0=py37h7f8727e_0 + - mkl_fft=1.3.1=py37hd3c417c_0 + - mkl_random=1.2.2=py37h51133e4_0 + - ncurses=6.1=he6710b0_1 + - numexpr=2.7.3=py37h22e1b3c_1 + - numpy-base=1.21.2=py37h79a1101_0 + - opencv=3.4.2=py37h6fd60c2_1 + - openssl=1.1.1h=h516909a_0 + - pandas=1.3.3=py37h8c16a72_0 + - pcre=8.45=h295c915_0 + - pip=19.3.1=py37_0 + - pixman=0.40.0=h7f8727e_1 + - protobuf=3.17.2=py37h295c915_0 + - py-opencv=3.4.2=py37hb342d67_1 + - pycosat=0.6.3=py37h14c3975_0 + - pycparser=2.19=py37_0 + - pyopenssl=19.0.0=py37_0 + - pysocks=1.7.1=py37_0 + - python=3.7.4=h265db76_1 + - python-dateutil=2.8.2=pyhd3eb1b0_0 + - python_abi=3.7=2_cp37m + - pytz=2021.3=pyhd3eb1b0_0 + - readline=7.0=h7b6447c_5 + - requests=2.22.0=py37_0 + - ruamel_yaml=0.15.46=py37h14c3975_0 + - scipy=1.7.1=py37h292c36d_2 + - setuptools=41.4.0=py37_0 + - six=1.12.0=py37_0 + - sqlite=3.30.0=h7b6447c_0 + - tensorboardx=2.2=pyhd3eb1b0_0 + - tk=8.6.8=hbc83047_0 + - tqdm=4.36.1=py_0 + - urllib3=1.24.2=py37_0 + - wheel=0.33.6=py37_0 + - xz=5.2.4=h14c3975_4 + - yaml=0.1.7=had09818_2 + - zlib=1.2.11=h7b6447c_3 + - zstd=1.3.7=h0b5b093_0 + - pip: + - absl-py==1.0.0 + - cachetools==4.2.4 + - conda-pack==0.6.0 + - google-auth==2.3.3 + - google-auth-oauthlib==0.4.6 + - grpcio==1.42.0 + - importlib-metadata==4.8.2 + - markdown==3.3.6 + - numpy==1.21.3 + - oauthlib==3.1.1 + - pillow==8.4.0 + - pyasn1==0.4.8 + - pyasn1-modules==0.2.8 + - requests-oauthlib==1.3.0 + - rsa==4.8 + - tensorboard==2.7.0 + - tensorboard-data-server==0.6.1 + - tensorboard-plugin-wit==1.8.0 + - torch==1.8.1+cu111 + - torchvision==0.9.1+cu111 + - typing-extensions==3.10.0.2 + - werkzeug==2.0.2 + - zipp==3.6.0 +prefix: /home/chenjin/miniconda3/envs/Py3.7_torch1.8 diff --git a/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py b/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py new file mode 100644 index 0000000000000000000000000000000000000000..0c15d4f45e5f1c469c28af62477bea344593ecfc --- /dev/null +++ b/Meta-causal/code-withStyleAttack/main_my_joint_v13_auto.py @@ -0,0 +1,672 @@ + +''' +训练 base 模型 +''' + +import torch +import torch.nn as nn +import torch.nn.functional as F +import itertools +from torch import optim +from torch.utils.data import DataLoader, RandomSampler +from torchvision import models +from torchvision.datasets import CIFAR10 +from torchvision.utils import make_grid +import torchvision.transforms as transforms +from tensorboardX import SummaryWriter +from torch.cuda.amp import autocast,GradScaler + +import os +import click +import time +import numpy as np + +from network import mnist_net_my as mnist_net +from mnist_net_multiblock import ConvNetMultiblock +from network import wideresnet as wideresnet +from network import resnet as resnet +from backbone_multiblock import resnet18Multiblock +from network import adaptor_v2 + +from tools import causalaugment_v3 as causalaugment +import data_loader_joint_v3 as data_loader +from tool_func import * +# from utils import set_requires_grad + + +HOME = os.environ['HOME'] + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择gpu') +@click.option('--data', type=str, default='mnist', help='数据集名称') +@click.option('--ntr', type=int, default=None, help='选择训练集前ntr个样本') +@click.option('--translate', type=float, default=None, help='随机平移数据增强') +@click.option('--autoaug', type=str, default=None, help='AA FastAA RA') +@click.option('--n', type=int, default=3, help='选择多少个factor生成RA') +@click.option('--stride', type=int, default=5, help='if autoaug==CA_multiple, stride is used') +@click.option('--factor_num', type=int, default=16, help='the first n factors') +@click.option('--epochs', type=int, default=100) +@click.option('--nbatch', type=int, default=100, help='每个epoch中batch的数量') +@click.option('--batchsize', type=int, default=128, help='每个batch中样本的数量') +@click.option('--lr', type=float, default=1e-3) +@click.option('--lr_scheduler', type=str, default='none', help='是否选择学习率衰减策略') +@click.option('--svroot', type=str, default='./saved', help='项目文件保存路径') +@click.option('--clsadapt', type=bool, default=True, help='映射后是否用分类损失') +@click.option('--lambda_causal', type=float, default=1, help='the weight of reconstruction during mapping and causal ') +@click.option('--lambda_re', type=float, default=1, help='the weight of reconstruction during mapping and causal ') +@click.option('--randm', type=bool, default=True, help='m取值是否randm') +@click.option('--randn', type=bool, default=False, help='原始特征是否detach') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') + + + + +def experiment(gpu, data, ntr, translate, autoaug,n,stride, factor_num, epochs, nbatch, batchsize, lr, lr_scheduler, svroot, clsadapt, lambda_causal,lambda_re,randm,randn,network): + print('here2') + settings = locals().copy() + print(settings) + + # 全局设置 + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + if not os.path.exists(svroot): + os.makedirs(svroot) + log_file = open(svroot+os.sep+'log.log',"w") + log_file.write(str(settings)+'\n') + writer = SummaryWriter(svroot) + CA = causalaugment.MultiCounterfactualAugment(factor_num,stride) + # FA = causalaugment.FactualAugment(m=4, factor_num=factor_num, randm=True) + # 加载数据集和模型 + if data in ['mnist', 'mnist_t']: + if data == 'mnist': + trset = data_loader.load_mnist('train', translate=translate,twox=True, ntr=ntr, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + elif data == 'mnist_t': + trset = data_loader.load_mnist_t('train', translate=translate, ntr=ntr) + teset = data_loader.load_mnist('test') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=0, \ + sampler=RandomSampler(trset, True, nbatch*batchsize)) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=0, shuffle=False) + #cls_net = mnist_net.ConvNet().cuda() + cls_net = ConvNetMultiblock().cuda() + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(1024,512,1024,2).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + parameter_list.append({'params':cls_net.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + #print("---------------------------------------------------------------------------------------") + opt = optim.Adam(parameter_list, lr=lr) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.95) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=int(epochs*0.8)) + # print("------------------------------------opt_mapping---------------------------------------------------") + # for param_group in opt_mapping.param_groups: + # print(param_group.keys()) + # # print(type(param_group)) + # print([type(value) for value in param_group.values()]) + # print('lr: ',param_group['lr']) + + # print("------------------------------------opt_causal---------------------------------------------------") + # for param_group in opt_causal.param_groups: + # print(param_group.keys()) + # # print(type(param_group)) + # print([type(value) for value in param_group.values()]) + # print('lr: ',param_group['lr']) + + elif data == 'cifar10': + # 加载数据集 + trset = data_loader.load_cifar10(split='train',twox=True, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + teset = data_loader.load_cifar10(split='test') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=4, shuffle=True, drop_last=True) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=4, shuffle=False) + cls_net = wideresnet.WideResNet(16, 10, 4).cuda() + # cls_opt = optim.SGD(cls_net.parameters(), lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(256,512,256,4).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + parameter_list.append({'params':cls_net.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + #print("---------------------------------------------------------------------------------------") + #opt = optim.Adam(parameter_list) + opt = optim.SGD(parameter_list, lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.95) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=int(epochs*0.8)) + + + elif data in ['art_painting', 'cartoon', 'photo', 'sketch']: + # 加载数据集 + trset = data_loader.load_pacs(domain=data, split='train', twox=True, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + teset = data_loader.load_pacs(domain=data, split='val') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=4, shuffle=True, drop_last=True) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=4, shuffle=False) + if network == 'resnet18': + #cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + cls_net = resnet18Multiblock(classes=7,c_dim=2048).cuda() + input_dim = 2048 + # for param in cls_net.features.parameters(): + # param.requires_grad = False + # for name, parms in cls_net.named_parameters(): + # print('-->name:', name) + # print('-->grad_requirs:',parms.requires_grad) + # cls_opt = optim.SGD(cls_net.parameters(), lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + # print(cls_net.state_dict()) + + classifier_param = list(map(id, cls_net.class_classifier.parameters())) + classifierAttack_param = list(map(id, cls_net.classifier.parameters())) + #backbone_param = filter(lambda p: id(p) not in classifier_param and p.requires_grad, cls_net.parameters()) + backbone_param = filter(lambda p: id(p) not in classifier_param and id(p) not in classifierAttack_param and p.requires_grad, cls_net.parameters()) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(input_dim,1024,input_dim,4).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(7,70,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(7,70,1).cuda() + parameter_list.append({'params':backbone_param,'lr':0.01*lr}) + parameter_list.append({'params':cls_net.class_classifier.parameters(),'lr':lr}) + parameter_list.append({'params':cls_net.classifier.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + + #print("---------------------------------------------------------------------------------------") + #opt = optim.Adam(parameter_list) #version2 + opt = optim.SGD(parameter_list, momentum=0.9, nesterov=True, weight_decay=5e-4) #version1, inital + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.99999) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=15) + + elif 'synthia' in data: + # 加载数据集 + branch = data.split('_')[1] + trset = data_loader.load_synthia(branch) + trloader = DataLoader(trset, batch_size=batchsize, num_workers=8, shuffle=True) + teloader = DataLoader(trset, batch_size=batchsize, num_workers=8, shuffle=True) + imsize = [192, 320] + nclass = 14 + # 加载模型 + cls_net = fcn.FCN_resnet50(nclass=nclass).cuda() + cls_opt = optim.Adam(cls_net.parameters(), lr=lr)#, weight_decay=1e-4) # 对于synthia 加上weigh_decay会掉1-2个点 + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(cls_opt, epochs*len(trloader)) + + + cls_criterion = nn.CrossEntropyLoss() + adapt_criterion = nn.MSELoss() + # 开始训练 + best_acc = 0 + best_acc_t = 0 + scaler = GradScaler() + for epoch in range(epochs): + for name, param in cls_net.classifier.named_parameters(): + if param.requires_grad: + print(f'Epoch {epoch+1}, {name}, value: {param.data}, grad: {param.grad}') + + t1 = time.time() + loss_list = [] + cls_net.train() + # unloader = transforms.ToPILImage() + print(len(trloader)) + for i, (x_four,y) in enumerate(trloader): + b_sample_num = y.size(0) + x, x_RA, x_FA, x_CA, y = x_four[0].cuda(), x_four[1].cuda(), x_four[2].cuda(), x_four[3].cuda(), y.cuda() + b, c, h, w = x.shape + # x_FA_ = x_FA.transpose(1,2) + x_FA = x_FA.reshape(b*factor_num, c, h, w) + x_CA = x_CA.reshape(b*factor_num*var_num, c, h, w) + #learning mapping + y_repeat = y.unsqueeze(0).reshape(b_sample_num,1).repeat((1,factor_num)).reshape(1,b_sample_num*factor_num).squeeze() + # x_FA = FA(x).cuda().detach() + # x_CA = CA(x_RA).cuda().detach() + with autocast(): + # print('x:', x.shape) + # print('cls_net', cls_net) + # p,f = cls_net(x) + ''' + StyleAttack part + given: cls_net, cls_net.classiifer for gradients, input x + output: cls_loss, p, loss_adv, p_adv, loss_cls_ori + ''' + epsilon_list = [0.8, 0.08, 0.008] #eps1 + #epsilon_list = [0.2, 0.02, 0.002] #eps2 + #epsilon_list = [0.1, 0.01, 0.001] + #epsilon_list = [0.08, 0.008, 0.0008] + #epsilon_list = [0.02, 0.002, 0.0002] + ################################################################## + # 0. first cp x_adv from x_ori + #x_adv = x + x_adv = x_RA + #x_adv = x_FA + #x_adv = x_CA + + ################################################################## + # 1. styleAdv + cls_net.eval() + + #adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x, y, epsilon_list) + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_RA, y, epsilon_list) + #adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_FA, y, epsilon_list) + #adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = adversarial_attack_Incre(cls_net, cls_criterion, x_CA, y, epsilon_list) + + cls_net.zero_grad() + + ################################################################# + # 2. forward and get loss + cls_net.train() + + # forward x_ori + #L1, L2, L3, L4, p,f = cls_net(x) + L1, L2, L3, L4, p,f = cls_net(x_RA) + #L1, L2, L3, L4, p,f = cls_net(x_FA) + #L1, L2, L3, L4, p,f = cls_net(x_CA) + + cls_loss = cls_criterion(p, y) + + # ori cls global loss + scores_cls_ori = cls_net.classifier.forward(f) + loss_cls_ori = cls_criterion(scores_cls_ori, y) + + # forward x_adv + x_adv = x_adv.cuda() + x_adv_block1 = cls_net.forward_block1(x_adv) + x_adv_block1_newStyle = changeNewAdvStyle(x_adv_block1, adv_style_mean_block1, adv_style_std_block1, p_thred = P_THRED) + x_adv_block2 = cls_net.forward_block2(x_adv_block1_newStyle) + x_adv_block2_newStyle = changeNewAdvStyle(x_adv_block2, adv_style_mean_block2, adv_style_std_block2, p_thred = P_THRED) + x_adv_block3 = cls_net.forward_block3(x_adv_block2_newStyle) + x_adv_block3_newStyle = changeNewAdvStyle(x_adv_block3, adv_style_mean_block3, adv_style_std_block3, p_thred = P_THRED) + x_adv_block4 = cls_net.forward_block4(x_adv_block3_newStyle) + p_adv, x_adv_fea = cls_net.forward_rest(x_adv_block4) + loss_adv = cls_criterion(p_adv, y) + + loss_KL = consistency_loss(p, p_adv, 'KL3') + loss_styleattack = loss_adv + loss_KL + loss_cls_ori + + + ''' + original metaCausal + ''' + L1, L2, L3, L4, p,f = cls_net(x) + L1_FA, L2_FA, L3_FA, L4_FA, _,f_FA = cls_net(x_FA) + L1_RA, L2_RA, L3_RA, L4_RA, p_RA,f_RA = cls_net(x_RA) + L1_CA, L2_CA, L3_CA, L4_CA, p_CA,_ = cls_net(x_CA) + #learning mapping + f_repeat = f.repeat((1,factor_num)).reshape(f_FA.shape) + f_adapt = torch.zeros(f_FA.shape).cuda() + for b in range(b_sample_num): + for j in range(factor_num): + f_adapt[b*factor_num+j] = AdaptNet[j](f_FA[b*factor_num+j]) + p_adapt = cls_net(f_adapt, mode='c') + + #learning causality + if autoaug == 'CA_multiple': + p_RA_repeat = p_RA.repeat((1,factor_num*var_num)).reshape(p_CA.shape) + effect_context = p_RA_repeat - p_CA + effect_context = effect_context.reshape(b_sample_num,factor_num,var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*factor_num,-1) + # print("effect_context.shape:",effect_context.shape) + else: + p_RA_repeat = p_RA.repeat((1,factor_num)).reshape(p_CA.shape) + effect_context = p_RA_repeat - p_CA + weight = E_to_W(effect_context) + # weight = E_to_W(effect_context.detach()) + weight = weight.reshape(b_sample_num,factor_num) + alphas = F.softmax(weight,dim=1) + + f_adapt_RA = torch.zeros(f_RA.shape).cuda() + for b in range(b_sample_num): + for j in range(factor_num): + f_adapt_RA[b] = f_adapt_RA[b]+ alphas[b,j]*AdaptNet[j](f_RA[b]) + p_adapt_RA = cls_net(f_adapt_RA, mode='c') + + #cls_loss = cls_criterion(p, y) + re_mapping = adapt_criterion(f_adapt,f_repeat) + re_causal = adapt_criterion(f_adapt_RA,f) + cls_loss_mapping = cls_criterion(p_adapt, y_repeat) + cls_loss_causal = cls_criterion(p_adapt_RA, y) + + loss_metacausal = cls_loss + cls_loss_mapping + lambda_re*re_mapping + lambda_causal*(lambda_re*re_causal + cls_loss_causal) + + #print('loss_styleattack:', loss_styleattack, 'loss_metacausal:', loss_metacausal) + + loss = loss_styleattack + loss_metacausal + + + opt.zero_grad() + scaler.scale(loss).backward() + scaler.step(opt) + scaler.update() + loss_list.append([cls_loss.item(), cls_loss_mapping.item(),cls_loss_causal.item(), re_mapping.item(), re_causal.item()]) + + # 调整学习率 + if lr_scheduler in ['cosine', 'Exp', 'Step']: + writer.add_scalar('scalar/lr', opt.param_groups[0]["lr"], epoch) + print(opt.param_groups[0]["lr"]) + print("changing lr") + scheduler.step() + cls_loss, cls_loss_mapping, cls_loss_causal, re_mapping, re_causal = np.mean(loss_list, 0) + + # 测试，并保存最优模型 + cls_net.eval() + if data in ['mnist', 'mnist_t', 'cifar10', 'mnistvis', 'art_painting', 'cartoon', 'photo', 'sketch']: + teacc = evaluate(cls_net, teloader) + + elif 'synthia' in data: + teacc = evaluate_seg(cls_net, teloader, nclass) # 这里算的其实是 miou + + if best_acc < teacc: + print(f'---------------------saving model at epoch {epoch}----------------------------------------------------') + log_file.write(f'saving model at epoch {epoch}\n') + + best_acc = teacc + torch.save(cls_net.state_dict(),os.path.join(svroot, 'best_cls_net.pkl')) + for j in range(factor_num): + torch.save(AdaptNet[j].state_dict(),os.path.join(svroot, 'best_mapping_'+str(j)+'.pkl')) + torch.save(E_to_W.state_dict(), os.path.join(svroot, 'best_E_to_W.pkl')) + + # 保存日志 + t2 = time.time() + print(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f} cls_loss_mapping {cls_loss_mapping:.4f} cls_loss_causal {cls_loss_causal:.4f} re_mapping {re_mapping:.4f} re_causal {re_causal:.4f} /// teacc {teacc:2.2f} lr {opt.param_groups[0]["lr"]:.8f}') + log_file.write(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f} cls_loss_mapping {cls_loss_mapping:.4f} cls_loss_causal {cls_loss_causal:.4f} re_mapping {re_mapping:.4f} re_causal {re_causal:.4f} /// teacc {teacc:2.2f} lr {opt.param_groups[0]["lr"]:.8f} \n') + writer.add_scalar('scalar/cls_loss', cls_loss, epoch) + writer.add_scalar('scalar/cls_loss_mapping', cls_loss_mapping, epoch) + writer.add_scalar('scalar/cls_loss_causal', cls_loss_causal, epoch) + writer.add_scalar('scalar/re_mapping', re_mapping, epoch) + writer.add_scalar('scalar/re_causal', re_causal, epoch) + writer.add_scalar('scalar/teacc', teacc, epoch) + print(f'---------------------saving last model at epoch {epoch}----------------------------------------------------') + log_file.write(f'saving last model at epoch {epoch}\n') + torch.save(cls_net.state_dict(),os.path.join(svroot, 'last_cls_net.pkl')) + for j in range(factor_num): + torch.save(AdaptNet[j].state_dict(),os.path.join(svroot, 'last_mapping_'+str(j)+'.pkl')) + torch.save(E_to_W.state_dict(), os.path.join(svroot, 'last_E_to_W.pkl')) + + writer.close() + +def evalute_pacs(source_domain,cls_net,CA,AdaptNet,E_to_W): + cls_net.eval() + data_total = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in data_total if i!=source_domain] + acc_CA = np.zeros(len(target)) + for idx, data in enumerate(target): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=6, num_workers=0) + # 计算评价指标 + acc_CA[idx] = evaluate_causal(cls_net, teloader, CA, AdaptNet, E_to_W) + acc_avg_CA = sum(acc_CA)/len(target) + return acc_avg_CA,acc_CA + + +def evaluate(net, teloader): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + with torch.no_grad(): + x1 = x1.cuda() + #p1,_ = net(x1, mode='fc') + _, _, _, _, p1,_ = net(x1, mode='fc') + p1 = p1.argmax(dim=1) + ps.append(p1.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +def extract_feature(net, teloader, savedir): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + img_class = y1[0].cpu().numpy() + save_path = os.path.join(savedir,str(img_class)) + if not os.path.exists(save_path): + os.makedirs(save_path) + + with torch.no_grad(): + x1 = x1.cuda() + #p1,f1 = net(x1, mode='fc') + _, _, _, _, p1,f1 = net(x1, mode='fc') + save_name = save_path+os.sep+str(i)+'.npy' + np.save(save_name,f1.cpu()) + p1 = p1.argmax(dim=1) + ps.append(p1.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +def evaluate_causal(net, teloader, CA, AdaptNet, E_to_W): + ps = [] + ys = [] + p_orig = [] + y_orig = [] + for i,(x1, y1) in enumerate(teloader): + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + #p1,f_x1 = net(x1, mode='fc') + _, _, _, _, p1,f_x1 = net(x1, mode='fc') + x1_CA = CA(x1).cuda() + #p1_CA,_ = net(x1_CA, mode='fc') + _, _, _, _, p1_CA,_ = net(x1_CA, mode='fc') + p1_repeat = p1.repeat((1,CA.factor_num*CA.var_num)).reshape(p1_CA.shape) + effect_context = p1_repeat - p1_CA + effect_context = effect_context.reshape(b_sample_num,CA.factor_num,CA.var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*CA.factor_num,-1) + weight = E_to_W(effect_context) + weight = weight.reshape(b_sample_num,CA.factor_num) + alphas = F.softmax(weight,dim=1) + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(CA.factor_num): + f_adapt[b] = f_adapt[b]+ alphas[b,j]*AdaptNet[j](f_x1[b]) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +def extract_feature_do(net, teloader, CA, AdaptNet, E_to_W, savedir_base, savedir,source_flag): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + img_class = y1[0].cpu().numpy() + save_path_base = os.path.join(savedir_base,str(img_class)) + save_path = os.path.join(savedir,str(img_class)) + if not os.path.exists(save_path_base): + os.makedirs(save_path_base) + if not os.path.exists(save_path): + os.makedirs(save_path) + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + #p1,f_x1 = net(x1, mode='fc') + _, _, _, _, p1,f_x1 = net(x1, mode='fc') + save_name_base = save_path_base+os.sep+str(i)+'_base.npy' + print(save_name_base) + np.save(save_name_base,f_x1.cpu()) + x1_CA = CA(x1).cuda() + #p1_CA,_ = net(x1_CA, mode='fc') + _, _, _, _, p1_CA,_ = net(x1_CA, mode='fc') + p1_repeat = p1.repeat((1,CA.factor_num*CA.var_num)).reshape(p1_CA.shape) + effect_context = p1_repeat - p1_CA + effect_context = effect_context.reshape(b_sample_num,CA.factor_num,CA.var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*CA.factor_num,-1) + weight = E_to_W(effect_context) + weight = weight.reshape(b_sample_num,CA.factor_num) + alphas = F.softmax(weight,dim=1) + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(CA.factor_num): + f_adapt[b] = f_adapt[b]+ alphas[b,j]*AdaptNet[j](f_x1[b]) + if not source_flag: + save_name = save_path+os.sep+str(i)+'.npy' + print(save_name) + np.save(save_name,f_adapt.cpu()) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + + +def evaluate_mapping(net, teloader, FA, AdaptNet, source=False): + correct, count = 0, 0 + ps = [] + ys = [] + pt = [] + yt = [] + factor_num = FA.factor_num + for j in range(factor_num): + ps.append([]) + ys.append([]) + pt.append([]) + yt.append([]) + ps.append([]) + ys.append([]) + # print(len(ps),len(ys)) + for i,(x1, y1) in enumerate(teloader): + with torch.no_grad(): + x1 = x1.cuda() + b = x1.size(0) + if source: + x_FA = FA(x1).cuda() + #_, f = net(x_FA, mode='fc') + _, _, _, _, _, f = net(x_FA, mode='fc') + #p,_ = net(x1, mode='fc') + _, _, _, _, p,_ = net(x1, mode='fc') + p = p.argmax(dim=1) + ps[-1].append(p.detach().cpu().numpy()) + ys[-1].append(y1.numpy()) + else: + #p, f = net(x1, mode='fc') + _, _, _, _, p, f = net(x1, mode='fc') + f = f.repeat((1,factor_num)).reshape((-1,f.size(1))) + p = p.argmax(dim=1) + ps[-1].append(p.detach().cpu().numpy()) + ys[-1].append(y1.numpy()) + for b_ in range(b): + for j in range(factor_num): + f_adapt = AdaptNet[j](f[b_*factor_num+j]) + #f_adapt = torch.mm(AdaptNet[j].W1,f_FA[b_*factor_num+j].unsqueeze(1)).squeeze() + p1 = net(f_adapt, mode='c') + p1 = p1.argmax(dim=0) + ps[j].append(p1.detach().cpu()) + ys[j].append(y1[b_]) + p1_t = net(f[b_*factor_num+j], mode='c') + # print("p1_t.shape:",p1_t.shape) + p1_t = p1_t.argmax(dim=0) + pt[j].append(p1_t.detach().cpu()) + yt[j].append(y1[b_]) + # 计算评价指标 + acc = np.zeros(factor_num+1) + acc_t = np.zeros(factor_num+1) + for j in range(factor_num): + pred = torch.stack(ps[j]) + label = torch.stack(ys[j]) + acc[j] = (pred==label).sum()/float(len(ys[j]))*100 + predt = torch.stack(pt[j]) + labelt = torch.stack(yt[j]) + acc_t[j] = (predt==labelt).sum()/float(len(yt[j]))*100 + pred = np.concatenate(ps[-1]) + label = np.concatenate(ys[-1]) + acc[-1] = np.mean(pred==label)*100 + # print("acc:",acc) + return acc, acc_t + +def evaluate_causal_with_entropy(net, teloader, CA, AdaptNet): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + #p1,f_x1 = net(x1, mode='fc') + _, _, _, _, p1,f_x1 = net(x1, mode='fc') + x1_CA = CA(x1).cuda() + #p1_CA, _ = net(x1_CA, mode='fc') + _, _, _, _, p1_CA, _ = net(x1_CA, mode='fc') + p1_repeat = p1.repeat((1,CA.factor_num*CA.var_num)).reshape(p1_CA.shape) + effect_context = p1_repeat - p1_CA + effect_context = effect_context.reshape(b_sample_num,CA.factor_num,CA.var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*CA.factor_num,-1) + effect_context = F.softmax(effect_context,dim=1) + # weight = calc_ent(effect_context) + weight = torch.sum(-effect_context*(torch.log2(effect_context)),dim=1) + weight = weight.reshape(b_sample_num,CA.factor_num) + alphas = F.softmax(-weight,dim=1) + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(CA.factor_num): + f_adapt[b] = f_adapt[b]+ alphas[b,j]*AdaptNet[j](f_x1[b]) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc +def evaluate_causal_with_average(net, teloader, factor_num, AdaptNet): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + #p1,f_x1 = net(x1, mode='fc') + _, _, _, _, p1,f_x1 = net(x1, mode='fc') + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(factor_num): + f_adapt[b] = f_adapt[b]+ float(1/factor_num)*AdaptNet[j](f_x1[b]) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +if __name__=='__main__': + print('here1') + experiment() \ No newline at end of file diff --git a/Meta-causal/code-withStyleAttack/main_test_digit_v13.py b/Meta-causal/code-withStyleAttack/main_test_digit_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..b591743d989d0b593cfee714221972cf714c4789 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/main_test_digit_v13.py @@ -0,0 +1,146 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import mnist_net_my as mnist_net +from mnist_net_multiblock import ConvNetMultiblock +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate,evaluate_causal,evaluate_causal_with_entropy,evaluate_mapping,evaluate_causal_with_average +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--channels', type=int, default=3) +@click.option('--factor_num', type=int, default=16) +@click.option('--stride', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--eval_mapping', type=bool, default=True, help='是否查看mapping学习效果') +def main(gpu, svroot, svpath, channels, factor_num,stride, epoch, eval_mapping): + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + +def evaluate_digit(gpu, svroot, svpath, channels=3, factor_num=16,stride=5,epoch='best', eval_mapping=True): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if channels == 3: + #cls_net = mnist_net.ConvNet().cuda() + cls_net = ConvNetMultiblock().cuda() + elif channels == 1: + #cls_net = mnist_net.ConvNet(imdim=channels).cuda() + cls_net = ConvNetMultiblock(imdim=channels).cuda() + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + cls_net.load_state_dict(saved_weight) + #cls_net.eval() + # 加载adaptation模型 + FA = causalaugment.FactualAugment(m=4, factor_num=factor_num) + CA = causalaugment.MultiCounterfactualAugment(factor_num,stride) + # Color_mapping = adaptor.mapping().cuda() + # Contrast_mapping = adaptor.mapping().cuda() + # Brightness_mapping = adaptor.mapping().cuda() + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) + # saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + mapping = adaptor_v2.mapping(1024,512,1024,2).cuda() + mapping.load_state_dict(saved_weight) + AdaptNet.append(mapping) + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) + + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + # Color_mapping.load_state_dict(saved_weight['Color_mapping']) + # Contrast_mapping.load_state_dict(saved_weight['Contrast_mapping']) + # Brightness_mapping.load_state_dict(saved_weight['Brightness_mapping']) + # saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + E_to_W.load_state_dict(saved_weight) + + # 测试 + str2fun = { + 'mnist': data_loader.load_mnist, + 'mnist_m': data_loader.load_mnist_m, + 'usps': data_loader.load_usps, + 'svhn': data_loader.load_svhn, + 'syndigit': data_loader.load_syndigit, + } + columns = ['mnist', 'svhn', 'mnist_m', 'syndigit','usps'] + target = ['svhn', 'mnist_m', 'syndigit','usps'] + if eval_mapping: + index = FA.factor_list + index.append('w/o do (original x)') + else: + index = ['w/o do (original x)'] + index_ours = ['do'] + data_result = {} + data_result_ours = {} + cls_net.eval() + for idx, data in enumerate(columns): + teset = str2fun[data]('test', channels=channels) + teloader = DataLoader(teset, batch_size=8, num_workers=0) + # 计算评价指标 + acc_CA = evaluate_causal(cls_net, teloader, CA, AdaptNet, E_to_W) + data_result_ours[data] = acc_CA + #最后一维度是原始数据 + if eval_mapping: + if data == 'mnist': + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=True) + acc_avg = np.zeros(teacc_FA_aftermapping.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=False) + acc_avg = acc_avg + teacc_FA_aftermapping + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data]=teacc_FA_aftermapping + data_result[data+'_FA'] = acc_FA + else: + teacc = evaluate(cls_net, teloader) + if data == 'mnist': + acc_avg = np.zeros(teacc.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + acc_avg = acc_avg + teacc + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + acc_avg_CA = acc_avg_CA/float(len(target)) + + data_result['Avg'] = acc_avg + data_result_ours['Avg'] = acc_avg_CA + + df = pd.DataFrame(data_result,index = index) + df_ours = pd.DataFrame(data_result_ours,index = index_ours) + print(df) + print(df_ours) + if svpath is not None: + df.to_csv(svpath) + df_ours.to_csv(svpath, mode='a') + +if __name__=='__main__': + main() + diff --git a/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py b/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..8ab72c2e98d8f3c1a2bad66263b7e0444dabfc1e --- /dev/null +++ b/Meta-causal/code-withStyleAttack/main_test_pacs_v13.py @@ -0,0 +1,141 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import resnet as resnet +from backbone_multiblock import resnet18Multiblock +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate,evaluate_causal,evaluate_causal_with_entropy,evaluate_mapping,evaluate_causal_with_average +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--source_domain', type=str, default='art_painting', help='source domain') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--factor_num', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--stride', type=int, default=5) +@click.option('--eval_mapping', type=bool, default=False, help='是否查看mapping学习效果') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') +def main(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network): + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + +def evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num=16, epoch='best', stride=5,eval_mapping=False, network='resnet18'): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if network == 'resnet18': + #cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + cls_net = resnet18Multiblock(classes=7,c_dim=2048).cuda() + input_dim = 2048 + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + cls_net.load_state_dict(saved_weight) + cls_net.eval() + # 加载adaptation模型 + FA = causalaugment.FactualAugment(m=4, factor_num=factor_num) + CA = causalaugment.MultiCounterfactualAugment(factor_num,stride) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) + # saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + mapping = adaptor_v2.mapping(input_dim,1024,input_dim,4).cuda() + mapping.load_state_dict(saved_weight) + AdaptNet.append(mapping) + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) + E_to_W = adaptor_v2.effect_to_weight(7,70,1).cuda() + # Color_mapping.load_state_dict(saved_weight['Color_mapping']) + # Contrast_mapping.load_state_dict(saved_weight['Contrast_mapping']) + # Brightness_mapping.load_state_dict(saved_weight['Brightness_mapping']) + # saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + E_to_W.load_state_dict(saved_weight) + + # 测试 + # str2fun = { + # 'art_painting': data_loader.load_pacs, + # 'cartoon': data_loader.load_pacs, + # 'photo': data_loader.load_pacs, + # 'sketch': data_loader.load_pacs, + # } + columns = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in columns if i!=source_domain] + columns = [source_domain] + target + print("columns:",columns) + if eval_mapping: + index = FA.factor_list + index.append('w/o do (original x)') + else: + index = ['w/o do (original x)'] + index_ours = ['do'] + data_result = {} + data_result_ours = {} + + for idx, data in enumerate(columns): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=4, num_workers=0) + # 计算评价指标 + acc_CA = evaluate_causal(cls_net, teloader, CA, AdaptNet, E_to_W) + data_result_ours[data] = acc_CA + #最后一维度是原始数据 + if eval_mapping: + if data == source_domain: + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=True) + acc_avg = np.zeros(teacc_FA_aftermapping.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=False) + acc_avg = acc_avg + teacc_FA_aftermapping + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data]=teacc_FA_aftermapping + data_result[data+'_FA'] = acc_FA + else: + teacc = evaluate(cls_net, teloader) + if data == source_domain: + acc_avg = np.zeros(teacc.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + acc_avg = acc_avg + teacc + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + acc_avg_CA = acc_avg_CA/float(len(target)) + + data_result['Avg'] = acc_avg + data_result_ours['Avg'] = acc_avg_CA + + df = pd.DataFrame(data_result,index = index) + df_ours = pd.DataFrame(data_result_ours,index = index_ours) + print(df) + print(df_ours) + if svpath is not None: + df.to_csv(svpath) + df_ours.to_csv(svpath, mode='a') +if __name__=='__main__': + main() + diff --git a/Meta-causal/code-withStyleAttack/mnist_net_multiblock.py b/Meta-causal/code-withStyleAttack/mnist_net_multiblock.py new file mode 100644 index 0000000000000000000000000000000000000000..603853616aba78111bb96226c6502af1abf5890c --- /dev/null +++ b/Meta-causal/code-withStyleAttack/mnist_net_multiblock.py @@ -0,0 +1,96 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F + +class ConvNetMultiblock(nn.Module): + def __init__(self, imdim=3): + super(ConvNetMultiblock, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 1024) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(1024, 10) + # self.cls_head_tgt = nn.Linear(1024, 10) + # self.pro_head = nn.Linear(1024, 128) + + # for style attacking + self.classifier = nn.Linear(1024, 10) + + + def forward(self,x,mode='fc'): + if mode == 'c': + return self.class_classifier(x) + else: + layer1 = self.forward_block1(x) + layer2 = self.forward_block2(layer1) + layer3 = self.forward_block3(layer2) + layer4 = self.forward_block4(layer3) + p, f= self.forward_rest(layer4) + return layer1, layer2, layer3, layer4, p, f + + + def forward_block1(self, x): + self.in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + return out1 + + def forward_block2(self, out1): + out2 = self.mp(self.relu2(self.conv2(out1))) + return out2 + + def forward_block3(self, out2): + out2 = out2.view(self.in_size, -1) + out3 = self.relu3(self.fc1(out2)) + return out3 + + def forward_block4(self,out3): + out4_worelu = self.fc2(out3) + return out4_worelu + + def forward_rest(self, out4_worelu): + out4 = self.relu4(out4_worelu) + p = self.cls_head_src(out4) + return p, out4_worelu + + + def forward(self, x, mode='fc'): + if mode == 'c': + out4 = self.relu4(x) + p = self.cls_head_src(out4) + return p + else: + layer1 = self.forward_block1(x) + layer2 = self.forward_block2(layer1) + layer3 = self.forward_block3(layer2) + layer4 = self.forward_block4(layer3) + p, out4_worelu = self.forward_rest(layer4) + return layer1, layer2, layer3, layer4, p, out4_worelu + + + + +if __name__ =='__main__': + print('%'*100) + print('---test ConvNet original--') + from network.mnist_net_my import ConvNet + cls_net = ConvNet().cuda() + print('cls_net:', cls_net) + x = torch.randn([16,3,32,32]).cuda() + p, f = cls_net(x) + print(p.shape, f.shape) + + + print('---test RN18 multiblock--') + cls_net = ConvNetMultiblock().cuda() + print('cls_net:', cls_net) + x = torch.randn([16,3,32,32]).cuda() + L1, L2, L3, L4, p, f = cls_net(x) + print(L1.shape, L2.shape, L3.shape, L4.shape, p.shape, f.shape) diff --git a/Meta-causal/code-withStyleAttack/network/adaptor_v2.py b/Meta-causal/code-withStyleAttack/network/adaptor_v2.py new file mode 100644 index 0000000000000000000000000000000000000000..ce47dbd1a24f9e2f741d8a82061b62b86d3dba41 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/network/adaptor_v2.py @@ -0,0 +1,63 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +import numpy as np + +class mapping(nn.Module): + def __init__(self, input_dim=1024, hidden_dim = 512, out_dim=1024, layernum=4): + ''' + ''' + super().__init__() + self.layernum = layernum + if layernum == 4: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim) + self.fc3 = nn.Linear(hidden_dim, hidden_dim) + self.fc4 = nn.Linear(hidden_dim, out_dim) + elif layernum == 2: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, out_dim) + self.relu = nn.ReLU(inplace=True) + + def forward(self, x): + ''' x ''' + if self.layernum == 4: + x = self.relu(self.fc1(x)) + x = self.relu(self.fc2(x)) + x = self.relu(self.fc3(x)) + x = self.fc4(x) + elif self.layernum == 2: + x = self.relu(self.fc1(x)) + x = self.fc2(x) + return x + + +class effect_to_weight(nn.Module): + def __init__(self, input_dim = 512, hidden_dim = 256, out_dim = 1, layernum=2, hidden_dim2 = 128): + ''' + ''' + super().__init__() + + self.layernum = layernum + if layernum == 2: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, out_dim) + elif layernum == 3: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim2) + self.fc3 = nn.Linear(hidden_dim2, out_dim) + self.relu = nn.ReLU(inplace=True) + + def forward(self, x): + ''' x ''' + if self.layernum == 2: + x = self.relu(self.fc1(x)) + x = self.fc2(x) + else: + x = self.relu(self.fc1(x)) + x = self.relu(self.fc2(x)) + x = self.fc3(x) + return x + + diff --git a/Meta-causal/code-withStyleAttack/network/mnist_net_my.py b/Meta-causal/code-withStyleAttack/network/mnist_net_my.py new file mode 100644 index 0000000000000000000000000000000000000000..15e2e677280fdd2211b559f9f1bafd2fb66b5ef4 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/network/mnist_net_my.py @@ -0,0 +1,104 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F + +class ConvNet(nn.Module): + ''' 网络结构和cvpr2020的 M-ADA 方法一致 ''' + def __init__(self, imdim=3): + super(ConvNet, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 1024) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(1024, 10) + # self.cls_head_tgt = nn.Linear(1024, 10) + # self.pro_head = nn.Linear(1024, 128) + + def forward(self, x, mode='fc'): + if mode == 'c': + out4 = self.relu4(x) + p = self.cls_head_src(out4) + return p + elif mode == 'fc': + in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + out2 = self.mp(self.relu2(self.conv2(out1))) + out2 = out2.view(in_size, -1) + out3 = self.relu3(self.fc1(out2)) + out4_worelu = self.fc2(out3) + out4 = self.relu4(out4_worelu) + p = self.cls_head_src(out4) + return p, out4_worelu + + # if mode == 'test': + # p = self.cls_head_src(out4) + # return p + # elif mode == 'train': + # p = self.cls_head_src(out4) + # # z = self.pro_head(out4) + # # z = F.normalize(z) + # return p,out4_worelu + # elif mode == 'p_f': + # p = self.cls_head_src(out4) + # return p, out4 + #elif mode == 'target': + # p = self.cls_head_tgt(out4) + # z = self.pro_head(out4) + # z = F.normalize(z) + # return p,z + +class ConvNetVis(nn.Module): + ''' 方便可视化，特征提取器输出2-d特征 + ''' + def __init__(self, imdim=3): + super(ConvNetVis, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 2) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(2, 10) + self.cls_head_tgt = nn.Linear(2, 10) + self.pro_head = nn.Linear(2, 128) + + def forward(self, x, mode='test'): + + in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + out2 = self.mp(self.relu2(self.conv2(out1))) + out2 = out2.view(in_size, -1) + out3 = self.relu3(self.fc1(out2)) + out4 = self.relu4(self.fc2(out3)) + + if mode == 'test': + p = self.cls_head_src(out4) + return p + elif mode == 'train': + p = self.cls_head_src(out4) + z = self.pro_head(out4) + z = F.normalize(z) + return p,z + elif mode == 'p_f': + p = self.cls_head_src(out4) + return p, out4 + #elif mode == 'target': + # p = self.cls_head_tgt(out4) + # z = self.pro_head(out4) + # z = F.normalize(z) + # return p,z + + diff --git a/Meta-causal/code-withStyleAttack/network/resnet.py b/Meta-causal/code-withStyleAttack/network/resnet.py new file mode 100644 index 0000000000000000000000000000000000000000..a0beda0f3e0ac68574f3194e368737e79854b934 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/network/resnet.py @@ -0,0 +1,102 @@ +from torch import nn +from torch.utils import model_zoo +#from torchvision.models.resnet import BasicBlock, model_urls, Bottleneck +from torchvision.models.resnet import BasicBlock, Bottleneck + +import torch +import ssl +# from torch import nn as nn +# from utils.util import * + +ssl._create_default_https_context = ssl._create_unverified_context + +all = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101','resnet152'] + +model_urls = { +'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', +'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', +'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', +'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', +'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth', +} + + +class ResNet(nn.Module): + def __init__(self, block, layers,classes=7,c_dim=512): + self.inplanes = 64 + super(ResNet, self).__init__() + self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, + bias=False) + self.bn1 = nn.BatchNorm2d(64) + self.relu = nn.ReLU(inplace=True) + self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + self.layer1 = self._make_layer(block, 64, layers[0]) + self.layer2 = self._make_layer(block, 128, layers[1], stride=2) + self.layer3 = self._make_layer(block, 256, layers[2], stride=2) + self.layer4 = self._make_layer(block, 512, layers[3], stride=2) + self.avgpool = nn.AvgPool2d(7, stride=1) + self.class_classifier = nn.Linear(c_dim, classes) + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + nn.init.constant_(m.weight, 1) + nn.init.constant_(m.bias, 0) + + def _make_layer(self, block, planes, blocks, stride=1): + downsample = None + if stride != 1 or self.inplanes != planes * block.expansion: + downsample = nn.Sequential( + nn.Conv2d(self.inplanes, planes * block.expansion, + kernel_size=1, stride=stride, bias=False), + nn.BatchNorm2d(planes * block.expansion), + ) + + layers = [] + layers.append(block(self.inplanes, planes, stride, downsample)) + self.inplanes = planes * block.expansion + for i in range(1, blocks): + layers.append(block(self.inplanes, planes)) + + return nn.Sequential(*layers) + + def forward(self, x, mode='fc'): + if mode == 'c': + return self.class_classifier(x) + else: + x = self.conv1(x) + x = self.bn1(x) + x = self.relu(x) + x = self.maxpool(x) + + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + x = self.avgpool(x) + x = x.view(x.size(0), -1) + # print("x.shape:",x.shape) + return self.class_classifier(x), x + + +def resnet18(pretrained=True, **kwargs): + """Constructs a ResNet-18 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet18']), strict=False) + return model + +def resnet50(pretrained=True, **kwargs): + """Constructs a ResNet-50 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet50']), strict=False) + return model diff --git a/Meta-causal/code-withStyleAttack/network/wideresnet.py b/Meta-causal/code-withStyleAttack/network/wideresnet.py new file mode 100644 index 0000000000000000000000000000000000000000..1ca130a5f278c3b63f43b589db6ebd18d6e91593 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/network/wideresnet.py @@ -0,0 +1,86 @@ +import math +import torch +import torch.nn as nn +import torch.nn.functional as F + + +class BasicBlock(nn.Module): + def __init__(self, in_planes, out_planes, stride, dropRate=0.0): + super(BasicBlock, self).__init__() + self.bn1 = nn.BatchNorm2d(in_planes) + self.relu1 = nn.ReLU(inplace=True) + self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, + padding=1, bias=False) + self.bn2 = nn.BatchNorm2d(out_planes) + self.relu2 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=3, stride=1, + padding=1, bias=False) + self.droprate = dropRate + self.equalInOut = (in_planes == out_planes) + self.convShortcut = (not self.equalInOut) and nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, + padding=0, bias=False) or None + def forward(self, x): + if not self.equalInOut: + x = self.relu1(self.bn1(x)) + else: + out = self.relu1(self.bn1(x)) + out = self.relu2(self.bn2(self.conv1(out if self.equalInOut else x))) + if self.droprate > 0: + out = F.dropout(out, p=self.droprate, training=self.training) + out = self.conv2(out) + return torch.add(x if self.equalInOut else self.convShortcut(x), out) + +class NetworkBlock(nn.Module): + def __init__(self, nb_layers, in_planes, out_planes, block, stride, dropRate=0.0): + super(NetworkBlock, self).__init__() + self.layer = self._make_layer(block, in_planes, out_planes, nb_layers, stride, dropRate) + def _make_layer(self, block, in_planes, out_planes, nb_layers, stride, dropRate): + layers = [] + for i in range(int(nb_layers)): + layers.append(block(i == 0 and in_planes or out_planes, out_planes, i == 0 and stride or 1, dropRate)) + return nn.Sequential(*layers) + def forward(self, x): + return self.layer(x) + +class WideResNet(nn.Module): + def __init__(self, depth, num_classes, widen_factor=1, dropRate=0.0): + super(WideResNet, self).__init__() + nChannels = [16, 16*widen_factor, 32*widen_factor, 64*widen_factor] + assert((depth - 4) % 6 == 0) + n = (depth - 4) / 6 + block = BasicBlock + # 1st conv before any network block + self.conv1 = nn.Conv2d(3, nChannels[0], kernel_size=3, stride=1, + padding=1, bias=False) + # 1st block + self.block1 = NetworkBlock(n, nChannels[0], nChannels[1], block, 1, dropRate) + # 2nd block + self.block2 = NetworkBlock(n, nChannels[1], nChannels[2], block, 2, dropRate) + # 3rd block + self.block3 = NetworkBlock(n, nChannels[2], nChannels[3], block, 2, dropRate) + # global average pooling and classifier + self.bn1 = nn.BatchNorm2d(nChannels[3]) + self.relu = nn.ReLU(inplace=True) + self.fc = nn.Linear(nChannels[3], num_classes) + self.nChannels = nChannels[3] + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + m.weight.data.fill_(1) + m.bias.data.zero_() + elif isinstance(m, nn.Linear): + m.bias.data.zero_() + def forward(self, x, mode='fc'): + if mode == 'c': + return self.fc(x) + else: + out = self.conv1(x) + out = self.block1(out) + out = self.block2(out) + out = self.block3(out) + out = self.relu(self.bn1(out)) + out = F.avg_pool2d(out, 8) + out = out.view(-1, self.nChannels) + return self.fc(out), out diff --git a/Meta-causal/code-withStyleAttack/run_PACS/run_my_joint_v13_test.sh b/Meta-causal/code-withStyleAttack/run_PACS/run_my_joint_v13_test.sh new file mode 100644 index 0000000000000000000000000000000000000000..a2008370982092645507504273e7079568f716c6 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/run_PACS/run_my_joint_v13_test.sh @@ -0,0 +1,47 @@ + +# $1 gpuid +# $2 runid + +# base方法 +cd .. +epochs=70 +clsadapt=True +lr=0.01 +factor_num=16 +lr_scheduler=cosine +lambda_causal=1 +lambda_re=1 +batchsize=6 +stride=5 +randm=True +randn=True +autoaug=CA_multiple +network=resnet18 +#UniqueExpName=WithStyleAttackExp1 +#UniqueExpName=WithStyleAttackExp1_eps5_RA_SGD +#UniqueExpName=WithStyleAttackExp1_eps5_RA_Adam +#UniqueExpName=WithStyleAttackExp1_eps1_RA +#UniqueExpName=WithStyleAttackExp1_eps2 +UniqueExpName=WithStyleAttackExp1_eps1_RA_repeat + +root=/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS/ +#data=art_painting +#data=cartoon +data=photo +#data=sketch +svroot=$root/${data}/${autoaug}_${factor_num}fa_v2_ep${epochs}_lr${lr}_${lr_scheduler}_base0.01_bs${batchsize}_lamCa_${lambda_causal}_lamRe${lambda_re}_adt4_cls1_EW2_70_rm${randm}_rn${randn}_str${stride}_${UniqueExpName} + +python3 main_my_joint_v13_auto.py --gpu $1 --data ${data} --epochs $epochs --autoaug $autoaug --lambda_causal ${lambda_causal} --lambda_re ${lambda_re} --lr $lr --svroot $svroot --clsadapt $clsadapt --factor_num $factor_num --lr_scheduler ${lr_scheduler} --batchsize ${batchsize} --network ${network} --randm ${randm} --randn ${randn} --stride ${stride} + +test_epoch=best +python3 main_test_pacs_v13.py --gpu $1 --source_domain $data --svroot $svroot --svpath $svroot/${data}_${factor_num}factor_${test_epoch}_test_check.csv --factor_num $factor_num --epoch $test_epoch \ + --network ${network} --stride ${stride} + + +test_epoch=last +python3 main_test_pacs_v13.py --gpu $1 --source_domain $data --svroot $svroot --svpath $svroot/${data}_${factor_num}factor_${test_epoch}_test_check.csv --factor_num $factor_num --epoch $test_epoch \ + --network ${network} --stride ${stride} + + + + diff --git a/Meta-causal/code-withStyleAttack/run_digits/run_my_joint_test.sh b/Meta-causal/code-withStyleAttack/run_digits/run_my_joint_test.sh new file mode 100644 index 0000000000000000000000000000000000000000..7786bfcd65b79e536da79bc1add2fc390e429bb0 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/run_digits/run_my_joint_test.sh @@ -0,0 +1,55 @@ + +# $1 gpuid + +cd .. +epochs=500 +#epochs=250 +clsadapt=True +lr=1e-4 +lr_scheduler=Step +factor_num=14 +lambda_causal=1 +lambda_re=1 +batchsize=32 +stride=3 +randm=True +randn=True +autoaug=CA_multiple +#UniqueExpName=WithStyleAttackExp1_epoch250 +#UniqueExpName=WithStyleAttackExp1 +#UniqueExpName=WithStyleAttackExp1_RA +#UniqueExpName=WithStyleAttackExp1_FA +#UniqueExpName=WithStyleAttackExp1_CA +#UniqueExpName=WithStyleAttackExp1_onlyblock1 +#UniqueExpName=WithStyleAttackExp1_onlyblock2 +#UniqueExpName=WithStyleAttackExp1_adam +#UniqueExpName=WithStyleAttackExp1_eps2 +#UniqueExpName=WithStyleAttackExp1_eps3 +#UniqueExpName=WithStyleAttackExp1_eps4 +#UniqueExpName=WithStyleAttackExp1_eps5 +#UniqueExpName=WithStyleAttackExp1_skip2 +#UniqueExpName=WithStyleAttackExp1_skip3 +#UniqueExpName=WithStyleAttackExp1_skip4 + +#UniqueExpName=WithStyleAttackExp1_eps5_repeat +#UniqueExpName=WithStyleAttackExp1_eps5_RA + +UniqueExpName=WithStyleAttackExp1_eps1_RA + + +root=/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit +svroot=$root/${autoaug}_${factor_num}fa_all_ep${epochs}_lr${lr}_lr_scheduler${lr_scheduler}0.8_bs${batchsize}_lamCa_${lambda_causal}_lamRe_${lambda_re}_cls1_adt2_EW2_100_rm${randm}_rn${randn}_str${stride}_${UniqueExpName} + +python3 main_my_joint_v13_auto.py --gpu $1 --data mnist --epochs $epochs --autoaug $autoaug --lambda_causal ${lambda_causal} --lambda_re ${lambda_re} --lr $lr --lr_scheduler $lr_scheduler --svroot $svroot --clsadapt $clsadapt --factor_num $factor_num --batchsize ${batchsize} --randm ${randm} --randn ${randn} --stride ${stride} + +test_epoch=best +python3 main_test_digit_v13.py --gpu $1 --svroot $svroot --svpath $svroot/${factor_num}factor_${test_epoch}.csv --factor_num $factor_num --epoch $test_epoch --stride ${stride} + +test_epoch=last +python3 main_test_digit_v13.py --gpu $1 --svroot $svroot --svpath $svroot/${factor_num}factor_${test_epoch}.csv --factor_num $factor_num --epoch $test_epoch --stride ${stride} + + + + + + diff --git a/Meta-causal/code-withStyleAttack/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala b/Meta-causal/code-withStyleAttack/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala new file mode 100644 index 0000000000000000000000000000000000000000..aa44ae0c513b57a8501e9bb1af27dc442b72f7d7 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a44a49f0a1b3c59b9763c67ea85708ef8b56cae5fe4336f0383f5f71ba0dac84 +size 40 diff --git a/Meta-causal/code-withStyleAttack/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log b/Meta-causal/code-withStyleAttack/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log new file mode 100644 index 0000000000000000000000000000000000000000..f26feaaef352ae5821e49b7fbc4b1720f8735f38 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log @@ -0,0 +1 @@ +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': 'saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} diff --git a/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala new file mode 100644 index 0000000000000000000000000000000000000000..5ec21b3afdf0e11651cc768f4f55ea6269b887f5 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7946f93077ec2136f75fc090a5762ce810be71cc78d5201e8a671217a678c563 +size 40 diff --git a/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala new file mode 100644 index 0000000000000000000000000000000000000000..620f9ba109e77ed90b7676c138933f814245e7f1 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2021c61739fbe1f9c066067b4e5903d8d2d6c1c44865e1e9c61449eb3d90327 +size 40 diff --git a/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala new file mode 100644 index 0000000000000000000000000000000000000000..3144b1448112cff1aa0c26e0d825b50698f41d65 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fafb4b17d350157735eb6174ff44bafcea7ab8bf86948df3421447ef45ffcae3 +size 40 diff --git a/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log new file mode 100644 index 0000000000000000000000000000000000000000..f4c211545f0d2b537d3dcf980579f604a33419a7 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log @@ -0,0 +1 @@ +{'gpu': '0çç', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': 'saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} diff --git a/Meta-causal/code-withStyleAttack/submit_digits.sh b/Meta-causal/code-withStyleAttack/submit_digits.sh new file mode 100644 index 0000000000000000000000000000000000000000..c768ad7fcc05ecaf8978e73448341c9671de92f4 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/submit_digits.sh @@ -0,0 +1,23 @@ +#!/bin/bash +#SBATCH --job-name=metatrainRN +#SBATCH --nodes=1 # Request 1 node +#SBATCH --ntasks=1 # Number of tasks (total) +#SBATCH --cpus-per-task=8 # Number of CPU cores (threads) per task +#SBATCH --mem-per-cpu=4G # Memory limit per CPU core (there is no --mem-per-task) +#SBATCH --time=96:00:00 # Job timeout +#SBATCH --gpus-per-node=l4-24g:1 +#SBATCH --nodelist=gcpl4-eu-1 +#SBATCH --output=%j.log # Redirect stdout to a log file +#SBATCH --error=%j.error # Redirect stderr to a separate error log file + + +srun --nodes 1 --ntasks-per-node 1 -- \ +mkenv -f ../env_mc.yml -- \ +sh -c "cd run_digits +bash run_my_joint_test.sh 0 +" + + + + + diff --git a/Meta-causal/code-withStyleAttack/submit_pacs.sh b/Meta-causal/code-withStyleAttack/submit_pacs.sh new file mode 100644 index 0000000000000000000000000000000000000000..d7c22cec7befd329e4eac93fde6845fa09b5cedc --- /dev/null +++ b/Meta-causal/code-withStyleAttack/submit_pacs.sh @@ -0,0 +1,23 @@ +#!/bin/bash +#SBATCH --job-name=metatrainRN +#SBATCH --nodes=1 # Request 1 node +#SBATCH --ntasks=1 # Number of tasks (total) +#SBATCH --cpus-per-task=8 # Number of CPU cores (threads) per task +#SBATCH --mem-per-cpu=4G # Memory limit per CPU core (there is no --mem-per-task) +#SBATCH --time=96:00:00 # Job timeout +#SBATCH --gpus-per-node=l4-24g:1 +#SBATCH --nodelist=gcpl4-eu-4 +#SBATCH --output=%j.log # Redirect stdout to a log file +#SBATCH --error=%j.error # Redirect stderr to a separate error log file + + +srun --nodes 1 --ntasks-per-node 1 -- \ +mkenv -f ../env_mc.yml -- \ +sh -c "cd run_PACS +bash run_my_joint_v13_test.sh 0 +" + + + + + diff --git a/Meta-causal/code-withStyleAttack/tool_func.py b/Meta-causal/code-withStyleAttack/tool_func.py new file mode 100644 index 0000000000000000000000000000000000000000..0dc0220a5fd285312e0139f4601efefe2c34af90 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/tool_func.py @@ -0,0 +1,217 @@ + +import torch +import torch.nn as nn +import random +import numpy as np +import torch.nn.functional as F + +EPS=0.00001 +P_THRED = 0.4 +#P_THRED = 0.2 +#P_THRED = 0.6 +#P_THRED = 0.8 +START_EPS = 16/255 + + +def calc_mean_std(feat, eps=1e-5): + # eps is a small value added to the variance to avoid divide-by-zero. + size = feat.size() + assert (len(size) == 4) + N, C = size[:2] + feat_var = feat.view(N, C, -1).var(dim=2) + eps + feat_std = feat_var.sqrt().view(N, C, 1, 1) + feat_mean = feat.view(N, C, -1).mean(dim=2).view(N, C, 1, 1) + return feat_mean, feat_std + + +def fgsm_attack(init_input, epsilon, data_grad): + # random start init_input + init_input = init_input + torch.empty_like(init_input).uniform_(START_EPS, START_EPS) + + sign_data_grad = data_grad.sign() + adv_input = init_input + epsilon*sign_data_grad + return adv_input + + +def changeNewAdvStyle(input_fea, new_styleAug_mean, new_styleAug_std, p_thred): + if(new_styleAug_mean=='None'): + return input_fea + + p = np.random.uniform() + if( p < p_thred): + return input_fea + + feat_size = input_fea.size() + ori_style_mean, ori_style_std = calc_mean_std(input_fea) + normalized_fea = (input_fea - ori_style_mean.expand(feat_size)) / ori_style_std.expand(feat_size) + styleAug_fea = normalized_fea * new_styleAug_std.expand(feat_size) + new_styleAug_mean.expand(feat_size) + return styleAug_fea + + +def consistency_loss(scoresM1, scoresM2, type='euclidean'): + if(type=='euclidean'): + avg_pro = (scoresM1 + scoresM2)/2.0 + matrix1 = torch.sqrt(torch.sum((scoresM1 - avg_pro)**2,dim=1)) + matrix2 = torch.sqrt(torch.sum((scoresM2 - avg_pro)**2,dim=1)) + dis1 = torch.mean(matrix1) + dis2 = torch.mean(matrix2) + dis = (dis1+dis2)/2.0 + elif(type=='KL1'): + avg_pro = (scoresM1 + scoresM2)/2.0 + matrix1 = torch.sum( F.softmax(scoresM1,dim=-1) * (F.log_softmax(scoresM1, dim=-1) - F.log_softmax(avg_pro,dim=-1)), 1) + matrix2 = torch.sum( F.softmax(scoresM2,dim=-1) * (F.log_softmax(scoresM2, dim=-1) - F.log_softmax(avg_pro,dim=-1)), 1) + dis1 = torch.mean(matrix1) + dis2 = torch.mean(matrix2) + dis = (dis1+dis2)/2.0 + elif(type=='KL2'): + matrix = torch.sum( F.softmax(scoresM2,dim=-1) * (F.log_softmax(scoresM2, dim=-1) - F.log_softmax(scoresM1,dim=-1)), 1) + dis = torch.mean(matrix) + elif(type=='KL3'): + matrix = torch.sum( F.softmax(scoresM1,dim=-1) * (F.log_softmax(scoresM1, dim=-1) - F.log_softmax(scoresM2,dim=-1)), 1) + dis = torch.mean(matrix) + else: + return + return dis + + +def adversarial_attack_Incre(cls_net, cls_criterion, x_ori, y_ori, epsilon_list): + x_ori = x_ori.cuda() + y_ori = y_ori.cuda() + x_size = x_ori.size() + + # if not adv, set defalut = 'None' + adv_style_mean_block1, adv_style_std_block1 = 'None', 'None' + adv_style_mean_block2, adv_style_std_block2 = 'None', 'None' + adv_style_mean_block3, adv_style_std_block3 = 'None', 'None' + + # forward and set the grad = True + blocklist = 'block123' #for PACS + #blocklist = 'block12' #for digits, exp1 + #blocklist = 'block1' #for digits, exp1_onlyblock1 + #blocklist = 'block2' #for digits, exp1_onlyblock1 + + if('1' in blocklist and epsilon_list[0] != 0 ): + # forward block1 + x_ori_block1 = cls_net.forward_block1(x_ori) + feat_size_block1 = x_ori_block1.size() + ori_style_mean_block1, ori_style_std_block1 = calc_mean_std(x_ori_block1) + # set them as learnable parameters + ori_style_mean_block1 = torch.nn.Parameter(ori_style_mean_block1) + ori_style_std_block1 = torch.nn.Parameter(ori_style_std_block1) + ori_style_mean_block1.requires_grad_() + ori_style_std_block1.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block1 = (x_ori_block1 - ori_style_mean_block1.detach().expand(feat_size_block1)) / ori_style_std_block1.detach().expand(feat_size_block1) + x_ori_block1 = x_normalized_block1 * ori_style_std_block1.expand(feat_size_block1) + ori_style_mean_block1.expand(feat_size_block1) + + # pass the rest model + x_ori_block2 = cls_net.forward_block2(x_ori_block1) + x_ori_block3 = cls_net.forward_block3(x_ori_block2) + x_ori_block4 = cls_net.forward_block4(x_ori_block3) + _, x_ori_fea = cls_net.forward_rest(x_ori_block4) + x_ori_output = cls_net.classifier.forward(x_ori_fea) + + # calculate initial pred, loss and acc + ori_loss = cls_criterion(x_ori_output, y_ori) + + # zero all the existing gradients + cls_net.zero_grad() + + # backward loss + ori_loss.backward() + + # collect datagrad + grad_ori_style_mean_block1 = ori_style_mean_block1.grad.detach() + grad_ori_style_std_block1 = ori_style_std_block1.grad.detach() + + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + + adv_style_mean_block1 = fgsm_attack(ori_style_mean_block1, epsilon, grad_ori_style_mean_block1) + adv_style_std_block1 = fgsm_attack(ori_style_std_block1, epsilon, grad_ori_style_std_block1) + + # add zero_grad + cls_net.zero_grad() + + if('2' in blocklist and epsilon_list[1] != 0): + # forward block1 + x_ori_block1 = cls_net.forward_block1(x_ori) + # update adv_block1 + x_adv_block1 = changeNewAdvStyle(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + # forward block2 + x_ori_block2 = cls_net.forward_block2(x_adv_block1) + # calculate mean and std + feat_size_block2 = x_ori_block2.size() + ori_style_mean_block2, ori_style_std_block2 = calc_mean_std(x_ori_block2) + # set them as learnable parameters + ori_style_mean_block2 = torch.nn.Parameter(ori_style_mean_block2) + ori_style_std_block2 = torch.nn.Parameter(ori_style_std_block2) + ori_style_mean_block2.requires_grad_() + ori_style_std_block2.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block2 = (x_ori_block2 - ori_style_mean_block2.detach().expand(feat_size_block2)) / ori_style_std_block2.detach().expand(feat_size_block2) + x_ori_block2 = x_normalized_block2 * ori_style_std_block2.expand(feat_size_block2) + ori_style_mean_block2.expand(feat_size_block2) + # pass the rest model + x_ori_block3 = cls_net.forward_block3(x_ori_block2) + x_ori_block4 = cls_net.forward_block4(x_ori_block3) + _, x_ori_fea = cls_net.forward_rest(x_ori_block4) + x_ori_output = cls_net.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_loss = cls_criterion(x_ori_output, y_ori) + # zero all the existing gradients + cls_net.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block2 = ori_style_mean_block2.grad.detach() + grad_ori_style_std_block2 = ori_style_std_block2.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block2 = fgsm_attack(ori_style_mean_block2, epsilon, grad_ori_style_mean_block2) + adv_style_std_block2 = fgsm_attack(ori_style_std_block2, epsilon, grad_ori_style_std_block2) + + # add zero_grad + cls_net.zero_grad() + + if('3' in blocklist and epsilon_list[2] != 0): + # forward block1, block2, block3 + x_ori_block1 = cls_net.forward_block1(x_ori) + x_adv_block1 = changeNewAdvStyle(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + x_ori_block2 = cls_net.forward_block2(x_adv_block1) + x_adv_block2 = changeNewAdvStyle(x_ori_block2, adv_style_mean_block2, adv_style_std_block2, p_thred=0) + x_ori_block3 = cls_net.forward_block3(x_adv_block2) + # calculate mean and std + feat_size_block3 = x_ori_block3.size() + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3) + # set them as learnable parameters + ori_style_mean_block3 = torch.nn.Parameter(ori_style_mean_block3) + ori_style_std_block3 = torch.nn.Parameter(ori_style_std_block3) + ori_style_mean_block3.requires_grad_() + ori_style_std_block3.requires_grad_() + # contain ori_style_mean_block3 in the graph + x_normalized_block3 = (x_ori_block3 - ori_style_mean_block3.detach().expand(feat_size_block3)) / ori_style_std_block3.detach().expand(feat_size_block3) + x_ori_block3 = x_normalized_block3 * ori_style_std_block3.expand(feat_size_block3) + ori_style_mean_block3.expand(feat_size_block3) + # pass the rest model + x_ori_block4 = cls_net.forward_block4(x_ori_block3) + _, x_ori_fea = cls_net.forward_rest(x_ori_block4) + x_ori_output = cls_net.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_loss = cls_criterion(x_ori_output, y_ori) + # zero all the existing gradients + cls_net.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block3 = ori_style_mean_block3.grad.detach() + grad_ori_style_std_block3 = ori_style_std_block3.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block3 = fgsm_attack(ori_style_mean_block3, epsilon, grad_ori_style_mean_block3) + adv_style_std_block3 = fgsm_attack(ori_style_std_block3, epsilon, grad_ori_style_std_block3) + + return adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 + + \ No newline at end of file diff --git a/Meta-causal/code-withStyleAttack/tools/autoaugment.py b/Meta-causal/code-withStyleAttack/tools/autoaugment.py new file mode 100644 index 0000000000000000000000000000000000000000..76c6bc4ebd5c59b76a58a8dca196f22d41fbf114 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/tools/autoaugment.py @@ -0,0 +1,234 @@ +from PIL import Image, ImageEnhance, ImageOps +import numpy as np +import random + + +class ImageNetPolicy(object): + """ Randomly choose one of the best 24 Sub-policies on ImageNet. + + Example: + >>> policy = ImageNetPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> ImageNetPolicy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.4, "posterize", 8, 0.6, "rotate", 9, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "posterize", 7, 0.6, "posterize", 6, fillcolor), + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + + SubPolicy(0.4, "equalize", 4, 0.8, "rotate", 8, fillcolor), + SubPolicy(0.6, "solarize", 3, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.8, "posterize", 5, 1.0, "equalize", 2, fillcolor), + SubPolicy(0.2, "rotate", 3, 0.6, "solarize", 8, fillcolor), + SubPolicy(0.6, "equalize", 8, 0.4, "posterize", 6, fillcolor), + + SubPolicy(0.8, "rotate", 8, 0.4, "color", 0, fillcolor), + SubPolicy(0.4, "rotate", 9, 0.6, "equalize", 2, fillcolor), + SubPolicy(0.0, "equalize", 7, 0.8, "equalize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + + SubPolicy(0.8, "rotate", 8, 1.0, "color", 2, fillcolor), + SubPolicy(0.8, "color", 8, 0.8, "solarize", 7, fillcolor), + SubPolicy(0.4, "sharpness", 7, 0.6, "invert", 8, fillcolor), + SubPolicy(0.6, "shearX", 5, 1.0, "equalize", 9, fillcolor), + SubPolicy(0.4, "color", 0, 0.6, "equalize", 3, fillcolor), + + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment ImageNet Policy" + + +class CIFAR10Policy(object): + """ Randomly choose one of the best 25 Sub-policies on CIFAR10. + + Example: + >>> policy = CIFAR10Policy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> CIFAR10Policy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.1, "invert", 7, 0.2, "contrast", 6, fillcolor), + SubPolicy(0.7, "rotate", 2, 0.3, "translateX", 9, fillcolor), + SubPolicy(0.8, "sharpness", 1, 0.9, "sharpness", 3, fillcolor), + SubPolicy(0.5, "shearY", 8, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.5, "autocontrast", 8, 0.9, "equalize", 2, fillcolor), + + SubPolicy(0.2, "shearY", 7, 0.3, "posterize", 7, fillcolor), + SubPolicy(0.4, "color", 3, 0.6, "brightness", 7, fillcolor), + SubPolicy(0.3, "sharpness", 9, 0.7, "brightness", 9, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.5, "equalize", 1, fillcolor), + SubPolicy(0.6, "contrast", 7, 0.6, "sharpness", 5, fillcolor), + + SubPolicy(0.7, "color", 7, 0.5, "translateX", 8, fillcolor), + SubPolicy(0.3, "equalize", 7, 0.4, "autocontrast", 8, fillcolor), + SubPolicy(0.4, "translateY", 3, 0.2, "sharpness", 6, fillcolor), + SubPolicy(0.9, "brightness", 6, 0.2, "color", 8, fillcolor), + SubPolicy(0.5, "solarize", 2, 0.0, "invert", 3, fillcolor), + + SubPolicy(0.2, "equalize", 0, 0.6, "autocontrast", 0, fillcolor), + SubPolicy(0.2, "equalize", 8, 0.6, "equalize", 4, fillcolor), + SubPolicy(0.9, "color", 9, 0.6, "equalize", 6, fillcolor), + SubPolicy(0.8, "autocontrast", 4, 0.2, "solarize", 8, fillcolor), + SubPolicy(0.1, "brightness", 3, 0.7, "color", 0, fillcolor), + + SubPolicy(0.4, "solarize", 5, 0.9, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "translateY", 9, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.9, "autocontrast", 2, 0.8, "solarize", 3, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.1, "invert", 3, fillcolor), + SubPolicy(0.7, "translateY", 9, 0.9, "autocontrast", 1, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment CIFAR10 Policy" + + +class SVHNPolicy(object): + """ Randomly choose one of the best 25 Sub-policies on SVHN. + + Example: + >>> policy = SVHNPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> SVHNPolicy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.9, "shearX", 4, 0.2, "invert", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.7, "invert", 5, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.6, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 3, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "equalize", 1, 0.9, "rotate", 3, fillcolor), + + SubPolicy(0.9, "shearX", 4, 0.8, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.4, "invert", 5, fillcolor), + SubPolicy(0.9, "shearY", 5, 0.2, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 6, 0.8, "autocontrast", 1, fillcolor), + SubPolicy(0.6, "equalize", 3, 0.9, "rotate", 3, fillcolor), + + SubPolicy(0.9, "shearX", 4, 0.3, "solarize", 3, fillcolor), + SubPolicy(0.8, "shearY", 8, 0.7, "invert", 4, fillcolor), + SubPolicy(0.9, "equalize", 5, 0.6, "translateY", 6, fillcolor), + SubPolicy(0.9, "invert", 4, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.3, "contrast", 3, 0.8, "rotate", 4, fillcolor), + + SubPolicy(0.8, "invert", 5, 0.0, "translateY", 2, fillcolor), + SubPolicy(0.7, "shearY", 6, 0.4, "solarize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 0.8, "rotate", 4, fillcolor), + SubPolicy(0.3, "shearY", 7, 0.9, "translateX", 3, fillcolor), + SubPolicy(0.1, "shearX", 6, 0.6, "invert", 5, fillcolor), + + SubPolicy(0.7, "solarize", 2, 0.6, "translateY", 7, fillcolor), + SubPolicy(0.8, "shearY", 4, 0.8, "invert", 8, fillcolor), + SubPolicy(0.7, "shearX", 9, 0.8, "translateY", 3, fillcolor), + SubPolicy(0.8, "shearY", 5, 0.7, "autocontrast", 3, fillcolor), + SubPolicy(0.7, "shearX", 2, 0.1, "invert", 5, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment SVHN Policy" + + +class SubPolicy(object): + def __init__(self, p1, operation1, magnitude_idx1, p2, operation2, magnitude_idx2, fillcolor=(128, 128, 128)): + ranges = { + "shearX": np.linspace(0, 0.3, 10), + "shearY": np.linspace(0, 0.3, 10), + "translateX": np.linspace(0, 150 / 331, 10), + "translateY": np.linspace(0, 150 / 331, 10), + "rotate": np.linspace(0, 30, 10), + "color": np.linspace(0.0, 0.9, 10), + "posterize": np.round(np.linspace(8, 4, 10), 0).astype(np.int), + "solarize": np.linspace(256, 0, 10), + "contrast": np.linspace(0.0, 0.9, 10), + "sharpness": np.linspace(0.0, 0.9, 10), + "brightness": np.linspace(0.0, 0.9, 10), + "autocontrast": [0] * 10, + "equalize": [0] * 10, + "invert": [0] * 10 + } + + # from https://stackoverflow.com/questions/5252170/specify-image-filling-color-when-rotating-in-python-with-pil-and-setting-expand + def rotate_with_fill(img, magnitude): + rot = img.convert("RGBA").rotate(magnitude) + return Image.composite(rot, Image.new("RGBA", rot.size, (128,) * 4), rot).convert(img.mode) + + func = { + "shearX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, magnitude * random.choice([-1, 1]), 0, 0, 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "shearY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, magnitude * random.choice([-1, 1]), 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "translateX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, magnitude * img.size[0] * random.choice([-1, 1]), 0, 1, 0), + fillcolor=fillcolor), + "translateY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, 0, 1, magnitude * img.size[1] * random.choice([-1, 1])), + fillcolor=fillcolor), + "rotate": lambda img, magnitude: rotate_with_fill(img, magnitude), + "color": lambda img, magnitude: ImageEnhance.Color(img).enhance(1 + magnitude * random.choice([-1, 1])), + "posterize": lambda img, magnitude: ImageOps.posterize(img, magnitude), + "solarize": lambda img, magnitude: ImageOps.solarize(img, magnitude), + "contrast": lambda img, magnitude: ImageEnhance.Contrast(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "sharpness": lambda img, magnitude: ImageEnhance.Sharpness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "brightness": lambda img, magnitude: ImageEnhance.Brightness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "autocontrast": lambda img, magnitude: ImageOps.autocontrast(img), + "equalize": lambda img, magnitude: ImageOps.equalize(img), + "invert": lambda img, magnitude: ImageOps.invert(img) + } + + self.p1 = p1 + self.operation1 = func[operation1] + self.magnitude1 = ranges[operation1][magnitude_idx1] + self.p2 = p2 + self.operation2 = func[operation2] + self.magnitude2 = ranges[operation2][magnitude_idx2] + + + def __call__(self, img): + if random.random() < self.p1: img = self.operation1(img, self.magnitude1) + if random.random() < self.p2: img = self.operation2(img, self.magnitude2) + return img \ No newline at end of file diff --git a/Meta-causal/code-withStyleAttack/tools/causalaugment_v3.py b/Meta-causal/code-withStyleAttack/tools/causalaugment_v3.py new file mode 100644 index 0000000000000000000000000000000000000000..a375b7ebe5a83c3dba5b88f48f23a4326dec77e1 --- /dev/null +++ b/Meta-causal/code-withStyleAttack/tools/causalaugment_v3.py @@ -0,0 +1,694 @@ +# code in this file is adpated from rpmcruz/autoaugment +# https://github.com/rpmcruz/autoaugment/blob/master/transformations.py +import random + +import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw +import numpy as np +import torch +from PIL import Image,ImageStat +#import cv2 +from torchvision import transforms + +# def tensor2img(tensor): +# transform = transforms.Compose() + +def ShearX(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + +def DoShearX(img, v): # [-0.3, 0.3] + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + +def ShearY(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + +def DoShearY(img, v): # [-0.3, 0.3] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + +def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[0] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + +def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) +def DoTranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + +def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[1] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) +def DoTranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + +def Rotate(img, v): # [-30, 30] + assert -30 <= v <= 30 + if random.random() > 0.5: + v = -v + return img.rotate(v) +def DoRotate(img, v): # [-30, 30] + return img.rotate(v) + + +def AutoContrast(img, v): + return PIL.ImageOps.autocontrast(img, v) +def DoAutoContrast(img, v): + return PIL.ImageOps.autocontrast(img, v) + +def Invert(img, _): + return PIL.ImageOps.invert(img) +def DoInvert(img, _): + return PIL.ImageOps.invert(img) + + +def Equalize(img, _): + return PIL.ImageOps.equalize(img) +def DoEqualize(img, _): + return PIL.ImageOps.equalize(img) + +def Flip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + +def DoFlip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + + +def Solarize(img, v): # [0, 256] + assert 0 <= v <= 256 + return PIL.ImageOps.solarize(img, v) +def DoSolarize(img, v): # [0, 256] + return PIL.ImageOps.solarize(img, v) + +def SolarizeAdd(img, addition=0, threshold=128): + #img_np = np.array(img).astype(np.int) + img_np = np.array(img).astype(np.int32) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) +def DoSolarizeAdd(img, addition=0, threshold=128): + #img_np = np.array(img).astype(np.int) + img_np = np.array(img).astype(np.int32) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) + +def Posterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) +def DoPosterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) + + +def Contrast(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Contrast(img).enhance(v) + +def DoContrast(img, v): + return PIL.ImageEnhance.Contrast(img).enhance(v) + +def Color(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Color(img).enhance(v) + +def DoColor(img, v): + stat =ImageStat.Stat(img) + return PIL.ImageEnhance.Color(img).enhance(v) + + +def Brightness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Brightness(img).enhance(v) + +def DoBrightness(img, v): # obtain the brightness of image + return PIL.ImageEnhance.Brightness(img).enhance(v) + + +def Sharpness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Sharpness(img).enhance(v) + +def DoSharpness(img, v): + return PIL.ImageEnhance.Sharpness(img).enhance(v) + +def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] + assert 0.0 <= v <= 0.2 + if v <= 0.: + return img + + v = v * img.size[0] + return CutoutAbs(img, v) + + +def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img +def DoCutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img + + +def SamplePairing(imgs): # [0, 0.4] + def f(img1, v): + i = np.random.choice(len(imgs)) + img2 = PIL.Image.fromarray(imgs[i]) + return PIL.Image.blend(img1, img2, v) + + return f + + +def Identity(img, v): + return img + +def NoiseSalt(img, noise_rate): + """增加椒盐噪声 + args: + noise_rate (float): noise rate + """ + img_ = np.array(img).copy() + h, w, c = img_.shape + signal_pct = 1 - noise_rate + mask = np.random.choice((0, 1, 2), size=(h, w, 1), p=[signal_pct, noise_rate/2., noise_rate/2.]) + mask = np.repeat(mask, c, axis=2) + img_[mask == 1] = 255 # 盐噪声 + img_[mask == 2] = 0 # 椒噪声 + return Image.fromarray(img_.astype('uint8')) + +def DoNoiseSalt(img, noise_rate): + """增加椒盐噪声 + args: + noise_rate (float): noise rate + """ + img_ = np.array(img).copy() + h, w, c = img_.shape + signal_pct = 1 - noise_rate + mask = np.random.choice((0, 1, 2), size=(h, w, 1), p=[signal_pct, noise_rate/2., noise_rate/2.]) + mask = np.repeat(mask, c, axis=2) + img_[mask == 1] = 255 # 盐噪声 + img_[mask == 2] = 0 # 椒噪声 + return Image.fromarray(img_.astype('uint8')) +def NoiseGaussian(img, sigma): + """增加高斯噪声 + 传入: + img : 原图 + mean : 均值默认0 + sigma : 标准差 + 返回: + gaussian_out : 噪声处理后的图片 + """ + # 将图片灰度标准化 + img_ = np.array(img).copy() + img_ = img_ / 255.0 + # 产生高斯 noise + noise = np.random.normal(0, sigma, img_.shape) + # 将噪声和图片叠加 + gaussian_out = img_ + noise + # 将超过 1 的置 1，低于 0 的置 0 + gaussian_out = np.clip(gaussian_out, 0, 1) + # 将图片灰度范围的恢复为 0-255 + gaussian_out = np.uint8(gaussian_out*255) + # 将噪声范围搞为 0-255 + # noise = np.uint8(noise*255) + return Image.fromarray(gaussian_out) + +def DoNoiseGaussian(img, sigma): + """增加高斯噪声 + 传入: + img : 原图 + mean : 均值默认0 + sigma : 标准差 + 返回: + gaussian_out : 噪声处理后的图片 + """ + # 将图片灰度标准化 + img_ = np.array(img).copy() + img_ = img_ / 255.0 + # 产生高斯 noise + noise = np.random.normal(0, sigma, img_.shape) + # 将噪声和图片叠加 + gaussian_out = img_ + noise + # 将超过 1 的置 1，低于 0 的置 0 + gaussian_out = np.clip(gaussian_out, 0, 1) + # 将图片灰度范围的恢复为 0-255 + gaussian_out = np.uint8(gaussian_out*255) + # 将噪声范围搞为 0-255 + # noise = np.uint8(noise*255) + return Image.fromarray(gaussian_out) + +# def factor_list(factor_num): +# l = [ +# 'AutoContrast', +# 'Invert', +# 'Equalize', +# 'Solarize', +# 'SolarizeAdd', +# 'Posterize', +# 'Contrast', +# 'Color', +# 'Brightness', +# 'Sharpness', +# 'NoiseSalt', +# 'NoiseGaussian', +# ] +# return l[:factor_num] + +# def causal_list(factor_num): # 16 oeprations and their ranges +# l = [ +# (AutoContrast, 0, 100), +# (Invert, 0, 1), +# (Equalize, 0, 1), +# (Solarize, 0, 256), +# (SolarizeAdd, 0, 110), +# (Posterize, 0, 4), +# (Contrast, 0.1, 1.9), +# (Color, 0.1, 1.9), +# (Brightness, 0.1, 1.9), +# (Sharpness, 0.1, 1.9), +# (NoiseSalt,0.0,0.1), +# (NoiseGaussian,0.0,0.1), +# ] + +# return l[:factor_num] + + +# def factor_list(factor_num): +# l = [ +# 'ShearX', +# 'ShearY', +# 'Rotate', +# 'Flip' +# ] +# return l[:factor_num] + +# def causal_list(factor_num): # 16 oeprations and their ranges +# l = [ +# (ShearX, 0., 0.3), +# (ShearY, 0., 0.3), +# (Rotate, 0, 30), +# (Flip, 0, 1), +# ] + +# return l[:factor_num] + +def factor_list(factor_num): + l = [ + 'ShearX', + 'ShearY', + 'AutoContrast', + 'Invert', + 'Equalize', + 'Solarize', + 'SolarizeAdd', + 'Posterize', + 'Contrast', + 'Color', + 'Brightness', + 'Sharpness', + 'NoiseSalt', + 'NoiseGaussian', + 'Rotate', + 'Flip' + ] + return l[:factor_num] + +def causal_list(factor_num): # 16 oeprations and their ranges + l = [ + (ShearX, 0., 0.3), + (ShearY, 0., 0.3), + (AutoContrast, 0, 100), + (Invert, 0, 1), + (Equalize, 0, 1), + (Solarize, 0, 256), + (SolarizeAdd, 0, 110), + (Posterize, 0, 4), + (Contrast, 0.1, 1.9), + (Color, 0.1, 1.9), + (Brightness, 0.1, 1.9), + (Sharpness, 0.1, 1.9), + (NoiseSalt,0.0,0.1), + (NoiseGaussian,0.0,0.1), + (Rotate, 0, 30), + (Flip, 0, 1), + ] + + return l[:factor_num] + +class Lighting(object): + """Lighting noise(AlexNet - style PCA - based noise)""" + + def __init__(self, alphastd, eigval, eigvec): + self.alphastd = alphastd + self.eigval = torch.Tensor(eigval) + self.eigvec = torch.Tensor(eigvec) + + def __call__(self, img): + if self.alphastd == 0: + return img + + alpha = img.new().resize_(3).normal_(0, self.alphastd) + rgb = self.eigvec.type_as(img).clone() \ + .mul(alpha.view(1, 3).expand(3, 3)) \ + .mul(self.eigval.view(1, 3).expand(3, 3)) \ + .sum(1).squeeze() + + return img.add(rgb.view(3, 1, 1).expand_as(img)) + + +class CutoutDefault(object): + """ + Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py + """ + def __init__(self, length): + self.length = length + + def __call__(self, img): + h, w = img.size(1), img.size(2) + mask = np.ones((h, w), np.float32) + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + mask[y1: y2, x1: x2] = 0. + mask = torch.from_numpy(mask) + mask = mask.expand_as(img) + img *= mask + return img + + +class RandAugment_incausal: + def __init__(self, n, m, factor_num, randm=False, randn=False): + self.n = n + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + self.randn = randn + self.factor_num = factor_num + print("randm:",self.randm) + print("randn:",self.randn) + print("n:",self.n) + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + if self.randn: + self.n = random.randint(1,self.factor_num) + + ops = random.choices(self.causal_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img +class RandAugment_all: + def __init__(self, m, factor_num, randm=False): + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + self.factor_num = factor_num + + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + factor_choice = np.random.randint(0,2,self.factor_num) + # ops = random.choices(self.causal_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + if factor_choice[index] == 0: + continue + else: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img +class RandAugment_incausal_label: + def __init__(self, n, m, factor_num, randm=False): + self.n = n + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + self.factor_num = factor_num + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + print("randm:",self.randm) + + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + #op_labels = np.random.randint(0,self.factor_num-1,self.n) + op_labels = random.sample(range(0, self.factor_num), self.n) + ops = [li for index, li in enumerate(self.causal_list) if index in op_labels] + #ops = random.choices(self.causal_list, k=self.n) + # print(self.causal_list) + # print("op_labels:",op_labels) + # print("select_op:",ops) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img, np.array(op_labels) +class FactualAugment_incausal: + def __init__(self, m, factor_num, randm=False): + self.m = m + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.randm = randm + print("randm:",self.randm) + def __call__(self, img): + # ops = random.choices(self.causal_list, k=1) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + val = (float(self.m) / 30) * float(maxval - minval) + minval + if index == 0: + imgs = np.array(op(img, val)) + else: + imgs = np.concatenate((imgs, op(img, val)),-1) + # print("imgs",imgs.shape) + return imgs +class CounterfactualAugment_incausal: + def __init__(self,factor_num): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + def __call__(self, img): + # index = 0 + # b, c, h, w = img.shape + # imgs = torch.zeros(b*self.factor_num, c, h, w) + # for b_ in range(32): + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + if index == 0: + imgs = np.array(op(img, maxval)) + else: + imgs = np.concatenate((imgs, op(img, maxval)),-1) + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs +class MultiCounterfactualAugment_incausal: + def __init__(self, factor_num, stride): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.stride = stride + + def __call__(self, img): + # index = 0 + # b, c, h, w = img.shape + # imgs = torch.zeros(b*self.factor_num, c, h, w) + # for b_ in range(32): + # 0,5,10,15,20,25,30 + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + for i in range(0, 31, self.stride): + val = (float(i) / 30) * float(maxval - minval) + minval + if index == 0 and i == 0: + imgs = np.array(op(img, val)) + else: + imgs = np.concatenate((imgs, op(img, val)),-1) + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs +class MultiCounterfactualAugment: + def __init__(self, factor_num, stride=5): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.stride = stride + self.var_num = len(list(range(0, 31, self.stride))) + print("stride:",stride) + def __call__(self, img): + # index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num*self.var_num, c, h, w) + # for b_ in range(32): + # 0,5,10,15,20,25,30 + # print(img.shape) + for b_ in range(b): + img0 = transforms.ToPILImage()(imgs[b_]) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + i_index = 0 + for i in range(0, 31, self.stride): + val = (float(i) / 30) * float(maxval - minval) + minval + img1 = op(img0, val) + img1 = transforms.ToTensor()(img1) + #print(f'batch {b_} factor {index} stride {i} i_index {i_index} total {b_*self.factor_num*self.var_num+index*self.var_num+i_index}') + imgs[b_*self.factor_num*self.var_num+index*self.var_num+i_index] = img1 + i_index = i_index + 1 + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs + + +class FactualAugment: + def __init__(self, m, factor_num, randm=False): + self.m = m + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.randm = randm + print("randm:",randm) + def __call__(self, img): + index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num, c, h, w) + + img = img.cpu() + for b_ in range(b): + imgs[b_*self.factor_num:(b_+1)*self.factor_num] = self.get_item(img[b_]) + return imgs + def get_item(self, img): + index = 0 + # print("input_dim:",img.shape) + c, h, w = img.shape + imgs = torch.zeros(self.factor_num, c, h, w) + # img = img.squeeze(0) + # print(img.shape) + img = transforms.ToPILImage()(img) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval(self.factor_list[index]) + val = (float(self.m) / 30) * float(maxval - minval) + minval + img1 = op(img, val) + img1 = transforms.ToTensor()(img1) + imgs[index] = img1 + return imgs +class CounterfactualAugment: + def __init__(self,factor_num): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + + def __call__(self, img): + index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num, c, h, w) + + img = img.cpu() + for b_ in range(b): + imgs[b_*self.factor_num:(b_+1)*self.factor_num] = self.get_item(img[b_]) + return imgs + def get_item(self, img): + index = 0 + c, h, w = img.shape + imgs = torch.ones(self.factor_num, c, h, w) + # img = img.squeeze(0) + img = transforms.ToPILImage()(img) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + img1 = op(img, maxval) + # img1.save('test'+str(index)+'.png') + img1 = transforms.ToTensor()(img1) + imgs[index] = img1 + return imgs + +class Avg_statistic: + def __init__(self): + self.do_list = do_list() + self.statistic_num = len(self.do_list) + self.avg_val = np.zeros(self.statistic_num) + self.img_num = 0 + + def get_item(self,img): + # ops = self.statistic_list + do_index = 0 + for op in self.do_list: + val=op(img) + self.avg_val[do_index] += val + self.img_num = self.img_num + 1 + + def compute_average(self): + self.avg_val = self.avg_val/self.img_num + + def get_infor(self): + return self.avg_val, self.img_num + + + + diff --git a/Meta-causal/code-withStyleAttack/tools/randaugment.py b/Meta-causal/code-withStyleAttack/tools/randaugment.py new file mode 100644 index 0000000000000000000000000000000000000000..f3bbdf11541df078144fa0ced8d693d4c98507ad --- /dev/null +++ b/Meta-causal/code-withStyleAttack/tools/randaugment.py @@ -0,0 +1,248 @@ +# code in this file is adpated from rpmcruz/autoaugment +# https://github.com/rpmcruz/autoaugment/blob/master/transformations.py +import random + +import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw +import numpy as np +import torch +from PIL import Image + + +def ShearX(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + + +def ShearY(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + + +def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[0] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + + +def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + + +def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[1] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def Rotate(img, v): # [-30, 30] + assert -30 <= v <= 30 + if random.random() > 0.5: + v = -v + return img.rotate(v) + + +def AutoContrast(img, _): + return PIL.ImageOps.autocontrast(img) + + +def Invert(img, _): + return PIL.ImageOps.invert(img) + + +def Equalize(img, _): + return PIL.ImageOps.equalize(img) + + +def Flip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + + +def Solarize(img, v): # [0, 256] + assert 0 <= v <= 256 + return PIL.ImageOps.solarize(img, v) + + +def SolarizeAdd(img, addition=0, threshold=128): + img_np = np.array(img).astype(np.int) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) + + +def Posterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) + + +def Contrast(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Contrast(img).enhance(v) + + +def Color(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Color(img).enhance(v) + + +def Brightness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Brightness(img).enhance(v) + + +def Sharpness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Sharpness(img).enhance(v) + + +def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] + assert 0.0 <= v <= 0.2 + if v <= 0.: + return img + + v = v * img.size[0] + return CutoutAbs(img, v) + + +def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img + + +def SamplePairing(imgs): # [0, 0.4] + def f(img1, v): + i = np.random.choice(len(imgs)) + img2 = PIL.Image.fromarray(imgs[i]) + return PIL.Image.blend(img1, img2, v) + + return f + + +def Identity(img, v): + return img + + +def augment_list(): # 16 oeprations and their ranges + + # https://github.com/tensorflow/tpu/blob/8462d083dd89489a79e3200bcc8d4063bf362186/models/official/efficientnet/autoaugment.py#L505 + l = [ + (AutoContrast, 0, 1), + (Equalize, 0, 1), + (Invert, 0, 1), + (Rotate, 0, 30), + (Posterize, 0, 4), + (Solarize, 0, 256), + (SolarizeAdd, 0, 110), + (Color, 0.1, 1.9), + (Contrast, 0.1, 1.9), + (Brightness, 0.1, 1.9), + (Sharpness, 0.1, 1.9), + (ShearX, 0., 0.3), + (ShearY, 0., 0.3), + (CutoutAbs, 0, 40), + (TranslateXabs, 0., 100), + (TranslateYabs, 0., 100), + ] + + return l + + +class Lighting(object): + """Lighting noise(AlexNet - style PCA - based noise)""" + + def __init__(self, alphastd, eigval, eigvec): + self.alphastd = alphastd + self.eigval = torch.Tensor(eigval) + self.eigvec = torch.Tensor(eigvec) + + def __call__(self, img): + if self.alphastd == 0: + return img + + alpha = img.new().resize_(3).normal_(0, self.alphastd) + rgb = self.eigvec.type_as(img).clone() \ + .mul(alpha.view(1, 3).expand(3, 3)) \ + .mul(self.eigval.view(1, 3).expand(3, 3)) \ + .sum(1).squeeze() + + return img.add(rgb.view(3, 1, 1).expand_as(img)) + + +class CutoutDefault(object): + """ + Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py + """ + def __init__(self, length): + self.length = length + + def __call__(self, img): + h, w = img.size(1), img.size(2) + mask = np.ones((h, w), np.float32) + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + mask[y1: y2, x1: x2] = 0. + mask = torch.from_numpy(mask) + mask = mask.expand_as(img) + img *= mask + return img + + +class RandAugment: + def __init__(self, n, m, randm=False): + self.n = n + self.m = m # [0, 30] + self.augment_list = augment_list() + self.randm = randm + + def __call__(self, img): + ops = random.choices(self.augment_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + + return img diff --git a/Meta-causal/code/56717.error b/Meta-causal/code/56717.error new file mode 100644 index 0000000000000000000000000000000000000000..f4d95947c7a86339e1d04481c9ef0f88fee09876 --- /dev/null +++ b/Meta-causal/code/56717.error @@ -0,0 +1 @@ +run_my_joint_v13_test.sh: line 27: m}: command not found diff --git a/Meta-causal/code/56717.log b/Meta-causal/code/56717.log new file mode 100644 index 0000000000000000000000000000000000000000..342d449cbbc0cf96ab603cfcc9a39a8178c93297 --- /dev/null +++ b/Meta-causal/code/56717.log @@ -0,0 +1,334 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_train.hdf5 torch.Size([1840, 3, 227, 227]) torch.Size([1840]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_val.hdf5 torch.Size([208, 3, 227, 227]) torch.Size([208]) +-------------------------------------loading pretrain weights---------------------------------- +306 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 396.56, cls_loss 6.7564 cls_loss_mapping 1.5193 cls_loss_causal 1.7521 re_mapping 1.0575 re_causal 1.0584 /// teacc 81.25 lr 0.00999497 +306 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 415.32, cls_loss 2.1970 cls_loss_mapping 0.9096 cls_loss_causal 1.4403 re_mapping 0.7024 re_causal 0.7051 /// teacc 83.65 lr 0.00997987 +306 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 457.96, cls_loss 1.3065 cls_loss_mapping 0.6322 cls_loss_causal 1.2780 re_mapping 0.6032 re_causal 0.6057 /// teacc 88.46 lr 0.00995475 +306 +0.009954748808839675 +changing lr +epoch 3, time 451.75, cls_loss 0.5818 cls_loss_mapping 0.5055 cls_loss_causal 1.1465 re_mapping 0.5267 re_causal 0.5293 /// teacc 87.02 lr 0.00991965 +306 +0.009919647942993149 +changing lr +epoch 4, time 451.48, cls_loss 0.3909 cls_loss_mapping 0.4012 cls_loss_causal 1.0889 re_mapping 0.4649 re_causal 0.4683 /// teacc 84.62 lr 0.00987464 +306 +0.009874639560909117 +changing lr +epoch 5, time 441.59, cls_loss 0.3191 cls_loss_mapping 0.3555 cls_loss_causal 1.0670 re_mapping 0.3968 re_causal 0.4013 /// teacc 86.06 lr 0.00981981 +306 +0.009819814303479266 +changing lr +epoch 6, time 432.93, cls_loss 0.1327 cls_loss_mapping 0.2760 cls_loss_causal 1.0002 re_mapping 0.3232 re_causal 0.3278 /// teacc 83.17 lr 0.00975528 +306 +0.009755282581475767 +changing lr +epoch 7, time 444.85, cls_loss 0.0411 cls_loss_mapping 0.2236 cls_loss_causal 0.9368 re_mapping 0.2592 re_causal 0.2641 /// teacc 88.46 lr 0.00968117 +306 +0.009681174353198686 +changing lr +epoch 8, time 448.36, cls_loss 0.0723 cls_loss_mapping 0.2492 cls_loss_causal 0.9911 re_mapping 0.2174 re_causal 0.2224 /// teacc 86.54 lr 0.00959764 +306 +0.009597638862757255 +changing lr +epoch 9, time 446.26, cls_loss 0.0174 cls_loss_mapping 0.1853 cls_loss_causal 0.8733 re_mapping 0.1873 re_causal 0.1925 /// teacc 86.54 lr 0.00950484 +306 +0.009504844339512096 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 457.12, cls_loss 0.0358 cls_loss_mapping 0.1781 cls_loss_causal 0.8735 re_mapping 0.1610 re_causal 0.1661 /// teacc 89.90 lr 0.00940298 +306 +0.009402977659283692 +changing lr +epoch 11, time 443.50, cls_loss 0.0162 cls_loss_mapping 0.1514 cls_loss_causal 0.8453 re_mapping 0.1432 re_causal 0.1486 /// teacc 89.90 lr 0.00929224 +306 +0.009292243968009333 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 453.53, cls_loss 0.0101 cls_loss_mapping 0.1383 cls_loss_causal 0.8002 re_mapping 0.1270 re_causal 0.1328 /// teacc 90.87 lr 0.00917287 +306 +0.009172866268606516 +changing lr +epoch 13, time 466.47, cls_loss 0.0092 cls_loss_mapping 0.1432 cls_loss_causal 0.8412 re_mapping 0.1167 re_causal 0.1224 /// teacc 90.38 lr 0.00904508 +306 +0.00904508497187474 +changing lr +epoch 14, time 448.78, cls_loss 0.0063 cls_loss_mapping 0.1207 cls_loss_causal 0.7912 re_mapping 0.1077 re_causal 0.1140 /// teacc 90.38 lr 0.00890916 +306 +0.008909157412340152 +changing lr +epoch 15, time 442.01, cls_loss 0.0075 cls_loss_mapping 0.1148 cls_loss_causal 0.7640 re_mapping 0.0982 re_causal 0.1047 /// teacc 89.90 lr 0.00876536 +306 +0.00876535733001806 +changing lr +epoch 16, time 451.70, cls_loss 0.0050 cls_loss_mapping 0.1000 cls_loss_causal 0.7562 re_mapping 0.0898 re_causal 0.0964 /// teacc 90.87 lr 0.00861397 +306 +0.008613974319136962 +changing lr +epoch 17, time 454.36, cls_loss 0.0084 cls_loss_mapping 0.0986 cls_loss_causal 0.7422 re_mapping 0.0817 re_causal 0.0883 /// teacc 89.90 lr 0.00845531 +306 +0.008455313244934327 +changing lr +epoch 18, time 450.70, cls_loss 0.0033 cls_loss_mapping 0.0951 cls_loss_causal 0.7426 re_mapping 0.0760 re_causal 0.0827 /// teacc 89.42 lr 0.00828969 +306 +0.008289693629698565 +changing lr +epoch 19, time 456.56, cls_loss 0.0051 cls_loss_mapping 0.0938 cls_loss_causal 0.7288 re_mapping 0.0711 re_causal 0.0787 /// teacc 88.94 lr 0.00811745 +306 +0.00811744900929367 +changing lr +epoch 20, time 444.31, cls_loss 0.0025 cls_loss_mapping 0.0920 cls_loss_causal 0.7432 re_mapping 0.0652 re_causal 0.0723 /// teacc 89.90 lr 0.00793893 +306 +0.007938926261462368 +changing lr +epoch 21, time 436.20, cls_loss 0.0028 cls_loss_mapping 0.0782 cls_loss_causal 0.7226 re_mapping 0.0605 re_causal 0.0677 /// teacc 90.87 lr 0.00775448 +306 +0.007754484907260515 +changing lr +epoch 22, time 447.42, cls_loss 0.0020 cls_loss_mapping 0.0778 cls_loss_causal 0.6694 re_mapping 0.0571 re_causal 0.0641 /// teacc 90.38 lr 0.00756450 +306 +0.007564496387029534 +changing lr +epoch 23, time 443.40, cls_loss 0.0019 cls_loss_mapping 0.0766 cls_loss_causal 0.7606 re_mapping 0.0533 re_causal 0.0621 /// teacc 89.42 lr 0.00736934 +306 +0.007369343312364995 +changing lr +epoch 24, time 439.80, cls_loss 0.0045 cls_loss_mapping 0.0782 cls_loss_causal 0.7261 re_mapping 0.0521 re_causal 0.0608 /// teacc 90.38 lr 0.00716942 +306 +0.0071694186955877925 +changing lr +epoch 25, time 430.50, cls_loss 0.0020 cls_loss_mapping 0.0645 cls_loss_causal 0.7059 re_mapping 0.0500 re_causal 0.0593 /// teacc 90.87 lr 0.00696513 +306 +0.0069651251582696205 +changing lr +epoch 26, time 444.21, cls_loss 0.0008 cls_loss_mapping 0.0529 cls_loss_causal 0.6660 re_mapping 0.0448 re_causal 0.0527 /// teacc 90.87 lr 0.00675687 +306 +0.006756874120406716 +changing lr +epoch 27, time 451.19, cls_loss 0.0027 cls_loss_mapping 0.0633 cls_loss_causal 0.7457 re_mapping 0.0430 re_causal 0.0520 /// teacc 90.87 lr 0.00654508 +306 +0.00654508497187474 +changing lr +---------------------saving model at epoch 28---------------------------------------------------- +epoch 28, time 444.91, cls_loss 0.0045 cls_loss_mapping 0.0630 cls_loss_causal 0.6839 re_mapping 0.0409 re_causal 0.0485 /// teacc 91.35 lr 0.00633018 +306 +0.006330184227833378 +changing lr +epoch 29, time 454.38, cls_loss 0.0030 cls_loss_mapping 0.0528 cls_loss_causal 0.6373 re_mapping 0.0388 re_causal 0.0468 /// teacc 88.94 lr 0.00611260 +306 +0.006112604669781575 +changing lr +epoch 30, time 455.36, cls_loss 0.0023 cls_loss_mapping 0.0479 cls_loss_causal 0.6459 re_mapping 0.0382 re_causal 0.0462 /// teacc 91.35 lr 0.00589278 +306 +0.005892784473993186 +changing lr +epoch 31, time 447.61, cls_loss 0.0014 cls_loss_mapping 0.0532 cls_loss_causal 0.6553 re_mapping 0.0365 re_causal 0.0447 /// teacc 91.35 lr 0.00567117 +306 +0.00567116632908828 +changing lr +epoch 32, time 455.74, cls_loss 0.0019 cls_loss_mapping 0.0470 cls_loss_causal 0.6156 re_mapping 0.0346 re_causal 0.0422 /// teacc 90.38 lr 0.00544820 +306 +0.00544819654451717 +changing lr +epoch 33, time 458.62, cls_loss 0.0026 cls_loss_mapping 0.0475 cls_loss_causal 0.6128 re_mapping 0.0336 re_causal 0.0415 /// teacc 91.35 lr 0.00522432 +306 +0.005224324151752577 +changing lr +epoch 34, time 443.89, cls_loss 0.0034 cls_loss_mapping 0.0503 cls_loss_causal 0.6216 re_mapping 0.0331 re_causal 0.0412 /// teacc 90.87 lr 0.00500000 +306 +0.005000000000000003 +changing lr +---------------------saving model at epoch 35---------------------------------------------------- +epoch 35, time 474.23, cls_loss 0.0025 cls_loss_mapping 0.0398 cls_loss_causal 0.5884 re_mapping 0.0317 re_causal 0.0397 /// teacc 91.83 lr 0.00477568 +306 +0.004775675848247429 +changing lr +epoch 36, time 456.46, cls_loss 0.0023 cls_loss_mapping 0.0434 cls_loss_causal 0.6319 re_mapping 0.0308 re_causal 0.0386 /// teacc 91.35 lr 0.00455180 +306 +0.004551803455482836 +changing lr +epoch 37, time 460.36, cls_loss 0.0024 cls_loss_mapping 0.0376 cls_loss_causal 0.6052 re_mapping 0.0290 re_causal 0.0364 /// teacc 90.87 lr 0.00432883 +306 +0.004328833670911726 +changing lr +epoch 38, time 456.58, cls_loss 0.0013 cls_loss_mapping 0.0368 cls_loss_causal 0.6265 re_mapping 0.0276 re_causal 0.0354 /// teacc 90.38 lr 0.00410722 +306 +0.0041072155260068206 +changing lr +epoch 39, time 468.90, cls_loss 0.0019 cls_loss_mapping 0.0310 cls_loss_causal 0.6240 re_mapping 0.0264 re_causal 0.0344 /// teacc 90.87 lr 0.00388740 +306 +0.0038873953302184317 +changing lr +epoch 40, time 457.96, cls_loss 0.0020 cls_loss_mapping 0.0328 cls_loss_causal 0.6230 re_mapping 0.0257 re_causal 0.0335 /// teacc 90.87 lr 0.00366982 +306 +0.003669815772166629 +changing lr +---------------------saving model at epoch 41---------------------------------------------------- +epoch 41, time 469.29, cls_loss 0.0023 cls_loss_mapping 0.0376 cls_loss_causal 0.6061 re_mapping 0.0249 re_causal 0.0320 /// teacc 92.31 lr 0.00345492 +306 +0.0034549150281252667 +changing lr +epoch 42, time 475.72, cls_loss 0.0025 cls_loss_mapping 0.0311 cls_loss_causal 0.6195 re_mapping 0.0243 re_causal 0.0322 /// teacc 90.87 lr 0.00324313 +306 +0.0032431258795932905 +changing lr +epoch 43, time 450.85, cls_loss 0.0018 cls_loss_mapping 0.0341 cls_loss_causal 0.6223 re_mapping 0.0235 re_causal 0.0310 /// teacc 90.87 lr 0.00303487 +306 +0.0030348748417303863 +changing lr +epoch 44, time 441.78, cls_loss 0.0019 cls_loss_mapping 0.0317 cls_loss_causal 0.6072 re_mapping 0.0228 re_causal 0.0304 /// teacc 90.38 lr 0.00283058 +306 +0.0028305813044122124 +changing lr +---------------------saving model at epoch 45---------------------------------------------------- +epoch 45, time 462.98, cls_loss 0.0013 cls_loss_mapping 0.0307 cls_loss_causal 0.5641 re_mapping 0.0222 re_causal 0.0291 /// teacc 93.75 lr 0.00263066 +306 +0.0026306566876350096 +changing lr +epoch 46, time 474.81, cls_loss 0.0028 cls_loss_mapping 0.0323 cls_loss_causal 0.6004 re_mapping 0.0218 re_causal 0.0287 /// teacc 91.83 lr 0.00243550 +306 +0.0024355036129704724 +changing lr +epoch 47, time 465.56, cls_loss 0.0013 cls_loss_mapping 0.0291 cls_loss_causal 0.6082 re_mapping 0.0213 re_causal 0.0289 /// teacc 92.31 lr 0.00224552 +306 +0.00224551509273949 +changing lr +epoch 48, time 458.33, cls_loss 0.0011 cls_loss_mapping 0.0269 cls_loss_causal 0.6051 re_mapping 0.0208 re_causal 0.0289 /// teacc 91.35 lr 0.00206107 +306 +0.002061073738537637 +changing lr +epoch 49, time 450.51, cls_loss 0.0012 cls_loss_mapping 0.0242 cls_loss_causal 0.5558 re_mapping 0.0200 re_causal 0.0273 /// teacc 91.35 lr 0.00188255 +306 +0.0018825509907063344 +changing lr +epoch 50, time 462.46, cls_loss 0.0009 cls_loss_mapping 0.0237 cls_loss_causal 0.5775 re_mapping 0.0194 re_causal 0.0261 /// teacc 90.38 lr 0.00171031 +306 +0.0017103063703014388 +changing lr +epoch 51, time 458.67, cls_loss 0.0017 cls_loss_mapping 0.0239 cls_loss_causal 0.5359 re_mapping 0.0184 re_causal 0.0244 /// teacc 91.35 lr 0.00154469 +306 +0.0015446867550656784 +changing lr +epoch 52, time 439.55, cls_loss 0.0016 cls_loss_mapping 0.0239 cls_loss_causal 0.5782 re_mapping 0.0180 re_causal 0.0248 /// teacc 92.31 lr 0.00138603 +306 +0.001386025680863044 +changing lr +epoch 53, time 468.39, cls_loss 0.0011 cls_loss_mapping 0.0221 cls_loss_causal 0.5797 re_mapping 0.0174 re_causal 0.0241 /// teacc 90.38 lr 0.00123464 +306 +0.0012346426699819469 +changing lr +epoch 54, time 478.52, cls_loss 0.0011 cls_loss_mapping 0.0208 cls_loss_causal 0.5323 re_mapping 0.0171 re_causal 0.0233 /// teacc 91.35 lr 0.00109084 +306 +0.0010908425876598518 +changing lr +epoch 55, time 451.23, cls_loss 0.0018 cls_loss_mapping 0.0228 cls_loss_causal 0.5217 re_mapping 0.0167 re_causal 0.0227 /// teacc 91.35 lr 0.00095492 +306 +0.000954915028125264 +changing lr +epoch 56, time 455.62, cls_loss 0.0008 cls_loss_mapping 0.0185 cls_loss_causal 0.5520 re_mapping 0.0165 re_causal 0.0225 /// teacc 90.87 lr 0.00082713 +306 +0.0008271337313934874 +changing lr +epoch 57, time 455.64, cls_loss 0.0015 cls_loss_mapping 0.0242 cls_loss_causal 0.5776 re_mapping 0.0162 re_causal 0.0225 /// teacc 90.87 lr 0.00070776 +306 +0.00070775603199067 +changing lr +epoch 58, time 446.78, cls_loss 0.0009 cls_loss_mapping 0.0185 cls_loss_causal 0.5541 re_mapping 0.0158 re_causal 0.0221 /// teacc 91.35 lr 0.00059702 +306 +0.0005970223407163104 +changing lr +epoch 59, time 451.88, cls_loss 0.0025 cls_loss_mapping 0.0193 cls_loss_causal 0.5280 re_mapping 0.0156 re_causal 0.0217 /// teacc 92.31 lr 0.00049516 +306 +0.0004951556604879052 +changing lr +epoch 60, time 459.80, cls_loss 0.0019 cls_loss_mapping 0.0191 cls_loss_causal 0.5650 re_mapping 0.0154 re_causal 0.0212 /// teacc 91.83 lr 0.00040236 +306 +0.00040236113724274745 +changing lr +epoch 61, time 456.30, cls_loss 0.0013 cls_loss_mapping 0.0195 cls_loss_causal 0.5573 re_mapping 0.0151 re_causal 0.0209 /// teacc 90.87 lr 0.00031883 +306 +0.00031882564680131423 +changing lr +epoch 62, time 461.25, cls_loss 0.0016 cls_loss_mapping 0.0184 cls_loss_causal 0.5320 re_mapping 0.0149 re_causal 0.0203 /// teacc 91.83 lr 0.00024472 +306 +0.0002447174185242325 +changing lr +epoch 63, time 461.95, cls_loss 0.0025 cls_loss_mapping 0.0234 cls_loss_causal 0.5478 re_mapping 0.0148 re_causal 0.0203 /// teacc 91.35 lr 0.00018019 +306 +0.0001801856965207339 +changing lr +epoch 64, time 443.04, cls_loss 0.0012 cls_loss_mapping 0.0208 cls_loss_causal 0.5022 re_mapping 0.0147 re_causal 0.0200 /// teacc 91.35 lr 0.00012536 +306 +0.000125360439090882 +changing lr +epoch 65, time 454.35, cls_loss 0.0012 cls_loss_mapping 0.0176 cls_loss_causal 0.5745 re_mapping 0.0147 re_causal 0.0203 /// teacc 91.83 lr 0.00008035 +306 +8.03520570068517e-05 +changing lr +epoch 66, time 462.74, cls_loss 0.0018 cls_loss_mapping 0.0228 cls_loss_causal 0.5579 re_mapping 0.0147 re_causal 0.0201 /// teacc 91.35 lr 0.00004525 +306 +4.5251191160326525e-05 +changing lr +epoch 67, time 470.10, cls_loss 0.0012 cls_loss_mapping 0.0186 cls_loss_causal 0.5288 re_mapping 0.0147 re_causal 0.0205 /// teacc 92.31 lr 0.00002013 +306 +2.0128530023804673e-05 +changing lr +epoch 68, time 446.31, cls_loss 0.0011 cls_loss_mapping 0.0165 cls_loss_causal 0.5339 re_mapping 0.0146 re_causal 0.0202 /// teacc 89.42 lr 0.00000503 +306 +5.034667293427056e-06 +changing lr +epoch 69, time 458.08, cls_loss 0.0013 cls_loss_mapping 0.0148 cls_loss_causal 0.5422 re_mapping 0.0146 re_causal 0.0204 /// teacc 92.31 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'art_painting', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/art_painting_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['art_painting', 'cartoon', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + art_painting cartoon photo sketch Avg +w/o do (original x) 99.169922 65.784983 95.209581 64.596589 75.197051 + art_painting cartoon photo sketch Avg +do 99.072266 64.803754 95.269461 64.342072 74.805096 diff --git a/Meta-causal/code/56718.error b/Meta-causal/code/56718.error new file mode 100644 index 0000000000000000000000000000000000000000..f26c68e6c5fa980b508c7bd532627e6b75b149fa --- /dev/null +++ b/Meta-causal/code/56718.error @@ -0,0 +1,2 @@ +bash: run_my_joint_v13_test.sh: No such file or directory +srun: error: gcpl4-eu-1: task 0: Exited with exit code 127 diff --git a/Meta-causal/code/56718.log b/Meta-causal/code/56718.log new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code/56719.error b/Meta-causal/code/56719.error new file mode 100644 index 0000000000000000000000000000000000000000..444e676738c3b4b1c880f3c832cec125757b1b1b --- /dev/null +++ b/Meta-causal/code/56719.error @@ -0,0 +1,25 @@ +Traceback (most recent call last): + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/main_test_digit_v13.py", line 142, in + main() + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1157, in __call__ + return self.main(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1078, in main + rv = self.invoke(ctx) + ^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 1434, in invoke + return ctx.invoke(self.callback, **ctx.params) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/scratch/yuqian_fu/micromamba/envs/auto-uvapqvk3mmem/lib/python3.11/site-packages/click/core.py", line 783, in invoke + return __callback(*args, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/main_test_digit_v13.py", line 28, in main + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/main_test_digit_v13.py", line 101, in evaluate_digit + teset = str2fun[data]('test', channels=channels) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yuqian_fu/Projects/CausalStyleAdv/Meta-causal/code/data_loader_joint_v3.py", line 722, in load_mnist_m + with open(path, 'rb') as f: + ^^^^^^^^^^^^^^^^ +FileNotFoundError: [Errno 2] No such file or directory: 'data/mnist_m-test.pkl' +srun: error: gcpl4-eu-1: task 0: Exited with exit code 1 diff --git a/Meta-causal/code/56719.log b/Meta-causal/code/56719.log new file mode 100644 index 0000000000000000000000000000000000000000..877231eeeaf1d132b8ddf33e5c8762b82226e64b --- /dev/null +++ b/Meta-causal/code/56719.log @@ -0,0 +1,2066 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 3 +--------------------------CA_multiple-------------------------- +---------------------------14 factors----------------- +randm: True +randn: True +n: 3 +randm: False +100 +0.0001 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 260.22, cls_loss 1.1168 cls_loss_mapping 1.7217 cls_loss_causal 2.1730 re_mapping 0.1107 re_causal 0.1210 /// teacc 88.60 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 265.91, cls_loss 0.3310 cls_loss_mapping 0.6635 cls_loss_causal 1.7775 re_mapping 0.1227 re_causal 0.1643 /// teacc 94.13 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 274.59, cls_loss 0.2190 cls_loss_mapping 0.3836 cls_loss_causal 1.5398 re_mapping 0.0889 re_causal 0.1349 /// teacc 95.73 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 274.87, cls_loss 0.1538 cls_loss_mapping 0.2464 cls_loss_causal 1.3205 re_mapping 0.0726 re_causal 0.1133 /// teacc 96.67 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 274.40, cls_loss 0.1333 cls_loss_mapping 0.2005 cls_loss_causal 1.2889 re_mapping 0.0565 re_causal 0.0967 /// teacc 96.76 lr 0.00010000 +100 +0.0001 +changing lr +epoch 5, time 273.58, cls_loss 0.1192 cls_loss_mapping 0.1777 cls_loss_causal 1.1780 re_mapping 0.0494 re_causal 0.0858 /// teacc 96.69 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 6---------------------------------------------------- +epoch 6, time 274.48, cls_loss 0.1015 cls_loss_mapping 0.1485 cls_loss_causal 1.1906 re_mapping 0.0407 re_causal 0.0792 /// teacc 97.65 lr 0.00010000 +100 +0.0001 +changing lr +epoch 7, time 273.51, cls_loss 0.0994 cls_loss_mapping 0.1401 cls_loss_causal 1.0640 re_mapping 0.0373 re_causal 0.0706 /// teacc 97.62 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 8---------------------------------------------------- +epoch 8, time 274.20, cls_loss 0.0851 cls_loss_mapping 0.1189 cls_loss_causal 1.0603 re_mapping 0.0328 re_causal 0.0659 /// teacc 97.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 9, time 273.45, cls_loss 0.0854 cls_loss_mapping 0.1226 cls_loss_causal 1.0207 re_mapping 0.0298 re_causal 0.0623 /// teacc 97.90 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 10---------------------------------------------------- +epoch 10, time 273.93, cls_loss 0.0650 cls_loss_mapping 0.0935 cls_loss_causal 0.9621 re_mapping 0.0281 re_causal 0.0602 /// teacc 98.02 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 11---------------------------------------------------- +epoch 11, time 274.43, cls_loss 0.0669 cls_loss_mapping 0.0951 cls_loss_causal 0.9560 re_mapping 0.0255 re_causal 0.0558 /// teacc 98.22 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 12---------------------------------------------------- +epoch 12, time 274.42, cls_loss 0.0667 cls_loss_mapping 0.0970 cls_loss_causal 0.9466 re_mapping 0.0245 re_causal 0.0554 /// teacc 98.28 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 13---------------------------------------------------- +epoch 13, time 273.96, cls_loss 0.0591 cls_loss_mapping 0.0844 cls_loss_causal 0.9433 re_mapping 0.0231 re_causal 0.0545 /// teacc 98.31 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 14---------------------------------------------------- +epoch 14, time 273.58, cls_loss 0.0548 cls_loss_mapping 0.0830 cls_loss_causal 0.8947 re_mapping 0.0220 re_causal 0.0519 /// teacc 98.41 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 15---------------------------------------------------- +epoch 15, time 269.38, cls_loss 0.0418 cls_loss_mapping 0.0628 cls_loss_causal 0.9005 re_mapping 0.0207 re_causal 0.0518 /// teacc 98.44 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 265.41, cls_loss 0.0544 cls_loss_mapping 0.0769 cls_loss_causal 0.8831 re_mapping 0.0197 re_causal 0.0493 /// teacc 98.48 lr 0.00010000 +100 +0.0001 +changing lr +epoch 17, time 261.91, cls_loss 0.0525 cls_loss_mapping 0.0776 cls_loss_causal 0.8870 re_mapping 0.0197 re_causal 0.0493 /// teacc 98.32 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 18---------------------------------------------------- +epoch 18, time 262.02, cls_loss 0.0382 cls_loss_mapping 0.0581 cls_loss_causal 0.8764 re_mapping 0.0184 re_causal 0.0472 /// teacc 98.51 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 19---------------------------------------------------- +epoch 19, time 262.11, cls_loss 0.0374 cls_loss_mapping 0.0573 cls_loss_causal 0.7987 re_mapping 0.0184 re_causal 0.0452 /// teacc 98.54 lr 0.00010000 +100 +0.0001 +changing lr +epoch 20, time 261.89, cls_loss 0.0342 cls_loss_mapping 0.0538 cls_loss_causal 0.7636 re_mapping 0.0178 re_causal 0.0453 /// teacc 98.52 lr 0.00010000 +100 +0.0001 +changing lr +epoch 21, time 261.79, cls_loss 0.0292 cls_loss_mapping 0.0457 cls_loss_causal 0.7961 re_mapping 0.0171 re_causal 0.0436 /// teacc 98.43 lr 0.00010000 +100 +0.0001 +changing lr +epoch 22, time 261.58, cls_loss 0.0277 cls_loss_mapping 0.0426 cls_loss_causal 0.8074 re_mapping 0.0162 re_causal 0.0421 /// teacc 98.49 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 23---------------------------------------------------- +epoch 23, time 262.62, cls_loss 0.0333 cls_loss_mapping 0.0530 cls_loss_causal 0.7916 re_mapping 0.0156 re_causal 0.0414 /// teacc 98.64 lr 0.00010000 +100 +0.0001 +changing lr +epoch 24, time 262.08, cls_loss 0.0296 cls_loss_mapping 0.0474 cls_loss_causal 0.7989 re_mapping 0.0151 re_causal 0.0402 /// teacc 98.45 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 25---------------------------------------------------- +epoch 25, time 263.12, cls_loss 0.0246 cls_loss_mapping 0.0418 cls_loss_causal 0.7816 re_mapping 0.0149 re_causal 0.0393 /// teacc 98.75 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 26---------------------------------------------------- +epoch 26, time 262.71, cls_loss 0.0229 cls_loss_mapping 0.0378 cls_loss_causal 0.7518 re_mapping 0.0141 re_causal 0.0374 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +epoch 27, time 261.78, cls_loss 0.0247 cls_loss_mapping 0.0419 cls_loss_causal 0.7570 re_mapping 0.0147 re_causal 0.0376 /// teacc 98.74 lr 0.00010000 +100 +0.0001 +changing lr +epoch 28, time 262.28, cls_loss 0.0212 cls_loss_mapping 0.0304 cls_loss_causal 0.7520 re_mapping 0.0141 re_causal 0.0367 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 29---------------------------------------------------- +epoch 29, time 262.94, cls_loss 0.0295 cls_loss_mapping 0.0448 cls_loss_causal 0.7504 re_mapping 0.0136 re_causal 0.0360 /// teacc 98.83 lr 0.00010000 +100 +0.0001 +changing lr +epoch 30, time 262.56, cls_loss 0.0240 cls_loss_mapping 0.0389 cls_loss_causal 0.7479 re_mapping 0.0136 re_causal 0.0364 /// teacc 98.76 lr 0.00010000 +100 +0.0001 +changing lr +epoch 31, time 262.55, cls_loss 0.0208 cls_loss_mapping 0.0348 cls_loss_causal 0.7169 re_mapping 0.0130 re_causal 0.0347 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 32, time 262.24, cls_loss 0.0193 cls_loss_mapping 0.0329 cls_loss_causal 0.6995 re_mapping 0.0122 re_causal 0.0327 /// teacc 98.66 lr 0.00010000 +100 +0.0001 +changing lr +epoch 33, time 262.86, cls_loss 0.0189 cls_loss_mapping 0.0334 cls_loss_causal 0.7307 re_mapping 0.0124 re_causal 0.0343 /// teacc 98.57 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 34---------------------------------------------------- +epoch 34, time 263.86, cls_loss 0.0187 cls_loss_mapping 0.0314 cls_loss_causal 0.7412 re_mapping 0.0121 re_causal 0.0325 /// teacc 98.88 lr 0.00010000 +100 +0.0001 +changing lr +epoch 35, time 262.90, cls_loss 0.0162 cls_loss_mapping 0.0290 cls_loss_causal 0.7096 re_mapping 0.0120 re_causal 0.0328 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 36, time 263.18, cls_loss 0.0130 cls_loss_mapping 0.0216 cls_loss_causal 0.6816 re_mapping 0.0117 re_causal 0.0312 /// teacc 98.71 lr 0.00010000 +100 +0.0001 +changing lr +epoch 37, time 263.13, cls_loss 0.0150 cls_loss_mapping 0.0245 cls_loss_causal 0.6711 re_mapping 0.0119 re_causal 0.0316 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +epoch 38, time 262.71, cls_loss 0.0171 cls_loss_mapping 0.0291 cls_loss_causal 0.6826 re_mapping 0.0114 re_causal 0.0303 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 39---------------------------------------------------- +epoch 39, time 263.42, cls_loss 0.0148 cls_loss_mapping 0.0251 cls_loss_causal 0.6789 re_mapping 0.0111 re_causal 0.0298 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 40---------------------------------------------------- +epoch 40, time 263.43, cls_loss 0.0165 cls_loss_mapping 0.0301 cls_loss_causal 0.6877 re_mapping 0.0113 re_causal 0.0297 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 41, time 262.57, cls_loss 0.0161 cls_loss_mapping 0.0290 cls_loss_causal 0.6867 re_mapping 0.0103 re_causal 0.0283 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 42, time 261.54, cls_loss 0.0124 cls_loss_mapping 0.0221 cls_loss_causal 0.6524 re_mapping 0.0104 re_causal 0.0276 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 43, time 261.64, cls_loss 0.0121 cls_loss_mapping 0.0236 cls_loss_causal 0.6499 re_mapping 0.0107 re_causal 0.0281 /// teacc 98.87 lr 0.00010000 +100 +0.0001 +changing lr +epoch 44, time 261.70, cls_loss 0.0126 cls_loss_mapping 0.0222 cls_loss_causal 0.6472 re_mapping 0.0107 re_causal 0.0277 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 45, time 262.23, cls_loss 0.0139 cls_loss_mapping 0.0248 cls_loss_causal 0.6458 re_mapping 0.0097 re_causal 0.0267 /// teacc 98.72 lr 0.00010000 +100 +0.0001 +changing lr +epoch 46, time 262.35, cls_loss 0.0128 cls_loss_mapping 0.0236 cls_loss_causal 0.6192 re_mapping 0.0103 re_causal 0.0264 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +epoch 47, time 262.98, cls_loss 0.0120 cls_loss_mapping 0.0198 cls_loss_causal 0.6455 re_mapping 0.0097 re_causal 0.0258 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 48, time 262.58, cls_loss 0.0116 cls_loss_mapping 0.0229 cls_loss_causal 0.6623 re_mapping 0.0099 re_causal 0.0264 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 49, time 262.10, cls_loss 0.0109 cls_loss_mapping 0.0222 cls_loss_causal 0.6632 re_mapping 0.0094 re_causal 0.0260 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 50, time 262.49, cls_loss 0.0107 cls_loss_mapping 0.0186 cls_loss_causal 0.6425 re_mapping 0.0094 re_causal 0.0260 /// teacc 98.74 lr 0.00010000 +100 +0.0001 +changing lr +epoch 51, time 261.98, cls_loss 0.0105 cls_loss_mapping 0.0196 cls_loss_causal 0.6062 re_mapping 0.0099 re_causal 0.0249 /// teacc 98.77 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 52---------------------------------------------------- +epoch 52, time 262.70, cls_loss 0.0123 cls_loss_mapping 0.0222 cls_loss_causal 0.6539 re_mapping 0.0090 re_causal 0.0243 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 53, time 262.04, cls_loss 0.0082 cls_loss_mapping 0.0165 cls_loss_causal 0.5830 re_mapping 0.0095 re_causal 0.0243 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 54, time 262.03, cls_loss 0.0134 cls_loss_mapping 0.0238 cls_loss_causal 0.6506 re_mapping 0.0092 re_causal 0.0241 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 55, time 262.23, cls_loss 0.0092 cls_loss_mapping 0.0175 cls_loss_causal 0.6151 re_mapping 0.0094 re_causal 0.0241 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 56, time 261.68, cls_loss 0.0083 cls_loss_mapping 0.0146 cls_loss_causal 0.6247 re_mapping 0.0093 re_causal 0.0250 /// teacc 98.78 lr 0.00010000 +100 +0.0001 +changing lr +epoch 57, time 262.04, cls_loss 0.0094 cls_loss_mapping 0.0173 cls_loss_causal 0.6450 re_mapping 0.0082 re_causal 0.0236 /// teacc 98.80 lr 0.00010000 +100 +0.0001 +changing lr +epoch 58, time 262.22, cls_loss 0.0082 cls_loss_mapping 0.0182 cls_loss_causal 0.5940 re_mapping 0.0090 re_causal 0.0236 /// teacc 98.76 lr 0.00010000 +100 +0.0001 +changing lr +epoch 59, time 261.98, cls_loss 0.0107 cls_loss_mapping 0.0187 cls_loss_causal 0.6018 re_mapping 0.0082 re_causal 0.0217 /// teacc 98.85 lr 0.00010000 +100 +0.0001 +changing lr +epoch 60, time 262.64, cls_loss 0.0096 cls_loss_mapping 0.0165 cls_loss_causal 0.6197 re_mapping 0.0079 re_causal 0.0227 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 61, time 262.62, cls_loss 0.0077 cls_loss_mapping 0.0133 cls_loss_causal 0.6104 re_mapping 0.0077 re_causal 0.0216 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 62, time 261.55, cls_loss 0.0094 cls_loss_mapping 0.0177 cls_loss_causal 0.6325 re_mapping 0.0077 re_causal 0.0211 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 63, time 261.66, cls_loss 0.0096 cls_loss_mapping 0.0173 cls_loss_causal 0.6390 re_mapping 0.0075 re_causal 0.0211 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 64, time 262.39, cls_loss 0.0089 cls_loss_mapping 0.0176 cls_loss_causal 0.6220 re_mapping 0.0080 re_causal 0.0211 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 65, time 262.81, cls_loss 0.0054 cls_loss_mapping 0.0089 cls_loss_causal 0.5919 re_mapping 0.0081 re_causal 0.0215 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 66, time 262.62, cls_loss 0.0072 cls_loss_mapping 0.0145 cls_loss_causal 0.5995 re_mapping 0.0079 re_causal 0.0213 /// teacc 98.80 lr 0.00010000 +100 +0.0001 +changing lr +epoch 67, time 262.54, cls_loss 0.0065 cls_loss_mapping 0.0115 cls_loss_causal 0.5839 re_mapping 0.0082 re_causal 0.0214 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 68, time 262.66, cls_loss 0.0082 cls_loss_mapping 0.0151 cls_loss_causal 0.6010 re_mapping 0.0072 re_causal 0.0203 /// teacc 98.68 lr 0.00010000 +100 +0.0001 +changing lr +epoch 69, time 262.04, cls_loss 0.0072 cls_loss_mapping 0.0131 cls_loss_causal 0.5964 re_mapping 0.0075 re_causal 0.0202 /// teacc 98.86 lr 0.00010000 +100 +0.0001 +changing lr +epoch 70, time 262.42, cls_loss 0.0068 cls_loss_mapping 0.0141 cls_loss_causal 0.6231 re_mapping 0.0076 re_causal 0.0214 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 71, time 261.65, cls_loss 0.0077 cls_loss_mapping 0.0151 cls_loss_causal 0.5752 re_mapping 0.0073 re_causal 0.0197 /// teacc 98.89 lr 0.00010000 +100 +0.0001 +changing lr +epoch 72, time 262.42, cls_loss 0.0077 cls_loss_mapping 0.0130 cls_loss_causal 0.5860 re_mapping 0.0073 re_causal 0.0197 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 73---------------------------------------------------- +epoch 73, time 263.56, cls_loss 0.0074 cls_loss_mapping 0.0145 cls_loss_causal 0.5783 re_mapping 0.0071 re_causal 0.0192 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 74, time 262.97, cls_loss 0.0064 cls_loss_mapping 0.0111 cls_loss_causal 0.5752 re_mapping 0.0073 re_causal 0.0198 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 75, time 262.29, cls_loss 0.0067 cls_loss_mapping 0.0128 cls_loss_causal 0.5738 re_mapping 0.0072 re_causal 0.0190 /// teacc 98.84 lr 0.00010000 +100 +0.0001 +changing lr +epoch 76, time 262.02, cls_loss 0.0060 cls_loss_mapping 0.0107 cls_loss_causal 0.5944 re_mapping 0.0071 re_causal 0.0198 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 77, time 262.25, cls_loss 0.0055 cls_loss_mapping 0.0086 cls_loss_causal 0.5748 re_mapping 0.0066 re_causal 0.0185 /// teacc 98.83 lr 0.00010000 +100 +0.0001 +changing lr +epoch 78, time 262.33, cls_loss 0.0048 cls_loss_mapping 0.0105 cls_loss_causal 0.5729 re_mapping 0.0071 re_causal 0.0200 /// teacc 98.86 lr 0.00010000 +100 +0.0001 +changing lr +epoch 79, time 262.67, cls_loss 0.0054 cls_loss_mapping 0.0086 cls_loss_causal 0.5782 re_mapping 0.0067 re_causal 0.0187 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 80, time 262.54, cls_loss 0.0048 cls_loss_mapping 0.0092 cls_loss_causal 0.5620 re_mapping 0.0067 re_causal 0.0185 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 81, time 262.50, cls_loss 0.0063 cls_loss_mapping 0.0131 cls_loss_causal 0.6240 re_mapping 0.0067 re_causal 0.0190 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 82, time 261.76, cls_loss 0.0077 cls_loss_mapping 0.0136 cls_loss_causal 0.5922 re_mapping 0.0067 re_causal 0.0178 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 83, time 262.17, cls_loss 0.0064 cls_loss_mapping 0.0120 cls_loss_causal 0.5514 re_mapping 0.0073 re_causal 0.0188 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 84, time 262.19, cls_loss 0.0056 cls_loss_mapping 0.0093 cls_loss_causal 0.5766 re_mapping 0.0065 re_causal 0.0180 /// teacc 98.85 lr 0.00010000 +100 +0.0001 +changing lr +epoch 85, time 262.14, cls_loss 0.0050 cls_loss_mapping 0.0080 cls_loss_causal 0.5528 re_mapping 0.0063 re_causal 0.0174 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 86, time 261.86, cls_loss 0.0051 cls_loss_mapping 0.0088 cls_loss_causal 0.5929 re_mapping 0.0063 re_causal 0.0178 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 87, time 261.74, cls_loss 0.0050 cls_loss_mapping 0.0087 cls_loss_causal 0.5941 re_mapping 0.0063 re_causal 0.0177 /// teacc 98.80 lr 0.00010000 +100 +0.0001 +changing lr +epoch 88, time 260.51, cls_loss 0.0048 cls_loss_mapping 0.0085 cls_loss_causal 0.5624 re_mapping 0.0064 re_causal 0.0177 /// teacc 98.88 lr 0.00010000 +100 +0.0001 +changing lr +epoch 89, time 250.06, cls_loss 0.0047 cls_loss_mapping 0.0084 cls_loss_causal 0.5650 re_mapping 0.0065 re_causal 0.0173 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 90, time 250.78, cls_loss 0.0049 cls_loss_mapping 0.0091 cls_loss_causal 0.5613 re_mapping 0.0060 re_causal 0.0167 /// teacc 98.88 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 91---------------------------------------------------- +epoch 91, time 251.20, cls_loss 0.0041 cls_loss_mapping 0.0070 cls_loss_causal 0.5382 re_mapping 0.0064 re_causal 0.0165 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 92, time 251.25, cls_loss 0.0051 cls_loss_mapping 0.0108 cls_loss_causal 0.6002 re_mapping 0.0059 re_causal 0.0168 /// teacc 98.89 lr 0.00010000 +100 +0.0001 +changing lr +epoch 93, time 250.57, cls_loss 0.0049 cls_loss_mapping 0.0096 cls_loss_causal 0.5548 re_mapping 0.0063 re_causal 0.0168 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 94, time 250.77, cls_loss 0.0047 cls_loss_mapping 0.0096 cls_loss_causal 0.5460 re_mapping 0.0063 re_causal 0.0163 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 95, time 250.37, cls_loss 0.0045 cls_loss_mapping 0.0074 cls_loss_causal 0.5265 re_mapping 0.0064 re_causal 0.0160 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 96, time 249.18, cls_loss 0.0037 cls_loss_mapping 0.0063 cls_loss_causal 0.5633 re_mapping 0.0062 re_causal 0.0172 /// teacc 98.82 lr 0.00010000 +100 +0.0001 +changing lr +epoch 97, time 249.34, cls_loss 0.0051 cls_loss_mapping 0.0080 cls_loss_causal 0.5467 re_mapping 0.0057 re_causal 0.0156 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 98, time 249.24, cls_loss 0.0043 cls_loss_mapping 0.0077 cls_loss_causal 0.5665 re_mapping 0.0061 re_causal 0.0163 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 99, time 247.66, cls_loss 0.0042 cls_loss_mapping 0.0055 cls_loss_causal 0.5559 re_mapping 0.0059 re_causal 0.0160 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 100, time 247.43, cls_loss 0.0039 cls_loss_mapping 0.0072 cls_loss_causal 0.5491 re_mapping 0.0059 re_causal 0.0159 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 101, time 247.17, cls_loss 0.0036 cls_loss_mapping 0.0062 cls_loss_causal 0.5947 re_mapping 0.0058 re_causal 0.0166 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 102, time 247.21, cls_loss 0.0041 cls_loss_mapping 0.0065 cls_loss_causal 0.5484 re_mapping 0.0057 re_causal 0.0155 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 103, time 247.20, cls_loss 0.0047 cls_loss_mapping 0.0077 cls_loss_causal 0.5315 re_mapping 0.0056 re_causal 0.0149 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 104, time 247.26, cls_loss 0.0047 cls_loss_mapping 0.0076 cls_loss_causal 0.5507 re_mapping 0.0055 re_causal 0.0148 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 105, time 247.09, cls_loss 0.0040 cls_loss_mapping 0.0063 cls_loss_causal 0.5417 re_mapping 0.0052 re_causal 0.0150 /// teacc 98.90 lr 0.00010000 +100 +0.0001 +changing lr +epoch 106, time 247.22, cls_loss 0.0046 cls_loss_mapping 0.0085 cls_loss_causal 0.5688 re_mapping 0.0053 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 107, time 247.40, cls_loss 0.0039 cls_loss_mapping 0.0085 cls_loss_causal 0.5396 re_mapping 0.0057 re_causal 0.0155 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 108, time 247.52, cls_loss 0.0047 cls_loss_mapping 0.0094 cls_loss_causal 0.5722 re_mapping 0.0056 re_causal 0.0150 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 109, time 247.28, cls_loss 0.0036 cls_loss_mapping 0.0055 cls_loss_causal 0.5219 re_mapping 0.0055 re_causal 0.0145 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 110, time 247.31, cls_loss 0.0033 cls_loss_mapping 0.0053 cls_loss_causal 0.5339 re_mapping 0.0056 re_causal 0.0153 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 111---------------------------------------------------- +epoch 111, time 248.20, cls_loss 0.0044 cls_loss_mapping 0.0070 cls_loss_causal 0.5686 re_mapping 0.0051 re_causal 0.0146 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 112, time 247.44, cls_loss 0.0037 cls_loss_mapping 0.0064 cls_loss_causal 0.5641 re_mapping 0.0053 re_causal 0.0150 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 113, time 247.39, cls_loss 0.0037 cls_loss_mapping 0.0063 cls_loss_causal 0.5414 re_mapping 0.0054 re_causal 0.0149 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 114, time 247.52, cls_loss 0.0039 cls_loss_mapping 0.0082 cls_loss_causal 0.5541 re_mapping 0.0051 re_causal 0.0144 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 115, time 247.69, cls_loss 0.0040 cls_loss_mapping 0.0066 cls_loss_causal 0.5456 re_mapping 0.0054 re_causal 0.0145 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 116, time 247.22, cls_loss 0.0031 cls_loss_mapping 0.0053 cls_loss_causal 0.5168 re_mapping 0.0053 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 117---------------------------------------------------- +epoch 117, time 248.08, cls_loss 0.0053 cls_loss_mapping 0.0090 cls_loss_causal 0.5568 re_mapping 0.0053 re_causal 0.0148 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 118, time 247.60, cls_loss 0.0033 cls_loss_mapping 0.0064 cls_loss_causal 0.5252 re_mapping 0.0052 re_causal 0.0147 /// teacc 98.92 lr 0.00010000 +100 +0.0001 +changing lr +epoch 119, time 247.65, cls_loss 0.0033 cls_loss_mapping 0.0068 cls_loss_causal 0.5163 re_mapping 0.0053 re_causal 0.0149 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 120---------------------------------------------------- +epoch 120, time 249.12, cls_loss 0.0041 cls_loss_mapping 0.0073 cls_loss_causal 0.5428 re_mapping 0.0048 re_causal 0.0140 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 121, time 247.46, cls_loss 0.0038 cls_loss_mapping 0.0055 cls_loss_causal 0.5502 re_mapping 0.0047 re_causal 0.0134 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 122, time 247.34, cls_loss 0.0040 cls_loss_mapping 0.0070 cls_loss_causal 0.5413 re_mapping 0.0049 re_causal 0.0141 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 123, time 247.75, cls_loss 0.0029 cls_loss_mapping 0.0051 cls_loss_causal 0.5293 re_mapping 0.0052 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 124, time 247.49, cls_loss 0.0039 cls_loss_mapping 0.0059 cls_loss_causal 0.5299 re_mapping 0.0048 re_causal 0.0137 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 125, time 247.82, cls_loss 0.0035 cls_loss_mapping 0.0055 cls_loss_causal 0.5164 re_mapping 0.0052 re_causal 0.0143 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 126, time 247.56, cls_loss 0.0033 cls_loss_mapping 0.0056 cls_loss_causal 0.5298 re_mapping 0.0050 re_causal 0.0141 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 127, time 247.90, cls_loss 0.0032 cls_loss_mapping 0.0058 cls_loss_causal 0.5069 re_mapping 0.0051 re_causal 0.0140 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 128, time 247.44, cls_loss 0.0035 cls_loss_mapping 0.0061 cls_loss_causal 0.5469 re_mapping 0.0046 re_causal 0.0133 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 129, time 247.22, cls_loss 0.0035 cls_loss_mapping 0.0046 cls_loss_causal 0.5124 re_mapping 0.0049 re_causal 0.0131 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 130, time 247.53, cls_loss 0.0041 cls_loss_mapping 0.0070 cls_loss_causal 0.5574 re_mapping 0.0048 re_causal 0.0133 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 131, time 247.26, cls_loss 0.0034 cls_loss_mapping 0.0052 cls_loss_causal 0.5246 re_mapping 0.0049 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 132, time 247.37, cls_loss 0.0034 cls_loss_mapping 0.0064 cls_loss_causal 0.5529 re_mapping 0.0047 re_causal 0.0142 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 133, time 247.41, cls_loss 0.0035 cls_loss_mapping 0.0043 cls_loss_causal 0.5204 re_mapping 0.0046 re_causal 0.0126 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 134, time 247.49, cls_loss 0.0033 cls_loss_mapping 0.0055 cls_loss_causal 0.5262 re_mapping 0.0045 re_causal 0.0127 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 135, time 247.43, cls_loss 0.0031 cls_loss_mapping 0.0054 cls_loss_causal 0.5655 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 136, time 247.17, cls_loss 0.0030 cls_loss_mapping 0.0045 cls_loss_causal 0.5369 re_mapping 0.0046 re_causal 0.0129 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 137---------------------------------------------------- +epoch 137, time 247.90, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.4877 re_mapping 0.0049 re_causal 0.0136 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 138, time 247.42, cls_loss 0.0034 cls_loss_mapping 0.0051 cls_loss_causal 0.5592 re_mapping 0.0045 re_causal 0.0132 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 139, time 247.30, cls_loss 0.0035 cls_loss_mapping 0.0068 cls_loss_causal 0.4932 re_mapping 0.0048 re_causal 0.0133 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 140, time 247.43, cls_loss 0.0030 cls_loss_mapping 0.0041 cls_loss_causal 0.5293 re_mapping 0.0046 re_causal 0.0132 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 141, time 247.95, cls_loss 0.0024 cls_loss_mapping 0.0036 cls_loss_causal 0.5366 re_mapping 0.0045 re_causal 0.0134 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 142, time 247.33, cls_loss 0.0033 cls_loss_mapping 0.0058 cls_loss_causal 0.5003 re_mapping 0.0042 re_causal 0.0122 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 143, time 247.18, cls_loss 0.0030 cls_loss_mapping 0.0053 cls_loss_causal 0.5321 re_mapping 0.0044 re_causal 0.0128 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 144, time 247.52, cls_loss 0.0032 cls_loss_mapping 0.0051 cls_loss_causal 0.4899 re_mapping 0.0044 re_causal 0.0121 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 145, time 247.52, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.5202 re_mapping 0.0047 re_causal 0.0134 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 146, time 248.01, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.4945 re_mapping 0.0044 re_causal 0.0126 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 147, time 247.07, cls_loss 0.0025 cls_loss_mapping 0.0037 cls_loss_causal 0.5273 re_mapping 0.0047 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 148, time 247.44, cls_loss 0.0029 cls_loss_mapping 0.0042 cls_loss_causal 0.5309 re_mapping 0.0046 re_causal 0.0126 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 149, time 246.82, cls_loss 0.0030 cls_loss_mapping 0.0043 cls_loss_causal 0.5280 re_mapping 0.0042 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 150, time 247.07, cls_loss 0.0025 cls_loss_mapping 0.0038 cls_loss_causal 0.5050 re_mapping 0.0042 re_causal 0.0121 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 151, time 247.12, cls_loss 0.0030 cls_loss_mapping 0.0058 cls_loss_causal 0.5175 re_mapping 0.0042 re_causal 0.0118 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 152, time 247.31, cls_loss 0.0028 cls_loss_mapping 0.0039 cls_loss_causal 0.5003 re_mapping 0.0041 re_causal 0.0114 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 153, time 247.30, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.5100 re_mapping 0.0043 re_causal 0.0120 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 154, time 247.18, cls_loss 0.0028 cls_loss_mapping 0.0035 cls_loss_causal 0.5038 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 155, time 247.28, cls_loss 0.0030 cls_loss_mapping 0.0046 cls_loss_causal 0.5092 re_mapping 0.0045 re_causal 0.0121 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 156, time 247.53, cls_loss 0.0050 cls_loss_mapping 0.0085 cls_loss_causal 0.5153 re_mapping 0.0044 re_causal 0.0121 /// teacc 98.89 lr 0.00010000 +100 +0.0001 +changing lr +epoch 157, time 247.52, cls_loss 0.0027 cls_loss_mapping 0.0043 cls_loss_causal 0.5363 re_mapping 0.0044 re_causal 0.0125 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 158, time 247.53, cls_loss 0.0020 cls_loss_mapping 0.0042 cls_loss_causal 0.4788 re_mapping 0.0043 re_causal 0.0124 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 159, time 247.44, cls_loss 0.0027 cls_loss_mapping 0.0053 cls_loss_causal 0.5289 re_mapping 0.0040 re_causal 0.0117 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 160, time 247.25, cls_loss 0.0031 cls_loss_mapping 0.0043 cls_loss_causal 0.4845 re_mapping 0.0040 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 161, time 247.20, cls_loss 0.0023 cls_loss_mapping 0.0031 cls_loss_causal 0.5342 re_mapping 0.0042 re_causal 0.0119 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 162, time 247.36, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.5377 re_mapping 0.0041 re_causal 0.0118 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 163, time 247.20, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5306 re_mapping 0.0040 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 164, time 247.35, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5117 re_mapping 0.0042 re_causal 0.0119 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 165, time 247.03, cls_loss 0.0026 cls_loss_mapping 0.0040 cls_loss_causal 0.5038 re_mapping 0.0040 re_causal 0.0114 /// teacc 98.93 lr 0.00010000 +100 +0.0001 +changing lr +epoch 166, time 247.07, cls_loss 0.0025 cls_loss_mapping 0.0041 cls_loss_causal 0.5101 re_mapping 0.0042 re_causal 0.0118 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 167, time 247.31, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.5069 re_mapping 0.0042 re_causal 0.0115 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 168, time 247.20, cls_loss 0.0025 cls_loss_mapping 0.0036 cls_loss_causal 0.5038 re_mapping 0.0041 re_causal 0.0114 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 169, time 247.25, cls_loss 0.0029 cls_loss_mapping 0.0037 cls_loss_causal 0.5111 re_mapping 0.0041 re_causal 0.0109 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 170, time 247.40, cls_loss 0.0030 cls_loss_mapping 0.0044 cls_loss_causal 0.5222 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 171, time 247.19, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.5095 re_mapping 0.0039 re_causal 0.0115 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 172, time 247.11, cls_loss 0.0023 cls_loss_mapping 0.0030 cls_loss_causal 0.5020 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 173, time 246.80, cls_loss 0.0024 cls_loss_mapping 0.0035 cls_loss_causal 0.5326 re_mapping 0.0038 re_causal 0.0111 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 174, time 247.13, cls_loss 0.0024 cls_loss_mapping 0.0026 cls_loss_causal 0.5236 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 175, time 247.24, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.4945 re_mapping 0.0037 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 176, time 247.09, cls_loss 0.0024 cls_loss_mapping 0.0042 cls_loss_causal 0.5163 re_mapping 0.0039 re_causal 0.0114 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 177, time 246.73, cls_loss 0.0025 cls_loss_mapping 0.0033 cls_loss_causal 0.5106 re_mapping 0.0037 re_causal 0.0110 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 178, time 247.11, cls_loss 0.0022 cls_loss_mapping 0.0025 cls_loss_causal 0.4798 re_mapping 0.0040 re_causal 0.0111 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 179, time 246.93, cls_loss 0.0022 cls_loss_mapping 0.0039 cls_loss_causal 0.5108 re_mapping 0.0040 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 180, time 247.21, cls_loss 0.0030 cls_loss_mapping 0.0037 cls_loss_causal 0.5233 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 181, time 247.09, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.5065 re_mapping 0.0039 re_causal 0.0113 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 182, time 246.91, cls_loss 0.0024 cls_loss_mapping 0.0037 cls_loss_causal 0.5588 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 183, time 247.64, cls_loss 0.0026 cls_loss_mapping 0.0037 cls_loss_causal 0.5331 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 184, time 247.11, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5064 re_mapping 0.0039 re_causal 0.0110 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 185, time 247.25, cls_loss 0.0025 cls_loss_mapping 0.0035 cls_loss_causal 0.4997 re_mapping 0.0038 re_causal 0.0109 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 186, time 247.56, cls_loss 0.0023 cls_loss_mapping 0.0033 cls_loss_causal 0.5319 re_mapping 0.0038 re_causal 0.0112 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 187, time 246.96, cls_loss 0.0027 cls_loss_mapping 0.0039 cls_loss_causal 0.5077 re_mapping 0.0035 re_causal 0.0098 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 188, time 247.08, cls_loss 0.0018 cls_loss_mapping 0.0029 cls_loss_causal 0.4799 re_mapping 0.0039 re_causal 0.0109 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 189, time 247.25, cls_loss 0.0017 cls_loss_mapping 0.0028 cls_loss_causal 0.4788 re_mapping 0.0040 re_causal 0.0112 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 190, time 246.87, cls_loss 0.0025 cls_loss_mapping 0.0032 cls_loss_causal 0.4994 re_mapping 0.0035 re_causal 0.0103 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 191, time 247.17, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.4759 re_mapping 0.0039 re_causal 0.0106 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 192, time 246.82, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5043 re_mapping 0.0039 re_causal 0.0108 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 193, time 247.13, cls_loss 0.0025 cls_loss_mapping 0.0043 cls_loss_causal 0.5180 re_mapping 0.0035 re_causal 0.0103 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 194, time 246.74, cls_loss 0.0021 cls_loss_mapping 0.0038 cls_loss_causal 0.5164 re_mapping 0.0037 re_causal 0.0113 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 195, time 246.90, cls_loss 0.0028 cls_loss_mapping 0.0044 cls_loss_causal 0.5003 re_mapping 0.0037 re_causal 0.0105 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 196, time 246.96, cls_loss 0.0024 cls_loss_mapping 0.0043 cls_loss_causal 0.5004 re_mapping 0.0037 re_causal 0.0107 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 197, time 247.11, cls_loss 0.0020 cls_loss_mapping 0.0031 cls_loss_causal 0.4946 re_mapping 0.0038 re_causal 0.0107 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 198, time 247.15, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5043 re_mapping 0.0036 re_causal 0.0107 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 199, time 246.86, cls_loss 0.0023 cls_loss_mapping 0.0034 cls_loss_causal 0.4853 re_mapping 0.0036 re_causal 0.0102 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 200, time 247.43, cls_loss 0.0021 cls_loss_mapping 0.0037 cls_loss_causal 0.4856 re_mapping 0.0034 re_causal 0.0103 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 201, time 247.24, cls_loss 0.0019 cls_loss_mapping 0.0023 cls_loss_causal 0.5071 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 202, time 247.65, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.5178 re_mapping 0.0035 re_causal 0.0104 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 203, time 247.52, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4825 re_mapping 0.0033 re_causal 0.0094 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 204, time 247.32, cls_loss 0.0018 cls_loss_mapping 0.0021 cls_loss_causal 0.4940 re_mapping 0.0036 re_causal 0.0101 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 205, time 247.61, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.5107 re_mapping 0.0039 re_causal 0.0105 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 206, time 247.53, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.4936 re_mapping 0.0036 re_causal 0.0099 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 207, time 247.27, cls_loss 0.0024 cls_loss_mapping 0.0033 cls_loss_causal 0.4938 re_mapping 0.0033 re_causal 0.0095 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 208, time 247.31, cls_loss 0.0020 cls_loss_mapping 0.0024 cls_loss_causal 0.4524 re_mapping 0.0035 re_causal 0.0101 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 209, time 247.73, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4973 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 210, time 247.40, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4702 re_mapping 0.0036 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 211, time 246.89, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5196 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 212, time 247.35, cls_loss 0.0021 cls_loss_mapping 0.0034 cls_loss_causal 0.4901 re_mapping 0.0034 re_causal 0.0096 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 213, time 247.22, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5019 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 214, time 247.39, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4779 re_mapping 0.0034 re_causal 0.0100 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 215, time 247.27, cls_loss 0.0024 cls_loss_mapping 0.0040 cls_loss_causal 0.4916 re_mapping 0.0033 re_causal 0.0095 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 216, time 247.37, cls_loss 0.0016 cls_loss_mapping 0.0017 cls_loss_causal 0.4976 re_mapping 0.0033 re_causal 0.0097 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 217, time 247.45, cls_loss 0.0023 cls_loss_mapping 0.0035 cls_loss_causal 0.4677 re_mapping 0.0034 re_causal 0.0095 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 218, time 247.74, cls_loss 0.0020 cls_loss_mapping 0.0023 cls_loss_causal 0.4740 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 219, time 247.00, cls_loss 0.0020 cls_loss_mapping 0.0024 cls_loss_causal 0.4902 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 220, time 247.26, cls_loss 0.0015 cls_loss_mapping 0.0015 cls_loss_causal 0.4984 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 221, time 247.21, cls_loss 0.0018 cls_loss_mapping 0.0018 cls_loss_causal 0.4791 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 222, time 247.27, cls_loss 0.0019 cls_loss_mapping 0.0025 cls_loss_causal 0.4897 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 223, time 247.18, cls_loss 0.0022 cls_loss_mapping 0.0027 cls_loss_causal 0.5187 re_mapping 0.0031 re_causal 0.0093 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 224, time 247.18, cls_loss 0.0020 cls_loss_mapping 0.0025 cls_loss_causal 0.4952 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 225, time 247.25, cls_loss 0.0021 cls_loss_mapping 0.0027 cls_loss_causal 0.4951 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 226, time 247.42, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.5013 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 227, time 248.14, cls_loss 0.0016 cls_loss_mapping 0.0030 cls_loss_causal 0.5144 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 228, time 247.56, cls_loss 0.0020 cls_loss_mapping 0.0027 cls_loss_causal 0.5000 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 229---------------------------------------------------- +epoch 229, time 247.85, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.5045 re_mapping 0.0033 re_causal 0.0098 /// teacc 99.21 lr 0.00010000 +100 +0.0001 +changing lr +epoch 230, time 247.45, cls_loss 0.0020 cls_loss_mapping 0.0033 cls_loss_causal 0.5028 re_mapping 0.0032 re_causal 0.0097 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 231, time 247.60, cls_loss 0.0024 cls_loss_mapping 0.0033 cls_loss_causal 0.5090 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 232, time 247.14, cls_loss 0.0024 cls_loss_mapping 0.0037 cls_loss_causal 0.4987 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.18 lr 0.00010000 +100 +0.0001 +changing lr +epoch 233, time 247.34, cls_loss 0.0015 cls_loss_mapping 0.0025 cls_loss_causal 0.5306 re_mapping 0.0032 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 234, time 247.70, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4880 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 235, time 247.42, cls_loss 0.0017 cls_loss_mapping 0.0020 cls_loss_causal 0.4734 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 236, time 248.01, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.4746 re_mapping 0.0032 re_causal 0.0093 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 237, time 247.40, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4826 re_mapping 0.0029 re_causal 0.0088 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 238, time 247.38, cls_loss 0.0017 cls_loss_mapping 0.0016 cls_loss_causal 0.5047 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 239, time 248.73, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.5064 re_mapping 0.0030 re_causal 0.0092 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 240, time 248.15, cls_loss 0.0018 cls_loss_mapping 0.0034 cls_loss_causal 0.5029 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 241, time 247.23, cls_loss 0.0017 cls_loss_mapping 0.0023 cls_loss_causal 0.4986 re_mapping 0.0032 re_causal 0.0096 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 242, time 247.18, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.4912 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 243, time 247.37, cls_loss 0.0017 cls_loss_mapping 0.0024 cls_loss_causal 0.4714 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 244, time 247.36, cls_loss 0.0018 cls_loss_mapping 0.0030 cls_loss_causal 0.4707 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 245, time 247.07, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4907 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 246, time 247.63, cls_loss 0.0017 cls_loss_mapping 0.0039 cls_loss_causal 0.5042 re_mapping 0.0032 re_causal 0.0098 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 247, time 247.28, cls_loss 0.0026 cls_loss_mapping 0.0037 cls_loss_causal 0.4860 re_mapping 0.0031 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 248, time 247.15, cls_loss 0.0021 cls_loss_mapping 0.0031 cls_loss_causal 0.4790 re_mapping 0.0033 re_causal 0.0095 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 249, time 247.23, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4878 re_mapping 0.0031 re_causal 0.0088 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 250, time 247.02, cls_loss 0.0015 cls_loss_mapping 0.0016 cls_loss_causal 0.4962 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 251, time 247.51, cls_loss 0.0018 cls_loss_mapping 0.0028 cls_loss_causal 0.4979 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 252, time 247.35, cls_loss 0.0020 cls_loss_mapping 0.0024 cls_loss_causal 0.4525 re_mapping 0.0031 re_causal 0.0088 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 253, time 247.04, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4552 re_mapping 0.0030 re_causal 0.0087 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 254, time 247.26, cls_loss 0.0022 cls_loss_mapping 0.0036 cls_loss_causal 0.4710 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 255, time 247.16, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4512 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 256, time 247.38, cls_loss 0.0018 cls_loss_mapping 0.0023 cls_loss_causal 0.4975 re_mapping 0.0028 re_causal 0.0086 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 257, time 247.20, cls_loss 0.0016 cls_loss_mapping 0.0018 cls_loss_causal 0.4632 re_mapping 0.0032 re_causal 0.0094 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 258, time 247.12, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4664 re_mapping 0.0032 re_causal 0.0091 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 259, time 247.32, cls_loss 0.0018 cls_loss_mapping 0.0025 cls_loss_causal 0.4997 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 260, time 247.19, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4738 re_mapping 0.0029 re_causal 0.0088 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 261, time 246.87, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4996 re_mapping 0.0032 re_causal 0.0095 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 262, time 246.91, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5120 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 263, time 247.18, cls_loss 0.0016 cls_loss_mapping 0.0025 cls_loss_causal 0.4762 re_mapping 0.0031 re_causal 0.0087 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 264, time 246.94, cls_loss 0.0015 cls_loss_mapping 0.0019 cls_loss_causal 0.4728 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 265, time 246.92, cls_loss 0.0017 cls_loss_mapping 0.0016 cls_loss_causal 0.4729 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 266, time 246.97, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.4830 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 267, time 247.19, cls_loss 0.0016 cls_loss_mapping 0.0028 cls_loss_causal 0.4905 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 268, time 247.41, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4688 re_mapping 0.0031 re_causal 0.0091 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 269, time 247.14, cls_loss 0.0021 cls_loss_mapping 0.0027 cls_loss_causal 0.5079 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.18 lr 0.00010000 +100 +0.0001 +changing lr +epoch 270, time 247.08, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4751 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 271, time 247.23, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.4807 re_mapping 0.0029 re_causal 0.0087 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 272, time 247.25, cls_loss 0.0021 cls_loss_mapping 0.0029 cls_loss_causal 0.4811 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 273, time 247.12, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4693 re_mapping 0.0030 re_causal 0.0088 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 274, time 247.52, cls_loss 0.0018 cls_loss_mapping 0.0019 cls_loss_causal 0.4625 re_mapping 0.0030 re_causal 0.0081 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 275, time 247.32, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4594 re_mapping 0.0030 re_causal 0.0086 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 276, time 247.31, cls_loss 0.0013 cls_loss_mapping 0.0030 cls_loss_causal 0.4717 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 277, time 247.59, cls_loss 0.0021 cls_loss_mapping 0.0022 cls_loss_causal 0.4800 re_mapping 0.0029 re_causal 0.0084 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 278, time 247.25, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.4832 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 279, time 247.30, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4871 re_mapping 0.0030 re_causal 0.0091 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 280---------------------------------------------------- +epoch 280, time 248.43, cls_loss 0.0016 cls_loss_mapping 0.0017 cls_loss_causal 0.4657 re_mapping 0.0030 re_causal 0.0084 /// teacc 99.26 lr 0.00010000 +100 +0.0001 +changing lr +epoch 281, time 247.76, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4639 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.25 lr 0.00010000 +100 +0.0001 +changing lr +epoch 282, time 247.74, cls_loss 0.0014 cls_loss_mapping 0.0019 cls_loss_causal 0.4466 re_mapping 0.0029 re_causal 0.0082 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 283, time 247.60, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4532 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 284, time 247.58, cls_loss 0.0020 cls_loss_mapping 0.0031 cls_loss_causal 0.4614 re_mapping 0.0029 re_causal 0.0086 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 285, time 247.57, cls_loss 0.0022 cls_loss_mapping 0.0026 cls_loss_causal 0.5009 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 286, time 247.67, cls_loss 0.0017 cls_loss_mapping 0.0019 cls_loss_causal 0.4442 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.15 lr 0.00010000 +100 +0.0001 +changing lr +epoch 287, time 247.52, cls_loss 0.0018 cls_loss_mapping 0.0041 cls_loss_causal 0.4619 re_mapping 0.0029 re_causal 0.0083 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 288, time 246.94, cls_loss 0.0012 cls_loss_mapping 0.0019 cls_loss_causal 0.4668 re_mapping 0.0030 re_causal 0.0090 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 289, time 247.41, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.4698 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 290, time 247.31, cls_loss 0.0016 cls_loss_mapping 0.0021 cls_loss_causal 0.4558 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 291, time 247.85, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4896 re_mapping 0.0030 re_causal 0.0089 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 292, time 247.58, cls_loss 0.0018 cls_loss_mapping 0.0017 cls_loss_causal 0.4845 re_mapping 0.0025 re_causal 0.0075 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 293, time 246.85, cls_loss 0.0018 cls_loss_mapping 0.0019 cls_loss_causal 0.4797 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 294, time 247.06, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4528 re_mapping 0.0029 re_causal 0.0080 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +---------------------saving model at epoch 295---------------------------------------------------- +epoch 295, time 248.20, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4663 re_mapping 0.0029 re_causal 0.0085 /// teacc 99.27 lr 0.00010000 +100 +0.0001 +changing lr +epoch 296, time 247.05, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4457 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.19 lr 0.00010000 +100 +0.0001 +changing lr +epoch 297, time 247.01, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.4646 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 298, time 247.03, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4582 re_mapping 0.0027 re_causal 0.0081 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 299, time 247.08, cls_loss 0.0015 cls_loss_mapping 0.0017 cls_loss_causal 0.4958 re_mapping 0.0026 re_causal 0.0083 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 300, time 247.04, cls_loss 0.0012 cls_loss_mapping 0.0012 cls_loss_causal 0.4689 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 301, time 247.00, cls_loss 0.0016 cls_loss_mapping 0.0018 cls_loss_causal 0.4784 re_mapping 0.0027 re_causal 0.0083 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 302, time 246.87, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4964 re_mapping 0.0026 re_causal 0.0082 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 303, time 246.97, cls_loss 0.0014 cls_loss_mapping 0.0016 cls_loss_causal 0.4547 re_mapping 0.0027 re_causal 0.0077 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 304, time 246.93, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4448 re_mapping 0.0027 re_causal 0.0081 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 305, time 247.00, cls_loss 0.0011 cls_loss_mapping 0.0009 cls_loss_causal 0.4617 re_mapping 0.0027 re_causal 0.0080 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 306, time 246.97, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4790 re_mapping 0.0027 re_causal 0.0085 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 307, time 247.08, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4900 re_mapping 0.0026 re_causal 0.0081 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 308, time 246.70, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4600 re_mapping 0.0026 re_causal 0.0078 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 309, time 246.77, cls_loss 0.0014 cls_loss_mapping 0.0018 cls_loss_causal 0.4756 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 310, time 247.15, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.4717 re_mapping 0.0028 re_causal 0.0082 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 311, time 247.54, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.4607 re_mapping 0.0027 re_causal 0.0078 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 312, time 247.41, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4517 re_mapping 0.0027 re_causal 0.0080 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 313, time 247.18, cls_loss 0.0010 cls_loss_mapping 0.0012 cls_loss_causal 0.4551 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 314, time 247.40, cls_loss 0.0012 cls_loss_mapping 0.0013 cls_loss_causal 0.4355 re_mapping 0.0028 re_causal 0.0079 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 315, time 247.58, cls_loss 0.0015 cls_loss_mapping 0.0020 cls_loss_causal 0.4555 re_mapping 0.0026 re_causal 0.0075 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 316, time 247.22, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4448 re_mapping 0.0026 re_causal 0.0075 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 317, time 247.06, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.4914 re_mapping 0.0027 re_causal 0.0078 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 318, time 246.94, cls_loss 0.0012 cls_loss_mapping 0.0016 cls_loss_causal 0.4779 re_mapping 0.0027 re_causal 0.0080 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 319, time 247.06, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4348 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 320, time 247.23, cls_loss 0.0015 cls_loss_mapping 0.0018 cls_loss_causal 0.4390 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 321, time 246.84, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4578 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 322, time 247.00, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.4698 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 323, time 247.23, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4447 re_mapping 0.0027 re_causal 0.0079 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 324, time 246.90, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4695 re_mapping 0.0028 re_causal 0.0077 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 325, time 246.85, cls_loss 0.0016 cls_loss_mapping 0.0019 cls_loss_causal 0.4536 re_mapping 0.0027 re_causal 0.0078 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 326, time 247.13, cls_loss 0.0018 cls_loss_mapping 0.0017 cls_loss_causal 0.4503 re_mapping 0.0026 re_causal 0.0073 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 327, time 246.92, cls_loss 0.0014 cls_loss_mapping 0.0014 cls_loss_causal 0.4610 re_mapping 0.0027 re_causal 0.0078 /// teacc 98.96 lr 0.00010000 +100 +0.0001 +changing lr +epoch 328, time 247.01, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.4952 re_mapping 0.0026 re_causal 0.0081 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 329, time 247.26, cls_loss 0.0012 cls_loss_mapping 0.0013 cls_loss_causal 0.4556 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 330, time 247.18, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4647 re_mapping 0.0024 re_causal 0.0073 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 331, time 247.30, cls_loss 0.0012 cls_loss_mapping 0.0017 cls_loss_causal 0.4686 re_mapping 0.0027 re_causal 0.0082 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 332, time 247.31, cls_loss 0.0012 cls_loss_mapping 0.0011 cls_loss_causal 0.4722 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 333, time 247.49, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4423 re_mapping 0.0026 re_causal 0.0077 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 334, time 247.27, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4602 re_mapping 0.0023 re_causal 0.0074 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 335, time 247.17, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4384 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 336, time 247.21, cls_loss 0.0021 cls_loss_mapping 0.0031 cls_loss_causal 0.4611 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.20 lr 0.00010000 +100 +0.0001 +changing lr +epoch 337, time 247.10, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4444 re_mapping 0.0025 re_causal 0.0077 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 338, time 247.26, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4533 re_mapping 0.0023 re_causal 0.0076 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 339, time 247.32, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.4566 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 340, time 247.16, cls_loss 0.0010 cls_loss_mapping 0.0010 cls_loss_causal 0.4598 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 341, time 247.15, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4526 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 342, time 247.51, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.5016 re_mapping 0.0026 re_causal 0.0084 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 343, time 247.38, cls_loss 0.0015 cls_loss_mapping 0.0015 cls_loss_causal 0.4960 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 344, time 247.23, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4559 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 345, time 247.50, cls_loss 0.0012 cls_loss_mapping 0.0015 cls_loss_causal 0.4610 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 346, time 247.38, cls_loss 0.0017 cls_loss_mapping 0.0023 cls_loss_causal 0.4869 re_mapping 0.0025 re_causal 0.0075 /// teacc 98.91 lr 0.00010000 +100 +0.0001 +changing lr +epoch 347, time 247.37, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4465 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 348, time 247.38, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4634 re_mapping 0.0026 re_causal 0.0077 /// teacc 98.95 lr 0.00010000 +100 +0.0001 +changing lr +epoch 349, time 247.41, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4596 re_mapping 0.0025 re_causal 0.0073 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 350, time 246.99, cls_loss 0.0013 cls_loss_mapping 0.0013 cls_loss_causal 0.4557 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 351, time 247.25, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4623 re_mapping 0.0025 re_causal 0.0075 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 352, time 247.18, cls_loss 0.0016 cls_loss_mapping 0.0016 cls_loss_causal 0.4614 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.15 lr 0.00010000 +100 +0.0001 +changing lr +epoch 353, time 246.88, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4559 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 354, time 247.59, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4315 re_mapping 0.0025 re_causal 0.0079 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 355, time 247.21, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4696 re_mapping 0.0025 re_causal 0.0078 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 356, time 247.56, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.4666 re_mapping 0.0025 re_causal 0.0077 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 357, time 247.29, cls_loss 0.0010 cls_loss_mapping 0.0014 cls_loss_causal 0.4475 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 358, time 247.22, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4710 re_mapping 0.0025 re_causal 0.0077 /// teacc 99.17 lr 0.00010000 +100 +0.0001 +changing lr +epoch 359, time 247.26, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4392 re_mapping 0.0024 re_causal 0.0073 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 360, time 247.41, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4256 re_mapping 0.0024 re_causal 0.0071 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 361, time 247.29, cls_loss 0.0011 cls_loss_mapping 0.0011 cls_loss_causal 0.4301 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 362, time 247.23, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4589 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 363, time 247.09, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4652 re_mapping 0.0026 re_causal 0.0077 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 364, time 246.92, cls_loss 0.0014 cls_loss_mapping 0.0011 cls_loss_causal 0.4869 re_mapping 0.0024 re_causal 0.0071 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 365, time 247.20, cls_loss 0.0020 cls_loss_mapping 0.0026 cls_loss_causal 0.4712 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 366, time 247.23, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4724 re_mapping 0.0025 re_causal 0.0078 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 367, time 247.13, cls_loss 0.0015 cls_loss_mapping 0.0021 cls_loss_causal 0.4755 re_mapping 0.0023 re_causal 0.0072 /// teacc 99.08 lr 0.00010000 +100 +0.0001 +changing lr +epoch 368, time 247.27, cls_loss 0.0020 cls_loss_mapping 0.0022 cls_loss_causal 0.4718 re_mapping 0.0024 re_causal 0.0073 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 369, time 247.12, cls_loss 0.0019 cls_loss_mapping 0.0024 cls_loss_causal 0.4716 re_mapping 0.0025 re_causal 0.0074 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 370, time 246.83, cls_loss 0.0014 cls_loss_mapping 0.0017 cls_loss_causal 0.4717 re_mapping 0.0026 re_causal 0.0074 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 371, time 246.80, cls_loss 0.0009 cls_loss_mapping 0.0011 cls_loss_causal 0.4637 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.11 lr 0.00010000 +100 +0.0001 +changing lr +epoch 372, time 246.84, cls_loss 0.0010 cls_loss_mapping 0.0013 cls_loss_causal 0.4744 re_mapping 0.0026 re_causal 0.0080 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 373, time 246.58, cls_loss 0.0011 cls_loss_mapping 0.0014 cls_loss_causal 0.4420 re_mapping 0.0024 re_causal 0.0071 /// teacc 99.03 lr 0.00010000 +100 +0.0001 +changing lr +epoch 374, time 246.67, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4422 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 375, time 246.70, cls_loss 0.0011 cls_loss_mapping 0.0012 cls_loss_causal 0.4266 re_mapping 0.0024 re_causal 0.0072 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 376, time 247.08, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4453 re_mapping 0.0023 re_causal 0.0071 /// teacc 98.97 lr 0.00010000 +100 +0.0001 +changing lr +epoch 377, time 247.12, cls_loss 0.0011 cls_loss_mapping 0.0016 cls_loss_causal 0.4735 re_mapping 0.0024 re_causal 0.0076 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 378, time 247.23, cls_loss 0.0013 cls_loss_mapping 0.0014 cls_loss_causal 0.4365 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 379, time 247.06, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4635 re_mapping 0.0023 re_causal 0.0070 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 380, time 247.45, cls_loss 0.0008 cls_loss_mapping 0.0011 cls_loss_causal 0.4711 re_mapping 0.0024 re_causal 0.0078 /// teacc 98.98 lr 0.00010000 +100 +0.0001 +changing lr +epoch 381, time 247.57, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4574 re_mapping 0.0022 re_causal 0.0069 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 382, time 247.44, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4559 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 383, time 247.33, cls_loss 0.0013 cls_loss_mapping 0.0031 cls_loss_causal 0.4570 re_mapping 0.0026 re_causal 0.0074 /// teacc 98.99 lr 0.00010000 +100 +0.0001 +changing lr +epoch 384, time 247.28, cls_loss 0.0020 cls_loss_mapping 0.0025 cls_loss_causal 0.4635 re_mapping 0.0024 re_causal 0.0074 /// teacc 98.94 lr 0.00010000 +100 +0.0001 +changing lr +epoch 385, time 247.44, cls_loss 0.0013 cls_loss_mapping 0.0017 cls_loss_causal 0.4457 re_mapping 0.0024 re_causal 0.0070 /// teacc 99.00 lr 0.00010000 +100 +0.0001 +changing lr +epoch 386, time 247.08, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4753 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.05 lr 0.00010000 +100 +0.0001 +changing lr +epoch 387, time 247.01, cls_loss 0.0013 cls_loss_mapping 0.0018 cls_loss_causal 0.4634 re_mapping 0.0025 re_causal 0.0076 /// teacc 99.02 lr 0.00010000 +100 +0.0001 +changing lr +epoch 388, time 247.20, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4897 re_mapping 0.0023 re_causal 0.0072 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 389, time 247.17, cls_loss 0.0012 cls_loss_mapping 0.0020 cls_loss_causal 0.4658 re_mapping 0.0022 re_causal 0.0068 /// teacc 99.04 lr 0.00010000 +100 +0.0001 +changing lr +epoch 390, time 247.07, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4615 re_mapping 0.0023 re_causal 0.0072 /// teacc 99.13 lr 0.00010000 +100 +0.0001 +changing lr +epoch 391, time 247.09, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.4522 re_mapping 0.0022 re_causal 0.0068 /// teacc 99.10 lr 0.00010000 +100 +0.0001 +changing lr +epoch 392, time 247.11, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4690 re_mapping 0.0023 re_causal 0.0074 /// teacc 99.09 lr 0.00010000 +100 +0.0001 +changing lr +epoch 393, time 246.80, cls_loss 0.0009 cls_loss_mapping 0.0013 cls_loss_causal 0.4457 re_mapping 0.0024 re_causal 0.0075 /// teacc 99.14 lr 0.00010000 +100 +0.0001 +changing lr +epoch 394, time 247.16, cls_loss 0.0011 cls_loss_mapping 0.0013 cls_loss_causal 0.4379 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.01 lr 0.00010000 +100 +0.0001 +changing lr +epoch 395, time 247.64, cls_loss 0.0012 cls_loss_mapping 0.0014 cls_loss_causal 0.4345 re_mapping 0.0024 re_causal 0.0074 /// teacc 99.12 lr 0.00010000 +100 +0.0001 +changing lr +epoch 396, time 247.37, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4401 re_mapping 0.0024 re_causal 0.0072 /// teacc 99.07 lr 0.00010000 +100 +0.0001 +changing lr +epoch 397, time 247.54, cls_loss 0.0011 cls_loss_mapping 0.0015 cls_loss_causal 0.4345 re_mapping 0.0022 re_causal 0.0070 /// teacc 99.06 lr 0.00010000 +100 +0.0001 +changing lr +epoch 398, time 247.34, cls_loss 0.0010 cls_loss_mapping 0.0011 cls_loss_causal 0.4310 re_mapping 0.0023 re_causal 0.0071 /// teacc 99.16 lr 0.00010000 +100 +0.0001 +changing lr +epoch 399, time 246.86, cls_loss 0.0008 cls_loss_mapping 0.0010 cls_loss_causal 0.4484 re_mapping 0.0023 re_causal 0.0074 /// teacc 99.11 lr 0.00001000 +100 +1e-05 +changing lr +epoch 400, time 247.11, cls_loss 0.0010 cls_loss_mapping 0.0015 cls_loss_causal 0.4272 re_mapping 0.0023 re_causal 0.0071 /// teacc 99.14 lr 0.00001000 +100 +1e-05 +changing lr +epoch 401, time 247.02, cls_loss 0.0009 cls_loss_mapping 0.0009 cls_loss_causal 0.4241 re_mapping 0.0022 re_causal 0.0068 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 402, time 247.20, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4582 re_mapping 0.0020 re_causal 0.0069 /// teacc 99.17 lr 0.00001000 +100 +1e-05 +changing lr +epoch 403, time 247.43, cls_loss 0.0009 cls_loss_mapping 0.0007 cls_loss_causal 0.4785 re_mapping 0.0020 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 404, time 246.87, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4336 re_mapping 0.0020 re_causal 0.0066 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 405, time 247.03, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4119 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.18 lr 0.00001000 +100 +1e-05 +changing lr +epoch 406, time 247.20, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4492 re_mapping 0.0020 re_causal 0.0067 /// teacc 99.19 lr 0.00001000 +100 +1e-05 +changing lr +epoch 407, time 247.14, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4428 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 408, time 246.95, cls_loss 0.0008 cls_loss_mapping 0.0006 cls_loss_causal 0.4292 re_mapping 0.0019 re_causal 0.0065 /// teacc 99.19 lr 0.00001000 +100 +1e-05 +changing lr +epoch 409, time 247.07, cls_loss 0.0009 cls_loss_mapping 0.0006 cls_loss_causal 0.4267 re_mapping 0.0019 re_causal 0.0063 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 410, time 246.79, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4532 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 411, time 247.28, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4191 re_mapping 0.0018 re_causal 0.0065 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 412, time 247.11, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4227 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 413, time 247.24, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4239 re_mapping 0.0019 re_causal 0.0067 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 414, time 246.99, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4236 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 415, time 247.09, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4173 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 416, time 246.96, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4130 re_mapping 0.0019 re_causal 0.0066 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 417, time 247.21, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4219 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 418, time 247.16, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4107 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 419, time 247.16, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4217 re_mapping 0.0018 re_causal 0.0064 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 420, time 247.03, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4252 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 421, time 246.83, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4170 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 422, time 247.05, cls_loss 0.0007 cls_loss_mapping 0.0005 cls_loss_causal 0.4611 re_mapping 0.0017 re_causal 0.0067 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 423, time 247.12, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4178 re_mapping 0.0017 re_causal 0.0064 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 424, time 246.46, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4336 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 425, time 246.57, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4000 re_mapping 0.0017 re_causal 0.0060 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 426, time 246.61, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4221 re_mapping 0.0017 re_causal 0.0062 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 427, time 246.74, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4261 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 428, time 247.08, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4231 re_mapping 0.0017 re_causal 0.0062 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 429, time 247.14, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4139 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 430, time 247.05, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4485 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 431, time 246.93, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.3967 re_mapping 0.0016 re_causal 0.0059 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 432, time 246.91, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4022 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 433, time 246.92, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4053 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 434, time 246.75, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4163 re_mapping 0.0017 re_causal 0.0063 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 435, time 247.10, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4253 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 436, time 247.14, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4303 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 437, time 246.97, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4144 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 438, time 247.00, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4147 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 439, time 247.17, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4282 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 440, time 246.86, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.3817 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 441, time 246.73, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4127 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 442, time 247.03, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4147 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 443, time 247.06, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4173 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 444, time 246.70, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4060 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 445, time 246.69, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3875 re_mapping 0.0016 re_causal 0.0061 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +---------------------saving model at epoch 446---------------------------------------------------- +epoch 446, time 247.46, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4331 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 447, time 246.73, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4148 re_mapping 0.0016 re_causal 0.0063 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 448, time 247.05, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4421 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 449, time 246.60, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3929 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +---------------------saving model at epoch 450---------------------------------------------------- +epoch 450, time 247.44, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4173 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.31 lr 0.00001000 +100 +1e-05 +changing lr +epoch 451, time 246.85, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4352 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 452, time 246.56, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4248 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 453, time 246.71, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4208 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 454, time 247.02, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4349 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 455, time 246.71, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4133 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 456, time 246.84, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4097 re_mapping 0.0015 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 457, time 246.87, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3905 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 458, time 246.86, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.3884 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 459, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3983 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 460, time 246.95, cls_loss 0.0006 cls_loss_mapping 0.0004 cls_loss_causal 0.4410 re_mapping 0.0016 re_causal 0.0062 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 461, time 246.76, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4042 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 462, time 247.25, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4222 re_mapping 0.0016 re_causal 0.0060 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 463, time 247.13, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4288 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 464, time 247.10, cls_loss 0.0010 cls_loss_mapping 0.0005 cls_loss_causal 0.4350 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 465, time 247.08, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4409 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.29 lr 0.00001000 +100 +1e-05 +changing lr +epoch 466, time 246.90, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.3901 re_mapping 0.0015 re_causal 0.0056 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 467, time 247.04, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4076 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 468, time 246.84, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.3978 re_mapping 0.0015 re_causal 0.0056 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 469, time 246.98, cls_loss 0.0008 cls_loss_mapping 0.0005 cls_loss_causal 0.4259 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 470, time 246.99, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4459 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 471, time 247.05, cls_loss 0.0009 cls_loss_mapping 0.0005 cls_loss_causal 0.4229 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 472, time 246.98, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4319 re_mapping 0.0014 re_causal 0.0058 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 473, time 246.98, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4292 re_mapping 0.0014 re_causal 0.0058 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 474, time 247.17, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4197 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 475, time 247.03, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.3734 re_mapping 0.0015 re_causal 0.0057 /// teacc 99.29 lr 0.00001000 +100 +1e-05 +changing lr +epoch 476, time 247.30, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3885 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 477, time 248.54, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4097 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 478, time 247.02, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4113 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.23 lr 0.00001000 +100 +1e-05 +changing lr +epoch 479, time 247.12, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.3727 re_mapping 0.0014 re_causal 0.0054 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 480, time 246.85, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4400 re_mapping 0.0015 re_causal 0.0060 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 481, time 246.83, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4142 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 482, time 246.97, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4060 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.24 lr 0.00001000 +100 +1e-05 +changing lr +epoch 483, time 246.78, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4004 re_mapping 0.0014 re_causal 0.0055 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 484, time 246.96, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4422 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.28 lr 0.00001000 +100 +1e-05 +changing lr +epoch 485, time 246.70, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3912 re_mapping 0.0014 re_causal 0.0055 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 486, time 247.02, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.3976 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 487, time 246.77, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4313 re_mapping 0.0015 re_causal 0.0060 /// teacc 99.26 lr 0.00001000 +100 +1e-05 +changing lr +epoch 488, time 246.95, cls_loss 0.0007 cls_loss_mapping 0.0003 cls_loss_causal 0.4221 re_mapping 0.0015 re_causal 0.0059 /// teacc 99.27 lr 0.00001000 +100 +1e-05 +changing lr +epoch 489, time 246.76, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4082 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.25 lr 0.00001000 +100 +1e-05 +changing lr +epoch 490, time 246.83, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4112 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 491, time 246.79, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3886 re_mapping 0.0015 re_causal 0.0055 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 492, time 246.78, cls_loss 0.0006 cls_loss_mapping 0.0003 cls_loss_causal 0.4168 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 493, time 246.88, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4221 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.20 lr 0.00001000 +100 +1e-05 +changing lr +epoch 494, time 246.71, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4312 re_mapping 0.0015 re_causal 0.0058 /// teacc 99.19 lr 0.00001000 +100 +1e-05 +changing lr +epoch 495, time 246.90, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.4413 re_mapping 0.0014 re_causal 0.0059 /// teacc 99.21 lr 0.00001000 +100 +1e-05 +changing lr +epoch 496, time 246.63, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4036 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.22 lr 0.00001000 +100 +1e-05 +changing lr +epoch 497, time 246.91, cls_loss 0.0009 cls_loss_mapping 0.0004 cls_loss_causal 0.4220 re_mapping 0.0014 re_causal 0.0057 /// teacc 99.18 lr 0.00001000 +100 +1e-05 +changing lr +epoch 498, time 247.19, cls_loss 0.0008 cls_loss_mapping 0.0004 cls_loss_causal 0.4161 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.17 lr 0.00001000 +100 +1e-05 +changing lr +epoch 499, time 247.07, cls_loss 0.0007 cls_loss_mapping 0.0004 cls_loss_causal 0.3912 re_mapping 0.0014 re_causal 0.0056 /// teacc 99.19 lr 0.00001000 +---------------------saving last model at epoch 499---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_ReProduceMetaCausal', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_ReProduceMetaCausal/14factor_best.csv', 'channels': 3, 'factor_num': 14, 'stride': 3, 'epoch': 'best', 'eval_mapping': True} +loading weight of best +randm: False +stride: 3 +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +loading weight of best +Using downloaded and verified file: /home/yuqian_fu/.pytorch/SVHN/test_32x32.mat diff --git a/Meta-causal/code/56720.error b/Meta-causal/code/56720.error new file mode 100644 index 0000000000000000000000000000000000000000..4c741962fda5fd145618ae7373555295b05ff9de --- /dev/null +++ b/Meta-causal/code/56720.error @@ -0,0 +1 @@ +run_my_joint_v13_test.sh: line 28: andn}: command not found diff --git a/Meta-causal/code/56720.log b/Meta-causal/code/56720.log new file mode 100644 index 0000000000000000000000000000000000000000..5e281586589ea1303704167b5264e28ec7b696f5 --- /dev/null +++ b/Meta-causal/code/56720.log @@ -0,0 +1,336 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'cartoon', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_train.hdf5 torch.Size([2107, 3, 227, 227]) torch.Size([2107]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_val.hdf5 torch.Size([237, 3, 227, 227]) torch.Size([237]) +-------------------------------------loading pretrain weights---------------------------------- +351 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 500.68, cls_loss 5.0126 cls_loss_mapping 1.4019 cls_loss_causal 1.7210 re_mapping 1.0578 re_causal 1.0584 /// teacc 83.12 lr 0.00999497 +351 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 530.56, cls_loss 2.0306 cls_loss_mapping 0.7946 cls_loss_causal 1.3288 re_mapping 0.6527 re_causal 0.6538 /// teacc 87.76 lr 0.00997987 +351 +0.009979871469976196 +changing lr +---------------------saving model at epoch 2---------------------------------------------------- +epoch 2, time 536.46, cls_loss 0.6382 cls_loss_mapping 0.4834 cls_loss_causal 1.1278 re_mapping 0.3952 re_causal 0.3957 /// teacc 91.98 lr 0.00995475 +351 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 514.73, cls_loss 0.2115 cls_loss_mapping 0.3043 cls_loss_causal 0.9479 re_mapping 0.2605 re_causal 0.2608 /// teacc 92.41 lr 0.00991965 +351 +0.009919647942993149 +changing lr +epoch 4, time 518.58, cls_loss 0.1048 cls_loss_mapping 0.2504 cls_loss_causal 0.8913 re_mapping 0.2075 re_causal 0.2080 /// teacc 92.41 lr 0.00987464 +351 +0.009874639560909117 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 522.88, cls_loss 0.0517 cls_loss_mapping 0.2038 cls_loss_causal 0.8571 re_mapping 0.1746 re_causal 0.1753 /// teacc 95.36 lr 0.00981981 +351 +0.009819814303479266 +changing lr +epoch 6, time 515.51, cls_loss 0.0244 cls_loss_mapping 0.1830 cls_loss_causal 0.7905 re_mapping 0.1502 re_causal 0.1512 /// teacc 94.51 lr 0.00975528 +351 +0.009755282581475767 +changing lr +epoch 7, time 516.68, cls_loss 0.0226 cls_loss_mapping 0.1536 cls_loss_causal 0.7386 re_mapping 0.1335 re_causal 0.1347 /// teacc 94.94 lr 0.00968117 +351 +0.009681174353198686 +changing lr +epoch 8, time 512.83, cls_loss 0.0311 cls_loss_mapping 0.1488 cls_loss_causal 0.7284 re_mapping 0.1200 re_causal 0.1218 /// teacc 91.56 lr 0.00959764 +351 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 515.75, cls_loss 0.0257 cls_loss_mapping 0.1258 cls_loss_causal 0.7038 re_mapping 0.1090 re_causal 0.1110 /// teacc 95.78 lr 0.00950484 +351 +0.009504844339512096 +changing lr +epoch 10, time 508.97, cls_loss 0.0086 cls_loss_mapping 0.1049 cls_loss_causal 0.7078 re_mapping 0.0973 re_causal 0.0997 /// teacc 94.09 lr 0.00940298 +351 +0.009402977659283692 +changing lr +epoch 11, time 522.72, cls_loss 0.0121 cls_loss_mapping 0.1017 cls_loss_causal 0.6880 re_mapping 0.0899 re_causal 0.0929 /// teacc 95.36 lr 0.00929224 +351 +0.009292243968009333 +changing lr +epoch 12, time 511.13, cls_loss 0.0138 cls_loss_mapping 0.0946 cls_loss_causal 0.7011 re_mapping 0.0820 re_causal 0.0855 /// teacc 94.94 lr 0.00917287 +351 +0.009172866268606516 +changing lr +epoch 13, time 522.62, cls_loss 0.0104 cls_loss_mapping 0.0844 cls_loss_causal 0.6675 re_mapping 0.0747 re_causal 0.0784 /// teacc 94.51 lr 0.00904508 +351 +0.00904508497187474 +changing lr +epoch 14, time 530.18, cls_loss 0.0122 cls_loss_mapping 0.0736 cls_loss_causal 0.6363 re_mapping 0.0698 re_causal 0.0745 /// teacc 95.78 lr 0.00890916 +351 +0.008909157412340152 +changing lr +epoch 15, time 513.03, cls_loss 0.0108 cls_loss_mapping 0.0735 cls_loss_causal 0.6055 re_mapping 0.0623 re_causal 0.0673 /// teacc 94.94 lr 0.00876536 +351 +0.00876535733001806 +changing lr +---------------------saving model at epoch 16---------------------------------------------------- +epoch 16, time 515.62, cls_loss 0.0097 cls_loss_mapping 0.0626 cls_loss_causal 0.6328 re_mapping 0.0572 re_causal 0.0629 /// teacc 97.05 lr 0.00861397 +351 +0.008613974319136962 +changing lr +epoch 17, time 534.26, cls_loss 0.0145 cls_loss_mapping 0.0706 cls_loss_causal 0.6484 re_mapping 0.0533 re_causal 0.0603 /// teacc 96.20 lr 0.00845531 +351 +0.008455313244934327 +changing lr +epoch 18, time 532.03, cls_loss 0.0106 cls_loss_mapping 0.0571 cls_loss_causal 0.5705 re_mapping 0.0492 re_causal 0.0567 /// teacc 96.62 lr 0.00828969 +351 +0.008289693629698565 +changing lr +epoch 19, time 518.07, cls_loss 0.0076 cls_loss_mapping 0.0474 cls_loss_causal 0.5525 re_mapping 0.0441 re_causal 0.0513 /// teacc 95.78 lr 0.00811745 +351 +0.00811744900929367 +changing lr +epoch 20, time 530.26, cls_loss 0.0081 cls_loss_mapping 0.0546 cls_loss_causal 0.5926 re_mapping 0.0409 re_causal 0.0491 /// teacc 97.05 lr 0.00793893 +351 +0.007938926261462368 +changing lr +epoch 21, time 534.00, cls_loss 0.0104 cls_loss_mapping 0.0511 cls_loss_causal 0.5469 re_mapping 0.0373 re_causal 0.0451 /// teacc 95.78 lr 0.00775448 +351 +0.007754484907260515 +changing lr +epoch 22, time 534.87, cls_loss 0.0148 cls_loss_mapping 0.0474 cls_loss_causal 0.5694 re_mapping 0.0353 re_causal 0.0430 /// teacc 95.36 lr 0.00756450 +351 +0.007564496387029534 +changing lr +epoch 23, time 515.04, cls_loss 0.0053 cls_loss_mapping 0.0395 cls_loss_causal 0.5557 re_mapping 0.0324 re_causal 0.0409 /// teacc 96.62 lr 0.00736934 +351 +0.007369343312364995 +changing lr +epoch 24, time 527.73, cls_loss 0.0083 cls_loss_mapping 0.0487 cls_loss_causal 0.5594 re_mapping 0.0306 re_causal 0.0402 /// teacc 94.94 lr 0.00716942 +351 +0.0071694186955877925 +changing lr +epoch 25, time 521.10, cls_loss 0.0080 cls_loss_mapping 0.0392 cls_loss_causal 0.5600 re_mapping 0.0291 re_causal 0.0390 /// teacc 97.05 lr 0.00696513 +351 +0.0069651251582696205 +changing lr +epoch 26, time 528.30, cls_loss 0.0054 cls_loss_mapping 0.0316 cls_loss_causal 0.5380 re_mapping 0.0270 re_causal 0.0366 /// teacc 96.20 lr 0.00675687 +351 +0.006756874120406716 +changing lr +epoch 27, time 526.16, cls_loss 0.0075 cls_loss_mapping 0.0354 cls_loss_causal 0.5384 re_mapping 0.0251 re_causal 0.0347 /// teacc 97.05 lr 0.00654508 +351 +0.00654508497187474 +changing lr +epoch 28, time 520.46, cls_loss 0.0066 cls_loss_mapping 0.0281 cls_loss_causal 0.5043 re_mapping 0.0240 re_causal 0.0354 /// teacc 96.62 lr 0.00633018 +351 +0.006330184227833378 +changing lr +epoch 29, time 536.80, cls_loss 0.0074 cls_loss_mapping 0.0305 cls_loss_causal 0.5296 re_mapping 0.0227 re_causal 0.0341 /// teacc 95.78 lr 0.00611260 +351 +0.006112604669781575 +changing lr +epoch 30, time 530.96, cls_loss 0.0062 cls_loss_mapping 0.0301 cls_loss_causal 0.5251 re_mapping 0.0214 re_causal 0.0317 /// teacc 96.20 lr 0.00589278 +351 +0.005892784473993186 +changing lr +epoch 31, time 528.74, cls_loss 0.0051 cls_loss_mapping 0.0263 cls_loss_causal 0.5350 re_mapping 0.0205 re_causal 0.0317 /// teacc 96.20 lr 0.00567117 +351 +0.00567116632908828 +changing lr +---------------------saving model at epoch 32---------------------------------------------------- +epoch 32, time 517.42, cls_loss 0.0051 cls_loss_mapping 0.0225 cls_loss_causal 0.5060 re_mapping 0.0197 re_causal 0.0305 /// teacc 97.47 lr 0.00544820 +351 +0.00544819654451717 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 532.29, cls_loss 0.0050 cls_loss_mapping 0.0196 cls_loss_causal 0.5099 re_mapping 0.0185 re_causal 0.0291 /// teacc 97.89 lr 0.00522432 +351 +0.005224324151752577 +changing lr +epoch 34, time 521.23, cls_loss 0.0079 cls_loss_mapping 0.0235 cls_loss_causal 0.5058 re_mapping 0.0177 re_causal 0.0285 /// teacc 97.89 lr 0.00500000 +351 +0.005000000000000003 +changing lr +epoch 35, time 521.42, cls_loss 0.0054 cls_loss_mapping 0.0236 cls_loss_causal 0.4683 re_mapping 0.0178 re_causal 0.0281 /// teacc 97.05 lr 0.00477568 +351 +0.004775675848247429 +changing lr +epoch 36, time 526.29, cls_loss 0.0057 cls_loss_mapping 0.0231 cls_loss_causal 0.5159 re_mapping 0.0172 re_causal 0.0278 /// teacc 97.05 lr 0.00455180 +351 +0.004551803455482836 +changing lr +epoch 37, time 535.59, cls_loss 0.0063 cls_loss_mapping 0.0199 cls_loss_causal 0.4658 re_mapping 0.0163 re_causal 0.0267 /// teacc 97.47 lr 0.00432883 +351 +0.004328833670911726 +changing lr +epoch 38, time 512.58, cls_loss 0.0045 cls_loss_mapping 0.0199 cls_loss_causal 0.4925 re_mapping 0.0155 re_causal 0.0258 /// teacc 97.05 lr 0.00410722 +351 +0.0041072155260068206 +changing lr +epoch 39, time 532.69, cls_loss 0.0056 cls_loss_mapping 0.0220 cls_loss_causal 0.4772 re_mapping 0.0150 re_causal 0.0253 /// teacc 97.47 lr 0.00388740 +351 +0.0038873953302184317 +changing lr +epoch 40, time 536.18, cls_loss 0.0044 cls_loss_mapping 0.0185 cls_loss_causal 0.4992 re_mapping 0.0146 re_causal 0.0241 /// teacc 97.47 lr 0.00366982 +351 +0.003669815772166629 +changing lr +epoch 41, time 531.87, cls_loss 0.0044 cls_loss_mapping 0.0147 cls_loss_causal 0.4840 re_mapping 0.0144 re_causal 0.0246 /// teacc 97.89 lr 0.00345492 +351 +0.0034549150281252667 +changing lr +---------------------saving model at epoch 42---------------------------------------------------- +epoch 42, time 509.65, cls_loss 0.0045 cls_loss_mapping 0.0164 cls_loss_causal 0.4600 re_mapping 0.0136 re_causal 0.0224 /// teacc 98.31 lr 0.00324313 +351 +0.0032431258795932905 +changing lr +epoch 43, time 520.56, cls_loss 0.0051 cls_loss_mapping 0.0169 cls_loss_causal 0.5021 re_mapping 0.0137 re_causal 0.0235 /// teacc 97.47 lr 0.00303487 +351 +0.0030348748417303863 +changing lr +---------------------saving model at epoch 44---------------------------------------------------- +epoch 44, time 532.35, cls_loss 0.0042 cls_loss_mapping 0.0153 cls_loss_causal 0.4512 re_mapping 0.0131 re_causal 0.0230 /// teacc 98.73 lr 0.00283058 +351 +0.0028305813044122124 +changing lr +epoch 45, time 523.83, cls_loss 0.0053 cls_loss_mapping 0.0159 cls_loss_causal 0.4523 re_mapping 0.0130 re_causal 0.0219 /// teacc 97.89 lr 0.00263066 +351 +0.0026306566876350096 +changing lr +epoch 46, time 536.05, cls_loss 0.0050 cls_loss_mapping 0.0148 cls_loss_causal 0.4521 re_mapping 0.0125 re_causal 0.0215 /// teacc 96.62 lr 0.00243550 +351 +0.0024355036129704724 +changing lr +epoch 47, time 509.13, cls_loss 0.0043 cls_loss_mapping 0.0159 cls_loss_causal 0.4864 re_mapping 0.0121 re_causal 0.0214 /// teacc 97.89 lr 0.00224552 +351 +0.00224551509273949 +changing lr +epoch 48, time 524.58, cls_loss 0.0037 cls_loss_mapping 0.0109 cls_loss_causal 0.4474 re_mapping 0.0120 re_causal 0.0208 /// teacc 98.31 lr 0.00206107 +351 +0.002061073738537637 +changing lr +epoch 49, time 517.27, cls_loss 0.0033 cls_loss_mapping 0.0125 cls_loss_causal 0.4527 re_mapping 0.0117 re_causal 0.0205 /// teacc 97.89 lr 0.00188255 +351 +0.0018825509907063344 +changing lr +epoch 50, time 516.76, cls_loss 0.0039 cls_loss_mapping 0.0142 cls_loss_causal 0.4602 re_mapping 0.0116 re_causal 0.0204 /// teacc 97.47 lr 0.00171031 +351 +0.0017103063703014388 +changing lr +epoch 51, time 513.81, cls_loss 0.0025 cls_loss_mapping 0.0098 cls_loss_causal 0.4081 re_mapping 0.0116 re_causal 0.0197 /// teacc 98.31 lr 0.00154469 +351 +0.0015446867550656784 +changing lr +epoch 52, time 514.01, cls_loss 0.0042 cls_loss_mapping 0.0125 cls_loss_causal 0.4603 re_mapping 0.0114 re_causal 0.0195 /// teacc 97.89 lr 0.00138603 +351 +0.001386025680863044 +changing lr +epoch 53, time 524.35, cls_loss 0.0051 cls_loss_mapping 0.0127 cls_loss_causal 0.4572 re_mapping 0.0111 re_causal 0.0193 /// teacc 97.89 lr 0.00123464 +351 +0.0012346426699819469 +changing lr +epoch 54, time 514.44, cls_loss 0.0044 cls_loss_mapping 0.0127 cls_loss_causal 0.4353 re_mapping 0.0111 re_causal 0.0187 /// teacc 97.47 lr 0.00109084 +351 +0.0010908425876598518 +changing lr +epoch 55, time 522.77, cls_loss 0.0037 cls_loss_mapping 0.0112 cls_loss_causal 0.4375 re_mapping 0.0109 re_causal 0.0188 /// teacc 98.31 lr 0.00095492 +351 +0.000954915028125264 +changing lr +epoch 56, time 523.02, cls_loss 0.0041 cls_loss_mapping 0.0109 cls_loss_causal 0.4403 re_mapping 0.0108 re_causal 0.0186 /// teacc 97.05 lr 0.00082713 +351 +0.0008271337313934874 +changing lr +epoch 57, time 527.11, cls_loss 0.0028 cls_loss_mapping 0.0091 cls_loss_causal 0.4157 re_mapping 0.0108 re_causal 0.0176 /// teacc 97.47 lr 0.00070776 +351 +0.00070775603199067 +changing lr +epoch 58, time 504.49, cls_loss 0.0031 cls_loss_mapping 0.0086 cls_loss_causal 0.4095 re_mapping 0.0108 re_causal 0.0171 /// teacc 97.89 lr 0.00059702 +351 +0.0005970223407163104 +changing lr +epoch 59, time 497.53, cls_loss 0.0053 cls_loss_mapping 0.0115 cls_loss_causal 0.4429 re_mapping 0.0105 re_causal 0.0172 /// teacc 97.05 lr 0.00049516 +351 +0.0004951556604879052 +changing lr +epoch 60, time 507.85, cls_loss 0.0043 cls_loss_mapping 0.0108 cls_loss_causal 0.4240 re_mapping 0.0103 re_causal 0.0166 /// teacc 98.31 lr 0.00040236 +351 +0.00040236113724274745 +changing lr +epoch 61, time 489.10, cls_loss 0.0040 cls_loss_mapping 0.0104 cls_loss_causal 0.4613 re_mapping 0.0103 re_causal 0.0175 /// teacc 97.05 lr 0.00031883 +351 +0.00031882564680131423 +changing lr +epoch 62, time 487.44, cls_loss 0.0040 cls_loss_mapping 0.0101 cls_loss_causal 0.4445 re_mapping 0.0102 re_causal 0.0167 /// teacc 98.31 lr 0.00024472 +351 +0.0002447174185242325 +changing lr +epoch 63, time 492.60, cls_loss 0.0030 cls_loss_mapping 0.0067 cls_loss_causal 0.3786 re_mapping 0.0102 re_causal 0.0165 /// teacc 97.89 lr 0.00018019 +351 +0.0001801856965207339 +changing lr +epoch 64, time 493.59, cls_loss 0.0040 cls_loss_mapping 0.0106 cls_loss_causal 0.4459 re_mapping 0.0101 re_causal 0.0165 /// teacc 96.62 lr 0.00012536 +351 +0.000125360439090882 +changing lr +epoch 65, time 485.22, cls_loss 0.0051 cls_loss_mapping 0.0094 cls_loss_causal 0.4355 re_mapping 0.0101 re_causal 0.0162 /// teacc 97.47 lr 0.00008035 +351 +8.03520570068517e-05 +changing lr +epoch 66, time 475.77, cls_loss 0.0036 cls_loss_mapping 0.0086 cls_loss_causal 0.4274 re_mapping 0.0101 re_causal 0.0165 /// teacc 97.89 lr 0.00004525 +351 +4.5251191160326525e-05 +changing lr +epoch 67, time 483.43, cls_loss 0.0043 cls_loss_mapping 0.0107 cls_loss_causal 0.4531 re_mapping 0.0102 re_causal 0.0168 /// teacc 96.62 lr 0.00002013 +351 +2.0128530023804673e-05 +changing lr +epoch 68, time 484.93, cls_loss 0.0030 cls_loss_mapping 0.0073 cls_loss_causal 0.4376 re_mapping 0.0102 re_causal 0.0166 /// teacc 97.89 lr 0.00000503 +351 +5.034667293427056e-06 +changing lr +epoch 69, time 479.93, cls_loss 0.0041 cls_loss_mapping 0.0089 cls_loss_causal 0.4412 re_mapping 0.0100 re_causal 0.0165 /// teacc 96.20 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'cartoon', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//cartoon/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/cartoon_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['cartoon', 'art_painting', 'photo', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + cartoon art_painting photo sketch Avg +w/o do (original x) 99.616041 76.806641 89.700599 72.613897 79.707045 + cartoon art_painting photo sketch Avg +do 99.573379 75.537109 89.760479 73.631967 79.643185 diff --git a/Meta-causal/code/56721.error b/Meta-causal/code/56721.error new file mode 100644 index 0000000000000000000000000000000000000000..fd313270a3ba847b383c7eb4ae546600fd872b6a --- /dev/null +++ b/Meta-causal/code/56721.error @@ -0,0 +1 @@ +run_my_joint_v13_test.sh: line 29: de: command not found diff --git a/Meta-causal/code/56721.log b/Meta-causal/code/56721.log new file mode 100644 index 0000000000000000000000000000000000000000..e0e3660d35bfb4347dcbaf5d7e601b60419518bc --- /dev/null +++ b/Meta-causal/code/56721.log @@ -0,0 +1,329 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'photo', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_train.hdf5 torch.Size([1499, 3, 227, 227]) torch.Size([1499]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_val.hdf5 torch.Size([171, 3, 227, 227]) torch.Size([171]) +-------------------------------------loading pretrain weights---------------------------------- +249 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 330.26, cls_loss 2.1901 cls_loss_mapping 1.0966 cls_loss_causal 1.5411 re_mapping 1.2660 re_causal 1.2689 /// teacc 95.32 lr 0.00999497 +249 +0.009994965332706574 +changing lr +epoch 1, time 329.57, cls_loss 1.1155 cls_loss_mapping 0.7787 cls_loss_causal 1.4901 re_mapping 0.9558 re_causal 0.9646 /// teacc 93.57 lr 0.00997987 +249 +0.009979871469976196 +changing lr +epoch 2, time 330.39, cls_loss 0.9288 cls_loss_mapping 0.7032 cls_loss_causal 1.4747 re_mapping 0.7723 re_causal 0.7847 /// teacc 75.44 lr 0.00995475 +249 +0.009954748808839675 +changing lr +epoch 3, time 328.90, cls_loss 1.2627 cls_loss_mapping 0.6321 cls_loss_causal 1.5390 re_mapping 0.6502 re_causal 0.6690 /// teacc 85.96 lr 0.00991965 +249 +0.009919647942993149 +changing lr +epoch 4, time 327.35, cls_loss 0.9500 cls_loss_mapping 0.7241 cls_loss_causal 1.5912 re_mapping 0.6164 re_causal 0.6345 /// teacc 93.57 lr 0.00987464 +249 +0.009874639560909117 +changing lr +epoch 5, time 326.62, cls_loss 1.4824 cls_loss_mapping 0.9000 cls_loss_causal 1.6461 re_mapping 0.5127 re_causal 0.5278 /// teacc 79.53 lr 0.00981981 +249 +0.009819814303479266 +changing lr +epoch 6, time 329.86, cls_loss 0.5391 cls_loss_mapping 0.6994 cls_loss_causal 1.5297 re_mapping 0.4445 re_causal 0.4580 /// teacc 91.23 lr 0.00975528 +249 +0.009755282581475767 +changing lr +epoch 7, time 326.96, cls_loss 0.4282 cls_loss_mapping 0.7985 cls_loss_causal 1.5762 re_mapping 0.4031 re_causal 0.4239 /// teacc 90.06 lr 0.00968117 +249 +0.009681174353198686 +changing lr +epoch 8, time 324.18, cls_loss 0.4582 cls_loss_mapping 0.6437 cls_loss_causal 1.4929 re_mapping 0.3704 re_causal 0.4016 /// teacc 92.40 lr 0.00959764 +249 +0.009597638862757255 +changing lr +epoch 9, time 329.22, cls_loss 0.4196 cls_loss_mapping 0.6434 cls_loss_causal 1.5206 re_mapping 0.3366 re_causal 0.3732 /// teacc 91.23 lr 0.00950484 +249 +0.009504844339512096 +changing lr +epoch 10, time 326.22, cls_loss 0.6899 cls_loss_mapping 0.6931 cls_loss_causal 1.4531 re_mapping 0.3138 re_causal 0.3465 /// teacc 91.81 lr 0.00940298 +249 +0.009402977659283692 +changing lr +epoch 11, time 332.53, cls_loss 0.2100 cls_loss_mapping 0.5305 cls_loss_causal 1.2812 re_mapping 0.2652 re_causal 0.3015 /// teacc 94.15 lr 0.00929224 +249 +0.009292243968009333 +changing lr +epoch 12, time 328.24, cls_loss 1.9157 cls_loss_mapping 1.1807 cls_loss_causal 1.8542 re_mapping 0.2875 re_causal 0.3153 /// teacc 81.29 lr 0.00917287 +249 +0.009172866268606516 +changing lr +epoch 13, time 330.89, cls_loss 0.5559 cls_loss_mapping 0.9412 cls_loss_causal 1.5863 re_mapping 0.2804 re_causal 0.3010 /// teacc 88.30 lr 0.00904508 +249 +0.00904508497187474 +changing lr +epoch 14, time 327.08, cls_loss 0.2945 cls_loss_mapping 0.7027 cls_loss_causal 1.4399 re_mapping 0.2493 re_causal 0.2637 /// teacc 89.47 lr 0.00890916 +249 +0.008909157412340152 +changing lr +epoch 15, time 327.50, cls_loss 0.1556 cls_loss_mapping 0.5735 cls_loss_causal 1.3348 re_mapping 0.2367 re_causal 0.2499 /// teacc 90.64 lr 0.00876536 +249 +0.00876535733001806 +changing lr +epoch 16, time 325.93, cls_loss 0.5865 cls_loss_mapping 0.6469 cls_loss_causal 1.4535 re_mapping 0.2249 re_causal 0.2442 /// teacc 83.63 lr 0.00861397 +249 +0.008613974319136962 +changing lr +epoch 17, time 325.67, cls_loss 0.2541 cls_loss_mapping 0.5530 cls_loss_causal 1.3152 re_mapping 0.1981 re_causal 0.2108 /// teacc 90.64 lr 0.00845531 +249 +0.008455313244934327 +changing lr +epoch 18, time 328.07, cls_loss 0.1021 cls_loss_mapping 0.4746 cls_loss_causal 1.2840 re_mapping 0.1724 re_causal 0.1940 /// teacc 91.81 lr 0.00828969 +249 +0.008289693629698565 +changing lr +epoch 19, time 327.66, cls_loss 0.2583 cls_loss_mapping 0.4658 cls_loss_causal 1.3477 re_mapping 0.1511 re_causal 0.1725 /// teacc 85.38 lr 0.00811745 +249 +0.00811744900929367 +changing lr +epoch 20, time 330.32, cls_loss 0.2436 cls_loss_mapping 0.4640 cls_loss_causal 1.2885 re_mapping 0.1358 re_causal 0.1612 /// teacc 89.47 lr 0.00793893 +249 +0.007938926261462368 +changing lr +epoch 21, time 327.13, cls_loss 0.0809 cls_loss_mapping 0.3624 cls_loss_causal 1.1645 re_mapping 0.1276 re_causal 0.1497 /// teacc 92.40 lr 0.00775448 +249 +0.007754484907260515 +changing lr +---------------------saving model at epoch 22---------------------------------------------------- +epoch 22, time 327.51, cls_loss 0.0782 cls_loss_mapping 0.2983 cls_loss_causal 1.0244 re_mapping 0.1161 re_causal 0.1302 /// teacc 95.91 lr 0.00756450 +249 +0.007564496387029534 +changing lr +epoch 23, time 328.44, cls_loss 0.0508 cls_loss_mapping 0.2665 cls_loss_causal 1.0062 re_mapping 0.1035 re_causal 0.1238 /// teacc 92.40 lr 0.00736934 +249 +0.007369343312364995 +changing lr +epoch 24, time 326.85, cls_loss 0.0439 cls_loss_mapping 0.2489 cls_loss_causal 0.9377 re_mapping 0.0935 re_causal 0.1083 /// teacc 93.57 lr 0.00716942 +249 +0.0071694186955877925 +changing lr +epoch 25, time 328.40, cls_loss 0.0447 cls_loss_mapping 0.2510 cls_loss_causal 0.9697 re_mapping 0.0891 re_causal 0.1042 /// teacc 95.32 lr 0.00696513 +249 +0.0069651251582696205 +changing lr +epoch 26, time 326.49, cls_loss 0.0183 cls_loss_mapping 0.2090 cls_loss_causal 0.9070 re_mapping 0.0889 re_causal 0.1054 /// teacc 94.15 lr 0.00675687 +249 +0.006756874120406716 +changing lr +epoch 27, time 329.18, cls_loss 0.0199 cls_loss_mapping 0.2252 cls_loss_causal 0.9563 re_mapping 0.0849 re_causal 0.1040 /// teacc 92.40 lr 0.00654508 +249 +0.00654508497187474 +changing lr +epoch 28, time 331.00, cls_loss 0.0349 cls_loss_mapping 0.1811 cls_loss_causal 0.8829 re_mapping 0.0737 re_causal 0.0947 /// teacc 94.15 lr 0.00633018 +249 +0.006330184227833378 +changing lr +epoch 29, time 330.93, cls_loss 0.0173 cls_loss_mapping 0.1582 cls_loss_causal 0.8307 re_mapping 0.0685 re_causal 0.0870 /// teacc 95.91 lr 0.00611260 +249 +0.006112604669781575 +changing lr +---------------------saving model at epoch 30---------------------------------------------------- +epoch 30, time 333.04, cls_loss 0.0136 cls_loss_mapping 0.1520 cls_loss_causal 0.8025 re_mapping 0.0632 re_causal 0.0809 /// teacc 97.08 lr 0.00589278 +249 +0.005892784473993186 +changing lr +epoch 31, time 328.71, cls_loss 0.0093 cls_loss_mapping 0.1464 cls_loss_causal 0.7705 re_mapping 0.0664 re_causal 0.0860 /// teacc 95.32 lr 0.00567117 +249 +0.00567116632908828 +changing lr +epoch 32, time 331.65, cls_loss 0.0048 cls_loss_mapping 0.1322 cls_loss_causal 0.7072 re_mapping 0.0552 re_causal 0.0736 /// teacc 95.32 lr 0.00544820 +249 +0.00544819654451717 +changing lr +---------------------saving model at epoch 33---------------------------------------------------- +epoch 33, time 331.93, cls_loss 0.0196 cls_loss_mapping 0.1406 cls_loss_causal 0.7016 re_mapping 0.0551 re_causal 0.0790 /// teacc 97.66 lr 0.00522432 +249 +0.005224324151752577 +changing lr +epoch 34, time 326.51, cls_loss 0.0110 cls_loss_mapping 0.1272 cls_loss_causal 0.7379 re_mapping 0.0532 re_causal 0.0753 /// teacc 95.91 lr 0.00500000 +249 +0.005000000000000003 +changing lr +epoch 35, time 326.79, cls_loss 0.0039 cls_loss_mapping 0.1204 cls_loss_causal 0.7016 re_mapping 0.0500 re_causal 0.0750 /// teacc 96.49 lr 0.00477568 +249 +0.004775675848247429 +changing lr +epoch 36, time 328.75, cls_loss 0.0098 cls_loss_mapping 0.1122 cls_loss_causal 0.6372 re_mapping 0.0458 re_causal 0.0661 /// teacc 95.32 lr 0.00455180 +249 +0.004551803455482836 +changing lr +epoch 37, time 333.82, cls_loss 0.0088 cls_loss_mapping 0.1083 cls_loss_causal 0.6648 re_mapping 0.0459 re_causal 0.0701 /// teacc 95.91 lr 0.00432883 +249 +0.004328833670911726 +changing lr +epoch 38, time 328.36, cls_loss 0.0111 cls_loss_mapping 0.1082 cls_loss_causal 0.6774 re_mapping 0.0479 re_causal 0.0716 /// teacc 94.74 lr 0.00410722 +249 +0.0041072155260068206 +changing lr +epoch 39, time 329.81, cls_loss 0.0019 cls_loss_mapping 0.0890 cls_loss_causal 0.6447 re_mapping 0.0461 re_causal 0.0699 /// teacc 95.32 lr 0.00388740 +249 +0.0038873953302184317 +changing lr +epoch 40, time 329.55, cls_loss 0.0031 cls_loss_mapping 0.0853 cls_loss_causal 0.5882 re_mapping 0.0445 re_causal 0.0632 /// teacc 94.74 lr 0.00366982 +249 +0.003669815772166629 +changing lr +epoch 41, time 330.31, cls_loss 0.0050 cls_loss_mapping 0.0811 cls_loss_causal 0.5662 re_mapping 0.0384 re_causal 0.0568 /// teacc 95.32 lr 0.00345492 +249 +0.0034549150281252667 +changing lr +epoch 42, time 333.18, cls_loss 0.0062 cls_loss_mapping 0.0839 cls_loss_causal 0.6104 re_mapping 0.0375 re_causal 0.0582 /// teacc 95.91 lr 0.00324313 +249 +0.0032431258795932905 +changing lr +epoch 43, time 329.10, cls_loss 0.0014 cls_loss_mapping 0.0792 cls_loss_causal 0.5998 re_mapping 0.0385 re_causal 0.0578 /// teacc 96.49 lr 0.00303487 +249 +0.0030348748417303863 +changing lr +epoch 44, time 327.44, cls_loss 0.0038 cls_loss_mapping 0.0816 cls_loss_causal 0.5993 re_mapping 0.0363 re_causal 0.0564 /// teacc 96.49 lr 0.00283058 +249 +0.0028305813044122124 +changing lr +epoch 45, time 328.69, cls_loss 0.0064 cls_loss_mapping 0.0724 cls_loss_causal 0.5434 re_mapping 0.0350 re_causal 0.0566 /// teacc 97.08 lr 0.00263066 +249 +0.0026306566876350096 +changing lr +epoch 46, time 329.11, cls_loss 0.0036 cls_loss_mapping 0.0732 cls_loss_causal 0.6550 re_mapping 0.0336 re_causal 0.0560 /// teacc 97.66 lr 0.00243550 +249 +0.0024355036129704724 +changing lr +epoch 47, time 330.95, cls_loss 0.0028 cls_loss_mapping 0.0696 cls_loss_causal 0.5213 re_mapping 0.0347 re_causal 0.0540 /// teacc 95.32 lr 0.00224552 +249 +0.00224551509273949 +changing lr +epoch 48, time 329.49, cls_loss 0.0022 cls_loss_mapping 0.0614 cls_loss_causal 0.5186 re_mapping 0.0319 re_causal 0.0531 /// teacc 97.08 lr 0.00206107 +249 +0.002061073738537637 +changing lr +epoch 49, time 327.39, cls_loss 0.0030 cls_loss_mapping 0.0631 cls_loss_causal 0.5368 re_mapping 0.0315 re_causal 0.0477 /// teacc 97.08 lr 0.00188255 +249 +0.0018825509907063344 +changing lr +epoch 50, time 330.68, cls_loss 0.0025 cls_loss_mapping 0.0624 cls_loss_causal 0.5418 re_mapping 0.0308 re_causal 0.0501 /// teacc 95.91 lr 0.00171031 +249 +0.0017103063703014388 +changing lr +epoch 51, time 331.11, cls_loss 0.0024 cls_loss_mapping 0.0666 cls_loss_causal 0.6219 re_mapping 0.0303 re_causal 0.0463 /// teacc 94.15 lr 0.00154469 +249 +0.0015446867550656784 +changing lr +epoch 52, time 329.80, cls_loss 0.0037 cls_loss_mapping 0.0624 cls_loss_causal 0.5204 re_mapping 0.0305 re_causal 0.0459 /// teacc 95.91 lr 0.00138603 +249 +0.001386025680863044 +changing lr +epoch 53, time 330.30, cls_loss 0.0021 cls_loss_mapping 0.0573 cls_loss_causal 0.4976 re_mapping 0.0330 re_causal 0.0522 /// teacc 96.49 lr 0.00123464 +249 +0.0012346426699819469 +changing lr +epoch 54, time 328.15, cls_loss 0.0037 cls_loss_mapping 0.0636 cls_loss_causal 0.5476 re_mapping 0.0300 re_causal 0.0478 /// teacc 94.74 lr 0.00109084 +249 +0.0010908425876598518 +changing lr +epoch 55, time 330.82, cls_loss 0.0019 cls_loss_mapping 0.0573 cls_loss_causal 0.4965 re_mapping 0.0298 re_causal 0.0464 /// teacc 94.74 lr 0.00095492 +249 +0.000954915028125264 +changing lr +epoch 56, time 327.34, cls_loss 0.0026 cls_loss_mapping 0.0569 cls_loss_causal 0.5251 re_mapping 0.0303 re_causal 0.0466 /// teacc 95.91 lr 0.00082713 +249 +0.0008271337313934874 +changing lr +epoch 57, time 333.58, cls_loss 0.0042 cls_loss_mapping 0.0546 cls_loss_causal 0.5309 re_mapping 0.0287 re_causal 0.0428 /// teacc 95.32 lr 0.00070776 +249 +0.00070775603199067 +changing lr +epoch 58, time 328.86, cls_loss 0.0031 cls_loss_mapping 0.0587 cls_loss_causal 0.5149 re_mapping 0.0288 re_causal 0.0456 /// teacc 96.49 lr 0.00059702 +249 +0.0005970223407163104 +changing lr +epoch 59, time 328.86, cls_loss 0.0046 cls_loss_mapping 0.0559 cls_loss_causal 0.5242 re_mapping 0.0292 re_causal 0.0461 /// teacc 95.32 lr 0.00049516 +249 +0.0004951556604879052 +changing lr +epoch 60, time 329.33, cls_loss 0.0035 cls_loss_mapping 0.0531 cls_loss_causal 0.5105 re_mapping 0.0286 re_causal 0.0415 /// teacc 94.74 lr 0.00040236 +249 +0.00040236113724274745 +changing lr +epoch 61, time 329.57, cls_loss 0.0024 cls_loss_mapping 0.0552 cls_loss_causal 0.5395 re_mapping 0.0269 re_causal 0.0440 /// teacc 95.91 lr 0.00031883 +249 +0.00031882564680131423 +changing lr +epoch 62, time 333.79, cls_loss 0.0025 cls_loss_mapping 0.0505 cls_loss_causal 0.5307 re_mapping 0.0257 re_causal 0.0430 /// teacc 95.32 lr 0.00024472 +249 +0.0002447174185242325 +changing lr +epoch 63, time 325.49, cls_loss 0.0033 cls_loss_mapping 0.0561 cls_loss_causal 0.5009 re_mapping 0.0285 re_causal 0.0429 /// teacc 96.49 lr 0.00018019 +249 +0.0001801856965207339 +changing lr +epoch 64, time 325.60, cls_loss 0.0020 cls_loss_mapping 0.0478 cls_loss_causal 0.5195 re_mapping 0.0274 re_causal 0.0416 /// teacc 95.32 lr 0.00012536 +249 +0.000125360439090882 +changing lr +epoch 65, time 329.45, cls_loss 0.0022 cls_loss_mapping 0.0502 cls_loss_causal 0.4924 re_mapping 0.0274 re_causal 0.0425 /// teacc 94.15 lr 0.00008035 +249 +8.03520570068517e-05 +changing lr +epoch 66, time 331.82, cls_loss 0.0036 cls_loss_mapping 0.0536 cls_loss_causal 0.5226 re_mapping 0.0276 re_causal 0.0429 /// teacc 95.91 lr 0.00004525 +249 +4.5251191160326525e-05 +changing lr +epoch 67, time 328.42, cls_loss 0.0030 cls_loss_mapping 0.0563 cls_loss_causal 0.5390 re_mapping 0.0282 re_causal 0.0435 /// teacc 95.32 lr 0.00002013 +249 +2.0128530023804673e-05 +changing lr +epoch 68, time 331.35, cls_loss 0.0034 cls_loss_mapping 0.0501 cls_loss_causal 0.5100 re_mapping 0.0269 re_causal 0.0424 /// teacc 95.32 lr 0.00000503 +249 +5.034667293427056e-06 +changing lr +epoch 69, time 332.59, cls_loss 0.0023 cls_loss_mapping 0.0540 cls_loss_causal 0.5166 re_mapping 0.0279 re_causal 0.0451 /// teacc 97.08 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'photo', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/photo_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['photo', 'art_painting', 'cartoon', 'sketch'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) + photo art_painting cartoon sketch Avg +w/o do (original x) 99.700599 60.253906 43.515358 57.724612 53.831292 + photo art_painting cartoon sketch Avg +do 99.760479 60.15625 49.274744 60.57521 56.668735 diff --git a/Meta-causal/code/56722.error b/Meta-causal/code/56722.error new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/Meta-causal/code/56722.log b/Meta-causal/code/56722.log new file mode 100644 index 0000000000000000000000000000000000000000..7f87ef44ae593d26a6c9da1bc6c04bcc59b5b290 --- /dev/null +++ b/Meta-causal/code/56722.log @@ -0,0 +1,333 @@ +/home/yuqian_fu +{'gpu': '0', 'data': 'sketch', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} +stride: 5 +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531]) +--------------------------CA_multiple-------------------------- +---------------------------16 factors----------------- +randm: True +randn: True +n: 3 +randm: False +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398]) +-------------------------------------loading pretrain weights---------------------------------- +588 +0.01 +changing lr +---------------------saving model at epoch 0---------------------------------------------------- +epoch 0, time 845.95, cls_loss 3.6738 cls_loss_mapping 1.1243 cls_loss_causal 1.4459 re_mapping 0.6948 re_causal 0.6950 /// teacc 87.69 lr 0.00999497 +588 +0.009994965332706574 +changing lr +---------------------saving model at epoch 1---------------------------------------------------- +epoch 1, time 854.64, cls_loss 0.5577 cls_loss_mapping 0.4581 cls_loss_causal 1.0021 re_mapping 0.2545 re_causal 0.2541 /// teacc 90.20 lr 0.00997987 +588 +0.009979871469976196 +changing lr +epoch 2, time 878.75, cls_loss 0.1988 cls_loss_mapping 0.2884 cls_loss_causal 0.8433 re_mapping 0.1588 re_causal 0.1584 /// teacc 89.70 lr 0.00995475 +588 +0.009954748808839675 +changing lr +---------------------saving model at epoch 3---------------------------------------------------- +epoch 3, time 892.14, cls_loss 0.1337 cls_loss_mapping 0.2165 cls_loss_causal 0.7708 re_mapping 0.1261 re_causal 0.1263 /// teacc 92.96 lr 0.00991965 +588 +0.009919647942993149 +changing lr +---------------------saving model at epoch 4---------------------------------------------------- +epoch 4, time 852.27, cls_loss 0.0720 cls_loss_mapping 0.1604 cls_loss_causal 0.6995 re_mapping 0.1031 re_causal 0.1040 /// teacc 93.22 lr 0.00987464 +588 +0.009874639560909117 +changing lr +---------------------saving model at epoch 5---------------------------------------------------- +epoch 5, time 847.00, cls_loss 0.0390 cls_loss_mapping 0.1253 cls_loss_causal 0.6512 re_mapping 0.0839 re_causal 0.0858 /// teacc 93.72 lr 0.00981981 +588 +0.009819814303479266 +changing lr +epoch 6, time 868.82, cls_loss 0.0280 cls_loss_mapping 0.1074 cls_loss_causal 0.6153 re_mapping 0.0698 re_causal 0.0724 /// teacc 92.46 lr 0.00975528 +588 +0.009755282581475767 +changing lr +---------------------saving model at epoch 7---------------------------------------------------- +epoch 7, time 845.22, cls_loss 0.0251 cls_loss_mapping 0.0936 cls_loss_causal 0.5946 re_mapping 0.0582 re_causal 0.0616 /// teacc 93.97 lr 0.00968117 +588 +0.009681174353198686 +changing lr +epoch 8, time 850.90, cls_loss 0.0209 cls_loss_mapping 0.0815 cls_loss_causal 0.5643 re_mapping 0.0500 re_causal 0.0549 /// teacc 92.71 lr 0.00959764 +588 +0.009597638862757255 +changing lr +---------------------saving model at epoch 9---------------------------------------------------- +epoch 9, time 861.45, cls_loss 0.0196 cls_loss_mapping 0.0758 cls_loss_causal 0.5799 re_mapping 0.0425 re_causal 0.0492 /// teacc 94.97 lr 0.00950484 +588 +0.009504844339512096 +changing lr +epoch 10, time 887.24, cls_loss 0.0124 cls_loss_mapping 0.0557 cls_loss_causal 0.5474 re_mapping 0.0349 re_causal 0.0424 /// teacc 94.22 lr 0.00940298 +588 +0.009402977659283692 +changing lr +epoch 11, time 860.82, cls_loss 0.0155 cls_loss_mapping 0.0521 cls_loss_causal 0.5327 re_mapping 0.0303 re_causal 0.0397 /// teacc 92.96 lr 0.00929224 +588 +0.009292243968009333 +changing lr +epoch 12, time 882.17, cls_loss 0.0122 cls_loss_mapping 0.0491 cls_loss_causal 0.5306 re_mapping 0.0254 re_causal 0.0353 /// teacc 92.46 lr 0.00917287 +588 +0.009172866268606516 +changing lr +epoch 13, time 886.31, cls_loss 0.0123 cls_loss_mapping 0.0488 cls_loss_causal 0.5114 re_mapping 0.0231 re_causal 0.0341 /// teacc 93.97 lr 0.00904508 +588 +0.00904508497187474 +changing lr +epoch 14, time 860.16, cls_loss 0.0097 cls_loss_mapping 0.0391 cls_loss_causal 0.5384 re_mapping 0.0201 re_causal 0.0320 /// teacc 94.47 lr 0.00890916 +588 +0.008909157412340152 +changing lr +epoch 15, time 851.33, cls_loss 0.0083 cls_loss_mapping 0.0374 cls_loss_causal 0.4965 re_mapping 0.0180 re_causal 0.0299 /// teacc 92.71 lr 0.00876536 +588 +0.00876535733001806 +changing lr +epoch 16, time 856.91, cls_loss 0.0087 cls_loss_mapping 0.0368 cls_loss_causal 0.4716 re_mapping 0.0163 re_causal 0.0278 /// teacc 91.96 lr 0.00861397 +588 +0.008613974319136962 +changing lr +epoch 17, time 881.17, cls_loss 0.0070 cls_loss_mapping 0.0307 cls_loss_causal 0.4890 re_mapping 0.0146 re_causal 0.0269 /// teacc 93.47 lr 0.00845531 +588 +0.008455313244934327 +changing lr +epoch 18, time 868.14, cls_loss 0.0060 cls_loss_mapping 0.0262 cls_loss_causal 0.4835 re_mapping 0.0132 re_causal 0.0256 /// teacc 93.47 lr 0.00828969 +588 +0.008289693629698565 +changing lr +epoch 19, time 883.13, cls_loss 0.0067 cls_loss_mapping 0.0258 cls_loss_causal 0.4667 re_mapping 0.0123 re_causal 0.0250 /// teacc 93.72 lr 0.00811745 +588 +0.00811744900929367 +changing lr +epoch 20, time 884.58, cls_loss 0.0064 cls_loss_mapping 0.0251 cls_loss_causal 0.4629 re_mapping 0.0117 re_causal 0.0239 /// teacc 94.72 lr 0.00793893 +588 +0.007938926261462368 +changing lr +epoch 21, time 875.05, cls_loss 0.0051 cls_loss_mapping 0.0202 cls_loss_causal 0.4715 re_mapping 0.0107 re_causal 0.0234 /// teacc 92.46 lr 0.00775448 +588 +0.007754484907260515 +changing lr +epoch 22, time 896.60, cls_loss 0.0054 cls_loss_mapping 0.0194 cls_loss_causal 0.4351 re_mapping 0.0099 re_causal 0.0214 /// teacc 94.97 lr 0.00756450 +588 +0.007564496387029534 +changing lr +epoch 23, time 860.93, cls_loss 0.0049 cls_loss_mapping 0.0175 cls_loss_causal 0.4279 re_mapping 0.0094 re_causal 0.0210 /// teacc 92.71 lr 0.00736934 +588 +0.007369343312364995 +changing lr +epoch 24, time 870.01, cls_loss 0.0046 cls_loss_mapping 0.0183 cls_loss_causal 0.4499 re_mapping 0.0094 re_causal 0.0216 /// teacc 94.22 lr 0.00716942 +588 +0.0071694186955877925 +changing lr +epoch 25, time 881.12, cls_loss 0.0059 cls_loss_mapping 0.0212 cls_loss_causal 0.4502 re_mapping 0.0092 re_causal 0.0210 /// teacc 94.22 lr 0.00696513 +588 +0.0069651251582696205 +changing lr +epoch 26, time 883.73, cls_loss 0.0052 cls_loss_mapping 0.0151 cls_loss_causal 0.4330 re_mapping 0.0088 re_causal 0.0207 /// teacc 94.47 lr 0.00675687 +588 +0.006756874120406716 +changing lr +epoch 27, time 876.67, cls_loss 0.0050 cls_loss_mapping 0.0183 cls_loss_causal 0.4334 re_mapping 0.0082 re_causal 0.0200 /// teacc 93.22 lr 0.00654508 +588 +0.00654508497187474 +changing lr +epoch 28, time 849.09, cls_loss 0.0067 cls_loss_mapping 0.0154 cls_loss_causal 0.4283 re_mapping 0.0084 re_causal 0.0204 /// teacc 93.47 lr 0.00633018 +588 +0.006330184227833378 +changing lr +epoch 29, time 851.32, cls_loss 0.0044 cls_loss_mapping 0.0147 cls_loss_causal 0.3901 re_mapping 0.0077 re_causal 0.0185 /// teacc 92.96 lr 0.00611260 +588 +0.006112604669781575 +changing lr +epoch 30, time 854.67, cls_loss 0.0034 cls_loss_mapping 0.0126 cls_loss_causal 0.4241 re_mapping 0.0076 re_causal 0.0193 /// teacc 93.72 lr 0.00589278 +588 +0.005892784473993186 +changing lr +epoch 31, time 861.52, cls_loss 0.0048 cls_loss_mapping 0.0151 cls_loss_causal 0.4106 re_mapping 0.0072 re_causal 0.0186 /// teacc 93.22 lr 0.00567117 +588 +0.00567116632908828 +changing lr +epoch 32, time 886.70, cls_loss 0.0034 cls_loss_mapping 0.0119 cls_loss_causal 0.4174 re_mapping 0.0070 re_causal 0.0183 /// teacc 93.72 lr 0.00544820 +588 +0.00544819654451717 +changing lr +epoch 33, time 865.62, cls_loss 0.0038 cls_loss_mapping 0.0111 cls_loss_causal 0.4096 re_mapping 0.0068 re_causal 0.0178 /// teacc 92.96 lr 0.00522432 +588 +0.005224324151752577 +changing lr +epoch 34, time 853.80, cls_loss 0.0039 cls_loss_mapping 0.0117 cls_loss_causal 0.4176 re_mapping 0.0066 re_causal 0.0176 /// teacc 93.22 lr 0.00500000 +588 +0.005000000000000003 +changing lr +epoch 35, time 873.63, cls_loss 0.0043 cls_loss_mapping 0.0126 cls_loss_causal 0.4324 re_mapping 0.0065 re_causal 0.0176 /// teacc 93.22 lr 0.00477568 +588 +0.004775675848247429 +changing lr +epoch 36, time 847.64, cls_loss 0.0035 cls_loss_mapping 0.0099 cls_loss_causal 0.4156 re_mapping 0.0062 re_causal 0.0166 /// teacc 93.72 lr 0.00455180 +588 +0.004551803455482836 +changing lr +epoch 37, time 821.88, cls_loss 0.0038 cls_loss_mapping 0.0099 cls_loss_causal 0.4130 re_mapping 0.0059 re_causal 0.0165 /// teacc 94.22 lr 0.00432883 +588 +0.004328833670911726 +changing lr +epoch 38, time 833.52, cls_loss 0.0039 cls_loss_mapping 0.0113 cls_loss_causal 0.3887 re_mapping 0.0059 re_causal 0.0166 /// teacc 94.97 lr 0.00410722 +588 +0.0041072155260068206 +changing lr +epoch 39, time 803.24, cls_loss 0.0032 cls_loss_mapping 0.0079 cls_loss_causal 0.4193 re_mapping 0.0058 re_causal 0.0165 /// teacc 94.72 lr 0.00388740 +588 +0.0038873953302184317 +changing lr +epoch 40, time 810.38, cls_loss 0.0034 cls_loss_mapping 0.0082 cls_loss_causal 0.3832 re_mapping 0.0056 re_causal 0.0154 /// teacc 93.47 lr 0.00366982 +588 +0.003669815772166629 +changing lr +epoch 41, time 798.30, cls_loss 0.0038 cls_loss_mapping 0.0093 cls_loss_causal 0.3853 re_mapping 0.0054 re_causal 0.0152 /// teacc 93.72 lr 0.00345492 +588 +0.0034549150281252667 +changing lr +epoch 42, time 770.71, cls_loss 0.0038 cls_loss_mapping 0.0078 cls_loss_causal 0.4206 re_mapping 0.0052 re_causal 0.0155 /// teacc 93.22 lr 0.00324313 +588 +0.0032431258795932905 +changing lr +epoch 43, time 769.73, cls_loss 0.0032 cls_loss_mapping 0.0085 cls_loss_causal 0.3786 re_mapping 0.0052 re_causal 0.0147 /// teacc 94.22 lr 0.00303487 +588 +0.0030348748417303863 +changing lr +epoch 44, time 781.30, cls_loss 0.0030 cls_loss_mapping 0.0066 cls_loss_causal 0.3762 re_mapping 0.0052 re_causal 0.0141 /// teacc 92.96 lr 0.00283058 +588 +0.0028305813044122124 +changing lr +epoch 45, time 763.23, cls_loss 0.0028 cls_loss_mapping 0.0060 cls_loss_causal 0.3935 re_mapping 0.0050 re_causal 0.0143 /// teacc 93.97 lr 0.00263066 +588 +0.0026306566876350096 +changing lr +epoch 46, time 756.78, cls_loss 0.0030 cls_loss_mapping 0.0072 cls_loss_causal 0.3847 re_mapping 0.0049 re_causal 0.0141 /// teacc 94.47 lr 0.00243550 +588 +0.0024355036129704724 +changing lr +epoch 47, time 753.45, cls_loss 0.0027 cls_loss_mapping 0.0062 cls_loss_causal 0.3732 re_mapping 0.0048 re_causal 0.0134 /// teacc 93.72 lr 0.00224552 +588 +0.00224551509273949 +changing lr +epoch 48, time 753.93, cls_loss 0.0029 cls_loss_mapping 0.0050 cls_loss_causal 0.3621 re_mapping 0.0047 re_causal 0.0131 /// teacc 94.22 lr 0.00206107 +588 +0.002061073738537637 +changing lr +epoch 49, time 760.37, cls_loss 0.0028 cls_loss_mapping 0.0057 cls_loss_causal 0.3736 re_mapping 0.0048 re_causal 0.0132 /// teacc 93.47 lr 0.00188255 +588 +0.0018825509907063344 +changing lr +---------------------saving model at epoch 50---------------------------------------------------- +epoch 50, time 761.62, cls_loss 0.0025 cls_loss_mapping 0.0047 cls_loss_causal 0.3886 re_mapping 0.0047 re_causal 0.0133 /// teacc 95.48 lr 0.00171031 +588 +0.0017103063703014388 +changing lr +epoch 51, time 757.83, cls_loss 0.0026 cls_loss_mapping 0.0051 cls_loss_causal 0.3723 re_mapping 0.0047 re_causal 0.0131 /// teacc 92.21 lr 0.00154469 +588 +0.0015446867550656784 +changing lr +epoch 52, time 756.68, cls_loss 0.0027 cls_loss_mapping 0.0047 cls_loss_causal 0.3874 re_mapping 0.0046 re_causal 0.0131 /// teacc 92.71 lr 0.00138603 +588 +0.001386025680863044 +changing lr +epoch 53, time 758.80, cls_loss 0.0027 cls_loss_mapping 0.0049 cls_loss_causal 0.3915 re_mapping 0.0046 re_causal 0.0130 /// teacc 93.72 lr 0.00123464 +588 +0.0012346426699819469 +changing lr +epoch 54, time 754.88, cls_loss 0.0026 cls_loss_mapping 0.0049 cls_loss_causal 0.3825 re_mapping 0.0045 re_causal 0.0130 /// teacc 94.22 lr 0.00109084 +588 +0.0010908425876598518 +changing lr +epoch 55, time 758.50, cls_loss 0.0030 cls_loss_mapping 0.0050 cls_loss_causal 0.3839 re_mapping 0.0045 re_causal 0.0126 /// teacc 93.22 lr 0.00095492 +588 +0.000954915028125264 +changing lr +epoch 56, time 757.06, cls_loss 0.0025 cls_loss_mapping 0.0044 cls_loss_causal 0.3577 re_mapping 0.0045 re_causal 0.0122 /// teacc 94.22 lr 0.00082713 +588 +0.0008271337313934874 +changing lr +epoch 57, time 758.09, cls_loss 0.0023 cls_loss_mapping 0.0046 cls_loss_causal 0.3461 re_mapping 0.0046 re_causal 0.0118 /// teacc 93.47 lr 0.00070776 +588 +0.00070775603199067 +changing lr +epoch 58, time 760.13, cls_loss 0.0023 cls_loss_mapping 0.0039 cls_loss_causal 0.3523 re_mapping 0.0046 re_causal 0.0118 /// teacc 94.47 lr 0.00059702 +588 +0.0005970223407163104 +changing lr +epoch 59, time 756.85, cls_loss 0.0021 cls_loss_mapping 0.0035 cls_loss_causal 0.3762 re_mapping 0.0045 re_causal 0.0121 /// teacc 93.22 lr 0.00049516 +588 +0.0004951556604879052 +changing lr +epoch 60, time 754.41, cls_loss 0.0023 cls_loss_mapping 0.0041 cls_loss_causal 0.3579 re_mapping 0.0044 re_causal 0.0115 /// teacc 92.71 lr 0.00040236 +588 +0.00040236113724274745 +changing lr +epoch 61, time 758.79, cls_loss 0.0026 cls_loss_mapping 0.0042 cls_loss_causal 0.3682 re_mapping 0.0044 re_causal 0.0114 /// teacc 93.72 lr 0.00031883 +588 +0.00031882564680131423 +changing lr +epoch 62, time 752.52, cls_loss 0.0025 cls_loss_mapping 0.0039 cls_loss_causal 0.3746 re_mapping 0.0044 re_causal 0.0117 /// teacc 92.46 lr 0.00024472 +588 +0.0002447174185242325 +changing lr +epoch 63, time 758.14, cls_loss 0.0025 cls_loss_mapping 0.0034 cls_loss_causal 0.3751 re_mapping 0.0044 re_causal 0.0115 /// teacc 93.72 lr 0.00018019 +588 +0.0001801856965207339 +changing lr +epoch 64, time 751.51, cls_loss 0.0024 cls_loss_mapping 0.0034 cls_loss_causal 0.3590 re_mapping 0.0044 re_causal 0.0114 /// teacc 93.47 lr 0.00012536 +588 +0.000125360439090882 +changing lr +epoch 65, time 760.25, cls_loss 0.0030 cls_loss_mapping 0.0049 cls_loss_causal 0.3519 re_mapping 0.0044 re_causal 0.0112 /// teacc 93.97 lr 0.00008035 +588 +8.03520570068517e-05 +changing lr +epoch 66, time 759.68, cls_loss 0.0023 cls_loss_mapping 0.0036 cls_loss_causal 0.3571 re_mapping 0.0044 re_causal 0.0114 /// teacc 92.96 lr 0.00004525 +588 +4.5251191160326525e-05 +changing lr +epoch 67, time 758.83, cls_loss 0.0025 cls_loss_mapping 0.0043 cls_loss_causal 0.3541 re_mapping 0.0044 re_causal 0.0113 /// teacc 92.21 lr 0.00002013 +588 +2.0128530023804673e-05 +changing lr +epoch 68, time 755.82, cls_loss 0.0021 cls_loss_mapping 0.0032 cls_loss_causal 0.3385 re_mapping 0.0044 re_causal 0.0113 /// teacc 93.22 lr 0.00000503 +588 +5.034667293427056e-06 +changing lr +epoch 69, time 757.80, cls_loss 0.0027 cls_loss_mapping 0.0040 cls_loss_causal 0.3563 re_mapping 0.0044 re_causal 0.0113 /// teacc 93.22 lr 0.00000000 +---------------------saving last model at epoch 69---------------------------------------------------- +/home/yuqian_fu +{'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_ReProduceMetaCausal/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} +-------------------------------------loading pretrain weights---------------------------------- +loading weight of last +randm: False +stride: 5 +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +loading weight of last +columns: ['sketch', 'art_painting', 'cartoon', 'photo'] +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) +/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) + sketch art_painting cartoon photo Avg +w/o do (original x) 99.312802 55.029297 67.491468 57.964072 60.161612 + sketch art_painting cartoon photo Avg +do 99.312802 49.804688 63.90785 55.449102 56.387213 diff --git a/Meta-causal/code/data_loader_joint_v3.py b/Meta-causal/code/data_loader_joint_v3.py new file mode 100644 index 0000000000000000000000000000000000000000..dd2de2de850d89657507ff9a0e348c94c0e070d0 --- /dev/null +++ b/Meta-causal/code/data_loader_joint_v3.py @@ -0,0 +1,743 @@ +''' Digit 实验 +''' +import torch +import torch.nn.functional as F +from torch.utils.data import Dataset, TensorDataset +from torchvision import transforms +from torchvision.datasets import MNIST, SVHN, CIFAR10, STL10, USPS + +import os +import pickle +import numpy as np +import h5py +#import cv2 +from scipy.io import loadmat +from PIL import Image + +from tools.autoaugment import SVHNPolicy, CIFAR10Policy +from tools.randaugment import RandAugment +from tools.causalaugment_v3 import RandAugment_incausal, FactualAugment_incausal, CounterfactualAugment_incausal, MultiCounterfactualAugment_incausal + +class myTensorDataset(Dataset): + def __init__(self, x, y, transform=None, transform2=None, transform3=None, twox=False): + self.x = x + self.y = y + self.transform = transform + self.transform2 = transform2 + self.transform3 = transform3 + self.twox = twox + def __len__(self): + return len(self.x) + def __getitem__(self, index): + x = self.x[index] + y = self.y[index] + c, h, w =x.shape + # print("x.shape:",x.shape) + if self.transform is not None: + x_RA = self.transform(x) + # print("x_RA.shape:",x_RA.shape) + if self.transform3 is not None: + x_CA = self.transform3(x_RA) + x_CA = x_CA.reshape(-1,c,h,w) + # print("x_CA.shape:",x_CA.shape) + if self.transform2 is not None: + x_FA = self.transform2(x) + # x_FA = x_FA.view(c,13,h,w) + x_FA = x_FA.reshape(-1,c,h,w) + # print("x_FA_in getitem.shape:",x_FA.shape) + # print("x_FA.shape:",x_FA.shape) + return (x, x_RA, x_FA, x_CA), y + else: + return (x, x_RA, x_CA), y + else: + if self.transform2 is not None: + x_FA = self.transform2(x) + x_FA = x_FA.reshape(-1,c,h,w) + return (x, x_RA, x_FA), y + else: + if self.twox: + return (x, x_RA), y + else: + return x_RA, y + +HOME = os.environ['HOME'] +print(HOME) +def resize_imgs(x, size): + ''' 目前只能处理单通道 + x [n, 28, 28] + size int + ''' + resize_x = np.zeros([x.shape[0], size, size]) + for i, im in enumerate(x): + im = Image.fromarray(im) + im = im.resize([size, size], Image.ANTIALIAS) + resize_x[i] = np.asarray(im) + return resize_x + +def load_mnist(split='train', translate=None, twox=False, ntr=None, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + ''' + autoaug == 'AA', AutoAugment + 'FastAA', Fast AutoAugment + 'RA', RandAugment + channels == 3 默认返回 rgb 3通道图像 + 1 返回单通道图像 + ''' + #path = f'data/mnist-{split}.pkl' + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/minst-{split}.pkl' + if not os.path.exists(path): + dataset = MNIST(f'{HOME}/.pytorch/MNIST', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + if split=='train': + x, y = x[0:10000], y[0:10000] + x = torch.tensor(resize_imgs(x.numpy(), 32)) + x = (x.float()/255.).unsqueeze(1).repeat(1,3,1,1) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + with open(path, 'rb') as f: + # print("reading!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!") + x, y = pickle.load(f) + if channels == 1: + x = x[:,0:1,:,:] + + if ntr is not None: + x, y = x[0:ntr], y[0:ntr] + + # 如果没有数据增强 + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + + # 数据增强管道 + transform = [transforms.ToPILImage()] + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def load_cifar10(split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + dataset = CIFAR10(f'{HOME}/.pytorch/CIFAR10', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y) + x = x.float()/255. + print(x.shape,y.shape) + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset +def load_IMG(task='S-U', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + # path = f'data/img2vid/{domain}/stanford40_12.npz' + if task == 'S-U': + path = f'data/img2vid/{task}/stanford40_12.npz' + elif task == 'E-H': + path = f'data/img2vid/{task}/EAD50_13.npz' + print(path) + dataset = np.load(path) + x, y = dataset['x'], dataset['y'] + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def load_VID(task='S-U',split='1'): + if task == 'S-U': + path = f'data/img2vid/{task}/ucf101_12_frame_sample8_{split}.npz' + elif task == 'E-H': + path = f'data/img2vid/{task}/hmdb51_13_frame_sample8_{split}.npz' + dataset = np.load(path) + print(path) + x, y = dataset['x'], dataset['y'] + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + dataset = TensorDataset(x, y) + return dataset + +def load_pacs(domain='photo', split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + #path = f'data/PACS/{domain}_{split}.hdf5' + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/{domain}_{split}.hdf5' + dataset = h5py.File(path, 'r') + x, y = dataset['images'], dataset['labels'] + #for i in range(20): + # cv2.imwrite('data/PACS/debug_images/img_cv2_'+domain+'_'+split+'_'+str(i)+'.png', x[i]) + b, g, r = np.split(x,3,axis=-1) + x = np.concatenate((r,g,b),axis=-1) + x = x.transpose(0,3,1,2) + x, y = torch.tensor(x), torch.tensor(y, dtype=torch.long) + y = y - 1 + x = x.float()/255. + print(path,x.shape,y.shape) + # for i in range(20): + # img_temp = transforms.ToPILImage()(x[i]) + # img_temp.save('data/PACS/debug_images/img_pil_'+domain+'_'+split+'_'+str(i)+'.png') + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + +def read_dataset(domain, split): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/{domain}_{split}.hdf5' + dataset = h5py.File(path, 'r') + x_temp, y_temp = dataset['images'], dataset['labels'] + b, g, r = np.split(x_temp,3,axis=-1) + x_temp = np.concatenate((r,g,b),axis=-1) + x_temp = x_temp.transpose(0,3,1,2) + x_temp, y_temp = torch.tensor(x_temp), torch.tensor(y_temp, dtype=torch.long) + y_temp = y_temp - 1 + x_temp = x_temp.float()/255. + return x_temp, y_temp + +def load_pacs_multi(target_domain=['photo'], split='train', translate=None, twox=False, autoaug=None, factor_num=16, randm=False,randn=False,channels=3,n=3,stride=5): + domains = ['art_painting', 'cartoon', 'photo', 'sketch'] + source_domain = [i for i in domains if i != target_domain] + for i in range(len(source_domain)): + x_temp, y_temp = read_dataset(source_domain[i],split=split) + print(x_temp.shape,y_temp.shape) + if i == 0: + x = x_temp.clone() + y = y_temp.clone() + else: + x = torch.cat([x,x_temp],0) + y = torch.cat([y,y_temp],0) + print(x.shape,y.shape) + if (translate is None) and (autoaug is None): + dataset = TensorDataset(x, y) + return dataset + #x.transpose(0,3,1,2) + + # 数据增强管道 + transform = [transforms.ToPILImage()] + if autoaug != 'CA_multiple_noSingle': + transform_single_factor = [transforms.ToPILImage()] + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA = [transforms.ToPILImage()] + if translate is not None: + transform.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug == 'CA' or autoaug == 'CA_multiple' or autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.RandomAffine(0, [translate, translate])) + if autoaug is not None: + if autoaug == 'CA': + print("--------------------------CA--------------------------") + print("n:",n) + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(CounterfactualAugment_incausal(factor_num)) + elif autoaug == 'CA_multiple': + print("--------------------------CA_multiple--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'CA_multiple_noSingle': + print("--------------------------CA_multiple_noSingle--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + # transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + transform_CA.append(MultiCounterfactualAugment_incausal(factor_num, stride)) + elif autoaug == 'Ours_A': + print("--------------------------Ours_Augment--------------------------") + transform.append(RandAugment_incausal(n,4,factor_num, randm=randm,randn=randn)) + transform_single_factor.append(FactualAugment_incausal(4, factor_num, randm=False)) + + transform.append(transforms.ToTensor()) + transform = transforms.Compose(transform) + if autoaug != 'CA_multiple_noSingle': + transform_single_factor.append(transforms.ToTensor()) + transform_single_factor = transforms.Compose(transform_single_factor) + if autoaug == 'CA' or autoaug == 'CA_multiple': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, transform3=transform_CA,twox=twox) + elif autoaug == 'CA_multiple_noSingle': + transform_CA.append(transforms.ToTensor()) + transform_CA = transforms.Compose(transform_CA) + dataset = myTensorDataset(x, y, transform=transform, transform3=transform_CA,twox=twox) + else: + dataset = myTensorDataset(x, y, transform=transform, transform2=transform_single_factor, twox=twox) + # print(x.shape) + # print(y.shape) + return dataset + + +def load_cifar10_c_level1(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level1.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level1") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[0:10000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level1") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level2(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level2.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level2") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[10000:20000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level2") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level3(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level3.pkl' + if not os.path.exists(path): + print("generating cifar10_c_level3") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[20000:30000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level3") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level4(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level4.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level4") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[30000:40000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level4") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c_level5(dataroot): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/cifar10_c_level5.pkl' + if not os.path.exists(path): + print("genenrating cifar10_c_level5") + labels = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = labels[0:10000] + x = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y = y_single + else: + y = np.hstack((y,y_single)) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + imgs = np.load(os.path.join(dataroot,filename)) + imgs = imgs.transpose(0,3,1,2) + imgs = torch.tensor(imgs) + imgs = imgs.float()/255. + print(imgs.shape) + x[index*10000:(index+1)*10000] = imgs[40000:50000] + index = index + 1 + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + else: + print("reading cifar10_c_level5") + with open(path, 'rb') as f: + x, y = pickle.load(f) + dataset = TensorDataset(x, y) + return dataset +def load_cifar10_c(dataroot): + y = np.load(os.path.join(dataroot, 'labels.npy')) + print("y.shape:",y.shape) + y_single = y[0:10000] + x1 = torch.zeros((190000,3,32,32)) + x2 = torch.zeros((190000,3,32,32)) + x3 = torch.zeros((190000,3,32,32)) + x4 = torch.zeros((190000,3,32,32)) + x5 = torch.zeros((190000,3,32,32)) + for j in range(19): + if j == 0: + y_total = y_single + else: + y_total = np.hstack((y_total,y_single)) + print("y_total.shape:",y_total.shape) + index = 0 + for filename in os.listdir(dataroot): + if filename=='labels.npy': + continue + else: + x = np.load(os.path.join(dataroot,filename)) + x = x.transpose(0,3,1,2) + x = torch.tensor(x) + x = x.float()/255. + print(x.shape) + x1[index*10000:(index+1)*10000] = x[0:10000] + x2[index*10000:(index+1)*10000] = x[10000:20000] + x3[index*10000:(index+1)*10000] = x[20000:30000] + x4[index*10000:(index+1)*10000] = x[30000:40000] + x5[index*10000:(index+1)*10000] = x[40000:50000] + index = index + 1 + # x1, x2, x3, x4, x5, y_total = torch.tensor(x1), torch.tensor(x2), torch.tensor(x3),\ + # torch.tensor(x4),torch.tensor(x5),torch.tensor(y_total) + y_total = torch.tensor(y_total) + dataset1 = TensorDataset(x1, y_total) + dataset2 = TensorDataset(x2, y_total) + dataset3 = TensorDataset(x3, y_total) + dataset4 = TensorDataset(x4, y_total) + dataset5 = TensorDataset(x5, y_total) + return dataset1,dataset2,dataset3,dataset4,dataset5 + +def load_cifar10_c_class(dataroot,CORRUPTIONS): + y = np.load(os.path.join(dataroot, 'labels.npy')) + y_single = y[0:10000] + y_single = torch.tensor(y_single) + print("y.shape:",y.shape) + x = np.load(os.path.join(dataroot,CORRUPTIONS+'.npy')) + print("loading data of",os.path.join(dataroot,CORRUPTIONS+'.npy')) + x = x.transpose(0,3,1,2) + x = torch.tensor(x) + x = x.float()/255. + dataset = [] + for i in range(5): + x_single = x[i*10000:(i+1)*10000] + dataset.append(TensorDataset(x_single, y_single)) + return dataset + +def load_usps(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/usps-{split}.pkl' + if not os.path.exists(path): + dataset = USPS(f'{HOME}/.pytorch/USPS', train=(split=='train'), download=True) + x, y = dataset.data, dataset.targets + x = torch.tensor(resize_imgs(x, 32)) + x = (x.float()/255.).unsqueeze(1).repeat(1,3,1,1) + y = torch.tensor(y) + with open(path, 'wb') as f: + pickle.dump([x, y], f) + with open(path, 'rb') as f: + x, y = pickle.load(f) + if channels == 1: + x = x[:,0:1,:,:] + dataset = TensorDataset(x, y) + return dataset + +def load_svhn(split='train', channels=3): + dataset = SVHN(f'{HOME}/.pytorch/SVHN', split=split, download=True) + x, y = dataset.data, dataset.labels + x = x.astype('float32')/255. + x, y = torch.tensor(x), torch.tensor(y) + if channels == 1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + + +def load_syndigit(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/synth_{split}_32x32.mat' + data = loadmat(path) + x, y = data['X'], data['y'] + x = np.transpose(x, [3, 2, 0, 1]).astype('float32')/255. + y = y.squeeze() + x, y = torch.tensor(x), torch.tensor(y) + if channels == 1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + +def load_mnist_m(split='train', channels=3): + path = f'/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/mnist_m-{split}.pkl' + with open(path, 'rb') as f: + x, y = pickle.load(f) + x, y = torch.tensor(x.astype('float32')/255.), torch.tensor(y) + if channels==1: + x = x.mean(1, keepdim=True) + dataset = TensorDataset(x, y) + return dataset + +if __name__=='__main__': + dataset = load_mnist(split='train') + print('mnist train', len(dataset)) + dataset = load_mnist('test') + print('mnist test', len(dataset)) + dataset = load_mnist_m('test') + print('mnsit_m test', len(dataset)) + dataset = load_svhn(split='test') + print('svhn', len(dataset)) + dataset = load_usps(split='test') + print('usps', len(dataset)) + dataset = load_syndigit(split='test') + print('syndigit', len(dataset)) + diff --git a/Meta-causal/code/env.yaml b/Meta-causal/code/env.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b0bd424fb7c5aa818f10a82173549eb0dd3199c7 --- /dev/null +++ b/Meta-causal/code/env.yaml @@ -0,0 +1,119 @@ +name: Py3.7_torch1.8 +channels: + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ + - conda-forge + - bioconda + - defaults +dependencies: + - _libgcc_mutex=0.1=main + - asn1crypto=1.2.0=py37_0 + - blas=1.0=mkl + - bottleneck=1.3.2=py37heb32a55_1 + - bzip2=1.0.8=h7b6447c_0 + - ca-certificates=2021.10.8=ha878542_0 + - cairo=1.14.12=h8948797_3 + - certifi=2021.10.8=py37h89c1867_1 + - cffi=1.13.0=py37h2e261b9_0 + - chardet=3.0.4=py37_1003 + - click=8.0.3=pyhd3eb1b0_0 + - conda-package-handling=1.6.0=py37h7b6447c_0 + - cryptography=2.8=py37h1ba5d50_0 + - ffmpeg=4.0=hcdf2ecd_0 + - fontconfig=2.13.0=h9420a91_0 + - freeglut=3.0.0=hf484d3e_5 + - freetype=2.11.0=h70c0345_0 + - glib=2.63.1=h5a9c865_0 + - graphite2=1.3.14=h23475e2_0 + - h5py=2.8.0=py37h3010b51_1003 + - harfbuzz=1.8.8=hffaf4a1_0 + - hdf5=1.10.2=hba1933b_1 + - icu=58.2=he6710b0_3 + - idna=2.8=py37_0 + - intel-openmp=2021.3.0=h06a4308_3350 + - jasper=2.0.14=hd8c5072_2 + - jpeg=9d=h7f8727e_0 + - libedit=3.1.20181209=hc058e9b_0 + - libffi=3.2.1=hd88cf55_4 + - libgcc-ng=9.1.0=hdf63c60_0 + - libgfortran-ng=7.5.0=ha8ba4b0_17 + - libgfortran4=7.5.0=ha8ba4b0_17 + - libglu=9.0.0=hf484d3e_1 + - libopencv=3.4.2=hb342d67_1 + - libopus=1.3.1=h7b6447c_0 + - libpng=1.6.37=hbc83047_0 + - libprotobuf=3.17.2=h4ff587b_1 + - libstdcxx-ng=9.1.0=hdf63c60_0 + - libtiff=4.1.0=h2733197_0 + - libuuid=1.0.3=h7f8727e_2 + - libvpx=1.7.0=h439df22_0 + - libxcb=1.14=h7b6447c_0 + - libxml2=2.9.9=hea5a465_1 + - mkl=2021.3.0=h06a4308_520 + - mkl-service=2.4.0=py37h7f8727e_0 + - mkl_fft=1.3.1=py37hd3c417c_0 + - mkl_random=1.2.2=py37h51133e4_0 + - ncurses=6.1=he6710b0_1 + - numexpr=2.7.3=py37h22e1b3c_1 + - numpy-base=1.21.2=py37h79a1101_0 + - opencv=3.4.2=py37h6fd60c2_1 + - openssl=1.1.1h=h516909a_0 + - pandas=1.3.3=py37h8c16a72_0 + - pcre=8.45=h295c915_0 + - pip=19.3.1=py37_0 + - pixman=0.40.0=h7f8727e_1 + - protobuf=3.17.2=py37h295c915_0 + - py-opencv=3.4.2=py37hb342d67_1 + - pycosat=0.6.3=py37h14c3975_0 + - pycparser=2.19=py37_0 + - pyopenssl=19.0.0=py37_0 + - pysocks=1.7.1=py37_0 + - python=3.7.4=h265db76_1 + - python-dateutil=2.8.2=pyhd3eb1b0_0 + - python_abi=3.7=2_cp37m + - pytz=2021.3=pyhd3eb1b0_0 + - readline=7.0=h7b6447c_5 + - requests=2.22.0=py37_0 + - ruamel_yaml=0.15.46=py37h14c3975_0 + - scipy=1.7.1=py37h292c36d_2 + - setuptools=41.4.0=py37_0 + - six=1.12.0=py37_0 + - sqlite=3.30.0=h7b6447c_0 + - tensorboardx=2.2=pyhd3eb1b0_0 + - tk=8.6.8=hbc83047_0 + - tqdm=4.36.1=py_0 + - urllib3=1.24.2=py37_0 + - wheel=0.33.6=py37_0 + - xz=5.2.4=h14c3975_4 + - yaml=0.1.7=had09818_2 + - zlib=1.2.11=h7b6447c_3 + - zstd=1.3.7=h0b5b093_0 + - pip: + - absl-py==1.0.0 + - cachetools==4.2.4 + - conda-pack==0.6.0 + - google-auth==2.3.3 + - google-auth-oauthlib==0.4.6 + - grpcio==1.42.0 + - importlib-metadata==4.8.2 + - markdown==3.3.6 + - numpy==1.21.3 + - oauthlib==3.1.1 + - pillow==8.4.0 + - pyasn1==0.4.8 + - pyasn1-modules==0.2.8 + - requests-oauthlib==1.3.0 + - rsa==4.8 + - tensorboard==2.7.0 + - tensorboard-data-server==0.6.1 + - tensorboard-plugin-wit==1.8.0 + - torch==1.8.1+cu111 + - torchvision==0.9.1+cu111 + - typing-extensions==3.10.0.2 + - werkzeug==2.0.2 + - zipp==3.6.0 +prefix: /home/chenjin/miniconda3/envs/Py3.7_torch1.8 diff --git a/Meta-causal/code/main_my_joint_v13_auto.py b/Meta-causal/code/main_my_joint_v13_auto.py new file mode 100644 index 0000000000000000000000000000000000000000..c1b572ceb4a3b6aee91675c82c6c316757a92792 --- /dev/null +++ b/Meta-causal/code/main_my_joint_v13_auto.py @@ -0,0 +1,568 @@ + +''' +训练 base 模型 +''' + +import torch +import torch.nn as nn +import torch.nn.functional as F +import itertools +from torch import optim +from torch.utils.data import DataLoader, RandomSampler +from torchvision import models +from torchvision.datasets import CIFAR10 +from torchvision.utils import make_grid +import torchvision.transforms as transforms +from tensorboardX import SummaryWriter +from torch.cuda.amp import autocast,GradScaler + +import os +import click +import time +import numpy as np + +from network import mnist_net_my as mnist_net +from network import wideresnet as wideresnet +from network import resnet as resnet +from network import adaptor_v2 + +from tools import causalaugment_v3 as causalaugment +import data_loader_joint_v3 as data_loader +# from utils import set_requires_grad + +HOME = os.environ['HOME'] + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择gpu') +@click.option('--data', type=str, default='mnist', help='数据集名称') +@click.option('--ntr', type=int, default=None, help='选择训练集前ntr个样本') +@click.option('--translate', type=float, default=None, help='随机平移数据增强') +@click.option('--autoaug', type=str, default=None, help='AA FastAA RA') +@click.option('--n', type=int, default=3, help='选择多少个factor生成RA') +@click.option('--stride', type=int, default=5, help='if autoaug==CA_multiple, stride is used') +@click.option('--factor_num', type=int, default=16, help='the first n factors') +@click.option('--epochs', type=int, default=100) +@click.option('--nbatch', type=int, default=100, help='每个epoch中batch的数量') +@click.option('--batchsize', type=int, default=128, help='每个batch中样本的数量') +@click.option('--lr', type=float, default=1e-3) +@click.option('--lr_scheduler', type=str, default='none', help='是否选择学习率衰减策略') +@click.option('--svroot', type=str, default='./saved', help='项目文件保存路径') +@click.option('--clsadapt', type=bool, default=True, help='映射后是否用分类损失') +@click.option('--lambda_causal', type=float, default=1, help='the weight of reconstruction during mapping and causal ') +@click.option('--lambda_re', type=float, default=1, help='the weight of reconstruction during mapping and causal ') +@click.option('--randm', type=bool, default=True, help='m取值是否randm') +@click.option('--randn', type=bool, default=False, help='原始特征是否detach') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') +def experiment(gpu, data, ntr, translate, autoaug,n,stride, factor_num, epochs, nbatch, batchsize, lr, lr_scheduler, svroot, clsadapt, lambda_causal,lambda_re,randm,randn,network): + + settings = locals().copy() + print(settings) + + # 全局设置 + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + if not os.path.exists(svroot): + os.makedirs(svroot) + log_file = open(svroot+os.sep+'log.log',"w") + log_file.write(str(settings)+'\n') + writer = SummaryWriter(svroot) + CA = causalaugment.MultiCounterfactualAugment(factor_num,stride) + # FA = causalaugment.FactualAugment(m=4, factor_num=factor_num, randm=True) + # 加载数据集和模型 + if data in ['mnist', 'mnist_t']: + if data == 'mnist': + trset = data_loader.load_mnist('train', translate=translate,twox=True, ntr=ntr, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + elif data == 'mnist_t': + trset = data_loader.load_mnist_t('train', translate=translate, ntr=ntr) + teset = data_loader.load_mnist('test') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=0, \ + sampler=RandomSampler(trset, True, nbatch*batchsize)) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=0, shuffle=False) + cls_net = mnist_net.ConvNet().cuda() + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(1024,512,1024,2).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + parameter_list.append({'params':cls_net.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + #print("---------------------------------------------------------------------------------------") + opt = optim.Adam(parameter_list, lr=lr) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.95) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=int(epochs*0.8)) + # print("------------------------------------opt_mapping---------------------------------------------------") + # for param_group in opt_mapping.param_groups: + # print(param_group.keys()) + # # print(type(param_group)) + # print([type(value) for value in param_group.values()]) + # print('lr: ',param_group['lr']) + + # print("------------------------------------opt_causal---------------------------------------------------") + # for param_group in opt_causal.param_groups: + # print(param_group.keys()) + # # print(type(param_group)) + # print([type(value) for value in param_group.values()]) + # print('lr: ',param_group['lr']) + + elif data == 'cifar10': + # 加载数据集 + trset = data_loader.load_cifar10(split='train',twox=True, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + teset = data_loader.load_cifar10(split='test') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=4, shuffle=True, drop_last=True) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=4, shuffle=False) + cls_net = wideresnet.WideResNet(16, 10, 4).cuda() + # cls_opt = optim.SGD(cls_net.parameters(), lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(256,512,256,4).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + parameter_list.append({'params':cls_net.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + #print("---------------------------------------------------------------------------------------") + # opt = optim.Adam(parameter_list) + opt = optim.SGD(parameter_list, lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.95) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=int(epochs*0.8)) + elif data in ['art_painting', 'cartoon', 'photo', 'sketch']: + # 加载数据集 + trset = data_loader.load_pacs(domain=data, split='train', twox=True, factor_num=factor_num,autoaug=autoaug,randm=randm,randn=randn,n=n,stride=stride) + teset = data_loader.load_pacs(domain=data, split='val') + trloader = DataLoader(trset, batch_size=batchsize, num_workers=4, shuffle=True, drop_last=True) + teloader = DataLoader(teset, batch_size=batchsize, num_workers=4, shuffle=False) + if network == 'resnet18': + cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + input_dim = 2048 + # for param in cls_net.features.parameters(): + # param.requires_grad = False + # for name, parms in cls_net.named_parameters(): + # print('-->name:', name) + # print('-->grad_requirs:',parms.requires_grad) + # cls_opt = optim.SGD(cls_net.parameters(), lr=lr, momentum=0.9, nesterov=True, weight_decay=5e-4) + # print(cls_net.state_dict()) + + classifier_param = list(map(id, cls_net.class_classifier.parameters())) + backbone_param = filter(lambda p: id(p) not in classifier_param and p.requires_grad, cls_net.parameters()) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + mapping = adaptor_v2.mapping(input_dim,1024,input_dim,4).cuda() + AdaptNet.append(mapping) + parameter_list.append({'params':mapping.parameters(),'lr':lr}) + if autoaug == 'CA_multiple': + var_num = len(list(range(0, 31, stride))) + E_to_W = adaptor_v2.effect_to_weight(7,70,1).cuda() + else: + E_to_W = adaptor_v2.effect_to_weight(7,70,1).cuda() + parameter_list.append({'params':backbone_param,'lr':0.01*lr}) + parameter_list.append({'params':cls_net.class_classifier.parameters(),'lr':lr}) + parameter_list.append({'params':E_to_W.parameters(),'lr':lr}) + #print("---------------------------------------------------------------------------------------") + # opt = optim.Adam(parameter_list) + + opt = optim.SGD(parameter_list, momentum=0.9, nesterov=True, weight_decay=5e-4) + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(opt, epochs) + elif lr_scheduler == 'Exp': + scheduler = optim.lr_scheduler.ExponentialLR(opt, gamma=0.99999) + elif lr_scheduler == 'Step': + scheduler = optim.lr_scheduler.StepLR(opt, step_size=15) + elif 'synthia' in data: + # 加载数据集 + branch = data.split('_')[1] + trset = data_loader.load_synthia(branch) + trloader = DataLoader(trset, batch_size=batchsize, num_workers=8, shuffle=True) + teloader = DataLoader(trset, batch_size=batchsize, num_workers=8, shuffle=True) + imsize = [192, 320] + nclass = 14 + # 加载模型 + cls_net = fcn.FCN_resnet50(nclass=nclass).cuda() + cls_opt = optim.Adam(cls_net.parameters(), lr=lr)#, weight_decay=1e-4) # 对于synthia 加上weigh_decay会掉1-2个点 + if lr_scheduler == 'cosine': + scheduler = optim.lr_scheduler.CosineAnnealingLR(cls_opt, epochs*len(trloader)) + + cls_criterion = nn.CrossEntropyLoss() + adapt_criterion = nn.MSELoss() + # 开始训练 + best_acc = 0 + best_acc_t = 0 + scaler = GradScaler() + for epoch in range(epochs): + t1 = time.time() + loss_list = [] + cls_net.train() + # unloader = transforms.ToPILImage() + print(len(trloader)) + for i, (x_four,y) in enumerate(trloader): + b_sample_num = y.size(0) + x, x_RA, x_FA, x_CA, y = x_four[0].cuda(), x_four[1].cuda(), x_four[2].cuda(), x_four[3].cuda(), y.cuda() + b, c, h, w = x.shape + # x_FA_ = x_FA.transpose(1,2) + x_FA = x_FA.reshape(b*factor_num, c, h, w) + x_CA = x_CA.reshape(b*factor_num*var_num, c, h, w) + #learning mapping + y_repeat = y.unsqueeze(0).reshape(b_sample_num,1).repeat((1,factor_num)).reshape(1,b_sample_num*factor_num).squeeze() + # x_FA = FA(x).cuda().detach() + # x_CA = CA(x_RA).cuda().detach() + with autocast(): + p,f = cls_net(x) + # print("x.shape:",x.shape) + # print("x_FA.shape:",x_FA.shape) + _,f_FA = cls_net(x_FA) + p_RA,f_RA = cls_net(x_RA) + p_CA,_ = cls_net(x_CA) + # print("f.shape:",f.shape) + # print("f_FA.shape:",f_FA.shape) + #learning mapping + f_repeat = f.repeat((1,factor_num)).reshape(f_FA.shape) + f_adapt = torch.zeros(f_FA.shape).cuda() + for b in range(b_sample_num): + for j in range(factor_num): + f_adapt[b*factor_num+j] = AdaptNet[j](f_FA[b*factor_num+j]) + p_adapt = cls_net(f_adapt, mode='c') + + #learning causality + if autoaug == 'CA_multiple': + p_RA_repeat = p_RA.repeat((1,factor_num*var_num)).reshape(p_CA.shape) + effect_context = p_RA_repeat - p_CA + effect_context = effect_context.reshape(b_sample_num,factor_num,var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*factor_num,-1) + # print("effect_context.shape:",effect_context.shape) + else: + p_RA_repeat = p_RA.repeat((1,factor_num)).reshape(p_CA.shape) + effect_context = p_RA_repeat - p_CA + weight = E_to_W(effect_context) + # weight = E_to_W(effect_context.detach()) + weight = weight.reshape(b_sample_num,factor_num) + alphas = F.softmax(weight,dim=1) + + f_adapt_RA = torch.zeros(f_RA.shape).cuda() + for b in range(b_sample_num): + for j in range(factor_num): + f_adapt_RA[b] = f_adapt_RA[b]+ alphas[b,j]*AdaptNet[j](f_RA[b]) + p_adapt_RA = cls_net(f_adapt_RA, mode='c') + + cls_loss = cls_criterion(p, y) + re_mapping = adapt_criterion(f_adapt,f_repeat) + re_causal = adapt_criterion(f_adapt_RA,f) + cls_loss_mapping = cls_criterion(p_adapt, y_repeat) + cls_loss_causal = cls_criterion(p_adapt_RA, y) + + loss = cls_loss + cls_loss_mapping + lambda_re*re_mapping + lambda_causal*(lambda_re*re_causal + cls_loss_causal) + + opt.zero_grad() + scaler.scale(loss).backward() + scaler.step(opt) + scaler.update() + loss_list.append([cls_loss.item(), cls_loss_mapping.item(),cls_loss_causal.item(), re_mapping.item(), re_causal.item()]) + + # 调整学习率 + if lr_scheduler in ['cosine', 'Exp', 'Step']: + writer.add_scalar('scalar/lr', opt.param_groups[0]["lr"], epoch) + print(opt.param_groups[0]["lr"]) + print("changing lr") + scheduler.step() + cls_loss, cls_loss_mapping, cls_loss_causal, re_mapping, re_causal = np.mean(loss_list, 0) + + # 测试，并保存最优模型 + cls_net.eval() + if data in ['mnist', 'mnist_t', 'cifar10', 'mnistvis', 'art_painting', 'cartoon', 'photo', 'sketch']: + teacc = evaluate(cls_net, teloader) + + elif 'synthia' in data: + teacc = evaluate_seg(cls_net, teloader, nclass) # 这里算的其实是 miou + + if best_acc < teacc: + print(f'---------------------saving model at epoch {epoch}----------------------------------------------------') + log_file.write(f'saving model at epoch {epoch}\n') + + best_acc = teacc + torch.save(cls_net.state_dict(),os.path.join(svroot, 'best_cls_net.pkl')) + for j in range(factor_num): + torch.save(AdaptNet[j].state_dict(),os.path.join(svroot, 'best_mapping_'+str(j)+'.pkl')) + torch.save(E_to_W.state_dict(), os.path.join(svroot, 'best_E_to_W.pkl')) + + # 保存日志 + t2 = time.time() + print(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f} cls_loss_mapping {cls_loss_mapping:.4f} cls_loss_causal {cls_loss_causal:.4f} re_mapping {re_mapping:.4f} re_causal {re_causal:.4f} /// teacc {teacc:2.2f} lr {opt.param_groups[0]["lr"]:.8f}') + log_file.write(f'epoch {epoch}, time {t2-t1:.2f}, cls_loss {cls_loss:.4f} cls_loss_mapping {cls_loss_mapping:.4f} cls_loss_causal {cls_loss_causal:.4f} re_mapping {re_mapping:.4f} re_causal {re_causal:.4f} /// teacc {teacc:2.2f} lr {opt.param_groups[0]["lr"]:.8f} \n') + writer.add_scalar('scalar/cls_loss', cls_loss, epoch) + writer.add_scalar('scalar/cls_loss_mapping', cls_loss_mapping, epoch) + writer.add_scalar('scalar/cls_loss_causal', cls_loss_causal, epoch) + writer.add_scalar('scalar/re_mapping', re_mapping, epoch) + writer.add_scalar('scalar/re_causal', re_causal, epoch) + writer.add_scalar('scalar/teacc', teacc, epoch) + print(f'---------------------saving last model at epoch {epoch}----------------------------------------------------') + log_file.write(f'saving last model at epoch {epoch}\n') + torch.save(cls_net.state_dict(),os.path.join(svroot, 'last_cls_net.pkl')) + for j in range(factor_num): + torch.save(AdaptNet[j].state_dict(),os.path.join(svroot, 'last_mapping_'+str(j)+'.pkl')) + torch.save(E_to_W.state_dict(), os.path.join(svroot, 'last_E_to_W.pkl')) + + writer.close() +def evalute_pacs(source_domain,cls_net,CA,AdaptNet,E_to_W): + cls_net.eval() + data_total = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in data_total if i!=source_domain] + acc_CA = np.zeros(len(target)) + for idx, data in enumerate(target): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=6, num_workers=0) + # 计算评价指标 + acc_CA[idx] = evaluate_causal(cls_net, teloader, CA, AdaptNet, E_to_W) + acc_avg_CA = sum(acc_CA)/len(target) + return acc_avg_CA,acc_CA + + +def evaluate(net, teloader): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + with torch.no_grad(): + x1 = x1.cuda() + p1,_ = net(x1, mode='fc') + p1 = p1.argmax(dim=1) + ps.append(p1.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc +def extract_feature(net, teloader, savedir): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + img_class = y1[0].cpu().numpy() + save_path = os.path.join(savedir,str(img_class)) + if not os.path.exists(save_path): + os.makedirs(save_path) + + with torch.no_grad(): + x1 = x1.cuda() + p1,f1 = net(x1, mode='fc') + save_name = save_path+os.sep+str(i)+'.npy' + np.save(save_name,f1.cpu()) + p1 = p1.argmax(dim=1) + ps.append(p1.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +def evaluate_causal(net, teloader, CA, AdaptNet, E_to_W): + ps = [] + ys = [] + p_orig = [] + y_orig = [] + for i,(x1, y1) in enumerate(teloader): + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + p1,f_x1 = net(x1, mode='fc') + x1_CA = CA(x1).cuda() + p1_CA,_ = net(x1_CA, mode='fc') + p1_repeat = p1.repeat((1,CA.factor_num*CA.var_num)).reshape(p1_CA.shape) + effect_context = p1_repeat - p1_CA + effect_context = effect_context.reshape(b_sample_num,CA.factor_num,CA.var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*CA.factor_num,-1) + weight = E_to_W(effect_context) + weight = weight.reshape(b_sample_num,CA.factor_num) + alphas = F.softmax(weight,dim=1) + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(CA.factor_num): + f_adapt[b] = f_adapt[b]+ alphas[b,j]*AdaptNet[j](f_x1[b]) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + +def extract_feature_do(net, teloader, CA, AdaptNet, E_to_W, savedir_base, savedir,source_flag): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + img_class = y1[0].cpu().numpy() + save_path_base = os.path.join(savedir_base,str(img_class)) + save_path = os.path.join(savedir,str(img_class)) + if not os.path.exists(save_path_base): + os.makedirs(save_path_base) + if not os.path.exists(save_path): + os.makedirs(save_path) + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + p1,f_x1 = net(x1, mode='fc') + save_name_base = save_path_base+os.sep+str(i)+'_base.npy' + print(save_name_base) + np.save(save_name_base,f_x1.cpu()) + x1_CA = CA(x1).cuda() + p1_CA,_ = net(x1_CA, mode='fc') + p1_repeat = p1.repeat((1,CA.factor_num*CA.var_num)).reshape(p1_CA.shape) + effect_context = p1_repeat - p1_CA + effect_context = effect_context.reshape(b_sample_num,CA.factor_num,CA.var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*CA.factor_num,-1) + weight = E_to_W(effect_context) + weight = weight.reshape(b_sample_num,CA.factor_num) + alphas = F.softmax(weight,dim=1) + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(CA.factor_num): + f_adapt[b] = f_adapt[b]+ alphas[b,j]*AdaptNet[j](f_x1[b]) + if not source_flag: + save_name = save_path+os.sep+str(i)+'.npy' + print(save_name) + np.save(save_name,f_adapt.cpu()) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc + + +def evaluate_mapping(net, teloader, FA, AdaptNet, source=False): + correct, count = 0, 0 + ps = [] + ys = [] + pt = [] + yt = [] + factor_num = FA.factor_num + for j in range(factor_num): + ps.append([]) + ys.append([]) + pt.append([]) + yt.append([]) + ps.append([]) + ys.append([]) + # print(len(ps),len(ys)) + for i,(x1, y1) in enumerate(teloader): + with torch.no_grad(): + x1 = x1.cuda() + b = x1.size(0) + if source: + x_FA = FA(x1).cuda() + _, f = net(x_FA, mode='fc') + p,_ = net(x1, mode='fc') + p = p.argmax(dim=1) + ps[-1].append(p.detach().cpu().numpy()) + ys[-1].append(y1.numpy()) + else: + p, f = net(x1, mode='fc') + f = f.repeat((1,factor_num)).reshape((-1,f.size(1))) + p = p.argmax(dim=1) + ps[-1].append(p.detach().cpu().numpy()) + ys[-1].append(y1.numpy()) + for b_ in range(b): + for j in range(factor_num): + f_adapt = AdaptNet[j](f[b_*factor_num+j]) + #f_adapt = torch.mm(AdaptNet[j].W1,f_FA[b_*factor_num+j].unsqueeze(1)).squeeze() + p1 = net(f_adapt, mode='c') + p1 = p1.argmax(dim=0) + ps[j].append(p1.detach().cpu()) + ys[j].append(y1[b_]) + p1_t = net(f[b_*factor_num+j], mode='c') + # print("p1_t.shape:",p1_t.shape) + p1_t = p1_t.argmax(dim=0) + pt[j].append(p1_t.detach().cpu()) + yt[j].append(y1[b_]) + # 计算评价指标 + acc = np.zeros(factor_num+1) + acc_t = np.zeros(factor_num+1) + for j in range(factor_num): + pred = torch.stack(ps[j]) + label = torch.stack(ys[j]) + acc[j] = (pred==label).sum()/float(len(ys[j]))*100 + predt = torch.stack(pt[j]) + labelt = torch.stack(yt[j]) + acc_t[j] = (predt==labelt).sum()/float(len(yt[j]))*100 + pred = np.concatenate(ps[-1]) + label = np.concatenate(ys[-1]) + acc[-1] = np.mean(pred==label)*100 + # print("acc:",acc) + return acc, acc_t +def evaluate_causal_with_entropy(net, teloader, CA, AdaptNet): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + p1,f_x1 = net(x1, mode='fc') + + x1_CA = CA(x1).cuda() + p1_CA, _ = net(x1_CA, mode='fc') + p1_repeat = p1.repeat((1,CA.factor_num*CA.var_num)).reshape(p1_CA.shape) + effect_context = p1_repeat - p1_CA + effect_context = effect_context.reshape(b_sample_num,CA.factor_num,CA.var_num,-1) + effect_context = effect_context.mean(axis=2).reshape(b_sample_num*CA.factor_num,-1) + effect_context = F.softmax(effect_context,dim=1) + # weight = calc_ent(effect_context) + weight = torch.sum(-effect_context*(torch.log2(effect_context)),dim=1) + weight = weight.reshape(b_sample_num,CA.factor_num) + alphas = F.softmax(-weight,dim=1) + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(CA.factor_num): + f_adapt[b] = f_adapt[b]+ alphas[b,j]*AdaptNet[j](f_x1[b]) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc +def evaluate_causal_with_average(net, teloader, factor_num, AdaptNet): + ps = [] + ys = [] + for i,(x1, y1) in enumerate(teloader): + b_sample_num = x1.size(0) + with torch.no_grad(): + x1 = x1.cuda() + p1,f_x1 = net(x1, mode='fc') + f_adapt = torch.zeros(f_x1.shape).cuda() + for b in range(b_sample_num): + for j in range(factor_num): + f_adapt[b] = f_adapt[b]+ float(1/factor_num)*AdaptNet[j](f_x1[b]) + p_adapt = net(f_adapt, mode='c') + p_adapt = p_adapt.argmax(dim=1) + ps.append(p_adapt.detach().cpu().numpy()) + ys.append(y1.numpy()) + # 计算评价指标 + ps = np.concatenate(ps) + ys = np.concatenate(ys) + acc = np.mean(ys==ps)*100 + return acc +if __name__=='__main__': + experiment() \ No newline at end of file diff --git a/Meta-causal/code/main_test_digit_v13.py b/Meta-causal/code/main_test_digit_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..ffe683c3846ddf8ce34a4d834d089b6868e19dcc --- /dev/null +++ b/Meta-causal/code/main_test_digit_v13.py @@ -0,0 +1,143 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import mnist_net_my as mnist_net +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate,evaluate_causal,evaluate_causal_with_entropy,evaluate_mapping,evaluate_causal_with_average +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--channels', type=int, default=3) +@click.option('--factor_num', type=int, default=16) +@click.option('--stride', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--eval_mapping', type=bool, default=True, help='是否查看mapping学习效果') +def main(gpu, svroot, svpath, channels, factor_num,stride, epoch, eval_mapping): + evaluate_digit(gpu, svroot, svpath, channels, factor_num, stride,epoch, eval_mapping) + +def evaluate_digit(gpu, svroot, svpath, channels=3, factor_num=16,stride=5,epoch='best', eval_mapping=True): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if channels == 3: + cls_net = mnist_net.ConvNet().cuda() + elif channels == 1: + cls_net = mnist_net.ConvNet(imdim=channels).cuda() + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + cls_net.load_state_dict(saved_weight) + #cls_net.eval() + # 加载adaptation模型 + FA = causalaugment.FactualAugment(m=4, factor_num=factor_num) + CA = causalaugment.MultiCounterfactualAugment(factor_num,stride) + # Color_mapping = adaptor.mapping().cuda() + # Contrast_mapping = adaptor.mapping().cuda() + # Brightness_mapping = adaptor.mapping().cuda() + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) + # saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + mapping = adaptor_v2.mapping(1024,512,1024,2).cuda() + mapping.load_state_dict(saved_weight) + AdaptNet.append(mapping) + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) + + E_to_W = adaptor_v2.effect_to_weight(10,100,1).cuda() + # Color_mapping.load_state_dict(saved_weight['Color_mapping']) + # Contrast_mapping.load_state_dict(saved_weight['Contrast_mapping']) + # Brightness_mapping.load_state_dict(saved_weight['Brightness_mapping']) + # saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + E_to_W.load_state_dict(saved_weight) + + # 测试 + str2fun = { + 'mnist': data_loader.load_mnist, + 'mnist_m': data_loader.load_mnist_m, + 'usps': data_loader.load_usps, + 'svhn': data_loader.load_svhn, + 'syndigit': data_loader.load_syndigit, + } + columns = ['mnist', 'svhn', 'mnist_m', 'syndigit','usps'] + target = ['svhn', 'mnist_m', 'syndigit','usps'] + if eval_mapping: + index = FA.factor_list + index.append('w/o do (original x)') + else: + index = ['w/o do (original x)'] + index_ours = ['do'] + data_result = {} + data_result_ours = {} + cls_net.eval() + for idx, data in enumerate(columns): + teset = str2fun[data]('test', channels=channels) + teloader = DataLoader(teset, batch_size=8, num_workers=0) + # 计算评价指标 + acc_CA = evaluate_causal(cls_net, teloader, CA, AdaptNet, E_to_W) + data_result_ours[data] = acc_CA + #最后一维度是原始数据 + if eval_mapping: + if data == 'mnist': + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=True) + acc_avg = np.zeros(teacc_FA_aftermapping.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=False) + acc_avg = acc_avg + teacc_FA_aftermapping + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data]=teacc_FA_aftermapping + data_result[data+'_FA'] = acc_FA + else: + teacc = evaluate(cls_net, teloader) + if data == 'mnist': + acc_avg = np.zeros(teacc.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + acc_avg = acc_avg + teacc + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + acc_avg_CA = acc_avg_CA/float(len(target)) + + data_result['Avg'] = acc_avg + data_result_ours['Avg'] = acc_avg_CA + + df = pd.DataFrame(data_result,index = index) + df_ours = pd.DataFrame(data_result_ours,index = index_ours) + print(df) + print(df_ours) + if svpath is not None: + df.to_csv(svpath) + df_ours.to_csv(svpath, mode='a') + +if __name__=='__main__': + main() + diff --git a/Meta-causal/code/main_test_pacs_v13.py b/Meta-causal/code/main_test_pacs_v13.py new file mode 100644 index 0000000000000000000000000000000000000000..e671f80903d98050eee7ea006ccc3abfdd2c5f44 --- /dev/null +++ b/Meta-causal/code/main_test_pacs_v13.py @@ -0,0 +1,139 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.utils.data import DataLoader + +import os +import numpy as np +import click +import pandas as pd + +from network import resnet as resnet +from network import adaptor_v2 +from tools import causalaugment_v3 as causalaugment +from main_my_joint_v13_auto import evaluate,evaluate_causal,evaluate_causal_with_entropy,evaluate_mapping,evaluate_causal_with_average +import data_loader_joint_v3 as data_loader + +@click.command() +@click.option('--gpu', type=str, default='0', help='选择GPU编号') +@click.option('--svroot', type=str, default='./saved') +@click.option('--source_domain', type=str, default='art_painting', help='source domain') +@click.option('--svpath', type=str, default=None, help='保存日志的路径') +@click.option('--factor_num', type=int, default=16) +@click.option('--epoch', type=str, default='best') +@click.option('--stride', type=int, default=5) +@click.option('--eval_mapping', type=bool, default=False, help='是否查看mapping学习效果') +@click.option('--network', type=str, default='resnet18', help='项目文件保存路径') +def main(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network): + evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num, epoch, stride,eval_mapping, network) + +def evaluate_pacs(gpu, svroot, source_domain, svpath, factor_num=16, epoch='best', stride=5,eval_mapping=False, network='resnet18'): + settings = locals().copy() + print(settings) + os.environ['CUDA_VISIBLE_DEVICES'] = gpu + + # 加载分类模型 + if network == 'resnet18': + cls_net = resnet.resnet18(classes=7,c_dim=2048).cuda() + input_dim = 2048 + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_cls_net.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_cls_net.pkl')) + cls_net.load_state_dict(saved_weight) + cls_net.eval() + # 加载adaptation模型 + FA = causalaugment.FactualAugment(m=4, factor_num=factor_num) + CA = causalaugment.MultiCounterfactualAugment(factor_num,stride) + AdaptNet = [] + parameter_list = [] + for i in range(factor_num): + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_mapping_'+str(i)+'.pkl')) + # saved_weight = torch.load(os.path.join(svroot, 'best_mapping_'+str(i)+'.pkl')) + mapping = adaptor_v2.mapping(input_dim,1024,input_dim,4).cuda() + mapping.load_state_dict(saved_weight) + AdaptNet.append(mapping) + if epoch == 'best': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + elif epoch == 'last': + print("loading weight of %s"%(epoch)) + saved_weight = torch.load(os.path.join(svroot, 'last_E_to_W.pkl')) + E_to_W = adaptor_v2.effect_to_weight(7,70,1).cuda() + # Color_mapping.load_state_dict(saved_weight['Color_mapping']) + # Contrast_mapping.load_state_dict(saved_weight['Contrast_mapping']) + # Brightness_mapping.load_state_dict(saved_weight['Brightness_mapping']) + # saved_weight = torch.load(os.path.join(svroot, 'best_E_to_W.pkl')) + E_to_W.load_state_dict(saved_weight) + + # 测试 + # str2fun = { + # 'art_painting': data_loader.load_pacs, + # 'cartoon': data_loader.load_pacs, + # 'photo': data_loader.load_pacs, + # 'sketch': data_loader.load_pacs, + # } + columns = ['art_painting', 'cartoon', 'photo', 'sketch'] + target = [i for i in columns if i!=source_domain] + columns = [source_domain] + target + print("columns:",columns) + if eval_mapping: + index = FA.factor_list + index.append('w/o do (original x)') + else: + index = ['w/o do (original x)'] + index_ours = ['do'] + data_result = {} + data_result_ours = {} + + for idx, data in enumerate(columns): + teset = data_loader.load_pacs(data, 'test') + teloader = DataLoader(teset, batch_size=4, num_workers=0) + # 计算评价指标 + acc_CA = evaluate_causal(cls_net, teloader, CA, AdaptNet, E_to_W) + data_result_ours[data] = acc_CA + #最后一维度是原始数据 + if eval_mapping: + if data == source_domain: + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=True) + acc_avg = np.zeros(teacc_FA_aftermapping.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + teacc_FA_aftermapping, acc_FA = evaluate_mapping(cls_net, teloader, FA, AdaptNet, source=False) + acc_avg = acc_avg + teacc_FA_aftermapping + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data]=teacc_FA_aftermapping + data_result[data+'_FA'] = acc_FA + else: + teacc = evaluate(cls_net, teloader) + if data == source_domain: + acc_avg = np.zeros(teacc.shape) + acc_avg_CA = np.zeros(acc_CA.shape) + else: + acc_avg = acc_avg + teacc + acc_avg_CA = acc_avg_CA + acc_CA + data_result[data] = teacc + acc_avg = acc_avg/float(len(target)) + acc_avg_CA = acc_avg_CA/float(len(target)) + + data_result['Avg'] = acc_avg + data_result_ours['Avg'] = acc_avg_CA + + df = pd.DataFrame(data_result,index = index) + df_ours = pd.DataFrame(data_result_ours,index = index_ours) + print(df) + print(df_ours) + if svpath is not None: + df.to_csv(svpath) + df_ours.to_csv(svpath, mode='a') +if __name__=='__main__': + main() + diff --git a/Meta-causal/code/network/adaptor_v2.py b/Meta-causal/code/network/adaptor_v2.py new file mode 100644 index 0000000000000000000000000000000000000000..ce47dbd1a24f9e2f741d8a82061b62b86d3dba41 --- /dev/null +++ b/Meta-causal/code/network/adaptor_v2.py @@ -0,0 +1,63 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F +import numpy as np + +class mapping(nn.Module): + def __init__(self, input_dim=1024, hidden_dim = 512, out_dim=1024, layernum=4): + ''' + ''' + super().__init__() + self.layernum = layernum + if layernum == 4: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim) + self.fc3 = nn.Linear(hidden_dim, hidden_dim) + self.fc4 = nn.Linear(hidden_dim, out_dim) + elif layernum == 2: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, out_dim) + self.relu = nn.ReLU(inplace=True) + + def forward(self, x): + ''' x ''' + if self.layernum == 4: + x = self.relu(self.fc1(x)) + x = self.relu(self.fc2(x)) + x = self.relu(self.fc3(x)) + x = self.fc4(x) + elif self.layernum == 2: + x = self.relu(self.fc1(x)) + x = self.fc2(x) + return x + + +class effect_to_weight(nn.Module): + def __init__(self, input_dim = 512, hidden_dim = 256, out_dim = 1, layernum=2, hidden_dim2 = 128): + ''' + ''' + super().__init__() + + self.layernum = layernum + if layernum == 2: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, out_dim) + elif layernum == 3: + self.fc1 = nn.Linear(input_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim2) + self.fc3 = nn.Linear(hidden_dim2, out_dim) + self.relu = nn.ReLU(inplace=True) + + def forward(self, x): + ''' x ''' + if self.layernum == 2: + x = self.relu(self.fc1(x)) + x = self.fc2(x) + else: + x = self.relu(self.fc1(x)) + x = self.relu(self.fc2(x)) + x = self.fc3(x) + return x + + diff --git a/Meta-causal/code/network/mnist_net_my.py b/Meta-causal/code/network/mnist_net_my.py new file mode 100644 index 0000000000000000000000000000000000000000..15e2e677280fdd2211b559f9f1bafd2fb66b5ef4 --- /dev/null +++ b/Meta-causal/code/network/mnist_net_my.py @@ -0,0 +1,104 @@ + +import torch +import torch.nn as nn +import torch.nn.functional as F + +class ConvNet(nn.Module): + ''' 网络结构和cvpr2020的 M-ADA 方法一致 ''' + def __init__(self, imdim=3): + super(ConvNet, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 1024) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(1024, 10) + # self.cls_head_tgt = nn.Linear(1024, 10) + # self.pro_head = nn.Linear(1024, 128) + + def forward(self, x, mode='fc'): + if mode == 'c': + out4 = self.relu4(x) + p = self.cls_head_src(out4) + return p + elif mode == 'fc': + in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + out2 = self.mp(self.relu2(self.conv2(out1))) + out2 = out2.view(in_size, -1) + out3 = self.relu3(self.fc1(out2)) + out4_worelu = self.fc2(out3) + out4 = self.relu4(out4_worelu) + p = self.cls_head_src(out4) + return p, out4_worelu + + # if mode == 'test': + # p = self.cls_head_src(out4) + # return p + # elif mode == 'train': + # p = self.cls_head_src(out4) + # # z = self.pro_head(out4) + # # z = F.normalize(z) + # return p,out4_worelu + # elif mode == 'p_f': + # p = self.cls_head_src(out4) + # return p, out4 + #elif mode == 'target': + # p = self.cls_head_tgt(out4) + # z = self.pro_head(out4) + # z = F.normalize(z) + # return p,z + +class ConvNetVis(nn.Module): + ''' 方便可视化，特征提取器输出2-d特征 + ''' + def __init__(self, imdim=3): + super(ConvNetVis, self).__init__() + + self.conv1 = nn.Conv2d(imdim, 64, kernel_size=5, stride=1, padding=0) + self.mp = nn.MaxPool2d(2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(64, 128, kernel_size=5, stride=1, padding=0) + self.relu2 = nn.ReLU(inplace=True) + self.fc1 = nn.Linear(128*5*5, 1024) + self.relu3 = nn.ReLU(inplace=True) + self.fc2 = nn.Linear(1024, 2) + self.relu4 = nn.ReLU(inplace=True) + + self.cls_head_src = nn.Linear(2, 10) + self.cls_head_tgt = nn.Linear(2, 10) + self.pro_head = nn.Linear(2, 128) + + def forward(self, x, mode='test'): + + in_size = x.size(0) + out1 = self.mp(self.relu1(self.conv1(x))) + out2 = self.mp(self.relu2(self.conv2(out1))) + out2 = out2.view(in_size, -1) + out3 = self.relu3(self.fc1(out2)) + out4 = self.relu4(self.fc2(out3)) + + if mode == 'test': + p = self.cls_head_src(out4) + return p + elif mode == 'train': + p = self.cls_head_src(out4) + z = self.pro_head(out4) + z = F.normalize(z) + return p,z + elif mode == 'p_f': + p = self.cls_head_src(out4) + return p, out4 + #elif mode == 'target': + # p = self.cls_head_tgt(out4) + # z = self.pro_head(out4) + # z = F.normalize(z) + # return p,z + + diff --git a/Meta-causal/code/network/resnet.py b/Meta-causal/code/network/resnet.py new file mode 100644 index 0000000000000000000000000000000000000000..925410b6cc064aba01d1f86efa8eb7fdd592ecee --- /dev/null +++ b/Meta-causal/code/network/resnet.py @@ -0,0 +1,101 @@ +from torch import nn +from torch.utils import model_zoo +#from torchvision.models.resnet import BasicBlock, model_urls, Bottleneck +from torchvision.models.resnet import BasicBlock, Bottleneck + +import torch +import ssl +# from torch import nn as nn +# from utils.util import * + +ssl._create_default_https_context = ssl._create_unverified_context + +all = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101','resnet152'] + +model_urls = { +'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', +'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', +'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', +'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', +'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth', +} + + +class ResNet(nn.Module): + def __init__(self, block, layers,classes=7,c_dim=512): + self.inplanes = 64 + super(ResNet, self).__init__() + self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, + bias=False) + self.bn1 = nn.BatchNorm2d(64) + self.relu = nn.ReLU(inplace=True) + self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + self.layer1 = self._make_layer(block, 64, layers[0]) + self.layer2 = self._make_layer(block, 128, layers[1], stride=2) + self.layer3 = self._make_layer(block, 256, layers[2], stride=2) + self.layer4 = self._make_layer(block, 512, layers[3], stride=2) + self.avgpool = nn.AvgPool2d(7, stride=1) + self.class_classifier = nn.Linear(c_dim, classes) + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + nn.init.constant_(m.weight, 1) + nn.init.constant_(m.bias, 0) + + def _make_layer(self, block, planes, blocks, stride=1): + downsample = None + if stride != 1 or self.inplanes != planes * block.expansion: + downsample = nn.Sequential( + nn.Conv2d(self.inplanes, planes * block.expansion, + kernel_size=1, stride=stride, bias=False), + nn.BatchNorm2d(planes * block.expansion), + ) + + layers = [] + layers.append(block(self.inplanes, planes, stride, downsample)) + self.inplanes = planes * block.expansion + for i in range(1, blocks): + layers.append(block(self.inplanes, planes)) + + return nn.Sequential(*layers) + def forward(self, x, mode='fc'): + if mode == 'c': + return self.class_classifier(x) + else: + x = self.conv1(x) + x = self.bn1(x) + x = self.relu(x) + x = self.maxpool(x) + + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + x = self.avgpool(x) + x = x.view(x.size(0), -1) + # print("x.shape:",x.shape) + return self.class_classifier(x), x + + +def resnet18(pretrained=True, **kwargs): + """Constructs a ResNet-18 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet18']), strict=False) + return model + +def resnet50(pretrained=True, **kwargs): + """Constructs a ResNet-50 model. + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + """ + model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs) + if pretrained: + print("-------------------------------------loading pretrain weights----------------------------------") + model.load_state_dict(model_zoo.load_url(model_urls['resnet50']), strict=False) + return model diff --git a/Meta-causal/code/network/wideresnet.py b/Meta-causal/code/network/wideresnet.py new file mode 100644 index 0000000000000000000000000000000000000000..1ca130a5f278c3b63f43b589db6ebd18d6e91593 --- /dev/null +++ b/Meta-causal/code/network/wideresnet.py @@ -0,0 +1,86 @@ +import math +import torch +import torch.nn as nn +import torch.nn.functional as F + + +class BasicBlock(nn.Module): + def __init__(self, in_planes, out_planes, stride, dropRate=0.0): + super(BasicBlock, self).__init__() + self.bn1 = nn.BatchNorm2d(in_planes) + self.relu1 = nn.ReLU(inplace=True) + self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, + padding=1, bias=False) + self.bn2 = nn.BatchNorm2d(out_planes) + self.relu2 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=3, stride=1, + padding=1, bias=False) + self.droprate = dropRate + self.equalInOut = (in_planes == out_planes) + self.convShortcut = (not self.equalInOut) and nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, + padding=0, bias=False) or None + def forward(self, x): + if not self.equalInOut: + x = self.relu1(self.bn1(x)) + else: + out = self.relu1(self.bn1(x)) + out = self.relu2(self.bn2(self.conv1(out if self.equalInOut else x))) + if self.droprate > 0: + out = F.dropout(out, p=self.droprate, training=self.training) + out = self.conv2(out) + return torch.add(x if self.equalInOut else self.convShortcut(x), out) + +class NetworkBlock(nn.Module): + def __init__(self, nb_layers, in_planes, out_planes, block, stride, dropRate=0.0): + super(NetworkBlock, self).__init__() + self.layer = self._make_layer(block, in_planes, out_planes, nb_layers, stride, dropRate) + def _make_layer(self, block, in_planes, out_planes, nb_layers, stride, dropRate): + layers = [] + for i in range(int(nb_layers)): + layers.append(block(i == 0 and in_planes or out_planes, out_planes, i == 0 and stride or 1, dropRate)) + return nn.Sequential(*layers) + def forward(self, x): + return self.layer(x) + +class WideResNet(nn.Module): + def __init__(self, depth, num_classes, widen_factor=1, dropRate=0.0): + super(WideResNet, self).__init__() + nChannels = [16, 16*widen_factor, 32*widen_factor, 64*widen_factor] + assert((depth - 4) % 6 == 0) + n = (depth - 4) / 6 + block = BasicBlock + # 1st conv before any network block + self.conv1 = nn.Conv2d(3, nChannels[0], kernel_size=3, stride=1, + padding=1, bias=False) + # 1st block + self.block1 = NetworkBlock(n, nChannels[0], nChannels[1], block, 1, dropRate) + # 2nd block + self.block2 = NetworkBlock(n, nChannels[1], nChannels[2], block, 2, dropRate) + # 3rd block + self.block3 = NetworkBlock(n, nChannels[2], nChannels[3], block, 2, dropRate) + # global average pooling and classifier + self.bn1 = nn.BatchNorm2d(nChannels[3]) + self.relu = nn.ReLU(inplace=True) + self.fc = nn.Linear(nChannels[3], num_classes) + self.nChannels = nChannels[3] + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, nn.BatchNorm2d): + m.weight.data.fill_(1) + m.bias.data.zero_() + elif isinstance(m, nn.Linear): + m.bias.data.zero_() + def forward(self, x, mode='fc'): + if mode == 'c': + return self.fc(x) + else: + out = self.conv1(x) + out = self.block1(out) + out = self.block2(out) + out = self.block3(out) + out = self.relu(self.bn1(out)) + out = F.avg_pool2d(out, 8) + out = out.view(-1, self.nChannels) + return self.fc(out), out diff --git a/Meta-causal/code/run_PACS/run_my_joint_v13_test.sh b/Meta-causal/code/run_PACS/run_my_joint_v13_test.sh new file mode 100644 index 0000000000000000000000000000000000000000..4cc6202fbc2602b146fd3ec25521a7383e60248a --- /dev/null +++ b/Meta-causal/code/run_PACS/run_my_joint_v13_test.sh @@ -0,0 +1,39 @@ + +# $1 gpuid +# $2 runid + +# base方法 +cd .. +epochs=70 +clsadapt=True +lr=0.01 +factor_num=16 +lr_scheduler=cosine +lambda_causal=1 +lambda_re=1 +batchsize=6 +stride=5 +randm=True +randn=True +autoaug=CA_multiple +network=resnet18 +UniqueExpName=ReProduceMetaCausal + +root=/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS/ +#data=art_painting +#data=cartoon +#data=photo +data=sketch +svroot=$root/${data}/${autoaug}_${factor_num}fa_v2_ep${epochs}_lr${lr}_${lr_scheduler}_base0.01_bs${batchsize}_lamCa_${lambda_causal}_lamRe${lambda_re}_adt4_cls1_EW2_70_rm${randm}_rn${randn}_str${stride}_${UniqueExpName} + +#python3 main_my_joint_v13_auto.py --gpu $1 --data ${data} --epochs $epochs --autoaug $autoaug --lambda_causal ${lambda_causal} --lambda_re ${lambda_re} --lr $lr --svroot $svroot --clsadapt $clsadapt --factor_num $factor_num --lr_scheduler ${lr_scheduler} --batchsize ${batchsize} --network ${network} --randm ${randm} --randn ${randn} --stride ${stride} + +test_epoch=best +python3 main_test_pacs_v13.py --gpu $1 --source_domain $data --svroot $svroot --svpath $svroot/${data}_${factor_num}factor_${test_epoch}_test_check.csv --factor_num $factor_num --epoch $test_epoch \ + --network ${network} --stride ${stride} + + + + + + diff --git a/Meta-causal/code/run_digits/run_my_joint_test.sh b/Meta-causal/code/run_digits/run_my_joint_test.sh new file mode 100644 index 0000000000000000000000000000000000000000..93d3b0700388d4d274f60f805669b9a559cb6e38 --- /dev/null +++ b/Meta-causal/code/run_digits/run_my_joint_test.sh @@ -0,0 +1,35 @@ + +# $1 gpuid + +cd .. +epochs=500 +clsadapt=True +lr=1e-4 +lr_scheduler=Step +factor_num=14 +#test_epoch=best +test_epoch=last +lambda_causal=1 +lambda_re=1 +batchsize=32 +stride=3 +randm=True +randn=True +autoaug=CA_multiple +UniqueExpName=ReProduceMetaCausal + + +root=/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit +svroot=$root/${autoaug}_${factor_num}fa_all_ep${epochs}_lr${lr}_lr_scheduler${lr_scheduler}0.8_bs${batchsize}_lamCa_${lambda_causal}_lamRe_${lambda_re}_cls1_adt2_EW2_100_rm${randm}_rn${randn}_str${stride}_${UniqueExpName} + +#python3 main_my_joint_v13_auto.py --gpu $1 --data mnist --epochs $epochs --autoaug $autoaug --lambda_causal ${lambda_causal} --lambda_re ${lambda_re} --lr $lr --lr_scheduler $lr_scheduler --svroot $svroot --clsadapt $clsadapt --factor_num $factor_num --batchsize ${batchsize} --randm ${randm} --randn ${randn} --stride ${stride} + +python3 main_test_digit_v13.py --gpu $1 --svroot $svroot --svpath $svroot/${factor_num}factor_${test_epoch}.csv --factor_num $factor_num --epoch $test_epoch \ + --stride ${stride} + + + + + + + diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_last_test_check.csv b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_last_test_check.csv new file mode 100644 index 0000000000000000000000000000000000000000..0770500f310525fc23467dd4e63d812f74b78ffc --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/art_painting_16factor_last_test_check.csv @@ -0,0 +1,4 @@ +,art_painting,cartoon,photo,sketch,Avg +w/o do (original x),98.92578125,67.61945392491468,94.67065868263474,71.69763298549249,77.99591519768063 +,art_painting,cartoon,photo,sketch,Avg +do,99.0234375,65.61433447098976,95.1497005988024,70.65411046067702,77.13938184348973 diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_E_to_W.pkl b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_E_to_W.pkl new file mode 100644 index 0000000000000000000000000000000000000000..ce7963c5715838bd6cd4e5ce703ba15c045ad50e --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_E_to_W.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c358d11a7d411a03dd882a7c61ba3fdf2756b8d6103f4ce417af8ba547d72c28 +size 4243 diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_cls_net.pkl b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_cls_net.pkl new file mode 100644 index 0000000000000000000000000000000000000000..8d63f5c6bc13fb241bef0a5170ea3106c7ccca25 --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_cls_net.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28f92eb28b9b902797d43492e74605652ad7c7db42b65d66d9b94237ab103fbd +size 44844729 diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_0.pkl b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_0.pkl new file mode 100644 index 0000000000000000000000000000000000000000..b270f1c6b36a77be4740368618b2da3bb101f268 --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_0.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1abf2abe4e5df3d56315d11f0b59904ef92af11213b51fd9a17f244e994ae610 +size 25189303 diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_1.pkl b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_1.pkl new file mode 100644 index 0000000000000000000000000000000000000000..e6d94a20ce0588bbbd1fb84e433233ac73173c7d --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_1.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:089f78530f6f4a4c505e96ce81f7f29570c848aa8c56341b5bdf4e07793a3c8a +size 25189303 diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_10.pkl b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_10.pkl new file mode 100644 index 0000000000000000000000000000000000000000..1e55dd9203712d4ea8d491df3cc3361b75913720 --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_10.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c4f047a7dafc4c491679a156df62bd8e701923fd4a2b5acf00260061604450e +size 25189303 diff --git a/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_11.pkl b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_11.pkl new file mode 100644 index 0000000000000000000000000000000000000000..6ef2aa4f01320ff85eda149380b2fa451d4dd8c8 --- /dev/null +++ b/Meta-causal/code/run_digits/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/last_mapping_11.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44d0650c82b3c77494af2714c74ddfb9604c17e1e57756bf9bb555bbddc0fde1 +size 11534336 diff --git a/Meta-causal/code/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala b/Meta-causal/code/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala new file mode 100644 index 0000000000000000000000000000000000000000..aa44ae0c513b57a8501e9bb1af27dc442b72f7d7 --- /dev/null +++ b/Meta-causal/code/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/events.out.tfevents.1719926752.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a44a49f0a1b3c59b9763c67ea85708ef8b56cae5fe4336f0383f5f71ba0dac84 +size 40 diff --git a/Meta-causal/code/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log b/Meta-causal/code/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log new file mode 100644 index 0000000000000000000000000000000000000000..f26feaaef352ae5821e49b7fbc4b1720f8735f38 --- /dev/null +++ b/Meta-causal/code/saved-PACS/art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5/log.log @@ -0,0 +1 @@ +{'gpu': '0', 'data': 'art_painting', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': 'saved-PACS//art_painting/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} diff --git a/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala new file mode 100644 index 0000000000000000000000000000000000000000..5ec21b3afdf0e11651cc768f4f55ea6269b887f5 --- /dev/null +++ b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925086.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7946f93077ec2136f75fc090a5762ce810be71cc78d5201e8a671217a678c563 +size 40 diff --git a/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala new file mode 100644 index 0000000000000000000000000000000000000000..620f9ba109e77ed90b7676c138933f814245e7f1 --- /dev/null +++ b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925314.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2021c61739fbe1f9c066067b4e5903d8d2d6c1c44865e1e9c61449eb3d90327 +size 40 diff --git a/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala new file mode 100644 index 0000000000000000000000000000000000000000..3144b1448112cff1aa0c26e0d825b50698f41d65 --- /dev/null +++ b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/events.out.tfevents.1719925652.hala @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fafb4b17d350157735eb6174ff44bafcea7ab8bf86948df3421447ef45ffcae3 +size 40 diff --git a/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log new file mode 100644 index 0000000000000000000000000000000000000000..f4c211545f0d2b537d3dcf980579f604a33419a7 --- /dev/null +++ b/Meta-causal/code/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3/log.log @@ -0,0 +1 @@ +{'gpu': '0çç', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': 'saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} diff --git a/Meta-causal/code/submit_digits.sh b/Meta-causal/code/submit_digits.sh new file mode 100644 index 0000000000000000000000000000000000000000..7a1824ca5314d87ecb2eb8caa14e677e07fbd37e --- /dev/null +++ b/Meta-causal/code/submit_digits.sh @@ -0,0 +1,22 @@ +#!/bin/bash +#SBATCH --job-name=metatrainRN +#SBATCH --nodes=1 # Request 1 node +#SBATCH --ntasks=1 # Number of tasks (total) +#SBATCH --cpus-per-task=8 # Number of CPU cores (threads) per task +#SBATCH --mem-per-cpu=4G # Memory limit per CPU core (there is no --mem-per-task) +#SBATCH --time=36:00:00 # Job timeout +#SBATCH --gpus-per-node=l4-24g:1 +#SBATCH --nodelist=gcpl4-eu-1 +#SBATCH --output=%j.log # Redirect stdout to a log file +#SBATCH --error=%j.error # Redirect stderr to a separate error log file + +srun --nodes 1 --ntasks-per-node 1 -- \ +mkenv -f ../env_mc.yml -- \ +sh -c "cd run_digits +bash run_my_joint_test.sh 0 +" + + + + + diff --git a/Meta-causal/code/submit_pacs.sh b/Meta-causal/code/submit_pacs.sh new file mode 100644 index 0000000000000000000000000000000000000000..9625e061ddd513d6896b9e7f1f735aab3e4b1ce1 --- /dev/null +++ b/Meta-causal/code/submit_pacs.sh @@ -0,0 +1,22 @@ +#!/bin/bash +#SBATCH --job-name=metatrainRN +#SBATCH --nodes=1 # Request 1 node +#SBATCH --ntasks=1 # Number of tasks (total) +#SBATCH --cpus-per-task=8 # Number of CPU cores (threads) per task +#SBATCH --mem-per-cpu=4G # Memory limit per CPU core (there is no --mem-per-task) +#SBATCH --time=36:00:00 # Job timeout +#SBATCH --gpus-per-node=l4-24g:1 +#SBATCH --nodelist=gcpl4-eu-1 +#SBATCH --output=%j.log # Redirect stdout to a log file +#SBATCH --error=%j.error # Redirect stderr to a separate error log file + +srun --nodes 1 --ntasks-per-node 1 -- \ +mkenv -f ../env_mc.yml -- \ +sh -c "cd run_PACS +bash run_my_joint_v13_test.sh 0 +" + + + + + diff --git a/Meta-causal/code/tools/autoaugment.py b/Meta-causal/code/tools/autoaugment.py new file mode 100644 index 0000000000000000000000000000000000000000..76c6bc4ebd5c59b76a58a8dca196f22d41fbf114 --- /dev/null +++ b/Meta-causal/code/tools/autoaugment.py @@ -0,0 +1,234 @@ +from PIL import Image, ImageEnhance, ImageOps +import numpy as np +import random + + +class ImageNetPolicy(object): + """ Randomly choose one of the best 24 Sub-policies on ImageNet. + + Example: + >>> policy = ImageNetPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> ImageNetPolicy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.4, "posterize", 8, 0.6, "rotate", 9, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "posterize", 7, 0.6, "posterize", 6, fillcolor), + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + + SubPolicy(0.4, "equalize", 4, 0.8, "rotate", 8, fillcolor), + SubPolicy(0.6, "solarize", 3, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.8, "posterize", 5, 1.0, "equalize", 2, fillcolor), + SubPolicy(0.2, "rotate", 3, 0.6, "solarize", 8, fillcolor), + SubPolicy(0.6, "equalize", 8, 0.4, "posterize", 6, fillcolor), + + SubPolicy(0.8, "rotate", 8, 0.4, "color", 0, fillcolor), + SubPolicy(0.4, "rotate", 9, 0.6, "equalize", 2, fillcolor), + SubPolicy(0.0, "equalize", 7, 0.8, "equalize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + + SubPolicy(0.8, "rotate", 8, 1.0, "color", 2, fillcolor), + SubPolicy(0.8, "color", 8, 0.8, "solarize", 7, fillcolor), + SubPolicy(0.4, "sharpness", 7, 0.6, "invert", 8, fillcolor), + SubPolicy(0.6, "shearX", 5, 1.0, "equalize", 9, fillcolor), + SubPolicy(0.4, "color", 0, 0.6, "equalize", 3, fillcolor), + + SubPolicy(0.4, "equalize", 7, 0.2, "solarize", 4, fillcolor), + SubPolicy(0.6, "solarize", 5, 0.6, "autocontrast", 5, fillcolor), + SubPolicy(0.6, "invert", 4, 1.0, "equalize", 8, fillcolor), + SubPolicy(0.6, "color", 4, 1.0, "contrast", 8, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.6, "equalize", 3, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment ImageNet Policy" + + +class CIFAR10Policy(object): + """ Randomly choose one of the best 25 Sub-policies on CIFAR10. + + Example: + >>> policy = CIFAR10Policy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> CIFAR10Policy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.1, "invert", 7, 0.2, "contrast", 6, fillcolor), + SubPolicy(0.7, "rotate", 2, 0.3, "translateX", 9, fillcolor), + SubPolicy(0.8, "sharpness", 1, 0.9, "sharpness", 3, fillcolor), + SubPolicy(0.5, "shearY", 8, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.5, "autocontrast", 8, 0.9, "equalize", 2, fillcolor), + + SubPolicy(0.2, "shearY", 7, 0.3, "posterize", 7, fillcolor), + SubPolicy(0.4, "color", 3, 0.6, "brightness", 7, fillcolor), + SubPolicy(0.3, "sharpness", 9, 0.7, "brightness", 9, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.5, "equalize", 1, fillcolor), + SubPolicy(0.6, "contrast", 7, 0.6, "sharpness", 5, fillcolor), + + SubPolicy(0.7, "color", 7, 0.5, "translateX", 8, fillcolor), + SubPolicy(0.3, "equalize", 7, 0.4, "autocontrast", 8, fillcolor), + SubPolicy(0.4, "translateY", 3, 0.2, "sharpness", 6, fillcolor), + SubPolicy(0.9, "brightness", 6, 0.2, "color", 8, fillcolor), + SubPolicy(0.5, "solarize", 2, 0.0, "invert", 3, fillcolor), + + SubPolicy(0.2, "equalize", 0, 0.6, "autocontrast", 0, fillcolor), + SubPolicy(0.2, "equalize", 8, 0.6, "equalize", 4, fillcolor), + SubPolicy(0.9, "color", 9, 0.6, "equalize", 6, fillcolor), + SubPolicy(0.8, "autocontrast", 4, 0.2, "solarize", 8, fillcolor), + SubPolicy(0.1, "brightness", 3, 0.7, "color", 0, fillcolor), + + SubPolicy(0.4, "solarize", 5, 0.9, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "translateY", 9, 0.7, "translateY", 9, fillcolor), + SubPolicy(0.9, "autocontrast", 2, 0.8, "solarize", 3, fillcolor), + SubPolicy(0.8, "equalize", 8, 0.1, "invert", 3, fillcolor), + SubPolicy(0.7, "translateY", 9, 0.9, "autocontrast", 1, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment CIFAR10 Policy" + + +class SVHNPolicy(object): + """ Randomly choose one of the best 25 Sub-policies on SVHN. + + Example: + >>> policy = SVHNPolicy() + >>> transformed = policy(image) + + Example as a PyTorch Transform: + >>> transform=transforms.Compose([ + >>> transforms.Resize(256), + >>> SVHNPolicy(), + >>> transforms.ToTensor()]) + """ + def __init__(self, fillcolor=(128, 128, 128)): + self.policies = [ + SubPolicy(0.9, "shearX", 4, 0.2, "invert", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.7, "invert", 5, fillcolor), + SubPolicy(0.6, "equalize", 5, 0.6, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 3, 0.6, "equalize", 3, fillcolor), + SubPolicy(0.6, "equalize", 1, 0.9, "rotate", 3, fillcolor), + + SubPolicy(0.9, "shearX", 4, 0.8, "autocontrast", 3, fillcolor), + SubPolicy(0.9, "shearY", 8, 0.4, "invert", 5, fillcolor), + SubPolicy(0.9, "shearY", 5, 0.2, "solarize", 6, fillcolor), + SubPolicy(0.9, "invert", 6, 0.8, "autocontrast", 1, fillcolor), + SubPolicy(0.6, "equalize", 3, 0.9, "rotate", 3, fillcolor), + + SubPolicy(0.9, "shearX", 4, 0.3, "solarize", 3, fillcolor), + SubPolicy(0.8, "shearY", 8, 0.7, "invert", 4, fillcolor), + SubPolicy(0.9, "equalize", 5, 0.6, "translateY", 6, fillcolor), + SubPolicy(0.9, "invert", 4, 0.6, "equalize", 7, fillcolor), + SubPolicy(0.3, "contrast", 3, 0.8, "rotate", 4, fillcolor), + + SubPolicy(0.8, "invert", 5, 0.0, "translateY", 2, fillcolor), + SubPolicy(0.7, "shearY", 6, 0.4, "solarize", 8, fillcolor), + SubPolicy(0.6, "invert", 4, 0.8, "rotate", 4, fillcolor), + SubPolicy(0.3, "shearY", 7, 0.9, "translateX", 3, fillcolor), + SubPolicy(0.1, "shearX", 6, 0.6, "invert", 5, fillcolor), + + SubPolicy(0.7, "solarize", 2, 0.6, "translateY", 7, fillcolor), + SubPolicy(0.8, "shearY", 4, 0.8, "invert", 8, fillcolor), + SubPolicy(0.7, "shearX", 9, 0.8, "translateY", 3, fillcolor), + SubPolicy(0.8, "shearY", 5, 0.7, "autocontrast", 3, fillcolor), + SubPolicy(0.7, "shearX", 2, 0.1, "invert", 5, fillcolor) + ] + + + def __call__(self, img): + policy_idx = random.randint(0, len(self.policies) - 1) + return self.policies[policy_idx](img) + + def __repr__(self): + return "AutoAugment SVHN Policy" + + +class SubPolicy(object): + def __init__(self, p1, operation1, magnitude_idx1, p2, operation2, magnitude_idx2, fillcolor=(128, 128, 128)): + ranges = { + "shearX": np.linspace(0, 0.3, 10), + "shearY": np.linspace(0, 0.3, 10), + "translateX": np.linspace(0, 150 / 331, 10), + "translateY": np.linspace(0, 150 / 331, 10), + "rotate": np.linspace(0, 30, 10), + "color": np.linspace(0.0, 0.9, 10), + "posterize": np.round(np.linspace(8, 4, 10), 0).astype(np.int), + "solarize": np.linspace(256, 0, 10), + "contrast": np.linspace(0.0, 0.9, 10), + "sharpness": np.linspace(0.0, 0.9, 10), + "brightness": np.linspace(0.0, 0.9, 10), + "autocontrast": [0] * 10, + "equalize": [0] * 10, + "invert": [0] * 10 + } + + # from https://stackoverflow.com/questions/5252170/specify-image-filling-color-when-rotating-in-python-with-pil-and-setting-expand + def rotate_with_fill(img, magnitude): + rot = img.convert("RGBA").rotate(magnitude) + return Image.composite(rot, Image.new("RGBA", rot.size, (128,) * 4), rot).convert(img.mode) + + func = { + "shearX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, magnitude * random.choice([-1, 1]), 0, 0, 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "shearY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, magnitude * random.choice([-1, 1]), 1, 0), + Image.BICUBIC, fillcolor=fillcolor), + "translateX": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, magnitude * img.size[0] * random.choice([-1, 1]), 0, 1, 0), + fillcolor=fillcolor), + "translateY": lambda img, magnitude: img.transform( + img.size, Image.AFFINE, (1, 0, 0, 0, 1, magnitude * img.size[1] * random.choice([-1, 1])), + fillcolor=fillcolor), + "rotate": lambda img, magnitude: rotate_with_fill(img, magnitude), + "color": lambda img, magnitude: ImageEnhance.Color(img).enhance(1 + magnitude * random.choice([-1, 1])), + "posterize": lambda img, magnitude: ImageOps.posterize(img, magnitude), + "solarize": lambda img, magnitude: ImageOps.solarize(img, magnitude), + "contrast": lambda img, magnitude: ImageEnhance.Contrast(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "sharpness": lambda img, magnitude: ImageEnhance.Sharpness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "brightness": lambda img, magnitude: ImageEnhance.Brightness(img).enhance( + 1 + magnitude * random.choice([-1, 1])), + "autocontrast": lambda img, magnitude: ImageOps.autocontrast(img), + "equalize": lambda img, magnitude: ImageOps.equalize(img), + "invert": lambda img, magnitude: ImageOps.invert(img) + } + + self.p1 = p1 + self.operation1 = func[operation1] + self.magnitude1 = ranges[operation1][magnitude_idx1] + self.p2 = p2 + self.operation2 = func[operation2] + self.magnitude2 = ranges[operation2][magnitude_idx2] + + + def __call__(self, img): + if random.random() < self.p1: img = self.operation1(img, self.magnitude1) + if random.random() < self.p2: img = self.operation2(img, self.magnitude2) + return img \ No newline at end of file diff --git a/Meta-causal/code/tools/causalaugment_v3.py b/Meta-causal/code/tools/causalaugment_v3.py new file mode 100644 index 0000000000000000000000000000000000000000..a375b7ebe5a83c3dba5b88f48f23a4326dec77e1 --- /dev/null +++ b/Meta-causal/code/tools/causalaugment_v3.py @@ -0,0 +1,694 @@ +# code in this file is adpated from rpmcruz/autoaugment +# https://github.com/rpmcruz/autoaugment/blob/master/transformations.py +import random + +import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw +import numpy as np +import torch +from PIL import Image,ImageStat +#import cv2 +from torchvision import transforms + +# def tensor2img(tensor): +# transform = transforms.Compose() + +def ShearX(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + +def DoShearX(img, v): # [-0.3, 0.3] + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + +def ShearY(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + +def DoShearY(img, v): # [-0.3, 0.3] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + +def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[0] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + +def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) +def DoTranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + +def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[1] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) +def DoTranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + +def Rotate(img, v): # [-30, 30] + assert -30 <= v <= 30 + if random.random() > 0.5: + v = -v + return img.rotate(v) +def DoRotate(img, v): # [-30, 30] + return img.rotate(v) + + +def AutoContrast(img, v): + return PIL.ImageOps.autocontrast(img, v) +def DoAutoContrast(img, v): + return PIL.ImageOps.autocontrast(img, v) + +def Invert(img, _): + return PIL.ImageOps.invert(img) +def DoInvert(img, _): + return PIL.ImageOps.invert(img) + + +def Equalize(img, _): + return PIL.ImageOps.equalize(img) +def DoEqualize(img, _): + return PIL.ImageOps.equalize(img) + +def Flip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + +def DoFlip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + + +def Solarize(img, v): # [0, 256] + assert 0 <= v <= 256 + return PIL.ImageOps.solarize(img, v) +def DoSolarize(img, v): # [0, 256] + return PIL.ImageOps.solarize(img, v) + +def SolarizeAdd(img, addition=0, threshold=128): + #img_np = np.array(img).astype(np.int) + img_np = np.array(img).astype(np.int32) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) +def DoSolarizeAdd(img, addition=0, threshold=128): + #img_np = np.array(img).astype(np.int) + img_np = np.array(img).astype(np.int32) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) + +def Posterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) +def DoPosterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) + + +def Contrast(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Contrast(img).enhance(v) + +def DoContrast(img, v): + return PIL.ImageEnhance.Contrast(img).enhance(v) + +def Color(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Color(img).enhance(v) + +def DoColor(img, v): + stat =ImageStat.Stat(img) + return PIL.ImageEnhance.Color(img).enhance(v) + + +def Brightness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Brightness(img).enhance(v) + +def DoBrightness(img, v): # obtain the brightness of image + return PIL.ImageEnhance.Brightness(img).enhance(v) + + +def Sharpness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Sharpness(img).enhance(v) + +def DoSharpness(img, v): + return PIL.ImageEnhance.Sharpness(img).enhance(v) + +def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] + assert 0.0 <= v <= 0.2 + if v <= 0.: + return img + + v = v * img.size[0] + return CutoutAbs(img, v) + + +def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img +def DoCutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img + + +def SamplePairing(imgs): # [0, 0.4] + def f(img1, v): + i = np.random.choice(len(imgs)) + img2 = PIL.Image.fromarray(imgs[i]) + return PIL.Image.blend(img1, img2, v) + + return f + + +def Identity(img, v): + return img + +def NoiseSalt(img, noise_rate): + """增加椒盐噪声 + args: + noise_rate (float): noise rate + """ + img_ = np.array(img).copy() + h, w, c = img_.shape + signal_pct = 1 - noise_rate + mask = np.random.choice((0, 1, 2), size=(h, w, 1), p=[signal_pct, noise_rate/2., noise_rate/2.]) + mask = np.repeat(mask, c, axis=2) + img_[mask == 1] = 255 # 盐噪声 + img_[mask == 2] = 0 # 椒噪声 + return Image.fromarray(img_.astype('uint8')) + +def DoNoiseSalt(img, noise_rate): + """增加椒盐噪声 + args: + noise_rate (float): noise rate + """ + img_ = np.array(img).copy() + h, w, c = img_.shape + signal_pct = 1 - noise_rate + mask = np.random.choice((0, 1, 2), size=(h, w, 1), p=[signal_pct, noise_rate/2., noise_rate/2.]) + mask = np.repeat(mask, c, axis=2) + img_[mask == 1] = 255 # 盐噪声 + img_[mask == 2] = 0 # 椒噪声 + return Image.fromarray(img_.astype('uint8')) +def NoiseGaussian(img, sigma): + """增加高斯噪声 + 传入: + img : 原图 + mean : 均值默认0 + sigma : 标准差 + 返回: + gaussian_out : 噪声处理后的图片 + """ + # 将图片灰度标准化 + img_ = np.array(img).copy() + img_ = img_ / 255.0 + # 产生高斯 noise + noise = np.random.normal(0, sigma, img_.shape) + # 将噪声和图片叠加 + gaussian_out = img_ + noise + # 将超过 1 的置 1，低于 0 的置 0 + gaussian_out = np.clip(gaussian_out, 0, 1) + # 将图片灰度范围的恢复为 0-255 + gaussian_out = np.uint8(gaussian_out*255) + # 将噪声范围搞为 0-255 + # noise = np.uint8(noise*255) + return Image.fromarray(gaussian_out) + +def DoNoiseGaussian(img, sigma): + """增加高斯噪声 + 传入: + img : 原图 + mean : 均值默认0 + sigma : 标准差 + 返回: + gaussian_out : 噪声处理后的图片 + """ + # 将图片灰度标准化 + img_ = np.array(img).copy() + img_ = img_ / 255.0 + # 产生高斯 noise + noise = np.random.normal(0, sigma, img_.shape) + # 将噪声和图片叠加 + gaussian_out = img_ + noise + # 将超过 1 的置 1，低于 0 的置 0 + gaussian_out = np.clip(gaussian_out, 0, 1) + # 将图片灰度范围的恢复为 0-255 + gaussian_out = np.uint8(gaussian_out*255) + # 将噪声范围搞为 0-255 + # noise = np.uint8(noise*255) + return Image.fromarray(gaussian_out) + +# def factor_list(factor_num): +# l = [ +# 'AutoContrast', +# 'Invert', +# 'Equalize', +# 'Solarize', +# 'SolarizeAdd', +# 'Posterize', +# 'Contrast', +# 'Color', +# 'Brightness', +# 'Sharpness', +# 'NoiseSalt', +# 'NoiseGaussian', +# ] +# return l[:factor_num] + +# def causal_list(factor_num): # 16 oeprations and their ranges +# l = [ +# (AutoContrast, 0, 100), +# (Invert, 0, 1), +# (Equalize, 0, 1), +# (Solarize, 0, 256), +# (SolarizeAdd, 0, 110), +# (Posterize, 0, 4), +# (Contrast, 0.1, 1.9), +# (Color, 0.1, 1.9), +# (Brightness, 0.1, 1.9), +# (Sharpness, 0.1, 1.9), +# (NoiseSalt,0.0,0.1), +# (NoiseGaussian,0.0,0.1), +# ] + +# return l[:factor_num] + + +# def factor_list(factor_num): +# l = [ +# 'ShearX', +# 'ShearY', +# 'Rotate', +# 'Flip' +# ] +# return l[:factor_num] + +# def causal_list(factor_num): # 16 oeprations and their ranges +# l = [ +# (ShearX, 0., 0.3), +# (ShearY, 0., 0.3), +# (Rotate, 0, 30), +# (Flip, 0, 1), +# ] + +# return l[:factor_num] + +def factor_list(factor_num): + l = [ + 'ShearX', + 'ShearY', + 'AutoContrast', + 'Invert', + 'Equalize', + 'Solarize', + 'SolarizeAdd', + 'Posterize', + 'Contrast', + 'Color', + 'Brightness', + 'Sharpness', + 'NoiseSalt', + 'NoiseGaussian', + 'Rotate', + 'Flip' + ] + return l[:factor_num] + +def causal_list(factor_num): # 16 oeprations and their ranges + l = [ + (ShearX, 0., 0.3), + (ShearY, 0., 0.3), + (AutoContrast, 0, 100), + (Invert, 0, 1), + (Equalize, 0, 1), + (Solarize, 0, 256), + (SolarizeAdd, 0, 110), + (Posterize, 0, 4), + (Contrast, 0.1, 1.9), + (Color, 0.1, 1.9), + (Brightness, 0.1, 1.9), + (Sharpness, 0.1, 1.9), + (NoiseSalt,0.0,0.1), + (NoiseGaussian,0.0,0.1), + (Rotate, 0, 30), + (Flip, 0, 1), + ] + + return l[:factor_num] + +class Lighting(object): + """Lighting noise(AlexNet - style PCA - based noise)""" + + def __init__(self, alphastd, eigval, eigvec): + self.alphastd = alphastd + self.eigval = torch.Tensor(eigval) + self.eigvec = torch.Tensor(eigvec) + + def __call__(self, img): + if self.alphastd == 0: + return img + + alpha = img.new().resize_(3).normal_(0, self.alphastd) + rgb = self.eigvec.type_as(img).clone() \ + .mul(alpha.view(1, 3).expand(3, 3)) \ + .mul(self.eigval.view(1, 3).expand(3, 3)) \ + .sum(1).squeeze() + + return img.add(rgb.view(3, 1, 1).expand_as(img)) + + +class CutoutDefault(object): + """ + Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py + """ + def __init__(self, length): + self.length = length + + def __call__(self, img): + h, w = img.size(1), img.size(2) + mask = np.ones((h, w), np.float32) + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + mask[y1: y2, x1: x2] = 0. + mask = torch.from_numpy(mask) + mask = mask.expand_as(img) + img *= mask + return img + + +class RandAugment_incausal: + def __init__(self, n, m, factor_num, randm=False, randn=False): + self.n = n + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + self.randn = randn + self.factor_num = factor_num + print("randm:",self.randm) + print("randn:",self.randn) + print("n:",self.n) + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + if self.randn: + self.n = random.randint(1,self.factor_num) + + ops = random.choices(self.causal_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img +class RandAugment_all: + def __init__(self, m, factor_num, randm=False): + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + self.factor_num = factor_num + + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + factor_choice = np.random.randint(0,2,self.factor_num) + # ops = random.choices(self.causal_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + if factor_choice[index] == 0: + continue + else: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img +class RandAugment_incausal_label: + def __init__(self, n, m, factor_num, randm=False): + self.n = n + self.m = m # [0, 30] + self.causal_list = causal_list(factor_num) + self.factor_num = factor_num + print("---------------------------%d factors-----------------"%(len(self.causal_list))) + self.randm = randm + print("randm:",self.randm) + + def __call__(self, img): + # print("%d factors-----------------"%(len(self.causal_list))) + #op_labels = np.random.randint(0,self.factor_num-1,self.n) + op_labels = random.sample(range(0, self.factor_num), self.n) + ops = [li for index, li in enumerate(self.causal_list) if index in op_labels] + #ops = random.choices(self.causal_list, k=self.n) + # print(self.causal_list) + # print("op_labels:",op_labels) + # print("select_op:",ops) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + return img, np.array(op_labels) +class FactualAugment_incausal: + def __init__(self, m, factor_num, randm=False): + self.m = m + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.randm = randm + print("randm:",self.randm) + def __call__(self, img): + # ops = random.choices(self.causal_list, k=1) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + val = (float(self.m) / 30) * float(maxval - minval) + minval + if index == 0: + imgs = np.array(op(img, val)) + else: + imgs = np.concatenate((imgs, op(img, val)),-1) + # print("imgs",imgs.shape) + return imgs +class CounterfactualAugment_incausal: + def __init__(self,factor_num): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + def __call__(self, img): + # index = 0 + # b, c, h, w = img.shape + # imgs = torch.zeros(b*self.factor_num, c, h, w) + # for b_ in range(32): + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + if index == 0: + imgs = np.array(op(img, maxval)) + else: + imgs = np.concatenate((imgs, op(img, maxval)),-1) + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs +class MultiCounterfactualAugment_incausal: + def __init__(self, factor_num, stride): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.stride = stride + + def __call__(self, img): + # index = 0 + # b, c, h, w = img.shape + # imgs = torch.zeros(b*self.factor_num, c, h, w) + # for b_ in range(32): + # 0,5,10,15,20,25,30 + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + for i in range(0, 31, self.stride): + val = (float(i) / 30) * float(maxval - minval) + minval + if index == 0 and i == 0: + imgs = np.array(op(img, val)) + else: + imgs = np.concatenate((imgs, op(img, val)),-1) + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs +class MultiCounterfactualAugment: + def __init__(self, factor_num, stride=5): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.stride = stride + self.var_num = len(list(range(0, 31, self.stride))) + print("stride:",stride) + def __call__(self, img): + # index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num*self.var_num, c, h, w) + # for b_ in range(32): + # 0,5,10,15,20,25,30 + # print(img.shape) + for b_ in range(b): + img0 = transforms.ToPILImage()(imgs[b_]) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + i_index = 0 + for i in range(0, 31, self.stride): + val = (float(i) / 30) * float(maxval - minval) + minval + img1 = op(img0, val) + img1 = transforms.ToTensor()(img1) + #print(f'batch {b_} factor {index} stride {i} i_index {i_index} total {b_*self.factor_num*self.var_num+index*self.var_num+i_index}') + imgs[b_*self.factor_num*self.var_num+index*self.var_num+i_index] = img1 + i_index = i_index + 1 + # img = op(img, maxval) + # imgs[b_*factor_num+index] = op(img[b_], maxval) + return imgs + + +class FactualAugment: + def __init__(self, m, factor_num, randm=False): + self.m = m + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + self.randm = randm + print("randm:",randm) + def __call__(self, img): + index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num, c, h, w) + + img = img.cpu() + for b_ in range(b): + imgs[b_*self.factor_num:(b_+1)*self.factor_num] = self.get_item(img[b_]) + return imgs + def get_item(self, img): + index = 0 + # print("input_dim:",img.shape) + c, h, w = img.shape + imgs = torch.zeros(self.factor_num, c, h, w) + # img = img.squeeze(0) + # print(img.shape) + img = transforms.ToPILImage()(img) + if self.randm: + self.m = random.randint(0,30) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval(self.factor_list[index]) + val = (float(self.m) / 30) * float(maxval - minval) + minval + img1 = op(img, val) + img1 = transforms.ToTensor()(img1) + imgs[index] = img1 + return imgs +class CounterfactualAugment: + def __init__(self,factor_num): + self.causal_list = causal_list(factor_num) + self.factor_list = factor_list(factor_num) + self.factor_num = factor_num + + def __call__(self, img): + index = 0 + b, c, h, w = img.shape + imgs = torch.zeros(b*self.factor_num, c, h, w) + + img = img.cpu() + for b_ in range(b): + imgs[b_*self.factor_num:(b_+1)*self.factor_num] = self.get_item(img[b_]) + return imgs + def get_item(self, img): + index = 0 + c, h, w = img.shape + imgs = torch.ones(self.factor_num, c, h, w) + # img = img.squeeze(0) + img = transforms.ToPILImage()(img) + for index, (op, minval, maxval) in enumerate(self.causal_list): + op = eval('Do'+self.factor_list[index]) + img1 = op(img, maxval) + # img1.save('test'+str(index)+'.png') + img1 = transforms.ToTensor()(img1) + imgs[index] = img1 + return imgs + +class Avg_statistic: + def __init__(self): + self.do_list = do_list() + self.statistic_num = len(self.do_list) + self.avg_val = np.zeros(self.statistic_num) + self.img_num = 0 + + def get_item(self,img): + # ops = self.statistic_list + do_index = 0 + for op in self.do_list: + val=op(img) + self.avg_val[do_index] += val + self.img_num = self.img_num + 1 + + def compute_average(self): + self.avg_val = self.avg_val/self.img_num + + def get_infor(self): + return self.avg_val, self.img_num + + + + diff --git a/Meta-causal/code/tools/randaugment.py b/Meta-causal/code/tools/randaugment.py new file mode 100644 index 0000000000000000000000000000000000000000..f3bbdf11541df078144fa0ced8d693d4c98507ad --- /dev/null +++ b/Meta-causal/code/tools/randaugment.py @@ -0,0 +1,248 @@ +# code in this file is adpated from rpmcruz/autoaugment +# https://github.com/rpmcruz/autoaugment/blob/master/transformations.py +import random + +import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw +import numpy as np +import torch +from PIL import Image + + +def ShearX(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) + + +def ShearY(img, v): # [-0.3, 0.3] + assert -0.3 <= v <= 0.3 + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) + + +def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[0] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + + +def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) + + +def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert -0.45 <= v <= 0.45 + if random.random() > 0.5: + v = -v + v = v * img.size[1] + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] + assert 0 <= v + if random.random() > 0.5: + v = -v + return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) + + +def Rotate(img, v): # [-30, 30] + assert -30 <= v <= 30 + if random.random() > 0.5: + v = -v + return img.rotate(v) + + +def AutoContrast(img, _): + return PIL.ImageOps.autocontrast(img) + + +def Invert(img, _): + return PIL.ImageOps.invert(img) + + +def Equalize(img, _): + return PIL.ImageOps.equalize(img) + + +def Flip(img, _): # not from the paper + return PIL.ImageOps.mirror(img) + + +def Solarize(img, v): # [0, 256] + assert 0 <= v <= 256 + return PIL.ImageOps.solarize(img, v) + + +def SolarizeAdd(img, addition=0, threshold=128): + img_np = np.array(img).astype(np.int) + img_np = img_np + addition + img_np = np.clip(img_np, 0, 255) + img_np = img_np.astype(np.uint8) + img = Image.fromarray(img_np) + return PIL.ImageOps.solarize(img, threshold) + + +def Posterize(img, v): # [4, 8] + v = int(v) + v = max(1, v) + return PIL.ImageOps.posterize(img, v) + + +def Contrast(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Contrast(img).enhance(v) + + +def Color(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Color(img).enhance(v) + + +def Brightness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Brightness(img).enhance(v) + + +def Sharpness(img, v): # [0.1,1.9] + assert 0.1 <= v <= 1.9 + return PIL.ImageEnhance.Sharpness(img).enhance(v) + + +def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] + assert 0.0 <= v <= 0.2 + if v <= 0.: + return img + + v = v * img.size[0] + return CutoutAbs(img, v) + + +def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] + # assert 0 <= v <= 20 + if v < 0: + return img + w, h = img.size + x0 = np.random.uniform(w) + y0 = np.random.uniform(h) + + x0 = int(max(0, x0 - v / 2.)) + y0 = int(max(0, y0 - v / 2.)) + x1 = min(w, x0 + v) + y1 = min(h, y0 + v) + + xy = (x0, y0, x1, y1) + color = (125, 123, 114) + # color = (0, 0, 0) + img = img.copy() + PIL.ImageDraw.Draw(img).rectangle(xy, color) + return img + + +def SamplePairing(imgs): # [0, 0.4] + def f(img1, v): + i = np.random.choice(len(imgs)) + img2 = PIL.Image.fromarray(imgs[i]) + return PIL.Image.blend(img1, img2, v) + + return f + + +def Identity(img, v): + return img + + +def augment_list(): # 16 oeprations and their ranges + + # https://github.com/tensorflow/tpu/blob/8462d083dd89489a79e3200bcc8d4063bf362186/models/official/efficientnet/autoaugment.py#L505 + l = [ + (AutoContrast, 0, 1), + (Equalize, 0, 1), + (Invert, 0, 1), + (Rotate, 0, 30), + (Posterize, 0, 4), + (Solarize, 0, 256), + (SolarizeAdd, 0, 110), + (Color, 0.1, 1.9), + (Contrast, 0.1, 1.9), + (Brightness, 0.1, 1.9), + (Sharpness, 0.1, 1.9), + (ShearX, 0., 0.3), + (ShearY, 0., 0.3), + (CutoutAbs, 0, 40), + (TranslateXabs, 0., 100), + (TranslateYabs, 0., 100), + ] + + return l + + +class Lighting(object): + """Lighting noise(AlexNet - style PCA - based noise)""" + + def __init__(self, alphastd, eigval, eigvec): + self.alphastd = alphastd + self.eigval = torch.Tensor(eigval) + self.eigvec = torch.Tensor(eigvec) + + def __call__(self, img): + if self.alphastd == 0: + return img + + alpha = img.new().resize_(3).normal_(0, self.alphastd) + rgb = self.eigvec.type_as(img).clone() \ + .mul(alpha.view(1, 3).expand(3, 3)) \ + .mul(self.eigval.view(1, 3).expand(3, 3)) \ + .sum(1).squeeze() + + return img.add(rgb.view(3, 1, 1).expand_as(img)) + + +class CutoutDefault(object): + """ + Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py + """ + def __init__(self, length): + self.length = length + + def __call__(self, img): + h, w = img.size(1), img.size(2) + mask = np.ones((h, w), np.float32) + y = np.random.randint(h) + x = np.random.randint(w) + + y1 = np.clip(y - self.length // 2, 0, h) + y2 = np.clip(y + self.length // 2, 0, h) + x1 = np.clip(x - self.length // 2, 0, w) + x2 = np.clip(x + self.length // 2, 0, w) + + mask[y1: y2, x1: x2] = 0. + mask = torch.from_numpy(mask) + mask = mask.expand_as(img) + img *= mask + return img + + +class RandAugment: + def __init__(self, n, m, randm=False): + self.n = n + self.m = m # [0, 30] + self.augment_list = augment_list() + self.randm = randm + + def __call__(self, img): + ops = random.choices(self.augment_list, k=self.n) + if self.randm: + self.m = random.randint(0,30) + for op, minval, maxval in ops: + val = (float(self.m) / 30) * float(maxval - minval) + minval + # print("val:",val) + img = op(img, val) + + return img diff --git a/Meta-causal/env_mc.yml b/Meta-causal/env_mc.yml new file mode 100644 index 0000000000000000000000000000000000000000..f57a97c20b545aceafecd9ec08c6cbb55038a01f --- /dev/null +++ b/Meta-causal/env_mc.yml @@ -0,0 +1,22 @@ +name: py36 +channels: + - pytorch + - nvidia + - conda-forge +dependencies: + - python=3.11.* + - torchvision + - pandas + - pip + - pytorch-cuda=12.1 + - click + - pip: + - scipy>=1.3.2 + - tensorboardX>=1.4 + - h5py>=2.9.0 + - tensorboard + - timm + - opencv-python==4.5.5.62 + - ml-collections + - numpy + diff --git a/Meta-causal/metacaEnv.yml b/Meta-causal/metacaEnv.yml new file mode 100644 index 0000000000000000000000000000000000000000..b0bd424fb7c5aa818f10a82173549eb0dd3199c7 --- /dev/null +++ b/Meta-causal/metacaEnv.yml @@ -0,0 +1,119 @@ +name: Py3.7_torch1.8 +channels: + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ + - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ + - conda-forge + - bioconda + - defaults +dependencies: + - _libgcc_mutex=0.1=main + - asn1crypto=1.2.0=py37_0 + - blas=1.0=mkl + - bottleneck=1.3.2=py37heb32a55_1 + - bzip2=1.0.8=h7b6447c_0 + - ca-certificates=2021.10.8=ha878542_0 + - cairo=1.14.12=h8948797_3 + - certifi=2021.10.8=py37h89c1867_1 + - cffi=1.13.0=py37h2e261b9_0 + - chardet=3.0.4=py37_1003 + - click=8.0.3=pyhd3eb1b0_0 + - conda-package-handling=1.6.0=py37h7b6447c_0 + - cryptography=2.8=py37h1ba5d50_0 + - ffmpeg=4.0=hcdf2ecd_0 + - fontconfig=2.13.0=h9420a91_0 + - freeglut=3.0.0=hf484d3e_5 + - freetype=2.11.0=h70c0345_0 + - glib=2.63.1=h5a9c865_0 + - graphite2=1.3.14=h23475e2_0 + - h5py=2.8.0=py37h3010b51_1003 + - harfbuzz=1.8.8=hffaf4a1_0 + - hdf5=1.10.2=hba1933b_1 + - icu=58.2=he6710b0_3 + - idna=2.8=py37_0 + - intel-openmp=2021.3.0=h06a4308_3350 + - jasper=2.0.14=hd8c5072_2 + - jpeg=9d=h7f8727e_0 + - libedit=3.1.20181209=hc058e9b_0 + - libffi=3.2.1=hd88cf55_4 + - libgcc-ng=9.1.0=hdf63c60_0 + - libgfortran-ng=7.5.0=ha8ba4b0_17 + - libgfortran4=7.5.0=ha8ba4b0_17 + - libglu=9.0.0=hf484d3e_1 + - libopencv=3.4.2=hb342d67_1 + - libopus=1.3.1=h7b6447c_0 + - libpng=1.6.37=hbc83047_0 + - libprotobuf=3.17.2=h4ff587b_1 + - libstdcxx-ng=9.1.0=hdf63c60_0 + - libtiff=4.1.0=h2733197_0 + - libuuid=1.0.3=h7f8727e_2 + - libvpx=1.7.0=h439df22_0 + - libxcb=1.14=h7b6447c_0 + - libxml2=2.9.9=hea5a465_1 + - mkl=2021.3.0=h06a4308_520 + - mkl-service=2.4.0=py37h7f8727e_0 + - mkl_fft=1.3.1=py37hd3c417c_0 + - mkl_random=1.2.2=py37h51133e4_0 + - ncurses=6.1=he6710b0_1 + - numexpr=2.7.3=py37h22e1b3c_1 + - numpy-base=1.21.2=py37h79a1101_0 + - opencv=3.4.2=py37h6fd60c2_1 + - openssl=1.1.1h=h516909a_0 + - pandas=1.3.3=py37h8c16a72_0 + - pcre=8.45=h295c915_0 + - pip=19.3.1=py37_0 + - pixman=0.40.0=h7f8727e_1 + - protobuf=3.17.2=py37h295c915_0 + - py-opencv=3.4.2=py37hb342d67_1 + - pycosat=0.6.3=py37h14c3975_0 + - pycparser=2.19=py37_0 + - pyopenssl=19.0.0=py37_0 + - pysocks=1.7.1=py37_0 + - python=3.7.4=h265db76_1 + - python-dateutil=2.8.2=pyhd3eb1b0_0 + - python_abi=3.7=2_cp37m + - pytz=2021.3=pyhd3eb1b0_0 + - readline=7.0=h7b6447c_5 + - requests=2.22.0=py37_0 + - ruamel_yaml=0.15.46=py37h14c3975_0 + - scipy=1.7.1=py37h292c36d_2 + - setuptools=41.4.0=py37_0 + - six=1.12.0=py37_0 + - sqlite=3.30.0=h7b6447c_0 + - tensorboardx=2.2=pyhd3eb1b0_0 + - tk=8.6.8=hbc83047_0 + - tqdm=4.36.1=py_0 + - urllib3=1.24.2=py37_0 + - wheel=0.33.6=py37_0 + - xz=5.2.4=h14c3975_4 + - yaml=0.1.7=had09818_2 + - zlib=1.2.11=h7b6447c_3 + - zstd=1.3.7=h0b5b093_0 + - pip: + - absl-py==1.0.0 + - cachetools==4.2.4 + - conda-pack==0.6.0 + - google-auth==2.3.3 + - google-auth-oauthlib==0.4.6 + - grpcio==1.42.0 + - importlib-metadata==4.8.2 + - markdown==3.3.6 + - numpy==1.21.3 + - oauthlib==3.1.1 + - pillow==8.4.0 + - pyasn1==0.4.8 + - pyasn1-modules==0.2.8 + - requests-oauthlib==1.3.0 + - rsa==4.8 + - tensorboard==2.7.0 + - tensorboard-data-server==0.6.1 + - tensorboard-plugin-wit==1.8.0 + - torch==1.8.1+cu111 + - torchvision==0.9.1+cu111 + - typing-extensions==3.10.0.2 + - werkzeug==2.0.2 + - zipp==3.6.0 +prefix: /home/chenjin/miniconda3/envs/Py3.7_torch1.8 diff --git a/Meta-causal/readme.md b/Meta-causal/readme.md new file mode 100644 index 0000000000000000000000000000000000000000..e796ef747933d931c5f825afd689e43953ff640b --- /dev/null +++ b/Meta-causal/readme.md @@ -0,0 +1,33 @@ +# Meta-causal + +The code for **Meta-causal Learning for Single Domain Generalization [CVPR2023]**. Our code is based on the method of PDEN(https://github.com/lileicv/PDEN/). + +### Dataset + +- Download the data and model from [Baidu Cloud Disk](https://pan.baidu.com/s/14pdVbNAHWKeC4AE7QqtFmw) (password:pxvt ). +- Place the dataset files in the path `./data/` and the model files in the path `./` + +### Environment + +Please refer to `env.yaml` + +### Train and Test +- For digit, run the command `bash run_my_joint_test.sh 0` under the path `./run_digits/` . +- For PACS, when using art_painting as the source domain, run the command `bash run_my_joint_v13_test.sh 0` under the path `./run_PACS/` . + +### If this code is helpful, please cite our paper + +``` +@InProceedings{Chen_2023_CVPR, + author = {Chen, Jin and Gao, Zhi and Wu, Xinxiao and Luo, Jiebo}, + title = {Meta-Causal Learning for Single Domain Generalization}, + booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, + month = {June}, + year = {2023}, + pages = {7683-7692} +} +``` + +### Contact +gaozhi_2017@126.com + diff --git a/config_bscdfsl_dir.py b/config_bscdfsl_dir.py new file mode 100644 index 0000000000000000000000000000000000000000..828669d5221bf05d03c5bb670f636af7196a3513 --- /dev/null +++ b/config_bscdfsl_dir.py @@ -0,0 +1,4 @@ +EuroSAT_path = "/scratch/yuqian_fu/Data/CDFSL/EuroSAT" +ChestX_path = "/scratch/yuqian_fu/Data/CDFSL/ChestX" +CropDisease_path = "/scratch/yuqian_fu/Data/CDFSL/CropDisease" +ISIC_path = "/scratch/yuqian_fu/Data/CDFSL/ISIC" diff --git a/env.yml b/env.yml new file mode 100644 index 0000000000000000000000000000000000000000..559c7e06ec8499c78a67528b2907b658adebbddf --- /dev/null +++ b/env.yml @@ -0,0 +1,23 @@ +name: py36 +channels: + - pytorch + - nvidia + - conda-forge +dependencies: + - python=3.11.* + - numpy + - pytorch + - torchvision + - pandas + - pip + - pytorch-cuda=12.1 + - pip: + - scipy>=1.3.2 + - tensorboardX>=1.4 + - h5py>=2.9.0 + - tensorboard + - timm + - opencv-python==4.5.5.62 + - ml-collections + + diff --git a/finetune_StyleAdv_RN.py b/finetune_StyleAdv_RN.py new file mode 100644 index 0000000000000000000000000000000000000000..8d615b447381fdfb9f2a772207c2b1a23fd475d4 --- /dev/null +++ b/finetune_StyleAdv_RN.py @@ -0,0 +1,114 @@ +import os +import numpy as np +import torch +import torch.nn as nn +import torch.optim +import random +from options import parse_args +from utils.PSG import PseudoSampleGenerator +from methods.backbone_multiblock import model_dict +from methods.StyleAdv_RN_GNN import StyleAdvGNN +from data.datamgr import SetDataManager +from data import ISIC_few_shot, EuroSAT_few_shot, CropDisease_few_shot, Chest_few_shot + +def finetune(novel_loader, n_pseudo=75, n_way=5, n_support=5): + iter_num = len(novel_loader) + acc_all = [] + + checkpoint_dir = '%s/checkpoints/%s/best_model.tar' % (params.save_dir, params.resume_dir) + state = torch.load(checkpoint_dir)['state'] + for ti, (x, _) in enumerate(novel_loader): # x:(5, 20, 3, 224, 224) + model = StyleAdvGNN(model_dict[params.model], n_way=n_way, n_support=n_support).cuda() + model.load_state_dict(state, strict = True) + x = x.cuda() + # Finetune components initialization + xs = x[:, :n_support].reshape(-1, *x.size()[2:]) # (25, 3, 224, 224) + pseudo_q_genrator = PseudoSampleGenerator(n_way, n_support, n_pseudo) + loss_fun = nn.CrossEntropyLoss().cuda() + #opt = torch.optim.Adam(model.parameters()) + opt = torch.optim.Adam(model.parameters(), lr = 0.005) + #opt = torch.optim.Adam(model.parameters(), lr=0.0005) #lr version 2 + #opt = torch.optim.Adam(model.parameters(), lr=5e-5) #lr version3, for cvpr2023 + # Finetune process + n_query = n_pseudo//n_way + pseudo_set_y = torch.from_numpy(np.repeat(range(n_way), n_query)).cuda() + model.n_query = n_query + model.train() + for epoch in range(params.finetune_epoch): + opt.zero_grad() + pseudo_set = pseudo_q_genrator.generate(xs) # (5, n_support+n_query, 3, 224, 224) + scores = model.set_forward(pseudo_set) # (5*n_query, 5) + loss = loss_fun(scores, pseudo_set_y) + loss.backward() + opt.step() + del pseudo_set, scores, loss + torch.cuda.empty_cache() + # Inference process + n_query = x.size(1) - n_support + model.n_query = n_query + yq = np.repeat(range(n_way), n_query) + with torch.no_grad(): + scores = model.set_forward(x) # (80, 5) + _, topk_labels = scores.data.topk(1, 1, True, True) + topk_ind = topk_labels.cpu().numpy() # (80, 1) + top1_correct = np.sum(topk_ind[:,0]==yq) + acc = top1_correct*100./(n_way*n_query) + acc_all.append(acc) + del scores, topk_labels + torch.cuda.empty_cache() + print('Task %d : %4.2f%%, mean Acc: %4.2f'%(ti, acc, np.mean(np.array(acc_all)))) + + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print('Test Acc = %4.2f +- %4.2f%%'%(acc_mean, 1.96*acc_std/np.sqrt(iter_num))) + +if __name__=='__main__': + seed = 0 + print("set seed = %d" % seed) + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) + torch.cuda.manual_seed_all(seed) + torch.backends.cudnn.deterministic = True + torch.backends.cudnn.benchmark = False + + params = parse_args('train') + + image_size = 224 + iter_num = 1000 + n_query = 16 + n_pseudo = 75 + + print('Loading target dataset!:', params.testset) + if params.testset in ['cub', 'cars', 'places', 'plantae']: + novel_file = os.path.join(params.data_dir, params.testset, 'novel.json') + datamgr = SetDataManager(image_size, n_query=n_query, n_way=params.test_n_way, n_support=params.n_shot, n_eposide=iter_num) + novel_loader = datamgr.get_data_loader(novel_file, aug=False) + + else: + few_shot_params = dict(n_way = params.test_n_way , n_support = params.n_shot) + if params.testset in ["ISIC"]: + datamgr = ISIC_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + elif params.testset in ["EuroSAT"]: + datamgr = EuroSAT_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + elif params.testset in ["CropDisease"]: + datamgr = CropDisease_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + elif params.testset in ["ChestX"]: + datamgr = Chest_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + import time + #start = time.clock() + start =time.perf_counter() + finetune(novel_loader, n_pseudo=n_pseudo, n_way=params.test_n_way, n_support=params.n_shot) + #end = time.clock() + end = time.perf_counter() + print('Running time: %s Seconds: %s Min: %s Min per epoch'%(end-start, (end-start)/60, (end-start)/60/iter_num)) + diff --git a/finetune_StyleAdv_ViT.py b/finetune_StyleAdv_ViT.py new file mode 100644 index 0000000000000000000000000000000000000000..51a3f2e0d27611bcd02be675bc2e8b647df5868e --- /dev/null +++ b/finetune_StyleAdv_ViT.py @@ -0,0 +1,233 @@ +import os +import numpy as np +import torch +import torch.nn as nn +import torch.optim +import random +from methods.backbone import model_dict +from data.datamgr import SetDataManager +from options import parse_args +#from methods.matchingnet import MatchingNet +#from methods.relationnet import RelationNet +#from methods.protonet import ProtoNet +#from methods.gnnnet import GnnNet +#from methods.tpn import TPN +#from PSG import PseudoSampleGenerator +from utils.PSG import PseudoSampleGenerator + +from data import ISIC_few_shot, EuroSAT_few_shot, CropDisease_few_shot, Chest_few_shot + +#from cvpr2023_startup_20221026 import * +#from cvpr2023_load_models_20221102 import load_ViTsmall +from methods.load_ViT_models import load_ViTsmall +#from models.pmf_protonet import ProtoNet +#from methods.pmf_protonet import ProtoNet +from methods.protonet import ProtoNet + +#PMF_metatrained = False +PMF_metatrained = True +FINAL_FEAT_DIM = 384 +FINETUNE_ALL = True +#FINETUNE_ALL = False + +#tune_lr = 0.01 +#tune_lr = 0.001 +#tune_lr = 0.0001 +tune_lr = 5e-5 + +def load_model(): + vit_model = load_ViTsmall() + model = ProtoNet(vit_model) + + if PMF_metatrained: + #pmf_pretrained_ckp = 'outputs/20221103-styleAdv_metatrain_vit_protonet_trainEpoch20_exp0_lr0/checkpoint.pth' + #pmf_pretrained_ckp = 'outputs/20221103-styleAdv_metatrain_vit_protonet_trainEpoch20_exp1_lr1/checkpoint.pth' + #pmf_pretrained_ckp = 'outputs/20221103-styleAdv_metatrain_vit_protonet_trainEpoch20_exp2_lr2/checkpoint.pth' + #pmf_pretrained_ckp = 'outputs/20221103-styleAdv_metatrain_vit_protonet_trainEpoch20_exp3_lr3/checkpoint.pth' + + # 1shot + #pmf_pretrained_ckp = 'outputs/20221106-styleAdv_metatrain_vit_protonet_trainEpoch20_1shot_exp0_lr0_saveBestPth/checkpoint.pth' + pmf_pretrained_ckp = 'output/20221106-styleAdv_metatrain_vit_protonet_trainEpoch20_1shot_exp2_lr2_saveBestPth/checkpoint.pth' + #pmf_pretrained_ckp = 'outputs/20221106-styleAdv_metatrain_vit_protonet_trainEpoch20_1shot_exp0_lr0_saveBestPth_PthreDot4/checkpoint.pth' + #pmf_pretrained_ckp = 'outputs/20221106-styleAdv_metatrain_vit_protonet_trainEpoch20_1shot_exp2_lr2_saveBestPth_PthreDot4/checkpoint.pth' + + #pmf_pretrained_ckp = 'outputs/20221106-withoutstyleAdv_metatrain_vit_protonet_exp0_1shot/best.pth' + + state_pmf = torch.load(pmf_pretrained_ckp)['model'] + + # + state_new = state_pmf + state_keys = list(state_pmf.keys()) + for i, key in enumerate(state_keys): + if 'feature.' in key: + newkey = key.replace("feature.","backbone.") + state_new[newkey] = state_pmf.pop(key) + if 'classifier.' in key: + state_new.pop(key) + else: + pass + model.load_state_dict(state_new) + model.train().cuda() + return model + + +def set_forward_ViTProtonet(model, x): + n_way = x.size()[0] + n_query = 15 + n_support = x.size()[1] - n_query + + SupportTensor = x[:, :n_support, :, :, :] + QueryTensor = x[:, n_support:, :, :, :] + SupportLabel = torch.from_numpy(np.repeat(range(n_way), n_support)).cuda() + QueryLabel = torch.from_numpy(np.repeat(range(n_way), n_query)).cuda() + + SupportTensor = SupportTensor.contiguous().view(-1, n_way*n_support, 3, 224, 224) + QueryTensor = QueryTensor.contiguous().view(-1, n_way*n_query, 3, 224, 224) + SupportLabel = SupportLabel.contiguous().view(-1, n_way*n_support) + QueryLabel = QueryLabel.contiguous().view(-1, n_way*n_query) + #print(SupportTensor.size(), SupportLabel.size(), QueryTensor.size()) + output = model(SupportTensor, SupportLabel, QueryTensor) + output = output.view(n_way*n_query,n_way) + return output + +def finetune(novel_loader, n_pseudo=75, n_way=5, n_support=5): + iter_num = len(novel_loader) + acc_all = [] + + #checkpoint_dir = '%s/checkpoints/%s/best_model.tar' % (params.save_dir, params.name) + #checkpoint_dir = '%s/checkpoints/%s/best_model.tar' % (params.save_dir, params.resume_dir) + #state = torch.load(checkpoint_dir)['state'] + for ti, (x, _) in enumerate(novel_loader): # x:(5, 20, 3, 224, 224) + ''' + # Model + if params.method == 'MatchingNet': + model = MatchingNet(model_dict[params.model], n_way=n_way, n_support=n_support).cuda() + elif params.method == 'RelationNet': + model = RelationNet(model_dict[params.model], n_way=n_way, n_support=n_support).cuda() + elif params.method == 'ProtoNet': + model = ProtoNet(model_dict[params.model], n_way=n_way, n_support=n_support).cuda() + elif params.method == 'GNN': + model = GnnNet(model_dict[params.model], n_way=n_way, n_support=n_support).cuda() + elif params.method == 'TPN': + model = TPN(model_dict[params.model], n_way=n_way, n_support=n_support).cuda() + else: + print("Please specify the method!") + assert (False) + # Update model + if 'FWT' in params.name: + model_params = model.state_dict() + pretrained_dict = {k: v for k, v in state.items() if k in model_params} + model_params.update(pretrained_dict) + model.load_state_dict(model_params) + else: + model.load_state_dict(state, strict = False) + ''' + model = load_model() + x = x.cuda() + # Finetune components initialization + xs = x[:, :n_support].reshape(-1, *x.size()[2:]) # (25, 3, 224, 224) + #print('xs:', xs.size()) + pseudo_q_genrator = PseudoSampleGenerator(n_way, n_support, n_pseudo) + loss_fun = nn.CrossEntropyLoss().cuda() + #opt = torch.optim.Adam(model.parameters()) + #opt = torch.optim.Adam(model.parameters(), lr=0.0005) #lr version 2 + opt = torch.optim.SGD(model.parameters(), lr = tune_lr, momentum=0.9, weight_decay=0,) #pmf opt + + # Finetune process + n_query = n_pseudo//n_way + pseudo_set_y = torch.from_numpy(np.repeat(range(n_way), n_query)).cuda() + model.n_query = n_query + model.train() + for epoch in range(params.finetune_epoch): + opt.zero_grad() + pseudo_set = pseudo_q_genrator.generate(xs) # (5, n_support+n_query, 3, 224, 224) + #scores = model.set_forward(pseudo_set) # (5*n_query, 5) + scores = set_forward_ViTProtonet(model, pseudo_set) + loss = loss_fun(scores, pseudo_set_y) + loss.backward() + opt.step() + del pseudo_set, scores, loss + torch.cuda.empty_cache() + + # Inference process + n_query = x.size(1) - n_support + model.n_query = n_query + yq = np.repeat(range(n_way), n_query) + with torch.no_grad(): + #scores = model.set_forward(x) # (80, 5) + scores = set_forward_ViTProtonet(model, x) + _, topk_labels = scores.data.topk(1, 1, True, True) + topk_ind = topk_labels.cpu().numpy() # (80, 1) + top1_correct = np.sum(topk_ind[:,0]==yq) + acc = top1_correct*100./(n_way*n_query) + acc_all.append(acc) + del scores, topk_labels + torch.cuda.empty_cache() + #print('Task %d : %4.2f%%'%(ti, acc)) + #print('Task %d : %4.2f%%, mean Acc: %4.2f'%(ti, acc, np.mean(np.array(acc_all)))) + if(ti%50==0): + print('Task %d : %4.2f%%, mean Acc: %4.2f'%(ti, acc, np.mean(np.array(acc_all)))) + + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print('Test Acc = %4.2f +- %4.2f%%'%(acc_mean, 1.96*acc_std/np.sqrt(iter_num))) + +def run_single_testset(params): + seed = 0 + #print("set seed = %d" % seed) + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) + torch.cuda.manual_seed_all(seed) + torch.backends.cudnn.deterministic = True + torch.backends.cudnn.benchmark = False + + #np.random.seed(10) + #params = parse_args('train') + + #params = parse_args() + + image_size = 224 + iter_num = 1000 + n_query = 15 + n_pseudo = 75 + #print('n_pseudo: ', n_pseudo) + + print('Loading target dataset!:', params.testset) + if params.testset in ['cub', 'cars', 'places', 'plantae']: + novel_file = os.path.join(params.data_dir, params.testset, 'novel.json') + datamgr = SetDataManager(image_size, n_query=n_query, n_way=params.test_n_way, n_support=params.n_shot, n_eposide=iter_num) + novel_loader = datamgr.get_data_loader(novel_file, aug=False) + + else: + few_shot_params = dict(n_way = params.test_n_way , n_support = params.n_shot) + if params.testset in ["ISIC"]: + datamgr = ISIC_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + elif params.testset in ["EuroSAT"]: + datamgr = EuroSAT_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + elif params.testset in ["CropDisease"]: + datamgr = CropDisease_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + elif params.testset in ["ChestX"]: + datamgr = Chest_few_shot.SetDataManager(image_size, n_eposide = iter_num, n_query = n_query, **few_shot_params) + novel_loader = datamgr.get_data_loader(aug = False ) + + finetune(novel_loader, n_pseudo=n_pseudo, n_way=params.test_n_way, n_support=params.n_shot) + +if __name__=='__main__': + params = parse_args(script='train') + #for tmp_testset in ['cub', 'cars', 'places', 'plantae', 'ChestX', 'ISIC', 'EuroSAT', 'CropDisease']: + #for tmp_testset in ['EuroSAT', 'CropDisease']: + #for tmp_testset in ['CropDisease']: + #for tmp_testset in ['EuroSAT', 'plantae']: + #for tmp_testset in ['ISIC']: + #for tmp_testset in ['ChestX', 'ISIC']: + for tmp_testset in ['EuroSAT']: + params.testset = tmp_testset + run_single_testset(params) diff --git a/gdown.yml b/gdown.yml new file mode 100644 index 0000000000000000000000000000000000000000..7b692c34eb3dc1232e7d0cead95740a5483ead99 --- /dev/null +++ b/gdown.yml @@ -0,0 +1,11 @@ +name: gdown +channels: + - nvidia + - conda-forge +dependencies: + - python=3.11.* + - pip + - pip: + - gdown + + diff --git a/kaggle.yml b/kaggle.yml new file mode 100644 index 0000000000000000000000000000000000000000..bbb5d0efdf182c6d7d4da86bfcc21dcbbd3cebc6 --- /dev/null +++ b/kaggle.yml @@ -0,0 +1,11 @@ +name: kaggle +channels: + - nvidia + - conda-forge +dependencies: + - python=3.11.* + - pip + - pip: + - kaggle + + diff --git a/metatrain_CausalStyle_RN.py b/metatrain_CausalStyle_RN.py new file mode 100644 index 0000000000000000000000000000000000000000..13215867c0866c1bf44ceb7ca9ee009c4f6f7b1c --- /dev/null +++ b/metatrain_CausalStyle_RN.py @@ -0,0 +1,163 @@ +import numpy as np +import torch +import torch.optim +import os +import random + +from methods import backbone +from methods.backbone_multiblock import model_dict +from data.datamgr import SimpleDataManager, SetDataManager +#from methods.StyleAdv_RN_GNN import StyleAdvGNN +from methods.CausalStyle_RN_GNN import CausalStyleGNN + +from options import parse_args, get_resume_file, load_warmup_state +from test_function_fwt_benchmark import test_bestmodel +from test_function_bscdfsl_benchmark import test_bestmodel_bscdfsl + + +def train(base_loader, val_loader, model, start_epoch, stop_epoch, params): + + # get optimizer and checkpoint path + optimizer = torch.optim.Adam(model.parameters()) + if not os.path.isdir(params.checkpoint_dir): + os.makedirs(params.checkpoint_dir) + + # for validation + max_acc = 0 + total_it = 0 + + # start + for epoch in range(start_epoch, stop_epoch): + model.train() + total_it = model.train_loop(epoch, base_loader, optimizer, total_it) #model are called by reference, no need to return + model.eval() + + acc = model.test_loop( val_loader) + if acc > max_acc : + print("best model! save...") + max_acc = acc + outfile = os.path.join(params.checkpoint_dir, 'best_model.tar') + torch.save({'epoch':epoch, 'state':model.state_dict()}, outfile) + else: + print("GG! best accuracy {:f}".format(max_acc)) + + #if ((epoch + 1) % params.save_freq==0) or (epoch==stop_epoch-1): + if(epoch == stop_epoch-1): + outfile = os.path.join(params.checkpoint_dir, '{:d}.tar'.format(epoch)) + torch.save({'epoch':epoch, 'state':model.state_dict()}, outfile) + + return model + + +def record_test_result(params): + acc_file_path = os.path.join(params.checkpoint_dir, 'acc.txt') + acc_file = open(acc_file_path,'w') + epoch_id = -1 + print('epoch', epoch_id, 'miniImagenet:', 'cub:', 'cars:', 'places:', 'plantae:', file = acc_file) + name = params.name + n_shot = params.n_shot + method = params.method + #test_bestmodel(acc_file, name, method, 'miniImagenet', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'cub', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'cars', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'places', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'plantae', n_shot, epoch_id) + + acc_file.close() + return + + +def record_test_result_bscdfsl(params): + print('hhhhhhh testing for bscdfsl') + acc_file_path = os.path.join(params.checkpoint_dir, 'acc_bscdfsl.txt') + acc_file = open(acc_file_path,'w') + epoch_id = -1 + print('epoch', epoch_id, 'ChestX:', 'ISIC:', 'EuroSAT:', 'CropDisease', file = acc_file) + name = params.name + n_shot = params.n_shot + method = params.method + test_bestmodel_bscdfsl(acc_file, name, method, 'ChestX', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, 'ISIC', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, 'EuroSAT', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, 'CropDisease', n_shot, epoch_id) + + acc_file.close() + return + + +# --- main function --- +if __name__=='__main__': + #fix seed + seed = 0 + print("set seed = %d" % seed) + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) + torch.cuda.manual_seed_all(seed) + torch.backends.cudnn.deterministic = True + torch.backends.cudnn.benchmark = False + + # parser argument + params = parse_args('train') + + # output and tensorboard dir + params.tf_dir = '%s/log/%s'%(params.save_dir, params.name) + params.checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if not os.path.isdir(params.checkpoint_dir): + os.makedirs(params.checkpoint_dir) + + # dataloader + print('\n--- prepare dataloader ---') + print(' train with single seen domain {}'.format(params.dataset)) + base_file = os.path.join(params.data_dir, params.dataset, 'base.json') + val_file = os.path.join(params.data_dir, params.dataset, 'val.json') + + # model + print('\n--- build model ---') + image_size = 224 + + #if test_n_way is smaller than train_n_way, reduce n_query to keep batch size small + n_query = max(1, int(16* params.test_n_way/params.train_n_way)) + + train_few_shot_params = dict(n_way = params.train_n_way, n_support = params.n_shot) + base_datamgr = SetDataManager(image_size, n_query = n_query, **train_few_shot_params) + base_loader = base_datamgr.get_data_loader( base_file , aug = params.train_aug ) + + test_few_shot_params = dict(n_way = params.test_n_way, n_support = params.n_shot) + val_datamgr = SetDataManager(image_size, n_query = n_query, **test_few_shot_params) + val_loader = val_datamgr.get_data_loader( val_file, aug = False) + + model = CausalStyleGNN( model_dict[params.model], tf_path=params.tf_dir, **train_few_shot_params) + model = model.cuda() + + # load model + start_epoch = params.start_epoch + stop_epoch = params.stop_epoch + if params.resume != '': + resume_file = get_resume_file('%s/checkpoints/%s'%(params.save_dir, params.resume), params.resume_epoch) + if resume_file is not None: + tmp = torch.load(resume_file) + start_epoch = tmp['epoch']+1 + model.load_state_dict(tmp['state']) + print(' resume the training with at {} epoch (model file {})'.format(start_epoch, params.resume)) + else: + if params.warmup == 'gg3b0': + raise Exception('Must provide the pre-trained feature encoder file using --warmup option!') + state = load_warmup_state('%s/checkpoints/%s'%(params.save_dir, params.warmup), params.method) + model.feature.load_state_dict(state, strict=False) + + import time + #start =time.clock() + start =time.perf_counter() + # training + print('\n--- start the training ---') + model = train(base_loader, val_loader, model, start_epoch, stop_epoch, params) + #end=time.clock() + end =time.perf_counter() + print('Running time: %s Seconds: %s Min: %s Min per epoch'%(end-start, (end-start)/60, (end-start)/60/params.stop_epoch)) + + # testing + record_test_result(params) + # testing bscdfsl + record_test_result_bscdfsl(params) + diff --git a/metatrain_StyleAdv_RN.py b/metatrain_StyleAdv_RN.py new file mode 100644 index 0000000000000000000000000000000000000000..97af228ae70f8e77d87abc2cbbc5eff0763ab3d7 --- /dev/null +++ b/metatrain_StyleAdv_RN.py @@ -0,0 +1,162 @@ +import numpy as np +import torch +import torch.optim +import os +import random + +from methods import backbone +from methods.backbone_multiblock import model_dict +from data.datamgr import SimpleDataManager, SetDataManager +from methods.StyleAdv_RN_GNN import StyleAdvGNN + +from options import parse_args, get_resume_file, load_warmup_state +from test_function_fwt_benchmark import test_bestmodel +from test_function_bscdfsl_benchmark import test_bestmodel_bscdfsl + + +def train(base_loader, val_loader, model, start_epoch, stop_epoch, params): + + # get optimizer and checkpoint path + optimizer = torch.optim.Adam(model.parameters()) + if not os.path.isdir(params.checkpoint_dir): + os.makedirs(params.checkpoint_dir) + + # for validation + max_acc = 0 + total_it = 0 + + # start + for epoch in range(start_epoch, stop_epoch): + model.train() + total_it = model.train_loop(epoch, base_loader, optimizer, total_it) #model are called by reference, no need to return + model.eval() + + acc = model.test_loop( val_loader) + if acc > max_acc : + print("best model! save...") + max_acc = acc + outfile = os.path.join(params.checkpoint_dir, 'best_model.tar') + torch.save({'epoch':epoch, 'state':model.state_dict()}, outfile) + else: + print("GG! best accuracy {:f}".format(max_acc)) + + #if ((epoch + 1) % params.save_freq==0) or (epoch==stop_epoch-1): + if(epoch == stop_epoch-1): + outfile = os.path.join(params.checkpoint_dir, '{:d}.tar'.format(epoch)) + torch.save({'epoch':epoch, 'state':model.state_dict()}, outfile) + + return model + + +def record_test_result(params): + acc_file_path = os.path.join(params.checkpoint_dir, 'acc.txt') + acc_file = open(acc_file_path,'w') + epoch_id = -1 + print('epoch', epoch_id, 'miniImagenet:', 'cub:', 'cars:', 'places:', 'plantae:', file = acc_file) + name = params.name + n_shot = params.n_shot + method = params.method + test_bestmodel(acc_file, name, method, 'miniImagenet', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'cub', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'cars', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'places', n_shot, epoch_id) + test_bestmodel(acc_file, name, method, 'plantae', n_shot, epoch_id) + + acc_file.close() + return + + +def record_test_result_bscdfsl(params): + print('hhhhhhh testing for bscdfsl') + acc_file_path = os.path.join(params.checkpoint_dir, 'acc_bscdfsl.txt') + acc_file = open(acc_file_path,'w') + epoch_id = -1 + print('epoch', epoch_id, 'ChestX:', 'ISIC:', 'EuroSAT:', 'CropDisease', file = acc_file) + name = params.name + n_shot = params.n_shot + method = params.method + test_bestmodel_bscdfsl(acc_file, name, method, 'ChestX', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, 'ISIC', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, 'EuroSAT', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, 'CropDisease', n_shot, epoch_id) + + acc_file.close() + return + + +# --- main function --- +if __name__=='__main__': + #fix seed + seed = 0 + print("set seed = %d" % seed) + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) + torch.cuda.manual_seed_all(seed) + torch.backends.cudnn.deterministic = True + torch.backends.cudnn.benchmark = False + + # parser argument + params = parse_args('train') + + # output and tensorboard dir + params.tf_dir = '%s/log/%s'%(params.save_dir, params.name) + params.checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if not os.path.isdir(params.checkpoint_dir): + os.makedirs(params.checkpoint_dir) + + # dataloader + print('\n--- prepare dataloader ---') + print(' train with single seen domain {}'.format(params.dataset)) + base_file = os.path.join(params.data_dir, params.dataset, 'base.json') + val_file = os.path.join(params.data_dir, params.dataset, 'val.json') + + # model + print('\n--- build model ---') + image_size = 224 + + #if test_n_way is smaller than train_n_way, reduce n_query to keep batch size small + n_query = max(1, int(16* params.test_n_way/params.train_n_way)) + + train_few_shot_params = dict(n_way = params.train_n_way, n_support = params.n_shot) + base_datamgr = SetDataManager(image_size, n_query = n_query, **train_few_shot_params) + base_loader = base_datamgr.get_data_loader( base_file , aug = params.train_aug ) + + test_few_shot_params = dict(n_way = params.test_n_way, n_support = params.n_shot) + val_datamgr = SetDataManager(image_size, n_query = n_query, **test_few_shot_params) + val_loader = val_datamgr.get_data_loader( val_file, aug = False) + + model = StyleAdvGNN( model_dict[params.model], tf_path=params.tf_dir, **train_few_shot_params) + model = model.cuda() + + # load model + start_epoch = params.start_epoch + stop_epoch = params.stop_epoch + if params.resume != '': + resume_file = get_resume_file('%s/checkpoints/%s'%(params.save_dir, params.resume), params.resume_epoch) + if resume_file is not None: + tmp = torch.load(resume_file) + start_epoch = tmp['epoch']+1 + model.load_state_dict(tmp['state']) + print(' resume the training with at {} epoch (model file {})'.format(start_epoch, params.resume)) + else: + if params.warmup == 'gg3b0': + raise Exception('Must provide the pre-trained feature encoder file using --warmup option!') + state = load_warmup_state('%s/checkpoints/%s'%(params.save_dir, params.warmup), params.method) + model.feature.load_state_dict(state, strict=False) + + import time + #start =time.clock() + start =time.perf_counter() + # training + print('\n--- start the training ---') + model = train(base_loader, val_loader, model, start_epoch, stop_epoch, params) + #end=time.clock() + end =time.perf_counter() + print('Running time: %s Seconds: %s Min: %s Min per epoch'%(end-start, (end-start)/60, (end-start)/60/params.stop_epoch)) + + # testing + #record_test_result(params) + # testing bscdfsl + #record_test_result_bscdfsl(params) + diff --git a/metatrain_StyleAdv_ViT.py b/metatrain_StyleAdv_ViT.py new file mode 100644 index 0000000000000000000000000000000000000000..6fe4c4765d9448a37f53607411e1f36c71be3c51 --- /dev/null +++ b/metatrain_StyleAdv_ViT.py @@ -0,0 +1,240 @@ +import sys +import datetime +import random +import numpy as np +import time +import torch +import torch.backends.cudnn as cudnn +import json + +from pathlib import Path +from torch.utils.tensorboard import SummaryWriter + +from timm.data import Mixup +from timm.loss import LabelSmoothingCrossEntropy, SoftTargetCrossEntropy +from timm.scheduler import create_scheduler +from timm.optim import create_optimizer +from timm.utils import NativeScaler, get_state_dict, ModelEma + +#from models.pmf_engine import train_one_epoch, evaluate +#from models.pmf_engine_styleAdv_20221102 import train_one_epoch_styleAdv, evaluate +#from methods.pmf_engine_styleAdv_20221102 import train_one_epoch_styleAdv, evaluate +from methods.engine_StyleAdv_ViT import train_one_epoch_styleAdv, evaluate +#import pmf_utils.deit_util as utils +import utils.deit_util as utils +#from pmf_datasets import get_loaders +#from pmf_datasets import get_loaders_withGlobalID +from data.pmf_datasets import get_loaders_withGlobalID +#from pmf_utils.args import get_args_parser +from utils.args import get_args_parser +#from models import get_model +#from methods.cvpr2023_load_models_20221102 import get_model +from methods.load_ViT_models import get_model + +#lr_classifier = 5e-5 +#lr_classifier = 0.01 +lr_classifier = 0.001 +#lr_classifier = 0.0001 + +def main(args): + utils.init_distributed_mode(args) + + print(args) + device = torch.device(args.device) + + # fix the seed for reproducibility + seed = args.seed + utils.get_rank() + args.seed = seed + torch.manual_seed(seed) + np.random.seed(seed) + random.seed(seed) + + cudnn.benchmark = True + + output_dir = Path(args.output_dir) + if utils.is_main_process(): + output_dir.mkdir(parents=True, exist_ok=True) + with (output_dir / "log.txt").open("a") as f: + f.write(" ".join(sys.argv) + "\n") + + ############################################## + # Data loaders + num_tasks = utils.get_world_size() + global_rank = utils.get_rank() + data_loader_train, data_loader_val = get_loaders_withGlobalID(args, num_tasks, global_rank) + + ############################################## + # Mixup regularization (by default OFF) + mixup_fn = None + mixup_active = args.mixup > 0 or args.cutmix > 0. or args.cutmix_minmax is not None + if mixup_active: + mixup_fn = Mixup( + mixup_alpha=args.mixup, cutmix_alpha=args.cutmix, cutmix_minmax=args.cutmix_minmax, + prob=args.mixup_prob, switch_prob=args.mixup_switch_prob, mode=args.mixup_mode, + label_smoothing=args.smoothing, num_classes=args.nClsEpisode) + + ############################################## + # Model + print(f"Creating model: ProtoNet {args.arch}") + model = get_model(backbone = 'vit_small', classifier='protonet', styleAdv=True) + #model = get_model(args) + model.to(device) + + model_ema = None # (by default OFF) + if args.model_ema: + # Important to create EMA model after cuda(), DP wrapper, and AMP but before SyncBN and DDP wrapper + model_ema = ModelEma( + model, + decay=args.model_ema_decay, + device='cpu' if args.model_ema_force_cpu else '', + resume='') + + model_without_ddp = model + if args.distributed: + model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], + find_unused_parameters=args.unused_params) + model_without_ddp = model.module + n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad) + print('number of params:', n_parameters) + + ############################################## + # Optimizer & scheduler & criterion + if args.fp16: + scale = 1 / 8 # the default lr is for 8 GPUs + linear_scaled_lr = args.lr * utils.get_world_size() * scale + args.lr = linear_scaled_lr + + loss_scaler = NativeScaler() if args.fp16 else None + + #optimizer = create_optimizer(args, model_without_ddp) + ''' + optimizer = torch.optim.SGD( + [p for p in model_without_ddp.parameters() if p.requires_grad], + args.lr, + momentum=args.momentum, + weight_decay=0, # no weight decay for fine-tuning + ) + ''' + optimizer = torch.optim.SGD( + [ {'params': p for p in model_without_ddp.feature.parameters() if p.requires_grad}, + {'params': model_without_ddp.classifier.parameters(), 'lr': lr_classifier}], + args.lr, + momentum=args.momentum, + weight_decay=0, # no weight decay for fine-tuning + ) + lr_scheduler, _ = create_scheduler(args, optimizer) + + if args.mixup > 0.: + # smoothing is handled with mixup label transform + criterion = SoftTargetCrossEntropy() + elif args.smoothing: + criterion = LabelSmoothingCrossEntropy(smoothing=args.smoothing) + else: + criterion = torch.nn.CrossEntropyLoss() + + ############################################## + # Resume training from ckpt (model, optimizer, lr_scheduler, epoch, model_ema, scaler) + if args.resume: + if args.resume.startswith('https'): + checkpoint = torch.hub.load_state_dict_from_url( + args.resume, map_location='cpu', check_hash=True) + else: + checkpoint = torch.load(args.resume, map_location='cpu') + + model_without_ddp.load_state_dict(checkpoint['model']) + + if not args.eval and 'optimizer' in checkpoint and 'lr_scheduler' in checkpoint and 'epoch' in checkpoint: + optimizer.load_state_dict(checkpoint['optimizer']) + lr_scheduler.load_state_dict(checkpoint['lr_scheduler']) + args.start_epoch = checkpoint['epoch'] + 1 + if args.model_ema: + utils._load_checkpoint_for_ema(model_ema, checkpoint['model_ema']) + if 'scaler' in checkpoint: + loss_scaler.load_state_dict(checkpoint['scaler']) + + print(f'Resume from {args.resume} at epoch {args.start_epoch}.') + + + ############################################## + # Test + test_stats = evaluate(data_loader_val, model, criterion, device, args.seed+10000) + print(f"Accuracy of the network on dataset_val: {test_stats['acc1']:.1f}%") + if args.output_dir and utils.is_main_process(): + test_stats['epoch'] = -1 + with (output_dir / "log.txt").open("a") as f: + f.write(json.dumps(test_stats) + "\n") + + if args.eval: + return + + ############################################## + # Training + if utils.is_main_process(): + writer = SummaryWriter(log_dir=str(output_dir)) + else: + writer = None + + print(f"Start training for {args.epochs} epochs") + start_time = time.time() + #max_accuracy = test_stats['acc1'] + max_accuracy = 0.0 + + for epoch in range(args.start_epoch, args.epochs): + print('args.start_epoch:', args.start_epoch, 'args.epochs:', args.epochs, 'tmp epoch:', epoch) + train_stats = train_one_epoch_styleAdv( + data_loader_train, model, criterion, optimizer, epoch, device, + loss_scaler, args.fp16, args.clip_grad, model_ema, mixup_fn, writer, + set_training_mode=False # TODO: may need eval mode for finetuning + ) + + lr_scheduler.step(epoch) + + test_stats = evaluate(data_loader_val, model, criterion, device, args.seed+10000) + + log_stats = {**{f'train_{k}': v for k, v in train_stats.items()}, + **{f'test_{k}': v for k, v in test_stats.items()}, + 'epoch': epoch, + 'n_parameters': n_parameters} + + if args.output_dir: + checkpoint_paths = [output_dir / 'checkpoint.pth', output_dir / 'best.pth'] + for checkpoint_path in checkpoint_paths: + state_dict = { + 'model': model_without_ddp.state_dict(), + 'optimizer': optimizer.state_dict(), + 'lr_scheduler': lr_scheduler.state_dict(), + 'epoch': epoch, + 'model_ema': get_state_dict(model_ema) if args.model_ema else None, + 'args': args, + } + if loss_scaler is not None: + state_dict['scalar'] = loss_scaler.state_dict() + utils.save_on_master(state_dict, checkpoint_path) + + if test_stats["acc1"] <= max_accuracy: + break # do not save best.pth + + print(f"Accuracy of the network on dataset_val: {test_stats['acc1']:.1f}%") + max_accuracy = max(max_accuracy, test_stats["acc1"]) + print(f'Max accuracy: {max_accuracy:.2f}%') + + if args.output_dir and utils.is_main_process(): + log_stats['best_test_acc'] = max_accuracy + with (output_dir / "log.txt").open("a") as f: + f.write(json.dumps(log_stats) + "\n") + + total_time = time.time() - start_time + total_time_str = str(datetime.timedelta(seconds=int(total_time))) + print('Training time {}'.format(total_time_str)) + + if utils.is_main_process(): + writer.close() + import tables + tables.file._open_files.close_all() + + +if __name__ == '__main__': + parser = get_args_parser() + args = parser.parse_args() + + main(args) diff --git a/methods/CausalStyle_RN_GNN.py b/methods/CausalStyle_RN_GNN.py new file mode 100644 index 0000000000000000000000000000000000000000..c4aa132ea4e586c8e0f83168bc66eb5713027c3a --- /dev/null +++ b/methods/CausalStyle_RN_GNN.py @@ -0,0 +1,368 @@ +import torch +import torch.nn as nn +import numpy as np +import random + +from methods.gnn import GNN_nl +from methods import backbone_multiblock +from methods.tool_func import * +#from methods.meta_template_StyleAdv_RN_GNN import MetaTemplate +from methods.meta_template_CausalStyle_RN_GNN import MetaTemplate + + +class CausalStyleGNN(MetaTemplate): + maml=False + def __init__(self, model_func, n_way, n_support, tf_path=None): + super(CausalStyleGNN, self).__init__(model_func, n_way, n_support, tf_path=tf_path) + + # loss function + self.loss_fn = nn.CrossEntropyLoss() + + # metric function + self.fc = nn.Sequential(nn.Linear(self.feat_dim, 128), nn.BatchNorm1d(128, track_running_stats=False)) if not self.maml else nn.Sequential(backbone.Linear_fw(self.feat_dim, 128), backbone.BatchNorm1d_fw(128, track_running_stats=False)) + self.gnn = GNN_nl(128 + self.n_way, 96, self.n_way) + + # for global classifier + self.method = 'GnnNet' + self.classifier = nn.Linear(self.feature.final_feat_dim, 64) + + # for global domain noises + mean = 0 + std = 0.1 + #std = 0.05 + #std = 0.02 + # domains as gaussian noise + #domains = torch.randn((85,512)) * std + mean #fixed version + #domains = nn.Parameter(torch.randn((85,512)) * std + mean) #learnable version + domains = nn.Parameter(torch.randn((85,512))) #learnable version but not with gauss + self.domains = domains.cuda() + + # fix label for training the metric function 1*nw(1 + ns)*nw + support_label = torch.from_numpy(np.repeat(range(self.n_way), self.n_support)).unsqueeze(1) + support_label = torch.zeros(self.n_way*self.n_support, self.n_way).scatter(1, support_label, 1).view(self.n_way, self.n_support, self.n_way) + support_label = torch.cat([support_label, torch.zeros(self.n_way, 1, n_way)], dim=1) + self.support_label = support_label.view(1, -1, self.n_way) + + def cuda(self): + self.feature.cuda() + self.fc.cuda() + self.gnn.cuda() + self.classifier.cuda() + self.support_label = self.support_label.cuda() + return self + + def set_forward(self,x,is_feature=False): + x = x.cuda() + + if is_feature: + # reshape the feature tensor: n_way * n_s + 15 * f + assert(x.size(1) == self.n_support + 15) + z = self.fc(x.view(-1, *x.size()[2:])) + z = z.view(self.n_way, -1, z.size(1)) + else: + # get feature using encoder + x = x.view(-1, *x.size()[2:]) + z = self.fc(self.feature(x)) + z = z.view(self.n_way, -1, z.size(1)) + + # stack the feature for metric function: n_way * n_s + n_q * f -> n_q * [1 * n_way(n_s + 1) * f] + z_stack = [torch.cat([z[:, :self.n_support], z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, z.size(2)) for i in range(self.n_query)] + assert(z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores = self.forward_gnn(z_stack) + return scores + + + + def forward_gnn(self, zs): + # gnn inp: n_q * n_way(n_s + 1) * f + nodes = torch.cat([torch.cat([z, self.support_label], dim=2) for z in zs], dim=0) + scores = self.gnn(nodes) + + # n_q * n_way(n_s + 1) * n_way -> (n_way * n_q) * n_way + scores = scores.view(self.n_query, self.n_way, self.n_support + 1, self.n_way)[:, :, -1].permute(1, 0, 2).contiguous().view(-1, self.n_way) + return scores + + + def set_forward_loss(self, x): + y_query = torch.from_numpy(np.repeat(range( self.n_way ), self.n_query)) + y_query = y_query.cuda() + scores = self.set_forward(x) + loss = self.loss_fn(scores, y_query) + return scores, loss + + + def adversarial_attack_Incre(self, x_ori, y_ori, epsilon_list): + x_ori = x_ori.cuda() + y_ori = y_ori.cuda() + x_size = x_ori.size() + x_ori = x_ori.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + y_ori = y_ori.view(x_size[0]*x_size[1]) + + # if not adv, set defalut = 'None' + adv_style_mean_block1, adv_style_std_block1 = 'None', 'None' + adv_style_mean_block2, adv_style_std_block2 = 'None', 'None' + adv_style_mean_block3, adv_style_std_block3 = 'None', 'None' + + # forward and set the grad = True + blocklist = 'block123' + + if('1' in blocklist and epsilon_list[0] != 0 ): + # forward block1 + x_ori_block1 = self.feature.forward_block1(x_ori) + feat_size_block1 = x_ori_block1.size() + ori_style_mean_block1, ori_style_std_block1 = calc_mean_std(x_ori_block1) + # set them as learnable parameters + ori_style_mean_block1 = torch.nn.Parameter(ori_style_mean_block1) + ori_style_std_block1 = torch.nn.Parameter(ori_style_std_block1) + ori_style_mean_block1.requires_grad_() + ori_style_std_block1.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block1 = (x_ori_block1 - ori_style_mean_block1.detach().expand(feat_size_block1)) / ori_style_std_block1.detach().expand(feat_size_block1) + x_ori_block1 = x_normalized_block1 * ori_style_std_block1.expand(feat_size_block1) + ori_style_mean_block1.expand(feat_size_block1) + + # pass the rest model + x_ori_block2 = self.feature.forward_block2(x_ori_block1) + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + + # backward loss + ori_loss.backward() + + # collect datagrad + grad_ori_style_mean_block1 = ori_style_mean_block1.grad.detach() + grad_ori_style_std_block1 = ori_style_std_block1.grad.detach() + + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + + adv_style_mean_block1 = fgsm_attack(ori_style_mean_block1, epsilon, grad_ori_style_mean_block1) + adv_style_std_block1 = fgsm_attack(ori_style_std_block1, epsilon, grad_ori_style_std_block1) + + # add zero_grad + self.feature.zero_grad() + self.classifier.zero_grad() + + if('2' in blocklist and epsilon_list[1] != 0): + # forward block1 + x_ori_block1 = self.feature.forward_block1(x_ori) + # update adv_block1 + x_adv_block1 = changeNewAdvStyle(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + # forward block2 + x_ori_block2 = self.feature.forward_block2(x_adv_block1) + # calculate mean and std + feat_size_block2 = x_ori_block2.size() + ori_style_mean_block2, ori_style_std_block2 = calc_mean_std(x_ori_block2) + # set them as learnable parameters + ori_style_mean_block2 = torch.nn.Parameter(ori_style_mean_block2) + ori_style_std_block2 = torch.nn.Parameter(ori_style_std_block2) + ori_style_mean_block2.requires_grad_() + ori_style_std_block2.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block2 = (x_ori_block2 - ori_style_mean_block2.detach().expand(feat_size_block2)) / ori_style_std_block2.detach().expand(feat_size_block2) + x_ori_block2 = x_normalized_block2 * ori_style_std_block2.expand(feat_size_block2) + ori_style_mean_block2.expand(feat_size_block2) + # pass the rest model + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block2 = ori_style_mean_block2.grad.detach() + grad_ori_style_std_block2 = ori_style_std_block2.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block2 = fgsm_attack(ori_style_mean_block2, epsilon, grad_ori_style_mean_block2) + adv_style_std_block2 = fgsm_attack(ori_style_std_block2, epsilon, grad_ori_style_std_block2) + + # add zero_grad + self.feature.zero_grad() + self.classifier.zero_grad() + + if('3' in blocklist and epsilon_list[2] != 0): + # forward block1, block2, block3 + x_ori_block1 = self.feature.forward_block1(x_ori) + x_adv_block1 = changeNewAdvStyle(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + x_ori_block2 = self.feature.forward_block2(x_adv_block1) + x_adv_block2 = changeNewAdvStyle(x_ori_block2, adv_style_mean_block2, adv_style_std_block2, p_thred=0) + x_ori_block3 = self.feature.forward_block3(x_adv_block2) + # calculate mean and std + feat_size_block3 = x_ori_block3.size() + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3) + # set them as learnable parameters + ori_style_mean_block3 = torch.nn.Parameter(ori_style_mean_block3) + ori_style_std_block3 = torch.nn.Parameter(ori_style_std_block3) + ori_style_mean_block3.requires_grad_() + ori_style_std_block3.requires_grad_() + # contain ori_style_mean_block3 in the graph + x_normalized_block3 = (x_ori_block3 - ori_style_mean_block3.detach().expand(feat_size_block3)) / ori_style_std_block3.detach().expand(feat_size_block3) + x_ori_block3 = x_normalized_block3 * ori_style_std_block3.expand(feat_size_block3) + ori_style_mean_block3.expand(feat_size_block3) + # pass the rest model + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block3 = ori_style_mean_block3.grad.detach() + grad_ori_style_std_block3 = ori_style_std_block3.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block3 = fgsm_attack(ori_style_mean_block3, epsilon, grad_ori_style_mean_block3) + adv_style_std_block3 = fgsm_attack(ori_style_std_block3, epsilon, grad_ori_style_std_block3) + + return adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 + + + def set_statues_of_modules(self, flag): + if(flag=='eval'): + self.feature.eval() + self.fc.eval() + self.gnn.eval() + self.classifier.eval() + elif(flag=='train'): + self.feature.train() + self.fc.train() + self.gnn.train() + self.classifier.train() + return + + + + def perturb_via_latent_domains_v1(self, ori_fea): + ''' + add local gaussion noises + ''' + #print('ori_fea:', ori_fea.size()) #ori_fea: torch.Size([85, 512]) + mean = 0 + std = 0.1 + #std = 0.05 + #std = 0.02 + # domains as gaussian noise + domains = torch.randn(ori_fea.size()) * std + mean + domains = domains.cuda() + dom_fea = ori_fea + domains + return dom_fea + + def perturb_via_latent_domains_v2(self, ori_fea): + dom_fea = ori_fea + self.domains + return dom_fea + + def set_forward_loss_CausalStyle(self, x_ori, global_y, epsilon_list): + ################################################################## + # 0. first cp x_adv from x_ori + x_adv = x_ori + + ################################################################## + # 1. styleAdv + self.set_statues_of_modules('eval') + + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = self.adversarial_attack_Incre(x_ori, global_y, epsilon_list) + + self.feature.zero_grad() + self.fc.zero_grad() + self.classifier.zero_grad() + self.gnn.zero_grad() + + ################################################################# + # 2. forward and get loss + self.set_statues_of_modules('train') + + # define y_query for FSL + y_query = torch.from_numpy(np.repeat(range( self.n_way ), self.n_query)) + y_query = y_query.cuda() + + # forward x_ori + x_ori = x_ori.cuda() + x_size = x_ori.size() + x_ori = x_ori.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + global_y = global_y.view(x_size[0]*x_size[1]).cuda() + x_ori_block1 = self.feature.forward_block1(x_ori) + x_ori_block2 = self.feature.forward_block2(x_ori_block1) + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + + # ori cls global loss + scores_cls_ori = self.classifier.forward(x_ori_fea) + loss_cls_ori = self.loss_fn(scores_cls_ori, global_y) + acc_cls_ori = ( scores_cls_ori.max(1, keepdim=True)[1] == global_y ).type(torch.float).sum().item() / global_y.size()[0] + + # ori FSL scores and losses + x_ori_z = self.fc(x_ori_fea) + x_ori_z = x_ori_z.view(self.n_way, -1, x_ori_z.size(1)) + x_ori_z_stack = [torch.cat([x_ori_z[:, :self.n_support], x_ori_z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, x_ori_z.size(2)) for i in range(self.n_query)] + assert(x_ori_z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores_fsl_ori = self.forward_gnn(x_ori_z_stack) + loss_fsl_ori = self.loss_fn(scores_fsl_ori, y_query) + + # apply domain perturb + #x_dom_fea = self.perturb_via_latent_domains_v1(x_ori_fea) + x_dom_fea = self.perturb_via_latent_domains_v2(x_ori_fea) + + # forward x_dom + x_dom_z = self.fc(x_dom_fea) + x_dom_z = x_dom_z.view(self.n_way, -1, x_dom_z.size(1)) + x_dom_z_stack = [torch.cat([x_dom_z[:, :self.n_support], x_dom_z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, x_dom_z.size(2)) for i in range(self.n_query)] + assert(x_dom_z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores_fsl_dom = self.forward_gnn(x_dom_z_stack) + loss_fsl_dom = self.loss_fn(scores_fsl_dom, y_query) + + + # forward x_adv + x_adv = x_adv.cuda() + x_adv = x_adv.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + x_adv_block1 = self.feature.forward_block1(x_adv) + + x_adv_block1_newStyle = changeNewAdvStyle(x_adv_block1, adv_style_mean_block1, adv_style_std_block1, p_thred = P_THRED) + x_adv_block2 = self.feature.forward_block2(x_adv_block1_newStyle) + x_adv_block2_newStyle = changeNewAdvStyle(x_adv_block2, adv_style_mean_block2, adv_style_std_block2, p_thred = P_THRED) + x_adv_block3 = self.feature.forward_block3(x_adv_block2_newStyle) + x_adv_block3_newStyle = changeNewAdvStyle(x_adv_block3, adv_style_mean_block3, adv_style_std_block3, p_thred = P_THRED) + x_adv_block4 = self.feature.forward_block4(x_adv_block3_newStyle) + x_adv_fea = self.feature.forward_rest(x_adv_block4) + + # adv cls gloabl loss + scores_cls_adv = self.classifier.forward(x_adv_fea) + loss_cls_adv = self.loss_fn(scores_cls_adv, global_y) + acc_cls_adv = ( scores_cls_adv.max(1, keepdim=True)[1] == global_y ).type(torch.float).sum().item() / global_y.size()[0] + + # adv FSL scores and losses + x_adv_z = self.fc(x_adv_fea) + x_adv_z = x_adv_z.view(self.n_way, -1, x_adv_z.size(1)) + x_adv_z_stack = [torch.cat([x_adv_z[:, :self.n_support], x_adv_z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, x_adv_z.size(2)) for i in range(self.n_query)] + assert(x_adv_z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores_fsl_adv = self.forward_gnn(x_adv_z_stack) + loss_fsl_adv = self.loss_fn(scores_fsl_adv, y_query) + + #print('scores_fsl_adv:', scores_fsl_adv.mean(), 'loss_fsl_adv:', loss_fsl_adv, 'scores_cls_adv:', scores_cls_adv.mean(), 'loss_cls_adv:', loss_cls_adv) + #return scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv + return scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv, scores_fsl_dom, loss_fsl_dom diff --git a/methods/StyleAdv_RN_GNN.py b/methods/StyleAdv_RN_GNN.py new file mode 100644 index 0000000000000000000000000000000000000000..97f75018413cf43bffdc2a0098f5a13a276ece06 --- /dev/null +++ b/methods/StyleAdv_RN_GNN.py @@ -0,0 +1,322 @@ +import torch +import torch.nn as nn +import numpy as np +import random + +from methods.gnn import GNN_nl +from methods import backbone_multiblock +from methods.tool_func import * +from methods.meta_template_StyleAdv_RN_GNN import MetaTemplate + + +class StyleAdvGNN(MetaTemplate): + maml=False + def __init__(self, model_func, n_way, n_support, tf_path=None): + super(StyleAdvGNN, self).__init__(model_func, n_way, n_support, tf_path=tf_path) + + # loss function + self.loss_fn = nn.CrossEntropyLoss() + + # metric function + self.fc = nn.Sequential(nn.Linear(self.feat_dim, 128), nn.BatchNorm1d(128, track_running_stats=False)) if not self.maml else nn.Sequential(backbone.Linear_fw(self.feat_dim, 128), backbone.BatchNorm1d_fw(128, track_running_stats=False)) + self.gnn = GNN_nl(128 + self.n_way, 96, self.n_way) + + # for global classifier + self.method = 'GnnNet' + self.classifier = nn.Linear(self.feature.final_feat_dim, 64) + + # fix label for training the metric function 1*nw(1 + ns)*nw + support_label = torch.from_numpy(np.repeat(range(self.n_way), self.n_support)).unsqueeze(1) + support_label = torch.zeros(self.n_way*self.n_support, self.n_way).scatter(1, support_label, 1).view(self.n_way, self.n_support, self.n_way) + support_label = torch.cat([support_label, torch.zeros(self.n_way, 1, n_way)], dim=1) + self.support_label = support_label.view(1, -1, self.n_way) + + def cuda(self): + self.feature.cuda() + self.fc.cuda() + self.gnn.cuda() + self.classifier.cuda() + self.support_label = self.support_label.cuda() + return self + + def set_forward(self,x,is_feature=False): + x = x.cuda() + + if is_feature: + # reshape the feature tensor: n_way * n_s + 15 * f + assert(x.size(1) == self.n_support + 15) + z = self.fc(x.view(-1, *x.size()[2:])) + z = z.view(self.n_way, -1, z.size(1)) + else: + # get feature using encoder + x = x.view(-1, *x.size()[2:]) + z = self.fc(self.feature(x)) + z = z.view(self.n_way, -1, z.size(1)) + + # stack the feature for metric function: n_way * n_s + n_q * f -> n_q * [1 * n_way(n_s + 1) * f] + z_stack = [torch.cat([z[:, :self.n_support], z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, z.size(2)) for i in range(self.n_query)] + assert(z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores = self.forward_gnn(z_stack) + return scores + + + + def forward_gnn(self, zs): + # gnn inp: n_q * n_way(n_s + 1) * f + nodes = torch.cat([torch.cat([z, self.support_label], dim=2) for z in zs], dim=0) + scores = self.gnn(nodes) + + # n_q * n_way(n_s + 1) * n_way -> (n_way * n_q) * n_way + scores = scores.view(self.n_query, self.n_way, self.n_support + 1, self.n_way)[:, :, -1].permute(1, 0, 2).contiguous().view(-1, self.n_way) + return scores + + + def set_forward_loss(self, x): + y_query = torch.from_numpy(np.repeat(range( self.n_way ), self.n_query)) + y_query = y_query.cuda() + scores = self.set_forward(x) + loss = self.loss_fn(scores, y_query) + return scores, loss + + + def adversarial_attack_Incre(self, x_ori, y_ori, epsilon_list): + x_ori = x_ori.cuda() + y_ori = y_ori.cuda() + x_size = x_ori.size() + x_ori = x_ori.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + y_ori = y_ori.view(x_size[0]*x_size[1]) + + # if not adv, set defalut = 'None' + adv_style_mean_block1, adv_style_std_block1 = 'None', 'None' + adv_style_mean_block2, adv_style_std_block2 = 'None', 'None' + adv_style_mean_block3, adv_style_std_block3 = 'None', 'None' + + # forward and set the grad = True + blocklist = 'block123' + + if('1' in blocklist and epsilon_list[0] != 0 ): + # forward block1 + x_ori_block1 = self.feature.forward_block1(x_ori) + feat_size_block1 = x_ori_block1.size() + ori_style_mean_block1, ori_style_std_block1 = calc_mean_std(x_ori_block1) + # set them as learnable parameters + ori_style_mean_block1 = torch.nn.Parameter(ori_style_mean_block1) + ori_style_std_block1 = torch.nn.Parameter(ori_style_std_block1) + ori_style_mean_block1.requires_grad_() + ori_style_std_block1.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block1 = (x_ori_block1 - ori_style_mean_block1.detach().expand(feat_size_block1)) / ori_style_std_block1.detach().expand(feat_size_block1) + x_ori_block1 = x_normalized_block1 * ori_style_std_block1.expand(feat_size_block1) + ori_style_mean_block1.expand(feat_size_block1) + + # pass the rest model + x_ori_block2 = self.feature.forward_block2(x_ori_block1) + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + + # backward loss + ori_loss.backward() + + # collect datagrad + grad_ori_style_mean_block1 = ori_style_mean_block1.grad.detach() + grad_ori_style_std_block1 = ori_style_std_block1.grad.detach() + + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + + adv_style_mean_block1 = fgsm_attack(ori_style_mean_block1, epsilon, grad_ori_style_mean_block1) + adv_style_std_block1 = fgsm_attack(ori_style_std_block1, epsilon, grad_ori_style_std_block1) + + # add zero_grad + self.feature.zero_grad() + self.classifier.zero_grad() + + if('2' in blocklist and epsilon_list[1] != 0): + # forward block1 + x_ori_block1 = self.feature.forward_block1(x_ori) + # update adv_block1 + x_adv_block1 = changeNewAdvStyle(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + # forward block2 + x_ori_block2 = self.feature.forward_block2(x_adv_block1) + # calculate mean and std + feat_size_block2 = x_ori_block2.size() + ori_style_mean_block2, ori_style_std_block2 = calc_mean_std(x_ori_block2) + # set them as learnable parameters + ori_style_mean_block2 = torch.nn.Parameter(ori_style_mean_block2) + ori_style_std_block2 = torch.nn.Parameter(ori_style_std_block2) + ori_style_mean_block2.requires_grad_() + ori_style_std_block2.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block2 = (x_ori_block2 - ori_style_mean_block2.detach().expand(feat_size_block2)) / ori_style_std_block2.detach().expand(feat_size_block2) + x_ori_block2 = x_normalized_block2 * ori_style_std_block2.expand(feat_size_block2) + ori_style_mean_block2.expand(feat_size_block2) + # pass the rest model + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block2 = ori_style_mean_block2.grad.detach() + grad_ori_style_std_block2 = ori_style_std_block2.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block2 = fgsm_attack(ori_style_mean_block2, epsilon, grad_ori_style_mean_block2) + adv_style_std_block2 = fgsm_attack(ori_style_std_block2, epsilon, grad_ori_style_std_block2) + + # add zero_grad + self.feature.zero_grad() + self.classifier.zero_grad() + + if('3' in blocklist and epsilon_list[2] != 0): + # forward block1, block2, block3 + x_ori_block1 = self.feature.forward_block1(x_ori) + x_adv_block1 = changeNewAdvStyle(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + x_ori_block2 = self.feature.forward_block2(x_adv_block1) + x_adv_block2 = changeNewAdvStyle(x_ori_block2, adv_style_mean_block2, adv_style_std_block2, p_thred=0) + x_ori_block3 = self.feature.forward_block3(x_adv_block2) + # calculate mean and std + feat_size_block3 = x_ori_block3.size() + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3) + # set them as learnable parameters + ori_style_mean_block3 = torch.nn.Parameter(ori_style_mean_block3) + ori_style_std_block3 = torch.nn.Parameter(ori_style_std_block3) + ori_style_mean_block3.requires_grad_() + ori_style_std_block3.requires_grad_() + # contain ori_style_mean_block3 in the graph + x_normalized_block3 = (x_ori_block3 - ori_style_mean_block3.detach().expand(feat_size_block3)) / ori_style_std_block3.detach().expand(feat_size_block3) + x_ori_block3 = x_normalized_block3 * ori_style_std_block3.expand(feat_size_block3) + ori_style_mean_block3.expand(feat_size_block3) + # pass the rest model + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block3 = ori_style_mean_block3.grad.detach() + grad_ori_style_std_block3 = ori_style_std_block3.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block3 = fgsm_attack(ori_style_mean_block3, epsilon, grad_ori_style_mean_block3) + adv_style_std_block3 = fgsm_attack(ori_style_std_block3, epsilon, grad_ori_style_std_block3) + + return adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 + + + def set_statues_of_modules(self, flag): + if(flag=='eval'): + self.feature.eval() + self.fc.eval() + self.gnn.eval() + self.classifier.eval() + elif(flag=='train'): + self.feature.train() + self.fc.train() + self.gnn.train() + self.classifier.train() + return + + + def set_forward_loss_StyAdv(self, x_ori, global_y, epsilon_list): + ################################################################## + # 0. first cp x_adv from x_ori + x_adv = x_ori + + ################################################################## + # 1. styleAdv + self.set_statues_of_modules('eval') + + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = self.adversarial_attack_Incre(x_ori, global_y, epsilon_list) + + self.feature.zero_grad() + self.fc.zero_grad() + self.classifier.zero_grad() + self.gnn.zero_grad() + + ################################################################# + # 2. forward and get loss + self.set_statues_of_modules('train') + + # define y_query for FSL + y_query = torch.from_numpy(np.repeat(range( self.n_way ), self.n_query)) + y_query = y_query.cuda() + + # forward x_ori + x_ori = x_ori.cuda() + x_size = x_ori.size() + x_ori = x_ori.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + global_y = global_y.view(x_size[0]*x_size[1]).cuda() + x_ori_block1 = self.feature.forward_block1(x_ori) + x_ori_block2 = self.feature.forward_block2(x_ori_block1) + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + + # ori cls global loss + scores_cls_ori = self.classifier.forward(x_ori_fea) + loss_cls_ori = self.loss_fn(scores_cls_ori, global_y) + acc_cls_ori = ( scores_cls_ori.max(1, keepdim=True)[1] == global_y ).type(torch.float).sum().item() / global_y.size()[0] + + # ori FSL scores and losses + x_ori_z = self.fc(x_ori_fea) + x_ori_z = x_ori_z.view(self.n_way, -1, x_ori_z.size(1)) + x_ori_z_stack = [torch.cat([x_ori_z[:, :self.n_support], x_ori_z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, x_ori_z.size(2)) for i in range(self.n_query)] + assert(x_ori_z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores_fsl_ori = self.forward_gnn(x_ori_z_stack) + loss_fsl_ori = self.loss_fn(scores_fsl_ori, y_query) + + # forward x_adv + x_adv = x_adv.cuda() + x_adv = x_adv.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + x_adv_block1 = self.feature.forward_block1(x_adv) + + x_adv_block1_newStyle = changeNewAdvStyle(x_adv_block1, adv_style_mean_block1, adv_style_std_block1, p_thred = P_THRED) + x_adv_block2 = self.feature.forward_block2(x_adv_block1_newStyle) + x_adv_block2_newStyle = changeNewAdvStyle(x_adv_block2, adv_style_mean_block2, adv_style_std_block2, p_thred = P_THRED) + x_adv_block3 = self.feature.forward_block3(x_adv_block2_newStyle) + x_adv_block3_newStyle = changeNewAdvStyle(x_adv_block3, adv_style_mean_block3, adv_style_std_block3, p_thred = P_THRED) + x_adv_block4 = self.feature.forward_block4(x_adv_block3_newStyle) + x_adv_fea = self.feature.forward_rest(x_adv_block4) + + # adv cls gloabl loss + scores_cls_adv = self.classifier.forward(x_adv_fea) + loss_cls_adv = self.loss_fn(scores_cls_adv, global_y) + acc_cls_adv = ( scores_cls_adv.max(1, keepdim=True)[1] == global_y ).type(torch.float).sum().item() / global_y.size()[0] + + # adv FSL scores and losses + x_adv_z = self.fc(x_adv_fea) + x_adv_z = x_adv_z.view(self.n_way, -1, x_adv_z.size(1)) + x_adv_z_stack = [torch.cat([x_adv_z[:, :self.n_support], x_adv_z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, x_adv_z.size(2)) for i in range(self.n_query)] + assert(x_adv_z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + scores_fsl_adv = self.forward_gnn(x_adv_z_stack) + loss_fsl_adv = self.loss_fn(scores_fsl_adv, y_query) + + #print('scores_fsl_adv:', scores_fsl_adv.mean(), 'loss_fsl_adv:', loss_fsl_adv, 'scores_cls_adv:', scores_cls_adv.mean(), 'loss_cls_adv:', loss_cls_adv) + return scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv diff --git a/methods/StyleAdv_ViT_protonet.py b/methods/StyleAdv_ViT_protonet.py new file mode 100644 index 0000000000000000000000000000000000000000..2ca270bd774f82c6eaf94ce69df0aefaa5e9f877 --- /dev/null +++ b/methods/StyleAdv_ViT_protonet.py @@ -0,0 +1,357 @@ +import torch +import torch.nn as nn +import torch.nn.functional as F +import numpy as np + +from methods.tool_func import * + + +def preprocessing(x_fea): + # x_fea: [B, 197, 384] --> x_cls_fea [B, 1, 384], x_patch_fea [B, 384, 14, 14] + B, num, dim = x_fea.size()[0], x_fea.size()[1], x_fea.size()[2] + x_cls_fea = x_fea[:, :1, :] + x_patch_fea = x_fea[:,1:, :] + x_patch_fea = x_patch_fea.contiguous().view(B,dim,num-1).view(B, dim, 14, 14) + return x_cls_fea, x_patch_fea + +def postprocessing(x_cls_fea, x_patch_fea): + # x_cls_fea [B, 1, 384], x_patch_fea [B, 384, 14, 14] --> x_fea: [B, 197, 384] + B, num, dim = x_patch_fea.size()[0], x_patch_fea.size()[2]*x_patch_fea.size()[3]+1, x_patch_fea.size()[1] + x_patch_fea = x_patch_fea.contiguous().view(B,dim,num-1).view(B,num-1,dim) + x_fea = torch.cat((x_cls_fea, x_patch_fea), 1) + return x_fea + +def changeNewAdvStyle_ViT(vit_fea, new_styleAug_mean, new_styleAug_std, p_thred): + if(new_styleAug_mean=='None'): + return vit_fea + + #final + p = np.random.uniform() + if( p < p_thred): + return vit_fea + + cls_fea, input_fea = preprocessing(vit_fea) + feat_size = input_fea.size() + ori_style_mean, ori_style_std = calc_mean_std(input_fea) + #print('ori mean:', ori_style_mean.mean(), 'ori std:', ori_style_std.mean()) + #print('adv mean:', new_styleAug_mean.mean(), 'adv std:', new_styleAug_std.mean()) + #print('mean diff:', new_styleAug_mean.mean() - ori_style_mean.mean(), 'std diff:', new_styleAug_std.mean() - ori_style_std.mean()) + normalized_fea = (input_fea - ori_style_mean.expand(feat_size)) / ori_style_std.expand(feat_size) + styleAug_fea = normalized_fea * new_styleAug_std.expand(feat_size) + new_styleAug_mean.expand(feat_size) + styleAug_fea_vit = postprocessing(cls_fea, styleAug_fea) + return styleAug_fea_vit + +class ProtoNet(nn.Module): + def __init__(self, backbone): + super().__init__() + + # bias & scale of cosine classifier + self.bias = nn.Parameter(torch.FloatTensor(1).fill_(0), requires_grad=True) + self.scale_cls = nn.Parameter(torch.FloatTensor(1).fill_(10), requires_grad=True) + + # backbone + self.feature = backbone + final_feat_dim = 384 + self.classifier = nn.Linear(final_feat_dim, 64) + + self.loss_fn = nn.CrossEntropyLoss() + + def cos_classifier(self, w, f): + """ + w.shape = B, nC, d + f.shape = B, M, d + """ + f = F.normalize(f, p=2, dim=f.dim()-1, eps=1e-12) + w = F.normalize(w, p=2, dim=w.dim()-1, eps=1e-12) + + cls_scores = f @ w.transpose(1, 2) # B, M, nC + cls_scores = self.scale_cls * (cls_scores + self.bias) + return cls_scores + + def forward(self, supp_x, supp_y, x): + """ + supp_x.shape = [B, nSupp, C, H, W] + supp_y.shape = [B, nSupp] + x.shape = [B, nQry, C, H, W] + """ + num_classes = supp_y.max() + 1 # NOTE: assume B==1 + B, nSupp, C, H, W = supp_x.shape + supp_f = self.feature.forward(supp_x.contiguous().view(-1, C, H, W)) + supp_f = supp_f.view(B, nSupp, -1) + supp_y_1hot = F.one_hot(supp_y, num_classes).transpose(1, 2) # B, nC, nSupp + + # B, nC, nSupp x B, nSupp, d = B, nC, d + prototypes = torch.bmm(supp_y_1hot.float(), supp_f) + prototypes = prototypes / supp_y_1hot.sum(dim=2, keepdim=True) # NOTE: may div 0 if some classes got 0 images + + feat = self.feature.forward(x.view(-1, C, H, W)) + feat = feat.view(B, x.shape[1], -1) # B, nQry, d + + logits = self.cos_classifier(prototypes, feat) # B, nQry, nC + return logits + + def set_statues_of_modules(self, flag): + if(flag=='eval'): + self.feature.eval() + self.classifier.eval() + #self.scale_cls.eval() + #self.bias.eval() + elif(flag=='train'): + self.feature.train() + self.classifier.train() + #self.scale_cls.train() + #self.bias.train() + return + + + def forward_protonet(self, episode_f,supp_y, B, nSupp, nQuery, num_classes): + #print('episode_f:', episode_f.size()) + episode_f = episode_f.view(num_classes, nSupp + nQuery, -1) + #print('episode_f:', episode_f.size()) + fea_dim = episode_f.size()[-1] + supp_f = episode_f[:, :nSupp, :].contiguous().view(-1, fea_dim).unsqueeze(0) + query_f = episode_f[:, nSupp:, :].contiguous().view(-1, fea_dim).unsqueeze(0) + supp_y_1hot = F.one_hot(supp_y, num_classes).transpose(1, 2) # B, nC, nSupp + # B, nC, nSupp x B, nSupp, d = B, nC, d + prototypes = torch.bmm(supp_y_1hot.float(), supp_f) + prototypes = prototypes / supp_y_1hot.sum(dim=2, keepdim=True) # NOTE: may div 0 if some classes got 0 images + logits = self.cos_classifier(prototypes, query_f) # B, nQry, nC + return logits + + def adversarial_attack_Incre(self, x_ori, y_ori, epsilon_list): + x_ori = x_ori.cuda() + y_ori = y_ori.cuda() + x_size = x_ori.size() + x_ori = x_ori.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + y_ori = y_ori.view(x_size[0]*x_size[1]) + + # if not adv, set defalut = 'None' + adv_style_mean_block1, adv_style_std_block1 = 'None', 'None' + adv_style_mean_block2, adv_style_std_block2 = 'None', 'None' + adv_style_mean_block3, adv_style_std_block3 = 'None', 'None' + + # forward and set the grad = True + blocklist = 'block123' + + if('1' in blocklist and epsilon_list[0] != 0 ): + x_ori_block1 = self.feature.forward_block1(x_ori) + x_ori_block1_cls, x_ori_block1_P = preprocessing(x_ori_block1) + feat_size_block1 = x_ori_block1_P.size() + #print('x_ori_block1:', x_ori_block1.size(), x_ori_block1_P.size()) + ori_style_mean_block1, ori_style_std_block1 = calc_mean_std(x_ori_block1_P) + # set them as learnable parameters + ori_style_mean_block1 = torch.nn.Parameter(ori_style_mean_block1) + ori_style_std_block1 = torch.nn.Parameter(ori_style_std_block1) + ori_style_mean_block1.requires_grad_() + ori_style_std_block1.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block1 = (x_ori_block1_P - ori_style_mean_block1.detach().expand(feat_size_block1)) / ori_style_std_block1.detach().expand(feat_size_block1) + x_ori_block1_P = x_normalized_block1 * ori_style_std_block1.expand(feat_size_block1) + ori_style_mean_block1.expand(feat_size_block1) + x_ori_block1 = postprocessing(x_ori_block1_cls, x_ori_block1_P) + #print('x_ori_block1:', x_ori_block1.size()) + + # pass the rest model + x_ori_block2 = self.feature.forward_block2(x_ori_block1) + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + + # backward loss + ori_loss.backward() + + # collect datagrad + grad_ori_style_mean_block1 = ori_style_mean_block1.grad.detach() + grad_ori_style_std_block1 = ori_style_std_block1.grad.detach() + + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + + adv_style_mean_block1 = fgsm_attack(ori_style_mean_block1, epsilon, grad_ori_style_mean_block1) + adv_style_std_block1 = fgsm_attack(ori_style_std_block1, epsilon, grad_ori_style_std_block1) + + # add zero_grad + self.feature.zero_grad() + self.classifier.zero_grad() + + if('2' in blocklist and epsilon_list[1] != 0): + x_ori_block1 = self.feature.forward_block1(x_ori) + # update adv_block1 + x_adv_block1 = changeNewAdvStyle_ViT(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + # forward block2 + x_ori_block2 = self.feature.forward_block2(x_adv_block1) + # calculate mean and std + x_ori_block2_cls , x_ori_block2_P = preprocessing(x_ori_block2) + feat_size_block2 = x_ori_block2_P.size() + ori_style_mean_block2, ori_style_std_block2 = calc_mean_std(x_ori_block2_P) + # set them as learnable parameters + ori_style_mean_block2 = torch.nn.Parameter(ori_style_mean_block2) + ori_style_std_block2 = torch.nn.Parameter(ori_style_std_block2) + ori_style_mean_block2.requires_grad_() + ori_style_std_block2.requires_grad_() + # contain ori_style_mean_block1 in the graph + x_normalized_block2 = (x_ori_block2_P - ori_style_mean_block2.detach().expand(feat_size_block2)) / ori_style_std_block2.detach().expand(feat_size_block2) + x_ori_block2_P = x_normalized_block2 * ori_style_std_block2.expand(feat_size_block2) + ori_style_mean_block2.expand(feat_size_block2) + x_ori_block2 = postprocessing(x_ori_block2_cls, x_ori_block2_P) + # pass the rest model + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + #print('ori_pred:', ori_pred, 'ori_loss:', ori_loss, 'ori_acc:', ori_acc) + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block2 = ori_style_mean_block2.grad.detach() + grad_ori_style_std_block2 = ori_style_std_block2.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block2 = fgsm_attack(ori_style_mean_block2, epsilon, grad_ori_style_mean_block2) + adv_style_std_block2 = fgsm_attack(ori_style_std_block2, epsilon, grad_ori_style_std_block2) + #print('adv_style_mean_block2:', adv_style_mean_block2.size(), 'adv_style_std_block2:', adv_style_std_block2.size()) + + # add zero_grad + self.feature.zero_grad() + self.classifier.zero_grad() + + if('3' in blocklist and epsilon_list[2] != 0): + x_ori_block1 = self.feature.forward_block1(x_ori) + x_adv_block1 = changeNewAdvStyle_ViT(x_ori_block1, adv_style_mean_block1, adv_style_std_block1, p_thred=0) + x_ori_block2 = self.feature.forward_block2(x_adv_block1) + x_adv_block2 = changeNewAdvStyle_ViT(x_ori_block2, adv_style_mean_block2, adv_style_std_block2, p_thred=0) + x_ori_block3 = self.feature.forward_block3(x_adv_block2) + x_ori_block3_cls, x_ori_block3_P = preprocessing(x_ori_block3) + # calculate mean and std + feat_size_block3 = x_ori_block3_P.size() + ori_style_mean_block3, ori_style_std_block3 = calc_mean_std(x_ori_block3_P) + # set them as learnable parameters + ori_style_mean_block3 = torch.nn.Parameter(ori_style_mean_block3) + ori_style_std_block3 = torch.nn.Parameter(ori_style_std_block3) + ori_style_mean_block3.requires_grad_() + ori_style_std_block3.requires_grad_() + # contain ori_style_mean_block3 in the graph + x_normalized_block3 = (x_ori_block3_P - ori_style_mean_block3.detach().expand(feat_size_block3)) / ori_style_std_block3.detach().expand(feat_size_block3) + x_ori_block3_P = x_normalized_block3 * ori_style_std_block3.expand(feat_size_block3) + ori_style_mean_block3.expand(feat_size_block3) + x_ori_block3 = postprocessing(x_ori_block3_cls, x_ori_block3_P) + # pass the rest model + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + x_ori_output = self.classifier.forward(x_ori_fea) + # calculate initial pred, loss and acc + ori_pred = x_ori_output.max(1, keepdim=True)[1] + ori_loss = self.loss_fn(x_ori_output, y_ori) + ori_acc = (ori_pred == y_ori).type(torch.float).sum().item() / y_ori.size()[0] + # zero all the existing gradients + self.feature.zero_grad() + self.classifier.zero_grad() + # backward loss + ori_loss.backward() + # collect datagrad + grad_ori_style_mean_block3 = ori_style_mean_block3.grad.detach() + grad_ori_style_std_block3 = ori_style_std_block3.grad.detach() + # fgsm style attack + index = torch.randint(0, len(epsilon_list), (1, ))[0] + epsilon = epsilon_list[index] + adv_style_mean_block3 = fgsm_attack(ori_style_mean_block3, epsilon, grad_ori_style_mean_block3) + adv_style_std_block3 = fgsm_attack(ori_style_std_block3, epsilon, grad_ori_style_std_block3) + return adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 + + + + + def set_forward_loss_StyAdv(self, SupportTensor,QueryTensor,SupportLabel, QueryLabel, GlobalID_S,GlobalID_Q, epsilon_list): + ################################################################## + ''' + supp_x.shape = [B, nSupp, C, H, W] + supp_y.shape = [B, nSupp] + x.shape = [B, nQry, C, H, W] + + # to tacke the input data + x_ori: [5, 21, 3, 224, 224], global_y: [5, 21] + ''' + # to resize as x_ori: torch.Size([5, 21, 3, 224, 224]) global_y: torch.Size([5, 21]) + B = SupportTensor.size()[0] + num_classes = SupportLabel.max() + 1 # NOTE: assume B==1 + SupportTensor = SupportTensor.squeeze().view(num_classes, -1, 3, 224, 224) + QueryTensor = QueryTensor.squeeze().view(num_classes, -1, 3, 224, 224) + nSupp = SupportTensor.size()[1] + nQuery = QueryTensor.size()[1] + + x_ori = torch.cat((SupportTensor, QueryTensor), dim=1) + global_y = torch.cat((GlobalID_S.view(num_classes, nSupp), GlobalID_Q.view(num_classes, nQuery)), dim=1) + #print('x_ori:', x_ori.size(), 'global_y:', global_y.size()) + ################################################################## + + # 0. first cp x_adv from x_ori + x_adv = x_ori + + # 1. styleAdv + self.set_statues_of_modules('eval') + adv_style_mean_block1, adv_style_std_block1, adv_style_mean_block2, adv_style_std_block2, adv_style_mean_block3, adv_style_std_block3 = self.adversarial_attack_Incre(x_ori, global_y, epsilon_list) + self.feature.zero_grad() + self.classifier.zero_grad() + + # 2. forward and get loss + self.set_statues_of_modules('train') + x_ori = x_ori.cuda() + x_size = x_ori.size() + x_ori = x_ori.view(num_classes*(nSupp+nQuery), 3, 224, 224) + global_y = global_y.view(num_classes*(nSupp+nQuery)).cuda() + x_ori_block1 = self.feature.forward_block1(x_ori) + x_ori_block2 = self.feature.forward_block2(x_ori_block1) + x_ori_block3 = self.feature.forward_block3(x_ori_block2) + x_ori_block4 = self.feature.forward_block4(x_ori_block3) + x_ori_fea = self.feature.forward_rest(x_ori_block4) + + # 3. ori cls global loss + scores_cls_ori = self.classifier.forward(x_ori_fea) + loss_cls_ori = self.loss_fn(scores_cls_ori, global_y) + + # 4. ori FSL scores and losses + scores_fsl_ori = self.forward_protonet(x_ori_fea, SupportLabel,B, nSupp, nQuery, num_classes) + scores_fsl_ori = scores_fsl_ori.view(num_classes*nQuery, -1) + QueryLabel = QueryLabel.view(-1) + loss_fsl_ori = self.loss_fn(scores_fsl_ori, QueryLabel) + + # 5. forward StyleAdv + x_adv = x_adv.cuda() + x_adv = x_adv.view(x_size[0]*x_size[1], x_size[2], x_size[3], x_size[4]) + x_adv_block1 = self.feature.forward_block1(x_adv) + x_adv_block1_newStyle = changeNewAdvStyle_ViT(x_adv_block1, adv_style_mean_block1, adv_style_std_block1, p_thred = P_THRED) + x_adv_block2 = self.feature.forward_block2(x_adv_block1_newStyle) + x_adv_block2_newStyle = changeNewAdvStyle_ViT(x_adv_block2, adv_style_mean_block2, adv_style_std_block2, p_thred = P_THRED) + x_adv_block3 = self.feature.forward_block3(x_adv_block2_newStyle) + x_adv_block3_newStyle = changeNewAdvStyle_ViT(x_adv_block3, adv_style_mean_block3, adv_style_std_block3, p_thred = P_THRED) + x_adv_block4 = self.feature.forward_block4(x_adv_block3_newStyle) + x_adv_fea = self.feature.forward_rest(x_adv_block4) + + # 6. adv cls gloabl loss + scores_cls_adv = self.classifier.forward(x_adv_fea) + loss_cls_adv = self.loss_fn(scores_cls_adv, global_y) + + # 7. adv FSL scores and losses + scores_fsl_adv = self.forward_protonet(x_adv_fea, SupportLabel,B, nSupp, nQuery, num_classes) + scores_fsl_adv = scores_fsl_adv.view(num_classes*nQuery, -1) + loss_fsl_adv = self.loss_fn(scores_fsl_adv, QueryLabel) + + return scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv + + diff --git a/methods/ViT.py b/methods/ViT.py new file mode 100644 index 0000000000000000000000000000000000000000..b58099e1dddea9d1ae4de4f9e678864e655ab277 --- /dev/null +++ b/methods/ViT.py @@ -0,0 +1,284 @@ +import torch +import torch.nn as nn + +import math +from functools import partial +from .model_utils import trunc_normal_ + + +def drop_path(x, drop_prob: float = 0., training: bool = False): + if drop_prob == 0. or not training: + return x + keep_prob = 1 - drop_prob + shape = (x.shape[0],) + (1,) * (x.ndim - 1) # work with diff dim tensors, not just 2D ConvNets + random_tensor = keep_prob + torch.rand(shape, dtype=x.dtype, device=x.device) + random_tensor.floor_() # binarize + output = x.div(keep_prob) * random_tensor + return output + + +class DropPath(nn.Module): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + """ + def __init__(self, drop_prob=None): + super(DropPath, self).__init__() + self.drop_prob = drop_prob + + def forward(self, x): + return drop_path(x, self.drop_prob, self.training) + + +class Mlp(nn.Module): + def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.act = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + + def forward(self, x): + x = self.fc1(x) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class Attention(nn.Module): + def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.): + super().__init__() + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim ** -0.5 + + self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + self.proj_drop = nn.Dropout(proj_drop) + + def forward(self, x): + B, N, C = x.shape + qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4) + q, k, v = qkv[0], qkv[1], qkv[2] + + attn = (q @ k.transpose(-2, -1)) * self.scale + attn = attn.softmax(dim=-1) + attn = self.attn_drop(attn) + + x = (attn @ v).transpose(1, 2).reshape(B, N, C) + x = self.proj(x) + x = self.proj_drop(x) + return x, attn + + +class Block(nn.Module): + def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0., + drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm): + super().__init__() + self.norm1 = norm_layer(dim) + self.attn = Attention( + dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop) + self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() + self.norm2 = norm_layer(dim) + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop) + + def forward(self, x, return_attention=False): + y, attn = self.attn(self.norm1(x)) + if return_attention: + return attn + x = x + self.drop_path(y) + x = x + self.drop_path(self.mlp(self.norm2(x))) + return x + + +class PatchEmbed(nn.Module): + """ Image to Patch Embedding + """ + def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768): + super().__init__() + num_patches = (img_size // patch_size) * (img_size // patch_size) + self.img_size = img_size + self.patch_size = patch_size + self.num_patches = num_patches + + self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size) + + def forward(self, x): + B, C, H, W = x.shape + x = self.proj(x).flatten(2).transpose(1, 2) + return x + + +class VisionTransformer(nn.Module): + """ Vision Transformer """ + def __init__(self, img_size=[224], patch_size=16, in_chans=3, num_classes=0, embed_dim=768, depth=12, + num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop_rate=0., attn_drop_rate=0., + drop_path_rate=0., norm_layer=nn.LayerNorm, **kwargs): + super().__init__() + self.num_features = self.embed_dim = embed_dim + + self.patch_embed = PatchEmbed( + img_size=img_size[0], patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim) + num_patches = self.patch_embed.num_patches + + self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim)) + self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim)) + self.pos_drop = nn.Dropout(p=drop_rate) + + dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)] # stochastic depth decay rule + self.blocks = nn.ModuleList([ + Block( + dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, + drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer) + for i in range(depth)]) + self.norm = norm_layer(embed_dim) + + # Classifier head + self.head = nn.Linear(embed_dim, num_classes) if num_classes > 0 else nn.Identity() + + trunc_normal_(self.pos_embed, std=.02) + trunc_normal_(self.cls_token, std=.02) + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight, std=.02) + if isinstance(m, nn.Linear) and m.bias is not None: + nn.init.constant_(m.bias, 0) + elif isinstance(m, nn.LayerNorm): + nn.init.constant_(m.bias, 0) + nn.init.constant_(m.weight, 1.0) + + def interpolate_pos_encoding(self, x, w, h): + npatch = x.shape[1] - 1 + N = self.pos_embed.shape[1] - 1 + if npatch == N and w == h: + return self.pos_embed + class_pos_embed = self.pos_embed[:, 0] + patch_pos_embed = self.pos_embed[:, 1:] + dim = x.shape[-1] + w0 = w // self.patch_embed.patch_size + h0 = h // self.patch_embed.patch_size + # we add a small number to avoid floating point error in the interpolation + # see discussion at https://github.com/facebookresearch/dino/issues/8 + w0, h0 = w0 + 0.1, h0 + 0.1 + patch_pos_embed = nn.functional.interpolate( + patch_pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(0, 3, 1, 2), + scale_factor=(w0 / math.sqrt(N), h0 / math.sqrt(N)), + mode='bicubic', + align_corners=False, + recompute_scale_factor=False + ) + assert int(w0) == patch_pos_embed.shape[-2] and int(h0) == patch_pos_embed.shape[-1] + patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim) + return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1) + + def prepare_tokens(self, x, ada_token=None): + B, nc, w, h = x.shape + x = self.patch_embed(x) # patch linear embedding + + # add the [CLS] token to the embed patch tokens + cls_tokens = self.cls_token.expand(B, -1, -1) + x = torch.cat((cls_tokens, x), dim=1) + + # add positional encoding to each token + x = x + self.interpolate_pos_encoding(x, w, h) + + if ada_token is not None: + ada_tokens = ada_token.expand(B, -1, -1) # B, p, d + x = torch.cat((x, ada_tokens), dim=1) + + return self.pos_drop(x) + + def forward(self, x, ada_token=None, use_patches=False): + #print('1:', x.size()) + x = self.prepare_tokens(x, ada_token) + #print('2:', x.size()) + for blk in self.blocks: + x = blk(x) + #print('blk:', x.size()) + #print('3:', x.size()) + x = self.norm(x) + #print('x:', x.size()) + + if use_patches: + return x[:, 1:] + else: + return x[:, 0] + + def forward_block1(self, x, ada_token=None, use_patches=False): + x = self.prepare_tokens(x, ada_token) + num_units = len(self.blocks)//4 + for blk in self.blocks[:num_units]: + x = blk(x) + return x + + def forward_block2(self, x, ada_token=None, use_patches=False): + num_units = len(self.blocks)//4 + for blk in self.blocks[num_units:2*num_units]: + x = blk(x) + return x + + def forward_block3(self, x, ada_token=None, use_patches=False): + num_units = len(self.blocks)//4 + for blk in self.blocks[2*num_units:3*num_units]: + x = blk(x) + return x + + def forward_block4(self, x, ada_token=None, use_patches=False): + num_units = len(self.blocks)//4 + for blk in self.blocks[3*num_units:]: + x = blk(x) + return x + + def forward_rest(self, x, ada_token=None, use_patches=False): + x = self.norm(x) + if use_patches: + return x[:, 1:] + else: + return x[:, 0] + + + def get_last_selfattention(self, x): + x = self.prepare_tokens(x) + for i, blk in enumerate(self.blocks): + if i < len(self.blocks) - 1: + x = blk(x) + else: + # return attention of the last block + return blk(x, return_attention=True) + + def get_intermediate_layers(self, x, n=1): + x = self.prepare_tokens(x) + # we return the output tokens from the `n` last blocks + output = [] + for i, blk in enumerate(self.blocks): + x = blk(x) + if len(self.blocks) - i <= n: + output.append(self.norm(x)) + return output + + +def vit_tiny(patch_size=16, **kwargs): + model = VisionTransformer( + patch_size=patch_size, embed_dim=192, depth=12, num_heads=3, mlp_ratio=4, + qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs) + return model + + +def vit_small(patch_size=16, **kwargs): + model = VisionTransformer( + patch_size=patch_size, embed_dim=384, depth=12, num_heads=6, mlp_ratio=4, + qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs) + return model + + +def vit_base(patch_size=16, **kwargs): + model = VisionTransformer( + patch_size=patch_size, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, + qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs) + return model diff --git a/methods/backbone.py b/methods/backbone.py new file mode 100644 index 0000000000000000000000000000000000000000..c82eb6508cf25032dcaa3942ce847b82097b40af --- /dev/null +++ b/methods/backbone.py @@ -0,0 +1,465 @@ +# This code is modified from https://github.com/facebookresearch/low-shot-shrink-hallucinate + +import torch +import torch.nn as nn +import math +import torch.nn.functional as F +from torch.nn.utils import weight_norm + +# --- gaussian initialize --- +def init_layer(L): + # Initialization using fan-in + if isinstance(L, nn.Conv2d): + n = L.kernel_size[0]*L.kernel_size[1]*L.out_channels + L.weight.data.normal_(0,math.sqrt(2.0/float(n))) + elif isinstance(L, nn.BatchNorm2d): + L.weight.data.fill_(1) + L.bias.data.fill_(0) + +class distLinear(nn.Module): + def __init__(self, indim, outdim): + super(distLinear, self).__init__() + self.L = weight_norm(nn.Linear(indim, outdim, bias=False), name='weight', dim=0) + self.relu = nn.ReLU() + + def forward(self, x): + x_norm = torch.norm(x, p=2, dim =1).unsqueeze(1).expand_as(x) + x_normalized = x.div(x_norm + 0.00001) + L_norm = torch.norm(self.L.weight.data, p=2, dim =1).unsqueeze(1).expand_as(self.L.weight.data) + self.L.weight.data = self.L.weight.data.div(L_norm + 0.00001) + cos_dist = self.L(x_normalized) + scores = 10 * cos_dist + return scores + +# --- flatten tensor --- +class Flatten(nn.Module): + def __init__(self): + super(Flatten, self).__init__() + + def forward(self, x): + return x.view(x.size(0), -1) + +# --- LSTMCell module for matchingnet --- +class LSTMCell(nn.Module): + maml = False + def __init__(self, input_size, hidden_size, bias=True): + super(LSTMCell, self).__init__() + self.input_size = input_size + self.hidden_size = hidden_size + self.bias = bias + if self.maml: + self.x2h = Linear_fw(input_size, 4 * hidden_size, bias=bias) + self.h2h = Linear_fw(hidden_size, 4 * hidden_size, bias=bias) + else: + self.x2h = nn.Linear(input_size, 4 * hidden_size, bias=bias) + self.h2h = nn.Linear(hidden_size, 4 * hidden_size, bias=bias) + self.reset_parameters() + + def reset_parameters(self): + std = 1.0 / math.sqrt(self.hidden_size) + for w in self.parameters(): + w.data.uniform_(-std, std) + + def forward(self, x, hidden=None): + if hidden is None: + hx = torch.zeors_like(x) + cx = torch.zeros_like(x) + else: + hx, cx = hidden + + gates = self.x2h(x) + self.h2h(hx) + ingate, forgetgate, cellgate, outgate = torch.split(gates, self.hidden_size, dim=1) + + ingate = torch.sigmoid(ingate) + forgetgate = torch.sigmoid(forgetgate) + cellgate = torch.tanh(cellgate) + outgate = torch.sigmoid(outgate) + + cy = torch.mul(cx, forgetgate) + torch.mul(ingate, cellgate) + hy = torch.mul(outgate, torch.tanh(cy)) + return (hy, cy) + +# --- LSTM module for matchingnet --- +class LSTM(nn.Module): + def __init__(self, input_size, hidden_size, num_layers=1, bias=True, batch_first=False, bidirectional=False): + super(LSTM, self).__init__() + + self.input_size = input_size + self.hidden_size = hidden_size + self.num_layers = num_layers + self.bias = bias + self.batch_first = batch_first + self.num_directions = 2 if bidirectional else 1 + assert(self.num_layers == 1) + + self.lstm = LSTMCell(input_size, hidden_size, self.bias) + + def forward(self, x, hidden=None): + # swap axis if batch first + if self.batch_first: + x = x.permute(1, 0 ,2) + + # hidden state + if hidden is None: + h0 = torch.zeros(self.num_directions, x.size(1), self.hidden_size, dtype=x.dtype, device=x.device) + c0 = torch.zeros(self.num_directions, x.size(1), self.hidden_size, dtype=x.dtype, device=x.device) + else: + h0, c0 = hidden + + # forward + outs = [] + hn = h0[0] + cn = c0[0] + for seq in range(x.size(0)): + hn, cn = self.lstm(x[seq], (hn, cn)) + outs.append(hn.unsqueeze(0)) + outs = torch.cat(outs, dim=0) + + # reverse foward + if self.num_directions == 2: + outs_reverse = [] + hn = h0[1] + cn = c0[1] + for seq in range(x.size(0)): + seq = x.size(1) - 1 - seq + hn, cn = self.lstm(x[seq], (hn, cn)) + outs_reverse.append(hn.unsqueeze(0)) + outs_reverse = torch.cat(outs_reverse, dim=0) + outs = torch.cat([outs, outs_reverse], dim=2) + + # swap axis if batch first + if self.batch_first: + outs = outs.permute(1, 0, 2) + return outs + +# --- Linear module --- +class Linear_fw(nn.Linear): #used in MAML to forward input with fast weight + def __init__(self, in_features, out_features, bias=True): + super(Linear_fw, self).__init__(in_features, out_features, bias=bias) + self.weight.fast = None #Lazy hack to add fast weight link + self.bias.fast = None + + def forward(self, x): + if self.weight.fast is not None and self.bias.fast is not None: + out = F.linear(x, self.weight.fast, self.bias.fast) + else: + out = super(Linear_fw, self).forward(x) + return out + +# --- Conv2d module --- +class Conv2d_fw(nn.Conv2d): #used in MAML to forward input with fast weight + def __init__(self, in_channels, out_channels, kernel_size, stride=1,padding=0, bias = True): + super(Conv2d_fw, self).__init__(in_channels, out_channels, kernel_size, stride=stride, padding=padding, bias=bias) + self.weight.fast = None + if not self.bias is None: + self.bias.fast = None + + def forward(self, x): + if self.bias is None: + if self.weight.fast is not None: + out = F.conv2d(x, self.weight.fast, None, stride= self.stride, padding=self.padding) + else: + out = super(Conv2d_fw, self).forward(x) + else: + if self.weight.fast is not None and self.bias.fast is not None: + out = F.conv2d(x, self.weight.fast, self.bias.fast, stride= self.stride, padding=self.padding) + else: + out = super(Conv2d_fw, self).forward(x) + return out + +# --- softplus module --- +def softplus(x): + return torch.nn.functional.softplus(x, beta=100) + +# --- feature-wise transformation layer --- +class FeatureWiseTransformation2d_fw(nn.BatchNorm2d): + feature_augment = False + def __init__(self, num_features, momentum=0.1, track_running_stats=True): + super(FeatureWiseTransformation2d_fw, self).__init__(num_features, momentum=momentum, track_running_stats=track_running_stats) + self.weight.fast = None + self.bias.fast = None + if self.track_running_stats: + self.register_buffer('running_mean', torch.zeros(num_features)) + self.register_buffer('running_var', torch.zeros(num_features)) + if self.feature_augment: # initialize {gamma, beta} with {0.3, 0.5} + self.gamma = torch.nn.Parameter(torch.ones(1, num_features, 1, 1)*0.3) + self.beta = torch.nn.Parameter(torch.ones(1, num_features, 1, 1)*0.5) + self.reset_parameters() + + def reset_running_stats(self): + if self.track_running_stats: + self.running_mean.zero_() + self.running_var.fill_(1) + + def forward(self, x, step=0): + if self.weight.fast is not None and self.bias.fast is not None: + weight = self.weight.fast + bias = self.bias.fast + else: + weight = self.weight + bias = self.bias + if self.track_running_stats: + out = F.batch_norm(x, self.running_mean, self.running_var, weight, bias, training=self.training, momentum=self.momentum) + else: + out = F.batch_norm(x, torch.zeros_like(x), torch.ones_like(x), weight, bias, training=True, momentum=1) + + # apply feature-wise transformation + if self.feature_augment and self.training: + gamma = (1 + torch.randn(1, self.num_features, 1, 1, dtype=self.gamma.dtype, device=self.gamma.device)*softplus(self.gamma)).expand_as(out) + beta = (torch.randn(1, self.num_features, 1, 1, dtype=self.beta.dtype, device=self.beta.device)*softplus(self.beta)).expand_as(out) + out = gamma*out + beta + return out + +# --- BatchNorm2d --- +class BatchNorm2d_fw(nn.BatchNorm2d): + def __init__(self, num_features, momentum=0.1, track_running_stats=True): + super(BatchNorm2d_fw, self).__init__(num_features, momentum=momentum, track_running_stats=track_running_stats) + self.weight.fast = None + self.bias.fast = None + if self.track_running_stats: + self.register_buffer('running_mean', torch.zeros(num_features)) + self.register_buffer('running_var', torch.zeros(num_features)) + self.reset_parameters() + + def reset_running_stats(self): + if self.track_running_stats: + self.running_mean.zero_() + self.running_var.fill_(1) + + def forward(self, x, step=0): + if self.weight.fast is not None and self.bias.fast is not None: + weight = self.weight.fast + bias = self.bias.fast + else: + weight = self.weight + bias = self.bias + if self.track_running_stats: + out = F.batch_norm(x, self.running_mean, self.running_var, weight, bias, training=self.training, momentum=self.momentum) + else: + out = F.batch_norm(x, torch.zeros(x.size(1), dtype=x.dtype, device=x.device), torch.ones(x.size(1), dtype=x.dtype, device=x.device), weight, bias, training=True, momentum=1) + return out + +# --- BatchNorm1d --- +class BatchNorm1d_fw(nn.BatchNorm1d): + def __init__(self, num_features, momentum=0.1, track_running_stats=True): + super(BatchNorm1d_fw, self).__init__(num_features, momentum=momentum, track_running_stats=track_running_stats) + self.weight.fast = None + self.bias.fast = None + if self.track_running_stats: + self.register_buffer('running_mean', torch.zeros(num_features)) + self.register_buffer('running_var', torch.zeros(num_features)) + self.reset_parameters() + + def reset_running_stats(self): + if self.track_running_stats: + self.running_mean.zero_() + self.running_var.fill_(1) + + def forward(self, x, step=0): + if self.weight.fast is not None and self.bias.fast is not None: + weight = self.weight.fast + bias = self.bias.fast + else: + weight = self.weight + bias = self.bias + if self.track_running_stats: + out = F.batch_norm(x, self.running_mean, self.running_var, weight, bias, training=self.training, momentum=self.momentum) + else: + out = F.batch_norm(x, torch.zeros(x.size(1), dtype=x.dtype, device=x.device), torch.ones(x.size(1), dtype=x.dtype, device=x.device), weight, bias, training=True, momentum=1) + return out + +# --- Simple Conv Block --- +class ConvBlock(nn.Module): + maml = False + def __init__(self, indim, outdim, pool = True, padding = 1): + super(ConvBlock, self).__init__() + self.indim = indim + self.outdim = outdim + if self.maml: + self.C = Conv2d_fw(indim, outdim, 3, padding = padding) + self.BN = FeatureWiseTransformation2d_fw(outdim) + else: + self.C = nn.Conv2d(indim, outdim, 3, padding= padding) + self.BN = nn.BatchNorm2d(outdim) + self.relu = nn.ReLU(inplace=True) + + self.parametrized_layers = [self.C, self.BN, self.relu] + if pool: + self.pool = nn.MaxPool2d(2) + self.parametrized_layers.append(self.pool) + + for layer in self.parametrized_layers: + init_layer(layer) + self.trunk = nn.Sequential(*self.parametrized_layers) + + def forward(self,x): + out = self.trunk(x) + return out + +# --- Simple ResNet Block --- +class SimpleBlock(nn.Module): + maml = False + def __init__(self, indim, outdim, half_res, leaky=False): + super(SimpleBlock, self).__init__() + self.indim = indim + self.outdim = outdim + if self.maml: + self.C1 = Conv2d_fw(indim, outdim, kernel_size=3, stride=2 if half_res else 1, padding=1, bias=False) + self.BN1 = BatchNorm2d_fw(outdim) + self.C2 = Conv2d_fw(outdim, outdim,kernel_size=3, padding=1,bias=False) + self.BN2 = FeatureWiseTransformation2d_fw(outdim) # feature-wise transformation at the end of each residual block + else: + self.C1 = nn.Conv2d(indim, outdim, kernel_size=3, stride=2 if half_res else 1, padding=1, bias=False) + self.BN1 = nn.BatchNorm2d(outdim) + self.C2 = nn.Conv2d(outdim, outdim,kernel_size=3, padding=1,bias=False) + self.BN2 = nn.BatchNorm2d(outdim) + self.relu1 = nn.ReLU(inplace=True) if not leaky else nn.LeakyReLU(0.2, inplace=True) + self.relu2 = nn.ReLU(inplace=True) if not leaky else nn.LeakyReLU(0.2, inplace=True) + + self.parametrized_layers = [self.C1, self.C2, self.BN1, self.BN2] + + self.half_res = half_res + + # if the input number of channels is not equal to the output, then need a 1x1 convolution + if indim!=outdim: + if self.maml: + self.shortcut = Conv2d_fw(indim, outdim, 1, 2 if half_res else 1, bias=False) + self.BNshortcut = FeatureWiseTransformation2d_fw(outdim) + else: + self.shortcut = nn.Conv2d(indim, outdim, 1, 2 if half_res else 1, bias=False) + self.BNshortcut = nn.BatchNorm2d(outdim) + + self.parametrized_layers.append(self.shortcut) + self.parametrized_layers.append(self.BNshortcut) + self.shortcut_type = '1x1' + else: + self.shortcut_type = 'identity' + + for layer in self.parametrized_layers: + init_layer(layer) + + def forward(self, x): + out = self.C1(x) + out = self.BN1(out) + out = self.relu1(out) + out = self.C2(out) + out = self.BN2(out) + short_out = x if self.shortcut_type == 'identity' else self.BNshortcut(self.shortcut(x)) + out = out + short_out + out = self.relu2(out) + return out + +# --- ConvNet module --- +class ConvNet(nn.Module): + def __init__(self, depth, flatten = True): + super(ConvNet,self).__init__() + self.grads = [] + self.fmaps = [] + trunk = [] + for i in range(depth): + indim = 3 if i == 0 else 64 + outdim = 64 + B = ConvBlock(indim, outdim, pool = ( i <4 ) ) #only pooling for fist 4 layers + trunk.append(B) + + if flatten: + trunk.append(Flatten()) + + self.trunk = nn.Sequential(*trunk) + self.final_feat_dim = 1600 + + def forward(self,x): + out = self.trunk(x) + return out + +# --- ConvNetNopool module --- +class ConvNetNopool(nn.Module): #Relation net use a 4 layer conv with pooling in only first two layers, else no pooling + def __init__(self, depth): + super(ConvNetNopool,self).__init__() + self.grads = [] + self.fmaps = [] + trunk = [] + for i in range(depth): + indim = 3 if i == 0 else 64 + outdim = 64 + B = ConvBlock(indim, outdim, pool = ( i in [0,1] ), padding = 0 if i in[0,1] else 1 ) #only first two layer has pooling and no padding + trunk.append(B) + + self.trunk = nn.Sequential(*trunk) + self.final_feat_dim = [64,19,19] + + def forward(self,x): + out = self.trunk(x) + return out + +# --- ResNet module --- +class ResNet(nn.Module): + maml = False + print('backbone:', 'maml:', maml) + def __init__(self,block,list_of_num_layers, list_of_out_dims, flatten=True, leakyrelu=False): + # list_of_num_layers specifies number of layers in each stage + # list_of_out_dims specifies number of output channel for each stage + super(ResNet,self).__init__() + self.grads = [] + self.fmaps = [] + assert len(list_of_num_layers)==4, 'Can have only four stages' + if self.maml: + conv1 = Conv2d_fw(3, 64, kernel_size=7, stride=2, padding=3, bias=False) + bn1 = BatchNorm2d_fw(64) + else: + conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) + bn1 = nn.BatchNorm2d(64) + + relu = nn.ReLU(inplace=True) if not leakyrelu else nn.LeakyReLU(0.2, inplace=True) + pool1 = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + + init_layer(conv1) + init_layer(bn1) + + trunk = [conv1, bn1, relu, pool1] + + indim = 64 + for i in range(4): + for j in range(list_of_num_layers[i]): + half_res = (i>=1) and (j==0) + B = block(indim, list_of_out_dims[i], half_res, leaky=leakyrelu) + trunk.append(B) + indim = list_of_out_dims[i] + + if flatten: + avgpool = nn.AvgPool2d(7) + trunk.append(avgpool) + trunk.append(Flatten()) + self.final_feat_dim = indim + else: + self.final_feat_dim = [ indim, 7, 7] + + self.trunk = nn.Sequential(*trunk) + + def forward(self,x): + out = self.trunk(x) + return out + +# --- Conv networks --- +def Conv4(): + return ConvNet(4) +def Conv6(): + return ConvNet(6) +def Conv4NP(): + return ConvNetNopool(4) +def Conv6NP(): + return ConvNetNopool(6) + +# --- ResNet networks --- +def ResNet10(flatten=True, leakyrelu=False): + return ResNet(SimpleBlock, [1,1,1,1],[64,128,256,512], flatten, leakyrelu) +def ResNet18(flatten=True, leakyrelu=False): + return ResNet(SimpleBlock, [2,2,2,2],[64,128,256,512], flatten, leakyrelu) +def ResNet34(flatten=True, leakyrelu=False): + return ResNet(SimpleBlock, [3,4,6,3],[64,128,256,512], flatten, leakyrelu) + +model_dict = dict(Conv4 = Conv4, + Conv6 = Conv6, + ResNet10 = ResNet10, + ResNet18 = ResNet18, + ResNet34 = ResNet34) diff --git a/methods/backbone_multiblock.py b/methods/backbone_multiblock.py new file mode 100644 index 0000000000000000000000000000000000000000..607e2938767a913b24426014be2ea4dd8da2ef9d --- /dev/null +++ b/methods/backbone_multiblock.py @@ -0,0 +1,593 @@ +# This code is modified from https://github.com/facebookresearch/low-shot-shrink-hallucinate + +import torch +import torch.nn as nn +import math +import torch.nn.functional as F +from torch.nn.utils import weight_norm + +# --- gaussian initialize --- +def init_layer(L): + # Initialization using fan-in + if isinstance(L, nn.Conv2d): + n = L.kernel_size[0]*L.kernel_size[1]*L.out_channels + L.weight.data.normal_(0,math.sqrt(2.0/float(n))) + elif isinstance(L, nn.BatchNorm2d): + L.weight.data.fill_(1) + L.bias.data.fill_(0) + +class distLinear(nn.Module): + def __init__(self, indim, outdim): + super(distLinear, self).__init__() + self.L = weight_norm(nn.Linear(indim, outdim, bias=False), name='weight', dim=0) + self.relu = nn.ReLU() + + def forward(self, x): + x_norm = torch.norm(x, p=2, dim =1).unsqueeze(1).expand_as(x) + x_normalized = x.div(x_norm + 0.00001) + L_norm = torch.norm(self.L.weight.data, p=2, dim =1).unsqueeze(1).expand_as(self.L.weight.data) + self.L.weight.data = self.L.weight.data.div(L_norm + 0.00001) + cos_dist = self.L(x_normalized) + scores = 10 * cos_dist + return scores + +# --- flatten tensor --- +class Flatten(nn.Module): + def __init__(self): + super(Flatten, self).__init__() + + def forward(self, x): + return x.view(x.size(0), -1) + +# --- LSTMCell module for matchingnet --- +class LSTMCell(nn.Module): + maml = False + def __init__(self, input_size, hidden_size, bias=True): + super(LSTMCell, self).__init__() + self.input_size = input_size + self.hidden_size = hidden_size + self.bias = bias + if self.maml: + self.x2h = Linear_fw(input_size, 4 * hidden_size, bias=bias) + self.h2h = Linear_fw(hidden_size, 4 * hidden_size, bias=bias) + else: + self.x2h = nn.Linear(input_size, 4 * hidden_size, bias=bias) + self.h2h = nn.Linear(hidden_size, 4 * hidden_size, bias=bias) + self.reset_parameters() + + def reset_parameters(self): + std = 1.0 / math.sqrt(self.hidden_size) + for w in self.parameters(): + w.data.uniform_(-std, std) + + def forward(self, x, hidden=None): + if hidden is None: + hx = torch.zeors_like(x) + cx = torch.zeros_like(x) + else: + hx, cx = hidden + + gates = self.x2h(x) + self.h2h(hx) + ingate, forgetgate, cellgate, outgate = torch.split(gates, self.hidden_size, dim=1) + + ingate = torch.sigmoid(ingate) + forgetgate = torch.sigmoid(forgetgate) + cellgate = torch.tanh(cellgate) + outgate = torch.sigmoid(outgate) + + cy = torch.mul(cx, forgetgate) + torch.mul(ingate, cellgate) + hy = torch.mul(outgate, torch.tanh(cy)) + return (hy, cy) + +# --- LSTM module for matchingnet --- +class LSTM(nn.Module): + def __init__(self, input_size, hidden_size, num_layers=1, bias=True, batch_first=False, bidirectional=False): + super(LSTM, self).__init__() + + self.input_size = input_size + self.hidden_size = hidden_size + self.num_layers = num_layers + self.bias = bias + self.batch_first = batch_first + self.num_directions = 2 if bidirectional else 1 + assert(self.num_layers == 1) + + self.lstm = LSTMCell(input_size, hidden_size, self.bias) + + def forward(self, x, hidden=None): + # swap axis if batch first + if self.batch_first: + x = x.permute(1, 0 ,2) + + # hidden state + if hidden is None: + h0 = torch.zeros(self.num_directions, x.size(1), self.hidden_size, dtype=x.dtype, device=x.device) + c0 = torch.zeros(self.num_directions, x.size(1), self.hidden_size, dtype=x.dtype, device=x.device) + else: + h0, c0 = hidden + + # forward + outs = [] + hn = h0[0] + cn = c0[0] + for seq in range(x.size(0)): + hn, cn = self.lstm(x[seq], (hn, cn)) + outs.append(hn.unsqueeze(0)) + outs = torch.cat(outs, dim=0) + + # reverse foward + if self.num_directions == 2: + outs_reverse = [] + hn = h0[1] + cn = c0[1] + for seq in range(x.size(0)): + seq = x.size(1) - 1 - seq + hn, cn = self.lstm(x[seq], (hn, cn)) + outs_reverse.append(hn.unsqueeze(0)) + outs_reverse = torch.cat(outs_reverse, dim=0) + outs = torch.cat([outs, outs_reverse], dim=2) + + # swap axis if batch first + if self.batch_first: + outs = outs.permute(1, 0, 2) + return outs + +# --- Linear module --- +class Linear_fw(nn.Linear): #used in MAML to forward input with fast weight + def __init__(self, in_features, out_features, bias=True): + super(Linear_fw, self).__init__(in_features, out_features, bias=bias) + self.weight.fast = None #Lazy hack to add fast weight link + self.bias.fast = None + + def forward(self, x): + if self.weight.fast is not None and self.bias.fast is not None: + out = F.linear(x, self.weight.fast, self.bias.fast) + else: + out = super(Linear_fw, self).forward(x) + return out + +# --- Conv2d module --- +class Conv2d_fw(nn.Conv2d): #used in MAML to forward input with fast weight + def __init__(self, in_channels, out_channels, kernel_size, stride=1,padding=0, bias = True): + super(Conv2d_fw, self).__init__(in_channels, out_channels, kernel_size, stride=stride, padding=padding, bias=bias) + self.weight.fast = None + if not self.bias is None: + self.bias.fast = None + + def forward(self, x): + if self.bias is None: + if self.weight.fast is not None: + out = F.conv2d(x, self.weight.fast, None, stride= self.stride, padding=self.padding) + else: + out = super(Conv2d_fw, self).forward(x) + else: + if self.weight.fast is not None and self.bias.fast is not None: + out = F.conv2d(x, self.weight.fast, self.bias.fast, stride= self.stride, padding=self.padding) + else: + out = super(Conv2d_fw, self).forward(x) + return out + +# --- softplus module --- +def softplus(x): + return torch.nn.functional.softplus(x, beta=100) + +# --- feature-wise transformation layer --- +class FeatureWiseTransformation2d_fw(nn.BatchNorm2d): + feature_augment = False + def __init__(self, num_features, momentum=0.1, track_running_stats=True): + super(FeatureWiseTransformation2d_fw, self).__init__(num_features, momentum=momentum, track_running_stats=track_running_stats) + self.weight.fast = None + self.bias.fast = None + if self.track_running_stats: + self.register_buffer('running_mean', torch.zeros(num_features)) + self.register_buffer('running_var', torch.zeros(num_features)) + if self.feature_augment: # initialize {gamma, beta} with {0.3, 0.5} + self.gamma = torch.nn.Parameter(torch.ones(1, num_features, 1, 1)*0.3) + self.beta = torch.nn.Parameter(torch.ones(1, num_features, 1, 1)*0.5) + self.reset_parameters() + + def reset_running_stats(self): + if self.track_running_stats: + self.running_mean.zero_() + self.running_var.fill_(1) + + def forward(self, x, step=0): + if self.weight.fast is not None and self.bias.fast is not None: + weight = self.weight.fast + bias = self.bias.fast + else: + weight = self.weight + bias = self.bias + if self.track_running_stats: + out = F.batch_norm(x, self.running_mean, self.running_var, weight, bias, training=self.training, momentum=self.momentum) + else: + out = F.batch_norm(x, torch.zeros_like(x), torch.ones_like(x), weight, bias, training=True, momentum=1) + + # apply feature-wise transformation + if self.feature_augment and self.training: + gamma = (1 + torch.randn(1, self.num_features, 1, 1, dtype=self.gamma.dtype, device=self.gamma.device)*softplus(self.gamma)).expand_as(out) + beta = (torch.randn(1, self.num_features, 1, 1, dtype=self.beta.dtype, device=self.beta.device)*softplus(self.beta)).expand_as(out) + out = gamma*out + beta + return out + +# --- BatchNorm2d --- +class BatchNorm2d_fw(nn.BatchNorm2d): + def __init__(self, num_features, momentum=0.1, track_running_stats=True): + super(BatchNorm2d_fw, self).__init__(num_features, momentum=momentum, track_running_stats=track_running_stats) + self.weight.fast = None + self.bias.fast = None + if self.track_running_stats: + self.register_buffer('running_mean', torch.zeros(num_features)) + self.register_buffer('running_var', torch.zeros(num_features)) + self.reset_parameters() + + def reset_running_stats(self): + if self.track_running_stats: + self.running_mean.zero_() + self.running_var.fill_(1) + + def forward(self, x, step=0): + if self.weight.fast is not None and self.bias.fast is not None: + weight = self.weight.fast + bias = self.bias.fast + else: + weight = self.weight + bias = self.bias + if self.track_running_stats: + out = F.batch_norm(x, self.running_mean, self.running_var, weight, bias, training=self.training, momentum=self.momentum) + else: + out = F.batch_norm(x, torch.zeros(x.size(1), dtype=x.dtype, device=x.device), torch.ones(x.size(1), dtype=x.dtype, device=x.device), weight, bias, training=True, momentum=1) + return out + +# --- BatchNorm1d --- +class BatchNorm1d_fw(nn.BatchNorm1d): + def __init__(self, num_features, momentum=0.1, track_running_stats=True): + super(BatchNorm1d_fw, self).__init__(num_features, momentum=momentum, track_running_stats=track_running_stats) + self.weight.fast = None + self.bias.fast = None + if self.track_running_stats: + self.register_buffer('running_mean', torch.zeros(num_features)) + self.register_buffer('running_var', torch.zeros(num_features)) + self.reset_parameters() + + def reset_running_stats(self): + if self.track_running_stats: + self.running_mean.zero_() + self.running_var.fill_(1) + + def forward(self, x, step=0): + if self.weight.fast is not None and self.bias.fast is not None: + weight = self.weight.fast + bias = self.bias.fast + else: + weight = self.weight + bias = self.bias + if self.track_running_stats: + out = F.batch_norm(x, self.running_mean, self.running_var, weight, bias, training=self.training, momentum=self.momentum) + else: + out = F.batch_norm(x, torch.zeros(x.size(1), dtype=x.dtype, device=x.device), torch.ones(x.size(1), dtype=x.dtype, device=x.device), weight, bias, training=True, momentum=1) + return out + +# --- Simple Conv Block --- +class ConvBlock(nn.Module): + maml = False + def __init__(self, indim, outdim, pool = True, padding = 1): + super(ConvBlock, self).__init__() + self.indim = indim + self.outdim = outdim + if self.maml: + self.C = Conv2d_fw(indim, outdim, 3, padding = padding) + self.BN = FeatureWiseTransformation2d_fw(outdim) + else: + self.C = nn.Conv2d(indim, outdim, 3, padding= padding) + self.BN = nn.BatchNorm2d(outdim) + self.relu = nn.ReLU(inplace=True) + + self.parametrized_layers = [self.C, self.BN, self.relu] + if pool: + self.pool = nn.MaxPool2d(2) + self.parametrized_layers.append(self.pool) + + for layer in self.parametrized_layers: + init_layer(layer) + self.trunk = nn.Sequential(*self.parametrized_layers) + + def forward(self,x): + out = self.trunk(x) + return out + +# --- Simple ResNet Block --- +class SimpleBlock(nn.Module): + maml = False + def __init__(self, indim, outdim, half_res, leaky=False): + super(SimpleBlock, self).__init__() + self.indim = indim + self.outdim = outdim + if self.maml: + self.C1 = Conv2d_fw(indim, outdim, kernel_size=3, stride=2 if half_res else 1, padding=1, bias=False) + self.BN1 = BatchNorm2d_fw(outdim) + self.C2 = Conv2d_fw(outdim, outdim,kernel_size=3, padding=1,bias=False) + self.BN2 = FeatureWiseTransformation2d_fw(outdim) # feature-wise transformation at the end of each residual block + else: + self.C1 = nn.Conv2d(indim, outdim, kernel_size=3, stride=2 if half_res else 1, padding=1, bias=False) + self.BN1 = nn.BatchNorm2d(outdim) + self.C2 = nn.Conv2d(outdim, outdim,kernel_size=3, padding=1,bias=False) + self.BN2 = nn.BatchNorm2d(outdim) + self.relu1 = nn.ReLU(inplace=True) if not leaky else nn.LeakyReLU(0.2, inplace=True) + self.relu2 = nn.ReLU(inplace=True) if not leaky else nn.LeakyReLU(0.2, inplace=True) + + self.parametrized_layers = [self.C1, self.C2, self.BN1, self.BN2] + + self.half_res = half_res + + # if the input number of channels is not equal to the output, then need a 1x1 convolution + if indim!=outdim: + if self.maml: + self.shortcut = Conv2d_fw(indim, outdim, 1, 2 if half_res else 1, bias=False) + self.BNshortcut = FeatureWiseTransformation2d_fw(outdim) + else: + self.shortcut = nn.Conv2d(indim, outdim, 1, 2 if half_res else 1, bias=False) + self.BNshortcut = nn.BatchNorm2d(outdim) + + self.parametrized_layers.append(self.shortcut) + self.parametrized_layers.append(self.BNshortcut) + self.shortcut_type = '1x1' + else: + self.shortcut_type = 'identity' + + for layer in self.parametrized_layers: + init_layer(layer) + + def forward(self, x): + out = self.C1(x) + out = self.BN1(out) + out = self.relu1(out) + out = self.C2(out) + out = self.BN2(out) + short_out = x if self.shortcut_type == 'identity' else self.BNshortcut(self.shortcut(x)) + out = out + short_out + out = self.relu2(out) + return out + +# --- ConvNet module --- +class ConvNet(nn.Module): + def __init__(self, depth, flatten = True): + super(ConvNet,self).__init__() + self.grads = [] + self.fmaps = [] + trunk = [] + for i in range(depth): + indim = 3 if i == 0 else 64 + outdim = 64 + B = ConvBlock(indim, outdim, pool = ( i <4 ) ) #only pooling for fist 4 layers + trunk.append(B) + + if flatten: + trunk.append(Flatten()) + + self.trunk = nn.Sequential(*trunk) + self.final_feat_dim = 1600 + + def forward(self,x): + out = self.trunk(x) + return out + +# --- ConvNetNopool module --- +class ConvNetNopool(nn.Module): #Relation net use a 4 layer conv with pooling in only first two layers, else no pooling + def __init__(self, depth): + super(ConvNetNopool,self).__init__() + self.grads = [] + self.fmaps = [] + trunk = [] + for i in range(depth): + indim = 3 if i == 0 else 64 + outdim = 64 + B = ConvBlock(indim, outdim, pool = ( i in [0,1] ), padding = 0 if i in[0,1] else 1 ) #only first two layer has pooling and no padding + trunk.append(B) + + self.trunk = nn.Sequential(*trunk) + self.final_feat_dim = [64,19,19] + + def forward(self,x): + out = self.trunk(x) + return out + +# --- ResNet module --- +class ResNet(nn.Module): + maml = False + def __init__(self,block,list_of_num_layers, list_of_out_dims, flatten=True, leakyrelu=False): + # list_of_num_layers specifies number of layers in each stage + # list_of_out_dims specifies number of output channel for each stage + super(ResNet,self).__init__() + self.grads = [] + self.fmaps = [] + assert len(list_of_num_layers)==4, 'Can have only four stages' + if self.maml: + conv1 = Conv2d_fw(3, 64, kernel_size=7, stride=2, padding=3, bias=False) + bn1 = BatchNorm2d_fw(64) + else: + conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) + bn1 = nn.BatchNorm2d(64) + + relu = nn.ReLU(inplace=True) if not leakyrelu else nn.LeakyReLU(0.2, inplace=True) + pool1 = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + + init_layer(conv1) + init_layer(bn1) + + trunk = [conv1, bn1, relu, pool1] + + indim = 64 + for i in range(4): + for j in range(list_of_num_layers[i]): + half_res = (i>=1) and (j==0) + B = block(indim, list_of_out_dims[i], half_res, leaky=leakyrelu) + trunk.append(B) + indim = list_of_out_dims[i] + + if flatten: + avgpool = nn.AvgPool2d(7) + trunk.append(avgpool) + trunk.append(Flatten()) + self.final_feat_dim = indim + else: + self.final_feat_dim = [ indim, 7, 7] + + self.trunk = nn.Sequential(*trunk) + + def forward(self,x): + out = self.trunk(x) + return out + + + def forward_block1(self, x): + out = self.trunk[:5](x) + return out + + def forward_block2(self, x): + out = self.trunk[5:6](x) + return out + + def forward_block3(self, x): + out = self.trunk[6:7](x) + return out + + def forward_block4(self, x): + out = self.trunk[7:8](x) + return out + ''' + def forward_block5(self, x): + out = self.trunk[8:](x) + return out + ''' + def forward_rest(self,x): + out = self.trunk[8:](x) + return out + +# ----ResNet-multi module --------------- +class ResNet_Multi(nn.Module): + maml = False + def __init__(self,block,list_of_num_layers, list_of_out_dims, flatten=True, leakyrelu=False): + # list_of_num_layers specifies number of layers in each stage + # list_of_out_dims specifies number of output channel for each stage + super(ResNet_Multi,self).__init__() + self.grads = [] + self.fmaps = [] + assert len(list_of_num_layers)==4, 'Can have only four stages' + if self.maml: + conv1 = Conv2d_fw(3, 64, kernel_size=7, stride=2, padding=3, bias=False) + bn1 = BatchNorm2d_fw(64) + else: + conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) + bn1 = nn.BatchNorm2d(64) + + relu = nn.ReLU(inplace=True) if not leakyrelu else nn.LeakyReLU(0.2, inplace=True) + pool1 = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + + init_layer(conv1) + init_layer(bn1) + + trunk = [conv1, bn1, relu, pool1] + + indim = 64 + for i in range(4): + for j in range(list_of_num_layers[i]): + half_res = (i>=1) and (j==0) + B = block(indim, list_of_out_dims[i], half_res, leaky=leakyrelu) + trunk.append(B) + indim = list_of_out_dims[i] + + if flatten: + avgpool = nn.AvgPool2d(7) + trunk.append(avgpool) + trunk.append(Flatten()) + self.final_feat_dim = indim + else: + self.final_feat_dim = [ indim, 7, 7] + + self.trunk = nn.Sequential(*trunk) + + def forward(self,x): + #out = self.trunk(x) + layer1 = self.trunk[:5](x) + #print('layer1:', layer1.size()) + layer2 = self.trunk[5:6](layer1) + #print('layer2:', layer2.size()) + layer3 = self.trunk[6:7](layer2) + #print('layer3:', layer3.size()) + layer4 = self.trunk[7:8](layer3) + #print('layer4:', layer4.size()) + out = self.trunk[8:](layer4) + #print('out:', out.size()) + return layer1, layer2, layer3, layer4, out + + +# --- Conv networks --- +def Conv4(): + return ConvNet(4) +def Conv6(): + return ConvNet(6) +def Conv4NP(): + return ConvNetNopool(4) +def Conv6NP(): + return ConvNetNopool(6) + +# --- ResNet networks --- +def ResNet10(flatten=True, leakyrelu=False): + print('backbone:', 'return resnet10') + return ResNet(SimpleBlock, [1,1,1,1],[64,128,256,512], flatten, leakyrelu) +def ResNet10_Multi(flatten=True, leakyrelu=False): + print('this is resnet10-multi') + return ResNet_Multi(SimpleBlock, [1,1,1,1],[64,128,256,512], flatten, leakyrelu) +def ResNet18(flatten=True, leakyrelu=False): + return ResNet(SimpleBlock, [2,2,2,2],[64,128,256,512], flatten, leakyrelu) +def ResNet34(flatten=True, leakyrelu=False): + return ResNet(SimpleBlock, [3,4,6,3],[64,128,256,512], flatten, leakyrelu) + +model_dict = dict(Conv4 = Conv4, + Conv6 = Conv6, + ResNet10 = ResNet10, + ResNet10_Multi = ResNet10_Multi, + ResNet18 = ResNet18, + ResNet34 = ResNet34) + + +if __name__ == '__main__': + model_func = model_dict['ResNet10'] + net = model_func(flatten = True, leakyrelu= False) + from torch.autograd import Variable + x = Variable(torch.randn([16,3,224,224])) + out = net(x) + print(out.size()) + + # ---------------- multi ResNet + model_func_2 = model_dict['ResNet10_Multi'] + net2 = model_func_2(flatten = True, leakyrelu = False) + layer1, layer2, layer3, layer4, out2 = net2(x) + print('net-multi:', layer1.size(), layer2.size(), layer3.size(), layer4.size(), out2.size()) + + + + print('------------------') + model_func = model_dict['ResNet10'] + net = model_func(flatten = True, leakyrelu= False) + from torch.autograd import Variable + x = Variable(torch.randn([16,3,224,224])) + out = net(x) + print(out.size()) + + print(net) + block1 = net.forward_block1(x) + print('block1:', block1.size()) + + block2 = net.forward_block2(block1) + print('block2:', block2.size()) + + block3 = net.forward_block3(block2) + print('block3:', block3.size()) + + block4 = net.forward_block4(block3) + print('block4:', block4.size()) + + block5 = net.forward_block5(block4) + print('block5:', block5.size()) diff --git a/methods/engine_StyleAdv_ViT.py b/methods/engine_StyleAdv_ViT.py new file mode 100644 index 0000000000000000000000000000000000000000..17a0f590aba2f3901bc618c2299bf10a6b929bee --- /dev/null +++ b/methods/engine_StyleAdv_ViT.py @@ -0,0 +1,196 @@ +import math +import sys +import warnings +from typing import Iterable, Optional + +import torch +from torch.utils.tensorboard import SummaryWriter + +from timm.data import Mixup +from timm.utils import accuracy, ModelEma + +#import pmf_utils.deit_util as utils +#from pmf_utils import AverageMeter, to_device +from utils import AverageMeter, to_device +import utils.deit_util as utils + +import numpy as np + +#from methods.meta_template_StyleAdvIncrem_v10_epsilonFromList_RandomStartFGSM_20220501 import consistency_loss +#from methods.meta_template_StyleAdv_RN_GNN import consistency_loss +from methods.tool_func import consistency_loss + +def train_one_epoch_styleAdv(data_loader: Iterable, + model: torch.nn.Module, + criterion: torch.nn.Module, + optimizer: torch.optim.Optimizer, + epoch: int, + device: torch.device, + loss_scaler = None, + fp16: bool = False, + max_norm: float = 0, # clip_grad + model_ema: Optional[ModelEma] = None, + mixup_fn: Optional[Mixup] = None, + writer: Optional[SummaryWriter] = None, + set_training_mode=True): + + global_step = epoch * len(data_loader) + + metric_logger = utils.MetricLogger(delimiter=" ") + metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}')) + metric_logger.add_meter('n_ways', utils.SmoothedValue(window_size=1, fmt='{value:d}')) + metric_logger.add_meter('n_imgs', utils.SmoothedValue(window_size=1, fmt='{value:d}')) + header = 'Epoch: [{}]'.format(epoch) + print_freq = 10 + + model.train(set_training_mode) + + for batch in metric_logger.log_every(data_loader, print_freq, header): + batch = to_device(batch, device) + SupportTensor, SupportLabel, QueryTensor, QueryLabel, GlobalID_S, GlobalID_Q = batch + #print('SupportTensor:', SupportTensor.size(), 'SupportLabel:', SupportLabel, 'x:', x.size(), 'y:', y.size()) + + epsilon_list = [0.8, 0.08, 0.008] + # forward + with torch.cuda.amp.autocast(fp16): + #output = model(SupportTensor, SupportLabel, x) + scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv = model.set_forward_loss_StyAdv(SupportTensor,QueryTensor,SupportLabel, QueryLabel, GlobalID_S,GlobalID_Q, epsilon_list) + if(scores_fsl_ori.equal(scores_fsl_adv)): + loss_fsl_KL = 0 + else: + loss_fsl_KL = consistency_loss(scores_fsl_ori, scores_fsl_adv, 'KL3') + if(scores_cls_ori.equal(scores_cls_adv)): + loss_cls_KL = 0 + else: + loss_cls_KL = consistency_loss(scores_cls_ori, scores_cls_adv,'KL3') + + k1, k2, k3, k4, k5, k6 = 1, 1, 1, 1, 0, 0 + loss = k1 * loss_fsl_ori + k2 * loss_fsl_adv + k3 * loss_fsl_KL + k4 * loss_cls_ori + k5 * loss_cls_adv + k6 * loss_cls_KL + #print('loss_fsl_ori:', loss_fsl_ori, 'loss_fsl_adv:', loss_fsl_adv, 'loss_fsl_KL:', loss_fsl_KL, 'loss_cls_ori:', loss_cls_ori, 'loss_cls_adv:',loss_cls_adv, 'loss_cls_adv') + #output = output.view(QueryTensor.shape[0] * QueryTensor.shape[1], -1) + #QueryLabel = QueryLabel.view(-1) + #loss = criterion(output, QueryLabel) + loss_value = loss.item() + + if not math.isfinite(loss_value): + print("Loss is {}, stopping training".format(loss_value)) + sys.exit(1) + + optimizer.zero_grad() + + if fp16: + # this attribute is added by timm on one optimizer (adahessian) + is_second_order = hasattr(optimizer, 'is_second_order') and optimizer.is_second_order + loss_scaler(loss, optimizer, clip_grad=max_norm, + parameters=model.parameters(), create_graph=is_second_order) + else: + loss.backward() + optimizer.step() + + torch.cuda.synchronize() + if model_ema is not None: + model_ema.update(model) + + lr = optimizer.param_groups[0]["lr"] + metric_logger.update(loss=loss_value) + metric_logger.update(lr=lr) + metric_logger.update(n_ways=SupportLabel.max()+1) + metric_logger.update(n_imgs=SupportTensor.shape[1] + QueryTensor.shape[1]) + + # tensorboard + if utils.is_main_process() and global_step % print_freq == 0: + writer.add_scalar("train/loss", scalar_value=loss_value, global_step=global_step) + writer.add_scalar("train/lr", scalar_value=lr, global_step=global_step) + + global_step += 1 + + # gather the stats from all processes + metric_logger.synchronize_between_processes() + print("Averaged stats:", metric_logger) + return {k: meter.global_avg for k, meter in metric_logger.meters.items()} + + +def evaluate(data_loaders, model, criterion, device, seed=None, ep=None): + if isinstance(data_loaders, dict): + test_stats_lst = {} + test_stats_glb = {} + + for j, (source, data_loader) in enumerate(data_loaders.items()): + print(f'* Evaluating {source}:') + seed_j = seed + j if seed else None + test_stats = _evaluate(data_loader, model, criterion, device, seed_j) + test_stats_lst[source] = test_stats + test_stats_glb[source] = test_stats['acc1'] + + # apart from individual's acc1, accumulate metrics over all domains to compute mean + for k in test_stats_lst[source].keys(): + test_stats_glb[k] = torch.tensor([test_stats[k] for test_stats in test_stats_lst.values()]).mean().item() + + return test_stats_glb + elif isinstance(data_loaders, torch.utils.data.DataLoader): # when args.eval = True + return _evaluate(data_loaders, model, criterion, device, seed, ep) + else: + warnings.warn(f'The structure of {data_loaders} is not recognizable.') + return _evaluate(data_loaders, model, criterion, device, seed) + + +@torch.no_grad() +def _evaluate(data_loader, model, criterion, device, seed=None, ep=None): + metric_logger = utils.MetricLogger(delimiter=" ") + metric_logger.add_meter('n_ways', utils.SmoothedValue(window_size=1, fmt='{value:d}')) + metric_logger.add_meter('n_imgs', utils.SmoothedValue(window_size=1, fmt='{value:d}')) + metric_logger.add_meter('acc1', utils.SmoothedValue(window_size=len(data_loader.dataset))) + metric_logger.add_meter('acc5', utils.SmoothedValue(window_size=len(data_loader.dataset))) + # added for debug + #metric_logger.add_meter('loss', utils.SmoothedValue(window_size=len(data_loader.dataset))) + header = 'Test:' + + # switch to evaluation mode + model.eval() + + if seed is not None: + data_loader.generator.manual_seed(seed) + + for ii, batch in enumerate(metric_logger.log_every(data_loader, 10, header)): + if ep is not None: + if ii > ep: + break + + batch = to_device(batch, device) + SupportTensor, SupportLabel, x, y = batch + #print('SupportTensor:', SupportTensor.size(), 'SupportLabel:', SupportLabel, 'x:', x.size(), 'y:', y.size()) + + # compute output + with torch.cuda.amp.autocast(): + output = model(SupportTensor, SupportLabel, x) + + output = output.view(x.shape[0] * x.shape[1], -1) + y = y.view(-1) + loss = criterion(output, y) + acc1, acc5 = accuracy(output, y, topk=(1, 5)) + + batch_size = x.shape[0] + metric_logger.update(loss=loss.item()) + # for debug + #metric_logger.meters['loss'].update(loss.item(), n=batch_size) + metric_logger.meters['acc1'].update(acc1.item(), n=batch_size) + metric_logger.meters['acc5'].update(acc5.item(), n=batch_size) + metric_logger.update(n_ways=SupportLabel.max()+1) + metric_logger.update(n_imgs=SupportTensor.shape[1] + x.shape[1]) + + # gather the stats from all processes + metric_logger.synchronize_between_processes() + + # initial + #print('* Acc@1 {top1.global_avg:.3f} Acc@5 {top5.global_avg:.3f} loss {losses.global_avg:.3f}' + # .format(top1=metric_logger.acc1, top5=metric_logger.acc5, losses=metric_logger.loss)) + + ret_dict = {k: meter.global_avg for k, meter in metric_logger.meters.items()} + ret_dict['acc_std'] = metric_logger.meters['acc1'].std + print('ret dict:', ret_dict['acc_std'], metric_logger.meters['acc1'], metric_logger.meters['acc1'].std) + + ''' + # debug for test BSCDFSL + ret_dict['acc_std'] = metric_logger.meters['acc1'].std + ''' + return ret_dict diff --git a/methods/gnn.py b/methods/gnn.py new file mode 100644 index 0000000000000000000000000000000000000000..1a33229918e2aef099cf87d634067befe061d555 --- /dev/null +++ b/methods/gnn.py @@ -0,0 +1,167 @@ +# This code is modified from https://github.com/vgsatorras/few-shot-gnn/blob/master/models/gnn_iclr.py + +import torch +import torch.nn as nn +from torch.autograd import Variable +import torch.nn.functional as F +from methods.backbone import Linear_fw, Conv2d_fw, BatchNorm2d_fw, BatchNorm1d_fw + +if torch.cuda.is_available(): + dtype = torch.cuda.FloatTensor + dtype_l = torch.cuda.LongTensor +else: + dtype = torch.FloatTensor + dtype_l = torch.cuda.LongTensor + +def gmul(input): + W, x = input + # x is a tensor of size (bs, N, num_features) + # W is a tensor of size (bs, N, N, J) + #x_size = x.size() + W_size = W.size() + N = W_size[-2] + W = W.split(1, 3) + W = torch.cat(W, 1).squeeze(3) # W is now a tensor of size (bs, J*N, N) + output = torch.bmm(W, x) # output has size (bs, J*N, num_features) + output = output.split(N, 1) + output = torch.cat(output, 2) # output has size (bs, N, J*num_features) + return output + +class Gconv(nn.Module): + maml = False + def __init__(self, nf_input, nf_output, J, bn_bool=True): + super(Gconv, self).__init__() + self.J = J + self.num_inputs = J*nf_input + self.num_outputs = nf_output + self.fc = nn.Linear(self.num_inputs, self.num_outputs) if not self.maml else Linear_fw(self.num_inputs, self.num_outputs) + + self.bn_bool = bn_bool + if self.bn_bool: + self.bn = nn.BatchNorm1d(self.num_outputs, track_running_stats=False) if not self.maml else BatchNorm1d_fw(self.num_outputs, track_running_stats=False) + + def forward(self, input): + W = input[0] + x = gmul(input) # out has size (bs, N, num_inputs) + #if self.J == 1: + # x = torch.abs(x) + x_size = x.size() + x = x.contiguous() + x = x.view(-1, self.num_inputs) + x = self.fc(x) # has size (bs*N, num_outputs) + + if self.bn_bool: + x = self.bn(x) + x = x.view(*x_size[:-1], self.num_outputs) + return W, x + +class Wcompute(nn.Module): + maml = False + def __init__(self, input_features, nf, operator='J2', activation='softmax', ratio=[2,2,1,1], num_operators=1, drop=False): + super(Wcompute, self).__init__() + self.num_features = nf + self.operator = operator + self.conv2d_1 = nn.Conv2d(input_features, int(nf * ratio[0]), 1, stride=1) if not self.maml else Conv2d_fw(input_features, int(nf * ratio[0]), 1, stride=1) + self.bn_1 = nn.BatchNorm2d(int(nf * ratio[0]), track_running_stats=False) if not self.maml else BatchNorm2d_fw(int(nf * ratio[0]), track_running_stats=False) + self.drop = drop + if self.drop: + self.dropout = nn.Dropout(0.3) + self.conv2d_2 = nn.Conv2d(int(nf * ratio[0]), int(nf * ratio[1]), 1, stride=1) if not self.maml else Conv2d_fw(int(nf * ratio[0]), int(nf * ratio[1]), 1, stride=1) + self.bn_2 = nn.BatchNorm2d(int(nf * ratio[1]), track_running_stats=False) if not self.maml else BatchNorm2d_fw(int(nf * ratio[1]), track_running_stats=False) + self.conv2d_3 = nn.Conv2d(int(nf * ratio[1]), nf*ratio[2], 1, stride=1) if not self.maml else Conv2d_fw(int(nf * ratio[1]), nf*ratio[2], 1, stride=1) + self.bn_3 = nn.BatchNorm2d(nf*ratio[2], track_running_stats=False) if not self.maml else BatchNorm2d_fw(nf*ratio[2], track_running_stats=False) + self.conv2d_4 = nn.Conv2d(nf*ratio[2], nf*ratio[3], 1, stride=1) if not self.maml else Conv2d_fw(nf*ratio[2], nf*ratio[3], 1, stride=1) + self.bn_4 = nn.BatchNorm2d(nf*ratio[3], track_running_stats=False) if not self.maml else BatchNorm2d_fw(nf*ratio[3], track_running_stats=False) + self.conv2d_last = nn.Conv2d(nf, num_operators, 1, stride=1) if not self.maml else Conv2d_fw(nf, num_operators, 1, stride=1) + self.activation = activation + + def forward(self, x, W_id): + W1 = x.unsqueeze(2) + W2 = torch.transpose(W1, 1, 2) #size: bs x N x N x num_features + W_new = torch.abs(W1 - W2) #size: bs x N x N x num_features + W_new = torch.transpose(W_new, 1, 3) #size: bs x num_features x N x N + + W_new = self.conv2d_1(W_new) + W_new = self.bn_1(W_new) + W_new = F.leaky_relu(W_new) + if self.drop: + W_new = self.dropout(W_new) + + W_new = self.conv2d_2(W_new) + W_new = self.bn_2(W_new) + W_new = F.leaky_relu(W_new) + + W_new = self.conv2d_3(W_new) + W_new = self.bn_3(W_new) + W_new = F.leaky_relu(W_new) + + W_new = self.conv2d_4(W_new) + W_new = self.bn_4(W_new) + W_new = F.leaky_relu(W_new) + + W_new = self.conv2d_last(W_new) + W_new = torch.transpose(W_new, 1, 3) #size: bs x N x N x 1 + + if self.activation == 'softmax': + W_new = W_new - W_id.expand_as(W_new) * 1e8 + W_new = torch.transpose(W_new, 2, 3) + # Applying Softmax + W_new = W_new.contiguous() + W_new_size = W_new.size() + W_new = W_new.view(-1, W_new.size(3)) + W_new = F.softmax(W_new, dim=1) + W_new = W_new.view(W_new_size) + # Softmax applied + W_new = torch.transpose(W_new, 2, 3) + + elif self.activation == 'sigmoid': + W_new = F.sigmoid(W_new) + W_new *= (1 - W_id) + elif self.activation == 'none': + W_new *= (1 - W_id) + else: + raise (NotImplementedError) + + if self.operator == 'laplace': + W_new = W_id - W_new + elif self.operator == 'J2': + W_new = torch.cat([W_id, W_new], 3) + else: + raise(NotImplementedError) + + return W_new + +class GNN_nl(nn.Module): + def __init__(self, input_features, nf, train_N_way): + super(GNN_nl, self).__init__() + self.input_features = input_features + self.nf = nf + self.num_layers = 2 + + for i in range(self.num_layers): + if i == 0: + module_w = Wcompute(self.input_features, nf, operator='J2', activation='softmax', ratio=[2, 2, 1, 1]) + module_l = Gconv(self.input_features, int(nf / 2), 2) + else: + module_w = Wcompute(self.input_features + int(nf / 2) * i, nf, operator='J2', activation='softmax', ratio=[2, 2, 1, 1]) + module_l = Gconv(self.input_features + int(nf / 2) * i, int(nf / 2), 2) + self.add_module('layer_w{}'.format(i), module_w) + self.add_module('layer_l{}'.format(i), module_l) + + self.w_comp_last = Wcompute(self.input_features + int(self.nf / 2) * self.num_layers, nf, operator='J2', activation='softmax', ratio=[2, 2, 1, 1]) + self.layer_last = Gconv(self.input_features + int(self.nf / 2) * self.num_layers, train_N_way, 2, bn_bool=False) + + def forward(self, x): + W_init = torch.eye(x.size(1), device=x.device).unsqueeze(0).repeat(x.size(0), 1, 1).unsqueeze(3) + + for i in range(self.num_layers): + Wi = self._modules['layer_w{}'.format(i)](x, W_init) + #print('Wi:', Wi.size()) + + x_new = F.leaky_relu(self._modules['layer_l{}'.format(i)]([Wi, x])[1]) + x = torch.cat([x, x_new], 2) + + Wl=self.w_comp_last(x, W_init) + out = self.layer_last([Wl, x])[1] + + return out diff --git a/methods/gnnnet.py b/methods/gnnnet.py new file mode 100644 index 0000000000000000000000000000000000000000..f74135f15f6e073ed1dcac650a004df21575e102 --- /dev/null +++ b/methods/gnnnet.py @@ -0,0 +1,75 @@ +import torch +import torch.nn as nn +import numpy as np +from methods.meta_template import MetaTemplate +from methods.gnn import GNN_nl +from methods import backbone + +class GnnNet(MetaTemplate): + maml=False + def __init__(self, model_func, n_way, n_support, tf_path=None): + super(GnnNet, self).__init__(model_func, n_way, n_support, tf_path=tf_path) + + # loss function + self.loss_fn = nn.CrossEntropyLoss() + + # metric function + self.fc = nn.Sequential(nn.Linear(self.feat_dim, 128), nn.BatchNorm1d(128, track_running_stats=False)) if not self.maml else nn.Sequential(backbone.Linear_fw(self.feat_dim, 128), backbone.BatchNorm1d_fw(128, track_running_stats=False)) + self.gnn = GNN_nl(128 + self.n_way, 96, self.n_way) + self.method = 'GnnNet' + + # fix label for training the metric function 1*nw(1 + ns)*nw + support_label = torch.from_numpy(np.repeat(range(self.n_way), self.n_support)).unsqueeze(1) + support_label = torch.zeros(self.n_way*self.n_support, self.n_way).scatter(1, support_label, 1).view(self.n_way, self.n_support, self.n_way) + support_label = torch.cat([support_label, torch.zeros(self.n_way, 1, n_way)], dim=1) + self.support_label = support_label.view(1, -1, self.n_way) + + def cuda(self): + self.feature.cuda() + self.fc.cuda() + self.gnn.cuda() + self.support_label = self.support_label.cuda() + return self + + def set_forward(self,x,is_feature=False): + x = x.cuda() + + if is_feature: + # reshape the feature tensor: n_way * n_s + 15 * f + assert(x.size(1) == self.n_support + 15) + z = self.fc(x.view(-1, *x.size()[2:])) + z = z.view(self.n_way, -1, z.size(1)) + else: + # get feature using encoder + x = x.view(-1, *x.size()[2:]) + z = self.fc(self.feature(x)) + z = z.view(self.n_way, -1, z.size(1)) + #print('z:', z.size()) + # stack the feature for metric function: n_way * n_s + n_q * f -> n_q * [1 * n_way(n_s + 1) * f] + z_stack = [torch.cat([z[:, :self.n_support], z[:, self.n_support + i:self.n_support + i + 1]], dim=1).view(1, -1, z.size(2)) for i in range(self.n_query)] + assert(z_stack[0].size(1) == self.n_way*(self.n_support + 1)) + #print('z_stack:', 'len:', len(z_stack), 'z_stack[0]:', z_stack[0].size()) + scores = self.forward_gnn(z_stack) + return scores + + def forward_gnn(self, zs): + # gnn inp: n_q * n_way(n_s + 1) * f + nodes = torch.cat([torch.cat([z, self.support_label], dim=2) for z in zs], dim=0) + #print('nodes:', nodes.size()) + scores = self.gnn(nodes) + + # n_q * n_way(n_s + 1) * n_way -> (n_way * n_q) * n_way + scores = scores.view(self.n_query, self.n_way, self.n_support + 1, self.n_way)[:, :, -1].permute(1, 0, 2).contiguous().view(-1, self.n_way) + return scores + + def set_forward_loss(self, x): + #print('gnnnet:', 'set forward loss:') + #print('1: x:', x.size()) + y_query = torch.from_numpy(np.repeat(range( self.n_way ), self.n_query)) + #print('2: y_query:', y_query) + y_query = y_query.cuda() + scores = self.set_forward(x) + #print('3: scores:', scores.size()) + loss = self.loss_fn(scores, y_query) + #print('4: loss:', loss) + return scores, loss diff --git a/methods/load_ViT_models.py b/methods/load_ViT_models.py new file mode 100644 index 0000000000000000000000000000000000000000..30f9d612e88ce03c87559a0476c08e90edd7bb27 --- /dev/null +++ b/methods/load_ViT_models.py @@ -0,0 +1,94 @@ +import torch +#from models import vision_transformer as vit +#from models import vision_transformer_multiBlocks_20221030 as vit +#from methods import vision_transformer_multiBlocks_20221030 as vit +from methods import ViT as vit +#import vision_transformer_multiBlocks_20221030 as vit +#from models.pmf_protonet import ProtoNet +#from methods.pmf_protonet import ProtoNet +from methods.protonet import ProtoNet +#from pmf_protonet import ProtoNet +#from models.cvpr2023_gnnnet_20221102 import GnnNet +#from methods.cvpr2023_gnnnet_20221102 import GnnNet +#from cvpr2023_gnnnet_20221102 import GnnNet + +def load_ViTsmall(no_pretrain=False): + model = vit.__dict__['vit_small'](patch_size=16, num_classes=0) + if(not no_pretrain): + url = "dino_deitsmall16_pretrain/dino_deitsmall16_pretrain.pth" + state_dict = torch.hub.load_state_dict_from_url(url="https://dl.fbaipublicfiles.com/dino/" + url) + model.load_state_dict(state_dict, strict=True) + #print('Pretrained weights found at {}'.format(url)) + #print('model defined.') + return model + +def load_ViTbase(no_pretrain=False): + model = vit.__dict__['vit_base'](patch_size=16, num_classes=0) + if(not no_pretrain): + url = "dino_vitbase16_pretrain/dino_vitbase16_pretrain.pth" + state_dict = torch.hub.load_state_dict_from_url(url="https://dl.fbaipublicfiles.com/dino/" + url) + model.load_state_dict(state_dict, strict=True) + print('Pretrained weights found at {}'.format(url)) + print('model defined.') + return model + + +def load_ResNet50(no_pretrain=False): + from torchvision.models.resnet import resnet50 + pretrained = not no_pretrain + model = resnet50(pretrained=pretrained) + model.fc = torch.nn.Identity() + print('model defined.') + return model + +def load_ResNet50_dino(no_pretrain=False): + from torchvision.models.resnet import resnet50 + model = resnet50(pretrained=False) + model.fc = torch.nn.Identity() + if not no_pretrain: + state_dict = torch.hub.load_state_dict_from_url(url="https://dl.fbaipublicfiles.com/dino/dino_resnet50_pretrain/dino_resnet50_pretrain.pth",map_location="cpu",) + model.load_state_dict(state_dict, strict=False) + return model + +def load_ResNet50_clip(no_pretrain=False): + from models import clip + model, _ = clip.load('RN50', 'cpu') + return model + + +def get_model(backbone='vit_small', classifier='protonet', args=None, styleAdv=False): + if(backbone=='vit_small' and classifier == 'protonet'): + extractor = load_ViTsmall() + if(not styleAdv): + #from models.pmf_protonet import ProtoNet + from methods.protonet import ProtoNet + model = ProtoNet(extractor) + else: + #from models.pmf_protonet_metatrain_vit_protonet_20221102 import ProtoNet + #from methods.pmf_protonet_metatrain_vit_protonet_20221102 import ProtoNet + from methods.StyleAdv_ViT_protonet import ProtoNet + model = ProtoNet(extractor) + + if(backbone=='resnet50' and classifier == 'protonet'): + extractor = load_ResNet50_dino() + model = ProtoNet(extractor) + + if(backbone=='vit_small' and classifier == 'gnnnet'): + extractor = load_ViTsmall() + model = GnnNet(extractor, backbone_flag='vit_small', n_way = 5, n_support = args.nSupport) + + if(backbone=='resnet50' and classifier == 'gnnnet'): + extractor = load_ResNet50_dino() + model = GnnNet(extractor, backbone_flag='resnet50', n_way = 5, n_support = args.nSupport) + return model + + + + +if __name__ == '__main__': + input = torch.randn(16, 3, 224, 224) + print('input:', input.size()) + model = load_ViTsmall() + out = model(input) + print('out:', out.size()) + diff --git a/methods/meta_template.py b/methods/meta_template.py new file mode 100644 index 0000000000000000000000000000000000000000..a8bf030366a7b3b220b69cc45b365e313f60d3fd --- /dev/null +++ b/methods/meta_template.py @@ -0,0 +1,92 @@ +import torch.nn as nn +import numpy as np +from abc import abstractmethod +from tensorboardX import SummaryWriter +print('Ori meta template.') +class MetaTemplate(nn.Module): + def __init__(self, model_func, n_way, n_support, flatten=True, leakyrelu=False, tf_path=None, change_way=True): + super(MetaTemplate, self).__init__() + self.n_way = n_way + self.n_support = n_support + self.n_query = -1 #(change depends on input) + self.feature = model_func(flatten=flatten, leakyrelu=leakyrelu) + self.feat_dim = self.feature.final_feat_dim + self.change_way = change_way #some methods allow different_way classification during training and test + self.tf_writer = SummaryWriter(log_dir=tf_path) if tf_path is not None else None + + @abstractmethod + def set_forward(self,x,is_feature): + pass + + @abstractmethod + def set_forward_loss(self, x): + pass + + def forward(self,x): + out = self.feature.forward(x) + return out + + def parse_feature(self,x,is_feature): + x = x.cuda() + if is_feature: + z_all = x + else: + x = x.contiguous().view( self.n_way * (self.n_support + self.n_query), *x.size()[2:]) + z_all = self.feature.forward(x) + z_all = z_all.view( self.n_way, self.n_support + self.n_query, -1) + z_support = z_all[:, :self.n_support] + z_query = z_all[:, self.n_support:] + + return z_support, z_query + + def correct(self, x): + scores, loss = self.set_forward_loss(x) + y_query = np.repeat(range( self.n_way ), self.n_query ) + + topk_scores, topk_labels = scores.data.topk(1, 1, True, True) + topk_ind = topk_labels.cpu().numpy() + top1_correct = np.sum(topk_ind[:,0] == y_query) + return float(top1_correct), len(y_query), loss.item()*len(y_query) + + def train_loop(self, epoch, train_loader, optimizer, total_it): + print_freq = len(train_loader) // 10 + avg_loss=0 + for i, (x,_ ) in enumerate(train_loader): + self.n_query = x.size(1) - self.n_support + if self.change_way: + self.n_way = x.size(0) + optimizer.zero_grad() + _, loss = self.set_forward_loss(x) + loss.backward() + optimizer.step() + avg_loss = avg_loss+loss.item() + + if (i + 1) % print_freq==0: + print('Epoch {:d} | Batch {:d}/{:d} | Loss {:f}'.format(epoch, i + 1, len(train_loader), avg_loss/float(i+1))) + if (total_it + 1) % 10 == 0 and self.tf_writer is not None: + self.tf_writer.add_scalar(self.method + '/query_loss', loss.item(), total_it + 1) + total_it += 1 + return total_it + + def test_loop(self, test_loader, record = None): + loss = 0. + count = 0 + acc_all = [] + + iter_num = len(test_loader) + for i, (x,_) in enumerate(test_loader): + self.n_query = x.size(1) - self.n_support + if self.change_way: + self.n_way = x.size(0) + correct_this, count_this, loss_this = self.correct(x) + acc_all.append(correct_this/ count_this*100 ) + loss += loss_this + count += count_this + + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print('--- %d Loss = %.6f ---' %(iter_num, loss/count)) + print('--- %d Test Acc = %4.2f%% +- %4.2f%% ---' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + + return acc_mean diff --git a/methods/meta_template_CausalStyle_RN_GNN.py b/methods/meta_template_CausalStyle_RN_GNN.py new file mode 100644 index 0000000000000000000000000000000000000000..e2391d502e52163ee422e80e9c6d46a87efa6103 --- /dev/null +++ b/methods/meta_template_CausalStyle_RN_GNN.py @@ -0,0 +1,138 @@ +import torch.nn as nn +import torch +import numpy as np +from abc import abstractmethod +from tensorboardX import SummaryWriter +import torch.nn.functional as F +from methods.tool_func import consistency_loss + +print("hi this is causal style") + +class MetaTemplate(nn.Module): + def __init__(self, model_func, n_way, n_support, flatten=True, leakyrelu=False, tf_path=None, change_way=True): + super(MetaTemplate, self).__init__() + self.n_way = n_way + self.n_support = n_support + self.n_query = -1 #(change depends on input) + self.feature = model_func(flatten=flatten, leakyrelu=leakyrelu) + self.feat_dim = self.feature.final_feat_dim + self.change_way = change_way #some methods allow different_way classification during training and test + self.tf_writer = SummaryWriter(log_dir=tf_path) if tf_path is not None else None + + @abstractmethod + def set_forward(self,x,is_feature): + pass + + @abstractmethod + def set_forward_loss(self, x): + pass + + def forward(self,x): + out = self.feature.forward(x) + return out + + def parse_feature(self,x,is_feature): + x = x.cuda() + if is_feature: + z_all = x + else: + x = x.contiguous().view( self.n_way * (self.n_support + self.n_query), *x.size()[2:]) + z_all = self.feature.forward(x) + z_all = z_all.view( self.n_way, self.n_support + self.n_query, -1) + z_support = z_all[:, :self.n_support] + z_query = z_all[:, self.n_support:] + + return z_support, z_query + + def correct(self, x): + scores, loss = self.set_forward_loss(x) + y_query = np.repeat(range( self.n_way ), self.n_query ) + + topk_scores, topk_labels = scores.data.topk(1, 1, True, True) + topk_ind = topk_labels.cpu().numpy() + top1_correct = np.sum(topk_ind[:,0] == y_query) + return float(top1_correct), len(y_query), loss.item()*len(y_query) + + + def train_loop(self, epoch, train_loader_ori, optimizer, total_it): + print_freq = len(train_loader_ori) // 10 + avg_loss=0 + for i, (x_ori, global_y ) in enumerate(train_loader_ori): + self.n_query = x_ori.size(1) - self.n_support + if self.change_way: + self.n_way = x_ori.size(0) + optimizer.zero_grad() + + epsilon_list = [0.8, 0.08, 0.008] + + scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv, scores_fsl_x, loss_fsl_x = self.set_forward_loss_CausalStyle(x_ori, global_y, epsilon_list) + + # consistency loss between initial and styleAdv + if(scores_fsl_ori.equal(scores_fsl_adv)): + loss_fsl_KL = 0 + else: + loss_fsl_KL = consistency_loss(scores_fsl_ori, scores_fsl_adv, 'KL3') + + if(scores_cls_ori.equal(scores_cls_adv)): + loss_cls_KL = 0 + else: + loss_cls_KL = consistency_loss(scores_cls_ori, scores_cls_adv,'KL3') + + # consistency loss between initial and x + if(scores_fsl_ori.equal(scores_fsl_x)): + loss_fsl_KL_x = 0 + else: + loss_fsl_KL_x = consistency_loss(scores_fsl_ori, scores_fsl_x, 'KL3') + + + # final loss + #k1, k2, k3, k4, k5, k6 = 1, 1, 1, 1, 0, 0 + #loss = k1 * loss_fsl_ori + k2 * loss_fsl_adv + k3 * loss_fsl_KL + k4 * loss_cls_ori + k5 * loss_cls_adv + k6 * loss_cls_KL + + print('loss_fsl_x:', loss_fsl_x, 'loss_fsl_KL_x:', loss_fsl_KL_x) + loss = loss_fsl_ori + loss_fsl_adv + loss_fsl_KL + loss_cls_ori + loss_fsl_x + loss_fsl_KL_x + + loss.backward() + optimizer.step() + avg_loss = avg_loss+loss.item() + + if (i + 1) % print_freq==0: + print('Epoch {:d} | Batch {:d}/{:d} | Loss {:f}'.format(epoch, i + 1, len(train_loader_ori), avg_loss/float(i+1))) + if (total_it + 1) % 10 == 0 and self.tf_writer is not None: + self.tf_writer.add_scalar('loss_fsl_ori:', loss_fsl_ori.item(), total_it +1) + self.tf_writer.add_scalar('loss_fsl_adv:', loss_fsl_adv.item(), total_it +1) + self.tf_writer.add_scalar('loss_fsl_x:', loss_fsl_adv.item(), total_it +1) + #self.tf_writer.add_scalar('loss_fsl_KL:', loss_fsl_KL.item(), total_it +1) + self.tf_writer.add_scalar('loss_cls_ori:', loss_cls_ori.item(), total_it +1) + #self.tf_writer.add_scalar('loss_cls_adv:', loss_cls_adv.item(), total_it +1) + self.tf_writer.add_scalar('loss_fsl_Kl:', loss_fsl_KL, total_it +1) + self.tf_writer.add_scalar('loss_fsl_Kl_x:', loss_fsl_KL_x, total_it +1) + self.tf_writer.add_scalar('total_loss:', loss.item(), total_it +1) + # intial + self.tf_writer.add_scalar(self.method + '/query_loss', loss.item(), total_it + 1) + + total_it += 1 + return total_it + + def test_loop(self, test_loader, record = None): + loss = 0. + count = 0 + acc_all = [] + + iter_num = len(test_loader) + for i, (x,_) in enumerate(test_loader): + self.n_query = x.size(1) - self.n_support + if self.change_way: + self.n_way = x.size(0) + correct_this, count_this, loss_this = self.correct(x) + acc_all.append(correct_this/ count_this*100 ) + loss += loss_this + count += count_this + + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print('--- %d Loss = %.6f ---' %(iter_num, loss/count)) + print('--- %d Test Acc = %4.2f%% +- %4.2f%% ---' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + + return acc_mean diff --git a/methods/meta_template_StyleAdv_RN_GNN.py b/methods/meta_template_StyleAdv_RN_GNN.py new file mode 100644 index 0000000000000000000000000000000000000000..b0915645228d1f6126fdaafa1d29115bc589e23e --- /dev/null +++ b/methods/meta_template_StyleAdv_RN_GNN.py @@ -0,0 +1,126 @@ +import torch.nn as nn +import torch +import numpy as np +from abc import abstractmethod +from tensorboardX import SummaryWriter +import torch.nn.functional as F +from methods.tool_func import consistency_loss + + +class MetaTemplate(nn.Module): + def __init__(self, model_func, n_way, n_support, flatten=True, leakyrelu=False, tf_path=None, change_way=True): + super(MetaTemplate, self).__init__() + self.n_way = n_way + self.n_support = n_support + self.n_query = -1 #(change depends on input) + self.feature = model_func(flatten=flatten, leakyrelu=leakyrelu) + self.feat_dim = self.feature.final_feat_dim + self.change_way = change_way #some methods allow different_way classification during training and test + self.tf_writer = SummaryWriter(log_dir=tf_path) if tf_path is not None else None + + @abstractmethod + def set_forward(self,x,is_feature): + pass + + @abstractmethod + def set_forward_loss(self, x): + pass + + def forward(self,x): + out = self.feature.forward(x) + return out + + def parse_feature(self,x,is_feature): + x = x.cuda() + if is_feature: + z_all = x + else: + x = x.contiguous().view( self.n_way * (self.n_support + self.n_query), *x.size()[2:]) + z_all = self.feature.forward(x) + z_all = z_all.view( self.n_way, self.n_support + self.n_query, -1) + z_support = z_all[:, :self.n_support] + z_query = z_all[:, self.n_support:] + + return z_support, z_query + + def correct(self, x): + scores, loss = self.set_forward_loss(x) + y_query = np.repeat(range( self.n_way ), self.n_query ) + + topk_scores, topk_labels = scores.data.topk(1, 1, True, True) + topk_ind = topk_labels.cpu().numpy() + top1_correct = np.sum(topk_ind[:,0] == y_query) + return float(top1_correct), len(y_query), loss.item()*len(y_query) + + + def train_loop(self, epoch, train_loader_ori, optimizer, total_it): + print_freq = len(train_loader_ori) // 10 + avg_loss=0 + for i, (x_ori, global_y ) in enumerate(train_loader_ori): + self.n_query = x_ori.size(1) - self.n_support + if self.change_way: + self.n_way = x_ori.size(0) + optimizer.zero_grad() + + epsilon_list = [0.8, 0.08, 0.008] + + scores_fsl_ori, loss_fsl_ori, scores_cls_ori, loss_cls_ori, scores_fsl_adv, loss_fsl_adv, scores_cls_adv, loss_cls_adv = self.set_forward_loss_StyAdv(x_ori, global_y, epsilon_list) + + # consistency loss between initial and styleAdv + if(scores_fsl_ori.equal(scores_fsl_adv)): + loss_fsl_KL = 0 + else: + loss_fsl_KL = consistency_loss(scores_fsl_ori, scores_fsl_adv, 'KL3') + + if(scores_cls_ori.equal(scores_cls_adv)): + loss_cls_KL = 0 + else: + loss_cls_KL = consistency_loss(scores_cls_ori, scores_cls_adv,'KL3') + + + # final loss + k1, k2, k3, k4, k5, k6 = 1, 1, 1, 1, 0, 0 + loss = k1 * loss_fsl_ori + k2 * loss_fsl_adv + k3 * loss_fsl_KL + k4 * loss_cls_ori + k5 * loss_cls_adv + k6 * loss_cls_KL + #loss = loss_fsl_ori + loss.backward() + optimizer.step() + avg_loss = avg_loss+loss.item() + + if (i + 1) % print_freq==0: + print('Epoch {:d} | Batch {:d}/{:d} | Loss {:f}'.format(epoch, i + 1, len(train_loader_ori), avg_loss/float(i+1))) + if (total_it + 1) % 10 == 0 and self.tf_writer is not None: + self.tf_writer.add_scalar('loss_fsl_ori:', loss_fsl_ori.item(), total_it +1) + self.tf_writer.add_scalar('loss_fsl_adv:', loss_fsl_adv.item(), total_it +1) + #self.tf_writer.add_scalar('loss_fsl_KL:', loss_fsl_KL.item(), total_it +1) + self.tf_writer.add_scalar('loss_cls_ori:', loss_cls_ori.item(), total_it +1) + self.tf_writer.add_scalar('loss_cls_adv:', loss_cls_adv.item(), total_it +1) + #self.tf_writer.add_scalar('loss_cls_Kl:', loss_cls_KL.item(), total_it +1) + self.tf_writer.add_scalar('total_loss:', loss.item(), total_it +1) + # intial + self.tf_writer.add_scalar(self.method + '/query_loss', loss.item(), total_it + 1) + + total_it += 1 + return total_it + + def test_loop(self, test_loader, record = None): + loss = 0. + count = 0 + acc_all = [] + + iter_num = len(test_loader) + for i, (x,_) in enumerate(test_loader): + self.n_query = x.size(1) - self.n_support + if self.change_way: + self.n_way = x.size(0) + correct_this, count_this, loss_this = self.correct(x) + acc_all.append(correct_this/ count_this*100 ) + loss += loss_this + count += count_this + + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print('--- %d Loss = %.6f ---' %(iter_num, loss/count)) + print('--- %d Test Acc = %4.2f%% +- %4.2f%% ---' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + + return acc_mean diff --git a/methods/model_utils.py b/methods/model_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..0ed25646780b6ddf9a8c582e82f6ef55e333abcc --- /dev/null +++ b/methods/model_utils.py @@ -0,0 +1,238 @@ +import math +import torch +import warnings +import ml_collections +import random +import torch.nn.functional as F + + +def DiffAugment(x, types=[], prob = 0.5, detach=True): + """ + x.shape = B, C, H, W + """ + if random.random() < prob: + with torch.set_grad_enabled(not detach): + x = random_hflip(x, prob=0.5) + for p in types: + for f in AUGMENT_FNS[p]: + x = f(x) + x = x.contiguous() + return x + + +def random_hflip(tensor, prob): + if prob > random.random(): + return tensor + return torch.flip(tensor, dims=(3,)) + +def rand_brightness(x): + x = x + (torch.rand(x.size(0), 1, 1, 1, dtype=x.dtype, device=x.device) - 0.5) + return x + +def rand_saturation(x): + x_mean = x.mean(dim=1, keepdim=True) + x = (x - x_mean) * (torch.rand(x.size(0), 1, 1, 1, dtype=x.dtype, device=x.device) * 2) + x_mean + return x + +def rand_contrast(x): + x_mean = x.mean(dim=[1, 2, 3], keepdim=True) + x = (x - x_mean) * (torch.rand(x.size(0), 1, 1, 1, dtype=x.dtype, device=x.device) + 0.5) + x_mean + return x + +def rand_translation(x, ratio=0.125): + shift_x, shift_y = int(x.size(2) * ratio + 0.5), int(x.size(3) * ratio + 0.5) + translation_x = torch.randint(-shift_x, shift_x + 1, size=[x.size(0), 1, 1], device=x.device) + translation_y = torch.randint(-shift_y, shift_y + 1, size=[x.size(0), 1, 1], device=x.device) + grid_batch, grid_x, grid_y = torch.meshgrid( + torch.arange(x.size(0), dtype=torch.long, device=x.device), + torch.arange(x.size(2), dtype=torch.long, device=x.device), + torch.arange(x.size(3), dtype=torch.long, device=x.device), + ) + grid_x = torch.clamp(grid_x + translation_x + 1, 0, x.size(2) + 1) + grid_y = torch.clamp(grid_y + translation_y + 1, 0, x.size(3) + 1) + x_pad = F.pad(x, [1, 1, 1, 1, 0, 0, 0, 0]) + x = x_pad.permute(0, 2, 3, 1).contiguous()[grid_batch, grid_x, grid_y].permute(0, 3, 1, 2) + return x + +def rand_offset(x, ratio=1, ratio_h=1, ratio_v=1): + w, h = x.size(2), x.size(3) + + imgs = [] + for img in x.unbind(dim = 0): + max_h = int(w * ratio * ratio_h) + max_v = int(h * ratio * ratio_v) + + value_h = random.randint(0, max_h) * 2 - max_h + value_v = random.randint(0, max_v) * 2 - max_v + + if abs(value_h) > 0: + img = torch.roll(img, value_h, 2) + + if abs(value_v) > 0: + img = torch.roll(img, value_v, 1) + + imgs.append(img) + + return torch.stack(imgs) + +def rand_offset_h(x, ratio=1): + return rand_offset(x, ratio=1, ratio_h=ratio, ratio_v=0) + +def rand_offset_v(x, ratio=1): + return rand_offset(x, ratio=1, ratio_h=0, ratio_v=ratio) + +def rand_cutout(x, ratio=0.5): + cutout_size = int(x.size(2) * ratio + 0.5), int(x.size(3) * ratio + 0.5) + offset_x = torch.randint(0, x.size(2) + (1 - cutout_size[0] % 2), size=[x.size(0), 1, 1], device=x.device) + offset_y = torch.randint(0, x.size(3) + (1 - cutout_size[1] % 2), size=[x.size(0), 1, 1], device=x.device) + grid_batch, grid_x, grid_y = torch.meshgrid( + torch.arange(x.size(0), dtype=torch.long, device=x.device), + torch.arange(cutout_size[0], dtype=torch.long, device=x.device), + torch.arange(cutout_size[1], dtype=torch.long, device=x.device), + ) + grid_x = torch.clamp(grid_x + offset_x - cutout_size[0] // 2, min=0, max=x.size(2) - 1) + grid_y = torch.clamp(grid_y + offset_y - cutout_size[1] // 2, min=0, max=x.size(3) - 1) + mask = torch.ones(x.size(0), x.size(2), x.size(3), dtype=x.dtype, device=x.device) + mask[grid_batch, grid_x, grid_y] = 0 + x = x * mask.unsqueeze(1) + return x + + +AUGMENT_FNS = { + 'color': [rand_brightness, rand_saturation, rand_contrast], + 'offset': [rand_offset], + 'offset_h': [rand_offset_h], + 'offset_v': [rand_offset_v], + 'translation': [rand_translation], + 'cutout': [rand_cutout], +} + + +def _no_grad_trunc_normal_(tensor, mean, std, a, b): + # Cut & paste from PyTorch official master until it's in a few official releases - RW + # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf + def norm_cdf(x): + # Computes standard normal cumulative distribution function + return (1. + math.erf(x / math.sqrt(2.))) / 2. + + if (mean < a - 2 * std) or (mean > b + 2 * std): + warnings.warn("mean is more than 2 std from [a, b] in nn.init.trunc_normal_. " + "The distribution of values may be incorrect.", + stacklevel=2) + + with torch.no_grad(): + # Values are generated by using a truncated uniform distribution and + # then using the inverse CDF for the normal distribution. + # Get upper and lower cdf values + l = norm_cdf((a - mean) / std) + u = norm_cdf((b - mean) / std) + + # Uniformly fill tensor with values from [l, u], then translate to + # [2l-1, 2u-1]. + tensor.uniform_(2 * l - 1, 2 * u - 1) + + # Use inverse cdf transform for normal distribution to get truncated + # standard normal + tensor.erfinv_() + + # Transform to proper mean, std + tensor.mul_(std * math.sqrt(2.)) + tensor.add_(mean) + + # Clamp to ensure it's in the proper range + tensor.clamp_(min=a, max=b) + return tensor + + +def trunc_normal_(tensor, mean=0., std=1., a=-2., b=2.): + # type: (Tensor, float, float, float, float) -> Tensor + return _no_grad_trunc_normal_(tensor, mean, std, a, b) + + +def get_testing(): + """Returns a minimal configuration for testing.""" + config = ml_collections.ConfigDict() + config.patches = ml_collections.ConfigDict({'size': (16, 16)}) + config.hidden_size = 1 + config.transformer = ml_collections.ConfigDict() + config.transformer.mlp_dim = 1 + config.transformer.num_heads = 1 + config.transformer.num_layers = 1 + config.transformer.attention_dropout_rate = 0.0 + config.transformer.dropout_rate = 0.1 + config.classifier = 'token' + config.representation_size = None + return config + + +def get_b16_config(): + """Returns the ViT-B/16 configuration.""" + config = ml_collections.ConfigDict() + config.patches = ml_collections.ConfigDict({'size': (16, 16)}) + config.hidden_size = 768 + config.transformer = ml_collections.ConfigDict() + config.transformer.mlp_dim = 3072 + config.transformer.num_heads = 12 + config.transformer.num_layers = 12 + config.transformer.attention_dropout_rate = 0.0 + config.transformer.dropout_rate = 0.1 + config.classifier = 'token' + config.representation_size = None + return config + + +def get_r50_b16_config(): + """Returns the Resnet50 + ViT-B/16 configuration.""" + config = get_b16_config() + del config.patches.size + config.patches.grid = (14, 14) + config.resnet = ml_collections.ConfigDict() + config.resnet.num_layers = (3, 4, 9) + config.resnet.width_factor = 1 + return config + + +def get_b32_config(): + """Returns the ViT-B/32 configuration.""" + config = get_b16_config() + config.patches.size = (32, 32) + return config + + +def get_l16_config(): + """Returns the ViT-L/16 configuration.""" + config = ml_collections.ConfigDict() + config.patches = ml_collections.ConfigDict({'size': (16, 16)}) + config.hidden_size = 1024 + config.transformer = ml_collections.ConfigDict() + config.transformer.mlp_dim = 4096 + config.transformer.num_heads = 16 + config.transformer.num_layers = 24 + config.transformer.attention_dropout_rate = 0.0 + config.transformer.dropout_rate = 0.1 + config.classifier = 'token' + config.representation_size = None + return config + + +def get_l32_config(): + """Returns the ViT-L/32 configuration.""" + config = get_l16_config() + config.patches.size = (32, 32) + return config + + +def get_h14_config(): + """Returns the ViT-L/16 configuration.""" + config = ml_collections.ConfigDict() + config.patches = ml_collections.ConfigDict({'size': (14, 14)}) + config.hidden_size = 1280 + config.transformer = ml_collections.ConfigDict() + config.transformer.mlp_dim = 5120 + config.transformer.num_heads = 16 + config.transformer.num_layers = 32 + config.transformer.attention_dropout_rate = 0.0 + config.transformer.dropout_rate = 0.1 + config.classifier = 'token' + config.representation_size = None + return config diff --git a/methods/protonet.py b/methods/protonet.py new file mode 100644 index 0000000000000000000000000000000000000000..15e1ceb9776a40a059469fdbe4d66771478e3738 --- /dev/null +++ b/methods/protonet.py @@ -0,0 +1,57 @@ +import torch +import torch.nn as nn +import torch.nn.functional as F + + +class ProtoNet(nn.Module): + def __init__(self, backbone): + super().__init__() + + # bias & scale of cosine classifier + self.bias = nn.Parameter(torch.FloatTensor(1).fill_(0), requires_grad=True) + self.scale_cls = nn.Parameter(torch.FloatTensor(1).fill_(10), requires_grad=True) + + # backbone + self.backbone = backbone + + def cos_classifier(self, w, f): + """ + w.shape = B, nC, d + f.shape = B, M, d + """ + f = F.normalize(f, p=2, dim=f.dim()-1, eps=1e-12) + w = F.normalize(w, p=2, dim=w.dim()-1, eps=1e-12) + + cls_scores = f @ w.transpose(1, 2) # B, M, nC + cls_scores = self.scale_cls * (cls_scores + self.bias) + return cls_scores + + def forward(self, supp_x, supp_y, x): + """ + supp_x.shape = [B, nSupp, C, H, W] + supp_y.shape = [B, nSupp] + x.shape = [B, nQry, C, H, W] + """ + num_classes = supp_y.max() + 1 # NOTE: assume B==1 + B, nSupp, C, H, W = supp_x.shape + supp_f = self.backbone.forward(supp_x.contiguous().view(-1, C, H, W)) + ''' + # for test vit_multiBlock (equals to forward()) + supp_f = self.backbone.forward_block1(supp_x.contiguous().view(-1, C, H, W)) + supp_f = self.backbone.forward_block2(supp_f) + supp_f = self.backbone.forward_block3(supp_f) + supp_f = self.backbone.forward_block4(supp_f) + supp_f = self.backbone.forward_rest(supp_f) + ''' + supp_f = supp_f.view(B, nSupp, -1) + supp_y_1hot = F.one_hot(supp_y, num_classes).transpose(1, 2) # B, nC, nSupp + + # B, nC, nSupp x B, nSupp, d = B, nC, d + prototypes = torch.bmm(supp_y_1hot.float(), supp_f) + prototypes = prototypes / supp_y_1hot.sum(dim=2, keepdim=True) # NOTE: may div 0 if some classes got 0 images + + feat = self.backbone.forward(x.view(-1, C, H, W)) + feat = feat.view(B, x.shape[1], -1) # B, nQry, d + + logits = self.cos_classifier(prototypes, feat) # B, nQry, nC + return logits diff --git a/methods/tool_func.py b/methods/tool_func.py new file mode 100644 index 0000000000000000000000000000000000000000..92f7761dc8daaac4c1e184b10d0768fd752739ac --- /dev/null +++ b/methods/tool_func.py @@ -0,0 +1,68 @@ + +import torch +import torch.nn as nn +import random +import numpy as np +import torch.nn.functional as F + +EPS=0.00001 +#P_THRED = 0.2 +P_THRED = 0.4 +START_EPS = 16/255 + +def calc_mean_std(feat, eps=1e-5): + # eps is a small value added to the variance to avoid divide-by-zero. + size = feat.size() + assert (len(size) == 4) + N, C = size[:2] + feat_var = feat.view(N, C, -1).var(dim=2) + eps + feat_std = feat_var.sqrt().view(N, C, 1, 1) + feat_mean = feat.view(N, C, -1).mean(dim=2).view(N, C, 1, 1) + return feat_mean, feat_std + +def fgsm_attack(init_input, epsilon, data_grad): + # random start init_input + init_input = init_input + torch.empty_like(init_input).uniform_(START_EPS, START_EPS) + + sign_data_grad = data_grad.sign() + adv_input = init_input + epsilon*sign_data_grad + return adv_input + +def changeNewAdvStyle(input_fea, new_styleAug_mean, new_styleAug_std, p_thred): + if(new_styleAug_mean=='None'): + return input_fea + + p = np.random.uniform() + if( p < p_thred): + return input_fea + + feat_size = input_fea.size() + ori_style_mean, ori_style_std = calc_mean_std(input_fea) + normalized_fea = (input_fea - ori_style_mean.expand(feat_size)) / ori_style_std.expand(feat_size) + styleAug_fea = normalized_fea * new_styleAug_std.expand(feat_size) + new_styleAug_mean.expand(feat_size) + return styleAug_fea + +def consistency_loss(scoresM1, scoresM2, type='euclidean'): + if(type=='euclidean'): + avg_pro = (scoresM1 + scoresM2)/2.0 + matrix1 = torch.sqrt(torch.sum((scoresM1 - avg_pro)**2,dim=1)) + matrix2 = torch.sqrt(torch.sum((scoresM2 - avg_pro)**2,dim=1)) + dis1 = torch.mean(matrix1) + dis2 = torch.mean(matrix2) + dis = (dis1+dis2)/2.0 + elif(type=='KL1'): + avg_pro = (scoresM1 + scoresM2)/2.0 + matrix1 = torch.sum( F.softmax(scoresM1,dim=-1) * (F.log_softmax(scoresM1, dim=-1) - F.log_softmax(avg_pro,dim=-1)), 1) + matrix2 = torch.sum( F.softmax(scoresM2,dim=-1) * (F.log_softmax(scoresM2, dim=-1) - F.log_softmax(avg_pro,dim=-1)), 1) + dis1 = torch.mean(matrix1) + dis2 = torch.mean(matrix2) + dis = (dis1+dis2)/2.0 + elif(type=='KL2'): + matrix = torch.sum( F.softmax(scoresM2,dim=-1) * (F.log_softmax(scoresM2, dim=-1) - F.log_softmax(scoresM1,dim=-1)), 1) + dis = torch.mean(matrix) + elif(type=='KL3'): + matrix = torch.sum( F.softmax(scoresM1,dim=-1) * (F.log_softmax(scoresM1, dim=-1) - F.log_softmax(scoresM2,dim=-1)), 1) + dis = torch.mean(matrix) + else: + return + return dis diff --git a/options.py b/options.py new file mode 100644 index 0000000000000000000000000000000000000000..f7b4bb9ccb6ffb1a9cf57ae1970d10d4e12be975 --- /dev/null +++ b/options.py @@ -0,0 +1,126 @@ +import numpy as np +import os +import glob +import torch +import argparse + +def parse_args(script): + parser = argparse.ArgumentParser(description= 'few-shot script %s' %(script)) + parser.add_argument('--dataset', default='multi', help='miniImagenet/cub/cars/places/plantae, specify multi for training with multiple domains') + parser.add_argument('--testset', default='cub', help='cub/cars/places/plantae, valid only when dataset=multi') + parser.add_argument('--model', default='ResNet10', help='model: Conv{4|6} / ResNet{10|18|34}') # we use ResNet10 in the paper + parser.add_argument('--method', default='baseline', help='baseline/baseline++/protonet/matchingnet/relationnet{_softmax}/gnnnet') + parser.add_argument('--train_n_way' , default=5, type=int, help='class num to classify for training') + parser.add_argument('--test_n_way' , default=5, type=int, help='class num to classify for testing (validation) ') + parser.add_argument('--n_shot' , default=5, type=int, help='number of labeled data in each class, same as n_support') + parser.add_argument('--train_aug' , action='store_true', help='perform data augmentation or not during training ') + parser.add_argument('--name' , default='tmp', type=str, help='') + parser.add_argument('--save_dir' , default='./output', type=str, help='') + #parser.add_argument('--data_dir' , default='./filelists', type=str, help='') + parser.add_argument('--data_dir' , default='/scratch/yuqian_fu/Data/CDFSL/', type=str, help='') + + # for finetuning + parser.add_argument('--finetune_epoch', default=50, type=int, help='') + parser.add_argument('--resume_dir', default='Pretrain', type=str, help='continue from previous trained model with largest epoch') + + if script == 'train': + #parser.add_argument('--num_classes' , default=200, type=int, help='total number of classes in softmax, only used in baseline') + # bug fixed + parser.add_argument('--num_classes' , default=64, type=int, help='total number of classes in softmax, only used in baseline') + parser.add_argument('--save_freq' , default=100, type=int, help='Save frequency') + parser.add_argument('--target_set', default='cub', help='cub/cars/places/plantae, use the extremely labeled target data') + parser.add_argument('--target_num_label', default=5, type=int, help='number of labeled target base images per class') + parser.add_argument('--start_epoch' , default=0, type=int,help ='Starting epoch') + parser.add_argument('--stop_epoch' , default=200, type=int, help ='Stopping epoch') + parser.add_argument('--resume' , default='', type=str, help='continue from previous trained model with largest epoch') + parser.add_argument('--resume_epoch', default=-1, type=int, help='') + parser.add_argument('--warmup' , default='gg3b0', type=str, help='continue from baseline, neglected if resume is true') + elif script == 'test': + parser.add_argument('--split' , default='novel', help='base/val/novel') + parser.add_argument('--save_epoch', default=400, type=int,help ='load the model trained in x epoch, use the best model if x is -1') + parser.add_argument('--warmup' , default='gg3bo', type = str, help = 'just for insert the test function into the training.') + #parser.add_argument(''--target_set', default='cub', help=' + parser.add_argument('--stop_epoch' , default=400, type=int, help ='Stopping epoch') + + else: + raise ValueError('Unknown script') + + return parser.parse_args() + +def get_assigned_file(checkpoint_dir,num): + assign_file = os.path.join(checkpoint_dir, '{:d}.tar'.format(num)) + return assign_file + +def get_resume_file(checkpoint_dir, resume_epoch=-1): + filelist = glob.glob(os.path.join(checkpoint_dir, '*.tar')) + if len(filelist) == 0: + return None + + filelist = [ x for x in filelist if os.path.basename(x) != 'best_model.tar' ] + epochs = np.array([int(os.path.splitext(os.path.basename(x))[0]) for x in filelist]) + max_epoch = np.max(epochs) + epoch = max_epoch if resume_epoch == -1 else resume_epoch + resume_file = os.path.join(checkpoint_dir, '{:d}.tar'.format(epoch)) + return resume_file + +def get_best_file(checkpoint_dir): + best_file = os.path.join(checkpoint_dir, 'best_model.tar') + if os.path.isfile(best_file): + return best_file + else: + return get_resume_file(checkpoint_dir) + +def load_warmup_state(filename, method): + print(' load pre-trained model file: {}'.format(filename)) + warmup_resume_file = get_resume_file(filename) + print(' warmup_resume_file:', warmup_resume_file) + tmp = torch.load(warmup_resume_file) + if tmp is not None: + state = tmp['state'] + state_keys = list(state.keys()) + for i, key in enumerate(state_keys): + if 'relationnet' in method and "feature." in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + elif method == 'gnnnet' and 'feature.' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + elif method == 'matchingnet' and 'feature.' in key and '.7.' not in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + elif method == 'tpn' and 'feature.' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + else: + state.pop(key) + else: + raise ValueError(' No pre-trained encoder file found!') + return state + + + +# added in 20210110 +def load_warmup_state_speci(filename, method): + print(' load pre-trained model file: {}'.format(filename)) + #warmup_resume_file = get_resume_file(filename) + warmup_resume_file = filename + print(' warmup_resume_file:', warmup_resume_file) + tmp = torch.load(warmup_resume_file) + if tmp is not None: + state = tmp['state'] + state_keys = list(state.keys()) + for i, key in enumerate(state_keys): + if 'relationnet' in method and "feature." in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + elif method == 'gnnnet' and 'feature.' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + elif method == 'matchingnet' and 'feature.' in key and '.7.' not in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + else: + state.pop(key) + else: + raise ValueError(' No pre-trained encoder file found!') + return state diff --git a/tackle_dataset.py b/tackle_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..de8152636b164857e8e148178e87904f2c30c751 --- /dev/null +++ b/tackle_dataset.py @@ -0,0 +1,204 @@ +import os +import json + +mini_global_path = '/scratch/yuqian_fu/Data/CDFSL/miniImagenet/' + +#imagenet base +def miniImg(flag='test'): + file_path = mini_global_path + flag + '.json' + with open(file_path, 'r') as file: + data = json.load(file) + + print(data.keys()) + image_names = data['image_names'] + lengthes = len(image_names) + print(lengthes) + + for i in range(lengthes): + print(image_names[i]) + splits = image_names[i].split('\n') + print(splits) + if (flag == 'novel'): + new_line = mini_global_path + 'test' + '/n' + splits[-2] + '/n' + splits[-1] + else: + new_line = mini_global_path + flag + '/n' + splits[-2] + '/n' + splits[-1] + print(new_line) + if os.path.exists(new_line): + pass + else: + print('file:', new_line, 'does not exist!') + break + data['image_names'][i] = new_line + + with open(file_path, 'w') as file: + json.dump(data, file, indent=4) + + + with open(file_path, 'r') as file: + data = json.load(file) + + for i in range(lengthes): + print(data['image_names'][i]) + return + + +#cub +mini_global_path_cub = '/scratch/yuqian_fu/Data/CDFSL/cub/' +def cub(flag='base'): + file_path = mini_global_path_cub + flag + '.json' + with open(file_path, 'r') as file: + data = json.load(file) + + print(data.keys()) + image_names = data['image_names'] + lengthes = len(image_names) + print(lengthes) + + for i in range(lengthes): + print(image_names[i]) + new_line = image_names[i].replace('/wy_test/qiuxingyu/CrossDomainFewShot/filelists/cub/', mini_global_path_cub) + print(new_line) + if os.path.exists(new_line): + pass + else: + print('file:', new_line, 'does not exist!') + break + data['image_names'][i] = new_line + + with open(file_path, 'w') as file: + json.dump(data, file, indent=4) + + with open(file_path, 'r') as file: + data = json.load(file) + for i in range(lengthes): + print(data['image_names'][i]) + return + +#cars +mini_global_path_cars = '/scratch/yuqian_fu/Data/CDFSL/cars/' +def cars(flag='base'): + file_path = mini_global_path_cars + flag + '.json' + with open(file_path, 'r') as file: + data = json.load(file) + + print(data.keys()) + image_names = data['image_names'] + lengthes = len(image_names) + print(lengthes) + + for i in range(lengthes): + print(image_names[i]) + new_line = image_names[i].replace('/wy_test/qiuxingyu/CrossDomainFewShot/filelists/cars/', mini_global_path_cars) + print(new_line) + if os.path.exists(new_line): + pass + else: + print('file:', new_line, 'does not exist!') + break + data['image_names'][i] = new_line + + with open(file_path, 'w') as file: + json.dump(data, file, indent=4) + + with open(file_path, 'r') as file: + data = json.load(file) + for i in range(lengthes): + print(data['image_names'][i]) + return + + + +#places +#mini_global_path_places = '/scratch/yuqian_fu/Data/CDFSL/places365_standard/' +mini_global_path_places = '/scratch/yuqian_fu/Data/CDFSL/places/' +def places(flag='base'): + file_path = mini_global_path_places + flag + '.json' + with open(file_path, 'r') as file: + data = json.load(file) + + print(data.keys()) + image_names = data['image_names'] + lengthes = len(image_names) + print(lengthes) + + for i in range(lengthes): + print(image_names[i]) + #new_line = image_names[i].replace('/wy_test/qiuxingyu/CrossDomainFewShot/filelists/places/source/places365_standard/', mini_global_path_places) + new_line = image_names[i].replace('places365_standard', 'places') + print(new_line) + if os.path.exists(new_line): + pass + else: + print('file:', new_line, 'does not exist!') + break + data['image_names'][i] = new_line + + with open(file_path, 'w') as file: + json.dump(data, file, indent=4) + + with open(file_path, 'r') as file: + data = json.load(file) + for i in range(lengthes): + print(data['image_names'][i]) + return + +#places +#mini_global_path_places = '/scratch/yuqian_fu/Data/CDFSL/places365_standard/' +mini_global_path_plantae = '/scratch/yuqian_fu/Data/CDFSL/plantae/' +def plantae(flag='base'): + file_path = mini_global_path_plantae + flag + '.json' + with open(file_path, 'r') as file: + data = json.load(file) + + print(data.keys()) + image_names = data['image_names'] + lengthes = len(image_names) + print(lengthes) + + for i in range(lengthes): + print(image_names[i]) + #new_line = image_names[i].replace('/wy_test/qiuxingyu/CrossDomainFewShot/filelists/places/source/places365_standard/', mini_global_path_places) + new_line = image_names[i].replace('Plantae', 'plantae') + print(new_line) + if os.path.exists(new_line): + pass + else: + print('file:', new_line, 'does not exist!') + break + data['image_names'][i] = new_line + + with open(file_path, 'w') as file: + json.dump(data, file, indent=4) + + with open(file_path, 'r') as file: + data = json.load(file) + for i in range(lengthes): + print(data['image_names'][i]) + return + + + +if __name__ == '__main__': + #miniImg(flag='base') + #miniImg(flag='val') + #miniImg(flag='novel') + + # cub(flag='base') + # cub(flag='val') + # cub(flag='novel') + + # cars(flag='base') + # cars(flag='val') + # cars(flag='novel') + + # places(flag='base') + # places(flag='val') + # places(flag='novel') + + plantae(flag='base') + plantae(flag='val') + plantae(flag='novel') + + + + diff --git a/test.py b/test.py new file mode 100644 index 0000000000000000000000000000000000000000..7c22cd5c891c3b76259a9b50d8c112d2ded9ca7e --- /dev/null +++ b/test.py @@ -0,0 +1,82 @@ +import numpy as np +import torch +import torch.optim +import os +import random + +from methods import backbone +from methods.backbone_multiblock import model_dict +from data.datamgr import SimpleDataManager, SetDataManager +from methods.StyleAdv_RN_GNN import StyleAdvGNN + +from options import parse_args, get_resume_file, load_warmup_state +from test_function_fwt_benchmark import test_bestmodel +from test_function_bscdfsl_benchmark import test_bestmodel_bscdfsl + + +def record_test_result(params): + acc_file_path = "tmp2.txt" + acc_file = open(acc_file_path, "w") + epoch_id = -1 + print( + "epoch", + epoch_id, + "miniImagenet:", + "cub:", + "cars:", + "places:", + "plantae:", + file=acc_file, + ) + name = params.name + n_shot = params.n_shot + method = params.method + test_bestmodel(acc_file, name, method, "miniImagenet", n_shot, epoch_id) + # test_bestmodel(acc_file, name, method, 'cub', n_shot, epoch_id) + # test_bestmodel(acc_file, name, method, 'cars', n_shot, epoch_id) + # test_bestmodel(acc_file, name, method, 'places', n_shot, epoch_id) + # test_bestmodel(acc_file, name, method, 'plantae', n_shot, epoch_id) + + acc_file.close() + return + + +def record_test_result_bscdfsl(params): + print("hhhhhhh testing for bscdfsl") + acc_file_path = "tmp_bscdfsl2.txt" + acc_file = open(acc_file_path, "w") + epoch_id = -1 + print( + "epoch", epoch_id, "ChestX:", "ISIC:", "EuroSAT:", "CropDisease", file=acc_file + ) + name = params.name + n_shot = params.n_shot + method = params.method + # test_bestmodel_bscdfsl(acc_file, name, method, 'ChestX', n_shot, epoch_id) + # test_bestmodel_bscdfsl(acc_file, name, method, 'ISIC', n_shot, epoch_id) + test_bestmodel_bscdfsl(acc_file, name, method, "EuroSAT", n_shot, epoch_id) + # test_bestmodel_bscdfsl(acc_file, name, method, 'CropDisease', n_shot, epoch_id) + + acc_file.close() + return + + +# --- main function --- +if __name__ == "__main__": + # fix seed + seed = 0 + print("set seed = %d" % seed) + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) + torch.cuda.manual_seed_all(seed) + torch.backends.cudnn.deterministic = True + torch.backends.cudnn.benchmark = False + + # parser argument + params = parse_args("train") + + # testing + # record_test_result(params) + # testing bscdfsl + record_test_result_bscdfsl(params) diff --git a/test_function_bscdfsl_benchmark.py b/test_function_bscdfsl_benchmark.py new file mode 100644 index 0000000000000000000000000000000000000000..1bac7157c0014d75f006fdcce314e2d22397bdc4 --- /dev/null +++ b/test_function_bscdfsl_benchmark.py @@ -0,0 +1,188 @@ +import torch +import os +import h5py +import random +import numpy as np +from options import parse_args, get_best_file, get_assigned_file +import data.feature_loader as feat_loader +from data.datamgr import SimpleDataManager +from data import ISIC_few_shot, EuroSAT_few_shot, CropDisease_few_shot, Chest_few_shot +from methods.backbone_multiblock import model_dict +from methods.StyleAdv_RN_GNN import StyleAdvGNN + +# extract and save image features +def save_features(model, data_loader, featurefile): + f = h5py.File(featurefile, 'w') + max_count = len(data_loader)*data_loader.batch_size + all_labels = f.create_dataset('all_labels',(max_count,), dtype='i') + all_feats=None + count=0 + for i, (x,y) in enumerate(data_loader): + if (i % 10) == 0: + print(' {:d}/{:d}'.format(i, len(data_loader))) + x = x.cuda() + feats = model(x) + if all_feats is None: + all_feats = f.create_dataset('all_feats', [max_count] + list( feats.size()[1:]) , dtype='f') + all_feats[count:count+feats.size(0)] = feats.data.cpu().numpy() + all_labels[count:count+feats.size(0)] = y.cpu().numpy() + count = count + feats.size(0) + + count_var = f.create_dataset('count', (1,), dtype='i') + count_var[0] = count + f.close() + +# evaluate using features +def feature_evaluation(cl_data_file, model, n_way = 5, n_support = 5, n_query = 15): + class_list = cl_data_file.keys() + class_list = list(class_list) + select_class = random.sample(class_list,n_way) + z_all = [] + for cl in select_class: + img_feat = cl_data_file[cl] + perm_ids = np.random.permutation(len(img_feat)).tolist() + z_all.append( [ np.squeeze( img_feat[perm_ids[i]]) for i in range(n_support+n_query) ] ) + z_all = torch.from_numpy(np.array(z_all) ) + + model.n_query = n_query + scores = model.set_forward(z_all, is_feature = True) + pred = scores.data.cpu().numpy().argmax(axis = 1) + y = np.repeat(range( n_way ), n_query ) + acc = np.mean(pred == y)*100 + return acc + + + +def test_bestmodel_bscdfsl(acc_file, name, method, dataset,n_shot, save_epoch=-1): + # parse argument + params = parse_args('test') + params.n_shot = n_shot + params.dataset = dataset + params.method = method + params.name = name + params.save_epoch = save_epoch #-1 = best + print('Testing! {} shots on {} dataset with {} epochs of {}({})'.format(params.n_shot, params.dataset, params.save_epoch, params.name, params.method)) + remove_featurefile = True + + print('\nStage 1: saving features') + # dataset + print(' build dataset') + image_size = 224 + split = params.split + if(params.dataset in ["miniImagenet", "cub", "cars", "places", "plantae"]): + loadfile = os.path.join(params.data_dir, params.dataset, split + '.json') + print('load file:', loadfile) + datamgr = SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(loadfile, aug = False) + + else: + if params.dataset in ["ISIC"]: + datamgr = ISIC_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + elif params.dataset in ["EuroSAT"]: + datamgr = EuroSAT_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + elif params.dataset in ["CropDisease"]: + datamgr = CropDisease_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + elif params.dataset in ["ChestX"]: + datamgr = Chest_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + + print(' build feature encoder') + # feature encoder + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir,params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + model = model_dict[params.model]() + model = model.cuda() + tmp = torch.load(modelfile) + try: + state = tmp['state'] + except KeyError: + state = tmp['model_state'] + except: + raise + state_keys = list(state.keys()) + print('state_keys:', state_keys, len(state_keys)) + for i, key in enumerate(state_keys): + if "feature." in key and not 'gamma' in key and not 'beta' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + else: + state.pop(key) + print('state keys:', list(state.keys()), len(list(state.keys()))) + + model.load_state_dict(state) + model.eval() + + # save feature file + print(' extract and save features...') + if params.save_epoch != -1: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + "_" + str(params.save_epoch)+ ".hdf5") + else: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + ".hdf5") + dirname = os.path.dirname(featurefile) + if not os.path.isdir(dirname): + os.makedirs(dirname) + save_features(model, data_loader, featurefile) + + print('\nStage 2: evaluate') + acc_all = [] + iter_num = 1000 + few_shot_params = dict(n_way = params.test_n_way , n_support = params.n_shot) + # model + print(' build metric-based model') + model = StyleAdvGNN( model_dict[params.model], **few_shot_params) + model = model.cuda() + model.eval() + + # load model + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir, params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + if modelfile is not None: + tmp = torch.load(modelfile) + try: + model.load_state_dict(tmp['state']) + except RuntimeError: + print('warning! RuntimeError when load_state_dict()!') + model.load_state_dict(tmp['state'], strict=False) + except KeyError: + for k in tmp['model_state']: ##### revise latter + if 'running' in k: + tmp['model_state'][k] = tmp['model_state'][k].squeeze() + model.load_state_dict(tmp['model_state'], strict=False) + except: + raise + + # load feature file + print(' load saved feature file') + cl_data_file = feat_loader.init_loader(featurefile) + #print('cl_data_file:', cl_data_file) + + # start evaluate + print(' evaluate') + for i in range(iter_num): + acc = feature_evaluation(cl_data_file, model, n_query=15, **few_shot_params) + acc_all.append(acc) + + # statics + print(' get statics') + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num)), file = acc_file) + + # remove feature files [optional] + if remove_featurefile: + os.remove(featurefile) diff --git a/test_function_bscdfsl_benchmark2.py b/test_function_bscdfsl_benchmark2.py new file mode 100644 index 0000000000000000000000000000000000000000..9601865dc5ece3306e23122181abe9d074702d07 --- /dev/null +++ b/test_function_bscdfsl_benchmark2.py @@ -0,0 +1,226 @@ +import torch +import os +import h5py + +from methods import backbone +from methods.backbone import model_dict +from data.datamgr import SimpleDataManager +from options import parse_args, get_best_file, get_assigned_file + +#from methods.protonet import ProtoNet +#from methods.matchingnet import MatchingNet +from methods.gnnnet import GnnNet +#from methods.relationnet import RelationNet +import data.feature_loader as feat_loader +import random +import numpy as np + +from data import ISIC_few_shot, EuroSAT_few_shot, CropDisease_few_shot, Chest_few_shot + +# extract and save image features +def save_features(model, data_loader, featurefile): + f = h5py.File(featurefile, 'w') + max_count = len(data_loader)*data_loader.batch_size + all_labels = f.create_dataset('all_labels',(max_count,), dtype='i') + all_feats=None + count=0 + for i, (x,y) in enumerate(data_loader): + if (i % 10) == 0: + print(' {:d}/{:d}'.format(i, len(data_loader))) + x = x.cuda() + feats = model(x) + if all_feats is None: + all_feats = f.create_dataset('all_feats', [max_count] + list( feats.size()[1:]) , dtype='f') + all_feats[count:count+feats.size(0)] = feats.data.cpu().numpy() + all_labels[count:count+feats.size(0)] = y.cpu().numpy() + count = count + feats.size(0) + + count_var = f.create_dataset('count', (1,), dtype='i') + count_var[0] = count + f.close() + +# evaluate using features +def feature_evaluation(cl_data_file, model, n_way = 5, n_support = 5, n_query = 15): + class_list = cl_data_file.keys() + select_class = random.sample(class_list,n_way) + z_all = [] + for cl in select_class: + img_feat = cl_data_file[cl] + perm_ids = np.random.permutation(len(img_feat)).tolist() + z_all.append( [ np.squeeze( img_feat[perm_ids[i]]) for i in range(n_support+n_query) ] ) + z_all = torch.from_numpy(np.array(z_all) ) + + model.n_query = n_query + scores = model.set_forward(z_all, is_feature = True) + pred = scores.data.cpu().numpy().argmax(axis = 1) + y = np.repeat(range( n_way ), n_query ) + acc = np.mean(pred == y)*100 + return acc + + + +def test_bestmodel_bscdfsl(acc_file, name, method, dataset,n_shot, save_epoch=-1): + # parse argument + print('hi, test model 1') + params = parse_args('test') + print (' hi, test model 2') + params.n_shot = n_shot + params.dataset = dataset + params.method = method + params.name = name + params.save_epoch = save_epoch #-1 = best + print('Testing! {} shots on {} dataset with {} epochs of {}({})'.format(params.n_shot, params.dataset, params.save_epoch, params.name, params.method)) + remove_featurefile = True + + print('\nStage 1: saving features') + # dataset + print(' build dataset') + if 'Conv' in params.model: + image_size = 84 + else: + image_size = 224 + + split = params.split + print(split) + if(params.dataset in ["miniImagenet", "cub", "cars", "places", "plantae"]): + loadfile = os.path.join(params.data_dir, params.dataset, split + '.json') + print('load file:', loadfile) + datamgr = SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(loadfile, aug = False) + + else: + if params.dataset in ["ISIC"]: + datamgr = ISIC_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + elif params.dataset in ["EuroSAT"]: + + datamgr = EuroSAT_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + elif params.dataset in ["CropDisease"]: + datamgr = CropDisease_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + elif params.dataset in ["ChestX"]: + datamgr = Chest_few_shot.SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(aug = False ) + + + print(' build feature encoder') + # feature encoder + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir,params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + if params.method in ['relationnet', 'relationnet_softmax']: + if params.model == 'Conv4': + model = backbone.Conv4NP() + elif params.model == 'Conv6': + model = backbone.Conv6NP() + else: + model = model_dict[params.model]( flatten = False ) + else: + model = model_dict[params.model]() + model = model.cuda() + tmp = torch.load(modelfile) + try: + state = tmp['state'] + except KeyError: + state = tmp['model_state'] + except: + raise + state_keys = list(state.keys()) + print('state_keys:', state_keys, len(state_keys)) + for i, key in enumerate(state_keys): + if "feature." in key and not 'gamma' in key and not 'beta' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + else: + state.pop(key) + print('state keys:', list(state.keys()), len(list(state.keys()))) + + model.load_state_dict(state) + model.eval() + + # save feature file + print(' extract and save features...') + if params.save_epoch != -1: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + "_" + str(params.save_epoch)+ ".hdf5") + else: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + ".hdf5") + dirname = os.path.dirname(featurefile) + if not os.path.isdir(dirname): + os.makedirs(dirname) + save_features(model, data_loader, featurefile) + + print('\nStage 2: evaluate') + acc_all = [] + iter_num = 1000 + few_shot_params = dict(n_way = params.test_n_way , n_support = params.n_shot) + # model + print(' build metric-based model') + if params.method == 'protonet': + model = ProtoNet( model_dict[params.model], **few_shot_params) + elif params.method == 'matchingnet': + model = MatchingNet( model_dict[params.model], **few_shot_params ) + elif params.method == 'gnnnet': + model = GnnNet( model_dict[params.model], **few_shot_params) + elif params.method in ['relationnet', 'relationnet_softmax']: + if params.model == 'Conv4': + feature_model = backbone.Conv4NP + elif params.model == 'Conv6': + feature_model = backbone.Conv6NP + else: + feature_model = model_dict[params.model] + loss_type = 'mse' if params.method == 'relationnet' else 'softmax' + model = RelationNet( feature_model, loss_type = loss_type , **few_shot_params ) + else: + raise ValueError('Unknown method') + model = model.cuda() + model.eval() + + # load model + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir, params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + if modelfile is not None: + tmp = torch.load(modelfile) + try: + model.load_state_dict(tmp['state']) + except RuntimeError: + print('warning! RuntimeError when load_state_dict()!') + model.load_state_dict(tmp['state'], strict=False) + except KeyError: + for k in tmp['model_state']: ##### revise latter + if 'running' in k: + tmp['model_state'][k] = tmp['model_state'][k].squeeze() + model.load_state_dict(tmp['model_state'], strict=False) + except: + raise + + # load feature file + print(' load saved feature file') + cl_data_file = feat_loader.init_loader(featurefile) + #print('cl_data_file:', cl_data_file) + + # start evaluate + print(' evaluate') + for i in range(iter_num): + acc = feature_evaluation(cl_data_file, model, n_query=15, **few_shot_params) + acc_all.append(acc) + + # statics + print(' get statics') + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num)), file = acc_file) + + # remove feature files [optional] + if remove_featurefile: + os.remove(featurefile) diff --git a/test_function_fwt_benchmark.py b/test_function_fwt_benchmark.py new file mode 100644 index 0000000000000000000000000000000000000000..cd70727b5b59b2186e8e06197355e024b5f1d272 --- /dev/null +++ b/test_function_fwt_benchmark.py @@ -0,0 +1,176 @@ +import torch +import os +import h5py + +#from methods import backbone +#from methods.backbone import model_dict +from methods.backbone_multiblock import model_dict +from data.datamgr import SimpleDataManager +from options import parse_args, get_best_file, get_assigned_file + +#from methods.gnnnet import GnnNet +from methods.StyleAdv_RN_GNN import StyleAdvGNN +import data.feature_loader as feat_loader +import random +import numpy as np + +# extract and save image features +def save_features(model, data_loader, featurefile): + f = h5py.File(featurefile, 'w') + max_count = len(data_loader)*data_loader.batch_size + all_labels = f.create_dataset('all_labels',(max_count,), dtype='i') + all_feats=None + count=0 + for i, (x,y) in enumerate(data_loader): + if (i % 10) == 0: + print(' {:d}/{:d}'.format(i, len(data_loader))) + x = x.cuda() + feats = model(x) + if all_feats is None: + all_feats = f.create_dataset('all_feats', [max_count] + list( feats.size()[1:]) , dtype='f') + all_feats[count:count+feats.size(0)] = feats.data.cpu().numpy() + all_labels[count:count+feats.size(0)] = y.cpu().numpy() + count = count + feats.size(0) + + count_var = f.create_dataset('count', (1,), dtype='i') + count_var[0] = count + f.close() + +# evaluate using features +def feature_evaluation(cl_data_file, model, n_way = 5, n_support = 5, n_query = 15): + + class_list = cl_data_file.keys() + class_list = list(class_list) + select_class = random.sample(class_list,n_way) + z_all = [] + for cl in select_class: + img_feat = cl_data_file[cl] + perm_ids = np.random.permutation(len(img_feat)).tolist() + z_all.append( [ np.squeeze( img_feat[perm_ids[i]]) for i in range(n_support+n_query) ] ) + z_all = torch.from_numpy(np.array(z_all) ) + #print('z_all:', z_all.size()) + + model.n_query = n_query + scores = model.set_forward(z_all, is_feature = True) + pred = scores.data.cpu().numpy().argmax(axis = 1) + y = np.repeat(range( n_way ), n_query ) + acc = np.mean(pred == y)*100 + return acc + + + +def test_bestmodel(acc_file, name, method, dataset,n_shot, save_epoch=-1): + # parse argument + params = parse_args('test') + params.n_shot = n_shot + params.dataset = dataset + params.method = method + params.name = name + params.save_epoch = save_epoch #-1 = best + print('Testing! {} shots on {} dataset with {} epochs of {}({})'.format(params.n_shot, params.dataset, params.save_epoch, params.name, params.method)) + remove_featurefile = True + + print('\nStage 1: saving features') + # dataset + print(' build dataset') + image_size = 224 + split = params.split + loadfile = os.path.join(params.data_dir, params.dataset, split + '.json') + print('load file:', loadfile) + datamgr = SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(loadfile, aug = False) + + print(' build feature encoder') + # feature encoder + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir,params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + print(modelfile) + model = model_dict[params.model]() + model = model.cuda() + tmp = torch.load(modelfile) + try: + state = tmp['state'] + except KeyError: + state = tmp['model_state'] + except: + raise + state_keys = list(state.keys()) + print('state_keys:', state_keys, len(state_keys)) + for i, key in enumerate(state_keys): + if "feature." in key and not 'gamma' in key and not 'beta' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + else: + state.pop(key) + print('state keys:', list(state.keys()), len(list(state.keys()))) + + model.load_state_dict(state) + model.eval() + + # save feature file + print(' extract and save features...') + if params.save_epoch != -1: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + "_" + str(params.save_epoch)+ ".hdf5") + else: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + ".hdf5") + dirname = os.path.dirname(featurefile) + if not os.path.isdir(dirname): + os.makedirs(dirname) + save_features(model, data_loader, featurefile) + + print('\nStage 2: evaluate') + acc_all = [] + iter_num = 1000 + few_shot_params = dict(n_way = params.test_n_way , n_support = params.n_shot) + # model + print(' build metric-based model') + #model = GnnNet( model_dict[params.model], **few_shot_params) + model = StyleAdvGNN(model_dict[params.model], **few_shot_params) + model = model.cuda() + model.eval() + + # load model + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir, params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + if modelfile is not None: + tmp = torch.load(modelfile) + try: + model.load_state_dict(tmp['state']) + except RuntimeError: + print('warning! RuntimeError when load_state_dict()!') + model.load_state_dict(tmp['state'], strict=False) + except KeyError: + for k in tmp['model_state']: ##### revise latter + if 'running' in k: + tmp['model_state'][k] = tmp['model_state'][k].squeeze() + model.load_state_dict(tmp['model_state'], strict=False) + except: + raise + + # load feature file + print(' load saved feature file') + cl_data_file = feat_loader.init_loader(featurefile) + + # start evaluate + print(' evaluate') + for i in range(iter_num): + acc = feature_evaluation(cl_data_file, model, n_query=15, **few_shot_params) + acc_all.append(acc) + + # statics + print(' get statics') + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num)), file = acc_file) + + # remove feature files [optional] + if remove_featurefile: + os.remove(featurefile) diff --git a/test_function_fwt_benchmark2.py b/test_function_fwt_benchmark2.py new file mode 100644 index 0000000000000000000000000000000000000000..5d5b4b689e5a3aef32f7a462a9c8f4f5fe73bafe --- /dev/null +++ b/test_function_fwt_benchmark2.py @@ -0,0 +1,211 @@ +import torch +import os +import h5py + +from methods import backbone +from methods.backbone import model_dict +from data.datamgr import SimpleDataManager +from options import parse_args, get_best_file, get_assigned_file + +#from methods.protonet import ProtoNet +#from methods.matchingnet import MatchingNet +from methods.gnnnet import GnnNet +#from methods.relationnet import RelationNet +#from methods.tpn import TPN +import data.feature_loader as feat_loader +import random +import numpy as np + +# extract and save image features +def save_features(model, data_loader, featurefile): + f = h5py.File(featurefile, 'w') + max_count = len(data_loader)*data_loader.batch_size + all_labels = f.create_dataset('all_labels',(max_count,), dtype='i') + all_feats=None + count=0 + for i, (x,y) in enumerate(data_loader): + if (i % 10) == 0: + print(' {:d}/{:d}'.format(i, len(data_loader))) + x = x.cuda() + feats = model(x) + if all_feats is None: + all_feats = f.create_dataset('all_feats', [max_count] + list( feats.size()[1:]) , dtype='f') + all_feats[count:count+feats.size(0)] = feats.data.cpu().numpy() + all_labels[count:count+feats.size(0)] = y.cpu().numpy() + count = count + feats.size(0) + + count_var = f.create_dataset('count', (1,), dtype='i') + count_var[0] = count + f.close() + +# evaluate using features +def feature_evaluation(cl_data_file, model, n_way = 5, n_support = 5, n_query = 15): + class_list = cl_data_file.keys() + select_class = random.sample(class_list,n_way) + z_all = [] + for cl in select_class: + img_feat = cl_data_file[cl] + perm_ids = np.random.permutation(len(img_feat)).tolist() + z_all.append( [ np.squeeze( img_feat[perm_ids[i]]) for i in range(n_support+n_query) ] ) + z_all = torch.from_numpy(np.array(z_all) ) + #print('z_all:', z_all.size()) + + model.n_query = n_query + scores = model.set_forward(z_all, is_feature = True) + pred = scores.data.cpu().numpy().argmax(axis = 1) + y = np.repeat(range( n_way ), n_query ) + acc = np.mean(pred == y)*100 + return acc + + + +def test_bestmodel(acc_file, name, method, dataset,n_shot, save_epoch=-1): + # parse argument + print('hi, test model 1') + params = parse_args('test') + print (' hi, test model 2') + params.n_shot = n_shot + params.dataset = dataset + params.method = method + params.name = name + params.save_epoch = save_epoch #-1 = best + print('Testing! {} shots on {} dataset with {} epochs of {}({})'.format(params.n_shot, params.dataset, params.save_epoch, params.name, params.method)) + remove_featurefile = True + + print('\nStage 1: saving features') + # dataset + print(' build dataset') + if 'Conv' in params.model: + image_size = 84 + else: + image_size = 224 + split = params.split + loadfile = os.path.join(params.data_dir, params.dataset, split + '.json') + print('load file:', loadfile) + datamgr = SimpleDataManager(image_size, batch_size = 64) + data_loader = datamgr.get_data_loader(loadfile, aug = False) + + print(' build feature encoder') + # feature encoder + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir,params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + if params.method in ['relationnet', 'relationnet_softmax','tpn']: + if params.model == 'Conv4': + model = backbone.Conv4NP() + elif params.model == 'Conv6': + model = backbone.Conv6NP() + else: + model = model_dict[params.model]( flatten = False ) + else: + model = model_dict[params.model]() + model = model.cuda() + tmp = torch.load(modelfile) + try: + state = tmp['state'] + except KeyError: + state = tmp['model_state'] + except: + raise + state_keys = list(state.keys()) + print('state_keys:', state_keys, len(state_keys)) + for i, key in enumerate(state_keys): + if "feature." in key and not 'gamma' in key and not 'beta' in key: + newkey = key.replace("feature.","") + state[newkey] = state.pop(key) + else: + state.pop(key) + print('state keys:', list(state.keys()), len(list(state.keys()))) + + model.load_state_dict(state) + if( params.method != 'tpn'): + model.eval() + + # save feature file + print(' extract and save features...') + if params.save_epoch != -1: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + "_" + str(params.save_epoch)+ ".hdf5") + else: + featurefile = os.path.join( checkpoint_dir.replace("checkpoints","features"), split + ".hdf5") + dirname = os.path.dirname(featurefile) + if not os.path.isdir(dirname): + os.makedirs(dirname) + save_features(model, data_loader, featurefile) + + print('\nStage 2: evaluate') + acc_all = [] + iter_num = 1000 + few_shot_params = dict(n_way = params.test_n_way , n_support = params.n_shot) + # model + print(' build metric-based model') + if params.method == 'protonet': + model = ProtoNet( model_dict[params.model], **few_shot_params) + elif params.method == 'matchingnet': + model = MatchingNet( model_dict[params.model], **few_shot_params ) + elif params.method == 'gnnnet': + model = GnnNet( model_dict[params.model], **few_shot_params) + elif params.method in ['relationnet', 'relationnet_softmax']: + if params.model == 'Conv4': + feature_model = backbone.Conv4NP + elif params.model == 'Conv6': + feature_model = backbone.Conv6NP + else: + feature_model = model_dict[params.model] + loss_type = 'mse' if params.method == 'relationnet' else 'softmax' + model = RelationNet( feature_model, loss_type = loss_type , **few_shot_params ) + elif params.method == 'tpn': + model = TPN( model_dict[params.model], **few_shot_params) + #backbone = model_dict[params.model](flatten=False) + #model = TPN(backbone, **few_shot_params) + else: + raise ValueError('Unknown method') + model = model.cuda() + + if( params.method != 'tpn'): + model.eval() + + # load model + checkpoint_dir = '%s/checkpoints/%s'%(params.save_dir, params.name) + if params.save_epoch != -1: + modelfile = get_assigned_file(checkpoint_dir, params.save_epoch) + else: + modelfile = get_best_file(checkpoint_dir) + if modelfile is not None: + tmp = torch.load(modelfile) + try: + model.load_state_dict(tmp['state']) + except RuntimeError: + print('warning! RuntimeError when load_state_dict()!') + model.load_state_dict(tmp['state'], strict=False) + except KeyError: + for k in tmp['model_state']: ##### revise latter + if 'running' in k: + tmp['model_state'][k] = tmp['model_state'][k].squeeze() + model.load_state_dict(tmp['model_state'], strict=False) + except: + raise + + # load feature file + print(' load saved feature file') + cl_data_file = feat_loader.init_loader(featurefile) + #print('cl_data_file:', cl_data_file) + + # start evaluate + print(' evaluate') + for i in range(iter_num): + acc = feature_evaluation(cl_data_file, model, n_query=15, **few_shot_params) + acc_all.append(acc) + + # statics + print(' get statics') + acc_all = np.asarray(acc_all) + acc_mean = np.mean(acc_all) + acc_std = np.std(acc_all) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num))) + print(' %d test iterations: Acc = %4.2f%% +- %4.2f%%' %(iter_num, acc_mean, 1.96* acc_std/np.sqrt(iter_num)), file = acc_file) + + # remove feature files [optional] + if remove_featurefile: + os.remove(featurefile) diff --git a/tmp.txt b/tmp.txt new file mode 100644 index 0000000000000000000000000000000000000000..6d06227ad8275023b53cc3f30a83be4f294109f4 --- /dev/null +++ b/tmp.txt @@ -0,0 +1,6 @@ +epoch -1 miniImagenet: cub: cars: places: plantae: + 1000 test iterations: Acc = 67.36% +- 0.80% + 1000 test iterations: Acc = 48.34% +- 0.70% + 1000 test iterations: Acc = 34.30% +- 0.59% + 1000 test iterations: Acc = 58.53% +- 0.80% + 1000 test iterations: Acc = 41.29% +- 0.66% diff --git a/tmp2.txt b/tmp2.txt new file mode 100644 index 0000000000000000000000000000000000000000..0bb334f0c779571649dce2dde3a0c97001ed3b48 --- /dev/null +++ b/tmp2.txt @@ -0,0 +1,2 @@ +epoch -1 miniImagenet: cub: cars: places: plantae: + 1000 test iterations: Acc = 67.36% +- 0.80% diff --git a/tmp_bscdfsl.txt b/tmp_bscdfsl.txt new file mode 100644 index 0000000000000000000000000000000000000000..b10a3c9a626ff98f3858548788c576c0d0b2bcd8 --- /dev/null +++ b/tmp_bscdfsl.txt @@ -0,0 +1,5 @@ +epoch -1 ChestX: ISIC: EuroSAT: CropDisease + 1000 test iterations: Acc = 22.19% +- 0.35% + 1000 test iterations: Acc = 34.22% +- 0.56% + 1000 test iterations: Acc = 70.66% +- 0.84% + 1000 test iterations: Acc = 73.82% +- 0.75% diff --git a/tmp_bscdfsl2.txt b/tmp_bscdfsl2.txt new file mode 100644 index 0000000000000000000000000000000000000000..de30e4b9b03dde7fddb0be2fd7f748bf45ae2c7f --- /dev/null +++ b/tmp_bscdfsl2.txt @@ -0,0 +1,2 @@ +epoch -1 ChestX: ISIC: EuroSAT: CropDisease + 1000 test iterations: Acc = 70.69% +- 0.85% diff --git a/utils/PSG.py b/utils/PSG.py new file mode 100644 index 0000000000000000000000000000000000000000..863265cac2cdacd9fe4cdbe421b8b7ccf6995e39 --- /dev/null +++ b/utils/PSG.py @@ -0,0 +1,94 @@ +import torch +import torchvision.transforms as transforms +import random + +def gamma_correction(x, gamma): + minv = torch.min(x) + x = x - minv + + maxv = torch.max(x) + x = x / maxv + + x = x**gamma + x = x * maxv + minv + return x + +def random_aug(x): + #print('x1:', x.size()) + # gamma correction + if random.random() <= 0.3: + gamma = random.uniform(1.0, 1.5) + x = gamma_correction(x, gamma) + # random erasing with mean value + mean_v = tuple(x.view(x.size(0), -1).mean(-1)) + re = transforms.RandomErasing(p=0.5, value=mean_v) + x = re(x) + # color channel shuffle + if random.random() <= 0.3: + l = [0,1,2] + random.shuffle(l) + x_c = torch.zeros_like(x) + x_c[l] = x + x = x_c + # horizontal flip or vertical flip + if random.random() <= 0.5: + if random.random() <= 0.5: + x = torch.flip(x, [1]) + else: + x = torch.flip(x, [2]) + # rotate 90, 180 or 270 degree + if random.random() <= 0.5: + degree = [90, 180, 270] + d = random.choice(degree) + x = torch.rot90(x, d//90, [1, 2]) + #print('x2:', x.size()) + return x + +class PseudoSampleGenerator(object): + def __init__(self, n_way, n_support, n_pseudo): + super(PseudoSampleGenerator, self).__init__() + self.n_way = n_way + self.n_support = n_support + self.n_pseudo = n_pseudo + self.n_pseudo_per_way = self.n_pseudo//self.n_way + + def generate(self, support_set): # (5*n_support, 3, 224, 224) + #default ATA: 1-shot/5-shot + if(self.n_support<=5): + times = self.n_pseudo//(self.n_way*self.n_support)+1 + psedo_list = [] + for i in range(support_set.size(0)): + psedo_list.append(support_set[i]) + for j in range(1, times): + cur_x = support_set[i] + cur_x = random_aug(cur_x) + psedo_list.append(cur_x) + psedo_set = torch.stack(psedo_list) + #print('psedo_set:', psedo_set.size()) + psedo_set = psedo_set.reshape([self.n_way, self.n_pseudo_per_way+ self.n_support]+list(psedo_set.size()[1:])) + + # adapt ata to 20/50 shots + else: + #random select 15 support images from 20/50shot + support_set = support_set.view(self.n_way, self.n_support, 3, 224, 224) + #print('support_set:', support_set.size()) + perm = torch.randperm(self.n_support) + idx = perm[:15] + #print('idx:', idx) + selected_support_set = support_set[:, idx, :, :, :] + #print('selected_support_set:', selected_support_set.size()) + selected_support_set = selected_support_set.view(self.n_way*15, 3, 224, 224) + # use the selected_support_set to generate pesudo query + times =1 + psedo_query_list = [] + for i in range(selected_support_set.size(0)): + for j in range(0, times): + cur_x = selected_support_set[i] + cur_x = random_aug(cur_x) + psedo_query_list.append(cur_x) + psedo_query_list = torch.stack(psedo_query_list) + psedo_query_set = psedo_query_list.view(self.n_way, 15, 3, 224, 224) + #print('psedo_query_set:', psedo_query_set.size()) + psedo_set = torch.cat((support_set, psedo_query_set), dim = 1) + #print("psedo_set:", psedo_set.size()) + return psedo_set diff --git a/utils/__init__.py b/utils/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..d12c02d894b95820ad7fa7a575717eb7cafb5095 --- /dev/null +++ b/utils/__init__.py @@ -0,0 +1,65 @@ +import os +import torch +import numpy as np +import random +import collections + + +class AverageMeter(object): + """Computes and stores the average and current value""" + def __init__(self): + self.reset() + + def reset(self): + self.val = 0 + self.avg = 0 + self.sum = 0 + self.count = 0 + + def update(self, val, n=1): + self.val = val + self.sum += val * n + self.count += n + self.avg = self.sum / self.count + + +def simple_accuracy(preds, labels): + return (preds == labels).mean() + + +def save_model(args, model): + model_to_save = model.module if hasattr(model, 'module') else model + model_checkpoint = os.path.join(args.output_dir, "%s_checkpoint.bin" % args.name) + torch.save(model_to_save.state_dict(), model_checkpoint) + + +def load_model(args, model): + model_to_save = model.module if hasattr(model, 'module') else model + model_checkpoint = os.path.join(args.output_dir, "%s_checkpoint.bin" % args.name) + model.load_state_dict(torch.load(model_checkpoint, map_location='cpu')) + + +def count_parameters(model): + params = sum(p.numel() for p in model.parameters() if p.requires_grad) + return params/1000000 + + +def set_seed(args): + random.seed(args.seed) + np.random.seed(args.seed) + torch.manual_seed(args.seed) + if args.gpus > 0: + torch.cuda.manual_seed_all(args.seed) + + +def to_device(input, device): + if torch.is_tensor(input): + return input.to(device=device, non_blocking=True) + elif isinstance(input, str): + return input + elif isinstance(input, collections.Mapping): + return {k: to_device(sample, device=device) for k, sample in input.items()} + elif isinstance(input, collections.Sequence): + return [to_device(sample, device=device) for sample in input] + else: + raise TypeError("Input must contain tensor, dict or list, found {type(input)}") diff --git a/utils/args.py b/utils/args.py new file mode 100644 index 0000000000000000000000000000000000000000..3a491eca1c1183ce23df644e4b49fd85a9c013e5 --- /dev/null +++ b/utils/args.py @@ -0,0 +1,231 @@ +import argparse +import numpy as np + + +def get_args_parser(): + parser = argparse.ArgumentParser('Few-shot learning script', add_help=False) + # General + parser.add_argument('--batch-size', default=1, type=int) + parser.add_argument('--num_classes', default=1000, type=int) + parser.add_argument('--epochs', default=100, type=int) + parser.add_argument('--fp16', action='store_true', + help="Whether to use 16-bit float precision instead of 32-bit") + parser.set_defaults(fp16=True) + parser.add_argument('--output_dir', default='output/tmp', + help='path where to save, empty for no saving') + parser.add_argument('--device', default='cuda', + help='cuda:gpu_id for single GPU training') + parser.add_argument('--seed', default=0, type=int) + + # Dataset parameters + parser.add_argument('--data-path', default='/datasets01/imagenet_full_size/061417/', type=str, + help='dataset path') + parser.add_argument('--pretrained-checkpoint-path', default='.', type=str, + help='path which contains the directories pretrained_ckpts and pretrained_ckpts_converted') + parser.add_argument("--dataset", choices=["cifar_fs_elite", "cifar_fs", "mini_imagenet", + "meta_dataset", "meta_dataset_h5", "imagenet_h5", + "full_meta_dataset", "imagewise_meta_dataset"], + default="cifar_fs", + help="Which few-shot dataset.") + + # Few-shot parameters (Mini-ImageNet & CIFAR-FS) + parser.add_argument("--nClsEpisode", default=5, type=int, + help="Number of categories in each episode.") + parser.add_argument("--nSupport", default=1, type=int, + help="Number of samples per category in the support set.") + parser.add_argument("--nQuery", default=15, type=int, + help="Number of samples per category in the query set.") + parser.add_argument("--nValEpisode", default=120, type=int, + help="Number of episodes for validation.") + parser.add_argument("--nEpisode", default=2000, type=int, + help="Number of episodes for training / testing.") + + # MetaDataset parameters + parser.add_argument('--image_size', type=int, default=128, + help='Images will be resized to this value') + parser.add_argument('--base_sources', nargs="+", default=['aircraft', 'cu_birds', 'dtd', 'fungi', 'ilsvrc_2012', 'omniglot', 'quickdraw', 'vgg_flower'], + help='List of datasets to use for training') + parser.add_argument('--val_sources', nargs="+", default=['aircraft', 'cu_birds', 'dtd', 'fungi', 'ilsvrc_2012', 'omniglot', 'quickdraw', 'vgg_flower'], + help='List of datasets to use for validation') + parser.add_argument('--test_sources', nargs="+", default=['traffic_sign', 'mscoco', 'ilsvrc_2012', 'omniglot', 'aircraft', 'cu_birds', 'dtd', 'quickdraw', 'fungi', 'vgg_flower'], + help='List of datasets to use for meta-testing') + parser.add_argument('--shuffle', type=bool, default=True, + help='Whether or not to shuffle data for TFRecordDataset') + parser.add_argument('--train_transforms', nargs="+", default=['random_resized_crop', 'jitter', 'random_flip', 'to_tensor', 'normalize'], + help='Transforms applied to training data',) + parser.add_argument('--test_transforms', nargs="+", default=['resize', 'center_crop', 'to_tensor', 'normalize'], + help='Transforms applied to test data',) + parser.add_argument('--num_ways', type=int, default=None, + help='Set it if you want a fixed # of ways per task') + parser.add_argument('--num_support', type=int, default=None, + help='Set it if you want a fixed # of support samples per class') + parser.add_argument('--num_query', type=int, default=None, + help='Set it if you want a fixed # of query samples per class') + parser.add_argument('--min_ways', type=int, default=5, + help='Minimum # of ways per task') + parser.add_argument('--max_ways_upper_bound', type=int, default=50, + help='Maximum # of ways per task') + parser.add_argument('--max_num_query', type=int, default=10, + help='Maximum # of query samples') + parser.add_argument('--max_support_set_size', type=int, default=500, + help='Maximum # of support samples') + parser.add_argument('--max_support_size_contrib_per_class', type=int, default=100, + help='Maximum # of support samples per class') + parser.add_argument('--min_examples_in_class', type=int, default=0, + help='Classes that have less samples will be skipped') + parser.add_argument('--min_log_weight', type=float, default=np.log(0.5), + help='Do not touch, used to randomly sample support set') + parser.add_argument('--max_log_weight', type=float, default=np.log(2), + help='Do not touch, used to randomly sample support set') + parser.add_argument('--ignore_bilevel_ontology', action='store_true', + help='Whether or not to use superclass for BiLevel datasets (e.g Omniglot)') + parser.add_argument('--ignore_dag_ontology', action='store_true', + help='Whether to ignore ImageNet DAG ontology when sampling \ + classes from it. This has no effect if ImageNet is not \ + part of the benchmark.') + parser.add_argument('--ignore_hierarchy_probability', type=float, default=0., + help='if using a hierarchy, this flag makes the sampler \ + ignore the hierarchy for this proportion of episodes \ + and instead sample categories uniformly.') + + # CDFSL parameters + parser.add_argument('--test_n_way' , default=5, type=int, help='class num to classify for testing (validation) ') + parser.add_argument('--n_shot' , default=5, type=int, help='number of labeled data in each class, same as n_support') + parser.add_argument('--cdfsl_domains', nargs="+", default=['EuroSAT', 'ISIC', 'CropDisease', 'ChestX'], help='CDFSL datasets') + + # Model params + parser.add_argument('--arch', default='dino_base_patch16_224', type=str, + help='Architecture of the backbone.') + parser.add_argument('--patch_size', default=16, type=int, help='Patch resolution of the model.') + parser.add_argument('--pretrained_weights', default='', type=str, help="Path to pretrained weights to evaluate.") + parser.add_argument("--checkpoint_key", default="teacher", type=str, help='Key to use in the checkpoint (example: "teacher")') + parser.add_argument('--unused_params', action='store_true') + parser.add_argument('--no-pretrain', action='store_true') + + # Deployment params + parser.add_argument("--deploy", type=str, default="vanilla", + help="Which few-shot model to be deployed for meta-testing.") + parser.add_argument('--num_adapters', default=1, type=int, help='Number of adapter tokens') + parser.add_argument('--ada_steps', default=40, type=int, help='Number of feature adaptation steps') + parser.add_argument('--ada_lr', default=5e-2, type=float, help='Learning rate of feature adaptation') + parser.add_argument('--aug_prob', default=0.9, type=float, help='Probability of applying data augmentation during meta-testing') + parser.add_argument('--aug_types', nargs="+", default=['color', 'translation'], + help='color, offset, offset_h, offset_v, translation, cutout') + + # Other model parameters + parser.add_argument('--img-size', default=224, type=int, help='images input size') + + parser.add_argument('--drop', type=float, default=0.0, metavar='PCT', + help='Dropout rate (default: 0.)') + parser.add_argument('--drop-path', type=float, default=0.1, metavar='PCT', + help='Drop path rate (default: 0.1)') + + parser.add_argument('--model-ema', action='store_true') + parser.add_argument('--no-model-ema', action='store_false', dest='model_ema') + parser.set_defaults(model_ema=False) + parser.add_argument('--model-ema-decay', type=float, default=0.99996, help='') + parser.add_argument('--model-ema-force-cpu', action='store_true', default=False, help='') + + # Optimizer parameters + parser.add_argument('--opt', default='adamw', type=str, metavar='OPTIMIZER', + help='Optimizer (default: "adamw"') + parser.add_argument('--opt-eps', default=1e-8, type=float, metavar='EPSILON', + help='Optimizer Epsilon (default: 1e-8)') + parser.add_argument('--opt-betas', default=None, type=float, nargs='+', metavar='BETA', + help='Optimizer Betas (default: None, use opt default)') + parser.add_argument('--clip-grad', type=float, default=None, metavar='NORM', + help='Clip gradient norm (default: None, no clipping)') + parser.add_argument('--momentum', type=float, default=0.9, metavar='M', + help='SGD momentum (default: 0.9)') + parser.add_argument('--weight-decay', type=float, default=0.05, + help='weight decay (default: 0.05)') + + # Learning rate schedule parameters + parser.add_argument('--sched', default='cosine', type=str, metavar='SCHEDULER', + help='LR scheduler (default: "cosine"') + parser.add_argument('--lr', type=float, default=5e-5, metavar='LR', + help='learning rate (default: 5e-4)') + parser.add_argument('--lr-noise', type=float, nargs='+', default=None, metavar='pct, pct', + help='learning rate noise on/off epoch percentages') + parser.add_argument('--lr-noise-pct', type=float, default=0.67, metavar='PERCENT', + help='learning rate noise limit percent (default: 0.67)') + parser.add_argument('--lr-noise-std', type=float, default=1.0, metavar='STDDEV', + help='learning rate noise std-dev (default: 1.0)') + parser.add_argument('--warmup-lr', type=float, default=1e-6, metavar='LR', + help='warmup learning rate (default: 1e-6)') + parser.add_argument('--min-lr', type=float, default=1e-6, metavar='LR', + help='lower lr bound for cyclic schedulers that hit 0 (1e-5)') + + parser.add_argument('--decay-epochs', type=float, default=30, metavar='N', + help='epoch interval to decay LR (step scheduler)') + parser.add_argument('--warmup-epochs', type=int, default=5, metavar='N', + help='epochs to warmup LR, if scheduler supports') + parser.add_argument('--cooldown-epochs', type=int, default=10, metavar='N', + help='epochs to cooldown LR at min_lr, after cyclic schedule ends') + parser.add_argument('--patience-epochs', type=int, default=10, metavar='N', + help='patience epochs for Plateau LR scheduler (default: 10') + parser.add_argument('--decay-rate', '--dr', type=float, default=0.1, metavar='RATE', + help='LR decay rate (default: 0.1)') + + # Augmentation parameters + parser.add_argument('--color-jitter', type=float, default=0.4, metavar='PCT', + help='Color jitter factor (default: 0.4)') + parser.add_argument('--aa', type=str, default='rand-m9-mstd0.5-inc1', metavar='NAME', + help='Use AutoAugment policy. "v0" or "original". " + \ + "(default: rand-m9-mstd0.5-inc1)'), + parser.add_argument('--smoothing', type=float, default=0.0, help='Label smoothing (default: 0.1)') + parser.add_argument('--train-interpolation', type=str, default='bicubic', + help='Training interpolation (random, bilinear, bicubic default: "bicubic")') + + parser.add_argument('--repeated-aug', action='store_true') + + # * Random Erase params + parser.add_argument('--reprob', type=float, default=0.25, metavar='PCT', + help='Random erase prob (default: 0.25)') + parser.add_argument('--remode', type=str, default='pixel', + help='Random erase mode (default: "pixel")') + parser.add_argument('--recount', type=int, default=1, + help='Random erase count (default: 1)') + parser.add_argument('--resplit', action='store_true', default=False, + help='Do not random erase first (clean) augmentation split') + + # * Mixup params + parser.add_argument('--mixup', type=float, default=0., + help='mixup alpha, mixup enabled if > 0. (default: 0.8)') + parser.add_argument('--cutmix', type=float, default=0., + help='cutmix alpha, cutmix enabled if > 0. (default: 1.0)') + parser.add_argument('--cutmix-minmax', type=float, nargs='+', default=None, + help='cutmix min/max ratio, overrides alpha and enables cutmix if set (default: None)') + parser.add_argument('--mixup-prob', type=float, default=1.0, + help='Probability of performing mixup or cutmix when either/both is enabled') + parser.add_argument('--mixup-switch-prob', type=float, default=0.5, + help='Probability of switching to cutmix when both mixup and cutmix enabled') + parser.add_argument('--mixup-mode', type=str, default='batch', + help='How to apply mixup/cutmix params. Per "batch", "pair", or "elem"') + + # Distillation parameters + parser.add_argument('--teacher-model', default='regnety_160', type=str, metavar='MODEL', + help='Name of teacher model to train (default: "regnety_160"') + parser.add_argument('--teacher-path', type=str, default='') + parser.add_argument('--distillation-type', default='none', choices=['none', 'soft', 'hard'], type=str, help="") + parser.add_argument('--distillation-alpha', default=0.5, type=float, help="") + parser.add_argument('--distillation-tau', default=1.0, type=float, help="") + + # Misc + parser.add_argument('--resume', default='', help='resume from checkpoint') + parser.add_argument('--start_epoch', default=0, type=int, metavar='N', + help='start epoch') + parser.add_argument('--eval', action='store_true', help='Perform evaluation only') + parser.add_argument('--dist-eval', action='store_true', default=False, help='Enabling distributed evaluation') + parser.add_argument('--num_workers', default=10, type=int) + parser.add_argument('--pin-mem', action='store_true', + help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.') + parser.add_argument('--no-pin-mem', action='store_false', dest='pin_mem', + help='') + parser.set_defaults(pin_mem=True) + + # distributed training parameters + parser.add_argument('--world_size', default=1, type=int, + help='number of distributed processes') + parser.add_argument('--dist_url', default='env://', help='url used to set up distributed training') + return parser diff --git a/utils/deit_util.py b/utils/deit_util.py new file mode 100644 index 0000000000000000000000000000000000000000..f12e92787bfd7f91bc9d1891f952ffa19727d35f --- /dev/null +++ b/utils/deit_util.py @@ -0,0 +1,246 @@ +# Copyright (c) 2015-present, Facebook, Inc. +# All rights reserved. +""" +Misc functions, including distributed helpers. + +Mostly copy-paste from torchvision references. +""" +import io +import os +import time +from collections import defaultdict, deque +import datetime + +import torch +import torch.distributed as dist + + +class SmoothedValue(object): + """Track a series of values and provide access to smoothed values over a + window or the global series average. + """ + + def __init__(self, window_size=20, fmt=None): + if fmt is None: + fmt = "{median:.4f} ({global_avg:.4f})" + self.deque = deque(maxlen=window_size) + self.total = 0.0 + self.count = 0 + self.fmt = fmt + + def update(self, value, n=1): + self.deque.append(value) + self.count += n + self.total += value * n + + def synchronize_between_processes(self): + """ + Warning: does not synchronize the deque! + """ + if not is_dist_avail_and_initialized(): + return + t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda') + dist.barrier() + dist.all_reduce(t) + t = t.tolist() + self.count = int(t[0]) + self.total = t[1] + + @property + def median(self): + d = torch.tensor(list(self.deque)) + return d.median().item() + + @property + def avg(self): + d = torch.tensor(list(self.deque), dtype=torch.float32) + return d.mean().item() + + @property + def std(self): + d = torch.tensor(list(self.deque), dtype=torch.float32) + return d.std().item() + + @property + def global_avg(self): + return self.total / self.count + + @property + def max(self): + return max(self.deque) + + @property + def value(self): + return self.deque[-1] + + def __str__(self): + return self.fmt.format( + median=self.median, + avg=self.avg, + global_avg=self.global_avg, + max=self.max, + value=self.value) + + +class MetricLogger(object): + def __init__(self, delimiter="\t"): + self.meters = defaultdict(SmoothedValue) + self.delimiter = delimiter + + def update(self, **kwargs): + for k, v in kwargs.items(): + if isinstance(v, torch.Tensor): + v = v.item() + assert isinstance(v, (float, int)) + self.meters[k].update(v) + + def __getattr__(self, attr): + if attr in self.meters: + return self.meters[attr] + if attr in self.__dict__: + return self.__dict__[attr] + raise AttributeError("'{}' object has no attribute '{}'".format( + type(self).__name__, attr)) + + def __str__(self): + loss_str = [] + for name, meter in self.meters.items(): + loss_str.append( + "{}: {}".format(name, str(meter)) + ) + return self.delimiter.join(loss_str) + + def synchronize_between_processes(self): + for meter in self.meters.values(): + meter.synchronize_between_processes() + + def add_meter(self, name, meter): + self.meters[name] = meter + + def log_every(self, iterable, print_freq, header=None): + i = 0 + if not header: + header = '' + start_time = time.time() + end = time.time() + iter_time = SmoothedValue(fmt='{avg:.4f}') + data_time = SmoothedValue(fmt='{avg:.4f}') + space_fmt = ':' + str(len(str(len(iterable)))) + 'd' + log_msg = [ + header, + '[{0' + space_fmt + '}/{1}]', + 'eta: {eta}', + '{meters}', + 'time: {time}', + 'data: {data}' + ] + if torch.cuda.is_available(): + log_msg.append('max mem: {memory:.0f}') + log_msg = self.delimiter.join(log_msg) + MB = 1024.0 * 1024.0 + for obj in iterable: + data_time.update(time.time() - end) + yield obj + iter_time.update(time.time() - end) + if i % print_freq == 0 or i == len(iterable) - 1: + eta_seconds = iter_time.global_avg * (len(iterable) - i) + eta_string = str(datetime.timedelta(seconds=int(eta_seconds))) + if torch.cuda.is_available(): + print(log_msg.format( + i, len(iterable), eta=eta_string, + meters=str(self), + time=str(iter_time), data=str(data_time), + memory=torch.cuda.max_memory_allocated() / MB)) + else: + print(log_msg.format( + i, len(iterable), eta=eta_string, + meters=str(self), + time=str(iter_time), data=str(data_time))) + i += 1 + end = time.time() + total_time = time.time() - start_time + total_time_str = str(datetime.timedelta(seconds=int(total_time))) + print('{} Total time: {} ({:.4f} s / it)'.format( + header, total_time_str, total_time / len(iterable))) + + +def _load_checkpoint_for_ema(model_ema, checkpoint): + """ + Workaround for ModelEma._load_checkpoint to accept an already-loaded object + """ + mem_file = io.BytesIO() + torch.save(checkpoint, mem_file) + mem_file.seek(0) + model_ema._load_checkpoint(mem_file) + + +def setup_for_distributed(is_master): + """ + This function disables printing when not in master process + """ + import builtins as __builtin__ + builtin_print = __builtin__.print + + def print(*args, **kwargs): + force = kwargs.pop('force', False) + if is_master or force: + builtin_print(*args, **kwargs) + + __builtin__.print = print + + +def is_dist_avail_and_initialized(): + if not dist.is_available(): + return False + if not dist.is_initialized(): + return False + return True + + +def get_world_size(): + if not is_dist_avail_and_initialized(): + return 1 + return dist.get_world_size() + + +def get_rank(): + if not is_dist_avail_and_initialized(): + return 0 + return dist.get_rank() + + +def is_main_process(): + return get_rank() == 0 + + +def save_on_master(*args, **kwargs): + if is_main_process(): + torch.save(*args, **kwargs) + + +def init_distributed_mode(args): + if args.device != 'cuda': + print('Not using distributed mode') + args.distributed = False + return + + if 'RANK' in os.environ and 'WORLD_SIZE' in os.environ: + args.rank = int(os.environ["RANK"]) + args.world_size = int(os.environ['WORLD_SIZE']) + args.gpu = int(os.environ['LOCAL_RANK']) + elif 'SLURM_PROCID' in os.environ: + args.rank = int(os.environ['SLURM_PROCID']) + args.gpu = args.rank % torch.cuda.device_count() + + args.distributed = True + + args.dist_backend = 'nccl' + print('| distributed init (rank {}): {}'.format( + args.rank, args.dist_url), flush=True) + torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url, + world_size=args.world_size, rank=args.rank) + + #torch.distributed.barrier() + torch.distributed.barrier(device_ids=[args.gpu]) + torch.cuda.set_device(args.gpu) + setup_for_distributed(args.rank == 0) diff --git a/utils/dino_utils.py b/utils/dino_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..53b51175caf4c898e47839dbc0c3042c929f47b9 --- /dev/null +++ b/utils/dino_utils.py @@ -0,0 +1,578 @@ +# Copyright (c) Facebook, Inc. and its affiliates. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +""" +Misc functions. + +Mostly copy-paste from torchvision references or other public repos like DETR: +https://github.com/facebookresearch/detr/blob/master/util/misc.py +""" +import os +import sys +import time +import math +import random +import datetime +import subprocess +from collections import defaultdict, deque + +import numpy as np +import torch +from torch import nn +import torch.distributed as dist +from PIL import ImageFilter, ImageOps + + +class GaussianBlur(object): + """ + Apply Gaussian Blur to the PIL image. + """ + def __init__(self, p=0.5, radius_min=0.1, radius_max=2.): + self.prob = p + self.radius_min = radius_min + self.radius_max = radius_max + + def __call__(self, img): + do_it = random.random() <= self.prob + if not do_it: + return img + + return img.filter( + ImageFilter.GaussianBlur( + radius=random.uniform(self.radius_min, self.radius_max) + ) + ) + + +class Solarization(object): + """ + Apply Solarization to the PIL image. + """ + def __init__(self, p): + self.p = p + + def __call__(self, img): + if random.random() < self.p: + return ImageOps.solarize(img) + else: + return img + + +def load_pretrained_weights(model, pretrained_weights, checkpoint_key, model_name, patch_size): + if os.path.isfile(pretrained_weights): + state_dict = torch.load(pretrained_weights, map_location="cpu") + if checkpoint_key is not None and checkpoint_key in state_dict: + print(f"Take key {checkpoint_key} in provided checkpoint dict") + state_dict = state_dict[checkpoint_key] + # remove `module.` prefix + state_dict = {k.replace("module.", ""): v for k, v in state_dict.items()} + # remove `backbone.` prefix induced by multicrop wrapper + state_dict = {k.replace("backbone.", ""): v for k, v in state_dict.items()} + msg = model.load_state_dict(state_dict, strict=False) + print('Pretrained weights found at {} and loaded with msg: {}'.format(pretrained_weights, msg)) + else: + print("Please use the `--pretrained_weights` argument to indicate the path of the checkpoint to evaluate.") + url = None + if model_name == "vit_small" and patch_size == 16: + url = "dino_deitsmall16_pretrain/dino_deitsmall16_pretrain.pth" + elif model_name == "vit_small" and patch_size == 8: + url = "dino_deitsmall8_pretrain/dino_deitsmall8_pretrain.pth" + elif model_name == "vit_base" and patch_size == 16: + url = "dino_vitbase16_pretrain/dino_vitbase16_pretrain.pth" + elif model_name == "vit_base" and patch_size == 8: + url = "dino_vitbase8_pretrain/dino_vitbase8_pretrain.pth" + if url is not None: + print("Since no pretrained weights have been provided, we load the reference pretrained DINO weights.") + state_dict = torch.hub.load_state_dict_from_url(url="https://dl.fbaipublicfiles.com/dino/" + url) + model.load_state_dict(state_dict, strict=True) + else: + print("There is no reference weights available for this model => We use random weights.") + + +def clip_gradients(model, clip): + norms = [] + for name, p in model.named_parameters(): + if p.grad is not None: + param_norm = p.grad.data.norm(2) + norms.append(param_norm.item()) + clip_coef = clip / (param_norm + 1e-6) + if clip_coef < 1: + p.grad.data.mul_(clip_coef) + return norms + + +def cancel_gradients_last_layer(epoch, model, freeze_last_layer): + if epoch >= freeze_last_layer: + return + for n, p in model.named_parameters(): + if "last_layer" in n: + p.grad = None + + +def restart_from_checkpoint(ckp_path, run_variables=None, **kwargs): + """ + Re-start from checkpoint + """ + if not os.path.isfile(ckp_path): + return + print("Found checkpoint at {}".format(ckp_path)) + + # open checkpoint file + checkpoint = torch.load(ckp_path, map_location="cpu") + + # key is what to look for in the checkpoint file + # value is the object to load + # example: {'state_dict': model} + for key, value in kwargs.items(): + if key in checkpoint and value is not None: + try: + msg = value.load_state_dict(checkpoint[key], strict=False) + print("=> loaded {} from checkpoint '{}' with msg {}".format(key, ckp_path, msg)) + except TypeError: + try: + msg = value.load_state_dict(checkpoint[key]) + print("=> loaded {} from checkpoint '{}'".format(key, ckp_path)) + except ValueError: + print("=> failed to load {} from checkpoint '{}'".format(key, ckp_path)) + else: + print("=> failed to load {} from checkpoint '{}'".format(key, ckp_path)) + + # re load variable important for the run + if run_variables is not None: + for var_name in run_variables: + if var_name in checkpoint: + run_variables[var_name] = checkpoint[var_name] + + +def cosine_scheduler(base_value, final_value, epochs, niter_per_ep, warmup_epochs=0, start_warmup_value=0): + warmup_schedule = np.array([]) + warmup_iters = warmup_epochs * niter_per_ep + if warmup_epochs > 0: + warmup_schedule = np.linspace(start_warmup_value, base_value, warmup_iters) + + iters = np.arange(epochs * niter_per_ep - warmup_iters) + schedule = final_value + 0.5 * (base_value - final_value) * (1 + np.cos(np.pi * iters / len(iters))) + + schedule = np.concatenate((warmup_schedule, schedule)) + assert len(schedule) == epochs * niter_per_ep + return schedule + + +def bool_flag(s): + """ + Parse boolean arguments from the command line. + """ + FALSY_STRINGS = {"off", "false", "0"} + TRUTHY_STRINGS = {"on", "true", "1"} + if s.lower() in FALSY_STRINGS: + return False + elif s.lower() in TRUTHY_STRINGS: + return True + else: + raise argparse.ArgumentTypeError("invalid value for a boolean flag") + + +def fix_random_seeds(seed=31): + """ + Fix random seeds. + """ + torch.manual_seed(seed) + torch.cuda.manual_seed_all(seed) + np.random.seed(seed) + + +class SmoothedValue(object): + """Track a series of values and provide access to smoothed values over a + window or the global series average. + """ + + def __init__(self, window_size=20, fmt=None): + if fmt is None: + fmt = "{median:.6f} ({global_avg:.6f})" + self.deque = deque(maxlen=window_size) + self.total = 0.0 + self.count = 0 + self.fmt = fmt + + def update(self, value, n=1): + self.deque.append(value) + self.count += n + self.total += value * n + + def synchronize_between_processes(self): + """ + Warning: does not synchronize the deque! + """ + if not is_dist_avail_and_initialized(): + return + t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda') + dist.barrier() + dist.all_reduce(t) + t = t.tolist() + self.count = int(t[0]) + self.total = t[1] + + @property + def median(self): + d = torch.tensor(list(self.deque)) + return d.median().item() + + @property + def avg(self): + d = torch.tensor(list(self.deque), dtype=torch.float32) + return d.mean().item() + + @property + def global_avg(self): + return self.total / self.count + + @property + def max(self): + return max(self.deque) + + @property + def value(self): + return self.deque[-1] + + def __str__(self): + return self.fmt.format( + median=self.median, + avg=self.avg, + global_avg=self.global_avg, + max=self.max, + value=self.value) + + +def reduce_dict(input_dict, average=True): + """ + Args: + input_dict (dict): all the values will be reduced + average (bool): whether to do average or sum + Reduce the values in the dictionary from all processes so that all processes + have the averaged results. Returns a dict with the same fields as + input_dict, after reduction. + """ + world_size = get_world_size() + if world_size < 2: + return input_dict + with torch.no_grad(): + names = [] + values = [] + # sort the keys so that they are consistent across processes + for k in sorted(input_dict.keys()): + names.append(k) + values.append(input_dict[k]) + values = torch.stack(values, dim=0) + dist.all_reduce(values) + if average: + values /= world_size + reduced_dict = {k: v for k, v in zip(names, values)} + return reduced_dict + + +class MetricLogger(object): + def __init__(self, delimiter="\t"): + self.meters = defaultdict(SmoothedValue) + self.delimiter = delimiter + + def update(self, **kwargs): + for k, v in kwargs.items(): + if isinstance(v, torch.Tensor): + v = v.item() + assert isinstance(v, (float, int)) + self.meters[k].update(v) + + def __getattr__(self, attr): + if attr in self.meters: + return self.meters[attr] + if attr in self.__dict__: + return self.__dict__[attr] + raise AttributeError("'{}' object has no attribute '{}'".format( + type(self).__name__, attr)) + + def __str__(self): + loss_str = [] + for name, meter in self.meters.items(): + loss_str.append( + "{}: {}".format(name, str(meter)) + ) + return self.delimiter.join(loss_str) + + def synchronize_between_processes(self): + for meter in self.meters.values(): + meter.synchronize_between_processes() + + def add_meter(self, name, meter): + self.meters[name] = meter + + def log_every(self, iterable, print_freq, header=None): + i = 0 + if not header: + header = '' + start_time = time.time() + end = time.time() + iter_time = SmoothedValue(fmt='{avg:.6f}') + data_time = SmoothedValue(fmt='{avg:.6f}') + space_fmt = ':' + str(len(str(len(iterable)))) + 'd' + if torch.cuda.is_available(): + log_msg = self.delimiter.join([ + header, + '[{0' + space_fmt + '}/{1}]', + 'eta: {eta}', + '{meters}', + 'time: {time}', + 'data: {data}', + 'max mem: {memory:.0f}' + ]) + else: + log_msg = self.delimiter.join([ + header, + '[{0' + space_fmt + '}/{1}]', + 'eta: {eta}', + '{meters}', + 'time: {time}', + 'data: {data}' + ]) + MB = 1024.0 * 1024.0 + for obj in iterable: + data_time.update(time.time() - end) + yield obj + iter_time.update(time.time() - end) + if i % print_freq == 0 or i == len(iterable) - 1: + eta_seconds = iter_time.global_avg * (len(iterable) - i) + eta_string = str(datetime.timedelta(seconds=int(eta_seconds))) + if torch.cuda.is_available(): + print(log_msg.format( + i, len(iterable), eta=eta_string, + meters=str(self), + time=str(iter_time), data=str(data_time), + memory=torch.cuda.max_memory_allocated() / MB)) + else: + print(log_msg.format( + i, len(iterable), eta=eta_string, + meters=str(self), + time=str(iter_time), data=str(data_time))) + i += 1 + end = time.time() + total_time = time.time() - start_time + total_time_str = str(datetime.timedelta(seconds=int(total_time))) + print('{} Total time: {} ({:.6f} s / it)'.format( + header, total_time_str, total_time / len(iterable))) + + +def get_sha(): + cwd = os.path.dirname(os.path.abspath(__file__)) + + def _run(command): + return subprocess.check_output(command, cwd=cwd).decode('ascii').strip() + sha = 'N/A' + diff = "clean" + branch = 'N/A' + try: + sha = _run(['git', 'rev-parse', 'HEAD']) + subprocess.check_output(['git', 'diff'], cwd=cwd) + diff = _run(['git', 'diff-index', 'HEAD']) + diff = "has uncommited changes" if diff else "clean" + branch = _run(['git', 'rev-parse', '--abbrev-ref', 'HEAD']) + except Exception: + pass + message = f"sha: {sha}, status: {diff}, branch: {branch}" + return message + + +def is_dist_avail_and_initialized(): + if not dist.is_available(): + return False + if not dist.is_initialized(): + return False + return True + + +def get_world_size(): + if not is_dist_avail_and_initialized(): + return 1 + return dist.get_world_size() + + +def get_rank(): + if not is_dist_avail_and_initialized(): + return 0 + return dist.get_rank() + + +def is_main_process(): + return get_rank() == 0 + + +def save_on_master(*args, **kwargs): + if is_main_process(): + torch.save(*args, **kwargs) + + +def setup_for_distributed(is_master): + """ + This function disables printing when not in master process + """ + import builtins as __builtin__ + builtin_print = __builtin__.print + + def print(*args, **kwargs): + force = kwargs.pop('force', False) + if is_master or force: + builtin_print(*args, **kwargs) + + __builtin__.print = print + + +def init_distributed_mode(args): + # launched with torch.distributed.launch + if 'RANK' in os.environ and 'WORLD_SIZE' in os.environ: + args.rank = int(os.environ["RANK"]) + args.world_size = int(os.environ['WORLD_SIZE']) + args.gpu = int(os.environ['LOCAL_RANK']) + # launched with submitit on a slurm cluster + elif 'SLURM_PROCID' in os.environ: + args.rank = int(os.environ['SLURM_PROCID']) + args.gpu = args.rank % torch.cuda.device_count() + # launched naively with `python main_dino.py` + # we manually add MASTER_ADDR and MASTER_PORT to env variables + elif torch.cuda.is_available(): + print('Will run the code on one GPU.') + args.rank, args.gpu, args.world_size = 0, 0, 1 + os.environ['MASTER_ADDR'] = '127.0.0.1' + os.environ['MASTER_PORT'] = '29500' + else: + print('Does not support training without GPU.') + sys.exit(1) + + dist.init_process_group( + backend="nccl", + init_method=args.dist_url, + world_size=args.world_size, + rank=args.rank, + ) + + torch.cuda.set_device(args.gpu) + print('| distributed init (rank {}): {}'.format( + args.rank, args.dist_url), flush=True) + dist.barrier() + setup_for_distributed(args.rank == 0) + + +def accuracy(output, target, topk=(1,)): + """Computes the accuracy over the k top predictions for the specified values of k""" + maxk = max(topk) + batch_size = target.size(0) + _, pred = output.topk(maxk, 1, True, True) + pred = pred.t() + correct = pred.eq(target.reshape(1, -1).expand_as(pred)) + return [correct[:k].reshape(-1).float().sum(0) * 100. / batch_size for k in topk] + + +class LARS(torch.optim.Optimizer): + """ + Almost copy-paste from https://github.com/facebookresearch/barlowtwins/blob/main/main.py + """ + def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, eta=0.001, + weight_decay_filter=None, lars_adaptation_filter=None): + defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, + eta=eta, weight_decay_filter=weight_decay_filter, + lars_adaptation_filter=lars_adaptation_filter) + super().__init__(params, defaults) + + @torch.no_grad() + def step(self): + for g in self.param_groups: + for p in g['params']: + dp = p.grad + + if dp is None: + continue + + if p.ndim != 1: + dp = dp.add(p, alpha=g['weight_decay']) + + if p.ndim != 1: + param_norm = torch.norm(p) + update_norm = torch.norm(dp) + one = torch.ones_like(param_norm) + q = torch.where(param_norm > 0., + torch.where(update_norm > 0, + (g['eta'] * param_norm / update_norm), one), one) + dp = dp.mul(q) + + param_state = self.state[p] + if 'mu' not in param_state: + param_state['mu'] = torch.zeros_like(p) + mu = param_state['mu'] + mu.mul_(g['momentum']).add_(dp) + + p.add_(mu, alpha=-g['lr']) + + +class MultiCropWrapper(nn.Module): + """ + Perform forward pass separately on each resolution input. + The inputs corresponding to a single resolution are clubbed and single + forward is run on the same resolution inputs. Hence we do several + forward passes = number of different resolutions used. We then + concatenate all the output features and run the head forward on these + concatenated features. + """ + def __init__(self, backbone, head): + super(MultiCropWrapper, self).__init__() + # disable layers dedicated to ImageNet labels classification + backbone.fc, backbone.head = nn.Identity(), nn.Identity() + self.backbone = backbone + self.head = head + + def forward(self, x): + # convert to list + if not isinstance(x, list): + x = [x] + idx_crops = torch.cumsum(torch.unique_consecutive( + torch.tensor([inp.shape[-1] for inp in x]), + return_counts=True, + )[1], 0) + start_idx = 0 + for end_idx in idx_crops: + _out = self.backbone(torch.cat(x[start_idx: end_idx])) + if start_idx == 0: + output = _out + else: + output = torch.cat((output, _out)) + start_idx = end_idx + # Run the head forward on the concatenated features. + return self.head(output) + + +def get_params_groups(model): + regularized = [] + not_regularized = [] + for name, param in model.named_parameters(): + if not param.requires_grad: + continue + # we do not regularize biases nor Norm parameters + if name.endswith(".bias") or len(param.shape) == 1: + not_regularized.append(param) + else: + regularized.append(param) + return [{'params': regularized}, {'params': not_regularized, 'weight_decay': 0.}] + + +def has_batchnorms(model): + bn_types = (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d, nn.SyncBatchNorm) + for name, module in model.named_modules(): + if isinstance(module, bn_types): + return True + return False diff --git a/utils/dist_util.py b/utils/dist_util.py new file mode 100644 index 0000000000000000000000000000000000000000..f98d4847608b836607c9b942284da736e37afe67 --- /dev/null +++ b/utils/dist_util.py @@ -0,0 +1,97 @@ +import io +import os + +import torch +import torch.distributed as dist + + +def _load_checkpoint_for_ema(model_ema, checkpoint): + """ + Workaround for ModelEma._load_checkpoint to accept an already-loaded object + """ + mem_file = io.BytesIO() + torch.save(checkpoint, mem_file) + mem_file.seek(0) + model_ema._load_checkpoint(mem_file) + + +def setup_for_distributed(is_master): + """ + This function disables printing when not in master process + """ + import builtins as __builtin__ + builtin_print = __builtin__.print + + def print(*args, **kwargs): + force = kwargs.pop('force', False) + if is_master or force: + builtin_print(*args, **kwargs) + + __builtin__.print = print + + +def is_dist_avail_and_initialized(): + if not dist.is_available(): + return False + if not dist.is_initialized(): + return False + return True + + +def get_world_size(): + if not is_dist_avail_and_initialized(): + return 1 + return dist.get_world_size() + + +def get_rank(): + if not is_dist_avail_and_initialized(): + return 0 + return dist.get_rank() + + +def is_main_process(): + return get_rank() == 0 + + +def save_on_master(*args, **kwargs): + if is_main_process(): + torch.save(*args, **kwargs) + + +def init_distributed_mode(args): + if 'RANK' in os.environ and 'WORLD_SIZE' in os.environ: + args.rank = int(os.environ["RANK"]) + args.world_size = int(os.environ['WORLD_SIZE']) + args.gpu = int(os.environ['LOCAL_RANK']) + elif 'SLURM_PROCID' in os.environ: + args.rank = int(os.environ['SLURM_PROCID']) + args.gpu = args.rank % torch.cuda.device_count() + else: + print('Not using distributed mode') + args.distributed = False + return + + args.distributed = True + + torch.cuda.set_device(args.gpu) + args.dist_backend = 'nccl' + print('| distributed init (rank {}): {}'.format( + args.rank, args.dist_url), flush=True) + torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url, + world_size=args.world_size, rank=args.rank) + torch.distributed.barrier() + setup_for_distributed(args.rank == 0) + + +def format_step(step): + if isinstance(step, str): + return step + s = "" + if len(step) > 0: + s += "Training Epoch: {} ".format(step[0]) + if len(step) > 1: + s += "Training Iteration: {} ".format(step[1]) + if len(step) > 2: + s += "Validation Iteration: {} ".format(step[2]) + return s diff --git a/utils/scheduler.py b/utils/scheduler.py new file mode 100644 index 0000000000000000000000000000000000000000..9daaf6e7ebca8a0de1477b66cd431981ef1f1461 --- /dev/null +++ b/utils/scheduler.py @@ -0,0 +1,63 @@ +import logging +import math + +from torch.optim.lr_scheduler import LambdaLR + +logger = logging.getLogger(__name__) + +class ConstantLRSchedule(LambdaLR): + """ Constant learning rate schedule. + """ + def __init__(self, optimizer, last_epoch=-1): + super(ConstantLRSchedule, self).__init__(optimizer, lambda _: 1.0, last_epoch=last_epoch) + + +class WarmupConstantSchedule(LambdaLR): + """ Linear warmup and then constant. + Linearly increases learning rate schedule from 0 to 1 over `warmup_steps` training steps. + Keeps learning rate schedule equal to 1. after warmup_steps. + """ + def __init__(self, optimizer, warmup_steps, last_epoch=-1): + self.warmup_steps = warmup_steps + super(WarmupConstantSchedule, self).__init__(optimizer, self.lr_lambda, last_epoch=last_epoch) + + def lr_lambda(self, step): + if step < self.warmup_steps: + return float(step) / float(max(1.0, self.warmup_steps)) + return 1. + + +class WarmupLinearSchedule(LambdaLR): + """ Linear warmup and then linear decay. + Linearly increases learning rate from 0 to 1 over `warmup_steps` training steps. + Linearly decreases learning rate from 1. to 0. over remaining `t_total - warmup_steps` steps. + """ + def __init__(self, optimizer, warmup_steps, t_total, last_epoch=-1): + self.warmup_steps = warmup_steps + self.t_total = t_total + super(WarmupLinearSchedule, self).__init__(optimizer, self.lr_lambda, last_epoch=last_epoch) + + def lr_lambda(self, step): + if step < self.warmup_steps: + return float(step) / float(max(1, self.warmup_steps)) + return max(0.0, float(self.t_total - step) / float(max(1.0, self.t_total - self.warmup_steps))) + + +class WarmupCosineSchedule(LambdaLR): + """ Linear warmup and then cosine decay. + Linearly increases learning rate from 0 to 1 over `warmup_steps` training steps. + Decreases learning rate from 1. to 0. over remaining `t_total - warmup_steps` steps following a cosine curve. + If `cycles` (default=0.5) is different from default, learning rate follows cosine function after warmup. + """ + def __init__(self, optimizer, warmup_steps, t_total, cycles=.5, last_epoch=-1): + self.warmup_steps = warmup_steps + self.t_total = t_total + self.cycles = cycles + super(WarmupCosineSchedule, self).__init__(optimizer, self.lr_lambda, last_epoch=last_epoch) + + def lr_lambda(self, step): + if step < self.warmup_steps: + return float(step) / float(max(1.0, self.warmup_steps)) + # progress after warmup + progress = float(step - self.warmup_steps) / float(max(1, self.t_total - self.warmup_steps)) + return max(0.0, 0.5 * (1. + math.cos(math.pi * float(self.cycles) * 2.0 * progress))) diff --git a/write_plantae_filelist.py b/write_plantae_filelist.py new file mode 100644 index 0000000000000000000000000000000000000000..619fd6b6ee6bc2880071461e28ea7a47481ee2e6 --- /dev/null +++ b/write_plantae_filelist.py @@ -0,0 +1,76 @@ +import numpy as np +from os import listdir +from os.path import isfile, isdir, join +import os +#import json +import random +from subprocess import call + +cwd = '/home/yuqian_fu/Data/CDFSL/Plantae' +source_path = cwd +data_path = join(cwd,'images') +if not os.path.exists(data_path): + os.makedirs(data_path) +savedir = './' +dataset_list = ['base','val','novel'] + + +folder_list = [f for f in listdir(source_path) if isdir(join(source_path, f))] +#folder_list.sort() +folder_list_count = np.array([len(listdir(join(source_path, f))) for f in folder_list]) +folder_list_idx = np.argsort(folder_list_count) +folder_list = np.array(folder_list)[folder_list_idx[-200:]].tolist() +label_dict = dict(zip(folder_list,range(0,len(folder_list)))) + +classfile_list_all = [] + +for i, folder in enumerate(folder_list): + source_folder_path = join(source_path, folder) + folder_path = join(data_path, folder) + classfile_list_all.append( [ cf for cf in listdir(source_folder_path) if (isfile(join(source_folder_path,cf)) and cf[0] != '.')]) + random.shuffle(classfile_list_all[i]) + classfile_list_all[i] = classfile_list_all[i][:min(len(classfile_list_all[i]), 600)] + + call('mkdir ' + folder_path, shell=True) + for cf in classfile_list_all[i]: + call('cp ' + join(source_folder_path, cf) + ' ' + join(folder_path, cf), shell=True) + classfile_list_all[i] = [join(folder_path, cf) for cf in classfile_list_all[i]] + +for dataset in dataset_list: + file_list = [] + label_list = [] + for i, classfile_list in enumerate(classfile_list_all): + if 'base' in dataset: + if (i%2 == 0): + file_list = file_list + classfile_list + label_list = label_list + np.repeat(i, len(classfile_list)).tolist() + if 'val' in dataset: + if (i%4 == 1): + file_list = file_list + classfile_list + label_list = label_list + np.repeat(i, len(classfile_list)).tolist() + if 'novel' in dataset: + if (i%4 == 3): + file_list = file_list + classfile_list + label_list = label_list + np.repeat(i, len(classfile_list)).tolist() + + fo = open(savedir + dataset + ".json", "w") + fo.write('{"label_names": [') + fo.writelines(['"%s",' % item for item in folder_list]) + fo.seek(0, os.SEEK_END) + fo.seek(fo.tell()-1, os.SEEK_SET) + fo.write('],') + + fo.write('"image_names": [') + fo.writelines(['"%s",' % item for item in file_list]) + fo.seek(0, os.SEEK_END) + fo.seek(fo.tell()-1, os.SEEK_SET) + fo.write('],') + + fo.write('"image_labels": [') + fo.writelines(['%d,' % item for item in label_list]) + fo.seek(0, os.SEEK_END) + fo.seek(fo.tell()-1, os.SEEK_SET) + fo.write(']}') + + fo.close() + print("%s -OK" %dataset) \ No newline at end of file